From 7cd468a3d7dee7d6c92f69a0bb7061ae208ec727 Mon Sep 17 00:00:00 2001 From: Damjan Marion Date: Mon, 19 Dec 2016 23:05:39 +0100 Subject: Reorganize source tree to use single autotools instance Change-Id: I7b51f88292e057c6443b12224486f2d0c9f8ae23 Signed-off-by: Damjan Marion --- src/Makefile.am | 101 ++++++++++++++++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 101 insertions(+) create mode 100644 src/Makefile.am (limited to 'src/Makefile.am') diff --git a/src/Makefile.am b/src/Makefile.am new file mode 100644 index 00000000..e691a539 --- /dev/null +++ b/src/Makefile.am @@ -0,0 +1,101 @@ +# Copyright (c) 2016 Cisco and/or its affiliates. +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at: +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +############################################################################### +# Global Defines +############################################################################### + +AUTOMAKE_OPTIONS = foreign subdir-objects +ACLOCAL_AMFLAGS = -I m4 + +AM_CFLAGS = -Wall + +SUBDIRS = . +SUFFIXES = .api.h .api .api.json +API_FILES = +noinst_HEADERS = +dist_bin_SCRIPTS = +lib_LTLIBRARIES = +BUILT_SOURCES = + +############################################################################### +# DPDK +############################################################################### + +if WITH_DPDK +if ENABLE_DPDK_SHARED +DPDK_LD_FLAGS = -Wl,--whole-archive,-ldpdk,--no-whole-archive +else +DPDK_LD_FLAGS = -Wl,--whole-archive,-l:libdpdk.a,--no-whole-archive +endif +if WITH_DPDK_CRYPTO +DPDK_LD_ADD = -L$(AESNI_MULTI_BUFFER_LIB_PATH) -lIPSec_MB +endif +if WITH_DPDK_MLX5_PMD +DPDK_LD_FLAGS += -libverbs -lmlx5 -lnuma +endif +else +DPDK_LD_FLAGS = +DPDK_LD_ADD = +endif + +############################################################################### +# Components +############################################################################### + +include vppinfra.am +include vppapigen.am + +if ENABLE_PERFTOOL +include perftool.am +endif + +if ENABLE_G2 +include g2.am +endif + +if ENABLE_SVM +include svm.am +endif + +if ENABLE_VLIB +include vlib.am +endif + +if ENABLE_SVM +if ENABLE_VLIB +include vlib-api.am +include vnet.am +include vpp.am +include vpp-api-test.am + +SUBDIRS += vpp-api/python plugins + +############################################################################### +# API +############################################################################### + +include suffix-rules.mk + +# Set the suffix list +apidir = $(prefix)/share/vpp/api/core + +api_DATA = \ + $(patsubst %.api,%.api.json,$(API_FILES)) + +BUILT_SOURCES += \ + $(patsubst %.api,%.api.json,$(API_FILES)) \ + $(patsubst %.api,%.api.h,$(API_FILES)) + +endif # if ENABLE_VLIB +endif # if ENABLE_SVM -- cgit 1.2.3-korg From cb034b9b374927c7552e36dcbc306d8456b2a0cb Mon Sep 17 00:00:00 2001 From: Damjan Marion Date: Wed, 28 Dec 2016 18:38:59 +0100 Subject: Move java,lua api and remaining plugins to src/ Change-Id: I1c3b87e886603678368428ae56a6bd3327cbc90d Signed-off-by: Damjan Marion --- .gitignore | 10 +- Makefile | 21 +- build-data/packages/plugins.mk | 34 - build-data/packages/vnet.mk | 47 - build-data/packages/vpp-api.mk | 9 - build-data/platforms.mk | 20 +- build-data/platforms/vpp.mk | 4 +- build-data/platforms/vpp_lite.mk | 2 +- build-root/deb/debian/.gitignore | 4 +- build-root/deb/debian/control | 16 +- build-root/packages/tools.mk | 2 +- build-root/rpm/vpp.spec | 70 +- build-root/scripts/find-api-core-contents | 2 +- build-root/scripts/find-dev-contents | 8 +- build-root/scripts/find-plugins-contents | 4 +- build-root/scripts/find-python-api-contents | 8 - build-root/scripts/find-vpp-api-java-contents | 8 + build-root/scripts/find-vpp-api-lua-contents | 6 + build-root/scripts/find-vpp-api-python-contents | 8 + doxygen/Makefile | 7 +- plugins/Makefile.am | 15 - plugins/acl-plugin/Makefile.am | 114 -- plugins/acl-plugin/acl/acl.api | 444 ----- plugins/acl-plugin/acl/acl.c | 1901 ------------------- plugins/acl-plugin/acl/acl.h | 148 -- plugins/acl-plugin/acl/acl_all_api_h.h | 321 ---- plugins/acl-plugin/acl/acl_msg_enum.h | 28 - plugins/acl-plugin/acl/acl_test.c | 1024 ---------- .../fd/vpp/jvpp/acl/test/AclExpectedDumpData.java | 135 -- .../jvpp/io/fd/vpp/jvpp/acl/test/AclTestData.java | 101 - .../io/fd/vpp/jvpp/acl/test/AclTestRequests.java | 141 -- .../io/fd/vpp/jvpp/acl/test/FutureApiTest.java | 68 - .../acl/jvpp/io/fd/vpp/jvpp/acl/test/Readme.txt | 1 - plugins/acl-plugin/acl/jvpp_acl.c | 124 -- plugins/acl-plugin/acl/jvpp_acl.h | 45 - plugins/acl-plugin/acl/l2sess.c | 243 --- plugins/acl-plugin/acl/l2sess.h | 150 -- plugins/acl-plugin/acl/l2sess_node.c | 816 -------- plugins/acl-plugin/acl/node_in.c | 168 -- plugins/acl-plugin/acl/node_in.h | 12 - plugins/acl-plugin/acl/node_out.c | 175 -- plugins/acl-plugin/acl/node_out.h | 12 - plugins/acl-plugin/configure.ac | 24 - plugins/acl-plugin/test/run-python | 28 - plugins/acl-plugin/test/run-scapy | 26 - plugins/acl-plugin/test/test_acl_plugin.py | 118 -- plugins/configure.ac | 5 - plugins/ioam-plugin/Makefile.am | 338 ---- plugins/ioam-plugin/configure.ac | 25 - plugins/ioam-plugin/ioam/dir.dox | 18 - plugins/ioam-plugin/ioam/encap/ip6_ioam_e2e.c | 232 --- plugins/ioam-plugin/ioam/encap/ip6_ioam_e2e.h | 47 - plugins/ioam-plugin/ioam/encap/ip6_ioam_pot.c | 276 --- plugins/ioam-plugin/ioam/encap/ip6_ioam_seqno.c | 109 -- plugins/ioam-plugin/ioam/encap/ip6_ioam_seqno.h | 70 - .../ioam/encap/ip6_ioam_seqno_analyse.c | 141 -- plugins/ioam-plugin/ioam/encap/ip6_ioam_trace.c | 438 ----- .../ioam-plugin/ioam/export-common/ioam_export.h | 616 ------ .../export-vxlan-gpe/vxlan_gpe_ioam_export.api | 42 - .../ioam/export-vxlan-gpe/vxlan_gpe_ioam_export.c | 271 --- .../vxlan_gpe_ioam_export_all_api_h.h | 16 - .../vxlan_gpe_ioam_export_msg_enum.h | 28 - .../export-vxlan-gpe/vxlan_gpe_ioam_export_test.c | 215 --- .../vxlan_gpe_ioam_export_thread.c | 49 - .../ioam/export-vxlan-gpe/vxlan_gpe_node.c | 162 -- plugins/ioam-plugin/ioam/export/ioam_export.api | 42 - plugins/ioam-plugin/ioam/export/ioam_export.c | 282 --- .../ioam/export/ioam_export_all_api_h.h | 16 - .../ioam-plugin/ioam/export/ioam_export_msg_enum.h | 28 - plugins/ioam-plugin/ioam/export/ioam_export_test.c | 206 --- .../ioam-plugin/ioam/export/ioam_export_thread.c | 38 - plugins/ioam-plugin/ioam/export/jvpp_ioam_export.c | 124 -- plugins/ioam-plugin/ioam/export/jvpp_ioam_export.h | 45 - plugins/ioam-plugin/ioam/export/node.c | 151 -- plugins/ioam-plugin/ioam/ioam_plugin_doc.md | 464 ----- .../jvpp/ioamexport/test/IoamExportApiTest.java | 56 - .../jvpp/io/fd/vpp/jvpp/ioamexport/test/Readme.txt | 1 - .../fd/vpp/jvpp/ioampot/test/IoamPotApiTest.java | 75 - .../jvpp/io/fd/vpp/jvpp/ioampot/test/Readme.txt | 1 - .../vpp/jvpp/ioamtrace/test/IoamTraceApiTest.java | 77 - .../jvpp/io/fd/vpp/jvpp/ioamtrace/test/Readme.txt | 1 - plugins/ioam-plugin/ioam/lib-pot/jvpp_ioam_pot.c | 124 -- plugins/ioam-plugin/ioam/lib-pot/jvpp_ioam_pot.h | 45 - plugins/ioam-plugin/ioam/lib-pot/math64.h | 159 -- plugins/ioam-plugin/ioam/lib-pot/pot.api | 133 -- plugins/ioam-plugin/ioam/lib-pot/pot_all_api_h.h | 16 - plugins/ioam-plugin/ioam/lib-pot/pot_api.c | 292 --- plugins/ioam-plugin/ioam/lib-pot/pot_msg_enum.h | 28 - plugins/ioam-plugin/ioam/lib-pot/pot_test.c | 365 ---- plugins/ioam-plugin/ioam/lib-pot/pot_util.c | 445 ----- plugins/ioam-plugin/ioam/lib-pot/pot_util.h | 195 -- .../ioam-plugin/ioam/lib-trace/jvpp_ioam_trace.c | 124 -- .../ioam-plugin/ioam/lib-trace/jvpp_ioam_trace.h | 45 - plugins/ioam-plugin/ioam/lib-trace/trace.api | 92 - .../ioam-plugin/ioam/lib-trace/trace_all_api_h.h | 16 - plugins/ioam-plugin/ioam/lib-trace/trace_api.c | 252 --- .../ioam-plugin/ioam/lib-trace/trace_msg_enum.h | 28 - plugins/ioam-plugin/ioam/lib-trace/trace_test.c | 292 --- plugins/ioam-plugin/ioam/lib-trace/trace_util.c | 206 --- plugins/ioam-plugin/ioam/lib-trace/trace_util.h | 247 --- .../ioam-plugin/ioam/lib-vxlan-gpe/ioam_decap.c | 223 --- .../ioam-plugin/ioam/lib-vxlan-gpe/ioam_encap.c | 194 -- plugins/ioam-plugin/ioam/lib-vxlan-gpe/ioam_pop.c | 353 ---- .../ioam-plugin/ioam/lib-vxlan-gpe/ioam_transit.c | 188 -- .../ioam-plugin/ioam/lib-vxlan-gpe/vxlan_gpe.api | 181 -- .../ioam/lib-vxlan-gpe/vxlan_gpe_all_api_h.h | 16 - .../ioam-plugin/ioam/lib-vxlan-gpe/vxlan_gpe_api.c | 378 ---- .../ioam/lib-vxlan-gpe/vxlan_gpe_ioam.c | 773 -------- .../ioam/lib-vxlan-gpe/vxlan_gpe_ioam.h | 183 -- .../ioam/lib-vxlan-gpe/vxlan_gpe_ioam_packet.h | 61 - .../ioam/lib-vxlan-gpe/vxlan_gpe_ioam_trace.c | 552 ------ .../ioam/lib-vxlan-gpe/vxlan_gpe_ioam_util.h | 172 -- .../ioam/lib-vxlan-gpe/vxlan_gpe_msg_enum.h | 28 - .../ioam/lib-vxlan-gpe/vxlan_gpe_test.c | 600 ------ plugins/lb-plugin/Makefile.am | 55 - plugins/lb-plugin/configure.ac | 9 - plugins/lb-plugin/lb/api.c | 228 --- plugins/lb-plugin/lb/cli.c | 250 --- plugins/lb-plugin/lb/lb.api | 71 - plugins/lb-plugin/lb/lb.c | 844 --------- plugins/lb-plugin/lb/lb.h | 333 ---- plugins/lb-plugin/lb/lb_test.c | 293 --- plugins/lb-plugin/lb/lbhash.h | 216 --- plugins/lb-plugin/lb/node.c | 419 ----- plugins/lb-plugin/lb/refcount.c | 41 - plugins/lb-plugin/lb/refcount.h | 67 - plugins/lb-plugin/lb/util.c | 72 - plugins/lb-plugin/lb/util.h | 40 - plugins/lb-plugin/lb_plugin_doc.md | 141 -- plugins/sample-plugin/Makefile.am | 56 - plugins/sample-plugin/configure.ac | 9 - plugins/sample-plugin/sample/node.c | 295 --- plugins/sample-plugin/sample/sample.api | 39 - plugins/sample-plugin/sample/sample.c | 255 --- plugins/sample-plugin/sample/sample.h | 40 - plugins/sample-plugin/sample/sample_all_api_h.h | 16 - plugins/sample-plugin/sample/sample_msg_enum.h | 28 - plugins/sample-plugin/sample/sample_test.c | 213 --- plugins/snat-plugin/Makefile.am | 113 -- plugins/snat-plugin/configure.ac | 32 - plugins/snat-plugin/snat/in2out.c | 1597 ---------------- .../io/fd/vpp/jvpp/snat/test/CallbackApiTest.java | 68 - .../snat/jvpp/io/fd/vpp/jvpp/snat/test/Readme.txt | 1 - plugins/snat-plugin/snat/jvpp_snat.c | 124 -- plugins/snat-plugin/snat/jvpp_snat.h | 45 - plugins/snat-plugin/snat/out2in.c | 1261 ------------- plugins/snat-plugin/snat/snat.api | 283 --- plugins/snat-plugin/snat/snat.c | 1957 -------------------- plugins/snat-plugin/snat/snat.h | 259 --- plugins/snat-plugin/snat/snat_all_api_h.h | 19 - plugins/snat-plugin/snat/snat_msg_enum.h | 31 - plugins/snat-plugin/snat/snat_test.c | 602 ------ src/Makefile.am | 10 +- src/configure.ac | 42 +- src/examples/sample-plugin/Makefile.am | 56 + src/examples/sample-plugin/configure.ac | 9 + src/examples/sample-plugin/sample/node.c | 295 +++ src/examples/sample-plugin/sample/sample.api | 39 + src/examples/sample-plugin/sample/sample.c | 255 +++ src/examples/sample-plugin/sample/sample.h | 40 + .../sample-plugin/sample/sample_all_api_h.h | 16 + .../sample-plugin/sample/sample_msg_enum.h | 28 + src/examples/sample-plugin/sample/sample_test.c | 213 +++ src/m4/ax_vpp_find_jdk8.m4 | 29 + src/plugins/Makefile.am | 18 + src/plugins/acl.am | 35 + src/plugins/acl/acl.api | 444 +++++ src/plugins/acl/acl.c | 1901 +++++++++++++++++++ src/plugins/acl/acl.h | 148 ++ src/plugins/acl/acl_all_api_h.h | 321 ++++ src/plugins/acl/acl_msg_enum.h | 28 + src/plugins/acl/acl_test.c | 1024 ++++++++++ src/plugins/acl/l2sess.c | 243 +++ src/plugins/acl/l2sess.h | 150 ++ src/plugins/acl/l2sess_node.c | 816 ++++++++ src/plugins/acl/node_in.c | 168 ++ src/plugins/acl/node_in.h | 12 + src/plugins/acl/node_out.c | 175 ++ src/plugins/acl/node_out.h | 12 + src/plugins/acl/test/run-python | 28 + src/plugins/acl/test/run-scapy | 26 + src/plugins/acl/test/test_acl_plugin.py | 118 ++ src/plugins/ioam.am | 150 ++ src/plugins/ioam/dir.dox | 18 + src/plugins/ioam/encap/ip6_ioam_e2e.c | 232 +++ src/plugins/ioam/encap/ip6_ioam_e2e.h | 47 + src/plugins/ioam/encap/ip6_ioam_pot.c | 276 +++ src/plugins/ioam/encap/ip6_ioam_seqno.c | 109 ++ src/plugins/ioam/encap/ip6_ioam_seqno.h | 70 + src/plugins/ioam/encap/ip6_ioam_seqno_analyse.c | 141 ++ src/plugins/ioam/encap/ip6_ioam_trace.c | 438 +++++ src/plugins/ioam/export-common/ioam_export.h | 616 ++++++ .../export-vxlan-gpe/vxlan_gpe_ioam_export.api | 42 + .../ioam/export-vxlan-gpe/vxlan_gpe_ioam_export.c | 271 +++ .../vxlan_gpe_ioam_export_all_api_h.h | 16 + .../vxlan_gpe_ioam_export_msg_enum.h | 28 + .../export-vxlan-gpe/vxlan_gpe_ioam_export_test.c | 215 +++ .../vxlan_gpe_ioam_export_thread.c | 49 + src/plugins/ioam/export-vxlan-gpe/vxlan_gpe_node.c | 162 ++ src/plugins/ioam/export/ioam_export.api | 42 + src/plugins/ioam/export/ioam_export.c | 282 +++ src/plugins/ioam/export/ioam_export_all_api_h.h | 16 + src/plugins/ioam/export/ioam_export_msg_enum.h | 28 + src/plugins/ioam/export/ioam_export_test.c | 206 +++ src/plugins/ioam/export/ioam_export_thread.c | 38 + src/plugins/ioam/export/node.c | 151 ++ src/plugins/ioam/ioam_plugin_doc.md | 464 +++++ src/plugins/ioam/lib-pot/math64.h | 159 ++ src/plugins/ioam/lib-pot/pot.api | 133 ++ src/plugins/ioam/lib-pot/pot_all_api_h.h | 16 + src/plugins/ioam/lib-pot/pot_api.c | 292 +++ src/plugins/ioam/lib-pot/pot_msg_enum.h | 28 + src/plugins/ioam/lib-pot/pot_test.c | 365 ++++ src/plugins/ioam/lib-pot/pot_util.c | 445 +++++ src/plugins/ioam/lib-pot/pot_util.h | 195 ++ src/plugins/ioam/lib-trace/trace.api | 92 + src/plugins/ioam/lib-trace/trace_all_api_h.h | 16 + src/plugins/ioam/lib-trace/trace_api.c | 252 +++ src/plugins/ioam/lib-trace/trace_msg_enum.h | 28 + src/plugins/ioam/lib-trace/trace_test.c | 292 +++ src/plugins/ioam/lib-trace/trace_util.c | 206 +++ src/plugins/ioam/lib-trace/trace_util.h | 247 +++ src/plugins/ioam/lib-vxlan-gpe/ioam_decap.c | 223 +++ src/plugins/ioam/lib-vxlan-gpe/ioam_encap.c | 194 ++ src/plugins/ioam/lib-vxlan-gpe/ioam_pop.c | 353 ++++ src/plugins/ioam/lib-vxlan-gpe/ioam_transit.c | 188 ++ src/plugins/ioam/lib-vxlan-gpe/ioam_vxlan_gpe.api | 181 ++ .../ioam/lib-vxlan-gpe/vxlan_gpe_all_api_h.h | 16 + src/plugins/ioam/lib-vxlan-gpe/vxlan_gpe_api.c | 378 ++++ src/plugins/ioam/lib-vxlan-gpe/vxlan_gpe_ioam.c | 773 ++++++++ src/plugins/ioam/lib-vxlan-gpe/vxlan_gpe_ioam.h | 183 ++ .../ioam/lib-vxlan-gpe/vxlan_gpe_ioam_packet.h | 61 + .../ioam/lib-vxlan-gpe/vxlan_gpe_ioam_trace.c | 552 ++++++ .../ioam/lib-vxlan-gpe/vxlan_gpe_ioam_util.h | 172 ++ .../ioam/lib-vxlan-gpe/vxlan_gpe_msg_enum.h | 28 + src/plugins/ioam/lib-vxlan-gpe/vxlan_gpe_test.c | 600 ++++++ src/plugins/lb.am | 42 + src/plugins/lb/api.c | 228 +++ src/plugins/lb/cli.c | 250 +++ src/plugins/lb/lb.api | 71 + src/plugins/lb/lb.c | 844 +++++++++ src/plugins/lb/lb.h | 333 ++++ src/plugins/lb/lb_plugin_doc.md | 141 ++ src/plugins/lb/lb_test.c | 293 +++ src/plugins/lb/lbhash.h | 216 +++ src/plugins/lb/node.c | 419 +++++ src/plugins/lb/refcount.c | 41 + src/plugins/lb/refcount.h | 67 + src/plugins/lb/util.c | 72 + src/plugins/lb/util.h | 40 + src/plugins/snat.am | 33 + src/plugins/snat/in2out.c | 1597 ++++++++++++++++ src/plugins/snat/out2in.c | 1261 +++++++++++++ src/plugins/snat/snat.api | 283 +++ src/plugins/snat/snat.c | 1957 ++++++++++++++++++++ src/plugins/snat/snat.h | 259 +++ src/plugins/snat/snat_all_api_h.h | 19 + src/plugins/snat/snat_msg_enum.h | 31 + src/plugins/snat/snat_test.c | 602 ++++++ src/suffix-rules.mk | 6 +- src/tools/g2/configure.ac | 12 - src/tools/perftool/configure.ac | 12 - src/tools/vppapigen/configure.ac | 14 - src/vpp-api/java/Makefile.am | 199 ++ src/vpp-api/java/Readme.txt | 236 +++ .../fd/vpp/jvpp/acl/test/AclExpectedDumpData.java | 135 ++ .../io/fd/vpp/jvpp/acl/test/AclTestData.java | 101 + .../io/fd/vpp/jvpp/acl/test/AclTestRequests.java | 141 ++ .../io/fd/vpp/jvpp/acl/test/FutureApiTest.java | 68 + .../jvpp-acl/io/fd/vpp/jvpp/acl/test/Readme.txt | 1 + src/vpp-api/java/jvpp-acl/jvpp_acl.c | 124 ++ src/vpp-api/java/jvpp-acl/jvpp_acl.h | 45 + src/vpp-api/java/jvpp-common/jvpp_common.c | 65 + src/vpp-api/java/jvpp-common/jvpp_common.h | 67 + .../io/fd/vpp/jvpp/core/test/CallbackApiTest.java | 96 + .../test/CallbackJVppFacadeNotificationTest.java | 87 + .../vpp/jvpp/core/test/CallbackJVppFacadeTest.java | 103 ++ .../core/test/CallbackNotificationApiTest.java | 94 + .../io/fd/vpp/jvpp/core/test/ControlPingTest.java | 68 + .../vpp/jvpp/core/test/CreateSubInterfaceTest.java | 120 ++ .../jvpp/core/test/FutureApiNotificationTest.java | 55 + .../io/fd/vpp/jvpp/core/test/FutureApiTest.java | 123 ++ .../io/fd/vpp/jvpp/core/test/L2AclTest.java | 195 ++ .../fd/vpp/jvpp/core/test/LispAdjacencyTest.java | 124 ++ .../fd/vpp/jvpp/core/test/NotificationUtils.java | 53 + .../jvpp-core/io/fd/vpp/jvpp/core/test/Readme.txt | 17 + src/vpp-api/java/jvpp-core/jvpp_core.c | 117 ++ .../jvpp/ioamexport/test/IoamExportApiTest.java | 56 + .../io/fd/vpp/jvpp/ioamexport/test/Readme.txt | 1 + .../java/jvpp-ioamexport/jvpp_ioam_export.c | 124 ++ .../java/jvpp-ioamexport/jvpp_ioam_export.h | 45 + .../fd/vpp/jvpp/ioampot/test/IoamPotApiTest.java | 75 + .../io/fd/vpp/jvpp/ioampot/test/Readme.txt | 1 + src/vpp-api/java/jvpp-ioampot/jvpp_ioam_pot.c | 124 ++ src/vpp-api/java/jvpp-ioampot/jvpp_ioam_pot.h | 45 + .../vpp/jvpp/ioamtrace/test/IoamTraceApiTest.java | 77 + .../io/fd/vpp/jvpp/ioamtrace/test/Readme.txt | 1 + src/vpp-api/java/jvpp-ioamtrace/jvpp_ioam_trace.c | 124 ++ src/vpp-api/java/jvpp-ioamtrace/jvpp_ioam_trace.h | 45 + .../java/jvpp-registry/io/fd/vpp/jvpp/JVpp.java | 56 + .../jvpp-registry/io/fd/vpp/jvpp/JVppRegistry.java | 76 + .../io/fd/vpp/jvpp/JVppRegistryImpl.java | 147 ++ .../io/fd/vpp/jvpp/NativeLibraryLoader.java | 73 + .../io/fd/vpp/jvpp/VppBaseCallException.java | 60 + .../io/fd/vpp/jvpp/VppCallbackException.java | 47 + .../io/fd/vpp/jvpp/VppConnection.java | 45 + .../io/fd/vpp/jvpp/VppInvocationException.java | 33 + .../io/fd/vpp/jvpp/VppJNIConnection.java | 137 ++ .../fd/vpp/jvpp/callback/ControlPingCallback.java | 29 + .../io/fd/vpp/jvpp/callback/JVppCallback.java | 29 + .../jvpp/callback/JVppNotificationCallback.java | 24 + .../io/fd/vpp/jvpp/dto/ControlPing.java | 34 + .../io/fd/vpp/jvpp/dto/ControlPingReply.java | 58 + .../jvpp-registry/io/fd/vpp/jvpp/dto/JVppDump.java | 24 + .../io/fd/vpp/jvpp/dto/JVppNotification.java | 23 + .../io/fd/vpp/jvpp/dto/JVppReply.java | 24 + .../io/fd/vpp/jvpp/dto/JVppReplyDump.java | 25 + .../io/fd/vpp/jvpp/dto/JVppRequest.java | 34 + .../vpp/jvpp/future/AbstractFutureJVppInvoker.java | 141 ++ .../io/fd/vpp/jvpp/future/FutureJVppInvoker.java | 49 + .../jvpp/notification/NotificationRegistry.java | 25 + .../notification/NotificationRegistryProvider.java | 28 + .../io/fd/vpp/jvpp/test/ConnectionTest.java | 44 + src/vpp-api/java/jvpp-registry/jvpp_registry.c | 352 ++++ .../io/fd/vpp/jvpp/snat/test/CallbackApiTest.java | 68 + .../jvpp-snat/io/fd/vpp/jvpp/snat/test/Readme.txt | 1 + src/vpp-api/java/jvpp-snat/jvpp_snat.c | 124 ++ src/vpp-api/java/jvpp-snat/jvpp_snat.h | 45 + src/vpp-api/java/jvpp/gen/jvpp_gen.py | 185 ++ src/vpp-api/java/jvpp/gen/jvppgen/__init__.py | 0 src/vpp-api/java/jvpp/gen/jvppgen/callback_gen.py | 105 ++ src/vpp-api/java/jvpp/gen/jvppgen/dto_gen.py | 308 +++ src/vpp-api/java/jvpp/gen/jvppgen/jni_gen.py | 295 +++ src/vpp-api/java/jvpp/gen/jvppgen/jvpp_c_gen.py | 343 ++++ .../jvpp/gen/jvppgen/jvpp_callback_facade_gen.py | 324 ++++ .../jvpp/gen/jvppgen/jvpp_future_facade_gen.py | 331 ++++ src/vpp-api/java/jvpp/gen/jvppgen/jvpp_impl_gen.py | 219 +++ .../java/jvpp/gen/jvppgen/notification_gen.py | 199 ++ src/vpp-api/java/jvpp/gen/jvppgen/types_gen.py | 227 +++ src/vpp-api/java/jvpp/gen/jvppgen/util.py | 220 +++ src/vpp-api/lua/README.md | 50 + src/vpp-api/lua/bench.lua | 70 + src/vpp-api/lua/examples/cli/README.md | 5 + src/vpp-api/lua/examples/cli/lua-cli.lua | 747 ++++++++ src/vpp-api/lua/examples/example-acl-plugin.lua | 110 ++ src/vpp-api/lua/examples/example-classifier.lua | 51 + src/vpp-api/lua/examples/example-cli.lua | 44 + src/vpp-api/lua/examples/lute/README.md | 66 + src/vpp-api/lua/examples/lute/lute.lua | 777 ++++++++ .../lua/examples/lute/script-inout-acl-noacl.lute | 329 ++++ .../lua/examples/lute/script-inout-acl-old.lute | 329 ++++ .../lua/examples/lute/script-inout-acl.lute | 329 ++++ src/vpp-api/lua/examples/lute/script.lute | 7 + src/vpp-api/lua/examples/lute/sessions-acl.lute | 308 +++ src/vpp-api/lua/vpp-lapi.lua | 989 ++++++++++ src/vpp.am | 2 +- src/vppapigen.am | 6 +- vpp-api/Makefile.am | 4 - vpp-api/configure.ac | 12 - vpp-api/java/Makefile.am | 116 -- vpp-api/java/Readme.txt | 236 --- vpp-api/java/configure.ac | 24 - vpp-api/java/jvpp-common/jvpp_common.c | 65 - vpp-api/java/jvpp-common/jvpp_common.h | 67 - .../io/fd/vpp/jvpp/core/test/CallbackApiTest.java | 96 - .../test/CallbackJVppFacadeNotificationTest.java | 87 - .../vpp/jvpp/core/test/CallbackJVppFacadeTest.java | 103 -- .../core/test/CallbackNotificationApiTest.java | 94 - .../io/fd/vpp/jvpp/core/test/ControlPingTest.java | 68 - .../vpp/jvpp/core/test/CreateSubInterfaceTest.java | 120 -- .../jvpp/core/test/FutureApiNotificationTest.java | 55 - .../io/fd/vpp/jvpp/core/test/FutureApiTest.java | 123 -- .../io/fd/vpp/jvpp/core/test/L2AclTest.java | 195 -- .../fd/vpp/jvpp/core/test/LispAdjacencyTest.java | 124 -- .../fd/vpp/jvpp/core/test/NotificationUtils.java | 53 - .../jvpp-core/io/fd/vpp/jvpp/core/test/Readme.txt | 17 - vpp-api/java/jvpp-core/jvpp_core.c | 117 -- .../java/jvpp-registry/io/fd/vpp/jvpp/JVpp.java | 56 - .../jvpp-registry/io/fd/vpp/jvpp/JVppRegistry.java | 76 - .../io/fd/vpp/jvpp/JVppRegistryImpl.java | 147 -- .../io/fd/vpp/jvpp/NativeLibraryLoader.java | 73 - .../io/fd/vpp/jvpp/VppBaseCallException.java | 60 - .../io/fd/vpp/jvpp/VppCallbackException.java | 47 - .../io/fd/vpp/jvpp/VppConnection.java | 45 - .../io/fd/vpp/jvpp/VppInvocationException.java | 33 - .../io/fd/vpp/jvpp/VppJNIConnection.java | 137 -- .../fd/vpp/jvpp/callback/ControlPingCallback.java | 29 - .../io/fd/vpp/jvpp/callback/JVppCallback.java | 29 - .../jvpp/callback/JVppNotificationCallback.java | 24 - .../io/fd/vpp/jvpp/dto/ControlPing.java | 34 - .../io/fd/vpp/jvpp/dto/ControlPingReply.java | 58 - .../jvpp-registry/io/fd/vpp/jvpp/dto/JVppDump.java | 24 - .../io/fd/vpp/jvpp/dto/JVppNotification.java | 23 - .../io/fd/vpp/jvpp/dto/JVppReply.java | 24 - .../io/fd/vpp/jvpp/dto/JVppReplyDump.java | 25 - .../io/fd/vpp/jvpp/dto/JVppRequest.java | 34 - .../vpp/jvpp/future/AbstractFutureJVppInvoker.java | 141 -- .../io/fd/vpp/jvpp/future/FutureJVppInvoker.java | 49 - .../jvpp/notification/NotificationRegistry.java | 25 - .../notification/NotificationRegistryProvider.java | 28 - .../io/fd/vpp/jvpp/test/ConnectionTest.java | 44 - vpp-api/java/jvpp-registry/jvpp_registry.c | 352 ---- vpp-api/java/jvpp/gen/jvpp_gen.py | 171 -- vpp-api/java/jvpp/gen/jvppgen/__init__.py | 0 vpp-api/java/jvpp/gen/jvppgen/callback_gen.py | 105 -- vpp-api/java/jvpp/gen/jvppgen/dto_gen.py | 308 --- vpp-api/java/jvpp/gen/jvppgen/jni_gen.py | 295 --- vpp-api/java/jvpp/gen/jvppgen/jvpp_c_gen.py | 343 ---- .../jvpp/gen/jvppgen/jvpp_callback_facade_gen.py | 324 ---- .../jvpp/gen/jvppgen/jvpp_future_facade_gen.py | 331 ---- vpp-api/java/jvpp/gen/jvppgen/jvpp_impl_gen.py | 219 --- vpp-api/java/jvpp/gen/jvppgen/notification_gen.py | 199 -- vpp-api/java/jvpp/gen/jvppgen/types_gen.py | 227 --- vpp-api/java/jvpp/gen/jvppgen/util.py | 220 --- vpp-api/java/m4/ax_check_java_home.m4 | 80 - vpp-api/java/m4/ax_check_java_plugin.m4 | 101 - vpp-api/java/m4/ax_java_check_class.m4 | 85 - vpp-api/java/m4/ax_java_options.m4 | 48 - vpp-api/java/m4/ax_libgcj_jar.m4 | 83 - vpp-api/java/m4/ax_prog_jar.m4 | 49 - vpp-api/java/m4/ax_prog_java.m4 | 115 -- vpp-api/java/m4/ax_prog_java_cc.m4 | 104 -- vpp-api/java/m4/ax_prog_java_works.m4 | 134 -- vpp-api/java/m4/ax_prog_javac.m4 | 79 - vpp-api/java/m4/ax_prog_javac_works.m4 | 72 - vpp-api/java/m4/ax_prog_javadoc.m4 | 50 - vpp-api/java/m4/ax_prog_javah.m4 | 64 - vpp-api/java/m4/ax_try_compile_java.m4 | 55 - vpp-api/java/m4/ax_try_run_java.m4 | 56 - vpp-api/lua/README.md | 50 - vpp-api/lua/bench.lua | 70 - vpp-api/lua/examples/cli/README.md | 5 - vpp-api/lua/examples/cli/lua-cli.lua | 747 -------- vpp-api/lua/examples/example-acl-plugin.lua | 110 -- vpp-api/lua/examples/example-classifier.lua | 51 - vpp-api/lua/examples/example-cli.lua | 44 - vpp-api/lua/examples/lute/README.md | 66 - vpp-api/lua/examples/lute/lute.lua | 777 -------- .../lua/examples/lute/script-inout-acl-noacl.lute | 329 ---- .../lua/examples/lute/script-inout-acl-old.lute | 329 ---- vpp-api/lua/examples/lute/script-inout-acl.lute | 329 ---- vpp-api/lua/examples/lute/script.lute | 7 - vpp-api/lua/examples/lute/sessions-acl.lute | 308 --- vpp-api/lua/vpp-lapi.lua | 989 ---------- 444 files changed, 38673 insertions(+), 40252 deletions(-) delete mode 100644 build-data/packages/plugins.mk delete mode 100644 build-data/packages/vnet.mk delete mode 100644 build-data/packages/vpp-api.mk delete mode 100755 build-root/scripts/find-python-api-contents create mode 100755 build-root/scripts/find-vpp-api-java-contents create mode 100755 build-root/scripts/find-vpp-api-lua-contents create mode 100755 build-root/scripts/find-vpp-api-python-contents delete mode 100644 plugins/acl-plugin/Makefile.am delete mode 100644 plugins/acl-plugin/acl/acl.api delete mode 100644 plugins/acl-plugin/acl/acl.c delete mode 100644 plugins/acl-plugin/acl/acl.h delete mode 100644 plugins/acl-plugin/acl/acl_all_api_h.h delete mode 100644 plugins/acl-plugin/acl/acl_msg_enum.h delete mode 100644 plugins/acl-plugin/acl/acl_test.c delete mode 100644 plugins/acl-plugin/acl/jvpp/io/fd/vpp/jvpp/acl/test/AclExpectedDumpData.java delete mode 100644 plugins/acl-plugin/acl/jvpp/io/fd/vpp/jvpp/acl/test/AclTestData.java delete mode 100644 plugins/acl-plugin/acl/jvpp/io/fd/vpp/jvpp/acl/test/AclTestRequests.java delete mode 100644 plugins/acl-plugin/acl/jvpp/io/fd/vpp/jvpp/acl/test/FutureApiTest.java delete mode 100644 plugins/acl-plugin/acl/jvpp/io/fd/vpp/jvpp/acl/test/Readme.txt delete mode 100644 plugins/acl-plugin/acl/jvpp_acl.c delete mode 100644 plugins/acl-plugin/acl/jvpp_acl.h delete mode 100644 plugins/acl-plugin/acl/l2sess.c delete mode 100644 plugins/acl-plugin/acl/l2sess.h delete mode 100644 plugins/acl-plugin/acl/l2sess_node.c delete mode 100644 plugins/acl-plugin/acl/node_in.c delete mode 100644 plugins/acl-plugin/acl/node_in.h delete mode 100644 plugins/acl-plugin/acl/node_out.c delete mode 100644 plugins/acl-plugin/acl/node_out.h delete mode 100644 plugins/acl-plugin/configure.ac delete mode 100755 plugins/acl-plugin/test/run-python delete mode 100755 plugins/acl-plugin/test/run-scapy delete mode 100644 plugins/acl-plugin/test/test_acl_plugin.py delete mode 100644 plugins/ioam-plugin/Makefile.am delete mode 100644 plugins/ioam-plugin/configure.ac delete mode 100644 plugins/ioam-plugin/ioam/dir.dox delete mode 100644 plugins/ioam-plugin/ioam/encap/ip6_ioam_e2e.c delete mode 100644 plugins/ioam-plugin/ioam/encap/ip6_ioam_e2e.h delete mode 100644 plugins/ioam-plugin/ioam/encap/ip6_ioam_pot.c delete mode 100644 plugins/ioam-plugin/ioam/encap/ip6_ioam_seqno.c delete mode 100644 plugins/ioam-plugin/ioam/encap/ip6_ioam_seqno.h delete mode 100644 plugins/ioam-plugin/ioam/encap/ip6_ioam_seqno_analyse.c delete mode 100644 plugins/ioam-plugin/ioam/encap/ip6_ioam_trace.c delete mode 100644 plugins/ioam-plugin/ioam/export-common/ioam_export.h delete mode 100644 plugins/ioam-plugin/ioam/export-vxlan-gpe/vxlan_gpe_ioam_export.api delete mode 100644 plugins/ioam-plugin/ioam/export-vxlan-gpe/vxlan_gpe_ioam_export.c delete mode 100644 plugins/ioam-plugin/ioam/export-vxlan-gpe/vxlan_gpe_ioam_export_all_api_h.h delete mode 100644 plugins/ioam-plugin/ioam/export-vxlan-gpe/vxlan_gpe_ioam_export_msg_enum.h delete mode 100644 plugins/ioam-plugin/ioam/export-vxlan-gpe/vxlan_gpe_ioam_export_test.c delete mode 100644 plugins/ioam-plugin/ioam/export-vxlan-gpe/vxlan_gpe_ioam_export_thread.c delete mode 100644 plugins/ioam-plugin/ioam/export-vxlan-gpe/vxlan_gpe_node.c delete mode 100644 plugins/ioam-plugin/ioam/export/ioam_export.api delete mode 100644 plugins/ioam-plugin/ioam/export/ioam_export.c delete mode 100644 plugins/ioam-plugin/ioam/export/ioam_export_all_api_h.h delete mode 100644 plugins/ioam-plugin/ioam/export/ioam_export_msg_enum.h delete mode 100644 plugins/ioam-plugin/ioam/export/ioam_export_test.c delete mode 100644 plugins/ioam-plugin/ioam/export/ioam_export_thread.c delete mode 100644 plugins/ioam-plugin/ioam/export/jvpp_ioam_export.c delete mode 100644 plugins/ioam-plugin/ioam/export/jvpp_ioam_export.h delete mode 100644 plugins/ioam-plugin/ioam/export/node.c delete mode 100644 plugins/ioam-plugin/ioam/ioam_plugin_doc.md delete mode 100644 plugins/ioam-plugin/ioam/jvpp/io/fd/vpp/jvpp/ioamexport/test/IoamExportApiTest.java delete mode 100644 plugins/ioam-plugin/ioam/jvpp/io/fd/vpp/jvpp/ioamexport/test/Readme.txt delete mode 100644 plugins/ioam-plugin/ioam/jvpp/io/fd/vpp/jvpp/ioampot/test/IoamPotApiTest.java delete mode 100644 plugins/ioam-plugin/ioam/jvpp/io/fd/vpp/jvpp/ioampot/test/Readme.txt delete mode 100644 plugins/ioam-plugin/ioam/jvpp/io/fd/vpp/jvpp/ioamtrace/test/IoamTraceApiTest.java delete mode 100644 plugins/ioam-plugin/ioam/jvpp/io/fd/vpp/jvpp/ioamtrace/test/Readme.txt delete mode 100644 plugins/ioam-plugin/ioam/lib-pot/jvpp_ioam_pot.c delete mode 100644 plugins/ioam-plugin/ioam/lib-pot/jvpp_ioam_pot.h delete mode 100644 plugins/ioam-plugin/ioam/lib-pot/math64.h delete mode 100644 plugins/ioam-plugin/ioam/lib-pot/pot.api delete mode 100644 plugins/ioam-plugin/ioam/lib-pot/pot_all_api_h.h delete mode 100644 plugins/ioam-plugin/ioam/lib-pot/pot_api.c delete mode 100644 plugins/ioam-plugin/ioam/lib-pot/pot_msg_enum.h delete mode 100644 plugins/ioam-plugin/ioam/lib-pot/pot_test.c delete mode 100644 plugins/ioam-plugin/ioam/lib-pot/pot_util.c delete mode 100644 plugins/ioam-plugin/ioam/lib-pot/pot_util.h delete mode 100644 plugins/ioam-plugin/ioam/lib-trace/jvpp_ioam_trace.c delete mode 100644 plugins/ioam-plugin/ioam/lib-trace/jvpp_ioam_trace.h delete mode 100644 plugins/ioam-plugin/ioam/lib-trace/trace.api delete mode 100644 plugins/ioam-plugin/ioam/lib-trace/trace_all_api_h.h delete mode 100644 plugins/ioam-plugin/ioam/lib-trace/trace_api.c delete mode 100644 plugins/ioam-plugin/ioam/lib-trace/trace_msg_enum.h delete mode 100644 plugins/ioam-plugin/ioam/lib-trace/trace_test.c delete mode 100644 plugins/ioam-plugin/ioam/lib-trace/trace_util.c delete mode 100644 plugins/ioam-plugin/ioam/lib-trace/trace_util.h delete mode 100644 plugins/ioam-plugin/ioam/lib-vxlan-gpe/ioam_decap.c delete mode 100644 plugins/ioam-plugin/ioam/lib-vxlan-gpe/ioam_encap.c delete mode 100644 plugins/ioam-plugin/ioam/lib-vxlan-gpe/ioam_pop.c delete mode 100644 plugins/ioam-plugin/ioam/lib-vxlan-gpe/ioam_transit.c delete mode 100644 plugins/ioam-plugin/ioam/lib-vxlan-gpe/vxlan_gpe.api delete mode 100644 plugins/ioam-plugin/ioam/lib-vxlan-gpe/vxlan_gpe_all_api_h.h delete mode 100644 plugins/ioam-plugin/ioam/lib-vxlan-gpe/vxlan_gpe_api.c delete mode 100644 plugins/ioam-plugin/ioam/lib-vxlan-gpe/vxlan_gpe_ioam.c delete mode 100644 plugins/ioam-plugin/ioam/lib-vxlan-gpe/vxlan_gpe_ioam.h delete mode 100644 plugins/ioam-plugin/ioam/lib-vxlan-gpe/vxlan_gpe_ioam_packet.h delete mode 100644 plugins/ioam-plugin/ioam/lib-vxlan-gpe/vxlan_gpe_ioam_trace.c delete mode 100644 plugins/ioam-plugin/ioam/lib-vxlan-gpe/vxlan_gpe_ioam_util.h delete mode 100644 plugins/ioam-plugin/ioam/lib-vxlan-gpe/vxlan_gpe_msg_enum.h delete mode 100644 plugins/ioam-plugin/ioam/lib-vxlan-gpe/vxlan_gpe_test.c delete mode 100644 plugins/lb-plugin/Makefile.am delete mode 100644 plugins/lb-plugin/configure.ac delete mode 100644 plugins/lb-plugin/lb/api.c delete mode 100644 plugins/lb-plugin/lb/cli.c delete mode 100644 plugins/lb-plugin/lb/lb.api delete mode 100644 plugins/lb-plugin/lb/lb.c delete mode 100644 plugins/lb-plugin/lb/lb.h delete mode 100644 plugins/lb-plugin/lb/lb_test.c delete mode 100644 plugins/lb-plugin/lb/lbhash.h delete mode 100644 plugins/lb-plugin/lb/node.c delete mode 100644 plugins/lb-plugin/lb/refcount.c delete mode 100644 plugins/lb-plugin/lb/refcount.h delete mode 100644 plugins/lb-plugin/lb/util.c delete mode 100644 plugins/lb-plugin/lb/util.h delete mode 100644 plugins/lb-plugin/lb_plugin_doc.md delete mode 100644 plugins/sample-plugin/Makefile.am delete mode 100644 plugins/sample-plugin/configure.ac delete mode 100644 plugins/sample-plugin/sample/node.c delete mode 100644 plugins/sample-plugin/sample/sample.api delete mode 100644 plugins/sample-plugin/sample/sample.c delete mode 100644 plugins/sample-plugin/sample/sample.h delete mode 100644 plugins/sample-plugin/sample/sample_all_api_h.h delete mode 100644 plugins/sample-plugin/sample/sample_msg_enum.h delete mode 100644 plugins/sample-plugin/sample/sample_test.c delete mode 100644 plugins/snat-plugin/Makefile.am delete mode 100644 plugins/snat-plugin/configure.ac delete mode 100644 plugins/snat-plugin/snat/in2out.c delete mode 100644 plugins/snat-plugin/snat/jvpp/io/fd/vpp/jvpp/snat/test/CallbackApiTest.java delete mode 100644 plugins/snat-plugin/snat/jvpp/io/fd/vpp/jvpp/snat/test/Readme.txt delete mode 100644 plugins/snat-plugin/snat/jvpp_snat.c delete mode 100644 plugins/snat-plugin/snat/jvpp_snat.h delete mode 100644 plugins/snat-plugin/snat/out2in.c delete mode 100644 plugins/snat-plugin/snat/snat.api delete mode 100644 plugins/snat-plugin/snat/snat.c delete mode 100644 plugins/snat-plugin/snat/snat.h delete mode 100644 plugins/snat-plugin/snat/snat_all_api_h.h delete mode 100644 plugins/snat-plugin/snat/snat_msg_enum.h delete mode 100644 plugins/snat-plugin/snat/snat_test.c create mode 100644 src/examples/sample-plugin/Makefile.am create mode 100644 src/examples/sample-plugin/configure.ac create mode 100644 src/examples/sample-plugin/sample/node.c create mode 100644 src/examples/sample-plugin/sample/sample.api create mode 100644 src/examples/sample-plugin/sample/sample.c create mode 100644 src/examples/sample-plugin/sample/sample.h create mode 100644 src/examples/sample-plugin/sample/sample_all_api_h.h create mode 100644 src/examples/sample-plugin/sample/sample_msg_enum.h create mode 100644 src/examples/sample-plugin/sample/sample_test.c create mode 100644 src/m4/ax_vpp_find_jdk8.m4 create mode 100644 src/plugins/acl.am create mode 100644 src/plugins/acl/acl.api create mode 100644 src/plugins/acl/acl.c create mode 100644 src/plugins/acl/acl.h create mode 100644 src/plugins/acl/acl_all_api_h.h create mode 100644 src/plugins/acl/acl_msg_enum.h create mode 100644 src/plugins/acl/acl_test.c create mode 100644 src/plugins/acl/l2sess.c create mode 100644 src/plugins/acl/l2sess.h create mode 100644 src/plugins/acl/l2sess_node.c create mode 100644 src/plugins/acl/node_in.c create mode 100644 src/plugins/acl/node_in.h create mode 100644 src/plugins/acl/node_out.c create mode 100644 src/plugins/acl/node_out.h create mode 100755 src/plugins/acl/test/run-python create mode 100755 src/plugins/acl/test/run-scapy create mode 100644 src/plugins/acl/test/test_acl_plugin.py create mode 100644 src/plugins/ioam.am create mode 100644 src/plugins/ioam/dir.dox create mode 100644 src/plugins/ioam/encap/ip6_ioam_e2e.c create mode 100644 src/plugins/ioam/encap/ip6_ioam_e2e.h create mode 100644 src/plugins/ioam/encap/ip6_ioam_pot.c create mode 100644 src/plugins/ioam/encap/ip6_ioam_seqno.c create mode 100644 src/plugins/ioam/encap/ip6_ioam_seqno.h create mode 100644 src/plugins/ioam/encap/ip6_ioam_seqno_analyse.c create mode 100644 src/plugins/ioam/encap/ip6_ioam_trace.c create mode 100644 src/plugins/ioam/export-common/ioam_export.h create mode 100644 src/plugins/ioam/export-vxlan-gpe/vxlan_gpe_ioam_export.api create mode 100644 src/plugins/ioam/export-vxlan-gpe/vxlan_gpe_ioam_export.c create mode 100644 src/plugins/ioam/export-vxlan-gpe/vxlan_gpe_ioam_export_all_api_h.h create mode 100644 src/plugins/ioam/export-vxlan-gpe/vxlan_gpe_ioam_export_msg_enum.h create mode 100644 src/plugins/ioam/export-vxlan-gpe/vxlan_gpe_ioam_export_test.c create mode 100644 src/plugins/ioam/export-vxlan-gpe/vxlan_gpe_ioam_export_thread.c create mode 100644 src/plugins/ioam/export-vxlan-gpe/vxlan_gpe_node.c create mode 100644 src/plugins/ioam/export/ioam_export.api create mode 100644 src/plugins/ioam/export/ioam_export.c create mode 100644 src/plugins/ioam/export/ioam_export_all_api_h.h create mode 100644 src/plugins/ioam/export/ioam_export_msg_enum.h create mode 100644 src/plugins/ioam/export/ioam_export_test.c create mode 100644 src/plugins/ioam/export/ioam_export_thread.c create mode 100644 src/plugins/ioam/export/node.c create mode 100644 src/plugins/ioam/ioam_plugin_doc.md create mode 100644 src/plugins/ioam/lib-pot/math64.h create mode 100644 src/plugins/ioam/lib-pot/pot.api create mode 100644 src/plugins/ioam/lib-pot/pot_all_api_h.h create mode 100644 src/plugins/ioam/lib-pot/pot_api.c create mode 100644 src/plugins/ioam/lib-pot/pot_msg_enum.h create mode 100644 src/plugins/ioam/lib-pot/pot_test.c create mode 100644 src/plugins/ioam/lib-pot/pot_util.c create mode 100644 src/plugins/ioam/lib-pot/pot_util.h create mode 100644 src/plugins/ioam/lib-trace/trace.api create mode 100644 src/plugins/ioam/lib-trace/trace_all_api_h.h create mode 100644 src/plugins/ioam/lib-trace/trace_api.c create mode 100644 src/plugins/ioam/lib-trace/trace_msg_enum.h create mode 100644 src/plugins/ioam/lib-trace/trace_test.c create mode 100644 src/plugins/ioam/lib-trace/trace_util.c create mode 100644 src/plugins/ioam/lib-trace/trace_util.h create mode 100644 src/plugins/ioam/lib-vxlan-gpe/ioam_decap.c create mode 100644 src/plugins/ioam/lib-vxlan-gpe/ioam_encap.c create mode 100644 src/plugins/ioam/lib-vxlan-gpe/ioam_pop.c create mode 100644 src/plugins/ioam/lib-vxlan-gpe/ioam_transit.c create mode 100644 src/plugins/ioam/lib-vxlan-gpe/ioam_vxlan_gpe.api create mode 100644 src/plugins/ioam/lib-vxlan-gpe/vxlan_gpe_all_api_h.h create mode 100644 src/plugins/ioam/lib-vxlan-gpe/vxlan_gpe_api.c create mode 100644 src/plugins/ioam/lib-vxlan-gpe/vxlan_gpe_ioam.c create mode 100644 src/plugins/ioam/lib-vxlan-gpe/vxlan_gpe_ioam.h create mode 100644 src/plugins/ioam/lib-vxlan-gpe/vxlan_gpe_ioam_packet.h create mode 100644 src/plugins/ioam/lib-vxlan-gpe/vxlan_gpe_ioam_trace.c create mode 100644 src/plugins/ioam/lib-vxlan-gpe/vxlan_gpe_ioam_util.h create mode 100644 src/plugins/ioam/lib-vxlan-gpe/vxlan_gpe_msg_enum.h create mode 100644 src/plugins/ioam/lib-vxlan-gpe/vxlan_gpe_test.c create mode 100644 src/plugins/lb.am create mode 100644 src/plugins/lb/api.c create mode 100644 src/plugins/lb/cli.c create mode 100644 src/plugins/lb/lb.api create mode 100644 src/plugins/lb/lb.c create mode 100644 src/plugins/lb/lb.h create mode 100644 src/plugins/lb/lb_plugin_doc.md create mode 100644 src/plugins/lb/lb_test.c create mode 100644 src/plugins/lb/lbhash.h create mode 100644 src/plugins/lb/node.c create mode 100644 src/plugins/lb/refcount.c create mode 100644 src/plugins/lb/refcount.h create mode 100644 src/plugins/lb/util.c create mode 100644 src/plugins/lb/util.h create mode 100644 src/plugins/snat.am create mode 100644 src/plugins/snat/in2out.c create mode 100644 src/plugins/snat/out2in.c create mode 100644 src/plugins/snat/snat.api create mode 100644 src/plugins/snat/snat.c create mode 100644 src/plugins/snat/snat.h create mode 100644 src/plugins/snat/snat_all_api_h.h create mode 100644 src/plugins/snat/snat_msg_enum.h create mode 100644 src/plugins/snat/snat_test.c delete mode 100644 src/tools/g2/configure.ac delete mode 100644 src/tools/perftool/configure.ac delete mode 100644 src/tools/vppapigen/configure.ac create mode 100644 src/vpp-api/java/Makefile.am create mode 100644 src/vpp-api/java/Readme.txt create mode 100644 src/vpp-api/java/jvpp-acl/io/fd/vpp/jvpp/acl/test/AclExpectedDumpData.java create mode 100644 src/vpp-api/java/jvpp-acl/io/fd/vpp/jvpp/acl/test/AclTestData.java create mode 100644 src/vpp-api/java/jvpp-acl/io/fd/vpp/jvpp/acl/test/AclTestRequests.java create mode 100644 src/vpp-api/java/jvpp-acl/io/fd/vpp/jvpp/acl/test/FutureApiTest.java create mode 100644 src/vpp-api/java/jvpp-acl/io/fd/vpp/jvpp/acl/test/Readme.txt create mode 100644 src/vpp-api/java/jvpp-acl/jvpp_acl.c create mode 100644 src/vpp-api/java/jvpp-acl/jvpp_acl.h create mode 100644 src/vpp-api/java/jvpp-common/jvpp_common.c create mode 100644 src/vpp-api/java/jvpp-common/jvpp_common.h create mode 100644 src/vpp-api/java/jvpp-core/io/fd/vpp/jvpp/core/test/CallbackApiTest.java create mode 100644 src/vpp-api/java/jvpp-core/io/fd/vpp/jvpp/core/test/CallbackJVppFacadeNotificationTest.java create mode 100644 src/vpp-api/java/jvpp-core/io/fd/vpp/jvpp/core/test/CallbackJVppFacadeTest.java create mode 100644 src/vpp-api/java/jvpp-core/io/fd/vpp/jvpp/core/test/CallbackNotificationApiTest.java create mode 100644 src/vpp-api/java/jvpp-core/io/fd/vpp/jvpp/core/test/ControlPingTest.java create mode 100644 src/vpp-api/java/jvpp-core/io/fd/vpp/jvpp/core/test/CreateSubInterfaceTest.java create mode 100644 src/vpp-api/java/jvpp-core/io/fd/vpp/jvpp/core/test/FutureApiNotificationTest.java create mode 100644 src/vpp-api/java/jvpp-core/io/fd/vpp/jvpp/core/test/FutureApiTest.java create mode 100644 src/vpp-api/java/jvpp-core/io/fd/vpp/jvpp/core/test/L2AclTest.java create mode 100644 src/vpp-api/java/jvpp-core/io/fd/vpp/jvpp/core/test/LispAdjacencyTest.java create mode 100644 src/vpp-api/java/jvpp-core/io/fd/vpp/jvpp/core/test/NotificationUtils.java create mode 100644 src/vpp-api/java/jvpp-core/io/fd/vpp/jvpp/core/test/Readme.txt create mode 100644 src/vpp-api/java/jvpp-core/jvpp_core.c create mode 100644 src/vpp-api/java/jvpp-ioamexport/io/fd/vpp/jvpp/ioamexport/test/IoamExportApiTest.java create mode 100644 src/vpp-api/java/jvpp-ioamexport/io/fd/vpp/jvpp/ioamexport/test/Readme.txt create mode 100644 src/vpp-api/java/jvpp-ioamexport/jvpp_ioam_export.c create mode 100644 src/vpp-api/java/jvpp-ioamexport/jvpp_ioam_export.h create mode 100644 src/vpp-api/java/jvpp-ioampot/io/fd/vpp/jvpp/ioampot/test/IoamPotApiTest.java create mode 100644 src/vpp-api/java/jvpp-ioampot/io/fd/vpp/jvpp/ioampot/test/Readme.txt create mode 100644 src/vpp-api/java/jvpp-ioampot/jvpp_ioam_pot.c create mode 100644 src/vpp-api/java/jvpp-ioampot/jvpp_ioam_pot.h create mode 100644 src/vpp-api/java/jvpp-ioamtrace/io/fd/vpp/jvpp/ioamtrace/test/IoamTraceApiTest.java create mode 100644 src/vpp-api/java/jvpp-ioamtrace/io/fd/vpp/jvpp/ioamtrace/test/Readme.txt create mode 100644 src/vpp-api/java/jvpp-ioamtrace/jvpp_ioam_trace.c create mode 100644 src/vpp-api/java/jvpp-ioamtrace/jvpp_ioam_trace.h create mode 100644 src/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/JVpp.java create mode 100644 src/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/JVppRegistry.java create mode 100644 src/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/JVppRegistryImpl.java create mode 100644 src/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/NativeLibraryLoader.java create mode 100644 src/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/VppBaseCallException.java create mode 100644 src/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/VppCallbackException.java create mode 100644 src/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/VppConnection.java create mode 100644 src/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/VppInvocationException.java create mode 100644 src/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/VppJNIConnection.java create mode 100644 src/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/callback/ControlPingCallback.java create mode 100644 src/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/callback/JVppCallback.java create mode 100644 src/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/callback/JVppNotificationCallback.java create mode 100644 src/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/dto/ControlPing.java create mode 100644 src/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/dto/ControlPingReply.java create mode 100644 src/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/dto/JVppDump.java create mode 100644 src/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/dto/JVppNotification.java create mode 100644 src/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/dto/JVppReply.java create mode 100644 src/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/dto/JVppReplyDump.java create mode 100644 src/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/dto/JVppRequest.java create mode 100644 src/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/future/AbstractFutureJVppInvoker.java create mode 100644 src/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/future/FutureJVppInvoker.java create mode 100644 src/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/notification/NotificationRegistry.java create mode 100644 src/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/notification/NotificationRegistryProvider.java create mode 100644 src/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/test/ConnectionTest.java create mode 100644 src/vpp-api/java/jvpp-registry/jvpp_registry.c create mode 100644 src/vpp-api/java/jvpp-snat/io/fd/vpp/jvpp/snat/test/CallbackApiTest.java create mode 100644 src/vpp-api/java/jvpp-snat/io/fd/vpp/jvpp/snat/test/Readme.txt create mode 100644 src/vpp-api/java/jvpp-snat/jvpp_snat.c create mode 100644 src/vpp-api/java/jvpp-snat/jvpp_snat.h create mode 100755 src/vpp-api/java/jvpp/gen/jvpp_gen.py create mode 100644 src/vpp-api/java/jvpp/gen/jvppgen/__init__.py create mode 100644 src/vpp-api/java/jvpp/gen/jvppgen/callback_gen.py create mode 100644 src/vpp-api/java/jvpp/gen/jvppgen/dto_gen.py create mode 100644 src/vpp-api/java/jvpp/gen/jvppgen/jni_gen.py create mode 100644 src/vpp-api/java/jvpp/gen/jvppgen/jvpp_c_gen.py create mode 100644 src/vpp-api/java/jvpp/gen/jvppgen/jvpp_callback_facade_gen.py create mode 100644 src/vpp-api/java/jvpp/gen/jvppgen/jvpp_future_facade_gen.py create mode 100644 src/vpp-api/java/jvpp/gen/jvppgen/jvpp_impl_gen.py create mode 100644 src/vpp-api/java/jvpp/gen/jvppgen/notification_gen.py create mode 100644 src/vpp-api/java/jvpp/gen/jvppgen/types_gen.py create mode 100644 src/vpp-api/java/jvpp/gen/jvppgen/util.py create mode 100644 src/vpp-api/lua/README.md create mode 100644 src/vpp-api/lua/bench.lua create mode 100644 src/vpp-api/lua/examples/cli/README.md create mode 100644 src/vpp-api/lua/examples/cli/lua-cli.lua create mode 100644 src/vpp-api/lua/examples/example-acl-plugin.lua create mode 100644 src/vpp-api/lua/examples/example-classifier.lua create mode 100644 src/vpp-api/lua/examples/example-cli.lua create mode 100644 src/vpp-api/lua/examples/lute/README.md create mode 100644 src/vpp-api/lua/examples/lute/lute.lua create mode 100644 src/vpp-api/lua/examples/lute/script-inout-acl-noacl.lute create mode 100644 src/vpp-api/lua/examples/lute/script-inout-acl-old.lute create mode 100644 src/vpp-api/lua/examples/lute/script-inout-acl.lute create mode 100644 src/vpp-api/lua/examples/lute/script.lute create mode 100644 src/vpp-api/lua/examples/lute/sessions-acl.lute create mode 100644 src/vpp-api/lua/vpp-lapi.lua delete mode 100644 vpp-api/Makefile.am delete mode 100644 vpp-api/configure.ac delete mode 100644 vpp-api/java/Makefile.am delete mode 100644 vpp-api/java/Readme.txt delete mode 100644 vpp-api/java/configure.ac delete mode 100644 vpp-api/java/jvpp-common/jvpp_common.c delete mode 100644 vpp-api/java/jvpp-common/jvpp_common.h delete mode 100644 vpp-api/java/jvpp-core/io/fd/vpp/jvpp/core/test/CallbackApiTest.java delete mode 100644 vpp-api/java/jvpp-core/io/fd/vpp/jvpp/core/test/CallbackJVppFacadeNotificationTest.java delete mode 100644 vpp-api/java/jvpp-core/io/fd/vpp/jvpp/core/test/CallbackJVppFacadeTest.java delete mode 100644 vpp-api/java/jvpp-core/io/fd/vpp/jvpp/core/test/CallbackNotificationApiTest.java delete mode 100644 vpp-api/java/jvpp-core/io/fd/vpp/jvpp/core/test/ControlPingTest.java delete mode 100644 vpp-api/java/jvpp-core/io/fd/vpp/jvpp/core/test/CreateSubInterfaceTest.java delete mode 100644 vpp-api/java/jvpp-core/io/fd/vpp/jvpp/core/test/FutureApiNotificationTest.java delete mode 100644 vpp-api/java/jvpp-core/io/fd/vpp/jvpp/core/test/FutureApiTest.java delete mode 100644 vpp-api/java/jvpp-core/io/fd/vpp/jvpp/core/test/L2AclTest.java delete mode 100644 vpp-api/java/jvpp-core/io/fd/vpp/jvpp/core/test/LispAdjacencyTest.java delete mode 100644 vpp-api/java/jvpp-core/io/fd/vpp/jvpp/core/test/NotificationUtils.java delete mode 100644 vpp-api/java/jvpp-core/io/fd/vpp/jvpp/core/test/Readme.txt delete mode 100644 vpp-api/java/jvpp-core/jvpp_core.c delete mode 100644 vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/JVpp.java delete mode 100644 vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/JVppRegistry.java delete mode 100644 vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/JVppRegistryImpl.java delete mode 100644 vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/NativeLibraryLoader.java delete mode 100644 vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/VppBaseCallException.java delete mode 100644 vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/VppCallbackException.java delete mode 100644 vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/VppConnection.java delete mode 100644 vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/VppInvocationException.java delete mode 100644 vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/VppJNIConnection.java delete mode 100644 vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/callback/ControlPingCallback.java delete mode 100644 vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/callback/JVppCallback.java delete mode 100644 vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/callback/JVppNotificationCallback.java delete mode 100644 vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/dto/ControlPing.java delete mode 100644 vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/dto/ControlPingReply.java delete mode 100644 vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/dto/JVppDump.java delete mode 100644 vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/dto/JVppNotification.java delete mode 100644 vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/dto/JVppReply.java delete mode 100644 vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/dto/JVppReplyDump.java delete mode 100644 vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/dto/JVppRequest.java delete mode 100644 vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/future/AbstractFutureJVppInvoker.java delete mode 100644 vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/future/FutureJVppInvoker.java delete mode 100644 vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/notification/NotificationRegistry.java delete mode 100644 vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/notification/NotificationRegistryProvider.java delete mode 100644 vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/test/ConnectionTest.java delete mode 100644 vpp-api/java/jvpp-registry/jvpp_registry.c delete mode 100755 vpp-api/java/jvpp/gen/jvpp_gen.py delete mode 100644 vpp-api/java/jvpp/gen/jvppgen/__init__.py delete mode 100644 vpp-api/java/jvpp/gen/jvppgen/callback_gen.py delete mode 100644 vpp-api/java/jvpp/gen/jvppgen/dto_gen.py delete mode 100644 vpp-api/java/jvpp/gen/jvppgen/jni_gen.py delete mode 100644 vpp-api/java/jvpp/gen/jvppgen/jvpp_c_gen.py delete mode 100644 vpp-api/java/jvpp/gen/jvppgen/jvpp_callback_facade_gen.py delete mode 100644 vpp-api/java/jvpp/gen/jvppgen/jvpp_future_facade_gen.py delete mode 100644 vpp-api/java/jvpp/gen/jvppgen/jvpp_impl_gen.py delete mode 100644 vpp-api/java/jvpp/gen/jvppgen/notification_gen.py delete mode 100644 vpp-api/java/jvpp/gen/jvppgen/types_gen.py delete mode 100644 vpp-api/java/jvpp/gen/jvppgen/util.py delete mode 100644 vpp-api/java/m4/ax_check_java_home.m4 delete mode 100644 vpp-api/java/m4/ax_check_java_plugin.m4 delete mode 100644 vpp-api/java/m4/ax_java_check_class.m4 delete mode 100644 vpp-api/java/m4/ax_java_options.m4 delete mode 100644 vpp-api/java/m4/ax_libgcj_jar.m4 delete mode 100644 vpp-api/java/m4/ax_prog_jar.m4 delete mode 100644 vpp-api/java/m4/ax_prog_java.m4 delete mode 100644 vpp-api/java/m4/ax_prog_java_cc.m4 delete mode 100644 vpp-api/java/m4/ax_prog_java_works.m4 delete mode 100644 vpp-api/java/m4/ax_prog_javac.m4 delete mode 100644 vpp-api/java/m4/ax_prog_javac_works.m4 delete mode 100644 vpp-api/java/m4/ax_prog_javadoc.m4 delete mode 100644 vpp-api/java/m4/ax_prog_javah.m4 delete mode 100644 vpp-api/java/m4/ax_try_compile_java.m4 delete mode 100644 vpp-api/java/m4/ax_try_run_java.m4 delete mode 100644 vpp-api/lua/README.md delete mode 100644 vpp-api/lua/bench.lua delete mode 100644 vpp-api/lua/examples/cli/README.md delete mode 100644 vpp-api/lua/examples/cli/lua-cli.lua delete mode 100644 vpp-api/lua/examples/example-acl-plugin.lua delete mode 100644 vpp-api/lua/examples/example-classifier.lua delete mode 100644 vpp-api/lua/examples/example-cli.lua delete mode 100644 vpp-api/lua/examples/lute/README.md delete mode 100644 vpp-api/lua/examples/lute/lute.lua delete mode 100644 vpp-api/lua/examples/lute/script-inout-acl-noacl.lute delete mode 100644 vpp-api/lua/examples/lute/script-inout-acl-old.lute delete mode 100644 vpp-api/lua/examples/lute/script-inout-acl.lute delete mode 100644 vpp-api/lua/examples/lute/script.lute delete mode 100644 vpp-api/lua/examples/lute/sessions-acl.lute delete mode 100644 vpp-api/lua/vpp-lapi.lua (limited to 'src/Makefile.am') diff --git a/.gitignore b/.gitignore index c12eedfb..61c407ff 100644 --- a/.gitignore +++ b/.gitignore @@ -80,11 +80,11 @@ GTAGS *.pyc # Python api generator -/vpp-api/python/build -/vpp-api/python/dist -/vpp-api/python/vpp_papi.egg-info -/vpp-api/python/vpp_papi/memclnt.py -/vpp-api/python/vpp_papi/vpe.py +/src/vpp-api/python/build +/src/vpp-api/python/dist +/src/vpp-api/python/vpp_papi.egg-info +/src/vpp-api/python/vpp_papi/memclnt.py +/src/vpp-api/python/vpp_papi/vpe.py # Build files in the test directory /test/*.ok diff --git a/Makefile b/Makefile index eeb9837a..9187404c 100644 --- a/Makefile +++ b/Makefile @@ -57,7 +57,7 @@ endif .PHONY: help bootstrap wipe wipe-release build build-release rebuild rebuild-release .PHONY: run run-release debug debug-release build-vat run-vat pkg-deb pkg-rpm -.PHONY: ctags cscope plugins plugins-release build-vpp-api +.PHONY: ctags cscope .PHONY: test test-debug retest retest-debug test-doc test-wipe-doc test-help test-wipe .PHONY: test-cov test-wipe-cov @@ -69,8 +69,6 @@ help: @echo " wipe-release - wipe all products of release build " @echo " build - build debug binaries" @echo " build-release - build release binaries" - @echo " plugins - build debug plugin binaries" - @echo " plugins-release - build release plugin binaries" @echo " rebuild - wipe and build debug binares" @echo " rebuild-release - wipe and build release binares" @echo " run - run debug binary" @@ -83,8 +81,6 @@ help: @echo " retest - run functional tests" @echo " retest-debug - run functional tests (debug build)" @echo " test-help - show help on test framework" - @echo " build-vat - build vpp-api-test tool" - @echo " build-vpp-api - build vpp-api" @echo " run-vat - run vpp-api-test tool" @echo " pkg-deb - build DEB packages" @echo " pkg-rpm - build RPM packages" @@ -206,19 +202,10 @@ wipe-release: $(BR)/.bootstrap.ok rebuild-release: wipe-release build-release -plugins: $(BR)/.bootstrap.ok - $(call make,$(PLATFORM)_debug,plugins-install) - -plugins-release: $(BR)/.bootstrap.ok - $(call make,$(PLATFORM),plugins-install) - -build-vpp-api: $(BR)/.bootstrap.ok - $(call make,$(PLATFORM)_debug,vpp-api-install) - VPP_PYTHON_PREFIX=$(BR)/python define test - $(if $(filter-out $(3),retest),make -C $(BR) PLATFORM=$(1) TAG=$(2) vpp-api-install plugins-install vpp-install,) + $(if $(filter-out $(3),retest),make -C $(BR) PLATFORM=$(1) TAG=$(2) vpp-install,) make -C test \ BR=$(BR) \ VPP_TEST_BUILD_DIR=$(BR)/build-$(2)-native \ @@ -265,12 +252,12 @@ define run @echo "WARNING: STARTUP_CONF not defined or file doesn't exist." @echo " Running with minimal startup config: $(MINIMAL_STARTUP_CONF)\n" @cd $(STARTUP_DIR) && \ - sudo $(2) $(1)/vpp/bin/vpp $(MINIMAL_STARTUP_CONF) plugin_path $(1)/plugins/lib64/vpp_plugins + sudo $(2) $(1)/vpp/bin/vpp $(MINIMAL_STARTUP_CONF) plugin_path $(1)/vpp/lib64/vpp_plugins endef else define run @cd $(STARTUP_DIR) && \ - sudo $(2) $(1)/vpp/bin/vpp $(shell cat $(STARTUP_CONF) | sed -e 's/#.*//') plugin_path $(1)/plugins/lib64/vpp_plugins + sudo $(2) $(1)/vpp/bin/vpp $(shell cat $(STARTUP_CONF) | sed -e 's/#.*//') plugin_path $(1)/vpp/lib64/vpp_plugins endef endif diff --git a/build-data/packages/plugins.mk b/build-data/packages/plugins.mk deleted file mode 100644 index b4d67a28..00000000 --- a/build-data/packages/plugins.mk +++ /dev/null @@ -1,34 +0,0 @@ -plugins_configure_depend = \ - vpp-api-install \ - vpp-install - -plugins_CPPFLAGS = $(call installed_includes_fn, \ - vpp \ - vpp-api) - -plugins_LDFLAGS = $(call installed_libs_fn, \ - vpp) - -ifeq ($($(PLATFORM)_enable_tests),yes) -plugins_configure_args += --enable-tests -endif - -# Platform dependent configure flags -plugins_configure_args += $(plugins_configure_args_$(PLATFORM)) - -# include & link with openssl only if needed -ifneq ($($(PLATFORM)_uses_openssl),no) -plugins_CPPFLAGS += $(call installed_includes_fn, openssl) -plugins_LDFLAGS += $(call installed_libs_fn, openssl) -endif - -ifneq ($($(PLATFORM)_uses_dpdk),no) -ifeq ($($(PLATFORM)_uses_external_dpdk),yes) -plugins_CPPFLAGS += -I$($(PLATFORM)_dpdk_inc_dir) -plugins_LDFLAGS += -L$($(PLATFORM)_dpdk_lib_dir) -else -plugins_configure_depend += dpdk-install -plugins_CPPFLAGS += $(call installed_includes_fn, dpdk) -plugins_LDFLAGS += $(call installed_libs_fn, dpdk) -endif -endif diff --git a/build-data/packages/vnet.mk b/build-data/packages/vnet.mk deleted file mode 100644 index 57c444e8..00000000 --- a/build-data/packages/vnet.mk +++ /dev/null @@ -1,47 +0,0 @@ -vnet_configure_depend = \ - vppinfra-install \ - svm-install \ - vlib-api-install \ - vlib-install - -vnet_CPPFLAGS = $(call installed_includes_fn, \ - vppinfra \ - svm \ - vlib \ - vlib-api) - -vnet_LDFLAGS = $(call installed_libs_fn, \ - vppinfra \ - svm \ - vlib \ - vlib-api) - -ifeq ($($(PLATFORM)_enable_tests),yes) -vnet_configure_args += --enable-tests -endif - -# Platform dependent configure flags -vnet_configure_args += $(vnet_configure_args_$(PLATFORM)) - -# include & link with openssl only if needed -ifneq ($($(PLATFORM)_uses_openssl),no) -vnet_CPPFLAGS += $(call installed_includes_fn, openssl) -vnet_LDFLAGS += $(call installed_libs_fn, openssl) -endif - -ifneq ($($(PLATFORM)_uses_dpdk),no) -ifeq ($($(PLATFORM)_uses_external_dpdk),yes) -vnet_CPPFLAGS += -I$($(PLATFORM)_dpdk_inc_dir) -vnet_LDFLAGS += -L$($(PLATFORM)_dpdk_lib_dir) -else -vnet_configure_depend += dpdk-install -vnet_CPPFLAGS += $(call installed_includes_fn, dpdk) -vnet_LDFLAGS += $(call installed_libs_fn, dpdk) -endif -ifeq ($($(PLATFORM)_uses_dpdk_cryptodev),yes) -vnet_configure_args += --with-dpdk-crypto -endif -ifeq ($($(PLATFORM)_uses_dpdk_mlx5_pmd),yes) -vnet_configure_args += --with-dpdk-mlx5-pmd -endif -endif diff --git a/build-data/packages/vpp-api.mk b/build-data/packages/vpp-api.mk deleted file mode 100644 index d9e8d72e..00000000 --- a/build-data/packages/vpp-api.mk +++ /dev/null @@ -1,9 +0,0 @@ -vpp-api_configure_depend = \ - vpp-install - -vpp-api_CPPFLAGS = $(call installed_includes_fn, \ - vpp) - -vpp-api_LDFLAGS = - -vpp-api_CPPFLAGS += -I/usr/lib/jvm/java-8-openjdk-amd64/include diff --git a/build-data/platforms.mk b/build-data/platforms.mk index 88dd3ed8..6f21b6dc 100644 --- a/build-data/platforms.mk +++ b/build-data/platforms.mk @@ -25,7 +25,7 @@ install-deb: $(patsubst %,%-find-source,$(ROOT_PACKAGES)) \ : generate file manifests ; \ find $(INSTALL_PREFIX)$(ARCH)/*/bin -type f -print \ - | sed -e 's:.*:../& /usr/bin:' \ + | sed -e 's:.*:../& /usr/bin:' | grep -v vppapigen \ > deb/debian/vpp.install ; \ \ : core api definitions ; \ @@ -51,9 +51,17 @@ install-deb: $(patsubst %,%-find-source,$(ROOT_PACKAGES)) ./scripts/find-plugins-contents $(INSTALL_PREFIX)$(ARCH) \ deb/debian/vpp-plugins.install ; \ \ - : python-api package ; \ - ./scripts/find-python-api-contents $(INSTALL_PREFIX)$(ARCH) \ - deb/debian/vpp-python-api.install ; \ + : vpp-api-lua package ; \ + ./scripts/find-vpp-api-lua-contents $(INSTALL_PREFIX)$(ARCH) \ + deb/debian/vpp-api-lua.install ; \ + \ + : vpp-api-java package ; \ + ./scripts/find-vpp-api-java-contents $(INSTALL_PREFIX)$(ARCH) \ + deb/debian/vpp-api-java.install ; \ + \ + : vpp-api-python package ; \ + ./scripts/find-vpp-api-python-contents $(INSTALL_PREFIX)$(ARCH) \ + deb/debian/vpp-api-python.install ; \ \ : dpdk headers ; \ ./scripts/find-dpdk-contents $(INSTALL_PREFIX)$(ARCH) \ @@ -70,9 +78,9 @@ install-deb: $(patsubst %,%-find-source,$(ROOT_PACKAGES)) : dev package needs a couple of additions ; \ echo ../build-tool-native/tools/vppapigen /usr/bin \ >> deb/debian/vpp-dev.install ; \ - echo ../../vpp-api/java/jvpp/gen/jvpp_gen.py /usr/bin \ + echo ../../src/vpp-api/java/jvpp/gen/jvpp_gen.py /usr/bin \ >> deb/debian/vpp-dev.install ; \ - for i in $$(ls ../vpp-api/java/jvpp/gen/jvppgen/*.py); do \ + for i in $$(ls ../src/vpp-api/java/jvpp/gen/jvppgen/*.py); do \ echo ../$${i} /usr/lib/python2.7/dist-packages/jvppgen \ >> deb/debian/vpp-dev.install; \ done; \ diff --git a/build-data/platforms/vpp.mk b/build-data/platforms/vpp.mk index 513a4db4..c9a214ac 100644 --- a/build-data/platforms/vpp.mk +++ b/build-data/platforms/vpp.mk @@ -29,15 +29,13 @@ vpp_uses_dpdk = yes # Uncoment to enable building unit tests # vpp_enable_tests = yes -vpp_root_packages = vpp vpp-api gmod plugins +vpp_root_packages = vpp gmod vpp_configure_args_vpp = --with-dpdk # Set these parameters carefully. The vlib_buffer_t is 128 bytes, i.e. vlib_configure_args_vpp = --with-pre-data=128 -plugins_configure_args_vpp = --with-dpdk - # DPDK configuration parameters # vpp_uses_dpdk_cryptodev = yes # vpp_uses_dpdk_mlx5_pmd = yes diff --git a/build-data/platforms/vpp_lite.mk b/build-data/platforms/vpp_lite.mk index 55805d10..a556b487 100644 --- a/build-data/platforms/vpp_lite.mk +++ b/build-data/platforms/vpp_lite.mk @@ -27,7 +27,7 @@ vpp_lite_uses_dpdk = no # Uncoment to enable building unit tests #vpp_lite_enable_tests = yes -vpp_lite_root_packages = vpp vpp-api gmod +vpp_lite_root_packages = vpp gmod vlib_configure_args_vpp_lite = --with-pre-data=128 diff --git a/build-root/deb/debian/.gitignore b/build-root/deb/debian/.gitignore index 75d8fbbc..c5e915a8 100644 --- a/build-root/deb/debian/.gitignore +++ b/build-root/deb/debian/.gitignore @@ -11,4 +11,6 @@ vpp-dpdk-dev/ vpp-dpdk-dkms/ vpp-dbg/ vppctl/ -vpp-python-api/ +vpp-api-lua/ +vpp-api-java/ +vpp-api-python/ diff --git a/build-root/deb/debian/control b/build-root/deb/debian/control index 643774e3..de48c903 100644 --- a/build-root/deb/debian/control +++ b/build-root/deb/debian/control @@ -60,7 +60,21 @@ Depends: ${shlibs:Depends}, ${misc:Depends} Description: DPDK 2.1 igb_uio_driver This package contains Linux kernel modules distributed with DPDK. -Package: vpp-python-api +Package: vpp-api-lua +Architecture: any +Depends: ${python:Depends}, ${misc:Depends}, vpp (= ${source:Version}) +Description: VPP LUA API bindings + This package contains VPP lua api bindings + . + +Package: vpp-api-java +Architecture: any +Depends: ${python:Depends}, ${misc:Depends}, vpp (= ${source:Version}) +Description: VPP Java API bindings + This package contains VPP java api bindings + . + +Package: vpp-api-python Architecture: any Depends: ${python:Depends}, ${misc:Depends}, vpp (= ${source:Version}) Description: VPP Python API bindings diff --git a/build-root/packages/tools.mk b/build-root/packages/tools.mk index 506e024b..d64a7959 100644 --- a/build-root/packages/tools.mk +++ b/build-root/packages/tools.mk @@ -1,3 +1,3 @@ tools_source = src -tools_configure_args = --disable-vlib +tools_configure_args = --disable-vlib --disable-svm --disable-japi diff --git a/build-root/rpm/vpp.spec b/build-root/rpm/vpp.spec index 654424c0..149ac51c 100644 --- a/build-root/rpm/vpp.spec +++ b/build-root/rpm/vpp.spec @@ -71,12 +71,28 @@ Requires: vpp = %{_version}-%{_release} %description plugins This package contains VPP plugins -%package python-api +%package api-lua +Summary: VPP api lua bindings +Group: Development/Libraries +Requires: vpp = %{_version}-%{_release}, vpp-lib = %{_version}-%{_release} + +%description api-lua +This package contains the lua bindings for the vpp api + +%package api-java +Summary: VPP api java bindings +Group: Development/Libraries +Requires: vpp = %{_version}-%{_release}, vpp-lib = %{_version}-%{_release} + +%description api-java +This package contains the java bindings for the vpp api + +%package api-python Summary: VPP api python bindings Group: Development/Libraries Requires: vpp = %{_version}-%{_release}, vpp-lib = %{_version}-%{_release}, python-setuptools -%description python-api +%description api-python This package contains the python bindings for the vpp api %prep @@ -131,6 +147,22 @@ do install -p -m 644 $file %{buildroot}/usr/share/vpp/api done +# Lua bindings +mkdir -p -m755 %{buildroot}/usr/share/doc/vpp/examples/lua/examples/cli +mkdir -p -m755 %{buildroot}/usr/share/doc/vpp/examples/lua/examples/lute +for file in $(cd %{_mu_build_dir}/%{_vpp_install_dir}/../../src/vpp-api/lua && git ls-files .) +do + install -p -m 644 %{_mu_build_dir}/%{_vpp_install_dir}/../../src/vpp-api/lua/$file \ + %{buildroot}/usr/share/doc/vpp/examples/lua/$file +done + +# Java bindings +mkdir -p -m755 %{buildroot}/usr/share/java +for file in $(find %{_mu_build_dir}/%{_vpp_install_dir}/vpp/share/java -type f -name '*.jar' -print ) +do + install -p -m 644 $file %{buildroot}/usr/share/java +done + # Python bindings mkdir -p -m755 %{buildroot}%{python2_sitelib} install -p -m 666 %{_mu_build_dir}/%{_vpp_install_dir}/*/lib/python2.7/site-packages/vpp_papi-*.egg %{buildroot}%{python2_sitelib} @@ -151,16 +183,16 @@ do done mkdir -p -m755 %{buildroot}%{python2_sitelib}/jvppgen -install -p -m755 %{_mu_build_dir}/../vpp-api/java/jvpp/gen/jvpp_gen.py %{buildroot}/usr/bin -for i in $(ls %{_mu_build_dir}/../vpp-api/java/jvpp/gen/jvppgen/*.py); do +install -p -m755 %{_mu_build_dir}/../src/vpp-api/java/jvpp/gen/jvpp_gen.py %{buildroot}/usr/bin +for i in $(ls %{_mu_build_dir}/../src/vpp-api/java/jvpp/gen/jvppgen/*.py); do install -p -m666 ${i} %{buildroot}%{python2_sitelib}/jvppgen done; # sample plugin mkdir -p -m755 %{buildroot}/usr/share/doc/vpp/examples/sample-plugin/sample -for file in $(cd %{_mu_build_dir}/%{_vpp_install_dir}/../../plugins/sample-plugin && git ls-files .) +for file in $(cd %{_mu_build_dir}/%{_vpp_install_dir}/../../src/examples/sample-plugin && git ls-files .) do - install -p -m 644 %{_mu_build_dir}/%{_vpp_install_dir}/../../plugins/sample-plugin/$file \ + install -p -m 644 %{_mu_build_dir}/%{_vpp_install_dir}/../../src/examples/sample-plugin/$file \ %{buildroot}/usr/share/doc/vpp/examples/sample-plugin/$file done @@ -170,24 +202,12 @@ done # mkdir -p -m755 %{buildroot}/usr/lib/vpp_plugins mkdir -p -m755 %{buildroot}/usr/lib/vpp_api_test_plugins -for file in $(cd %{_mu_build_dir}/%{_vpp_install_dir}/plugins/lib64/vpp_plugins && find -type f -print) -do - install -p -m 644 %{_mu_build_dir}/%{_vpp_install_dir}/plugins/lib64/vpp_plugins/$file \ - %{buildroot}/usr/lib/vpp_plugins/$file -done - for file in $(cd %{_mu_build_dir}/%{_vpp_install_dir}/vpp/lib64/vpp_plugins && find -type f -print) do install -p -m 644 %{_mu_build_dir}/%{_vpp_install_dir}/vpp/lib64/vpp_plugins/$file \ %{buildroot}/usr/lib/vpp_plugins/$file done -for file in $(cd %{_mu_build_dir}/%{_vpp_install_dir}/plugins/lib64/vpp_api_test_plugins && find -type f -print) -do - install -p -m 644 %{_mu_build_dir}/%{_vpp_install_dir}/plugins/lib64/vpp_api_test_plugins/$file \ - %{buildroot}/usr/lib/vpp_api_test_plugins/$file -done - for file in $(cd %{_mu_build_dir}/%{_vpp_install_dir}/vpp/lib64/vpp_api_test_plugins && find -type f -print) do install -p -m 644 %{_mu_build_dir}/%{_vpp_install_dir}/vpp/lib64/vpp_api_test_plugins/$file \ @@ -208,13 +228,13 @@ done sysctl --system %systemd_post vpp.service -%post python-api +%post api-python easy_install -z %{python2_sitelib}/vpp_papi-*.egg %preun %systemd_preun vpp.service -%preun python-api +%preun api-python easy_install -mxNq vpp_papi %postun @@ -255,7 +275,15 @@ fi %{_libdir}/* /usr/share/vpp/api/* -%files python-api +%files api-lua +%defattr(644,root,root) +/usr/share/doc/vpp/examples/lua + +%files api-java +%defattr(644,root,root) +/usr/share/java/* + +%files api-python %defattr(644,root,root) %{python2_sitelib}/vpp_papi-*.egg diff --git a/build-root/scripts/find-api-core-contents b/build-root/scripts/find-api-core-contents index f1f96f1f..c1af69db 100755 --- a/build-root/scripts/find-api-core-contents +++ b/build-root/scripts/find-api-core-contents @@ -1,6 +1,6 @@ #!/bin/bash -for i in $(find ${1}/vpp -name *.api.json -type f -print); do +for i in $(find ${1}/vpp/share/vpp/api/core -name *.api.json -type f -print); do echo ../${i} /usr/share/vpp/api/ >> ${2} done for i in $(find ${1}/vlib-api -name *.api.json -type f -print); do diff --git a/build-root/scripts/find-dev-contents b/build-root/scripts/find-dev-contents index 2dc6cc4d..d4f7b63f 100755 --- a/build-root/scripts/find-dev-contents +++ b/build-root/scripts/find-dev-contents @@ -16,16 +16,16 @@ do done # sample plugin -paths=`(cd ..; find plugins/sample-plugin -type f -print | grep -v autom4te)` +paths=`(cd ..; find src/examples/sample-plugin -type f -print | grep -v autom4te)` for path in $paths do - relpath=`echo $path | sed -e 's:.*plugins/::'` + relpath=`echo $path | sed -e 's:.*src/examples/::'` dir=`dirname $relpath` if [ $dir = "sample-plugin" ] ; then - echo ../../$path /usr/share/doc/vpp/examples/plugins/sample-plugin >> $2 + echo ../../$path /usr/share/doc/vpp/examples/sample-plugin/ >> $2 else echo ../../$path \ - /usr/share/doc/vpp/examples/plugins/$dir >> $2 + /usr/share/doc/vpp/examples/$dir >> $2 fi done diff --git a/build-root/scripts/find-plugins-contents b/build-root/scripts/find-plugins-contents index 4108f790..ae2a4271 100755 --- a/build-root/scripts/find-plugins-contents +++ b/build-root/scripts/find-plugins-contents @@ -2,11 +2,11 @@ rm -f $2 -for i in ${1}/{plugins,vpp}/lib64/vpp_plugins/*.so; do +for i in ${1}/vpp/lib64/vpp_plugins/*.so; do echo ../${i} /usr/lib/vpp_plugins >> ${2} done -for i in ${1}/{plugins,vpp}/lib64/vpp_api_test_plugins/*.so; do +for i in ${1}/vpp/lib64/vpp_api_test_plugins/*.so; do echo ../${i} /usr/lib/vpp_api_test_plugins >> ${2} done diff --git a/build-root/scripts/find-python-api-contents b/build-root/scripts/find-python-api-contents deleted file mode 100755 index 24e8532c..00000000 --- a/build-root/scripts/find-python-api-contents +++ /dev/null @@ -1,8 +0,0 @@ -#!/bin/bash - -rm -f $2 - -for i in $(find ${1}/{vpp,vpp-api}/lib/python2.7/site-packages/ -type f -print); do - echo ../${i} /usr/lib/python2.7/site-packages/vpp_papi >> ${2} -done - diff --git a/build-root/scripts/find-vpp-api-java-contents b/build-root/scripts/find-vpp-api-java-contents new file mode 100755 index 00000000..5f1bf197 --- /dev/null +++ b/build-root/scripts/find-vpp-api-java-contents @@ -0,0 +1,8 @@ +#!/bin/bash + +rm -f $2 + +for i in $(find ${1}/vpp/share/java/ -type f -print); do + echo ../${i} /usr/share/java >> ${2} +done + diff --git a/build-root/scripts/find-vpp-api-lua-contents b/build-root/scripts/find-vpp-api-lua-contents new file mode 100755 index 00000000..f1173db8 --- /dev/null +++ b/build-root/scripts/find-vpp-api-lua-contents @@ -0,0 +1,6 @@ +#!/bin/bash + +#i for now put everything into examples directory + +echo ../../src/vpp-api/lua /usr/share/vpp/examples > ${2} + diff --git a/build-root/scripts/find-vpp-api-python-contents b/build-root/scripts/find-vpp-api-python-contents new file mode 100755 index 00000000..819c9122 --- /dev/null +++ b/build-root/scripts/find-vpp-api-python-contents @@ -0,0 +1,8 @@ +#!/bin/bash + +rm -f $2 + +for i in $(find ${1}/vpp/lib/python2.7/site-packages/ -type f -print); do + echo ../${i} /usr/lib/python2.7/site-packages/vpp_papi >> ${2} +done + diff --git a/doxygen/Makefile b/doxygen/Makefile index ffce3c13..face5b44 100644 --- a/doxygen/Makefile +++ b/doxygen/Makefile @@ -52,8 +52,7 @@ DOXY_SRC_DIRECTORIES = \ $(DOXY_SRC)/vlibsocket \ $(DOXY_SRC)/vnet \ $(DOXY_SRC)/vpp \ - $(DOXY_SRC)/vpp-api \ - vpp-api + $(DOXY_SRC)/vpp-api # Input directories and files DOXY_INPUT ?= \ @@ -74,8 +73,8 @@ DOXY_INPUT := $(subst $(WS_ROOT)/,,$(DOXY_INPUT)) DOXY_EXCLUDE ?= \ $(DOXY_SRC)/vlib/vlib/buffer.c \ $(DOXY_SRC)/vlib/example \ - plugins/sample-plugin \ - vpp-api/lua + $(DOXY_SRC)/vpp-api/lua \ + plugins/sample-plugin # Generate a regexp for filenames to exclude DOXY_EXCLUDE_REGEXP = ($(subst .,\.,$(shell echo '$(strip $(DOXY_EXCLUDE))' | sed -e 's/ /|/g'))) diff --git a/plugins/Makefile.am b/plugins/Makefile.am index a101e47f..20676fc8 100644 --- a/plugins/Makefile.am +++ b/plugins/Makefile.am @@ -28,22 +28,7 @@ if ENABLE_sample_PLUGIN SUBDIRS += sample-plugin endif -if ENABLE_ioam_PLUGIN -SUBDIRS += ioam-plugin -endif - if ENABLE_vcgn_PLUGIN SUBDIRS += vcgn-plugin endif -if ENABLE_snat_PLUGIN -SUBDIRS += snat-plugin -endif - -if ENABLE_lb_PLUGIN -SUBDIRS += lb-plugin -endif - -if ENABLE_acl_PLUGIN -SUBDIRS += acl-plugin -endif diff --git a/plugins/acl-plugin/Makefile.am b/plugins/acl-plugin/Makefile.am deleted file mode 100644 index 798a32bb..00000000 --- a/plugins/acl-plugin/Makefile.am +++ /dev/null @@ -1,114 +0,0 @@ -# Copyright (c) 2016 Cisco and/or its affiliates. -# Licensed under the Apache License, Version 2.0 (the "License"); -# you may not use this file except in compliance with the License. -# You may obtain a copy of the License at: -# -# http://www.apache.org/licenses/LICENSE-2.0 -# -# Unless required by applicable law or agreed to in writing, software -# distributed under the License is distributed on an "AS IS" BASIS, -# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. -# See the License for the specific language governing permissions and -# limitations under the License. - -AUTOMAKE_OPTIONS = foreign subdir-objects - -AM_CFLAGS = -Wall -AM_LDFLAGS = -module -shared -avoid-version - -vppapitestpluginsdir = ${libdir}/vpp_api_test_plugins -vpppluginsdir = ${libdir}/vpp_plugins - -vppapitestplugins_LTLIBRARIES = acl_test_plugin.la -vppplugins_LTLIBRARIES = acl_plugin.la - -acl_plugin_la_SOURCES = \ - acl/acl.c \ - acl/node_in.c \ - acl/node_out.c \ - acl/l2sess.c \ - acl/l2sess_node.c \ - acl/l2sess.h \ - acl/acl_plugin.api.h - -BUILT_SOURCES = acl/acl.api.h acl/acl.api.json - -SUFFIXES = .api.h .api .api.json - -# -# ACL API -# -%.api.h: %.api - mkdir -p `dirname $@` ; \ - $(CC) $(CPPFLAGS) -E -P -C -x c $^ \ - | vppapigen --input - --output $@ --show-name $@ - -%.api.json: %.api - @echo " JSON APIGEN " $@ ; \ - mkdir -p `dirname $@` ; \ - $(CC) $(CPPFLAGS) -E -P -C -x c $^ \ - | vppapigen --input - --json $@ - -apidir = $(prefix)/acl -api_DATA = acl/acl.api.json - -noinst_HEADERS = \ - acl/acl_all_api_h.h \ - acl/acl_msg_enum.h \ - acl/acl.api.h - -acl_test_plugin_la_SOURCES = acl/acl_test.c acl/acl_plugin.api.h - -# Remove *.la files -install-data-hook: - @(cd $(vpppluginsdir) && $(RM) $(vppplugins_LTLIBRARIES)) - @(cd $(vppapitestpluginsdir) && $(RM) $(vppapitestplugins_LTLIBRARIES)) - -# -# Java code generation -# -jvpp_registry_root = ../../vpp-api/java -jvpp_registry_version = 17.04 -jacl_jarfile = jvpp-acl-$(PACKAGE_VERSION).jar -jvpp_package_dir = io/fd/vpp/jvpp/acl -jvpp_root = acl/jvpp -jvpp_target_dir = target -jvpp_target = $(jvpp_root)/$(jvpp_target_dir) - -lib_LTLIBRARIES = libjvpp_acl.la -libjvpp_acl_la_SOURCES = acl/acl.api.h acl/jvpp_acl.c acl/jvpp/io_fd_vpp_jvpp_acl_JVppAclImpl.h -libjvpp_acl_la_LIBADD = -lvlibmemoryclient -lvlibapi -lvppinfra \ - -lpthread -lm -lrt -L$(jvpp_registry_root)/.libs -ljvpp_common -libjvpp_acl_la_LDFLAGS = -module -libjvpp_acl_la_CPPFLAGS = -I$(JAVA_HOME)/include -I$(JAVA_HOME)/include/linux -I../ -I$(srcdir)/../ - -BUILT_SOURCES += $(jvpp_root)/io_fd_vpp_jvpp_acl_JVppAclImpl.h - -$(jvpp_root)/io_fd_vpp_jvpp_acl_JVppAclImpl.h: acl/acl.api.json - dir=`pwd`; \ - mkdir -p $(jvpp_target); \ - mkdir -p $(jvpp_root)/$(jvpp_package_dir); \ - cd $(jvpp_root)/$(jvpp_package_dir); \ - mkdir -p types dto future callfacade callback notification test; \ - @srcdir@/$(jvpp_registry_root)/jvpp/gen/jvpp_gen.py -i $${dir}/acl/acl.api.json --plugin_name acl; \ - cd -; \ - mv -f $(jvpp_root)/$(jvpp_package_dir)/jvpp_acl_gen.h $(jvpp_root)/jvpp_acl_gen.h; \ - cp $(srcdir)/$(jvpp_root)/$(jvpp_package_dir)/test/*.java $(jvpp_root)/$(jvpp_package_dir)/test/; \ - cd $(jvpp_root); \ - $(JAVAC) -classpath .:$(jvpp_target_dir):../../$(jvpp_registry_root)/jvpp-registry-$(jvpp_registry_version).jar -d $(jvpp_target_dir) $(jvpp_package_dir)/*.java \ - $(jvpp_package_dir)/types/*.java \ - $(jvpp_package_dir)/dto/*.java \ - $(jvpp_package_dir)/callback/*.java \ - $(jvpp_package_dir)/notification/*.java \ - $(jvpp_package_dir)/future/*.java \ - $(jvpp_package_dir)/callfacade/*.java \ - $(jvpp_package_dir)/test/*.java \ - || (echo "acl jvpp compilation failed: $$?"; exit 1); \ - $(JAVAH) -classpath .:$(jvpp_target_dir):../../$(jvpp_registry_root)/jvpp-registry-$(jvpp_registry_version).jar -d . io.fd.vpp.jvpp.acl.JVppAclImpl ; - -$(jacl_jarfile): libjvpp_acl.la - cp .libs/libjvpp_acl.so.0.0.0 $(jvpp_target); \ - cd $(jvpp_target); \ - $(JAR) cfv $(JARFLAGS) ../../../$@ libjvpp_acl.so.0.0.0 $(jvpp_package_dir)/* ; cd ..; - -all-local: $(jacl_jarfile) diff --git a/plugins/acl-plugin/acl/acl.api b/plugins/acl-plugin/acl/acl.api deleted file mode 100644 index 58a5a171..00000000 --- a/plugins/acl-plugin/acl/acl.api +++ /dev/null @@ -1,444 +0,0 @@ -/* Hey Emacs use -*- mode: C -*- */ -/* - * Copyright (c) 2016 Cisco and/or its affiliates. - * Licensed under the Apache License, Version 2.0 (the "License"); - * you may not use this file except in compliance with the License. - * You may obtain a copy of the License at: - * - * http://www.apache.org/licenses/LICENSE-2.0 - * - * Unless required by applicable law or agreed to in writing, software - * distributed under the License is distributed on an "AS IS" BASIS, - * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. - * See the License for the specific language governing permissions and - * limitations under the License. - */ - -/** \file - This file defines the vpp control-plane API messages - used to control the ACL plugin -*/ - - -/** \brief Get the plugin version - @param client_index - opaque cookie to identify the sender - @param context - sender context, to match reply w/ request -*/ - -define acl_plugin_get_version -{ - u32 client_index; - u32 context; -}; - -/** \brief Reply to get the plugin version - @param context - returned sender context, to match reply w/ request - @param major - Incremented every time a known breaking behavior change is introduced - @param minor - Incremented with small changes, may be used to avoid buggy versions -*/ - -define acl_plugin_get_version_reply -{ - u32 context; - u32 major; - u32 minor; -}; - -/** \brief Access List Rule entry - @param is_permit - deny (0), permit (1), or permit+reflect(2) action on this rule. - @param is_ipv6 - IP addresses in this rule are IPv6 (1) or IPv4 (0) - @param src_ip_addr - Source prefix value - @param src_ip_prefix_len - Source prefix length - @param dst_ip_addr - Destination prefix value - @param dst_ip_prefix_len - Destination prefix length - @param proto - L4 protocol (http://www.iana.org/assignments/protocol-numbers/protocol-numbers.xhtml) - @param srcport_or_icmptype_first - beginning of source port or ICMP4/6 type range - @param srcport_or_icmptype_last - end of source port or ICMP4/6 type range - @param dstport_or_icmpcode_first - beginning of destination port or ICMP4/6 code range - @param dstport_or_icmpcode_last - end of destination port or ICMP4/6 code range - @param tcp_flags_mask - if proto==6, match masked TCP flags with this value - @param tcp_flags_value - if proto==6, mask to AND the TCP flags in the packet with -*/ - -typeonly manual_print manual_endian define acl_rule -{ - u8 is_permit; - u8 is_ipv6; - u8 src_ip_addr[16]; - u8 src_ip_prefix_len; - u8 dst_ip_addr[16]; - u8 dst_ip_prefix_len; -/* - * L4 protocol. IANA number. 1 = ICMP, 58 = ICMPv6, 6 = TCP, 17 = UDP. - * 0 => ignore L4 and ignore the ports/tcpflags when matching. - */ - u8 proto; -/* - * If the L4 protocol is TCP or UDP, the below - * hold ranges of ports, else if the L4 is ICMP/ICMPv6 - * they hold ranges of ICMP(v6) types/codes. - * - * Ranges are inclusive, i.e. to match "any" TCP/UDP port, - * use first=0,last=65535. For ICMP(v6), - * use first=0,last=255. - */ - u16 srcport_or_icmptype_first; - u16 srcport_or_icmptype_last; - u16 dstport_or_icmpcode_first; - u16 dstport_or_icmpcode_last; -/* - * for proto = 6, this matches if the - * TCP flags in the packet, ANDed with tcp_flags_mask, - * is equal to tcp_flags_value. - */ - u8 tcp_flags_mask; - u8 tcp_flags_value; -}; - -/** \brief MACIP Access List Rule entry - @param is_permit - deny (0), permit (1) action on this rule. - @param is_ipv6 - IP addresses in this rule are IPv6 (1) or IPv4 (0) - @param src_mac - match masked source MAC address against this value - @param src_mac_mask - AND source MAC address with this value before matching - @param src_ip_addr - Source prefix value - @param src_ip_prefix_len - Source prefix length -*/ - -typeonly manual_print manual_endian define macip_acl_rule -{ - u8 is_permit; - u8 is_ipv6; -/* - * The source mac of the packet ANDed with src_mac_mask. - * The source ip[46] address in the packet is matched - * against src_ip_addr, with src_ip_prefix_len set to 0. - * - * For better performance, minimize the number of - * (src_mac_mask, src_ip_prefix_len) combinations - * in a MACIP ACL. - */ - u8 src_mac[6]; - u8 src_mac_mask[6]; - u8 src_ip_addr[16]; - u8 src_ip_prefix_len; -}; - -/** \brief Replace an existing ACL in-place or create a new ACL - @param client_index - opaque cookie to identify the sender - @param context - sender context, to match reply w/ request - @param acl_index - an existing ACL entry (0..0xfffffffe) to replace, or 0xffffffff to make new ACL - @param tag - a string value stored along with the ACL, for descriptive purposes - @param count - number of ACL rules - @r - Rules for this access-list -*/ - -manual_print manual_endian define acl_add_replace -{ - u32 client_index; - u32 context; - u32 acl_index; /* ~0 to add, existing ACL# to replace */ - u8 tag[64]; /* What gets in here gets out in the corresponding tag field when dumping the ACLs. */ - u32 count; - vl_api_acl_rule_t r[count]; -}; - -/** \brief Reply to add/replace ACL - @param context - returned sender context, to match reply w/ request - @param acl_index - index of the updated or newly created ACL - @param retval 0 - no error -*/ - -define acl_add_replace_reply -{ - u32 context; - u32 acl_index; - i32 retval; -}; - -/** \brief Delete an ACL - @param client_index - opaque cookie to identify the sender - @param context - sender context, to match reply w/ request - @param acl_index - ACL index to delete -*/ - -define acl_del -{ - u32 client_index; - u32 context; - u32 acl_index; -}; - -/** \brief Reply to delete the ACL - @param context - returned sender context, to match reply w/ request - @param retval 0 - no error -*/ - -define acl_del_reply -{ - u32 context; - i32 retval; -}; - -/* acl_interface_add_del(_reply) to be deprecated in lieu of acl_interface_set_acl_list */ -/** \brief Use acl_interface_set_acl_list instead - Append/remove an ACL index to/from the list of ACLs checked for an interface - @param client_index - opaque cookie to identify the sender - @param context - sender context, to match reply w/ request - @param is_add - add or delete the ACL index from the list - @param is_input - check the ACL on input (1) or output (0) - @param sw_if_index - the interface to alter the list of ACLs on - @param acl_index - index of ACL for the operation -*/ - -define acl_interface_add_del -{ - u32 client_index; - u32 context; - u8 is_add; -/* - * is_input = 0 => ACL applied on interface egress - * is_input = 1 => ACL applied on interface ingress - */ - u8 is_input; - u32 sw_if_index; - u32 acl_index; -}; - -/** \brief Reply to alter the ACL list - @param context - returned sender context, to match reply w/ request - @param retval 0 - no error -*/ - -define acl_interface_add_del_reply -{ - u32 context; - i32 retval; -}; - -/** \brief Set the vector of input/output ACLs checked for an interface - @param client_index - opaque cookie to identify the sender - @param context - sender context, to match reply w/ request - @param sw_if_index - the interface to alter the list of ACLs on - @param count - total number of ACL indices in the vector - @param n_input - this many first elements correspond to input ACLs, the rest - output - @param acls - vector of ACL indices -*/ - -manual_endian define acl_interface_set_acl_list -{ - u32 client_index; - u32 context; - u32 sw_if_index; - u8 count; - u8 n_input; /* First n_input ACLs are set as a list of input ACLs, the rest are applied as output */ - u32 acls[count]; -}; - -/** \brief Reply to set the ACL list on an interface - @param context - returned sender context, to match reply w/ request - @param retval 0 - no error -*/ - -define acl_interface_set_acl_list_reply -{ - u32 context; - i32 retval; -}; - -/** \brief Dump the specific ACL contents or all of the ACLs' contents - @param client_index - opaque cookie to identify the sender - @param context - sender context, to match reply w/ request - @param acl_index - ACL index to dump, ~0 to dump all ACLs -*/ - -define acl_dump -{ - u32 client_index; - u32 context; - u32 acl_index; /* ~0 for all ACLs */ -}; - -/** \brief Details about a single ACL contents - @param context - returned sender context, to match reply w/ request - @param acl_index - ACL index whose contents are being sent in this message - @param tag - Descriptive tag value which was supplied at ACL creation - @param count - Number of rules in this ACL - @param r - Array of rules within this ACL -*/ - -manual_print manual_endian define acl_details -{ - u32 context; - u32 acl_index; - u8 tag[64]; /* Same blob that was supplied to us when creating the ACL, one hopes. */ - u32 count; - vl_api_acl_rule_t r[count]; -}; - -/** \brief Dump the list(s) of ACL applied to specific or all interfaces - @param client_index - opaque cookie to identify the sender - @param context - sender context, to match reply w/ request - @param sw_if_index - interface to dump the ACL list for -*/ - -define acl_interface_list_dump -{ - u32 client_index; - u32 context; - u32 sw_if_index; /* ~0 for all interfaces */ -}; - -/** \brief Details about a single ACL contents - @param context - returned sender context, to match reply w/ request - @param sw_if_index - interface for which the list of ACLs is applied - @param count - total length of acl indices vector - @param n_input - this many of indices in the beginning are input ACLs, the rest - output - @param acls - the vector of ACL indices -*/ - -manual_endian define acl_interface_list_details -{ - u32 context; - u32 sw_if_index; - u8 count; - u8 n_input; - u32 acls[count]; -}; - -/** \brief Add a MACIP ACL - @param client_index - opaque cookie to identify the sender - @param context - sender context, to match reply w/ request - @param tag - descriptive value for this MACIP ACL - @param count - number of rules in this ACL - @param r - vector of MACIP ACL rules -*/ - -manual_print manual_endian define macip_acl_add -{ - u32 client_index; - u32 context; - u8 tag[64]; - u32 count; - vl_api_macip_acl_rule_t r[count]; -}; - -/** \brief Reply to add MACIP ACL - @param context - returned sender context, to match reply w/ request - @param acl_index - index of the newly created ACL - @param retval 0 - no error -*/ - -define macip_acl_add_reply -{ - u32 context; - u32 acl_index; - i32 retval; -}; - -/** \brief Delete a MACIP ACL - @param client_index - opaque cookie to identify the sender - @param context - sender context, to match reply w/ request - @param acl_index - MACIP ACL index to delete -*/ - -define macip_acl_del -{ - u32 client_index; - u32 context; - u32 acl_index; -}; - -/** \brief Reply to delete the MACIP ACL - @param context - returned sender context, to match reply w/ request - @param retval 0 - no error -*/ - -define macip_acl_del_reply -{ - u32 context; - i32 retval; -}; - -/** \brief Add or delete a MACIP ACL to/from interface - @param client_index - opaque cookie to identify the sender - @param context - sender context, to match reply w/ request - @param is_add - add (1) or delete (0) ACL from being used on an interface - @param sw_if_index - interface to apply the action to - @param acl_index - MACIP ACL index -*/ - -define macip_acl_interface_add_del -{ - u32 client_index; - u32 context; - u8 is_add; - /* macip ACLs are always input */ - u32 sw_if_index; - u32 acl_index; -}; - -/** \brief Reply to apply/unapply the MACIP ACL - @param context - returned sender context, to match reply w/ request - @param retval 0 - no error -*/ - -define macip_acl_interface_add_del_reply -{ - u32 context; - i32 retval; -}; - -/** \brief Dump one or all defined MACIP ACLs - @param client_index - opaque cookie to identify the sender - @param context - sender context, to match reply w/ request - @param acl_index - MACIP ACL index or ~0 to dump all ACLs -*/ - -define macip_acl_dump -{ - u32 client_index; - u32 context; - u32 acl_index; /* ~0 for all ACLs */ -}; - -/** \brief Details about one MACIP ACL - @param context - returned sender context, to match reply w/ request - @param acl_index - index of this MACIP ACL - @param tag - descriptive tag which was supplied during the creation - @param count - length of the vector of MACIP ACL rules - @param r - rules comprising this ACL -*/ - -manual_print manual_endian define macip_acl_details -{ - u32 context; - u32 acl_index; - u8 tag[64]; - u32 count; - vl_api_macip_acl_rule_t r[count]; -}; - -/** \brief Get the vector of MACIP ACL IDs applied to the interfaces - @param client_index - opaque cookie to identify the sender - @param context - sender context, to match reply w/ request -*/ - -define macip_acl_interface_get -{ - u32 client_index; - u32 context; -}; - -/** \brief Reply with the vector of MACIP ACLs by sw_if_index - @param context - returned sender context, to match reply w/ request - @param count - total number of elements in the vector - @param acls - the vector of active MACACL indices per sw_if_index -*/ - -define macip_acl_interface_get_reply -{ - u32 context; - u32 count; - u32 acls[count]; -}; - diff --git a/plugins/acl-plugin/acl/acl.c b/plugins/acl-plugin/acl/acl.c deleted file mode 100644 index 8ff5a6b7..00000000 --- a/plugins/acl-plugin/acl/acl.c +++ /dev/null @@ -1,1901 +0,0 @@ -/* - * Copyright (c) 2016 Cisco and/or its affiliates. - * Licensed under the Apache License, Version 2.0 (the "License"); - * you may not use this file except in compliance with the License. - * You may obtain a copy of the License at: - * - * http://www.apache.org/licenses/LICENSE-2.0 - * - * Unless required by applicable law or agreed to in writing, software - * distributed under the License is distributed on an "AS IS" BASIS, - * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. - * See the License for the specific language governing permissions and - * limitations under the License. - */ - -#include - -#include -#include -#include -#include - -#include -#include - -#include -#include -#include - -/* define message IDs */ -#include - -/* define message structures */ -#define vl_typedefs -#include -#undef vl_typedefs - -/* define generated endian-swappers */ -#define vl_endianfun -#include -#undef vl_endianfun - -/* instantiate all the print functions we know about */ -#define vl_print(handle, ...) vlib_cli_output (handle, __VA_ARGS__) -#define vl_printfun -#include -#undef vl_printfun - -/* Get the API version number */ -#define vl_api_version(n,v) static u32 api_version=(v); -#include -#undef vl_api_version - -#include "node_in.h" -#include "node_out.h" - -acl_main_t acl_main; - -/* - * A handy macro to set up a message reply. - * Assumes that the following variables are available: - * mp - pointer to request message - * rmp - pointer to reply message type - * rv - return value - */ - -#define REPLY_MACRO(t) \ -do { \ - unix_shared_memory_queue_t * q = \ - vl_api_client_index_to_input_queue (mp->client_index); \ - if (!q) \ - return; \ - \ - rmp = vl_msg_api_alloc (sizeof (*rmp)); \ - rmp->_vl_msg_id = ntohs((t)+sm->msg_id_base); \ - rmp->context = mp->context; \ - rmp->retval = ntohl(rv); \ - \ - vl_msg_api_send_shmem (q, (u8 *)&rmp); \ -} while(0); - -#define REPLY_MACRO2(t, body) \ -do { \ - unix_shared_memory_queue_t * q; \ - rv = vl_msg_api_pd_handler (mp, rv); \ - q = vl_api_client_index_to_input_queue (mp->client_index); \ - if (!q) \ - return; \ - \ - rmp = vl_msg_api_alloc (sizeof (*rmp)); \ - rmp->_vl_msg_id = ntohs((t)+am->msg_id_base); \ - rmp->context = mp->context; \ - rmp->retval = ntohl(rv); \ - do {body;} while (0); \ - vl_msg_api_send_shmem (q, (u8 *)&rmp); \ -} while(0); - -#define REPLY_MACRO3(t, n, body) \ -do { \ - unix_shared_memory_queue_t * q; \ - rv = vl_msg_api_pd_handler (mp, rv); \ - q = vl_api_client_index_to_input_queue (mp->client_index); \ - if (!q) \ - return; \ - \ - rmp = vl_msg_api_alloc (sizeof (*rmp) + n); \ - rmp->_vl_msg_id = ntohs((t)+am->msg_id_base); \ - rmp->context = mp->context; \ - rmp->retval = ntohl(rv); \ - do {body;} while (0); \ - vl_msg_api_send_shmem (q, (u8 *)&rmp); \ -} while(0); - - -/* List of message types that this plugin understands */ - -#define foreach_acl_plugin_api_msg \ -_(ACL_PLUGIN_GET_VERSION, acl_plugin_get_version) \ -_(ACL_ADD_REPLACE, acl_add_replace) \ -_(ACL_DEL, acl_del) \ -_(ACL_INTERFACE_ADD_DEL, acl_interface_add_del) \ -_(ACL_INTERFACE_SET_ACL_LIST, acl_interface_set_acl_list) \ -_(ACL_DUMP, acl_dump) \ -_(ACL_INTERFACE_LIST_DUMP, acl_interface_list_dump) \ -_(MACIP_ACL_ADD, macip_acl_add) \ -_(MACIP_ACL_DEL, macip_acl_del) \ -_(MACIP_ACL_INTERFACE_ADD_DEL, macip_acl_interface_add_del) \ -_(MACIP_ACL_DUMP, macip_acl_dump) \ -_(MACIP_ACL_INTERFACE_GET, macip_acl_interface_get) - -/* - * This routine exists to convince the vlib plugin framework that - * we haven't accidentally copied a random .dll into the plugin directory. - * - * Also collects global variable pointers passed from the vpp engine - */ - -clib_error_t * -vlib_plugin_register (vlib_main_t * vm, vnet_plugin_handoff_t * h, - int from_early_init) -{ - acl_main_t *am = &acl_main; - clib_error_t *error = 0; - - am->vlib_main = vm; - am->vnet_main = h->vnet_main; - am->ethernet_main = h->ethernet_main; - - l2sess_vlib_plugin_register(vm, h, from_early_init); - - return error; -} - - -static void -vl_api_acl_plugin_get_version_t_handler (vl_api_acl_plugin_get_version_t * mp) -{ - acl_main_t *am = &acl_main; - vl_api_acl_plugin_get_version_reply_t *rmp; - int msg_size = sizeof (*rmp); - unix_shared_memory_queue_t *q; - - q = vl_api_client_index_to_input_queue (mp->client_index); - if (q == 0) - { - return; - } - - rmp = vl_msg_api_alloc (msg_size); - memset (rmp, 0, msg_size); - rmp->_vl_msg_id = - ntohs (VL_API_ACL_PLUGIN_GET_VERSION_REPLY + am->msg_id_base); - rmp->context = mp->context; - rmp->major = htonl (ACL_PLUGIN_VERSION_MAJOR); - rmp->minor = htonl (ACL_PLUGIN_VERSION_MINOR); - - vl_msg_api_send_shmem (q, (u8 *) & rmp); -} - - -static int -acl_add_list (u32 count, vl_api_acl_rule_t rules[], - u32 * acl_list_index, u8 * tag) -{ - acl_main_t *am = &acl_main; - acl_list_t *a; - acl_rule_t *r; - acl_rule_t *acl_new_rules; - int i; - - if (*acl_list_index != ~0) - { - /* They supplied some number, let's see if this ACL exists */ - if (pool_is_free_index (am->acls, *acl_list_index)) - { - /* tried to replace a non-existent ACL, no point doing anything */ - return -1; - } - } - - /* Create and populate the rules */ - acl_new_rules = clib_mem_alloc_aligned (sizeof (acl_rule_t) * count, - CLIB_CACHE_LINE_BYTES); - if (!acl_new_rules) - { - /* Could not allocate rules. New or existing ACL - bail out regardless */ - return -1; - } - - for (i = 0; i < count; i++) - { - r = &acl_new_rules[i]; - r->is_permit = rules[i].is_permit; - r->is_ipv6 = rules[i].is_ipv6; - if (r->is_ipv6) - { - memcpy (&r->src, rules[i].src_ip_addr, sizeof (r->src)); - memcpy (&r->dst, rules[i].dst_ip_addr, sizeof (r->dst)); - } - else - { - memcpy (&r->src.ip4, rules[i].src_ip_addr, sizeof (r->src.ip4)); - memcpy (&r->dst.ip4, rules[i].dst_ip_addr, sizeof (r->dst.ip4)); - } - r->src_prefixlen = rules[i].src_ip_prefix_len; - r->dst_prefixlen = rules[i].dst_ip_prefix_len; - r->proto = rules[i].proto; - r->src_port_or_type_first = rules[i].srcport_or_icmptype_first; - r->src_port_or_type_last = rules[i].srcport_or_icmptype_last; - r->dst_port_or_code_first = rules[i].dstport_or_icmpcode_first; - r->dst_port_or_code_last = rules[i].dstport_or_icmpcode_last; - r->tcp_flags_value = rules[i].tcp_flags_value; - r->tcp_flags_mask = rules[i].tcp_flags_mask; - } - - if (~0 == *acl_list_index) - { - /* Get ACL index */ - pool_get_aligned (am->acls, a, CLIB_CACHE_LINE_BYTES); - memset (a, 0, sizeof (*a)); - /* Will return the newly allocated ACL index */ - *acl_list_index = a - am->acls; - } - else - { - a = am->acls + *acl_list_index; - /* Get rid of the old rules */ - clib_mem_free (a->rules); - } - a->rules = acl_new_rules; - a->count = count; - memcpy (a->tag, tag, sizeof (a->tag)); - - return 0; -} - -static int -acl_del_list (u32 acl_list_index) -{ - acl_main_t *am = &acl_main; - acl_list_t *a; - int i, ii; - if (pool_is_free_index (am->acls, acl_list_index)) - { - return -1; - } - - /* delete any references to the ACL */ - for (i = 0; i < vec_len (am->output_acl_vec_by_sw_if_index); i++) - { - for (ii = 0; ii < vec_len (am->output_acl_vec_by_sw_if_index[i]); - /* see body */ ) - { - if (acl_list_index == am->output_acl_vec_by_sw_if_index[i][ii]) - { - vec_del1 (am->output_acl_vec_by_sw_if_index[i], ii); - } - else - { - ii++; - } - } - } - for (i = 0; i < vec_len (am->input_acl_vec_by_sw_if_index); i++) - { - for (ii = 0; ii < vec_len (am->input_acl_vec_by_sw_if_index[i]); - /* see body */ ) - { - if (acl_list_index == am->input_acl_vec_by_sw_if_index[i][ii]) - { - vec_del1 (am->input_acl_vec_by_sw_if_index[i], ii); - } - else - { - ii++; - } - } - } - - /* now we can delete the ACL itself */ - a = &am->acls[acl_list_index]; - if (a->rules) - { - clib_mem_free (a->rules); - } - pool_put (am->acls, a); - return 0; -} - -/* Some aids in ASCII graphing the content */ -#define XX "\377" -#define __ "\000" -#define _(x) -#define v - -u8 ip4_5tuple_mask[] = -_(" dmac smac etype ") -_(ether) __ __ __ __ __ __ v __ __ __ __ __ __ v __ __ v - _(" v ihl totlen ") - _(0x0000) - __ __ __ __ - _(" ident fl+fo ") - _(0x0004) - __ __ __ __ - _(" ttl pr checksum ") - _(0x0008) - __ XX __ __ - _(" src address ") - _(0x000C) - XX XX XX XX - _(" dst address ") - _(0x0010) - XX XX XX XX - _("L4 T/U sport dport ") - _(tcpudp) - XX XX XX XX - _(padpad) - __ __ __ __ - _(padpad) - __ __ __ __ - _(padeth) - __ __; - - u8 ip6_5tuple_mask[] = - _(" dmac smac etype ") - _(ether) __ __ __ __ __ __ v __ __ __ __ __ __ v __ __ v - _(" v tc + flow ") - _(0x0000) __ __ __ __ - _(" plen nh hl ") - _(0x0004) __ __ XX __ - _(" src address ") - _(0x0008) XX XX XX XX - _(0x000C) XX XX XX XX - _(0x0010) XX XX XX XX - _(0x0014) XX XX XX XX - _(" dst address ") - _(0x0018) XX XX XX XX - _(0x001C) XX XX XX XX - _(0x0020) XX XX XX XX - _(0x0024) XX XX XX XX - _("L4T/U sport dport ") - _(tcpudp) XX XX XX XX _(padpad) __ __ __ __ _(padeth) __ __; - -#undef XX -#undef __ -#undef _ -#undef v - - static int count_skip (u8 * p, u32 size) -{ - u64 *p64 = (u64 *) p; - /* Be tolerant to null pointer */ - if (0 == p) - return 0; - - while ((0ULL == *p64) && ((u8 *) p64 - p) < size) - { - p64++; - } - return (p64 - (u64 *) p) / 2; -} - -static int -acl_classify_add_del_table_big (vnet_classify_main_t * cm, u8 * mask, - u32 mask_len, u32 next_table_index, - u32 miss_next_index, u32 * table_index, - int is_add) -{ - u32 nbuckets = 65536; - u32 memory_size = 2 << 30; - u32 skip = count_skip (mask, mask_len); - u32 match = (mask_len / 16) - skip; - u8 *skip_mask_ptr = mask + 16 * skip; - u32 current_data_flag = 0; - int current_data_offset = 0; - - if (0 == match) - match = 1; - - return vnet_classify_add_del_table (cm, skip_mask_ptr, nbuckets, - memory_size, skip, match, - next_table_index, miss_next_index, - table_index, current_data_flag, - current_data_offset, is_add, - 1 /* delete_chain */); -} - -static int -acl_classify_add_del_table_small (vnet_classify_main_t * cm, u8 * mask, - u32 mask_len, u32 next_table_index, - u32 miss_next_index, u32 * table_index, - int is_add) -{ - u32 nbuckets = 32; - u32 memory_size = 2 << 20; - u32 skip = count_skip (mask, mask_len); - u32 match = (mask_len / 16) - skip; - u8 *skip_mask_ptr = mask + 16 * skip; - u32 current_data_flag = 0; - int current_data_offset = 0; - - if (0 == match) - match = 1; - - return vnet_classify_add_del_table (cm, skip_mask_ptr, nbuckets, - memory_size, skip, match, - next_table_index, miss_next_index, - table_index, current_data_flag, - current_data_offset, is_add, - 1 /* delete_chain */); -} - - -static int -acl_unhook_l2_input_classify (acl_main_t * am, u32 sw_if_index) -{ - vnet_classify_main_t *cm = &vnet_classify_main; - u32 ip4_table_index = ~0; - u32 ip6_table_index = ~0; - - vec_validate_init_empty (am->acl_ip4_input_classify_table_by_sw_if_index, - sw_if_index, ~0); - vec_validate_init_empty (am->acl_ip6_input_classify_table_by_sw_if_index, - sw_if_index, ~0); - - vnet_l2_input_classify_enable_disable (sw_if_index, 0); - - if (am->acl_ip4_input_classify_table_by_sw_if_index[sw_if_index] != ~0) - { - ip4_table_index = - am->acl_ip4_input_classify_table_by_sw_if_index[sw_if_index]; - am->acl_ip4_input_classify_table_by_sw_if_index[sw_if_index] = ~0; - acl_classify_add_del_table_big (cm, ip4_5tuple_mask, - sizeof (ip4_5tuple_mask) - 1, ~0, - am->l2_input_classify_next_acl, - &ip4_table_index, 0); - } - if (am->acl_ip6_input_classify_table_by_sw_if_index[sw_if_index] != ~0) - { - ip6_table_index = - am->acl_ip6_input_classify_table_by_sw_if_index[sw_if_index]; - am->acl_ip6_input_classify_table_by_sw_if_index[sw_if_index] = ~0; - acl_classify_add_del_table_big (cm, ip6_5tuple_mask, - sizeof (ip6_5tuple_mask) - 1, ~0, - am->l2_input_classify_next_acl, - &ip6_table_index, 0); - } - - return 0; -} - -static int -acl_unhook_l2_output_classify (acl_main_t * am, u32 sw_if_index) -{ - vnet_classify_main_t *cm = &vnet_classify_main; - u32 ip4_table_index = ~0; - u32 ip6_table_index = ~0; - - vec_validate_init_empty (am->acl_ip4_output_classify_table_by_sw_if_index, - sw_if_index, ~0); - vec_validate_init_empty (am->acl_ip6_output_classify_table_by_sw_if_index, - sw_if_index, ~0); - - vnet_l2_output_classify_enable_disable (sw_if_index, 0); - - if (am->acl_ip4_output_classify_table_by_sw_if_index[sw_if_index] != ~0) - { - ip4_table_index = - am->acl_ip4_output_classify_table_by_sw_if_index[sw_if_index]; - am->acl_ip4_output_classify_table_by_sw_if_index[sw_if_index] = ~0; - acl_classify_add_del_table_big (cm, ip4_5tuple_mask, - sizeof (ip4_5tuple_mask) - 1, ~0, - am->l2_output_classify_next_acl, - &ip4_table_index, 0); - } - if (am->acl_ip6_output_classify_table_by_sw_if_index[sw_if_index] != ~0) - { - ip6_table_index = - am->acl_ip6_output_classify_table_by_sw_if_index[sw_if_index]; - am->acl_ip6_output_classify_table_by_sw_if_index[sw_if_index] = ~0; - acl_classify_add_del_table_big (cm, ip6_5tuple_mask, - sizeof (ip6_5tuple_mask) - 1, ~0, - am->l2_output_classify_next_acl, - &ip6_table_index, 0); - } - - return 0; -} - -static int -acl_hook_l2_input_classify (acl_main_t * am, u32 sw_if_index) -{ - vnet_classify_main_t *cm = &vnet_classify_main; - u32 ip4_table_index = ~0; - u32 ip6_table_index = ~0; - int rv; - - /* in case there were previous tables attached */ - acl_unhook_l2_input_classify (am, sw_if_index); - rv = - acl_classify_add_del_table_big (cm, ip4_5tuple_mask, - sizeof (ip4_5tuple_mask) - 1, ~0, - am->l2_input_classify_next_acl, - &ip4_table_index, 1); - if (rv) - return rv; - rv = - acl_classify_add_del_table_big (cm, ip6_5tuple_mask, - sizeof (ip6_5tuple_mask) - 1, ~0, - am->l2_input_classify_next_acl, - &ip6_table_index, 1); - if (rv) - { - acl_classify_add_del_table_big (cm, ip4_5tuple_mask, - sizeof (ip4_5tuple_mask) - 1, ~0, - am->l2_input_classify_next_acl, - &ip4_table_index, 0); - return rv; - } - rv = - vnet_l2_input_classify_set_tables (sw_if_index, ip4_table_index, - ip6_table_index, ~0); - clib_warning - ("ACL enabling on interface sw_if_index %d, setting tables to the following: ip4: %d ip6: %d\n", - sw_if_index, ip4_table_index, ip6_table_index); - if (rv) - { - acl_classify_add_del_table_big (cm, ip6_5tuple_mask, - sizeof (ip6_5tuple_mask) - 1, ~0, - am->l2_input_classify_next_acl, - &ip6_table_index, 0); - acl_classify_add_del_table_big (cm, ip4_5tuple_mask, - sizeof (ip4_5tuple_mask) - 1, ~0, - am->l2_input_classify_next_acl, - &ip4_table_index, 0); - return rv; - } - - am->acl_ip4_input_classify_table_by_sw_if_index[sw_if_index] = - ip4_table_index; - am->acl_ip6_input_classify_table_by_sw_if_index[sw_if_index] = - ip6_table_index; - - vnet_l2_input_classify_enable_disable (sw_if_index, 1); - return rv; -} - -static int -acl_hook_l2_output_classify (acl_main_t * am, u32 sw_if_index) -{ - vnet_classify_main_t *cm = &vnet_classify_main; - u32 ip4_table_index = ~0; - u32 ip6_table_index = ~0; - int rv; - - /* in case there were previous tables attached */ - acl_unhook_l2_output_classify (am, sw_if_index); - rv = - acl_classify_add_del_table_big (cm, ip4_5tuple_mask, - sizeof (ip4_5tuple_mask) - 1, ~0, - am->l2_output_classify_next_acl, - &ip4_table_index, 1); - if (rv) - return rv; - rv = - acl_classify_add_del_table_big (cm, ip6_5tuple_mask, - sizeof (ip6_5tuple_mask) - 1, ~0, - am->l2_output_classify_next_acl, - &ip6_table_index, 1); - if (rv) - { - acl_classify_add_del_table_big (cm, ip4_5tuple_mask, - sizeof (ip4_5tuple_mask) - 1, ~0, - am->l2_output_classify_next_acl, - &ip4_table_index, 0); - return rv; - } - rv = - vnet_l2_output_classify_set_tables (sw_if_index, ip4_table_index, - ip6_table_index, ~0); - clib_warning - ("ACL enabling on interface sw_if_index %d, setting tables to the following: ip4: %d ip6: %d\n", - sw_if_index, ip4_table_index, ip6_table_index); - if (rv) - { - acl_classify_add_del_table_big (cm, ip6_5tuple_mask, - sizeof (ip6_5tuple_mask) - 1, ~0, - am->l2_output_classify_next_acl, - &ip6_table_index, 0); - acl_classify_add_del_table_big (cm, ip4_5tuple_mask, - sizeof (ip4_5tuple_mask) - 1, ~0, - am->l2_output_classify_next_acl, - &ip4_table_index, 0); - return rv; - } - - am->acl_ip4_output_classify_table_by_sw_if_index[sw_if_index] = - ip4_table_index; - am->acl_ip6_output_classify_table_by_sw_if_index[sw_if_index] = - ip6_table_index; - - vnet_l2_output_classify_enable_disable (sw_if_index, 1); - return rv; -} - - -int -acl_interface_in_enable_disable (acl_main_t * am, u32 sw_if_index, - int enable_disable) -{ - int rv; - - /* Utterly wrong? */ - if (pool_is_free_index (am->vnet_main->interface_main.sw_interfaces, - sw_if_index)) - return VNET_API_ERROR_INVALID_SW_IF_INDEX; - - if (enable_disable) - { - rv = acl_hook_l2_input_classify (am, sw_if_index); - } - else - { - rv = acl_unhook_l2_input_classify (am, sw_if_index); - } - - return rv; -} - -int -acl_interface_out_enable_disable (acl_main_t * am, u32 sw_if_index, - int enable_disable) -{ - int rv; - - /* Utterly wrong? */ - if (pool_is_free_index (am->vnet_main->interface_main.sw_interfaces, - sw_if_index)) - return VNET_API_ERROR_INVALID_SW_IF_INDEX; - - if (enable_disable) - { - rv = acl_hook_l2_output_classify (am, sw_if_index); - } - else - { - rv = acl_unhook_l2_output_classify (am, sw_if_index); - } - - return rv; -} - - -static int -acl_interface_add_inout_acl (u32 sw_if_index, u8 is_input, u32 acl_list_index) -{ - acl_main_t *am = &acl_main; - if (is_input) - { - vec_validate (am->input_acl_vec_by_sw_if_index, sw_if_index); - vec_add (am->input_acl_vec_by_sw_if_index[sw_if_index], &acl_list_index, - 1); - acl_interface_in_enable_disable (am, sw_if_index, 1); - } - else - { - vec_validate (am->output_acl_vec_by_sw_if_index, sw_if_index); - vec_add (am->output_acl_vec_by_sw_if_index[sw_if_index], - &acl_list_index, 1); - acl_interface_out_enable_disable (am, sw_if_index, 1); - } - return 0; -} - -static int -acl_interface_del_inout_acl (u32 sw_if_index, u8 is_input, u32 acl_list_index) -{ - acl_main_t *am = &acl_main; - int i; - int rv = -1; - if (is_input) - { - vec_validate (am->input_acl_vec_by_sw_if_index, sw_if_index); - for (i = 0; i < vec_len (am->input_acl_vec_by_sw_if_index[sw_if_index]); - i++) - { - if (acl_list_index == - am->input_acl_vec_by_sw_if_index[sw_if_index][i]) - { - vec_del1 (am->input_acl_vec_by_sw_if_index[sw_if_index], i); - rv = 0; - break; - } - } - if (0 == vec_len (am->input_acl_vec_by_sw_if_index[sw_if_index])) - { - acl_interface_in_enable_disable (am, sw_if_index, 0); - } - } - else - { - vec_validate (am->output_acl_vec_by_sw_if_index, sw_if_index); - for (i = 0; - i < vec_len (am->output_acl_vec_by_sw_if_index[sw_if_index]); i++) - { - if (acl_list_index == - am->output_acl_vec_by_sw_if_index[sw_if_index][i]) - { - vec_del1 (am->output_acl_vec_by_sw_if_index[sw_if_index], i); - rv = 0; - break; - } - } - if (0 == vec_len (am->output_acl_vec_by_sw_if_index[sw_if_index])) - { - acl_interface_out_enable_disable (am, sw_if_index, 0); - } - } - return rv; -} - -static void -acl_interface_reset_inout_acls (u32 sw_if_index, u8 is_input) -{ - acl_main_t *am = &acl_main; - if (is_input) - { - acl_interface_in_enable_disable (am, sw_if_index, 0); - vec_validate (am->input_acl_vec_by_sw_if_index, sw_if_index); - vec_reset_length (am->input_acl_vec_by_sw_if_index[sw_if_index]); - } - else - { - acl_interface_out_enable_disable (am, sw_if_index, 0); - vec_validate (am->output_acl_vec_by_sw_if_index, sw_if_index); - vec_reset_length (am->output_acl_vec_by_sw_if_index[sw_if_index]); - } -} - -static int -acl_interface_add_del_inout_acl (u32 sw_if_index, u8 is_add, u8 is_input, - u32 acl_list_index) -{ - int rv = -1; - if (is_add) - { - rv = - acl_interface_add_inout_acl (sw_if_index, is_input, acl_list_index); - } - else - { - rv = - acl_interface_del_inout_acl (sw_if_index, is_input, acl_list_index); - } - return rv; -} - - -static void * -get_ptr_to_offset (vlib_buffer_t * b0, int offset) -{ - u8 *p = vlib_buffer_get_current (b0) + offset; - return p; -} - -static u8 -acl_get_l4_proto (vlib_buffer_t * b0, int node_is_ip6) -{ - u8 proto; - int proto_offset; - if (node_is_ip6) - { - proto_offset = 20; - } - else - { - proto_offset = 23; - } - proto = *((u8 *) vlib_buffer_get_current (b0) + proto_offset); - return proto; -} - -static int -acl_match_addr (ip46_address_t * addr1, ip46_address_t * addr2, int prefixlen, - int is_ip6) -{ - if (prefixlen == 0) - { - /* match any always succeeds */ - return 1; - } - if (is_ip6) - { - if (memcmp (addr1, addr2, prefixlen / 8)) - { - /* If the starting full bytes do not match, no point in bittwidling the thumbs further */ - return 0; - } - if (prefixlen % 8) - { - u8 b1 = *((u8 *) addr1 + 1 + prefixlen / 8); - u8 b2 = *((u8 *) addr2 + 1 + prefixlen / 8); - u8 mask0 = (0xff - ((1 << (8 - (prefixlen % 8))) - 1)); - return (b1 & mask0) == b2; - } - else - { - /* The prefix fits into integer number of bytes, so nothing left to do */ - return 1; - } - } - else - { - uint32_t a1 = ntohl (addr1->ip4.as_u32); - uint32_t a2 = ntohl (addr2->ip4.as_u32); - uint32_t mask0 = 0xffffffff - ((1 << (32 - prefixlen)) - 1); - return (a1 & mask0) == a2; - } -} - -static int -acl_match_port (u16 port, u16 port_first, u16 port_last, int is_ip6) -{ - return ((port >= port_first) && (port <= port_last)); -} - -static int -acl_packet_match (acl_main_t * am, u32 acl_index, vlib_buffer_t * b0, - u8 * r_action, int *r_is_ip6, u32 * r_acl_match_p, - u32 * r_rule_match_p, u32 * trace_bitmap) -{ - ethernet_header_t *h0; - u16 type0; - - ip46_address_t src, dst; - int is_ip6; - int is_ip4; - u8 proto; - u16 src_port; - u16 dst_port; - u8 tcp_flags = 0; - int i; - acl_list_t *a; - acl_rule_t *r; - - h0 = vlib_buffer_get_current (b0); - type0 = clib_net_to_host_u16 (h0->type); - is_ip4 = (type0 == ETHERNET_TYPE_IP4); - is_ip6 = (type0 == ETHERNET_TYPE_IP6); - - if (!(is_ip4 || is_ip6)) - { - return 0; - } - /* The bunch of hardcoded offsets here is intentional to get rid of them - ASAP, when getting to a faster matching code */ - if (is_ip4) - { - clib_memcpy (&src.ip4, get_ptr_to_offset (b0, 26), 4); - clib_memcpy (&dst.ip4, get_ptr_to_offset (b0, 30), 4); - proto = acl_get_l4_proto (b0, 0); - if (1 == proto) - { - *trace_bitmap |= 0x00000001; - /* type */ - src_port = *(u8 *) get_ptr_to_offset (b0, 34); - /* code */ - dst_port = *(u8 *) get_ptr_to_offset (b0, 35); - } - else - { - /* assume TCP/UDP */ - src_port = (*(u16 *) get_ptr_to_offset (b0, 34)); - dst_port = (*(u16 *) get_ptr_to_offset (b0, 36)); - /* UDP gets ability to check on an oddball data byte as a bonus */ - tcp_flags = *(u8 *) get_ptr_to_offset (b0, 14 + 20 + 13); - } - } - else /* is_ipv6 implicitly */ - { - clib_memcpy (&src, get_ptr_to_offset (b0, 22), 16); - clib_memcpy (&dst, get_ptr_to_offset (b0, 38), 16); - proto = acl_get_l4_proto (b0, 1); - if (58 == proto) - { - *trace_bitmap |= 0x00000002; - /* type */ - src_port = *(u8 *) get_ptr_to_offset (b0, 54); - /* code */ - dst_port = *(u8 *) get_ptr_to_offset (b0, 55); - } - else - { - /* assume TCP/UDP */ - src_port = (*(u16 *) get_ptr_to_offset (b0, 54)); - dst_port = (*(u16 *) get_ptr_to_offset (b0, 56)); - tcp_flags = *(u8 *) get_ptr_to_offset (b0, 14 + 40 + 13); - } - } - if (pool_is_free_index (am->acls, acl_index)) - { - if (r_acl_match_p) - *r_acl_match_p = acl_index; - if (r_rule_match_p) - *r_rule_match_p = -1; - /* the ACL does not exist but is used for policy. Block traffic. */ - return 0; - } - a = am->acls + acl_index; - for (i = 0; i < a->count; i++) - { - r = a->rules + i; - if (is_ip6 != r->is_ipv6) - { - continue; - } - if (!acl_match_addr (&dst, &r->dst, r->dst_prefixlen, is_ip6)) - continue; - if (!acl_match_addr (&src, &r->src, r->src_prefixlen, is_ip6)) - continue; - if (r->proto) - { - if (proto != r->proto) - continue; - if (!acl_match_port - (src_port, r->src_port_or_type_first, r->src_port_or_type_last, - is_ip6)) - continue; - if (!acl_match_port - (dst_port, r->dst_port_or_code_first, r->dst_port_or_code_last, - is_ip6)) - continue; - /* No need for check of proto == TCP, since in other rules both fields should be zero, so this match will succeed */ - if ((tcp_flags & r->tcp_flags_mask) != r->tcp_flags_value) - continue; - } - /* everything matches! */ - *r_action = r->is_permit; - *r_is_ip6 = is_ip6; - if (r_acl_match_p) - *r_acl_match_p = acl_index; - if (r_rule_match_p) - *r_rule_match_p = i; - return 1; - } - return 0; -} - -void -input_acl_packet_match (u32 sw_if_index, vlib_buffer_t * b0, u32 * nextp, - u32 * acl_match_p, u32 * rule_match_p, - u32 * trace_bitmap) -{ - acl_main_t *am = &acl_main; - uint8_t action = 0; - int is_ip6 = 0; - int i; - vec_validate (am->input_acl_vec_by_sw_if_index, sw_if_index); - for (i = 0; i < vec_len (am->input_acl_vec_by_sw_if_index[sw_if_index]); - i++) - { - if (acl_packet_match - (am, am->input_acl_vec_by_sw_if_index[sw_if_index][i], b0, &action, - &is_ip6, acl_match_p, rule_match_p, trace_bitmap)) - { - if (is_ip6) - { - *nextp = am->acl_in_ip6_match_next[action]; - } - else - { - *nextp = am->acl_in_ip4_match_next[action]; - } - return; - } - } - if (vec_len (am->input_acl_vec_by_sw_if_index[sw_if_index]) > 0) - { - /* If there are ACLs and none matched, deny by default */ - *nextp = 0; - } - -} - -void -output_acl_packet_match (u32 sw_if_index, vlib_buffer_t * b0, u32 * nextp, - u32 * acl_match_p, u32 * rule_match_p, - u32 * trace_bitmap) -{ - acl_main_t *am = &acl_main; - uint8_t action = 0; - int is_ip6 = 0; - int i; - vec_validate (am->output_acl_vec_by_sw_if_index, sw_if_index); - for (i = 0; i < vec_len (am->output_acl_vec_by_sw_if_index[sw_if_index]); - i++) - { - if (acl_packet_match - (am, am->output_acl_vec_by_sw_if_index[sw_if_index][i], b0, &action, - &is_ip6, acl_match_p, rule_match_p, trace_bitmap)) - { - if (is_ip6) - { - *nextp = am->acl_out_ip6_match_next[action]; - } - else - { - *nextp = am->acl_out_ip4_match_next[action]; - } - return; - } - } - if (vec_len (am->output_acl_vec_by_sw_if_index[sw_if_index]) > 0) - { - /* If there are ACLs and none matched, deny by default */ - *nextp = 0; - } -} - -typedef struct -{ - u8 is_ipv6; - u8 mac_mask[6]; - u8 prefix_len; - u32 count; - u32 table_index; -} macip_match_type_t; - -static u32 -macip_find_match_type (macip_match_type_t * mv, u8 * mac_mask, u8 prefix_len, - u8 is_ipv6) -{ - u32 i; - if (mv) - { - for (i = 0; i < vec_len (mv); i++) - { - if ((mv[i].prefix_len == prefix_len) && (mv[i].is_ipv6 == is_ipv6) - && (0 == memcmp (mv[i].mac_mask, mac_mask, 6))) - { - return i; - } - } - } - return ~0; -} - - -/* Get metric used to sort match types. - The more specific and the more often seen - the bigger the metric */ -static int -match_type_metric (macip_match_type_t * m) -{ - /* FIXME: count the ones in the MAC mask as well, check how well this heuristic works in real life */ - return m->prefix_len + m->is_ipv6 + 10 * m->count; -} - -static int -match_type_compare (macip_match_type_t * m1, macip_match_type_t * m2) -{ - /* Ascending sort based on the metric values */ - return match_type_metric (m1) - match_type_metric (m2); -} - -/* Get the offset of L3 source within ethernet packet */ -static int -get_l3_src_offset(int is6) -{ - if(is6) - return (sizeof(ethernet_header_t) + offsetof(ip6_header_t, src_address)); - else - return (sizeof(ethernet_header_t) + offsetof(ip4_header_t, src_address)); -} - -static int -macip_create_classify_tables (acl_main_t * am, u32 macip_acl_index) -{ - macip_match_type_t *mvec = NULL; - macip_match_type_t *mt; - macip_acl_list_t *a = &am->macip_acls[macip_acl_index]; - int i; - u32 match_type_index; - u32 last_table; - u8 mask[5 * 16]; - vnet_classify_main_t *cm = &vnet_classify_main; - - /* Count the number of different types of rules */ - for (i = 0; i < a->count; i++) - { - if (~0 == - (match_type_index = - macip_find_match_type (mvec, a->rules[i].src_mac_mask, - a->rules[i].src_prefixlen, - a->rules[i].is_ipv6))) - { - match_type_index = vec_len (mvec); - vec_validate (mvec, match_type_index); - memcpy (mvec[match_type_index].mac_mask, - a->rules[match_type_index].src_mac_mask, 6); - mvec[match_type_index].prefix_len = a->rules[i].src_prefixlen; - mvec[match_type_index].is_ipv6 = a->rules[i].is_ipv6; - mvec[match_type_index].table_index = ~0; - } - mvec[match_type_index].count++; - } - /* Put the most frequently used tables last in the list so we can create classifier tables in reverse order */ - vec_sort_with_function (mvec, match_type_compare); - /* Create the classifier tables */ - last_table = ~0; - vec_foreach (mt, mvec) - { - int mask_len; - int is6 = mt->is_ipv6; - int l3_src_offs = get_l3_src_offset(is6); - memset (mask, 0, sizeof (mask)); - memcpy (&mask[6], mt->mac_mask, 6); - for (i = 0; i < (mt->prefix_len / 8); i++) - { - mask[l3_src_offs + i] = 0xff; - } - if (mt->prefix_len % 8) - { - mask[l3_src_offs + (mt->prefix_len / 8)] = - 0xff - ((1 << (8 - mt->prefix_len % 8)) - 1); - } - /* - * Round-up the number of bytes needed to store the prefix, - * and round up the number of vectors too - */ - mask_len = ((l3_src_offs + ((mt->prefix_len+7) / 8) + - (sizeof (u32x4)-1))/sizeof(u32x4)) * sizeof (u32x4); - acl_classify_add_del_table_small (cm, mask, mask_len, last_table, - (~0 == last_table) ? 0 : ~0, &mt->table_index, - 1); - last_table = mt->table_index; - } - a->ip4_table_index = ~0; - a->ip6_table_index = ~0; - a->l2_table_index = last_table; - - /* Populate the classifier tables with rules from the MACIP ACL */ - for (i = 0; i < a->count; i++) - { - u32 action = 0; - u32 metadata = 0; - int is6 = a->rules[i].is_ipv6; - int l3_src_offs = get_l3_src_offset(is6); - memset (mask, 0, sizeof (mask)); - memcpy (&mask[6], a->rules[i].src_mac, 6); - if (is6) - { - memcpy (&mask[l3_src_offs], &a->rules[i].src_ip_addr.ip6, 16); - } - else - { - memcpy (&mask[l3_src_offs], &a->rules[i].src_ip_addr.ip4, 4); - } - match_type_index = - macip_find_match_type (mvec, a->rules[i].src_mac_mask, - a->rules[i].src_prefixlen, - a->rules[i].is_ipv6); - /* add session to table mvec[match_type_index].table_index; */ - vnet_classify_add_del_session (cm, mvec[match_type_index].table_index, - mask, a->rules[i].is_permit ? ~0 : 0, i, - 0, action, metadata, 1); - } - return 0; -} - -static void -macip_destroy_classify_tables (acl_main_t * am, u32 macip_acl_index) -{ - vnet_classify_main_t *cm = &vnet_classify_main; - macip_acl_list_t *a = &am->macip_acls[macip_acl_index]; - - if (a->ip4_table_index != ~0) - { - acl_classify_add_del_table_small (cm, 0, ~0, ~0, ~0, &a->ip4_table_index, 0); - a->ip4_table_index = ~0; - } - if (a->ip6_table_index != ~0) - { - acl_classify_add_del_table_small (cm, 0, ~0, ~0, ~0, &a->ip6_table_index, 0); - a->ip6_table_index = ~0; - } - if (a->l2_table_index != ~0) - { - acl_classify_add_del_table_small (cm, 0, ~0, ~0, ~0, &a->l2_table_index, 0); - a->l2_table_index = ~0; - } -} - -static int -macip_acl_add_list (u32 count, vl_api_macip_acl_rule_t rules[], - u32 * acl_list_index, u8 * tag) -{ - acl_main_t *am = &acl_main; - macip_acl_list_t *a; - macip_acl_rule_t *r; - macip_acl_rule_t *acl_new_rules; - int i; - - /* Create and populate the rules */ - acl_new_rules = clib_mem_alloc_aligned (sizeof (macip_acl_rule_t) * count, - CLIB_CACHE_LINE_BYTES); - if (!acl_new_rules) - { - /* Could not allocate rules. New or existing ACL - bail out regardless */ - return -1; - } - - for (i = 0; i < count; i++) - { - r = &acl_new_rules[i]; - r->is_permit = rules[i].is_permit; - r->is_ipv6 = rules[i].is_ipv6; - memcpy (&r->src_mac, rules[i].src_mac, 6); - memcpy (&r->src_mac_mask, rules[i].src_mac_mask, 6); - if(rules[i].is_ipv6) - memcpy (&r->src_ip_addr.ip6, rules[i].src_ip_addr, 16); - else - memcpy (&r->src_ip_addr.ip4, rules[i].src_ip_addr, 4); - r->src_prefixlen = rules[i].src_ip_prefix_len; - } - - /* Get ACL index */ - pool_get_aligned (am->macip_acls, a, CLIB_CACHE_LINE_BYTES); - memset (a, 0, sizeof (*a)); - /* Will return the newly allocated ACL index */ - *acl_list_index = a - am->macip_acls; - - a->rules = acl_new_rules; - a->count = count; - memcpy (a->tag, tag, sizeof (a->tag)); - - /* Create and populate the classifer tables */ - macip_create_classify_tables (am, *acl_list_index); - - return 0; -} - - -/* No check for validity of sw_if_index - the callers were supposed to validate */ - -static int -macip_acl_interface_del_acl (acl_main_t * am, u32 sw_if_index) -{ - int rv; - u32 macip_acl_index; - macip_acl_list_t *a; - vec_validate_init_empty (am->macip_acl_by_sw_if_index, sw_if_index, ~0); - macip_acl_index = am->macip_acl_by_sw_if_index[sw_if_index]; - /* No point in deleting MACIP ACL which is not applied */ - if (~0 == macip_acl_index) - return -1; - a = &am->macip_acls[macip_acl_index]; - /* remove the classifier tables off the interface L2 ACL */ - rv = - vnet_set_input_acl_intfc (am->vlib_main, sw_if_index, a->ip4_table_index, - a->ip6_table_index, a->l2_table_index, 0); - /* Unset the MACIP ACL index */ - am->macip_acl_by_sw_if_index[sw_if_index] = ~0; - return rv; -} - -/* No check for validity of sw_if_index - the callers were supposed to validate */ - -static int -macip_acl_interface_add_acl (acl_main_t * am, u32 sw_if_index, - u32 macip_acl_index) -{ - macip_acl_list_t *a; - int rv; - if (pool_is_free_index (am->macip_acls, macip_acl_index)) - { - return -1; - } - a = &am->macip_acls[macip_acl_index]; - vec_validate_init_empty (am->macip_acl_by_sw_if_index, sw_if_index, ~0); - /* If there already a MACIP ACL applied, unapply it */ - if (~0 != am->macip_acl_by_sw_if_index[sw_if_index]) - macip_acl_interface_del_acl(am, sw_if_index); - am->macip_acl_by_sw_if_index[sw_if_index] = macip_acl_index; - /* Apply the classifier tables for L2 ACLs */ - rv = - vnet_set_input_acl_intfc (am->vlib_main, sw_if_index, a->ip4_table_index, - a->ip6_table_index, a->l2_table_index, 1); - return rv; -} - -static int -macip_acl_del_list (u32 acl_list_index) -{ - acl_main_t *am = &acl_main; - macip_acl_list_t *a; - int i; - if (pool_is_free_index (am->macip_acls, acl_list_index)) - { - return -1; - } - - /* delete any references to the ACL */ - for (i = 0; i < vec_len (am->macip_acl_by_sw_if_index); i++) - { - if (am->macip_acl_by_sw_if_index[i] == acl_list_index) - { - macip_acl_interface_del_acl (am, i); - } - } - - /* Now that classifier tables are detached, clean them up */ - macip_destroy_classify_tables (am, acl_list_index); - - /* now we can delete the ACL itself */ - a = &am->macip_acls[acl_list_index]; - if (a->rules) - { - clib_mem_free (a->rules); - } - pool_put (am->macip_acls, a); - return 0; -} - - -static int -macip_acl_interface_add_del_acl (u32 sw_if_index, u8 is_add, - u32 acl_list_index) -{ - acl_main_t *am = &acl_main; - int rv = -1; - if (is_add) - { - rv = macip_acl_interface_add_acl (am, sw_if_index, acl_list_index); - } - else - { - rv = macip_acl_interface_del_acl (am, sw_if_index); - } - return rv; -} - -/* API message handler */ -static void -vl_api_acl_add_replace_t_handler (vl_api_acl_add_replace_t * mp) -{ - vl_api_acl_add_replace_reply_t *rmp; - acl_main_t *am = &acl_main; - int rv; - u32 acl_list_index = ntohl (mp->acl_index); - - rv = acl_add_list (ntohl (mp->count), mp->r, &acl_list_index, mp->tag); - - /* *INDENT-OFF* */ - REPLY_MACRO2(VL_API_ACL_ADD_REPLACE_REPLY, - ({ - rmp->acl_index = htonl(acl_list_index); - })); - /* *INDENT-ON* */ -} - -static void -vl_api_acl_del_t_handler (vl_api_acl_del_t * mp) -{ - acl_main_t *sm = &acl_main; - vl_api_acl_del_reply_t *rmp; - int rv; - - rv = acl_del_list (ntohl (mp->acl_index)); - - REPLY_MACRO (VL_API_ACL_DEL_REPLY); -} - -static void -vl_api_acl_interface_add_del_t_handler (vl_api_acl_interface_add_del_t * mp) -{ - acl_main_t *sm = &acl_main; - vnet_interface_main_t *im = &sm->vnet_main->interface_main; - u32 sw_if_index = ntohl (mp->sw_if_index); - vl_api_acl_interface_add_del_reply_t *rmp; - int rv = -1; - - if (pool_is_free_index(im->sw_interfaces, sw_if_index)) - rv = VNET_API_ERROR_INVALID_SW_IF_INDEX; - else - rv = - acl_interface_add_del_inout_acl (sw_if_index, mp->is_add, - mp->is_input, ntohl (mp->acl_index)); - - REPLY_MACRO (VL_API_ACL_INTERFACE_ADD_DEL_REPLY); -} - -static void -vl_api_acl_interface_set_acl_list_t_handler - (vl_api_acl_interface_set_acl_list_t * mp) -{ - acl_main_t *sm = &acl_main; - vl_api_acl_interface_set_acl_list_reply_t *rmp; - int rv = 0; - int i; - vnet_interface_main_t *im = &sm->vnet_main->interface_main; - u32 sw_if_index = ntohl (mp->sw_if_index); - - if (pool_is_free_index(im->sw_interfaces, sw_if_index)) - rv = VNET_API_ERROR_INVALID_SW_IF_INDEX; - else - { - acl_interface_reset_inout_acls (sw_if_index, 0); - acl_interface_reset_inout_acls (sw_if_index, 1); - - for (i = 0; i < mp->count; i++) - { - acl_interface_add_del_inout_acl (sw_if_index, 1, (i < mp->n_input), - ntohl (mp->acls[i])); - } - } - - REPLY_MACRO (VL_API_ACL_INTERFACE_SET_ACL_LIST_REPLY); -} - -static void -copy_acl_rule_to_api_rule (vl_api_acl_rule_t * api_rule, acl_rule_t * r) -{ - api_rule->is_permit = r->is_permit; - api_rule->is_ipv6 = r->is_ipv6; - if(r->is_ipv6) - { - memcpy (api_rule->src_ip_addr, &r->src, sizeof (r->src)); - memcpy (api_rule->dst_ip_addr, &r->dst, sizeof (r->dst)); - } - else - { - memcpy (api_rule->src_ip_addr, &r->src.ip4, sizeof (r->src.ip4)); - memcpy (api_rule->dst_ip_addr, &r->dst.ip4, sizeof (r->dst.ip4)); - } - api_rule->src_ip_prefix_len = r->src_prefixlen; - api_rule->dst_ip_prefix_len = r->dst_prefixlen; - api_rule->proto = r->proto; - api_rule->srcport_or_icmptype_first = r->src_port_or_type_first; - api_rule->srcport_or_icmptype_last = r->src_port_or_type_last; - api_rule->dstport_or_icmpcode_first = r->dst_port_or_code_first; - api_rule->dstport_or_icmpcode_last = r->dst_port_or_code_last; - api_rule->tcp_flags_mask = r->tcp_flags_mask; - api_rule->tcp_flags_value = r->tcp_flags_value; -} - -static void -send_acl_details (acl_main_t * am, unix_shared_memory_queue_t * q, - acl_list_t * acl, u32 context) -{ - vl_api_acl_details_t *mp; - vl_api_acl_rule_t *rules; - int i; - int msg_size = sizeof (*mp) + sizeof (mp->r[0]) * acl->count; - - mp = vl_msg_api_alloc (msg_size); - memset (mp, 0, msg_size); - mp->_vl_msg_id = ntohs (VL_API_ACL_DETAILS + am->msg_id_base); - - /* fill in the message */ - mp->context = context; - mp->count = htonl (acl->count); - mp->acl_index = htonl (acl - am->acls); - memcpy (mp->tag, acl->tag, sizeof (mp->tag)); - // clib_memcpy (mp->r, acl->rules, acl->count * sizeof(acl->rules[0])); - rules = mp->r; - for (i = 0; i < acl->count; i++) - { - copy_acl_rule_to_api_rule (&rules[i], &acl->rules[i]); - } - - clib_warning("Sending acl details for ACL index %d", ntohl(mp->acl_index)); - vl_msg_api_send_shmem (q, (u8 *) & mp); -} - - -static void -vl_api_acl_dump_t_handler (vl_api_acl_dump_t * mp) -{ - acl_main_t *am = &acl_main; - u32 acl_index; - acl_list_t *acl; - - int rv = -1; - unix_shared_memory_queue_t *q; - - q = vl_api_client_index_to_input_queue (mp->client_index); - if (q == 0) - { - return; - } - - if (mp->acl_index == ~0) - { - /* *INDENT-OFF* */ - /* Just dump all ACLs */ - pool_foreach (acl, am->acls, - ({ - send_acl_details(am, q, acl, mp->context); - })); - /* *INDENT-ON* */ - } - else - { - acl_index = ntohl (mp->acl_index); - if (!pool_is_free_index (am->acls, acl_index)) - { - acl = &am->acls[acl_index]; - send_acl_details (am, q, acl, mp->context); - } - } - - if (rv == -1) - { - /* FIXME API: should we signal an error here at all ? */ - return; - } -} - -static void -send_acl_interface_list_details (acl_main_t * am, - unix_shared_memory_queue_t * q, - u32 sw_if_index, u32 context) -{ - vl_api_acl_interface_list_details_t *mp; - int msg_size; - int n_input; - int n_output; - int count; - int i = 0; - - vec_validate (am->input_acl_vec_by_sw_if_index, sw_if_index); - vec_validate (am->output_acl_vec_by_sw_if_index, sw_if_index); - - n_input = vec_len (am->input_acl_vec_by_sw_if_index[sw_if_index]); - n_output = vec_len (am->output_acl_vec_by_sw_if_index[sw_if_index]); - count = n_input + n_output; - - msg_size = sizeof (*mp); - msg_size += sizeof (mp->acls[0]) * count; - - mp = vl_msg_api_alloc (msg_size); - memset (mp, 0, msg_size); - mp->_vl_msg_id = - ntohs (VL_API_ACL_INTERFACE_LIST_DETAILS + am->msg_id_base); - - /* fill in the message */ - mp->context = context; - mp->sw_if_index = htonl (sw_if_index); - mp->count = count; - mp->n_input = n_input; - for (i = 0; i < n_input; i++) - { - mp->acls[i] = htonl (am->input_acl_vec_by_sw_if_index[sw_if_index][i]); - } - for (i = 0; i < n_output; i++) - { - mp->acls[n_input + i] = - htonl (am->output_acl_vec_by_sw_if_index[sw_if_index][i]); - } - - vl_msg_api_send_shmem (q, (u8 *) & mp); -} - -static void -vl_api_acl_interface_list_dump_t_handler (vl_api_acl_interface_list_dump_t * - mp) -{ - acl_main_t *am = &acl_main; - vnet_sw_interface_t *swif; - vnet_interface_main_t *im = &am->vnet_main->interface_main; - - u32 sw_if_index; - unix_shared_memory_queue_t *q; - - q = vl_api_client_index_to_input_queue (mp->client_index); - if (q == 0) - { - return; - } - - if (mp->sw_if_index == ~0) - { - /* *INDENT-OFF* */ - pool_foreach (swif, im->sw_interfaces, - ({ - send_acl_interface_list_details(am, q, swif->sw_if_index, mp->context); - })); - /* *INDENT-ON* */ - } - else - { - sw_if_index = ntohl (mp->sw_if_index); - if (!pool_is_free_index(im->sw_interfaces, sw_if_index)) - send_acl_interface_list_details (am, q, sw_if_index, mp->context); - } -} - -/* MACIP ACL API handlers */ - -static void -vl_api_macip_acl_add_t_handler (vl_api_macip_acl_add_t * mp) -{ - vl_api_macip_acl_add_reply_t *rmp; - acl_main_t *am = &acl_main; - int rv; - u32 acl_list_index = ~0; - - rv = - macip_acl_add_list (ntohl (mp->count), mp->r, &acl_list_index, mp->tag); - - /* *INDENT-OFF* */ - REPLY_MACRO2(VL_API_MACIP_ACL_ADD_REPLY, - ({ - rmp->acl_index = htonl(acl_list_index); - })); - /* *INDENT-ON* */ -} - -static void -vl_api_macip_acl_del_t_handler (vl_api_macip_acl_del_t * mp) -{ - acl_main_t *sm = &acl_main; - vl_api_macip_acl_del_reply_t *rmp; - int rv; - - rv = macip_acl_del_list (ntohl (mp->acl_index)); - - REPLY_MACRO (VL_API_MACIP_ACL_DEL_REPLY); -} - -static void -vl_api_macip_acl_interface_add_del_t_handler - (vl_api_macip_acl_interface_add_del_t * mp) -{ - acl_main_t *sm = &acl_main; - vl_api_macip_acl_interface_add_del_reply_t *rmp; - int rv = -1; - vnet_interface_main_t *im = &sm->vnet_main->interface_main; - u32 sw_if_index = ntohl (mp->sw_if_index); - - if (pool_is_free_index(im->sw_interfaces, sw_if_index)) - rv = VNET_API_ERROR_INVALID_SW_IF_INDEX; - else - rv = - macip_acl_interface_add_del_acl (ntohl (mp->sw_if_index), mp->is_add, - ntohl (mp->acl_index)); - - REPLY_MACRO (VL_API_MACIP_ACL_INTERFACE_ADD_DEL_REPLY); -} - -static void -send_macip_acl_details (acl_main_t * am, unix_shared_memory_queue_t * q, - macip_acl_list_t * acl, u32 context) -{ - vl_api_macip_acl_details_t *mp; - vl_api_macip_acl_rule_t *rules; - macip_acl_rule_t *r; - int i; - int msg_size = sizeof (*mp) + (acl ? sizeof (mp->r[0]) * acl->count : 0); - - mp = vl_msg_api_alloc (msg_size); - memset (mp, 0, msg_size); - mp->_vl_msg_id = ntohs (VL_API_MACIP_ACL_DETAILS + am->msg_id_base); - - /* fill in the message */ - mp->context = context; - if (acl) - { - memcpy (mp->tag, acl->tag, sizeof (mp->tag)); - mp->count = htonl (acl->count); - mp->acl_index = htonl (acl - am->macip_acls); - rules = mp->r; - for (i = 0; i < acl->count; i++) - { - r = &acl->rules[i]; - rules[i].is_permit = r->is_permit; - rules[i].is_ipv6 = r->is_ipv6; - memcpy (rules[i].src_mac, &r->src_mac, sizeof (r->src_mac)); - memcpy (rules[i].src_mac_mask, &r->src_mac_mask, - sizeof (r->src_mac_mask)); - if (r->is_ipv6) - memcpy (rules[i].src_ip_addr, &r->src_ip_addr.ip6, - sizeof (r->src_ip_addr.ip6)); - else - memcpy (rules[i].src_ip_addr, &r->src_ip_addr.ip4, - sizeof (r->src_ip_addr.ip4)); - rules[i].src_ip_prefix_len = r->src_prefixlen; - } - } - else - { - /* No martini, no party - no ACL applied to this interface. */ - mp->acl_index = ~0; - mp->count = 0; - } - - vl_msg_api_send_shmem (q, (u8 *) & mp); -} - - -static void -vl_api_macip_acl_dump_t_handler (vl_api_macip_acl_dump_t * mp) -{ - acl_main_t *am = &acl_main; - macip_acl_list_t *acl; - - unix_shared_memory_queue_t *q; - - q = vl_api_client_index_to_input_queue (mp->client_index); - if (q == 0) - { - return; - } - - if (mp->acl_index == ~0) - { - /* Just dump all ACLs for now, with sw_if_index = ~0 */ - pool_foreach (acl, am->macip_acls, ( - { - send_macip_acl_details (am, q, acl, - mp-> - context);} - )); - /* *INDENT-ON* */ - } - else - { - u32 acl_index = ntohl (mp->acl_index); - if (!pool_is_free_index (am->macip_acls, acl_index)) - { - acl = &am->macip_acls[acl_index]; - send_macip_acl_details (am, q, acl, mp->context); - } - } -} - -static void -vl_api_macip_acl_interface_get_t_handler (vl_api_macip_acl_interface_get_t * - mp) -{ - acl_main_t *am = &acl_main; - vl_api_macip_acl_interface_get_reply_t *rmp; - u32 count = vec_len (am->macip_acl_by_sw_if_index); - int msg_size = sizeof (*rmp) + sizeof (rmp->acls[0]) * count; - unix_shared_memory_queue_t *q; - int i; - - q = vl_api_client_index_to_input_queue (mp->client_index); - if (q == 0) - { - return; - } - - rmp = vl_msg_api_alloc (msg_size); - memset (rmp, 0, msg_size); - rmp->_vl_msg_id = - ntohs (VL_API_MACIP_ACL_INTERFACE_GET_REPLY + am->msg_id_base); - rmp->context = mp->context; - rmp->count = htonl (count); - for (i = 0; i < count; i++) - { - rmp->acls[i] = htonl (am->macip_acl_by_sw_if_index[i]); - } - - vl_msg_api_send_shmem (q, (u8 *) & rmp); -} - - - -/* Set up the API message handling tables */ -static clib_error_t * -acl_plugin_api_hookup (vlib_main_t * vm) -{ - acl_main_t *sm = &acl_main; -#define _(N,n) \ - vl_msg_api_set_handlers((VL_API_##N + sm->msg_id_base), \ - #n, \ - vl_api_##n##_t_handler, \ - vl_noop_handler, \ - vl_api_##n##_t_endian, \ - vl_api_##n##_t_print, \ - sizeof(vl_api_##n##_t), 1); - foreach_acl_plugin_api_msg; -#undef _ - - return 0; -} - -#define vl_msg_name_crc_list -#include -#undef vl_msg_name_crc_list - -static void -setup_message_id_table (acl_main_t * sm, api_main_t * am) -{ -#define _(id,n,crc) \ - vl_msg_api_add_msg_name_crc (am, #n "_" #crc, id + sm->msg_id_base); - foreach_vl_msg_name_crc_acl; -#undef _ -} - -u32 -register_match_action_nexts (u32 next_in_ip4, u32 next_in_ip6, - u32 next_out_ip4, u32 next_out_ip6) -{ - acl_main_t *am = &acl_main; - u32 act = am->n_match_actions; - if (am->n_match_actions == 255) - { - return ~0; - } - am->n_match_actions++; - am->acl_in_ip4_match_next[act] = next_in_ip4; - am->acl_in_ip6_match_next[act] = next_in_ip6; - am->acl_out_ip4_match_next[act] = next_out_ip4; - am->acl_out_ip6_match_next[act] = next_out_ip6; - return act; -} - -void -acl_setup_nodes (void) -{ - vlib_main_t *vm = vlib_get_main (); - acl_main_t *am = &acl_main; - vlib_node_t *n; - - n = vlib_get_node_by_name (vm, (u8 *) "l2-input-classify"); - am->l2_input_classify_next_acl = - vlib_node_add_next_with_slot (vm, n->index, acl_in_node.index, ~0); - n = vlib_get_node_by_name (vm, (u8 *) "l2-output-classify"); - am->l2_output_classify_next_acl = - vlib_node_add_next_with_slot (vm, n->index, acl_out_node.index, ~0); - - feat_bitmap_init_next_nodes (vm, acl_in_node.index, L2INPUT_N_FEAT, - l2input_get_feat_names (), - am->acl_in_node_input_next_node_index); - - memset (&am->acl_in_ip4_match_next[0], 0, - sizeof (am->acl_in_ip4_match_next)); - memset (&am->acl_in_ip6_match_next[0], 0, - sizeof (am->acl_in_ip6_match_next)); - memset (&am->acl_out_ip4_match_next[0], 0, - sizeof (am->acl_out_ip4_match_next)); - memset (&am->acl_out_ip6_match_next[0], 0, - sizeof (am->acl_out_ip6_match_next)); - am->n_match_actions = 0; - - register_match_action_nexts (0, 0, 0, 0); /* drop */ - register_match_action_nexts (~0, ~0, ~0, ~0); /* permit */ - register_match_action_nexts (ACL_IN_L2S_INPUT_IP4_ADD, ACL_IN_L2S_INPUT_IP6_ADD, ACL_OUT_L2S_OUTPUT_IP4_ADD, ACL_OUT_L2S_OUTPUT_IP6_ADD); /* permit + create session */ -} - - - -static clib_error_t * -acl_init (vlib_main_t * vm) -{ - acl_main_t *am = &acl_main; - clib_error_t *error = 0; - memset (am, 0, sizeof (*am)); - am->vlib_main = vm; - am->vnet_main = vnet_get_main (); - - u8 *name = format (0, "acl_%08x%c", api_version, 0); - - /* Ask for a correctly-sized block of API message decode slots */ - am->msg_id_base = vl_msg_api_get_msg_ids ((char *) name, - VL_MSG_FIRST_AVAILABLE); - - error = acl_plugin_api_hookup (vm); - acl_setup_nodes (); - - /* Add our API messages to the global name_crc hash table */ - setup_message_id_table (am, &api_main); - - vec_free (name); - - return error; -} - -VLIB_INIT_FUNCTION (acl_init); diff --git a/plugins/acl-plugin/acl/acl.h b/plugins/acl-plugin/acl/acl.h deleted file mode 100644 index afc9b289..00000000 --- a/plugins/acl-plugin/acl/acl.h +++ /dev/null @@ -1,148 +0,0 @@ -/* - * Copyright (c) 2016 Cisco and/or its affiliates. - * Licensed under the Apache License, Version 2.0 (the "License"); - * you may not use this file except in compliance with the License. - * You may obtain a copy of the License at: - * - * http://www.apache.org/licenses/LICENSE-2.0 - * - * Unless required by applicable law or agreed to in writing, software - * distributed under the License is distributed on an "AS IS" BASIS, - * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. - * See the License for the specific language governing permissions and - * limitations under the License. - */ -#ifndef included_acl_h -#define included_acl_h - -#include -#include -#include -#include - - -#include -#include -#include - -#define ACL_PLUGIN_VERSION_MAJOR 1 -#define ACL_PLUGIN_VERSION_MINOR 1 - -extern vlib_node_registration_t acl_in_node; -extern vlib_node_registration_t acl_out_node; - -void input_acl_packet_match(u32 sw_if_index, vlib_buffer_t * b0, u32 *nextp, u32 *acl_match_p, u32 *rule_match_p, u32 *trace_bitmap); -void output_acl_packet_match(u32 sw_if_index, vlib_buffer_t * b0, u32 *nextp, u32 *acl_match_p, u32 *rule_match_p, u32 *trace_bitmap); - -enum address_e { IP4, IP6 }; -typedef struct -{ - enum address_e type; - union { - ip6_address_t ip6; - ip4_address_t ip4; - } addr; -} address_t; - -/* - * ACL rules - */ -typedef struct -{ - u8 is_permit; - u8 is_ipv6; - ip46_address_t src; - u8 src_prefixlen; - ip46_address_t dst; - u8 dst_prefixlen; - u8 proto; - u16 src_port_or_type_first; - u16 src_port_or_type_last; - u16 dst_port_or_code_first; - u16 dst_port_or_code_last; - u8 tcp_flags_value; - u8 tcp_flags_mask; -} acl_rule_t; - -typedef struct -{ - u8 is_permit; - u8 is_ipv6; - u8 src_mac[6]; - u8 src_mac_mask[6]; - ip46_address_t src_ip_addr; - u8 src_prefixlen; -} macip_acl_rule_t; - -/* - * ACL - */ -typedef struct -{ - u8 tag[64]; - u32 count; - acl_rule_t *rules; -} acl_list_t; - -typedef struct -{ - u8 tag[64]; - u32 count; - macip_acl_rule_t *rules; - /* References to the classifier tables that will enforce the rules */ - u32 ip4_table_index; - u32 ip6_table_index; - u32 l2_table_index; -} macip_acl_list_t; - -typedef struct { - /* API message ID base */ - u16 msg_id_base; - - acl_list_t *acls; /* Pool of ACLs */ - macip_acl_list_t *macip_acls; /* Pool of MAC-IP ACLs */ - - /* ACLs associated with interfaces */ - u32 **input_acl_vec_by_sw_if_index; - u32 **output_acl_vec_by_sw_if_index; - - /* - * Classify tables used to grab the packets for the ACL check, - * and serving as the 5-tuple session tables at the same time - */ - u32 *acl_ip4_input_classify_table_by_sw_if_index; - u32 *acl_ip6_input_classify_table_by_sw_if_index; - u32 *acl_ip4_output_classify_table_by_sw_if_index; - u32 *acl_ip6_output_classify_table_by_sw_if_index; - - /* MACIP (input) ACLs associated with the interfaces */ - u32 *macip_acl_by_sw_if_index; - - /* next indices for our nodes in the l2-classify tables */ - u32 l2_input_classify_next_acl; - u32 l2_output_classify_next_acl; - - /* next node indices for feature bitmap */ - u32 acl_in_node_input_next_node_index[32]; - /* the respective thing for the output feature */ - l2_output_next_nodes_st acl_out_output_next_nodes; - - /* ACL match actions (must be coherent across in/out ACLs to next indices (can differ) */ - - u32 acl_in_ip4_match_next[256]; - u32 acl_in_ip6_match_next[256]; - u32 acl_out_ip4_match_next[256]; - u32 acl_out_ip6_match_next[256]; - u32 n_match_actions; - - - /* convenience */ - vlib_main_t * vlib_main; - vnet_main_t * vnet_main; - ethernet_main_t * ethernet_main; -} acl_main_t; - -extern acl_main_t acl_main; - - -#endif diff --git a/plugins/acl-plugin/acl/acl_all_api_h.h b/plugins/acl-plugin/acl/acl_all_api_h.h deleted file mode 100644 index 96eca56d..00000000 --- a/plugins/acl-plugin/acl/acl_all_api_h.h +++ /dev/null @@ -1,321 +0,0 @@ -/* - * Copyright (c) 2016 Cisco and/or its affiliates. - * Licensed under the Apache License, Version 2.0 (the "License"); - * you may not use this file except in compliance with the License. - * You may obtain a copy of the License at: - * - * http://www.apache.org/licenses/LICENSE-2.0 - * - * Unless required by applicable law or agreed to in writing, software - * distributed under the License is distributed on an "AS IS" BASIS, - * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. - * See the License for the specific language governing permissions and - * limitations under the License. - */ -/* Include the generated file, see BUILT_SOURCES in Makefile.am */ -#include - -#ifdef vl_printfun - -#ifdef LP64 -#define _uword_fmt "%lld" -#define _uword_cast (long long) -#else -#define _uword_fmt "%ld" -#define _uword_cast long -#endif - -static inline void * -vl_api_acl_rule_t_print (vl_api_acl_rule_t * a, void *handle) -{ - vl_print (handle, "vl_api_acl_rule_t:\n"); - vl_print (handle, "is_permit: %u\n", (unsigned) a->is_permit); - vl_print (handle, "is_ipv6: %u\n", (unsigned) a->is_ipv6); - { - int _i; - for (_i = 0; _i < 16; _i++) - { - vl_print (handle, "src_ip_addr[%d]: %u\n", _i, a->src_ip_addr[_i]); - } - } - vl_print (handle, "src_ip_prefix_len: %u\n", - (unsigned) a->src_ip_prefix_len); - { - int _i; - for (_i = 0; _i < 16; _i++) - { - vl_print (handle, "dst_ip_addr[%d]: %u\n", _i, a->dst_ip_addr[_i]); - } - } - vl_print (handle, "dst_ip_prefix_len: %u\n", - (unsigned) a->dst_ip_prefix_len); - vl_print (handle, "proto: %u\n", (unsigned) a->proto); - vl_print (handle, "srcport_or_icmptype_first: %u\n", - (unsigned) a->srcport_or_icmptype_first); - vl_print (handle, "srcport_or_icmptype_last: %u\n", - (unsigned) a->srcport_or_icmptype_last); - vl_print (handle, "dstport_or_icmpcode_first: %u\n", - (unsigned) a->dstport_or_icmpcode_first); - vl_print (handle, "dstport_or_icmpcode_last: %u\n", - (unsigned) a->dstport_or_icmpcode_last); - vl_print (handle, "tcp_flags_mask: %u\n", (unsigned) a->tcp_flags_mask); - vl_print (handle, "tcp_flags_value: %u\n", (unsigned) a->tcp_flags_value); - return handle; -} - -static inline void * -vl_api_acl_add_replace_t_print (vl_api_acl_add_replace_t * a, void *handle) -{ - int i; - vl_print (handle, "vl_api_acl_add_replace_t:\n"); - vl_print (handle, "_vl_msg_id: %u\n", (unsigned) a->_vl_msg_id); - vl_print (handle, "client_index: %u\n", (unsigned) a->client_index); - vl_print (handle, "context: %u\n", (unsigned) a->context); - vl_print (handle, "acl_index: %u\n", (unsigned) a->acl_index); - vl_print (handle, "count: %u\n", (unsigned) a->count); - vl_print (handle, "r ----- \n"); - for (i = 0; i < a->count; i++) - { - vl_print (handle, " r[%d]:\n", i); - vl_api_acl_rule_t_print (&a->r[i], handle); - } - vl_print (handle, "r ----- END \n"); - return handle; -} - - -static inline void *vl_api_acl_details_t_print (vl_api_acl_details_t *a,void *handle) -{ - vl_print(handle, "vl_api_acl_details_t:\n"); - vl_print(handle, "_vl_msg_id: %u\n", (unsigned) a->_vl_msg_id); - vl_print(handle, "context: %u\n", (unsigned) a->context); - vl_print(handle, "acl_index: %u\n", (unsigned) a->acl_index); - { - int _i; - for (_i = 0; _i < 64; _i++) { - vl_print(handle, "tag[%d]: %u\n", _i, a->tag[_i]); - } - } - vl_print(handle, "count: %u\n", (unsigned) a->count); - vl_print(handle, "r ----- \n"); - // FIXME vl_api_acl_rule_t_print(&a->r, handle); - vl_print(handle, "r ----- END \n"); - return handle; -} - -static inline void * -vl_api_macip_acl_rule_t_print (vl_api_macip_acl_rule_t * a, void *handle) -{ - vl_print (handle, "vl_api_macip_acl_rule_t:\n"); - vl_print (handle, "is_permit: %u\n", (unsigned) a->is_permit); - vl_print (handle, "is_ipv6: %u\n", (unsigned) a->is_ipv6); - { - int _i; - for (_i = 0; _i < 6; _i++) - { - vl_print (handle, "src_mac[%d]: %u\n", _i, a->src_mac[_i]); - } - } - { - int _i; - for (_i = 0; _i < 6; _i++) - { - vl_print (handle, "src_mac_mask[%d]: %u\n", _i, a->src_mac_mask[_i]); - } - } - { - int _i; - for (_i = 0; _i < 16; _i++) - { - vl_print (handle, "src_ip_addr[%d]: %u\n", _i, a->src_ip_addr[_i]); - } - } - vl_print (handle, "src_ip_prefix_len: %u\n", - (unsigned) a->src_ip_prefix_len); - return handle; -} - -static inline void * -vl_api_macip_acl_add_t_print (vl_api_macip_acl_add_t * a, void *handle) -{ - int i; - vl_print (handle, "vl_api_macip_acl_add_t:\n"); - vl_print (handle, "_vl_msg_id: %u\n", (unsigned) a->_vl_msg_id); - vl_print (handle, "client_index: %u\n", (unsigned) a->client_index); - vl_print (handle, "context: %u\n", (unsigned) a->context); - vl_print (handle, "count: %u\n", (unsigned) a->count); - vl_print (handle, "r ----- \n"); - for (i = 0; i < a->count; i++) - { - vl_print (handle, " r[%d]:\n", i); - vl_api_macip_acl_rule_t_print (&a->r[i], handle); - } - vl_print (handle, "r ----- END \n"); - return handle; -} - -static inline void *vl_api_macip_acl_details_t_print (vl_api_macip_acl_details_t *a,void *handle) -{ - int i; - vl_print(handle, "vl_api_macip_acl_details_t:\n"); - vl_print(handle, "_vl_msg_id: %u\n", (unsigned) a->_vl_msg_id); - vl_print(handle, "context: %u\n", (unsigned) a->context); - vl_print(handle, "acl_index: %u\n", (unsigned) a->acl_index); - { - int _i; - for (_i = 0; _i < 64; _i++) { - vl_print(handle, "tag[%d]: %u\n", _i, a->tag[_i]); - } - } - vl_print(handle, "count: %u\n", (unsigned) a->count); - vl_print(handle, "r ----- \n"); - for (i = 0; i < a->count; i++) - { - vl_print (handle, " r[%d]:\n", i); - vl_api_macip_acl_rule_t_print (&a->r[i], handle); - } - vl_print(handle, "r ----- END \n"); - return handle; -} - -#endif /* vl_printfun */ - - -#ifdef vl_endianfun - -#undef clib_net_to_host_uword -#ifdef LP64 -#define clib_net_to_host_uword clib_net_to_host_u64 -#else -#define clib_net_to_host_uword clib_net_to_host_u32 -#endif - -/* - * Manual endian/print functions created by copypasting the automatically - * generated ones with small required adjustments. Appears the codegen - * can't make code to print the contents of custom-type array. - */ - -static inline void -vl_api_acl_rule_t_endian (vl_api_acl_rule_t * a) -{ - /* a->is_permit = a->is_permit (no-op) */ - /* a->is_ipv6 = a->is_ipv6 (no-op) */ - /* a->src_ip_addr[0..15] = a->src_ip_addr[0..15] (no-op) */ - /* a->src_ip_prefix_len = a->src_ip_prefix_len (no-op) */ - /* a->dst_ip_addr[0..15] = a->dst_ip_addr[0..15] (no-op) */ - /* a->dst_ip_prefix_len = a->dst_ip_prefix_len (no-op) */ - /* a->proto = a->proto (no-op) */ - a->srcport_or_icmptype_first = - clib_net_to_host_u16 (a->srcport_or_icmptype_first); - a->srcport_or_icmptype_last = - clib_net_to_host_u16 (a->srcport_or_icmptype_last); - a->dstport_or_icmpcode_first = - clib_net_to_host_u16 (a->dstport_or_icmpcode_first); - a->dstport_or_icmpcode_last = - clib_net_to_host_u16 (a->dstport_or_icmpcode_last); - /* a->tcp_flags_mask = a->tcp_flags_mask (no-op) */ - /* a->tcp_flags_value = a->tcp_flags_value (no-op) */ -} - -static inline void -vl_api_acl_add_replace_t_endian (vl_api_acl_add_replace_t * a) -{ - int i; - a->_vl_msg_id = clib_net_to_host_u16 (a->_vl_msg_id); - a->client_index = clib_net_to_host_u32 (a->client_index); - a->context = clib_net_to_host_u32 (a->context); - a->acl_index = clib_net_to_host_u32 (a->acl_index); - a->count = clib_net_to_host_u32 (a->count); - for (i = 0; i < a->count; i++) - { - vl_api_acl_rule_t_endian (&a->r[i]); - } -} - -static inline void vl_api_acl_details_t_endian (vl_api_acl_details_t *a) -{ - int i; - a->_vl_msg_id = clib_net_to_host_u16(a->_vl_msg_id); - a->context = clib_net_to_host_u32(a->context); - a->acl_index = clib_net_to_host_u32(a->acl_index); - /* a->tag[0..63] = a->tag[0..63] (no-op) */ - a->count = clib_net_to_host_u32(a->count); - for (i = 0; i < a->count; i++) - { - vl_api_acl_rule_t_endian (&a->r[i]); - } -} - -static inline void vl_api_acl_interface_list_details_t_endian (vl_api_acl_interface_list_details_t *a) -{ - int i; - a->_vl_msg_id = clib_net_to_host_u16(a->_vl_msg_id); - a->context = clib_net_to_host_u32(a->context); - a->sw_if_index = clib_net_to_host_u32(a->sw_if_index); - /* a->count = a->count (no-op) */ - /* a->n_input = a->n_input (no-op) */ - for(i=0; icount; i++) { - a->acls[i] = clib_net_to_host_u32(a->acls[i]); - } -} - -static inline void vl_api_acl_interface_set_acl_list_t_endian (vl_api_acl_interface_set_acl_list_t *a) -{ - int i; - a->_vl_msg_id = clib_net_to_host_u16(a->_vl_msg_id); - a->client_index = clib_net_to_host_u32(a->client_index); - a->context = clib_net_to_host_u32(a->context); - a->sw_if_index = clib_net_to_host_u32(a->sw_if_index); - /* a->count = a->count (no-op) */ - /* a->n_input = a->n_input (no-op) */ - for(i=0; icount; i++) { - a->acls[i] = clib_net_to_host_u32(a->acls[i]); - } -} - -static inline void -vl_api_macip_acl_rule_t_endian (vl_api_macip_acl_rule_t * a) -{ - /* a->is_permit = a->is_permit (no-op) */ - /* a->is_ipv6 = a->is_ipv6 (no-op) */ - /* a->src_mac[0..5] = a->src_mac[0..5] (no-op) */ - /* a->src_mac_mask[0..5] = a->src_mac_mask[0..5] (no-op) */ - /* a->src_ip_addr[0..15] = a->src_ip_addr[0..15] (no-op) */ - /* a->src_ip_prefix_len = a->src_ip_prefix_len (no-op) */ -} - -static inline void -vl_api_macip_acl_add_t_endian (vl_api_macip_acl_add_t * a) -{ - int i; - a->_vl_msg_id = clib_net_to_host_u16 (a->_vl_msg_id); - a->client_index = clib_net_to_host_u32 (a->client_index); - a->context = clib_net_to_host_u32 (a->context); - a->count = clib_net_to_host_u32 (a->count); - for (i = 0; i < a->count; i++) - { - vl_api_macip_acl_rule_t_endian (&a->r[i]); - } -} - -static inline void vl_api_macip_acl_details_t_endian (vl_api_macip_acl_details_t *a) -{ - int i; - a->_vl_msg_id = clib_net_to_host_u16(a->_vl_msg_id); - a->context = clib_net_to_host_u32(a->context); - a->acl_index = clib_net_to_host_u32(a->acl_index); - /* a->tag[0..63] = a->tag[0..63] (no-op) */ - a->count = clib_net_to_host_u32(a->count); - for (i = 0; i < a->count; i++) - { - vl_api_macip_acl_rule_t_endian (&a->r[i]); - } -} - - - - -#endif /* vl_printfun */ - - diff --git a/plugins/acl-plugin/acl/acl_msg_enum.h b/plugins/acl-plugin/acl/acl_msg_enum.h deleted file mode 100644 index 14d8b48c..00000000 --- a/plugins/acl-plugin/acl/acl_msg_enum.h +++ /dev/null @@ -1,28 +0,0 @@ -/* - * Copyright (c) 2016 Cisco and/or its affiliates. - * Licensed under the Apache License, Version 2.0 (the "License"); - * you may not use this file except in compliance with the License. - * You may obtain a copy of the License at: - * - * http://www.apache.org/licenses/LICENSE-2.0 - * - * Unless required by applicable law or agreed to in writing, software - * distributed under the License is distributed on an "AS IS" BASIS, - * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. - * See the License for the specific language governing permissions and - * limitations under the License. - */ -#ifndef included_acl_msg_enum_h -#define included_acl_msg_enum_h - -#include - -#define vl_msg_id(n,h) n, -typedef enum { -#include - /* We'll want to know how many messages IDs we need... */ - VL_MSG_FIRST_AVAILABLE, -} vl_msg_id_t; -#undef vl_msg_id - -#endif diff --git a/plugins/acl-plugin/acl/acl_test.c b/plugins/acl-plugin/acl/acl_test.c deleted file mode 100644 index a0e413e1..00000000 --- a/plugins/acl-plugin/acl/acl_test.c +++ /dev/null @@ -1,1024 +0,0 @@ -/* - * Copyright (c) 2015 Cisco and/or its affiliates. - * Licensed under the Apache License, Version 2.0 (the "License"); - * you may not use this file except in compliance with the License. - * You may obtain a copy of the License at: - * - * http://www.apache.org/licenses/LICENSE-2.0 - * - * Unless required by applicable law or agreed to in writing, software - * distributed under the License is distributed on an "AS IS" BASIS, - * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. - * See the License for the specific language governing permissions and - * limitations under the License. - */ -/* - *------------------------------------------------------------------ - * acl_test.c - test harness plugin - *------------------------------------------------------------------ - */ - -#include -#include -#include -#include -#include -#include -#include - -uword unformat_sw_if_index (unformat_input_t * input, va_list * args); - -/* Declare message IDs */ -#include - -/* define message structures */ -#define vl_typedefs -#include -#undef vl_typedefs - -/* define message structures */ -#define vl_endianfun -#include -#undef vl_endianfun - -/* instantiate all the print functions we know about */ -#define vl_print(handle, ...) -#define vl_printfun -#include -#undef vl_printfun - -/* Get the API version number. */ -#define vl_api_version(n,v) static u32 api_version=(v); -#include -#undef vl_api_version - -typedef struct { - /* API message ID base */ - u16 msg_id_base; - vat_main_t *vat_main; -} acl_test_main_t; - -acl_test_main_t acl_test_main; - -#define foreach_standard_reply_retval_handler \ -_(acl_del_reply) \ -_(acl_interface_add_del_reply) \ -_(macip_acl_interface_add_del_reply) \ -_(acl_interface_set_acl_list_reply) \ -_(macip_acl_del_reply) - -#define foreach_reply_retval_aclindex_handler \ -_(acl_add_replace_reply) \ -_(macip_acl_add_reply) - -#define _(n) \ - static void vl_api_##n##_t_handler \ - (vl_api_##n##_t * mp) \ - { \ - vat_main_t * vam = acl_test_main.vat_main; \ - i32 retval = ntohl(mp->retval); \ - if (vam->async_mode) { \ - vam->async_errors += (retval < 0); \ - } else { \ - vam->retval = retval; \ - vam->result_ready = 1; \ - } \ - } -foreach_standard_reply_retval_handler; -#undef _ - -#define _(n) \ - static void vl_api_##n##_t_handler \ - (vl_api_##n##_t * mp) \ - { \ - vat_main_t * vam = acl_test_main.vat_main; \ - i32 retval = ntohl(mp->retval); \ - if (vam->async_mode) { \ - vam->async_errors += (retval < 0); \ - } else { \ - clib_warning("ACL index: %d", ntohl(mp->acl_index)); \ - vam->retval = retval; \ - vam->result_ready = 1; \ - } \ - } -foreach_reply_retval_aclindex_handler; -#undef _ - -/* These two ought to be in a library somewhere but they aren't */ -static uword -my_unformat_mac_address (unformat_input_t * input, va_list * args) -{ - u8 *a = va_arg (*args, u8 *); - return unformat (input, "%x:%x:%x:%x:%x:%x", &a[0], &a[1], &a[2], &a[3], - &a[4], &a[5]); -} - -static u8 * -my_format_mac_address (u8 * s, va_list * args) -{ - u8 *a = va_arg (*args, u8 *); - return format (s, "%02x:%02x:%02x:%02x:%02x:%02x", - a[0], a[1], a[2], a[3], a[4], a[5]); -} - - - -static void vl_api_acl_plugin_get_version_reply_t_handler - (vl_api_acl_plugin_get_version_reply_t * mp) - { - vat_main_t * vam = acl_test_main.vat_main; - clib_warning("ACL plugin version: %d.%d", ntohl(mp->major), ntohl(mp->minor)); - vam->result_ready = 1; - } - -static void vl_api_acl_interface_list_details_t_handler - (vl_api_acl_interface_list_details_t * mp) - { - int i; - vat_main_t * vam = acl_test_main.vat_main; - u8 *out = 0; - vl_api_acl_interface_list_details_t_endian(mp); - out = format(out, "sw_if_index: %d, count: %d, n_input: %d\n", mp->sw_if_index, mp->count, mp->n_input); - out = format(out, " input "); - for(i=0; icount; i++) { - out = format(out, "%d ", mp->acls[i]); - if (i == mp->n_input-1) - out = format(out, "\n output "); - } - out = format(out, "\n"); - clib_warning("%s", out); - vec_free(out); - vam->result_ready = 1; - } - - -static inline u8 * -vl_api_acl_rule_t_pretty_format (u8 *out, vl_api_acl_rule_t * a) -{ - int af = a->is_ipv6 ? AF_INET6 : AF_INET; - u8 src[INET6_ADDRSTRLEN]; - u8 dst[INET6_ADDRSTRLEN]; - inet_ntop(af, a->src_ip_addr, (void *)src, sizeof(src)); - inet_ntop(af, a->dst_ip_addr, (void *)dst, sizeof(dst)); - - out = format(out, "%s action %d src %s/%d dst %s/%d proto %d sport %d-%d dport %d-%d tcpflags %d %d", - a->is_ipv6 ? "ipv6" : "ipv4", a->is_permit, - src, a->src_ip_prefix_len, - dst, a->dst_ip_prefix_len, - a->proto, - a->srcport_or_icmptype_first, a->srcport_or_icmptype_last, - a->dstport_or_icmpcode_first, a->dstport_or_icmpcode_last, - a->tcp_flags_mask, a->tcp_flags_value); - return(out); -} - - - -static void vl_api_acl_details_t_handler - (vl_api_acl_details_t * mp) - { - int i; - vat_main_t * vam = acl_test_main.vat_main; - vl_api_acl_details_t_endian(mp); - u8 *out = 0; - out = format(0, "acl_index: %d, count: %d\n tag {%s}\n", mp->acl_index, mp->count, mp->tag); - for(i=0; icount; i++) { - out = format(out, " "); - out = vl_api_acl_rule_t_pretty_format(out, &mp->r[i]); - out = format(out, "%s\n", icount-1 ? "," : ""); - } - clib_warning("%s", out); - vec_free(out); - vam->result_ready = 1; - } - -static inline u8 * -vl_api_macip_acl_rule_t_pretty_format (u8 *out, vl_api_macip_acl_rule_t * a) -{ - int af = a->is_ipv6 ? AF_INET6 : AF_INET; - u8 src[INET6_ADDRSTRLEN]; - inet_ntop(af, a->src_ip_addr, (void *)src, sizeof(src)); - - out = format(out, "%s action %d ip %s/%d mac %U mask %U", - a->is_ipv6 ? "ipv6" : "ipv4", a->is_permit, - src, a->src_ip_prefix_len, - my_format_mac_address, a->src_mac, - my_format_mac_address, a->src_mac_mask); - return(out); -} - - -static void vl_api_macip_acl_details_t_handler - (vl_api_macip_acl_details_t * mp) - { - int i; - vat_main_t * vam = acl_test_main.vat_main; - vl_api_macip_acl_details_t_endian(mp); - u8 *out = format(0,"MACIP acl_index: %d, count: %d\n tag {%s}\n", mp->acl_index, mp->count, mp->tag); - for(i=0; icount; i++) { - out = format(out, " "); - out = vl_api_macip_acl_rule_t_pretty_format(out, &mp->r[i]); - out = format(out, "%s\n", icount-1 ? "," : ""); - } - clib_warning("%s", out); - vec_free(out); - vam->result_ready = 1; - } - -static void vl_api_macip_acl_interface_get_reply_t_handler - (vl_api_macip_acl_interface_get_reply_t * mp) - { - int i; - vat_main_t * vam = acl_test_main.vat_main; - u8 *out = format(0, "sw_if_index with MACIP ACL count: %d\n", ntohl(mp->count)); - for(i=0; icount); i++) { - out = format(out, " macip_acl_interface_add_del sw_if_index %d add acl %d\n", i, ntohl(mp->acls[i])); - } - out = format(out, "\n"); - clib_warning("%s", out); - vec_free(out); - vam->result_ready = 1; - } - - -/* - * Table of message reply handlers, must include boilerplate handlers - * we just generated - */ -#define foreach_vpe_api_reply_msg \ -_(ACL_ADD_REPLACE_REPLY, acl_add_replace_reply) \ -_(ACL_DEL_REPLY, acl_del_reply) \ -_(ACL_INTERFACE_ADD_DEL_REPLY, acl_interface_add_del_reply) \ -_(ACL_INTERFACE_SET_ACL_LIST_REPLY, acl_interface_set_acl_list_reply) \ -_(ACL_INTERFACE_LIST_DETAILS, acl_interface_list_details) \ -_(ACL_DETAILS, acl_details) \ -_(MACIP_ACL_ADD_REPLY, macip_acl_add_reply) \ -_(MACIP_ACL_DEL_REPLY, macip_acl_del_reply) \ -_(MACIP_ACL_DETAILS, macip_acl_details) \ -_(MACIP_ACL_INTERFACE_ADD_DEL_REPLY, macip_acl_interface_add_del_reply) \ -_(MACIP_ACL_INTERFACE_GET_REPLY, macip_acl_interface_get_reply) \ -_(ACL_PLUGIN_GET_VERSION_REPLY, acl_plugin_get_version_reply) - -/* M: construct, but don't yet send a message */ - -#define M(T,t) \ -do { \ - vam->result_ready = 0; \ - mp = vl_msg_api_alloc(sizeof(*mp)); \ - memset (mp, 0, sizeof (*mp)); \ - mp->_vl_msg_id = ntohs (VL_API_##T + sm->msg_id_base); \ - mp->client_index = vam->my_client_index; \ -} while(0); - -#define M2(T,t,n) \ -do { \ - vam->result_ready = 0; \ - mp = vl_msg_api_alloc(sizeof(*mp)+(n)); \ - memset (mp, 0, sizeof (*mp)); \ - mp->_vl_msg_id = ntohs (VL_API_##T + sm->msg_id_base); \ - mp->client_index = vam->my_client_index; \ -} while(0); - -/* S: send a message */ -#define S (vl_msg_api_send_shmem (vam->vl_input_queue, (u8 *)&mp)) - -/* W: wait for results, with timeout */ -#define W \ -do { \ - timeout = vat_time_now (vam) + 1.0; \ - \ - while (vat_time_now (vam) < timeout) { \ - if (vam->result_ready == 1) { \ - return (vam->retval); \ - } \ - } \ - return -99; \ -} while(0); - -static int api_acl_plugin_get_version (vat_main_t * vam) -{ - acl_test_main_t * sm = &acl_test_main; - vl_api_acl_plugin_get_version_t * mp; - u32 msg_size = sizeof(*mp); - f64 timeout; - - vam->result_ready = 0; - mp = vl_msg_api_alloc_as_if_client(msg_size); - memset (mp, 0, msg_size); - mp->_vl_msg_id = ntohs (VL_API_ACL_PLUGIN_GET_VERSION + sm->msg_id_base); - mp->client_index = vam->my_client_index; - - /* send it... */ - S; - - /* Wait for a reply... */ - W; - - return 0; -} - -static int api_macip_acl_interface_get (vat_main_t * vam) -{ - acl_test_main_t * sm = &acl_test_main; - vl_api_acl_plugin_get_version_t * mp; - u32 msg_size = sizeof(*mp); - f64 timeout; - - vam->result_ready = 0; - mp = vl_msg_api_alloc_as_if_client(msg_size); - memset (mp, 0, msg_size); - mp->_vl_msg_id = ntohs (VL_API_MACIP_ACL_INTERFACE_GET + sm->msg_id_base); - mp->client_index = vam->my_client_index; - - /* send it... */ - S; - - /* Wait for a reply... */ - W; - - return 0; -} - -#define vec_validate_acl_rules(v, idx) \ - do { \ - if (vec_len(v) < idx+1) { \ - vec_validate(v, idx); \ - v[idx].is_permit = 0x1; \ - v[idx].srcport_or_icmptype_last = 0xffff; \ - v[idx].dstport_or_icmpcode_last = 0xffff; \ - } \ - } while (0) - - -static int api_acl_add_replace (vat_main_t * vam) -{ - acl_test_main_t * sm = &acl_test_main; - unformat_input_t * i = vam->input; - f64 timeout; - vl_api_acl_add_replace_t * mp; - u32 acl_index = ~0; - u32 msg_size = sizeof (*mp); /* without the rules */ - - vl_api_acl_rule_t *rules = 0; - int rule_idx = 0; - int n_rules = 0; - u32 proto = 0; - u32 port1 = 0; - u32 port2 = 0; - u32 action = 0; - u32 tcpflags, tcpmask; - u32 src_prefix_length = 0, dst_prefix_length = 0; - ip4_address_t src_v4address, dst_v4address; - ip6_address_t src_v6address, dst_v6address; - u8 *tag = 0; - - if (!unformat (i, "%d", &acl_index)) { - /* Just assume -1 */ - } - - while (unformat_check_input (i) != UNFORMAT_END_OF_INPUT) - { - if (unformat (i, "ipv6")) - { - vec_validate_acl_rules(rules, rule_idx); - rules[rule_idx].is_ipv6 = 1; - } - else if (unformat (i, "ipv4")) - { - vec_validate_acl_rules(rules, rule_idx); - rules[rule_idx].is_ipv6 = 0; - } - else if (unformat (i, "permit+reflect")) - { - vec_validate_acl_rules(rules, rule_idx); - rules[rule_idx].is_permit = 2; - } - else if (unformat (i, "permit")) - { - vec_validate_acl_rules(rules, rule_idx); - rules[rule_idx].is_permit = 1; - } - else if (unformat (i, "action %d", &action)) - { - vec_validate_acl_rules(rules, rule_idx); - rules[rule_idx].is_permit = action; - } - else if (unformat (i, "src %U/%d", - unformat_ip4_address, &src_v4address, &src_prefix_length)) - { - vec_validate_acl_rules(rules, rule_idx); - memcpy (rules[rule_idx].src_ip_addr, &src_v4address, 4); - rules[rule_idx].src_ip_prefix_len = src_prefix_length; - rules[rule_idx].is_ipv6 = 0; - } - else if (unformat (i, "src %U/%d", - unformat_ip6_address, &src_v6address, &src_prefix_length)) - { - vec_validate_acl_rules(rules, rule_idx); - memcpy (rules[rule_idx].src_ip_addr, &src_v6address, 16); - rules[rule_idx].src_ip_prefix_len = src_prefix_length; - rules[rule_idx].is_ipv6 = 1; - } - else if (unformat (i, "dst %U/%d", - unformat_ip4_address, &dst_v4address, &dst_prefix_length)) - { - vec_validate_acl_rules(rules, rule_idx); - memcpy (rules[rule_idx].dst_ip_addr, &dst_v4address, 4); - rules[rule_idx].dst_ip_prefix_len = dst_prefix_length; - rules[rule_idx].is_ipv6 = 0; - } - else if (unformat (i, "dst %U/%d", - unformat_ip6_address, &dst_v6address, &dst_prefix_length)) - { - vec_validate_acl_rules(rules, rule_idx); - memcpy (rules[rule_idx].dst_ip_addr, &dst_v6address, 16); - rules[rule_idx].dst_ip_prefix_len = dst_prefix_length; - rules[rule_idx].is_ipv6 = 1; - } - else if (unformat (i, "sport %d-%d", &port1, &port2)) - { - vec_validate_acl_rules(rules, rule_idx); - rules[rule_idx].srcport_or_icmptype_first = htons(port1); - rules[rule_idx].srcport_or_icmptype_last = htons(port2); - } - else if (unformat (i, "sport %d", &port1)) - { - vec_validate_acl_rules(rules, rule_idx); - rules[rule_idx].srcport_or_icmptype_first = htons(port1); - rules[rule_idx].srcport_or_icmptype_last = htons(port1); - } - else if (unformat (i, "dport %d-%d", &port1, &port2)) - { - vec_validate_acl_rules(rules, rule_idx); - rules[rule_idx].dstport_or_icmpcode_first = htons(port1); - rules[rule_idx].dstport_or_icmpcode_last = htons(port2); - } - else if (unformat (i, "dport %d", &port1)) - { - vec_validate_acl_rules(rules, rule_idx); - rules[rule_idx].dstport_or_icmpcode_first = htons(port1); - rules[rule_idx].dstport_or_icmpcode_last = htons(port1); - } - else if (unformat (i, "tcpflags %d %d", &tcpflags, &tcpmask)) - { - vec_validate_acl_rules(rules, rule_idx); - rules[rule_idx].tcp_flags_value = tcpflags; - rules[rule_idx].tcp_flags_mask = tcpmask; - } - else if (unformat (i, "proto %d", &proto)) - { - vec_validate_acl_rules(rules, rule_idx); - rules[rule_idx].proto = proto; - } - else if (unformat (i, "tag %s", &tag)) - { - } - else if (unformat (i, ",")) - { - rule_idx++; - vec_validate_acl_rules(rules, rule_idx); - } - else - break; - } - - /* Construct the API message */ - vam->result_ready = 0; - - if(rules) - n_rules = vec_len(rules); - else - n_rules = 0; - - msg_size += n_rules*sizeof(rules[0]); - - mp = vl_msg_api_alloc_as_if_client(msg_size); - memset (mp, 0, msg_size); - mp->_vl_msg_id = ntohs (VL_API_ACL_ADD_REPLACE + sm->msg_id_base); - mp->client_index = vam->my_client_index; - if (n_rules > 0) - clib_memcpy(mp->r, rules, n_rules*sizeof (vl_api_acl_rule_t)); - if (tag) - { - if (vec_len(tag) >= sizeof(mp->tag)) - { - tag[sizeof(mp->tag)-1] = 0; - _vec_len(tag) = sizeof(mp->tag); - } - clib_memcpy(mp->tag, tag, vec_len(tag)); - vec_free(tag); - } - mp->acl_index = ntohl(acl_index); - mp->count = htonl(n_rules); - - /* send it... */ - S; - - /* Wait for a reply... */ - W; -} - -static int api_acl_del (vat_main_t * vam) -{ - acl_test_main_t * sm = &acl_test_main; - unformat_input_t * i = vam->input; - f64 timeout; - vl_api_acl_del_t * mp; - u32 acl_index = ~0; - - if (!unformat (i, "%d", &acl_index)) { - errmsg ("missing acl index\n"); - return -99; - } - - /* Construct the API message */ - M(ACL_DEL, acl_del); - mp->acl_index = ntohl(acl_index); - - /* send it... */ - S; - - /* Wait for a reply... */ - W; -} - -static int api_macip_acl_del (vat_main_t * vam) -{ - acl_test_main_t * sm = &acl_test_main; - unformat_input_t * i = vam->input; - f64 timeout; - vl_api_acl_del_t * mp; - u32 acl_index = ~0; - - if (!unformat (i, "%d", &acl_index)) { - errmsg ("missing acl index\n"); - return -99; - } - - /* Construct the API message */ - M(MACIP_ACL_DEL, acl_del); - mp->acl_index = ntohl(acl_index); - - /* send it... */ - S; - - /* Wait for a reply... */ - W; -} - -static int api_acl_interface_add_del (vat_main_t * vam) -{ - acl_test_main_t * sm = &acl_test_main; - unformat_input_t * i = vam->input; - f64 timeout; - vl_api_acl_interface_add_del_t * mp; - u32 sw_if_index = ~0; - u32 acl_index = ~0; - u8 is_input = 0; - u8 is_add = 0; - -// acl_interface_add_del | sw_if_index acl_index [out] [del] - - while (unformat_check_input (i) != UNFORMAT_END_OF_INPUT) - { - if (unformat (i, "%d", &acl_index)) - ; - else - break; - } - - - /* Parse args required to build the message */ - while (unformat_check_input (i) != UNFORMAT_END_OF_INPUT) { - if (unformat (i, "%U", unformat_sw_if_index, vam, &sw_if_index)) - ; - else if (unformat (i, "sw_if_index %d", &sw_if_index)) - ; - else if (unformat (i, "add")) - is_add = 1; - else if (unformat (i, "del")) - is_add = 0; - else if (unformat (i, "acl %d", &acl_index)) - ; - else if (unformat (i, "input")) - is_input = 1; - else if (unformat (i, "output")) - is_input = 0; - else - break; - } - - if (sw_if_index == ~0) { - errmsg ("missing interface name / explicit sw_if_index number \n"); - return -99; - } - - if (acl_index == ~0) { - errmsg ("missing ACL index\n"); - return -99; - } - - - - /* Construct the API message */ - M(ACL_INTERFACE_ADD_DEL, acl_interface_add_del); - mp->acl_index = ntohl(acl_index); - mp->sw_if_index = ntohl(sw_if_index); - mp->is_add = is_add; - mp->is_input = is_input; - - /* send it... */ - S; - - /* Wait for a reply... */ - W; -} - -static int api_macip_acl_interface_add_del (vat_main_t * vam) -{ - acl_test_main_t * sm = &acl_test_main; - unformat_input_t * i = vam->input; - f64 timeout; - vl_api_macip_acl_interface_add_del_t * mp; - u32 sw_if_index = ~0; - u32 acl_index = ~0; - u8 is_add = 0; - - /* Parse args required to build the message */ - while (unformat_check_input (i) != UNFORMAT_END_OF_INPUT) { - if (unformat (i, "%U", unformat_sw_if_index, vam, &sw_if_index)) - ; - else if (unformat (i, "sw_if_index %d", &sw_if_index)) - ; - else if (unformat (i, "add")) - is_add = 1; - else if (unformat (i, "del")) - is_add = 0; - else if (unformat (i, "acl %d", &acl_index)) - ; - else - break; - } - - if (sw_if_index == ~0) { - errmsg ("missing interface name / explicit sw_if_index number \n"); - return -99; - } - - if (acl_index == ~0) { - errmsg ("missing ACL index\n"); - return -99; - } - - - - /* Construct the API message */ - M(MACIP_ACL_INTERFACE_ADD_DEL, macip_acl_interface_add_del); - mp->acl_index = ntohl(acl_index); - mp->sw_if_index = ntohl(sw_if_index); - mp->is_add = is_add; - - /* send it... */ - S; - - /* Wait for a reply... */ - W; -} - -static int api_acl_interface_set_acl_list (vat_main_t * vam) -{ - acl_test_main_t * sm = &acl_test_main; - unformat_input_t * i = vam->input; - f64 timeout; - vl_api_acl_interface_set_acl_list_t * mp; - u32 sw_if_index = ~0; - u32 acl_index = ~0; - u32 *inacls = 0; - u32 *outacls = 0; - u8 is_input = 0; - -// acl_interface_set_acl_list | sw_if_index input [acl-idx list] output [acl-idx list] - - /* Parse args required to build the message */ - while (unformat_check_input (i) != UNFORMAT_END_OF_INPUT) { - if (unformat (i, "%U", unformat_sw_if_index, vam, &sw_if_index)) - ; - else if (unformat (i, "sw_if_index %d", &sw_if_index)) - ; - else if (unformat (i, "%d", &acl_index)) - { - if(is_input) - vec_add1(inacls, htonl(acl_index)); - else - vec_add1(outacls, htonl(acl_index)); - } - else if (unformat (i, "acl %d", &acl_index)) - ; - else if (unformat (i, "input")) - is_input = 1; - else if (unformat (i, "output")) - is_input = 0; - else - break; - } - - if (sw_if_index == ~0) { - errmsg ("missing interface name / explicit sw_if_index number \n"); - return -99; - } - - /* Construct the API message */ - M2(ACL_INTERFACE_SET_ACL_LIST, acl_interface_set_acl_list, sizeof(u32) * (vec_len(inacls) + vec_len(outacls))); - mp->sw_if_index = ntohl(sw_if_index); - mp->n_input = vec_len(inacls); - mp->count = vec_len(inacls) + vec_len(outacls); - vec_append(inacls, outacls); - if (vec_len(inacls) > 0) - clib_memcpy(mp->acls, inacls, vec_len(inacls)*sizeof(u32)); - - /* send it... */ - S; - - /* Wait for a reply... */ - W; -} - - -static int api_acl_interface_list_dump (vat_main_t * vam) -{ - acl_test_main_t * sm = &acl_test_main; - unformat_input_t * i = vam->input; - f64 timeout; - u32 sw_if_index = ~0; - vl_api_acl_interface_list_dump_t * mp; - - /* Parse args required to build the message */ - while (unformat_check_input (i) != UNFORMAT_END_OF_INPUT) { - if (unformat (i, "%U", unformat_sw_if_index, vam, &sw_if_index)) - ; - else if (unformat (i, "sw_if_index %d", &sw_if_index)) - ; - else - break; - } - - /* Construct the API message */ - M(ACL_INTERFACE_LIST_DUMP, acl_interface_list_dump); - mp->sw_if_index = ntohl (sw_if_index); - - /* send it... */ - S; - - /* Wait for a reply... */ - W; -} - -static int api_acl_dump (vat_main_t * vam) -{ - acl_test_main_t * sm = &acl_test_main; - unformat_input_t * i = vam->input; - f64 timeout; - u32 acl_index = ~0; - vl_api_acl_dump_t * mp; - - /* Parse args required to build the message */ - while (unformat_check_input (i) != UNFORMAT_END_OF_INPUT) { - if (unformat (i, "%d", &acl_index)) - ; - else - break; - } - - /* Construct the API message */ - M(ACL_DUMP, acl_dump); - mp->acl_index = ntohl (acl_index); - - /* send it... */ - S; - - /* Wait for a reply... */ - W; -} - -static int api_macip_acl_dump (vat_main_t * vam) -{ - acl_test_main_t * sm = &acl_test_main; - unformat_input_t * i = vam->input; - f64 timeout; - u32 acl_index = ~0; - vl_api_acl_dump_t * mp; - - /* Parse args required to build the message */ - while (unformat_check_input (i) != UNFORMAT_END_OF_INPUT) { - if (unformat (i, "%d", &acl_index)) - ; - else - break; - } - - /* Construct the API message */ - M(MACIP_ACL_DUMP, macip_acl_dump); - mp->acl_index = ntohl (acl_index); - - /* send it... */ - S; - - /* Wait for a reply... */ - W; -} - -#define vec_validate_macip_acl_rules(v, idx) \ - do { \ - if (vec_len(v) < idx+1) { \ - vec_validate(v, idx); \ - v[idx].is_permit = 0x1; \ - } \ - } while (0) - - -static int api_macip_acl_add (vat_main_t * vam) -{ - acl_test_main_t * sm = &acl_test_main; - unformat_input_t * i = vam->input; - f64 timeout; - vl_api_macip_acl_add_t * mp; - u32 msg_size = sizeof (*mp); /* without the rules */ - - vl_api_macip_acl_rule_t *rules = 0; - int rule_idx = 0; - int n_rules = 0; - u32 src_prefix_length = 0; - u32 action = 0; - ip4_address_t src_v4address; - ip6_address_t src_v6address; - u8 src_mac[6]; - u8 *tag = 0; - u8 mac_mask_all_1[6] = { 0xff, 0xff, 0xff, 0xff, 0xff, 0xff }; - - while (unformat_check_input (i) != UNFORMAT_END_OF_INPUT) - { - if (unformat (i, "ipv6")) - { - vec_validate_macip_acl_rules(rules, rule_idx); - rules[rule_idx].is_ipv6 = 1; - } - else if (unformat (i, "ipv4")) - { - vec_validate_macip_acl_rules(rules, rule_idx); - rules[rule_idx].is_ipv6 = 1; - } - else if (unformat (i, "permit")) - { - vec_validate_macip_acl_rules(rules, rule_idx); - rules[rule_idx].is_permit = 1; - } - else if (unformat (i, "deny")) - { - vec_validate_macip_acl_rules(rules, rule_idx); - rules[rule_idx].is_permit = 0; - } - else if (unformat (i, "action %d", &action)) - { - vec_validate_macip_acl_rules(rules, rule_idx); - rules[rule_idx].is_permit = action; - } - else if (unformat (i, "ip %U/%d", - unformat_ip4_address, &src_v4address, &src_prefix_length)) - { - vec_validate_macip_acl_rules(rules, rule_idx); - memcpy (rules[rule_idx].src_ip_addr, &src_v4address, 4); - rules[rule_idx].src_ip_prefix_len = src_prefix_length; - rules[rule_idx].is_ipv6 = 0; - } - else if (unformat (i, "ip %U/%d", - unformat_ip6_address, &src_v6address, &src_prefix_length)) - { - vec_validate_macip_acl_rules(rules, rule_idx); - memcpy (rules[rule_idx].src_ip_addr, &src_v6address, 16); - rules[rule_idx].src_ip_prefix_len = src_prefix_length; - rules[rule_idx].is_ipv6 = 1; - } - else if (unformat (i, "mac %U", - my_unformat_mac_address, &src_mac)) - { - vec_validate_macip_acl_rules(rules, rule_idx); - memcpy (rules[rule_idx].src_mac, &src_mac, 6); - memcpy (rules[rule_idx].src_mac_mask, &mac_mask_all_1, 6); - } - else if (unformat (i, "mask %U", - my_unformat_mac_address, &src_mac)) - { - vec_validate_macip_acl_rules(rules, rule_idx); - memcpy (rules[rule_idx].src_mac_mask, &src_mac, 6); - } - else if (unformat (i, "tag %s", &tag)) - { - } - else if (unformat (i, ",")) - { - rule_idx++; - vec_validate_macip_acl_rules(rules, rule_idx); - } - else - break; - } - - /* Construct the API message */ - vam->result_ready = 0; - - if(rules) - n_rules = vec_len(rules); - else - n_rules = 0; - - msg_size += n_rules*sizeof(rules[0]); - - mp = vl_msg_api_alloc_as_if_client(msg_size); - memset (mp, 0, msg_size); - mp->_vl_msg_id = ntohs (VL_API_MACIP_ACL_ADD + sm->msg_id_base); - mp->client_index = vam->my_client_index; - if (n_rules > 0) - clib_memcpy(mp->r, rules, n_rules*sizeof (mp->r[0])); - if (tag) - { - if (vec_len(tag) >= sizeof(mp->tag)) - { - tag[sizeof(mp->tag)-1] = 0; - _vec_len(tag) = sizeof(mp->tag); - } - clib_memcpy(mp->tag, tag, vec_len(tag)); - vec_free(tag); - } - - mp->count = htonl(n_rules); - - /* send it... */ - S; - - /* Wait for a reply... */ - W; -} - -/* - * List of messages that the api test plugin sends, - * and that the data plane plugin processes - */ -#define foreach_vpe_api_msg \ -_(acl_plugin_get_version, "") \ -_(acl_add_replace, " [ [src IP/plen] [dst IP/plen] [sport X-Y] [dport X-Y] [proto P] [tcpflags FL MASK], ... , ...") \ -_(acl_del, "") \ -_(acl_dump, "[]") \ -_(acl_interface_add_del, " | sw_if_index [add|del] [input|output] acl ") \ -_(acl_interface_set_acl_list, " | sw_if_index input [acl-idx list] output [acl-idx list]") \ -_(acl_interface_list_dump, "[ | sw_if_index ]") \ -_(macip_acl_add, "...") \ -_(macip_acl_del, "")\ -_(macip_acl_dump, "[]") \ -_(macip_acl_interface_add_del, " | sw_if_index [add|del] acl ") \ -_(macip_acl_interface_get, "") - - - -void vat_api_hookup (vat_main_t *vam) -{ - acl_test_main_t * sm = &acl_test_main; - /* Hook up handlers for replies from the data plane plug-in */ -#define _(N,n) \ - vl_msg_api_set_handlers((VL_API_##N + sm->msg_id_base), \ - #n, \ - vl_api_##n##_t_handler, \ - vl_noop_handler, \ - vl_api_##n##_t_endian, \ - vl_api_##n##_t_print, \ - sizeof(vl_api_##n##_t), 1); - foreach_vpe_api_reply_msg; -#undef _ - - /* API messages we can send */ -#define _(n,h) hash_set_mem (vam->function_by_name, #n, api_##n); - foreach_vpe_api_msg; -#undef _ - - /* Help strings */ -#define _(n,h) hash_set_mem (vam->help_by_name, #n, h); - foreach_vpe_api_msg; -#undef _ -} - -clib_error_t * vat_plugin_register (vat_main_t *vam) -{ - acl_test_main_t * sm = &acl_test_main; - u8 * name; - - sm->vat_main = vam; - - name = format (0, "acl_%08x%c", api_version, 0); - sm->msg_id_base = vl_client_get_first_plugin_msg_id ((char *) name); - - if (sm->msg_id_base != (u16) ~0) - vat_api_hookup (vam); - - vec_free(name); - - return 0; -} diff --git a/plugins/acl-plugin/acl/jvpp/io/fd/vpp/jvpp/acl/test/AclExpectedDumpData.java b/plugins/acl-plugin/acl/jvpp/io/fd/vpp/jvpp/acl/test/AclExpectedDumpData.java deleted file mode 100644 index 979edbc4..00000000 --- a/plugins/acl-plugin/acl/jvpp/io/fd/vpp/jvpp/acl/test/AclExpectedDumpData.java +++ /dev/null @@ -1,135 +0,0 @@ -/* - * Copyright (c) 2016 Cisco and/or its affiliates. - * - * Licensed under the Apache License, Version 2.0 (the "License"); - * you may not use this file except in compliance with the License. - * You may obtain a copy of the License at: - * - * http://www.apache.org/licenses/LICENSE-2.0 - * - * Unless required by applicable law or agreed to in writing, software - * distributed under the License is distributed on an "AS IS" BASIS, - * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. - * See the License for the specific language governing permissions and - * limitations under the License. - */ - -package io.fd.vpp.jvpp.acl.test; - - -import static io.fd.vpp.jvpp.acl.test.AclTestData.FIRST_RULE_ADDRESS_2_AS_ARRAY; -import static io.fd.vpp.jvpp.acl.test.AclTestData.FIRST_RULE_ADDRESS_AS_ARRAY; -import static io.fd.vpp.jvpp.acl.test.AclTestData.FIRST_RULE_DST_ICMP_TYPE_END; -import static io.fd.vpp.jvpp.acl.test.AclTestData.FIRST_RULE_DST_ICMP_TYPE_START; -import static io.fd.vpp.jvpp.acl.test.AclTestData.FIRST_RULE_MAC; -import static io.fd.vpp.jvpp.acl.test.AclTestData.FIRST_RULE_MAC_MASK; -import static io.fd.vpp.jvpp.acl.test.AclTestData.FIRST_RULE_PREFIX; -import static io.fd.vpp.jvpp.acl.test.AclTestData.FIRST_RULE_PREFIX_2; -import static io.fd.vpp.jvpp.acl.test.AclTestData.FIRST_RULE_SRC_ICMP_TYPE_END; -import static io.fd.vpp.jvpp.acl.test.AclTestData.FIRST_RULE_SRC_ICMP_TYPE_START; -import static io.fd.vpp.jvpp.acl.test.AclTestData.ICMP_PROTOCOL; -import static io.fd.vpp.jvpp.acl.test.AclTestData.SECOND_RULE_ADDRESS_2_AS_ARRAY; -import static io.fd.vpp.jvpp.acl.test.AclTestData.SECOND_RULE_ADDRESS_AS_ARRAY; -import static io.fd.vpp.jvpp.acl.test.AclTestData.SECOND_RULE_DST_PORT_RANGE_END; -import static io.fd.vpp.jvpp.acl.test.AclTestData.SECOND_RULE_DST_PORT_RANGE_START; -import static io.fd.vpp.jvpp.acl.test.AclTestData.SECOND_RULE_MAC; -import static io.fd.vpp.jvpp.acl.test.AclTestData.SECOND_RULE_MAC_MASK; -import static io.fd.vpp.jvpp.acl.test.AclTestData.SECOND_RULE_PREFIX; -import static io.fd.vpp.jvpp.acl.test.AclTestData.SECOND_RULE_PREFIX_2; -import static io.fd.vpp.jvpp.acl.test.AclTestData.SECOND_RULE_SRC_PORT_RANGE_END; -import static io.fd.vpp.jvpp.acl.test.AclTestData.SECOND_RULE_SRC_PORT_RANGE_START; -import static io.fd.vpp.jvpp.acl.test.AclTestData.UDP_PROTOCOL; - -import io.fd.vpp.jvpp.acl.dto.AclDetails; -import io.fd.vpp.jvpp.acl.dto.AclInterfaceListDetails; -import io.fd.vpp.jvpp.acl.dto.MacipAclDetails; -import io.fd.vpp.jvpp.acl.types.AclRule; -import io.fd.vpp.jvpp.acl.types.MacipAclRule; -import java.util.Arrays; - -class AclExpectedDumpData { - - static void verifyMacIpDump(final MacipAclDetails macipAclDetails) { - // asserting data create by previous call - assertEquals(0, macipAclDetails.aclIndex); - assertEquals(2, macipAclDetails.count); - - final MacipAclRule currentIpv4Rule = macipAclDetails.r[0]; - final MacipAclRule currentIpv6Rule = macipAclDetails.r[1]; - - // Comparing one property at the time to better pointer if something is wrong - //Ipv4 rule - assertEquals(0, currentIpv4Rule.isIpv6); - assertEquals(1, currentIpv4Rule.isPermit); - - // cutting expected ipv4 to 4 bytes,vpp sends it as 16 always - assertArrays(FIRST_RULE_ADDRESS_AS_ARRAY, Arrays.copyOfRange(currentIpv4Rule.srcIpAddr, 0, 4)); - assertEquals(FIRST_RULE_PREFIX, currentIpv4Rule.srcIpPrefixLen); - assertArrays(FIRST_RULE_MAC, currentIpv4Rule.srcMac); - assertArrays(FIRST_RULE_MAC_MASK, currentIpv4Rule.srcMacMask); - - //Ipv6 rule - assertEquals(1, currentIpv6Rule.isIpv6); - assertEquals(0, currentIpv6Rule.isPermit); - assertArrays(SECOND_RULE_ADDRESS_AS_ARRAY, currentIpv6Rule.srcIpAddr); - assertEquals(SECOND_RULE_PREFIX, currentIpv6Rule.srcIpPrefixLen); - assertArrays(SECOND_RULE_MAC, currentIpv6Rule.srcMac); - assertArrays(SECOND_RULE_MAC_MASK, currentIpv6Rule.srcMacMask); - } - - static void verifyAclDump(final AclDetails aclDetails) { - assertEquals(0, aclDetails.aclIndex); - assertEquals(2, aclDetails.count); - - final AclRule currentIpv4Rule = aclDetails.r[0]; - final AclRule currentIpv6Rule = aclDetails.r[1]; - - // Comparing one property at the time to better pointer if something is wrong - //Ipv4 rule - assertEquals(0, currentIpv4Rule.isIpv6); - assertEquals(1, currentIpv4Rule.isPermit); - - // cutting expected ipv4 to 4 bytes,vpp sends it as 16 always - assertArrays(FIRST_RULE_ADDRESS_AS_ARRAY, Arrays.copyOfRange(currentIpv4Rule.srcIpAddr, 0, 4)); - assertEquals(FIRST_RULE_PREFIX, currentIpv4Rule.srcIpPrefixLen); - assertArrays(FIRST_RULE_ADDRESS_2_AS_ARRAY, Arrays.copyOfRange(currentIpv4Rule.dstIpAddr, 0, 4)); - assertEquals(FIRST_RULE_PREFIX_2, currentIpv4Rule.dstIpPrefixLen); - - assertEquals(ICMP_PROTOCOL, currentIpv4Rule.proto); - assertEquals(FIRST_RULE_SRC_ICMP_TYPE_START, currentIpv4Rule.srcportOrIcmptypeFirst); - assertEquals(FIRST_RULE_SRC_ICMP_TYPE_END, currentIpv4Rule.srcportOrIcmptypeLast); - assertEquals(FIRST_RULE_DST_ICMP_TYPE_START, currentIpv4Rule.dstportOrIcmpcodeFirst); - assertEquals(FIRST_RULE_DST_ICMP_TYPE_END, currentIpv4Rule.dstportOrIcmpcodeLast); - - assertArrays(SECOND_RULE_ADDRESS_AS_ARRAY, currentIpv6Rule.srcIpAddr); - assertEquals(SECOND_RULE_PREFIX, currentIpv6Rule.srcIpPrefixLen); - assertArrays(SECOND_RULE_ADDRESS_2_AS_ARRAY, currentIpv6Rule.dstIpAddr); - assertEquals(SECOND_RULE_PREFIX_2, currentIpv6Rule.dstIpPrefixLen); - - assertEquals(UDP_PROTOCOL, currentIpv6Rule.proto); - assertEquals(SECOND_RULE_SRC_PORT_RANGE_START, currentIpv6Rule.srcportOrIcmptypeFirst); - assertEquals(SECOND_RULE_SRC_PORT_RANGE_END, currentIpv6Rule.srcportOrIcmptypeLast); - assertEquals(SECOND_RULE_DST_PORT_RANGE_START, currentIpv6Rule.dstportOrIcmpcodeFirst); - assertEquals(SECOND_RULE_DST_PORT_RANGE_END, currentIpv6Rule.dstportOrIcmpcodeLast); - } - - static void verifyAclInterfaceList(final AclInterfaceListDetails aclInterfaceListDetails) { - assertEquals(1, aclInterfaceListDetails.count); - assertEquals(1, aclInterfaceListDetails.acls[0]); - assertEquals(0, aclInterfaceListDetails.nInput); - assertEquals(0, aclInterfaceListDetails.swIfIndex); - } - - private static void assertArrays(final byte[] expected, final byte[] actual) { - if (!Arrays.equals(expected, actual)) { - throw new IllegalArgumentException( - String.format("Expected[%s]/Actual[%s]", Arrays.toString(expected), Arrays.toString(actual))); - } - } - - private static void assertEquals(final int expected, final int actual) { - if (expected != actual) { - throw new IllegalArgumentException(String.format("Expected[%s]/Actual[%s]", expected, actual)); - } - } -} diff --git a/plugins/acl-plugin/acl/jvpp/io/fd/vpp/jvpp/acl/test/AclTestData.java b/plugins/acl-plugin/acl/jvpp/io/fd/vpp/jvpp/acl/test/AclTestData.java deleted file mode 100644 index 5d228eea..00000000 --- a/plugins/acl-plugin/acl/jvpp/io/fd/vpp/jvpp/acl/test/AclTestData.java +++ /dev/null @@ -1,101 +0,0 @@ -/* - * Copyright (c) 2016 Cisco and/or its affiliates. - * - * Licensed under the Apache License, Version 2.0 (the "License"); - * you may not use this file except in compliance with the License. - * You may obtain a copy of the License at: - * - * http://www.apache.org/licenses/LICENSE-2.0 - * - * Unless required by applicable law or agreed to in writing, software - * distributed under the License is distributed on an "AS IS" BASIS, - * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. - * See the License for the specific language governing permissions and - * limitations under the License. - */ - -package io.fd.vpp.jvpp.acl.test; - - -import io.fd.vpp.jvpp.acl.types.AclRule; -import io.fd.vpp.jvpp.acl.types.MacipAclRule; - -class AclTestData { - - static final byte[] FIRST_RULE_ADDRESS_AS_ARRAY = {-64, -88, 2, 1}; - static final byte[] FIRST_RULE_ADDRESS_2_AS_ARRAY = {-64, -88, 2, 3}; - static final byte[] SECOND_RULE_ADDRESS_AS_ARRAY = - {32, 1, 13, -72, 10, 11, 18, -16, 0, 0, 0, 0, 0, 0, 0, 1}; - static final byte[] SECOND_RULE_ADDRESS_2_AS_ARRAY = - {32, 1, 13, -72, 10, 11, 18, -16, 0, 0, 0, 0, 0, 0, 0, 1}; - static final byte[] FIRST_RULE_MAC = {11, 11, 11, 11, 11, 11}; - static final byte[] FIRST_RULE_MAC_MASK = {0, 0, 0, 0, 0, 0}; - static final byte[] SECOND_RULE_MAC = {11, 12, 11, 11, 12, 11}; - static final byte[] SECOND_RULE_MAC_MASK = {(byte) 170, 0, 0, 0, 0, 0}; - static final int FIRST_RULE_PREFIX = 32; - static final int FIRST_RULE_PREFIX_2 = 24; - static final int SECOND_RULE_PREFIX = 64; - static final int SECOND_RULE_PREFIX_2 = 62; - static final int FIRST_RULE_DST_ICMP_TYPE_START = 0; - static final int FIRST_RULE_DST_ICMP_TYPE_END = 8; - static final int FIRST_RULE_SRC_ICMP_TYPE_START = 1; - static final int FIRST_RULE_SRC_ICMP_TYPE_END = 7; - static final int ICMP_PROTOCOL = 1; - static final int SECOND_RULE_DST_PORT_RANGE_START = 2000; - static final int SECOND_RULE_DST_PORT_RANGE_END = 6000; - static final int SECOND_RULE_SRC_PORT_RANGE_START = 400; - static final int SECOND_RULE_SRC_PORT_RANGE_END = 2047; - static final int UDP_PROTOCOL = 17; - - - static MacipAclRule[] createMacipRules() { - MacipAclRule ruleOne = new MacipAclRule(); - ruleOne.isIpv6 = 0; - ruleOne.isPermit = 1; - ruleOne.srcIpAddr = FIRST_RULE_ADDRESS_AS_ARRAY; - ruleOne.srcIpPrefixLen = FIRST_RULE_PREFIX; - ruleOne.srcMac = FIRST_RULE_MAC; - ruleOne.srcMacMask = FIRST_RULE_MAC_MASK;// no mask - - MacipAclRule ruleTwo = new MacipAclRule(); - ruleTwo.isIpv6 = 1; - ruleTwo.isPermit = 0; - ruleTwo.srcIpAddr = SECOND_RULE_ADDRESS_AS_ARRAY; - ruleTwo.srcIpPrefixLen = SECOND_RULE_PREFIX; - ruleTwo.srcMac = SECOND_RULE_MAC; - ruleTwo.srcMacMask = SECOND_RULE_MAC_MASK; - - return new MacipAclRule[]{ruleOne, ruleTwo}; - } - - static AclRule[] createAclRules() { - AclRule ruleOne = new AclRule(); - - ruleOne.isIpv6 = 0; - ruleOne.isPermit = 1; - ruleOne.srcIpAddr = FIRST_RULE_ADDRESS_AS_ARRAY; - ruleOne.srcIpPrefixLen = FIRST_RULE_PREFIX; - ruleOne.dstIpAddr = FIRST_RULE_ADDRESS_2_AS_ARRAY; - ruleOne.dstIpPrefixLen = FIRST_RULE_PREFIX_2; - ruleOne.dstportOrIcmpcodeFirst = FIRST_RULE_DST_ICMP_TYPE_START; - ruleOne.dstportOrIcmpcodeLast = FIRST_RULE_DST_ICMP_TYPE_END; - ruleOne.srcportOrIcmptypeFirst = FIRST_RULE_SRC_ICMP_TYPE_START; - ruleOne.srcportOrIcmptypeLast = FIRST_RULE_SRC_ICMP_TYPE_END; - ruleOne.proto = ICMP_PROTOCOL; //ICMP - - AclRule ruleTwo = new AclRule(); - ruleTwo.isIpv6 = 1; - ruleTwo.isPermit = 0; - ruleTwo.srcIpAddr = SECOND_RULE_ADDRESS_AS_ARRAY; - ruleTwo.srcIpPrefixLen = SECOND_RULE_PREFIX; - ruleTwo.dstIpAddr = SECOND_RULE_ADDRESS_2_AS_ARRAY; - ruleTwo.dstIpPrefixLen = SECOND_RULE_PREFIX_2; - ruleTwo.dstportOrIcmpcodeFirst = SECOND_RULE_DST_PORT_RANGE_START; - ruleTwo.dstportOrIcmpcodeLast = SECOND_RULE_DST_PORT_RANGE_END; - ruleTwo.srcportOrIcmptypeFirst = SECOND_RULE_SRC_PORT_RANGE_START; - ruleTwo.srcportOrIcmptypeLast = SECOND_RULE_SRC_PORT_RANGE_END; - ruleTwo.proto = UDP_PROTOCOL; //UDP - - return new AclRule[]{ruleOne, ruleTwo}; - } -} diff --git a/plugins/acl-plugin/acl/jvpp/io/fd/vpp/jvpp/acl/test/AclTestRequests.java b/plugins/acl-plugin/acl/jvpp/io/fd/vpp/jvpp/acl/test/AclTestRequests.java deleted file mode 100644 index b580ee8c..00000000 --- a/plugins/acl-plugin/acl/jvpp/io/fd/vpp/jvpp/acl/test/AclTestRequests.java +++ /dev/null @@ -1,141 +0,0 @@ -/* - * Copyright (c) 2016 Cisco and/or its affiliates. - * - * Licensed under the Apache License, Version 2.0 (the "License"); - * you may not use this file except in compliance with the License. - * You may obtain a copy of the License at: - * - * http://www.apache.org/licenses/LICENSE-2.0 - * - * Unless required by applicable law or agreed to in writing, software - * distributed under the License is distributed on an "AS IS" BASIS, - * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. - * See the License for the specific language governing permissions and - * limitations under the License. - */ - -package io.fd.vpp.jvpp.acl.test; - -import static io.fd.vpp.jvpp.acl.test.AclTestData.createAclRules; -import static io.fd.vpp.jvpp.acl.test.AclTestData.createMacipRules; - -import io.fd.vpp.jvpp.VppInvocationException; -import io.fd.vpp.jvpp.acl.dto.AclAddReplace; -import io.fd.vpp.jvpp.acl.dto.AclAddReplaceReply; -import io.fd.vpp.jvpp.acl.dto.AclDel; -import io.fd.vpp.jvpp.acl.dto.AclDelReply; -import io.fd.vpp.jvpp.acl.dto.AclDetailsReplyDump; -import io.fd.vpp.jvpp.acl.dto.AclDump; -import io.fd.vpp.jvpp.acl.dto.AclInterfaceListDetailsReplyDump; -import io.fd.vpp.jvpp.acl.dto.AclInterfaceListDump; -import io.fd.vpp.jvpp.acl.dto.AclInterfaceSetAclList; -import io.fd.vpp.jvpp.acl.dto.AclInterfaceSetAclListReply; -import io.fd.vpp.jvpp.acl.dto.MacipAclAdd; -import io.fd.vpp.jvpp.acl.dto.MacipAclAddReply; -import io.fd.vpp.jvpp.acl.dto.MacipAclDel; -import io.fd.vpp.jvpp.acl.dto.MacipAclDelReply; -import io.fd.vpp.jvpp.acl.dto.MacipAclDetailsReplyDump; -import io.fd.vpp.jvpp.acl.dto.MacipAclDump; -import io.fd.vpp.jvpp.acl.future.FutureJVppAclFacade; -import java.util.concurrent.ExecutionException; - -class AclTestRequests { - - static MacipAclDetailsReplyDump sendMacIpDumpRequest(final FutureJVppAclFacade jvpp) - throws ExecutionException, InterruptedException { - System.out.println("Sending MacipAclDump request..."); - MacipAclDetailsReplyDump dump = jvpp.macipAclDump(new MacipAclDump()).toCompletableFuture().get(); - System.out.println("MacipAclDump returned"); - return dump; - } - - static void sendMacIpAddRequest(final FutureJVppAclFacade jvpp) throws InterruptedException, ExecutionException { - final MacipAclAdd request = createMacIpAddRequest(); - System.out.printf("Sending MacipAclAdd request %s%n", request.toString()); - final MacipAclAddReply reply = jvpp.macipAclAdd(createMacIpAddRequest()).toCompletableFuture().get(); - System.out.printf("MacipAclAdd send result = %s%n", reply); - } - - static void sendMacIpDelRequest(final FutureJVppAclFacade jvpp) throws InterruptedException, ExecutionException { - final MacipAclDel request = new MacipAclDel(); - request.aclIndex = 0; - System.out.printf("Sending MacipAclDel request %s%n", request.toString()); - final MacipAclDelReply reply = jvpp.macipAclDel(request).toCompletableFuture().get(); - System.out.printf("MacipAclDel send result = %s%n", reply); - } - - static void sendAclAddRequest(final FutureJVppAclFacade jvpp) throws InterruptedException, ExecutionException { - final AclAddReplace request = createAclAddRequest(); - System.out.printf("Sending AclAddReplace request %s%n", request.toString()); - final AclAddReplaceReply reply = jvpp.aclAddReplace(request).toCompletableFuture().get(); - System.out.printf("AclAddReplace send result = %s%n", reply); - } - - static AclDetailsReplyDump sendAclDumpRequest(final FutureJVppAclFacade jvpp) - throws InterruptedException, VppInvocationException, ExecutionException { - System.out.println("Sending AclDump request..."); - final AclDetailsReplyDump dump = jvpp.aclDump(new AclDump()).toCompletableFuture().get(); - System.out.printf("AclDump send result = %s%n", dump); - return dump; - } - - static void sendAclDelRequest(final FutureJVppAclFacade jvpp) throws InterruptedException, ExecutionException { - final AclDel request = new AclDel(); - request.aclIndex = 0; - System.out.printf("Sending AclDel request %s%n", request.toString()); - final AclDelReply reply = jvpp.aclDel(request).toCompletableFuture().get(); - System.out.printf("AclDel send result = %s%n", reply); - } - - static AclInterfaceListDetailsReplyDump sendAclInterfaceListDumpRequest(final FutureJVppAclFacade jvpp) - throws InterruptedException, ExecutionException { - final AclInterfaceListDump request = new AclInterfaceListDump(); - request.swIfIndex = 0; - System.out.printf("Sending AclInterfaceListDump request %s%n", request.toString()); - final AclInterfaceListDetailsReplyDump dump = jvpp.aclInterfaceListDump(request).toCompletableFuture().get(); - System.out.printf("AclInterfaceListDump send result = %s%n", dump); - return dump; - } - - static void sendAclInterfaceSetAclList(final FutureJVppAclFacade jvpp) - throws InterruptedException, ExecutionException { - final AclInterfaceSetAclList request = new AclInterfaceSetAclList(); - request.count = 1; - request.acls = new int[]{1}; - request.swIfIndex = 0; - request.nInput = 0; - System.out.printf("Sending AclInterfaceSetAclList request %s%n", request.toString()); - final AclInterfaceSetAclListReply reply = jvpp.aclInterfaceSetAclList(request).toCompletableFuture().get(); - System.out.printf("AclInterfaceSetAclList send result = %s%n", reply); - } - - static void sendAclInterfaceDeleteList(final FutureJVppAclFacade jvpp) - throws InterruptedException, ExecutionException { - // uses same api but sets list to empty - final AclInterfaceSetAclList request = new AclInterfaceSetAclList(); - request.count = 0; - request.acls = new int[]{}; - request.swIfIndex = 0; - request.nInput = 0; - System.out.printf("Sending AclInterfaceSetAclList(Delete) request %s%n", request.toString()); - final AclInterfaceSetAclListReply reply = jvpp.aclInterfaceSetAclList(request).toCompletableFuture().get(); - System.out.printf("AclInterfaceSetAclList(Delete) send result = %s%n", reply); - } - - private static MacipAclAdd createMacIpAddRequest() { - MacipAclAdd request = new MacipAclAdd(); - - request.count = 2; - request.r = createMacipRules(); - return request; - } - - private static AclAddReplace createAclAddRequest() { - AclAddReplace request = new AclAddReplace(); - - request.aclIndex = -1;// to define new one - request.count = 2; - request.r = createAclRules(); - return request; - } -} diff --git a/plugins/acl-plugin/acl/jvpp/io/fd/vpp/jvpp/acl/test/FutureApiTest.java b/plugins/acl-plugin/acl/jvpp/io/fd/vpp/jvpp/acl/test/FutureApiTest.java deleted file mode 100644 index 94490193..00000000 --- a/plugins/acl-plugin/acl/jvpp/io/fd/vpp/jvpp/acl/test/FutureApiTest.java +++ /dev/null @@ -1,68 +0,0 @@ -/* - * Copyright (c) 2016 Cisco and/or its affiliates. - * - * Licensed under the Apache License, Version 2.0 (the "License"); - * you may not use this file except in compliance with the License. - * You may obtain a copy of the License at: - * - * http://www.apache.org/licenses/LICENSE-2.0 - * - * Unless required by applicable law or agreed to in writing, software - * distributed under the License is distributed on an "AS IS" BASIS, - * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. - * See the License for the specific language governing permissions and - * limitations under the License. - */ - -package io.fd.vpp.jvpp.acl.test; - -import static io.fd.vpp.jvpp.acl.test.AclExpectedDumpData.verifyAclDump; -import static io.fd.vpp.jvpp.acl.test.AclExpectedDumpData.verifyAclInterfaceList; -import static io.fd.vpp.jvpp.acl.test.AclExpectedDumpData.verifyMacIpDump; -import static io.fd.vpp.jvpp.acl.test.AclTestRequests.sendAclAddRequest; -import static io.fd.vpp.jvpp.acl.test.AclTestRequests.sendAclDelRequest; -import static io.fd.vpp.jvpp.acl.test.AclTestRequests.sendAclDumpRequest; -import static io.fd.vpp.jvpp.acl.test.AclTestRequests.sendAclInterfaceDeleteList; -import static io.fd.vpp.jvpp.acl.test.AclTestRequests.sendAclInterfaceListDumpRequest; -import static io.fd.vpp.jvpp.acl.test.AclTestRequests.sendAclInterfaceSetAclList; -import static io.fd.vpp.jvpp.acl.test.AclTestRequests.sendMacIpAddRequest; -import static io.fd.vpp.jvpp.acl.test.AclTestRequests.sendMacIpDelRequest; -import static io.fd.vpp.jvpp.acl.test.AclTestRequests.sendMacIpDumpRequest; - -import io.fd.vpp.jvpp.JVppRegistry; -import io.fd.vpp.jvpp.JVppRegistryImpl; -import io.fd.vpp.jvpp.acl.JVppAclImpl; -import io.fd.vpp.jvpp.acl.future.FutureJVppAclFacade; - -public class FutureApiTest { - - public static void main(String[] args) throws Exception { - testCallbackApi(); - } - - private static void testCallbackApi() throws Exception { - System.out.println("Testing Java callback API for acl plugin"); - try (final JVppRegistry registry = new JVppRegistryImpl("macipAclAddTest"); - final FutureJVppAclFacade jvpp = new FutureJVppAclFacade(registry, new JVppAclImpl())) { - - // adds,dump and verifies Mac-Ip acl - sendMacIpAddRequest(jvpp); - verifyMacIpDump(sendMacIpDumpRequest(jvpp).macipAclDetails.get(0)); - - // adds,dumps and verifies Acl acl - sendAclAddRequest(jvpp); - verifyAclDump(sendAclDumpRequest(jvpp).aclDetails.get(0)); - - // adds,dumps and verifies Interface for acl - sendAclInterfaceSetAclList(jvpp); - verifyAclInterfaceList(sendAclInterfaceListDumpRequest(jvpp).aclInterfaceListDetails.get(0)); - - // deletes all created data - sendAclInterfaceDeleteList(jvpp); - sendAclDelRequest(jvpp); - sendMacIpDelRequest(jvpp); - - System.out.println("Disconnecting..."); - } - } -} diff --git a/plugins/acl-plugin/acl/jvpp/io/fd/vpp/jvpp/acl/test/Readme.txt b/plugins/acl-plugin/acl/jvpp/io/fd/vpp/jvpp/acl/test/Readme.txt deleted file mode 100644 index f68e7aba..00000000 --- a/plugins/acl-plugin/acl/jvpp/io/fd/vpp/jvpp/acl/test/Readme.txt +++ /dev/null @@ -1 +0,0 @@ -sudo java -cp build-vpp-native/vpp-api/java/jvpp-registry-17.01.jar:build-vpp-native/plugins/acl-plugin/jvpp-acl-1.0.jar io.fd.vpp.jvpp.acl.test.FutureApiTest diff --git a/plugins/acl-plugin/acl/jvpp_acl.c b/plugins/acl-plugin/acl/jvpp_acl.c deleted file mode 100644 index 0af53bc9..00000000 --- a/plugins/acl-plugin/acl/jvpp_acl.c +++ /dev/null @@ -1,124 +0,0 @@ -/* - * Copyright (c) 2016 Cisco and/or its affiliates. - * Licensed under the Apache License, Version 2.0 (the "License"); - * you may not use this file except in compliance with the License. - * You may obtain a copy of the License at: - * - * http://www.apache.org/licenses/LICENSE-2.0 - * - * Unless required by applicable law or agreed to in writing, software - * distributed under the License is distributed on an "AS IS" BASIS, - * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. - * See the License for the specific language governing permissions and - * limitations under the License. - */ - -#include - -#include -#define vl_typedefs /* define message structures */ -#include -#undef vl_typedefs - -#define vl_endianfun -#include -#undef vl_endianfun - -#define vl_print(handle, ...) -#define vl_printfun -#include -#undef vl_printfun - -/* Get the API version number */ -#define vl_api_version(n,v) static u32 api_version=(v); -#include -#undef vl_api_version - -#include -#include -#include - -#if VPPJNI_DEBUG == 1 - #define DEBUG_LOG(...) clib_warning(__VA_ARGS__) -#else - #define DEBUG_LOG(...) -#endif - -#include - -#include "acl/jvpp/io_fd_vpp_jvpp_acl_JVppAclImpl.h" -#include "jvpp_acl.h" -#include "acl/jvpp/jvpp_acl_gen.h" - -/* - * Class: io_fd_vpp_jvpp_acl_JVppaclImpl - * Method: init0 - * Signature: (JI)V - */ -JNIEXPORT void JNICALL Java_io_fd_vpp_jvpp_acl_JVppAclImpl_init0 - (JNIEnv *env, jclass clazz, jobject callback, jlong queue_address, jint my_client_index) { - acl_main_t * plugin_main = &acl_main; - u8 * name; - clib_warning ("Java_io_fd_vpp_jvpp_acl_JVppAclImpl_init0"); - - plugin_main->my_client_index = my_client_index; - plugin_main->vl_input_queue = (unix_shared_memory_queue_t *)queue_address; - - name = format (0, "acl_%08x%c", api_version, 0); - plugin_main->msg_id_base = vl_client_get_first_plugin_msg_id ((char *) name); - - if (plugin_main->msg_id_base == (u16) ~0) { - jclass exClass = (*env)->FindClass(env, "java/lang/IllegalStateException"); - (*env)->ThrowNew(env, exClass, "acl plugin is not loaded in VPP"); - } else { - plugin_main->callbackObject = (*env)->NewGlobalRef(env, callback); - plugin_main->callbackClass = (jclass)(*env)->NewGlobalRef(env, (*env)->GetObjectClass(env, callback)); - - #define _(N,n) \ - vl_msg_api_set_handlers(VL_API_##N + plugin_main->msg_id_base, #n, \ - vl_api_##n##_t_handler, \ - vl_noop_handler, \ - vl_api_##n##_t_endian, \ - vl_api_##n##_t_print, \ - sizeof(vl_api_##n##_t), 1); - foreach_api_reply_handler; - #undef _ - } -} - -JNIEXPORT void JNICALL Java_io_fd_vpp_jvpp_acl_JVppAclImpl_close0 -(JNIEnv *env, jclass clazz) { - acl_main_t * plugin_main = &acl_main; - - // cleanup: - (*env)->DeleteGlobalRef(env, plugin_main->callbackClass); - (*env)->DeleteGlobalRef(env, plugin_main->callbackObject); - - plugin_main->callbackClass = NULL; - plugin_main->callbackObject = NULL; -} - -/* Attach thread to JVM and cache class references when initiating JVPP ACL */ -jint JNI_OnLoad(JavaVM *vm, void *reserved) { - JNIEnv* env; - - if ((*vm)->GetEnv(vm, (void**) &env, JNI_VERSION_1_8) != JNI_OK) { - return JNI_EVERSION; - } - - if (cache_class_references(env) != 0) { - clib_warning ("Failed to cache class references\n"); - return JNI_ERR; - } - - return JNI_VERSION_1_8; -} - -/* Clean up cached references when disposing JVPP ACL */ -void JNI_OnUnload(JavaVM *vm, void *reserved) { - JNIEnv* env; - if ((*vm)->GetEnv(vm, (void**) &env, JNI_VERSION_1_8) != JNI_OK) { - return; - } - delete_class_references(env); -} diff --git a/plugins/acl-plugin/acl/jvpp_acl.h b/plugins/acl-plugin/acl/jvpp_acl.h deleted file mode 100644 index 2b73d672..00000000 --- a/plugins/acl-plugin/acl/jvpp_acl.h +++ /dev/null @@ -1,45 +0,0 @@ -/* - * Copyright (c) 2016 Cisco and/or its affiliates. - * Licensed under the Apache License, Version 2.0 (the "License"); - * you may not use this file except in compliance with the License. - * You may obtain a copy of the License at: - * - * http://www.apache.org/licenses/LICENSE-2.0 - * - * Unless required by applicable law or agreed to in writing, software - * distributed under the License is distributed on an "AS IS" BASIS, - * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. - * See the License for the specific language governing permissions and - * limitations under the License. - */ -#ifndef __included_jvpp_acl_h__ -#define __included_jvpp_acl_h__ - -#include -#include -#include -#include -#include -#include - -/* Global state for JVPP-acl */ -typedef struct { - /* Base message index for the acl plugin */ - u16 msg_id_base; - - /* Pointer to shared memory queue */ - unix_shared_memory_queue_t * vl_input_queue; - - /* VPP api client index */ - u32 my_client_index; - - /* Callback object and class references enabling asynchronous Java calls */ - jobject callbackObject; - jclass callbackClass; - -} acl_main_t; - -acl_main_t acl_main __attribute__((aligned (64))); - - -#endif /* __included_jvpp_acl_h__ */ diff --git a/plugins/acl-plugin/acl/l2sess.c b/plugins/acl-plugin/acl/l2sess.c deleted file mode 100644 index cc9bde44..00000000 --- a/plugins/acl-plugin/acl/l2sess.c +++ /dev/null @@ -1,243 +0,0 @@ -/* - * Copyright (c) 2016 Cisco and/or its affiliates. - * Licensed under the Apache License, Version 2.0 (the "License"); - * you may not use this file except in compliance with the License. - * You may obtain a copy of the License at: - * - * http://www.apache.org/licenses/LICENSE-2.0 - * - * Unless required by applicable law or agreed to in writing, software - * distributed under the License is distributed on an "AS IS" BASIS, - * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. - * See the License for the specific language governing permissions and - * limitations under the License. - */ -/* - *------------------------------------------------------------------ - * l2sess.c - simple MAC-swap API / debug CLI handling - *------------------------------------------------------------------ - */ - -#include -#include -#include - -#include -#include -#include -#include - -#include -#include - -void -l2sess_vlib_plugin_register (vlib_main_t * vm, void* hh, - int from_early_init) -{ - l2sess_main_t *sm = &l2sess_main; - vnet_plugin_handoff_t * h = hh; - memset (sm, 0, sizeof (*sm)); - - sm->vlib_main = vm; - sm->vnet_main = h->vnet_main; - sm->ethernet_main = h->ethernet_main; -} - -void -l2sess_init_next_features_input (vlib_main_t * vm, l2sess_main_t * sm) -{ -#define _(node_name, node_var, is_out, is_ip6, is_track) \ - if (!is_out) feat_bitmap_init_next_nodes(vm, node_var.index, L2INPUT_N_FEAT, l2input_get_feat_names (), sm->node_var ## _input_next_node_index); - foreach_l2sess_node -#undef _ -} - -void -l2sess_add_our_next_nodes (vlib_main_t * vm, l2sess_main_t * sm, - u8 * prev_node_name, int add_output_nodes) -{ - vlib_node_t *n; - n = vlib_get_node_by_name (vm, prev_node_name); -#define _(node_name, node_var, is_out, is_ip6, is_track) \ - if (is_out == add_output_nodes) { \ - u32 idx = vlib_node_add_next_with_slot(vm, n->index, node_var.index, ~0); \ - if (is_track) { \ - sm->next_slot_track_node_by_is_ip6_is_out[is_ip6][is_out] = idx; \ - } \ - } - foreach_l2sess_node -#undef _ -} - -void -l2sess_setup_nodes (void) -{ - vlib_main_t *vm = vlib_get_main (); - l2sess_main_t *sm = &l2sess_main; - - l2sess_init_next_features_input (vm, sm); - - l2sess_add_our_next_nodes (vm, sm, (u8 *) "l2-input-classify", 0); - l2sess_add_our_next_nodes (vm, sm, (u8 *) "l2-output-classify", 1); - -} - -static char * -get_l4_proto_str (int is_ip6, uint8_t l4_proto) -{ - switch (l4_proto) - { - case 6: - return "tcp"; - case 17: - return "udp"; - case 1: - return "icmp"; - case 58: - return "icmp6"; - default: - return ""; - } -} - -static clib_error_t * -l2sess_show_command_fn (vlib_main_t * vm, - unformat_input_t * input, vlib_cli_command_t * cmd) -{ - l2sess_main_t *sm = &l2sess_main; - clib_time_t *ct = &vm->clib_time; - l2s_session_t *s; - u64 now = clib_cpu_time_now (); - - vlib_cli_output (vm, "Timing wheel info: \n%U", format_timing_wheel, - &sm->timing_wheel, 255); - - pool_foreach (s, sm->sessions, ( - { - f64 ctime = - (now - - s->create_time) * ct->seconds_per_clock; - f64 atime0 = - (now - - s->side[0].active_time) * - ct->seconds_per_clock; - f64 atime1 = - (now - - s->side[1].active_time) * - ct->seconds_per_clock; -/* - f64 ctime = (s->create_time - vm->cpu_time_main_loop_start) * ct->seconds_per_clock; - f64 atime0 = (s->side[0].active_time - vm->cpu_time_main_loop_start) * ct->seconds_per_clock; - f64 atime1 = (s->side[1].active_time - vm->cpu_time_main_loop_start) * ct->seconds_per_clock; -*/ - u8 * out0 = - format (0, - "%5d: create time: %U pkts/bytes/active time: [ %ld %ld %U : %ld %ld %U ]\n", - (s - sm->sessions), - format_time_interval, "h:m:s:u", - ctime, s->side[0].n_packets, - s->side[0].n_bytes, - format_time_interval, "h:m:s:u", - atime0, s->side[1].n_packets, - s->side[1].n_bytes, - format_time_interval, "h:m:s:u", - atime1); u8 * out1 = 0; - if (s->is_ip6) - { - out1 = - format (0, "%s %U :%u <-> %U :%u", - get_l4_proto_str (s->is_ip6, - s->l4_proto), - format_ip6_address, - &s->side[0].addr.ip6, - s->side[0].port, - format_ip6_address, - &s->side[1].addr.ip6, - s->side[1].port);} - else - { - out1 = - format (0, "%s %U :%u <-> %U :%u", - get_l4_proto_str (s->is_ip6, - s->l4_proto), - format_ip4_address, - &s->side[0].addr.ip4, - s->side[0].port, - format_ip4_address, - &s->side[1].addr.ip4, - s->side[1].port);} - vlib_cli_output (vm, "%s %s", out0, - out1); vec_free (out0); - vec_free (out1);} - )); - return 0; -} - -static clib_error_t * -l2sess_show_count_command_fn (vlib_main_t * vm, - unformat_input_t * input, - vlib_cli_command_t * cmd) -{ - l2sess_main_t *sm = &l2sess_main; - - vlib_cli_output (vm, "Timing wheel info: \n%U", format_timing_wheel, - &sm->timing_wheel, 255); - vlib_cli_output (vm, "session pool len: %d, pool elts: %d", - pool_len (sm->sessions), pool_elts (sm->sessions)); - vlib_cli_output (vm, - "attempted to delete sessions which were already free: %d", - sm->counter_attempted_delete_free_session); - return 0; -} - - -/* *INDENT-OFF* */ -VLIB_CLI_COMMAND (l2sess_show_command, static) = { - .path = "show l2sess", - .short_help = "show l2sess", - .function = l2sess_show_command_fn, -}; - -VLIB_CLI_COMMAND (l2sess_show_count_command, static) = { - .path = "show l2sess count", - .short_help = "show l2sess count", - .function = l2sess_show_count_command_fn, -}; -/* *INDENT-OFF* */ - -static inline u64 -time_sec_to_clock( clib_time_t *ct, f64 sec) -{ - return (u64)(((f64)sec)/ct->seconds_per_clock); -} - -static clib_error_t * l2sess_init (vlib_main_t * vm) -{ - l2sess_main_t * sm = &l2sess_main; - clib_error_t * error = 0; - u64 cpu_time_now = clib_cpu_time_now(); - - - clib_time_t *ct = &vm->clib_time; - sm->udp_session_idle_timeout = time_sec_to_clock(ct, UDP_SESSION_IDLE_TIMEOUT_SEC); - sm->tcp_session_idle_timeout = time_sec_to_clock(ct, TCP_SESSION_IDLE_TIMEOUT_SEC); - sm->tcp_session_transient_timeout = time_sec_to_clock(ct, TCP_SESSION_TRANSIENT_TIMEOUT_SEC); - - /* The min sched time of 10e-1 causes erroneous behavior... */ - sm->timing_wheel.min_sched_time = 10e-2; - sm->timing_wheel.max_sched_time = 3600.0*48.0; - timing_wheel_init (&sm->timing_wheel, cpu_time_now, vm->clib_time.clocks_per_second); - sm->timer_wheel_next_expiring_time = 0; - sm->timer_wheel_tick = time_sec_to_clock(ct, sm->timing_wheel.min_sched_time); - /* Pre-allocate expired nodes. */ - vec_alloc (sm->data_from_advancing_timing_wheel, 32); - - l2sess_setup_nodes(); - l2output_init_output_node_vec (&sm->output_next_nodes.output_node_index_vec); - - return error; -} - -VLIB_INIT_FUNCTION (l2sess_init); - - diff --git a/plugins/acl-plugin/acl/l2sess.h b/plugins/acl-plugin/acl/l2sess.h deleted file mode 100644 index db899917..00000000 --- a/plugins/acl-plugin/acl/l2sess.h +++ /dev/null @@ -1,150 +0,0 @@ -/* - * Copyright (c) 2016 Cisco and/or its affiliates. - * Licensed under the Apache License, Version 2.0 (the "License"); - * you may not use this file except in compliance with the License. - * You may obtain a copy of the License at: - * - * http://www.apache.org/licenses/LICENSE-2.0 - * - * Unless required by applicable law or agreed to in writing, software - * distributed under the License is distributed on an "AS IS" BASIS, - * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. - * See the License for the specific language governing permissions and - * limitations under the License. - */ -#ifndef __included_l2sess_h__ -#define __included_l2sess_h__ - -#include -#include -#include - -#include -#include -#include -#include - -#include -#include - -#define _(node_name, node_var, is_out, is_ip6, is_track) -#undef _ -#define foreach_l2sess_node \ - _("aclp-l2s-input-ip4-add", l2sess_in_ip4_add, 0, 0, 0) \ - _("aclp-l2s-input-ip6-add", l2sess_in_ip6_add, 0, 1, 0) \ - _("aclp-l2s-output-ip4-add", l2sess_out_ip4_add, 1, 0, 0) \ - _("aclp-l2s-output-ip6-add", l2sess_out_ip6_add, 1, 1, 0) \ - _("aclp-l2s-input-ip4-track", l2sess_in_ip4_track, 0, 0, 1) \ - _("aclp-l2s-input-ip6-track", l2sess_in_ip6_track, 0, 1, 1) \ - _("aclp-l2s-output-ip4-track",l2sess_out_ip4_track, 1, 0, 1) \ - _("aclp-l2s-output-ip6-track", l2sess_out_ip6_track, 1, 1, 1) - -#define _(node_name, node_var, is_out, is_ip6, is_track) \ - extern vlib_node_registration_t node_var; -foreach_l2sess_node -#undef _ - -#define TCP_FLAG_FIN 0x01 -#define TCP_FLAG_SYN 0x02 -#define TCP_FLAG_RST 0x04 -#define TCP_FLAG_PUSH 0x08 -#define TCP_FLAG_ACK 0x10 -#define TCP_FLAG_URG 0x20 -#define TCP_FLAG_ECE 0x40 -#define TCP_FLAG_CWR 0x80 -#define TCP_FLAGS_RSTFINACKSYN (TCP_FLAG_RST + TCP_FLAG_FIN + TCP_FLAG_SYN + TCP_FLAG_ACK) -#define TCP_FLAGS_ACKSYN (TCP_FLAG_SYN + TCP_FLAG_ACK) - -typedef struct { - ip46_address_t addr; - u64 active_time; - u64 n_packets; - u64 n_bytes; - u16 port; -} l2s_session_side_t; - -enum { - L2S_SESSION_SIDE_IN = 0, - L2S_SESSION_SIDE_OUT, - L2S_N_SESSION_SIDES -}; - -typedef struct { - u64 create_time; - l2s_session_side_t side[L2S_N_SESSION_SIDES]; - u8 l4_proto; - u8 is_ip6; - u16 tcp_flags_seen; /* u16 because of two sides */ -} l2s_session_t; - -#define PROD -#ifdef PROD -#define UDP_SESSION_IDLE_TIMEOUT_SEC 600 -#define TCP_SESSION_IDLE_TIMEOUT_SEC (3600*24) -#define TCP_SESSION_TRANSIENT_TIMEOUT_SEC 120 -#else -#define UDP_SESSION_IDLE_TIMEOUT_SEC 15 -#define TCP_SESSION_IDLE_TIMEOUT_SEC 15 -#define TCP_SESSION_TRANSIENT_TIMEOUT_SEC 5 -#endif - -typedef struct { - /* - * the next two fields are present for all nodes, but - * only one of them is used per node - depending - * on whether the node is an input or output one. - */ -#define _(node_name, node_var, is_out, is_ip6, is_track) \ - u32 node_var ## _input_next_node_index[32]; \ - l2_output_next_nodes_st node_var ## _next_nodes; -foreach_l2sess_node -#undef _ - l2_output_next_nodes_st output_next_nodes; - - /* Next indices of the tracker nodes */ - u32 next_slot_track_node_by_is_ip6_is_out[2][2]; - - /* - * Pairing of "forward" and "reverse" tables by table index. - * Each relationship has two entries - for one and the other table, - * so it is bidirectional. - */ - - u32 *fwd_to_rev_by_table_index; - - /* - * The vector of per-interface session pools - */ - - l2s_session_t *sessions; - - /* The session timeouts */ - u64 tcp_session_transient_timeout; - u64 tcp_session_idle_timeout; - u64 udp_session_idle_timeout; - - /* Timing wheel to time out the idle sessions */ - timing_wheel_t timing_wheel; - u32 *data_from_advancing_timing_wheel; - u64 timer_wheel_next_expiring_time; - u64 timer_wheel_tick; - - /* convenience */ - vlib_main_t * vlib_main; - vnet_main_t * vnet_main; - ethernet_main_t * ethernet_main; - - /* Counter(s) */ - u64 counter_attempted_delete_free_session; -} l2sess_main_t; - -l2sess_main_t l2sess_main; - -/* Just exposed for acl.c */ - -void -l2sess_vlib_plugin_register (vlib_main_t * vm, void * hh, - int from_early_init); - - -#endif /* __included_l2sess_h__ */ diff --git a/plugins/acl-plugin/acl/l2sess_node.c b/plugins/acl-plugin/acl/l2sess_node.c deleted file mode 100644 index 520e5929..00000000 --- a/plugins/acl-plugin/acl/l2sess_node.c +++ /dev/null @@ -1,816 +0,0 @@ -/* - * Copyright (c) 2016 Cisco and/or its affiliates. - * Licensed under the Apache License, Version 2.0 (the "License"); - * you may not use this file except in compliance with the License. - * You may obtain a copy of the License at: - * - * http://www.apache.org/licenses/LICENSE-2.0 - * - * Unless required by applicable law or agreed to in writing, software - * distributed under the License is distributed on an "AS IS" BASIS, - * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. - * See the License for the specific language governing permissions and - * limitations under the License. - */ -#include -#include -#include -#include -#include -#include -#include - - -typedef struct -{ - u32 next_index; - u32 sw_if_index; - u32 trace_flags; - u32 session_tables[2]; - u32 session_nexts[2]; - u8 l4_proto; -} l2sess_trace_t; - -/* packet trace format function */ - -#define _(node_name, node_var, is_out, is_ip6, is_track) \ -static u8 * format_## node_var ##_trace (u8 * s, va_list * args) \ -{ \ - CLIB_UNUSED (vlib_main_t * vm) = va_arg (*args, vlib_main_t *); \ - CLIB_UNUSED (vlib_node_t * node) = va_arg (*args, vlib_node_t *); \ - l2sess_trace_t * t = va_arg (*args, l2sess_trace_t *); \ - \ - s = format (s, node_name ": sw_if_index %d, next index %d trace_flags %08x L4 proto %d\n" \ - " tables [ %d, %d ] nexts [ %d, %d ]", \ - t->sw_if_index, t->next_index, t->trace_flags, t->l4_proto, \ - t->session_tables[0], t->session_tables[1], \ - t->session_nexts[0], t->session_nexts[1]); \ - return s; \ -} -foreach_l2sess_node -#undef _ -#define foreach_l2sess_error \ -_(SWAPPED, "Mac swap packets processed") - typedef enum -{ -#define _(sym,str) L2SESS_ERROR_##sym, - foreach_l2sess_error -#undef _ - L2SESS_N_ERROR, -} l2sess_error_t; - -static char *l2sess_error_strings[] = { -#define _(sym,string) string, - foreach_l2sess_error -#undef _ -}; - -typedef enum -{ - L2SESS_NEXT_DROP, - L2SESS_N_NEXT, -} l2sess_next_t; - -u8 -l2sess_get_l4_proto (vlib_buffer_t * b0, int node_is_ip6) -{ - u8 proto; - int proto_offset; - if (node_is_ip6) - { - proto_offset = 20; - } - else - { - proto_offset = 23; - } - proto = *((u8 *) vlib_buffer_get_current (b0) + proto_offset); - return proto; -} - - -u8 -l2sess_get_tcp_flags (vlib_buffer_t * b0, int node_is_ip6) -{ - u8 flags; - int flags_offset; - if (node_is_ip6) - { - flags_offset = 14 + 40 + 13; /* FIXME: no extension headers assumed */ - } - else - { - flags_offset = 14 + 20 + 13; - } - flags = *((u8 *) vlib_buffer_get_current (b0) + flags_offset); - return flags; -} - -static inline int -l4_tcp_or_udp (u8 proto) -{ - return ((proto == 6) || (proto == 17)); -} - -void -l2sess_get_session_tables (l2sess_main_t * sm, u32 sw_if_index, - int node_is_out, int node_is_ip6, u8 l4_proto, - u32 * session_tables) -{ -/* - * Based on the direction, l3 and l4 protocol, fill a u32[2] array: - * [0] is index for the "direct match" path, [1] is for "mirrored match". - * Store the indices of the tables to add the session to in session_tables[] - */ - l2_output_classify_main_t *l2om = &l2_output_classify_main; - l2_input_classify_main_t *l2im = &l2_input_classify_main; - - u32 output_table_index; - u32 input_table_index; - - if (!l4_tcp_or_udp (l4_proto)) - { - return; - } - - if (node_is_ip6) - { - vec_validate_init_empty (l2im-> - classify_table_index_by_sw_if_index - [L2_INPUT_CLASSIFY_TABLE_IP6], sw_if_index, - ~0); - input_table_index = - l2im-> - classify_table_index_by_sw_if_index[L2_INPUT_CLASSIFY_TABLE_IP6] - [sw_if_index]; - vec_validate_init_empty (l2om-> - classify_table_index_by_sw_if_index - [L2_OUTPUT_CLASSIFY_TABLE_IP6], sw_if_index, - ~0); - output_table_index = - l2om-> - classify_table_index_by_sw_if_index[L2_OUTPUT_CLASSIFY_TABLE_IP6] - [sw_if_index]; - } - else - { - vec_validate_init_empty (l2im-> - classify_table_index_by_sw_if_index - [L2_INPUT_CLASSIFY_TABLE_IP4], sw_if_index, - ~0); - input_table_index = - l2im-> - classify_table_index_by_sw_if_index[L2_INPUT_CLASSIFY_TABLE_IP4] - [sw_if_index]; - vec_validate_init_empty (l2om-> - classify_table_index_by_sw_if_index - [L2_OUTPUT_CLASSIFY_TABLE_IP4], sw_if_index, - ~0); - output_table_index = - l2om-> - classify_table_index_by_sw_if_index[L2_OUTPUT_CLASSIFY_TABLE_IP4] - [sw_if_index]; - } - - if (node_is_out) - { - session_tables[0] = output_table_index; - session_tables[1] = input_table_index; - } - else - { - session_tables[0] = input_table_index; - session_tables[1] = output_table_index; - } -} - -void -l2sess_get_session_nexts (l2sess_main_t * sm, u32 sw_if_index, - int node_is_out, int node_is_ip6, u8 l4_proto, - u32 * session_nexts) -{ -/* - * Based on the direction, l3 and l4 protocol, fill a u32[2] array: - * [0] is the index for the "direct match" path, [1] is for "mirrored match". - * Store the match_next_index in session_nexts[] for a new session entry which is being added to session tables. - */ - u32 input_node_index; - u32 output_node_index; - - if (!l4_tcp_or_udp (l4_proto)) - { - return; - } - - input_node_index = - sm->next_slot_track_node_by_is_ip6_is_out[node_is_ip6][0]; - output_node_index = - sm->next_slot_track_node_by_is_ip6_is_out[node_is_ip6][1]; - - if (node_is_out) - { - session_nexts[0] = output_node_index; - session_nexts[1] = input_node_index; - } - else - { - session_nexts[0] = input_node_index; - session_nexts[1] = output_node_index; - } -} - - -static inline void -swap_bytes (vlib_buffer_t * b0, int off_a, int off_b, int nbytes) -{ - u8 tmp; - u8 *pa = vlib_buffer_get_current (b0) + off_a; - u8 *pb = vlib_buffer_get_current (b0) + off_b; - while (nbytes--) - { - tmp = *pa; - *pa++ = *pb; - *pb++ = tmp; - } -} - -/* - * This quite pro[bv]ably is a terrible idea performance wise. Moreso doing it twice. - * Would having a long (ish) chunk of memory work better for this ? - * We will see when we get to the performance of this. - */ -void -l2sess_flip_l3l4_fields (vlib_buffer_t * b0, int node_is_ip6, u8 l4_proto) -{ - if (!l4_tcp_or_udp (l4_proto)) - { - return; - } - if (node_is_ip6) - { - swap_bytes (b0, 22, 38, 16); /* L3 */ - swap_bytes (b0, 54, 56, 2); /* L4 (when no EH!) */ - } - else - { - swap_bytes (b0, 26, 30, 4); /* L3 */ - swap_bytes (b0, 34, 36, 2); /* L4 */ - } -} - -void -l2sess_add_session (vlib_buffer_t * b0, int node_is_out, int node_is_ip6, - u32 session_table, u32 session_match_next, - u32 opaque_index) -{ - vnet_classify_main_t *cm = &vnet_classify_main; - u32 action = 0; - u32 metadata = 0; - -#ifdef DEBUG_SESSIONS - printf ("Adding session to table %d with next %d\n", session_table, - session_match_next); -#endif - vnet_classify_add_del_session (cm, session_table, - vlib_buffer_get_current (b0), - session_match_next, opaque_index, 0, action, - metadata, 1); -} - - - -static void * -get_ptr_to_offset (vlib_buffer_t * b0, int offset) -{ - u8 *p = vlib_buffer_get_current (b0) + offset; - return p; -} - - -/* - * FIXME: Hardcoded offsets are ugly, although if casting to structs one - * would need to take care about alignment.. So let's for now be naive and simple. - */ - -void -session_store_ip4_l3l4_info (vlib_buffer_t * b0, l2s_session_t * sess, - int node_is_out) -{ - clib_memcpy (&sess->side[1 - node_is_out].addr.ip4, - get_ptr_to_offset (b0, 26), 4); - clib_memcpy (&sess->side[node_is_out].addr.ip4, get_ptr_to_offset (b0, 30), - 4); - sess->side[1 - node_is_out].port = - ntohs (*(u16 *) get_ptr_to_offset (b0, 34)); - sess->side[node_is_out].port = ntohs (*(u16 *) get_ptr_to_offset (b0, 36)); -} - -void -session_store_ip6_l3l4_info (vlib_buffer_t * b0, l2s_session_t * sess, - int node_is_out) -{ - clib_memcpy (&sess->side[1 - node_is_out].addr.ip6, - get_ptr_to_offset (b0, 22), 16); - clib_memcpy (&sess->side[node_is_out].addr.ip4, get_ptr_to_offset (b0, 38), - 16); - sess->side[1 - node_is_out].port = - ntohs (*(u16 *) get_ptr_to_offset (b0, 54)); - sess->side[node_is_out].port = ntohs (*(u16 *) get_ptr_to_offset (b0, 56)); -} - -static void -build_match_from_session (l2sess_main_t * sm, u8 * match, - l2s_session_t * sess, int is_out) -{ - if (sess->is_ip6) - { - match[20] = sess->l4_proto; - clib_memcpy (&match[22], &sess->side[1 - is_out].addr.ip6, 16); - clib_memcpy (&match[38], &sess->side[is_out].addr.ip4, 16); - *(u16 *) & match[54] = htons (sess->side[1 - is_out].port); - *(u16 *) & match[56] = htons (sess->side[is_out].port); - } - else - { - match[23] = sess->l4_proto; - clib_memcpy (&match[26], &sess->side[1 - is_out].addr.ip6, 4); - clib_memcpy (&match[30], &sess->side[is_out].addr.ip4, 4); - *(u16 *) & match[34] = htons (sess->side[1 - is_out].port); - *(u16 *) & match[36] = htons (sess->side[is_out].port); - } -} - -static void -delete_session (l2sess_main_t * sm, u32 sw_if_index, u32 session_index) -{ - vnet_classify_main_t *cm = &vnet_classify_main; - u8 match[5 * 16]; /* For building the mock of the packet to delete the classifier session */ - u32 session_tables[2] = { ~0, ~0 }; - l2s_session_t *sess = sm->sessions + session_index; - if (pool_is_free (sm->sessions, sess)) - { - sm->counter_attempted_delete_free_session++; - return; - } - l2sess_get_session_tables (sm, sw_if_index, 0, sess->is_ip6, sess->l4_proto, - session_tables); - if (session_tables[1] != ~0) - { - build_match_from_session (sm, match, sess, 1); - vnet_classify_add_del_session (cm, session_tables[1], match, 0, 0, 0, 0, - 0, 0); - } - if (session_tables[1] != ~0) - { - build_match_from_session (sm, match, sess, 1); - vnet_classify_add_del_session (cm, session_tables[1], match, 0, 0, 0, 0, - 0, 0); - } - pool_put (sm->sessions, sess); -} - -static void -udp_session_account_buffer (vlib_buffer_t * b0, l2s_session_t * s, - int which_side, u64 now) -{ - l2s_session_side_t *ss = &s->side[which_side]; - ss->active_time = now; - ss->n_packets++; - ss->n_bytes += b0->current_data + b0->current_length; -} - -static inline u64 -udp_session_get_timeout (l2sess_main_t * sm, l2s_session_t * sess, u64 now) -{ - return (sm->udp_session_idle_timeout); -} - -static void -tcp_session_account_buffer (vlib_buffer_t * b0, l2s_session_t * s, - int which_side, u64 now) -{ - l2s_session_side_t *ss = &s->side[which_side]; - ss->active_time = now; - ss->n_packets++; - ss->n_bytes += b0->current_data + b0->current_length; - /* Very very lightweight TCP state tracking: just record which flags were seen */ - s->tcp_flags_seen |= - l2sess_get_tcp_flags (b0, s->is_ip6) << (8 * which_side); -} - -/* - * Since we are tracking for the purposes of timing the sessions out, - * we mostly care about two states: established (maximize the idle timeouts) - * and transient (halfopen/halfclosed/reset) - we need to have a reasonably short timeout to - * quickly get rid of sessions but not short enough to violate the TCP specs. - */ - -static inline u64 -tcp_session_get_timeout (l2sess_main_t * sm, l2s_session_t * sess, u64 now) -{ - /* seen both SYNs and ACKs but not FINs means we are in establshed state */ - u16 masked_flags = - sess->tcp_flags_seen & ((TCP_FLAGS_RSTFINACKSYN << 8) + - TCP_FLAGS_RSTFINACKSYN); - if (((TCP_FLAGS_ACKSYN << 8) + TCP_FLAGS_ACKSYN) == masked_flags) - { - return (sm->tcp_session_idle_timeout); - } - else - { - return (sm->tcp_session_transient_timeout); - } -} - -static inline u64 -session_get_timeout (l2sess_main_t * sm, l2s_session_t * sess, u64 now) -{ - u64 timeout; - - switch (sess->l4_proto) - { - case 6: - timeout = tcp_session_get_timeout (sm, sess, now); - break; - case 17: - timeout = udp_session_get_timeout (sm, sess, now); - break; - default: - timeout = 0; - } - - return timeout; -} - -static inline u64 -get_session_last_active_time(l2s_session_t * sess) -{ - u64 last_active = - sess->side[0].active_time > - sess->side[1].active_time ? sess->side[0].active_time : sess->side[1]. - active_time; - return last_active; -} - -static int -session_is_alive (l2sess_main_t * sm, l2s_session_t * sess, u64 now, u64 *last_active_cache) -{ - u64 last_active = get_session_last_active_time(sess); - u64 timeout = session_get_timeout (sm, sess, now); - int is_alive = ((now - last_active) < timeout); - if (last_active_cache) - *last_active_cache = last_active; - return is_alive; -} - -static void -check_idle_sessions (l2sess_main_t * sm, u32 sw_if_index, u64 now) -{ - sm->timer_wheel_next_expiring_time = 0; - sm->data_from_advancing_timing_wheel - = - timing_wheel_advance (&sm->timing_wheel, now, - sm->data_from_advancing_timing_wheel, - &sm->timer_wheel_next_expiring_time); -#ifdef DEBUG_SESSIONS_VERBOSE - { - clib_time_t *ct = &sm->vlib_main->clib_time; - f64 ctime; - ctime = now * ct->seconds_per_clock; - clib_warning ("Now : %U", format_time_interval, "h:m:s:u", ctime); - ctime = sm->timer_wheel_next_expiring_time * ct->seconds_per_clock; - clib_warning ("Next expire: %U", format_time_interval, "h:m:s:u", ctime); - clib_warning ("Expired items: %d", - (int) vec_len (sm->data_from_advancing_timing_wheel)); - } -#endif - - sm->timer_wheel_next_expiring_time = now + sm->timer_wheel_tick; - if (PREDICT_FALSE ( 0 == sm->data_from_advancing_timing_wheel )) { - return; - } - - if (PREDICT_FALSE (_vec_len (sm->data_from_advancing_timing_wheel) > 0)) - { - uword i; - for (i = 0; i < _vec_len (sm->data_from_advancing_timing_wheel); i++) - { - u32 session_index = sm->data_from_advancing_timing_wheel[i]; - if (!pool_is_free_index (sm->sessions, session_index)) - { - l2s_session_t *sess = sm->sessions + session_index; - u64 last_active; - if (session_is_alive (sm, sess, now, &last_active)) - { -#ifdef DEBUG_SESSIONS - clib_warning ("Restarting timer for session %d", (int) session_index); -#endif - /* Pretend we did this in the past, at last_active moment */ - timing_wheel_insert (&sm->timing_wheel, - last_active + session_get_timeout (sm, sess, - last_active), - session_index); - } - else - { -#ifdef DEBUG_SESSIONS - clib_warning ("Deleting session %d", (int) session_index); -#endif - delete_session (sm, sw_if_index, session_index); - } - } - } - _vec_len (sm->data_from_advancing_timing_wheel) = 0; - } -} - -static uword -l2sess_node_fn (vlib_main_t * vm, - vlib_node_runtime_t * node, vlib_frame_t * frame) -{ - u32 n_left_from, *from, *to_next; - l2sess_next_t next_index; - u32 pkts_swapped = 0; - u32 cached_sw_if_index = (u32) ~ 0; - u32 cached_next_index = (u32) ~ 0; - u32 feature_bitmap0; - u32 trace_flags0; - - l2sess_main_t *sm = &l2sess_main; - - from = vlib_frame_vector_args (frame); - n_left_from = frame->n_vectors; - next_index = node->cached_next_index; - - while (n_left_from > 0) - { - u32 n_left_to_next; - - vlib_get_next_frame (vm, node, next_index, to_next, n_left_to_next); - - /* Only a single loop for now for simplicity */ - - while (n_left_from > 0 && n_left_to_next > 0) - { - u32 bi0; - vlib_buffer_t *b0; - u32 next0 = L2SESS_NEXT_DROP; - u32 sw_if_index0; - //ethernet_header_t *en0; - - /* speculatively enqueue b0 to the current next frame */ - bi0 = from[0]; - to_next[0] = bi0; - from += 1; - to_next += 1; - n_left_from -= 1; - n_left_to_next -= 1; - - b0 = vlib_get_buffer (vm, bi0); - //en0 = vlib_buffer_get_current (b0); - -/* - * The non-boilerplate is in the block below. - * Note first a magic macro block that sets up the behavior qualifiers: - * node_is_out : 1 = is output, 0 = is input - * node_is_ip6 : 1 = is ip6, 0 = is ip4 - * node_is_track : 1 = is a state tracking node, 0 - is a session addition node - * - * Subsequently the code adjusts its behavior depending on these variables. - * It's most probably not great performance wise but much easier to work with. - * - */ - { - int node_is_out = -1; - CLIB_UNUSED (int node_is_ip6) = -1; - CLIB_UNUSED (int node_is_track) = -1; - u32 node_index = 0; - u32 session_tables[2] = { ~0, ~0 }; - u32 session_nexts[2] = { ~0, ~0 }; - l2_output_next_nodes_st *next_nodes = 0; - u32 *input_feat_next_node_index; - u8 l4_proto; - u64 now = clib_cpu_time_now (); - -/* - * Set the variables according to which of the 8 nodes we are. - * Hopefully the compiler is smart enough to eliminate the extraneous. - */ -#define _(node_name, node_var, is_out, is_ip6, is_track) \ -if(node_var.index == node->node_index) \ - { \ - node_is_out = is_out; \ - node_is_ip6 = is_ip6; \ - node_is_track = is_track; \ - node_index = node_var.index; \ - next_nodes = &sm->node_var ## _next_nodes; \ - input_feat_next_node_index = sm->node_var ## _input_next_node_index; \ - } - foreach_l2sess_node -#undef _ - trace_flags0 = 0; - if (node_is_out) - { - sw_if_index0 = vnet_buffer (b0)->sw_if_index[VLIB_TX]; - } - else - { - sw_if_index0 = vnet_buffer (b0)->sw_if_index[VLIB_RX]; - } - /* potentially also remove the nodes here */ - feature_bitmap0 = vnet_buffer (b0)->l2.feature_bitmap; - - if (node_is_track) - { - u32 sess_index = vnet_buffer (b0)->l2_classify.opaque_index; - l2s_session_t *sess = sm->sessions + sess_index; - l4_proto = sess->l4_proto; - - if (session_is_alive (sm, sess, now, 0)) - { - if (6 == l4_proto) - { - tcp_session_account_buffer (b0, sess, node_is_out, - now); - } - else - { - udp_session_account_buffer (b0, sess, node_is_out, - now); - } - } - else - { - timing_wheel_delete (&sm->timing_wheel, sess_index); - delete_session (sm, sw_if_index0, sess_index); - /* FIXME: drop the packet that hit the obsolete node, for now. We really ought to recycle it. */ - next0 = 0; - } - } - else - { - /* - * "-add" node: take l2opaque which arrived to us, and deduce - * the tables out of that. ~0 means the topmost classifier table - * applied for this AF on the RX(for input)/TX(for output)) sw_if_index. - * Also add the mirrored session to the paired table. - */ - l2s_session_t *sess; - u32 sess_index; - - l4_proto = l2sess_get_l4_proto (b0, node_is_ip6); - - pool_get (sm->sessions, sess); - sess_index = sess - sm->sessions; - sess->create_time = now; - sess->side[node_is_out].active_time = now; - sess->side[1 - node_is_out].active_time = now; - sess->l4_proto = l4_proto; - sess->is_ip6 = node_is_ip6; - if (node_is_ip6) - { - session_store_ip6_l3l4_info (b0, sess, node_is_out); - } - else - { - session_store_ip4_l3l4_info (b0, sess, node_is_out); - } - - l2sess_get_session_tables (sm, sw_if_index0, node_is_out, - node_is_ip6, l4_proto, - session_tables); - l2sess_get_session_nexts (sm, sw_if_index0, node_is_out, - node_is_ip6, l4_proto, - session_nexts); - l2sess_flip_l3l4_fields (b0, node_is_ip6, l4_proto); - if (session_tables[1] != ~0) - { - l2sess_add_session (b0, node_is_out, node_is_ip6, - session_tables[1], session_nexts[1], - sess_index); - } - l2sess_flip_l3l4_fields (b0, node_is_ip6, l4_proto); - if (session_tables[0] != ~0) - { - l2sess_add_session (b0, node_is_out, node_is_ip6, - session_tables[0], session_nexts[0], - sess_index); - } - if (6 == sess->l4_proto) - { - tcp_session_account_buffer (b0, sess, node_is_out, now); - } - else - { - udp_session_account_buffer (b0, sess, node_is_out, now); - } - timing_wheel_insert (&sm->timing_wheel, - now + session_get_timeout (sm, sess, - now), - sess_index); - } - - if (now >= sm->timer_wheel_next_expiring_time) - { - check_idle_sessions (sm, sw_if_index0, now); - } - - if (node_is_out) - { - if (feature_bitmap0) - { - trace_flags0 |= 0x10; - } - if (sw_if_index0 == cached_sw_if_index) - { - trace_flags0 |= 0x20; - } - l2_output_dispatch (sm->vlib_main, - sm->vnet_main, - node, - node_index, - &cached_sw_if_index, - &cached_next_index, - next_nodes, - b0, sw_if_index0, feature_bitmap0, - &next0); - trace_flags0 |= 2; - - } - else - { - next0 = - feat_bitmap_get_next_node_index (input_feat_next_node_index, - feature_bitmap0); - trace_flags0 |= 4; - - } - - - - if (next0 >= node->n_next_nodes) - { - trace_flags0 |= 1; - } - - if (PREDICT_FALSE ((node->flags & VLIB_NODE_FLAG_TRACE) - && (b0->flags & VLIB_BUFFER_IS_TRACED))) - { - l2sess_trace_t *t = - vlib_add_trace (vm, node, b0, sizeof (*t)); - t->sw_if_index = sw_if_index0; - t->next_index = next0; - t->trace_flags = trace_flags0; - t->l4_proto = l4_proto; - t->session_tables[0] = session_tables[0]; - t->session_tables[1] = session_tables[1]; - t->session_nexts[0] = session_nexts[0]; - t->session_nexts[1] = session_nexts[1]; - } - - } - pkts_swapped += 1; - if (next0 >= node->n_next_nodes) - { - next0 = 0; - } - - /* verify speculative enqueue, maybe switch current next frame */ - vlib_validate_buffer_enqueue_x1 (vm, node, next_index, - to_next, n_left_to_next, - bi0, next0); - } - - vlib_put_next_frame (vm, node, next_index, n_left_to_next); - } - vlib_node_increment_counter (vm, node->node_index, - L2SESS_ERROR_SWAPPED, pkts_swapped); - return frame->n_vectors; -} - - -#define _(node_name, node_var, is_out, is_ip6, is_track) \ -static uword \ -node_var ## node_fn (vlib_main_t * vm, \ - vlib_node_runtime_t * node, \ - vlib_frame_t * frame) \ -{ \ - return l2sess_node_fn(vm, node, frame); \ -} \ -VLIB_REGISTER_NODE (node_var) = { \ - .function = node_var ## node_fn, \ - .name = node_name, \ - .vector_size = sizeof (u32), \ - .format_trace = format_ ## node_var ## _trace, \ - .type = VLIB_NODE_TYPE_INTERNAL, \ - \ - .n_errors = ARRAY_LEN(l2sess_error_strings), \ - .error_strings = l2sess_error_strings, \ - \ - .n_next_nodes = L2SESS_N_NEXT, \ - .next_nodes = { \ - [L2SESS_NEXT_DROP] = "error-drop", \ - }, \ -}; -foreach_l2sess_node -#undef _ diff --git a/plugins/acl-plugin/acl/node_in.c b/plugins/acl-plugin/acl/node_in.c deleted file mode 100644 index 2a5199a9..00000000 --- a/plugins/acl-plugin/acl/node_in.c +++ /dev/null @@ -1,168 +0,0 @@ -/* - * Copyright (c) 2016 Cisco and/or its affiliates. - * Licensed under the Apache License, Version 2.0 (the "License"); - * you may not use this file except in compliance with the License. - * You may obtain a copy of the License at: - * - * http://www.apache.org/licenses/LICENSE-2.0 - * - * Unless required by applicable law or agreed to in writing, software - * distributed under the License is distributed on an "AS IS" BASIS, - * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. - * See the License for the specific language governing permissions and - * limitations under the License. - */ -#include -#include -#include -#include -#include -#include "node_in.h" - -typedef struct -{ - u32 next_index; - u32 sw_if_index; - u32 match_acl_index; - u32 match_rule_index; - u32 trace_bitmap; -} acl_in_trace_t; - -/* packet trace format function */ -static u8 * -format_acl_in_trace (u8 * s, va_list * args) -{ - CLIB_UNUSED (vlib_main_t * vm) = va_arg (*args, vlib_main_t *); - CLIB_UNUSED (vlib_node_t * node) = va_arg (*args, vlib_node_t *); - acl_in_trace_t *t = va_arg (*args, acl_in_trace_t *); - - s = - format (s, - "ACL_IN: sw_if_index %d, next index %d, match: inacl %d rule %d trace_bits %08x", - t->sw_if_index, t->next_index, t->match_acl_index, - t->match_rule_index, t->trace_bitmap); - return s; -} - -vlib_node_registration_t acl_in_node; - -#define foreach_acl_in_error \ -_(ACL_CHECK, "InACL check packets processed") - -typedef enum -{ -#define _(sym,str) ACL_IN_ERROR_##sym, - foreach_acl_in_error -#undef _ - ACL_IN_N_ERROR, -} acl_in_error_t; - -static char *acl_in_error_strings[] = { -#define _(sym,string) string, - foreach_acl_in_error -#undef _ -}; - -static uword -acl_in_node_fn (vlib_main_t * vm, - vlib_node_runtime_t * node, vlib_frame_t * frame) -{ - u32 n_left_from, *from, *to_next; - acl_in_next_t next_index; - u32 pkts_acl_checked = 0; - u32 feature_bitmap0; - u32 trace_bitmap = 0; - u32 *input_feat_next_node_index = - acl_main.acl_in_node_input_next_node_index; - - from = vlib_frame_vector_args (frame); - n_left_from = frame->n_vectors; - next_index = node->cached_next_index; - - while (n_left_from > 0) - { - u32 n_left_to_next; - - vlib_get_next_frame (vm, node, next_index, to_next, n_left_to_next); - - while (n_left_from > 0 && n_left_to_next > 0) - { - u32 bi0; - vlib_buffer_t *b0; - u32 next0 = ~0; - u32 sw_if_index0; - u32 next = ~0; - u32 match_acl_index = ~0; - u32 match_rule_index = ~0; - - /* speculatively enqueue b0 to the current next frame */ - bi0 = from[0]; - to_next[0] = bi0; - from += 1; - to_next += 1; - n_left_from -= 1; - n_left_to_next -= 1; - - b0 = vlib_get_buffer (vm, bi0); - - - sw_if_index0 = vnet_buffer (b0)->sw_if_index[VLIB_RX]; - feature_bitmap0 = vnet_buffer (b0)->l2.feature_bitmap; - - input_acl_packet_match (sw_if_index0, b0, &next, &match_acl_index, - &match_rule_index, &trace_bitmap); - if (next != ~0) - { - next0 = next; - } - if (next0 == ~0) - { - next0 = - feat_bitmap_get_next_node_index (input_feat_next_node_index, - feature_bitmap0); - } - - if (PREDICT_FALSE ((node->flags & VLIB_NODE_FLAG_TRACE) - && (b0->flags & VLIB_BUFFER_IS_TRACED))) - { - acl_in_trace_t *t = vlib_add_trace (vm, node, b0, sizeof (*t)); - t->sw_if_index = sw_if_index0; - t->next_index = next0; - t->match_acl_index = match_acl_index; - t->match_rule_index = match_rule_index; - t->trace_bitmap = trace_bitmap; - } - - next0 = next0 < node->n_next_nodes ? next0 : 0; - - pkts_acl_checked += 1; - - /* verify speculative enqueue, maybe switch current next frame */ - vlib_validate_buffer_enqueue_x1 (vm, node, next_index, - to_next, n_left_to_next, - bi0, next0); - } - - vlib_put_next_frame (vm, node, next_index, n_left_to_next); - } - - vlib_node_increment_counter (vm, acl_in_node.index, - ACL_IN_ERROR_ACL_CHECK, pkts_acl_checked); - return frame->n_vectors; -} - -VLIB_REGISTER_NODE (acl_in_node) = -{ - .function = acl_in_node_fn,.name = "acl-plugin-in",.vector_size = - sizeof (u32),.format_trace = format_acl_in_trace,.type = - VLIB_NODE_TYPE_INTERNAL,.n_errors = - ARRAY_LEN (acl_in_error_strings),.error_strings = - acl_in_error_strings,.n_next_nodes = ACL_IN_N_NEXT, - /* edit / add dispositions here */ - .next_nodes = - { - [ACL_IN_ERROR_DROP] = "error-drop", - [ACL_IN_ETHERNET_INPUT] = "ethernet-input", - [ACL_IN_L2S_INPUT_IP4_ADD] = "aclp-l2s-input-ip4-add", - [ACL_IN_L2S_INPUT_IP6_ADD] = "aclp-l2s-input-ip6-add",} -,}; diff --git a/plugins/acl-plugin/acl/node_in.h b/plugins/acl-plugin/acl/node_in.h deleted file mode 100644 index 502bbf8d..00000000 --- a/plugins/acl-plugin/acl/node_in.h +++ /dev/null @@ -1,12 +0,0 @@ -#ifndef _NODE_IN_H_ -#define _NODE_IN_H_ - -typedef enum { - ACL_IN_ERROR_DROP, - ACL_IN_ETHERNET_INPUT, - ACL_IN_L2S_INPUT_IP4_ADD, - ACL_IN_L2S_INPUT_IP6_ADD, - ACL_IN_N_NEXT, -} acl_in_next_t; - -#endif diff --git a/plugins/acl-plugin/acl/node_out.c b/plugins/acl-plugin/acl/node_out.c deleted file mode 100644 index 50af3679..00000000 --- a/plugins/acl-plugin/acl/node_out.c +++ /dev/null @@ -1,175 +0,0 @@ -/* - * Copyright (c) 2016 Cisco and/or its affiliates. - * Licensed under the Apache License, Version 2.0 (the "License"); - * you may not use this file except in compliance with the License. - * You may obtain a copy of the License at: - * - * http://www.apache.org/licenses/LICENSE-2.0 - * - * Unless required by applicable law or agreed to in writing, software - * distributed under the License is distributed on an "AS IS" BASIS, - * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. - * See the License for the specific language governing permissions and - * limitations under the License. - */ -#include -#include -#include -#include -#include - -#include "node_out.h" - -typedef struct -{ - u32 next_index; - u32 sw_if_index; - u32 match_acl_index; - u32 match_rule_index; - u32 trace_bitmap; -} acl_out_trace_t; - -/* packet trace format function */ -static u8 * -format_acl_out_trace (u8 * s, va_list * args) -{ - CLIB_UNUSED (vlib_main_t * vm) = va_arg (*args, vlib_main_t *); - CLIB_UNUSED (vlib_node_t * node) = va_arg (*args, vlib_node_t *); - acl_out_trace_t *t = va_arg (*args, acl_out_trace_t *); - s = - format (s, - "ACL_OUT: sw_if_index %d, next index %d, match: outacl %d rule %d trace_bits %08x", - t->sw_if_index, t->next_index, t->match_acl_index, - t->match_rule_index, t->trace_bitmap); - return s; -} - -vlib_node_registration_t acl_out_node; - -#define foreach_acl_out_error \ -_(ACL_CHECK, "OutACL check packets processed") - -typedef enum -{ -#define _(sym,str) ACL_OUT_ERROR_##sym, - foreach_acl_out_error -#undef _ - ACL_OUT_N_ERROR, -} acl_out_error_t; - -static char *acl_out_error_strings[] = { -#define _(sym,string) string, - foreach_acl_out_error -#undef _ -}; - -static uword -acl_out_node_fn (vlib_main_t * vm, - vlib_node_runtime_t * node, vlib_frame_t * frame) -{ - acl_main_t *am = &acl_main; - l2_output_next_nodes_st *next_nodes = &am->acl_out_output_next_nodes; - u32 n_left_from, *from, *to_next; - acl_out_next_t next_index; - u32 pkts_acl_checked = 0; - u32 feature_bitmap0; - u32 cached_sw_if_index = (u32) ~ 0; - u32 cached_next_index = (u32) ~ 0; - u32 match_acl_index = ~0; - u32 match_rule_index = ~0; - u32 trace_bitmap = 0; - - from = vlib_frame_vector_args (frame); - n_left_from = frame->n_vectors; - next_index = node->cached_next_index; - - while (n_left_from > 0) - { - u32 n_left_to_next; - - vlib_get_next_frame (vm, node, next_index, to_next, n_left_to_next); - - while (n_left_from > 0 && n_left_to_next > 0) - { - u32 bi0; - vlib_buffer_t *b0; - u32 next0 = ~0; - u32 next = 0; - u32 sw_if_index0; - - /* speculatively enqueue b0 to the current next frame */ - bi0 = from[0]; - to_next[0] = bi0; - from += 1; - to_next += 1; - n_left_from -= 1; - n_left_to_next -= 1; - - b0 = vlib_get_buffer (vm, bi0); - - - sw_if_index0 = vnet_buffer (b0)->sw_if_index[VLIB_TX]; - feature_bitmap0 = vnet_buffer (b0)->l2.feature_bitmap; - - output_acl_packet_match (sw_if_index0, b0, &next, &match_acl_index, - &match_rule_index, &trace_bitmap); - if (next != ~0) - { - next0 = next; - } - if (next0 == ~0) - { - l2_output_dispatch (vm, - am->vnet_main, - node, - acl_out_node.index, - &cached_sw_if_index, - &cached_next_index, - next_nodes, - b0, sw_if_index0, feature_bitmap0, &next0); - } - - - - if (PREDICT_FALSE ((node->flags & VLIB_NODE_FLAG_TRACE) - && (b0->flags & VLIB_BUFFER_IS_TRACED))) - { - acl_out_trace_t *t = vlib_add_trace (vm, node, b0, sizeof (*t)); - t->sw_if_index = sw_if_index0; - t->next_index = next0; - t->match_acl_index = match_acl_index; - t->match_rule_index = match_rule_index; - t->trace_bitmap = trace_bitmap; - } - - pkts_acl_checked += 1; - - /* verify speculative enqueue, maybe switch current next frame */ - vlib_validate_buffer_enqueue_x1 (vm, node, next_index, - to_next, n_left_to_next, - bi0, next0); - } - - vlib_put_next_frame (vm, node, next_index, n_left_to_next); - } - - vlib_node_increment_counter (vm, acl_out_node.index, - ACL_OUT_ERROR_ACL_CHECK, pkts_acl_checked); - return frame->n_vectors; -} - -VLIB_REGISTER_NODE (acl_out_node) = -{ - .function = acl_out_node_fn,.name = "acl-plugin-out",.vector_size = - sizeof (u32),.format_trace = format_acl_out_trace,.type = - VLIB_NODE_TYPE_INTERNAL,.n_errors = - ARRAY_LEN (acl_out_error_strings),.error_strings = - acl_out_error_strings,.n_next_nodes = ACL_OUT_N_NEXT, - /* edit / add dispositions here */ - .next_nodes = - { - [ACL_OUT_ERROR_DROP] = "error-drop", - [ACL_OUT_INTERFACE_OUTPUT] = "interface-output", - [ACL_OUT_L2S_OUTPUT_IP4_ADD] = "aclp-l2s-output-ip4-add", - [ACL_OUT_L2S_OUTPUT_IP6_ADD] = "aclp-l2s-output-ip6-add",} -,}; diff --git a/plugins/acl-plugin/acl/node_out.h b/plugins/acl-plugin/acl/node_out.h deleted file mode 100644 index c919f3b7..00000000 --- a/plugins/acl-plugin/acl/node_out.h +++ /dev/null @@ -1,12 +0,0 @@ -#ifndef _NODE_OUT_H_ -#define _NODE_OUT_H_ - -typedef enum { - ACL_OUT_ERROR_DROP, - ACL_OUT_INTERFACE_OUTPUT, - ACL_OUT_L2S_OUTPUT_IP4_ADD, - ACL_OUT_L2S_OUTPUT_IP6_ADD, - ACL_OUT_N_NEXT, -} acl_out_next_t; - -#endif diff --git a/plugins/acl-plugin/configure.ac b/plugins/acl-plugin/configure.ac deleted file mode 100644 index 204f4e4d..00000000 --- a/plugins/acl-plugin/configure.ac +++ /dev/null @@ -1,24 +0,0 @@ -AC_INIT(acl_plugin, 1.0) -AC_CONFIG_MACRO_DIR([../../vpp-api/java/m4]) -LT_INIT -AM_INIT_AUTOMAKE -AM_SILENT_RULES([yes]) -AC_PREFIX_DEFAULT([/usr]) - -AC_PROG_CC - -if test -f /usr/bin/lsb_release && test `lsb_release -si` == "Ubuntu" && test `lsb_release -sr` == "14.04" && test -d /usr/lib/jvm/java-8-openjdk-amd64/ ; then - JAVA_HOME=/usr/lib/jvm/java-8-openjdk-amd64/ - JAVAC=${JAVA_HOME}/bin/javac - PATH=${JAVA_HOME}/bin/:${PATH} - break -fi - -AX_CHECK_JAVA_HOME -AX_PROG_JAVAC -AX_PROG_JAVAH -AX_PROG_JAR -AX_PROG_JAVADOC -AX_PROG_JAVA - -AC_OUTPUT([Makefile]) diff --git a/plugins/acl-plugin/test/run-python b/plugins/acl-plugin/test/run-python deleted file mode 100755 index 215eb17a..00000000 --- a/plugins/acl-plugin/test/run-python +++ /dev/null @@ -1,28 +0,0 @@ -#!/bin/sh -# -# Do all the legwork to run a scapy shell with APIs available for load -# -CURR_DIR=`pwd` -ROOT_DIR=`git rev-parse --show-toplevel` -cd $ROOT_DIR -sudo apt-get install -y python-virtualenv -# uncomment the line below to enable build of plugins and api each time -# make plugins && make build-vpp-api || exit -virtualenv virtualenv -virtualenv/bin/pip install ipaddress -virtualenv/bin/pip install scapy -# install the python API into the virtualenv -cd $ROOT_DIR/vpp-api/python/ -$ROOT_DIR/virtualenv/bin/python setup.py install -# install the python ACL plugin API into the virtualenv -ACL_PLUGIN_SETUP_DIR=`find $ROOT_DIR/build-root -name acl-plugin` -cd $ACL_PLUGIN_SETUP_DIR; -$ROOT_DIR/virtualenv/bin/python setup.py install -cd $ROOT_DIR -# figure out the shared library path and start scapy -export LD_LIBRARY_PATH=`pwd`/`find . -name "libpneum.so" -exec dirname {} \; | grep lib64 | head -n 1` -cd $CURR_DIR -sudo LD_LIBRARY_PATH=$LD_LIBRARY_PATH $ROOT_DIR/virtualenv/bin/python $1 $2 $3 $4 $5 $6 $7 $8 $9 - - - diff --git a/plugins/acl-plugin/test/run-scapy b/plugins/acl-plugin/test/run-scapy deleted file mode 100755 index 266f07d1..00000000 --- a/plugins/acl-plugin/test/run-scapy +++ /dev/null @@ -1,26 +0,0 @@ -#!/bin/sh -# -# Do all the legwork to run a scapy shell with APIs available for load -# -ROOT_DIR=`git rev-parse --show-toplevel` -cd $ROOT_DIR -sudo apt-get install -y python-virtualenv -# uncomment the line below to enable the build of plugins and API each time.. -# make plugins && make build-vpp-api || exit -virtualenv virtualenv -virtualenv/bin/pip install ipaddress -virtualenv/bin/pip install scapy -# install the python API into the virtualenv -cd $ROOT_DIR/vpp-api/python/ -$ROOT_DIR/virtualenv/bin/python setup.py install -# install the python ACL plugin API into the virtualenv -ACL_PLUGIN_SETUP_DIR=`find $ROOT_DIR/build-root -name acl-plugin` -cd $ACL_PLUGIN_SETUP_DIR; -$ROOT_DIR/virtualenv/bin/python setup.py install -cd $ROOT_DIR -# figure out the shared library path and start scapy -export LD_LIBRARY_PATH=`pwd`/`find . -name "libpneum.so" -exec dirname {} \; | grep lib64 | head -n 1` -sudo LD_LIBRARY_PATH=$LD_LIBRARY_PATH virtualenv/bin/scapy - - - diff --git a/plugins/acl-plugin/test/test_acl_plugin.py b/plugins/acl-plugin/test/test_acl_plugin.py deleted file mode 100644 index 7fc72d67..00000000 --- a/plugins/acl-plugin/test/test_acl_plugin.py +++ /dev/null @@ -1,118 +0,0 @@ -from __future__ import print_function -import unittest, sys, time, threading, struct, logging, os -import vpp_papi -# import vpp_papi_plugins.acl -from ipaddress import * -papi_event = threading.Event() -print(vpp_papi.vpe.VL_API_SW_INTERFACE_SET_FLAGS) -def papi_event_handler(result): - if result.vl_msg_id == vpp_papi.vpe.VL_API_SW_INTERFACE_SET_FLAGS: - return - if result.vl_msg_id == vpp_papi.vpe.VL_API_VNET_INTERFACE_COUNTERS: - print('Interface counters', result) - return - if result.vl_msg_id == vpp_papi.vpe.VL_API_VNET_IP6_FIB_COUNTERS: - print('IPv6 FIB counters', result) - papi_event.set() - return - - print('Unknown message id:', result.vl_msg_id) - -import glob, subprocess -class TestAclPlugin(unittest.TestCase): - @classmethod - def setUpClass(cls): - print("Setup") - @classmethod - def tearDownClass(cls): - print("Teardown") - - def setUp(self): - print("Connecting API") - r = vpp_papi.connect("test_papi") - self.assertEqual(r, 0) - - def tearDown(self): - r = vpp_papi.disconnect() - self.assertEqual(r, 0) - - # - # The tests themselves - # - - # - # Basic request / reply - # - def test_show_version(self): - t = vpp_papi.show_version() - print('T', t); - program = t.program.decode().rstrip('\x00') - self.assertEqual('vpe', program) - - def x_test_acl_add(self): - print("Test ACL add") - self.assertEqual(1, 1) - - # - # Details / Dump - # - def x_test_details_dump(self): - t = vpp_papi.sw_interface_dump(0, b'') - print('Dump/details T', t) - - # - # Arrays - # - def x_test_arrays(self): - t = vpp_papi.vnet_get_summary_stats() - print('Summary stats', t) - print('Packets:', t.total_pkts[0]) - print('Packets:', t.total_pkts[1]) - # - # Variable sized arrays and counters - # - #@unittest.skip("stats") - def x_test_want_stats(self): - pid = 123 - vpp_papi.register_event_callback(papi_event_handler) - papi_event.clear() - - # Need to configure IPv6 to get som IPv6 FIB stats - t = vpp_papi.create_loopback('') - print(t) - self.assertEqual(t.retval, 0) - - ifindex = t.sw_if_index - addr = str(IPv6Address(u'1::1').packed) - t = vpp_papi.sw_interface_add_del_address(ifindex, 1, 1, 0, 16, addr) - print(t) - self.assertEqual(t.retval, 0) - - # Check if interface is up - # XXX: Add new API to query interface state based on ifindex, instead of dump all. - t = vpp_papi.sw_interface_set_flags(ifindex, 1, 1, 0) - self.assertEqual(t.retval, 0) - - t = vpp_papi.want_stats(True, pid) - - print (t) - - # - # Wait for some stats - # - self.assertEqual(papi_event.wait(15), True) - t = vpp_papi.want_stats(False, pid) - print (t) - - - # - # Plugins? - # - -if __name__ == '__main__' or __name__ == '__builtin__': - print("This is main") - suite = unittest.TestLoader().loadTestsFromTestCase(TestAclPlugin) - unittest.TextTestRunner(verbosity=2).run(suite) - #logging.basicConfig(level=logging.DEBUG) - # unittest.main() - diff --git a/plugins/configure.ac b/plugins/configure.ac index 9c631634..b6ba4152 100644 --- a/plugins/configure.ac +++ b/plugins/configure.ac @@ -53,11 +53,6 @@ AM_CONDITIONAL(ENABLE_$1_PLUGIN, test "$enable_the_plugin" = "1") # SUBDIRS += new-plugin # endif -PLUGIN_ENABLED(ioam) -PLUGIN_ENABLED(snat) -PLUGIN_ENABLED(lb) -PLUGIN_ENABLED(acl) - # Disabled plugins, require --enable-XXX-plugin PLUGIN_DISABLED(vcgn) PLUGIN_DISABLED(sample) diff --git a/plugins/ioam-plugin/Makefile.am b/plugins/ioam-plugin/Makefile.am deleted file mode 100644 index b00b088c..00000000 --- a/plugins/ioam-plugin/Makefile.am +++ /dev/null @@ -1,338 +0,0 @@ -# Copyright (c) 2015 Cisco and/or its affiliates. -# Licensed under the Apache License, Version 2.0 (the "License"); -# you may not use this file except in compliance with the License. -# You may obtain a copy of the License at: -# -# http://www.apache.org/licenses/LICENSE-2.0 -# -# Unless required by applicable law or agreed to in writing, software -# distributed under the License is distributed on an "AS IS" BASIS, -# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. -# See the License for the specific language governing permissions and -# limitations under the License. - -AUTOMAKE_OPTIONS = foreign subdir-objects - -AM_CFLAGS = -Wall -AM_LDFLAGS = -module -shared -avoid-version - -######################################## -# iOAM Proof of Transit -######################################## - -ioam_pot_plugin_la_SOURCES = \ - ioam/lib-pot/pot_util.c \ - ioam/encap/ip6_ioam_pot.c \ - ioam/lib-pot/pot_util.h \ - ioam/lib-pot/math64.h \ - ioam/lib-pot/pot_api.c - -BUILT_SOURCES = \ - ioam/lib-pot/pot.api.h \ - ioam/lib-pot/pot.api.json - -SUFFIXES = .api.h .api - -%.api.h: %.api - mkdir -p `dirname $@` ; \ - $(CC) $(CPPFLAGS) -E -P -C -x c $^ \ - | vppapigen --input - --output $@ --show-name $@ - -%.api.json: %.api - @echo " JSON APIGEN " $@ ; \ - mkdir -p `dirname $@` ; \ - $(CC) $(CPPFLAGS) -E -P -C -x c $^ \ - | vppapigen --input - --json $@ - -apidir = $(prefix)/ioam/ -api_DATA = \ - ioam/lib-pot/pot.api.json \ - ioam/lib-trace/trace.api.json \ - ioam/export/ioam_export.api.json - -noinst_HEADERS = \ - ioam/lib-pot/pot_all_api_h.h \ - ioam/lib-pot/pot_msg_enum.h \ - ioam/lib-pot/pot.api.h \ - ioam/lib-pot/pot_util.h \ - ioam/lib-pot/math64.h - -ioam_pot_test_plugin_la_SOURCES = \ - ioam/lib-pot/pot_test.c \ - ioam/lib-pot/pot_plugin.api.h - -vppapitestpluginsdir = ${libdir}/vpp_api_test_plugins -vpppluginsdir = ${libdir}/vpp_plugins - -vppapitestplugins_LTLIBRARIES = ioam_pot_test_plugin.la -vppplugins_LTLIBRARIES = ioam_pot_plugin.la - -######################################## -# iOAM trace export for IPv6 -######################################## - -ioam_export_plugin_la_SOURCES = \ -ioam/export/ioam_export.c \ -ioam/export/node.c \ -ioam/export/ioam_export.api.h \ -ioam/export/ioam_export_thread.c - -BUILT_SOURCES += \ - ioam/export/ioam_export.api.h \ - ioam/export/ioam_export.api.json - -noinst_HEADERS += \ - ioam/export/ioam_export_all_api_h.h \ - ioam/export/ioam_export_msg_enum.h \ - ioam/export/ioam_export.api.h - -ioam_export_test_plugin_la_SOURCES = \ - ioam/export/ioam_export_test.c \ - ioam/export/ioam_export_plugin.api.h - -vppapitestplugins_LTLIBRARIES += ioam_export_test_plugin.la -vppplugins_LTLIBRARIES += ioam_export_plugin.la - -######################################## -# iOAM Trace -######################################## -libioam_trace_plugin_la_SOURCES = \ - ioam/lib-trace/trace_util.c \ - ioam/encap/ip6_ioam_trace.c \ - ioam/lib-trace/trace_util.h \ - ioam/lib-trace/trace_api.c - -BUILT_SOURCES += \ - ioam/lib-trace/trace.api.h \ - ioam/lib-trace/trace.api.json - -noinst_HEADERS += \ - ioam/export/ioam_export_all_api_h.h \ - ioam/lib-trace/trace_all_api_h.h \ - ioam/lib-trace/trace_msg_enum.h \ - ioam/lib-trace/trace.api.h \ - ioam/lib-trace/trace_util.h - -ioam_trace_test_plugin_la_SOURCES = \ - ioam/lib-trace/trace_test.c \ - ioam/lib-trace/trace_plugin.api.h - -vppapitestplugins_LTLIBRARIES += ioam_trace_test_plugin.la -vppplugins_LTLIBRARIES += libioam_trace_plugin.la - -######################################## -# VxLAN-GPE -######################################## -libioam_vxlan_gpe_plugin_la_SOURCES = \ - ioam/lib-vxlan-gpe/ioam_encap.c \ - ioam/lib-vxlan-gpe/ioam_decap.c \ - ioam/lib-vxlan-gpe/ioam_transit.c \ - ioam/lib-vxlan-gpe/ioam_pop.c \ - ioam/lib-vxlan-gpe/vxlan_gpe_api.c \ - ioam/lib-vxlan-gpe/vxlan_gpe_ioam_trace.c \ - ioam/lib-vxlan-gpe/vxlan_gpe_ioam.c \ - ioam/export-vxlan-gpe/vxlan_gpe_ioam_export.c \ - ioam/export-vxlan-gpe/vxlan_gpe_node.c \ - ioam/export-vxlan-gpe/vxlan_gpe_ioam_export.api.h\ - ioam/export-vxlan-gpe/vxlan_gpe_ioam_export_thread.c - -BUILT_SOURCES += \ - ioam/lib-vxlan-gpe/vxlan_gpe.api.h \ - ioam/lib-vxlan-gpe/vxlan_gpe.api.json \ - ioam/export-vxlan-gpe/vxlan_gpe_ioam_export.api.h \ - ioam/export-vxlan-gpe/vxlan_gpe_ioam_export.api.json - -noinst_HEADERS += \ - ioam/export/ioam_export_all_api_h.h \ - ioam/lib-vxlan-gpe/vxlan_gpe_all_api_h.h \ - ioam/lib-vxlan-gpe/vxlan_gpe_msg_enum.h \ - ioam/lib-vxlan-gpe/vxlan_gpe.api.h \ - ioam/lib-vxlan-gpe/vxlan_gpe_ioam_util.h \ - ioam/lib-vxlan-gpe/vxlan_gpe_ioam_packet.h \ - ioam/lib-vxlan-gpe/vxlan_gpe_ioam.h \ - ioam/export-vxlan-gpe/vxlan_gpe_ioam_export_all_api_h.h \ - ioam/export-vxlan-gpe/vxlan_gpe_ioam_export_msg_enum.h \ - ioam/export-vxlan-gpe/vxlan_gpe_ioam_export.api.h - -ioam_vxlan_gpe_test_plugin_la_SOURCES = \ - ioam/lib-vxlan-gpe/vxlan_gpe_test.c \ - ioam/lib-vxlan-gpe/vxlan_gpe_plugin.api.h - -libioam_vxlan_gpe_plugin_la_LIBADD = libioam_trace_plugin.la - -vppapitestplugins_LTLIBRARIES += ioam_vxlan_gpe_test_plugin.la -vppplugins_LTLIBRARIES += libioam_vxlan_gpe_plugin.la - -vxlan_gpe_ioam_export_test_plugin_la_SOURCES = \ - ioam/export-vxlan-gpe/vxlan_gpe_ioam_export_test.c \ - ioam/export-vxlan-gpe/vxlan_gpe_ioam_export_plugin.api.h - -vppapitestplugins_LTLIBRARIES += vxlan_gpe_ioam_export_test_plugin.la - -######################################## -# iOAM E2E plugin -######################################## - -ioam_e2e_plugin_la_SOURCES = \ - ioam/encap/ip6_ioam_e2e.c \ - ioam/encap/ip6_ioam_seqno.c \ - ioam/encap/ip6_ioam_seqno_analyse.c - -noinst_HEADERS += \ - ioam/encap/ip6_ioam_e2e.h \ - ioam/encap/ip6_ioam_seqno.h - -vppplugins_LTLIBRARIES += ioam_e2e_plugin.la - -# Remove *.la files -install-data-hook: - @(cd $(vpppluginsdir) && $(RM) $(vppplugins_LTLIBRARIES)) - @(cd $(vppapitestpluginsdir) && $(RM) $(vppapitestplugins_LTLIBRARIES)) - - -# -# Java code generation -# -jvpp_registry_root = ../../vpp-api/java -jvpp_registry_version = 17.04 -jioam_trace_jarfile = jvpp-ioam-trace-$(PACKAGE_VERSION).jar -jvpp_trace_package_dir = io/fd/vpp/jvpp/ioamtrace -jvpp_root = ioam/jvpp -jvpp_target_dir = target -jvpp_target = $(jvpp_root)/$(jvpp_target_dir) -trace_api_file=$(srcdir)/ioam/lib-trace/trace.api - - -lib_LTLIBRARIES = libjvpp_ioamtrace.la -libjvpp_ioamtrace_la_SOURCES = ioam/lib-trace/trace.api.h ioam/lib-trace/jvpp_ioam_trace.c ioam/jvpp/io_fd_vpp_jvpp_ioam_trace_JVppIoamTraceImpl.h -libjvpp_ioamtrace_la_LIBADD = -lvlibmemoryclient -lvlibapi -lvppinfra \ - -lpthread -lm -lrt -L$(jvpp_registry_root)/.libs -ljvpp_common -libjvpp_ioamtrace_la_LDFLAGS = -module -libjvpp_ioamtrace_la_CPPFLAGS = -I$(JAVA_HOME)/include -I$(JAVA_HOME)/include/linux -I../ -I$(srcdir)/../ - -BUILT_SOURCES += $(jvpp_root)/io_fd_vpp_jvpp_ioamtrace_JVppIoamtraceImpl.h - -$(jvpp_root)/io_fd_vpp_jvpp_ioamtrace_JVppIoamtraceImpl.h: ioam_trace.api.json - dir=`pwd`; \ - mkdir -p $(jvpp_target); \ - mkdir -p $(jvpp_root)/$(jvpp_trace_package_dir); \ - cd $(jvpp_root)/$(jvpp_trace_package_dir); \ - mkdir -p dto future callfacade callback notification test; \ - @srcdir@/$(jvpp_registry_root)/jvpp/gen/jvpp_gen.py -i $${dir}/ioam_trace.api.json --plugin_name ioamtrace; \ - cd -; \ - mv -f $(jvpp_root)/$(jvpp_trace_package_dir)/jvpp_ioamtrace_gen.h $(jvpp_root)/jvpp_ioam_trace_gen.h; \ - cp $(srcdir)/$(jvpp_root)/$(jvpp_trace_package_dir)/test/*.java $(jvpp_root)/$(jvpp_trace_package_dir)/test/; \ - cd $(jvpp_root); \ - $(JAVAC) -classpath .:$(jvpp_target_dir):../../$(jvpp_registry_root)/jvpp-registry-$(jvpp_registry_version).jar -d $(jvpp_target_dir) $(jvpp_trace_package_dir)/*.java \ - $(jvpp_trace_package_dir)/dto/*.java \ - $(jvpp_trace_package_dir)/callback/*.java \ - $(jvpp_trace_package_dir)/notification/*.java \ - $(jvpp_trace_package_dir)/future/*.java \ - $(jvpp_trace_package_dir)/callfacade/*.java \ - $(jvpp_trace_package_dir)/test/*.java \ - || (echo "ioam trace jvpp compilation failed: $$?"; exit 1); \ - $(JAVAH) -classpath .:$(jvpp_target_dir):../../$(jvpp_registry_root)/jvpp-registry-$(jvpp_registry_version).jar -d . io.fd.vpp.jvpp.ioamtrace.JVppIoamtraceImpl ; - -$(jioam_trace_jarfile): libjvpp_ioamtrace.la - cp .libs/libjvpp_ioamtrace.so.0.0.0 $(jvpp_target); \ - cd $(jvpp_target); \ - $(JAR) cfv $(JARFLAGS) ../../../$@ libjvpp_ioamtrace.so.0.0.0 $(jvpp_trace_package_dir)/* ; cd ..; - -ioam_trace.api.json: - @echo " jIoam_trace API"; \ - vppapigen --input $(trace_api_file) --json ioam_trace.api.json; - -all-local: $(jioam_trace_jarfile) - - -jioam_pot_jarfile = jvpp-ioam-pot-$(PACKAGE_VERSION).jar -jvpp_pot_package_dir = io/fd/vpp/jvpp/ioampot -pot_api_file=$(srcdir)/ioam/lib-pot/pot.api - - -lib_LTLIBRARIES += libjvpp_ioampot.la -libjvpp_ioampot_la_SOURCES = ioam/lib-pot/pot.api.h ioam/lib-pot/jvpp_ioam_pot.c ioam/jvpp/io_fd_vpp_jvpp_ioam_pot_JVppIoamPotImpl.h -libjvpp_ioampot_la_LIBADD = -lvlibmemoryclient -lvlibapi -lvppinfra \ - -lpthread -lm -lrt -L$(jvpp_registry_root)/.libs -ljvpp_common -libjvpp_ioampot_la_LDFLAGS = -module -libjvpp_ioampot_la_CPPFLAGS = -I$(JAVA_HOME)/include -I$(JAVA_HOME)/include/linux -I../ -I$(srcdir)/../ - -BUILT_SOURCES += $(jvpp_root)/io_fd_vpp_jvpp_ioampot_JVppIoampotImpl.h - -$(jvpp_root)/io_fd_vpp_jvpp_ioampot_JVppIoampotImpl.h: ioam_pot.api.json - dir=`pwd`; \ - mkdir -p $(jvpp_target); \ - mkdir -p $(jvpp_root)/$(jvpp_pot_package_dir); \ - cd $(jvpp_root)/$(jvpp_pot_package_dir); \ - mkdir -p dto future callfacade callback notification test; \ - @srcdir@/$(jvpp_registry_root)/jvpp/gen/jvpp_gen.py -i $${dir}/ioam_pot.api.json --plugin_name ioampot; \ - cd -; \ - mv -f $(jvpp_root)/$(jvpp_pot_package_dir)/jvpp_ioampot_gen.h $(jvpp_root)/jvpp_ioam_pot_gen.h; \ - cp $(srcdir)/$(jvpp_root)/$(jvpp_pot_package_dir)/test/*.java $(jvpp_root)/$(jvpp_pot_package_dir)/test/; \ - cd $(jvpp_root); \ - $(JAVAC) -classpath .:$(jvpp_target_dir):../../$(jvpp_registry_root)/jvpp-registry-$(jvpp_registry_version).jar -d $(jvpp_target_dir) $(jvpp_pot_package_dir)/*.java \ - $(jvpp_pot_package_dir)/dto/*.java \ - $(jvpp_pot_package_dir)/callback/*.java \ - $(jvpp_pot_package_dir)/notification/*.java \ - $(jvpp_pot_package_dir)/future/*.java \ - $(jvpp_pot_package_dir)/callfacade/*.java \ - $(jvpp_pot_package_dir)/test/*.java \ - || (echo "ioam pot jvpp compilation failed: $$?"; exit 1); \ - $(JAVAH) -classpath .:$(jvpp_target_dir):../../$(jvpp_registry_root)/jvpp-registry-$(jvpp_registry_version).jar -d . io.fd.vpp.jvpp.ioampot.JVppIoampotImpl ; - -$(jioam_pot_jarfile): libjvpp_ioampot.la - cp .libs/libjvpp_ioampot.so.0.0.0 $(jvpp_target); \ - cd $(jvpp_target); \ - $(JAR) cfv $(JARFLAGS) ../../../$@ libjvpp_ioampot.so.0.0.0 $(jvpp_pot_package_dir)/* ; cd ..; - -ioam_pot.api.json: - @echo " jIoam_pot API"; \ - vppapigen --input $(pot_api_file) --json ioam_pot.api.json; - -all-local: $(jioam_pot_jarfile) - -jioam_export_jarfile = jvpp-ioam-export-$(PACKAGE_VERSION).jar -jvpp_export_package_dir = io/fd/vpp/jvpp/ioamexport -export_api_file=$(srcdir)/ioam/export/ioam_export.api - - -lib_LTLIBRARIES += libjvpp_ioamexport.la -libjvpp_ioamexport_la_SOURCES = ioam/export/export.api.h ioam/export/jvpp_ioam_export.c ioam/jvpp/io_fd_vpp_jvpp_ioam_export_JVppIoamexportImpl.h -libjvpp_ioamexport_la_LIBADD = -lvlibmemoryclient -lvlibapi -lvppinfra \ - -lpthread -lm -lrt -L$(jvpp_registry_root)/.libs -ljvpp_common -libjvpp_ioamexport_la_LDFLAGS = -module -libjvpp_ioamexport_la_CPPFLAGS = -I$(JAVA_HOME)/include -I$(JAVA_HOME)/include/linux -I../ -I$(srcdir)/../ - -BUILT_SOURCES += $(jvpp_root)/io_fd_vpp_jvpp_ioamexport_JVppIoamexportImpl.h - -$(jvpp_root)/io_fd_vpp_jvpp_ioamexport_JVppIoamexportImpl.h: ioam_export.api.json - dir=`pwd`; \ - mkdir -p $(jvpp_target); \ - mkdir -p $(jvpp_root)/$(jvpp_export_package_dir); \ - cd $(jvpp_root)/$(jvpp_export_package_dir); \ - mkdir -p dto future callfacade callback notification test; \ - @srcdir@/$(jvpp_registry_root)/jvpp/gen/jvpp_gen.py -i $${dir}/ioam_export.api.json --plugin_name ioamexport; \ - cd -; \ - mv -f $(jvpp_root)/$(jvpp_export_package_dir)/jvpp_ioamexport_gen.h $(jvpp_root)/jvpp_ioam_export_gen.h; \ - cp $(srcdir)/$(jvpp_root)/$(jvpp_export_package_dir)/test/*.java $(jvpp_root)/$(jvpp_export_package_dir)/test/; \ - cd $(jvpp_root); \ - $(JAVAC) -classpath .:$(jvpp_target_dir):../../$(jvpp_registry_root)/jvpp-registry-$(jvpp_registry_version).jar -d $(jvpp_target_dir) $(jvpp_export_package_dir)/*.java \ - $(jvpp_export_package_dir)/dto/*.java \ - $(jvpp_export_package_dir)/callback/*.java \ - $(jvpp_export_package_dir)/notification/*.java \ - $(jvpp_export_package_dir)/future/*.java \ - $(jvpp_export_package_dir)/callfacade/*.java \ - $(jvpp_export_package_dir)/test/*.java \ - || (echo "ioam export jvpp compilation failed: $$?"; exit 1); \ - $(JAVAH) -classpath .:$(jvpp_target_dir):../../$(jvpp_registry_root)/jvpp-registry-$(jvpp_registry_version).jar -d . io.fd.vpp.jvpp.ioamexport.JVppIoamexportImpl ; - -$(jioam_export_jarfile): libjvpp_ioamexport.la - cp .libs/libjvpp_ioamexport.so.0.0.0 $(jvpp_target); \ - cd $(jvpp_target); \ - $(JAR) cfv $(JARFLAGS) ../../../$@ libjvpp_ioamexport.so.0.0.0 $(jvpp_export_package_dir)/* ; cd ..; - -ioam_export.api.json: - @echo " jIoam_export API"; \ - vppapigen --input $(export_api_file) --json ioam_export.api.json; - -all-local: $(jioam_export_jarfile) diff --git a/plugins/ioam-plugin/configure.ac b/plugins/ioam-plugin/configure.ac deleted file mode 100644 index 287bcaab..00000000 --- a/plugins/ioam-plugin/configure.ac +++ /dev/null @@ -1,25 +0,0 @@ -AC_INIT(ioam_plugin, 1.0) -LT_INIT -AC_CONFIG_MACRO_DIR([../../vpp-api/java/m4]) -AM_INIT_AUTOMAKE -AC_PREFIX_DEFAULT([/usr]) -AM_SILENT_RULES([yes]) - -AC_PROG_CC - -if test -f /usr/bin/lsb_release && test `lsb_release -si` == "Ubuntu" && test `lsb_release -sr` == "14.04" && test -d /usr/lib/jvm/java-8-openjdk-amd64/ ; then - JAVA_HOME=/usr/lib/jvm/java-8-openjdk-amd64/ - JAVAC=${JAVA_HOME}/bin/javac - PATH=${JAVA_HOME}/bin/:${PATH} - break -fi - -AX_CHECK_JAVA_HOME -AX_PROG_JAVAC -AX_PROG_JAVAH -AX_PROG_JAR -AX_PROG_JAVADOC -AX_PROG_JAVA - - -AC_OUTPUT([Makefile]) diff --git a/plugins/ioam-plugin/ioam/dir.dox b/plugins/ioam-plugin/ioam/dir.dox deleted file mode 100644 index f3389b52..00000000 --- a/plugins/ioam-plugin/ioam/dir.dox +++ /dev/null @@ -1,18 +0,0 @@ -/* - * Copyright (c) 2016 Cisco and/or its affiliates. - * Licensed under the Apache License, Version 2.0 (the "License"); - * you may not use this file except in compliance with the License. - * You may obtain a copy of the License at: - * - * http://www.apache.org/licenses/LICENSE-2.0 - * - * Unless required by applicable law or agreed to in writing, software - * distributed under the License is distributed on an "AS IS" BASIS, - * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. - * See the License for the specific language governing permissions and - * limitations under the License. - */ -/** - @dir - @brief Inband OAM (iOAM) implementation -*/ diff --git a/plugins/ioam-plugin/ioam/encap/ip6_ioam_e2e.c b/plugins/ioam-plugin/ioam/encap/ip6_ioam_e2e.c deleted file mode 100644 index 0839cdce..00000000 --- a/plugins/ioam-plugin/ioam/encap/ip6_ioam_e2e.c +++ /dev/null @@ -1,232 +0,0 @@ -/* - * Copyright (c) 2016 Cisco and/or its affiliates. - * Licensed under the Apache License, Version 2.0 (the "License"); - * you may not use this file except in compliance with the License. - * You may obtain a copy of the License at: - * - * http://www.apache.org/licenses/LICENSE-2.0 - * - * Unless required by applicable law or agreed to in writing, software - * distributed under the License is distributed on an "AS IS" BASIS, - * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. - * See the License for the specific language governing permissions and - * limitations under the License. - */ - -#include -#include -#include -#include - -#include - -#include -#include -#include - -#include -#include - -#include "ip6_ioam_e2e.h" - -ioam_e2e_main_t ioam_e2e_main; - -static u8 * ioam_e2e_trace_handler (u8 * s, - ip6_hop_by_hop_option_t *opt) -{ - ioam_e2e_option_t * e2e = (ioam_e2e_option_t *)opt; - u32 seqno = 0; - - if (e2e) - { - seqno = clib_net_to_host_u32 (e2e->e2e_data); - } - - s = format (s, "SeqNo = 0x%Lx", seqno); - return s; -} - -int -ioam_e2e_config_handler (void *data, u8 disable) -{ - int *analyse = data; - - /* Register hanlders if enabled */ - if (!disable) - { - /* If encap node register for encap handler */ - if (0 == *analyse) - { - if (ip6_hbh_register_option(HBH_OPTION_TYPE_IOAM_EDGE_TO_EDGE, - ioam_seqno_encap_handler, - ioam_e2e_trace_handler) < 0) - { - return (-1); - } - } - /* If analyze node then register for decap handler */ - else - { - if (ip6_hbh_pop_register_option(HBH_OPTION_TYPE_IOAM_EDGE_TO_EDGE, - ioam_seqno_decap_handler) < 0) - { - return (-1); - } - } - return 0; - } - - /* UnRegister handlers */ - (void) ip6_hbh_unregister_option(HBH_OPTION_TYPE_IOAM_EDGE_TO_EDGE); - (void) ip6_hbh_pop_unregister_option(HBH_OPTION_TYPE_IOAM_EDGE_TO_EDGE); - return 0; -} - -int -ioam_e2e_rewrite_handler (u8 *rewrite_string, - u8 *rewrite_size) -{ - ioam_e2e_option_t *e2e_option; - - if (rewrite_string && *rewrite_size == sizeof(ioam_e2e_option_t)) - { - e2e_option = (ioam_e2e_option_t *)rewrite_string; - e2e_option->hdr.type = HBH_OPTION_TYPE_IOAM_EDGE_TO_EDGE - | HBH_OPTION_TYPE_SKIP_UNKNOWN; - e2e_option->hdr.length = sizeof (ioam_e2e_option_t) - - sizeof (ip6_hop_by_hop_option_t); - return(0); - } - return(-1); -} - -u32 -ioam_e2e_flow_handler (u32 ctx, u8 add) -{ - ioam_e2e_data_t *data; - u16 i; - - if (add) - { - pool_get(ioam_e2e_main.e2e_data, data); - data->flow_ctx = ctx; - ioam_seqno_init_bitmap(&data->seqno_data); - return ((u32) (data - ioam_e2e_main.e2e_data)); - } - - /* Delete case */ - for (i = 0; i < vec_len(ioam_e2e_main.e2e_data); i++) - { - if (pool_is_free_index(ioam_e2e_main.e2e_data, i)) - continue; - - data = pool_elt_at_index(ioam_e2e_main.e2e_data, i); - if (data && (data->flow_ctx == ctx)) - { - pool_put_index(ioam_e2e_main.e2e_data, i); - return (0); - } - } - return 0; -} - -static clib_error_t * -ioam_show_e2e_cmd_fn (vlib_main_t * vm, - unformat_input_t * input, - vlib_cli_command_t * cmd) -{ - ioam_e2e_data_t *e2e_data; - u8 *s = 0; - int i; - - vec_reset_length(s); - - s = format(0, "IOAM E2E information: \n"); - for (i = 0; i < vec_len(ioam_e2e_main.e2e_data); i++) - { - if (pool_is_free_index(ioam_e2e_main.e2e_data, i)) - continue; - - e2e_data = pool_elt_at_index(ioam_e2e_main.e2e_data, i); - s = format(s, "Flow name: %s\n", get_flow_name_from_flow_ctx(e2e_data->flow_ctx)); - - s = show_ioam_seqno_cmd_fn(s, - &e2e_data->seqno_data, - !IOAM_DEAP_ENABLED(e2e_data->flow_ctx)); - } - - vlib_cli_output(vm, "%v", s); - return 0; -} - - -VLIB_CLI_COMMAND (ioam_show_e2e_cmd, static) = { - .path = "show ioam e2e ", - .short_help = "show ioam e2e information", - .function = ioam_show_e2e_cmd_fn, -}; - -/* - * This routine exists to convince the vlib plugin framework that - * we haven't accidentally copied a random .dll into the plugin directory. - * - * Also collects global variable pointers passed from the vpp engine - */ -clib_error_t * -vlib_plugin_register (vlib_main_t * vm, vnet_plugin_handoff_t * h, - int from_early_init) -{ - clib_error_t * error = 0; - - ioam_e2e_main.vlib_main = vm; - ioam_e2e_main.vnet_main = h->vnet_main; - return error; -} - -/* - * Init handler E2E headet handling. - * Init hanlder registers encap, decap, trace and Rewrite handlers. - */ -static clib_error_t * -ioam_e2e_init (vlib_main_t * vm) -{ - clib_error_t * error; - - if ((error = vlib_call_init_function (vm, ip6_hop_by_hop_ioam_init))) - { - return(error); - } - - /* - * As of now we have only PPC under E2E header. - */ - if (ip6_hbh_config_handler_register(HBH_OPTION_TYPE_IOAM_EDGE_TO_EDGE, - ioam_e2e_config_handler) < 0) - { - return (clib_error_create("Registration of " - "HBH_OPTION_TYPE_IOAM_EDGE_TO_EDGE for rewrite failed")); - } - - if (ip6_hbh_add_register_option(HBH_OPTION_TYPE_IOAM_EDGE_TO_EDGE, - sizeof(ioam_e2e_option_t), - ioam_e2e_rewrite_handler) < 0) - { - return (clib_error_create("Registration of " - "HBH_OPTION_TYPE_IOAM_EDGE_TO_EDGE for rewrite failed")); - } - - if (ip6_hbh_flow_handler_register(HBH_OPTION_TYPE_IOAM_EDGE_TO_EDGE, - ioam_e2e_flow_handler) < 0) - { - return (clib_error_create("Registration of " - "HBH_OPTION_TYPE_IOAM_EDGE_TO_EDGE Flow handler failed")); - } - - return (0); -} - -/* - * Init function for the E2E lib. - * ip6_hop_by_hop_ioam_e2e_init gets called during init. - */ -VLIB_INIT_FUNCTION (ioam_e2e_init); diff --git a/plugins/ioam-plugin/ioam/encap/ip6_ioam_e2e.h b/plugins/ioam-plugin/ioam/encap/ip6_ioam_e2e.h deleted file mode 100644 index 18f35f80..00000000 --- a/plugins/ioam-plugin/ioam/encap/ip6_ioam_e2e.h +++ /dev/null @@ -1,47 +0,0 @@ -/* - * Copyright (c) 2016 Cisco and/or its affiliates. - * Licensed under the Apache License, Version 2.0 (the "License"); - * you may not use this file except in compliance with the License. - * You may obtain a copy of the License at: - * - * http://www.apache.org/licenses/LICENSE-2.0 - * - * Unless required by applicable law or agreed to in writing, software - * distributed under the License is distributed on an "AS IS" BASIS, - * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. - * See the License for the specific language governing permissions and - * limitations under the License. - */ - -#ifndef __included_ip6_ioam_e2e_h__ -#define __included_ip6_ioam_e2e_h__ - -#include "ip6_ioam_seqno.h" - -typedef struct ioam_e2e_data_t_ { - u32 flow_ctx; - u32 pad; - ioam_seqno_data seqno_data; -} ioam_e2e_data_t; - -typedef struct { - ioam_e2e_data_t *e2e_data; - vlib_main_t *vlib_main; - vnet_main_t *vnet_main; -} ioam_e2e_main_t; - -extern ioam_e2e_main_t ioam_e2e_main; - -static inline ioam_seqno_data * -ioam_e2ec_get_seqno_data_from_flow_ctx (u32 flow_ctx) -{ - ioam_e2e_data_t *data = NULL; - u32 index; - - index = get_flow_data_from_flow_ctx(flow_ctx, - HBH_OPTION_TYPE_IOAM_EDGE_TO_EDGE); - data = &ioam_e2e_main.e2e_data[index]; - return &(data->seqno_data); -} - -#endif /* __included_ioam_e2e_h__ */ diff --git a/plugins/ioam-plugin/ioam/encap/ip6_ioam_pot.c b/plugins/ioam-plugin/ioam/encap/ip6_ioam_pot.c deleted file mode 100644 index 05f42c91..00000000 --- a/plugins/ioam-plugin/ioam/encap/ip6_ioam_pot.c +++ /dev/null @@ -1,276 +0,0 @@ -/* - * Copyright (c) 2016 Cisco and/or its affiliates. - * Licensed under the Apache License, Version 2.0 (the "License"); - * you may not use this file except in compliance with the License. - * You may obtain a copy of the License at: - * - * http://www.apache.org/licenses/LICENSE-2.0 - * - * Unless required by applicable law or agreed to in writing, software - * distributed under the License is distributed on an "AS IS" BASIS, - * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. - * See the License for the specific language governing permissions and - * limitations under the License. - */ -#include -#include -#include -#include - -#include -#include -#include - -#include -#include -#include - -#include - -typedef CLIB_PACKED(struct { - ip6_hop_by_hop_option_t hdr; - u8 pot_type; -#define PROFILE_ID_MASK 0xF - u8 reserved_profile_id; /* 4 bits reserved, 4 bits to carry profile id */ - u64 random; - u64 cumulative; -}) ioam_pot_option_t; - -#define foreach_ip6_hop_by_hop_ioam_pot_stats \ - _(PROCESSED, "Pkts with ip6 hop-by-hop pot options") \ - _(PROFILE_MISS, "Pkts with ip6 hop-by-hop pot options but no profile set") \ - _(PASSED, "Pkts with POT in Policy") \ - _(FAILED, "Pkts with POT out of Policy") - -static char * ip6_hop_by_hop_ioam_pot_stats_strings[] = { -#define _(sym,string) string, - foreach_ip6_hop_by_hop_ioam_pot_stats -#undef _ -}; - -typedef enum { -#define _(sym,str) IP6_IOAM_POT_##sym, - foreach_ip6_hop_by_hop_ioam_pot_stats -#undef _ - IP6_IOAM_POT_N_STATS, -} ip6_ioam_pot_stats_t; - -typedef struct { - /* stats */ - u64 counters[ARRAY_LEN(ip6_hop_by_hop_ioam_pot_stats_strings)]; - - /* convenience */ - vlib_main_t * vlib_main; - vnet_main_t * vnet_main; -} ip6_hop_by_hop_ioam_pot_main_t; - -ip6_hop_by_hop_ioam_pot_main_t ip6_hop_by_hop_ioam_pot_main; - -always_inline void -ip6_ioam_stats_increment_counter (u32 counter_index, u64 increment) -{ - ip6_hop_by_hop_ioam_pot_main_t *hm = &ip6_hop_by_hop_ioam_pot_main; - - hm->counters[counter_index] += increment; -} - - -static u8 * format_ioam_pot (u8 * s, va_list * args) -{ - ioam_pot_option_t * pot0 = va_arg (*args, ioam_pot_option_t *); - u64 random, cumulative; - random = cumulative = 0; - if (pot0) - { - random = clib_net_to_host_u64 (pot0->random); - cumulative = clib_net_to_host_u64 (pot0->cumulative); - } - - s = format (s, "random = 0x%Lx, Cumulative = 0x%Lx, Index = 0x%x", - random, cumulative, pot0 ? pot0->reserved_profile_id : ~0); - return s; -} - -u8 * -ip6_hbh_ioam_proof_of_transit_trace_handler (u8 *s, ip6_hop_by_hop_option_t *opt) -{ - ioam_pot_option_t *pot; - - s = format (s, " POT opt present\n"); - pot = (ioam_pot_option_t *) opt; - s = format (s, " %U\n", format_ioam_pot, pot); - return (s); -} - -int -ip6_hbh_ioam_proof_of_transit_handler (vlib_buffer_t *b, - ip6_header_t *ip, - ip6_hop_by_hop_option_t *opt0) -{ - ioam_pot_option_t * pot0; - u64 random = 0, cumulative = 0; - int rv = 0; - u8 pot_profile_index; - pot_profile *pot_profile = 0, *new_profile = 0; - u8 pot_encap = 0; - - pot0 = (ioam_pot_option_t *) opt0; - pot_encap = (pot0->random == 0); - pot_profile_index = pot_profile_get_active_id(); - pot_profile = pot_profile_get_active(); - if (pot_encap && PREDICT_FALSE(!pot_profile)) - { - ip6_ioam_stats_increment_counter (IP6_IOAM_POT_PROFILE_MISS, 1); - return(-1); - } - if (pot_encap) - { - pot0->reserved_profile_id = - pot_profile_index & PROFILE_ID_MASK; - pot_profile_incr_usage_stats(pot_profile); - } - else - { /* Non encap node */ - if (PREDICT_FALSE(pot0->reserved_profile_id != - pot_profile_index || pot_profile == 0)) - { - /* New profile announced by encap node. */ - new_profile = - pot_profile_find(pot0->reserved_profile_id); - if (PREDICT_FALSE(new_profile == 0 || - new_profile->valid == 0)) - { - ip6_ioam_stats_increment_counter (IP6_IOAM_POT_PROFILE_MISS, 1); - return(-1); - } - else - { - pot_profile_index = pot0->reserved_profile_id; - pot_profile = new_profile; - pot_profile_set_active(pot_profile_index); - pot_profile_reset_usage_stats(pot_profile); - } - } - pot_profile_incr_usage_stats(pot_profile); - } - - if (pot0->random == 0) - { - pot0->random = clib_host_to_net_u64(pot_generate_random(pot_profile)); - pot0->cumulative = 0; - } - random = clib_net_to_host_u64(pot0->random); - cumulative = clib_net_to_host_u64(pot0->cumulative); - pot0->cumulative = clib_host_to_net_u64( - pot_update_cumulative(pot_profile, - cumulative, - random)); - ip6_ioam_stats_increment_counter (IP6_IOAM_POT_PROCESSED, 1); - - return (rv); -} - -int -ip6_hbh_ioam_proof_of_transit_pop_handler (vlib_buffer_t *b, ip6_header_t *ip, - ip6_hop_by_hop_option_t *opt0) -{ - ioam_pot_option_t * pot0; - u64 random = 0; - u64 cumulative = 0; - int rv = 0; - pot_profile *pot_profile = 0; - u8 result = 0; - - pot0 = (ioam_pot_option_t *) opt0; - random = clib_net_to_host_u64(pot0->random); - cumulative = clib_net_to_host_u64(pot0->cumulative); - pot_profile = pot_profile_get_active(); - result = pot_validate (pot_profile, - cumulative, random); - - if (result == 1) - { - ip6_ioam_stats_increment_counter (IP6_IOAM_POT_PASSED, 1); - } - else - { - ip6_ioam_stats_increment_counter (IP6_IOAM_POT_FAILED, 1); - } - return (rv); -} - -int ip6_hop_by_hop_ioam_pot_rewrite_handler (u8 *rewrite_string, u8 *rewrite_size) -{ - ioam_pot_option_t * pot_option; - if (rewrite_string && *rewrite_size == sizeof(ioam_pot_option_t)) - { - pot_option = (ioam_pot_option_t *)rewrite_string; - pot_option->hdr.type = HBH_OPTION_TYPE_IOAM_PROOF_OF_TRANSIT - | HBH_OPTION_TYPE_DATA_CHANGE_ENROUTE; - pot_option->hdr.length = sizeof (ioam_pot_option_t) - - sizeof (ip6_hop_by_hop_option_t); - return(0); - } - return(-1); -} - -static clib_error_t * -ip6_show_ioam_pot_cmd_fn (vlib_main_t * vm, - unformat_input_t * input, - vlib_cli_command_t * cmd) -{ - ip6_hop_by_hop_ioam_pot_main_t *hm = &ip6_hop_by_hop_ioam_pot_main; - u8 *s = 0; - int i = 0; - - for ( i = 0; i < IP6_IOAM_POT_N_STATS; i++) - { - s = format(s, " %s - %lu\n", ip6_hop_by_hop_ioam_pot_stats_strings[i], - hm->counters[i]); - } - - vlib_cli_output(vm, "%v", s); - vec_free(s); - return 0; -} - - -VLIB_CLI_COMMAND (ip6_show_ioam_pot_cmd, static) = { - .path = "show ioam pot", - .short_help = "iOAM pot statistics", - .function = ip6_show_ioam_pot_cmd_fn, -}; - - -static clib_error_t * -ip6_hop_by_hop_ioam_pot_init (vlib_main_t * vm) -{ - ip6_hop_by_hop_ioam_pot_main_t * hm = &ip6_hop_by_hop_ioam_pot_main; - clib_error_t * error; - - if ((error = vlib_call_init_function (vm, ip6_hop_by_hop_ioam_init))) - return(error); - - hm->vlib_main = vm; - hm->vnet_main = vnet_get_main(); - memset(hm->counters, 0, sizeof(hm->counters)); - - if (ip6_hbh_register_option(HBH_OPTION_TYPE_IOAM_PROOF_OF_TRANSIT, ip6_hbh_ioam_proof_of_transit_handler, - ip6_hbh_ioam_proof_of_transit_trace_handler) < 0) - return (clib_error_create("registration of HBH_OPTION_TYPE_IOAM_PROOF_OF_TRANSIT failed")); - - if (ip6_hbh_add_register_option(HBH_OPTION_TYPE_IOAM_PROOF_OF_TRANSIT, - sizeof(ioam_pot_option_t), - ip6_hop_by_hop_ioam_pot_rewrite_handler) < 0) - return (clib_error_create("registration of HBH_OPTION_TYPE_IOAM_PROOF_OF_TRANSIT for rewrite failed")); - - if (ip6_hbh_pop_register_option(HBH_OPTION_TYPE_IOAM_PROOF_OF_TRANSIT, - ip6_hbh_ioam_proof_of_transit_pop_handler) < 0) - return (clib_error_create("registration of HBH_OPTION_TYPE_IOAM_PROOF_OF_TRANSIT POP failed")); - - return (0); -} - -VLIB_INIT_FUNCTION (ip6_hop_by_hop_ioam_pot_init); - - diff --git a/plugins/ioam-plugin/ioam/encap/ip6_ioam_seqno.c b/plugins/ioam-plugin/ioam/encap/ip6_ioam_seqno.c deleted file mode 100644 index 0b4d4192..00000000 --- a/plugins/ioam-plugin/ioam/encap/ip6_ioam_seqno.c +++ /dev/null @@ -1,109 +0,0 @@ -/* - * Copyright (c) 2016 Cisco and/or its affiliates. - * Licensed under the Apache License, Version 2.0 (the "License"); - * you may not use this file except in compliance with the License. - * You may obtain a copy of the License at: - * - * http://www.apache.org/licenses/LICENSE-2.0 - * - * Unless required by applicable law or agreed to in writing, software - * distributed under the License is distributed on an "AS IS" BASIS, - * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. - * See the License for the specific language governing permissions and - * limitations under the License. - */ - - -#include -#include -#include -#include - -#include - -#include -#include -#include - -#include "ip6_ioam_seqno.h" -#include "ip6_ioam_e2e.h" - -ioam_seqno_data_main_t ioam_seqno_main; - -void ioam_seqno_init_bitmap (ioam_seqno_data *data) -{ - seqno_bitmap *bitmap = &data->seqno_rx.bitmap; - bitmap->window_size = SEQNO_WINDOW_SIZE; - bitmap->array_size = SEQNO_WINDOW_ARRAY_SIZE; - bitmap->mask = 32 * SEQNO_WINDOW_ARRAY_SIZE - 1; - bitmap->array[0] = 0x00000000;/* pretend we haven seen sequence numbers 0*/ - bitmap->highest = 0; - - data->seq_num = 0; - return ; -} - -/* - * This Routine gets called from IPv6 hop-by-hop option handling. - * Only if we are encap node, then add PPC data. - * On a Transit(MID) node we dont do anything with E2E headers. - * On decap node decap is handled by seperate function. - */ -int -ioam_seqno_encap_handler (vlib_buffer_t *b, ip6_header_t *ip, - ip6_hop_by_hop_option_t *opt) -{ - u32 opaque_index = vnet_buffer(b)->l2_classify.opaque_index; - ioam_e2e_option_t * e2e; - int rv = 0; - ioam_seqno_data *data; - - data = ioam_e2ec_get_seqno_data_from_flow_ctx(opaque_index); - e2e = (ioam_e2e_option_t *) opt; - e2e->e2e_data = clib_host_to_net_u32(++data->seq_num); - - return (rv); -} - -/* - * This Routine gets called on POP/Decap node. - */ -int -ioam_seqno_decap_handler (vlib_buffer_t *b, ip6_header_t *ip, - ip6_hop_by_hop_option_t *opt) -{ - u32 opaque_index = vnet_buffer(b)->l2_classify.opaque_index; - ioam_e2e_option_t * e2e; - int rv = 0; - ioam_seqno_data *data; - - data = ioam_e2ec_get_seqno_data_from_flow_ctx(opaque_index); - e2e = (ioam_e2e_option_t *) opt; - ioam_analyze_seqno(&data->seqno_rx, (u64) clib_net_to_host_u32(e2e->e2e_data)); - - return (rv); -} - -u8 * -show_ioam_seqno_cmd_fn (u8 *s, ioam_seqno_data *seqno_data, u8 enc) -{ - seqno_rx_info *rx; - - s = format(s, "SeqNo Data:\n"); - if (enc) - { - s = format(s, " Current Seq. Number : %llu\n", seqno_data->seq_num); - } - else - { - rx = &seqno_data->seqno_rx; - s = format(s, " Highest Seq. Number : %llu\n", rx->bitmap.highest); - s = format(s, " Packets received : %llu\n", rx->rx_packets); - s = format(s, " Lost packets : %llu\n", rx->lost_packets); - s = format(s, " Reordered packets : %llu\n", rx->reordered_packets); - s = format(s, " Duplicate packets : %llu\n", rx->dup_packets); - } - - format(s, "\n"); - return s; -} diff --git a/plugins/ioam-plugin/ioam/encap/ip6_ioam_seqno.h b/plugins/ioam-plugin/ioam/encap/ip6_ioam_seqno.h deleted file mode 100644 index 13a84db0..00000000 --- a/plugins/ioam-plugin/ioam/encap/ip6_ioam_seqno.h +++ /dev/null @@ -1,70 +0,0 @@ -/* - * Copyright (c) 2016 Cisco and/or its affiliates. - * Licensed under the Apache License, Version 2.0 (the "License"); - * you may not use this file except in compliance with the License. - * You may obtain a copy of the License at: - * - * http://www.apache.org/licenses/LICENSE-2.0 - * - * Unless required by applicable law or agreed to in writing, software - * distributed under the License is distributed on an "AS IS" BASIS, - * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. - * See the License for the specific language governing permissions and - * limitations under the License. - */ - -#ifndef __included_ip6_ioam_seqno_h__ -#define __included_ip6_ioam_seqno_h__ - -#include -#include - -#define SEQ_CHECK_VALUE 0x80000000 /* for seq number wraparound detection */ - -#define SEQNO_WINDOW_SIZE 2048 -#define SEQNO_WINDOW_ARRAY_SIZE 64 - -typedef struct seqno_bitmap_ { - u32 window_size; - u32 array_size; - u32 mask; - u32 pad; - u64 highest; - u64 array[SEQNO_WINDOW_ARRAY_SIZE]; /* Will be alloc to array_size */ -} seqno_bitmap; - -typedef struct seqno_rx_info_ { - u64 rx_packets; - u64 lost_packets; - u64 reordered_packets; - u64 dup_packets; - seqno_bitmap bitmap; -} seqno_rx_info; - -/* This structure is 64-byte aligned */ -typedef struct ioam_seqno_data_ { - union { - u32 seq_num; /* Useful only for encap node */ - seqno_rx_info seqno_rx; - }; -} ioam_seqno_data; - -typedef struct ioam_seqno_data_main_t_ { - ioam_seqno_data *seqno_data; -} ioam_seqno_data_main_t; - -void ioam_seqno_init_bitmap(ioam_seqno_data *data); - -int ioam_seqno_encap_handler(vlib_buffer_t *b, ip6_header_t *ip, - ip6_hop_by_hop_option_t *opt); - -int -ioam_seqno_decap_handler(vlib_buffer_t *b, ip6_header_t *ip, - ip6_hop_by_hop_option_t *opt); - -void ioam_analyze_seqno(seqno_rx_info *ppc_rx, u64 seqno); - -u8 * -show_ioam_seqno_cmd_fn(u8 *s, ioam_seqno_data *seqno_data, u8 enc); - -#endif diff --git a/plugins/ioam-plugin/ioam/encap/ip6_ioam_seqno_analyse.c b/plugins/ioam-plugin/ioam/encap/ip6_ioam_seqno_analyse.c deleted file mode 100644 index 4638871c..00000000 --- a/plugins/ioam-plugin/ioam/encap/ip6_ioam_seqno_analyse.c +++ /dev/null @@ -1,141 +0,0 @@ -/* - * Copyright (c) 2016 Cisco and/or its affiliates. - * Licensed under the Apache License, Version 2.0 (the "License"); - * you may not use this file except in compliance with the License. - * You may obtain a copy of the License at: - * - * http://www.apache.org/licenses/LICENSE-2.0 - * - * Unless required by applicable law or agreed to in writing, software - * distributed under the License is distributed on an "AS IS" BASIS, - * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. - * See the License for the specific language governing permissions and - * limitations under the License. - */ - -#include -#include "ip6_ioam_seqno.h" - -static inline void BIT_SET (u64 *p, u32 n) -{ - p[ n>>5 ] |= (1 << (n&31)); -} - -static inline int BIT_TEST (u64 *p, u32 n) -{ - return p[ n>>5 ] & (1 << (n&31)); -} - -static void BIT_CLEAR (u64 *p, u64 start, int num_bits, u32 mask) -{ - int n, t; - int start_index = (start >> 5); - int mask_index = (mask >> 5); - - start_index &= mask_index; - if (start & 0x1f) - { - int start_bit = (start & 0x1f); - - n = (1 << start_bit)-1; - t = start_bit + num_bits; - if (t < 32) - { - n |= ~((1 << t)-1); - p[ start_index ] &= n; - return; - } - p[ start_index ] &= n; - start_index = (start_index + 1) & mask_index; - num_bits -= (32 - start_bit); - } - while (num_bits >= 32) - { - p[ start_index ] = 0; - start_index = (start_index + 1) & mask_index; - num_bits -= 32; - } - n = ~((1 << num_bits) - 1); - p[ start_index ] &= n; -} - -static inline u8 seqno_check_wraparound(u32 a, u32 b) -{ - if ((a != b) && (a > b) && ((a - b) > SEQ_CHECK_VALUE)) - { - return 1; - } - return 0; -} - -/* - * Function to analyze the PPC value recevied. - * - Updates the bitmap with received sequence number - * - counts the received/lost/duplicate/reordered packets - */ -void ioam_analyze_seqno (seqno_rx_info *seqno_rx, u64 seqno) -{ - int diff; - static int peer_dead_count; - seqno_bitmap *bitmap = &seqno_rx->bitmap; - - seqno_rx->rx_packets++; - - if (seqno > bitmap->highest) - { /* new larger sequence number */ - peer_dead_count = 0; - diff = seqno - bitmap->highest; - if (diff < bitmap->window_size) - { - if (diff > 1) - { /* diff==1 is *such* a common case it's a win to optimize it */ - BIT_CLEAR(bitmap->array, bitmap->highest+1, diff-1, bitmap->mask); - seqno_rx->lost_packets += diff -1; - } - } - else - { - seqno_rx->lost_packets += diff -1; - memset( bitmap->array, 0, bitmap->array_size * sizeof(u64) ); - } - BIT_SET(bitmap->array, seqno & bitmap->mask); - bitmap->highest = seqno; - return; - } - - /* we've seen a bigger seq number before */ - diff = bitmap->highest - seqno; - if (diff >= bitmap->window_size) - { - if (seqno_check_wraparound(bitmap->highest, seqno)) - { - memset( bitmap->array, 0, bitmap->array_size * sizeof(u64)); - BIT_SET(bitmap->array, seqno & bitmap->mask); - bitmap->highest = seqno; - return; - } - else - { - peer_dead_count++; - if (peer_dead_count > 25) - { - peer_dead_count = 0; - memset( bitmap->array, 0, bitmap->array_size * sizeof(u64) ); - BIT_SET(bitmap->array, seqno & bitmap->mask); - bitmap->highest = seqno; - } - //ppc_rx->reordered_packets++; - } - return; - } - - if (BIT_TEST(bitmap->array, seqno & bitmap->mask)) - { - seqno_rx->dup_packets++; - return; /* Already seen */ - } - seqno_rx->reordered_packets++; - seqno_rx->lost_packets--; - BIT_SET(bitmap->array, seqno & bitmap->mask); - return; -} diff --git a/plugins/ioam-plugin/ioam/encap/ip6_ioam_trace.c b/plugins/ioam-plugin/ioam/encap/ip6_ioam_trace.c deleted file mode 100644 index e63db6e4..00000000 --- a/plugins/ioam-plugin/ioam/encap/ip6_ioam_trace.c +++ /dev/null @@ -1,438 +0,0 @@ -/* - * Copyright (c) 2016 Cisco and/or its affiliates. - * Licensed under the Apache License, Version 2.0 (the "License"); - * you may not use this file except in compliance with the License. - * You may obtain a copy of the License at: - * - * http://www.apache.org/licenses/LICENSE-2.0 - * - * Unless required by applicable law or agreed to in writing, software - * distributed under the License is distributed on an "AS IS" BASIS, - * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. - * See the License for the specific language governing permissions and - * limitations under the License. - */ -#include -#include -#include -#include - -#include -#include -#include - -#include -#include -#include - -#include - -/* Timestamp precision multipliers for seconds, milliseconds, microseconds - * and nanoseconds respectively. - */ -static f64 trace_tsp_mul[4] = { 1, 1e3, 1e6, 1e9 }; - -typedef union -{ - u64 as_u64; - u32 as_u32[2]; -} time_u64_t; - -/* *INDENT-OFF* */ -typedef CLIB_PACKED(struct { - ip6_hop_by_hop_option_t hdr; - u8 ioam_trace_type; - u8 data_list_elts_left; - u32 elts[0]; /* Variable type. So keep it generic */ -}) ioam_trace_option_t; -/* *INDENT-ON* */ - - -extern ip6_hop_by_hop_ioam_main_t ip6_hop_by_hop_ioam_main; -extern ip6_main_t ip6_main; - -#define foreach_ip6_hop_by_hop_ioam_trace_stats \ - _(PROCESSED, "Pkts with ip6 hop-by-hop trace options") \ - _(PROFILE_MISS, "Pkts with ip6 hop-by-hop trace options but no profile set") \ - _(UPDATED, "Pkts with trace updated") \ - _(FULL, "Pkts with trace options but no space") - -static char *ip6_hop_by_hop_ioam_trace_stats_strings[] = { -#define _(sym,string) string, - foreach_ip6_hop_by_hop_ioam_trace_stats -#undef _ -}; - -typedef enum -{ -#define _(sym,str) IP6_IOAM_TRACE_##sym, - foreach_ip6_hop_by_hop_ioam_trace_stats -#undef _ - IP6_IOAM_TRACE_N_STATS, -} ip6_ioam_trace_stats_t; - - -typedef struct -{ - /* stats */ - u64 counters[ARRAY_LEN (ip6_hop_by_hop_ioam_trace_stats_strings)]; - - /* convenience */ - vlib_main_t *vlib_main; - vnet_main_t *vnet_main; -} ip6_hop_by_hop_ioam_trace_main_t; - -ip6_hop_by_hop_ioam_trace_main_t ip6_hop_by_hop_ioam_trace_main; - -always_inline void -ip6_ioam_trace_stats_increment_counter (u32 counter_index, u64 increment) -{ - ip6_hop_by_hop_ioam_trace_main_t *hm = &ip6_hop_by_hop_ioam_trace_main; - - hm->counters[counter_index] += increment; -} - - -static u8 * -format_ioam_data_list_element (u8 * s, va_list * args) -{ - u32 *elt = va_arg (*args, u32 *); - u8 *trace_type_p = va_arg (*args, u8 *); - u8 trace_type = *trace_type_p; - - - if (trace_type & BIT_TTL_NODEID) - { - u32 ttl_node_id_host_byte_order = clib_net_to_host_u32 (*elt); - s = format (s, "ttl 0x%x node id 0x%x ", - ttl_node_id_host_byte_order >> 24, - ttl_node_id_host_byte_order & 0x00FFFFFF); - - elt++; - } - - if (trace_type & BIT_ING_INTERFACE && trace_type & BIT_ING_INTERFACE) - { - u32 ingress_host_byte_order = clib_net_to_host_u32 (*elt); - s = format (s, "ingress 0x%x egress 0x%x ", - ingress_host_byte_order >> 16, - ingress_host_byte_order & 0xFFFF); - elt++; - } - - if (trace_type & BIT_TIMESTAMP) - { - u32 ts_in_host_byte_order = clib_net_to_host_u32 (*elt); - s = format (s, "ts 0x%x \n", ts_in_host_byte_order); - elt++; - } - - if (trace_type & BIT_APPDATA) - { - u32 appdata_in_host_byte_order = clib_net_to_host_u32 (*elt); - s = format (s, "app 0x%x ", appdata_in_host_byte_order); - elt++; - } - - return s; -} - - -int -ip6_ioam_trace_get_sizeof_handler (u32 * result) -{ - u16 size = 0; - u8 trace_data_size = 0; - trace_profile *profile = NULL; - - *result = 0; - - profile = trace_profile_find (); - - if (PREDICT_FALSE (!profile)) - { - ip6_ioam_trace_stats_increment_counter (IP6_IOAM_TRACE_PROFILE_MISS, 1); - return (-1); - } - - trace_data_size = fetch_trace_data_size (profile->trace_type); - if (PREDICT_FALSE (trace_data_size == 0)) - return VNET_API_ERROR_INVALID_VALUE; - - if (PREDICT_FALSE (profile->num_elts * trace_data_size > 254)) - return VNET_API_ERROR_INVALID_VALUE; - - size += - sizeof (ioam_trace_option_t) + (profile->num_elts * trace_data_size); - *result = size; - - return 0; -} - - - -int -ip6_hop_by_hop_ioam_trace_rewrite_handler (u8 * rewrite_string, - u8 * rewrite_size) -{ - ioam_trace_option_t *trace_option = NULL; - u8 trace_data_size = 0; - u8 trace_option_elts = 0; - trace_profile *profile = NULL; - - - profile = trace_profile_find (); - - if (PREDICT_FALSE (!profile)) - { - ip6_ioam_trace_stats_increment_counter (IP6_IOAM_TRACE_PROFILE_MISS, 1); - return (-1); - } - - if (PREDICT_FALSE (!rewrite_string)) - return -1; - - trace_option_elts = profile->num_elts; - trace_data_size = fetch_trace_data_size (profile->trace_type); - trace_option = (ioam_trace_option_t *) rewrite_string; - trace_option->hdr.type = HBH_OPTION_TYPE_IOAM_TRACE_DATA_LIST | - HBH_OPTION_TYPE_DATA_CHANGE_ENROUTE; - trace_option->hdr.length = 2 /*ioam_trace_type,data_list_elts_left */ + - trace_option_elts * trace_data_size; - trace_option->ioam_trace_type = profile->trace_type & TRACE_TYPE_MASK; - trace_option->data_list_elts_left = trace_option_elts; - *rewrite_size = - sizeof (ioam_trace_option_t) + (trace_option_elts * trace_data_size); - - return 0; -} - - -int -ip6_hbh_ioam_trace_data_list_handler (vlib_buffer_t * b, ip6_header_t * ip, - ip6_hop_by_hop_option_t * opt) -{ - ip6_main_t *im = &ip6_main; - ip_lookup_main_t *lm = &im->lookup_main; - ip6_hop_by_hop_ioam_main_t *hm = &ip6_hop_by_hop_ioam_main; - u8 elt_index = 0; - ioam_trace_option_t *trace = (ioam_trace_option_t *) opt; - u32 adj_index = vnet_buffer (b)->ip.adj_index[VLIB_TX]; - ip_adjacency_t *adj = ip_get_adjacency (lm, adj_index); - time_u64_t time_u64; - u32 *elt; - int rv = 0; - trace_profile *profile = NULL; - - - profile = trace_profile_find (); - - if (PREDICT_FALSE (!profile)) - { - ip6_ioam_trace_stats_increment_counter (IP6_IOAM_TRACE_PROFILE_MISS, 1); - return (-1); - } - - - time_u64.as_u64 = 0; - - if (PREDICT_TRUE (trace->data_list_elts_left)) - { - trace->data_list_elts_left--; - /* fetch_trace_data_size returns in bytes. Convert it to 4-bytes - * to skip to this node's location. - */ - elt_index = - trace->data_list_elts_left * - fetch_trace_data_size (trace->ioam_trace_type) / 4; - elt = &trace->elts[elt_index]; - if (trace->ioam_trace_type & BIT_TTL_NODEID) - { - *elt = - clib_host_to_net_u32 ((ip->hop_limit << 24) | profile->node_id); - elt++; - } - - if (trace->ioam_trace_type & BIT_ING_INTERFACE) - { - *elt = - (vnet_buffer (b)->sw_if_index[VLIB_RX] & 0xFFFF) << 16 | - (adj->rewrite_header.sw_if_index & 0xFFFF); - *elt = clib_host_to_net_u32 (*elt); - elt++; - } - - if (trace->ioam_trace_type & BIT_TIMESTAMP) - { - /* Send least significant 32 bits */ - f64 time_f64 = - (f64) (((f64) hm->unix_time_0) + - (vlib_time_now (hm->vlib_main) - hm->vlib_time_0)); - - time_u64.as_u64 = time_f64 * trace_tsp_mul[profile->trace_tsp]; - *elt = clib_host_to_net_u32 (time_u64.as_u32[0]); - elt++; - } - - if (trace->ioam_trace_type & BIT_APPDATA) - { - /* $$$ set elt0->app_data */ - *elt = clib_host_to_net_u32 (profile->app_data); - elt++; - } - ip6_ioam_trace_stats_increment_counter (IP6_IOAM_TRACE_UPDATED, 1); - } - else - { - ip6_ioam_trace_stats_increment_counter (IP6_IOAM_TRACE_FULL, 1); - } - return (rv); -} - -u8 * -ip6_hbh_ioam_trace_data_list_trace_handler (u8 * s, - ip6_hop_by_hop_option_t * opt) -{ - ioam_trace_option_t *trace; - u8 trace_data_size_in_words = 0; - u32 *elt; - int elt_index = 0; - - trace = (ioam_trace_option_t *) opt; - s = - format (s, " Trace Type 0x%x , %d elts left\n", trace->ioam_trace_type, - trace->data_list_elts_left); - trace_data_size_in_words = - fetch_trace_data_size (trace->ioam_trace_type) / 4; - elt = &trace->elts[0]; - while ((u8 *) elt < ((u8 *) (&trace->elts[0]) + trace->hdr.length - 2 - /* -2 accounts for ioam_trace_type,elts_left */ )) - { - s = format (s, " [%d] %U\n", elt_index, - format_ioam_data_list_element, - elt, &trace->ioam_trace_type); - elt_index++; - elt += trace_data_size_in_words; - } - return (s); -} - - -static clib_error_t * -ip6_show_ioam_trace_cmd_fn (vlib_main_t * vm, - unformat_input_t * input, - vlib_cli_command_t * cmd) -{ - ip6_hop_by_hop_ioam_trace_main_t *hm = &ip6_hop_by_hop_ioam_trace_main; - u8 *s = 0; - int i = 0; - - for (i = 0; i < IP6_IOAM_TRACE_N_STATS; i++) - { - s = - format (s, " %s - %lu\n", ip6_hop_by_hop_ioam_trace_stats_strings[i], - hm->counters[i]); - } - - vlib_cli_output (vm, "%v", s); - vec_free (s); - return 0; -} - - -/* *INDENT-OFF* */ -VLIB_CLI_COMMAND (ip6_show_ioam_trace_cmd, static) = { - .path = "show ioam trace", - .short_help = "iOAM trace statistics", - .function = ip6_show_ioam_trace_cmd_fn, -}; -/* *INDENT-ON* */ - - -static clib_error_t * -ip6_hop_by_hop_ioam_trace_init (vlib_main_t * vm) -{ - ip6_hop_by_hop_ioam_trace_main_t *hm = &ip6_hop_by_hop_ioam_trace_main; - clib_error_t *error; - - if ((error = vlib_call_init_function (vm, ip_main_init))) - return (error); - - if ((error = vlib_call_init_function (vm, ip6_lookup_init))) - return error; - - if ((error = vlib_call_init_function (vm, ip6_hop_by_hop_ioam_init))) - return (error); - - hm->vlib_main = vm; - hm->vnet_main = vnet_get_main (); - memset (hm->counters, 0, sizeof (hm->counters)); - - - if (ip6_hbh_register_option - (HBH_OPTION_TYPE_IOAM_TRACE_DATA_LIST, - ip6_hbh_ioam_trace_data_list_handler, - ip6_hbh_ioam_trace_data_list_trace_handler) < 0) - return (clib_error_create - ("registration of HBH_OPTION_TYPE_IOAM_TRACE_DATA_LIST failed")); - - - if (ip6_hbh_add_register_option (HBH_OPTION_TYPE_IOAM_TRACE_DATA_LIST, - sizeof (ioam_trace_option_t), - ip6_hop_by_hop_ioam_trace_rewrite_handler) - < 0) - return (clib_error_create - ("registration of HBH_OPTION_TYPE_IOAM_TRACE_DATA_LIST for rewrite failed")); - - - return (0); -} - -int -ip6_trace_profile_cleanup (void) -{ - ip6_hop_by_hop_ioam_main_t *hm = &ip6_hop_by_hop_ioam_main; - - hm->options_size[HBH_OPTION_TYPE_IOAM_TRACE_DATA_LIST] = 0; - - return 0; - -} - - -int -ip6_trace_profile_setup (void) -{ - u32 trace_size = 0; - ip6_hop_by_hop_ioam_main_t *hm = &ip6_hop_by_hop_ioam_main; - - trace_profile *profile = NULL; - - - profile = trace_profile_find (); - - if (PREDICT_FALSE (!profile)) - { - ip6_ioam_trace_stats_increment_counter (IP6_IOAM_TRACE_PROFILE_MISS, 1); - return (-1); - } - - - if (ip6_ioam_trace_get_sizeof_handler (&trace_size) < 0) - return (-1); - - hm->options_size[HBH_OPTION_TYPE_IOAM_TRACE_DATA_LIST] = trace_size; - - return (0); -} - - -VLIB_INIT_FUNCTION (ip6_hop_by_hop_ioam_trace_init); - -/* - * fd.io coding-style-patch-verification: ON - * - * Local Variables: - * eval: (c-set-style "gnu") - * End: - */ diff --git a/plugins/ioam-plugin/ioam/export-common/ioam_export.h b/plugins/ioam-plugin/ioam/export-common/ioam_export.h deleted file mode 100644 index a9ec8d00..00000000 --- a/plugins/ioam-plugin/ioam/export-common/ioam_export.h +++ /dev/null @@ -1,616 +0,0 @@ -/* - * Copyright (c) 2016 Cisco and/or its affiliates. - * Licensed under the Apache License, Version 2.0 (the "License"); - * you may not use this file except in compliance with the License. - * You may obtain a copy of the License at: - * - * http://www.apache.org/licenses/LICENSE-2.0 - * - * Unless required by applicable law or agreed to in writing, software - * distributed under the License is distributed on an "AS IS" BASIS, - * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. - * See the License for the specific language governing permissions and - * limitations under the License. - */ -#ifndef __included_ioam_export_h__ -#define __included_ioam_export_h__ - -#include -#include -#include -#include -#include -#include -#include - -#include -#include -#include -#include - -#include - -typedef struct ioam_export_buffer -{ - /* Allocated buffer */ - u32 buffer_index; - u64 touched_at; - u8 records_in_this_buffer; -} ioam_export_buffer_t; - - -typedef struct -{ - /* API message ID base */ - u16 msg_id_base; - - /* TODO: to support multiple collectors all this has to be grouped and create a vector here */ - u8 *record_header; - u32 sequence_number; - u32 domain_id; - - /* ipfix collector, our ip address */ - ip4_address_t ipfix_collector; - ip4_address_t src_address; - - /* Pool of ioam_export_buffer_t */ - ioam_export_buffer_t *buffer_pool; - /* Vector of per thread ioam_export_buffer_t to buffer pool index */ - u32 *buffer_per_thread; - /* Lock per thread to swap buffers between worker and timer process */ - volatile u32 **lockp; - - /* time scale transform */ - u32 unix_time_0; - f64 vlib_time_0; - - /* convenience */ - vlib_main_t *vlib_main; - vnet_main_t *vnet_main; - u32 ip4_lookup_node_index; - - uword my_hbh_slot; - u32 export_process_node_index; -} ioam_export_main_t; - -ioam_export_main_t ioam_export_main; -ioam_export_main_t vxlan_gpe_ioam_export_main; - -vlib_node_registration_t export_node; - -#define DEFAULT_EXPORT_SIZE (3 * CLIB_CACHE_LINE_BYTES) -/* - * Number of records in a buffer - * ~(MTU (1500) - [ip hdr(40) + UDP(8) + ipfix (24)]) / DEFAULT_EXPORT_SIZE - */ -#define DEFAULT_EXPORT_RECORDS 7 - -always_inline ioam_export_buffer_t * -ioam_export_get_my_buffer (ioam_export_main_t * em, u32 thread_id) -{ - - if (vec_len (em->buffer_per_thread) > thread_id) - return (pool_elt_at_index - (em->buffer_pool, em->buffer_per_thread[thread_id])); - return (0); -} - -inline static int -ioam_export_buffer_add_header (ioam_export_main_t * em, vlib_buffer_t * b0) -{ - clib_memcpy (b0->data, em->record_header, vec_len (em->record_header)); - b0->current_data = 0; - b0->current_length = vec_len (em->record_header); - b0->flags |= VLIB_BUFFER_TOTAL_LENGTH_VALID; - return (1); -} - -inline static int -ioam_export_init_buffer (ioam_export_main_t * em, vlib_main_t * vm, - ioam_export_buffer_t * eb) -{ - vlib_buffer_t *b = 0; - - if (!eb) - return (-1); - /* TODO: Perhaps buffer init from template here */ - if (vlib_buffer_alloc (vm, &(eb->buffer_index), 1) != 1) - return (-2); - eb->records_in_this_buffer = 0; - eb->touched_at = vlib_time_now (vm); - b = vlib_get_buffer (vm, eb->buffer_index); - (void) ioam_export_buffer_add_header (em, b); - vnet_buffer (b)->sw_if_index[VLIB_RX] = 0; - vnet_buffer (b)->sw_if_index[VLIB_TX] = ~0; - return (1); -} - -inline static void -ioam_export_thread_buffer_free (ioam_export_main_t * em) -{ - vlib_main_t *vm = em->vlib_main; - ioam_export_buffer_t *eb = 0; - int i; - for (i = 0; i < vec_len (em->buffer_per_thread); i++) - { - eb = pool_elt_at_index (em->buffer_pool, em->buffer_per_thread[i]); - if (eb) - vlib_buffer_free (vm, &(eb->buffer_index), 1); - } - for (i = 0; i < vec_len (em->lockp); i++) - clib_mem_free ((void *) em->lockp[i]); - vec_free (em->buffer_per_thread); - pool_free (em->buffer_pool); - vec_free (em->lockp); - em->buffer_per_thread = 0; - em->buffer_pool = 0; - em->lockp = 0; -} - -inline static int -ioam_export_thread_buffer_init (ioam_export_main_t * em, vlib_main_t * vm) -{ - int no_of_threads = vec_len (vlib_worker_threads); - int i; - ioam_export_buffer_t *eb = 0; - vlib_node_t *ip4_lookup_node; - - pool_alloc_aligned (em->buffer_pool, - no_of_threads - 1, CLIB_CACHE_LINE_BYTES); - vec_validate_aligned (em->buffer_per_thread, - no_of_threads - 1, CLIB_CACHE_LINE_BYTES); - vec_validate_aligned (em->lockp, no_of_threads - 1, CLIB_CACHE_LINE_BYTES); - ip4_lookup_node = vlib_get_node_by_name (vm, (u8 *) "ip4-lookup"); - em->ip4_lookup_node_index = ip4_lookup_node->index; - if (!em->buffer_per_thread || !em->buffer_pool || !em->lockp) - { - return (-1); - } - for (i = 0; i < no_of_threads; i++) - { - eb = 0; - pool_get_aligned (em->buffer_pool, eb, CLIB_CACHE_LINE_BYTES); - memset (eb, 0, sizeof (*eb)); - em->buffer_per_thread[i] = eb - em->buffer_pool; - if (ioam_export_init_buffer (em, vm, eb) != 1) - { - ioam_export_thread_buffer_free (em); - return (-2); - } - em->lockp[i] = clib_mem_alloc_aligned (CLIB_CACHE_LINE_BYTES, - CLIB_CACHE_LINE_BYTES); - memset ((void *) em->lockp[i], 0, CLIB_CACHE_LINE_BYTES); - } - return (1); -} - -#define IPFIX_IOAM_EXPORT_ID 272 - -/* Used to build the rewrite */ -/* data set packet */ -typedef struct -{ - ipfix_message_header_t h; - ipfix_set_header_t s; -} ipfix_data_packet_t; - -typedef struct -{ - ip4_header_t ip4; - udp_header_t udp; - ipfix_data_packet_t ipfix; -} ip4_ipfix_data_packet_t; - - -inline static void -ioam_export_header_cleanup (ioam_export_main_t * em, - ip4_address_t * collector_address, - ip4_address_t * src_address) -{ - vec_free (em->record_header); - em->record_header = 0; -} - -inline static int -ioam_export_header_create (ioam_export_main_t * em, - ip4_address_t * collector_address, - ip4_address_t * src_address) -{ - ip4_header_t *ip; - udp_header_t *udp; - ipfix_message_header_t *h; - ipfix_set_header_t *s; - u8 *rewrite = 0; - ip4_ipfix_data_packet_t *tp; - - - /* allocate rewrite space */ - vec_validate_aligned (rewrite, - sizeof (ip4_ipfix_data_packet_t) - 1, - CLIB_CACHE_LINE_BYTES); - - tp = (ip4_ipfix_data_packet_t *) rewrite; - ip = (ip4_header_t *) & tp->ip4; - udp = (udp_header_t *) (ip + 1); - h = (ipfix_message_header_t *) (udp + 1); - s = (ipfix_set_header_t *) (h + 1); - - ip->ip_version_and_header_length = 0x45; - ip->ttl = 254; - ip->protocol = IP_PROTOCOL_UDP; - ip->src_address.as_u32 = src_address->as_u32; - ip->dst_address.as_u32 = collector_address->as_u32; - udp->src_port = clib_host_to_net_u16 (4939 /* $$FIXME */ ); - udp->dst_port = clib_host_to_net_u16 (4939); - /* FIXUP: UDP length */ - udp->length = clib_host_to_net_u16 (vec_len (rewrite) + - (DEFAULT_EXPORT_RECORDS * - DEFAULT_EXPORT_SIZE) - sizeof (*ip)); - - /* FIXUP: message header export_time */ - /* FIXUP: message header sequence_number */ - h->domain_id = clib_host_to_net_u32 (em->domain_id); - - /*FIXUP: Setid length in octets if records exported are not default */ - s->set_id_length = ipfix_set_id_length (IPFIX_IOAM_EXPORT_ID, - (sizeof (*s) + - (DEFAULT_EXPORT_RECORDS * - DEFAULT_EXPORT_SIZE))); - - /* FIXUP: h version and length length in octets if records exported are not default */ - h->version_length = version_length (sizeof (*h) + - (sizeof (*s) + - (DEFAULT_EXPORT_RECORDS * - DEFAULT_EXPORT_SIZE))); - - /* FIXUP: ip length if records exported are not default */ - /* FIXUP: ip checksum if records exported are not default */ - ip->length = clib_host_to_net_u16 (vec_len (rewrite) + - (DEFAULT_EXPORT_RECORDS * - DEFAULT_EXPORT_SIZE)); - ip->checksum = ip4_header_checksum (ip); - _vec_len (rewrite) = sizeof (ip4_ipfix_data_packet_t); - em->record_header = rewrite; - return (1); -} - -inline static int -ioam_export_send_buffer (ioam_export_main_t * em, vlib_main_t * vm, - ioam_export_buffer_t * eb) -{ - ip4_header_t *ip; - udp_header_t *udp; - ipfix_message_header_t *h; - ipfix_set_header_t *s; - ip4_ipfix_data_packet_t *tp; - vlib_buffer_t *b0; - u16 new_l0, old_l0; - ip_csum_t sum0; - vlib_frame_t *nf = 0; - u32 *to_next; - - b0 = vlib_get_buffer (vm, eb->buffer_index); - tp = vlib_buffer_get_current (b0); - ip = (ip4_header_t *) & tp->ip4; - udp = (udp_header_t *) (ip + 1); - h = (ipfix_message_header_t *) (udp + 1); - s = (ipfix_set_header_t *) (h + 1); - - /* FIXUP: message header export_time */ - h->export_time = clib_host_to_net_u32 ((u32) - (((f64) em->unix_time_0) + - (vlib_time_now (em->vlib_main) - - em->vlib_time_0))); - - /* FIXUP: message header sequence_number */ - h->sequence_number = clib_host_to_net_u32 (em->sequence_number++); - - /* FIXUP: lengths if different from default */ - if (PREDICT_FALSE (eb->records_in_this_buffer != DEFAULT_EXPORT_RECORDS)) - { - s->set_id_length = - ipfix_set_id_length (IPFIX_IOAM_EXPORT_ID /* set_id */ , - b0->current_length - (sizeof (*ip) + - sizeof (*udp) + - sizeof (*h))); - h->version_length = - version_length (b0->current_length - (sizeof (*ip) + sizeof (*udp))); - sum0 = ip->checksum; - old_l0 = ip->length; - new_l0 = clib_host_to_net_u16 ((u16) b0->current_length); - sum0 = ip_csum_update (sum0, old_l0, new_l0, ip4_header_t, - length /* changed member */ ); - ip->checksum = ip_csum_fold (sum0); - ip->length = new_l0; - udp->length = clib_host_to_net_u16 (b0->current_length - sizeof (*ip)); - } - - /* Enqueue pkts to ip4-lookup */ - - nf = vlib_get_frame_to_node (vm, em->ip4_lookup_node_index); - nf->n_vectors = 0; - to_next = vlib_frame_vector_args (nf); - nf->n_vectors = 1; - to_next[0] = eb->buffer_index; - vlib_put_frame_to_node (vm, em->ip4_lookup_node_index, nf); - return (1); - -} - -#define EXPORT_TIMEOUT (20.0) -#define THREAD_PERIOD (30.0) -inline static uword -ioam_export_process_common (ioam_export_main_t * em, vlib_main_t * vm, - vlib_node_runtime_t * rt, vlib_frame_t * f, - u32 index) -{ - f64 now; - f64 timeout = 30.0; - uword event_type; - uword *event_data = 0; - int i; - ioam_export_buffer_t *eb = 0, *new_eb = 0; - u32 *vec_buffer_indices = 0; - u32 *vec_buffer_to_be_sent = 0; - u32 *thread_index = 0; - u32 new_pool_index = 0; - - em->export_process_node_index = index; - /* Wait for Godot... */ - vlib_process_wait_for_event_or_clock (vm, 1e9); - event_type = vlib_process_get_events (vm, &event_data); - if (event_type != 1) - clib_warning ("bogus kickoff event received, %d", event_type); - vec_reset_length (event_data); - - while (1) - { - vlib_process_wait_for_event_or_clock (vm, timeout); - event_type = vlib_process_get_events (vm, &event_data); - switch (event_type) - { - case 2: /* Stop and Wait for kickoff again */ - timeout = 1e9; - break; - case 1: /* kickoff : Check for unsent buffers */ - timeout = THREAD_PERIOD; - break; - case ~0: /* timeout */ - break; - } - vec_reset_length (event_data); - now = vlib_time_now (vm); - /* - * Create buffers for threads that are not active enough - * to send out the export records - */ - for (i = 0; i < vec_len (em->buffer_per_thread); i++) - { - /* If the worker thread is processing export records ignore further checks */ - if (*em->lockp[i] == 1) - continue; - eb = pool_elt_at_index (em->buffer_pool, em->buffer_per_thread[i]); - if (eb->records_in_this_buffer > 0 - && now > (eb->touched_at + EXPORT_TIMEOUT)) - { - pool_get_aligned (em->buffer_pool, new_eb, - CLIB_CACHE_LINE_BYTES); - memset (new_eb, 0, sizeof (*new_eb)); - if (ioam_export_init_buffer (em, vm, new_eb) == 1) - { - new_pool_index = new_eb - em->buffer_pool; - vec_add (vec_buffer_indices, &new_pool_index, 1); - vec_add (vec_buffer_to_be_sent, &em->buffer_per_thread[i], - 1); - vec_add (thread_index, &i, 1); - } - else - { - pool_put (em->buffer_pool, new_eb); - /*Give up */ - goto CLEANUP; - } - } - } - if (vec_len (thread_index) != 0) - { - /* - * Now swap the buffers out - */ - for (i = 0; i < vec_len (thread_index); i++) - { - while (__sync_lock_test_and_set (em->lockp[thread_index[i]], 1)) - ; - em->buffer_per_thread[thread_index[i]] = - vec_pop (vec_buffer_indices); - *em->lockp[thread_index[i]] = 0; - } - - /* Send the buffers */ - for (i = 0; i < vec_len (vec_buffer_to_be_sent); i++) - { - eb = - pool_elt_at_index (em->buffer_pool, vec_buffer_to_be_sent[i]); - ioam_export_send_buffer (em, vm, eb); - pool_put (em->buffer_pool, eb); - } - } - - CLEANUP: - /* Free any leftover/unused buffers and everything that was allocated */ - for (i = 0; i < vec_len (vec_buffer_indices); i++) - { - new_eb = pool_elt_at_index (em->buffer_pool, vec_buffer_indices[i]); - vlib_buffer_free (vm, &new_eb->buffer_index, 1); - pool_put (em->buffer_pool, new_eb); - } - vec_free (vec_buffer_indices); - vec_free (vec_buffer_to_be_sent); - vec_free (thread_index); - } - return 0; /* not so much */ -} - -#define ioam_export_node_common(EM, VM, N, F, HTYPE, L, V, NEXT) \ -do { \ - u32 n_left_from, *from, *to_next; \ - export_next_t next_index; \ - u32 pkts_recorded = 0; \ - ioam_export_buffer_t *my_buf = 0; \ - vlib_buffer_t *eb0 = 0; \ - u32 ebi0 = 0; \ - from = vlib_frame_vector_args (F); \ - n_left_from = (F)->n_vectors; \ - next_index = (N)->cached_next_index; \ - while (__sync_lock_test_and_set ((EM)->lockp[(VM)->cpu_index], 1)); \ - my_buf = ioam_export_get_my_buffer (EM, (VM)->cpu_index); \ - my_buf->touched_at = vlib_time_now (VM); \ - while (n_left_from > 0) \ - { \ - u32 n_left_to_next; \ - vlib_get_next_frame (VM, N, next_index, to_next, n_left_to_next); \ - while (n_left_from >= 4 && n_left_to_next >= 2) \ - { \ - u32 next0 = NEXT; \ - u32 next1 = NEXT; \ - u32 bi0, bi1; \ - HTYPE *ip0, *ip1; \ - vlib_buffer_t *p0, *p1; \ - u32 ip_len0, ip_len1; \ - { \ - vlib_buffer_t *p2, *p3; \ - p2 = vlib_get_buffer (VM, from[2]); \ - p3 = vlib_get_buffer (VM, from[3]); \ - vlib_prefetch_buffer_header (p2, LOAD); \ - vlib_prefetch_buffer_header (p3, LOAD); \ - CLIB_PREFETCH (p2->data, 3 * CLIB_CACHE_LINE_BYTES, LOAD); \ - CLIB_PREFETCH (p3->data, 3 * CLIB_CACHE_LINE_BYTES, LOAD); \ - } \ - to_next[0] = bi0 = from[0]; \ - to_next[1] = bi1 = from[1]; \ - from += 2; \ - to_next += 2; \ - n_left_from -= 2; \ - n_left_to_next -= 2; \ - p0 = vlib_get_buffer (VM, bi0); \ - p1 = vlib_get_buffer (VM, bi1); \ - ip0 = vlib_buffer_get_current (p0); \ - ip1 = vlib_buffer_get_current (p1); \ - ip_len0 = \ - clib_net_to_host_u16 (ip0->L) + sizeof (HTYPE); \ - ip_len1 = \ - clib_net_to_host_u16 (ip1->L) + sizeof (HTYPE); \ - ebi0 = my_buf->buffer_index; \ - eb0 = vlib_get_buffer (VM, ebi0); \ - if (PREDICT_FALSE (eb0 == 0)) \ - goto NO_BUFFER1; \ - ip_len0 = \ - ip_len0 > DEFAULT_EXPORT_SIZE ? DEFAULT_EXPORT_SIZE : ip_len0; \ - ip_len1 = \ - ip_len1 > DEFAULT_EXPORT_SIZE ? DEFAULT_EXPORT_SIZE : ip_len1; \ - copy3cachelines (eb0->data + eb0->current_length, ip0, ip_len0); \ - eb0->current_length += DEFAULT_EXPORT_SIZE; \ - my_buf->records_in_this_buffer++; \ - if (my_buf->records_in_this_buffer >= DEFAULT_EXPORT_RECORDS) \ - { \ - ioam_export_send_buffer (EM, VM, my_buf); \ - ioam_export_init_buffer (EM, VM, my_buf); \ - } \ - ebi0 = my_buf->buffer_index; \ - eb0 = vlib_get_buffer (VM, ebi0); \ - if (PREDICT_FALSE (eb0 == 0)) \ - goto NO_BUFFER1; \ - copy3cachelines (eb0->data + eb0->current_length, ip1, ip_len1); \ - eb0->current_length += DEFAULT_EXPORT_SIZE; \ - my_buf->records_in_this_buffer++; \ - if (my_buf->records_in_this_buffer >= DEFAULT_EXPORT_RECORDS) \ - { \ - ioam_export_send_buffer (EM, VM, my_buf); \ - ioam_export_init_buffer (EM, VM, my_buf); \ - } \ - pkts_recorded += 2; \ - if (PREDICT_FALSE (((node)->flags & VLIB_NODE_FLAG_TRACE))) \ - { \ - if (p0->flags & VLIB_BUFFER_IS_TRACED) \ - { \ - export_trace_t *t = \ - vlib_add_trace (VM, node, p0, sizeof (*t)); \ - t->flow_label = \ - clib_net_to_host_u32 (ip0->V); \ - t->next_index = next0; \ - } \ - if (p1->flags & VLIB_BUFFER_IS_TRACED) \ - { \ - export_trace_t *t = \ - vlib_add_trace (VM, N, p1, sizeof (*t)); \ - t->flow_label = \ - clib_net_to_host_u32 (ip1->V); \ - t->next_index = next1; \ - } \ - } \ - NO_BUFFER1: \ - vlib_validate_buffer_enqueue_x2 (VM, N, next_index, \ - to_next, n_left_to_next, \ - bi0, bi1, next0, next1); \ - } \ - while (n_left_from > 0 && n_left_to_next > 0) \ - { \ - u32 bi0; \ - vlib_buffer_t *p0; \ - u32 next0 = NEXT; \ - HTYPE *ip0; \ - u32 ip_len0; \ - bi0 = from[0]; \ - to_next[0] = bi0; \ - from += 1; \ - to_next += 1; \ - n_left_from -= 1; \ - n_left_to_next -= 1; \ - p0 = vlib_get_buffer (VM, bi0); \ - ip0 = vlib_buffer_get_current (p0); \ - ip_len0 = \ - clib_net_to_host_u16 (ip0->L) + sizeof (HTYPE); \ - ebi0 = my_buf->buffer_index; \ - eb0 = vlib_get_buffer (VM, ebi0); \ - if (PREDICT_FALSE (eb0 == 0)) \ - goto NO_BUFFER; \ - ip_len0 = \ - ip_len0 > DEFAULT_EXPORT_SIZE ? DEFAULT_EXPORT_SIZE : ip_len0; \ - copy3cachelines (eb0->data + eb0->current_length, ip0, ip_len0); \ - eb0->current_length += DEFAULT_EXPORT_SIZE; \ - my_buf->records_in_this_buffer++; \ - if (my_buf->records_in_this_buffer >= DEFAULT_EXPORT_RECORDS) \ - { \ - ioam_export_send_buffer (EM, VM, my_buf); \ - ioam_export_init_buffer (EM, VM, my_buf); \ - } \ - if (PREDICT_FALSE (((N)->flags & VLIB_NODE_FLAG_TRACE) \ - && (p0->flags & VLIB_BUFFER_IS_TRACED))) \ - { \ - export_trace_t *t = vlib_add_trace (VM, (N), p0, sizeof (*t)); \ - t->flow_label = \ - clib_net_to_host_u32 (ip0->V); \ - t->next_index = next0; \ - } \ - pkts_recorded += 1; \ - NO_BUFFER: \ - vlib_validate_buffer_enqueue_x1 (VM, N, next_index, \ - to_next, n_left_to_next, \ - bi0, next0); \ - } \ - vlib_put_next_frame (VM, N, next_index, n_left_to_next); \ - } \ - vlib_node_increment_counter (VM, export_node.index, \ - EXPORT_ERROR_RECORDED, pkts_recorded); \ - *(EM)->lockp[(VM)->cpu_index] = 0; \ -} while(0) - -#endif /* __included_ioam_export_h__ */ - -/* - * fd.io coding-style-patch-verification: ON - * - * Local Variables: - * eval: (c-set-style "gnu") - * End: - */ diff --git a/plugins/ioam-plugin/ioam/export-vxlan-gpe/vxlan_gpe_ioam_export.api b/plugins/ioam-plugin/ioam/export-vxlan-gpe/vxlan_gpe_ioam_export.api deleted file mode 100644 index 7b17c3f7..00000000 --- a/plugins/ioam-plugin/ioam/export-vxlan-gpe/vxlan_gpe_ioam_export.api +++ /dev/null @@ -1,42 +0,0 @@ -/* Hey Emacs use -*- mode: C -*- */ -/* - * Copyright (c) 2016 Cisco and/or its affiliates. - * Licensed under the Apache License, Version 2.0 (the "License"); - * you may not use this file except in compliance with the License. - * You may obtain a copy of the License at: - * - * http://www.apache.org/licenses/LICENSE-2.0 - * - * Unless required by applicable law or agreed to in writing, software - * distributed under the License is distributed on an "AS IS" BASIS, - * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. - * See the License for the specific language governing permissions and - * limitations under the License. - */ - -/* Define a simple binary API to control the feature */ - -define vxlan_gpe_ioam_export_enable_disable { - /* Client identifier, set from api_main.my_client_index */ - u32 client_index; - - /* Arbitrary context, so client can match reply to request */ - u32 context; - - /* Enable / disable the feature */ - u8 is_disable; - - /* Collector ip address */ - u8 collector_address[4]; - u8 src_address[4]; - - /* Src ip address */ -}; - -define vxlan_gpe_ioam_export_enable_disable_reply { - /* From the request */ - u32 context; - - /* Return value, zero means all OK */ - i32 retval; -}; \ No newline at end of file diff --git a/plugins/ioam-plugin/ioam/export-vxlan-gpe/vxlan_gpe_ioam_export.c b/plugins/ioam-plugin/ioam/export-vxlan-gpe/vxlan_gpe_ioam_export.c deleted file mode 100644 index bab8d977..00000000 --- a/plugins/ioam-plugin/ioam/export-vxlan-gpe/vxlan_gpe_ioam_export.c +++ /dev/null @@ -1,271 +0,0 @@ -/* - * Copyright (c) 2016 Cisco and/or its affiliates. - * Licensed under the Apache License, Version 2.0 (the "License"); - * you may not use this file except in compliance with the License. - * You may obtain a copy of the License at: - * - * http://www.apache.org/licenses/LICENSE-2.0 - * - * Unless required by applicable law or agreed to in writing, software - * distributed under the License is distributed on an "AS IS" BASIS, - * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. - * See the License for the specific language governing permissions and - * limitations under the License. - */ -/* - *------------------------------------------------------------------ - * vxlan_gpe_ioam_export.c - ioam export API / debug CLI handling - *------------------------------------------------------------------ - */ - -#include -#include -#include -#include - -#include -#include -#include - -#include - -/* define message IDs */ -#include - -/* define message structures */ -#define vl_typedefs -#include -#undef vl_typedefs - -/* define generated endian-swappers */ -#define vl_endianfun -#include -#undef vl_endianfun - -/* instantiate all the print functions we know about */ -#define vl_print(handle, ...) vlib_cli_output (handle, __VA_ARGS__) -#define vl_printfun -#include -#undef vl_printfun - -/* Get the API version number */ -#define vl_api_version(n,v) static u32 api_version=(v); -#include -#undef vl_api_version - -/* - * A handy macro to set up a message reply. - * Assumes that the following variables are available: - * mp - pointer to request message - * rmp - pointer to reply message type - * rv - return value - */ - -#define REPLY_MACRO(t) \ -do { \ - unix_shared_memory_queue_t * q = \ - vl_api_client_index_to_input_queue (mp->client_index); \ - if (!q) \ - return; \ - \ - rmp = vl_msg_api_alloc (sizeof (*rmp)); \ - rmp->_vl_msg_id = ntohs((t)+sm->msg_id_base); \ - rmp->context = mp->context; \ - rmp->retval = ntohl(rv); \ - \ - vl_msg_api_send_shmem (q, (u8 *)&rmp); \ -} while(0); - - -/* List of message types that this plugin understands */ - - -#define foreach_vxlan_gpe_ioam_export_plugin_api_msg \ -_(VXLAN_GPE_IOAM_EXPORT_ENABLE_DISABLE, vxlan_gpe_ioam_export_enable_disable) - -extern void vxlan_gpe_set_next_override (uword next); -/* Action function shared between message handler and debug CLI */ -int -vxlan_gpe_ioam_export_enable_disable (ioam_export_main_t * em, - u8 is_disable, - ip4_address_t * collector_address, - ip4_address_t * src_address) -{ - vlib_main_t *vm = em->vlib_main; - u32 node_index = export_node.index; - vlib_node_t *vxlan_gpe_decap_ioam_node = NULL; - - if (is_disable == 0) - { - if (em->my_hbh_slot == ~0) - { - /* Hook this export node to vxlan-gpe-decap-ioam-v4 */ - vxlan_gpe_decap_ioam_node = - vlib_get_node_by_name (vm, (u8 *) "vxlan-gpe-decap-ioam-v4"); - if (!vxlan_gpe_decap_ioam_node) - { - /* node does not exist give up */ - return (-1); - } - em->my_hbh_slot = - vlib_node_add_next (vm, vxlan_gpe_decap_ioam_node->index, - node_index); - } - if (1 == ioam_export_header_create (em, collector_address, src_address)) - { - ioam_export_thread_buffer_init (em, vm); - vxlan_gpe_set_next_override (em->my_hbh_slot); - /* Turn on the export buffer check process */ - vlib_process_signal_event (vm, em->export_process_node_index, 1, 0); - - } - else - { - return (-2); - } - } - else - { - vxlan_gpe_set_next_override (VXLAN_GPE_DECAP_IOAM_V4_NEXT_POP); - ioam_export_header_cleanup (em, collector_address, src_address); - ioam_export_thread_buffer_free (em); - /* Turn off the export buffer check process */ - vlib_process_signal_event (vm, em->export_process_node_index, 2, 0); - - } - - return 0; -} - -/* API message handler */ -static void vl_api_vxlan_gpe_ioam_export_enable_disable_t_handler - (vl_api_vxlan_gpe_ioam_export_enable_disable_t * mp) -{ - vl_api_vxlan_gpe_ioam_export_enable_disable_reply_t *rmp; - ioam_export_main_t *sm = &vxlan_gpe_ioam_export_main; - int rv; - - rv = vxlan_gpe_ioam_export_enable_disable (sm, (int) (mp->is_disable), - (ip4_address_t *) - mp->collector_address, - (ip4_address_t *) - mp->src_address); - - REPLY_MACRO (VL_API_VXLAN_GPE_IOAM_EXPORT_ENABLE_DISABLE_REPLY); -} /* API message handler */ - - - -/* Set up the API message handling tables */ -static clib_error_t * -vxlan_gpe_ioam_export_plugin_api_hookup (vlib_main_t * vm) -{ - ioam_export_main_t *sm = &vxlan_gpe_ioam_export_main; -#define _(N,n) \ - vl_msg_api_set_handlers((VL_API_##N + sm->msg_id_base), \ - #n, \ - vl_api_##n##_t_handler, \ - vl_noop_handler, \ - vl_api_##n##_t_endian, \ - vl_api_##n##_t_print, \ - sizeof(vl_api_##n##_t), 1); - foreach_vxlan_gpe_ioam_export_plugin_api_msg; -#undef _ - - return 0; -} - - -static clib_error_t * -set_vxlan_gpe_ioam_export_ipfix_command_fn (vlib_main_t * vm, - unformat_input_t * input, - vlib_cli_command_t * cmd) -{ - ioam_export_main_t *em = &vxlan_gpe_ioam_export_main; - ip4_address_t collector, src; - u8 is_disable = 0; - - collector.as_u32 = 0; - src.as_u32 = 0; - - while (unformat_check_input (input) != UNFORMAT_END_OF_INPUT) - { - if (unformat (input, "collector %U", unformat_ip4_address, &collector)) - ; - else if (unformat (input, "src %U", unformat_ip4_address, &src)) - ; - else if (unformat (input, "disable")) - is_disable = 1; - else - break; - } - - if (collector.as_u32 == 0) - return clib_error_return (0, "collector address required"); - - if (src.as_u32 == 0) - return clib_error_return (0, "src address required"); - - em->ipfix_collector.as_u32 = collector.as_u32; - em->src_address.as_u32 = src.as_u32; - - vlib_cli_output (vm, "Collector %U, src address %U", - format_ip4_address, &em->ipfix_collector, - format_ip4_address, &em->src_address); - - /* Turn on the export timer process */ - // vlib_process_signal_event (vm, flow_report_process_node.index, - //1, 0); - if (0 != - vxlan_gpe_ioam_export_enable_disable (em, is_disable, &collector, &src)) - { - return clib_error_return (0, "Unable to set ioam vxlan-gpe export"); - } - - return 0; -} - -/* *INDENT-OFF* */ -VLIB_CLI_COMMAND (set_vxlan_gpe_ioam_ipfix_command, static) = -{ -.path = "set vxlan-gpe-ioam export ipfix", -.short_help = "set vxlan-gpe-ioam export ipfix collector src ", -.function = set_vxlan_gpe_ioam_export_ipfix_command_fn, -}; -/* *INDENT-ON* */ - - -static clib_error_t * -vxlan_gpe_ioam_export_init (vlib_main_t * vm) -{ - ioam_export_main_t *em = &vxlan_gpe_ioam_export_main; - clib_error_t *error = 0; - u8 *name; - - name = format (0, "vxlan_gpe_ioam_export_%08x%c", api_version, 0); - - /* Ask for a correctly-sized block of API message decode slots */ - em->msg_id_base = vl_msg_api_get_msg_ids - ((char *) name, VL_MSG_FIRST_AVAILABLE); - em->unix_time_0 = (u32) time (0); /* Store starting time */ - em->vlib_time_0 = vlib_time_now (vm); - - error = vxlan_gpe_ioam_export_plugin_api_hookup (vm); - em->my_hbh_slot = ~0; - em->vlib_main = vm; - em->vnet_main = vnet_get_main (); - vec_free (name); - - return error; -} - -VLIB_INIT_FUNCTION (vxlan_gpe_ioam_export_init); - - -/* - * fd.io coding-style-patch-verification: ON - * - * Local Variables: - * eval: (c-set-style "gnu") - * End: - */ diff --git a/plugins/ioam-plugin/ioam/export-vxlan-gpe/vxlan_gpe_ioam_export_all_api_h.h b/plugins/ioam-plugin/ioam/export-vxlan-gpe/vxlan_gpe_ioam_export_all_api_h.h deleted file mode 100644 index 6d93f093..00000000 --- a/plugins/ioam-plugin/ioam/export-vxlan-gpe/vxlan_gpe_ioam_export_all_api_h.h +++ /dev/null @@ -1,16 +0,0 @@ -/* - * Copyright (c) 2016 Cisco and/or its affiliates. - * Licensed under the Apache License, Version 2.0 (the "License"); - * you may not use this file except in compliance with the License. - * You may obtain a copy of the License at: - * - * http://www.apache.org/licenses/LICENSE-2.0 - * - * Unless required by applicable law or agreed to in writing, software - * distributed under the License is distributed on an "AS IS" BASIS, - * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. - * See the License for the specific language governing permissions and - * limitations under the License. - */ -/* Include the generated file, see BUILT_SOURCES in Makefile.am */ -#include diff --git a/plugins/ioam-plugin/ioam/export-vxlan-gpe/vxlan_gpe_ioam_export_msg_enum.h b/plugins/ioam-plugin/ioam/export-vxlan-gpe/vxlan_gpe_ioam_export_msg_enum.h deleted file mode 100644 index cc5698de..00000000 --- a/plugins/ioam-plugin/ioam/export-vxlan-gpe/vxlan_gpe_ioam_export_msg_enum.h +++ /dev/null @@ -1,28 +0,0 @@ -/* - * Copyright (c) 2016 Cisco and/or its affiliates. - * Licensed under the Apache License, Version 2.0 (the "License"); - * you may not use this file except in compliance with the License. - * You may obtain a copy of the License at: - * - * http://www.apache.org/licenses/LICENSE-2.0 - * - * Unless required by applicable law or agreed to in writing, software - * distributed under the License is distributed on an "AS IS" BASIS, - * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. - * See the License for the specific language governing permissions and - * limitations under the License. - */ -#ifndef included_vxlan_gpe_ioam_export_msg_enum_h -#define included_vxlan_gpe_ioam_export_msg_enum_h - -#include - -#define vl_msg_id(n,h) n, -typedef enum { -#include - /* We'll want to know how many messages IDs we need... */ - VL_MSG_FIRST_AVAILABLE, -} vl_msg_id_t; -#undef vl_msg_id - -#endif /* included_vxlan_gpe_ioam_export_msg_enum_h */ diff --git a/plugins/ioam-plugin/ioam/export-vxlan-gpe/vxlan_gpe_ioam_export_test.c b/plugins/ioam-plugin/ioam/export-vxlan-gpe/vxlan_gpe_ioam_export_test.c deleted file mode 100644 index 494263d9..00000000 --- a/plugins/ioam-plugin/ioam/export-vxlan-gpe/vxlan_gpe_ioam_export_test.c +++ /dev/null @@ -1,215 +0,0 @@ -/* - * Copyright (c) 2016 Cisco and/or its affiliates. - * Licensed under the Apache License, Version 2.0 (the "License"); - * you may not use this file except in compliance with the License. - * You may obtain a copy of the License at: - * - * http://www.apache.org/licenses/LICENSE-2.0 - * - * Unless required by applicable law or agreed to in writing, software - * distributed under the License is distributed on an "AS IS" BASIS, - * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. - * See the License for the specific language governing permissions and - * limitations under the License. - */ -/* - *------------------------------------------------------------------ - * vxlan_gpe_ioam_export_test.c - test harness plugin - *------------------------------------------------------------------ - */ - -#include -#include -#include -#include -#include - - -/* Declare message IDs */ -#include - -/* define message structures */ -#define vl_typedefs -#include -#undef vl_typedefs - -/* declare message handlers for each api */ - -#define vl_endianfun /* define message structures */ -#include -#undef vl_endianfun - -/* instantiate all the print functions we know about */ -#define vl_print(handle, ...) -#define vl_printfun -#include -#undef vl_printfun - -/* Get the API version number. */ -#define vl_api_version(n,v) static u32 api_version=(v); -#include -#undef vl_api_version - - -typedef struct -{ - /* API message ID base */ - u16 msg_id_base; - vat_main_t *vat_main; -} export_test_main_t; - -export_test_main_t export_test_main; - -#define foreach_standard_reply_retval_handler \ -_(vxlan_gpe_ioam_export_enable_disable_reply) - -#define _(n) \ - static void vl_api_##n##_t_handler \ - (vl_api_##n##_t * mp) \ - { \ - vat_main_t * vam = export_test_main.vat_main; \ - i32 retval = ntohl(mp->retval); \ - if (vam->async_mode) { \ - vam->async_errors += (retval < 0); \ - } else { \ - vam->retval = retval; \ - vam->result_ready = 1; \ - } \ - } -foreach_standard_reply_retval_handler; -#undef _ - -/* - * Table of message reply handlers, must include boilerplate handlers - * we just generated - */ -#define foreach_vpe_api_reply_msg \ -_(VXLAN_GPE_IOAM_EXPORT_ENABLE_DISABLE_REPLY, vxlan_gpe_ioam_export_enable_disable_reply) - - -/* M: construct, but don't yet send a message */ - -#define M(T,t) \ -do { \ - vam->result_ready = 0; \ - mp = vl_msg_api_alloc(sizeof(*mp)); \ - memset (mp, 0, sizeof (*mp)); \ - mp->_vl_msg_id = ntohs (VL_API_##T + sm->msg_id_base); \ - mp->client_index = vam->my_client_index; \ -} while(0); - -#define M2(T,t,n) \ -do { \ - vam->result_ready = 0; \ - mp = vl_msg_api_alloc(sizeof(*mp)+(n)); \ - memset (mp, 0, sizeof (*mp)); \ - mp->_vl_msg_id = ntohs (VL_API_##T + sm->msg_id_base); \ - mp->client_index = vam->my_client_index; \ -} while(0); - -/* S: send a message */ -#define S (vl_msg_api_send_shmem (vam->vl_input_queue, (u8 *)&mp)) - -/* W: wait for results, with timeout */ -#define W \ -do { \ - timeout = vat_time_now (vam) + 1.0; \ - \ - while (vat_time_now (vam) < timeout) { \ - if (vam->result_ready == 1) { \ - return (vam->retval); \ - } \ - } \ - return -99; \ -} while(0); - -static int -api_vxlan_gpe_ioam_export_enable_disable (vat_main_t * vam) -{ - export_test_main_t *sm = &export_test_main; - unformat_input_t *i = vam->input; - f64 timeout; - int is_disable = 0; - vl_api_vxlan_gpe_ioam_export_enable_disable_t *mp; - - /* Parse args required to build the message */ - while (unformat_check_input (i) != UNFORMAT_END_OF_INPUT) - { - if (unformat (i, "disable")) - is_disable = 1; - else - break; - } - - /* Construct the API message */ - M (VXLAN_GPE_IOAM_EXPORT_ENABLE_DISABLE, - vxlan_gpe_ioam_export_enable_disable); - mp->is_disable = is_disable; - - /* send it... */ - S; - - /* Wait for a reply... */ - W; -} - -/* - * List of messages that the api test plugin sends, - * and that the data plane plugin processes - */ -#define foreach_vpe_api_msg \ -_(vxlan_gpe_ioam_export_enable_disable, " [disable]") - -void -vat_api_hookup (vat_main_t * vam) -{ - export_test_main_t *sm = &export_test_main; - /* Hook up handlers for replies from the data plane plug-in */ -#define _(N,n) \ - vl_msg_api_set_handlers((VL_API_##N + sm->msg_id_base), \ - #n, \ - vl_api_##n##_t_handler, \ - vl_noop_handler, \ - vl_api_##n##_t_endian, \ - vl_api_##n##_t_print, \ - sizeof(vl_api_##n##_t), 1); - foreach_vpe_api_reply_msg; -#undef _ - - /* API messages we can send */ -#define _(n,h) hash_set_mem (vam->function_by_name, #n, api_##n); - foreach_vpe_api_msg; -#undef _ - - /* Help strings */ -#define _(n,h) hash_set_mem (vam->help_by_name, #n, h); - foreach_vpe_api_msg; -#undef _ -} - -clib_error_t * -vat_plugin_register (vat_main_t * vam) -{ - export_test_main_t *sm = &export_test_main; - u8 *name; - - sm->vat_main = vam; - - name = format (0, "vxlan_gpe_ioam_export_%08x%c", api_version, 0); - sm->msg_id_base = vl_client_get_first_plugin_msg_id ((char *) name); - - if (sm->msg_id_base != (u16) ~ 0) - vat_api_hookup (vam); - - vec_free (name); - - return 0; -} - -/* - * fd.io coding-style-patch-verification: ON - * - * Local Variables: - * eval: (c-set-style "gnu") - * End: - */ diff --git a/plugins/ioam-plugin/ioam/export-vxlan-gpe/vxlan_gpe_ioam_export_thread.c b/plugins/ioam-plugin/ioam/export-vxlan-gpe/vxlan_gpe_ioam_export_thread.c deleted file mode 100644 index 58508ebf..00000000 --- a/plugins/ioam-plugin/ioam/export-vxlan-gpe/vxlan_gpe_ioam_export_thread.c +++ /dev/null @@ -1,49 +0,0 @@ -/* - * Copyright (c) 2016 Cisco and/or its affiliates. - * Licensed under the Apache License, Version 2.0 (the "License"); - * you may not use this file except in compliance with the License. - * You may obtain a copy of the License at: - * - * http://www.apache.org/licenses/LICENSE-2.0 - * - * Unless required by applicable law or agreed to in writing, software - * distributed under the License is distributed on an "AS IS" BASIS, - * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. - * See the License for the specific language governing permissions and - * limitations under the License. - */ -/* - * ioam_export_thread.c - */ -#include -#include -#include - -static vlib_node_registration_t vxlan_gpe_ioam_export_process_node; - -static uword -vxlan_gpe_ioam_export_process (vlib_main_t * vm, - vlib_node_runtime_t * rt, vlib_frame_t * f) -{ - return (ioam_export_process_common (&vxlan_gpe_ioam_export_main, - vm, rt, f, - vxlan_gpe_ioam_export_process_node.index)); -} - - -/* *INDENT-OFF* */ -VLIB_REGISTER_NODE (vxlan_gpe_ioam_export_process_node, static) = -{ - .function = vxlan_gpe_ioam_export_process, - .type = VLIB_NODE_TYPE_PROCESS, - .name = "vxlan-gpe-ioam-export-process", -}; -/* *INDENT-ON* */ - -/* - * fd.io coding-style-patch-verification: ON - * - * Local Variables: - * eval: (c-set-style "gnu") - * End: - */ diff --git a/plugins/ioam-plugin/ioam/export-vxlan-gpe/vxlan_gpe_node.c b/plugins/ioam-plugin/ioam/export-vxlan-gpe/vxlan_gpe_node.c deleted file mode 100644 index 722c2b06..00000000 --- a/plugins/ioam-plugin/ioam/export-vxlan-gpe/vxlan_gpe_node.c +++ /dev/null @@ -1,162 +0,0 @@ -/* - * Copyright (c) 2016 Cisco and/or its affiliates. - * Licensed under the Apache License, Version 2.0 (the "License"); - * you may not use this file except in compliance with the License. - * You may obtain a copy of the License at: - * - * http://www.apache.org/licenses/LICENSE-2.0 - * - * Unless required by applicable law or agreed to in writing, software - * distributed under the License is distributed on an "AS IS" BASIS, - * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. - * See the License for the specific language governing permissions and - * limitations under the License. - */ -#include -#include -#include -#include -#include -#include -#include -#include - -typedef struct -{ - u32 next_index; - u32 flow_label; -} export_trace_t; - -/* packet trace format function */ -static u8 * -format_export_trace (u8 * s, va_list * args) -{ - CLIB_UNUSED (vlib_main_t * vm) = va_arg (*args, vlib_main_t *); - CLIB_UNUSED (vlib_node_t * node) = va_arg (*args, vlib_node_t *); - export_trace_t *t = va_arg (*args, export_trace_t *); - - s = format (s, "EXPORT: flow_label %d, next index %d", - t->flow_label, t->next_index); - return s; -} - -vlib_node_registration_t export_node; - -#define foreach_export_error \ -_(RECORDED, "Packets recorded for export") - -typedef enum -{ -#define _(sym,str) EXPORT_ERROR_##sym, - foreach_export_error -#undef _ - EXPORT_N_ERROR, -} export_error_t; - -static char *export_error_strings[] = { -#define _(sym,string) string, - foreach_export_error -#undef _ -}; - -typedef enum -{ - EXPORT_NEXT_VXLAN_GPE_INPUT, - EXPORT_N_NEXT, -} export_next_t; - -always_inline void -copy3cachelines (void *dst, const void *src, size_t n) -{ -#if 0 - if (PREDICT_FALSE (n < DEFAULT_EXPORT_SIZE)) - { - /* Copy only the first 1/2 cache lines whatever is available */ - if (n >= 64) - clib_mov64 ((u8 *) dst, (const u8 *) src); - if (n >= 128) - clib_mov64 ((u8 *) dst + 64, (const u8 *) src + 64); - return; - } - clib_mov64 ((u8 *) dst, (const u8 *) src); - clib_mov64 ((u8 *) dst + 64, (const u8 *) src + 64); - clib_mov64 ((u8 *) dst + 128, (const u8 *) src + 128); -#endif -#if 1 - - u64 *copy_dst, *copy_src; - int i; - copy_dst = (u64 *) dst; - copy_src = (u64 *) src; - if (PREDICT_FALSE (n < DEFAULT_EXPORT_SIZE)) - { - for (i = 0; i < n / 64; i++) - { - copy_dst[0] = copy_src[0]; - copy_dst[1] = copy_src[1]; - copy_dst[2] = copy_src[2]; - copy_dst[3] = copy_src[3]; - copy_dst[4] = copy_src[4]; - copy_dst[5] = copy_src[5]; - copy_dst[6] = copy_src[6]; - copy_dst[7] = copy_src[7]; - copy_dst += 8; - copy_src += 8; - } - return; - } - for (i = 0; i < 3; i++) - { - copy_dst[0] = copy_src[0]; - copy_dst[1] = copy_src[1]; - copy_dst[2] = copy_src[2]; - copy_dst[3] = copy_src[3]; - copy_dst[4] = copy_src[4]; - copy_dst[5] = copy_src[5]; - copy_dst[6] = copy_src[6]; - copy_dst[7] = copy_src[7]; - copy_dst += 8; - copy_src += 8; - } -#endif -} - - -static uword -vxlan_gpe_export_node_fn (vlib_main_t * vm, - vlib_node_runtime_t * node, vlib_frame_t * frame) -{ - ioam_export_main_t *em = &vxlan_gpe_ioam_export_main; - ioam_export_node_common (em, vm, node, frame, ip4_header_t, length, - ip_version_and_header_length, - EXPORT_NEXT_VXLAN_GPE_INPUT); - return frame->n_vectors; -} - -/* - * Node for VXLAN-GPE export - */ -/* *INDENT-OFF* */ -VLIB_REGISTER_NODE (export_node) = -{ - .function = vxlan_gpe_export_node_fn, - .name = "vxlan-gpe-ioam-export", - .vector_size = sizeof (u32), - .format_trace = format_export_trace, - .type = VLIB_NODE_TYPE_INTERNAL, - .n_errors = ARRAY_LEN (export_error_strings), - .error_strings = export_error_strings, - .n_next_nodes = EXPORT_N_NEXT, - /* edit / add dispositions here */ - .next_nodes = - {[EXPORT_NEXT_VXLAN_GPE_INPUT] = "vxlan-gpe-pop-ioam-v4"}, -}; -/* *INDENT-ON* */ - -/* - * fd.io coding-style-patch-verification: ON - * - * Local Variables: - * eval: (c-set-style "gnu") - * End: - */ diff --git a/plugins/ioam-plugin/ioam/export/ioam_export.api b/plugins/ioam-plugin/ioam/export/ioam_export.api deleted file mode 100644 index f22d9fc8..00000000 --- a/plugins/ioam-plugin/ioam/export/ioam_export.api +++ /dev/null @@ -1,42 +0,0 @@ -/* Hey Emacs use -*- mode: C -*- */ -/* - * Copyright (c) 2016 Cisco and/or its affiliates. - * Licensed under the Apache License, Version 2.0 (the "License"); - * you may not use this file except in compliance with the License. - * You may obtain a copy of the License at: - * - * http://www.apache.org/licenses/LICENSE-2.0 - * - * Unless required by applicable law or agreed to in writing, software - * distributed under the License is distributed on an "AS IS" BASIS, - * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. - * See the License for the specific language governing permissions and - * limitations under the License. - */ - -/* Define a simple binary API to control the feature */ - -define ioam_export_ip6_enable_disable { - /* Client identifier, set from api_main.my_client_index */ - u32 client_index; - - /* Arbitrary context, so client can match reply to request */ - u32 context; - - /* Enable / disable the feature */ - u8 is_disable; - - /* Collector ip address */ - u8 collector_address[4]; - u8 src_address[4]; - - /* Src ip address */ -}; - -define ioam_export_ip6_enable_disable_reply { - /* From the request */ - u32 context; - - /* Return value, zero means all OK */ - i32 retval; -}; diff --git a/plugins/ioam-plugin/ioam/export/ioam_export.c b/plugins/ioam-plugin/ioam/export/ioam_export.c deleted file mode 100644 index b122e445..00000000 --- a/plugins/ioam-plugin/ioam/export/ioam_export.c +++ /dev/null @@ -1,282 +0,0 @@ -/* - * Copyright (c) 2016 Cisco and/or its affiliates. - * Licensed under the Apache License, Version 2.0 (the "License"); - * you may not use this file except in compliance with the License. - * You may obtain a copy of the License at: - * - * http://www.apache.org/licenses/LICENSE-2.0 - * - * Unless required by applicable law or agreed to in writing, software - * distributed under the License is distributed on an "AS IS" BASIS, - * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. - * See the License for the specific language governing permissions and - * limitations under the License. - */ -/* - *------------------------------------------------------------------ - * ioam_export.c - ioam export API / debug CLI handling - *------------------------------------------------------------------ - */ - -#include -#include -#include - -#include -#include -#include -#include - - -/* define message IDs */ -#include - -/* define message structures */ -#define vl_typedefs -#include -#undef vl_typedefs - -/* define generated endian-swappers */ -#define vl_endianfun -#include -#undef vl_endianfun - -/* instantiate all the print functions we know about */ -#define vl_print(handle, ...) vlib_cli_output (handle, __VA_ARGS__) -#define vl_printfun -#include -#undef vl_printfun - -/* Get the API version number */ -#define vl_api_version(n,v) static u32 api_version=(v); -#include -#undef vl_api_version - -/* - * A handy macro to set up a message reply. - * Assumes that the following variables are available: - * mp - pointer to request message - * rmp - pointer to reply message type - * rv - return value - */ - -#define REPLY_MACRO(t) \ -do { \ - unix_shared_memory_queue_t * q = \ - vl_api_client_index_to_input_queue (mp->client_index); \ - if (!q) \ - return; \ - \ - rmp = vl_msg_api_alloc (sizeof (*rmp)); \ - rmp->_vl_msg_id = ntohs((t)+sm->msg_id_base); \ - rmp->context = mp->context; \ - rmp->retval = ntohl(rv); \ - \ - vl_msg_api_send_shmem (q, (u8 *)&rmp); \ -} while(0); - - -/* List of message types that this plugin understands */ - -#define foreach_ioam_export_plugin_api_msg \ -_(IOAM_EXPORT_IP6_ENABLE_DISABLE, ioam_export_ip6_enable_disable) - -/* - * This routine exists to convince the vlib plugin framework that - * we haven't accidentally copied a random .dll into the plugin directory. - * - * Also collects global variable pointers passed from the vpp engine - */ - -clib_error_t * -vlib_plugin_register (vlib_main_t * vm, vnet_plugin_handoff_t * h, - int from_early_init) -{ - ioam_export_main_t *em = &ioam_export_main; - clib_error_t *error = 0; - - em->vlib_main = vm; - em->vnet_main = h->vnet_main; - - return error; -} - -/* Action function shared between message handler and debug CLI */ - -int -ioam_export_ip6_enable_disable (ioam_export_main_t * em, - u8 is_disable, - ip4_address_t * collector_address, - ip4_address_t * src_address) -{ - vlib_main_t *vm = em->vlib_main; - - if (is_disable == 0) - { - if (1 == ioam_export_header_create (em, collector_address, src_address)) - { - ioam_export_thread_buffer_init (em, vm); - ip6_hbh_set_next_override (em->my_hbh_slot); - /* Turn on the export buffer check process */ - vlib_process_signal_event (vm, em->export_process_node_index, 1, 0); - - } - else - { - return (-2); - } - } - else - { - ip6_hbh_set_next_override (IP6_LOOKUP_NEXT_POP_HOP_BY_HOP); - ioam_export_header_cleanup (em, collector_address, src_address); - ioam_export_thread_buffer_free (em); - /* Turn off the export buffer check process */ - vlib_process_signal_event (vm, em->export_process_node_index, 2, 0); - - } - - return 0; -} - -/* API message handler */ -static void vl_api_ioam_export_ip6_enable_disable_t_handler - (vl_api_ioam_export_ip6_enable_disable_t * mp) -{ - vl_api_ioam_export_ip6_enable_disable_reply_t *rmp; - ioam_export_main_t *sm = &ioam_export_main; - int rv; - - rv = ioam_export_ip6_enable_disable (sm, (int) (mp->is_disable), - (ip4_address_t *) - mp->collector_address, - (ip4_address_t *) mp->src_address); - - REPLY_MACRO (VL_API_IOAM_EXPORT_IP6_ENABLE_DISABLE_REPLY); -} - -/* Set up the API message handling tables */ -static clib_error_t * -ioam_export_plugin_api_hookup (vlib_main_t * vm) -{ - ioam_export_main_t *sm = &ioam_export_main; -#define _(N,n) \ - vl_msg_api_set_handlers((VL_API_##N + sm->msg_id_base), \ - #n, \ - vl_api_##n##_t_handler, \ - vl_noop_handler, \ - vl_api_##n##_t_endian, \ - vl_api_##n##_t_print, \ - sizeof(vl_api_##n##_t), 1); - foreach_ioam_export_plugin_api_msg; -#undef _ - - return 0; -} - -#define vl_msg_name_crc_list -#include -#undef vl_msg_name_crc_list - -static void -setup_message_id_table (ioam_export_main_t * sm, api_main_t * am) -{ -#define _(id,n,crc) \ - vl_msg_api_add_msg_name_crc (am, #n "_" #crc, id + sm->msg_id_base); - foreach_vl_msg_name_crc_ioam_export; -#undef _ -} - -static clib_error_t * -set_ioam_export_ipfix_command_fn (vlib_main_t * vm, - unformat_input_t * input, - vlib_cli_command_t * cmd) -{ - ioam_export_main_t *em = &ioam_export_main; - ip4_address_t collector, src; - u8 is_disable = 0; - - collector.as_u32 = 0; - src.as_u32 = 0; - - while (unformat_check_input (input) != UNFORMAT_END_OF_INPUT) - { - if (unformat (input, "collector %U", unformat_ip4_address, &collector)) - ; - else if (unformat (input, "src %U", unformat_ip4_address, &src)) - ; - else if (unformat (input, "disable")) - is_disable = 1; - else - break; - } - - if (collector.as_u32 == 0) - return clib_error_return (0, "collector address required"); - - if (src.as_u32 == 0) - return clib_error_return (0, "src address required"); - - em->ipfix_collector.as_u32 = collector.as_u32; - em->src_address.as_u32 = src.as_u32; - - vlib_cli_output (vm, "Collector %U, src address %U", - format_ip4_address, &em->ipfix_collector, - format_ip4_address, &em->src_address); - - /* Turn on the export timer process */ - // vlib_process_signal_event (vm, flow_report_process_node.index, - //1, 0); - ioam_export_ip6_enable_disable (em, is_disable, &collector, &src); - - return 0; -} - -/* *INDENT-OFF* */ -VLIB_CLI_COMMAND (set_ipfix_command, static) = -{ -.path = "set ioam export ipfix",.short_help = - "set ioam export ipfix collector src ",. - function = set_ioam_export_ipfix_command_fn,}; -/* *INDENT-ON* */ - - -static clib_error_t * -ioam_export_init (vlib_main_t * vm) -{ - ioam_export_main_t *em = &ioam_export_main; - clib_error_t *error = 0; - u8 *name; - u32 node_index = export_node.index; - vlib_node_t *ip6_hbyh_node = NULL; - - name = format (0, "ioam_export_%08x%c", api_version, 0); - - /* Ask for a correctly-sized block of API message decode slots */ - em->msg_id_base = vl_msg_api_get_msg_ids - ((char *) name, VL_MSG_FIRST_AVAILABLE); - em->unix_time_0 = (u32) time (0); /* Store starting time */ - em->vlib_time_0 = vlib_time_now (vm); - - error = ioam_export_plugin_api_hookup (vm); - - /* Add our API messages to the global name_crc hash table */ - setup_message_id_table (em, &api_main); - - /* Hook this export node to ip6-hop-by-hop */ - ip6_hbyh_node = vlib_get_node_by_name (vm, (u8 *) "ip6-hop-by-hop"); - em->my_hbh_slot = vlib_node_add_next (vm, ip6_hbyh_node->index, node_index); - vec_free (name); - - return error; -} - -VLIB_INIT_FUNCTION (ioam_export_init); - -/* - * fd.io coding-style-patch-verification: ON - * - * Local Variables: - * eval: (c-set-style "gnu") - * End: - */ diff --git a/plugins/ioam-plugin/ioam/export/ioam_export_all_api_h.h b/plugins/ioam-plugin/ioam/export/ioam_export_all_api_h.h deleted file mode 100644 index bc4368f2..00000000 --- a/plugins/ioam-plugin/ioam/export/ioam_export_all_api_h.h +++ /dev/null @@ -1,16 +0,0 @@ -/* - * Copyright (c) 2016 Cisco and/or its affiliates. - * Licensed under the Apache License, Version 2.0 (the "License"); - * you may not use this file except in compliance with the License. - * You may obtain a copy of the License at: - * - * http://www.apache.org/licenses/LICENSE-2.0 - * - * Unless required by applicable law or agreed to in writing, software - * distributed under the License is distributed on an "AS IS" BASIS, - * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. - * See the License for the specific language governing permissions and - * limitations under the License. - */ -/* Include the generated file, see BUILT_SOURCES in Makefile.am */ -#include diff --git a/plugins/ioam-plugin/ioam/export/ioam_export_msg_enum.h b/plugins/ioam-plugin/ioam/export/ioam_export_msg_enum.h deleted file mode 100644 index c2de7988..00000000 --- a/plugins/ioam-plugin/ioam/export/ioam_export_msg_enum.h +++ /dev/null @@ -1,28 +0,0 @@ -/* - * Copyright (c) 2016 Cisco and/or its affiliates. - * Licensed under the Apache License, Version 2.0 (the "License"); - * you may not use this file except in compliance with the License. - * You may obtain a copy of the License at: - * - * http://www.apache.org/licenses/LICENSE-2.0 - * - * Unless required by applicable law or agreed to in writing, software - * distributed under the License is distributed on an "AS IS" BASIS, - * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. - * See the License for the specific language governing permissions and - * limitations under the License. - */ -#ifndef included_ioam_export_msg_enum_h -#define included_ioam_export_msg_enum_h - -#include - -#define vl_msg_id(n,h) n, -typedef enum { -#include - /* We'll want to know how many messages IDs we need... */ - VL_MSG_FIRST_AVAILABLE, -} vl_msg_id_t; -#undef vl_msg_id - -#endif /* included_ioam_export_msg_enum_h */ diff --git a/plugins/ioam-plugin/ioam/export/ioam_export_test.c b/plugins/ioam-plugin/ioam/export/ioam_export_test.c deleted file mode 100644 index f991fc0c..00000000 --- a/plugins/ioam-plugin/ioam/export/ioam_export_test.c +++ /dev/null @@ -1,206 +0,0 @@ -/* - * Copyright (c) 2016 Cisco and/or its affiliates. - * Licensed under the Apache License, Version 2.0 (the "License"); - * you may not use this file except in compliance with the License. - * You may obtain a copy of the License at: - * - * http://www.apache.org/licenses/LICENSE-2.0 - * - * Unless required by applicable law or agreed to in writing, software - * distributed under the License is distributed on an "AS IS" BASIS, - * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. - * See the License for the specific language governing permissions and - * limitations under the License. - */ -/* - *------------------------------------------------------------------ - * ioam_export_test.c - test harness plugin - *------------------------------------------------------------------ - */ - -#include -#include -#include -#include -#include - - -/* Declare message IDs */ -#include - -/* define message structures */ -#define vl_typedefs -#include -#undef vl_typedefs - -/* declare message handlers for each api */ - -#define vl_endianfun /* define message structures */ -#include -#undef vl_endianfun - -/* instantiate all the print functions we know about */ -#define vl_print(handle, ...) -#define vl_printfun -#include -#undef vl_printfun - -/* Get the API version number. */ -#define vl_api_version(n,v) static u32 api_version=(v); -#include -#undef vl_api_version - - -typedef struct -{ - /* API message ID base */ - u16 msg_id_base; - vat_main_t *vat_main; -} export_test_main_t; - -export_test_main_t export_test_main; - -#define foreach_standard_reply_retval_handler \ -_(ioam_export_ip6_enable_disable_reply) - -#define _(n) \ - static void vl_api_##n##_t_handler \ - (vl_api_##n##_t * mp) \ - { \ - vat_main_t * vam = export_test_main.vat_main; \ - i32 retval = ntohl(mp->retval); \ - if (vam->async_mode) { \ - vam->async_errors += (retval < 0); \ - } else { \ - vam->retval = retval; \ - vam->result_ready = 1; \ - } \ - } -foreach_standard_reply_retval_handler; -#undef _ - -/* - * Table of message reply handlers, must include boilerplate handlers - * we just generated - */ -#define foreach_vpe_api_reply_msg \ -_(IOAM_EXPORT_IP6_ENABLE_DISABLE_REPLY, ioam_export_ip6_enable_disable_reply) - - -/* M: construct, but don't yet send a message */ - -#define M(T,t) \ -do { \ - vam->result_ready = 0; \ - mp = vl_msg_api_alloc(sizeof(*mp)); \ - memset (mp, 0, sizeof (*mp)); \ - mp->_vl_msg_id = ntohs (VL_API_##T + sm->msg_id_base); \ - mp->client_index = vam->my_client_index; \ -} while(0); - -#define M2(T,t,n) \ -do { \ - vam->result_ready = 0; \ - mp = vl_msg_api_alloc(sizeof(*mp)+(n)); \ - memset (mp, 0, sizeof (*mp)); \ - mp->_vl_msg_id = ntohs (VL_API_##T + sm->msg_id_base); \ - mp->client_index = vam->my_client_index; \ -} while(0); - -/* S: send a message */ -#define S (vl_msg_api_send_shmem (vam->vl_input_queue, (u8 *)&mp)) - -/* W: wait for results, with timeout */ -#define W \ -do { \ - timeout = vat_time_now (vam) + 1.0; \ - \ - while (vat_time_now (vam) < timeout) { \ - if (vam->result_ready == 1) { \ - return (vam->retval); \ - } \ - } \ - return -99; \ -} while(0); - -static int -api_ioam_export_ip6_enable_disable (vat_main_t * vam) -{ - export_test_main_t *sm = &export_test_main; - unformat_input_t *i = vam->input; - f64 timeout; - int is_disable = 0; - vl_api_ioam_export_ip6_enable_disable_t *mp; - - /* Parse args required to build the message */ - while (unformat_check_input (i) != UNFORMAT_END_OF_INPUT) - { - if (unformat (i, "disable")) - is_disable = 1; - else - break; - } - - /* Construct the API message */ - M (IOAM_EXPORT_IP6_ENABLE_DISABLE, ioam_export_ip6_enable_disable); - mp->is_disable = is_disable; - - /* send it... */ - S; - - /* Wait for a reply... */ - W; -} - -/* - * List of messages that the api test plugin sends, - * and that the data plane plugin processes - */ -#define foreach_vpe_api_msg \ -_(ioam_export_ip6_enable_disable, " [disable]") - -void -vat_api_hookup (vat_main_t * vam) -{ - export_test_main_t *sm = &export_test_main; - /* Hook up handlers for replies from the data plane plug-in */ -#define _(N,n) \ - vl_msg_api_set_handlers((VL_API_##N + sm->msg_id_base), \ - #n, \ - vl_api_##n##_t_handler, \ - vl_noop_handler, \ - vl_api_##n##_t_endian, \ - vl_api_##n##_t_print, \ - sizeof(vl_api_##n##_t), 1); - foreach_vpe_api_reply_msg; -#undef _ - - /* API messages we can send */ -#define _(n,h) hash_set_mem (vam->function_by_name, #n, api_##n); - foreach_vpe_api_msg; -#undef _ - - /* Help strings */ -#define _(n,h) hash_set_mem (vam->help_by_name, #n, h); - foreach_vpe_api_msg; -#undef _ -} - -clib_error_t * -vat_plugin_register (vat_main_t * vam) -{ - export_test_main_t *sm = &export_test_main; - u8 *name; - - sm->vat_main = vam; - - name = format (0, "ioam_export_%08x%c", api_version, 0); - sm->msg_id_base = vl_client_get_first_plugin_msg_id ((char *) name); - - if (sm->msg_id_base != (u16) ~ 0) - vat_api_hookup (vam); - - vec_free (name); - - return 0; -} diff --git a/plugins/ioam-plugin/ioam/export/ioam_export_thread.c b/plugins/ioam-plugin/ioam/export/ioam_export_thread.c deleted file mode 100644 index d2eb2009..00000000 --- a/plugins/ioam-plugin/ioam/export/ioam_export_thread.c +++ /dev/null @@ -1,38 +0,0 @@ -/* - * Copyright (c) 2016 Cisco and/or its affiliates. - * Licensed under the Apache License, Version 2.0 (the "License"); - * you may not use this file except in compliance with the License. - * You may obtain a copy of the License at: - * - * http://www.apache.org/licenses/LICENSE-2.0 - * - * Unless required by applicable law or agreed to in writing, software - * distributed under the License is distributed on an "AS IS" BASIS, - * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. - * See the License for the specific language governing permissions and - * limitations under the License. - */ -/* - * ioam_export_thread.c - */ -#include -#include -#include - -static vlib_node_registration_t ioam_export_process_node; - -static uword -ioam_export_process (vlib_main_t * vm, - vlib_node_runtime_t * rt, vlib_frame_t * f) -{ - return (ioam_export_process_common(&ioam_export_main, - vm, rt, f, - ioam_export_process_node.index)); -} - -VLIB_REGISTER_NODE (ioam_export_process_node, static) = -{ - .function = ioam_export_process, - .type = VLIB_NODE_TYPE_PROCESS, - .name = "ioam-export-process", -}; diff --git a/plugins/ioam-plugin/ioam/export/jvpp_ioam_export.c b/plugins/ioam-plugin/ioam/export/jvpp_ioam_export.c deleted file mode 100644 index 27d3e214..00000000 --- a/plugins/ioam-plugin/ioam/export/jvpp_ioam_export.c +++ /dev/null @@ -1,124 +0,0 @@ -/* - * Copyright (c) 2016 Cisco and/or its affiliates. - * Licensed under the Apache License, Version 2.0 (the "License"); - * you may not use this file except in compliance with the License. - * You may obtain a copy of the License at: - * - * http://www.apache.org/licenses/LICENSE-2.0 - * - * Unless required by applicable law or agreed to in writing, software - * distributed under the License is distributed on an "AS IS" BASIS, - * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. - * See the License for the specific language governing permissions and - * limitations under the License. - */ - -#include - -#include -#define vl_typedefs /* define message structures */ -#include -#undef vl_typedefs - -#define vl_endianfun -#include -#undef vl_endianfun - -#define vl_print(handle, ...) -#define vl_printfun -#include -#undef vl_printfun - -/* Get the API version number */ -#define vl_api_version(n,v) static u32 api_version=(v); -#include -#undef vl_api_version - -#include -#include -#include - -#if VPPJNI_DEBUG == 1 - #define DEBUG_LOG(...) clib_warning(__VA_ARGS__) -#else - #define DEBUG_LOG(...) -#endif - -#include - -#include "ioam/jvpp/io_fd_vpp_jvpp_ioamexport_JVppIoamexportImpl.h" -#include "jvpp_ioam_export.h" -#include "ioam/jvpp/jvpp_ioam_export_gen.h" - -/* - * Class: io_fd_vpp_jvpp_ioamexport_JVppIoamexportImpl - * Method: init0 - * Signature: (JI)V - */ -JNIEXPORT void JNICALL Java_io_fd_vpp_jvpp_ioamexport_JVppIoamexportImpl_init0 - (JNIEnv *env, jclass clazz, jobject callback, jlong queue_address, jint my_client_index) { - ioamexport_main_t * plugin_main = &ioamexport_main; - u8 * name; - clib_warning ("Java_io_fd_vpp_jvpp_ioamexport_JVppIoamexportImpl_init0"); - - plugin_main->my_client_index = my_client_index; - plugin_main->vl_input_queue = (unix_shared_memory_queue_t *)queue_address; - - name = format (0, "ioam_export_%08x%c", api_version, 0); - plugin_main->msg_id_base = vl_client_get_first_plugin_msg_id ((char *) name); - - if (plugin_main->msg_id_base == (u16) ~0) { - jclass exClass = (*env)->FindClass(env, "java/lang/IllegalStateException"); - (*env)->ThrowNew(env, exClass, "ioam_export plugin is not loaded in VPP"); - } else { - plugin_main->callbackObject = (*env)->NewGlobalRef(env, callback); - plugin_main->callbackClass = (jclass)(*env)->NewGlobalRef(env, (*env)->GetObjectClass(env, callback)); - - #define _(N,n) \ - vl_msg_api_set_handlers(VL_API_##N + plugin_main->msg_id_base, #n, \ - vl_api_##n##_t_handler, \ - vl_noop_handler, \ - vl_api_##n##_t_endian, \ - vl_api_##n##_t_print, \ - sizeof(vl_api_##n##_t), 1); - foreach_api_reply_handler; - #undef _ - } -} - -JNIEXPORT void JNICALL Java_io_fd_vpp_jvpp_ioamexport_JVppIoamexportImpl_close0 -(JNIEnv *env, jclass clazz) { - ioamexport_main_t * plugin_main = &ioamexport_main; - - // cleanup: - (*env)->DeleteGlobalRef(env, plugin_main->callbackClass); - (*env)->DeleteGlobalRef(env, plugin_main->callbackObject); - - plugin_main->callbackClass = NULL; - plugin_main->callbackObject = NULL; -} - -/* Attach thread to JVM and cache class references when initiating JVPP iOAM EXPORT */ -jint JNI_OnLoad(JavaVM *vm, void *reserved) { - JNIEnv* env; - - if ((*vm)->GetEnv(vm, (void**) &env, JNI_VERSION_1_8) != JNI_OK) { - return JNI_EVERSION; - } - - if (cache_class_references(env) != 0) { - clib_warning ("Failed to cache class references\n"); - return JNI_ERR; - } - - return JNI_VERSION_1_8; -} - -/* Clean up cached references when disposing JVPP iOAM EXPORT */ -void JNI_OnUnload(JavaVM *vm, void *reserved) { - JNIEnv* env; - if ((*vm)->GetEnv(vm, (void**) &env, JNI_VERSION_1_8) != JNI_OK) { - return; - } - delete_class_references(env); -} diff --git a/plugins/ioam-plugin/ioam/export/jvpp_ioam_export.h b/plugins/ioam-plugin/ioam/export/jvpp_ioam_export.h deleted file mode 100644 index b6c0c16e..00000000 --- a/plugins/ioam-plugin/ioam/export/jvpp_ioam_export.h +++ /dev/null @@ -1,45 +0,0 @@ -/* - * Copyright (c) 2016 Cisco and/or its affiliates. - * Licensed under the Apache License, Version 2.0 (the "License"); - * you may not use this file except in compliance with the License. - * You may obtain a copy of the License at: - * - * http://www.apache.org/licenses/LICENSE-2.0 - * - * Unless required by applicable law or agreed to in writing, software - * distributed under the License is distributed on an "AS IS" BASIS, - * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. - * See the License for the specific language governing permissions and - * limitations under the License. - */ -#ifndef __included_jvpp_ioam_export_h__ -#define __included_jvpp_ioam_export_h__ - -#include -#include -#include -#include -#include -#include - -/* Global state for JVPP-IOAM-EXPORT */ -typedef struct { - /* Base message index for the export plugin */ - u16 msg_id_base; - - /* Pointer to shared memory queue */ - unix_shared_memory_queue_t * vl_input_queue; - - /* VPP api client index */ - u32 my_client_index; - - /* Callback object and class references enabling asynchronous Java calls */ - jobject callbackObject; - jclass callbackClass; - -} ioamexport_main_t; - -ioamexport_main_t ioamexport_main __attribute__((aligned (64))); - - -#endif /* __included_jvpp_ioam_export_h__ */ diff --git a/plugins/ioam-plugin/ioam/export/node.c b/plugins/ioam-plugin/ioam/export/node.c deleted file mode 100644 index 19f143df..00000000 --- a/plugins/ioam-plugin/ioam/export/node.c +++ /dev/null @@ -1,151 +0,0 @@ -/* - * Copyright (c) 2016 Cisco and/or its affiliates. - * Licensed under the Apache License, Version 2.0 (the "License"); - * you may not use this file except in compliance with the License. - * You may obtain a copy of the License at: - * - * http://www.apache.org/licenses/LICENSE-2.0 - * - * Unless required by applicable law or agreed to in writing, software - * distributed under the License is distributed on an "AS IS" BASIS, - * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. - * See the License for the specific language governing permissions and - * limitations under the License. - */ -#include -#include -#include -#include -#include -#include - -typedef struct -{ - u32 next_index; - u32 flow_label; -} export_trace_t; - -/* packet trace format function */ -static u8 * -format_export_trace (u8 * s, va_list * args) -{ - CLIB_UNUSED (vlib_main_t * vm) = va_arg (*args, vlib_main_t *); - CLIB_UNUSED (vlib_node_t * node) = va_arg (*args, vlib_node_t *); - export_trace_t *t = va_arg (*args, export_trace_t *); - - s = format (s, "EXPORT: flow_label %d, next index %d", - t->flow_label, t->next_index); - return s; -} - -vlib_node_registration_t export_node; - -#define foreach_export_error \ -_(RECORDED, "Packets recorded for export") - -typedef enum -{ -#define _(sym,str) EXPORT_ERROR_##sym, - foreach_export_error -#undef _ - EXPORT_N_ERROR, -} export_error_t; - -static char *export_error_strings[] = { -#define _(sym,string) string, - foreach_export_error -#undef _ -}; - -typedef enum -{ - EXPORT_NEXT_POP_HBYH, - EXPORT_N_NEXT, -} export_next_t; - -always_inline void -copy3cachelines (void *dst, const void *src, size_t n) -{ -#if 0 - if (PREDICT_FALSE (n < DEFAULT_EXPORT_SIZE)) - { - /* Copy only the first 1/2 cache lines whatever is available */ - if (n >= 64) - clib_mov64 ((u8 *) dst, (const u8 *) src); - if (n >= 128) - clib_mov64 ((u8 *) dst + 64, (const u8 *) src + 64); - return; - } - clib_mov64 ((u8 *) dst, (const u8 *) src); - clib_mov64 ((u8 *) dst + 64, (const u8 *) src + 64); - clib_mov64 ((u8 *) dst + 128, (const u8 *) src + 128); -#endif -#if 1 - - u64 *copy_dst, *copy_src; - int i; - copy_dst = (u64 *) dst; - copy_src = (u64 *) src; - if (PREDICT_FALSE (n < DEFAULT_EXPORT_SIZE)) - { - for (i = 0; i < n / 64; i++) - { - copy_dst[0] = copy_src[0]; - copy_dst[1] = copy_src[1]; - copy_dst[2] = copy_src[2]; - copy_dst[3] = copy_src[3]; - copy_dst[4] = copy_src[4]; - copy_dst[5] = copy_src[5]; - copy_dst[6] = copy_src[6]; - copy_dst[7] = copy_src[7]; - copy_dst += 8; - copy_src += 8; - } - return; - } - for (i = 0; i < 3; i++) - { - copy_dst[0] = copy_src[0]; - copy_dst[1] = copy_src[1]; - copy_dst[2] = copy_src[2]; - copy_dst[3] = copy_src[3]; - copy_dst[4] = copy_src[4]; - copy_dst[5] = copy_src[5]; - copy_dst[6] = copy_src[6]; - copy_dst[7] = copy_src[7]; - copy_dst += 8; - copy_src += 8; - } -#endif -} - -static uword -ip6_export_node_fn (vlib_main_t * vm, - vlib_node_runtime_t * node, vlib_frame_t * frame) -{ - ioam_export_main_t *em = &ioam_export_main; - ioam_export_node_common(em, vm, node, frame, ip6_header_t, payload_length, - ip_version_traffic_class_and_flow_label, - EXPORT_NEXT_POP_HBYH); - return frame->n_vectors; -} - -/* - * Node for IP6 export - */ -VLIB_REGISTER_NODE (export_node) = -{ - .function = ip6_export_node_fn, - .name = "ip6-export", - .vector_size = sizeof (u32), - .format_trace = format_export_trace, - .type = VLIB_NODE_TYPE_INTERNAL, - .n_errors = ARRAY_LEN (export_error_strings), - .error_strings = export_error_strings, - .n_next_nodes = EXPORT_N_NEXT, - /* edit / add dispositions here */ - .next_nodes = - { - [EXPORT_NEXT_POP_HBYH] = "ip6-pop-hop-by-hop" - }, -}; diff --git a/plugins/ioam-plugin/ioam/ioam_plugin_doc.md b/plugins/ioam-plugin/ioam/ioam_plugin_doc.md deleted file mode 100644 index 343abcf7..00000000 --- a/plugins/ioam-plugin/ioam/ioam_plugin_doc.md +++ /dev/null @@ -1,464 +0,0 @@ -## VPP Inband OAM (iOAM) {#ioam_plugin_doc} - -In-band OAM (iOAM) is an implementation study to record operational -information in the packet while the packet traverses a path between -two points in the network. - -Overview of iOAM can be found in [iOAM-Devnet] page. -The following IETF drafts detail the motivation and mechanism for -recording operational information: - - [iOAM-ietf-requirements] - Describes motivation and usecases for iOAM - - [iOAM-ietf-data] - Describes data records that can be collected using iOAM - - [iOAM-ietf-transport] - Lists out the transport protocols - and mechanism to carry iOAM data records - - [iOAM-ietf-proof-of-transit] - Describes the idea of Proof of Transit (POT) - and mechanisms to operationalize the idea - -## Terminology -In-band OAM is expected to be deployed in a specific domain rather -than on the overall Internet. The part of the network which employs in-band OAM -is referred to as **"in-band OAM-domain"**. - -In-band OAM data is added to a packet on entering the in-band OAM-domain -and is removed from the packet when exiting the domain. -Within the in-band OAM-domain, network nodes that the packet traverses -may update the in-band OAM data records. - -- The node which adds in-band OAM data to the packet is called the -**"in-band OAM encapsulating node"**. - -- The node which removes the in-band OAM data is referred to as the -**"in-band OAM decapsulating node"**. - -- Nodes within the domain which are aware of in-band OAM data and read -and/or write or process the in-band OAM data are called -**"in-band OAM transit nodes"**. - -## Features supported in the current release -VPP can function as in-band OAM encapsulating, transit and decapsulating node. -In this version of VPP in-band OAM data is transported as options in an -IPv6 hop-by-hop extension header. Hence in-band OAM can be enabled -for IPv6 traffic. - -The following iOAM features are supported: - -- **In-band OAM Tracing** : In-band OAM supports multiple data records to be -recorded in the packet as the packet traverses the network. -These data records offer insights into the operational behavior of the network. -The following information can be collected in the tracing -data from the nodes a packet traverses: - - Node ID - - Ingress interface ID - - Egress interface ID - - Timestamp - - Pre-configured application data - -- **In-band OAM Proof of Transit (POT)**: Proof of transit iOAM data is -added to every packet for verifying that a packet traverses a specific -set of nodes. -In-band OAM data is updated at every node that is enabled with iOAM -proof of transit and is used to verify whether a packet traversed -all the specified nodes. When the verifier receives each packet, -it can validate whether the packet traversed the specified nodes. - - -## Configuration -Configuring iOAM involves: -- Selecting the packets for which iOAM data must be inserted, updated or removed - - Selection of packets for iOAM data insertion on iOAM encapsulating node. - Selection of packets is done by 5-tuple based classification - - Selection of packets for updating iOAM data is implicitly done on the - presence of iOAM options in the packet - - Selection of packets for removing the iOAM data is done on 5-tuple - based classification -- The kind of data to be collected - - Tracing data - - Proof of transit -- Additional details for processing iOAM data to be collected - - For trace data - trace type, number of nodes to be recorded in the trace, - time stamp precision, etc. - - For POT data - configuration of POT profile required to process the POT data - -The CLI for configuring iOAM is explained here followed by detailed steps -and examples to deploy iOAM on VPP as an encapsulating, transit or -decapsulating iOAM node in the subsequent sub-sections. - -VPP iOAM configuration for enabling trace and POT is as follows: - - set ioam rewrite trace-type <0x1f|0x7|0x9|0x11|0x19> - trace-elts trace-tsp <0|1|2|3> - node-id app-data [pot] - -A description of each of the options of the CLI follows: -- trace-type : An entry in the "Node data List" array of the trace option -can have different formats, following the needs of the a deployment. -For example: Some deployments might only be interested -in recording the node identifiers, whereas others might be interested -in recording node identifier and timestamp. -The following types are currently supported: - - 0x1f : Node data to include hop limit (8 bits), node ID (24 bits), - ingress and egress interface IDs (16 bits each), timestamp (32 bits), - application data (32 bits) - - 0x7 : Node data to include hop limit (8 bits), node ID (24 bits), - ingress and egress interface IDs (16 bits each) - - 0x9 : Node data to include hop limit (8 bits), node ID (24 bits), - timestamp (32 bits) - - 0x11: Node data to include hop limit (8 bits), node ID (24 bits), - application data (32 bits) - - 0x19: Node data to include hop limit (8 bits), node ID (24 bits), - timestamp (32 bits), application data (32 bits) -- trace-elts : Defines the length of the node data array in the trace option. -- trace-tsp : Defines the timestamp precision to use with the enumerated value - for precision as follows: - - 0 : 32bits timestamp in seconds - - 1 : 32bits timestamp in milliseconds - - 2 : 32bits timestamp in microseconds - - 3 : 32bits timestamp in nanoseconds -- node-id : Unique identifier for the node, included in the node ID - field of the node data in trace option. -- app-data : The value configured here is included as is in -application data field of node data in trace option. -- pot : Enables POT option to be included in the iOAM options. - -### Trace configuration - -#### On in-band OAM encapsulating node - - **Configure classifier and apply ACL** to select packets for - iOAM data insertion - - Example to enable iOAM data insertion for all the packets - towards IPv6 address db06::06: - - vpp# classify table miss-next node ip6-lookup mask l3 ip6 dst - - vpp# classify session acl-hit-next node ip6-add-hop-by-hop - table-index 0 match l3 ip6 dst db06::06 - - vpp# set int input acl intfc GigabitEthernet0/0/0 ip6-table 0 - - - **Enable tracing** : Specify node ID, maximum number of nodes for which - trace data should be recorded, type of data to be included for recording, - optionally application data to be included - - Example to enable tracing with a maximum of 4 nodes recorded - and the data to be recorded to include - hop limit, node id, - ingress and egress interface IDs, timestamp (millisecond precision), - application data (0x1234): - - - vpp# set ioam rewrite trace-type 0x1f trace-elts 4 trace-tsp 1 - node-id 0x1 app-data 0x1234 - - - -#### On in-band OAM transit node -- The transit node requires trace type, timestamp precision, node ID and -optionally application data to be configured, -to update its node data in the trace option. - -Example: - - vpp# set ioam rewrite trace-type 0x1f trace-elts 4 trace-tsp 1 - node-id 0x2 app-data 0x1234 - -#### On the In-band OAM decapsulating node -- The decapsulating node similar to encapsulating node requires -**classification** of the packets to remove iOAM data from. - - Example to decapsulate iOAM data for packets towards - db06::06, configure classifier and enable it as an ACL as follows: - - - vpp# classify table miss-next node ip6-lookup mask l3 ip6 dst - - vpp# classify session acl-hit-next node ip6-lookup table-index 0 - match l3 ip6 dst db06::06 opaque-index 100 - - vpp# set int input acl intfc GigabitEthernet0/0/0 ip6-table 0 - - -- Decapsulating node requires trace type, timestamp precision, -node ID and optionally application data to be configured, -to update its node data in the trace option before it is decapsulated. - -Example: - - vpp# set ioam rewrite trace-type 0x1f trace-elts 4 - trace-tsp 1 node-id 0x3 app-data 0x1234 - - -### Proof of Transit configuration - -For details on proof-of-transit, -see the IETF draft [iOAM-ietf-proof-of-transit]. -To enable Proof of Transit all the nodes that participate -and hence are verified for transit need a proof of transit profile. -A script to generate a proof of transit profile as per the mechanism -described in [iOAM-ietf-proof-of-transit] will be available at [iOAM-Devnet]. - -The Proof of transit mechanism implemented here is based on -Shamir's Secret Sharing algorithm. -The overall algorithm uses two polynomials -POLY-1 and POLY-2. The degree of polynomials depends on number of nodes -to be verified for transit. -POLY-1 is secret and constant. Each node gets a point on POLY-1 -at setup-time and keeps it secret. -POLY-2 is public, random and per packet. -Each node is assigned a point on POLY-1 and POLY-2 with the same x index. -Each node derives its point on POLY-2 each time a packet arrives at it. -A node then contributes its points on POLY-1 and POLY-2 to construct -POLY-3 (POLY-3 = POLY-1 + POLY-2) using lagrange extrapolation and -forwards it towards the verifier by updating POT data in the packet. -The verifier constructs POLY-3 from the accumulated value from all the nodes -and its own points on POLY-1 and POLY-2 and verifies whether -POLY-3 = POLY-1 + POLY-2. Only the verifier knows POLY-1. -The solution leverages finite field arithmetic in a field of size "prime number" -for reasons explained in description of Shamir's secret sharing algorithm. - -Here is an explanation of POT profile list and profile configuration CLI to -realize the above mechanism. -It is best to use the script provided at [iOAM-Devnet] to generate -this configuration. -- **Create POT profile** : set pot profile name id [0-1] -[validator-key 0xu64] prime-number 0xu64 secret_share 0xu64 -lpc 0xu64 polynomial2 0xu64 bits-in-random [0-64] - - name : Profile list name. - - id : Profile id, it can be 0 or 1. - A maximum of two profiles can be configured per profile list. - - validator-key : Secret key configured only on the - verifier/decapsulating node used to compare and verify proof of transit. - - prime-number : Prime number for finite field arithmetic as required by the - proof of transit mechanism. - - secret_share : Unique point for each node on the secret polynomial POLY-1. - - lpc : Lagrange Polynomial Constant(LPC) calculated per node based on - its point (x value used for evaluating the points on the polynomial) - on the polynomial used in lagrange extrapolation - for reconstructing polynomial (POLY-3). - - polynomial2 : Is the pre-evaluated value of the point on - 2nd polynomial(POLY-2). This is unique for each node. - It is pre-evaluated for all the coefficients of POLY-2 except - for the constant part of the polynomial that changes per packet - and is received as part of the POT data in the packet. - - bits-in-random : To control the size of the random number to be - generated. This number has to match the other numbers generated and used - in the profile as per the algorithm. - -- **Set a configured profile as active/in-use** : -set pot profile-active name ID [0-1] - - name : Name of the profile list to be used for computing - POT data per packet. - - ID : Identifier of the profile within the list to be used. - -#### On In-band OAM encapsulating node - - Configure the classifier and apply ACL to select packets for iOAM data insertion. - - Example to enable iOAM data insertion for all the packet towards - IPv6 address db06::06 - - - - vpp# classify table miss-next node ip6-lookup mask l3 ip6 dst - - vpp# classify session acl-hit-next node - ip6-add-hop-by-hop table-index 0 match l3 ip6 dst db06::06 - - vpp# set int input acl intfc GigabitEthernet0/0/0 ip6-table 0 - - - - Configure the proof of transit profile list with profiles. -Each profile list referred to by a name can contain 2 profiles, -only one is in use for updating proof of transit data at any time. - - Example profile list example with a profile generated from the - script to verify transit through 3 nodes is: - - - vpp# set pot profile name example id 0 prime-number 0x7fff0000fa884685 - secret_share 0x6c22eff0f45ec56d lpc 0x7fff0000fa884682 - polynomial2 0xffb543d4a9c bits-in-random 63 - - - Enable one of the profiles from the configured profile list as active - so that is will be used for calculating proof of transit - -Example enable profile ID 0 from profile list example configured above: - - - vpp# set pot profile-active name example ID 0 - - - - Enable POT option to be inserted - - - vpp# set ioam rewrite pot - - -#### On in-band OAM transit node - - Configure the proof of transit profile list with profiles for transit node. -Example: - - - vpp# set pot profile name example id 0 prime-number 0x7fff0000fa884685 - secret_share 0x564cdbdec4eb625d lpc 0x1 - polynomial2 0x23f3a227186a bits-in-random 63 - -#### On in-band OAM decapsulating node / verifier -- The decapsulating node, similar to the encapsulating node requires -classification of the packets to remove iOAM data from. - - Example to decapsulate iOAM data for packets towards db06::06 - configure classifier and enable it as an ACL as follows: - - - vpp# classify table miss-next node ip6-lookup mask l3 ip6 dst - - vpp# classify session acl-hit-next node ip6-lookup table-index 0 - match l3 ip6 dst db06::06 opaque-index 100 - - vpp# set int input acl intfc GigabitEthernet0/0/0 ip6-table 0 - -- To update and verify the proof of transit, POT profile list should be configured. - - Example POT profile list configured as follows: - - vpp# set pot profile name example id 0 validate-key 0x7fff0000fa88465d - prime-number 0x7fff0000fa884685 secret_share 0x7a08fbfc5b93116d lpc 0x3 - polynomial2 0x3ff738597ce bits-in-random 63 - -## Operational data - -Following CLIs are available to check iOAM operation: -- To check iOAM configuration that are effective use "show ioam summary" - -Example: - - vpp# show ioam summary - REWRITE FLOW CONFIGS - Not configured - HOP BY HOP OPTIONS - TRACE CONFIG - - Trace Type : 0x1f (31) - Trace timestamp precision : 1 (Milliseconds) - Num of trace nodes : 4 - Node-id : 0x2 (2) - App Data : 0x1234 (4660) - POT OPTION - 1 (Enabled) - Try 'show ioam pot and show pot profile' for more information - -- To find statistics about packets for which iOAM options were -added (encapsulating node) and removed (decapsulating node) execute -*show errors* - -Example on encapsulating node: - - - vpp# show error - Count Node Reason - 1208804706 ip6-inacl input ACL hits - 1208804706 ip6-add-hop-by-hop Pkts w/ added ip6 hop-by-hop options - -Example on decapsulating node: - - vpp# show error - Count Node Reason - 69508569 ip6-inacl input ACL hits - 69508569 ip6-pop-hop-by-hop Pkts w/ removed ip6 hop-by-hop options - -- To check the POT profiles use "show pot profile" - -Example: - - vpp# show pot profile - Profile list in use : example - POT Profile at index: 0 - ID : 0 - Validator : False (0) - Secret share : 0x564cdbdec4eb625d (6218586935324795485) - Prime number : 0x7fff0000fa884685 (9223090566081300101) - 2nd polynomial(eval) : 0x23f3a227186a (39529304496234) - LPC : 0x1 (1) - Bit mask : 0x7fffffffffffffff (9223372036854775807) - Profile index in use: 0 - Pkts passed : 0x36 (54) - -- To get statistics of POT for packets use "show ioam pot" - -Example at encapsulating or transit node: - - vpp# show ioam pot - Pkts with ip6 hop-by-hop POT options - 54 - Pkts with ip6 hop-by-hop POT options but no profile set - 0 - Pkts with POT in Policy - 0 - Pkts with POT out of Policy - 0 - - -Example at decapsulating/verification node: - - - vpp# show ioam pot - Pkts with ip6 hop-by-hop POT options - 54 - Pkts with ip6 hop-by-hop POT options but no profile set - 0 - Pkts with POT in Policy - 54 - Pkts with POT out of Policy - 0 - -- Tracing - enable trace of IPv6 packets to view the data inserted and -collected. - -Example when the nodes are receiving data over a DPDK interface: -Enable tracing using "trace add dpdk-input 20" and -execute "show trace" to view the iOAM data collected: - - - vpp# trace add dpdk-input 20 - - vpp# show trace - - ------------------- Start of thread 0 vpp_main ------------------- - - Packet 1 - - 00:00:19:294697: dpdk-input - GigabitEthernetb/0/0 rx queue 0 - buffer 0x10e6b: current data 0, length 214, free-list 0, totlen-nifb 0, trace 0x0 - PKT MBUF: port 0, nb_segs 1, pkt_len 214 - buf_len 2176, data_len 214, ol_flags 0x0, data_off 128, phys_addr 0xe9a35a00 - packet_type 0x0 - IP6: 00:50:56:9c:df:72 -> 00:50:56:9c:be:55 - IP6_HOP_BY_HOP_OPTIONS: db05::2 -> db06::6 - tos 0x00, flow label 0x0, hop limit 63, payload length 160 - 00:00:19:294737: ethernet-input - IP6: 00:50:56:9c:df:72 -> 00:50:56:9c:be:55 - 00:00:19:294753: ip6-input - IP6_HOP_BY_HOP_OPTIONS: db05::2 -> db06::6 - tos 0x00, flow label 0x0, hop limit 63, payload length 160 - 00:00:19:294757: ip6-lookup - fib 0 adj-idx 15 : indirect via db05::2 flow hash: 0x00000000 - IP6_HOP_BY_HOP_OPTIONS: db05::2 -> db06::6 - tos 0x00, flow label 0x0, hop limit 63, payload length 160 - 00:00:19:294802: ip6-hop-by-hop - IP6_HOP_BY_HOP: next index 5 len 96 traced 96 Trace Type 0x1f , 1 elts left - [0] ttl 0x0 node ID 0x0 ingress 0x0 egress 0x0 ts 0x0 - app 0x0 - [1] ttl 0x3e node ID 0x3 ingress 0x1 egress 0x2 ts 0xb68c2213 - app 0x1234 - [2] ttl 0x3f node ID 0x2 ingress 0x1 egress 0x2 ts 0xb68c2204 - app 0x1234 - [3] ttl 0x40 node ID 0x1 ingress 0x5 egress 0x6 ts 0xb68c2200 - app 0x1234 - POT opt present - random = 0x577a916946071950, Cumulative = 0x10b46e78a35a392d, Index = 0x0 - 00:00:19:294810: ip6-rewrite - tx_sw_if_index 1 adj-idx 14 : GigabitEthernetb/0/0 - IP6: 00:50:56:9c:be:55 -> 00:50:56:9c:df:72 flow hash: 0x00000000 - IP6: 00:50:56:9c:be:55 -> 00:50:56:9c:df:72 - IP6_HOP_BY_HOP_OPTIONS: db05::2 -> db06::6 - tos 0x00, flow label 0x0, hop limit 62, payload length 160 - 00:00:19:294814: GigabitEthernetb/0/0-output - GigabitEthernetb/0/0 - IP6: 00:50:56:9c:be:55 -> 00:50:56:9c:df:72 - IP6_HOP_BY_HOP_OPTIONS: db05::2 -> db06::6 - tos 0x00, flow label 0x0, hop limit 62, payload length 160 - 00:00:19:294820: GigabitEthernetb/0/0-tx - GigabitEthernetb/0/0 tx queue 0 - buffer 0x10e6b: current data 0, length 214, free-list 0, totlen-nifb 0, trace 0x0 - IP6: 00:50:56:9c:be:55 -> 00:50:56:9c:df:72 - - IP6_HOP_BY_HOP_OPTIONS: db05::2 -> db06::6 - - tos 0x00, flow label 0x0, hop limit 62, payload length 160 - - -[iOAM-Devnet]: -[iOAM-ietf-requirements]: -[iOAM-ietf-transport]: -[iOAM-ietf-data]: -[iOAM-ietf-proof-of-transit]: diff --git a/plugins/ioam-plugin/ioam/jvpp/io/fd/vpp/jvpp/ioamexport/test/IoamExportApiTest.java b/plugins/ioam-plugin/ioam/jvpp/io/fd/vpp/jvpp/ioamexport/test/IoamExportApiTest.java deleted file mode 100644 index cb85f005..00000000 --- a/plugins/ioam-plugin/ioam/jvpp/io/fd/vpp/jvpp/ioamexport/test/IoamExportApiTest.java +++ /dev/null @@ -1,56 +0,0 @@ -/* - * Copyright (c) 2016 Cisco and/or its affiliates. - * - * Licensed under the Apache License, Version 2.0 (the "License"); - * you may not use this file except in compliance with the License. - * You may obtain a copy of the License at: - * - * http://www.apache.org/licenses/LICENSE-2.0 - * - * Unless required by applicable law or agreed to in writing, software - * distributed under the License is distributed on an "AS IS" BASIS, - * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. - * See the License for the specific language governing permissions and - * limitations under the License. - */ - -package io.fd.vpp.jvpp.ioamexport.test; - -import java.net.InetAddress; - -import io.fd.vpp.jvpp.JVpp; -import io.fd.vpp.jvpp.JVppRegistry; -import io.fd.vpp.jvpp.JVppRegistryImpl; -import io.fd.vpp.jvpp.VppCallbackException; -import io.fd.vpp.jvpp.ioamexport.JVppIoamexportImpl; -import io.fd.vpp.jvpp.ioamexport.future.FutureJVppIoamexportFacade; -import io.fd.vpp.jvpp.ioamexport.dto.IoamExportIp6EnableDisable; -import io.fd.vpp.jvpp.ioamexport.dto.IoamExportIp6EnableDisableReply; - -public class IoamExportApiTest { - - public static void main(String[] args) throws Exception { - ioamExportTestApi(); - } - - private static void ioamExportTestApi() throws Exception { - System.out.println("Testing Java API for ioam export plugin"); - try (final JVppRegistry registry = new JVppRegistryImpl("ioamExportApiTest"); - final JVpp jvpp = new JVppIoamexportImpl()) { - FutureJVppIoamexportFacade ioamexportJvpp = new FutureJVppIoamexportFacade(registry,jvpp); - System.out.println("Sending ioam export request..."); - IoamExportIp6EnableDisable request = new IoamExportIp6EnableDisable(); - request.isDisable = 0; - InetAddress collectorAddress = InetAddress.getByName("2001:0DB8:AC10:FE01:0000:0000:0000:0000"); - InetAddress srcAddress = InetAddress.getByName("2001:0DB8:AC10:FE01:0000:0000:0000:0001"); - request.collectorAddress = collectorAddress.getAddress(); - request.srcAddress = srcAddress.getAddress(); - IoamExportIp6EnableDisableReply reply = ioamexportJvpp.ioamExportIp6EnableDisable(request).toCompletableFuture().get(); - System.out.printf("IoamExportIp6EnableDisableReply = "+reply.toString()+"%n"); - - Thread.sleep(1000); - - System.out.println("Disconnecting..."); - } - } -} diff --git a/plugins/ioam-plugin/ioam/jvpp/io/fd/vpp/jvpp/ioamexport/test/Readme.txt b/plugins/ioam-plugin/ioam/jvpp/io/fd/vpp/jvpp/ioamexport/test/Readme.txt deleted file mode 100644 index 1b38c285..00000000 --- a/plugins/ioam-plugin/ioam/jvpp/io/fd/vpp/jvpp/ioamexport/test/Readme.txt +++ /dev/null @@ -1 +0,0 @@ -sudo java -cp build-vpp_debug-native/vpp-api/java/jvpp-registry-17.01.jar:build-vpp_debug-native/plugins/ioam-plugin/jvpp-ioam-export-1.0.jar io.fd.vpp.jvpp.ioamexport.test.IoamExportApiTest diff --git a/plugins/ioam-plugin/ioam/jvpp/io/fd/vpp/jvpp/ioampot/test/IoamPotApiTest.java b/plugins/ioam-plugin/ioam/jvpp/io/fd/vpp/jvpp/ioampot/test/IoamPotApiTest.java deleted file mode 100644 index 74eb86a1..00000000 --- a/plugins/ioam-plugin/ioam/jvpp/io/fd/vpp/jvpp/ioampot/test/IoamPotApiTest.java +++ /dev/null @@ -1,75 +0,0 @@ -/* - * Copyright (c) 2016 Cisco and/or its affiliates. - * - * Licensed under the Apache License, Version 2.0 (the "License"); - * you may not use this file except in compliance with the License. - * You may obtain a copy of the License at: - * - * http://www.apache.org/licenses/LICENSE-2.0 - * - * Unless required by applicable law or agreed to in writing, software - * distributed under the License is distributed on an "AS IS" BASIS, - * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. - * See the License for the specific language governing permissions and - * limitations under the License. - */ - -package io.fd.vpp.jvpp.ioampot.test; - -import io.fd.vpp.jvpp.JVpp; -import io.fd.vpp.jvpp.JVppRegistry; -import io.fd.vpp.jvpp.JVppRegistryImpl; -import io.fd.vpp.jvpp.VppCallbackException; -import io.fd.vpp.jvpp.ioampot.JVppIoampotImpl; -import io.fd.vpp.jvpp.ioampot.callback.PotProfileAddCallback; -import io.fd.vpp.jvpp.ioampot.dto.PotProfileAdd; -import io.fd.vpp.jvpp.ioampot.dto.PotProfileAddReply; - -public class IoamPotApiTest { - - static class IoamPotTestCallback implements PotProfileAddCallback { - - @Override - public void onPotProfileAddReply(final PotProfileAddReply reply) { - System.out.printf("Received PotProfileAddReply reply: context=%d%n", - reply.context); - } - - @Override - public void onError(VppCallbackException ex) { - System.out.printf("Received onError exception: call=%s, context=%d, retval=%d%n", ex.getMethodName(), - ex.getCtxId(), ex.getErrorCode()); - } - } - - public static void main(String[] args) throws Exception { - ioamPotTestApi(); - } - - private static void ioamPotTestApi() throws Exception { - System.out.println("Testing Java API for ioam pot plugin"); - try (final JVppRegistry registry = new JVppRegistryImpl("ioamPotApiTest"); - final JVpp jvpp = new JVppIoampotImpl()) { - registry.register(jvpp, new IoamPotTestCallback()); - - System.out.println("Sending ioam pot profile add request..."); - PotProfileAdd request = new PotProfileAdd(); - request.id = 0; - request.validator = 4; - request.secretKey = 1; - request.secretShare = 2; - request.prime = 1234; - request.maxBits = 53; - request.lpc = 1234; - request.polynomialPublic = 1234; - request.listNameLen = (byte)"test pot profile".getBytes().length; - request.listName = "test pot profile".getBytes(); - final int result = jvpp.send(request); - System.out.printf("PotProfileAdd send result = %d%n", result); - - Thread.sleep(1000); - - System.out.println("Disconnecting..."); - } - } -} diff --git a/plugins/ioam-plugin/ioam/jvpp/io/fd/vpp/jvpp/ioampot/test/Readme.txt b/plugins/ioam-plugin/ioam/jvpp/io/fd/vpp/jvpp/ioampot/test/Readme.txt deleted file mode 100644 index 2323494d..00000000 --- a/plugins/ioam-plugin/ioam/jvpp/io/fd/vpp/jvpp/ioampot/test/Readme.txt +++ /dev/null @@ -1 +0,0 @@ -sudo java -cp build-vpp_debug-native/vpp-api/java/jvpp-registry-16.12.jar:build-vpp_debug-native/plugins/ioam-plugin/jvpp-ioam-pot-1.0.jar io.fd.vpp.jvpp.ioampot.test.IoamPotApiTest diff --git a/plugins/ioam-plugin/ioam/jvpp/io/fd/vpp/jvpp/ioamtrace/test/IoamTraceApiTest.java b/plugins/ioam-plugin/ioam/jvpp/io/fd/vpp/jvpp/ioamtrace/test/IoamTraceApiTest.java deleted file mode 100644 index bc8c1c3a..00000000 --- a/plugins/ioam-plugin/ioam/jvpp/io/fd/vpp/jvpp/ioamtrace/test/IoamTraceApiTest.java +++ /dev/null @@ -1,77 +0,0 @@ -/* - * Copyright (c) 2016 Cisco and/or its affiliates. - * - * Licensed under the Apache License, Version 2.0 (the "License"); - * you may not use this file except in compliance with the License. - * You may obtain a copy of the License at: - * - * http://www.apache.org/licenses/LICENSE-2.0 - * - * Unless required by applicable law or agreed to in writing, software - * distributed under the License is distributed on an "AS IS" BASIS, - * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. - * See the License for the specific language governing permissions and - * limitations under the License. - */ - -package io.fd.vpp.jvpp.ioamtrace.test; - -import io.fd.vpp.jvpp.JVpp; -import io.fd.vpp.jvpp.JVppRegistry; -import io.fd.vpp.jvpp.JVppRegistryImpl; -import io.fd.vpp.jvpp.VppCallbackException; -import io.fd.vpp.jvpp.ioamtrace.future.FutureJVppIoamtraceFacade; -import io.fd.vpp.jvpp.ioamtrace.JVppIoamtraceImpl; -import io.fd.vpp.jvpp.ioamtrace.callback.TraceProfileAddCallback; -import io.fd.vpp.jvpp.ioamtrace.dto.TraceProfileAdd; -import io.fd.vpp.jvpp.ioamtrace.dto.TraceProfileAddReply; -import io.fd.vpp.jvpp.ioamtrace.dto.TraceProfileShowConfig; -import io.fd.vpp.jvpp.ioamtrace.dto.TraceProfileShowConfigReply; - -public class IoamTraceApiTest { - - static class IoamTraceTestCallback implements TraceProfileAddCallback { - - @Override - public void onTraceProfileAddReply(final TraceProfileAddReply reply) { - System.out.printf("Received TraceProfileAddReply reply: context=%d%n", - reply.context); - } - - @Override - public void onError(VppCallbackException ex) { - System.out.printf("Received onError exception: call=%s, context=%d, retval=%d%n", ex.getMethodName(), - ex.getCtxId(), ex.getErrorCode()); - } - } - - public static void main(String[] args) throws Exception { - ioamTraceTestApi(); - } - - private static void ioamTraceTestApi() throws Exception { - System.out.println("Testing Java API for ioam trace plugin"); - try (final JVppRegistry registry = new JVppRegistryImpl("ioamTraceApiTest"); - final JVpp jvpp = new JVppIoamtraceImpl()) { - FutureJVppIoamtraceFacade ioamtraceJvpp = new FutureJVppIoamtraceFacade(registry,jvpp); - - System.out.println("Sending ioam trace profile add request..."); - TraceProfileAdd request = new TraceProfileAdd(); - request.traceType = 0x1f; - request.numElts = 4; - request.nodeId = 1; - request.traceTsp = 2; - request.appData = 1234; - final int result = jvpp.send(request); - System.out.printf("TraceProfileAdd send result = %d%n", result); - - Thread.sleep(1000); - - TraceProfileShowConfig showRequest = new TraceProfileShowConfig(); - TraceProfileShowConfigReply reply = ioamtraceJvpp.traceProfileShowConfig(showRequest).toCompletableFuture().get(); - System.out.printf("TraceProfileShowConfig result = "+ reply.toString()); - - System.out.println("Disconnecting..."); - } - } -} diff --git a/plugins/ioam-plugin/ioam/jvpp/io/fd/vpp/jvpp/ioamtrace/test/Readme.txt b/plugins/ioam-plugin/ioam/jvpp/io/fd/vpp/jvpp/ioamtrace/test/Readme.txt deleted file mode 100644 index 17e45a81..00000000 --- a/plugins/ioam-plugin/ioam/jvpp/io/fd/vpp/jvpp/ioamtrace/test/Readme.txt +++ /dev/null @@ -1 +0,0 @@ -sudo java -cp build-vpp-native/vpp-api/java/jvpp-registry-17.01.jar:build-vpp-native/plugins/ioam-plugin/jvpp-ioam-trace-1.0.jar io.fd.vpp.jvpp.ioamtrace.test.IoamTraceApiTest diff --git a/plugins/ioam-plugin/ioam/lib-pot/jvpp_ioam_pot.c b/plugins/ioam-plugin/ioam/lib-pot/jvpp_ioam_pot.c deleted file mode 100644 index a60ae60f..00000000 --- a/plugins/ioam-plugin/ioam/lib-pot/jvpp_ioam_pot.c +++ /dev/null @@ -1,124 +0,0 @@ -/* - * Copyright (c) 2016 Cisco and/or its affiliates. - * Licensed under the Apache License, Version 2.0 (the "License"); - * you may not use this file except in compliance with the License. - * You may obtain a copy of the License at: - * - * http://www.apache.org/licenses/LICENSE-2.0 - * - * Unless required by applicable law or agreed to in writing, software - * distributed under the License is distributed on an "AS IS" BASIS, - * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. - * See the License for the specific language governing permissions and - * limitations under the License. - */ - -#include - -#include -#define vl_typedefs /* define message structures */ -#include -#undef vl_typedefs - -#define vl_endianfun -#include -#undef vl_endianfun - -#define vl_print(handle, ...) -#define vl_printfun -#include -#undef vl_printfun - -/* Get the API version number */ -#define vl_api_version(n,v) static u32 api_version=(v); -#include -#undef vl_api_version - -#include -#include -#include - -#if VPPJNI_DEBUG == 1 - #define DEBUG_LOG(...) clib_warning(__VA_ARGS__) -#else - #define DEBUG_LOG(...) -#endif - -#include - -#include "ioam/jvpp/io_fd_vpp_jvpp_ioampot_JVppIoampotImpl.h" -#include "jvpp_ioam_pot.h" -#include "ioam/jvpp/jvpp_ioam_pot_gen.h" - -/* - * Class: io_fd_vpp_jvpp_ioampot_JVppIoampotImpl - * Method: init0 - * Signature: (JI)V - */ -JNIEXPORT void JNICALL Java_io_fd_vpp_jvpp_ioampot_JVppIoampotImpl_init0 - (JNIEnv *env, jclass clazz, jobject callback, jlong queue_address, jint my_client_index) { - ioampot_main_t * plugin_main = &ioampot_main; - u8 * name; - clib_warning ("Java_io_fd_vpp_jvpp_ioampot_JVppIoampotImpl_init0"); - - plugin_main->my_client_index = my_client_index; - plugin_main->vl_input_queue = (unix_shared_memory_queue_t *)queue_address; - - name = format (0, "ioam_pot_%08x%c", api_version, 0); - plugin_main->msg_id_base = vl_client_get_first_plugin_msg_id ((char *) name); - - if (plugin_main->msg_id_base == (u16) ~0) { - jclass exClass = (*env)->FindClass(env, "java/lang/IllegalStateException"); - (*env)->ThrowNew(env, exClass, "ioam_pot plugin is not loaded in VPP"); - } else { - plugin_main->callbackObject = (*env)->NewGlobalRef(env, callback); - plugin_main->callbackClass = (jclass)(*env)->NewGlobalRef(env, (*env)->GetObjectClass(env, callback)); - - #define _(N,n) \ - vl_msg_api_set_handlers(VL_API_##N + plugin_main->msg_id_base, #n, \ - vl_api_##n##_t_handler, \ - vl_noop_handler, \ - vl_api_##n##_t_endian, \ - vl_api_##n##_t_print, \ - sizeof(vl_api_##n##_t), 1); - foreach_api_reply_handler; - #undef _ - } -} - -JNIEXPORT void JNICALL Java_io_fd_vpp_jvpp_ioampot_JVppIoampotImpl_close0 -(JNIEnv *env, jclass clazz) { - ioampot_main_t * plugin_main = &ioampot_main; - - // cleanup: - (*env)->DeleteGlobalRef(env, plugin_main->callbackClass); - (*env)->DeleteGlobalRef(env, plugin_main->callbackObject); - - plugin_main->callbackClass = NULL; - plugin_main->callbackObject = NULL; -} - -/* Attach thread to JVM and cache class references when initiating JVPP iOAM POT */ -jint JNI_OnLoad(JavaVM *vm, void *reserved) { - JNIEnv* env; - - if ((*vm)->GetEnv(vm, (void**) &env, JNI_VERSION_1_8) != JNI_OK) { - return JNI_EVERSION; - } - - if (cache_class_references(env) != 0) { - clib_warning ("Failed to cache class references\n"); - return JNI_ERR; - } - - return JNI_VERSION_1_8; -} - -/* Clean up cached references when disposing JVPP iOAM POT */ -void JNI_OnUnload(JavaVM *vm, void *reserved) { - JNIEnv* env; - if ((*vm)->GetEnv(vm, (void**) &env, JNI_VERSION_1_8) != JNI_OK) { - return; - } - delete_class_references(env); -} diff --git a/plugins/ioam-plugin/ioam/lib-pot/jvpp_ioam_pot.h b/plugins/ioam-plugin/ioam/lib-pot/jvpp_ioam_pot.h deleted file mode 100644 index 00aa51db..00000000 --- a/plugins/ioam-plugin/ioam/lib-pot/jvpp_ioam_pot.h +++ /dev/null @@ -1,45 +0,0 @@ -/* - * Copyright (c) 2016 Cisco and/or its affiliates. - * Licensed under the Apache License, Version 2.0 (the "License"); - * you may not use this file except in compliance with the License. - * You may obtain a copy of the License at: - * - * http://www.apache.org/licenses/LICENSE-2.0 - * - * Unless required by applicable law or agreed to in writing, software - * distributed under the License is distributed on an "AS IS" BASIS, - * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. - * See the License for the specific language governing permissions and - * limitations under the License. - */ -#ifndef __included_jvpp_ioam_pot_h__ -#define __included_jvpp_ioam_pot_h__ - -#include -#include -#include -#include -#include -#include - -/* Global state for JVPP-IOAM-POT */ -typedef struct { - /* Base message index for the pot plugin */ - u16 msg_id_base; - - /* Pointer to shared memory queue */ - unix_shared_memory_queue_t * vl_input_queue; - - /* VPP api client index */ - u32 my_client_index; - - /* Callback object and class references enabling asynchronous Java calls */ - jobject callbackObject; - jclass callbackClass; - -} ioampot_main_t; - -ioampot_main_t ioampot_main __attribute__((aligned (64))); - - -#endif /* __included_jvpp_ioam_pot_h__ */ diff --git a/plugins/ioam-plugin/ioam/lib-pot/math64.h b/plugins/ioam-plugin/ioam/lib-pot/math64.h deleted file mode 100644 index 4c608a37..00000000 --- a/plugins/ioam-plugin/ioam/lib-pot/math64.h +++ /dev/null @@ -1,159 +0,0 @@ -/* - * math64.h provides the 64 bit unsigned integer add, multiply followed by modulo operation - * The linux/math64.h provides divide and multiply 64 bit integers but: - * 1. multiply: mul_u64_u64_shr - only returns 64 bits of the result and has to be called - * twice to get the complete 128 bits of the result. - * 2. Modulo operation of the result of addition and multiplication of u64 that may result - * in integers > 64 bits is not supported - * Hence this header to combine add/multiply followed by modulo of u64 integrers - * always resulting in u64. - * - * Copyright (c) 2016 Cisco and/or its affiliates. - * Licensed under the Apache License, Version 2.0 (the "License"); - * you may not use this file except in compliance with the License. - * You may obtain a copy of the License at: - * - * http://www.apache.org/licenses/LICENSE-2.0 - * - * Unless required by applicable law or agreed to in writing, software - * distributed under the License is distributed on an "AS IS" BASIS, - * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. - * See the License for the specific language governing permissions and - * limitations under the License. - */ -#ifndef include_vnet_math64_h -#define include_vnet_math64_h -#include - -/* - * multiplies and returns result in hi and lo - */ -static inline void mul64by64(u64 a, u64 b, u64 * hi, u64 * lo) -{ - u64 a_lo = (u64) (uint32_t) a; - u64 a_hi = a >> 32; - u64 b_lo = (u64) (u32) b; - u64 b_hi = b >> 32; - - u64 p0 = a_lo * b_lo; - u64 p1 = a_lo * b_hi; - u64 p2 = a_hi * b_lo; - u64 p3 = a_hi * b_hi; - - u32 cy = (u32) (((p0 >> 32) + (u32) p1 + (u32) p2) >> 32); - - *lo = p0 + (p1 << 32) + (p2 << 32); - *hi = p3 + (p1 >> 32) + (p2 >> 32) + cy; - return; -} - -#define TWO64 18446744073709551616.0 - -static inline u64 mod128by64(u64 x, u64 y, u64 m, double di) -{ - u64 q1, q2, q; - u64 p1, p0; - double dq; - - /* calculate quotient first pass 53 bits */ - dq = (TWO64 * (double)x + (double)y) * di; - - if (dq >= TWO64) - q1 = 0xfffffffffffff800L; - else - q1 = dq; - - /* q1 * m to compare the product to the dividend. */ - mul64by64(q1, m, &p1, &p0); - - /* Adjust quotient. is it > actual result: */ - if (x < p1 || (x == p1 && y < p0)) - { - /* q1 > quotient. calculate abs remainder */ - x = p1 - (x + (p0 < y)); - y = p0 - y; - - /* use the remainder as new dividend to adjust quotient */ - q2 = (u64) ((TWO64 * (double)x + (double)y) * di); - mul64by64(q2, m, &p1, &p0); - - q = q1 - q2; - if (x < p1 || (x == p1 && y <= p0)) - { - y = p0 - y; - } - else - { - y = p0 - y; - y += m; - q--; - } - } - else - { - x = x - (p1 + (y < p0)); - y = y - p0; - - q2 = (u64) ((TWO64 * (double)x + (double)y) * di); - mul64by64(q2, m, &p1, &p0); - - q = q1 + q2; - if (x < p1 || (x == p1 && y < p0)) - { - y = y - p0; - y += m; - q--; - } - else - { - y = y - p0; - if (y >= m) - { - y -= m; - q++; - } - } - } - - return y; -} - -/* - * returns a % p - */ -static inline u64 mod64by64(u64 a, u64 p, u64 primeinv) -{ - return (mod128by64(0, a, p, primeinv)); -} - -static inline void add64(u64 a, u64 b, u64 * whi, u64 * wlo) -{ - *wlo = a + b; - if (*wlo < a) - *whi = 1; - -} - -/* - * returns (a + b)%p - */ -static inline u64 add64_mod(u64 a, u64 b, u64 p, double pi) -{ - u64 shi = 0, slo = 0; - - add64(a, b, &shi, &slo); - return (mod128by64(shi, slo, p, pi)); -} - -/* - * returns (ab) % p - */ -static inline u64 mul64_mod(u64 a, u64 b, u64 p, double pi) -{ - u64 phi = 0, plo = 0; - - mul64by64(a, b, &phi, &plo); - return (mod128by64(phi, plo, p, pi)); -} - -#endif diff --git a/plugins/ioam-plugin/ioam/lib-pot/pot.api b/plugins/ioam-plugin/ioam/lib-pot/pot.api deleted file mode 100644 index fa2fc126..00000000 --- a/plugins/ioam-plugin/ioam/lib-pot/pot.api +++ /dev/null @@ -1,133 +0,0 @@ -/* Hey Emacs use -*- mode: C -*- */ -/* - * Copyright (c) 2016 Cisco and/or its affiliates. - * Licensed under the Apache License, Version 2.0 (the "License"); - * you may not use this file except in compliance with the License. - * You may obtain a copy of the License at: - * - * http://www.apache.org/licenses/LICENSE-2.0 - * - * Unless required by applicable law or agreed to in writing, software - * distributed under the License is distributed on an "AS IS" BASIS, - * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. - * See the License for the specific language governing permissions and - * limitations under the License. - */ - - -/** \brief Proof of Transit(POT): Set POT profile - @param id - id of the profile - @param validator - True/False to indicate if this is verifier - @param secret_key - Verification key - @param secret_share - Share of the 1st polynomial - @param prime - Prime number used for modulo operation - @param max_bits - Max bits to be used for Random number generation - @param lpc - Lagrange basis polynomial - @param polynomial_public - pre-evaluated public polynomial - @param list_name_len - length of the name of this profile list - @param list_name - name of this profile list -*/ -define pot_profile_add { - u32 client_index; - u32 context; - u8 id; - u8 validator; - u64 secret_key; - u64 secret_share; - u64 prime; - u8 max_bits; - u64 lpc; - u64 polynomial_public; - u8 list_name_len; - u8 list_name[0]; -}; - -/** \brief Proof of Transit profile add / del response - @param context - sender context, to match reply w/ request - @param retval - return value for request -*/ -define pot_profile_add_reply { - u32 context; - i32 retval; -}; - - -/** \brief Proof of Transit(POT): Activate POT profile in the list - @param id - id of the profile - @param list_name_len - length of the name of this profile list - @param list_name - name of this profile list -*/ -define pot_profile_activate { - u32 client_index; - u32 context; - u8 id; - u8 list_name_len; - u8 list_name[0]; -}; - -/** \brief Proof of Transit profile activate response - @param context - sender context, to match reply w/ request - @param retval - return value for request -*/ -define pot_profile_activate_reply { - u32 context; - i32 retval; -}; - -/** \brief Delete POT Profile - @param client_index - opaque cookie to identify the sender - @param context - sender context, to match reply w/ request - @param list_name_len - length of the name of the profile list - @param list_name - name of profile list to delete -*/ -define pot_profile_del { - u32 client_index; - u32 context; - u8 list_name_len; - u8 list_name[0]; -}; - -/** \brief Proof of Transit profile add / del response - @param context - sender context, to match reply w/ request - @param retval - return value for request -*/ -define pot_profile_del_reply { - u32 context; - i32 retval; -}; - -/** \brief Show POT Profiles - @param client_index - opaque cookie to identify the sender - @param context - sender context, to match reply w/ request - @param id - id of the profile -*/ -define pot_profile_show_config_dump { - u32 client_index; - u32 context; - u8 id; -}; - -/** \brief Show POT profile reply - @param id - id of the profile - @param validator - True/False to indicate if this is verifier - @param secret_key - Verification key - @param secret_share - Share of the 1st polynomial - @param prime - Prime number used for modulo operation - @param max_bits - Max bits to be used for Random number generation - @param lpc - Lagrange basis polynomial - @param polynomial_public - pre-evaluated public polynomial - @param list_name_len - length of the name of this profile list - @param list_name - name of this profile list -*/ -define pot_profile_show_config_details { - u32 context; - i32 retval; - u8 id; - u8 validator; - u64 secret_key; - u64 secret_share; - u64 prime; - u64 bit_mask; - u64 lpc; - u64 polynomial_public; -}; diff --git a/plugins/ioam-plugin/ioam/lib-pot/pot_all_api_h.h b/plugins/ioam-plugin/ioam/lib-pot/pot_all_api_h.h deleted file mode 100644 index 63967c45..00000000 --- a/plugins/ioam-plugin/ioam/lib-pot/pot_all_api_h.h +++ /dev/null @@ -1,16 +0,0 @@ -/* - * Copyright (c) 2016 Cisco and/or its affiliates. - * Licensed under the Apache License, Version 2.0 (the "License"); - * you may not use this file except in compliance with the License. - * You may obtain a copy of the License at: - * - * http://www.apache.org/licenses/LICENSE-2.0 - * - * Unless required by applicable law or agreed to in writing, software - * distributed under the License is distributed on an "AS IS" BASIS, - * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. - * See the License for the specific language governing permissions and - * limitations under the License. - */ -/* Include the generated file, see BUILT_SOURCES in Makefile.am */ -#include diff --git a/plugins/ioam-plugin/ioam/lib-pot/pot_api.c b/plugins/ioam-plugin/ioam/lib-pot/pot_api.c deleted file mode 100644 index d3af7b40..00000000 --- a/plugins/ioam-plugin/ioam/lib-pot/pot_api.c +++ /dev/null @@ -1,292 +0,0 @@ -/* - * Copyright (c) 2016 Cisco and/or its affiliates. - * Licensed under the Apache License, Version 2.0 (the "License"); - * you may not use this file except in compliance with the License. - * You may obtain a copy of the License at: - * - * http://www.apache.org/licenses/LICENSE-2.0 - * - * Unless required by applicable law or agreed to in writing, software - * distributed under the License is distributed on an "AS IS" BASIS, - * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. - * See the License for the specific language governing permissions and - * limitations under the License. - */ -/* - *------------------------------------------------------------------ - * pot_api.c - Proof of Transit related APIs to create - * and maintain profiles - *------------------------------------------------------------------ - */ - -#include -#include -#include - -#include -#include -#include - -/* define message IDs */ -#include - -/* define message structures */ -#define vl_typedefs -#include -#undef vl_typedefs - -/* define generated endian-swappers */ -#define vl_endianfun -#include -#undef vl_endianfun - -/* instantiate all the print functions we know about */ -#define vl_print(handle, ...) vlib_cli_output (handle, __VA_ARGS__) -#define vl_printfun -#include -#undef vl_printfun - -/* Get the API version number */ -#define vl_api_version(n,v) static u32 api_version=(v); -#include -#undef vl_api_version - -/* - * A handy macro to set up a message reply. - * Assumes that the following variables are available: - * mp - pointer to request message - * rmp - pointer to reply message type - * rv - return value - */ - -#define REPLY_MACRO(t) \ -do { \ - unix_shared_memory_queue_t * q = \ - vl_api_client_index_to_input_queue (mp->client_index); \ - if (!q) \ - return; \ - \ - rmp = vl_msg_api_alloc (sizeof (*rmp)); \ - rmp->_vl_msg_id = ntohs((t)+sm->msg_id_base); \ - rmp->context = mp->context; \ - rmp->retval = ntohl(rv); \ - \ - vl_msg_api_send_shmem (q, (u8 *)&rmp); \ -} while(0); - -#define REPLY_MACRO2(t, body) \ -do { \ - unix_shared_memory_queue_t * q; \ - rv = vl_msg_api_pd_handler (mp, rv); \ - q = vl_api_client_index_to_input_queue (mp->client_index); \ - if (!q) \ - return; \ - \ - rmp = vl_msg_api_alloc (sizeof (*rmp)); \ - rmp->_vl_msg_id = ntohs((t)+sm->msg_id_base); \ - rmp->context = mp->context; \ - rmp->retval = ntohl(rv); \ - do {body;} while (0); \ - vl_msg_api_send_shmem (q, (u8 *)&rmp); \ -} while(0); - -/* List of message types that this plugin understands */ - -#define foreach_pot_plugin_api_msg \ -_(POT_PROFILE_ADD, pot_profile_add) \ -_(POT_PROFILE_ACTIVATE, pot_profile_activate) \ -_(POT_PROFILE_DEL, pot_profile_del) \ -_(POT_PROFILE_SHOW_CONFIG_DUMP, pot_profile_show_config_dump) \ - -static void vl_api_pot_profile_add_t_handler -(vl_api_pot_profile_add_t *mp) -{ - pot_main_t * sm = &pot_main; - int rv = 0; - vl_api_pot_profile_add_reply_t * rmp; - u8 id; - pot_profile *profile = NULL; - u8 *name = 0; - - if (mp->list_name_len) - name = format(0, "%s", mp->list_name); - - pot_profile_list_init(name); - id = mp->id; - profile = pot_profile_find(id); - if (profile) { - rv = pot_profile_create(profile, - clib_net_to_host_u64(mp->prime), - clib_net_to_host_u64(mp->polynomial_public), - clib_net_to_host_u64(mp->lpc), - clib_net_to_host_u64(mp->secret_share)); - if (rv != 0) - goto ERROROUT; - if (1 == mp->validator) - (void)pot_set_validator(profile, clib_net_to_host_u64(mp->secret_key)); - (void)pot_profile_set_bit_mask(profile, mp->max_bits); - } else { - rv = -3; - } - ERROROUT: - vec_free(name); - REPLY_MACRO(VL_API_POT_PROFILE_ADD_REPLY); -} - -static void send_pot_profile_details(vl_api_pot_profile_show_config_dump_t *mp, u8 id) -{ - vl_api_pot_profile_show_config_details_t * rmp; - pot_main_t * sm = &pot_main; - pot_profile *profile = pot_profile_find(id); - int rv = 0; - if(profile){ - REPLY_MACRO2(VL_API_POT_PROFILE_SHOW_CONFIG_DETAILS, - rmp->id=id; - rmp->validator=profile->validator; - rmp->secret_key=clib_host_to_net_u64(profile->secret_key); - rmp->secret_share=clib_host_to_net_u64(profile->secret_share); - rmp->prime=clib_host_to_net_u64(profile->prime); - rmp->bit_mask=clib_host_to_net_u64(profile->bit_mask); - rmp->lpc=clib_host_to_net_u64(profile->lpc); - rmp->polynomial_public=clib_host_to_net_u64(profile->poly_pre_eval); - ); - } - else{ - REPLY_MACRO2(VL_API_POT_PROFILE_SHOW_CONFIG_DETAILS, - rmp->id=id; - rmp->validator=0; - rmp->secret_key=0; - rmp->secret_share=0; - rmp->prime=0; - rmp->bit_mask=0; - rmp->lpc=0; - rmp->polynomial_public=0; - ); - } -} - -static void vl_api_pot_profile_show_config_dump_t_handler -(vl_api_pot_profile_show_config_dump_t *mp) -{ - u8 id = mp->id; - u8 dump_call_id = ~0; - if(dump_call_id==id){ - for(id=0;idlist_name_len) - name = format(0, "%s", mp->list_name); - if (!pot_profile_list_is_enabled(name)) { - rv = -1; - } else { - id = mp->id; - rv = pot_profile_set_active(id); - } - - vec_free(name); - REPLY_MACRO(VL_API_POT_PROFILE_ACTIVATE_REPLY); -} - - -static void vl_api_pot_profile_del_t_handler -(vl_api_pot_profile_del_t *mp) -{ - pot_main_t * sm = &pot_main; - int rv = 0; - vl_api_pot_profile_del_reply_t * rmp; - - clear_pot_profiles(); - - REPLY_MACRO(VL_API_POT_PROFILE_DEL_REPLY); -} - - -/* - * This routine exists to convince the vlib plugin framework that - * we haven't accidentally copied a random .dll into the plugin directory. - * - * Also collects global variable pointers passed from the vpp engine - */ - -clib_error_t * -vlib_plugin_register (vlib_main_t * vm, vnet_plugin_handoff_t * h, - int from_early_init) -{ - pot_main_t * sm = &pot_main; - clib_error_t * error = 0; - - sm->vlib_main = vm; - sm->vnet_main = h->vnet_main; - return error; -} - -/* Set up the API message handling tables */ -static clib_error_t * -pot_plugin_api_hookup (vlib_main_t *vm) -{ - pot_main_t * sm = &pot_main; -#define _(N,n) \ - vl_msg_api_set_handlers((VL_API_##N + sm->msg_id_base), \ - #n, \ - vl_api_##n##_t_handler, \ - vl_noop_handler, \ - vl_api_##n##_t_endian, \ - vl_api_##n##_t_print, \ - sizeof(vl_api_##n##_t), 1); - foreach_pot_plugin_api_msg; -#undef _ - - return 0; -} - -#define vl_msg_name_crc_list -#include -#undef vl_msg_name_crc_list - -static void -setup_message_id_table (pot_main_t * sm, api_main_t * am) -{ -#define _(id,n,crc) \ - vl_msg_api_add_msg_name_crc (am, #n "_" #crc, id + sm->msg_id_base); - foreach_vl_msg_name_crc_pot; -#undef _ -} - -static clib_error_t * pot_init (vlib_main_t * vm) -{ - pot_main_t * sm = &pot_main; - clib_error_t * error = 0; - u8 * name; - - bzero(sm, sizeof(pot_main)); - (void)pot_util_init(); - name = format (0, "ioam_pot_%08x%c", api_version, 0); - - /* Ask for a correctly-sized block of API message decode slots */ - sm->msg_id_base = vl_msg_api_get_msg_ids - ((char *) name, VL_MSG_FIRST_AVAILABLE); - - error = pot_plugin_api_hookup (vm); - - /* Add our API messages to the global name_crc hash table */ - setup_message_id_table (sm, &api_main); - - vec_free(name); - - return error; -} - -VLIB_INIT_FUNCTION (pot_init); diff --git a/plugins/ioam-plugin/ioam/lib-pot/pot_msg_enum.h b/plugins/ioam-plugin/ioam/lib-pot/pot_msg_enum.h deleted file mode 100644 index a4a88bed..00000000 --- a/plugins/ioam-plugin/ioam/lib-pot/pot_msg_enum.h +++ /dev/null @@ -1,28 +0,0 @@ -/* - * Copyright (c) 2016 Cisco and/or its affiliates. - * Licensed under the Apache License, Version 2.0 (the "License"); - * you may not use this file except in compliance with the License. - * You may obtain a copy of the License at: - * - * http://www.apache.org/licenses/LICENSE-2.0 - * - * Unless required by applicable law or agreed to in writing, software - * distributed under the License is distributed on an "AS IS" BASIS, - * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. - * See the License for the specific language governing permissions and - * limitations under the License. - */ -#ifndef included_pot_msg_enum_h -#define included_pot_msg_enum_h - -#include - -#define vl_msg_id(n,h) n, -typedef enum { -#include - /* We'll want to know how many messages IDs we need... */ - VL_MSG_FIRST_AVAILABLE, -} vl_msg_id_t; -#undef vl_msg_id - -#endif /* included_pot_msg_enum_h */ diff --git a/plugins/ioam-plugin/ioam/lib-pot/pot_test.c b/plugins/ioam-plugin/ioam/lib-pot/pot_test.c deleted file mode 100644 index 2e870238..00000000 --- a/plugins/ioam-plugin/ioam/lib-pot/pot_test.c +++ /dev/null @@ -1,365 +0,0 @@ -/* - * Copyright (c) 2016 Cisco and/or its affiliates. - * Licensed under the Apache License, Version 2.0 (the "License"); - * you may not use this file except in compliance with the License. - * You may obtain a copy of the License at: - * - * http://www.apache.org/licenses/LICENSE-2.0 - * - * Unless required by applicable law or agreed to in writing, software - * distributed under the License is distributed on an "AS IS" BASIS, - * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. - * See the License for the specific language governing permissions and - * limitations under the License. - */ -/* - *------------------------------------------------------------------ - * pot_test.c - test harness for pot plugin - *------------------------------------------------------------------ - */ - -#include -#include -#include -#include -#include - -/* Declare message IDs */ -#include - -/* define message structures */ -#define vl_typedefs -#include -#undef vl_typedefs - -/* declare message handlers for each api */ - -#define vl_endianfun /* define message structures */ -#include -#undef vl_endianfun - -/* instantiate all the print functions we know about */ -#define vl_print(handle, ...) -#define vl_printfun -#include -#undef vl_printfun - -/* Get the API version number. */ -#define vl_api_version(n,v) static u32 api_version=(v); -#include -#undef vl_api_version - - -typedef struct { - /* API message ID base */ - u16 msg_id_base; - vat_main_t *vat_main; -} pot_test_main_t; - -pot_test_main_t pot_test_main; - -#define foreach_standard_reply_retval_handler \ -_(pot_profile_add_reply) \ -_(pot_profile_activate_reply) \ -_(pot_profile_del_reply) - -#define foreach_custom_reply_retval_handler \ -_(pot_profile_show_config_details, \ - errmsg(" ID:%d\n",mp->id); \ - errmsg(" Validator:%d\n",mp->validator); \ - errmsg(" secret_key:%Lx\n",clib_net_to_host_u64(mp->secret_key)); \ - errmsg(" secret_share:%Lx\n",clib_net_to_host_u64(mp->secret_share)); \ - errmsg(" prime:%Lx\n",clib_net_to_host_u64(mp->prime)); \ - errmsg(" bitmask:%Lx\n",clib_net_to_host_u64(mp->bit_mask)); \ - errmsg(" lpc:%Lx\n",clib_net_to_host_u64(mp->lpc)); \ - errmsg(" public poly:%Lx\n",clib_net_to_host_u64(mp->polynomial_public)); \ - ) - -#define _(n) \ - static void vl_api_##n##_t_handler \ - (vl_api_##n##_t * mp) \ - { \ - vat_main_t * vam = pot_test_main.vat_main; \ - i32 retval = ntohl(mp->retval); \ - if (vam->async_mode) { \ - vam->async_errors += (retval < 0); \ - } else { \ - vam->retval = retval; \ - vam->result_ready = 1; \ - } \ - } -foreach_standard_reply_retval_handler; -#undef _ - -#define _(n,body) \ - static void vl_api_##n##_t_handler \ - (vl_api_##n##_t * mp) \ - { \ - vat_main_t * vam = pot_test_main.vat_main; \ - i32 retval = ntohl(mp->retval); \ - if (vam->async_mode) { \ - vam->async_errors += (retval < 0); \ - } else { \ - vam->retval = retval; \ - vam->result_ready = 1; \ - } \ - do{body;}while(0); \ - } -foreach_custom_reply_retval_handler; -#undef _ - -/* - * Table of message reply handlers, must include boilerplate handlers - * we just generated - */ -#define foreach_vpe_api_reply_msg \ -_(POT_PROFILE_ADD_REPLY, pot_profile_add_reply) \ -_(POT_PROFILE_ACTIVATE_REPLY, pot_profile_activate_reply) \ -_(POT_PROFILE_DEL_REPLY, pot_profile_del_reply) \ -_(POT_PROFILE_SHOW_CONFIG_DETAILS, pot_profile_show_config_details) - - -/* M: construct, but don't yet send a message */ - -#define M(T,t) \ -do { \ - vam->result_ready = 0; \ - mp = vl_msg_api_alloc(sizeof(*mp)); \ - memset (mp, 0, sizeof (*mp)); \ - mp->_vl_msg_id = ntohs (VL_API_##T + sm->msg_id_base); \ - mp->client_index = vam->my_client_index; \ -} while(0); - -#define M2(T,t,n) \ -do { \ - vam->result_ready = 0; \ - mp = vl_msg_api_alloc(sizeof(*mp)+(n)); \ - memset (mp, 0, sizeof (*mp)); \ - mp->_vl_msg_id = ntohs (VL_API_##T + sm->msg_id_base); \ - mp->client_index = vam->my_client_index; \ -} while(0); - -/* S: send a message */ -#define S (vl_msg_api_send_shmem (vam->vl_input_queue, (u8 *)&mp)) - -/* W: wait for results, with timeout */ -#define W \ -do { \ - timeout = vat_time_now (vam) + 1.0; \ - \ - while (vat_time_now (vam) < timeout) { \ - if (vam->result_ready == 1) { \ - return (vam->retval); \ - } \ - } \ - return -99; \ -} while(0); - - -static int api_pot_profile_add (vat_main_t *vam) -{ -#define MAX_BITS 64 - pot_test_main_t * sm = &pot_test_main; - unformat_input_t *input = vam->input; - vl_api_pot_profile_add_t *mp; - u8 *name = NULL; - u64 prime = 0; - u64 secret_share = 0; - u64 secret_key = 0; - u32 bits = MAX_BITS; - u64 lpc = 0, poly2 = 0; - f64 timeout; - u8 id = 0; - int rv = 0; - - while (unformat_check_input(input) != UNFORMAT_END_OF_INPUT) - { - if (unformat(input, "name %s", &name)) - ; - else if(unformat(input, "id %d", &id)) - ; - else if (unformat(input, "validator-key 0x%Lx", &secret_key)) - ; - else if (unformat(input, "prime-number 0x%Lx", &prime)) - ; - else if (unformat(input, "secret-share 0x%Lx", &secret_share)) - ; - else if (unformat(input, "polynomial-public 0x%Lx", &poly2)) - ; - else if (unformat(input, "lpc 0x%Lx", &lpc)) - ; - else if (unformat(input, "bits-in-random %u", &bits)) - { - if (bits > MAX_BITS) - bits = MAX_BITS; - } - else - break; - } - - if (!name) - { - errmsg ("name required\n"); - rv = -99; - goto OUT; - } - - M2(POT_PROFILE_ADD, pot_profile_add, vec_len(name)); - - mp->list_name_len = vec_len(name); - clib_memcpy(mp->list_name, name, mp->list_name_len); - mp->secret_share = clib_host_to_net_u64(secret_share); - mp->polynomial_public = clib_host_to_net_u64(poly2); - mp->lpc = clib_host_to_net_u64(lpc); - mp->prime = clib_host_to_net_u64(prime); - if (secret_key != 0) - { - mp->secret_key = clib_host_to_net_u64(secret_key); - mp->validator = 1; - } - else - { - mp->validator = 0; - } - mp->id = id; - mp->max_bits = bits; - - S; W; - -OUT: - vec_free(name); - return(rv); -} - -static int api_pot_profile_activate (vat_main_t *vam) -{ -#define MAX_BITS 64 - pot_test_main_t * sm = &pot_test_main; - unformat_input_t *input = vam->input; - vl_api_pot_profile_activate_t *mp; - u8 *name = NULL; - u8 id = 0; - int rv = 0; - f64 timeout; - - while (unformat_check_input(input) != UNFORMAT_END_OF_INPUT) - { - if (unformat(input, "name %s", &name)) - ; - else if(unformat(input, "id %d", &id)) - ; - else - break; - } - - if (!name) - { - errmsg ("name required\n"); - rv = -99; - goto OUT; - } - - M2(POT_PROFILE_ACTIVATE, pot_profile_activate, vec_len(name)); - - mp->list_name_len = vec_len(name); - clib_memcpy(mp->list_name, name, mp->list_name_len); - mp->id = id; - - S; W; - -OUT: - vec_free(name); - return(rv); -} - - -static int api_pot_profile_del (vat_main_t *vam) -{ - pot_test_main_t * sm = &pot_test_main; - vl_api_pot_profile_del_t *mp; - f64 timeout; - - M(POT_PROFILE_DEL, pot_profile_del); - mp->list_name_len = 0; - S; W; - return 0; -} - -static int api_pot_profile_show_config_dump (vat_main_t *vam) -{ - pot_test_main_t * sm = &pot_test_main; - unformat_input_t *input = vam->input; - vl_api_pot_profile_show_config_dump_t *mp; - f64 timeout; - u8 id = 0; - - while(unformat_check_input(input) != UNFORMAT_END_OF_INPUT) - { - if(unformat(input,"id %d",&id)); - else - break; - } - M(POT_PROFILE_SHOW_CONFIG_DUMP, pot_profile_show_config_dump); - - mp->id = id; - - S; W; - return 0; -} - -/* - * List of messages that the api test plugin sends, - * and that the data plane plugin processes - */ -#define foreach_vpe_api_msg \ -_(pot_profile_add, "name id [0-1] " \ - "prime-number <0xu64> bits-in-random [0-64] " \ - "secret-share <0xu64> lpc <0xu64> polynomial-public <0xu64> " \ - "[validator-key <0xu64>] [validity <0xu64>]") \ -_(pot_profile_activate, "name id [0-1] ") \ -_(pot_profile_del, "[id ]") \ -_(pot_profile_show_config_dump, "id [0-1]") - -void vat_api_hookup (vat_main_t *vam) -{ - pot_test_main_t * sm = &pot_test_main; - /* Hook up handlers for replies from the data plane plug-in */ -#define _(N,n) \ - vl_msg_api_set_handlers((VL_API_##N + sm->msg_id_base), \ - #n, \ - vl_api_##n##_t_handler, \ - vl_noop_handler, \ - vl_api_##n##_t_endian, \ - vl_api_##n##_t_print, \ - sizeof(vl_api_##n##_t), 1); - foreach_vpe_api_reply_msg; -#undef _ - - /* API messages we can send */ -#define _(n,h) hash_set_mem (vam->function_by_name, #n, api_##n); - foreach_vpe_api_msg; -#undef _ - - /* Help strings */ -#define _(n,h) hash_set_mem (vam->help_by_name, #n, h); - foreach_vpe_api_msg; -#undef _ -} - -clib_error_t * vat_plugin_register (vat_main_t *vam) -{ - pot_test_main_t * sm = &pot_test_main; - u8 * name; - - sm->vat_main = vam; - - name = format (0, "ioam_pot_%08x%c", api_version, 0); - sm->msg_id_base = vl_client_get_first_plugin_msg_id ((char *) name); - - if (sm->msg_id_base != (u16) ~0) - vat_api_hookup (vam); - - vec_free(name); - - return 0; -} diff --git a/plugins/ioam-plugin/ioam/lib-pot/pot_util.c b/plugins/ioam-plugin/ioam/lib-pot/pot_util.c deleted file mode 100644 index a253ad41..00000000 --- a/plugins/ioam-plugin/ioam/lib-pot/pot_util.c +++ /dev/null @@ -1,445 +0,0 @@ -/* - * Copyright (c) 2016 Cisco and/or its affiliates. - * Licensed under the Apache License, Version 2.0 (the "License"); - * you may not use this file except in compliance with the License. - * You may obtain a copy of the License at: - * - * http://www.apache.org/licenses/LICENSE-2.0 - * - * Unless required by applicable law or agreed to in writing, software - * distributed under the License is distributed on an "AS IS" BASIS, - * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. - * See the License for the specific language governing permissions and - * limitations under the License. - */ -#include -#include -#include -#include -#include -#include "math64.h" -#include "pot_util.h" - -pot_main_t pot_main; - -static void pot_profile_cleanup(pot_profile *profile); - -static void pot_main_profiles_reset (void) -{ - pot_main_t *sm = &pot_main; - int i = 0; - - for (i = 0; i < MAX_POT_PROFILES; i++) - { - pot_profile_cleanup(&(sm->profile_list[i])); - } - sm->active_profile_id = 0; - if (sm->profile_list_name) - vec_free(sm->profile_list_name); - sm->profile_list_name = NULL; -} - -int pot_util_init (void) -{ - pot_main_profiles_reset(); - - return(0); -} - -static void pot_profile_init(pot_profile * new, u8 id) -{ - if (new) - { - memset(new, 0, sizeof(pot_profile)); - new->id = id; - } -} - -pot_profile *pot_profile_find(u8 id) -{ - pot_main_t *sm = &pot_main; - - if (id < MAX_POT_PROFILES) - { - return (&(sm->profile_list[id])); - } - return (NULL); -} -static int pot_profile_name_equal (u8 *name0, u8 *name1) -{ - int len0, len1; - - len0 = vec_len (name0); - len1 = vec_len (name1); - if (len0 != len1) - return(0); - return (0==strncmp ((char *) name0, (char *)name1, len0)); -} - -int pot_profile_list_is_enabled (u8 *name) -{ - pot_main_t *sm = &pot_main; - return (pot_profile_name_equal(sm->profile_list_name, name)); -} - -void pot_profile_list_init(u8 * profile_list_name) -{ - pot_main_t *sm = &pot_main; - int i = 0; - - /* If it is the same profile list skip reset */ - if (pot_profile_name_equal(sm->profile_list_name, profile_list_name)) - { - return; - } - - pot_main_profiles_reset(); - if (vec_len(profile_list_name)) - sm->profile_list_name = (u8 *)vec_dup(profile_list_name); - else - sm->profile_list_name = 0; - sm->active_profile_id = 0; - - for (i = 0; i < MAX_POT_PROFILES; i++) - { - pot_profile_init(&(sm->profile_list[i]), i); - } -} - -static void pot_profile_cleanup(pot_profile * profile) -{ - u16 id = profile->id; - - memset(profile, 0, sizeof(pot_profile)); - profile->id = id; /* Restore id alone */ -} - -int pot_profile_create(pot_profile * profile, u64 prime, - u64 poly2, u64 lpc, u64 secret_share) -{ - if (profile && !profile->in_use) - { - pot_profile_cleanup(profile); - profile->prime = prime; - profile->primeinv = 1.0 / prime; - profile->lpc = lpc; - profile->poly_pre_eval = poly2; - profile->secret_share = secret_share; - profile->total_pkts_using_this_profile = 0; - profile->valid = 1; - return(0); - } - - return(-1); -} - -int pot_set_validator(pot_profile * profile, u64 key) -{ - if (profile && !profile->in_use) - { - profile->validator = 1; - profile->secret_key = key; - return(0); - } - return(-1); -} - -always_inline u64 pot_update_cumulative_inline(u64 cumulative, u64 random, - u64 secret_share, u64 prime, u64 lpc, u64 pre_split, double prime_inv) -{ - u64 share_random = 0; - u64 cumulative_new = 0; - - /* - * calculate split share for random - */ - share_random = add64_mod(pre_split, random, prime, prime_inv); - - /* - * lpc * (share_secret + share_random) - */ - share_random = add64_mod(share_random, secret_share, prime, prime_inv); - share_random = mul64_mod(share_random, lpc, prime, prime_inv); - - cumulative_new = add64_mod(cumulative, share_random, prime, prime_inv); - - return (cumulative_new); -} - -u64 pot_update_cumulative(pot_profile * profile, u64 cumulative, u64 random) -{ - if (profile && profile->valid != 0) - { - return (pot_update_cumulative_inline(cumulative, random, profile->secret_share, - profile->prime, profile->lpc, profile->poly_pre_eval, - profile->primeinv)); - } - return (0); -} - -always_inline u8 pot_validate_inline(u64 secret, u64 prime, double prime_inv, - u64 cumulative, u64 random) -{ - if (cumulative == (random + secret)) - { - return (1); - } - else if (cumulative == add64_mod(random, secret, prime, prime_inv)) - { - return (1); - } - return (0); -} - -/* - * return True if the cumulative matches secret from a profile - */ -u8 pot_validate(pot_profile * profile, u64 cumulative, u64 random) -{ - if (profile && profile->validator) - { - return (pot_validate_inline(profile->secret_key, profile->prime, - profile->primeinv, cumulative, random)); - } - return (0); -} - -/* - * Utility function to get random number per pack - */ -u64 pot_generate_random(pot_profile * profile) -{ - u64 random = 0; - int32_t second_half; - static u32 seed = 0; - - if (PREDICT_FALSE(!seed)) - seed = random_default_seed(); - - /* - * Upper 4 bytes seconds - */ - random = (u64) time(NULL); - - random &= 0xffffffff; - random = random << 32; - /* - * Lower 4 bytes random number - */ - second_half = random_u32(&seed); - - random |= second_half; - - if (PREDICT_TRUE(profile != NULL)) - { - random &= profile->bit_mask; - } - return (random); -} - -int pot_profile_set_bit_mask(pot_profile * profile, u16 bits) -{ - int sizeInBits; - - if (profile && !profile->in_use) - { - sizeInBits = sizeof(profile->bit_mask) * 8; - profile->bit_mask = - (bits >= - sizeInBits ? (u64) - 1 : (u64) ((u64) 1 << (u64) bits) - 1); - return(0); - } - return(-1); -} - -clib_error_t *clear_pot_profile_command_fn(vlib_main_t * vm, - unformat_input_t * input, vlib_cli_command_t * cmd) -{ - - pot_main_profiles_reset(); - - return 0; -} - -void clear_pot_profiles() -{ - clear_pot_profile_command_fn(0, 0, 0); -} - -VLIB_CLI_COMMAND(clear_pot_profile_command) = -{ -.path = "clear pot profile", -.short_help = "clear pot profile [|all]", -.function = clear_pot_profile_command_fn, -}; - -static clib_error_t *set_pot_profile_command_fn(vlib_main_t * vm, - unformat_input_t * input, vlib_cli_command_t * cmd) -{ - u64 prime; - u64 secret_share; - u64 secret_key; - u8 validator = 0; - u32 profile_id = ~0; - u32 bits; - u64 lpc = 0, poly2 = 0; - pot_profile *profile = NULL; - u8 *profile_list_name = NULL; - - bits = MAX_BITS; - - while (unformat_check_input(input) != UNFORMAT_END_OF_INPUT) - { - if (unformat(input, "name %s", - &profile_list_name)); - else if (unformat(input, "id %d", &profile_id)) - ; - else if (unformat(input, "validate-key 0x%Lx", &secret_key)) - validator = 1; - else if (unformat(input, "prime-number 0x%Lx", &prime)) - ; - else if (unformat(input, "secret_share 0x%Lx", &secret_share)) - ; - else if (unformat(input, "polynomial2 0x%Lx", &poly2)) - ; - else if (unformat(input, "lpc 0x%Lx", &lpc)) - ; - else if (unformat(input, "bits-in-random %d", &bits)) - { - if (bits > MAX_BITS) - bits = MAX_BITS; - } - else - break; - } - if (profile_list_name == 0) - { - return clib_error_return(0, "Name cannot be null"); - } - pot_profile_list_init(profile_list_name); - profile = pot_profile_find(profile_id); - - if (profile) - { - pot_profile_create(profile, prime, poly2, lpc, secret_share); - if (validator) - pot_set_validator(profile, secret_key); - pot_profile_set_bit_mask(profile, bits); - } - vec_free(profile_list_name); - return 0; -} - -VLIB_CLI_COMMAND(set_pot_profile_command) = -{ -.path = "set pot profile", -.short_help = "set pot profile name id [0-1] [validator-key 0xu64] \ - prime-number 0xu64 secret_share 0xu64 lpc 0xu64 \ - polynomial2 0xu64 bits-in-random [0-64] ", -.function = set_pot_profile_command_fn, -}; - -static clib_error_t *set_pot_profile_activate_command_fn(vlib_main_t * vm, - unformat_input_t * input, vlib_cli_command_t * cmd) -{ - pot_main_t *sm = &pot_main; - u8 *profile_list_name = NULL; - u32 id = 0; - clib_error_t *result = NULL; - - while (unformat_check_input(input) != UNFORMAT_END_OF_INPUT) - { - if (unformat(input, "name %s", - &profile_list_name)); - else if (unformat(input, "id %d", &id)) - ; - else - return clib_error_return(0, "unknown input `%U'", - format_unformat_error, input); - } - if (profile_list_name == 0) - { - return clib_error_return(0, "Name cannot be null"); - } - - if (!pot_profile_list_is_enabled(profile_list_name)) { - result = clib_error_return(0, "%s list is not enabled, profile in use %s", - profile_list_name, sm->profile_list_name); - } else if (0 != pot_profile_set_active((u8)id)) { - result = clib_error_return(0, "Profile %d not defined in %s", - id, sm->profile_list_name); - } - vec_free(profile_list_name); - return result; -} - -VLIB_CLI_COMMAND(set_pot_profile_activate_command) = -{ -.path = "set pot profile-active", -.short_help = "set pot profile-active name id [0-1]", -.function = set_pot_profile_activate_command_fn, -}; - -static clib_error_t *show_pot_profile_command_fn(vlib_main_t * vm, - unformat_input_t * input, vlib_cli_command_t * cmd) -{ - pot_main_t *sm = &pot_main; - pot_profile *p = NULL; - u16 i; - u8 *s = 0; - - if (vec_len(sm->profile_list_name) == 0) - { - s = format(s, "POT Profiles not configured\n"); - vlib_cli_output(vm, "%v", s); - return 0; - } - s = format(s, "Profile list in use : %s\n",sm->profile_list_name); - for (i = 0; i < MAX_POT_PROFILES; i++) - { - p = pot_profile_find(i); - if (p->valid == 0) - continue; - s = format(s, "POT Profile at index: %d\n", i); - s = format(s, " Id : %d\n", p->id); - s = format(s, " Validator : %s (%d)\n", - (p->validator) ? "True" : "False", p->validator); - if (p->validator == 1) - s = format(s, " Secret key : 0x%Lx (%Ld)\n", - p->secret_key, p->secret_key); - s = format(s, " Secret share : 0x%Lx (%Ld)\n", - p->secret_share, p->secret_share); - s = format(s, " Prime number : 0x%Lx (%Ld)\n", - p->prime, p->prime); - s = format(s, "2nd polynomial(eval) : 0x%Lx (%Ld)\n", - p->poly_pre_eval, p->poly_pre_eval); - s = format(s, " LPC : 0x%Lx (%Ld)\n", p->lpc, p->lpc); - - s = format(s, " Bit mask : 0x%Lx (%Ld)\n", - p->bit_mask, p->bit_mask); - } - - p = pot_profile_find(sm->active_profile_id); - - if (p && p->valid && p->in_use) { - s = format(s, "\nProfile index in use: %d\n", sm->active_profile_id); - s = format(s, "Pkts passed : 0x%Lx (%Ld)\n", - p->total_pkts_using_this_profile, - p->total_pkts_using_this_profile); - if (pot_is_decap(p)) - s = format(s, " This is Decap node. \n"); - } else { - s = format(s, "\nProfile index in use: None\n"); - } - vlib_cli_output(vm, "%v", s); - vec_free(s); - - return 0; -} - -VLIB_CLI_COMMAND(show_pot_profile_command) = -{ -.path = "show pot profile", -.short_help = "show pot profile", -.function = show_pot_profile_command_fn, -}; diff --git a/plugins/ioam-plugin/ioam/lib-pot/pot_util.h b/plugins/ioam-plugin/ioam/lib-pot/pot_util.h deleted file mode 100644 index 9df31fae..00000000 --- a/plugins/ioam-plugin/ioam/lib-pot/pot_util.h +++ /dev/null @@ -1,195 +0,0 @@ -/* - * pot_util.h -- Proof Of Transit Utility Header - * - * Copyright (c) 2016 Cisco and/or its affiliates. - * Licensed under the Apache License, Version 2.0 (the "License"); - * you may not use this file except in compliance with the License. - * You may obtain a copy of the License at: - * - * http://www.apache.org/licenses/LICENSE-2.0 - * - * Unless required by applicable law or agreed to in writing, software - * distributed under the License is distributed on an "AS IS" BASIS, - * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. - * See the License for the specific language governing permissions and - * limitations under the License. - */ - -#ifndef include_vnet_pot_util_h -#define include_vnet_pot_util_h - -#include -#define debug_ioam debug_ioam_fn -/* Dont change this size 256. This is there across multiple components */ -#define PATH_NAME_SIZE 256 - -/* Ring size. this should be same as the one in ODL. Do not change this - without change in ODL. */ -#define MAX_POT_PROFILES 2 - -/** - * Usage: - * - * On any node that participates in Proof of Transit: - * - * Step 1: Initialize this library by calling pot_init() - * Step 2: Setup a proof of transit profile that contains all the parameters needed to compute cumulative: - * Call these functions: - * pot_profile_find - * pot_profile_create - * pot_profile_set_bit_mask - To setup how large we want the numbers used in the computation and random number <= 64 bits - * Step 2a: For validator do this: - * pot_set_validator - * Step 2b: On initial node enable the profile to be used: - * pot_profile_set_active / pot_profile_get_active will return the profile - * Step 3a: At the initial node to generate Random number that will be read by all other nodes: - * pot_generate_random - * Step 3b: At all nodes including initial and verifier call this to compute cumulative: - * pot_update_cumulative - * Step 4: At the verifier: - * pot_validate - * - */ - -typedef struct pot_profile_ -{ - u8 id : 1; - u8 valid : 1; - u8 in_use : 1; - u64 random; - u8 validator; - u64 secret_key; - u64 secret_share; - u64 prime; - u64 lpc; - u64 poly_pre_eval; - u64 bit_mask; - u64 limit; - double primeinv; - u64 total_pkts_using_this_profile; -} pot_profile; - -typedef struct { - /* Name of the default profile list in use*/ - u8 *profile_list_name; - pot_profile profile_list[MAX_POT_PROFILES]; - /* number of profiles in the list */ - u8 active_profile_id : 1; - - /* API message ID base */ - u16 msg_id_base; - - /* convenience */ - vlib_main_t * vlib_main; - vnet_main_t * vnet_main; -} pot_main_t; - -extern pot_main_t pot_main; - -/* - * Initialize proof of transit - */ -int pot_util_init(void); -void pot_profile_list_init(u8 * name); - - -/* - * Find a pot profile by ID - */ -pot_profile *pot_profile_find(u8 id); - -static inline u16 pot_profile_get_id(pot_profile * profile) -{ - if (profile) - { - return (profile->id); - } - return (0); -} - -/* setup and clean up profile */ -int pot_profile_create(pot_profile * profile, u64 prime, - u64 poly2, u64 lpc, u64 secret_share); -/* - * Setup profile as a validator - */ -int pot_set_validator(pot_profile * profile, u64 key); - -/* - * Setup max bits to be used for random number generation - */ -#define MAX_BITS 64 -int pot_profile_set_bit_mask(pot_profile * profile, u16 bits); - -/* - * Given a random and cumulative compute the new cumulative for a given profile - */ -u64 pot_update_cumulative(pot_profile * profile, u64 cumulative, u64 random); - -/* - * return True if the cumulative matches secret from a profile - */ -u8 pot_validate(pot_profile * profile, u64 cumulative, u64 random); - -/* - * Utility function to get random number per pack - */ -u64 pot_generate_random(pot_profile * profile); - - -extern void clear_pot_profiles(); -extern int pot_profile_list_is_enabled(u8 *name); - -static inline u8 pot_is_decap(pot_profile * p) -{ - return (p->validator == 1); -} - -static inline int pot_profile_set_active (u8 id) -{ - pot_main_t *sm = &pot_main; - pot_profile *profile = NULL; - pot_profile *current_active_prof = NULL; - - current_active_prof = pot_profile_find(sm->active_profile_id); - profile = pot_profile_find(id); - if (profile && profile->valid) { - sm->active_profile_id = id; - current_active_prof->in_use = 0; - profile->in_use = 1; - return(0); - } - return(-1); -} -static inline u8 pot_profile_get_active_id (void) -{ - pot_main_t *sm = &pot_main; - return (sm->active_profile_id); -} - -static inline pot_profile * pot_profile_get_active (void) -{ - pot_main_t *sm = &pot_main; - pot_profile *profile = NULL; - profile = pot_profile_find(sm->active_profile_id); - if (profile && profile->in_use) - return(profile); - return (NULL); -} - -static inline void pot_profile_reset_usage_stats (pot_profile *pow) -{ - if (pow) { - pow->total_pkts_using_this_profile = 0; - } -} - -static inline void pot_profile_incr_usage_stats (pot_profile *pow) -{ - if (pow) { - pow->total_pkts_using_this_profile++; - } -} - - -#endif diff --git a/plugins/ioam-plugin/ioam/lib-trace/jvpp_ioam_trace.c b/plugins/ioam-plugin/ioam/lib-trace/jvpp_ioam_trace.c deleted file mode 100644 index 1d878ea3..00000000 --- a/plugins/ioam-plugin/ioam/lib-trace/jvpp_ioam_trace.c +++ /dev/null @@ -1,124 +0,0 @@ -/* - * Copyright (c) 2016 Cisco and/or its affiliates. - * Licensed under the Apache License, Version 2.0 (the "License"); - * you may not use this file except in compliance with the License. - * You may obtain a copy of the License at: - * - * http://www.apache.org/licenses/LICENSE-2.0 - * - * Unless required by applicable law or agreed to in writing, software - * distributed under the License is distributed on an "AS IS" BASIS, - * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. - * See the License for the specific language governing permissions and - * limitations under the License. - */ - -#include - -#include -#define vl_typedefs /* define message structures */ -#include -#undef vl_typedefs - -#define vl_endianfun -#include -#undef vl_endianfun - -#define vl_print(handle, ...) -#define vl_printfun -#include -#undef vl_printfun - -/* Get the API version number */ -#define vl_api_version(n,v) static u32 api_version=(v); -#include -#undef vl_api_version - -#include -#include -#include - -#if VPPJNI_DEBUG == 1 - #define DEBUG_LOG(...) clib_warning(__VA_ARGS__) -#else - #define DEBUG_LOG(...) -#endif - -#include - -#include "ioam/jvpp/io_fd_vpp_jvpp_ioamtrace_JVppIoamtraceImpl.h" -#include "jvpp_ioam_trace.h" -#include "ioam/jvpp/jvpp_ioam_trace_gen.h" - -/* - * Class: io_fd_vpp_jvpp_ioamtrace_JVppIoamtraceImpl - * Method: init0 - * Signature: (JI)V - */ -JNIEXPORT void JNICALL Java_io_fd_vpp_jvpp_ioamtrace_JVppIoamtraceImpl_init0 - (JNIEnv *env, jclass clazz, jobject callback, jlong queue_address, jint my_client_index) { - ioamtrace_main_t * plugin_main = &ioamtrace_main; - u8 * name; - clib_warning ("Java_io_fd_vpp_jvpp_ioamtrace_JVppIoamtraceImpl_init0"); - - plugin_main->my_client_index = my_client_index; - plugin_main->vl_input_queue = (unix_shared_memory_queue_t *)queue_address; - - name = format (0, "ioam_trace_%08x%c", api_version, 0); - plugin_main->msg_id_base = vl_client_get_first_plugin_msg_id ((char *) name); - - if (plugin_main->msg_id_base == (u16) ~0) { - jclass exClass = (*env)->FindClass(env, "java/lang/IllegalStateException"); - (*env)->ThrowNew(env, exClass, "ioam_trace plugin is not loaded in VPP"); - } else { - plugin_main->callbackObject = (*env)->NewGlobalRef(env, callback); - plugin_main->callbackClass = (jclass)(*env)->NewGlobalRef(env, (*env)->GetObjectClass(env, callback)); - - #define _(N,n) \ - vl_msg_api_set_handlers(VL_API_##N + plugin_main->msg_id_base, #n, \ - vl_api_##n##_t_handler, \ - vl_noop_handler, \ - vl_api_##n##_t_endian, \ - vl_api_##n##_t_print, \ - sizeof(vl_api_##n##_t), 1); - foreach_api_reply_handler; - #undef _ - } -} - -JNIEXPORT void JNICALL Java_io_fd_vpp_jvpp_ioamtrace_JVppIoamtraceImpl_close0 -(JNIEnv *env, jclass clazz) { - ioamtrace_main_t * plugin_main = &ioamtrace_main; - - // cleanup: - (*env)->DeleteGlobalRef(env, plugin_main->callbackClass); - (*env)->DeleteGlobalRef(env, plugin_main->callbackObject); - - plugin_main->callbackClass = NULL; - plugin_main->callbackObject = NULL; -} - -/* Attach thread to JVM and cache class references when initiating JVPP iOAM Trace */ -jint JNI_OnLoad(JavaVM *vm, void *reserved) { - JNIEnv* env; - - if ((*vm)->GetEnv(vm, (void**) &env, JNI_VERSION_1_8) != JNI_OK) { - return JNI_EVERSION; - } - - if (cache_class_references(env) != 0) { - clib_warning ("Failed to cache class references\n"); - return JNI_ERR; - } - - return JNI_VERSION_1_8; -} - -/* Clean up cached references when disposing JVPP iOAM Trace */ -void JNI_OnUnload(JavaVM *vm, void *reserved) { - JNIEnv* env; - if ((*vm)->GetEnv(vm, (void**) &env, JNI_VERSION_1_8) != JNI_OK) { - return; - } - delete_class_references(env); -} diff --git a/plugins/ioam-plugin/ioam/lib-trace/jvpp_ioam_trace.h b/plugins/ioam-plugin/ioam/lib-trace/jvpp_ioam_trace.h deleted file mode 100644 index 9fc16c15..00000000 --- a/plugins/ioam-plugin/ioam/lib-trace/jvpp_ioam_trace.h +++ /dev/null @@ -1,45 +0,0 @@ -/* - * Copyright (c) 2016 Cisco and/or its affiliates. - * Licensed under the Apache License, Version 2.0 (the "License"); - * you may not use this file except in compliance with the License. - * You may obtain a copy of the License at: - * - * http://www.apache.org/licenses/LICENSE-2.0 - * - * Unless required by applicable law or agreed to in writing, software - * distributed under the License is distributed on an "AS IS" BASIS, - * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. - * See the License for the specific language governing permissions and - * limitations under the License. - */ -#ifndef __included_jvpp_ioam_trace_h__ -#define __included_jvpp_ioam_trace_h__ - -#include -#include -#include -#include -#include -#include - -/* Global state for JVPP-IOAM-TRACE */ -typedef struct { - /* Base message index for the trace plugin */ - u16 msg_id_base; - - /* Pointer to shared memory queue */ - unix_shared_memory_queue_t * vl_input_queue; - - /* VPP api client index */ - u32 my_client_index; - - /* Callback object and class references enabling asynchronous Java calls */ - jobject callbackObject; - jclass callbackClass; - -} ioamtrace_main_t; - -ioamtrace_main_t ioamtrace_main __attribute__((aligned (64))); - - -#endif /* __included_jvpp_ioam_trace_h__ */ diff --git a/plugins/ioam-plugin/ioam/lib-trace/trace.api b/plugins/ioam-plugin/ioam/lib-trace/trace.api deleted file mode 100644 index cb958325..00000000 --- a/plugins/ioam-plugin/ioam/lib-trace/trace.api +++ /dev/null @@ -1,92 +0,0 @@ -/* Hey Emacs use -*- mode: C -*- */ -/* - * Copyright (c) 2016 Cisco and/or its affiliates. - * Licensed under the Apache License, Version 2.0 (the "License"); - * you may not use this file except in compliance with the License. - * You may obtain a copy of the License at: - * - * http://www.apache.org/licenses/LICENSE-2.0 - * - * Unless required by applicable law or agreed to in writing, software - * distributed under the License is distributed on an "AS IS" BASIS, - * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. - * See the License for the specific language governing permissions and - * limitations under the License. - */ - - -/** \brief iOAM6 Trace - Set the iOAM6 trace profile - @param trace_type - Type of trace requested - @param num_elts - Number of trace elements to be inserted - @param node_id - Trace Node ID - @param trace_tsp- Timestamp resolution - @param app_data - Application specific opaque -*/ -define trace_profile_add { - u32 client_index; - u32 context; - u8 trace_type; - u8 num_elts; - u8 trace_tsp; - u32 node_id; - u32 app_data; -}; - -/** \brief Trace profile add / del response - @param context - sender context, to match reply w/ request - @param retval - return value for request -*/ -define trace_profile_add_reply { - u32 context; - i32 retval; -}; - - - -/** \brief Delete trace Profile - @param client_index - opaque cookie to identify the sender - @param context - sender context, to match reply w/ request -*/ -define trace_profile_del { - u32 client_index; - u32 context; -}; - -/** \brief Trace profile add / del response - @param context - sender context, to match reply w/ request - @param retval - return value for request -*/ -define trace_profile_del_reply { - u32 context; - i32 retval; -}; - - - -/** \brief Show trace Profile - @param client_index - opaque cookie to identify the sender - @param context - sender context, to match reply w/ request -*/ -define trace_profile_show_config { - u32 client_index; - u32 context; -}; - -/** \brief Show trace config response - @param context - sender context, to match reply w/ request - @param retval - return value for request - @param trace_type - Type of trace requested - @param num_elts - Number of trace elements to be inserted - @param node_id - Trace Node ID - @param trace_tsp- Timestamp resolution - @param app_data - Application specific opaque -*/ -define trace_profile_show_config_reply { - u32 context; - i32 retval; - u8 trace_type; - u8 num_elts; - u8 trace_tsp; - u32 node_id; - u32 app_data; -}; diff --git a/plugins/ioam-plugin/ioam/lib-trace/trace_all_api_h.h b/plugins/ioam-plugin/ioam/lib-trace/trace_all_api_h.h deleted file mode 100644 index 223f9545..00000000 --- a/plugins/ioam-plugin/ioam/lib-trace/trace_all_api_h.h +++ /dev/null @@ -1,16 +0,0 @@ -/* - * Copyright (c) 2016 Cisco and/or its affiliates. - * Licensed under the Apache License, Version 2.0 (the "License"); - * you may not use this file except in compliance with the License. - * You may obtain a copy of the License at: - * - * http://www.apache.org/licenses/LICENSE-2.0 - * - * Unless required by applicable law or agreed to in writing, software - * distributed under the License is distributed on an "AS IS" BASIS, - * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. - * See the License for the specific language governing permissions and - * limitations under the License. - */ -/* Include the generated file, see BUILT_SOURCES in Makefile.am */ -#include diff --git a/plugins/ioam-plugin/ioam/lib-trace/trace_api.c b/plugins/ioam-plugin/ioam/lib-trace/trace_api.c deleted file mode 100644 index 7e0d708e..00000000 --- a/plugins/ioam-plugin/ioam/lib-trace/trace_api.c +++ /dev/null @@ -1,252 +0,0 @@ -/* - * Copyright (c) 2016 Cisco and/or its affiliates. - * Licensed under the Apache License, Version 2.0 (the "License"); - * you may not use this file except in compliance with the License. - * You may obtain a copy of the License at: - * - * http://www.apache.org/licenses/LICENSE-2.0 - * - * Unless required by applicable law or agreed to in writing, software - * distributed under the License is distributed on an "AS IS" BASIS, - * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. - * See the License for the specific language governing permissions and - * limitations under the License. - */ -/* - *------------------------------------------------------------------ - * trace_api.c - iOAM Trace related APIs to create - * and maintain profiles - *------------------------------------------------------------------ - */ - -#include -#include -#include - -#include -#include -#include - -/* define message IDs */ -#include - -/* define message structures */ -#define vl_typedefs -#include -#undef vl_typedefs - -/* define generated endian-swappers */ -#define vl_endianfun -#include -#undef vl_endianfun - -/* instantiate all the print functions we know about */ -#define vl_print(handle, ...) vlib_cli_output (handle, __VA_ARGS__) -#define vl_printfun -#include -#undef vl_printfun - -/* Get the API version number */ -#define vl_api_version(n,v) static u32 api_version=(v); -#include -#undef vl_api_version - -/* - * A handy macro to set up a message reply. - * Assumes that the following variables are available: - * mp - pointer to request message - * rmp - pointer to reply message type - * rv - return value - */ - -#define TRACE_REPLY_MACRO(t) \ -do { \ - unix_shared_memory_queue_t * q = \ - vl_api_client_index_to_input_queue (mp->client_index); \ - if (!q) \ - return; \ - \ - rmp = vl_msg_api_alloc (sizeof (*rmp)); \ - rmp->_vl_msg_id = ntohs((t)+sm->msg_id_base); \ - rmp->context = mp->context; \ - rmp->retval = ntohl(rv); \ - \ - vl_msg_api_send_shmem (q, (u8 *)&rmp); \ -} while(0); - -/* *INDENT-OFF* */ -#define TRACE_REPLY_MACRO2(t, body) \ -do { \ - unix_shared_memory_queue_t * q; \ - rv = vl_msg_api_pd_handler (mp, rv); \ - q = vl_api_client_index_to_input_queue (mp->client_index); \ - if (!q) \ - return; \ - \ - rmp = vl_msg_api_alloc (sizeof (*rmp)); \ - rmp->_vl_msg_id = ntohs((t)+sm->msg_id_base); \ - rmp->context = mp->context; \ - rmp->retval = ntohl(rv); \ - do {body;} while (0); \ - vl_msg_api_send_shmem (q, (u8 *)&rmp); \ -} while(0); -/* *INDENT-ON* */ - -/* List of message types that this plugin understands */ - -#define foreach_trace_plugin_api_msg \ -_(TRACE_PROFILE_ADD, trace_profile_add) \ -_(TRACE_PROFILE_DEL, trace_profile_del) \ -_(TRACE_PROFILE_SHOW_CONFIG, trace_profile_show_config) - -static void vl_api_trace_profile_add_t_handler - (vl_api_trace_profile_add_t * mp) -{ - trace_main_t *sm = &trace_main; - int rv = 0; - vl_api_trace_profile_add_reply_t *rmp; - trace_profile *profile = NULL; - - profile = trace_profile_find (); - if (profile) - { - rv = - trace_profile_create (profile, mp->trace_type, mp->num_elts, - mp->trace_tsp, ntohl (mp->node_id), - ntohl (mp->app_data)); - if (rv != 0) - goto ERROROUT; - } - else - { - rv = -3; - } -ERROROUT: - TRACE_REPLY_MACRO (VL_API_TRACE_PROFILE_ADD_REPLY); -} - - -static void vl_api_trace_profile_del_t_handler - (vl_api_trace_profile_del_t * mp) -{ - trace_main_t *sm = &trace_main; - int rv = 0; - vl_api_trace_profile_del_reply_t *rmp; - - clear_trace_profiles (); - - TRACE_REPLY_MACRO (VL_API_TRACE_PROFILE_DEL_REPLY); -} - -static void vl_api_trace_profile_show_config_t_handler - (vl_api_trace_profile_show_config_t * mp) -{ - trace_main_t *sm = &trace_main; - vl_api_trace_profile_show_config_reply_t *rmp; - int rv = 0; - trace_profile *profile = trace_profile_find (); - if (profile->valid) - { - TRACE_REPLY_MACRO2 (VL_API_TRACE_PROFILE_SHOW_CONFIG_REPLY, - rmp->trace_type = profile->trace_type; - rmp->num_elts = profile->num_elts; - rmp->trace_tsp = profile->trace_tsp; - rmp->node_id = htonl (profile->node_id); - rmp->app_data = htonl (profile->app_data); - ); - } - else - { - TRACE_REPLY_MACRO2 (VL_API_TRACE_PROFILE_SHOW_CONFIG_REPLY, - rmp->trace_type = 0; - rmp->num_elts = 0; rmp->trace_tsp = 0; - rmp->node_id = 0; rmp->app_data = 0; - ); - } -} - -/* - * This routine exists to convince the vlib plugin framework that - * we haven't accidentally copied a random .dll into the plugin directory. - * - * Also collects global variable pointers passed from the vpp engine - */ - -clib_error_t * -vlib_plugin_register (vlib_main_t * vm, vnet_plugin_handoff_t * h, - int from_early_init) -{ - trace_main_t *sm = &trace_main; - clib_error_t *error = 0; - - sm->vlib_main = vm; - sm->vnet_main = h->vnet_main; - return error; -} - -/* Set up the API message handling tables */ -static clib_error_t * -trace_plugin_api_hookup (vlib_main_t * vm) -{ - trace_main_t *sm = &trace_main; -#define _(N,n) \ - vl_msg_api_set_handlers((VL_API_##N + sm->msg_id_base), \ - #n, \ - vl_api_##n##_t_handler, \ - vl_noop_handler, \ - vl_api_##n##_t_endian, \ - vl_api_##n##_t_print, \ - sizeof(vl_api_##n##_t), 1); - foreach_trace_plugin_api_msg; -#undef _ - - return 0; -} - -#define vl_msg_name_crc_list -#include -#undef vl_msg_name_crc_list - -static void -setup_message_id_table (trace_main_t * sm, api_main_t * am) -{ -#define _(id,n,crc) \ - vl_msg_api_add_msg_name_crc (am, #n "_" #crc, id + sm->msg_id_base); - foreach_vl_msg_name_crc_trace; -#undef _ -} - -static clib_error_t * -trace_init (vlib_main_t * vm) -{ - trace_main_t *sm = &trace_main; - clib_error_t *error = 0; - u8 *name; - - bzero (sm, sizeof (trace_main)); - (void) trace_util_init (); - name = format (0, "ioam_trace_%08x%c", api_version, 0); - - /* Ask for a correctly-sized block of API message decode slots */ - sm->msg_id_base = vl_msg_api_get_msg_ids - ((char *) name, VL_MSG_FIRST_AVAILABLE); - - error = trace_plugin_api_hookup (vm); - - /* Add our API messages to the global name_crc hash table */ - setup_message_id_table (sm, &api_main); - - vec_free (name); - - return error; -} - -VLIB_INIT_FUNCTION (trace_init); - -/* - * fd.io coding-style-patch-verification: ON - * - * Local Variables: - * eval: (c-set-style "gnu") - * End: - */ diff --git a/plugins/ioam-plugin/ioam/lib-trace/trace_msg_enum.h b/plugins/ioam-plugin/ioam/lib-trace/trace_msg_enum.h deleted file mode 100644 index 78c35665..00000000 --- a/plugins/ioam-plugin/ioam/lib-trace/trace_msg_enum.h +++ /dev/null @@ -1,28 +0,0 @@ -/* - * Copyright (c) 2016 Cisco and/or its affiliates. - * Licensed under the Apache License, Version 2.0 (the "License"); - * you may not use this file except in compliance with the License. - * You may obtain a copy of the License at: - * - * http://www.apache.org/licenses/LICENSE-2.0 - * - * Unless required by applicable law or agreed to in writing, software - * distributed under the License is distributed on an "AS IS" BASIS, - * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. - * See the License for the specific language governing permissions and - * limitations under the License. - */ -#ifndef included_trace_msg_enum_h -#define included_trace_msg_enum_h - -#include - -#define vl_msg_id(n,h) n, -typedef enum { -#include - /* We'll want to know how many messages IDs we need... */ - VL_MSG_FIRST_AVAILABLE, -} vl_msg_id_t; -#undef vl_msg_id - -#endif /* included_trace_msg_enum_h */ diff --git a/plugins/ioam-plugin/ioam/lib-trace/trace_test.c b/plugins/ioam-plugin/ioam/lib-trace/trace_test.c deleted file mode 100644 index 111dd461..00000000 --- a/plugins/ioam-plugin/ioam/lib-trace/trace_test.c +++ /dev/null @@ -1,292 +0,0 @@ -/* - * Copyright (c) 2016 Cisco and/or its affiliates. - * Licensed under the Apache License, Version 2.0 (the "License"); - * you may not use this file except in compliance with the License. - * You may obtain a copy of the License at: - * - * http://www.apache.org/licenses/LICENSE-2.0 - * - * Unless required by applicable law or agreed to in writing, software - * distributed under the License is distributed on an "AS IS" BASIS, - * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. - * See the License for the specific language governing permissions and - * limitations under the License. - */ -/* - *------------------------------------------------------------------ - * trace_test.c - test harness for trace plugin - *------------------------------------------------------------------ - */ - -#include -#include -#include -#include -#include - -/* Declare message IDs */ -#include - -/* define message structures */ -#define vl_typedefs -#include -#undef vl_typedefs - -/* declare message handlers for each api */ - -#define vl_endianfun /* define message structures */ -#include -#undef vl_endianfun - -/* instantiate all the print functions we know about */ -#define vl_print(handle, ...) -#define vl_printfun -#include -#undef vl_printfun - -/* Get the API version number. */ -#define vl_api_version(n,v) static u32 api_version=(v); -#include -#undef vl_api_version - - -typedef struct -{ - /* API message ID base */ - u16 msg_id_base; - vat_main_t *vat_main; -} trace_test_main_t; - -trace_test_main_t trace_test_main; - -#define foreach_standard_reply_retval_handler \ -_(trace_profile_add_reply) \ -_(trace_profile_del_reply) - -#define foreach_custom_reply_handler \ -_(trace_profile_show_config_reply, \ - if(mp->trace_type) \ - { \ - errmsg(" Trace Type : 0x%x (%d)\n",mp->trace_type, mp->trace_type); \ - errmsg(" Trace timestamp precision : %d \n",mp->trace_tsp); \ - errmsg(" Node Id : 0x%x (%d)\n",htonl(mp->node_id), htonl(mp->node_id)); \ - errmsg(" App Data : 0x%x (%d)\n",htonl(mp->app_data), htonl(mp->app_data)); \ - } \ - else errmsg("No valid trace profile configuration found\n");) -#define _(n) \ - static void vl_api_##n##_t_handler \ - (vl_api_##n##_t * mp) \ - { \ - vat_main_t * vam = trace_test_main.vat_main; \ - i32 retval = ntohl(mp->retval); \ - if (vam->async_mode) { \ - vam->async_errors += (retval < 0); \ - } else { \ - vam->retval = retval; \ - vam->result_ready = 1; \ - } \ - } -foreach_standard_reply_retval_handler; -#undef _ - -#define _(n,body) \ - static void vl_api_##n##_t_handler \ - (vl_api_##n##_t * mp) \ - { \ - vat_main_t * vam = trace_test_main.vat_main; \ - i32 retval = ntohl(mp->retval); \ - if (vam->async_mode) { \ - vam->async_errors += (retval < 0); \ - } else { \ - vam->retval = retval; \ - vam->result_ready = 1; \ - } \ - if(retval>=0)do{body;} while(0); \ - else errmsg("Error, retval: %d",retval); \ - } -foreach_custom_reply_handler; -#undef _ -/* - * Table of message reply handlers, must include boilerplate handlers - * we just generated - */ -#define foreach_vpe_api_reply_msg \ -_(TRACE_PROFILE_ADD_REPLY, trace_profile_add_reply) \ -_(TRACE_PROFILE_DEL_REPLY, trace_profile_del_reply) \ -_(TRACE_PROFILE_SHOW_CONFIG_REPLY, trace_profile_show_config_reply) - - -/* M: construct, but don't yet send a message */ - -#define M(T,t) \ -do { \ - vam->result_ready = 0; \ - mp = vl_msg_api_alloc(sizeof(*mp)); \ - memset (mp, 0, sizeof (*mp)); \ - mp->_vl_msg_id = ntohs (VL_API_##T + sm->msg_id_base); \ - mp->client_index = vam->my_client_index; \ -} while(0); - -#define M2(T,t,n) \ -do { \ - vam->result_ready = 0; \ - mp = vl_msg_api_alloc(sizeof(*mp)+(n)); \ - memset (mp, 0, sizeof (*mp)); \ - mp->_vl_msg_id = ntohs (VL_API_##T + sm->msg_id_base); \ - mp->client_index = vam->my_client_index; \ -} while(0); - -/* S: send a message */ -#define S (vl_msg_api_send_shmem (vam->vl_input_queue, (u8 *)&mp)) - -/* W: wait for results, with timeout */ -#define W \ -do { \ - timeout = vat_time_now (vam) + 1.0; \ - \ - while (vat_time_now (vam) < timeout) { \ - if (vam->result_ready == 1) { \ - return (vam->retval); \ - } \ - } \ - return -99; \ -} while(0); - - -static int -api_trace_profile_add (vat_main_t * vam) -{ - trace_test_main_t *sm = &trace_test_main; - unformat_input_t *input = vam->input; - vl_api_trace_profile_add_t *mp; - u8 trace_type = 0; - u8 num_elts = 0; - int rv = 0; - u32 node_id = 0; - u32 app_data = 0; - u8 trace_tsp = 0; - f64 timeout; - - while (unformat_check_input (input) != UNFORMAT_END_OF_INPUT) - { - if (unformat (input, "trace-type 0x%x", &trace_type)) - ; - else if (unformat (input, "trace-elts %d", &num_elts)) - ; - else if (unformat (input, "trace-tsp %d", &trace_tsp)) - ; - else if (unformat (input, "node-id 0x%x", &node_id)) - ; - else if (unformat (input, "app-data 0x%x", &app_data)) - ; - - else - break; - } - - - M (TRACE_PROFILE_ADD, trace_profile_add); - - mp->trace_type = trace_type; - mp->trace_tsp = trace_tsp; - mp->node_id = htonl (node_id); - mp->app_data = htonl (app_data); - mp->num_elts = num_elts; - - S; - W; - - return (rv); -} - - - -static int -api_trace_profile_del (vat_main_t * vam) -{ - trace_test_main_t *sm = &trace_test_main; - vl_api_trace_profile_del_t *mp; - f64 timeout; - - M (TRACE_PROFILE_DEL, trace_profile_del); - S; - W; - return 0; -} - -static int -api_trace_profile_show_config (vat_main_t * vam) -{ - trace_test_main_t *sm = &trace_test_main; - vl_api_trace_profile_show_config_t *mp; - f64 timeout; - M (TRACE_PROFILE_SHOW_CONFIG, trace_profile_show_config); - S; - W; - return 0; -} - -/* - * List of messages that the api test plugin sends, - * and that the data plane plugin processes - */ -#define foreach_vpe_api_msg \ -_(trace_profile_add, ""\ - "trace-type <0x1f|0x3|0x9|0x11|0x19> trace-elts trace-tsp <0|1|2|3> node-id app-data ") \ -_(trace_profile_del, "[id ]") \ -_(trace_profile_show_config, "[id ]") - - -void -vat_api_hookup (vat_main_t * vam) -{ - trace_test_main_t *sm = &trace_test_main; - /* Hook up handlers for replies from the data plane plug-in */ -#define _(N,n) \ - vl_msg_api_set_handlers((VL_API_##N + sm->msg_id_base), \ - #n, \ - vl_api_##n##_t_handler, \ - vl_noop_handler, \ - vl_api_##n##_t_endian, \ - vl_api_##n##_t_print, \ - sizeof(vl_api_##n##_t), 1); - foreach_vpe_api_reply_msg; -#undef _ - - /* API messages we can send */ -#define _(n,h) hash_set_mem (vam->function_by_name, #n, api_##n); - foreach_vpe_api_msg; -#undef _ - - /* Help strings */ -#define _(n,h) hash_set_mem (vam->help_by_name, #n, h); - foreach_vpe_api_msg; -#undef _ -} - -clib_error_t * -vat_plugin_register (vat_main_t * vam) -{ - trace_test_main_t *sm = &trace_test_main; - u8 *name; - - sm->vat_main = vam; - - name = format (0, "ioam_trace_%08x%c", api_version, 0); - sm->msg_id_base = vl_client_get_first_plugin_msg_id ((char *) name); - - if (sm->msg_id_base != (u16) ~ 0) - vat_api_hookup (vam); - - vec_free (name); - - return 0; -} - -/* - * fd.io coding-style-patch-verification: ON - * - * Local Variables: - * eval: (c-set-style "gnu") - * End: - */ diff --git a/plugins/ioam-plugin/ioam/lib-trace/trace_util.c b/plugins/ioam-plugin/ioam/lib-trace/trace_util.c deleted file mode 100644 index 5c7f1eef..00000000 --- a/plugins/ioam-plugin/ioam/lib-trace/trace_util.c +++ /dev/null @@ -1,206 +0,0 @@ -/* - * Copyright (c) 2016 Cisco and/or its affiliates. - * Licensed under the Apache License, Version 2.0 (the "License"); - * you may not use this file except in compliance with the License. - * You may obtain a copy of the License at: - * - * http://www.apache.org/licenses/LICENSE-2.0 - * - * Unless required by applicable law or agreed to in writing, software - * distributed under the License is distributed on an "AS IS" BASIS, - * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. - * See the License for the specific language governing permissions and - * limitations under the License. - */ -#include -#include -#include -#include -#include -#include "trace_util.h" - -trace_main_t trace_main; - -static int -trace_profile_cleanup (trace_profile * profile) -{ - - memset (profile, 0, sizeof (trace_profile)); - profile->trace_tsp = TSP_MICROSECONDS; /* Micro seconds */ - ip6_trace_profile_cleanup (); /* lib-trace_TODO: Remove this once IOAM-IPv6 transport is a plugin */ - return 0; - -} - -static int -trace_main_profiles_reset (void) -{ - int rv; - - trace_main_t *sm = &trace_main; - rv = trace_profile_cleanup (&(sm->profile)); - return (rv); -} - -int -trace_util_init (void) -{ - int rv; - - rv = trace_main_profiles_reset (); - return (rv); -} - - -int -trace_profile_create (trace_profile * profile, u8 trace_type, u8 num_elts, - u32 trace_tsp, u32 node_id, u32 app_data) -{ - - if (!trace_type || !num_elts || !(node_id)) - { - return (-1); - } - if (profile && !profile->valid) - { - //rv = trace_profile_cleanup (profile); - profile->trace_type = trace_type; - profile->num_elts = num_elts; - profile->trace_tsp = trace_tsp; - profile->node_id = node_id; - profile->app_data = app_data; - profile->valid = 1; - - /* lib-trace_TODO: Remove this once IOAM-IPv6 transport is a plugin */ - ip6_trace_profile_setup (); - return (0); - } - - return (-1); -} - - - -clib_error_t * -clear_trace_profile_command_fn (vlib_main_t * vm, - unformat_input_t * input, - vlib_cli_command_t * cmd) -{ - - trace_main_profiles_reset (); - return 0; -} - -void -clear_trace_profiles (void) -{ - clear_trace_profile_command_fn (0, 0, 0); -} - -/* *INDENT-OFF* */ -VLIB_CLI_COMMAND(clear_trace_profile_command) = -{ -.path = "clear ioam-trace profile", -.short_help = "clear ioam-trace profile [|all]", -.function = clear_trace_profile_command_fn, -}; -/* *INDENT-ON* */ - -static clib_error_t * -set_trace_profile_command_fn (vlib_main_t * vm, - unformat_input_t * input, - vlib_cli_command_t * cmd) -{ - u8 trace_type = 0; - u8 num_elts = 0; - u32 node_id = 0; - u32 app_data = 0; - u32 trace_tsp = 0; - trace_profile *profile = NULL; - while (unformat_check_input (input) != UNFORMAT_END_OF_INPUT) - { - if (unformat (input, "trace-type 0x%x", &trace_type)); - else if (unformat (input, "trace-elts %d", &num_elts)); - else if (unformat (input, "trace-tsp %d", &trace_tsp)); - else if (unformat (input, "node-id 0x%x", &node_id)); - else if (unformat (input, "app-data 0x%x", &app_data)); - else - break; - } - profile = trace_profile_find (); - if (profile) - { - trace_profile_create (profile, trace_type, num_elts, trace_tsp, - node_id, app_data); - } - return 0; -} - -/* *INDENT-OFF* */ -VLIB_CLI_COMMAND (set_trace_profile_command, static) = -{ -.path = "set ioam-trace profile", -.short_help = "set ioam-trace \ - trace-type <0x1f|0x3|0x9|0x11|0x19> trace-elts trace-tsp <0|1|2|3> \ - node-id app-data ", -.function = set_trace_profile_command_fn, -}; -/* *INDENT-ON* */ - -static clib_error_t * -show_trace_profile_command_fn (vlib_main_t * vm, - unformat_input_t * input, - vlib_cli_command_t * cmd) -{ - trace_profile *p = NULL; - u8 *s = 0; - p = trace_profile_find (); - if (!(p && p->valid)) - { - s = format (s, "\nTrace configuration not valid\n"); - vlib_cli_output (vm, "%v", s); - vec_free (s); - return 0; - } - s = format (s, " HOP BY HOP OPTIONS - TRACE CONFIG - \n"); - s = format (s, " Trace Type : 0x%x (%d)\n", - p->trace_type, p->trace_type); - s = - format (s, " Trace timestamp precision : %d (%s)\n", - p->trace_tsp, - (p->trace_tsp == - TSP_SECONDS) ? "Seconds" : ((p->trace_tsp == - TSP_MILLISECONDS) ? - "Milliseconds" - : (((p->trace_tsp == - TSP_MICROSECONDS) ? - "Microseconds" : - "Nanoseconds")))); - s = format (s, " Num of trace nodes : %d\n", p->num_elts); - s = - format (s, " Node-id : 0x%x (%d)\n", - p->node_id, p->node_id); - s = - format (s, " App Data : 0x%x (%d)\n", - p->app_data, p->app_data); - vlib_cli_output (vm, "%v", s); - vec_free (s); - return 0; -} - -/* *INDENT-OFF* */ -VLIB_CLI_COMMAND (show_trace_profile_command, static) = -{ -.path = "show ioam-trace profile", -.short_help = "show ioam-trace profile", -.function = show_trace_profile_command_fn, -}; -/* *INDENT-ON* */ - -/* - * fd.io coding-style-patch-verification: ON - * - * Local Variables: - * eval: (c-set-style "gnu") - * End: - */ diff --git a/plugins/ioam-plugin/ioam/lib-trace/trace_util.h b/plugins/ioam-plugin/ioam/lib-trace/trace_util.h deleted file mode 100644 index 556f07ee..00000000 --- a/plugins/ioam-plugin/ioam/lib-trace/trace_util.h +++ /dev/null @@ -1,247 +0,0 @@ -/* - * trace_util.h -- Trace Profile Utility header - * - * Copyright (c) 2016 Cisco and/or its affiliates. - * Licensed under the Apache License, Version 2.0 (the "License"); - * you may not use this file except in compliance with the License. - * You may obtain a copy of the License at: - * - * http://www.apache.org/licenses/LICENSE-2.0 - * - * Unless required by applicable law or agreed to in writing, software - * distributed under the License is distributed on an "AS IS" BASIS, - * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. - * See the License for the specific language governing permissions and - * limitations under the License. - */ - -#ifndef include_vnet_trace_util_h -#define include_vnet_trace_util_h - -#define debug_ioam debug_ioam_fn - - -/** - * Usage: - * - * On any node that participates in iOAM Trace. - * - * Step 1: Initialize this library by calling trace_init() - * Step 2: Setup a trace profile that contains all the parameters needed to compute cumulative: - * Call these functions: - * trace_profile_find - * trace_profile_create - * Step 2a: On initial node enable the profile to be used: - * trace_profile_set_active / trace_profile_get_active will return the profile - * Step 4: TBD - * trace_validate - * - */ - -typedef struct trace_profile_ -{ - u8 valid:1; - u8 trace_type; - u8 num_elts; - /* Configured node-id */ - u32 node_id; - u32 app_data; - u32 trace_tsp; -} trace_profile; - -typedef struct -{ - /* Name of the default profile list in use */ - trace_profile profile; - - /* API message ID base */ - u16 msg_id_base; - - /* convenience */ - vlib_main_t *vlib_main; - vnet_main_t *vnet_main; -} trace_main_t; - -extern trace_main_t trace_main; - -/* - * Initialize Trace profile - */ -int trace_util_init (void); - - -/* - * Find a trace profile - */ - -always_inline trace_profile * -trace_profile_find (void) -{ - trace_main_t *sm = &trace_main; - - return (&(sm->profile)); -} - - -/* setup and clean up profile */ -int trace_profile_create (trace_profile * profile, u8 trace_type, u8 num_elts, - u32 trace_tsp, u32 node_id, u32 app_data); - -void clear_trace_profiles (void); - - - -#define BIT_TTL_NODEID (1<<0) -#define BIT_ING_INTERFACE (1<<1) -#define BIT_EGR_INTERFACE (1<<2) -#define BIT_TIMESTAMP (1<<3) -#define BIT_APPDATA (1<<4) -#define TRACE_TYPE_MASK 0x1F /* Mask of all above bits */ - -/* - 0x00011111 iOAM-trace-type is 0x00011111 then the format of node - data is: - - 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 - +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ - | Hop_Lim | node_id | - +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ - | ingress_if_id | egress_if_id | - +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ - + timestamp + - +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ - | app_data | - +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ - -*/ -#define TRACE_TYPE_IF_TS_APP 0x1f -typedef struct -{ - u32 ttl_node_id; - u16 ingress_if; - u16 egress_if; - u32 timestamp; - u32 app_data; -} ioam_trace_if_ts_app_t; - -/* - 0x00000111 iOAM-trace-type is 0x00000111 then the format is: - - 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 - +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ - | Hop_Lim | node_id | - +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ - | ingress_if_id | egress_if_id | - +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ - -*/ - -#define TRACE_TYPE_IF 0x03 -typedef struct -{ - u32 ttl_node_id; - u16 ingress_if; - u16 egress_if; -} ioam_trace_if_t; - -/* - 0x00001001 iOAM-trace-type is 0x00001001 then the format is: - - 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 - +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ - | Hop_Lim | node_id | - +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ - + timestamp + - +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ - -*/ - -#define TRACE_TYPE_TS 0x09 -typedef struct -{ - u32 ttl_node_id; - u32 timestamp; -} ioam_trace_ts_t; - -/* - 0x00010001 iOAM-trace-type is 0x00010001 then the format is: - - - 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 - +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ - | Hop_Lim | node_id | - +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ - | app_data | - +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ - -*/ - - -#define TRACE_TYPE_APP 0x11 -typedef struct -{ - u32 ttl_node_id; - u32 app_data; -} ioam_trace_app_t; - -/* - - 0x00011001 iOAM-trace-type is 0x00011001 then the format is: - - 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 - +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ - | Hop_Lim | node_id | - +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ - + timestamp + - +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ - | app_data | - +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ -*/ - -#define TRACE_TYPE_TS_APP 0x19 -typedef struct -{ - u32 ttl_node_id; - u32 timestamp; - u32 app_data; -} ioam_trace_ts_app_t; - - - -static inline u8 -fetch_trace_data_size (u8 trace_type) -{ - u8 trace_data_size = 0; - - if (trace_type == TRACE_TYPE_IF_TS_APP) - trace_data_size = sizeof (ioam_trace_if_ts_app_t); - else if (trace_type == TRACE_TYPE_IF) - trace_data_size = sizeof (ioam_trace_if_t); - else if (trace_type == TRACE_TYPE_TS) - trace_data_size = sizeof (ioam_trace_ts_t); - else if (trace_type == TRACE_TYPE_APP) - trace_data_size = sizeof (ioam_trace_app_t); - else if (trace_type == TRACE_TYPE_TS_APP) - trace_data_size = sizeof (ioam_trace_ts_app_t); - - return trace_data_size; -} - -int ioam_trace_get_sizeof_handler (u32 * result); -int ip6_trace_profile_setup (void); -int ip6_trace_profile_cleanup (void); - -#define TSP_SECONDS 0 -#define TSP_MILLISECONDS 1 -#define TSP_MICROSECONDS 2 -#define TSP_NANOSECONDS 3 - -#endif - -/* - * fd.io coding-style-patch-verification: ON - * - * Local Variables: - * eval: (c-set-style "gnu") - * End: - */ diff --git a/plugins/ioam-plugin/ioam/lib-vxlan-gpe/ioam_decap.c b/plugins/ioam-plugin/ioam/lib-vxlan-gpe/ioam_decap.c deleted file mode 100644 index fd308657..00000000 --- a/plugins/ioam-plugin/ioam/lib-vxlan-gpe/ioam_decap.c +++ /dev/null @@ -1,223 +0,0 @@ -/* - * Copyright (c) 2015 Cisco and/or its affiliates. - * Licensed under the Apache License, Version 2.0 (the "License"); - * you may not use this file except in compliance with the License. - * You may obtain a copy of the License at: - * - * http://www.apache.org/licenses/LICENSE-2.0 - * - * Unless required by applicable law or agreed to in writing, software - * distributed under the License is distributed on an "AS IS" BASIS, - * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. - * See the License for the specific language governing permissions and - * limitations under the License. - */ -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include - -/* Statistics (not really errors) */ -#define foreach_vxlan_gpe_decap_ioam_v4_error \ -_(DECAPSULATED, "good packets decapsulated") - -static char *vxlan_gpe_decap_ioam_v4_error_strings[] = { -#define _(sym,string) string, - foreach_vxlan_gpe_decap_ioam_v4_error -#undef _ -}; - -typedef enum -{ -#define _(sym,str) VXLAN_GPE_DECAP_IOAM_V4_ERROR_##sym, - foreach_vxlan_gpe_decap_ioam_v4_error -#undef _ - VXLAN_GPE_DECAP_IOAM_V4_N_ERROR, -} vxlan_gpe_decap_ioam_v4_error_t; - - -always_inline void -vxlan_gpe_decap_ioam_v4_two_inline (vlib_main_t * vm, - vlib_node_runtime_t * node, - vxlan_gpe_main_t * ngm, - vlib_buffer_t * b0, vlib_buffer_t * b1, - u32 * next0, u32 * next1) -{ - vxlan_gpe_ioam_main_t *hm = &vxlan_gpe_ioam_main; - - next0[0] = next1[0] = hm->decap_v4_next_override; - vxlan_gpe_encap_decap_ioam_v4_one_inline (vm, node, b0, &next0[0], - VXLAN_GPE_DECAP_IOAM_V4_NEXT_DROP, - 0 /* use_adj */ ); - vxlan_gpe_encap_decap_ioam_v4_one_inline (vm, node, b1, &next0[1], - VXLAN_GPE_DECAP_IOAM_V4_NEXT_DROP, - 0 /* use_adj */ ); -} - - - -static uword -vxlan_gpe_decap_ioam (vlib_main_t * vm, - vlib_node_runtime_t * node, - vlib_frame_t * from_frame, u8 is_ipv6) -{ - u32 n_left_from, next_index, *from, *to_next; - vxlan_gpe_main_t *ngm = &vxlan_gpe_main; - vxlan_gpe_ioam_main_t *hm = &vxlan_gpe_ioam_main; - - from = vlib_frame_vector_args (from_frame); - n_left_from = from_frame->n_vectors; - - next_index = node->cached_next_index; - - while (n_left_from > 0) - { - u32 n_left_to_next; - - vlib_get_next_frame (vm, node, next_index, to_next, n_left_to_next); - - while (n_left_from >= 4 && n_left_to_next >= 2) - { - u32 bi0, bi1; - vlib_buffer_t *b0, *b1; - u32 next0, next1; - - next0 = next1 = hm->decap_v4_next_override; - - /* Prefetch next iteration. */ - { - vlib_buffer_t *p2, *p3; - - p2 = vlib_get_buffer (vm, from[2]); - p3 = vlib_get_buffer (vm, from[3]); - - vlib_prefetch_buffer_header (p2, LOAD); - vlib_prefetch_buffer_header (p3, LOAD); - - CLIB_PREFETCH (p2->data, 2 * CLIB_CACHE_LINE_BYTES, LOAD); - CLIB_PREFETCH (p3->data, 2 * CLIB_CACHE_LINE_BYTES, LOAD); - } - - bi0 = from[0]; - bi1 = from[1]; - to_next[0] = bi0; - to_next[1] = bi1; - from += 2; - to_next += 2; - n_left_to_next -= 2; - n_left_from -= 2; - - b0 = vlib_get_buffer (vm, bi0); - b1 = vlib_get_buffer (vm, bi1); - - - vlib_buffer_advance (b0, - -(word) (sizeof (udp_header_t) + - sizeof (ip4_header_t) + - sizeof (vxlan_gpe_header_t))); - vlib_buffer_advance (b1, - -(word) (sizeof (udp_header_t) + - sizeof (ip4_header_t) + - sizeof (vxlan_gpe_header_t))); - - vxlan_gpe_decap_ioam_v4_two_inline (vm, node, ngm, b0, b1, - &next0, &next1); - - - vlib_validate_buffer_enqueue_x2 (vm, node, next_index, to_next, - n_left_to_next, bi0, bi1, next0, - next1); - - if (PREDICT_FALSE (b0->flags & VLIB_BUFFER_IS_TRACED)) - { - vxlan_gpe_ioam_v4_trace_t *tr = vlib_add_trace (vm, node, b0, - sizeof (*tr)); - } - } - - while (n_left_from > 0 && n_left_to_next > 0) - { - u32 bi0; - vlib_buffer_t *b0; - u32 next0 = hm->decap_v4_next_override; - - bi0 = from[0]; - to_next[0] = bi0; - from += 1; - to_next += 1; - n_left_from -= 1; - n_left_to_next -= 1; - - b0 = vlib_get_buffer (vm, bi0); - - - vlib_buffer_advance (b0, - -(word) (sizeof (udp_header_t) + - sizeof (ip4_header_t) + - sizeof (vxlan_gpe_header_t))); - - next0 = hm->decap_v4_next_override; - vxlan_gpe_encap_decap_ioam_v4_one_inline (vm, node, b0, - &next0, - VXLAN_GPE_DECAP_IOAM_V4_NEXT_DROP, - 0 /* use_adj */ ); - - if (PREDICT_FALSE (b0->flags & VLIB_BUFFER_IS_TRACED)) - { - vxlan_gpe_ioam_v4_trace_t *tr = vlib_add_trace (vm, node, b0, - sizeof (*tr)); - } - vlib_validate_buffer_enqueue_x1 (vm, node, next_index, to_next, - n_left_to_next, bi0, next0); - } - - vlib_put_next_frame (vm, node, next_index, n_left_to_next); - } - - return from_frame->n_vectors; -} - - -static uword -vxlan_gpe_decap_ioam_v4 (vlib_main_t * vm, - vlib_node_runtime_t * node, - vlib_frame_t * from_frame) -{ - return vxlan_gpe_decap_ioam (vm, node, from_frame, 0); -} - - -/* *INDENT-OFF* */ -VLIB_REGISTER_NODE (vxlan_gpe_decap_ioam_v4_node) = { - .function = vxlan_gpe_decap_ioam_v4, - .name = "vxlan-gpe-decap-ioam-v4", - .vector_size = sizeof (u32), - .format_trace = format_vxlan_gpe_ioam_v4_trace, - .type = VLIB_NODE_TYPE_INTERNAL, - - .n_errors = ARRAY_LEN(vxlan_gpe_decap_ioam_v4_error_strings), - .error_strings = vxlan_gpe_decap_ioam_v4_error_strings, - - .n_next_nodes = VXLAN_GPE_DECAP_IOAM_V4_N_NEXT, - - .next_nodes = { - [VXLAN_GPE_DECAP_IOAM_V4_NEXT_POP] = "vxlan-gpe-pop-ioam-v4", - [VXLAN_GPE_DECAP_IOAM_V4_NEXT_DROP] = "error-drop", - }, -}; -/* *INDENT-ON* */ - - -/* - * fd.io coding-style-patch-verification: ON - * - * Local Variables: - * eval: (c-set-style "gnu") - * End: - */ diff --git a/plugins/ioam-plugin/ioam/lib-vxlan-gpe/ioam_encap.c b/plugins/ioam-plugin/ioam/lib-vxlan-gpe/ioam_encap.c deleted file mode 100644 index 4b18bfea..00000000 --- a/plugins/ioam-plugin/ioam/lib-vxlan-gpe/ioam_encap.c +++ /dev/null @@ -1,194 +0,0 @@ -/* - * Copyright (c) 2015 Cisco and/or its affiliates. - * Licensed under the Apache License, Version 2.0 (the "License"); - * you may not use this file except in compliance with the License. - * You may obtain a copy of the License at: - * - * http://www.apache.org/licenses/LICENSE-2.0 - * - * Unless required by applicable law or agreed to in writing, software - * distributed under the License is distributed on an "AS IS" BASIS, - * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. - * See the License for the specific language governing permissions and - * limitations under the License. - */ -#include -#include -#include -#include -#include -#include -#include -#include -#include - -/* Statistics (not really errors) */ -#define foreach_vxlan_gpe_encap_ioam_v4_error \ -_(ENCAPSULATED, "good packets encapsulated") - -static char *vxlan_gpe_encap_ioam_v4_error_strings[] = { -#define _(sym,string) string, - foreach_vxlan_gpe_encap_ioam_v4_error -#undef _ -}; - -typedef enum -{ -#define _(sym,str) VXLAN_GPE_ENCAP_IOAM_V4_ERROR_##sym, - foreach_vxlan_gpe_encap_ioam_v4_error -#undef _ - VXLAN_GPE_ENCAP_IOAM_V4_N_ERROR, -} vxlan_gpe_encap_ioam_v4_error_t; - -typedef enum -{ - VXLAN_GPE_ENCAP_IOAM_V4_NEXT_IP4_LOOKUP, - VXLAN_GPE_ENCAP_IOAM_V4_NEXT_DROP, - VXLAN_GPE_ENCAP_IOAM_V4_N_NEXT -} vxlan_gpe_encap_ioam_v4_next_t; - - -always_inline void -vxlan_gpe_encap_ioam_v4_two_inline (vlib_main_t * vm, - vlib_node_runtime_t * node, - vxlan_gpe_main_t * ngm, - vlib_buffer_t * b0, vlib_buffer_t * b1, - u32 * next0, u32 * next1) -{ - next0[0] = next1[0] = VXLAN_GPE_ENCAP_IOAM_V4_NEXT_IP4_LOOKUP; - vxlan_gpe_encap_decap_ioam_v4_one_inline (vm, node, b0, next0, - VXLAN_GPE_ENCAP_IOAM_V4_NEXT_DROP, - 0 /* use_adj */ ); - vxlan_gpe_encap_decap_ioam_v4_one_inline (vm, node, b1, next1, - VXLAN_GPE_ENCAP_IOAM_V4_NEXT_DROP, - 0 /* use_adj */ ); -} - - -static uword -vxlan_gpe_encap_ioam_v4 (vlib_main_t * vm, - vlib_node_runtime_t * node, - vlib_frame_t * from_frame) -{ - u32 n_left_from, next_index, *from, *to_next; - vxlan_gpe_main_t *ngm = &vxlan_gpe_main; - - from = vlib_frame_vector_args (from_frame); - n_left_from = from_frame->n_vectors; - - next_index = node->cached_next_index; - - while (n_left_from > 0) - { - u32 n_left_to_next; - - vlib_get_next_frame (vm, node, next_index, to_next, n_left_to_next); - - while (n_left_from >= 4 && n_left_to_next >= 2) - { - u32 bi0, bi1; - vlib_buffer_t *b0, *b1; - u32 next0, next1; - - next0 = next1 = VXLAN_GPE_ENCAP_IOAM_V4_NEXT_IP4_LOOKUP; - - /* Prefetch next iteration. */ - { - vlib_buffer_t *p2, *p3; - - p2 = vlib_get_buffer (vm, from[2]); - p3 = vlib_get_buffer (vm, from[3]); - - vlib_prefetch_buffer_header (p2, LOAD); - vlib_prefetch_buffer_header (p3, LOAD); - - CLIB_PREFETCH (p2->data, 2 * CLIB_CACHE_LINE_BYTES, LOAD); - CLIB_PREFETCH (p3->data, 2 * CLIB_CACHE_LINE_BYTES, LOAD); - } - - bi0 = from[0]; - bi1 = from[1]; - to_next[0] = bi0; - to_next[1] = bi1; - from += 2; - to_next += 2; - n_left_to_next -= 2; - n_left_from -= 2; - - b0 = vlib_get_buffer (vm, bi0); - b1 = vlib_get_buffer (vm, bi1); - - vxlan_gpe_encap_ioam_v4_two_inline (vm, node, ngm, b0, b1, - &next0, &next1); - - - vlib_validate_buffer_enqueue_x2 (vm, node, next_index, to_next, - n_left_to_next, bi0, bi1, next0, - next1); - } - - while (n_left_from > 0 && n_left_to_next > 0) - { - u32 bi0; - vlib_buffer_t *b0; - u32 next0 = VXLAN_GPE_ENCAP_IOAM_V4_NEXT_IP4_LOOKUP; - - bi0 = from[0]; - to_next[0] = bi0; - from += 1; - to_next += 1; - n_left_from -= 1; - n_left_to_next -= 1; - - b0 = vlib_get_buffer (vm, bi0); - - vxlan_gpe_encap_decap_ioam_v4_one_inline (vm, node, b0, - &next0, - VXLAN_GPE_ENCAP_IOAM_V4_NEXT_DROP, - 0 /* use_adj */ ); - - if (PREDICT_FALSE (b0->flags & VLIB_BUFFER_IS_TRACED)) - { - vxlan_gpe_ioam_v4_trace_t *tr = vlib_add_trace (vm, node, b0, - sizeof (*tr)); - } - vlib_validate_buffer_enqueue_x1 (vm, node, next_index, to_next, - n_left_to_next, bi0, next0); - } - - vlib_put_next_frame (vm, node, next_index, n_left_to_next); - } - - return from_frame->n_vectors; -} - - - -/* *INDENT-OFF* */ -VLIB_REGISTER_NODE (vxlan_gpe_encap_ioam_v4_node) = { - .function = vxlan_gpe_encap_ioam_v4, - .name = "vxlan-gpe-encap-ioam-v4", - .vector_size = sizeof (u32), - .format_trace = format_vxlan_gpe_ioam_v4_trace, - .type = VLIB_NODE_TYPE_INTERNAL, - - .n_errors = ARRAY_LEN(vxlan_gpe_encap_ioam_v4_error_strings), - .error_strings = vxlan_gpe_encap_ioam_v4_error_strings, - - .n_next_nodes = VXLAN_GPE_ENCAP_IOAM_V4_N_NEXT, - - .next_nodes = { - [VXLAN_GPE_ENCAP_IOAM_V4_NEXT_IP4_LOOKUP] = "ip4-lookup", - [VXLAN_GPE_ENCAP_IOAM_V4_NEXT_DROP] = "error-drop", - }, -}; -/* *INDENT-ON* */ - - -/* - * fd.io coding-style-patch-verification: ON - * - * Local Variables: - * eval: (c-set-style "gnu") - * End: - */ diff --git a/plugins/ioam-plugin/ioam/lib-vxlan-gpe/ioam_pop.c b/plugins/ioam-plugin/ioam/lib-vxlan-gpe/ioam_pop.c deleted file mode 100644 index 55c33b14..00000000 --- a/plugins/ioam-plugin/ioam/lib-vxlan-gpe/ioam_pop.c +++ /dev/null @@ -1,353 +0,0 @@ -/* - * Copyright (c) 2015 Cisco and/or its affiliates. - * Licensed under the Apache License, Version 2.0 (the "License"); - * you may not use this file except in compliance with the License. - * You may obtain a copy of the License at: - * - * http://www.apache.org/licenses/LICENSE-2.0 - * - * Unless required by applicable law or agreed to in writing, software - * distributed under the License is distributed on an "AS IS" BASIS, - * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. - * See the License for the specific language governing permissions and - * limitations under the License. - */ -#include -#include -#include -#include -#include -#include -#include - -/* Statistics (not really errors) */ -#define foreach_vxlan_gpe_pop_ioam_v4_error \ -_(POPPED, "good packets popped") - -static char *vxlan_gpe_pop_ioam_v4_error_strings[] = { -#define _(sym,string) string, - foreach_vxlan_gpe_pop_ioam_v4_error -#undef _ -}; - -typedef enum -{ -#define _(sym,str) VXLAN_GPE_POP_IOAM_V4_ERROR_##sym, - foreach_vxlan_gpe_pop_ioam_v4_error -#undef _ - VXLAN_GPE_POP_IOAM_V4_N_ERROR, -} vxlan_gpe_pop_ioam_v4_error_t; - -typedef struct -{ - ioam_trace_t fmt_trace; -} vxlan_gpe_pop_ioam_v4_trace_t; - - -u8 * -format_vxlan_gpe_pop_ioam_v4_trace (u8 * s, va_list * args) -{ - CLIB_UNUSED (vlib_main_t * vm) = va_arg (*args, vlib_main_t *); - CLIB_UNUSED (vlib_node_t * node) = va_arg (*args, vlib_node_t *); - vxlan_gpe_pop_ioam_v4_trace_t *t1 - = va_arg (*args, vxlan_gpe_pop_ioam_v4_trace_t *); - ioam_trace_t *t = &(t1->fmt_trace); - vxlan_gpe_ioam_option_t *fmt_trace0; - vxlan_gpe_ioam_option_t *opt0, *limit0; - vxlan_gpe_ioam_main_t *hm = &vxlan_gpe_ioam_main; - - u8 type0; - - fmt_trace0 = (vxlan_gpe_ioam_option_t *) t->option_data; - - s = format (s, "VXLAN_GPE_IOAM_POP: next_index %d len %d traced %d", - t->next_index, fmt_trace0->length, t->trace_len); - - opt0 = (vxlan_gpe_ioam_option_t *) (fmt_trace0 + 1); - limit0 = (vxlan_gpe_ioam_option_t *) ((u8 *) fmt_trace0) + t->trace_len; - - while (opt0 < limit0) - { - type0 = opt0->type; - switch (type0) - { - case 0: /* Pad, just stop */ - opt0 = (vxlan_gpe_ioam_option_t *) ((u8 *) opt0) + 1; - break; - - default: - if (hm->trace[type0]) - { - s = (*hm->trace[type0]) (s, opt0); - } - else - { - s = - format (s, "\n unrecognized option %d length %d", type0, - opt0->length); - } - opt0 = - (vxlan_gpe_ioam_option_t *) (((u8 *) opt0) + opt0->length + - sizeof (vxlan_gpe_ioam_option_t)); - break; - } - } - - return s; -} - -always_inline void -vxlan_gpe_ioam_pop_v4 (vlib_main_t * vm, vlib_node_runtime_t * node, - vlib_buffer_t * b0) -{ - ip4_header_t *ip0; - udp_header_t *udp_hdr0; - vxlan_gpe_header_t *gpe_hdr0; - vxlan_gpe_ioam_hdr_t *gpe_ioam0; - - ip0 = vlib_buffer_get_current (b0); - - udp_hdr0 = (udp_header_t *) (ip0 + 1); - gpe_hdr0 = (vxlan_gpe_header_t *) (udp_hdr0 + 1); - gpe_ioam0 = (vxlan_gpe_ioam_hdr_t *) (gpe_hdr0 + 1); - - /* Pop the iOAM data */ - vlib_buffer_advance (b0, - (word) (sizeof (udp_header_t) + - sizeof (ip4_header_t) + - sizeof (vxlan_gpe_header_t) + - gpe_ioam0->length)); - - return; -} - - - -always_inline void -vxlan_gpe_pop_ioam_v4_one_inline (vlib_main_t * vm, - vlib_node_runtime_t * node, - vxlan_gpe_main_t * ngm, - vlib_buffer_t * b0, u32 * next0) -{ - CLIB_UNUSED (ip4_header_t * ip0); - CLIB_UNUSED (udp_header_t * udp_hdr0); - CLIB_UNUSED (vxlan_gpe_header_t * gpe_hdr0); - CLIB_UNUSED (vxlan_gpe_ioam_hdr_t * gpe_ioam0); - CLIB_UNUSED (vxlan_gpe_ioam_option_t * opt0); - CLIB_UNUSED (vxlan_gpe_ioam_option_t * limit0); - vxlan_gpe_ioam_main_t *hm = &vxlan_gpe_ioam_main; - - - /* Pop the iOAM header */ - ip0 = vlib_buffer_get_current (b0); - udp_hdr0 = (udp_header_t *) (ip0 + 1); - gpe_hdr0 = (vxlan_gpe_header_t *) (udp_hdr0 + 1); - gpe_ioam0 = (vxlan_gpe_ioam_hdr_t *) (gpe_hdr0 + 1); - opt0 = (vxlan_gpe_ioam_option_t *) (gpe_ioam0 + 1); - limit0 = (vxlan_gpe_ioam_option_t *) ((u8 *) gpe_ioam0 + gpe_ioam0->length); - - /* - * Basic validity checks - */ - if (gpe_ioam0->length > clib_net_to_host_u16 (ip0->length)) - { - next0[0] = VXLAN_GPE_INPUT_NEXT_DROP; - goto trace00; - } - - /* Scan the set of h-b-h options, process ones that we understand */ - while (opt0 < limit0) - { - u8 type0; - type0 = opt0->type; - switch (type0) - { - case 0: /* Pad1 */ - opt0 = (vxlan_gpe_ioam_option_t *) ((u8 *) opt0) + 1; - continue; - case 1: /* PadN */ - break; - default: - if (hm->pop_options[type0]) - { - if ((*hm->pop_options[type0]) (ip0, opt0) < 0) - { - next0[0] = VXLAN_GPE_INPUT_NEXT_DROP; - goto trace00; - } - } - break; - } - opt0 = - (vxlan_gpe_ioam_option_t *) (((u8 *) opt0) + opt0->length + - sizeof (vxlan_gpe_ioam_hdr_t)); - } - - - next0[0] = - (gpe_ioam0->protocol < VXLAN_GPE_PROTOCOL_MAX) ? - ngm-> - decap_next_node_list[gpe_ioam0->protocol] : VXLAN_GPE_INPUT_NEXT_DROP; - -trace00: - if (PREDICT_FALSE (b0->flags & VLIB_BUFFER_IS_TRACED)) - { - vxlan_gpe_pop_ioam_v4_trace_t *t = - vlib_add_trace (vm, node, b0, sizeof (*t)); - u32 trace_len = gpe_ioam0->length; - t->fmt_trace.next_index = next0[0]; - /* Capture the h-b-h option verbatim */ - trace_len = - trace_len < - ARRAY_LEN (t->fmt_trace. - option_data) ? trace_len : ARRAY_LEN (t->fmt_trace. - option_data); - t->fmt_trace.trace_len = trace_len; - clib_memcpy (&(t->fmt_trace.option_data), gpe_ioam0, trace_len); - } - - /* Remove the iOAM header inside the VxLAN-GPE header */ - vxlan_gpe_ioam_pop_v4 (vm, node, b0); - return; -} - -always_inline void -vxlan_gpe_pop_ioam_v4_two_inline (vlib_main_t * vm, - vlib_node_runtime_t * node, - vxlan_gpe_main_t * ngm, - vlib_buffer_t * b0, vlib_buffer_t * b1, - u32 * next0, u32 * next1) -{ - - vxlan_gpe_pop_ioam_v4_one_inline (vm, node, ngm, b0, next0); - vxlan_gpe_pop_ioam_v4_one_inline (vm, node, ngm, b1, next1); -} - - - -static uword -vxlan_gpe_pop_ioam (vlib_main_t * vm, - vlib_node_runtime_t * node, - vlib_frame_t * from_frame, u8 is_ipv6) -{ - u32 n_left_from, next_index, *from, *to_next; - vxlan_gpe_main_t *ngm = &vxlan_gpe_main; - - from = vlib_frame_vector_args (from_frame); - n_left_from = from_frame->n_vectors; - - next_index = node->cached_next_index; - - while (n_left_from > 0) - { - u32 n_left_to_next; - - vlib_get_next_frame (vm, node, next_index, to_next, n_left_to_next); - - while (n_left_from >= 4 && n_left_to_next >= 2) - { - u32 bi0, bi1; - vlib_buffer_t *b0, *b1; - u32 next0, next1; - - /* Prefetch next iteration. */ - { - vlib_buffer_t *p2, *p3; - - p2 = vlib_get_buffer (vm, from[2]); - p3 = vlib_get_buffer (vm, from[3]); - - vlib_prefetch_buffer_header (p2, LOAD); - vlib_prefetch_buffer_header (p3, LOAD); - - CLIB_PREFETCH (p2->data, 2 * CLIB_CACHE_LINE_BYTES, LOAD); - CLIB_PREFETCH (p3->data, 2 * CLIB_CACHE_LINE_BYTES, LOAD); - } - - bi0 = from[0]; - bi1 = from[1]; - to_next[0] = bi0; - to_next[1] = bi1; - from += 2; - to_next += 2; - n_left_to_next -= 2; - n_left_from -= 2; - - b0 = vlib_get_buffer (vm, bi0); - b1 = vlib_get_buffer (vm, bi1); - - vxlan_gpe_pop_ioam_v4_two_inline (vm, node, ngm, b0, b1, &next0, - &next1); - - - vlib_validate_buffer_enqueue_x2 (vm, node, next_index, to_next, - n_left_to_next, bi0, bi1, next0, - next1); - } - - while (n_left_from > 0 && n_left_to_next > 0) - { - u32 bi0; - vlib_buffer_t *b0; - u32 next0; - - bi0 = from[0]; - to_next[0] = bi0; - from += 1; - to_next += 1; - n_left_from -= 1; - n_left_to_next -= 1; - - b0 = vlib_get_buffer (vm, bi0); - - vxlan_gpe_pop_ioam_v4_one_inline (vm, node, ngm, b0, &next0); - - - vlib_validate_buffer_enqueue_x1 (vm, node, next_index, to_next, - n_left_to_next, bi0, next0); - } - - vlib_put_next_frame (vm, node, next_index, n_left_to_next); - } - - return from_frame->n_vectors; -} - - -static uword -vxlan_gpe_pop_ioam_v4 (vlib_main_t * vm, - vlib_node_runtime_t * node, vlib_frame_t * from_frame) -{ - return vxlan_gpe_pop_ioam (vm, node, from_frame, 0); -} - -/* *INDENT-OFF* */ -VLIB_REGISTER_NODE (vxlan_gpe_pop_ioam_v4_node) = { - .function = vxlan_gpe_pop_ioam_v4, - .name = "vxlan-gpe-pop-ioam-v4", - .vector_size = sizeof (u32), - .format_trace = format_vxlan_gpe_pop_ioam_v4_trace, - .type = VLIB_NODE_TYPE_INTERNAL, - - .n_errors = ARRAY_LEN(vxlan_gpe_pop_ioam_v4_error_strings), - .error_strings = vxlan_gpe_pop_ioam_v4_error_strings, - - .n_next_nodes = VXLAN_GPE_INPUT_N_NEXT, - - .next_nodes = { -#define _(s,n) [VXLAN_GPE_INPUT_NEXT_##s] = n, - foreach_vxlan_gpe_input_next -#undef _ - }, -}; -/* *INDENT-ON* */ - - - -/* - * fd.io coding-style-patch-verification: ON - * - * Local Variables: - * eval: (c-set-style "gnu") - * End: - */ diff --git a/plugins/ioam-plugin/ioam/lib-vxlan-gpe/ioam_transit.c b/plugins/ioam-plugin/ioam/lib-vxlan-gpe/ioam_transit.c deleted file mode 100644 index b42c357c..00000000 --- a/plugins/ioam-plugin/ioam/lib-vxlan-gpe/ioam_transit.c +++ /dev/null @@ -1,188 +0,0 @@ - /* - * Copyright (c) 2015 Cisco and/or its affiliates. - * Licensed under the Apache License, Version 2.0 (the "License"); - * you may not use this file except in compliance with the License. - * You may obtain a copy of the License at: - * - * http://www.apache.org/licenses/LICENSE-2.0 - * - * Unless required by applicable law or agreed to in writing, software - * distributed under the License is distributed on an "AS IS" BASIS, - * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. - * See the License for the specific language governing permissions and - * limitations under the License. - */ -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include - -/* Statistics (not really errors) */ -#define foreach_vxlan_gpe_transit_ioam_error \ -_(ENCAPSULATED, "good packets encapsulated") - -static char *vxlan_gpe_transit_ioam_error_strings[] = { -#define _(sym,string) string, - foreach_vxlan_gpe_transit_ioam_error -#undef _ -}; - -typedef enum -{ -#define _(sym,str) VXLAN_GPE_TRANSIT_IOAM_ERROR_##sym, - foreach_vxlan_gpe_transit_ioam_error -#undef _ - VXLAN_GPE_TRANSIT_IOAM_N_ERROR, -} vxlan_gpe_transit_ioam_error_t; - -typedef enum -{ - VXLAN_GPE_TRANSIT_IOAM_NEXT_OUTPUT, - VXLAN_GPE_TRANSIT_IOAM_NEXT_DROP, - VXLAN_GPE_TRANSIT_IOAM_N_NEXT -} vxlan_gpe_transit_ioam_next_t; - - -/* *INDENT-OFF* */ -VNET_FEATURE_INIT (vxlan_gpe_transit_ioam, static) = -{ - .arc_name = "ip4-output", - .node_name = "vxlan-gpe-transit-ioam", - .runs_before = VNET_FEATURES ("interface-output"), -}; -/* *INDENT-ON* */ - -static uword -vxlan_gpe_transit_ioam (vlib_main_t * vm, - vlib_node_runtime_t * node, vlib_frame_t * from_frame) -{ - u32 n_left_from, next_index, *from, *to_next; - - from = vlib_frame_vector_args (from_frame); - n_left_from = from_frame->n_vectors; - - next_index = node->cached_next_index; - - while (n_left_from > 0) - { - u32 n_left_to_next; - - vlib_get_next_frame (vm, node, next_index, to_next, n_left_to_next); - - - while (n_left_from > 0 && n_left_to_next > 0) - { - u32 bi0; - vlib_buffer_t *b0; - u32 next0 = VXLAN_GPE_TRANSIT_IOAM_NEXT_OUTPUT; - - bi0 = from[0]; - to_next[0] = bi0; - from += 1; - to_next += 1; - n_left_from -= 1; - n_left_to_next -= 1; - ip4_header_t *ip0; - u32 iph_offset = 0; - - b0 = vlib_get_buffer (vm, bi0); - iph_offset = vnet_buffer (b0)->ip.save_rewrite_length; - ip0 = (ip4_header_t *) ((u8 *) vlib_buffer_get_current (b0) - + iph_offset); - - /* just forward non ipv4 packets */ - if (PREDICT_FALSE - ((ip0->ip_version_and_header_length & 0xF0) == 0x40)) - { - /* ipv4 packets */ - udp_header_t *udp_hdr0 = (udp_header_t *) (ip0 + 1); - if (PREDICT_FALSE - ((ip0->protocol == IP_PROTOCOL_UDP) && - (clib_net_to_host_u16 (udp_hdr0->dst_port) == - UDP_DST_PORT_vxlan_gpe))) - { - - /* Check the iOAM header */ - vxlan_gpe_header_t *gpe_hdr0 = - (vxlan_gpe_header_t *) (udp_hdr0 + 1); - - if (PREDICT_FALSE - (gpe_hdr0->protocol == VXLAN_GPE_PROTOCOL_IOAM)) - { - uword *t = NULL; - vxlan_gpe_ioam_main_t *hm = &vxlan_gpe_ioam_main; - fib_prefix_t key4; - memset (&key4, 0, sizeof (key4)); - key4.fp_proto = FIB_PROTOCOL_IP4; - key4.fp_addr.ip4.as_u32 = ip0->dst_address.as_u32; - t = hash_get_mem (hm->dst_by_ip4, &key4); - if (t) - { - - - vlib_buffer_advance (b0, - (word) (sizeof - (ethernet_header_t))); - vxlan_gpe_encap_decap_ioam_v4_one_inline (vm, node, - b0, - &next0, - VXLAN_GPE_TRANSIT_IOAM_NEXT_DROP, - 1 - /* use_adj */ - ); - vlib_buffer_advance (b0, - -(word) (sizeof - (ethernet_header_t))); - } - } - } - } - - vlib_validate_buffer_enqueue_x1 (vm, node, next_index, to_next, - n_left_to_next, bi0, next0); - } - - vlib_put_next_frame (vm, node, next_index, n_left_to_next); - } - - return from_frame->n_vectors; -} - -/* *INDENT-OFF* */ -VLIB_REGISTER_NODE (vxlan_gpe_transit_ioam_node) = { - .function = vxlan_gpe_transit_ioam, - .name = "vxlan-gpe-transit-ioam", - .vector_size = sizeof (u32), - .format_trace = format_vxlan_gpe_ioam_v4_trace, - .type = VLIB_NODE_TYPE_INTERNAL, - - .n_errors = ARRAY_LEN(vxlan_gpe_transit_ioam_error_strings), - .error_strings = vxlan_gpe_transit_ioam_error_strings, - - .n_next_nodes = VXLAN_GPE_TRANSIT_IOAM_N_NEXT, - - .next_nodes = { - [VXLAN_GPE_TRANSIT_IOAM_NEXT_OUTPUT] = "interface-output", - [VXLAN_GPE_TRANSIT_IOAM_NEXT_DROP] = "error-drop", - }, - -}; -/* *INDENT-ON* */ - - -/* - * fd.io coding-style-patch-verification: ON - * - * Local Variables: - * eval: (c-set-style "gnu") - * End: - */ diff --git a/plugins/ioam-plugin/ioam/lib-vxlan-gpe/vxlan_gpe.api b/plugins/ioam-plugin/ioam/lib-vxlan-gpe/vxlan_gpe.api deleted file mode 100644 index 056529a4..00000000 --- a/plugins/ioam-plugin/ioam/lib-vxlan-gpe/vxlan_gpe.api +++ /dev/null @@ -1,181 +0,0 @@ -/* Hey Emacs use -*- mode: C -*- */ -/* - * Copyright (c) 2016 Cisco and/or its affiliates. - * Licensed under the Apache License, Version 2.0 (the "License"); - * you may not use this file except in compliance with the License. - * You may obtain a copy of the License at: - * - * http://www.apache.org/licenses/LICENSE-2.0 - * - * Unless required by applicable law or agreed to in writing, software - * distributed under the License is distributed on an "AS IS" BASIS, - * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. - * See the License for the specific language governing permissions and - * limitations under the License. - */ - - -/** \brief iOAM Over VxLAN-GPE - Set iOAM transport for VxLAN-GPE - @param client_index - opaque cookie to identify the sender - @param context - sender context, to match reply w/ request - @param id - profile id - @param trace_ppc - Trace PPC (none/encap/decap) - @param pow_enable - Proof of Work enabled or not flag - @param trace_enable - iOAM Trace enabled or not flag - -*/ -define vxlan_gpe_ioam_enable { - u32 client_index; - u32 context; - u16 id; - u8 trace_ppc; - u8 pow_enable; - u8 trace_enable; -}; - -/** \brief iOAM Over VxLAN-GPE - Set iOAM transport for VXLAN-GPE reply - @param context - sender context, to match reply w/ request - @param retval - return value for request -*/ -define vxlan_gpe_ioam_enable_reply { - u32 context; - i32 retval; -}; - - -/** \brief iOAM for VxLAN-GPE disable - @param client_index - opaque cookie to identify the sender - @param context - sender context, to match reply w/ request - @param id - profile id -*/ -define vxlan_gpe_ioam_disable -{ - u32 client_index; - u32 context; - u16 id; -}; - -/** \brief vxlan_gpe_ioam disable response - @param context - sender context, to match reply w/ request - @param retval - return value for request -*/ -define vxlan_gpe_ioam_disable_reply -{ - u32 context; - i32 retval; -}; - -/** \brief Enable iOAM for a VNI (VXLAN-GPE) - @param client_index - opaque cookie to identify the sender - @param context - sender context, to match reply w/ request - @param vni - VXLAN-GPE VNI - @param local - IPv4/6 Address of the local VTEP - @param remote - IPv4/6 Address of the remote VTEP - -*/ -define vxlan_gpe_ioam_vni_enable { - u32 client_index; - u32 context; - u32 vni; - u8 local[16]; - u8 remote[16]; - u8 is_ipv6; -}; - -/** \brief Reply to enable iOAM for a VNI (VXLAN-GPE) - @param client_index - opaque cookie to identify the sender - @param context - sender context, to match reply w/ request - @param retval - return value for request - -*/ -define vxlan_gpe_ioam_vni_enable_reply { - u32 client_index; - u32 context; - i32 retval; -}; - -/** \brief Disable iOAM for a VNI (VXLAN-GPE) - @param client_index - opaque cookie to identify the sender - @param context - sender context, to match reply w/ request - @param vni - VXLAN-GPE VNI - @param local - IPv4/6 Address of the local VTEP - @param remote - IPv4/6 Address of the remote VTEP - -*/ -define vxlan_gpe_ioam_vni_disable { - u32 client_index; - u32 context; - u32 vni; - u8 local[16]; - u8 remote[16]; - u8 is_ipv6; -}; - -/** \brief Reply to disable iOAM for a VNI (VXLAN-GPE) - @param client_index - opaque cookie to identify the sender - @param context - sender context, to match reply w/ request - @param retval - return value for request - -*/ -define vxlan_gpe_ioam_vni_disable_reply { - u32 client_index; - u32 context; - i32 retval; -}; - - -/** \brief Enable iOAM for a VXLAN-GPE transit - @param client_index - opaque cookie to identify the sender - @param context - sender context, to match reply w/ request - @param dst_addr - IPv4/6 Address of the local VTEP - @param outer_fib_index- FIB index - -*/ -define vxlan_gpe_ioam_transit_enable { - u32 client_index; - u32 context; - u32 outer_fib_index; - u8 dst_addr[16]; - u8 is_ipv6; -}; - -/** \brief Reply to enable iOAM for VXLAN-GPE transit - @param client_index - opaque cookie to identify the sender - @param context - sender context, to match reply w/ request - @param retval - return value for request - -*/ -define vxlan_gpe_ioam_transit_enable_reply { - u32 client_index; - u32 context; - i32 retval; -}; - -/** \brief Disable iOAM for VXLAN-GPE transit - @param client_index - opaque cookie to identify the sender - @param context - sender context, to match reply w/ request - @param dst_addr - IPv4/6 Address of the local VTEP - @param outer_fib_index- FIB index - -*/ -define vxlan_gpe_ioam_transit_disable { - u32 client_index; - u32 context; - u32 outer_fib_index; - u8 dst_addr[16]; - u8 is_ipv6; -}; - -/** \brief Reply to disable iOAM for VXLAN-GPE transit - @param client_index - opaque cookie to identify the sender - @param context - sender context, to match reply w/ request - @param retval - return value for request - -*/ -define vxlan_gpe_ioam_transit_disable_reply { - u32 client_index; - u32 context; - i32 retval; -}; - - diff --git a/plugins/ioam-plugin/ioam/lib-vxlan-gpe/vxlan_gpe_all_api_h.h b/plugins/ioam-plugin/ioam/lib-vxlan-gpe/vxlan_gpe_all_api_h.h deleted file mode 100644 index bbf2c101..00000000 --- a/plugins/ioam-plugin/ioam/lib-vxlan-gpe/vxlan_gpe_all_api_h.h +++ /dev/null @@ -1,16 +0,0 @@ -/* - * Copyright (c) 2016 Cisco and/or its affiliates. - * Licensed under the Apache License, Version 2.0 (the "License"); - * you may not use this file except in compliance with the License. - * You may obtain a copy of the License at: - * - * http://www.apache.org/licenses/LICENSE-2.0 - * - * Unless required by applicable law or agreed to in writing, software - * distributed under the License is distributed on an "AS IS" BASIS, - * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. - * See the License for the specific language governing permissions and - * limitations under the License. - */ -/* Include the generated file, see BUILT_SOURCES in Makefile.am */ -#include diff --git a/plugins/ioam-plugin/ioam/lib-vxlan-gpe/vxlan_gpe_api.c b/plugins/ioam-plugin/ioam/lib-vxlan-gpe/vxlan_gpe_api.c deleted file mode 100644 index 68752365..00000000 --- a/plugins/ioam-plugin/ioam/lib-vxlan-gpe/vxlan_gpe_api.c +++ /dev/null @@ -1,378 +0,0 @@ -/* - * Copyright (c) 2016 Cisco and/or its affiliates. - * Licensed under the Apache License, Version 2.0 (the "License"); - * you may not use this file except in compliance with the License. - * You may obtain a copy of the License at: - * - * http://www.apache.org/licenses/LICENSE-2.0 - * - * Unless required by applicable law or agreed to in writing, software - * distributed under the License is distributed on an "AS IS" BASIS, - * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. - * See the License for the specific language governing permissions and - * limitations under the License. - */ -/* - *------------------------------------------------------------------ - * vxlan_gpe_api.c - iOAM VxLAN-GPE related APIs to create - * and maintain profiles - *------------------------------------------------------------------ - */ - -#include -#include -#include - -#include -#include -#include - -/* define message IDs */ -#include - -/* define message structures */ -#define vl_typedefs -#include -#undef vl_typedefs - -/* define generated endian-swappers */ -#define vl_endianfun -#include -#undef vl_endianfun - -/* instantiate all the print functions we know about */ -#define vl_print(handle, ...) vlib_cli_output (handle, __VA_ARGS__) -#define vl_printfun -#include -#undef vl_printfun - -/* Get the API version number */ -#define vl_api_version(n,v) static u32 api_version=(v); -#include -#undef vl_api_version - -/* - * A handy macro to set up a message reply. - * Assumes that the following variables are available: - * mp - pointer to request message - * rmp - pointer to reply message type - * rv - return value - */ - -#define VXLAN_GPE_REPLY_MACRO(t) \ -do { \ - unix_shared_memory_queue_t * q = \ - vl_api_client_index_to_input_queue (mp->client_index); \ - if (!q) \ - return; \ - \ - rmp = vl_msg_api_alloc (sizeof (*rmp)); \ - rmp->_vl_msg_id = ntohs((t)+sm->msg_id_base); \ - rmp->context = mp->context; \ - rmp->retval = ntohl(rv); \ - \ - vl_msg_api_send_shmem (q, (u8 *)&rmp); \ -} while(0); - -/* *INDENT-OFF* */ -#define VXLAN_GPE_REPLY_MACRO2(t, body) \ -do { \ - unix_shared_memory_queue_t * q; \ - rv = vl_msg_api_pd_handler (mp, rv); \ - q = vl_api_client_index_to_input_queue (mp->client_index); \ - if (!q) \ - return; \ - \ - rmp = vl_msg_api_alloc (sizeof (*rmp)); \ - rmp->_vl_msg_id = ntohs((t)); \ - rmp->context = mp->context; \ - rmp->retval = ntohl(rv); \ - do {body;} while (0); \ - vl_msg_api_send_shmem (q, (u8 *)&rmp); \ -} while(0); -/* *INDENT-ON* */ - -/* List of message types that this plugin understands */ - -#define foreach_vxlan_gpe_plugin_api_msg \ -_(VXLAN_GPE_IOAM_ENABLE, vxlan_gpe_ioam_enable) \ -_(VXLAN_GPE_IOAM_DISABLE, vxlan_gpe_ioam_disable) \ -_(VXLAN_GPE_IOAM_VNI_ENABLE, vxlan_gpe_ioam_vni_enable) \ -_(VXLAN_GPE_IOAM_VNI_DISABLE, vxlan_gpe_ioam_vni_disable) \ -_(VXLAN_GPE_IOAM_TRANSIT_ENABLE, vxlan_gpe_ioam_transit_enable) \ -_(VXLAN_GPE_IOAM_TRANSIT_DISABLE, vxlan_gpe_ioam_transit_disable) \ - - -static void vl_api_vxlan_gpe_ioam_enable_t_handler - (vl_api_vxlan_gpe_ioam_enable_t * mp) -{ - int rv = 0; - vl_api_vxlan_gpe_ioam_enable_reply_t *rmp; - clib_error_t *error; - vxlan_gpe_ioam_main_t *sm = &vxlan_gpe_ioam_main; - - /* Ignoring the profile id as currently a single profile - * is supported */ - error = - vxlan_gpe_ioam_enable (mp->trace_enable, mp->pow_enable, mp->trace_ppc); - if (error) - { - clib_error_report (error); - rv = clib_error_get_code (error); - } - - VXLAN_GPE_REPLY_MACRO (VL_API_VXLAN_GPE_IOAM_ENABLE_REPLY); -} - -static void vl_api_vxlan_gpe_ioam_disable_t_handler - (vl_api_vxlan_gpe_ioam_disable_t * mp) -{ - int rv = 0; - vl_api_vxlan_gpe_ioam_disable_reply_t *rmp; - clib_error_t *error; - vxlan_gpe_ioam_main_t *sm = &vxlan_gpe_ioam_main; - - /* Ignoring the profile id as currently a single profile - * is supported */ - error = vxlan_gpe_ioam_disable (0, 0, 0); - if (error) - { - clib_error_report (error); - rv = clib_error_get_code (error); - } - - VXLAN_GPE_REPLY_MACRO (VL_API_VXLAN_GPE_IOAM_DISABLE_REPLY); -} - -static void vl_api_vxlan_gpe_ioam_vni_enable_t_handler - (vl_api_vxlan_gpe_ioam_vni_enable_t * mp) -{ - int rv = 0; - vl_api_vxlan_gpe_ioam_vni_enable_reply_t *rmp; - clib_error_t *error; - vxlan_gpe_ioam_main_t *sm = &vxlan_gpe_ioam_main; - vxlan4_gpe_tunnel_key_t key4; - uword *p = NULL; - vxlan_gpe_main_t *gm = &vxlan_gpe_main; - vxlan_gpe_tunnel_t *t = 0; - vxlan_gpe_ioam_main_t *hm = &vxlan_gpe_ioam_main; - u32 vni; - - - if (!mp->is_ipv6) - { - clib_memcpy (&key4.local, &mp->local, sizeof (key4.local)); - clib_memcpy (&key4.remote, &mp->remote, sizeof (key4.remote)); - vni = clib_net_to_host_u32 (mp->vni); - key4.vni = clib_host_to_net_u32 (vni << 8); - key4.pad = 0; - - p = hash_get_mem (gm->vxlan4_gpe_tunnel_by_key, &key4); - } - else - { - return; - } - - if (!p) - return; - - t = pool_elt_at_index (gm->tunnels, p[0]); - - error = vxlan_gpe_ioam_set (t, hm->has_trace_option, - hm->has_pot_option, - hm->has_ppc_option, mp->is_ipv6); - - - if (error) - { - clib_error_report (error); - rv = clib_error_get_code (error); - } - - VXLAN_GPE_REPLY_MACRO (VL_API_VXLAN_GPE_IOAM_VNI_ENABLE_REPLY); -} - - -static void vl_api_vxlan_gpe_ioam_vni_disable_t_handler - (vl_api_vxlan_gpe_ioam_vni_disable_t * mp) -{ - int rv = 0; - vl_api_vxlan_gpe_ioam_vni_enable_reply_t *rmp; - clib_error_t *error; - vxlan_gpe_ioam_main_t *sm = &vxlan_gpe_ioam_main; - vxlan4_gpe_tunnel_key_t key4; - uword *p = NULL; - vxlan_gpe_main_t *gm = &vxlan_gpe_main; - vxlan_gpe_tunnel_t *t = 0; - u32 vni; - - - if (!mp->is_ipv6) - { - clib_memcpy (&key4.local, &mp->local, sizeof (key4.local)); - clib_memcpy (&key4.remote, &mp->remote, sizeof (key4.remote)); - vni = clib_net_to_host_u32 (mp->vni); - key4.vni = clib_host_to_net_u32 (vni << 8); - key4.pad = 0; - - p = hash_get_mem (gm->vxlan4_gpe_tunnel_by_key, &key4); - } - else - { - return; - } - - if (!p) - return; - - t = pool_elt_at_index (gm->tunnels, p[0]); - - error = vxlan_gpe_ioam_clear (t, 0, 0, 0, 0); - - - if (error) - { - clib_error_report (error); - rv = clib_error_get_code (error); - } - - - VXLAN_GPE_REPLY_MACRO (VL_API_VXLAN_GPE_IOAM_VNI_DISABLE_REPLY); -} - -static void vl_api_vxlan_gpe_ioam_transit_enable_t_handler - (vl_api_vxlan_gpe_ioam_transit_enable_t * mp) -{ - int rv = 0; - vl_api_vxlan_gpe_ioam_transit_enable_reply_t *rmp; - vxlan_gpe_ioam_main_t *sm = &vxlan_gpe_ioam_main; - ip46_address_t dst_addr; - - memset (&dst_addr.ip4, 0, sizeof (dst_addr.ip4)); - if (!mp->is_ipv6) - { - clib_memcpy (&dst_addr.ip4, &mp->dst_addr, sizeof (dst_addr.ip4)); - } - rv = vxlan_gpe_enable_disable_ioam_for_dest (sm->vlib_main, - dst_addr, - ntohl (mp->outer_fib_index), - mp->is_ipv6 ? 0 : 1, - 1 /* is_add */ ); - - VXLAN_GPE_REPLY_MACRO (VL_API_VXLAN_GPE_IOAM_TRANSIT_ENABLE_REPLY); -} - -static void vl_api_vxlan_gpe_ioam_transit_disable_t_handler - (vl_api_vxlan_gpe_ioam_transit_disable_t * mp) -{ - int rv = 0; - vl_api_vxlan_gpe_ioam_transit_disable_reply_t *rmp; - vxlan_gpe_ioam_main_t *sm = &vxlan_gpe_ioam_main; - ip46_address_t dst_addr; - - memset (&dst_addr.ip4, 0, sizeof (dst_addr.ip4)); - if (!mp->is_ipv6) - { - clib_memcpy (&dst_addr.ip4, &mp->dst_addr, sizeof (dst_addr.ip4)); - } - - rv = vxlan_gpe_ioam_disable_for_dest (sm->vlib_main, - dst_addr, - ntohl (mp->outer_fib_index), - mp->is_ipv6 ? 0 : 1); - VXLAN_GPE_REPLY_MACRO (VL_API_VXLAN_GPE_IOAM_TRANSIT_DISABLE_REPLY); -} - - -/* - * This routine exists to convince the vlib plugin framework that - * we haven't accidentally copied a random .dll into the plugin directory. - * - * Also collects global variable pointers passed from the vpp engine - */ - -clib_error_t * -vlib_plugin_register (vlib_main_t * vm, vnet_plugin_handoff_t * h, - int from_early_init) -{ - vxlan_gpe_ioam_main_t *sm = &vxlan_gpe_ioam_main; - clib_error_t *error = 0; - - sm->vlib_main = vm; - sm->vnet_main = h->vnet_main; - sm->unix_time_0 = (u32) time (0); /* Store starting time */ - sm->vlib_time_0 = vlib_time_now (vm); - return error; -} - -/* Set up the API message handling tables */ -static clib_error_t * -vxlan_gpe_plugin_api_hookup (vlib_main_t * vm) -{ - vxlan_gpe_ioam_main_t *sm = &vxlan_gpe_ioam_main; -#define _(N,n) \ - vl_msg_api_set_handlers((VL_API_##N + sm->msg_id_base), \ - #n, \ - vl_api_##n##_t_handler, \ - vl_noop_handler, \ - vl_api_##n##_t_endian, \ - vl_api_##n##_t_print, \ - sizeof(vl_api_##n##_t), 1); - foreach_vxlan_gpe_plugin_api_msg; -#undef _ - - return 0; -} - -static clib_error_t * -vxlan_gpe_init (vlib_main_t * vm) -{ - vxlan_gpe_ioam_main_t *sm = &vxlan_gpe_ioam_main; - clib_error_t *error = 0; - u8 *name; - u32 encap_node_index = vxlan_gpe_encap_ioam_v4_node.index; - u32 decap_node_index = vxlan_gpe_decap_ioam_v4_node.index; - vlib_node_t *vxlan_gpe_encap_node = NULL; - vlib_node_t *vxlan_gpe_decap_node = NULL; - uword next_node = 0; - - name = format (0, "ioam_vxlan_gpe_%08x%c", api_version, 0); - - /* Ask for a correctly-sized block of API message decode slots */ - sm->msg_id_base = vl_msg_api_get_msg_ids - ((char *) name, VL_MSG_FIRST_AVAILABLE); - - error = vxlan_gpe_plugin_api_hookup (vm); - - /* Hook the ioam-encap node to vxlan-gpe-encap */ - vxlan_gpe_encap_node = vlib_get_node_by_name (vm, (u8 *) "vxlan-gpe-encap"); - sm->encap_v4_next_node = - vlib_node_add_next (vm, vxlan_gpe_encap_node->index, encap_node_index); - - vxlan_gpe_decap_node = - vlib_get_node_by_name (vm, (u8 *) "vxlan4-gpe-input"); - next_node = - vlib_node_add_next (vm, vxlan_gpe_decap_node->index, decap_node_index); - vxlan_gpe_register_decap_protocol (VXLAN_GPE_PROTOCOL_IOAM, next_node); - - vec_new (vxlan_gpe_ioam_sw_interface_t, pool_elts (sm->sw_interfaces)); - sm->dst_by_ip4 = hash_create_mem (0, sizeof (fib_prefix_t), sizeof (uword)); - - sm->dst_by_ip6 = hash_create_mem (0, sizeof (fib_prefix_t), sizeof (uword)); - - vxlan_gpe_ioam_interface_init (); - vec_free (name); - - return error; -} - -VLIB_INIT_FUNCTION (vxlan_gpe_init); - -/* - * fd.io coding-style-patch-verification: ON - * - * Local Variables: - * eval: (c-set-style "gnu") - * End: - */ diff --git a/plugins/ioam-plugin/ioam/lib-vxlan-gpe/vxlan_gpe_ioam.c b/plugins/ioam-plugin/ioam/lib-vxlan-gpe/vxlan_gpe_ioam.c deleted file mode 100644 index 6c04d9af..00000000 --- a/plugins/ioam-plugin/ioam/lib-vxlan-gpe/vxlan_gpe_ioam.c +++ /dev/null @@ -1,773 +0,0 @@ -/* - * Copyright (c) 2016 Cisco and/or its affiliates. - * Licensed under the Apache License, Version 2.0 (the "License"); - * you may not use this file except in compliance with the License. - * You may obtain a copy of the License at: - * - * http://www.apache.org/licenses/LICENSE-2.0 - * - * Unless required by applicable law or agreed to in writing, software - * distributed under the License is distributed on an "AS IS" BASIS, - * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. - * See the License for the specific language governing permissions and - * limitations under the License. - */ -#include -#include -#include -#include -#include -#include -#include - -vxlan_gpe_ioam_main_t vxlan_gpe_ioam_main; - -int -vxlan_gpe_ioam_set_rewrite (vxlan_gpe_tunnel_t * t, int has_trace_option, - int has_pot_option, int has_ppc_option, - u8 ipv6_set) -{ - vxlan_gpe_ioam_main_t *hm = &vxlan_gpe_ioam_main; - u32 size; - vxlan_gpe_ioam_hdr_t *vxlan_gpe_ioam_hdr; - u8 *current; - u8 trace_data_size = 0; - u8 pot_data_size = 0; - - if (has_trace_option == 0 && has_pot_option == 0) - return -1; - - /* Work out how much space we need */ - size = sizeof (vxlan_gpe_ioam_hdr_t); - - if (has_trace_option - && hm->add_options[VXLAN_GPE_OPTION_TYPE_IOAM_TRACE] != 0) - { - size += sizeof (vxlan_gpe_ioam_option_t); - size += hm->options_size[VXLAN_GPE_OPTION_TYPE_IOAM_TRACE]; - } - if (has_pot_option - && hm->add_options[VXLAN_GPE_OPTION_TYPE_IOAM_PROOF_OF_TRANSIT] != 0) - { - size += sizeof (vxlan_gpe_ioam_option_t); - size += hm->options_size[VXLAN_GPE_OPTION_TYPE_IOAM_PROOF_OF_TRANSIT]; - } - - t->rewrite_size = size; - - if (!ipv6_set) - { - vxlan4_gpe_rewrite (t, size, VXLAN_GPE_PROTOCOL_IOAM, - hm->encap_v4_next_node); - vxlan_gpe_ioam_hdr = - (vxlan_gpe_ioam_hdr_t *) (t->rewrite + - sizeof (ip4_vxlan_gpe_header_t)); - } - else - { - vxlan6_gpe_rewrite (t, size, VXLAN_GPE_PROTOCOL_IOAM, - VXLAN_GPE_ENCAP_NEXT_IP6_LOOKUP); - vxlan_gpe_ioam_hdr = - (vxlan_gpe_ioam_hdr_t *) (t->rewrite + - sizeof (ip6_vxlan_gpe_header_t)); - } - - - vxlan_gpe_ioam_hdr->type = VXLAN_GPE_PROTOCOL_IOAM; - /* Length of the header in octets */ - vxlan_gpe_ioam_hdr->length = size; - vxlan_gpe_ioam_hdr->protocol = t->protocol; - current = (u8 *) vxlan_gpe_ioam_hdr + sizeof (vxlan_gpe_ioam_hdr_t); - - if (has_trace_option - && hm->add_options[VXLAN_GPE_OPTION_TYPE_IOAM_TRACE] != 0) - { - if (0 != hm->add_options[VXLAN_GPE_OPTION_TYPE_IOAM_TRACE] (current, - &trace_data_size)) - return -1; - current += trace_data_size; - } - if (has_pot_option - && hm->add_options[VXLAN_GPE_OPTION_TYPE_IOAM_PROOF_OF_TRANSIT] != 0) - { - pot_data_size = - hm->options_size[VXLAN_GPE_OPTION_TYPE_IOAM_PROOF_OF_TRANSIT]; - if (0 == - hm->add_options[VXLAN_GPE_OPTION_TYPE_IOAM_PROOF_OF_TRANSIT] - (current, &pot_data_size)) - current += pot_data_size; - } - - return 0; -} - -int -vxlan_gpe_ioam_clear_rewrite (vxlan_gpe_tunnel_t * t, int has_trace_option, - int has_pot_option, int has_ppc_option, - u8 ipv6_set) -{ - - t->rewrite_size = 0; - - if (!ipv6_set) - { - vxlan4_gpe_rewrite (t, 0, 0, VXLAN_GPE_ENCAP_NEXT_IP4_LOOKUP); - } - else - { - vxlan6_gpe_rewrite (t, 0, 0, VXLAN_GPE_ENCAP_NEXT_IP6_LOOKUP); - } - - - return 0; -} - -clib_error_t * -vxlan_gpe_ioam_clear (vxlan_gpe_tunnel_t * t, - int has_trace_option, int has_pot_option, - int has_ppc_option, u8 ipv6_set) -{ - int rv; - rv = vxlan_gpe_ioam_clear_rewrite (t, 0, 0, 0, 0); - - if (rv == 0) - { - return (0); - } - else - { - return clib_error_return_code (0, rv, 0, - "vxlan_gpe_ioam_clear_rewrite returned %d", - rv); - } - -} - - -clib_error_t * -vxlan_gpe_ioam_set (vxlan_gpe_tunnel_t * t, - int has_trace_option, int has_pot_option, - int has_ppc_option, u8 ipv6_set) -{ - int rv; - rv = vxlan_gpe_ioam_set_rewrite (t, has_trace_option, - has_pot_option, has_ppc_option, ipv6_set); - - if (rv == 0) - { - return (0); - } - else - { - return clib_error_return_code (0, rv, 0, - "vxlan_gpe_ioam_set_rewrite returned %d", - rv); - } - -} - -static void -vxlan_gpe_set_clear_output_feature_on_intf (vlib_main_t * vm, - u32 sw_if_index0, u8 is_add) -{ - - - - vnet_feature_enable_disable ("ip4-output", "vxlan-gpe-transit-ioam", - sw_if_index0, is_add, - 0 /* void *feature_config */ , - 0 /* u32 n_feature_config_bytes */ ); - return; -} - -void -vxlan_gpe_clear_output_feature_on_all_intfs (vlib_main_t * vm) -{ - vnet_sw_interface_t *si = 0; - vnet_main_t *vnm = vnet_get_main (); - vnet_interface_main_t *im = &vnm->interface_main; - - pool_foreach (si, im->sw_interfaces, ( - { - vxlan_gpe_set_clear_output_feature_on_intf - (vm, si->sw_if_index, 0); - })); - return; -} - - -extern fib_forward_chain_type_t -fib_entry_get_default_chain_type (const fib_entry_t * fib_entry); - -int -vxlan_gpe_enable_disable_ioam_for_dest (vlib_main_t * vm, - ip46_address_t dst_addr, - u32 outer_fib_index, - u8 is_ipv4, u8 is_add) -{ - vxlan_gpe_ioam_main_t *hm = &vxlan_gpe_ioam_main; - u32 fib_index0 = 0; - u32 sw_if_index0 = ~0; - - fib_node_index_t fei = ~0; - fib_entry_t *fib_entry; - u32 adj_index0; - ip_adjacency_t *adj0; - fib_prefix_t fib_prefix; - //fib_forward_chain_type_t fct; - load_balance_t *lb_m, *lb_b; - const dpo_id_t *dpo0, *dpo1; - u32 i, j; - //vnet_hw_interface_t *hw; - - if (is_ipv4) - { - memset (&fib_prefix, 0, sizeof (fib_prefix_t)); - fib_prefix.fp_len = 32; - fib_prefix.fp_proto = FIB_PROTOCOL_IP4; - fib_prefix.fp_addr = dst_addr; - } - else - { - return 0; - } - - fei = fib_table_lookup (fib_index0, &fib_prefix); - fib_entry = fib_entry_get (fei); - - //fct = fib_entry_get_default_chain_type (fib_entry); - - if (!dpo_id_is_valid (&fib_entry->fe_lb /*[fct] */ )) - { - return (-1); - } - - lb_m = load_balance_get (fib_entry->fe_lb /*[fct] */ .dpoi_index); - - for (i = 0; i < lb_m->lb_n_buckets; i++) - { - dpo0 = load_balance_get_bucket_i (lb_m, i); - - if (dpo0->dpoi_type == DPO_LOAD_BALANCE) - { - lb_b = load_balance_get (dpo0->dpoi_index); - - for (j = 0; j < lb_b->lb_n_buckets; j++) - { - dpo1 = load_balance_get_bucket_i (lb_b, j); - - if (dpo1->dpoi_type == DPO_ADJACENCY) - { - adj_index0 = dpo1->dpoi_index; - - if (ADJ_INDEX_INVALID == adj_index0) - { - continue; - } - - adj0 = - ip_get_adjacency (&(ip4_main.lookup_main), adj_index0); - sw_if_index0 = adj0->rewrite_header.sw_if_index; - - if (~0 == sw_if_index0) - { - continue; - } - - - if (is_add) - { - vnet_feature_enable_disable ("ip4-output", - "vxlan-gpe-transit-ioam", - sw_if_index0, is_add, 0 - /* void *feature_config */ - , 0 /* u32 n_feature_config_bytes */ - ); - - vec_validate_init_empty (hm->bool_ref_by_sw_if_index, - sw_if_index0, ~0); - hm->bool_ref_by_sw_if_index[sw_if_index0] = 1; - } - else - { - hm->bool_ref_by_sw_if_index[sw_if_index0] = ~0; - } - } - } - } - } - - if (is_ipv4) - { - - uword *t = NULL; - vxlan_gpe_ioam_dest_tunnels_t *t1; - fib_prefix_t key4, *key4_copy; - hash_pair_t *hp; - memset (&key4, 0, sizeof (key4)); - key4.fp_proto = FIB_PROTOCOL_IP4; - key4.fp_addr.ip4.as_u32 = fib_prefix.fp_addr.ip4.as_u32; - t = hash_get_mem (hm->dst_by_ip4, &key4); - if (is_add) - { - if (t) - { - return 0; - } - pool_get_aligned (hm->dst_tunnels, t1, CLIB_CACHE_LINE_BYTES); - memset (t1, 0, sizeof (*t1)); - t1->fp_proto = FIB_PROTOCOL_IP4; - t1->dst_addr.ip4.as_u32 = fib_prefix.fp_addr.ip4.as_u32; - key4_copy = clib_mem_alloc (sizeof (*key4_copy)); - clib_memcpy (key4_copy, &key4, sizeof (*key4_copy)); - hash_set_mem (hm->dst_by_ip4, key4_copy, t1 - hm->dst_tunnels); - /* - * Attach to the FIB entry for the VxLAN-GPE destination - * and become its child. The dest route will invoke a callback - * when the fib entry changes, it can be used to - * re-program the output feature on the egress interface. - */ - - const fib_prefix_t tun_dst_pfx = { - .fp_len = 32, - .fp_proto = FIB_PROTOCOL_IP4, - .fp_addr = {.ip4 = t1->dst_addr.ip4,} - }; - - t1->fib_entry_index = - fib_table_entry_special_add (outer_fib_index, - &tun_dst_pfx, - FIB_SOURCE_RR, - FIB_ENTRY_FLAG_NONE, - ADJ_INDEX_INVALID); - t1->sibling_index = - fib_entry_child_add (t1->fib_entry_index, - hm->fib_entry_type, t1 - hm->dst_tunnels); - t1->outer_fib_index = outer_fib_index; - - } - else - { - if (!t) - { - return 0; - } - t1 = pool_elt_at_index (hm->dst_tunnels, t[0]); - hp = hash_get_pair (hm->dst_by_ip4, &key4); - key4_copy = (void *) (hp->key); - hash_unset_mem (hm->dst_by_ip4, &key4); - clib_mem_free (key4_copy); - pool_put (hm->dst_tunnels, t1); - } - } - else - { - // TBD for IPv6 - } - - return 0; -} - -void -vxlan_gpe_refresh_output_feature_on_all_dest (void) -{ - vxlan_gpe_ioam_main_t *hm = &vxlan_gpe_ioam_main; - vxlan_gpe_ioam_dest_tunnels_t *t; - u32 i; - if (pool_elts (hm->dst_tunnels) == 0) - return; - vxlan_gpe_clear_output_feature_on_all_intfs (hm->vlib_main); - i = vec_len (hm->bool_ref_by_sw_if_index); - vec_free (hm->bool_ref_by_sw_if_index); - vec_validate_init_empty (hm->bool_ref_by_sw_if_index, i, ~0); - pool_foreach (t, hm->dst_tunnels, ( - { - vxlan_gpe_enable_disable_ioam_for_dest - (hm->vlib_main, - t->dst_addr, - t->outer_fib_index, - (t->fp_proto == FIB_PROTOCOL_IP4), 1 - /* is_add */ - ); - } - )); - return; -} - -void -vxlan_gpe_clear_output_feature_on_select_intfs (void) -{ - vxlan_gpe_ioam_main_t *hm = &vxlan_gpe_ioam_main; - u32 sw_if_index0 = 0; - for (sw_if_index0 = 0; - sw_if_index0 < vec_len (hm->bool_ref_by_sw_if_index); sw_if_index0++) - { - if (hm->bool_ref_by_sw_if_index[sw_if_index0] == 0xFF) - { - vxlan_gpe_set_clear_output_feature_on_intf - (hm->vlib_main, sw_if_index0, 0); - } - } - - return; -} - -static clib_error_t * -vxlan_gpe_set_ioam_rewrite_command_fn (vlib_main_t * - vm, - unformat_input_t - * input, vlib_cli_command_t * cmd) -{ - vxlan_gpe_ioam_main_t *hm = &vxlan_gpe_ioam_main; - ip46_address_t local, remote; - u8 local_set = 0; - u8 remote_set = 0; - u8 ipv4_set = 0; - u8 ipv6_set = 0; - u32 vni; - u8 vni_set = 0; - u8 disable = 0; - clib_error_t *rv = 0; - vxlan4_gpe_tunnel_key_t key4; - vxlan6_gpe_tunnel_key_t key6; - uword *p; - vxlan_gpe_main_t *gm = &vxlan_gpe_main; - vxlan_gpe_tunnel_t *t = 0; - while (unformat_check_input (input) != UNFORMAT_END_OF_INPUT) - { - if (unformat (input, "local %U", unformat_ip4_address, &local.ip4)) - { - local_set = 1; - ipv4_set = 1; - } - else - if (unformat (input, "remote %U", unformat_ip4_address, &remote.ip4)) - { - remote_set = 1; - ipv4_set = 1; - } - else if (unformat (input, "local %U", unformat_ip6_address, &local.ip6)) - { - local_set = 1; - ipv6_set = 1; - } - else - if (unformat (input, "remote %U", unformat_ip6_address, &remote.ip6)) - { - remote_set = 1; - ipv6_set = 1; - } - else if (unformat (input, "vni %d", &vni)) - vni_set = 1; - else if (unformat (input, "disable")) - disable = 1; - else - break; - } - - if (local_set == 0) - return clib_error_return (0, "tunnel local address not specified"); - if (remote_set == 0) - return clib_error_return (0, "tunnel remote address not specified"); - if (ipv4_set && ipv6_set) - return clib_error_return (0, "both IPv4 and IPv6 addresses specified"); - if ((ipv4_set - && memcmp (&local.ip4, &remote.ip4, - sizeof (local.ip4)) == 0) || (ipv6_set - && - memcmp - (&local.ip6, - &remote.ip6, - sizeof (local.ip6)) == 0)) - return clib_error_return (0, "src and dst addresses are identical"); - if (vni_set == 0) - return clib_error_return (0, "vni not specified"); - if (!ipv6_set) - { - key4.local = local.ip4.as_u32; - key4.remote = remote.ip4.as_u32; - key4.vni = clib_host_to_net_u32 (vni << 8); - key4.pad = 0; - p = hash_get_mem (gm->vxlan4_gpe_tunnel_by_key, &key4); - } - else - { - key6.local.as_u64[0] = local.ip6.as_u64[0]; - key6.local.as_u64[1] = local.ip6.as_u64[1]; - key6.remote.as_u64[0] = remote.ip6.as_u64[0]; - key6.remote.as_u64[1] = remote.ip6.as_u64[1]; - key6.vni = clib_host_to_net_u32 (vni << 8); - p = hash_get_mem (gm->vxlan6_gpe_tunnel_by_key, &key6); - } - - if (!p) - return clib_error_return (0, "VxLAN Tunnel not found"); - t = pool_elt_at_index (gm->tunnels, p[0]); - if (!disable) - { - rv = - vxlan_gpe_ioam_set (t, hm->has_trace_option, - hm->has_pot_option, hm->has_ppc_option, ipv6_set); - } - else - { - rv = vxlan_gpe_ioam_clear (t, 0, 0, 0, 0); - } - return rv; -} - - -/* *INDENT-OFF* */ -VLIB_CLI_COMMAND (vxlan_gpe_set_ioam_rewrite_cmd, static) = { - .path = "set vxlan-gpe-ioam", - .short_help = "set vxlan-gpe-ioam vxlan [disable]", - .function = vxlan_gpe_set_ioam_rewrite_command_fn, -}; -/* *INDENT-ON* */ - - - -clib_error_t * -vxlan_gpe_ioam_enable (int has_trace_option, - int has_pot_option, int has_ppc_option) -{ - vxlan_gpe_ioam_main_t *hm = &vxlan_gpe_ioam_main; - hm->has_trace_option = has_trace_option; - hm->has_pot_option = has_pot_option; - hm->has_ppc_option = has_ppc_option; - if (hm->has_trace_option) - { - vxlan_gpe_trace_profile_setup (); - } - - return 0; -} - -clib_error_t * -vxlan_gpe_ioam_disable (int - has_trace_option, - int has_pot_option, int has_ppc_option) -{ - vxlan_gpe_ioam_main_t *hm = &vxlan_gpe_ioam_main; - hm->has_trace_option = has_trace_option; - hm->has_pot_option = has_pot_option; - hm->has_ppc_option = has_ppc_option; - if (!hm->has_trace_option) - { - vxlan_gpe_trace_profile_cleanup (); - } - - return 0; -} - -void -vxlan_gpe_set_next_override (uword next) -{ - vxlan_gpe_ioam_main_t *hm = &vxlan_gpe_ioam_main; - hm->decap_v4_next_override = next; - return; -} - -static clib_error_t * -vxlan_gpe_set_ioam_flags_command_fn (vlib_main_t * vm, - unformat_input_t - * input, vlib_cli_command_t * cmd) -{ - int has_trace_option = 0; - int has_pot_option = 0; - int has_ppc_option = 0; - clib_error_t *rv = 0; - while (unformat_check_input (input) != UNFORMAT_END_OF_INPUT) - { - if (unformat (input, "trace")) - has_trace_option = 1; - else if (unformat (input, "pot")) - has_pot_option = 1; - else if (unformat (input, "ppc encap")) - has_ppc_option = PPC_ENCAP; - else if (unformat (input, "ppc decap")) - has_ppc_option = PPC_DECAP; - else if (unformat (input, "ppc none")) - has_ppc_option = PPC_NONE; - else - break; - } - - - rv = - vxlan_gpe_ioam_enable (has_trace_option, has_pot_option, has_ppc_option); - return rv; -} - -/* *INDENT-OFF* */ -VLIB_CLI_COMMAND (vxlan_gpe_set_ioam_flags_cmd, static) = -{ -.path = "set vxlan-gpe-ioam rewrite", -.short_help = "set vxlan-gpe-ioam [trace] [pot] [ppc ]", -.function = vxlan_gpe_set_ioam_flags_command_fn,}; -/* *INDENT-ON* */ - - -int vxlan_gpe_ioam_disable_for_dest - (vlib_main_t * vm, ip46_address_t dst_addr, u32 outer_fib_index, - u8 ipv4_set) -{ - vxlan_gpe_ioam_dest_tunnels_t *t; - vxlan_gpe_ioam_main_t *hm = &vxlan_gpe_ioam_main; - - vxlan_gpe_enable_disable_ioam_for_dest (hm->vlib_main, - dst_addr, outer_fib_index, ipv4_set, - 0); - if (pool_elts (hm->dst_tunnels) == 0) - { - vxlan_gpe_clear_output_feature_on_select_intfs (); - return 0; - } - - pool_foreach (t, hm->dst_tunnels, ( - { - vxlan_gpe_enable_disable_ioam_for_dest - (hm->vlib_main, - t->dst_addr, - t->outer_fib_index, - (t->fp_proto == - FIB_PROTOCOL_IP4), 1 /* is_add */ ); - } - )); - vxlan_gpe_clear_output_feature_on_select_intfs (); - return (0); - -} - -static clib_error_t *vxlan_gpe_set_ioam_transit_rewrite_command_fn - (vlib_main_t * vm, unformat_input_t * input, vlib_cli_command_t * cmd) -{ - vxlan_gpe_ioam_main_t *hm = &vxlan_gpe_ioam_main; - ip46_address_t dst_addr; - u8 dst_addr_set = 0; - u8 ipv4_set = 0; - u8 ipv6_set = 0; - u8 disable = 0; - clib_error_t *rv = 0; - u32 outer_fib_index = 0; - while (unformat_check_input (input) != UNFORMAT_END_OF_INPUT) - { - if (unformat (input, "dst-ip %U", unformat_ip4_address, &dst_addr.ip4)) - { - dst_addr_set = 1; - ipv4_set = 1; - } - else - if (unformat - (input, "dst-ip %U", unformat_ip6_address, &dst_addr.ip6)) - { - dst_addr_set = 1; - ipv6_set = 1; - } - else if (unformat (input, "outer-fib-index %d", &outer_fib_index)) - { - } - - else if (unformat (input, "disable")) - disable = 1; - else - break; - } - - if (dst_addr_set == 0) - return clib_error_return (0, "tunnel destination address not specified"); - if (ipv4_set && ipv6_set) - return clib_error_return (0, "both IPv4 and IPv6 addresses specified"); - if (!disable) - { - vxlan_gpe_enable_disable_ioam_for_dest (hm->vlib_main, - dst_addr, outer_fib_index, - ipv4_set, 1); - } - else - { - vxlan_gpe_ioam_disable_for_dest - (vm, dst_addr, outer_fib_index, ipv4_set); - } - return rv; -} - - /* *INDENT-OFF* */ -VLIB_CLI_COMMAND (vxlan_gpe_set_ioam_transit_rewrite_cmd, static) = { - .path = "set vxlan-gpe-ioam-transit", - .short_help = "set vxlan-gpe-ioam-transit dst-ip [outer-fib-index ] [disable]", - .function = vxlan_gpe_set_ioam_transit_rewrite_command_fn, -}; -/* *INDENT-ON* */ - -clib_error_t *clear_vxlan_gpe_ioam_rewrite_command_fn - (vlib_main_t * vm, unformat_input_t * input, vlib_cli_command_t * cmd) -{ - return (vxlan_gpe_ioam_disable (0, 0, 0)); -} - -/* *INDENT-OFF* */ -VLIB_CLI_COMMAND (vxlan_gpe_clear_ioam_flags_cmd, static) = -{ -.path = "clear vxlan-gpe-ioam rewrite", -.short_help = "clear vxlan-gpe-ioam rewrite", -.function = clear_vxlan_gpe_ioam_rewrite_command_fn, -}; -/* *INDENT-ON* */ - - -/** - * Function definition to backwalk a FIB node - */ -static fib_node_back_walk_rc_t -vxlan_gpe_ioam_back_walk (fib_node_t * node, fib_node_back_walk_ctx_t * ctx) -{ - vxlan_gpe_refresh_output_feature_on_all_dest (); - return (FIB_NODE_BACK_WALK_CONTINUE); -} - -/** - * Function definition to get a FIB node from its index - */ -static fib_node_t * -vxlan_gpe_ioam_fib_node_get (fib_node_index_t index) -{ - vxlan_gpe_ioam_main_t *hm = &vxlan_gpe_ioam_main; - return (&hm->node); -} - -/** - * Function definition to inform the FIB node that its last lock has gone. - */ -static void -vxlan_gpe_ioam_last_lock_gone (fib_node_t * node) -{ - ASSERT (0); -} - - -/* - * Virtual function table registered by MPLS GRE tunnels - * for participation in the FIB object graph. - */ -const static fib_node_vft_t vxlan_gpe_ioam_vft = { - .fnv_get = vxlan_gpe_ioam_fib_node_get, - .fnv_last_lock = vxlan_gpe_ioam_last_lock_gone, - .fnv_back_walk = vxlan_gpe_ioam_back_walk, -}; - -void -vxlan_gpe_ioam_interface_init (void) -{ - vxlan_gpe_ioam_main_t *hm = &vxlan_gpe_ioam_main; - hm->fib_entry_type = fib_node_register_new_type (&vxlan_gpe_ioam_vft); - return; -} - -/* - * fd.io coding-style-patch-verification: ON - * - * Local Variables: - * eval: (c-set-style "gnu") - * End: - */ diff --git a/plugins/ioam-plugin/ioam/lib-vxlan-gpe/vxlan_gpe_ioam.h b/plugins/ioam-plugin/ioam/lib-vxlan-gpe/vxlan_gpe_ioam.h deleted file mode 100644 index 3b7d72cf..00000000 --- a/plugins/ioam-plugin/ioam/lib-vxlan-gpe/vxlan_gpe_ioam.h +++ /dev/null @@ -1,183 +0,0 @@ -/* - * Copyright (c) 2015 Cisco and/or its affiliates. - * Licensed under the Apache License, Version 2.0 (the "License"); - * you may not use this file except in compliance with the License. - * You may obtain a copy of the License at: - * - * http://www.apache.org/licenses/LICENSE-2.0 - * - * Unless required by applicable law or agreed to in writing, software - * distributed under the License is distributed on an "AS IS" BASIS, - * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. - * See the License for the specific language governing permissions and - * limitations under the License. - */ -#ifndef __included_vxlan_gpe_ioam_h__ -#define __included_vxlan_gpe_ioam_h__ - -#include -#include -#include -#include - - -typedef struct vxlan_gpe_sw_interface_ -{ - u32 sw_if_index; -} vxlan_gpe_ioam_sw_interface_t; - -typedef struct vxlan_gpe_dest_tunnels_ -{ - ip46_address_t dst_addr; - u32 fp_proto; - u32 sibling_index; - fib_node_index_t fib_entry_index; - u32 outer_fib_index; -} vxlan_gpe_ioam_dest_tunnels_t; - -typedef struct vxlan_gpe_ioam_main_ -{ - /** - * Linkage into the FIB object graph - */ - fib_node_t node; - - /* time scale transform. Joy. */ - u32 unix_time_0; - f64 vlib_time_0; - - - /* Trace option */ - u8 has_trace_option; - - /* Pot option */ - u8 has_pot_option; - -#define PPC_NONE 0 -#define PPC_ENCAP 1 -#define PPC_DECAP 2 - u8 has_ppc_option; - -#define TSP_SECONDS 0 -#define TSP_MILLISECONDS 1 -#define TSP_MICROSECONDS 2 -#define TSP_NANOSECONDS 3 - - /* Array of function pointers to ADD and POP VxLAN-GPE iOAM option handling routines */ - u8 options_size[256]; - int (*add_options[256]) (u8 * rewrite_string, u8 * rewrite_size); - int (*pop_options[256]) (ip4_header_t * ip, vxlan_gpe_ioam_option_t * opt); - - /* Array of function pointers to iOAM option handling routines */ - int (*options[256]) (vlib_buffer_t * b, vxlan_gpe_ioam_option_t * opt, - u8 is_ipv4, u8 use_adj); - u8 *(*trace[256]) (u8 * s, vxlan_gpe_ioam_option_t * opt); - - /* API message ID base */ - u16 msg_id_base; - - /* Override to export for iOAM */ - uword decap_v4_next_override; - uword decap_v6_next_override; - - /* sequence of node graph for encap */ - uword encap_v4_next_node; - uword encap_v6_next_node; - - /* Software interfaces. */ - vxlan_gpe_ioam_sw_interface_t *sw_interfaces; - - /* hash ip4/ip6 -> list of destinations for doing transit iOAM operation */ - vxlan_gpe_ioam_dest_tunnels_t *dst_tunnels; - uword *dst_by_ip4; - uword *dst_by_ip6; - - /** per sw_if_index, to maintain bitmap */ - u8 *bool_ref_by_sw_if_index; - fib_node_type_t fib_entry_type; - - /** State convenience vlib_main_t */ - vlib_main_t *vlib_main; - /** State convenience vnet_main_t */ - vnet_main_t *vnet_main; - - -} vxlan_gpe_ioam_main_t; -extern vxlan_gpe_ioam_main_t vxlan_gpe_ioam_main; - -/* - * Primary h-b-h handler trace support - */ -typedef struct -{ - u32 next_index; - u32 trace_len; - u8 option_data[256]; -} ioam_trace_t; - - -vlib_node_registration_t vxlan_gpe_encap_ioam_v4_node; -vlib_node_registration_t vxlan_gpe_decap_ioam_v4_node; -vlib_node_registration_t vxlan_gpe_transit_ioam_v4_node; - -clib_error_t *vxlan_gpe_ioam_enable (int has_trace_option, int has_pot_option, - int has_ppc_option); - -clib_error_t *vxlan_gpe_ioam_disable (int has_trace_option, - int has_pot_option, int has_ppc_option); - -clib_error_t *vxlan_gpe_ioam_set (vxlan_gpe_tunnel_t * t, - int has_trace_option, - int has_pot_option, - int has_ppc_option, u8 ipv6_set); -clib_error_t *vxlan_gpe_ioam_clear (vxlan_gpe_tunnel_t * t, - int has_trace_option, int has_pot_option, - int has_ppc_option, u8 ipv6_set); - -int vxlan_gpe_ioam_add_register_option (u8 option, - u8 size, - int rewrite_options (u8 * - rewrite_string, - u8 * - rewrite_size)); - -int vxlan_gpe_add_unregister_option (u8 option); - -int vxlan_gpe_ioam_register_option (u8 option, - int options (vlib_buffer_t * b, - vxlan_gpe_ioam_option_t * - opt, u8 is_ipv4, u8 use_adj), - u8 * trace (u8 * s, - vxlan_gpe_ioam_option_t * - opt)); -int vxlan_gpe_ioam_unregister_option (u8 option); - -int vxlan_gpe_trace_profile_setup (void); - -int vxlan_gpe_trace_profile_cleanup (void); -extern void vxlan_gpe_ioam_interface_init (void); -int -vxlan_gpe_enable_disable_ioam_for_dest (vlib_main_t * vm, - ip46_address_t dst_addr, - u32 outer_fib_index, - u8 is_ipv4, u8 is_add); -int vxlan_gpe_ioam_disable_for_dest - (vlib_main_t * vm, ip46_address_t dst_addr, u32 outer_fib_index, - u8 ipv4_set); - -typedef enum -{ - VXLAN_GPE_DECAP_IOAM_V4_NEXT_POP, - VXLAN_GPE_DECAP_IOAM_V4_NEXT_DROP, - VXLAN_GPE_DECAP_IOAM_V4_N_NEXT -} vxlan_gpe_decap_ioam_v4_next_t; - -#endif - -/* - * fd.io coding-style-patch-verification: ON - * - * Local Variables: - * eval: (c-set-style "gnu") - * End: - */ diff --git a/plugins/ioam-plugin/ioam/lib-vxlan-gpe/vxlan_gpe_ioam_packet.h b/plugins/ioam-plugin/ioam/lib-vxlan-gpe/vxlan_gpe_ioam_packet.h deleted file mode 100644 index a7ef859e..00000000 --- a/plugins/ioam-plugin/ioam/lib-vxlan-gpe/vxlan_gpe_ioam_packet.h +++ /dev/null @@ -1,61 +0,0 @@ -/* - * Copyright (c) 2015 Cisco and/or its affiliates. - * Licensed under the Apache License, Version 2.0 (the "License"); - * you may not use this file except in compliance with the License. - * You may obtain a copy of the License at: - * - * http://www.apache.org/licenses/LICENSE-2.0 - * - * Unless required by applicable law or agreed to in writing, software - * distributed under the License is distributed on an "AS IS" BASIS, - * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. - * See the License for the specific language governing permissions and - * limitations under the License. - */ -#ifndef __included_vxlan_gpe_ioam_packet_h__ -#define __included_vxlan_gpe_ioam_packet_h__ - -#include -#include -#include - - - -#define VXLAN_GPE_OPTION_TYPE_IOAM_TRACE 59 -#define VXLAN_GPE_OPTION_TYPE_IOAM_PROOF_OF_TRANSIT 60 - -/** - * @brief VXLAN GPE Extension (iOAM) Header definition - */ -typedef struct -{ - u8 type; - u8 length; - /** Reserved */ - u8 reserved; - /** see vxlan_gpe_protocol_t */ - u8 protocol; -} vxlan_gpe_ioam_hdr_t; - -/* - * @brief VxLAN GPE iOAM Option definition - */ -typedef struct -{ - /* Option Type */ - u8 type; - /* Length in octets of the option data field */ - u8 length; -} vxlan_gpe_ioam_option_t; - - -#endif - - -/* - * fd.io coding-style-patch-verification: ON - * - * Local Variables: - * eval: (c-set-style "gnu") - * End: - */ diff --git a/plugins/ioam-plugin/ioam/lib-vxlan-gpe/vxlan_gpe_ioam_trace.c b/plugins/ioam-plugin/ioam/lib-vxlan-gpe/vxlan_gpe_ioam_trace.c deleted file mode 100644 index e37b1642..00000000 --- a/plugins/ioam-plugin/ioam/lib-vxlan-gpe/vxlan_gpe_ioam_trace.c +++ /dev/null @@ -1,552 +0,0 @@ -/* - * Copyright (c) 2016 Cisco and/or its affiliates. - * Licensed under the Apache License, Version 2.0 (the "License"); - * you may not use this file except in compliance with the License. - * You may obtain a copy of the License at: - * - * http://www.apache.org/licenses/LICENSE-2.0 - * - * Unless required by applicable law or agreed to in writing, software - * distributed under the License is distributed on an "AS IS" BASIS, - * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. - * See the License for the specific language governing permissions and - * limitations under the License. - */ -#include -#include -#include -#include - -#include -#include - -#include -#include -#include - -#include -#include - -/* Timestamp precision multipliers for seconds, milliseconds, microseconds - * and nanoseconds respectively. - */ -static f64 trace_tsp_mul[4] = { 1, 1e3, 1e6, 1e9 }; - -typedef union -{ - u64 as_u64; - u32 as_u32[2]; -} time_u64_t; - - -/* *INDENT-OFF* */ -typedef CLIB_PACKED(struct { - vxlan_gpe_ioam_option_t hdr; - u8 ioam_trace_type; - u8 data_list_elts_left; - u32 elts[0]; /* Variable type. So keep it generic */ -}) vxlan_gpe_ioam_trace_option_t; -/* *INDENT-ON* */ - - -#define foreach_vxlan_gpe_ioam_trace_stats \ - _(SUCCESS, "Pkts updated with TRACE records") \ - _(FAILED, "Errors in TRACE due to lack of TRACE records") - -static char *vxlan_gpe_ioam_trace_stats_strings[] = { -#define _(sym,string) string, - foreach_vxlan_gpe_ioam_trace_stats -#undef _ -}; - -typedef enum -{ -#define _(sym,str) VXLAN_GPE_IOAM_TRACE_##sym, - foreach_vxlan_gpe_ioam_trace_stats -#undef _ - VXLAN_GPE_IOAM_TRACE_N_STATS, -} vxlan_gpe_ioam_trace_stats_t; - - -typedef struct -{ - /* stats */ - u64 counters[ARRAY_LEN (vxlan_gpe_ioam_trace_stats_strings)]; - - /* convenience */ - vlib_main_t *vlib_main; - vnet_main_t *vnet_main; -} vxlan_gpe_ioam_trace_main_t; - -vxlan_gpe_ioam_trace_main_t vxlan_gpe_ioam_trace_main; - -int -vxlan_gpe_ioam_add_register_option (u8 option, - u8 size, - int rewrite_options (u8 * rewrite_string, - u8 * rewrite_size)) -{ - vxlan_gpe_ioam_main_t *hm = &vxlan_gpe_ioam_main; - - ASSERT (option < ARRAY_LEN (hm->add_options)); - - /* Already registered */ - if (hm->add_options[option]) - return (-1); - - hm->add_options[option] = rewrite_options; - hm->options_size[option] = size; - - return (0); -} - -int -vxlan_gpe_add_unregister_option (u8 option) -{ - vxlan_gpe_ioam_main_t *hm = &vxlan_gpe_ioam_main; - - ASSERT (option < ARRAY_LEN (hm->add_options)); - - /* Not registered */ - if (!hm->add_options[option]) - return (-1); - - hm->add_options[option] = NULL; - hm->options_size[option] = 0; - return (0); -} - - -int -vxlan_gpe_ioam_register_option (u8 option, - int options (vlib_buffer_t * b, - vxlan_gpe_ioam_option_t * opt, - u8 is_ipv4, u8 use_adj), - u8 * trace (u8 * s, - vxlan_gpe_ioam_option_t * opt)) -{ - vxlan_gpe_ioam_main_t *im = &vxlan_gpe_ioam_main; - - ASSERT (option < ARRAY_LEN (im->options)); - - /* Already registered */ - if (im->options[option]) - return (-1); - - im->options[option] = options; - im->trace[option] = trace; - - return (0); -} - -int -vxlan_gpe_ioam_unregister_option (u8 option) -{ - vxlan_gpe_ioam_main_t *hm = &vxlan_gpe_ioam_main; - - ASSERT (option < ARRAY_LEN (hm->options)); - - /* Not registered */ - if (!hm->options[option]) - return (-1); - - hm->options[option] = NULL; - hm->trace[option] = NULL; - - return (0); -} - - -always_inline void -vxlan_gpe_ioam_trace_stats_increment_counter (u32 counter_index, - u64 increment) -{ - vxlan_gpe_ioam_trace_main_t *hm = &vxlan_gpe_ioam_trace_main; - - hm->counters[counter_index] += increment; -} - - -static u8 * -format_ioam_data_list_element (u8 * s, va_list * args) -{ - u32 *elt = va_arg (*args, u32 *); - u8 *trace_type_p = va_arg (*args, u8 *); - u8 trace_type = *trace_type_p; - - - if (trace_type & BIT_TTL_NODEID) - { - u32 ttl_node_id_host_byte_order = clib_net_to_host_u32 (*elt); - s = format (s, "ttl 0x%x node id 0x%x ", - ttl_node_id_host_byte_order >> 24, - ttl_node_id_host_byte_order & 0x00FFFFFF); - - elt++; - } - - if (trace_type & BIT_ING_INTERFACE && trace_type & BIT_ING_INTERFACE) - { - u32 ingress_host_byte_order = clib_net_to_host_u32 (*elt); - s = format (s, "ingress 0x%x egress 0x%x ", - ingress_host_byte_order >> 16, - ingress_host_byte_order & 0xFFFF); - elt++; - } - - if (trace_type & BIT_TIMESTAMP) - { - u32 ts_in_host_byte_order = clib_net_to_host_u32 (*elt); - s = format (s, "ts 0x%x \n", ts_in_host_byte_order); - elt++; - } - - if (trace_type & BIT_APPDATA) - { - u32 appdata_in_host_byte_order = clib_net_to_host_u32 (*elt); - s = format (s, "app 0x%x ", appdata_in_host_byte_order); - elt++; - } - - return s; -} - - - -int -vxlan_gpe_ioam_trace_rewrite_handler (u8 * rewrite_string, u8 * rewrite_size) -{ - vxlan_gpe_ioam_trace_option_t *trace_option = NULL; - u8 trace_data_size = 0; - u8 trace_option_elts = 0; - trace_profile *profile = NULL; - - - profile = trace_profile_find (); - - if (PREDICT_FALSE (!profile)) - { - return (-1); - } - - if (PREDICT_FALSE (!rewrite_string)) - return -1; - - trace_option_elts = profile->num_elts; - trace_data_size = fetch_trace_data_size (profile->trace_type); - trace_option = (vxlan_gpe_ioam_trace_option_t *) rewrite_string; - trace_option->hdr.type = VXLAN_GPE_OPTION_TYPE_IOAM_TRACE; - trace_option->hdr.length = 2 /*ioam_trace_type,data_list_elts_left */ + - trace_option_elts * trace_data_size; - trace_option->ioam_trace_type = profile->trace_type & TRACE_TYPE_MASK; - trace_option->data_list_elts_left = trace_option_elts; - *rewrite_size = - sizeof (vxlan_gpe_ioam_trace_option_t) + - (trace_option_elts * trace_data_size); - - return 0; -} - - -int -vxlan_gpe_ioam_trace_data_list_handler (vlib_buffer_t * b, - vxlan_gpe_ioam_option_t * opt, - u8 is_ipv4, u8 use_adj) -{ - u8 elt_index = 0; - vxlan_gpe_ioam_trace_option_t *trace = - (vxlan_gpe_ioam_trace_option_t *) opt; - time_u64_t time_u64; - u32 *elt; - int rv = 0; - trace_profile *profile = NULL; - vxlan_gpe_ioam_main_t *hm = &vxlan_gpe_ioam_main; - - - profile = trace_profile_find (); - - if (PREDICT_FALSE (!profile)) - { - return (-1); - } - - - time_u64.as_u64 = 0; - - if (PREDICT_TRUE (trace->data_list_elts_left)) - { - trace->data_list_elts_left--; - /* fetch_trace_data_size returns in bytes. Convert it to 4-bytes - * to skip to this node's location. - */ - elt_index = - trace->data_list_elts_left * - fetch_trace_data_size (trace->ioam_trace_type) / 4; - elt = &trace->elts[elt_index]; - if (is_ipv4) - { - if (trace->ioam_trace_type & BIT_TTL_NODEID) - { - ip4_header_t *ip0 = vlib_buffer_get_current (b); - /* The transit case is the only case where the TTL decrement happens - * before iOAM processing. For now, use the use_adj flag as an overload. - * We can probably use a separate flag instead of overloading the use_adj flag. - */ - *elt = clib_host_to_net_u32 (((ip0->ttl - 1 + use_adj) << 24) | - profile->node_id); - elt++; - } - - if (trace->ioam_trace_type & BIT_ING_INTERFACE) - { - u16 tx_if = 0; - u32 adj_index = vnet_buffer (b)->ip.adj_index[VLIB_TX]; - ip4_main_t *im4 = &ip4_main; - ip_lookup_main_t *lm = &im4->lookup_main; - if (use_adj) - { - ip_adjacency_t *adj = ip_get_adjacency (lm, adj_index); - tx_if = adj->rewrite_header.sw_if_index & 0xFFFF; - } - - *elt = - (vnet_buffer (b)->sw_if_index[VLIB_RX] & 0xFFFF) << 16 | - tx_if; - *elt = clib_host_to_net_u32 (*elt); - elt++; - } - } - else - { - if (trace->ioam_trace_type & BIT_TTL_NODEID) - { - ip6_header_t *ip0 = vlib_buffer_get_current (b); - *elt = clib_host_to_net_u32 ((ip0->hop_limit << 24) | - profile->node_id); - elt++; - } - if (trace->ioam_trace_type & BIT_ING_INTERFACE) - { - u16 tx_if = 0; - u32 adj_index = vnet_buffer (b)->ip.adj_index[VLIB_TX]; - ip6_main_t *im6 = &ip6_main; - ip_lookup_main_t *lm = &im6->lookup_main; - if (use_adj) - { - ip_adjacency_t *adj = ip_get_adjacency (lm, adj_index); - tx_if = adj->rewrite_header.sw_if_index & 0xFFFF; - } - - *elt = - (vnet_buffer (b)->sw_if_index[VLIB_RX] & 0xFFFF) << 16 | - tx_if; - *elt = clib_host_to_net_u32 (*elt); - elt++; - } - } - - if (trace->ioam_trace_type & BIT_TIMESTAMP) - { - /* Send least significant 32 bits */ - f64 time_f64 = - (f64) (((f64) hm->unix_time_0) + - (vlib_time_now (hm->vlib_main) - hm->vlib_time_0)); - - time_u64.as_u64 = time_f64 * trace_tsp_mul[profile->trace_tsp]; - *elt = clib_host_to_net_u32 (time_u64.as_u32[0]); - elt++; - } - - if (trace->ioam_trace_type & BIT_APPDATA) - { - /* $$$ set elt0->app_data */ - *elt = clib_host_to_net_u32 (profile->app_data); - elt++; - } - vxlan_gpe_ioam_trace_stats_increment_counter - (VXLAN_GPE_IOAM_TRACE_SUCCESS, 1); - } - else - { - vxlan_gpe_ioam_trace_stats_increment_counter - (VXLAN_GPE_IOAM_TRACE_FAILED, 1); - } - return (rv); -} - -u8 * -vxlan_gpe_ioam_trace_data_list_trace_handler (u8 * s, - vxlan_gpe_ioam_option_t * opt) -{ - vxlan_gpe_ioam_trace_option_t *trace; - u8 trace_data_size_in_words = 0; - u32 *elt; - int elt_index = 0; - - trace = (vxlan_gpe_ioam_trace_option_t *) opt; - s = - format (s, " Trace Type 0x%x , %d elts left\n", trace->ioam_trace_type, - trace->data_list_elts_left); - trace_data_size_in_words = - fetch_trace_data_size (trace->ioam_trace_type) / 4; - elt = &trace->elts[0]; - while ((u8 *) elt < ((u8 *) (&trace->elts[0]) + trace->hdr.length - 2 - /* -2 accounts for ioam_trace_type,elts_left */ )) - { - s = format (s, " [%d] %U\n", elt_index, - format_ioam_data_list_element, - elt, &trace->ioam_trace_type); - elt_index++; - elt += trace_data_size_in_words; - } - return (s); -} - - -static clib_error_t * -vxlan_gpe_show_ioam_trace_cmd_fn (vlib_main_t * vm, - unformat_input_t * input, - vlib_cli_command_t * cmd) -{ - vxlan_gpe_ioam_trace_main_t *hm = &vxlan_gpe_ioam_trace_main; - u8 *s = 0; - int i = 0; - - for (i = 0; i < VXLAN_GPE_IOAM_TRACE_N_STATS; i++) - { - s = format (s, " %s - %lu\n", vxlan_gpe_ioam_trace_stats_strings[i], - hm->counters[i]); - } - - vlib_cli_output (vm, "%v", s); - vec_free (s); - return 0; -} - - -/* *INDENT-OFF* */ -VLIB_CLI_COMMAND (vxlan_gpe_show_ioam_trace_cmd, static) = { - .path = "show ioam vxlan-gpe trace", - .short_help = "iOAM trace statistics", - .function = vxlan_gpe_show_ioam_trace_cmd_fn, -}; -/* *INDENT-ON* */ - - -static clib_error_t * -vxlan_gpe_ioam_trace_init (vlib_main_t * vm) -{ - vxlan_gpe_ioam_trace_main_t *hm = &vxlan_gpe_ioam_trace_main; - clib_error_t *error; - - if ((error = vlib_call_init_function (vm, ip_main_init))) - return (error); - - if ((error = vlib_call_init_function (vm, ip6_lookup_init))) - return error; - - if ((error = vlib_call_init_function (vm, vxlan_gpe_init))) - return (error); - - hm->vlib_main = vm; - hm->vnet_main = vnet_get_main (); - memset (hm->counters, 0, sizeof (hm->counters)); - - if (vxlan_gpe_ioam_register_option - (VXLAN_GPE_OPTION_TYPE_IOAM_TRACE, - vxlan_gpe_ioam_trace_data_list_handler, - vxlan_gpe_ioam_trace_data_list_trace_handler) < 0) - return (clib_error_create - ("registration of VXLAN_GPE_OPTION_TYPE_IOAM_TRACE failed")); - - - if (vxlan_gpe_ioam_add_register_option - (VXLAN_GPE_OPTION_TYPE_IOAM_TRACE, - sizeof (vxlan_gpe_ioam_trace_option_t), - vxlan_gpe_ioam_trace_rewrite_handler) < 0) - return (clib_error_create - ("registration of VXLAN_GPE_OPTION_TYPE_IOAM_TRACE for rewrite failed")); - - - return (0); -} - -VLIB_INIT_FUNCTION (vxlan_gpe_ioam_trace_init); - -int -vxlan_gpe_trace_profile_cleanup (void) -{ - vxlan_gpe_ioam_main_t *hm = &vxlan_gpe_ioam_main; - - hm->options_size[VXLAN_GPE_OPTION_TYPE_IOAM_TRACE] = 0; - - return 0; - -} - -static int -vxlan_gpe_ioam_trace_get_sizeof_handler (u32 * result) -{ - u16 size = 0; - u8 trace_data_size = 0; - trace_profile *profile = NULL; - - *result = 0; - - profile = trace_profile_find (); - - if (PREDICT_FALSE (!profile)) - { - return (-1); - } - - trace_data_size = fetch_trace_data_size (profile->trace_type); - if (PREDICT_FALSE (trace_data_size == 0)) - return VNET_API_ERROR_INVALID_VALUE; - - if (PREDICT_FALSE (profile->num_elts * trace_data_size > 254)) - return VNET_API_ERROR_INVALID_VALUE; - - size += - sizeof (vxlan_gpe_ioam_trace_option_t) + - profile->num_elts * trace_data_size; - *result = size; - - return 0; -} - - -int -vxlan_gpe_trace_profile_setup (void) -{ - u32 trace_size = 0; - vxlan_gpe_ioam_main_t *hm = &vxlan_gpe_ioam_main; - - trace_profile *profile = NULL; - - - profile = trace_profile_find (); - - if (PREDICT_FALSE (!profile)) - { - return (-1); - } - - - if (vxlan_gpe_ioam_trace_get_sizeof_handler (&trace_size) < 0) - return (-1); - - hm->options_size[VXLAN_GPE_OPTION_TYPE_IOAM_TRACE] = trace_size; - - return (0); -} - - - -/* - * fd.io coding-style-patch-verification: ON - * - * Local Variables: - * eval: (c-set-style "gnu") - * End: - */ diff --git a/plugins/ioam-plugin/ioam/lib-vxlan-gpe/vxlan_gpe_ioam_util.h b/plugins/ioam-plugin/ioam/lib-vxlan-gpe/vxlan_gpe_ioam_util.h deleted file mode 100644 index c0ad8d9d..00000000 --- a/plugins/ioam-plugin/ioam/lib-vxlan-gpe/vxlan_gpe_ioam_util.h +++ /dev/null @@ -1,172 +0,0 @@ -/* - * Copyright (c) 2015 Cisco and/or its affiliates. - * Licensed under the Apache License, Version 2.0 (the "License"); - * you may not use this file except in compliance with the License. - * You may obtain a copy of the License at: - * - * http://www.apache.org/licenses/LICENSE-2.0 - * - * Unless required by applicable law or agreed to in writing, software - * distributed under the License is distributed on an "AS IS" BASIS, - * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. - * See the License for the specific language governing permissions and - * limitations under the License. - */ -#ifndef __included_vxlan_gpe_ioam_util_h__ -#define __included_vxlan_gpe_ioam_util_h__ - -#include -#include -#include - - -typedef struct -{ - u32 tunnel_index; - ioam_trace_t fmt_trace; -} vxlan_gpe_ioam_v4_trace_t; - - -static u8 * -format_vxlan_gpe_ioam_v4_trace (u8 * s, va_list * args) -{ - CLIB_UNUSED (vlib_main_t * vm) = va_arg (*args, vlib_main_t *); - CLIB_UNUSED (vlib_node_t * node) = va_arg (*args, vlib_node_t *); - vxlan_gpe_ioam_v4_trace_t *t1 = va_arg (*args, vxlan_gpe_ioam_v4_trace_t *); - ioam_trace_t *t = &(t1->fmt_trace); - vxlan_gpe_ioam_option_t *fmt_trace0; - vxlan_gpe_ioam_option_t *opt0, *limit0; - vxlan_gpe_ioam_main_t *hm = &vxlan_gpe_ioam_main; - - u8 type0; - - fmt_trace0 = (vxlan_gpe_ioam_option_t *) t->option_data; - - s = format (s, "VXLAN-GPE-IOAM: next_index %d len %d traced %d", - t->next_index, fmt_trace0->length, t->trace_len); - - opt0 = (vxlan_gpe_ioam_option_t *) (fmt_trace0 + 1); - limit0 = (vxlan_gpe_ioam_option_t *) ((u8 *) fmt_trace0) + t->trace_len; - - while (opt0 < limit0) - { - type0 = opt0->type; - switch (type0) - { - case 0: /* Pad, just stop */ - opt0 = (vxlan_gpe_ioam_option_t *) ((u8 *) opt0) + 1; - break; - - default: - if (hm->trace[type0]) - { - s = (*hm->trace[type0]) (s, opt0); - } - else - { - s = - format (s, "\n unrecognized option %d length %d", type0, - opt0->length); - } - opt0 = - (vxlan_gpe_ioam_option_t *) (((u8 *) opt0) + opt0->length + - sizeof (vxlan_gpe_ioam_option_t)); - break; - } - } - - s = format (s, "VXLAN-GPE-IOAM: tunnel %d", t1->tunnel_index); - return s; -} - - -always_inline void -vxlan_gpe_encap_decap_ioam_v4_one_inline (vlib_main_t * vm, - vlib_node_runtime_t * node, - vlib_buffer_t * b0, - u32 * next0, u32 drop_node_val, - u8 use_adj) -{ - ip4_header_t *ip0; - udp_header_t *udp_hdr0; - vxlan_gpe_header_t *gpe_hdr0; - vxlan_gpe_ioam_hdr_t *gpe_ioam0; - vxlan_gpe_ioam_option_t *opt0; - vxlan_gpe_ioam_option_t *limit0; - vxlan_gpe_ioam_main_t *hm = &vxlan_gpe_ioam_main; - - /* Populate the iOAM header */ - ip0 = vlib_buffer_get_current (b0); - udp_hdr0 = (udp_header_t *) (ip0 + 1); - gpe_hdr0 = (vxlan_gpe_header_t *) (udp_hdr0 + 1); - gpe_ioam0 = (vxlan_gpe_ioam_hdr_t *) (gpe_hdr0 + 1); - opt0 = (vxlan_gpe_ioam_option_t *) (gpe_ioam0 + 1); - limit0 = (vxlan_gpe_ioam_option_t *) ((u8 *) gpe_ioam0 + gpe_ioam0->length); - - /* - * Basic validity checks - */ - if (gpe_ioam0->length > clib_net_to_host_u16 (ip0->length)) - { - *next0 = drop_node_val; - return; - } - - /* Scan the set of h-b-h options, process ones that we understand */ - while (opt0 < limit0) - { - u8 type0; - type0 = opt0->type; - switch (type0) - { - case 0: /* Pad1 */ - opt0 = (vxlan_gpe_ioam_option_t *) ((u8 *) opt0) + 1; - continue; - case 1: /* PadN */ - break; - default: - if (hm->options[type0]) - { - if ((*hm->options[type0]) (b0, opt0, 1 /* is_ipv4 */ , - use_adj) < 0) - { - *next0 = drop_node_val; - return; - } - } - break; - } - opt0 = - (vxlan_gpe_ioam_option_t *) (((u8 *) opt0) + opt0->length + - sizeof (vxlan_gpe_ioam_hdr_t)); - } - - - if (PREDICT_FALSE (b0->flags & VLIB_BUFFER_IS_TRACED)) - { - vxlan_gpe_ioam_v4_trace_t *t = - vlib_add_trace (vm, node, b0, sizeof (*t)); - u32 trace_len = gpe_ioam0->length; - t->fmt_trace.next_index = *next0; - /* Capture the ioam option verbatim */ - trace_len = - trace_len < - ARRAY_LEN (t->fmt_trace. - option_data) ? trace_len : ARRAY_LEN (t->fmt_trace. - option_data); - t->fmt_trace.trace_len = trace_len; - clib_memcpy (&(t->fmt_trace.option_data), gpe_ioam0, trace_len); - } - return; -} - - -#endif - -/* - * fd.io coding-style-patch-verification: ON - * - * Local Variables: - * eval: (c-set-style "gnu") - * End: - */ diff --git a/plugins/ioam-plugin/ioam/lib-vxlan-gpe/vxlan_gpe_msg_enum.h b/plugins/ioam-plugin/ioam/lib-vxlan-gpe/vxlan_gpe_msg_enum.h deleted file mode 100644 index cc0a10a3..00000000 --- a/plugins/ioam-plugin/ioam/lib-vxlan-gpe/vxlan_gpe_msg_enum.h +++ /dev/null @@ -1,28 +0,0 @@ -/* - * Copyright (c) 2016 Cisco and/or its affiliates. - * Licensed under the Apache License, Version 2.0 (the "License"); - * you may not use this file except in compliance with the License. - * You may obtain a copy of the License at: - * - * http://www.apache.org/licenses/LICENSE-2.0 - * - * Unless required by applicable law or agreed to in writing, software - * distributed under the License is distributed on an "AS IS" BASIS, - * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. - * See the License for the specific language governing permissions and - * limitations under the License. - */ -#ifndef included_vxlan_gpe_msg_enum_h -#define included_vxlan_gpe_msg_enum_h - -#include - -#define vl_msg_id(n,h) n, -typedef enum { -#include - /* We'll want to know how many messages IDs we need... */ - VL_MSG_FIRST_AVAILABLE, -} vl_msg_id_t; -#undef vl_msg_id - -#endif /* included_vxlan_gpe_msg_enum_h */ diff --git a/plugins/ioam-plugin/ioam/lib-vxlan-gpe/vxlan_gpe_test.c b/plugins/ioam-plugin/ioam/lib-vxlan-gpe/vxlan_gpe_test.c deleted file mode 100644 index 47253eb6..00000000 --- a/plugins/ioam-plugin/ioam/lib-vxlan-gpe/vxlan_gpe_test.c +++ /dev/null @@ -1,600 +0,0 @@ -/* - * Copyright (c) 2016 Cisco and/or its affiliates. - * Licensed under the Apache License, Version 2.0 (the "License"); - * you may not use this file except in compliance with the License. - * You may obtain a copy of the License at: - * - * http://www.apache.org/licenses/LICENSE-2.0 - * - * Unless required by applicable law or agreed to in writing, software - * distributed under the License is distributed on an "AS IS" BASIS, - * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. - * See the License for the specific language governing permissions and - * limitations under the License. - */ -/* - *------------------------------------------------------------------ - * vxlan_gpe_test.c - test harness for vxlan_gpe plugin - *------------------------------------------------------------------ - */ - -#include -#include -#include -#include -#include - -/* Declare message IDs */ -#include - -/* define message structures */ -#define vl_typedefs -#include -#undef vl_typedefs - -/* declare message handlers for each api */ - -#define vl_endianfun /* define message structures */ -#include -#undef vl_endianfun - -/* instantiate all the print functions we know about */ -#define vl_print(handle, ...) -#define vl_printfun -#include -#undef vl_printfun - -/* Get the API version number. */ -#define vl_api_version(n,v) static u32 api_version=(v); -#include -#undef vl_api_version -#include -#include - -typedef struct -{ - /* API message ID base */ - u16 msg_id_base; - vat_main_t *vat_main; -} vxlan_gpe_test_main_t; - -vxlan_gpe_test_main_t vxlan_gpe_test_main; - -#define foreach_standard_reply_retval_handler \ -_(vxlan_gpe_ioam_enable_reply) \ -_(vxlan_gpe_ioam_disable_reply) \ -_(vxlan_gpe_ioam_vni_enable_reply) \ -_(vxlan_gpe_ioam_vni_disable_reply) \ -_(vxlan_gpe_ioam_transit_enable_reply) \ -_(vxlan_gpe_ioam_transit_disable_reply) - -#define _(n) \ - static void vl_api_##n##_t_handler \ - (vl_api_##n##_t * mp) \ - { \ - vat_main_t * vam = vxlan_gpe_test_main.vat_main; \ - i32 retval = ntohl(mp->retval); \ - if (vam->async_mode) { \ - vam->async_errors += (retval < 0); \ - } else { \ - vam->retval = retval; \ - vam->result_ready = 1; \ - } \ - } -foreach_standard_reply_retval_handler; -#undef _ - -/* - * Table of message reply handlers, must include boilerplate handlers - * we just generated - */ -#define foreach_vpe_api_reply_msg \ -_(VXLAN_GPE_IOAM_ENABLE_REPLY, vxlan_gpe_ioam_enable_reply) \ -_(VXLAN_GPE_IOAM_DISABLE_REPLY, vxlan_gpe_ioam_disable_reply) \ -_(VXLAN_GPE_IOAM_VNI_ENABLE_REPLY, vxlan_gpe_ioam_vni_enable_reply) \ -_(VXLAN_GPE_IOAM_VNI_DISABLE_REPLY, vxlan_gpe_ioam_vni_disable_reply) \ -_(VXLAN_GPE_IOAM_TRANSIT_ENABLE_REPLY, vxlan_gpe_ioam_transit_enable_reply) \ -_(VXLAN_GPE_IOAM_TRANSIT_DISABLE_REPLY, vxlan_gpe_ioam_transit_disable_reply) \ - - -/* M: construct, but don't yet send a message */ - -#define M(T,t) \ -do { \ - vam->result_ready = 0; \ - mp = vl_msg_api_alloc(sizeof(*mp)); \ - memset (mp, 0, sizeof (*mp)); \ - mp->_vl_msg_id = ntohs (VL_API_##T + sm->msg_id_base); \ - mp->client_index = vam->my_client_index; \ -} while(0); - -#define M2(T,t,n) \ -do { \ - vam->result_ready = 0; \ - mp = vl_msg_api_alloc(sizeof(*mp)+(n)); \ - memset (mp, 0, sizeof (*mp)); \ - mp->_vl_msg_id = ntohs (VL_API_##T + sm->msg_id_base); \ - mp->client_index = vam->my_client_index; \ -} while(0); - -/* S: send a message */ -#define S (vl_msg_api_send_shmem (vam->vl_input_queue, (u8 *)&mp)) - -/* W: wait for results, with timeout */ -#define W \ -do { \ - timeout = vat_time_now (vam) + 1.0; \ - \ - while (vat_time_now (vam) < timeout) { \ - if (vam->result_ready == 1) { \ - return (vam->retval); \ - } \ - } \ - return -99; \ -} while(0); - - -static int -api_vxlan_gpe_ioam_enable (vat_main_t * vam) -{ - vxlan_gpe_test_main_t *sm = &vxlan_gpe_test_main; - - unformat_input_t *input = vam->input; - vl_api_vxlan_gpe_ioam_enable_t *mp; - f64 timeout; - u32 id = 0; - int has_trace_option = 0; - int has_pow_option = 0; - int has_ppc_option = 0; - - while (unformat_check_input (input) != UNFORMAT_END_OF_INPUT) - { - if (unformat (input, "trace")) - has_trace_option = 1; - else if (unformat (input, "pow")) - has_pow_option = 1; - else if (unformat (input, "ppc encap")) - has_ppc_option = PPC_ENCAP; - else if (unformat (input, "ppc decap")) - has_ppc_option = PPC_DECAP; - else if (unformat (input, "ppc none")) - has_ppc_option = PPC_NONE; - else - break; - } - M (VXLAN_GPE_IOAM_ENABLE, vxlan_gpe_ioam_enable); - mp->id = htons (id); - mp->trace_ppc = has_ppc_option; - mp->pow_enable = has_pow_option; - mp->trace_enable = has_trace_option; - - - S; - W; - - return (0); -} - - -static int -api_vxlan_gpe_ioam_disable (vat_main_t * vam) -{ - vxlan_gpe_test_main_t *sm = &vxlan_gpe_test_main; - vl_api_vxlan_gpe_ioam_disable_t *mp; - f64 timeout; - - M (VXLAN_GPE_IOAM_DISABLE, vxlan_gpe_ioam_disable); - S; - W; - return 0; -} - -static int -api_vxlan_gpe_ioam_vni_enable (vat_main_t * vam) -{ - vxlan_gpe_test_main_t *sm = &vxlan_gpe_test_main; - - unformat_input_t *line_input = vam->input; - vl_api_vxlan_gpe_ioam_vni_enable_t *mp; - ip4_address_t local4, remote4; - ip6_address_t local6, remote6; - u8 ipv4_set = 0, ipv6_set = 0; - u8 local_set = 0; - u8 remote_set = 0; - u32 vni; - u8 vni_set = 0; - f64 timeout; - - - while (unformat_check_input (line_input) != UNFORMAT_END_OF_INPUT) - { - if (unformat (line_input, "local %U", unformat_ip4_address, &local4)) - { - local_set = 1; - ipv4_set = 1; - } - else if (unformat (line_input, "remote %U", - unformat_ip4_address, &remote4)) - { - remote_set = 1; - ipv4_set = 1; - } - else if (unformat (line_input, "local %U", - unformat_ip6_address, &local6)) - { - local_set = 1; - ipv6_set = 1; - } - else if (unformat (line_input, "remote %U", - unformat_ip6_address, &remote6)) - { - remote_set = 1; - ipv6_set = 1; - } - - else if (unformat (line_input, "vni %d", &vni)) - vni_set = 1; - else - { - errmsg ("parse error '%U'\n", format_unformat_error, line_input); - return -99; - } - } - - if (local_set == 0) - { - errmsg ("tunnel local address not specified\n"); - return -99; - } - if (remote_set == 0) - { - errmsg ("tunnel remote address not specified\n"); - return -99; - } - if (ipv4_set && ipv6_set) - { - errmsg ("both IPv4 and IPv6 addresses specified"); - return -99; - } - - if (vni_set == 0) - { - errmsg ("vni not specified\n"); - return -99; - } - - M (VXLAN_GPE_IOAM_VNI_ENABLE, vxlan_gpe_ioam_vni_enable); - - - if (ipv6_set) - { - clib_memcpy (&mp->local, &local6, sizeof (local6)); - clib_memcpy (&mp->remote, &remote6, sizeof (remote6)); - } - else - { - clib_memcpy (&mp->local, &local4, sizeof (local4)); - clib_memcpy (&mp->remote, &remote4, sizeof (remote4)); - } - - mp->vni = ntohl (vni); - mp->is_ipv6 = ipv6_set; - - S; - W; - - return (0); -} - -static int -api_vxlan_gpe_ioam_vni_disable (vat_main_t * vam) -{ - vxlan_gpe_test_main_t *sm = &vxlan_gpe_test_main; - - unformat_input_t *line_input = vam->input; - vl_api_vxlan_gpe_ioam_vni_disable_t *mp; - ip4_address_t local4, remote4; - ip6_address_t local6, remote6; - u8 ipv4_set = 0, ipv6_set = 0; - u8 local_set = 0; - u8 remote_set = 0; - u32 vni; - u8 vni_set = 0; - f64 timeout; - - - while (unformat_check_input (line_input) != UNFORMAT_END_OF_INPUT) - { - if (unformat (line_input, "local %U", unformat_ip4_address, &local4)) - { - local_set = 1; - ipv4_set = 1; - } - else if (unformat (line_input, "remote %U", - unformat_ip4_address, &remote4)) - { - remote_set = 1; - ipv4_set = 1; - } - else if (unformat (line_input, "local %U", - unformat_ip6_address, &local6)) - { - local_set = 1; - ipv6_set = 1; - } - else if (unformat (line_input, "remote %U", - unformat_ip6_address, &remote6)) - { - remote_set = 1; - ipv6_set = 1; - } - - else if (unformat (line_input, "vni %d", &vni)) - vni_set = 1; - else - { - errmsg ("parse error '%U'\n", format_unformat_error, line_input); - return -99; - } - } - - if (local_set == 0) - { - errmsg ("tunnel local address not specified\n"); - return -99; - } - if (remote_set == 0) - { - errmsg ("tunnel remote address not specified\n"); - return -99; - } - if (ipv4_set && ipv6_set) - { - errmsg ("both IPv4 and IPv6 addresses specified"); - return -99; - } - - if (vni_set == 0) - { - errmsg ("vni not specified\n"); - return -99; - } - - M (VXLAN_GPE_IOAM_VNI_DISABLE, vxlan_gpe_ioam_vni_disable); - - - if (ipv6_set) - { - clib_memcpy (&mp->local, &local6, sizeof (local6)); - clib_memcpy (&mp->remote, &remote6, sizeof (remote6)); - } - else - { - clib_memcpy (&mp->local, &local4, sizeof (local4)); - clib_memcpy (&mp->remote, &remote4, sizeof (remote4)); - } - - mp->vni = ntohl (vni); - mp->is_ipv6 = ipv6_set; - - S; - W; - - return 0; -} - -static int -api_vxlan_gpe_ioam_transit_enable (vat_main_t * vam) -{ - vxlan_gpe_test_main_t *sm = &vxlan_gpe_test_main; - - unformat_input_t *line_input = vam->input; - vl_api_vxlan_gpe_ioam_transit_enable_t *mp; - ip4_address_t local4; - ip6_address_t local6; - u8 ipv4_set = 0, ipv6_set = 0; - u8 local_set = 0; - u32 outer_fib_index = 0; - f64 timeout; - - - while (unformat_check_input (line_input) != UNFORMAT_END_OF_INPUT) - { - if (unformat (line_input, "dst-ip %U", unformat_ip4_address, &local4)) - { - local_set = 1; - ipv4_set = 1; - } - else if (unformat (line_input, "dst-ip %U", - unformat_ip6_address, &local6)) - { - local_set = 1; - ipv6_set = 1; - } - - else if (unformat (line_input, "outer-fib-index %d", &outer_fib_index)) - ; - else - { - errmsg ("parse error '%U'\n", format_unformat_error, line_input); - return -99; - } - } - - if (local_set == 0) - { - errmsg ("destination address not specified\n"); - return -99; - } - if (ipv4_set && ipv6_set) - { - errmsg ("both IPv4 and IPv6 addresses specified"); - return -99; - } - - - M (VXLAN_GPE_IOAM_TRANSIT_ENABLE, vxlan_gpe_ioam_transit_enable); - - - if (ipv6_set) - { - errmsg ("IPv6 currently unsupported"); - return -1; - } - else - { - clib_memcpy (&mp->dst_addr, &local4, sizeof (local4)); - } - - mp->outer_fib_index = htonl (outer_fib_index); - mp->is_ipv6 = ipv6_set; - - S; - W; - - return (0); -} - -static int -api_vxlan_gpe_ioam_transit_disable (vat_main_t * vam) -{ - vxlan_gpe_test_main_t *sm = &vxlan_gpe_test_main; - - unformat_input_t *line_input = vam->input; - vl_api_vxlan_gpe_ioam_transit_disable_t *mp; - ip4_address_t local4; - ip6_address_t local6; - u8 ipv4_set = 0, ipv6_set = 0; - u8 local_set = 0; - u32 outer_fib_index; - f64 timeout; - - - while (unformat_check_input (line_input) != UNFORMAT_END_OF_INPUT) - { - if (unformat (line_input, "dst-ip %U", unformat_ip4_address, &local4)) - { - local_set = 1; - ipv4_set = 1; - } - else if (unformat (line_input, "dst-ip %U", - unformat_ip6_address, &local6)) - { - local_set = 1; - ipv6_set = 1; - } - - else if (unformat (line_input, "outer-fib-index %d", &outer_fib_index)) - ; - else - { - errmsg ("parse error '%U'\n", format_unformat_error, line_input); - return -99; - } - } - - if (local_set == 0) - { - errmsg ("destination address not specified\n"); - return -99; - } - if (ipv4_set && ipv6_set) - { - errmsg ("both IPv4 and IPv6 addresses specified"); - return -99; - } - - - M (VXLAN_GPE_IOAM_TRANSIT_DISABLE, vxlan_gpe_ioam_transit_disable); - - - if (ipv6_set) - { - return -1; - } - else - { - clib_memcpy (&mp->dst_addr, &local4, sizeof (local4)); - } - - mp->outer_fib_index = htonl (outer_fib_index); - mp->is_ipv6 = ipv6_set; - - S; - W; - - - return (0); -} - -/* - * List of messages that the api test plugin sends, - * and that the data plane plugin processes - */ -#define foreach_vpe_api_msg \ -_(vxlan_gpe_ioam_enable, ""\ - "[trace] [pow] [ppc ]") \ -_(vxlan_gpe_ioam_disable, "") \ -_(vxlan_gpe_ioam_vni_enable, ""\ - "local remote vni ") \ -_(vxlan_gpe_ioam_vni_disable, ""\ - "local remote vni ") \ -_(vxlan_gpe_ioam_transit_enable, ""\ - "dst-ip [outer-fib-index ]") \ -_(vxlan_gpe_ioam_transit_disable, ""\ - "dst-ip [outer-fib-index ]") \ - - -void -vat_api_hookup (vat_main_t * vam) -{ - vxlan_gpe_test_main_t *sm = &vxlan_gpe_test_main; - /* Hook up handlers for replies from the data plane plug-in */ -#define _(N,n) \ - vl_msg_api_set_handlers((VL_API_##N + sm->msg_id_base), \ - #n, \ - vl_api_##n##_t_handler, \ - vl_noop_handler, \ - vl_api_##n##_t_endian, \ - vl_api_##n##_t_print, \ - sizeof(vl_api_##n##_t), 1); - foreach_vpe_api_reply_msg; -#undef _ - - /* API messages we can send */ -#define _(n,h) hash_set_mem (vam->function_by_name, #n, api_##n); - foreach_vpe_api_msg; -#undef _ - - /* Help strings */ -#define _(n,h) hash_set_mem (vam->help_by_name, #n, h); - foreach_vpe_api_msg; -#undef _ -} - -clib_error_t * -vat_plugin_register (vat_main_t * vam) -{ - vxlan_gpe_test_main_t *sm = &vxlan_gpe_test_main; - u8 *name; - - sm->vat_main = vam; - - name = format (0, "ioam_vxlan_gpe_%08x%c", api_version, 0); - sm->msg_id_base = vl_client_get_first_plugin_msg_id ((char *) name); - - if (sm->msg_id_base != (u16) ~ 0) - vat_api_hookup (vam); - - vec_free (name); - - return 0; -} - -/* - * fd.io coding-style-patch-verification: ON - * - * Local Variables: - * eval: (c-set-style "gnu") - * End: - */ diff --git a/plugins/lb-plugin/Makefile.am b/plugins/lb-plugin/Makefile.am deleted file mode 100644 index 8e360279..00000000 --- a/plugins/lb-plugin/Makefile.am +++ /dev/null @@ -1,55 +0,0 @@ -# Copyright (c) 2016 Cisco Systems, Inc. -# Licensed under the Apache License, Version 2.0 (the "License"); -# you may not use this file except in compliance with the License. -# You may obtain a copy of the License at: -# -# http://www.apache.org/licenses/LICENSE-2.0 -# -# Unless required by applicable law or agreed to in writing, software -# distributed under the License is distributed on an "AS IS" BASIS, -# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. -# See the License for the specific language governing permissions and -# limitations under the License. - -AUTOMAKE_OPTIONS = foreign subdir-objects - -AM_CFLAGS = -Wall -AM_LDFLAGS = -module -shared -avoid-version - -vppapitestpluginsdir = ${libdir}/vpp_api_test_plugins -vpppluginsdir = ${libdir}/vpp_plugins - -vppapitestplugins_LTLIBRARIES = lb_test_plugin.la -vppplugins_LTLIBRARIES = lb_plugin.la - -lb_plugin_la_SOURCES = lb/lb.c lb/node.c lb/cli.c lb/util.c lb/refcount.c lb/api.c - -BUILT_SOURCES = \ - lb/lb.api.h \ - lb/lb.api.json - -SUFFIXES = .api.h .api .api.json - -%.api.h: %.api - mkdir -p `dirname $@` ; \ - $(CC) $(CPPFLAGS) -E -P -C -x c $^ \ - | vppapigen --input - --output $@ --show-name $@ - -%.api.json: %.api - @echo " JSON APIGEN " $@ ; \ - mkdir -p `dirname $@` ; \ - $(CC) $(CPPFLAGS) -E -P -C -x c $^ \ - | vppapigen --input - --json $@ - -apidir = $(prefix)/lb/ -api_DATA = lb/lb.api.json - -noinst_HEADERS = lb/lb.h lb/util.h lb/refcount.h lb/lbhash.h lb/lb.api.h - -lb_test_plugin_la_SOURCES = \ - lb/lb_test.c lb/lb_plugin.api.h - -# Remove *.la files -install-data-hook: - @(cd $(vpppluginsdir) && $(RM) $(vppplugins_LTLIBRARIES)) - @(cd $(vppapitestpluginsdir) && $(RM) $(vppapitestplugins_LTLIBRARIES)) diff --git a/plugins/lb-plugin/configure.ac b/plugins/lb-plugin/configure.ac deleted file mode 100644 index 1b02e54f..00000000 --- a/plugins/lb-plugin/configure.ac +++ /dev/null @@ -1,9 +0,0 @@ -AC_INIT(lb_plugin, 1.0) -AM_INIT_AUTOMAKE -AM_SILENT_RULES([yes]) -AC_PREFIX_DEFAULT([/usr]) - -AC_PROG_LIBTOOL -AC_PROG_CC - -AC_OUTPUT([Makefile]) diff --git a/plugins/lb-plugin/lb/api.c b/plugins/lb-plugin/lb/api.c deleted file mode 100644 index 06c53fa1..00000000 --- a/plugins/lb-plugin/lb/api.c +++ /dev/null @@ -1,228 +0,0 @@ -/* - * Copyright (c) 2016 Cisco and/or its affiliates. - * Licensed under the Apache License, Version 2.0 (the "License"); - * you may not use this file except in compliance with the License. - * You may obtain a copy of the License at: - * - * http://www.apache.org/licenses/LICENSE-2.0 - * - * Unless required by applicable law or agreed to in writing, software - * distributed under the License is distributed on an "AS IS" BASIS, - * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. - * See the License for the specific language governing permissions and - * limitations under the License. - */ - -#include - -#include -#include -#include -#include -#include - -#define vl_msg_id(n,h) n, -typedef enum { -#include - /* We'll want to know how many messages IDs we need... */ - VL_MSG_FIRST_AVAILABLE, -} vl_msg_id_t; -#undef vl_msg_id - - -/* define message structures */ -#define vl_typedefs -#include -#undef vl_typedefs - -/* define generated endian-swappers */ -#define vl_endianfun -#include -#undef vl_endianfun - -#define vl_print(handle, ...) vlib_cli_output (handle, __VA_ARGS__) - -/* Get the API version number */ -#define vl_api_version(n,v) static u32 api_version=(v); -#include -#undef vl_api_version - -#define vl_msg_name_crc_list -#include -#undef vl_msg_name_crc_list - -static void -setup_message_id_table (lb_main_t * lbm, api_main_t * am) -{ -#define _(id,n,crc) \ - vl_msg_api_add_msg_name_crc (am, #n "_" #crc, id + lbm->msg_id_base); - foreach_vl_msg_name_crc_lb; -#undef _ -} - -/* Macro to finish up custom dump fns */ -#define FINISH \ - vec_add1 (s, 0); \ - vl_print (handle, (char *)s); \ - vec_free (s); \ - return handle; - -/* - * A handy macro to set up a message reply. - * Assumes that the following variables are available: - * mp - pointer to request message - * rmp - pointer to reply message type - * rv - return value - */ - -#define REPLY_MACRO(t) \ -do { \ - unix_shared_memory_queue_t * q = \ - vl_api_client_index_to_input_queue (mp->client_index); \ - if (!q) \ - return; \ - \ - rmp = vl_msg_api_alloc (sizeof (*rmp)); \ - rmp->_vl_msg_id = ntohs((t)+lbm->msg_id_base); \ - rmp->context = mp->context; \ - rmp->retval = ntohl(rv); \ - \ - vl_msg_api_send_shmem (q, (u8 *)&rmp); \ -} while(0); - -static void -vl_api_lb_conf_t_handler -(vl_api_lb_conf_t * mp) -{ - lb_main_t *lbm = &lb_main; - vl_api_lb_conf_reply_t * rmp; - int rv = 0; - - rv = lb_conf((ip4_address_t *)&mp->ip4_src_address, - (ip6_address_t *)mp->ip6_src_address, - mp->sticky_buckets_per_core, - mp->flow_timeout); - - REPLY_MACRO (VL_API_LB_CONF_REPLY); -} - -static void *vl_api_lb_conf_t_print -(vl_api_lb_conf_t *mp, void * handle) -{ - u8 * s; - s = format (0, "SCRIPT: lb_conf "); - s = format (s, "%U ", format_ip4_address, (ip4_address_t *)&mp->ip4_src_address); - s = format (s, "%U ", format_ip6_address, (ip6_address_t *)mp->ip6_src_address); - s = format (s, "%u ", mp->sticky_buckets_per_core); - s = format (s, "%u ", mp->flow_timeout); - FINISH; -} - - -static void -vl_api_lb_add_del_vip_t_handler -(vl_api_lb_add_del_vip_t * mp) -{ - lb_main_t *lbm = &lb_main; - vl_api_lb_conf_reply_t * rmp; - int rv = 0; - ip46_address_t prefix; - memcpy(&prefix.ip6, mp->ip_prefix, sizeof(prefix.ip6)); - - if (mp->is_del) { - u32 vip_index; - if (!(rv = lb_vip_find_index(&prefix, mp->prefix_length, &vip_index))) - rv = lb_vip_del(vip_index); - } else { - u32 vip_index; - lb_vip_type_t type; - if (ip46_prefix_is_ip4(&prefix, mp->prefix_length)) { - type = mp->is_gre4?LB_VIP_TYPE_IP4_GRE4:LB_VIP_TYPE_IP4_GRE6; - } else { - type = mp->is_gre4?LB_VIP_TYPE_IP6_GRE4:LB_VIP_TYPE_IP6_GRE6; - } - - rv = lb_vip_add(&prefix, mp->prefix_length, type, - mp->new_flows_table_length, &vip_index); - } - REPLY_MACRO (VL_API_LB_CONF_REPLY); -} - -static void *vl_api_lb_add_del_vip_t_print -(vl_api_lb_add_del_vip_t *mp, void * handle) -{ - u8 * s; - s = format (0, "SCRIPT: lb_add_del_vip "); - s = format (s, "%U ", format_ip46_prefix, - (ip46_address_t *)mp->ip_prefix, mp->prefix_length, IP46_TYPE_ANY); - s = format (s, "%s ", mp->is_gre4?"gre4":"gre6"); - s = format (s, "%u ", mp->new_flows_table_length); - s = format (s, "%s ", mp->is_del?"del":"add"); - FINISH; -} - -static void -vl_api_lb_add_del_as_t_handler -(vl_api_lb_add_del_as_t * mp) -{ - lb_main_t *lbm = &lb_main; - vl_api_lb_conf_reply_t * rmp; - int rv = 0; - u32 vip_index; - if ((rv = lb_vip_find_index((ip46_address_t *)mp->vip_ip_prefix, - mp->vip_prefix_length, &vip_index))) - goto done; - - if (mp->is_del) - rv = lb_vip_del_ass(vip_index, (ip46_address_t *)mp->as_address, 1); - else - rv = lb_vip_add_ass(vip_index, (ip46_address_t *)mp->as_address, 1); - -done: - REPLY_MACRO (VL_API_LB_CONF_REPLY); -} - -static void *vl_api_lb_add_del_as_t_print -(vl_api_lb_add_del_as_t *mp, void * handle) -{ - u8 * s; - s = format (0, "SCRIPT: lb_add_del_as "); - s = format (s, "%U ", format_ip46_prefix, - (ip46_address_t *)mp->vip_ip_prefix, mp->vip_prefix_length, IP46_TYPE_ANY); - s = format (s, "%U ", format_ip46_address, - (ip46_address_t *)mp->as_address, IP46_TYPE_ANY); - s = format (s, "%s ", mp->is_del?"del":"add"); - FINISH; -} - -/* List of message types that this plugin understands */ -#define foreach_lb_plugin_api_msg \ -_(LB_CONF, lb_conf) \ -_(LB_ADD_DEL_VIP, lb_add_del_vip) \ -_(LB_ADD_DEL_AS, lb_add_del_as) - -static clib_error_t * lb_api_init (vlib_main_t * vm) -{ - lb_main_t *lbm = &lb_main; - u8 *name = format (0, "lb_%08x%c", api_version, 0); - lbm->msg_id_base = vl_msg_api_get_msg_ids - ((char *) name, VL_MSG_FIRST_AVAILABLE); - -#define _(N,n) \ - vl_msg_api_set_handlers((VL_API_##N + lbm->msg_id_base), \ - #n, \ - vl_api_##n##_t_handler, \ - vl_noop_handler, \ - vl_api_##n##_t_endian, \ - vl_api_##n##_t_print, \ - sizeof(vl_api_##n##_t), 1); - foreach_lb_plugin_api_msg; -#undef _ - - /* Add our API messages to the global name_crc hash table */ - setup_message_id_table (lbm, &api_main); - - return 0; -} - -VLIB_INIT_FUNCTION (lb_api_init); diff --git a/plugins/lb-plugin/lb/cli.c b/plugins/lb-plugin/lb/cli.c deleted file mode 100644 index b59c6426..00000000 --- a/plugins/lb-plugin/lb/cli.c +++ /dev/null @@ -1,250 +0,0 @@ -/* - * Copyright (c) 2016 Cisco and/or its affiliates. - * Licensed under the Apache License, Version 2.0 (the "License"); - * you may not use this file except in compliance with the License. - * You may obtain a copy of the License at: - * - * http://www.apache.org/licenses/LICENSE-2.0 - * - * Unless required by applicable law or agreed to in writing, software - * distributed under the License is distributed on an "AS IS" BASIS, - * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. - * See the License for the specific language governing permissions and - * limitations under the License. - */ - -#include -#include - -static clib_error_t * -lb_vip_command_fn (vlib_main_t * vm, - unformat_input_t * input, vlib_cli_command_t * cmd) -{ - unformat_input_t _line_input, *line_input = &_line_input; - ip46_address_t prefix; - u8 plen; - u32 new_len = 1024; - u8 del = 0; - int ret; - u32 gre4 = 0; - lb_vip_type_t type; - - if (!unformat_user (input, unformat_line_input, line_input)) - return 0; - - if (!unformat(line_input, "%U", unformat_ip46_prefix, &prefix, &plen, IP46_TYPE_ANY, &plen)) - return clib_error_return (0, "invalid vip prefix: '%U'", - format_unformat_error, line_input); - - while (unformat_check_input (line_input) != UNFORMAT_END_OF_INPUT) - { - if (unformat(line_input, "new_len %d", &new_len)) - ; - else if (unformat(line_input, "del")) - del = 1; - else if (unformat(line_input, "encap gre4")) - gre4 = 1; - else if (unformat(line_input, "encap gre6")) - gre4 = 0; - else - return clib_error_return (0, "parse error: '%U'", - format_unformat_error, line_input); - } - - unformat_free (line_input); - - - if (ip46_prefix_is_ip4(&prefix, plen)) { - type = (gre4)?LB_VIP_TYPE_IP4_GRE4:LB_VIP_TYPE_IP4_GRE6; - } else { - type = (gre4)?LB_VIP_TYPE_IP6_GRE4:LB_VIP_TYPE_IP6_GRE6; - } - - lb_garbage_collection(); - - u32 index; - if (!del) { - if ((ret = lb_vip_add(&prefix, plen, type, new_len, &index))) { - return clib_error_return (0, "lb_vip_add error %d", ret); - } else { - vlib_cli_output(vm, "lb_vip_add ok %d", index); - } - } else { - if ((ret = lb_vip_find_index(&prefix, plen, &index))) - return clib_error_return (0, "lb_vip_find_index error %d", ret); - else if ((ret = lb_vip_del(index))) - return clib_error_return (0, "lb_vip_del error %d", ret); - } - return NULL; -} - -VLIB_CLI_COMMAND (lb_vip_command, static) = -{ - .path = "lb vip", - .short_help = "lb vip [encap (gre6|gre4)] [new_len ] [del]", - .function = lb_vip_command_fn, -}; - -static clib_error_t * -lb_as_command_fn (vlib_main_t * vm, - unformat_input_t * input, vlib_cli_command_t * cmd) -{ - unformat_input_t _line_input, *line_input = &_line_input; - ip46_address_t vip_prefix, as_addr; - u8 vip_plen; - ip46_address_t *as_array = 0; - u32 vip_index; - u8 del = 0; - int ret; - - if (!unformat_user (input, unformat_line_input, line_input)) - return 0; - - if (!unformat(line_input, "%U", unformat_ip46_prefix, &vip_prefix, &vip_plen, IP46_TYPE_ANY)) - return clib_error_return (0, "invalid as address: '%U'", - format_unformat_error, line_input); - - if ((ret = lb_vip_find_index(&vip_prefix, vip_plen, &vip_index))) - return clib_error_return (0, "lb_vip_find_index error %d", ret); - - while (unformat_check_input (line_input) != UNFORMAT_END_OF_INPUT) - { - if (unformat(line_input, "%U", unformat_ip46_address, &as_addr, IP46_TYPE_ANY)) { - vec_add1(as_array, as_addr); - } else if (unformat(line_input, "del")) { - del = 1; - } else { - vec_free(as_array); - return clib_error_return (0, "parse error: '%U'", - format_unformat_error, line_input); - } - } - - if (!vec_len(as_array)) { - vec_free(as_array); - return clib_error_return (0, "No AS address provided"); - } - - lb_garbage_collection(); - clib_warning("vip index is %d", vip_index); - - if (del) { - if ((ret = lb_vip_del_ass(vip_index, as_array, vec_len(as_array)))) { - vec_free(as_array); - return clib_error_return (0, "lb_vip_del_ass error %d", ret); - } - } else { - if ((ret = lb_vip_add_ass(vip_index, as_array, vec_len(as_array)))) { - vec_free(as_array); - return clib_error_return (0, "lb_vip_add_ass error %d", ret); - } - } - - vec_free(as_array); - return 0; -} - -VLIB_CLI_COMMAND (lb_as_command, static) = -{ - .path = "lb as", - .short_help = "lb as [
[
[...]]] [del]", - .function = lb_as_command_fn, -}; - -static clib_error_t * -lb_conf_command_fn (vlib_main_t * vm, - unformat_input_t * input, vlib_cli_command_t * cmd) -{ - lb_main_t *lbm = &lb_main; - unformat_input_t _line_input, *line_input = &_line_input; - ip4_address_t ip4 = lbm->ip4_src_address; - ip6_address_t ip6 = lbm->ip6_src_address; - u32 per_cpu_sticky_buckets = lbm->per_cpu_sticky_buckets; - u32 per_cpu_sticky_buckets_log2 = 0; - u32 flow_timeout = lbm->flow_timeout; - int ret; - - if (!unformat_user (input, unformat_line_input, line_input)) - return 0; - - while (unformat_check_input (line_input) != UNFORMAT_END_OF_INPUT) - { - if (unformat(line_input, "ip4-src-address %U", unformat_ip4_address, &ip4)) - ; - else if (unformat(line_input, "ip6-src-address %U", unformat_ip6_address, &ip6)) - ; - else if (unformat(line_input, "buckets %d", &per_cpu_sticky_buckets)) - ; - else if (unformat(line_input, "buckets-log2 %d", &per_cpu_sticky_buckets_log2)) { - if (per_cpu_sticky_buckets_log2 >= 32) - return clib_error_return (0, "buckets-log2 value is too high"); - per_cpu_sticky_buckets = 1 << per_cpu_sticky_buckets_log2; - } else if (unformat(line_input, "timeout %d", &flow_timeout)) - ; - else - return clib_error_return (0, "parse error: '%U'", - format_unformat_error, line_input); - } - - unformat_free (line_input); - - lb_garbage_collection(); - - if ((ret = lb_conf(&ip4, &ip6, per_cpu_sticky_buckets, flow_timeout))) - return clib_error_return (0, "lb_conf error %d", ret); - - return NULL; -} - -VLIB_CLI_COMMAND (lb_conf_command, static) = -{ - .path = "lb conf", - .short_help = "lb conf [ip4-src-address ] [ip6-src-address ] [buckets ] [timeout ]", - .function = lb_conf_command_fn, -}; - -static clib_error_t * -lb_show_command_fn (vlib_main_t * vm, - unformat_input_t * input, vlib_cli_command_t * cmd) -{ - vlib_cli_output(vm, "%U", format_lb_main); - return NULL; -} - - -VLIB_CLI_COMMAND (lb_show_command, static) = -{ - .path = "show lb", - .short_help = "show lb", - .function = lb_show_command_fn, -}; - -static clib_error_t * -lb_show_vips_command_fn (vlib_main_t * vm, - unformat_input_t * input, vlib_cli_command_t * cmd) -{ - unformat_input_t line_input; - lb_main_t *lbm = &lb_main; - lb_vip_t *vip; - u8 verbose = 0; - - if (!unformat_user (input, unformat_line_input, &line_input)) - return 0; - - if (unformat(&line_input, "verbose")) - verbose = 1; - - pool_foreach(vip, lbm->vips, { - vlib_cli_output(vm, "%U\n", verbose?format_lb_vip_detailed:format_lb_vip, vip); - }); - - unformat_free (&line_input); - return NULL; -} - -VLIB_CLI_COMMAND (lb_show_vips_command, static) = -{ - .path = "show lb vips", - .short_help = "show lb vips [verbose]", - .function = lb_show_vips_command_fn, -}; diff --git a/plugins/lb-plugin/lb/lb.api b/plugins/lb-plugin/lb/lb.api deleted file mode 100644 index 39ee3c8f..00000000 --- a/plugins/lb-plugin/lb/lb.api +++ /dev/null @@ -1,71 +0,0 @@ -/** \brief Configure Load-Balancer global parameters - @param client_index - opaque cookie to identify the sender - @param context - sender context, to match reply w/ request - @param ip4_src_address - IPv4 address to be used as source for IPv4 GRE traffic. - @param ip6_src_address - IPv6 address to be used as source for IPv6 GRE traffic. - @param n_sticky_buckets - Number of buckets *per worker thread* in the - established flow table (must be power of 2). - @param flow_timeout - Time in seconds after which, if no packet is received - for a given flow, the flow is removed from the established flow table. -*/ -define lb_conf -{ - u32 client_index; - u32 context; - u32 ip4_src_address; - u8 ip6_src_address[16]; - u32 sticky_buckets_per_core; - u32 flow_timeout; -}; - -define lb_conf_reply { - u32 context; - i32 retval; -}; - -/** \brief Add a virtual address (or prefix) - @param client_index - opaque cookie to identify the sender - @param context - sender context, to match reply w/ request - @param ip_prefix - IP address (IPv4 in lower order 32 bits). - @param prefix_length - IP prefix length (96 + 'IPv4 prefix length' for IPv4). - @param is_gre4 - Encap is ip4 GRE (ip6 GRE otherwise). - @param new_flows_table_length - Size of the new connections flow table used - for this VIP (must be power of 2). - @param is_del - The VIP should be removed. -*/ -define lb_add_del_vip { - u32 client_index; - u32 context; - u8 ip_prefix[16]; - u8 prefix_length; - u8 is_gre4; - u32 new_flows_table_length; - u8 is_del; -}; - -define lb_add_del_vip_reply { - u32 context; - i32 retval; -}; - -/** \brief Add an application server for a given VIP - @param client_index - opaque cookie to identify the sender - @param context - sender context, to match reply w/ request - @param vip_ip_prefix - VIP IP address (IPv4 in lower order 32 bits). - @param vip_ip_prefix - VIP IP prefix length (96 + 'IPv4 prefix length' for IPv4). - @param as_address - The application server address (IPv4 in lower order 32 bits). - @param is_del - The AS should be removed. -*/ -define lb_add_del_as { - u32 client_index; - u32 context; - u8 vip_ip_prefix[16]; - u8 vip_prefix_length; - u8 as_address[16]; - u8 is_del; -}; - -define lb_add_del_as_reply { - u32 context; - i32 retval; -}; diff --git a/plugins/lb-plugin/lb/lb.c b/plugins/lb-plugin/lb/lb.c deleted file mode 100644 index 1d9b9870..00000000 --- a/plugins/lb-plugin/lb/lb.c +++ /dev/null @@ -1,844 +0,0 @@ -/* - * Copyright (c) 2016 Cisco and/or its affiliates. - * Licensed under the Apache License, Version 2.0 (the "License"); - * you may not use this file except in compliance with the License. - * You may obtain a copy of the License at: - * - * http://www.apache.org/licenses/LICENSE-2.0 - * - * Unless required by applicable law or agreed to in writing, software - * distributed under the License is distributed on an "AS IS" BASIS, - * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. - * See the License for the specific language governing permissions and - * limitations under the License. - */ - -#include -#include -#include - -//GC runs at most once every so many seconds -#define LB_GARBAGE_RUN 60 - -//After so many seconds. It is assumed that inter-core race condition will not occur. -#define LB_CONCURRENCY_TIMEOUT 10 - -lb_main_t lb_main; - -#define lb_get_writer_lock() do {} while(__sync_lock_test_and_set (lb_main.writer_lock, 1)) -#define lb_put_writer_lock() lb_main.writer_lock[0] = 0 - -static void lb_as_stack (lb_as_t *as); - - -const static char * const lb_dpo_gre4_ip4[] = { "lb4-gre4" , NULL }; -const static char * const lb_dpo_gre4_ip6[] = { "lb6-gre4" , NULL }; -const static char* const * const lb_dpo_gre4_nodes[DPO_PROTO_NUM] = - { - [DPO_PROTO_IP4] = lb_dpo_gre4_ip4, - [DPO_PROTO_IP6] = lb_dpo_gre4_ip6, - }; - -const static char * const lb_dpo_gre6_ip4[] = { "lb4-gre6" , NULL }; -const static char * const lb_dpo_gre6_ip6[] = { "lb6-gre6" , NULL }; -const static char* const * const lb_dpo_gre6_nodes[DPO_PROTO_NUM] = - { - [DPO_PROTO_IP4] = lb_dpo_gre6_ip4, - [DPO_PROTO_IP6] = lb_dpo_gre6_ip6, - }; - -u32 lb_hash_time_now(vlib_main_t * vm) -{ - return (u32) (vlib_time_now(vm) + 10000); -} - -u8 *format_lb_main (u8 * s, va_list * args) -{ - vlib_thread_main_t *tm = vlib_get_thread_main(); - lb_main_t *lbm = &lb_main; - s = format(s, "lb_main"); - s = format(s, " ip4-src-address: %U \n", format_ip4_address, &lbm->ip4_src_address); - s = format(s, " ip6-src-address: %U \n", format_ip6_address, &lbm->ip6_src_address); - s = format(s, " #vips: %u\n", pool_elts(lbm->vips)); - s = format(s, " #ass: %u\n", pool_elts(lbm->ass) - 1); - - u32 cpu_index; - for(cpu_index = 0; cpu_index < tm->n_vlib_mains; cpu_index++ ) { - lb_hash_t *h = lbm->per_cpu[cpu_index].sticky_ht; - if (h) { - s = format(s, "core %d\n", cpu_index); - s = format(s, " timeout: %ds\n", h->timeout); - s = format(s, " usage: %d / %d\n", lb_hash_elts(h, lb_hash_time_now(vlib_get_main())), lb_hash_size(h)); - } - } - - return s; -} - -static char *lb_vip_type_strings[] = { - [LB_VIP_TYPE_IP6_GRE6] = "ip6-gre6", - [LB_VIP_TYPE_IP6_GRE4] = "ip6-gre4", - [LB_VIP_TYPE_IP4_GRE6] = "ip4-gre6", - [LB_VIP_TYPE_IP4_GRE4] = "ip4-gre4", -}; - -u8 *format_lb_vip_type (u8 * s, va_list * args) -{ - lb_vip_type_t vipt = va_arg (*args, lb_vip_type_t); - u32 i; - for (i=0; itype, - format_ip46_prefix, &vip->prefix, vip->plen, IP46_TYPE_ANY, - vip->new_flow_table_mask + 1, - pool_elts(vip->as_indexes), - (vip->flags & LB_VIP_FLAGS_USED)?"":" removed"); -} - -u8 *format_lb_as (u8 * s, va_list * args) -{ - lb_as_t *as = va_arg (*args, lb_as_t *); - return format(s, "%U %s", format_ip46_address, - &as->address, IP46_TYPE_ANY, - (as->flags & LB_AS_FLAGS_USED)?"used":"removed"); -} - -u8 *format_lb_vip_detailed (u8 * s, va_list * args) -{ - lb_main_t *lbm = &lb_main; - lb_vip_t *vip = va_arg (*args, lb_vip_t *); - uword indent = format_get_indent (s); - - s = format(s, "%U %U [%u] %U%s\n" - "%U new_size:%u\n", - format_white_space, indent, - format_lb_vip_type, vip->type, - vip - lbm->vips, format_ip46_prefix, &vip->prefix, vip->plen, IP46_TYPE_ANY, - (vip->flags & LB_VIP_FLAGS_USED)?"":" removed", - format_white_space, indent, - vip->new_flow_table_mask + 1); - - //Print counters - s = format(s, "%U counters:\n", - format_white_space, indent); - u32 i; - for (i=0; ivip_counters[i].name, - vlib_get_simple_counter(&lbm->vip_counters[i], vip - lbm->vips)); - - - s = format(s, "%U #as:%u\n", - format_white_space, indent, - pool_elts(vip->as_indexes)); - - //Let's count the buckets for each AS - u32 *count = 0; - vec_validate(count, pool_len(lbm->ass)); //Possibly big alloc for not much... - lb_new_flow_entry_t *nfe; - vec_foreach(nfe, vip->new_flow_table) - count[nfe->as_index]++; - - lb_as_t *as; - u32 *as_index; - pool_foreach(as_index, vip->as_indexes, { - as = &lbm->ass[*as_index]; - s = format(s, "%U %U %d buckets %d flows dpo:%u %s\n", - format_white_space, indent, - format_ip46_address, &as->address, IP46_TYPE_ANY, - count[as - lbm->ass], - vlib_refcount_get(&lbm->as_refcount, as - lbm->ass), - as->dpo.dpoi_index, - (as->flags & LB_AS_FLAGS_USED)?"used":" removed"); - }); - - vec_free(count); - - /* - s = format(s, "%U new flows table:\n", format_white_space, indent); - lb_new_flow_entry_t *nfe; - vec_foreach(nfe, vip->new_flow_table) { - s = format(s, "%U %d: %d\n", format_white_space, indent, nfe - vip->new_flow_table, nfe->as_index); - } - */ - return s; -} - -typedef struct { - u32 as_index; - u32 last; - u32 skip; -} lb_pseudorand_t; - -static int lb_pseudorand_compare(void *a, void *b) -{ - lb_as_t *asa, *asb; - lb_main_t *lbm = &lb_main; - asa = &lbm->ass[((lb_pseudorand_t *)a)->as_index]; - asb = &lbm->ass[((lb_pseudorand_t *)b)->as_index]; - return memcmp(&asa->address, &asb->address, sizeof(asb->address)); -} - -static void lb_vip_garbage_collection(lb_vip_t *vip) -{ - lb_main_t *lbm = &lb_main; - ASSERT (lbm->writer_lock[0]); - - u32 now = (u32) vlib_time_now(vlib_get_main()); - if (!clib_u32_loop_gt(now, vip->last_garbage_collection + LB_GARBAGE_RUN)) - return; - - vip->last_garbage_collection = now; - lb_as_t *as; - u32 *as_index; - pool_foreach(as_index, vip->as_indexes, { - as = &lbm->ass[*as_index]; - if (!(as->flags & LB_AS_FLAGS_USED) && //Not used - clib_u32_loop_gt(now, as->last_used + LB_CONCURRENCY_TIMEOUT) && //Not recently used - (vlib_refcount_get(&lbm->as_refcount, as - lbm->ass) == 0)) - { //Not referenced - fib_entry_child_remove(as->next_hop_fib_entry_index, - as->next_hop_child_index); - fib_table_entry_delete_index(as->next_hop_fib_entry_index, - FIB_SOURCE_RR); - as->next_hop_fib_entry_index = FIB_NODE_INDEX_INVALID; - - pool_put(vip->as_indexes, as_index); - pool_put(lbm->ass, as); - } - }); -} - -void lb_garbage_collection() -{ - lb_main_t *lbm = &lb_main; - lb_get_writer_lock(); - lb_vip_t *vip; - u32 *to_be_removed_vips = 0, *i; - pool_foreach(vip, lbm->vips, { - lb_vip_garbage_collection(vip); - - if (!(vip->flags & LB_VIP_FLAGS_USED) && - (pool_elts(vip->as_indexes) == 0)) { - vec_add1(to_be_removed_vips, vip - lbm->vips); - } - }); - - vec_foreach(i, to_be_removed_vips) { - vip = &lbm->vips[*i]; - pool_put(lbm->vips, vip); - pool_free(vip->as_indexes); - } - - vec_free(to_be_removed_vips); - lb_put_writer_lock(); -} - -static void lb_vip_update_new_flow_table(lb_vip_t *vip) -{ - lb_main_t *lbm = &lb_main; - lb_new_flow_entry_t *old_table; - u32 i, *as_index; - lb_new_flow_entry_t *new_flow_table = 0; - lb_as_t *as; - lb_pseudorand_t *pr, *sort_arr = 0; - u32 count; - - ASSERT (lbm->writer_lock[0]); //We must have the lock - - //Check if some AS is configured or not - i = 0; - pool_foreach(as_index, vip->as_indexes, { - as = &lbm->ass[*as_index]; - if (as->flags & LB_AS_FLAGS_USED) { //Not used anymore - i = 1; - goto out; //Not sure 'break' works in this macro-loop - } - }); - -out: - if (i == 0) { - //Only the default. i.e. no AS - vec_validate(new_flow_table, vip->new_flow_table_mask); - for (i=0; ias_indexes)); - - i = 0; - pool_foreach(as_index, vip->as_indexes, { - as = &lbm->ass[*as_index]; - if (!(as->flags & LB_AS_FLAGS_USED)) //Not used anymore - continue; - - sort_arr[i].as_index = as - lbm->ass; - i++; - }); - _vec_len(sort_arr) = i; - - vec_sort_with_function(sort_arr, lb_pseudorand_compare); - - //Now let's pseudo-randomly generate permutations - vec_foreach(pr, sort_arr) { - lb_as_t *as = &lbm->ass[pr->as_index]; - - u64 seed = clib_xxhash(as->address.as_u64[0] ^ - as->address.as_u64[1]); - /* We have 2^n buckets. - * skip must be prime with 2^n. - * So skip must be odd. - * MagLev actually state that M should be prime, - * but this has a big computation cost (% operation). - * Using 2^n is more better (& operation). - */ - pr->skip = ((seed & 0xffffffff) | 1) & vip->new_flow_table_mask; - pr->last = (seed >> 32) & vip->new_flow_table_mask; - } - - //Let's create a new flow table - vec_validate(new_flow_table, vip->new_flow_table_mask); - for (i=0; ilast; - pr->last = (pr->last + pr->skip) & vip->new_flow_table_mask; - if (new_flow_table[last].as_index == ~0) { - new_flow_table[last].as_index = pr->as_index; - break; - } - } - done++; - if (done == vec_len(new_flow_table)) - goto finished; - } - } - - vec_free(sort_arr); - -finished: - -//Count number of changed entries - count = 0; - for (i=0; inew_flow_table == 0 || - new_flow_table[i].as_index != vip->new_flow_table[i].as_index) - count++; - - old_table = vip->new_flow_table; - vip->new_flow_table = new_flow_table; - vec_free(old_table); -} - -int lb_conf(ip4_address_t *ip4_address, ip6_address_t *ip6_address, - u32 per_cpu_sticky_buckets, u32 flow_timeout) -{ - lb_main_t *lbm = &lb_main; - - if (!is_pow2(per_cpu_sticky_buckets)) - return VNET_API_ERROR_INVALID_MEMORY_SIZE; - - lb_get_writer_lock(); //Not exactly necessary but just a reminder that it exists for my future self - lbm->ip4_src_address = *ip4_address; - lbm->ip6_src_address = *ip6_address; - lbm->per_cpu_sticky_buckets = per_cpu_sticky_buckets; - lbm->flow_timeout = flow_timeout; - lb_put_writer_lock(); - return 0; -} - -static -int lb_vip_find_index_with_lock(ip46_address_t *prefix, u8 plen, u32 *vip_index) -{ - lb_main_t *lbm = &lb_main; - lb_vip_t *vip; - ASSERT (lbm->writer_lock[0]); //This must be called with the lock owned - ip46_prefix_normalize(prefix, plen); - pool_foreach(vip, lbm->vips, { - if ((vip->flags & LB_AS_FLAGS_USED) && - vip->plen == plen && - vip->prefix.as_u64[0] == prefix->as_u64[0] && - vip->prefix.as_u64[1] == prefix->as_u64[1]) { - *vip_index = vip - lbm->vips; - return 0; - } - }); - return VNET_API_ERROR_NO_SUCH_ENTRY; -} - -int lb_vip_find_index(ip46_address_t *prefix, u8 plen, u32 *vip_index) -{ - int ret; - lb_get_writer_lock(); - ret = lb_vip_find_index_with_lock(prefix, plen, vip_index); - lb_put_writer_lock(); - return ret; -} - -static int lb_as_find_index_vip(lb_vip_t *vip, ip46_address_t *address, u32 *as_index) -{ - lb_main_t *lbm = &lb_main; - ASSERT (lbm->writer_lock[0]); //This must be called with the lock owned - lb_as_t *as; - u32 *asi; - pool_foreach(asi, vip->as_indexes, { - as = &lbm->ass[*asi]; - if (as->vip_index == (vip - lbm->vips) && - as->address.as_u64[0] == address->as_u64[0] && - as->address.as_u64[1] == address->as_u64[1]) { - *as_index = as - lbm->ass; - return 0; - } - }); - return -1; -} - -int lb_vip_add_ass(u32 vip_index, ip46_address_t *addresses, u32 n) -{ - lb_main_t *lbm = &lb_main; - lb_get_writer_lock(); - lb_vip_t *vip; - if (!(vip = lb_vip_get_by_index(vip_index))) { - lb_put_writer_lock(); - return VNET_API_ERROR_NO_SUCH_ENTRY; - } - - ip46_type_t type = lb_vip_is_gre4(vip)?IP46_TYPE_IP4:IP46_TYPE_IP6; - u32 *to_be_added = 0; - u32 *to_be_updated = 0; - u32 i; - u32 *ip; - - //Sanity check - while (n--) { - - if (!lb_as_find_index_vip(vip, &addresses[n], &i)) { - if (lbm->ass[i].flags & LB_AS_FLAGS_USED) { - vec_free(to_be_added); - vec_free(to_be_updated); - lb_put_writer_lock(); - return VNET_API_ERROR_VALUE_EXIST; - } - vec_add1(to_be_updated, i); - goto next; - } - - if (ip46_address_type(&addresses[n]) != type) { - vec_free(to_be_added); - vec_free(to_be_updated); - lb_put_writer_lock(); - return VNET_API_ERROR_INVALID_ADDRESS_FAMILY; - } - - if (n) { - u32 n2 = n; - while(n2--) //Check for duplicates - if (addresses[n2].as_u64[0] == addresses[n].as_u64[0] && - addresses[n2].as_u64[1] == addresses[n].as_u64[1]) - goto next; - } - - vec_add1(to_be_added, n); - -next: - continue; - } - - //Update reused ASs - vec_foreach(ip, to_be_updated) { - lbm->ass[*ip].flags = LB_AS_FLAGS_USED; - } - vec_free(to_be_updated); - - //Create those who have to be created - vec_foreach(ip, to_be_added) { - lb_as_t *as; - u32 *as_index; - pool_get(lbm->ass, as); - as->address = addresses[*ip]; - as->flags = LB_AS_FLAGS_USED; - as->vip_index = vip_index; - pool_get(vip->as_indexes, as_index); - *as_index = as - lbm->ass; - - /* - * become a child of the FIB entry - * so we are informed when its forwarding changes - */ - fib_prefix_t nh = {}; - if (lb_vip_is_gre4(vip)) { - nh.fp_addr.ip4 = as->address.ip4; - nh.fp_len = 32; - nh.fp_proto = FIB_PROTOCOL_IP4; - } else { - nh.fp_addr.ip6 = as->address.ip6; - nh.fp_len = 128; - nh.fp_proto = FIB_PROTOCOL_IP6; - } - - as->next_hop_fib_entry_index = - fib_table_entry_special_add(0, - &nh, - FIB_SOURCE_RR, - FIB_ENTRY_FLAG_NONE, - ADJ_INDEX_INVALID); - as->next_hop_child_index = - fib_entry_child_add(as->next_hop_fib_entry_index, - lbm->fib_node_type, - as - lbm->ass); - - lb_as_stack(as); - } - vec_free(to_be_added); - - //Recompute flows - lb_vip_update_new_flow_table(vip); - - //Garbage collection maybe - lb_vip_garbage_collection(vip); - - lb_put_writer_lock(); - return 0; -} - -int lb_vip_del_ass_withlock(u32 vip_index, ip46_address_t *addresses, u32 n) -{ - lb_main_t *lbm = &lb_main; - u32 now = (u32) vlib_time_now(vlib_get_main()); - u32 *ip = 0; - - lb_vip_t *vip; - if (!(vip = lb_vip_get_by_index(vip_index))) { - return VNET_API_ERROR_NO_SUCH_ENTRY; - } - - u32 *indexes = NULL; - while (n--) { - u32 i; - if (lb_as_find_index_vip(vip, &addresses[n], &i)) { - vec_free(indexes); - return VNET_API_ERROR_NO_SUCH_ENTRY; - } - - if (n) { //Check for duplicates - u32 n2 = n - 1; - while(n2--) { - if (addresses[n2].as_u64[0] == addresses[n].as_u64[0] && - addresses[n2].as_u64[1] == addresses[n].as_u64[1]) - goto next; - } - } - - vec_add1(indexes, i); -next: - continue; - } - - //Garbage collection maybe - lb_vip_garbage_collection(vip); - - if (indexes != NULL) { - vec_foreach(ip, indexes) { - lbm->ass[*ip].flags &= ~LB_AS_FLAGS_USED; - lbm->ass[*ip].last_used = now; - } - - //Recompute flows - lb_vip_update_new_flow_table(vip); - } - - vec_free(indexes); - return 0; -} - -int lb_vip_del_ass(u32 vip_index, ip46_address_t *addresses, u32 n) -{ - lb_get_writer_lock(); - int ret = lb_vip_del_ass_withlock(vip_index, addresses, n); - lb_put_writer_lock(); - return ret; -} - -/** - * Add the VIP adjacency to the ip4 or ip6 fib - */ -static void lb_vip_add_adjacency(lb_main_t *lbm, lb_vip_t *vip) -{ - dpo_proto_t proto = 0; - dpo_id_t dpo = DPO_INVALID; - fib_prefix_t pfx = {}; - if (lb_vip_is_ip4(vip)) { - pfx.fp_addr.ip4 = vip->prefix.ip4; - pfx.fp_len = vip->plen - 96; - pfx.fp_proto = FIB_PROTOCOL_IP4; - proto = DPO_PROTO_IP4; - } else { - pfx.fp_addr.ip6 = vip->prefix.ip6; - pfx.fp_len = vip->plen; - pfx.fp_proto = FIB_PROTOCOL_IP6; - proto = DPO_PROTO_IP6; - } - dpo_set(&dpo, lb_vip_is_gre4(vip)?lbm->dpo_gre4_type:lbm->dpo_gre6_type, - proto, vip - lbm->vips); - fib_table_entry_special_dpo_add(0, - &pfx, - FIB_SOURCE_PLUGIN_HI, - FIB_ENTRY_FLAG_EXCLUSIVE, - &dpo); - dpo_reset(&dpo); -} - -/** - * Deletes the adjacency associated with the VIP - */ -static void lb_vip_del_adjacency(lb_main_t *lbm, lb_vip_t *vip) -{ - fib_prefix_t pfx = {}; - if (lb_vip_is_ip4(vip)) { - pfx.fp_addr.ip4 = vip->prefix.ip4; - pfx.fp_len = vip->plen - 96; - pfx.fp_proto = FIB_PROTOCOL_IP4; - } else { - pfx.fp_addr.ip6 = vip->prefix.ip6; - pfx.fp_len = vip->plen; - pfx.fp_proto = FIB_PROTOCOL_IP6; - } - fib_table_entry_special_remove(0, &pfx, FIB_SOURCE_PLUGIN_HI); -} - -int lb_vip_add(ip46_address_t *prefix, u8 plen, lb_vip_type_t type, u32 new_length, u32 *vip_index) -{ - lb_main_t *lbm = &lb_main; - lb_vip_t *vip; - lb_get_writer_lock(); - ip46_prefix_normalize(prefix, plen); - - if (!lb_vip_find_index_with_lock(prefix, plen, vip_index)) { - lb_put_writer_lock(); - return VNET_API_ERROR_VALUE_EXIST; - } - - if (!is_pow2(new_length)) { - lb_put_writer_lock(); - return VNET_API_ERROR_INVALID_MEMORY_SIZE; - } - - if (ip46_prefix_is_ip4(prefix, plen) && - (type != LB_VIP_TYPE_IP4_GRE4) && - (type != LB_VIP_TYPE_IP4_GRE6)) - return VNET_API_ERROR_INVALID_ADDRESS_FAMILY; - - - //Allocate - pool_get(lbm->vips, vip); - - //Init - vip->prefix = *prefix; - vip->plen = plen; - vip->last_garbage_collection = (u32) vlib_time_now(vlib_get_main()); - vip->type = type; - vip->flags = LB_VIP_FLAGS_USED; - vip->as_indexes = 0; - - //Validate counters - u32 i; - for (i = 0; i < LB_N_VIP_COUNTERS; i++) { - vlib_validate_simple_counter(&lbm->vip_counters[i], vip - lbm->vips); - vlib_zero_simple_counter(&lbm->vip_counters[i], vip - lbm->vips); - } - - //Configure new flow table - vip->new_flow_table_mask = new_length - 1; - vip->new_flow_table = 0; - - //Create a new flow hash table full of the default entry - lb_vip_update_new_flow_table(vip); - - //Create adjacency to direct traffic - lb_vip_add_adjacency(lbm, vip); - - //Return result - *vip_index = vip - lbm->vips; - - lb_put_writer_lock(); - return 0; -} - -int lb_vip_del(u32 vip_index) -{ - lb_main_t *lbm = &lb_main; - lb_vip_t *vip; - lb_get_writer_lock(); - if (!(vip = lb_vip_get_by_index(vip_index))) { - lb_put_writer_lock(); - return VNET_API_ERROR_NO_SUCH_ENTRY; - } - - //FIXME: This operation is actually not working - //We will need to remove state before performing this. - - { - //Remove all ASs - ip46_address_t *ass = 0; - lb_as_t *as; - u32 *as_index; - pool_foreach(as_index, vip->as_indexes, { - as = &lbm->ass[*as_index]; - vec_add1(ass, as->address); - }); - if (vec_len(ass)) - lb_vip_del_ass_withlock(vip_index, ass, vec_len(ass)); - vec_free(ass); - } - - //Delete adjacency - lb_vip_del_adjacency(lbm, vip); - - //Set the VIP as unused - vip->flags &= ~LB_VIP_FLAGS_USED; - - lb_put_writer_lock(); - return 0; -} - -clib_error_t * -vlib_plugin_register (vlib_main_t * vm, - vnet_plugin_handoff_t * h, - int from_early_init) -{ - clib_error_t *error = 0; - return error; -} - - -u8 *format_lb_dpo (u8 * s, va_list * va) -{ - index_t index = va_arg (*va, index_t); - CLIB_UNUSED(u32 indent) = va_arg (*va, u32); - lb_main_t *lbm = &lb_main; - lb_vip_t *vip = pool_elt_at_index (lbm->vips, index); - return format (s, "%U", format_lb_vip, vip); -} - -static void lb_dpo_lock (dpo_id_t *dpo) {} -static void lb_dpo_unlock (dpo_id_t *dpo) {} - -static fib_node_t * -lb_fib_node_get_node (fib_node_index_t index) -{ - lb_main_t *lbm = &lb_main; - lb_as_t *as = pool_elt_at_index (lbm->ass, index); - return (&as->fib_node); -} - -static void -lb_fib_node_last_lock_gone (fib_node_t *node) -{ -} - -static lb_as_t * -lb_as_from_fib_node (fib_node_t *node) -{ - return ((lb_as_t*)(((char*)node) - - STRUCT_OFFSET_OF(lb_as_t, fib_node))); -} - -static void -lb_as_stack (lb_as_t *as) -{ - lb_main_t *lbm = &lb_main; - lb_vip_t *vip = &lbm->vips[as->vip_index]; - dpo_stack(lb_vip_is_gre4(vip)?lbm->dpo_gre4_type:lbm->dpo_gre6_type, - lb_vip_is_ip4(vip)?DPO_PROTO_IP4:DPO_PROTO_IP6, - &as->dpo, - fib_entry_contribute_ip_forwarding( - as->next_hop_fib_entry_index)); -} - -static fib_node_back_walk_rc_t -lb_fib_node_back_walk_notify (fib_node_t *node, - fib_node_back_walk_ctx_t *ctx) -{ - lb_as_stack(lb_as_from_fib_node(node)); - return (FIB_NODE_BACK_WALK_CONTINUE); -} - -clib_error_t * -lb_init (vlib_main_t * vm) -{ - vlib_thread_main_t *tm = vlib_get_thread_main (); - lb_main_t *lbm = &lb_main; - lb_as_t *default_as; - fib_node_vft_t lb_fib_node_vft = { - .fnv_get = lb_fib_node_get_node, - .fnv_last_lock = lb_fib_node_last_lock_gone, - .fnv_back_walk = lb_fib_node_back_walk_notify, - }; - dpo_vft_t lb_vft = { - .dv_lock = lb_dpo_lock, - .dv_unlock = lb_dpo_unlock, - .dv_format = format_lb_dpo, - }; - - lbm->vips = 0; - lbm->per_cpu = 0; - vec_validate(lbm->per_cpu, tm->n_vlib_mains - 1); - lbm->writer_lock = clib_mem_alloc_aligned (CLIB_CACHE_LINE_BYTES, CLIB_CACHE_LINE_BYTES); - lbm->writer_lock[0] = 0; - lbm->per_cpu_sticky_buckets = LB_DEFAULT_PER_CPU_STICKY_BUCKETS; - lbm->flow_timeout = LB_DEFAULT_FLOW_TIMEOUT; - lbm->ip4_src_address.as_u32 = 0xffffffff; - lbm->ip6_src_address.as_u64[0] = 0xffffffffffffffffL; - lbm->ip6_src_address.as_u64[1] = 0xffffffffffffffffL; - lbm->dpo_gre4_type = dpo_register_new_type(&lb_vft, lb_dpo_gre4_nodes); - lbm->dpo_gre6_type = dpo_register_new_type(&lb_vft, lb_dpo_gre6_nodes); - lbm->fib_node_type = fib_node_register_new_type(&lb_fib_node_vft); - - //Init AS reference counters - vlib_refcount_init(&lbm->as_refcount); - - //Allocate and init default AS. - lbm->ass = 0; - pool_get(lbm->ass, default_as); - default_as->flags = 0; - default_as->dpo.dpoi_next_node = LB_NEXT_DROP; - default_as->vip_index = ~0; - default_as->address.ip6.as_u64[0] = 0xffffffffffffffffL; - default_as->address.ip6.as_u64[1] = 0xffffffffffffffffL; - -#define _(a,b,c) lbm->vip_counters[c].name = b; - lb_foreach_vip_counter -#undef _ - return NULL; -} - -VLIB_INIT_FUNCTION (lb_init); diff --git a/plugins/lb-plugin/lb/lb.h b/plugins/lb-plugin/lb/lb.h deleted file mode 100644 index 882b9b30..00000000 --- a/plugins/lb-plugin/lb/lb.h +++ /dev/null @@ -1,333 +0,0 @@ -/* - * Copyright (c) 2016 Cisco and/or its affiliates. - * Licensed under the Apache License, Version 2.0 (the "License"); - * you may not use this file except in compliance with the License. - * You may obtain a copy of the License at: - * - * http://www.apache.org/licenses/LICENSE-2.0 - * - * Unless required by applicable law or agreed to in writing, software - * distributed under the License is distributed on an "AS IS" BASIS, - * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. - * See the License for the specific language governing permissions and - * limitations under the License. - */ - -/** - * lb-plugin implements a MagLev-like load balancer. - * http://research.google.com/pubs/pub44824.html - * - * It hasn't been tested for interoperability with the original MagLev - * but intends to provide similar functionality. - * The load-balancer receives traffic destined to VIP (Virtual IP) - * addresses from one or multiple(ECMP) routers. - * The load-balancer tunnels the traffic toward many application servers - * ensuring session stickyness (i.e. that a single sessions is tunneled - * towards a single application server). - * - */ - -#ifndef LB_PLUGIN_LB_LB_H_ -#define LB_PLUGIN_LB_LB_H_ - -#include -#include - -#include -#include -#include -#include - -#include - -#define LB_DEFAULT_PER_CPU_STICKY_BUCKETS 1 << 10 -#define LB_DEFAULT_FLOW_TIMEOUT 40 - -typedef enum { - LB_NEXT_DROP, - LB_N_NEXT, -} lb_next_t; - -/** - * Each VIP is configured with a set of - * application server. - */ -typedef struct { - /** - * Registration to FIB event. - */ - fib_node_t fib_node; - - /** - * Destination address used to tunnel traffic towards - * that application server. - * The address is also used as ID and pseudo-random - * seed for the load-balancing process. - */ - ip46_address_t address; - - /** - * ASs are indexed by address and VIP Index. - * Which means there will be duplicated if the same server - * address is used for multiple VIPs. - */ - u32 vip_index; - - /** - * Some per-AS flags. - * For now only LB_AS_FLAGS_USED is defined. - */ - u8 flags; - -#define LB_AS_FLAGS_USED 0x1 - - /** - * Rotating timestamp of when LB_AS_FLAGS_USED flag was last set. - * - * AS removal is based on garbage collection and reference counting. - * When an AS is removed, there is a race between configuration core - * and worker cores which may still add a reference while it should not - * be used. This timestamp is used to not remove the AS while a race condition - * may happen. - */ - u32 last_used; - - /** - * The FIB entry index for the next-hop - */ - fib_node_index_t next_hop_fib_entry_index; - - /** - * The child index on the FIB entry - */ - u32 next_hop_child_index; - - /** - * The next DPO in the graph to follow. - */ - dpo_id_t dpo; - -} lb_as_t; - -format_function_t format_lb_as; - -typedef struct { - u32 as_index; -} lb_new_flow_entry_t; - -#define lb_foreach_vip_counter \ - _(NEXT_PACKET, "packet from existing sessions", 0) \ - _(FIRST_PACKET, "first session packet", 1) \ - _(UNTRACKED_PACKET, "untracked packet", 2) \ - _(NO_SERVER, "no server configured", 3) - -typedef enum { -#define _(a,b,c) LB_VIP_COUNTER_##a = c, - lb_foreach_vip_counter -#undef _ - LB_N_VIP_COUNTERS -} lb_vip_counter_t; - -/** - * The load balancer supports IPv4 and IPv6 traffic - * and GRE4 and GRE6 encap. - */ -typedef enum { - LB_VIP_TYPE_IP6_GRE6, - LB_VIP_TYPE_IP6_GRE4, - LB_VIP_TYPE_IP4_GRE6, - LB_VIP_TYPE_IP4_GRE4, - LB_VIP_N_TYPES, -} lb_vip_type_t; - -format_function_t format_lb_vip_type; -unformat_function_t unformat_lb_vip_type; - -/** - * Load balancing service is provided per VIP. - * In this data model, a VIP can be a whole prefix. - * But load balancing only - * occurs on a per-source-address/port basis. Meaning that if a given source - * reuses the same port for multiple destinations within the same VIP, - * they will be considered as a single flow. - */ -typedef struct { - - //Runtime - - /** - * Vector mapping (flow-hash & new_connect_table_mask) to AS index. - * This is used for new flows. - */ - lb_new_flow_entry_t *new_flow_table; - - /** - * New flows table length - 1 - * (length MUST be a power of 2) - */ - u32 new_flow_table_mask; - - /** - * Last time garbage collection was run to free the ASs. - */ - u32 last_garbage_collection; - - //Not runtime - - /** - * A Virtual IP represents a given service delivered - * by a set of application servers. It can be a single - * address or a prefix. - * IPv4 prefixes are encoded using IPv4-in-IPv6 embedded address - * (i.e. ::/96 prefix). - */ - ip46_address_t prefix; - - /** - * The VIP prefix length. - * In case of IPv4, plen = 96 + ip4_plen. - */ - u8 plen; - - /** - * The type of traffic for this. - * LB_TYPE_UNDEFINED if unknown. - */ - lb_vip_type_t type; - - /** - * Flags related to this VIP. - * LB_VIP_FLAGS_USED means the VIP is active. - * When it is not set, the VIP in the process of being removed. - * We cannot immediately remove a VIP because the VIP index still may be stored - * in the adjacency index. - */ - u8 flags; -#define LB_VIP_FLAGS_USED 0x1 - - /** - * Pool of AS indexes used for this VIP. - * This also includes ASs that have been removed (but are still referenced). - */ - u32 *as_indexes; -} lb_vip_t; - -#define lb_vip_is_ip4(vip) ((vip)->type == LB_VIP_TYPE_IP4_GRE6 || (vip)->type == LB_VIP_TYPE_IP4_GRE4) -#define lb_vip_is_gre4(vip) ((vip)->type == LB_VIP_TYPE_IP6_GRE4 || (vip)->type == LB_VIP_TYPE_IP4_GRE4) -format_function_t format_lb_vip; -format_function_t format_lb_vip_detailed; - -typedef struct { - /** - * Each CPU has its own sticky flow hash table. - * One single table is used for all VIPs. - */ - lb_hash_t *sticky_ht; -} lb_per_cpu_t; - -typedef struct { - /** - * Pool of all Virtual IPs - */ - lb_vip_t *vips; - - /** - * Pool of ASs. - * ASs are referenced by address and vip index. - * The first element (index 0) is special and used only to fill - * new_flow_tables when no AS has been configured. - */ - lb_as_t *ass; - - /** - * Each AS has an associated reference counter. - * As ass[0] has a special meaning, its associated counter - * starts at 0 and is decremented instead. i.e. do not use it. - */ - vlib_refcount_t as_refcount; - - /** - * Some global data is per-cpu - */ - lb_per_cpu_t *per_cpu; - - /** - * Node next index for IP adjacencies, for each of the traffic types. - */ - u32 ip_lookup_next_index[LB_VIP_N_TYPES]; - - /** - * Source address used in IPv6 encapsulated traffic - */ - ip6_address_t ip6_src_address; - - /** - * Source address used for IPv4 encapsulated traffic - */ - ip4_address_t ip4_src_address; - - /** - * Number of buckets in the per-cpu sticky hash table. - */ - u32 per_cpu_sticky_buckets; - - /** - * Flow timeout in seconds. - */ - u32 flow_timeout; - - /** - * Per VIP counter - */ - vlib_simple_counter_main_t vip_counters[LB_N_VIP_COUNTERS]; - - /** - * DPO used to send packet from IP4/6 lookup to LB node. - */ - dpo_type_t dpo_gre4_type; - dpo_type_t dpo_gre6_type; - - /** - * Node type for registering to fib changes. - */ - fib_node_type_t fib_node_type; - - /** - * API dynamically registered base ID. - */ - u16 msg_id_base; - - volatile u32 *writer_lock; -} lb_main_t; - -extern lb_main_t lb_main; -extern vlib_node_registration_t lb6_node; -extern vlib_node_registration_t lb4_node; - -/** - * Fix global load-balancer parameters. - * @param ip4_address IPv4 source address used for encapsulated traffic - * @param ip6_address IPv6 source address used for encapsulated traffic - * @return 0 on success. VNET_LB_ERR_XXX on error - */ -int lb_conf(ip4_address_t *ip4_address, ip6_address_t *ip6_address, - u32 sticky_buckets, u32 flow_timeout); - -int lb_vip_add(ip46_address_t *prefix, u8 plen, lb_vip_type_t type, - u32 new_length, u32 *vip_index); -int lb_vip_del(u32 vip_index); - -int lb_vip_find_index(ip46_address_t *prefix, u8 plen, u32 *vip_index); - -#define lb_vip_get_by_index(index) (pool_is_free_index(lb_main.vips, index)?NULL:pool_elt_at_index(lb_main.vips, index)) - -int lb_vip_add_ass(u32 vip_index, ip46_address_t *addresses, u32 n); -int lb_vip_del_ass(u32 vip_index, ip46_address_t *addresses, u32 n); - -u32 lb_hash_time_now(vlib_main_t * vm); - -void lb_garbage_collection(); - -format_function_t format_lb_main; - -#endif /* LB_PLUGIN_LB_LB_H_ */ diff --git a/plugins/lb-plugin/lb/lb_test.c b/plugins/lb-plugin/lb/lb_test.c deleted file mode 100644 index 8c2eaa91..00000000 --- a/plugins/lb-plugin/lb/lb_test.c +++ /dev/null @@ -1,293 +0,0 @@ -/* - * Copyright (c) 2016 Cisco and/or its affiliates. - * Licensed under the Apache License, Version 2.0 (the "License"); - * you may not use this file except in compliance with the License. - * You may obtain a copy of the License at: - * - * http://www.apache.org/licenses/LICENSE-2.0 - * - * Unless required by applicable law or agreed to in writing, software - * distributed under the License is distributed on an "AS IS" BASIS, - * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. - * See the License for the specific language governing permissions and - * limitations under the License. - */ - -#include -#include -#include -#include -#include -#include - -//TODO: Move that to vat/plugin_api.c -////////////////////////// -uword unformat_ip46_address (unformat_input_t * input, va_list * args) -{ - ip46_address_t *ip46 = va_arg (*args, ip46_address_t *); - ip46_type_t type = va_arg (*args, ip46_type_t); - if ((type != IP46_TYPE_IP6) && - unformat(input, "%U", unformat_ip4_address, &ip46->ip4)) { - ip46_address_mask_ip4(ip46); - return 1; - } else if ((type != IP46_TYPE_IP4) && - unformat(input, "%U", unformat_ip6_address, &ip46->ip6)) { - return 1; - } - return 0; -} -uword unformat_ip46_prefix (unformat_input_t * input, va_list * args) -{ - ip46_address_t *ip46 = va_arg (*args, ip46_address_t *); - u8 *len = va_arg (*args, u8 *); - ip46_type_t type = va_arg (*args, ip46_type_t); - - u32 l; - if ((type != IP46_TYPE_IP6) && unformat(input, "%U/%u", unformat_ip4_address, &ip46->ip4, &l)) { - if (l > 32) - return 0; - *len = l + 96; - ip46->pad[0] = ip46->pad[1] = ip46->pad[2] = 0; - } else if ((type != IP46_TYPE_IP4) && unformat(input, "%U/%u", unformat_ip6_address, &ip46->ip6, &l)) { - if (l > 128) - return 0; - *len = l; - } else { - return 0; - } - return 1; -} -///////////////////////// - -#define vl_msg_id(n,h) n, -typedef enum { -#include - /* We'll want to know how many messages IDs we need... */ - VL_MSG_FIRST_AVAILABLE, -} vl_msg_id_t; -#undef vl_msg_id - -/* define message structures */ -#define vl_typedefs -#include -#undef vl_typedefs - -/* declare message handlers for each api */ - -#define vl_endianfun /* define message structures */ -#include -#undef vl_endianfun - -/* instantiate all the print functions we know about */ -#define vl_print(handle, ...) -#define vl_printfun -#include -#undef vl_printfun - -/* Get the API version number. */ -#define vl_api_version(n,v) static u32 api_version=(v); -#include -#undef vl_api_version - -typedef struct { - /* API message ID base */ - u16 msg_id_base; - vat_main_t *vat_main; -} lb_test_main_t; - -lb_test_main_t lb_test_main; - -#define foreach_standard_reply_retval_handler \ -_(lb_conf_reply) \ -_(lb_add_del_vip_reply) \ -_(lb_add_del_as_reply) - -#define _(n) \ - static void vl_api_##n##_t_handler \ - (vl_api_##n##_t * mp) \ - { \ - vat_main_t * vam = lb_test_main.vat_main; \ - i32 retval = ntohl(mp->retval); \ - if (vam->async_mode) { \ - vam->async_errors += (retval < 0); \ - } else { \ - vam->retval = retval; \ - vam->result_ready = 1; \ - } \ - } -foreach_standard_reply_retval_handler; -#undef _ - -/* - * Table of message reply handlers, must include boilerplate handlers - * we just generated - */ -#define foreach_vpe_api_reply_msg \ - _(LB_CONF_REPLY, lb_conf_reply) \ - _(LB_ADD_DEL_VIP_REPLY, lb_add_del_vip_reply) \ - _(LB_ADD_DEL_AS_REPLY, lb_add_del_as_reply) - -/* M: construct, but don't yet send a message */ -#define M(T,t) \ -do { \ - vam->result_ready = 0; \ - mp = vl_msg_api_alloc(sizeof(*mp)); \ - memcpy (mp, &mps, sizeof (*mp)); \ - mp->_vl_msg_id = ntohs (VL_API_##T + lbtm->msg_id_base); \ - mp->client_index = vam->my_client_index; \ -} while(0); - -/* S: send a message */ -#define S (vl_msg_api_send_shmem (vam->vl_input_queue, (u8 *)&mp)) - -/* W: wait for results, with timeout */ -#define W \ -do { \ - timeout = vat_time_now (vam) + 1.0; \ - \ - while (vat_time_now (vam) < timeout) { \ - if (vam->result_ready == 1) { \ - return (vam->retval); \ - } \ - } \ - return -99; \ -} while(0); - -static int api_lb_conf (vat_main_t * vam) -{ - lb_test_main_t *lbtm = &lb_test_main; - unformat_input_t *i = vam->input; - f64 timeout; - vl_api_lb_conf_t mps, *mp; - - if (!unformat(i, "%U %U %u %u", - unformat_ip4_address, &mps.ip4_src_address, - unformat_ip6_address, mps.ip6_src_address, - &mps.sticky_buckets_per_core, - &mps.flow_timeout)) { - errmsg ("invalid arguments\n"); - return -99; - } - - M(LB_CONF, lb_conf); S; W; - - /* NOTREACHED */ - return 0; -} - -static int api_lb_add_del_vip (vat_main_t * vam) -{ - lb_test_main_t *lbtm = &lb_test_main; - unformat_input_t * i = vam->input; - f64 timeout; - vl_api_lb_add_del_vip_t mps, *mp; - mps.is_del = 0; - mps.is_gre4 = 0; - - if (!unformat(i, "%U", - unformat_ip46_prefix, mps.ip_prefix, &mps.prefix_length, IP46_TYPE_ANY)) { - errmsg ("invalid prefix\n"); - return -99; - } - - if (unformat(i, "gre4")) { - mps.is_gre4 = 1; - } else if (unformat(i, "gre6")) { - mps.is_gre4 = 0; - } else { - errmsg ("no encap\n"); - return -99; - } - - if (!unformat(i, "%d", &mps.new_flows_table_length)) { - errmsg ("no table lentgh\n"); - return -99; - } - - if (unformat(i, "del")) { - mps.is_del = 1; - } - - M(LB_ADD_DEL_VIP, lb_add_del_vip); S; W; - /* NOTREACHED */ - return 0; -} - -static int api_lb_add_del_as (vat_main_t * vam) -{ - lb_test_main_t *lbtm = &lb_test_main; - unformat_input_t * i = vam->input; - f64 timeout; - vl_api_lb_add_del_as_t mps, *mp; - mps.is_del = 0; - - if (!unformat(i, "%U %U", - unformat_ip46_prefix, mps.vip_ip_prefix, &mps.vip_prefix_length, IP46_TYPE_ANY, - unformat_ip46_address, mps.as_address)) { - errmsg ("invalid prefix or address\n"); - return -99; - } - - if (unformat(i, "del")) { - mps.is_del = 1; - } - - M(LB_ADD_DEL_AS, lb_add_del_as); S; W; - /* NOTREACHED */ - return 0; -} - -/* - * List of messages that the api test plugin sends, - * and that the data plane plugin processes - */ -#define foreach_vpe_api_msg \ -_(lb_conf, " ") \ -_(lb_add_del_vip, " [gre4|gre6] [del]") \ -_(lb_add_del_as, "
[del]") - -void vat_api_hookup (vat_main_t *vam) -{ - lb_test_main_t * lbtm = &lb_test_main; - /* Hook up handlers for replies from the data plane plug-in */ -#define _(N,n) \ - vl_msg_api_set_handlers((VL_API_##N + lbtm->msg_id_base), \ - #n, \ - vl_api_##n##_t_handler, \ - vl_noop_handler, \ - vl_api_##n##_t_endian, \ - vl_api_##n##_t_print, \ - sizeof(vl_api_##n##_t), 1); - foreach_vpe_api_reply_msg; -#undef _ - - /* API messages we can send */ -#define _(n,h) hash_set_mem (vam->function_by_name, #n, api_##n); - foreach_vpe_api_msg; -#undef _ - - /* Help strings */ -#define _(n,h) hash_set_mem (vam->help_by_name, #n, h); - foreach_vpe_api_msg; -#undef _ -} - -clib_error_t * vat_plugin_register (vat_main_t *vam) -{ - lb_test_main_t * lbtm = &lb_test_main; - - u8 * name; - - lbtm->vat_main = vam; - - /* Ask the vpp engine for the first assigned message-id */ - name = format (0, "lb_%08x%c", api_version, 0); - lbtm->msg_id_base = vl_client_get_first_plugin_msg_id ((char *) name); - - if (lbtm->msg_id_base != (u16) ~0) - vat_api_hookup (vam); - - vec_free(name); - - return 0; -} diff --git a/plugins/lb-plugin/lb/lbhash.h b/plugins/lb-plugin/lb/lbhash.h deleted file mode 100644 index ca3cc143..00000000 --- a/plugins/lb-plugin/lb/lbhash.h +++ /dev/null @@ -1,216 +0,0 @@ -/* - * Copyright (c) 2012 Cisco and/or its affiliates. - * Licensed under the Apache License, Version 2.0 (the "License"); - * you may not use this file except in compliance with the License. - * You may obtain a copy of the License at: - * - * http://www.apache.org/licenses/LICENSE-2.0 - * - * Unless required by applicable law or agreed to in writing, software - * distributed under the License is distributed on an "AS IS" BASIS, - * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. - * See the License for the specific language governing permissions and - * limitations under the License. - */ - -/** - * vppinfra already includes tons of different hash tables. - * MagLev flow table is a bit different. It has to be very efficient - * for both writing and reading operations. But it does not need to - * be 100% reliable (write can fail). It also needs to recycle - * old entries in a lazy way. - * - * This hash table is the most dummy hash table you can do. - * Fixed total size, fixed bucket size. - * Advantage is that it could be very efficient (maybe). - * - */ - -#ifndef LB_PLUGIN_LB_LBHASH_H_ -#define LB_PLUGIN_LB_LBHASH_H_ - -#include - -#if defined (__SSE4_2__) -#include -#endif - -/* - * @brief Number of entries per bucket. - */ -#define LBHASH_ENTRY_PER_BUCKET 4 - -#define LB_HASH_DO_NOT_USE_SSE_BUCKETS 0 - -/* - * @brief One bucket contains 4 entries. - * Each bucket takes one 64B cache line in memory. - */ -typedef struct { - CLIB_CACHE_LINE_ALIGN_MARK (cacheline0); - u32 hash[LBHASH_ENTRY_PER_BUCKET]; - u32 timeout[LBHASH_ENTRY_PER_BUCKET]; - u32 vip[LBHASH_ENTRY_PER_BUCKET]; - u32 value[LBHASH_ENTRY_PER_BUCKET]; -} lb_hash_bucket_t; - -typedef struct { - u32 buckets_mask; - u32 timeout; - lb_hash_bucket_t buckets[]; -} lb_hash_t; - -#define lb_hash_nbuckets(h) (((h)->buckets_mask) + 1) -#define lb_hash_size(h) ((h)->buckets_mask + LBHASH_ENTRY_PER_BUCKET) - -#define lb_hash_foreach_bucket(h, bucket) \ - for (bucket = (h)->buckets; \ - bucket < (h)->buckets + lb_hash_nbuckets(h); \ - bucket++) - -#define lb_hash_foreach_entry(h, bucket, i) \ - lb_hash_foreach_bucket(h, bucket) \ - for (i = 0; i < LBHASH_ENTRY_PER_BUCKET; i++) - -#define lb_hash_foreach_valid_entry(h, bucket, i, now) \ - lb_hash_foreach_entry(h, bucket, i) \ - if (!clib_u32_loop_gt((now), bucket->timeout[i])) - -static_always_inline -lb_hash_t *lb_hash_alloc(u32 buckets, u32 timeout) -{ - if (!is_pow2(buckets)) - return NULL; - - // Allocate 1 more bucket for prefetch - u32 size = ((u64)&((lb_hash_t *)(0))->buckets[0]) + - sizeof(lb_hash_bucket_t) * (buckets + 1); - u8 *mem = 0; - lb_hash_t *h; - vec_alloc_aligned(mem, size, CLIB_CACHE_LINE_BYTES); - h = (lb_hash_t *)mem; - h->buckets_mask = (buckets - 1); - h->timeout = timeout; - return h; -} - -static_always_inline -void lb_hash_free(lb_hash_t *h) -{ - u8 *mem = (u8 *)h; - vec_free(mem); -} - -#if __SSE4_2__ -static_always_inline -u32 lb_hash_hash(u64 k0, u64 k1, u64 k2, u64 k3, u64 k4) -{ - u64 val = 0; - val = _mm_crc32_u64(val, k0); - val = _mm_crc32_u64(val, k1); - val = _mm_crc32_u64(val, k2); - val = _mm_crc32_u64(val, k3); - val = _mm_crc32_u64(val, k4); - return (u32) val; -} -#else -static_always_inline -u32 lb_hash_hash(u64 k0, u64 k1, u64 k2, u64 k3, u64 k4) -{ - u64 tmp = k0 ^ k1 ^ k2 ^ k3 ^ k4; - return (u32)clib_xxhash (tmp); -} -#endif - -static_always_inline -void lb_hash_prefetch_bucket(lb_hash_t *ht, u32 hash) -{ - lb_hash_bucket_t *bucket = &ht->buckets[hash & ht->buckets_mask]; - CLIB_PREFETCH(bucket, sizeof(*bucket), READ); -} - -static_always_inline -void lb_hash_get(lb_hash_t *ht, u32 hash, u32 vip, u32 time_now, - u32 *available_index, u32 *found_value) -{ - lb_hash_bucket_t *bucket = &ht->buckets[hash & ht->buckets_mask]; - *found_value = ~0; - *available_index = ~0; -#if __SSE4_2__ && LB_HASH_DO_NOT_USE_SSE_BUCKETS == 0 - u32 bitmask, found_index; - __m128i mask; - - // mask[*] = timeout[*] > now - mask = _mm_cmpgt_epi32(_mm_loadu_si128 ((__m128i *) bucket->timeout), - _mm_set1_epi32 (time_now)); - // bitmask[*] = now <= timeout[*/4] - bitmask = (~_mm_movemask_epi8(mask)) & 0xffff; - // Get first index with now <= timeout[*], if any. - *available_index = (bitmask)?__builtin_ctz(bitmask)/4:*available_index; - - // mask[*] = (timeout[*] > now) && (hash[*] == hash) - mask = _mm_and_si128(mask, - _mm_cmpeq_epi32( - _mm_loadu_si128 ((__m128i *) bucket->hash), - _mm_set1_epi32 (hash))); - - // Load the array of vip values - // mask[*] = (timeout[*] > now) && (hash[*] == hash) && (vip[*] == vip) - mask = _mm_and_si128(mask, - _mm_cmpeq_epi32( - _mm_loadu_si128 ((__m128i *) bucket->vip), - _mm_set1_epi32 (vip))); - - // mask[*] = (timeout[*x4] > now) && (hash[*x4] == hash) && (vip[*x4] == vip) - bitmask = _mm_movemask_epi8(mask); - // Get first index, if any - found_index = (bitmask)?__builtin_ctzll(bitmask)/4:0; - ASSERT(found_index < 4); - *found_value = (bitmask)?bucket->value[found_index]:*found_value; - bucket->timeout[found_index] = - (bitmask)?time_now + ht->timeout:bucket->timeout[found_index]; -#else - u32 i; - for (i = 0; i < LBHASH_ENTRY_PER_BUCKET; i++) { - u8 cmp = (bucket->hash[i] == hash && bucket->vip[i] == vip); - u8 timeouted = clib_u32_loop_gt(time_now, bucket->timeout[i]); - *found_value = (cmp || timeouted)?*found_value:bucket->value[i]; - bucket->timeout[i] = (cmp || timeouted)?time_now + ht->timeout:bucket->timeout[i]; - *available_index = (timeouted && (*available_index == ~0))?i:*available_index; - - if (!cmp) - return; - } -#endif -} - -static_always_inline -u32 lb_hash_available_value(lb_hash_t *h, u32 hash, u32 available_index) -{ - return h->buckets[hash & h->buckets_mask].value[available_index]; -} - -static_always_inline -void lb_hash_put(lb_hash_t *h, u32 hash, u32 value, u32 vip, - u32 available_index, u32 time_now) -{ - lb_hash_bucket_t *bucket = &h->buckets[hash & h->buckets_mask]; - bucket->hash[available_index] = hash; - bucket->value[available_index] = value; - bucket->timeout[available_index] = time_now + h->timeout; - bucket->vip[available_index] = vip; -} - -static_always_inline -u32 lb_hash_elts(lb_hash_t *h, u32 time_now) -{ - u32 tot = 0; - lb_hash_bucket_t *bucket; - u32 i; - lb_hash_foreach_valid_entry(h, bucket, i, time_now) { - tot++; - } - return tot; -} - -#endif /* LB_PLUGIN_LB_LBHASH_H_ */ diff --git a/plugins/lb-plugin/lb/node.c b/plugins/lb-plugin/lb/node.c deleted file mode 100644 index 8b763c53..00000000 --- a/plugins/lb-plugin/lb/node.c +++ /dev/null @@ -1,419 +0,0 @@ -/* - * Copyright (c) 2016 Cisco and/or its affiliates. - * Licensed under the Apache License, Version 2.0 (the "License"); - * you may not use this file except in compliance with the License. - * You may obtain a copy of the License at: - * - * http://www.apache.org/licenses/LICENSE-2.0 - * - * Unless required by applicable law or agreed to in writing, software - * distributed under the License is distributed on an "AS IS" BASIS, - * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. - * See the License for the specific language governing permissions and - * limitations under the License. - */ - -#include - -#include -#include - -#define foreach_lb_error \ - _(NONE, "no error") \ - _(PROTO_NOT_SUPPORTED, "protocol not supported") - -typedef enum { -#define _(sym,str) LB_ERROR_##sym, - foreach_lb_error -#undef _ - LB_N_ERROR, -} lb_error_t; - -static char *lb_error_strings[] = { -#define _(sym,string) string, - foreach_lb_error -#undef _ -}; - -typedef struct { - u32 vip_index; - u32 as_index; -} lb_trace_t; - -u8 * -format_lb_trace (u8 * s, va_list * args) -{ - lb_main_t *lbm = &lb_main; - CLIB_UNUSED (vlib_main_t * vm) = va_arg (*args, vlib_main_t *); - CLIB_UNUSED (vlib_node_t * node) = va_arg (*args, vlib_node_t *); - lb_trace_t *t = va_arg (*args, lb_trace_t *); - if (pool_is_free_index(lbm->vips, t->vip_index)) { - s = format(s, "lb vip[%d]: This VIP was freed since capture\n"); - } else { - s = format(s, "lb vip[%d]: %U\n", t->vip_index, format_lb_vip, &lbm->vips[t->vip_index]); - } - if (pool_is_free_index(lbm->ass, t->as_index)) { - s = format(s, "lb as[%d]: This AS was freed since capture\n"); - } else { - s = format(s, "lb as[%d]: %U\n", t->as_index, format_lb_as, &lbm->ass[t->as_index]); - } - return s; -} - -lb_hash_t *lb_get_sticky_table(u32 cpu_index) -{ - lb_main_t *lbm = &lb_main; - lb_hash_t *sticky_ht = lbm->per_cpu[cpu_index].sticky_ht; - //Check if size changed - if (PREDICT_FALSE(sticky_ht && (lbm->per_cpu_sticky_buckets != lb_hash_nbuckets(sticky_ht)))) - { - //Dereference everything in there - lb_hash_bucket_t *b; - u32 i; - lb_hash_foreach_entry(sticky_ht, b, i) { - vlib_refcount_add(&lbm->as_refcount, cpu_index, b->value[i], -1); - vlib_refcount_add(&lbm->as_refcount, cpu_index, 0, 1); - } - - lb_hash_free(sticky_ht); - sticky_ht = NULL; - } - - //Create if necessary - if (PREDICT_FALSE(sticky_ht == NULL)) { - lbm->per_cpu[cpu_index].sticky_ht = lb_hash_alloc(lbm->per_cpu_sticky_buckets, lbm->flow_timeout); - sticky_ht = lbm->per_cpu[cpu_index].sticky_ht; - clib_warning("Regenerated sticky table %p", sticky_ht); - } - - ASSERT(sticky_ht); - - //Update timeout - sticky_ht->timeout = lbm->flow_timeout; - return sticky_ht; -} - -u64 -lb_node_get_other_ports4(ip4_header_t *ip40) -{ - return 0; -} - -u64 -lb_node_get_other_ports6(ip6_header_t *ip60) -{ - return 0; -} - -static_always_inline u32 -lb_node_get_hash(vlib_buffer_t *p, u8 is_input_v4) -{ - u32 hash; - if (is_input_v4) - { - ip4_header_t *ip40; - u64 ports; - ip40 = vlib_buffer_get_current (p); - if (PREDICT_TRUE (ip40->protocol == IP_PROTOCOL_TCP || - ip40->protocol == IP_PROTOCOL_UDP)) - ports = ((u64)((udp_header_t *)(ip40 + 1))->src_port << 16) | - ((u64)((udp_header_t *)(ip40 + 1))->dst_port); - else - ports = lb_node_get_other_ports4(ip40); - - hash = lb_hash_hash(*((u64 *)&ip40->address_pair), ports, - 0, 0, 0); - } - else - { - ip6_header_t *ip60; - ip60 = vlib_buffer_get_current (p); - u64 ports; - if (PREDICT_TRUE (ip60->protocol == IP_PROTOCOL_TCP || - ip60->protocol == IP_PROTOCOL_UDP)) - ports = ((u64)((udp_header_t *)(ip60 + 1))->src_port << 16) | - ((u64)((udp_header_t *)(ip60 + 1))->dst_port); - else - ports = lb_node_get_other_ports6(ip60); - - hash = lb_hash_hash(ip60->src_address.as_u64[0], - ip60->src_address.as_u64[1], - ip60->dst_address.as_u64[0], - ip60->dst_address.as_u64[1], - ports); - } - return hash; -} - -static_always_inline uword -lb_node_fn (vlib_main_t * vm, - vlib_node_runtime_t * node, vlib_frame_t * frame, - u8 is_input_v4, //Compile-time parameter stating that is input is v4 (or v6) - u8 is_encap_v4) //Compile-time parameter stating that is GRE encap is v4 (or v6) -{ - lb_main_t *lbm = &lb_main; - u32 n_left_from, *from, next_index, *to_next, n_left_to_next; - u32 cpu_index = os_get_cpu_number(); - u32 lb_time = lb_hash_time_now(vm); - - lb_hash_t *sticky_ht = lb_get_sticky_table(cpu_index); - from = vlib_frame_vector_args (frame); - n_left_from = frame->n_vectors; - next_index = node->cached_next_index; - - u32 nexthash0 = 0; - if (PREDICT_TRUE(n_left_from > 0)) - nexthash0 = lb_node_get_hash(vlib_get_buffer (vm, from[0]), is_input_v4); - - while (n_left_from > 0) - { - vlib_get_next_frame (vm, node, next_index, to_next, n_left_to_next); - while (n_left_from > 0 && n_left_to_next > 0) - { - u32 pi0; - vlib_buffer_t *p0; - lb_vip_t *vip0; - u32 asindex0; - u16 len0; - u32 available_index0; - u8 counter = 0; - u32 hash0 = nexthash0; - - if (PREDICT_TRUE(n_left_from > 1)) - { - vlib_buffer_t *p1 = vlib_get_buffer (vm, from[1]); - //Compute next hash and prefetch bucket - nexthash0 = lb_node_get_hash(p1, is_input_v4); - lb_hash_prefetch_bucket(sticky_ht, nexthash0); - //Prefetch for encap, next - CLIB_PREFETCH (vlib_buffer_get_current(p1) - 64, 64, STORE); - } - - if (PREDICT_TRUE(n_left_from > 2)) - { - vlib_buffer_t *p2; - p2 = vlib_get_buffer(vm, from[2]); - /* prefetch packet header and data */ - vlib_prefetch_buffer_header(p2, STORE); - CLIB_PREFETCH (vlib_buffer_get_current(p2), 64, STORE); - } - - pi0 = to_next[0] = from[0]; - from += 1; - n_left_from -= 1; - to_next += 1; - n_left_to_next -= 1; - - p0 = vlib_get_buffer (vm, pi0); - vip0 = pool_elt_at_index (lbm->vips, - vnet_buffer (p0)->ip.adj_index[VLIB_TX]); - - if (is_input_v4) - { - ip4_header_t *ip40; - ip40 = vlib_buffer_get_current (p0); - len0 = clib_net_to_host_u16(ip40->length); - } - else - { - ip6_header_t *ip60; - ip60 = vlib_buffer_get_current (p0); - len0 = clib_net_to_host_u16(ip60->payload_length) + sizeof(ip6_header_t); - } - - lb_hash_get(sticky_ht, hash0, vnet_buffer (p0)->ip.adj_index[VLIB_TX], - lb_time, &available_index0, &asindex0); - - if (PREDICT_TRUE(asindex0 != ~0)) - { - //Found an existing entry - counter = LB_VIP_COUNTER_NEXT_PACKET; - } - else if (PREDICT_TRUE(available_index0 != ~0)) - { - //There is an available slot for a new flow - asindex0 = vip0->new_flow_table[hash0 & vip0->new_flow_table_mask].as_index; - counter = LB_VIP_COUNTER_FIRST_PACKET; - counter = (asindex0 == 0)?LB_VIP_COUNTER_NO_SERVER:counter; - - //TODO: There are race conditions with as0 and vip0 manipulation. - //Configuration may be changed, vectors resized, etc... - - //Dereference previously used - vlib_refcount_add(&lbm->as_refcount, cpu_index, - lb_hash_available_value(sticky_ht, hash0, available_index0), -1); - vlib_refcount_add(&lbm->as_refcount, cpu_index, - asindex0, 1); - - //Add sticky entry - //Note that when there is no AS configured, an entry is configured anyway. - //But no configured AS is not something that should happen - lb_hash_put(sticky_ht, hash0, asindex0, - vnet_buffer (p0)->ip.adj_index[VLIB_TX], - available_index0, lb_time); - } - else - { - //Could not store new entry in the table - asindex0 = vip0->new_flow_table[hash0 & vip0->new_flow_table_mask].as_index; - counter = LB_VIP_COUNTER_UNTRACKED_PACKET; - } - - vlib_increment_simple_counter(&lbm->vip_counters[counter], - cpu_index, - vnet_buffer (p0)->ip.adj_index[VLIB_TX], - 1); - - //Now let's encap - { - gre_header_t *gre0; - if (is_encap_v4) - { - ip4_header_t *ip40; - vlib_buffer_advance(p0, - sizeof(ip4_header_t) - sizeof(gre_header_t)); - ip40 = vlib_buffer_get_current(p0); - gre0 = (gre_header_t *)(ip40 + 1); - ip40->src_address = lbm->ip4_src_address; - ip40->dst_address = lbm->ass[asindex0].address.ip4; - ip40->ip_version_and_header_length = 0x45; - ip40->ttl = 128; - ip40->length = clib_host_to_net_u16(len0 + sizeof(gre_header_t) + sizeof(ip4_header_t)); - ip40->protocol = IP_PROTOCOL_GRE; - ip40->checksum = ip4_header_checksum (ip40); - } - else - { - ip6_header_t *ip60; - vlib_buffer_advance(p0, - sizeof(ip6_header_t) - sizeof(gre_header_t)); - ip60 = vlib_buffer_get_current(p0); - gre0 = (gre_header_t *)(ip60 + 1); - ip60->dst_address = lbm->ass[asindex0].address.ip6; - ip60->src_address = lbm->ip6_src_address; - ip60->hop_limit = 128; - ip60->ip_version_traffic_class_and_flow_label = clib_host_to_net_u32 (0x6<<28); - ip60->payload_length = clib_host_to_net_u16(len0 + sizeof(gre_header_t)); - ip60->protocol = IP_PROTOCOL_GRE; - } - - gre0->flags_and_version = 0; - gre0->protocol = (is_input_v4)? - clib_host_to_net_u16(0x0800): - clib_host_to_net_u16(0x86DD); - } - - if (PREDICT_FALSE (p0->flags & VLIB_BUFFER_IS_TRACED)) - { - lb_trace_t *tr = vlib_add_trace (vm, node, p0, sizeof (*tr)); - tr->as_index = asindex0; - tr->vip_index = vnet_buffer (p0)->ip.adj_index[VLIB_TX]; - } - - //Enqueue to next - //Note that this is going to error if asindex0 == 0 - vnet_buffer (p0)->ip.adj_index[VLIB_TX] = lbm->ass[asindex0].dpo.dpoi_index; - vlib_validate_buffer_enqueue_x1 (vm, node, next_index, to_next, - n_left_to_next, pi0, - lbm->ass[asindex0].dpo.dpoi_next_node); - } - vlib_put_next_frame (vm, node, next_index, n_left_to_next); - } - - return frame->n_vectors; -} - -static uword -lb6_gre6_node_fn (vlib_main_t * vm, - vlib_node_runtime_t * node, vlib_frame_t * frame) -{ - return lb_node_fn(vm, node, frame, 0, 0); -} - -static uword -lb6_gre4_node_fn (vlib_main_t * vm, - vlib_node_runtime_t * node, vlib_frame_t * frame) -{ - return lb_node_fn(vm, node, frame, 0, 1); -} - -static uword -lb4_gre6_node_fn (vlib_main_t * vm, - vlib_node_runtime_t * node, vlib_frame_t * frame) -{ - return lb_node_fn(vm, node, frame, 1, 0); -} - -static uword -lb4_gre4_node_fn (vlib_main_t * vm, - vlib_node_runtime_t * node, vlib_frame_t * frame) -{ - return lb_node_fn(vm, node, frame, 1, 1); -} - -VLIB_REGISTER_NODE (lb6_gre6_node) = -{ - .function = lb6_gre6_node_fn, - .name = "lb6-gre6", - .vector_size = sizeof (u32), - .format_trace = format_lb_trace, - - .n_errors = LB_N_ERROR, - .error_strings = lb_error_strings, - - .n_next_nodes = LB_N_NEXT, - .next_nodes = - { - [LB_NEXT_DROP] = "error-drop" - }, -}; - -VLIB_REGISTER_NODE (lb6_gre4_node) = -{ - .function = lb6_gre4_node_fn, - .name = "lb6-gre4", - .vector_size = sizeof (u32), - .format_trace = format_lb_trace, - - .n_errors = LB_N_ERROR, - .error_strings = lb_error_strings, - - .n_next_nodes = LB_N_NEXT, - .next_nodes = - { - [LB_NEXT_DROP] = "error-drop" - }, -}; - -VLIB_REGISTER_NODE (lb4_gre6_node) = -{ - .function = lb4_gre6_node_fn, - .name = "lb4-gre6", - .vector_size = sizeof (u32), - .format_trace = format_lb_trace, - - .n_errors = LB_N_ERROR, - .error_strings = lb_error_strings, - - .n_next_nodes = LB_N_NEXT, - .next_nodes = - { - [LB_NEXT_DROP] = "error-drop" - }, -}; - -VLIB_REGISTER_NODE (lb4_gre4_node) = -{ - .function = lb4_gre4_node_fn, - .name = "lb4-gre4", - .vector_size = sizeof (u32), - .format_trace = format_lb_trace, - - .n_errors = LB_N_ERROR, - .error_strings = lb_error_strings, - - .n_next_nodes = LB_N_NEXT, - .next_nodes = - { - [LB_NEXT_DROP] = "error-drop" - }, -}; - diff --git a/plugins/lb-plugin/lb/refcount.c b/plugins/lb-plugin/lb/refcount.c deleted file mode 100644 index 22415c88..00000000 --- a/plugins/lb-plugin/lb/refcount.c +++ /dev/null @@ -1,41 +0,0 @@ -/* - * Copyright (c) 2016 Cisco and/or its affiliates. - * Licensed under the Apache License, Version 2.0 (the "License"); - * you may not use this file except in compliance with the License. - * You may obtain a copy of the License at: - * - * http://www.apache.org/licenses/LICENSE-2.0 - * - * Unless required by applicable law or agreed to in writing, software - * distributed under the License is distributed on an "AS IS" BASIS, - * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. - * See the License for the specific language governing permissions and - * limitations under the License. - */ - -#include - -void __vlib_refcount_resize(vlib_refcount_per_cpu_t *per_cpu, u32 size) -{ - u32 *new_counter = 0, *old_counter; - vec_validate(new_counter, size); - memcpy(new_counter, per_cpu->counters, per_cpu->length); - old_counter = per_cpu->counters; - per_cpu->counters = new_counter; - CLIB_MEMORY_BARRIER(); - per_cpu->length = vec_len(new_counter); - vec_free(old_counter); -} - -u64 vlib_refcount_get(vlib_refcount_t *r, u32 index) -{ - u64 count = 0; - vlib_thread_main_t *tm = vlib_get_thread_main (); - u32 cpu_index; - for (cpu_index = 0; cpu_index < tm->n_vlib_mains; cpu_index++) { - if (r->per_cpu[cpu_index].length > index) - count += r->per_cpu[cpu_index].counters[index]; - } - return count; -} - diff --git a/plugins/lb-plugin/lb/refcount.h b/plugins/lb-plugin/lb/refcount.h deleted file mode 100644 index 8c26e7be..00000000 --- a/plugins/lb-plugin/lb/refcount.h +++ /dev/null @@ -1,67 +0,0 @@ -/* - * Copyright (c) 2016 Cisco and/or its affiliates. - * Licensed under the Apache License, Version 2.0 (the "License"); - * you may not use this file except in compliance with the License. - * You may obtain a copy of the License at: - * - * http://www.apache.org/licenses/LICENSE-2.0 - * - * Unless required by applicable law or agreed to in writing, software - * distributed under the License is distributed on an "AS IS" BASIS, - * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. - * See the License for the specific language governing permissions and - * limitations under the License. - */ - -/* - * vlib provides lock-free counters but those - * - Have 16bits per-CPU counter, which may overflow. - * - Would only increment. - * - * This is very similar to vlib counters, but may be used to count reference. - * Such a counter includes an arbitrary number of counters. Each counter - * is identified by its index. This is used to aggregate per-cpu memory. - * - * Warning: - * This reference counter is lock-free but is not race-condition free. - * The counting result is approximate and another mechanism needs to be used - * in order to ensure that an object may be freed. - * - */ - -#include - -typedef struct { - u32 *counters; - u32 length; - u32 *reader_lengths; - CLIB_CACHE_LINE_ALIGN_MARK(o); -} vlib_refcount_per_cpu_t; - -typedef struct { - vlib_refcount_per_cpu_t *per_cpu; -} vlib_refcount_t; - -void __vlib_refcount_resize(vlib_refcount_per_cpu_t *per_cpu, u32 size); - -static_always_inline -void vlib_refcount_add(vlib_refcount_t *r, u32 cpu_index, u32 counter_index, i32 v) -{ - vlib_refcount_per_cpu_t *per_cpu = &r->per_cpu[cpu_index]; - if (PREDICT_FALSE(counter_index >= per_cpu->length)) - __vlib_refcount_resize(per_cpu, clib_max(counter_index + 16, per_cpu->length * 2)); - - per_cpu->counters[counter_index] += v; -} - -u64 vlib_refcount_get(vlib_refcount_t *r, u32 index); - -static_always_inline -void vlib_refcount_init(vlib_refcount_t *r) -{ - vlib_thread_main_t *tm = vlib_get_thread_main (); - r->per_cpu = 0; - vec_validate (r->per_cpu, tm->n_vlib_mains - 1); -} - - diff --git a/plugins/lb-plugin/lb/util.c b/plugins/lb-plugin/lb/util.c deleted file mode 100644 index d969d168..00000000 --- a/plugins/lb-plugin/lb/util.c +++ /dev/null @@ -1,72 +0,0 @@ -/* - * Copyright (c) 2016 Cisco and/or its affiliates. - * Licensed under the Apache License, Version 2.0 (the "License"); - * you may not use this file except in compliance with the License. - * You may obtain a copy of the License at: - * - * http://www.apache.org/licenses/LICENSE-2.0 - * - * Unless required by applicable law or agreed to in writing, software - * distributed under the License is distributed on an "AS IS" BASIS, - * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. - * See the License for the specific language governing permissions and - * limitations under the License. - */ - -#include - -void ip46_prefix_normalize(ip46_address_t *prefix, u8 plen) -{ - if (plen == 0) { - prefix->as_u64[0] = 0; - prefix->as_u64[1] = 0; - } else if (plen <= 64) { - prefix->as_u64[0] &= clib_host_to_net_u64(0xffffffffffffffffL << (64 - plen)); - prefix->as_u64[1] = 0; - } else { - prefix->as_u64[1] &= clib_host_to_net_u64(0xffffffffffffffffL << (128 - plen)); - } - -} - -uword unformat_ip46_prefix (unformat_input_t * input, va_list * args) -{ - ip46_address_t *ip46 = va_arg (*args, ip46_address_t *); - u8 *len = va_arg (*args, u8 *); - ip46_type_t type = va_arg (*args, ip46_type_t); - - u32 l; - if ((type != IP46_TYPE_IP6) && unformat(input, "%U/%u", unformat_ip4_address, &ip46->ip4, &l)) { - if (l > 32) - return 0; - *len = l + 96; - ip46->pad[0] = ip46->pad[1] = ip46->pad[2] = 0; - } else if ((type != IP46_TYPE_IP4) && unformat(input, "%U/%u", unformat_ip6_address, &ip46->ip6, &l)) { - if (l > 128) - return 0; - *len = l; - } else { - return 0; - } - return 1; -} - -u8 *format_ip46_prefix (u8 * s, va_list * args) -{ - ip46_address_t *ip46 = va_arg (*args, ip46_address_t *); - u32 len = va_arg (*args, u32); //va_arg cannot use u8 or u16 - ip46_type_t type = va_arg (*args, ip46_type_t); - - int is_ip4 = 0; - if (type == IP46_TYPE_IP4) - is_ip4 = 1; - else if (type == IP46_TYPE_IP6) - is_ip4 = 0; - else - is_ip4 = (len >= 96) && ip46_address_is_ip4(ip46); - - return is_ip4 ? - format(s, "%U/%d", format_ip4_address, &ip46->ip4, len - 96): - format(s, "%U/%d", format_ip6_address, &ip46->ip6, len); -} - diff --git a/plugins/lb-plugin/lb/util.h b/plugins/lb-plugin/lb/util.h deleted file mode 100644 index 3f082310..00000000 --- a/plugins/lb-plugin/lb/util.h +++ /dev/null @@ -1,40 +0,0 @@ -/* - * Copyright (c) 2016 Cisco and/or its affiliates. - * Licensed under the Apache License, Version 2.0 (the "License"); - * you may not use this file except in compliance with the License. - * You may obtain a copy of the License at: - * - * http://www.apache.org/licenses/LICENSE-2.0 - * - * Unless required by applicable law or agreed to in writing, software - * distributed under the License is distributed on an "AS IS" BASIS, - * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. - * See the License for the specific language governing permissions and - * limitations under the License. - */ - -/* - * Non-LB specific stuff comes here - */ - -#ifndef LB_PLUGIN_LB_UTIL_H_ -#define LB_PLUGIN_LB_UTIL_H_ - -#include -#include - -#define ip46_address_type(ip46) (ip46_address_is_ip4(ip46)?IP46_TYPE_IP4:IP46_TYPE_IP6) -#define ip46_prefix_is_ip4(ip46, len) ((len) >= 96 && ip46_address_is_ip4(ip46)) -#define ip46_prefix_type(ip46, len) (ip46_prefix_is_ip4(ip46, len)?IP46_TYPE_IP4:IP46_TYPE_IP6) - -void ip46_prefix_normalize(ip46_address_t *prefix, u8 plen); -uword unformat_ip46_prefix (unformat_input_t * input, va_list * args); -u8 *format_ip46_prefix (u8 * s, va_list * args); - -/** - * 32 bits integer comparison for running values. - * 1 > 0 is true. But 1 > 0xffffffff also is. - */ -#define clib_u32_loop_gt(a, b) (((u32)(a)) - ((u32)(b)) < 0x7fffffff) - -#endif /* LB_PLUGIN_LB_UTIL_H_ */ diff --git a/plugins/lb-plugin/lb_plugin_doc.md b/plugins/lb-plugin/lb_plugin_doc.md deleted file mode 100644 index c7885ffb..00000000 --- a/plugins/lb-plugin/lb_plugin_doc.md +++ /dev/null @@ -1,141 +0,0 @@ -# Load Balancer plugin for VPP {#lb_plugin_doc} - -## Version - -The load balancer plugin is currently in *beta* version. -Both CLIs and APIs are subject to *heavy* changes. -Wich also means feedback is really welcome regarding features, apis, etc... - -## Overview - -This plugin provides load balancing for VPP in a way that is largely inspired -from Google's MagLev: http://research.google.com/pubs/pub44824.html - -The load balancer is configured with a set of Virtual IPs (VIP, which can be -prefixes), and for each VIP, with a set of Application Server addresses (ASs). - -Traffic received for a given VIP (or VIP prefix) is tunneled using GRE towards -the different ASs in a way that (tries to) ensure that a given session will -always be tunneled to the same AS. - -Both VIPs or ASs can be IPv4 or IPv6, but for a given VIP, all ASs must be using -the same encap. type (i.e. IPv4+GRE or IPv6+GRE). Meaning that for a given VIP, -all AS addresses must be of the same family. - -## Performances - -The load balancer has been tested up to 1 millions flows and still forwards more -than 3Mpps per core in such circumstances. -Although 3Mpps seems already good, it is likely that performances will be improved -in next versions. - -## Configuration - -### Global LB parameters - -The load balancer needs to be configured with some parameters: - - lb conf [ip4-src-address ] [ip6-src-address ] - [buckets ] [timeout ] - -ip4-src-address: the source address used to send encap. packets using IPv4. - -ip6-src-address: the source address used to send encap. packets using IPv6. - -buckets: the *per-thread* established-connexions-table number of buckets. - -timeout: the number of seconds a connection will remain in the - established-connexions-table while no packet for this flow - is received. - - -### Configure the VIPs - - lb vip [encap (gre6|gre4)] [new_len ] [del] - -new_len is the size of the new-connection-table. It should be 1 or 2 orders of -magnitude bigger than the number of ASs for the VIP in order to ensure a good -load balancing. - -Examples: - - lb vip 2002::/16 encap gre6 new_len 1024 - lb vip 2003::/16 encap gre4 new_len 2048 - lb vip 80.0.0.0/8 encap gre6 new_len 16 - lb vip 90.0.0.0/8 encap gre4 new_len 1024 - -### Configure the ASs (for each VIP) - - lb as [
[
[...]]] [del] - -You can add (or delete) as many ASs at a time (for a single VIP). -Note that the AS address family must correspond to the VIP encap. IP family. - -Examples: - - lb as 2002::/16 2001::2 2001::3 2001::4 - lb as 2003::/16 10.0.0.1 10.0.0.2 - lb as 80.0.0.0/8 2001::2 - lb as 90.0.0.0/8 10.0.0.1 - - - -## Monitoring - -The plugin provides quite a bunch of counters and information. -These are still subject to quite significant changes. - - show lb - show lb vip - show lb vip verbose - - show node counters - - -## Design notes - -### Multi-Threading - -MagLev is a distributed system which pseudo-randomly generates a -new-connections-table based on AS names such that each server configured with -the same set of ASs ends up with the same table. Connection stickyness is then -ensured with an established-connections-table. Using ECMP, it is assumed (but -not relied on) that servers will mostly receive traffic for different flows. - -This implementation pushes the parallelism a little bit further by using -one established-connections table per thread. This is equivalent to assuming -that RSS will make a job similar to ECMP, and is pretty useful as threads don't -need to get a lock in order to write in the table. - -### Hash Table - -A load balancer requires an efficient read and write hash table. The hash table -used by ip6-forward is very read-efficient, but not so much for writing. In -addition, it is not a big deal if writing into the hash table fails (again, -MagLev uses a flow table but does not heaviliy relies on it). - -The plugin therefore uses a very specific (and stupid) hash table. - - Fixed (and power of 2) number of buckets (configured at runtime) - - Fixed (and power of 2) elements per buckets (configured at compilation time) - -### Reference counting - -When an AS is removed, there is two possible ways to react. - - Keep using the AS for established connections - - Change AS for established connections (likely to cause error for TCP) - -In the first case, although an AS is removed from the configuration, its -associated state needs to stay around as long as it is used by at least one -thread. - -In order to avoid locks, a specific reference counter is used. The design is quite -similar to clib counters but: - - It is possible to decrease the value - - Summing will not zero the per-thread counters - - Only the thread can reallocate its own counters vector (to avoid concurrency issues) - -This reference counter is lock free, but reading a count of 0 does not mean -the value can be freed unless it is ensured by *other* means that no other thread -is concurrently referencing the object. In the case of this plugin, it is assumed -that no concurrent event will take place after a few seconds. - diff --git a/plugins/sample-plugin/Makefile.am b/plugins/sample-plugin/Makefile.am deleted file mode 100644 index e221f8c1..00000000 --- a/plugins/sample-plugin/Makefile.am +++ /dev/null @@ -1,56 +0,0 @@ -# Copyright (c) 2015 Cisco and/or its affiliates. -# Licensed under the Apache License, Version 2.0 (the "License"); -# you may not use this file except in compliance with the License. -# You may obtain a copy of the License at: -# -# http://www.apache.org/licenses/LICENSE-2.0 -# -# Unless required by applicable law or agreed to in writing, software -# distributed under the License is distributed on an "AS IS" BASIS, -# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. -# See the License for the specific language governing permissions and -# limitations under the License. - -AUTOMAKE_OPTIONS = foreign subdir-objects - -AM_CFLAGS = -Wall -AM_LDFLAGS = -module -shared -avoid-version - -vppapitestpluginsdir = ${libdir}/vpp_api_test_plugins -vpppluginsdir = ${libdir}/vpp_plugins - -vppapitestplugins_LTLIBRARIES = sample_test_plugin.la -vppplugins_LTLIBRARIES = sample_plugin.la - -sample_plugin_la_SOURCES = sample/sample.c sample/node.c \ - sample/sample_plugin.api.h - -BUILT_SOURCES = sample/sample.api.h sample/sample.api.json - -SUFFIXES = .api.h .api - -%.api.h: %.api - mkdir -p `dirname $@` ; \ - $(CC) $(CPPFLAGS) -E -P -C -x c $^ \ - | vppapigen --input - --output $@ --show-name $@ - -%.api.json: %.api - @echo " JSON APIGEN " $@ ; \ - mkdir -p `dirname $@` ; \ - $(CC) $(CPPFLAGS) -E -P -C -x c $^ \ - | vppapigen --input - --json $@ - -apidir = $(prefix)/sample/ -api_DATA = sample.api.json - -noinst_HEADERS = \ - sample/sample_all_api_h.h \ - sample/sample_msg_enum.h \ - sample/sample.api.h - -sample_test_plugin_la_SOURCES = sample/sample_test.c sample/sample_plugin.api.h - -# Remove *.la files -install-data-hook: - @(cd $(vpppluginsdir) && $(RM) $(vppplugins_LTLIBRARIES)) - @(cd $(vppapitestpluginsdir) && $(RM) $(vppapitestplugins_LTLIBRARIES)) diff --git a/plugins/sample-plugin/configure.ac b/plugins/sample-plugin/configure.ac deleted file mode 100644 index 43642732..00000000 --- a/plugins/sample-plugin/configure.ac +++ /dev/null @@ -1,9 +0,0 @@ -AC_INIT(sample_plugin, 1.0) -LT_INIT -AM_INIT_AUTOMAKE -AM_SILENT_RULES([yes]) -AC_PREFIX_DEFAULT([/usr]) - -AC_PROG_CC - -AC_OUTPUT([Makefile]) diff --git a/plugins/sample-plugin/sample/node.c b/plugins/sample-plugin/sample/node.c deleted file mode 100644 index 94c1706b..00000000 --- a/plugins/sample-plugin/sample/node.c +++ /dev/null @@ -1,295 +0,0 @@ -/* - * Copyright (c) 2015 Cisco and/or its affiliates. - * Licensed under the Apache License, Version 2.0 (the "License"); - * you may not use this file except in compliance with the License. - * You may obtain a copy of the License at: - * - * http://www.apache.org/licenses/LICENSE-2.0 - * - * Unless required by applicable law or agreed to in writing, software - * distributed under the License is distributed on an "AS IS" BASIS, - * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. - * See the License for the specific language governing permissions and - * limitations under the License. - */ -#include -#include -#include -#include -#include - -typedef struct { - u32 next_index; - u32 sw_if_index; - u8 new_src_mac[6]; - u8 new_dst_mac[6]; -} sample_trace_t; - -static u8 * -format_mac_address (u8 * s, va_list * args) -{ - u8 *a = va_arg (*args, u8 *); - return format (s, "%02x:%02x:%02x:%02x:%02x:%02x", - a[0], a[1], a[2], a[3], a[4], a[5]); -} - -/* packet trace format function */ -static u8 * format_sample_trace (u8 * s, va_list * args) -{ - CLIB_UNUSED (vlib_main_t * vm) = va_arg (*args, vlib_main_t *); - CLIB_UNUSED (vlib_node_t * node) = va_arg (*args, vlib_node_t *); - sample_trace_t * t = va_arg (*args, sample_trace_t *); - - s = format (s, "SAMPLE: sw_if_index %d, next index %d\n", - t->sw_if_index, t->next_index); - s = format (s, " new src %U -> new dst %U", - format_mac_address, t->new_src_mac, - format_mac_address, t->new_dst_mac); - - return s; -} - -vlib_node_registration_t sample_node; - -#define foreach_sample_error \ -_(SWAPPED, "Mac swap packets processed") - -typedef enum { -#define _(sym,str) SAMPLE_ERROR_##sym, - foreach_sample_error -#undef _ - SAMPLE_N_ERROR, -} sample_error_t; - -static char * sample_error_strings[] = { -#define _(sym,string) string, - foreach_sample_error -#undef _ -}; - -typedef enum { - SAMPLE_NEXT_INTERFACE_OUTPUT, - SAMPLE_N_NEXT, -} sample_next_t; - -#define foreach_mac_address_offset \ -_(0) \ -_(1) \ -_(2) \ -_(3) \ -_(4) \ -_(5) - -static uword -sample_node_fn (vlib_main_t * vm, - vlib_node_runtime_t * node, - vlib_frame_t * frame) -{ - u32 n_left_from, * from, * to_next; - sample_next_t next_index; - u32 pkts_swapped = 0; - - from = vlib_frame_vector_args (frame); - n_left_from = frame->n_vectors; - next_index = node->cached_next_index; - - while (n_left_from > 0) - { - u32 n_left_to_next; - - vlib_get_next_frame (vm, node, next_index, - to_next, n_left_to_next); - - while (n_left_from >= 4 && n_left_to_next >= 2) - { - u32 next0 = SAMPLE_NEXT_INTERFACE_OUTPUT; - u32 next1 = SAMPLE_NEXT_INTERFACE_OUTPUT; - u32 sw_if_index0, sw_if_index1; - u8 tmp0[6], tmp1[6]; - ethernet_header_t *en0, *en1; - u32 bi0, bi1; - vlib_buffer_t * b0, * b1; - - /* Prefetch next iteration. */ - { - vlib_buffer_t * p2, * p3; - - p2 = vlib_get_buffer (vm, from[2]); - p3 = vlib_get_buffer (vm, from[3]); - - vlib_prefetch_buffer_header (p2, LOAD); - vlib_prefetch_buffer_header (p3, LOAD); - - CLIB_PREFETCH (p2->data, CLIB_CACHE_LINE_BYTES, STORE); - CLIB_PREFETCH (p3->data, CLIB_CACHE_LINE_BYTES, STORE); - } - - /* speculatively enqueue b0 and b1 to the current next frame */ - to_next[0] = bi0 = from[0]; - to_next[1] = bi1 = from[1]; - from += 2; - to_next += 2; - n_left_from -= 2; - n_left_to_next -= 2; - - b0 = vlib_get_buffer (vm, bi0); - b1 = vlib_get_buffer (vm, bi1); - - ASSERT (b0->current_data == 0); - ASSERT (b1->current_data == 0); - - en0 = vlib_buffer_get_current (b0); - en1 = vlib_buffer_get_current (b1); - - /* This is not the fastest way to swap src + dst mac addresses */ -#define _(a) tmp0[a] = en0->src_address[a]; - foreach_mac_address_offset; -#undef _ -#define _(a) en0->src_address[a] = en0->dst_address[a]; - foreach_mac_address_offset; -#undef _ -#define _(a) en0->dst_address[a] = tmp0[a]; - foreach_mac_address_offset; -#undef _ - -#define _(a) tmp1[a] = en1->src_address[a]; - foreach_mac_address_offset; -#undef _ -#define _(a) en1->src_address[a] = en1->dst_address[a]; - foreach_mac_address_offset; -#undef _ -#define _(a) en1->dst_address[a] = tmp1[a]; - foreach_mac_address_offset; -#undef _ - - - - sw_if_index0 = vnet_buffer(b0)->sw_if_index[VLIB_RX]; - sw_if_index1 = vnet_buffer(b1)->sw_if_index[VLIB_RX]; - - /* Send pkt back out the RX interface */ - vnet_buffer(b0)->sw_if_index[VLIB_TX] = sw_if_index0; - vnet_buffer(b1)->sw_if_index[VLIB_TX] = sw_if_index1; - - pkts_swapped += 2; - - if (PREDICT_FALSE((node->flags & VLIB_NODE_FLAG_TRACE))) - { - if (b0->flags & VLIB_BUFFER_IS_TRACED) - { - sample_trace_t *t = - vlib_add_trace (vm, node, b0, sizeof (*t)); - t->sw_if_index = sw_if_index0; - t->next_index = next0; - clib_memcpy (t->new_src_mac, en0->src_address, - sizeof (t->new_src_mac)); - clib_memcpy (t->new_dst_mac, en0->dst_address, - sizeof (t->new_dst_mac)); - - } - if (b1->flags & VLIB_BUFFER_IS_TRACED) - { - sample_trace_t *t = - vlib_add_trace (vm, node, b1, sizeof (*t)); - t->sw_if_index = sw_if_index1; - t->next_index = next1; - clib_memcpy (t->new_src_mac, en1->src_address, - sizeof (t->new_src_mac)); - clib_memcpy (t->new_dst_mac, en1->dst_address, - sizeof (t->new_dst_mac)); - } - } - - /* verify speculative enqueues, maybe switch current next frame */ - vlib_validate_buffer_enqueue_x2 (vm, node, next_index, - to_next, n_left_to_next, - bi0, bi1, next0, next1); - } - - while (n_left_from > 0 && n_left_to_next > 0) - { - u32 bi0; - vlib_buffer_t * b0; - u32 next0 = SAMPLE_NEXT_INTERFACE_OUTPUT; - u32 sw_if_index0; - u8 tmp0[6]; - ethernet_header_t *en0; - - /* speculatively enqueue b0 to the current next frame */ - bi0 = from[0]; - to_next[0] = bi0; - from += 1; - to_next += 1; - n_left_from -= 1; - n_left_to_next -= 1; - - b0 = vlib_get_buffer (vm, bi0); - /* - * Direct from the driver, we should be at offset 0 - * aka at &b0->data[0] - */ - ASSERT (b0->current_data == 0); - - en0 = vlib_buffer_get_current (b0); - - /* This is not the fastest way to swap src + dst mac addresses */ -#define _(a) tmp0[a] = en0->src_address[a]; - foreach_mac_address_offset; -#undef _ -#define _(a) en0->src_address[a] = en0->dst_address[a]; - foreach_mac_address_offset; -#undef _ -#define _(a) en0->dst_address[a] = tmp0[a]; - foreach_mac_address_offset; -#undef _ - - sw_if_index0 = vnet_buffer(b0)->sw_if_index[VLIB_RX]; - - /* Send pkt back out the RX interface */ - vnet_buffer(b0)->sw_if_index[VLIB_TX] = sw_if_index0; - - if (PREDICT_FALSE((node->flags & VLIB_NODE_FLAG_TRACE) - && (b0->flags & VLIB_BUFFER_IS_TRACED))) { - sample_trace_t *t = - vlib_add_trace (vm, node, b0, sizeof (*t)); - t->sw_if_index = sw_if_index0; - t->next_index = next0; - clib_memcpy (t->new_src_mac, en0->src_address, - sizeof (t->new_src_mac)); - clib_memcpy (t->new_dst_mac, en0->dst_address, - sizeof (t->new_dst_mac)); - } - - pkts_swapped += 1; - - /* verify speculative enqueue, maybe switch current next frame */ - vlib_validate_buffer_enqueue_x1 (vm, node, next_index, - to_next, n_left_to_next, - bi0, next0); - } - - vlib_put_next_frame (vm, node, next_index, n_left_to_next); - } - - vlib_node_increment_counter (vm, sample_node.index, - SAMPLE_ERROR_SWAPPED, pkts_swapped); - return frame->n_vectors; -} - -VLIB_REGISTER_NODE (sample_node) = { - .function = sample_node_fn, - .name = "sample", - .vector_size = sizeof (u32), - .format_trace = format_sample_trace, - .type = VLIB_NODE_TYPE_INTERNAL, - - .n_errors = ARRAY_LEN(sample_error_strings), - .error_strings = sample_error_strings, - - .n_next_nodes = SAMPLE_N_NEXT, - - /* edit / add dispositions here */ - .next_nodes = { - [SAMPLE_NEXT_INTERFACE_OUTPUT] = "interface-output", - }, -}; diff --git a/plugins/sample-plugin/sample/sample.api b/plugins/sample-plugin/sample/sample.api deleted file mode 100644 index f99cdb38..00000000 --- a/plugins/sample-plugin/sample/sample.api +++ /dev/null @@ -1,39 +0,0 @@ -/* Hey Emacs use -*- mode: C -*- */ -/* - * Copyright (c) 2015 Cisco and/or its affiliates. - * Licensed under the Apache License, Version 2.0 (the "License"); - * you may not use this file except in compliance with the License. - * You may obtain a copy of the License at: - * - * http://www.apache.org/licenses/LICENSE-2.0 - * - * Unless required by applicable law or agreed to in writing, software - * distributed under the License is distributed on an "AS IS" BASIS, - * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. - * See the License for the specific language governing permissions and - * limitations under the License. - */ - -/* Define a simple binary API to control the feature */ - -define sample_macswap_enable_disable { - /* Client identifier, set from api_main.my_client_index */ - u32 client_index; - - /* Arbitrary context, so client can match reply to request */ - u32 context; - - /* Enable / disable the feature */ - u8 enable_disable; - - /* Interface handle */ - u32 sw_if_index; -}; - -define sample_macswap_enable_disable_reply { - /* From the request */ - u32 context; - - /* Return value, zero means all OK */ - i32 retval; -}; diff --git a/plugins/sample-plugin/sample/sample.c b/plugins/sample-plugin/sample/sample.c deleted file mode 100644 index 603cb2d0..00000000 --- a/plugins/sample-plugin/sample/sample.c +++ /dev/null @@ -1,255 +0,0 @@ -/* - * Copyright (c) 2015 Cisco and/or its affiliates. - * Licensed under the Apache License, Version 2.0 (the "License"); - * you may not use this file except in compliance with the License. - * You may obtain a copy of the License at: - * - * http://www.apache.org/licenses/LICENSE-2.0 - * - * Unless required by applicable law or agreed to in writing, software - * distributed under the License is distributed on an "AS IS" BASIS, - * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. - * See the License for the specific language governing permissions and - * limitations under the License. - */ -/* - *------------------------------------------------------------------ - * sample.c - simple MAC-swap API / debug CLI handling - *------------------------------------------------------------------ - */ - -#include -#include -#include - -#include -#include -#include - -/* define message IDs */ -#include - -/* define message structures */ -#define vl_typedefs -#include -#undef vl_typedefs - -/* define generated endian-swappers */ -#define vl_endianfun -#include -#undef vl_endianfun - -/* instantiate all the print functions we know about */ -#define vl_print(handle, ...) vlib_cli_output (handle, __VA_ARGS__) -#define vl_printfun -#include -#undef vl_printfun - -/* Get the API version number */ -#define vl_api_version(n,v) static u32 api_version=(v); -#include -#undef vl_api_version - -/* - * A handy macro to set up a message reply. - * Assumes that the following variables are available: - * mp - pointer to request message - * rmp - pointer to reply message type - * rv - return value - */ - -#define REPLY_MACRO(t) \ -do { \ - unix_shared_memory_queue_t * q = \ - vl_api_client_index_to_input_queue (mp->client_index); \ - if (!q) \ - return; \ - \ - rmp = vl_msg_api_alloc (sizeof (*rmp)); \ - rmp->_vl_msg_id = ntohs((t)+sm->msg_id_base); \ - rmp->context = mp->context; \ - rmp->retval = ntohl(rv); \ - \ - vl_msg_api_send_shmem (q, (u8 *)&rmp); \ -} while(0); - - -/* List of message types that this plugin understands */ - -#define foreach_sample_plugin_api_msg \ -_(SAMPLE_MACSWAP_ENABLE_DISABLE, sample_macswap_enable_disable) - -/* - * This routine exists to convince the vlib plugin framework that - * we haven't accidentally copied a random .dll into the plugin directory. - * - * Also collects global variable pointers passed from the vpp engine - */ - -clib_error_t * -vlib_plugin_register (vlib_main_t * vm, vnet_plugin_handoff_t * h, - int from_early_init) -{ - sample_main_t * sm = &sample_main; - clib_error_t * error = 0; - - sm->vlib_main = vm; - sm->vnet_main = h->vnet_main; - sm->ethernet_main = h->ethernet_main; - - return error; -} - -/* Action function shared between message handler and debug CLI */ - -int sample_macswap_enable_disable (sample_main_t * sm, u32 sw_if_index, - int enable_disable) -{ - vnet_sw_interface_t * sw; - int rv = 0; - - /* Utterly wrong? */ - if (pool_is_free_index (sm->vnet_main->interface_main.sw_interfaces, - sw_if_index)) - return VNET_API_ERROR_INVALID_SW_IF_INDEX; - - /* Not a physical port? */ - sw = vnet_get_sw_interface (sm->vnet_main, sw_if_index); - if (sw->type != VNET_SW_INTERFACE_TYPE_HARDWARE) - return VNET_API_ERROR_INVALID_SW_IF_INDEX; - - vnet_feature_enable_disable ("device-input", "sample", - sw_if_index, enable_disable, 0, 0); - - return rv; -} - -static clib_error_t * -macswap_enable_disable_command_fn (vlib_main_t * vm, - unformat_input_t * input, - vlib_cli_command_t * cmd) -{ - sample_main_t * sm = &sample_main; - u32 sw_if_index = ~0; - int enable_disable = 1; - - int rv; - - while (unformat_check_input (input) != UNFORMAT_END_OF_INPUT) { - if (unformat (input, "disable")) - enable_disable = 0; - else if (unformat (input, "%U", unformat_vnet_sw_interface, - sm->vnet_main, &sw_if_index)) - ; - else - break; - } - - if (sw_if_index == ~0) - return clib_error_return (0, "Please specify an interface..."); - - rv = sample_macswap_enable_disable (sm, sw_if_index, enable_disable); - - switch(rv) { - case 0: - break; - - case VNET_API_ERROR_INVALID_SW_IF_INDEX: - return clib_error_return - (0, "Invalid interface, only works on physical ports"); - break; - - case VNET_API_ERROR_UNIMPLEMENTED: - return clib_error_return (0, "Device driver doesn't support redirection"); - break; - - default: - return clib_error_return (0, "sample_macswap_enable_disable returned %d", - rv); - } - return 0; -} - -VLIB_CLI_COMMAND (sr_content_command, static) = { - .path = "sample macswap", - .short_help = - "sample macswap [disable]", - .function = macswap_enable_disable_command_fn, -}; - -/* API message handler */ -static void vl_api_sample_macswap_enable_disable_t_handler -(vl_api_sample_macswap_enable_disable_t * mp) -{ - vl_api_sample_macswap_enable_disable_reply_t * rmp; - sample_main_t * sm = &sample_main; - int rv; - - rv = sample_macswap_enable_disable (sm, ntohl(mp->sw_if_index), - (int) (mp->enable_disable)); - - REPLY_MACRO(VL_API_SAMPLE_MACSWAP_ENABLE_DISABLE_REPLY); -} - -/* Set up the API message handling tables */ -static clib_error_t * -sample_plugin_api_hookup (vlib_main_t *vm) -{ - sample_main_t * sm = &sample_main; -#define _(N,n) \ - vl_msg_api_set_handlers((VL_API_##N + sm->msg_id_base), \ - #n, \ - vl_api_##n##_t_handler, \ - vl_noop_handler, \ - vl_api_##n##_t_endian, \ - vl_api_##n##_t_print, \ - sizeof(vl_api_##n##_t), 1); - foreach_sample_plugin_api_msg; -#undef _ - - return 0; -} - -#define vl_msg_name_crc_list -#include -#undef vl_msg_name_crc_list - -static void -setup_message_id_table (sample_main_t * sm, api_main_t *am) -{ -#define _(id,n,crc) \ - vl_msg_api_add_msg_name_crc (am, #n "_" #crc, id + sm->msg_id_base); - foreach_vl_msg_name_crc_sample; -#undef _ -} - -static clib_error_t * sample_init (vlib_main_t * vm) -{ - sample_main_t * sm = &sample_main; - clib_error_t * error = 0; - u8 * name; - - name = format (0, "sample_%08x%c", api_version, 0); - - /* Ask for a correctly-sized block of API message decode slots */ - sm->msg_id_base = vl_msg_api_get_msg_ids - ((char *) name, VL_MSG_FIRST_AVAILABLE); - - error = sample_plugin_api_hookup (vm); - - /* Add our API messages to the global name_crc hash table */ - setup_message_id_table (sm, &api_main); - - vec_free(name); - - return error; -} - -VLIB_INIT_FUNCTION (sample_init); - -VNET_FEATURE_INIT (sample, static) = -{ - .arc_name = "device-input", - .node_name = "sample", - .runs_before = VNET_FEATURES ("ethernet-input"), -}; diff --git a/plugins/sample-plugin/sample/sample.h b/plugins/sample-plugin/sample/sample.h deleted file mode 100644 index d268d482..00000000 --- a/plugins/sample-plugin/sample/sample.h +++ /dev/null @@ -1,40 +0,0 @@ -/* - * Copyright (c) 2015 Cisco and/or its affiliates. - * Licensed under the Apache License, Version 2.0 (the "License"); - * you may not use this file except in compliance with the License. - * You may obtain a copy of the License at: - * - * http://www.apache.org/licenses/LICENSE-2.0 - * - * Unless required by applicable law or agreed to in writing, software - * distributed under the License is distributed on an "AS IS" BASIS, - * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. - * See the License for the specific language governing permissions and - * limitations under the License. - */ -#ifndef __included_sample_h__ -#define __included_sample_h__ - -#include -#include -#include - -#include -#include -#include - -typedef struct { - /* API message ID base */ - u16 msg_id_base; - - /* convenience */ - vlib_main_t * vlib_main; - vnet_main_t * vnet_main; - ethernet_main_t * ethernet_main; -} sample_main_t; - -sample_main_t sample_main; - -vlib_node_registration_t sample_node; - -#endif /* __included_sample_h__ */ diff --git a/plugins/sample-plugin/sample/sample_all_api_h.h b/plugins/sample-plugin/sample/sample_all_api_h.h deleted file mode 100644 index 774d782f..00000000 --- a/plugins/sample-plugin/sample/sample_all_api_h.h +++ /dev/null @@ -1,16 +0,0 @@ -/* - * Copyright (c) 2015 Cisco and/or its affiliates. - * Licensed under the Apache License, Version 2.0 (the "License"); - * you may not use this file except in compliance with the License. - * You may obtain a copy of the License at: - * - * http://www.apache.org/licenses/LICENSE-2.0 - * - * Unless required by applicable law or agreed to in writing, software - * distributed under the License is distributed on an "AS IS" BASIS, - * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. - * See the License for the specific language governing permissions and - * limitations under the License. - */ -/* Include the generated file, see BUILT_SOURCES in Makefile.am */ -#include diff --git a/plugins/sample-plugin/sample/sample_msg_enum.h b/plugins/sample-plugin/sample/sample_msg_enum.h deleted file mode 100644 index af4172f7..00000000 --- a/plugins/sample-plugin/sample/sample_msg_enum.h +++ /dev/null @@ -1,28 +0,0 @@ -/* - * Copyright (c) 2015 Cisco and/or its affiliates. - * Licensed under the Apache License, Version 2.0 (the "License"); - * you may not use this file except in compliance with the License. - * You may obtain a copy of the License at: - * - * http://www.apache.org/licenses/LICENSE-2.0 - * - * Unless required by applicable law or agreed to in writing, software - * distributed under the License is distributed on an "AS IS" BASIS, - * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. - * See the License for the specific language governing permissions and - * limitations under the License. - */ -#ifndef included_sample_msg_enum_h -#define included_sample_msg_enum_h - -#include - -#define vl_msg_id(n,h) n, -typedef enum { -#include - /* We'll want to know how many messages IDs we need... */ - VL_MSG_FIRST_AVAILABLE, -} vl_msg_id_t; -#undef vl_msg_id - -#endif /* included_sample_msg_enum_h */ diff --git a/plugins/sample-plugin/sample/sample_test.c b/plugins/sample-plugin/sample/sample_test.c deleted file mode 100644 index dd1b0215..00000000 --- a/plugins/sample-plugin/sample/sample_test.c +++ /dev/null @@ -1,213 +0,0 @@ -/* - * Copyright (c) 2015 Cisco and/or its affiliates. - * Licensed under the Apache License, Version 2.0 (the "License"); - * you may not use this file except in compliance with the License. - * You may obtain a copy of the License at: - * - * http://www.apache.org/licenses/LICENSE-2.0 - * - * Unless required by applicable law or agreed to in writing, software - * distributed under the License is distributed on an "AS IS" BASIS, - * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. - * See the License for the specific language governing permissions and - * limitations under the License. - */ -/* - *------------------------------------------------------------------ - * sample_test.c - test harness plugin - *------------------------------------------------------------------ - */ - -#include -#include -#include -#include -#include - -uword unformat_sw_if_index (unformat_input_t * input, va_list * args); - -/* Declare message IDs */ -#include - -/* define message structures */ -#define vl_typedefs -#include -#undef vl_typedefs - -/* declare message handlers for each api */ - -#define vl_endianfun /* define message structures */ -#include -#undef vl_endianfun - -/* instantiate all the print functions we know about */ -#define vl_print(handle, ...) -#define vl_printfun -#include -#undef vl_printfun - -/* Get the API version number. */ -#define vl_api_version(n,v) static u32 api_version=(v); -#include -#undef vl_api_version - - -typedef struct { - /* API message ID base */ - u16 msg_id_base; - vat_main_t *vat_main; -} sample_test_main_t; - -sample_test_main_t sample_test_main; - -#define foreach_standard_reply_retval_handler \ -_(sample_macswap_enable_disable_reply) - -#define _(n) \ - static void vl_api_##n##_t_handler \ - (vl_api_##n##_t * mp) \ - { \ - vat_main_t * vam = sample_test_main.vat_main; \ - i32 retval = ntohl(mp->retval); \ - if (vam->async_mode) { \ - vam->async_errors += (retval < 0); \ - } else { \ - vam->retval = retval; \ - vam->result_ready = 1; \ - } \ - } -foreach_standard_reply_retval_handler; -#undef _ - -/* - * Table of message reply handlers, must include boilerplate handlers - * we just generated - */ -#define foreach_vpe_api_reply_msg \ -_(SAMPLE_MACSWAP_ENABLE_DISABLE_REPLY, sample_macswap_enable_disable_reply) - - -/* M: construct, but don't yet send a message */ - -#define M(T,t) \ -do { \ - vam->result_ready = 0; \ - mp = vl_msg_api_alloc(sizeof(*mp)); \ - memset (mp, 0, sizeof (*mp)); \ - mp->_vl_msg_id = ntohs (VL_API_##T + sm->msg_id_base); \ - mp->client_index = vam->my_client_index; \ -} while(0); - -#define M2(T,t,n) \ -do { \ - vam->result_ready = 0; \ - mp = vl_msg_api_alloc(sizeof(*mp)+(n)); \ - memset (mp, 0, sizeof (*mp)); \ - mp->_vl_msg_id = ntohs (VL_API_##T + sm->msg_id_base); \ - mp->client_index = vam->my_client_index; \ -} while(0); - -/* S: send a message */ -#define S (vl_msg_api_send_shmem (vam->vl_input_queue, (u8 *)&mp)) - -/* W: wait for results, with timeout */ -#define W \ -do { \ - timeout = vat_time_now (vam) + 1.0; \ - \ - while (vat_time_now (vam) < timeout) { \ - if (vam->result_ready == 1) { \ - return (vam->retval); \ - } \ - } \ - return -99; \ -} while(0); - -static int api_sample_macswap_enable_disable (vat_main_t * vam) -{ - sample_test_main_t * sm = &sample_test_main; - unformat_input_t * i = vam->input; - f64 timeout; - int enable_disable = 1; - u32 sw_if_index = ~0; - vl_api_sample_macswap_enable_disable_t * mp; - - /* Parse args required to build the message */ - while (unformat_check_input (i) != UNFORMAT_END_OF_INPUT) { - if (unformat (i, "%U", unformat_sw_if_index, vam, &sw_if_index)) - ; - else if (unformat (i, "sw_if_index %d", &sw_if_index)) - ; - else if (unformat (i, "disable")) - enable_disable = 0; - else - break; - } - - if (sw_if_index == ~0) { - errmsg ("missing interface name / explicit sw_if_index number \n"); - return -99; - } - - /* Construct the API message */ - M(SAMPLE_MACSWAP_ENABLE_DISABLE, sample_macswap_enable_disable); - mp->sw_if_index = ntohl (sw_if_index); - mp->enable_disable = enable_disable; - - /* send it... */ - S; - - /* Wait for a reply... */ - W; -} - -/* - * List of messages that the api test plugin sends, - * and that the data plane plugin processes - */ -#define foreach_vpe_api_msg \ -_(sample_macswap_enable_disable, " [disable]") - -void vat_api_hookup (vat_main_t *vam) -{ - sample_test_main_t * sm = &sample_test_main; - /* Hook up handlers for replies from the data plane plug-in */ -#define _(N,n) \ - vl_msg_api_set_handlers((VL_API_##N + sm->msg_id_base), \ - #n, \ - vl_api_##n##_t_handler, \ - vl_noop_handler, \ - vl_api_##n##_t_endian, \ - vl_api_##n##_t_print, \ - sizeof(vl_api_##n##_t), 1); - foreach_vpe_api_reply_msg; -#undef _ - - /* API messages we can send */ -#define _(n,h) hash_set_mem (vam->function_by_name, #n, api_##n); - foreach_vpe_api_msg; -#undef _ - - /* Help strings */ -#define _(n,h) hash_set_mem (vam->help_by_name, #n, h); - foreach_vpe_api_msg; -#undef _ -} - -clib_error_t * vat_plugin_register (vat_main_t *vam) -{ - sample_test_main_t * sm = &sample_test_main; - u8 * name; - - sm->vat_main = vam; - - name = format (0, "sample_%08x%c", api_version, 0); - sm->msg_id_base = vl_client_get_first_plugin_msg_id ((char *) name); - - if (sm->msg_id_base != (u16) ~0) - vat_api_hookup (vam); - - vec_free(name); - - return 0; -} diff --git a/plugins/snat-plugin/Makefile.am b/plugins/snat-plugin/Makefile.am deleted file mode 100644 index e1922eb0..00000000 --- a/plugins/snat-plugin/Makefile.am +++ /dev/null @@ -1,113 +0,0 @@ - -# Copyright (c) -# Licensed under the Apache License, Version 2.0 (the "License"); -# you may not use this file except in compliance with the License. -# You may obtain a copy of the License at: -# -# http://www.apache.org/licenses/LICENSE-2.0 -# -# Unless required by applicable law or agreed to in writing, software -# distributed under the License is distributed on an "AS IS" BASIS, -# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. -# See the License for the specific language governing permissions and -# limitations under the License. - -AUTOMAKE_OPTIONS = foreign subdir-objects - -AM_CFLAGS = -Wall @DPDK@ -AM_LDFLAGS = -module -shared -avoid-version - -vppapitestpluginsdir = ${libdir}/vpp_api_test_plugins -vpppluginsdir = ${libdir}/vpp_plugins - -vppapitestplugins_LTLIBRARIES = snat_test_plugin.la -vppplugins_LTLIBRARIES = snat_plugin.la - -snat_plugin_la_SOURCES = snat/snat.c \ - snat/in2out.c \ - snat/out2in.c \ - snat/snat_plugin.api.h - -BUILT_SOURCES = snat/snat.api.h snat/snat.api.json - -SUFFIXES = .api.h .api - -%.api.h: %.api - mkdir -p `dirname $@` ; \ - $(CC) $(CPPFLAGS) -E -P -C -x c $^ \ - | vppapigen --input - --output $@ --show-name $@ - -%.api.json: %.api - @echo " JSON APIGEN " $@ ; \ - mkdir -p `dirname $@` ; \ - $(CC) $(CPPFLAGS) -E -P -C -x c $^ \ - | vppapigen --input - --json $@ - -apidir = $(prefix)/snat/ -api_DATA = snat.api.json - -noinst_HEADERS = \ - snat/snat_all_api_h.h \ - snat/snat_msg_enum.h \ - snat/snat.api.h - -snat_test_plugin_la_SOURCES = \ - snat/snat_test.c snat/snat_plugin.api.h - -# Remove *.la files -install-data-hook: - @(cd $(vpppluginsdir) && $(RM) $(vppplugins_LTLIBRARIES)) - @(cd $(vppapitestpluginsdir) && $(RM) $(vppapitestplugins_LTLIBRARIES)) - -# -# Java code generation -# -jvpp_registry_root = ../../vpp-api/java -jvpp_registry_version = 17.04 -jsnat_jarfile = jvpp-snat-$(PACKAGE_VERSION).jar -jvpp_package_dir = io/fd/vpp/jvpp/snat -jvpp_root = snat/jvpp -jvpp_target_dir = target -jvpp_target = $(jvpp_root)/$(jvpp_target_dir) -api_file=$(srcdir)/snat/snat.api - -lib_LTLIBRARIES = libjvpp_snat.la -libjvpp_snat_la_SOURCES = snat/snat.api.h snat/jvpp_snat.c snat/jvpp/io_fd_vpp_jvpp_snat_JVppSnatImpl.h -libjvpp_snat_la_LIBADD = -lvlibmemoryclient -lvlibapi -lvppinfra \ - -lpthread -lm -lrt -L$(jvpp_registry_root)/.libs -ljvpp_common -libjvpp_snat_la_LDFLAGS = -module -libjvpp_snat_la_CPPFLAGS = -I$(JAVA_HOME)/include -I$(JAVA_HOME)/include/linux -I../ -I$(srcdir)/../ - -BUILT_SOURCES += $(jvpp_root)/io_fd_vpp_jvpp_snat_JVppSnatImpl.h - -$(jvpp_root)/io_fd_vpp_jvpp_snat_JVppSnatImpl.h: snat.api.json - dir=`pwd`; \ - mkdir -p $(jvpp_target); \ - mkdir -p $(jvpp_root)/$(jvpp_package_dir); \ - cd $(jvpp_root)/$(jvpp_package_dir); \ - mkdir -p dto future callfacade callback notification test; \ - @srcdir@/$(jvpp_registry_root)/jvpp/gen/jvpp_gen.py -i $${dir}/snat.api.json --plugin_name snat; \ - cd -; \ - mv -f $(jvpp_root)/$(jvpp_package_dir)/jvpp_snat_gen.h $(jvpp_root)/jvpp_snat_gen.h; \ - cp $(srcdir)/$(jvpp_root)/$(jvpp_package_dir)/test/*.java $(jvpp_root)/$(jvpp_package_dir)/test/; \ - cd $(jvpp_root); \ - $(JAVAC) -classpath .:$(jvpp_target_dir):../../$(jvpp_registry_root)/jvpp-registry-$(jvpp_registry_version).jar -d $(jvpp_target_dir) $(jvpp_package_dir)/*.java \ - $(jvpp_package_dir)/dto/*.java \ - $(jvpp_package_dir)/callback/*.java \ - $(jvpp_package_dir)/notification/*.java \ - $(jvpp_package_dir)/future/*.java \ - $(jvpp_package_dir)/callfacade/*.java \ - $(jvpp_package_dir)/test/*.java \ - || (echo "snat jvpp compilation failed: $$?"; exit 1); \ - $(JAVAH) -classpath .:$(jvpp_target_dir):../../$(jvpp_registry_root)/jvpp-registry-$(jvpp_registry_version).jar -d . io.fd.vpp.jvpp.snat.JVppSnatImpl ; - -$(jsnat_jarfile): libjvpp_snat.la - cp .libs/libjvpp_snat.so.0.0.0 $(jvpp_target); \ - cd $(jvpp_target); \ - $(JAR) cfv $(JARFLAGS) ../../../$@ libjvpp_snat.so.0.0.0 $(jvpp_package_dir)/* ; cd ..; - -snat.api.json: - @echo " jSnat_sfc API"; \ - vppapigen --input $(api_file) --json snat.api.json; - -all-local: $(jsnat_jarfile) diff --git a/plugins/snat-plugin/configure.ac b/plugins/snat-plugin/configure.ac deleted file mode 100644 index 011246cd..00000000 --- a/plugins/snat-plugin/configure.ac +++ /dev/null @@ -1,32 +0,0 @@ -AC_INIT(snat_plugin, 1.0) -AC_CONFIG_MACRO_DIR([../../vpp-api/java/m4]) -AM_INIT_AUTOMAKE -AM_SILENT_RULES([yes]) -AC_PREFIX_DEFAULT([/usr]) - -AC_PROG_LIBTOOL -AC_PROG_CC - -if test -f /usr/bin/lsb_release && test `lsb_release -si` == "Ubuntu" && test `lsb_release -sr` == "14.04" && test -d /usr/lib/jvm/java-8-openjdk-amd64/ ; then - JAVA_HOME=/usr/lib/jvm/java-8-openjdk-amd64/ - JAVAC=${JAVA_HOME}/bin/javac - PATH=${JAVA_HOME}/bin/:${PATH} - break -fi - -AX_CHECK_JAVA_HOME -AX_PROG_JAVAC -AX_PROG_JAVAH -AX_PROG_JAR -AX_PROG_JAVADOC -AX_PROG_JAVA - -AC_ARG_WITH(dpdk, - AC_HELP_STRING([--with-dpdk],[Use DPDK]), - [with_dpdk=1], - [with_dpdk=0]) - -AM_CONDITIONAL(WITH_DPDK, test "$with_dpdk" = "1") -AC_SUBST(DPDK,["-DDPDK=${with_dpdk}"]) - -AC_OUTPUT([Makefile]) diff --git a/plugins/snat-plugin/snat/in2out.c b/plugins/snat-plugin/snat/in2out.c deleted file mode 100644 index c78fdd76..00000000 --- a/plugins/snat-plugin/snat/in2out.c +++ /dev/null @@ -1,1597 +0,0 @@ -/* - * Copyright (c) 2016 Cisco and/or its affiliates. - * Licensed under the Apache License, Version 2.0 (the "License"); - * you may not use this file except in compliance with the License. - * You may obtain a copy of the License at: - * - * http://www.apache.org/licenses/LICENSE-2.0 - * - * Unless required by applicable law or agreed to in writing, software - * distributed under the License is distributed on an "AS IS" BASIS, - * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. - * See the License for the specific language governing permissions and - * limitations under the License. - */ - -#include -#include -#include -#include - -#include -#include -#include -#include - -#include -#include -#include - -typedef struct { - u32 sw_if_index; - u32 next_index; - u32 session_index; - u32 is_slow_path; -} snat_in2out_trace_t; - -typedef struct { - u32 next_worker_index; - u8 do_handoff; -} snat_in2out_worker_handoff_trace_t; - -/* packet trace format function */ -static u8 * format_snat_in2out_trace (u8 * s, va_list * args) -{ - CLIB_UNUSED (vlib_main_t * vm) = va_arg (*args, vlib_main_t *); - CLIB_UNUSED (vlib_node_t * node) = va_arg (*args, vlib_node_t *); - snat_in2out_trace_t * t = va_arg (*args, snat_in2out_trace_t *); - char * tag; - - tag = t->is_slow_path ? "SNAT_IN2OUT_SLOW_PATH" : "SNAT_IN2OUT_FAST_PATH"; - - s = format (s, "%s: sw_if_index %d, next index %d, session %d", tag, - t->sw_if_index, t->next_index, t->session_index); - - return s; -} - -static u8 * format_snat_in2out_fast_trace (u8 * s, va_list * args) -{ - CLIB_UNUSED (vlib_main_t * vm) = va_arg (*args, vlib_main_t *); - CLIB_UNUSED (vlib_node_t * node) = va_arg (*args, vlib_node_t *); - snat_in2out_trace_t * t = va_arg (*args, snat_in2out_trace_t *); - - s = format (s, "SANT_IN2OUT_FAST: sw_if_index %d, next index %d", - t->sw_if_index, t->next_index); - - return s; -} - -static u8 * format_snat_in2out_worker_handoff_trace (u8 * s, va_list * args) -{ - CLIB_UNUSED (vlib_main_t * vm) = va_arg (*args, vlib_main_t *); - CLIB_UNUSED (vlib_node_t * node) = va_arg (*args, vlib_node_t *); - snat_in2out_worker_handoff_trace_t * t = - va_arg (*args, snat_in2out_worker_handoff_trace_t *); - char * m; - - m = t->do_handoff ? "next worker" : "same worker"; - s = format (s, "SNAT_IN2OUT_WORKER_HANDOFF: %s %d", m, t->next_worker_index); - - return s; -} - -vlib_node_registration_t snat_in2out_node; -vlib_node_registration_t snat_in2out_slowpath_node; -vlib_node_registration_t snat_in2out_fast_node; -vlib_node_registration_t snat_in2out_worker_handoff_node; - -#define foreach_snat_in2out_error \ -_(UNSUPPORTED_PROTOCOL, "Unsupported protocol") \ -_(IN2OUT_PACKETS, "Good in2out packets processed") \ -_(OUT_OF_PORTS, "Out of ports") \ -_(BAD_OUTSIDE_FIB, "Outside VRF ID not found") \ -_(BAD_ICMP_TYPE, "icmp type not echo-request") \ -_(NO_TRANSLATION, "No translation") - -typedef enum { -#define _(sym,str) SNAT_IN2OUT_ERROR_##sym, - foreach_snat_in2out_error -#undef _ - SNAT_IN2OUT_N_ERROR, -} snat_in2out_error_t; - -static char * snat_in2out_error_strings[] = { -#define _(sym,string) string, - foreach_snat_in2out_error -#undef _ -}; - -typedef enum { - SNAT_IN2OUT_NEXT_LOOKUP, - SNAT_IN2OUT_NEXT_DROP, - SNAT_IN2OUT_NEXT_SLOW_PATH, - SNAT_IN2OUT_N_NEXT, -} snat_in2out_next_t; - -static u32 slow_path (snat_main_t *sm, vlib_buffer_t *b0, - ip4_header_t * ip0, - u32 rx_fib_index0, - snat_session_key_t * key0, - snat_session_t ** sessionp, - vlib_node_runtime_t * node, - u32 next0, - u32 cpu_index) -{ - snat_user_t *u; - snat_user_key_t user_key; - snat_session_t *s; - clib_bihash_kv_8_8_t kv0, value0; - u32 oldest_per_user_translation_list_index; - dlist_elt_t * oldest_per_user_translation_list_elt; - dlist_elt_t * per_user_translation_list_elt; - dlist_elt_t * per_user_list_head_elt; - u32 session_index; - snat_session_key_t key1; - u32 address_index = ~0; - u32 outside_fib_index; - uword * p; - snat_static_mapping_key_t worker_by_out_key; - - p = hash_get (sm->ip4_main->fib_index_by_table_id, sm->outside_vrf_id); - if (! p) - { - b0->error = node->errors[SNAT_IN2OUT_ERROR_BAD_OUTSIDE_FIB]; - return SNAT_IN2OUT_NEXT_DROP; - } - outside_fib_index = p[0]; - - user_key.addr = ip0->src_address; - user_key.fib_index = rx_fib_index0; - kv0.key = user_key.as_u64; - - /* Ever heard of the "user" = src ip4 address before? */ - if (clib_bihash_search_8_8 (&sm->user_hash, &kv0, &value0)) - { - /* no, make a new one */ - pool_get (sm->per_thread_data[cpu_index].users, u); - memset (u, 0, sizeof (*u)); - u->addr = ip0->src_address; - - pool_get (sm->per_thread_data[cpu_index].list_pool, per_user_list_head_elt); - - u->sessions_per_user_list_head_index = per_user_list_head_elt - - sm->per_thread_data[cpu_index].list_pool; - - clib_dlist_init (sm->per_thread_data[cpu_index].list_pool, - u->sessions_per_user_list_head_index); - - kv0.value = u - sm->per_thread_data[cpu_index].users; - - /* add user */ - clib_bihash_add_del_8_8 (&sm->user_hash, &kv0, 1 /* is_add */); - } - else - { - u = pool_elt_at_index (sm->per_thread_data[cpu_index].users, - value0.value); - } - - /* Over quota? Recycle the least recently used dynamic translation */ - if (u->nsessions >= sm->max_translations_per_user) - { - /* Remove the oldest dynamic translation */ - do { - oldest_per_user_translation_list_index = - clib_dlist_remove_head (sm->per_thread_data[cpu_index].list_pool, - u->sessions_per_user_list_head_index); - - ASSERT (oldest_per_user_translation_list_index != ~0); - - /* add it back to the end of the LRU list */ - clib_dlist_addtail (sm->per_thread_data[cpu_index].list_pool, - u->sessions_per_user_list_head_index, - oldest_per_user_translation_list_index); - /* Get the list element */ - oldest_per_user_translation_list_elt = - pool_elt_at_index (sm->per_thread_data[cpu_index].list_pool, - oldest_per_user_translation_list_index); - - /* Get the session index from the list element */ - session_index = oldest_per_user_translation_list_elt->value; - - /* Get the session */ - s = pool_elt_at_index (sm->per_thread_data[cpu_index].sessions, - session_index); - } while (snat_is_session_static (s)); - - /* Remove in2out, out2in keys */ - kv0.key = s->in2out.as_u64; - if (clib_bihash_add_del_8_8 (&sm->in2out, &kv0, 0 /* is_add */)) - clib_warning ("in2out key delete failed"); - kv0.key = s->out2in.as_u64; - if (clib_bihash_add_del_8_8 (&sm->out2in, &kv0, 0 /* is_add */)) - clib_warning ("out2in key delete failed"); - - snat_free_outside_address_and_port - (sm, &s->out2in, s->outside_address_index); - s->outside_address_index = ~0; - - if (snat_alloc_outside_address_and_port (sm, &key1, &address_index)) - { - ASSERT(0); - - b0->error = node->errors[SNAT_IN2OUT_ERROR_OUT_OF_PORTS]; - return SNAT_IN2OUT_NEXT_DROP; - } - s->outside_address_index = address_index; - } - else - { - u8 static_mapping = 1; - - /* First try to match static mapping by local address and port */ - if (snat_static_mapping_match (sm, *key0, &key1, 0)) - { - static_mapping = 0; - /* Try to create dynamic translation */ - if (snat_alloc_outside_address_and_port (sm, &key1, &address_index)) - { - b0->error = node->errors[SNAT_IN2OUT_ERROR_OUT_OF_PORTS]; - return SNAT_IN2OUT_NEXT_DROP; - } - } - - /* Create a new session */ - pool_get (sm->per_thread_data[cpu_index].sessions, s); - memset (s, 0, sizeof (*s)); - - s->outside_address_index = address_index; - - if (static_mapping) - { - u->nstaticsessions++; - s->flags |= SNAT_SESSION_FLAG_STATIC_MAPPING; - } - else - { - u->nsessions++; - } - - /* Create list elts */ - pool_get (sm->per_thread_data[cpu_index].list_pool, - per_user_translation_list_elt); - clib_dlist_init (sm->per_thread_data[cpu_index].list_pool, - per_user_translation_list_elt - - sm->per_thread_data[cpu_index].list_pool); - - per_user_translation_list_elt->value = - s - sm->per_thread_data[cpu_index].sessions; - s->per_user_index = per_user_translation_list_elt - - sm->per_thread_data[cpu_index].list_pool; - s->per_user_list_head_index = u->sessions_per_user_list_head_index; - - clib_dlist_addtail (sm->per_thread_data[cpu_index].list_pool, - s->per_user_list_head_index, - per_user_translation_list_elt - - sm->per_thread_data[cpu_index].list_pool); - } - - s->in2out = *key0; - s->out2in = key1; - s->out2in.protocol = key0->protocol; - s->out2in.fib_index = outside_fib_index; - *sessionp = s; - - /* Add to translation hashes */ - kv0.key = s->in2out.as_u64; - kv0.value = s - sm->per_thread_data[cpu_index].sessions; - if (clib_bihash_add_del_8_8 (&sm->in2out, &kv0, 1 /* is_add */)) - clib_warning ("in2out key add failed"); - - kv0.key = s->out2in.as_u64; - kv0.value = s - sm->per_thread_data[cpu_index].sessions; - - if (clib_bihash_add_del_8_8 (&sm->out2in, &kv0, 1 /* is_add */)) - clib_warning ("out2in key add failed"); - - /* Add to translated packets worker lookup */ - worker_by_out_key.addr = s->out2in.addr; - worker_by_out_key.port = s->out2in.port; - worker_by_out_key.fib_index = s->out2in.fib_index; - kv0.key = worker_by_out_key.as_u64; - kv0.value = cpu_index; - clib_bihash_add_del_8_8 (&sm->worker_by_out, &kv0, 1); - return next0; -} - -static inline u32 icmp_in2out_slow_path (snat_main_t *sm, - vlib_buffer_t * b0, - ip4_header_t * ip0, - icmp46_header_t * icmp0, - u32 sw_if_index0, - u32 rx_fib_index0, - vlib_node_runtime_t * node, - u32 next0, - f64 now, - u32 cpu_index) -{ - snat_session_key_t key0; - icmp_echo_header_t *echo0; - clib_bihash_kv_8_8_t kv0, value0; - snat_session_t * s0; - u32 new_addr0, old_addr0; - u16 old_id0, new_id0; - ip_csum_t sum0; - snat_runtime_t * rt = (snat_runtime_t *)node->runtime_data; - - if (PREDICT_FALSE(icmp0->type != ICMP4_echo_request)) - { - b0->error = node->errors[SNAT_IN2OUT_ERROR_BAD_ICMP_TYPE]; - return SNAT_IN2OUT_NEXT_DROP; - } - - echo0 = (icmp_echo_header_t *)(icmp0+1); - - key0.addr = ip0->src_address; - key0.port = echo0->identifier; - key0.protocol = SNAT_PROTOCOL_ICMP; - key0.fib_index = rx_fib_index0; - - kv0.key = key0.as_u64; - - if (clib_bihash_search_8_8 (&sm->in2out, &kv0, &value0)) - { - ip4_address_t * first_int_addr; - - if (PREDICT_FALSE(rt->cached_sw_if_index != sw_if_index0)) - { - first_int_addr = - ip4_interface_first_address (sm->ip4_main, sw_if_index0, - 0 /* just want the address */); - rt->cached_sw_if_index = sw_if_index0; - rt->cached_ip4_address = first_int_addr->as_u32; - } - - /* Don't NAT packet aimed at the intfc address */ - if (PREDICT_FALSE(ip0->dst_address.as_u32 == - rt->cached_ip4_address)) - return next0; - - next0 = slow_path (sm, b0, ip0, rx_fib_index0, &key0, - &s0, node, next0, cpu_index); - - if (PREDICT_FALSE (next0 == SNAT_IN2OUT_NEXT_DROP)) - return next0; - } - else - s0 = pool_elt_at_index (sm->per_thread_data[cpu_index].sessions, - value0.value); - - old_addr0 = ip0->src_address.as_u32; - ip0->src_address = s0->out2in.addr; - new_addr0 = ip0->src_address.as_u32; - vnet_buffer(b0)->sw_if_index[VLIB_TX] = s0->out2in.fib_index; - - sum0 = ip0->checksum; - sum0 = ip_csum_update (sum0, old_addr0, new_addr0, - ip4_header_t, - src_address /* changed member */); - ip0->checksum = ip_csum_fold (sum0); - - old_id0 = echo0->identifier; - new_id0 = s0->out2in.port; - echo0->identifier = new_id0; - - sum0 = icmp0->checksum; - sum0 = ip_csum_update (sum0, old_id0, new_id0, icmp_echo_header_t, - identifier); - icmp0->checksum = ip_csum_fold (sum0); - - /* Accounting */ - s0->last_heard = now; - s0->total_pkts++; - s0->total_bytes += vlib_buffer_length_in_chain (sm->vlib_main, b0); - /* Per-user LRU list maintenance for dynamic translations */ - if (!snat_is_session_static (s0)) - { - clib_dlist_remove (sm->per_thread_data[cpu_index].list_pool, - s0->per_user_index); - clib_dlist_addtail (sm->per_thread_data[cpu_index].list_pool, - s0->per_user_list_head_index, - s0->per_user_index); - } - - return next0; -} - -/** - * @brief Hairpinning - * - * Hairpinning allows two endpoints on the internal side of the NAT to - * communicate even if they only use each other's external IP addresses - * and ports. - * - * @param sm SNAT main. - * @param b0 Vlib buffer. - * @param ip0 IP header. - * @param udp0 UDP header. - * @param tcp0 TCP header. - * @param proto0 SNAT protocol. - */ -static inline void -snat_hairpinning (snat_main_t *sm, - vlib_buffer_t * b0, - ip4_header_t * ip0, - udp_header_t * udp0, - tcp_header_t * tcp0, - u32 proto0) -{ - snat_session_key_t key0, sm0; - snat_static_mapping_key_t k0; - snat_session_t * s0; - clib_bihash_kv_8_8_t kv0, value0; - ip_csum_t sum0; - u32 new_dst_addr0 = 0, old_dst_addr0, ti = 0, si; - u16 new_dst_port0, old_dst_port0; - - key0.addr = ip0->dst_address; - key0.port = udp0->dst_port; - key0.protocol = proto0; - key0.fib_index = sm->outside_fib_index; - kv0.key = key0.as_u64; - - /* Check if destination is in active sessions */ - if (clib_bihash_search_8_8 (&sm->out2in, &kv0, &value0)) - { - /* or static mappings */ - if (!snat_static_mapping_match(sm, key0, &sm0, 1)) - { - new_dst_addr0 = sm0.addr.as_u32; - new_dst_port0 = sm0.port; - vnet_buffer(b0)->sw_if_index[VLIB_TX] = sm0.fib_index; - } - } - else - { - si = value0.value; - if (sm->num_workers > 1) - { - k0.addr = ip0->dst_address; - k0.port = udp0->dst_port; - k0.fib_index = sm->outside_fib_index; - kv0.key = k0.as_u64; - if (clib_bihash_search_8_8 (&sm->worker_by_out, &kv0, &value0)) - ASSERT(0); - else - ti = value0.value; - } - else - ti = sm->num_workers; - - s0 = pool_elt_at_index (sm->per_thread_data[ti].sessions, si); - new_dst_addr0 = s0->in2out.addr.as_u32; - new_dst_port0 = s0->in2out.port; - vnet_buffer(b0)->sw_if_index[VLIB_TX] = s0->in2out.fib_index; - } - - /* Destination is behind the same NAT, use internal address and port */ - if (new_dst_addr0) - { - old_dst_addr0 = ip0->dst_address.as_u32; - ip0->dst_address.as_u32 = new_dst_addr0; - sum0 = ip0->checksum; - sum0 = ip_csum_update (sum0, old_dst_addr0, new_dst_addr0, - ip4_header_t, dst_address); - ip0->checksum = ip_csum_fold (sum0); - - old_dst_port0 = tcp0->ports.dst; - if (PREDICT_TRUE(new_dst_port0 != old_dst_port0)) - { - if (PREDICT_TRUE(proto0 == SNAT_PROTOCOL_TCP)) - { - tcp0->ports.dst = new_dst_port0; - sum0 = tcp0->checksum; - sum0 = ip_csum_update (sum0, old_dst_addr0, new_dst_addr0, - ip4_header_t, dst_address); - sum0 = ip_csum_update (sum0, old_dst_port0, new_dst_port0, - ip4_header_t /* cheat */, length); - tcp0->checksum = ip_csum_fold(sum0); - } - else - { - udp0->dst_port = new_dst_port0; - udp0->checksum = 0; - } - } - } -} - -static inline uword -snat_in2out_node_fn_inline (vlib_main_t * vm, - vlib_node_runtime_t * node, - vlib_frame_t * frame, int is_slow_path) -{ - u32 n_left_from, * from, * to_next; - snat_in2out_next_t next_index; - u32 pkts_processed = 0; - snat_main_t * sm = &snat_main; - snat_runtime_t * rt = (snat_runtime_t *)node->runtime_data; - f64 now = vlib_time_now (vm); - u32 stats_node_index; - u32 cpu_index = os_get_cpu_number (); - - stats_node_index = is_slow_path ? snat_in2out_slowpath_node.index : - snat_in2out_node.index; - - from = vlib_frame_vector_args (frame); - n_left_from = frame->n_vectors; - next_index = node->cached_next_index; - - while (n_left_from > 0) - { - u32 n_left_to_next; - - vlib_get_next_frame (vm, node, next_index, - to_next, n_left_to_next); - - while (n_left_from >= 4 && n_left_to_next >= 2) - { - u32 bi0, bi1; - vlib_buffer_t * b0, * b1; - u32 next0, next1; - u32 sw_if_index0, sw_if_index1; - ip4_header_t * ip0, * ip1; - ip_csum_t sum0, sum1; - u32 new_addr0, old_addr0, new_addr1, old_addr1; - u16 old_port0, new_port0, old_port1, new_port1; - udp_header_t * udp0, * udp1; - tcp_header_t * tcp0, * tcp1; - icmp46_header_t * icmp0, * icmp1; - snat_session_key_t key0, key1; - u32 rx_fib_index0, rx_fib_index1; - u32 proto0, proto1; - snat_session_t * s0 = 0, * s1 = 0; - clib_bihash_kv_8_8_t kv0, value0, kv1, value1; - - /* Prefetch next iteration. */ - { - vlib_buffer_t * p2, * p3; - - p2 = vlib_get_buffer (vm, from[2]); - p3 = vlib_get_buffer (vm, from[3]); - - vlib_prefetch_buffer_header (p2, LOAD); - vlib_prefetch_buffer_header (p3, LOAD); - - CLIB_PREFETCH (p2->data, CLIB_CACHE_LINE_BYTES, STORE); - CLIB_PREFETCH (p3->data, CLIB_CACHE_LINE_BYTES, STORE); - } - - /* speculatively enqueue b0 and b1 to the current next frame */ - to_next[0] = bi0 = from[0]; - to_next[1] = bi1 = from[1]; - from += 2; - to_next += 2; - n_left_from -= 2; - n_left_to_next -= 2; - - b0 = vlib_get_buffer (vm, bi0); - b1 = vlib_get_buffer (vm, bi1); - - ip0 = vlib_buffer_get_current (b0); - udp0 = ip4_next_header (ip0); - tcp0 = (tcp_header_t *) udp0; - icmp0 = (icmp46_header_t *) udp0; - - sw_if_index0 = vnet_buffer(b0)->sw_if_index[VLIB_RX]; - rx_fib_index0 = vec_elt (sm->ip4_main->fib_index_by_sw_if_index, - sw_if_index0); - - next0 = next1 = SNAT_IN2OUT_NEXT_LOOKUP; - - proto0 = ~0; - proto0 = (ip0->protocol == IP_PROTOCOL_UDP) - ? SNAT_PROTOCOL_UDP : proto0; - proto0 = (ip0->protocol == IP_PROTOCOL_TCP) - ? SNAT_PROTOCOL_TCP : proto0; - proto0 = (ip0->protocol == IP_PROTOCOL_ICMP) - ? SNAT_PROTOCOL_ICMP : proto0; - - /* Next configured feature, probably ip4-lookup */ - if (is_slow_path) - { - if (PREDICT_FALSE (proto0 == ~0)) - goto trace00; - - if (PREDICT_FALSE (proto0 == SNAT_PROTOCOL_ICMP)) - { - next0 = icmp_in2out_slow_path - (sm, b0, ip0, icmp0, sw_if_index0, rx_fib_index0, - node, next0, now, cpu_index); - goto trace00; - } - } - else - { - if (PREDICT_FALSE (proto0 == ~0 || proto0 == SNAT_PROTOCOL_ICMP)) - { - next0 = SNAT_IN2OUT_NEXT_SLOW_PATH; - goto trace00; - } - } - - key0.addr = ip0->src_address; - key0.port = udp0->src_port; - key0.protocol = proto0; - key0.fib_index = rx_fib_index0; - - kv0.key = key0.as_u64; - - if (PREDICT_FALSE (clib_bihash_search_8_8 (&sm->in2out, &kv0, &value0) != 0)) - { - if (is_slow_path) - { - ip4_address_t * first_int_addr; - - if (PREDICT_FALSE(rt->cached_sw_if_index != sw_if_index0)) - { - first_int_addr = - ip4_interface_first_address (sm->ip4_main, sw_if_index0, - 0 /* just want the address */); - rt->cached_sw_if_index = sw_if_index0; - rt->cached_ip4_address = first_int_addr->as_u32; - } - - /* Don't NAT packet aimed at the intfc address */ - if (PREDICT_FALSE(ip0->dst_address.as_u32 == - rt->cached_ip4_address)) - goto trace00; - - next0 = slow_path (sm, b0, ip0, rx_fib_index0, &key0, - &s0, node, next0, cpu_index); - if (PREDICT_FALSE (next0 == SNAT_IN2OUT_NEXT_DROP)) - goto trace00; - } - else - { - next0 = SNAT_IN2OUT_NEXT_SLOW_PATH; - goto trace00; - } - } - else - s0 = pool_elt_at_index (sm->per_thread_data[cpu_index].sessions, - value0.value); - - old_addr0 = ip0->src_address.as_u32; - ip0->src_address = s0->out2in.addr; - new_addr0 = ip0->src_address.as_u32; - vnet_buffer(b0)->sw_if_index[VLIB_TX] = s0->out2in.fib_index; - - sum0 = ip0->checksum; - sum0 = ip_csum_update (sum0, old_addr0, new_addr0, - ip4_header_t, - src_address /* changed member */); - ip0->checksum = ip_csum_fold (sum0); - - if (PREDICT_TRUE(proto0 == SNAT_PROTOCOL_TCP)) - { - old_port0 = tcp0->ports.src; - tcp0->ports.src = s0->out2in.port; - new_port0 = tcp0->ports.src; - - sum0 = tcp0->checksum; - sum0 = ip_csum_update (sum0, old_addr0, new_addr0, - ip4_header_t, - dst_address /* changed member */); - sum0 = ip_csum_update (sum0, old_port0, new_port0, - ip4_header_t /* cheat */, - length /* changed member */); - tcp0->checksum = ip_csum_fold(sum0); - } - else - { - old_port0 = udp0->src_port; - udp0->src_port = s0->out2in.port; - udp0->checksum = 0; - } - - /* Hairpinning */ - snat_hairpinning (sm, b0, ip0, udp0, tcp0, proto0); - - /* Accounting */ - s0->last_heard = now; - s0->total_pkts++; - s0->total_bytes += vlib_buffer_length_in_chain (vm, b0); - /* Per-user LRU list maintenance for dynamic translation */ - if (!snat_is_session_static (s0)) - { - clib_dlist_remove (sm->per_thread_data[cpu_index].list_pool, - s0->per_user_index); - clib_dlist_addtail (sm->per_thread_data[cpu_index].list_pool, - s0->per_user_list_head_index, - s0->per_user_index); - } - trace00: - - if (PREDICT_FALSE((node->flags & VLIB_NODE_FLAG_TRACE) - && (b0->flags & VLIB_BUFFER_IS_TRACED))) - { - snat_in2out_trace_t *t = - vlib_add_trace (vm, node, b0, sizeof (*t)); - t->is_slow_path = is_slow_path; - t->sw_if_index = sw_if_index0; - t->next_index = next0; - t->session_index = ~0; - if (s0) - t->session_index = s0 - sm->per_thread_data[cpu_index].sessions; - } - - pkts_processed += next0 != SNAT_IN2OUT_NEXT_DROP; - - ip1 = vlib_buffer_get_current (b1); - udp1 = ip4_next_header (ip1); - tcp1 = (tcp_header_t *) udp1; - icmp1 = (icmp46_header_t *) udp1; - - sw_if_index1 = vnet_buffer(b1)->sw_if_index[VLIB_RX]; - rx_fib_index1 = vec_elt (sm->ip4_main->fib_index_by_sw_if_index, - sw_if_index1); - - proto1 = ~0; - proto1 = (ip1->protocol == IP_PROTOCOL_UDP) - ? SNAT_PROTOCOL_UDP : proto1; - proto1 = (ip1->protocol == IP_PROTOCOL_TCP) - ? SNAT_PROTOCOL_TCP : proto1; - proto1 = (ip1->protocol == IP_PROTOCOL_ICMP) - ? SNAT_PROTOCOL_ICMP : proto1; - - /* Next configured feature, probably ip4-lookup */ - if (is_slow_path) - { - if (PREDICT_FALSE (proto1 == ~0)) - goto trace01; - - if (PREDICT_FALSE (proto1 == SNAT_PROTOCOL_ICMP)) - { - next1 = icmp_in2out_slow_path - (sm, b1, ip1, icmp1, sw_if_index1, rx_fib_index1, node, - next1, now, cpu_index); - goto trace01; - } - } - else - { - if (PREDICT_FALSE (proto1 == ~0 || proto1 == SNAT_PROTOCOL_ICMP)) - { - next1 = SNAT_IN2OUT_NEXT_SLOW_PATH; - goto trace01; - } - } - - key1.addr = ip1->src_address; - key1.port = udp1->src_port; - key1.protocol = proto1; - key1.fib_index = rx_fib_index1; - - kv1.key = key1.as_u64; - - if (PREDICT_FALSE(clib_bihash_search_8_8 (&sm->in2out, &kv1, &value1) != 0)) - { - if (is_slow_path) - { - ip4_address_t * first_int_addr; - - if (PREDICT_FALSE(rt->cached_sw_if_index != sw_if_index1)) - { - first_int_addr = - ip4_interface_first_address (sm->ip4_main, sw_if_index1, - 0 /* just want the address */); - rt->cached_sw_if_index = sw_if_index1; - rt->cached_ip4_address = first_int_addr->as_u32; - } - - /* Don't NAT packet aimed at the intfc address */ - if (PREDICT_FALSE(ip1->dst_address.as_u32 == - rt->cached_ip4_address)) - goto trace01; - - next1 = slow_path (sm, b1, ip1, rx_fib_index1, &key1, - &s1, node, next1, cpu_index); - if (PREDICT_FALSE (next1 == SNAT_IN2OUT_NEXT_DROP)) - goto trace01; - } - else - { - next1 = SNAT_IN2OUT_NEXT_SLOW_PATH; - goto trace01; - } - } - else - s1 = pool_elt_at_index (sm->per_thread_data[cpu_index].sessions, - value1.value); - - old_addr1 = ip1->src_address.as_u32; - ip1->src_address = s1->out2in.addr; - new_addr1 = ip1->src_address.as_u32; - vnet_buffer(b1)->sw_if_index[VLIB_TX] = s1->out2in.fib_index; - - sum1 = ip1->checksum; - sum1 = ip_csum_update (sum1, old_addr1, new_addr1, - ip4_header_t, - src_address /* changed member */); - ip1->checksum = ip_csum_fold (sum1); - - if (PREDICT_TRUE(proto1 == SNAT_PROTOCOL_TCP)) - { - old_port1 = tcp1->ports.src; - tcp1->ports.src = s1->out2in.port; - new_port1 = tcp1->ports.src; - - sum1 = tcp1->checksum; - sum1 = ip_csum_update (sum1, old_addr1, new_addr1, - ip4_header_t, - dst_address /* changed member */); - sum1 = ip_csum_update (sum1, old_port1, new_port1, - ip4_header_t /* cheat */, - length /* changed member */); - tcp1->checksum = ip_csum_fold(sum1); - } - else - { - old_port1 = udp1->src_port; - udp1->src_port = s1->out2in.port; - udp1->checksum = 0; - } - - /* Hairpinning */ - snat_hairpinning (sm, b1, ip1, udp1, tcp1, proto1); - - /* Accounting */ - s1->last_heard = now; - s1->total_pkts++; - s1->total_bytes += vlib_buffer_length_in_chain (vm, b1); - /* Per-user LRU list maintenance for dynamic translation */ - if (!snat_is_session_static (s1)) - { - clib_dlist_remove (sm->per_thread_data[cpu_index].list_pool, - s1->per_user_index); - clib_dlist_addtail (sm->per_thread_data[cpu_index].list_pool, - s1->per_user_list_head_index, - s1->per_user_index); - } - trace01: - - if (PREDICT_FALSE((node->flags & VLIB_NODE_FLAG_TRACE) - && (b1->flags & VLIB_BUFFER_IS_TRACED))) - { - snat_in2out_trace_t *t = - vlib_add_trace (vm, node, b1, sizeof (*t)); - t->sw_if_index = sw_if_index1; - t->next_index = next1; - t->session_index = ~0; - if (s1) - t->session_index = s1 - sm->per_thread_data[cpu_index].sessions; - } - - pkts_processed += next1 != SNAT_IN2OUT_NEXT_DROP; - - /* verify speculative enqueues, maybe switch current next frame */ - vlib_validate_buffer_enqueue_x2 (vm, node, next_index, - to_next, n_left_to_next, - bi0, bi1, next0, next1); - } - - while (n_left_from > 0 && n_left_to_next > 0) - { - u32 bi0; - vlib_buffer_t * b0; - u32 next0; - u32 sw_if_index0; - ip4_header_t * ip0; - ip_csum_t sum0; - u32 new_addr0, old_addr0; - u16 old_port0, new_port0; - udp_header_t * udp0; - tcp_header_t * tcp0; - icmp46_header_t * icmp0; - snat_session_key_t key0; - u32 rx_fib_index0; - u32 proto0; - snat_session_t * s0 = 0; - clib_bihash_kv_8_8_t kv0, value0; - - /* speculatively enqueue b0 to the current next frame */ - bi0 = from[0]; - to_next[0] = bi0; - from += 1; - to_next += 1; - n_left_from -= 1; - n_left_to_next -= 1; - - b0 = vlib_get_buffer (vm, bi0); - next0 = SNAT_IN2OUT_NEXT_LOOKUP; - - ip0 = vlib_buffer_get_current (b0); - udp0 = ip4_next_header (ip0); - tcp0 = (tcp_header_t *) udp0; - icmp0 = (icmp46_header_t *) udp0; - - sw_if_index0 = vnet_buffer(b0)->sw_if_index[VLIB_RX]; - rx_fib_index0 = vec_elt (sm->ip4_main->fib_index_by_sw_if_index, - sw_if_index0); - - proto0 = ~0; - proto0 = (ip0->protocol == IP_PROTOCOL_UDP) - ? SNAT_PROTOCOL_UDP : proto0; - proto0 = (ip0->protocol == IP_PROTOCOL_TCP) - ? SNAT_PROTOCOL_TCP : proto0; - proto0 = (ip0->protocol == IP_PROTOCOL_ICMP) - ? SNAT_PROTOCOL_ICMP : proto0; - - /* Next configured feature, probably ip4-lookup */ - if (is_slow_path) - { - if (PREDICT_FALSE (proto0 == ~0)) - goto trace0; - - if (PREDICT_FALSE (proto0 == SNAT_PROTOCOL_ICMP)) - { - next0 = icmp_in2out_slow_path - (sm, b0, ip0, icmp0, sw_if_index0, rx_fib_index0, node, - next0, now, cpu_index); - goto trace0; - } - } - else - { - if (PREDICT_FALSE (proto0 == ~0 || proto0 == SNAT_PROTOCOL_ICMP)) - { - next0 = SNAT_IN2OUT_NEXT_SLOW_PATH; - goto trace0; - } - } - - key0.addr = ip0->src_address; - key0.port = udp0->src_port; - key0.protocol = proto0; - key0.fib_index = rx_fib_index0; - - kv0.key = key0.as_u64; - - if (clib_bihash_search_8_8 (&sm->in2out, &kv0, &value0)) - { - if (is_slow_path) - { - ip4_address_t * first_int_addr; - - if (PREDICT_FALSE(rt->cached_sw_if_index != sw_if_index0)) - { - first_int_addr = - ip4_interface_first_address (sm->ip4_main, sw_if_index0, - 0 /* just want the address */); - rt->cached_sw_if_index = sw_if_index0; - rt->cached_ip4_address = first_int_addr->as_u32; - } - - /* Don't NAT packet aimed at the intfc address */ - if (PREDICT_FALSE(ip0->dst_address.as_u32 == - rt->cached_ip4_address)) - goto trace0; - - next0 = slow_path (sm, b0, ip0, rx_fib_index0, &key0, - &s0, node, next0, cpu_index); - if (PREDICT_FALSE (next0 == SNAT_IN2OUT_NEXT_DROP)) - goto trace0; - } - else - { - next0 = SNAT_IN2OUT_NEXT_SLOW_PATH; - goto trace0; - } - } - else - s0 = pool_elt_at_index (sm->per_thread_data[cpu_index].sessions, - value0.value); - - old_addr0 = ip0->src_address.as_u32; - ip0->src_address = s0->out2in.addr; - new_addr0 = ip0->src_address.as_u32; - vnet_buffer(b0)->sw_if_index[VLIB_TX] = s0->out2in.fib_index; - - sum0 = ip0->checksum; - sum0 = ip_csum_update (sum0, old_addr0, new_addr0, - ip4_header_t, - src_address /* changed member */); - ip0->checksum = ip_csum_fold (sum0); - - if (PREDICT_TRUE(proto0 == SNAT_PROTOCOL_TCP)) - { - old_port0 = tcp0->ports.src; - tcp0->ports.src = s0->out2in.port; - new_port0 = tcp0->ports.src; - - sum0 = tcp0->checksum; - sum0 = ip_csum_update (sum0, old_addr0, new_addr0, - ip4_header_t, - dst_address /* changed member */); - sum0 = ip_csum_update (sum0, old_port0, new_port0, - ip4_header_t /* cheat */, - length /* changed member */); - tcp0->checksum = ip_csum_fold(sum0); - } - else - { - old_port0 = udp0->src_port; - udp0->src_port = s0->out2in.port; - udp0->checksum = 0; - } - - /* Hairpinning */ - snat_hairpinning (sm, b0, ip0, udp0, tcp0, proto0); - - /* Accounting */ - s0->last_heard = now; - s0->total_pkts++; - s0->total_bytes += vlib_buffer_length_in_chain (vm, b0); - /* Per-user LRU list maintenance for dynamic translation */ - if (!snat_is_session_static (s0)) - { - clib_dlist_remove (sm->per_thread_data[cpu_index].list_pool, - s0->per_user_index); - clib_dlist_addtail (sm->per_thread_data[cpu_index].list_pool, - s0->per_user_list_head_index, - s0->per_user_index); - } - - trace0: - if (PREDICT_FALSE((node->flags & VLIB_NODE_FLAG_TRACE) - && (b0->flags & VLIB_BUFFER_IS_TRACED))) - { - snat_in2out_trace_t *t = - vlib_add_trace (vm, node, b0, sizeof (*t)); - t->is_slow_path = is_slow_path; - t->sw_if_index = sw_if_index0; - t->next_index = next0; - t->session_index = ~0; - if (s0) - t->session_index = s0 - sm->per_thread_data[cpu_index].sessions; - } - - pkts_processed += next0 != SNAT_IN2OUT_NEXT_DROP; - - /* verify speculative enqueue, maybe switch current next frame */ - vlib_validate_buffer_enqueue_x1 (vm, node, next_index, - to_next, n_left_to_next, - bi0, next0); - } - - vlib_put_next_frame (vm, node, next_index, n_left_to_next); - } - - vlib_node_increment_counter (vm, stats_node_index, - SNAT_IN2OUT_ERROR_IN2OUT_PACKETS, - pkts_processed); - return frame->n_vectors; -} - -static uword -snat_in2out_fast_path_fn (vlib_main_t * vm, - vlib_node_runtime_t * node, - vlib_frame_t * frame) -{ - return snat_in2out_node_fn_inline (vm, node, frame, 0 /* is_slow_path */); -} - -VLIB_REGISTER_NODE (snat_in2out_node) = { - .function = snat_in2out_fast_path_fn, - .name = "snat-in2out", - .vector_size = sizeof (u32), - .format_trace = format_snat_in2out_trace, - .type = VLIB_NODE_TYPE_INTERNAL, - - .n_errors = ARRAY_LEN(snat_in2out_error_strings), - .error_strings = snat_in2out_error_strings, - - .runtime_data_bytes = sizeof (snat_runtime_t), - - .n_next_nodes = SNAT_IN2OUT_N_NEXT, - - /* edit / add dispositions here */ - .next_nodes = { - [SNAT_IN2OUT_NEXT_DROP] = "error-drop", - [SNAT_IN2OUT_NEXT_LOOKUP] = "ip4-lookup", - [SNAT_IN2OUT_NEXT_SLOW_PATH] = "snat-in2out-slowpath", - }, -}; - -VLIB_NODE_FUNCTION_MULTIARCH (snat_in2out_node, snat_in2out_fast_path_fn); - -static uword -snat_in2out_slow_path_fn (vlib_main_t * vm, - vlib_node_runtime_t * node, - vlib_frame_t * frame) -{ - return snat_in2out_node_fn_inline (vm, node, frame, 1 /* is_slow_path */); -} - -VLIB_REGISTER_NODE (snat_in2out_slowpath_node) = { - .function = snat_in2out_slow_path_fn, - .name = "snat-in2out-slowpath", - .vector_size = sizeof (u32), - .format_trace = format_snat_in2out_trace, - .type = VLIB_NODE_TYPE_INTERNAL, - - .n_errors = ARRAY_LEN(snat_in2out_error_strings), - .error_strings = snat_in2out_error_strings, - - .runtime_data_bytes = sizeof (snat_runtime_t), - - .n_next_nodes = SNAT_IN2OUT_N_NEXT, - - /* edit / add dispositions here */ - .next_nodes = { - [SNAT_IN2OUT_NEXT_DROP] = "error-drop", - [SNAT_IN2OUT_NEXT_LOOKUP] = "ip4-lookup", - [SNAT_IN2OUT_NEXT_SLOW_PATH] = "snat-in2out-slowpath", - }, -}; - -VLIB_NODE_FUNCTION_MULTIARCH (snat_in2out_slowpath_node, snat_in2out_slow_path_fn); - -static uword -snat_in2out_worker_handoff_fn (vlib_main_t * vm, - vlib_node_runtime_t * node, - vlib_frame_t * frame) -{ - snat_main_t *sm = &snat_main; - vlib_thread_main_t *tm = vlib_get_thread_main (); - u32 n_left_from, *from, *to_next = 0; - static __thread vlib_frame_queue_elt_t **handoff_queue_elt_by_worker_index; - static __thread vlib_frame_queue_t **congested_handoff_queue_by_worker_index - = 0; - vlib_frame_queue_elt_t *hf = 0; - vlib_frame_t *f = 0; - int i; - u32 n_left_to_next_worker = 0, *to_next_worker = 0; - u32 next_worker_index = 0; - u32 current_worker_index = ~0; - u32 cpu_index = os_get_cpu_number (); - - ASSERT (vec_len (sm->workers)); - - if (PREDICT_FALSE (handoff_queue_elt_by_worker_index == 0)) - { - vec_validate (handoff_queue_elt_by_worker_index, tm->n_vlib_mains - 1); - - vec_validate_init_empty (congested_handoff_queue_by_worker_index, - sm->first_worker_index + sm->num_workers - 1, - (vlib_frame_queue_t *) (~0)); - } - - from = vlib_frame_vector_args (frame); - n_left_from = frame->n_vectors; - - while (n_left_from > 0) - { - u32 bi0; - vlib_buffer_t *b0; - u32 sw_if_index0; - u32 rx_fib_index0; - ip4_header_t * ip0; - snat_user_key_t key0; - clib_bihash_kv_8_8_t kv0, value0; - u8 do_handoff; - - bi0 = from[0]; - from += 1; - n_left_from -= 1; - - b0 = vlib_get_buffer (vm, bi0); - - sw_if_index0 = vnet_buffer (b0)->sw_if_index[VLIB_RX]; - rx_fib_index0 = ip4_fib_table_get_index_for_sw_if_index(sw_if_index0); - - ip0 = vlib_buffer_get_current (b0); - - key0.addr = ip0->src_address; - key0.fib_index = rx_fib_index0; - - kv0.key = key0.as_u64; - - /* Ever heard of of the "user" before? */ - if (clib_bihash_search_8_8 (&sm->worker_by_in, &kv0, &value0)) - { - /* No, assign next available worker (RR) */ - next_worker_index = sm->first_worker_index + - sm->workers[sm->next_worker++ % vec_len (sm->workers)]; - - /* add non-traslated packets worker lookup */ - kv0.value = next_worker_index; - clib_bihash_add_del_8_8 (&sm->worker_by_in, &kv0, 1); - } - else - next_worker_index = value0.value; - - if (PREDICT_FALSE (next_worker_index != cpu_index)) - { - do_handoff = 1; - - if (next_worker_index != current_worker_index) - { - if (hf) - hf->n_vectors = VLIB_FRAME_SIZE - n_left_to_next_worker; - - hf = vlib_get_worker_handoff_queue_elt (sm->fq_in2out_index, - next_worker_index, - handoff_queue_elt_by_worker_index); - - n_left_to_next_worker = VLIB_FRAME_SIZE - hf->n_vectors; - to_next_worker = &hf->buffer_index[hf->n_vectors]; - current_worker_index = next_worker_index; - } - - /* enqueue to correct worker thread */ - to_next_worker[0] = bi0; - to_next_worker++; - n_left_to_next_worker--; - - if (n_left_to_next_worker == 0) - { - hf->n_vectors = VLIB_FRAME_SIZE; - vlib_put_frame_queue_elt (hf); - current_worker_index = ~0; - handoff_queue_elt_by_worker_index[next_worker_index] = 0; - hf = 0; - } - } - else - { - do_handoff = 0; - /* if this is 1st frame */ - if (!f) - { - f = vlib_get_frame_to_node (vm, snat_in2out_node.index); - to_next = vlib_frame_vector_args (f); - } - - to_next[0] = bi0; - to_next += 1; - f->n_vectors++; - } - - if (PREDICT_FALSE ((node->flags & VLIB_NODE_FLAG_TRACE) - && (b0->flags & VLIB_BUFFER_IS_TRACED))) - { - snat_in2out_worker_handoff_trace_t *t = - vlib_add_trace (vm, node, b0, sizeof (*t)); - t->next_worker_index = next_worker_index; - t->do_handoff = do_handoff; - } - } - - if (f) - vlib_put_frame_to_node (vm, snat_in2out_node.index, f); - - if (hf) - hf->n_vectors = VLIB_FRAME_SIZE - n_left_to_next_worker; - - /* Ship frames to the worker nodes */ - for (i = 0; i < vec_len (handoff_queue_elt_by_worker_index); i++) - { - if (handoff_queue_elt_by_worker_index[i]) - { - hf = handoff_queue_elt_by_worker_index[i]; - /* - * It works better to let the handoff node - * rate-adapt, always ship the handoff queue element. - */ - if (1 || hf->n_vectors == hf->last_n_vectors) - { - vlib_put_frame_queue_elt (hf); - handoff_queue_elt_by_worker_index[i] = 0; - } - else - hf->last_n_vectors = hf->n_vectors; - } - congested_handoff_queue_by_worker_index[i] = - (vlib_frame_queue_t *) (~0); - } - hf = 0; - current_worker_index = ~0; - return frame->n_vectors; -} - -VLIB_REGISTER_NODE (snat_in2out_worker_handoff_node) = { - .function = snat_in2out_worker_handoff_fn, - .name = "snat-in2out-worker-handoff", - .vector_size = sizeof (u32), - .format_trace = format_snat_in2out_worker_handoff_trace, - .type = VLIB_NODE_TYPE_INTERNAL, - - .n_next_nodes = 1, - - .next_nodes = { - [0] = "error-drop", - }, -}; - -VLIB_NODE_FUNCTION_MULTIARCH (snat_in2out_worker_handoff_node, snat_in2out_worker_handoff_fn); - -static inline u32 icmp_in2out_static_map (snat_main_t *sm, - vlib_buffer_t * b0, - ip4_header_t * ip0, - icmp46_header_t * icmp0, - u32 sw_if_index0, - vlib_node_runtime_t * node, - u32 next0, - u32 rx_fib_index0) -{ - snat_session_key_t key0, sm0; - icmp_echo_header_t *echo0; - u32 new_addr0, old_addr0; - u16 old_id0, new_id0; - ip_csum_t sum0; - snat_runtime_t * rt = (snat_runtime_t *)node->runtime_data; - - echo0 = (icmp_echo_header_t *)(icmp0+1); - - key0.addr = ip0->src_address; - key0.port = echo0->identifier; - key0.fib_index = rx_fib_index0; - - if (snat_static_mapping_match(sm, key0, &sm0, 0)) - { - ip4_address_t * first_int_addr; - - if (PREDICT_FALSE(rt->cached_sw_if_index != sw_if_index0)) - { - first_int_addr = - ip4_interface_first_address (sm->ip4_main, sw_if_index0, - 0 /* just want the address */); - rt->cached_sw_if_index = sw_if_index0; - rt->cached_ip4_address = first_int_addr->as_u32; - } - - /* Don't NAT packet aimed at the intfc address */ - if (PREDICT_FALSE(ip0->dst_address.as_u32 == - rt->cached_ip4_address)) - return next0; - - b0->error = node->errors[SNAT_IN2OUT_ERROR_NO_TRANSLATION]; - return SNAT_IN2OUT_NEXT_DROP; - } - - new_addr0 = sm0.addr.as_u32; - new_id0 = sm0.port; - vnet_buffer(b0)->sw_if_index[VLIB_TX] = sm0.fib_index; - old_addr0 = ip0->src_address.as_u32; - ip0->src_address.as_u32 = new_addr0; - - sum0 = ip0->checksum; - sum0 = ip_csum_update (sum0, old_addr0, new_addr0, - ip4_header_t, - src_address /* changed member */); - ip0->checksum = ip_csum_fold (sum0); - - if (PREDICT_FALSE(new_id0 != echo0->identifier)) - { - old_id0 = echo0->identifier; - echo0->identifier = new_id0; - - sum0 = icmp0->checksum; - sum0 = ip_csum_update (sum0, old_id0, new_id0, icmp_echo_header_t, - identifier); - icmp0->checksum = ip_csum_fold (sum0); - } - - return next0; -} - -static uword -snat_in2out_fast_static_map_fn (vlib_main_t * vm, - vlib_node_runtime_t * node, - vlib_frame_t * frame) -{ - u32 n_left_from, * from, * to_next; - snat_in2out_next_t next_index; - u32 pkts_processed = 0; - snat_main_t * sm = &snat_main; - snat_runtime_t * rt = (snat_runtime_t *)node->runtime_data; - u32 stats_node_index; - - stats_node_index = snat_in2out_fast_node.index; - - from = vlib_frame_vector_args (frame); - n_left_from = frame->n_vectors; - next_index = node->cached_next_index; - - while (n_left_from > 0) - { - u32 n_left_to_next; - - vlib_get_next_frame (vm, node, next_index, - to_next, n_left_to_next); - - while (n_left_from > 0 && n_left_to_next > 0) - { - u32 bi0; - vlib_buffer_t * b0; - u32 next0; - u32 sw_if_index0; - ip4_header_t * ip0; - ip_csum_t sum0; - u32 new_addr0, old_addr0; - u16 old_port0, new_port0; - udp_header_t * udp0; - tcp_header_t * tcp0; - icmp46_header_t * icmp0; - snat_session_key_t key0, sm0; - u32 proto0; - u32 rx_fib_index0; - - /* speculatively enqueue b0 to the current next frame */ - bi0 = from[0]; - to_next[0] = bi0; - from += 1; - to_next += 1; - n_left_from -= 1; - n_left_to_next -= 1; - - b0 = vlib_get_buffer (vm, bi0); - next0 = SNAT_IN2OUT_NEXT_LOOKUP; - - ip0 = vlib_buffer_get_current (b0); - udp0 = ip4_next_header (ip0); - tcp0 = (tcp_header_t *) udp0; - icmp0 = (icmp46_header_t *) udp0; - - sw_if_index0 = vnet_buffer(b0)->sw_if_index[VLIB_RX]; - rx_fib_index0 = ip4_fib_table_get_index_for_sw_if_index(sw_if_index0); - - proto0 = ~0; - proto0 = (ip0->protocol == IP_PROTOCOL_UDP) - ? SNAT_PROTOCOL_UDP : proto0; - proto0 = (ip0->protocol == IP_PROTOCOL_TCP) - ? SNAT_PROTOCOL_TCP : proto0; - proto0 = (ip0->protocol == IP_PROTOCOL_ICMP) - ? SNAT_PROTOCOL_ICMP : proto0; - - if (PREDICT_FALSE (proto0 == ~0)) - goto trace0; - - if (PREDICT_FALSE (proto0 == SNAT_PROTOCOL_ICMP)) - { - ip4_address_t * first_int_addr; - - if (PREDICT_FALSE(rt->cached_sw_if_index != sw_if_index0)) - { - first_int_addr = - ip4_interface_first_address (sm->ip4_main, sw_if_index0, - 0 /* just want the address */); - rt->cached_sw_if_index = sw_if_index0; - rt->cached_ip4_address = first_int_addr->as_u32; - } - - /* Don't NAT packet aimed at the intfc address */ - if (PREDICT_FALSE(ip0->dst_address.as_u32 == - rt->cached_ip4_address)) - goto trace0; - - next0 = icmp_in2out_static_map - (sm, b0, ip0, icmp0, sw_if_index0, node, next0, rx_fib_index0); - goto trace0; - } - - key0.addr = ip0->src_address; - key0.port = udp0->src_port; - key0.fib_index = rx_fib_index0; - - if (snat_static_mapping_match(sm, key0, &sm0, 0)) - { - b0->error = node->errors[SNAT_IN2OUT_ERROR_NO_TRANSLATION]; - next0= SNAT_IN2OUT_NEXT_DROP; - goto trace0; - } - - new_addr0 = sm0.addr.as_u32; - new_port0 = sm0.port; - vnet_buffer(b0)->sw_if_index[VLIB_TX] = sm0.fib_index; - old_addr0 = ip0->src_address.as_u32; - ip0->src_address.as_u32 = new_addr0; - - sum0 = ip0->checksum; - sum0 = ip_csum_update (sum0, old_addr0, new_addr0, - ip4_header_t, - src_address /* changed member */); - ip0->checksum = ip_csum_fold (sum0); - - if (PREDICT_FALSE(new_port0 != udp0->dst_port)) - { - if (PREDICT_TRUE(proto0 == SNAT_PROTOCOL_TCP)) - { - old_port0 = tcp0->ports.src; - tcp0->ports.src = new_port0; - - sum0 = tcp0->checksum; - sum0 = ip_csum_update (sum0, old_addr0, new_addr0, - ip4_header_t, - dst_address /* changed member */); - sum0 = ip_csum_update (sum0, old_port0, new_port0, - ip4_header_t /* cheat */, - length /* changed member */); - tcp0->checksum = ip_csum_fold(sum0); - } - else - { - old_port0 = udp0->src_port; - udp0->src_port = new_port0; - udp0->checksum = 0; - } - } - else - { - if (PREDICT_TRUE(proto0 == SNAT_PROTOCOL_TCP)) - { - sum0 = tcp0->checksum; - sum0 = ip_csum_update (sum0, old_addr0, new_addr0, - ip4_header_t, - dst_address /* changed member */); - tcp0->checksum = ip_csum_fold(sum0); - } - } - - /* Hairpinning */ - snat_hairpinning (sm, b0, ip0, udp0, tcp0, proto0); - - trace0: - if (PREDICT_FALSE((node->flags & VLIB_NODE_FLAG_TRACE) - && (b0->flags & VLIB_BUFFER_IS_TRACED))) - { - snat_in2out_trace_t *t = - vlib_add_trace (vm, node, b0, sizeof (*t)); - t->sw_if_index = sw_if_index0; - t->next_index = next0; - } - - pkts_processed += next0 != SNAT_IN2OUT_NEXT_DROP; - - /* verify speculative enqueue, maybe switch current next frame */ - vlib_validate_buffer_enqueue_x1 (vm, node, next_index, - to_next, n_left_to_next, - bi0, next0); - } - - vlib_put_next_frame (vm, node, next_index, n_left_to_next); - } - - vlib_node_increment_counter (vm, stats_node_index, - SNAT_IN2OUT_ERROR_IN2OUT_PACKETS, - pkts_processed); - return frame->n_vectors; -} - - -VLIB_REGISTER_NODE (snat_in2out_fast_node) = { - .function = snat_in2out_fast_static_map_fn, - .name = "snat-in2out-fast", - .vector_size = sizeof (u32), - .format_trace = format_snat_in2out_fast_trace, - .type = VLIB_NODE_TYPE_INTERNAL, - - .n_errors = ARRAY_LEN(snat_in2out_error_strings), - .error_strings = snat_in2out_error_strings, - - .runtime_data_bytes = sizeof (snat_runtime_t), - - .n_next_nodes = SNAT_IN2OUT_N_NEXT, - - /* edit / add dispositions here */ - .next_nodes = { - [SNAT_IN2OUT_NEXT_DROP] = "error-drop", - [SNAT_IN2OUT_NEXT_LOOKUP] = "ip4-lookup", - [SNAT_IN2OUT_NEXT_SLOW_PATH] = "snat-in2out-slowpath", - }, -}; - -VLIB_NODE_FUNCTION_MULTIARCH (snat_in2out_fast_node, snat_in2out_fast_static_map_fn); diff --git a/plugins/snat-plugin/snat/jvpp/io/fd/vpp/jvpp/snat/test/CallbackApiTest.java b/plugins/snat-plugin/snat/jvpp/io/fd/vpp/jvpp/snat/test/CallbackApiTest.java deleted file mode 100644 index 32165d96..00000000 --- a/plugins/snat-plugin/snat/jvpp/io/fd/vpp/jvpp/snat/test/CallbackApiTest.java +++ /dev/null @@ -1,68 +0,0 @@ -/* - * Copyright (c) 2016 Cisco and/or its affiliates. - * - * Licensed under the Apache License, Version 2.0 (the "License"); - * you may not use this file except in compliance with the License. - * You may obtain a copy of the License at: - * - * http://www.apache.org/licenses/LICENSE-2.0 - * - * Unless required by applicable law or agreed to in writing, software - * distributed under the License is distributed on an "AS IS" BASIS, - * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. - * See the License for the specific language governing permissions and - * limitations under the License. - */ - -package io.fd.vpp.jvpp.snat.test; - -import io.fd.vpp.jvpp.JVpp; -import io.fd.vpp.jvpp.JVppRegistry; -import io.fd.vpp.jvpp.JVppRegistryImpl; -import io.fd.vpp.jvpp.VppCallbackException; -import io.fd.vpp.jvpp.snat.JVppSnatImpl; -import io.fd.vpp.jvpp.snat.callback.SnatInterfaceAddDelFeatureCallback; -import io.fd.vpp.jvpp.snat.dto.SnatInterfaceAddDelFeature; -import io.fd.vpp.jvpp.snat.dto.SnatInterfaceAddDelFeatureReply; - -public class CallbackApiTest { - - static class TestCallback implements SnatInterfaceAddDelFeatureCallback { - - @Override - public void onSnatInterfaceAddDelFeatureReply(final SnatInterfaceAddDelFeatureReply msg) { - System.out.printf("Received SnatInterfaceAddDelFeatureReply: context=%d%n", - msg.context); - } - - @Override - public void onError(VppCallbackException ex) { - System.out.printf("Received onError exception: call=%s, context=%d, retval=%d%n", ex.getMethodName(), - ex.getCtxId(), ex.getErrorCode()); - } - } - - public static void main(String[] args) throws Exception { - testCallbackApi(); - } - - private static void testCallbackApi() throws Exception { - System.out.println("Testing Java callback API for snat plugin"); - try (final JVppRegistry registry = new JVppRegistryImpl("SnatCallbackApiTest"); - final JVpp jvpp = new JVppSnatImpl()) { - registry.register(jvpp, new TestCallback()); - - System.out.println("Sending SnatInterfaceAddDelFeature request..."); - SnatInterfaceAddDelFeature request = new SnatInterfaceAddDelFeature(); - request.isAdd = 1; - request.isInside = 1; - request.swIfIndex = 1; - final int result = jvpp.send(request); - System.out.printf("SnatInterfaceAddDelFeature send result = %d%n", result); - - Thread.sleep(1000); - - System.out.println("Disconnecting..."); - } - } -} diff --git a/plugins/snat-plugin/snat/jvpp/io/fd/vpp/jvpp/snat/test/Readme.txt b/plugins/snat-plugin/snat/jvpp/io/fd/vpp/jvpp/snat/test/Readme.txt deleted file mode 100644 index a2b0c41f..00000000 --- a/plugins/snat-plugin/snat/jvpp/io/fd/vpp/jvpp/snat/test/Readme.txt +++ /dev/null @@ -1 +0,0 @@ -sudo java -cp build-vpp-native/vpp-api/java/jvpp-registry-17.01.jar:build-vpp-native/plugins/snat-plugin/jvpp-snat-1.0.jar io.fd.vpp.jvpp.snat.test.CallbackApiTest diff --git a/plugins/snat-plugin/snat/jvpp_snat.c b/plugins/snat-plugin/snat/jvpp_snat.c deleted file mode 100644 index fd72ddb1..00000000 --- a/plugins/snat-plugin/snat/jvpp_snat.c +++ /dev/null @@ -1,124 +0,0 @@ -/* - * Copyright (c) 2016 Cisco and/or its affiliates. - * Licensed under the Apache License, Version 2.0 (the "License"); - * you may not use this file except in compliance with the License. - * You may obtain a copy of the License at: - * - * http://www.apache.org/licenses/LICENSE-2.0 - * - * Unless required by applicable law or agreed to in writing, software - * distributed under the License is distributed on an "AS IS" BASIS, - * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. - * See the License for the specific language governing permissions and - * limitations under the License. - */ - -#include - -#include -#define vl_typedefs /* define message structures */ -#include -#undef vl_typedefs - -#define vl_endianfun -#include -#undef vl_endianfun - -#define vl_print(handle, ...) -#define vl_printfun -#include -#undef vl_printfun - -/* Get the API version number */ -#define vl_api_version(n,v) static u32 api_version=(v); -#include -#undef vl_api_version - -#include -#include -#include - -#if VPPJNI_DEBUG == 1 - #define DEBUG_LOG(...) clib_warning(__VA_ARGS__) -#else - #define DEBUG_LOG(...) -#endif - -#include - -#include "snat/jvpp/io_fd_vpp_jvpp_snat_JVppSnatImpl.h" -#include "jvpp_snat.h" -#include "snat/jvpp/jvpp_snat_gen.h" - -/* - * Class: io_fd_vpp_jvpp_snat_JVppsnatImpl - * Method: init0 - * Signature: (JI)V - */ -JNIEXPORT void JNICALL Java_io_fd_vpp_jvpp_snat_JVppSnatImpl_init0 - (JNIEnv *env, jclass clazz, jobject callback, jlong queue_address, jint my_client_index) { - snat_main_t * plugin_main = &snat_main; - u8 * name; - clib_warning ("Java_io_fd_vpp_jvpp_snat_JVppSnatImpl_init0"); - - plugin_main->my_client_index = my_client_index; - plugin_main->vl_input_queue = (unix_shared_memory_queue_t *)queue_address; - - name = format (0, "snat_%08x%c", api_version, 0); - plugin_main->msg_id_base = vl_client_get_first_plugin_msg_id ((char *) name); - - if (plugin_main->msg_id_base == (u16) ~0) { - jclass exClass = (*env)->FindClass(env, "java/lang/IllegalStateException"); - (*env)->ThrowNew(env, exClass, "snat plugin is not loaded in VPP"); - } else { - plugin_main->callbackObject = (*env)->NewGlobalRef(env, callback); - plugin_main->callbackClass = (jclass)(*env)->NewGlobalRef(env, (*env)->GetObjectClass(env, callback)); - - #define _(N,n) \ - vl_msg_api_set_handlers(VL_API_##N + plugin_main->msg_id_base, #n, \ - vl_api_##n##_t_handler, \ - vl_noop_handler, \ - vl_api_##n##_t_endian, \ - vl_api_##n##_t_print, \ - sizeof(vl_api_##n##_t), 1); - foreach_api_reply_handler; - #undef _ - } -} - -JNIEXPORT void JNICALL Java_io_fd_vpp_jvpp_snat_JVppSnatImpl_close0 -(JNIEnv *env, jclass clazz) { - snat_main_t * plugin_main = &snat_main; - - // cleanup: - (*env)->DeleteGlobalRef(env, plugin_main->callbackClass); - (*env)->DeleteGlobalRef(env, plugin_main->callbackObject); - - plugin_main->callbackClass = NULL; - plugin_main->callbackObject = NULL; -} - -/* Attach thread to JVM and cache class references when initiating JVPP SNAT */ -jint JNI_OnLoad(JavaVM *vm, void *reserved) { - JNIEnv* env; - - if ((*vm)->GetEnv(vm, (void**) &env, JNI_VERSION_1_8) != JNI_OK) { - return JNI_EVERSION; - } - - if (cache_class_references(env) != 0) { - clib_warning ("Failed to cache class references\n"); - return JNI_ERR; - } - - return JNI_VERSION_1_8; -} - -/* Clean up cached references when disposing JVPP SNAT */ -void JNI_OnUnload(JavaVM *vm, void *reserved) { - JNIEnv* env; - if ((*vm)->GetEnv(vm, (void**) &env, JNI_VERSION_1_8) != JNI_OK) { - return; - } - delete_class_references(env); -} diff --git a/plugins/snat-plugin/snat/jvpp_snat.h b/plugins/snat-plugin/snat/jvpp_snat.h deleted file mode 100644 index 6426bda8..00000000 --- a/plugins/snat-plugin/snat/jvpp_snat.h +++ /dev/null @@ -1,45 +0,0 @@ -/* - * Copyright (c) 2016 Cisco and/or its affiliates. - * Licensed under the Apache License, Version 2.0 (the "License"); - * you may not use this file except in compliance with the License. - * You may obtain a copy of the License at: - * - * http://www.apache.org/licenses/LICENSE-2.0 - * - * Unless required by applicable law or agreed to in writing, software - * distributed under the License is distributed on an "AS IS" BASIS, - * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. - * See the License for the specific language governing permissions and - * limitations under the License. - */ -#ifndef __included_jvpp_snat_h__ -#define __included_jvpp_snat_h__ - -#include -#include -#include -#include -#include -#include - -/* Global state for JVPP-SNAT */ -typedef struct { - /* Base message index for the nsh plugin */ - u16 msg_id_base; - - /* Pointer to shared memory queue */ - unix_shared_memory_queue_t * vl_input_queue; - - /* VPP api client index */ - u32 my_client_index; - - /* Callback object and class references enabling asynchronous Java calls */ - jobject callbackObject; - jclass callbackClass; - -} snat_main_t; - -snat_main_t snat_main __attribute__((aligned (64))); - - -#endif /* __included_jvpp_snat_h__ */ diff --git a/plugins/snat-plugin/snat/out2in.c b/plugins/snat-plugin/snat/out2in.c deleted file mode 100644 index f1f4159c..00000000 --- a/plugins/snat-plugin/snat/out2in.c +++ /dev/null @@ -1,1261 +0,0 @@ -/* - * Copyright (c) 2016 Cisco and/or its affiliates. - * Licensed under the Apache License, Version 2.0 (the "License"); - * you may not use this file except in compliance with the License. - * You may obtain a copy of the License at: - * - * http://www.apache.org/licenses/LICENSE-2.0 - * - * Unless required by applicable law or agreed to in writing, software - * distributed under the License is distributed on an "AS IS" BASIS, - * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. - * See the License for the specific language governing permissions and - * limitations under the License. - */ - -#include -#include -#include -#include - -#include -#include -#include -#include - -#include -#include -#include - -typedef struct { - u32 sw_if_index; - u32 next_index; - u32 session_index; -} snat_out2in_trace_t; - -typedef struct { - u32 next_worker_index; - u8 do_handoff; -} snat_out2in_worker_handoff_trace_t; - -/* packet trace format function */ -static u8 * format_snat_out2in_trace (u8 * s, va_list * args) -{ - CLIB_UNUSED (vlib_main_t * vm) = va_arg (*args, vlib_main_t *); - CLIB_UNUSED (vlib_node_t * node) = va_arg (*args, vlib_node_t *); - snat_out2in_trace_t * t = va_arg (*args, snat_out2in_trace_t *); - - s = format (s, "SNAT_OUT2IN: sw_if_index %d, next index %d, session index %d", - t->sw_if_index, t->next_index, t->session_index); - return s; -} - -static u8 * format_snat_out2in_fast_trace (u8 * s, va_list * args) -{ - CLIB_UNUSED (vlib_main_t * vm) = va_arg (*args, vlib_main_t *); - CLIB_UNUSED (vlib_node_t * node) = va_arg (*args, vlib_node_t *); - snat_out2in_trace_t * t = va_arg (*args, snat_out2in_trace_t *); - - s = format (s, "SNAT_OUT2IN_FAST: sw_if_index %d, next index %d", - t->sw_if_index, t->next_index); - return s; -} - -static u8 * format_snat_out2in_worker_handoff_trace (u8 * s, va_list * args) -{ - CLIB_UNUSED (vlib_main_t * vm) = va_arg (*args, vlib_main_t *); - CLIB_UNUSED (vlib_node_t * node) = va_arg (*args, vlib_node_t *); - snat_out2in_worker_handoff_trace_t * t = - va_arg (*args, snat_out2in_worker_handoff_trace_t *); - char * m; - - m = t->do_handoff ? "next worker" : "same worker"; - s = format (s, "SNAT_OUT2IN_WORKER_HANDOFF: %s %d", m, t->next_worker_index); - - return s; -} - -vlib_node_registration_t snat_out2in_node; -vlib_node_registration_t snat_out2in_fast_node; -vlib_node_registration_t snat_out2in_worker_handoff_node; - -#define foreach_snat_out2in_error \ -_(UNSUPPORTED_PROTOCOL, "Unsupported protocol") \ -_(OUT2IN_PACKETS, "Good out2in packets processed") \ -_(BAD_ICMP_TYPE, "icmp type not echo-reply") \ -_(NO_TRANSLATION, "No translation") - -typedef enum { -#define _(sym,str) SNAT_OUT2IN_ERROR_##sym, - foreach_snat_out2in_error -#undef _ - SNAT_OUT2IN_N_ERROR, -} snat_out2in_error_t; - -static char * snat_out2in_error_strings[] = { -#define _(sym,string) string, - foreach_snat_out2in_error -#undef _ -}; - -typedef enum { - SNAT_OUT2IN_NEXT_DROP, - SNAT_OUT2IN_NEXT_LOOKUP, - SNAT_OUT2IN_N_NEXT, -} snat_out2in_next_t; - -/** - * @brief Create session for static mapping. - * - * Create NAT session initiated by host from external network with static - * mapping. - * - * @param sm SNAT main. - * @param b0 Vlib buffer. - * @param in2out In2out SNAT session key. - * @param out2in Out2in SNAT session key. - * @param node Vlib node. - * - * @returns SNAT session if successfully created otherwise 0. - */ -static inline snat_session_t * -create_session_for_static_mapping (snat_main_t *sm, - vlib_buffer_t *b0, - snat_session_key_t in2out, - snat_session_key_t out2in, - vlib_node_runtime_t * node, - u32 cpu_index) -{ - snat_user_t *u; - snat_user_key_t user_key; - snat_session_t *s; - clib_bihash_kv_8_8_t kv0, value0; - dlist_elt_t * per_user_translation_list_elt; - dlist_elt_t * per_user_list_head_elt; - - user_key.addr = in2out.addr; - user_key.fib_index = in2out.fib_index; - kv0.key = user_key.as_u64; - - /* Ever heard of the "user" = inside ip4 address before? */ - if (clib_bihash_search_8_8 (&sm->user_hash, &kv0, &value0)) - { - /* no, make a new one */ - pool_get (sm->per_thread_data[cpu_index].users, u); - memset (u, 0, sizeof (*u)); - u->addr = in2out.addr; - - pool_get (sm->per_thread_data[cpu_index].list_pool, - per_user_list_head_elt); - - u->sessions_per_user_list_head_index = per_user_list_head_elt - - sm->per_thread_data[cpu_index].list_pool; - - clib_dlist_init (sm->per_thread_data[cpu_index].list_pool, - u->sessions_per_user_list_head_index); - - kv0.value = u - sm->per_thread_data[cpu_index].users; - - /* add user */ - clib_bihash_add_del_8_8 (&sm->user_hash, &kv0, 1 /* is_add */); - - /* add non-traslated packets worker lookup */ - kv0.value = cpu_index; - clib_bihash_add_del_8_8 (&sm->worker_by_in, &kv0, 1); - } - else - { - u = pool_elt_at_index (sm->per_thread_data[cpu_index].users, - value0.value); - } - - pool_get (sm->per_thread_data[cpu_index].sessions, s); - memset (s, 0, sizeof (*s)); - - s->outside_address_index = ~0; - s->flags |= SNAT_SESSION_FLAG_STATIC_MAPPING; - u->nstaticsessions++; - - /* Create list elts */ - pool_get (sm->per_thread_data[cpu_index].list_pool, - per_user_translation_list_elt); - clib_dlist_init (sm->per_thread_data[cpu_index].list_pool, - per_user_translation_list_elt - - sm->per_thread_data[cpu_index].list_pool); - - per_user_translation_list_elt->value = - s - sm->per_thread_data[cpu_index].sessions; - s->per_user_index = - per_user_translation_list_elt - sm->per_thread_data[cpu_index].list_pool; - s->per_user_list_head_index = u->sessions_per_user_list_head_index; - - clib_dlist_addtail (sm->per_thread_data[cpu_index].list_pool, - s->per_user_list_head_index, - per_user_translation_list_elt - - sm->per_thread_data[cpu_index].list_pool); - - s->in2out = in2out; - s->out2in = out2in; - s->in2out.protocol = out2in.protocol; - - /* Add to translation hashes */ - kv0.key = s->in2out.as_u64; - kv0.value = s - sm->per_thread_data[cpu_index].sessions; - if (clib_bihash_add_del_8_8 (&sm->in2out, &kv0, 1 /* is_add */)) - clib_warning ("in2out key add failed"); - - kv0.key = s->out2in.as_u64; - kv0.value = s - sm->per_thread_data[cpu_index].sessions; - - if (clib_bihash_add_del_8_8 (&sm->out2in, &kv0, 1 /* is_add */)) - clib_warning ("out2in key add failed"); - - return s; -} - -static inline u32 icmp_out2in_slow_path (snat_main_t *sm, - vlib_buffer_t * b0, - ip4_header_t * ip0, - icmp46_header_t * icmp0, - u32 sw_if_index0, - u32 rx_fib_index0, - vlib_node_runtime_t * node, - u32 next0, f64 now, - u32 cpu_index) -{ - snat_session_key_t key0, sm0; - icmp_echo_header_t *echo0; - clib_bihash_kv_8_8_t kv0, value0; - snat_session_t * s0; - u32 new_addr0, old_addr0; - u16 old_id0, new_id0; - ip_csum_t sum0; - snat_runtime_t * rt = (snat_runtime_t *)node->runtime_data; - - echo0 = (icmp_echo_header_t *)(icmp0+1); - - key0.addr = ip0->dst_address; - key0.port = echo0->identifier; - key0.protocol = SNAT_PROTOCOL_ICMP; - key0.fib_index = rx_fib_index0; - - kv0.key = key0.as_u64; - - if (clib_bihash_search_8_8 (&sm->out2in, &kv0, &value0)) - { - /* Try to match static mapping by external address and port, - destination address and port in packet */ - if (snat_static_mapping_match(sm, key0, &sm0, 1)) - { - ip4_address_t * first_int_addr; - - if (PREDICT_FALSE(rt->cached_sw_if_index != sw_if_index0)) - { - first_int_addr = - ip4_interface_first_address (sm->ip4_main, sw_if_index0, - 0 /* just want the address */); - rt->cached_sw_if_index = sw_if_index0; - rt->cached_ip4_address = first_int_addr->as_u32; - } - - /* Don't NAT packet aimed at the intfc address */ - if (PREDICT_FALSE(ip0->dst_address.as_u32 == - rt->cached_ip4_address)) - return next0; - - b0->error = node->errors[SNAT_OUT2IN_ERROR_NO_TRANSLATION]; - return SNAT_OUT2IN_NEXT_DROP; - } - - /* Create session initiated by host from external network */ - s0 = create_session_for_static_mapping(sm, b0, sm0, key0, - node, cpu_index); - if (!s0) - return SNAT_OUT2IN_NEXT_DROP; - } - else - s0 = pool_elt_at_index (sm->per_thread_data[cpu_index].sessions, - value0.value); - - old_addr0 = ip0->dst_address.as_u32; - ip0->dst_address = s0->in2out.addr; - new_addr0 = ip0->dst_address.as_u32; - vnet_buffer(b0)->sw_if_index[VLIB_TX] = s0->in2out.fib_index; - - sum0 = ip0->checksum; - sum0 = ip_csum_update (sum0, old_addr0, new_addr0, - ip4_header_t, - dst_address /* changed member */); - ip0->checksum = ip_csum_fold (sum0); - - old_id0 = echo0->identifier; - new_id0 = s0->in2out.port; - echo0->identifier = new_id0; - - sum0 = icmp0->checksum; - sum0 = ip_csum_update (sum0, old_id0, new_id0, icmp_echo_header_t, - identifier); - icmp0->checksum = ip_csum_fold (sum0); - - /* Accounting */ - s0->last_heard = now; - s0->total_pkts++; - s0->total_bytes += vlib_buffer_length_in_chain (sm->vlib_main, b0); - /* Per-user LRU list maintenance for dynamic translation */ - if (!snat_is_session_static (s0)) - { - clib_dlist_remove (sm->per_thread_data[cpu_index].list_pool, - s0->per_user_index); - clib_dlist_addtail (sm->per_thread_data[cpu_index].list_pool, - s0->per_user_list_head_index, - s0->per_user_index); - } - - return next0; -} - -static uword -snat_out2in_node_fn (vlib_main_t * vm, - vlib_node_runtime_t * node, - vlib_frame_t * frame) -{ - u32 n_left_from, * from, * to_next; - snat_out2in_next_t next_index; - u32 pkts_processed = 0; - snat_main_t * sm = &snat_main; - f64 now = vlib_time_now (vm); - u32 cpu_index = os_get_cpu_number (); - - from = vlib_frame_vector_args (frame); - n_left_from = frame->n_vectors; - next_index = node->cached_next_index; - - while (n_left_from > 0) - { - u32 n_left_to_next; - - vlib_get_next_frame (vm, node, next_index, - to_next, n_left_to_next); - - while (n_left_from >= 4 && n_left_to_next >= 2) - { - u32 bi0, bi1; - vlib_buffer_t * b0, * b1; - u32 next0 = SNAT_OUT2IN_NEXT_LOOKUP; - u32 next1 = SNAT_OUT2IN_NEXT_LOOKUP; - u32 sw_if_index0, sw_if_index1; - ip4_header_t * ip0, *ip1; - ip_csum_t sum0, sum1; - u32 new_addr0, old_addr0; - u16 new_port0, old_port0; - u32 new_addr1, old_addr1; - u16 new_port1, old_port1; - udp_header_t * udp0, * udp1; - tcp_header_t * tcp0, * tcp1; - icmp46_header_t * icmp0, * icmp1; - snat_session_key_t key0, key1, sm0, sm1; - u32 rx_fib_index0, rx_fib_index1; - u32 proto0, proto1; - snat_session_t * s0 = 0, * s1 = 0; - clib_bihash_kv_8_8_t kv0, kv1, value0, value1; - - /* Prefetch next iteration. */ - { - vlib_buffer_t * p2, * p3; - - p2 = vlib_get_buffer (vm, from[2]); - p3 = vlib_get_buffer (vm, from[3]); - - vlib_prefetch_buffer_header (p2, LOAD); - vlib_prefetch_buffer_header (p3, LOAD); - - CLIB_PREFETCH (p2->data, CLIB_CACHE_LINE_BYTES, STORE); - CLIB_PREFETCH (p3->data, CLIB_CACHE_LINE_BYTES, STORE); - } - - /* speculatively enqueue b0 and b1 to the current next frame */ - to_next[0] = bi0 = from[0]; - to_next[1] = bi1 = from[1]; - from += 2; - to_next += 2; - n_left_from -= 2; - n_left_to_next -= 2; - - b0 = vlib_get_buffer (vm, bi0); - b1 = vlib_get_buffer (vm, bi1); - - ip0 = vlib_buffer_get_current (b0); - udp0 = ip4_next_header (ip0); - tcp0 = (tcp_header_t *) udp0; - icmp0 = (icmp46_header_t *) udp0; - - sw_if_index0 = vnet_buffer(b0)->sw_if_index[VLIB_RX]; - rx_fib_index0 = vec_elt (sm->ip4_main->fib_index_by_sw_if_index, - sw_if_index0); - - proto0 = ~0; - proto0 = (ip0->protocol == IP_PROTOCOL_UDP) - ? SNAT_PROTOCOL_UDP : proto0; - proto0 = (ip0->protocol == IP_PROTOCOL_TCP) - ? SNAT_PROTOCOL_TCP : proto0; - proto0 = (ip0->protocol == IP_PROTOCOL_ICMP) - ? SNAT_PROTOCOL_ICMP : proto0; - - if (PREDICT_FALSE (proto0 == ~0)) - goto trace0; - - if (PREDICT_FALSE (proto0 == SNAT_PROTOCOL_ICMP)) - { - next0 = icmp_out2in_slow_path - (sm, b0, ip0, icmp0, sw_if_index0, rx_fib_index0, node, - next0, now, cpu_index); - goto trace0; - } - - key0.addr = ip0->dst_address; - key0.port = udp0->dst_port; - key0.protocol = proto0; - key0.fib_index = rx_fib_index0; - - kv0.key = key0.as_u64; - - if (clib_bihash_search_8_8 (&sm->out2in, &kv0, &value0)) - { - /* Try to match static mapping by external address and port, - destination address and port in packet */ - if (snat_static_mapping_match(sm, key0, &sm0, 1)) - { - b0->error = node->errors[SNAT_OUT2IN_ERROR_NO_TRANSLATION]; - goto trace0; - } - - /* Create session initiated by host from external network */ - s0 = create_session_for_static_mapping(sm, b0, sm0, key0, node, - cpu_index); - if (!s0) - goto trace0; - } - else - s0 = pool_elt_at_index (sm->per_thread_data[cpu_index].sessions, - value0.value); - - old_addr0 = ip0->dst_address.as_u32; - ip0->dst_address = s0->in2out.addr; - new_addr0 = ip0->dst_address.as_u32; - vnet_buffer(b0)->sw_if_index[VLIB_TX] = s0->in2out.fib_index; - - sum0 = ip0->checksum; - sum0 = ip_csum_update (sum0, old_addr0, new_addr0, - ip4_header_t, - dst_address /* changed member */); - ip0->checksum = ip_csum_fold (sum0); - - if (PREDICT_TRUE(proto0 == SNAT_PROTOCOL_TCP)) - { - old_port0 = tcp0->ports.dst; - tcp0->ports.dst = s0->in2out.port; - new_port0 = tcp0->ports.dst; - - sum0 = tcp0->checksum; - sum0 = ip_csum_update (sum0, old_addr0, new_addr0, - ip4_header_t, - dst_address /* changed member */); - - sum0 = ip_csum_update (sum0, old_port0, new_port0, - ip4_header_t /* cheat */, - length /* changed member */); - tcp0->checksum = ip_csum_fold(sum0); - } - else - { - old_port0 = udp0->dst_port; - udp0->dst_port = s0->in2out.port; - udp0->checksum = 0; - } - - /* Accounting */ - s0->last_heard = now; - s0->total_pkts++; - s0->total_bytes += vlib_buffer_length_in_chain (vm, b0); - /* Per-user LRU list maintenance for dynamic translation */ - if (!snat_is_session_static (s0)) - { - clib_dlist_remove (sm->per_thread_data[cpu_index].list_pool, - s0->per_user_index); - clib_dlist_addtail (sm->per_thread_data[cpu_index].list_pool, - s0->per_user_list_head_index, - s0->per_user_index); - } - trace0: - - if (PREDICT_FALSE((node->flags & VLIB_NODE_FLAG_TRACE) - && (b0->flags & VLIB_BUFFER_IS_TRACED))) - { - snat_out2in_trace_t *t = - vlib_add_trace (vm, node, b0, sizeof (*t)); - t->sw_if_index = sw_if_index0; - t->next_index = next0; - t->session_index = ~0; - if (s0) - t->session_index = s0 - sm->per_thread_data[cpu_index].sessions; - } - - pkts_processed += next0 != SNAT_OUT2IN_NEXT_DROP; - - - ip1 = vlib_buffer_get_current (b1); - udp1 = ip4_next_header (ip1); - tcp1 = (tcp_header_t *) udp1; - icmp1 = (icmp46_header_t *) udp1; - - sw_if_index1 = vnet_buffer(b1)->sw_if_index[VLIB_RX]; - rx_fib_index1 = vec_elt (sm->ip4_main->fib_index_by_sw_if_index, - sw_if_index1); - - proto1 = ~0; - proto1 = (ip1->protocol == IP_PROTOCOL_UDP) - ? SNAT_PROTOCOL_UDP : proto1; - proto1 = (ip1->protocol == IP_PROTOCOL_TCP) - ? SNAT_PROTOCOL_TCP : proto1; - proto1 = (ip1->protocol == IP_PROTOCOL_ICMP) - ? SNAT_PROTOCOL_ICMP : proto1; - - if (PREDICT_FALSE (proto1 == ~0)) - goto trace1; - - if (PREDICT_FALSE (proto1 == SNAT_PROTOCOL_ICMP)) - { - next1 = icmp_out2in_slow_path - (sm, b1, ip1, icmp1, sw_if_index1, rx_fib_index1, node, - next1, now, cpu_index); - goto trace1; - } - - key1.addr = ip1->dst_address; - key1.port = udp1->dst_port; - key1.protocol = proto1; - key1.fib_index = rx_fib_index1; - - kv1.key = key1.as_u64; - - if (clib_bihash_search_8_8 (&sm->out2in, &kv1, &value1)) - { - /* Try to match static mapping by external address and port, - destination address and port in packet */ - if (snat_static_mapping_match(sm, key1, &sm1, 1)) - { - b1->error = node->errors[SNAT_OUT2IN_ERROR_NO_TRANSLATION]; - goto trace1; - } - - /* Create session initiated by host from external network */ - s1 = create_session_for_static_mapping(sm, b1, sm1, key1, node, - cpu_index); - if (!s1) - goto trace1; - } - else - s1 = pool_elt_at_index (sm->per_thread_data[cpu_index].sessions, - value1.value); - - old_addr1 = ip1->dst_address.as_u32; - ip1->dst_address = s1->in2out.addr; - new_addr1 = ip1->dst_address.as_u32; - vnet_buffer(b1)->sw_if_index[VLIB_TX] = s1->in2out.fib_index; - - sum1 = ip1->checksum; - sum1 = ip_csum_update (sum1, old_addr1, new_addr1, - ip4_header_t, - dst_address /* changed member */); - ip1->checksum = ip_csum_fold (sum1); - - if (PREDICT_TRUE(proto1 == SNAT_PROTOCOL_TCP)) - { - old_port1 = tcp1->ports.dst; - tcp1->ports.dst = s1->in2out.port; - new_port1 = tcp1->ports.dst; - - sum1 = tcp1->checksum; - sum1 = ip_csum_update (sum1, old_addr1, new_addr1, - ip4_header_t, - dst_address /* changed member */); - - sum1 = ip_csum_update (sum1, old_port1, new_port1, - ip4_header_t /* cheat */, - length /* changed member */); - tcp1->checksum = ip_csum_fold(sum1); - } - else - { - old_port1 = udp1->dst_port; - udp1->dst_port = s1->in2out.port; - udp1->checksum = 0; - } - - /* Accounting */ - s1->last_heard = now; - s1->total_pkts++; - s1->total_bytes += vlib_buffer_length_in_chain (vm, b1); - /* Per-user LRU list maintenance for dynamic translation */ - if (!snat_is_session_static (s1)) - { - clib_dlist_remove (sm->per_thread_data[cpu_index].list_pool, - s1->per_user_index); - clib_dlist_addtail (sm->per_thread_data[cpu_index].list_pool, - s1->per_user_list_head_index, - s1->per_user_index); - } - trace1: - - if (PREDICT_FALSE((node->flags & VLIB_NODE_FLAG_TRACE) - && (b1->flags & VLIB_BUFFER_IS_TRACED))) - { - snat_out2in_trace_t *t = - vlib_add_trace (vm, node, b1, sizeof (*t)); - t->sw_if_index = sw_if_index1; - t->next_index = next1; - t->session_index = ~0; - if (s1) - t->session_index = s1 - sm->per_thread_data[cpu_index].sessions; - } - - pkts_processed += next1 != SNAT_OUT2IN_NEXT_DROP; - - /* verify speculative enqueues, maybe switch current next frame */ - vlib_validate_buffer_enqueue_x2 (vm, node, next_index, - to_next, n_left_to_next, - bi0, bi1, next0, next1); - } - - while (n_left_from > 0 && n_left_to_next > 0) - { - u32 bi0; - vlib_buffer_t * b0; - u32 next0 = SNAT_OUT2IN_NEXT_LOOKUP; - u32 sw_if_index0; - ip4_header_t * ip0; - ip_csum_t sum0; - u32 new_addr0, old_addr0; - u16 new_port0, old_port0; - udp_header_t * udp0; - tcp_header_t * tcp0; - icmp46_header_t * icmp0; - snat_session_key_t key0, sm0; - u32 rx_fib_index0; - u32 proto0; - snat_session_t * s0 = 0; - clib_bihash_kv_8_8_t kv0, value0; - - /* speculatively enqueue b0 to the current next frame */ - bi0 = from[0]; - to_next[0] = bi0; - from += 1; - to_next += 1; - n_left_from -= 1; - n_left_to_next -= 1; - - b0 = vlib_get_buffer (vm, bi0); - - ip0 = vlib_buffer_get_current (b0); - udp0 = ip4_next_header (ip0); - tcp0 = (tcp_header_t *) udp0; - icmp0 = (icmp46_header_t *) udp0; - - sw_if_index0 = vnet_buffer(b0)->sw_if_index[VLIB_RX]; - rx_fib_index0 = vec_elt (sm->ip4_main->fib_index_by_sw_if_index, - sw_if_index0); - - proto0 = ~0; - proto0 = (ip0->protocol == IP_PROTOCOL_UDP) - ? SNAT_PROTOCOL_UDP : proto0; - proto0 = (ip0->protocol == IP_PROTOCOL_TCP) - ? SNAT_PROTOCOL_TCP : proto0; - proto0 = (ip0->protocol == IP_PROTOCOL_ICMP) - ? SNAT_PROTOCOL_ICMP : proto0; - - if (PREDICT_FALSE (proto0 == ~0)) - goto trace00; - - if (PREDICT_FALSE (proto0 == SNAT_PROTOCOL_ICMP)) - { - next0 = icmp_out2in_slow_path - (sm, b0, ip0, icmp0, sw_if_index0, rx_fib_index0, node, - next0, now, cpu_index); - goto trace00; - } - - key0.addr = ip0->dst_address; - key0.port = udp0->dst_port; - key0.protocol = proto0; - key0.fib_index = rx_fib_index0; - - kv0.key = key0.as_u64; - - if (clib_bihash_search_8_8 (&sm->out2in, &kv0, &value0)) - { - /* Try to match static mapping by external address and port, - destination address and port in packet */ - if (snat_static_mapping_match(sm, key0, &sm0, 1)) - { - b0->error = node->errors[SNAT_OUT2IN_ERROR_NO_TRANSLATION]; - goto trace00; - } - - /* Create session initiated by host from external network */ - s0 = create_session_for_static_mapping(sm, b0, sm0, key0, node, - cpu_index); - if (!s0) - goto trace00; - } - else - s0 = pool_elt_at_index (sm->per_thread_data[cpu_index].sessions, - value0.value); - - old_addr0 = ip0->dst_address.as_u32; - ip0->dst_address = s0->in2out.addr; - new_addr0 = ip0->dst_address.as_u32; - vnet_buffer(b0)->sw_if_index[VLIB_TX] = s0->in2out.fib_index; - - sum0 = ip0->checksum; - sum0 = ip_csum_update (sum0, old_addr0, new_addr0, - ip4_header_t, - dst_address /* changed member */); - ip0->checksum = ip_csum_fold (sum0); - - if (PREDICT_TRUE(proto0 == SNAT_PROTOCOL_TCP)) - { - old_port0 = tcp0->ports.dst; - tcp0->ports.dst = s0->in2out.port; - new_port0 = tcp0->ports.dst; - - sum0 = tcp0->checksum; - sum0 = ip_csum_update (sum0, old_addr0, new_addr0, - ip4_header_t, - dst_address /* changed member */); - - sum0 = ip_csum_update (sum0, old_port0, new_port0, - ip4_header_t /* cheat */, - length /* changed member */); - tcp0->checksum = ip_csum_fold(sum0); - } - else - { - old_port0 = udp0->dst_port; - udp0->dst_port = s0->in2out.port; - udp0->checksum = 0; - } - - /* Accounting */ - s0->last_heard = now; - s0->total_pkts++; - s0->total_bytes += vlib_buffer_length_in_chain (vm, b0); - /* Per-user LRU list maintenance for dynamic translation */ - if (!snat_is_session_static (s0)) - { - clib_dlist_remove (sm->per_thread_data[cpu_index].list_pool, - s0->per_user_index); - clib_dlist_addtail (sm->per_thread_data[cpu_index].list_pool, - s0->per_user_list_head_index, - s0->per_user_index); - } - trace00: - - if (PREDICT_FALSE((node->flags & VLIB_NODE_FLAG_TRACE) - && (b0->flags & VLIB_BUFFER_IS_TRACED))) - { - snat_out2in_trace_t *t = - vlib_add_trace (vm, node, b0, sizeof (*t)); - t->sw_if_index = sw_if_index0; - t->next_index = next0; - t->session_index = ~0; - if (s0) - t->session_index = s0 - sm->per_thread_data[cpu_index].sessions; - } - - pkts_processed += next0 != SNAT_OUT2IN_NEXT_DROP; - - /* verify speculative enqueue, maybe switch current next frame */ - vlib_validate_buffer_enqueue_x1 (vm, node, next_index, - to_next, n_left_to_next, - bi0, next0); - } - - vlib_put_next_frame (vm, node, next_index, n_left_to_next); - } - - vlib_node_increment_counter (vm, snat_out2in_node.index, - SNAT_OUT2IN_ERROR_OUT2IN_PACKETS, - pkts_processed); - return frame->n_vectors; -} - -VLIB_REGISTER_NODE (snat_out2in_node) = { - .function = snat_out2in_node_fn, - .name = "snat-out2in", - .vector_size = sizeof (u32), - .format_trace = format_snat_out2in_trace, - .type = VLIB_NODE_TYPE_INTERNAL, - - .n_errors = ARRAY_LEN(snat_out2in_error_strings), - .error_strings = snat_out2in_error_strings, - - .runtime_data_bytes = sizeof (snat_runtime_t), - - .n_next_nodes = SNAT_OUT2IN_N_NEXT, - - /* edit / add dispositions here */ - .next_nodes = { - [SNAT_OUT2IN_NEXT_DROP] = "error-drop", - [SNAT_OUT2IN_NEXT_LOOKUP] = "ip4-lookup", - }, -}; -VLIB_NODE_FUNCTION_MULTIARCH (snat_out2in_node, snat_out2in_node_fn); - -static uword -snat_out2in_worker_handoff_fn (vlib_main_t * vm, - vlib_node_runtime_t * node, - vlib_frame_t * frame) -{ - snat_main_t *sm = &snat_main; - vlib_thread_main_t *tm = vlib_get_thread_main (); - u32 n_left_from, *from, *to_next = 0; - static __thread vlib_frame_queue_elt_t **handoff_queue_elt_by_worker_index; - static __thread vlib_frame_queue_t **congested_handoff_queue_by_worker_index - = 0; - vlib_frame_queue_elt_t *hf = 0; - vlib_frame_t *f = 0; - int i; - u32 n_left_to_next_worker = 0, *to_next_worker = 0; - u32 next_worker_index = 0; - u32 current_worker_index = ~0; - u32 cpu_index = os_get_cpu_number (); - - ASSERT (vec_len (sm->workers)); - - if (PREDICT_FALSE (handoff_queue_elt_by_worker_index == 0)) - { - vec_validate (handoff_queue_elt_by_worker_index, tm->n_vlib_mains - 1); - - vec_validate_init_empty (congested_handoff_queue_by_worker_index, - sm->first_worker_index + sm->num_workers - 1, - (vlib_frame_queue_t *) (~0)); - } - - from = vlib_frame_vector_args (frame); - n_left_from = frame->n_vectors; - - while (n_left_from > 0) - { - u32 bi0; - vlib_buffer_t *b0; - u32 sw_if_index0; - u32 rx_fib_index0; - ip4_header_t * ip0; - udp_header_t * udp0; - snat_static_mapping_key_t key0; - clib_bihash_kv_8_8_t kv0, value0; - u8 do_handoff; - - bi0 = from[0]; - from += 1; - n_left_from -= 1; - - b0 = vlib_get_buffer (vm, bi0); - - sw_if_index0 = vnet_buffer (b0)->sw_if_index[VLIB_RX]; - rx_fib_index0 = ip4_fib_table_get_index_for_sw_if_index(sw_if_index0); - - ip0 = vlib_buffer_get_current (b0); - udp0 = ip4_next_header (ip0); - - key0.addr = ip0->dst_address; - key0.port = udp0->dst_port; - key0.fib_index = rx_fib_index0; - - kv0.key = key0.as_u64; - - /* Ever heard of of the "user" before? */ - if (clib_bihash_search_8_8 (&sm->worker_by_out, &kv0, &value0)) - { - key0.port = 0; - kv0.key = key0.as_u64; - - if (clib_bihash_search_8_8 (&sm->worker_by_out, &kv0, &value0)) - { - /* No, assign next available worker (RR) */ - next_worker_index = sm->first_worker_index + - sm->workers[sm->next_worker++ % vec_len (sm->workers)]; - } - else - { - /* Static mapping without port */ - next_worker_index = value0.value; - } - - /* Add to translated packets worker lookup */ - kv0.value = next_worker_index; - clib_bihash_add_del_8_8 (&sm->worker_by_out, &kv0, 1); - } - else - next_worker_index = value0.value; - - if (PREDICT_FALSE (next_worker_index != cpu_index)) - { - do_handoff = 1; - - if (next_worker_index != current_worker_index) - { - if (hf) - hf->n_vectors = VLIB_FRAME_SIZE - n_left_to_next_worker; - - hf = vlib_get_worker_handoff_queue_elt (sm->fq_out2in_index, - next_worker_index, - handoff_queue_elt_by_worker_index); - - n_left_to_next_worker = VLIB_FRAME_SIZE - hf->n_vectors; - to_next_worker = &hf->buffer_index[hf->n_vectors]; - current_worker_index = next_worker_index; - } - - /* enqueue to correct worker thread */ - to_next_worker[0] = bi0; - to_next_worker++; - n_left_to_next_worker--; - - if (n_left_to_next_worker == 0) - { - hf->n_vectors = VLIB_FRAME_SIZE; - vlib_put_frame_queue_elt (hf); - current_worker_index = ~0; - handoff_queue_elt_by_worker_index[next_worker_index] = 0; - hf = 0; - } - } - else - { - do_handoff = 0; - /* if this is 1st frame */ - if (!f) - { - f = vlib_get_frame_to_node (vm, snat_out2in_node.index); - to_next = vlib_frame_vector_args (f); - } - - to_next[0] = bi0; - to_next += 1; - f->n_vectors++; - } - - if (PREDICT_FALSE ((node->flags & VLIB_NODE_FLAG_TRACE) - && (b0->flags & VLIB_BUFFER_IS_TRACED))) - { - snat_out2in_worker_handoff_trace_t *t = - vlib_add_trace (vm, node, b0, sizeof (*t)); - t->next_worker_index = next_worker_index; - t->do_handoff = do_handoff; - } - } - - if (f) - vlib_put_frame_to_node (vm, snat_out2in_node.index, f); - - if (hf) - hf->n_vectors = VLIB_FRAME_SIZE - n_left_to_next_worker; - - /* Ship frames to the worker nodes */ - for (i = 0; i < vec_len (handoff_queue_elt_by_worker_index); i++) - { - if (handoff_queue_elt_by_worker_index[i]) - { - hf = handoff_queue_elt_by_worker_index[i]; - /* - * It works better to let the handoff node - * rate-adapt, always ship the handoff queue element. - */ - if (1 || hf->n_vectors == hf->last_n_vectors) - { - vlib_put_frame_queue_elt (hf); - handoff_queue_elt_by_worker_index[i] = 0; - } - else - hf->last_n_vectors = hf->n_vectors; - } - congested_handoff_queue_by_worker_index[i] = - (vlib_frame_queue_t *) (~0); - } - hf = 0; - current_worker_index = ~0; - return frame->n_vectors; -} - -VLIB_REGISTER_NODE (snat_out2in_worker_handoff_node) = { - .function = snat_out2in_worker_handoff_fn, - .name = "snat-out2in-worker-handoff", - .vector_size = sizeof (u32), - .format_trace = format_snat_out2in_worker_handoff_trace, - .type = VLIB_NODE_TYPE_INTERNAL, - - .n_next_nodes = 1, - - .next_nodes = { - [0] = "error-drop", - }, -}; - -VLIB_NODE_FUNCTION_MULTIARCH (snat_out2in_worker_handoff_node, snat_out2in_worker_handoff_fn); - -static inline u32 icmp_out2in_fast (snat_main_t *sm, - vlib_buffer_t * b0, - ip4_header_t * ip0, - icmp46_header_t * icmp0, - u32 sw_if_index0, - vlib_node_runtime_t * node, - u32 next0, - u32 rx_fib_index0) -{ - snat_session_key_t key0, sm0; - icmp_echo_header_t *echo0; - u32 new_addr0, old_addr0; - u16 old_id0, new_id0; - ip_csum_t sum0; - snat_runtime_t * rt = (snat_runtime_t *)node->runtime_data; - - echo0 = (icmp_echo_header_t *)(icmp0+1); - - key0.addr = ip0->dst_address; - key0.port = echo0->identifier; - key0.fib_index = rx_fib_index0; - - if (snat_static_mapping_match(sm, key0, &sm0, 1)) - { - ip4_address_t * first_int_addr; - - if (PREDICT_FALSE(rt->cached_sw_if_index != sw_if_index0)) - { - first_int_addr = - ip4_interface_first_address (sm->ip4_main, sw_if_index0, - 0 /* just want the address */); - rt->cached_sw_if_index = sw_if_index0; - rt->cached_ip4_address = first_int_addr->as_u32; - } - - /* Don't NAT packet aimed at the intfc address */ - if (PREDICT_FALSE(ip0->dst_address.as_u32 == - rt->cached_ip4_address)) - return next0; - - b0->error = node->errors[SNAT_OUT2IN_ERROR_NO_TRANSLATION]; - return SNAT_OUT2IN_NEXT_DROP; - } - - new_addr0 = sm0.addr.as_u32; - new_id0 = sm0.port; - vnet_buffer(b0)->sw_if_index[VLIB_TX] = sm0.fib_index; - - old_addr0 = ip0->dst_address.as_u32; - ip0->dst_address.as_u32 = new_addr0; - - sum0 = ip0->checksum; - sum0 = ip_csum_update (sum0, old_addr0, new_addr0, - ip4_header_t, - dst_address /* changed member */); - ip0->checksum = ip_csum_fold (sum0); - - if (PREDICT_FALSE(new_id0 != echo0->identifier)) - { - old_id0 = echo0->identifier; - echo0->identifier = new_id0; - - sum0 = icmp0->checksum; - sum0 = ip_csum_update (sum0, old_id0, new_id0, icmp_echo_header_t, - identifier); - icmp0->checksum = ip_csum_fold (sum0); - } - - return next0; -} - -static uword -snat_out2in_fast_node_fn (vlib_main_t * vm, - vlib_node_runtime_t * node, - vlib_frame_t * frame) -{ - u32 n_left_from, * from, * to_next; - snat_out2in_next_t next_index; - u32 pkts_processed = 0; - snat_main_t * sm = &snat_main; - - from = vlib_frame_vector_args (frame); - n_left_from = frame->n_vectors; - next_index = node->cached_next_index; - - while (n_left_from > 0) - { - u32 n_left_to_next; - - vlib_get_next_frame (vm, node, next_index, - to_next, n_left_to_next); - - while (n_left_from > 0 && n_left_to_next > 0) - { - u32 bi0; - vlib_buffer_t * b0; - u32 next0 = SNAT_OUT2IN_NEXT_DROP; - u32 sw_if_index0; - ip4_header_t * ip0; - ip_csum_t sum0; - u32 new_addr0, old_addr0; - u16 new_port0, old_port0; - udp_header_t * udp0; - tcp_header_t * tcp0; - icmp46_header_t * icmp0; - snat_session_key_t key0, sm0; - u32 proto0; - u32 rx_fib_index0; - - /* speculatively enqueue b0 to the current next frame */ - bi0 = from[0]; - to_next[0] = bi0; - from += 1; - to_next += 1; - n_left_from -= 1; - n_left_to_next -= 1; - - b0 = vlib_get_buffer (vm, bi0); - - ip0 = vlib_buffer_get_current (b0); - udp0 = ip4_next_header (ip0); - tcp0 = (tcp_header_t *) udp0; - icmp0 = (icmp46_header_t *) udp0; - - sw_if_index0 = vnet_buffer(b0)->sw_if_index[VLIB_RX]; - rx_fib_index0 = ip4_fib_table_get_index_for_sw_if_index(sw_if_index0); - - vnet_feature_next (sw_if_index0, &next0, b0); - - proto0 = ~0; - proto0 = (ip0->protocol == IP_PROTOCOL_UDP) - ? SNAT_PROTOCOL_UDP : proto0; - proto0 = (ip0->protocol == IP_PROTOCOL_TCP) - ? SNAT_PROTOCOL_TCP : proto0; - proto0 = (ip0->protocol == IP_PROTOCOL_ICMP) - ? SNAT_PROTOCOL_ICMP : proto0; - - if (PREDICT_FALSE (proto0 == ~0)) - goto trace00; - - if (PREDICT_FALSE (proto0 == SNAT_PROTOCOL_ICMP)) - { - next0 = icmp_out2in_fast - (sm, b0, ip0, icmp0, sw_if_index0, node, next0, rx_fib_index0); - goto trace00; - } - - key0.addr = ip0->dst_address; - key0.port = udp0->dst_port; - key0.fib_index = rx_fib_index0; - - if (snat_static_mapping_match(sm, key0, &sm0, 1)) - { - b0->error = node->errors[SNAT_OUT2IN_ERROR_NO_TRANSLATION]; - goto trace00; - } - - new_addr0 = sm0.addr.as_u32; - new_port0 = sm0.port; - vnet_buffer(b0)->sw_if_index[VLIB_TX] = sm0.fib_index; - old_addr0 = ip0->dst_address.as_u32; - ip0->dst_address.as_u32 = new_addr0; - - sum0 = ip0->checksum; - sum0 = ip_csum_update (sum0, old_addr0, new_addr0, - ip4_header_t, - dst_address /* changed member */); - ip0->checksum = ip_csum_fold (sum0); - - if (PREDICT_FALSE(new_port0 != udp0->dst_port)) - { - if (PREDICT_TRUE(proto0 == SNAT_PROTOCOL_TCP)) - { - old_port0 = tcp0->ports.dst; - tcp0->ports.dst = new_port0; - - sum0 = tcp0->checksum; - sum0 = ip_csum_update (sum0, old_addr0, new_addr0, - ip4_header_t, - dst_address /* changed member */); - - sum0 = ip_csum_update (sum0, old_port0, new_port0, - ip4_header_t /* cheat */, - length /* changed member */); - tcp0->checksum = ip_csum_fold(sum0); - } - else - { - old_port0 = udp0->dst_port; - udp0->dst_port = new_port0; - udp0->checksum = 0; - } - } - else - { - if (PREDICT_TRUE(proto0 == SNAT_PROTOCOL_TCP)) - { - sum0 = tcp0->checksum; - sum0 = ip_csum_update (sum0, old_addr0, new_addr0, - ip4_header_t, - dst_address /* changed member */); - - tcp0->checksum = ip_csum_fold(sum0); - } - } - - trace00: - - if (PREDICT_FALSE((node->flags & VLIB_NODE_FLAG_TRACE) - && (b0->flags & VLIB_BUFFER_IS_TRACED))) - { - snat_out2in_trace_t *t = - vlib_add_trace (vm, node, b0, sizeof (*t)); - t->sw_if_index = sw_if_index0; - t->next_index = next0; - } - - pkts_processed += next0 != SNAT_OUT2IN_NEXT_DROP; - - /* verify speculative enqueue, maybe switch current next frame */ - vlib_validate_buffer_enqueue_x1 (vm, node, next_index, - to_next, n_left_to_next, - bi0, next0); - } - - vlib_put_next_frame (vm, node, next_index, n_left_to_next); - } - - vlib_node_increment_counter (vm, snat_out2in_fast_node.index, - SNAT_OUT2IN_ERROR_OUT2IN_PACKETS, - pkts_processed); - return frame->n_vectors; -} - -VLIB_REGISTER_NODE (snat_out2in_fast_node) = { - .function = snat_out2in_fast_node_fn, - .name = "snat-out2in-fast", - .vector_size = sizeof (u32), - .format_trace = format_snat_out2in_fast_trace, - .type = VLIB_NODE_TYPE_INTERNAL, - - .n_errors = ARRAY_LEN(snat_out2in_error_strings), - .error_strings = snat_out2in_error_strings, - - .runtime_data_bytes = sizeof (snat_runtime_t), - - .n_next_nodes = SNAT_OUT2IN_N_NEXT, - - /* edit / add dispositions here */ - .next_nodes = { - [SNAT_OUT2IN_NEXT_LOOKUP] = "ip4-lookup", - [SNAT_OUT2IN_NEXT_DROP] = "error-drop", - }, -}; -VLIB_NODE_FUNCTION_MULTIARCH (snat_out2in_fast_node, snat_out2in_fast_node_fn); diff --git a/plugins/snat-plugin/snat/snat.api b/plugins/snat-plugin/snat/snat.api deleted file mode 100644 index a191eed5..00000000 --- a/plugins/snat-plugin/snat/snat.api +++ /dev/null @@ -1,283 +0,0 @@ -/* - * Copyright (c) 2016 Cisco and/or its affiliates. - * Licensed under the Apache License, Version 2.0 (the "License"); - * you may not use this file except in compliance with the License. - * You may obtain a copy of the License at: - * - * http://www.apache.org/licenses/LICENSE-2.0 - * - * Unless required by applicable law or agreed to in writing, software - * distributed under the License is distributed on an "AS IS" BASIS, - * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. - * See the License for the specific language governing permissions and - * limitations under the License. - */ -/** - * @file snat.api - * @brief VPP control-plane API messages. - * - * This file defines VPP control-plane API messages which are generally - * called through a shared memory interface. - */ - -/** \brief Add/del S-NAT address range - @param client_index - opaque cookie to identify the sender - @param context - sender context, to match reply w/ request - @param is_ip4 - 1 if address type is IPv4 - @first_ip_address - first IP address - @last_ip_address - last IP address - @is_add - 1 if add, 0 if delete -*/ -define snat_add_address_range { - u32 client_index; - u32 context; - u8 is_ip4; - u8 first_ip_address[16]; - u8 last_ip_address[16]; - u8 is_add; -}; - -/** \brief Add S-NAT address range reply - @param client_index - opaque cookie to identify the sender - @param context - sender context, to match reply w/ request - @param retval - return code -*/ -define snat_add_address_range_reply { - u32 context; - i32 retval; -}; - -/** \brief Dump S-NAT addresses - @param client_index - opaque cookie to identify the sender - @param context - sender context, to match reply w/ request -*/ -define snat_address_dump { - u32 client_index; - u32 context; -}; - -/** \brief S-NAT address details response - @param context - sender context, to match reply w/ request - @param is_ip4 - 1 if address type is IPv4 - @param ip_address - IP address -*/ -define snat_address_details { - u32 context; - u8 is_ip4; - u8 ip_address[16]; -}; - -/** \brief Enable/disable S-NAT feature on the interface - @param client_index - opaque cookie to identify the sender - @param context - sender context, to match reply w/ request - @param is_add - 1 if add, 0 if delete - @param is_inside - 1 if inside, 0 if outside - @param sw_if_index - software index of the interface -*/ -define snat_interface_add_del_feature { - u32 client_index; - u32 context; - u8 is_add; - u8 is_inside; - u32 sw_if_index; -}; - -/** \brief Enable/disable S-NAT feature on the interface reply - @param client_index - opaque cookie to identify the sender - @param context - sender context, to match reply w/ request - @param retval - return code -*/ -define snat_interface_add_del_feature_reply { - u32 context; - i32 retval; -}; - -/** \brief Dump interfaces with S-NAT feature - @param client_index - opaque cookie to identify the sender - @param context - sender context, to match reply w/ request -*/ -define snat_interface_dump { - u32 client_index; - u32 context; -}; - -/** \brief S-NAT interface details response - @param context - sender context, to match reply w/ request - @param is_inside - 1 if inside, 0 if outside - @param sw_if_index - software index of the interface -*/ -define snat_interface_details { - u32 context; - u8 is_inside; - u32 sw_if_index; -}; - -/** \brief Add/delete S-NAT static mapping - @param client_index - opaque cookie to identify the sender - @param context - sender context, to match reply w/ request - @param is_add - 1 if add, 0 if delete - @param is_ip4 - 1 if address type is IPv4 - @param addr_only - 1 if address only mapping - @param local_ip_address - local IP address - @param external_ip_address - external IP address - @param local_port - local port number - @param external_port - external port number - @param vfr_id - VRF ID -*/ -define snat_add_static_mapping { - u32 client_index; - u32 context; - u8 is_add; - u8 is_ip4; - u8 addr_only; - u8 local_ip_address[16]; - u8 external_ip_address[16]; - u16 local_port; - u16 external_port; - u32 vrf_id; -}; - -/** \brief Add/delete S-NAT static mapping reply - @param client_index - opaque cookie to identify the sender - @param context - sender context, to match reply w/ request - @param retval - return code -*/ -define snat_add_static_mapping_reply { - u32 context; - i32 retval; -}; - -/** \brief Dump S-NAT static mappings - @param client_index - opaque cookie to identify the sender - @param context - sender context, to match reply w/ request -*/ -define snat_static_mapping_dump { - u32 client_index; - u32 context; -}; - -/** \brief S-NAT static mapping details response - @param context - sender context, to match reply w/ request - @param is_ip4 - 1 if address type is IPv4 - @param addr_only - 1 if address only mapping - @param local_ip_address - local IP address - @param external_ip_address - external IP address - @param local_port - local port number - @param external_port - external port number - @param vfr_id - VRF ID -*/ -define snat_static_mapping_details { - u32 context; - u8 is_ip4; - u8 addr_only; - u8 local_ip_address[16]; - u8 external_ip_address[16]; - u16 local_port; - u16 external_port; - u32 vrf_id; -}; - -/** \brief Control ping from client to api server request - @param client_index - opaque cookie to identify the sender - @param context - sender context, to match reply w/ request -*/ -define snat_control_ping -{ - u32 client_index; - u32 context; -}; - -/** \brief Control ping from the client to the server response - @param client_index - opaque cookie to identify the sender - @param context - sender context, to match reply w/ request - @param retval - return code for the request - @param vpe_pid - the pid of the vpe, returned by the server -*/ -define snat_control_ping_reply -{ - u32 context; - i32 retval; - u32 client_index; - u32 vpe_pid; -}; - -/** \brief Show S-NAT plugin startup config - @param client_index - opaque cookie to identify the sender - @param context - sender context, to match reply w/ request -*/ -define snat_show_config -{ - u32 client_index; - u32 context; -}; - -/** \brief Show S-NAT plugin startup config reply - @param context - sender context, to match reply w/ request - @param retval - return code for the request - @param static_mapping_only - if 1 dynamic translations disabled - @param static_mapping_connection_tracking - if 1 create session data - @param translation_buckets - number of translation hash buckets - @param translation_memory_size - translation hash memory size - @param user_buckets - number of user hash buckets - @param user_memory_size - user hash memory size - @param max_translations_per_user - maximum number of translations per user - @param outside_vrf_id - outside VRF id - @param inside_vrf_id - default inside VRF id -*/ -define snat_show_config_reply -{ - u32 context; - i32 retval; - u8 static_mapping_only; - u8 static_mapping_connection_tracking; - u32 translation_buckets; - u32 translation_memory_size; - u32 user_buckets; - u32 user_memory_size; - u32 max_translations_per_user; - u32 outside_vrf_id; - u32 inside_vrf_id; -}; - -/** \brief Set S-NAT workers - @param client_index - opaque cookie to identify the sender - @param context - sender context, to match reply w/ request - @param worker_mask - S-NAT workers mask -*/ -define snat_set_workers { - u32 client_index; - u32 context; - u64 worker_mask; -}; - -/** \brief Set S-NAT workers reply - @param client_index - opaque cookie to identify the sender - @param context - sender context, to match reply w/ request - @param retval - return code -*/ -define snat_set_workers_reply { - u32 context; - i32 retval; -}; - -/** \brief Dump S-NAT workers - @param client_index - opaque cookie to identify the sender - @param context - sender context, to match reply w/ request -*/ -define snat_worker_dump { - u32 client_index; - u32 context; -}; - -/** \brief S-NAT workers details response - @param context - sender context, to match reply w/ request - @param worker_index - worker index - @param lcore_id - lcore ID - @param name - worker name -*/ -define snat_worker_details { - u32 context; - u32 worker_index; - u32 lcore_id; - u8 name[64]; -}; diff --git a/plugins/snat-plugin/snat/snat.c b/plugins/snat-plugin/snat/snat.c deleted file mode 100644 index bc995684..00000000 --- a/plugins/snat-plugin/snat/snat.c +++ /dev/null @@ -1,1957 +0,0 @@ -/* - * snat.c - simple nat plugin - * - * Copyright (c) 2016 Cisco and/or its affiliates. - * Licensed under the Apache License, Version 2.0 (the "License"); - * you may not use this file except in compliance with the License. - * You may obtain a copy of the License at: - * - * http://www.apache.org/licenses/LICENSE-2.0 - * - * Unless required by applicable law or agreed to in writing, software - * distributed under the License is distributed on an "AS IS" BASIS, - * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. - * See the License for the specific language governing permissions and - * limitations under the License. - */ - -#include -#include -#include -#include - -#include -#include -#include - -snat_main_t snat_main; - -/* define message IDs */ -#include - -/* define message structures */ -#define vl_typedefs -#include -#undef vl_typedefs - -/* define generated endian-swappers */ -#define vl_endianfun -#include -#undef vl_endianfun - -#define vl_print(handle, ...) vlib_cli_output (handle, __VA_ARGS__) - -/* Get the API version number */ -#define vl_api_version(n,v) static u32 api_version=(v); -#include -#undef vl_api_version - -/* Macro to finish up custom dump fns */ -#define FINISH \ - vec_add1 (s, 0); \ - vl_print (handle, (char *)s); \ - vec_free (s); \ - return handle; - -/* - * A handy macro to set up a message reply. - * Assumes that the following variables are available: - * mp - pointer to request message - * rmp - pointer to reply message type - * rv - return value - */ - -#define REPLY_MACRO(t) \ -do { \ - unix_shared_memory_queue_t * q = \ - vl_api_client_index_to_input_queue (mp->client_index); \ - if (!q) \ - return; \ - \ - rmp = vl_msg_api_alloc (sizeof (*rmp)); \ - rmp->_vl_msg_id = ntohs((t)+sm->msg_id_base); \ - rmp->context = mp->context; \ - rmp->retval = ntohl(rv); \ - \ - vl_msg_api_send_shmem (q, (u8 *)&rmp); \ -} while(0); - -#define REPLY_MACRO2(t, body) \ -do { \ - unix_shared_memory_queue_t * q = \ - vl_api_client_index_to_input_queue (mp->client_index); \ - if (!q) \ - return; \ - \ - rmp = vl_msg_api_alloc (sizeof (*rmp)); \ - rmp->_vl_msg_id = ntohs((t)+sm->msg_id_base); \ - rmp->context = mp->context; \ - rmp->retval = ntohl(rv); \ - do {body;} while (0); \ - vl_msg_api_send_shmem (q, (u8 *)&rmp); \ -} while(0); - - -/* Hook up input features */ -VNET_FEATURE_INIT (ip4_snat_in2out, static) = { - .arc_name = "ip4-unicast", - .node_name = "snat-in2out", - .runs_before = VNET_FEATURES ("snat-out2in"), -}; -VNET_FEATURE_INIT (ip4_snat_out2in, static) = { - .arc_name = "ip4-unicast", - .node_name = "snat-out2in", - .runs_before = VNET_FEATURES ("ip4-lookup"), -}; -VNET_FEATURE_INIT (ip4_snat_in2out_worker_handoff, static) = { - .arc_name = "ip4-unicast", - .node_name = "snat-in2out-worker-handoff", - .runs_before = VNET_FEATURES ("snat-out2in-worker-handoff"), -}; -VNET_FEATURE_INIT (ip4_snat_out2in_worker_handoff, static) = { - .arc_name = "ip4-unicast", - .node_name = "snat-out2in-worker-handoff", - .runs_before = VNET_FEATURES ("ip4-lookup"), -}; -VNET_FEATURE_INIT (ip4_snat_in2out_fast, static) = { - .arc_name = "ip4-unicast", - .node_name = "snat-in2out-fast", - .runs_before = VNET_FEATURES ("snat-out2in-fast"), -}; -VNET_FEATURE_INIT (ip4_snat_out2in_fast, static) = { - .arc_name = "ip4-unicast", - .node_name = "snat-out2in-fast", - .runs_before = VNET_FEATURES ("ip4-lookup"), -}; - - -/* - * This routine exists to convince the vlib plugin framework that - * we haven't accidentally copied a random .dll into the plugin directory. - * - * Also collects global variable pointers passed from the vpp engine - */ - -clib_error_t * -vlib_plugin_register (vlib_main_t * vm, vnet_plugin_handoff_t * h, - int from_early_init) -{ - snat_main_t * sm = &snat_main; - clib_error_t * error = 0; - - sm->vlib_main = vm; - sm->vnet_main = h->vnet_main; - sm->ethernet_main = h->ethernet_main; - - return error; -} - -/*$$$$$ move to an installed header file */ -#if (1 || CLIB_DEBUG > 0) /* "trust, but verify" */ - -#define VALIDATE_SW_IF_INDEX(mp) \ - do { u32 __sw_if_index = ntohl(mp->sw_if_index); \ - vnet_main_t *__vnm = vnet_get_main(); \ - if (pool_is_free_index(__vnm->interface_main.sw_interfaces, \ - __sw_if_index)) { \ - rv = VNET_API_ERROR_INVALID_SW_IF_INDEX; \ - goto bad_sw_if_index; \ - } \ -} while(0); - -#define BAD_SW_IF_INDEX_LABEL \ -do { \ -bad_sw_if_index: \ - ; \ -} while (0); - -#define VALIDATE_RX_SW_IF_INDEX(mp) \ - do { u32 __rx_sw_if_index = ntohl(mp->rx_sw_if_index); \ - vnet_main_t *__vnm = vnet_get_main(); \ - if (pool_is_free_index(__vnm->interface_main.sw_interfaces, \ - __rx_sw_if_index)) { \ - rv = VNET_API_ERROR_INVALID_SW_IF_INDEX; \ - goto bad_rx_sw_if_index; \ - } \ -} while(0); - -#define BAD_RX_SW_IF_INDEX_LABEL \ -do { \ -bad_rx_sw_if_index: \ - ; \ -} while (0); - -#define VALIDATE_TX_SW_IF_INDEX(mp) \ - do { u32 __tx_sw_if_index = ntohl(mp->tx_sw_if_index); \ - vnet_main_t *__vnm = vnet_get_main(); \ - if (pool_is_free_index(__vnm->interface_main.sw_interfaces, \ - __tx_sw_if_index)) { \ - rv = VNET_API_ERROR_INVALID_SW_IF_INDEX; \ - goto bad_tx_sw_if_index; \ - } \ -} while(0); - -#define BAD_TX_SW_IF_INDEX_LABEL \ -do { \ -bad_tx_sw_if_index: \ - ; \ -} while (0); - -#else - -#define VALIDATE_SW_IF_INDEX(mp) -#define BAD_SW_IF_INDEX_LABEL -#define VALIDATE_RX_SW_IF_INDEX(mp) -#define BAD_RX_SW_IF_INDEX_LABEL -#define VALIDATE_TX_SW_IF_INDEX(mp) -#define BAD_TX_SW_IF_INDEX_LABEL - -#endif /* CLIB_DEBUG > 0 */ - -void snat_add_address (snat_main_t *sm, ip4_address_t *addr) -{ - snat_address_t * ap; - - /* Check if address already exists */ - vec_foreach (ap, sm->addresses) - { - if (ap->addr.as_u32 == addr->as_u32) - return; - } - - vec_add2 (sm->addresses, ap, 1); - ap->addr = *addr; - clib_bitmap_alloc (ap->busy_port_bitmap, 65535); -} - -static int is_snat_address_used_in_static_mapping (snat_main_t *sm, - ip4_address_t addr) -{ - snat_static_mapping_t *m; - pool_foreach (m, sm->static_mappings, - ({ - if (m->external_addr.as_u32 == addr.as_u32) - return 1; - })); - - return 0; -} - -int snat_del_address (snat_main_t *sm, ip4_address_t addr) -{ - snat_address_t *a = 0; - snat_session_t *ses; - u32 *ses_to_be_removed = 0, *ses_index; - clib_bihash_kv_8_8_t kv, value; - snat_user_key_t user_key; - snat_user_t *u; - snat_main_per_thread_data_t *tsm; - - int i; - - /* Find SNAT address */ - for (i=0; i < vec_len (sm->addresses); i++) - { - if (sm->addresses[i].addr.as_u32 == addr.as_u32) - { - a = sm->addresses + i; - break; - } - } - if (!a) - return VNET_API_ERROR_NO_SUCH_ENTRY; - - /* Check if address is used in some static mapping */ - if (is_snat_address_used_in_static_mapping(sm, addr)) - { - clib_warning ("address used in static mapping"); - return VNET_API_ERROR_UNSPECIFIED; - } - - /* Delete sessions using address */ - if (a->busy_ports) - { - vec_foreach (tsm, sm->per_thread_data) - { - pool_foreach (ses, tsm->sessions, ({ - if (ses->out2in.addr.as_u32 == addr.as_u32) - { - vec_add1 (ses_to_be_removed, ses - tsm->sessions); - kv.key = ses->in2out.as_u64; - clib_bihash_add_del_8_8 (&sm->in2out, &kv, 0); - kv.key = ses->out2in.as_u64; - clib_bihash_add_del_8_8 (&sm->out2in, &kv, 0); - clib_dlist_remove (tsm->list_pool, ses->per_user_index); - user_key.addr = ses->in2out.addr; - user_key.fib_index = ses->in2out.fib_index; - kv.key = user_key.as_u64; - if (!clib_bihash_search_8_8 (&sm->user_hash, &kv, &value)) - { - u = pool_elt_at_index (tsm->users, value.value); - u->nsessions--; - } - } - })); - - vec_foreach (ses_index, ses_to_be_removed) - pool_put_index (tsm->sessions, ses_index[0]); - - vec_free (ses_to_be_removed); - } - } - - vec_del1 (sm->addresses, i); - - return 0; -} - -static void increment_v4_address (ip4_address_t * a) -{ - u32 v; - - v = clib_net_to_host_u32(a->as_u32) + 1; - a->as_u32 = clib_host_to_net_u32(v); -} - -/** - * @brief Add static mapping. - * - * Create static mapping between local addr+port and external addr+port. - * - * @param l_addr Local IPv4 address. - * @param e_addr External IPv4 address. - * @param l_port Local port number. - * @param e_port External port number. - * @param vrf_id VRF ID. - * @param addr_only If 0 address port and pair mapping, otherwise address only. - * @param is_add If 0 delete static mapping, otherwise add. - * - * @returns - */ -int snat_add_static_mapping(ip4_address_t l_addr, ip4_address_t e_addr, - u16 l_port, u16 e_port, u32 vrf_id, int addr_only, - int is_add) -{ - snat_main_t * sm = &snat_main; - snat_static_mapping_t *m; - snat_static_mapping_key_t m_key; - clib_bihash_kv_8_8_t kv, value; - snat_address_t *a = 0; - u32 fib_index = ~0; - uword * p; - int i; - - /* If outside FIB index is not resolved yet */ - if (sm->outside_fib_index == ~0) - { - p = hash_get (sm->ip4_main->fib_index_by_table_id, sm->outside_vrf_id); - if (!p) - return VNET_API_ERROR_NO_SUCH_FIB; - sm->outside_fib_index = p[0]; - } - - m_key.addr = e_addr; - m_key.port = addr_only ? 0 : e_port; - m_key.fib_index = sm->outside_fib_index; - kv.key = m_key.as_u64; - if (clib_bihash_search_8_8 (&sm->static_mapping_by_external, &kv, &value)) - m = 0; - else - m = pool_elt_at_index (sm->static_mappings, value.value); - - if (is_add) - { - if (m) - return VNET_API_ERROR_VALUE_EXIST; - - /* Convert VRF id to FIB index */ - if (vrf_id != ~0) - { - p = hash_get (sm->ip4_main->fib_index_by_table_id, vrf_id); - if (!p) - return VNET_API_ERROR_NO_SUCH_FIB; - fib_index = p[0]; - } - /* If not specified use inside VRF id from SNAT plugin startup config */ - else - { - if (sm->inside_fib_index == ~0) - { - p = hash_get (sm->ip4_main->fib_index_by_table_id, sm->inside_vrf_id); - if (!p) - return VNET_API_ERROR_NO_SUCH_FIB; - fib_index = p[0]; - sm->inside_fib_index = fib_index; - } - else - fib_index = sm->inside_fib_index; - - vrf_id = sm->inside_vrf_id; - } - - /* Find external address in allocated addresses and reserve port for - address and port pair mapping when dynamic translations enabled */ - if (!addr_only && !(sm->static_mapping_only)) - { - for (i = 0; i < vec_len (sm->addresses); i++) - { - if (sm->addresses[i].addr.as_u32 == e_addr.as_u32) - { - a = sm->addresses + i; - /* External port must be unused */ - if (clib_bitmap_get_no_check (a->busy_port_bitmap, e_port)) - return VNET_API_ERROR_INVALID_VALUE; - clib_bitmap_set_no_check (a->busy_port_bitmap, e_port, 1); - if (e_port > 1024) - a->busy_ports++; - - break; - } - } - /* External address must be allocated */ - if (!a) - return VNET_API_ERROR_NO_SUCH_ENTRY; - } - - pool_get (sm->static_mappings, m); - memset (m, 0, sizeof (*m)); - m->local_addr = l_addr; - m->external_addr = e_addr; - m->addr_only = addr_only; - m->vrf_id = vrf_id; - m->fib_index = fib_index; - if (!addr_only) - { - m->local_port = l_port; - m->external_port = e_port; - } - - m_key.addr = m->local_addr; - m_key.port = m->local_port; - m_key.fib_index = m->fib_index; - kv.key = m_key.as_u64; - kv.value = m - sm->static_mappings; - clib_bihash_add_del_8_8(&sm->static_mapping_by_local, &kv, 1); - - m_key.addr = m->external_addr; - m_key.port = m->external_port; - m_key.fib_index = sm->outside_fib_index; - kv.key = m_key.as_u64; - kv.value = m - sm->static_mappings; - clib_bihash_add_del_8_8(&sm->static_mapping_by_external, &kv, 1); - - /* Assign worker */ - if (sm->workers) - { - snat_user_key_t w_key0; - snat_static_mapping_key_t w_key1; - - w_key0.addr = m->local_addr; - w_key0.fib_index = m->fib_index; - kv.key = w_key0.as_u64; - - if (clib_bihash_search_8_8 (&sm->worker_by_in, &kv, &value)) - { - kv.value = sm->first_worker_index + - sm->workers[sm->next_worker++ % vec_len (sm->workers)]; - - clib_bihash_add_del_8_8 (&sm->worker_by_in, &kv, 1); - } - else - { - kv.value = value.value; - } - - w_key1.addr = m->external_addr; - w_key1.port = clib_host_to_net_u16 (m->external_port); - w_key1.fib_index = sm->outside_fib_index; - kv.key = w_key1.as_u64; - clib_bihash_add_del_8_8 (&sm->worker_by_out, &kv, 1); - } - } - else - { - if (!m) - return VNET_API_ERROR_NO_SUCH_ENTRY; - - /* Free external address port */ - if (!addr_only && !(sm->static_mapping_only)) - { - for (i = 0; i < vec_len (sm->addresses); i++) - { - if (sm->addresses[i].addr.as_u32 == e_addr.as_u32) - { - a = sm->addresses + i; - clib_bitmap_set_no_check (a->busy_port_bitmap, e_port, 0); - a->busy_ports--; - - break; - } - } - } - - m_key.addr = m->local_addr; - m_key.port = m->local_port; - m_key.fib_index = m->fib_index; - kv.key = m_key.as_u64; - clib_bihash_add_del_8_8(&sm->static_mapping_by_local, &kv, 0); - - m_key.addr = m->external_addr; - m_key.port = m->external_port; - m_key.fib_index = sm->outside_fib_index; - kv.key = m_key.as_u64; - clib_bihash_add_del_8_8(&sm->static_mapping_by_external, &kv, 0); - - /* Delete session(s) for static mapping if exist */ - if (!(sm->static_mapping_only) || - (sm->static_mapping_only && sm->static_mapping_connection_tracking)) - { - snat_user_key_t u_key; - snat_user_t *u; - dlist_elt_t * head, * elt; - u32 elt_index, head_index, del_elt_index; - u32 ses_index; - u64 user_index; - snat_session_t * s; - snat_main_per_thread_data_t *tsm; - - u_key.addr = m->local_addr; - u_key.fib_index = m->fib_index; - kv.key = u_key.as_u64; - if (!clib_bihash_search_8_8 (&sm->user_hash, &kv, &value)) - { - user_index = value.value; - if (!clib_bihash_search_8_8 (&sm->worker_by_in, &kv, &value)) - tsm = vec_elt_at_index (sm->per_thread_data, value.value); - else - tsm = vec_elt_at_index (sm->per_thread_data, sm->num_workers); - u = pool_elt_at_index (tsm->users, user_index); - if (u->nstaticsessions) - { - head_index = u->sessions_per_user_list_head_index; - head = pool_elt_at_index (tsm->list_pool, head_index); - elt_index = head->next; - elt = pool_elt_at_index (tsm->list_pool, elt_index); - ses_index = elt->value; - while (ses_index != ~0) - { - s = pool_elt_at_index (tsm->sessions, ses_index); - del_elt_index = elt_index; - elt_index = elt->next; - elt = pool_elt_at_index (tsm->list_pool, elt_index); - ses_index = elt->value; - - if (!addr_only) - { - if ((s->out2in.addr.as_u32 != e_addr.as_u32) && - (clib_net_to_host_u16 (s->out2in.port) != e_port)) - continue; - } - - value.key = s->in2out.as_u64; - clib_bihash_add_del_8_8 (&sm->in2out, &value, 0); - value.key = s->out2in.as_u64; - clib_bihash_add_del_8_8 (&sm->out2in, &value, 0); - pool_put (tsm->sessions, s); - - clib_dlist_remove (tsm->list_pool, del_elt_index); - pool_put_index (tsm->list_pool, del_elt_index); - u->nstaticsessions--; - - if (!addr_only) - break; - } - if (addr_only) - { - pool_put (tsm->users, u); - clib_bihash_add_del_8_8 (&sm->user_hash, &kv, 0); - } - } - } - } - - /* Delete static mapping from pool */ - pool_put (sm->static_mappings, m); - } - - return 0; -} - -static int snat_interface_add_del (u32 sw_if_index, u8 is_inside, int is_del) -{ - snat_main_t *sm = &snat_main; - snat_interface_t *i; - const char * feature_name; - - if (sm->static_mapping_only && !(sm->static_mapping_connection_tracking)) - feature_name = is_inside ? "snat-in2out-fast" : "snat-out2in-fast"; - else - { - if (sm->num_workers > 1) - feature_name = is_inside ? "snat-in2out-worker-handoff" : "snat-out2in-worker-handoff"; - else - feature_name = is_inside ? "snat-in2out" : "snat-out2in"; - } - - vnet_feature_enable_disable ("ip4-unicast", feature_name, sw_if_index, - !is_del, 0, 0); - - if (sm->fq_in2out_index == ~0) - sm->fq_in2out_index = vlib_frame_queue_main_init (snat_in2out_node.index, 0); - - if (sm->fq_out2in_index == ~0) - sm->fq_out2in_index = vlib_frame_queue_main_init (snat_out2in_node.index, 0); - - pool_foreach (i, sm->interfaces, - ({ - if (i->sw_if_index == sw_if_index) - { - if (is_del) - pool_put (sm->interfaces, i); - else - return VNET_API_ERROR_VALUE_EXIST; - - return 0; - } - })); - - if (is_del) - return VNET_API_ERROR_NO_SUCH_ENTRY; - - pool_get (sm->interfaces, i); - i->sw_if_index = sw_if_index; - i->is_inside = is_inside; - - return 0; -} - -static int snat_set_workers (uword * bitmap) -{ - snat_main_t *sm = &snat_main; - int i; - - if (sm->num_workers < 2) - return VNET_API_ERROR_FEATURE_DISABLED; - - if (clib_bitmap_last_set (bitmap) >= sm->num_workers) - return VNET_API_ERROR_INVALID_WORKER; - - vec_free (sm->workers); - clib_bitmap_foreach (i, bitmap, - ({ - vec_add1(sm->workers, i); - })); - - return 0; -} - -static void -vl_api_snat_add_address_range_t_handler -(vl_api_snat_add_address_range_t * mp) -{ - snat_main_t * sm = &snat_main; - vl_api_snat_add_address_range_reply_t * rmp; - ip4_address_t this_addr; - u32 start_host_order, end_host_order; - int i, count; - int rv = 0; - u32 * tmp; - - if (mp->is_ip4 != 1) - { - rv = VNET_API_ERROR_UNIMPLEMENTED; - goto send_reply; - } - - if (sm->static_mapping_only) - { - rv = VNET_API_ERROR_FEATURE_DISABLED; - goto send_reply; - } - - tmp = (u32 *) mp->first_ip_address; - start_host_order = clib_host_to_net_u32 (tmp[0]); - tmp = (u32 *) mp->last_ip_address; - end_host_order = clib_host_to_net_u32 (tmp[0]); - - count = (end_host_order - start_host_order) + 1; - - if (count > 1024) - clib_warning ("%U - %U, %d addresses...", - format_ip4_address, mp->first_ip_address, - format_ip4_address, mp->last_ip_address, - count); - - memcpy (&this_addr.as_u8, mp->first_ip_address, 4); - - for (i = 0; i < count; i++) - { - if (mp->is_add) - snat_add_address (sm, &this_addr); - else - rv = snat_del_address (sm, this_addr); - - if (rv) - goto send_reply; - - increment_v4_address (&this_addr); - } - - send_reply: - REPLY_MACRO (VL_API_SNAT_ADD_ADDRESS_RANGE_REPLY); -} - -static void *vl_api_snat_add_address_range_t_print -(vl_api_snat_add_address_range_t *mp, void * handle) -{ - u8 * s; - - s = format (0, "SCRIPT: snat_add_address_range "); - s = format (s, "%U ", format_ip4_address, mp->first_ip_address); - if (memcmp (mp->first_ip_address, mp->last_ip_address, 4)) - { - s = format (s, " - %U ", format_ip4_address, mp->last_ip_address); - } - FINISH; -} - -static void -send_snat_address_details -(snat_address_t * a, unix_shared_memory_queue_t * q, u32 context) -{ - vl_api_snat_address_details_t *rmp; - snat_main_t * sm = &snat_main; - - rmp = vl_msg_api_alloc (sizeof (*rmp)); - memset (rmp, 0, sizeof (*rmp)); - rmp->_vl_msg_id = ntohs (VL_API_SNAT_ADDRESS_DETAILS+sm->msg_id_base); - rmp->is_ip4 = 1; - clib_memcpy (rmp->ip_address, &(a->addr), 4); - rmp->context = context; - - vl_msg_api_send_shmem (q, (u8 *) & rmp); -} - -static void -vl_api_snat_address_dump_t_handler -(vl_api_snat_address_dump_t * mp) -{ - unix_shared_memory_queue_t *q; - snat_main_t * sm = &snat_main; - snat_address_t * a; - - q = vl_api_client_index_to_input_queue (mp->client_index); - if (q == 0) - return; - - vec_foreach (a, sm->addresses) - send_snat_address_details (a, q, mp->context); -} - -static void *vl_api_snat_address_dump_t_print -(vl_api_snat_address_dump_t *mp, void * handle) -{ - u8 *s; - - s = format (0, "SCRIPT: snat_address_dump "); - - FINISH; -} - -static void -vl_api_snat_interface_add_del_feature_t_handler -(vl_api_snat_interface_add_del_feature_t * mp) -{ - snat_main_t * sm = &snat_main; - vl_api_snat_interface_add_del_feature_reply_t * rmp; - u8 is_del = mp->is_add == 0; - u32 sw_if_index = ntohl(mp->sw_if_index); - int rv = 0; - - VALIDATE_SW_IF_INDEX(mp); - - rv = snat_interface_add_del (sw_if_index, mp->is_inside, is_del); - - BAD_SW_IF_INDEX_LABEL; - - REPLY_MACRO(VL_API_SNAT_INTERFACE_ADD_DEL_FEATURE_REPLY); -} - -static void *vl_api_snat_interface_add_del_feature_t_print -(vl_api_snat_interface_add_del_feature_t * mp, void *handle) -{ - u8 * s; - - s = format (0, "SCRIPT: snat_interface_add_del_feature "); - s = format (s, "sw_if_index %d %s %s", - clib_host_to_net_u32(mp->sw_if_index), - mp->is_inside ? "in":"out", - mp->is_add ? "" : "del"); - - FINISH; -} - -static void -send_snat_interface_details -(snat_interface_t * i, unix_shared_memory_queue_t * q, u32 context) -{ - vl_api_snat_interface_details_t *rmp; - snat_main_t * sm = &snat_main; - - rmp = vl_msg_api_alloc (sizeof (*rmp)); - memset (rmp, 0, sizeof (*rmp)); - rmp->_vl_msg_id = ntohs (VL_API_SNAT_INTERFACE_DETAILS+sm->msg_id_base); - rmp->sw_if_index = ntohl (i->sw_if_index); - rmp->is_inside = i->is_inside; - rmp->context = context; - - vl_msg_api_send_shmem (q, (u8 *) & rmp); -} - -static void -vl_api_snat_interface_dump_t_handler -(vl_api_snat_interface_dump_t * mp) -{ - unix_shared_memory_queue_t *q; - snat_main_t * sm = &snat_main; - snat_interface_t * i; - - q = vl_api_client_index_to_input_queue (mp->client_index); - if (q == 0) - return; - - pool_foreach (i, sm->interfaces, - ({ - send_snat_interface_details(i, q, mp->context); - })); -} - -static void *vl_api_snat_interface_dump_t_print -(vl_api_snat_interface_dump_t *mp, void * handle) -{ - u8 *s; - - s = format (0, "SCRIPT: snat_interface_dump "); - - FINISH; -}static void - -vl_api_snat_add_static_mapping_t_handler -(vl_api_snat_add_static_mapping_t * mp) -{ - snat_main_t * sm = &snat_main; - vl_api_snat_add_static_mapping_reply_t * rmp; - ip4_address_t local_addr, external_addr; - u16 local_port = 0, external_port = 0; - u32 vrf_id; - int rv = 0; - - if (mp->is_ip4 != 1) - { - rv = VNET_API_ERROR_UNIMPLEMENTED; - goto send_reply; - } - - memcpy (&local_addr.as_u8, mp->local_ip_address, 4); - memcpy (&external_addr.as_u8, mp->external_ip_address, 4); - if (mp->addr_only == 0) - { - local_port = clib_net_to_host_u16 (mp->local_port); - external_port = clib_net_to_host_u16 (mp->external_port); - } - vrf_id = clib_net_to_host_u32 (mp->vrf_id); - - rv = snat_add_static_mapping(local_addr, external_addr, local_port, - external_port, vrf_id, mp->addr_only, - mp->is_add); - - send_reply: - REPLY_MACRO (VL_API_SNAT_ADD_ADDRESS_RANGE_REPLY); -} - -static void *vl_api_snat_add_static_mapping_t_print -(vl_api_snat_add_static_mapping_t *mp, void * handle) -{ - u8 * s; - - s = format (0, "SCRIPT: snat_add_static_mapping "); - s = format (s, "local_addr %U external_addr %U ", - format_ip4_address, mp->local_ip_address, - format_ip4_address, mp->external_ip_address); - - if (mp->addr_only == 0) - s = format (s, "local_port %d external_port %d ", - clib_net_to_host_u16 (mp->local_port), - clib_net_to_host_u16 (mp->external_port)); - - if (mp->vrf_id != ~0) - s = format (s, "vrf %d", clib_net_to_host_u32 (mp->vrf_id)); - - FINISH; -} - -static void -send_snat_static_mapping_details -(snat_static_mapping_t * m, unix_shared_memory_queue_t * q, u32 context) -{ - vl_api_snat_static_mapping_details_t *rmp; - snat_main_t * sm = &snat_main; - - rmp = vl_msg_api_alloc (sizeof (*rmp)); - memset (rmp, 0, sizeof (*rmp)); - rmp->_vl_msg_id = ntohs (VL_API_SNAT_STATIC_MAPPING_DETAILS+sm->msg_id_base); - rmp->is_ip4 = 1; - rmp->addr_only = m->addr_only; - clib_memcpy (rmp->local_ip_address, &(m->local_addr), 4); - clib_memcpy (rmp->external_ip_address, &(m->external_addr), 4); - rmp->local_port = htons (m->local_port); - rmp->external_port = htons (m->external_port); - rmp->vrf_id = htonl (m->vrf_id); - rmp->context = context; - - vl_msg_api_send_shmem (q, (u8 *) & rmp); -} - -static void -vl_api_snat_static_mapping_dump_t_handler -(vl_api_snat_static_mapping_dump_t * mp) -{ - unix_shared_memory_queue_t *q; - snat_main_t * sm = &snat_main; - snat_static_mapping_t * m; - - q = vl_api_client_index_to_input_queue (mp->client_index); - if (q == 0) - return; - - pool_foreach (m, sm->static_mappings, - ({ - send_snat_static_mapping_details (m, q, mp->context); - })); -} - -static void *vl_api_snat_static_mapping_dump_t_print -(vl_api_snat_static_mapping_dump_t *mp, void * handle) -{ - u8 *s; - - s = format (0, "SCRIPT: snat_static_mapping_dump "); - - FINISH; -} - -static void -vl_api_snat_control_ping_t_handler -(vl_api_snat_control_ping_t * mp) -{ - vl_api_snat_control_ping_reply_t *rmp; - snat_main_t * sm = &snat_main; - int rv = 0; - - REPLY_MACRO2(VL_API_SNAT_CONTROL_PING_REPLY, - ({ - rmp->vpe_pid = ntohl (getpid()); - })); -} - -static void *vl_api_snat_control_ping_t_print -(vl_api_snat_control_ping_t *mp, void * handle) -{ - u8 *s; - - s = format (0, "SCRIPT: snat_control_ping "); - - FINISH; -} - -static void -vl_api_snat_show_config_t_handler -(vl_api_snat_show_config_t * mp) -{ - vl_api_snat_show_config_reply_t *rmp; - snat_main_t * sm = &snat_main; - int rv = 0; - - REPLY_MACRO2(VL_API_SNAT_SHOW_CONFIG_REPLY, - ({ - rmp->translation_buckets = htonl (sm->translation_buckets); - rmp->translation_memory_size = htonl (sm->translation_memory_size); - rmp->user_buckets = htonl (sm->user_buckets); - rmp->user_memory_size = htonl (sm->user_memory_size); - rmp->max_translations_per_user = htonl (sm->max_translations_per_user); - rmp->outside_vrf_id = htonl (sm->outside_vrf_id); - rmp->inside_vrf_id = htonl (sm->inside_vrf_id); - rmp->static_mapping_only = sm->static_mapping_only; - rmp->static_mapping_connection_tracking = - sm->static_mapping_connection_tracking; - })); -} - -static void *vl_api_snat_show_config_t_print -(vl_api_snat_show_config_t *mp, void * handle) -{ - u8 *s; - - s = format (0, "SCRIPT: snat_show_config "); - - FINISH; -} - -static void -vl_api_snat_set_workers_t_handler -(vl_api_snat_set_workers_t * mp) -{ - snat_main_t * sm = &snat_main; - vl_api_snat_set_workers_reply_t * rmp; - int rv = 0; - uword *bitmap = 0; - u64 mask = clib_net_to_host_u64 (mp->worker_mask); - - if (sm->num_workers < 2) - { - rv = VNET_API_ERROR_FEATURE_DISABLED; - goto send_reply; - } - - bitmap = clib_bitmap_set_multiple (bitmap, 0, mask, BITS (mask)); - rv = snat_set_workers(bitmap); - clib_bitmap_free (bitmap); - - send_reply: - REPLY_MACRO (VL_API_SNAT_SET_WORKERS_REPLY); -} - -static void *vl_api_snat_set_workers_t_print -(vl_api_snat_set_workers_t *mp, void * handle) -{ - u8 * s; - uword *bitmap = 0; - u8 first = 1; - int i; - u64 mask = clib_net_to_host_u64 (mp->worker_mask); - - s = format (0, "SCRIPT: snat_set_workers "); - bitmap = clib_bitmap_set_multiple (bitmap, 0, mask, BITS (mask)); - clib_bitmap_foreach (i, bitmap, - ({ - if (first) - s = format (s, "%d", i); - else - s = format (s, ",%d", i); - first = 0; - })); - clib_bitmap_free (bitmap); - FINISH; -} - -static void -send_snat_worker_details -(u32 worker_index, unix_shared_memory_queue_t * q, u32 context) -{ - vl_api_snat_worker_details_t *rmp; - snat_main_t * sm = &snat_main; - vlib_worker_thread_t *w = - vlib_worker_threads + worker_index + sm->first_worker_index; - - rmp = vl_msg_api_alloc (sizeof (*rmp)); - memset (rmp, 0, sizeof (*rmp)); - rmp->_vl_msg_id = ntohs (VL_API_SNAT_WORKER_DETAILS+sm->msg_id_base); - rmp->context = context; - rmp->worker_index = htonl (worker_index); - rmp->lcore_id = htonl (w->lcore_id); - strncpy ((char *) rmp->name, (char *) w->name, ARRAY_LEN (rmp->name) - 1); - - vl_msg_api_send_shmem (q, (u8 *) & rmp); -} - -static void -vl_api_snat_worker_dump_t_handler -(vl_api_snat_worker_dump_t * mp) -{ - unix_shared_memory_queue_t *q; - snat_main_t * sm = &snat_main; - u32 * worker_index; - - q = vl_api_client_index_to_input_queue (mp->client_index); - if (q == 0) - return; - - vec_foreach (worker_index, sm->workers) - { - send_snat_worker_details(*worker_index, q, mp->context); - } -} - -static void *vl_api_snat_worker_dump_t_print -(vl_api_snat_worker_dump_t *mp, void * handle) -{ - u8 *s; - - s = format (0, "SCRIPT: snat_worker_dump "); - - FINISH; -} - -/* List of message types that this plugin understands */ -#define foreach_snat_plugin_api_msg \ -_(SNAT_ADD_ADDRESS_RANGE, snat_add_address_range) \ -_(SNAT_INTERFACE_ADD_DEL_FEATURE, snat_interface_add_del_feature) \ -_(SNAT_ADD_STATIC_MAPPING, snat_add_static_mapping) \ -_(SNAT_CONTROL_PING, snat_control_ping) \ -_(SNAT_STATIC_MAPPING_DUMP, snat_static_mapping_dump) \ -_(SNAT_SHOW_CONFIG, snat_show_config) \ -_(SNAT_ADDRESS_DUMP, snat_address_dump) \ -_(SNAT_INTERFACE_DUMP, snat_interface_dump) \ -_(SNAT_SET_WORKERS, snat_set_workers) \ -_(SNAT_WORKER_DUMP, snat_worker_dump) - -/* Set up the API message handling tables */ -static clib_error_t * -snat_plugin_api_hookup (vlib_main_t *vm) -{ - snat_main_t * sm __attribute__ ((unused)) = &snat_main; -#define _(N,n) \ - vl_msg_api_set_handlers((VL_API_##N + sm->msg_id_base), \ - #n, \ - vl_api_##n##_t_handler, \ - vl_noop_handler, \ - vl_api_##n##_t_endian, \ - vl_api_##n##_t_print, \ - sizeof(vl_api_##n##_t), 1); - foreach_snat_plugin_api_msg; -#undef _ - - return 0; -} - -#define vl_msg_name_crc_list -#include -#undef vl_msg_name_crc_list - -static void -setup_message_id_table (snat_main_t * sm, api_main_t * am) -{ -#define _(id,n,crc) \ - vl_msg_api_add_msg_name_crc (am, #n "_" #crc, id + sm->msg_id_base); - foreach_vl_msg_name_crc_snat; -#undef _ -} - -static void plugin_custom_dump_configure (snat_main_t * sm) -{ -#define _(n,f) sm->api_main->msg_print_handlers \ - [VL_API_##n + sm->msg_id_base] \ - = (void *) vl_api_##f##_t_print; - foreach_snat_plugin_api_msg; -#undef _ -} - -static clib_error_t * snat_init (vlib_main_t * vm) -{ - snat_main_t * sm = &snat_main; - clib_error_t * error = 0; - ip4_main_t * im = &ip4_main; - ip_lookup_main_t * lm = &im->lookup_main; - u8 * name; - uword *p; - vlib_thread_registration_t *tr; - vlib_thread_main_t *tm = vlib_get_thread_main (); - uword *bitmap = 0; - u32 i; - - name = format (0, "snat_%08x%c", api_version, 0); - - /* Ask for a correctly-sized block of API message decode slots */ - sm->msg_id_base = vl_msg_api_get_msg_ids - ((char *) name, VL_MSG_FIRST_AVAILABLE); - - sm->vlib_main = vm; - sm->vnet_main = vnet_get_main(); - sm->ip4_main = im; - sm->ip4_lookup_main = lm; - sm->api_main = &api_main; - sm->first_worker_index = 0; - sm->next_worker = 0; - sm->num_workers = 0; - sm->workers = 0; - sm->fq_in2out_index = ~0; - sm->fq_out2in_index = ~0; - - p = hash_get_mem (tm->thread_registrations_by_name, "workers"); - if (p) - { - tr = (vlib_thread_registration_t *) p[0]; - if (tr) - { - sm->num_workers = tr->count; - sm->first_worker_index = tr->first_index; - } - } - - /* Use all available workers by default */ - if (sm->num_workers > 1) - { - for (i=0; i < sm->num_workers; i++) - bitmap = clib_bitmap_set (bitmap, i, 1); - snat_set_workers(bitmap); - clib_bitmap_free (bitmap); - } - - error = snat_plugin_api_hookup (vm); - - /* Add our API messages to the global name_crc hash table */ - setup_message_id_table (sm, &api_main); - - plugin_custom_dump_configure (sm); - vec_free(name); - - return error; -} - -VLIB_INIT_FUNCTION (snat_init); - -void snat_free_outside_address_and_port (snat_main_t * sm, - snat_session_key_t * k, - u32 address_index) -{ - snat_address_t *a; - u16 port_host_byte_order = clib_net_to_host_u16 (k->port); - - ASSERT (address_index < vec_len (sm->addresses)); - - a = sm->addresses + address_index; - - ASSERT (clib_bitmap_get_no_check (a->busy_port_bitmap, - port_host_byte_order) == 1); - - clib_bitmap_set_no_check (a->busy_port_bitmap, port_host_byte_order, 0); - a->busy_ports--; -} - -/** - * @brief Match SNAT static mapping. - * - * @param sm SNAT main. - * @param match Address and port to match. - * @param mapping External or local address and port of the matched mapping. - * @param by_external If 0 match by local address otherwise match by external - * address. - * - * @returns 0 if match found otherwise 1. - */ -int snat_static_mapping_match (snat_main_t * sm, - snat_session_key_t match, - snat_session_key_t * mapping, - u8 by_external) -{ - clib_bihash_kv_8_8_t kv, value; - snat_static_mapping_t *m; - snat_static_mapping_key_t m_key; - clib_bihash_8_8_t *mapping_hash = &sm->static_mapping_by_local; - - if (by_external) - mapping_hash = &sm->static_mapping_by_external; - - m_key.addr = match.addr; - m_key.port = clib_net_to_host_u16 (match.port); - m_key.fib_index = match.fib_index; - - kv.key = m_key.as_u64; - - if (clib_bihash_search_8_8 (mapping_hash, &kv, &value)) - { - /* Try address only mapping */ - m_key.port = 0; - kv.key = m_key.as_u64; - if (clib_bihash_search_8_8 (mapping_hash, &kv, &value)) - return 1; - } - - m = pool_elt_at_index (sm->static_mappings, value.value); - - if (by_external) - { - mapping->addr = m->local_addr; - /* Address only mapping doesn't change port */ - mapping->port = m->addr_only ? match.port - : clib_host_to_net_u16 (m->local_port); - mapping->fib_index = m->fib_index; - } - else - { - mapping->addr = m->external_addr; - /* Address only mapping doesn't change port */ - mapping->port = m->addr_only ? match.port - : clib_host_to_net_u16 (m->external_port); - mapping->fib_index = sm->outside_fib_index; - } - - return 0; -} - -int snat_alloc_outside_address_and_port (snat_main_t * sm, - snat_session_key_t * k, - u32 * address_indexp) -{ - int i; - snat_address_t *a; - u32 portnum; - - for (i = 0; i < vec_len (sm->addresses); i++) - { - if (sm->addresses[i].busy_ports < (65535-1024)) - { - a = sm->addresses + i; - - while (1) - { - portnum = random_u32 (&sm->random_seed); - portnum &= 0xFFFF; - if (portnum < 1024) - continue; - if (clib_bitmap_get_no_check (a->busy_port_bitmap, portnum)) - continue; - clib_bitmap_set_no_check (a->busy_port_bitmap, portnum, 1); - a->busy_ports++; - /* Caller sets protocol and fib index */ - k->addr = a->addr; - k->port = clib_host_to_net_u16(portnum); - *address_indexp = i; - return 0; - } - } - } - /* Totally out of translations to use... */ - return 1; -} - - -static clib_error_t * -add_address_command_fn (vlib_main_t * vm, - unformat_input_t * input, - vlib_cli_command_t * cmd) -{ - unformat_input_t _line_input, *line_input = &_line_input; - snat_main_t * sm = &snat_main; - ip4_address_t start_addr, end_addr, this_addr; - u32 start_host_order, end_host_order; - int i, count; - int is_add = 1; - int rv = 0; - - /* Get a line of input. */ - if (!unformat_user (input, unformat_line_input, line_input)) - return 0; - - while (unformat_check_input (line_input) != UNFORMAT_END_OF_INPUT) - { - if (unformat (line_input, "%U - %U", - unformat_ip4_address, &start_addr, - unformat_ip4_address, &end_addr)) - ; - else if (unformat (line_input, "%U", unformat_ip4_address, &start_addr)) - end_addr = start_addr; - else if (unformat (line_input, "del")) - is_add = 0; - else - return clib_error_return (0, "unknown input '%U'", - format_unformat_error, input); - } - unformat_free (line_input); - - if (sm->static_mapping_only) - return clib_error_return (0, "static mapping only mode"); - - start_host_order = clib_host_to_net_u32 (start_addr.as_u32); - end_host_order = clib_host_to_net_u32 (end_addr.as_u32); - - if (end_host_order < start_host_order) - return clib_error_return (0, "end address less than start address"); - - count = (end_host_order - start_host_order) + 1; - - if (count > 1024) - clib_warning ("%U - %U, %d addresses...", - format_ip4_address, &start_addr, - format_ip4_address, &end_addr, - count); - - this_addr = start_addr; - - for (i = 0; i < count; i++) - { - if (is_add) - snat_add_address (sm, &this_addr); - else - rv = snat_del_address (sm, this_addr); - - switch (rv) - { - case VNET_API_ERROR_NO_SUCH_ENTRY: - return clib_error_return (0, "S-NAT address not exist."); - break; - case VNET_API_ERROR_UNSPECIFIED: - return clib_error_return (0, "S-NAT address used in static mapping."); - break; - default: - break; - } - - increment_v4_address (&this_addr); - } - - return 0; -} - -VLIB_CLI_COMMAND (add_address_command, static) = { - .path = "snat add address", - .short_help = "snat add addresses [- ] [del]", - .function = add_address_command_fn, -}; - -static clib_error_t * -snat_feature_command_fn (vlib_main_t * vm, - unformat_input_t * input, - vlib_cli_command_t * cmd) -{ - unformat_input_t _line_input, *line_input = &_line_input; - vnet_main_t * vnm = vnet_get_main(); - clib_error_t * error = 0; - u32 sw_if_index; - u32 * inside_sw_if_indices = 0; - u32 * outside_sw_if_indices = 0; - int is_del = 0; - int i; - - sw_if_index = ~0; - - /* Get a line of input. */ - if (!unformat_user (input, unformat_line_input, line_input)) - return 0; - - while (unformat_check_input (line_input) != UNFORMAT_END_OF_INPUT) - { - if (unformat (line_input, "in %U", unformat_vnet_sw_interface, - vnm, &sw_if_index)) - vec_add1 (inside_sw_if_indices, sw_if_index); - else if (unformat (line_input, "out %U", unformat_vnet_sw_interface, - vnm, &sw_if_index)) - vec_add1 (outside_sw_if_indices, sw_if_index); - else if (unformat (line_input, "del")) - is_del = 1; - else - return clib_error_return (0, "unknown input '%U'", - format_unformat_error, input); - } - unformat_free (line_input); - - if (vec_len (inside_sw_if_indices)) - { - for (i = 0; i < vec_len(inside_sw_if_indices); i++) - { - sw_if_index = inside_sw_if_indices[i]; - snat_interface_add_del (sw_if_index, 1, is_del); - } - } - - if (vec_len (outside_sw_if_indices)) - { - for (i = 0; i < vec_len(outside_sw_if_indices); i++) - { - sw_if_index = outside_sw_if_indices[i]; - snat_interface_add_del (sw_if_index, 0, is_del); - } - } - - vec_free (inside_sw_if_indices); - vec_free (outside_sw_if_indices); - - return error; -} - -VLIB_CLI_COMMAND (set_interface_snat_command, static) = { - .path = "set interface snat", - .function = snat_feature_command_fn, - .short_help = "set interface snat in out [del]", -}; - -static clib_error_t * -add_static_mapping_command_fn (vlib_main_t * vm, - unformat_input_t * input, - vlib_cli_command_t * cmd) -{ - unformat_input_t _line_input, *line_input = &_line_input; - clib_error_t * error = 0; - ip4_address_t l_addr, e_addr; - u32 l_port = 0, e_port = 0, vrf_id = ~0; - int is_add = 1; - int addr_only = 1; - int rv; - - /* Get a line of input. */ - if (!unformat_user (input, unformat_line_input, line_input)) - return 0; - - while (unformat_check_input (line_input) != UNFORMAT_END_OF_INPUT) - { - if (unformat (line_input, "local %U %u", unformat_ip4_address, &l_addr, - &l_port)) - addr_only = 0; - else if (unformat (line_input, "local %U", unformat_ip4_address, &l_addr)) - ; - else if (unformat (line_input, "external %U %u", unformat_ip4_address, - &e_addr, &e_port)) - addr_only = 0; - else if (unformat (line_input, "external %U", unformat_ip4_address, - &e_addr)) - ; - else if (unformat (line_input, "vrf %u", &vrf_id)) - ; - else if (unformat (line_input, "del")) - is_add = 0; - else - return clib_error_return (0, "unknown input: '%U'", - format_unformat_error, line_input); - } - unformat_free (line_input); - - rv = snat_add_static_mapping(l_addr, e_addr, (u16) l_port, (u16) e_port, - vrf_id, addr_only, is_add); - - switch (rv) - { - case VNET_API_ERROR_INVALID_VALUE: - return clib_error_return (0, "External port already in use."); - break; - case VNET_API_ERROR_NO_SUCH_ENTRY: - if (is_add) - return clib_error_return (0, "External addres must be allocated."); - else - return clib_error_return (0, "Mapping not exist."); - break; - case VNET_API_ERROR_NO_SUCH_FIB: - return clib_error_return (0, "No such VRF id."); - case VNET_API_ERROR_VALUE_EXIST: - return clib_error_return (0, "Mapping already exist."); - default: - break; - } - - return error; -} - -/*? - * @cliexpar - * @cliexstart{snat add static mapping} - * Static mapping allows hosts on the external network to initiate connection - * to to the local network host. - * To create static mapping between local host address 10.0.0.3 port 6303 and - * external address 4.4.4.4 port 3606 use: - * vpp# snat add static mapping local 10.0.0.3 6303 external 4.4.4.4 3606 - * If not runnig "static mapping only" S-NAT plugin mode use before: - * vpp# snat add address 4.4.4.4 - * To create static mapping between local and external address use: - * vpp# snat add static mapping local 10.0.0.3 external 4.4.4.4 - * @cliexend -?*/ -VLIB_CLI_COMMAND (add_static_mapping_command, static) = { - .path = "snat add static mapping", - .function = add_static_mapping_command_fn, - .short_help = - "snat add static mapping local [] external [] [vrf ] [del]", -}; - -static clib_error_t * -set_workers_command_fn (vlib_main_t * vm, - unformat_input_t * input, - vlib_cli_command_t * cmd) -{ - unformat_input_t _line_input, *line_input = &_line_input; - uword *bitmap = 0; - int rv = 0; - - /* Get a line of input. */ - if (!unformat_user (input, unformat_line_input, line_input)) - return 0; - - while (unformat_check_input (line_input) != UNFORMAT_END_OF_INPUT) - { - if (unformat (line_input, "%U", unformat_bitmap_list, &bitmap)) - ; - else - return clib_error_return (0, "unknown input '%U'", - format_unformat_error, input); - } - unformat_free (line_input); - - if (bitmap == 0) - return clib_error_return (0, "List of workers must be specified."); - - rv = snat_set_workers(bitmap); - - clib_bitmap_free (bitmap); - - switch (rv) - { - case VNET_API_ERROR_INVALID_WORKER: - return clib_error_return (0, "Invalid worker(s)."); - break; - case VNET_API_ERROR_FEATURE_DISABLED: - return clib_error_return (0, - "Supported only if 2 or more workes available."); - break; - default: - break; - } - - return 0; -} - -/*? - * @cliexpar - * @cliexstart{set snat workers} - * Set SNAT workers if 2 or more workers available, use: - * vpp# set snat workers 0-2,5 - * @cliexend -?*/ -VLIB_CLI_COMMAND (set_workers_command, static) = { - .path = "set snat workers", - .function = set_workers_command_fn, - .short_help = - "set snat workers ", -}; - -static clib_error_t * -snat_config (vlib_main_t * vm, unformat_input_t * input) -{ - snat_main_t * sm = &snat_main; - u32 translation_buckets = 1024; - u32 translation_memory_size = 128<<20; - u32 user_buckets = 128; - u32 user_memory_size = 64<<20; - u32 max_translations_per_user = 100; - u32 outside_vrf_id = 0; - u32 inside_vrf_id = 0; - u32 static_mapping_buckets = 1024; - u32 static_mapping_memory_size = 64<<20; - u8 static_mapping_only = 0; - u8 static_mapping_connection_tracking = 0; - vlib_thread_main_t *tm = vlib_get_thread_main (); - - while (unformat_check_input (input) != UNFORMAT_END_OF_INPUT) - { - if (unformat (input, "translation hash buckets %d", &translation_buckets)) - ; - else if (unformat (input, "translation hash memory %d", - &translation_memory_size)); - else if (unformat (input, "user hash buckets %d", &user_buckets)) - ; - else if (unformat (input, "user hash memory %d", - &user_memory_size)) - ; - else if (unformat (input, "max translations per user %d", - &max_translations_per_user)) - ; - else if (unformat (input, "outside VRF id %d", - &outside_vrf_id)) - ; - else if (unformat (input, "inside VRF id %d", - &inside_vrf_id)) - ; - else if (unformat (input, "static mapping only")) - { - static_mapping_only = 1; - if (unformat (input, "connection tracking")) - static_mapping_connection_tracking = 1; - } - else - return clib_error_return (0, "unknown input '%U'", - format_unformat_error, input); - } - - /* for show commands, etc. */ - sm->translation_buckets = translation_buckets; - sm->translation_memory_size = translation_memory_size; - sm->user_buckets = user_buckets; - sm->user_memory_size = user_memory_size; - sm->max_translations_per_user = max_translations_per_user; - sm->outside_vrf_id = outside_vrf_id; - sm->outside_fib_index = ~0; - sm->inside_vrf_id = inside_vrf_id; - sm->inside_fib_index = ~0; - sm->static_mapping_only = static_mapping_only; - sm->static_mapping_connection_tracking = static_mapping_connection_tracking; - - if (!static_mapping_only || - (static_mapping_only && static_mapping_connection_tracking)) - { - clib_bihash_init_8_8 (&sm->worker_by_in, "worker-by-in", user_buckets, - user_memory_size); - - clib_bihash_init_8_8 (&sm->worker_by_out, "worker-by-out", user_buckets, - user_memory_size); - - vec_validate (sm->per_thread_data, tm->n_vlib_mains - 1); - - clib_bihash_init_8_8 (&sm->in2out, "in2out", translation_buckets, - translation_memory_size); - - clib_bihash_init_8_8 (&sm->out2in, "out2in", translation_buckets, - translation_memory_size); - - clib_bihash_init_8_8 (&sm->user_hash, "users", user_buckets, - user_memory_size); - } - clib_bihash_init_8_8 (&sm->static_mapping_by_local, - "static_mapping_by_local", static_mapping_buckets, - static_mapping_memory_size); - - clib_bihash_init_8_8 (&sm->static_mapping_by_external, - "static_mapping_by_external", static_mapping_buckets, - static_mapping_memory_size); - return 0; -} - -VLIB_CONFIG_FUNCTION (snat_config, "snat"); - -u8 * format_snat_key (u8 * s, va_list * args) -{ - snat_session_key_t * key = va_arg (*args, snat_session_key_t *); - char * protocol_string = "unknown"; - static char *protocol_strings[] = { - "UDP", - "TCP", - "ICMP", - }; - - if (key->protocol < ARRAY_LEN(protocol_strings)) - protocol_string = protocol_strings[key->protocol]; - - s = format (s, "%U proto %s port %d fib %d", - format_ip4_address, &key->addr, protocol_string, - clib_net_to_host_u16 (key->port), key->fib_index); - return s; -} - -u8 * format_snat_session (u8 * s, va_list * args) -{ - snat_main_t * sm __attribute__((unused)) = va_arg (*args, snat_main_t *); - snat_session_t * sess = va_arg (*args, snat_session_t *); - - s = format (s, " i2o %U\n", format_snat_key, &sess->in2out); - s = format (s, " o2i %U\n", format_snat_key, &sess->out2in); - s = format (s, " last heard %.2f\n", sess->last_heard); - s = format (s, " total pkts %d, total bytes %lld\n", - sess->total_pkts, sess->total_bytes); - if (snat_is_session_static (sess)) - s = format (s, " static translation\n"); - else - s = format (s, " dynamic translation\n"); - - return s; -} - -u8 * format_snat_user (u8 * s, va_list * args) -{ - snat_main_per_thread_data_t * sm = va_arg (*args, snat_main_per_thread_data_t *); - snat_user_t * u = va_arg (*args, snat_user_t *); - int verbose = va_arg (*args, int); - dlist_elt_t * head, * elt; - u32 elt_index, head_index; - u32 session_index; - snat_session_t * sess; - - s = format (s, "%U: %d dynamic translations, %d static translations\n", - format_ip4_address, &u->addr, u->nsessions, u->nstaticsessions); - - if (verbose == 0) - return s; - - if (u->nsessions || u->nstaticsessions) - { - head_index = u->sessions_per_user_list_head_index; - head = pool_elt_at_index (sm->list_pool, head_index); - - elt_index = head->next; - elt = pool_elt_at_index (sm->list_pool, elt_index); - session_index = elt->value; - - while (session_index != ~0) - { - sess = pool_elt_at_index (sm->sessions, session_index); - - s = format (s, " %U\n", format_snat_session, sm, sess); - - elt_index = elt->next; - elt = pool_elt_at_index (sm->list_pool, elt_index); - session_index = elt->value; - } - } - - return s; -} - -u8 * format_snat_static_mapping (u8 * s, va_list * args) -{ - snat_static_mapping_t *m = va_arg (*args, snat_static_mapping_t *); - - if (m->addr_only) - s = format (s, "local %U external %U vrf %d", - format_ip4_address, &m->local_addr, - format_ip4_address, &m->external_addr, - m->vrf_id); - else - s = format (s, "local %U:%d external %U:%d vrf %d", - format_ip4_address, &m->local_addr, m->local_port, - format_ip4_address, &m->external_addr, m->external_port, - m->vrf_id); - - return s; -} - -static clib_error_t * -show_snat_command_fn (vlib_main_t * vm, - unformat_input_t * input, - vlib_cli_command_t * cmd) -{ - int verbose = 0; - snat_main_t * sm = &snat_main; - snat_user_t * u; - snat_static_mapping_t *m; - snat_interface_t *i; - snat_address_t * ap; - vnet_main_t *vnm = vnet_get_main(); - snat_main_per_thread_data_t *tsm; - u32 users_num = 0, sessions_num = 0, *worker; - uword j = 0; - - if (unformat (input, "detail")) - verbose = 1; - else if (unformat (input, "verbose")) - verbose = 2; - - if (sm->static_mapping_only) - { - if (sm->static_mapping_connection_tracking) - vlib_cli_output (vm, "SNAT mode: static mapping only connection " - "tracking"); - else - vlib_cli_output (vm, "SNAT mode: static mapping only"); - } - else - { - vlib_cli_output (vm, "SNAT mode: dynamic translations enabled"); - } - - if (verbose > 0) - { - pool_foreach (i, sm->interfaces, - ({ - vlib_cli_output (vm, "%U %s", format_vnet_sw_interface_name, vnm, - vnet_get_sw_interface (vnm, i->sw_if_index), - i->is_inside ? "in" : "out"); - })); - - vec_foreach (ap, sm->addresses) - { - u8 * s = format (0, ""); - vlib_cli_output (vm, "%U", format_ip4_address, &ap->addr); - clib_bitmap_foreach (j, ap->busy_port_bitmap, - ({ - s = format (s, " %d", j); - })); - vlib_cli_output (vm, " %d busy ports:%v", ap->busy_ports, s); - } - } - - if (sm->num_workers > 1) - { - vlib_cli_output (vm, "%d workers", vec_len (sm->workers)); - if (verbose > 0) - { - vec_foreach (worker, sm->workers) - { - vlib_worker_thread_t *w = - vlib_worker_threads + *worker + sm->first_worker_index; - vlib_cli_output (vm, " %v", w->name); - } - } - } - - if (sm->static_mapping_only && !(sm->static_mapping_connection_tracking)) - { - vlib_cli_output (vm, "%d static mappings", - pool_elts (sm->static_mappings)); - - if (verbose > 0) - { - pool_foreach (m, sm->static_mappings, - ({ - vlib_cli_output (vm, "%U", format_snat_static_mapping, m); - })); - } - } - else - { - vec_foreach (tsm, sm->per_thread_data) - { - users_num += pool_elts (tsm->users); - sessions_num += pool_elts (tsm->sessions); - } - - vlib_cli_output (vm, "%d users, %d outside addresses, %d active sessions," - " %d static mappings", - users_num, - vec_len (sm->addresses), - sessions_num, - pool_elts (sm->static_mappings)); - - if (verbose > 0) - { - vlib_cli_output (vm, "%U", format_bihash_8_8, &sm->in2out, - verbose - 1); - vlib_cli_output (vm, "%U", format_bihash_8_8, &sm->out2in, - verbose - 1); - vlib_cli_output (vm, "%U", format_bihash_8_8, &sm->worker_by_in, - verbose - 1); - vlib_cli_output (vm, "%U", format_bihash_8_8, &sm->worker_by_out, - verbose - 1); - vec_foreach_index (j, sm->per_thread_data) - { - tsm = vec_elt_at_index (sm->per_thread_data, j); - - if (pool_elts (tsm->users) == 0) - continue; - - vlib_worker_thread_t *w = vlib_worker_threads + j; - vlib_cli_output (vm, "Thread %d (%v at lcore %u):", j, w->name, - w->lcore_id); - vlib_cli_output (vm, " %d list pool elements", - pool_elts (tsm->list_pool)); - - pool_foreach (u, tsm->users, - ({ - vlib_cli_output (vm, " %U", format_snat_user, tsm, u, - verbose - 1); - })); - } - - if (pool_elts (sm->static_mappings)) - { - vlib_cli_output (vm, "static mappings:"); - pool_foreach (m, sm->static_mappings, - ({ - vlib_cli_output (vm, "%U", format_snat_static_mapping, m); - })); - } - } - } - - return 0; -} - -VLIB_CLI_COMMAND (show_snat_command, static) = { - .path = "show snat", - .short_help = "show snat", - .function = show_snat_command_fn, -}; diff --git a/plugins/snat-plugin/snat/snat.h b/plugins/snat-plugin/snat/snat.h deleted file mode 100644 index cb31dc51..00000000 --- a/plugins/snat-plugin/snat/snat.h +++ /dev/null @@ -1,259 +0,0 @@ - -/* - * snat.h - simple nat definitions - * - * Copyright (c) 2016 Cisco and/or its affiliates. - * Licensed under the Apache License, Version 2.0 (the "License"); - * you may not use this file except in compliance with the License. - * You may obtain a copy of the License at: - * - * http://www.apache.org/licenses/LICENSE-2.0 - * - * Unless required by applicable law or agreed to in writing, software - * distributed under the License is distributed on an "AS IS" BASIS, - * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. - * See the License for the specific language governing permissions and - * limitations under the License. - */ -#ifndef __included_snat_h__ -#define __included_snat_h__ - -#include -#include -#include -#include -#include -#include -#include -#include -#include - -/* Key */ -typedef struct { - union - { - struct - { - ip4_address_t addr; - u16 port; - u16 protocol:3, - fib_index:13; - }; - u64 as_u64; - }; -} snat_session_key_t; - -typedef struct { - union - { - struct - { - ip4_address_t addr; - u32 fib_index; - }; - u64 as_u64; - }; -} snat_user_key_t; - -typedef struct { - union - { - struct - { - ip4_address_t addr; - u16 port; - u16 fib_index; - }; - u64 as_u64; - }; -} snat_static_mapping_key_t; - - -typedef enum { - SNAT_PROTOCOL_UDP = 0, - SNAT_PROTOCOL_TCP, - SNAT_PROTOCOL_ICMP, -} snat_protocol_t; - - -#define SNAT_SESSION_FLAG_STATIC_MAPPING 1 - -typedef CLIB_PACKED(struct { - snat_session_key_t out2in; /* 0-15 */ - - snat_session_key_t in2out; /* 16-31 */ - - u32 flags; /* 32-35 */ - - /* per-user translations */ - u32 per_user_index; /* 36-39 */ - - u32 per_user_list_head_index; /* 40-43 */ - - /* Last heard timer */ - f64 last_heard; /* 44-51 */ - - u64 total_bytes; /* 52-59 */ - - u32 total_pkts; /* 60-63 */ - - /* Outside address */ - u32 outside_address_index; /* 64-67 */ - -}) snat_session_t; - - -typedef struct { - ip4_address_t addr; - u32 sessions_per_user_list_head_index; - u32 nsessions; - u32 nstaticsessions; -} snat_user_t; - -typedef struct { - ip4_address_t addr; - u32 busy_ports; - uword * busy_port_bitmap; -} snat_address_t; - -typedef struct { - ip4_address_t local_addr; - ip4_address_t external_addr; - u16 local_port; - u16 external_port; - u8 addr_only; - u32 vrf_id; - u32 fib_index; -} snat_static_mapping_t; - -typedef struct { - u32 sw_if_index; - u8 is_inside; -} snat_interface_t; - -typedef struct { - /* User pool */ - snat_user_t * users; - - /* Session pool */ - snat_session_t * sessions; - - /* Pool of doubly-linked list elements */ - dlist_elt_t * list_pool; -} snat_main_per_thread_data_t; - -typedef struct { - /* Main lookup tables */ - clib_bihash_8_8_t out2in; - clib_bihash_8_8_t in2out; - - /* Find-a-user => src address lookup */ - clib_bihash_8_8_t user_hash; - - /* Non-translated packets worker lookup => src address + VRF */ - clib_bihash_8_8_t worker_by_in; - - /* Translated packets worker lookup => IP address + port number */ - clib_bihash_8_8_t worker_by_out; - - u32 num_workers; - u32 first_worker_index; - u32 next_worker; - u32 * workers; - - /* Per thread data */ - snat_main_per_thread_data_t * per_thread_data; - - /* Find a static mapping by local */ - clib_bihash_8_8_t static_mapping_by_local; - - /* Find a static mapping by external */ - clib_bihash_8_8_t static_mapping_by_external; - - /* Static mapping pool */ - snat_static_mapping_t * static_mappings; - - /* Interface pool */ - snat_interface_t * interfaces; - - /* Vector of outside addresses */ - snat_address_t * addresses; - - /* Randomize port allocation order */ - u32 random_seed; - - /* Worker handoff index */ - u32 fq_in2out_index; - u32 fq_out2in_index; - - /* Config parameters */ - u8 static_mapping_only; - u8 static_mapping_connection_tracking; - u32 translation_buckets; - u32 translation_memory_size; - u32 user_buckets; - u32 user_memory_size; - u32 max_translations_per_user; - u32 outside_vrf_id; - u32 outside_fib_index; - u32 inside_vrf_id; - u32 inside_fib_index; - - /* API message ID base */ - u16 msg_id_base; - - /* convenience */ - vlib_main_t * vlib_main; - vnet_main_t * vnet_main; - ip4_main_t * ip4_main; - ip_lookup_main_t * ip4_lookup_main; - ethernet_main_t * ethernet_main; - api_main_t * api_main; -} snat_main_t; - -extern snat_main_t snat_main; -extern vlib_node_registration_t snat_in2out_node; -extern vlib_node_registration_t snat_out2in_node; -extern vlib_node_registration_t snat_in2out_fast_node; -extern vlib_node_registration_t snat_out2in_fast_node; -extern vlib_node_registration_t snat_in2out_worker_handoff_node; -extern vlib_node_registration_t snat_out2in_worker_handoff_node; - -void snat_free_outside_address_and_port (snat_main_t * sm, - snat_session_key_t * k, - u32 address_index); - -int snat_alloc_outside_address_and_port (snat_main_t * sm, - snat_session_key_t * k, - u32 * address_indexp); - -int snat_static_mapping_match (snat_main_t * sm, - snat_session_key_t match, - snat_session_key_t * mapping, - u8 by_external); - -format_function_t format_snat_user; - -typedef struct { - u32 cached_sw_if_index; - u32 cached_ip4_address; -} snat_runtime_t; - -/** \brief Check if SNAT session is created from static mapping. - @param s SNAT session - @return 1 if SNAT session is created from static mapping otherwise 0 -*/ -#define snat_is_session_static(s) s->flags & SNAT_SESSION_FLAG_STATIC_MAPPING - -/* - * Why is this here? Because we don't need to touch this layer to - * simply reply to an icmp. We need to change id to a unique - * value to NAT an echo request/reply. - */ - -typedef struct { - u16 identifier; - u16 sequence; -} icmp_echo_header_t; - -#endif /* __included_snat_h__ */ diff --git a/plugins/snat-plugin/snat/snat_all_api_h.h b/plugins/snat-plugin/snat/snat_all_api_h.h deleted file mode 100644 index 49017700..00000000 --- a/plugins/snat-plugin/snat/snat_all_api_h.h +++ /dev/null @@ -1,19 +0,0 @@ - -/* - * snat_all_api_h.h - skeleton vpp engine plug-in api #include file - * - * Copyright (c) - * Licensed under the Apache License, Version 2.0 (the "License"); - * you may not use this file except in compliance with the License. - * You may obtain a copy of the License at: - * - * http://www.apache.org/licenses/LICENSE-2.0 - * - * Unless required by applicable law or agreed to in writing, software - * distributed under the License is distributed on an "AS IS" BASIS, - * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. - * See the License for the specific language governing permissions and - * limitations under the License. - */ -/* Include the generated file, see BUILT_SOURCES in Makefile.am */ -#include diff --git a/plugins/snat-plugin/snat/snat_msg_enum.h b/plugins/snat-plugin/snat/snat_msg_enum.h deleted file mode 100644 index 2c76fd51..00000000 --- a/plugins/snat-plugin/snat/snat_msg_enum.h +++ /dev/null @@ -1,31 +0,0 @@ - -/* - * snat_msg_enum.h - skeleton vpp engine plug-in message enumeration - * - * Copyright (c) - * Licensed under the Apache License, Version 2.0 (the "License"); - * you may not use this file except in compliance with the License. - * You may obtain a copy of the License at: - * - * http://www.apache.org/licenses/LICENSE-2.0 - * - * Unless required by applicable law or agreed to in writing, software - * distributed under the License is distributed on an "AS IS" BASIS, - * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. - * See the License for the specific language governing permissions and - * limitations under the License. - */ -#ifndef included_snat_msg_enum_h -#define included_snat_msg_enum_h - -#include - -#define vl_msg_id(n,h) n, -typedef enum { -#include - /* We'll want to know how many messages IDs we need... */ - VL_MSG_FIRST_AVAILABLE, -} vl_msg_id_t; -#undef vl_msg_id - -#endif /* included_snat_msg_enum_h */ diff --git a/plugins/snat-plugin/snat/snat_test.c b/plugins/snat-plugin/snat/snat_test.c deleted file mode 100644 index 2a003ba6..00000000 --- a/plugins/snat-plugin/snat/snat_test.c +++ /dev/null @@ -1,602 +0,0 @@ - -/* - * snat.c - skeleton vpp-api-test plug-in - * - * Copyright (c) - * Licensed under the Apache License, Version 2.0 (the "License"); - * you may not use this file except in compliance with the License. - * You may obtain a copy of the License at: - * - * http://www.apache.org/licenses/LICENSE-2.0 - * - * Unless required by applicable law or agreed to in writing, software - * distributed under the License is distributed on an "AS IS" BASIS, - * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. - * See the License for the specific language governing permissions and - * limitations under the License. - */ -#include -#include -#include -#include -#include -#include - -uword unformat_sw_if_index (unformat_input_t * input, va_list * args); - -/* Declare message IDs */ -#include - -/* define message structures */ -#define vl_typedefs -#include -#undef vl_typedefs - -/* declare message handlers for each api */ - -#define vl_endianfun /* define message structures */ -#include -#undef vl_endianfun - -/* instantiate all the print functions we know about */ -#define vl_print(handle, ...) -#define vl_printfun -#include -#undef vl_printfun - -/* Get the API version number. */ -#define vl_api_version(n,v) static u32 api_version=(v); -#include -#undef vl_api_version - -typedef struct { - /* API message ID base */ - u16 msg_id_base; - vat_main_t *vat_main; -} snat_test_main_t; - -snat_test_main_t snat_test_main; - -#define foreach_standard_reply_retval_handler \ -_(snat_add_address_range_reply) \ -_(snat_interface_add_del_feature_reply) \ -_(snat_add_static_mapping_reply) \ -_(snat_set_workers_reply) - -#define _(n) \ - static void vl_api_##n##_t_handler \ - (vl_api_##n##_t * mp) \ - { \ - vat_main_t * vam = snat_test_main.vat_main; \ - i32 retval = ntohl(mp->retval); \ - if (vam->async_mode) { \ - vam->async_errors += (retval < 0); \ - } else { \ - vam->retval = retval; \ - vam->result_ready = 1; \ - } \ - } -foreach_standard_reply_retval_handler; -#undef _ - -/* - * Table of message reply handlers, must include boilerplate handlers - * we just generated - */ -#define foreach_vpe_api_reply_msg \ -_(SNAT_ADD_ADDRESS_RANGE_REPLY, snat_add_address_range_reply) \ -_(SNAT_INTERFACE_ADD_DEL_FEATURE_REPLY, \ - snat_interface_add_del_feature_reply) \ -_(SNAT_ADD_STATIC_MAPPING_REPLY, snat_add_static_mapping_reply) \ -_(SNAT_CONTROL_PING_REPLY, snat_control_ping_reply) \ -_(SNAT_STATIC_MAPPING_DETAILS, snat_static_mapping_details) \ -_(SNAT_SHOW_CONFIG_REPLY, snat_show_config_reply) \ -_(SNAT_ADDRESS_DETAILS, snat_address_details) \ -_(SNAT_INTERFACE_DETAILS, snat_interface_details) \ -_(SNAT_SET_WORKERS_REPLY, snat_set_workers_reply) \ -_(SNAT_WORKER_DETAILS, snat_worker_details) - -/* M: construct, but don't yet send a message */ -#define M(T,t) \ -do { \ - vam->result_ready = 0; \ - mp = vl_msg_api_alloc(sizeof(*mp)); \ - memset (mp, 0, sizeof (*mp)); \ - mp->_vl_msg_id = ntohs (VL_API_##T + sm->msg_id_base); \ - mp->client_index = vam->my_client_index; \ -} while(0); - -#define M2(T,t,n) \ -do { \ - vam->result_ready = 0; \ - mp = vl_msg_api_alloc(sizeof(*mp)+(n)); \ - memset (mp, 0, sizeof (*mp)); \ - mp->_vl_msg_id = ntohs (VL_API_##T + sm->msg_id_base); \ - mp->client_index = vam->my_client_index; \ -} while(0); - -/* S: send a message */ -#define S (vl_msg_api_send_shmem (vam->vl_input_queue, (u8 *)&mp)) - -/* W: wait for results, with timeout */ -#define W \ -do { \ - timeout = vat_time_now (vam) + 1.0; \ - \ - while (vat_time_now (vam) < timeout) { \ - if (vam->result_ready == 1) { \ - return (vam->retval); \ - } \ - } \ - return -99; \ -} while(0); - -static int api_snat_add_address_range (vat_main_t * vam) -{ - snat_test_main_t * sm = &snat_test_main; - unformat_input_t * i = vam->input; - f64 timeout; - ip4_address_t start_addr, end_addr; - u32 start_host_order, end_host_order; - vl_api_snat_add_address_range_t * mp; - u8 is_add = 1; - int count; - - while (unformat_check_input (i) != UNFORMAT_END_OF_INPUT) - { - if (unformat (i, "%U - %U", - unformat_ip4_address, &start_addr, - unformat_ip4_address, &end_addr)) - ; - else if (unformat (i, "%U", unformat_ip4_address, &start_addr)) - end_addr = start_addr; - else if (unformat (i, "del")) - is_add = 0; - else - { - clib_warning("unknown input '%U'", format_unformat_error, i); - return -99; - } - } - - start_host_order = clib_host_to_net_u32 (start_addr.as_u32); - end_host_order = clib_host_to_net_u32 (end_addr.as_u32); - - if (end_host_order < start_host_order) - { - errmsg ("end address less than start address\n"); - return -99; - } - - count = (end_host_order - start_host_order) + 1; - - if (count > 1024) - { - errmsg ("%U - %U, %d addresses...\n", - format_ip4_address, &start_addr, - format_ip4_address, &end_addr, - count); - } - - M(SNAT_ADD_ADDRESS_RANGE, snat_add_address_range); - - memcpy (mp->first_ip_address, &start_addr, 4); - memcpy (mp->last_ip_address, &end_addr, 4); - mp->is_ip4 = 1; - mp->is_add = is_add; - - S; W; - - /* NOTREACHED */ - return 0; -} - -static int api_snat_interface_add_del_feature (vat_main_t * vam) -{ - snat_test_main_t * sm = &snat_test_main; - unformat_input_t * i = vam->input; - f64 timeout; - vl_api_snat_interface_add_del_feature_t * mp; - u32 sw_if_index; - u8 sw_if_index_set = 0; - u8 is_inside = 1; - u8 is_add = 1; - - while (unformat_check_input (i) != UNFORMAT_END_OF_INPUT) - { - if (unformat (i, "%U", unformat_sw_if_index, vam, &sw_if_index)) - sw_if_index_set = 1; - else if (unformat (i, "sw_if_index %d", &sw_if_index)) - sw_if_index_set = 1; - else if (unformat (i, "out")) - is_inside = 0; - else if (unformat (i, "in")) - is_inside = 1; - else if (unformat (i, "del")) - is_add = 0; - else - { - clib_warning("unknown input '%U'", format_unformat_error, i); - return -99; - } - } - - if (sw_if_index_set == 0) - { - errmsg ("interface / sw_if_index required\n"); - return -99; - } - - M(SNAT_INTERFACE_ADD_DEL_FEATURE, snat_interface_add_del_feature); - mp->sw_if_index = ntohl(sw_if_index); - mp->is_add = is_add; - mp->is_inside = is_inside; - - S; W; - /* NOTREACHED */ - return 0; -} - -static int api_snat_add_static_mapping(vat_main_t * vam) -{ - snat_test_main_t * sm = &snat_test_main; - unformat_input_t * i = vam->input; - f64 timeout; - vl_api_snat_add_static_mapping_t * mp; - u8 addr_set_n = 0; - u8 is_add = 1; - u8 addr_only = 1; - ip4_address_t local_addr, external_addr; - u32 local_port = 0, external_port = 0, vrf_id = ~0; - - while (unformat_check_input (i) != UNFORMAT_END_OF_INPUT) - { - if (unformat (i, "local_addr %U", unformat_ip4_address, &local_addr)) - addr_set_n++; - else if (unformat (i, "external_addr %U", unformat_ip4_address, - &external_addr)) - addr_set_n++; - else if (unformat (i, "local_port %u", &local_port)) - addr_only = 0; - else if (unformat (i, "external_port %u", &external_port)) - addr_only = 0; - else if (unformat (i, "vrf %u", &vrf_id)) - ; - else if (unformat (i, "del")) - is_add = 0; - else - { - clib_warning("unknown input '%U'", format_unformat_error, i); - return -99; - } - } - - if (addr_set_n != 2) - { - errmsg ("local_addr and remote_addr required\n"); - return -99; - } - - M(SNAT_ADD_STATIC_MAPPING, snat_add_static_mapping); - mp->is_add = is_add; - mp->is_ip4 = 1; - mp->addr_only = addr_only; - mp->local_port = ntohs ((u16) local_port); - mp->external_port = ntohs ((u16) external_port); - mp->vrf_id = ntohl (vrf_id); - memcpy (mp->local_ip_address, &local_addr, 4); - memcpy (mp->external_ip_address, &external_addr, 4); - - S; W; - /* NOTREACHED */ - return 0; -} - -static void vl_api_snat_control_ping_reply_t_handler - (vl_api_snat_control_ping_reply_t * mp) -{ - vat_main_t *vam = &vat_main; - i32 retval = ntohl (mp->retval); - if (vam->async_mode) - { - vam->async_errors += (retval < 0); - } - else - { - vam->retval = retval; - vam->result_ready = 1; - } -} - -static void vl_api_snat_static_mapping_details_t_handler - (vl_api_snat_static_mapping_details_t *mp) -{ - snat_test_main_t * sm = &snat_test_main; - vat_main_t *vam = sm->vat_main; - - if (mp->addr_only) - fformat (vam->ofp, "%15U%6s%15U%6s%11d\n", - format_ip4_address, &mp->local_ip_address, "", - format_ip4_address, &mp->external_ip_address, "", - ntohl (mp->vrf_id)); - else - fformat (vam->ofp, "%15U%6d%15U%6d%11d\n", - format_ip4_address, &mp->local_ip_address, - ntohs (mp->local_port), - format_ip4_address, &mp->external_ip_address, - ntohs (mp->external_port), - ntohl (mp->vrf_id)); - -} - -static int api_snat_static_mapping_dump(vat_main_t * vam) -{ - snat_test_main_t * sm = &snat_test_main; - f64 timeout; - vl_api_snat_static_mapping_dump_t * mp; - - if (vam->json_output) - { - clib_warning ("JSON output not supported for snat_static_mapping_dump"); - return -99; - } - - fformat (vam->ofp, "%21s%21s\n", "local", "external"); - fformat (vam->ofp, "%15s%6s%15s%6s%11s\n", "address", "port", "address", - "port", "vrf"); - - M(SNAT_STATIC_MAPPING_DUMP, snat_static_mapping_dump); - S; - /* Use a control ping for synchronization */ - { - vl_api_snat_control_ping_t *mp; - M (SNAT_CONTROL_PING, snat_control_ping); - S; - } - W; - /* NOTREACHED */ - return 0; -} - -static void vl_api_snat_show_config_reply_t_handler - (vl_api_snat_show_config_reply_t *mp) -{ - snat_test_main_t * sm = &snat_test_main; - vat_main_t *vam = sm->vat_main; - i32 retval = ntohl (mp->retval); - - if (retval >= 0) - { - fformat (vam->ofp, "translation hash buckets %d\n", - ntohl (mp->translation_buckets)); - fformat (vam->ofp, "translation hash memory %d\n", - ntohl (mp->translation_memory_size)); - fformat (vam->ofp, "user hash buckets %d\n", ntohl (mp->user_buckets)); - fformat (vam->ofp, "user hash memory %d\n", ntohl (mp->user_memory_size)); - fformat (vam->ofp, "max translations per user %d\n", - ntohl (mp->max_translations_per_user)); - fformat (vam->ofp, "outside VRF id %d\n", ntohl (mp->outside_vrf_id)); - fformat (vam->ofp, "inside VRF id %d\n", ntohl (mp->inside_vrf_id)); - if (mp->static_mapping_only) - { - fformat (vam->ofp, "static mapping only"); - if (mp->static_mapping_connection_tracking) - fformat (vam->ofp, " connection tracking"); - fformat (vam->ofp, "\n"); - } - } - vam->retval = retval; - vam->result_ready = 1; -} - -static int api_snat_show_config(vat_main_t * vam) -{ - snat_test_main_t * sm = &snat_test_main; - f64 timeout; - vl_api_snat_show_config_t * mp; - - if (vam->json_output) - { - clib_warning ("JSON output not supported for snat_show_config"); - return -99; - } - - M(SNAT_SHOW_CONFIG, snat_show_config); - S; W; - /* NOTREACHED */ - return 0; -} - -static void vl_api_snat_address_details_t_handler - (vl_api_snat_address_details_t *mp) -{ - snat_test_main_t * sm = &snat_test_main; - vat_main_t *vam = sm->vat_main; - - fformat (vam->ofp, "%U\n", format_ip4_address, &mp->ip_address); -} - -static int api_snat_address_dump(vat_main_t * vam) -{ - snat_test_main_t * sm = &snat_test_main; - f64 timeout; - vl_api_snat_address_dump_t * mp; - - if (vam->json_output) - { - clib_warning ("JSON output not supported for snat_address_dump"); - return -99; - } - - M(SNAT_ADDRESS_DUMP, snat_address_dump); - S; - /* Use a control ping for synchronization */ - { - vl_api_snat_control_ping_t *mp; - M (SNAT_CONTROL_PING, snat_control_ping); - S; - } - W; - /* NOTREACHED */ - return 0; -} - -static void vl_api_snat_interface_details_t_handler - (vl_api_snat_interface_details_t *mp) -{ - snat_test_main_t * sm = &snat_test_main; - vat_main_t *vam = sm->vat_main; - - fformat (vam->ofp, "sw_if_index %d %s\n", ntohl (mp->sw_if_index), - mp->is_inside ? "in" : "out"); -} - -static int api_snat_interface_dump(vat_main_t * vam) -{ - snat_test_main_t * sm = &snat_test_main; - f64 timeout; - vl_api_snat_interface_dump_t * mp; - - if (vam->json_output) - { - clib_warning ("JSON output not supported for snat_address_dump"); - return -99; - } - - M(SNAT_INTERFACE_DUMP, snat_interface_dump); - S; - /* Use a control ping for synchronization */ - { - vl_api_snat_control_ping_t *mp; - M (SNAT_CONTROL_PING, snat_control_ping); - S; - } - W; - /* NOTREACHED */ - return 0; -} - -static int api_snat_set_workers (vat_main_t * vam) -{ - snat_test_main_t * sm = &snat_test_main; - unformat_input_t * i = vam->input; - f64 timeout; - vl_api_snat_set_workers_t * mp; - uword *bitmap; - - while (unformat_check_input (i) != UNFORMAT_END_OF_INPUT) - { - if (unformat (i, "%U", unformat_bitmap_list, &bitmap)) - ; - else - { - clib_warning("unknown input '%U'", format_unformat_error, i); - return -99; - } - } - - M(SNAT_SET_WORKERS, snat_set_workers); - mp->worker_mask = clib_host_to_net_u64 (bitmap[0]); - - S; W; - - /* NOTREACHED */ - return 0; -} - -static void vl_api_snat_worker_details_t_handler - (vl_api_snat_worker_details_t *mp) -{ - snat_test_main_t * sm = &snat_test_main; - vat_main_t *vam = sm->vat_main; - - fformat (vam->ofp, "worker_index %d (%s at lcore %u)\n", - ntohl (mp->worker_index), mp->name, ntohl (mp->lcore_id)); -} - -static int api_snat_worker_dump(vat_main_t * vam) -{ - snat_test_main_t * sm = &snat_test_main; - f64 timeout; - vl_api_snat_worker_dump_t * mp; - - if (vam->json_output) - { - clib_warning ("JSON output not supported for snat_address_dump"); - return -99; - } - - M(SNAT_WORKER_DUMP, snat_worker_dump); - S; - /* Use a control ping for synchronization */ - { - vl_api_snat_control_ping_t *mp; - M (SNAT_CONTROL_PING, snat_control_ping); - S; - } - W; - /* NOTREACHED */ - return 0; -} - -/* - * List of messages that the api test plugin sends, - * and that the data plane plugin processes - */ -#define foreach_vpe_api_msg \ -_(snat_add_address_range, " [- | sw_if_index [in] [out] [del]") \ -_(snat_add_static_mapping, "local_addr external_addr " \ - "[local_port ] [external_port ] [vrf ] [del]") \ -_(snat_set_workers, "") \ -_(snat_static_mapping_dump, "") \ -_(snat_show_config, "") \ -_(snat_address_dump, "") \ -_(snat_interface_dump, "") \ -_(snat_worker_dump, "") - -void vat_api_hookup (vat_main_t *vam) -{ - snat_test_main_t * sm __attribute__((unused)) = &snat_test_main; - /* Hook up handlers for replies from the data plane plug-in */ -#define _(N,n) \ - vl_msg_api_set_handlers((VL_API_##N + sm->msg_id_base), \ - #n, \ - vl_api_##n##_t_handler, \ - vl_noop_handler, \ - vl_api_##n##_t_endian, \ - vl_api_##n##_t_print, \ - sizeof(vl_api_##n##_t), 1); - foreach_vpe_api_reply_msg; -#undef _ - - /* API messages we can send */ -#define _(n,h) hash_set_mem (vam->function_by_name, #n, api_##n); - foreach_vpe_api_msg; -#undef _ - - /* Help strings */ -#define _(n,h) hash_set_mem (vam->help_by_name, #n, h); - foreach_vpe_api_msg; -#undef _ -} - -clib_error_t * vat_plugin_register (vat_main_t *vam) -{ - snat_test_main_t * sm = &snat_test_main; - u8 * name; - - sm->vat_main = vam; - - /* Ask the vpp engine for the first assigned message-id */ - name = format (0, "snat_%08x%c", api_version, 0); - sm->msg_id_base = vl_client_get_first_plugin_msg_id ((char *) name); - - if (sm->msg_id_base != (u16) ~0) - vat_api_hookup (vam); - - vec_free(name); - - return 0; -} diff --git a/src/Makefile.am b/src/Makefile.am index e691a539..0fc437a4 100644 --- a/src/Makefile.am +++ b/src/Makefile.am @@ -79,7 +79,15 @@ include vnet.am include vpp.am include vpp-api-test.am -SUBDIRS += vpp-api/python plugins +SUBDIRS += plugins + +if ENABLE_PAPI +SUBDIRS += vpp-api/python +endif + +if ENABLE_JAPI +SUBDIRS += vpp-api/java +endif ############################################################################### # API diff --git a/src/configure.ac b/src/configure.ac index f8e4d94f..4ed55705 100644 --- a/src/configure.ac +++ b/src/configure.ac @@ -3,7 +3,8 @@ LT_INIT AC_CONFIG_AUX_DIR([.]) AM_INIT_AUTOMAKE([subdir-objects]) AM_SILENT_RULES([yes]) -AC_CONFIG_FILES([Makefile plugins/Makefile vpp-api/python/Makefile]) +AC_CONFIG_FILES([Makefile plugins/Makefile vpp-api/python/Makefile vpp-api/java/Makefile]) +AC_CONFIG_MACRO_DIR([m4]) AC_PROG_CC AM_PROG_AS @@ -89,6 +90,8 @@ ENABLE_ARG(g2, [Enable g2]) # --disable-X DISABLE_ARG(vlib, [Disable vlib and dependant libs and binaries]) DISABLE_ARG(svm, [Disable svm and dependant libs and binaries]) +DISABLE_ARG(papi, [Disable Python API bindings]) +DISABLE_ARG(japi, [Disable Java API bindings]) # --with-X WITH_ARG(dpdk, [Use use DPDK]) @@ -131,6 +134,25 @@ AC_DEFINE_UNQUOTED(DPDK_CRYPTO, [${n_with_dpdk_crypto}]) AC_DEFINE_UNQUOTED(IPSEC, [${n_with_ipsec}]) AC_DEFINE_UNQUOTED(IPV6SR, [${n_with_ipv6sr}]) + +# Silence following noise: +# ar: `u' modifier ignored since `D' is the default (see `U') +AR_FLAGS=cr +AC_SUBST(AR_FLAGS) + +############################################################################### +# Plugins +############################################################################### + +# Please keep alphabetical order +PLUGIN_ENABLED(acl) +PLUGIN_ENABLED(flowperpkt) +PLUGIN_ENABLED(ila) +PLUGIN_ENABLED(ioam) +PLUGIN_ENABLED(lb) +PLUGIN_ENABLED(sixrd) +PLUGIN_ENABLED(snat) + ############################################################################### # Dependency checks ############################################################################### @@ -151,12 +173,17 @@ AM_COND_IF([ENABLE_G2], ]) ############################################################################### -# Plugins +# JAVA ############################################################################### -PLUGIN_ENABLED(sixrd) -PLUGIN_ENABLED(ila) -PLUGIN_ENABLED(flowperpkt) +AM_COND_IF([ENABLE_JAPI], +[ + AX_VPP_FIND_JDK8 + AC_SUBST(JAVA_HOME) + AC_SUBST(JAVAC) + AC_SUBST(JAVAH) + AC_SUBST(JAR) +]) ############################################################################### # Output @@ -172,6 +199,11 @@ PRINT_VAL([includedir], ${includedir}) PRINT_VAL([CFLAGS], ${CFLAGS}) PRINT_VAL([CPPFLAGS], ${CPPFLAGS}) PRINT_VAL([LDFLAGS], ${LDFLAGS}) +AM_COND_IF([ENABLE_JAPI], +[ + PRINT_VAL([JAVA_VERSION], ${JAVA_VERSION}) + PRINT_VAL([JAVA_HOME], ${JAVA_HOME}) +]) AC_MSG_RESULT([]) AC_MSG_RESULT([with:]) diff --git a/src/examples/sample-plugin/Makefile.am b/src/examples/sample-plugin/Makefile.am new file mode 100644 index 00000000..e221f8c1 --- /dev/null +++ b/src/examples/sample-plugin/Makefile.am @@ -0,0 +1,56 @@ +# Copyright (c) 2015 Cisco and/or its affiliates. +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at: +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +AUTOMAKE_OPTIONS = foreign subdir-objects + +AM_CFLAGS = -Wall +AM_LDFLAGS = -module -shared -avoid-version + +vppapitestpluginsdir = ${libdir}/vpp_api_test_plugins +vpppluginsdir = ${libdir}/vpp_plugins + +vppapitestplugins_LTLIBRARIES = sample_test_plugin.la +vppplugins_LTLIBRARIES = sample_plugin.la + +sample_plugin_la_SOURCES = sample/sample.c sample/node.c \ + sample/sample_plugin.api.h + +BUILT_SOURCES = sample/sample.api.h sample/sample.api.json + +SUFFIXES = .api.h .api + +%.api.h: %.api + mkdir -p `dirname $@` ; \ + $(CC) $(CPPFLAGS) -E -P -C -x c $^ \ + | vppapigen --input - --output $@ --show-name $@ + +%.api.json: %.api + @echo " JSON APIGEN " $@ ; \ + mkdir -p `dirname $@` ; \ + $(CC) $(CPPFLAGS) -E -P -C -x c $^ \ + | vppapigen --input - --json $@ + +apidir = $(prefix)/sample/ +api_DATA = sample.api.json + +noinst_HEADERS = \ + sample/sample_all_api_h.h \ + sample/sample_msg_enum.h \ + sample/sample.api.h + +sample_test_plugin_la_SOURCES = sample/sample_test.c sample/sample_plugin.api.h + +# Remove *.la files +install-data-hook: + @(cd $(vpppluginsdir) && $(RM) $(vppplugins_LTLIBRARIES)) + @(cd $(vppapitestpluginsdir) && $(RM) $(vppapitestplugins_LTLIBRARIES)) diff --git a/src/examples/sample-plugin/configure.ac b/src/examples/sample-plugin/configure.ac new file mode 100644 index 00000000..43642732 --- /dev/null +++ b/src/examples/sample-plugin/configure.ac @@ -0,0 +1,9 @@ +AC_INIT(sample_plugin, 1.0) +LT_INIT +AM_INIT_AUTOMAKE +AM_SILENT_RULES([yes]) +AC_PREFIX_DEFAULT([/usr]) + +AC_PROG_CC + +AC_OUTPUT([Makefile]) diff --git a/src/examples/sample-plugin/sample/node.c b/src/examples/sample-plugin/sample/node.c new file mode 100644 index 00000000..94c1706b --- /dev/null +++ b/src/examples/sample-plugin/sample/node.c @@ -0,0 +1,295 @@ +/* + * Copyright (c) 2015 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ +#include +#include +#include +#include +#include + +typedef struct { + u32 next_index; + u32 sw_if_index; + u8 new_src_mac[6]; + u8 new_dst_mac[6]; +} sample_trace_t; + +static u8 * +format_mac_address (u8 * s, va_list * args) +{ + u8 *a = va_arg (*args, u8 *); + return format (s, "%02x:%02x:%02x:%02x:%02x:%02x", + a[0], a[1], a[2], a[3], a[4], a[5]); +} + +/* packet trace format function */ +static u8 * format_sample_trace (u8 * s, va_list * args) +{ + CLIB_UNUSED (vlib_main_t * vm) = va_arg (*args, vlib_main_t *); + CLIB_UNUSED (vlib_node_t * node) = va_arg (*args, vlib_node_t *); + sample_trace_t * t = va_arg (*args, sample_trace_t *); + + s = format (s, "SAMPLE: sw_if_index %d, next index %d\n", + t->sw_if_index, t->next_index); + s = format (s, " new src %U -> new dst %U", + format_mac_address, t->new_src_mac, + format_mac_address, t->new_dst_mac); + + return s; +} + +vlib_node_registration_t sample_node; + +#define foreach_sample_error \ +_(SWAPPED, "Mac swap packets processed") + +typedef enum { +#define _(sym,str) SAMPLE_ERROR_##sym, + foreach_sample_error +#undef _ + SAMPLE_N_ERROR, +} sample_error_t; + +static char * sample_error_strings[] = { +#define _(sym,string) string, + foreach_sample_error +#undef _ +}; + +typedef enum { + SAMPLE_NEXT_INTERFACE_OUTPUT, + SAMPLE_N_NEXT, +} sample_next_t; + +#define foreach_mac_address_offset \ +_(0) \ +_(1) \ +_(2) \ +_(3) \ +_(4) \ +_(5) + +static uword +sample_node_fn (vlib_main_t * vm, + vlib_node_runtime_t * node, + vlib_frame_t * frame) +{ + u32 n_left_from, * from, * to_next; + sample_next_t next_index; + u32 pkts_swapped = 0; + + from = vlib_frame_vector_args (frame); + n_left_from = frame->n_vectors; + next_index = node->cached_next_index; + + while (n_left_from > 0) + { + u32 n_left_to_next; + + vlib_get_next_frame (vm, node, next_index, + to_next, n_left_to_next); + + while (n_left_from >= 4 && n_left_to_next >= 2) + { + u32 next0 = SAMPLE_NEXT_INTERFACE_OUTPUT; + u32 next1 = SAMPLE_NEXT_INTERFACE_OUTPUT; + u32 sw_if_index0, sw_if_index1; + u8 tmp0[6], tmp1[6]; + ethernet_header_t *en0, *en1; + u32 bi0, bi1; + vlib_buffer_t * b0, * b1; + + /* Prefetch next iteration. */ + { + vlib_buffer_t * p2, * p3; + + p2 = vlib_get_buffer (vm, from[2]); + p3 = vlib_get_buffer (vm, from[3]); + + vlib_prefetch_buffer_header (p2, LOAD); + vlib_prefetch_buffer_header (p3, LOAD); + + CLIB_PREFETCH (p2->data, CLIB_CACHE_LINE_BYTES, STORE); + CLIB_PREFETCH (p3->data, CLIB_CACHE_LINE_BYTES, STORE); + } + + /* speculatively enqueue b0 and b1 to the current next frame */ + to_next[0] = bi0 = from[0]; + to_next[1] = bi1 = from[1]; + from += 2; + to_next += 2; + n_left_from -= 2; + n_left_to_next -= 2; + + b0 = vlib_get_buffer (vm, bi0); + b1 = vlib_get_buffer (vm, bi1); + + ASSERT (b0->current_data == 0); + ASSERT (b1->current_data == 0); + + en0 = vlib_buffer_get_current (b0); + en1 = vlib_buffer_get_current (b1); + + /* This is not the fastest way to swap src + dst mac addresses */ +#define _(a) tmp0[a] = en0->src_address[a]; + foreach_mac_address_offset; +#undef _ +#define _(a) en0->src_address[a] = en0->dst_address[a]; + foreach_mac_address_offset; +#undef _ +#define _(a) en0->dst_address[a] = tmp0[a]; + foreach_mac_address_offset; +#undef _ + +#define _(a) tmp1[a] = en1->src_address[a]; + foreach_mac_address_offset; +#undef _ +#define _(a) en1->src_address[a] = en1->dst_address[a]; + foreach_mac_address_offset; +#undef _ +#define _(a) en1->dst_address[a] = tmp1[a]; + foreach_mac_address_offset; +#undef _ + + + + sw_if_index0 = vnet_buffer(b0)->sw_if_index[VLIB_RX]; + sw_if_index1 = vnet_buffer(b1)->sw_if_index[VLIB_RX]; + + /* Send pkt back out the RX interface */ + vnet_buffer(b0)->sw_if_index[VLIB_TX] = sw_if_index0; + vnet_buffer(b1)->sw_if_index[VLIB_TX] = sw_if_index1; + + pkts_swapped += 2; + + if (PREDICT_FALSE((node->flags & VLIB_NODE_FLAG_TRACE))) + { + if (b0->flags & VLIB_BUFFER_IS_TRACED) + { + sample_trace_t *t = + vlib_add_trace (vm, node, b0, sizeof (*t)); + t->sw_if_index = sw_if_index0; + t->next_index = next0; + clib_memcpy (t->new_src_mac, en0->src_address, + sizeof (t->new_src_mac)); + clib_memcpy (t->new_dst_mac, en0->dst_address, + sizeof (t->new_dst_mac)); + + } + if (b1->flags & VLIB_BUFFER_IS_TRACED) + { + sample_trace_t *t = + vlib_add_trace (vm, node, b1, sizeof (*t)); + t->sw_if_index = sw_if_index1; + t->next_index = next1; + clib_memcpy (t->new_src_mac, en1->src_address, + sizeof (t->new_src_mac)); + clib_memcpy (t->new_dst_mac, en1->dst_address, + sizeof (t->new_dst_mac)); + } + } + + /* verify speculative enqueues, maybe switch current next frame */ + vlib_validate_buffer_enqueue_x2 (vm, node, next_index, + to_next, n_left_to_next, + bi0, bi1, next0, next1); + } + + while (n_left_from > 0 && n_left_to_next > 0) + { + u32 bi0; + vlib_buffer_t * b0; + u32 next0 = SAMPLE_NEXT_INTERFACE_OUTPUT; + u32 sw_if_index0; + u8 tmp0[6]; + ethernet_header_t *en0; + + /* speculatively enqueue b0 to the current next frame */ + bi0 = from[0]; + to_next[0] = bi0; + from += 1; + to_next += 1; + n_left_from -= 1; + n_left_to_next -= 1; + + b0 = vlib_get_buffer (vm, bi0); + /* + * Direct from the driver, we should be at offset 0 + * aka at &b0->data[0] + */ + ASSERT (b0->current_data == 0); + + en0 = vlib_buffer_get_current (b0); + + /* This is not the fastest way to swap src + dst mac addresses */ +#define _(a) tmp0[a] = en0->src_address[a]; + foreach_mac_address_offset; +#undef _ +#define _(a) en0->src_address[a] = en0->dst_address[a]; + foreach_mac_address_offset; +#undef _ +#define _(a) en0->dst_address[a] = tmp0[a]; + foreach_mac_address_offset; +#undef _ + + sw_if_index0 = vnet_buffer(b0)->sw_if_index[VLIB_RX]; + + /* Send pkt back out the RX interface */ + vnet_buffer(b0)->sw_if_index[VLIB_TX] = sw_if_index0; + + if (PREDICT_FALSE((node->flags & VLIB_NODE_FLAG_TRACE) + && (b0->flags & VLIB_BUFFER_IS_TRACED))) { + sample_trace_t *t = + vlib_add_trace (vm, node, b0, sizeof (*t)); + t->sw_if_index = sw_if_index0; + t->next_index = next0; + clib_memcpy (t->new_src_mac, en0->src_address, + sizeof (t->new_src_mac)); + clib_memcpy (t->new_dst_mac, en0->dst_address, + sizeof (t->new_dst_mac)); + } + + pkts_swapped += 1; + + /* verify speculative enqueue, maybe switch current next frame */ + vlib_validate_buffer_enqueue_x1 (vm, node, next_index, + to_next, n_left_to_next, + bi0, next0); + } + + vlib_put_next_frame (vm, node, next_index, n_left_to_next); + } + + vlib_node_increment_counter (vm, sample_node.index, + SAMPLE_ERROR_SWAPPED, pkts_swapped); + return frame->n_vectors; +} + +VLIB_REGISTER_NODE (sample_node) = { + .function = sample_node_fn, + .name = "sample", + .vector_size = sizeof (u32), + .format_trace = format_sample_trace, + .type = VLIB_NODE_TYPE_INTERNAL, + + .n_errors = ARRAY_LEN(sample_error_strings), + .error_strings = sample_error_strings, + + .n_next_nodes = SAMPLE_N_NEXT, + + /* edit / add dispositions here */ + .next_nodes = { + [SAMPLE_NEXT_INTERFACE_OUTPUT] = "interface-output", + }, +}; diff --git a/src/examples/sample-plugin/sample/sample.api b/src/examples/sample-plugin/sample/sample.api new file mode 100644 index 00000000..f99cdb38 --- /dev/null +++ b/src/examples/sample-plugin/sample/sample.api @@ -0,0 +1,39 @@ +/* Hey Emacs use -*- mode: C -*- */ +/* + * Copyright (c) 2015 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +/* Define a simple binary API to control the feature */ + +define sample_macswap_enable_disable { + /* Client identifier, set from api_main.my_client_index */ + u32 client_index; + + /* Arbitrary context, so client can match reply to request */ + u32 context; + + /* Enable / disable the feature */ + u8 enable_disable; + + /* Interface handle */ + u32 sw_if_index; +}; + +define sample_macswap_enable_disable_reply { + /* From the request */ + u32 context; + + /* Return value, zero means all OK */ + i32 retval; +}; diff --git a/src/examples/sample-plugin/sample/sample.c b/src/examples/sample-plugin/sample/sample.c new file mode 100644 index 00000000..603cb2d0 --- /dev/null +++ b/src/examples/sample-plugin/sample/sample.c @@ -0,0 +1,255 @@ +/* + * Copyright (c) 2015 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ +/* + *------------------------------------------------------------------ + * sample.c - simple MAC-swap API / debug CLI handling + *------------------------------------------------------------------ + */ + +#include +#include +#include + +#include +#include +#include + +/* define message IDs */ +#include + +/* define message structures */ +#define vl_typedefs +#include +#undef vl_typedefs + +/* define generated endian-swappers */ +#define vl_endianfun +#include +#undef vl_endianfun + +/* instantiate all the print functions we know about */ +#define vl_print(handle, ...) vlib_cli_output (handle, __VA_ARGS__) +#define vl_printfun +#include +#undef vl_printfun + +/* Get the API version number */ +#define vl_api_version(n,v) static u32 api_version=(v); +#include +#undef vl_api_version + +/* + * A handy macro to set up a message reply. + * Assumes that the following variables are available: + * mp - pointer to request message + * rmp - pointer to reply message type + * rv - return value + */ + +#define REPLY_MACRO(t) \ +do { \ + unix_shared_memory_queue_t * q = \ + vl_api_client_index_to_input_queue (mp->client_index); \ + if (!q) \ + return; \ + \ + rmp = vl_msg_api_alloc (sizeof (*rmp)); \ + rmp->_vl_msg_id = ntohs((t)+sm->msg_id_base); \ + rmp->context = mp->context; \ + rmp->retval = ntohl(rv); \ + \ + vl_msg_api_send_shmem (q, (u8 *)&rmp); \ +} while(0); + + +/* List of message types that this plugin understands */ + +#define foreach_sample_plugin_api_msg \ +_(SAMPLE_MACSWAP_ENABLE_DISABLE, sample_macswap_enable_disable) + +/* + * This routine exists to convince the vlib plugin framework that + * we haven't accidentally copied a random .dll into the plugin directory. + * + * Also collects global variable pointers passed from the vpp engine + */ + +clib_error_t * +vlib_plugin_register (vlib_main_t * vm, vnet_plugin_handoff_t * h, + int from_early_init) +{ + sample_main_t * sm = &sample_main; + clib_error_t * error = 0; + + sm->vlib_main = vm; + sm->vnet_main = h->vnet_main; + sm->ethernet_main = h->ethernet_main; + + return error; +} + +/* Action function shared between message handler and debug CLI */ + +int sample_macswap_enable_disable (sample_main_t * sm, u32 sw_if_index, + int enable_disable) +{ + vnet_sw_interface_t * sw; + int rv = 0; + + /* Utterly wrong? */ + if (pool_is_free_index (sm->vnet_main->interface_main.sw_interfaces, + sw_if_index)) + return VNET_API_ERROR_INVALID_SW_IF_INDEX; + + /* Not a physical port? */ + sw = vnet_get_sw_interface (sm->vnet_main, sw_if_index); + if (sw->type != VNET_SW_INTERFACE_TYPE_HARDWARE) + return VNET_API_ERROR_INVALID_SW_IF_INDEX; + + vnet_feature_enable_disable ("device-input", "sample", + sw_if_index, enable_disable, 0, 0); + + return rv; +} + +static clib_error_t * +macswap_enable_disable_command_fn (vlib_main_t * vm, + unformat_input_t * input, + vlib_cli_command_t * cmd) +{ + sample_main_t * sm = &sample_main; + u32 sw_if_index = ~0; + int enable_disable = 1; + + int rv; + + while (unformat_check_input (input) != UNFORMAT_END_OF_INPUT) { + if (unformat (input, "disable")) + enable_disable = 0; + else if (unformat (input, "%U", unformat_vnet_sw_interface, + sm->vnet_main, &sw_if_index)) + ; + else + break; + } + + if (sw_if_index == ~0) + return clib_error_return (0, "Please specify an interface..."); + + rv = sample_macswap_enable_disable (sm, sw_if_index, enable_disable); + + switch(rv) { + case 0: + break; + + case VNET_API_ERROR_INVALID_SW_IF_INDEX: + return clib_error_return + (0, "Invalid interface, only works on physical ports"); + break; + + case VNET_API_ERROR_UNIMPLEMENTED: + return clib_error_return (0, "Device driver doesn't support redirection"); + break; + + default: + return clib_error_return (0, "sample_macswap_enable_disable returned %d", + rv); + } + return 0; +} + +VLIB_CLI_COMMAND (sr_content_command, static) = { + .path = "sample macswap", + .short_help = + "sample macswap [disable]", + .function = macswap_enable_disable_command_fn, +}; + +/* API message handler */ +static void vl_api_sample_macswap_enable_disable_t_handler +(vl_api_sample_macswap_enable_disable_t * mp) +{ + vl_api_sample_macswap_enable_disable_reply_t * rmp; + sample_main_t * sm = &sample_main; + int rv; + + rv = sample_macswap_enable_disable (sm, ntohl(mp->sw_if_index), + (int) (mp->enable_disable)); + + REPLY_MACRO(VL_API_SAMPLE_MACSWAP_ENABLE_DISABLE_REPLY); +} + +/* Set up the API message handling tables */ +static clib_error_t * +sample_plugin_api_hookup (vlib_main_t *vm) +{ + sample_main_t * sm = &sample_main; +#define _(N,n) \ + vl_msg_api_set_handlers((VL_API_##N + sm->msg_id_base), \ + #n, \ + vl_api_##n##_t_handler, \ + vl_noop_handler, \ + vl_api_##n##_t_endian, \ + vl_api_##n##_t_print, \ + sizeof(vl_api_##n##_t), 1); + foreach_sample_plugin_api_msg; +#undef _ + + return 0; +} + +#define vl_msg_name_crc_list +#include +#undef vl_msg_name_crc_list + +static void +setup_message_id_table (sample_main_t * sm, api_main_t *am) +{ +#define _(id,n,crc) \ + vl_msg_api_add_msg_name_crc (am, #n "_" #crc, id + sm->msg_id_base); + foreach_vl_msg_name_crc_sample; +#undef _ +} + +static clib_error_t * sample_init (vlib_main_t * vm) +{ + sample_main_t * sm = &sample_main; + clib_error_t * error = 0; + u8 * name; + + name = format (0, "sample_%08x%c", api_version, 0); + + /* Ask for a correctly-sized block of API message decode slots */ + sm->msg_id_base = vl_msg_api_get_msg_ids + ((char *) name, VL_MSG_FIRST_AVAILABLE); + + error = sample_plugin_api_hookup (vm); + + /* Add our API messages to the global name_crc hash table */ + setup_message_id_table (sm, &api_main); + + vec_free(name); + + return error; +} + +VLIB_INIT_FUNCTION (sample_init); + +VNET_FEATURE_INIT (sample, static) = +{ + .arc_name = "device-input", + .node_name = "sample", + .runs_before = VNET_FEATURES ("ethernet-input"), +}; diff --git a/src/examples/sample-plugin/sample/sample.h b/src/examples/sample-plugin/sample/sample.h new file mode 100644 index 00000000..d268d482 --- /dev/null +++ b/src/examples/sample-plugin/sample/sample.h @@ -0,0 +1,40 @@ +/* + * Copyright (c) 2015 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ +#ifndef __included_sample_h__ +#define __included_sample_h__ + +#include +#include +#include + +#include +#include +#include + +typedef struct { + /* API message ID base */ + u16 msg_id_base; + + /* convenience */ + vlib_main_t * vlib_main; + vnet_main_t * vnet_main; + ethernet_main_t * ethernet_main; +} sample_main_t; + +sample_main_t sample_main; + +vlib_node_registration_t sample_node; + +#endif /* __included_sample_h__ */ diff --git a/src/examples/sample-plugin/sample/sample_all_api_h.h b/src/examples/sample-plugin/sample/sample_all_api_h.h new file mode 100644 index 00000000..774d782f --- /dev/null +++ b/src/examples/sample-plugin/sample/sample_all_api_h.h @@ -0,0 +1,16 @@ +/* + * Copyright (c) 2015 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ +/* Include the generated file, see BUILT_SOURCES in Makefile.am */ +#include diff --git a/src/examples/sample-plugin/sample/sample_msg_enum.h b/src/examples/sample-plugin/sample/sample_msg_enum.h new file mode 100644 index 00000000..af4172f7 --- /dev/null +++ b/src/examples/sample-plugin/sample/sample_msg_enum.h @@ -0,0 +1,28 @@ +/* + * Copyright (c) 2015 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ +#ifndef included_sample_msg_enum_h +#define included_sample_msg_enum_h + +#include + +#define vl_msg_id(n,h) n, +typedef enum { +#include + /* We'll want to know how many messages IDs we need... */ + VL_MSG_FIRST_AVAILABLE, +} vl_msg_id_t; +#undef vl_msg_id + +#endif /* included_sample_msg_enum_h */ diff --git a/src/examples/sample-plugin/sample/sample_test.c b/src/examples/sample-plugin/sample/sample_test.c new file mode 100644 index 00000000..dd1b0215 --- /dev/null +++ b/src/examples/sample-plugin/sample/sample_test.c @@ -0,0 +1,213 @@ +/* + * Copyright (c) 2015 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ +/* + *------------------------------------------------------------------ + * sample_test.c - test harness plugin + *------------------------------------------------------------------ + */ + +#include +#include +#include +#include +#include + +uword unformat_sw_if_index (unformat_input_t * input, va_list * args); + +/* Declare message IDs */ +#include + +/* define message structures */ +#define vl_typedefs +#include +#undef vl_typedefs + +/* declare message handlers for each api */ + +#define vl_endianfun /* define message structures */ +#include +#undef vl_endianfun + +/* instantiate all the print functions we know about */ +#define vl_print(handle, ...) +#define vl_printfun +#include +#undef vl_printfun + +/* Get the API version number. */ +#define vl_api_version(n,v) static u32 api_version=(v); +#include +#undef vl_api_version + + +typedef struct { + /* API message ID base */ + u16 msg_id_base; + vat_main_t *vat_main; +} sample_test_main_t; + +sample_test_main_t sample_test_main; + +#define foreach_standard_reply_retval_handler \ +_(sample_macswap_enable_disable_reply) + +#define _(n) \ + static void vl_api_##n##_t_handler \ + (vl_api_##n##_t * mp) \ + { \ + vat_main_t * vam = sample_test_main.vat_main; \ + i32 retval = ntohl(mp->retval); \ + if (vam->async_mode) { \ + vam->async_errors += (retval < 0); \ + } else { \ + vam->retval = retval; \ + vam->result_ready = 1; \ + } \ + } +foreach_standard_reply_retval_handler; +#undef _ + +/* + * Table of message reply handlers, must include boilerplate handlers + * we just generated + */ +#define foreach_vpe_api_reply_msg \ +_(SAMPLE_MACSWAP_ENABLE_DISABLE_REPLY, sample_macswap_enable_disable_reply) + + +/* M: construct, but don't yet send a message */ + +#define M(T,t) \ +do { \ + vam->result_ready = 0; \ + mp = vl_msg_api_alloc(sizeof(*mp)); \ + memset (mp, 0, sizeof (*mp)); \ + mp->_vl_msg_id = ntohs (VL_API_##T + sm->msg_id_base); \ + mp->client_index = vam->my_client_index; \ +} while(0); + +#define M2(T,t,n) \ +do { \ + vam->result_ready = 0; \ + mp = vl_msg_api_alloc(sizeof(*mp)+(n)); \ + memset (mp, 0, sizeof (*mp)); \ + mp->_vl_msg_id = ntohs (VL_API_##T + sm->msg_id_base); \ + mp->client_index = vam->my_client_index; \ +} while(0); + +/* S: send a message */ +#define S (vl_msg_api_send_shmem (vam->vl_input_queue, (u8 *)&mp)) + +/* W: wait for results, with timeout */ +#define W \ +do { \ + timeout = vat_time_now (vam) + 1.0; \ + \ + while (vat_time_now (vam) < timeout) { \ + if (vam->result_ready == 1) { \ + return (vam->retval); \ + } \ + } \ + return -99; \ +} while(0); + +static int api_sample_macswap_enable_disable (vat_main_t * vam) +{ + sample_test_main_t * sm = &sample_test_main; + unformat_input_t * i = vam->input; + f64 timeout; + int enable_disable = 1; + u32 sw_if_index = ~0; + vl_api_sample_macswap_enable_disable_t * mp; + + /* Parse args required to build the message */ + while (unformat_check_input (i) != UNFORMAT_END_OF_INPUT) { + if (unformat (i, "%U", unformat_sw_if_index, vam, &sw_if_index)) + ; + else if (unformat (i, "sw_if_index %d", &sw_if_index)) + ; + else if (unformat (i, "disable")) + enable_disable = 0; + else + break; + } + + if (sw_if_index == ~0) { + errmsg ("missing interface name / explicit sw_if_index number \n"); + return -99; + } + + /* Construct the API message */ + M(SAMPLE_MACSWAP_ENABLE_DISABLE, sample_macswap_enable_disable); + mp->sw_if_index = ntohl (sw_if_index); + mp->enable_disable = enable_disable; + + /* send it... */ + S; + + /* Wait for a reply... */ + W; +} + +/* + * List of messages that the api test plugin sends, + * and that the data plane plugin processes + */ +#define foreach_vpe_api_msg \ +_(sample_macswap_enable_disable, " [disable]") + +void vat_api_hookup (vat_main_t *vam) +{ + sample_test_main_t * sm = &sample_test_main; + /* Hook up handlers for replies from the data plane plug-in */ +#define _(N,n) \ + vl_msg_api_set_handlers((VL_API_##N + sm->msg_id_base), \ + #n, \ + vl_api_##n##_t_handler, \ + vl_noop_handler, \ + vl_api_##n##_t_endian, \ + vl_api_##n##_t_print, \ + sizeof(vl_api_##n##_t), 1); + foreach_vpe_api_reply_msg; +#undef _ + + /* API messages we can send */ +#define _(n,h) hash_set_mem (vam->function_by_name, #n, api_##n); + foreach_vpe_api_msg; +#undef _ + + /* Help strings */ +#define _(n,h) hash_set_mem (vam->help_by_name, #n, h); + foreach_vpe_api_msg; +#undef _ +} + +clib_error_t * vat_plugin_register (vat_main_t *vam) +{ + sample_test_main_t * sm = &sample_test_main; + u8 * name; + + sm->vat_main = vam; + + name = format (0, "sample_%08x%c", api_version, 0); + sm->msg_id_base = vl_client_get_first_plugin_msg_id ((char *) name); + + if (sm->msg_id_base != (u16) ~0) + vat_api_hookup (vam); + + vec_free(name); + + return 0; +} diff --git a/src/m4/ax_vpp_find_jdk8.m4 b/src/m4/ax_vpp_find_jdk8.m4 new file mode 100644 index 00000000..29459822 --- /dev/null +++ b/src/m4/ax_vpp_find_jdk8.m4 @@ -0,0 +1,29 @@ + + +AC_DEFUN([AX_VPP_FIND_JDK8], +[ +while true +do + test "${JAVA_HOME+set}" = set && break + + for dir in $(find /usr/lib/jvm/* -maxdepth 0 -type d); do + AC_MSG_CHECKING([${dir} for Java 8 compiler]) + JAVA_VERSION=$(${dir}/bin/javac -source 8 -version 2>&1) + if test 0 -eq "$?"; then + JAVA_VERSION=$(echo "${JAVA_VERSION}" | cut -d\ -f2) + JAVA_HOME=${dir} + JAVAC=${dir}/bin/javac + JAVAH=${dir}/bin/javah + JAR=${dir}/bin/jar + AC_MSG_RESULT([found version $JAVA_VERSION]) + break + else + JAVA_VERSION="" + AC_MSG_RESULT([no]) + fi + done + + test "${JAVA_HOME}set" = set && AC_MSG_ERROR([Could not find Java 8 compiler]) + break +done +]) diff --git a/src/plugins/Makefile.am b/src/plugins/Makefile.am index ffc4b3ab..987310b7 100644 --- a/src/plugins/Makefile.am +++ b/src/plugins/Makefile.am @@ -22,10 +22,15 @@ BUILT_SOURCES = vppplugins_LTLIBRARIES = vppapitestplugins_LTLIBRARIES = noinst_HEADERS = +nobase_apiinclude_HEADERS = vppapitestpluginsdir = ${libdir}/vpp_api_test_plugins vpppluginsdir = ${libdir}/vpp_plugins +if ENABLE_ACL_PLUGIN +include acl.am +endif + if ENABLE_FLOWPERPKT_PLUGIN include flowperpkt.am endif @@ -34,10 +39,22 @@ if ENABLE_ILA_PLUGIN include ila.am endif +if ENABLE_IOAM_PLUGIN +include ioam.am +endif + +if ENABLE_LB_PLUGIN +include lb.am +endif + if ENABLE_SIXRD_PLUGIN include sixrd.am endif +if ENABLE_SNAT_PLUGIN +include snat.am +endif + include ../suffix-rules.mk # Remove *.la files @@ -50,6 +67,7 @@ install-data-hook: ############################################################################### apidir = $(prefix)/share/vpp/api/plugins +apiincludedir = ${includedir}/vpp_plugins api_DATA = \ $(patsubst %.api,%.api.json,$(API_FILES)) diff --git a/src/plugins/acl.am b/src/plugins/acl.am new file mode 100644 index 00000000..efed31c2 --- /dev/null +++ b/src/plugins/acl.am @@ -0,0 +1,35 @@ +# Copyright (c) 2016 Cisco and/or its affiliates. +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at: +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +vppapitestplugins_LTLIBRARIES += acl_test_plugin.la +vppplugins_LTLIBRARIES += acl_plugin.la + +acl_plugin_la_SOURCES = \ + acl/acl.c \ + acl/node_in.c \ + acl/node_out.c \ + acl/l2sess.c \ + acl/l2sess_node.c \ + acl/l2sess.h \ + acl/acl_plugin.api.h + +API_FILES += acl/acl.api + +nobase_apiinclude_HEADERS += \ + acl/acl_all_api_h.h \ + acl/acl_msg_enum.h \ + acl/acl.api.h + +acl_test_plugin_la_SOURCES = acl/acl_test.c acl/acl_plugin.api.h + +# vi:syntax=automake diff --git a/src/plugins/acl/acl.api b/src/plugins/acl/acl.api new file mode 100644 index 00000000..58a5a171 --- /dev/null +++ b/src/plugins/acl/acl.api @@ -0,0 +1,444 @@ +/* Hey Emacs use -*- mode: C -*- */ +/* + * Copyright (c) 2016 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +/** \file + This file defines the vpp control-plane API messages + used to control the ACL plugin +*/ + + +/** \brief Get the plugin version + @param client_index - opaque cookie to identify the sender + @param context - sender context, to match reply w/ request +*/ + +define acl_plugin_get_version +{ + u32 client_index; + u32 context; +}; + +/** \brief Reply to get the plugin version + @param context - returned sender context, to match reply w/ request + @param major - Incremented every time a known breaking behavior change is introduced + @param minor - Incremented with small changes, may be used to avoid buggy versions +*/ + +define acl_plugin_get_version_reply +{ + u32 context; + u32 major; + u32 minor; +}; + +/** \brief Access List Rule entry + @param is_permit - deny (0), permit (1), or permit+reflect(2) action on this rule. + @param is_ipv6 - IP addresses in this rule are IPv6 (1) or IPv4 (0) + @param src_ip_addr - Source prefix value + @param src_ip_prefix_len - Source prefix length + @param dst_ip_addr - Destination prefix value + @param dst_ip_prefix_len - Destination prefix length + @param proto - L4 protocol (http://www.iana.org/assignments/protocol-numbers/protocol-numbers.xhtml) + @param srcport_or_icmptype_first - beginning of source port or ICMP4/6 type range + @param srcport_or_icmptype_last - end of source port or ICMP4/6 type range + @param dstport_or_icmpcode_first - beginning of destination port or ICMP4/6 code range + @param dstport_or_icmpcode_last - end of destination port or ICMP4/6 code range + @param tcp_flags_mask - if proto==6, match masked TCP flags with this value + @param tcp_flags_value - if proto==6, mask to AND the TCP flags in the packet with +*/ + +typeonly manual_print manual_endian define acl_rule +{ + u8 is_permit; + u8 is_ipv6; + u8 src_ip_addr[16]; + u8 src_ip_prefix_len; + u8 dst_ip_addr[16]; + u8 dst_ip_prefix_len; +/* + * L4 protocol. IANA number. 1 = ICMP, 58 = ICMPv6, 6 = TCP, 17 = UDP. + * 0 => ignore L4 and ignore the ports/tcpflags when matching. + */ + u8 proto; +/* + * If the L4 protocol is TCP or UDP, the below + * hold ranges of ports, else if the L4 is ICMP/ICMPv6 + * they hold ranges of ICMP(v6) types/codes. + * + * Ranges are inclusive, i.e. to match "any" TCP/UDP port, + * use first=0,last=65535. For ICMP(v6), + * use first=0,last=255. + */ + u16 srcport_or_icmptype_first; + u16 srcport_or_icmptype_last; + u16 dstport_or_icmpcode_first; + u16 dstport_or_icmpcode_last; +/* + * for proto = 6, this matches if the + * TCP flags in the packet, ANDed with tcp_flags_mask, + * is equal to tcp_flags_value. + */ + u8 tcp_flags_mask; + u8 tcp_flags_value; +}; + +/** \brief MACIP Access List Rule entry + @param is_permit - deny (0), permit (1) action on this rule. + @param is_ipv6 - IP addresses in this rule are IPv6 (1) or IPv4 (0) + @param src_mac - match masked source MAC address against this value + @param src_mac_mask - AND source MAC address with this value before matching + @param src_ip_addr - Source prefix value + @param src_ip_prefix_len - Source prefix length +*/ + +typeonly manual_print manual_endian define macip_acl_rule +{ + u8 is_permit; + u8 is_ipv6; +/* + * The source mac of the packet ANDed with src_mac_mask. + * The source ip[46] address in the packet is matched + * against src_ip_addr, with src_ip_prefix_len set to 0. + * + * For better performance, minimize the number of + * (src_mac_mask, src_ip_prefix_len) combinations + * in a MACIP ACL. + */ + u8 src_mac[6]; + u8 src_mac_mask[6]; + u8 src_ip_addr[16]; + u8 src_ip_prefix_len; +}; + +/** \brief Replace an existing ACL in-place or create a new ACL + @param client_index - opaque cookie to identify the sender + @param context - sender context, to match reply w/ request + @param acl_index - an existing ACL entry (0..0xfffffffe) to replace, or 0xffffffff to make new ACL + @param tag - a string value stored along with the ACL, for descriptive purposes + @param count - number of ACL rules + @r - Rules for this access-list +*/ + +manual_print manual_endian define acl_add_replace +{ + u32 client_index; + u32 context; + u32 acl_index; /* ~0 to add, existing ACL# to replace */ + u8 tag[64]; /* What gets in here gets out in the corresponding tag field when dumping the ACLs. */ + u32 count; + vl_api_acl_rule_t r[count]; +}; + +/** \brief Reply to add/replace ACL + @param context - returned sender context, to match reply w/ request + @param acl_index - index of the updated or newly created ACL + @param retval 0 - no error +*/ + +define acl_add_replace_reply +{ + u32 context; + u32 acl_index; + i32 retval; +}; + +/** \brief Delete an ACL + @param client_index - opaque cookie to identify the sender + @param context - sender context, to match reply w/ request + @param acl_index - ACL index to delete +*/ + +define acl_del +{ + u32 client_index; + u32 context; + u32 acl_index; +}; + +/** \brief Reply to delete the ACL + @param context - returned sender context, to match reply w/ request + @param retval 0 - no error +*/ + +define acl_del_reply +{ + u32 context; + i32 retval; +}; + +/* acl_interface_add_del(_reply) to be deprecated in lieu of acl_interface_set_acl_list */ +/** \brief Use acl_interface_set_acl_list instead + Append/remove an ACL index to/from the list of ACLs checked for an interface + @param client_index - opaque cookie to identify the sender + @param context - sender context, to match reply w/ request + @param is_add - add or delete the ACL index from the list + @param is_input - check the ACL on input (1) or output (0) + @param sw_if_index - the interface to alter the list of ACLs on + @param acl_index - index of ACL for the operation +*/ + +define acl_interface_add_del +{ + u32 client_index; + u32 context; + u8 is_add; +/* + * is_input = 0 => ACL applied on interface egress + * is_input = 1 => ACL applied on interface ingress + */ + u8 is_input; + u32 sw_if_index; + u32 acl_index; +}; + +/** \brief Reply to alter the ACL list + @param context - returned sender context, to match reply w/ request + @param retval 0 - no error +*/ + +define acl_interface_add_del_reply +{ + u32 context; + i32 retval; +}; + +/** \brief Set the vector of input/output ACLs checked for an interface + @param client_index - opaque cookie to identify the sender + @param context - sender context, to match reply w/ request + @param sw_if_index - the interface to alter the list of ACLs on + @param count - total number of ACL indices in the vector + @param n_input - this many first elements correspond to input ACLs, the rest - output + @param acls - vector of ACL indices +*/ + +manual_endian define acl_interface_set_acl_list +{ + u32 client_index; + u32 context; + u32 sw_if_index; + u8 count; + u8 n_input; /* First n_input ACLs are set as a list of input ACLs, the rest are applied as output */ + u32 acls[count]; +}; + +/** \brief Reply to set the ACL list on an interface + @param context - returned sender context, to match reply w/ request + @param retval 0 - no error +*/ + +define acl_interface_set_acl_list_reply +{ + u32 context; + i32 retval; +}; + +/** \brief Dump the specific ACL contents or all of the ACLs' contents + @param client_index - opaque cookie to identify the sender + @param context - sender context, to match reply w/ request + @param acl_index - ACL index to dump, ~0 to dump all ACLs +*/ + +define acl_dump +{ + u32 client_index; + u32 context; + u32 acl_index; /* ~0 for all ACLs */ +}; + +/** \brief Details about a single ACL contents + @param context - returned sender context, to match reply w/ request + @param acl_index - ACL index whose contents are being sent in this message + @param tag - Descriptive tag value which was supplied at ACL creation + @param count - Number of rules in this ACL + @param r - Array of rules within this ACL +*/ + +manual_print manual_endian define acl_details +{ + u32 context; + u32 acl_index; + u8 tag[64]; /* Same blob that was supplied to us when creating the ACL, one hopes. */ + u32 count; + vl_api_acl_rule_t r[count]; +}; + +/** \brief Dump the list(s) of ACL applied to specific or all interfaces + @param client_index - opaque cookie to identify the sender + @param context - sender context, to match reply w/ request + @param sw_if_index - interface to dump the ACL list for +*/ + +define acl_interface_list_dump +{ + u32 client_index; + u32 context; + u32 sw_if_index; /* ~0 for all interfaces */ +}; + +/** \brief Details about a single ACL contents + @param context - returned sender context, to match reply w/ request + @param sw_if_index - interface for which the list of ACLs is applied + @param count - total length of acl indices vector + @param n_input - this many of indices in the beginning are input ACLs, the rest - output + @param acls - the vector of ACL indices +*/ + +manual_endian define acl_interface_list_details +{ + u32 context; + u32 sw_if_index; + u8 count; + u8 n_input; + u32 acls[count]; +}; + +/** \brief Add a MACIP ACL + @param client_index - opaque cookie to identify the sender + @param context - sender context, to match reply w/ request + @param tag - descriptive value for this MACIP ACL + @param count - number of rules in this ACL + @param r - vector of MACIP ACL rules +*/ + +manual_print manual_endian define macip_acl_add +{ + u32 client_index; + u32 context; + u8 tag[64]; + u32 count; + vl_api_macip_acl_rule_t r[count]; +}; + +/** \brief Reply to add MACIP ACL + @param context - returned sender context, to match reply w/ request + @param acl_index - index of the newly created ACL + @param retval 0 - no error +*/ + +define macip_acl_add_reply +{ + u32 context; + u32 acl_index; + i32 retval; +}; + +/** \brief Delete a MACIP ACL + @param client_index - opaque cookie to identify the sender + @param context - sender context, to match reply w/ request + @param acl_index - MACIP ACL index to delete +*/ + +define macip_acl_del +{ + u32 client_index; + u32 context; + u32 acl_index; +}; + +/** \brief Reply to delete the MACIP ACL + @param context - returned sender context, to match reply w/ request + @param retval 0 - no error +*/ + +define macip_acl_del_reply +{ + u32 context; + i32 retval; +}; + +/** \brief Add or delete a MACIP ACL to/from interface + @param client_index - opaque cookie to identify the sender + @param context - sender context, to match reply w/ request + @param is_add - add (1) or delete (0) ACL from being used on an interface + @param sw_if_index - interface to apply the action to + @param acl_index - MACIP ACL index +*/ + +define macip_acl_interface_add_del +{ + u32 client_index; + u32 context; + u8 is_add; + /* macip ACLs are always input */ + u32 sw_if_index; + u32 acl_index; +}; + +/** \brief Reply to apply/unapply the MACIP ACL + @param context - returned sender context, to match reply w/ request + @param retval 0 - no error +*/ + +define macip_acl_interface_add_del_reply +{ + u32 context; + i32 retval; +}; + +/** \brief Dump one or all defined MACIP ACLs + @param client_index - opaque cookie to identify the sender + @param context - sender context, to match reply w/ request + @param acl_index - MACIP ACL index or ~0 to dump all ACLs +*/ + +define macip_acl_dump +{ + u32 client_index; + u32 context; + u32 acl_index; /* ~0 for all ACLs */ +}; + +/** \brief Details about one MACIP ACL + @param context - returned sender context, to match reply w/ request + @param acl_index - index of this MACIP ACL + @param tag - descriptive tag which was supplied during the creation + @param count - length of the vector of MACIP ACL rules + @param r - rules comprising this ACL +*/ + +manual_print manual_endian define macip_acl_details +{ + u32 context; + u32 acl_index; + u8 tag[64]; + u32 count; + vl_api_macip_acl_rule_t r[count]; +}; + +/** \brief Get the vector of MACIP ACL IDs applied to the interfaces + @param client_index - opaque cookie to identify the sender + @param context - sender context, to match reply w/ request +*/ + +define macip_acl_interface_get +{ + u32 client_index; + u32 context; +}; + +/** \brief Reply with the vector of MACIP ACLs by sw_if_index + @param context - returned sender context, to match reply w/ request + @param count - total number of elements in the vector + @param acls - the vector of active MACACL indices per sw_if_index +*/ + +define macip_acl_interface_get_reply +{ + u32 context; + u32 count; + u32 acls[count]; +}; + diff --git a/src/plugins/acl/acl.c b/src/plugins/acl/acl.c new file mode 100644 index 00000000..8ff5a6b7 --- /dev/null +++ b/src/plugins/acl/acl.c @@ -0,0 +1,1901 @@ +/* + * Copyright (c) 2016 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +#include + +#include +#include +#include +#include + +#include +#include + +#include +#include +#include + +/* define message IDs */ +#include + +/* define message structures */ +#define vl_typedefs +#include +#undef vl_typedefs + +/* define generated endian-swappers */ +#define vl_endianfun +#include +#undef vl_endianfun + +/* instantiate all the print functions we know about */ +#define vl_print(handle, ...) vlib_cli_output (handle, __VA_ARGS__) +#define vl_printfun +#include +#undef vl_printfun + +/* Get the API version number */ +#define vl_api_version(n,v) static u32 api_version=(v); +#include +#undef vl_api_version + +#include "node_in.h" +#include "node_out.h" + +acl_main_t acl_main; + +/* + * A handy macro to set up a message reply. + * Assumes that the following variables are available: + * mp - pointer to request message + * rmp - pointer to reply message type + * rv - return value + */ + +#define REPLY_MACRO(t) \ +do { \ + unix_shared_memory_queue_t * q = \ + vl_api_client_index_to_input_queue (mp->client_index); \ + if (!q) \ + return; \ + \ + rmp = vl_msg_api_alloc (sizeof (*rmp)); \ + rmp->_vl_msg_id = ntohs((t)+sm->msg_id_base); \ + rmp->context = mp->context; \ + rmp->retval = ntohl(rv); \ + \ + vl_msg_api_send_shmem (q, (u8 *)&rmp); \ +} while(0); + +#define REPLY_MACRO2(t, body) \ +do { \ + unix_shared_memory_queue_t * q; \ + rv = vl_msg_api_pd_handler (mp, rv); \ + q = vl_api_client_index_to_input_queue (mp->client_index); \ + if (!q) \ + return; \ + \ + rmp = vl_msg_api_alloc (sizeof (*rmp)); \ + rmp->_vl_msg_id = ntohs((t)+am->msg_id_base); \ + rmp->context = mp->context; \ + rmp->retval = ntohl(rv); \ + do {body;} while (0); \ + vl_msg_api_send_shmem (q, (u8 *)&rmp); \ +} while(0); + +#define REPLY_MACRO3(t, n, body) \ +do { \ + unix_shared_memory_queue_t * q; \ + rv = vl_msg_api_pd_handler (mp, rv); \ + q = vl_api_client_index_to_input_queue (mp->client_index); \ + if (!q) \ + return; \ + \ + rmp = vl_msg_api_alloc (sizeof (*rmp) + n); \ + rmp->_vl_msg_id = ntohs((t)+am->msg_id_base); \ + rmp->context = mp->context; \ + rmp->retval = ntohl(rv); \ + do {body;} while (0); \ + vl_msg_api_send_shmem (q, (u8 *)&rmp); \ +} while(0); + + +/* List of message types that this plugin understands */ + +#define foreach_acl_plugin_api_msg \ +_(ACL_PLUGIN_GET_VERSION, acl_plugin_get_version) \ +_(ACL_ADD_REPLACE, acl_add_replace) \ +_(ACL_DEL, acl_del) \ +_(ACL_INTERFACE_ADD_DEL, acl_interface_add_del) \ +_(ACL_INTERFACE_SET_ACL_LIST, acl_interface_set_acl_list) \ +_(ACL_DUMP, acl_dump) \ +_(ACL_INTERFACE_LIST_DUMP, acl_interface_list_dump) \ +_(MACIP_ACL_ADD, macip_acl_add) \ +_(MACIP_ACL_DEL, macip_acl_del) \ +_(MACIP_ACL_INTERFACE_ADD_DEL, macip_acl_interface_add_del) \ +_(MACIP_ACL_DUMP, macip_acl_dump) \ +_(MACIP_ACL_INTERFACE_GET, macip_acl_interface_get) + +/* + * This routine exists to convince the vlib plugin framework that + * we haven't accidentally copied a random .dll into the plugin directory. + * + * Also collects global variable pointers passed from the vpp engine + */ + +clib_error_t * +vlib_plugin_register (vlib_main_t * vm, vnet_plugin_handoff_t * h, + int from_early_init) +{ + acl_main_t *am = &acl_main; + clib_error_t *error = 0; + + am->vlib_main = vm; + am->vnet_main = h->vnet_main; + am->ethernet_main = h->ethernet_main; + + l2sess_vlib_plugin_register(vm, h, from_early_init); + + return error; +} + + +static void +vl_api_acl_plugin_get_version_t_handler (vl_api_acl_plugin_get_version_t * mp) +{ + acl_main_t *am = &acl_main; + vl_api_acl_plugin_get_version_reply_t *rmp; + int msg_size = sizeof (*rmp); + unix_shared_memory_queue_t *q; + + q = vl_api_client_index_to_input_queue (mp->client_index); + if (q == 0) + { + return; + } + + rmp = vl_msg_api_alloc (msg_size); + memset (rmp, 0, msg_size); + rmp->_vl_msg_id = + ntohs (VL_API_ACL_PLUGIN_GET_VERSION_REPLY + am->msg_id_base); + rmp->context = mp->context; + rmp->major = htonl (ACL_PLUGIN_VERSION_MAJOR); + rmp->minor = htonl (ACL_PLUGIN_VERSION_MINOR); + + vl_msg_api_send_shmem (q, (u8 *) & rmp); +} + + +static int +acl_add_list (u32 count, vl_api_acl_rule_t rules[], + u32 * acl_list_index, u8 * tag) +{ + acl_main_t *am = &acl_main; + acl_list_t *a; + acl_rule_t *r; + acl_rule_t *acl_new_rules; + int i; + + if (*acl_list_index != ~0) + { + /* They supplied some number, let's see if this ACL exists */ + if (pool_is_free_index (am->acls, *acl_list_index)) + { + /* tried to replace a non-existent ACL, no point doing anything */ + return -1; + } + } + + /* Create and populate the rules */ + acl_new_rules = clib_mem_alloc_aligned (sizeof (acl_rule_t) * count, + CLIB_CACHE_LINE_BYTES); + if (!acl_new_rules) + { + /* Could not allocate rules. New or existing ACL - bail out regardless */ + return -1; + } + + for (i = 0; i < count; i++) + { + r = &acl_new_rules[i]; + r->is_permit = rules[i].is_permit; + r->is_ipv6 = rules[i].is_ipv6; + if (r->is_ipv6) + { + memcpy (&r->src, rules[i].src_ip_addr, sizeof (r->src)); + memcpy (&r->dst, rules[i].dst_ip_addr, sizeof (r->dst)); + } + else + { + memcpy (&r->src.ip4, rules[i].src_ip_addr, sizeof (r->src.ip4)); + memcpy (&r->dst.ip4, rules[i].dst_ip_addr, sizeof (r->dst.ip4)); + } + r->src_prefixlen = rules[i].src_ip_prefix_len; + r->dst_prefixlen = rules[i].dst_ip_prefix_len; + r->proto = rules[i].proto; + r->src_port_or_type_first = rules[i].srcport_or_icmptype_first; + r->src_port_or_type_last = rules[i].srcport_or_icmptype_last; + r->dst_port_or_code_first = rules[i].dstport_or_icmpcode_first; + r->dst_port_or_code_last = rules[i].dstport_or_icmpcode_last; + r->tcp_flags_value = rules[i].tcp_flags_value; + r->tcp_flags_mask = rules[i].tcp_flags_mask; + } + + if (~0 == *acl_list_index) + { + /* Get ACL index */ + pool_get_aligned (am->acls, a, CLIB_CACHE_LINE_BYTES); + memset (a, 0, sizeof (*a)); + /* Will return the newly allocated ACL index */ + *acl_list_index = a - am->acls; + } + else + { + a = am->acls + *acl_list_index; + /* Get rid of the old rules */ + clib_mem_free (a->rules); + } + a->rules = acl_new_rules; + a->count = count; + memcpy (a->tag, tag, sizeof (a->tag)); + + return 0; +} + +static int +acl_del_list (u32 acl_list_index) +{ + acl_main_t *am = &acl_main; + acl_list_t *a; + int i, ii; + if (pool_is_free_index (am->acls, acl_list_index)) + { + return -1; + } + + /* delete any references to the ACL */ + for (i = 0; i < vec_len (am->output_acl_vec_by_sw_if_index); i++) + { + for (ii = 0; ii < vec_len (am->output_acl_vec_by_sw_if_index[i]); + /* see body */ ) + { + if (acl_list_index == am->output_acl_vec_by_sw_if_index[i][ii]) + { + vec_del1 (am->output_acl_vec_by_sw_if_index[i], ii); + } + else + { + ii++; + } + } + } + for (i = 0; i < vec_len (am->input_acl_vec_by_sw_if_index); i++) + { + for (ii = 0; ii < vec_len (am->input_acl_vec_by_sw_if_index[i]); + /* see body */ ) + { + if (acl_list_index == am->input_acl_vec_by_sw_if_index[i][ii]) + { + vec_del1 (am->input_acl_vec_by_sw_if_index[i], ii); + } + else + { + ii++; + } + } + } + + /* now we can delete the ACL itself */ + a = &am->acls[acl_list_index]; + if (a->rules) + { + clib_mem_free (a->rules); + } + pool_put (am->acls, a); + return 0; +} + +/* Some aids in ASCII graphing the content */ +#define XX "\377" +#define __ "\000" +#define _(x) +#define v + +u8 ip4_5tuple_mask[] = +_(" dmac smac etype ") +_(ether) __ __ __ __ __ __ v __ __ __ __ __ __ v __ __ v + _(" v ihl totlen ") + _(0x0000) + __ __ __ __ + _(" ident fl+fo ") + _(0x0004) + __ __ __ __ + _(" ttl pr checksum ") + _(0x0008) + __ XX __ __ + _(" src address ") + _(0x000C) + XX XX XX XX + _(" dst address ") + _(0x0010) + XX XX XX XX + _("L4 T/U sport dport ") + _(tcpudp) + XX XX XX XX + _(padpad) + __ __ __ __ + _(padpad) + __ __ __ __ + _(padeth) + __ __; + + u8 ip6_5tuple_mask[] = + _(" dmac smac etype ") + _(ether) __ __ __ __ __ __ v __ __ __ __ __ __ v __ __ v + _(" v tc + flow ") + _(0x0000) __ __ __ __ + _(" plen nh hl ") + _(0x0004) __ __ XX __ + _(" src address ") + _(0x0008) XX XX XX XX + _(0x000C) XX XX XX XX + _(0x0010) XX XX XX XX + _(0x0014) XX XX XX XX + _(" dst address ") + _(0x0018) XX XX XX XX + _(0x001C) XX XX XX XX + _(0x0020) XX XX XX XX + _(0x0024) XX XX XX XX + _("L4T/U sport dport ") + _(tcpudp) XX XX XX XX _(padpad) __ __ __ __ _(padeth) __ __; + +#undef XX +#undef __ +#undef _ +#undef v + + static int count_skip (u8 * p, u32 size) +{ + u64 *p64 = (u64 *) p; + /* Be tolerant to null pointer */ + if (0 == p) + return 0; + + while ((0ULL == *p64) && ((u8 *) p64 - p) < size) + { + p64++; + } + return (p64 - (u64 *) p) / 2; +} + +static int +acl_classify_add_del_table_big (vnet_classify_main_t * cm, u8 * mask, + u32 mask_len, u32 next_table_index, + u32 miss_next_index, u32 * table_index, + int is_add) +{ + u32 nbuckets = 65536; + u32 memory_size = 2 << 30; + u32 skip = count_skip (mask, mask_len); + u32 match = (mask_len / 16) - skip; + u8 *skip_mask_ptr = mask + 16 * skip; + u32 current_data_flag = 0; + int current_data_offset = 0; + + if (0 == match) + match = 1; + + return vnet_classify_add_del_table (cm, skip_mask_ptr, nbuckets, + memory_size, skip, match, + next_table_index, miss_next_index, + table_index, current_data_flag, + current_data_offset, is_add, + 1 /* delete_chain */); +} + +static int +acl_classify_add_del_table_small (vnet_classify_main_t * cm, u8 * mask, + u32 mask_len, u32 next_table_index, + u32 miss_next_index, u32 * table_index, + int is_add) +{ + u32 nbuckets = 32; + u32 memory_size = 2 << 20; + u32 skip = count_skip (mask, mask_len); + u32 match = (mask_len / 16) - skip; + u8 *skip_mask_ptr = mask + 16 * skip; + u32 current_data_flag = 0; + int current_data_offset = 0; + + if (0 == match) + match = 1; + + return vnet_classify_add_del_table (cm, skip_mask_ptr, nbuckets, + memory_size, skip, match, + next_table_index, miss_next_index, + table_index, current_data_flag, + current_data_offset, is_add, + 1 /* delete_chain */); +} + + +static int +acl_unhook_l2_input_classify (acl_main_t * am, u32 sw_if_index) +{ + vnet_classify_main_t *cm = &vnet_classify_main; + u32 ip4_table_index = ~0; + u32 ip6_table_index = ~0; + + vec_validate_init_empty (am->acl_ip4_input_classify_table_by_sw_if_index, + sw_if_index, ~0); + vec_validate_init_empty (am->acl_ip6_input_classify_table_by_sw_if_index, + sw_if_index, ~0); + + vnet_l2_input_classify_enable_disable (sw_if_index, 0); + + if (am->acl_ip4_input_classify_table_by_sw_if_index[sw_if_index] != ~0) + { + ip4_table_index = + am->acl_ip4_input_classify_table_by_sw_if_index[sw_if_index]; + am->acl_ip4_input_classify_table_by_sw_if_index[sw_if_index] = ~0; + acl_classify_add_del_table_big (cm, ip4_5tuple_mask, + sizeof (ip4_5tuple_mask) - 1, ~0, + am->l2_input_classify_next_acl, + &ip4_table_index, 0); + } + if (am->acl_ip6_input_classify_table_by_sw_if_index[sw_if_index] != ~0) + { + ip6_table_index = + am->acl_ip6_input_classify_table_by_sw_if_index[sw_if_index]; + am->acl_ip6_input_classify_table_by_sw_if_index[sw_if_index] = ~0; + acl_classify_add_del_table_big (cm, ip6_5tuple_mask, + sizeof (ip6_5tuple_mask) - 1, ~0, + am->l2_input_classify_next_acl, + &ip6_table_index, 0); + } + + return 0; +} + +static int +acl_unhook_l2_output_classify (acl_main_t * am, u32 sw_if_index) +{ + vnet_classify_main_t *cm = &vnet_classify_main; + u32 ip4_table_index = ~0; + u32 ip6_table_index = ~0; + + vec_validate_init_empty (am->acl_ip4_output_classify_table_by_sw_if_index, + sw_if_index, ~0); + vec_validate_init_empty (am->acl_ip6_output_classify_table_by_sw_if_index, + sw_if_index, ~0); + + vnet_l2_output_classify_enable_disable (sw_if_index, 0); + + if (am->acl_ip4_output_classify_table_by_sw_if_index[sw_if_index] != ~0) + { + ip4_table_index = + am->acl_ip4_output_classify_table_by_sw_if_index[sw_if_index]; + am->acl_ip4_output_classify_table_by_sw_if_index[sw_if_index] = ~0; + acl_classify_add_del_table_big (cm, ip4_5tuple_mask, + sizeof (ip4_5tuple_mask) - 1, ~0, + am->l2_output_classify_next_acl, + &ip4_table_index, 0); + } + if (am->acl_ip6_output_classify_table_by_sw_if_index[sw_if_index] != ~0) + { + ip6_table_index = + am->acl_ip6_output_classify_table_by_sw_if_index[sw_if_index]; + am->acl_ip6_output_classify_table_by_sw_if_index[sw_if_index] = ~0; + acl_classify_add_del_table_big (cm, ip6_5tuple_mask, + sizeof (ip6_5tuple_mask) - 1, ~0, + am->l2_output_classify_next_acl, + &ip6_table_index, 0); + } + + return 0; +} + +static int +acl_hook_l2_input_classify (acl_main_t * am, u32 sw_if_index) +{ + vnet_classify_main_t *cm = &vnet_classify_main; + u32 ip4_table_index = ~0; + u32 ip6_table_index = ~0; + int rv; + + /* in case there were previous tables attached */ + acl_unhook_l2_input_classify (am, sw_if_index); + rv = + acl_classify_add_del_table_big (cm, ip4_5tuple_mask, + sizeof (ip4_5tuple_mask) - 1, ~0, + am->l2_input_classify_next_acl, + &ip4_table_index, 1); + if (rv) + return rv; + rv = + acl_classify_add_del_table_big (cm, ip6_5tuple_mask, + sizeof (ip6_5tuple_mask) - 1, ~0, + am->l2_input_classify_next_acl, + &ip6_table_index, 1); + if (rv) + { + acl_classify_add_del_table_big (cm, ip4_5tuple_mask, + sizeof (ip4_5tuple_mask) - 1, ~0, + am->l2_input_classify_next_acl, + &ip4_table_index, 0); + return rv; + } + rv = + vnet_l2_input_classify_set_tables (sw_if_index, ip4_table_index, + ip6_table_index, ~0); + clib_warning + ("ACL enabling on interface sw_if_index %d, setting tables to the following: ip4: %d ip6: %d\n", + sw_if_index, ip4_table_index, ip6_table_index); + if (rv) + { + acl_classify_add_del_table_big (cm, ip6_5tuple_mask, + sizeof (ip6_5tuple_mask) - 1, ~0, + am->l2_input_classify_next_acl, + &ip6_table_index, 0); + acl_classify_add_del_table_big (cm, ip4_5tuple_mask, + sizeof (ip4_5tuple_mask) - 1, ~0, + am->l2_input_classify_next_acl, + &ip4_table_index, 0); + return rv; + } + + am->acl_ip4_input_classify_table_by_sw_if_index[sw_if_index] = + ip4_table_index; + am->acl_ip6_input_classify_table_by_sw_if_index[sw_if_index] = + ip6_table_index; + + vnet_l2_input_classify_enable_disable (sw_if_index, 1); + return rv; +} + +static int +acl_hook_l2_output_classify (acl_main_t * am, u32 sw_if_index) +{ + vnet_classify_main_t *cm = &vnet_classify_main; + u32 ip4_table_index = ~0; + u32 ip6_table_index = ~0; + int rv; + + /* in case there were previous tables attached */ + acl_unhook_l2_output_classify (am, sw_if_index); + rv = + acl_classify_add_del_table_big (cm, ip4_5tuple_mask, + sizeof (ip4_5tuple_mask) - 1, ~0, + am->l2_output_classify_next_acl, + &ip4_table_index, 1); + if (rv) + return rv; + rv = + acl_classify_add_del_table_big (cm, ip6_5tuple_mask, + sizeof (ip6_5tuple_mask) - 1, ~0, + am->l2_output_classify_next_acl, + &ip6_table_index, 1); + if (rv) + { + acl_classify_add_del_table_big (cm, ip4_5tuple_mask, + sizeof (ip4_5tuple_mask) - 1, ~0, + am->l2_output_classify_next_acl, + &ip4_table_index, 0); + return rv; + } + rv = + vnet_l2_output_classify_set_tables (sw_if_index, ip4_table_index, + ip6_table_index, ~0); + clib_warning + ("ACL enabling on interface sw_if_index %d, setting tables to the following: ip4: %d ip6: %d\n", + sw_if_index, ip4_table_index, ip6_table_index); + if (rv) + { + acl_classify_add_del_table_big (cm, ip6_5tuple_mask, + sizeof (ip6_5tuple_mask) - 1, ~0, + am->l2_output_classify_next_acl, + &ip6_table_index, 0); + acl_classify_add_del_table_big (cm, ip4_5tuple_mask, + sizeof (ip4_5tuple_mask) - 1, ~0, + am->l2_output_classify_next_acl, + &ip4_table_index, 0); + return rv; + } + + am->acl_ip4_output_classify_table_by_sw_if_index[sw_if_index] = + ip4_table_index; + am->acl_ip6_output_classify_table_by_sw_if_index[sw_if_index] = + ip6_table_index; + + vnet_l2_output_classify_enable_disable (sw_if_index, 1); + return rv; +} + + +int +acl_interface_in_enable_disable (acl_main_t * am, u32 sw_if_index, + int enable_disable) +{ + int rv; + + /* Utterly wrong? */ + if (pool_is_free_index (am->vnet_main->interface_main.sw_interfaces, + sw_if_index)) + return VNET_API_ERROR_INVALID_SW_IF_INDEX; + + if (enable_disable) + { + rv = acl_hook_l2_input_classify (am, sw_if_index); + } + else + { + rv = acl_unhook_l2_input_classify (am, sw_if_index); + } + + return rv; +} + +int +acl_interface_out_enable_disable (acl_main_t * am, u32 sw_if_index, + int enable_disable) +{ + int rv; + + /* Utterly wrong? */ + if (pool_is_free_index (am->vnet_main->interface_main.sw_interfaces, + sw_if_index)) + return VNET_API_ERROR_INVALID_SW_IF_INDEX; + + if (enable_disable) + { + rv = acl_hook_l2_output_classify (am, sw_if_index); + } + else + { + rv = acl_unhook_l2_output_classify (am, sw_if_index); + } + + return rv; +} + + +static int +acl_interface_add_inout_acl (u32 sw_if_index, u8 is_input, u32 acl_list_index) +{ + acl_main_t *am = &acl_main; + if (is_input) + { + vec_validate (am->input_acl_vec_by_sw_if_index, sw_if_index); + vec_add (am->input_acl_vec_by_sw_if_index[sw_if_index], &acl_list_index, + 1); + acl_interface_in_enable_disable (am, sw_if_index, 1); + } + else + { + vec_validate (am->output_acl_vec_by_sw_if_index, sw_if_index); + vec_add (am->output_acl_vec_by_sw_if_index[sw_if_index], + &acl_list_index, 1); + acl_interface_out_enable_disable (am, sw_if_index, 1); + } + return 0; +} + +static int +acl_interface_del_inout_acl (u32 sw_if_index, u8 is_input, u32 acl_list_index) +{ + acl_main_t *am = &acl_main; + int i; + int rv = -1; + if (is_input) + { + vec_validate (am->input_acl_vec_by_sw_if_index, sw_if_index); + for (i = 0; i < vec_len (am->input_acl_vec_by_sw_if_index[sw_if_index]); + i++) + { + if (acl_list_index == + am->input_acl_vec_by_sw_if_index[sw_if_index][i]) + { + vec_del1 (am->input_acl_vec_by_sw_if_index[sw_if_index], i); + rv = 0; + break; + } + } + if (0 == vec_len (am->input_acl_vec_by_sw_if_index[sw_if_index])) + { + acl_interface_in_enable_disable (am, sw_if_index, 0); + } + } + else + { + vec_validate (am->output_acl_vec_by_sw_if_index, sw_if_index); + for (i = 0; + i < vec_len (am->output_acl_vec_by_sw_if_index[sw_if_index]); i++) + { + if (acl_list_index == + am->output_acl_vec_by_sw_if_index[sw_if_index][i]) + { + vec_del1 (am->output_acl_vec_by_sw_if_index[sw_if_index], i); + rv = 0; + break; + } + } + if (0 == vec_len (am->output_acl_vec_by_sw_if_index[sw_if_index])) + { + acl_interface_out_enable_disable (am, sw_if_index, 0); + } + } + return rv; +} + +static void +acl_interface_reset_inout_acls (u32 sw_if_index, u8 is_input) +{ + acl_main_t *am = &acl_main; + if (is_input) + { + acl_interface_in_enable_disable (am, sw_if_index, 0); + vec_validate (am->input_acl_vec_by_sw_if_index, sw_if_index); + vec_reset_length (am->input_acl_vec_by_sw_if_index[sw_if_index]); + } + else + { + acl_interface_out_enable_disable (am, sw_if_index, 0); + vec_validate (am->output_acl_vec_by_sw_if_index, sw_if_index); + vec_reset_length (am->output_acl_vec_by_sw_if_index[sw_if_index]); + } +} + +static int +acl_interface_add_del_inout_acl (u32 sw_if_index, u8 is_add, u8 is_input, + u32 acl_list_index) +{ + int rv = -1; + if (is_add) + { + rv = + acl_interface_add_inout_acl (sw_if_index, is_input, acl_list_index); + } + else + { + rv = + acl_interface_del_inout_acl (sw_if_index, is_input, acl_list_index); + } + return rv; +} + + +static void * +get_ptr_to_offset (vlib_buffer_t * b0, int offset) +{ + u8 *p = vlib_buffer_get_current (b0) + offset; + return p; +} + +static u8 +acl_get_l4_proto (vlib_buffer_t * b0, int node_is_ip6) +{ + u8 proto; + int proto_offset; + if (node_is_ip6) + { + proto_offset = 20; + } + else + { + proto_offset = 23; + } + proto = *((u8 *) vlib_buffer_get_current (b0) + proto_offset); + return proto; +} + +static int +acl_match_addr (ip46_address_t * addr1, ip46_address_t * addr2, int prefixlen, + int is_ip6) +{ + if (prefixlen == 0) + { + /* match any always succeeds */ + return 1; + } + if (is_ip6) + { + if (memcmp (addr1, addr2, prefixlen / 8)) + { + /* If the starting full bytes do not match, no point in bittwidling the thumbs further */ + return 0; + } + if (prefixlen % 8) + { + u8 b1 = *((u8 *) addr1 + 1 + prefixlen / 8); + u8 b2 = *((u8 *) addr2 + 1 + prefixlen / 8); + u8 mask0 = (0xff - ((1 << (8 - (prefixlen % 8))) - 1)); + return (b1 & mask0) == b2; + } + else + { + /* The prefix fits into integer number of bytes, so nothing left to do */ + return 1; + } + } + else + { + uint32_t a1 = ntohl (addr1->ip4.as_u32); + uint32_t a2 = ntohl (addr2->ip4.as_u32); + uint32_t mask0 = 0xffffffff - ((1 << (32 - prefixlen)) - 1); + return (a1 & mask0) == a2; + } +} + +static int +acl_match_port (u16 port, u16 port_first, u16 port_last, int is_ip6) +{ + return ((port >= port_first) && (port <= port_last)); +} + +static int +acl_packet_match (acl_main_t * am, u32 acl_index, vlib_buffer_t * b0, + u8 * r_action, int *r_is_ip6, u32 * r_acl_match_p, + u32 * r_rule_match_p, u32 * trace_bitmap) +{ + ethernet_header_t *h0; + u16 type0; + + ip46_address_t src, dst; + int is_ip6; + int is_ip4; + u8 proto; + u16 src_port; + u16 dst_port; + u8 tcp_flags = 0; + int i; + acl_list_t *a; + acl_rule_t *r; + + h0 = vlib_buffer_get_current (b0); + type0 = clib_net_to_host_u16 (h0->type); + is_ip4 = (type0 == ETHERNET_TYPE_IP4); + is_ip6 = (type0 == ETHERNET_TYPE_IP6); + + if (!(is_ip4 || is_ip6)) + { + return 0; + } + /* The bunch of hardcoded offsets here is intentional to get rid of them + ASAP, when getting to a faster matching code */ + if (is_ip4) + { + clib_memcpy (&src.ip4, get_ptr_to_offset (b0, 26), 4); + clib_memcpy (&dst.ip4, get_ptr_to_offset (b0, 30), 4); + proto = acl_get_l4_proto (b0, 0); + if (1 == proto) + { + *trace_bitmap |= 0x00000001; + /* type */ + src_port = *(u8 *) get_ptr_to_offset (b0, 34); + /* code */ + dst_port = *(u8 *) get_ptr_to_offset (b0, 35); + } + else + { + /* assume TCP/UDP */ + src_port = (*(u16 *) get_ptr_to_offset (b0, 34)); + dst_port = (*(u16 *) get_ptr_to_offset (b0, 36)); + /* UDP gets ability to check on an oddball data byte as a bonus */ + tcp_flags = *(u8 *) get_ptr_to_offset (b0, 14 + 20 + 13); + } + } + else /* is_ipv6 implicitly */ + { + clib_memcpy (&src, get_ptr_to_offset (b0, 22), 16); + clib_memcpy (&dst, get_ptr_to_offset (b0, 38), 16); + proto = acl_get_l4_proto (b0, 1); + if (58 == proto) + { + *trace_bitmap |= 0x00000002; + /* type */ + src_port = *(u8 *) get_ptr_to_offset (b0, 54); + /* code */ + dst_port = *(u8 *) get_ptr_to_offset (b0, 55); + } + else + { + /* assume TCP/UDP */ + src_port = (*(u16 *) get_ptr_to_offset (b0, 54)); + dst_port = (*(u16 *) get_ptr_to_offset (b0, 56)); + tcp_flags = *(u8 *) get_ptr_to_offset (b0, 14 + 40 + 13); + } + } + if (pool_is_free_index (am->acls, acl_index)) + { + if (r_acl_match_p) + *r_acl_match_p = acl_index; + if (r_rule_match_p) + *r_rule_match_p = -1; + /* the ACL does not exist but is used for policy. Block traffic. */ + return 0; + } + a = am->acls + acl_index; + for (i = 0; i < a->count; i++) + { + r = a->rules + i; + if (is_ip6 != r->is_ipv6) + { + continue; + } + if (!acl_match_addr (&dst, &r->dst, r->dst_prefixlen, is_ip6)) + continue; + if (!acl_match_addr (&src, &r->src, r->src_prefixlen, is_ip6)) + continue; + if (r->proto) + { + if (proto != r->proto) + continue; + if (!acl_match_port + (src_port, r->src_port_or_type_first, r->src_port_or_type_last, + is_ip6)) + continue; + if (!acl_match_port + (dst_port, r->dst_port_or_code_first, r->dst_port_or_code_last, + is_ip6)) + continue; + /* No need for check of proto == TCP, since in other rules both fields should be zero, so this match will succeed */ + if ((tcp_flags & r->tcp_flags_mask) != r->tcp_flags_value) + continue; + } + /* everything matches! */ + *r_action = r->is_permit; + *r_is_ip6 = is_ip6; + if (r_acl_match_p) + *r_acl_match_p = acl_index; + if (r_rule_match_p) + *r_rule_match_p = i; + return 1; + } + return 0; +} + +void +input_acl_packet_match (u32 sw_if_index, vlib_buffer_t * b0, u32 * nextp, + u32 * acl_match_p, u32 * rule_match_p, + u32 * trace_bitmap) +{ + acl_main_t *am = &acl_main; + uint8_t action = 0; + int is_ip6 = 0; + int i; + vec_validate (am->input_acl_vec_by_sw_if_index, sw_if_index); + for (i = 0; i < vec_len (am->input_acl_vec_by_sw_if_index[sw_if_index]); + i++) + { + if (acl_packet_match + (am, am->input_acl_vec_by_sw_if_index[sw_if_index][i], b0, &action, + &is_ip6, acl_match_p, rule_match_p, trace_bitmap)) + { + if (is_ip6) + { + *nextp = am->acl_in_ip6_match_next[action]; + } + else + { + *nextp = am->acl_in_ip4_match_next[action]; + } + return; + } + } + if (vec_len (am->input_acl_vec_by_sw_if_index[sw_if_index]) > 0) + { + /* If there are ACLs and none matched, deny by default */ + *nextp = 0; + } + +} + +void +output_acl_packet_match (u32 sw_if_index, vlib_buffer_t * b0, u32 * nextp, + u32 * acl_match_p, u32 * rule_match_p, + u32 * trace_bitmap) +{ + acl_main_t *am = &acl_main; + uint8_t action = 0; + int is_ip6 = 0; + int i; + vec_validate (am->output_acl_vec_by_sw_if_index, sw_if_index); + for (i = 0; i < vec_len (am->output_acl_vec_by_sw_if_index[sw_if_index]); + i++) + { + if (acl_packet_match + (am, am->output_acl_vec_by_sw_if_index[sw_if_index][i], b0, &action, + &is_ip6, acl_match_p, rule_match_p, trace_bitmap)) + { + if (is_ip6) + { + *nextp = am->acl_out_ip6_match_next[action]; + } + else + { + *nextp = am->acl_out_ip4_match_next[action]; + } + return; + } + } + if (vec_len (am->output_acl_vec_by_sw_if_index[sw_if_index]) > 0) + { + /* If there are ACLs and none matched, deny by default */ + *nextp = 0; + } +} + +typedef struct +{ + u8 is_ipv6; + u8 mac_mask[6]; + u8 prefix_len; + u32 count; + u32 table_index; +} macip_match_type_t; + +static u32 +macip_find_match_type (macip_match_type_t * mv, u8 * mac_mask, u8 prefix_len, + u8 is_ipv6) +{ + u32 i; + if (mv) + { + for (i = 0; i < vec_len (mv); i++) + { + if ((mv[i].prefix_len == prefix_len) && (mv[i].is_ipv6 == is_ipv6) + && (0 == memcmp (mv[i].mac_mask, mac_mask, 6))) + { + return i; + } + } + } + return ~0; +} + + +/* Get metric used to sort match types. + The more specific and the more often seen - the bigger the metric */ +static int +match_type_metric (macip_match_type_t * m) +{ + /* FIXME: count the ones in the MAC mask as well, check how well this heuristic works in real life */ + return m->prefix_len + m->is_ipv6 + 10 * m->count; +} + +static int +match_type_compare (macip_match_type_t * m1, macip_match_type_t * m2) +{ + /* Ascending sort based on the metric values */ + return match_type_metric (m1) - match_type_metric (m2); +} + +/* Get the offset of L3 source within ethernet packet */ +static int +get_l3_src_offset(int is6) +{ + if(is6) + return (sizeof(ethernet_header_t) + offsetof(ip6_header_t, src_address)); + else + return (sizeof(ethernet_header_t) + offsetof(ip4_header_t, src_address)); +} + +static int +macip_create_classify_tables (acl_main_t * am, u32 macip_acl_index) +{ + macip_match_type_t *mvec = NULL; + macip_match_type_t *mt; + macip_acl_list_t *a = &am->macip_acls[macip_acl_index]; + int i; + u32 match_type_index; + u32 last_table; + u8 mask[5 * 16]; + vnet_classify_main_t *cm = &vnet_classify_main; + + /* Count the number of different types of rules */ + for (i = 0; i < a->count; i++) + { + if (~0 == + (match_type_index = + macip_find_match_type (mvec, a->rules[i].src_mac_mask, + a->rules[i].src_prefixlen, + a->rules[i].is_ipv6))) + { + match_type_index = vec_len (mvec); + vec_validate (mvec, match_type_index); + memcpy (mvec[match_type_index].mac_mask, + a->rules[match_type_index].src_mac_mask, 6); + mvec[match_type_index].prefix_len = a->rules[i].src_prefixlen; + mvec[match_type_index].is_ipv6 = a->rules[i].is_ipv6; + mvec[match_type_index].table_index = ~0; + } + mvec[match_type_index].count++; + } + /* Put the most frequently used tables last in the list so we can create classifier tables in reverse order */ + vec_sort_with_function (mvec, match_type_compare); + /* Create the classifier tables */ + last_table = ~0; + vec_foreach (mt, mvec) + { + int mask_len; + int is6 = mt->is_ipv6; + int l3_src_offs = get_l3_src_offset(is6); + memset (mask, 0, sizeof (mask)); + memcpy (&mask[6], mt->mac_mask, 6); + for (i = 0; i < (mt->prefix_len / 8); i++) + { + mask[l3_src_offs + i] = 0xff; + } + if (mt->prefix_len % 8) + { + mask[l3_src_offs + (mt->prefix_len / 8)] = + 0xff - ((1 << (8 - mt->prefix_len % 8)) - 1); + } + /* + * Round-up the number of bytes needed to store the prefix, + * and round up the number of vectors too + */ + mask_len = ((l3_src_offs + ((mt->prefix_len+7) / 8) + + (sizeof (u32x4)-1))/sizeof(u32x4)) * sizeof (u32x4); + acl_classify_add_del_table_small (cm, mask, mask_len, last_table, + (~0 == last_table) ? 0 : ~0, &mt->table_index, + 1); + last_table = mt->table_index; + } + a->ip4_table_index = ~0; + a->ip6_table_index = ~0; + a->l2_table_index = last_table; + + /* Populate the classifier tables with rules from the MACIP ACL */ + for (i = 0; i < a->count; i++) + { + u32 action = 0; + u32 metadata = 0; + int is6 = a->rules[i].is_ipv6; + int l3_src_offs = get_l3_src_offset(is6); + memset (mask, 0, sizeof (mask)); + memcpy (&mask[6], a->rules[i].src_mac, 6); + if (is6) + { + memcpy (&mask[l3_src_offs], &a->rules[i].src_ip_addr.ip6, 16); + } + else + { + memcpy (&mask[l3_src_offs], &a->rules[i].src_ip_addr.ip4, 4); + } + match_type_index = + macip_find_match_type (mvec, a->rules[i].src_mac_mask, + a->rules[i].src_prefixlen, + a->rules[i].is_ipv6); + /* add session to table mvec[match_type_index].table_index; */ + vnet_classify_add_del_session (cm, mvec[match_type_index].table_index, + mask, a->rules[i].is_permit ? ~0 : 0, i, + 0, action, metadata, 1); + } + return 0; +} + +static void +macip_destroy_classify_tables (acl_main_t * am, u32 macip_acl_index) +{ + vnet_classify_main_t *cm = &vnet_classify_main; + macip_acl_list_t *a = &am->macip_acls[macip_acl_index]; + + if (a->ip4_table_index != ~0) + { + acl_classify_add_del_table_small (cm, 0, ~0, ~0, ~0, &a->ip4_table_index, 0); + a->ip4_table_index = ~0; + } + if (a->ip6_table_index != ~0) + { + acl_classify_add_del_table_small (cm, 0, ~0, ~0, ~0, &a->ip6_table_index, 0); + a->ip6_table_index = ~0; + } + if (a->l2_table_index != ~0) + { + acl_classify_add_del_table_small (cm, 0, ~0, ~0, ~0, &a->l2_table_index, 0); + a->l2_table_index = ~0; + } +} + +static int +macip_acl_add_list (u32 count, vl_api_macip_acl_rule_t rules[], + u32 * acl_list_index, u8 * tag) +{ + acl_main_t *am = &acl_main; + macip_acl_list_t *a; + macip_acl_rule_t *r; + macip_acl_rule_t *acl_new_rules; + int i; + + /* Create and populate the rules */ + acl_new_rules = clib_mem_alloc_aligned (sizeof (macip_acl_rule_t) * count, + CLIB_CACHE_LINE_BYTES); + if (!acl_new_rules) + { + /* Could not allocate rules. New or existing ACL - bail out regardless */ + return -1; + } + + for (i = 0; i < count; i++) + { + r = &acl_new_rules[i]; + r->is_permit = rules[i].is_permit; + r->is_ipv6 = rules[i].is_ipv6; + memcpy (&r->src_mac, rules[i].src_mac, 6); + memcpy (&r->src_mac_mask, rules[i].src_mac_mask, 6); + if(rules[i].is_ipv6) + memcpy (&r->src_ip_addr.ip6, rules[i].src_ip_addr, 16); + else + memcpy (&r->src_ip_addr.ip4, rules[i].src_ip_addr, 4); + r->src_prefixlen = rules[i].src_ip_prefix_len; + } + + /* Get ACL index */ + pool_get_aligned (am->macip_acls, a, CLIB_CACHE_LINE_BYTES); + memset (a, 0, sizeof (*a)); + /* Will return the newly allocated ACL index */ + *acl_list_index = a - am->macip_acls; + + a->rules = acl_new_rules; + a->count = count; + memcpy (a->tag, tag, sizeof (a->tag)); + + /* Create and populate the classifer tables */ + macip_create_classify_tables (am, *acl_list_index); + + return 0; +} + + +/* No check for validity of sw_if_index - the callers were supposed to validate */ + +static int +macip_acl_interface_del_acl (acl_main_t * am, u32 sw_if_index) +{ + int rv; + u32 macip_acl_index; + macip_acl_list_t *a; + vec_validate_init_empty (am->macip_acl_by_sw_if_index, sw_if_index, ~0); + macip_acl_index = am->macip_acl_by_sw_if_index[sw_if_index]; + /* No point in deleting MACIP ACL which is not applied */ + if (~0 == macip_acl_index) + return -1; + a = &am->macip_acls[macip_acl_index]; + /* remove the classifier tables off the interface L2 ACL */ + rv = + vnet_set_input_acl_intfc (am->vlib_main, sw_if_index, a->ip4_table_index, + a->ip6_table_index, a->l2_table_index, 0); + /* Unset the MACIP ACL index */ + am->macip_acl_by_sw_if_index[sw_if_index] = ~0; + return rv; +} + +/* No check for validity of sw_if_index - the callers were supposed to validate */ + +static int +macip_acl_interface_add_acl (acl_main_t * am, u32 sw_if_index, + u32 macip_acl_index) +{ + macip_acl_list_t *a; + int rv; + if (pool_is_free_index (am->macip_acls, macip_acl_index)) + { + return -1; + } + a = &am->macip_acls[macip_acl_index]; + vec_validate_init_empty (am->macip_acl_by_sw_if_index, sw_if_index, ~0); + /* If there already a MACIP ACL applied, unapply it */ + if (~0 != am->macip_acl_by_sw_if_index[sw_if_index]) + macip_acl_interface_del_acl(am, sw_if_index); + am->macip_acl_by_sw_if_index[sw_if_index] = macip_acl_index; + /* Apply the classifier tables for L2 ACLs */ + rv = + vnet_set_input_acl_intfc (am->vlib_main, sw_if_index, a->ip4_table_index, + a->ip6_table_index, a->l2_table_index, 1); + return rv; +} + +static int +macip_acl_del_list (u32 acl_list_index) +{ + acl_main_t *am = &acl_main; + macip_acl_list_t *a; + int i; + if (pool_is_free_index (am->macip_acls, acl_list_index)) + { + return -1; + } + + /* delete any references to the ACL */ + for (i = 0; i < vec_len (am->macip_acl_by_sw_if_index); i++) + { + if (am->macip_acl_by_sw_if_index[i] == acl_list_index) + { + macip_acl_interface_del_acl (am, i); + } + } + + /* Now that classifier tables are detached, clean them up */ + macip_destroy_classify_tables (am, acl_list_index); + + /* now we can delete the ACL itself */ + a = &am->macip_acls[acl_list_index]; + if (a->rules) + { + clib_mem_free (a->rules); + } + pool_put (am->macip_acls, a); + return 0; +} + + +static int +macip_acl_interface_add_del_acl (u32 sw_if_index, u8 is_add, + u32 acl_list_index) +{ + acl_main_t *am = &acl_main; + int rv = -1; + if (is_add) + { + rv = macip_acl_interface_add_acl (am, sw_if_index, acl_list_index); + } + else + { + rv = macip_acl_interface_del_acl (am, sw_if_index); + } + return rv; +} + +/* API message handler */ +static void +vl_api_acl_add_replace_t_handler (vl_api_acl_add_replace_t * mp) +{ + vl_api_acl_add_replace_reply_t *rmp; + acl_main_t *am = &acl_main; + int rv; + u32 acl_list_index = ntohl (mp->acl_index); + + rv = acl_add_list (ntohl (mp->count), mp->r, &acl_list_index, mp->tag); + + /* *INDENT-OFF* */ + REPLY_MACRO2(VL_API_ACL_ADD_REPLACE_REPLY, + ({ + rmp->acl_index = htonl(acl_list_index); + })); + /* *INDENT-ON* */ +} + +static void +vl_api_acl_del_t_handler (vl_api_acl_del_t * mp) +{ + acl_main_t *sm = &acl_main; + vl_api_acl_del_reply_t *rmp; + int rv; + + rv = acl_del_list (ntohl (mp->acl_index)); + + REPLY_MACRO (VL_API_ACL_DEL_REPLY); +} + +static void +vl_api_acl_interface_add_del_t_handler (vl_api_acl_interface_add_del_t * mp) +{ + acl_main_t *sm = &acl_main; + vnet_interface_main_t *im = &sm->vnet_main->interface_main; + u32 sw_if_index = ntohl (mp->sw_if_index); + vl_api_acl_interface_add_del_reply_t *rmp; + int rv = -1; + + if (pool_is_free_index(im->sw_interfaces, sw_if_index)) + rv = VNET_API_ERROR_INVALID_SW_IF_INDEX; + else + rv = + acl_interface_add_del_inout_acl (sw_if_index, mp->is_add, + mp->is_input, ntohl (mp->acl_index)); + + REPLY_MACRO (VL_API_ACL_INTERFACE_ADD_DEL_REPLY); +} + +static void +vl_api_acl_interface_set_acl_list_t_handler + (vl_api_acl_interface_set_acl_list_t * mp) +{ + acl_main_t *sm = &acl_main; + vl_api_acl_interface_set_acl_list_reply_t *rmp; + int rv = 0; + int i; + vnet_interface_main_t *im = &sm->vnet_main->interface_main; + u32 sw_if_index = ntohl (mp->sw_if_index); + + if (pool_is_free_index(im->sw_interfaces, sw_if_index)) + rv = VNET_API_ERROR_INVALID_SW_IF_INDEX; + else + { + acl_interface_reset_inout_acls (sw_if_index, 0); + acl_interface_reset_inout_acls (sw_if_index, 1); + + for (i = 0; i < mp->count; i++) + { + acl_interface_add_del_inout_acl (sw_if_index, 1, (i < mp->n_input), + ntohl (mp->acls[i])); + } + } + + REPLY_MACRO (VL_API_ACL_INTERFACE_SET_ACL_LIST_REPLY); +} + +static void +copy_acl_rule_to_api_rule (vl_api_acl_rule_t * api_rule, acl_rule_t * r) +{ + api_rule->is_permit = r->is_permit; + api_rule->is_ipv6 = r->is_ipv6; + if(r->is_ipv6) + { + memcpy (api_rule->src_ip_addr, &r->src, sizeof (r->src)); + memcpy (api_rule->dst_ip_addr, &r->dst, sizeof (r->dst)); + } + else + { + memcpy (api_rule->src_ip_addr, &r->src.ip4, sizeof (r->src.ip4)); + memcpy (api_rule->dst_ip_addr, &r->dst.ip4, sizeof (r->dst.ip4)); + } + api_rule->src_ip_prefix_len = r->src_prefixlen; + api_rule->dst_ip_prefix_len = r->dst_prefixlen; + api_rule->proto = r->proto; + api_rule->srcport_or_icmptype_first = r->src_port_or_type_first; + api_rule->srcport_or_icmptype_last = r->src_port_or_type_last; + api_rule->dstport_or_icmpcode_first = r->dst_port_or_code_first; + api_rule->dstport_or_icmpcode_last = r->dst_port_or_code_last; + api_rule->tcp_flags_mask = r->tcp_flags_mask; + api_rule->tcp_flags_value = r->tcp_flags_value; +} + +static void +send_acl_details (acl_main_t * am, unix_shared_memory_queue_t * q, + acl_list_t * acl, u32 context) +{ + vl_api_acl_details_t *mp; + vl_api_acl_rule_t *rules; + int i; + int msg_size = sizeof (*mp) + sizeof (mp->r[0]) * acl->count; + + mp = vl_msg_api_alloc (msg_size); + memset (mp, 0, msg_size); + mp->_vl_msg_id = ntohs (VL_API_ACL_DETAILS + am->msg_id_base); + + /* fill in the message */ + mp->context = context; + mp->count = htonl (acl->count); + mp->acl_index = htonl (acl - am->acls); + memcpy (mp->tag, acl->tag, sizeof (mp->tag)); + // clib_memcpy (mp->r, acl->rules, acl->count * sizeof(acl->rules[0])); + rules = mp->r; + for (i = 0; i < acl->count; i++) + { + copy_acl_rule_to_api_rule (&rules[i], &acl->rules[i]); + } + + clib_warning("Sending acl details for ACL index %d", ntohl(mp->acl_index)); + vl_msg_api_send_shmem (q, (u8 *) & mp); +} + + +static void +vl_api_acl_dump_t_handler (vl_api_acl_dump_t * mp) +{ + acl_main_t *am = &acl_main; + u32 acl_index; + acl_list_t *acl; + + int rv = -1; + unix_shared_memory_queue_t *q; + + q = vl_api_client_index_to_input_queue (mp->client_index); + if (q == 0) + { + return; + } + + if (mp->acl_index == ~0) + { + /* *INDENT-OFF* */ + /* Just dump all ACLs */ + pool_foreach (acl, am->acls, + ({ + send_acl_details(am, q, acl, mp->context); + })); + /* *INDENT-ON* */ + } + else + { + acl_index = ntohl (mp->acl_index); + if (!pool_is_free_index (am->acls, acl_index)) + { + acl = &am->acls[acl_index]; + send_acl_details (am, q, acl, mp->context); + } + } + + if (rv == -1) + { + /* FIXME API: should we signal an error here at all ? */ + return; + } +} + +static void +send_acl_interface_list_details (acl_main_t * am, + unix_shared_memory_queue_t * q, + u32 sw_if_index, u32 context) +{ + vl_api_acl_interface_list_details_t *mp; + int msg_size; + int n_input; + int n_output; + int count; + int i = 0; + + vec_validate (am->input_acl_vec_by_sw_if_index, sw_if_index); + vec_validate (am->output_acl_vec_by_sw_if_index, sw_if_index); + + n_input = vec_len (am->input_acl_vec_by_sw_if_index[sw_if_index]); + n_output = vec_len (am->output_acl_vec_by_sw_if_index[sw_if_index]); + count = n_input + n_output; + + msg_size = sizeof (*mp); + msg_size += sizeof (mp->acls[0]) * count; + + mp = vl_msg_api_alloc (msg_size); + memset (mp, 0, msg_size); + mp->_vl_msg_id = + ntohs (VL_API_ACL_INTERFACE_LIST_DETAILS + am->msg_id_base); + + /* fill in the message */ + mp->context = context; + mp->sw_if_index = htonl (sw_if_index); + mp->count = count; + mp->n_input = n_input; + for (i = 0; i < n_input; i++) + { + mp->acls[i] = htonl (am->input_acl_vec_by_sw_if_index[sw_if_index][i]); + } + for (i = 0; i < n_output; i++) + { + mp->acls[n_input + i] = + htonl (am->output_acl_vec_by_sw_if_index[sw_if_index][i]); + } + + vl_msg_api_send_shmem (q, (u8 *) & mp); +} + +static void +vl_api_acl_interface_list_dump_t_handler (vl_api_acl_interface_list_dump_t * + mp) +{ + acl_main_t *am = &acl_main; + vnet_sw_interface_t *swif; + vnet_interface_main_t *im = &am->vnet_main->interface_main; + + u32 sw_if_index; + unix_shared_memory_queue_t *q; + + q = vl_api_client_index_to_input_queue (mp->client_index); + if (q == 0) + { + return; + } + + if (mp->sw_if_index == ~0) + { + /* *INDENT-OFF* */ + pool_foreach (swif, im->sw_interfaces, + ({ + send_acl_interface_list_details(am, q, swif->sw_if_index, mp->context); + })); + /* *INDENT-ON* */ + } + else + { + sw_if_index = ntohl (mp->sw_if_index); + if (!pool_is_free_index(im->sw_interfaces, sw_if_index)) + send_acl_interface_list_details (am, q, sw_if_index, mp->context); + } +} + +/* MACIP ACL API handlers */ + +static void +vl_api_macip_acl_add_t_handler (vl_api_macip_acl_add_t * mp) +{ + vl_api_macip_acl_add_reply_t *rmp; + acl_main_t *am = &acl_main; + int rv; + u32 acl_list_index = ~0; + + rv = + macip_acl_add_list (ntohl (mp->count), mp->r, &acl_list_index, mp->tag); + + /* *INDENT-OFF* */ + REPLY_MACRO2(VL_API_MACIP_ACL_ADD_REPLY, + ({ + rmp->acl_index = htonl(acl_list_index); + })); + /* *INDENT-ON* */ +} + +static void +vl_api_macip_acl_del_t_handler (vl_api_macip_acl_del_t * mp) +{ + acl_main_t *sm = &acl_main; + vl_api_macip_acl_del_reply_t *rmp; + int rv; + + rv = macip_acl_del_list (ntohl (mp->acl_index)); + + REPLY_MACRO (VL_API_MACIP_ACL_DEL_REPLY); +} + +static void +vl_api_macip_acl_interface_add_del_t_handler + (vl_api_macip_acl_interface_add_del_t * mp) +{ + acl_main_t *sm = &acl_main; + vl_api_macip_acl_interface_add_del_reply_t *rmp; + int rv = -1; + vnet_interface_main_t *im = &sm->vnet_main->interface_main; + u32 sw_if_index = ntohl (mp->sw_if_index); + + if (pool_is_free_index(im->sw_interfaces, sw_if_index)) + rv = VNET_API_ERROR_INVALID_SW_IF_INDEX; + else + rv = + macip_acl_interface_add_del_acl (ntohl (mp->sw_if_index), mp->is_add, + ntohl (mp->acl_index)); + + REPLY_MACRO (VL_API_MACIP_ACL_INTERFACE_ADD_DEL_REPLY); +} + +static void +send_macip_acl_details (acl_main_t * am, unix_shared_memory_queue_t * q, + macip_acl_list_t * acl, u32 context) +{ + vl_api_macip_acl_details_t *mp; + vl_api_macip_acl_rule_t *rules; + macip_acl_rule_t *r; + int i; + int msg_size = sizeof (*mp) + (acl ? sizeof (mp->r[0]) * acl->count : 0); + + mp = vl_msg_api_alloc (msg_size); + memset (mp, 0, msg_size); + mp->_vl_msg_id = ntohs (VL_API_MACIP_ACL_DETAILS + am->msg_id_base); + + /* fill in the message */ + mp->context = context; + if (acl) + { + memcpy (mp->tag, acl->tag, sizeof (mp->tag)); + mp->count = htonl (acl->count); + mp->acl_index = htonl (acl - am->macip_acls); + rules = mp->r; + for (i = 0; i < acl->count; i++) + { + r = &acl->rules[i]; + rules[i].is_permit = r->is_permit; + rules[i].is_ipv6 = r->is_ipv6; + memcpy (rules[i].src_mac, &r->src_mac, sizeof (r->src_mac)); + memcpy (rules[i].src_mac_mask, &r->src_mac_mask, + sizeof (r->src_mac_mask)); + if (r->is_ipv6) + memcpy (rules[i].src_ip_addr, &r->src_ip_addr.ip6, + sizeof (r->src_ip_addr.ip6)); + else + memcpy (rules[i].src_ip_addr, &r->src_ip_addr.ip4, + sizeof (r->src_ip_addr.ip4)); + rules[i].src_ip_prefix_len = r->src_prefixlen; + } + } + else + { + /* No martini, no party - no ACL applied to this interface. */ + mp->acl_index = ~0; + mp->count = 0; + } + + vl_msg_api_send_shmem (q, (u8 *) & mp); +} + + +static void +vl_api_macip_acl_dump_t_handler (vl_api_macip_acl_dump_t * mp) +{ + acl_main_t *am = &acl_main; + macip_acl_list_t *acl; + + unix_shared_memory_queue_t *q; + + q = vl_api_client_index_to_input_queue (mp->client_index); + if (q == 0) + { + return; + } + + if (mp->acl_index == ~0) + { + /* Just dump all ACLs for now, with sw_if_index = ~0 */ + pool_foreach (acl, am->macip_acls, ( + { + send_macip_acl_details (am, q, acl, + mp-> + context);} + )); + /* *INDENT-ON* */ + } + else + { + u32 acl_index = ntohl (mp->acl_index); + if (!pool_is_free_index (am->macip_acls, acl_index)) + { + acl = &am->macip_acls[acl_index]; + send_macip_acl_details (am, q, acl, mp->context); + } + } +} + +static void +vl_api_macip_acl_interface_get_t_handler (vl_api_macip_acl_interface_get_t * + mp) +{ + acl_main_t *am = &acl_main; + vl_api_macip_acl_interface_get_reply_t *rmp; + u32 count = vec_len (am->macip_acl_by_sw_if_index); + int msg_size = sizeof (*rmp) + sizeof (rmp->acls[0]) * count; + unix_shared_memory_queue_t *q; + int i; + + q = vl_api_client_index_to_input_queue (mp->client_index); + if (q == 0) + { + return; + } + + rmp = vl_msg_api_alloc (msg_size); + memset (rmp, 0, msg_size); + rmp->_vl_msg_id = + ntohs (VL_API_MACIP_ACL_INTERFACE_GET_REPLY + am->msg_id_base); + rmp->context = mp->context; + rmp->count = htonl (count); + for (i = 0; i < count; i++) + { + rmp->acls[i] = htonl (am->macip_acl_by_sw_if_index[i]); + } + + vl_msg_api_send_shmem (q, (u8 *) & rmp); +} + + + +/* Set up the API message handling tables */ +static clib_error_t * +acl_plugin_api_hookup (vlib_main_t * vm) +{ + acl_main_t *sm = &acl_main; +#define _(N,n) \ + vl_msg_api_set_handlers((VL_API_##N + sm->msg_id_base), \ + #n, \ + vl_api_##n##_t_handler, \ + vl_noop_handler, \ + vl_api_##n##_t_endian, \ + vl_api_##n##_t_print, \ + sizeof(vl_api_##n##_t), 1); + foreach_acl_plugin_api_msg; +#undef _ + + return 0; +} + +#define vl_msg_name_crc_list +#include +#undef vl_msg_name_crc_list + +static void +setup_message_id_table (acl_main_t * sm, api_main_t * am) +{ +#define _(id,n,crc) \ + vl_msg_api_add_msg_name_crc (am, #n "_" #crc, id + sm->msg_id_base); + foreach_vl_msg_name_crc_acl; +#undef _ +} + +u32 +register_match_action_nexts (u32 next_in_ip4, u32 next_in_ip6, + u32 next_out_ip4, u32 next_out_ip6) +{ + acl_main_t *am = &acl_main; + u32 act = am->n_match_actions; + if (am->n_match_actions == 255) + { + return ~0; + } + am->n_match_actions++; + am->acl_in_ip4_match_next[act] = next_in_ip4; + am->acl_in_ip6_match_next[act] = next_in_ip6; + am->acl_out_ip4_match_next[act] = next_out_ip4; + am->acl_out_ip6_match_next[act] = next_out_ip6; + return act; +} + +void +acl_setup_nodes (void) +{ + vlib_main_t *vm = vlib_get_main (); + acl_main_t *am = &acl_main; + vlib_node_t *n; + + n = vlib_get_node_by_name (vm, (u8 *) "l2-input-classify"); + am->l2_input_classify_next_acl = + vlib_node_add_next_with_slot (vm, n->index, acl_in_node.index, ~0); + n = vlib_get_node_by_name (vm, (u8 *) "l2-output-classify"); + am->l2_output_classify_next_acl = + vlib_node_add_next_with_slot (vm, n->index, acl_out_node.index, ~0); + + feat_bitmap_init_next_nodes (vm, acl_in_node.index, L2INPUT_N_FEAT, + l2input_get_feat_names (), + am->acl_in_node_input_next_node_index); + + memset (&am->acl_in_ip4_match_next[0], 0, + sizeof (am->acl_in_ip4_match_next)); + memset (&am->acl_in_ip6_match_next[0], 0, + sizeof (am->acl_in_ip6_match_next)); + memset (&am->acl_out_ip4_match_next[0], 0, + sizeof (am->acl_out_ip4_match_next)); + memset (&am->acl_out_ip6_match_next[0], 0, + sizeof (am->acl_out_ip6_match_next)); + am->n_match_actions = 0; + + register_match_action_nexts (0, 0, 0, 0); /* drop */ + register_match_action_nexts (~0, ~0, ~0, ~0); /* permit */ + register_match_action_nexts (ACL_IN_L2S_INPUT_IP4_ADD, ACL_IN_L2S_INPUT_IP6_ADD, ACL_OUT_L2S_OUTPUT_IP4_ADD, ACL_OUT_L2S_OUTPUT_IP6_ADD); /* permit + create session */ +} + + + +static clib_error_t * +acl_init (vlib_main_t * vm) +{ + acl_main_t *am = &acl_main; + clib_error_t *error = 0; + memset (am, 0, sizeof (*am)); + am->vlib_main = vm; + am->vnet_main = vnet_get_main (); + + u8 *name = format (0, "acl_%08x%c", api_version, 0); + + /* Ask for a correctly-sized block of API message decode slots */ + am->msg_id_base = vl_msg_api_get_msg_ids ((char *) name, + VL_MSG_FIRST_AVAILABLE); + + error = acl_plugin_api_hookup (vm); + acl_setup_nodes (); + + /* Add our API messages to the global name_crc hash table */ + setup_message_id_table (am, &api_main); + + vec_free (name); + + return error; +} + +VLIB_INIT_FUNCTION (acl_init); diff --git a/src/plugins/acl/acl.h b/src/plugins/acl/acl.h new file mode 100644 index 00000000..afc9b289 --- /dev/null +++ b/src/plugins/acl/acl.h @@ -0,0 +1,148 @@ +/* + * Copyright (c) 2016 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ +#ifndef included_acl_h +#define included_acl_h + +#include +#include +#include +#include + + +#include +#include +#include + +#define ACL_PLUGIN_VERSION_MAJOR 1 +#define ACL_PLUGIN_VERSION_MINOR 1 + +extern vlib_node_registration_t acl_in_node; +extern vlib_node_registration_t acl_out_node; + +void input_acl_packet_match(u32 sw_if_index, vlib_buffer_t * b0, u32 *nextp, u32 *acl_match_p, u32 *rule_match_p, u32 *trace_bitmap); +void output_acl_packet_match(u32 sw_if_index, vlib_buffer_t * b0, u32 *nextp, u32 *acl_match_p, u32 *rule_match_p, u32 *trace_bitmap); + +enum address_e { IP4, IP6 }; +typedef struct +{ + enum address_e type; + union { + ip6_address_t ip6; + ip4_address_t ip4; + } addr; +} address_t; + +/* + * ACL rules + */ +typedef struct +{ + u8 is_permit; + u8 is_ipv6; + ip46_address_t src; + u8 src_prefixlen; + ip46_address_t dst; + u8 dst_prefixlen; + u8 proto; + u16 src_port_or_type_first; + u16 src_port_or_type_last; + u16 dst_port_or_code_first; + u16 dst_port_or_code_last; + u8 tcp_flags_value; + u8 tcp_flags_mask; +} acl_rule_t; + +typedef struct +{ + u8 is_permit; + u8 is_ipv6; + u8 src_mac[6]; + u8 src_mac_mask[6]; + ip46_address_t src_ip_addr; + u8 src_prefixlen; +} macip_acl_rule_t; + +/* + * ACL + */ +typedef struct +{ + u8 tag[64]; + u32 count; + acl_rule_t *rules; +} acl_list_t; + +typedef struct +{ + u8 tag[64]; + u32 count; + macip_acl_rule_t *rules; + /* References to the classifier tables that will enforce the rules */ + u32 ip4_table_index; + u32 ip6_table_index; + u32 l2_table_index; +} macip_acl_list_t; + +typedef struct { + /* API message ID base */ + u16 msg_id_base; + + acl_list_t *acls; /* Pool of ACLs */ + macip_acl_list_t *macip_acls; /* Pool of MAC-IP ACLs */ + + /* ACLs associated with interfaces */ + u32 **input_acl_vec_by_sw_if_index; + u32 **output_acl_vec_by_sw_if_index; + + /* + * Classify tables used to grab the packets for the ACL check, + * and serving as the 5-tuple session tables at the same time + */ + u32 *acl_ip4_input_classify_table_by_sw_if_index; + u32 *acl_ip6_input_classify_table_by_sw_if_index; + u32 *acl_ip4_output_classify_table_by_sw_if_index; + u32 *acl_ip6_output_classify_table_by_sw_if_index; + + /* MACIP (input) ACLs associated with the interfaces */ + u32 *macip_acl_by_sw_if_index; + + /* next indices for our nodes in the l2-classify tables */ + u32 l2_input_classify_next_acl; + u32 l2_output_classify_next_acl; + + /* next node indices for feature bitmap */ + u32 acl_in_node_input_next_node_index[32]; + /* the respective thing for the output feature */ + l2_output_next_nodes_st acl_out_output_next_nodes; + + /* ACL match actions (must be coherent across in/out ACLs to next indices (can differ) */ + + u32 acl_in_ip4_match_next[256]; + u32 acl_in_ip6_match_next[256]; + u32 acl_out_ip4_match_next[256]; + u32 acl_out_ip6_match_next[256]; + u32 n_match_actions; + + + /* convenience */ + vlib_main_t * vlib_main; + vnet_main_t * vnet_main; + ethernet_main_t * ethernet_main; +} acl_main_t; + +extern acl_main_t acl_main; + + +#endif diff --git a/src/plugins/acl/acl_all_api_h.h b/src/plugins/acl/acl_all_api_h.h new file mode 100644 index 00000000..96eca56d --- /dev/null +++ b/src/plugins/acl/acl_all_api_h.h @@ -0,0 +1,321 @@ +/* + * Copyright (c) 2016 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ +/* Include the generated file, see BUILT_SOURCES in Makefile.am */ +#include + +#ifdef vl_printfun + +#ifdef LP64 +#define _uword_fmt "%lld" +#define _uword_cast (long long) +#else +#define _uword_fmt "%ld" +#define _uword_cast long +#endif + +static inline void * +vl_api_acl_rule_t_print (vl_api_acl_rule_t * a, void *handle) +{ + vl_print (handle, "vl_api_acl_rule_t:\n"); + vl_print (handle, "is_permit: %u\n", (unsigned) a->is_permit); + vl_print (handle, "is_ipv6: %u\n", (unsigned) a->is_ipv6); + { + int _i; + for (_i = 0; _i < 16; _i++) + { + vl_print (handle, "src_ip_addr[%d]: %u\n", _i, a->src_ip_addr[_i]); + } + } + vl_print (handle, "src_ip_prefix_len: %u\n", + (unsigned) a->src_ip_prefix_len); + { + int _i; + for (_i = 0; _i < 16; _i++) + { + vl_print (handle, "dst_ip_addr[%d]: %u\n", _i, a->dst_ip_addr[_i]); + } + } + vl_print (handle, "dst_ip_prefix_len: %u\n", + (unsigned) a->dst_ip_prefix_len); + vl_print (handle, "proto: %u\n", (unsigned) a->proto); + vl_print (handle, "srcport_or_icmptype_first: %u\n", + (unsigned) a->srcport_or_icmptype_first); + vl_print (handle, "srcport_or_icmptype_last: %u\n", + (unsigned) a->srcport_or_icmptype_last); + vl_print (handle, "dstport_or_icmpcode_first: %u\n", + (unsigned) a->dstport_or_icmpcode_first); + vl_print (handle, "dstport_or_icmpcode_last: %u\n", + (unsigned) a->dstport_or_icmpcode_last); + vl_print (handle, "tcp_flags_mask: %u\n", (unsigned) a->tcp_flags_mask); + vl_print (handle, "tcp_flags_value: %u\n", (unsigned) a->tcp_flags_value); + return handle; +} + +static inline void * +vl_api_acl_add_replace_t_print (vl_api_acl_add_replace_t * a, void *handle) +{ + int i; + vl_print (handle, "vl_api_acl_add_replace_t:\n"); + vl_print (handle, "_vl_msg_id: %u\n", (unsigned) a->_vl_msg_id); + vl_print (handle, "client_index: %u\n", (unsigned) a->client_index); + vl_print (handle, "context: %u\n", (unsigned) a->context); + vl_print (handle, "acl_index: %u\n", (unsigned) a->acl_index); + vl_print (handle, "count: %u\n", (unsigned) a->count); + vl_print (handle, "r ----- \n"); + for (i = 0; i < a->count; i++) + { + vl_print (handle, " r[%d]:\n", i); + vl_api_acl_rule_t_print (&a->r[i], handle); + } + vl_print (handle, "r ----- END \n"); + return handle; +} + + +static inline void *vl_api_acl_details_t_print (vl_api_acl_details_t *a,void *handle) +{ + vl_print(handle, "vl_api_acl_details_t:\n"); + vl_print(handle, "_vl_msg_id: %u\n", (unsigned) a->_vl_msg_id); + vl_print(handle, "context: %u\n", (unsigned) a->context); + vl_print(handle, "acl_index: %u\n", (unsigned) a->acl_index); + { + int _i; + for (_i = 0; _i < 64; _i++) { + vl_print(handle, "tag[%d]: %u\n", _i, a->tag[_i]); + } + } + vl_print(handle, "count: %u\n", (unsigned) a->count); + vl_print(handle, "r ----- \n"); + // FIXME vl_api_acl_rule_t_print(&a->r, handle); + vl_print(handle, "r ----- END \n"); + return handle; +} + +static inline void * +vl_api_macip_acl_rule_t_print (vl_api_macip_acl_rule_t * a, void *handle) +{ + vl_print (handle, "vl_api_macip_acl_rule_t:\n"); + vl_print (handle, "is_permit: %u\n", (unsigned) a->is_permit); + vl_print (handle, "is_ipv6: %u\n", (unsigned) a->is_ipv6); + { + int _i; + for (_i = 0; _i < 6; _i++) + { + vl_print (handle, "src_mac[%d]: %u\n", _i, a->src_mac[_i]); + } + } + { + int _i; + for (_i = 0; _i < 6; _i++) + { + vl_print (handle, "src_mac_mask[%d]: %u\n", _i, a->src_mac_mask[_i]); + } + } + { + int _i; + for (_i = 0; _i < 16; _i++) + { + vl_print (handle, "src_ip_addr[%d]: %u\n", _i, a->src_ip_addr[_i]); + } + } + vl_print (handle, "src_ip_prefix_len: %u\n", + (unsigned) a->src_ip_prefix_len); + return handle; +} + +static inline void * +vl_api_macip_acl_add_t_print (vl_api_macip_acl_add_t * a, void *handle) +{ + int i; + vl_print (handle, "vl_api_macip_acl_add_t:\n"); + vl_print (handle, "_vl_msg_id: %u\n", (unsigned) a->_vl_msg_id); + vl_print (handle, "client_index: %u\n", (unsigned) a->client_index); + vl_print (handle, "context: %u\n", (unsigned) a->context); + vl_print (handle, "count: %u\n", (unsigned) a->count); + vl_print (handle, "r ----- \n"); + for (i = 0; i < a->count; i++) + { + vl_print (handle, " r[%d]:\n", i); + vl_api_macip_acl_rule_t_print (&a->r[i], handle); + } + vl_print (handle, "r ----- END \n"); + return handle; +} + +static inline void *vl_api_macip_acl_details_t_print (vl_api_macip_acl_details_t *a,void *handle) +{ + int i; + vl_print(handle, "vl_api_macip_acl_details_t:\n"); + vl_print(handle, "_vl_msg_id: %u\n", (unsigned) a->_vl_msg_id); + vl_print(handle, "context: %u\n", (unsigned) a->context); + vl_print(handle, "acl_index: %u\n", (unsigned) a->acl_index); + { + int _i; + for (_i = 0; _i < 64; _i++) { + vl_print(handle, "tag[%d]: %u\n", _i, a->tag[_i]); + } + } + vl_print(handle, "count: %u\n", (unsigned) a->count); + vl_print(handle, "r ----- \n"); + for (i = 0; i < a->count; i++) + { + vl_print (handle, " r[%d]:\n", i); + vl_api_macip_acl_rule_t_print (&a->r[i], handle); + } + vl_print(handle, "r ----- END \n"); + return handle; +} + +#endif /* vl_printfun */ + + +#ifdef vl_endianfun + +#undef clib_net_to_host_uword +#ifdef LP64 +#define clib_net_to_host_uword clib_net_to_host_u64 +#else +#define clib_net_to_host_uword clib_net_to_host_u32 +#endif + +/* + * Manual endian/print functions created by copypasting the automatically + * generated ones with small required adjustments. Appears the codegen + * can't make code to print the contents of custom-type array. + */ + +static inline void +vl_api_acl_rule_t_endian (vl_api_acl_rule_t * a) +{ + /* a->is_permit = a->is_permit (no-op) */ + /* a->is_ipv6 = a->is_ipv6 (no-op) */ + /* a->src_ip_addr[0..15] = a->src_ip_addr[0..15] (no-op) */ + /* a->src_ip_prefix_len = a->src_ip_prefix_len (no-op) */ + /* a->dst_ip_addr[0..15] = a->dst_ip_addr[0..15] (no-op) */ + /* a->dst_ip_prefix_len = a->dst_ip_prefix_len (no-op) */ + /* a->proto = a->proto (no-op) */ + a->srcport_or_icmptype_first = + clib_net_to_host_u16 (a->srcport_or_icmptype_first); + a->srcport_or_icmptype_last = + clib_net_to_host_u16 (a->srcport_or_icmptype_last); + a->dstport_or_icmpcode_first = + clib_net_to_host_u16 (a->dstport_or_icmpcode_first); + a->dstport_or_icmpcode_last = + clib_net_to_host_u16 (a->dstport_or_icmpcode_last); + /* a->tcp_flags_mask = a->tcp_flags_mask (no-op) */ + /* a->tcp_flags_value = a->tcp_flags_value (no-op) */ +} + +static inline void +vl_api_acl_add_replace_t_endian (vl_api_acl_add_replace_t * a) +{ + int i; + a->_vl_msg_id = clib_net_to_host_u16 (a->_vl_msg_id); + a->client_index = clib_net_to_host_u32 (a->client_index); + a->context = clib_net_to_host_u32 (a->context); + a->acl_index = clib_net_to_host_u32 (a->acl_index); + a->count = clib_net_to_host_u32 (a->count); + for (i = 0; i < a->count; i++) + { + vl_api_acl_rule_t_endian (&a->r[i]); + } +} + +static inline void vl_api_acl_details_t_endian (vl_api_acl_details_t *a) +{ + int i; + a->_vl_msg_id = clib_net_to_host_u16(a->_vl_msg_id); + a->context = clib_net_to_host_u32(a->context); + a->acl_index = clib_net_to_host_u32(a->acl_index); + /* a->tag[0..63] = a->tag[0..63] (no-op) */ + a->count = clib_net_to_host_u32(a->count); + for (i = 0; i < a->count; i++) + { + vl_api_acl_rule_t_endian (&a->r[i]); + } +} + +static inline void vl_api_acl_interface_list_details_t_endian (vl_api_acl_interface_list_details_t *a) +{ + int i; + a->_vl_msg_id = clib_net_to_host_u16(a->_vl_msg_id); + a->context = clib_net_to_host_u32(a->context); + a->sw_if_index = clib_net_to_host_u32(a->sw_if_index); + /* a->count = a->count (no-op) */ + /* a->n_input = a->n_input (no-op) */ + for(i=0; icount; i++) { + a->acls[i] = clib_net_to_host_u32(a->acls[i]); + } +} + +static inline void vl_api_acl_interface_set_acl_list_t_endian (vl_api_acl_interface_set_acl_list_t *a) +{ + int i; + a->_vl_msg_id = clib_net_to_host_u16(a->_vl_msg_id); + a->client_index = clib_net_to_host_u32(a->client_index); + a->context = clib_net_to_host_u32(a->context); + a->sw_if_index = clib_net_to_host_u32(a->sw_if_index); + /* a->count = a->count (no-op) */ + /* a->n_input = a->n_input (no-op) */ + for(i=0; icount; i++) { + a->acls[i] = clib_net_to_host_u32(a->acls[i]); + } +} + +static inline void +vl_api_macip_acl_rule_t_endian (vl_api_macip_acl_rule_t * a) +{ + /* a->is_permit = a->is_permit (no-op) */ + /* a->is_ipv6 = a->is_ipv6 (no-op) */ + /* a->src_mac[0..5] = a->src_mac[0..5] (no-op) */ + /* a->src_mac_mask[0..5] = a->src_mac_mask[0..5] (no-op) */ + /* a->src_ip_addr[0..15] = a->src_ip_addr[0..15] (no-op) */ + /* a->src_ip_prefix_len = a->src_ip_prefix_len (no-op) */ +} + +static inline void +vl_api_macip_acl_add_t_endian (vl_api_macip_acl_add_t * a) +{ + int i; + a->_vl_msg_id = clib_net_to_host_u16 (a->_vl_msg_id); + a->client_index = clib_net_to_host_u32 (a->client_index); + a->context = clib_net_to_host_u32 (a->context); + a->count = clib_net_to_host_u32 (a->count); + for (i = 0; i < a->count; i++) + { + vl_api_macip_acl_rule_t_endian (&a->r[i]); + } +} + +static inline void vl_api_macip_acl_details_t_endian (vl_api_macip_acl_details_t *a) +{ + int i; + a->_vl_msg_id = clib_net_to_host_u16(a->_vl_msg_id); + a->context = clib_net_to_host_u32(a->context); + a->acl_index = clib_net_to_host_u32(a->acl_index); + /* a->tag[0..63] = a->tag[0..63] (no-op) */ + a->count = clib_net_to_host_u32(a->count); + for (i = 0; i < a->count; i++) + { + vl_api_macip_acl_rule_t_endian (&a->r[i]); + } +} + + + + +#endif /* vl_printfun */ + + diff --git a/src/plugins/acl/acl_msg_enum.h b/src/plugins/acl/acl_msg_enum.h new file mode 100644 index 00000000..14d8b48c --- /dev/null +++ b/src/plugins/acl/acl_msg_enum.h @@ -0,0 +1,28 @@ +/* + * Copyright (c) 2016 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ +#ifndef included_acl_msg_enum_h +#define included_acl_msg_enum_h + +#include + +#define vl_msg_id(n,h) n, +typedef enum { +#include + /* We'll want to know how many messages IDs we need... */ + VL_MSG_FIRST_AVAILABLE, +} vl_msg_id_t; +#undef vl_msg_id + +#endif diff --git a/src/plugins/acl/acl_test.c b/src/plugins/acl/acl_test.c new file mode 100644 index 00000000..a0e413e1 --- /dev/null +++ b/src/plugins/acl/acl_test.c @@ -0,0 +1,1024 @@ +/* + * Copyright (c) 2015 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ +/* + *------------------------------------------------------------------ + * acl_test.c - test harness plugin + *------------------------------------------------------------------ + */ + +#include +#include +#include +#include +#include +#include +#include + +uword unformat_sw_if_index (unformat_input_t * input, va_list * args); + +/* Declare message IDs */ +#include + +/* define message structures */ +#define vl_typedefs +#include +#undef vl_typedefs + +/* define message structures */ +#define vl_endianfun +#include +#undef vl_endianfun + +/* instantiate all the print functions we know about */ +#define vl_print(handle, ...) +#define vl_printfun +#include +#undef vl_printfun + +/* Get the API version number. */ +#define vl_api_version(n,v) static u32 api_version=(v); +#include +#undef vl_api_version + +typedef struct { + /* API message ID base */ + u16 msg_id_base; + vat_main_t *vat_main; +} acl_test_main_t; + +acl_test_main_t acl_test_main; + +#define foreach_standard_reply_retval_handler \ +_(acl_del_reply) \ +_(acl_interface_add_del_reply) \ +_(macip_acl_interface_add_del_reply) \ +_(acl_interface_set_acl_list_reply) \ +_(macip_acl_del_reply) + +#define foreach_reply_retval_aclindex_handler \ +_(acl_add_replace_reply) \ +_(macip_acl_add_reply) + +#define _(n) \ + static void vl_api_##n##_t_handler \ + (vl_api_##n##_t * mp) \ + { \ + vat_main_t * vam = acl_test_main.vat_main; \ + i32 retval = ntohl(mp->retval); \ + if (vam->async_mode) { \ + vam->async_errors += (retval < 0); \ + } else { \ + vam->retval = retval; \ + vam->result_ready = 1; \ + } \ + } +foreach_standard_reply_retval_handler; +#undef _ + +#define _(n) \ + static void vl_api_##n##_t_handler \ + (vl_api_##n##_t * mp) \ + { \ + vat_main_t * vam = acl_test_main.vat_main; \ + i32 retval = ntohl(mp->retval); \ + if (vam->async_mode) { \ + vam->async_errors += (retval < 0); \ + } else { \ + clib_warning("ACL index: %d", ntohl(mp->acl_index)); \ + vam->retval = retval; \ + vam->result_ready = 1; \ + } \ + } +foreach_reply_retval_aclindex_handler; +#undef _ + +/* These two ought to be in a library somewhere but they aren't */ +static uword +my_unformat_mac_address (unformat_input_t * input, va_list * args) +{ + u8 *a = va_arg (*args, u8 *); + return unformat (input, "%x:%x:%x:%x:%x:%x", &a[0], &a[1], &a[2], &a[3], + &a[4], &a[5]); +} + +static u8 * +my_format_mac_address (u8 * s, va_list * args) +{ + u8 *a = va_arg (*args, u8 *); + return format (s, "%02x:%02x:%02x:%02x:%02x:%02x", + a[0], a[1], a[2], a[3], a[4], a[5]); +} + + + +static void vl_api_acl_plugin_get_version_reply_t_handler + (vl_api_acl_plugin_get_version_reply_t * mp) + { + vat_main_t * vam = acl_test_main.vat_main; + clib_warning("ACL plugin version: %d.%d", ntohl(mp->major), ntohl(mp->minor)); + vam->result_ready = 1; + } + +static void vl_api_acl_interface_list_details_t_handler + (vl_api_acl_interface_list_details_t * mp) + { + int i; + vat_main_t * vam = acl_test_main.vat_main; + u8 *out = 0; + vl_api_acl_interface_list_details_t_endian(mp); + out = format(out, "sw_if_index: %d, count: %d, n_input: %d\n", mp->sw_if_index, mp->count, mp->n_input); + out = format(out, " input "); + for(i=0; icount; i++) { + out = format(out, "%d ", mp->acls[i]); + if (i == mp->n_input-1) + out = format(out, "\n output "); + } + out = format(out, "\n"); + clib_warning("%s", out); + vec_free(out); + vam->result_ready = 1; + } + + +static inline u8 * +vl_api_acl_rule_t_pretty_format (u8 *out, vl_api_acl_rule_t * a) +{ + int af = a->is_ipv6 ? AF_INET6 : AF_INET; + u8 src[INET6_ADDRSTRLEN]; + u8 dst[INET6_ADDRSTRLEN]; + inet_ntop(af, a->src_ip_addr, (void *)src, sizeof(src)); + inet_ntop(af, a->dst_ip_addr, (void *)dst, sizeof(dst)); + + out = format(out, "%s action %d src %s/%d dst %s/%d proto %d sport %d-%d dport %d-%d tcpflags %d %d", + a->is_ipv6 ? "ipv6" : "ipv4", a->is_permit, + src, a->src_ip_prefix_len, + dst, a->dst_ip_prefix_len, + a->proto, + a->srcport_or_icmptype_first, a->srcport_or_icmptype_last, + a->dstport_or_icmpcode_first, a->dstport_or_icmpcode_last, + a->tcp_flags_mask, a->tcp_flags_value); + return(out); +} + + + +static void vl_api_acl_details_t_handler + (vl_api_acl_details_t * mp) + { + int i; + vat_main_t * vam = acl_test_main.vat_main; + vl_api_acl_details_t_endian(mp); + u8 *out = 0; + out = format(0, "acl_index: %d, count: %d\n tag {%s}\n", mp->acl_index, mp->count, mp->tag); + for(i=0; icount; i++) { + out = format(out, " "); + out = vl_api_acl_rule_t_pretty_format(out, &mp->r[i]); + out = format(out, "%s\n", icount-1 ? "," : ""); + } + clib_warning("%s", out); + vec_free(out); + vam->result_ready = 1; + } + +static inline u8 * +vl_api_macip_acl_rule_t_pretty_format (u8 *out, vl_api_macip_acl_rule_t * a) +{ + int af = a->is_ipv6 ? AF_INET6 : AF_INET; + u8 src[INET6_ADDRSTRLEN]; + inet_ntop(af, a->src_ip_addr, (void *)src, sizeof(src)); + + out = format(out, "%s action %d ip %s/%d mac %U mask %U", + a->is_ipv6 ? "ipv6" : "ipv4", a->is_permit, + src, a->src_ip_prefix_len, + my_format_mac_address, a->src_mac, + my_format_mac_address, a->src_mac_mask); + return(out); +} + + +static void vl_api_macip_acl_details_t_handler + (vl_api_macip_acl_details_t * mp) + { + int i; + vat_main_t * vam = acl_test_main.vat_main; + vl_api_macip_acl_details_t_endian(mp); + u8 *out = format(0,"MACIP acl_index: %d, count: %d\n tag {%s}\n", mp->acl_index, mp->count, mp->tag); + for(i=0; icount; i++) { + out = format(out, " "); + out = vl_api_macip_acl_rule_t_pretty_format(out, &mp->r[i]); + out = format(out, "%s\n", icount-1 ? "," : ""); + } + clib_warning("%s", out); + vec_free(out); + vam->result_ready = 1; + } + +static void vl_api_macip_acl_interface_get_reply_t_handler + (vl_api_macip_acl_interface_get_reply_t * mp) + { + int i; + vat_main_t * vam = acl_test_main.vat_main; + u8 *out = format(0, "sw_if_index with MACIP ACL count: %d\n", ntohl(mp->count)); + for(i=0; icount); i++) { + out = format(out, " macip_acl_interface_add_del sw_if_index %d add acl %d\n", i, ntohl(mp->acls[i])); + } + out = format(out, "\n"); + clib_warning("%s", out); + vec_free(out); + vam->result_ready = 1; + } + + +/* + * Table of message reply handlers, must include boilerplate handlers + * we just generated + */ +#define foreach_vpe_api_reply_msg \ +_(ACL_ADD_REPLACE_REPLY, acl_add_replace_reply) \ +_(ACL_DEL_REPLY, acl_del_reply) \ +_(ACL_INTERFACE_ADD_DEL_REPLY, acl_interface_add_del_reply) \ +_(ACL_INTERFACE_SET_ACL_LIST_REPLY, acl_interface_set_acl_list_reply) \ +_(ACL_INTERFACE_LIST_DETAILS, acl_interface_list_details) \ +_(ACL_DETAILS, acl_details) \ +_(MACIP_ACL_ADD_REPLY, macip_acl_add_reply) \ +_(MACIP_ACL_DEL_REPLY, macip_acl_del_reply) \ +_(MACIP_ACL_DETAILS, macip_acl_details) \ +_(MACIP_ACL_INTERFACE_ADD_DEL_REPLY, macip_acl_interface_add_del_reply) \ +_(MACIP_ACL_INTERFACE_GET_REPLY, macip_acl_interface_get_reply) \ +_(ACL_PLUGIN_GET_VERSION_REPLY, acl_plugin_get_version_reply) + +/* M: construct, but don't yet send a message */ + +#define M(T,t) \ +do { \ + vam->result_ready = 0; \ + mp = vl_msg_api_alloc(sizeof(*mp)); \ + memset (mp, 0, sizeof (*mp)); \ + mp->_vl_msg_id = ntohs (VL_API_##T + sm->msg_id_base); \ + mp->client_index = vam->my_client_index; \ +} while(0); + +#define M2(T,t,n) \ +do { \ + vam->result_ready = 0; \ + mp = vl_msg_api_alloc(sizeof(*mp)+(n)); \ + memset (mp, 0, sizeof (*mp)); \ + mp->_vl_msg_id = ntohs (VL_API_##T + sm->msg_id_base); \ + mp->client_index = vam->my_client_index; \ +} while(0); + +/* S: send a message */ +#define S (vl_msg_api_send_shmem (vam->vl_input_queue, (u8 *)&mp)) + +/* W: wait for results, with timeout */ +#define W \ +do { \ + timeout = vat_time_now (vam) + 1.0; \ + \ + while (vat_time_now (vam) < timeout) { \ + if (vam->result_ready == 1) { \ + return (vam->retval); \ + } \ + } \ + return -99; \ +} while(0); + +static int api_acl_plugin_get_version (vat_main_t * vam) +{ + acl_test_main_t * sm = &acl_test_main; + vl_api_acl_plugin_get_version_t * mp; + u32 msg_size = sizeof(*mp); + f64 timeout; + + vam->result_ready = 0; + mp = vl_msg_api_alloc_as_if_client(msg_size); + memset (mp, 0, msg_size); + mp->_vl_msg_id = ntohs (VL_API_ACL_PLUGIN_GET_VERSION + sm->msg_id_base); + mp->client_index = vam->my_client_index; + + /* send it... */ + S; + + /* Wait for a reply... */ + W; + + return 0; +} + +static int api_macip_acl_interface_get (vat_main_t * vam) +{ + acl_test_main_t * sm = &acl_test_main; + vl_api_acl_plugin_get_version_t * mp; + u32 msg_size = sizeof(*mp); + f64 timeout; + + vam->result_ready = 0; + mp = vl_msg_api_alloc_as_if_client(msg_size); + memset (mp, 0, msg_size); + mp->_vl_msg_id = ntohs (VL_API_MACIP_ACL_INTERFACE_GET + sm->msg_id_base); + mp->client_index = vam->my_client_index; + + /* send it... */ + S; + + /* Wait for a reply... */ + W; + + return 0; +} + +#define vec_validate_acl_rules(v, idx) \ + do { \ + if (vec_len(v) < idx+1) { \ + vec_validate(v, idx); \ + v[idx].is_permit = 0x1; \ + v[idx].srcport_or_icmptype_last = 0xffff; \ + v[idx].dstport_or_icmpcode_last = 0xffff; \ + } \ + } while (0) + + +static int api_acl_add_replace (vat_main_t * vam) +{ + acl_test_main_t * sm = &acl_test_main; + unformat_input_t * i = vam->input; + f64 timeout; + vl_api_acl_add_replace_t * mp; + u32 acl_index = ~0; + u32 msg_size = sizeof (*mp); /* without the rules */ + + vl_api_acl_rule_t *rules = 0; + int rule_idx = 0; + int n_rules = 0; + u32 proto = 0; + u32 port1 = 0; + u32 port2 = 0; + u32 action = 0; + u32 tcpflags, tcpmask; + u32 src_prefix_length = 0, dst_prefix_length = 0; + ip4_address_t src_v4address, dst_v4address; + ip6_address_t src_v6address, dst_v6address; + u8 *tag = 0; + + if (!unformat (i, "%d", &acl_index)) { + /* Just assume -1 */ + } + + while (unformat_check_input (i) != UNFORMAT_END_OF_INPUT) + { + if (unformat (i, "ipv6")) + { + vec_validate_acl_rules(rules, rule_idx); + rules[rule_idx].is_ipv6 = 1; + } + else if (unformat (i, "ipv4")) + { + vec_validate_acl_rules(rules, rule_idx); + rules[rule_idx].is_ipv6 = 0; + } + else if (unformat (i, "permit+reflect")) + { + vec_validate_acl_rules(rules, rule_idx); + rules[rule_idx].is_permit = 2; + } + else if (unformat (i, "permit")) + { + vec_validate_acl_rules(rules, rule_idx); + rules[rule_idx].is_permit = 1; + } + else if (unformat (i, "action %d", &action)) + { + vec_validate_acl_rules(rules, rule_idx); + rules[rule_idx].is_permit = action; + } + else if (unformat (i, "src %U/%d", + unformat_ip4_address, &src_v4address, &src_prefix_length)) + { + vec_validate_acl_rules(rules, rule_idx); + memcpy (rules[rule_idx].src_ip_addr, &src_v4address, 4); + rules[rule_idx].src_ip_prefix_len = src_prefix_length; + rules[rule_idx].is_ipv6 = 0; + } + else if (unformat (i, "src %U/%d", + unformat_ip6_address, &src_v6address, &src_prefix_length)) + { + vec_validate_acl_rules(rules, rule_idx); + memcpy (rules[rule_idx].src_ip_addr, &src_v6address, 16); + rules[rule_idx].src_ip_prefix_len = src_prefix_length; + rules[rule_idx].is_ipv6 = 1; + } + else if (unformat (i, "dst %U/%d", + unformat_ip4_address, &dst_v4address, &dst_prefix_length)) + { + vec_validate_acl_rules(rules, rule_idx); + memcpy (rules[rule_idx].dst_ip_addr, &dst_v4address, 4); + rules[rule_idx].dst_ip_prefix_len = dst_prefix_length; + rules[rule_idx].is_ipv6 = 0; + } + else if (unformat (i, "dst %U/%d", + unformat_ip6_address, &dst_v6address, &dst_prefix_length)) + { + vec_validate_acl_rules(rules, rule_idx); + memcpy (rules[rule_idx].dst_ip_addr, &dst_v6address, 16); + rules[rule_idx].dst_ip_prefix_len = dst_prefix_length; + rules[rule_idx].is_ipv6 = 1; + } + else if (unformat (i, "sport %d-%d", &port1, &port2)) + { + vec_validate_acl_rules(rules, rule_idx); + rules[rule_idx].srcport_or_icmptype_first = htons(port1); + rules[rule_idx].srcport_or_icmptype_last = htons(port2); + } + else if (unformat (i, "sport %d", &port1)) + { + vec_validate_acl_rules(rules, rule_idx); + rules[rule_idx].srcport_or_icmptype_first = htons(port1); + rules[rule_idx].srcport_or_icmptype_last = htons(port1); + } + else if (unformat (i, "dport %d-%d", &port1, &port2)) + { + vec_validate_acl_rules(rules, rule_idx); + rules[rule_idx].dstport_or_icmpcode_first = htons(port1); + rules[rule_idx].dstport_or_icmpcode_last = htons(port2); + } + else if (unformat (i, "dport %d", &port1)) + { + vec_validate_acl_rules(rules, rule_idx); + rules[rule_idx].dstport_or_icmpcode_first = htons(port1); + rules[rule_idx].dstport_or_icmpcode_last = htons(port1); + } + else if (unformat (i, "tcpflags %d %d", &tcpflags, &tcpmask)) + { + vec_validate_acl_rules(rules, rule_idx); + rules[rule_idx].tcp_flags_value = tcpflags; + rules[rule_idx].tcp_flags_mask = tcpmask; + } + else if (unformat (i, "proto %d", &proto)) + { + vec_validate_acl_rules(rules, rule_idx); + rules[rule_idx].proto = proto; + } + else if (unformat (i, "tag %s", &tag)) + { + } + else if (unformat (i, ",")) + { + rule_idx++; + vec_validate_acl_rules(rules, rule_idx); + } + else + break; + } + + /* Construct the API message */ + vam->result_ready = 0; + + if(rules) + n_rules = vec_len(rules); + else + n_rules = 0; + + msg_size += n_rules*sizeof(rules[0]); + + mp = vl_msg_api_alloc_as_if_client(msg_size); + memset (mp, 0, msg_size); + mp->_vl_msg_id = ntohs (VL_API_ACL_ADD_REPLACE + sm->msg_id_base); + mp->client_index = vam->my_client_index; + if (n_rules > 0) + clib_memcpy(mp->r, rules, n_rules*sizeof (vl_api_acl_rule_t)); + if (tag) + { + if (vec_len(tag) >= sizeof(mp->tag)) + { + tag[sizeof(mp->tag)-1] = 0; + _vec_len(tag) = sizeof(mp->tag); + } + clib_memcpy(mp->tag, tag, vec_len(tag)); + vec_free(tag); + } + mp->acl_index = ntohl(acl_index); + mp->count = htonl(n_rules); + + /* send it... */ + S; + + /* Wait for a reply... */ + W; +} + +static int api_acl_del (vat_main_t * vam) +{ + acl_test_main_t * sm = &acl_test_main; + unformat_input_t * i = vam->input; + f64 timeout; + vl_api_acl_del_t * mp; + u32 acl_index = ~0; + + if (!unformat (i, "%d", &acl_index)) { + errmsg ("missing acl index\n"); + return -99; + } + + /* Construct the API message */ + M(ACL_DEL, acl_del); + mp->acl_index = ntohl(acl_index); + + /* send it... */ + S; + + /* Wait for a reply... */ + W; +} + +static int api_macip_acl_del (vat_main_t * vam) +{ + acl_test_main_t * sm = &acl_test_main; + unformat_input_t * i = vam->input; + f64 timeout; + vl_api_acl_del_t * mp; + u32 acl_index = ~0; + + if (!unformat (i, "%d", &acl_index)) { + errmsg ("missing acl index\n"); + return -99; + } + + /* Construct the API message */ + M(MACIP_ACL_DEL, acl_del); + mp->acl_index = ntohl(acl_index); + + /* send it... */ + S; + + /* Wait for a reply... */ + W; +} + +static int api_acl_interface_add_del (vat_main_t * vam) +{ + acl_test_main_t * sm = &acl_test_main; + unformat_input_t * i = vam->input; + f64 timeout; + vl_api_acl_interface_add_del_t * mp; + u32 sw_if_index = ~0; + u32 acl_index = ~0; + u8 is_input = 0; + u8 is_add = 0; + +// acl_interface_add_del | sw_if_index acl_index [out] [del] + + while (unformat_check_input (i) != UNFORMAT_END_OF_INPUT) + { + if (unformat (i, "%d", &acl_index)) + ; + else + break; + } + + + /* Parse args required to build the message */ + while (unformat_check_input (i) != UNFORMAT_END_OF_INPUT) { + if (unformat (i, "%U", unformat_sw_if_index, vam, &sw_if_index)) + ; + else if (unformat (i, "sw_if_index %d", &sw_if_index)) + ; + else if (unformat (i, "add")) + is_add = 1; + else if (unformat (i, "del")) + is_add = 0; + else if (unformat (i, "acl %d", &acl_index)) + ; + else if (unformat (i, "input")) + is_input = 1; + else if (unformat (i, "output")) + is_input = 0; + else + break; + } + + if (sw_if_index == ~0) { + errmsg ("missing interface name / explicit sw_if_index number \n"); + return -99; + } + + if (acl_index == ~0) { + errmsg ("missing ACL index\n"); + return -99; + } + + + + /* Construct the API message */ + M(ACL_INTERFACE_ADD_DEL, acl_interface_add_del); + mp->acl_index = ntohl(acl_index); + mp->sw_if_index = ntohl(sw_if_index); + mp->is_add = is_add; + mp->is_input = is_input; + + /* send it... */ + S; + + /* Wait for a reply... */ + W; +} + +static int api_macip_acl_interface_add_del (vat_main_t * vam) +{ + acl_test_main_t * sm = &acl_test_main; + unformat_input_t * i = vam->input; + f64 timeout; + vl_api_macip_acl_interface_add_del_t * mp; + u32 sw_if_index = ~0; + u32 acl_index = ~0; + u8 is_add = 0; + + /* Parse args required to build the message */ + while (unformat_check_input (i) != UNFORMAT_END_OF_INPUT) { + if (unformat (i, "%U", unformat_sw_if_index, vam, &sw_if_index)) + ; + else if (unformat (i, "sw_if_index %d", &sw_if_index)) + ; + else if (unformat (i, "add")) + is_add = 1; + else if (unformat (i, "del")) + is_add = 0; + else if (unformat (i, "acl %d", &acl_index)) + ; + else + break; + } + + if (sw_if_index == ~0) { + errmsg ("missing interface name / explicit sw_if_index number \n"); + return -99; + } + + if (acl_index == ~0) { + errmsg ("missing ACL index\n"); + return -99; + } + + + + /* Construct the API message */ + M(MACIP_ACL_INTERFACE_ADD_DEL, macip_acl_interface_add_del); + mp->acl_index = ntohl(acl_index); + mp->sw_if_index = ntohl(sw_if_index); + mp->is_add = is_add; + + /* send it... */ + S; + + /* Wait for a reply... */ + W; +} + +static int api_acl_interface_set_acl_list (vat_main_t * vam) +{ + acl_test_main_t * sm = &acl_test_main; + unformat_input_t * i = vam->input; + f64 timeout; + vl_api_acl_interface_set_acl_list_t * mp; + u32 sw_if_index = ~0; + u32 acl_index = ~0; + u32 *inacls = 0; + u32 *outacls = 0; + u8 is_input = 0; + +// acl_interface_set_acl_list | sw_if_index input [acl-idx list] output [acl-idx list] + + /* Parse args required to build the message */ + while (unformat_check_input (i) != UNFORMAT_END_OF_INPUT) { + if (unformat (i, "%U", unformat_sw_if_index, vam, &sw_if_index)) + ; + else if (unformat (i, "sw_if_index %d", &sw_if_index)) + ; + else if (unformat (i, "%d", &acl_index)) + { + if(is_input) + vec_add1(inacls, htonl(acl_index)); + else + vec_add1(outacls, htonl(acl_index)); + } + else if (unformat (i, "acl %d", &acl_index)) + ; + else if (unformat (i, "input")) + is_input = 1; + else if (unformat (i, "output")) + is_input = 0; + else + break; + } + + if (sw_if_index == ~0) { + errmsg ("missing interface name / explicit sw_if_index number \n"); + return -99; + } + + /* Construct the API message */ + M2(ACL_INTERFACE_SET_ACL_LIST, acl_interface_set_acl_list, sizeof(u32) * (vec_len(inacls) + vec_len(outacls))); + mp->sw_if_index = ntohl(sw_if_index); + mp->n_input = vec_len(inacls); + mp->count = vec_len(inacls) + vec_len(outacls); + vec_append(inacls, outacls); + if (vec_len(inacls) > 0) + clib_memcpy(mp->acls, inacls, vec_len(inacls)*sizeof(u32)); + + /* send it... */ + S; + + /* Wait for a reply... */ + W; +} + + +static int api_acl_interface_list_dump (vat_main_t * vam) +{ + acl_test_main_t * sm = &acl_test_main; + unformat_input_t * i = vam->input; + f64 timeout; + u32 sw_if_index = ~0; + vl_api_acl_interface_list_dump_t * mp; + + /* Parse args required to build the message */ + while (unformat_check_input (i) != UNFORMAT_END_OF_INPUT) { + if (unformat (i, "%U", unformat_sw_if_index, vam, &sw_if_index)) + ; + else if (unformat (i, "sw_if_index %d", &sw_if_index)) + ; + else + break; + } + + /* Construct the API message */ + M(ACL_INTERFACE_LIST_DUMP, acl_interface_list_dump); + mp->sw_if_index = ntohl (sw_if_index); + + /* send it... */ + S; + + /* Wait for a reply... */ + W; +} + +static int api_acl_dump (vat_main_t * vam) +{ + acl_test_main_t * sm = &acl_test_main; + unformat_input_t * i = vam->input; + f64 timeout; + u32 acl_index = ~0; + vl_api_acl_dump_t * mp; + + /* Parse args required to build the message */ + while (unformat_check_input (i) != UNFORMAT_END_OF_INPUT) { + if (unformat (i, "%d", &acl_index)) + ; + else + break; + } + + /* Construct the API message */ + M(ACL_DUMP, acl_dump); + mp->acl_index = ntohl (acl_index); + + /* send it... */ + S; + + /* Wait for a reply... */ + W; +} + +static int api_macip_acl_dump (vat_main_t * vam) +{ + acl_test_main_t * sm = &acl_test_main; + unformat_input_t * i = vam->input; + f64 timeout; + u32 acl_index = ~0; + vl_api_acl_dump_t * mp; + + /* Parse args required to build the message */ + while (unformat_check_input (i) != UNFORMAT_END_OF_INPUT) { + if (unformat (i, "%d", &acl_index)) + ; + else + break; + } + + /* Construct the API message */ + M(MACIP_ACL_DUMP, macip_acl_dump); + mp->acl_index = ntohl (acl_index); + + /* send it... */ + S; + + /* Wait for a reply... */ + W; +} + +#define vec_validate_macip_acl_rules(v, idx) \ + do { \ + if (vec_len(v) < idx+1) { \ + vec_validate(v, idx); \ + v[idx].is_permit = 0x1; \ + } \ + } while (0) + + +static int api_macip_acl_add (vat_main_t * vam) +{ + acl_test_main_t * sm = &acl_test_main; + unformat_input_t * i = vam->input; + f64 timeout; + vl_api_macip_acl_add_t * mp; + u32 msg_size = sizeof (*mp); /* without the rules */ + + vl_api_macip_acl_rule_t *rules = 0; + int rule_idx = 0; + int n_rules = 0; + u32 src_prefix_length = 0; + u32 action = 0; + ip4_address_t src_v4address; + ip6_address_t src_v6address; + u8 src_mac[6]; + u8 *tag = 0; + u8 mac_mask_all_1[6] = { 0xff, 0xff, 0xff, 0xff, 0xff, 0xff }; + + while (unformat_check_input (i) != UNFORMAT_END_OF_INPUT) + { + if (unformat (i, "ipv6")) + { + vec_validate_macip_acl_rules(rules, rule_idx); + rules[rule_idx].is_ipv6 = 1; + } + else if (unformat (i, "ipv4")) + { + vec_validate_macip_acl_rules(rules, rule_idx); + rules[rule_idx].is_ipv6 = 1; + } + else if (unformat (i, "permit")) + { + vec_validate_macip_acl_rules(rules, rule_idx); + rules[rule_idx].is_permit = 1; + } + else if (unformat (i, "deny")) + { + vec_validate_macip_acl_rules(rules, rule_idx); + rules[rule_idx].is_permit = 0; + } + else if (unformat (i, "action %d", &action)) + { + vec_validate_macip_acl_rules(rules, rule_idx); + rules[rule_idx].is_permit = action; + } + else if (unformat (i, "ip %U/%d", + unformat_ip4_address, &src_v4address, &src_prefix_length)) + { + vec_validate_macip_acl_rules(rules, rule_idx); + memcpy (rules[rule_idx].src_ip_addr, &src_v4address, 4); + rules[rule_idx].src_ip_prefix_len = src_prefix_length; + rules[rule_idx].is_ipv6 = 0; + } + else if (unformat (i, "ip %U/%d", + unformat_ip6_address, &src_v6address, &src_prefix_length)) + { + vec_validate_macip_acl_rules(rules, rule_idx); + memcpy (rules[rule_idx].src_ip_addr, &src_v6address, 16); + rules[rule_idx].src_ip_prefix_len = src_prefix_length; + rules[rule_idx].is_ipv6 = 1; + } + else if (unformat (i, "mac %U", + my_unformat_mac_address, &src_mac)) + { + vec_validate_macip_acl_rules(rules, rule_idx); + memcpy (rules[rule_idx].src_mac, &src_mac, 6); + memcpy (rules[rule_idx].src_mac_mask, &mac_mask_all_1, 6); + } + else if (unformat (i, "mask %U", + my_unformat_mac_address, &src_mac)) + { + vec_validate_macip_acl_rules(rules, rule_idx); + memcpy (rules[rule_idx].src_mac_mask, &src_mac, 6); + } + else if (unformat (i, "tag %s", &tag)) + { + } + else if (unformat (i, ",")) + { + rule_idx++; + vec_validate_macip_acl_rules(rules, rule_idx); + } + else + break; + } + + /* Construct the API message */ + vam->result_ready = 0; + + if(rules) + n_rules = vec_len(rules); + else + n_rules = 0; + + msg_size += n_rules*sizeof(rules[0]); + + mp = vl_msg_api_alloc_as_if_client(msg_size); + memset (mp, 0, msg_size); + mp->_vl_msg_id = ntohs (VL_API_MACIP_ACL_ADD + sm->msg_id_base); + mp->client_index = vam->my_client_index; + if (n_rules > 0) + clib_memcpy(mp->r, rules, n_rules*sizeof (mp->r[0])); + if (tag) + { + if (vec_len(tag) >= sizeof(mp->tag)) + { + tag[sizeof(mp->tag)-1] = 0; + _vec_len(tag) = sizeof(mp->tag); + } + clib_memcpy(mp->tag, tag, vec_len(tag)); + vec_free(tag); + } + + mp->count = htonl(n_rules); + + /* send it... */ + S; + + /* Wait for a reply... */ + W; +} + +/* + * List of messages that the api test plugin sends, + * and that the data plane plugin processes + */ +#define foreach_vpe_api_msg \ +_(acl_plugin_get_version, "") \ +_(acl_add_replace, " [ [src IP/plen] [dst IP/plen] [sport X-Y] [dport X-Y] [proto P] [tcpflags FL MASK], ... , ...") \ +_(acl_del, "") \ +_(acl_dump, "[]") \ +_(acl_interface_add_del, " | sw_if_index [add|del] [input|output] acl ") \ +_(acl_interface_set_acl_list, " | sw_if_index input [acl-idx list] output [acl-idx list]") \ +_(acl_interface_list_dump, "[ | sw_if_index ]") \ +_(macip_acl_add, "...") \ +_(macip_acl_del, "")\ +_(macip_acl_dump, "[]") \ +_(macip_acl_interface_add_del, " | sw_if_index [add|del] acl ") \ +_(macip_acl_interface_get, "") + + + +void vat_api_hookup (vat_main_t *vam) +{ + acl_test_main_t * sm = &acl_test_main; + /* Hook up handlers for replies from the data plane plug-in */ +#define _(N,n) \ + vl_msg_api_set_handlers((VL_API_##N + sm->msg_id_base), \ + #n, \ + vl_api_##n##_t_handler, \ + vl_noop_handler, \ + vl_api_##n##_t_endian, \ + vl_api_##n##_t_print, \ + sizeof(vl_api_##n##_t), 1); + foreach_vpe_api_reply_msg; +#undef _ + + /* API messages we can send */ +#define _(n,h) hash_set_mem (vam->function_by_name, #n, api_##n); + foreach_vpe_api_msg; +#undef _ + + /* Help strings */ +#define _(n,h) hash_set_mem (vam->help_by_name, #n, h); + foreach_vpe_api_msg; +#undef _ +} + +clib_error_t * vat_plugin_register (vat_main_t *vam) +{ + acl_test_main_t * sm = &acl_test_main; + u8 * name; + + sm->vat_main = vam; + + name = format (0, "acl_%08x%c", api_version, 0); + sm->msg_id_base = vl_client_get_first_plugin_msg_id ((char *) name); + + if (sm->msg_id_base != (u16) ~0) + vat_api_hookup (vam); + + vec_free(name); + + return 0; +} diff --git a/src/plugins/acl/l2sess.c b/src/plugins/acl/l2sess.c new file mode 100644 index 00000000..cc9bde44 --- /dev/null +++ b/src/plugins/acl/l2sess.c @@ -0,0 +1,243 @@ +/* + * Copyright (c) 2016 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ +/* + *------------------------------------------------------------------ + * l2sess.c - simple MAC-swap API / debug CLI handling + *------------------------------------------------------------------ + */ + +#include +#include +#include + +#include +#include +#include +#include + +#include +#include + +void +l2sess_vlib_plugin_register (vlib_main_t * vm, void* hh, + int from_early_init) +{ + l2sess_main_t *sm = &l2sess_main; + vnet_plugin_handoff_t * h = hh; + memset (sm, 0, sizeof (*sm)); + + sm->vlib_main = vm; + sm->vnet_main = h->vnet_main; + sm->ethernet_main = h->ethernet_main; +} + +void +l2sess_init_next_features_input (vlib_main_t * vm, l2sess_main_t * sm) +{ +#define _(node_name, node_var, is_out, is_ip6, is_track) \ + if (!is_out) feat_bitmap_init_next_nodes(vm, node_var.index, L2INPUT_N_FEAT, l2input_get_feat_names (), sm->node_var ## _input_next_node_index); + foreach_l2sess_node +#undef _ +} + +void +l2sess_add_our_next_nodes (vlib_main_t * vm, l2sess_main_t * sm, + u8 * prev_node_name, int add_output_nodes) +{ + vlib_node_t *n; + n = vlib_get_node_by_name (vm, prev_node_name); +#define _(node_name, node_var, is_out, is_ip6, is_track) \ + if (is_out == add_output_nodes) { \ + u32 idx = vlib_node_add_next_with_slot(vm, n->index, node_var.index, ~0); \ + if (is_track) { \ + sm->next_slot_track_node_by_is_ip6_is_out[is_ip6][is_out] = idx; \ + } \ + } + foreach_l2sess_node +#undef _ +} + +void +l2sess_setup_nodes (void) +{ + vlib_main_t *vm = vlib_get_main (); + l2sess_main_t *sm = &l2sess_main; + + l2sess_init_next_features_input (vm, sm); + + l2sess_add_our_next_nodes (vm, sm, (u8 *) "l2-input-classify", 0); + l2sess_add_our_next_nodes (vm, sm, (u8 *) "l2-output-classify", 1); + +} + +static char * +get_l4_proto_str (int is_ip6, uint8_t l4_proto) +{ + switch (l4_proto) + { + case 6: + return "tcp"; + case 17: + return "udp"; + case 1: + return "icmp"; + case 58: + return "icmp6"; + default: + return ""; + } +} + +static clib_error_t * +l2sess_show_command_fn (vlib_main_t * vm, + unformat_input_t * input, vlib_cli_command_t * cmd) +{ + l2sess_main_t *sm = &l2sess_main; + clib_time_t *ct = &vm->clib_time; + l2s_session_t *s; + u64 now = clib_cpu_time_now (); + + vlib_cli_output (vm, "Timing wheel info: \n%U", format_timing_wheel, + &sm->timing_wheel, 255); + + pool_foreach (s, sm->sessions, ( + { + f64 ctime = + (now - + s->create_time) * ct->seconds_per_clock; + f64 atime0 = + (now - + s->side[0].active_time) * + ct->seconds_per_clock; + f64 atime1 = + (now - + s->side[1].active_time) * + ct->seconds_per_clock; +/* + f64 ctime = (s->create_time - vm->cpu_time_main_loop_start) * ct->seconds_per_clock; + f64 atime0 = (s->side[0].active_time - vm->cpu_time_main_loop_start) * ct->seconds_per_clock; + f64 atime1 = (s->side[1].active_time - vm->cpu_time_main_loop_start) * ct->seconds_per_clock; +*/ + u8 * out0 = + format (0, + "%5d: create time: %U pkts/bytes/active time: [ %ld %ld %U : %ld %ld %U ]\n", + (s - sm->sessions), + format_time_interval, "h:m:s:u", + ctime, s->side[0].n_packets, + s->side[0].n_bytes, + format_time_interval, "h:m:s:u", + atime0, s->side[1].n_packets, + s->side[1].n_bytes, + format_time_interval, "h:m:s:u", + atime1); u8 * out1 = 0; + if (s->is_ip6) + { + out1 = + format (0, "%s %U :%u <-> %U :%u", + get_l4_proto_str (s->is_ip6, + s->l4_proto), + format_ip6_address, + &s->side[0].addr.ip6, + s->side[0].port, + format_ip6_address, + &s->side[1].addr.ip6, + s->side[1].port);} + else + { + out1 = + format (0, "%s %U :%u <-> %U :%u", + get_l4_proto_str (s->is_ip6, + s->l4_proto), + format_ip4_address, + &s->side[0].addr.ip4, + s->side[0].port, + format_ip4_address, + &s->side[1].addr.ip4, + s->side[1].port);} + vlib_cli_output (vm, "%s %s", out0, + out1); vec_free (out0); + vec_free (out1);} + )); + return 0; +} + +static clib_error_t * +l2sess_show_count_command_fn (vlib_main_t * vm, + unformat_input_t * input, + vlib_cli_command_t * cmd) +{ + l2sess_main_t *sm = &l2sess_main; + + vlib_cli_output (vm, "Timing wheel info: \n%U", format_timing_wheel, + &sm->timing_wheel, 255); + vlib_cli_output (vm, "session pool len: %d, pool elts: %d", + pool_len (sm->sessions), pool_elts (sm->sessions)); + vlib_cli_output (vm, + "attempted to delete sessions which were already free: %d", + sm->counter_attempted_delete_free_session); + return 0; +} + + +/* *INDENT-OFF* */ +VLIB_CLI_COMMAND (l2sess_show_command, static) = { + .path = "show l2sess", + .short_help = "show l2sess", + .function = l2sess_show_command_fn, +}; + +VLIB_CLI_COMMAND (l2sess_show_count_command, static) = { + .path = "show l2sess count", + .short_help = "show l2sess count", + .function = l2sess_show_count_command_fn, +}; +/* *INDENT-OFF* */ + +static inline u64 +time_sec_to_clock( clib_time_t *ct, f64 sec) +{ + return (u64)(((f64)sec)/ct->seconds_per_clock); +} + +static clib_error_t * l2sess_init (vlib_main_t * vm) +{ + l2sess_main_t * sm = &l2sess_main; + clib_error_t * error = 0; + u64 cpu_time_now = clib_cpu_time_now(); + + + clib_time_t *ct = &vm->clib_time; + sm->udp_session_idle_timeout = time_sec_to_clock(ct, UDP_SESSION_IDLE_TIMEOUT_SEC); + sm->tcp_session_idle_timeout = time_sec_to_clock(ct, TCP_SESSION_IDLE_TIMEOUT_SEC); + sm->tcp_session_transient_timeout = time_sec_to_clock(ct, TCP_SESSION_TRANSIENT_TIMEOUT_SEC); + + /* The min sched time of 10e-1 causes erroneous behavior... */ + sm->timing_wheel.min_sched_time = 10e-2; + sm->timing_wheel.max_sched_time = 3600.0*48.0; + timing_wheel_init (&sm->timing_wheel, cpu_time_now, vm->clib_time.clocks_per_second); + sm->timer_wheel_next_expiring_time = 0; + sm->timer_wheel_tick = time_sec_to_clock(ct, sm->timing_wheel.min_sched_time); + /* Pre-allocate expired nodes. */ + vec_alloc (sm->data_from_advancing_timing_wheel, 32); + + l2sess_setup_nodes(); + l2output_init_output_node_vec (&sm->output_next_nodes.output_node_index_vec); + + return error; +} + +VLIB_INIT_FUNCTION (l2sess_init); + + diff --git a/src/plugins/acl/l2sess.h b/src/plugins/acl/l2sess.h new file mode 100644 index 00000000..db899917 --- /dev/null +++ b/src/plugins/acl/l2sess.h @@ -0,0 +1,150 @@ +/* + * Copyright (c) 2016 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ +#ifndef __included_l2sess_h__ +#define __included_l2sess_h__ + +#include +#include +#include + +#include +#include +#include +#include + +#include +#include + +#define _(node_name, node_var, is_out, is_ip6, is_track) +#undef _ +#define foreach_l2sess_node \ + _("aclp-l2s-input-ip4-add", l2sess_in_ip4_add, 0, 0, 0) \ + _("aclp-l2s-input-ip6-add", l2sess_in_ip6_add, 0, 1, 0) \ + _("aclp-l2s-output-ip4-add", l2sess_out_ip4_add, 1, 0, 0) \ + _("aclp-l2s-output-ip6-add", l2sess_out_ip6_add, 1, 1, 0) \ + _("aclp-l2s-input-ip4-track", l2sess_in_ip4_track, 0, 0, 1) \ + _("aclp-l2s-input-ip6-track", l2sess_in_ip6_track, 0, 1, 1) \ + _("aclp-l2s-output-ip4-track",l2sess_out_ip4_track, 1, 0, 1) \ + _("aclp-l2s-output-ip6-track", l2sess_out_ip6_track, 1, 1, 1) + +#define _(node_name, node_var, is_out, is_ip6, is_track) \ + extern vlib_node_registration_t node_var; +foreach_l2sess_node +#undef _ + +#define TCP_FLAG_FIN 0x01 +#define TCP_FLAG_SYN 0x02 +#define TCP_FLAG_RST 0x04 +#define TCP_FLAG_PUSH 0x08 +#define TCP_FLAG_ACK 0x10 +#define TCP_FLAG_URG 0x20 +#define TCP_FLAG_ECE 0x40 +#define TCP_FLAG_CWR 0x80 +#define TCP_FLAGS_RSTFINACKSYN (TCP_FLAG_RST + TCP_FLAG_FIN + TCP_FLAG_SYN + TCP_FLAG_ACK) +#define TCP_FLAGS_ACKSYN (TCP_FLAG_SYN + TCP_FLAG_ACK) + +typedef struct { + ip46_address_t addr; + u64 active_time; + u64 n_packets; + u64 n_bytes; + u16 port; +} l2s_session_side_t; + +enum { + L2S_SESSION_SIDE_IN = 0, + L2S_SESSION_SIDE_OUT, + L2S_N_SESSION_SIDES +}; + +typedef struct { + u64 create_time; + l2s_session_side_t side[L2S_N_SESSION_SIDES]; + u8 l4_proto; + u8 is_ip6; + u16 tcp_flags_seen; /* u16 because of two sides */ +} l2s_session_t; + +#define PROD +#ifdef PROD +#define UDP_SESSION_IDLE_TIMEOUT_SEC 600 +#define TCP_SESSION_IDLE_TIMEOUT_SEC (3600*24) +#define TCP_SESSION_TRANSIENT_TIMEOUT_SEC 120 +#else +#define UDP_SESSION_IDLE_TIMEOUT_SEC 15 +#define TCP_SESSION_IDLE_TIMEOUT_SEC 15 +#define TCP_SESSION_TRANSIENT_TIMEOUT_SEC 5 +#endif + +typedef struct { + /* + * the next two fields are present for all nodes, but + * only one of them is used per node - depending + * on whether the node is an input or output one. + */ +#define _(node_name, node_var, is_out, is_ip6, is_track) \ + u32 node_var ## _input_next_node_index[32]; \ + l2_output_next_nodes_st node_var ## _next_nodes; +foreach_l2sess_node +#undef _ + l2_output_next_nodes_st output_next_nodes; + + /* Next indices of the tracker nodes */ + u32 next_slot_track_node_by_is_ip6_is_out[2][2]; + + /* + * Pairing of "forward" and "reverse" tables by table index. + * Each relationship has two entries - for one and the other table, + * so it is bidirectional. + */ + + u32 *fwd_to_rev_by_table_index; + + /* + * The vector of per-interface session pools + */ + + l2s_session_t *sessions; + + /* The session timeouts */ + u64 tcp_session_transient_timeout; + u64 tcp_session_idle_timeout; + u64 udp_session_idle_timeout; + + /* Timing wheel to time out the idle sessions */ + timing_wheel_t timing_wheel; + u32 *data_from_advancing_timing_wheel; + u64 timer_wheel_next_expiring_time; + u64 timer_wheel_tick; + + /* convenience */ + vlib_main_t * vlib_main; + vnet_main_t * vnet_main; + ethernet_main_t * ethernet_main; + + /* Counter(s) */ + u64 counter_attempted_delete_free_session; +} l2sess_main_t; + +l2sess_main_t l2sess_main; + +/* Just exposed for acl.c */ + +void +l2sess_vlib_plugin_register (vlib_main_t * vm, void * hh, + int from_early_init); + + +#endif /* __included_l2sess_h__ */ diff --git a/src/plugins/acl/l2sess_node.c b/src/plugins/acl/l2sess_node.c new file mode 100644 index 00000000..520e5929 --- /dev/null +++ b/src/plugins/acl/l2sess_node.c @@ -0,0 +1,816 @@ +/* + * Copyright (c) 2016 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ +#include +#include +#include +#include +#include +#include +#include + + +typedef struct +{ + u32 next_index; + u32 sw_if_index; + u32 trace_flags; + u32 session_tables[2]; + u32 session_nexts[2]; + u8 l4_proto; +} l2sess_trace_t; + +/* packet trace format function */ + +#define _(node_name, node_var, is_out, is_ip6, is_track) \ +static u8 * format_## node_var ##_trace (u8 * s, va_list * args) \ +{ \ + CLIB_UNUSED (vlib_main_t * vm) = va_arg (*args, vlib_main_t *); \ + CLIB_UNUSED (vlib_node_t * node) = va_arg (*args, vlib_node_t *); \ + l2sess_trace_t * t = va_arg (*args, l2sess_trace_t *); \ + \ + s = format (s, node_name ": sw_if_index %d, next index %d trace_flags %08x L4 proto %d\n" \ + " tables [ %d, %d ] nexts [ %d, %d ]", \ + t->sw_if_index, t->next_index, t->trace_flags, t->l4_proto, \ + t->session_tables[0], t->session_tables[1], \ + t->session_nexts[0], t->session_nexts[1]); \ + return s; \ +} +foreach_l2sess_node +#undef _ +#define foreach_l2sess_error \ +_(SWAPPED, "Mac swap packets processed") + typedef enum +{ +#define _(sym,str) L2SESS_ERROR_##sym, + foreach_l2sess_error +#undef _ + L2SESS_N_ERROR, +} l2sess_error_t; + +static char *l2sess_error_strings[] = { +#define _(sym,string) string, + foreach_l2sess_error +#undef _ +}; + +typedef enum +{ + L2SESS_NEXT_DROP, + L2SESS_N_NEXT, +} l2sess_next_t; + +u8 +l2sess_get_l4_proto (vlib_buffer_t * b0, int node_is_ip6) +{ + u8 proto; + int proto_offset; + if (node_is_ip6) + { + proto_offset = 20; + } + else + { + proto_offset = 23; + } + proto = *((u8 *) vlib_buffer_get_current (b0) + proto_offset); + return proto; +} + + +u8 +l2sess_get_tcp_flags (vlib_buffer_t * b0, int node_is_ip6) +{ + u8 flags; + int flags_offset; + if (node_is_ip6) + { + flags_offset = 14 + 40 + 13; /* FIXME: no extension headers assumed */ + } + else + { + flags_offset = 14 + 20 + 13; + } + flags = *((u8 *) vlib_buffer_get_current (b0) + flags_offset); + return flags; +} + +static inline int +l4_tcp_or_udp (u8 proto) +{ + return ((proto == 6) || (proto == 17)); +} + +void +l2sess_get_session_tables (l2sess_main_t * sm, u32 sw_if_index, + int node_is_out, int node_is_ip6, u8 l4_proto, + u32 * session_tables) +{ +/* + * Based on the direction, l3 and l4 protocol, fill a u32[2] array: + * [0] is index for the "direct match" path, [1] is for "mirrored match". + * Store the indices of the tables to add the session to in session_tables[] + */ + l2_output_classify_main_t *l2om = &l2_output_classify_main; + l2_input_classify_main_t *l2im = &l2_input_classify_main; + + u32 output_table_index; + u32 input_table_index; + + if (!l4_tcp_or_udp (l4_proto)) + { + return; + } + + if (node_is_ip6) + { + vec_validate_init_empty (l2im-> + classify_table_index_by_sw_if_index + [L2_INPUT_CLASSIFY_TABLE_IP6], sw_if_index, + ~0); + input_table_index = + l2im-> + classify_table_index_by_sw_if_index[L2_INPUT_CLASSIFY_TABLE_IP6] + [sw_if_index]; + vec_validate_init_empty (l2om-> + classify_table_index_by_sw_if_index + [L2_OUTPUT_CLASSIFY_TABLE_IP6], sw_if_index, + ~0); + output_table_index = + l2om-> + classify_table_index_by_sw_if_index[L2_OUTPUT_CLASSIFY_TABLE_IP6] + [sw_if_index]; + } + else + { + vec_validate_init_empty (l2im-> + classify_table_index_by_sw_if_index + [L2_INPUT_CLASSIFY_TABLE_IP4], sw_if_index, + ~0); + input_table_index = + l2im-> + classify_table_index_by_sw_if_index[L2_INPUT_CLASSIFY_TABLE_IP4] + [sw_if_index]; + vec_validate_init_empty (l2om-> + classify_table_index_by_sw_if_index + [L2_OUTPUT_CLASSIFY_TABLE_IP4], sw_if_index, + ~0); + output_table_index = + l2om-> + classify_table_index_by_sw_if_index[L2_OUTPUT_CLASSIFY_TABLE_IP4] + [sw_if_index]; + } + + if (node_is_out) + { + session_tables[0] = output_table_index; + session_tables[1] = input_table_index; + } + else + { + session_tables[0] = input_table_index; + session_tables[1] = output_table_index; + } +} + +void +l2sess_get_session_nexts (l2sess_main_t * sm, u32 sw_if_index, + int node_is_out, int node_is_ip6, u8 l4_proto, + u32 * session_nexts) +{ +/* + * Based on the direction, l3 and l4 protocol, fill a u32[2] array: + * [0] is the index for the "direct match" path, [1] is for "mirrored match". + * Store the match_next_index in session_nexts[] for a new session entry which is being added to session tables. + */ + u32 input_node_index; + u32 output_node_index; + + if (!l4_tcp_or_udp (l4_proto)) + { + return; + } + + input_node_index = + sm->next_slot_track_node_by_is_ip6_is_out[node_is_ip6][0]; + output_node_index = + sm->next_slot_track_node_by_is_ip6_is_out[node_is_ip6][1]; + + if (node_is_out) + { + session_nexts[0] = output_node_index; + session_nexts[1] = input_node_index; + } + else + { + session_nexts[0] = input_node_index; + session_nexts[1] = output_node_index; + } +} + + +static inline void +swap_bytes (vlib_buffer_t * b0, int off_a, int off_b, int nbytes) +{ + u8 tmp; + u8 *pa = vlib_buffer_get_current (b0) + off_a; + u8 *pb = vlib_buffer_get_current (b0) + off_b; + while (nbytes--) + { + tmp = *pa; + *pa++ = *pb; + *pb++ = tmp; + } +} + +/* + * This quite pro[bv]ably is a terrible idea performance wise. Moreso doing it twice. + * Would having a long (ish) chunk of memory work better for this ? + * We will see when we get to the performance of this. + */ +void +l2sess_flip_l3l4_fields (vlib_buffer_t * b0, int node_is_ip6, u8 l4_proto) +{ + if (!l4_tcp_or_udp (l4_proto)) + { + return; + } + if (node_is_ip6) + { + swap_bytes (b0, 22, 38, 16); /* L3 */ + swap_bytes (b0, 54, 56, 2); /* L4 (when no EH!) */ + } + else + { + swap_bytes (b0, 26, 30, 4); /* L3 */ + swap_bytes (b0, 34, 36, 2); /* L4 */ + } +} + +void +l2sess_add_session (vlib_buffer_t * b0, int node_is_out, int node_is_ip6, + u32 session_table, u32 session_match_next, + u32 opaque_index) +{ + vnet_classify_main_t *cm = &vnet_classify_main; + u32 action = 0; + u32 metadata = 0; + +#ifdef DEBUG_SESSIONS + printf ("Adding session to table %d with next %d\n", session_table, + session_match_next); +#endif + vnet_classify_add_del_session (cm, session_table, + vlib_buffer_get_current (b0), + session_match_next, opaque_index, 0, action, + metadata, 1); +} + + + +static void * +get_ptr_to_offset (vlib_buffer_t * b0, int offset) +{ + u8 *p = vlib_buffer_get_current (b0) + offset; + return p; +} + + +/* + * FIXME: Hardcoded offsets are ugly, although if casting to structs one + * would need to take care about alignment.. So let's for now be naive and simple. + */ + +void +session_store_ip4_l3l4_info (vlib_buffer_t * b0, l2s_session_t * sess, + int node_is_out) +{ + clib_memcpy (&sess->side[1 - node_is_out].addr.ip4, + get_ptr_to_offset (b0, 26), 4); + clib_memcpy (&sess->side[node_is_out].addr.ip4, get_ptr_to_offset (b0, 30), + 4); + sess->side[1 - node_is_out].port = + ntohs (*(u16 *) get_ptr_to_offset (b0, 34)); + sess->side[node_is_out].port = ntohs (*(u16 *) get_ptr_to_offset (b0, 36)); +} + +void +session_store_ip6_l3l4_info (vlib_buffer_t * b0, l2s_session_t * sess, + int node_is_out) +{ + clib_memcpy (&sess->side[1 - node_is_out].addr.ip6, + get_ptr_to_offset (b0, 22), 16); + clib_memcpy (&sess->side[node_is_out].addr.ip4, get_ptr_to_offset (b0, 38), + 16); + sess->side[1 - node_is_out].port = + ntohs (*(u16 *) get_ptr_to_offset (b0, 54)); + sess->side[node_is_out].port = ntohs (*(u16 *) get_ptr_to_offset (b0, 56)); +} + +static void +build_match_from_session (l2sess_main_t * sm, u8 * match, + l2s_session_t * sess, int is_out) +{ + if (sess->is_ip6) + { + match[20] = sess->l4_proto; + clib_memcpy (&match[22], &sess->side[1 - is_out].addr.ip6, 16); + clib_memcpy (&match[38], &sess->side[is_out].addr.ip4, 16); + *(u16 *) & match[54] = htons (sess->side[1 - is_out].port); + *(u16 *) & match[56] = htons (sess->side[is_out].port); + } + else + { + match[23] = sess->l4_proto; + clib_memcpy (&match[26], &sess->side[1 - is_out].addr.ip6, 4); + clib_memcpy (&match[30], &sess->side[is_out].addr.ip4, 4); + *(u16 *) & match[34] = htons (sess->side[1 - is_out].port); + *(u16 *) & match[36] = htons (sess->side[is_out].port); + } +} + +static void +delete_session (l2sess_main_t * sm, u32 sw_if_index, u32 session_index) +{ + vnet_classify_main_t *cm = &vnet_classify_main; + u8 match[5 * 16]; /* For building the mock of the packet to delete the classifier session */ + u32 session_tables[2] = { ~0, ~0 }; + l2s_session_t *sess = sm->sessions + session_index; + if (pool_is_free (sm->sessions, sess)) + { + sm->counter_attempted_delete_free_session++; + return; + } + l2sess_get_session_tables (sm, sw_if_index, 0, sess->is_ip6, sess->l4_proto, + session_tables); + if (session_tables[1] != ~0) + { + build_match_from_session (sm, match, sess, 1); + vnet_classify_add_del_session (cm, session_tables[1], match, 0, 0, 0, 0, + 0, 0); + } + if (session_tables[1] != ~0) + { + build_match_from_session (sm, match, sess, 1); + vnet_classify_add_del_session (cm, session_tables[1], match, 0, 0, 0, 0, + 0, 0); + } + pool_put (sm->sessions, sess); +} + +static void +udp_session_account_buffer (vlib_buffer_t * b0, l2s_session_t * s, + int which_side, u64 now) +{ + l2s_session_side_t *ss = &s->side[which_side]; + ss->active_time = now; + ss->n_packets++; + ss->n_bytes += b0->current_data + b0->current_length; +} + +static inline u64 +udp_session_get_timeout (l2sess_main_t * sm, l2s_session_t * sess, u64 now) +{ + return (sm->udp_session_idle_timeout); +} + +static void +tcp_session_account_buffer (vlib_buffer_t * b0, l2s_session_t * s, + int which_side, u64 now) +{ + l2s_session_side_t *ss = &s->side[which_side]; + ss->active_time = now; + ss->n_packets++; + ss->n_bytes += b0->current_data + b0->current_length; + /* Very very lightweight TCP state tracking: just record which flags were seen */ + s->tcp_flags_seen |= + l2sess_get_tcp_flags (b0, s->is_ip6) << (8 * which_side); +} + +/* + * Since we are tracking for the purposes of timing the sessions out, + * we mostly care about two states: established (maximize the idle timeouts) + * and transient (halfopen/halfclosed/reset) - we need to have a reasonably short timeout to + * quickly get rid of sessions but not short enough to violate the TCP specs. + */ + +static inline u64 +tcp_session_get_timeout (l2sess_main_t * sm, l2s_session_t * sess, u64 now) +{ + /* seen both SYNs and ACKs but not FINs means we are in establshed state */ + u16 masked_flags = + sess->tcp_flags_seen & ((TCP_FLAGS_RSTFINACKSYN << 8) + + TCP_FLAGS_RSTFINACKSYN); + if (((TCP_FLAGS_ACKSYN << 8) + TCP_FLAGS_ACKSYN) == masked_flags) + { + return (sm->tcp_session_idle_timeout); + } + else + { + return (sm->tcp_session_transient_timeout); + } +} + +static inline u64 +session_get_timeout (l2sess_main_t * sm, l2s_session_t * sess, u64 now) +{ + u64 timeout; + + switch (sess->l4_proto) + { + case 6: + timeout = tcp_session_get_timeout (sm, sess, now); + break; + case 17: + timeout = udp_session_get_timeout (sm, sess, now); + break; + default: + timeout = 0; + } + + return timeout; +} + +static inline u64 +get_session_last_active_time(l2s_session_t * sess) +{ + u64 last_active = + sess->side[0].active_time > + sess->side[1].active_time ? sess->side[0].active_time : sess->side[1]. + active_time; + return last_active; +} + +static int +session_is_alive (l2sess_main_t * sm, l2s_session_t * sess, u64 now, u64 *last_active_cache) +{ + u64 last_active = get_session_last_active_time(sess); + u64 timeout = session_get_timeout (sm, sess, now); + int is_alive = ((now - last_active) < timeout); + if (last_active_cache) + *last_active_cache = last_active; + return is_alive; +} + +static void +check_idle_sessions (l2sess_main_t * sm, u32 sw_if_index, u64 now) +{ + sm->timer_wheel_next_expiring_time = 0; + sm->data_from_advancing_timing_wheel + = + timing_wheel_advance (&sm->timing_wheel, now, + sm->data_from_advancing_timing_wheel, + &sm->timer_wheel_next_expiring_time); +#ifdef DEBUG_SESSIONS_VERBOSE + { + clib_time_t *ct = &sm->vlib_main->clib_time; + f64 ctime; + ctime = now * ct->seconds_per_clock; + clib_warning ("Now : %U", format_time_interval, "h:m:s:u", ctime); + ctime = sm->timer_wheel_next_expiring_time * ct->seconds_per_clock; + clib_warning ("Next expire: %U", format_time_interval, "h:m:s:u", ctime); + clib_warning ("Expired items: %d", + (int) vec_len (sm->data_from_advancing_timing_wheel)); + } +#endif + + sm->timer_wheel_next_expiring_time = now + sm->timer_wheel_tick; + if (PREDICT_FALSE ( 0 == sm->data_from_advancing_timing_wheel )) { + return; + } + + if (PREDICT_FALSE (_vec_len (sm->data_from_advancing_timing_wheel) > 0)) + { + uword i; + for (i = 0; i < _vec_len (sm->data_from_advancing_timing_wheel); i++) + { + u32 session_index = sm->data_from_advancing_timing_wheel[i]; + if (!pool_is_free_index (sm->sessions, session_index)) + { + l2s_session_t *sess = sm->sessions + session_index; + u64 last_active; + if (session_is_alive (sm, sess, now, &last_active)) + { +#ifdef DEBUG_SESSIONS + clib_warning ("Restarting timer for session %d", (int) session_index); +#endif + /* Pretend we did this in the past, at last_active moment */ + timing_wheel_insert (&sm->timing_wheel, + last_active + session_get_timeout (sm, sess, + last_active), + session_index); + } + else + { +#ifdef DEBUG_SESSIONS + clib_warning ("Deleting session %d", (int) session_index); +#endif + delete_session (sm, sw_if_index, session_index); + } + } + } + _vec_len (sm->data_from_advancing_timing_wheel) = 0; + } +} + +static uword +l2sess_node_fn (vlib_main_t * vm, + vlib_node_runtime_t * node, vlib_frame_t * frame) +{ + u32 n_left_from, *from, *to_next; + l2sess_next_t next_index; + u32 pkts_swapped = 0; + u32 cached_sw_if_index = (u32) ~ 0; + u32 cached_next_index = (u32) ~ 0; + u32 feature_bitmap0; + u32 trace_flags0; + + l2sess_main_t *sm = &l2sess_main; + + from = vlib_frame_vector_args (frame); + n_left_from = frame->n_vectors; + next_index = node->cached_next_index; + + while (n_left_from > 0) + { + u32 n_left_to_next; + + vlib_get_next_frame (vm, node, next_index, to_next, n_left_to_next); + + /* Only a single loop for now for simplicity */ + + while (n_left_from > 0 && n_left_to_next > 0) + { + u32 bi0; + vlib_buffer_t *b0; + u32 next0 = L2SESS_NEXT_DROP; + u32 sw_if_index0; + //ethernet_header_t *en0; + + /* speculatively enqueue b0 to the current next frame */ + bi0 = from[0]; + to_next[0] = bi0; + from += 1; + to_next += 1; + n_left_from -= 1; + n_left_to_next -= 1; + + b0 = vlib_get_buffer (vm, bi0); + //en0 = vlib_buffer_get_current (b0); + +/* + * The non-boilerplate is in the block below. + * Note first a magic macro block that sets up the behavior qualifiers: + * node_is_out : 1 = is output, 0 = is input + * node_is_ip6 : 1 = is ip6, 0 = is ip4 + * node_is_track : 1 = is a state tracking node, 0 - is a session addition node + * + * Subsequently the code adjusts its behavior depending on these variables. + * It's most probably not great performance wise but much easier to work with. + * + */ + { + int node_is_out = -1; + CLIB_UNUSED (int node_is_ip6) = -1; + CLIB_UNUSED (int node_is_track) = -1; + u32 node_index = 0; + u32 session_tables[2] = { ~0, ~0 }; + u32 session_nexts[2] = { ~0, ~0 }; + l2_output_next_nodes_st *next_nodes = 0; + u32 *input_feat_next_node_index; + u8 l4_proto; + u64 now = clib_cpu_time_now (); + +/* + * Set the variables according to which of the 8 nodes we are. + * Hopefully the compiler is smart enough to eliminate the extraneous. + */ +#define _(node_name, node_var, is_out, is_ip6, is_track) \ +if(node_var.index == node->node_index) \ + { \ + node_is_out = is_out; \ + node_is_ip6 = is_ip6; \ + node_is_track = is_track; \ + node_index = node_var.index; \ + next_nodes = &sm->node_var ## _next_nodes; \ + input_feat_next_node_index = sm->node_var ## _input_next_node_index; \ + } + foreach_l2sess_node +#undef _ + trace_flags0 = 0; + if (node_is_out) + { + sw_if_index0 = vnet_buffer (b0)->sw_if_index[VLIB_TX]; + } + else + { + sw_if_index0 = vnet_buffer (b0)->sw_if_index[VLIB_RX]; + } + /* potentially also remove the nodes here */ + feature_bitmap0 = vnet_buffer (b0)->l2.feature_bitmap; + + if (node_is_track) + { + u32 sess_index = vnet_buffer (b0)->l2_classify.opaque_index; + l2s_session_t *sess = sm->sessions + sess_index; + l4_proto = sess->l4_proto; + + if (session_is_alive (sm, sess, now, 0)) + { + if (6 == l4_proto) + { + tcp_session_account_buffer (b0, sess, node_is_out, + now); + } + else + { + udp_session_account_buffer (b0, sess, node_is_out, + now); + } + } + else + { + timing_wheel_delete (&sm->timing_wheel, sess_index); + delete_session (sm, sw_if_index0, sess_index); + /* FIXME: drop the packet that hit the obsolete node, for now. We really ought to recycle it. */ + next0 = 0; + } + } + else + { + /* + * "-add" node: take l2opaque which arrived to us, and deduce + * the tables out of that. ~0 means the topmost classifier table + * applied for this AF on the RX(for input)/TX(for output)) sw_if_index. + * Also add the mirrored session to the paired table. + */ + l2s_session_t *sess; + u32 sess_index; + + l4_proto = l2sess_get_l4_proto (b0, node_is_ip6); + + pool_get (sm->sessions, sess); + sess_index = sess - sm->sessions; + sess->create_time = now; + sess->side[node_is_out].active_time = now; + sess->side[1 - node_is_out].active_time = now; + sess->l4_proto = l4_proto; + sess->is_ip6 = node_is_ip6; + if (node_is_ip6) + { + session_store_ip6_l3l4_info (b0, sess, node_is_out); + } + else + { + session_store_ip4_l3l4_info (b0, sess, node_is_out); + } + + l2sess_get_session_tables (sm, sw_if_index0, node_is_out, + node_is_ip6, l4_proto, + session_tables); + l2sess_get_session_nexts (sm, sw_if_index0, node_is_out, + node_is_ip6, l4_proto, + session_nexts); + l2sess_flip_l3l4_fields (b0, node_is_ip6, l4_proto); + if (session_tables[1] != ~0) + { + l2sess_add_session (b0, node_is_out, node_is_ip6, + session_tables[1], session_nexts[1], + sess_index); + } + l2sess_flip_l3l4_fields (b0, node_is_ip6, l4_proto); + if (session_tables[0] != ~0) + { + l2sess_add_session (b0, node_is_out, node_is_ip6, + session_tables[0], session_nexts[0], + sess_index); + } + if (6 == sess->l4_proto) + { + tcp_session_account_buffer (b0, sess, node_is_out, now); + } + else + { + udp_session_account_buffer (b0, sess, node_is_out, now); + } + timing_wheel_insert (&sm->timing_wheel, + now + session_get_timeout (sm, sess, + now), + sess_index); + } + + if (now >= sm->timer_wheel_next_expiring_time) + { + check_idle_sessions (sm, sw_if_index0, now); + } + + if (node_is_out) + { + if (feature_bitmap0) + { + trace_flags0 |= 0x10; + } + if (sw_if_index0 == cached_sw_if_index) + { + trace_flags0 |= 0x20; + } + l2_output_dispatch (sm->vlib_main, + sm->vnet_main, + node, + node_index, + &cached_sw_if_index, + &cached_next_index, + next_nodes, + b0, sw_if_index0, feature_bitmap0, + &next0); + trace_flags0 |= 2; + + } + else + { + next0 = + feat_bitmap_get_next_node_index (input_feat_next_node_index, + feature_bitmap0); + trace_flags0 |= 4; + + } + + + + if (next0 >= node->n_next_nodes) + { + trace_flags0 |= 1; + } + + if (PREDICT_FALSE ((node->flags & VLIB_NODE_FLAG_TRACE) + && (b0->flags & VLIB_BUFFER_IS_TRACED))) + { + l2sess_trace_t *t = + vlib_add_trace (vm, node, b0, sizeof (*t)); + t->sw_if_index = sw_if_index0; + t->next_index = next0; + t->trace_flags = trace_flags0; + t->l4_proto = l4_proto; + t->session_tables[0] = session_tables[0]; + t->session_tables[1] = session_tables[1]; + t->session_nexts[0] = session_nexts[0]; + t->session_nexts[1] = session_nexts[1]; + } + + } + pkts_swapped += 1; + if (next0 >= node->n_next_nodes) + { + next0 = 0; + } + + /* verify speculative enqueue, maybe switch current next frame */ + vlib_validate_buffer_enqueue_x1 (vm, node, next_index, + to_next, n_left_to_next, + bi0, next0); + } + + vlib_put_next_frame (vm, node, next_index, n_left_to_next); + } + vlib_node_increment_counter (vm, node->node_index, + L2SESS_ERROR_SWAPPED, pkts_swapped); + return frame->n_vectors; +} + + +#define _(node_name, node_var, is_out, is_ip6, is_track) \ +static uword \ +node_var ## node_fn (vlib_main_t * vm, \ + vlib_node_runtime_t * node, \ + vlib_frame_t * frame) \ +{ \ + return l2sess_node_fn(vm, node, frame); \ +} \ +VLIB_REGISTER_NODE (node_var) = { \ + .function = node_var ## node_fn, \ + .name = node_name, \ + .vector_size = sizeof (u32), \ + .format_trace = format_ ## node_var ## _trace, \ + .type = VLIB_NODE_TYPE_INTERNAL, \ + \ + .n_errors = ARRAY_LEN(l2sess_error_strings), \ + .error_strings = l2sess_error_strings, \ + \ + .n_next_nodes = L2SESS_N_NEXT, \ + .next_nodes = { \ + [L2SESS_NEXT_DROP] = "error-drop", \ + }, \ +}; +foreach_l2sess_node +#undef _ diff --git a/src/plugins/acl/node_in.c b/src/plugins/acl/node_in.c new file mode 100644 index 00000000..2a5199a9 --- /dev/null +++ b/src/plugins/acl/node_in.c @@ -0,0 +1,168 @@ +/* + * Copyright (c) 2016 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ +#include +#include +#include +#include +#include +#include "node_in.h" + +typedef struct +{ + u32 next_index; + u32 sw_if_index; + u32 match_acl_index; + u32 match_rule_index; + u32 trace_bitmap; +} acl_in_trace_t; + +/* packet trace format function */ +static u8 * +format_acl_in_trace (u8 * s, va_list * args) +{ + CLIB_UNUSED (vlib_main_t * vm) = va_arg (*args, vlib_main_t *); + CLIB_UNUSED (vlib_node_t * node) = va_arg (*args, vlib_node_t *); + acl_in_trace_t *t = va_arg (*args, acl_in_trace_t *); + + s = + format (s, + "ACL_IN: sw_if_index %d, next index %d, match: inacl %d rule %d trace_bits %08x", + t->sw_if_index, t->next_index, t->match_acl_index, + t->match_rule_index, t->trace_bitmap); + return s; +} + +vlib_node_registration_t acl_in_node; + +#define foreach_acl_in_error \ +_(ACL_CHECK, "InACL check packets processed") + +typedef enum +{ +#define _(sym,str) ACL_IN_ERROR_##sym, + foreach_acl_in_error +#undef _ + ACL_IN_N_ERROR, +} acl_in_error_t; + +static char *acl_in_error_strings[] = { +#define _(sym,string) string, + foreach_acl_in_error +#undef _ +}; + +static uword +acl_in_node_fn (vlib_main_t * vm, + vlib_node_runtime_t * node, vlib_frame_t * frame) +{ + u32 n_left_from, *from, *to_next; + acl_in_next_t next_index; + u32 pkts_acl_checked = 0; + u32 feature_bitmap0; + u32 trace_bitmap = 0; + u32 *input_feat_next_node_index = + acl_main.acl_in_node_input_next_node_index; + + from = vlib_frame_vector_args (frame); + n_left_from = frame->n_vectors; + next_index = node->cached_next_index; + + while (n_left_from > 0) + { + u32 n_left_to_next; + + vlib_get_next_frame (vm, node, next_index, to_next, n_left_to_next); + + while (n_left_from > 0 && n_left_to_next > 0) + { + u32 bi0; + vlib_buffer_t *b0; + u32 next0 = ~0; + u32 sw_if_index0; + u32 next = ~0; + u32 match_acl_index = ~0; + u32 match_rule_index = ~0; + + /* speculatively enqueue b0 to the current next frame */ + bi0 = from[0]; + to_next[0] = bi0; + from += 1; + to_next += 1; + n_left_from -= 1; + n_left_to_next -= 1; + + b0 = vlib_get_buffer (vm, bi0); + + + sw_if_index0 = vnet_buffer (b0)->sw_if_index[VLIB_RX]; + feature_bitmap0 = vnet_buffer (b0)->l2.feature_bitmap; + + input_acl_packet_match (sw_if_index0, b0, &next, &match_acl_index, + &match_rule_index, &trace_bitmap); + if (next != ~0) + { + next0 = next; + } + if (next0 == ~0) + { + next0 = + feat_bitmap_get_next_node_index (input_feat_next_node_index, + feature_bitmap0); + } + + if (PREDICT_FALSE ((node->flags & VLIB_NODE_FLAG_TRACE) + && (b0->flags & VLIB_BUFFER_IS_TRACED))) + { + acl_in_trace_t *t = vlib_add_trace (vm, node, b0, sizeof (*t)); + t->sw_if_index = sw_if_index0; + t->next_index = next0; + t->match_acl_index = match_acl_index; + t->match_rule_index = match_rule_index; + t->trace_bitmap = trace_bitmap; + } + + next0 = next0 < node->n_next_nodes ? next0 : 0; + + pkts_acl_checked += 1; + + /* verify speculative enqueue, maybe switch current next frame */ + vlib_validate_buffer_enqueue_x1 (vm, node, next_index, + to_next, n_left_to_next, + bi0, next0); + } + + vlib_put_next_frame (vm, node, next_index, n_left_to_next); + } + + vlib_node_increment_counter (vm, acl_in_node.index, + ACL_IN_ERROR_ACL_CHECK, pkts_acl_checked); + return frame->n_vectors; +} + +VLIB_REGISTER_NODE (acl_in_node) = +{ + .function = acl_in_node_fn,.name = "acl-plugin-in",.vector_size = + sizeof (u32),.format_trace = format_acl_in_trace,.type = + VLIB_NODE_TYPE_INTERNAL,.n_errors = + ARRAY_LEN (acl_in_error_strings),.error_strings = + acl_in_error_strings,.n_next_nodes = ACL_IN_N_NEXT, + /* edit / add dispositions here */ + .next_nodes = + { + [ACL_IN_ERROR_DROP] = "error-drop", + [ACL_IN_ETHERNET_INPUT] = "ethernet-input", + [ACL_IN_L2S_INPUT_IP4_ADD] = "aclp-l2s-input-ip4-add", + [ACL_IN_L2S_INPUT_IP6_ADD] = "aclp-l2s-input-ip6-add",} +,}; diff --git a/src/plugins/acl/node_in.h b/src/plugins/acl/node_in.h new file mode 100644 index 00000000..502bbf8d --- /dev/null +++ b/src/plugins/acl/node_in.h @@ -0,0 +1,12 @@ +#ifndef _NODE_IN_H_ +#define _NODE_IN_H_ + +typedef enum { + ACL_IN_ERROR_DROP, + ACL_IN_ETHERNET_INPUT, + ACL_IN_L2S_INPUT_IP4_ADD, + ACL_IN_L2S_INPUT_IP6_ADD, + ACL_IN_N_NEXT, +} acl_in_next_t; + +#endif diff --git a/src/plugins/acl/node_out.c b/src/plugins/acl/node_out.c new file mode 100644 index 00000000..50af3679 --- /dev/null +++ b/src/plugins/acl/node_out.c @@ -0,0 +1,175 @@ +/* + * Copyright (c) 2016 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ +#include +#include +#include +#include +#include + +#include "node_out.h" + +typedef struct +{ + u32 next_index; + u32 sw_if_index; + u32 match_acl_index; + u32 match_rule_index; + u32 trace_bitmap; +} acl_out_trace_t; + +/* packet trace format function */ +static u8 * +format_acl_out_trace (u8 * s, va_list * args) +{ + CLIB_UNUSED (vlib_main_t * vm) = va_arg (*args, vlib_main_t *); + CLIB_UNUSED (vlib_node_t * node) = va_arg (*args, vlib_node_t *); + acl_out_trace_t *t = va_arg (*args, acl_out_trace_t *); + s = + format (s, + "ACL_OUT: sw_if_index %d, next index %d, match: outacl %d rule %d trace_bits %08x", + t->sw_if_index, t->next_index, t->match_acl_index, + t->match_rule_index, t->trace_bitmap); + return s; +} + +vlib_node_registration_t acl_out_node; + +#define foreach_acl_out_error \ +_(ACL_CHECK, "OutACL check packets processed") + +typedef enum +{ +#define _(sym,str) ACL_OUT_ERROR_##sym, + foreach_acl_out_error +#undef _ + ACL_OUT_N_ERROR, +} acl_out_error_t; + +static char *acl_out_error_strings[] = { +#define _(sym,string) string, + foreach_acl_out_error +#undef _ +}; + +static uword +acl_out_node_fn (vlib_main_t * vm, + vlib_node_runtime_t * node, vlib_frame_t * frame) +{ + acl_main_t *am = &acl_main; + l2_output_next_nodes_st *next_nodes = &am->acl_out_output_next_nodes; + u32 n_left_from, *from, *to_next; + acl_out_next_t next_index; + u32 pkts_acl_checked = 0; + u32 feature_bitmap0; + u32 cached_sw_if_index = (u32) ~ 0; + u32 cached_next_index = (u32) ~ 0; + u32 match_acl_index = ~0; + u32 match_rule_index = ~0; + u32 trace_bitmap = 0; + + from = vlib_frame_vector_args (frame); + n_left_from = frame->n_vectors; + next_index = node->cached_next_index; + + while (n_left_from > 0) + { + u32 n_left_to_next; + + vlib_get_next_frame (vm, node, next_index, to_next, n_left_to_next); + + while (n_left_from > 0 && n_left_to_next > 0) + { + u32 bi0; + vlib_buffer_t *b0; + u32 next0 = ~0; + u32 next = 0; + u32 sw_if_index0; + + /* speculatively enqueue b0 to the current next frame */ + bi0 = from[0]; + to_next[0] = bi0; + from += 1; + to_next += 1; + n_left_from -= 1; + n_left_to_next -= 1; + + b0 = vlib_get_buffer (vm, bi0); + + + sw_if_index0 = vnet_buffer (b0)->sw_if_index[VLIB_TX]; + feature_bitmap0 = vnet_buffer (b0)->l2.feature_bitmap; + + output_acl_packet_match (sw_if_index0, b0, &next, &match_acl_index, + &match_rule_index, &trace_bitmap); + if (next != ~0) + { + next0 = next; + } + if (next0 == ~0) + { + l2_output_dispatch (vm, + am->vnet_main, + node, + acl_out_node.index, + &cached_sw_if_index, + &cached_next_index, + next_nodes, + b0, sw_if_index0, feature_bitmap0, &next0); + } + + + + if (PREDICT_FALSE ((node->flags & VLIB_NODE_FLAG_TRACE) + && (b0->flags & VLIB_BUFFER_IS_TRACED))) + { + acl_out_trace_t *t = vlib_add_trace (vm, node, b0, sizeof (*t)); + t->sw_if_index = sw_if_index0; + t->next_index = next0; + t->match_acl_index = match_acl_index; + t->match_rule_index = match_rule_index; + t->trace_bitmap = trace_bitmap; + } + + pkts_acl_checked += 1; + + /* verify speculative enqueue, maybe switch current next frame */ + vlib_validate_buffer_enqueue_x1 (vm, node, next_index, + to_next, n_left_to_next, + bi0, next0); + } + + vlib_put_next_frame (vm, node, next_index, n_left_to_next); + } + + vlib_node_increment_counter (vm, acl_out_node.index, + ACL_OUT_ERROR_ACL_CHECK, pkts_acl_checked); + return frame->n_vectors; +} + +VLIB_REGISTER_NODE (acl_out_node) = +{ + .function = acl_out_node_fn,.name = "acl-plugin-out",.vector_size = + sizeof (u32),.format_trace = format_acl_out_trace,.type = + VLIB_NODE_TYPE_INTERNAL,.n_errors = + ARRAY_LEN (acl_out_error_strings),.error_strings = + acl_out_error_strings,.n_next_nodes = ACL_OUT_N_NEXT, + /* edit / add dispositions here */ + .next_nodes = + { + [ACL_OUT_ERROR_DROP] = "error-drop", + [ACL_OUT_INTERFACE_OUTPUT] = "interface-output", + [ACL_OUT_L2S_OUTPUT_IP4_ADD] = "aclp-l2s-output-ip4-add", + [ACL_OUT_L2S_OUTPUT_IP6_ADD] = "aclp-l2s-output-ip6-add",} +,}; diff --git a/src/plugins/acl/node_out.h b/src/plugins/acl/node_out.h new file mode 100644 index 00000000..c919f3b7 --- /dev/null +++ b/src/plugins/acl/node_out.h @@ -0,0 +1,12 @@ +#ifndef _NODE_OUT_H_ +#define _NODE_OUT_H_ + +typedef enum { + ACL_OUT_ERROR_DROP, + ACL_OUT_INTERFACE_OUTPUT, + ACL_OUT_L2S_OUTPUT_IP4_ADD, + ACL_OUT_L2S_OUTPUT_IP6_ADD, + ACL_OUT_N_NEXT, +} acl_out_next_t; + +#endif diff --git a/src/plugins/acl/test/run-python b/src/plugins/acl/test/run-python new file mode 100755 index 00000000..215eb17a --- /dev/null +++ b/src/plugins/acl/test/run-python @@ -0,0 +1,28 @@ +#!/bin/sh +# +# Do all the legwork to run a scapy shell with APIs available for load +# +CURR_DIR=`pwd` +ROOT_DIR=`git rev-parse --show-toplevel` +cd $ROOT_DIR +sudo apt-get install -y python-virtualenv +# uncomment the line below to enable build of plugins and api each time +# make plugins && make build-vpp-api || exit +virtualenv virtualenv +virtualenv/bin/pip install ipaddress +virtualenv/bin/pip install scapy +# install the python API into the virtualenv +cd $ROOT_DIR/vpp-api/python/ +$ROOT_DIR/virtualenv/bin/python setup.py install +# install the python ACL plugin API into the virtualenv +ACL_PLUGIN_SETUP_DIR=`find $ROOT_DIR/build-root -name acl-plugin` +cd $ACL_PLUGIN_SETUP_DIR; +$ROOT_DIR/virtualenv/bin/python setup.py install +cd $ROOT_DIR +# figure out the shared library path and start scapy +export LD_LIBRARY_PATH=`pwd`/`find . -name "libpneum.so" -exec dirname {} \; | grep lib64 | head -n 1` +cd $CURR_DIR +sudo LD_LIBRARY_PATH=$LD_LIBRARY_PATH $ROOT_DIR/virtualenv/bin/python $1 $2 $3 $4 $5 $6 $7 $8 $9 + + + diff --git a/src/plugins/acl/test/run-scapy b/src/plugins/acl/test/run-scapy new file mode 100755 index 00000000..266f07d1 --- /dev/null +++ b/src/plugins/acl/test/run-scapy @@ -0,0 +1,26 @@ +#!/bin/sh +# +# Do all the legwork to run a scapy shell with APIs available for load +# +ROOT_DIR=`git rev-parse --show-toplevel` +cd $ROOT_DIR +sudo apt-get install -y python-virtualenv +# uncomment the line below to enable the build of plugins and API each time.. +# make plugins && make build-vpp-api || exit +virtualenv virtualenv +virtualenv/bin/pip install ipaddress +virtualenv/bin/pip install scapy +# install the python API into the virtualenv +cd $ROOT_DIR/vpp-api/python/ +$ROOT_DIR/virtualenv/bin/python setup.py install +# install the python ACL plugin API into the virtualenv +ACL_PLUGIN_SETUP_DIR=`find $ROOT_DIR/build-root -name acl-plugin` +cd $ACL_PLUGIN_SETUP_DIR; +$ROOT_DIR/virtualenv/bin/python setup.py install +cd $ROOT_DIR +# figure out the shared library path and start scapy +export LD_LIBRARY_PATH=`pwd`/`find . -name "libpneum.so" -exec dirname {} \; | grep lib64 | head -n 1` +sudo LD_LIBRARY_PATH=$LD_LIBRARY_PATH virtualenv/bin/scapy + + + diff --git a/src/plugins/acl/test/test_acl_plugin.py b/src/plugins/acl/test/test_acl_plugin.py new file mode 100644 index 00000000..7fc72d67 --- /dev/null +++ b/src/plugins/acl/test/test_acl_plugin.py @@ -0,0 +1,118 @@ +from __future__ import print_function +import unittest, sys, time, threading, struct, logging, os +import vpp_papi +# import vpp_papi_plugins.acl +from ipaddress import * +papi_event = threading.Event() +print(vpp_papi.vpe.VL_API_SW_INTERFACE_SET_FLAGS) +def papi_event_handler(result): + if result.vl_msg_id == vpp_papi.vpe.VL_API_SW_INTERFACE_SET_FLAGS: + return + if result.vl_msg_id == vpp_papi.vpe.VL_API_VNET_INTERFACE_COUNTERS: + print('Interface counters', result) + return + if result.vl_msg_id == vpp_papi.vpe.VL_API_VNET_IP6_FIB_COUNTERS: + print('IPv6 FIB counters', result) + papi_event.set() + return + + print('Unknown message id:', result.vl_msg_id) + +import glob, subprocess +class TestAclPlugin(unittest.TestCase): + @classmethod + def setUpClass(cls): + print("Setup") + @classmethod + def tearDownClass(cls): + print("Teardown") + + def setUp(self): + print("Connecting API") + r = vpp_papi.connect("test_papi") + self.assertEqual(r, 0) + + def tearDown(self): + r = vpp_papi.disconnect() + self.assertEqual(r, 0) + + # + # The tests themselves + # + + # + # Basic request / reply + # + def test_show_version(self): + t = vpp_papi.show_version() + print('T', t); + program = t.program.decode().rstrip('\x00') + self.assertEqual('vpe', program) + + def x_test_acl_add(self): + print("Test ACL add") + self.assertEqual(1, 1) + + # + # Details / Dump + # + def x_test_details_dump(self): + t = vpp_papi.sw_interface_dump(0, b'') + print('Dump/details T', t) + + # + # Arrays + # + def x_test_arrays(self): + t = vpp_papi.vnet_get_summary_stats() + print('Summary stats', t) + print('Packets:', t.total_pkts[0]) + print('Packets:', t.total_pkts[1]) + # + # Variable sized arrays and counters + # + #@unittest.skip("stats") + def x_test_want_stats(self): + pid = 123 + vpp_papi.register_event_callback(papi_event_handler) + papi_event.clear() + + # Need to configure IPv6 to get som IPv6 FIB stats + t = vpp_papi.create_loopback('') + print(t) + self.assertEqual(t.retval, 0) + + ifindex = t.sw_if_index + addr = str(IPv6Address(u'1::1').packed) + t = vpp_papi.sw_interface_add_del_address(ifindex, 1, 1, 0, 16, addr) + print(t) + self.assertEqual(t.retval, 0) + + # Check if interface is up + # XXX: Add new API to query interface state based on ifindex, instead of dump all. + t = vpp_papi.sw_interface_set_flags(ifindex, 1, 1, 0) + self.assertEqual(t.retval, 0) + + t = vpp_papi.want_stats(True, pid) + + print (t) + + # + # Wait for some stats + # + self.assertEqual(papi_event.wait(15), True) + t = vpp_papi.want_stats(False, pid) + print (t) + + + # + # Plugins? + # + +if __name__ == '__main__' or __name__ == '__builtin__': + print("This is main") + suite = unittest.TestLoader().loadTestsFromTestCase(TestAclPlugin) + unittest.TextTestRunner(verbosity=2).run(suite) + #logging.basicConfig(level=logging.DEBUG) + # unittest.main() + diff --git a/src/plugins/ioam.am b/src/plugins/ioam.am new file mode 100644 index 00000000..a4984b18 --- /dev/null +++ b/src/plugins/ioam.am @@ -0,0 +1,150 @@ +# Copyright (c) 2015 Cisco and/or its affiliates. +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at: +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + + +######################################## +# iOAM Proof of Transit +######################################## + +ioam_pot_plugin_la_SOURCES = \ + ioam/lib-pot/pot_util.c \ + ioam/encap/ip6_ioam_pot.c \ + ioam/lib-pot/pot_util.h \ + ioam/lib-pot/math64.h \ + ioam/lib-pot/pot_api.c + +noinst_HEADERS += \ + ioam/lib-pot/pot_all_api_h.h \ + ioam/lib-pot/pot_msg_enum.h \ + ioam/lib-pot/pot.api.h \ + ioam/lib-pot/pot_util.h \ + ioam/lib-pot/math64.h + +API_FILES += ioam/lib-pot/pot.api + +ioam_pot_test_plugin_la_SOURCES = \ + ioam/lib-pot/pot_test.c \ + ioam/lib-pot/pot_plugin.api.h + +vppapitestplugins_LTLIBRARIES += ioam_pot_test_plugin.la +vppplugins_LTLIBRARIES += ioam_pot_plugin.la + +######################################## +# iOAM trace export for IPv6 +######################################## + +ioam_export_plugin_la_SOURCES = \ +ioam/export/ioam_export.c \ +ioam/export/node.c \ +ioam/export/ioam_export.api.h \ +ioam/export/ioam_export_thread.c + +noinst_HEADERS += \ + ioam/export/ioam_export_all_api_h.h \ + ioam/export/ioam_export_msg_enum.h \ + ioam/export/ioam_export.api.h + +API_FILES += ioam/export/ioam_export.api + +ioam_export_test_plugin_la_SOURCES = \ + ioam/export/ioam_export_test.c \ + ioam/export/ioam_export_plugin.api.h + +vppapitestplugins_LTLIBRARIES += ioam_export_test_plugin.la +vppplugins_LTLIBRARIES += ioam_export_plugin.la + +######################################## +# iOAM Trace +######################################## +libioam_trace_plugin_la_SOURCES = \ + ioam/lib-trace/trace_util.c \ + ioam/encap/ip6_ioam_trace.c \ + ioam/lib-trace/trace_util.h \ + ioam/lib-trace/trace_api.c + +noinst_HEADERS += \ + ioam/export/ioam_export_all_api_h.h \ + ioam/lib-trace/trace_all_api_h.h \ + ioam/lib-trace/trace_msg_enum.h \ + ioam/lib-trace/trace.api.h \ + ioam/lib-trace/trace_util.h + +API_FILES += ioam/lib-trace/trace.api + +ioam_trace_test_plugin_la_SOURCES = \ + ioam/lib-trace/trace_test.c \ + ioam/lib-trace/trace_plugin.api.h + +vppapitestplugins_LTLIBRARIES += ioam_trace_test_plugin.la +vppplugins_LTLIBRARIES += libioam_trace_plugin.la + +######################################## +# VxLAN-GPE +######################################## +libioam_vxlan_gpe_plugin_la_SOURCES = \ + ioam/lib-vxlan-gpe/ioam_encap.c \ + ioam/lib-vxlan-gpe/ioam_decap.c \ + ioam/lib-vxlan-gpe/ioam_transit.c \ + ioam/lib-vxlan-gpe/ioam_pop.c \ + ioam/lib-vxlan-gpe/vxlan_gpe_api.c \ + ioam/lib-vxlan-gpe/vxlan_gpe_ioam_trace.c \ + ioam/lib-vxlan-gpe/vxlan_gpe_ioam.c \ + ioam/export-vxlan-gpe/vxlan_gpe_ioam_export.c \ + ioam/export-vxlan-gpe/vxlan_gpe_node.c \ + ioam/export-vxlan-gpe/vxlan_gpe_ioam_export.api.h\ + ioam/export-vxlan-gpe/vxlan_gpe_ioam_export_thread.c + +noinst_HEADERS += \ + ioam/export/ioam_export_all_api_h.h \ + ioam/lib-vxlan-gpe/vxlan_gpe_all_api_h.h \ + ioam/lib-vxlan-gpe/vxlan_gpe_msg_enum.h \ + ioam/lib-vxlan-gpe/ioam_vxlan_gpe.api.h \ + ioam/lib-vxlan-gpe/vxlan_gpe_ioam_util.h \ + ioam/lib-vxlan-gpe/vxlan_gpe_ioam_packet.h \ + ioam/lib-vxlan-gpe/vxlan_gpe_ioam.h \ + ioam/export-vxlan-gpe/vxlan_gpe_ioam_export_all_api_h.h \ + ioam/export-vxlan-gpe/vxlan_gpe_ioam_export_msg_enum.h \ + ioam/export-vxlan-gpe/vxlan_gpe_ioam_export.api.h + +API_FILES += ioam/lib-vxlan-gpe/ioam_vxlan_gpe.api +API_FILES += ioam/export-vxlan-gpe/vxlan_gpe_ioam_export.api + +ioam_vxlan_gpe_test_plugin_la_SOURCES = \ + ioam/lib-vxlan-gpe/vxlan_gpe_test.c \ + ioam/lib-vxlan-gpe/vxlan_gpe_plugin.api.h + +vppapitestplugins_LTLIBRARIES += ioam_vxlan_gpe_test_plugin.la +vppplugins_LTLIBRARIES += libioam_vxlan_gpe_plugin.la + +vxlan_gpe_ioam_export_test_plugin_la_SOURCES = \ + ioam/export-vxlan-gpe/vxlan_gpe_ioam_export_test.c \ + ioam/export-vxlan-gpe/vxlan_gpe_ioam_export_plugin.api.h + +vppapitestplugins_LTLIBRARIES += vxlan_gpe_ioam_export_test_plugin.la + +######################################## +# iOAM E2E plugin +######################################## + +ioam_e2e_plugin_la_SOURCES = \ + ioam/encap/ip6_ioam_e2e.c \ + ioam/encap/ip6_ioam_seqno.c \ + ioam/encap/ip6_ioam_seqno_analyse.c + +noinst_HEADERS += \ + ioam/encap/ip6_ioam_e2e.h \ + ioam/encap/ip6_ioam_seqno.h + +vppplugins_LTLIBRARIES += ioam_e2e_plugin.la + +# vi:syntax=automake diff --git a/src/plugins/ioam/dir.dox b/src/plugins/ioam/dir.dox new file mode 100644 index 00000000..f3389b52 --- /dev/null +++ b/src/plugins/ioam/dir.dox @@ -0,0 +1,18 @@ +/* + * Copyright (c) 2016 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ +/** + @dir + @brief Inband OAM (iOAM) implementation +*/ diff --git a/src/plugins/ioam/encap/ip6_ioam_e2e.c b/src/plugins/ioam/encap/ip6_ioam_e2e.c new file mode 100644 index 00000000..0839cdce --- /dev/null +++ b/src/plugins/ioam/encap/ip6_ioam_e2e.c @@ -0,0 +1,232 @@ +/* + * Copyright (c) 2016 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +#include +#include +#include +#include + +#include + +#include +#include +#include + +#include +#include + +#include "ip6_ioam_e2e.h" + +ioam_e2e_main_t ioam_e2e_main; + +static u8 * ioam_e2e_trace_handler (u8 * s, + ip6_hop_by_hop_option_t *opt) +{ + ioam_e2e_option_t * e2e = (ioam_e2e_option_t *)opt; + u32 seqno = 0; + + if (e2e) + { + seqno = clib_net_to_host_u32 (e2e->e2e_data); + } + + s = format (s, "SeqNo = 0x%Lx", seqno); + return s; +} + +int +ioam_e2e_config_handler (void *data, u8 disable) +{ + int *analyse = data; + + /* Register hanlders if enabled */ + if (!disable) + { + /* If encap node register for encap handler */ + if (0 == *analyse) + { + if (ip6_hbh_register_option(HBH_OPTION_TYPE_IOAM_EDGE_TO_EDGE, + ioam_seqno_encap_handler, + ioam_e2e_trace_handler) < 0) + { + return (-1); + } + } + /* If analyze node then register for decap handler */ + else + { + if (ip6_hbh_pop_register_option(HBH_OPTION_TYPE_IOAM_EDGE_TO_EDGE, + ioam_seqno_decap_handler) < 0) + { + return (-1); + } + } + return 0; + } + + /* UnRegister handlers */ + (void) ip6_hbh_unregister_option(HBH_OPTION_TYPE_IOAM_EDGE_TO_EDGE); + (void) ip6_hbh_pop_unregister_option(HBH_OPTION_TYPE_IOAM_EDGE_TO_EDGE); + return 0; +} + +int +ioam_e2e_rewrite_handler (u8 *rewrite_string, + u8 *rewrite_size) +{ + ioam_e2e_option_t *e2e_option; + + if (rewrite_string && *rewrite_size == sizeof(ioam_e2e_option_t)) + { + e2e_option = (ioam_e2e_option_t *)rewrite_string; + e2e_option->hdr.type = HBH_OPTION_TYPE_IOAM_EDGE_TO_EDGE + | HBH_OPTION_TYPE_SKIP_UNKNOWN; + e2e_option->hdr.length = sizeof (ioam_e2e_option_t) - + sizeof (ip6_hop_by_hop_option_t); + return(0); + } + return(-1); +} + +u32 +ioam_e2e_flow_handler (u32 ctx, u8 add) +{ + ioam_e2e_data_t *data; + u16 i; + + if (add) + { + pool_get(ioam_e2e_main.e2e_data, data); + data->flow_ctx = ctx; + ioam_seqno_init_bitmap(&data->seqno_data); + return ((u32) (data - ioam_e2e_main.e2e_data)); + } + + /* Delete case */ + for (i = 0; i < vec_len(ioam_e2e_main.e2e_data); i++) + { + if (pool_is_free_index(ioam_e2e_main.e2e_data, i)) + continue; + + data = pool_elt_at_index(ioam_e2e_main.e2e_data, i); + if (data && (data->flow_ctx == ctx)) + { + pool_put_index(ioam_e2e_main.e2e_data, i); + return (0); + } + } + return 0; +} + +static clib_error_t * +ioam_show_e2e_cmd_fn (vlib_main_t * vm, + unformat_input_t * input, + vlib_cli_command_t * cmd) +{ + ioam_e2e_data_t *e2e_data; + u8 *s = 0; + int i; + + vec_reset_length(s); + + s = format(0, "IOAM E2E information: \n"); + for (i = 0; i < vec_len(ioam_e2e_main.e2e_data); i++) + { + if (pool_is_free_index(ioam_e2e_main.e2e_data, i)) + continue; + + e2e_data = pool_elt_at_index(ioam_e2e_main.e2e_data, i); + s = format(s, "Flow name: %s\n", get_flow_name_from_flow_ctx(e2e_data->flow_ctx)); + + s = show_ioam_seqno_cmd_fn(s, + &e2e_data->seqno_data, + !IOAM_DEAP_ENABLED(e2e_data->flow_ctx)); + } + + vlib_cli_output(vm, "%v", s); + return 0; +} + + +VLIB_CLI_COMMAND (ioam_show_e2e_cmd, static) = { + .path = "show ioam e2e ", + .short_help = "show ioam e2e information", + .function = ioam_show_e2e_cmd_fn, +}; + +/* + * This routine exists to convince the vlib plugin framework that + * we haven't accidentally copied a random .dll into the plugin directory. + * + * Also collects global variable pointers passed from the vpp engine + */ +clib_error_t * +vlib_plugin_register (vlib_main_t * vm, vnet_plugin_handoff_t * h, + int from_early_init) +{ + clib_error_t * error = 0; + + ioam_e2e_main.vlib_main = vm; + ioam_e2e_main.vnet_main = h->vnet_main; + return error; +} + +/* + * Init handler E2E headet handling. + * Init hanlder registers encap, decap, trace and Rewrite handlers. + */ +static clib_error_t * +ioam_e2e_init (vlib_main_t * vm) +{ + clib_error_t * error; + + if ((error = vlib_call_init_function (vm, ip6_hop_by_hop_ioam_init))) + { + return(error); + } + + /* + * As of now we have only PPC under E2E header. + */ + if (ip6_hbh_config_handler_register(HBH_OPTION_TYPE_IOAM_EDGE_TO_EDGE, + ioam_e2e_config_handler) < 0) + { + return (clib_error_create("Registration of " + "HBH_OPTION_TYPE_IOAM_EDGE_TO_EDGE for rewrite failed")); + } + + if (ip6_hbh_add_register_option(HBH_OPTION_TYPE_IOAM_EDGE_TO_EDGE, + sizeof(ioam_e2e_option_t), + ioam_e2e_rewrite_handler) < 0) + { + return (clib_error_create("Registration of " + "HBH_OPTION_TYPE_IOAM_EDGE_TO_EDGE for rewrite failed")); + } + + if (ip6_hbh_flow_handler_register(HBH_OPTION_TYPE_IOAM_EDGE_TO_EDGE, + ioam_e2e_flow_handler) < 0) + { + return (clib_error_create("Registration of " + "HBH_OPTION_TYPE_IOAM_EDGE_TO_EDGE Flow handler failed")); + } + + return (0); +} + +/* + * Init function for the E2E lib. + * ip6_hop_by_hop_ioam_e2e_init gets called during init. + */ +VLIB_INIT_FUNCTION (ioam_e2e_init); diff --git a/src/plugins/ioam/encap/ip6_ioam_e2e.h b/src/plugins/ioam/encap/ip6_ioam_e2e.h new file mode 100644 index 00000000..18f35f80 --- /dev/null +++ b/src/plugins/ioam/encap/ip6_ioam_e2e.h @@ -0,0 +1,47 @@ +/* + * Copyright (c) 2016 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +#ifndef __included_ip6_ioam_e2e_h__ +#define __included_ip6_ioam_e2e_h__ + +#include "ip6_ioam_seqno.h" + +typedef struct ioam_e2e_data_t_ { + u32 flow_ctx; + u32 pad; + ioam_seqno_data seqno_data; +} ioam_e2e_data_t; + +typedef struct { + ioam_e2e_data_t *e2e_data; + vlib_main_t *vlib_main; + vnet_main_t *vnet_main; +} ioam_e2e_main_t; + +extern ioam_e2e_main_t ioam_e2e_main; + +static inline ioam_seqno_data * +ioam_e2ec_get_seqno_data_from_flow_ctx (u32 flow_ctx) +{ + ioam_e2e_data_t *data = NULL; + u32 index; + + index = get_flow_data_from_flow_ctx(flow_ctx, + HBH_OPTION_TYPE_IOAM_EDGE_TO_EDGE); + data = &ioam_e2e_main.e2e_data[index]; + return &(data->seqno_data); +} + +#endif /* __included_ioam_e2e_h__ */ diff --git a/src/plugins/ioam/encap/ip6_ioam_pot.c b/src/plugins/ioam/encap/ip6_ioam_pot.c new file mode 100644 index 00000000..05f42c91 --- /dev/null +++ b/src/plugins/ioam/encap/ip6_ioam_pot.c @@ -0,0 +1,276 @@ +/* + * Copyright (c) 2016 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ +#include +#include +#include +#include + +#include +#include +#include + +#include +#include +#include + +#include + +typedef CLIB_PACKED(struct { + ip6_hop_by_hop_option_t hdr; + u8 pot_type; +#define PROFILE_ID_MASK 0xF + u8 reserved_profile_id; /* 4 bits reserved, 4 bits to carry profile id */ + u64 random; + u64 cumulative; +}) ioam_pot_option_t; + +#define foreach_ip6_hop_by_hop_ioam_pot_stats \ + _(PROCESSED, "Pkts with ip6 hop-by-hop pot options") \ + _(PROFILE_MISS, "Pkts with ip6 hop-by-hop pot options but no profile set") \ + _(PASSED, "Pkts with POT in Policy") \ + _(FAILED, "Pkts with POT out of Policy") + +static char * ip6_hop_by_hop_ioam_pot_stats_strings[] = { +#define _(sym,string) string, + foreach_ip6_hop_by_hop_ioam_pot_stats +#undef _ +}; + +typedef enum { +#define _(sym,str) IP6_IOAM_POT_##sym, + foreach_ip6_hop_by_hop_ioam_pot_stats +#undef _ + IP6_IOAM_POT_N_STATS, +} ip6_ioam_pot_stats_t; + +typedef struct { + /* stats */ + u64 counters[ARRAY_LEN(ip6_hop_by_hop_ioam_pot_stats_strings)]; + + /* convenience */ + vlib_main_t * vlib_main; + vnet_main_t * vnet_main; +} ip6_hop_by_hop_ioam_pot_main_t; + +ip6_hop_by_hop_ioam_pot_main_t ip6_hop_by_hop_ioam_pot_main; + +always_inline void +ip6_ioam_stats_increment_counter (u32 counter_index, u64 increment) +{ + ip6_hop_by_hop_ioam_pot_main_t *hm = &ip6_hop_by_hop_ioam_pot_main; + + hm->counters[counter_index] += increment; +} + + +static u8 * format_ioam_pot (u8 * s, va_list * args) +{ + ioam_pot_option_t * pot0 = va_arg (*args, ioam_pot_option_t *); + u64 random, cumulative; + random = cumulative = 0; + if (pot0) + { + random = clib_net_to_host_u64 (pot0->random); + cumulative = clib_net_to_host_u64 (pot0->cumulative); + } + + s = format (s, "random = 0x%Lx, Cumulative = 0x%Lx, Index = 0x%x", + random, cumulative, pot0 ? pot0->reserved_profile_id : ~0); + return s; +} + +u8 * +ip6_hbh_ioam_proof_of_transit_trace_handler (u8 *s, ip6_hop_by_hop_option_t *opt) +{ + ioam_pot_option_t *pot; + + s = format (s, " POT opt present\n"); + pot = (ioam_pot_option_t *) opt; + s = format (s, " %U\n", format_ioam_pot, pot); + return (s); +} + +int +ip6_hbh_ioam_proof_of_transit_handler (vlib_buffer_t *b, + ip6_header_t *ip, + ip6_hop_by_hop_option_t *opt0) +{ + ioam_pot_option_t * pot0; + u64 random = 0, cumulative = 0; + int rv = 0; + u8 pot_profile_index; + pot_profile *pot_profile = 0, *new_profile = 0; + u8 pot_encap = 0; + + pot0 = (ioam_pot_option_t *) opt0; + pot_encap = (pot0->random == 0); + pot_profile_index = pot_profile_get_active_id(); + pot_profile = pot_profile_get_active(); + if (pot_encap && PREDICT_FALSE(!pot_profile)) + { + ip6_ioam_stats_increment_counter (IP6_IOAM_POT_PROFILE_MISS, 1); + return(-1); + } + if (pot_encap) + { + pot0->reserved_profile_id = + pot_profile_index & PROFILE_ID_MASK; + pot_profile_incr_usage_stats(pot_profile); + } + else + { /* Non encap node */ + if (PREDICT_FALSE(pot0->reserved_profile_id != + pot_profile_index || pot_profile == 0)) + { + /* New profile announced by encap node. */ + new_profile = + pot_profile_find(pot0->reserved_profile_id); + if (PREDICT_FALSE(new_profile == 0 || + new_profile->valid == 0)) + { + ip6_ioam_stats_increment_counter (IP6_IOAM_POT_PROFILE_MISS, 1); + return(-1); + } + else + { + pot_profile_index = pot0->reserved_profile_id; + pot_profile = new_profile; + pot_profile_set_active(pot_profile_index); + pot_profile_reset_usage_stats(pot_profile); + } + } + pot_profile_incr_usage_stats(pot_profile); + } + + if (pot0->random == 0) + { + pot0->random = clib_host_to_net_u64(pot_generate_random(pot_profile)); + pot0->cumulative = 0; + } + random = clib_net_to_host_u64(pot0->random); + cumulative = clib_net_to_host_u64(pot0->cumulative); + pot0->cumulative = clib_host_to_net_u64( + pot_update_cumulative(pot_profile, + cumulative, + random)); + ip6_ioam_stats_increment_counter (IP6_IOAM_POT_PROCESSED, 1); + + return (rv); +} + +int +ip6_hbh_ioam_proof_of_transit_pop_handler (vlib_buffer_t *b, ip6_header_t *ip, + ip6_hop_by_hop_option_t *opt0) +{ + ioam_pot_option_t * pot0; + u64 random = 0; + u64 cumulative = 0; + int rv = 0; + pot_profile *pot_profile = 0; + u8 result = 0; + + pot0 = (ioam_pot_option_t *) opt0; + random = clib_net_to_host_u64(pot0->random); + cumulative = clib_net_to_host_u64(pot0->cumulative); + pot_profile = pot_profile_get_active(); + result = pot_validate (pot_profile, + cumulative, random); + + if (result == 1) + { + ip6_ioam_stats_increment_counter (IP6_IOAM_POT_PASSED, 1); + } + else + { + ip6_ioam_stats_increment_counter (IP6_IOAM_POT_FAILED, 1); + } + return (rv); +} + +int ip6_hop_by_hop_ioam_pot_rewrite_handler (u8 *rewrite_string, u8 *rewrite_size) +{ + ioam_pot_option_t * pot_option; + if (rewrite_string && *rewrite_size == sizeof(ioam_pot_option_t)) + { + pot_option = (ioam_pot_option_t *)rewrite_string; + pot_option->hdr.type = HBH_OPTION_TYPE_IOAM_PROOF_OF_TRANSIT + | HBH_OPTION_TYPE_DATA_CHANGE_ENROUTE; + pot_option->hdr.length = sizeof (ioam_pot_option_t) - + sizeof (ip6_hop_by_hop_option_t); + return(0); + } + return(-1); +} + +static clib_error_t * +ip6_show_ioam_pot_cmd_fn (vlib_main_t * vm, + unformat_input_t * input, + vlib_cli_command_t * cmd) +{ + ip6_hop_by_hop_ioam_pot_main_t *hm = &ip6_hop_by_hop_ioam_pot_main; + u8 *s = 0; + int i = 0; + + for ( i = 0; i < IP6_IOAM_POT_N_STATS; i++) + { + s = format(s, " %s - %lu\n", ip6_hop_by_hop_ioam_pot_stats_strings[i], + hm->counters[i]); + } + + vlib_cli_output(vm, "%v", s); + vec_free(s); + return 0; +} + + +VLIB_CLI_COMMAND (ip6_show_ioam_pot_cmd, static) = { + .path = "show ioam pot", + .short_help = "iOAM pot statistics", + .function = ip6_show_ioam_pot_cmd_fn, +}; + + +static clib_error_t * +ip6_hop_by_hop_ioam_pot_init (vlib_main_t * vm) +{ + ip6_hop_by_hop_ioam_pot_main_t * hm = &ip6_hop_by_hop_ioam_pot_main; + clib_error_t * error; + + if ((error = vlib_call_init_function (vm, ip6_hop_by_hop_ioam_init))) + return(error); + + hm->vlib_main = vm; + hm->vnet_main = vnet_get_main(); + memset(hm->counters, 0, sizeof(hm->counters)); + + if (ip6_hbh_register_option(HBH_OPTION_TYPE_IOAM_PROOF_OF_TRANSIT, ip6_hbh_ioam_proof_of_transit_handler, + ip6_hbh_ioam_proof_of_transit_trace_handler) < 0) + return (clib_error_create("registration of HBH_OPTION_TYPE_IOAM_PROOF_OF_TRANSIT failed")); + + if (ip6_hbh_add_register_option(HBH_OPTION_TYPE_IOAM_PROOF_OF_TRANSIT, + sizeof(ioam_pot_option_t), + ip6_hop_by_hop_ioam_pot_rewrite_handler) < 0) + return (clib_error_create("registration of HBH_OPTION_TYPE_IOAM_PROOF_OF_TRANSIT for rewrite failed")); + + if (ip6_hbh_pop_register_option(HBH_OPTION_TYPE_IOAM_PROOF_OF_TRANSIT, + ip6_hbh_ioam_proof_of_transit_pop_handler) < 0) + return (clib_error_create("registration of HBH_OPTION_TYPE_IOAM_PROOF_OF_TRANSIT POP failed")); + + return (0); +} + +VLIB_INIT_FUNCTION (ip6_hop_by_hop_ioam_pot_init); + + diff --git a/src/plugins/ioam/encap/ip6_ioam_seqno.c b/src/plugins/ioam/encap/ip6_ioam_seqno.c new file mode 100644 index 00000000..0b4d4192 --- /dev/null +++ b/src/plugins/ioam/encap/ip6_ioam_seqno.c @@ -0,0 +1,109 @@ +/* + * Copyright (c) 2016 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + + +#include +#include +#include +#include + +#include + +#include +#include +#include + +#include "ip6_ioam_seqno.h" +#include "ip6_ioam_e2e.h" + +ioam_seqno_data_main_t ioam_seqno_main; + +void ioam_seqno_init_bitmap (ioam_seqno_data *data) +{ + seqno_bitmap *bitmap = &data->seqno_rx.bitmap; + bitmap->window_size = SEQNO_WINDOW_SIZE; + bitmap->array_size = SEQNO_WINDOW_ARRAY_SIZE; + bitmap->mask = 32 * SEQNO_WINDOW_ARRAY_SIZE - 1; + bitmap->array[0] = 0x00000000;/* pretend we haven seen sequence numbers 0*/ + bitmap->highest = 0; + + data->seq_num = 0; + return ; +} + +/* + * This Routine gets called from IPv6 hop-by-hop option handling. + * Only if we are encap node, then add PPC data. + * On a Transit(MID) node we dont do anything with E2E headers. + * On decap node decap is handled by seperate function. + */ +int +ioam_seqno_encap_handler (vlib_buffer_t *b, ip6_header_t *ip, + ip6_hop_by_hop_option_t *opt) +{ + u32 opaque_index = vnet_buffer(b)->l2_classify.opaque_index; + ioam_e2e_option_t * e2e; + int rv = 0; + ioam_seqno_data *data; + + data = ioam_e2ec_get_seqno_data_from_flow_ctx(opaque_index); + e2e = (ioam_e2e_option_t *) opt; + e2e->e2e_data = clib_host_to_net_u32(++data->seq_num); + + return (rv); +} + +/* + * This Routine gets called on POP/Decap node. + */ +int +ioam_seqno_decap_handler (vlib_buffer_t *b, ip6_header_t *ip, + ip6_hop_by_hop_option_t *opt) +{ + u32 opaque_index = vnet_buffer(b)->l2_classify.opaque_index; + ioam_e2e_option_t * e2e; + int rv = 0; + ioam_seqno_data *data; + + data = ioam_e2ec_get_seqno_data_from_flow_ctx(opaque_index); + e2e = (ioam_e2e_option_t *) opt; + ioam_analyze_seqno(&data->seqno_rx, (u64) clib_net_to_host_u32(e2e->e2e_data)); + + return (rv); +} + +u8 * +show_ioam_seqno_cmd_fn (u8 *s, ioam_seqno_data *seqno_data, u8 enc) +{ + seqno_rx_info *rx; + + s = format(s, "SeqNo Data:\n"); + if (enc) + { + s = format(s, " Current Seq. Number : %llu\n", seqno_data->seq_num); + } + else + { + rx = &seqno_data->seqno_rx; + s = format(s, " Highest Seq. Number : %llu\n", rx->bitmap.highest); + s = format(s, " Packets received : %llu\n", rx->rx_packets); + s = format(s, " Lost packets : %llu\n", rx->lost_packets); + s = format(s, " Reordered packets : %llu\n", rx->reordered_packets); + s = format(s, " Duplicate packets : %llu\n", rx->dup_packets); + } + + format(s, "\n"); + return s; +} diff --git a/src/plugins/ioam/encap/ip6_ioam_seqno.h b/src/plugins/ioam/encap/ip6_ioam_seqno.h new file mode 100644 index 00000000..13a84db0 --- /dev/null +++ b/src/plugins/ioam/encap/ip6_ioam_seqno.h @@ -0,0 +1,70 @@ +/* + * Copyright (c) 2016 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +#ifndef __included_ip6_ioam_seqno_h__ +#define __included_ip6_ioam_seqno_h__ + +#include +#include + +#define SEQ_CHECK_VALUE 0x80000000 /* for seq number wraparound detection */ + +#define SEQNO_WINDOW_SIZE 2048 +#define SEQNO_WINDOW_ARRAY_SIZE 64 + +typedef struct seqno_bitmap_ { + u32 window_size; + u32 array_size; + u32 mask; + u32 pad; + u64 highest; + u64 array[SEQNO_WINDOW_ARRAY_SIZE]; /* Will be alloc to array_size */ +} seqno_bitmap; + +typedef struct seqno_rx_info_ { + u64 rx_packets; + u64 lost_packets; + u64 reordered_packets; + u64 dup_packets; + seqno_bitmap bitmap; +} seqno_rx_info; + +/* This structure is 64-byte aligned */ +typedef struct ioam_seqno_data_ { + union { + u32 seq_num; /* Useful only for encap node */ + seqno_rx_info seqno_rx; + }; +} ioam_seqno_data; + +typedef struct ioam_seqno_data_main_t_ { + ioam_seqno_data *seqno_data; +} ioam_seqno_data_main_t; + +void ioam_seqno_init_bitmap(ioam_seqno_data *data); + +int ioam_seqno_encap_handler(vlib_buffer_t *b, ip6_header_t *ip, + ip6_hop_by_hop_option_t *opt); + +int +ioam_seqno_decap_handler(vlib_buffer_t *b, ip6_header_t *ip, + ip6_hop_by_hop_option_t *opt); + +void ioam_analyze_seqno(seqno_rx_info *ppc_rx, u64 seqno); + +u8 * +show_ioam_seqno_cmd_fn(u8 *s, ioam_seqno_data *seqno_data, u8 enc); + +#endif diff --git a/src/plugins/ioam/encap/ip6_ioam_seqno_analyse.c b/src/plugins/ioam/encap/ip6_ioam_seqno_analyse.c new file mode 100644 index 00000000..4638871c --- /dev/null +++ b/src/plugins/ioam/encap/ip6_ioam_seqno_analyse.c @@ -0,0 +1,141 @@ +/* + * Copyright (c) 2016 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +#include +#include "ip6_ioam_seqno.h" + +static inline void BIT_SET (u64 *p, u32 n) +{ + p[ n>>5 ] |= (1 << (n&31)); +} + +static inline int BIT_TEST (u64 *p, u32 n) +{ + return p[ n>>5 ] & (1 << (n&31)); +} + +static void BIT_CLEAR (u64 *p, u64 start, int num_bits, u32 mask) +{ + int n, t; + int start_index = (start >> 5); + int mask_index = (mask >> 5); + + start_index &= mask_index; + if (start & 0x1f) + { + int start_bit = (start & 0x1f); + + n = (1 << start_bit)-1; + t = start_bit + num_bits; + if (t < 32) + { + n |= ~((1 << t)-1); + p[ start_index ] &= n; + return; + } + p[ start_index ] &= n; + start_index = (start_index + 1) & mask_index; + num_bits -= (32 - start_bit); + } + while (num_bits >= 32) + { + p[ start_index ] = 0; + start_index = (start_index + 1) & mask_index; + num_bits -= 32; + } + n = ~((1 << num_bits) - 1); + p[ start_index ] &= n; +} + +static inline u8 seqno_check_wraparound(u32 a, u32 b) +{ + if ((a != b) && (a > b) && ((a - b) > SEQ_CHECK_VALUE)) + { + return 1; + } + return 0; +} + +/* + * Function to analyze the PPC value recevied. + * - Updates the bitmap with received sequence number + * - counts the received/lost/duplicate/reordered packets + */ +void ioam_analyze_seqno (seqno_rx_info *seqno_rx, u64 seqno) +{ + int diff; + static int peer_dead_count; + seqno_bitmap *bitmap = &seqno_rx->bitmap; + + seqno_rx->rx_packets++; + + if (seqno > bitmap->highest) + { /* new larger sequence number */ + peer_dead_count = 0; + diff = seqno - bitmap->highest; + if (diff < bitmap->window_size) + { + if (diff > 1) + { /* diff==1 is *such* a common case it's a win to optimize it */ + BIT_CLEAR(bitmap->array, bitmap->highest+1, diff-1, bitmap->mask); + seqno_rx->lost_packets += diff -1; + } + } + else + { + seqno_rx->lost_packets += diff -1; + memset( bitmap->array, 0, bitmap->array_size * sizeof(u64) ); + } + BIT_SET(bitmap->array, seqno & bitmap->mask); + bitmap->highest = seqno; + return; + } + + /* we've seen a bigger seq number before */ + diff = bitmap->highest - seqno; + if (diff >= bitmap->window_size) + { + if (seqno_check_wraparound(bitmap->highest, seqno)) + { + memset( bitmap->array, 0, bitmap->array_size * sizeof(u64)); + BIT_SET(bitmap->array, seqno & bitmap->mask); + bitmap->highest = seqno; + return; + } + else + { + peer_dead_count++; + if (peer_dead_count > 25) + { + peer_dead_count = 0; + memset( bitmap->array, 0, bitmap->array_size * sizeof(u64) ); + BIT_SET(bitmap->array, seqno & bitmap->mask); + bitmap->highest = seqno; + } + //ppc_rx->reordered_packets++; + } + return; + } + + if (BIT_TEST(bitmap->array, seqno & bitmap->mask)) + { + seqno_rx->dup_packets++; + return; /* Already seen */ + } + seqno_rx->reordered_packets++; + seqno_rx->lost_packets--; + BIT_SET(bitmap->array, seqno & bitmap->mask); + return; +} diff --git a/src/plugins/ioam/encap/ip6_ioam_trace.c b/src/plugins/ioam/encap/ip6_ioam_trace.c new file mode 100644 index 00000000..e63db6e4 --- /dev/null +++ b/src/plugins/ioam/encap/ip6_ioam_trace.c @@ -0,0 +1,438 @@ +/* + * Copyright (c) 2016 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ +#include +#include +#include +#include + +#include +#include +#include + +#include +#include +#include + +#include + +/* Timestamp precision multipliers for seconds, milliseconds, microseconds + * and nanoseconds respectively. + */ +static f64 trace_tsp_mul[4] = { 1, 1e3, 1e6, 1e9 }; + +typedef union +{ + u64 as_u64; + u32 as_u32[2]; +} time_u64_t; + +/* *INDENT-OFF* */ +typedef CLIB_PACKED(struct { + ip6_hop_by_hop_option_t hdr; + u8 ioam_trace_type; + u8 data_list_elts_left; + u32 elts[0]; /* Variable type. So keep it generic */ +}) ioam_trace_option_t; +/* *INDENT-ON* */ + + +extern ip6_hop_by_hop_ioam_main_t ip6_hop_by_hop_ioam_main; +extern ip6_main_t ip6_main; + +#define foreach_ip6_hop_by_hop_ioam_trace_stats \ + _(PROCESSED, "Pkts with ip6 hop-by-hop trace options") \ + _(PROFILE_MISS, "Pkts with ip6 hop-by-hop trace options but no profile set") \ + _(UPDATED, "Pkts with trace updated") \ + _(FULL, "Pkts with trace options but no space") + +static char *ip6_hop_by_hop_ioam_trace_stats_strings[] = { +#define _(sym,string) string, + foreach_ip6_hop_by_hop_ioam_trace_stats +#undef _ +}; + +typedef enum +{ +#define _(sym,str) IP6_IOAM_TRACE_##sym, + foreach_ip6_hop_by_hop_ioam_trace_stats +#undef _ + IP6_IOAM_TRACE_N_STATS, +} ip6_ioam_trace_stats_t; + + +typedef struct +{ + /* stats */ + u64 counters[ARRAY_LEN (ip6_hop_by_hop_ioam_trace_stats_strings)]; + + /* convenience */ + vlib_main_t *vlib_main; + vnet_main_t *vnet_main; +} ip6_hop_by_hop_ioam_trace_main_t; + +ip6_hop_by_hop_ioam_trace_main_t ip6_hop_by_hop_ioam_trace_main; + +always_inline void +ip6_ioam_trace_stats_increment_counter (u32 counter_index, u64 increment) +{ + ip6_hop_by_hop_ioam_trace_main_t *hm = &ip6_hop_by_hop_ioam_trace_main; + + hm->counters[counter_index] += increment; +} + + +static u8 * +format_ioam_data_list_element (u8 * s, va_list * args) +{ + u32 *elt = va_arg (*args, u32 *); + u8 *trace_type_p = va_arg (*args, u8 *); + u8 trace_type = *trace_type_p; + + + if (trace_type & BIT_TTL_NODEID) + { + u32 ttl_node_id_host_byte_order = clib_net_to_host_u32 (*elt); + s = format (s, "ttl 0x%x node id 0x%x ", + ttl_node_id_host_byte_order >> 24, + ttl_node_id_host_byte_order & 0x00FFFFFF); + + elt++; + } + + if (trace_type & BIT_ING_INTERFACE && trace_type & BIT_ING_INTERFACE) + { + u32 ingress_host_byte_order = clib_net_to_host_u32 (*elt); + s = format (s, "ingress 0x%x egress 0x%x ", + ingress_host_byte_order >> 16, + ingress_host_byte_order & 0xFFFF); + elt++; + } + + if (trace_type & BIT_TIMESTAMP) + { + u32 ts_in_host_byte_order = clib_net_to_host_u32 (*elt); + s = format (s, "ts 0x%x \n", ts_in_host_byte_order); + elt++; + } + + if (trace_type & BIT_APPDATA) + { + u32 appdata_in_host_byte_order = clib_net_to_host_u32 (*elt); + s = format (s, "app 0x%x ", appdata_in_host_byte_order); + elt++; + } + + return s; +} + + +int +ip6_ioam_trace_get_sizeof_handler (u32 * result) +{ + u16 size = 0; + u8 trace_data_size = 0; + trace_profile *profile = NULL; + + *result = 0; + + profile = trace_profile_find (); + + if (PREDICT_FALSE (!profile)) + { + ip6_ioam_trace_stats_increment_counter (IP6_IOAM_TRACE_PROFILE_MISS, 1); + return (-1); + } + + trace_data_size = fetch_trace_data_size (profile->trace_type); + if (PREDICT_FALSE (trace_data_size == 0)) + return VNET_API_ERROR_INVALID_VALUE; + + if (PREDICT_FALSE (profile->num_elts * trace_data_size > 254)) + return VNET_API_ERROR_INVALID_VALUE; + + size += + sizeof (ioam_trace_option_t) + (profile->num_elts * trace_data_size); + *result = size; + + return 0; +} + + + +int +ip6_hop_by_hop_ioam_trace_rewrite_handler (u8 * rewrite_string, + u8 * rewrite_size) +{ + ioam_trace_option_t *trace_option = NULL; + u8 trace_data_size = 0; + u8 trace_option_elts = 0; + trace_profile *profile = NULL; + + + profile = trace_profile_find (); + + if (PREDICT_FALSE (!profile)) + { + ip6_ioam_trace_stats_increment_counter (IP6_IOAM_TRACE_PROFILE_MISS, 1); + return (-1); + } + + if (PREDICT_FALSE (!rewrite_string)) + return -1; + + trace_option_elts = profile->num_elts; + trace_data_size = fetch_trace_data_size (profile->trace_type); + trace_option = (ioam_trace_option_t *) rewrite_string; + trace_option->hdr.type = HBH_OPTION_TYPE_IOAM_TRACE_DATA_LIST | + HBH_OPTION_TYPE_DATA_CHANGE_ENROUTE; + trace_option->hdr.length = 2 /*ioam_trace_type,data_list_elts_left */ + + trace_option_elts * trace_data_size; + trace_option->ioam_trace_type = profile->trace_type & TRACE_TYPE_MASK; + trace_option->data_list_elts_left = trace_option_elts; + *rewrite_size = + sizeof (ioam_trace_option_t) + (trace_option_elts * trace_data_size); + + return 0; +} + + +int +ip6_hbh_ioam_trace_data_list_handler (vlib_buffer_t * b, ip6_header_t * ip, + ip6_hop_by_hop_option_t * opt) +{ + ip6_main_t *im = &ip6_main; + ip_lookup_main_t *lm = &im->lookup_main; + ip6_hop_by_hop_ioam_main_t *hm = &ip6_hop_by_hop_ioam_main; + u8 elt_index = 0; + ioam_trace_option_t *trace = (ioam_trace_option_t *) opt; + u32 adj_index = vnet_buffer (b)->ip.adj_index[VLIB_TX]; + ip_adjacency_t *adj = ip_get_adjacency (lm, adj_index); + time_u64_t time_u64; + u32 *elt; + int rv = 0; + trace_profile *profile = NULL; + + + profile = trace_profile_find (); + + if (PREDICT_FALSE (!profile)) + { + ip6_ioam_trace_stats_increment_counter (IP6_IOAM_TRACE_PROFILE_MISS, 1); + return (-1); + } + + + time_u64.as_u64 = 0; + + if (PREDICT_TRUE (trace->data_list_elts_left)) + { + trace->data_list_elts_left--; + /* fetch_trace_data_size returns in bytes. Convert it to 4-bytes + * to skip to this node's location. + */ + elt_index = + trace->data_list_elts_left * + fetch_trace_data_size (trace->ioam_trace_type) / 4; + elt = &trace->elts[elt_index]; + if (trace->ioam_trace_type & BIT_TTL_NODEID) + { + *elt = + clib_host_to_net_u32 ((ip->hop_limit << 24) | profile->node_id); + elt++; + } + + if (trace->ioam_trace_type & BIT_ING_INTERFACE) + { + *elt = + (vnet_buffer (b)->sw_if_index[VLIB_RX] & 0xFFFF) << 16 | + (adj->rewrite_header.sw_if_index & 0xFFFF); + *elt = clib_host_to_net_u32 (*elt); + elt++; + } + + if (trace->ioam_trace_type & BIT_TIMESTAMP) + { + /* Send least significant 32 bits */ + f64 time_f64 = + (f64) (((f64) hm->unix_time_0) + + (vlib_time_now (hm->vlib_main) - hm->vlib_time_0)); + + time_u64.as_u64 = time_f64 * trace_tsp_mul[profile->trace_tsp]; + *elt = clib_host_to_net_u32 (time_u64.as_u32[0]); + elt++; + } + + if (trace->ioam_trace_type & BIT_APPDATA) + { + /* $$$ set elt0->app_data */ + *elt = clib_host_to_net_u32 (profile->app_data); + elt++; + } + ip6_ioam_trace_stats_increment_counter (IP6_IOAM_TRACE_UPDATED, 1); + } + else + { + ip6_ioam_trace_stats_increment_counter (IP6_IOAM_TRACE_FULL, 1); + } + return (rv); +} + +u8 * +ip6_hbh_ioam_trace_data_list_trace_handler (u8 * s, + ip6_hop_by_hop_option_t * opt) +{ + ioam_trace_option_t *trace; + u8 trace_data_size_in_words = 0; + u32 *elt; + int elt_index = 0; + + trace = (ioam_trace_option_t *) opt; + s = + format (s, " Trace Type 0x%x , %d elts left\n", trace->ioam_trace_type, + trace->data_list_elts_left); + trace_data_size_in_words = + fetch_trace_data_size (trace->ioam_trace_type) / 4; + elt = &trace->elts[0]; + while ((u8 *) elt < ((u8 *) (&trace->elts[0]) + trace->hdr.length - 2 + /* -2 accounts for ioam_trace_type,elts_left */ )) + { + s = format (s, " [%d] %U\n", elt_index, + format_ioam_data_list_element, + elt, &trace->ioam_trace_type); + elt_index++; + elt += trace_data_size_in_words; + } + return (s); +} + + +static clib_error_t * +ip6_show_ioam_trace_cmd_fn (vlib_main_t * vm, + unformat_input_t * input, + vlib_cli_command_t * cmd) +{ + ip6_hop_by_hop_ioam_trace_main_t *hm = &ip6_hop_by_hop_ioam_trace_main; + u8 *s = 0; + int i = 0; + + for (i = 0; i < IP6_IOAM_TRACE_N_STATS; i++) + { + s = + format (s, " %s - %lu\n", ip6_hop_by_hop_ioam_trace_stats_strings[i], + hm->counters[i]); + } + + vlib_cli_output (vm, "%v", s); + vec_free (s); + return 0; +} + + +/* *INDENT-OFF* */ +VLIB_CLI_COMMAND (ip6_show_ioam_trace_cmd, static) = { + .path = "show ioam trace", + .short_help = "iOAM trace statistics", + .function = ip6_show_ioam_trace_cmd_fn, +}; +/* *INDENT-ON* */ + + +static clib_error_t * +ip6_hop_by_hop_ioam_trace_init (vlib_main_t * vm) +{ + ip6_hop_by_hop_ioam_trace_main_t *hm = &ip6_hop_by_hop_ioam_trace_main; + clib_error_t *error; + + if ((error = vlib_call_init_function (vm, ip_main_init))) + return (error); + + if ((error = vlib_call_init_function (vm, ip6_lookup_init))) + return error; + + if ((error = vlib_call_init_function (vm, ip6_hop_by_hop_ioam_init))) + return (error); + + hm->vlib_main = vm; + hm->vnet_main = vnet_get_main (); + memset (hm->counters, 0, sizeof (hm->counters)); + + + if (ip6_hbh_register_option + (HBH_OPTION_TYPE_IOAM_TRACE_DATA_LIST, + ip6_hbh_ioam_trace_data_list_handler, + ip6_hbh_ioam_trace_data_list_trace_handler) < 0) + return (clib_error_create + ("registration of HBH_OPTION_TYPE_IOAM_TRACE_DATA_LIST failed")); + + + if (ip6_hbh_add_register_option (HBH_OPTION_TYPE_IOAM_TRACE_DATA_LIST, + sizeof (ioam_trace_option_t), + ip6_hop_by_hop_ioam_trace_rewrite_handler) + < 0) + return (clib_error_create + ("registration of HBH_OPTION_TYPE_IOAM_TRACE_DATA_LIST for rewrite failed")); + + + return (0); +} + +int +ip6_trace_profile_cleanup (void) +{ + ip6_hop_by_hop_ioam_main_t *hm = &ip6_hop_by_hop_ioam_main; + + hm->options_size[HBH_OPTION_TYPE_IOAM_TRACE_DATA_LIST] = 0; + + return 0; + +} + + +int +ip6_trace_profile_setup (void) +{ + u32 trace_size = 0; + ip6_hop_by_hop_ioam_main_t *hm = &ip6_hop_by_hop_ioam_main; + + trace_profile *profile = NULL; + + + profile = trace_profile_find (); + + if (PREDICT_FALSE (!profile)) + { + ip6_ioam_trace_stats_increment_counter (IP6_IOAM_TRACE_PROFILE_MISS, 1); + return (-1); + } + + + if (ip6_ioam_trace_get_sizeof_handler (&trace_size) < 0) + return (-1); + + hm->options_size[HBH_OPTION_TYPE_IOAM_TRACE_DATA_LIST] = trace_size; + + return (0); +} + + +VLIB_INIT_FUNCTION (ip6_hop_by_hop_ioam_trace_init); + +/* + * fd.io coding-style-patch-verification: ON + * + * Local Variables: + * eval: (c-set-style "gnu") + * End: + */ diff --git a/src/plugins/ioam/export-common/ioam_export.h b/src/plugins/ioam/export-common/ioam_export.h new file mode 100644 index 00000000..fbd86180 --- /dev/null +++ b/src/plugins/ioam/export-common/ioam_export.h @@ -0,0 +1,616 @@ +/* + * Copyright (c) 2016 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ +#ifndef __included_ioam_export_h__ +#define __included_ioam_export_h__ + +#include +#include +#include +#include +#include +#include +#include + +#include +#include +#include +#include + +#include + +typedef struct ioam_export_buffer +{ + /* Allocated buffer */ + u32 buffer_index; + u64 touched_at; + u8 records_in_this_buffer; +} ioam_export_buffer_t; + + +typedef struct +{ + /* API message ID base */ + u16 msg_id_base; + + /* TODO: to support multiple collectors all this has to be grouped and create a vector here */ + u8 *record_header; + u32 sequence_number; + u32 domain_id; + + /* ipfix collector, our ip address */ + ip4_address_t ipfix_collector; + ip4_address_t src_address; + + /* Pool of ioam_export_buffer_t */ + ioam_export_buffer_t *buffer_pool; + /* Vector of per thread ioam_export_buffer_t to buffer pool index */ + u32 *buffer_per_thread; + /* Lock per thread to swap buffers between worker and timer process */ + volatile u32 **lockp; + + /* time scale transform */ + u32 unix_time_0; + f64 vlib_time_0; + + /* convenience */ + vlib_main_t *vlib_main; + vnet_main_t *vnet_main; + u32 ip4_lookup_node_index; + + uword my_hbh_slot; + u32 export_process_node_index; +} ioam_export_main_t; + +ioam_export_main_t ioam_export_main; +ioam_export_main_t vxlan_gpe_ioam_export_main; + +extern vlib_node_registration_t export_node; + +#define DEFAULT_EXPORT_SIZE (3 * CLIB_CACHE_LINE_BYTES) +/* + * Number of records in a buffer + * ~(MTU (1500) - [ip hdr(40) + UDP(8) + ipfix (24)]) / DEFAULT_EXPORT_SIZE + */ +#define DEFAULT_EXPORT_RECORDS 7 + +always_inline ioam_export_buffer_t * +ioam_export_get_my_buffer (ioam_export_main_t * em, u32 thread_id) +{ + + if (vec_len (em->buffer_per_thread) > thread_id) + return (pool_elt_at_index + (em->buffer_pool, em->buffer_per_thread[thread_id])); + return (0); +} + +inline static int +ioam_export_buffer_add_header (ioam_export_main_t * em, vlib_buffer_t * b0) +{ + clib_memcpy (b0->data, em->record_header, vec_len (em->record_header)); + b0->current_data = 0; + b0->current_length = vec_len (em->record_header); + b0->flags |= VLIB_BUFFER_TOTAL_LENGTH_VALID; + return (1); +} + +inline static int +ioam_export_init_buffer (ioam_export_main_t * em, vlib_main_t * vm, + ioam_export_buffer_t * eb) +{ + vlib_buffer_t *b = 0; + + if (!eb) + return (-1); + /* TODO: Perhaps buffer init from template here */ + if (vlib_buffer_alloc (vm, &(eb->buffer_index), 1) != 1) + return (-2); + eb->records_in_this_buffer = 0; + eb->touched_at = vlib_time_now (vm); + b = vlib_get_buffer (vm, eb->buffer_index); + (void) ioam_export_buffer_add_header (em, b); + vnet_buffer (b)->sw_if_index[VLIB_RX] = 0; + vnet_buffer (b)->sw_if_index[VLIB_TX] = ~0; + return (1); +} + +inline static void +ioam_export_thread_buffer_free (ioam_export_main_t * em) +{ + vlib_main_t *vm = em->vlib_main; + ioam_export_buffer_t *eb = 0; + int i; + for (i = 0; i < vec_len (em->buffer_per_thread); i++) + { + eb = pool_elt_at_index (em->buffer_pool, em->buffer_per_thread[i]); + if (eb) + vlib_buffer_free (vm, &(eb->buffer_index), 1); + } + for (i = 0; i < vec_len (em->lockp); i++) + clib_mem_free ((void *) em->lockp[i]); + vec_free (em->buffer_per_thread); + pool_free (em->buffer_pool); + vec_free (em->lockp); + em->buffer_per_thread = 0; + em->buffer_pool = 0; + em->lockp = 0; +} + +inline static int +ioam_export_thread_buffer_init (ioam_export_main_t * em, vlib_main_t * vm) +{ + int no_of_threads = vec_len (vlib_worker_threads); + int i; + ioam_export_buffer_t *eb = 0; + vlib_node_t *ip4_lookup_node; + + pool_alloc_aligned (em->buffer_pool, + no_of_threads - 1, CLIB_CACHE_LINE_BYTES); + vec_validate_aligned (em->buffer_per_thread, + no_of_threads - 1, CLIB_CACHE_LINE_BYTES); + vec_validate_aligned (em->lockp, no_of_threads - 1, CLIB_CACHE_LINE_BYTES); + ip4_lookup_node = vlib_get_node_by_name (vm, (u8 *) "ip4-lookup"); + em->ip4_lookup_node_index = ip4_lookup_node->index; + if (!em->buffer_per_thread || !em->buffer_pool || !em->lockp) + { + return (-1); + } + for (i = 0; i < no_of_threads; i++) + { + eb = 0; + pool_get_aligned (em->buffer_pool, eb, CLIB_CACHE_LINE_BYTES); + memset (eb, 0, sizeof (*eb)); + em->buffer_per_thread[i] = eb - em->buffer_pool; + if (ioam_export_init_buffer (em, vm, eb) != 1) + { + ioam_export_thread_buffer_free (em); + return (-2); + } + em->lockp[i] = clib_mem_alloc_aligned (CLIB_CACHE_LINE_BYTES, + CLIB_CACHE_LINE_BYTES); + memset ((void *) em->lockp[i], 0, CLIB_CACHE_LINE_BYTES); + } + return (1); +} + +#define IPFIX_IOAM_EXPORT_ID 272 + +/* Used to build the rewrite */ +/* data set packet */ +typedef struct +{ + ipfix_message_header_t h; + ipfix_set_header_t s; +} ipfix_data_packet_t; + +typedef struct +{ + ip4_header_t ip4; + udp_header_t udp; + ipfix_data_packet_t ipfix; +} ip4_ipfix_data_packet_t; + + +inline static void +ioam_export_header_cleanup (ioam_export_main_t * em, + ip4_address_t * collector_address, + ip4_address_t * src_address) +{ + vec_free (em->record_header); + em->record_header = 0; +} + +inline static int +ioam_export_header_create (ioam_export_main_t * em, + ip4_address_t * collector_address, + ip4_address_t * src_address) +{ + ip4_header_t *ip; + udp_header_t *udp; + ipfix_message_header_t *h; + ipfix_set_header_t *s; + u8 *rewrite = 0; + ip4_ipfix_data_packet_t *tp; + + + /* allocate rewrite space */ + vec_validate_aligned (rewrite, + sizeof (ip4_ipfix_data_packet_t) - 1, + CLIB_CACHE_LINE_BYTES); + + tp = (ip4_ipfix_data_packet_t *) rewrite; + ip = (ip4_header_t *) & tp->ip4; + udp = (udp_header_t *) (ip + 1); + h = (ipfix_message_header_t *) (udp + 1); + s = (ipfix_set_header_t *) (h + 1); + + ip->ip_version_and_header_length = 0x45; + ip->ttl = 254; + ip->protocol = IP_PROTOCOL_UDP; + ip->src_address.as_u32 = src_address->as_u32; + ip->dst_address.as_u32 = collector_address->as_u32; + udp->src_port = clib_host_to_net_u16 (4939 /* $$FIXME */ ); + udp->dst_port = clib_host_to_net_u16 (4939); + /* FIXUP: UDP length */ + udp->length = clib_host_to_net_u16 (vec_len (rewrite) + + (DEFAULT_EXPORT_RECORDS * + DEFAULT_EXPORT_SIZE) - sizeof (*ip)); + + /* FIXUP: message header export_time */ + /* FIXUP: message header sequence_number */ + h->domain_id = clib_host_to_net_u32 (em->domain_id); + + /*FIXUP: Setid length in octets if records exported are not default */ + s->set_id_length = ipfix_set_id_length (IPFIX_IOAM_EXPORT_ID, + (sizeof (*s) + + (DEFAULT_EXPORT_RECORDS * + DEFAULT_EXPORT_SIZE))); + + /* FIXUP: h version and length length in octets if records exported are not default */ + h->version_length = version_length (sizeof (*h) + + (sizeof (*s) + + (DEFAULT_EXPORT_RECORDS * + DEFAULT_EXPORT_SIZE))); + + /* FIXUP: ip length if records exported are not default */ + /* FIXUP: ip checksum if records exported are not default */ + ip->length = clib_host_to_net_u16 (vec_len (rewrite) + + (DEFAULT_EXPORT_RECORDS * + DEFAULT_EXPORT_SIZE)); + ip->checksum = ip4_header_checksum (ip); + _vec_len (rewrite) = sizeof (ip4_ipfix_data_packet_t); + em->record_header = rewrite; + return (1); +} + +inline static int +ioam_export_send_buffer (ioam_export_main_t * em, vlib_main_t * vm, + ioam_export_buffer_t * eb) +{ + ip4_header_t *ip; + udp_header_t *udp; + ipfix_message_header_t *h; + ipfix_set_header_t *s; + ip4_ipfix_data_packet_t *tp; + vlib_buffer_t *b0; + u16 new_l0, old_l0; + ip_csum_t sum0; + vlib_frame_t *nf = 0; + u32 *to_next; + + b0 = vlib_get_buffer (vm, eb->buffer_index); + tp = vlib_buffer_get_current (b0); + ip = (ip4_header_t *) & tp->ip4; + udp = (udp_header_t *) (ip + 1); + h = (ipfix_message_header_t *) (udp + 1); + s = (ipfix_set_header_t *) (h + 1); + + /* FIXUP: message header export_time */ + h->export_time = clib_host_to_net_u32 ((u32) + (((f64) em->unix_time_0) + + (vlib_time_now (em->vlib_main) - + em->vlib_time_0))); + + /* FIXUP: message header sequence_number */ + h->sequence_number = clib_host_to_net_u32 (em->sequence_number++); + + /* FIXUP: lengths if different from default */ + if (PREDICT_FALSE (eb->records_in_this_buffer != DEFAULT_EXPORT_RECORDS)) + { + s->set_id_length = + ipfix_set_id_length (IPFIX_IOAM_EXPORT_ID /* set_id */ , + b0->current_length - (sizeof (*ip) + + sizeof (*udp) + + sizeof (*h))); + h->version_length = + version_length (b0->current_length - (sizeof (*ip) + sizeof (*udp))); + sum0 = ip->checksum; + old_l0 = ip->length; + new_l0 = clib_host_to_net_u16 ((u16) b0->current_length); + sum0 = ip_csum_update (sum0, old_l0, new_l0, ip4_header_t, + length /* changed member */ ); + ip->checksum = ip_csum_fold (sum0); + ip->length = new_l0; + udp->length = clib_host_to_net_u16 (b0->current_length - sizeof (*ip)); + } + + /* Enqueue pkts to ip4-lookup */ + + nf = vlib_get_frame_to_node (vm, em->ip4_lookup_node_index); + nf->n_vectors = 0; + to_next = vlib_frame_vector_args (nf); + nf->n_vectors = 1; + to_next[0] = eb->buffer_index; + vlib_put_frame_to_node (vm, em->ip4_lookup_node_index, nf); + return (1); + +} + +#define EXPORT_TIMEOUT (20.0) +#define THREAD_PERIOD (30.0) +inline static uword +ioam_export_process_common (ioam_export_main_t * em, vlib_main_t * vm, + vlib_node_runtime_t * rt, vlib_frame_t * f, + u32 index) +{ + f64 now; + f64 timeout = 30.0; + uword event_type; + uword *event_data = 0; + int i; + ioam_export_buffer_t *eb = 0, *new_eb = 0; + u32 *vec_buffer_indices = 0; + u32 *vec_buffer_to_be_sent = 0; + u32 *thread_index = 0; + u32 new_pool_index = 0; + + em->export_process_node_index = index; + /* Wait for Godot... */ + vlib_process_wait_for_event_or_clock (vm, 1e9); + event_type = vlib_process_get_events (vm, &event_data); + if (event_type != 1) + clib_warning ("bogus kickoff event received, %d", event_type); + vec_reset_length (event_data); + + while (1) + { + vlib_process_wait_for_event_or_clock (vm, timeout); + event_type = vlib_process_get_events (vm, &event_data); + switch (event_type) + { + case 2: /* Stop and Wait for kickoff again */ + timeout = 1e9; + break; + case 1: /* kickoff : Check for unsent buffers */ + timeout = THREAD_PERIOD; + break; + case ~0: /* timeout */ + break; + } + vec_reset_length (event_data); + now = vlib_time_now (vm); + /* + * Create buffers for threads that are not active enough + * to send out the export records + */ + for (i = 0; i < vec_len (em->buffer_per_thread); i++) + { + /* If the worker thread is processing export records ignore further checks */ + if (*em->lockp[i] == 1) + continue; + eb = pool_elt_at_index (em->buffer_pool, em->buffer_per_thread[i]); + if (eb->records_in_this_buffer > 0 + && now > (eb->touched_at + EXPORT_TIMEOUT)) + { + pool_get_aligned (em->buffer_pool, new_eb, + CLIB_CACHE_LINE_BYTES); + memset (new_eb, 0, sizeof (*new_eb)); + if (ioam_export_init_buffer (em, vm, new_eb) == 1) + { + new_pool_index = new_eb - em->buffer_pool; + vec_add (vec_buffer_indices, &new_pool_index, 1); + vec_add (vec_buffer_to_be_sent, &em->buffer_per_thread[i], + 1); + vec_add (thread_index, &i, 1); + } + else + { + pool_put (em->buffer_pool, new_eb); + /*Give up */ + goto CLEANUP; + } + } + } + if (vec_len (thread_index) != 0) + { + /* + * Now swap the buffers out + */ + for (i = 0; i < vec_len (thread_index); i++) + { + while (__sync_lock_test_and_set (em->lockp[thread_index[i]], 1)) + ; + em->buffer_per_thread[thread_index[i]] = + vec_pop (vec_buffer_indices); + *em->lockp[thread_index[i]] = 0; + } + + /* Send the buffers */ + for (i = 0; i < vec_len (vec_buffer_to_be_sent); i++) + { + eb = + pool_elt_at_index (em->buffer_pool, vec_buffer_to_be_sent[i]); + ioam_export_send_buffer (em, vm, eb); + pool_put (em->buffer_pool, eb); + } + } + + CLEANUP: + /* Free any leftover/unused buffers and everything that was allocated */ + for (i = 0; i < vec_len (vec_buffer_indices); i++) + { + new_eb = pool_elt_at_index (em->buffer_pool, vec_buffer_indices[i]); + vlib_buffer_free (vm, &new_eb->buffer_index, 1); + pool_put (em->buffer_pool, new_eb); + } + vec_free (vec_buffer_indices); + vec_free (vec_buffer_to_be_sent); + vec_free (thread_index); + } + return 0; /* not so much */ +} + +#define ioam_export_node_common(EM, VM, N, F, HTYPE, L, V, NEXT) \ +do { \ + u32 n_left_from, *from, *to_next; \ + export_next_t next_index; \ + u32 pkts_recorded = 0; \ + ioam_export_buffer_t *my_buf = 0; \ + vlib_buffer_t *eb0 = 0; \ + u32 ebi0 = 0; \ + from = vlib_frame_vector_args (F); \ + n_left_from = (F)->n_vectors; \ + next_index = (N)->cached_next_index; \ + while (__sync_lock_test_and_set ((EM)->lockp[(VM)->cpu_index], 1)); \ + my_buf = ioam_export_get_my_buffer (EM, (VM)->cpu_index); \ + my_buf->touched_at = vlib_time_now (VM); \ + while (n_left_from > 0) \ + { \ + u32 n_left_to_next; \ + vlib_get_next_frame (VM, N, next_index, to_next, n_left_to_next); \ + while (n_left_from >= 4 && n_left_to_next >= 2) \ + { \ + u32 next0 = NEXT; \ + u32 next1 = NEXT; \ + u32 bi0, bi1; \ + HTYPE *ip0, *ip1; \ + vlib_buffer_t *p0, *p1; \ + u32 ip_len0, ip_len1; \ + { \ + vlib_buffer_t *p2, *p3; \ + p2 = vlib_get_buffer (VM, from[2]); \ + p3 = vlib_get_buffer (VM, from[3]); \ + vlib_prefetch_buffer_header (p2, LOAD); \ + vlib_prefetch_buffer_header (p3, LOAD); \ + CLIB_PREFETCH (p2->data, 3 * CLIB_CACHE_LINE_BYTES, LOAD); \ + CLIB_PREFETCH (p3->data, 3 * CLIB_CACHE_LINE_BYTES, LOAD); \ + } \ + to_next[0] = bi0 = from[0]; \ + to_next[1] = bi1 = from[1]; \ + from += 2; \ + to_next += 2; \ + n_left_from -= 2; \ + n_left_to_next -= 2; \ + p0 = vlib_get_buffer (VM, bi0); \ + p1 = vlib_get_buffer (VM, bi1); \ + ip0 = vlib_buffer_get_current (p0); \ + ip1 = vlib_buffer_get_current (p1); \ + ip_len0 = \ + clib_net_to_host_u16 (ip0->L) + sizeof (HTYPE); \ + ip_len1 = \ + clib_net_to_host_u16 (ip1->L) + sizeof (HTYPE); \ + ebi0 = my_buf->buffer_index; \ + eb0 = vlib_get_buffer (VM, ebi0); \ + if (PREDICT_FALSE (eb0 == 0)) \ + goto NO_BUFFER1; \ + ip_len0 = \ + ip_len0 > DEFAULT_EXPORT_SIZE ? DEFAULT_EXPORT_SIZE : ip_len0; \ + ip_len1 = \ + ip_len1 > DEFAULT_EXPORT_SIZE ? DEFAULT_EXPORT_SIZE : ip_len1; \ + copy3cachelines (eb0->data + eb0->current_length, ip0, ip_len0); \ + eb0->current_length += DEFAULT_EXPORT_SIZE; \ + my_buf->records_in_this_buffer++; \ + if (my_buf->records_in_this_buffer >= DEFAULT_EXPORT_RECORDS) \ + { \ + ioam_export_send_buffer (EM, VM, my_buf); \ + ioam_export_init_buffer (EM, VM, my_buf); \ + } \ + ebi0 = my_buf->buffer_index; \ + eb0 = vlib_get_buffer (VM, ebi0); \ + if (PREDICT_FALSE (eb0 == 0)) \ + goto NO_BUFFER1; \ + copy3cachelines (eb0->data + eb0->current_length, ip1, ip_len1); \ + eb0->current_length += DEFAULT_EXPORT_SIZE; \ + my_buf->records_in_this_buffer++; \ + if (my_buf->records_in_this_buffer >= DEFAULT_EXPORT_RECORDS) \ + { \ + ioam_export_send_buffer (EM, VM, my_buf); \ + ioam_export_init_buffer (EM, VM, my_buf); \ + } \ + pkts_recorded += 2; \ + if (PREDICT_FALSE (((node)->flags & VLIB_NODE_FLAG_TRACE))) \ + { \ + if (p0->flags & VLIB_BUFFER_IS_TRACED) \ + { \ + export_trace_t *t = \ + vlib_add_trace (VM, node, p0, sizeof (*t)); \ + t->flow_label = \ + clib_net_to_host_u32 (ip0->V); \ + t->next_index = next0; \ + } \ + if (p1->flags & VLIB_BUFFER_IS_TRACED) \ + { \ + export_trace_t *t = \ + vlib_add_trace (VM, N, p1, sizeof (*t)); \ + t->flow_label = \ + clib_net_to_host_u32 (ip1->V); \ + t->next_index = next1; \ + } \ + } \ + NO_BUFFER1: \ + vlib_validate_buffer_enqueue_x2 (VM, N, next_index, \ + to_next, n_left_to_next, \ + bi0, bi1, next0, next1); \ + } \ + while (n_left_from > 0 && n_left_to_next > 0) \ + { \ + u32 bi0; \ + vlib_buffer_t *p0; \ + u32 next0 = NEXT; \ + HTYPE *ip0; \ + u32 ip_len0; \ + bi0 = from[0]; \ + to_next[0] = bi0; \ + from += 1; \ + to_next += 1; \ + n_left_from -= 1; \ + n_left_to_next -= 1; \ + p0 = vlib_get_buffer (VM, bi0); \ + ip0 = vlib_buffer_get_current (p0); \ + ip_len0 = \ + clib_net_to_host_u16 (ip0->L) + sizeof (HTYPE); \ + ebi0 = my_buf->buffer_index; \ + eb0 = vlib_get_buffer (VM, ebi0); \ + if (PREDICT_FALSE (eb0 == 0)) \ + goto NO_BUFFER; \ + ip_len0 = \ + ip_len0 > DEFAULT_EXPORT_SIZE ? DEFAULT_EXPORT_SIZE : ip_len0; \ + copy3cachelines (eb0->data + eb0->current_length, ip0, ip_len0); \ + eb0->current_length += DEFAULT_EXPORT_SIZE; \ + my_buf->records_in_this_buffer++; \ + if (my_buf->records_in_this_buffer >= DEFAULT_EXPORT_RECORDS) \ + { \ + ioam_export_send_buffer (EM, VM, my_buf); \ + ioam_export_init_buffer (EM, VM, my_buf); \ + } \ + if (PREDICT_FALSE (((N)->flags & VLIB_NODE_FLAG_TRACE) \ + && (p0->flags & VLIB_BUFFER_IS_TRACED))) \ + { \ + export_trace_t *t = vlib_add_trace (VM, (N), p0, sizeof (*t)); \ + t->flow_label = \ + clib_net_to_host_u32 (ip0->V); \ + t->next_index = next0; \ + } \ + pkts_recorded += 1; \ + NO_BUFFER: \ + vlib_validate_buffer_enqueue_x1 (VM, N, next_index, \ + to_next, n_left_to_next, \ + bi0, next0); \ + } \ + vlib_put_next_frame (VM, N, next_index, n_left_to_next); \ + } \ + vlib_node_increment_counter (VM, export_node.index, \ + EXPORT_ERROR_RECORDED, pkts_recorded); \ + *(EM)->lockp[(VM)->cpu_index] = 0; \ +} while(0) + +#endif /* __included_ioam_export_h__ */ + +/* + * fd.io coding-style-patch-verification: ON + * + * Local Variables: + * eval: (c-set-style "gnu") + * End: + */ diff --git a/src/plugins/ioam/export-vxlan-gpe/vxlan_gpe_ioam_export.api b/src/plugins/ioam/export-vxlan-gpe/vxlan_gpe_ioam_export.api new file mode 100644 index 00000000..7b17c3f7 --- /dev/null +++ b/src/plugins/ioam/export-vxlan-gpe/vxlan_gpe_ioam_export.api @@ -0,0 +1,42 @@ +/* Hey Emacs use -*- mode: C -*- */ +/* + * Copyright (c) 2016 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +/* Define a simple binary API to control the feature */ + +define vxlan_gpe_ioam_export_enable_disable { + /* Client identifier, set from api_main.my_client_index */ + u32 client_index; + + /* Arbitrary context, so client can match reply to request */ + u32 context; + + /* Enable / disable the feature */ + u8 is_disable; + + /* Collector ip address */ + u8 collector_address[4]; + u8 src_address[4]; + + /* Src ip address */ +}; + +define vxlan_gpe_ioam_export_enable_disable_reply { + /* From the request */ + u32 context; + + /* Return value, zero means all OK */ + i32 retval; +}; \ No newline at end of file diff --git a/src/plugins/ioam/export-vxlan-gpe/vxlan_gpe_ioam_export.c b/src/plugins/ioam/export-vxlan-gpe/vxlan_gpe_ioam_export.c new file mode 100644 index 00000000..bab8d977 --- /dev/null +++ b/src/plugins/ioam/export-vxlan-gpe/vxlan_gpe_ioam_export.c @@ -0,0 +1,271 @@ +/* + * Copyright (c) 2016 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ +/* + *------------------------------------------------------------------ + * vxlan_gpe_ioam_export.c - ioam export API / debug CLI handling + *------------------------------------------------------------------ + */ + +#include +#include +#include +#include + +#include +#include +#include + +#include + +/* define message IDs */ +#include + +/* define message structures */ +#define vl_typedefs +#include +#undef vl_typedefs + +/* define generated endian-swappers */ +#define vl_endianfun +#include +#undef vl_endianfun + +/* instantiate all the print functions we know about */ +#define vl_print(handle, ...) vlib_cli_output (handle, __VA_ARGS__) +#define vl_printfun +#include +#undef vl_printfun + +/* Get the API version number */ +#define vl_api_version(n,v) static u32 api_version=(v); +#include +#undef vl_api_version + +/* + * A handy macro to set up a message reply. + * Assumes that the following variables are available: + * mp - pointer to request message + * rmp - pointer to reply message type + * rv - return value + */ + +#define REPLY_MACRO(t) \ +do { \ + unix_shared_memory_queue_t * q = \ + vl_api_client_index_to_input_queue (mp->client_index); \ + if (!q) \ + return; \ + \ + rmp = vl_msg_api_alloc (sizeof (*rmp)); \ + rmp->_vl_msg_id = ntohs((t)+sm->msg_id_base); \ + rmp->context = mp->context; \ + rmp->retval = ntohl(rv); \ + \ + vl_msg_api_send_shmem (q, (u8 *)&rmp); \ +} while(0); + + +/* List of message types that this plugin understands */ + + +#define foreach_vxlan_gpe_ioam_export_plugin_api_msg \ +_(VXLAN_GPE_IOAM_EXPORT_ENABLE_DISABLE, vxlan_gpe_ioam_export_enable_disable) + +extern void vxlan_gpe_set_next_override (uword next); +/* Action function shared between message handler and debug CLI */ +int +vxlan_gpe_ioam_export_enable_disable (ioam_export_main_t * em, + u8 is_disable, + ip4_address_t * collector_address, + ip4_address_t * src_address) +{ + vlib_main_t *vm = em->vlib_main; + u32 node_index = export_node.index; + vlib_node_t *vxlan_gpe_decap_ioam_node = NULL; + + if (is_disable == 0) + { + if (em->my_hbh_slot == ~0) + { + /* Hook this export node to vxlan-gpe-decap-ioam-v4 */ + vxlan_gpe_decap_ioam_node = + vlib_get_node_by_name (vm, (u8 *) "vxlan-gpe-decap-ioam-v4"); + if (!vxlan_gpe_decap_ioam_node) + { + /* node does not exist give up */ + return (-1); + } + em->my_hbh_slot = + vlib_node_add_next (vm, vxlan_gpe_decap_ioam_node->index, + node_index); + } + if (1 == ioam_export_header_create (em, collector_address, src_address)) + { + ioam_export_thread_buffer_init (em, vm); + vxlan_gpe_set_next_override (em->my_hbh_slot); + /* Turn on the export buffer check process */ + vlib_process_signal_event (vm, em->export_process_node_index, 1, 0); + + } + else + { + return (-2); + } + } + else + { + vxlan_gpe_set_next_override (VXLAN_GPE_DECAP_IOAM_V4_NEXT_POP); + ioam_export_header_cleanup (em, collector_address, src_address); + ioam_export_thread_buffer_free (em); + /* Turn off the export buffer check process */ + vlib_process_signal_event (vm, em->export_process_node_index, 2, 0); + + } + + return 0; +} + +/* API message handler */ +static void vl_api_vxlan_gpe_ioam_export_enable_disable_t_handler + (vl_api_vxlan_gpe_ioam_export_enable_disable_t * mp) +{ + vl_api_vxlan_gpe_ioam_export_enable_disable_reply_t *rmp; + ioam_export_main_t *sm = &vxlan_gpe_ioam_export_main; + int rv; + + rv = vxlan_gpe_ioam_export_enable_disable (sm, (int) (mp->is_disable), + (ip4_address_t *) + mp->collector_address, + (ip4_address_t *) + mp->src_address); + + REPLY_MACRO (VL_API_VXLAN_GPE_IOAM_EXPORT_ENABLE_DISABLE_REPLY); +} /* API message handler */ + + + +/* Set up the API message handling tables */ +static clib_error_t * +vxlan_gpe_ioam_export_plugin_api_hookup (vlib_main_t * vm) +{ + ioam_export_main_t *sm = &vxlan_gpe_ioam_export_main; +#define _(N,n) \ + vl_msg_api_set_handlers((VL_API_##N + sm->msg_id_base), \ + #n, \ + vl_api_##n##_t_handler, \ + vl_noop_handler, \ + vl_api_##n##_t_endian, \ + vl_api_##n##_t_print, \ + sizeof(vl_api_##n##_t), 1); + foreach_vxlan_gpe_ioam_export_plugin_api_msg; +#undef _ + + return 0; +} + + +static clib_error_t * +set_vxlan_gpe_ioam_export_ipfix_command_fn (vlib_main_t * vm, + unformat_input_t * input, + vlib_cli_command_t * cmd) +{ + ioam_export_main_t *em = &vxlan_gpe_ioam_export_main; + ip4_address_t collector, src; + u8 is_disable = 0; + + collector.as_u32 = 0; + src.as_u32 = 0; + + while (unformat_check_input (input) != UNFORMAT_END_OF_INPUT) + { + if (unformat (input, "collector %U", unformat_ip4_address, &collector)) + ; + else if (unformat (input, "src %U", unformat_ip4_address, &src)) + ; + else if (unformat (input, "disable")) + is_disable = 1; + else + break; + } + + if (collector.as_u32 == 0) + return clib_error_return (0, "collector address required"); + + if (src.as_u32 == 0) + return clib_error_return (0, "src address required"); + + em->ipfix_collector.as_u32 = collector.as_u32; + em->src_address.as_u32 = src.as_u32; + + vlib_cli_output (vm, "Collector %U, src address %U", + format_ip4_address, &em->ipfix_collector, + format_ip4_address, &em->src_address); + + /* Turn on the export timer process */ + // vlib_process_signal_event (vm, flow_report_process_node.index, + //1, 0); + if (0 != + vxlan_gpe_ioam_export_enable_disable (em, is_disable, &collector, &src)) + { + return clib_error_return (0, "Unable to set ioam vxlan-gpe export"); + } + + return 0; +} + +/* *INDENT-OFF* */ +VLIB_CLI_COMMAND (set_vxlan_gpe_ioam_ipfix_command, static) = +{ +.path = "set vxlan-gpe-ioam export ipfix", +.short_help = "set vxlan-gpe-ioam export ipfix collector src ", +.function = set_vxlan_gpe_ioam_export_ipfix_command_fn, +}; +/* *INDENT-ON* */ + + +static clib_error_t * +vxlan_gpe_ioam_export_init (vlib_main_t * vm) +{ + ioam_export_main_t *em = &vxlan_gpe_ioam_export_main; + clib_error_t *error = 0; + u8 *name; + + name = format (0, "vxlan_gpe_ioam_export_%08x%c", api_version, 0); + + /* Ask for a correctly-sized block of API message decode slots */ + em->msg_id_base = vl_msg_api_get_msg_ids + ((char *) name, VL_MSG_FIRST_AVAILABLE); + em->unix_time_0 = (u32) time (0); /* Store starting time */ + em->vlib_time_0 = vlib_time_now (vm); + + error = vxlan_gpe_ioam_export_plugin_api_hookup (vm); + em->my_hbh_slot = ~0; + em->vlib_main = vm; + em->vnet_main = vnet_get_main (); + vec_free (name); + + return error; +} + +VLIB_INIT_FUNCTION (vxlan_gpe_ioam_export_init); + + +/* + * fd.io coding-style-patch-verification: ON + * + * Local Variables: + * eval: (c-set-style "gnu") + * End: + */ diff --git a/src/plugins/ioam/export-vxlan-gpe/vxlan_gpe_ioam_export_all_api_h.h b/src/plugins/ioam/export-vxlan-gpe/vxlan_gpe_ioam_export_all_api_h.h new file mode 100644 index 00000000..6d93f093 --- /dev/null +++ b/src/plugins/ioam/export-vxlan-gpe/vxlan_gpe_ioam_export_all_api_h.h @@ -0,0 +1,16 @@ +/* + * Copyright (c) 2016 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ +/* Include the generated file, see BUILT_SOURCES in Makefile.am */ +#include diff --git a/src/plugins/ioam/export-vxlan-gpe/vxlan_gpe_ioam_export_msg_enum.h b/src/plugins/ioam/export-vxlan-gpe/vxlan_gpe_ioam_export_msg_enum.h new file mode 100644 index 00000000..cc5698de --- /dev/null +++ b/src/plugins/ioam/export-vxlan-gpe/vxlan_gpe_ioam_export_msg_enum.h @@ -0,0 +1,28 @@ +/* + * Copyright (c) 2016 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ +#ifndef included_vxlan_gpe_ioam_export_msg_enum_h +#define included_vxlan_gpe_ioam_export_msg_enum_h + +#include + +#define vl_msg_id(n,h) n, +typedef enum { +#include + /* We'll want to know how many messages IDs we need... */ + VL_MSG_FIRST_AVAILABLE, +} vl_msg_id_t; +#undef vl_msg_id + +#endif /* included_vxlan_gpe_ioam_export_msg_enum_h */ diff --git a/src/plugins/ioam/export-vxlan-gpe/vxlan_gpe_ioam_export_test.c b/src/plugins/ioam/export-vxlan-gpe/vxlan_gpe_ioam_export_test.c new file mode 100644 index 00000000..494263d9 --- /dev/null +++ b/src/plugins/ioam/export-vxlan-gpe/vxlan_gpe_ioam_export_test.c @@ -0,0 +1,215 @@ +/* + * Copyright (c) 2016 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ +/* + *------------------------------------------------------------------ + * vxlan_gpe_ioam_export_test.c - test harness plugin + *------------------------------------------------------------------ + */ + +#include +#include +#include +#include +#include + + +/* Declare message IDs */ +#include + +/* define message structures */ +#define vl_typedefs +#include +#undef vl_typedefs + +/* declare message handlers for each api */ + +#define vl_endianfun /* define message structures */ +#include +#undef vl_endianfun + +/* instantiate all the print functions we know about */ +#define vl_print(handle, ...) +#define vl_printfun +#include +#undef vl_printfun + +/* Get the API version number. */ +#define vl_api_version(n,v) static u32 api_version=(v); +#include +#undef vl_api_version + + +typedef struct +{ + /* API message ID base */ + u16 msg_id_base; + vat_main_t *vat_main; +} export_test_main_t; + +export_test_main_t export_test_main; + +#define foreach_standard_reply_retval_handler \ +_(vxlan_gpe_ioam_export_enable_disable_reply) + +#define _(n) \ + static void vl_api_##n##_t_handler \ + (vl_api_##n##_t * mp) \ + { \ + vat_main_t * vam = export_test_main.vat_main; \ + i32 retval = ntohl(mp->retval); \ + if (vam->async_mode) { \ + vam->async_errors += (retval < 0); \ + } else { \ + vam->retval = retval; \ + vam->result_ready = 1; \ + } \ + } +foreach_standard_reply_retval_handler; +#undef _ + +/* + * Table of message reply handlers, must include boilerplate handlers + * we just generated + */ +#define foreach_vpe_api_reply_msg \ +_(VXLAN_GPE_IOAM_EXPORT_ENABLE_DISABLE_REPLY, vxlan_gpe_ioam_export_enable_disable_reply) + + +/* M: construct, but don't yet send a message */ + +#define M(T,t) \ +do { \ + vam->result_ready = 0; \ + mp = vl_msg_api_alloc(sizeof(*mp)); \ + memset (mp, 0, sizeof (*mp)); \ + mp->_vl_msg_id = ntohs (VL_API_##T + sm->msg_id_base); \ + mp->client_index = vam->my_client_index; \ +} while(0); + +#define M2(T,t,n) \ +do { \ + vam->result_ready = 0; \ + mp = vl_msg_api_alloc(sizeof(*mp)+(n)); \ + memset (mp, 0, sizeof (*mp)); \ + mp->_vl_msg_id = ntohs (VL_API_##T + sm->msg_id_base); \ + mp->client_index = vam->my_client_index; \ +} while(0); + +/* S: send a message */ +#define S (vl_msg_api_send_shmem (vam->vl_input_queue, (u8 *)&mp)) + +/* W: wait for results, with timeout */ +#define W \ +do { \ + timeout = vat_time_now (vam) + 1.0; \ + \ + while (vat_time_now (vam) < timeout) { \ + if (vam->result_ready == 1) { \ + return (vam->retval); \ + } \ + } \ + return -99; \ +} while(0); + +static int +api_vxlan_gpe_ioam_export_enable_disable (vat_main_t * vam) +{ + export_test_main_t *sm = &export_test_main; + unformat_input_t *i = vam->input; + f64 timeout; + int is_disable = 0; + vl_api_vxlan_gpe_ioam_export_enable_disable_t *mp; + + /* Parse args required to build the message */ + while (unformat_check_input (i) != UNFORMAT_END_OF_INPUT) + { + if (unformat (i, "disable")) + is_disable = 1; + else + break; + } + + /* Construct the API message */ + M (VXLAN_GPE_IOAM_EXPORT_ENABLE_DISABLE, + vxlan_gpe_ioam_export_enable_disable); + mp->is_disable = is_disable; + + /* send it... */ + S; + + /* Wait for a reply... */ + W; +} + +/* + * List of messages that the api test plugin sends, + * and that the data plane plugin processes + */ +#define foreach_vpe_api_msg \ +_(vxlan_gpe_ioam_export_enable_disable, " [disable]") + +void +vat_api_hookup (vat_main_t * vam) +{ + export_test_main_t *sm = &export_test_main; + /* Hook up handlers for replies from the data plane plug-in */ +#define _(N,n) \ + vl_msg_api_set_handlers((VL_API_##N + sm->msg_id_base), \ + #n, \ + vl_api_##n##_t_handler, \ + vl_noop_handler, \ + vl_api_##n##_t_endian, \ + vl_api_##n##_t_print, \ + sizeof(vl_api_##n##_t), 1); + foreach_vpe_api_reply_msg; +#undef _ + + /* API messages we can send */ +#define _(n,h) hash_set_mem (vam->function_by_name, #n, api_##n); + foreach_vpe_api_msg; +#undef _ + + /* Help strings */ +#define _(n,h) hash_set_mem (vam->help_by_name, #n, h); + foreach_vpe_api_msg; +#undef _ +} + +clib_error_t * +vat_plugin_register (vat_main_t * vam) +{ + export_test_main_t *sm = &export_test_main; + u8 *name; + + sm->vat_main = vam; + + name = format (0, "vxlan_gpe_ioam_export_%08x%c", api_version, 0); + sm->msg_id_base = vl_client_get_first_plugin_msg_id ((char *) name); + + if (sm->msg_id_base != (u16) ~ 0) + vat_api_hookup (vam); + + vec_free (name); + + return 0; +} + +/* + * fd.io coding-style-patch-verification: ON + * + * Local Variables: + * eval: (c-set-style "gnu") + * End: + */ diff --git a/src/plugins/ioam/export-vxlan-gpe/vxlan_gpe_ioam_export_thread.c b/src/plugins/ioam/export-vxlan-gpe/vxlan_gpe_ioam_export_thread.c new file mode 100644 index 00000000..58508ebf --- /dev/null +++ b/src/plugins/ioam/export-vxlan-gpe/vxlan_gpe_ioam_export_thread.c @@ -0,0 +1,49 @@ +/* + * Copyright (c) 2016 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ +/* + * ioam_export_thread.c + */ +#include +#include +#include + +static vlib_node_registration_t vxlan_gpe_ioam_export_process_node; + +static uword +vxlan_gpe_ioam_export_process (vlib_main_t * vm, + vlib_node_runtime_t * rt, vlib_frame_t * f) +{ + return (ioam_export_process_common (&vxlan_gpe_ioam_export_main, + vm, rt, f, + vxlan_gpe_ioam_export_process_node.index)); +} + + +/* *INDENT-OFF* */ +VLIB_REGISTER_NODE (vxlan_gpe_ioam_export_process_node, static) = +{ + .function = vxlan_gpe_ioam_export_process, + .type = VLIB_NODE_TYPE_PROCESS, + .name = "vxlan-gpe-ioam-export-process", +}; +/* *INDENT-ON* */ + +/* + * fd.io coding-style-patch-verification: ON + * + * Local Variables: + * eval: (c-set-style "gnu") + * End: + */ diff --git a/src/plugins/ioam/export-vxlan-gpe/vxlan_gpe_node.c b/src/plugins/ioam/export-vxlan-gpe/vxlan_gpe_node.c new file mode 100644 index 00000000..722c2b06 --- /dev/null +++ b/src/plugins/ioam/export-vxlan-gpe/vxlan_gpe_node.c @@ -0,0 +1,162 @@ +/* + * Copyright (c) 2016 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ +#include +#include +#include +#include +#include +#include +#include +#include + +typedef struct +{ + u32 next_index; + u32 flow_label; +} export_trace_t; + +/* packet trace format function */ +static u8 * +format_export_trace (u8 * s, va_list * args) +{ + CLIB_UNUSED (vlib_main_t * vm) = va_arg (*args, vlib_main_t *); + CLIB_UNUSED (vlib_node_t * node) = va_arg (*args, vlib_node_t *); + export_trace_t *t = va_arg (*args, export_trace_t *); + + s = format (s, "EXPORT: flow_label %d, next index %d", + t->flow_label, t->next_index); + return s; +} + +vlib_node_registration_t export_node; + +#define foreach_export_error \ +_(RECORDED, "Packets recorded for export") + +typedef enum +{ +#define _(sym,str) EXPORT_ERROR_##sym, + foreach_export_error +#undef _ + EXPORT_N_ERROR, +} export_error_t; + +static char *export_error_strings[] = { +#define _(sym,string) string, + foreach_export_error +#undef _ +}; + +typedef enum +{ + EXPORT_NEXT_VXLAN_GPE_INPUT, + EXPORT_N_NEXT, +} export_next_t; + +always_inline void +copy3cachelines (void *dst, const void *src, size_t n) +{ +#if 0 + if (PREDICT_FALSE (n < DEFAULT_EXPORT_SIZE)) + { + /* Copy only the first 1/2 cache lines whatever is available */ + if (n >= 64) + clib_mov64 ((u8 *) dst, (const u8 *) src); + if (n >= 128) + clib_mov64 ((u8 *) dst + 64, (const u8 *) src + 64); + return; + } + clib_mov64 ((u8 *) dst, (const u8 *) src); + clib_mov64 ((u8 *) dst + 64, (const u8 *) src + 64); + clib_mov64 ((u8 *) dst + 128, (const u8 *) src + 128); +#endif +#if 1 + + u64 *copy_dst, *copy_src; + int i; + copy_dst = (u64 *) dst; + copy_src = (u64 *) src; + if (PREDICT_FALSE (n < DEFAULT_EXPORT_SIZE)) + { + for (i = 0; i < n / 64; i++) + { + copy_dst[0] = copy_src[0]; + copy_dst[1] = copy_src[1]; + copy_dst[2] = copy_src[2]; + copy_dst[3] = copy_src[3]; + copy_dst[4] = copy_src[4]; + copy_dst[5] = copy_src[5]; + copy_dst[6] = copy_src[6]; + copy_dst[7] = copy_src[7]; + copy_dst += 8; + copy_src += 8; + } + return; + } + for (i = 0; i < 3; i++) + { + copy_dst[0] = copy_src[0]; + copy_dst[1] = copy_src[1]; + copy_dst[2] = copy_src[2]; + copy_dst[3] = copy_src[3]; + copy_dst[4] = copy_src[4]; + copy_dst[5] = copy_src[5]; + copy_dst[6] = copy_src[6]; + copy_dst[7] = copy_src[7]; + copy_dst += 8; + copy_src += 8; + } +#endif +} + + +static uword +vxlan_gpe_export_node_fn (vlib_main_t * vm, + vlib_node_runtime_t * node, vlib_frame_t * frame) +{ + ioam_export_main_t *em = &vxlan_gpe_ioam_export_main; + ioam_export_node_common (em, vm, node, frame, ip4_header_t, length, + ip_version_and_header_length, + EXPORT_NEXT_VXLAN_GPE_INPUT); + return frame->n_vectors; +} + +/* + * Node for VXLAN-GPE export + */ +/* *INDENT-OFF* */ +VLIB_REGISTER_NODE (export_node) = +{ + .function = vxlan_gpe_export_node_fn, + .name = "vxlan-gpe-ioam-export", + .vector_size = sizeof (u32), + .format_trace = format_export_trace, + .type = VLIB_NODE_TYPE_INTERNAL, + .n_errors = ARRAY_LEN (export_error_strings), + .error_strings = export_error_strings, + .n_next_nodes = EXPORT_N_NEXT, + /* edit / add dispositions here */ + .next_nodes = + {[EXPORT_NEXT_VXLAN_GPE_INPUT] = "vxlan-gpe-pop-ioam-v4"}, +}; +/* *INDENT-ON* */ + +/* + * fd.io coding-style-patch-verification: ON + * + * Local Variables: + * eval: (c-set-style "gnu") + * End: + */ diff --git a/src/plugins/ioam/export/ioam_export.api b/src/plugins/ioam/export/ioam_export.api new file mode 100644 index 00000000..f22d9fc8 --- /dev/null +++ b/src/plugins/ioam/export/ioam_export.api @@ -0,0 +1,42 @@ +/* Hey Emacs use -*- mode: C -*- */ +/* + * Copyright (c) 2016 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +/* Define a simple binary API to control the feature */ + +define ioam_export_ip6_enable_disable { + /* Client identifier, set from api_main.my_client_index */ + u32 client_index; + + /* Arbitrary context, so client can match reply to request */ + u32 context; + + /* Enable / disable the feature */ + u8 is_disable; + + /* Collector ip address */ + u8 collector_address[4]; + u8 src_address[4]; + + /* Src ip address */ +}; + +define ioam_export_ip6_enable_disable_reply { + /* From the request */ + u32 context; + + /* Return value, zero means all OK */ + i32 retval; +}; diff --git a/src/plugins/ioam/export/ioam_export.c b/src/plugins/ioam/export/ioam_export.c new file mode 100644 index 00000000..b122e445 --- /dev/null +++ b/src/plugins/ioam/export/ioam_export.c @@ -0,0 +1,282 @@ +/* + * Copyright (c) 2016 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ +/* + *------------------------------------------------------------------ + * ioam_export.c - ioam export API / debug CLI handling + *------------------------------------------------------------------ + */ + +#include +#include +#include + +#include +#include +#include +#include + + +/* define message IDs */ +#include + +/* define message structures */ +#define vl_typedefs +#include +#undef vl_typedefs + +/* define generated endian-swappers */ +#define vl_endianfun +#include +#undef vl_endianfun + +/* instantiate all the print functions we know about */ +#define vl_print(handle, ...) vlib_cli_output (handle, __VA_ARGS__) +#define vl_printfun +#include +#undef vl_printfun + +/* Get the API version number */ +#define vl_api_version(n,v) static u32 api_version=(v); +#include +#undef vl_api_version + +/* + * A handy macro to set up a message reply. + * Assumes that the following variables are available: + * mp - pointer to request message + * rmp - pointer to reply message type + * rv - return value + */ + +#define REPLY_MACRO(t) \ +do { \ + unix_shared_memory_queue_t * q = \ + vl_api_client_index_to_input_queue (mp->client_index); \ + if (!q) \ + return; \ + \ + rmp = vl_msg_api_alloc (sizeof (*rmp)); \ + rmp->_vl_msg_id = ntohs((t)+sm->msg_id_base); \ + rmp->context = mp->context; \ + rmp->retval = ntohl(rv); \ + \ + vl_msg_api_send_shmem (q, (u8 *)&rmp); \ +} while(0); + + +/* List of message types that this plugin understands */ + +#define foreach_ioam_export_plugin_api_msg \ +_(IOAM_EXPORT_IP6_ENABLE_DISABLE, ioam_export_ip6_enable_disable) + +/* + * This routine exists to convince the vlib plugin framework that + * we haven't accidentally copied a random .dll into the plugin directory. + * + * Also collects global variable pointers passed from the vpp engine + */ + +clib_error_t * +vlib_plugin_register (vlib_main_t * vm, vnet_plugin_handoff_t * h, + int from_early_init) +{ + ioam_export_main_t *em = &ioam_export_main; + clib_error_t *error = 0; + + em->vlib_main = vm; + em->vnet_main = h->vnet_main; + + return error; +} + +/* Action function shared between message handler and debug CLI */ + +int +ioam_export_ip6_enable_disable (ioam_export_main_t * em, + u8 is_disable, + ip4_address_t * collector_address, + ip4_address_t * src_address) +{ + vlib_main_t *vm = em->vlib_main; + + if (is_disable == 0) + { + if (1 == ioam_export_header_create (em, collector_address, src_address)) + { + ioam_export_thread_buffer_init (em, vm); + ip6_hbh_set_next_override (em->my_hbh_slot); + /* Turn on the export buffer check process */ + vlib_process_signal_event (vm, em->export_process_node_index, 1, 0); + + } + else + { + return (-2); + } + } + else + { + ip6_hbh_set_next_override (IP6_LOOKUP_NEXT_POP_HOP_BY_HOP); + ioam_export_header_cleanup (em, collector_address, src_address); + ioam_export_thread_buffer_free (em); + /* Turn off the export buffer check process */ + vlib_process_signal_event (vm, em->export_process_node_index, 2, 0); + + } + + return 0; +} + +/* API message handler */ +static void vl_api_ioam_export_ip6_enable_disable_t_handler + (vl_api_ioam_export_ip6_enable_disable_t * mp) +{ + vl_api_ioam_export_ip6_enable_disable_reply_t *rmp; + ioam_export_main_t *sm = &ioam_export_main; + int rv; + + rv = ioam_export_ip6_enable_disable (sm, (int) (mp->is_disable), + (ip4_address_t *) + mp->collector_address, + (ip4_address_t *) mp->src_address); + + REPLY_MACRO (VL_API_IOAM_EXPORT_IP6_ENABLE_DISABLE_REPLY); +} + +/* Set up the API message handling tables */ +static clib_error_t * +ioam_export_plugin_api_hookup (vlib_main_t * vm) +{ + ioam_export_main_t *sm = &ioam_export_main; +#define _(N,n) \ + vl_msg_api_set_handlers((VL_API_##N + sm->msg_id_base), \ + #n, \ + vl_api_##n##_t_handler, \ + vl_noop_handler, \ + vl_api_##n##_t_endian, \ + vl_api_##n##_t_print, \ + sizeof(vl_api_##n##_t), 1); + foreach_ioam_export_plugin_api_msg; +#undef _ + + return 0; +} + +#define vl_msg_name_crc_list +#include +#undef vl_msg_name_crc_list + +static void +setup_message_id_table (ioam_export_main_t * sm, api_main_t * am) +{ +#define _(id,n,crc) \ + vl_msg_api_add_msg_name_crc (am, #n "_" #crc, id + sm->msg_id_base); + foreach_vl_msg_name_crc_ioam_export; +#undef _ +} + +static clib_error_t * +set_ioam_export_ipfix_command_fn (vlib_main_t * vm, + unformat_input_t * input, + vlib_cli_command_t * cmd) +{ + ioam_export_main_t *em = &ioam_export_main; + ip4_address_t collector, src; + u8 is_disable = 0; + + collector.as_u32 = 0; + src.as_u32 = 0; + + while (unformat_check_input (input) != UNFORMAT_END_OF_INPUT) + { + if (unformat (input, "collector %U", unformat_ip4_address, &collector)) + ; + else if (unformat (input, "src %U", unformat_ip4_address, &src)) + ; + else if (unformat (input, "disable")) + is_disable = 1; + else + break; + } + + if (collector.as_u32 == 0) + return clib_error_return (0, "collector address required"); + + if (src.as_u32 == 0) + return clib_error_return (0, "src address required"); + + em->ipfix_collector.as_u32 = collector.as_u32; + em->src_address.as_u32 = src.as_u32; + + vlib_cli_output (vm, "Collector %U, src address %U", + format_ip4_address, &em->ipfix_collector, + format_ip4_address, &em->src_address); + + /* Turn on the export timer process */ + // vlib_process_signal_event (vm, flow_report_process_node.index, + //1, 0); + ioam_export_ip6_enable_disable (em, is_disable, &collector, &src); + + return 0; +} + +/* *INDENT-OFF* */ +VLIB_CLI_COMMAND (set_ipfix_command, static) = +{ +.path = "set ioam export ipfix",.short_help = + "set ioam export ipfix collector src ",. + function = set_ioam_export_ipfix_command_fn,}; +/* *INDENT-ON* */ + + +static clib_error_t * +ioam_export_init (vlib_main_t * vm) +{ + ioam_export_main_t *em = &ioam_export_main; + clib_error_t *error = 0; + u8 *name; + u32 node_index = export_node.index; + vlib_node_t *ip6_hbyh_node = NULL; + + name = format (0, "ioam_export_%08x%c", api_version, 0); + + /* Ask for a correctly-sized block of API message decode slots */ + em->msg_id_base = vl_msg_api_get_msg_ids + ((char *) name, VL_MSG_FIRST_AVAILABLE); + em->unix_time_0 = (u32) time (0); /* Store starting time */ + em->vlib_time_0 = vlib_time_now (vm); + + error = ioam_export_plugin_api_hookup (vm); + + /* Add our API messages to the global name_crc hash table */ + setup_message_id_table (em, &api_main); + + /* Hook this export node to ip6-hop-by-hop */ + ip6_hbyh_node = vlib_get_node_by_name (vm, (u8 *) "ip6-hop-by-hop"); + em->my_hbh_slot = vlib_node_add_next (vm, ip6_hbyh_node->index, node_index); + vec_free (name); + + return error; +} + +VLIB_INIT_FUNCTION (ioam_export_init); + +/* + * fd.io coding-style-patch-verification: ON + * + * Local Variables: + * eval: (c-set-style "gnu") + * End: + */ diff --git a/src/plugins/ioam/export/ioam_export_all_api_h.h b/src/plugins/ioam/export/ioam_export_all_api_h.h new file mode 100644 index 00000000..bc4368f2 --- /dev/null +++ b/src/plugins/ioam/export/ioam_export_all_api_h.h @@ -0,0 +1,16 @@ +/* + * Copyright (c) 2016 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ +/* Include the generated file, see BUILT_SOURCES in Makefile.am */ +#include diff --git a/src/plugins/ioam/export/ioam_export_msg_enum.h b/src/plugins/ioam/export/ioam_export_msg_enum.h new file mode 100644 index 00000000..c2de7988 --- /dev/null +++ b/src/plugins/ioam/export/ioam_export_msg_enum.h @@ -0,0 +1,28 @@ +/* + * Copyright (c) 2016 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ +#ifndef included_ioam_export_msg_enum_h +#define included_ioam_export_msg_enum_h + +#include + +#define vl_msg_id(n,h) n, +typedef enum { +#include + /* We'll want to know how many messages IDs we need... */ + VL_MSG_FIRST_AVAILABLE, +} vl_msg_id_t; +#undef vl_msg_id + +#endif /* included_ioam_export_msg_enum_h */ diff --git a/src/plugins/ioam/export/ioam_export_test.c b/src/plugins/ioam/export/ioam_export_test.c new file mode 100644 index 00000000..f991fc0c --- /dev/null +++ b/src/plugins/ioam/export/ioam_export_test.c @@ -0,0 +1,206 @@ +/* + * Copyright (c) 2016 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ +/* + *------------------------------------------------------------------ + * ioam_export_test.c - test harness plugin + *------------------------------------------------------------------ + */ + +#include +#include +#include +#include +#include + + +/* Declare message IDs */ +#include + +/* define message structures */ +#define vl_typedefs +#include +#undef vl_typedefs + +/* declare message handlers for each api */ + +#define vl_endianfun /* define message structures */ +#include +#undef vl_endianfun + +/* instantiate all the print functions we know about */ +#define vl_print(handle, ...) +#define vl_printfun +#include +#undef vl_printfun + +/* Get the API version number. */ +#define vl_api_version(n,v) static u32 api_version=(v); +#include +#undef vl_api_version + + +typedef struct +{ + /* API message ID base */ + u16 msg_id_base; + vat_main_t *vat_main; +} export_test_main_t; + +export_test_main_t export_test_main; + +#define foreach_standard_reply_retval_handler \ +_(ioam_export_ip6_enable_disable_reply) + +#define _(n) \ + static void vl_api_##n##_t_handler \ + (vl_api_##n##_t * mp) \ + { \ + vat_main_t * vam = export_test_main.vat_main; \ + i32 retval = ntohl(mp->retval); \ + if (vam->async_mode) { \ + vam->async_errors += (retval < 0); \ + } else { \ + vam->retval = retval; \ + vam->result_ready = 1; \ + } \ + } +foreach_standard_reply_retval_handler; +#undef _ + +/* + * Table of message reply handlers, must include boilerplate handlers + * we just generated + */ +#define foreach_vpe_api_reply_msg \ +_(IOAM_EXPORT_IP6_ENABLE_DISABLE_REPLY, ioam_export_ip6_enable_disable_reply) + + +/* M: construct, but don't yet send a message */ + +#define M(T,t) \ +do { \ + vam->result_ready = 0; \ + mp = vl_msg_api_alloc(sizeof(*mp)); \ + memset (mp, 0, sizeof (*mp)); \ + mp->_vl_msg_id = ntohs (VL_API_##T + sm->msg_id_base); \ + mp->client_index = vam->my_client_index; \ +} while(0); + +#define M2(T,t,n) \ +do { \ + vam->result_ready = 0; \ + mp = vl_msg_api_alloc(sizeof(*mp)+(n)); \ + memset (mp, 0, sizeof (*mp)); \ + mp->_vl_msg_id = ntohs (VL_API_##T + sm->msg_id_base); \ + mp->client_index = vam->my_client_index; \ +} while(0); + +/* S: send a message */ +#define S (vl_msg_api_send_shmem (vam->vl_input_queue, (u8 *)&mp)) + +/* W: wait for results, with timeout */ +#define W \ +do { \ + timeout = vat_time_now (vam) + 1.0; \ + \ + while (vat_time_now (vam) < timeout) { \ + if (vam->result_ready == 1) { \ + return (vam->retval); \ + } \ + } \ + return -99; \ +} while(0); + +static int +api_ioam_export_ip6_enable_disable (vat_main_t * vam) +{ + export_test_main_t *sm = &export_test_main; + unformat_input_t *i = vam->input; + f64 timeout; + int is_disable = 0; + vl_api_ioam_export_ip6_enable_disable_t *mp; + + /* Parse args required to build the message */ + while (unformat_check_input (i) != UNFORMAT_END_OF_INPUT) + { + if (unformat (i, "disable")) + is_disable = 1; + else + break; + } + + /* Construct the API message */ + M (IOAM_EXPORT_IP6_ENABLE_DISABLE, ioam_export_ip6_enable_disable); + mp->is_disable = is_disable; + + /* send it... */ + S; + + /* Wait for a reply... */ + W; +} + +/* + * List of messages that the api test plugin sends, + * and that the data plane plugin processes + */ +#define foreach_vpe_api_msg \ +_(ioam_export_ip6_enable_disable, " [disable]") + +void +vat_api_hookup (vat_main_t * vam) +{ + export_test_main_t *sm = &export_test_main; + /* Hook up handlers for replies from the data plane plug-in */ +#define _(N,n) \ + vl_msg_api_set_handlers((VL_API_##N + sm->msg_id_base), \ + #n, \ + vl_api_##n##_t_handler, \ + vl_noop_handler, \ + vl_api_##n##_t_endian, \ + vl_api_##n##_t_print, \ + sizeof(vl_api_##n##_t), 1); + foreach_vpe_api_reply_msg; +#undef _ + + /* API messages we can send */ +#define _(n,h) hash_set_mem (vam->function_by_name, #n, api_##n); + foreach_vpe_api_msg; +#undef _ + + /* Help strings */ +#define _(n,h) hash_set_mem (vam->help_by_name, #n, h); + foreach_vpe_api_msg; +#undef _ +} + +clib_error_t * +vat_plugin_register (vat_main_t * vam) +{ + export_test_main_t *sm = &export_test_main; + u8 *name; + + sm->vat_main = vam; + + name = format (0, "ioam_export_%08x%c", api_version, 0); + sm->msg_id_base = vl_client_get_first_plugin_msg_id ((char *) name); + + if (sm->msg_id_base != (u16) ~ 0) + vat_api_hookup (vam); + + vec_free (name); + + return 0; +} diff --git a/src/plugins/ioam/export/ioam_export_thread.c b/src/plugins/ioam/export/ioam_export_thread.c new file mode 100644 index 00000000..d2eb2009 --- /dev/null +++ b/src/plugins/ioam/export/ioam_export_thread.c @@ -0,0 +1,38 @@ +/* + * Copyright (c) 2016 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ +/* + * ioam_export_thread.c + */ +#include +#include +#include + +static vlib_node_registration_t ioam_export_process_node; + +static uword +ioam_export_process (vlib_main_t * vm, + vlib_node_runtime_t * rt, vlib_frame_t * f) +{ + return (ioam_export_process_common(&ioam_export_main, + vm, rt, f, + ioam_export_process_node.index)); +} + +VLIB_REGISTER_NODE (ioam_export_process_node, static) = +{ + .function = ioam_export_process, + .type = VLIB_NODE_TYPE_PROCESS, + .name = "ioam-export-process", +}; diff --git a/src/plugins/ioam/export/node.c b/src/plugins/ioam/export/node.c new file mode 100644 index 00000000..19f143df --- /dev/null +++ b/src/plugins/ioam/export/node.c @@ -0,0 +1,151 @@ +/* + * Copyright (c) 2016 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ +#include +#include +#include +#include +#include +#include + +typedef struct +{ + u32 next_index; + u32 flow_label; +} export_trace_t; + +/* packet trace format function */ +static u8 * +format_export_trace (u8 * s, va_list * args) +{ + CLIB_UNUSED (vlib_main_t * vm) = va_arg (*args, vlib_main_t *); + CLIB_UNUSED (vlib_node_t * node) = va_arg (*args, vlib_node_t *); + export_trace_t *t = va_arg (*args, export_trace_t *); + + s = format (s, "EXPORT: flow_label %d, next index %d", + t->flow_label, t->next_index); + return s; +} + +vlib_node_registration_t export_node; + +#define foreach_export_error \ +_(RECORDED, "Packets recorded for export") + +typedef enum +{ +#define _(sym,str) EXPORT_ERROR_##sym, + foreach_export_error +#undef _ + EXPORT_N_ERROR, +} export_error_t; + +static char *export_error_strings[] = { +#define _(sym,string) string, + foreach_export_error +#undef _ +}; + +typedef enum +{ + EXPORT_NEXT_POP_HBYH, + EXPORT_N_NEXT, +} export_next_t; + +always_inline void +copy3cachelines (void *dst, const void *src, size_t n) +{ +#if 0 + if (PREDICT_FALSE (n < DEFAULT_EXPORT_SIZE)) + { + /* Copy only the first 1/2 cache lines whatever is available */ + if (n >= 64) + clib_mov64 ((u8 *) dst, (const u8 *) src); + if (n >= 128) + clib_mov64 ((u8 *) dst + 64, (const u8 *) src + 64); + return; + } + clib_mov64 ((u8 *) dst, (const u8 *) src); + clib_mov64 ((u8 *) dst + 64, (const u8 *) src + 64); + clib_mov64 ((u8 *) dst + 128, (const u8 *) src + 128); +#endif +#if 1 + + u64 *copy_dst, *copy_src; + int i; + copy_dst = (u64 *) dst; + copy_src = (u64 *) src; + if (PREDICT_FALSE (n < DEFAULT_EXPORT_SIZE)) + { + for (i = 0; i < n / 64; i++) + { + copy_dst[0] = copy_src[0]; + copy_dst[1] = copy_src[1]; + copy_dst[2] = copy_src[2]; + copy_dst[3] = copy_src[3]; + copy_dst[4] = copy_src[4]; + copy_dst[5] = copy_src[5]; + copy_dst[6] = copy_src[6]; + copy_dst[7] = copy_src[7]; + copy_dst += 8; + copy_src += 8; + } + return; + } + for (i = 0; i < 3; i++) + { + copy_dst[0] = copy_src[0]; + copy_dst[1] = copy_src[1]; + copy_dst[2] = copy_src[2]; + copy_dst[3] = copy_src[3]; + copy_dst[4] = copy_src[4]; + copy_dst[5] = copy_src[5]; + copy_dst[6] = copy_src[6]; + copy_dst[7] = copy_src[7]; + copy_dst += 8; + copy_src += 8; + } +#endif +} + +static uword +ip6_export_node_fn (vlib_main_t * vm, + vlib_node_runtime_t * node, vlib_frame_t * frame) +{ + ioam_export_main_t *em = &ioam_export_main; + ioam_export_node_common(em, vm, node, frame, ip6_header_t, payload_length, + ip_version_traffic_class_and_flow_label, + EXPORT_NEXT_POP_HBYH); + return frame->n_vectors; +} + +/* + * Node for IP6 export + */ +VLIB_REGISTER_NODE (export_node) = +{ + .function = ip6_export_node_fn, + .name = "ip6-export", + .vector_size = sizeof (u32), + .format_trace = format_export_trace, + .type = VLIB_NODE_TYPE_INTERNAL, + .n_errors = ARRAY_LEN (export_error_strings), + .error_strings = export_error_strings, + .n_next_nodes = EXPORT_N_NEXT, + /* edit / add dispositions here */ + .next_nodes = + { + [EXPORT_NEXT_POP_HBYH] = "ip6-pop-hop-by-hop" + }, +}; diff --git a/src/plugins/ioam/ioam_plugin_doc.md b/src/plugins/ioam/ioam_plugin_doc.md new file mode 100644 index 00000000..343abcf7 --- /dev/null +++ b/src/plugins/ioam/ioam_plugin_doc.md @@ -0,0 +1,464 @@ +## VPP Inband OAM (iOAM) {#ioam_plugin_doc} + +In-band OAM (iOAM) is an implementation study to record operational +information in the packet while the packet traverses a path between +two points in the network. + +Overview of iOAM can be found in [iOAM-Devnet] page. +The following IETF drafts detail the motivation and mechanism for +recording operational information: + - [iOAM-ietf-requirements] - Describes motivation and usecases for iOAM + - [iOAM-ietf-data] - Describes data records that can be collected using iOAM + - [iOAM-ietf-transport] - Lists out the transport protocols + and mechanism to carry iOAM data records + - [iOAM-ietf-proof-of-transit] - Describes the idea of Proof of Transit (POT) + and mechanisms to operationalize the idea + +## Terminology +In-band OAM is expected to be deployed in a specific domain rather +than on the overall Internet. The part of the network which employs in-band OAM +is referred to as **"in-band OAM-domain"**. + +In-band OAM data is added to a packet on entering the in-band OAM-domain +and is removed from the packet when exiting the domain. +Within the in-band OAM-domain, network nodes that the packet traverses +may update the in-band OAM data records. + +- The node which adds in-band OAM data to the packet is called the +**"in-band OAM encapsulating node"**. + +- The node which removes the in-band OAM data is referred to as the +**"in-band OAM decapsulating node"**. + +- Nodes within the domain which are aware of in-band OAM data and read +and/or write or process the in-band OAM data are called +**"in-band OAM transit nodes"**. + +## Features supported in the current release +VPP can function as in-band OAM encapsulating, transit and decapsulating node. +In this version of VPP in-band OAM data is transported as options in an +IPv6 hop-by-hop extension header. Hence in-band OAM can be enabled +for IPv6 traffic. + +The following iOAM features are supported: + +- **In-band OAM Tracing** : In-band OAM supports multiple data records to be +recorded in the packet as the packet traverses the network. +These data records offer insights into the operational behavior of the network. +The following information can be collected in the tracing +data from the nodes a packet traverses: + - Node ID + - Ingress interface ID + - Egress interface ID + - Timestamp + - Pre-configured application data + +- **In-band OAM Proof of Transit (POT)**: Proof of transit iOAM data is +added to every packet for verifying that a packet traverses a specific +set of nodes. +In-band OAM data is updated at every node that is enabled with iOAM +proof of transit and is used to verify whether a packet traversed +all the specified nodes. When the verifier receives each packet, +it can validate whether the packet traversed the specified nodes. + + +## Configuration +Configuring iOAM involves: +- Selecting the packets for which iOAM data must be inserted, updated or removed + - Selection of packets for iOAM data insertion on iOAM encapsulating node. + Selection of packets is done by 5-tuple based classification + - Selection of packets for updating iOAM data is implicitly done on the + presence of iOAM options in the packet + - Selection of packets for removing the iOAM data is done on 5-tuple + based classification +- The kind of data to be collected + - Tracing data + - Proof of transit +- Additional details for processing iOAM data to be collected + - For trace data - trace type, number of nodes to be recorded in the trace, + time stamp precision, etc. + - For POT data - configuration of POT profile required to process the POT data + +The CLI for configuring iOAM is explained here followed by detailed steps +and examples to deploy iOAM on VPP as an encapsulating, transit or +decapsulating iOAM node in the subsequent sub-sections. + +VPP iOAM configuration for enabling trace and POT is as follows: + + set ioam rewrite trace-type <0x1f|0x7|0x9|0x11|0x19> + trace-elts trace-tsp <0|1|2|3> + node-id app-data [pot] + +A description of each of the options of the CLI follows: +- trace-type : An entry in the "Node data List" array of the trace option +can have different formats, following the needs of the a deployment. +For example: Some deployments might only be interested +in recording the node identifiers, whereas others might be interested +in recording node identifier and timestamp. +The following types are currently supported: + - 0x1f : Node data to include hop limit (8 bits), node ID (24 bits), + ingress and egress interface IDs (16 bits each), timestamp (32 bits), + application data (32 bits) + - 0x7 : Node data to include hop limit (8 bits), node ID (24 bits), + ingress and egress interface IDs (16 bits each) + - 0x9 : Node data to include hop limit (8 bits), node ID (24 bits), + timestamp (32 bits) + - 0x11: Node data to include hop limit (8 bits), node ID (24 bits), + application data (32 bits) + - 0x19: Node data to include hop limit (8 bits), node ID (24 bits), + timestamp (32 bits), application data (32 bits) +- trace-elts : Defines the length of the node data array in the trace option. +- trace-tsp : Defines the timestamp precision to use with the enumerated value + for precision as follows: + - 0 : 32bits timestamp in seconds + - 1 : 32bits timestamp in milliseconds + - 2 : 32bits timestamp in microseconds + - 3 : 32bits timestamp in nanoseconds +- node-id : Unique identifier for the node, included in the node ID + field of the node data in trace option. +- app-data : The value configured here is included as is in +application data field of node data in trace option. +- pot : Enables POT option to be included in the iOAM options. + +### Trace configuration + +#### On in-band OAM encapsulating node + - **Configure classifier and apply ACL** to select packets for + iOAM data insertion + - Example to enable iOAM data insertion for all the packets + towards IPv6 address db06::06: + + vpp# classify table miss-next node ip6-lookup mask l3 ip6 dst + + vpp# classify session acl-hit-next node ip6-add-hop-by-hop + table-index 0 match l3 ip6 dst db06::06 + + vpp# set int input acl intfc GigabitEthernet0/0/0 ip6-table 0 + + - **Enable tracing** : Specify node ID, maximum number of nodes for which + trace data should be recorded, type of data to be included for recording, + optionally application data to be included + - Example to enable tracing with a maximum of 4 nodes recorded + and the data to be recorded to include - hop limit, node id, + ingress and egress interface IDs, timestamp (millisecond precision), + application data (0x1234): + + + vpp# set ioam rewrite trace-type 0x1f trace-elts 4 trace-tsp 1 + node-id 0x1 app-data 0x1234 + + + +#### On in-band OAM transit node +- The transit node requires trace type, timestamp precision, node ID and +optionally application data to be configured, +to update its node data in the trace option. + +Example: + + vpp# set ioam rewrite trace-type 0x1f trace-elts 4 trace-tsp 1 + node-id 0x2 app-data 0x1234 + +#### On the In-band OAM decapsulating node +- The decapsulating node similar to encapsulating node requires +**classification** of the packets to remove iOAM data from. + - Example to decapsulate iOAM data for packets towards + db06::06, configure classifier and enable it as an ACL as follows: + + + vpp# classify table miss-next node ip6-lookup mask l3 ip6 dst + + vpp# classify session acl-hit-next node ip6-lookup table-index 0 + match l3 ip6 dst db06::06 opaque-index 100 + + vpp# set int input acl intfc GigabitEthernet0/0/0 ip6-table 0 + + +- Decapsulating node requires trace type, timestamp precision, +node ID and optionally application data to be configured, +to update its node data in the trace option before it is decapsulated. + +Example: + + vpp# set ioam rewrite trace-type 0x1f trace-elts 4 + trace-tsp 1 node-id 0x3 app-data 0x1234 + + +### Proof of Transit configuration + +For details on proof-of-transit, +see the IETF draft [iOAM-ietf-proof-of-transit]. +To enable Proof of Transit all the nodes that participate +and hence are verified for transit need a proof of transit profile. +A script to generate a proof of transit profile as per the mechanism +described in [iOAM-ietf-proof-of-transit] will be available at [iOAM-Devnet]. + +The Proof of transit mechanism implemented here is based on +Shamir's Secret Sharing algorithm. +The overall algorithm uses two polynomials +POLY-1 and POLY-2. The degree of polynomials depends on number of nodes +to be verified for transit. +POLY-1 is secret and constant. Each node gets a point on POLY-1 +at setup-time and keeps it secret. +POLY-2 is public, random and per packet. +Each node is assigned a point on POLY-1 and POLY-2 with the same x index. +Each node derives its point on POLY-2 each time a packet arrives at it. +A node then contributes its points on POLY-1 and POLY-2 to construct +POLY-3 (POLY-3 = POLY-1 + POLY-2) using lagrange extrapolation and +forwards it towards the verifier by updating POT data in the packet. +The verifier constructs POLY-3 from the accumulated value from all the nodes +and its own points on POLY-1 and POLY-2 and verifies whether +POLY-3 = POLY-1 + POLY-2. Only the verifier knows POLY-1. +The solution leverages finite field arithmetic in a field of size "prime number" +for reasons explained in description of Shamir's secret sharing algorithm. + +Here is an explanation of POT profile list and profile configuration CLI to +realize the above mechanism. +It is best to use the script provided at [iOAM-Devnet] to generate +this configuration. +- **Create POT profile** : set pot profile name id [0-1] +[validator-key 0xu64] prime-number 0xu64 secret_share 0xu64 +lpc 0xu64 polynomial2 0xu64 bits-in-random [0-64] + - name : Profile list name. + - id : Profile id, it can be 0 or 1. + A maximum of two profiles can be configured per profile list. + - validator-key : Secret key configured only on the + verifier/decapsulating node used to compare and verify proof of transit. + - prime-number : Prime number for finite field arithmetic as required by the + proof of transit mechanism. + - secret_share : Unique point for each node on the secret polynomial POLY-1. + - lpc : Lagrange Polynomial Constant(LPC) calculated per node based on + its point (x value used for evaluating the points on the polynomial) + on the polynomial used in lagrange extrapolation + for reconstructing polynomial (POLY-3). + - polynomial2 : Is the pre-evaluated value of the point on + 2nd polynomial(POLY-2). This is unique for each node. + It is pre-evaluated for all the coefficients of POLY-2 except + for the constant part of the polynomial that changes per packet + and is received as part of the POT data in the packet. + - bits-in-random : To control the size of the random number to be + generated. This number has to match the other numbers generated and used + in the profile as per the algorithm. + +- **Set a configured profile as active/in-use** : +set pot profile-active name ID [0-1] + - name : Name of the profile list to be used for computing + POT data per packet. + - ID : Identifier of the profile within the list to be used. + +#### On In-band OAM encapsulating node + - Configure the classifier and apply ACL to select packets for iOAM data insertion. + - Example to enable iOAM data insertion for all the packet towards + IPv6 address db06::06 - + + + vpp# classify table miss-next node ip6-lookup mask l3 ip6 dst + + vpp# classify session acl-hit-next node + ip6-add-hop-by-hop table-index 0 match l3 ip6 dst db06::06 + + vpp# set int input acl intfc GigabitEthernet0/0/0 ip6-table 0 + + + - Configure the proof of transit profile list with profiles. +Each profile list referred to by a name can contain 2 profiles, +only one is in use for updating proof of transit data at any time. + - Example profile list example with a profile generated from the + script to verify transit through 3 nodes is: + + + vpp# set pot profile name example id 0 prime-number 0x7fff0000fa884685 + secret_share 0x6c22eff0f45ec56d lpc 0x7fff0000fa884682 + polynomial2 0xffb543d4a9c bits-in-random 63 + + - Enable one of the profiles from the configured profile list as active + so that is will be used for calculating proof of transit + +Example enable profile ID 0 from profile list example configured above: + + + vpp# set pot profile-active name example ID 0 + + + - Enable POT option to be inserted + + + vpp# set ioam rewrite pot + + +#### On in-band OAM transit node + - Configure the proof of transit profile list with profiles for transit node. +Example: + + + vpp# set pot profile name example id 0 prime-number 0x7fff0000fa884685 + secret_share 0x564cdbdec4eb625d lpc 0x1 + polynomial2 0x23f3a227186a bits-in-random 63 + +#### On in-band OAM decapsulating node / verifier +- The decapsulating node, similar to the encapsulating node requires +classification of the packets to remove iOAM data from. + - Example to decapsulate iOAM data for packets towards db06::06 + configure classifier and enable it as an ACL as follows: + + + vpp# classify table miss-next node ip6-lookup mask l3 ip6 dst + + vpp# classify session acl-hit-next node ip6-lookup table-index 0 + match l3 ip6 dst db06::06 opaque-index 100 + + vpp# set int input acl intfc GigabitEthernet0/0/0 ip6-table 0 + +- To update and verify the proof of transit, POT profile list should be configured. + - Example POT profile list configured as follows: + + vpp# set pot profile name example id 0 validate-key 0x7fff0000fa88465d + prime-number 0x7fff0000fa884685 secret_share 0x7a08fbfc5b93116d lpc 0x3 + polynomial2 0x3ff738597ce bits-in-random 63 + +## Operational data + +Following CLIs are available to check iOAM operation: +- To check iOAM configuration that are effective use "show ioam summary" + +Example: + + vpp# show ioam summary + REWRITE FLOW CONFIGS - Not configured + HOP BY HOP OPTIONS - TRACE CONFIG - + Trace Type : 0x1f (31) + Trace timestamp precision : 1 (Milliseconds) + Num of trace nodes : 4 + Node-id : 0x2 (2) + App Data : 0x1234 (4660) + POT OPTION - 1 (Enabled) + Try 'show ioam pot and show pot profile' for more information + +- To find statistics about packets for which iOAM options were +added (encapsulating node) and removed (decapsulating node) execute +*show errors* + +Example on encapsulating node: + + + vpp# show error + Count Node Reason + 1208804706 ip6-inacl input ACL hits + 1208804706 ip6-add-hop-by-hop Pkts w/ added ip6 hop-by-hop options + +Example on decapsulating node: + + vpp# show error + Count Node Reason + 69508569 ip6-inacl input ACL hits + 69508569 ip6-pop-hop-by-hop Pkts w/ removed ip6 hop-by-hop options + +- To check the POT profiles use "show pot profile" + +Example: + + vpp# show pot profile + Profile list in use : example + POT Profile at index: 0 + ID : 0 + Validator : False (0) + Secret share : 0x564cdbdec4eb625d (6218586935324795485) + Prime number : 0x7fff0000fa884685 (9223090566081300101) + 2nd polynomial(eval) : 0x23f3a227186a (39529304496234) + LPC : 0x1 (1) + Bit mask : 0x7fffffffffffffff (9223372036854775807) + Profile index in use: 0 + Pkts passed : 0x36 (54) + +- To get statistics of POT for packets use "show ioam pot" + +Example at encapsulating or transit node: + + vpp# show ioam pot + Pkts with ip6 hop-by-hop POT options - 54 + Pkts with ip6 hop-by-hop POT options but no profile set - 0 + Pkts with POT in Policy - 0 + Pkts with POT out of Policy - 0 + + +Example at decapsulating/verification node: + + + vpp# show ioam pot + Pkts with ip6 hop-by-hop POT options - 54 + Pkts with ip6 hop-by-hop POT options but no profile set - 0 + Pkts with POT in Policy - 54 + Pkts with POT out of Policy - 0 + +- Tracing - enable trace of IPv6 packets to view the data inserted and +collected. + +Example when the nodes are receiving data over a DPDK interface: +Enable tracing using "trace add dpdk-input 20" and +execute "show trace" to view the iOAM data collected: + + + vpp# trace add dpdk-input 20 + + vpp# show trace + + ------------------- Start of thread 0 vpp_main ------------------- + + Packet 1 + + 00:00:19:294697: dpdk-input + GigabitEthernetb/0/0 rx queue 0 + buffer 0x10e6b: current data 0, length 214, free-list 0, totlen-nifb 0, trace 0x0 + PKT MBUF: port 0, nb_segs 1, pkt_len 214 + buf_len 2176, data_len 214, ol_flags 0x0, data_off 128, phys_addr 0xe9a35a00 + packet_type 0x0 + IP6: 00:50:56:9c:df:72 -> 00:50:56:9c:be:55 + IP6_HOP_BY_HOP_OPTIONS: db05::2 -> db06::6 + tos 0x00, flow label 0x0, hop limit 63, payload length 160 + 00:00:19:294737: ethernet-input + IP6: 00:50:56:9c:df:72 -> 00:50:56:9c:be:55 + 00:00:19:294753: ip6-input + IP6_HOP_BY_HOP_OPTIONS: db05::2 -> db06::6 + tos 0x00, flow label 0x0, hop limit 63, payload length 160 + 00:00:19:294757: ip6-lookup + fib 0 adj-idx 15 : indirect via db05::2 flow hash: 0x00000000 + IP6_HOP_BY_HOP_OPTIONS: db05::2 -> db06::6 + tos 0x00, flow label 0x0, hop limit 63, payload length 160 + 00:00:19:294802: ip6-hop-by-hop + IP6_HOP_BY_HOP: next index 5 len 96 traced 96 Trace Type 0x1f , 1 elts left + [0] ttl 0x0 node ID 0x0 ingress 0x0 egress 0x0 ts 0x0 + app 0x0 + [1] ttl 0x3e node ID 0x3 ingress 0x1 egress 0x2 ts 0xb68c2213 + app 0x1234 + [2] ttl 0x3f node ID 0x2 ingress 0x1 egress 0x2 ts 0xb68c2204 + app 0x1234 + [3] ttl 0x40 node ID 0x1 ingress 0x5 egress 0x6 ts 0xb68c2200 + app 0x1234 + POT opt present + random = 0x577a916946071950, Cumulative = 0x10b46e78a35a392d, Index = 0x0 + 00:00:19:294810: ip6-rewrite + tx_sw_if_index 1 adj-idx 14 : GigabitEthernetb/0/0 + IP6: 00:50:56:9c:be:55 -> 00:50:56:9c:df:72 flow hash: 0x00000000 + IP6: 00:50:56:9c:be:55 -> 00:50:56:9c:df:72 + IP6_HOP_BY_HOP_OPTIONS: db05::2 -> db06::6 + tos 0x00, flow label 0x0, hop limit 62, payload length 160 + 00:00:19:294814: GigabitEthernetb/0/0-output + GigabitEthernetb/0/0 + IP6: 00:50:56:9c:be:55 -> 00:50:56:9c:df:72 + IP6_HOP_BY_HOP_OPTIONS: db05::2 -> db06::6 + tos 0x00, flow label 0x0, hop limit 62, payload length 160 + 00:00:19:294820: GigabitEthernetb/0/0-tx + GigabitEthernetb/0/0 tx queue 0 + buffer 0x10e6b: current data 0, length 214, free-list 0, totlen-nifb 0, trace 0x0 + IP6: 00:50:56:9c:be:55 -> 00:50:56:9c:df:72 + + IP6_HOP_BY_HOP_OPTIONS: db05::2 -> db06::6 + + tos 0x00, flow label 0x0, hop limit 62, payload length 160 + + +[iOAM-Devnet]: +[iOAM-ietf-requirements]: +[iOAM-ietf-transport]: +[iOAM-ietf-data]: +[iOAM-ietf-proof-of-transit]: diff --git a/src/plugins/ioam/lib-pot/math64.h b/src/plugins/ioam/lib-pot/math64.h new file mode 100644 index 00000000..4c608a37 --- /dev/null +++ b/src/plugins/ioam/lib-pot/math64.h @@ -0,0 +1,159 @@ +/* + * math64.h provides the 64 bit unsigned integer add, multiply followed by modulo operation + * The linux/math64.h provides divide and multiply 64 bit integers but: + * 1. multiply: mul_u64_u64_shr - only returns 64 bits of the result and has to be called + * twice to get the complete 128 bits of the result. + * 2. Modulo operation of the result of addition and multiplication of u64 that may result + * in integers > 64 bits is not supported + * Hence this header to combine add/multiply followed by modulo of u64 integrers + * always resulting in u64. + * + * Copyright (c) 2016 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ +#ifndef include_vnet_math64_h +#define include_vnet_math64_h +#include + +/* + * multiplies and returns result in hi and lo + */ +static inline void mul64by64(u64 a, u64 b, u64 * hi, u64 * lo) +{ + u64 a_lo = (u64) (uint32_t) a; + u64 a_hi = a >> 32; + u64 b_lo = (u64) (u32) b; + u64 b_hi = b >> 32; + + u64 p0 = a_lo * b_lo; + u64 p1 = a_lo * b_hi; + u64 p2 = a_hi * b_lo; + u64 p3 = a_hi * b_hi; + + u32 cy = (u32) (((p0 >> 32) + (u32) p1 + (u32) p2) >> 32); + + *lo = p0 + (p1 << 32) + (p2 << 32); + *hi = p3 + (p1 >> 32) + (p2 >> 32) + cy; + return; +} + +#define TWO64 18446744073709551616.0 + +static inline u64 mod128by64(u64 x, u64 y, u64 m, double di) +{ + u64 q1, q2, q; + u64 p1, p0; + double dq; + + /* calculate quotient first pass 53 bits */ + dq = (TWO64 * (double)x + (double)y) * di; + + if (dq >= TWO64) + q1 = 0xfffffffffffff800L; + else + q1 = dq; + + /* q1 * m to compare the product to the dividend. */ + mul64by64(q1, m, &p1, &p0); + + /* Adjust quotient. is it > actual result: */ + if (x < p1 || (x == p1 && y < p0)) + { + /* q1 > quotient. calculate abs remainder */ + x = p1 - (x + (p0 < y)); + y = p0 - y; + + /* use the remainder as new dividend to adjust quotient */ + q2 = (u64) ((TWO64 * (double)x + (double)y) * di); + mul64by64(q2, m, &p1, &p0); + + q = q1 - q2; + if (x < p1 || (x == p1 && y <= p0)) + { + y = p0 - y; + } + else + { + y = p0 - y; + y += m; + q--; + } + } + else + { + x = x - (p1 + (y < p0)); + y = y - p0; + + q2 = (u64) ((TWO64 * (double)x + (double)y) * di); + mul64by64(q2, m, &p1, &p0); + + q = q1 + q2; + if (x < p1 || (x == p1 && y < p0)) + { + y = y - p0; + y += m; + q--; + } + else + { + y = y - p0; + if (y >= m) + { + y -= m; + q++; + } + } + } + + return y; +} + +/* + * returns a % p + */ +static inline u64 mod64by64(u64 a, u64 p, u64 primeinv) +{ + return (mod128by64(0, a, p, primeinv)); +} + +static inline void add64(u64 a, u64 b, u64 * whi, u64 * wlo) +{ + *wlo = a + b; + if (*wlo < a) + *whi = 1; + +} + +/* + * returns (a + b)%p + */ +static inline u64 add64_mod(u64 a, u64 b, u64 p, double pi) +{ + u64 shi = 0, slo = 0; + + add64(a, b, &shi, &slo); + return (mod128by64(shi, slo, p, pi)); +} + +/* + * returns (ab) % p + */ +static inline u64 mul64_mod(u64 a, u64 b, u64 p, double pi) +{ + u64 phi = 0, plo = 0; + + mul64by64(a, b, &phi, &plo); + return (mod128by64(phi, plo, p, pi)); +} + +#endif diff --git a/src/plugins/ioam/lib-pot/pot.api b/src/plugins/ioam/lib-pot/pot.api new file mode 100644 index 00000000..fa2fc126 --- /dev/null +++ b/src/plugins/ioam/lib-pot/pot.api @@ -0,0 +1,133 @@ +/* Hey Emacs use -*- mode: C -*- */ +/* + * Copyright (c) 2016 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + + +/** \brief Proof of Transit(POT): Set POT profile + @param id - id of the profile + @param validator - True/False to indicate if this is verifier + @param secret_key - Verification key + @param secret_share - Share of the 1st polynomial + @param prime - Prime number used for modulo operation + @param max_bits - Max bits to be used for Random number generation + @param lpc - Lagrange basis polynomial + @param polynomial_public - pre-evaluated public polynomial + @param list_name_len - length of the name of this profile list + @param list_name - name of this profile list +*/ +define pot_profile_add { + u32 client_index; + u32 context; + u8 id; + u8 validator; + u64 secret_key; + u64 secret_share; + u64 prime; + u8 max_bits; + u64 lpc; + u64 polynomial_public; + u8 list_name_len; + u8 list_name[0]; +}; + +/** \brief Proof of Transit profile add / del response + @param context - sender context, to match reply w/ request + @param retval - return value for request +*/ +define pot_profile_add_reply { + u32 context; + i32 retval; +}; + + +/** \brief Proof of Transit(POT): Activate POT profile in the list + @param id - id of the profile + @param list_name_len - length of the name of this profile list + @param list_name - name of this profile list +*/ +define pot_profile_activate { + u32 client_index; + u32 context; + u8 id; + u8 list_name_len; + u8 list_name[0]; +}; + +/** \brief Proof of Transit profile activate response + @param context - sender context, to match reply w/ request + @param retval - return value for request +*/ +define pot_profile_activate_reply { + u32 context; + i32 retval; +}; + +/** \brief Delete POT Profile + @param client_index - opaque cookie to identify the sender + @param context - sender context, to match reply w/ request + @param list_name_len - length of the name of the profile list + @param list_name - name of profile list to delete +*/ +define pot_profile_del { + u32 client_index; + u32 context; + u8 list_name_len; + u8 list_name[0]; +}; + +/** \brief Proof of Transit profile add / del response + @param context - sender context, to match reply w/ request + @param retval - return value for request +*/ +define pot_profile_del_reply { + u32 context; + i32 retval; +}; + +/** \brief Show POT Profiles + @param client_index - opaque cookie to identify the sender + @param context - sender context, to match reply w/ request + @param id - id of the profile +*/ +define pot_profile_show_config_dump { + u32 client_index; + u32 context; + u8 id; +}; + +/** \brief Show POT profile reply + @param id - id of the profile + @param validator - True/False to indicate if this is verifier + @param secret_key - Verification key + @param secret_share - Share of the 1st polynomial + @param prime - Prime number used for modulo operation + @param max_bits - Max bits to be used for Random number generation + @param lpc - Lagrange basis polynomial + @param polynomial_public - pre-evaluated public polynomial + @param list_name_len - length of the name of this profile list + @param list_name - name of this profile list +*/ +define pot_profile_show_config_details { + u32 context; + i32 retval; + u8 id; + u8 validator; + u64 secret_key; + u64 secret_share; + u64 prime; + u64 bit_mask; + u64 lpc; + u64 polynomial_public; +}; diff --git a/src/plugins/ioam/lib-pot/pot_all_api_h.h b/src/plugins/ioam/lib-pot/pot_all_api_h.h new file mode 100644 index 00000000..63967c45 --- /dev/null +++ b/src/plugins/ioam/lib-pot/pot_all_api_h.h @@ -0,0 +1,16 @@ +/* + * Copyright (c) 2016 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ +/* Include the generated file, see BUILT_SOURCES in Makefile.am */ +#include diff --git a/src/plugins/ioam/lib-pot/pot_api.c b/src/plugins/ioam/lib-pot/pot_api.c new file mode 100644 index 00000000..d3af7b40 --- /dev/null +++ b/src/plugins/ioam/lib-pot/pot_api.c @@ -0,0 +1,292 @@ +/* + * Copyright (c) 2016 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ +/* + *------------------------------------------------------------------ + * pot_api.c - Proof of Transit related APIs to create + * and maintain profiles + *------------------------------------------------------------------ + */ + +#include +#include +#include + +#include +#include +#include + +/* define message IDs */ +#include + +/* define message structures */ +#define vl_typedefs +#include +#undef vl_typedefs + +/* define generated endian-swappers */ +#define vl_endianfun +#include +#undef vl_endianfun + +/* instantiate all the print functions we know about */ +#define vl_print(handle, ...) vlib_cli_output (handle, __VA_ARGS__) +#define vl_printfun +#include +#undef vl_printfun + +/* Get the API version number */ +#define vl_api_version(n,v) static u32 api_version=(v); +#include +#undef vl_api_version + +/* + * A handy macro to set up a message reply. + * Assumes that the following variables are available: + * mp - pointer to request message + * rmp - pointer to reply message type + * rv - return value + */ + +#define REPLY_MACRO(t) \ +do { \ + unix_shared_memory_queue_t * q = \ + vl_api_client_index_to_input_queue (mp->client_index); \ + if (!q) \ + return; \ + \ + rmp = vl_msg_api_alloc (sizeof (*rmp)); \ + rmp->_vl_msg_id = ntohs((t)+sm->msg_id_base); \ + rmp->context = mp->context; \ + rmp->retval = ntohl(rv); \ + \ + vl_msg_api_send_shmem (q, (u8 *)&rmp); \ +} while(0); + +#define REPLY_MACRO2(t, body) \ +do { \ + unix_shared_memory_queue_t * q; \ + rv = vl_msg_api_pd_handler (mp, rv); \ + q = vl_api_client_index_to_input_queue (mp->client_index); \ + if (!q) \ + return; \ + \ + rmp = vl_msg_api_alloc (sizeof (*rmp)); \ + rmp->_vl_msg_id = ntohs((t)+sm->msg_id_base); \ + rmp->context = mp->context; \ + rmp->retval = ntohl(rv); \ + do {body;} while (0); \ + vl_msg_api_send_shmem (q, (u8 *)&rmp); \ +} while(0); + +/* List of message types that this plugin understands */ + +#define foreach_pot_plugin_api_msg \ +_(POT_PROFILE_ADD, pot_profile_add) \ +_(POT_PROFILE_ACTIVATE, pot_profile_activate) \ +_(POT_PROFILE_DEL, pot_profile_del) \ +_(POT_PROFILE_SHOW_CONFIG_DUMP, pot_profile_show_config_dump) \ + +static void vl_api_pot_profile_add_t_handler +(vl_api_pot_profile_add_t *mp) +{ + pot_main_t * sm = &pot_main; + int rv = 0; + vl_api_pot_profile_add_reply_t * rmp; + u8 id; + pot_profile *profile = NULL; + u8 *name = 0; + + if (mp->list_name_len) + name = format(0, "%s", mp->list_name); + + pot_profile_list_init(name); + id = mp->id; + profile = pot_profile_find(id); + if (profile) { + rv = pot_profile_create(profile, + clib_net_to_host_u64(mp->prime), + clib_net_to_host_u64(mp->polynomial_public), + clib_net_to_host_u64(mp->lpc), + clib_net_to_host_u64(mp->secret_share)); + if (rv != 0) + goto ERROROUT; + if (1 == mp->validator) + (void)pot_set_validator(profile, clib_net_to_host_u64(mp->secret_key)); + (void)pot_profile_set_bit_mask(profile, mp->max_bits); + } else { + rv = -3; + } + ERROROUT: + vec_free(name); + REPLY_MACRO(VL_API_POT_PROFILE_ADD_REPLY); +} + +static void send_pot_profile_details(vl_api_pot_profile_show_config_dump_t *mp, u8 id) +{ + vl_api_pot_profile_show_config_details_t * rmp; + pot_main_t * sm = &pot_main; + pot_profile *profile = pot_profile_find(id); + int rv = 0; + if(profile){ + REPLY_MACRO2(VL_API_POT_PROFILE_SHOW_CONFIG_DETAILS, + rmp->id=id; + rmp->validator=profile->validator; + rmp->secret_key=clib_host_to_net_u64(profile->secret_key); + rmp->secret_share=clib_host_to_net_u64(profile->secret_share); + rmp->prime=clib_host_to_net_u64(profile->prime); + rmp->bit_mask=clib_host_to_net_u64(profile->bit_mask); + rmp->lpc=clib_host_to_net_u64(profile->lpc); + rmp->polynomial_public=clib_host_to_net_u64(profile->poly_pre_eval); + ); + } + else{ + REPLY_MACRO2(VL_API_POT_PROFILE_SHOW_CONFIG_DETAILS, + rmp->id=id; + rmp->validator=0; + rmp->secret_key=0; + rmp->secret_share=0; + rmp->prime=0; + rmp->bit_mask=0; + rmp->lpc=0; + rmp->polynomial_public=0; + ); + } +} + +static void vl_api_pot_profile_show_config_dump_t_handler +(vl_api_pot_profile_show_config_dump_t *mp) +{ + u8 id = mp->id; + u8 dump_call_id = ~0; + if(dump_call_id==id){ + for(id=0;idlist_name_len) + name = format(0, "%s", mp->list_name); + if (!pot_profile_list_is_enabled(name)) { + rv = -1; + } else { + id = mp->id; + rv = pot_profile_set_active(id); + } + + vec_free(name); + REPLY_MACRO(VL_API_POT_PROFILE_ACTIVATE_REPLY); +} + + +static void vl_api_pot_profile_del_t_handler +(vl_api_pot_profile_del_t *mp) +{ + pot_main_t * sm = &pot_main; + int rv = 0; + vl_api_pot_profile_del_reply_t * rmp; + + clear_pot_profiles(); + + REPLY_MACRO(VL_API_POT_PROFILE_DEL_REPLY); +} + + +/* + * This routine exists to convince the vlib plugin framework that + * we haven't accidentally copied a random .dll into the plugin directory. + * + * Also collects global variable pointers passed from the vpp engine + */ + +clib_error_t * +vlib_plugin_register (vlib_main_t * vm, vnet_plugin_handoff_t * h, + int from_early_init) +{ + pot_main_t * sm = &pot_main; + clib_error_t * error = 0; + + sm->vlib_main = vm; + sm->vnet_main = h->vnet_main; + return error; +} + +/* Set up the API message handling tables */ +static clib_error_t * +pot_plugin_api_hookup (vlib_main_t *vm) +{ + pot_main_t * sm = &pot_main; +#define _(N,n) \ + vl_msg_api_set_handlers((VL_API_##N + sm->msg_id_base), \ + #n, \ + vl_api_##n##_t_handler, \ + vl_noop_handler, \ + vl_api_##n##_t_endian, \ + vl_api_##n##_t_print, \ + sizeof(vl_api_##n##_t), 1); + foreach_pot_plugin_api_msg; +#undef _ + + return 0; +} + +#define vl_msg_name_crc_list +#include +#undef vl_msg_name_crc_list + +static void +setup_message_id_table (pot_main_t * sm, api_main_t * am) +{ +#define _(id,n,crc) \ + vl_msg_api_add_msg_name_crc (am, #n "_" #crc, id + sm->msg_id_base); + foreach_vl_msg_name_crc_pot; +#undef _ +} + +static clib_error_t * pot_init (vlib_main_t * vm) +{ + pot_main_t * sm = &pot_main; + clib_error_t * error = 0; + u8 * name; + + bzero(sm, sizeof(pot_main)); + (void)pot_util_init(); + name = format (0, "ioam_pot_%08x%c", api_version, 0); + + /* Ask for a correctly-sized block of API message decode slots */ + sm->msg_id_base = vl_msg_api_get_msg_ids + ((char *) name, VL_MSG_FIRST_AVAILABLE); + + error = pot_plugin_api_hookup (vm); + + /* Add our API messages to the global name_crc hash table */ + setup_message_id_table (sm, &api_main); + + vec_free(name); + + return error; +} + +VLIB_INIT_FUNCTION (pot_init); diff --git a/src/plugins/ioam/lib-pot/pot_msg_enum.h b/src/plugins/ioam/lib-pot/pot_msg_enum.h new file mode 100644 index 00000000..a4a88bed --- /dev/null +++ b/src/plugins/ioam/lib-pot/pot_msg_enum.h @@ -0,0 +1,28 @@ +/* + * Copyright (c) 2016 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ +#ifndef included_pot_msg_enum_h +#define included_pot_msg_enum_h + +#include + +#define vl_msg_id(n,h) n, +typedef enum { +#include + /* We'll want to know how many messages IDs we need... */ + VL_MSG_FIRST_AVAILABLE, +} vl_msg_id_t; +#undef vl_msg_id + +#endif /* included_pot_msg_enum_h */ diff --git a/src/plugins/ioam/lib-pot/pot_test.c b/src/plugins/ioam/lib-pot/pot_test.c new file mode 100644 index 00000000..2e870238 --- /dev/null +++ b/src/plugins/ioam/lib-pot/pot_test.c @@ -0,0 +1,365 @@ +/* + * Copyright (c) 2016 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ +/* + *------------------------------------------------------------------ + * pot_test.c - test harness for pot plugin + *------------------------------------------------------------------ + */ + +#include +#include +#include +#include +#include + +/* Declare message IDs */ +#include + +/* define message structures */ +#define vl_typedefs +#include +#undef vl_typedefs + +/* declare message handlers for each api */ + +#define vl_endianfun /* define message structures */ +#include +#undef vl_endianfun + +/* instantiate all the print functions we know about */ +#define vl_print(handle, ...) +#define vl_printfun +#include +#undef vl_printfun + +/* Get the API version number. */ +#define vl_api_version(n,v) static u32 api_version=(v); +#include +#undef vl_api_version + + +typedef struct { + /* API message ID base */ + u16 msg_id_base; + vat_main_t *vat_main; +} pot_test_main_t; + +pot_test_main_t pot_test_main; + +#define foreach_standard_reply_retval_handler \ +_(pot_profile_add_reply) \ +_(pot_profile_activate_reply) \ +_(pot_profile_del_reply) + +#define foreach_custom_reply_retval_handler \ +_(pot_profile_show_config_details, \ + errmsg(" ID:%d\n",mp->id); \ + errmsg(" Validator:%d\n",mp->validator); \ + errmsg(" secret_key:%Lx\n",clib_net_to_host_u64(mp->secret_key)); \ + errmsg(" secret_share:%Lx\n",clib_net_to_host_u64(mp->secret_share)); \ + errmsg(" prime:%Lx\n",clib_net_to_host_u64(mp->prime)); \ + errmsg(" bitmask:%Lx\n",clib_net_to_host_u64(mp->bit_mask)); \ + errmsg(" lpc:%Lx\n",clib_net_to_host_u64(mp->lpc)); \ + errmsg(" public poly:%Lx\n",clib_net_to_host_u64(mp->polynomial_public)); \ + ) + +#define _(n) \ + static void vl_api_##n##_t_handler \ + (vl_api_##n##_t * mp) \ + { \ + vat_main_t * vam = pot_test_main.vat_main; \ + i32 retval = ntohl(mp->retval); \ + if (vam->async_mode) { \ + vam->async_errors += (retval < 0); \ + } else { \ + vam->retval = retval; \ + vam->result_ready = 1; \ + } \ + } +foreach_standard_reply_retval_handler; +#undef _ + +#define _(n,body) \ + static void vl_api_##n##_t_handler \ + (vl_api_##n##_t * mp) \ + { \ + vat_main_t * vam = pot_test_main.vat_main; \ + i32 retval = ntohl(mp->retval); \ + if (vam->async_mode) { \ + vam->async_errors += (retval < 0); \ + } else { \ + vam->retval = retval; \ + vam->result_ready = 1; \ + } \ + do{body;}while(0); \ + } +foreach_custom_reply_retval_handler; +#undef _ + +/* + * Table of message reply handlers, must include boilerplate handlers + * we just generated + */ +#define foreach_vpe_api_reply_msg \ +_(POT_PROFILE_ADD_REPLY, pot_profile_add_reply) \ +_(POT_PROFILE_ACTIVATE_REPLY, pot_profile_activate_reply) \ +_(POT_PROFILE_DEL_REPLY, pot_profile_del_reply) \ +_(POT_PROFILE_SHOW_CONFIG_DETAILS, pot_profile_show_config_details) + + +/* M: construct, but don't yet send a message */ + +#define M(T,t) \ +do { \ + vam->result_ready = 0; \ + mp = vl_msg_api_alloc(sizeof(*mp)); \ + memset (mp, 0, sizeof (*mp)); \ + mp->_vl_msg_id = ntohs (VL_API_##T + sm->msg_id_base); \ + mp->client_index = vam->my_client_index; \ +} while(0); + +#define M2(T,t,n) \ +do { \ + vam->result_ready = 0; \ + mp = vl_msg_api_alloc(sizeof(*mp)+(n)); \ + memset (mp, 0, sizeof (*mp)); \ + mp->_vl_msg_id = ntohs (VL_API_##T + sm->msg_id_base); \ + mp->client_index = vam->my_client_index; \ +} while(0); + +/* S: send a message */ +#define S (vl_msg_api_send_shmem (vam->vl_input_queue, (u8 *)&mp)) + +/* W: wait for results, with timeout */ +#define W \ +do { \ + timeout = vat_time_now (vam) + 1.0; \ + \ + while (vat_time_now (vam) < timeout) { \ + if (vam->result_ready == 1) { \ + return (vam->retval); \ + } \ + } \ + return -99; \ +} while(0); + + +static int api_pot_profile_add (vat_main_t *vam) +{ +#define MAX_BITS 64 + pot_test_main_t * sm = &pot_test_main; + unformat_input_t *input = vam->input; + vl_api_pot_profile_add_t *mp; + u8 *name = NULL; + u64 prime = 0; + u64 secret_share = 0; + u64 secret_key = 0; + u32 bits = MAX_BITS; + u64 lpc = 0, poly2 = 0; + f64 timeout; + u8 id = 0; + int rv = 0; + + while (unformat_check_input(input) != UNFORMAT_END_OF_INPUT) + { + if (unformat(input, "name %s", &name)) + ; + else if(unformat(input, "id %d", &id)) + ; + else if (unformat(input, "validator-key 0x%Lx", &secret_key)) + ; + else if (unformat(input, "prime-number 0x%Lx", &prime)) + ; + else if (unformat(input, "secret-share 0x%Lx", &secret_share)) + ; + else if (unformat(input, "polynomial-public 0x%Lx", &poly2)) + ; + else if (unformat(input, "lpc 0x%Lx", &lpc)) + ; + else if (unformat(input, "bits-in-random %u", &bits)) + { + if (bits > MAX_BITS) + bits = MAX_BITS; + } + else + break; + } + + if (!name) + { + errmsg ("name required\n"); + rv = -99; + goto OUT; + } + + M2(POT_PROFILE_ADD, pot_profile_add, vec_len(name)); + + mp->list_name_len = vec_len(name); + clib_memcpy(mp->list_name, name, mp->list_name_len); + mp->secret_share = clib_host_to_net_u64(secret_share); + mp->polynomial_public = clib_host_to_net_u64(poly2); + mp->lpc = clib_host_to_net_u64(lpc); + mp->prime = clib_host_to_net_u64(prime); + if (secret_key != 0) + { + mp->secret_key = clib_host_to_net_u64(secret_key); + mp->validator = 1; + } + else + { + mp->validator = 0; + } + mp->id = id; + mp->max_bits = bits; + + S; W; + +OUT: + vec_free(name); + return(rv); +} + +static int api_pot_profile_activate (vat_main_t *vam) +{ +#define MAX_BITS 64 + pot_test_main_t * sm = &pot_test_main; + unformat_input_t *input = vam->input; + vl_api_pot_profile_activate_t *mp; + u8 *name = NULL; + u8 id = 0; + int rv = 0; + f64 timeout; + + while (unformat_check_input(input) != UNFORMAT_END_OF_INPUT) + { + if (unformat(input, "name %s", &name)) + ; + else if(unformat(input, "id %d", &id)) + ; + else + break; + } + + if (!name) + { + errmsg ("name required\n"); + rv = -99; + goto OUT; + } + + M2(POT_PROFILE_ACTIVATE, pot_profile_activate, vec_len(name)); + + mp->list_name_len = vec_len(name); + clib_memcpy(mp->list_name, name, mp->list_name_len); + mp->id = id; + + S; W; + +OUT: + vec_free(name); + return(rv); +} + + +static int api_pot_profile_del (vat_main_t *vam) +{ + pot_test_main_t * sm = &pot_test_main; + vl_api_pot_profile_del_t *mp; + f64 timeout; + + M(POT_PROFILE_DEL, pot_profile_del); + mp->list_name_len = 0; + S; W; + return 0; +} + +static int api_pot_profile_show_config_dump (vat_main_t *vam) +{ + pot_test_main_t * sm = &pot_test_main; + unformat_input_t *input = vam->input; + vl_api_pot_profile_show_config_dump_t *mp; + f64 timeout; + u8 id = 0; + + while(unformat_check_input(input) != UNFORMAT_END_OF_INPUT) + { + if(unformat(input,"id %d",&id)); + else + break; + } + M(POT_PROFILE_SHOW_CONFIG_DUMP, pot_profile_show_config_dump); + + mp->id = id; + + S; W; + return 0; +} + +/* + * List of messages that the api test plugin sends, + * and that the data plane plugin processes + */ +#define foreach_vpe_api_msg \ +_(pot_profile_add, "name id [0-1] " \ + "prime-number <0xu64> bits-in-random [0-64] " \ + "secret-share <0xu64> lpc <0xu64> polynomial-public <0xu64> " \ + "[validator-key <0xu64>] [validity <0xu64>]") \ +_(pot_profile_activate, "name id [0-1] ") \ +_(pot_profile_del, "[id ]") \ +_(pot_profile_show_config_dump, "id [0-1]") + +void vat_api_hookup (vat_main_t *vam) +{ + pot_test_main_t * sm = &pot_test_main; + /* Hook up handlers for replies from the data plane plug-in */ +#define _(N,n) \ + vl_msg_api_set_handlers((VL_API_##N + sm->msg_id_base), \ + #n, \ + vl_api_##n##_t_handler, \ + vl_noop_handler, \ + vl_api_##n##_t_endian, \ + vl_api_##n##_t_print, \ + sizeof(vl_api_##n##_t), 1); + foreach_vpe_api_reply_msg; +#undef _ + + /* API messages we can send */ +#define _(n,h) hash_set_mem (vam->function_by_name, #n, api_##n); + foreach_vpe_api_msg; +#undef _ + + /* Help strings */ +#define _(n,h) hash_set_mem (vam->help_by_name, #n, h); + foreach_vpe_api_msg; +#undef _ +} + +clib_error_t * vat_plugin_register (vat_main_t *vam) +{ + pot_test_main_t * sm = &pot_test_main; + u8 * name; + + sm->vat_main = vam; + + name = format (0, "ioam_pot_%08x%c", api_version, 0); + sm->msg_id_base = vl_client_get_first_plugin_msg_id ((char *) name); + + if (sm->msg_id_base != (u16) ~0) + vat_api_hookup (vam); + + vec_free(name); + + return 0; +} diff --git a/src/plugins/ioam/lib-pot/pot_util.c b/src/plugins/ioam/lib-pot/pot_util.c new file mode 100644 index 00000000..a253ad41 --- /dev/null +++ b/src/plugins/ioam/lib-pot/pot_util.c @@ -0,0 +1,445 @@ +/* + * Copyright (c) 2016 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ +#include +#include +#include +#include +#include +#include "math64.h" +#include "pot_util.h" + +pot_main_t pot_main; + +static void pot_profile_cleanup(pot_profile *profile); + +static void pot_main_profiles_reset (void) +{ + pot_main_t *sm = &pot_main; + int i = 0; + + for (i = 0; i < MAX_POT_PROFILES; i++) + { + pot_profile_cleanup(&(sm->profile_list[i])); + } + sm->active_profile_id = 0; + if (sm->profile_list_name) + vec_free(sm->profile_list_name); + sm->profile_list_name = NULL; +} + +int pot_util_init (void) +{ + pot_main_profiles_reset(); + + return(0); +} + +static void pot_profile_init(pot_profile * new, u8 id) +{ + if (new) + { + memset(new, 0, sizeof(pot_profile)); + new->id = id; + } +} + +pot_profile *pot_profile_find(u8 id) +{ + pot_main_t *sm = &pot_main; + + if (id < MAX_POT_PROFILES) + { + return (&(sm->profile_list[id])); + } + return (NULL); +} +static int pot_profile_name_equal (u8 *name0, u8 *name1) +{ + int len0, len1; + + len0 = vec_len (name0); + len1 = vec_len (name1); + if (len0 != len1) + return(0); + return (0==strncmp ((char *) name0, (char *)name1, len0)); +} + +int pot_profile_list_is_enabled (u8 *name) +{ + pot_main_t *sm = &pot_main; + return (pot_profile_name_equal(sm->profile_list_name, name)); +} + +void pot_profile_list_init(u8 * profile_list_name) +{ + pot_main_t *sm = &pot_main; + int i = 0; + + /* If it is the same profile list skip reset */ + if (pot_profile_name_equal(sm->profile_list_name, profile_list_name)) + { + return; + } + + pot_main_profiles_reset(); + if (vec_len(profile_list_name)) + sm->profile_list_name = (u8 *)vec_dup(profile_list_name); + else + sm->profile_list_name = 0; + sm->active_profile_id = 0; + + for (i = 0; i < MAX_POT_PROFILES; i++) + { + pot_profile_init(&(sm->profile_list[i]), i); + } +} + +static void pot_profile_cleanup(pot_profile * profile) +{ + u16 id = profile->id; + + memset(profile, 0, sizeof(pot_profile)); + profile->id = id; /* Restore id alone */ +} + +int pot_profile_create(pot_profile * profile, u64 prime, + u64 poly2, u64 lpc, u64 secret_share) +{ + if (profile && !profile->in_use) + { + pot_profile_cleanup(profile); + profile->prime = prime; + profile->primeinv = 1.0 / prime; + profile->lpc = lpc; + profile->poly_pre_eval = poly2; + profile->secret_share = secret_share; + profile->total_pkts_using_this_profile = 0; + profile->valid = 1; + return(0); + } + + return(-1); +} + +int pot_set_validator(pot_profile * profile, u64 key) +{ + if (profile && !profile->in_use) + { + profile->validator = 1; + profile->secret_key = key; + return(0); + } + return(-1); +} + +always_inline u64 pot_update_cumulative_inline(u64 cumulative, u64 random, + u64 secret_share, u64 prime, u64 lpc, u64 pre_split, double prime_inv) +{ + u64 share_random = 0; + u64 cumulative_new = 0; + + /* + * calculate split share for random + */ + share_random = add64_mod(pre_split, random, prime, prime_inv); + + /* + * lpc * (share_secret + share_random) + */ + share_random = add64_mod(share_random, secret_share, prime, prime_inv); + share_random = mul64_mod(share_random, lpc, prime, prime_inv); + + cumulative_new = add64_mod(cumulative, share_random, prime, prime_inv); + + return (cumulative_new); +} + +u64 pot_update_cumulative(pot_profile * profile, u64 cumulative, u64 random) +{ + if (profile && profile->valid != 0) + { + return (pot_update_cumulative_inline(cumulative, random, profile->secret_share, + profile->prime, profile->lpc, profile->poly_pre_eval, + profile->primeinv)); + } + return (0); +} + +always_inline u8 pot_validate_inline(u64 secret, u64 prime, double prime_inv, + u64 cumulative, u64 random) +{ + if (cumulative == (random + secret)) + { + return (1); + } + else if (cumulative == add64_mod(random, secret, prime, prime_inv)) + { + return (1); + } + return (0); +} + +/* + * return True if the cumulative matches secret from a profile + */ +u8 pot_validate(pot_profile * profile, u64 cumulative, u64 random) +{ + if (profile && profile->validator) + { + return (pot_validate_inline(profile->secret_key, profile->prime, + profile->primeinv, cumulative, random)); + } + return (0); +} + +/* + * Utility function to get random number per pack + */ +u64 pot_generate_random(pot_profile * profile) +{ + u64 random = 0; + int32_t second_half; + static u32 seed = 0; + + if (PREDICT_FALSE(!seed)) + seed = random_default_seed(); + + /* + * Upper 4 bytes seconds + */ + random = (u64) time(NULL); + + random &= 0xffffffff; + random = random << 32; + /* + * Lower 4 bytes random number + */ + second_half = random_u32(&seed); + + random |= second_half; + + if (PREDICT_TRUE(profile != NULL)) + { + random &= profile->bit_mask; + } + return (random); +} + +int pot_profile_set_bit_mask(pot_profile * profile, u16 bits) +{ + int sizeInBits; + + if (profile && !profile->in_use) + { + sizeInBits = sizeof(profile->bit_mask) * 8; + profile->bit_mask = + (bits >= + sizeInBits ? (u64) - 1 : (u64) ((u64) 1 << (u64) bits) - 1); + return(0); + } + return(-1); +} + +clib_error_t *clear_pot_profile_command_fn(vlib_main_t * vm, + unformat_input_t * input, vlib_cli_command_t * cmd) +{ + + pot_main_profiles_reset(); + + return 0; +} + +void clear_pot_profiles() +{ + clear_pot_profile_command_fn(0, 0, 0); +} + +VLIB_CLI_COMMAND(clear_pot_profile_command) = +{ +.path = "clear pot profile", +.short_help = "clear pot profile [|all]", +.function = clear_pot_profile_command_fn, +}; + +static clib_error_t *set_pot_profile_command_fn(vlib_main_t * vm, + unformat_input_t * input, vlib_cli_command_t * cmd) +{ + u64 prime; + u64 secret_share; + u64 secret_key; + u8 validator = 0; + u32 profile_id = ~0; + u32 bits; + u64 lpc = 0, poly2 = 0; + pot_profile *profile = NULL; + u8 *profile_list_name = NULL; + + bits = MAX_BITS; + + while (unformat_check_input(input) != UNFORMAT_END_OF_INPUT) + { + if (unformat(input, "name %s", + &profile_list_name)); + else if (unformat(input, "id %d", &profile_id)) + ; + else if (unformat(input, "validate-key 0x%Lx", &secret_key)) + validator = 1; + else if (unformat(input, "prime-number 0x%Lx", &prime)) + ; + else if (unformat(input, "secret_share 0x%Lx", &secret_share)) + ; + else if (unformat(input, "polynomial2 0x%Lx", &poly2)) + ; + else if (unformat(input, "lpc 0x%Lx", &lpc)) + ; + else if (unformat(input, "bits-in-random %d", &bits)) + { + if (bits > MAX_BITS) + bits = MAX_BITS; + } + else + break; + } + if (profile_list_name == 0) + { + return clib_error_return(0, "Name cannot be null"); + } + pot_profile_list_init(profile_list_name); + profile = pot_profile_find(profile_id); + + if (profile) + { + pot_profile_create(profile, prime, poly2, lpc, secret_share); + if (validator) + pot_set_validator(profile, secret_key); + pot_profile_set_bit_mask(profile, bits); + } + vec_free(profile_list_name); + return 0; +} + +VLIB_CLI_COMMAND(set_pot_profile_command) = +{ +.path = "set pot profile", +.short_help = "set pot profile name id [0-1] [validator-key 0xu64] \ + prime-number 0xu64 secret_share 0xu64 lpc 0xu64 \ + polynomial2 0xu64 bits-in-random [0-64] ", +.function = set_pot_profile_command_fn, +}; + +static clib_error_t *set_pot_profile_activate_command_fn(vlib_main_t * vm, + unformat_input_t * input, vlib_cli_command_t * cmd) +{ + pot_main_t *sm = &pot_main; + u8 *profile_list_name = NULL; + u32 id = 0; + clib_error_t *result = NULL; + + while (unformat_check_input(input) != UNFORMAT_END_OF_INPUT) + { + if (unformat(input, "name %s", + &profile_list_name)); + else if (unformat(input, "id %d", &id)) + ; + else + return clib_error_return(0, "unknown input `%U'", + format_unformat_error, input); + } + if (profile_list_name == 0) + { + return clib_error_return(0, "Name cannot be null"); + } + + if (!pot_profile_list_is_enabled(profile_list_name)) { + result = clib_error_return(0, "%s list is not enabled, profile in use %s", + profile_list_name, sm->profile_list_name); + } else if (0 != pot_profile_set_active((u8)id)) { + result = clib_error_return(0, "Profile %d not defined in %s", + id, sm->profile_list_name); + } + vec_free(profile_list_name); + return result; +} + +VLIB_CLI_COMMAND(set_pot_profile_activate_command) = +{ +.path = "set pot profile-active", +.short_help = "set pot profile-active name id [0-1]", +.function = set_pot_profile_activate_command_fn, +}; + +static clib_error_t *show_pot_profile_command_fn(vlib_main_t * vm, + unformat_input_t * input, vlib_cli_command_t * cmd) +{ + pot_main_t *sm = &pot_main; + pot_profile *p = NULL; + u16 i; + u8 *s = 0; + + if (vec_len(sm->profile_list_name) == 0) + { + s = format(s, "POT Profiles not configured\n"); + vlib_cli_output(vm, "%v", s); + return 0; + } + s = format(s, "Profile list in use : %s\n",sm->profile_list_name); + for (i = 0; i < MAX_POT_PROFILES; i++) + { + p = pot_profile_find(i); + if (p->valid == 0) + continue; + s = format(s, "POT Profile at index: %d\n", i); + s = format(s, " Id : %d\n", p->id); + s = format(s, " Validator : %s (%d)\n", + (p->validator) ? "True" : "False", p->validator); + if (p->validator == 1) + s = format(s, " Secret key : 0x%Lx (%Ld)\n", + p->secret_key, p->secret_key); + s = format(s, " Secret share : 0x%Lx (%Ld)\n", + p->secret_share, p->secret_share); + s = format(s, " Prime number : 0x%Lx (%Ld)\n", + p->prime, p->prime); + s = format(s, "2nd polynomial(eval) : 0x%Lx (%Ld)\n", + p->poly_pre_eval, p->poly_pre_eval); + s = format(s, " LPC : 0x%Lx (%Ld)\n", p->lpc, p->lpc); + + s = format(s, " Bit mask : 0x%Lx (%Ld)\n", + p->bit_mask, p->bit_mask); + } + + p = pot_profile_find(sm->active_profile_id); + + if (p && p->valid && p->in_use) { + s = format(s, "\nProfile index in use: %d\n", sm->active_profile_id); + s = format(s, "Pkts passed : 0x%Lx (%Ld)\n", + p->total_pkts_using_this_profile, + p->total_pkts_using_this_profile); + if (pot_is_decap(p)) + s = format(s, " This is Decap node. \n"); + } else { + s = format(s, "\nProfile index in use: None\n"); + } + vlib_cli_output(vm, "%v", s); + vec_free(s); + + return 0; +} + +VLIB_CLI_COMMAND(show_pot_profile_command) = +{ +.path = "show pot profile", +.short_help = "show pot profile", +.function = show_pot_profile_command_fn, +}; diff --git a/src/plugins/ioam/lib-pot/pot_util.h b/src/plugins/ioam/lib-pot/pot_util.h new file mode 100644 index 00000000..9df31fae --- /dev/null +++ b/src/plugins/ioam/lib-pot/pot_util.h @@ -0,0 +1,195 @@ +/* + * pot_util.h -- Proof Of Transit Utility Header + * + * Copyright (c) 2016 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +#ifndef include_vnet_pot_util_h +#define include_vnet_pot_util_h + +#include +#define debug_ioam debug_ioam_fn +/* Dont change this size 256. This is there across multiple components */ +#define PATH_NAME_SIZE 256 + +/* Ring size. this should be same as the one in ODL. Do not change this + without change in ODL. */ +#define MAX_POT_PROFILES 2 + +/** + * Usage: + * + * On any node that participates in Proof of Transit: + * + * Step 1: Initialize this library by calling pot_init() + * Step 2: Setup a proof of transit profile that contains all the parameters needed to compute cumulative: + * Call these functions: + * pot_profile_find + * pot_profile_create + * pot_profile_set_bit_mask - To setup how large we want the numbers used in the computation and random number <= 64 bits + * Step 2a: For validator do this: + * pot_set_validator + * Step 2b: On initial node enable the profile to be used: + * pot_profile_set_active / pot_profile_get_active will return the profile + * Step 3a: At the initial node to generate Random number that will be read by all other nodes: + * pot_generate_random + * Step 3b: At all nodes including initial and verifier call this to compute cumulative: + * pot_update_cumulative + * Step 4: At the verifier: + * pot_validate + * + */ + +typedef struct pot_profile_ +{ + u8 id : 1; + u8 valid : 1; + u8 in_use : 1; + u64 random; + u8 validator; + u64 secret_key; + u64 secret_share; + u64 prime; + u64 lpc; + u64 poly_pre_eval; + u64 bit_mask; + u64 limit; + double primeinv; + u64 total_pkts_using_this_profile; +} pot_profile; + +typedef struct { + /* Name of the default profile list in use*/ + u8 *profile_list_name; + pot_profile profile_list[MAX_POT_PROFILES]; + /* number of profiles in the list */ + u8 active_profile_id : 1; + + /* API message ID base */ + u16 msg_id_base; + + /* convenience */ + vlib_main_t * vlib_main; + vnet_main_t * vnet_main; +} pot_main_t; + +extern pot_main_t pot_main; + +/* + * Initialize proof of transit + */ +int pot_util_init(void); +void pot_profile_list_init(u8 * name); + + +/* + * Find a pot profile by ID + */ +pot_profile *pot_profile_find(u8 id); + +static inline u16 pot_profile_get_id(pot_profile * profile) +{ + if (profile) + { + return (profile->id); + } + return (0); +} + +/* setup and clean up profile */ +int pot_profile_create(pot_profile * profile, u64 prime, + u64 poly2, u64 lpc, u64 secret_share); +/* + * Setup profile as a validator + */ +int pot_set_validator(pot_profile * profile, u64 key); + +/* + * Setup max bits to be used for random number generation + */ +#define MAX_BITS 64 +int pot_profile_set_bit_mask(pot_profile * profile, u16 bits); + +/* + * Given a random and cumulative compute the new cumulative for a given profile + */ +u64 pot_update_cumulative(pot_profile * profile, u64 cumulative, u64 random); + +/* + * return True if the cumulative matches secret from a profile + */ +u8 pot_validate(pot_profile * profile, u64 cumulative, u64 random); + +/* + * Utility function to get random number per pack + */ +u64 pot_generate_random(pot_profile * profile); + + +extern void clear_pot_profiles(); +extern int pot_profile_list_is_enabled(u8 *name); + +static inline u8 pot_is_decap(pot_profile * p) +{ + return (p->validator == 1); +} + +static inline int pot_profile_set_active (u8 id) +{ + pot_main_t *sm = &pot_main; + pot_profile *profile = NULL; + pot_profile *current_active_prof = NULL; + + current_active_prof = pot_profile_find(sm->active_profile_id); + profile = pot_profile_find(id); + if (profile && profile->valid) { + sm->active_profile_id = id; + current_active_prof->in_use = 0; + profile->in_use = 1; + return(0); + } + return(-1); +} +static inline u8 pot_profile_get_active_id (void) +{ + pot_main_t *sm = &pot_main; + return (sm->active_profile_id); +} + +static inline pot_profile * pot_profile_get_active (void) +{ + pot_main_t *sm = &pot_main; + pot_profile *profile = NULL; + profile = pot_profile_find(sm->active_profile_id); + if (profile && profile->in_use) + return(profile); + return (NULL); +} + +static inline void pot_profile_reset_usage_stats (pot_profile *pow) +{ + if (pow) { + pow->total_pkts_using_this_profile = 0; + } +} + +static inline void pot_profile_incr_usage_stats (pot_profile *pow) +{ + if (pow) { + pow->total_pkts_using_this_profile++; + } +} + + +#endif diff --git a/src/plugins/ioam/lib-trace/trace.api b/src/plugins/ioam/lib-trace/trace.api new file mode 100644 index 00000000..cb958325 --- /dev/null +++ b/src/plugins/ioam/lib-trace/trace.api @@ -0,0 +1,92 @@ +/* Hey Emacs use -*- mode: C -*- */ +/* + * Copyright (c) 2016 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + + +/** \brief iOAM6 Trace - Set the iOAM6 trace profile + @param trace_type - Type of trace requested + @param num_elts - Number of trace elements to be inserted + @param node_id - Trace Node ID + @param trace_tsp- Timestamp resolution + @param app_data - Application specific opaque +*/ +define trace_profile_add { + u32 client_index; + u32 context; + u8 trace_type; + u8 num_elts; + u8 trace_tsp; + u32 node_id; + u32 app_data; +}; + +/** \brief Trace profile add / del response + @param context - sender context, to match reply w/ request + @param retval - return value for request +*/ +define trace_profile_add_reply { + u32 context; + i32 retval; +}; + + + +/** \brief Delete trace Profile + @param client_index - opaque cookie to identify the sender + @param context - sender context, to match reply w/ request +*/ +define trace_profile_del { + u32 client_index; + u32 context; +}; + +/** \brief Trace profile add / del response + @param context - sender context, to match reply w/ request + @param retval - return value for request +*/ +define trace_profile_del_reply { + u32 context; + i32 retval; +}; + + + +/** \brief Show trace Profile + @param client_index - opaque cookie to identify the sender + @param context - sender context, to match reply w/ request +*/ +define trace_profile_show_config { + u32 client_index; + u32 context; +}; + +/** \brief Show trace config response + @param context - sender context, to match reply w/ request + @param retval - return value for request + @param trace_type - Type of trace requested + @param num_elts - Number of trace elements to be inserted + @param node_id - Trace Node ID + @param trace_tsp- Timestamp resolution + @param app_data - Application specific opaque +*/ +define trace_profile_show_config_reply { + u32 context; + i32 retval; + u8 trace_type; + u8 num_elts; + u8 trace_tsp; + u32 node_id; + u32 app_data; +}; diff --git a/src/plugins/ioam/lib-trace/trace_all_api_h.h b/src/plugins/ioam/lib-trace/trace_all_api_h.h new file mode 100644 index 00000000..223f9545 --- /dev/null +++ b/src/plugins/ioam/lib-trace/trace_all_api_h.h @@ -0,0 +1,16 @@ +/* + * Copyright (c) 2016 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ +/* Include the generated file, see BUILT_SOURCES in Makefile.am */ +#include diff --git a/src/plugins/ioam/lib-trace/trace_api.c b/src/plugins/ioam/lib-trace/trace_api.c new file mode 100644 index 00000000..7e0d708e --- /dev/null +++ b/src/plugins/ioam/lib-trace/trace_api.c @@ -0,0 +1,252 @@ +/* + * Copyright (c) 2016 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ +/* + *------------------------------------------------------------------ + * trace_api.c - iOAM Trace related APIs to create + * and maintain profiles + *------------------------------------------------------------------ + */ + +#include +#include +#include + +#include +#include +#include + +/* define message IDs */ +#include + +/* define message structures */ +#define vl_typedefs +#include +#undef vl_typedefs + +/* define generated endian-swappers */ +#define vl_endianfun +#include +#undef vl_endianfun + +/* instantiate all the print functions we know about */ +#define vl_print(handle, ...) vlib_cli_output (handle, __VA_ARGS__) +#define vl_printfun +#include +#undef vl_printfun + +/* Get the API version number */ +#define vl_api_version(n,v) static u32 api_version=(v); +#include +#undef vl_api_version + +/* + * A handy macro to set up a message reply. + * Assumes that the following variables are available: + * mp - pointer to request message + * rmp - pointer to reply message type + * rv - return value + */ + +#define TRACE_REPLY_MACRO(t) \ +do { \ + unix_shared_memory_queue_t * q = \ + vl_api_client_index_to_input_queue (mp->client_index); \ + if (!q) \ + return; \ + \ + rmp = vl_msg_api_alloc (sizeof (*rmp)); \ + rmp->_vl_msg_id = ntohs((t)+sm->msg_id_base); \ + rmp->context = mp->context; \ + rmp->retval = ntohl(rv); \ + \ + vl_msg_api_send_shmem (q, (u8 *)&rmp); \ +} while(0); + +/* *INDENT-OFF* */ +#define TRACE_REPLY_MACRO2(t, body) \ +do { \ + unix_shared_memory_queue_t * q; \ + rv = vl_msg_api_pd_handler (mp, rv); \ + q = vl_api_client_index_to_input_queue (mp->client_index); \ + if (!q) \ + return; \ + \ + rmp = vl_msg_api_alloc (sizeof (*rmp)); \ + rmp->_vl_msg_id = ntohs((t)+sm->msg_id_base); \ + rmp->context = mp->context; \ + rmp->retval = ntohl(rv); \ + do {body;} while (0); \ + vl_msg_api_send_shmem (q, (u8 *)&rmp); \ +} while(0); +/* *INDENT-ON* */ + +/* List of message types that this plugin understands */ + +#define foreach_trace_plugin_api_msg \ +_(TRACE_PROFILE_ADD, trace_profile_add) \ +_(TRACE_PROFILE_DEL, trace_profile_del) \ +_(TRACE_PROFILE_SHOW_CONFIG, trace_profile_show_config) + +static void vl_api_trace_profile_add_t_handler + (vl_api_trace_profile_add_t * mp) +{ + trace_main_t *sm = &trace_main; + int rv = 0; + vl_api_trace_profile_add_reply_t *rmp; + trace_profile *profile = NULL; + + profile = trace_profile_find (); + if (profile) + { + rv = + trace_profile_create (profile, mp->trace_type, mp->num_elts, + mp->trace_tsp, ntohl (mp->node_id), + ntohl (mp->app_data)); + if (rv != 0) + goto ERROROUT; + } + else + { + rv = -3; + } +ERROROUT: + TRACE_REPLY_MACRO (VL_API_TRACE_PROFILE_ADD_REPLY); +} + + +static void vl_api_trace_profile_del_t_handler + (vl_api_trace_profile_del_t * mp) +{ + trace_main_t *sm = &trace_main; + int rv = 0; + vl_api_trace_profile_del_reply_t *rmp; + + clear_trace_profiles (); + + TRACE_REPLY_MACRO (VL_API_TRACE_PROFILE_DEL_REPLY); +} + +static void vl_api_trace_profile_show_config_t_handler + (vl_api_trace_profile_show_config_t * mp) +{ + trace_main_t *sm = &trace_main; + vl_api_trace_profile_show_config_reply_t *rmp; + int rv = 0; + trace_profile *profile = trace_profile_find (); + if (profile->valid) + { + TRACE_REPLY_MACRO2 (VL_API_TRACE_PROFILE_SHOW_CONFIG_REPLY, + rmp->trace_type = profile->trace_type; + rmp->num_elts = profile->num_elts; + rmp->trace_tsp = profile->trace_tsp; + rmp->node_id = htonl (profile->node_id); + rmp->app_data = htonl (profile->app_data); + ); + } + else + { + TRACE_REPLY_MACRO2 (VL_API_TRACE_PROFILE_SHOW_CONFIG_REPLY, + rmp->trace_type = 0; + rmp->num_elts = 0; rmp->trace_tsp = 0; + rmp->node_id = 0; rmp->app_data = 0; + ); + } +} + +/* + * This routine exists to convince the vlib plugin framework that + * we haven't accidentally copied a random .dll into the plugin directory. + * + * Also collects global variable pointers passed from the vpp engine + */ + +clib_error_t * +vlib_plugin_register (vlib_main_t * vm, vnet_plugin_handoff_t * h, + int from_early_init) +{ + trace_main_t *sm = &trace_main; + clib_error_t *error = 0; + + sm->vlib_main = vm; + sm->vnet_main = h->vnet_main; + return error; +} + +/* Set up the API message handling tables */ +static clib_error_t * +trace_plugin_api_hookup (vlib_main_t * vm) +{ + trace_main_t *sm = &trace_main; +#define _(N,n) \ + vl_msg_api_set_handlers((VL_API_##N + sm->msg_id_base), \ + #n, \ + vl_api_##n##_t_handler, \ + vl_noop_handler, \ + vl_api_##n##_t_endian, \ + vl_api_##n##_t_print, \ + sizeof(vl_api_##n##_t), 1); + foreach_trace_plugin_api_msg; +#undef _ + + return 0; +} + +#define vl_msg_name_crc_list +#include +#undef vl_msg_name_crc_list + +static void +setup_message_id_table (trace_main_t * sm, api_main_t * am) +{ +#define _(id,n,crc) \ + vl_msg_api_add_msg_name_crc (am, #n "_" #crc, id + sm->msg_id_base); + foreach_vl_msg_name_crc_trace; +#undef _ +} + +static clib_error_t * +trace_init (vlib_main_t * vm) +{ + trace_main_t *sm = &trace_main; + clib_error_t *error = 0; + u8 *name; + + bzero (sm, sizeof (trace_main)); + (void) trace_util_init (); + name = format (0, "ioam_trace_%08x%c", api_version, 0); + + /* Ask for a correctly-sized block of API message decode slots */ + sm->msg_id_base = vl_msg_api_get_msg_ids + ((char *) name, VL_MSG_FIRST_AVAILABLE); + + error = trace_plugin_api_hookup (vm); + + /* Add our API messages to the global name_crc hash table */ + setup_message_id_table (sm, &api_main); + + vec_free (name); + + return error; +} + +VLIB_INIT_FUNCTION (trace_init); + +/* + * fd.io coding-style-patch-verification: ON + * + * Local Variables: + * eval: (c-set-style "gnu") + * End: + */ diff --git a/src/plugins/ioam/lib-trace/trace_msg_enum.h b/src/plugins/ioam/lib-trace/trace_msg_enum.h new file mode 100644 index 00000000..78c35665 --- /dev/null +++ b/src/plugins/ioam/lib-trace/trace_msg_enum.h @@ -0,0 +1,28 @@ +/* + * Copyright (c) 2016 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ +#ifndef included_trace_msg_enum_h +#define included_trace_msg_enum_h + +#include + +#define vl_msg_id(n,h) n, +typedef enum { +#include + /* We'll want to know how many messages IDs we need... */ + VL_MSG_FIRST_AVAILABLE, +} vl_msg_id_t; +#undef vl_msg_id + +#endif /* included_trace_msg_enum_h */ diff --git a/src/plugins/ioam/lib-trace/trace_test.c b/src/plugins/ioam/lib-trace/trace_test.c new file mode 100644 index 00000000..111dd461 --- /dev/null +++ b/src/plugins/ioam/lib-trace/trace_test.c @@ -0,0 +1,292 @@ +/* + * Copyright (c) 2016 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ +/* + *------------------------------------------------------------------ + * trace_test.c - test harness for trace plugin + *------------------------------------------------------------------ + */ + +#include +#include +#include +#include +#include + +/* Declare message IDs */ +#include + +/* define message structures */ +#define vl_typedefs +#include +#undef vl_typedefs + +/* declare message handlers for each api */ + +#define vl_endianfun /* define message structures */ +#include +#undef vl_endianfun + +/* instantiate all the print functions we know about */ +#define vl_print(handle, ...) +#define vl_printfun +#include +#undef vl_printfun + +/* Get the API version number. */ +#define vl_api_version(n,v) static u32 api_version=(v); +#include +#undef vl_api_version + + +typedef struct +{ + /* API message ID base */ + u16 msg_id_base; + vat_main_t *vat_main; +} trace_test_main_t; + +trace_test_main_t trace_test_main; + +#define foreach_standard_reply_retval_handler \ +_(trace_profile_add_reply) \ +_(trace_profile_del_reply) + +#define foreach_custom_reply_handler \ +_(trace_profile_show_config_reply, \ + if(mp->trace_type) \ + { \ + errmsg(" Trace Type : 0x%x (%d)\n",mp->trace_type, mp->trace_type); \ + errmsg(" Trace timestamp precision : %d \n",mp->trace_tsp); \ + errmsg(" Node Id : 0x%x (%d)\n",htonl(mp->node_id), htonl(mp->node_id)); \ + errmsg(" App Data : 0x%x (%d)\n",htonl(mp->app_data), htonl(mp->app_data)); \ + } \ + else errmsg("No valid trace profile configuration found\n");) +#define _(n) \ + static void vl_api_##n##_t_handler \ + (vl_api_##n##_t * mp) \ + { \ + vat_main_t * vam = trace_test_main.vat_main; \ + i32 retval = ntohl(mp->retval); \ + if (vam->async_mode) { \ + vam->async_errors += (retval < 0); \ + } else { \ + vam->retval = retval; \ + vam->result_ready = 1; \ + } \ + } +foreach_standard_reply_retval_handler; +#undef _ + +#define _(n,body) \ + static void vl_api_##n##_t_handler \ + (vl_api_##n##_t * mp) \ + { \ + vat_main_t * vam = trace_test_main.vat_main; \ + i32 retval = ntohl(mp->retval); \ + if (vam->async_mode) { \ + vam->async_errors += (retval < 0); \ + } else { \ + vam->retval = retval; \ + vam->result_ready = 1; \ + } \ + if(retval>=0)do{body;} while(0); \ + else errmsg("Error, retval: %d",retval); \ + } +foreach_custom_reply_handler; +#undef _ +/* + * Table of message reply handlers, must include boilerplate handlers + * we just generated + */ +#define foreach_vpe_api_reply_msg \ +_(TRACE_PROFILE_ADD_REPLY, trace_profile_add_reply) \ +_(TRACE_PROFILE_DEL_REPLY, trace_profile_del_reply) \ +_(TRACE_PROFILE_SHOW_CONFIG_REPLY, trace_profile_show_config_reply) + + +/* M: construct, but don't yet send a message */ + +#define M(T,t) \ +do { \ + vam->result_ready = 0; \ + mp = vl_msg_api_alloc(sizeof(*mp)); \ + memset (mp, 0, sizeof (*mp)); \ + mp->_vl_msg_id = ntohs (VL_API_##T + sm->msg_id_base); \ + mp->client_index = vam->my_client_index; \ +} while(0); + +#define M2(T,t,n) \ +do { \ + vam->result_ready = 0; \ + mp = vl_msg_api_alloc(sizeof(*mp)+(n)); \ + memset (mp, 0, sizeof (*mp)); \ + mp->_vl_msg_id = ntohs (VL_API_##T + sm->msg_id_base); \ + mp->client_index = vam->my_client_index; \ +} while(0); + +/* S: send a message */ +#define S (vl_msg_api_send_shmem (vam->vl_input_queue, (u8 *)&mp)) + +/* W: wait for results, with timeout */ +#define W \ +do { \ + timeout = vat_time_now (vam) + 1.0; \ + \ + while (vat_time_now (vam) < timeout) { \ + if (vam->result_ready == 1) { \ + return (vam->retval); \ + } \ + } \ + return -99; \ +} while(0); + + +static int +api_trace_profile_add (vat_main_t * vam) +{ + trace_test_main_t *sm = &trace_test_main; + unformat_input_t *input = vam->input; + vl_api_trace_profile_add_t *mp; + u8 trace_type = 0; + u8 num_elts = 0; + int rv = 0; + u32 node_id = 0; + u32 app_data = 0; + u8 trace_tsp = 0; + f64 timeout; + + while (unformat_check_input (input) != UNFORMAT_END_OF_INPUT) + { + if (unformat (input, "trace-type 0x%x", &trace_type)) + ; + else if (unformat (input, "trace-elts %d", &num_elts)) + ; + else if (unformat (input, "trace-tsp %d", &trace_tsp)) + ; + else if (unformat (input, "node-id 0x%x", &node_id)) + ; + else if (unformat (input, "app-data 0x%x", &app_data)) + ; + + else + break; + } + + + M (TRACE_PROFILE_ADD, trace_profile_add); + + mp->trace_type = trace_type; + mp->trace_tsp = trace_tsp; + mp->node_id = htonl (node_id); + mp->app_data = htonl (app_data); + mp->num_elts = num_elts; + + S; + W; + + return (rv); +} + + + +static int +api_trace_profile_del (vat_main_t * vam) +{ + trace_test_main_t *sm = &trace_test_main; + vl_api_trace_profile_del_t *mp; + f64 timeout; + + M (TRACE_PROFILE_DEL, trace_profile_del); + S; + W; + return 0; +} + +static int +api_trace_profile_show_config (vat_main_t * vam) +{ + trace_test_main_t *sm = &trace_test_main; + vl_api_trace_profile_show_config_t *mp; + f64 timeout; + M (TRACE_PROFILE_SHOW_CONFIG, trace_profile_show_config); + S; + W; + return 0; +} + +/* + * List of messages that the api test plugin sends, + * and that the data plane plugin processes + */ +#define foreach_vpe_api_msg \ +_(trace_profile_add, ""\ + "trace-type <0x1f|0x3|0x9|0x11|0x19> trace-elts trace-tsp <0|1|2|3> node-id app-data ") \ +_(trace_profile_del, "[id ]") \ +_(trace_profile_show_config, "[id ]") + + +void +vat_api_hookup (vat_main_t * vam) +{ + trace_test_main_t *sm = &trace_test_main; + /* Hook up handlers for replies from the data plane plug-in */ +#define _(N,n) \ + vl_msg_api_set_handlers((VL_API_##N + sm->msg_id_base), \ + #n, \ + vl_api_##n##_t_handler, \ + vl_noop_handler, \ + vl_api_##n##_t_endian, \ + vl_api_##n##_t_print, \ + sizeof(vl_api_##n##_t), 1); + foreach_vpe_api_reply_msg; +#undef _ + + /* API messages we can send */ +#define _(n,h) hash_set_mem (vam->function_by_name, #n, api_##n); + foreach_vpe_api_msg; +#undef _ + + /* Help strings */ +#define _(n,h) hash_set_mem (vam->help_by_name, #n, h); + foreach_vpe_api_msg; +#undef _ +} + +clib_error_t * +vat_plugin_register (vat_main_t * vam) +{ + trace_test_main_t *sm = &trace_test_main; + u8 *name; + + sm->vat_main = vam; + + name = format (0, "ioam_trace_%08x%c", api_version, 0); + sm->msg_id_base = vl_client_get_first_plugin_msg_id ((char *) name); + + if (sm->msg_id_base != (u16) ~ 0) + vat_api_hookup (vam); + + vec_free (name); + + return 0; +} + +/* + * fd.io coding-style-patch-verification: ON + * + * Local Variables: + * eval: (c-set-style "gnu") + * End: + */ diff --git a/src/plugins/ioam/lib-trace/trace_util.c b/src/plugins/ioam/lib-trace/trace_util.c new file mode 100644 index 00000000..5c7f1eef --- /dev/null +++ b/src/plugins/ioam/lib-trace/trace_util.c @@ -0,0 +1,206 @@ +/* + * Copyright (c) 2016 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ +#include +#include +#include +#include +#include +#include "trace_util.h" + +trace_main_t trace_main; + +static int +trace_profile_cleanup (trace_profile * profile) +{ + + memset (profile, 0, sizeof (trace_profile)); + profile->trace_tsp = TSP_MICROSECONDS; /* Micro seconds */ + ip6_trace_profile_cleanup (); /* lib-trace_TODO: Remove this once IOAM-IPv6 transport is a plugin */ + return 0; + +} + +static int +trace_main_profiles_reset (void) +{ + int rv; + + trace_main_t *sm = &trace_main; + rv = trace_profile_cleanup (&(sm->profile)); + return (rv); +} + +int +trace_util_init (void) +{ + int rv; + + rv = trace_main_profiles_reset (); + return (rv); +} + + +int +trace_profile_create (trace_profile * profile, u8 trace_type, u8 num_elts, + u32 trace_tsp, u32 node_id, u32 app_data) +{ + + if (!trace_type || !num_elts || !(node_id)) + { + return (-1); + } + if (profile && !profile->valid) + { + //rv = trace_profile_cleanup (profile); + profile->trace_type = trace_type; + profile->num_elts = num_elts; + profile->trace_tsp = trace_tsp; + profile->node_id = node_id; + profile->app_data = app_data; + profile->valid = 1; + + /* lib-trace_TODO: Remove this once IOAM-IPv6 transport is a plugin */ + ip6_trace_profile_setup (); + return (0); + } + + return (-1); +} + + + +clib_error_t * +clear_trace_profile_command_fn (vlib_main_t * vm, + unformat_input_t * input, + vlib_cli_command_t * cmd) +{ + + trace_main_profiles_reset (); + return 0; +} + +void +clear_trace_profiles (void) +{ + clear_trace_profile_command_fn (0, 0, 0); +} + +/* *INDENT-OFF* */ +VLIB_CLI_COMMAND(clear_trace_profile_command) = +{ +.path = "clear ioam-trace profile", +.short_help = "clear ioam-trace profile [|all]", +.function = clear_trace_profile_command_fn, +}; +/* *INDENT-ON* */ + +static clib_error_t * +set_trace_profile_command_fn (vlib_main_t * vm, + unformat_input_t * input, + vlib_cli_command_t * cmd) +{ + u8 trace_type = 0; + u8 num_elts = 0; + u32 node_id = 0; + u32 app_data = 0; + u32 trace_tsp = 0; + trace_profile *profile = NULL; + while (unformat_check_input (input) != UNFORMAT_END_OF_INPUT) + { + if (unformat (input, "trace-type 0x%x", &trace_type)); + else if (unformat (input, "trace-elts %d", &num_elts)); + else if (unformat (input, "trace-tsp %d", &trace_tsp)); + else if (unformat (input, "node-id 0x%x", &node_id)); + else if (unformat (input, "app-data 0x%x", &app_data)); + else + break; + } + profile = trace_profile_find (); + if (profile) + { + trace_profile_create (profile, trace_type, num_elts, trace_tsp, + node_id, app_data); + } + return 0; +} + +/* *INDENT-OFF* */ +VLIB_CLI_COMMAND (set_trace_profile_command, static) = +{ +.path = "set ioam-trace profile", +.short_help = "set ioam-trace \ + trace-type <0x1f|0x3|0x9|0x11|0x19> trace-elts trace-tsp <0|1|2|3> \ + node-id app-data ", +.function = set_trace_profile_command_fn, +}; +/* *INDENT-ON* */ + +static clib_error_t * +show_trace_profile_command_fn (vlib_main_t * vm, + unformat_input_t * input, + vlib_cli_command_t * cmd) +{ + trace_profile *p = NULL; + u8 *s = 0; + p = trace_profile_find (); + if (!(p && p->valid)) + { + s = format (s, "\nTrace configuration not valid\n"); + vlib_cli_output (vm, "%v", s); + vec_free (s); + return 0; + } + s = format (s, " HOP BY HOP OPTIONS - TRACE CONFIG - \n"); + s = format (s, " Trace Type : 0x%x (%d)\n", + p->trace_type, p->trace_type); + s = + format (s, " Trace timestamp precision : %d (%s)\n", + p->trace_tsp, + (p->trace_tsp == + TSP_SECONDS) ? "Seconds" : ((p->trace_tsp == + TSP_MILLISECONDS) ? + "Milliseconds" + : (((p->trace_tsp == + TSP_MICROSECONDS) ? + "Microseconds" : + "Nanoseconds")))); + s = format (s, " Num of trace nodes : %d\n", p->num_elts); + s = + format (s, " Node-id : 0x%x (%d)\n", + p->node_id, p->node_id); + s = + format (s, " App Data : 0x%x (%d)\n", + p->app_data, p->app_data); + vlib_cli_output (vm, "%v", s); + vec_free (s); + return 0; +} + +/* *INDENT-OFF* */ +VLIB_CLI_COMMAND (show_trace_profile_command, static) = +{ +.path = "show ioam-trace profile", +.short_help = "show ioam-trace profile", +.function = show_trace_profile_command_fn, +}; +/* *INDENT-ON* */ + +/* + * fd.io coding-style-patch-verification: ON + * + * Local Variables: + * eval: (c-set-style "gnu") + * End: + */ diff --git a/src/plugins/ioam/lib-trace/trace_util.h b/src/plugins/ioam/lib-trace/trace_util.h new file mode 100644 index 00000000..556f07ee --- /dev/null +++ b/src/plugins/ioam/lib-trace/trace_util.h @@ -0,0 +1,247 @@ +/* + * trace_util.h -- Trace Profile Utility header + * + * Copyright (c) 2016 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +#ifndef include_vnet_trace_util_h +#define include_vnet_trace_util_h + +#define debug_ioam debug_ioam_fn + + +/** + * Usage: + * + * On any node that participates in iOAM Trace. + * + * Step 1: Initialize this library by calling trace_init() + * Step 2: Setup a trace profile that contains all the parameters needed to compute cumulative: + * Call these functions: + * trace_profile_find + * trace_profile_create + * Step 2a: On initial node enable the profile to be used: + * trace_profile_set_active / trace_profile_get_active will return the profile + * Step 4: TBD + * trace_validate + * + */ + +typedef struct trace_profile_ +{ + u8 valid:1; + u8 trace_type; + u8 num_elts; + /* Configured node-id */ + u32 node_id; + u32 app_data; + u32 trace_tsp; +} trace_profile; + +typedef struct +{ + /* Name of the default profile list in use */ + trace_profile profile; + + /* API message ID base */ + u16 msg_id_base; + + /* convenience */ + vlib_main_t *vlib_main; + vnet_main_t *vnet_main; +} trace_main_t; + +extern trace_main_t trace_main; + +/* + * Initialize Trace profile + */ +int trace_util_init (void); + + +/* + * Find a trace profile + */ + +always_inline trace_profile * +trace_profile_find (void) +{ + trace_main_t *sm = &trace_main; + + return (&(sm->profile)); +} + + +/* setup and clean up profile */ +int trace_profile_create (trace_profile * profile, u8 trace_type, u8 num_elts, + u32 trace_tsp, u32 node_id, u32 app_data); + +void clear_trace_profiles (void); + + + +#define BIT_TTL_NODEID (1<<0) +#define BIT_ING_INTERFACE (1<<1) +#define BIT_EGR_INTERFACE (1<<2) +#define BIT_TIMESTAMP (1<<3) +#define BIT_APPDATA (1<<4) +#define TRACE_TYPE_MASK 0x1F /* Mask of all above bits */ + +/* + 0x00011111 iOAM-trace-type is 0x00011111 then the format of node + data is: + + 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 + +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ + | Hop_Lim | node_id | + +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ + | ingress_if_id | egress_if_id | + +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ + + timestamp + + +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ + | app_data | + +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ + +*/ +#define TRACE_TYPE_IF_TS_APP 0x1f +typedef struct +{ + u32 ttl_node_id; + u16 ingress_if; + u16 egress_if; + u32 timestamp; + u32 app_data; +} ioam_trace_if_ts_app_t; + +/* + 0x00000111 iOAM-trace-type is 0x00000111 then the format is: + + 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 + +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ + | Hop_Lim | node_id | + +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ + | ingress_if_id | egress_if_id | + +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ + +*/ + +#define TRACE_TYPE_IF 0x03 +typedef struct +{ + u32 ttl_node_id; + u16 ingress_if; + u16 egress_if; +} ioam_trace_if_t; + +/* + 0x00001001 iOAM-trace-type is 0x00001001 then the format is: + + 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 + +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ + | Hop_Lim | node_id | + +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ + + timestamp + + +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ + +*/ + +#define TRACE_TYPE_TS 0x09 +typedef struct +{ + u32 ttl_node_id; + u32 timestamp; +} ioam_trace_ts_t; + +/* + 0x00010001 iOAM-trace-type is 0x00010001 then the format is: + + + 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 + +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ + | Hop_Lim | node_id | + +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ + | app_data | + +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ + +*/ + + +#define TRACE_TYPE_APP 0x11 +typedef struct +{ + u32 ttl_node_id; + u32 app_data; +} ioam_trace_app_t; + +/* + + 0x00011001 iOAM-trace-type is 0x00011001 then the format is: + + 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 + +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ + | Hop_Lim | node_id | + +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ + + timestamp + + +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ + | app_data | + +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ +*/ + +#define TRACE_TYPE_TS_APP 0x19 +typedef struct +{ + u32 ttl_node_id; + u32 timestamp; + u32 app_data; +} ioam_trace_ts_app_t; + + + +static inline u8 +fetch_trace_data_size (u8 trace_type) +{ + u8 trace_data_size = 0; + + if (trace_type == TRACE_TYPE_IF_TS_APP) + trace_data_size = sizeof (ioam_trace_if_ts_app_t); + else if (trace_type == TRACE_TYPE_IF) + trace_data_size = sizeof (ioam_trace_if_t); + else if (trace_type == TRACE_TYPE_TS) + trace_data_size = sizeof (ioam_trace_ts_t); + else if (trace_type == TRACE_TYPE_APP) + trace_data_size = sizeof (ioam_trace_app_t); + else if (trace_type == TRACE_TYPE_TS_APP) + trace_data_size = sizeof (ioam_trace_ts_app_t); + + return trace_data_size; +} + +int ioam_trace_get_sizeof_handler (u32 * result); +int ip6_trace_profile_setup (void); +int ip6_trace_profile_cleanup (void); + +#define TSP_SECONDS 0 +#define TSP_MILLISECONDS 1 +#define TSP_MICROSECONDS 2 +#define TSP_NANOSECONDS 3 + +#endif + +/* + * fd.io coding-style-patch-verification: ON + * + * Local Variables: + * eval: (c-set-style "gnu") + * End: + */ diff --git a/src/plugins/ioam/lib-vxlan-gpe/ioam_decap.c b/src/plugins/ioam/lib-vxlan-gpe/ioam_decap.c new file mode 100644 index 00000000..fd308657 --- /dev/null +++ b/src/plugins/ioam/lib-vxlan-gpe/ioam_decap.c @@ -0,0 +1,223 @@ +/* + * Copyright (c) 2015 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include + +/* Statistics (not really errors) */ +#define foreach_vxlan_gpe_decap_ioam_v4_error \ +_(DECAPSULATED, "good packets decapsulated") + +static char *vxlan_gpe_decap_ioam_v4_error_strings[] = { +#define _(sym,string) string, + foreach_vxlan_gpe_decap_ioam_v4_error +#undef _ +}; + +typedef enum +{ +#define _(sym,str) VXLAN_GPE_DECAP_IOAM_V4_ERROR_##sym, + foreach_vxlan_gpe_decap_ioam_v4_error +#undef _ + VXLAN_GPE_DECAP_IOAM_V4_N_ERROR, +} vxlan_gpe_decap_ioam_v4_error_t; + + +always_inline void +vxlan_gpe_decap_ioam_v4_two_inline (vlib_main_t * vm, + vlib_node_runtime_t * node, + vxlan_gpe_main_t * ngm, + vlib_buffer_t * b0, vlib_buffer_t * b1, + u32 * next0, u32 * next1) +{ + vxlan_gpe_ioam_main_t *hm = &vxlan_gpe_ioam_main; + + next0[0] = next1[0] = hm->decap_v4_next_override; + vxlan_gpe_encap_decap_ioam_v4_one_inline (vm, node, b0, &next0[0], + VXLAN_GPE_DECAP_IOAM_V4_NEXT_DROP, + 0 /* use_adj */ ); + vxlan_gpe_encap_decap_ioam_v4_one_inline (vm, node, b1, &next0[1], + VXLAN_GPE_DECAP_IOAM_V4_NEXT_DROP, + 0 /* use_adj */ ); +} + + + +static uword +vxlan_gpe_decap_ioam (vlib_main_t * vm, + vlib_node_runtime_t * node, + vlib_frame_t * from_frame, u8 is_ipv6) +{ + u32 n_left_from, next_index, *from, *to_next; + vxlan_gpe_main_t *ngm = &vxlan_gpe_main; + vxlan_gpe_ioam_main_t *hm = &vxlan_gpe_ioam_main; + + from = vlib_frame_vector_args (from_frame); + n_left_from = from_frame->n_vectors; + + next_index = node->cached_next_index; + + while (n_left_from > 0) + { + u32 n_left_to_next; + + vlib_get_next_frame (vm, node, next_index, to_next, n_left_to_next); + + while (n_left_from >= 4 && n_left_to_next >= 2) + { + u32 bi0, bi1; + vlib_buffer_t *b0, *b1; + u32 next0, next1; + + next0 = next1 = hm->decap_v4_next_override; + + /* Prefetch next iteration. */ + { + vlib_buffer_t *p2, *p3; + + p2 = vlib_get_buffer (vm, from[2]); + p3 = vlib_get_buffer (vm, from[3]); + + vlib_prefetch_buffer_header (p2, LOAD); + vlib_prefetch_buffer_header (p3, LOAD); + + CLIB_PREFETCH (p2->data, 2 * CLIB_CACHE_LINE_BYTES, LOAD); + CLIB_PREFETCH (p3->data, 2 * CLIB_CACHE_LINE_BYTES, LOAD); + } + + bi0 = from[0]; + bi1 = from[1]; + to_next[0] = bi0; + to_next[1] = bi1; + from += 2; + to_next += 2; + n_left_to_next -= 2; + n_left_from -= 2; + + b0 = vlib_get_buffer (vm, bi0); + b1 = vlib_get_buffer (vm, bi1); + + + vlib_buffer_advance (b0, + -(word) (sizeof (udp_header_t) + + sizeof (ip4_header_t) + + sizeof (vxlan_gpe_header_t))); + vlib_buffer_advance (b1, + -(word) (sizeof (udp_header_t) + + sizeof (ip4_header_t) + + sizeof (vxlan_gpe_header_t))); + + vxlan_gpe_decap_ioam_v4_two_inline (vm, node, ngm, b0, b1, + &next0, &next1); + + + vlib_validate_buffer_enqueue_x2 (vm, node, next_index, to_next, + n_left_to_next, bi0, bi1, next0, + next1); + + if (PREDICT_FALSE (b0->flags & VLIB_BUFFER_IS_TRACED)) + { + vxlan_gpe_ioam_v4_trace_t *tr = vlib_add_trace (vm, node, b0, + sizeof (*tr)); + } + } + + while (n_left_from > 0 && n_left_to_next > 0) + { + u32 bi0; + vlib_buffer_t *b0; + u32 next0 = hm->decap_v4_next_override; + + bi0 = from[0]; + to_next[0] = bi0; + from += 1; + to_next += 1; + n_left_from -= 1; + n_left_to_next -= 1; + + b0 = vlib_get_buffer (vm, bi0); + + + vlib_buffer_advance (b0, + -(word) (sizeof (udp_header_t) + + sizeof (ip4_header_t) + + sizeof (vxlan_gpe_header_t))); + + next0 = hm->decap_v4_next_override; + vxlan_gpe_encap_decap_ioam_v4_one_inline (vm, node, b0, + &next0, + VXLAN_GPE_DECAP_IOAM_V4_NEXT_DROP, + 0 /* use_adj */ ); + + if (PREDICT_FALSE (b0->flags & VLIB_BUFFER_IS_TRACED)) + { + vxlan_gpe_ioam_v4_trace_t *tr = vlib_add_trace (vm, node, b0, + sizeof (*tr)); + } + vlib_validate_buffer_enqueue_x1 (vm, node, next_index, to_next, + n_left_to_next, bi0, next0); + } + + vlib_put_next_frame (vm, node, next_index, n_left_to_next); + } + + return from_frame->n_vectors; +} + + +static uword +vxlan_gpe_decap_ioam_v4 (vlib_main_t * vm, + vlib_node_runtime_t * node, + vlib_frame_t * from_frame) +{ + return vxlan_gpe_decap_ioam (vm, node, from_frame, 0); +} + + +/* *INDENT-OFF* */ +VLIB_REGISTER_NODE (vxlan_gpe_decap_ioam_v4_node) = { + .function = vxlan_gpe_decap_ioam_v4, + .name = "vxlan-gpe-decap-ioam-v4", + .vector_size = sizeof (u32), + .format_trace = format_vxlan_gpe_ioam_v4_trace, + .type = VLIB_NODE_TYPE_INTERNAL, + + .n_errors = ARRAY_LEN(vxlan_gpe_decap_ioam_v4_error_strings), + .error_strings = vxlan_gpe_decap_ioam_v4_error_strings, + + .n_next_nodes = VXLAN_GPE_DECAP_IOAM_V4_N_NEXT, + + .next_nodes = { + [VXLAN_GPE_DECAP_IOAM_V4_NEXT_POP] = "vxlan-gpe-pop-ioam-v4", + [VXLAN_GPE_DECAP_IOAM_V4_NEXT_DROP] = "error-drop", + }, +}; +/* *INDENT-ON* */ + + +/* + * fd.io coding-style-patch-verification: ON + * + * Local Variables: + * eval: (c-set-style "gnu") + * End: + */ diff --git a/src/plugins/ioam/lib-vxlan-gpe/ioam_encap.c b/src/plugins/ioam/lib-vxlan-gpe/ioam_encap.c new file mode 100644 index 00000000..4b18bfea --- /dev/null +++ b/src/plugins/ioam/lib-vxlan-gpe/ioam_encap.c @@ -0,0 +1,194 @@ +/* + * Copyright (c) 2015 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ +#include +#include +#include +#include +#include +#include +#include +#include +#include + +/* Statistics (not really errors) */ +#define foreach_vxlan_gpe_encap_ioam_v4_error \ +_(ENCAPSULATED, "good packets encapsulated") + +static char *vxlan_gpe_encap_ioam_v4_error_strings[] = { +#define _(sym,string) string, + foreach_vxlan_gpe_encap_ioam_v4_error +#undef _ +}; + +typedef enum +{ +#define _(sym,str) VXLAN_GPE_ENCAP_IOAM_V4_ERROR_##sym, + foreach_vxlan_gpe_encap_ioam_v4_error +#undef _ + VXLAN_GPE_ENCAP_IOAM_V4_N_ERROR, +} vxlan_gpe_encap_ioam_v4_error_t; + +typedef enum +{ + VXLAN_GPE_ENCAP_IOAM_V4_NEXT_IP4_LOOKUP, + VXLAN_GPE_ENCAP_IOAM_V4_NEXT_DROP, + VXLAN_GPE_ENCAP_IOAM_V4_N_NEXT +} vxlan_gpe_encap_ioam_v4_next_t; + + +always_inline void +vxlan_gpe_encap_ioam_v4_two_inline (vlib_main_t * vm, + vlib_node_runtime_t * node, + vxlan_gpe_main_t * ngm, + vlib_buffer_t * b0, vlib_buffer_t * b1, + u32 * next0, u32 * next1) +{ + next0[0] = next1[0] = VXLAN_GPE_ENCAP_IOAM_V4_NEXT_IP4_LOOKUP; + vxlan_gpe_encap_decap_ioam_v4_one_inline (vm, node, b0, next0, + VXLAN_GPE_ENCAP_IOAM_V4_NEXT_DROP, + 0 /* use_adj */ ); + vxlan_gpe_encap_decap_ioam_v4_one_inline (vm, node, b1, next1, + VXLAN_GPE_ENCAP_IOAM_V4_NEXT_DROP, + 0 /* use_adj */ ); +} + + +static uword +vxlan_gpe_encap_ioam_v4 (vlib_main_t * vm, + vlib_node_runtime_t * node, + vlib_frame_t * from_frame) +{ + u32 n_left_from, next_index, *from, *to_next; + vxlan_gpe_main_t *ngm = &vxlan_gpe_main; + + from = vlib_frame_vector_args (from_frame); + n_left_from = from_frame->n_vectors; + + next_index = node->cached_next_index; + + while (n_left_from > 0) + { + u32 n_left_to_next; + + vlib_get_next_frame (vm, node, next_index, to_next, n_left_to_next); + + while (n_left_from >= 4 && n_left_to_next >= 2) + { + u32 bi0, bi1; + vlib_buffer_t *b0, *b1; + u32 next0, next1; + + next0 = next1 = VXLAN_GPE_ENCAP_IOAM_V4_NEXT_IP4_LOOKUP; + + /* Prefetch next iteration. */ + { + vlib_buffer_t *p2, *p3; + + p2 = vlib_get_buffer (vm, from[2]); + p3 = vlib_get_buffer (vm, from[3]); + + vlib_prefetch_buffer_header (p2, LOAD); + vlib_prefetch_buffer_header (p3, LOAD); + + CLIB_PREFETCH (p2->data, 2 * CLIB_CACHE_LINE_BYTES, LOAD); + CLIB_PREFETCH (p3->data, 2 * CLIB_CACHE_LINE_BYTES, LOAD); + } + + bi0 = from[0]; + bi1 = from[1]; + to_next[0] = bi0; + to_next[1] = bi1; + from += 2; + to_next += 2; + n_left_to_next -= 2; + n_left_from -= 2; + + b0 = vlib_get_buffer (vm, bi0); + b1 = vlib_get_buffer (vm, bi1); + + vxlan_gpe_encap_ioam_v4_two_inline (vm, node, ngm, b0, b1, + &next0, &next1); + + + vlib_validate_buffer_enqueue_x2 (vm, node, next_index, to_next, + n_left_to_next, bi0, bi1, next0, + next1); + } + + while (n_left_from > 0 && n_left_to_next > 0) + { + u32 bi0; + vlib_buffer_t *b0; + u32 next0 = VXLAN_GPE_ENCAP_IOAM_V4_NEXT_IP4_LOOKUP; + + bi0 = from[0]; + to_next[0] = bi0; + from += 1; + to_next += 1; + n_left_from -= 1; + n_left_to_next -= 1; + + b0 = vlib_get_buffer (vm, bi0); + + vxlan_gpe_encap_decap_ioam_v4_one_inline (vm, node, b0, + &next0, + VXLAN_GPE_ENCAP_IOAM_V4_NEXT_DROP, + 0 /* use_adj */ ); + + if (PREDICT_FALSE (b0->flags & VLIB_BUFFER_IS_TRACED)) + { + vxlan_gpe_ioam_v4_trace_t *tr = vlib_add_trace (vm, node, b0, + sizeof (*tr)); + } + vlib_validate_buffer_enqueue_x1 (vm, node, next_index, to_next, + n_left_to_next, bi0, next0); + } + + vlib_put_next_frame (vm, node, next_index, n_left_to_next); + } + + return from_frame->n_vectors; +} + + + +/* *INDENT-OFF* */ +VLIB_REGISTER_NODE (vxlan_gpe_encap_ioam_v4_node) = { + .function = vxlan_gpe_encap_ioam_v4, + .name = "vxlan-gpe-encap-ioam-v4", + .vector_size = sizeof (u32), + .format_trace = format_vxlan_gpe_ioam_v4_trace, + .type = VLIB_NODE_TYPE_INTERNAL, + + .n_errors = ARRAY_LEN(vxlan_gpe_encap_ioam_v4_error_strings), + .error_strings = vxlan_gpe_encap_ioam_v4_error_strings, + + .n_next_nodes = VXLAN_GPE_ENCAP_IOAM_V4_N_NEXT, + + .next_nodes = { + [VXLAN_GPE_ENCAP_IOAM_V4_NEXT_IP4_LOOKUP] = "ip4-lookup", + [VXLAN_GPE_ENCAP_IOAM_V4_NEXT_DROP] = "error-drop", + }, +}; +/* *INDENT-ON* */ + + +/* + * fd.io coding-style-patch-verification: ON + * + * Local Variables: + * eval: (c-set-style "gnu") + * End: + */ diff --git a/src/plugins/ioam/lib-vxlan-gpe/ioam_pop.c b/src/plugins/ioam/lib-vxlan-gpe/ioam_pop.c new file mode 100644 index 00000000..55c33b14 --- /dev/null +++ b/src/plugins/ioam/lib-vxlan-gpe/ioam_pop.c @@ -0,0 +1,353 @@ +/* + * Copyright (c) 2015 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ +#include +#include +#include +#include +#include +#include +#include + +/* Statistics (not really errors) */ +#define foreach_vxlan_gpe_pop_ioam_v4_error \ +_(POPPED, "good packets popped") + +static char *vxlan_gpe_pop_ioam_v4_error_strings[] = { +#define _(sym,string) string, + foreach_vxlan_gpe_pop_ioam_v4_error +#undef _ +}; + +typedef enum +{ +#define _(sym,str) VXLAN_GPE_POP_IOAM_V4_ERROR_##sym, + foreach_vxlan_gpe_pop_ioam_v4_error +#undef _ + VXLAN_GPE_POP_IOAM_V4_N_ERROR, +} vxlan_gpe_pop_ioam_v4_error_t; + +typedef struct +{ + ioam_trace_t fmt_trace; +} vxlan_gpe_pop_ioam_v4_trace_t; + + +u8 * +format_vxlan_gpe_pop_ioam_v4_trace (u8 * s, va_list * args) +{ + CLIB_UNUSED (vlib_main_t * vm) = va_arg (*args, vlib_main_t *); + CLIB_UNUSED (vlib_node_t * node) = va_arg (*args, vlib_node_t *); + vxlan_gpe_pop_ioam_v4_trace_t *t1 + = va_arg (*args, vxlan_gpe_pop_ioam_v4_trace_t *); + ioam_trace_t *t = &(t1->fmt_trace); + vxlan_gpe_ioam_option_t *fmt_trace0; + vxlan_gpe_ioam_option_t *opt0, *limit0; + vxlan_gpe_ioam_main_t *hm = &vxlan_gpe_ioam_main; + + u8 type0; + + fmt_trace0 = (vxlan_gpe_ioam_option_t *) t->option_data; + + s = format (s, "VXLAN_GPE_IOAM_POP: next_index %d len %d traced %d", + t->next_index, fmt_trace0->length, t->trace_len); + + opt0 = (vxlan_gpe_ioam_option_t *) (fmt_trace0 + 1); + limit0 = (vxlan_gpe_ioam_option_t *) ((u8 *) fmt_trace0) + t->trace_len; + + while (opt0 < limit0) + { + type0 = opt0->type; + switch (type0) + { + case 0: /* Pad, just stop */ + opt0 = (vxlan_gpe_ioam_option_t *) ((u8 *) opt0) + 1; + break; + + default: + if (hm->trace[type0]) + { + s = (*hm->trace[type0]) (s, opt0); + } + else + { + s = + format (s, "\n unrecognized option %d length %d", type0, + opt0->length); + } + opt0 = + (vxlan_gpe_ioam_option_t *) (((u8 *) opt0) + opt0->length + + sizeof (vxlan_gpe_ioam_option_t)); + break; + } + } + + return s; +} + +always_inline void +vxlan_gpe_ioam_pop_v4 (vlib_main_t * vm, vlib_node_runtime_t * node, + vlib_buffer_t * b0) +{ + ip4_header_t *ip0; + udp_header_t *udp_hdr0; + vxlan_gpe_header_t *gpe_hdr0; + vxlan_gpe_ioam_hdr_t *gpe_ioam0; + + ip0 = vlib_buffer_get_current (b0); + + udp_hdr0 = (udp_header_t *) (ip0 + 1); + gpe_hdr0 = (vxlan_gpe_header_t *) (udp_hdr0 + 1); + gpe_ioam0 = (vxlan_gpe_ioam_hdr_t *) (gpe_hdr0 + 1); + + /* Pop the iOAM data */ + vlib_buffer_advance (b0, + (word) (sizeof (udp_header_t) + + sizeof (ip4_header_t) + + sizeof (vxlan_gpe_header_t) + + gpe_ioam0->length)); + + return; +} + + + +always_inline void +vxlan_gpe_pop_ioam_v4_one_inline (vlib_main_t * vm, + vlib_node_runtime_t * node, + vxlan_gpe_main_t * ngm, + vlib_buffer_t * b0, u32 * next0) +{ + CLIB_UNUSED (ip4_header_t * ip0); + CLIB_UNUSED (udp_header_t * udp_hdr0); + CLIB_UNUSED (vxlan_gpe_header_t * gpe_hdr0); + CLIB_UNUSED (vxlan_gpe_ioam_hdr_t * gpe_ioam0); + CLIB_UNUSED (vxlan_gpe_ioam_option_t * opt0); + CLIB_UNUSED (vxlan_gpe_ioam_option_t * limit0); + vxlan_gpe_ioam_main_t *hm = &vxlan_gpe_ioam_main; + + + /* Pop the iOAM header */ + ip0 = vlib_buffer_get_current (b0); + udp_hdr0 = (udp_header_t *) (ip0 + 1); + gpe_hdr0 = (vxlan_gpe_header_t *) (udp_hdr0 + 1); + gpe_ioam0 = (vxlan_gpe_ioam_hdr_t *) (gpe_hdr0 + 1); + opt0 = (vxlan_gpe_ioam_option_t *) (gpe_ioam0 + 1); + limit0 = (vxlan_gpe_ioam_option_t *) ((u8 *) gpe_ioam0 + gpe_ioam0->length); + + /* + * Basic validity checks + */ + if (gpe_ioam0->length > clib_net_to_host_u16 (ip0->length)) + { + next0[0] = VXLAN_GPE_INPUT_NEXT_DROP; + goto trace00; + } + + /* Scan the set of h-b-h options, process ones that we understand */ + while (opt0 < limit0) + { + u8 type0; + type0 = opt0->type; + switch (type0) + { + case 0: /* Pad1 */ + opt0 = (vxlan_gpe_ioam_option_t *) ((u8 *) opt0) + 1; + continue; + case 1: /* PadN */ + break; + default: + if (hm->pop_options[type0]) + { + if ((*hm->pop_options[type0]) (ip0, opt0) < 0) + { + next0[0] = VXLAN_GPE_INPUT_NEXT_DROP; + goto trace00; + } + } + break; + } + opt0 = + (vxlan_gpe_ioam_option_t *) (((u8 *) opt0) + opt0->length + + sizeof (vxlan_gpe_ioam_hdr_t)); + } + + + next0[0] = + (gpe_ioam0->protocol < VXLAN_GPE_PROTOCOL_MAX) ? + ngm-> + decap_next_node_list[gpe_ioam0->protocol] : VXLAN_GPE_INPUT_NEXT_DROP; + +trace00: + if (PREDICT_FALSE (b0->flags & VLIB_BUFFER_IS_TRACED)) + { + vxlan_gpe_pop_ioam_v4_trace_t *t = + vlib_add_trace (vm, node, b0, sizeof (*t)); + u32 trace_len = gpe_ioam0->length; + t->fmt_trace.next_index = next0[0]; + /* Capture the h-b-h option verbatim */ + trace_len = + trace_len < + ARRAY_LEN (t->fmt_trace. + option_data) ? trace_len : ARRAY_LEN (t->fmt_trace. + option_data); + t->fmt_trace.trace_len = trace_len; + clib_memcpy (&(t->fmt_trace.option_data), gpe_ioam0, trace_len); + } + + /* Remove the iOAM header inside the VxLAN-GPE header */ + vxlan_gpe_ioam_pop_v4 (vm, node, b0); + return; +} + +always_inline void +vxlan_gpe_pop_ioam_v4_two_inline (vlib_main_t * vm, + vlib_node_runtime_t * node, + vxlan_gpe_main_t * ngm, + vlib_buffer_t * b0, vlib_buffer_t * b1, + u32 * next0, u32 * next1) +{ + + vxlan_gpe_pop_ioam_v4_one_inline (vm, node, ngm, b0, next0); + vxlan_gpe_pop_ioam_v4_one_inline (vm, node, ngm, b1, next1); +} + + + +static uword +vxlan_gpe_pop_ioam (vlib_main_t * vm, + vlib_node_runtime_t * node, + vlib_frame_t * from_frame, u8 is_ipv6) +{ + u32 n_left_from, next_index, *from, *to_next; + vxlan_gpe_main_t *ngm = &vxlan_gpe_main; + + from = vlib_frame_vector_args (from_frame); + n_left_from = from_frame->n_vectors; + + next_index = node->cached_next_index; + + while (n_left_from > 0) + { + u32 n_left_to_next; + + vlib_get_next_frame (vm, node, next_index, to_next, n_left_to_next); + + while (n_left_from >= 4 && n_left_to_next >= 2) + { + u32 bi0, bi1; + vlib_buffer_t *b0, *b1; + u32 next0, next1; + + /* Prefetch next iteration. */ + { + vlib_buffer_t *p2, *p3; + + p2 = vlib_get_buffer (vm, from[2]); + p3 = vlib_get_buffer (vm, from[3]); + + vlib_prefetch_buffer_header (p2, LOAD); + vlib_prefetch_buffer_header (p3, LOAD); + + CLIB_PREFETCH (p2->data, 2 * CLIB_CACHE_LINE_BYTES, LOAD); + CLIB_PREFETCH (p3->data, 2 * CLIB_CACHE_LINE_BYTES, LOAD); + } + + bi0 = from[0]; + bi1 = from[1]; + to_next[0] = bi0; + to_next[1] = bi1; + from += 2; + to_next += 2; + n_left_to_next -= 2; + n_left_from -= 2; + + b0 = vlib_get_buffer (vm, bi0); + b1 = vlib_get_buffer (vm, bi1); + + vxlan_gpe_pop_ioam_v4_two_inline (vm, node, ngm, b0, b1, &next0, + &next1); + + + vlib_validate_buffer_enqueue_x2 (vm, node, next_index, to_next, + n_left_to_next, bi0, bi1, next0, + next1); + } + + while (n_left_from > 0 && n_left_to_next > 0) + { + u32 bi0; + vlib_buffer_t *b0; + u32 next0; + + bi0 = from[0]; + to_next[0] = bi0; + from += 1; + to_next += 1; + n_left_from -= 1; + n_left_to_next -= 1; + + b0 = vlib_get_buffer (vm, bi0); + + vxlan_gpe_pop_ioam_v4_one_inline (vm, node, ngm, b0, &next0); + + + vlib_validate_buffer_enqueue_x1 (vm, node, next_index, to_next, + n_left_to_next, bi0, next0); + } + + vlib_put_next_frame (vm, node, next_index, n_left_to_next); + } + + return from_frame->n_vectors; +} + + +static uword +vxlan_gpe_pop_ioam_v4 (vlib_main_t * vm, + vlib_node_runtime_t * node, vlib_frame_t * from_frame) +{ + return vxlan_gpe_pop_ioam (vm, node, from_frame, 0); +} + +/* *INDENT-OFF* */ +VLIB_REGISTER_NODE (vxlan_gpe_pop_ioam_v4_node) = { + .function = vxlan_gpe_pop_ioam_v4, + .name = "vxlan-gpe-pop-ioam-v4", + .vector_size = sizeof (u32), + .format_trace = format_vxlan_gpe_pop_ioam_v4_trace, + .type = VLIB_NODE_TYPE_INTERNAL, + + .n_errors = ARRAY_LEN(vxlan_gpe_pop_ioam_v4_error_strings), + .error_strings = vxlan_gpe_pop_ioam_v4_error_strings, + + .n_next_nodes = VXLAN_GPE_INPUT_N_NEXT, + + .next_nodes = { +#define _(s,n) [VXLAN_GPE_INPUT_NEXT_##s] = n, + foreach_vxlan_gpe_input_next +#undef _ + }, +}; +/* *INDENT-ON* */ + + + +/* + * fd.io coding-style-patch-verification: ON + * + * Local Variables: + * eval: (c-set-style "gnu") + * End: + */ diff --git a/src/plugins/ioam/lib-vxlan-gpe/ioam_transit.c b/src/plugins/ioam/lib-vxlan-gpe/ioam_transit.c new file mode 100644 index 00000000..b42c357c --- /dev/null +++ b/src/plugins/ioam/lib-vxlan-gpe/ioam_transit.c @@ -0,0 +1,188 @@ + /* + * Copyright (c) 2015 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include + +/* Statistics (not really errors) */ +#define foreach_vxlan_gpe_transit_ioam_error \ +_(ENCAPSULATED, "good packets encapsulated") + +static char *vxlan_gpe_transit_ioam_error_strings[] = { +#define _(sym,string) string, + foreach_vxlan_gpe_transit_ioam_error +#undef _ +}; + +typedef enum +{ +#define _(sym,str) VXLAN_GPE_TRANSIT_IOAM_ERROR_##sym, + foreach_vxlan_gpe_transit_ioam_error +#undef _ + VXLAN_GPE_TRANSIT_IOAM_N_ERROR, +} vxlan_gpe_transit_ioam_error_t; + +typedef enum +{ + VXLAN_GPE_TRANSIT_IOAM_NEXT_OUTPUT, + VXLAN_GPE_TRANSIT_IOAM_NEXT_DROP, + VXLAN_GPE_TRANSIT_IOAM_N_NEXT +} vxlan_gpe_transit_ioam_next_t; + + +/* *INDENT-OFF* */ +VNET_FEATURE_INIT (vxlan_gpe_transit_ioam, static) = +{ + .arc_name = "ip4-output", + .node_name = "vxlan-gpe-transit-ioam", + .runs_before = VNET_FEATURES ("interface-output"), +}; +/* *INDENT-ON* */ + +static uword +vxlan_gpe_transit_ioam (vlib_main_t * vm, + vlib_node_runtime_t * node, vlib_frame_t * from_frame) +{ + u32 n_left_from, next_index, *from, *to_next; + + from = vlib_frame_vector_args (from_frame); + n_left_from = from_frame->n_vectors; + + next_index = node->cached_next_index; + + while (n_left_from > 0) + { + u32 n_left_to_next; + + vlib_get_next_frame (vm, node, next_index, to_next, n_left_to_next); + + + while (n_left_from > 0 && n_left_to_next > 0) + { + u32 bi0; + vlib_buffer_t *b0; + u32 next0 = VXLAN_GPE_TRANSIT_IOAM_NEXT_OUTPUT; + + bi0 = from[0]; + to_next[0] = bi0; + from += 1; + to_next += 1; + n_left_from -= 1; + n_left_to_next -= 1; + ip4_header_t *ip0; + u32 iph_offset = 0; + + b0 = vlib_get_buffer (vm, bi0); + iph_offset = vnet_buffer (b0)->ip.save_rewrite_length; + ip0 = (ip4_header_t *) ((u8 *) vlib_buffer_get_current (b0) + + iph_offset); + + /* just forward non ipv4 packets */ + if (PREDICT_FALSE + ((ip0->ip_version_and_header_length & 0xF0) == 0x40)) + { + /* ipv4 packets */ + udp_header_t *udp_hdr0 = (udp_header_t *) (ip0 + 1); + if (PREDICT_FALSE + ((ip0->protocol == IP_PROTOCOL_UDP) && + (clib_net_to_host_u16 (udp_hdr0->dst_port) == + UDP_DST_PORT_vxlan_gpe))) + { + + /* Check the iOAM header */ + vxlan_gpe_header_t *gpe_hdr0 = + (vxlan_gpe_header_t *) (udp_hdr0 + 1); + + if (PREDICT_FALSE + (gpe_hdr0->protocol == VXLAN_GPE_PROTOCOL_IOAM)) + { + uword *t = NULL; + vxlan_gpe_ioam_main_t *hm = &vxlan_gpe_ioam_main; + fib_prefix_t key4; + memset (&key4, 0, sizeof (key4)); + key4.fp_proto = FIB_PROTOCOL_IP4; + key4.fp_addr.ip4.as_u32 = ip0->dst_address.as_u32; + t = hash_get_mem (hm->dst_by_ip4, &key4); + if (t) + { + + + vlib_buffer_advance (b0, + (word) (sizeof + (ethernet_header_t))); + vxlan_gpe_encap_decap_ioam_v4_one_inline (vm, node, + b0, + &next0, + VXLAN_GPE_TRANSIT_IOAM_NEXT_DROP, + 1 + /* use_adj */ + ); + vlib_buffer_advance (b0, + -(word) (sizeof + (ethernet_header_t))); + } + } + } + } + + vlib_validate_buffer_enqueue_x1 (vm, node, next_index, to_next, + n_left_to_next, bi0, next0); + } + + vlib_put_next_frame (vm, node, next_index, n_left_to_next); + } + + return from_frame->n_vectors; +} + +/* *INDENT-OFF* */ +VLIB_REGISTER_NODE (vxlan_gpe_transit_ioam_node) = { + .function = vxlan_gpe_transit_ioam, + .name = "vxlan-gpe-transit-ioam", + .vector_size = sizeof (u32), + .format_trace = format_vxlan_gpe_ioam_v4_trace, + .type = VLIB_NODE_TYPE_INTERNAL, + + .n_errors = ARRAY_LEN(vxlan_gpe_transit_ioam_error_strings), + .error_strings = vxlan_gpe_transit_ioam_error_strings, + + .n_next_nodes = VXLAN_GPE_TRANSIT_IOAM_N_NEXT, + + .next_nodes = { + [VXLAN_GPE_TRANSIT_IOAM_NEXT_OUTPUT] = "interface-output", + [VXLAN_GPE_TRANSIT_IOAM_NEXT_DROP] = "error-drop", + }, + +}; +/* *INDENT-ON* */ + + +/* + * fd.io coding-style-patch-verification: ON + * + * Local Variables: + * eval: (c-set-style "gnu") + * End: + */ diff --git a/src/plugins/ioam/lib-vxlan-gpe/ioam_vxlan_gpe.api b/src/plugins/ioam/lib-vxlan-gpe/ioam_vxlan_gpe.api new file mode 100644 index 00000000..056529a4 --- /dev/null +++ b/src/plugins/ioam/lib-vxlan-gpe/ioam_vxlan_gpe.api @@ -0,0 +1,181 @@ +/* Hey Emacs use -*- mode: C -*- */ +/* + * Copyright (c) 2016 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + + +/** \brief iOAM Over VxLAN-GPE - Set iOAM transport for VxLAN-GPE + @param client_index - opaque cookie to identify the sender + @param context - sender context, to match reply w/ request + @param id - profile id + @param trace_ppc - Trace PPC (none/encap/decap) + @param pow_enable - Proof of Work enabled or not flag + @param trace_enable - iOAM Trace enabled or not flag + +*/ +define vxlan_gpe_ioam_enable { + u32 client_index; + u32 context; + u16 id; + u8 trace_ppc; + u8 pow_enable; + u8 trace_enable; +}; + +/** \brief iOAM Over VxLAN-GPE - Set iOAM transport for VXLAN-GPE reply + @param context - sender context, to match reply w/ request + @param retval - return value for request +*/ +define vxlan_gpe_ioam_enable_reply { + u32 context; + i32 retval; +}; + + +/** \brief iOAM for VxLAN-GPE disable + @param client_index - opaque cookie to identify the sender + @param context - sender context, to match reply w/ request + @param id - profile id +*/ +define vxlan_gpe_ioam_disable +{ + u32 client_index; + u32 context; + u16 id; +}; + +/** \brief vxlan_gpe_ioam disable response + @param context - sender context, to match reply w/ request + @param retval - return value for request +*/ +define vxlan_gpe_ioam_disable_reply +{ + u32 context; + i32 retval; +}; + +/** \brief Enable iOAM for a VNI (VXLAN-GPE) + @param client_index - opaque cookie to identify the sender + @param context - sender context, to match reply w/ request + @param vni - VXLAN-GPE VNI + @param local - IPv4/6 Address of the local VTEP + @param remote - IPv4/6 Address of the remote VTEP + +*/ +define vxlan_gpe_ioam_vni_enable { + u32 client_index; + u32 context; + u32 vni; + u8 local[16]; + u8 remote[16]; + u8 is_ipv6; +}; + +/** \brief Reply to enable iOAM for a VNI (VXLAN-GPE) + @param client_index - opaque cookie to identify the sender + @param context - sender context, to match reply w/ request + @param retval - return value for request + +*/ +define vxlan_gpe_ioam_vni_enable_reply { + u32 client_index; + u32 context; + i32 retval; +}; + +/** \brief Disable iOAM for a VNI (VXLAN-GPE) + @param client_index - opaque cookie to identify the sender + @param context - sender context, to match reply w/ request + @param vni - VXLAN-GPE VNI + @param local - IPv4/6 Address of the local VTEP + @param remote - IPv4/6 Address of the remote VTEP + +*/ +define vxlan_gpe_ioam_vni_disable { + u32 client_index; + u32 context; + u32 vni; + u8 local[16]; + u8 remote[16]; + u8 is_ipv6; +}; + +/** \brief Reply to disable iOAM for a VNI (VXLAN-GPE) + @param client_index - opaque cookie to identify the sender + @param context - sender context, to match reply w/ request + @param retval - return value for request + +*/ +define vxlan_gpe_ioam_vni_disable_reply { + u32 client_index; + u32 context; + i32 retval; +}; + + +/** \brief Enable iOAM for a VXLAN-GPE transit + @param client_index - opaque cookie to identify the sender + @param context - sender context, to match reply w/ request + @param dst_addr - IPv4/6 Address of the local VTEP + @param outer_fib_index- FIB index + +*/ +define vxlan_gpe_ioam_transit_enable { + u32 client_index; + u32 context; + u32 outer_fib_index; + u8 dst_addr[16]; + u8 is_ipv6; +}; + +/** \brief Reply to enable iOAM for VXLAN-GPE transit + @param client_index - opaque cookie to identify the sender + @param context - sender context, to match reply w/ request + @param retval - return value for request + +*/ +define vxlan_gpe_ioam_transit_enable_reply { + u32 client_index; + u32 context; + i32 retval; +}; + +/** \brief Disable iOAM for VXLAN-GPE transit + @param client_index - opaque cookie to identify the sender + @param context - sender context, to match reply w/ request + @param dst_addr - IPv4/6 Address of the local VTEP + @param outer_fib_index- FIB index + +*/ +define vxlan_gpe_ioam_transit_disable { + u32 client_index; + u32 context; + u32 outer_fib_index; + u8 dst_addr[16]; + u8 is_ipv6; +}; + +/** \brief Reply to disable iOAM for VXLAN-GPE transit + @param client_index - opaque cookie to identify the sender + @param context - sender context, to match reply w/ request + @param retval - return value for request + +*/ +define vxlan_gpe_ioam_transit_disable_reply { + u32 client_index; + u32 context; + i32 retval; +}; + + diff --git a/src/plugins/ioam/lib-vxlan-gpe/vxlan_gpe_all_api_h.h b/src/plugins/ioam/lib-vxlan-gpe/vxlan_gpe_all_api_h.h new file mode 100644 index 00000000..06fa0d2c --- /dev/null +++ b/src/plugins/ioam/lib-vxlan-gpe/vxlan_gpe_all_api_h.h @@ -0,0 +1,16 @@ +/* + * Copyright (c) 2016 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ +/* Include the generated file, see BUILT_SOURCES in Makefile.am */ +#include diff --git a/src/plugins/ioam/lib-vxlan-gpe/vxlan_gpe_api.c b/src/plugins/ioam/lib-vxlan-gpe/vxlan_gpe_api.c new file mode 100644 index 00000000..68752365 --- /dev/null +++ b/src/plugins/ioam/lib-vxlan-gpe/vxlan_gpe_api.c @@ -0,0 +1,378 @@ +/* + * Copyright (c) 2016 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ +/* + *------------------------------------------------------------------ + * vxlan_gpe_api.c - iOAM VxLAN-GPE related APIs to create + * and maintain profiles + *------------------------------------------------------------------ + */ + +#include +#include +#include + +#include +#include +#include + +/* define message IDs */ +#include + +/* define message structures */ +#define vl_typedefs +#include +#undef vl_typedefs + +/* define generated endian-swappers */ +#define vl_endianfun +#include +#undef vl_endianfun + +/* instantiate all the print functions we know about */ +#define vl_print(handle, ...) vlib_cli_output (handle, __VA_ARGS__) +#define vl_printfun +#include +#undef vl_printfun + +/* Get the API version number */ +#define vl_api_version(n,v) static u32 api_version=(v); +#include +#undef vl_api_version + +/* + * A handy macro to set up a message reply. + * Assumes that the following variables are available: + * mp - pointer to request message + * rmp - pointer to reply message type + * rv - return value + */ + +#define VXLAN_GPE_REPLY_MACRO(t) \ +do { \ + unix_shared_memory_queue_t * q = \ + vl_api_client_index_to_input_queue (mp->client_index); \ + if (!q) \ + return; \ + \ + rmp = vl_msg_api_alloc (sizeof (*rmp)); \ + rmp->_vl_msg_id = ntohs((t)+sm->msg_id_base); \ + rmp->context = mp->context; \ + rmp->retval = ntohl(rv); \ + \ + vl_msg_api_send_shmem (q, (u8 *)&rmp); \ +} while(0); + +/* *INDENT-OFF* */ +#define VXLAN_GPE_REPLY_MACRO2(t, body) \ +do { \ + unix_shared_memory_queue_t * q; \ + rv = vl_msg_api_pd_handler (mp, rv); \ + q = vl_api_client_index_to_input_queue (mp->client_index); \ + if (!q) \ + return; \ + \ + rmp = vl_msg_api_alloc (sizeof (*rmp)); \ + rmp->_vl_msg_id = ntohs((t)); \ + rmp->context = mp->context; \ + rmp->retval = ntohl(rv); \ + do {body;} while (0); \ + vl_msg_api_send_shmem (q, (u8 *)&rmp); \ +} while(0); +/* *INDENT-ON* */ + +/* List of message types that this plugin understands */ + +#define foreach_vxlan_gpe_plugin_api_msg \ +_(VXLAN_GPE_IOAM_ENABLE, vxlan_gpe_ioam_enable) \ +_(VXLAN_GPE_IOAM_DISABLE, vxlan_gpe_ioam_disable) \ +_(VXLAN_GPE_IOAM_VNI_ENABLE, vxlan_gpe_ioam_vni_enable) \ +_(VXLAN_GPE_IOAM_VNI_DISABLE, vxlan_gpe_ioam_vni_disable) \ +_(VXLAN_GPE_IOAM_TRANSIT_ENABLE, vxlan_gpe_ioam_transit_enable) \ +_(VXLAN_GPE_IOAM_TRANSIT_DISABLE, vxlan_gpe_ioam_transit_disable) \ + + +static void vl_api_vxlan_gpe_ioam_enable_t_handler + (vl_api_vxlan_gpe_ioam_enable_t * mp) +{ + int rv = 0; + vl_api_vxlan_gpe_ioam_enable_reply_t *rmp; + clib_error_t *error; + vxlan_gpe_ioam_main_t *sm = &vxlan_gpe_ioam_main; + + /* Ignoring the profile id as currently a single profile + * is supported */ + error = + vxlan_gpe_ioam_enable (mp->trace_enable, mp->pow_enable, mp->trace_ppc); + if (error) + { + clib_error_report (error); + rv = clib_error_get_code (error); + } + + VXLAN_GPE_REPLY_MACRO (VL_API_VXLAN_GPE_IOAM_ENABLE_REPLY); +} + +static void vl_api_vxlan_gpe_ioam_disable_t_handler + (vl_api_vxlan_gpe_ioam_disable_t * mp) +{ + int rv = 0; + vl_api_vxlan_gpe_ioam_disable_reply_t *rmp; + clib_error_t *error; + vxlan_gpe_ioam_main_t *sm = &vxlan_gpe_ioam_main; + + /* Ignoring the profile id as currently a single profile + * is supported */ + error = vxlan_gpe_ioam_disable (0, 0, 0); + if (error) + { + clib_error_report (error); + rv = clib_error_get_code (error); + } + + VXLAN_GPE_REPLY_MACRO (VL_API_VXLAN_GPE_IOAM_DISABLE_REPLY); +} + +static void vl_api_vxlan_gpe_ioam_vni_enable_t_handler + (vl_api_vxlan_gpe_ioam_vni_enable_t * mp) +{ + int rv = 0; + vl_api_vxlan_gpe_ioam_vni_enable_reply_t *rmp; + clib_error_t *error; + vxlan_gpe_ioam_main_t *sm = &vxlan_gpe_ioam_main; + vxlan4_gpe_tunnel_key_t key4; + uword *p = NULL; + vxlan_gpe_main_t *gm = &vxlan_gpe_main; + vxlan_gpe_tunnel_t *t = 0; + vxlan_gpe_ioam_main_t *hm = &vxlan_gpe_ioam_main; + u32 vni; + + + if (!mp->is_ipv6) + { + clib_memcpy (&key4.local, &mp->local, sizeof (key4.local)); + clib_memcpy (&key4.remote, &mp->remote, sizeof (key4.remote)); + vni = clib_net_to_host_u32 (mp->vni); + key4.vni = clib_host_to_net_u32 (vni << 8); + key4.pad = 0; + + p = hash_get_mem (gm->vxlan4_gpe_tunnel_by_key, &key4); + } + else + { + return; + } + + if (!p) + return; + + t = pool_elt_at_index (gm->tunnels, p[0]); + + error = vxlan_gpe_ioam_set (t, hm->has_trace_option, + hm->has_pot_option, + hm->has_ppc_option, mp->is_ipv6); + + + if (error) + { + clib_error_report (error); + rv = clib_error_get_code (error); + } + + VXLAN_GPE_REPLY_MACRO (VL_API_VXLAN_GPE_IOAM_VNI_ENABLE_REPLY); +} + + +static void vl_api_vxlan_gpe_ioam_vni_disable_t_handler + (vl_api_vxlan_gpe_ioam_vni_disable_t * mp) +{ + int rv = 0; + vl_api_vxlan_gpe_ioam_vni_enable_reply_t *rmp; + clib_error_t *error; + vxlan_gpe_ioam_main_t *sm = &vxlan_gpe_ioam_main; + vxlan4_gpe_tunnel_key_t key4; + uword *p = NULL; + vxlan_gpe_main_t *gm = &vxlan_gpe_main; + vxlan_gpe_tunnel_t *t = 0; + u32 vni; + + + if (!mp->is_ipv6) + { + clib_memcpy (&key4.local, &mp->local, sizeof (key4.local)); + clib_memcpy (&key4.remote, &mp->remote, sizeof (key4.remote)); + vni = clib_net_to_host_u32 (mp->vni); + key4.vni = clib_host_to_net_u32 (vni << 8); + key4.pad = 0; + + p = hash_get_mem (gm->vxlan4_gpe_tunnel_by_key, &key4); + } + else + { + return; + } + + if (!p) + return; + + t = pool_elt_at_index (gm->tunnels, p[0]); + + error = vxlan_gpe_ioam_clear (t, 0, 0, 0, 0); + + + if (error) + { + clib_error_report (error); + rv = clib_error_get_code (error); + } + + + VXLAN_GPE_REPLY_MACRO (VL_API_VXLAN_GPE_IOAM_VNI_DISABLE_REPLY); +} + +static void vl_api_vxlan_gpe_ioam_transit_enable_t_handler + (vl_api_vxlan_gpe_ioam_transit_enable_t * mp) +{ + int rv = 0; + vl_api_vxlan_gpe_ioam_transit_enable_reply_t *rmp; + vxlan_gpe_ioam_main_t *sm = &vxlan_gpe_ioam_main; + ip46_address_t dst_addr; + + memset (&dst_addr.ip4, 0, sizeof (dst_addr.ip4)); + if (!mp->is_ipv6) + { + clib_memcpy (&dst_addr.ip4, &mp->dst_addr, sizeof (dst_addr.ip4)); + } + rv = vxlan_gpe_enable_disable_ioam_for_dest (sm->vlib_main, + dst_addr, + ntohl (mp->outer_fib_index), + mp->is_ipv6 ? 0 : 1, + 1 /* is_add */ ); + + VXLAN_GPE_REPLY_MACRO (VL_API_VXLAN_GPE_IOAM_TRANSIT_ENABLE_REPLY); +} + +static void vl_api_vxlan_gpe_ioam_transit_disable_t_handler + (vl_api_vxlan_gpe_ioam_transit_disable_t * mp) +{ + int rv = 0; + vl_api_vxlan_gpe_ioam_transit_disable_reply_t *rmp; + vxlan_gpe_ioam_main_t *sm = &vxlan_gpe_ioam_main; + ip46_address_t dst_addr; + + memset (&dst_addr.ip4, 0, sizeof (dst_addr.ip4)); + if (!mp->is_ipv6) + { + clib_memcpy (&dst_addr.ip4, &mp->dst_addr, sizeof (dst_addr.ip4)); + } + + rv = vxlan_gpe_ioam_disable_for_dest (sm->vlib_main, + dst_addr, + ntohl (mp->outer_fib_index), + mp->is_ipv6 ? 0 : 1); + VXLAN_GPE_REPLY_MACRO (VL_API_VXLAN_GPE_IOAM_TRANSIT_DISABLE_REPLY); +} + + +/* + * This routine exists to convince the vlib plugin framework that + * we haven't accidentally copied a random .dll into the plugin directory. + * + * Also collects global variable pointers passed from the vpp engine + */ + +clib_error_t * +vlib_plugin_register (vlib_main_t * vm, vnet_plugin_handoff_t * h, + int from_early_init) +{ + vxlan_gpe_ioam_main_t *sm = &vxlan_gpe_ioam_main; + clib_error_t *error = 0; + + sm->vlib_main = vm; + sm->vnet_main = h->vnet_main; + sm->unix_time_0 = (u32) time (0); /* Store starting time */ + sm->vlib_time_0 = vlib_time_now (vm); + return error; +} + +/* Set up the API message handling tables */ +static clib_error_t * +vxlan_gpe_plugin_api_hookup (vlib_main_t * vm) +{ + vxlan_gpe_ioam_main_t *sm = &vxlan_gpe_ioam_main; +#define _(N,n) \ + vl_msg_api_set_handlers((VL_API_##N + sm->msg_id_base), \ + #n, \ + vl_api_##n##_t_handler, \ + vl_noop_handler, \ + vl_api_##n##_t_endian, \ + vl_api_##n##_t_print, \ + sizeof(vl_api_##n##_t), 1); + foreach_vxlan_gpe_plugin_api_msg; +#undef _ + + return 0; +} + +static clib_error_t * +vxlan_gpe_init (vlib_main_t * vm) +{ + vxlan_gpe_ioam_main_t *sm = &vxlan_gpe_ioam_main; + clib_error_t *error = 0; + u8 *name; + u32 encap_node_index = vxlan_gpe_encap_ioam_v4_node.index; + u32 decap_node_index = vxlan_gpe_decap_ioam_v4_node.index; + vlib_node_t *vxlan_gpe_encap_node = NULL; + vlib_node_t *vxlan_gpe_decap_node = NULL; + uword next_node = 0; + + name = format (0, "ioam_vxlan_gpe_%08x%c", api_version, 0); + + /* Ask for a correctly-sized block of API message decode slots */ + sm->msg_id_base = vl_msg_api_get_msg_ids + ((char *) name, VL_MSG_FIRST_AVAILABLE); + + error = vxlan_gpe_plugin_api_hookup (vm); + + /* Hook the ioam-encap node to vxlan-gpe-encap */ + vxlan_gpe_encap_node = vlib_get_node_by_name (vm, (u8 *) "vxlan-gpe-encap"); + sm->encap_v4_next_node = + vlib_node_add_next (vm, vxlan_gpe_encap_node->index, encap_node_index); + + vxlan_gpe_decap_node = + vlib_get_node_by_name (vm, (u8 *) "vxlan4-gpe-input"); + next_node = + vlib_node_add_next (vm, vxlan_gpe_decap_node->index, decap_node_index); + vxlan_gpe_register_decap_protocol (VXLAN_GPE_PROTOCOL_IOAM, next_node); + + vec_new (vxlan_gpe_ioam_sw_interface_t, pool_elts (sm->sw_interfaces)); + sm->dst_by_ip4 = hash_create_mem (0, sizeof (fib_prefix_t), sizeof (uword)); + + sm->dst_by_ip6 = hash_create_mem (0, sizeof (fib_prefix_t), sizeof (uword)); + + vxlan_gpe_ioam_interface_init (); + vec_free (name); + + return error; +} + +VLIB_INIT_FUNCTION (vxlan_gpe_init); + +/* + * fd.io coding-style-patch-verification: ON + * + * Local Variables: + * eval: (c-set-style "gnu") + * End: + */ diff --git a/src/plugins/ioam/lib-vxlan-gpe/vxlan_gpe_ioam.c b/src/plugins/ioam/lib-vxlan-gpe/vxlan_gpe_ioam.c new file mode 100644 index 00000000..6c04d9af --- /dev/null +++ b/src/plugins/ioam/lib-vxlan-gpe/vxlan_gpe_ioam.c @@ -0,0 +1,773 @@ +/* + * Copyright (c) 2016 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ +#include +#include +#include +#include +#include +#include +#include + +vxlan_gpe_ioam_main_t vxlan_gpe_ioam_main; + +int +vxlan_gpe_ioam_set_rewrite (vxlan_gpe_tunnel_t * t, int has_trace_option, + int has_pot_option, int has_ppc_option, + u8 ipv6_set) +{ + vxlan_gpe_ioam_main_t *hm = &vxlan_gpe_ioam_main; + u32 size; + vxlan_gpe_ioam_hdr_t *vxlan_gpe_ioam_hdr; + u8 *current; + u8 trace_data_size = 0; + u8 pot_data_size = 0; + + if (has_trace_option == 0 && has_pot_option == 0) + return -1; + + /* Work out how much space we need */ + size = sizeof (vxlan_gpe_ioam_hdr_t); + + if (has_trace_option + && hm->add_options[VXLAN_GPE_OPTION_TYPE_IOAM_TRACE] != 0) + { + size += sizeof (vxlan_gpe_ioam_option_t); + size += hm->options_size[VXLAN_GPE_OPTION_TYPE_IOAM_TRACE]; + } + if (has_pot_option + && hm->add_options[VXLAN_GPE_OPTION_TYPE_IOAM_PROOF_OF_TRANSIT] != 0) + { + size += sizeof (vxlan_gpe_ioam_option_t); + size += hm->options_size[VXLAN_GPE_OPTION_TYPE_IOAM_PROOF_OF_TRANSIT]; + } + + t->rewrite_size = size; + + if (!ipv6_set) + { + vxlan4_gpe_rewrite (t, size, VXLAN_GPE_PROTOCOL_IOAM, + hm->encap_v4_next_node); + vxlan_gpe_ioam_hdr = + (vxlan_gpe_ioam_hdr_t *) (t->rewrite + + sizeof (ip4_vxlan_gpe_header_t)); + } + else + { + vxlan6_gpe_rewrite (t, size, VXLAN_GPE_PROTOCOL_IOAM, + VXLAN_GPE_ENCAP_NEXT_IP6_LOOKUP); + vxlan_gpe_ioam_hdr = + (vxlan_gpe_ioam_hdr_t *) (t->rewrite + + sizeof (ip6_vxlan_gpe_header_t)); + } + + + vxlan_gpe_ioam_hdr->type = VXLAN_GPE_PROTOCOL_IOAM; + /* Length of the header in octets */ + vxlan_gpe_ioam_hdr->length = size; + vxlan_gpe_ioam_hdr->protocol = t->protocol; + current = (u8 *) vxlan_gpe_ioam_hdr + sizeof (vxlan_gpe_ioam_hdr_t); + + if (has_trace_option + && hm->add_options[VXLAN_GPE_OPTION_TYPE_IOAM_TRACE] != 0) + { + if (0 != hm->add_options[VXLAN_GPE_OPTION_TYPE_IOAM_TRACE] (current, + &trace_data_size)) + return -1; + current += trace_data_size; + } + if (has_pot_option + && hm->add_options[VXLAN_GPE_OPTION_TYPE_IOAM_PROOF_OF_TRANSIT] != 0) + { + pot_data_size = + hm->options_size[VXLAN_GPE_OPTION_TYPE_IOAM_PROOF_OF_TRANSIT]; + if (0 == + hm->add_options[VXLAN_GPE_OPTION_TYPE_IOAM_PROOF_OF_TRANSIT] + (current, &pot_data_size)) + current += pot_data_size; + } + + return 0; +} + +int +vxlan_gpe_ioam_clear_rewrite (vxlan_gpe_tunnel_t * t, int has_trace_option, + int has_pot_option, int has_ppc_option, + u8 ipv6_set) +{ + + t->rewrite_size = 0; + + if (!ipv6_set) + { + vxlan4_gpe_rewrite (t, 0, 0, VXLAN_GPE_ENCAP_NEXT_IP4_LOOKUP); + } + else + { + vxlan6_gpe_rewrite (t, 0, 0, VXLAN_GPE_ENCAP_NEXT_IP6_LOOKUP); + } + + + return 0; +} + +clib_error_t * +vxlan_gpe_ioam_clear (vxlan_gpe_tunnel_t * t, + int has_trace_option, int has_pot_option, + int has_ppc_option, u8 ipv6_set) +{ + int rv; + rv = vxlan_gpe_ioam_clear_rewrite (t, 0, 0, 0, 0); + + if (rv == 0) + { + return (0); + } + else + { + return clib_error_return_code (0, rv, 0, + "vxlan_gpe_ioam_clear_rewrite returned %d", + rv); + } + +} + + +clib_error_t * +vxlan_gpe_ioam_set (vxlan_gpe_tunnel_t * t, + int has_trace_option, int has_pot_option, + int has_ppc_option, u8 ipv6_set) +{ + int rv; + rv = vxlan_gpe_ioam_set_rewrite (t, has_trace_option, + has_pot_option, has_ppc_option, ipv6_set); + + if (rv == 0) + { + return (0); + } + else + { + return clib_error_return_code (0, rv, 0, + "vxlan_gpe_ioam_set_rewrite returned %d", + rv); + } + +} + +static void +vxlan_gpe_set_clear_output_feature_on_intf (vlib_main_t * vm, + u32 sw_if_index0, u8 is_add) +{ + + + + vnet_feature_enable_disable ("ip4-output", "vxlan-gpe-transit-ioam", + sw_if_index0, is_add, + 0 /* void *feature_config */ , + 0 /* u32 n_feature_config_bytes */ ); + return; +} + +void +vxlan_gpe_clear_output_feature_on_all_intfs (vlib_main_t * vm) +{ + vnet_sw_interface_t *si = 0; + vnet_main_t *vnm = vnet_get_main (); + vnet_interface_main_t *im = &vnm->interface_main; + + pool_foreach (si, im->sw_interfaces, ( + { + vxlan_gpe_set_clear_output_feature_on_intf + (vm, si->sw_if_index, 0); + })); + return; +} + + +extern fib_forward_chain_type_t +fib_entry_get_default_chain_type (const fib_entry_t * fib_entry); + +int +vxlan_gpe_enable_disable_ioam_for_dest (vlib_main_t * vm, + ip46_address_t dst_addr, + u32 outer_fib_index, + u8 is_ipv4, u8 is_add) +{ + vxlan_gpe_ioam_main_t *hm = &vxlan_gpe_ioam_main; + u32 fib_index0 = 0; + u32 sw_if_index0 = ~0; + + fib_node_index_t fei = ~0; + fib_entry_t *fib_entry; + u32 adj_index0; + ip_adjacency_t *adj0; + fib_prefix_t fib_prefix; + //fib_forward_chain_type_t fct; + load_balance_t *lb_m, *lb_b; + const dpo_id_t *dpo0, *dpo1; + u32 i, j; + //vnet_hw_interface_t *hw; + + if (is_ipv4) + { + memset (&fib_prefix, 0, sizeof (fib_prefix_t)); + fib_prefix.fp_len = 32; + fib_prefix.fp_proto = FIB_PROTOCOL_IP4; + fib_prefix.fp_addr = dst_addr; + } + else + { + return 0; + } + + fei = fib_table_lookup (fib_index0, &fib_prefix); + fib_entry = fib_entry_get (fei); + + //fct = fib_entry_get_default_chain_type (fib_entry); + + if (!dpo_id_is_valid (&fib_entry->fe_lb /*[fct] */ )) + { + return (-1); + } + + lb_m = load_balance_get (fib_entry->fe_lb /*[fct] */ .dpoi_index); + + for (i = 0; i < lb_m->lb_n_buckets; i++) + { + dpo0 = load_balance_get_bucket_i (lb_m, i); + + if (dpo0->dpoi_type == DPO_LOAD_BALANCE) + { + lb_b = load_balance_get (dpo0->dpoi_index); + + for (j = 0; j < lb_b->lb_n_buckets; j++) + { + dpo1 = load_balance_get_bucket_i (lb_b, j); + + if (dpo1->dpoi_type == DPO_ADJACENCY) + { + adj_index0 = dpo1->dpoi_index; + + if (ADJ_INDEX_INVALID == adj_index0) + { + continue; + } + + adj0 = + ip_get_adjacency (&(ip4_main.lookup_main), adj_index0); + sw_if_index0 = adj0->rewrite_header.sw_if_index; + + if (~0 == sw_if_index0) + { + continue; + } + + + if (is_add) + { + vnet_feature_enable_disable ("ip4-output", + "vxlan-gpe-transit-ioam", + sw_if_index0, is_add, 0 + /* void *feature_config */ + , 0 /* u32 n_feature_config_bytes */ + ); + + vec_validate_init_empty (hm->bool_ref_by_sw_if_index, + sw_if_index0, ~0); + hm->bool_ref_by_sw_if_index[sw_if_index0] = 1; + } + else + { + hm->bool_ref_by_sw_if_index[sw_if_index0] = ~0; + } + } + } + } + } + + if (is_ipv4) + { + + uword *t = NULL; + vxlan_gpe_ioam_dest_tunnels_t *t1; + fib_prefix_t key4, *key4_copy; + hash_pair_t *hp; + memset (&key4, 0, sizeof (key4)); + key4.fp_proto = FIB_PROTOCOL_IP4; + key4.fp_addr.ip4.as_u32 = fib_prefix.fp_addr.ip4.as_u32; + t = hash_get_mem (hm->dst_by_ip4, &key4); + if (is_add) + { + if (t) + { + return 0; + } + pool_get_aligned (hm->dst_tunnels, t1, CLIB_CACHE_LINE_BYTES); + memset (t1, 0, sizeof (*t1)); + t1->fp_proto = FIB_PROTOCOL_IP4; + t1->dst_addr.ip4.as_u32 = fib_prefix.fp_addr.ip4.as_u32; + key4_copy = clib_mem_alloc (sizeof (*key4_copy)); + clib_memcpy (key4_copy, &key4, sizeof (*key4_copy)); + hash_set_mem (hm->dst_by_ip4, key4_copy, t1 - hm->dst_tunnels); + /* + * Attach to the FIB entry for the VxLAN-GPE destination + * and become its child. The dest route will invoke a callback + * when the fib entry changes, it can be used to + * re-program the output feature on the egress interface. + */ + + const fib_prefix_t tun_dst_pfx = { + .fp_len = 32, + .fp_proto = FIB_PROTOCOL_IP4, + .fp_addr = {.ip4 = t1->dst_addr.ip4,} + }; + + t1->fib_entry_index = + fib_table_entry_special_add (outer_fib_index, + &tun_dst_pfx, + FIB_SOURCE_RR, + FIB_ENTRY_FLAG_NONE, + ADJ_INDEX_INVALID); + t1->sibling_index = + fib_entry_child_add (t1->fib_entry_index, + hm->fib_entry_type, t1 - hm->dst_tunnels); + t1->outer_fib_index = outer_fib_index; + + } + else + { + if (!t) + { + return 0; + } + t1 = pool_elt_at_index (hm->dst_tunnels, t[0]); + hp = hash_get_pair (hm->dst_by_ip4, &key4); + key4_copy = (void *) (hp->key); + hash_unset_mem (hm->dst_by_ip4, &key4); + clib_mem_free (key4_copy); + pool_put (hm->dst_tunnels, t1); + } + } + else + { + // TBD for IPv6 + } + + return 0; +} + +void +vxlan_gpe_refresh_output_feature_on_all_dest (void) +{ + vxlan_gpe_ioam_main_t *hm = &vxlan_gpe_ioam_main; + vxlan_gpe_ioam_dest_tunnels_t *t; + u32 i; + if (pool_elts (hm->dst_tunnels) == 0) + return; + vxlan_gpe_clear_output_feature_on_all_intfs (hm->vlib_main); + i = vec_len (hm->bool_ref_by_sw_if_index); + vec_free (hm->bool_ref_by_sw_if_index); + vec_validate_init_empty (hm->bool_ref_by_sw_if_index, i, ~0); + pool_foreach (t, hm->dst_tunnels, ( + { + vxlan_gpe_enable_disable_ioam_for_dest + (hm->vlib_main, + t->dst_addr, + t->outer_fib_index, + (t->fp_proto == FIB_PROTOCOL_IP4), 1 + /* is_add */ + ); + } + )); + return; +} + +void +vxlan_gpe_clear_output_feature_on_select_intfs (void) +{ + vxlan_gpe_ioam_main_t *hm = &vxlan_gpe_ioam_main; + u32 sw_if_index0 = 0; + for (sw_if_index0 = 0; + sw_if_index0 < vec_len (hm->bool_ref_by_sw_if_index); sw_if_index0++) + { + if (hm->bool_ref_by_sw_if_index[sw_if_index0] == 0xFF) + { + vxlan_gpe_set_clear_output_feature_on_intf + (hm->vlib_main, sw_if_index0, 0); + } + } + + return; +} + +static clib_error_t * +vxlan_gpe_set_ioam_rewrite_command_fn (vlib_main_t * + vm, + unformat_input_t + * input, vlib_cli_command_t * cmd) +{ + vxlan_gpe_ioam_main_t *hm = &vxlan_gpe_ioam_main; + ip46_address_t local, remote; + u8 local_set = 0; + u8 remote_set = 0; + u8 ipv4_set = 0; + u8 ipv6_set = 0; + u32 vni; + u8 vni_set = 0; + u8 disable = 0; + clib_error_t *rv = 0; + vxlan4_gpe_tunnel_key_t key4; + vxlan6_gpe_tunnel_key_t key6; + uword *p; + vxlan_gpe_main_t *gm = &vxlan_gpe_main; + vxlan_gpe_tunnel_t *t = 0; + while (unformat_check_input (input) != UNFORMAT_END_OF_INPUT) + { + if (unformat (input, "local %U", unformat_ip4_address, &local.ip4)) + { + local_set = 1; + ipv4_set = 1; + } + else + if (unformat (input, "remote %U", unformat_ip4_address, &remote.ip4)) + { + remote_set = 1; + ipv4_set = 1; + } + else if (unformat (input, "local %U", unformat_ip6_address, &local.ip6)) + { + local_set = 1; + ipv6_set = 1; + } + else + if (unformat (input, "remote %U", unformat_ip6_address, &remote.ip6)) + { + remote_set = 1; + ipv6_set = 1; + } + else if (unformat (input, "vni %d", &vni)) + vni_set = 1; + else if (unformat (input, "disable")) + disable = 1; + else + break; + } + + if (local_set == 0) + return clib_error_return (0, "tunnel local address not specified"); + if (remote_set == 0) + return clib_error_return (0, "tunnel remote address not specified"); + if (ipv4_set && ipv6_set) + return clib_error_return (0, "both IPv4 and IPv6 addresses specified"); + if ((ipv4_set + && memcmp (&local.ip4, &remote.ip4, + sizeof (local.ip4)) == 0) || (ipv6_set + && + memcmp + (&local.ip6, + &remote.ip6, + sizeof (local.ip6)) == 0)) + return clib_error_return (0, "src and dst addresses are identical"); + if (vni_set == 0) + return clib_error_return (0, "vni not specified"); + if (!ipv6_set) + { + key4.local = local.ip4.as_u32; + key4.remote = remote.ip4.as_u32; + key4.vni = clib_host_to_net_u32 (vni << 8); + key4.pad = 0; + p = hash_get_mem (gm->vxlan4_gpe_tunnel_by_key, &key4); + } + else + { + key6.local.as_u64[0] = local.ip6.as_u64[0]; + key6.local.as_u64[1] = local.ip6.as_u64[1]; + key6.remote.as_u64[0] = remote.ip6.as_u64[0]; + key6.remote.as_u64[1] = remote.ip6.as_u64[1]; + key6.vni = clib_host_to_net_u32 (vni << 8); + p = hash_get_mem (gm->vxlan6_gpe_tunnel_by_key, &key6); + } + + if (!p) + return clib_error_return (0, "VxLAN Tunnel not found"); + t = pool_elt_at_index (gm->tunnels, p[0]); + if (!disable) + { + rv = + vxlan_gpe_ioam_set (t, hm->has_trace_option, + hm->has_pot_option, hm->has_ppc_option, ipv6_set); + } + else + { + rv = vxlan_gpe_ioam_clear (t, 0, 0, 0, 0); + } + return rv; +} + + +/* *INDENT-OFF* */ +VLIB_CLI_COMMAND (vxlan_gpe_set_ioam_rewrite_cmd, static) = { + .path = "set vxlan-gpe-ioam", + .short_help = "set vxlan-gpe-ioam vxlan [disable]", + .function = vxlan_gpe_set_ioam_rewrite_command_fn, +}; +/* *INDENT-ON* */ + + + +clib_error_t * +vxlan_gpe_ioam_enable (int has_trace_option, + int has_pot_option, int has_ppc_option) +{ + vxlan_gpe_ioam_main_t *hm = &vxlan_gpe_ioam_main; + hm->has_trace_option = has_trace_option; + hm->has_pot_option = has_pot_option; + hm->has_ppc_option = has_ppc_option; + if (hm->has_trace_option) + { + vxlan_gpe_trace_profile_setup (); + } + + return 0; +} + +clib_error_t * +vxlan_gpe_ioam_disable (int + has_trace_option, + int has_pot_option, int has_ppc_option) +{ + vxlan_gpe_ioam_main_t *hm = &vxlan_gpe_ioam_main; + hm->has_trace_option = has_trace_option; + hm->has_pot_option = has_pot_option; + hm->has_ppc_option = has_ppc_option; + if (!hm->has_trace_option) + { + vxlan_gpe_trace_profile_cleanup (); + } + + return 0; +} + +void +vxlan_gpe_set_next_override (uword next) +{ + vxlan_gpe_ioam_main_t *hm = &vxlan_gpe_ioam_main; + hm->decap_v4_next_override = next; + return; +} + +static clib_error_t * +vxlan_gpe_set_ioam_flags_command_fn (vlib_main_t * vm, + unformat_input_t + * input, vlib_cli_command_t * cmd) +{ + int has_trace_option = 0; + int has_pot_option = 0; + int has_ppc_option = 0; + clib_error_t *rv = 0; + while (unformat_check_input (input) != UNFORMAT_END_OF_INPUT) + { + if (unformat (input, "trace")) + has_trace_option = 1; + else if (unformat (input, "pot")) + has_pot_option = 1; + else if (unformat (input, "ppc encap")) + has_ppc_option = PPC_ENCAP; + else if (unformat (input, "ppc decap")) + has_ppc_option = PPC_DECAP; + else if (unformat (input, "ppc none")) + has_ppc_option = PPC_NONE; + else + break; + } + + + rv = + vxlan_gpe_ioam_enable (has_trace_option, has_pot_option, has_ppc_option); + return rv; +} + +/* *INDENT-OFF* */ +VLIB_CLI_COMMAND (vxlan_gpe_set_ioam_flags_cmd, static) = +{ +.path = "set vxlan-gpe-ioam rewrite", +.short_help = "set vxlan-gpe-ioam [trace] [pot] [ppc ]", +.function = vxlan_gpe_set_ioam_flags_command_fn,}; +/* *INDENT-ON* */ + + +int vxlan_gpe_ioam_disable_for_dest + (vlib_main_t * vm, ip46_address_t dst_addr, u32 outer_fib_index, + u8 ipv4_set) +{ + vxlan_gpe_ioam_dest_tunnels_t *t; + vxlan_gpe_ioam_main_t *hm = &vxlan_gpe_ioam_main; + + vxlan_gpe_enable_disable_ioam_for_dest (hm->vlib_main, + dst_addr, outer_fib_index, ipv4_set, + 0); + if (pool_elts (hm->dst_tunnels) == 0) + { + vxlan_gpe_clear_output_feature_on_select_intfs (); + return 0; + } + + pool_foreach (t, hm->dst_tunnels, ( + { + vxlan_gpe_enable_disable_ioam_for_dest + (hm->vlib_main, + t->dst_addr, + t->outer_fib_index, + (t->fp_proto == + FIB_PROTOCOL_IP4), 1 /* is_add */ ); + } + )); + vxlan_gpe_clear_output_feature_on_select_intfs (); + return (0); + +} + +static clib_error_t *vxlan_gpe_set_ioam_transit_rewrite_command_fn + (vlib_main_t * vm, unformat_input_t * input, vlib_cli_command_t * cmd) +{ + vxlan_gpe_ioam_main_t *hm = &vxlan_gpe_ioam_main; + ip46_address_t dst_addr; + u8 dst_addr_set = 0; + u8 ipv4_set = 0; + u8 ipv6_set = 0; + u8 disable = 0; + clib_error_t *rv = 0; + u32 outer_fib_index = 0; + while (unformat_check_input (input) != UNFORMAT_END_OF_INPUT) + { + if (unformat (input, "dst-ip %U", unformat_ip4_address, &dst_addr.ip4)) + { + dst_addr_set = 1; + ipv4_set = 1; + } + else + if (unformat + (input, "dst-ip %U", unformat_ip6_address, &dst_addr.ip6)) + { + dst_addr_set = 1; + ipv6_set = 1; + } + else if (unformat (input, "outer-fib-index %d", &outer_fib_index)) + { + } + + else if (unformat (input, "disable")) + disable = 1; + else + break; + } + + if (dst_addr_set == 0) + return clib_error_return (0, "tunnel destination address not specified"); + if (ipv4_set && ipv6_set) + return clib_error_return (0, "both IPv4 and IPv6 addresses specified"); + if (!disable) + { + vxlan_gpe_enable_disable_ioam_for_dest (hm->vlib_main, + dst_addr, outer_fib_index, + ipv4_set, 1); + } + else + { + vxlan_gpe_ioam_disable_for_dest + (vm, dst_addr, outer_fib_index, ipv4_set); + } + return rv; +} + + /* *INDENT-OFF* */ +VLIB_CLI_COMMAND (vxlan_gpe_set_ioam_transit_rewrite_cmd, static) = { + .path = "set vxlan-gpe-ioam-transit", + .short_help = "set vxlan-gpe-ioam-transit dst-ip [outer-fib-index ] [disable]", + .function = vxlan_gpe_set_ioam_transit_rewrite_command_fn, +}; +/* *INDENT-ON* */ + +clib_error_t *clear_vxlan_gpe_ioam_rewrite_command_fn + (vlib_main_t * vm, unformat_input_t * input, vlib_cli_command_t * cmd) +{ + return (vxlan_gpe_ioam_disable (0, 0, 0)); +} + +/* *INDENT-OFF* */ +VLIB_CLI_COMMAND (vxlan_gpe_clear_ioam_flags_cmd, static) = +{ +.path = "clear vxlan-gpe-ioam rewrite", +.short_help = "clear vxlan-gpe-ioam rewrite", +.function = clear_vxlan_gpe_ioam_rewrite_command_fn, +}; +/* *INDENT-ON* */ + + +/** + * Function definition to backwalk a FIB node + */ +static fib_node_back_walk_rc_t +vxlan_gpe_ioam_back_walk (fib_node_t * node, fib_node_back_walk_ctx_t * ctx) +{ + vxlan_gpe_refresh_output_feature_on_all_dest (); + return (FIB_NODE_BACK_WALK_CONTINUE); +} + +/** + * Function definition to get a FIB node from its index + */ +static fib_node_t * +vxlan_gpe_ioam_fib_node_get (fib_node_index_t index) +{ + vxlan_gpe_ioam_main_t *hm = &vxlan_gpe_ioam_main; + return (&hm->node); +} + +/** + * Function definition to inform the FIB node that its last lock has gone. + */ +static void +vxlan_gpe_ioam_last_lock_gone (fib_node_t * node) +{ + ASSERT (0); +} + + +/* + * Virtual function table registered by MPLS GRE tunnels + * for participation in the FIB object graph. + */ +const static fib_node_vft_t vxlan_gpe_ioam_vft = { + .fnv_get = vxlan_gpe_ioam_fib_node_get, + .fnv_last_lock = vxlan_gpe_ioam_last_lock_gone, + .fnv_back_walk = vxlan_gpe_ioam_back_walk, +}; + +void +vxlan_gpe_ioam_interface_init (void) +{ + vxlan_gpe_ioam_main_t *hm = &vxlan_gpe_ioam_main; + hm->fib_entry_type = fib_node_register_new_type (&vxlan_gpe_ioam_vft); + return; +} + +/* + * fd.io coding-style-patch-verification: ON + * + * Local Variables: + * eval: (c-set-style "gnu") + * End: + */ diff --git a/src/plugins/ioam/lib-vxlan-gpe/vxlan_gpe_ioam.h b/src/plugins/ioam/lib-vxlan-gpe/vxlan_gpe_ioam.h new file mode 100644 index 00000000..0711b87a --- /dev/null +++ b/src/plugins/ioam/lib-vxlan-gpe/vxlan_gpe_ioam.h @@ -0,0 +1,183 @@ +/* + * Copyright (c) 2015 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ +#ifndef __included_vxlan_gpe_ioam_h__ +#define __included_vxlan_gpe_ioam_h__ + +#include +#include +#include +#include + + +typedef struct vxlan_gpe_sw_interface_ +{ + u32 sw_if_index; +} vxlan_gpe_ioam_sw_interface_t; + +typedef struct vxlan_gpe_dest_tunnels_ +{ + ip46_address_t dst_addr; + u32 fp_proto; + u32 sibling_index; + fib_node_index_t fib_entry_index; + u32 outer_fib_index; +} vxlan_gpe_ioam_dest_tunnels_t; + +typedef struct vxlan_gpe_ioam_main_ +{ + /** + * Linkage into the FIB object graph + */ + fib_node_t node; + + /* time scale transform. Joy. */ + u32 unix_time_0; + f64 vlib_time_0; + + + /* Trace option */ + u8 has_trace_option; + + /* Pot option */ + u8 has_pot_option; + +#define PPC_NONE 0 +#define PPC_ENCAP 1 +#define PPC_DECAP 2 + u8 has_ppc_option; + +#define TSP_SECONDS 0 +#define TSP_MILLISECONDS 1 +#define TSP_MICROSECONDS 2 +#define TSP_NANOSECONDS 3 + + /* Array of function pointers to ADD and POP VxLAN-GPE iOAM option handling routines */ + u8 options_size[256]; + int (*add_options[256]) (u8 * rewrite_string, u8 * rewrite_size); + int (*pop_options[256]) (ip4_header_t * ip, vxlan_gpe_ioam_option_t * opt); + + /* Array of function pointers to iOAM option handling routines */ + int (*options[256]) (vlib_buffer_t * b, vxlan_gpe_ioam_option_t * opt, + u8 is_ipv4, u8 use_adj); + u8 *(*trace[256]) (u8 * s, vxlan_gpe_ioam_option_t * opt); + + /* API message ID base */ + u16 msg_id_base; + + /* Override to export for iOAM */ + uword decap_v4_next_override; + uword decap_v6_next_override; + + /* sequence of node graph for encap */ + uword encap_v4_next_node; + uword encap_v6_next_node; + + /* Software interfaces. */ + vxlan_gpe_ioam_sw_interface_t *sw_interfaces; + + /* hash ip4/ip6 -> list of destinations for doing transit iOAM operation */ + vxlan_gpe_ioam_dest_tunnels_t *dst_tunnels; + uword *dst_by_ip4; + uword *dst_by_ip6; + + /** per sw_if_index, to maintain bitmap */ + u8 *bool_ref_by_sw_if_index; + fib_node_type_t fib_entry_type; + + /** State convenience vlib_main_t */ + vlib_main_t *vlib_main; + /** State convenience vnet_main_t */ + vnet_main_t *vnet_main; + + +} vxlan_gpe_ioam_main_t; +extern vxlan_gpe_ioam_main_t vxlan_gpe_ioam_main; + +/* + * Primary h-b-h handler trace support + */ +typedef struct +{ + u32 next_index; + u32 trace_len; + u8 option_data[256]; +} ioam_trace_t; + + +extern vlib_node_registration_t vxlan_gpe_encap_ioam_v4_node; +extern vlib_node_registration_t vxlan_gpe_decap_ioam_v4_node; +extern vlib_node_registration_t vxlan_gpe_transit_ioam_v4_node; + +clib_error_t *vxlan_gpe_ioam_enable (int has_trace_option, int has_pot_option, + int has_ppc_option); + +clib_error_t *vxlan_gpe_ioam_disable (int has_trace_option, + int has_pot_option, int has_ppc_option); + +clib_error_t *vxlan_gpe_ioam_set (vxlan_gpe_tunnel_t * t, + int has_trace_option, + int has_pot_option, + int has_ppc_option, u8 ipv6_set); +clib_error_t *vxlan_gpe_ioam_clear (vxlan_gpe_tunnel_t * t, + int has_trace_option, int has_pot_option, + int has_ppc_option, u8 ipv6_set); + +int vxlan_gpe_ioam_add_register_option (u8 option, + u8 size, + int rewrite_options (u8 * + rewrite_string, + u8 * + rewrite_size)); + +int vxlan_gpe_add_unregister_option (u8 option); + +int vxlan_gpe_ioam_register_option (u8 option, + int options (vlib_buffer_t * b, + vxlan_gpe_ioam_option_t * + opt, u8 is_ipv4, u8 use_adj), + u8 * trace (u8 * s, + vxlan_gpe_ioam_option_t * + opt)); +int vxlan_gpe_ioam_unregister_option (u8 option); + +int vxlan_gpe_trace_profile_setup (void); + +int vxlan_gpe_trace_profile_cleanup (void); +extern void vxlan_gpe_ioam_interface_init (void); +int +vxlan_gpe_enable_disable_ioam_for_dest (vlib_main_t * vm, + ip46_address_t dst_addr, + u32 outer_fib_index, + u8 is_ipv4, u8 is_add); +int vxlan_gpe_ioam_disable_for_dest + (vlib_main_t * vm, ip46_address_t dst_addr, u32 outer_fib_index, + u8 ipv4_set); + +typedef enum +{ + VXLAN_GPE_DECAP_IOAM_V4_NEXT_POP, + VXLAN_GPE_DECAP_IOAM_V4_NEXT_DROP, + VXLAN_GPE_DECAP_IOAM_V4_N_NEXT +} vxlan_gpe_decap_ioam_v4_next_t; + +#endif + +/* + * fd.io coding-style-patch-verification: ON + * + * Local Variables: + * eval: (c-set-style "gnu") + * End: + */ diff --git a/src/plugins/ioam/lib-vxlan-gpe/vxlan_gpe_ioam_packet.h b/src/plugins/ioam/lib-vxlan-gpe/vxlan_gpe_ioam_packet.h new file mode 100644 index 00000000..a7ef859e --- /dev/null +++ b/src/plugins/ioam/lib-vxlan-gpe/vxlan_gpe_ioam_packet.h @@ -0,0 +1,61 @@ +/* + * Copyright (c) 2015 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ +#ifndef __included_vxlan_gpe_ioam_packet_h__ +#define __included_vxlan_gpe_ioam_packet_h__ + +#include +#include +#include + + + +#define VXLAN_GPE_OPTION_TYPE_IOAM_TRACE 59 +#define VXLAN_GPE_OPTION_TYPE_IOAM_PROOF_OF_TRANSIT 60 + +/** + * @brief VXLAN GPE Extension (iOAM) Header definition + */ +typedef struct +{ + u8 type; + u8 length; + /** Reserved */ + u8 reserved; + /** see vxlan_gpe_protocol_t */ + u8 protocol; +} vxlan_gpe_ioam_hdr_t; + +/* + * @brief VxLAN GPE iOAM Option definition + */ +typedef struct +{ + /* Option Type */ + u8 type; + /* Length in octets of the option data field */ + u8 length; +} vxlan_gpe_ioam_option_t; + + +#endif + + +/* + * fd.io coding-style-patch-verification: ON + * + * Local Variables: + * eval: (c-set-style "gnu") + * End: + */ diff --git a/src/plugins/ioam/lib-vxlan-gpe/vxlan_gpe_ioam_trace.c b/src/plugins/ioam/lib-vxlan-gpe/vxlan_gpe_ioam_trace.c new file mode 100644 index 00000000..e37b1642 --- /dev/null +++ b/src/plugins/ioam/lib-vxlan-gpe/vxlan_gpe_ioam_trace.c @@ -0,0 +1,552 @@ +/* + * Copyright (c) 2016 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ +#include +#include +#include +#include + +#include +#include + +#include +#include +#include + +#include +#include + +/* Timestamp precision multipliers for seconds, milliseconds, microseconds + * and nanoseconds respectively. + */ +static f64 trace_tsp_mul[4] = { 1, 1e3, 1e6, 1e9 }; + +typedef union +{ + u64 as_u64; + u32 as_u32[2]; +} time_u64_t; + + +/* *INDENT-OFF* */ +typedef CLIB_PACKED(struct { + vxlan_gpe_ioam_option_t hdr; + u8 ioam_trace_type; + u8 data_list_elts_left; + u32 elts[0]; /* Variable type. So keep it generic */ +}) vxlan_gpe_ioam_trace_option_t; +/* *INDENT-ON* */ + + +#define foreach_vxlan_gpe_ioam_trace_stats \ + _(SUCCESS, "Pkts updated with TRACE records") \ + _(FAILED, "Errors in TRACE due to lack of TRACE records") + +static char *vxlan_gpe_ioam_trace_stats_strings[] = { +#define _(sym,string) string, + foreach_vxlan_gpe_ioam_trace_stats +#undef _ +}; + +typedef enum +{ +#define _(sym,str) VXLAN_GPE_IOAM_TRACE_##sym, + foreach_vxlan_gpe_ioam_trace_stats +#undef _ + VXLAN_GPE_IOAM_TRACE_N_STATS, +} vxlan_gpe_ioam_trace_stats_t; + + +typedef struct +{ + /* stats */ + u64 counters[ARRAY_LEN (vxlan_gpe_ioam_trace_stats_strings)]; + + /* convenience */ + vlib_main_t *vlib_main; + vnet_main_t *vnet_main; +} vxlan_gpe_ioam_trace_main_t; + +vxlan_gpe_ioam_trace_main_t vxlan_gpe_ioam_trace_main; + +int +vxlan_gpe_ioam_add_register_option (u8 option, + u8 size, + int rewrite_options (u8 * rewrite_string, + u8 * rewrite_size)) +{ + vxlan_gpe_ioam_main_t *hm = &vxlan_gpe_ioam_main; + + ASSERT (option < ARRAY_LEN (hm->add_options)); + + /* Already registered */ + if (hm->add_options[option]) + return (-1); + + hm->add_options[option] = rewrite_options; + hm->options_size[option] = size; + + return (0); +} + +int +vxlan_gpe_add_unregister_option (u8 option) +{ + vxlan_gpe_ioam_main_t *hm = &vxlan_gpe_ioam_main; + + ASSERT (option < ARRAY_LEN (hm->add_options)); + + /* Not registered */ + if (!hm->add_options[option]) + return (-1); + + hm->add_options[option] = NULL; + hm->options_size[option] = 0; + return (0); +} + + +int +vxlan_gpe_ioam_register_option (u8 option, + int options (vlib_buffer_t * b, + vxlan_gpe_ioam_option_t * opt, + u8 is_ipv4, u8 use_adj), + u8 * trace (u8 * s, + vxlan_gpe_ioam_option_t * opt)) +{ + vxlan_gpe_ioam_main_t *im = &vxlan_gpe_ioam_main; + + ASSERT (option < ARRAY_LEN (im->options)); + + /* Already registered */ + if (im->options[option]) + return (-1); + + im->options[option] = options; + im->trace[option] = trace; + + return (0); +} + +int +vxlan_gpe_ioam_unregister_option (u8 option) +{ + vxlan_gpe_ioam_main_t *hm = &vxlan_gpe_ioam_main; + + ASSERT (option < ARRAY_LEN (hm->options)); + + /* Not registered */ + if (!hm->options[option]) + return (-1); + + hm->options[option] = NULL; + hm->trace[option] = NULL; + + return (0); +} + + +always_inline void +vxlan_gpe_ioam_trace_stats_increment_counter (u32 counter_index, + u64 increment) +{ + vxlan_gpe_ioam_trace_main_t *hm = &vxlan_gpe_ioam_trace_main; + + hm->counters[counter_index] += increment; +} + + +static u8 * +format_ioam_data_list_element (u8 * s, va_list * args) +{ + u32 *elt = va_arg (*args, u32 *); + u8 *trace_type_p = va_arg (*args, u8 *); + u8 trace_type = *trace_type_p; + + + if (trace_type & BIT_TTL_NODEID) + { + u32 ttl_node_id_host_byte_order = clib_net_to_host_u32 (*elt); + s = format (s, "ttl 0x%x node id 0x%x ", + ttl_node_id_host_byte_order >> 24, + ttl_node_id_host_byte_order & 0x00FFFFFF); + + elt++; + } + + if (trace_type & BIT_ING_INTERFACE && trace_type & BIT_ING_INTERFACE) + { + u32 ingress_host_byte_order = clib_net_to_host_u32 (*elt); + s = format (s, "ingress 0x%x egress 0x%x ", + ingress_host_byte_order >> 16, + ingress_host_byte_order & 0xFFFF); + elt++; + } + + if (trace_type & BIT_TIMESTAMP) + { + u32 ts_in_host_byte_order = clib_net_to_host_u32 (*elt); + s = format (s, "ts 0x%x \n", ts_in_host_byte_order); + elt++; + } + + if (trace_type & BIT_APPDATA) + { + u32 appdata_in_host_byte_order = clib_net_to_host_u32 (*elt); + s = format (s, "app 0x%x ", appdata_in_host_byte_order); + elt++; + } + + return s; +} + + + +int +vxlan_gpe_ioam_trace_rewrite_handler (u8 * rewrite_string, u8 * rewrite_size) +{ + vxlan_gpe_ioam_trace_option_t *trace_option = NULL; + u8 trace_data_size = 0; + u8 trace_option_elts = 0; + trace_profile *profile = NULL; + + + profile = trace_profile_find (); + + if (PREDICT_FALSE (!profile)) + { + return (-1); + } + + if (PREDICT_FALSE (!rewrite_string)) + return -1; + + trace_option_elts = profile->num_elts; + trace_data_size = fetch_trace_data_size (profile->trace_type); + trace_option = (vxlan_gpe_ioam_trace_option_t *) rewrite_string; + trace_option->hdr.type = VXLAN_GPE_OPTION_TYPE_IOAM_TRACE; + trace_option->hdr.length = 2 /*ioam_trace_type,data_list_elts_left */ + + trace_option_elts * trace_data_size; + trace_option->ioam_trace_type = profile->trace_type & TRACE_TYPE_MASK; + trace_option->data_list_elts_left = trace_option_elts; + *rewrite_size = + sizeof (vxlan_gpe_ioam_trace_option_t) + + (trace_option_elts * trace_data_size); + + return 0; +} + + +int +vxlan_gpe_ioam_trace_data_list_handler (vlib_buffer_t * b, + vxlan_gpe_ioam_option_t * opt, + u8 is_ipv4, u8 use_adj) +{ + u8 elt_index = 0; + vxlan_gpe_ioam_trace_option_t *trace = + (vxlan_gpe_ioam_trace_option_t *) opt; + time_u64_t time_u64; + u32 *elt; + int rv = 0; + trace_profile *profile = NULL; + vxlan_gpe_ioam_main_t *hm = &vxlan_gpe_ioam_main; + + + profile = trace_profile_find (); + + if (PREDICT_FALSE (!profile)) + { + return (-1); + } + + + time_u64.as_u64 = 0; + + if (PREDICT_TRUE (trace->data_list_elts_left)) + { + trace->data_list_elts_left--; + /* fetch_trace_data_size returns in bytes. Convert it to 4-bytes + * to skip to this node's location. + */ + elt_index = + trace->data_list_elts_left * + fetch_trace_data_size (trace->ioam_trace_type) / 4; + elt = &trace->elts[elt_index]; + if (is_ipv4) + { + if (trace->ioam_trace_type & BIT_TTL_NODEID) + { + ip4_header_t *ip0 = vlib_buffer_get_current (b); + /* The transit case is the only case where the TTL decrement happens + * before iOAM processing. For now, use the use_adj flag as an overload. + * We can probably use a separate flag instead of overloading the use_adj flag. + */ + *elt = clib_host_to_net_u32 (((ip0->ttl - 1 + use_adj) << 24) | + profile->node_id); + elt++; + } + + if (trace->ioam_trace_type & BIT_ING_INTERFACE) + { + u16 tx_if = 0; + u32 adj_index = vnet_buffer (b)->ip.adj_index[VLIB_TX]; + ip4_main_t *im4 = &ip4_main; + ip_lookup_main_t *lm = &im4->lookup_main; + if (use_adj) + { + ip_adjacency_t *adj = ip_get_adjacency (lm, adj_index); + tx_if = adj->rewrite_header.sw_if_index & 0xFFFF; + } + + *elt = + (vnet_buffer (b)->sw_if_index[VLIB_RX] & 0xFFFF) << 16 | + tx_if; + *elt = clib_host_to_net_u32 (*elt); + elt++; + } + } + else + { + if (trace->ioam_trace_type & BIT_TTL_NODEID) + { + ip6_header_t *ip0 = vlib_buffer_get_current (b); + *elt = clib_host_to_net_u32 ((ip0->hop_limit << 24) | + profile->node_id); + elt++; + } + if (trace->ioam_trace_type & BIT_ING_INTERFACE) + { + u16 tx_if = 0; + u32 adj_index = vnet_buffer (b)->ip.adj_index[VLIB_TX]; + ip6_main_t *im6 = &ip6_main; + ip_lookup_main_t *lm = &im6->lookup_main; + if (use_adj) + { + ip_adjacency_t *adj = ip_get_adjacency (lm, adj_index); + tx_if = adj->rewrite_header.sw_if_index & 0xFFFF; + } + + *elt = + (vnet_buffer (b)->sw_if_index[VLIB_RX] & 0xFFFF) << 16 | + tx_if; + *elt = clib_host_to_net_u32 (*elt); + elt++; + } + } + + if (trace->ioam_trace_type & BIT_TIMESTAMP) + { + /* Send least significant 32 bits */ + f64 time_f64 = + (f64) (((f64) hm->unix_time_0) + + (vlib_time_now (hm->vlib_main) - hm->vlib_time_0)); + + time_u64.as_u64 = time_f64 * trace_tsp_mul[profile->trace_tsp]; + *elt = clib_host_to_net_u32 (time_u64.as_u32[0]); + elt++; + } + + if (trace->ioam_trace_type & BIT_APPDATA) + { + /* $$$ set elt0->app_data */ + *elt = clib_host_to_net_u32 (profile->app_data); + elt++; + } + vxlan_gpe_ioam_trace_stats_increment_counter + (VXLAN_GPE_IOAM_TRACE_SUCCESS, 1); + } + else + { + vxlan_gpe_ioam_trace_stats_increment_counter + (VXLAN_GPE_IOAM_TRACE_FAILED, 1); + } + return (rv); +} + +u8 * +vxlan_gpe_ioam_trace_data_list_trace_handler (u8 * s, + vxlan_gpe_ioam_option_t * opt) +{ + vxlan_gpe_ioam_trace_option_t *trace; + u8 trace_data_size_in_words = 0; + u32 *elt; + int elt_index = 0; + + trace = (vxlan_gpe_ioam_trace_option_t *) opt; + s = + format (s, " Trace Type 0x%x , %d elts left\n", trace->ioam_trace_type, + trace->data_list_elts_left); + trace_data_size_in_words = + fetch_trace_data_size (trace->ioam_trace_type) / 4; + elt = &trace->elts[0]; + while ((u8 *) elt < ((u8 *) (&trace->elts[0]) + trace->hdr.length - 2 + /* -2 accounts for ioam_trace_type,elts_left */ )) + { + s = format (s, " [%d] %U\n", elt_index, + format_ioam_data_list_element, + elt, &trace->ioam_trace_type); + elt_index++; + elt += trace_data_size_in_words; + } + return (s); +} + + +static clib_error_t * +vxlan_gpe_show_ioam_trace_cmd_fn (vlib_main_t * vm, + unformat_input_t * input, + vlib_cli_command_t * cmd) +{ + vxlan_gpe_ioam_trace_main_t *hm = &vxlan_gpe_ioam_trace_main; + u8 *s = 0; + int i = 0; + + for (i = 0; i < VXLAN_GPE_IOAM_TRACE_N_STATS; i++) + { + s = format (s, " %s - %lu\n", vxlan_gpe_ioam_trace_stats_strings[i], + hm->counters[i]); + } + + vlib_cli_output (vm, "%v", s); + vec_free (s); + return 0; +} + + +/* *INDENT-OFF* */ +VLIB_CLI_COMMAND (vxlan_gpe_show_ioam_trace_cmd, static) = { + .path = "show ioam vxlan-gpe trace", + .short_help = "iOAM trace statistics", + .function = vxlan_gpe_show_ioam_trace_cmd_fn, +}; +/* *INDENT-ON* */ + + +static clib_error_t * +vxlan_gpe_ioam_trace_init (vlib_main_t * vm) +{ + vxlan_gpe_ioam_trace_main_t *hm = &vxlan_gpe_ioam_trace_main; + clib_error_t *error; + + if ((error = vlib_call_init_function (vm, ip_main_init))) + return (error); + + if ((error = vlib_call_init_function (vm, ip6_lookup_init))) + return error; + + if ((error = vlib_call_init_function (vm, vxlan_gpe_init))) + return (error); + + hm->vlib_main = vm; + hm->vnet_main = vnet_get_main (); + memset (hm->counters, 0, sizeof (hm->counters)); + + if (vxlan_gpe_ioam_register_option + (VXLAN_GPE_OPTION_TYPE_IOAM_TRACE, + vxlan_gpe_ioam_trace_data_list_handler, + vxlan_gpe_ioam_trace_data_list_trace_handler) < 0) + return (clib_error_create + ("registration of VXLAN_GPE_OPTION_TYPE_IOAM_TRACE failed")); + + + if (vxlan_gpe_ioam_add_register_option + (VXLAN_GPE_OPTION_TYPE_IOAM_TRACE, + sizeof (vxlan_gpe_ioam_trace_option_t), + vxlan_gpe_ioam_trace_rewrite_handler) < 0) + return (clib_error_create + ("registration of VXLAN_GPE_OPTION_TYPE_IOAM_TRACE for rewrite failed")); + + + return (0); +} + +VLIB_INIT_FUNCTION (vxlan_gpe_ioam_trace_init); + +int +vxlan_gpe_trace_profile_cleanup (void) +{ + vxlan_gpe_ioam_main_t *hm = &vxlan_gpe_ioam_main; + + hm->options_size[VXLAN_GPE_OPTION_TYPE_IOAM_TRACE] = 0; + + return 0; + +} + +static int +vxlan_gpe_ioam_trace_get_sizeof_handler (u32 * result) +{ + u16 size = 0; + u8 trace_data_size = 0; + trace_profile *profile = NULL; + + *result = 0; + + profile = trace_profile_find (); + + if (PREDICT_FALSE (!profile)) + { + return (-1); + } + + trace_data_size = fetch_trace_data_size (profile->trace_type); + if (PREDICT_FALSE (trace_data_size == 0)) + return VNET_API_ERROR_INVALID_VALUE; + + if (PREDICT_FALSE (profile->num_elts * trace_data_size > 254)) + return VNET_API_ERROR_INVALID_VALUE; + + size += + sizeof (vxlan_gpe_ioam_trace_option_t) + + profile->num_elts * trace_data_size; + *result = size; + + return 0; +} + + +int +vxlan_gpe_trace_profile_setup (void) +{ + u32 trace_size = 0; + vxlan_gpe_ioam_main_t *hm = &vxlan_gpe_ioam_main; + + trace_profile *profile = NULL; + + + profile = trace_profile_find (); + + if (PREDICT_FALSE (!profile)) + { + return (-1); + } + + + if (vxlan_gpe_ioam_trace_get_sizeof_handler (&trace_size) < 0) + return (-1); + + hm->options_size[VXLAN_GPE_OPTION_TYPE_IOAM_TRACE] = trace_size; + + return (0); +} + + + +/* + * fd.io coding-style-patch-verification: ON + * + * Local Variables: + * eval: (c-set-style "gnu") + * End: + */ diff --git a/src/plugins/ioam/lib-vxlan-gpe/vxlan_gpe_ioam_util.h b/src/plugins/ioam/lib-vxlan-gpe/vxlan_gpe_ioam_util.h new file mode 100644 index 00000000..c0ad8d9d --- /dev/null +++ b/src/plugins/ioam/lib-vxlan-gpe/vxlan_gpe_ioam_util.h @@ -0,0 +1,172 @@ +/* + * Copyright (c) 2015 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ +#ifndef __included_vxlan_gpe_ioam_util_h__ +#define __included_vxlan_gpe_ioam_util_h__ + +#include +#include +#include + + +typedef struct +{ + u32 tunnel_index; + ioam_trace_t fmt_trace; +} vxlan_gpe_ioam_v4_trace_t; + + +static u8 * +format_vxlan_gpe_ioam_v4_trace (u8 * s, va_list * args) +{ + CLIB_UNUSED (vlib_main_t * vm) = va_arg (*args, vlib_main_t *); + CLIB_UNUSED (vlib_node_t * node) = va_arg (*args, vlib_node_t *); + vxlan_gpe_ioam_v4_trace_t *t1 = va_arg (*args, vxlan_gpe_ioam_v4_trace_t *); + ioam_trace_t *t = &(t1->fmt_trace); + vxlan_gpe_ioam_option_t *fmt_trace0; + vxlan_gpe_ioam_option_t *opt0, *limit0; + vxlan_gpe_ioam_main_t *hm = &vxlan_gpe_ioam_main; + + u8 type0; + + fmt_trace0 = (vxlan_gpe_ioam_option_t *) t->option_data; + + s = format (s, "VXLAN-GPE-IOAM: next_index %d len %d traced %d", + t->next_index, fmt_trace0->length, t->trace_len); + + opt0 = (vxlan_gpe_ioam_option_t *) (fmt_trace0 + 1); + limit0 = (vxlan_gpe_ioam_option_t *) ((u8 *) fmt_trace0) + t->trace_len; + + while (opt0 < limit0) + { + type0 = opt0->type; + switch (type0) + { + case 0: /* Pad, just stop */ + opt0 = (vxlan_gpe_ioam_option_t *) ((u8 *) opt0) + 1; + break; + + default: + if (hm->trace[type0]) + { + s = (*hm->trace[type0]) (s, opt0); + } + else + { + s = + format (s, "\n unrecognized option %d length %d", type0, + opt0->length); + } + opt0 = + (vxlan_gpe_ioam_option_t *) (((u8 *) opt0) + opt0->length + + sizeof (vxlan_gpe_ioam_option_t)); + break; + } + } + + s = format (s, "VXLAN-GPE-IOAM: tunnel %d", t1->tunnel_index); + return s; +} + + +always_inline void +vxlan_gpe_encap_decap_ioam_v4_one_inline (vlib_main_t * vm, + vlib_node_runtime_t * node, + vlib_buffer_t * b0, + u32 * next0, u32 drop_node_val, + u8 use_adj) +{ + ip4_header_t *ip0; + udp_header_t *udp_hdr0; + vxlan_gpe_header_t *gpe_hdr0; + vxlan_gpe_ioam_hdr_t *gpe_ioam0; + vxlan_gpe_ioam_option_t *opt0; + vxlan_gpe_ioam_option_t *limit0; + vxlan_gpe_ioam_main_t *hm = &vxlan_gpe_ioam_main; + + /* Populate the iOAM header */ + ip0 = vlib_buffer_get_current (b0); + udp_hdr0 = (udp_header_t *) (ip0 + 1); + gpe_hdr0 = (vxlan_gpe_header_t *) (udp_hdr0 + 1); + gpe_ioam0 = (vxlan_gpe_ioam_hdr_t *) (gpe_hdr0 + 1); + opt0 = (vxlan_gpe_ioam_option_t *) (gpe_ioam0 + 1); + limit0 = (vxlan_gpe_ioam_option_t *) ((u8 *) gpe_ioam0 + gpe_ioam0->length); + + /* + * Basic validity checks + */ + if (gpe_ioam0->length > clib_net_to_host_u16 (ip0->length)) + { + *next0 = drop_node_val; + return; + } + + /* Scan the set of h-b-h options, process ones that we understand */ + while (opt0 < limit0) + { + u8 type0; + type0 = opt0->type; + switch (type0) + { + case 0: /* Pad1 */ + opt0 = (vxlan_gpe_ioam_option_t *) ((u8 *) opt0) + 1; + continue; + case 1: /* PadN */ + break; + default: + if (hm->options[type0]) + { + if ((*hm->options[type0]) (b0, opt0, 1 /* is_ipv4 */ , + use_adj) < 0) + { + *next0 = drop_node_val; + return; + } + } + break; + } + opt0 = + (vxlan_gpe_ioam_option_t *) (((u8 *) opt0) + opt0->length + + sizeof (vxlan_gpe_ioam_hdr_t)); + } + + + if (PREDICT_FALSE (b0->flags & VLIB_BUFFER_IS_TRACED)) + { + vxlan_gpe_ioam_v4_trace_t *t = + vlib_add_trace (vm, node, b0, sizeof (*t)); + u32 trace_len = gpe_ioam0->length; + t->fmt_trace.next_index = *next0; + /* Capture the ioam option verbatim */ + trace_len = + trace_len < + ARRAY_LEN (t->fmt_trace. + option_data) ? trace_len : ARRAY_LEN (t->fmt_trace. + option_data); + t->fmt_trace.trace_len = trace_len; + clib_memcpy (&(t->fmt_trace.option_data), gpe_ioam0, trace_len); + } + return; +} + + +#endif + +/* + * fd.io coding-style-patch-verification: ON + * + * Local Variables: + * eval: (c-set-style "gnu") + * End: + */ diff --git a/src/plugins/ioam/lib-vxlan-gpe/vxlan_gpe_msg_enum.h b/src/plugins/ioam/lib-vxlan-gpe/vxlan_gpe_msg_enum.h new file mode 100644 index 00000000..cc0a10a3 --- /dev/null +++ b/src/plugins/ioam/lib-vxlan-gpe/vxlan_gpe_msg_enum.h @@ -0,0 +1,28 @@ +/* + * Copyright (c) 2016 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ +#ifndef included_vxlan_gpe_msg_enum_h +#define included_vxlan_gpe_msg_enum_h + +#include + +#define vl_msg_id(n,h) n, +typedef enum { +#include + /* We'll want to know how many messages IDs we need... */ + VL_MSG_FIRST_AVAILABLE, +} vl_msg_id_t; +#undef vl_msg_id + +#endif /* included_vxlan_gpe_msg_enum_h */ diff --git a/src/plugins/ioam/lib-vxlan-gpe/vxlan_gpe_test.c b/src/plugins/ioam/lib-vxlan-gpe/vxlan_gpe_test.c new file mode 100644 index 00000000..47253eb6 --- /dev/null +++ b/src/plugins/ioam/lib-vxlan-gpe/vxlan_gpe_test.c @@ -0,0 +1,600 @@ +/* + * Copyright (c) 2016 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ +/* + *------------------------------------------------------------------ + * vxlan_gpe_test.c - test harness for vxlan_gpe plugin + *------------------------------------------------------------------ + */ + +#include +#include +#include +#include +#include + +/* Declare message IDs */ +#include + +/* define message structures */ +#define vl_typedefs +#include +#undef vl_typedefs + +/* declare message handlers for each api */ + +#define vl_endianfun /* define message structures */ +#include +#undef vl_endianfun + +/* instantiate all the print functions we know about */ +#define vl_print(handle, ...) +#define vl_printfun +#include +#undef vl_printfun + +/* Get the API version number. */ +#define vl_api_version(n,v) static u32 api_version=(v); +#include +#undef vl_api_version +#include +#include + +typedef struct +{ + /* API message ID base */ + u16 msg_id_base; + vat_main_t *vat_main; +} vxlan_gpe_test_main_t; + +vxlan_gpe_test_main_t vxlan_gpe_test_main; + +#define foreach_standard_reply_retval_handler \ +_(vxlan_gpe_ioam_enable_reply) \ +_(vxlan_gpe_ioam_disable_reply) \ +_(vxlan_gpe_ioam_vni_enable_reply) \ +_(vxlan_gpe_ioam_vni_disable_reply) \ +_(vxlan_gpe_ioam_transit_enable_reply) \ +_(vxlan_gpe_ioam_transit_disable_reply) + +#define _(n) \ + static void vl_api_##n##_t_handler \ + (vl_api_##n##_t * mp) \ + { \ + vat_main_t * vam = vxlan_gpe_test_main.vat_main; \ + i32 retval = ntohl(mp->retval); \ + if (vam->async_mode) { \ + vam->async_errors += (retval < 0); \ + } else { \ + vam->retval = retval; \ + vam->result_ready = 1; \ + } \ + } +foreach_standard_reply_retval_handler; +#undef _ + +/* + * Table of message reply handlers, must include boilerplate handlers + * we just generated + */ +#define foreach_vpe_api_reply_msg \ +_(VXLAN_GPE_IOAM_ENABLE_REPLY, vxlan_gpe_ioam_enable_reply) \ +_(VXLAN_GPE_IOAM_DISABLE_REPLY, vxlan_gpe_ioam_disable_reply) \ +_(VXLAN_GPE_IOAM_VNI_ENABLE_REPLY, vxlan_gpe_ioam_vni_enable_reply) \ +_(VXLAN_GPE_IOAM_VNI_DISABLE_REPLY, vxlan_gpe_ioam_vni_disable_reply) \ +_(VXLAN_GPE_IOAM_TRANSIT_ENABLE_REPLY, vxlan_gpe_ioam_transit_enable_reply) \ +_(VXLAN_GPE_IOAM_TRANSIT_DISABLE_REPLY, vxlan_gpe_ioam_transit_disable_reply) \ + + +/* M: construct, but don't yet send a message */ + +#define M(T,t) \ +do { \ + vam->result_ready = 0; \ + mp = vl_msg_api_alloc(sizeof(*mp)); \ + memset (mp, 0, sizeof (*mp)); \ + mp->_vl_msg_id = ntohs (VL_API_##T + sm->msg_id_base); \ + mp->client_index = vam->my_client_index; \ +} while(0); + +#define M2(T,t,n) \ +do { \ + vam->result_ready = 0; \ + mp = vl_msg_api_alloc(sizeof(*mp)+(n)); \ + memset (mp, 0, sizeof (*mp)); \ + mp->_vl_msg_id = ntohs (VL_API_##T + sm->msg_id_base); \ + mp->client_index = vam->my_client_index; \ +} while(0); + +/* S: send a message */ +#define S (vl_msg_api_send_shmem (vam->vl_input_queue, (u8 *)&mp)) + +/* W: wait for results, with timeout */ +#define W \ +do { \ + timeout = vat_time_now (vam) + 1.0; \ + \ + while (vat_time_now (vam) < timeout) { \ + if (vam->result_ready == 1) { \ + return (vam->retval); \ + } \ + } \ + return -99; \ +} while(0); + + +static int +api_vxlan_gpe_ioam_enable (vat_main_t * vam) +{ + vxlan_gpe_test_main_t *sm = &vxlan_gpe_test_main; + + unformat_input_t *input = vam->input; + vl_api_vxlan_gpe_ioam_enable_t *mp; + f64 timeout; + u32 id = 0; + int has_trace_option = 0; + int has_pow_option = 0; + int has_ppc_option = 0; + + while (unformat_check_input (input) != UNFORMAT_END_OF_INPUT) + { + if (unformat (input, "trace")) + has_trace_option = 1; + else if (unformat (input, "pow")) + has_pow_option = 1; + else if (unformat (input, "ppc encap")) + has_ppc_option = PPC_ENCAP; + else if (unformat (input, "ppc decap")) + has_ppc_option = PPC_DECAP; + else if (unformat (input, "ppc none")) + has_ppc_option = PPC_NONE; + else + break; + } + M (VXLAN_GPE_IOAM_ENABLE, vxlan_gpe_ioam_enable); + mp->id = htons (id); + mp->trace_ppc = has_ppc_option; + mp->pow_enable = has_pow_option; + mp->trace_enable = has_trace_option; + + + S; + W; + + return (0); +} + + +static int +api_vxlan_gpe_ioam_disable (vat_main_t * vam) +{ + vxlan_gpe_test_main_t *sm = &vxlan_gpe_test_main; + vl_api_vxlan_gpe_ioam_disable_t *mp; + f64 timeout; + + M (VXLAN_GPE_IOAM_DISABLE, vxlan_gpe_ioam_disable); + S; + W; + return 0; +} + +static int +api_vxlan_gpe_ioam_vni_enable (vat_main_t * vam) +{ + vxlan_gpe_test_main_t *sm = &vxlan_gpe_test_main; + + unformat_input_t *line_input = vam->input; + vl_api_vxlan_gpe_ioam_vni_enable_t *mp; + ip4_address_t local4, remote4; + ip6_address_t local6, remote6; + u8 ipv4_set = 0, ipv6_set = 0; + u8 local_set = 0; + u8 remote_set = 0; + u32 vni; + u8 vni_set = 0; + f64 timeout; + + + while (unformat_check_input (line_input) != UNFORMAT_END_OF_INPUT) + { + if (unformat (line_input, "local %U", unformat_ip4_address, &local4)) + { + local_set = 1; + ipv4_set = 1; + } + else if (unformat (line_input, "remote %U", + unformat_ip4_address, &remote4)) + { + remote_set = 1; + ipv4_set = 1; + } + else if (unformat (line_input, "local %U", + unformat_ip6_address, &local6)) + { + local_set = 1; + ipv6_set = 1; + } + else if (unformat (line_input, "remote %U", + unformat_ip6_address, &remote6)) + { + remote_set = 1; + ipv6_set = 1; + } + + else if (unformat (line_input, "vni %d", &vni)) + vni_set = 1; + else + { + errmsg ("parse error '%U'\n", format_unformat_error, line_input); + return -99; + } + } + + if (local_set == 0) + { + errmsg ("tunnel local address not specified\n"); + return -99; + } + if (remote_set == 0) + { + errmsg ("tunnel remote address not specified\n"); + return -99; + } + if (ipv4_set && ipv6_set) + { + errmsg ("both IPv4 and IPv6 addresses specified"); + return -99; + } + + if (vni_set == 0) + { + errmsg ("vni not specified\n"); + return -99; + } + + M (VXLAN_GPE_IOAM_VNI_ENABLE, vxlan_gpe_ioam_vni_enable); + + + if (ipv6_set) + { + clib_memcpy (&mp->local, &local6, sizeof (local6)); + clib_memcpy (&mp->remote, &remote6, sizeof (remote6)); + } + else + { + clib_memcpy (&mp->local, &local4, sizeof (local4)); + clib_memcpy (&mp->remote, &remote4, sizeof (remote4)); + } + + mp->vni = ntohl (vni); + mp->is_ipv6 = ipv6_set; + + S; + W; + + return (0); +} + +static int +api_vxlan_gpe_ioam_vni_disable (vat_main_t * vam) +{ + vxlan_gpe_test_main_t *sm = &vxlan_gpe_test_main; + + unformat_input_t *line_input = vam->input; + vl_api_vxlan_gpe_ioam_vni_disable_t *mp; + ip4_address_t local4, remote4; + ip6_address_t local6, remote6; + u8 ipv4_set = 0, ipv6_set = 0; + u8 local_set = 0; + u8 remote_set = 0; + u32 vni; + u8 vni_set = 0; + f64 timeout; + + + while (unformat_check_input (line_input) != UNFORMAT_END_OF_INPUT) + { + if (unformat (line_input, "local %U", unformat_ip4_address, &local4)) + { + local_set = 1; + ipv4_set = 1; + } + else if (unformat (line_input, "remote %U", + unformat_ip4_address, &remote4)) + { + remote_set = 1; + ipv4_set = 1; + } + else if (unformat (line_input, "local %U", + unformat_ip6_address, &local6)) + { + local_set = 1; + ipv6_set = 1; + } + else if (unformat (line_input, "remote %U", + unformat_ip6_address, &remote6)) + { + remote_set = 1; + ipv6_set = 1; + } + + else if (unformat (line_input, "vni %d", &vni)) + vni_set = 1; + else + { + errmsg ("parse error '%U'\n", format_unformat_error, line_input); + return -99; + } + } + + if (local_set == 0) + { + errmsg ("tunnel local address not specified\n"); + return -99; + } + if (remote_set == 0) + { + errmsg ("tunnel remote address not specified\n"); + return -99; + } + if (ipv4_set && ipv6_set) + { + errmsg ("both IPv4 and IPv6 addresses specified"); + return -99; + } + + if (vni_set == 0) + { + errmsg ("vni not specified\n"); + return -99; + } + + M (VXLAN_GPE_IOAM_VNI_DISABLE, vxlan_gpe_ioam_vni_disable); + + + if (ipv6_set) + { + clib_memcpy (&mp->local, &local6, sizeof (local6)); + clib_memcpy (&mp->remote, &remote6, sizeof (remote6)); + } + else + { + clib_memcpy (&mp->local, &local4, sizeof (local4)); + clib_memcpy (&mp->remote, &remote4, sizeof (remote4)); + } + + mp->vni = ntohl (vni); + mp->is_ipv6 = ipv6_set; + + S; + W; + + return 0; +} + +static int +api_vxlan_gpe_ioam_transit_enable (vat_main_t * vam) +{ + vxlan_gpe_test_main_t *sm = &vxlan_gpe_test_main; + + unformat_input_t *line_input = vam->input; + vl_api_vxlan_gpe_ioam_transit_enable_t *mp; + ip4_address_t local4; + ip6_address_t local6; + u8 ipv4_set = 0, ipv6_set = 0; + u8 local_set = 0; + u32 outer_fib_index = 0; + f64 timeout; + + + while (unformat_check_input (line_input) != UNFORMAT_END_OF_INPUT) + { + if (unformat (line_input, "dst-ip %U", unformat_ip4_address, &local4)) + { + local_set = 1; + ipv4_set = 1; + } + else if (unformat (line_input, "dst-ip %U", + unformat_ip6_address, &local6)) + { + local_set = 1; + ipv6_set = 1; + } + + else if (unformat (line_input, "outer-fib-index %d", &outer_fib_index)) + ; + else + { + errmsg ("parse error '%U'\n", format_unformat_error, line_input); + return -99; + } + } + + if (local_set == 0) + { + errmsg ("destination address not specified\n"); + return -99; + } + if (ipv4_set && ipv6_set) + { + errmsg ("both IPv4 and IPv6 addresses specified"); + return -99; + } + + + M (VXLAN_GPE_IOAM_TRANSIT_ENABLE, vxlan_gpe_ioam_transit_enable); + + + if (ipv6_set) + { + errmsg ("IPv6 currently unsupported"); + return -1; + } + else + { + clib_memcpy (&mp->dst_addr, &local4, sizeof (local4)); + } + + mp->outer_fib_index = htonl (outer_fib_index); + mp->is_ipv6 = ipv6_set; + + S; + W; + + return (0); +} + +static int +api_vxlan_gpe_ioam_transit_disable (vat_main_t * vam) +{ + vxlan_gpe_test_main_t *sm = &vxlan_gpe_test_main; + + unformat_input_t *line_input = vam->input; + vl_api_vxlan_gpe_ioam_transit_disable_t *mp; + ip4_address_t local4; + ip6_address_t local6; + u8 ipv4_set = 0, ipv6_set = 0; + u8 local_set = 0; + u32 outer_fib_index; + f64 timeout; + + + while (unformat_check_input (line_input) != UNFORMAT_END_OF_INPUT) + { + if (unformat (line_input, "dst-ip %U", unformat_ip4_address, &local4)) + { + local_set = 1; + ipv4_set = 1; + } + else if (unformat (line_input, "dst-ip %U", + unformat_ip6_address, &local6)) + { + local_set = 1; + ipv6_set = 1; + } + + else if (unformat (line_input, "outer-fib-index %d", &outer_fib_index)) + ; + else + { + errmsg ("parse error '%U'\n", format_unformat_error, line_input); + return -99; + } + } + + if (local_set == 0) + { + errmsg ("destination address not specified\n"); + return -99; + } + if (ipv4_set && ipv6_set) + { + errmsg ("both IPv4 and IPv6 addresses specified"); + return -99; + } + + + M (VXLAN_GPE_IOAM_TRANSIT_DISABLE, vxlan_gpe_ioam_transit_disable); + + + if (ipv6_set) + { + return -1; + } + else + { + clib_memcpy (&mp->dst_addr, &local4, sizeof (local4)); + } + + mp->outer_fib_index = htonl (outer_fib_index); + mp->is_ipv6 = ipv6_set; + + S; + W; + + + return (0); +} + +/* + * List of messages that the api test plugin sends, + * and that the data plane plugin processes + */ +#define foreach_vpe_api_msg \ +_(vxlan_gpe_ioam_enable, ""\ + "[trace] [pow] [ppc ]") \ +_(vxlan_gpe_ioam_disable, "") \ +_(vxlan_gpe_ioam_vni_enable, ""\ + "local remote vni ") \ +_(vxlan_gpe_ioam_vni_disable, ""\ + "local remote vni ") \ +_(vxlan_gpe_ioam_transit_enable, ""\ + "dst-ip [outer-fib-index ]") \ +_(vxlan_gpe_ioam_transit_disable, ""\ + "dst-ip [outer-fib-index ]") \ + + +void +vat_api_hookup (vat_main_t * vam) +{ + vxlan_gpe_test_main_t *sm = &vxlan_gpe_test_main; + /* Hook up handlers for replies from the data plane plug-in */ +#define _(N,n) \ + vl_msg_api_set_handlers((VL_API_##N + sm->msg_id_base), \ + #n, \ + vl_api_##n##_t_handler, \ + vl_noop_handler, \ + vl_api_##n##_t_endian, \ + vl_api_##n##_t_print, \ + sizeof(vl_api_##n##_t), 1); + foreach_vpe_api_reply_msg; +#undef _ + + /* API messages we can send */ +#define _(n,h) hash_set_mem (vam->function_by_name, #n, api_##n); + foreach_vpe_api_msg; +#undef _ + + /* Help strings */ +#define _(n,h) hash_set_mem (vam->help_by_name, #n, h); + foreach_vpe_api_msg; +#undef _ +} + +clib_error_t * +vat_plugin_register (vat_main_t * vam) +{ + vxlan_gpe_test_main_t *sm = &vxlan_gpe_test_main; + u8 *name; + + sm->vat_main = vam; + + name = format (0, "ioam_vxlan_gpe_%08x%c", api_version, 0); + sm->msg_id_base = vl_client_get_first_plugin_msg_id ((char *) name); + + if (sm->msg_id_base != (u16) ~ 0) + vat_api_hookup (vam); + + vec_free (name); + + return 0; +} + +/* + * fd.io coding-style-patch-verification: ON + * + * Local Variables: + * eval: (c-set-style "gnu") + * End: + */ diff --git a/src/plugins/lb.am b/src/plugins/lb.am new file mode 100644 index 00000000..352358fa --- /dev/null +++ b/src/plugins/lb.am @@ -0,0 +1,42 @@ +# Copyright (c) 2016 Cisco Systems, Inc. +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at: +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +vppapitestplugins_LTLIBRARIES += lb_test_plugin.la +vppplugins_LTLIBRARIES += lb_plugin.la + +lb_plugin_la_SOURCES = \ + lb/lb.c \ + lb/node.c \ + lb/cli.c \ + lb/util.c \ + lb/refcount.c \ + lb/api.c + +BUILT_SOURCES += \ + lb/lb.api.h \ + lb/lb.api.json + +API_FILES += lb/lb.api + +noinst_HEADERS += \ + lb/lb.h \ + lb/util.h \ + lb/refcount.h \ + lb/lbhash.h \ + lb/lb.api.h + +lb_test_plugin_la_SOURCES = \ + lb/lb_test.c \ + lb/lb_plugin.api.h + +# vi:syntax=automake diff --git a/src/plugins/lb/api.c b/src/plugins/lb/api.c new file mode 100644 index 00000000..06c53fa1 --- /dev/null +++ b/src/plugins/lb/api.c @@ -0,0 +1,228 @@ +/* + * Copyright (c) 2016 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +#include + +#include +#include +#include +#include +#include + +#define vl_msg_id(n,h) n, +typedef enum { +#include + /* We'll want to know how many messages IDs we need... */ + VL_MSG_FIRST_AVAILABLE, +} vl_msg_id_t; +#undef vl_msg_id + + +/* define message structures */ +#define vl_typedefs +#include +#undef vl_typedefs + +/* define generated endian-swappers */ +#define vl_endianfun +#include +#undef vl_endianfun + +#define vl_print(handle, ...) vlib_cli_output (handle, __VA_ARGS__) + +/* Get the API version number */ +#define vl_api_version(n,v) static u32 api_version=(v); +#include +#undef vl_api_version + +#define vl_msg_name_crc_list +#include +#undef vl_msg_name_crc_list + +static void +setup_message_id_table (lb_main_t * lbm, api_main_t * am) +{ +#define _(id,n,crc) \ + vl_msg_api_add_msg_name_crc (am, #n "_" #crc, id + lbm->msg_id_base); + foreach_vl_msg_name_crc_lb; +#undef _ +} + +/* Macro to finish up custom dump fns */ +#define FINISH \ + vec_add1 (s, 0); \ + vl_print (handle, (char *)s); \ + vec_free (s); \ + return handle; + +/* + * A handy macro to set up a message reply. + * Assumes that the following variables are available: + * mp - pointer to request message + * rmp - pointer to reply message type + * rv - return value + */ + +#define REPLY_MACRO(t) \ +do { \ + unix_shared_memory_queue_t * q = \ + vl_api_client_index_to_input_queue (mp->client_index); \ + if (!q) \ + return; \ + \ + rmp = vl_msg_api_alloc (sizeof (*rmp)); \ + rmp->_vl_msg_id = ntohs((t)+lbm->msg_id_base); \ + rmp->context = mp->context; \ + rmp->retval = ntohl(rv); \ + \ + vl_msg_api_send_shmem (q, (u8 *)&rmp); \ +} while(0); + +static void +vl_api_lb_conf_t_handler +(vl_api_lb_conf_t * mp) +{ + lb_main_t *lbm = &lb_main; + vl_api_lb_conf_reply_t * rmp; + int rv = 0; + + rv = lb_conf((ip4_address_t *)&mp->ip4_src_address, + (ip6_address_t *)mp->ip6_src_address, + mp->sticky_buckets_per_core, + mp->flow_timeout); + + REPLY_MACRO (VL_API_LB_CONF_REPLY); +} + +static void *vl_api_lb_conf_t_print +(vl_api_lb_conf_t *mp, void * handle) +{ + u8 * s; + s = format (0, "SCRIPT: lb_conf "); + s = format (s, "%U ", format_ip4_address, (ip4_address_t *)&mp->ip4_src_address); + s = format (s, "%U ", format_ip6_address, (ip6_address_t *)mp->ip6_src_address); + s = format (s, "%u ", mp->sticky_buckets_per_core); + s = format (s, "%u ", mp->flow_timeout); + FINISH; +} + + +static void +vl_api_lb_add_del_vip_t_handler +(vl_api_lb_add_del_vip_t * mp) +{ + lb_main_t *lbm = &lb_main; + vl_api_lb_conf_reply_t * rmp; + int rv = 0; + ip46_address_t prefix; + memcpy(&prefix.ip6, mp->ip_prefix, sizeof(prefix.ip6)); + + if (mp->is_del) { + u32 vip_index; + if (!(rv = lb_vip_find_index(&prefix, mp->prefix_length, &vip_index))) + rv = lb_vip_del(vip_index); + } else { + u32 vip_index; + lb_vip_type_t type; + if (ip46_prefix_is_ip4(&prefix, mp->prefix_length)) { + type = mp->is_gre4?LB_VIP_TYPE_IP4_GRE4:LB_VIP_TYPE_IP4_GRE6; + } else { + type = mp->is_gre4?LB_VIP_TYPE_IP6_GRE4:LB_VIP_TYPE_IP6_GRE6; + } + + rv = lb_vip_add(&prefix, mp->prefix_length, type, + mp->new_flows_table_length, &vip_index); + } + REPLY_MACRO (VL_API_LB_CONF_REPLY); +} + +static void *vl_api_lb_add_del_vip_t_print +(vl_api_lb_add_del_vip_t *mp, void * handle) +{ + u8 * s; + s = format (0, "SCRIPT: lb_add_del_vip "); + s = format (s, "%U ", format_ip46_prefix, + (ip46_address_t *)mp->ip_prefix, mp->prefix_length, IP46_TYPE_ANY); + s = format (s, "%s ", mp->is_gre4?"gre4":"gre6"); + s = format (s, "%u ", mp->new_flows_table_length); + s = format (s, "%s ", mp->is_del?"del":"add"); + FINISH; +} + +static void +vl_api_lb_add_del_as_t_handler +(vl_api_lb_add_del_as_t * mp) +{ + lb_main_t *lbm = &lb_main; + vl_api_lb_conf_reply_t * rmp; + int rv = 0; + u32 vip_index; + if ((rv = lb_vip_find_index((ip46_address_t *)mp->vip_ip_prefix, + mp->vip_prefix_length, &vip_index))) + goto done; + + if (mp->is_del) + rv = lb_vip_del_ass(vip_index, (ip46_address_t *)mp->as_address, 1); + else + rv = lb_vip_add_ass(vip_index, (ip46_address_t *)mp->as_address, 1); + +done: + REPLY_MACRO (VL_API_LB_CONF_REPLY); +} + +static void *vl_api_lb_add_del_as_t_print +(vl_api_lb_add_del_as_t *mp, void * handle) +{ + u8 * s; + s = format (0, "SCRIPT: lb_add_del_as "); + s = format (s, "%U ", format_ip46_prefix, + (ip46_address_t *)mp->vip_ip_prefix, mp->vip_prefix_length, IP46_TYPE_ANY); + s = format (s, "%U ", format_ip46_address, + (ip46_address_t *)mp->as_address, IP46_TYPE_ANY); + s = format (s, "%s ", mp->is_del?"del":"add"); + FINISH; +} + +/* List of message types that this plugin understands */ +#define foreach_lb_plugin_api_msg \ +_(LB_CONF, lb_conf) \ +_(LB_ADD_DEL_VIP, lb_add_del_vip) \ +_(LB_ADD_DEL_AS, lb_add_del_as) + +static clib_error_t * lb_api_init (vlib_main_t * vm) +{ + lb_main_t *lbm = &lb_main; + u8 *name = format (0, "lb_%08x%c", api_version, 0); + lbm->msg_id_base = vl_msg_api_get_msg_ids + ((char *) name, VL_MSG_FIRST_AVAILABLE); + +#define _(N,n) \ + vl_msg_api_set_handlers((VL_API_##N + lbm->msg_id_base), \ + #n, \ + vl_api_##n##_t_handler, \ + vl_noop_handler, \ + vl_api_##n##_t_endian, \ + vl_api_##n##_t_print, \ + sizeof(vl_api_##n##_t), 1); + foreach_lb_plugin_api_msg; +#undef _ + + /* Add our API messages to the global name_crc hash table */ + setup_message_id_table (lbm, &api_main); + + return 0; +} + +VLIB_INIT_FUNCTION (lb_api_init); diff --git a/src/plugins/lb/cli.c b/src/plugins/lb/cli.c new file mode 100644 index 00000000..b59c6426 --- /dev/null +++ b/src/plugins/lb/cli.c @@ -0,0 +1,250 @@ +/* + * Copyright (c) 2016 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +#include +#include + +static clib_error_t * +lb_vip_command_fn (vlib_main_t * vm, + unformat_input_t * input, vlib_cli_command_t * cmd) +{ + unformat_input_t _line_input, *line_input = &_line_input; + ip46_address_t prefix; + u8 plen; + u32 new_len = 1024; + u8 del = 0; + int ret; + u32 gre4 = 0; + lb_vip_type_t type; + + if (!unformat_user (input, unformat_line_input, line_input)) + return 0; + + if (!unformat(line_input, "%U", unformat_ip46_prefix, &prefix, &plen, IP46_TYPE_ANY, &plen)) + return clib_error_return (0, "invalid vip prefix: '%U'", + format_unformat_error, line_input); + + while (unformat_check_input (line_input) != UNFORMAT_END_OF_INPUT) + { + if (unformat(line_input, "new_len %d", &new_len)) + ; + else if (unformat(line_input, "del")) + del = 1; + else if (unformat(line_input, "encap gre4")) + gre4 = 1; + else if (unformat(line_input, "encap gre6")) + gre4 = 0; + else + return clib_error_return (0, "parse error: '%U'", + format_unformat_error, line_input); + } + + unformat_free (line_input); + + + if (ip46_prefix_is_ip4(&prefix, plen)) { + type = (gre4)?LB_VIP_TYPE_IP4_GRE4:LB_VIP_TYPE_IP4_GRE6; + } else { + type = (gre4)?LB_VIP_TYPE_IP6_GRE4:LB_VIP_TYPE_IP6_GRE6; + } + + lb_garbage_collection(); + + u32 index; + if (!del) { + if ((ret = lb_vip_add(&prefix, plen, type, new_len, &index))) { + return clib_error_return (0, "lb_vip_add error %d", ret); + } else { + vlib_cli_output(vm, "lb_vip_add ok %d", index); + } + } else { + if ((ret = lb_vip_find_index(&prefix, plen, &index))) + return clib_error_return (0, "lb_vip_find_index error %d", ret); + else if ((ret = lb_vip_del(index))) + return clib_error_return (0, "lb_vip_del error %d", ret); + } + return NULL; +} + +VLIB_CLI_COMMAND (lb_vip_command, static) = +{ + .path = "lb vip", + .short_help = "lb vip [encap (gre6|gre4)] [new_len ] [del]", + .function = lb_vip_command_fn, +}; + +static clib_error_t * +lb_as_command_fn (vlib_main_t * vm, + unformat_input_t * input, vlib_cli_command_t * cmd) +{ + unformat_input_t _line_input, *line_input = &_line_input; + ip46_address_t vip_prefix, as_addr; + u8 vip_plen; + ip46_address_t *as_array = 0; + u32 vip_index; + u8 del = 0; + int ret; + + if (!unformat_user (input, unformat_line_input, line_input)) + return 0; + + if (!unformat(line_input, "%U", unformat_ip46_prefix, &vip_prefix, &vip_plen, IP46_TYPE_ANY)) + return clib_error_return (0, "invalid as address: '%U'", + format_unformat_error, line_input); + + if ((ret = lb_vip_find_index(&vip_prefix, vip_plen, &vip_index))) + return clib_error_return (0, "lb_vip_find_index error %d", ret); + + while (unformat_check_input (line_input) != UNFORMAT_END_OF_INPUT) + { + if (unformat(line_input, "%U", unformat_ip46_address, &as_addr, IP46_TYPE_ANY)) { + vec_add1(as_array, as_addr); + } else if (unformat(line_input, "del")) { + del = 1; + } else { + vec_free(as_array); + return clib_error_return (0, "parse error: '%U'", + format_unformat_error, line_input); + } + } + + if (!vec_len(as_array)) { + vec_free(as_array); + return clib_error_return (0, "No AS address provided"); + } + + lb_garbage_collection(); + clib_warning("vip index is %d", vip_index); + + if (del) { + if ((ret = lb_vip_del_ass(vip_index, as_array, vec_len(as_array)))) { + vec_free(as_array); + return clib_error_return (0, "lb_vip_del_ass error %d", ret); + } + } else { + if ((ret = lb_vip_add_ass(vip_index, as_array, vec_len(as_array)))) { + vec_free(as_array); + return clib_error_return (0, "lb_vip_add_ass error %d", ret); + } + } + + vec_free(as_array); + return 0; +} + +VLIB_CLI_COMMAND (lb_as_command, static) = +{ + .path = "lb as", + .short_help = "lb as [
[
[...]]] [del]", + .function = lb_as_command_fn, +}; + +static clib_error_t * +lb_conf_command_fn (vlib_main_t * vm, + unformat_input_t * input, vlib_cli_command_t * cmd) +{ + lb_main_t *lbm = &lb_main; + unformat_input_t _line_input, *line_input = &_line_input; + ip4_address_t ip4 = lbm->ip4_src_address; + ip6_address_t ip6 = lbm->ip6_src_address; + u32 per_cpu_sticky_buckets = lbm->per_cpu_sticky_buckets; + u32 per_cpu_sticky_buckets_log2 = 0; + u32 flow_timeout = lbm->flow_timeout; + int ret; + + if (!unformat_user (input, unformat_line_input, line_input)) + return 0; + + while (unformat_check_input (line_input) != UNFORMAT_END_OF_INPUT) + { + if (unformat(line_input, "ip4-src-address %U", unformat_ip4_address, &ip4)) + ; + else if (unformat(line_input, "ip6-src-address %U", unformat_ip6_address, &ip6)) + ; + else if (unformat(line_input, "buckets %d", &per_cpu_sticky_buckets)) + ; + else if (unformat(line_input, "buckets-log2 %d", &per_cpu_sticky_buckets_log2)) { + if (per_cpu_sticky_buckets_log2 >= 32) + return clib_error_return (0, "buckets-log2 value is too high"); + per_cpu_sticky_buckets = 1 << per_cpu_sticky_buckets_log2; + } else if (unformat(line_input, "timeout %d", &flow_timeout)) + ; + else + return clib_error_return (0, "parse error: '%U'", + format_unformat_error, line_input); + } + + unformat_free (line_input); + + lb_garbage_collection(); + + if ((ret = lb_conf(&ip4, &ip6, per_cpu_sticky_buckets, flow_timeout))) + return clib_error_return (0, "lb_conf error %d", ret); + + return NULL; +} + +VLIB_CLI_COMMAND (lb_conf_command, static) = +{ + .path = "lb conf", + .short_help = "lb conf [ip4-src-address ] [ip6-src-address ] [buckets ] [timeout ]", + .function = lb_conf_command_fn, +}; + +static clib_error_t * +lb_show_command_fn (vlib_main_t * vm, + unformat_input_t * input, vlib_cli_command_t * cmd) +{ + vlib_cli_output(vm, "%U", format_lb_main); + return NULL; +} + + +VLIB_CLI_COMMAND (lb_show_command, static) = +{ + .path = "show lb", + .short_help = "show lb", + .function = lb_show_command_fn, +}; + +static clib_error_t * +lb_show_vips_command_fn (vlib_main_t * vm, + unformat_input_t * input, vlib_cli_command_t * cmd) +{ + unformat_input_t line_input; + lb_main_t *lbm = &lb_main; + lb_vip_t *vip; + u8 verbose = 0; + + if (!unformat_user (input, unformat_line_input, &line_input)) + return 0; + + if (unformat(&line_input, "verbose")) + verbose = 1; + + pool_foreach(vip, lbm->vips, { + vlib_cli_output(vm, "%U\n", verbose?format_lb_vip_detailed:format_lb_vip, vip); + }); + + unformat_free (&line_input); + return NULL; +} + +VLIB_CLI_COMMAND (lb_show_vips_command, static) = +{ + .path = "show lb vips", + .short_help = "show lb vips [verbose]", + .function = lb_show_vips_command_fn, +}; diff --git a/src/plugins/lb/lb.api b/src/plugins/lb/lb.api new file mode 100644 index 00000000..39ee3c8f --- /dev/null +++ b/src/plugins/lb/lb.api @@ -0,0 +1,71 @@ +/** \brief Configure Load-Balancer global parameters + @param client_index - opaque cookie to identify the sender + @param context - sender context, to match reply w/ request + @param ip4_src_address - IPv4 address to be used as source for IPv4 GRE traffic. + @param ip6_src_address - IPv6 address to be used as source for IPv6 GRE traffic. + @param n_sticky_buckets - Number of buckets *per worker thread* in the + established flow table (must be power of 2). + @param flow_timeout - Time in seconds after which, if no packet is received + for a given flow, the flow is removed from the established flow table. +*/ +define lb_conf +{ + u32 client_index; + u32 context; + u32 ip4_src_address; + u8 ip6_src_address[16]; + u32 sticky_buckets_per_core; + u32 flow_timeout; +}; + +define lb_conf_reply { + u32 context; + i32 retval; +}; + +/** \brief Add a virtual address (or prefix) + @param client_index - opaque cookie to identify the sender + @param context - sender context, to match reply w/ request + @param ip_prefix - IP address (IPv4 in lower order 32 bits). + @param prefix_length - IP prefix length (96 + 'IPv4 prefix length' for IPv4). + @param is_gre4 - Encap is ip4 GRE (ip6 GRE otherwise). + @param new_flows_table_length - Size of the new connections flow table used + for this VIP (must be power of 2). + @param is_del - The VIP should be removed. +*/ +define lb_add_del_vip { + u32 client_index; + u32 context; + u8 ip_prefix[16]; + u8 prefix_length; + u8 is_gre4; + u32 new_flows_table_length; + u8 is_del; +}; + +define lb_add_del_vip_reply { + u32 context; + i32 retval; +}; + +/** \brief Add an application server for a given VIP + @param client_index - opaque cookie to identify the sender + @param context - sender context, to match reply w/ request + @param vip_ip_prefix - VIP IP address (IPv4 in lower order 32 bits). + @param vip_ip_prefix - VIP IP prefix length (96 + 'IPv4 prefix length' for IPv4). + @param as_address - The application server address (IPv4 in lower order 32 bits). + @param is_del - The AS should be removed. +*/ +define lb_add_del_as { + u32 client_index; + u32 context; + u8 vip_ip_prefix[16]; + u8 vip_prefix_length; + u8 as_address[16]; + u8 is_del; +}; + +define lb_add_del_as_reply { + u32 context; + i32 retval; +}; diff --git a/src/plugins/lb/lb.c b/src/plugins/lb/lb.c new file mode 100644 index 00000000..1d9b9870 --- /dev/null +++ b/src/plugins/lb/lb.c @@ -0,0 +1,844 @@ +/* + * Copyright (c) 2016 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +#include +#include +#include + +//GC runs at most once every so many seconds +#define LB_GARBAGE_RUN 60 + +//After so many seconds. It is assumed that inter-core race condition will not occur. +#define LB_CONCURRENCY_TIMEOUT 10 + +lb_main_t lb_main; + +#define lb_get_writer_lock() do {} while(__sync_lock_test_and_set (lb_main.writer_lock, 1)) +#define lb_put_writer_lock() lb_main.writer_lock[0] = 0 + +static void lb_as_stack (lb_as_t *as); + + +const static char * const lb_dpo_gre4_ip4[] = { "lb4-gre4" , NULL }; +const static char * const lb_dpo_gre4_ip6[] = { "lb6-gre4" , NULL }; +const static char* const * const lb_dpo_gre4_nodes[DPO_PROTO_NUM] = + { + [DPO_PROTO_IP4] = lb_dpo_gre4_ip4, + [DPO_PROTO_IP6] = lb_dpo_gre4_ip6, + }; + +const static char * const lb_dpo_gre6_ip4[] = { "lb4-gre6" , NULL }; +const static char * const lb_dpo_gre6_ip6[] = { "lb6-gre6" , NULL }; +const static char* const * const lb_dpo_gre6_nodes[DPO_PROTO_NUM] = + { + [DPO_PROTO_IP4] = lb_dpo_gre6_ip4, + [DPO_PROTO_IP6] = lb_dpo_gre6_ip6, + }; + +u32 lb_hash_time_now(vlib_main_t * vm) +{ + return (u32) (vlib_time_now(vm) + 10000); +} + +u8 *format_lb_main (u8 * s, va_list * args) +{ + vlib_thread_main_t *tm = vlib_get_thread_main(); + lb_main_t *lbm = &lb_main; + s = format(s, "lb_main"); + s = format(s, " ip4-src-address: %U \n", format_ip4_address, &lbm->ip4_src_address); + s = format(s, " ip6-src-address: %U \n", format_ip6_address, &lbm->ip6_src_address); + s = format(s, " #vips: %u\n", pool_elts(lbm->vips)); + s = format(s, " #ass: %u\n", pool_elts(lbm->ass) - 1); + + u32 cpu_index; + for(cpu_index = 0; cpu_index < tm->n_vlib_mains; cpu_index++ ) { + lb_hash_t *h = lbm->per_cpu[cpu_index].sticky_ht; + if (h) { + s = format(s, "core %d\n", cpu_index); + s = format(s, " timeout: %ds\n", h->timeout); + s = format(s, " usage: %d / %d\n", lb_hash_elts(h, lb_hash_time_now(vlib_get_main())), lb_hash_size(h)); + } + } + + return s; +} + +static char *lb_vip_type_strings[] = { + [LB_VIP_TYPE_IP6_GRE6] = "ip6-gre6", + [LB_VIP_TYPE_IP6_GRE4] = "ip6-gre4", + [LB_VIP_TYPE_IP4_GRE6] = "ip4-gre6", + [LB_VIP_TYPE_IP4_GRE4] = "ip4-gre4", +}; + +u8 *format_lb_vip_type (u8 * s, va_list * args) +{ + lb_vip_type_t vipt = va_arg (*args, lb_vip_type_t); + u32 i; + for (i=0; itype, + format_ip46_prefix, &vip->prefix, vip->plen, IP46_TYPE_ANY, + vip->new_flow_table_mask + 1, + pool_elts(vip->as_indexes), + (vip->flags & LB_VIP_FLAGS_USED)?"":" removed"); +} + +u8 *format_lb_as (u8 * s, va_list * args) +{ + lb_as_t *as = va_arg (*args, lb_as_t *); + return format(s, "%U %s", format_ip46_address, + &as->address, IP46_TYPE_ANY, + (as->flags & LB_AS_FLAGS_USED)?"used":"removed"); +} + +u8 *format_lb_vip_detailed (u8 * s, va_list * args) +{ + lb_main_t *lbm = &lb_main; + lb_vip_t *vip = va_arg (*args, lb_vip_t *); + uword indent = format_get_indent (s); + + s = format(s, "%U %U [%u] %U%s\n" + "%U new_size:%u\n", + format_white_space, indent, + format_lb_vip_type, vip->type, + vip - lbm->vips, format_ip46_prefix, &vip->prefix, vip->plen, IP46_TYPE_ANY, + (vip->flags & LB_VIP_FLAGS_USED)?"":" removed", + format_white_space, indent, + vip->new_flow_table_mask + 1); + + //Print counters + s = format(s, "%U counters:\n", + format_white_space, indent); + u32 i; + for (i=0; ivip_counters[i].name, + vlib_get_simple_counter(&lbm->vip_counters[i], vip - lbm->vips)); + + + s = format(s, "%U #as:%u\n", + format_white_space, indent, + pool_elts(vip->as_indexes)); + + //Let's count the buckets for each AS + u32 *count = 0; + vec_validate(count, pool_len(lbm->ass)); //Possibly big alloc for not much... + lb_new_flow_entry_t *nfe; + vec_foreach(nfe, vip->new_flow_table) + count[nfe->as_index]++; + + lb_as_t *as; + u32 *as_index; + pool_foreach(as_index, vip->as_indexes, { + as = &lbm->ass[*as_index]; + s = format(s, "%U %U %d buckets %d flows dpo:%u %s\n", + format_white_space, indent, + format_ip46_address, &as->address, IP46_TYPE_ANY, + count[as - lbm->ass], + vlib_refcount_get(&lbm->as_refcount, as - lbm->ass), + as->dpo.dpoi_index, + (as->flags & LB_AS_FLAGS_USED)?"used":" removed"); + }); + + vec_free(count); + + /* + s = format(s, "%U new flows table:\n", format_white_space, indent); + lb_new_flow_entry_t *nfe; + vec_foreach(nfe, vip->new_flow_table) { + s = format(s, "%U %d: %d\n", format_white_space, indent, nfe - vip->new_flow_table, nfe->as_index); + } + */ + return s; +} + +typedef struct { + u32 as_index; + u32 last; + u32 skip; +} lb_pseudorand_t; + +static int lb_pseudorand_compare(void *a, void *b) +{ + lb_as_t *asa, *asb; + lb_main_t *lbm = &lb_main; + asa = &lbm->ass[((lb_pseudorand_t *)a)->as_index]; + asb = &lbm->ass[((lb_pseudorand_t *)b)->as_index]; + return memcmp(&asa->address, &asb->address, sizeof(asb->address)); +} + +static void lb_vip_garbage_collection(lb_vip_t *vip) +{ + lb_main_t *lbm = &lb_main; + ASSERT (lbm->writer_lock[0]); + + u32 now = (u32) vlib_time_now(vlib_get_main()); + if (!clib_u32_loop_gt(now, vip->last_garbage_collection + LB_GARBAGE_RUN)) + return; + + vip->last_garbage_collection = now; + lb_as_t *as; + u32 *as_index; + pool_foreach(as_index, vip->as_indexes, { + as = &lbm->ass[*as_index]; + if (!(as->flags & LB_AS_FLAGS_USED) && //Not used + clib_u32_loop_gt(now, as->last_used + LB_CONCURRENCY_TIMEOUT) && //Not recently used + (vlib_refcount_get(&lbm->as_refcount, as - lbm->ass) == 0)) + { //Not referenced + fib_entry_child_remove(as->next_hop_fib_entry_index, + as->next_hop_child_index); + fib_table_entry_delete_index(as->next_hop_fib_entry_index, + FIB_SOURCE_RR); + as->next_hop_fib_entry_index = FIB_NODE_INDEX_INVALID; + + pool_put(vip->as_indexes, as_index); + pool_put(lbm->ass, as); + } + }); +} + +void lb_garbage_collection() +{ + lb_main_t *lbm = &lb_main; + lb_get_writer_lock(); + lb_vip_t *vip; + u32 *to_be_removed_vips = 0, *i; + pool_foreach(vip, lbm->vips, { + lb_vip_garbage_collection(vip); + + if (!(vip->flags & LB_VIP_FLAGS_USED) && + (pool_elts(vip->as_indexes) == 0)) { + vec_add1(to_be_removed_vips, vip - lbm->vips); + } + }); + + vec_foreach(i, to_be_removed_vips) { + vip = &lbm->vips[*i]; + pool_put(lbm->vips, vip); + pool_free(vip->as_indexes); + } + + vec_free(to_be_removed_vips); + lb_put_writer_lock(); +} + +static void lb_vip_update_new_flow_table(lb_vip_t *vip) +{ + lb_main_t *lbm = &lb_main; + lb_new_flow_entry_t *old_table; + u32 i, *as_index; + lb_new_flow_entry_t *new_flow_table = 0; + lb_as_t *as; + lb_pseudorand_t *pr, *sort_arr = 0; + u32 count; + + ASSERT (lbm->writer_lock[0]); //We must have the lock + + //Check if some AS is configured or not + i = 0; + pool_foreach(as_index, vip->as_indexes, { + as = &lbm->ass[*as_index]; + if (as->flags & LB_AS_FLAGS_USED) { //Not used anymore + i = 1; + goto out; //Not sure 'break' works in this macro-loop + } + }); + +out: + if (i == 0) { + //Only the default. i.e. no AS + vec_validate(new_flow_table, vip->new_flow_table_mask); + for (i=0; ias_indexes)); + + i = 0; + pool_foreach(as_index, vip->as_indexes, { + as = &lbm->ass[*as_index]; + if (!(as->flags & LB_AS_FLAGS_USED)) //Not used anymore + continue; + + sort_arr[i].as_index = as - lbm->ass; + i++; + }); + _vec_len(sort_arr) = i; + + vec_sort_with_function(sort_arr, lb_pseudorand_compare); + + //Now let's pseudo-randomly generate permutations + vec_foreach(pr, sort_arr) { + lb_as_t *as = &lbm->ass[pr->as_index]; + + u64 seed = clib_xxhash(as->address.as_u64[0] ^ + as->address.as_u64[1]); + /* We have 2^n buckets. + * skip must be prime with 2^n. + * So skip must be odd. + * MagLev actually state that M should be prime, + * but this has a big computation cost (% operation). + * Using 2^n is more better (& operation). + */ + pr->skip = ((seed & 0xffffffff) | 1) & vip->new_flow_table_mask; + pr->last = (seed >> 32) & vip->new_flow_table_mask; + } + + //Let's create a new flow table + vec_validate(new_flow_table, vip->new_flow_table_mask); + for (i=0; ilast; + pr->last = (pr->last + pr->skip) & vip->new_flow_table_mask; + if (new_flow_table[last].as_index == ~0) { + new_flow_table[last].as_index = pr->as_index; + break; + } + } + done++; + if (done == vec_len(new_flow_table)) + goto finished; + } + } + + vec_free(sort_arr); + +finished: + +//Count number of changed entries + count = 0; + for (i=0; inew_flow_table == 0 || + new_flow_table[i].as_index != vip->new_flow_table[i].as_index) + count++; + + old_table = vip->new_flow_table; + vip->new_flow_table = new_flow_table; + vec_free(old_table); +} + +int lb_conf(ip4_address_t *ip4_address, ip6_address_t *ip6_address, + u32 per_cpu_sticky_buckets, u32 flow_timeout) +{ + lb_main_t *lbm = &lb_main; + + if (!is_pow2(per_cpu_sticky_buckets)) + return VNET_API_ERROR_INVALID_MEMORY_SIZE; + + lb_get_writer_lock(); //Not exactly necessary but just a reminder that it exists for my future self + lbm->ip4_src_address = *ip4_address; + lbm->ip6_src_address = *ip6_address; + lbm->per_cpu_sticky_buckets = per_cpu_sticky_buckets; + lbm->flow_timeout = flow_timeout; + lb_put_writer_lock(); + return 0; +} + +static +int lb_vip_find_index_with_lock(ip46_address_t *prefix, u8 plen, u32 *vip_index) +{ + lb_main_t *lbm = &lb_main; + lb_vip_t *vip; + ASSERT (lbm->writer_lock[0]); //This must be called with the lock owned + ip46_prefix_normalize(prefix, plen); + pool_foreach(vip, lbm->vips, { + if ((vip->flags & LB_AS_FLAGS_USED) && + vip->plen == plen && + vip->prefix.as_u64[0] == prefix->as_u64[0] && + vip->prefix.as_u64[1] == prefix->as_u64[1]) { + *vip_index = vip - lbm->vips; + return 0; + } + }); + return VNET_API_ERROR_NO_SUCH_ENTRY; +} + +int lb_vip_find_index(ip46_address_t *prefix, u8 plen, u32 *vip_index) +{ + int ret; + lb_get_writer_lock(); + ret = lb_vip_find_index_with_lock(prefix, plen, vip_index); + lb_put_writer_lock(); + return ret; +} + +static int lb_as_find_index_vip(lb_vip_t *vip, ip46_address_t *address, u32 *as_index) +{ + lb_main_t *lbm = &lb_main; + ASSERT (lbm->writer_lock[0]); //This must be called with the lock owned + lb_as_t *as; + u32 *asi; + pool_foreach(asi, vip->as_indexes, { + as = &lbm->ass[*asi]; + if (as->vip_index == (vip - lbm->vips) && + as->address.as_u64[0] == address->as_u64[0] && + as->address.as_u64[1] == address->as_u64[1]) { + *as_index = as - lbm->ass; + return 0; + } + }); + return -1; +} + +int lb_vip_add_ass(u32 vip_index, ip46_address_t *addresses, u32 n) +{ + lb_main_t *lbm = &lb_main; + lb_get_writer_lock(); + lb_vip_t *vip; + if (!(vip = lb_vip_get_by_index(vip_index))) { + lb_put_writer_lock(); + return VNET_API_ERROR_NO_SUCH_ENTRY; + } + + ip46_type_t type = lb_vip_is_gre4(vip)?IP46_TYPE_IP4:IP46_TYPE_IP6; + u32 *to_be_added = 0; + u32 *to_be_updated = 0; + u32 i; + u32 *ip; + + //Sanity check + while (n--) { + + if (!lb_as_find_index_vip(vip, &addresses[n], &i)) { + if (lbm->ass[i].flags & LB_AS_FLAGS_USED) { + vec_free(to_be_added); + vec_free(to_be_updated); + lb_put_writer_lock(); + return VNET_API_ERROR_VALUE_EXIST; + } + vec_add1(to_be_updated, i); + goto next; + } + + if (ip46_address_type(&addresses[n]) != type) { + vec_free(to_be_added); + vec_free(to_be_updated); + lb_put_writer_lock(); + return VNET_API_ERROR_INVALID_ADDRESS_FAMILY; + } + + if (n) { + u32 n2 = n; + while(n2--) //Check for duplicates + if (addresses[n2].as_u64[0] == addresses[n].as_u64[0] && + addresses[n2].as_u64[1] == addresses[n].as_u64[1]) + goto next; + } + + vec_add1(to_be_added, n); + +next: + continue; + } + + //Update reused ASs + vec_foreach(ip, to_be_updated) { + lbm->ass[*ip].flags = LB_AS_FLAGS_USED; + } + vec_free(to_be_updated); + + //Create those who have to be created + vec_foreach(ip, to_be_added) { + lb_as_t *as; + u32 *as_index; + pool_get(lbm->ass, as); + as->address = addresses[*ip]; + as->flags = LB_AS_FLAGS_USED; + as->vip_index = vip_index; + pool_get(vip->as_indexes, as_index); + *as_index = as - lbm->ass; + + /* + * become a child of the FIB entry + * so we are informed when its forwarding changes + */ + fib_prefix_t nh = {}; + if (lb_vip_is_gre4(vip)) { + nh.fp_addr.ip4 = as->address.ip4; + nh.fp_len = 32; + nh.fp_proto = FIB_PROTOCOL_IP4; + } else { + nh.fp_addr.ip6 = as->address.ip6; + nh.fp_len = 128; + nh.fp_proto = FIB_PROTOCOL_IP6; + } + + as->next_hop_fib_entry_index = + fib_table_entry_special_add(0, + &nh, + FIB_SOURCE_RR, + FIB_ENTRY_FLAG_NONE, + ADJ_INDEX_INVALID); + as->next_hop_child_index = + fib_entry_child_add(as->next_hop_fib_entry_index, + lbm->fib_node_type, + as - lbm->ass); + + lb_as_stack(as); + } + vec_free(to_be_added); + + //Recompute flows + lb_vip_update_new_flow_table(vip); + + //Garbage collection maybe + lb_vip_garbage_collection(vip); + + lb_put_writer_lock(); + return 0; +} + +int lb_vip_del_ass_withlock(u32 vip_index, ip46_address_t *addresses, u32 n) +{ + lb_main_t *lbm = &lb_main; + u32 now = (u32) vlib_time_now(vlib_get_main()); + u32 *ip = 0; + + lb_vip_t *vip; + if (!(vip = lb_vip_get_by_index(vip_index))) { + return VNET_API_ERROR_NO_SUCH_ENTRY; + } + + u32 *indexes = NULL; + while (n--) { + u32 i; + if (lb_as_find_index_vip(vip, &addresses[n], &i)) { + vec_free(indexes); + return VNET_API_ERROR_NO_SUCH_ENTRY; + } + + if (n) { //Check for duplicates + u32 n2 = n - 1; + while(n2--) { + if (addresses[n2].as_u64[0] == addresses[n].as_u64[0] && + addresses[n2].as_u64[1] == addresses[n].as_u64[1]) + goto next; + } + } + + vec_add1(indexes, i); +next: + continue; + } + + //Garbage collection maybe + lb_vip_garbage_collection(vip); + + if (indexes != NULL) { + vec_foreach(ip, indexes) { + lbm->ass[*ip].flags &= ~LB_AS_FLAGS_USED; + lbm->ass[*ip].last_used = now; + } + + //Recompute flows + lb_vip_update_new_flow_table(vip); + } + + vec_free(indexes); + return 0; +} + +int lb_vip_del_ass(u32 vip_index, ip46_address_t *addresses, u32 n) +{ + lb_get_writer_lock(); + int ret = lb_vip_del_ass_withlock(vip_index, addresses, n); + lb_put_writer_lock(); + return ret; +} + +/** + * Add the VIP adjacency to the ip4 or ip6 fib + */ +static void lb_vip_add_adjacency(lb_main_t *lbm, lb_vip_t *vip) +{ + dpo_proto_t proto = 0; + dpo_id_t dpo = DPO_INVALID; + fib_prefix_t pfx = {}; + if (lb_vip_is_ip4(vip)) { + pfx.fp_addr.ip4 = vip->prefix.ip4; + pfx.fp_len = vip->plen - 96; + pfx.fp_proto = FIB_PROTOCOL_IP4; + proto = DPO_PROTO_IP4; + } else { + pfx.fp_addr.ip6 = vip->prefix.ip6; + pfx.fp_len = vip->plen; + pfx.fp_proto = FIB_PROTOCOL_IP6; + proto = DPO_PROTO_IP6; + } + dpo_set(&dpo, lb_vip_is_gre4(vip)?lbm->dpo_gre4_type:lbm->dpo_gre6_type, + proto, vip - lbm->vips); + fib_table_entry_special_dpo_add(0, + &pfx, + FIB_SOURCE_PLUGIN_HI, + FIB_ENTRY_FLAG_EXCLUSIVE, + &dpo); + dpo_reset(&dpo); +} + +/** + * Deletes the adjacency associated with the VIP + */ +static void lb_vip_del_adjacency(lb_main_t *lbm, lb_vip_t *vip) +{ + fib_prefix_t pfx = {}; + if (lb_vip_is_ip4(vip)) { + pfx.fp_addr.ip4 = vip->prefix.ip4; + pfx.fp_len = vip->plen - 96; + pfx.fp_proto = FIB_PROTOCOL_IP4; + } else { + pfx.fp_addr.ip6 = vip->prefix.ip6; + pfx.fp_len = vip->plen; + pfx.fp_proto = FIB_PROTOCOL_IP6; + } + fib_table_entry_special_remove(0, &pfx, FIB_SOURCE_PLUGIN_HI); +} + +int lb_vip_add(ip46_address_t *prefix, u8 plen, lb_vip_type_t type, u32 new_length, u32 *vip_index) +{ + lb_main_t *lbm = &lb_main; + lb_vip_t *vip; + lb_get_writer_lock(); + ip46_prefix_normalize(prefix, plen); + + if (!lb_vip_find_index_with_lock(prefix, plen, vip_index)) { + lb_put_writer_lock(); + return VNET_API_ERROR_VALUE_EXIST; + } + + if (!is_pow2(new_length)) { + lb_put_writer_lock(); + return VNET_API_ERROR_INVALID_MEMORY_SIZE; + } + + if (ip46_prefix_is_ip4(prefix, plen) && + (type != LB_VIP_TYPE_IP4_GRE4) && + (type != LB_VIP_TYPE_IP4_GRE6)) + return VNET_API_ERROR_INVALID_ADDRESS_FAMILY; + + + //Allocate + pool_get(lbm->vips, vip); + + //Init + vip->prefix = *prefix; + vip->plen = plen; + vip->last_garbage_collection = (u32) vlib_time_now(vlib_get_main()); + vip->type = type; + vip->flags = LB_VIP_FLAGS_USED; + vip->as_indexes = 0; + + //Validate counters + u32 i; + for (i = 0; i < LB_N_VIP_COUNTERS; i++) { + vlib_validate_simple_counter(&lbm->vip_counters[i], vip - lbm->vips); + vlib_zero_simple_counter(&lbm->vip_counters[i], vip - lbm->vips); + } + + //Configure new flow table + vip->new_flow_table_mask = new_length - 1; + vip->new_flow_table = 0; + + //Create a new flow hash table full of the default entry + lb_vip_update_new_flow_table(vip); + + //Create adjacency to direct traffic + lb_vip_add_adjacency(lbm, vip); + + //Return result + *vip_index = vip - lbm->vips; + + lb_put_writer_lock(); + return 0; +} + +int lb_vip_del(u32 vip_index) +{ + lb_main_t *lbm = &lb_main; + lb_vip_t *vip; + lb_get_writer_lock(); + if (!(vip = lb_vip_get_by_index(vip_index))) { + lb_put_writer_lock(); + return VNET_API_ERROR_NO_SUCH_ENTRY; + } + + //FIXME: This operation is actually not working + //We will need to remove state before performing this. + + { + //Remove all ASs + ip46_address_t *ass = 0; + lb_as_t *as; + u32 *as_index; + pool_foreach(as_index, vip->as_indexes, { + as = &lbm->ass[*as_index]; + vec_add1(ass, as->address); + }); + if (vec_len(ass)) + lb_vip_del_ass_withlock(vip_index, ass, vec_len(ass)); + vec_free(ass); + } + + //Delete adjacency + lb_vip_del_adjacency(lbm, vip); + + //Set the VIP as unused + vip->flags &= ~LB_VIP_FLAGS_USED; + + lb_put_writer_lock(); + return 0; +} + +clib_error_t * +vlib_plugin_register (vlib_main_t * vm, + vnet_plugin_handoff_t * h, + int from_early_init) +{ + clib_error_t *error = 0; + return error; +} + + +u8 *format_lb_dpo (u8 * s, va_list * va) +{ + index_t index = va_arg (*va, index_t); + CLIB_UNUSED(u32 indent) = va_arg (*va, u32); + lb_main_t *lbm = &lb_main; + lb_vip_t *vip = pool_elt_at_index (lbm->vips, index); + return format (s, "%U", format_lb_vip, vip); +} + +static void lb_dpo_lock (dpo_id_t *dpo) {} +static void lb_dpo_unlock (dpo_id_t *dpo) {} + +static fib_node_t * +lb_fib_node_get_node (fib_node_index_t index) +{ + lb_main_t *lbm = &lb_main; + lb_as_t *as = pool_elt_at_index (lbm->ass, index); + return (&as->fib_node); +} + +static void +lb_fib_node_last_lock_gone (fib_node_t *node) +{ +} + +static lb_as_t * +lb_as_from_fib_node (fib_node_t *node) +{ + return ((lb_as_t*)(((char*)node) - + STRUCT_OFFSET_OF(lb_as_t, fib_node))); +} + +static void +lb_as_stack (lb_as_t *as) +{ + lb_main_t *lbm = &lb_main; + lb_vip_t *vip = &lbm->vips[as->vip_index]; + dpo_stack(lb_vip_is_gre4(vip)?lbm->dpo_gre4_type:lbm->dpo_gre6_type, + lb_vip_is_ip4(vip)?DPO_PROTO_IP4:DPO_PROTO_IP6, + &as->dpo, + fib_entry_contribute_ip_forwarding( + as->next_hop_fib_entry_index)); +} + +static fib_node_back_walk_rc_t +lb_fib_node_back_walk_notify (fib_node_t *node, + fib_node_back_walk_ctx_t *ctx) +{ + lb_as_stack(lb_as_from_fib_node(node)); + return (FIB_NODE_BACK_WALK_CONTINUE); +} + +clib_error_t * +lb_init (vlib_main_t * vm) +{ + vlib_thread_main_t *tm = vlib_get_thread_main (); + lb_main_t *lbm = &lb_main; + lb_as_t *default_as; + fib_node_vft_t lb_fib_node_vft = { + .fnv_get = lb_fib_node_get_node, + .fnv_last_lock = lb_fib_node_last_lock_gone, + .fnv_back_walk = lb_fib_node_back_walk_notify, + }; + dpo_vft_t lb_vft = { + .dv_lock = lb_dpo_lock, + .dv_unlock = lb_dpo_unlock, + .dv_format = format_lb_dpo, + }; + + lbm->vips = 0; + lbm->per_cpu = 0; + vec_validate(lbm->per_cpu, tm->n_vlib_mains - 1); + lbm->writer_lock = clib_mem_alloc_aligned (CLIB_CACHE_LINE_BYTES, CLIB_CACHE_LINE_BYTES); + lbm->writer_lock[0] = 0; + lbm->per_cpu_sticky_buckets = LB_DEFAULT_PER_CPU_STICKY_BUCKETS; + lbm->flow_timeout = LB_DEFAULT_FLOW_TIMEOUT; + lbm->ip4_src_address.as_u32 = 0xffffffff; + lbm->ip6_src_address.as_u64[0] = 0xffffffffffffffffL; + lbm->ip6_src_address.as_u64[1] = 0xffffffffffffffffL; + lbm->dpo_gre4_type = dpo_register_new_type(&lb_vft, lb_dpo_gre4_nodes); + lbm->dpo_gre6_type = dpo_register_new_type(&lb_vft, lb_dpo_gre6_nodes); + lbm->fib_node_type = fib_node_register_new_type(&lb_fib_node_vft); + + //Init AS reference counters + vlib_refcount_init(&lbm->as_refcount); + + //Allocate and init default AS. + lbm->ass = 0; + pool_get(lbm->ass, default_as); + default_as->flags = 0; + default_as->dpo.dpoi_next_node = LB_NEXT_DROP; + default_as->vip_index = ~0; + default_as->address.ip6.as_u64[0] = 0xffffffffffffffffL; + default_as->address.ip6.as_u64[1] = 0xffffffffffffffffL; + +#define _(a,b,c) lbm->vip_counters[c].name = b; + lb_foreach_vip_counter +#undef _ + return NULL; +} + +VLIB_INIT_FUNCTION (lb_init); diff --git a/src/plugins/lb/lb.h b/src/plugins/lb/lb.h new file mode 100644 index 00000000..882b9b30 --- /dev/null +++ b/src/plugins/lb/lb.h @@ -0,0 +1,333 @@ +/* + * Copyright (c) 2016 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +/** + * lb-plugin implements a MagLev-like load balancer. + * http://research.google.com/pubs/pub44824.html + * + * It hasn't been tested for interoperability with the original MagLev + * but intends to provide similar functionality. + * The load-balancer receives traffic destined to VIP (Virtual IP) + * addresses from one or multiple(ECMP) routers. + * The load-balancer tunnels the traffic toward many application servers + * ensuring session stickyness (i.e. that a single sessions is tunneled + * towards a single application server). + * + */ + +#ifndef LB_PLUGIN_LB_LB_H_ +#define LB_PLUGIN_LB_LB_H_ + +#include +#include + +#include +#include +#include +#include + +#include + +#define LB_DEFAULT_PER_CPU_STICKY_BUCKETS 1 << 10 +#define LB_DEFAULT_FLOW_TIMEOUT 40 + +typedef enum { + LB_NEXT_DROP, + LB_N_NEXT, +} lb_next_t; + +/** + * Each VIP is configured with a set of + * application server. + */ +typedef struct { + /** + * Registration to FIB event. + */ + fib_node_t fib_node; + + /** + * Destination address used to tunnel traffic towards + * that application server. + * The address is also used as ID and pseudo-random + * seed for the load-balancing process. + */ + ip46_address_t address; + + /** + * ASs are indexed by address and VIP Index. + * Which means there will be duplicated if the same server + * address is used for multiple VIPs. + */ + u32 vip_index; + + /** + * Some per-AS flags. + * For now only LB_AS_FLAGS_USED is defined. + */ + u8 flags; + +#define LB_AS_FLAGS_USED 0x1 + + /** + * Rotating timestamp of when LB_AS_FLAGS_USED flag was last set. + * + * AS removal is based on garbage collection and reference counting. + * When an AS is removed, there is a race between configuration core + * and worker cores which may still add a reference while it should not + * be used. This timestamp is used to not remove the AS while a race condition + * may happen. + */ + u32 last_used; + + /** + * The FIB entry index for the next-hop + */ + fib_node_index_t next_hop_fib_entry_index; + + /** + * The child index on the FIB entry + */ + u32 next_hop_child_index; + + /** + * The next DPO in the graph to follow. + */ + dpo_id_t dpo; + +} lb_as_t; + +format_function_t format_lb_as; + +typedef struct { + u32 as_index; +} lb_new_flow_entry_t; + +#define lb_foreach_vip_counter \ + _(NEXT_PACKET, "packet from existing sessions", 0) \ + _(FIRST_PACKET, "first session packet", 1) \ + _(UNTRACKED_PACKET, "untracked packet", 2) \ + _(NO_SERVER, "no server configured", 3) + +typedef enum { +#define _(a,b,c) LB_VIP_COUNTER_##a = c, + lb_foreach_vip_counter +#undef _ + LB_N_VIP_COUNTERS +} lb_vip_counter_t; + +/** + * The load balancer supports IPv4 and IPv6 traffic + * and GRE4 and GRE6 encap. + */ +typedef enum { + LB_VIP_TYPE_IP6_GRE6, + LB_VIP_TYPE_IP6_GRE4, + LB_VIP_TYPE_IP4_GRE6, + LB_VIP_TYPE_IP4_GRE4, + LB_VIP_N_TYPES, +} lb_vip_type_t; + +format_function_t format_lb_vip_type; +unformat_function_t unformat_lb_vip_type; + +/** + * Load balancing service is provided per VIP. + * In this data model, a VIP can be a whole prefix. + * But load balancing only + * occurs on a per-source-address/port basis. Meaning that if a given source + * reuses the same port for multiple destinations within the same VIP, + * they will be considered as a single flow. + */ +typedef struct { + + //Runtime + + /** + * Vector mapping (flow-hash & new_connect_table_mask) to AS index. + * This is used for new flows. + */ + lb_new_flow_entry_t *new_flow_table; + + /** + * New flows table length - 1 + * (length MUST be a power of 2) + */ + u32 new_flow_table_mask; + + /** + * Last time garbage collection was run to free the ASs. + */ + u32 last_garbage_collection; + + //Not runtime + + /** + * A Virtual IP represents a given service delivered + * by a set of application servers. It can be a single + * address or a prefix. + * IPv4 prefixes are encoded using IPv4-in-IPv6 embedded address + * (i.e. ::/96 prefix). + */ + ip46_address_t prefix; + + /** + * The VIP prefix length. + * In case of IPv4, plen = 96 + ip4_plen. + */ + u8 plen; + + /** + * The type of traffic for this. + * LB_TYPE_UNDEFINED if unknown. + */ + lb_vip_type_t type; + + /** + * Flags related to this VIP. + * LB_VIP_FLAGS_USED means the VIP is active. + * When it is not set, the VIP in the process of being removed. + * We cannot immediately remove a VIP because the VIP index still may be stored + * in the adjacency index. + */ + u8 flags; +#define LB_VIP_FLAGS_USED 0x1 + + /** + * Pool of AS indexes used for this VIP. + * This also includes ASs that have been removed (but are still referenced). + */ + u32 *as_indexes; +} lb_vip_t; + +#define lb_vip_is_ip4(vip) ((vip)->type == LB_VIP_TYPE_IP4_GRE6 || (vip)->type == LB_VIP_TYPE_IP4_GRE4) +#define lb_vip_is_gre4(vip) ((vip)->type == LB_VIP_TYPE_IP6_GRE4 || (vip)->type == LB_VIP_TYPE_IP4_GRE4) +format_function_t format_lb_vip; +format_function_t format_lb_vip_detailed; + +typedef struct { + /** + * Each CPU has its own sticky flow hash table. + * One single table is used for all VIPs. + */ + lb_hash_t *sticky_ht; +} lb_per_cpu_t; + +typedef struct { + /** + * Pool of all Virtual IPs + */ + lb_vip_t *vips; + + /** + * Pool of ASs. + * ASs are referenced by address and vip index. + * The first element (index 0) is special and used only to fill + * new_flow_tables when no AS has been configured. + */ + lb_as_t *ass; + + /** + * Each AS has an associated reference counter. + * As ass[0] has a special meaning, its associated counter + * starts at 0 and is decremented instead. i.e. do not use it. + */ + vlib_refcount_t as_refcount; + + /** + * Some global data is per-cpu + */ + lb_per_cpu_t *per_cpu; + + /** + * Node next index for IP adjacencies, for each of the traffic types. + */ + u32 ip_lookup_next_index[LB_VIP_N_TYPES]; + + /** + * Source address used in IPv6 encapsulated traffic + */ + ip6_address_t ip6_src_address; + + /** + * Source address used for IPv4 encapsulated traffic + */ + ip4_address_t ip4_src_address; + + /** + * Number of buckets in the per-cpu sticky hash table. + */ + u32 per_cpu_sticky_buckets; + + /** + * Flow timeout in seconds. + */ + u32 flow_timeout; + + /** + * Per VIP counter + */ + vlib_simple_counter_main_t vip_counters[LB_N_VIP_COUNTERS]; + + /** + * DPO used to send packet from IP4/6 lookup to LB node. + */ + dpo_type_t dpo_gre4_type; + dpo_type_t dpo_gre6_type; + + /** + * Node type for registering to fib changes. + */ + fib_node_type_t fib_node_type; + + /** + * API dynamically registered base ID. + */ + u16 msg_id_base; + + volatile u32 *writer_lock; +} lb_main_t; + +extern lb_main_t lb_main; +extern vlib_node_registration_t lb6_node; +extern vlib_node_registration_t lb4_node; + +/** + * Fix global load-balancer parameters. + * @param ip4_address IPv4 source address used for encapsulated traffic + * @param ip6_address IPv6 source address used for encapsulated traffic + * @return 0 on success. VNET_LB_ERR_XXX on error + */ +int lb_conf(ip4_address_t *ip4_address, ip6_address_t *ip6_address, + u32 sticky_buckets, u32 flow_timeout); + +int lb_vip_add(ip46_address_t *prefix, u8 plen, lb_vip_type_t type, + u32 new_length, u32 *vip_index); +int lb_vip_del(u32 vip_index); + +int lb_vip_find_index(ip46_address_t *prefix, u8 plen, u32 *vip_index); + +#define lb_vip_get_by_index(index) (pool_is_free_index(lb_main.vips, index)?NULL:pool_elt_at_index(lb_main.vips, index)) + +int lb_vip_add_ass(u32 vip_index, ip46_address_t *addresses, u32 n); +int lb_vip_del_ass(u32 vip_index, ip46_address_t *addresses, u32 n); + +u32 lb_hash_time_now(vlib_main_t * vm); + +void lb_garbage_collection(); + +format_function_t format_lb_main; + +#endif /* LB_PLUGIN_LB_LB_H_ */ diff --git a/src/plugins/lb/lb_plugin_doc.md b/src/plugins/lb/lb_plugin_doc.md new file mode 100644 index 00000000..c7885ffb --- /dev/null +++ b/src/plugins/lb/lb_plugin_doc.md @@ -0,0 +1,141 @@ +# Load Balancer plugin for VPP {#lb_plugin_doc} + +## Version + +The load balancer plugin is currently in *beta* version. +Both CLIs and APIs are subject to *heavy* changes. +Wich also means feedback is really welcome regarding features, apis, etc... + +## Overview + +This plugin provides load balancing for VPP in a way that is largely inspired +from Google's MagLev: http://research.google.com/pubs/pub44824.html + +The load balancer is configured with a set of Virtual IPs (VIP, which can be +prefixes), and for each VIP, with a set of Application Server addresses (ASs). + +Traffic received for a given VIP (or VIP prefix) is tunneled using GRE towards +the different ASs in a way that (tries to) ensure that a given session will +always be tunneled to the same AS. + +Both VIPs or ASs can be IPv4 or IPv6, but for a given VIP, all ASs must be using +the same encap. type (i.e. IPv4+GRE or IPv6+GRE). Meaning that for a given VIP, +all AS addresses must be of the same family. + +## Performances + +The load balancer has been tested up to 1 millions flows and still forwards more +than 3Mpps per core in such circumstances. +Although 3Mpps seems already good, it is likely that performances will be improved +in next versions. + +## Configuration + +### Global LB parameters + +The load balancer needs to be configured with some parameters: + + lb conf [ip4-src-address ] [ip6-src-address ] + [buckets ] [timeout ] + +ip4-src-address: the source address used to send encap. packets using IPv4. + +ip6-src-address: the source address used to send encap. packets using IPv6. + +buckets: the *per-thread* established-connexions-table number of buckets. + +timeout: the number of seconds a connection will remain in the + established-connexions-table while no packet for this flow + is received. + + +### Configure the VIPs + + lb vip [encap (gre6|gre4)] [new_len ] [del] + +new_len is the size of the new-connection-table. It should be 1 or 2 orders of +magnitude bigger than the number of ASs for the VIP in order to ensure a good +load balancing. + +Examples: + + lb vip 2002::/16 encap gre6 new_len 1024 + lb vip 2003::/16 encap gre4 new_len 2048 + lb vip 80.0.0.0/8 encap gre6 new_len 16 + lb vip 90.0.0.0/8 encap gre4 new_len 1024 + +### Configure the ASs (for each VIP) + + lb as [
[
[...]]] [del] + +You can add (or delete) as many ASs at a time (for a single VIP). +Note that the AS address family must correspond to the VIP encap. IP family. + +Examples: + + lb as 2002::/16 2001::2 2001::3 2001::4 + lb as 2003::/16 10.0.0.1 10.0.0.2 + lb as 80.0.0.0/8 2001::2 + lb as 90.0.0.0/8 10.0.0.1 + + + +## Monitoring + +The plugin provides quite a bunch of counters and information. +These are still subject to quite significant changes. + + show lb + show lb vip + show lb vip verbose + + show node counters + + +## Design notes + +### Multi-Threading + +MagLev is a distributed system which pseudo-randomly generates a +new-connections-table based on AS names such that each server configured with +the same set of ASs ends up with the same table. Connection stickyness is then +ensured with an established-connections-table. Using ECMP, it is assumed (but +not relied on) that servers will mostly receive traffic for different flows. + +This implementation pushes the parallelism a little bit further by using +one established-connections table per thread. This is equivalent to assuming +that RSS will make a job similar to ECMP, and is pretty useful as threads don't +need to get a lock in order to write in the table. + +### Hash Table + +A load balancer requires an efficient read and write hash table. The hash table +used by ip6-forward is very read-efficient, but not so much for writing. In +addition, it is not a big deal if writing into the hash table fails (again, +MagLev uses a flow table but does not heaviliy relies on it). + +The plugin therefore uses a very specific (and stupid) hash table. + - Fixed (and power of 2) number of buckets (configured at runtime) + - Fixed (and power of 2) elements per buckets (configured at compilation time) + +### Reference counting + +When an AS is removed, there is two possible ways to react. + - Keep using the AS for established connections + - Change AS for established connections (likely to cause error for TCP) + +In the first case, although an AS is removed from the configuration, its +associated state needs to stay around as long as it is used by at least one +thread. + +In order to avoid locks, a specific reference counter is used. The design is quite +similar to clib counters but: + - It is possible to decrease the value + - Summing will not zero the per-thread counters + - Only the thread can reallocate its own counters vector (to avoid concurrency issues) + +This reference counter is lock free, but reading a count of 0 does not mean +the value can be freed unless it is ensured by *other* means that no other thread +is concurrently referencing the object. In the case of this plugin, it is assumed +that no concurrent event will take place after a few seconds. + diff --git a/src/plugins/lb/lb_test.c b/src/plugins/lb/lb_test.c new file mode 100644 index 00000000..8c2eaa91 --- /dev/null +++ b/src/plugins/lb/lb_test.c @@ -0,0 +1,293 @@ +/* + * Copyright (c) 2016 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +#include +#include +#include +#include +#include +#include + +//TODO: Move that to vat/plugin_api.c +////////////////////////// +uword unformat_ip46_address (unformat_input_t * input, va_list * args) +{ + ip46_address_t *ip46 = va_arg (*args, ip46_address_t *); + ip46_type_t type = va_arg (*args, ip46_type_t); + if ((type != IP46_TYPE_IP6) && + unformat(input, "%U", unformat_ip4_address, &ip46->ip4)) { + ip46_address_mask_ip4(ip46); + return 1; + } else if ((type != IP46_TYPE_IP4) && + unformat(input, "%U", unformat_ip6_address, &ip46->ip6)) { + return 1; + } + return 0; +} +uword unformat_ip46_prefix (unformat_input_t * input, va_list * args) +{ + ip46_address_t *ip46 = va_arg (*args, ip46_address_t *); + u8 *len = va_arg (*args, u8 *); + ip46_type_t type = va_arg (*args, ip46_type_t); + + u32 l; + if ((type != IP46_TYPE_IP6) && unformat(input, "%U/%u", unformat_ip4_address, &ip46->ip4, &l)) { + if (l > 32) + return 0; + *len = l + 96; + ip46->pad[0] = ip46->pad[1] = ip46->pad[2] = 0; + } else if ((type != IP46_TYPE_IP4) && unformat(input, "%U/%u", unformat_ip6_address, &ip46->ip6, &l)) { + if (l > 128) + return 0; + *len = l; + } else { + return 0; + } + return 1; +} +///////////////////////// + +#define vl_msg_id(n,h) n, +typedef enum { +#include + /* We'll want to know how many messages IDs we need... */ + VL_MSG_FIRST_AVAILABLE, +} vl_msg_id_t; +#undef vl_msg_id + +/* define message structures */ +#define vl_typedefs +#include +#undef vl_typedefs + +/* declare message handlers for each api */ + +#define vl_endianfun /* define message structures */ +#include +#undef vl_endianfun + +/* instantiate all the print functions we know about */ +#define vl_print(handle, ...) +#define vl_printfun +#include +#undef vl_printfun + +/* Get the API version number. */ +#define vl_api_version(n,v) static u32 api_version=(v); +#include +#undef vl_api_version + +typedef struct { + /* API message ID base */ + u16 msg_id_base; + vat_main_t *vat_main; +} lb_test_main_t; + +lb_test_main_t lb_test_main; + +#define foreach_standard_reply_retval_handler \ +_(lb_conf_reply) \ +_(lb_add_del_vip_reply) \ +_(lb_add_del_as_reply) + +#define _(n) \ + static void vl_api_##n##_t_handler \ + (vl_api_##n##_t * mp) \ + { \ + vat_main_t * vam = lb_test_main.vat_main; \ + i32 retval = ntohl(mp->retval); \ + if (vam->async_mode) { \ + vam->async_errors += (retval < 0); \ + } else { \ + vam->retval = retval; \ + vam->result_ready = 1; \ + } \ + } +foreach_standard_reply_retval_handler; +#undef _ + +/* + * Table of message reply handlers, must include boilerplate handlers + * we just generated + */ +#define foreach_vpe_api_reply_msg \ + _(LB_CONF_REPLY, lb_conf_reply) \ + _(LB_ADD_DEL_VIP_REPLY, lb_add_del_vip_reply) \ + _(LB_ADD_DEL_AS_REPLY, lb_add_del_as_reply) + +/* M: construct, but don't yet send a message */ +#define M(T,t) \ +do { \ + vam->result_ready = 0; \ + mp = vl_msg_api_alloc(sizeof(*mp)); \ + memcpy (mp, &mps, sizeof (*mp)); \ + mp->_vl_msg_id = ntohs (VL_API_##T + lbtm->msg_id_base); \ + mp->client_index = vam->my_client_index; \ +} while(0); + +/* S: send a message */ +#define S (vl_msg_api_send_shmem (vam->vl_input_queue, (u8 *)&mp)) + +/* W: wait for results, with timeout */ +#define W \ +do { \ + timeout = vat_time_now (vam) + 1.0; \ + \ + while (vat_time_now (vam) < timeout) { \ + if (vam->result_ready == 1) { \ + return (vam->retval); \ + } \ + } \ + return -99; \ +} while(0); + +static int api_lb_conf (vat_main_t * vam) +{ + lb_test_main_t *lbtm = &lb_test_main; + unformat_input_t *i = vam->input; + f64 timeout; + vl_api_lb_conf_t mps, *mp; + + if (!unformat(i, "%U %U %u %u", + unformat_ip4_address, &mps.ip4_src_address, + unformat_ip6_address, mps.ip6_src_address, + &mps.sticky_buckets_per_core, + &mps.flow_timeout)) { + errmsg ("invalid arguments\n"); + return -99; + } + + M(LB_CONF, lb_conf); S; W; + + /* NOTREACHED */ + return 0; +} + +static int api_lb_add_del_vip (vat_main_t * vam) +{ + lb_test_main_t *lbtm = &lb_test_main; + unformat_input_t * i = vam->input; + f64 timeout; + vl_api_lb_add_del_vip_t mps, *mp; + mps.is_del = 0; + mps.is_gre4 = 0; + + if (!unformat(i, "%U", + unformat_ip46_prefix, mps.ip_prefix, &mps.prefix_length, IP46_TYPE_ANY)) { + errmsg ("invalid prefix\n"); + return -99; + } + + if (unformat(i, "gre4")) { + mps.is_gre4 = 1; + } else if (unformat(i, "gre6")) { + mps.is_gre4 = 0; + } else { + errmsg ("no encap\n"); + return -99; + } + + if (!unformat(i, "%d", &mps.new_flows_table_length)) { + errmsg ("no table lentgh\n"); + return -99; + } + + if (unformat(i, "del")) { + mps.is_del = 1; + } + + M(LB_ADD_DEL_VIP, lb_add_del_vip); S; W; + /* NOTREACHED */ + return 0; +} + +static int api_lb_add_del_as (vat_main_t * vam) +{ + lb_test_main_t *lbtm = &lb_test_main; + unformat_input_t * i = vam->input; + f64 timeout; + vl_api_lb_add_del_as_t mps, *mp; + mps.is_del = 0; + + if (!unformat(i, "%U %U", + unformat_ip46_prefix, mps.vip_ip_prefix, &mps.vip_prefix_length, IP46_TYPE_ANY, + unformat_ip46_address, mps.as_address)) { + errmsg ("invalid prefix or address\n"); + return -99; + } + + if (unformat(i, "del")) { + mps.is_del = 1; + } + + M(LB_ADD_DEL_AS, lb_add_del_as); S; W; + /* NOTREACHED */ + return 0; +} + +/* + * List of messages that the api test plugin sends, + * and that the data plane plugin processes + */ +#define foreach_vpe_api_msg \ +_(lb_conf, " ") \ +_(lb_add_del_vip, " [gre4|gre6] [del]") \ +_(lb_add_del_as, "
[del]") + +void vat_api_hookup (vat_main_t *vam) +{ + lb_test_main_t * lbtm = &lb_test_main; + /* Hook up handlers for replies from the data plane plug-in */ +#define _(N,n) \ + vl_msg_api_set_handlers((VL_API_##N + lbtm->msg_id_base), \ + #n, \ + vl_api_##n##_t_handler, \ + vl_noop_handler, \ + vl_api_##n##_t_endian, \ + vl_api_##n##_t_print, \ + sizeof(vl_api_##n##_t), 1); + foreach_vpe_api_reply_msg; +#undef _ + + /* API messages we can send */ +#define _(n,h) hash_set_mem (vam->function_by_name, #n, api_##n); + foreach_vpe_api_msg; +#undef _ + + /* Help strings */ +#define _(n,h) hash_set_mem (vam->help_by_name, #n, h); + foreach_vpe_api_msg; +#undef _ +} + +clib_error_t * vat_plugin_register (vat_main_t *vam) +{ + lb_test_main_t * lbtm = &lb_test_main; + + u8 * name; + + lbtm->vat_main = vam; + + /* Ask the vpp engine for the first assigned message-id */ + name = format (0, "lb_%08x%c", api_version, 0); + lbtm->msg_id_base = vl_client_get_first_plugin_msg_id ((char *) name); + + if (lbtm->msg_id_base != (u16) ~0) + vat_api_hookup (vam); + + vec_free(name); + + return 0; +} diff --git a/src/plugins/lb/lbhash.h b/src/plugins/lb/lbhash.h new file mode 100644 index 00000000..ca3cc143 --- /dev/null +++ b/src/plugins/lb/lbhash.h @@ -0,0 +1,216 @@ +/* + * Copyright (c) 2012 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +/** + * vppinfra already includes tons of different hash tables. + * MagLev flow table is a bit different. It has to be very efficient + * for both writing and reading operations. But it does not need to + * be 100% reliable (write can fail). It also needs to recycle + * old entries in a lazy way. + * + * This hash table is the most dummy hash table you can do. + * Fixed total size, fixed bucket size. + * Advantage is that it could be very efficient (maybe). + * + */ + +#ifndef LB_PLUGIN_LB_LBHASH_H_ +#define LB_PLUGIN_LB_LBHASH_H_ + +#include + +#if defined (__SSE4_2__) +#include +#endif + +/* + * @brief Number of entries per bucket. + */ +#define LBHASH_ENTRY_PER_BUCKET 4 + +#define LB_HASH_DO_NOT_USE_SSE_BUCKETS 0 + +/* + * @brief One bucket contains 4 entries. + * Each bucket takes one 64B cache line in memory. + */ +typedef struct { + CLIB_CACHE_LINE_ALIGN_MARK (cacheline0); + u32 hash[LBHASH_ENTRY_PER_BUCKET]; + u32 timeout[LBHASH_ENTRY_PER_BUCKET]; + u32 vip[LBHASH_ENTRY_PER_BUCKET]; + u32 value[LBHASH_ENTRY_PER_BUCKET]; +} lb_hash_bucket_t; + +typedef struct { + u32 buckets_mask; + u32 timeout; + lb_hash_bucket_t buckets[]; +} lb_hash_t; + +#define lb_hash_nbuckets(h) (((h)->buckets_mask) + 1) +#define lb_hash_size(h) ((h)->buckets_mask + LBHASH_ENTRY_PER_BUCKET) + +#define lb_hash_foreach_bucket(h, bucket) \ + for (bucket = (h)->buckets; \ + bucket < (h)->buckets + lb_hash_nbuckets(h); \ + bucket++) + +#define lb_hash_foreach_entry(h, bucket, i) \ + lb_hash_foreach_bucket(h, bucket) \ + for (i = 0; i < LBHASH_ENTRY_PER_BUCKET; i++) + +#define lb_hash_foreach_valid_entry(h, bucket, i, now) \ + lb_hash_foreach_entry(h, bucket, i) \ + if (!clib_u32_loop_gt((now), bucket->timeout[i])) + +static_always_inline +lb_hash_t *lb_hash_alloc(u32 buckets, u32 timeout) +{ + if (!is_pow2(buckets)) + return NULL; + + // Allocate 1 more bucket for prefetch + u32 size = ((u64)&((lb_hash_t *)(0))->buckets[0]) + + sizeof(lb_hash_bucket_t) * (buckets + 1); + u8 *mem = 0; + lb_hash_t *h; + vec_alloc_aligned(mem, size, CLIB_CACHE_LINE_BYTES); + h = (lb_hash_t *)mem; + h->buckets_mask = (buckets - 1); + h->timeout = timeout; + return h; +} + +static_always_inline +void lb_hash_free(lb_hash_t *h) +{ + u8 *mem = (u8 *)h; + vec_free(mem); +} + +#if __SSE4_2__ +static_always_inline +u32 lb_hash_hash(u64 k0, u64 k1, u64 k2, u64 k3, u64 k4) +{ + u64 val = 0; + val = _mm_crc32_u64(val, k0); + val = _mm_crc32_u64(val, k1); + val = _mm_crc32_u64(val, k2); + val = _mm_crc32_u64(val, k3); + val = _mm_crc32_u64(val, k4); + return (u32) val; +} +#else +static_always_inline +u32 lb_hash_hash(u64 k0, u64 k1, u64 k2, u64 k3, u64 k4) +{ + u64 tmp = k0 ^ k1 ^ k2 ^ k3 ^ k4; + return (u32)clib_xxhash (tmp); +} +#endif + +static_always_inline +void lb_hash_prefetch_bucket(lb_hash_t *ht, u32 hash) +{ + lb_hash_bucket_t *bucket = &ht->buckets[hash & ht->buckets_mask]; + CLIB_PREFETCH(bucket, sizeof(*bucket), READ); +} + +static_always_inline +void lb_hash_get(lb_hash_t *ht, u32 hash, u32 vip, u32 time_now, + u32 *available_index, u32 *found_value) +{ + lb_hash_bucket_t *bucket = &ht->buckets[hash & ht->buckets_mask]; + *found_value = ~0; + *available_index = ~0; +#if __SSE4_2__ && LB_HASH_DO_NOT_USE_SSE_BUCKETS == 0 + u32 bitmask, found_index; + __m128i mask; + + // mask[*] = timeout[*] > now + mask = _mm_cmpgt_epi32(_mm_loadu_si128 ((__m128i *) bucket->timeout), + _mm_set1_epi32 (time_now)); + // bitmask[*] = now <= timeout[*/4] + bitmask = (~_mm_movemask_epi8(mask)) & 0xffff; + // Get first index with now <= timeout[*], if any. + *available_index = (bitmask)?__builtin_ctz(bitmask)/4:*available_index; + + // mask[*] = (timeout[*] > now) && (hash[*] == hash) + mask = _mm_and_si128(mask, + _mm_cmpeq_epi32( + _mm_loadu_si128 ((__m128i *) bucket->hash), + _mm_set1_epi32 (hash))); + + // Load the array of vip values + // mask[*] = (timeout[*] > now) && (hash[*] == hash) && (vip[*] == vip) + mask = _mm_and_si128(mask, + _mm_cmpeq_epi32( + _mm_loadu_si128 ((__m128i *) bucket->vip), + _mm_set1_epi32 (vip))); + + // mask[*] = (timeout[*x4] > now) && (hash[*x4] == hash) && (vip[*x4] == vip) + bitmask = _mm_movemask_epi8(mask); + // Get first index, if any + found_index = (bitmask)?__builtin_ctzll(bitmask)/4:0; + ASSERT(found_index < 4); + *found_value = (bitmask)?bucket->value[found_index]:*found_value; + bucket->timeout[found_index] = + (bitmask)?time_now + ht->timeout:bucket->timeout[found_index]; +#else + u32 i; + for (i = 0; i < LBHASH_ENTRY_PER_BUCKET; i++) { + u8 cmp = (bucket->hash[i] == hash && bucket->vip[i] == vip); + u8 timeouted = clib_u32_loop_gt(time_now, bucket->timeout[i]); + *found_value = (cmp || timeouted)?*found_value:bucket->value[i]; + bucket->timeout[i] = (cmp || timeouted)?time_now + ht->timeout:bucket->timeout[i]; + *available_index = (timeouted && (*available_index == ~0))?i:*available_index; + + if (!cmp) + return; + } +#endif +} + +static_always_inline +u32 lb_hash_available_value(lb_hash_t *h, u32 hash, u32 available_index) +{ + return h->buckets[hash & h->buckets_mask].value[available_index]; +} + +static_always_inline +void lb_hash_put(lb_hash_t *h, u32 hash, u32 value, u32 vip, + u32 available_index, u32 time_now) +{ + lb_hash_bucket_t *bucket = &h->buckets[hash & h->buckets_mask]; + bucket->hash[available_index] = hash; + bucket->value[available_index] = value; + bucket->timeout[available_index] = time_now + h->timeout; + bucket->vip[available_index] = vip; +} + +static_always_inline +u32 lb_hash_elts(lb_hash_t *h, u32 time_now) +{ + u32 tot = 0; + lb_hash_bucket_t *bucket; + u32 i; + lb_hash_foreach_valid_entry(h, bucket, i, time_now) { + tot++; + } + return tot; +} + +#endif /* LB_PLUGIN_LB_LBHASH_H_ */ diff --git a/src/plugins/lb/node.c b/src/plugins/lb/node.c new file mode 100644 index 00000000..8b763c53 --- /dev/null +++ b/src/plugins/lb/node.c @@ -0,0 +1,419 @@ +/* + * Copyright (c) 2016 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +#include + +#include +#include + +#define foreach_lb_error \ + _(NONE, "no error") \ + _(PROTO_NOT_SUPPORTED, "protocol not supported") + +typedef enum { +#define _(sym,str) LB_ERROR_##sym, + foreach_lb_error +#undef _ + LB_N_ERROR, +} lb_error_t; + +static char *lb_error_strings[] = { +#define _(sym,string) string, + foreach_lb_error +#undef _ +}; + +typedef struct { + u32 vip_index; + u32 as_index; +} lb_trace_t; + +u8 * +format_lb_trace (u8 * s, va_list * args) +{ + lb_main_t *lbm = &lb_main; + CLIB_UNUSED (vlib_main_t * vm) = va_arg (*args, vlib_main_t *); + CLIB_UNUSED (vlib_node_t * node) = va_arg (*args, vlib_node_t *); + lb_trace_t *t = va_arg (*args, lb_trace_t *); + if (pool_is_free_index(lbm->vips, t->vip_index)) { + s = format(s, "lb vip[%d]: This VIP was freed since capture\n"); + } else { + s = format(s, "lb vip[%d]: %U\n", t->vip_index, format_lb_vip, &lbm->vips[t->vip_index]); + } + if (pool_is_free_index(lbm->ass, t->as_index)) { + s = format(s, "lb as[%d]: This AS was freed since capture\n"); + } else { + s = format(s, "lb as[%d]: %U\n", t->as_index, format_lb_as, &lbm->ass[t->as_index]); + } + return s; +} + +lb_hash_t *lb_get_sticky_table(u32 cpu_index) +{ + lb_main_t *lbm = &lb_main; + lb_hash_t *sticky_ht = lbm->per_cpu[cpu_index].sticky_ht; + //Check if size changed + if (PREDICT_FALSE(sticky_ht && (lbm->per_cpu_sticky_buckets != lb_hash_nbuckets(sticky_ht)))) + { + //Dereference everything in there + lb_hash_bucket_t *b; + u32 i; + lb_hash_foreach_entry(sticky_ht, b, i) { + vlib_refcount_add(&lbm->as_refcount, cpu_index, b->value[i], -1); + vlib_refcount_add(&lbm->as_refcount, cpu_index, 0, 1); + } + + lb_hash_free(sticky_ht); + sticky_ht = NULL; + } + + //Create if necessary + if (PREDICT_FALSE(sticky_ht == NULL)) { + lbm->per_cpu[cpu_index].sticky_ht = lb_hash_alloc(lbm->per_cpu_sticky_buckets, lbm->flow_timeout); + sticky_ht = lbm->per_cpu[cpu_index].sticky_ht; + clib_warning("Regenerated sticky table %p", sticky_ht); + } + + ASSERT(sticky_ht); + + //Update timeout + sticky_ht->timeout = lbm->flow_timeout; + return sticky_ht; +} + +u64 +lb_node_get_other_ports4(ip4_header_t *ip40) +{ + return 0; +} + +u64 +lb_node_get_other_ports6(ip6_header_t *ip60) +{ + return 0; +} + +static_always_inline u32 +lb_node_get_hash(vlib_buffer_t *p, u8 is_input_v4) +{ + u32 hash; + if (is_input_v4) + { + ip4_header_t *ip40; + u64 ports; + ip40 = vlib_buffer_get_current (p); + if (PREDICT_TRUE (ip40->protocol == IP_PROTOCOL_TCP || + ip40->protocol == IP_PROTOCOL_UDP)) + ports = ((u64)((udp_header_t *)(ip40 + 1))->src_port << 16) | + ((u64)((udp_header_t *)(ip40 + 1))->dst_port); + else + ports = lb_node_get_other_ports4(ip40); + + hash = lb_hash_hash(*((u64 *)&ip40->address_pair), ports, + 0, 0, 0); + } + else + { + ip6_header_t *ip60; + ip60 = vlib_buffer_get_current (p); + u64 ports; + if (PREDICT_TRUE (ip60->protocol == IP_PROTOCOL_TCP || + ip60->protocol == IP_PROTOCOL_UDP)) + ports = ((u64)((udp_header_t *)(ip60 + 1))->src_port << 16) | + ((u64)((udp_header_t *)(ip60 + 1))->dst_port); + else + ports = lb_node_get_other_ports6(ip60); + + hash = lb_hash_hash(ip60->src_address.as_u64[0], + ip60->src_address.as_u64[1], + ip60->dst_address.as_u64[0], + ip60->dst_address.as_u64[1], + ports); + } + return hash; +} + +static_always_inline uword +lb_node_fn (vlib_main_t * vm, + vlib_node_runtime_t * node, vlib_frame_t * frame, + u8 is_input_v4, //Compile-time parameter stating that is input is v4 (or v6) + u8 is_encap_v4) //Compile-time parameter stating that is GRE encap is v4 (or v6) +{ + lb_main_t *lbm = &lb_main; + u32 n_left_from, *from, next_index, *to_next, n_left_to_next; + u32 cpu_index = os_get_cpu_number(); + u32 lb_time = lb_hash_time_now(vm); + + lb_hash_t *sticky_ht = lb_get_sticky_table(cpu_index); + from = vlib_frame_vector_args (frame); + n_left_from = frame->n_vectors; + next_index = node->cached_next_index; + + u32 nexthash0 = 0; + if (PREDICT_TRUE(n_left_from > 0)) + nexthash0 = lb_node_get_hash(vlib_get_buffer (vm, from[0]), is_input_v4); + + while (n_left_from > 0) + { + vlib_get_next_frame (vm, node, next_index, to_next, n_left_to_next); + while (n_left_from > 0 && n_left_to_next > 0) + { + u32 pi0; + vlib_buffer_t *p0; + lb_vip_t *vip0; + u32 asindex0; + u16 len0; + u32 available_index0; + u8 counter = 0; + u32 hash0 = nexthash0; + + if (PREDICT_TRUE(n_left_from > 1)) + { + vlib_buffer_t *p1 = vlib_get_buffer (vm, from[1]); + //Compute next hash and prefetch bucket + nexthash0 = lb_node_get_hash(p1, is_input_v4); + lb_hash_prefetch_bucket(sticky_ht, nexthash0); + //Prefetch for encap, next + CLIB_PREFETCH (vlib_buffer_get_current(p1) - 64, 64, STORE); + } + + if (PREDICT_TRUE(n_left_from > 2)) + { + vlib_buffer_t *p2; + p2 = vlib_get_buffer(vm, from[2]); + /* prefetch packet header and data */ + vlib_prefetch_buffer_header(p2, STORE); + CLIB_PREFETCH (vlib_buffer_get_current(p2), 64, STORE); + } + + pi0 = to_next[0] = from[0]; + from += 1; + n_left_from -= 1; + to_next += 1; + n_left_to_next -= 1; + + p0 = vlib_get_buffer (vm, pi0); + vip0 = pool_elt_at_index (lbm->vips, + vnet_buffer (p0)->ip.adj_index[VLIB_TX]); + + if (is_input_v4) + { + ip4_header_t *ip40; + ip40 = vlib_buffer_get_current (p0); + len0 = clib_net_to_host_u16(ip40->length); + } + else + { + ip6_header_t *ip60; + ip60 = vlib_buffer_get_current (p0); + len0 = clib_net_to_host_u16(ip60->payload_length) + sizeof(ip6_header_t); + } + + lb_hash_get(sticky_ht, hash0, vnet_buffer (p0)->ip.adj_index[VLIB_TX], + lb_time, &available_index0, &asindex0); + + if (PREDICT_TRUE(asindex0 != ~0)) + { + //Found an existing entry + counter = LB_VIP_COUNTER_NEXT_PACKET; + } + else if (PREDICT_TRUE(available_index0 != ~0)) + { + //There is an available slot for a new flow + asindex0 = vip0->new_flow_table[hash0 & vip0->new_flow_table_mask].as_index; + counter = LB_VIP_COUNTER_FIRST_PACKET; + counter = (asindex0 == 0)?LB_VIP_COUNTER_NO_SERVER:counter; + + //TODO: There are race conditions with as0 and vip0 manipulation. + //Configuration may be changed, vectors resized, etc... + + //Dereference previously used + vlib_refcount_add(&lbm->as_refcount, cpu_index, + lb_hash_available_value(sticky_ht, hash0, available_index0), -1); + vlib_refcount_add(&lbm->as_refcount, cpu_index, + asindex0, 1); + + //Add sticky entry + //Note that when there is no AS configured, an entry is configured anyway. + //But no configured AS is not something that should happen + lb_hash_put(sticky_ht, hash0, asindex0, + vnet_buffer (p0)->ip.adj_index[VLIB_TX], + available_index0, lb_time); + } + else + { + //Could not store new entry in the table + asindex0 = vip0->new_flow_table[hash0 & vip0->new_flow_table_mask].as_index; + counter = LB_VIP_COUNTER_UNTRACKED_PACKET; + } + + vlib_increment_simple_counter(&lbm->vip_counters[counter], + cpu_index, + vnet_buffer (p0)->ip.adj_index[VLIB_TX], + 1); + + //Now let's encap + { + gre_header_t *gre0; + if (is_encap_v4) + { + ip4_header_t *ip40; + vlib_buffer_advance(p0, - sizeof(ip4_header_t) - sizeof(gre_header_t)); + ip40 = vlib_buffer_get_current(p0); + gre0 = (gre_header_t *)(ip40 + 1); + ip40->src_address = lbm->ip4_src_address; + ip40->dst_address = lbm->ass[asindex0].address.ip4; + ip40->ip_version_and_header_length = 0x45; + ip40->ttl = 128; + ip40->length = clib_host_to_net_u16(len0 + sizeof(gre_header_t) + sizeof(ip4_header_t)); + ip40->protocol = IP_PROTOCOL_GRE; + ip40->checksum = ip4_header_checksum (ip40); + } + else + { + ip6_header_t *ip60; + vlib_buffer_advance(p0, - sizeof(ip6_header_t) - sizeof(gre_header_t)); + ip60 = vlib_buffer_get_current(p0); + gre0 = (gre_header_t *)(ip60 + 1); + ip60->dst_address = lbm->ass[asindex0].address.ip6; + ip60->src_address = lbm->ip6_src_address; + ip60->hop_limit = 128; + ip60->ip_version_traffic_class_and_flow_label = clib_host_to_net_u32 (0x6<<28); + ip60->payload_length = clib_host_to_net_u16(len0 + sizeof(gre_header_t)); + ip60->protocol = IP_PROTOCOL_GRE; + } + + gre0->flags_and_version = 0; + gre0->protocol = (is_input_v4)? + clib_host_to_net_u16(0x0800): + clib_host_to_net_u16(0x86DD); + } + + if (PREDICT_FALSE (p0->flags & VLIB_BUFFER_IS_TRACED)) + { + lb_trace_t *tr = vlib_add_trace (vm, node, p0, sizeof (*tr)); + tr->as_index = asindex0; + tr->vip_index = vnet_buffer (p0)->ip.adj_index[VLIB_TX]; + } + + //Enqueue to next + //Note that this is going to error if asindex0 == 0 + vnet_buffer (p0)->ip.adj_index[VLIB_TX] = lbm->ass[asindex0].dpo.dpoi_index; + vlib_validate_buffer_enqueue_x1 (vm, node, next_index, to_next, + n_left_to_next, pi0, + lbm->ass[asindex0].dpo.dpoi_next_node); + } + vlib_put_next_frame (vm, node, next_index, n_left_to_next); + } + + return frame->n_vectors; +} + +static uword +lb6_gre6_node_fn (vlib_main_t * vm, + vlib_node_runtime_t * node, vlib_frame_t * frame) +{ + return lb_node_fn(vm, node, frame, 0, 0); +} + +static uword +lb6_gre4_node_fn (vlib_main_t * vm, + vlib_node_runtime_t * node, vlib_frame_t * frame) +{ + return lb_node_fn(vm, node, frame, 0, 1); +} + +static uword +lb4_gre6_node_fn (vlib_main_t * vm, + vlib_node_runtime_t * node, vlib_frame_t * frame) +{ + return lb_node_fn(vm, node, frame, 1, 0); +} + +static uword +lb4_gre4_node_fn (vlib_main_t * vm, + vlib_node_runtime_t * node, vlib_frame_t * frame) +{ + return lb_node_fn(vm, node, frame, 1, 1); +} + +VLIB_REGISTER_NODE (lb6_gre6_node) = +{ + .function = lb6_gre6_node_fn, + .name = "lb6-gre6", + .vector_size = sizeof (u32), + .format_trace = format_lb_trace, + + .n_errors = LB_N_ERROR, + .error_strings = lb_error_strings, + + .n_next_nodes = LB_N_NEXT, + .next_nodes = + { + [LB_NEXT_DROP] = "error-drop" + }, +}; + +VLIB_REGISTER_NODE (lb6_gre4_node) = +{ + .function = lb6_gre4_node_fn, + .name = "lb6-gre4", + .vector_size = sizeof (u32), + .format_trace = format_lb_trace, + + .n_errors = LB_N_ERROR, + .error_strings = lb_error_strings, + + .n_next_nodes = LB_N_NEXT, + .next_nodes = + { + [LB_NEXT_DROP] = "error-drop" + }, +}; + +VLIB_REGISTER_NODE (lb4_gre6_node) = +{ + .function = lb4_gre6_node_fn, + .name = "lb4-gre6", + .vector_size = sizeof (u32), + .format_trace = format_lb_trace, + + .n_errors = LB_N_ERROR, + .error_strings = lb_error_strings, + + .n_next_nodes = LB_N_NEXT, + .next_nodes = + { + [LB_NEXT_DROP] = "error-drop" + }, +}; + +VLIB_REGISTER_NODE (lb4_gre4_node) = +{ + .function = lb4_gre4_node_fn, + .name = "lb4-gre4", + .vector_size = sizeof (u32), + .format_trace = format_lb_trace, + + .n_errors = LB_N_ERROR, + .error_strings = lb_error_strings, + + .n_next_nodes = LB_N_NEXT, + .next_nodes = + { + [LB_NEXT_DROP] = "error-drop" + }, +}; + diff --git a/src/plugins/lb/refcount.c b/src/plugins/lb/refcount.c new file mode 100644 index 00000000..22415c88 --- /dev/null +++ b/src/plugins/lb/refcount.c @@ -0,0 +1,41 @@ +/* + * Copyright (c) 2016 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +#include + +void __vlib_refcount_resize(vlib_refcount_per_cpu_t *per_cpu, u32 size) +{ + u32 *new_counter = 0, *old_counter; + vec_validate(new_counter, size); + memcpy(new_counter, per_cpu->counters, per_cpu->length); + old_counter = per_cpu->counters; + per_cpu->counters = new_counter; + CLIB_MEMORY_BARRIER(); + per_cpu->length = vec_len(new_counter); + vec_free(old_counter); +} + +u64 vlib_refcount_get(vlib_refcount_t *r, u32 index) +{ + u64 count = 0; + vlib_thread_main_t *tm = vlib_get_thread_main (); + u32 cpu_index; + for (cpu_index = 0; cpu_index < tm->n_vlib_mains; cpu_index++) { + if (r->per_cpu[cpu_index].length > index) + count += r->per_cpu[cpu_index].counters[index]; + } + return count; +} + diff --git a/src/plugins/lb/refcount.h b/src/plugins/lb/refcount.h new file mode 100644 index 00000000..8c26e7be --- /dev/null +++ b/src/plugins/lb/refcount.h @@ -0,0 +1,67 @@ +/* + * Copyright (c) 2016 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +/* + * vlib provides lock-free counters but those + * - Have 16bits per-CPU counter, which may overflow. + * - Would only increment. + * + * This is very similar to vlib counters, but may be used to count reference. + * Such a counter includes an arbitrary number of counters. Each counter + * is identified by its index. This is used to aggregate per-cpu memory. + * + * Warning: + * This reference counter is lock-free but is not race-condition free. + * The counting result is approximate and another mechanism needs to be used + * in order to ensure that an object may be freed. + * + */ + +#include + +typedef struct { + u32 *counters; + u32 length; + u32 *reader_lengths; + CLIB_CACHE_LINE_ALIGN_MARK(o); +} vlib_refcount_per_cpu_t; + +typedef struct { + vlib_refcount_per_cpu_t *per_cpu; +} vlib_refcount_t; + +void __vlib_refcount_resize(vlib_refcount_per_cpu_t *per_cpu, u32 size); + +static_always_inline +void vlib_refcount_add(vlib_refcount_t *r, u32 cpu_index, u32 counter_index, i32 v) +{ + vlib_refcount_per_cpu_t *per_cpu = &r->per_cpu[cpu_index]; + if (PREDICT_FALSE(counter_index >= per_cpu->length)) + __vlib_refcount_resize(per_cpu, clib_max(counter_index + 16, per_cpu->length * 2)); + + per_cpu->counters[counter_index] += v; +} + +u64 vlib_refcount_get(vlib_refcount_t *r, u32 index); + +static_always_inline +void vlib_refcount_init(vlib_refcount_t *r) +{ + vlib_thread_main_t *tm = vlib_get_thread_main (); + r->per_cpu = 0; + vec_validate (r->per_cpu, tm->n_vlib_mains - 1); +} + + diff --git a/src/plugins/lb/util.c b/src/plugins/lb/util.c new file mode 100644 index 00000000..d969d168 --- /dev/null +++ b/src/plugins/lb/util.c @@ -0,0 +1,72 @@ +/* + * Copyright (c) 2016 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +#include + +void ip46_prefix_normalize(ip46_address_t *prefix, u8 plen) +{ + if (plen == 0) { + prefix->as_u64[0] = 0; + prefix->as_u64[1] = 0; + } else if (plen <= 64) { + prefix->as_u64[0] &= clib_host_to_net_u64(0xffffffffffffffffL << (64 - plen)); + prefix->as_u64[1] = 0; + } else { + prefix->as_u64[1] &= clib_host_to_net_u64(0xffffffffffffffffL << (128 - plen)); + } + +} + +uword unformat_ip46_prefix (unformat_input_t * input, va_list * args) +{ + ip46_address_t *ip46 = va_arg (*args, ip46_address_t *); + u8 *len = va_arg (*args, u8 *); + ip46_type_t type = va_arg (*args, ip46_type_t); + + u32 l; + if ((type != IP46_TYPE_IP6) && unformat(input, "%U/%u", unformat_ip4_address, &ip46->ip4, &l)) { + if (l > 32) + return 0; + *len = l + 96; + ip46->pad[0] = ip46->pad[1] = ip46->pad[2] = 0; + } else if ((type != IP46_TYPE_IP4) && unformat(input, "%U/%u", unformat_ip6_address, &ip46->ip6, &l)) { + if (l > 128) + return 0; + *len = l; + } else { + return 0; + } + return 1; +} + +u8 *format_ip46_prefix (u8 * s, va_list * args) +{ + ip46_address_t *ip46 = va_arg (*args, ip46_address_t *); + u32 len = va_arg (*args, u32); //va_arg cannot use u8 or u16 + ip46_type_t type = va_arg (*args, ip46_type_t); + + int is_ip4 = 0; + if (type == IP46_TYPE_IP4) + is_ip4 = 1; + else if (type == IP46_TYPE_IP6) + is_ip4 = 0; + else + is_ip4 = (len >= 96) && ip46_address_is_ip4(ip46); + + return is_ip4 ? + format(s, "%U/%d", format_ip4_address, &ip46->ip4, len - 96): + format(s, "%U/%d", format_ip6_address, &ip46->ip6, len); +} + diff --git a/src/plugins/lb/util.h b/src/plugins/lb/util.h new file mode 100644 index 00000000..3f082310 --- /dev/null +++ b/src/plugins/lb/util.h @@ -0,0 +1,40 @@ +/* + * Copyright (c) 2016 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +/* + * Non-LB specific stuff comes here + */ + +#ifndef LB_PLUGIN_LB_UTIL_H_ +#define LB_PLUGIN_LB_UTIL_H_ + +#include +#include + +#define ip46_address_type(ip46) (ip46_address_is_ip4(ip46)?IP46_TYPE_IP4:IP46_TYPE_IP6) +#define ip46_prefix_is_ip4(ip46, len) ((len) >= 96 && ip46_address_is_ip4(ip46)) +#define ip46_prefix_type(ip46, len) (ip46_prefix_is_ip4(ip46, len)?IP46_TYPE_IP4:IP46_TYPE_IP6) + +void ip46_prefix_normalize(ip46_address_t *prefix, u8 plen); +uword unformat_ip46_prefix (unformat_input_t * input, va_list * args); +u8 *format_ip46_prefix (u8 * s, va_list * args); + +/** + * 32 bits integer comparison for running values. + * 1 > 0 is true. But 1 > 0xffffffff also is. + */ +#define clib_u32_loop_gt(a, b) (((u32)(a)) - ((u32)(b)) < 0x7fffffff) + +#endif /* LB_PLUGIN_LB_UTIL_H_ */ diff --git a/src/plugins/snat.am b/src/plugins/snat.am new file mode 100644 index 00000000..7ff2386e --- /dev/null +++ b/src/plugins/snat.am @@ -0,0 +1,33 @@ + +# Copyright (c) +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at: +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +vppapitestplugins_LTLIBRARIES += snat_test_plugin.la +vppplugins_LTLIBRARIES += snat_plugin.la + +snat_plugin_la_SOURCES = snat/snat.c \ + snat/in2out.c \ + snat/out2in.c \ + snat/snat_plugin.api.h + +API_FILES += snat/snat.api + +nobase_apiinclude_HEADERS += \ + snat/snat_all_api_h.h \ + snat/snat_msg_enum.h \ + snat/snat.api.h + +snat_test_plugin_la_SOURCES = \ + snat/snat_test.c snat/snat_plugin.api.h + +# vi:syntax=automake diff --git a/src/plugins/snat/in2out.c b/src/plugins/snat/in2out.c new file mode 100644 index 00000000..c78fdd76 --- /dev/null +++ b/src/plugins/snat/in2out.c @@ -0,0 +1,1597 @@ +/* + * Copyright (c) 2016 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +#include +#include +#include +#include + +#include +#include +#include +#include + +#include +#include +#include + +typedef struct { + u32 sw_if_index; + u32 next_index; + u32 session_index; + u32 is_slow_path; +} snat_in2out_trace_t; + +typedef struct { + u32 next_worker_index; + u8 do_handoff; +} snat_in2out_worker_handoff_trace_t; + +/* packet trace format function */ +static u8 * format_snat_in2out_trace (u8 * s, va_list * args) +{ + CLIB_UNUSED (vlib_main_t * vm) = va_arg (*args, vlib_main_t *); + CLIB_UNUSED (vlib_node_t * node) = va_arg (*args, vlib_node_t *); + snat_in2out_trace_t * t = va_arg (*args, snat_in2out_trace_t *); + char * tag; + + tag = t->is_slow_path ? "SNAT_IN2OUT_SLOW_PATH" : "SNAT_IN2OUT_FAST_PATH"; + + s = format (s, "%s: sw_if_index %d, next index %d, session %d", tag, + t->sw_if_index, t->next_index, t->session_index); + + return s; +} + +static u8 * format_snat_in2out_fast_trace (u8 * s, va_list * args) +{ + CLIB_UNUSED (vlib_main_t * vm) = va_arg (*args, vlib_main_t *); + CLIB_UNUSED (vlib_node_t * node) = va_arg (*args, vlib_node_t *); + snat_in2out_trace_t * t = va_arg (*args, snat_in2out_trace_t *); + + s = format (s, "SANT_IN2OUT_FAST: sw_if_index %d, next index %d", + t->sw_if_index, t->next_index); + + return s; +} + +static u8 * format_snat_in2out_worker_handoff_trace (u8 * s, va_list * args) +{ + CLIB_UNUSED (vlib_main_t * vm) = va_arg (*args, vlib_main_t *); + CLIB_UNUSED (vlib_node_t * node) = va_arg (*args, vlib_node_t *); + snat_in2out_worker_handoff_trace_t * t = + va_arg (*args, snat_in2out_worker_handoff_trace_t *); + char * m; + + m = t->do_handoff ? "next worker" : "same worker"; + s = format (s, "SNAT_IN2OUT_WORKER_HANDOFF: %s %d", m, t->next_worker_index); + + return s; +} + +vlib_node_registration_t snat_in2out_node; +vlib_node_registration_t snat_in2out_slowpath_node; +vlib_node_registration_t snat_in2out_fast_node; +vlib_node_registration_t snat_in2out_worker_handoff_node; + +#define foreach_snat_in2out_error \ +_(UNSUPPORTED_PROTOCOL, "Unsupported protocol") \ +_(IN2OUT_PACKETS, "Good in2out packets processed") \ +_(OUT_OF_PORTS, "Out of ports") \ +_(BAD_OUTSIDE_FIB, "Outside VRF ID not found") \ +_(BAD_ICMP_TYPE, "icmp type not echo-request") \ +_(NO_TRANSLATION, "No translation") + +typedef enum { +#define _(sym,str) SNAT_IN2OUT_ERROR_##sym, + foreach_snat_in2out_error +#undef _ + SNAT_IN2OUT_N_ERROR, +} snat_in2out_error_t; + +static char * snat_in2out_error_strings[] = { +#define _(sym,string) string, + foreach_snat_in2out_error +#undef _ +}; + +typedef enum { + SNAT_IN2OUT_NEXT_LOOKUP, + SNAT_IN2OUT_NEXT_DROP, + SNAT_IN2OUT_NEXT_SLOW_PATH, + SNAT_IN2OUT_N_NEXT, +} snat_in2out_next_t; + +static u32 slow_path (snat_main_t *sm, vlib_buffer_t *b0, + ip4_header_t * ip0, + u32 rx_fib_index0, + snat_session_key_t * key0, + snat_session_t ** sessionp, + vlib_node_runtime_t * node, + u32 next0, + u32 cpu_index) +{ + snat_user_t *u; + snat_user_key_t user_key; + snat_session_t *s; + clib_bihash_kv_8_8_t kv0, value0; + u32 oldest_per_user_translation_list_index; + dlist_elt_t * oldest_per_user_translation_list_elt; + dlist_elt_t * per_user_translation_list_elt; + dlist_elt_t * per_user_list_head_elt; + u32 session_index; + snat_session_key_t key1; + u32 address_index = ~0; + u32 outside_fib_index; + uword * p; + snat_static_mapping_key_t worker_by_out_key; + + p = hash_get (sm->ip4_main->fib_index_by_table_id, sm->outside_vrf_id); + if (! p) + { + b0->error = node->errors[SNAT_IN2OUT_ERROR_BAD_OUTSIDE_FIB]; + return SNAT_IN2OUT_NEXT_DROP; + } + outside_fib_index = p[0]; + + user_key.addr = ip0->src_address; + user_key.fib_index = rx_fib_index0; + kv0.key = user_key.as_u64; + + /* Ever heard of the "user" = src ip4 address before? */ + if (clib_bihash_search_8_8 (&sm->user_hash, &kv0, &value0)) + { + /* no, make a new one */ + pool_get (sm->per_thread_data[cpu_index].users, u); + memset (u, 0, sizeof (*u)); + u->addr = ip0->src_address; + + pool_get (sm->per_thread_data[cpu_index].list_pool, per_user_list_head_elt); + + u->sessions_per_user_list_head_index = per_user_list_head_elt - + sm->per_thread_data[cpu_index].list_pool; + + clib_dlist_init (sm->per_thread_data[cpu_index].list_pool, + u->sessions_per_user_list_head_index); + + kv0.value = u - sm->per_thread_data[cpu_index].users; + + /* add user */ + clib_bihash_add_del_8_8 (&sm->user_hash, &kv0, 1 /* is_add */); + } + else + { + u = pool_elt_at_index (sm->per_thread_data[cpu_index].users, + value0.value); + } + + /* Over quota? Recycle the least recently used dynamic translation */ + if (u->nsessions >= sm->max_translations_per_user) + { + /* Remove the oldest dynamic translation */ + do { + oldest_per_user_translation_list_index = + clib_dlist_remove_head (sm->per_thread_data[cpu_index].list_pool, + u->sessions_per_user_list_head_index); + + ASSERT (oldest_per_user_translation_list_index != ~0); + + /* add it back to the end of the LRU list */ + clib_dlist_addtail (sm->per_thread_data[cpu_index].list_pool, + u->sessions_per_user_list_head_index, + oldest_per_user_translation_list_index); + /* Get the list element */ + oldest_per_user_translation_list_elt = + pool_elt_at_index (sm->per_thread_data[cpu_index].list_pool, + oldest_per_user_translation_list_index); + + /* Get the session index from the list element */ + session_index = oldest_per_user_translation_list_elt->value; + + /* Get the session */ + s = pool_elt_at_index (sm->per_thread_data[cpu_index].sessions, + session_index); + } while (snat_is_session_static (s)); + + /* Remove in2out, out2in keys */ + kv0.key = s->in2out.as_u64; + if (clib_bihash_add_del_8_8 (&sm->in2out, &kv0, 0 /* is_add */)) + clib_warning ("in2out key delete failed"); + kv0.key = s->out2in.as_u64; + if (clib_bihash_add_del_8_8 (&sm->out2in, &kv0, 0 /* is_add */)) + clib_warning ("out2in key delete failed"); + + snat_free_outside_address_and_port + (sm, &s->out2in, s->outside_address_index); + s->outside_address_index = ~0; + + if (snat_alloc_outside_address_and_port (sm, &key1, &address_index)) + { + ASSERT(0); + + b0->error = node->errors[SNAT_IN2OUT_ERROR_OUT_OF_PORTS]; + return SNAT_IN2OUT_NEXT_DROP; + } + s->outside_address_index = address_index; + } + else + { + u8 static_mapping = 1; + + /* First try to match static mapping by local address and port */ + if (snat_static_mapping_match (sm, *key0, &key1, 0)) + { + static_mapping = 0; + /* Try to create dynamic translation */ + if (snat_alloc_outside_address_and_port (sm, &key1, &address_index)) + { + b0->error = node->errors[SNAT_IN2OUT_ERROR_OUT_OF_PORTS]; + return SNAT_IN2OUT_NEXT_DROP; + } + } + + /* Create a new session */ + pool_get (sm->per_thread_data[cpu_index].sessions, s); + memset (s, 0, sizeof (*s)); + + s->outside_address_index = address_index; + + if (static_mapping) + { + u->nstaticsessions++; + s->flags |= SNAT_SESSION_FLAG_STATIC_MAPPING; + } + else + { + u->nsessions++; + } + + /* Create list elts */ + pool_get (sm->per_thread_data[cpu_index].list_pool, + per_user_translation_list_elt); + clib_dlist_init (sm->per_thread_data[cpu_index].list_pool, + per_user_translation_list_elt - + sm->per_thread_data[cpu_index].list_pool); + + per_user_translation_list_elt->value = + s - sm->per_thread_data[cpu_index].sessions; + s->per_user_index = per_user_translation_list_elt - + sm->per_thread_data[cpu_index].list_pool; + s->per_user_list_head_index = u->sessions_per_user_list_head_index; + + clib_dlist_addtail (sm->per_thread_data[cpu_index].list_pool, + s->per_user_list_head_index, + per_user_translation_list_elt - + sm->per_thread_data[cpu_index].list_pool); + } + + s->in2out = *key0; + s->out2in = key1; + s->out2in.protocol = key0->protocol; + s->out2in.fib_index = outside_fib_index; + *sessionp = s; + + /* Add to translation hashes */ + kv0.key = s->in2out.as_u64; + kv0.value = s - sm->per_thread_data[cpu_index].sessions; + if (clib_bihash_add_del_8_8 (&sm->in2out, &kv0, 1 /* is_add */)) + clib_warning ("in2out key add failed"); + + kv0.key = s->out2in.as_u64; + kv0.value = s - sm->per_thread_data[cpu_index].sessions; + + if (clib_bihash_add_del_8_8 (&sm->out2in, &kv0, 1 /* is_add */)) + clib_warning ("out2in key add failed"); + + /* Add to translated packets worker lookup */ + worker_by_out_key.addr = s->out2in.addr; + worker_by_out_key.port = s->out2in.port; + worker_by_out_key.fib_index = s->out2in.fib_index; + kv0.key = worker_by_out_key.as_u64; + kv0.value = cpu_index; + clib_bihash_add_del_8_8 (&sm->worker_by_out, &kv0, 1); + return next0; +} + +static inline u32 icmp_in2out_slow_path (snat_main_t *sm, + vlib_buffer_t * b0, + ip4_header_t * ip0, + icmp46_header_t * icmp0, + u32 sw_if_index0, + u32 rx_fib_index0, + vlib_node_runtime_t * node, + u32 next0, + f64 now, + u32 cpu_index) +{ + snat_session_key_t key0; + icmp_echo_header_t *echo0; + clib_bihash_kv_8_8_t kv0, value0; + snat_session_t * s0; + u32 new_addr0, old_addr0; + u16 old_id0, new_id0; + ip_csum_t sum0; + snat_runtime_t * rt = (snat_runtime_t *)node->runtime_data; + + if (PREDICT_FALSE(icmp0->type != ICMP4_echo_request)) + { + b0->error = node->errors[SNAT_IN2OUT_ERROR_BAD_ICMP_TYPE]; + return SNAT_IN2OUT_NEXT_DROP; + } + + echo0 = (icmp_echo_header_t *)(icmp0+1); + + key0.addr = ip0->src_address; + key0.port = echo0->identifier; + key0.protocol = SNAT_PROTOCOL_ICMP; + key0.fib_index = rx_fib_index0; + + kv0.key = key0.as_u64; + + if (clib_bihash_search_8_8 (&sm->in2out, &kv0, &value0)) + { + ip4_address_t * first_int_addr; + + if (PREDICT_FALSE(rt->cached_sw_if_index != sw_if_index0)) + { + first_int_addr = + ip4_interface_first_address (sm->ip4_main, sw_if_index0, + 0 /* just want the address */); + rt->cached_sw_if_index = sw_if_index0; + rt->cached_ip4_address = first_int_addr->as_u32; + } + + /* Don't NAT packet aimed at the intfc address */ + if (PREDICT_FALSE(ip0->dst_address.as_u32 == + rt->cached_ip4_address)) + return next0; + + next0 = slow_path (sm, b0, ip0, rx_fib_index0, &key0, + &s0, node, next0, cpu_index); + + if (PREDICT_FALSE (next0 == SNAT_IN2OUT_NEXT_DROP)) + return next0; + } + else + s0 = pool_elt_at_index (sm->per_thread_data[cpu_index].sessions, + value0.value); + + old_addr0 = ip0->src_address.as_u32; + ip0->src_address = s0->out2in.addr; + new_addr0 = ip0->src_address.as_u32; + vnet_buffer(b0)->sw_if_index[VLIB_TX] = s0->out2in.fib_index; + + sum0 = ip0->checksum; + sum0 = ip_csum_update (sum0, old_addr0, new_addr0, + ip4_header_t, + src_address /* changed member */); + ip0->checksum = ip_csum_fold (sum0); + + old_id0 = echo0->identifier; + new_id0 = s0->out2in.port; + echo0->identifier = new_id0; + + sum0 = icmp0->checksum; + sum0 = ip_csum_update (sum0, old_id0, new_id0, icmp_echo_header_t, + identifier); + icmp0->checksum = ip_csum_fold (sum0); + + /* Accounting */ + s0->last_heard = now; + s0->total_pkts++; + s0->total_bytes += vlib_buffer_length_in_chain (sm->vlib_main, b0); + /* Per-user LRU list maintenance for dynamic translations */ + if (!snat_is_session_static (s0)) + { + clib_dlist_remove (sm->per_thread_data[cpu_index].list_pool, + s0->per_user_index); + clib_dlist_addtail (sm->per_thread_data[cpu_index].list_pool, + s0->per_user_list_head_index, + s0->per_user_index); + } + + return next0; +} + +/** + * @brief Hairpinning + * + * Hairpinning allows two endpoints on the internal side of the NAT to + * communicate even if they only use each other's external IP addresses + * and ports. + * + * @param sm SNAT main. + * @param b0 Vlib buffer. + * @param ip0 IP header. + * @param udp0 UDP header. + * @param tcp0 TCP header. + * @param proto0 SNAT protocol. + */ +static inline void +snat_hairpinning (snat_main_t *sm, + vlib_buffer_t * b0, + ip4_header_t * ip0, + udp_header_t * udp0, + tcp_header_t * tcp0, + u32 proto0) +{ + snat_session_key_t key0, sm0; + snat_static_mapping_key_t k0; + snat_session_t * s0; + clib_bihash_kv_8_8_t kv0, value0; + ip_csum_t sum0; + u32 new_dst_addr0 = 0, old_dst_addr0, ti = 0, si; + u16 new_dst_port0, old_dst_port0; + + key0.addr = ip0->dst_address; + key0.port = udp0->dst_port; + key0.protocol = proto0; + key0.fib_index = sm->outside_fib_index; + kv0.key = key0.as_u64; + + /* Check if destination is in active sessions */ + if (clib_bihash_search_8_8 (&sm->out2in, &kv0, &value0)) + { + /* or static mappings */ + if (!snat_static_mapping_match(sm, key0, &sm0, 1)) + { + new_dst_addr0 = sm0.addr.as_u32; + new_dst_port0 = sm0.port; + vnet_buffer(b0)->sw_if_index[VLIB_TX] = sm0.fib_index; + } + } + else + { + si = value0.value; + if (sm->num_workers > 1) + { + k0.addr = ip0->dst_address; + k0.port = udp0->dst_port; + k0.fib_index = sm->outside_fib_index; + kv0.key = k0.as_u64; + if (clib_bihash_search_8_8 (&sm->worker_by_out, &kv0, &value0)) + ASSERT(0); + else + ti = value0.value; + } + else + ti = sm->num_workers; + + s0 = pool_elt_at_index (sm->per_thread_data[ti].sessions, si); + new_dst_addr0 = s0->in2out.addr.as_u32; + new_dst_port0 = s0->in2out.port; + vnet_buffer(b0)->sw_if_index[VLIB_TX] = s0->in2out.fib_index; + } + + /* Destination is behind the same NAT, use internal address and port */ + if (new_dst_addr0) + { + old_dst_addr0 = ip0->dst_address.as_u32; + ip0->dst_address.as_u32 = new_dst_addr0; + sum0 = ip0->checksum; + sum0 = ip_csum_update (sum0, old_dst_addr0, new_dst_addr0, + ip4_header_t, dst_address); + ip0->checksum = ip_csum_fold (sum0); + + old_dst_port0 = tcp0->ports.dst; + if (PREDICT_TRUE(new_dst_port0 != old_dst_port0)) + { + if (PREDICT_TRUE(proto0 == SNAT_PROTOCOL_TCP)) + { + tcp0->ports.dst = new_dst_port0; + sum0 = tcp0->checksum; + sum0 = ip_csum_update (sum0, old_dst_addr0, new_dst_addr0, + ip4_header_t, dst_address); + sum0 = ip_csum_update (sum0, old_dst_port0, new_dst_port0, + ip4_header_t /* cheat */, length); + tcp0->checksum = ip_csum_fold(sum0); + } + else + { + udp0->dst_port = new_dst_port0; + udp0->checksum = 0; + } + } + } +} + +static inline uword +snat_in2out_node_fn_inline (vlib_main_t * vm, + vlib_node_runtime_t * node, + vlib_frame_t * frame, int is_slow_path) +{ + u32 n_left_from, * from, * to_next; + snat_in2out_next_t next_index; + u32 pkts_processed = 0; + snat_main_t * sm = &snat_main; + snat_runtime_t * rt = (snat_runtime_t *)node->runtime_data; + f64 now = vlib_time_now (vm); + u32 stats_node_index; + u32 cpu_index = os_get_cpu_number (); + + stats_node_index = is_slow_path ? snat_in2out_slowpath_node.index : + snat_in2out_node.index; + + from = vlib_frame_vector_args (frame); + n_left_from = frame->n_vectors; + next_index = node->cached_next_index; + + while (n_left_from > 0) + { + u32 n_left_to_next; + + vlib_get_next_frame (vm, node, next_index, + to_next, n_left_to_next); + + while (n_left_from >= 4 && n_left_to_next >= 2) + { + u32 bi0, bi1; + vlib_buffer_t * b0, * b1; + u32 next0, next1; + u32 sw_if_index0, sw_if_index1; + ip4_header_t * ip0, * ip1; + ip_csum_t sum0, sum1; + u32 new_addr0, old_addr0, new_addr1, old_addr1; + u16 old_port0, new_port0, old_port1, new_port1; + udp_header_t * udp0, * udp1; + tcp_header_t * tcp0, * tcp1; + icmp46_header_t * icmp0, * icmp1; + snat_session_key_t key0, key1; + u32 rx_fib_index0, rx_fib_index1; + u32 proto0, proto1; + snat_session_t * s0 = 0, * s1 = 0; + clib_bihash_kv_8_8_t kv0, value0, kv1, value1; + + /* Prefetch next iteration. */ + { + vlib_buffer_t * p2, * p3; + + p2 = vlib_get_buffer (vm, from[2]); + p3 = vlib_get_buffer (vm, from[3]); + + vlib_prefetch_buffer_header (p2, LOAD); + vlib_prefetch_buffer_header (p3, LOAD); + + CLIB_PREFETCH (p2->data, CLIB_CACHE_LINE_BYTES, STORE); + CLIB_PREFETCH (p3->data, CLIB_CACHE_LINE_BYTES, STORE); + } + + /* speculatively enqueue b0 and b1 to the current next frame */ + to_next[0] = bi0 = from[0]; + to_next[1] = bi1 = from[1]; + from += 2; + to_next += 2; + n_left_from -= 2; + n_left_to_next -= 2; + + b0 = vlib_get_buffer (vm, bi0); + b1 = vlib_get_buffer (vm, bi1); + + ip0 = vlib_buffer_get_current (b0); + udp0 = ip4_next_header (ip0); + tcp0 = (tcp_header_t *) udp0; + icmp0 = (icmp46_header_t *) udp0; + + sw_if_index0 = vnet_buffer(b0)->sw_if_index[VLIB_RX]; + rx_fib_index0 = vec_elt (sm->ip4_main->fib_index_by_sw_if_index, + sw_if_index0); + + next0 = next1 = SNAT_IN2OUT_NEXT_LOOKUP; + + proto0 = ~0; + proto0 = (ip0->protocol == IP_PROTOCOL_UDP) + ? SNAT_PROTOCOL_UDP : proto0; + proto0 = (ip0->protocol == IP_PROTOCOL_TCP) + ? SNAT_PROTOCOL_TCP : proto0; + proto0 = (ip0->protocol == IP_PROTOCOL_ICMP) + ? SNAT_PROTOCOL_ICMP : proto0; + + /* Next configured feature, probably ip4-lookup */ + if (is_slow_path) + { + if (PREDICT_FALSE (proto0 == ~0)) + goto trace00; + + if (PREDICT_FALSE (proto0 == SNAT_PROTOCOL_ICMP)) + { + next0 = icmp_in2out_slow_path + (sm, b0, ip0, icmp0, sw_if_index0, rx_fib_index0, + node, next0, now, cpu_index); + goto trace00; + } + } + else + { + if (PREDICT_FALSE (proto0 == ~0 || proto0 == SNAT_PROTOCOL_ICMP)) + { + next0 = SNAT_IN2OUT_NEXT_SLOW_PATH; + goto trace00; + } + } + + key0.addr = ip0->src_address; + key0.port = udp0->src_port; + key0.protocol = proto0; + key0.fib_index = rx_fib_index0; + + kv0.key = key0.as_u64; + + if (PREDICT_FALSE (clib_bihash_search_8_8 (&sm->in2out, &kv0, &value0) != 0)) + { + if (is_slow_path) + { + ip4_address_t * first_int_addr; + + if (PREDICT_FALSE(rt->cached_sw_if_index != sw_if_index0)) + { + first_int_addr = + ip4_interface_first_address (sm->ip4_main, sw_if_index0, + 0 /* just want the address */); + rt->cached_sw_if_index = sw_if_index0; + rt->cached_ip4_address = first_int_addr->as_u32; + } + + /* Don't NAT packet aimed at the intfc address */ + if (PREDICT_FALSE(ip0->dst_address.as_u32 == + rt->cached_ip4_address)) + goto trace00; + + next0 = slow_path (sm, b0, ip0, rx_fib_index0, &key0, + &s0, node, next0, cpu_index); + if (PREDICT_FALSE (next0 == SNAT_IN2OUT_NEXT_DROP)) + goto trace00; + } + else + { + next0 = SNAT_IN2OUT_NEXT_SLOW_PATH; + goto trace00; + } + } + else + s0 = pool_elt_at_index (sm->per_thread_data[cpu_index].sessions, + value0.value); + + old_addr0 = ip0->src_address.as_u32; + ip0->src_address = s0->out2in.addr; + new_addr0 = ip0->src_address.as_u32; + vnet_buffer(b0)->sw_if_index[VLIB_TX] = s0->out2in.fib_index; + + sum0 = ip0->checksum; + sum0 = ip_csum_update (sum0, old_addr0, new_addr0, + ip4_header_t, + src_address /* changed member */); + ip0->checksum = ip_csum_fold (sum0); + + if (PREDICT_TRUE(proto0 == SNAT_PROTOCOL_TCP)) + { + old_port0 = tcp0->ports.src; + tcp0->ports.src = s0->out2in.port; + new_port0 = tcp0->ports.src; + + sum0 = tcp0->checksum; + sum0 = ip_csum_update (sum0, old_addr0, new_addr0, + ip4_header_t, + dst_address /* changed member */); + sum0 = ip_csum_update (sum0, old_port0, new_port0, + ip4_header_t /* cheat */, + length /* changed member */); + tcp0->checksum = ip_csum_fold(sum0); + } + else + { + old_port0 = udp0->src_port; + udp0->src_port = s0->out2in.port; + udp0->checksum = 0; + } + + /* Hairpinning */ + snat_hairpinning (sm, b0, ip0, udp0, tcp0, proto0); + + /* Accounting */ + s0->last_heard = now; + s0->total_pkts++; + s0->total_bytes += vlib_buffer_length_in_chain (vm, b0); + /* Per-user LRU list maintenance for dynamic translation */ + if (!snat_is_session_static (s0)) + { + clib_dlist_remove (sm->per_thread_data[cpu_index].list_pool, + s0->per_user_index); + clib_dlist_addtail (sm->per_thread_data[cpu_index].list_pool, + s0->per_user_list_head_index, + s0->per_user_index); + } + trace00: + + if (PREDICT_FALSE((node->flags & VLIB_NODE_FLAG_TRACE) + && (b0->flags & VLIB_BUFFER_IS_TRACED))) + { + snat_in2out_trace_t *t = + vlib_add_trace (vm, node, b0, sizeof (*t)); + t->is_slow_path = is_slow_path; + t->sw_if_index = sw_if_index0; + t->next_index = next0; + t->session_index = ~0; + if (s0) + t->session_index = s0 - sm->per_thread_data[cpu_index].sessions; + } + + pkts_processed += next0 != SNAT_IN2OUT_NEXT_DROP; + + ip1 = vlib_buffer_get_current (b1); + udp1 = ip4_next_header (ip1); + tcp1 = (tcp_header_t *) udp1; + icmp1 = (icmp46_header_t *) udp1; + + sw_if_index1 = vnet_buffer(b1)->sw_if_index[VLIB_RX]; + rx_fib_index1 = vec_elt (sm->ip4_main->fib_index_by_sw_if_index, + sw_if_index1); + + proto1 = ~0; + proto1 = (ip1->protocol == IP_PROTOCOL_UDP) + ? SNAT_PROTOCOL_UDP : proto1; + proto1 = (ip1->protocol == IP_PROTOCOL_TCP) + ? SNAT_PROTOCOL_TCP : proto1; + proto1 = (ip1->protocol == IP_PROTOCOL_ICMP) + ? SNAT_PROTOCOL_ICMP : proto1; + + /* Next configured feature, probably ip4-lookup */ + if (is_slow_path) + { + if (PREDICT_FALSE (proto1 == ~0)) + goto trace01; + + if (PREDICT_FALSE (proto1 == SNAT_PROTOCOL_ICMP)) + { + next1 = icmp_in2out_slow_path + (sm, b1, ip1, icmp1, sw_if_index1, rx_fib_index1, node, + next1, now, cpu_index); + goto trace01; + } + } + else + { + if (PREDICT_FALSE (proto1 == ~0 || proto1 == SNAT_PROTOCOL_ICMP)) + { + next1 = SNAT_IN2OUT_NEXT_SLOW_PATH; + goto trace01; + } + } + + key1.addr = ip1->src_address; + key1.port = udp1->src_port; + key1.protocol = proto1; + key1.fib_index = rx_fib_index1; + + kv1.key = key1.as_u64; + + if (PREDICT_FALSE(clib_bihash_search_8_8 (&sm->in2out, &kv1, &value1) != 0)) + { + if (is_slow_path) + { + ip4_address_t * first_int_addr; + + if (PREDICT_FALSE(rt->cached_sw_if_index != sw_if_index1)) + { + first_int_addr = + ip4_interface_first_address (sm->ip4_main, sw_if_index1, + 0 /* just want the address */); + rt->cached_sw_if_index = sw_if_index1; + rt->cached_ip4_address = first_int_addr->as_u32; + } + + /* Don't NAT packet aimed at the intfc address */ + if (PREDICT_FALSE(ip1->dst_address.as_u32 == + rt->cached_ip4_address)) + goto trace01; + + next1 = slow_path (sm, b1, ip1, rx_fib_index1, &key1, + &s1, node, next1, cpu_index); + if (PREDICT_FALSE (next1 == SNAT_IN2OUT_NEXT_DROP)) + goto trace01; + } + else + { + next1 = SNAT_IN2OUT_NEXT_SLOW_PATH; + goto trace01; + } + } + else + s1 = pool_elt_at_index (sm->per_thread_data[cpu_index].sessions, + value1.value); + + old_addr1 = ip1->src_address.as_u32; + ip1->src_address = s1->out2in.addr; + new_addr1 = ip1->src_address.as_u32; + vnet_buffer(b1)->sw_if_index[VLIB_TX] = s1->out2in.fib_index; + + sum1 = ip1->checksum; + sum1 = ip_csum_update (sum1, old_addr1, new_addr1, + ip4_header_t, + src_address /* changed member */); + ip1->checksum = ip_csum_fold (sum1); + + if (PREDICT_TRUE(proto1 == SNAT_PROTOCOL_TCP)) + { + old_port1 = tcp1->ports.src; + tcp1->ports.src = s1->out2in.port; + new_port1 = tcp1->ports.src; + + sum1 = tcp1->checksum; + sum1 = ip_csum_update (sum1, old_addr1, new_addr1, + ip4_header_t, + dst_address /* changed member */); + sum1 = ip_csum_update (sum1, old_port1, new_port1, + ip4_header_t /* cheat */, + length /* changed member */); + tcp1->checksum = ip_csum_fold(sum1); + } + else + { + old_port1 = udp1->src_port; + udp1->src_port = s1->out2in.port; + udp1->checksum = 0; + } + + /* Hairpinning */ + snat_hairpinning (sm, b1, ip1, udp1, tcp1, proto1); + + /* Accounting */ + s1->last_heard = now; + s1->total_pkts++; + s1->total_bytes += vlib_buffer_length_in_chain (vm, b1); + /* Per-user LRU list maintenance for dynamic translation */ + if (!snat_is_session_static (s1)) + { + clib_dlist_remove (sm->per_thread_data[cpu_index].list_pool, + s1->per_user_index); + clib_dlist_addtail (sm->per_thread_data[cpu_index].list_pool, + s1->per_user_list_head_index, + s1->per_user_index); + } + trace01: + + if (PREDICT_FALSE((node->flags & VLIB_NODE_FLAG_TRACE) + && (b1->flags & VLIB_BUFFER_IS_TRACED))) + { + snat_in2out_trace_t *t = + vlib_add_trace (vm, node, b1, sizeof (*t)); + t->sw_if_index = sw_if_index1; + t->next_index = next1; + t->session_index = ~0; + if (s1) + t->session_index = s1 - sm->per_thread_data[cpu_index].sessions; + } + + pkts_processed += next1 != SNAT_IN2OUT_NEXT_DROP; + + /* verify speculative enqueues, maybe switch current next frame */ + vlib_validate_buffer_enqueue_x2 (vm, node, next_index, + to_next, n_left_to_next, + bi0, bi1, next0, next1); + } + + while (n_left_from > 0 && n_left_to_next > 0) + { + u32 bi0; + vlib_buffer_t * b0; + u32 next0; + u32 sw_if_index0; + ip4_header_t * ip0; + ip_csum_t sum0; + u32 new_addr0, old_addr0; + u16 old_port0, new_port0; + udp_header_t * udp0; + tcp_header_t * tcp0; + icmp46_header_t * icmp0; + snat_session_key_t key0; + u32 rx_fib_index0; + u32 proto0; + snat_session_t * s0 = 0; + clib_bihash_kv_8_8_t kv0, value0; + + /* speculatively enqueue b0 to the current next frame */ + bi0 = from[0]; + to_next[0] = bi0; + from += 1; + to_next += 1; + n_left_from -= 1; + n_left_to_next -= 1; + + b0 = vlib_get_buffer (vm, bi0); + next0 = SNAT_IN2OUT_NEXT_LOOKUP; + + ip0 = vlib_buffer_get_current (b0); + udp0 = ip4_next_header (ip0); + tcp0 = (tcp_header_t *) udp0; + icmp0 = (icmp46_header_t *) udp0; + + sw_if_index0 = vnet_buffer(b0)->sw_if_index[VLIB_RX]; + rx_fib_index0 = vec_elt (sm->ip4_main->fib_index_by_sw_if_index, + sw_if_index0); + + proto0 = ~0; + proto0 = (ip0->protocol == IP_PROTOCOL_UDP) + ? SNAT_PROTOCOL_UDP : proto0; + proto0 = (ip0->protocol == IP_PROTOCOL_TCP) + ? SNAT_PROTOCOL_TCP : proto0; + proto0 = (ip0->protocol == IP_PROTOCOL_ICMP) + ? SNAT_PROTOCOL_ICMP : proto0; + + /* Next configured feature, probably ip4-lookup */ + if (is_slow_path) + { + if (PREDICT_FALSE (proto0 == ~0)) + goto trace0; + + if (PREDICT_FALSE (proto0 == SNAT_PROTOCOL_ICMP)) + { + next0 = icmp_in2out_slow_path + (sm, b0, ip0, icmp0, sw_if_index0, rx_fib_index0, node, + next0, now, cpu_index); + goto trace0; + } + } + else + { + if (PREDICT_FALSE (proto0 == ~0 || proto0 == SNAT_PROTOCOL_ICMP)) + { + next0 = SNAT_IN2OUT_NEXT_SLOW_PATH; + goto trace0; + } + } + + key0.addr = ip0->src_address; + key0.port = udp0->src_port; + key0.protocol = proto0; + key0.fib_index = rx_fib_index0; + + kv0.key = key0.as_u64; + + if (clib_bihash_search_8_8 (&sm->in2out, &kv0, &value0)) + { + if (is_slow_path) + { + ip4_address_t * first_int_addr; + + if (PREDICT_FALSE(rt->cached_sw_if_index != sw_if_index0)) + { + first_int_addr = + ip4_interface_first_address (sm->ip4_main, sw_if_index0, + 0 /* just want the address */); + rt->cached_sw_if_index = sw_if_index0; + rt->cached_ip4_address = first_int_addr->as_u32; + } + + /* Don't NAT packet aimed at the intfc address */ + if (PREDICT_FALSE(ip0->dst_address.as_u32 == + rt->cached_ip4_address)) + goto trace0; + + next0 = slow_path (sm, b0, ip0, rx_fib_index0, &key0, + &s0, node, next0, cpu_index); + if (PREDICT_FALSE (next0 == SNAT_IN2OUT_NEXT_DROP)) + goto trace0; + } + else + { + next0 = SNAT_IN2OUT_NEXT_SLOW_PATH; + goto trace0; + } + } + else + s0 = pool_elt_at_index (sm->per_thread_data[cpu_index].sessions, + value0.value); + + old_addr0 = ip0->src_address.as_u32; + ip0->src_address = s0->out2in.addr; + new_addr0 = ip0->src_address.as_u32; + vnet_buffer(b0)->sw_if_index[VLIB_TX] = s0->out2in.fib_index; + + sum0 = ip0->checksum; + sum0 = ip_csum_update (sum0, old_addr0, new_addr0, + ip4_header_t, + src_address /* changed member */); + ip0->checksum = ip_csum_fold (sum0); + + if (PREDICT_TRUE(proto0 == SNAT_PROTOCOL_TCP)) + { + old_port0 = tcp0->ports.src; + tcp0->ports.src = s0->out2in.port; + new_port0 = tcp0->ports.src; + + sum0 = tcp0->checksum; + sum0 = ip_csum_update (sum0, old_addr0, new_addr0, + ip4_header_t, + dst_address /* changed member */); + sum0 = ip_csum_update (sum0, old_port0, new_port0, + ip4_header_t /* cheat */, + length /* changed member */); + tcp0->checksum = ip_csum_fold(sum0); + } + else + { + old_port0 = udp0->src_port; + udp0->src_port = s0->out2in.port; + udp0->checksum = 0; + } + + /* Hairpinning */ + snat_hairpinning (sm, b0, ip0, udp0, tcp0, proto0); + + /* Accounting */ + s0->last_heard = now; + s0->total_pkts++; + s0->total_bytes += vlib_buffer_length_in_chain (vm, b0); + /* Per-user LRU list maintenance for dynamic translation */ + if (!snat_is_session_static (s0)) + { + clib_dlist_remove (sm->per_thread_data[cpu_index].list_pool, + s0->per_user_index); + clib_dlist_addtail (sm->per_thread_data[cpu_index].list_pool, + s0->per_user_list_head_index, + s0->per_user_index); + } + + trace0: + if (PREDICT_FALSE((node->flags & VLIB_NODE_FLAG_TRACE) + && (b0->flags & VLIB_BUFFER_IS_TRACED))) + { + snat_in2out_trace_t *t = + vlib_add_trace (vm, node, b0, sizeof (*t)); + t->is_slow_path = is_slow_path; + t->sw_if_index = sw_if_index0; + t->next_index = next0; + t->session_index = ~0; + if (s0) + t->session_index = s0 - sm->per_thread_data[cpu_index].sessions; + } + + pkts_processed += next0 != SNAT_IN2OUT_NEXT_DROP; + + /* verify speculative enqueue, maybe switch current next frame */ + vlib_validate_buffer_enqueue_x1 (vm, node, next_index, + to_next, n_left_to_next, + bi0, next0); + } + + vlib_put_next_frame (vm, node, next_index, n_left_to_next); + } + + vlib_node_increment_counter (vm, stats_node_index, + SNAT_IN2OUT_ERROR_IN2OUT_PACKETS, + pkts_processed); + return frame->n_vectors; +} + +static uword +snat_in2out_fast_path_fn (vlib_main_t * vm, + vlib_node_runtime_t * node, + vlib_frame_t * frame) +{ + return snat_in2out_node_fn_inline (vm, node, frame, 0 /* is_slow_path */); +} + +VLIB_REGISTER_NODE (snat_in2out_node) = { + .function = snat_in2out_fast_path_fn, + .name = "snat-in2out", + .vector_size = sizeof (u32), + .format_trace = format_snat_in2out_trace, + .type = VLIB_NODE_TYPE_INTERNAL, + + .n_errors = ARRAY_LEN(snat_in2out_error_strings), + .error_strings = snat_in2out_error_strings, + + .runtime_data_bytes = sizeof (snat_runtime_t), + + .n_next_nodes = SNAT_IN2OUT_N_NEXT, + + /* edit / add dispositions here */ + .next_nodes = { + [SNAT_IN2OUT_NEXT_DROP] = "error-drop", + [SNAT_IN2OUT_NEXT_LOOKUP] = "ip4-lookup", + [SNAT_IN2OUT_NEXT_SLOW_PATH] = "snat-in2out-slowpath", + }, +}; + +VLIB_NODE_FUNCTION_MULTIARCH (snat_in2out_node, snat_in2out_fast_path_fn); + +static uword +snat_in2out_slow_path_fn (vlib_main_t * vm, + vlib_node_runtime_t * node, + vlib_frame_t * frame) +{ + return snat_in2out_node_fn_inline (vm, node, frame, 1 /* is_slow_path */); +} + +VLIB_REGISTER_NODE (snat_in2out_slowpath_node) = { + .function = snat_in2out_slow_path_fn, + .name = "snat-in2out-slowpath", + .vector_size = sizeof (u32), + .format_trace = format_snat_in2out_trace, + .type = VLIB_NODE_TYPE_INTERNAL, + + .n_errors = ARRAY_LEN(snat_in2out_error_strings), + .error_strings = snat_in2out_error_strings, + + .runtime_data_bytes = sizeof (snat_runtime_t), + + .n_next_nodes = SNAT_IN2OUT_N_NEXT, + + /* edit / add dispositions here */ + .next_nodes = { + [SNAT_IN2OUT_NEXT_DROP] = "error-drop", + [SNAT_IN2OUT_NEXT_LOOKUP] = "ip4-lookup", + [SNAT_IN2OUT_NEXT_SLOW_PATH] = "snat-in2out-slowpath", + }, +}; + +VLIB_NODE_FUNCTION_MULTIARCH (snat_in2out_slowpath_node, snat_in2out_slow_path_fn); + +static uword +snat_in2out_worker_handoff_fn (vlib_main_t * vm, + vlib_node_runtime_t * node, + vlib_frame_t * frame) +{ + snat_main_t *sm = &snat_main; + vlib_thread_main_t *tm = vlib_get_thread_main (); + u32 n_left_from, *from, *to_next = 0; + static __thread vlib_frame_queue_elt_t **handoff_queue_elt_by_worker_index; + static __thread vlib_frame_queue_t **congested_handoff_queue_by_worker_index + = 0; + vlib_frame_queue_elt_t *hf = 0; + vlib_frame_t *f = 0; + int i; + u32 n_left_to_next_worker = 0, *to_next_worker = 0; + u32 next_worker_index = 0; + u32 current_worker_index = ~0; + u32 cpu_index = os_get_cpu_number (); + + ASSERT (vec_len (sm->workers)); + + if (PREDICT_FALSE (handoff_queue_elt_by_worker_index == 0)) + { + vec_validate (handoff_queue_elt_by_worker_index, tm->n_vlib_mains - 1); + + vec_validate_init_empty (congested_handoff_queue_by_worker_index, + sm->first_worker_index + sm->num_workers - 1, + (vlib_frame_queue_t *) (~0)); + } + + from = vlib_frame_vector_args (frame); + n_left_from = frame->n_vectors; + + while (n_left_from > 0) + { + u32 bi0; + vlib_buffer_t *b0; + u32 sw_if_index0; + u32 rx_fib_index0; + ip4_header_t * ip0; + snat_user_key_t key0; + clib_bihash_kv_8_8_t kv0, value0; + u8 do_handoff; + + bi0 = from[0]; + from += 1; + n_left_from -= 1; + + b0 = vlib_get_buffer (vm, bi0); + + sw_if_index0 = vnet_buffer (b0)->sw_if_index[VLIB_RX]; + rx_fib_index0 = ip4_fib_table_get_index_for_sw_if_index(sw_if_index0); + + ip0 = vlib_buffer_get_current (b0); + + key0.addr = ip0->src_address; + key0.fib_index = rx_fib_index0; + + kv0.key = key0.as_u64; + + /* Ever heard of of the "user" before? */ + if (clib_bihash_search_8_8 (&sm->worker_by_in, &kv0, &value0)) + { + /* No, assign next available worker (RR) */ + next_worker_index = sm->first_worker_index + + sm->workers[sm->next_worker++ % vec_len (sm->workers)]; + + /* add non-traslated packets worker lookup */ + kv0.value = next_worker_index; + clib_bihash_add_del_8_8 (&sm->worker_by_in, &kv0, 1); + } + else + next_worker_index = value0.value; + + if (PREDICT_FALSE (next_worker_index != cpu_index)) + { + do_handoff = 1; + + if (next_worker_index != current_worker_index) + { + if (hf) + hf->n_vectors = VLIB_FRAME_SIZE - n_left_to_next_worker; + + hf = vlib_get_worker_handoff_queue_elt (sm->fq_in2out_index, + next_worker_index, + handoff_queue_elt_by_worker_index); + + n_left_to_next_worker = VLIB_FRAME_SIZE - hf->n_vectors; + to_next_worker = &hf->buffer_index[hf->n_vectors]; + current_worker_index = next_worker_index; + } + + /* enqueue to correct worker thread */ + to_next_worker[0] = bi0; + to_next_worker++; + n_left_to_next_worker--; + + if (n_left_to_next_worker == 0) + { + hf->n_vectors = VLIB_FRAME_SIZE; + vlib_put_frame_queue_elt (hf); + current_worker_index = ~0; + handoff_queue_elt_by_worker_index[next_worker_index] = 0; + hf = 0; + } + } + else + { + do_handoff = 0; + /* if this is 1st frame */ + if (!f) + { + f = vlib_get_frame_to_node (vm, snat_in2out_node.index); + to_next = vlib_frame_vector_args (f); + } + + to_next[0] = bi0; + to_next += 1; + f->n_vectors++; + } + + if (PREDICT_FALSE ((node->flags & VLIB_NODE_FLAG_TRACE) + && (b0->flags & VLIB_BUFFER_IS_TRACED))) + { + snat_in2out_worker_handoff_trace_t *t = + vlib_add_trace (vm, node, b0, sizeof (*t)); + t->next_worker_index = next_worker_index; + t->do_handoff = do_handoff; + } + } + + if (f) + vlib_put_frame_to_node (vm, snat_in2out_node.index, f); + + if (hf) + hf->n_vectors = VLIB_FRAME_SIZE - n_left_to_next_worker; + + /* Ship frames to the worker nodes */ + for (i = 0; i < vec_len (handoff_queue_elt_by_worker_index); i++) + { + if (handoff_queue_elt_by_worker_index[i]) + { + hf = handoff_queue_elt_by_worker_index[i]; + /* + * It works better to let the handoff node + * rate-adapt, always ship the handoff queue element. + */ + if (1 || hf->n_vectors == hf->last_n_vectors) + { + vlib_put_frame_queue_elt (hf); + handoff_queue_elt_by_worker_index[i] = 0; + } + else + hf->last_n_vectors = hf->n_vectors; + } + congested_handoff_queue_by_worker_index[i] = + (vlib_frame_queue_t *) (~0); + } + hf = 0; + current_worker_index = ~0; + return frame->n_vectors; +} + +VLIB_REGISTER_NODE (snat_in2out_worker_handoff_node) = { + .function = snat_in2out_worker_handoff_fn, + .name = "snat-in2out-worker-handoff", + .vector_size = sizeof (u32), + .format_trace = format_snat_in2out_worker_handoff_trace, + .type = VLIB_NODE_TYPE_INTERNAL, + + .n_next_nodes = 1, + + .next_nodes = { + [0] = "error-drop", + }, +}; + +VLIB_NODE_FUNCTION_MULTIARCH (snat_in2out_worker_handoff_node, snat_in2out_worker_handoff_fn); + +static inline u32 icmp_in2out_static_map (snat_main_t *sm, + vlib_buffer_t * b0, + ip4_header_t * ip0, + icmp46_header_t * icmp0, + u32 sw_if_index0, + vlib_node_runtime_t * node, + u32 next0, + u32 rx_fib_index0) +{ + snat_session_key_t key0, sm0; + icmp_echo_header_t *echo0; + u32 new_addr0, old_addr0; + u16 old_id0, new_id0; + ip_csum_t sum0; + snat_runtime_t * rt = (snat_runtime_t *)node->runtime_data; + + echo0 = (icmp_echo_header_t *)(icmp0+1); + + key0.addr = ip0->src_address; + key0.port = echo0->identifier; + key0.fib_index = rx_fib_index0; + + if (snat_static_mapping_match(sm, key0, &sm0, 0)) + { + ip4_address_t * first_int_addr; + + if (PREDICT_FALSE(rt->cached_sw_if_index != sw_if_index0)) + { + first_int_addr = + ip4_interface_first_address (sm->ip4_main, sw_if_index0, + 0 /* just want the address */); + rt->cached_sw_if_index = sw_if_index0; + rt->cached_ip4_address = first_int_addr->as_u32; + } + + /* Don't NAT packet aimed at the intfc address */ + if (PREDICT_FALSE(ip0->dst_address.as_u32 == + rt->cached_ip4_address)) + return next0; + + b0->error = node->errors[SNAT_IN2OUT_ERROR_NO_TRANSLATION]; + return SNAT_IN2OUT_NEXT_DROP; + } + + new_addr0 = sm0.addr.as_u32; + new_id0 = sm0.port; + vnet_buffer(b0)->sw_if_index[VLIB_TX] = sm0.fib_index; + old_addr0 = ip0->src_address.as_u32; + ip0->src_address.as_u32 = new_addr0; + + sum0 = ip0->checksum; + sum0 = ip_csum_update (sum0, old_addr0, new_addr0, + ip4_header_t, + src_address /* changed member */); + ip0->checksum = ip_csum_fold (sum0); + + if (PREDICT_FALSE(new_id0 != echo0->identifier)) + { + old_id0 = echo0->identifier; + echo0->identifier = new_id0; + + sum0 = icmp0->checksum; + sum0 = ip_csum_update (sum0, old_id0, new_id0, icmp_echo_header_t, + identifier); + icmp0->checksum = ip_csum_fold (sum0); + } + + return next0; +} + +static uword +snat_in2out_fast_static_map_fn (vlib_main_t * vm, + vlib_node_runtime_t * node, + vlib_frame_t * frame) +{ + u32 n_left_from, * from, * to_next; + snat_in2out_next_t next_index; + u32 pkts_processed = 0; + snat_main_t * sm = &snat_main; + snat_runtime_t * rt = (snat_runtime_t *)node->runtime_data; + u32 stats_node_index; + + stats_node_index = snat_in2out_fast_node.index; + + from = vlib_frame_vector_args (frame); + n_left_from = frame->n_vectors; + next_index = node->cached_next_index; + + while (n_left_from > 0) + { + u32 n_left_to_next; + + vlib_get_next_frame (vm, node, next_index, + to_next, n_left_to_next); + + while (n_left_from > 0 && n_left_to_next > 0) + { + u32 bi0; + vlib_buffer_t * b0; + u32 next0; + u32 sw_if_index0; + ip4_header_t * ip0; + ip_csum_t sum0; + u32 new_addr0, old_addr0; + u16 old_port0, new_port0; + udp_header_t * udp0; + tcp_header_t * tcp0; + icmp46_header_t * icmp0; + snat_session_key_t key0, sm0; + u32 proto0; + u32 rx_fib_index0; + + /* speculatively enqueue b0 to the current next frame */ + bi0 = from[0]; + to_next[0] = bi0; + from += 1; + to_next += 1; + n_left_from -= 1; + n_left_to_next -= 1; + + b0 = vlib_get_buffer (vm, bi0); + next0 = SNAT_IN2OUT_NEXT_LOOKUP; + + ip0 = vlib_buffer_get_current (b0); + udp0 = ip4_next_header (ip0); + tcp0 = (tcp_header_t *) udp0; + icmp0 = (icmp46_header_t *) udp0; + + sw_if_index0 = vnet_buffer(b0)->sw_if_index[VLIB_RX]; + rx_fib_index0 = ip4_fib_table_get_index_for_sw_if_index(sw_if_index0); + + proto0 = ~0; + proto0 = (ip0->protocol == IP_PROTOCOL_UDP) + ? SNAT_PROTOCOL_UDP : proto0; + proto0 = (ip0->protocol == IP_PROTOCOL_TCP) + ? SNAT_PROTOCOL_TCP : proto0; + proto0 = (ip0->protocol == IP_PROTOCOL_ICMP) + ? SNAT_PROTOCOL_ICMP : proto0; + + if (PREDICT_FALSE (proto0 == ~0)) + goto trace0; + + if (PREDICT_FALSE (proto0 == SNAT_PROTOCOL_ICMP)) + { + ip4_address_t * first_int_addr; + + if (PREDICT_FALSE(rt->cached_sw_if_index != sw_if_index0)) + { + first_int_addr = + ip4_interface_first_address (sm->ip4_main, sw_if_index0, + 0 /* just want the address */); + rt->cached_sw_if_index = sw_if_index0; + rt->cached_ip4_address = first_int_addr->as_u32; + } + + /* Don't NAT packet aimed at the intfc address */ + if (PREDICT_FALSE(ip0->dst_address.as_u32 == + rt->cached_ip4_address)) + goto trace0; + + next0 = icmp_in2out_static_map + (sm, b0, ip0, icmp0, sw_if_index0, node, next0, rx_fib_index0); + goto trace0; + } + + key0.addr = ip0->src_address; + key0.port = udp0->src_port; + key0.fib_index = rx_fib_index0; + + if (snat_static_mapping_match(sm, key0, &sm0, 0)) + { + b0->error = node->errors[SNAT_IN2OUT_ERROR_NO_TRANSLATION]; + next0= SNAT_IN2OUT_NEXT_DROP; + goto trace0; + } + + new_addr0 = sm0.addr.as_u32; + new_port0 = sm0.port; + vnet_buffer(b0)->sw_if_index[VLIB_TX] = sm0.fib_index; + old_addr0 = ip0->src_address.as_u32; + ip0->src_address.as_u32 = new_addr0; + + sum0 = ip0->checksum; + sum0 = ip_csum_update (sum0, old_addr0, new_addr0, + ip4_header_t, + src_address /* changed member */); + ip0->checksum = ip_csum_fold (sum0); + + if (PREDICT_FALSE(new_port0 != udp0->dst_port)) + { + if (PREDICT_TRUE(proto0 == SNAT_PROTOCOL_TCP)) + { + old_port0 = tcp0->ports.src; + tcp0->ports.src = new_port0; + + sum0 = tcp0->checksum; + sum0 = ip_csum_update (sum0, old_addr0, new_addr0, + ip4_header_t, + dst_address /* changed member */); + sum0 = ip_csum_update (sum0, old_port0, new_port0, + ip4_header_t /* cheat */, + length /* changed member */); + tcp0->checksum = ip_csum_fold(sum0); + } + else + { + old_port0 = udp0->src_port; + udp0->src_port = new_port0; + udp0->checksum = 0; + } + } + else + { + if (PREDICT_TRUE(proto0 == SNAT_PROTOCOL_TCP)) + { + sum0 = tcp0->checksum; + sum0 = ip_csum_update (sum0, old_addr0, new_addr0, + ip4_header_t, + dst_address /* changed member */); + tcp0->checksum = ip_csum_fold(sum0); + } + } + + /* Hairpinning */ + snat_hairpinning (sm, b0, ip0, udp0, tcp0, proto0); + + trace0: + if (PREDICT_FALSE((node->flags & VLIB_NODE_FLAG_TRACE) + && (b0->flags & VLIB_BUFFER_IS_TRACED))) + { + snat_in2out_trace_t *t = + vlib_add_trace (vm, node, b0, sizeof (*t)); + t->sw_if_index = sw_if_index0; + t->next_index = next0; + } + + pkts_processed += next0 != SNAT_IN2OUT_NEXT_DROP; + + /* verify speculative enqueue, maybe switch current next frame */ + vlib_validate_buffer_enqueue_x1 (vm, node, next_index, + to_next, n_left_to_next, + bi0, next0); + } + + vlib_put_next_frame (vm, node, next_index, n_left_to_next); + } + + vlib_node_increment_counter (vm, stats_node_index, + SNAT_IN2OUT_ERROR_IN2OUT_PACKETS, + pkts_processed); + return frame->n_vectors; +} + + +VLIB_REGISTER_NODE (snat_in2out_fast_node) = { + .function = snat_in2out_fast_static_map_fn, + .name = "snat-in2out-fast", + .vector_size = sizeof (u32), + .format_trace = format_snat_in2out_fast_trace, + .type = VLIB_NODE_TYPE_INTERNAL, + + .n_errors = ARRAY_LEN(snat_in2out_error_strings), + .error_strings = snat_in2out_error_strings, + + .runtime_data_bytes = sizeof (snat_runtime_t), + + .n_next_nodes = SNAT_IN2OUT_N_NEXT, + + /* edit / add dispositions here */ + .next_nodes = { + [SNAT_IN2OUT_NEXT_DROP] = "error-drop", + [SNAT_IN2OUT_NEXT_LOOKUP] = "ip4-lookup", + [SNAT_IN2OUT_NEXT_SLOW_PATH] = "snat-in2out-slowpath", + }, +}; + +VLIB_NODE_FUNCTION_MULTIARCH (snat_in2out_fast_node, snat_in2out_fast_static_map_fn); diff --git a/src/plugins/snat/out2in.c b/src/plugins/snat/out2in.c new file mode 100644 index 00000000..f1f4159c --- /dev/null +++ b/src/plugins/snat/out2in.c @@ -0,0 +1,1261 @@ +/* + * Copyright (c) 2016 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +#include +#include +#include +#include + +#include +#include +#include +#include + +#include +#include +#include + +typedef struct { + u32 sw_if_index; + u32 next_index; + u32 session_index; +} snat_out2in_trace_t; + +typedef struct { + u32 next_worker_index; + u8 do_handoff; +} snat_out2in_worker_handoff_trace_t; + +/* packet trace format function */ +static u8 * format_snat_out2in_trace (u8 * s, va_list * args) +{ + CLIB_UNUSED (vlib_main_t * vm) = va_arg (*args, vlib_main_t *); + CLIB_UNUSED (vlib_node_t * node) = va_arg (*args, vlib_node_t *); + snat_out2in_trace_t * t = va_arg (*args, snat_out2in_trace_t *); + + s = format (s, "SNAT_OUT2IN: sw_if_index %d, next index %d, session index %d", + t->sw_if_index, t->next_index, t->session_index); + return s; +} + +static u8 * format_snat_out2in_fast_trace (u8 * s, va_list * args) +{ + CLIB_UNUSED (vlib_main_t * vm) = va_arg (*args, vlib_main_t *); + CLIB_UNUSED (vlib_node_t * node) = va_arg (*args, vlib_node_t *); + snat_out2in_trace_t * t = va_arg (*args, snat_out2in_trace_t *); + + s = format (s, "SNAT_OUT2IN_FAST: sw_if_index %d, next index %d", + t->sw_if_index, t->next_index); + return s; +} + +static u8 * format_snat_out2in_worker_handoff_trace (u8 * s, va_list * args) +{ + CLIB_UNUSED (vlib_main_t * vm) = va_arg (*args, vlib_main_t *); + CLIB_UNUSED (vlib_node_t * node) = va_arg (*args, vlib_node_t *); + snat_out2in_worker_handoff_trace_t * t = + va_arg (*args, snat_out2in_worker_handoff_trace_t *); + char * m; + + m = t->do_handoff ? "next worker" : "same worker"; + s = format (s, "SNAT_OUT2IN_WORKER_HANDOFF: %s %d", m, t->next_worker_index); + + return s; +} + +vlib_node_registration_t snat_out2in_node; +vlib_node_registration_t snat_out2in_fast_node; +vlib_node_registration_t snat_out2in_worker_handoff_node; + +#define foreach_snat_out2in_error \ +_(UNSUPPORTED_PROTOCOL, "Unsupported protocol") \ +_(OUT2IN_PACKETS, "Good out2in packets processed") \ +_(BAD_ICMP_TYPE, "icmp type not echo-reply") \ +_(NO_TRANSLATION, "No translation") + +typedef enum { +#define _(sym,str) SNAT_OUT2IN_ERROR_##sym, + foreach_snat_out2in_error +#undef _ + SNAT_OUT2IN_N_ERROR, +} snat_out2in_error_t; + +static char * snat_out2in_error_strings[] = { +#define _(sym,string) string, + foreach_snat_out2in_error +#undef _ +}; + +typedef enum { + SNAT_OUT2IN_NEXT_DROP, + SNAT_OUT2IN_NEXT_LOOKUP, + SNAT_OUT2IN_N_NEXT, +} snat_out2in_next_t; + +/** + * @brief Create session for static mapping. + * + * Create NAT session initiated by host from external network with static + * mapping. + * + * @param sm SNAT main. + * @param b0 Vlib buffer. + * @param in2out In2out SNAT session key. + * @param out2in Out2in SNAT session key. + * @param node Vlib node. + * + * @returns SNAT session if successfully created otherwise 0. + */ +static inline snat_session_t * +create_session_for_static_mapping (snat_main_t *sm, + vlib_buffer_t *b0, + snat_session_key_t in2out, + snat_session_key_t out2in, + vlib_node_runtime_t * node, + u32 cpu_index) +{ + snat_user_t *u; + snat_user_key_t user_key; + snat_session_t *s; + clib_bihash_kv_8_8_t kv0, value0; + dlist_elt_t * per_user_translation_list_elt; + dlist_elt_t * per_user_list_head_elt; + + user_key.addr = in2out.addr; + user_key.fib_index = in2out.fib_index; + kv0.key = user_key.as_u64; + + /* Ever heard of the "user" = inside ip4 address before? */ + if (clib_bihash_search_8_8 (&sm->user_hash, &kv0, &value0)) + { + /* no, make a new one */ + pool_get (sm->per_thread_data[cpu_index].users, u); + memset (u, 0, sizeof (*u)); + u->addr = in2out.addr; + + pool_get (sm->per_thread_data[cpu_index].list_pool, + per_user_list_head_elt); + + u->sessions_per_user_list_head_index = per_user_list_head_elt - + sm->per_thread_data[cpu_index].list_pool; + + clib_dlist_init (sm->per_thread_data[cpu_index].list_pool, + u->sessions_per_user_list_head_index); + + kv0.value = u - sm->per_thread_data[cpu_index].users; + + /* add user */ + clib_bihash_add_del_8_8 (&sm->user_hash, &kv0, 1 /* is_add */); + + /* add non-traslated packets worker lookup */ + kv0.value = cpu_index; + clib_bihash_add_del_8_8 (&sm->worker_by_in, &kv0, 1); + } + else + { + u = pool_elt_at_index (sm->per_thread_data[cpu_index].users, + value0.value); + } + + pool_get (sm->per_thread_data[cpu_index].sessions, s); + memset (s, 0, sizeof (*s)); + + s->outside_address_index = ~0; + s->flags |= SNAT_SESSION_FLAG_STATIC_MAPPING; + u->nstaticsessions++; + + /* Create list elts */ + pool_get (sm->per_thread_data[cpu_index].list_pool, + per_user_translation_list_elt); + clib_dlist_init (sm->per_thread_data[cpu_index].list_pool, + per_user_translation_list_elt - + sm->per_thread_data[cpu_index].list_pool); + + per_user_translation_list_elt->value = + s - sm->per_thread_data[cpu_index].sessions; + s->per_user_index = + per_user_translation_list_elt - sm->per_thread_data[cpu_index].list_pool; + s->per_user_list_head_index = u->sessions_per_user_list_head_index; + + clib_dlist_addtail (sm->per_thread_data[cpu_index].list_pool, + s->per_user_list_head_index, + per_user_translation_list_elt - + sm->per_thread_data[cpu_index].list_pool); + + s->in2out = in2out; + s->out2in = out2in; + s->in2out.protocol = out2in.protocol; + + /* Add to translation hashes */ + kv0.key = s->in2out.as_u64; + kv0.value = s - sm->per_thread_data[cpu_index].sessions; + if (clib_bihash_add_del_8_8 (&sm->in2out, &kv0, 1 /* is_add */)) + clib_warning ("in2out key add failed"); + + kv0.key = s->out2in.as_u64; + kv0.value = s - sm->per_thread_data[cpu_index].sessions; + + if (clib_bihash_add_del_8_8 (&sm->out2in, &kv0, 1 /* is_add */)) + clib_warning ("out2in key add failed"); + + return s; +} + +static inline u32 icmp_out2in_slow_path (snat_main_t *sm, + vlib_buffer_t * b0, + ip4_header_t * ip0, + icmp46_header_t * icmp0, + u32 sw_if_index0, + u32 rx_fib_index0, + vlib_node_runtime_t * node, + u32 next0, f64 now, + u32 cpu_index) +{ + snat_session_key_t key0, sm0; + icmp_echo_header_t *echo0; + clib_bihash_kv_8_8_t kv0, value0; + snat_session_t * s0; + u32 new_addr0, old_addr0; + u16 old_id0, new_id0; + ip_csum_t sum0; + snat_runtime_t * rt = (snat_runtime_t *)node->runtime_data; + + echo0 = (icmp_echo_header_t *)(icmp0+1); + + key0.addr = ip0->dst_address; + key0.port = echo0->identifier; + key0.protocol = SNAT_PROTOCOL_ICMP; + key0.fib_index = rx_fib_index0; + + kv0.key = key0.as_u64; + + if (clib_bihash_search_8_8 (&sm->out2in, &kv0, &value0)) + { + /* Try to match static mapping by external address and port, + destination address and port in packet */ + if (snat_static_mapping_match(sm, key0, &sm0, 1)) + { + ip4_address_t * first_int_addr; + + if (PREDICT_FALSE(rt->cached_sw_if_index != sw_if_index0)) + { + first_int_addr = + ip4_interface_first_address (sm->ip4_main, sw_if_index0, + 0 /* just want the address */); + rt->cached_sw_if_index = sw_if_index0; + rt->cached_ip4_address = first_int_addr->as_u32; + } + + /* Don't NAT packet aimed at the intfc address */ + if (PREDICT_FALSE(ip0->dst_address.as_u32 == + rt->cached_ip4_address)) + return next0; + + b0->error = node->errors[SNAT_OUT2IN_ERROR_NO_TRANSLATION]; + return SNAT_OUT2IN_NEXT_DROP; + } + + /* Create session initiated by host from external network */ + s0 = create_session_for_static_mapping(sm, b0, sm0, key0, + node, cpu_index); + if (!s0) + return SNAT_OUT2IN_NEXT_DROP; + } + else + s0 = pool_elt_at_index (sm->per_thread_data[cpu_index].sessions, + value0.value); + + old_addr0 = ip0->dst_address.as_u32; + ip0->dst_address = s0->in2out.addr; + new_addr0 = ip0->dst_address.as_u32; + vnet_buffer(b0)->sw_if_index[VLIB_TX] = s0->in2out.fib_index; + + sum0 = ip0->checksum; + sum0 = ip_csum_update (sum0, old_addr0, new_addr0, + ip4_header_t, + dst_address /* changed member */); + ip0->checksum = ip_csum_fold (sum0); + + old_id0 = echo0->identifier; + new_id0 = s0->in2out.port; + echo0->identifier = new_id0; + + sum0 = icmp0->checksum; + sum0 = ip_csum_update (sum0, old_id0, new_id0, icmp_echo_header_t, + identifier); + icmp0->checksum = ip_csum_fold (sum0); + + /* Accounting */ + s0->last_heard = now; + s0->total_pkts++; + s0->total_bytes += vlib_buffer_length_in_chain (sm->vlib_main, b0); + /* Per-user LRU list maintenance for dynamic translation */ + if (!snat_is_session_static (s0)) + { + clib_dlist_remove (sm->per_thread_data[cpu_index].list_pool, + s0->per_user_index); + clib_dlist_addtail (sm->per_thread_data[cpu_index].list_pool, + s0->per_user_list_head_index, + s0->per_user_index); + } + + return next0; +} + +static uword +snat_out2in_node_fn (vlib_main_t * vm, + vlib_node_runtime_t * node, + vlib_frame_t * frame) +{ + u32 n_left_from, * from, * to_next; + snat_out2in_next_t next_index; + u32 pkts_processed = 0; + snat_main_t * sm = &snat_main; + f64 now = vlib_time_now (vm); + u32 cpu_index = os_get_cpu_number (); + + from = vlib_frame_vector_args (frame); + n_left_from = frame->n_vectors; + next_index = node->cached_next_index; + + while (n_left_from > 0) + { + u32 n_left_to_next; + + vlib_get_next_frame (vm, node, next_index, + to_next, n_left_to_next); + + while (n_left_from >= 4 && n_left_to_next >= 2) + { + u32 bi0, bi1; + vlib_buffer_t * b0, * b1; + u32 next0 = SNAT_OUT2IN_NEXT_LOOKUP; + u32 next1 = SNAT_OUT2IN_NEXT_LOOKUP; + u32 sw_if_index0, sw_if_index1; + ip4_header_t * ip0, *ip1; + ip_csum_t sum0, sum1; + u32 new_addr0, old_addr0; + u16 new_port0, old_port0; + u32 new_addr1, old_addr1; + u16 new_port1, old_port1; + udp_header_t * udp0, * udp1; + tcp_header_t * tcp0, * tcp1; + icmp46_header_t * icmp0, * icmp1; + snat_session_key_t key0, key1, sm0, sm1; + u32 rx_fib_index0, rx_fib_index1; + u32 proto0, proto1; + snat_session_t * s0 = 0, * s1 = 0; + clib_bihash_kv_8_8_t kv0, kv1, value0, value1; + + /* Prefetch next iteration. */ + { + vlib_buffer_t * p2, * p3; + + p2 = vlib_get_buffer (vm, from[2]); + p3 = vlib_get_buffer (vm, from[3]); + + vlib_prefetch_buffer_header (p2, LOAD); + vlib_prefetch_buffer_header (p3, LOAD); + + CLIB_PREFETCH (p2->data, CLIB_CACHE_LINE_BYTES, STORE); + CLIB_PREFETCH (p3->data, CLIB_CACHE_LINE_BYTES, STORE); + } + + /* speculatively enqueue b0 and b1 to the current next frame */ + to_next[0] = bi0 = from[0]; + to_next[1] = bi1 = from[1]; + from += 2; + to_next += 2; + n_left_from -= 2; + n_left_to_next -= 2; + + b0 = vlib_get_buffer (vm, bi0); + b1 = vlib_get_buffer (vm, bi1); + + ip0 = vlib_buffer_get_current (b0); + udp0 = ip4_next_header (ip0); + tcp0 = (tcp_header_t *) udp0; + icmp0 = (icmp46_header_t *) udp0; + + sw_if_index0 = vnet_buffer(b0)->sw_if_index[VLIB_RX]; + rx_fib_index0 = vec_elt (sm->ip4_main->fib_index_by_sw_if_index, + sw_if_index0); + + proto0 = ~0; + proto0 = (ip0->protocol == IP_PROTOCOL_UDP) + ? SNAT_PROTOCOL_UDP : proto0; + proto0 = (ip0->protocol == IP_PROTOCOL_TCP) + ? SNAT_PROTOCOL_TCP : proto0; + proto0 = (ip0->protocol == IP_PROTOCOL_ICMP) + ? SNAT_PROTOCOL_ICMP : proto0; + + if (PREDICT_FALSE (proto0 == ~0)) + goto trace0; + + if (PREDICT_FALSE (proto0 == SNAT_PROTOCOL_ICMP)) + { + next0 = icmp_out2in_slow_path + (sm, b0, ip0, icmp0, sw_if_index0, rx_fib_index0, node, + next0, now, cpu_index); + goto trace0; + } + + key0.addr = ip0->dst_address; + key0.port = udp0->dst_port; + key0.protocol = proto0; + key0.fib_index = rx_fib_index0; + + kv0.key = key0.as_u64; + + if (clib_bihash_search_8_8 (&sm->out2in, &kv0, &value0)) + { + /* Try to match static mapping by external address and port, + destination address and port in packet */ + if (snat_static_mapping_match(sm, key0, &sm0, 1)) + { + b0->error = node->errors[SNAT_OUT2IN_ERROR_NO_TRANSLATION]; + goto trace0; + } + + /* Create session initiated by host from external network */ + s0 = create_session_for_static_mapping(sm, b0, sm0, key0, node, + cpu_index); + if (!s0) + goto trace0; + } + else + s0 = pool_elt_at_index (sm->per_thread_data[cpu_index].sessions, + value0.value); + + old_addr0 = ip0->dst_address.as_u32; + ip0->dst_address = s0->in2out.addr; + new_addr0 = ip0->dst_address.as_u32; + vnet_buffer(b0)->sw_if_index[VLIB_TX] = s0->in2out.fib_index; + + sum0 = ip0->checksum; + sum0 = ip_csum_update (sum0, old_addr0, new_addr0, + ip4_header_t, + dst_address /* changed member */); + ip0->checksum = ip_csum_fold (sum0); + + if (PREDICT_TRUE(proto0 == SNAT_PROTOCOL_TCP)) + { + old_port0 = tcp0->ports.dst; + tcp0->ports.dst = s0->in2out.port; + new_port0 = tcp0->ports.dst; + + sum0 = tcp0->checksum; + sum0 = ip_csum_update (sum0, old_addr0, new_addr0, + ip4_header_t, + dst_address /* changed member */); + + sum0 = ip_csum_update (sum0, old_port0, new_port0, + ip4_header_t /* cheat */, + length /* changed member */); + tcp0->checksum = ip_csum_fold(sum0); + } + else + { + old_port0 = udp0->dst_port; + udp0->dst_port = s0->in2out.port; + udp0->checksum = 0; + } + + /* Accounting */ + s0->last_heard = now; + s0->total_pkts++; + s0->total_bytes += vlib_buffer_length_in_chain (vm, b0); + /* Per-user LRU list maintenance for dynamic translation */ + if (!snat_is_session_static (s0)) + { + clib_dlist_remove (sm->per_thread_data[cpu_index].list_pool, + s0->per_user_index); + clib_dlist_addtail (sm->per_thread_data[cpu_index].list_pool, + s0->per_user_list_head_index, + s0->per_user_index); + } + trace0: + + if (PREDICT_FALSE((node->flags & VLIB_NODE_FLAG_TRACE) + && (b0->flags & VLIB_BUFFER_IS_TRACED))) + { + snat_out2in_trace_t *t = + vlib_add_trace (vm, node, b0, sizeof (*t)); + t->sw_if_index = sw_if_index0; + t->next_index = next0; + t->session_index = ~0; + if (s0) + t->session_index = s0 - sm->per_thread_data[cpu_index].sessions; + } + + pkts_processed += next0 != SNAT_OUT2IN_NEXT_DROP; + + + ip1 = vlib_buffer_get_current (b1); + udp1 = ip4_next_header (ip1); + tcp1 = (tcp_header_t *) udp1; + icmp1 = (icmp46_header_t *) udp1; + + sw_if_index1 = vnet_buffer(b1)->sw_if_index[VLIB_RX]; + rx_fib_index1 = vec_elt (sm->ip4_main->fib_index_by_sw_if_index, + sw_if_index1); + + proto1 = ~0; + proto1 = (ip1->protocol == IP_PROTOCOL_UDP) + ? SNAT_PROTOCOL_UDP : proto1; + proto1 = (ip1->protocol == IP_PROTOCOL_TCP) + ? SNAT_PROTOCOL_TCP : proto1; + proto1 = (ip1->protocol == IP_PROTOCOL_ICMP) + ? SNAT_PROTOCOL_ICMP : proto1; + + if (PREDICT_FALSE (proto1 == ~0)) + goto trace1; + + if (PREDICT_FALSE (proto1 == SNAT_PROTOCOL_ICMP)) + { + next1 = icmp_out2in_slow_path + (sm, b1, ip1, icmp1, sw_if_index1, rx_fib_index1, node, + next1, now, cpu_index); + goto trace1; + } + + key1.addr = ip1->dst_address; + key1.port = udp1->dst_port; + key1.protocol = proto1; + key1.fib_index = rx_fib_index1; + + kv1.key = key1.as_u64; + + if (clib_bihash_search_8_8 (&sm->out2in, &kv1, &value1)) + { + /* Try to match static mapping by external address and port, + destination address and port in packet */ + if (snat_static_mapping_match(sm, key1, &sm1, 1)) + { + b1->error = node->errors[SNAT_OUT2IN_ERROR_NO_TRANSLATION]; + goto trace1; + } + + /* Create session initiated by host from external network */ + s1 = create_session_for_static_mapping(sm, b1, sm1, key1, node, + cpu_index); + if (!s1) + goto trace1; + } + else + s1 = pool_elt_at_index (sm->per_thread_data[cpu_index].sessions, + value1.value); + + old_addr1 = ip1->dst_address.as_u32; + ip1->dst_address = s1->in2out.addr; + new_addr1 = ip1->dst_address.as_u32; + vnet_buffer(b1)->sw_if_index[VLIB_TX] = s1->in2out.fib_index; + + sum1 = ip1->checksum; + sum1 = ip_csum_update (sum1, old_addr1, new_addr1, + ip4_header_t, + dst_address /* changed member */); + ip1->checksum = ip_csum_fold (sum1); + + if (PREDICT_TRUE(proto1 == SNAT_PROTOCOL_TCP)) + { + old_port1 = tcp1->ports.dst; + tcp1->ports.dst = s1->in2out.port; + new_port1 = tcp1->ports.dst; + + sum1 = tcp1->checksum; + sum1 = ip_csum_update (sum1, old_addr1, new_addr1, + ip4_header_t, + dst_address /* changed member */); + + sum1 = ip_csum_update (sum1, old_port1, new_port1, + ip4_header_t /* cheat */, + length /* changed member */); + tcp1->checksum = ip_csum_fold(sum1); + } + else + { + old_port1 = udp1->dst_port; + udp1->dst_port = s1->in2out.port; + udp1->checksum = 0; + } + + /* Accounting */ + s1->last_heard = now; + s1->total_pkts++; + s1->total_bytes += vlib_buffer_length_in_chain (vm, b1); + /* Per-user LRU list maintenance for dynamic translation */ + if (!snat_is_session_static (s1)) + { + clib_dlist_remove (sm->per_thread_data[cpu_index].list_pool, + s1->per_user_index); + clib_dlist_addtail (sm->per_thread_data[cpu_index].list_pool, + s1->per_user_list_head_index, + s1->per_user_index); + } + trace1: + + if (PREDICT_FALSE((node->flags & VLIB_NODE_FLAG_TRACE) + && (b1->flags & VLIB_BUFFER_IS_TRACED))) + { + snat_out2in_trace_t *t = + vlib_add_trace (vm, node, b1, sizeof (*t)); + t->sw_if_index = sw_if_index1; + t->next_index = next1; + t->session_index = ~0; + if (s1) + t->session_index = s1 - sm->per_thread_data[cpu_index].sessions; + } + + pkts_processed += next1 != SNAT_OUT2IN_NEXT_DROP; + + /* verify speculative enqueues, maybe switch current next frame */ + vlib_validate_buffer_enqueue_x2 (vm, node, next_index, + to_next, n_left_to_next, + bi0, bi1, next0, next1); + } + + while (n_left_from > 0 && n_left_to_next > 0) + { + u32 bi0; + vlib_buffer_t * b0; + u32 next0 = SNAT_OUT2IN_NEXT_LOOKUP; + u32 sw_if_index0; + ip4_header_t * ip0; + ip_csum_t sum0; + u32 new_addr0, old_addr0; + u16 new_port0, old_port0; + udp_header_t * udp0; + tcp_header_t * tcp0; + icmp46_header_t * icmp0; + snat_session_key_t key0, sm0; + u32 rx_fib_index0; + u32 proto0; + snat_session_t * s0 = 0; + clib_bihash_kv_8_8_t kv0, value0; + + /* speculatively enqueue b0 to the current next frame */ + bi0 = from[0]; + to_next[0] = bi0; + from += 1; + to_next += 1; + n_left_from -= 1; + n_left_to_next -= 1; + + b0 = vlib_get_buffer (vm, bi0); + + ip0 = vlib_buffer_get_current (b0); + udp0 = ip4_next_header (ip0); + tcp0 = (tcp_header_t *) udp0; + icmp0 = (icmp46_header_t *) udp0; + + sw_if_index0 = vnet_buffer(b0)->sw_if_index[VLIB_RX]; + rx_fib_index0 = vec_elt (sm->ip4_main->fib_index_by_sw_if_index, + sw_if_index0); + + proto0 = ~0; + proto0 = (ip0->protocol == IP_PROTOCOL_UDP) + ? SNAT_PROTOCOL_UDP : proto0; + proto0 = (ip0->protocol == IP_PROTOCOL_TCP) + ? SNAT_PROTOCOL_TCP : proto0; + proto0 = (ip0->protocol == IP_PROTOCOL_ICMP) + ? SNAT_PROTOCOL_ICMP : proto0; + + if (PREDICT_FALSE (proto0 == ~0)) + goto trace00; + + if (PREDICT_FALSE (proto0 == SNAT_PROTOCOL_ICMP)) + { + next0 = icmp_out2in_slow_path + (sm, b0, ip0, icmp0, sw_if_index0, rx_fib_index0, node, + next0, now, cpu_index); + goto trace00; + } + + key0.addr = ip0->dst_address; + key0.port = udp0->dst_port; + key0.protocol = proto0; + key0.fib_index = rx_fib_index0; + + kv0.key = key0.as_u64; + + if (clib_bihash_search_8_8 (&sm->out2in, &kv0, &value0)) + { + /* Try to match static mapping by external address and port, + destination address and port in packet */ + if (snat_static_mapping_match(sm, key0, &sm0, 1)) + { + b0->error = node->errors[SNAT_OUT2IN_ERROR_NO_TRANSLATION]; + goto trace00; + } + + /* Create session initiated by host from external network */ + s0 = create_session_for_static_mapping(sm, b0, sm0, key0, node, + cpu_index); + if (!s0) + goto trace00; + } + else + s0 = pool_elt_at_index (sm->per_thread_data[cpu_index].sessions, + value0.value); + + old_addr0 = ip0->dst_address.as_u32; + ip0->dst_address = s0->in2out.addr; + new_addr0 = ip0->dst_address.as_u32; + vnet_buffer(b0)->sw_if_index[VLIB_TX] = s0->in2out.fib_index; + + sum0 = ip0->checksum; + sum0 = ip_csum_update (sum0, old_addr0, new_addr0, + ip4_header_t, + dst_address /* changed member */); + ip0->checksum = ip_csum_fold (sum0); + + if (PREDICT_TRUE(proto0 == SNAT_PROTOCOL_TCP)) + { + old_port0 = tcp0->ports.dst; + tcp0->ports.dst = s0->in2out.port; + new_port0 = tcp0->ports.dst; + + sum0 = tcp0->checksum; + sum0 = ip_csum_update (sum0, old_addr0, new_addr0, + ip4_header_t, + dst_address /* changed member */); + + sum0 = ip_csum_update (sum0, old_port0, new_port0, + ip4_header_t /* cheat */, + length /* changed member */); + tcp0->checksum = ip_csum_fold(sum0); + } + else + { + old_port0 = udp0->dst_port; + udp0->dst_port = s0->in2out.port; + udp0->checksum = 0; + } + + /* Accounting */ + s0->last_heard = now; + s0->total_pkts++; + s0->total_bytes += vlib_buffer_length_in_chain (vm, b0); + /* Per-user LRU list maintenance for dynamic translation */ + if (!snat_is_session_static (s0)) + { + clib_dlist_remove (sm->per_thread_data[cpu_index].list_pool, + s0->per_user_index); + clib_dlist_addtail (sm->per_thread_data[cpu_index].list_pool, + s0->per_user_list_head_index, + s0->per_user_index); + } + trace00: + + if (PREDICT_FALSE((node->flags & VLIB_NODE_FLAG_TRACE) + && (b0->flags & VLIB_BUFFER_IS_TRACED))) + { + snat_out2in_trace_t *t = + vlib_add_trace (vm, node, b0, sizeof (*t)); + t->sw_if_index = sw_if_index0; + t->next_index = next0; + t->session_index = ~0; + if (s0) + t->session_index = s0 - sm->per_thread_data[cpu_index].sessions; + } + + pkts_processed += next0 != SNAT_OUT2IN_NEXT_DROP; + + /* verify speculative enqueue, maybe switch current next frame */ + vlib_validate_buffer_enqueue_x1 (vm, node, next_index, + to_next, n_left_to_next, + bi0, next0); + } + + vlib_put_next_frame (vm, node, next_index, n_left_to_next); + } + + vlib_node_increment_counter (vm, snat_out2in_node.index, + SNAT_OUT2IN_ERROR_OUT2IN_PACKETS, + pkts_processed); + return frame->n_vectors; +} + +VLIB_REGISTER_NODE (snat_out2in_node) = { + .function = snat_out2in_node_fn, + .name = "snat-out2in", + .vector_size = sizeof (u32), + .format_trace = format_snat_out2in_trace, + .type = VLIB_NODE_TYPE_INTERNAL, + + .n_errors = ARRAY_LEN(snat_out2in_error_strings), + .error_strings = snat_out2in_error_strings, + + .runtime_data_bytes = sizeof (snat_runtime_t), + + .n_next_nodes = SNAT_OUT2IN_N_NEXT, + + /* edit / add dispositions here */ + .next_nodes = { + [SNAT_OUT2IN_NEXT_DROP] = "error-drop", + [SNAT_OUT2IN_NEXT_LOOKUP] = "ip4-lookup", + }, +}; +VLIB_NODE_FUNCTION_MULTIARCH (snat_out2in_node, snat_out2in_node_fn); + +static uword +snat_out2in_worker_handoff_fn (vlib_main_t * vm, + vlib_node_runtime_t * node, + vlib_frame_t * frame) +{ + snat_main_t *sm = &snat_main; + vlib_thread_main_t *tm = vlib_get_thread_main (); + u32 n_left_from, *from, *to_next = 0; + static __thread vlib_frame_queue_elt_t **handoff_queue_elt_by_worker_index; + static __thread vlib_frame_queue_t **congested_handoff_queue_by_worker_index + = 0; + vlib_frame_queue_elt_t *hf = 0; + vlib_frame_t *f = 0; + int i; + u32 n_left_to_next_worker = 0, *to_next_worker = 0; + u32 next_worker_index = 0; + u32 current_worker_index = ~0; + u32 cpu_index = os_get_cpu_number (); + + ASSERT (vec_len (sm->workers)); + + if (PREDICT_FALSE (handoff_queue_elt_by_worker_index == 0)) + { + vec_validate (handoff_queue_elt_by_worker_index, tm->n_vlib_mains - 1); + + vec_validate_init_empty (congested_handoff_queue_by_worker_index, + sm->first_worker_index + sm->num_workers - 1, + (vlib_frame_queue_t *) (~0)); + } + + from = vlib_frame_vector_args (frame); + n_left_from = frame->n_vectors; + + while (n_left_from > 0) + { + u32 bi0; + vlib_buffer_t *b0; + u32 sw_if_index0; + u32 rx_fib_index0; + ip4_header_t * ip0; + udp_header_t * udp0; + snat_static_mapping_key_t key0; + clib_bihash_kv_8_8_t kv0, value0; + u8 do_handoff; + + bi0 = from[0]; + from += 1; + n_left_from -= 1; + + b0 = vlib_get_buffer (vm, bi0); + + sw_if_index0 = vnet_buffer (b0)->sw_if_index[VLIB_RX]; + rx_fib_index0 = ip4_fib_table_get_index_for_sw_if_index(sw_if_index0); + + ip0 = vlib_buffer_get_current (b0); + udp0 = ip4_next_header (ip0); + + key0.addr = ip0->dst_address; + key0.port = udp0->dst_port; + key0.fib_index = rx_fib_index0; + + kv0.key = key0.as_u64; + + /* Ever heard of of the "user" before? */ + if (clib_bihash_search_8_8 (&sm->worker_by_out, &kv0, &value0)) + { + key0.port = 0; + kv0.key = key0.as_u64; + + if (clib_bihash_search_8_8 (&sm->worker_by_out, &kv0, &value0)) + { + /* No, assign next available worker (RR) */ + next_worker_index = sm->first_worker_index + + sm->workers[sm->next_worker++ % vec_len (sm->workers)]; + } + else + { + /* Static mapping without port */ + next_worker_index = value0.value; + } + + /* Add to translated packets worker lookup */ + kv0.value = next_worker_index; + clib_bihash_add_del_8_8 (&sm->worker_by_out, &kv0, 1); + } + else + next_worker_index = value0.value; + + if (PREDICT_FALSE (next_worker_index != cpu_index)) + { + do_handoff = 1; + + if (next_worker_index != current_worker_index) + { + if (hf) + hf->n_vectors = VLIB_FRAME_SIZE - n_left_to_next_worker; + + hf = vlib_get_worker_handoff_queue_elt (sm->fq_out2in_index, + next_worker_index, + handoff_queue_elt_by_worker_index); + + n_left_to_next_worker = VLIB_FRAME_SIZE - hf->n_vectors; + to_next_worker = &hf->buffer_index[hf->n_vectors]; + current_worker_index = next_worker_index; + } + + /* enqueue to correct worker thread */ + to_next_worker[0] = bi0; + to_next_worker++; + n_left_to_next_worker--; + + if (n_left_to_next_worker == 0) + { + hf->n_vectors = VLIB_FRAME_SIZE; + vlib_put_frame_queue_elt (hf); + current_worker_index = ~0; + handoff_queue_elt_by_worker_index[next_worker_index] = 0; + hf = 0; + } + } + else + { + do_handoff = 0; + /* if this is 1st frame */ + if (!f) + { + f = vlib_get_frame_to_node (vm, snat_out2in_node.index); + to_next = vlib_frame_vector_args (f); + } + + to_next[0] = bi0; + to_next += 1; + f->n_vectors++; + } + + if (PREDICT_FALSE ((node->flags & VLIB_NODE_FLAG_TRACE) + && (b0->flags & VLIB_BUFFER_IS_TRACED))) + { + snat_out2in_worker_handoff_trace_t *t = + vlib_add_trace (vm, node, b0, sizeof (*t)); + t->next_worker_index = next_worker_index; + t->do_handoff = do_handoff; + } + } + + if (f) + vlib_put_frame_to_node (vm, snat_out2in_node.index, f); + + if (hf) + hf->n_vectors = VLIB_FRAME_SIZE - n_left_to_next_worker; + + /* Ship frames to the worker nodes */ + for (i = 0; i < vec_len (handoff_queue_elt_by_worker_index); i++) + { + if (handoff_queue_elt_by_worker_index[i]) + { + hf = handoff_queue_elt_by_worker_index[i]; + /* + * It works better to let the handoff node + * rate-adapt, always ship the handoff queue element. + */ + if (1 || hf->n_vectors == hf->last_n_vectors) + { + vlib_put_frame_queue_elt (hf); + handoff_queue_elt_by_worker_index[i] = 0; + } + else + hf->last_n_vectors = hf->n_vectors; + } + congested_handoff_queue_by_worker_index[i] = + (vlib_frame_queue_t *) (~0); + } + hf = 0; + current_worker_index = ~0; + return frame->n_vectors; +} + +VLIB_REGISTER_NODE (snat_out2in_worker_handoff_node) = { + .function = snat_out2in_worker_handoff_fn, + .name = "snat-out2in-worker-handoff", + .vector_size = sizeof (u32), + .format_trace = format_snat_out2in_worker_handoff_trace, + .type = VLIB_NODE_TYPE_INTERNAL, + + .n_next_nodes = 1, + + .next_nodes = { + [0] = "error-drop", + }, +}; + +VLIB_NODE_FUNCTION_MULTIARCH (snat_out2in_worker_handoff_node, snat_out2in_worker_handoff_fn); + +static inline u32 icmp_out2in_fast (snat_main_t *sm, + vlib_buffer_t * b0, + ip4_header_t * ip0, + icmp46_header_t * icmp0, + u32 sw_if_index0, + vlib_node_runtime_t * node, + u32 next0, + u32 rx_fib_index0) +{ + snat_session_key_t key0, sm0; + icmp_echo_header_t *echo0; + u32 new_addr0, old_addr0; + u16 old_id0, new_id0; + ip_csum_t sum0; + snat_runtime_t * rt = (snat_runtime_t *)node->runtime_data; + + echo0 = (icmp_echo_header_t *)(icmp0+1); + + key0.addr = ip0->dst_address; + key0.port = echo0->identifier; + key0.fib_index = rx_fib_index0; + + if (snat_static_mapping_match(sm, key0, &sm0, 1)) + { + ip4_address_t * first_int_addr; + + if (PREDICT_FALSE(rt->cached_sw_if_index != sw_if_index0)) + { + first_int_addr = + ip4_interface_first_address (sm->ip4_main, sw_if_index0, + 0 /* just want the address */); + rt->cached_sw_if_index = sw_if_index0; + rt->cached_ip4_address = first_int_addr->as_u32; + } + + /* Don't NAT packet aimed at the intfc address */ + if (PREDICT_FALSE(ip0->dst_address.as_u32 == + rt->cached_ip4_address)) + return next0; + + b0->error = node->errors[SNAT_OUT2IN_ERROR_NO_TRANSLATION]; + return SNAT_OUT2IN_NEXT_DROP; + } + + new_addr0 = sm0.addr.as_u32; + new_id0 = sm0.port; + vnet_buffer(b0)->sw_if_index[VLIB_TX] = sm0.fib_index; + + old_addr0 = ip0->dst_address.as_u32; + ip0->dst_address.as_u32 = new_addr0; + + sum0 = ip0->checksum; + sum0 = ip_csum_update (sum0, old_addr0, new_addr0, + ip4_header_t, + dst_address /* changed member */); + ip0->checksum = ip_csum_fold (sum0); + + if (PREDICT_FALSE(new_id0 != echo0->identifier)) + { + old_id0 = echo0->identifier; + echo0->identifier = new_id0; + + sum0 = icmp0->checksum; + sum0 = ip_csum_update (sum0, old_id0, new_id0, icmp_echo_header_t, + identifier); + icmp0->checksum = ip_csum_fold (sum0); + } + + return next0; +} + +static uword +snat_out2in_fast_node_fn (vlib_main_t * vm, + vlib_node_runtime_t * node, + vlib_frame_t * frame) +{ + u32 n_left_from, * from, * to_next; + snat_out2in_next_t next_index; + u32 pkts_processed = 0; + snat_main_t * sm = &snat_main; + + from = vlib_frame_vector_args (frame); + n_left_from = frame->n_vectors; + next_index = node->cached_next_index; + + while (n_left_from > 0) + { + u32 n_left_to_next; + + vlib_get_next_frame (vm, node, next_index, + to_next, n_left_to_next); + + while (n_left_from > 0 && n_left_to_next > 0) + { + u32 bi0; + vlib_buffer_t * b0; + u32 next0 = SNAT_OUT2IN_NEXT_DROP; + u32 sw_if_index0; + ip4_header_t * ip0; + ip_csum_t sum0; + u32 new_addr0, old_addr0; + u16 new_port0, old_port0; + udp_header_t * udp0; + tcp_header_t * tcp0; + icmp46_header_t * icmp0; + snat_session_key_t key0, sm0; + u32 proto0; + u32 rx_fib_index0; + + /* speculatively enqueue b0 to the current next frame */ + bi0 = from[0]; + to_next[0] = bi0; + from += 1; + to_next += 1; + n_left_from -= 1; + n_left_to_next -= 1; + + b0 = vlib_get_buffer (vm, bi0); + + ip0 = vlib_buffer_get_current (b0); + udp0 = ip4_next_header (ip0); + tcp0 = (tcp_header_t *) udp0; + icmp0 = (icmp46_header_t *) udp0; + + sw_if_index0 = vnet_buffer(b0)->sw_if_index[VLIB_RX]; + rx_fib_index0 = ip4_fib_table_get_index_for_sw_if_index(sw_if_index0); + + vnet_feature_next (sw_if_index0, &next0, b0); + + proto0 = ~0; + proto0 = (ip0->protocol == IP_PROTOCOL_UDP) + ? SNAT_PROTOCOL_UDP : proto0; + proto0 = (ip0->protocol == IP_PROTOCOL_TCP) + ? SNAT_PROTOCOL_TCP : proto0; + proto0 = (ip0->protocol == IP_PROTOCOL_ICMP) + ? SNAT_PROTOCOL_ICMP : proto0; + + if (PREDICT_FALSE (proto0 == ~0)) + goto trace00; + + if (PREDICT_FALSE (proto0 == SNAT_PROTOCOL_ICMP)) + { + next0 = icmp_out2in_fast + (sm, b0, ip0, icmp0, sw_if_index0, node, next0, rx_fib_index0); + goto trace00; + } + + key0.addr = ip0->dst_address; + key0.port = udp0->dst_port; + key0.fib_index = rx_fib_index0; + + if (snat_static_mapping_match(sm, key0, &sm0, 1)) + { + b0->error = node->errors[SNAT_OUT2IN_ERROR_NO_TRANSLATION]; + goto trace00; + } + + new_addr0 = sm0.addr.as_u32; + new_port0 = sm0.port; + vnet_buffer(b0)->sw_if_index[VLIB_TX] = sm0.fib_index; + old_addr0 = ip0->dst_address.as_u32; + ip0->dst_address.as_u32 = new_addr0; + + sum0 = ip0->checksum; + sum0 = ip_csum_update (sum0, old_addr0, new_addr0, + ip4_header_t, + dst_address /* changed member */); + ip0->checksum = ip_csum_fold (sum0); + + if (PREDICT_FALSE(new_port0 != udp0->dst_port)) + { + if (PREDICT_TRUE(proto0 == SNAT_PROTOCOL_TCP)) + { + old_port0 = tcp0->ports.dst; + tcp0->ports.dst = new_port0; + + sum0 = tcp0->checksum; + sum0 = ip_csum_update (sum0, old_addr0, new_addr0, + ip4_header_t, + dst_address /* changed member */); + + sum0 = ip_csum_update (sum0, old_port0, new_port0, + ip4_header_t /* cheat */, + length /* changed member */); + tcp0->checksum = ip_csum_fold(sum0); + } + else + { + old_port0 = udp0->dst_port; + udp0->dst_port = new_port0; + udp0->checksum = 0; + } + } + else + { + if (PREDICT_TRUE(proto0 == SNAT_PROTOCOL_TCP)) + { + sum0 = tcp0->checksum; + sum0 = ip_csum_update (sum0, old_addr0, new_addr0, + ip4_header_t, + dst_address /* changed member */); + + tcp0->checksum = ip_csum_fold(sum0); + } + } + + trace00: + + if (PREDICT_FALSE((node->flags & VLIB_NODE_FLAG_TRACE) + && (b0->flags & VLIB_BUFFER_IS_TRACED))) + { + snat_out2in_trace_t *t = + vlib_add_trace (vm, node, b0, sizeof (*t)); + t->sw_if_index = sw_if_index0; + t->next_index = next0; + } + + pkts_processed += next0 != SNAT_OUT2IN_NEXT_DROP; + + /* verify speculative enqueue, maybe switch current next frame */ + vlib_validate_buffer_enqueue_x1 (vm, node, next_index, + to_next, n_left_to_next, + bi0, next0); + } + + vlib_put_next_frame (vm, node, next_index, n_left_to_next); + } + + vlib_node_increment_counter (vm, snat_out2in_fast_node.index, + SNAT_OUT2IN_ERROR_OUT2IN_PACKETS, + pkts_processed); + return frame->n_vectors; +} + +VLIB_REGISTER_NODE (snat_out2in_fast_node) = { + .function = snat_out2in_fast_node_fn, + .name = "snat-out2in-fast", + .vector_size = sizeof (u32), + .format_trace = format_snat_out2in_fast_trace, + .type = VLIB_NODE_TYPE_INTERNAL, + + .n_errors = ARRAY_LEN(snat_out2in_error_strings), + .error_strings = snat_out2in_error_strings, + + .runtime_data_bytes = sizeof (snat_runtime_t), + + .n_next_nodes = SNAT_OUT2IN_N_NEXT, + + /* edit / add dispositions here */ + .next_nodes = { + [SNAT_OUT2IN_NEXT_LOOKUP] = "ip4-lookup", + [SNAT_OUT2IN_NEXT_DROP] = "error-drop", + }, +}; +VLIB_NODE_FUNCTION_MULTIARCH (snat_out2in_fast_node, snat_out2in_fast_node_fn); diff --git a/src/plugins/snat/snat.api b/src/plugins/snat/snat.api new file mode 100644 index 00000000..a191eed5 --- /dev/null +++ b/src/plugins/snat/snat.api @@ -0,0 +1,283 @@ +/* + * Copyright (c) 2016 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ +/** + * @file snat.api + * @brief VPP control-plane API messages. + * + * This file defines VPP control-plane API messages which are generally + * called through a shared memory interface. + */ + +/** \brief Add/del S-NAT address range + @param client_index - opaque cookie to identify the sender + @param context - sender context, to match reply w/ request + @param is_ip4 - 1 if address type is IPv4 + @first_ip_address - first IP address + @last_ip_address - last IP address + @is_add - 1 if add, 0 if delete +*/ +define snat_add_address_range { + u32 client_index; + u32 context; + u8 is_ip4; + u8 first_ip_address[16]; + u8 last_ip_address[16]; + u8 is_add; +}; + +/** \brief Add S-NAT address range reply + @param client_index - opaque cookie to identify the sender + @param context - sender context, to match reply w/ request + @param retval - return code +*/ +define snat_add_address_range_reply { + u32 context; + i32 retval; +}; + +/** \brief Dump S-NAT addresses + @param client_index - opaque cookie to identify the sender + @param context - sender context, to match reply w/ request +*/ +define snat_address_dump { + u32 client_index; + u32 context; +}; + +/** \brief S-NAT address details response + @param context - sender context, to match reply w/ request + @param is_ip4 - 1 if address type is IPv4 + @param ip_address - IP address +*/ +define snat_address_details { + u32 context; + u8 is_ip4; + u8 ip_address[16]; +}; + +/** \brief Enable/disable S-NAT feature on the interface + @param client_index - opaque cookie to identify the sender + @param context - sender context, to match reply w/ request + @param is_add - 1 if add, 0 if delete + @param is_inside - 1 if inside, 0 if outside + @param sw_if_index - software index of the interface +*/ +define snat_interface_add_del_feature { + u32 client_index; + u32 context; + u8 is_add; + u8 is_inside; + u32 sw_if_index; +}; + +/** \brief Enable/disable S-NAT feature on the interface reply + @param client_index - opaque cookie to identify the sender + @param context - sender context, to match reply w/ request + @param retval - return code +*/ +define snat_interface_add_del_feature_reply { + u32 context; + i32 retval; +}; + +/** \brief Dump interfaces with S-NAT feature + @param client_index - opaque cookie to identify the sender + @param context - sender context, to match reply w/ request +*/ +define snat_interface_dump { + u32 client_index; + u32 context; +}; + +/** \brief S-NAT interface details response + @param context - sender context, to match reply w/ request + @param is_inside - 1 if inside, 0 if outside + @param sw_if_index - software index of the interface +*/ +define snat_interface_details { + u32 context; + u8 is_inside; + u32 sw_if_index; +}; + +/** \brief Add/delete S-NAT static mapping + @param client_index - opaque cookie to identify the sender + @param context - sender context, to match reply w/ request + @param is_add - 1 if add, 0 if delete + @param is_ip4 - 1 if address type is IPv4 + @param addr_only - 1 if address only mapping + @param local_ip_address - local IP address + @param external_ip_address - external IP address + @param local_port - local port number + @param external_port - external port number + @param vfr_id - VRF ID +*/ +define snat_add_static_mapping { + u32 client_index; + u32 context; + u8 is_add; + u8 is_ip4; + u8 addr_only; + u8 local_ip_address[16]; + u8 external_ip_address[16]; + u16 local_port; + u16 external_port; + u32 vrf_id; +}; + +/** \brief Add/delete S-NAT static mapping reply + @param client_index - opaque cookie to identify the sender + @param context - sender context, to match reply w/ request + @param retval - return code +*/ +define snat_add_static_mapping_reply { + u32 context; + i32 retval; +}; + +/** \brief Dump S-NAT static mappings + @param client_index - opaque cookie to identify the sender + @param context - sender context, to match reply w/ request +*/ +define snat_static_mapping_dump { + u32 client_index; + u32 context; +}; + +/** \brief S-NAT static mapping details response + @param context - sender context, to match reply w/ request + @param is_ip4 - 1 if address type is IPv4 + @param addr_only - 1 if address only mapping + @param local_ip_address - local IP address + @param external_ip_address - external IP address + @param local_port - local port number + @param external_port - external port number + @param vfr_id - VRF ID +*/ +define snat_static_mapping_details { + u32 context; + u8 is_ip4; + u8 addr_only; + u8 local_ip_address[16]; + u8 external_ip_address[16]; + u16 local_port; + u16 external_port; + u32 vrf_id; +}; + +/** \brief Control ping from client to api server request + @param client_index - opaque cookie to identify the sender + @param context - sender context, to match reply w/ request +*/ +define snat_control_ping +{ + u32 client_index; + u32 context; +}; + +/** \brief Control ping from the client to the server response + @param client_index - opaque cookie to identify the sender + @param context - sender context, to match reply w/ request + @param retval - return code for the request + @param vpe_pid - the pid of the vpe, returned by the server +*/ +define snat_control_ping_reply +{ + u32 context; + i32 retval; + u32 client_index; + u32 vpe_pid; +}; + +/** \brief Show S-NAT plugin startup config + @param client_index - opaque cookie to identify the sender + @param context - sender context, to match reply w/ request +*/ +define snat_show_config +{ + u32 client_index; + u32 context; +}; + +/** \brief Show S-NAT plugin startup config reply + @param context - sender context, to match reply w/ request + @param retval - return code for the request + @param static_mapping_only - if 1 dynamic translations disabled + @param static_mapping_connection_tracking - if 1 create session data + @param translation_buckets - number of translation hash buckets + @param translation_memory_size - translation hash memory size + @param user_buckets - number of user hash buckets + @param user_memory_size - user hash memory size + @param max_translations_per_user - maximum number of translations per user + @param outside_vrf_id - outside VRF id + @param inside_vrf_id - default inside VRF id +*/ +define snat_show_config_reply +{ + u32 context; + i32 retval; + u8 static_mapping_only; + u8 static_mapping_connection_tracking; + u32 translation_buckets; + u32 translation_memory_size; + u32 user_buckets; + u32 user_memory_size; + u32 max_translations_per_user; + u32 outside_vrf_id; + u32 inside_vrf_id; +}; + +/** \brief Set S-NAT workers + @param client_index - opaque cookie to identify the sender + @param context - sender context, to match reply w/ request + @param worker_mask - S-NAT workers mask +*/ +define snat_set_workers { + u32 client_index; + u32 context; + u64 worker_mask; +}; + +/** \brief Set S-NAT workers reply + @param client_index - opaque cookie to identify the sender + @param context - sender context, to match reply w/ request + @param retval - return code +*/ +define snat_set_workers_reply { + u32 context; + i32 retval; +}; + +/** \brief Dump S-NAT workers + @param client_index - opaque cookie to identify the sender + @param context - sender context, to match reply w/ request +*/ +define snat_worker_dump { + u32 client_index; + u32 context; +}; + +/** \brief S-NAT workers details response + @param context - sender context, to match reply w/ request + @param worker_index - worker index + @param lcore_id - lcore ID + @param name - worker name +*/ +define snat_worker_details { + u32 context; + u32 worker_index; + u32 lcore_id; + u8 name[64]; +}; diff --git a/src/plugins/snat/snat.c b/src/plugins/snat/snat.c new file mode 100644 index 00000000..bc995684 --- /dev/null +++ b/src/plugins/snat/snat.c @@ -0,0 +1,1957 @@ +/* + * snat.c - simple nat plugin + * + * Copyright (c) 2016 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +#include +#include +#include +#include + +#include +#include +#include + +snat_main_t snat_main; + +/* define message IDs */ +#include + +/* define message structures */ +#define vl_typedefs +#include +#undef vl_typedefs + +/* define generated endian-swappers */ +#define vl_endianfun +#include +#undef vl_endianfun + +#define vl_print(handle, ...) vlib_cli_output (handle, __VA_ARGS__) + +/* Get the API version number */ +#define vl_api_version(n,v) static u32 api_version=(v); +#include +#undef vl_api_version + +/* Macro to finish up custom dump fns */ +#define FINISH \ + vec_add1 (s, 0); \ + vl_print (handle, (char *)s); \ + vec_free (s); \ + return handle; + +/* + * A handy macro to set up a message reply. + * Assumes that the following variables are available: + * mp - pointer to request message + * rmp - pointer to reply message type + * rv - return value + */ + +#define REPLY_MACRO(t) \ +do { \ + unix_shared_memory_queue_t * q = \ + vl_api_client_index_to_input_queue (mp->client_index); \ + if (!q) \ + return; \ + \ + rmp = vl_msg_api_alloc (sizeof (*rmp)); \ + rmp->_vl_msg_id = ntohs((t)+sm->msg_id_base); \ + rmp->context = mp->context; \ + rmp->retval = ntohl(rv); \ + \ + vl_msg_api_send_shmem (q, (u8 *)&rmp); \ +} while(0); + +#define REPLY_MACRO2(t, body) \ +do { \ + unix_shared_memory_queue_t * q = \ + vl_api_client_index_to_input_queue (mp->client_index); \ + if (!q) \ + return; \ + \ + rmp = vl_msg_api_alloc (sizeof (*rmp)); \ + rmp->_vl_msg_id = ntohs((t)+sm->msg_id_base); \ + rmp->context = mp->context; \ + rmp->retval = ntohl(rv); \ + do {body;} while (0); \ + vl_msg_api_send_shmem (q, (u8 *)&rmp); \ +} while(0); + + +/* Hook up input features */ +VNET_FEATURE_INIT (ip4_snat_in2out, static) = { + .arc_name = "ip4-unicast", + .node_name = "snat-in2out", + .runs_before = VNET_FEATURES ("snat-out2in"), +}; +VNET_FEATURE_INIT (ip4_snat_out2in, static) = { + .arc_name = "ip4-unicast", + .node_name = "snat-out2in", + .runs_before = VNET_FEATURES ("ip4-lookup"), +}; +VNET_FEATURE_INIT (ip4_snat_in2out_worker_handoff, static) = { + .arc_name = "ip4-unicast", + .node_name = "snat-in2out-worker-handoff", + .runs_before = VNET_FEATURES ("snat-out2in-worker-handoff"), +}; +VNET_FEATURE_INIT (ip4_snat_out2in_worker_handoff, static) = { + .arc_name = "ip4-unicast", + .node_name = "snat-out2in-worker-handoff", + .runs_before = VNET_FEATURES ("ip4-lookup"), +}; +VNET_FEATURE_INIT (ip4_snat_in2out_fast, static) = { + .arc_name = "ip4-unicast", + .node_name = "snat-in2out-fast", + .runs_before = VNET_FEATURES ("snat-out2in-fast"), +}; +VNET_FEATURE_INIT (ip4_snat_out2in_fast, static) = { + .arc_name = "ip4-unicast", + .node_name = "snat-out2in-fast", + .runs_before = VNET_FEATURES ("ip4-lookup"), +}; + + +/* + * This routine exists to convince the vlib plugin framework that + * we haven't accidentally copied a random .dll into the plugin directory. + * + * Also collects global variable pointers passed from the vpp engine + */ + +clib_error_t * +vlib_plugin_register (vlib_main_t * vm, vnet_plugin_handoff_t * h, + int from_early_init) +{ + snat_main_t * sm = &snat_main; + clib_error_t * error = 0; + + sm->vlib_main = vm; + sm->vnet_main = h->vnet_main; + sm->ethernet_main = h->ethernet_main; + + return error; +} + +/*$$$$$ move to an installed header file */ +#if (1 || CLIB_DEBUG > 0) /* "trust, but verify" */ + +#define VALIDATE_SW_IF_INDEX(mp) \ + do { u32 __sw_if_index = ntohl(mp->sw_if_index); \ + vnet_main_t *__vnm = vnet_get_main(); \ + if (pool_is_free_index(__vnm->interface_main.sw_interfaces, \ + __sw_if_index)) { \ + rv = VNET_API_ERROR_INVALID_SW_IF_INDEX; \ + goto bad_sw_if_index; \ + } \ +} while(0); + +#define BAD_SW_IF_INDEX_LABEL \ +do { \ +bad_sw_if_index: \ + ; \ +} while (0); + +#define VALIDATE_RX_SW_IF_INDEX(mp) \ + do { u32 __rx_sw_if_index = ntohl(mp->rx_sw_if_index); \ + vnet_main_t *__vnm = vnet_get_main(); \ + if (pool_is_free_index(__vnm->interface_main.sw_interfaces, \ + __rx_sw_if_index)) { \ + rv = VNET_API_ERROR_INVALID_SW_IF_INDEX; \ + goto bad_rx_sw_if_index; \ + } \ +} while(0); + +#define BAD_RX_SW_IF_INDEX_LABEL \ +do { \ +bad_rx_sw_if_index: \ + ; \ +} while (0); + +#define VALIDATE_TX_SW_IF_INDEX(mp) \ + do { u32 __tx_sw_if_index = ntohl(mp->tx_sw_if_index); \ + vnet_main_t *__vnm = vnet_get_main(); \ + if (pool_is_free_index(__vnm->interface_main.sw_interfaces, \ + __tx_sw_if_index)) { \ + rv = VNET_API_ERROR_INVALID_SW_IF_INDEX; \ + goto bad_tx_sw_if_index; \ + } \ +} while(0); + +#define BAD_TX_SW_IF_INDEX_LABEL \ +do { \ +bad_tx_sw_if_index: \ + ; \ +} while (0); + +#else + +#define VALIDATE_SW_IF_INDEX(mp) +#define BAD_SW_IF_INDEX_LABEL +#define VALIDATE_RX_SW_IF_INDEX(mp) +#define BAD_RX_SW_IF_INDEX_LABEL +#define VALIDATE_TX_SW_IF_INDEX(mp) +#define BAD_TX_SW_IF_INDEX_LABEL + +#endif /* CLIB_DEBUG > 0 */ + +void snat_add_address (snat_main_t *sm, ip4_address_t *addr) +{ + snat_address_t * ap; + + /* Check if address already exists */ + vec_foreach (ap, sm->addresses) + { + if (ap->addr.as_u32 == addr->as_u32) + return; + } + + vec_add2 (sm->addresses, ap, 1); + ap->addr = *addr; + clib_bitmap_alloc (ap->busy_port_bitmap, 65535); +} + +static int is_snat_address_used_in_static_mapping (snat_main_t *sm, + ip4_address_t addr) +{ + snat_static_mapping_t *m; + pool_foreach (m, sm->static_mappings, + ({ + if (m->external_addr.as_u32 == addr.as_u32) + return 1; + })); + + return 0; +} + +int snat_del_address (snat_main_t *sm, ip4_address_t addr) +{ + snat_address_t *a = 0; + snat_session_t *ses; + u32 *ses_to_be_removed = 0, *ses_index; + clib_bihash_kv_8_8_t kv, value; + snat_user_key_t user_key; + snat_user_t *u; + snat_main_per_thread_data_t *tsm; + + int i; + + /* Find SNAT address */ + for (i=0; i < vec_len (sm->addresses); i++) + { + if (sm->addresses[i].addr.as_u32 == addr.as_u32) + { + a = sm->addresses + i; + break; + } + } + if (!a) + return VNET_API_ERROR_NO_SUCH_ENTRY; + + /* Check if address is used in some static mapping */ + if (is_snat_address_used_in_static_mapping(sm, addr)) + { + clib_warning ("address used in static mapping"); + return VNET_API_ERROR_UNSPECIFIED; + } + + /* Delete sessions using address */ + if (a->busy_ports) + { + vec_foreach (tsm, sm->per_thread_data) + { + pool_foreach (ses, tsm->sessions, ({ + if (ses->out2in.addr.as_u32 == addr.as_u32) + { + vec_add1 (ses_to_be_removed, ses - tsm->sessions); + kv.key = ses->in2out.as_u64; + clib_bihash_add_del_8_8 (&sm->in2out, &kv, 0); + kv.key = ses->out2in.as_u64; + clib_bihash_add_del_8_8 (&sm->out2in, &kv, 0); + clib_dlist_remove (tsm->list_pool, ses->per_user_index); + user_key.addr = ses->in2out.addr; + user_key.fib_index = ses->in2out.fib_index; + kv.key = user_key.as_u64; + if (!clib_bihash_search_8_8 (&sm->user_hash, &kv, &value)) + { + u = pool_elt_at_index (tsm->users, value.value); + u->nsessions--; + } + } + })); + + vec_foreach (ses_index, ses_to_be_removed) + pool_put_index (tsm->sessions, ses_index[0]); + + vec_free (ses_to_be_removed); + } + } + + vec_del1 (sm->addresses, i); + + return 0; +} + +static void increment_v4_address (ip4_address_t * a) +{ + u32 v; + + v = clib_net_to_host_u32(a->as_u32) + 1; + a->as_u32 = clib_host_to_net_u32(v); +} + +/** + * @brief Add static mapping. + * + * Create static mapping between local addr+port and external addr+port. + * + * @param l_addr Local IPv4 address. + * @param e_addr External IPv4 address. + * @param l_port Local port number. + * @param e_port External port number. + * @param vrf_id VRF ID. + * @param addr_only If 0 address port and pair mapping, otherwise address only. + * @param is_add If 0 delete static mapping, otherwise add. + * + * @returns + */ +int snat_add_static_mapping(ip4_address_t l_addr, ip4_address_t e_addr, + u16 l_port, u16 e_port, u32 vrf_id, int addr_only, + int is_add) +{ + snat_main_t * sm = &snat_main; + snat_static_mapping_t *m; + snat_static_mapping_key_t m_key; + clib_bihash_kv_8_8_t kv, value; + snat_address_t *a = 0; + u32 fib_index = ~0; + uword * p; + int i; + + /* If outside FIB index is not resolved yet */ + if (sm->outside_fib_index == ~0) + { + p = hash_get (sm->ip4_main->fib_index_by_table_id, sm->outside_vrf_id); + if (!p) + return VNET_API_ERROR_NO_SUCH_FIB; + sm->outside_fib_index = p[0]; + } + + m_key.addr = e_addr; + m_key.port = addr_only ? 0 : e_port; + m_key.fib_index = sm->outside_fib_index; + kv.key = m_key.as_u64; + if (clib_bihash_search_8_8 (&sm->static_mapping_by_external, &kv, &value)) + m = 0; + else + m = pool_elt_at_index (sm->static_mappings, value.value); + + if (is_add) + { + if (m) + return VNET_API_ERROR_VALUE_EXIST; + + /* Convert VRF id to FIB index */ + if (vrf_id != ~0) + { + p = hash_get (sm->ip4_main->fib_index_by_table_id, vrf_id); + if (!p) + return VNET_API_ERROR_NO_SUCH_FIB; + fib_index = p[0]; + } + /* If not specified use inside VRF id from SNAT plugin startup config */ + else + { + if (sm->inside_fib_index == ~0) + { + p = hash_get (sm->ip4_main->fib_index_by_table_id, sm->inside_vrf_id); + if (!p) + return VNET_API_ERROR_NO_SUCH_FIB; + fib_index = p[0]; + sm->inside_fib_index = fib_index; + } + else + fib_index = sm->inside_fib_index; + + vrf_id = sm->inside_vrf_id; + } + + /* Find external address in allocated addresses and reserve port for + address and port pair mapping when dynamic translations enabled */ + if (!addr_only && !(sm->static_mapping_only)) + { + for (i = 0; i < vec_len (sm->addresses); i++) + { + if (sm->addresses[i].addr.as_u32 == e_addr.as_u32) + { + a = sm->addresses + i; + /* External port must be unused */ + if (clib_bitmap_get_no_check (a->busy_port_bitmap, e_port)) + return VNET_API_ERROR_INVALID_VALUE; + clib_bitmap_set_no_check (a->busy_port_bitmap, e_port, 1); + if (e_port > 1024) + a->busy_ports++; + + break; + } + } + /* External address must be allocated */ + if (!a) + return VNET_API_ERROR_NO_SUCH_ENTRY; + } + + pool_get (sm->static_mappings, m); + memset (m, 0, sizeof (*m)); + m->local_addr = l_addr; + m->external_addr = e_addr; + m->addr_only = addr_only; + m->vrf_id = vrf_id; + m->fib_index = fib_index; + if (!addr_only) + { + m->local_port = l_port; + m->external_port = e_port; + } + + m_key.addr = m->local_addr; + m_key.port = m->local_port; + m_key.fib_index = m->fib_index; + kv.key = m_key.as_u64; + kv.value = m - sm->static_mappings; + clib_bihash_add_del_8_8(&sm->static_mapping_by_local, &kv, 1); + + m_key.addr = m->external_addr; + m_key.port = m->external_port; + m_key.fib_index = sm->outside_fib_index; + kv.key = m_key.as_u64; + kv.value = m - sm->static_mappings; + clib_bihash_add_del_8_8(&sm->static_mapping_by_external, &kv, 1); + + /* Assign worker */ + if (sm->workers) + { + snat_user_key_t w_key0; + snat_static_mapping_key_t w_key1; + + w_key0.addr = m->local_addr; + w_key0.fib_index = m->fib_index; + kv.key = w_key0.as_u64; + + if (clib_bihash_search_8_8 (&sm->worker_by_in, &kv, &value)) + { + kv.value = sm->first_worker_index + + sm->workers[sm->next_worker++ % vec_len (sm->workers)]; + + clib_bihash_add_del_8_8 (&sm->worker_by_in, &kv, 1); + } + else + { + kv.value = value.value; + } + + w_key1.addr = m->external_addr; + w_key1.port = clib_host_to_net_u16 (m->external_port); + w_key1.fib_index = sm->outside_fib_index; + kv.key = w_key1.as_u64; + clib_bihash_add_del_8_8 (&sm->worker_by_out, &kv, 1); + } + } + else + { + if (!m) + return VNET_API_ERROR_NO_SUCH_ENTRY; + + /* Free external address port */ + if (!addr_only && !(sm->static_mapping_only)) + { + for (i = 0; i < vec_len (sm->addresses); i++) + { + if (sm->addresses[i].addr.as_u32 == e_addr.as_u32) + { + a = sm->addresses + i; + clib_bitmap_set_no_check (a->busy_port_bitmap, e_port, 0); + a->busy_ports--; + + break; + } + } + } + + m_key.addr = m->local_addr; + m_key.port = m->local_port; + m_key.fib_index = m->fib_index; + kv.key = m_key.as_u64; + clib_bihash_add_del_8_8(&sm->static_mapping_by_local, &kv, 0); + + m_key.addr = m->external_addr; + m_key.port = m->external_port; + m_key.fib_index = sm->outside_fib_index; + kv.key = m_key.as_u64; + clib_bihash_add_del_8_8(&sm->static_mapping_by_external, &kv, 0); + + /* Delete session(s) for static mapping if exist */ + if (!(sm->static_mapping_only) || + (sm->static_mapping_only && sm->static_mapping_connection_tracking)) + { + snat_user_key_t u_key; + snat_user_t *u; + dlist_elt_t * head, * elt; + u32 elt_index, head_index, del_elt_index; + u32 ses_index; + u64 user_index; + snat_session_t * s; + snat_main_per_thread_data_t *tsm; + + u_key.addr = m->local_addr; + u_key.fib_index = m->fib_index; + kv.key = u_key.as_u64; + if (!clib_bihash_search_8_8 (&sm->user_hash, &kv, &value)) + { + user_index = value.value; + if (!clib_bihash_search_8_8 (&sm->worker_by_in, &kv, &value)) + tsm = vec_elt_at_index (sm->per_thread_data, value.value); + else + tsm = vec_elt_at_index (sm->per_thread_data, sm->num_workers); + u = pool_elt_at_index (tsm->users, user_index); + if (u->nstaticsessions) + { + head_index = u->sessions_per_user_list_head_index; + head = pool_elt_at_index (tsm->list_pool, head_index); + elt_index = head->next; + elt = pool_elt_at_index (tsm->list_pool, elt_index); + ses_index = elt->value; + while (ses_index != ~0) + { + s = pool_elt_at_index (tsm->sessions, ses_index); + del_elt_index = elt_index; + elt_index = elt->next; + elt = pool_elt_at_index (tsm->list_pool, elt_index); + ses_index = elt->value; + + if (!addr_only) + { + if ((s->out2in.addr.as_u32 != e_addr.as_u32) && + (clib_net_to_host_u16 (s->out2in.port) != e_port)) + continue; + } + + value.key = s->in2out.as_u64; + clib_bihash_add_del_8_8 (&sm->in2out, &value, 0); + value.key = s->out2in.as_u64; + clib_bihash_add_del_8_8 (&sm->out2in, &value, 0); + pool_put (tsm->sessions, s); + + clib_dlist_remove (tsm->list_pool, del_elt_index); + pool_put_index (tsm->list_pool, del_elt_index); + u->nstaticsessions--; + + if (!addr_only) + break; + } + if (addr_only) + { + pool_put (tsm->users, u); + clib_bihash_add_del_8_8 (&sm->user_hash, &kv, 0); + } + } + } + } + + /* Delete static mapping from pool */ + pool_put (sm->static_mappings, m); + } + + return 0; +} + +static int snat_interface_add_del (u32 sw_if_index, u8 is_inside, int is_del) +{ + snat_main_t *sm = &snat_main; + snat_interface_t *i; + const char * feature_name; + + if (sm->static_mapping_only && !(sm->static_mapping_connection_tracking)) + feature_name = is_inside ? "snat-in2out-fast" : "snat-out2in-fast"; + else + { + if (sm->num_workers > 1) + feature_name = is_inside ? "snat-in2out-worker-handoff" : "snat-out2in-worker-handoff"; + else + feature_name = is_inside ? "snat-in2out" : "snat-out2in"; + } + + vnet_feature_enable_disable ("ip4-unicast", feature_name, sw_if_index, + !is_del, 0, 0); + + if (sm->fq_in2out_index == ~0) + sm->fq_in2out_index = vlib_frame_queue_main_init (snat_in2out_node.index, 0); + + if (sm->fq_out2in_index == ~0) + sm->fq_out2in_index = vlib_frame_queue_main_init (snat_out2in_node.index, 0); + + pool_foreach (i, sm->interfaces, + ({ + if (i->sw_if_index == sw_if_index) + { + if (is_del) + pool_put (sm->interfaces, i); + else + return VNET_API_ERROR_VALUE_EXIST; + + return 0; + } + })); + + if (is_del) + return VNET_API_ERROR_NO_SUCH_ENTRY; + + pool_get (sm->interfaces, i); + i->sw_if_index = sw_if_index; + i->is_inside = is_inside; + + return 0; +} + +static int snat_set_workers (uword * bitmap) +{ + snat_main_t *sm = &snat_main; + int i; + + if (sm->num_workers < 2) + return VNET_API_ERROR_FEATURE_DISABLED; + + if (clib_bitmap_last_set (bitmap) >= sm->num_workers) + return VNET_API_ERROR_INVALID_WORKER; + + vec_free (sm->workers); + clib_bitmap_foreach (i, bitmap, + ({ + vec_add1(sm->workers, i); + })); + + return 0; +} + +static void +vl_api_snat_add_address_range_t_handler +(vl_api_snat_add_address_range_t * mp) +{ + snat_main_t * sm = &snat_main; + vl_api_snat_add_address_range_reply_t * rmp; + ip4_address_t this_addr; + u32 start_host_order, end_host_order; + int i, count; + int rv = 0; + u32 * tmp; + + if (mp->is_ip4 != 1) + { + rv = VNET_API_ERROR_UNIMPLEMENTED; + goto send_reply; + } + + if (sm->static_mapping_only) + { + rv = VNET_API_ERROR_FEATURE_DISABLED; + goto send_reply; + } + + tmp = (u32 *) mp->first_ip_address; + start_host_order = clib_host_to_net_u32 (tmp[0]); + tmp = (u32 *) mp->last_ip_address; + end_host_order = clib_host_to_net_u32 (tmp[0]); + + count = (end_host_order - start_host_order) + 1; + + if (count > 1024) + clib_warning ("%U - %U, %d addresses...", + format_ip4_address, mp->first_ip_address, + format_ip4_address, mp->last_ip_address, + count); + + memcpy (&this_addr.as_u8, mp->first_ip_address, 4); + + for (i = 0; i < count; i++) + { + if (mp->is_add) + snat_add_address (sm, &this_addr); + else + rv = snat_del_address (sm, this_addr); + + if (rv) + goto send_reply; + + increment_v4_address (&this_addr); + } + + send_reply: + REPLY_MACRO (VL_API_SNAT_ADD_ADDRESS_RANGE_REPLY); +} + +static void *vl_api_snat_add_address_range_t_print +(vl_api_snat_add_address_range_t *mp, void * handle) +{ + u8 * s; + + s = format (0, "SCRIPT: snat_add_address_range "); + s = format (s, "%U ", format_ip4_address, mp->first_ip_address); + if (memcmp (mp->first_ip_address, mp->last_ip_address, 4)) + { + s = format (s, " - %U ", format_ip4_address, mp->last_ip_address); + } + FINISH; +} + +static void +send_snat_address_details +(snat_address_t * a, unix_shared_memory_queue_t * q, u32 context) +{ + vl_api_snat_address_details_t *rmp; + snat_main_t * sm = &snat_main; + + rmp = vl_msg_api_alloc (sizeof (*rmp)); + memset (rmp, 0, sizeof (*rmp)); + rmp->_vl_msg_id = ntohs (VL_API_SNAT_ADDRESS_DETAILS+sm->msg_id_base); + rmp->is_ip4 = 1; + clib_memcpy (rmp->ip_address, &(a->addr), 4); + rmp->context = context; + + vl_msg_api_send_shmem (q, (u8 *) & rmp); +} + +static void +vl_api_snat_address_dump_t_handler +(vl_api_snat_address_dump_t * mp) +{ + unix_shared_memory_queue_t *q; + snat_main_t * sm = &snat_main; + snat_address_t * a; + + q = vl_api_client_index_to_input_queue (mp->client_index); + if (q == 0) + return; + + vec_foreach (a, sm->addresses) + send_snat_address_details (a, q, mp->context); +} + +static void *vl_api_snat_address_dump_t_print +(vl_api_snat_address_dump_t *mp, void * handle) +{ + u8 *s; + + s = format (0, "SCRIPT: snat_address_dump "); + + FINISH; +} + +static void +vl_api_snat_interface_add_del_feature_t_handler +(vl_api_snat_interface_add_del_feature_t * mp) +{ + snat_main_t * sm = &snat_main; + vl_api_snat_interface_add_del_feature_reply_t * rmp; + u8 is_del = mp->is_add == 0; + u32 sw_if_index = ntohl(mp->sw_if_index); + int rv = 0; + + VALIDATE_SW_IF_INDEX(mp); + + rv = snat_interface_add_del (sw_if_index, mp->is_inside, is_del); + + BAD_SW_IF_INDEX_LABEL; + + REPLY_MACRO(VL_API_SNAT_INTERFACE_ADD_DEL_FEATURE_REPLY); +} + +static void *vl_api_snat_interface_add_del_feature_t_print +(vl_api_snat_interface_add_del_feature_t * mp, void *handle) +{ + u8 * s; + + s = format (0, "SCRIPT: snat_interface_add_del_feature "); + s = format (s, "sw_if_index %d %s %s", + clib_host_to_net_u32(mp->sw_if_index), + mp->is_inside ? "in":"out", + mp->is_add ? "" : "del"); + + FINISH; +} + +static void +send_snat_interface_details +(snat_interface_t * i, unix_shared_memory_queue_t * q, u32 context) +{ + vl_api_snat_interface_details_t *rmp; + snat_main_t * sm = &snat_main; + + rmp = vl_msg_api_alloc (sizeof (*rmp)); + memset (rmp, 0, sizeof (*rmp)); + rmp->_vl_msg_id = ntohs (VL_API_SNAT_INTERFACE_DETAILS+sm->msg_id_base); + rmp->sw_if_index = ntohl (i->sw_if_index); + rmp->is_inside = i->is_inside; + rmp->context = context; + + vl_msg_api_send_shmem (q, (u8 *) & rmp); +} + +static void +vl_api_snat_interface_dump_t_handler +(vl_api_snat_interface_dump_t * mp) +{ + unix_shared_memory_queue_t *q; + snat_main_t * sm = &snat_main; + snat_interface_t * i; + + q = vl_api_client_index_to_input_queue (mp->client_index); + if (q == 0) + return; + + pool_foreach (i, sm->interfaces, + ({ + send_snat_interface_details(i, q, mp->context); + })); +} + +static void *vl_api_snat_interface_dump_t_print +(vl_api_snat_interface_dump_t *mp, void * handle) +{ + u8 *s; + + s = format (0, "SCRIPT: snat_interface_dump "); + + FINISH; +}static void + +vl_api_snat_add_static_mapping_t_handler +(vl_api_snat_add_static_mapping_t * mp) +{ + snat_main_t * sm = &snat_main; + vl_api_snat_add_static_mapping_reply_t * rmp; + ip4_address_t local_addr, external_addr; + u16 local_port = 0, external_port = 0; + u32 vrf_id; + int rv = 0; + + if (mp->is_ip4 != 1) + { + rv = VNET_API_ERROR_UNIMPLEMENTED; + goto send_reply; + } + + memcpy (&local_addr.as_u8, mp->local_ip_address, 4); + memcpy (&external_addr.as_u8, mp->external_ip_address, 4); + if (mp->addr_only == 0) + { + local_port = clib_net_to_host_u16 (mp->local_port); + external_port = clib_net_to_host_u16 (mp->external_port); + } + vrf_id = clib_net_to_host_u32 (mp->vrf_id); + + rv = snat_add_static_mapping(local_addr, external_addr, local_port, + external_port, vrf_id, mp->addr_only, + mp->is_add); + + send_reply: + REPLY_MACRO (VL_API_SNAT_ADD_ADDRESS_RANGE_REPLY); +} + +static void *vl_api_snat_add_static_mapping_t_print +(vl_api_snat_add_static_mapping_t *mp, void * handle) +{ + u8 * s; + + s = format (0, "SCRIPT: snat_add_static_mapping "); + s = format (s, "local_addr %U external_addr %U ", + format_ip4_address, mp->local_ip_address, + format_ip4_address, mp->external_ip_address); + + if (mp->addr_only == 0) + s = format (s, "local_port %d external_port %d ", + clib_net_to_host_u16 (mp->local_port), + clib_net_to_host_u16 (mp->external_port)); + + if (mp->vrf_id != ~0) + s = format (s, "vrf %d", clib_net_to_host_u32 (mp->vrf_id)); + + FINISH; +} + +static void +send_snat_static_mapping_details +(snat_static_mapping_t * m, unix_shared_memory_queue_t * q, u32 context) +{ + vl_api_snat_static_mapping_details_t *rmp; + snat_main_t * sm = &snat_main; + + rmp = vl_msg_api_alloc (sizeof (*rmp)); + memset (rmp, 0, sizeof (*rmp)); + rmp->_vl_msg_id = ntohs (VL_API_SNAT_STATIC_MAPPING_DETAILS+sm->msg_id_base); + rmp->is_ip4 = 1; + rmp->addr_only = m->addr_only; + clib_memcpy (rmp->local_ip_address, &(m->local_addr), 4); + clib_memcpy (rmp->external_ip_address, &(m->external_addr), 4); + rmp->local_port = htons (m->local_port); + rmp->external_port = htons (m->external_port); + rmp->vrf_id = htonl (m->vrf_id); + rmp->context = context; + + vl_msg_api_send_shmem (q, (u8 *) & rmp); +} + +static void +vl_api_snat_static_mapping_dump_t_handler +(vl_api_snat_static_mapping_dump_t * mp) +{ + unix_shared_memory_queue_t *q; + snat_main_t * sm = &snat_main; + snat_static_mapping_t * m; + + q = vl_api_client_index_to_input_queue (mp->client_index); + if (q == 0) + return; + + pool_foreach (m, sm->static_mappings, + ({ + send_snat_static_mapping_details (m, q, mp->context); + })); +} + +static void *vl_api_snat_static_mapping_dump_t_print +(vl_api_snat_static_mapping_dump_t *mp, void * handle) +{ + u8 *s; + + s = format (0, "SCRIPT: snat_static_mapping_dump "); + + FINISH; +} + +static void +vl_api_snat_control_ping_t_handler +(vl_api_snat_control_ping_t * mp) +{ + vl_api_snat_control_ping_reply_t *rmp; + snat_main_t * sm = &snat_main; + int rv = 0; + + REPLY_MACRO2(VL_API_SNAT_CONTROL_PING_REPLY, + ({ + rmp->vpe_pid = ntohl (getpid()); + })); +} + +static void *vl_api_snat_control_ping_t_print +(vl_api_snat_control_ping_t *mp, void * handle) +{ + u8 *s; + + s = format (0, "SCRIPT: snat_control_ping "); + + FINISH; +} + +static void +vl_api_snat_show_config_t_handler +(vl_api_snat_show_config_t * mp) +{ + vl_api_snat_show_config_reply_t *rmp; + snat_main_t * sm = &snat_main; + int rv = 0; + + REPLY_MACRO2(VL_API_SNAT_SHOW_CONFIG_REPLY, + ({ + rmp->translation_buckets = htonl (sm->translation_buckets); + rmp->translation_memory_size = htonl (sm->translation_memory_size); + rmp->user_buckets = htonl (sm->user_buckets); + rmp->user_memory_size = htonl (sm->user_memory_size); + rmp->max_translations_per_user = htonl (sm->max_translations_per_user); + rmp->outside_vrf_id = htonl (sm->outside_vrf_id); + rmp->inside_vrf_id = htonl (sm->inside_vrf_id); + rmp->static_mapping_only = sm->static_mapping_only; + rmp->static_mapping_connection_tracking = + sm->static_mapping_connection_tracking; + })); +} + +static void *vl_api_snat_show_config_t_print +(vl_api_snat_show_config_t *mp, void * handle) +{ + u8 *s; + + s = format (0, "SCRIPT: snat_show_config "); + + FINISH; +} + +static void +vl_api_snat_set_workers_t_handler +(vl_api_snat_set_workers_t * mp) +{ + snat_main_t * sm = &snat_main; + vl_api_snat_set_workers_reply_t * rmp; + int rv = 0; + uword *bitmap = 0; + u64 mask = clib_net_to_host_u64 (mp->worker_mask); + + if (sm->num_workers < 2) + { + rv = VNET_API_ERROR_FEATURE_DISABLED; + goto send_reply; + } + + bitmap = clib_bitmap_set_multiple (bitmap, 0, mask, BITS (mask)); + rv = snat_set_workers(bitmap); + clib_bitmap_free (bitmap); + + send_reply: + REPLY_MACRO (VL_API_SNAT_SET_WORKERS_REPLY); +} + +static void *vl_api_snat_set_workers_t_print +(vl_api_snat_set_workers_t *mp, void * handle) +{ + u8 * s; + uword *bitmap = 0; + u8 first = 1; + int i; + u64 mask = clib_net_to_host_u64 (mp->worker_mask); + + s = format (0, "SCRIPT: snat_set_workers "); + bitmap = clib_bitmap_set_multiple (bitmap, 0, mask, BITS (mask)); + clib_bitmap_foreach (i, bitmap, + ({ + if (first) + s = format (s, "%d", i); + else + s = format (s, ",%d", i); + first = 0; + })); + clib_bitmap_free (bitmap); + FINISH; +} + +static void +send_snat_worker_details +(u32 worker_index, unix_shared_memory_queue_t * q, u32 context) +{ + vl_api_snat_worker_details_t *rmp; + snat_main_t * sm = &snat_main; + vlib_worker_thread_t *w = + vlib_worker_threads + worker_index + sm->first_worker_index; + + rmp = vl_msg_api_alloc (sizeof (*rmp)); + memset (rmp, 0, sizeof (*rmp)); + rmp->_vl_msg_id = ntohs (VL_API_SNAT_WORKER_DETAILS+sm->msg_id_base); + rmp->context = context; + rmp->worker_index = htonl (worker_index); + rmp->lcore_id = htonl (w->lcore_id); + strncpy ((char *) rmp->name, (char *) w->name, ARRAY_LEN (rmp->name) - 1); + + vl_msg_api_send_shmem (q, (u8 *) & rmp); +} + +static void +vl_api_snat_worker_dump_t_handler +(vl_api_snat_worker_dump_t * mp) +{ + unix_shared_memory_queue_t *q; + snat_main_t * sm = &snat_main; + u32 * worker_index; + + q = vl_api_client_index_to_input_queue (mp->client_index); + if (q == 0) + return; + + vec_foreach (worker_index, sm->workers) + { + send_snat_worker_details(*worker_index, q, mp->context); + } +} + +static void *vl_api_snat_worker_dump_t_print +(vl_api_snat_worker_dump_t *mp, void * handle) +{ + u8 *s; + + s = format (0, "SCRIPT: snat_worker_dump "); + + FINISH; +} + +/* List of message types that this plugin understands */ +#define foreach_snat_plugin_api_msg \ +_(SNAT_ADD_ADDRESS_RANGE, snat_add_address_range) \ +_(SNAT_INTERFACE_ADD_DEL_FEATURE, snat_interface_add_del_feature) \ +_(SNAT_ADD_STATIC_MAPPING, snat_add_static_mapping) \ +_(SNAT_CONTROL_PING, snat_control_ping) \ +_(SNAT_STATIC_MAPPING_DUMP, snat_static_mapping_dump) \ +_(SNAT_SHOW_CONFIG, snat_show_config) \ +_(SNAT_ADDRESS_DUMP, snat_address_dump) \ +_(SNAT_INTERFACE_DUMP, snat_interface_dump) \ +_(SNAT_SET_WORKERS, snat_set_workers) \ +_(SNAT_WORKER_DUMP, snat_worker_dump) + +/* Set up the API message handling tables */ +static clib_error_t * +snat_plugin_api_hookup (vlib_main_t *vm) +{ + snat_main_t * sm __attribute__ ((unused)) = &snat_main; +#define _(N,n) \ + vl_msg_api_set_handlers((VL_API_##N + sm->msg_id_base), \ + #n, \ + vl_api_##n##_t_handler, \ + vl_noop_handler, \ + vl_api_##n##_t_endian, \ + vl_api_##n##_t_print, \ + sizeof(vl_api_##n##_t), 1); + foreach_snat_plugin_api_msg; +#undef _ + + return 0; +} + +#define vl_msg_name_crc_list +#include +#undef vl_msg_name_crc_list + +static void +setup_message_id_table (snat_main_t * sm, api_main_t * am) +{ +#define _(id,n,crc) \ + vl_msg_api_add_msg_name_crc (am, #n "_" #crc, id + sm->msg_id_base); + foreach_vl_msg_name_crc_snat; +#undef _ +} + +static void plugin_custom_dump_configure (snat_main_t * sm) +{ +#define _(n,f) sm->api_main->msg_print_handlers \ + [VL_API_##n + sm->msg_id_base] \ + = (void *) vl_api_##f##_t_print; + foreach_snat_plugin_api_msg; +#undef _ +} + +static clib_error_t * snat_init (vlib_main_t * vm) +{ + snat_main_t * sm = &snat_main; + clib_error_t * error = 0; + ip4_main_t * im = &ip4_main; + ip_lookup_main_t * lm = &im->lookup_main; + u8 * name; + uword *p; + vlib_thread_registration_t *tr; + vlib_thread_main_t *tm = vlib_get_thread_main (); + uword *bitmap = 0; + u32 i; + + name = format (0, "snat_%08x%c", api_version, 0); + + /* Ask for a correctly-sized block of API message decode slots */ + sm->msg_id_base = vl_msg_api_get_msg_ids + ((char *) name, VL_MSG_FIRST_AVAILABLE); + + sm->vlib_main = vm; + sm->vnet_main = vnet_get_main(); + sm->ip4_main = im; + sm->ip4_lookup_main = lm; + sm->api_main = &api_main; + sm->first_worker_index = 0; + sm->next_worker = 0; + sm->num_workers = 0; + sm->workers = 0; + sm->fq_in2out_index = ~0; + sm->fq_out2in_index = ~0; + + p = hash_get_mem (tm->thread_registrations_by_name, "workers"); + if (p) + { + tr = (vlib_thread_registration_t *) p[0]; + if (tr) + { + sm->num_workers = tr->count; + sm->first_worker_index = tr->first_index; + } + } + + /* Use all available workers by default */ + if (sm->num_workers > 1) + { + for (i=0; i < sm->num_workers; i++) + bitmap = clib_bitmap_set (bitmap, i, 1); + snat_set_workers(bitmap); + clib_bitmap_free (bitmap); + } + + error = snat_plugin_api_hookup (vm); + + /* Add our API messages to the global name_crc hash table */ + setup_message_id_table (sm, &api_main); + + plugin_custom_dump_configure (sm); + vec_free(name); + + return error; +} + +VLIB_INIT_FUNCTION (snat_init); + +void snat_free_outside_address_and_port (snat_main_t * sm, + snat_session_key_t * k, + u32 address_index) +{ + snat_address_t *a; + u16 port_host_byte_order = clib_net_to_host_u16 (k->port); + + ASSERT (address_index < vec_len (sm->addresses)); + + a = sm->addresses + address_index; + + ASSERT (clib_bitmap_get_no_check (a->busy_port_bitmap, + port_host_byte_order) == 1); + + clib_bitmap_set_no_check (a->busy_port_bitmap, port_host_byte_order, 0); + a->busy_ports--; +} + +/** + * @brief Match SNAT static mapping. + * + * @param sm SNAT main. + * @param match Address and port to match. + * @param mapping External or local address and port of the matched mapping. + * @param by_external If 0 match by local address otherwise match by external + * address. + * + * @returns 0 if match found otherwise 1. + */ +int snat_static_mapping_match (snat_main_t * sm, + snat_session_key_t match, + snat_session_key_t * mapping, + u8 by_external) +{ + clib_bihash_kv_8_8_t kv, value; + snat_static_mapping_t *m; + snat_static_mapping_key_t m_key; + clib_bihash_8_8_t *mapping_hash = &sm->static_mapping_by_local; + + if (by_external) + mapping_hash = &sm->static_mapping_by_external; + + m_key.addr = match.addr; + m_key.port = clib_net_to_host_u16 (match.port); + m_key.fib_index = match.fib_index; + + kv.key = m_key.as_u64; + + if (clib_bihash_search_8_8 (mapping_hash, &kv, &value)) + { + /* Try address only mapping */ + m_key.port = 0; + kv.key = m_key.as_u64; + if (clib_bihash_search_8_8 (mapping_hash, &kv, &value)) + return 1; + } + + m = pool_elt_at_index (sm->static_mappings, value.value); + + if (by_external) + { + mapping->addr = m->local_addr; + /* Address only mapping doesn't change port */ + mapping->port = m->addr_only ? match.port + : clib_host_to_net_u16 (m->local_port); + mapping->fib_index = m->fib_index; + } + else + { + mapping->addr = m->external_addr; + /* Address only mapping doesn't change port */ + mapping->port = m->addr_only ? match.port + : clib_host_to_net_u16 (m->external_port); + mapping->fib_index = sm->outside_fib_index; + } + + return 0; +} + +int snat_alloc_outside_address_and_port (snat_main_t * sm, + snat_session_key_t * k, + u32 * address_indexp) +{ + int i; + snat_address_t *a; + u32 portnum; + + for (i = 0; i < vec_len (sm->addresses); i++) + { + if (sm->addresses[i].busy_ports < (65535-1024)) + { + a = sm->addresses + i; + + while (1) + { + portnum = random_u32 (&sm->random_seed); + portnum &= 0xFFFF; + if (portnum < 1024) + continue; + if (clib_bitmap_get_no_check (a->busy_port_bitmap, portnum)) + continue; + clib_bitmap_set_no_check (a->busy_port_bitmap, portnum, 1); + a->busy_ports++; + /* Caller sets protocol and fib index */ + k->addr = a->addr; + k->port = clib_host_to_net_u16(portnum); + *address_indexp = i; + return 0; + } + } + } + /* Totally out of translations to use... */ + return 1; +} + + +static clib_error_t * +add_address_command_fn (vlib_main_t * vm, + unformat_input_t * input, + vlib_cli_command_t * cmd) +{ + unformat_input_t _line_input, *line_input = &_line_input; + snat_main_t * sm = &snat_main; + ip4_address_t start_addr, end_addr, this_addr; + u32 start_host_order, end_host_order; + int i, count; + int is_add = 1; + int rv = 0; + + /* Get a line of input. */ + if (!unformat_user (input, unformat_line_input, line_input)) + return 0; + + while (unformat_check_input (line_input) != UNFORMAT_END_OF_INPUT) + { + if (unformat (line_input, "%U - %U", + unformat_ip4_address, &start_addr, + unformat_ip4_address, &end_addr)) + ; + else if (unformat (line_input, "%U", unformat_ip4_address, &start_addr)) + end_addr = start_addr; + else if (unformat (line_input, "del")) + is_add = 0; + else + return clib_error_return (0, "unknown input '%U'", + format_unformat_error, input); + } + unformat_free (line_input); + + if (sm->static_mapping_only) + return clib_error_return (0, "static mapping only mode"); + + start_host_order = clib_host_to_net_u32 (start_addr.as_u32); + end_host_order = clib_host_to_net_u32 (end_addr.as_u32); + + if (end_host_order < start_host_order) + return clib_error_return (0, "end address less than start address"); + + count = (end_host_order - start_host_order) + 1; + + if (count > 1024) + clib_warning ("%U - %U, %d addresses...", + format_ip4_address, &start_addr, + format_ip4_address, &end_addr, + count); + + this_addr = start_addr; + + for (i = 0; i < count; i++) + { + if (is_add) + snat_add_address (sm, &this_addr); + else + rv = snat_del_address (sm, this_addr); + + switch (rv) + { + case VNET_API_ERROR_NO_SUCH_ENTRY: + return clib_error_return (0, "S-NAT address not exist."); + break; + case VNET_API_ERROR_UNSPECIFIED: + return clib_error_return (0, "S-NAT address used in static mapping."); + break; + default: + break; + } + + increment_v4_address (&this_addr); + } + + return 0; +} + +VLIB_CLI_COMMAND (add_address_command, static) = { + .path = "snat add address", + .short_help = "snat add addresses [- ] [del]", + .function = add_address_command_fn, +}; + +static clib_error_t * +snat_feature_command_fn (vlib_main_t * vm, + unformat_input_t * input, + vlib_cli_command_t * cmd) +{ + unformat_input_t _line_input, *line_input = &_line_input; + vnet_main_t * vnm = vnet_get_main(); + clib_error_t * error = 0; + u32 sw_if_index; + u32 * inside_sw_if_indices = 0; + u32 * outside_sw_if_indices = 0; + int is_del = 0; + int i; + + sw_if_index = ~0; + + /* Get a line of input. */ + if (!unformat_user (input, unformat_line_input, line_input)) + return 0; + + while (unformat_check_input (line_input) != UNFORMAT_END_OF_INPUT) + { + if (unformat (line_input, "in %U", unformat_vnet_sw_interface, + vnm, &sw_if_index)) + vec_add1 (inside_sw_if_indices, sw_if_index); + else if (unformat (line_input, "out %U", unformat_vnet_sw_interface, + vnm, &sw_if_index)) + vec_add1 (outside_sw_if_indices, sw_if_index); + else if (unformat (line_input, "del")) + is_del = 1; + else + return clib_error_return (0, "unknown input '%U'", + format_unformat_error, input); + } + unformat_free (line_input); + + if (vec_len (inside_sw_if_indices)) + { + for (i = 0; i < vec_len(inside_sw_if_indices); i++) + { + sw_if_index = inside_sw_if_indices[i]; + snat_interface_add_del (sw_if_index, 1, is_del); + } + } + + if (vec_len (outside_sw_if_indices)) + { + for (i = 0; i < vec_len(outside_sw_if_indices); i++) + { + sw_if_index = outside_sw_if_indices[i]; + snat_interface_add_del (sw_if_index, 0, is_del); + } + } + + vec_free (inside_sw_if_indices); + vec_free (outside_sw_if_indices); + + return error; +} + +VLIB_CLI_COMMAND (set_interface_snat_command, static) = { + .path = "set interface snat", + .function = snat_feature_command_fn, + .short_help = "set interface snat in out [del]", +}; + +static clib_error_t * +add_static_mapping_command_fn (vlib_main_t * vm, + unformat_input_t * input, + vlib_cli_command_t * cmd) +{ + unformat_input_t _line_input, *line_input = &_line_input; + clib_error_t * error = 0; + ip4_address_t l_addr, e_addr; + u32 l_port = 0, e_port = 0, vrf_id = ~0; + int is_add = 1; + int addr_only = 1; + int rv; + + /* Get a line of input. */ + if (!unformat_user (input, unformat_line_input, line_input)) + return 0; + + while (unformat_check_input (line_input) != UNFORMAT_END_OF_INPUT) + { + if (unformat (line_input, "local %U %u", unformat_ip4_address, &l_addr, + &l_port)) + addr_only = 0; + else if (unformat (line_input, "local %U", unformat_ip4_address, &l_addr)) + ; + else if (unformat (line_input, "external %U %u", unformat_ip4_address, + &e_addr, &e_port)) + addr_only = 0; + else if (unformat (line_input, "external %U", unformat_ip4_address, + &e_addr)) + ; + else if (unformat (line_input, "vrf %u", &vrf_id)) + ; + else if (unformat (line_input, "del")) + is_add = 0; + else + return clib_error_return (0, "unknown input: '%U'", + format_unformat_error, line_input); + } + unformat_free (line_input); + + rv = snat_add_static_mapping(l_addr, e_addr, (u16) l_port, (u16) e_port, + vrf_id, addr_only, is_add); + + switch (rv) + { + case VNET_API_ERROR_INVALID_VALUE: + return clib_error_return (0, "External port already in use."); + break; + case VNET_API_ERROR_NO_SUCH_ENTRY: + if (is_add) + return clib_error_return (0, "External addres must be allocated."); + else + return clib_error_return (0, "Mapping not exist."); + break; + case VNET_API_ERROR_NO_SUCH_FIB: + return clib_error_return (0, "No such VRF id."); + case VNET_API_ERROR_VALUE_EXIST: + return clib_error_return (0, "Mapping already exist."); + default: + break; + } + + return error; +} + +/*? + * @cliexpar + * @cliexstart{snat add static mapping} + * Static mapping allows hosts on the external network to initiate connection + * to to the local network host. + * To create static mapping between local host address 10.0.0.3 port 6303 and + * external address 4.4.4.4 port 3606 use: + * vpp# snat add static mapping local 10.0.0.3 6303 external 4.4.4.4 3606 + * If not runnig "static mapping only" S-NAT plugin mode use before: + * vpp# snat add address 4.4.4.4 + * To create static mapping between local and external address use: + * vpp# snat add static mapping local 10.0.0.3 external 4.4.4.4 + * @cliexend +?*/ +VLIB_CLI_COMMAND (add_static_mapping_command, static) = { + .path = "snat add static mapping", + .function = add_static_mapping_command_fn, + .short_help = + "snat add static mapping local [] external [] [vrf ] [del]", +}; + +static clib_error_t * +set_workers_command_fn (vlib_main_t * vm, + unformat_input_t * input, + vlib_cli_command_t * cmd) +{ + unformat_input_t _line_input, *line_input = &_line_input; + uword *bitmap = 0; + int rv = 0; + + /* Get a line of input. */ + if (!unformat_user (input, unformat_line_input, line_input)) + return 0; + + while (unformat_check_input (line_input) != UNFORMAT_END_OF_INPUT) + { + if (unformat (line_input, "%U", unformat_bitmap_list, &bitmap)) + ; + else + return clib_error_return (0, "unknown input '%U'", + format_unformat_error, input); + } + unformat_free (line_input); + + if (bitmap == 0) + return clib_error_return (0, "List of workers must be specified."); + + rv = snat_set_workers(bitmap); + + clib_bitmap_free (bitmap); + + switch (rv) + { + case VNET_API_ERROR_INVALID_WORKER: + return clib_error_return (0, "Invalid worker(s)."); + break; + case VNET_API_ERROR_FEATURE_DISABLED: + return clib_error_return (0, + "Supported only if 2 or more workes available."); + break; + default: + break; + } + + return 0; +} + +/*? + * @cliexpar + * @cliexstart{set snat workers} + * Set SNAT workers if 2 or more workers available, use: + * vpp# set snat workers 0-2,5 + * @cliexend +?*/ +VLIB_CLI_COMMAND (set_workers_command, static) = { + .path = "set snat workers", + .function = set_workers_command_fn, + .short_help = + "set snat workers ", +}; + +static clib_error_t * +snat_config (vlib_main_t * vm, unformat_input_t * input) +{ + snat_main_t * sm = &snat_main; + u32 translation_buckets = 1024; + u32 translation_memory_size = 128<<20; + u32 user_buckets = 128; + u32 user_memory_size = 64<<20; + u32 max_translations_per_user = 100; + u32 outside_vrf_id = 0; + u32 inside_vrf_id = 0; + u32 static_mapping_buckets = 1024; + u32 static_mapping_memory_size = 64<<20; + u8 static_mapping_only = 0; + u8 static_mapping_connection_tracking = 0; + vlib_thread_main_t *tm = vlib_get_thread_main (); + + while (unformat_check_input (input) != UNFORMAT_END_OF_INPUT) + { + if (unformat (input, "translation hash buckets %d", &translation_buckets)) + ; + else if (unformat (input, "translation hash memory %d", + &translation_memory_size)); + else if (unformat (input, "user hash buckets %d", &user_buckets)) + ; + else if (unformat (input, "user hash memory %d", + &user_memory_size)) + ; + else if (unformat (input, "max translations per user %d", + &max_translations_per_user)) + ; + else if (unformat (input, "outside VRF id %d", + &outside_vrf_id)) + ; + else if (unformat (input, "inside VRF id %d", + &inside_vrf_id)) + ; + else if (unformat (input, "static mapping only")) + { + static_mapping_only = 1; + if (unformat (input, "connection tracking")) + static_mapping_connection_tracking = 1; + } + else + return clib_error_return (0, "unknown input '%U'", + format_unformat_error, input); + } + + /* for show commands, etc. */ + sm->translation_buckets = translation_buckets; + sm->translation_memory_size = translation_memory_size; + sm->user_buckets = user_buckets; + sm->user_memory_size = user_memory_size; + sm->max_translations_per_user = max_translations_per_user; + sm->outside_vrf_id = outside_vrf_id; + sm->outside_fib_index = ~0; + sm->inside_vrf_id = inside_vrf_id; + sm->inside_fib_index = ~0; + sm->static_mapping_only = static_mapping_only; + sm->static_mapping_connection_tracking = static_mapping_connection_tracking; + + if (!static_mapping_only || + (static_mapping_only && static_mapping_connection_tracking)) + { + clib_bihash_init_8_8 (&sm->worker_by_in, "worker-by-in", user_buckets, + user_memory_size); + + clib_bihash_init_8_8 (&sm->worker_by_out, "worker-by-out", user_buckets, + user_memory_size); + + vec_validate (sm->per_thread_data, tm->n_vlib_mains - 1); + + clib_bihash_init_8_8 (&sm->in2out, "in2out", translation_buckets, + translation_memory_size); + + clib_bihash_init_8_8 (&sm->out2in, "out2in", translation_buckets, + translation_memory_size); + + clib_bihash_init_8_8 (&sm->user_hash, "users", user_buckets, + user_memory_size); + } + clib_bihash_init_8_8 (&sm->static_mapping_by_local, + "static_mapping_by_local", static_mapping_buckets, + static_mapping_memory_size); + + clib_bihash_init_8_8 (&sm->static_mapping_by_external, + "static_mapping_by_external", static_mapping_buckets, + static_mapping_memory_size); + return 0; +} + +VLIB_CONFIG_FUNCTION (snat_config, "snat"); + +u8 * format_snat_key (u8 * s, va_list * args) +{ + snat_session_key_t * key = va_arg (*args, snat_session_key_t *); + char * protocol_string = "unknown"; + static char *protocol_strings[] = { + "UDP", + "TCP", + "ICMP", + }; + + if (key->protocol < ARRAY_LEN(protocol_strings)) + protocol_string = protocol_strings[key->protocol]; + + s = format (s, "%U proto %s port %d fib %d", + format_ip4_address, &key->addr, protocol_string, + clib_net_to_host_u16 (key->port), key->fib_index); + return s; +} + +u8 * format_snat_session (u8 * s, va_list * args) +{ + snat_main_t * sm __attribute__((unused)) = va_arg (*args, snat_main_t *); + snat_session_t * sess = va_arg (*args, snat_session_t *); + + s = format (s, " i2o %U\n", format_snat_key, &sess->in2out); + s = format (s, " o2i %U\n", format_snat_key, &sess->out2in); + s = format (s, " last heard %.2f\n", sess->last_heard); + s = format (s, " total pkts %d, total bytes %lld\n", + sess->total_pkts, sess->total_bytes); + if (snat_is_session_static (sess)) + s = format (s, " static translation\n"); + else + s = format (s, " dynamic translation\n"); + + return s; +} + +u8 * format_snat_user (u8 * s, va_list * args) +{ + snat_main_per_thread_data_t * sm = va_arg (*args, snat_main_per_thread_data_t *); + snat_user_t * u = va_arg (*args, snat_user_t *); + int verbose = va_arg (*args, int); + dlist_elt_t * head, * elt; + u32 elt_index, head_index; + u32 session_index; + snat_session_t * sess; + + s = format (s, "%U: %d dynamic translations, %d static translations\n", + format_ip4_address, &u->addr, u->nsessions, u->nstaticsessions); + + if (verbose == 0) + return s; + + if (u->nsessions || u->nstaticsessions) + { + head_index = u->sessions_per_user_list_head_index; + head = pool_elt_at_index (sm->list_pool, head_index); + + elt_index = head->next; + elt = pool_elt_at_index (sm->list_pool, elt_index); + session_index = elt->value; + + while (session_index != ~0) + { + sess = pool_elt_at_index (sm->sessions, session_index); + + s = format (s, " %U\n", format_snat_session, sm, sess); + + elt_index = elt->next; + elt = pool_elt_at_index (sm->list_pool, elt_index); + session_index = elt->value; + } + } + + return s; +} + +u8 * format_snat_static_mapping (u8 * s, va_list * args) +{ + snat_static_mapping_t *m = va_arg (*args, snat_static_mapping_t *); + + if (m->addr_only) + s = format (s, "local %U external %U vrf %d", + format_ip4_address, &m->local_addr, + format_ip4_address, &m->external_addr, + m->vrf_id); + else + s = format (s, "local %U:%d external %U:%d vrf %d", + format_ip4_address, &m->local_addr, m->local_port, + format_ip4_address, &m->external_addr, m->external_port, + m->vrf_id); + + return s; +} + +static clib_error_t * +show_snat_command_fn (vlib_main_t * vm, + unformat_input_t * input, + vlib_cli_command_t * cmd) +{ + int verbose = 0; + snat_main_t * sm = &snat_main; + snat_user_t * u; + snat_static_mapping_t *m; + snat_interface_t *i; + snat_address_t * ap; + vnet_main_t *vnm = vnet_get_main(); + snat_main_per_thread_data_t *tsm; + u32 users_num = 0, sessions_num = 0, *worker; + uword j = 0; + + if (unformat (input, "detail")) + verbose = 1; + else if (unformat (input, "verbose")) + verbose = 2; + + if (sm->static_mapping_only) + { + if (sm->static_mapping_connection_tracking) + vlib_cli_output (vm, "SNAT mode: static mapping only connection " + "tracking"); + else + vlib_cli_output (vm, "SNAT mode: static mapping only"); + } + else + { + vlib_cli_output (vm, "SNAT mode: dynamic translations enabled"); + } + + if (verbose > 0) + { + pool_foreach (i, sm->interfaces, + ({ + vlib_cli_output (vm, "%U %s", format_vnet_sw_interface_name, vnm, + vnet_get_sw_interface (vnm, i->sw_if_index), + i->is_inside ? "in" : "out"); + })); + + vec_foreach (ap, sm->addresses) + { + u8 * s = format (0, ""); + vlib_cli_output (vm, "%U", format_ip4_address, &ap->addr); + clib_bitmap_foreach (j, ap->busy_port_bitmap, + ({ + s = format (s, " %d", j); + })); + vlib_cli_output (vm, " %d busy ports:%v", ap->busy_ports, s); + } + } + + if (sm->num_workers > 1) + { + vlib_cli_output (vm, "%d workers", vec_len (sm->workers)); + if (verbose > 0) + { + vec_foreach (worker, sm->workers) + { + vlib_worker_thread_t *w = + vlib_worker_threads + *worker + sm->first_worker_index; + vlib_cli_output (vm, " %v", w->name); + } + } + } + + if (sm->static_mapping_only && !(sm->static_mapping_connection_tracking)) + { + vlib_cli_output (vm, "%d static mappings", + pool_elts (sm->static_mappings)); + + if (verbose > 0) + { + pool_foreach (m, sm->static_mappings, + ({ + vlib_cli_output (vm, "%U", format_snat_static_mapping, m); + })); + } + } + else + { + vec_foreach (tsm, sm->per_thread_data) + { + users_num += pool_elts (tsm->users); + sessions_num += pool_elts (tsm->sessions); + } + + vlib_cli_output (vm, "%d users, %d outside addresses, %d active sessions," + " %d static mappings", + users_num, + vec_len (sm->addresses), + sessions_num, + pool_elts (sm->static_mappings)); + + if (verbose > 0) + { + vlib_cli_output (vm, "%U", format_bihash_8_8, &sm->in2out, + verbose - 1); + vlib_cli_output (vm, "%U", format_bihash_8_8, &sm->out2in, + verbose - 1); + vlib_cli_output (vm, "%U", format_bihash_8_8, &sm->worker_by_in, + verbose - 1); + vlib_cli_output (vm, "%U", format_bihash_8_8, &sm->worker_by_out, + verbose - 1); + vec_foreach_index (j, sm->per_thread_data) + { + tsm = vec_elt_at_index (sm->per_thread_data, j); + + if (pool_elts (tsm->users) == 0) + continue; + + vlib_worker_thread_t *w = vlib_worker_threads + j; + vlib_cli_output (vm, "Thread %d (%v at lcore %u):", j, w->name, + w->lcore_id); + vlib_cli_output (vm, " %d list pool elements", + pool_elts (tsm->list_pool)); + + pool_foreach (u, tsm->users, + ({ + vlib_cli_output (vm, " %U", format_snat_user, tsm, u, + verbose - 1); + })); + } + + if (pool_elts (sm->static_mappings)) + { + vlib_cli_output (vm, "static mappings:"); + pool_foreach (m, sm->static_mappings, + ({ + vlib_cli_output (vm, "%U", format_snat_static_mapping, m); + })); + } + } + } + + return 0; +} + +VLIB_CLI_COMMAND (show_snat_command, static) = { + .path = "show snat", + .short_help = "show snat", + .function = show_snat_command_fn, +}; diff --git a/src/plugins/snat/snat.h b/src/plugins/snat/snat.h new file mode 100644 index 00000000..cb31dc51 --- /dev/null +++ b/src/plugins/snat/snat.h @@ -0,0 +1,259 @@ + +/* + * snat.h - simple nat definitions + * + * Copyright (c) 2016 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ +#ifndef __included_snat_h__ +#define __included_snat_h__ + +#include +#include +#include +#include +#include +#include +#include +#include +#include + +/* Key */ +typedef struct { + union + { + struct + { + ip4_address_t addr; + u16 port; + u16 protocol:3, + fib_index:13; + }; + u64 as_u64; + }; +} snat_session_key_t; + +typedef struct { + union + { + struct + { + ip4_address_t addr; + u32 fib_index; + }; + u64 as_u64; + }; +} snat_user_key_t; + +typedef struct { + union + { + struct + { + ip4_address_t addr; + u16 port; + u16 fib_index; + }; + u64 as_u64; + }; +} snat_static_mapping_key_t; + + +typedef enum { + SNAT_PROTOCOL_UDP = 0, + SNAT_PROTOCOL_TCP, + SNAT_PROTOCOL_ICMP, +} snat_protocol_t; + + +#define SNAT_SESSION_FLAG_STATIC_MAPPING 1 + +typedef CLIB_PACKED(struct { + snat_session_key_t out2in; /* 0-15 */ + + snat_session_key_t in2out; /* 16-31 */ + + u32 flags; /* 32-35 */ + + /* per-user translations */ + u32 per_user_index; /* 36-39 */ + + u32 per_user_list_head_index; /* 40-43 */ + + /* Last heard timer */ + f64 last_heard; /* 44-51 */ + + u64 total_bytes; /* 52-59 */ + + u32 total_pkts; /* 60-63 */ + + /* Outside address */ + u32 outside_address_index; /* 64-67 */ + +}) snat_session_t; + + +typedef struct { + ip4_address_t addr; + u32 sessions_per_user_list_head_index; + u32 nsessions; + u32 nstaticsessions; +} snat_user_t; + +typedef struct { + ip4_address_t addr; + u32 busy_ports; + uword * busy_port_bitmap; +} snat_address_t; + +typedef struct { + ip4_address_t local_addr; + ip4_address_t external_addr; + u16 local_port; + u16 external_port; + u8 addr_only; + u32 vrf_id; + u32 fib_index; +} snat_static_mapping_t; + +typedef struct { + u32 sw_if_index; + u8 is_inside; +} snat_interface_t; + +typedef struct { + /* User pool */ + snat_user_t * users; + + /* Session pool */ + snat_session_t * sessions; + + /* Pool of doubly-linked list elements */ + dlist_elt_t * list_pool; +} snat_main_per_thread_data_t; + +typedef struct { + /* Main lookup tables */ + clib_bihash_8_8_t out2in; + clib_bihash_8_8_t in2out; + + /* Find-a-user => src address lookup */ + clib_bihash_8_8_t user_hash; + + /* Non-translated packets worker lookup => src address + VRF */ + clib_bihash_8_8_t worker_by_in; + + /* Translated packets worker lookup => IP address + port number */ + clib_bihash_8_8_t worker_by_out; + + u32 num_workers; + u32 first_worker_index; + u32 next_worker; + u32 * workers; + + /* Per thread data */ + snat_main_per_thread_data_t * per_thread_data; + + /* Find a static mapping by local */ + clib_bihash_8_8_t static_mapping_by_local; + + /* Find a static mapping by external */ + clib_bihash_8_8_t static_mapping_by_external; + + /* Static mapping pool */ + snat_static_mapping_t * static_mappings; + + /* Interface pool */ + snat_interface_t * interfaces; + + /* Vector of outside addresses */ + snat_address_t * addresses; + + /* Randomize port allocation order */ + u32 random_seed; + + /* Worker handoff index */ + u32 fq_in2out_index; + u32 fq_out2in_index; + + /* Config parameters */ + u8 static_mapping_only; + u8 static_mapping_connection_tracking; + u32 translation_buckets; + u32 translation_memory_size; + u32 user_buckets; + u32 user_memory_size; + u32 max_translations_per_user; + u32 outside_vrf_id; + u32 outside_fib_index; + u32 inside_vrf_id; + u32 inside_fib_index; + + /* API message ID base */ + u16 msg_id_base; + + /* convenience */ + vlib_main_t * vlib_main; + vnet_main_t * vnet_main; + ip4_main_t * ip4_main; + ip_lookup_main_t * ip4_lookup_main; + ethernet_main_t * ethernet_main; + api_main_t * api_main; +} snat_main_t; + +extern snat_main_t snat_main; +extern vlib_node_registration_t snat_in2out_node; +extern vlib_node_registration_t snat_out2in_node; +extern vlib_node_registration_t snat_in2out_fast_node; +extern vlib_node_registration_t snat_out2in_fast_node; +extern vlib_node_registration_t snat_in2out_worker_handoff_node; +extern vlib_node_registration_t snat_out2in_worker_handoff_node; + +void snat_free_outside_address_and_port (snat_main_t * sm, + snat_session_key_t * k, + u32 address_index); + +int snat_alloc_outside_address_and_port (snat_main_t * sm, + snat_session_key_t * k, + u32 * address_indexp); + +int snat_static_mapping_match (snat_main_t * sm, + snat_session_key_t match, + snat_session_key_t * mapping, + u8 by_external); + +format_function_t format_snat_user; + +typedef struct { + u32 cached_sw_if_index; + u32 cached_ip4_address; +} snat_runtime_t; + +/** \brief Check if SNAT session is created from static mapping. + @param s SNAT session + @return 1 if SNAT session is created from static mapping otherwise 0 +*/ +#define snat_is_session_static(s) s->flags & SNAT_SESSION_FLAG_STATIC_MAPPING + +/* + * Why is this here? Because we don't need to touch this layer to + * simply reply to an icmp. We need to change id to a unique + * value to NAT an echo request/reply. + */ + +typedef struct { + u16 identifier; + u16 sequence; +} icmp_echo_header_t; + +#endif /* __included_snat_h__ */ diff --git a/src/plugins/snat/snat_all_api_h.h b/src/plugins/snat/snat_all_api_h.h new file mode 100644 index 00000000..49017700 --- /dev/null +++ b/src/plugins/snat/snat_all_api_h.h @@ -0,0 +1,19 @@ + +/* + * snat_all_api_h.h - skeleton vpp engine plug-in api #include file + * + * Copyright (c) + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ +/* Include the generated file, see BUILT_SOURCES in Makefile.am */ +#include diff --git a/src/plugins/snat/snat_msg_enum.h b/src/plugins/snat/snat_msg_enum.h new file mode 100644 index 00000000..2c76fd51 --- /dev/null +++ b/src/plugins/snat/snat_msg_enum.h @@ -0,0 +1,31 @@ + +/* + * snat_msg_enum.h - skeleton vpp engine plug-in message enumeration + * + * Copyright (c) + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ +#ifndef included_snat_msg_enum_h +#define included_snat_msg_enum_h + +#include + +#define vl_msg_id(n,h) n, +typedef enum { +#include + /* We'll want to know how many messages IDs we need... */ + VL_MSG_FIRST_AVAILABLE, +} vl_msg_id_t; +#undef vl_msg_id + +#endif /* included_snat_msg_enum_h */ diff --git a/src/plugins/snat/snat_test.c b/src/plugins/snat/snat_test.c new file mode 100644 index 00000000..2a003ba6 --- /dev/null +++ b/src/plugins/snat/snat_test.c @@ -0,0 +1,602 @@ + +/* + * snat.c - skeleton vpp-api-test plug-in + * + * Copyright (c) + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ +#include +#include +#include +#include +#include +#include + +uword unformat_sw_if_index (unformat_input_t * input, va_list * args); + +/* Declare message IDs */ +#include + +/* define message structures */ +#define vl_typedefs +#include +#undef vl_typedefs + +/* declare message handlers for each api */ + +#define vl_endianfun /* define message structures */ +#include +#undef vl_endianfun + +/* instantiate all the print functions we know about */ +#define vl_print(handle, ...) +#define vl_printfun +#include +#undef vl_printfun + +/* Get the API version number. */ +#define vl_api_version(n,v) static u32 api_version=(v); +#include +#undef vl_api_version + +typedef struct { + /* API message ID base */ + u16 msg_id_base; + vat_main_t *vat_main; +} snat_test_main_t; + +snat_test_main_t snat_test_main; + +#define foreach_standard_reply_retval_handler \ +_(snat_add_address_range_reply) \ +_(snat_interface_add_del_feature_reply) \ +_(snat_add_static_mapping_reply) \ +_(snat_set_workers_reply) + +#define _(n) \ + static void vl_api_##n##_t_handler \ + (vl_api_##n##_t * mp) \ + { \ + vat_main_t * vam = snat_test_main.vat_main; \ + i32 retval = ntohl(mp->retval); \ + if (vam->async_mode) { \ + vam->async_errors += (retval < 0); \ + } else { \ + vam->retval = retval; \ + vam->result_ready = 1; \ + } \ + } +foreach_standard_reply_retval_handler; +#undef _ + +/* + * Table of message reply handlers, must include boilerplate handlers + * we just generated + */ +#define foreach_vpe_api_reply_msg \ +_(SNAT_ADD_ADDRESS_RANGE_REPLY, snat_add_address_range_reply) \ +_(SNAT_INTERFACE_ADD_DEL_FEATURE_REPLY, \ + snat_interface_add_del_feature_reply) \ +_(SNAT_ADD_STATIC_MAPPING_REPLY, snat_add_static_mapping_reply) \ +_(SNAT_CONTROL_PING_REPLY, snat_control_ping_reply) \ +_(SNAT_STATIC_MAPPING_DETAILS, snat_static_mapping_details) \ +_(SNAT_SHOW_CONFIG_REPLY, snat_show_config_reply) \ +_(SNAT_ADDRESS_DETAILS, snat_address_details) \ +_(SNAT_INTERFACE_DETAILS, snat_interface_details) \ +_(SNAT_SET_WORKERS_REPLY, snat_set_workers_reply) \ +_(SNAT_WORKER_DETAILS, snat_worker_details) + +/* M: construct, but don't yet send a message */ +#define M(T,t) \ +do { \ + vam->result_ready = 0; \ + mp = vl_msg_api_alloc(sizeof(*mp)); \ + memset (mp, 0, sizeof (*mp)); \ + mp->_vl_msg_id = ntohs (VL_API_##T + sm->msg_id_base); \ + mp->client_index = vam->my_client_index; \ +} while(0); + +#define M2(T,t,n) \ +do { \ + vam->result_ready = 0; \ + mp = vl_msg_api_alloc(sizeof(*mp)+(n)); \ + memset (mp, 0, sizeof (*mp)); \ + mp->_vl_msg_id = ntohs (VL_API_##T + sm->msg_id_base); \ + mp->client_index = vam->my_client_index; \ +} while(0); + +/* S: send a message */ +#define S (vl_msg_api_send_shmem (vam->vl_input_queue, (u8 *)&mp)) + +/* W: wait for results, with timeout */ +#define W \ +do { \ + timeout = vat_time_now (vam) + 1.0; \ + \ + while (vat_time_now (vam) < timeout) { \ + if (vam->result_ready == 1) { \ + return (vam->retval); \ + } \ + } \ + return -99; \ +} while(0); + +static int api_snat_add_address_range (vat_main_t * vam) +{ + snat_test_main_t * sm = &snat_test_main; + unformat_input_t * i = vam->input; + f64 timeout; + ip4_address_t start_addr, end_addr; + u32 start_host_order, end_host_order; + vl_api_snat_add_address_range_t * mp; + u8 is_add = 1; + int count; + + while (unformat_check_input (i) != UNFORMAT_END_OF_INPUT) + { + if (unformat (i, "%U - %U", + unformat_ip4_address, &start_addr, + unformat_ip4_address, &end_addr)) + ; + else if (unformat (i, "%U", unformat_ip4_address, &start_addr)) + end_addr = start_addr; + else if (unformat (i, "del")) + is_add = 0; + else + { + clib_warning("unknown input '%U'", format_unformat_error, i); + return -99; + } + } + + start_host_order = clib_host_to_net_u32 (start_addr.as_u32); + end_host_order = clib_host_to_net_u32 (end_addr.as_u32); + + if (end_host_order < start_host_order) + { + errmsg ("end address less than start address\n"); + return -99; + } + + count = (end_host_order - start_host_order) + 1; + + if (count > 1024) + { + errmsg ("%U - %U, %d addresses...\n", + format_ip4_address, &start_addr, + format_ip4_address, &end_addr, + count); + } + + M(SNAT_ADD_ADDRESS_RANGE, snat_add_address_range); + + memcpy (mp->first_ip_address, &start_addr, 4); + memcpy (mp->last_ip_address, &end_addr, 4); + mp->is_ip4 = 1; + mp->is_add = is_add; + + S; W; + + /* NOTREACHED */ + return 0; +} + +static int api_snat_interface_add_del_feature (vat_main_t * vam) +{ + snat_test_main_t * sm = &snat_test_main; + unformat_input_t * i = vam->input; + f64 timeout; + vl_api_snat_interface_add_del_feature_t * mp; + u32 sw_if_index; + u8 sw_if_index_set = 0; + u8 is_inside = 1; + u8 is_add = 1; + + while (unformat_check_input (i) != UNFORMAT_END_OF_INPUT) + { + if (unformat (i, "%U", unformat_sw_if_index, vam, &sw_if_index)) + sw_if_index_set = 1; + else if (unformat (i, "sw_if_index %d", &sw_if_index)) + sw_if_index_set = 1; + else if (unformat (i, "out")) + is_inside = 0; + else if (unformat (i, "in")) + is_inside = 1; + else if (unformat (i, "del")) + is_add = 0; + else + { + clib_warning("unknown input '%U'", format_unformat_error, i); + return -99; + } + } + + if (sw_if_index_set == 0) + { + errmsg ("interface / sw_if_index required\n"); + return -99; + } + + M(SNAT_INTERFACE_ADD_DEL_FEATURE, snat_interface_add_del_feature); + mp->sw_if_index = ntohl(sw_if_index); + mp->is_add = is_add; + mp->is_inside = is_inside; + + S; W; + /* NOTREACHED */ + return 0; +} + +static int api_snat_add_static_mapping(vat_main_t * vam) +{ + snat_test_main_t * sm = &snat_test_main; + unformat_input_t * i = vam->input; + f64 timeout; + vl_api_snat_add_static_mapping_t * mp; + u8 addr_set_n = 0; + u8 is_add = 1; + u8 addr_only = 1; + ip4_address_t local_addr, external_addr; + u32 local_port = 0, external_port = 0, vrf_id = ~0; + + while (unformat_check_input (i) != UNFORMAT_END_OF_INPUT) + { + if (unformat (i, "local_addr %U", unformat_ip4_address, &local_addr)) + addr_set_n++; + else if (unformat (i, "external_addr %U", unformat_ip4_address, + &external_addr)) + addr_set_n++; + else if (unformat (i, "local_port %u", &local_port)) + addr_only = 0; + else if (unformat (i, "external_port %u", &external_port)) + addr_only = 0; + else if (unformat (i, "vrf %u", &vrf_id)) + ; + else if (unformat (i, "del")) + is_add = 0; + else + { + clib_warning("unknown input '%U'", format_unformat_error, i); + return -99; + } + } + + if (addr_set_n != 2) + { + errmsg ("local_addr and remote_addr required\n"); + return -99; + } + + M(SNAT_ADD_STATIC_MAPPING, snat_add_static_mapping); + mp->is_add = is_add; + mp->is_ip4 = 1; + mp->addr_only = addr_only; + mp->local_port = ntohs ((u16) local_port); + mp->external_port = ntohs ((u16) external_port); + mp->vrf_id = ntohl (vrf_id); + memcpy (mp->local_ip_address, &local_addr, 4); + memcpy (mp->external_ip_address, &external_addr, 4); + + S; W; + /* NOTREACHED */ + return 0; +} + +static void vl_api_snat_control_ping_reply_t_handler + (vl_api_snat_control_ping_reply_t * mp) +{ + vat_main_t *vam = &vat_main; + i32 retval = ntohl (mp->retval); + if (vam->async_mode) + { + vam->async_errors += (retval < 0); + } + else + { + vam->retval = retval; + vam->result_ready = 1; + } +} + +static void vl_api_snat_static_mapping_details_t_handler + (vl_api_snat_static_mapping_details_t *mp) +{ + snat_test_main_t * sm = &snat_test_main; + vat_main_t *vam = sm->vat_main; + + if (mp->addr_only) + fformat (vam->ofp, "%15U%6s%15U%6s%11d\n", + format_ip4_address, &mp->local_ip_address, "", + format_ip4_address, &mp->external_ip_address, "", + ntohl (mp->vrf_id)); + else + fformat (vam->ofp, "%15U%6d%15U%6d%11d\n", + format_ip4_address, &mp->local_ip_address, + ntohs (mp->local_port), + format_ip4_address, &mp->external_ip_address, + ntohs (mp->external_port), + ntohl (mp->vrf_id)); + +} + +static int api_snat_static_mapping_dump(vat_main_t * vam) +{ + snat_test_main_t * sm = &snat_test_main; + f64 timeout; + vl_api_snat_static_mapping_dump_t * mp; + + if (vam->json_output) + { + clib_warning ("JSON output not supported for snat_static_mapping_dump"); + return -99; + } + + fformat (vam->ofp, "%21s%21s\n", "local", "external"); + fformat (vam->ofp, "%15s%6s%15s%6s%11s\n", "address", "port", "address", + "port", "vrf"); + + M(SNAT_STATIC_MAPPING_DUMP, snat_static_mapping_dump); + S; + /* Use a control ping for synchronization */ + { + vl_api_snat_control_ping_t *mp; + M (SNAT_CONTROL_PING, snat_control_ping); + S; + } + W; + /* NOTREACHED */ + return 0; +} + +static void vl_api_snat_show_config_reply_t_handler + (vl_api_snat_show_config_reply_t *mp) +{ + snat_test_main_t * sm = &snat_test_main; + vat_main_t *vam = sm->vat_main; + i32 retval = ntohl (mp->retval); + + if (retval >= 0) + { + fformat (vam->ofp, "translation hash buckets %d\n", + ntohl (mp->translation_buckets)); + fformat (vam->ofp, "translation hash memory %d\n", + ntohl (mp->translation_memory_size)); + fformat (vam->ofp, "user hash buckets %d\n", ntohl (mp->user_buckets)); + fformat (vam->ofp, "user hash memory %d\n", ntohl (mp->user_memory_size)); + fformat (vam->ofp, "max translations per user %d\n", + ntohl (mp->max_translations_per_user)); + fformat (vam->ofp, "outside VRF id %d\n", ntohl (mp->outside_vrf_id)); + fformat (vam->ofp, "inside VRF id %d\n", ntohl (mp->inside_vrf_id)); + if (mp->static_mapping_only) + { + fformat (vam->ofp, "static mapping only"); + if (mp->static_mapping_connection_tracking) + fformat (vam->ofp, " connection tracking"); + fformat (vam->ofp, "\n"); + } + } + vam->retval = retval; + vam->result_ready = 1; +} + +static int api_snat_show_config(vat_main_t * vam) +{ + snat_test_main_t * sm = &snat_test_main; + f64 timeout; + vl_api_snat_show_config_t * mp; + + if (vam->json_output) + { + clib_warning ("JSON output not supported for snat_show_config"); + return -99; + } + + M(SNAT_SHOW_CONFIG, snat_show_config); + S; W; + /* NOTREACHED */ + return 0; +} + +static void vl_api_snat_address_details_t_handler + (vl_api_snat_address_details_t *mp) +{ + snat_test_main_t * sm = &snat_test_main; + vat_main_t *vam = sm->vat_main; + + fformat (vam->ofp, "%U\n", format_ip4_address, &mp->ip_address); +} + +static int api_snat_address_dump(vat_main_t * vam) +{ + snat_test_main_t * sm = &snat_test_main; + f64 timeout; + vl_api_snat_address_dump_t * mp; + + if (vam->json_output) + { + clib_warning ("JSON output not supported for snat_address_dump"); + return -99; + } + + M(SNAT_ADDRESS_DUMP, snat_address_dump); + S; + /* Use a control ping for synchronization */ + { + vl_api_snat_control_ping_t *mp; + M (SNAT_CONTROL_PING, snat_control_ping); + S; + } + W; + /* NOTREACHED */ + return 0; +} + +static void vl_api_snat_interface_details_t_handler + (vl_api_snat_interface_details_t *mp) +{ + snat_test_main_t * sm = &snat_test_main; + vat_main_t *vam = sm->vat_main; + + fformat (vam->ofp, "sw_if_index %d %s\n", ntohl (mp->sw_if_index), + mp->is_inside ? "in" : "out"); +} + +static int api_snat_interface_dump(vat_main_t * vam) +{ + snat_test_main_t * sm = &snat_test_main; + f64 timeout; + vl_api_snat_interface_dump_t * mp; + + if (vam->json_output) + { + clib_warning ("JSON output not supported for snat_address_dump"); + return -99; + } + + M(SNAT_INTERFACE_DUMP, snat_interface_dump); + S; + /* Use a control ping for synchronization */ + { + vl_api_snat_control_ping_t *mp; + M (SNAT_CONTROL_PING, snat_control_ping); + S; + } + W; + /* NOTREACHED */ + return 0; +} + +static int api_snat_set_workers (vat_main_t * vam) +{ + snat_test_main_t * sm = &snat_test_main; + unformat_input_t * i = vam->input; + f64 timeout; + vl_api_snat_set_workers_t * mp; + uword *bitmap; + + while (unformat_check_input (i) != UNFORMAT_END_OF_INPUT) + { + if (unformat (i, "%U", unformat_bitmap_list, &bitmap)) + ; + else + { + clib_warning("unknown input '%U'", format_unformat_error, i); + return -99; + } + } + + M(SNAT_SET_WORKERS, snat_set_workers); + mp->worker_mask = clib_host_to_net_u64 (bitmap[0]); + + S; W; + + /* NOTREACHED */ + return 0; +} + +static void vl_api_snat_worker_details_t_handler + (vl_api_snat_worker_details_t *mp) +{ + snat_test_main_t * sm = &snat_test_main; + vat_main_t *vam = sm->vat_main; + + fformat (vam->ofp, "worker_index %d (%s at lcore %u)\n", + ntohl (mp->worker_index), mp->name, ntohl (mp->lcore_id)); +} + +static int api_snat_worker_dump(vat_main_t * vam) +{ + snat_test_main_t * sm = &snat_test_main; + f64 timeout; + vl_api_snat_worker_dump_t * mp; + + if (vam->json_output) + { + clib_warning ("JSON output not supported for snat_address_dump"); + return -99; + } + + M(SNAT_WORKER_DUMP, snat_worker_dump); + S; + /* Use a control ping for synchronization */ + { + vl_api_snat_control_ping_t *mp; + M (SNAT_CONTROL_PING, snat_control_ping); + S; + } + W; + /* NOTREACHED */ + return 0; +} + +/* + * List of messages that the api test plugin sends, + * and that the data plane plugin processes + */ +#define foreach_vpe_api_msg \ +_(snat_add_address_range, " [- | sw_if_index [in] [out] [del]") \ +_(snat_add_static_mapping, "local_addr external_addr " \ + "[local_port ] [external_port ] [vrf ] [del]") \ +_(snat_set_workers, "") \ +_(snat_static_mapping_dump, "") \ +_(snat_show_config, "") \ +_(snat_address_dump, "") \ +_(snat_interface_dump, "") \ +_(snat_worker_dump, "") + +void vat_api_hookup (vat_main_t *vam) +{ + snat_test_main_t * sm __attribute__((unused)) = &snat_test_main; + /* Hook up handlers for replies from the data plane plug-in */ +#define _(N,n) \ + vl_msg_api_set_handlers((VL_API_##N + sm->msg_id_base), \ + #n, \ + vl_api_##n##_t_handler, \ + vl_noop_handler, \ + vl_api_##n##_t_endian, \ + vl_api_##n##_t_print, \ + sizeof(vl_api_##n##_t), 1); + foreach_vpe_api_reply_msg; +#undef _ + + /* API messages we can send */ +#define _(n,h) hash_set_mem (vam->function_by_name, #n, api_##n); + foreach_vpe_api_msg; +#undef _ + + /* Help strings */ +#define _(n,h) hash_set_mem (vam->help_by_name, #n, h); + foreach_vpe_api_msg; +#undef _ +} + +clib_error_t * vat_plugin_register (vat_main_t *vam) +{ + snat_test_main_t * sm = &snat_test_main; + u8 * name; + + sm->vat_main = vam; + + /* Ask the vpp engine for the first assigned message-id */ + name = format (0, "snat_%08x%c", api_version, 0); + sm->msg_id_base = vl_client_get_first_plugin_msg_id ((char *) name); + + if (sm->msg_id_base != (u16) ~0) + vat_api_hookup (vam); + + vec_free(name); + + return 0; +} diff --git a/src/suffix-rules.mk b/src/suffix-rules.mk index e3eeb922..f97b2276 100644 --- a/src/suffix-rules.mk +++ b/src/suffix-rules.mk @@ -18,10 +18,10 @@ @echo " APIGEN " $@ ; \ mkdir -p `dirname $@` ; \ $(CC) $(CPPFLAGS) -E -P -C -x c $^ \ - | vppapigen --input - --output $@ --show-name $@ + | vppapigen --input - --output $@ --show-name $@ > /dev/null %.api.json: %.api - @echo " JSON APIGEN " $@ ; \ + @echo " JSON API" $@ ; \ mkdir -p `dirname $@` ; \ $(CC) $(CPPFLAGS) -E -P -C -x c $^ \ - | vppapigen --input - --json $@ + | vppapigen --input - --json $@ > /dev/null diff --git a/src/tools/g2/configure.ac b/src/tools/g2/configure.ac deleted file mode 100644 index c8af7747..00000000 --- a/src/tools/g2/configure.ac +++ /dev/null @@ -1,12 +0,0 @@ -AC_INIT(g2, 3.0) -AM_INIT_AUTOMAKE -AM_SILENT_RULES([yes]) - -AC_CHECK_LIB([vppinfra], [clib_mem_get_page_size],, - AC_MSG_ERROR([Please install the vpp-lib package])) -AC_CHECK_HEADER([vppinfra/clib.h],, - AC_MSG_ERROR([Please install the vpp-dev package])) - -PKG_CHECK_MODULES(g2, gtk+-2.0) - -AC_OUTPUT([Makefile]) diff --git a/src/tools/perftool/configure.ac b/src/tools/perftool/configure.ac deleted file mode 100644 index f4a98697..00000000 --- a/src/tools/perftool/configure.ac +++ /dev/null @@ -1,12 +0,0 @@ -AC_INIT(perftool, 2.0) -AM_INIT_AUTOMAKE -AM_SILENT_RULES([yes]) - -AC_CHECK_LIB([vppinfra], [clib_mem_get_page_size],, - AC_MSG_ERROR([Please install the vpp-lib package])) -AC_CHECK_HEADER([vppinfra/clib.h],, - AC_MSG_ERROR([Please install the vpp-dev package])) - -AM_PROG_LIBTOOL - -AC_OUTPUT([Makefile]) diff --git a/src/tools/vppapigen/configure.ac b/src/tools/vppapigen/configure.ac deleted file mode 100644 index 16ad59d2..00000000 --- a/src/tools/vppapigen/configure.ac +++ /dev/null @@ -1,14 +0,0 @@ -# -*- Autoconf -*- -# Copyright (c) 2008 by cisco Systems, Inc. -# All rights reserved. -# Process this file with autoconf to produce a configure script. - -AC_INIT(vppapigen, 1.0) -AM_INIT_AUTOMAKE -AM_SILENT_RULES([yes]) - -# Checks for programs. -AC_PROG_CC -AC_PROG_YACC - -AC_OUTPUT([Makefile]) diff --git a/src/vpp-api/java/Makefile.am b/src/vpp-api/java/Makefile.am new file mode 100644 index 00000000..c7051ec1 --- /dev/null +++ b/src/vpp-api/java/Makefile.am @@ -0,0 +1,199 @@ +# Copyright (c) 2016 Cisco and/or its affiliates. +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at: +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +AUTOMAKE_OPTIONS = foreign +ACLOCAL_AMFLAGS = -I m4 + +AM_CFLAGS = -Wall -I${top_srcdir} -I${top_builddir} \ + -I$(JAVA_HOME)/include -I$(JAVA_HOME)/include/linux \ + -I@top_srcdir@/plugins -I@top_builddir@/plugins + +AM_LDFLAGS = -module -shared -avoid-version -rpath /none -no-undefined + +BUILT_SOURCES = +bin_PROGRAMS = +noinst_LTLIBRARIES = +JAR_FILES = +CLEANDIRS = + +# +# jvpp-common +# + +noinst_LTLIBRARIES += libjvpp_common.la +libjvpp_common_la_SOURCES = jvpp-common/jvpp_common.c +libjvpp_common_la_LDFLAGS = + +# +# jvpp-registry (connection management + plugin registry) +# + +noinst_LTLIBRARIES += libjvpp_registry.la + +libjvpp_registry_la_SOURCES = jvpp-registry/jvpp_registry.c +libjvpp_registry_la_CPPFLAGS = -Ijvpp-registry +libjvpp_registry_la_LIBAD = libjvpp_common.la + +packagedir_jvpp_registry = io/fd/vpp/jvpp +jvpp_registry_src_files := \ + $(wildcard @srcdir@/jvpp-registry/$(packagedir_jvpp_registry)/*.java) \ + $(wildcard @srcdir@/jvpp-registry/$(packagedir_jvpp_registry)/**/*.java) + +BUILT_SOURCES += jvpp-registry/io_fd_vpp_jvpp_VppJNIConnection.h +CLEANDIRS += jvpp-registry/target +JAR_FILES += jvpp-registry-$(PACKAGE_VERSION).jar + +jvpp_registry_ok = jvpp-registry/io_fd_vpp_jvpp_VppJNIConnection.h + +jvpp-registry/io_fd_vpp_jvpp_VppJNIConnection.h: $(jvpp_registry_src_files) + @echo " JAPIGEN $@" + @rm -rf jvpp-registry/target + @mkdir -p jvpp-registry/target + @$(JAVAC) -d jvpp-registry/target $^ + @$(JAVAH) -force -classpath jvpp-registry/target -d jvpp-registry io.fd.vpp.jvpp.VppJNIConnection + @$(JAVAH) -force -classpath jvpp-registry/target -d jvpp-registry io.fd.vpp.jvpp.JVppRegistryImpl + @touch jvpp-registry.ok + +define japigen + @echo " JAPIGEN $@" + @rm -rf jvpp-$(1)/target + @ @srcdir@/jvpp/gen/jvpp_gen.py --plugin_name $(1) --root_dir jvpp-$(1) \ + -i $(jvpp_$(1)_json_files) > /dev/null + @find jvpp-$(1)/target -name \*.java > jvpp-$(1).files + @$(JAVAC) -classpath jvpp-registry/target \ + -d jvpp-$(1)/target @jvpp-$(1).files + @$(JAVAH) -force \ + -classpath jvpp-registry/target:jvpp-$(1)/target \ + -d jvpp-$(1) io.fd.vpp.jvpp.$(1).$(2) +endef + +# +# jvpp-core (Java wrapper for vpe.api) +# +noinst_LTLIBRARIES += libjvpp_core.la +libjvpp_core_la_SOURCES = jvpp-core/jvpp_core.c jvpp-core/jvpp_core_gen.h +libjvpp_core_la_CPPFLAGS = -Ijvpp-registry -Ijvpp-core +BUILT_SOURCES += jvpp-core/io_fd_vpp_jvpp_core_JVppCoreImpl.h + +JAR_FILES += jvpp-core-$(PACKAGE_VERSION).jar +CLEANDIRS += jvpp-core/target +jvpp_core_json_files = $(shell find @top_builddir@/vnet/ -type f -name '*.api.json') + +jvpp-core/io_fd_vpp_jvpp_core_JVppCoreImpl.h: $(jvpp_registry_ok) $(jvpp_core_json_files) + $(call japigen,core,JVppCoreImpl) + +# +# ACL Plugin +# +if ENABLE_ACL_PLUGIN +noinst_LTLIBRARIES += libjvpp_acl.la +libjvpp_acl_la_SOURCES = jvpp-acl/jvpp_acl.c +libjvpp_acl_la_CPPFLAGS = -Ijvpp-acl + +BUILT_SOURCES += jvpp-acl/io_fd_vpp_jvpp_acl_JVppAclImpl.h + +JAR_FILES += jvpp-acl-$(PACKAGE_VERSION).jar +CLEANDIRS += jvpp-acl/target + +jvpp_acl_json_files = @top_builddir@/plugins/acl/acl.api.json + +jvpp-acl/io_fd_vpp_jvpp_acl_JVppAclImpl.h: $(jvpp_registry_ok) $(jvpp_acl_json_files) + $(call japigen,acl,JVppAclImpl) +endif + +# +# SNAT Plugin +# +if ENABLE_SNAT_PLUGIN +noinst_LTLIBRARIES += libjvpp_snat.la +libjvpp_snat_la_SOURCES = jvpp-snat/jvpp_snat.c +libjvpp_snat_la_CPPFLAGS = -Ijvpp-snat + +BUILT_SOURCES += jvpp-snat/io_fd_vpp_jvpp_snat_JVppSnatImpl.h + +JAR_FILES += jvpp-snat-$(PACKAGE_VERSION).jar +CLEANDIRS += jvpp-snat/target + +jvpp_snat_json_files = @top_builddir@/plugins/snat/snat.api.json + +jvpp-snat/io_fd_vpp_jvpp_snat_JVppSnatImpl.h: $(jvpp_registry_ok) $(jvpp_snat_json_files) + $(call japigen,snat,JVppSnatImpl) +endif + +# +# iOAM Trace Plugin +# +if ENABLE_IOAM_PLUGIN +noinst_LTLIBRARIES += libjvpp_ioamtrace.la +libjvpp_ioamtrace_la_SOURCES = jvpp-ioamtrace/jvpp_ioam_trace.c + +BUILT_SOURCES += jvpp-ioamtrace/io_fd_vpp_jvpp_ioamtrace_JVppIoamtraceImpl.h +JAR_FILES += jvpp-ioamtrace-$(PACKAGE_VERSION).jar +CLEANDIRS += jvpp-ioamtrace/target + +jvpp_ioamtrace_json_files = @top_builddir@/plugins/ioam/lib-trace/trace.api.json + +jvpp-ioamtrace/io_fd_vpp_jvpp_ioamtrace_JVppIoamtraceImpl.h: $(jvpp_registry_ok) $(jvpp_ioamtrace_json_files) + $(call japigen,ioamtrace,JVppIoamtraceImpl) + +# +# iOAM POT Plugin +# +noinst_LTLIBRARIES += libjvpp_ioampot.la +libjvpp_ioampot_la_SOURCES = jvpp-ioampot/jvpp_ioam_pot.c + +BUILT_SOURCES += jvpp-ioampot/io_fd_vpp_jvpp_ioampot_JVppIoampotImpl.h +JAR_FILES += jvpp-ioampot-$(PACKAGE_VERSION).jar +CLEANDIRS += jvpp-ioampot/target + +jvpp_ioampot_json_files = @top_builddir@/plugins/ioam/lib-pot/pot.api.json + +jvpp-ioampot/io_fd_vpp_jvpp_ioampot_JVppIoampotImpl.h: $(jvpp_registry_ok) $(jvpp_ioampot_json_files) + $(call japigen,ioampot,JVppIoampotImpl) + +# +# iOAM Export Plugin +# +noinst_LTLIBRARIES += libjvpp_ioamexport.la +libjvpp_ioamexport_la_SOURCES = jvpp-ioamexport/jvpp_ioam_export.c + +BUILT_SOURCES += jvpp-ioamexport/io_fd_vpp_jvpp_ioamexport_JVppIoamexportImpl.h +JAR_FILES += jvpp-ioamexport-$(PACKAGE_VERSION).jar +CLEANDIRS += jvpp-ioamexport/target + +jvpp_ioamexport_json_files = @top_builddir@/plugins/ioam/export/ioam_export.api.json + +jvpp-ioamexport/io_fd_vpp_jvpp_ioamexport_JVppIoamexportImpl.h: $(jvpp_registry_ok) $(jvpp_ioamexport_json_files) + $(call japigen,ioamexport,JVppIoamexportImpl) +endif + +# +# JAR creation +# +jvpp-%-$(PACKAGE_VERSION).jar: libjvpp_%.la + @echo " JAR $@" + @cp .libs/libjvpp_$*.so jvpp-$*/target + @$(JAR) cf $(JARFLAGS) $@ -C jvpp-$*/target . + +jardir = $(prefix)/share/java +jar_DATA = $(JAR_FILES) + +all-local: $(JAR_FILES) + +# +# Cleanup +# +CLEANFILES = jvpp-registry.ok $(JAR_FILES) $(BUILT_SOURCES) + +clean-local: + rm -rf $(CLEANDIRS) diff --git a/src/vpp-api/java/Readme.txt b/src/vpp-api/java/Readme.txt new file mode 100644 index 00000000..689b9b37 --- /dev/null +++ b/src/vpp-api/java/Readme.txt @@ -0,0 +1,236 @@ += JVpp + +JVpp is JNI based Java API for VPP. + +== Features +It is: + +* Asynchronous +* Fully generated +* Lightweight + +== Architecture + +=== Plugin support + + /-------------\ /--------------\ /---------------\ + | JvppPlugin1 +<-------+ JVppRegistry +--------->+ VppConnection | + \-------------/ inits \--+-----------/ uses \---------------/ + | + /-------------\ | + | JvppPlugin2 +<----------+ inits + \-------------/ | + | + ... | + | + /----------\ | + | JVppCore +<-------------+ + \----------/ + + +VppRegistry opens connection to vpp (VppConnection) and manages jvpp plugins. +Each plugin needs to be registered in the VppRegistry. Registration involves +plugin initialization (providing JNI implementation with JVppCallback reference, +vpp client identifier and vpp shared memory queue address). + +API user sends message by calling a method of appropriate plugin interface. +The call is delegated to JNI implementation provided by the particular plugin. +When JNI code receives reply, it invokes callback method of JVppCallback +that corresponds to the received message reply. + +=== JVppCore as an example of JVpp plugin architecture + + JVpp Java + + /--------------\ /----------\ /------------\ /------\ + | JVppRegistry | | JVppCore | | Callbacks | | DTOs | + \----+---------/ \----+-----/ \------+-----/ \------/ + ^ ^ ^ + | implements | implements | implements + /----+--------------\ /---+----------\ /-----+---------\ + | JVppRegistryImpl* +-------->+ JVppCoreImpl | | JVppCallback | + \-------+-----------/ inits \---+----------/ \-------+-------/ + | | ^ + | | uses | calls back + | | | +----------|--------------------------|-----------------------|--------------------- + | | | + C JNI | +-------------------+ | /-----------------\ + v | | +-->+ jvpp_core_gen.h | + /--------+--------\ | | | \-----------------/ + | jpp_registry.c* +---+ /--------+----+----\ | | | + \-----------------/ | | << shared lib >> | /-+--+---+------\ + + ->+ jvpp_common* <--------+ jvpp_core.c* | + uses \------------------/ uses \---------------/ + + +* Components marked with an asterisk contain manually crafted code, which in addition +to generated classes form jvpp. Exception applies to Callbacks and DTOs, since there are +manually crafted marker interfaces in callback and dto package (dto/JVppRequest, dto/JVppReply, +dto/JVppDump, dto/JVppReplyDump, callback/JVppCallback) + +Note: jvpp_core.c calls back the JVppCallback instance with every response. An instance of the +JVppCallback is provided to jvpp_core.c by JVppRegistryImpl on JVppCoreImpl initialization. + +Part of the JVpp is also Future facade. It is asynchronous API returning Future objects +on top of low level JVpp. It wraps dump reply messages in one DTO using control_ping message +(provided by JVppRegistry). + + +Future facade + + /----------------\ /---------------\ + | FutureJVppCore | +-->+ JVppRegistry* | + \-----+----------/ | \---------------/ + ^ | + | implements | uses + | | + /--------+-------------\ | /------------------------------\ + | FutureJVppCoreFacade +---+--->+ FutureJVppCoreFacadeCallback | + \---------+------------/ uses \-------+----------------------/ + | | +---------------|-----------------------------|------------------------------- + | uses | implements +JVpp Java | | + | | + /----------\ | | + | JVppCore +<-+ | + \----+-----/ | + ^ | + | implements v + /----+---------\ /--------+---------------\ + | JVppCoreImpl | | JVppCoreGlobalCallback | + \--------------/ \------------------------/ + + + +Another useful utility of the JVpp is Callback facade. It is asynchronous API +capable of calling specific callback instance (provided when performing a call) +per call. + + +Callback facade + + /------------------\ /---------------\ + | CallbackJVppCore | +-->+ JVppRegistry* | + \-----+------------/ | \---------------/ + ^ | + | implements | uses + | | + /--------+---------------\ | /--------------------------\ + | CallbackJVppCoreFacade +---+--->+ CallbackJVppCoreCallback | + \---------+--------------/ uses \-----+--------------------/ + | | +---------------|-----------------------------|------------------------------- + | uses | implements +JVpp Java | | + | | + /----------\ | | + | JVppCore +<-+ | + \----+-----/ | + ^ | + | implements v + /----+---------\ /----------+-------------\ + | JVppCoreImpl | | JVppCoreGlobalCallback | + \--------------/ \------------------------/ + + +== Package structure + +* *io.fd.vpp.jvpp* - top level package for generated JVpp interface+ implementation and hand-crafted +VppConnection interface + implementation - packaged as jvpp-registry-version.jar + +* *io.fd.vpp.jvpp.[plugin]* - top level package for generated JVpp interface + implementation ++ plugin's API tests - packaged as jvpp-[plugin]-version.jar + +** *dto* - package for DTOs generated from VPP API structures + base/marker hand-crafted interfaces +(in case of jvpp-registry) +** *callback* - package for low-level JVpp callbacks and a global callback interface implementing each of +the low-level JVppcallbacks +** *future* - package for future based facade on top of JVpp and callbacks +** *callfacade* - package for callback based facade on top of JVpp and callbacks. Allowing +users to provide callback per request +** *test* - package for JVpp standalone tests. Can also serve as samples for JVpp. + +C code is structured into modules: + +* *jvpp_common* - shared library that provides jvpp_main_t reference used by jvpp_registry and plugins. + +* *jvpp_registry* - native library used by JVppRegistryImpl, responsible for: + +** VPP connection open/close +** Rx thread to java thread attach +** control ping message handling + +* *jvpp_core* - native library used by jvpp core plugin: +** *jvpp_core.c* - contains hand crafted code for core plugin initialization +** *jvpp_core_gen.h* - contains generated JNI compatible handlers for all requests and replies defined in vpe.api + +== Code generators +All of the required code except the base/marker interfaces is generated using +simple python2 code generators. The generators use __defs_vpp_papi.py__ input +file produced by __vppapigen__ from vpe.api file. + +=== JNI compatible C code +Produces __jvpp_[plugin]_gen.h__ file containing JNI compatible handlers for each VPP +request and reply. + +[NOTE] +==== +Source: jvpp_c_gen.py +==== + +=== Request/Reply DTOs +For all the structures in __defs_vpp_papi.py__ a POJO DTO is produced. Logically, +there are 4 types of DTOs: + +* Request - requests that can be sent to VPP and only a single response is expected +* DumpRequest - requests that can be sent to VPP and a stream of responses is expected +* Reply - reply to a simple request or a single response from dump triggered response stream +* ReplyDump - collection of replies from a single dump request +* Notifications/Events - Not implemented yet + +[NOTE] +==== +Source: dto_gen.py +==== + +=== JVpp +Produces __JVpp.java__ and __JVppImpl.java__. This is the layer right above JNI compatible C +code. + +[NOTE] +==== +Source: jvpp_impl_gen.py +==== + +=== Callbacks +Produces callback interface for each VPP reply + a global callback interface called +__JVpp[plugin]GlobalCallback.java__ aggregating all of the callback interfaces. The JNI +compatible C code expects only a single instance of this global callback and calls +it with every reply. + +[NOTE] +==== +Source: callback_gen.py +==== + +=== Future facade +Produces an asynchronous facade on top of JVpp and callbacks, which returns a Future that provides +matching reply once VPP invocation finishes. Sources produced: +__FutureJVpp[plugin].java, FutureJVpp[plugin]Facade.java and FutureJVpp[plugin]Callback.java__ + +[NOTE] +==== +Source: jvpp_future_facade_gen.py +==== + +=== Callback facade +Similar to future facade, only this facade takes callback objects as part of the invocation +and the callback is called with result once VPP invocation finishes. Sources produced: +__CallbackJVpp[plugin].java, CallbackJVpp[plugin]Facade.java and CallbackJVpp[plugin]Callback.java__ + +[NOTE] +==== +Source: jvpp_callback_facade_gen.py +==== diff --git a/src/vpp-api/java/jvpp-acl/io/fd/vpp/jvpp/acl/test/AclExpectedDumpData.java b/src/vpp-api/java/jvpp-acl/io/fd/vpp/jvpp/acl/test/AclExpectedDumpData.java new file mode 100644 index 00000000..979edbc4 --- /dev/null +++ b/src/vpp-api/java/jvpp-acl/io/fd/vpp/jvpp/acl/test/AclExpectedDumpData.java @@ -0,0 +1,135 @@ +/* + * Copyright (c) 2016 Cisco and/or its affiliates. + * + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +package io.fd.vpp.jvpp.acl.test; + + +import static io.fd.vpp.jvpp.acl.test.AclTestData.FIRST_RULE_ADDRESS_2_AS_ARRAY; +import static io.fd.vpp.jvpp.acl.test.AclTestData.FIRST_RULE_ADDRESS_AS_ARRAY; +import static io.fd.vpp.jvpp.acl.test.AclTestData.FIRST_RULE_DST_ICMP_TYPE_END; +import static io.fd.vpp.jvpp.acl.test.AclTestData.FIRST_RULE_DST_ICMP_TYPE_START; +import static io.fd.vpp.jvpp.acl.test.AclTestData.FIRST_RULE_MAC; +import static io.fd.vpp.jvpp.acl.test.AclTestData.FIRST_RULE_MAC_MASK; +import static io.fd.vpp.jvpp.acl.test.AclTestData.FIRST_RULE_PREFIX; +import static io.fd.vpp.jvpp.acl.test.AclTestData.FIRST_RULE_PREFIX_2; +import static io.fd.vpp.jvpp.acl.test.AclTestData.FIRST_RULE_SRC_ICMP_TYPE_END; +import static io.fd.vpp.jvpp.acl.test.AclTestData.FIRST_RULE_SRC_ICMP_TYPE_START; +import static io.fd.vpp.jvpp.acl.test.AclTestData.ICMP_PROTOCOL; +import static io.fd.vpp.jvpp.acl.test.AclTestData.SECOND_RULE_ADDRESS_2_AS_ARRAY; +import static io.fd.vpp.jvpp.acl.test.AclTestData.SECOND_RULE_ADDRESS_AS_ARRAY; +import static io.fd.vpp.jvpp.acl.test.AclTestData.SECOND_RULE_DST_PORT_RANGE_END; +import static io.fd.vpp.jvpp.acl.test.AclTestData.SECOND_RULE_DST_PORT_RANGE_START; +import static io.fd.vpp.jvpp.acl.test.AclTestData.SECOND_RULE_MAC; +import static io.fd.vpp.jvpp.acl.test.AclTestData.SECOND_RULE_MAC_MASK; +import static io.fd.vpp.jvpp.acl.test.AclTestData.SECOND_RULE_PREFIX; +import static io.fd.vpp.jvpp.acl.test.AclTestData.SECOND_RULE_PREFIX_2; +import static io.fd.vpp.jvpp.acl.test.AclTestData.SECOND_RULE_SRC_PORT_RANGE_END; +import static io.fd.vpp.jvpp.acl.test.AclTestData.SECOND_RULE_SRC_PORT_RANGE_START; +import static io.fd.vpp.jvpp.acl.test.AclTestData.UDP_PROTOCOL; + +import io.fd.vpp.jvpp.acl.dto.AclDetails; +import io.fd.vpp.jvpp.acl.dto.AclInterfaceListDetails; +import io.fd.vpp.jvpp.acl.dto.MacipAclDetails; +import io.fd.vpp.jvpp.acl.types.AclRule; +import io.fd.vpp.jvpp.acl.types.MacipAclRule; +import java.util.Arrays; + +class AclExpectedDumpData { + + static void verifyMacIpDump(final MacipAclDetails macipAclDetails) { + // asserting data create by previous call + assertEquals(0, macipAclDetails.aclIndex); + assertEquals(2, macipAclDetails.count); + + final MacipAclRule currentIpv4Rule = macipAclDetails.r[0]; + final MacipAclRule currentIpv6Rule = macipAclDetails.r[1]; + + // Comparing one property at the time to better pointer if something is wrong + //Ipv4 rule + assertEquals(0, currentIpv4Rule.isIpv6); + assertEquals(1, currentIpv4Rule.isPermit); + + // cutting expected ipv4 to 4 bytes,vpp sends it as 16 always + assertArrays(FIRST_RULE_ADDRESS_AS_ARRAY, Arrays.copyOfRange(currentIpv4Rule.srcIpAddr, 0, 4)); + assertEquals(FIRST_RULE_PREFIX, currentIpv4Rule.srcIpPrefixLen); + assertArrays(FIRST_RULE_MAC, currentIpv4Rule.srcMac); + assertArrays(FIRST_RULE_MAC_MASK, currentIpv4Rule.srcMacMask); + + //Ipv6 rule + assertEquals(1, currentIpv6Rule.isIpv6); + assertEquals(0, currentIpv6Rule.isPermit); + assertArrays(SECOND_RULE_ADDRESS_AS_ARRAY, currentIpv6Rule.srcIpAddr); + assertEquals(SECOND_RULE_PREFIX, currentIpv6Rule.srcIpPrefixLen); + assertArrays(SECOND_RULE_MAC, currentIpv6Rule.srcMac); + assertArrays(SECOND_RULE_MAC_MASK, currentIpv6Rule.srcMacMask); + } + + static void verifyAclDump(final AclDetails aclDetails) { + assertEquals(0, aclDetails.aclIndex); + assertEquals(2, aclDetails.count); + + final AclRule currentIpv4Rule = aclDetails.r[0]; + final AclRule currentIpv6Rule = aclDetails.r[1]; + + // Comparing one property at the time to better pointer if something is wrong + //Ipv4 rule + assertEquals(0, currentIpv4Rule.isIpv6); + assertEquals(1, currentIpv4Rule.isPermit); + + // cutting expected ipv4 to 4 bytes,vpp sends it as 16 always + assertArrays(FIRST_RULE_ADDRESS_AS_ARRAY, Arrays.copyOfRange(currentIpv4Rule.srcIpAddr, 0, 4)); + assertEquals(FIRST_RULE_PREFIX, currentIpv4Rule.srcIpPrefixLen); + assertArrays(FIRST_RULE_ADDRESS_2_AS_ARRAY, Arrays.copyOfRange(currentIpv4Rule.dstIpAddr, 0, 4)); + assertEquals(FIRST_RULE_PREFIX_2, currentIpv4Rule.dstIpPrefixLen); + + assertEquals(ICMP_PROTOCOL, currentIpv4Rule.proto); + assertEquals(FIRST_RULE_SRC_ICMP_TYPE_START, currentIpv4Rule.srcportOrIcmptypeFirst); + assertEquals(FIRST_RULE_SRC_ICMP_TYPE_END, currentIpv4Rule.srcportOrIcmptypeLast); + assertEquals(FIRST_RULE_DST_ICMP_TYPE_START, currentIpv4Rule.dstportOrIcmpcodeFirst); + assertEquals(FIRST_RULE_DST_ICMP_TYPE_END, currentIpv4Rule.dstportOrIcmpcodeLast); + + assertArrays(SECOND_RULE_ADDRESS_AS_ARRAY, currentIpv6Rule.srcIpAddr); + assertEquals(SECOND_RULE_PREFIX, currentIpv6Rule.srcIpPrefixLen); + assertArrays(SECOND_RULE_ADDRESS_2_AS_ARRAY, currentIpv6Rule.dstIpAddr); + assertEquals(SECOND_RULE_PREFIX_2, currentIpv6Rule.dstIpPrefixLen); + + assertEquals(UDP_PROTOCOL, currentIpv6Rule.proto); + assertEquals(SECOND_RULE_SRC_PORT_RANGE_START, currentIpv6Rule.srcportOrIcmptypeFirst); + assertEquals(SECOND_RULE_SRC_PORT_RANGE_END, currentIpv6Rule.srcportOrIcmptypeLast); + assertEquals(SECOND_RULE_DST_PORT_RANGE_START, currentIpv6Rule.dstportOrIcmpcodeFirst); + assertEquals(SECOND_RULE_DST_PORT_RANGE_END, currentIpv6Rule.dstportOrIcmpcodeLast); + } + + static void verifyAclInterfaceList(final AclInterfaceListDetails aclInterfaceListDetails) { + assertEquals(1, aclInterfaceListDetails.count); + assertEquals(1, aclInterfaceListDetails.acls[0]); + assertEquals(0, aclInterfaceListDetails.nInput); + assertEquals(0, aclInterfaceListDetails.swIfIndex); + } + + private static void assertArrays(final byte[] expected, final byte[] actual) { + if (!Arrays.equals(expected, actual)) { + throw new IllegalArgumentException( + String.format("Expected[%s]/Actual[%s]", Arrays.toString(expected), Arrays.toString(actual))); + } + } + + private static void assertEquals(final int expected, final int actual) { + if (expected != actual) { + throw new IllegalArgumentException(String.format("Expected[%s]/Actual[%s]", expected, actual)); + } + } +} diff --git a/src/vpp-api/java/jvpp-acl/io/fd/vpp/jvpp/acl/test/AclTestData.java b/src/vpp-api/java/jvpp-acl/io/fd/vpp/jvpp/acl/test/AclTestData.java new file mode 100644 index 00000000..5d228eea --- /dev/null +++ b/src/vpp-api/java/jvpp-acl/io/fd/vpp/jvpp/acl/test/AclTestData.java @@ -0,0 +1,101 @@ +/* + * Copyright (c) 2016 Cisco and/or its affiliates. + * + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +package io.fd.vpp.jvpp.acl.test; + + +import io.fd.vpp.jvpp.acl.types.AclRule; +import io.fd.vpp.jvpp.acl.types.MacipAclRule; + +class AclTestData { + + static final byte[] FIRST_RULE_ADDRESS_AS_ARRAY = {-64, -88, 2, 1}; + static final byte[] FIRST_RULE_ADDRESS_2_AS_ARRAY = {-64, -88, 2, 3}; + static final byte[] SECOND_RULE_ADDRESS_AS_ARRAY = + {32, 1, 13, -72, 10, 11, 18, -16, 0, 0, 0, 0, 0, 0, 0, 1}; + static final byte[] SECOND_RULE_ADDRESS_2_AS_ARRAY = + {32, 1, 13, -72, 10, 11, 18, -16, 0, 0, 0, 0, 0, 0, 0, 1}; + static final byte[] FIRST_RULE_MAC = {11, 11, 11, 11, 11, 11}; + static final byte[] FIRST_RULE_MAC_MASK = {0, 0, 0, 0, 0, 0}; + static final byte[] SECOND_RULE_MAC = {11, 12, 11, 11, 12, 11}; + static final byte[] SECOND_RULE_MAC_MASK = {(byte) 170, 0, 0, 0, 0, 0}; + static final int FIRST_RULE_PREFIX = 32; + static final int FIRST_RULE_PREFIX_2 = 24; + static final int SECOND_RULE_PREFIX = 64; + static final int SECOND_RULE_PREFIX_2 = 62; + static final int FIRST_RULE_DST_ICMP_TYPE_START = 0; + static final int FIRST_RULE_DST_ICMP_TYPE_END = 8; + static final int FIRST_RULE_SRC_ICMP_TYPE_START = 1; + static final int FIRST_RULE_SRC_ICMP_TYPE_END = 7; + static final int ICMP_PROTOCOL = 1; + static final int SECOND_RULE_DST_PORT_RANGE_START = 2000; + static final int SECOND_RULE_DST_PORT_RANGE_END = 6000; + static final int SECOND_RULE_SRC_PORT_RANGE_START = 400; + static final int SECOND_RULE_SRC_PORT_RANGE_END = 2047; + static final int UDP_PROTOCOL = 17; + + + static MacipAclRule[] createMacipRules() { + MacipAclRule ruleOne = new MacipAclRule(); + ruleOne.isIpv6 = 0; + ruleOne.isPermit = 1; + ruleOne.srcIpAddr = FIRST_RULE_ADDRESS_AS_ARRAY; + ruleOne.srcIpPrefixLen = FIRST_RULE_PREFIX; + ruleOne.srcMac = FIRST_RULE_MAC; + ruleOne.srcMacMask = FIRST_RULE_MAC_MASK;// no mask + + MacipAclRule ruleTwo = new MacipAclRule(); + ruleTwo.isIpv6 = 1; + ruleTwo.isPermit = 0; + ruleTwo.srcIpAddr = SECOND_RULE_ADDRESS_AS_ARRAY; + ruleTwo.srcIpPrefixLen = SECOND_RULE_PREFIX; + ruleTwo.srcMac = SECOND_RULE_MAC; + ruleTwo.srcMacMask = SECOND_RULE_MAC_MASK; + + return new MacipAclRule[]{ruleOne, ruleTwo}; + } + + static AclRule[] createAclRules() { + AclRule ruleOne = new AclRule(); + + ruleOne.isIpv6 = 0; + ruleOne.isPermit = 1; + ruleOne.srcIpAddr = FIRST_RULE_ADDRESS_AS_ARRAY; + ruleOne.srcIpPrefixLen = FIRST_RULE_PREFIX; + ruleOne.dstIpAddr = FIRST_RULE_ADDRESS_2_AS_ARRAY; + ruleOne.dstIpPrefixLen = FIRST_RULE_PREFIX_2; + ruleOne.dstportOrIcmpcodeFirst = FIRST_RULE_DST_ICMP_TYPE_START; + ruleOne.dstportOrIcmpcodeLast = FIRST_RULE_DST_ICMP_TYPE_END; + ruleOne.srcportOrIcmptypeFirst = FIRST_RULE_SRC_ICMP_TYPE_START; + ruleOne.srcportOrIcmptypeLast = FIRST_RULE_SRC_ICMP_TYPE_END; + ruleOne.proto = ICMP_PROTOCOL; //ICMP + + AclRule ruleTwo = new AclRule(); + ruleTwo.isIpv6 = 1; + ruleTwo.isPermit = 0; + ruleTwo.srcIpAddr = SECOND_RULE_ADDRESS_AS_ARRAY; + ruleTwo.srcIpPrefixLen = SECOND_RULE_PREFIX; + ruleTwo.dstIpAddr = SECOND_RULE_ADDRESS_2_AS_ARRAY; + ruleTwo.dstIpPrefixLen = SECOND_RULE_PREFIX_2; + ruleTwo.dstportOrIcmpcodeFirst = SECOND_RULE_DST_PORT_RANGE_START; + ruleTwo.dstportOrIcmpcodeLast = SECOND_RULE_DST_PORT_RANGE_END; + ruleTwo.srcportOrIcmptypeFirst = SECOND_RULE_SRC_PORT_RANGE_START; + ruleTwo.srcportOrIcmptypeLast = SECOND_RULE_SRC_PORT_RANGE_END; + ruleTwo.proto = UDP_PROTOCOL; //UDP + + return new AclRule[]{ruleOne, ruleTwo}; + } +} diff --git a/src/vpp-api/java/jvpp-acl/io/fd/vpp/jvpp/acl/test/AclTestRequests.java b/src/vpp-api/java/jvpp-acl/io/fd/vpp/jvpp/acl/test/AclTestRequests.java new file mode 100644 index 00000000..b580ee8c --- /dev/null +++ b/src/vpp-api/java/jvpp-acl/io/fd/vpp/jvpp/acl/test/AclTestRequests.java @@ -0,0 +1,141 @@ +/* + * Copyright (c) 2016 Cisco and/or its affiliates. + * + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +package io.fd.vpp.jvpp.acl.test; + +import static io.fd.vpp.jvpp.acl.test.AclTestData.createAclRules; +import static io.fd.vpp.jvpp.acl.test.AclTestData.createMacipRules; + +import io.fd.vpp.jvpp.VppInvocationException; +import io.fd.vpp.jvpp.acl.dto.AclAddReplace; +import io.fd.vpp.jvpp.acl.dto.AclAddReplaceReply; +import io.fd.vpp.jvpp.acl.dto.AclDel; +import io.fd.vpp.jvpp.acl.dto.AclDelReply; +import io.fd.vpp.jvpp.acl.dto.AclDetailsReplyDump; +import io.fd.vpp.jvpp.acl.dto.AclDump; +import io.fd.vpp.jvpp.acl.dto.AclInterfaceListDetailsReplyDump; +import io.fd.vpp.jvpp.acl.dto.AclInterfaceListDump; +import io.fd.vpp.jvpp.acl.dto.AclInterfaceSetAclList; +import io.fd.vpp.jvpp.acl.dto.AclInterfaceSetAclListReply; +import io.fd.vpp.jvpp.acl.dto.MacipAclAdd; +import io.fd.vpp.jvpp.acl.dto.MacipAclAddReply; +import io.fd.vpp.jvpp.acl.dto.MacipAclDel; +import io.fd.vpp.jvpp.acl.dto.MacipAclDelReply; +import io.fd.vpp.jvpp.acl.dto.MacipAclDetailsReplyDump; +import io.fd.vpp.jvpp.acl.dto.MacipAclDump; +import io.fd.vpp.jvpp.acl.future.FutureJVppAclFacade; +import java.util.concurrent.ExecutionException; + +class AclTestRequests { + + static MacipAclDetailsReplyDump sendMacIpDumpRequest(final FutureJVppAclFacade jvpp) + throws ExecutionException, InterruptedException { + System.out.println("Sending MacipAclDump request..."); + MacipAclDetailsReplyDump dump = jvpp.macipAclDump(new MacipAclDump()).toCompletableFuture().get(); + System.out.println("MacipAclDump returned"); + return dump; + } + + static void sendMacIpAddRequest(final FutureJVppAclFacade jvpp) throws InterruptedException, ExecutionException { + final MacipAclAdd request = createMacIpAddRequest(); + System.out.printf("Sending MacipAclAdd request %s%n", request.toString()); + final MacipAclAddReply reply = jvpp.macipAclAdd(createMacIpAddRequest()).toCompletableFuture().get(); + System.out.printf("MacipAclAdd send result = %s%n", reply); + } + + static void sendMacIpDelRequest(final FutureJVppAclFacade jvpp) throws InterruptedException, ExecutionException { + final MacipAclDel request = new MacipAclDel(); + request.aclIndex = 0; + System.out.printf("Sending MacipAclDel request %s%n", request.toString()); + final MacipAclDelReply reply = jvpp.macipAclDel(request).toCompletableFuture().get(); + System.out.printf("MacipAclDel send result = %s%n", reply); + } + + static void sendAclAddRequest(final FutureJVppAclFacade jvpp) throws InterruptedException, ExecutionException { + final AclAddReplace request = createAclAddRequest(); + System.out.printf("Sending AclAddReplace request %s%n", request.toString()); + final AclAddReplaceReply reply = jvpp.aclAddReplace(request).toCompletableFuture().get(); + System.out.printf("AclAddReplace send result = %s%n", reply); + } + + static AclDetailsReplyDump sendAclDumpRequest(final FutureJVppAclFacade jvpp) + throws InterruptedException, VppInvocationException, ExecutionException { + System.out.println("Sending AclDump request..."); + final AclDetailsReplyDump dump = jvpp.aclDump(new AclDump()).toCompletableFuture().get(); + System.out.printf("AclDump send result = %s%n", dump); + return dump; + } + + static void sendAclDelRequest(final FutureJVppAclFacade jvpp) throws InterruptedException, ExecutionException { + final AclDel request = new AclDel(); + request.aclIndex = 0; + System.out.printf("Sending AclDel request %s%n", request.toString()); + final AclDelReply reply = jvpp.aclDel(request).toCompletableFuture().get(); + System.out.printf("AclDel send result = %s%n", reply); + } + + static AclInterfaceListDetailsReplyDump sendAclInterfaceListDumpRequest(final FutureJVppAclFacade jvpp) + throws InterruptedException, ExecutionException { + final AclInterfaceListDump request = new AclInterfaceListDump(); + request.swIfIndex = 0; + System.out.printf("Sending AclInterfaceListDump request %s%n", request.toString()); + final AclInterfaceListDetailsReplyDump dump = jvpp.aclInterfaceListDump(request).toCompletableFuture().get(); + System.out.printf("AclInterfaceListDump send result = %s%n", dump); + return dump; + } + + static void sendAclInterfaceSetAclList(final FutureJVppAclFacade jvpp) + throws InterruptedException, ExecutionException { + final AclInterfaceSetAclList request = new AclInterfaceSetAclList(); + request.count = 1; + request.acls = new int[]{1}; + request.swIfIndex = 0; + request.nInput = 0; + System.out.printf("Sending AclInterfaceSetAclList request %s%n", request.toString()); + final AclInterfaceSetAclListReply reply = jvpp.aclInterfaceSetAclList(request).toCompletableFuture().get(); + System.out.printf("AclInterfaceSetAclList send result = %s%n", reply); + } + + static void sendAclInterfaceDeleteList(final FutureJVppAclFacade jvpp) + throws InterruptedException, ExecutionException { + // uses same api but sets list to empty + final AclInterfaceSetAclList request = new AclInterfaceSetAclList(); + request.count = 0; + request.acls = new int[]{}; + request.swIfIndex = 0; + request.nInput = 0; + System.out.printf("Sending AclInterfaceSetAclList(Delete) request %s%n", request.toString()); + final AclInterfaceSetAclListReply reply = jvpp.aclInterfaceSetAclList(request).toCompletableFuture().get(); + System.out.printf("AclInterfaceSetAclList(Delete) send result = %s%n", reply); + } + + private static MacipAclAdd createMacIpAddRequest() { + MacipAclAdd request = new MacipAclAdd(); + + request.count = 2; + request.r = createMacipRules(); + return request; + } + + private static AclAddReplace createAclAddRequest() { + AclAddReplace request = new AclAddReplace(); + + request.aclIndex = -1;// to define new one + request.count = 2; + request.r = createAclRules(); + return request; + } +} diff --git a/src/vpp-api/java/jvpp-acl/io/fd/vpp/jvpp/acl/test/FutureApiTest.java b/src/vpp-api/java/jvpp-acl/io/fd/vpp/jvpp/acl/test/FutureApiTest.java new file mode 100644 index 00000000..94490193 --- /dev/null +++ b/src/vpp-api/java/jvpp-acl/io/fd/vpp/jvpp/acl/test/FutureApiTest.java @@ -0,0 +1,68 @@ +/* + * Copyright (c) 2016 Cisco and/or its affiliates. + * + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +package io.fd.vpp.jvpp.acl.test; + +import static io.fd.vpp.jvpp.acl.test.AclExpectedDumpData.verifyAclDump; +import static io.fd.vpp.jvpp.acl.test.AclExpectedDumpData.verifyAclInterfaceList; +import static io.fd.vpp.jvpp.acl.test.AclExpectedDumpData.verifyMacIpDump; +import static io.fd.vpp.jvpp.acl.test.AclTestRequests.sendAclAddRequest; +import static io.fd.vpp.jvpp.acl.test.AclTestRequests.sendAclDelRequest; +import static io.fd.vpp.jvpp.acl.test.AclTestRequests.sendAclDumpRequest; +import static io.fd.vpp.jvpp.acl.test.AclTestRequests.sendAclInterfaceDeleteList; +import static io.fd.vpp.jvpp.acl.test.AclTestRequests.sendAclInterfaceListDumpRequest; +import static io.fd.vpp.jvpp.acl.test.AclTestRequests.sendAclInterfaceSetAclList; +import static io.fd.vpp.jvpp.acl.test.AclTestRequests.sendMacIpAddRequest; +import static io.fd.vpp.jvpp.acl.test.AclTestRequests.sendMacIpDelRequest; +import static io.fd.vpp.jvpp.acl.test.AclTestRequests.sendMacIpDumpRequest; + +import io.fd.vpp.jvpp.JVppRegistry; +import io.fd.vpp.jvpp.JVppRegistryImpl; +import io.fd.vpp.jvpp.acl.JVppAclImpl; +import io.fd.vpp.jvpp.acl.future.FutureJVppAclFacade; + +public class FutureApiTest { + + public static void main(String[] args) throws Exception { + testCallbackApi(); + } + + private static void testCallbackApi() throws Exception { + System.out.println("Testing Java callback API for acl plugin"); + try (final JVppRegistry registry = new JVppRegistryImpl("macipAclAddTest"); + final FutureJVppAclFacade jvpp = new FutureJVppAclFacade(registry, new JVppAclImpl())) { + + // adds,dump and verifies Mac-Ip acl + sendMacIpAddRequest(jvpp); + verifyMacIpDump(sendMacIpDumpRequest(jvpp).macipAclDetails.get(0)); + + // adds,dumps and verifies Acl acl + sendAclAddRequest(jvpp); + verifyAclDump(sendAclDumpRequest(jvpp).aclDetails.get(0)); + + // adds,dumps and verifies Interface for acl + sendAclInterfaceSetAclList(jvpp); + verifyAclInterfaceList(sendAclInterfaceListDumpRequest(jvpp).aclInterfaceListDetails.get(0)); + + // deletes all created data + sendAclInterfaceDeleteList(jvpp); + sendAclDelRequest(jvpp); + sendMacIpDelRequest(jvpp); + + System.out.println("Disconnecting..."); + } + } +} diff --git a/src/vpp-api/java/jvpp-acl/io/fd/vpp/jvpp/acl/test/Readme.txt b/src/vpp-api/java/jvpp-acl/io/fd/vpp/jvpp/acl/test/Readme.txt new file mode 100644 index 00000000..f68e7aba --- /dev/null +++ b/src/vpp-api/java/jvpp-acl/io/fd/vpp/jvpp/acl/test/Readme.txt @@ -0,0 +1 @@ +sudo java -cp build-vpp-native/vpp-api/java/jvpp-registry-17.01.jar:build-vpp-native/plugins/acl-plugin/jvpp-acl-1.0.jar io.fd.vpp.jvpp.acl.test.FutureApiTest diff --git a/src/vpp-api/java/jvpp-acl/jvpp_acl.c b/src/vpp-api/java/jvpp-acl/jvpp_acl.c new file mode 100644 index 00000000..d56abe3d --- /dev/null +++ b/src/vpp-api/java/jvpp-acl/jvpp_acl.c @@ -0,0 +1,124 @@ +/* + * Copyright (c) 2016 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +#include + +#include +#define vl_typedefs /* define message structures */ +#include +#undef vl_typedefs + +#define vl_endianfun +#include +#undef vl_endianfun + +#define vl_print(handle, ...) +#define vl_printfun +#include +#undef vl_printfun + +/* Get the API version number */ +#define vl_api_version(n,v) static u32 api_version=(v); +#include +#undef vl_api_version + +#include +#include +#include + +#if VPPJNI_DEBUG == 1 + #define DEBUG_LOG(...) clib_warning(__VA_ARGS__) +#else + #define DEBUG_LOG(...) +#endif + +#include + +#include "jvpp-acl/io_fd_vpp_jvpp_acl_JVppAclImpl.h" +#include "jvpp_acl.h" +#include "jvpp-acl/jvpp_acl_gen.h" + +/* + * Class: io_fd_vpp_jvpp_acl_JVppaclImpl + * Method: init0 + * Signature: (JI)V + */ +JNIEXPORT void JNICALL Java_io_fd_vpp_jvpp_acl_JVppAclImpl_init0 + (JNIEnv *env, jclass clazz, jobject callback, jlong queue_address, jint my_client_index) { + acl_main_t * plugin_main = &acl_main; + u8 * name; + clib_warning ("Java_io_fd_vpp_jvpp_acl_JVppAclImpl_init0"); + + plugin_main->my_client_index = my_client_index; + plugin_main->vl_input_queue = (unix_shared_memory_queue_t *)queue_address; + + name = format (0, "acl_%08x%c", api_version, 0); + plugin_main->msg_id_base = vl_client_get_first_plugin_msg_id ((char *) name); + + if (plugin_main->msg_id_base == (u16) ~0) { + jclass exClass = (*env)->FindClass(env, "java/lang/IllegalStateException"); + (*env)->ThrowNew(env, exClass, "acl plugin is not loaded in VPP"); + } else { + plugin_main->callbackObject = (*env)->NewGlobalRef(env, callback); + plugin_main->callbackClass = (jclass)(*env)->NewGlobalRef(env, (*env)->GetObjectClass(env, callback)); + + #define _(N,n) \ + vl_msg_api_set_handlers(VL_API_##N + plugin_main->msg_id_base, #n, \ + vl_api_##n##_t_handler, \ + vl_noop_handler, \ + vl_api_##n##_t_endian, \ + vl_api_##n##_t_print, \ + sizeof(vl_api_##n##_t), 1); + foreach_api_reply_handler; + #undef _ + } +} + +JNIEXPORT void JNICALL Java_io_fd_vpp_jvpp_acl_JVppAclImpl_close0 +(JNIEnv *env, jclass clazz) { + acl_main_t * plugin_main = &acl_main; + + // cleanup: + (*env)->DeleteGlobalRef(env, plugin_main->callbackClass); + (*env)->DeleteGlobalRef(env, plugin_main->callbackObject); + + plugin_main->callbackClass = NULL; + plugin_main->callbackObject = NULL; +} + +/* Attach thread to JVM and cache class references when initiating JVPP ACL */ +jint JNI_OnLoad(JavaVM *vm, void *reserved) { + JNIEnv* env; + + if ((*vm)->GetEnv(vm, (void**) &env, JNI_VERSION_1_8) != JNI_OK) { + return JNI_EVERSION; + } + + if (cache_class_references(env) != 0) { + clib_warning ("Failed to cache class references\n"); + return JNI_ERR; + } + + return JNI_VERSION_1_8; +} + +/* Clean up cached references when disposing JVPP ACL */ +void JNI_OnUnload(JavaVM *vm, void *reserved) { + JNIEnv* env; + if ((*vm)->GetEnv(vm, (void**) &env, JNI_VERSION_1_8) != JNI_OK) { + return; + } + delete_class_references(env); +} diff --git a/src/vpp-api/java/jvpp-acl/jvpp_acl.h b/src/vpp-api/java/jvpp-acl/jvpp_acl.h new file mode 100644 index 00000000..2b73d672 --- /dev/null +++ b/src/vpp-api/java/jvpp-acl/jvpp_acl.h @@ -0,0 +1,45 @@ +/* + * Copyright (c) 2016 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ +#ifndef __included_jvpp_acl_h__ +#define __included_jvpp_acl_h__ + +#include +#include +#include +#include +#include +#include + +/* Global state for JVPP-acl */ +typedef struct { + /* Base message index for the acl plugin */ + u16 msg_id_base; + + /* Pointer to shared memory queue */ + unix_shared_memory_queue_t * vl_input_queue; + + /* VPP api client index */ + u32 my_client_index; + + /* Callback object and class references enabling asynchronous Java calls */ + jobject callbackObject; + jclass callbackClass; + +} acl_main_t; + +acl_main_t acl_main __attribute__((aligned (64))); + + +#endif /* __included_jvpp_acl_h__ */ diff --git a/src/vpp-api/java/jvpp-common/jvpp_common.c b/src/vpp-api/java/jvpp-common/jvpp_common.c new file mode 100644 index 00000000..a161c09c --- /dev/null +++ b/src/vpp-api/java/jvpp-common/jvpp_common.c @@ -0,0 +1,65 @@ +/* + * Copyright (c) 2016 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ +#define _GNU_SOURCE /* for strcasestr(3) */ + +#include "jvpp_common.h" + +#ifndef JVPP_DEBUG +#define JVPP_DEBUG 0 +#endif + +#if JVPP_DEBUG == 1 +#define DEBUG_LOG(...) clib_warning(__VA_ARGS__) +#else +#define DEBUG_LOG(...) +#endif + +/* shared jvpp main structure */ +jvpp_main_t jvpp_main __attribute__((aligned (64))); + +void call_on_error(const char* callName, int contextId, int retval, + jclass callbackClass, jobject callbackObject, + jclass callbackExceptionClass) { + DEBUG_LOG("\nCallOnError : callback=%s, retval=%d, context=%d\n", callName, + clib_net_to_host_u32(retval), clib_net_to_host_u32(context)); + JNIEnv *env = jvpp_main.jenv; + if (!callbackClass) { + DEBUG_LOG("CallOnError : jm->callbackClass is null!\n"); + return; + } + jmethodID excConstructor = (*env)->GetMethodID(env, callbackExceptionClass, + "", "(Ljava/lang/String;II)V"); + if (!excConstructor) { + DEBUG_LOG("CallOnError : excConstructor is null!\n"); + return; + } + jmethodID callbackExcMethod = (*env)->GetMethodID(env, callbackClass, + "onError", "(Lio/fd/vpp/jvpp/VppCallbackException;)V"); + if (!callbackExcMethod) { + DEBUG_LOG("CallOnError : callbackExcMethod is null!\n"); + return; + } + + jobject excObject = (*env)->NewObject(env, callbackExceptionClass, + excConstructor, (*env)->NewStringUTF(env, callName), + clib_net_to_host_u32(contextId), clib_net_to_host_u32(retval)); + if (!excObject) { + DEBUG_LOG("CallOnError : excObject is null!\n"); + return; + } + + (*env)->CallVoidMethod(env, callbackObject, callbackExcMethod, excObject); + DEBUG_LOG("CallOnError : Response sent\n"); +} diff --git a/src/vpp-api/java/jvpp-common/jvpp_common.h b/src/vpp-api/java/jvpp-common/jvpp_common.h new file mode 100644 index 00000000..bbb203ed --- /dev/null +++ b/src/vpp-api/java/jvpp-common/jvpp_common.h @@ -0,0 +1,67 @@ +/* + * Copyright (c) 2016 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ +#ifndef __included_jvpp_common_h__ +#define __included_jvpp_common_h__ +// +#include +#include +#include +#include + +typedef struct { + /* Unique identifier used for matching replays with requests */ + volatile u32 context_id; + + /* Spinlock */ + volatile u32 lock; + u32 tag; + + /* JNI Native Method Interface pointer for message handlers */ + JNIEnv *jenv; + + /* JNI Invoke Interface pointer for attachment of rx thread to java thread */ + JavaVM *jvm; + + /* Convenience */ + unix_shared_memory_queue_t * vl_input_queue; + u32 my_client_index; +} jvpp_main_t; + +extern jvpp_main_t jvpp_main __attribute__((aligned (64))); + +static_always_inline u32 vppjni_get_context_id(jvpp_main_t * jm) { + return __sync_add_and_fetch(&jm->context_id, 1); +} + +static_always_inline void vppjni_lock(jvpp_main_t * jm, u32 tag) { + while (__sync_lock_test_and_set(&jm->lock, 1)) + ; + jm->tag = tag; +} + +static_always_inline void vppjni_unlock(jvpp_main_t * jm) { + jm->tag = 0; + CLIB_MEMORY_BARRIER(); + jm->lock = 0; +} + +/** + * Calls onError callback on callbackObject reference. Passes instance of callbackExceptionClass as parameter. + */ +void call_on_error(const char* callName, int contextId, int retval, + jclass callbackClass, jobject callbackObject, + jclass callbackExceptionClass); + +#endif /* __included_jvpp_common_h__ */ diff --git a/src/vpp-api/java/jvpp-core/io/fd/vpp/jvpp/core/test/CallbackApiTest.java b/src/vpp-api/java/jvpp-core/io/fd/vpp/jvpp/core/test/CallbackApiTest.java new file mode 100644 index 00000000..986993b8 --- /dev/null +++ b/src/vpp-api/java/jvpp-core/io/fd/vpp/jvpp/core/test/CallbackApiTest.java @@ -0,0 +1,96 @@ +/* + * Copyright (c) 2016 Cisco and/or its affiliates. + * + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +package io.fd.vpp.jvpp.core.test; + +import io.fd.vpp.jvpp.JVpp; +import io.fd.vpp.jvpp.JVppRegistry; +import io.fd.vpp.jvpp.JVppRegistryImpl; +import io.fd.vpp.jvpp.VppCallbackException; +import io.fd.vpp.jvpp.core.JVppCoreImpl; +import io.fd.vpp.jvpp.core.callback.GetNodeIndexCallback; +import io.fd.vpp.jvpp.core.callback.ShowVersionCallback; +import io.fd.vpp.jvpp.core.callback.SwInterfaceCallback; +import io.fd.vpp.jvpp.core.dto.GetNodeIndex; +import io.fd.vpp.jvpp.core.dto.GetNodeIndexReply; +import io.fd.vpp.jvpp.core.dto.ShowVersion; +import io.fd.vpp.jvpp.core.dto.ShowVersionReply; +import io.fd.vpp.jvpp.core.dto.SwInterfaceDetails; +import io.fd.vpp.jvpp.core.dto.SwInterfaceDump; + +public class CallbackApiTest { + + public static void main(String[] args) throws Exception { + testCallbackApi(); + } + + private static void testCallbackApi() throws Exception { + System.out.println("Testing Java callback API with JVppRegistry"); + try (final JVppRegistry registry = new JVppRegistryImpl("CallbackApiTest"); + final JVpp jvpp = new JVppCoreImpl()) { + registry.register(jvpp, new TestCallback()); + + System.out.println("Sending ShowVersion request..."); + final int result = jvpp.send(new ShowVersion()); + System.out.printf("ShowVersion send result = %d%n", result); + + System.out.println("Sending GetNodeIndex request..."); + GetNodeIndex getNodeIndexRequest = new GetNodeIndex(); + getNodeIndexRequest.nodeName = "non-existing-node".getBytes(); + jvpp.send(getNodeIndexRequest); + + System.out.println("Sending SwInterfaceDump request..."); + SwInterfaceDump swInterfaceDumpRequest = new SwInterfaceDump(); + swInterfaceDumpRequest.nameFilterValid = 0; + swInterfaceDumpRequest.nameFilter = "".getBytes(); + jvpp.send(swInterfaceDumpRequest); + + Thread.sleep(1000); + System.out.println("Disconnecting..."); + } + Thread.sleep(1000); + } + + static class TestCallback implements GetNodeIndexCallback, ShowVersionCallback, SwInterfaceCallback { + + @Override + public void onGetNodeIndexReply(final GetNodeIndexReply msg) { + System.out.printf("Received GetNodeIndexReply: %s%n", msg); + } + + @Override + public void onShowVersionReply(final ShowVersionReply msg) { + System.out.printf("Received ShowVersionReply: context=%d, program=%s, version=%s, " + + "buildDate=%s, buildDirectory=%s%n", + msg.context, new String(msg.program), new String(msg.version), + new String(msg.buildDate), new String(msg.buildDirectory)); + } + + @Override + public void onSwInterfaceDetails(final SwInterfaceDetails msg) { + System.out.printf("Received SwInterfaceDetails: interfaceName=%s, l2AddressLength=%d, adminUpDown=%d, " + + "linkUpDown=%d, linkSpeed=%d, linkMtu=%d%n", + new String(msg.interfaceName), msg.l2AddressLength, msg.adminUpDown, + msg.linkUpDown, msg.linkSpeed, (int) msg.linkMtu); + } + + @Override + public void onError(VppCallbackException ex) { + System.out.printf("Received onError exception: call=%s, context=%d, retval=%d%n", ex.getMethodName(), + ex.getCtxId(), ex.getErrorCode()); + } + } +} diff --git a/src/vpp-api/java/jvpp-core/io/fd/vpp/jvpp/core/test/CallbackJVppFacadeNotificationTest.java b/src/vpp-api/java/jvpp-core/io/fd/vpp/jvpp/core/test/CallbackJVppFacadeNotificationTest.java new file mode 100644 index 00000000..d84cb034 --- /dev/null +++ b/src/vpp-api/java/jvpp-core/io/fd/vpp/jvpp/core/test/CallbackJVppFacadeNotificationTest.java @@ -0,0 +1,87 @@ +/* + * Copyright (c) 2016 Cisco and/or its affiliates. + * + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +package io.fd.vpp.jvpp.core.test; + +import io.fd.vpp.jvpp.JVppRegistry; +import io.fd.vpp.jvpp.JVppRegistryImpl; +import io.fd.vpp.jvpp.VppCallbackException; +import io.fd.vpp.jvpp.core.JVppCore; +import io.fd.vpp.jvpp.core.JVppCoreImpl; +import io.fd.vpp.jvpp.core.callback.WantInterfaceEventsCallback; +import io.fd.vpp.jvpp.core.callfacade.CallbackJVppCoreFacade; +import io.fd.vpp.jvpp.core.dto.WantInterfaceEventsReply; + +public class CallbackJVppFacadeNotificationTest { + + private static void testCallbackFacade() throws Exception { + System.out.println("Testing CallbackJVppFacade for notifications"); + + try (final JVppRegistry registry = new JVppRegistryImpl("CallbackFacadeTest"); + final JVppCore jvpp = new JVppCoreImpl()) { + final CallbackJVppCoreFacade jvppCallbackFacade = new CallbackJVppCoreFacade(registry, jvpp); + System.out.println("Successfully connected to VPP"); + + final AutoCloseable notificationListenerReg = + jvppCallbackFacade.getNotificationRegistry().registerSwInterfaceSetFlagsNotificationCallback( + NotificationUtils::printNotification + ); + + jvppCallbackFacade.wantInterfaceEvents(NotificationUtils.getEnableInterfaceNotificationsReq(), + new WantInterfaceEventsCallback() { + @Override + public void onWantInterfaceEventsReply(final WantInterfaceEventsReply reply) { + System.out.println("Interface events started"); + } + + @Override + public void onError(final VppCallbackException ex) { + System.out.printf("Received onError exception: call=%s, context=%d, retval=%d%n", + ex.getMethodName(), ex.getCtxId(), ex.getErrorCode()); + } + }); + + System.out.println("Changing interface configuration"); + NotificationUtils.getChangeInterfaceState().send(jvpp); + + Thread.sleep(1000); + + jvppCallbackFacade.wantInterfaceEvents(NotificationUtils.getDisableInterfaceNotificationsReq(), + new WantInterfaceEventsCallback() { + @Override + public void onWantInterfaceEventsReply(final WantInterfaceEventsReply reply) { + System.out.println("Interface events stopped"); + } + + @Override + public void onError(final VppCallbackException ex) { + System.out.printf("Received onError exception: call=%s, context=%d, retval=%d%n", + ex.getMethodName(), ex.getCtxId(), ex.getErrorCode()); + } + }); + + notificationListenerReg.close(); + + Thread.sleep(2000); + System.out.println("Disconnecting..."); + } + Thread.sleep(1000); + } + + public static void main(String[] args) throws Exception { + testCallbackFacade(); + } +} diff --git a/src/vpp-api/java/jvpp-core/io/fd/vpp/jvpp/core/test/CallbackJVppFacadeTest.java b/src/vpp-api/java/jvpp-core/io/fd/vpp/jvpp/core/test/CallbackJVppFacadeTest.java new file mode 100644 index 00000000..9f7cb8de --- /dev/null +++ b/src/vpp-api/java/jvpp-core/io/fd/vpp/jvpp/core/test/CallbackJVppFacadeTest.java @@ -0,0 +1,103 @@ +/* + * Copyright (c) 2016 Cisco and/or its affiliates. + * + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +package io.fd.vpp.jvpp.core.test; + +import io.fd.vpp.jvpp.JVppRegistry; +import io.fd.vpp.jvpp.JVppRegistryImpl; +import io.fd.vpp.jvpp.VppCallbackException; +import io.fd.vpp.jvpp.core.JVppCoreImpl; +import io.fd.vpp.jvpp.core.callback.GetNodeIndexCallback; +import io.fd.vpp.jvpp.core.callback.ShowVersionCallback; +import io.fd.vpp.jvpp.core.callfacade.CallbackJVppCoreFacade; +import io.fd.vpp.jvpp.core.dto.GetNodeIndex; +import io.fd.vpp.jvpp.core.dto.GetNodeIndexReply; +import io.fd.vpp.jvpp.core.dto.ShowVersionReply; + +/** + * CallbackJVppFacade together with CallbackJVppFacadeCallback allow for setting different callback for each request. + * This is more convenient than the approach shown in CallbackApiTest. + */ +public class CallbackJVppFacadeTest { + + private static ShowVersionCallback showVersionCallback1 = new ShowVersionCallback() { + @Override + public void onShowVersionReply(final ShowVersionReply msg) { + System.out.printf("ShowVersionCallback1 received ShowVersionReply: context=%d, program=%s," + + "version=%s, buildDate=%s, buildDirectory=%s%n", msg.context, new String(msg.program), + new String(msg.version), new String(msg.buildDate), new String(msg.buildDirectory)); + } + + @Override + public void onError(VppCallbackException ex) { + System.out.printf("Received onError exception in showVersionCallback1: call=%s, reply=%d, context=%d%n", + ex.getMethodName(), ex.getErrorCode(), ex.getCtxId()); + } + }; + + private static ShowVersionCallback showVersionCallback2 = new ShowVersionCallback() { + @Override + public void onShowVersionReply(final ShowVersionReply msg) { + System.out.printf("ShowVersionCallback2 received ShowVersionReply: context=%d, program=%s," + + "version=%s, buildDate=%s, buildDirectory=%s%n", msg.context, new String(msg.program), + new String(msg.version), new String(msg.buildDate), new String(msg.buildDirectory)); + } + + @Override + public void onError(VppCallbackException ex) { + System.out.printf("Received onError exception in showVersionCallback2: call=%s, reply=%d, context=%d%n", + ex.getMethodName(), ex.getErrorCode(), ex.getCtxId()); + } + + }; + + private static GetNodeIndexCallback getNodeIndexCallback = new GetNodeIndexCallback() { + @Override + public void onGetNodeIndexReply(final GetNodeIndexReply msg) { + System.out.printf("Received GetNodeIndexReply: %s%n", msg); + } + + @Override + public void onError(VppCallbackException ex) { + System.out.printf("Received onError exception in getNodeIndexCallback: call=%s, reply=%d, context=%d%n", + ex.getMethodName(), ex.getErrorCode(), ex.getCtxId()); + } + }; + + private static void testCallbackFacade() throws Exception { + System.out.println("Testing CallbackJVppFacade"); + + try (final JVppRegistry registry = new JVppRegistryImpl("CallbackFacadeTest"); + final CallbackJVppCoreFacade callbackFacade = new CallbackJVppCoreFacade(registry, new JVppCoreImpl())) { + System.out.println("Successfully connected to VPP"); + + callbackFacade.showVersion(showVersionCallback1); + callbackFacade.showVersion(showVersionCallback2); + + GetNodeIndex getNodeIndexRequest = new GetNodeIndex(); + getNodeIndexRequest.nodeName = "dummyNode".getBytes(); + callbackFacade.getNodeIndex(getNodeIndexRequest, getNodeIndexCallback); + + Thread.sleep(2000); + System.out.println("Disconnecting..."); + } + Thread.sleep(1000); + } + + public static void main(String[] args) throws Exception { + testCallbackFacade(); + } +} diff --git a/src/vpp-api/java/jvpp-core/io/fd/vpp/jvpp/core/test/CallbackNotificationApiTest.java b/src/vpp-api/java/jvpp-core/io/fd/vpp/jvpp/core/test/CallbackNotificationApiTest.java new file mode 100644 index 00000000..a9f71f11 --- /dev/null +++ b/src/vpp-api/java/jvpp-core/io/fd/vpp/jvpp/core/test/CallbackNotificationApiTest.java @@ -0,0 +1,94 @@ +/* + * Copyright (c) 2016 Cisco and/or its affiliates. + * + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +package io.fd.vpp.jvpp.core.test; + +import static io.fd.vpp.jvpp.core.test.NotificationUtils.getChangeInterfaceState; +import static io.fd.vpp.jvpp.core.test.NotificationUtils.getDisableInterfaceNotificationsReq; +import static io.fd.vpp.jvpp.core.test.NotificationUtils.getEnableInterfaceNotificationsReq; +import static io.fd.vpp.jvpp.core.test.NotificationUtils.printNotification; + +import io.fd.vpp.jvpp.JVpp; +import io.fd.vpp.jvpp.JVppRegistry; +import io.fd.vpp.jvpp.JVppRegistryImpl; +import io.fd.vpp.jvpp.VppCallbackException; +import io.fd.vpp.jvpp.core.JVppCoreImpl; +import io.fd.vpp.jvpp.core.callback.SwInterfaceSetFlagsCallback; +import io.fd.vpp.jvpp.core.callback.SwInterfaceSetFlagsNotificationCallback; +import io.fd.vpp.jvpp.core.callback.WantInterfaceEventsCallback; +import io.fd.vpp.jvpp.core.dto.SwInterfaceSetFlagsNotification; +import io.fd.vpp.jvpp.core.dto.SwInterfaceSetFlagsReply; +import io.fd.vpp.jvpp.core.dto.WantInterfaceEventsReply; + +public class CallbackNotificationApiTest { + + private static void testCallbackApi() throws Exception { + System.out.println("Testing Java callback API for notifications"); + try (final JVppRegistry registry = new JVppRegistryImpl("CallbackNotificationTest"); + final JVpp jvpp = new JVppCoreImpl()) { + registry.register(jvpp, new TestCallback()); + System.out.println("Successfully connected to VPP"); + + getEnableInterfaceNotificationsReq().send(jvpp); + System.out.println("Interface notifications started"); + // TODO test ifc dump which also triggers interface flags send + + System.out.println("Changing interface configuration"); + getChangeInterfaceState().send(jvpp); + + // Notifications are received + Thread.sleep(500); + + getDisableInterfaceNotificationsReq().send(jvpp); + System.out.println("Interface events stopped"); + + Thread.sleep(2000); + System.out.println("Disconnecting..."); + } + Thread.sleep(1000); + } + + public static void main(String[] args) throws Exception { + testCallbackApi(); + } + + private static class TestCallback implements SwInterfaceSetFlagsNotificationCallback, + WantInterfaceEventsCallback, SwInterfaceSetFlagsCallback { + + @Override + public void onSwInterfaceSetFlagsNotification( + final SwInterfaceSetFlagsNotification msg) { + printNotification(msg); + } + + @Override + public void onWantInterfaceEventsReply(final WantInterfaceEventsReply wantInterfaceEventsReply) { + System.out.println("Interface notification stream updated"); + } + + @Override + public void onSwInterfaceSetFlagsReply(final SwInterfaceSetFlagsReply swInterfaceSetFlagsReply) { + System.out.println("Interface flags set successfully"); + } + + @Override + public void onError(VppCallbackException ex) { + System.out.printf("Received onError exception in getNodeIndexCallback: call=%s, reply=%d, context=%d%n", + ex.getMethodName(), ex.getErrorCode(), ex.getCtxId()); + + } + } +} diff --git a/src/vpp-api/java/jvpp-core/io/fd/vpp/jvpp/core/test/ControlPingTest.java b/src/vpp-api/java/jvpp-core/io/fd/vpp/jvpp/core/test/ControlPingTest.java new file mode 100644 index 00000000..e97f4e3a --- /dev/null +++ b/src/vpp-api/java/jvpp-core/io/fd/vpp/jvpp/core/test/ControlPingTest.java @@ -0,0 +1,68 @@ +/* + * Copyright (c) 2016 Cisco and/or its affiliates. + * + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +package io.fd.vpp.jvpp.core.test; + +import io.fd.vpp.jvpp.JVpp; +import io.fd.vpp.jvpp.JVppRegistry; +import io.fd.vpp.jvpp.JVppRegistryImpl; +import io.fd.vpp.jvpp.VppCallbackException; +import io.fd.vpp.jvpp.callback.ControlPingCallback; +import io.fd.vpp.jvpp.core.JVppCoreImpl; +import io.fd.vpp.jvpp.dto.ControlPing; +import io.fd.vpp.jvpp.dto.ControlPingReply; + +public class ControlPingTest { + + private static void testControlPing() throws Exception { + System.out.println("Testing ControlPing using Java callback API"); + try (JVppRegistry registry = new JVppRegistryImpl("ControlPingTest"); + JVpp jvpp = new JVppCoreImpl()) { + + registry.register(jvpp, new ControlPingCallback() { + @Override + public void onControlPingReply(final ControlPingReply reply) { + System.out.printf("Received ControlPingReply: %s%n", reply); + } + + @Override + public void onError(VppCallbackException ex) { + System.out.printf("Received onError exception: call=%s, reply=%d, context=%d ", ex.getMethodName(), + ex.getErrorCode(), ex.getCtxId()); + } + + }); + System.out.println("Successfully connected to VPP"); + Thread.sleep(1000); + + System.out.println("Sending control ping using JVppRegistry"); + registry.controlPing(jvpp.getClass()); + + Thread.sleep(2000); + + System.out.println("Sending control ping using JVpp plugin"); + jvpp.send(new ControlPing()); + + Thread.sleep(2000); + System.out.println("Disconnecting..."); + } + Thread.sleep(1000); + } + + public static void main(String[] args) throws Exception { + testControlPing(); + } +} diff --git a/src/vpp-api/java/jvpp-core/io/fd/vpp/jvpp/core/test/CreateSubInterfaceTest.java b/src/vpp-api/java/jvpp-core/io/fd/vpp/jvpp/core/test/CreateSubInterfaceTest.java new file mode 100644 index 00000000..a96258f4 --- /dev/null +++ b/src/vpp-api/java/jvpp-core/io/fd/vpp/jvpp/core/test/CreateSubInterfaceTest.java @@ -0,0 +1,120 @@ +/* + * Copyright (c) 2016 Cisco and/or its affiliates. + * + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +package io.fd.vpp.jvpp.core.test; + +import static java.util.Objects.requireNonNull; + +import io.fd.vpp.jvpp.JVppRegistry; +import io.fd.vpp.jvpp.JVppRegistryImpl; +import io.fd.vpp.jvpp.core.JVppCoreImpl; +import io.fd.vpp.jvpp.core.dto.CreateSubif; +import io.fd.vpp.jvpp.core.dto.CreateSubifReply; +import io.fd.vpp.jvpp.core.dto.SwInterfaceDetailsReplyDump; +import io.fd.vpp.jvpp.core.dto.SwInterfaceDump; +import io.fd.vpp.jvpp.core.future.FutureJVppCoreFacade; + +/** + *

Tests sub-interface creation.
Equivalent to:
+ * + * 

{@code
+ * vppctl create sub GigabitEthernet0/9/0 1 dot1q 100 inner-dot1q any
+ * }
+ *
+ * + * To verify invoke:
+ * 
{@code
+ * vpp_api_test json
+ * vat# sw_interface_dump
+ * }
+ */
+public class CreateSubInterfaceTest {
+
+    private static SwInterfaceDump createSwInterfaceDumpRequest(final String ifaceName) {
+        SwInterfaceDump request = new SwInterfaceDump();
+        request.nameFilter = ifaceName.getBytes();
+        request.nameFilterValid = 1;
+        return request;
+    }
+
+    private static void requireSingleIface(final SwInterfaceDetailsReplyDump response, final String ifaceName) {
+        if (response.swInterfaceDetails.size() != 1) {
+            throw new IllegalStateException(
+                String.format("Expected one interface matching filter %s but was %d", ifaceName,
+                    response.swInterfaceDetails.size()));
+        }
+    }
+
+    private static CreateSubif createSubifRequest(final int swIfIndex, final int subId) {
+        CreateSubif request = new CreateSubif();
+        request.swIfIndex = swIfIndex; // super interface id
+        request.subId = subId;
+        request.noTags = 0;
+        request.oneTag = 0;
+        request.twoTags = 1;
+        request.dot1Ad = 0;
+        request.exactMatch = 1;
+        request.defaultSub = 0;
+        request.outerVlanIdAny = 0;
+        request.innerVlanIdAny = 1;
+        request.outerVlanId = 100;
+        request.innerVlanId = 0;
+        return request;
+    }
+
+    private static void print(CreateSubifReply reply) {
+        System.out.printf("CreateSubifReply: %s%n", reply);
+    }
+
+    private static void testCreateSubInterface() throws Exception {
+        System.out.println("Testing sub-interface creation using Java callback API");
+        try (final JVppRegistry registry = new JVppRegistryImpl("CreateSubInterface");
+             final FutureJVppCoreFacade jvppFacade = new FutureJVppCoreFacade(registry, new JVppCoreImpl())) {
+            System.out.println("Successfully connected to VPP");
+            Thread.sleep(1000);
+
+            final String ifaceName = "GigabitEthernet0/8/0";
+
+            final SwInterfaceDetailsReplyDump swInterfaceDetails =
+                jvppFacade.swInterfaceDump(createSwInterfaceDumpRequest(ifaceName)).toCompletableFuture().get();
+
+            requireNonNull(swInterfaceDetails, "swInterfaceDump returned null");
+            requireNonNull(swInterfaceDetails.swInterfaceDetails, "swInterfaceDetails is null");
+            requireSingleIface(swInterfaceDetails, ifaceName);
+
+            final int swIfIndex = swInterfaceDetails.swInterfaceDetails.get(0).swIfIndex;
+            final int subId = 1;
+
+            final CreateSubifReply createSubifReply =
+                jvppFacade.createSubif(createSubifRequest(swIfIndex, subId)).toCompletableFuture().get();
+            print(createSubifReply);
+
+            final String subIfaceName = "GigabitEthernet0/8/0." + subId;
+            final SwInterfaceDetailsReplyDump subIface =
+                jvppFacade.swInterfaceDump(createSwInterfaceDumpRequest(subIfaceName)).toCompletableFuture().get();
+            requireNonNull(swInterfaceDetails, "swInterfaceDump returned null");
+            requireNonNull(subIface.swInterfaceDetails, "swInterfaceDump returned null");
+            requireSingleIface(swInterfaceDetails, ifaceName);
+
+            System.out.println("Disconnecting...");
+        }
+        Thread.sleep(1000);
+    }
+
+    public static void main(String[] args) throws Exception {
+        testCreateSubInterface();
+    }
+}
diff --git a/src/vpp-api/java/jvpp-core/io/fd/vpp/jvpp/core/test/FutureApiNotificationTest.java b/src/vpp-api/java/jvpp-core/io/fd/vpp/jvpp/core/test/FutureApiNotificationTest.java
new file mode 100644
index 00000000..9efeae19
--- /dev/null
+++ b/src/vpp-api/java/jvpp-core/io/fd/vpp/jvpp/core/test/FutureApiNotificationTest.java
@@ -0,0 +1,55 @@
+/*
+ * Copyright (c) 2016 Cisco and/or its affiliates.
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at:
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+package io.fd.vpp.jvpp.core.test;
+
+import static io.fd.vpp.jvpp.core.test.NotificationUtils.getChangeInterfaceState;
+import static io.fd.vpp.jvpp.core.test.NotificationUtils.getDisableInterfaceNotificationsReq;
+import static io.fd.vpp.jvpp.core.test.NotificationUtils.getEnableInterfaceNotificationsReq;
+
+import io.fd.vpp.jvpp.JVppRegistry;
+import io.fd.vpp.jvpp.JVppRegistryImpl;
+import io.fd.vpp.jvpp.core.JVppCoreImpl;
+import io.fd.vpp.jvpp.core.future.FutureJVppCoreFacade;
+
+public class FutureApiNotificationTest {
+
+    private static void testFutureApi() throws Exception {
+        System.out.println("Testing Java future API for notifications");
+        try (final JVppRegistry registry = new JVppRegistryImpl("FutureApiNotificationTest");
+             final FutureJVppCoreFacade jvppFacade = new FutureJVppCoreFacade(registry, new JVppCoreImpl());
+             final AutoCloseable notificationListenerReg =
+                 jvppFacade.getNotificationRegistry()
+                     .registerSwInterfaceSetFlagsNotificationCallback(NotificationUtils::printNotification)) {
+            System.out.println("Successfully connected to VPP");
+            jvppFacade.wantInterfaceEvents(getEnableInterfaceNotificationsReq()).toCompletableFuture().get();
+            System.out.println("Interface events started");
+
+            System.out.println("Changing interface configuration");
+            jvppFacade.swInterfaceSetFlags(getChangeInterfaceState()).toCompletableFuture().get();
+
+            Thread.sleep(1000);
+
+            jvppFacade.wantInterfaceEvents(getDisableInterfaceNotificationsReq()).toCompletableFuture().get();
+            System.out.println("Interface events stopped");
+            System.out.println("Disconnecting...");
+        }
+    }
+
+    public static void main(String[] args) throws Exception {
+        testFutureApi();
+    }
+}
diff --git a/src/vpp-api/java/jvpp-core/io/fd/vpp/jvpp/core/test/FutureApiTest.java b/src/vpp-api/java/jvpp-core/io/fd/vpp/jvpp/core/test/FutureApiTest.java
new file mode 100644
index 00000000..f478bab4
--- /dev/null
+++ b/src/vpp-api/java/jvpp-core/io/fd/vpp/jvpp/core/test/FutureApiTest.java
@@ -0,0 +1,123 @@
+/*
+ * Copyright (c) 2016 Cisco and/or its affiliates.
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at:
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+package io.fd.vpp.jvpp.core.test;
+
+import io.fd.vpp.jvpp.JVppRegistry;
+import io.fd.vpp.jvpp.JVppRegistryImpl;
+import io.fd.vpp.jvpp.core.JVppCoreImpl;
+import io.fd.vpp.jvpp.core.dto.BridgeDomainDetailsReplyDump;
+import io.fd.vpp.jvpp.core.dto.BridgeDomainDump;
+import io.fd.vpp.jvpp.core.dto.GetNodeIndex;
+import io.fd.vpp.jvpp.core.dto.GetNodeIndexReply;
+import io.fd.vpp.jvpp.core.dto.ShowVersion;
+import io.fd.vpp.jvpp.core.dto.ShowVersionReply;
+import io.fd.vpp.jvpp.core.dto.SwInterfaceDetails;
+import io.fd.vpp.jvpp.core.dto.SwInterfaceDetailsReplyDump;
+import io.fd.vpp.jvpp.core.dto.SwInterfaceDump;
+import io.fd.vpp.jvpp.core.future.FutureJVppCoreFacade;
+import java.util.Objects;
+import java.util.concurrent.CompletableFuture;
+import java.util.concurrent.Future;
+import java.util.logging.Level;
+import java.util.logging.Logger;
+
+public class FutureApiTest {
+
+    private static final Logger LOG = Logger.getLogger(FutureApiTest.class.getName());
+
+    private static void testShowVersion(final FutureJVppCoreFacade jvpp) throws Exception {
+        LOG.info("Sending ShowVersion request...");
+        final Future replyFuture = jvpp.showVersion(new ShowVersion()).toCompletableFuture();
+        final ShowVersionReply reply = replyFuture.get();
+        LOG.info(
+            String.format(
+                "Received ShowVersionReply: context=%d, program=%s, version=%s, buildDate=%s, buildDirectory=%s%n",
+                reply.context, new String(reply.program), new String(reply.version), new String(reply.buildDate),
+                new String(reply.buildDirectory)));
+    }
+
+    private static void testEmptyBridgeDomainDump(final FutureJVppCoreFacade jvpp) throws Exception {
+        LOG.info("Sending ShowVersion request...");
+        final BridgeDomainDump request = new BridgeDomainDump();
+        request.bdId = -1; // dump call
+
+        final CompletableFuture
+            replyFuture = jvpp.bridgeDomainDump(request).toCompletableFuture();
+        final BridgeDomainDetailsReplyDump reply = replyFuture.get();
+
+        if (reply == null || reply.bridgeDomainDetails == null) {
+            LOG.severe("Received null response for empty dump: " + reply);
+        } else {
+            LOG.info(
+                String.format(
+                    "Received empty bridge-domain dump reply with list of bridge-domains: %s, %s",
+                    reply.bridgeDomainDetails, reply.bridgeDomainSwIfDetails));
+        }
+    }
+
+    private static void testGetNodeIndex(final FutureJVppCoreFacade jvpp) {
+        LOG.info("Sending GetNodeIndex request...");
+        final GetNodeIndex request = new GetNodeIndex();
+        request.nodeName = "non-existing-node".getBytes();
+        final Future replyFuture = jvpp.getNodeIndex(request).toCompletableFuture();
+        try {
+            final GetNodeIndexReply reply = replyFuture.get();
+            LOG.info(
+                String.format(
+                    "Received GetNodeIndexReply: context=%d, nodeIndex=%d%n", reply.context, reply.nodeIndex));
+        } catch (Exception e) {
+            LOG.log(Level.SEVERE, "GetNodeIndex request failed", e);
+        }
+    }
+
+    private static void testSwInterfaceDump(final FutureJVppCoreFacade jvpp) throws Exception {
+        LOG.info("Sending SwInterfaceDump request...");
+        final SwInterfaceDump request = new SwInterfaceDump();
+        request.nameFilterValid = 0;
+        request.nameFilter = "".getBytes();
+
+        final Future replyFuture = jvpp.swInterfaceDump(request).toCompletableFuture();
+        final SwInterfaceDetailsReplyDump reply = replyFuture.get();
+        for (SwInterfaceDetails details : reply.swInterfaceDetails) {
+            Objects.requireNonNull(details, "reply.swInterfaceDetails contains null element!");
+            LOG.info(
+                String.format("Received SwInterfaceDetails: interfaceName=%s, l2AddressLength=%d, adminUpDown=%d, "
+                        + "linkUpDown=%d, linkSpeed=%d, linkMtu=%d%n",
+                    new String(details.interfaceName), details.l2AddressLength, details.adminUpDown,
+                    details.linkUpDown, details.linkSpeed, (int) details.linkMtu));
+        }
+    }
+
+    private static void testFutureApi() throws Exception {
+        LOG.info("Testing Java future API");
+        try (final JVppRegistry registry = new JVppRegistryImpl("FutureApiTest");
+             final FutureJVppCoreFacade jvppFacade = new FutureJVppCoreFacade(registry, new JVppCoreImpl())) {
+            LOG.info("Successfully connected to VPP");
+
+            testEmptyBridgeDomainDump(jvppFacade);
+            testShowVersion(jvppFacade);
+            testGetNodeIndex(jvppFacade);
+            testSwInterfaceDump(jvppFacade);
+
+            LOG.info("Disconnecting...");
+        }
+    }
+
+    public static void main(String[] args) throws Exception {
+        testFutureApi();
+    }
+}
diff --git a/src/vpp-api/java/jvpp-core/io/fd/vpp/jvpp/core/test/L2AclTest.java b/src/vpp-api/java/jvpp-core/io/fd/vpp/jvpp/core/test/L2AclTest.java
new file mode 100644
index 00000000..6b3fa993
--- /dev/null
+++ b/src/vpp-api/java/jvpp-core/io/fd/vpp/jvpp/core/test/L2AclTest.java
@@ -0,0 +1,195 @@
+/*
+ * Copyright (c) 2016 Cisco and/or its affiliates.
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at:
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+package io.fd.vpp.jvpp.core.test;
+
+import io.fd.vpp.jvpp.JVppRegistry;
+import io.fd.vpp.jvpp.JVppRegistryImpl;
+import io.fd.vpp.jvpp.core.JVppCoreImpl;
+import io.fd.vpp.jvpp.core.dto.ClassifyAddDelSession;
+import io.fd.vpp.jvpp.core.dto.ClassifyAddDelSessionReply;
+import io.fd.vpp.jvpp.core.dto.ClassifyAddDelTable;
+import io.fd.vpp.jvpp.core.dto.ClassifyAddDelTableReply;
+import io.fd.vpp.jvpp.core.dto.ClassifySessionDetailsReplyDump;
+import io.fd.vpp.jvpp.core.dto.ClassifySessionDump;
+import io.fd.vpp.jvpp.core.dto.ClassifyTableByInterface;
+import io.fd.vpp.jvpp.core.dto.ClassifyTableByInterfaceReply;
+import io.fd.vpp.jvpp.core.dto.ClassifyTableIds;
+import io.fd.vpp.jvpp.core.dto.ClassifyTableIdsReply;
+import io.fd.vpp.jvpp.core.dto.ClassifyTableInfo;
+import io.fd.vpp.jvpp.core.dto.ClassifyTableInfoReply;
+import io.fd.vpp.jvpp.core.dto.InputAclSetInterface;
+import io.fd.vpp.jvpp.core.dto.InputAclSetInterfaceReply;
+import io.fd.vpp.jvpp.core.future.FutureJVppCoreFacade;
+import javax.xml.bind.DatatypeConverter;
+
+/**
+ * 
Tests L2 ACL creation and read.
 Equivalent to the following vppctl commands:

+ *
+ * 
{@code
+ * vppctl classify table mask l2 src
+ * vppctl classify session acl-hit-next deny opaque-index 0 table-index 0 match l2 src 01:02:03:04:05:06
+ * vppctl set int input acl intfc local0 l2-table 0
+ * vppctl sh class table verbose
+ * }
+ * 

+ */
+public class L2AclTest {
+
+    private static final int LOCAL0_IFACE_ID = 0;
+
+    private static ClassifyAddDelTable createClassifyTable() {
+        ClassifyAddDelTable request = new ClassifyAddDelTable();
+        request.isAdd = 1;
+        request.tableIndex = ~0; // default
+        request.nbuckets = 2;
+        request.memorySize = 2 << 20;
+        request.nextTableIndex = ~0; // default
+        request.missNextIndex = ~0; // default
+        request.skipNVectors = 0;
+        request.matchNVectors = 1;
+        request.mask =
+            new byte[] {0x00, 0x00, 0x00, 0x00, 0x00, 0x00, (byte) 0xff, (byte) 0xff, (byte) 0xff, (byte) 0xff,
+                (byte) 0xff, (byte) 0xff, 0x00, 0x00, 0x00, 0x00};
+        return request;
+    }
+
+    private static ClassifyTableInfo createClassifyTableInfoRequest(final int tableId) {
+        ClassifyTableInfo request = new ClassifyTableInfo();
+        request.tableId = tableId;
+        return request;
+    }
+
+    private static ClassifyAddDelSession createClassifySession(final int tableIndex) {
+        ClassifyAddDelSession request = new ClassifyAddDelSession();
+        request.isAdd = 1;
+        request.tableIndex = tableIndex;
+        request.hitNextIndex = 0; // deny
+        request.opaqueIndex = 0;
+        request.advance = 0; // default
+        // match 01:02:03:04:05:06 mac address
+        request.match =
+            new byte[] {0x00, 0x00, 0x00, 0x00, 0x00, 0x00, (byte) 0x01, (byte) 0x02, (byte) 0x03, (byte) 0x04,
+                (byte) 0x05, (byte) 0x06, 0x00, 0x00, 0x00, 0x00};
+        return request;
+    }
+
+    private static ClassifySessionDump createClassifySessionDumpRequest(final int newTableIndex) {
+        ClassifySessionDump request = new ClassifySessionDump();
+        request.tableId = newTableIndex;
+        return request;
+    }
+
+    private static InputAclSetInterface aclSetInterface() {
+        InputAclSetInterface request = new InputAclSetInterface();
+        request.isAdd = 1;
+        request.swIfIndex = LOCAL0_IFACE_ID;
+        request.ip4TableIndex = ~0; // skip
+        request.ip6TableIndex = ~0; // skip
+        request.l2TableIndex = 0;
+        return request;
+    }
+
+    private static ClassifyTableByInterface createClassifyTableByInterfaceRequest() {
+        ClassifyTableByInterface request = new ClassifyTableByInterface();
+        request.swIfIndex = LOCAL0_IFACE_ID;
+        return request;
+    }
+
+    private static void print(ClassifyAddDelTableReply reply) {
+        System.out.printf("ClassifyAddDelTableReply: %s%n", reply);
+    }
+
+    private static void print(ClassifyTableIdsReply reply) {
+        System.out.printf("ClassifyTableIdsReply: %s%n", reply);
+    }
+
+    private static void print(final ClassifyTableInfoReply reply) {
+        System.out.println(reply);
+        if (reply != null) {
+            System.out.println("Mask hex: " + DatatypeConverter.printHexBinary(reply.mask));
+        }
+    }
+
+    private static void print(ClassifyAddDelSessionReply reply) {
+        System.out.printf("ClassifyAddDelSessionReply: context=%s%n", reply);
+    }
+
+    private static void print(final ClassifySessionDetailsReplyDump reply) {
+        System.out.println(reply);
+        reply.classifySessionDetails.forEach(detail -> {
+            System.out.println(detail);
+            System.out.println("Match hex: " + DatatypeConverter.printHexBinary(detail.match));
+        });
+    }
+
+    private static void print(final InputAclSetInterfaceReply reply) {
+        System.out.printf("InputAclSetInterfaceReply: context=%s%n", reply);
+    }
+
+    private static void print(final ClassifyTableByInterfaceReply reply) {
+        System.out.printf("ClassifyAddDelTableReply: %s%n", reply);
+    }
+
+    private static void testL2Acl() throws Exception {
+        System.out.println("Testing L2 ACLs using Java callback API");
+        try (final JVppRegistry registry = new JVppRegistryImpl("L2AclTest");
+             final FutureJVppCoreFacade jvppFacade = new FutureJVppCoreFacade(registry, new JVppCoreImpl())) {
+
+            System.out.println("Successfully connected to VPP");
+            Thread.sleep(1000);
+
+            final ClassifyAddDelTableReply classifyAddDelTableReply =
+                jvppFacade.classifyAddDelTable(createClassifyTable()).toCompletableFuture().get();
+            print(classifyAddDelTableReply);
+
+            final ClassifyTableIdsReply classifyTableIdsReply =
+                jvppFacade.classifyTableIds(new ClassifyTableIds()).toCompletableFuture().get();
+            print(classifyTableIdsReply);
+
+            final ClassifyTableInfoReply classifyTableInfoReply =
+                jvppFacade.classifyTableInfo(createClassifyTableInfoRequest(classifyAddDelTableReply.newTableIndex))
+                    .toCompletableFuture().get();
+            print(classifyTableInfoReply);
+
+            final ClassifyAddDelSessionReply classifyAddDelSessionReply =
+                jvppFacade.classifyAddDelSession(createClassifySession(classifyAddDelTableReply.newTableIndex))
+                    .toCompletableFuture().get();
+            print(classifyAddDelSessionReply);
+
+            final ClassifySessionDetailsReplyDump classifySessionDetailsReplyDump =
+                jvppFacade.classifySessionDump(createClassifySessionDumpRequest(classifyAddDelTableReply.newTableIndex))
+                    .toCompletableFuture().get();
+            print(classifySessionDetailsReplyDump);
+
+            final InputAclSetInterfaceReply inputAclSetInterfaceReply =
+                jvppFacade.inputAclSetInterface(aclSetInterface()).toCompletableFuture().get();
+            print(inputAclSetInterfaceReply);
+
+            final ClassifyTableByInterfaceReply classifyTableByInterfaceReply =
+                jvppFacade.classifyTableByInterface(createClassifyTableByInterfaceRequest()).toCompletableFuture()
+                    .get();
+            print(classifyTableByInterfaceReply);
+
+            System.out.println("Disconnecting...");
+        }
+        Thread.sleep(1000);
+    }
+
+    public static void main(String[] args) throws Exception {
+        testL2Acl();
+    }
+}
diff --git a/src/vpp-api/java/jvpp-core/io/fd/vpp/jvpp/core/test/LispAdjacencyTest.java b/src/vpp-api/java/jvpp-core/io/fd/vpp/jvpp/core/test/LispAdjacencyTest.java
new file mode 100644
index 00000000..d7f5039b
--- /dev/null
+++ b/src/vpp-api/java/jvpp-core/io/fd/vpp/jvpp/core/test/LispAdjacencyTest.java
@@ -0,0 +1,124 @@
+/*
+ * Copyright (c) 2016 Cisco and/or its affiliates.
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at:
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+package io.fd.vpp.jvpp.core.test;
+
+import io.fd.vpp.jvpp.JVppRegistry;
+import io.fd.vpp.jvpp.JVppRegistryImpl;
+import io.fd.vpp.jvpp.core.JVppCoreImpl;
+import io.fd.vpp.jvpp.core.dto.LispAddDelAdjacency;
+import io.fd.vpp.jvpp.core.dto.LispAddDelLocalEid;
+import io.fd.vpp.jvpp.core.dto.LispAddDelLocatorSet;
+import io.fd.vpp.jvpp.core.dto.LispAddDelRemoteMapping;
+import io.fd.vpp.jvpp.core.dto.LispAdjacenciesGet;
+import io.fd.vpp.jvpp.core.dto.LispAdjacenciesGetReply;
+import io.fd.vpp.jvpp.core.dto.LispEnableDisable;
+import io.fd.vpp.jvpp.core.future.FutureJVppCoreFacade;
+import java.nio.charset.StandardCharsets;
+import java.util.concurrent.ExecutionException;
+import java.util.logging.Logger;
+
+/**
+ * Tests lisp adjacency creation and read (custom vpe.api type support showcase).
+ */
+public class LispAdjacencyTest {
+
+    private static final Logger LOG = Logger.getLogger(LispAdjacencyTest.class.getName());
+
+    private static void enableLisp(final FutureJVppCoreFacade jvpp) throws ExecutionException, InterruptedException {
+        final LispEnableDisable request = new LispEnableDisable();
+        request.isEn = 1;
+        jvpp.lispEnableDisable(request).toCompletableFuture().get();
+        LOG.info("Lisp enabled successfully");
+    }
+
+    private static void addLocatorSet(final FutureJVppCoreFacade jvpp) throws ExecutionException, InterruptedException {
+        final LispAddDelLocatorSet request = new LispAddDelLocatorSet();
+        request.isAdd = 1;
+        request.locatorSetName = "ls1".getBytes(StandardCharsets.UTF_8);
+        jvpp.lispAddDelLocatorSet(request).toCompletableFuture().get();
+        LOG.info("Locator set created successfully:" + request.toString());
+    }
+
+    private static void addLocalEid(final FutureJVppCoreFacade jvpp) throws ExecutionException, InterruptedException {
+        final LispAddDelLocalEid request = new LispAddDelLocalEid();
+        request.isAdd = 1;
+        request.locatorSetName = "ls1".getBytes(StandardCharsets.UTF_8);
+        request.eid = new byte[] {1, 2, 1, 10};
+        request.eidType = 0; // ip4
+        request.vni = 0;
+        request.prefixLen = 32;
+        jvpp.lispAddDelLocalEid(request).toCompletableFuture().get();
+        LOG.info("Local EID created successfully:" + request.toString());
+    }
+
+    private static void addRemoteMapping(final FutureJVppCoreFacade jvpp)
+        throws ExecutionException, InterruptedException {
+        final LispAddDelRemoteMapping request = new LispAddDelRemoteMapping();
+        request.isAdd = 1;
+        request.vni = 0;
+        request.eid = new byte[] {1, 2, 1, 20};
+        request.eidLen = 32;
+        request.rlocNum = 1;
+        request.rlocs = new byte[] {1, 1, 1, 1, 2, 1, 20, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0};
+        jvpp.lispAddDelRemoteMapping(request).toCompletableFuture().get();
+        LOG.info("Remote mapping created successfully:" + request.toString());
+    }
+
+    private static void addAdjacency(final FutureJVppCoreFacade jvpp) throws ExecutionException, InterruptedException {
+        final LispAddDelAdjacency request = new LispAddDelAdjacency();
+        request.isAdd = 1;
+        request.leid = new byte[] {1, 2, 1, 10};
+        request.leidLen = 32;
+        request.reid = new byte[] {1, 2, 1, 20};
+        request.reidLen = 32;
+        request.eidType = 0; // ip4
+        request.vni = 0;
+        jvpp.lispAddDelAdjacency(request).toCompletableFuture().get();
+        LOG.info("Lisp adjacency created successfully:" + request.toString());
+    }
+
+    private static void showAdjacencies(final FutureJVppCoreFacade jvpp)
+        throws ExecutionException, InterruptedException {
+        final LispAdjacenciesGetReply reply =
+            jvpp.lispAdjacenciesGet(new LispAdjacenciesGet()).toCompletableFuture().get();
+        LOG.info("Lisp adjacency received successfully:" + reply.toString());
+    }
+
+    private static void testAdjacency(final FutureJVppCoreFacade jvpp) throws Exception {
+        enableLisp(jvpp);
+        addLocatorSet(jvpp);
+        addLocalEid(jvpp);
+        addRemoteMapping(jvpp);
+        addAdjacency(jvpp);
+        showAdjacencies(jvpp);
+    }
+
+    private static void testFutureApi() throws Exception {
+        LOG.info("Create lisp adjacency test");
+        try (final JVppRegistry registry = new JVppRegistryImpl("LispAdjacencyTest");
+             final FutureJVppCoreFacade jvppFacade = new FutureJVppCoreFacade(registry, new JVppCoreImpl())) {
+            LOG.info("Successfully connected to VPP");
+
+            testAdjacency(jvppFacade);
+            LOG.info("Disconnecting...");
+        }
+    }
+
+    public static void main(String[] args) throws Exception {
+        testFutureApi();
+    }
+}
diff --git a/src/vpp-api/java/jvpp-core/io/fd/vpp/jvpp/core/test/NotificationUtils.java b/src/vpp-api/java/jvpp-core/io/fd/vpp/jvpp/core/test/NotificationUtils.java
new file mode 100644
index 00000000..f82946c3
--- /dev/null
+++ b/src/vpp-api/java/jvpp-core/io/fd/vpp/jvpp/core/test/NotificationUtils.java
@@ -0,0 +1,53 @@
+/*
+ * Copyright (c) 2016 Cisco and/or its affiliates.
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at:
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+package io.fd.vpp.jvpp.core.test;
+
+import java.io.PrintStream;
+import io.fd.vpp.jvpp.core.dto.SwInterfaceSetFlags;
+import io.fd.vpp.jvpp.core.dto.SwInterfaceSetFlagsNotification;
+import io.fd.vpp.jvpp.core.dto.WantInterfaceEvents;
+
+final class NotificationUtils {
+
+    private NotificationUtils() {}
+
+    static PrintStream printNotification(final SwInterfaceSetFlagsNotification msg) {
+        return System.out.printf("Received interface notification: ifc: %s%n", msg);
+    }
+
+    static SwInterfaceSetFlags getChangeInterfaceState() {
+        final SwInterfaceSetFlags swInterfaceSetFlags = new SwInterfaceSetFlags();
+        swInterfaceSetFlags.swIfIndex = 0;
+        swInterfaceSetFlags.adminUpDown = 1;
+        swInterfaceSetFlags.deleted = 0;
+        return swInterfaceSetFlags;
+    }
+
+    static WantInterfaceEvents getEnableInterfaceNotificationsReq() {
+        WantInterfaceEvents wantInterfaceEvents = new WantInterfaceEvents();
+        wantInterfaceEvents.pid = 1;
+        wantInterfaceEvents.enableDisable = 1;
+        return wantInterfaceEvents;
+    }
+
+    static WantInterfaceEvents getDisableInterfaceNotificationsReq() {
+        WantInterfaceEvents wantInterfaceEvents = new WantInterfaceEvents();
+        wantInterfaceEvents.pid = 1;
+        wantInterfaceEvents.enableDisable = 0;
+        return wantInterfaceEvents;
+    }
+}
diff --git a/src/vpp-api/java/jvpp-core/io/fd/vpp/jvpp/core/test/Readme.txt b/src/vpp-api/java/jvpp-core/io/fd/vpp/jvpp/core/test/Readme.txt
new file mode 100644
index 00000000..1344dc9e
--- /dev/null
+++ b/src/vpp-api/java/jvpp-core/io/fd/vpp/jvpp/core/test/Readme.txt
@@ -0,0 +1,17 @@
+This package contains basic tests for jvpp. To run the tests:
+
+- Make sure VPP is running
+- From VPP's build-root/ folder execute:
+  - sudo java -cp build-vpp-native/vpp-api/java/jvpp-registry-17.01.jar:build-vpp-native/vpp-api/java/jvpp-core-17.01.jar io.fd.vpp.jvpp.core.test.[test name]
+
+Available tests:
+CallbackApiTest - Similar to ControlPingTest, invokes more complex calls (e.g. interface dump) using low level JVpp APIs
+CallbackJVppFacadeNotificationTest - Tests interface notifications using Callback based JVpp facade
+CallbackJVppFacadeTest - Execution of more complex calls using Callback based JVpp facade
+CallbackNotificationApiTest - Tests interface notifications using low level JVpp APIs
+ControlPingTest - Simple test executing a single control ping using low level JVpp APIs
+CreateSubInterfaceTest - Tests sub-interface creation
+FutureApiNotificationTest - Tests interface notifications using Future based JVpp facade
+FutureApiTest - Execution of more complex calls using Future based JVpp facade
+L2AclTest - Tests L2 ACL creation
+LispAdjacencyTest - Tests lisp adjacency creation and read (custom vpe.api type support showcase)
diff --git a/src/vpp-api/java/jvpp-core/jvpp_core.c b/src/vpp-api/java/jvpp-core/jvpp_core.c
new file mode 100644
index 00000000..ef4cb8e3
--- /dev/null
+++ b/src/vpp-api/java/jvpp-core/jvpp_core.c
@@ -0,0 +1,117 @@
+/*
+ * Copyright (c) 2016 Cisco and/or its affiliates.
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at:
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#include 
+
+#include 
+#define vl_typedefs             /* define message structures */
+#include 
+#undef vl_typedefs
+
+#define vl_endianfun
+#include 
+#undef vl_endianfun
+
+#define vl_print(handle, ...)
+#define vl_printfun
+#include 
+#undef vl_printfun
+
+#include 
+#include 
+#include 
+#include 
+
+#include 
+
+// TODO: generate jvpp_plugin_name.c files (or at least reuse plugin's main structure)
+typedef struct {
+    /* Base message index for the jvpp-core plugin */
+    u16 msg_id_base;
+
+    /* Pointer to shared memory queue */
+    unix_shared_memory_queue_t * vl_input_queue;
+
+    /* VPP api client index */
+    u32 my_client_index;
+
+    /* Callback object and class references enabling asynchronous Java calls */
+    jobject callbackObject;
+    jclass callbackClass;
+
+} core_main_t;
+
+core_main_t core_main __attribute__((aligned (64)));
+
+#include "io_fd_vpp_jvpp_core_JVppCoreImpl.h"
+#include "jvpp_core_gen.h"
+
+JNIEXPORT void JNICALL Java_io_fd_vpp_jvpp_core_JVppCoreImpl_init0
+(JNIEnv * env, jclass clazz, jobject callback, jlong queue_address, jint my_client_index) {
+    core_main_t * plugin_main = &core_main;
+    plugin_main->my_client_index = my_client_index;
+    plugin_main->vl_input_queue = (unix_shared_memory_queue_t *)queue_address;
+
+    plugin_main->callbackObject = (*env)->NewGlobalRef(env, callback);
+    plugin_main->callbackClass = (jclass)(*env)->NewGlobalRef(env, (*env)->GetObjectClass(env, callback));
+
+    #define _(N,n)                                  \
+        vl_msg_api_set_handlers(VL_API_##N, #n,     \
+                vl_api_##n##_t_handler,             \
+                vl_noop_handler,                    \
+                vl_noop_handler,              \
+                vl_noop_handler,               \
+                sizeof(vl_api_##n##_t), 1);
+        foreach_api_reply_handler;
+    #undef _
+}
+
+JNIEXPORT void JNICALL Java_io_fd_vpp_jvpp_core_JVppCoreImpl_close0
+(JNIEnv *env, jclass clazz) {
+    core_main_t * plugin_main = &core_main;
+
+    // cleanup:
+    (*env)->DeleteGlobalRef(env, plugin_main->callbackClass);
+    (*env)->DeleteGlobalRef(env, plugin_main->callbackObject);
+
+    plugin_main->callbackClass = NULL;
+    plugin_main->callbackObject = NULL;
+}
+
+jint JNI_OnLoad(JavaVM *vm, void *reserved) {
+    JNIEnv* env;
+
+    if ((*vm)->GetEnv(vm, (void**) &env, JNI_VERSION_1_8) != JNI_OK) {
+        return JNI_EVERSION;
+    }
+
+    if (cache_class_references(env) != 0) {
+        clib_warning ("Failed to cache class references\n");
+        return JNI_ERR;
+    }
+
+    return JNI_VERSION_1_8;
+}
+
+void JNI_OnUnload(JavaVM *vm, void *reserved) {
+    JNIEnv* env;
+    if ((*vm)->GetEnv(vm, (void**) &env, JNI_VERSION_1_8) != JNI_OK) {
+        return;
+    }
+    delete_class_references(env);
+}
+
+
+
diff --git a/src/vpp-api/java/jvpp-ioamexport/io/fd/vpp/jvpp/ioamexport/test/IoamExportApiTest.java b/src/vpp-api/java/jvpp-ioamexport/io/fd/vpp/jvpp/ioamexport/test/IoamExportApiTest.java
new file mode 100644
index 00000000..cb85f005
--- /dev/null
+++ b/src/vpp-api/java/jvpp-ioamexport/io/fd/vpp/jvpp/ioamexport/test/IoamExportApiTest.java
@@ -0,0 +1,56 @@
+/*
+ * Copyright (c) 2016 Cisco and/or its affiliates.
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at:
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+package io.fd.vpp.jvpp.ioamexport.test;
+
+import java.net.InetAddress;
+
+import io.fd.vpp.jvpp.JVpp;
+import io.fd.vpp.jvpp.JVppRegistry;
+import io.fd.vpp.jvpp.JVppRegistryImpl;
+import io.fd.vpp.jvpp.VppCallbackException;
+import io.fd.vpp.jvpp.ioamexport.JVppIoamexportImpl;
+import io.fd.vpp.jvpp.ioamexport.future.FutureJVppIoamexportFacade;
+import io.fd.vpp.jvpp.ioamexport.dto.IoamExportIp6EnableDisable;
+import io.fd.vpp.jvpp.ioamexport.dto.IoamExportIp6EnableDisableReply;
+
+public class IoamExportApiTest {
+
+    public static void main(String[] args) throws Exception {
+        ioamExportTestApi();
+    }
+
+    private static void ioamExportTestApi() throws Exception {
+        System.out.println("Testing Java API for ioam export plugin");
+        try (final JVppRegistry registry = new JVppRegistryImpl("ioamExportApiTest");
+             final JVpp jvpp = new JVppIoamexportImpl()) {
+	    FutureJVppIoamexportFacade ioamexportJvpp = new FutureJVppIoamexportFacade(registry,jvpp);
+            System.out.println("Sending ioam export request...");
+	    IoamExportIp6EnableDisable request = new IoamExportIp6EnableDisable();
+            request.isDisable = 0;
+	    InetAddress collectorAddress = InetAddress.getByName("2001:0DB8:AC10:FE01:0000:0000:0000:0000");
+	    InetAddress srcAddress = InetAddress.getByName("2001:0DB8:AC10:FE01:0000:0000:0000:0001");
+	    request.collectorAddress = collectorAddress.getAddress();
+	    request.srcAddress = srcAddress.getAddress();
+	    IoamExportIp6EnableDisableReply reply = ioamexportJvpp.ioamExportIp6EnableDisable(request).toCompletableFuture().get();
+            System.out.printf("IoamExportIp6EnableDisableReply = "+reply.toString()+"%n");
+
+            Thread.sleep(1000);
+
+            System.out.println("Disconnecting...");
+        }
+    }
+}
diff --git a/src/vpp-api/java/jvpp-ioamexport/io/fd/vpp/jvpp/ioamexport/test/Readme.txt b/src/vpp-api/java/jvpp-ioamexport/io/fd/vpp/jvpp/ioamexport/test/Readme.txt
new file mode 100644
index 00000000..1b38c285
--- /dev/null
+++ b/src/vpp-api/java/jvpp-ioamexport/io/fd/vpp/jvpp/ioamexport/test/Readme.txt
@@ -0,0 +1 @@
+sudo java -cp build-vpp_debug-native/vpp-api/java/jvpp-registry-17.01.jar:build-vpp_debug-native/plugins/ioam-plugin/jvpp-ioam-export-1.0.jar io.fd.vpp.jvpp.ioamexport.test.IoamExportApiTest
diff --git a/src/vpp-api/java/jvpp-ioamexport/jvpp_ioam_export.c b/src/vpp-api/java/jvpp-ioamexport/jvpp_ioam_export.c
new file mode 100644
index 00000000..5cda89d1
--- /dev/null
+++ b/src/vpp-api/java/jvpp-ioamexport/jvpp_ioam_export.c
@@ -0,0 +1,124 @@
+/*
+ * Copyright (c) 2016 Cisco and/or its affiliates.
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at:
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#include 
+
+#include 
+#define vl_typedefs             /* define message structures */
+#include 
+#undef vl_typedefs
+
+#define vl_endianfun
+#include 
+#undef vl_endianfun
+
+#define vl_print(handle, ...)
+#define vl_printfun
+#include 
+#undef vl_printfun
+
+/* Get the API version number */
+#define vl_api_version(n,v) static u32 api_version=(v);
+#include 
+#undef vl_api_version
+
+#include 
+#include 
+#include 
+
+#if VPPJNI_DEBUG == 1
+  #define DEBUG_LOG(...) clib_warning(__VA_ARGS__)
+#else
+  #define DEBUG_LOG(...)
+#endif
+
+#include 
+
+#include "jvpp-ioamexport/io_fd_vpp_jvpp_ioamexport_JVppIoamexportImpl.h"
+#include "jvpp_ioam_export.h"
+#include "jvpp-ioamexport/jvpp_ioamexport_gen.h"
+
+/*
+ * Class:     io_fd_vpp_jvpp_ioamexport_JVppIoamexportImpl
+ * Method:    init0
+ * Signature: (JI)V
+ */
+JNIEXPORT void JNICALL Java_io_fd_vpp_jvpp_ioamexport_JVppIoamexportImpl_init0
+  (JNIEnv *env, jclass clazz, jobject callback, jlong queue_address, jint my_client_index) {
+  ioamexport_main_t * plugin_main = &ioamexport_main;
+  u8 * name;
+  clib_warning ("Java_io_fd_vpp_jvpp_ioamexport_JVppIoamexportImpl_init0");
+
+  plugin_main->my_client_index = my_client_index;
+  plugin_main->vl_input_queue = (unix_shared_memory_queue_t *)queue_address;
+
+  name = format (0, "ioam_export_%08x%c", api_version, 0);
+  plugin_main->msg_id_base = vl_client_get_first_plugin_msg_id ((char *) name);
+
+  if (plugin_main->msg_id_base == (u16) ~0) {
+    jclass exClass = (*env)->FindClass(env, "java/lang/IllegalStateException");
+    (*env)->ThrowNew(env, exClass, "ioam_export plugin is not loaded in VPP");
+  } else {
+    plugin_main->callbackObject = (*env)->NewGlobalRef(env, callback);
+    plugin_main->callbackClass = (jclass)(*env)->NewGlobalRef(env, (*env)->GetObjectClass(env, callback));
+
+    #define _(N,n)                                  \
+        vl_msg_api_set_handlers(VL_API_##N + plugin_main->msg_id_base, #n,     \
+                vl_api_##n##_t_handler,             \
+                vl_noop_handler,                    \
+                vl_api_##n##_t_endian,              \
+                vl_api_##n##_t_print,               \
+                sizeof(vl_api_##n##_t), 1);
+        foreach_api_reply_handler;
+    #undef _
+  }
+}
+
+JNIEXPORT void JNICALL Java_io_fd_vpp_jvpp_ioamexport_JVppIoamexportImpl_close0
+(JNIEnv *env, jclass clazz) {
+  ioamexport_main_t * plugin_main = &ioamexport_main;
+
+    // cleanup:
+    (*env)->DeleteGlobalRef(env, plugin_main->callbackClass);
+    (*env)->DeleteGlobalRef(env, plugin_main->callbackObject);
+
+    plugin_main->callbackClass = NULL;
+    plugin_main->callbackObject = NULL;
+}
+
+/* Attach thread to JVM and cache class references when initiating JVPP iOAM EXPORT */
+jint JNI_OnLoad(JavaVM *vm, void *reserved) {
+    JNIEnv* env;
+
+    if ((*vm)->GetEnv(vm, (void**) &env, JNI_VERSION_1_8) != JNI_OK) {
+        return JNI_EVERSION;
+    }
+
+    if (cache_class_references(env) != 0) {
+        clib_warning ("Failed to cache class references\n");
+        return JNI_ERR;
+    }
+
+    return JNI_VERSION_1_8;
+}
+
+/* Clean up cached references when disposing JVPP iOAM EXPORT */
+void JNI_OnUnload(JavaVM *vm, void *reserved) {
+    JNIEnv* env;
+    if ((*vm)->GetEnv(vm, (void**) &env, JNI_VERSION_1_8) != JNI_OK) {
+        return;
+    }
+    delete_class_references(env);
+}
diff --git a/src/vpp-api/java/jvpp-ioamexport/jvpp_ioam_export.h b/src/vpp-api/java/jvpp-ioamexport/jvpp_ioam_export.h
new file mode 100644
index 00000000..b6c0c16e
--- /dev/null
+++ b/src/vpp-api/java/jvpp-ioamexport/jvpp_ioam_export.h
@@ -0,0 +1,45 @@
+/*
+ * Copyright (c) 2016 Cisco and/or its affiliates.
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at:
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+#ifndef __included_jvpp_ioam_export_h__
+#define __included_jvpp_ioam_export_h__
+
+#include 
+#include 
+#include 
+#include 
+#include 
+#include 
+
+/* Global state for JVPP-IOAM-EXPORT */
+typedef struct {
+    /* Base message index for the export plugin */
+    u16 msg_id_base;
+
+    /* Pointer to shared memory queue */
+    unix_shared_memory_queue_t * vl_input_queue;
+
+    /* VPP api client index */
+    u32 my_client_index;
+
+    /* Callback object and class references enabling asynchronous Java calls */
+    jobject callbackObject;
+    jclass callbackClass;
+
+} ioamexport_main_t;
+
+ioamexport_main_t ioamexport_main __attribute__((aligned (64)));
+
+
+#endif /* __included_jvpp_ioam_export_h__ */
diff --git a/src/vpp-api/java/jvpp-ioampot/io/fd/vpp/jvpp/ioampot/test/IoamPotApiTest.java b/src/vpp-api/java/jvpp-ioampot/io/fd/vpp/jvpp/ioampot/test/IoamPotApiTest.java
new file mode 100644
index 00000000..74eb86a1
--- /dev/null
+++ b/src/vpp-api/java/jvpp-ioampot/io/fd/vpp/jvpp/ioampot/test/IoamPotApiTest.java
@@ -0,0 +1,75 @@
+/*
+ * Copyright (c) 2016 Cisco and/or its affiliates.
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at:
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+package io.fd.vpp.jvpp.ioampot.test;
+
+import io.fd.vpp.jvpp.JVpp;
+import io.fd.vpp.jvpp.JVppRegistry;
+import io.fd.vpp.jvpp.JVppRegistryImpl;
+import io.fd.vpp.jvpp.VppCallbackException;
+import io.fd.vpp.jvpp.ioampot.JVppIoampotImpl;
+import io.fd.vpp.jvpp.ioampot.callback.PotProfileAddCallback;
+import io.fd.vpp.jvpp.ioampot.dto.PotProfileAdd;
+import io.fd.vpp.jvpp.ioampot.dto.PotProfileAddReply;
+
+public class IoamPotApiTest {
+
+    static class IoamPotTestCallback implements PotProfileAddCallback {
+
+        @Override
+        public void onPotProfileAddReply(final PotProfileAddReply reply) {
+            System.out.printf("Received PotProfileAddReply reply: context=%d%n",
+                reply.context);
+        }
+
+        @Override
+        public void onError(VppCallbackException ex) {
+            System.out.printf("Received onError exception: call=%s, context=%d, retval=%d%n", ex.getMethodName(),
+                ex.getCtxId(), ex.getErrorCode());
+        }
+    }
+
+    public static void main(String[] args) throws Exception {
+        ioamPotTestApi();
+    }
+
+    private static void ioamPotTestApi() throws Exception {
+        System.out.println("Testing Java API for ioam pot plugin");
+        try (final JVppRegistry registry = new JVppRegistryImpl("ioamPotApiTest");
+             final JVpp jvpp = new JVppIoampotImpl()) {
+            registry.register(jvpp, new IoamPotTestCallback());
+
+            System.out.println("Sending ioam pot profile add request...");
+            PotProfileAdd request = new PotProfileAdd();
+            request.id = 0;
+            request.validator = 4;
+            request.secretKey = 1;
+            request.secretShare = 2;
+            request.prime = 1234;
+            request.maxBits = 53;
+            request.lpc = 1234;
+            request.polynomialPublic = 1234;
+            request.listNameLen = (byte)"test pot profile".getBytes().length;
+            request.listName = "test pot profile".getBytes();
+            final int result = jvpp.send(request);
+            System.out.printf("PotProfileAdd send result = %d%n", result);
+
+            Thread.sleep(1000);
+
+            System.out.println("Disconnecting...");
+        }
+    }
+}
diff --git a/src/vpp-api/java/jvpp-ioampot/io/fd/vpp/jvpp/ioampot/test/Readme.txt b/src/vpp-api/java/jvpp-ioampot/io/fd/vpp/jvpp/ioampot/test/Readme.txt
new file mode 100644
index 00000000..2323494d
--- /dev/null
+++ b/src/vpp-api/java/jvpp-ioampot/io/fd/vpp/jvpp/ioampot/test/Readme.txt
@@ -0,0 +1 @@
+sudo java -cp build-vpp_debug-native/vpp-api/java/jvpp-registry-16.12.jar:build-vpp_debug-native/plugins/ioam-plugin/jvpp-ioam-pot-1.0.jar io.fd.vpp.jvpp.ioampot.test.IoamPotApiTest
diff --git a/src/vpp-api/java/jvpp-ioampot/jvpp_ioam_pot.c b/src/vpp-api/java/jvpp-ioampot/jvpp_ioam_pot.c
new file mode 100644
index 00000000..9291dbba
--- /dev/null
+++ b/src/vpp-api/java/jvpp-ioampot/jvpp_ioam_pot.c
@@ -0,0 +1,124 @@
+/*
+ * Copyright (c) 2016 Cisco and/or its affiliates.
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at:
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#include 
+
+#include 
+#define vl_typedefs             /* define message structures */
+#include 
+#undef vl_typedefs
+
+#define vl_endianfun
+#include 
+#undef vl_endianfun
+
+#define vl_print(handle, ...)
+#define vl_printfun
+#include 
+#undef vl_printfun
+
+/* Get the API version number */
+#define vl_api_version(n,v) static u32 api_version=(v);
+#include 
+#undef vl_api_version
+
+#include 
+#include 
+#include 
+
+#if VPPJNI_DEBUG == 1
+  #define DEBUG_LOG(...) clib_warning(__VA_ARGS__)
+#else
+  #define DEBUG_LOG(...)
+#endif
+
+#include 
+
+#include "jvpp-ioampot/io_fd_vpp_jvpp_ioampot_JVppIoampotImpl.h"
+#include "jvpp_ioam_pot.h"
+#include "jvpp-ioampot/jvpp_ioampot_gen.h"
+
+/*
+ * Class:     io_fd_vpp_jvpp_ioampot_JVppIoampotImpl
+ * Method:    init0
+ * Signature: (JI)V
+ */
+JNIEXPORT void JNICALL Java_io_fd_vpp_jvpp_ioampot_JVppIoampotImpl_init0
+  (JNIEnv *env, jclass clazz, jobject callback, jlong queue_address, jint my_client_index) {
+  ioampot_main_t * plugin_main = &ioampot_main;
+  u8 * name;
+  clib_warning ("Java_io_fd_vpp_jvpp_ioampot_JVppIoampotImpl_init0");
+
+  plugin_main->my_client_index = my_client_index;
+  plugin_main->vl_input_queue = (unix_shared_memory_queue_t *)queue_address;
+
+  name = format (0, "ioam_pot_%08x%c", api_version, 0);
+  plugin_main->msg_id_base = vl_client_get_first_plugin_msg_id ((char *) name);
+
+  if (plugin_main->msg_id_base == (u16) ~0) {
+    jclass exClass = (*env)->FindClass(env, "java/lang/IllegalStateException");
+    (*env)->ThrowNew(env, exClass, "ioam_pot plugin is not loaded in VPP");
+  } else {
+    plugin_main->callbackObject = (*env)->NewGlobalRef(env, callback);
+    plugin_main->callbackClass = (jclass)(*env)->NewGlobalRef(env, (*env)->GetObjectClass(env, callback));
+
+    #define _(N,n)                                  \
+        vl_msg_api_set_handlers(VL_API_##N + plugin_main->msg_id_base, #n,     \
+                vl_api_##n##_t_handler,             \
+                vl_noop_handler,                    \
+                vl_api_##n##_t_endian,              \
+                vl_api_##n##_t_print,               \
+                sizeof(vl_api_##n##_t), 1);
+        foreach_api_reply_handler;
+    #undef _
+  }
+}
+
+JNIEXPORT void JNICALL Java_io_fd_vpp_jvpp_ioampot_JVppIoampotImpl_close0
+(JNIEnv *env, jclass clazz) {
+  ioampot_main_t * plugin_main = &ioampot_main;
+
+    // cleanup:
+    (*env)->DeleteGlobalRef(env, plugin_main->callbackClass);
+    (*env)->DeleteGlobalRef(env, plugin_main->callbackObject);
+
+    plugin_main->callbackClass = NULL;
+    plugin_main->callbackObject = NULL;
+}
+
+/* Attach thread to JVM and cache class references when initiating JVPP iOAM POT */
+jint JNI_OnLoad(JavaVM *vm, void *reserved) {
+    JNIEnv* env;
+
+    if ((*vm)->GetEnv(vm, (void**) &env, JNI_VERSION_1_8) != JNI_OK) {
+        return JNI_EVERSION;
+    }
+
+    if (cache_class_references(env) != 0) {
+        clib_warning ("Failed to cache class references\n");
+        return JNI_ERR;
+    }
+
+    return JNI_VERSION_1_8;
+}
+
+/* Clean up cached references when disposing JVPP iOAM POT */
+void JNI_OnUnload(JavaVM *vm, void *reserved) {
+    JNIEnv* env;
+    if ((*vm)->GetEnv(vm, (void**) &env, JNI_VERSION_1_8) != JNI_OK) {
+        return;
+    }
+    delete_class_references(env);
+}
diff --git a/src/vpp-api/java/jvpp-ioampot/jvpp_ioam_pot.h b/src/vpp-api/java/jvpp-ioampot/jvpp_ioam_pot.h
new file mode 100644
index 00000000..00aa51db
--- /dev/null
+++ b/src/vpp-api/java/jvpp-ioampot/jvpp_ioam_pot.h
@@ -0,0 +1,45 @@
+/*
+ * Copyright (c) 2016 Cisco and/or its affiliates.
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at:
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+#ifndef __included_jvpp_ioam_pot_h__
+#define __included_jvpp_ioam_pot_h__
+
+#include 
+#include 
+#include 
+#include 
+#include 
+#include 
+
+/* Global state for JVPP-IOAM-POT */
+typedef struct {
+    /* Base message index for the pot plugin */
+    u16 msg_id_base;
+
+    /* Pointer to shared memory queue */
+    unix_shared_memory_queue_t * vl_input_queue;
+
+    /* VPP api client index */
+    u32 my_client_index;
+
+    /* Callback object and class references enabling asynchronous Java calls */
+    jobject callbackObject;
+    jclass callbackClass;
+
+} ioampot_main_t;
+
+ioampot_main_t ioampot_main __attribute__((aligned (64)));
+
+
+#endif /* __included_jvpp_ioam_pot_h__ */
diff --git a/src/vpp-api/java/jvpp-ioamtrace/io/fd/vpp/jvpp/ioamtrace/test/IoamTraceApiTest.java b/src/vpp-api/java/jvpp-ioamtrace/io/fd/vpp/jvpp/ioamtrace/test/IoamTraceApiTest.java
new file mode 100644
index 00000000..bc8c1c3a
--- /dev/null
+++ b/src/vpp-api/java/jvpp-ioamtrace/io/fd/vpp/jvpp/ioamtrace/test/IoamTraceApiTest.java
@@ -0,0 +1,77 @@
+/*
+ * Copyright (c) 2016 Cisco and/or its affiliates.
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at:
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+package io.fd.vpp.jvpp.ioamtrace.test;
+
+import io.fd.vpp.jvpp.JVpp;
+import io.fd.vpp.jvpp.JVppRegistry;
+import io.fd.vpp.jvpp.JVppRegistryImpl;
+import io.fd.vpp.jvpp.VppCallbackException;
+import io.fd.vpp.jvpp.ioamtrace.future.FutureJVppIoamtraceFacade;
+import io.fd.vpp.jvpp.ioamtrace.JVppIoamtraceImpl;
+import io.fd.vpp.jvpp.ioamtrace.callback.TraceProfileAddCallback;
+import io.fd.vpp.jvpp.ioamtrace.dto.TraceProfileAdd;
+import io.fd.vpp.jvpp.ioamtrace.dto.TraceProfileAddReply;
+import io.fd.vpp.jvpp.ioamtrace.dto.TraceProfileShowConfig;
+import io.fd.vpp.jvpp.ioamtrace.dto.TraceProfileShowConfigReply;
+
+public class IoamTraceApiTest {
+
+    static class IoamTraceTestCallback implements TraceProfileAddCallback {
+
+        @Override
+        public void onTraceProfileAddReply(final TraceProfileAddReply reply) {
+            System.out.printf("Received TraceProfileAddReply reply: context=%d%n",
+                reply.context);
+        }
+
+        @Override
+        public void onError(VppCallbackException ex) {
+            System.out.printf("Received onError exception: call=%s, context=%d, retval=%d%n", ex.getMethodName(),
+                ex.getCtxId(), ex.getErrorCode());
+        }
+    }
+
+    public static void main(String[] args) throws Exception {
+        ioamTraceTestApi();
+    }
+
+    private static void ioamTraceTestApi() throws Exception {
+        System.out.println("Testing Java API for ioam trace plugin");
+        try (final JVppRegistry registry = new JVppRegistryImpl("ioamTraceApiTest");
+             final JVpp jvpp = new JVppIoamtraceImpl()) {
+	    FutureJVppIoamtraceFacade ioamtraceJvpp = new FutureJVppIoamtraceFacade(registry,jvpp);
+
+            System.out.println("Sending ioam trace profile add request...");
+            TraceProfileAdd request = new TraceProfileAdd();
+            request.traceType = 0x1f;
+            request.numElts = 4;
+            request.nodeId = 1;
+            request.traceTsp = 2;
+            request.appData = 1234;
+            final int result = jvpp.send(request);
+            System.out.printf("TraceProfileAdd send result = %d%n", result);
+
+            Thread.sleep(1000);
+
+	    TraceProfileShowConfig showRequest = new TraceProfileShowConfig();
+	    TraceProfileShowConfigReply reply = ioamtraceJvpp.traceProfileShowConfig(showRequest).toCompletableFuture().get();
+            System.out.printf("TraceProfileShowConfig result = "+ reply.toString());
+
+            System.out.println("Disconnecting...");
+        }
+    }
+}
diff --git a/src/vpp-api/java/jvpp-ioamtrace/io/fd/vpp/jvpp/ioamtrace/test/Readme.txt b/src/vpp-api/java/jvpp-ioamtrace/io/fd/vpp/jvpp/ioamtrace/test/Readme.txt
new file mode 100644
index 00000000..17e45a81
--- /dev/null
+++ b/src/vpp-api/java/jvpp-ioamtrace/io/fd/vpp/jvpp/ioamtrace/test/Readme.txt
@@ -0,0 +1 @@
+sudo java -cp build-vpp-native/vpp-api/java/jvpp-registry-17.01.jar:build-vpp-native/plugins/ioam-plugin/jvpp-ioam-trace-1.0.jar io.fd.vpp.jvpp.ioamtrace.test.IoamTraceApiTest
diff --git a/src/vpp-api/java/jvpp-ioamtrace/jvpp_ioam_trace.c b/src/vpp-api/java/jvpp-ioamtrace/jvpp_ioam_trace.c
new file mode 100644
index 00000000..0bf17889
--- /dev/null
+++ b/src/vpp-api/java/jvpp-ioamtrace/jvpp_ioam_trace.c
@@ -0,0 +1,124 @@
+/*
+ * Copyright (c) 2016 Cisco and/or its affiliates.
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at:
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#include 
+
+#include 
+#define vl_typedefs             /* define message structures */
+#include 
+#undef vl_typedefs
+
+#define vl_endianfun
+#include 
+#undef vl_endianfun
+
+#define vl_print(handle, ...)
+#define vl_printfun
+#include 
+#undef vl_printfun
+
+/* Get the API version number */
+#define vl_api_version(n,v) static u32 api_version=(v);
+#include 
+#undef vl_api_version
+
+#include 
+#include 
+#include 
+
+#if VPPJNI_DEBUG == 1
+  #define DEBUG_LOG(...) clib_warning(__VA_ARGS__)
+#else
+  #define DEBUG_LOG(...)
+#endif
+
+#include 
+
+#include "jvpp-ioamtrace/io_fd_vpp_jvpp_ioamtrace_JVppIoamtraceImpl.h"
+#include "jvpp_ioam_trace.h"
+#include "jvpp-ioamtrace/jvpp_ioamtrace_gen.h"
+
+/*
+ * Class:     io_fd_vpp_jvpp_ioamtrace_JVppIoamtraceImpl
+ * Method:    init0
+ * Signature: (JI)V
+ */
+JNIEXPORT void JNICALL Java_io_fd_vpp_jvpp_ioamtrace_JVppIoamtraceImpl_init0
+  (JNIEnv *env, jclass clazz, jobject callback, jlong queue_address, jint my_client_index) {
+  ioamtrace_main_t * plugin_main = &ioamtrace_main;
+  u8 * name;
+  clib_warning ("Java_io_fd_vpp_jvpp_ioamtrace_JVppIoamtraceImpl_init0");
+
+  plugin_main->my_client_index = my_client_index;
+  plugin_main->vl_input_queue = (unix_shared_memory_queue_t *)queue_address;
+
+  name = format (0, "ioam_trace_%08x%c", api_version, 0);
+  plugin_main->msg_id_base = vl_client_get_first_plugin_msg_id ((char *) name);
+
+  if (plugin_main->msg_id_base == (u16) ~0) {
+    jclass exClass = (*env)->FindClass(env, "java/lang/IllegalStateException");
+    (*env)->ThrowNew(env, exClass, "ioam_trace plugin is not loaded in VPP");
+  } else {
+    plugin_main->callbackObject = (*env)->NewGlobalRef(env, callback);
+    plugin_main->callbackClass = (jclass)(*env)->NewGlobalRef(env, (*env)->GetObjectClass(env, callback));
+
+    #define _(N,n)                                  \
+        vl_msg_api_set_handlers(VL_API_##N + plugin_main->msg_id_base, #n,     \
+                vl_api_##n##_t_handler,             \
+                vl_noop_handler,                    \
+                vl_api_##n##_t_endian,              \
+                vl_api_##n##_t_print,               \
+                sizeof(vl_api_##n##_t), 1);
+        foreach_api_reply_handler;
+    #undef _
+  }
+}
+
+JNIEXPORT void JNICALL Java_io_fd_vpp_jvpp_ioamtrace_JVppIoamtraceImpl_close0
+(JNIEnv *env, jclass clazz) {
+  ioamtrace_main_t * plugin_main = &ioamtrace_main;
+
+    // cleanup:
+    (*env)->DeleteGlobalRef(env, plugin_main->callbackClass);
+    (*env)->DeleteGlobalRef(env, plugin_main->callbackObject);
+
+    plugin_main->callbackClass = NULL;
+    plugin_main->callbackObject = NULL;
+}
+
+/* Attach thread to JVM and cache class references when initiating JVPP iOAM Trace */
+jint JNI_OnLoad(JavaVM *vm, void *reserved) {
+    JNIEnv* env;
+
+    if ((*vm)->GetEnv(vm, (void**) &env, JNI_VERSION_1_8) != JNI_OK) {
+        return JNI_EVERSION;
+    }
+
+    if (cache_class_references(env) != 0) {
+        clib_warning ("Failed to cache class references\n");
+        return JNI_ERR;
+    }
+
+    return JNI_VERSION_1_8;
+}
+
+/* Clean up cached references when disposing JVPP iOAM Trace */
+void JNI_OnUnload(JavaVM *vm, void *reserved) {
+    JNIEnv* env;
+    if ((*vm)->GetEnv(vm, (void**) &env, JNI_VERSION_1_8) != JNI_OK) {
+        return;
+    }
+    delete_class_references(env);
+}
diff --git a/src/vpp-api/java/jvpp-ioamtrace/jvpp_ioam_trace.h b/src/vpp-api/java/jvpp-ioamtrace/jvpp_ioam_trace.h
new file mode 100644
index 00000000..9fc16c15
--- /dev/null
+++ b/src/vpp-api/java/jvpp-ioamtrace/jvpp_ioam_trace.h
@@ -0,0 +1,45 @@
+/*
+ * Copyright (c) 2016 Cisco and/or its affiliates.
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at:
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+#ifndef __included_jvpp_ioam_trace_h__
+#define __included_jvpp_ioam_trace_h__
+
+#include 
+#include 
+#include 
+#include 
+#include 
+#include 
+
+/* Global state for JVPP-IOAM-TRACE */
+typedef struct {
+    /* Base message index for the trace plugin */
+    u16 msg_id_base;
+
+    /* Pointer to shared memory queue */
+    unix_shared_memory_queue_t * vl_input_queue;
+
+    /* VPP api client index */
+    u32 my_client_index;
+
+    /* Callback object and class references enabling asynchronous Java calls */
+    jobject callbackObject;
+    jclass callbackClass;
+
+} ioamtrace_main_t;
+
+ioamtrace_main_t ioamtrace_main __attribute__((aligned (64)));
+
+
+#endif /* __included_jvpp_ioam_trace_h__ */
diff --git a/src/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/JVpp.java b/src/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/JVpp.java
new file mode 100644
index 00000000..55f25a7b
--- /dev/null
+++ b/src/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/JVpp.java
@@ -0,0 +1,56 @@
+/*
+ * Copyright (c) 2016 Cisco and/or its affiliates.
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at:
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+package io.fd.vpp.jvpp;
+
+import io.fd.vpp.jvpp.callback.JVppCallback;
+import io.fd.vpp.jvpp.dto.ControlPing;
+import io.fd.vpp.jvpp.dto.JVppRequest;
+
+/**
+ * Base interface for plugin's Java API.
+ */
+public interface JVpp extends AutoCloseable {
+
+    /**
+     * Sends request to vpp.
+     *
+     * @param request request to be sent
+     * @return unique identifer of message in message queue
+     * @throws VppInvocationException when message could not be sent
+     */
+    int send(final JVppRequest request) throws VppInvocationException;
+
+    /**
+     * Initializes plugin's Java API.
+     *
+     * @param registry     plugin registry
+     * @param callback     called by vpe.api message handlers
+     * @param queueAddress address of vpp shared memory queue
+     * @param clientIndex  vpp client identifier
+     */
+    void init(final JVppRegistry registry, final JVppCallback callback, final long queueAddress,
+              final int clientIndex);
+
+    /**
+     * Sends control_ping message.
+     *
+     * @param controlPing request DTO
+     * @return unique identifer of message in message queue
+     * @throws VppInvocationException when message could not be sent
+     */
+    int controlPing(final ControlPing controlPing) throws VppInvocationException;
+}
diff --git a/src/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/JVppRegistry.java b/src/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/JVppRegistry.java
new file mode 100644
index 00000000..6535db02
--- /dev/null
+++ b/src/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/JVppRegistry.java
@@ -0,0 +1,76 @@
+/*
+ * Copyright (c) 2016 Cisco and/or its affiliates.
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at:
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+package io.fd.vpp.jvpp;
+
+import io.fd.vpp.jvpp.callback.JVppCallback;
+
+/**
+ * Manages VPP connection and stores plugin callbacks.
+ */
+public interface JVppRegistry extends AutoCloseable {
+
+    /**
+     * Vpp connection managed by the registry.
+     *
+     * @return representation of vpp connection
+     */
+    VppConnection getConnection();
+
+    /**
+     * Registers callback and initializes Java API for given plugin.
+     *
+     * @param jvpp     plugin name
+     * @param callback callback provided by the plugin
+     * @throws NullPointerException     if name or callback is null
+     * @throws IllegalArgumentException if plugin was already registered
+     */
+    void register(final JVpp jvpp, final JVppCallback callback);
+
+    /**
+     * Unregisters callback for the given plugin.
+     *
+     * @param name plugin name
+     * @throws NullPointerException     if name is null
+     * @throws IllegalArgumentException if plugin was not registered
+     */
+    void unregister(final String name);
+
+    /**
+     * Returns callback registered for the plugin.
+     *
+     * @param name plugin name
+     * @return callback provided by the plugin
+     * @throws NullPointerException     if name is null
+     * @throws IllegalArgumentException if plugin was not registered
+     */
+    JVppCallback get(final String name);
+
+    /**
+     * Sends control ping. Reply handler calls callback registered for give plugin.
+     *
+     * Control ping is used for initial RX thread to Java thread attachment
+     * that takes place in the plugin's JNI lib
+     * and to wrap dump message replies in one list.
+     *
+     * VPP plugins don't have to provide special control ping, therefore
+     * it is necessary to providing control ping support in JVppRegistry.
+
+     * @param clazz identifies plugin that should receive ping callback
+     * @return unique identifier of message in message queue
+     */
+    int controlPing(final Class clazz) throws VppInvocationException;
+}
diff --git a/src/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/JVppRegistryImpl.java b/src/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/JVppRegistryImpl.java
new file mode 100644
index 00000000..98ef1c15
--- /dev/null
+++ b/src/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/JVppRegistryImpl.java
@@ -0,0 +1,147 @@
+/*
+ * Copyright (c) 2016 Cisco and/or its affiliates.
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at:
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+package io.fd.vpp.jvpp;
+
+import static java.util.Objects.requireNonNull;
+
+import io.fd.vpp.jvpp.callback.ControlPingCallback;
+import io.fd.vpp.jvpp.callback.JVppCallback;
+import io.fd.vpp.jvpp.dto.ControlPingReply;
+import java.io.IOException;
+import java.util.HashMap;
+import java.util.Map;
+import java.util.concurrent.ConcurrentHashMap;
+import java.util.logging.Level;
+import java.util.logging.Logger;
+
+/**
+ * Default implementation of JVppRegistry.
+ */
+public final class JVppRegistryImpl implements JVppRegistry, ControlPingCallback {
+
+    private static final Logger LOG = Logger.getLogger(JVppRegistryImpl.class.getName());
+
+    private final VppJNIConnection connection;
+    // Unguarded concurrent map, no race conditions expected on top of that
+    private final Map pluginRegistry;
+    // Guarded by self
+    private final Map pingCalls;
+
+    public JVppRegistryImpl(final String clientName) throws IOException {
+        connection = new VppJNIConnection(clientName);
+        connection.connect();
+        pluginRegistry = new ConcurrentHashMap<>();
+        pingCalls = new HashMap<>();
+    }
+
+    @Override
+    public VppConnection getConnection() {
+        return connection;
+    }
+
+    @Override
+    public void register(final JVpp jvpp, final JVppCallback callback) {
+        requireNonNull(jvpp, "jvpp should not be null");
+        requireNonNull(callback, "Callback should not be null");
+        final String name = jvpp.getClass().getName();
+        if (pluginRegistry.containsKey(name)) {
+            throw new IllegalArgumentException(
+                String.format("Callback for plugin %s was already registered", name));
+        }
+        jvpp.init(this, callback, connection.getConnectionInfo().queueAddress,
+            connection.getConnectionInfo().clientIndex);
+        pluginRegistry.put(name, callback);
+    }
+
+    @Override
+    public void unregister(final String name) {
+        requireNonNull(name, "Plugin name should not be null");
+        final JVppCallback previous = pluginRegistry.remove(name);
+        assertPluginWasRegistered(name, previous);
+    }
+
+    @Override
+    public JVppCallback get(final String name) {
+        requireNonNull(name, "Plugin name should not be null");
+        JVppCallback value = pluginRegistry.get(name);
+        assertPluginWasRegistered(name, value);
+        return value;
+    }
+
+    private native int controlPing0() throws VppInvocationException;
+
+    @Override
+    public int controlPing(final Class clazz) throws VppInvocationException {
+        connection.checkActive();
+        final String name = clazz.getName();
+
+        final ControlPingCallback callback = (ControlPingCallback) pluginRegistry.get(clazz.getName());
+        assertPluginWasRegistered(name, callback);
+
+        synchronized (pingCalls) {
+            int context = controlPing0();
+            if (context < 0) {
+                throw new VppInvocationException("controlPing", context);
+            }
+
+            pingCalls.put(context, callback);
+            return context;
+        }
+    }
+
+    @Override
+    public void onControlPingReply(final ControlPingReply reply) {
+        final ControlPingCallback callback;
+        synchronized (pingCalls) {
+            callback = pingCalls.remove(reply.context);
+            if (callback == null) {
+                LOG.log(Level.WARNING, "No callback was registered for reply context=" + reply.context + " Contexts waiting="
+                    + pingCalls.keySet());
+                return;
+            }
+        }
+        // pass the reply to the callback registered by the ping caller
+        callback.onControlPingReply(reply);
+    }
+
+    @Override
+    public void onError(final VppCallbackException ex) {
+        final int ctxId = ex.getCtxId();
+        final ControlPingCallback callback;
+
+        synchronized (pingCalls) {
+            callback = pingCalls.get(ctxId);
+        }
+        if (callback == null) {
+            LOG.log(Level.WARNING, "No callback was registered for reply id={0} ", ctxId);
+            return;
+        }
+        // pass the error to the callback registered by the ping caller
+        callback.onError(ex);
+    }
+
+    private static void assertPluginWasRegistered(final String name, final JVppCallback value) {
+        if (value == null) {
+            throw new IllegalArgumentException(String.format("Callback for plugin %s is not registered", name));
+        }
+    }
+
+    @Override
+    public void close() throws Exception {
+        connection.close();
+    }
+}
diff --git a/src/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/NativeLibraryLoader.java b/src/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/NativeLibraryLoader.java
new file mode 100644
index 00000000..ce6d1bfc
--- /dev/null
+++ b/src/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/NativeLibraryLoader.java
@@ -0,0 +1,73 @@
+/*
+ * Copyright (c) 2016 Cisco and/or its affiliates.
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at:
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+package io.fd.vpp.jvpp;
+
+import java.io.IOException;
+import java.io.InputStream;
+import java.nio.file.Files;
+import java.nio.file.Path;
+import java.nio.file.StandardCopyOption;
+import java.nio.file.attribute.PosixFilePermission;
+import java.nio.file.attribute.PosixFilePermissions;
+import java.util.Set;
+import java.util.logging.Level;
+import java.util.logging.Logger;
+
+/**
+ * Utility class for loading JNI libraries.
+ */
+public final class NativeLibraryLoader {
+
+    private static final Logger LOG = Logger.getLogger(NativeLibraryLoader.class.getName());
+
+    private NativeLibraryLoader() {
+        throw new UnsupportedOperationException("This utility class cannot be instantiated.");
+    }
+
+    /**
+     * Loads JNI library using class loader of the given class.
+     *
+     * @param libName name of the library to be loaded
+     */
+    public static void loadLibrary(final String libName, final Class clazz) throws IOException {
+        java.util.Objects.requireNonNull(libName, "libName should not be null");
+        java.util.Objects.requireNonNull(clazz, "clazz should not be null");
+        try (final InputStream is = clazz.getResourceAsStream('/' + libName)) {
+            if (is == null) {
+                throw new IOException("Failed to open library resource " + libName);
+            }
+            loadStream(libName, is);
+        }
+    }
+
+    private static void loadStream(final String libName, final InputStream is) throws IOException {
+        final Set perms = PosixFilePermissions.fromString("rwxr-x---");
+        final Path p = Files.createTempFile(libName, null, PosixFilePermissions.asFileAttribute(perms));
+        try {
+            Files.copy(is, p, StandardCopyOption.REPLACE_EXISTING);
+            Runtime.getRuntime().load(p.toString());
+        } catch (Exception e) {
+            throw new IOException("Failed to load library " + p, e);
+        } finally {
+            try {
+                Files.deleteIfExists(p);
+            } catch (IOException e) {
+                LOG.log(Level.WARNING, String.format("Failed to delete temporary file %s.", p), e);
+            }
+        }
+    }
+}
diff --git a/src/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/VppBaseCallException.java b/src/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/VppBaseCallException.java
new file mode 100644
index 00000000..d71e3055
--- /dev/null
+++ b/src/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/VppBaseCallException.java
@@ -0,0 +1,60 @@
+/*
+ * Copyright (c) 2016 Cisco and/or its affiliates.
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at:
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+package io.fd.vpp.jvpp;
+
+/**
+ * Base exception representing failed operation of JVpp request call
+ */
+public abstract class VppBaseCallException extends Exception {
+    private final String methodName;
+    private final int errorCode;
+
+    /**
+     * Constructs an VppCallbackException with the specified api method name and error code.
+     *
+     * @param methodName name of a method, which invocation or execution failed
+     * @param errorCode  negative error code value associated with this failure
+     * @throws NullPointerException     if apiMethodName is null
+     */
+    public VppBaseCallException(final String methodName, final int errorCode) {
+        super(String.format("vppApi.%s failed with error code: %d", methodName, errorCode));
+        this.methodName = java.util.Objects.requireNonNull(methodName, "apiMethodName is null!");
+        this.errorCode = errorCode;
+        if(errorCode >= 0) {
+            throw new IllegalArgumentException("Error code must be < 0. Was " + errorCode +
+                    " for " + methodName );
+        }
+    }
+
+    /**
+     * Returns  name of a method, which invocation failed.
+     *
+     * @return method name
+     */
+    public String getMethodName() {
+        return methodName;
+    }
+
+    /**
+     * Returns the error code associated with this failure.
+     *
+     * @return a negative integer error code
+     */
+    public int getErrorCode() {
+        return errorCode;
+    }
+}
diff --git a/src/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/VppCallbackException.java b/src/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/VppCallbackException.java
new file mode 100644
index 00000000..ccfcbd3c
--- /dev/null
+++ b/src/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/VppCallbackException.java
@@ -0,0 +1,47 @@
+/*
+ * Copyright (c) 2016 Cisco and/or its affiliates.
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at:
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+package io.fd.vpp.jvpp;
+
+/**
+ * Callback Exception representing failed operation of JVpp request call
+ */
+public class VppCallbackException extends VppBaseCallException {
+    private final int ctxId;
+
+    /**
+     * Constructs an VppCallbackException with the specified api method name and error code.
+     *
+     * @param methodName  name of a method, which invocation failed.
+     * @param ctxId      api request context identifier
+     * @param errorCode  negative error code value associated with this failure
+     * @throws NullPointerException     if apiMethodName is null
+     */
+    public VppCallbackException(final String methodName, final int ctxId, final int errorCode ){
+        super(methodName, errorCode);
+        this.ctxId = ctxId;
+    }
+
+    /**
+     * Returns api request context identifier.
+     *
+     * @return value of context identifier
+     */
+    public int getCtxId() {
+        return ctxId;
+    }
+
+}
diff --git a/src/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/VppConnection.java b/src/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/VppConnection.java
new file mode 100644
index 00000000..e6fd3bdb
--- /dev/null
+++ b/src/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/VppConnection.java
@@ -0,0 +1,45 @@
+/*
+ * Copyright (c) 2016 Cisco and/or its affiliates.
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at:
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+package io.fd.vpp.jvpp;
+
+import java.io.IOException;
+
+/**
+ * Representation of a management connection to VPP.
+ */
+public interface VppConnection extends AutoCloseable {
+
+    /**
+     * Opens VppConnection for communication with VPP.
+     *
+     * @throws IOException if connection is not established
+     */
+    void connect() throws IOException;
+
+    /**
+     * Checks if this instance connection is active.
+     *
+     * @throws IllegalStateException if this instance was disconnected.
+     */
+    void checkActive();
+
+    /**
+     * Closes Vpp connection.
+     */
+    @Override
+    void close();
+}
diff --git a/src/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/VppInvocationException.java b/src/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/VppInvocationException.java
new file mode 100644
index 00000000..a7ccb197
--- /dev/null
+++ b/src/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/VppInvocationException.java
@@ -0,0 +1,33 @@
+/*
+ * Copyright (c) 2016 Cisco and/or its affiliates.
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at:
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+package io.fd.vpp.jvpp;
+
+/**
+ * Exception thrown when Vpp jAPI method invocation failed.
+ */
+public class VppInvocationException extends VppBaseCallException {
+    /**
+     * Constructs an VppApiInvocationFailedException with the specified api method name and error code.
+     *
+     * @param methodName  name of a method, which invocation failed.
+     * @param errorCode  negative error code value associated with this failure
+     * @throws NullPointerException     if apiMethodName is null
+     */
+    public VppInvocationException(final String methodName, final int errorCode) {
+        super(methodName, errorCode);
+    }
+}
diff --git a/src/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/VppJNIConnection.java b/src/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/VppJNIConnection.java
new file mode 100644
index 00000000..320c1283
--- /dev/null
+++ b/src/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/VppJNIConnection.java
@@ -0,0 +1,137 @@
+/*
+ * Copyright (c) 2016 Cisco and/or its affiliates.
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at:
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+package io.fd.vpp.jvpp;
+
+import static io.fd.vpp.jvpp.NativeLibraryLoader.loadLibrary;
+
+import java.io.IOException;
+import java.util.HashMap;
+import java.util.Map;
+import java.util.Objects;
+import java.util.logging.Level;
+import java.util.logging.Logger;
+
+/**
+ * JNI based representation of a management connection to VPP.
+ */
+public final class VppJNIConnection implements VppConnection {
+    private static final Logger LOG = Logger.getLogger(VppJNIConnection.class.getName());
+
+    static {
+        final String libName = "libjvpp_registry.so";
+        try {
+            loadLibrary(libName, VppJNIConnection.class);
+        } catch (IOException e) {
+            LOG.log(Level.SEVERE, String.format("Can't find vpp jni library: %s", libName), e);
+            throw new ExceptionInInitializerError(e);
+        }
+    }
+
+    private ConnectionInfo connectionInfo;
+
+    private final String clientName;
+    private volatile boolean disconnected = false;
+
+    /**
+     * Create VPPJNIConnection instance for client connecting to VPP.
+     *
+     * @param clientName client name instance to be used for communication. Single connection per clientName is
+     *                   allowed.
+     */
+    public VppJNIConnection(final String clientName) {
+        this.clientName = Objects.requireNonNull(clientName, "Null clientName");
+    }
+
+    /**
+     * Guarded by VppJNIConnection.class
+     */
+    private static final Map connections = new HashMap<>();
+
+    /**
+     * Initiate VPP connection for current instance
+     *
+     * Multiple instances are allowed since this class is not a singleton (VPP allows multiple management connections).
+     *
+     * However only a single connection per clientName is allowed.
+     *
+     * @throws IOException in case the connection could not be established
+     */
+
+    @Override
+    public void connect() throws IOException {
+        _connect();
+    }
+
+    private void _connect() throws IOException {
+        synchronized (VppJNIConnection.class) {
+            if (connections.containsKey(clientName)) {
+                throw new IOException("Client " + clientName + " already connected");
+            }
+
+            connectionInfo = clientConnect(clientName);
+            if (connectionInfo.status != 0) {
+                throw new IOException("Connection returned error " + connectionInfo.status);
+            }
+            connections.put(clientName, this);
+        }
+    }
+
+    @Override
+    public final void checkActive() {
+        if (disconnected) {
+            throw new IllegalStateException("Disconnected client " + clientName);
+        }
+    }
+
+    @Override
+    public final synchronized void close() {
+        if (!disconnected) {
+            disconnected = true;
+            try {
+                clientDisconnect();
+            } finally {
+                synchronized (VppJNIConnection.class) {
+                    connections.remove(clientName);
+                }
+            }
+        }
+    }
+
+    public ConnectionInfo getConnectionInfo() {
+        return connectionInfo;
+    }
+
+    /**
+     * VPP connection information used by plugins to reuse the connection.
+     */
+    public static final class ConnectionInfo {
+        public final long queueAddress;
+        public final int clientIndex;
+        public final int status; // FIXME throw exception instead
+
+        public ConnectionInfo(long queueAddress, int clientIndex, int status) {
+            this.queueAddress = queueAddress;
+            this.clientIndex = clientIndex;
+            this.status = status;
+        }
+    }
+
+    private static native ConnectionInfo clientConnect(String clientName);
+
+    private static native void clientDisconnect();
+
+}
diff --git a/src/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/callback/ControlPingCallback.java b/src/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/callback/ControlPingCallback.java
new file mode 100644
index 00000000..efddfdbb
--- /dev/null
+++ b/src/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/callback/ControlPingCallback.java
@@ -0,0 +1,29 @@
+/*
+ * Copyright (c) 2016 Cisco and/or its affiliates.
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at:
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+package io.fd.vpp.jvpp.callback;
+
+import io.fd.vpp.jvpp.dto.ControlPingReply;
+
+/**
+ * Represents callback for control_ping message.
+ */
+public interface ControlPingCallback extends JVppCallback {
+
+    void onControlPingReply(ControlPingReply reply);
+
+}
+
diff --git a/src/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/callback/JVppCallback.java b/src/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/callback/JVppCallback.java
new file mode 100644
index 00000000..ae02063b
--- /dev/null
+++ b/src/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/callback/JVppCallback.java
@@ -0,0 +1,29 @@
+/*
+ * Copyright (c) 2016 Cisco and/or its affiliates.
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at:
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+package io.fd.vpp.jvpp.callback;
+import io.fd.vpp.jvpp.VppCallbackException;
+
+/**
+ * Base JVppCallback interface
+ */
+public interface JVppCallback {
+    /**
+     * onError callback handler used to report failing operation
+     * @param ex VppCallbackException object containing details about failing operation
+     */
+    void onError(VppCallbackException ex);
+}
diff --git a/src/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/callback/JVppNotificationCallback.java b/src/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/callback/JVppNotificationCallback.java
new file mode 100644
index 00000000..8ab0cb21
--- /dev/null
+++ b/src/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/callback/JVppNotificationCallback.java
@@ -0,0 +1,24 @@
+/*
+ * Copyright (c) 2016 Cisco and/or its affiliates.
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at:
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+package io.fd.vpp.jvpp.callback;
+
+/**
+* Notification callback
+*/
+public interface JVppNotificationCallback {
+
+}
diff --git a/src/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/dto/ControlPing.java b/src/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/dto/ControlPing.java
new file mode 100644
index 00000000..984e1674
--- /dev/null
+++ b/src/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/dto/ControlPing.java
@@ -0,0 +1,34 @@
+/*
+ * Copyright (c) 2016 Cisco and/or its affiliates.
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at:
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+package io.fd.vpp.jvpp.dto;
+
+import io.fd.vpp.jvpp.JVpp;
+import io.fd.vpp.jvpp.VppInvocationException;
+
+/**
+ * Represents request DTO for control_ping message.
+ */
+public final class ControlPing implements JVppRequest {
+
+    @Override
+    public int send(final JVpp jvpp) throws VppInvocationException {
+        return jvpp.controlPing(this);
+    }
+
+}
+
+
diff --git a/src/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/dto/ControlPingReply.java b/src/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/dto/ControlPingReply.java
new file mode 100644
index 00000000..61e4d0e4
--- /dev/null
+++ b/src/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/dto/ControlPingReply.java
@@ -0,0 +1,58 @@
+/*
+ * Copyright (c) 2016 Cisco and/or its affiliates.
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at:
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+package io.fd.vpp.jvpp.dto;
+
+import java.util.Objects;
+
+/**
+ * Represents reply DTO for control_ping message.
+ */
+public final class ControlPingReply implements JVppReply {
+
+    public int context;
+    public int clientIndex;
+    public int vpePid;
+
+    @Override
+    public boolean equals(final Object o) {
+        if (this == o) {
+            return true;
+        }
+        if (o == null || getClass() != o.getClass()) {
+            return false;
+        }
+        final ControlPingReply that = (ControlPingReply) o;
+        return context == that.context &&
+                clientIndex == that.clientIndex &&
+                vpePid == that.vpePid;
+    }
+
+    @Override
+    public int hashCode() {
+        return Objects.hash(context, clientIndex, vpePid);
+    }
+
+    @Override
+    public String toString() {
+        return "ControlPingReply{" +
+                "context=" + context +
+                ", clientIndex=" + clientIndex +
+                ", vpePid=" + vpePid +
+                '}';
+    }
+}
+
diff --git a/src/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/dto/JVppDump.java b/src/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/dto/JVppDump.java
new file mode 100644
index 00000000..60b98984
--- /dev/null
+++ b/src/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/dto/JVppDump.java
@@ -0,0 +1,24 @@
+/*
+ * Copyright (c) 2016 Cisco and/or its affiliates.
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at:
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+package io.fd.vpp.jvpp.dto;
+
+/**
+* Base interface for all dump requests
+*/
+public interface JVppDump extends JVppRequest {
+
+}
diff --git a/src/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/dto/JVppNotification.java b/src/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/dto/JVppNotification.java
new file mode 100644
index 00000000..5554f501
--- /dev/null
+++ b/src/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/dto/JVppNotification.java
@@ -0,0 +1,23 @@
+/*
+ * Copyright (c) 2016 Cisco and/or its affiliates.
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at:
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+package io.fd.vpp.jvpp.dto;
+
+/**
+* Base interface for all notification DTOs
+*/
+public interface JVppNotification {
+}
diff --git a/src/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/dto/JVppReply.java b/src/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/dto/JVppReply.java
new file mode 100644
index 00000000..73f512d4
--- /dev/null
+++ b/src/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/dto/JVppReply.java
@@ -0,0 +1,24 @@
+/*
+ * Copyright (c) 2016 Cisco and/or its affiliates.
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at:
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+package io.fd.vpp.jvpp.dto;
+
+/**
+* Base interface for all reply DTOs
+*/
+public interface JVppReply {
+
+}
diff --git a/src/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/dto/JVppReplyDump.java b/src/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/dto/JVppReplyDump.java
new file mode 100644
index 00000000..15111395
--- /dev/null
+++ b/src/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/dto/JVppReplyDump.java
@@ -0,0 +1,25 @@
+/*
+ * Copyright (c) 2016 Cisco and/or its affiliates.
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at:
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+package io.fd.vpp.jvpp.dto;
+
+/**
+* Base interface for all dump replies
+*/
+public interface JVppReplyDump>
+    extends JVppReply {
+
+}
diff --git a/src/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/dto/JVppRequest.java b/src/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/dto/JVppRequest.java
new file mode 100644
index 00000000..9b301da2
--- /dev/null
+++ b/src/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/dto/JVppRequest.java
@@ -0,0 +1,34 @@
+/*
+ * Copyright (c) 2016 Cisco and/or its affiliates.
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at:
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+package io.fd.vpp.jvpp.dto;
+
+import io.fd.vpp.jvpp.JVpp;
+import io.fd.vpp.jvpp.VppInvocationException;
+
+/**
+* Base interface for all request DTOs
+*/
+public interface JVppRequest {
+
+    /**
+     * Invoke current operation asynchronously on VPP
+     *
+     * @return context id of this request. Can be used to track incoming response
+     */
+    int send(JVpp jvpp) throws VppInvocationException;
+
+}
diff --git a/src/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/future/AbstractFutureJVppInvoker.java b/src/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/future/AbstractFutureJVppInvoker.java
new file mode 100644
index 00000000..e7df528a
--- /dev/null
+++ b/src/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/future/AbstractFutureJVppInvoker.java
@@ -0,0 +1,141 @@
+/*
+ * Copyright (c) 2016 Cisco and/or its affiliates.
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at:
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+package io.fd.vpp.jvpp.future;
+
+
+import java.util.Map;
+import java.util.Objects;
+import java.util.concurrent.CompletableFuture;
+import java.util.concurrent.CompletionStage;
+import io.fd.vpp.jvpp.JVpp;
+import io.fd.vpp.jvpp.JVppRegistry;
+import io.fd.vpp.jvpp.VppInvocationException;
+import io.fd.vpp.jvpp.dto.JVppDump;
+import io.fd.vpp.jvpp.dto.JVppReply;
+import io.fd.vpp.jvpp.dto.JVppReplyDump;
+import io.fd.vpp.jvpp.dto.JVppRequest;
+
+/**
+ * Future facade on top of JVpp
+ */
+public abstract class AbstractFutureJVppInvoker implements FutureJVppInvoker {
+
+    private final JVpp jvpp;
+    private final JVppRegistry registry;
+
+    /**
+     * Guarded by self
+     */
+    private final Map>> requests;
+
+    protected AbstractFutureJVppInvoker(final JVpp jvpp, final JVppRegistry registry,
+                                        final Map>> requestMap) {
+        this.jvpp =  Objects.requireNonNull(jvpp, "jvpp should not be null");
+        this.registry =  Objects.requireNonNull(registry, "registry should not be null");
+        // Request map represents the shared state between this facade and it's callback
+        // where facade puts futures in and callback completes + removes them
+        this.requests = Objects.requireNonNull(requestMap, "Null requestMap");
+    }
+
+    protected final Map>> getRequests() {
+        synchronized (requests) {
+            return requests;
+        }
+    }
+
+    // TODO use Optional in Future, java8
+
+    @Override
+    @SuppressWarnings("unchecked")
+    public > CompletionStage send(REQ req) {
+        synchronized(requests) {
+            try {
+                final CompletableFuture replyCompletableFuture;
+                final int contextId = jvpp.send(req);
+
+                if(req instanceof JVppDump) {
+                    throw new IllegalArgumentException("Send with empty reply dump has to be used in case of dump calls");
+                }
+                replyCompletableFuture = new CompletableFuture<>();
+                requests.put(contextId, replyCompletableFuture);
+
+                // TODO in case of timeouts/missing replies, requests from the map are not removed
+                // consider adding cancel method, that would remove requests from the map and cancel
+                // associated replyCompletableFuture
+
+                return replyCompletableFuture;
+            } catch (VppInvocationException ex) {
+                final CompletableFuture replyCompletableFuture = new CompletableFuture<>();
+                replyCompletableFuture.completeExceptionally(ex);
+                return replyCompletableFuture;
+            }
+        }
+    }
+
+    @Override
+    @SuppressWarnings("unchecked")
+    public , DUMP extends JVppReplyDump> CompletionStage send(
+            REQ req, DUMP emptyReplyDump) {
+        synchronized(requests) {
+            try {
+                final CompletableDumpFuture replyCompletableFuture;
+                final int contextId = jvpp.send(req);
+
+                if(!(req instanceof JVppDump)) {
+                    throw new IllegalArgumentException("Send without empty reply dump has to be used in case of regular calls");
+                }
+                replyCompletableFuture = new CompletableDumpFuture<>(contextId, emptyReplyDump);
+
+                requests.put(contextId, replyCompletableFuture);
+                requests.put(registry.controlPing(jvpp.getClass()), replyCompletableFuture);
+
+                // TODO in case of timeouts/missing replies, requests from the map are not removed
+                // consider adding cancel method, that would remove requests from the map and cancel
+                // associated replyCompletableFuture
+
+                return replyCompletableFuture;
+            } catch (VppInvocationException ex) {
+                final CompletableFuture replyCompletableFuture = new CompletableFuture<>();
+                replyCompletableFuture.completeExceptionally(ex);
+                return replyCompletableFuture;
+            }
+        }
+    }
+
+    public static final class CompletableDumpFuture> extends CompletableFuture {
+        private final T replyDump;
+        private final int contextId;
+
+        public CompletableDumpFuture(final int contextId, final T emptyDump) {
+            this.contextId = contextId;
+            this.replyDump = emptyDump;
+        }
+
+        public int getContextId() {
+            return contextId;
+        }
+
+        public T getReplyDump() {
+            return replyDump;
+        }
+    }
+
+    @Override
+    public void close() throws Exception {
+        jvpp.close();
+    }
+}
diff --git a/src/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/future/FutureJVppInvoker.java b/src/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/future/FutureJVppInvoker.java
new file mode 100644
index 00000000..7a48e418
--- /dev/null
+++ b/src/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/future/FutureJVppInvoker.java
@@ -0,0 +1,49 @@
+/*
+ * Copyright (c) 2016 Cisco and/or its affiliates.
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at:
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+package io.fd.vpp.jvpp.future;
+
+
+import io.fd.vpp.jvpp.dto.JVppReply;
+import io.fd.vpp.jvpp.dto.JVppReplyDump;
+import io.fd.vpp.jvpp.dto.JVppRequest;
+
+import java.util.concurrent.CompletionStage;
+import io.fd.vpp.jvpp.notification.NotificationRegistryProvider;
+
+/**
+* Future facade on top of JVpp
+*/
+public interface FutureJVppInvoker extends NotificationRegistryProvider, AutoCloseable {
+
+    /**
+     * Invoke asynchronous operation on VPP
+     *
+     * @return CompletionStage with future result of an async VPP call
+     * @throws io.fd.vpp.jvpp.VppInvocationException when send request failed with details
+     */
+    > CompletionStage send(REQ req);
+
+
+    /**
+     * Invoke asynchronous dump operation on VPP
+     *
+     * @return CompletionStage with aggregated future result of an async VPP dump call
+     * @throws io.fd.vpp.jvpp.VppInvocationException when send request failed with details
+     */
+    , DUMP extends JVppReplyDump> CompletionStage send(
+            REQ req, DUMP emptyReplyDump);
+}
diff --git a/src/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/notification/NotificationRegistry.java b/src/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/notification/NotificationRegistry.java
new file mode 100644
index 00000000..3c72ff79
--- /dev/null
+++ b/src/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/notification/NotificationRegistry.java
@@ -0,0 +1,25 @@
+/*
+ * Copyright (c) 2016 Cisco and/or its affiliates.
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at:
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+package io.fd.vpp.jvpp.notification;
+
+/**
+ * Base registry for notification callbacks.
+ */
+public interface NotificationRegistry extends AutoCloseable {
+
+    void close();
+}
diff --git a/src/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/notification/NotificationRegistryProvider.java b/src/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/notification/NotificationRegistryProvider.java
new file mode 100644
index 00000000..4a6e06b7
--- /dev/null
+++ b/src/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/notification/NotificationRegistryProvider.java
@@ -0,0 +1,28 @@
+/*
+ * Copyright (c) 2016 Cisco and/or its affiliates.
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at:
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+package io.fd.vpp.jvpp.notification;
+
+/**
+ * Provides notification registry
+ */
+public interface NotificationRegistryProvider {
+
+    /**
+     * Get current notification registry instance
+     */
+    NotificationRegistry getNotificationRegistry();
+}
diff --git a/src/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/test/ConnectionTest.java b/src/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/test/ConnectionTest.java
new file mode 100644
index 00000000..27b4d29f
--- /dev/null
+++ b/src/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/test/ConnectionTest.java
@@ -0,0 +1,44 @@
+/*
+ * Copyright (c) 2016 Cisco and/or its affiliates.
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at:
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+package io.fd.vpp.jvpp.test;
+
+import io.fd.vpp.jvpp.JVppRegistry;
+import io.fd.vpp.jvpp.JVppRegistryImpl;
+
+/**
+ * Run using:
+ * sudo java -cp build-vpp-native/vpp-api/java/jvpp-registry-16.09.jar io.fd.vpp.jvpp.test.ConnectionTest
+ */
+public class ConnectionTest {
+
+    private static void testConnect() throws Exception {
+        System.out.println("Testing JNI connection with JVppRegistry");
+        final JVppRegistry registry = new JVppRegistryImpl("ConnectionTest");
+        try {
+            System.out.println("Successfully connected to vpp");
+            Thread.sleep(5000);
+            System.out.println("Disconnecting...");
+            Thread.sleep(1000);
+        } finally {
+            registry.close();
+        }
+    }
+
+    public static void main(String[] args) throws Exception {
+        testConnect();
+    }
+}
diff --git a/src/vpp-api/java/jvpp-registry/jvpp_registry.c b/src/vpp-api/java/jvpp-registry/jvpp_registry.c
new file mode 100644
index 00000000..cbd5e0ab
--- /dev/null
+++ b/src/vpp-api/java/jvpp-registry/jvpp_registry.c
@@ -0,0 +1,352 @@
+/*
+ * Copyright (c) 2016 Cisco and/or its affiliates.
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at:
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+#define _GNU_SOURCE /* for strcasestr(3) */
+#include 
+
+#define vl_api_version(n,v) static u32 vpe_api_version = (v);
+#include 
+#undef vl_api_version
+
+
+#include 
+#include 
+#include "io_fd_vpp_jvpp_VppJNIConnection.h"
+#include "io_fd_vpp_jvpp_JVppRegistryImpl.h"
+
+#include 
+#define vl_typedefs             /* define message structures */
+#include 
+#undef vl_typedefs
+
+#define vl_endianfun
+#include 
+#undef vl_endianfun
+
+/* instantiate all the print functions we know about */
+#define vl_print(handle, ...)
+#define vl_printfun
+#include 
+#undef vl_printfun
+
+vlib_main_t vlib_global_main;
+vlib_main_t **vlib_mains;
+
+/*
+ * The Java runtime isn't compile w/ -fstack-protector,
+ * so we have to supply missing external references for the
+ * regular vpp libraries.
+ */
+void __stack_chk_guard(void) __attribute__((weak));
+void __stack_chk_guard(void) {
+}
+
+typedef struct {
+    /* UThread attachment */
+    volatile u32 control_ping_result_ready;
+    volatile i32 control_ping_retval;
+
+    /* Control poing callback */
+    jobject registryObject;
+    jclass registryClass;
+    jclass controlPingReplyClass;
+    jclass callbackExceptionClass;
+
+    /* Thread cleanup */
+    pthread_key_t cleanup_rx_thread_key;
+
+    /* Connected indication */
+    volatile u8 is_connected;
+} jvpp_registry_main_t;
+
+jvpp_registry_main_t jvpp_registry_main __attribute__((aligned (64)));
+
+void vl_client_add_api_signatures(vl_api_memclnt_create_t *mp) {
+    /*
+     * Send the main API signature in slot 0. This bit of code must
+     * match the checks in ../vpe/api/api.c: vl_msg_api_version_check().
+     */
+    mp->api_versions[0] = clib_host_to_net_u32(vpe_api_version);
+}
+
+/* cleanup handler for RX thread */
+static_always_inline void cleanup_rx_thread(void *arg) {
+    jvpp_main_t * jm = &jvpp_main;
+    jvpp_registry_main_t * rm = &jvpp_registry_main;
+
+    vppjni_lock(jm, 99);
+
+    int getEnvStat = (*jm->jvm)->GetEnv(jm->jvm, (void **) &(jm->jenv),
+    JNI_VERSION_1_8);
+    if (getEnvStat == JNI_EVERSION) {
+        clib_warning("Unsupported JNI version\n");
+        rm->control_ping_retval = VNET_API_ERROR_UNSUPPORTED_JNI_VERSION;
+        goto out;
+    } else if (getEnvStat != JNI_EDETACHED) {
+        (*jm->jvm)->DetachCurrentThread(jm->jvm);
+    }
+    out: vppjni_unlock(jm);
+}
+
+static void vl_api_control_ping_reply_t_handler(
+        vl_api_control_ping_reply_t * mp) {
+    jvpp_main_t * jm = &jvpp_main;
+    jvpp_registry_main_t * rm = &jvpp_registry_main;
+    char was_thread_connected = 0;
+
+    // attach to java thread if not attached
+    int getEnvStat = (*jm->jvm)->GetEnv(jm->jvm, (void **) &(jm->jenv),
+            JNI_VERSION_1_8);
+    if (getEnvStat == JNI_EDETACHED) {
+        if ((*jm->jvm)->AttachCurrentThread(jm->jvm, (void **) &(jm->jenv),
+                NULL) != 0) {
+            clib_warning("Failed to attach thread\n");
+            rm->control_ping_retval =
+                    VNET_API_ERROR_FAILED_TO_ATTACH_TO_JAVA_THREAD;
+            goto out;
+        }
+
+        // workaround as we can't use pthread_cleanup_push
+        pthread_key_create(&rm->cleanup_rx_thread_key, cleanup_rx_thread);
+        // destructor is only called if the value of key is non null
+        pthread_setspecific(rm->cleanup_rx_thread_key, (void *) 1);
+        was_thread_connected = 1;
+    } else if (getEnvStat == JNI_EVERSION) {
+        clib_warning("Unsupported JNI version\n");
+        rm->control_ping_retval = VNET_API_ERROR_UNSUPPORTED_JNI_VERSION;
+        goto out;
+    }
+
+    if (was_thread_connected == 0) {
+        JNIEnv *env = jm->jenv;
+        if (mp->retval < 0) {
+            call_on_error("controlPing", mp->context, mp->retval,
+                    rm->registryClass, rm->registryObject,
+                    rm->callbackExceptionClass);
+        } else {
+            jmethodID constructor = (*env)->GetMethodID(env,
+                    rm->controlPingReplyClass, "", "()V");
+            jmethodID callbackMethod = (*env)->GetMethodID(env,
+                    rm->registryClass, "onControlPingReply",
+                    "(Lio/fd/vpp/jvpp/dto/ControlPingReply;)V");
+
+            jobject dto = (*env)->NewObject(env, rm->controlPingReplyClass,
+                    constructor);
+
+            jfieldID contextFieldId = (*env)->GetFieldID(env,
+                    rm->controlPingReplyClass, "context", "I");
+            (*env)->SetIntField(env, dto, contextFieldId,
+                    clib_net_to_host_u32(mp->context));
+
+            jfieldID clientIndexFieldId = (*env)->GetFieldID(env,
+                    rm->controlPingReplyClass, "clientIndex", "I");
+            (*env)->SetIntField(env, dto, clientIndexFieldId,
+                    clib_net_to_host_u32(mp->client_index));
+
+            jfieldID vpePidFieldId = (*env)->GetFieldID(env,
+                    rm->controlPingReplyClass, "vpePid", "I");
+            (*env)->SetIntField(env, dto, vpePidFieldId,
+                    clib_net_to_host_u32(mp->vpe_pid));
+
+            (*env)->CallVoidMethod(env, rm->registryObject, callbackMethod,
+                    dto);
+            (*env)->DeleteLocalRef(env, dto);
+        }
+    }
+
+    out: rm->control_ping_result_ready = 1;
+}
+
+static int send_initial_control_ping() {
+    f64 timeout;
+    clib_time_t clib_time;
+    vl_api_control_ping_t * mp;
+    jvpp_main_t * jm = &jvpp_main;
+    jvpp_registry_main_t * rm = &jvpp_registry_main;
+
+    clib_time_init(&clib_time);
+
+    rm->control_ping_result_ready = 0;
+    mp = vl_msg_api_alloc(sizeof(*mp));
+    memset(mp, 0, sizeof(*mp));
+    mp->_vl_msg_id = ntohs(VL_API_CONTROL_PING);
+    mp->client_index = jm->my_client_index;
+
+    // send message:
+    vl_msg_api_send_shmem(jm->vl_input_queue, (u8 *) &mp);
+
+    // wait for results: Current time + 10 seconds is the timeout
+    timeout = clib_time_now(&clib_time) + 10.0;
+    int rv = VNET_API_ERROR_RESPONSE_NOT_READY;
+    while (clib_time_now(&clib_time) < timeout) {
+        if (rm->control_ping_result_ready == 1) {
+            rv = rm->control_ping_retval;
+            break;
+        }
+    }
+
+    if (rv != 0) {
+        clib_warning("common: first control ping failed: %d", rv);
+    }
+
+    return rv;
+}
+
+static int connect_to_vpe(char *name) {
+    jvpp_main_t * jm = &jvpp_main;
+    api_main_t * am = &api_main;
+
+    if (vl_client_connect_to_vlib("/vpe-api", name, 32) < 0)
+        return -1;
+
+    jm->my_client_index = am->my_client_index;
+
+    jm->vl_input_queue = am->shmem_hdr->vl_input_queue;
+
+    vl_msg_api_set_handlers(VL_API_CONTROL_PING_REPLY, "control_ping_reply",
+            vl_api_control_ping_reply_t_handler, vl_noop_handler,
+            vl_api_control_ping_reply_t_endian,
+            vl_api_control_ping_reply_t_print,
+            sizeof(vl_api_control_ping_reply_t), 1);
+
+    return send_initial_control_ping();
+}
+
+JNIEXPORT jobject JNICALL Java_io_fd_vpp_jvpp_VppJNIConnection_clientConnect(
+        JNIEnv *env, jclass obj, jstring clientName) {
+    int rv;
+    const char *client_name;
+    void vl_msg_reply_handler_hookup(void);
+    jvpp_main_t * jm = &jvpp_main;
+    jvpp_registry_main_t * rm = &jvpp_registry_main;
+
+    jclass connectionInfoClass = (*env)->FindClass(env,
+            "io/fd/vpp/jvpp/VppJNIConnection$ConnectionInfo");
+    jmethodID connectionInfoConstructor = (*env)->GetMethodID(env,
+            connectionInfoClass, "", "(JII)V");
+
+    /*
+     * Bail out now if we're not running as root
+     */
+    if (geteuid() != 0) {
+        return (*env)->NewObject(env, connectionInfoClass,
+                connectionInfoConstructor, 0, 0,
+                VNET_API_ERROR_NOT_RUNNING_AS_ROOT);
+    }
+
+    if (rm->is_connected) {
+        return (*env)->NewObject(env, connectionInfoClass,
+                connectionInfoConstructor, 0, 0,
+                VNET_API_ERROR_ALREADY_CONNECTED);
+    }
+
+    client_name = (*env)->GetStringUTFChars(env, clientName, 0);
+    if (!client_name) {
+        return (*env)->NewObject(env, connectionInfoClass,
+                connectionInfoConstructor, 0, 0, VNET_API_ERROR_INVALID_VALUE);
+    }
+
+    rv = connect_to_vpe((char *) client_name);
+
+    if (rv < 0)
+        clib_warning("connection failed, rv %d", rv);
+
+    (*env)->ReleaseStringUTFChars(env, clientName, client_name);
+
+    return (*env)->NewObject(env, connectionInfoClass,
+            connectionInfoConstructor, (jlong) jm->vl_input_queue,
+            (jint) jm->my_client_index, (jint) rv);
+}
+
+JNIEXPORT jint JNICALL Java_io_fd_vpp_jvpp_JVppRegistryImpl_controlPing0(
+        JNIEnv *env, jobject regstryObject) {
+    jvpp_main_t * jm = &jvpp_main;
+    vl_api_control_ping_t * mp;
+    u32 my_context_id = vppjni_get_context_id(&jvpp_main);
+    jvpp_registry_main_t * rm = &jvpp_registry_main;
+
+    if (rm->registryObject == 0) {
+        rm->registryObject = (*env)->NewGlobalRef(env, regstryObject);
+    }
+    if (rm->registryClass == 0) {
+        rm->registryClass = (jclass) (*env)->NewGlobalRef(env,
+                (*env)->GetObjectClass(env, regstryObject));
+    }
+
+    mp = vl_msg_api_alloc(sizeof(*mp));
+    memset(mp, 0, sizeof(*mp));
+    mp->_vl_msg_id = ntohs(VL_API_CONTROL_PING);
+    mp->client_index = jm->my_client_index;
+    mp->context = clib_host_to_net_u32(my_context_id);
+
+    // send message:
+    vl_msg_api_send_shmem(jm->vl_input_queue, (u8 *) &mp);
+    return my_context_id;
+}
+
+JNIEXPORT void JNICALL Java_io_fd_vpp_jvpp_VppJNIConnection_clientDisconnect(
+        JNIEnv *env, jclass clazz) {
+    jvpp_registry_main_t * rm = &jvpp_registry_main;
+    rm->is_connected = 0; // TODO make thread safe
+    vl_client_disconnect_from_vlib();
+
+    // cleanup:
+    if (rm->registryObject) {
+        (*env)->DeleteGlobalRef(env, rm->registryObject);
+        rm->registryObject = 0;
+    }
+    if (rm->registryClass) {
+        (*env)->DeleteGlobalRef(env, rm->registryClass);
+        rm->registryClass = 0;
+    }
+}
+
+jint JNI_OnLoad(JavaVM *vm, void *reserved) {
+    jvpp_main_t * jm = &jvpp_main;
+    jvpp_registry_main_t * rm = &jvpp_registry_main;
+    JNIEnv* env;
+
+    if ((*vm)->GetEnv(vm, (void**) &env, JNI_VERSION_1_8) != JNI_OK) {
+        return JNI_EVERSION;
+    }
+
+    rm->controlPingReplyClass = (jclass) (*env)->NewGlobalRef(env,
+            (*env)->FindClass(env, "io/fd/vpp/jvpp/dto/ControlPingReply"));
+    if ((*env)->ExceptionCheck(env)) {
+        (*env)->ExceptionDescribe(env);
+        clib_warning("Failed to cache class references\n");
+        return JNI_ERR;
+    }
+
+    rm->callbackExceptionClass = (jclass) (*env)->NewGlobalRef(env,
+            (*env)->FindClass(env, "io/fd/vpp/jvpp/VppCallbackException"));
+    if ((*env)->ExceptionCheck(env)) {
+        (*env)->ExceptionDescribe(env);
+        return JNI_ERR;
+    }
+
+    jm->jvm = vm;
+    return JNI_VERSION_1_8;
+}
+
+void JNI_OnUnload(JavaVM *vm, void *reserved) {
+    jvpp_main_t * jm = &jvpp_main;
+    JNIEnv* env;
+    if ((*vm)->GetEnv(vm, (void**) &env, JNI_VERSION_1_8) != JNI_OK) {
+        return;
+    }
+
+    jm->jenv = NULL;
+    jm->jvm = NULL;
+}
diff --git a/src/vpp-api/java/jvpp-snat/io/fd/vpp/jvpp/snat/test/CallbackApiTest.java b/src/vpp-api/java/jvpp-snat/io/fd/vpp/jvpp/snat/test/CallbackApiTest.java
new file mode 100644
index 00000000..32165d96
--- /dev/null
+++ b/src/vpp-api/java/jvpp-snat/io/fd/vpp/jvpp/snat/test/CallbackApiTest.java
@@ -0,0 +1,68 @@
+/*
+ * Copyright (c) 2016 Cisco and/or its affiliates.
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at:
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+package io.fd.vpp.jvpp.snat.test;
+
+import io.fd.vpp.jvpp.JVpp;
+import io.fd.vpp.jvpp.JVppRegistry;
+import io.fd.vpp.jvpp.JVppRegistryImpl;
+import io.fd.vpp.jvpp.VppCallbackException;
+import io.fd.vpp.jvpp.snat.JVppSnatImpl;
+import io.fd.vpp.jvpp.snat.callback.SnatInterfaceAddDelFeatureCallback;
+import io.fd.vpp.jvpp.snat.dto.SnatInterfaceAddDelFeature;
+import io.fd.vpp.jvpp.snat.dto.SnatInterfaceAddDelFeatureReply;
+
+public class CallbackApiTest {
+
+    static class TestCallback implements SnatInterfaceAddDelFeatureCallback {
+
+        @Override
+        public void onSnatInterfaceAddDelFeatureReply(final SnatInterfaceAddDelFeatureReply msg) {
+            System.out.printf("Received SnatInterfaceAddDelFeatureReply: context=%d%n",
+                msg.context);
+        }
+
+        @Override
+        public void onError(VppCallbackException ex) {
+            System.out.printf("Received onError exception: call=%s, context=%d, retval=%d%n", ex.getMethodName(),
+                ex.getCtxId(), ex.getErrorCode());
+        }
+    }
+
+    public static void main(String[] args) throws Exception {
+        testCallbackApi();
+    }
+
+    private static void testCallbackApi() throws Exception {
+        System.out.println("Testing Java callback API for snat plugin");
+        try (final JVppRegistry registry = new JVppRegistryImpl("SnatCallbackApiTest");
+             final JVpp jvpp = new JVppSnatImpl()) {
+            registry.register(jvpp, new TestCallback());
+
+            System.out.println("Sending SnatInterfaceAddDelFeature request...");
+            SnatInterfaceAddDelFeature request = new SnatInterfaceAddDelFeature();
+            request.isAdd = 1;
+            request.isInside = 1;
+            request.swIfIndex = 1;
+            final int result = jvpp.send(request);
+            System.out.printf("SnatInterfaceAddDelFeature send result = %d%n", result);
+
+            Thread.sleep(1000);
+
+            System.out.println("Disconnecting...");
+        }
+    }
+}
diff --git a/src/vpp-api/java/jvpp-snat/io/fd/vpp/jvpp/snat/test/Readme.txt b/src/vpp-api/java/jvpp-snat/io/fd/vpp/jvpp/snat/test/Readme.txt
new file mode 100644
index 00000000..a2b0c41f
--- /dev/null
+++ b/src/vpp-api/java/jvpp-snat/io/fd/vpp/jvpp/snat/test/Readme.txt
@@ -0,0 +1 @@
+sudo java -cp build-vpp-native/vpp-api/java/jvpp-registry-17.01.jar:build-vpp-native/plugins/snat-plugin/jvpp-snat-1.0.jar io.fd.vpp.jvpp.snat.test.CallbackApiTest
diff --git a/src/vpp-api/java/jvpp-snat/jvpp_snat.c b/src/vpp-api/java/jvpp-snat/jvpp_snat.c
new file mode 100644
index 00000000..1095b6eb
--- /dev/null
+++ b/src/vpp-api/java/jvpp-snat/jvpp_snat.c
@@ -0,0 +1,124 @@
+/*
+ * Copyright (c) 2016 Cisco and/or its affiliates.
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at:
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#include 
+
+#include 
+#define vl_typedefs             /* define message structures */
+#include 
+#undef vl_typedefs
+
+#define vl_endianfun
+#include 
+#undef vl_endianfun
+
+#define vl_print(handle, ...)
+#define vl_printfun
+#include 
+#undef vl_printfun
+
+/* Get the API version number */
+#define vl_api_version(n,v) static u32 api_version=(v);
+#include 
+#undef vl_api_version
+
+#include 
+#include 
+#include 
+
+#if VPPJNI_DEBUG == 1
+  #define DEBUG_LOG(...) clib_warning(__VA_ARGS__)
+#else
+  #define DEBUG_LOG(...)
+#endif
+
+#include 
+
+#include "jvpp-snat/io_fd_vpp_jvpp_snat_JVppSnatImpl.h"
+#include "jvpp_snat.h"
+#include "jvpp-snat/jvpp_snat_gen.h"
+
+/*
+ * Class:     io_fd_vpp_jvpp_snat_JVppsnatImpl
+ * Method:    init0
+ * Signature: (JI)V
+ */
+JNIEXPORT void JNICALL Java_io_fd_vpp_jvpp_snat_JVppSnatImpl_init0
+  (JNIEnv *env, jclass clazz, jobject callback, jlong queue_address, jint my_client_index) {
+  snat_main_t * plugin_main = &snat_main;
+  u8 * name;
+  clib_warning ("Java_io_fd_vpp_jvpp_snat_JVppSnatImpl_init0");
+
+  plugin_main->my_client_index = my_client_index;
+  plugin_main->vl_input_queue = (unix_shared_memory_queue_t *)queue_address;
+
+  name = format (0, "snat_%08x%c", api_version, 0);
+  plugin_main->msg_id_base = vl_client_get_first_plugin_msg_id ((char *) name);
+
+  if (plugin_main->msg_id_base == (u16) ~0) {
+    jclass exClass = (*env)->FindClass(env, "java/lang/IllegalStateException");
+    (*env)->ThrowNew(env, exClass, "snat plugin is not loaded in VPP");
+  } else {
+    plugin_main->callbackObject = (*env)->NewGlobalRef(env, callback);
+    plugin_main->callbackClass = (jclass)(*env)->NewGlobalRef(env, (*env)->GetObjectClass(env, callback));
+
+    #define _(N,n)                                  \
+        vl_msg_api_set_handlers(VL_API_##N + plugin_main->msg_id_base, #n,     \
+                vl_api_##n##_t_handler,             \
+                vl_noop_handler,                    \
+                vl_api_##n##_t_endian,              \
+                vl_api_##n##_t_print,               \
+                sizeof(vl_api_##n##_t), 1);
+        foreach_api_reply_handler;
+    #undef _
+  }
+}
+
+JNIEXPORT void JNICALL Java_io_fd_vpp_jvpp_snat_JVppSnatImpl_close0
+(JNIEnv *env, jclass clazz) {
+  snat_main_t * plugin_main = &snat_main;
+
+    // cleanup:
+    (*env)->DeleteGlobalRef(env, plugin_main->callbackClass);
+    (*env)->DeleteGlobalRef(env, plugin_main->callbackObject);
+
+    plugin_main->callbackClass = NULL;
+    plugin_main->callbackObject = NULL;
+}
+
+/* Attach thread to JVM and cache class references when initiating JVPP SNAT */
+jint JNI_OnLoad(JavaVM *vm, void *reserved) {
+    JNIEnv* env;
+
+    if ((*vm)->GetEnv(vm, (void**) &env, JNI_VERSION_1_8) != JNI_OK) {
+        return JNI_EVERSION;
+    }
+
+    if (cache_class_references(env) != 0) {
+        clib_warning ("Failed to cache class references\n");
+        return JNI_ERR;
+    }
+
+    return JNI_VERSION_1_8;
+}
+
+/* Clean up cached references when disposing JVPP SNAT */
+void JNI_OnUnload(JavaVM *vm, void *reserved) {
+    JNIEnv* env;
+    if ((*vm)->GetEnv(vm, (void**) &env, JNI_VERSION_1_8) != JNI_OK) {
+        return;
+    }
+    delete_class_references(env);
+}
diff --git a/src/vpp-api/java/jvpp-snat/jvpp_snat.h b/src/vpp-api/java/jvpp-snat/jvpp_snat.h
new file mode 100644
index 00000000..6426bda8
--- /dev/null
+++ b/src/vpp-api/java/jvpp-snat/jvpp_snat.h
@@ -0,0 +1,45 @@
+/*
+ * Copyright (c) 2016 Cisco and/or its affiliates.
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at:
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+#ifndef __included_jvpp_snat_h__
+#define __included_jvpp_snat_h__
+
+#include 
+#include 
+#include 
+#include 
+#include 
+#include 
+
+/* Global state for JVPP-SNAT */
+typedef struct {
+    /* Base message index for the nsh plugin */
+    u16 msg_id_base;
+
+    /* Pointer to shared memory queue */
+    unix_shared_memory_queue_t * vl_input_queue;
+
+    /* VPP api client index */
+    u32 my_client_index;
+
+    /* Callback object and class references enabling asynchronous Java calls */
+    jobject callbackObject;
+    jclass callbackClass;
+
+} snat_main_t;
+
+snat_main_t snat_main __attribute__((aligned (64)));
+
+
+#endif /* __included_jvpp_snat_h__ */
diff --git a/src/vpp-api/java/jvpp/gen/jvpp_gen.py b/src/vpp-api/java/jvpp/gen/jvpp_gen.py
new file mode 100755
index 00000000..f51b11d0
--- /dev/null
+++ b/src/vpp-api/java/jvpp/gen/jvpp_gen.py
@@ -0,0 +1,185 @@
+#!/usr/bin/env python
+#
+# Copyright (c) 2016 Cisco and/or its affiliates.
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at:
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+# l
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+#
+
+import argparse
+import importlib
+import sys
+import os
+import json
+
+from jvppgen import types_gen
+from jvppgen import callback_gen
+from jvppgen import notification_gen
+from jvppgen import dto_gen
+from jvppgen import jvpp_callback_facade_gen
+from jvppgen import jvpp_future_facade_gen
+from jvppgen import jvpp_impl_gen
+from jvppgen import jvpp_c_gen
+from jvppgen import util
+
+blacklist = [ "memclnt.api", "flowperpkt.api" ]
+
+# Invocation:
+# ~/Projects/vpp/vpp-api/jvpp/gen$ mkdir -p java/io/fd/vpp/jvpp && cd java/io/fd/vpp/jvpp
+# ~/Projects/vpp/vpp-api/jvpp/gen/java/io/fd/vpp/jvpp$ ../../../../jvpp_gen.py -idefs_api_vpp_papi.py
+#
+# Compilation:
+# ~/Projects/vpp/vpp-api/jvpp/gen/java/io/fd/vpp/jvpp$ javac *.java dto/*.java callback/*.java
+#
+# where
+# defs_api_vpp_papi.py - vpe.api in python format (generated by vppapigen)
+
+parser = argparse.ArgumentParser(description='VPP Java API generator')
+parser.add_argument('-i', action="store", dest="inputfiles", nargs='+')
+parser.add_argument('--plugin_name', action="store", dest="plugin_name")
+parser.add_argument('--root_dir', action="store", dest="root_dir")
+args = parser.parse_args()
+
+sys.path.append(".")
+cwd = os.getcwd()
+
+print "Generating Java API for %s" % args.inputfiles
+print "inputfiles %s" % args.inputfiles
+plugin_name = args.plugin_name
+print "plugin_name %s" % plugin_name
+
+cfg = {}
+
+base_package = 'io.fd.vpp.jvpp'
+plugin_package = base_package + '.' + plugin_name
+root_dir = os.path.abspath(args.root_dir)
+print "root_dir %s" % root_dir
+work_dir = root_dir + "/target/" + plugin_package.replace(".","/")
+
+try:
+    os.makedirs(work_dir)
+except OSError:
+    if not os.path.isdir(work_dir):
+        raise
+
+os.chdir(work_dir)
+
+for inputfile in args.inputfiles:
+    if any(substring in inputfile for substring in blacklist):
+        print "WARNING: Imput file %s blacklisted" % inputfile
+        continue
+    _cfg = json.load(open(cwd + "/" + inputfile, 'r'))
+    if 'types' in cfg:
+        cfg['types'].extend(_cfg['types'])
+    else:
+        cfg['types'] = _cfg['types']
+    if 'messages' in cfg:
+        cfg['messages'].extend(_cfg['messages'])
+    else:
+        cfg['messages'] = _cfg['messages']
+
+
+def is_request_field(field_name):
+    return field_name not in {'_vl_msg_id', 'client_index', 'context'}
+
+
+def is_response_field(field_name):
+    return field_name not in {'_vl_msg_id'}
+
+
+def get_args(t, filter):
+    arg_list = []
+    for i in t:
+        if is_crc(i):
+            continue
+        if not filter(i[1]):
+            continue
+        arg_list.append(i[1])
+    return arg_list
+
+
+def get_types(t, filter):
+    types_list = []
+    lengths_list = []
+    crc = None
+    for i in t:
+        if is_crc(i):
+            crc = ('crc', i['crc'][2:])
+            continue
+        if not filter(i[1]):
+            continue
+        if len(i) is 3:  # array type
+            types_list.append(i[0] + '[]')
+            lengths_list.append((i[2], False))
+        elif len(i) is 4:  # variable length array type
+            types_list.append(i[0] + '[]')
+            lengths_list.append((i[3], True))
+        else:  # primitive type
+            types_list.append(i[0])
+            lengths_list.append((0, False))
+    return types_list, lengths_list, crc
+
+
+def is_crc(arg):
+    """ Check whether the argument inside message definition is just crc """
+    return 'crc' in arg
+
+
+def get_definitions(defs):
+    # Pass 1
+    func_list = []
+    func_name = {}
+    for a in defs:
+        java_name = util.underscore_to_camelcase(a[0])
+
+        # For replies include all the arguments except message_id
+        if util.is_reply(java_name):
+            types, lengths, crc = get_types(a[1:], is_response_field)
+            func_name[a[0]] = dict(
+                [('name', a[0]), ('java_name', java_name),
+                 ('args', get_args(a[1:], is_response_field)), ('full_args', get_args(a[1:], lambda x: True)),
+                 ('types', types), ('lengths', lengths), crc])
+        # For requests skip message_id, client_id and context
+        else:
+            types, lengths, crc = get_types(a[1:], is_request_field)
+            func_name[a[0]] = dict(
+                [('name', a[0]), ('java_name', java_name),
+                 ('args', get_args(a[1:], is_request_field)), ('full_args', get_args(a[1:], lambda x: True)),
+                 ('types', types), ('lengths', lengths), crc])
+
+        # Indexed by name
+        func_list.append(func_name[a[0]])
+    return func_list, func_name
+
+
+types_package = 'types'
+dto_package = 'dto'
+callback_package = 'callback'
+notification_package = 'notification'
+future_package = 'future'
+# TODO find better package name
+callback_facade_package = 'callfacade'
+
+types_list, types_name = get_definitions(cfg['types'])
+
+types_gen.generate_types(types_list, plugin_package, types_package, args.inputfiles)
+
+func_list, func_name = get_definitions(cfg['messages'])
+
+dto_gen.generate_dtos(func_list, base_package, plugin_package, plugin_name.title(), dto_package, args.inputfiles)
+jvpp_impl_gen.generate_jvpp(func_list, base_package, plugin_package, plugin_name, dto_package, args.inputfiles)
+callback_gen.generate_callbacks(func_list, base_package, plugin_package, plugin_name.title(), callback_package, dto_package, args.inputfiles)
+notification_gen.generate_notification_registry(func_list, base_package, plugin_package, plugin_name.title(), notification_package, callback_package, dto_package, args.inputfiles)
+jvpp_c_gen.generate_jvpp(func_list, plugin_name, args.inputfiles, root_dir)
+jvpp_future_facade_gen.generate_jvpp(func_list, base_package, plugin_package, plugin_name.title(), dto_package, callback_package, notification_package, future_package, args.inputfiles)
+jvpp_callback_facade_gen.generate_jvpp(func_list, base_package, plugin_package, plugin_name.title(), dto_package, callback_package, notification_package, callback_facade_package, args.inputfiles)
+
+print "Java API for %s generated successfully" % args.inputfiles
diff --git a/src/vpp-api/java/jvpp/gen/jvppgen/__init__.py b/src/vpp-api/java/jvpp/gen/jvppgen/__init__.py
new file mode 100644
index 00000000..e69de29b
diff --git a/src/vpp-api/java/jvpp/gen/jvppgen/callback_gen.py b/src/vpp-api/java/jvpp/gen/jvppgen/callback_gen.py
new file mode 100644
index 00000000..b3024b9c
--- /dev/null
+++ b/src/vpp-api/java/jvpp/gen/jvppgen/callback_gen.py
@@ -0,0 +1,105 @@
+#!/usr/bin/env python
+#
+# Copyright (c) 2016 Cisco and/or its affiliates.
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at:
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+import os
+import util
+from string import Template
+
+from util import remove_suffix
+
+callback_suffix = "Callback"
+
+callback_template = Template("""
+package $plugin_package.$callback_package;
+
+/**
+ * 
Represents callback for plugin's api file message.
+ * 
It was generated by callback_gen.py based on $inputfile preparsed data:
+ * 
+$docs
+ * 

+ */
+public interface $cls_name extends $base_package.$callback_package.$callback_type {
+
+    $callback_method
+
+}
+""")
+
+global_callback_template = Template("""
+package $plugin_package.$callback_package;
+
+/**
+ * 
Global aggregated callback interface.
+ * 
It was generated by callback_gen.py based on $inputfile
+ * 
(python representation of api file generated by vppapigen).
+ */
+public interface JVpp${plugin_name}GlobalCallback extends $base_package.$callback_package.ControlPingCallback, $callbacks {
+}
+""")
+
+
+def generate_callbacks(func_list, base_package, plugin_package, plugin_name, callback_package, dto_package, inputfile):
+    """ Generates callback interfaces """
+    print "Generating Callback interfaces"
+
+    if not os.path.exists(callback_package):
+        os.mkdir(callback_package)
+
+    callbacks = []
+    for func in func_list:
+
+        camel_case_name_with_suffix = util.underscore_to_camelcase_upper(func['name'])
+
+        if util.is_ignored(func['name']) or util.is_control_ping(camel_case_name_with_suffix):
+            continue
+        if not util.is_reply(camel_case_name_with_suffix) and not util.is_notification(func['name']):
+            continue
+
+        if util.is_reply(camel_case_name_with_suffix):
+            camel_case_name = util.remove_reply_suffix(camel_case_name_with_suffix)
+            callback_type = "JVppCallback"
+        else:
+            camel_case_name_with_suffix = util.add_notification_suffix(camel_case_name_with_suffix)
+            camel_case_name = camel_case_name_with_suffix
+            callback_type = "JVppNotificationCallback"
+
+        callbacks.append("{0}.{1}.{2}".format(plugin_package, callback_package, camel_case_name + callback_suffix))
+        callback_path = os.path.join(callback_package, camel_case_name + callback_suffix + ".java")
+        callback_file = open(callback_path, 'w')
+
+        reply_type = "%s.%s.%s" % (plugin_package, dto_package, camel_case_name_with_suffix)
+        method = "void on{0}({1} reply);".format(camel_case_name_with_suffix, reply_type)
+        callback_file.write(
+            callback_template.substitute(inputfile=inputfile,
+                                         docs=util.api_message_to_javadoc(func),
+                                         cls_name=camel_case_name + callback_suffix,
+                                         callback_method=method,
+                                         base_package=base_package,
+                                         plugin_package=plugin_package,
+                                         callback_package=callback_package,
+                                         callback_type=callback_type))
+        callback_file.flush()
+        callback_file.close()
+
+    callback_file = open(os.path.join(callback_package, "JVpp%sGlobalCallback.java" % plugin_name), 'w')
+    callback_file.write(global_callback_template.substitute(inputfile=inputfile,
+                                                            callbacks=", ".join(callbacks),
+                                                            base_package=base_package,
+                                                            plugin_package=plugin_package,
+                                                            plugin_name=plugin_name,
+                                                            callback_package=callback_package))
+    callback_file.flush()
+    callback_file.close()
diff --git a/src/vpp-api/java/jvpp/gen/jvppgen/dto_gen.py b/src/vpp-api/java/jvpp/gen/jvppgen/dto_gen.py
new file mode 100644
index 00000000..cfddb9ef
--- /dev/null
+++ b/src/vpp-api/java/jvpp/gen/jvppgen/dto_gen.py
@@ -0,0 +1,308 @@
+#!/usr/bin/env python
+#
+# Copyright (c) 2016 Cisco and/or its affiliates.
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at:
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+import os
+from string import Template
+
+import util
+
+dto_template = Template("""
+package $plugin_package.$dto_package;
+
+/**
+ * 
This class represents $description.
+ * 
It was generated by dto_gen.py based on $inputfile preparsed data:
+ * 
+$docs
+ * 

+ */
+public final class $cls_name implements $base_package.$dto_package.$base_type {
+
+$fields
+$methods
+}
+""")
+
+field_template = Template("""    public $type $name;\n""")
+
+send_template = Template("""    @Override
+    public int send(final $base_package.JVpp jvpp) throws io.fd.vpp.jvpp.VppInvocationException {
+        return (($plugin_package.JVpp${plugin_name})jvpp).$method_name($args);
+    }""")
+
+
+def generate_dtos(func_list, base_package, plugin_package, plugin_name, dto_package, inputfile):
+    """ Generates dto objects in a dedicated package """
+    print "Generating DTOs"
+
+    if not os.path.exists(dto_package):
+        os.mkdir(dto_package)
+
+    for func in func_list:
+        camel_case_dto_name = util.underscore_to_camelcase_upper(func['name'])
+        camel_case_method_name = util.underscore_to_camelcase(func['name'])
+        dto_path = os.path.join(dto_package, camel_case_dto_name + ".java")
+
+        if util.is_ignored(func['name']) or util.is_control_ping(camel_case_dto_name):
+            continue
+
+        fields = generate_dto_fields(camel_case_dto_name, func)
+        methods = generate_dto_base_methods(camel_case_dto_name, func)
+        base_type = ""
+
+        # Generate request/reply or dump/dumpReply even if structure can be used as notification
+        if not util.is_just_notification(func["name"]):
+            if util.is_reply(camel_case_dto_name):
+                description = "reply DTO"
+                request_dto_name = get_request_name(camel_case_dto_name, func['name'])
+                if util.is_details(camel_case_dto_name):
+                    # FIXME assumption that dump calls end with "Dump" suffix. Not enforced in vpe.api
+                    base_type += "JVppReply<%s.%s.%s>" % (plugin_package, dto_package, request_dto_name + "Dump")
+                    generate_dump_reply_dto(request_dto_name, base_package, plugin_package, dto_package,
+                                            camel_case_dto_name, camel_case_method_name, func)
+                else:
+                    base_type += "JVppReply<%s.%s.%s>" % (plugin_package, dto_package, request_dto_name)
+            else:
+                args = "" if fields is "" else "this"
+                methods += send_template.substitute(method_name=camel_case_method_name,
+                                                    base_package=base_package,
+                                                    plugin_package=plugin_package,
+                                                    plugin_name=plugin_name,
+                                                    args=args)
+                if util.is_dump(camel_case_dto_name):
+                    base_type += "JVppDump"
+                    description = "dump request DTO"
+                else:
+                    base_type += "JVppRequest"
+                    description = "request DTO"
+
+            write_dto_file(base_package, plugin_package, base_type, camel_case_dto_name, description, dto_package,
+                           dto_path, fields, func, inputfile, methods)
+
+        # for structures that are also used as notifications, generate dedicated notification DTO
+        if util.is_notification(func["name"]):
+            base_type = "JVppNotification"
+            description = "notification DTO"
+            camel_case_dto_name = util.add_notification_suffix(camel_case_dto_name)
+            dto_path = os.path.join(dto_package, camel_case_dto_name + ".java")
+            methods = generate_dto_base_methods(camel_case_dto_name, func)
+            write_dto_file(base_package, plugin_package, base_type, camel_case_dto_name, description, dto_package,
+                           dto_path, fields, func, inputfile, methods)
+
+    flush_dump_reply_dtos(inputfile)
+
+
+def generate_dto_base_methods(camel_case_dto_name, func):
+    methods = generate_dto_hash(func)
+    methods += generate_dto_equals(camel_case_dto_name, func)
+    methods += generate_dto_tostring(camel_case_dto_name, func)
+    return methods
+
+
+def generate_dto_fields(camel_case_dto_name, func):
+    fields = ""
+    for t in zip(func['types'], func['args']):
+        # for retval don't generate dto field in Reply
+        field_name = util.underscore_to_camelcase(t[1])
+        if util.is_reply(camel_case_dto_name) and util.is_retval_field(field_name):
+            continue
+        fields += field_template.substitute(type=util.jni_2_java_type_mapping[t[0]],
+                                            name=field_name)
+    return fields
+
+
+tostring_field_template = Template("""                \"$field_name=\" + $field_name + ", " +\n""")
+tostring_array_field_template = Template("""                \"$field_name=\" + java.util.Arrays.toString($field_name) + ", " +\n""")
+tostring_template = Template("""    @Override
+    public String toString() {
+        return "$cls_name{" +
+$fields_tostring "}";
+    }\n\n""")
+
+
+def generate_dto_tostring(camel_case_dto_name, func):
+    tostring_fields = ""
+    for t in zip(func['types'], func['args']):
+
+        field_name = util.underscore_to_camelcase(t[1])
+        # for retval don't generate dto field in Reply
+        if util.is_retval_field(field_name):
+            continue
+
+        # handle array types
+        if util.is_array(util.jni_2_java_type_mapping[t[0]]):
+            tostring_fields += tostring_array_field_template.substitute(field_name=field_name)
+        else:
+            tostring_fields += tostring_field_template.substitute(field_name=field_name)
+
+    return tostring_template.substitute(cls_name=camel_case_dto_name,
+                                        fields_tostring=tostring_fields[:-8])
+
+equals_other_template = Template("""
+        final $cls_name other = ($cls_name) o;
+\n""")
+equals_field_template = Template("""        if (!java.util.Objects.equals(this.$field_name, other.$field_name)) {
+            return false;
+        }\n""")
+equals_array_field_template = Template("""        if (!java.util.Arrays.equals(this.$field_name, other.$field_name)) {
+            return false;
+        }\n""")
+equals_template = Template("""    @Override
+    public boolean equals(final Object o) {
+        if (this == o) {
+            return true;
+        }
+        if (o == null || getClass() != o.getClass()) {
+            return false;
+        }
+$comparisons
+        return true;
+    }\n\n""")
+
+
+def generate_dto_equals(camel_case_dto_name, func):
+    equals_fields = ""
+    for t in zip(func['types'], func['args']):
+        field_name = util.underscore_to_camelcase(t[1])
+        # for retval don't generate dto field in Reply
+        if util.is_retval_field(field_name):
+            continue
+
+        # handle array types
+        if util.is_array(util.jni_2_java_type_mapping[t[0]]):
+            equals_fields += equals_array_field_template.substitute(field_name=field_name)
+        else:
+            equals_fields += equals_field_template.substitute(field_name=field_name)
+
+    if equals_fields != "":
+        equals_fields = equals_other_template.substitute(cls_name=camel_case_dto_name) + equals_fields
+
+    return equals_template.substitute(comparisons=equals_fields)
+
+
+hash_template = Template("""    @Override
+    public int hashCode() {
+        return java.util.Objects.hash($fields);
+    }\n\n""")
+hash_single_array_type_template = Template("""    @Override
+    public int hashCode() {
+        return java.util.Arrays.hashCode($fields);
+    }\n\n""")
+
+
+def generate_dto_hash(func):
+    hash_fields = ""
+
+    # Special handling for hashCode in case just a single array field is present. Cannot use Objects.equals since the
+    # array is mistaken for a varargs parameter. Instead use Arrays.hashCode in such case.
+    if len(func['args']) == 1:
+        single_type = func['types'][0]
+        single_type_name = func['args'][0]
+        if util.is_array(util.jni_2_java_type_mapping[single_type]):
+            return hash_single_array_type_template.substitute(fields=util.underscore_to_camelcase(single_type_name))
+
+    for t in zip(func['types'], func['args']):
+        field_name = util.underscore_to_camelcase(t[1])
+        # for retval don't generate dto field in Reply
+        if util.is_retval_field(field_name):
+            continue
+
+        hash_fields += field_name + ", "
+
+    return hash_template.substitute(fields=hash_fields[:-2])
+
+
+def write_dto_file(base_package, plugin_package, base_type, camel_case_dto_name, description, dto_package, dto_path,
+                   fields, func, inputfile, methods):
+    dto_file = open(dto_path, 'w')
+    dto_file.write(dto_template.substitute(inputfile=inputfile,
+                                           description=description,
+                                           docs=util.api_message_to_javadoc(func),
+                                           cls_name=camel_case_dto_name,
+                                           fields=fields,
+                                           methods=methods,
+                                           base_package=base_package,
+                                           plugin_package=plugin_package,
+                                           base_type=base_type,
+                                           dto_package=dto_package))
+    dto_file.flush()
+    dto_file.close()
+
+
+dump_dto_suffix = "ReplyDump"
+dump_reply_artificial_dtos = {}
+
+
+# Returns request name or special one from unconventional_naming_rep_req map
+def get_request_name(camel_case_dto_name, func_name):
+    return util.underscore_to_camelcase_upper(
+        util.unconventional_naming_rep_req[func_name]) if func_name in util.unconventional_naming_rep_req \
+        else util.remove_reply_suffix(camel_case_dto_name)
+
+
+def flush_dump_reply_dtos(inputfile):
+    for dump_reply_artificial_dto in dump_reply_artificial_dtos.values():
+        dto_path = os.path.join(dump_reply_artificial_dto['dto_package'],
+                                dump_reply_artificial_dto['cls_name'] + ".java")
+        dto_file = open(dto_path, 'w')
+        dto_file.write(dto_template.substitute(inputfile=inputfile,
+                                               description="dump reply wrapper",
+                                               docs=dump_reply_artificial_dto['docs'],
+                                               cls_name=dump_reply_artificial_dto['cls_name'],
+                                               fields=dump_reply_artificial_dto['fields'],
+                                               methods=dump_reply_artificial_dto['methods'],
+                                               plugin_package=dump_reply_artificial_dto['plugin_package'],
+                                               base_package=dump_reply_artificial_dto['base_package'],
+                                               base_type=dump_reply_artificial_dto['base_type'],
+                                               dto_package=dump_reply_artificial_dto['dto_package']))
+        dto_file.flush()
+        dto_file.close()
+
+
+def generate_dump_reply_dto(request_dto_name, base_package, plugin_package, dto_package, camel_case_dto_name,
+                            camel_case_method_name, func):
+    base_type = "JVppReplyDump<%s.%s.%s, %s.%s.%s>" % (
+        plugin_package, dto_package, util.remove_reply_suffix(camel_case_dto_name) + "Dump",
+        plugin_package, dto_package, camel_case_dto_name)
+    fields = "    public java.util.List<%s> %s = new java.util.ArrayList<>();" % (camel_case_dto_name, camel_case_method_name)
+    cls_name = camel_case_dto_name + dump_dto_suffix
+    # using artificial type for fields, just to bypass the is_array check in base methods generators
+    # the type is not really used
+    artificial_type = 'u8'
+
+    # In case of already existing artificial reply dump DTO, just update it
+    # Used for sub-dump dtos
+    if request_dto_name in dump_reply_artificial_dtos.keys():
+        dump_reply_artificial_dtos[request_dto_name]['fields'] += '\n' + fields
+        dump_reply_artificial_dtos[request_dto_name]['field_names'].append(func['name'])
+        dump_reply_artificial_dtos[request_dto_name]['field_types'].append(artificial_type)
+        methods = '\n' + generate_dto_base_methods(dump_reply_artificial_dtos[request_dto_name]['cls_name'],
+                                            {'args': dump_reply_artificial_dtos[request_dto_name]['field_names'],
+                                             'types': dump_reply_artificial_dtos[request_dto_name]['field_types']})
+        dump_reply_artificial_dtos[request_dto_name]['methods'] = methods
+    else:
+        methods = '\n' + generate_dto_base_methods(cls_name, {'args': [func['name']],
+                                                              'types': [artificial_type]})
+        dump_reply_artificial_dtos[request_dto_name] = ({'docs': util.api_message_to_javadoc(func),
+                                                         'cls_name': cls_name,
+                                                         'fields': fields,
+                                                         'field_names': [func['name']],
+                                                         'field_types': [artificial_type],
+                                                         # strip too many newlines at the end of base method block
+                                                         'methods': methods,
+                                                         'plugin_package': plugin_package,
+                                                         'base_package': base_package,
+                                                         'base_type': base_type,
+                                                         'dto_package': dto_package})
diff --git a/src/vpp-api/java/jvpp/gen/jvppgen/jni_gen.py b/src/vpp-api/java/jvpp/gen/jvppgen/jni_gen.py
new file mode 100644
index 00000000..328cc8d3
--- /dev/null
+++ b/src/vpp-api/java/jvpp/gen/jvppgen/jni_gen.py
@@ -0,0 +1,295 @@
+#!/usr/bin/env python
+#
+# Copyright (c) 2016 Cisco and/or its affiliates.
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at:
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+from string import Template
+
+import util
+
+variable_length_array_value_template = Template("""mp->${length_var_name}""")
+variable_length_array_template = Template("""clib_net_to_host_${length_field_type}(${value})""")
+
+dto_field_id_template = Template("""
+    jfieldID ${field_reference_name}FieldId = (*env)->GetFieldID(env, ${class_ref_name}Class, "${field_name}", "${jni_signature}");""")
+
+default_dto_field_setter_template = Template("""
+    (*env)->Set${jni_setter}(env, ${object_name}, ${field_reference_name}FieldId, mp->${c_name});
+""")
+
+variable_length_array_value_template = Template("""mp->${length_var_name}""")
+variable_length_array_template = Template("""clib_net_to_host_${length_field_type}(${value})""")
+
+u16_dto_field_setter_template = Template("""
+    (*env)->Set${jni_setter}(env, ${object_name}, ${field_reference_name}FieldId, clib_net_to_host_u16(mp->${c_name}));
+""")
+
+u32_dto_field_setter_template = Template("""
+    (*env)->Set${jni_setter}(env, ${object_name}, ${field_reference_name}FieldId, clib_net_to_host_u32(mp->${c_name}));
+""")
+
+u64_dto_field_setter_template = Template("""
+    (*env)->Set${jni_setter}(env, ${object_name}, ${field_reference_name}FieldId, clib_net_to_host_u64(mp->${c_name}));
+""")
+
+u8_array_dto_field_setter_template = Template("""
+    jbyteArray ${field_reference_name} = (*env)->NewByteArray(env, ${field_length});
+    (*env)->SetByteArrayRegion(env, ${field_reference_name}, 0, ${field_length}, (const jbyte*)mp->${c_name});
+    (*env)->SetObjectField(env, ${object_name}, ${field_reference_name}FieldId, ${field_reference_name});
+    (*env)->DeleteLocalRef(env, ${field_reference_name});
+""")
+
+u16_array_dto_field_setter_template = Template("""
+    {
+        jshortArray ${field_reference_name} = (*env)->NewShortArray(env, ${field_length});
+        jshort * ${field_reference_name}ArrayElements = (*env)->GetShortArrayElements(env, ${field_reference_name}, NULL);
+        unsigned int _i;
+        for (_i = 0; _i < ${field_length}; _i++) {
+            ${field_reference_name}ArrayElements[_i] = clib_net_to_host_u16(mp->${c_name}[_i]);
+        }
+
+        (*env)->ReleaseShortArrayElements(env,  ${field_reference_name}, ${field_reference_name}ArrayElements, 0);
+        (*env)->SetObjectField(env, ${object_name}, ${field_reference_name}FieldId, ${field_reference_name});
+        (*env)->DeleteLocalRef(env, ${field_reference_name});
+    }
+""")
+
+u32_array_dto_field_setter_template = Template("""
+    {
+        jintArray ${field_reference_name} = (*env)->NewIntArray(env, ${field_length});
+        jint * ${field_reference_name}ArrayElements = (*env)->GetIntArrayElements(env, ${field_reference_name}, NULL);
+        unsigned int _i;
+        for (_i = 0; _i < ${field_length}; _i++) {
+            ${field_reference_name}ArrayElements[_i] = clib_net_to_host_u32(mp->${c_name}[_i]);
+        }
+
+        (*env)->ReleaseIntArrayElements(env,  ${field_reference_name}, ${field_reference_name}ArrayElements, 0);
+        (*env)->SetObjectField(env, ${object_name}, ${field_reference_name}FieldId, ${field_reference_name});
+        (*env)->DeleteLocalRef(env, ${field_reference_name});
+    }
+""")
+
+# For each u64 array we get its elements. Then we convert values to host byte order.
+# All changes to  jlong* buffer are written to jlongArray (isCopy is set to NULL)
+u64_array_dto_field_setter_template = Template("""
+    {
+        jlongArray ${field_reference_name} = (*env)->NewLongArray(env, ${field_length});
+        jlong * ${field_reference_name}ArrayElements = (*env)->GetLongArrayElements(env, ${field_reference_name}, NULL);
+        unsigned int _i;
+        for (_i = 0; _i < ${field_length}; _i++) {
+            ${field_reference_name}ArrayElements[_i] = clib_net_to_host_u64(mp->${c_name}[_i]);
+        }
+
+        (*env)->ReleaseLongArrayElements(env,  ${field_reference_name}, ${field_reference_name}ArrayElements, 0);
+        (*env)->SetObjectField(env, ${object_name}, ${field_reference_name}FieldId, ${field_reference_name});
+        (*env)->DeleteLocalRef(env, ${field_reference_name});
+    }
+""")
+
+dto_field_setter_templates = {'u8': default_dto_field_setter_template,
+                              'u16': u16_dto_field_setter_template,
+                              'u32': u32_dto_field_setter_template,
+                              'i32': u32_dto_field_setter_template,
+                              'u64': u64_dto_field_setter_template,
+                              'f64': default_dto_field_setter_template,  # fixme
+                              'u8[]': u8_array_dto_field_setter_template,
+                              'u16[]': u16_array_dto_field_setter_template,
+                              'u32[]': u32_array_dto_field_setter_template,
+                              'u64[]': u64_array_dto_field_setter_template
+                              }
+
+
+def jni_reply_handler_for_type(handler_name, ref_name, field_type, c_name, field_reference_name,
+                               field_name, field_length, is_variable_len_array, length_field_type,
+                               object_name="dto"):
+    """
+    Generates jni code that initializes a field of java object (dto or custom type).
+    To be used in reply message handlers.
+    :param field_type: type of the field to be initialized (as defined in vpe.api)
+    :param c_name: name of the message struct member that stores initialization value
+    :param field_reference_name: name of the field reference in generated code
+    :param field_name: name of the field (camelcase)
+    :param field_length: integer or name of variable that stores field length
+    :param object_name: name of the object to be initialized
+    """
+
+    # todo move validation to vppapigen
+    if field_type.endswith('[]') and field_length == '0':
+        raise Exception('Variable array \'%s\' defined in \'%s\' '
+                        'should have defined length (e.g. \'%s[%s_length]\''
+                        % (c_name, handler_name, c_name, c_name))
+
+    if is_variable_len_array:
+        length_var_name = field_length
+        field_length = variable_length_array_value_template.substitute(length_var_name=length_var_name)
+        if length_field_type != 'u8':  # we need net to host conversion:
+            field_length = variable_length_array_template.substitute(
+                    length_field_type=length_field_type, value=field_length)
+
+    # for retval don't generate setters
+    if util.is_retval_field(c_name):
+        return ""
+
+    jni_signature = util.jni_2_signature_mapping[field_type]
+    jni_setter = util.jni_field_accessors[field_type]
+
+    result = dto_field_id_template.substitute(
+            field_reference_name=field_reference_name,
+            field_name=field_name,
+            class_ref_name=ref_name,
+            jni_signature=jni_signature)
+
+    dto_setter_template = dto_field_setter_templates[field_type]
+
+    result += dto_setter_template.substitute(
+            jni_signature=jni_signature,
+            object_name=object_name,
+            field_reference_name=field_reference_name,
+            c_name=c_name,
+            jni_setter=jni_setter,
+            field_length=field_length)
+    return result
+
+
+request_field_identifier_template = Template("""
+    jfieldID ${field_reference_name}FieldId = (*env)->GetFieldID(env, ${object_name}Class, "${field_name}", "${jni_signature}");
+    ${jni_type} ${field_reference_name} = (*env)->Get${jni_getter}(env, ${object_name}, ${field_reference_name}FieldId);
+    """)
+
+array_length_enforcement_template = Template("""
+        size_t max_size = ${field_length};
+        if (cnt > max_size) cnt = max_size;""")
+
+u8_struct_setter_template = Template("""
+    mp->${c_name} = ${field_reference_name};""")
+
+u16_struct_setter_template = Template("""
+    mp->${c_name} = clib_host_to_net_u16(${field_reference_name});""")
+
+u32_struct_setter_template = Template("""
+    mp->${c_name} = clib_host_to_net_u32(${field_reference_name});""")
+
+i32_struct_setter_template = Template("""
+    mp->${c_name} = clib_host_to_net_i32(${field_reference_name});!""")
+
+u64_struct_setter_template = Template("""
+    mp->${c_name} = clib_host_to_net_u64(${field_reference_name});""")
+
+array_length_enforcement_template = Template("""
+        size_t max_size = ${field_length};
+        if (cnt > max_size) cnt = max_size;""")
+
+u8_array_struct_setter_template = Template("""
+    if (${field_reference_name}) {
+        jsize cnt = (*env)->GetArrayLength (env, ${field_reference_name});
+        ${field_length_check}
+        (*env)->GetByteArrayRegion(env, ${field_reference_name}, 0, cnt, (jbyte *)mp->${c_name});
+    }
+""")
+
+u16_array_struct_setter_template = Template("""
+    if (${field_reference_name}) {
+        jshort * ${field_reference_name}ArrayElements = (*env)->GetShortArrayElements(env, ${field_reference_name}, NULL);
+        size_t _i;
+        jsize cnt = (*env)->GetArrayLength (env, ${field_reference_name});
+        ${field_length_check}
+        for (_i = 0; _i < cnt; _i++) {
+            mp->${c_name}[_i] = clib_host_to_net_u16(${field_reference_name}ArrayElements[_i]);
+        }
+        (*env)->ReleaseShortArrayElements (env, ${field_reference_name}, ${field_reference_name}ArrayElements, 0);
+    }
+    """)
+
+u32_array_struct_setter_template = Template("""
+    if (${field_reference_name}) {
+        jint * ${field_reference_name}ArrayElements = (*env)->GetIntArrayElements(env, ${field_reference_name}, NULL);
+        size_t _i;
+        jsize cnt = (*env)->GetArrayLength (env, ${field_reference_name});
+        ${field_length_check}
+        for (_i = 0; _i < cnt; _i++) {
+            mp->${c_name}[_i] = clib_host_to_net_u32(${field_reference_name}ArrayElements[_i]);
+        }
+        (*env)->ReleaseIntArrayElements (env, ${field_reference_name}, ${field_reference_name}ArrayElements, 0);
+    }
+    """)
+
+u64_array_struct_setter_template = Template("""
+    if (${field_reference_name}) {
+        jlong * ${field_reference_name}ArrayElements = (*env)->GetLongArrayElements(env, ${field_reference_name}, NULL);
+        size_t _i;
+        jsize cnt = (*env)->GetArrayLength (env, ${field_reference_name});
+        ${field_length_check}
+        for (_i = 0; _i < cnt; _i++) {
+            mp->${c_name}[_i] = clib_host_to_net_u64(${field_reference_name}ArrayElements[_i]);
+        }
+        (*env)->ReleaseLongArrayElements (env, ${field_reference_name}, ${field_reference_name}ArrayElements, 0);
+    }
+    """)
+
+struct_setter_templates = {'u8': u8_struct_setter_template,
+                           'u16': u16_struct_setter_template,
+                           'u32': u32_struct_setter_template,
+                           'i32': u32_struct_setter_template,
+                           'u64': u64_struct_setter_template,
+                           'u8[]': u8_array_struct_setter_template,
+                           'u16[]': u16_array_struct_setter_template,
+                           'u32[]': u32_array_struct_setter_template,
+                           'u64[]': u64_array_struct_setter_template
+                           }
+
+
+def jni_request_binding_for_type(field_type, c_name, field_reference_name, field_name, field_length,
+                                 is_variable_len_array, object_name="request"):
+    """
+    Generates jni code that initializes C structure that corresponds to a field of java object
+    (dto or custom type). To be used in request message handlers.
+    :param field_type: type of the field to be initialized (as defined in vpe.api)
+    :param c_name: name of the message struct member to be initialized
+    :param field_reference_name: name of the field reference in generated code
+    :param field_name: name of the field (camelcase)
+    :param field_length: integer or name of variable that stores field length
+    :param object_name: name of the object to be initialized
+    """
+    # field identifiers
+    jni_type = util.vpp_2_jni_type_mapping[field_type]
+    jni_signature = util.jni_2_signature_mapping[field_type]
+    jni_getter = util.jni_field_accessors[field_type]
+
+    # field identifier
+    msg_initialization = request_field_identifier_template.substitute(
+            jni_type=jni_type,
+            field_reference_name=field_reference_name,
+            field_name=field_name,
+            jni_signature=jni_signature,
+            jni_getter=jni_getter,
+            object_name=object_name)
+
+    # field setter
+    field_length_check = ""
+
+    # check if we are processing variable length array:
+    if is_variable_len_array:
+        field_length = util.underscore_to_camelcase(field_length)
+
+    # enforce max length if array has fixed length or uses variable length syntax
+    if str(field_length) != "0":
+        field_length_check = array_length_enforcement_template.substitute(field_length=field_length)
+
+    struct_setter_template = struct_setter_templates[field_type]
+
+    msg_initialization += struct_setter_template.substitute(
+            c_name=c_name,
+            field_reference_name=field_reference_name,
+            field_length_check=field_length_check)
+
+    return msg_initialization
diff --git a/src/vpp-api/java/jvpp/gen/jvppgen/jvpp_c_gen.py b/src/vpp-api/java/jvpp/gen/jvppgen/jvpp_c_gen.py
new file mode 100644
index 00000000..611171c4
--- /dev/null
+++ b/src/vpp-api/java/jvpp/gen/jvppgen/jvpp_c_gen.py
@@ -0,0 +1,343 @@
+#!/usr/bin/env python
+#
+# Copyright (c) 2016 Cisco and/or its affiliates.
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at:
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+# l
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+#
+
+import os, util
+from string import Template
+
+import jni_gen
+
+
+def is_manually_generated(f_name, plugin_name):
+    return f_name in {'control_ping_reply'}
+
+
+class_reference_template = Template("""jclass ${ref_name}Class;
+""")
+
+find_class_invocation_template = Template("""
+    ${ref_name}Class = (jclass)(*env)->NewGlobalRef(env, (*env)->FindClass(env, "io/fd/vpp/jvpp/${plugin_name}/dto/${class_name}"));
+    if ((*env)->ExceptionCheck(env)) {
+        (*env)->ExceptionDescribe(env);
+        return JNI_ERR;
+    }""")
+
+find_class_template = Template("""
+    ${ref_name}Class = (jclass)(*env)->NewGlobalRef(env, (*env)->FindClass(env, "${class_name}"));
+    if ((*env)->ExceptionCheck(env)) {
+        (*env)->ExceptionDescribe(env);
+        return JNI_ERR;
+    }""")
+
+delete_class_invocation_template = Template("""
+    if (${ref_name}Class) {
+        (*env)->DeleteGlobalRef(env, ${ref_name}Class);
+    }""")
+
+class_cache_template = Template("""
+$class_references
+static int cache_class_references(JNIEnv* env) {
+    $find_class_invocations
+    return 0;
+}
+
+static void delete_class_references(JNIEnv* env) {
+    $delete_class_invocations
+}""")
+
+
+def generate_class_cache(func_list, plugin_name):
+    class_references = []
+    find_class_invocations = []
+    delete_class_invocations = []
+    for f in func_list:
+        c_name = f['name']
+        class_name = util.underscore_to_camelcase_upper(c_name)
+        ref_name = util.underscore_to_camelcase(c_name)
+
+        if util.is_ignored(c_name) or util.is_control_ping(class_name):
+            continue
+
+        if util.is_reply(class_name):
+            class_references.append(class_reference_template.substitute(
+                ref_name=ref_name))
+            find_class_invocations.append(find_class_invocation_template.substitute(
+                plugin_name=plugin_name,
+                ref_name=ref_name,
+                class_name=class_name))
+            delete_class_invocations.append(delete_class_invocation_template.substitute(ref_name=ref_name))
+        elif util.is_notification(c_name):
+            class_references.append(class_reference_template.substitute(
+                ref_name=util.add_notification_suffix(ref_name)))
+            find_class_invocations.append(find_class_invocation_template.substitute(
+                plugin_name=plugin_name,
+                ref_name=util.add_notification_suffix(ref_name),
+                class_name=util.add_notification_suffix(class_name)))
+            delete_class_invocations.append(delete_class_invocation_template.substitute(
+                ref_name=util.add_notification_suffix(ref_name)))
+
+    # add exception class to class cache
+    ref_name = 'callbackException'
+    class_name = 'io/fd/vpp/jvpp/VppCallbackException'
+    class_references.append(class_reference_template.substitute(
+            ref_name=ref_name))
+    find_class_invocations.append(find_class_template.substitute(
+            ref_name=ref_name,
+            class_name=class_name))
+    delete_class_invocations.append(delete_class_invocation_template.substitute(ref_name=ref_name))
+
+    return class_cache_template.substitute(
+        class_references="".join(class_references), find_class_invocations="".join(find_class_invocations),
+        delete_class_invocations="".join(delete_class_invocations))
+
+
+# TODO: cache method and field identifiers to achieve better performance
+# https://jira.fd.io/browse/HONEYCOMB-42
+request_class_template = Template("""
+    jclass requestClass = (*env)->FindClass(env, "io/fd/vpp/jvpp/${plugin_name}/dto/${java_name_upper}");""")
+
+request_field_identifier_template = Template("""
+    jfieldID ${field_reference_name}FieldId = (*env)->GetFieldID(env, ${object_name}Class, "${field_name}", "${jni_signature}");
+    ${jni_type} ${field_reference_name} = (*env)->Get${jni_getter}(env, ${object_name}, ${field_reference_name}FieldId);
+    """)
+
+
+jni_impl_template = Template("""
+/**
+ * JNI binding for sending ${c_name} message.
+ * Generated based on $inputfile preparsed data:
+$api_data
+ */
+JNIEXPORT jint JNICALL Java_io_fd_vpp_jvpp_${plugin_name}_JVpp${java_plugin_name}Impl_${field_name}0
+(JNIEnv * env, jclass clazz$args) {
+    ${plugin_name}_main_t *plugin_main = &${plugin_name}_main;
+    vl_api_${c_name}_t * mp;
+    u32 my_context_id = vppjni_get_context_id (&jvpp_main);
+    $request_class
+
+    // create message:
+    mp = vl_msg_api_alloc(sizeof(*mp));
+    memset (mp, 0, sizeof (*mp));
+    mp->_vl_msg_id = ntohs (VL_API_${c_name_uppercase} + plugin_main->msg_id_base);
+    mp->client_index = plugin_main->my_client_index;
+    mp->context = clib_host_to_net_u32 (my_context_id);
+
+    $msg_initialization
+
+    // send message:
+    vl_msg_api_send_shmem (plugin_main->vl_input_queue, (u8 *)&mp);
+    if ((*env)->ExceptionCheck(env)) {
+        return JNI_ERR;
+    }
+    return my_context_id;
+}""")
+
+def generate_jni_impl(func_list, plugin_name, inputfile):
+    jni_impl = []
+    for f in func_list:
+        f_name = f['name']
+        camel_case_function_name = util.underscore_to_camelcase(f_name)
+        if is_manually_generated(f_name, plugin_name) or util.is_reply(camel_case_function_name) \
+                or util.is_ignored(f_name) or util.is_just_notification(f_name):
+            continue
+
+        arguments = ''
+        request_class = ''
+        msg_initialization = ''
+        f_name_uppercase = f_name.upper()
+
+        if f['args']:
+            arguments = ', jobject request'
+            camel_case_function_name_upper = util.underscore_to_camelcase_upper(f_name)
+
+            request_class = request_class_template.substitute(
+                    java_name_upper=camel_case_function_name_upper,
+                    plugin_name=plugin_name)
+
+            for t in zip(f['types'], f['args'], f['lengths']):
+                field_name = util.underscore_to_camelcase(t[1])
+                msg_initialization += jni_gen.jni_request_binding_for_type(field_type=t[0], c_name=t[1],
+                                                                           field_reference_name=field_name,
+                                                                           field_name=field_name,
+                                                                           field_length=t[2][0],
+                                                                           is_variable_len_array=t[2][1])
+
+        jni_impl.append(jni_impl_template.substitute(
+                inputfile=inputfile,
+                api_data=util.api_message_to_javadoc(f),
+                field_reference_name=camel_case_function_name,
+                field_name=camel_case_function_name,
+                c_name_uppercase=f_name_uppercase,
+                c_name=f_name,
+                plugin_name=plugin_name,
+                java_plugin_name=plugin_name.title(),
+                request_class=request_class,
+                msg_initialization=msg_initialization,
+                args=arguments))
+
+    return "\n".join(jni_impl)
+
+# code fragment for checking result of the operation before sending request reply
+callback_err_handler_template = Template("""
+    // for negative result don't send callback message but send error callback
+    if (mp->retval<0) {
+        call_on_error("${handler_name}", mp->context, mp->retval, plugin_main->callbackClass, plugin_main->callbackObject, callbackExceptionClass);
+        return;
+    }
+    if (mp->retval == VNET_API_ERROR_IN_PROGRESS) {
+        clib_warning("Result in progress");
+        return;
+    }
+""")
+
+msg_handler_template = Template("""
+/**
+ * Handler for ${handler_name} message.
+ * Generated based on $inputfile preparsed data:
+$api_data
+ */
+static void vl_api_${handler_name}_t_handler (vl_api_${handler_name}_t * mp)
+{
+    ${plugin_name}_main_t *plugin_main = &${plugin_name}_main;
+    JNIEnv *env = jvpp_main.jenv;
+
+    $err_handler
+
+    jmethodID constructor = (*env)->GetMethodID(env, ${class_ref_name}Class, "", "()V");
+    jmethodID callbackMethod = (*env)->GetMethodID(env, plugin_main->callbackClass, "on${dto_name}", "(Lio/fd/vpp/jvpp/${plugin_name}/dto/${dto_name};)V");
+
+    jobject dto = (*env)->NewObject(env, ${class_ref_name}Class, constructor);
+    $dto_setters
+
+    (*env)->CallVoidMethod(env, plugin_main->callbackObject, callbackMethod, dto);
+    // free DTO as per http://stackoverflow.com/questions/1340938/memory-leak-when-calling-java-code-from-c-using-jni
+    (*env)->DeleteLocalRef(env, dto);
+}""")
+
+
+def generate_msg_handlers(func_list, plugin_name, inputfile):
+    handlers = []
+    for f in func_list:
+        handler_name = f['name']
+        dto_name = util.underscore_to_camelcase_upper(handler_name)
+        ref_name = util.underscore_to_camelcase(handler_name)
+
+        if is_manually_generated(handler_name, plugin_name) or util.is_ignored(handler_name):
+            continue
+
+        if not util.is_reply(dto_name) and not util.is_notification(handler_name):
+            continue
+
+        if util.is_notification(handler_name):
+            dto_name = util.add_notification_suffix(dto_name)
+            ref_name = util.add_notification_suffix(ref_name)
+
+        dto_setters = ''
+        err_handler = ''
+        # dto setters
+        for t in zip(f['types'], f['args'], f['lengths']):
+            c_name = t[1]
+            java_name = util.underscore_to_camelcase(c_name)
+            field_length = t[2][0]
+            is_variable_len_array = t[2][1]
+            length_field_type = None
+            if is_variable_len_array:
+                length_field_type = f['types'][f['args'].index(field_length)]
+            dto_setters += jni_gen.jni_reply_handler_for_type(handler_name=handler_name, ref_name=ref_name,
+                                                              field_type=t[0], c_name=t[1],
+                                                              field_reference_name=java_name,
+                                                              field_name=java_name, field_length=field_length,
+                                                              is_variable_len_array=is_variable_len_array,
+                                                              length_field_type=length_field_type)
+
+            # for retval don't generate setters and generate retval check
+            if util.is_retval_field(c_name):
+                err_handler = callback_err_handler_template.substitute(
+                    handler_name=handler_name
+                )
+                continue
+
+        handlers.append(msg_handler_template.substitute(
+            inputfile=inputfile,
+            api_data=util.api_message_to_javadoc(f),
+            handler_name=handler_name,
+            plugin_name=plugin_name,
+            dto_name=dto_name,
+            class_ref_name=ref_name,
+            dto_setters=dto_setters,
+            err_handler=err_handler))
+
+    return "\n".join(handlers)
+
+
+handler_registration_template = Template("""_(${upercase_name}, ${name}) \\
+""")
+
+
+def generate_handler_registration(func_list):
+    handler_registration = ["#define foreach_api_reply_handler \\\n"]
+    for f in func_list:
+        name = f['name']
+        camelcase_name = util.underscore_to_camelcase(f['name'])
+
+        if (not util.is_reply(camelcase_name) and not util.is_notification(name)) or util.is_ignored(name) \
+                or util.is_control_ping(camelcase_name):
+            continue
+
+        handler_registration.append(handler_registration_template.substitute(
+            name=name,
+            upercase_name=name.upper()))
+
+    return "".join(handler_registration)
+
+
+jvpp_c_template = Template("""/**
+ * This file contains JNI bindings for jvpp Java API.
+ * It was generated by jvpp_c_gen.py based on $inputfile
+ * (python representation of api file generated by vppapigen).
+ */
+
+// JAVA class reference cache
+$class_cache
+
+// JNI bindings
+$jni_implementations
+
+// Message handlers
+$msg_handlers
+
+// Registration of message handlers in vlib
+$handler_registration
+""")
+
+def generate_jvpp(func_list, plugin_name, inputfile, path):
+    """ Generates jvpp C file """
+    print "Generating jvpp C"
+
+    class_cache = generate_class_cache(func_list, plugin_name)
+    jni_impl = generate_jni_impl(func_list, plugin_name, inputfile)
+    msg_handlers = generate_msg_handlers(func_list, plugin_name, inputfile)
+    handler_registration = generate_handler_registration(func_list)
+
+    jvpp_c_file = open("%s/jvpp_%s_gen.h" % (path, plugin_name), 'w')
+    jvpp_c_file.write(jvpp_c_template.substitute(
+            inputfile=inputfile,
+            class_cache=class_cache,
+            jni_implementations=jni_impl,
+            msg_handlers=msg_handlers,
+            handler_registration=handler_registration))
+    jvpp_c_file.flush()
+    jvpp_c_file.close()
+
diff --git a/src/vpp-api/java/jvpp/gen/jvppgen/jvpp_callback_facade_gen.py b/src/vpp-api/java/jvpp/gen/jvppgen/jvpp_callback_facade_gen.py
new file mode 100644
index 00000000..ac096a71
--- /dev/null
+++ b/src/vpp-api/java/jvpp/gen/jvppgen/jvpp_callback_facade_gen.py
@@ -0,0 +1,324 @@
+#!/usr/bin/env python
+#
+# Copyright (c) 2016 Cisco and/or its affiliates.
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at:
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+import os, util
+from string import Template
+
+import callback_gen
+import dto_gen
+
+jvpp_ifc_template = Template("""
+package $plugin_package.$callback_facade_package;
+
+/**
+ * 
Callback Java API representation of $plugin_package plugin.
+ * 
It was generated by jvpp_callback_facade_gen.py based on $inputfile
+ * 
(python representation of api file generated by vppapigen).
+ */
+public interface CallbackJVpp${plugin_name} extends $base_package.$notification_package.NotificationRegistryProvider, java.lang.AutoCloseable {
+
+    // TODO add send
+
+$methods
+}
+""")
+
+jvpp_impl_template = Template("""
+package $plugin_package.$callback_facade_package;
+
+/**
+ * 
Default implementation of Callback${plugin_name}JVpp interface.
+ * 
It was generated by jvpp_callback_facade_gen.py based on $inputfile
+ * 
(python representation of api file generated by vppapigen).
+ */
+public final class CallbackJVpp${plugin_name}Facade implements CallbackJVpp${plugin_name} {
+
+    private final $plugin_package.JVpp${plugin_name} jvpp;
+    private final java.util.Map callbacks;
+    private final $plugin_package.$notification_package.${plugin_name}NotificationRegistryImpl notificationRegistry = new $plugin_package.$notification_package.${plugin_name}NotificationRegistryImpl();
+    /**
+     * 
Create CallbackJVpp${plugin_name}Facade object for provided JVpp instance.
+     * Constructor internally creates CallbackJVppFacadeCallback class for processing callbacks
+     * and then connects to provided JVpp instance
+     *
+     * @param jvpp provided $base_package.JVpp instance
+     *
+     * @throws java.io.IOException in case instance cannot connect to JVPP
+     */
+    public CallbackJVpp${plugin_name}Facade(final $base_package.JVppRegistry registry, final $plugin_package.JVpp${plugin_name} jvpp) throws java.io.IOException {
+        this.jvpp = java.util.Objects.requireNonNull(jvpp,"jvpp is null");
+        this.callbacks = new java.util.HashMap<>();
+        java.util.Objects.requireNonNull(registry, "JVppRegistry should not be null");
+        registry.register(jvpp, new CallbackJVpp${plugin_name}FacadeCallback(this.callbacks, notificationRegistry));
+    }
+
+    @Override
+    public $plugin_package.$notification_package.${plugin_name}NotificationRegistry getNotificationRegistry() {
+        return notificationRegistry;
+    }
+
+    @Override
+    public void close() throws Exception {
+        jvpp.close();
+    }
+
+    // TODO add send()
+
+$methods
+}
+""")
+
+method_template = Template(
+    """    void $name($plugin_package.$dto_package.$request request, $plugin_package.$callback_package.$callback callback) throws $base_package.VppInvocationException;""")
+
+method_impl_template = Template("""    public final void $name($plugin_package.$dto_package.$request request, $plugin_package.$callback_package.$callback callback) throws $base_package.VppInvocationException {
+        synchronized (callbacks) {
+            callbacks.put(jvpp.$name(request), callback);
+        }
+    }
+""")
+
+no_arg_method_template = Template("""    void $name($plugin_package.$callback_package.$callback callback) throws $base_package.VppInvocationException;""")
+no_arg_method_impl_template = Template("""    public final void $name($plugin_package.$callback_package.$callback callback) throws $base_package.VppInvocationException {
+        synchronized (callbacks) {
+            callbacks.put(jvpp.$name(), callback);
+        }
+    }
+""")
+
+
+def generate_jvpp(func_list, base_package, plugin_package, plugin_name, dto_package, callback_package, notification_package, callback_facade_package, inputfile):
+    """ Generates callback facade """
+    print "Generating JVpp callback facade"
+
+    if os.path.exists(callback_facade_package):
+        util.remove_folder(callback_facade_package)
+
+    os.mkdir(callback_facade_package)
+
+    methods = []
+    methods_impl = []
+    for func in func_list:
+
+        if util.is_notification(func['name']) or util.is_ignored(func['name']):
+            continue
+
+        camel_case_name = util.underscore_to_camelcase(func['name'])
+        camel_case_name_upper = util.underscore_to_camelcase_upper(func['name'])
+        if util.is_reply(camel_case_name) or util.is_control_ping(camel_case_name):
+            continue
+
+        # Strip suffix for dump calls
+        callback_type = get_request_name(camel_case_name_upper, func['name']) + callback_gen.callback_suffix
+
+        if len(func['args']) == 0:
+            methods.append(no_arg_method_template.substitute(name=camel_case_name,
+                                                             base_package=base_package,
+                                                             plugin_package=plugin_package,
+                                                             dto_package=dto_package,
+                                                             callback_package=callback_package,
+                                                             callback=callback_type))
+            methods_impl.append(no_arg_method_impl_template.substitute(name=camel_case_name,
+                                                                       base_package=base_package,
+                                                                       plugin_package=plugin_package,
+                                                                       dto_package=dto_package,
+                                                                       callback_package=callback_package,
+                                                                       callback=callback_type))
+        else:
+            methods.append(method_template.substitute(name=camel_case_name,
+                                                      request=camel_case_name_upper,
+                                                      base_package=base_package,
+                                                      plugin_package=plugin_package,
+                                                      dto_package=dto_package,
+                                                      callback_package=callback_package,
+                                                      callback=callback_type))
+            methods_impl.append(method_impl_template.substitute(name=camel_case_name,
+                                                                request=camel_case_name_upper,
+                                                                base_package=base_package,
+                                                                plugin_package=plugin_package,
+                                                                dto_package=dto_package,
+                                                                callback_package=callback_package,
+                                                                callback=callback_type))
+
+    join = os.path.join(callback_facade_package, "CallbackJVpp%s.java" % plugin_name)
+    jvpp_file = open(join, 'w')
+    jvpp_file.write(
+        jvpp_ifc_template.substitute(inputfile=inputfile,
+                                     methods="\n".join(methods),
+                                     base_package=base_package,
+                                     plugin_package=plugin_package,
+                                     plugin_name=plugin_name,
+                                     dto_package=dto_package,
+                                     notification_package=notification_package,
+                                     callback_facade_package=callback_facade_package))
+    jvpp_file.flush()
+    jvpp_file.close()
+
+    jvpp_file = open(os.path.join(callback_facade_package, "CallbackJVpp%sFacade.java" % plugin_name), 'w')
+    jvpp_file.write(jvpp_impl_template.substitute(inputfile=inputfile,
+                                                  methods="\n".join(methods_impl),
+                                                  base_package=base_package,
+                                                  plugin_package=plugin_package,
+                                                  plugin_name=plugin_name,
+                                                  dto_package=dto_package,
+                                                  notification_package=notification_package,
+                                                  callback_package=callback_package,
+                                                  callback_facade_package=callback_facade_package))
+    jvpp_file.flush()
+    jvpp_file.close()
+
+    generate_callback(func_list, base_package, plugin_package, plugin_name, dto_package, callback_package, notification_package, callback_facade_package, inputfile)
+
+
+jvpp_facade_callback_template = Template("""
+package $plugin_package.$callback_facade_package;
+
+/**
+ * 
Implementation of JVppGlobalCallback interface for Java Callback API.
+ * 
It was generated by jvpp_callback_facade_gen.py based on $inputfile
+ * 
(python representation of api file generated by vppapigen).
+ */
+public final class CallbackJVpp${plugin_name}FacadeCallback implements $plugin_package.$callback_package.JVpp${plugin_name}GlobalCallback {
+
+    private final java.util.Map requests;
+    private final $plugin_package.$notification_package.Global${plugin_name}NotificationCallback notificationCallback;
+    private static final java.util.logging.Logger LOG = java.util.logging.Logger.getLogger(CallbackJVpp${plugin_name}FacadeCallback.class.getName());
+
+    public CallbackJVpp${plugin_name}FacadeCallback(final java.util.Map requestMap,
+                                      final $plugin_package.$notification_package.Global${plugin_name}NotificationCallback notificationCallback) {
+        this.requests = requestMap;
+        this.notificationCallback = notificationCallback;
+    }
+
+    @Override
+    public void onError($base_package.VppCallbackException reply) {
+
+        $base_package.$callback_package.JVppCallback failedCall;
+        synchronized(requests) {
+            failedCall = requests.remove(reply.getCtxId());
+        }
+
+        if(failedCall != null) {
+            try {
+                failedCall.onError(reply);
+            } catch(RuntimeException ex) {
+                ex.addSuppressed(reply);
+                LOG.log(java.util.logging.Level.WARNING, String.format("Callback: %s failed while handling exception: %s", failedCall, reply), ex);
+            }
+        }
+    }
+
+    @Override
+    @SuppressWarnings("unchecked")
+    public void onControlPingReply($base_package.$dto_package.ControlPingReply reply) {
+
+        $base_package.$callback_package.ControlPingCallback callback;
+        synchronized(requests) {
+            callback = ($base_package.$callback_package.ControlPingCallback) requests.remove(reply.context);
+        }
+
+        if(callback != null) {
+            callback.onControlPingReply(reply);
+        }
+    }
+
+$methods
+}
+""")
+
+jvpp_facade_callback_method_template = Template("""
+    @Override
+    @SuppressWarnings("unchecked")
+    public void on$callback_dto($plugin_package.$dto_package.$callback_dto reply) {
+
+        $plugin_package.$callback_package.$callback callback;
+        synchronized(requests) {
+            callback = ($plugin_package.$callback_package.$callback) requests.remove(reply.context);
+        }
+
+        if(callback != null) {
+            callback.on$callback_dto(reply);
+        }
+    }
+""")
+
+jvpp_facade_callback_notification_method_template = Template("""
+    @Override
+    @SuppressWarnings("unchecked")
+    public void on$callback_dto($plugin_package.$dto_package.$callback_dto notification) {
+        notificationCallback.on$callback_dto(notification);
+    }
+""")
+
+
+def generate_callback(func_list, base_package, plugin_package, plugin_name, dto_package, callback_package, notification_package, callback_facade_package, inputfile):
+    callbacks = []
+    for func in func_list:
+
+        camel_case_name_with_suffix = util.underscore_to_camelcase_upper(func['name'])
+
+        if util.is_ignored(func['name']) or util.is_control_ping(camel_case_name_with_suffix):
+            continue
+
+        if util.is_reply(camel_case_name_with_suffix):
+            callbacks.append(jvpp_facade_callback_method_template.substitute(plugin_package=plugin_package,
+                                                                             dto_package=dto_package,
+                                                                             callback_package=callback_package,
+                                                                             callback=util.remove_reply_suffix(camel_case_name_with_suffix) + callback_gen.callback_suffix,
+                                                                             callback_dto=camel_case_name_with_suffix))
+
+        if util.is_notification(func["name"]):
+            with_notification_suffix = util.add_notification_suffix(camel_case_name_with_suffix)
+            callbacks.append(jvpp_facade_callback_notification_method_template.substitute(plugin_package=plugin_package,
+                                                                             dto_package=dto_package,
+                                                                             callback_package=callback_package,
+                                                                             callback=with_notification_suffix + callback_gen.callback_suffix,
+                                                                             callback_dto=with_notification_suffix))
+
+    jvpp_file = open(os.path.join(callback_facade_package, "CallbackJVpp%sFacadeCallback.java" % plugin_name), 'w')
+    jvpp_file.write(jvpp_facade_callback_template.substitute(inputfile=inputfile,
+                                                             base_package=base_package,
+                                                             plugin_package=plugin_package,
+                                                             plugin_name=plugin_name,
+                                                             dto_package=dto_package,
+                                                             notification_package=notification_package,
+                                                             callback_package=callback_package,
+                                                             methods="".join(callbacks),
+                                                             callback_facade_package=callback_facade_package))
+    jvpp_file.flush()
+    jvpp_file.close()
+
+
+# Returns request name or special one from unconventional_naming_rep_req map
+def get_request_name(camel_case_dto_name, func_name):
+    if func_name in reverse_dict(util.unconventional_naming_rep_req):
+        request_name = util.underscore_to_camelcase_upper(reverse_dict(util.unconventional_naming_rep_req)[func_name])
+    else:
+        request_name = camel_case_dto_name
+    return remove_suffix(request_name)
+
+
+def reverse_dict(map):
+    return dict((v, k) for k, v in map.iteritems())
+
+
+def remove_suffix(name):
+    if util.is_reply(name):
+        return util.remove_reply_suffix(name)
+    else:
+        if util.is_dump(name):
+            return util.remove_suffix(name, util.dump_suffix)
+        else:
+            return name
diff --git a/src/vpp-api/java/jvpp/gen/jvppgen/jvpp_future_facade_gen.py b/src/vpp-api/java/jvpp/gen/jvppgen/jvpp_future_facade_gen.py
new file mode 100644
index 00000000..26b31e22
--- /dev/null
+++ b/src/vpp-api/java/jvpp/gen/jvppgen/jvpp_future_facade_gen.py
@@ -0,0 +1,331 @@
+#!/usr/bin/env python
+#
+# Copyright (c) 2016 Cisco and/or its affiliates.
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at:
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+import os
+from string import Template
+
+import dto_gen
+import util
+
+jvpp_facade_callback_template = Template("""
+package $plugin_package.$future_package;
+
+/**
+ * 
Async facade callback setting values to future objects
+ * 
It was generated by jvpp_future_facade_gen.py based on $inputfile
+ * 
(python representation of api file generated by vppapigen).
+ */
+public final class FutureJVpp${plugin_name}FacadeCallback implements $plugin_package.$callback_package.JVpp${plugin_name}GlobalCallback {
+
+    private final java.util.Map>> requests;
+    private final $plugin_package.$notification_package.Global${plugin_name}NotificationCallback notificationCallback;
+
+    public FutureJVpp${plugin_name}FacadeCallback(
+        final java.util.Map>> requestMap,
+        final $plugin_package.$notification_package.Global${plugin_name}NotificationCallback notificationCallback) {
+        this.requests = requestMap;
+        this.notificationCallback = notificationCallback;
+    }
+
+    @Override
+    @SuppressWarnings("unchecked")
+    public void onError($base_package.VppCallbackException reply) {
+        final java.util.concurrent.CompletableFuture<$base_package.$dto_package.JVppReply> completableFuture;
+
+        synchronized(requests) {
+            completableFuture = (java.util.concurrent.CompletableFuture<$base_package.$dto_package.JVppReply>) requests.get(reply.getCtxId());
+        }
+
+        if(completableFuture != null) {
+            completableFuture.completeExceptionally(reply);
+
+            synchronized(requests) {
+                requests.remove(reply.getCtxId());
+            }
+        }
+    }
+
+    @Override
+    @SuppressWarnings("unchecked")
+    public void onControlPingReply($base_package.$dto_package.ControlPingReply reply) {
+        final java.util.concurrent.CompletableFuture<$base_package.$dto_package.JVppReply> completableFuture;
+
+        synchronized(requests) {
+            completableFuture = (java.util.concurrent.CompletableFuture<$base_package.$dto_package.JVppReply>) requests.get(reply.context);
+        }
+
+        if(completableFuture != null) {
+            // Finish dump call
+            if (completableFuture instanceof $base_package.$future_package.AbstractFutureJVppInvoker.CompletableDumpFuture) {
+                completableFuture.complete((($base_package.$future_package.AbstractFutureJVppInvoker.CompletableDumpFuture) completableFuture).getReplyDump());
+                // Remove future mapped to dump call context id
+                synchronized(requests) {
+                    requests.remove((($base_package.$future_package.AbstractFutureJVppInvoker.CompletableDumpFuture) completableFuture).getContextId());
+                }
+            } else {
+                completableFuture.complete(reply);
+            }
+
+            synchronized(requests) {
+                requests.remove(reply.context);
+            }
+        }
+    }
+
+$methods
+}
+""")
+
+jvpp_facade_callback_method_template = Template("""
+    @Override
+    @SuppressWarnings("unchecked")
+    public void on$callback_dto($plugin_package.$dto_package.$callback_dto reply) {
+        final java.util.concurrent.CompletableFuture<$base_package.$dto_package.JVppReply> completableFuture;
+
+        synchronized(requests) {
+            completableFuture = (java.util.concurrent.CompletableFuture<$base_package.$dto_package.JVppReply>) requests.get(reply.context);
+        }
+
+        if(completableFuture != null) {
+            completableFuture.complete(reply);
+
+            synchronized(requests) {
+                requests.remove(reply.context);
+            }
+        }
+    }
+""")
+
+jvpp_facade_callback_notification_method_template = Template("""
+    @Override
+    public void on$callback_dto($plugin_package.$dto_package.$callback_dto notification) {
+        notificationCallback.on$callback_dto(notification);
+    }
+""")
+
+jvpp_facade_details_callback_method_template = Template("""
+    @Override
+    @SuppressWarnings("unchecked")
+    public void on$callback_dto($plugin_package.$dto_package.$callback_dto reply) {
+        final $base_package.$future_package.AbstractFutureJVppInvoker.CompletableDumpFuture<$plugin_package.$dto_package.$callback_dto_reply_dump> completableFuture;
+
+        synchronized(requests) {
+            completableFuture = ($base_package.$future_package.AbstractFutureJVppInvoker.CompletableDumpFuture<$plugin_package.$dto_package.$callback_dto_reply_dump>) requests.get(reply.context);
+        }
+
+        if(completableFuture != null) {
+            completableFuture.getReplyDump().$callback_dto_field.add(reply);
+        }
+    }
+""")
+
+
+def generate_jvpp(func_list, base_package, plugin_package, plugin_name, dto_package, callback_package, notification_package, future_facade_package, inputfile):
+    """ Generates JVpp interface and JNI implementation """
+    print "Generating JVpp future facade"
+
+    if not os.path.exists(future_facade_package):
+        os.mkdir(future_facade_package)
+
+    methods = []
+    methods_impl = []
+    callbacks = []
+    for func in func_list:
+        camel_case_name_with_suffix = util.underscore_to_camelcase_upper(func['name'])
+
+        if util.is_ignored(func['name']) or util.is_control_ping(camel_case_name_with_suffix):
+            continue
+
+        if not util.is_reply(camel_case_name_with_suffix) and not util.is_notification(func['name']):
+            continue
+
+        camel_case_method_name = util.underscore_to_camelcase(func['name'])
+
+        if not util.is_notification(func["name"]):
+            camel_case_request_method_name = util.remove_reply_suffix(util.underscore_to_camelcase(func['name']))
+            if util.is_details(camel_case_name_with_suffix):
+                camel_case_reply_name = get_standard_dump_reply_name(util.underscore_to_camelcase_upper(func['name']),
+                                                                     func['name'])
+                callbacks.append(jvpp_facade_details_callback_method_template.substitute(base_package=base_package,
+                                                                                         plugin_package=plugin_package,
+                                                                                         dto_package=dto_package,
+                                                                                         callback_dto=camel_case_name_with_suffix,
+                                                                                         callback_dto_field=camel_case_method_name,
+                                                                                         callback_dto_reply_dump=camel_case_reply_name + dto_gen.dump_dto_suffix,
+                                                                                         future_package=future_facade_package))
+
+                methods.append(future_jvpp_method_template.substitute(plugin_package=plugin_package,
+                                                                      dto_package=dto_package,
+                                                                      method_name=camel_case_request_method_name +
+                                                                                  util.underscore_to_camelcase_upper(util.dump_suffix),
+                                                                      reply_name=camel_case_reply_name + dto_gen.dump_dto_suffix,
+                                                                      request_name=util.remove_reply_suffix(camel_case_reply_name) +
+                                                                                   util.underscore_to_camelcase_upper(util.dump_suffix)))
+                methods_impl.append(future_jvpp_dump_method_impl_template.substitute(plugin_package=plugin_package,
+                                                                                     dto_package=dto_package,
+                                                                                     method_name=camel_case_request_method_name +
+                                                                                                 util.underscore_to_camelcase_upper(util.dump_suffix),
+                                                                                     reply_name=camel_case_reply_name + dto_gen.dump_dto_suffix,
+                                                                                     request_name=util.remove_reply_suffix(camel_case_reply_name) +
+                                                                                                  util.underscore_to_camelcase_upper(util.dump_suffix)))
+            else:
+                request_name = util.underscore_to_camelcase_upper(util.unconventional_naming_rep_req[func['name']]) \
+                    if func['name'] in util.unconventional_naming_rep_req else util.remove_reply_suffix(camel_case_name_with_suffix)
+
+                methods.append(future_jvpp_method_template.substitute(plugin_package=plugin_package,
+                                                                      dto_package=dto_package,
+                                                                      method_name=camel_case_request_method_name,
+                                                                      reply_name=camel_case_name_with_suffix,
+                                                                      request_name=request_name))
+                methods_impl.append(future_jvpp_method_impl_template.substitute(plugin_package=plugin_package,
+                                                                                dto_package=dto_package,
+                                                                                method_name=camel_case_request_method_name,
+                                                                                reply_name=camel_case_name_with_suffix,
+                                                                                request_name=request_name))
+
+                callbacks.append(jvpp_facade_callback_method_template.substitute(base_package=base_package,
+                                                                                 plugin_package=plugin_package,
+                                                                                 dto_package=dto_package,
+                                                                                 callback_dto=camel_case_name_with_suffix))
+
+        if util.is_notification(func["name"]):
+            callbacks.append(jvpp_facade_callback_notification_method_template.substitute(plugin_package=plugin_package,
+                                                                                          dto_package=dto_package,
+                                                                                          callback_dto=util.add_notification_suffix(camel_case_name_with_suffix)))
+
+    jvpp_file = open(os.path.join(future_facade_package, "FutureJVpp%sFacadeCallback.java" % plugin_name), 'w')
+    jvpp_file.write(jvpp_facade_callback_template.substitute(inputfile=inputfile,
+                                                             base_package=base_package,
+                                                             plugin_package=plugin_package,
+                                                             plugin_name=plugin_name,
+                                                             dto_package=dto_package,
+                                                             notification_package=notification_package,
+                                                             callback_package=callback_package,
+                                                             methods="".join(callbacks),
+                                                             future_package=future_facade_package))
+    jvpp_file.flush()
+    jvpp_file.close()
+
+    jvpp_file = open(os.path.join(future_facade_package, "FutureJVpp%s.java" % plugin_name), 'w')
+    jvpp_file.write(future_jvpp_template.substitute(inputfile=inputfile,
+                                                    base_package=base_package,
+                                                    plugin_package=plugin_package,
+                                                    plugin_name=plugin_name,
+                                                    notification_package=notification_package,
+                                                    methods="".join(methods),
+                                                    future_package=future_facade_package))
+    jvpp_file.flush()
+    jvpp_file.close()
+
+    jvpp_file = open(os.path.join(future_facade_package, "FutureJVpp%sFacade.java" % plugin_name), 'w')
+    jvpp_file.write(future_jvpp_facade_template.substitute(inputfile=inputfile,
+                                                           base_package=base_package,
+                                                           plugin_package=plugin_package,
+                                                           plugin_name=plugin_name,
+                                                           dto_package=dto_package,
+                                                           notification_package=notification_package,
+                                                           methods="".join(methods_impl),
+                                                           future_package=future_facade_package))
+    jvpp_file.flush()
+    jvpp_file.close()
+
+
+future_jvpp_template = Template('''
+package $plugin_package.$future_package;
+
+/**
+ * 
Async facade extension adding specific methods for each request invocation
+ * 
It was generated by jvpp_future_facade_gen.py based on $inputfile
+ * 
(python representation of api file generated by vppapigen).
+ */
+public interface FutureJVpp${plugin_name} extends $base_package.$future_package.FutureJVppInvoker {
+$methods
+
+    @Override
+    public $plugin_package.$notification_package.${plugin_name}NotificationRegistry getNotificationRegistry();
+
+}
+''')
+
+future_jvpp_method_template = Template('''
+    java.util.concurrent.CompletionStage<$plugin_package.$dto_package.$reply_name> $method_name($plugin_package.$dto_package.$request_name request);
+''')
+
+
+future_jvpp_facade_template = Template('''
+package $plugin_package.$future_package;
+
+/**
+ * 
Implementation of FutureJVpp based on AbstractFutureJVppInvoker
+ * 
It was generated by jvpp_future_facade_gen.py based on $inputfile
+ * 
(python representation of api file generated by vppapigen).
+ */
+public class FutureJVpp${plugin_name}Facade extends $base_package.$future_package.AbstractFutureJVppInvoker implements FutureJVpp${plugin_name} {
+
+    private final $plugin_package.$notification_package.${plugin_name}NotificationRegistryImpl notificationRegistry = new $plugin_package.$notification_package.${plugin_name}NotificationRegistryImpl();
+
+    /**
+     * 
Create FutureJVpp${plugin_name}Facade object for provided JVpp instance.
+     * Constructor internally creates FutureJVppFacadeCallback class for processing callbacks
+     * and then connects to provided JVpp instance
+     *
+     * @param jvpp provided $base_package.JVpp instance
+     *
+     * @throws java.io.IOException in case instance cannot connect to JVPP
+     */
+    public FutureJVpp${plugin_name}Facade(final $base_package.JVppRegistry registry, final $base_package.JVpp jvpp) throws java.io.IOException {
+        super(jvpp, registry, new java.util.HashMap<>());
+        java.util.Objects.requireNonNull(registry, "JVppRegistry should not be null");
+        registry.register(jvpp, new FutureJVpp${plugin_name}FacadeCallback(getRequests(), notificationRegistry));
+    }
+
+    @Override
+    public $plugin_package.$notification_package.${plugin_name}NotificationRegistry getNotificationRegistry() {
+        return notificationRegistry;
+    }
+
+$methods
+}
+''')
+
+future_jvpp_method_impl_template = Template('''
+    @Override
+    public java.util.concurrent.CompletionStage<$plugin_package.$dto_package.$reply_name> $method_name($plugin_package.$dto_package.$request_name request) {
+        return send(request);
+    }
+''')
+
+future_jvpp_dump_method_impl_template = Template('''
+    @Override
+    public java.util.concurrent.CompletionStage<$plugin_package.$dto_package.$reply_name> $method_name($plugin_package.$dto_package.$request_name request) {
+        return send(request, new $plugin_package.$dto_package.$reply_name());
+    }
+''')
+
+
+# Returns request name or special one from unconventional_naming_rep_req map
+def get_standard_dump_reply_name(camel_case_dto_name, func_name):
+    # FIXME this is a hotfix for sub-details callbacks
+    # FIXME also for L2FibTableEntry
+    # It's all because unclear mapping between
+    #  request -> reply,
+    #  dump -> reply, details,
+    #  notification_start -> reply, notifications
+
+    # vpe.api needs to be "standardized" so we can parse the information and create maps before generating java code
+    suffix = func_name.split("_")[-1]
+    return util.underscore_to_camelcase_upper(
+        util.unconventional_naming_rep_req[func_name]) + util.underscore_to_camelcase_upper(suffix) if func_name in util.unconventional_naming_rep_req \
+        else camel_case_dto_name
diff --git a/src/vpp-api/java/jvpp/gen/jvppgen/jvpp_impl_gen.py b/src/vpp-api/java/jvpp/gen/jvppgen/jvpp_impl_gen.py
new file mode 100644
index 00000000..7bf91138
--- /dev/null
+++ b/src/vpp-api/java/jvpp/gen/jvppgen/jvpp_impl_gen.py
@@ -0,0 +1,219 @@
+#!/usr/bin/env python
+#
+# Copyright (c) 2016 Cisco and/or its affiliates.
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at:
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+import os, util
+from string import Template
+
+jvpp_ifc_template = Template("""
+package $plugin_package;
+
+/**
+ * 
Java representation of plugin's api file.
+ * 
It was generated by jvpp_impl_gen.py based on $inputfile
+ * 
(python representation of api file generated by vppapigen).
+ */
+public interface JVpp${plugin_name} extends $base_package.JVpp {
+
+    /**
+     * Generic dispatch method for sending requests to VPP
+     *
+     * @throws io.fd.vpp.jvpp.VppInvocationException if send request had failed
+     */
+    int send($base_package.$dto_package.JVppRequest request) throws io.fd.vpp.jvpp.VppInvocationException;
+
+$methods
+}
+""")
+
+jvpp_impl_template = Template("""
+package $plugin_package;
+
+import java.io.IOException;
+import java.io.InputStream;
+import java.nio.file.Files;
+import java.nio.file.Path;
+import java.nio.file.StandardCopyOption;
+import java.nio.file.attribute.PosixFilePermission;
+import java.nio.file.attribute.PosixFilePermissions;
+import java.util.Set;
+import java.util.logging.Logger;
+import $base_package.callback.JVppCallback;
+import $base_package.VppConnection;
+import $base_package.JVppRegistry;
+
+/**
+ * 
Default implementation of JVpp interface.
+ * 
It was generated by jvpp_impl_gen.py based on $inputfile
+ * 
(python representation of api file generated by vppapigen).
+ */
+public final class JVpp${plugin_name}Impl implements $plugin_package.JVpp${plugin_name} {
+
+    private final static Logger LOG = Logger.getLogger(JVpp${plugin_name}Impl.class.getName());
+    private static final String LIBNAME = "libjvpp_${plugin_name_underscore}.so";
+
+    // FIXME using NativeLibraryLoader makes load fail could not find (WantInterfaceEventsReply).
+    static {
+        try {
+            loadLibrary();
+        } catch (Exception e) {
+            LOG.severe("Can't find jvpp jni library: " + LIBNAME);
+            throw new ExceptionInInitializerError(e);
+        }
+    }
+
+    private static void loadStream(final InputStream is) throws IOException {
+        final Set perms = PosixFilePermissions.fromString("rwxr-x---");
+        final Path p = Files.createTempFile(LIBNAME, null, PosixFilePermissions.asFileAttribute(perms));
+        try {
+            Files.copy(is, p, StandardCopyOption.REPLACE_EXISTING);
+
+            try {
+                Runtime.getRuntime().load(p.toString());
+            } catch (UnsatisfiedLinkError e) {
+                throw new IOException("Failed to load library " + p, e);
+            }
+        } finally {
+            try {
+                Files.deleteIfExists(p);
+            } catch (IOException e) {
+            }
+        }
+    }
+
+    private static void loadLibrary() throws IOException {
+        try (final InputStream is = JVpp${plugin_name}Impl.class.getResourceAsStream('/' + LIBNAME)) {
+            if (is == null) {
+                throw new IOException("Failed to open library resource " + LIBNAME);
+            }
+            loadStream(is);
+        }
+    }
+
+    private VppConnection connection;
+    private JVppRegistry registry;
+
+    private static native void init0(final JVppCallback callback, final long queueAddress, final int clientIndex);
+    @Override
+    public void init(final JVppRegistry registry, final JVppCallback callback, final long queueAddress, final int clientIndex) {
+        this.registry = java.util.Objects.requireNonNull(registry, "registry should not be null");
+        this.connection = java.util.Objects.requireNonNull(registry.getConnection(), "connection should not be null");
+        connection.checkActive();
+        init0(callback, queueAddress, clientIndex);
+    }
+
+    private static native void close0();
+    @Override
+    public void close() {
+        close0();
+    }
+
+    @Override
+    public int send($base_package.$dto_package.JVppRequest request) throws io.fd.vpp.jvpp.VppInvocationException {
+        return request.send(this);
+    }
+
+    @Override
+    public final int controlPing(final io.fd.vpp.jvpp.dto.ControlPing controlPing) throws io.fd.vpp.jvpp.VppInvocationException {
+        return registry.controlPing(JVpp${plugin_name}Impl.class);
+    }
+
+$methods
+}
+""")
+
+method_template = Template("""    int $name($plugin_package.$dto_package.$request request) throws io.fd.vpp.jvpp.VppInvocationException;""")
+method_native_template = Template(
+    """    private static native int ${name}0($plugin_package.$dto_package.$request request);""")
+method_impl_template = Template("""    public final int $name($plugin_package.$dto_package.$request request) throws io.fd.vpp.jvpp.VppInvocationException {
+        java.util.Objects.requireNonNull(request,"Null request object");
+        connection.checkActive();
+        int result=${name}0(request);
+        if(result<0){
+            throw new io.fd.vpp.jvpp.VppInvocationException("${name}",result);
+        }
+        return result;
+    }
+""")
+
+no_arg_method_template = Template("""    int $name() throws io.fd.vpp.jvpp.VppInvocationException;""")
+no_arg_method_native_template = Template("""    private static native int ${name}0() throws io.fd.vpp.jvpp.VppInvocationException;""")
+no_arg_method_impl_template = Template("""    public final int $name() throws io.fd.vpp.jvpp.VppInvocationException {
+        connection.checkActive();
+        int result=${name}0();
+        if(result<0){
+            throw new io.fd.vpp.jvpp.VppInvocationException("${name}",result);
+        }
+        return result;
+    }
+""")
+
+
+def generate_jvpp(func_list, base_package, plugin_package, plugin_name_underscore, dto_package, inputfile):
+    """ Generates JVpp interface and JNI implementation """
+    print "Generating JVpp"
+    plugin_name = util.underscore_to_camelcase_upper(plugin_name_underscore)
+
+    methods = []
+    methods_impl = []
+    for func in func_list:
+
+        # Skip structures that are used only as notifications
+        if util.is_just_notification(func['name']) or util.is_ignored(func['name']):
+            continue
+
+        camel_case_name = util.underscore_to_camelcase(func['name'])
+        camel_case_name_upper = util.underscore_to_camelcase_upper(func['name'])
+        if util.is_reply(camel_case_name):
+            continue
+
+        if len(func['args']) == 0:
+            methods.append(no_arg_method_template.substitute(name=camel_case_name))
+            methods_impl.append(no_arg_method_native_template.substitute(name=camel_case_name))
+            methods_impl.append(no_arg_method_impl_template.substitute(name=camel_case_name))
+        else:
+            methods.append(method_template.substitute(name=camel_case_name,
+                                                      request=camel_case_name_upper,
+                                                      plugin_package=plugin_package,
+                                                      dto_package=dto_package))
+            methods_impl.append(method_native_template.substitute(name=camel_case_name,
+                                                                  request=camel_case_name_upper,
+                                                                  plugin_package=plugin_package,
+                                                                  dto_package=dto_package))
+            methods_impl.append(method_impl_template.substitute(name=camel_case_name,
+                                                                request=camel_case_name_upper,
+                                                                plugin_package=plugin_package,
+                                                                dto_package=dto_package))
+
+    jvpp_file = open("JVpp%s.java" % plugin_name, 'w')
+    jvpp_file.write(
+        jvpp_ifc_template.substitute(inputfile=inputfile,
+                                     methods="\n".join(methods),
+                                     base_package=base_package,
+                                     plugin_package=plugin_package,
+                                     plugin_name=plugin_name,
+                                     dto_package=dto_package))
+    jvpp_file.flush()
+    jvpp_file.close()
+
+    jvpp_file = open("JVpp%sImpl.java" % plugin_name, 'w')
+    jvpp_file.write(jvpp_impl_template.substitute(inputfile=inputfile,
+                                                  methods="\n".join(methods_impl),
+                                                  base_package=base_package,
+                                                  plugin_package=plugin_package,
+                                                  plugin_name=plugin_name,
+                                                  plugin_name_underscore=plugin_name_underscore,
+                                                  dto_package=dto_package))
+    jvpp_file.flush()
+    jvpp_file.close()
diff --git a/src/vpp-api/java/jvpp/gen/jvppgen/notification_gen.py b/src/vpp-api/java/jvpp/gen/jvppgen/notification_gen.py
new file mode 100644
index 00000000..94302d56
--- /dev/null
+++ b/src/vpp-api/java/jvpp/gen/jvppgen/notification_gen.py
@@ -0,0 +1,199 @@
+#!/usr/bin/env python
+#
+# Copyright (c) 2016 Cisco and/or its affiliates.
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at:
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+import os
+
+import callback_gen
+import util
+from string import Template
+
+notification_registry_template = Template("""
+package $plugin_package.$notification_package;
+
+/**
+ * 
Registry for notification callbacks defined in ${plugin_name}.
+ * 
It was generated by notification_gen.py based on $inputfile
+ * 
(python representation of api file generated by vppapigen).
+ */
+public interface ${plugin_name}NotificationRegistry extends $base_package.$notification_package.NotificationRegistry {
+
+    $register_callback_methods
+
+    @Override
+    void close();
+}
+""")
+
+global_notification_callback_template = Template("""
+package $plugin_package.$notification_package;
+
+/**
+ * 
Aggregated callback interface for notifications only.
+ * 
It was generated by notification_gen.py based on $inputfile
+ * 
(python representation of api file generated by vppapigen).
+ */
+public interface Global${plugin_name}NotificationCallback$callbacks {
+
+}
+""")
+
+notification_registry_impl_template = Template("""
+package $plugin_package.$notification_package;
+
+/**
+ * 
Notification registry delegating notification processing to registered callbacks.
+ * 
It was generated by notification_gen.py based on $inputfile
+ * 
(python representation of api file generated by vppapigen).
+ */
+public final class ${plugin_name}NotificationRegistryImpl implements ${plugin_name}NotificationRegistry, Global${plugin_name}NotificationCallback {
+
+    // TODO add a special NotificationCallback interface and only allow those to be registered
+    private final java.util.concurrent.ConcurrentMap, $base_package.$callback_package.JVppNotificationCallback> registeredCallbacks =
+        new java.util.concurrent.ConcurrentHashMap<>();
+
+    $register_callback_methods
+    $handler_methods
+
+    @Override
+    public void close() {
+        registeredCallbacks.clear();
+    }
+}
+""")
+
+register_callback_impl_template = Template("""
+    public java.lang.AutoCloseable register$callback(final $plugin_package.$callback_package.$callback callback){
+        if(null != registeredCallbacks.putIfAbsent($plugin_package.$dto_package.$notification.class, callback)){
+            throw new IllegalArgumentException("Callback for " + $plugin_package.$dto_package.$notification.class +
+                "notification already registered");
+        }
+        return () -> registeredCallbacks.remove($plugin_package.$dto_package.$notification.class);
+    }
+""")
+
+handler_impl_template = Template("""
+    @Override
+    public void on$notification(
+        final $plugin_package.$dto_package.$notification notification) {
+        final $base_package.$callback_package.JVppNotificationCallback jVppNotificationCallback = registeredCallbacks.get($plugin_package.$dto_package.$notification.class);
+        if (null != jVppNotificationCallback) {
+            (($plugin_package.$callback_package.$callback) registeredCallbacks
+                .get($plugin_package.$dto_package.$notification.class))
+                .on$notification(notification);
+        }
+    }
+""")
+
+notification_provider_template = Template("""
+package $plugin_package.$notification_package;
+
+ /**
+ * Provides ${plugin_name}NotificationRegistry.
+ * 
The file was generated by notification_gen.py based on $inputfile
+ * 
(python representation of api file generated by vppapigen).
+ */
+public interface ${plugin_name}NotificationRegistryProvider extends $base_package.$notification_package.NotificationRegistryProvider {
+
+    @Override
+    public ${plugin_name}NotificationRegistry getNotificationRegistry();
+}
+""")
+
+
+def generate_notification_registry(func_list, base_package, plugin_package, plugin_name, notification_package, callback_package, dto_package, inputfile):
+    """ Generates notification registry interface and implementation """
+    print "Generating Notification interfaces and implementation"
+
+    if not os.path.exists(notification_package):
+        os.mkdir(notification_package)
+
+    callbacks = []
+    register_callback_methods = []
+    register_callback_methods_impl = []
+    handler_methods = []
+    for func in func_list:
+
+        if not util.is_notification(func['name']):
+            continue
+
+        camel_case_name_with_suffix = util.underscore_to_camelcase_upper(func['name'])
+        notification_dto = util.add_notification_suffix(camel_case_name_with_suffix)
+        callback_ifc = notification_dto + callback_gen.callback_suffix
+        fully_qualified_callback_ifc = "{0}.{1}.{2}".format(plugin_package, callback_package, callback_ifc)
+        callbacks.append(fully_qualified_callback_ifc)
+
+        # TODO create NotificationListenerRegistration and return that instead of AutoCloseable to better indicate
+        # that the registration should be closed
+        register_callback_methods.append("java.lang.AutoCloseable register{0}({1} callback);"
+                                         .format(callback_ifc, fully_qualified_callback_ifc))
+        register_callback_methods_impl.append(register_callback_impl_template.substitute(plugin_package=plugin_package,
+                                                                                         callback_package=callback_package,
+                                                                                         dto_package=dto_package,
+                                                                                         notification=notification_dto,
+                                                                                         callback=callback_ifc))
+        handler_methods.append(handler_impl_template.substitute(base_package=base_package,
+                                                                plugin_package=plugin_package,
+                                                                callback_package=callback_package,
+                                                                dto_package=dto_package,
+                                                                notification=notification_dto,
+                                                                callback=callback_ifc))
+
+
+    callback_file = open(os.path.join(notification_package, "%sNotificationRegistry.java" % plugin_name), 'w')
+    callback_file.write(notification_registry_template.substitute(inputfile=inputfile,
+                                                                register_callback_methods="\n    ".join(register_callback_methods),
+                                                                base_package=base_package,
+                                                                plugin_package=plugin_package,
+                                                                plugin_name=plugin_name,
+                                                                notification_package=notification_package))
+    callback_file.flush()
+    callback_file.close()
+
+    callback_file = open(os.path.join(notification_package, "Global%sNotificationCallback.java" % plugin_name), 'w')
+
+    global_notification_callback_callbacks = ""
+    if (callbacks):
+        global_notification_callback_callbacks = " extends " + ", ".join(callbacks)
+
+    callback_file.write(global_notification_callback_template.substitute(inputfile=inputfile,
+                                                                       callbacks=global_notification_callback_callbacks,
+                                                                       plugin_package=plugin_package,
+                                                                       plugin_name=plugin_name,
+                                                                       notification_package=notification_package))
+    callback_file.flush()
+    callback_file.close()
+
+    callback_file = open(os.path.join(notification_package, "%sNotificationRegistryImpl.java" % plugin_name), 'w')
+    callback_file.write(notification_registry_impl_template.substitute(inputfile=inputfile,
+                                                                     callback_package=callback_package,
+                                                                     dto_package=dto_package,
+                                                                     register_callback_methods="".join(register_callback_methods_impl),
+                                                                     handler_methods="".join(handler_methods),
+                                                                     base_package=base_package,
+                                                                     plugin_package=plugin_package,
+                                                                     plugin_name=plugin_name,
+                                                                     notification_package=notification_package))
+    callback_file.flush()
+    callback_file.close()
+
+    callback_file = open(os.path.join(notification_package, "%sNotificationRegistryProvider.java" % plugin_name), 'w')
+    callback_file.write(notification_provider_template.substitute(inputfile=inputfile,
+                                                                     base_package=base_package,
+                                                                     plugin_package=plugin_package,
+                                                                     plugin_name=plugin_name,
+                                                                     notification_package=notification_package))
+    callback_file.flush()
+    callback_file.close()
+
diff --git a/src/vpp-api/java/jvpp/gen/jvppgen/types_gen.py b/src/vpp-api/java/jvpp/gen/jvppgen/types_gen.py
new file mode 100644
index 00000000..7a5eec37
--- /dev/null
+++ b/src/vpp-api/java/jvpp/gen/jvppgen/types_gen.py
@@ -0,0 +1,227 @@
+#!/usr/bin/env python
+#
+# Copyright (c) 2016 Cisco and/or its affiliates.
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at:
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+import os
+from string import Template
+
+import util
+import jni_gen
+import dto_gen
+
+type_template = Template("""
+package $plugin_package.$type_package;
+
+/**
+ * 
This class represents $c_type_name type definition.
+ * 
It was generated by types_gen.py based on $inputfile preparsed data:
+ * 
+$docs
+ * 

+ */
+public final class $java_type_name {
+$fields
+$methods
+}
+""")
+
+field_template = Template("""    public $type $name;\n""")
+
+
+def generate_type_fields(type_definition):
+    """
+    Generates fields for class representing typeonly definition
+    :param type_definition: python representation of typeonly definition
+    :return: string representing class fields
+    """
+    fields = ""
+    for t in zip(type_definition['types'], type_definition['args']):
+        field_name = util.underscore_to_camelcase(t[1])
+        fields += field_template.substitute(type=util.jni_2_java_type_mapping[t[0]],
+                                            name=field_name)
+    return fields
+
+object_struct_setter_template = Template("""
+    {
+        jclass ${field_reference_name}Class = (*env)->FindClass(env, "${class_FQN}");
+        memset (&(mp->${c_name}), 0, sizeof (mp->${c_name}));
+        ${struct_initialization}
+    }
+""")
+
+object_array_struct_setter_template = Template("""
+    {
+        jclass ${field_reference_name}ArrayElementClass = (*env)->FindClass(env, "${class_FQN}");
+        if (${field_reference_name}) {
+            size_t _i;
+            jsize cnt = (*env)->GetArrayLength (env, ${field_reference_name});
+            ${field_length_check}
+            for (_i = 0; _i < cnt; _i++) {
+                jobject ${field_reference_name}ArrayElement = (*env)->GetObjectArrayElement(env, ${field_reference_name}, _i);
+                memset (&(mp->${c_name}[_i]), 0, sizeof (mp->${c_name}[_i]));
+                ${struct_initialization}
+            }
+        }
+    }
+""")
+
+object_dto_field_setter_template = Template("""
+    {
+        jclass ${field_reference_name}Class = (*env)->FindClass(env, "${class_FQN}");
+        jmethodID ${field_reference_name}Constructor = (*env)->GetMethodID(env, ${field_reference_name}Class, "", "()V");
+        jobject ${field_reference_name} = (*env)->NewObject(env, ${field_reference_name}Class,  ${field_reference_name}Constructor);
+        ${type_initialization}
+        (*env)->SetObjectField(env, dto, ${field_reference_name}FieldId, ${field_reference_name});
+    }
+""")
+
+object_array_dto_field_setter_template = Template("""
+    {
+        jclass ${field_reference_name}Class = (*env)->FindClass(env, "${class_FQN}");
+        jobjectArray ${field_reference_name} = (*env)->NewObjectArray(env, ${field_length}, ${field_reference_name}Class, 0);
+        unsigned int _i;
+        for (_i = 0; _i < ${field_length}; _i++) {
+            jmethodID ${field_reference_name}Constructor = (*env)->GetMethodID(env, ${field_reference_name}Class, "", "()V");
+            jobject ${field_reference_name}ArrayElement = (*env)->NewObject(env, ${field_reference_name}Class,  ${field_reference_name}Constructor);
+            ${type_initialization}
+            (*env)->SetObjectArrayElement(env, ${field_reference_name}, _i, ${field_reference_name}ArrayElement);
+        }
+        (*env)->SetObjectField(env, dto, ${field_reference_name}FieldId, ${field_reference_name});
+    }
+""")
+
+
+def generate_struct_initialization(type_def, c_name_prefix, object_name, indent):
+    struct_initialization = ""
+    # field identifiers
+    for t in zip(type_def['types'], type_def['args'], type_def['lengths']):
+        field_reference_name = "${c_name}" + util.underscore_to_camelcase_upper(t[1])
+        field_name = util.underscore_to_camelcase(t[1])
+        struct_initialization += jni_gen.jni_request_binding_for_type(field_type=t[0], c_name=c_name_prefix + t[1],
+                                                                     field_reference_name=field_reference_name,
+                                                                     field_name=field_name,
+                                                                     field_length=t[2][0],
+                                                                     is_variable_len_array=t[2][1],
+                                                                     object_name=object_name)
+    return indent + struct_initialization.replace('\n', '\n' + indent)
+
+
+def generate_type_setter(handler_name, type_def, c_name_prefix, object_name, indent):
+    type_initialization = ""
+    for t in zip(type_def['types'], type_def['args'], type_def['lengths']):
+        field_length = t[2][0]
+        is_variable_len_array = t[2][1]
+        length_field_type = None
+        if is_variable_len_array:
+            length_field_type = type_def['types'][type_def['args'].index(field_length)]
+        type_initialization += jni_gen.jni_reply_handler_for_type(handler_name=handler_name,
+                                                                  ref_name="${field_reference_name}",
+                                                                  field_type=t[0], c_name=c_name_prefix + t[1],
+                                                                  field_reference_name="${c_name}" + util.underscore_to_camelcase_upper(t[1]),
+                                                                  field_name=util.underscore_to_camelcase(t[1]),
+                                                                  field_length=field_length,
+                                                                  is_variable_len_array=is_variable_len_array,
+                                                                  length_field_type=length_field_type,
+                                                                  object_name=object_name)
+    return indent + type_initialization.replace('\n', '\n' + indent)
+
+
+def generate_types(types_list, plugin_package, types_package, inputfile):
+    """
+    Generates Java representation of custom types defined in api file.
+    """
+
+    #
+    if not types_list:
+        print "Skipping custom types generation (%s does not define custom types)." % inputfile
+        return
+
+    print "Generating custom types"
+
+    if not os.path.exists(types_package):
+        os.mkdir(types_package)
+
+    for type in types_list:
+        c_type_name = type['name']
+        java_type_name = util.underscore_to_camelcase_upper(type['name'])
+        dto_path = os.path.join(types_package, java_type_name + ".java")
+
+        fields = generate_type_fields(type)
+
+        dto_file = open(dto_path, 'w')
+        dto_file.write(type_template.substitute(plugin_package=plugin_package,
+                                                type_package=types_package,
+                                                c_type_name=c_type_name,
+                                                inputfile=inputfile,
+                                                docs=util.api_message_to_javadoc(type),
+                                                java_type_name=java_type_name,
+                                                fields=fields,
+                                                methods=dto_gen.generate_dto_base_methods(java_type_name, type)
+                                                ))
+
+        # update type mappings:
+        # todo fix vpe.api to use type_name instead of vl_api_type_name_t
+        type_name = "vl_api_" + c_type_name + "_t"
+        java_fqn = "%s.%s.%s" % (plugin_package, types_package, java_type_name)
+        util.vpp_2_jni_type_mapping[type_name] = "jobject"
+        util.vpp_2_jni_type_mapping[type_name + "[]"] = "jobjectArray"
+        util.jni_2_java_type_mapping[type_name] = java_fqn
+        util.jni_2_java_type_mapping[type_name + "[]"] = java_fqn + "[]"
+        jni_name = java_fqn.replace('.', "/")
+        jni_signature = "L" + jni_name + ";"
+        util.jni_2_signature_mapping[type_name] = "L" + jni_name + ";"
+        util.jni_2_signature_mapping[type_name + "[]"] = "[" + jni_signature
+        util.jni_field_accessors[type_name] = "ObjectField"
+        util.jni_field_accessors[type_name + "[]"] = "ObjectField"
+
+        jni_gen.struct_setter_templates[type_name] = Template(
+                object_struct_setter_template.substitute(
+                        c_name="${c_name}",
+                        field_reference_name="${field_reference_name}",
+                        class_FQN=jni_name,
+                        struct_initialization=generate_struct_initialization(type, "${c_name}.",
+                                                                           "${field_reference_name}", ' ' * 4))
+        )
+
+        jni_gen.struct_setter_templates[type_name+ "[]"] = Template(
+                object_array_struct_setter_template.substitute(
+                        c_name="${c_name}",
+                        field_reference_name="${field_reference_name}",
+                        field_length_check="${field_length_check}",
+                        class_FQN=jni_name,
+                        struct_initialization=generate_struct_initialization(type, "${c_name}[_i].",
+                                                                           "${field_reference_name}ArrayElement", ' ' * 8))
+        )
+
+        jni_gen.dto_field_setter_templates[type_name] = Template(
+                object_dto_field_setter_template.substitute(
+                        field_reference_name="${field_reference_name}",
+                        field_length="${field_length}",
+                        class_FQN=jni_name,
+                        type_initialization=generate_type_setter(c_type_name, type, "${c_name}.",
+                                                                 "${field_reference_name}", ' ' * 4))
+        )
+
+        jni_gen.dto_field_setter_templates[type_name + "[]"] = Template(
+                object_array_dto_field_setter_template.substitute(
+                        field_reference_name="${field_reference_name}",
+                        field_length="${field_length}",
+                        class_FQN=jni_name,
+                        type_initialization=generate_type_setter(c_type_name, type, "${c_name}[_i].",
+                                                                 "${field_reference_name}ArrayElement", ' ' * 8))
+        )
+
+        dto_file.flush()
+        dto_file.close()
+
diff --git a/src/vpp-api/java/jvpp/gen/jvppgen/util.py b/src/vpp-api/java/jvpp/gen/jvppgen/util.py
new file mode 100644
index 00000000..fc971c17
--- /dev/null
+++ b/src/vpp-api/java/jvpp/gen/jvppgen/util.py
@@ -0,0 +1,220 @@
+#!/usr/bin/env python
+#
+# Copyright (c) 2016 Cisco and/or its affiliates.
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at:
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+import os, pprint
+from os import removedirs
+
+
+def underscore_to_camelcase(name):
+    name = name.title().replace("_", "")
+    return name[0].lower() + name[1:]
+
+
+def underscore_to_camelcase_upper(name):
+    name = name.title().replace("_", "")
+    return name[0].upper() + name[1:]
+
+
+def remove_folder(folder):
+    """ Remove folder with all its files """
+    for root, dirs, files in os.walk(folder, topdown=False):
+        for name in files:
+            os.remove(os.path.join(root, name))
+        removedirs(folder)
+
+
+reply_suffixes = ("reply", "details", "l2fibtableentry")
+
+
+def is_reply(name):
+    return name.lower().endswith(reply_suffixes)
+
+
+def is_details(name):
+    return name.lower().endswith(reply_suffixes[1]) or name.lower().endswith(reply_suffixes[2])
+
+
+def is_retval_field(name):
+    return name == 'retval'
+
+dump_suffix = "dump"
+
+
+def is_dump(name):
+    return name.lower().endswith(dump_suffix)
+
+
+def get_reply_suffix(name):
+    for reply_suffix in reply_suffixes:
+        if name.lower().endswith(reply_suffix):
+            if reply_suffix == reply_suffixes[2]:
+                # FIXME workaround for l2_fib_table_entry
+                return 'entry'
+            else:
+                return reply_suffix
+
+# Mapping according to:
+# http://docs.oracle.com/javase/7/docs/technotes/guides/jni/spec/types.html
+#
+# Unsigned types are converted to signed java types that have the same size.
+# It is the API user responsibility to interpret them correctly.
+jni_2_java_type_mapping = {'u8': 'byte',
+                           'u8[]': 'byte[]',
+                           'i8': 'byte',
+                           'i8[]': 'byte[]',
+                           'u16': 'short',
+                           'u16[]': 'short[]',
+                           'i16': 'short',
+                           'i16[]': 'short[]',
+                           'u32': 'int',
+                           'u32[]': 'int[]',
+                           'i32': 'int',
+                           'i32[]': 'int[]',
+                           'u64': 'long',
+                           'u64[]': 'long[]',
+                           'i64': 'long',
+                           'i64[]': 'long[]',
+                           'f64': 'double',
+                           'f64[]': 'double[]'
+                           }
+
+vpp_2_jni_type_mapping = {'u8': 'jbyte',
+                          'u8[]': 'jbyteArray',
+                          'i8': 'jbyte',
+                          'u8[]': 'jbyteArray',
+                          'u16': 'jshort',
+                          'u16[]': 'jshortArray',
+                          'i16': 'jshort',
+                          'i16[]': 'jshortArray',
+                          'u32': 'jint',
+                          'u32[]': 'jintArray',
+                          'i32': 'jint',
+                          'i32[]': 'jintArray',
+                          'u64': 'jlong',
+                          'u64[]': 'jlongArray',
+                          'i64': 'jlong',
+                          'i64[]': 'jlongArray',
+                          'f64': 'jdouble',
+                          'f64[]': 'jdoubleArray'
+                          }
+
+# https://docs.oracle.com/javase/8/docs/technotes/guides/jni/spec/types.html#type_signatures
+jni_2_signature_mapping = {'u8': 'B',
+                           'u8[]': '[B',
+                           'i8': 'B',
+                           'i8[]': '[B',
+                           'u16': 'S',
+                           'u16[]': '[S',
+                           'i16': 'S',
+                           'i16[]': '[S',
+                           'u32': 'I',
+                           'u32[]': '[I',
+                           'i32': 'I',
+                           'i32[]': '[I',
+                           'u64': 'J',
+                           'u64[]': '[J',
+                           'i64': 'J',
+                           'i64[]': '[J',
+                           'f64': 'D',
+                           'f64[]': '[D'
+                           }
+
+# https://docs.oracle.com/javase/8/docs/technotes/guides/jni/spec/functions.html#Get_type_Field_routines
+jni_field_accessors =  {'u8': 'ByteField',
+                        'u8[]': 'ObjectField',
+                        'i8': 'ByteField',
+                        'i8[]': 'ObjectField',
+                        'u16': 'ShortField',
+                        'u16[]': 'ObjectField',
+                        'i16': 'ShortField',
+                        'i16[]': 'ObjectField',
+                        'u32': 'IntField',
+                        'u32[]': 'ObjectField',
+                        'i32': 'IntField',
+                        'i32[]': 'ObjectField',
+                        'u64': 'LongField',
+                        'u64[]': 'ObjectField',
+                        'i64': 'LongField',
+                        'i64[]': 'ObjectField',
+                        'f64': 'DoubleField',
+                        'f64[]': 'ObjectField'
+                        }
+
+
+# vpe.api calls that do not follow naming conventions and have to be handled exceptionally when finding reply -> request mapping
+# FIXME in vpe.api
+unconventional_naming_rep_req = {
+                                 'cli_reply': 'cli_request',
+                                 'vnet_summary_stats_reply': 'vnet_get_summary_stats',
+                                 # This below is actually a sub-details callback. We cannot derive the mapping of dump request
+                                 # belonging to this sub-details from naming conventions. We need special mapping
+                                 'bridge_domain_sw_if_details': 'bridge_domain',
+                                 # This is standard dump call + details reply. However it's not called details but entry
+                                 'l2_fib_table_entry': 'l2_fib_table'
+                                 }
+
+#
+# FIXME no convention in the naming of events (notifications) in vpe.api
+notifications_message_suffixes = ("event", "counters")
+notification_messages_reused = ["sw_interface_set_flags"]
+
+# messages that must be ignored. These messages are INSUFFICIENTLY marked as disabled in vpe.api
+# FIXME
+ignored_messages = ["is_address_reachable"]
+
+
+def is_notification(name):
+    """ Returns true if the structure is a notification regardless of its no other use """
+    return is_just_notification(name) or name.lower() in notification_messages_reused
+
+
+def is_just_notification(name):
+    """ Returns true if the structure is just a notification and has no other use """
+    return name.lower().endswith(notifications_message_suffixes)
+
+
+def is_ignored(param):
+    return param.lower() in ignored_messages
+
+
+def remove_reply_suffix(camel_case_name_with_suffix):
+    return remove_suffix(camel_case_name_with_suffix, get_reply_suffix(camel_case_name_with_suffix))
+
+
+def remove_suffix(camel_case_name_with_suffix, suffix):
+    suffix_length = len(suffix)
+    return camel_case_name_with_suffix[:-suffix_length] if suffix_length != 0 else camel_case_name_with_suffix
+
+
+def is_control_ping(camel_case_name_with_suffix):
+    return camel_case_name_with_suffix.lower().startswith("controlping");
+
+
+def api_message_to_javadoc(api_message):
+    """ Converts vpe.api message description to javadoc """
+    str = pprint.pformat(api_message, indent=4, width=120, depth=None)
+    return " * " + str.replace("\n", "\n * ")
+
+
+notification_dto_suffix = "Notification"
+
+
+def add_notification_suffix(camel_case_dto_name):
+    camel_case_dto_name += notification_dto_suffix
+    return camel_case_dto_name
+
+
+def is_array(java_type_as_string):
+    return java_type_as_string.endswith("[]")
diff --git a/src/vpp-api/lua/README.md b/src/vpp-api/lua/README.md
new file mode 100644
index 00000000..4ecdb34d
--- /dev/null
+++ b/src/vpp-api/lua/README.md
@@ -0,0 +1,50 @@
+This is the experimental version of Lua API, aimed for the luajit use.
+
+Please take a look and send the feedback to ayourtch@gmail.com.
+
+To run the examples here:
+
+1) install luajit - "sudo apt-get install luajit" on ubuntu
+
+2) "make build-vpp-api" in the top VPP directory
+
+3) "make run" in a separate terminal window
+   This ensures you have an instance of VPP running
+
+4) sudo luajit examples/example-cli.lua
+
+This will result in something like this:
+
+Version:
+00000000  00 00 00 00 00 00 00 00  00 00 00 00 00 00 00 00  ................
+00000010  00 00 00 00 00 00 00 00  00 00 00 00 00 00 00 00  ................
+
+{ [1] = { ["luaapi_message_name"] = show_version_reply,["program"] = vpe,["version"] = ,["build_date"] = Fri Nov 25 10:58:48 UTC 2016,["retval"] = 0,["build_directory"] = /home/ubuntu/vpp,["_vl_msg_id"] = 170,["context"] = 0,} ,}
+---
+{ [1] = { ["luaapi_message_name"] = cli_inband_reply,["_vl_msg_id"] = 94,["length"] = 66,["reply"] = vpp v built by ubuntu on vpp-toys at Fri Nov 25 10:58:48 UTC 2016
+,["retval"] = 0,["context"] = 0,} ,}
+---
+
+5) You can also run the performance test bench:
+
+$ sudo luajit bench.lua
+10001 iterations, average speed 5624LL per second
+10001 iterations, average speed 6650LL per second
+10001 iterations, average speed 6053LL per second
+10001 iterations, average speed 7056LL per second
+10001 iterations, average speed 6388LL per second
+10001 iterations, average speed 5849LL per second
+10001 iterations, average speed 6321LL per second
+10001 iterations, average speed 6368LL per second
+10001 iterations, average speed 5958LL per second
+10001 iterations, average speed 6482LL per second
+Average tps across the tests: 6274LL
+
+Note: the above is run in an lxd container running inside 2-core
+xhyve VM on a Macbook Pro, so I would not take the performance numbers for granted :)
+
+The "examples" directory contains a few naive examples, as well as a couple of more 
+advanced ones - a tab-completing CLI for VPP that can call both the APIs and CLI,
+and also a small test utility which I use for automating some small tests using
+VPP.
+
diff --git a/src/vpp-api/lua/bench.lua b/src/vpp-api/lua/bench.lua
new file mode 100644
index 00000000..8e5a0b4b
--- /dev/null
+++ b/src/vpp-api/lua/bench.lua
@@ -0,0 +1,70 @@
+--[[
+/*
+ * Copyright (c) 2016 Cisco and/or its affiliates.
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at:
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+]]
+
+local vpp = require "vpp-lapi"
+
+local ffi = require "ffi"
+
+ffi.cdef([[
+   struct timespec {
+               long tv_sec;        /* seconds */
+               long tv_nsec;       /* nanoseconds */
+           };
+
+   int clock_gettime(int clk_id, struct timespec *tp);
+]])
+
+
+local time_cache = ffi.new("struct timespec[1]")
+local time_cache_1 = time_cache[0]
+function get_ns()
+  ffi.C.clock_gettime(0, time_cache)
+  return time_cache_1.tv_nsec + 1000000000 * time_cache_1.tv_sec
+end
+
+function do_bench()
+  local cycle_start = get_ns()
+  local n_iterations =  10000
+  local count = 1
+  for i = 1,n_iterations do
+    -- print(i)
+    vpp:api_call("show_version")
+    count = count + 1
+    -- print(i, "done")
+  end
+  cycle_end = get_ns()
+  local tps = n_iterations*1000000000LL/(cycle_end - cycle_start)
+  print (tostring(count) .. " iterations, average speed " .. tostring(tps) .. " per second")
+  return tps
+end
+
+root_dir = "/home/ubuntu/vpp"
+pneum_path = root_dir .. "/build-root/install-vpp_lite_debug-native/vpp-api/lib64/libpneum.so"
+vpp:init({ pneum_path = pneum_path })
+vpp:json_api(root_dir .. "/build-root/install-vpp_lite_debug-native/vpp/vpp-api/vpe.api.json")
+
+vpp:connect("lua-bench")
+local n_tests = 10
+local tps_acc = 0LL
+for i=1,n_tests do
+  tps_acc = tps_acc + do_bench()
+end
+print("Average tps across the tests: " .. tostring(tps_acc/n_tests))
+
+vpp:disconnect()
+
+
diff --git a/src/vpp-api/lua/examples/cli/README.md b/src/vpp-api/lua/examples/cli/README.md
new file mode 100644
index 00000000..3a5f8ee9
--- /dev/null
+++ b/src/vpp-api/lua/examples/cli/README.md
@@ -0,0 +1,5 @@
+This is a small experiment to have a wrapper CLI which can call both API functions as well as debug CLI.
+
+To facilitate tab completion and help, the API call names are broken up with spaces replacing the underscores.
+
+
diff --git a/src/vpp-api/lua/examples/cli/lua-cli.lua b/src/vpp-api/lua/examples/cli/lua-cli.lua
new file mode 100644
index 00000000..b3a24d7d
--- /dev/null
+++ b/src/vpp-api/lua/examples/cli/lua-cli.lua
@@ -0,0 +1,747 @@
+--[[
+/*
+ * Copyright (c) 2016 Cisco and/or its affiliates.
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at:
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+]]
+
+-- Experimental prototype CLI using API to VPP, with tab completion
+--
+-- Written by Andrew Yourtchenko (ayourtch@cisco.com) 2010,2016
+--
+
+vpp = require "vpp-lapi"
+
+
+local dotdotdot = "..."
+
+-- First the "readline" routine
+
+readln = {
+split = function(str, pat)
+  local t = {}  -- NOTE: use {n = 0} in Lua-5.0
+  local fpat = "(.-)" .. pat
+  local last_end = 1
+  if str then
+    local s, e, cap = str:find(fpat, 1)
+    while s do
+      if s ~= 1 or cap ~= "" then
+        table.insert(t,cap)
+      end
+      last_end = e+1
+      s, e, cap = str:find(fpat, last_end)
+    end
+    if last_end <= #str then
+      cap = str:sub(last_end)
+      table.insert(t, cap)
+    end
+  end
+  return t
+end,
+
+reader = function()
+  local rl = {}
+
+  rl.init = function()
+    os.execute("stty -icanon min 1 -echo")
+    rl.rawmode = true
+  end
+
+  rl.done = function()
+    os.execute("stty icanon echo")
+    rl.rawmode = false
+  end
+
+  rl.prompt = ">"
+  rl.history = { "" }
+  rl.history_index = 1
+  rl.history_length = 1
+
+  rl.hide_cmd = function()
+    local bs = string.char(8) .. " " .. string.char(8)
+    for i = 1, #rl.command do
+      io.stdout:write(bs)
+    end
+  end
+
+  rl.show_cmd = function()
+    if rl.command then
+      io.stdout:write(rl.command)
+    end
+  end
+
+  rl.store_history = function(cmd)
+    if cmd == "" then
+      return
+    end
+    rl.history[rl.history_length] = cmd
+    rl.history_length = rl.history_length + 1
+    rl.history_index = rl.history_length
+    rl.history[rl.history_length] = ""
+  end
+
+  rl.readln = function()
+    local done = false
+    local need_prompt = true
+    rl.command = ""
+
+    if not rl.rawmode then
+      rl.init()
+    end
+
+    while not done do
+      if need_prompt then
+        io.stdout:write(rl.prompt)
+	io.stdout:write(rl.command)
+        need_prompt = false
+      end
+
+      local ch = io.stdin:read(1)
+      if ch:byte(1) == 27 then
+        -- CONTROL
+        local ch2 = io.stdin:read(1)
+        -- arrows
+        if ch2:byte(1) == 91 then
+          local ch3 = io.stdin:read(1)
+          local b = ch3:byte(1)
+          if b == 65 then
+            ch = "UP"
+          elseif b == 66 then
+            ch = "DOWN"
+          elseif b == 67 then
+            ch = "RIGHT"
+          elseif b == 68 then
+            ch = "LEFT"
+          end
+          -- print("Byte: " .. ch3:byte(1))
+          -- if ch3:byte(1)
+        end
+      end
+
+      if ch == "?" then
+        io.stdout:write(ch)
+        io.stdout:write("\n")
+        if rl.help then
+          rl.help(rl)
+        end
+        need_prompt = true
+      elseif ch == "\t" then
+        if rl.tab_complete then
+          rl.tab_complete(rl)
+        end
+        io.stdout:write("\n")
+        need_prompt = true
+      elseif ch == "\n" then
+        io.stdout:write(ch)
+        done = true
+      elseif ch == "\004" then
+        io.stdout:write("\n")
+        rl.command = nil
+	done = true
+      elseif ch == string.char(127) then
+        if rl.command ~= "" then
+          io.stdout:write(string.char(8) .. " " .. string.char(8))
+          rl.command = string.sub(rl.command, 1, -2)
+        end
+      elseif #ch > 1 then
+        -- control char
+        if ch == "UP" then
+          rl.hide_cmd()
+          if rl.history_index == #rl.history then
+            rl.history[rl.history_index] = rl.command
+          end
+          if rl.history_index > 1 then
+            rl.history_index = rl.history_index - 1
+            rl.command = rl.history[rl.history_index]
+          end
+          rl.show_cmd()
+        elseif ch == "DOWN" then
+          rl.hide_cmd()
+          if rl.history_index < rl.history_length then
+            rl.history_index = rl.history_index + 1
+            rl.command = rl.history[rl.history_index]
+          end
+          rl.show_cmd()
+        end
+      else
+        io.stdout:write(ch)
+        rl.command = rl.command .. ch
+      end
+    end
+    if rl.command then
+      rl.store_history(rl.command)
+    end
+    return rl.command
+  end
+  return rl
+end
+
+}
+
+--[[
+
+r = reader()
+
+local done = false
+
+while not done do
+  local cmd = r.readln()
+  print("Command: " .. tostring(cmd))
+  if not cmd or cmd == "quit" then
+    done = true
+  end
+end
+
+r.done()
+
+]]
+
+--------- MDS show tech parser
+
+local print_section = nil
+local list_sections = false
+
+local curr_section = "---"
+local curr_parser = nil
+
+-- by default operate in batch mode
+local batch_mode = true
+
+local db = {}
+local device = {}
+device.output = {}
+local seen_section = {}
+
+function start_collection(name)
+  device = {}
+  seen_section = {}
+end
+
+function print_error(errmsg)
+  print("@#$:" .. errmsg)
+end
+
+function keys(tbl)
+  local t = {}
+  for k, v in pairs(tbl) do
+    table.insert(t, k)
+  end
+  return t
+end
+
+function tset (parent, ...)
+
+  -- print ('set', ...)
+
+  local len = select ('#', ...)
+  local key, value = select (len-1, ...)
+  local cutpoint, cutkey
+
+  for i=1,len-2 do
+
+    local key = select (i, ...)
+    local child = parent[key]
+
+    if value == nil then
+      if child == nil then  return
+      elseif next (child, next (child)) then  cutpoint = nil  cutkey = nil
+      elseif cutpoint == nil then  cutpoint = parent  cutkey = key  end
+
+    elseif child == nil then  child = {}  parent[key] = child  end
+
+    parent = child
+    end
+
+  if value == nil and cutpoint then  cutpoint[cutkey] = nil
+  else  parent[key] = value  return value  end
+  end
+
+
+function tget (parent, ...)
+  local len = select ('#', ...)
+  for i=1,len do
+    parent = parent[select (i, ...)]
+    if parent == nil then  break  end
+    end
+  return parent
+  end
+
+
+local pager_lines = 23
+local pager_printed = 0
+local pager_skipping = false
+local pager_filter_pipe = nil
+
+function pager_reset()
+  pager_printed = 0
+  pager_skipping = false
+  if pager_filter_pipe then
+    pager_filter_pipe:close()
+    pager_filter_pipe = nil
+  end
+end
+
+
+function print_more()
+  io.stdout:write(" --More-- ")
+end
+
+function print_nomore()
+  local bs = string.char(8)
+  local bs10 = bs ..  bs ..  bs ..  bs ..  bs ..  bs ..  bs ..  bs ..  bs ..  bs
+  io.stdout:write(bs10 .. "          " .. bs10)
+end
+
+function print_line(txt)
+  if pager_filter_pipe then
+    pager_filter_pipe:write(txt .. "\n")
+    return
+  end
+  if pager_printed >= pager_lines then
+    print_more()
+    local ch = io.stdin:read(1)
+    if ch == " " then
+      pager_printed = 0
+    elseif ch == "\n" then
+      pager_printed = pager_printed - 1
+    elseif ch == "q" then
+      pager_printed = 0
+      pager_skipping = true
+    end
+    print_nomore()
+  end
+  if not pager_skipping then
+    print(txt)
+    pager_printed = pager_printed + 1
+  else
+    -- skip printing
+  end
+end
+
+function paged_write(text)
+  local t = readln.split(text, "[\n]")
+  if string.sub(text, -1) == "\n" then
+    table.insert(t, "")
+  end
+  for i, v in ipairs(t) do
+    if i < #t then
+      print_line(v)
+    else
+      if pager_filter_pipe then
+        pager_filter_pipe:write(v)
+      else
+        io.stdout:write(v)
+      end
+    end
+  end
+end
+
+
+
+
+
+function get_choices(tbl, key)
+  local res = {}
+  for k, v in pairs(tbl) do
+    if string.sub(k, 1, #key) == key then
+      table.insert(res, k)
+    elseif 0 < #key and dotdotdot == k then
+      table.insert(res, k)
+    end
+  end
+  return res
+end
+
+function get_exact_choice(choices, val)
+  local exact_idx = nil
+  local substr_idx = nil
+  local substr_seen = false
+
+  if #choices == 1 then
+    if choices[1] == dotdotdot then
+      return 1
+    elseif string.sub(choices[1], 1, #val) == val then
+      return 1
+    else
+      return nil
+    end
+  else
+    for i, v in ipairs(choices) do
+      if v == val then
+        exact_idx = i
+        substr_seen = true
+      elseif choices[i] ~= dotdotdot and string.sub(choices[i], 1, #val) == val then
+        if substr_seen then
+          substr_idx = nil
+        else
+          substr_idx = i
+          substr_seen = true
+        end
+      elseif choices[i] == dotdotdot then
+        if substr_seen then
+          substr_idx = nil
+        else
+          substr_idx = i
+          substr_seen = true
+        end
+      end
+    end
+  end
+  return exact_idx or substr_idx
+end
+
+function device_cli_help(rl)
+  local key = readln.split(rl.command, "[ ]+")
+  local tree = rl.tree
+  local keylen = #key
+  local fullcmd = ""
+  local error = false
+  local terse = true
+
+  if ((#rl.command >= 1) and (string.sub(rl.command, -1) == " ")) or (#rl.command == 0) then
+    table.insert(key, "")
+    terse = false
+  end
+
+  for i, v in ipairs(key) do
+    local choices = get_choices(tree, v)
+    local idx = get_exact_choice(choices, v)
+    if idx then
+      local choice = choices[idx]
+      tree = tree[choice]
+      fullcmd = fullcmd .. choice .. " "
+    else
+      if i < #key then
+        error = true
+      end
+    end
+
+    if i == #key and not error then
+      for j, w in ipairs(choices) do
+        if terse then
+          paged_write(w .. "\t")
+        else
+          paged_write("  " .. w .. "\n")
+        end
+      end
+      paged_write("\n")
+      if terse then
+        paged_write(" \n")
+      end
+    end
+  end
+  pager_reset()
+end
+
+function device_cli_tab_complete(rl)
+  local key = readln.split(rl.command, "[ ]+")
+  local tree = rl.tree
+  local keylen = #key
+  local fullcmd = ""
+  local error = false
+
+  for i, v in ipairs(key) do
+    local choices = get_choices(tree, v)
+    local idx = get_exact_choice(choices, v)
+    if idx and choices[idx] ~= dotdotdot then
+      local choice = choices[idx]
+      tree = tree[choice]
+      -- print("level " .. i .. " '" .. choice .. "'")
+      fullcmd = fullcmd .. choice .. " "
+    else
+      -- print("level " .. i .. " : " .. table.concat(choices, " ") .. " ")
+      error = true
+    end
+  end
+  if not error then
+    rl.command = fullcmd
+  else
+    -- print("\n\nerror\n")
+  end
+  pager_reset()
+end
+
+function device_cli_exec(rl)
+
+  local cmd_nopipe = rl.command
+  local cmd_pipe = nil
+
+  local pipe1, pipe2 = string.find(rl.command, "[|]")
+  if pipe1 then
+    cmd_nopipe = string.sub(rl.command, 1, pipe1-1)
+    cmd_pipe = string.sub(rl.command, pipe2+1, -1)
+  end
+
+  local key = readln.split(cmd_nopipe .. " ", "[ ]+")
+  local tree = rl.tree
+  local keylen = #key
+  local fullcmd = ""
+  local error = false
+  local func = nil
+
+  if cmd_pipe then
+    pager_filter_pipe = io.popen(cmd_pipe, "w")
+  end
+
+
+  rl.choices = {}
+
+  for i, v in ipairs(key) do
+    local choices = get_choices(tree, v)
+    local idx = get_exact_choice(choices, v)
+    if idx then
+      local choice = choices[idx]
+      if i == #key then
+        func = tree[choice]
+      else
+        if choice == dotdotdot then
+          -- keep the tree the same, update the choice value to match the input string
+          choices[idx] = v
+          choice = v
+        else
+          tree = tree[choice]
+        end
+      end
+      -- print("level " .. i .. " '" .. choice .. "'")
+      table.insert(rl.choices, choice)
+    else
+      -- print("level " .. i .. " : " .. table.concat(choices, " ") .. " ")
+      error = true
+      return nil
+    end
+  end
+  return func
+end
+
+function populate_tree(commands)
+  local tree = {}
+
+  for k, v in pairs(commands) do
+    local key = readln.split(k .. " ", "[ ]+")
+    local xtree = tree
+    for i, kk in ipairs(key) do
+      if i == 1 and kk == "sh" then
+        kk = "show"
+      end
+      if i == #key then
+        if type(v) == "function" then
+          xtree[kk] = v
+        else
+          xtree[kk] = function(rl) paged_write(table.concat(v, "\n") .. "\n") end
+        end
+      else
+        if not xtree[kk] then
+          xtree[kk] = {}
+        end
+        xtree = xtree[kk]
+      end
+    end
+  end
+  return tree
+end
+
+function trim (s)
+  return (string.gsub(s, "^%s*(.-)%s*$", "%1"))
+end
+
+
+function init_vpp(vpp)
+  local root_dir = "/home/ubuntu/vpp"
+  local pneum_path = root_dir .. "/build-root/install-vpp_lite_debug-native/vpp-api/lib64/libpneum.so"
+
+  vpp:init({ pneum_path = pneum_path })
+
+  vpp:init({ pneum_path = pneum_path })
+  vpp:json_api(root_dir .. "/build-root/install-vpp_lite_debug-native/vpp/vpp-api/vpe.api.json")
+
+
+
+  vpp:connect("lua_cli")
+end
+
+function run_cli(vpp, cli)
+  local reply = vpp:api_call("cli_inband", { cmd = cli })
+  if reply and #reply == 1 then
+    local rep = reply[1]
+    if 0 == rep.retval then
+      return rep.reply
+    else
+      return "XXXXXLUACLI: API RETVAL ERROR : " .. tostring(rep.retval)
+    end
+  else
+    return "XXXXXLUACLI ERROR, RAW REPLY: " .. vpp.dump(reply)
+  end
+end
+
+
+function toprintablestring(s)
+  if type(s) == "string" then
+    return "\n"..vpp.hex_dump(s)
+  else
+    return tostring(s)
+  end
+end
+
+function interactive_cli(r)
+  while not done do
+    pager_reset()
+    local cmd = r.readln()
+    if not cmd then
+      done = true
+    elseif cmd == "quit" or cmd == "exit" then
+      done = true
+    else
+      local func = device_cli_exec(r)
+      if func then
+	func(r)
+      else
+	if trim(cmd) == "" then
+	else
+	  for i = 1, #r.prompt do
+	    paged_write(" ")
+	  end
+	  paged_write("^\n% Invalid input detected at '^' marker.\n\n")
+	end
+      end
+    end
+  end
+end
+
+device = {}
+device.output = {}
+
+init_vpp(vpp)
+cmds_str = run_cli(vpp, "?")
+vpp_cmds = readln.split(cmds_str, "\n")
+vpp_clis = {}
+
+for linenum, line in ipairs(vpp_cmds) do
+  local m,h = string.match(line, "^  (.-)  (.*)$")
+  if m and #m > 0 then
+    table.insert(vpp_clis, m)
+    device.output["vpp debug cli " .. m] = function(rl)
+      -- print("ARBITRARY CLI" .. vpp.dump(rl.choices))
+      print("LUACLI command: " .. table.concat(rl.choices, " "))
+      local sub = {}
+      --
+      for i=4, #rl.choices -1 do
+        table.insert(sub, rl.choices[i])
+      end
+      local cli = table.concat(sub, " ")
+      print("Running CLI: " .. tostring(cli))
+      paged_write(run_cli(vpp, cli))
+    end
+    device.output["vpp debug cli " .. m .. " " .. dotdotdot] = function(rl)
+      print("ARGH")
+    end
+
+    local ret = run_cli(vpp, "help " .. m)
+    device.output["help vpp debug cli " .. m] = { ret }
+  end
+end
+
+for linenum, line in ipairs(vpp_clis) do
+  -- print(line, ret)
+end
+
+for msgnum, msgname in pairs(vpp.msg_number_to_name) do
+  local cli, numspaces = string.gsub(msgname, "_", " ")
+  device.output["call " .. cli .. " " .. dotdotdot] = function(rl)
+    print("ARGH")
+  end
+  device.output["call " .. cli] = function(rl)
+    print("LUACLI command: " .. table.concat(rl.choices, " "))
+    print("Running API: " .. msgname) -- vpp.dump(rl.choices))
+    local out = {}
+    local args = {}
+    local ntaken = 0
+    local argname = ""
+    for i=(1+1+numspaces+1), #rl.choices-1 do
+      -- print(i, rl.choices[i])
+      if ntaken > 0 then
+        ntaken = ntaken -1
+      else
+        local fieldname = rl.choices[i]
+        local field = vpp.msg_name_to_fields[msgname][fieldname]
+        if field then
+          local s = rl.choices[i+1]
+          s=s:gsub("\\x(%x%x)",function (x) return string.char(tonumber(x,16)) end)
+          args[fieldname] = s
+          ntaken = 1
+        end
+      end
+    end
+    -- print("ARGS: ", vpp.dump(args))
+    local ret = vpp:api_call(msgname, args)
+    for i, reply in ipairs(ret) do
+      table.insert(out, "=================== Entry #" .. tostring(i))
+      for k, v in pairs(reply) do
+        table.insert(out, "   " .. tostring(k) .. " : " .. toprintablestring(v))
+      end
+    end
+    -- paged_write(vpp.dump(ret) .. "\n\n")
+    paged_write(table.concat(out, "\n").."\n\n")
+  end
+  device.output["call " .. cli .. " help"] = function(rl)
+    local out = {}
+    for k, v in pairs(vpp.msg_name_to_fields[msgname]) do
+      table.insert(out, tostring(k) .. " : " .. v["ctype"] .. " ; " .. tostring(vpp.dump(v)) )
+    end
+    -- paged_write(vpp.dump(vpp.msg_name_to_fields[msgname]) .. "\n\n")
+    paged_write(table.concat(out, "\n").."\n\n")
+  end
+-- vpp.msg_name_to_number = {}
+end
+
+
+
+local r = readln.reader()
+local done = false
+
+r.prompt = "VPP(luaCLI)#"
+
+r.help = device_cli_help
+r.tab_complete = device_cli_tab_complete
+print("===== CLI view, use ^D to end =====")
+
+r.tree = populate_tree(device.output)
+-- readln.pretty("xxxx", r.tree)
+
+
+for idx, an_arg in ipairs(arg) do
+  local fname = an_arg
+  if fname == "-i" then
+    pager_lines = 23
+    interactive_cli(r)
+  else
+    pager_lines = 100000000
+    for line in io.lines(fname) do
+      r.command = line
+      local func = device_cli_exec(r)
+      if func then
+	func(r)
+      end
+    end
+  end
+end
+
+if #arg == 0 then
+  print("You should specify '-i' as an argument for the interactive session,")
+  print("but with no other sources of commands, we start interactive session now anyway")
+   interactive_cli(r)
+end
+
+vpp:disconnect()
+r.done()
+
+
diff --git a/src/vpp-api/lua/examples/example-acl-plugin.lua b/src/vpp-api/lua/examples/example-acl-plugin.lua
new file mode 100644
index 00000000..ca01f18d
--- /dev/null
+++ b/src/vpp-api/lua/examples/example-acl-plugin.lua
@@ -0,0 +1,110 @@
+--[[
+/*
+ * Copyright (c) 2016 Cisco and/or its affiliates.
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at:
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+]]
+
+
+vpp = require "vpp-lapi"
+
+root_dir = "/home/ubuntu/vpp"
+pneum_path = root_dir .. "/build-root/install-vpp_debug-native/vpp-api/lib64/libpneum.so"
+
+vpp:init({ pneum_path = pneum_path })
+
+vpp:consume_api(root_dir .. "/build-root/install-vpp_debug-native/vlib-api/vlibmemory/memclnt.api")
+vpp:consume_api(root_dir .. "/build-root/install-vpp_debug-native/vpp/vpp-api/vpe.api")
+vpp:connect("aytest")
+vpp:consume_api(root_dir .. "/plugins/acl-plugin/acl/acl.api", "acl")
+
+-- api calls
+reply = vpp:api_call("show_version")
+print("Version: ", reply[1].version)
+print(vpp.hex_dump(reply[1].version))
+print(vpp.dump(reply))
+print("---")
+
+reply = vpp:api_call("acl_del", { context = 42, acl_index = 230 })
+print(vpp.dump(reply))
+print("---")
+
+reply = vpp:api_call("acl_del", { context = 42, acl_index = 8 })
+print(vpp.dump(reply))
+print("---")
+
+reply = vpp:api_call("acl_del", { context = 42, acl_index = 15 })
+print(vpp.dump(reply))
+print("---")
+
+reply = vpp:api_call("acl_add", { context = 42, count = 2, r = { { is_permit = 1, is_ipv6 = 1 }, { is_permit = 0, is_ipv6 = 1 } } })
+print(vpp.dump(reply))
+print("---")
+interface_acl_in = reply[1].acl_index
+
+reply = vpp:api_call("acl_add", { context = 42, count = 3, r = { { is_permit = 1, is_ipv6 = 1 }, { is_permit = 0, is_ipv6 = 1 }, { is_permit = 1, is_ipv6 = 0 } } })
+print(vpp.dump(reply))
+print("---")
+interface_acl_out = reply[1].acl_index
+
+
+reply = vpp:api_call("acl_interface_add_del", { context = 42, sw_if_index = 0, is_add = 1, is_input = 1, acl_index = interface_acl_in })
+print(vpp.dump(reply))
+print("---")
+reply = vpp:api_call("acl_interface_add_del", { context = 42, sw_if_index = 0, is_add = 1, is_input = 1, acl_index = interface_acl_in })
+print(vpp.dump(reply))
+print("---")
+
+reply = vpp:api_call("acl_interface_add_del", { context = 42, sw_if_index = 0, is_add = 1, is_input = 0, acl_index = interface_acl_out })
+print(vpp.dump(reply))
+print("---")
+reply = vpp:api_call("acl_interface_add_del", { context = 42, sw_if_index = 0, is_add = 1, is_input = 0, acl_index = interface_acl_out })
+print(vpp.dump(reply))
+print("---")
+
+reply = vpp:api_call("acl_add", { context = 42, count = 0 })
+print(vpp.dump(reply))
+print("---")
+
+acl_index_to_delete = reply[1].acl_index
+print("Deleting " .. tostring(acl_index_to_delete))
+reply = vpp:api_call("acl_del", { context = 42, acl_index = acl_index_to_delete })
+print(vpp.dump(reply))
+print("---")
+
+reply = vpp:api_call("acl_dump", { context = 42, sw_if_index = 0})
+for ri, rv in ipairs(reply) do 
+  print("Reply message #" .. tostring(ri))
+  print(vpp.dump(rv))
+  for ai, av in ipairs(rv.r) do
+    print("ACL rule #" .. tostring(ai) .. " : " .. vpp.dump(av))
+  end
+   
+end
+print("---")
+
+reply = vpp:api_call("acl_del", { context = 42, acl_index = interface_acl_out })
+print(vpp.dump(reply))
+print("---")
+reply = vpp:api_call("acl_del", { context = 42, acl_index = interface_acl_in })
+print(vpp.dump(reply))
+print("---")
+
+reply = vpp:api_call("acl_dump", { context = 42, sw_if_index = 0})
+print(vpp.dump(reply))
+print("---")
+
+
+vpp:disconnect()
+
+
diff --git a/src/vpp-api/lua/examples/example-classifier.lua b/src/vpp-api/lua/examples/example-classifier.lua
new file mode 100644
index 00000000..ec9c3d3e
--- /dev/null
+++ b/src/vpp-api/lua/examples/example-classifier.lua
@@ -0,0 +1,51 @@
+--[[
+/*
+ * Copyright (c) 2016 Cisco and/or its affiliates.
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at:
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+]]
+
+
+local vpp = require "vpp-lapi"
+local bit = require("bit")
+
+root_dir = "/home/ubuntu/vpp"
+pneum_path = root_dir .. "/build-root/install-vpp_lite_debug-native/vpp-api/lib64/libpneum.so"
+
+
+vpp:init({ pneum_path = pneum_path })
+
+vpp:json_api(root_dir .. "/build-root/install-vpp_lite_debug-native/vpp/vpp-api/vpe.api.json")
+
+vpp:connect("aytest")
+
+-- api calls
+
+print("Calling API to add a new classifier table")
+reply = vpp:api_call("classify_add_del_table", {
+  context = 43,
+  memory_size = bit.lshift(2, 20),
+  client_index = 42,
+  is_add = 1,
+  nbuckets = 32,
+  skip_n_vectors = 0,
+  match_n_vectors = 1,
+  mask = "\255\255\255\255\255\255\255\255" .. "\255\255\255\255\255\255\255\255"
+})
+print(vpp.dump(reply))
+print("---")
+
+
+vpp:disconnect()
+
+
diff --git a/src/vpp-api/lua/examples/example-cli.lua b/src/vpp-api/lua/examples/example-cli.lua
new file mode 100644
index 00000000..8b84989f
--- /dev/null
+++ b/src/vpp-api/lua/examples/example-cli.lua
@@ -0,0 +1,44 @@
+--[[
+/*
+ * Copyright (c) 2016 Cisco and/or its affiliates.
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at:
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+]]
+
+vpp = require "vpp-lapi"
+
+root_dir = "/home/ubuntu/vpp"
+pneum_path = root_dir .. "/build-root/install-vpp_lite_debug-native/vpp-api/lib64/libpneum.so"
+
+vpp:init({ pneum_path = pneum_path })
+
+vpp:json_api(root_dir .. "/build-root/install-vpp_lite_debug-native/vpp/vpp-api/vpe.api.json")
+
+vpp:connect("aytest")
+
+-- api calls
+reply = vpp:api_call("show_version")
+print("Version: ", reply[1].version)
+print(vpp.hex_dump(reply[1].version))
+print(vpp.dump(reply))
+print("---")
+
+
+reply = vpp:api_call("cli_inband", { cmd = "show vers" }) 
+print(vpp.dump(reply))
+print("---")
+
+
+vpp:disconnect()
+
+
diff --git a/src/vpp-api/lua/examples/lute/README.md b/src/vpp-api/lua/examples/lute/README.md
new file mode 100644
index 00000000..8d37250a
--- /dev/null
+++ b/src/vpp-api/lua/examples/lute/README.md
@@ -0,0 +1,66 @@
+LUTE: Lua Unit Test Environment
+
+This is a small helper utility to automate some simple tests
+that one might need to do.
+
+Think of it as a hybrid of a screen and expect who
+also took some habits from HTML inline code.
+
+It is quite probably useless for building anything serious,
+but practice shows it is quite efficient at allowing
+convenient temporary quick tests, and for something
+that was written over a course of a couple of evenings it
+is quite a nice little helper tool.
+
+It allows do launch and drive multiple shell sessions,
+and by virtue of having been written in Lua, it of course
+also allows to add the business logic using the Lua code.
+
+If you launch the lute without parameters, it gives you
+the interactive shell to execute the commands in.
+
+If you launch it with an argument, it will attempt to
+read and execute the commands from the file.
+
+Commands:
+
+shell FOO
+
+  spawn a shell in a new PTY under the label FOO.
+
+run FOO bar
+
+  Send "bar" keystrokes followed by "ENTER" to the session FOO
+
+  Special case: "break" word on its own gets translated into ^C being sent.
+
+cd FOO
+
+  "change domain" into session FOO. All subsequent inputs will go,
+  line-buffered, into the session FOO. To jump back up, use ^D (Control-D),
+  or within the file, use ^D^D^D (caret D caret D caret D on its own line)
+
+expect FOO blablabla
+
+  Pause further interpretation of the batch mode until you see "blablabla"
+  in the output of session FOO, or until timeout happens.
+
+sleep N
+
+  Sleep an integer N seconds, if you are in batch mode.
+
+echo blabla
+
+  Echo the remainder of the line to standard output.
+
+For Lua code, there is a pre-existing pseudo-session called "lua",
+which accepts "run lua" command which does what you would expect
+(evaluate the rest of the string in Lua context - being the same
+as lute itself). Also you can do "cd lua" and get into a
+multiline-enabled interpreter shell.
+
+This way for the VPP case you can automate some of the things in your routine
+that you would have to have done manually, and test drive API as well
+as use the realistic native OS components to create the environment around it.
+
+
diff --git a/src/vpp-api/lua/examples/lute/lute.lua b/src/vpp-api/lua/examples/lute/lute.lua
new file mode 100644
index 00000000..89b9924b
--- /dev/null
+++ b/src/vpp-api/lua/examples/lute/lute.lua
@@ -0,0 +1,777 @@
+--[[
+version = 1
+/*
+ * Copyright (c) 2016 Cisco and/or its affiliates.
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at:
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+]]
+
+-- LUTE: Lua Unit Test Environment
+-- AKA what happens when screen tries to marry with lua and expect,
+-- but escapes mid-ceremony.
+--
+-- comments: @ayourtch
+
+ffi = require("ffi")
+
+vpp = {}
+function vpp.dump(o)
+   if type(o) == 'table' then
+      local s = '{ '
+      for k,v in pairs(o) do
+         if type(k) ~= 'number' then k = '"'..k..'"' end
+         s = s .. '['..k..'] = ' .. vpp.dump(v) .. ','
+      end
+      return s .. '} '
+   else
+      return tostring(o)
+   end
+end
+
+
+ffi.cdef([[
+
+int posix_openpt(int flags);
+int grantpt(int fd);
+int unlockpt(int fd);
+char *ptsname(int fd);
+
+typedef long pid_t;
+typedef long ssize_t;
+typedef long size_t;
+typedef int nfds_t;
+typedef long time_t;
+typedef long suseconds_t;
+
+pid_t fork(void);
+pid_t setsid(void);
+
+int close(int fd);
+int open(char *pathname, int flags);
+
+int dup2(int oldfd, int newfd);
+
+ssize_t read(int fd, void *buf, size_t count);
+ssize_t write(int fd, const void *buf, size_t count);
+
+struct pollfd {
+               int   fd;         /* file descriptor */
+               short events;     /* requested events */
+               short revents;    /* returned events */
+           };
+
+int poll(struct pollfd *fds, nfds_t nfds, int timeout);
+
+struct timeval {
+               time_t      tv_sec;     /* seconds */
+               suseconds_t tv_usec;    /* microseconds */
+           };
+
+int gettimeofday(struct timeval *tv, struct timezone *tz);
+
+int inet_pton(int af, const char *src, void *dst);
+
+]])
+
+ffi.cdef([[
+void *memset(void *s, int c, size_t n);
+void *memcpy(void *dest, void *src, size_t n);
+void *memmove(void *dest, const void *src, size_t n);
+void *memmem(const void *haystack, size_t haystacklen,
+        const void *needle, size_t needlelen);
+]])
+
+
+
+local O_RDWR = 2
+
+
+function os_time()
+  local tv = ffi.new("struct timeval[1]")
+  local ret = ffi.C.gettimeofday(tv, nil)
+  return tonumber(tv[0].tv_sec) + (tonumber(tv[0].tv_usec)/1000000.0)
+end
+
+function sleep(n)
+  local when_wakeup = os_time() + n
+  while os_time() <= when_wakeup do
+    ffi.C.poll(nil, 0, 10)
+  end
+end
+
+
+function c_str(text_in)
+  local text = text_in 
+  local c_str = ffi.new("char[?]", #text+1)
+  ffi.copy(c_str, text)
+  return c_str
+end
+
+function ip46(addr_text)
+  local out = ffi.new("char [200]")
+  local AF_INET6 = 10
+  local AF_INET = 2
+  local is_ip6 = ffi.C.inet_pton(AF_INET6, c_str(addr_text), out)
+  if is_ip6 == 1 then
+    return ffi.string(out, 16), true
+  end
+  local is_ip4 = ffi.C.inet_pton(AF_INET, c_str(addr_text), out)
+  if is_ip4 then
+    return (string.rep("4", 12).. ffi.string(out, 4)), false
+  end
+end
+
+function pty_master_open()
+  local fd = ffi.C.posix_openpt(O_RDWR)
+  ffi.C.grantpt(fd)
+  ffi.C.unlockpt(fd)
+  local p = ffi.C.ptsname(fd)
+  print("PTS:" .. ffi.string(p))
+  return fd, ffi.string(p)
+end
+
+function pty_run(cmd)
+  local master_fd, pts_name = pty_master_open()
+  local child_pid = ffi.C.fork()
+  if (child_pid == -1) then
+    print("Error fork()ing")
+    return -1
+  end 
+ 
+  if child_pid ~= 0 then
+    -- print("Parent")
+    return master_fd, child_pid
+  end
+
+  -- print("Child")
+  if (ffi.C.setsid() == -1) then
+    print("Child error setsid")
+    os.exit(-1)
+  end
+
+  ffi.C.close(master_fd)
+
+  local slave_fd = ffi.C.open(c_str(pts_name), O_RDWR)
+  if slave_fd == -1 then
+    print("Child can not open slave fd")
+    os.exit(-2)
+  end
+
+  ffi.C.dup2(slave_fd, 0)
+  ffi.C.dup2(slave_fd, 1)
+  ffi.C.dup2(slave_fd, 2)
+  os.execute(cmd)
+end
+
+function readch()
+  local buf = ffi.new("char[1]")
+  local nread= ffi.C.read(0, buf, 1)
+  -- print("\nREADCH : " .. string.char(buf[0]))
+  return string.char(buf[0])
+end
+
+function stdout_write(str)
+  ffi.C.write(1, c_str(str), #str)
+end
+
+
+readln = {
+split = function(str, pat)
+  local t = {}  -- NOTE: use {n = 0} in Lua-5.0
+  local fpat = "(.-)" .. pat
+  local last_end = 1
+  if str then
+    local s, e, cap = str:find(fpat, 1)
+    while s do
+      if s ~= 1 or cap ~= "" then
+        table.insert(t,cap)
+      end
+      last_end = e+1
+      s, e, cap = str:find(fpat, last_end)
+    end
+    if last_end <= #str then
+      cap = str:sub(last_end)
+      table.insert(t, cap)
+    end
+  end
+  return t
+end,
+
+reader = function()
+  local rl = {}
+
+  rl.init = function()
+    os.execute("stty -icanon min 1 -echo")
+    rl.rawmode = true
+  end
+
+  rl.done = function()
+    os.execute("stty icanon echo")
+    rl.rawmode = false
+  end
+
+  rl.prompt = ">"
+  rl.history = { "" }
+  rl.history_index = 1
+  rl.history_length = 1
+
+  rl.hide_cmd = function()
+    local bs = string.char(8) .. " " .. string.char(8)
+    for i = 1, #rl.command do
+      stdout_write(bs)
+    end
+  end
+
+  rl.show_cmd = function()
+    if rl.command then
+      stdout_write(rl.command)
+    end
+  end
+
+  rl.store_history = function(cmd)
+    if cmd == "" then
+      return
+    end
+    rl.history[rl.history_length] = cmd
+    rl.history_length = rl.history_length + 1
+    rl.history_index = rl.history_length
+    rl.history[rl.history_length] = ""
+  end
+
+  rl.readln = function(stdin_select_fn, batch_cmd, batch_when, batch_expect)
+    local done = false
+    local need_prompt = true
+    rl.command = ""
+
+    if not rl.rawmode then
+      rl.init()
+    end
+
+    while not done do
+      local indent_value = #rl.prompt + #rl.command
+      if need_prompt then
+        stdout_write(rl.prompt)
+        stdout_write(rl.command)
+        need_prompt = false
+      end
+      if type(stdin_select_fn) == "function" then
+        while not stdin_select_fn(indent_value, batch_cmd, batch_when, batch_expect) do
+          stdout_write(rl.prompt)
+          stdout_write(rl.command)
+          indent_value = #rl.prompt + #rl.command
+        end
+        if batch_cmd and ((os_time() > batch_when) or (batch_expect and expect_success(batch_expect, buf, 0))) then
+          stdout_write("\n" .. rl.prompt .. batch_cmd .. "\n")
+          if batch_expect then
+            expect_done(batch_expect)
+          end
+          return batch_cmd, batch_expect
+        end
+      end
+      local ch = readch()
+      if ch:byte(1) == 27 then
+        -- CONTROL
+        local ch2 = readch()
+        -- arrows
+        if ch2:byte(1) == 91 then
+          local ch3 = readch()
+          local b = ch3:byte(1)
+          if b == 65 then
+            ch = "UP"
+          elseif b == 66 then
+            ch = "DOWN"
+          elseif b == 67 then
+            ch = "RIGHT"
+          elseif b == 68 then
+            ch = "LEFT"
+          end
+          -- print("Byte: " .. ch3:byte(1))
+          -- if ch3:byte(1)
+        end
+      end
+
+      if ch == "?" then
+        stdout_write(ch)
+        stdout_write("\n")
+        if rl.help then
+          rl.help(rl)
+        end
+        need_prompt = true
+      elseif ch == "\t" then
+        if rl.tab_complete then
+          rl.tab_complete(rl)
+        end
+        stdout_write("\n")
+        need_prompt = true
+      elseif ch == "\n" then
+        stdout_write(ch)
+        done = true
+      elseif ch == "\004" then
+        stdout_write("\n")
+        rl.command = nil
+        done = true
+      elseif ch == string.char(127) then
+        if rl.command ~= "" then
+          stdout_write(string.char(8) .. " " .. string.char(8))
+          rl.command = string.sub(rl.command, 1, -2)
+        end
+      elseif #ch > 1 then
+        -- control char
+        if ch == "UP" then
+          rl.hide_cmd()
+          if rl.history_index == #rl.history then
+            rl.history[rl.history_index] = rl.command
+          end
+          if rl.history_index > 1 then
+            rl.history_index = rl.history_index - 1
+            rl.command = rl.history[rl.history_index]
+          end
+          rl.show_cmd()
+        elseif ch == "DOWN" then
+          rl.hide_cmd()
+          if rl.history_index < rl.history_length then
+            rl.history_index = rl.history_index + 1
+            rl.command = rl.history[rl.history_index]
+          end
+          rl.show_cmd()
+        end
+      else
+        stdout_write(ch)
+        rl.command = rl.command .. ch
+      end
+    end
+    if rl.command then
+      rl.store_history(rl.command)
+    end
+    return rl.command
+  end
+  return rl
+end
+
+}
+
+local select_fds = {}
+local sessions = {}
+
+local line_erased = false
+
+function erase_line(indent)
+  if not line_erased then
+    line_erased = true
+    stdout_write(string.rep(string.char(8), indent)..string.rep(" ", indent)..string.rep(string.char(8), indent))
+  end
+end
+
+function do_select_stdin(indent, batch_cmd, batch_when, batch_expect)
+  while true do
+    local nfds = 1+#select_fds
+    local pfds = ffi.new("struct pollfd[?]", nfds)
+    pfds[0].fd = 0;
+    pfds[0].events = 1;
+    pfds[0].revents = 0;
+    for i = 1,#select_fds do
+      pfds[i].fd = select_fds[i].fd
+      pfds[i].events = 1
+      pfds[i].revents = 0
+    end
+    if batch_cmd and ((os_time() > batch_when) or (batch_expect and expect_success(batch_expect, buf, 0))) then
+      return true
+    end
+    while ffi.C.poll(pfds, nfds, 10) == 0 do
+      if batch_cmd and ((os_time() > batch_when) or (batch_expect and expect_success(batch_expect, buf, 0))) then
+        return true
+      end
+      if line_erased then
+        line_erased = false
+        return false
+      end
+    end
+    if pfds[0].revents == 1 then
+      return true
+    end
+    for i = 1,#select_fds do
+      if(pfds[i].revents > 0) then
+        if pfds[i].fd ~= select_fds[i].fd then
+          print("File descriptors unequal", pfds[i].fd, select_fds[i].fd)
+        end
+        select_fds[i].cb(select_fds[i], pfds[i].revents, indent)
+      end
+    end
+  end
+end
+
+local buf = ffi.new("char [32768]")
+
+function session_stdout_write(prefix, data)
+  data = prefix .. data:gsub("\n", "\n"..prefix):gsub("\n"..prefix.."$", "\n")
+  
+  stdout_write(data)
+end
+
+function expect_success(sok, buf, nread)
+  local expect_buf_sz = ffi.sizeof(sok.expect_buf) - 128
+  local expect_buf_avail = expect_buf_sz - sok.expect_buf_idx
+  -- print("EXPECT_SUCCESS: nread ".. tostring(nread).. " expect_buf_idx: " .. tostring(sok.expect_buf_idx) .. " expect_buf_avail: " .. tostring(expect_buf_avail) )
+  if expect_buf_avail < 0 then
+    print "EXPECT BUFFER OVERRUN ALREADY"
+    os.exit(1)
+  end
+  if expect_buf_avail < nread then
+    if (nread >= ffi.sizeof(sok.expect_buf)) then
+      print("Read too large of a chunk to fit into expect buffer")
+      return nil
+    end
+    local delta = nread - expect_buf_avail
+
+    ffi.C.memmove(sok.expect_buf, sok.expect_buf + delta, expect_buf_sz - delta)
+    sok.expect_buf_idx = sok.expect_buf_idx - delta
+    expect_buf_avail = nread 
+  end
+  if sok.expect_buf_idx + nread > expect_buf_sz then
+    print("ERROR, I have just overrun the buffer !")
+    os.exit(1)
+  end
+  ffi.C.memcpy(sok.expect_buf + sok.expect_buf_idx, buf, nread)
+  sok.expect_buf_idx = sok.expect_buf_idx + nread
+  if sok.expect_str == nil then
+    return true
+  end
+  local match_p = ffi.C.memmem(sok.expect_buf, sok.expect_buf_idx, sok.expect_str, sok.expect_str_len)
+  if match_p ~= nil then
+    return true
+  end
+  return false
+end
+
+function expect_done(sok)
+  local expect_buf_sz = ffi.sizeof(sok.expect_buf) - 128
+  if not sok.expect_str then 
+    return false
+  end
+  local match_p = ffi.C.memmem(sok.expect_buf, sok.expect_buf_idx, sok.expect_str, sok.expect_str_len)
+  if match_p ~= nil then
+    if sok.expect_cb then
+      sok.expect_cb(sok)
+    end
+    local match_idx = ffi.cast("char *", match_p) - ffi.cast("char *", sok.expect_buf)
+    ffi.C.memmove(sok.expect_buf, ffi.cast("char *", match_p) + sok.expect_str_len, expect_buf_sz - match_idx - sok.expect_str_len)
+    sok.expect_buf_idx = match_idx + sok.expect_str_len
+    sok.expect_success = true
+
+    sok.expect_str = nil
+    sok.expect_str_len = 0
+    return true
+  end
+end
+
+function slave_events(sok, revents, indent)
+  local fd = sok.fd
+  local nread = ffi.C.read(fd, buf, ffi.sizeof(buf)-128)
+  local idx = nread - 1
+  while idx >= 0 and buf[idx] ~= 10 do
+    idx = idx - 1
+  end
+  if idx >= 0 then
+    erase_line(indent)
+    session_stdout_write(sok.prefix, sok.buf .. ffi.string(buf, idx+1))
+    sok.buf = ""
+  end
+  sok.buf = sok.buf .. ffi.string(buf+idx+1, nread-idx-1)
+  -- print("\nRead: " .. tostring(nread))
+  -- stdout_write(ffi.string(buf, nread))
+  if expect_success(sok, buf, nread) then
+    return true
+  end
+  return false
+end
+
+
+function start_session(name)
+  local mfd, cpid = pty_run("/bin/bash")
+  local sok =  { ["fd"] = mfd, ["cb"] = slave_events, ["buf"] = "", ["prefix"] = name .. ":", ["expect_buf"] = ffi.new("char [165536]"), ["expect_buf_idx"] = 0, ["expect_str"] = nil }
+  table.insert(select_fds, sok)
+  sessions[name] = sok
+end
+
+function command_transform(exe)
+  if exe == "break" then
+    exe = string.char(3)
+  end
+  return exe
+end
+
+function session_write(a_session, a_str)
+  if has_session(a_session) then
+    return tonumber(ffi.C.write(sessions[a_session].fd, c_str(a_str), #a_str))
+  else
+    return 0
+  end
+end
+
+function session_exec(a_session, a_cmd)
+  local exe = command_transform(a_cmd) .. "\n"
+  session_write(a_session, exe)
+end
+
+function session_cmd(ui, a_session, a_cmd)
+  if not has_session(a_session) then
+    stdout_write("ERR: No such session '" .. tostring(a_session) .. "'\n")
+    return nil
+  end
+  if a_session == "lua" then
+    local func, msg = loadstring(ui.lua_acc .. a_cmd)
+    -- stdout_write("LOADSTR: " .. vpp.dump({ ret, msg }) .. "\n")
+    if not func and string.match(msg, "") then
+      if a_session ~= ui.in_session then
+         stdout_write("ERR LOADSTR: " .. tostring(msg) .. "\n")
+         return nil
+      end
+      ui.lua_acc = ui.lua_acc .. a_cmd  .. "\n"
+      return true
+    end
+    ui.lua_acc = ""
+    local ret, msg = pcall(func)
+    if ret then
+      return true
+    else
+      stdout_write("ERR: " .. msg .. "\n") 
+      return nil
+    end
+  else
+    session_exec(a_session, a_cmd)
+    if ui.session_cmd_delay then
+      return { "delay", ui.session_cmd_delay }
+    end
+    return true
+  end
+end
+
+function has_session(a_session)
+  if a_session == "lua" then
+    return true
+  end
+  return (sessions[a_session] ~= nil)
+end
+
+function command_match(list, input, output)
+  for i, v in ipairs(list) do
+    local m = {}
+    m[1], m[2], m[3], m[4], m[5], m[6], m[7], m[8], m[9] = string.match(input, v[1])
+    -- print("MATCH: ", vpp.dump(m))
+    if m[1] then
+       output["result"] = m
+       output["result_index"] = i
+       return m
+    end 
+  end
+  return nil
+end
+
+function cmd_spawn_shell(ui, a_arg)
+  start_session(a_arg[1])
+  return true
+end
+
+function cmd_run_cmd(ui, a_arg)
+  local a_sess = a_arg[1]
+  local a_cmd = a_arg[2]
+  return session_cmd(ui, a_sess, a_cmd)
+end
+
+function cmd_cd(ui, a_arg)
+  local a_sess = a_arg[1]
+  if has_session(a_sess) then
+    ui.in_session = a_sess
+    return true
+  else
+    stdout_write("ERR: Unknown session '".. tostring(a_sess) .. "'\n")
+    return nil
+  end
+end
+
+function cmd_sleep(ui, a_arg)
+  return { "delay", tonumber(a_arg[1]) }
+end
+
+function cmd_expect(ui, a_arg)
+  local a_sess = a_arg[1]
+  local a_expect = a_arg[2]
+  local sok = sessions[a_sess]
+  if not sok then
+    stdout_write("ERR: unknown session '" .. tostring(a_sess) .. "'\n")
+    return nil
+  end
+  sok.expect_str = c_str(a_expect)
+  sok.expect_str_len = #a_expect
+  return { "expect", a_sess }
+end
+
+function cmd_info(ui, a_arg)
+  local a_sess = a_arg[1]
+  local sok = sessions[a_sess]
+  if not sok then
+    stdout_write("ERR: unknown session '" .. tostring(a_sess) .. "'\n")
+    return nil
+  end
+  print("Info for session " .. tostring(a_sess) .. "\n")
+  print("Expect buffer index: " .. tostring(sok.expect_buf_idx))
+  print("Expect buffer: '" .. tostring(ffi.string(sok.expect_buf, sok.expect_buf_idx)) .. "'\n")
+  if sok.expect_str then
+    print("Expect string: '" .. tostring(ffi.string(sok.expect_str, sok.expect_str_len)) .. "'\n")
+  else
+    print("Expect string not set\n")
+  end
+end
+
+function cmd_echo(ui, a_arg)
+  local a_data = a_arg[1]
+  print("ECHO: " .. tostring(a_data))
+end
+
+main_command_table = {
+  { "^shell ([a-zA-Z0-9_]+)$", cmd_spawn_shell },
+  { "^run ([a-zA-Z0-9_]+) (.+)$", cmd_run_cmd },
+  { "^cd ([a-zA-Z0-9_]+)$", cmd_cd  },
+  { "^sleep ([0-9]+)$", cmd_sleep },
+  { "^expect ([a-zA-Z0-9_]+) (.-)$", cmd_expect },
+  { "^info ([a-zA-Z0-9_]+)$", cmd_info },
+  { "^echo (.-)$", cmd_echo }
+}
+
+
+
+function ui_set_prompt(ui)
+  if ui.in_session then 
+    if ui.in_session == "lua" then
+      if #ui.lua_acc > 0 then
+        ui.r.prompt = ui.in_session .. ">>"
+      else
+        ui.r.prompt = ui.in_session .. ">"
+      end 
+    else
+      ui.r.prompt = ui.in_session .. "> "
+    end
+  else
+    ui.r.prompt = "> "
+  end
+  return ui.r.prompt
+end
+
+function ui_run_command(ui, cmd)
+  -- stdout_write("Command: " .. tostring(cmd) .. "\n")
+  local ret = false
+  if ui.in_session then
+    if cmd then
+      if cmd == "^D^D^D" then
+        ui.in_session = nil
+        ret = true
+      else
+        ret = session_cmd(ui, ui.in_session, cmd)
+      end
+    else
+      ui.in_session = nil
+      ret = true
+    end
+  else  
+    if cmd then
+      local out = {}
+      if cmd == "" then
+        ret = true
+      end
+      if command_match(main_command_table, cmd, out) then
+        local i = out.result_index
+        local m = out.result
+        if main_command_table[i][2] then
+          ret = main_command_table[i][2](ui, m)
+        end
+      end
+    end
+    if not cmd or cmd == "quit" then
+      return "quit"
+    end
+  end
+  return ret
+end
+
+local ui = {}
+ui.in_session = nil
+ui.r = readln.reader() 
+ui.lua_acc = ""
+ui.session_cmd_delay = 0.3
+
+local lines = ""
+
+local done = false
+-- a helper function which always returns nil
+local no_next_line = function() return nil end
+
+-- a function which returns the next batch line
+local next_line = no_next_line
+
+local batchfile = arg[1]
+
+if batchfile then
+  local f = io.lines(batchfile)
+  next_line = function() 
+    local line = f()
+    if line then 
+      return line
+    else 
+      next_line = no_next_line
+      session_stdout_write(batchfile .. ":", "End of batch\n")
+      return nil
+    end
+  end
+end
+
+
+local batch_when = 0
+local batch_expect = nil
+while not done do
+  local prompt = ui_set_prompt(ui)
+  local batch_cmd = next_line() 
+  local cmd, expect_sok = ui.r.readln(do_select_stdin, batch_cmd, batch_when, batch_expect)
+  if expect_sok and not expect_success(expect_sok, buf, 0) then
+    if not cmd_ret and next_line ~= no_next_line then
+      print("ERR: expect timeout\n")
+      next_line = no_next_line
+    end
+  else 
+    local cmd_ret = ui_run_command(ui, cmd)
+    if not cmd_ret and next_line ~= no_next_line then
+      print("ERR: Error during batch execution\n")
+      next_line = no_next_line
+    end
+
+    if cmd_ret  == "quit" then
+      done = true
+    end
+    batch_expect = nil
+    batch_when = 0
+    if type(cmd_ret) == "table" then
+      if cmd_ret[1] == "delay" then
+	batch_when = os_time() + tonumber(cmd_ret[2])
+      end
+      if cmd_ret[1] == "expect" then
+	batch_expect = sessions[cmd_ret[2]]
+	batch_when = os_time() + 15
+      end
+    end
+  end
+end
+ui.r.done()
+
+os.exit(1)
+
+
+
diff --git a/src/vpp-api/lua/examples/lute/script-inout-acl-noacl.lute b/src/vpp-api/lua/examples/lute/script-inout-acl-noacl.lute
new file mode 100644
index 00000000..a24d04bf
--- /dev/null
+++ b/src/vpp-api/lua/examples/lute/script-inout-acl-noacl.lute
@@ -0,0 +1,329 @@
+shell vppbuild
+run vppbuild stty -echo
+run vppbuild sudo -u ubuntu -i bash -c "(cd vpp && make plugins && echo ALLGOOD)"
+expect vppbuild ALLGOOD
+
+shell s0
+shell s1
+shell s2
+
+
+cd s1
+unshare -n /bin/bash
+/sbin/ifconfig -a
+^D^D^D
+
+cd s2
+unshare -n /bin/bash
+/sbin/ifconfig -a
+^D^D^D
+
+
+cd lua
+
+function session_get_bash_pid(s)
+  if not has_session(s) then
+    return nil
+  end
+  local fname = "/tmp/lute-"..s.."-pid.txt"
+
+  session_exec(s, "echo $$ >" .. fname)
+  -- it's a dirty hack but it's quick
+  sleep(0.5)
+  local pid = io.lines(fname)()
+  print("Got pid for " .. s .. " : " .. tostring(pid))
+  return(tonumber(pid))
+end
+
+function session_connect_with(s0, s1)
+  -- local pid0 = tostring(session_get_bash_pid(s0))
+  local pid1 = tostring(session_get_bash_pid(s1))
+  local eth_options = { "rx", "tx",  "sg", "tso", "ufo", "gso", "gro", "lro", "rxvlan", "txvlan", "rxhash" }
+  local this_end = s0 .. "_" .. s1
+  local other_end = s1 .. "_" .. s0
+  session_exec(s0, "ip link add name " .. this_end .. " type veth peer name " .. other_end)
+  session_exec(s0, "ip link set dev " .. this_end .. " up promisc on")
+  for i, option in ipairs(eth_options) do
+    session_exec(s0, "/sbin/ethtool --offload " .. this_end .. " " .. option .. " off")
+    session_exec(s0, "/sbin/ethtool --offload " .. other_end .. " " .. option .. " off")
+  end
+  session_exec(s0, "ip link set dev " .. other_end .. " up promisc on netns /proc/" .. pid1 .. "/ns/net")
+  sleep(0.5)
+end
+
+^D^D^D
+run lua session_connect_with("s0", "s1")
+run lua session_connect_with("s0", "s2")
+
+cd s1
+ip -6 addr add dev s1_s0 2001:db8:1::1/64
+ip -4 addr add dev s1_s0 192.0.2.1/24
+ip link set dev s1_s0 up promisc on
+^D^D^D
+
+cd s2
+ip -6 addr add dev s2_s0 2001:db8:1::2/64
+ip -6 addr add dev s2_s0 2001:db8:1::3/64
+ip -6 addr add dev s2_s0 2001:db8:1::4/64
+ip -4 addr add dev s2_s0 192.0.2.2/24
+ip -4 addr add dev s2_s0:1 192.0.2.3/24
+ip -4 addr add dev s2_s0:2 192.0.2.4/24
+ip link set dev s2_s0 up promisc on
+^D^D^D
+
+run s1 ip addr
+run s2 ip addr
+shell VPP
+cd VPP
+cd /home/ubuntu/vpp
+make debug 
+r
+^D^D^D
+expect VPP DBGvpp#
+
+cd lua
+-- Initialization of the Lua environment for talking to VPP
+vpp = require("vpp-lapi")
+root_dir = "/home/ubuntu/vpp"
+pneum_path = root_dir .. "/build-root/install-vpp_debug-native/vpp-api/lib64/libpneum.so"
+vpp:init({ pneum_path = pneum_path })
+vpp:consume_api(root_dir .. "/build-root/install-vpp_debug-native/vlib-api/vlibmemory/memclnt.api")
+vpp:consume_api(root_dir .. "/build-root/install-vpp_debug-native/vpp/vpp-api/vpe.api")
+vpp:connect("aytest")
+vpp:consume_api(root_dir .. "/plugins/acl-plugin/acl/acl.api", "acl")
+
+^D^D^D
+
+cd lua
+
+reply = vpp:api_call("af_packet_create", { host_if_name = "s0_s1", hw_addr = "AAAAAA" })
+vpp_if_to_s1 = reply[1].sw_if_index
+
+reply = vpp:api_call("af_packet_create", { host_if_name = "s0_s2", hw_addr = "AAAAAA" })
+vpp_if_to_s2 = reply[1].sw_if_index
+
+ifaces = { vpp_if_to_s1, vpp_if_to_s2 }
+
+reply = vpp:api_call("sw_interface_set_flags", { sw_if_index = vpp_if_to_s1, admin_up_down = 1, link_up_down = 1 })
+print(vpp.dump(reply))
+reply = vpp:api_call("sw_interface_set_flags", { sw_if_index = vpp_if_to_s2, admin_up_down = 1, link_up_down = 1 })
+print(vpp.dump(reply))
+
+bd_id = 42
+
+reply = vpp:api_call("bridge_domain_add_del", { bd_id = bd_id, flood = 1, uu_flood = 1, forward = 1, learn = 1, arp_term = 0, is_add = 1 })
+print(vpp.dump(reply))
+
+for i, v in ipairs(ifaces) do
+  reply = vpp:api_call("sw_interface_set_l2_bridge", { rx_sw_if_index = v, bd_id = bd_id, shg = 0, bvi = 0, enable = 1 } )
+  print(vpp.dump(reply))
+end
+
+^D^D^D
+
+run s1 ping -c 3 192.0.2.2
+expect s1 packet loss
+run s1 ping -c 3 192.0.2.3
+expect s1 packet loss
+run s1 ping -c 3 192.0.2.4
+expect s1 packet loss
+run s1 ping6 -c 3 2001:db8:1::2
+expect s1 packet loss
+run s1 ping6 -c 3 2001:db8:1::3
+expect s1 packet loss
+run s1 ping6 -c 3 2001:db8:1::4
+expect s1 packet loss
+
+
+cd lua
+--- ACL testing
+
+--[[ temporary comment out
+
+reply = vpp:api_call("acl_del", { context = 42, acl_index = 230 })
+print(vpp.dump(reply))
+print("---")
+
+reply = vpp:api_call("acl_del", { context = 42, acl_index = 8 })
+print(vpp.dump(reply))
+print("---")
+
+reply = vpp:api_call("acl_del", { context = 42, acl_index = 15 })
+print(vpp.dump(reply))
+print("---")
+
+reply = vpp:api_call("acl_add_replace", { context = 42, count = 2, r = { { is_permit = 1, is_ipv6 = 1 }, { is_permit = 0, is_ipv6 = 1 } } })
+print(vpp.dump(reply))
+print("---")
+interface_acl_in = reply[1].acl_index
+
+reply = vpp:api_call("acl_add_replace", { context = 42, count = 3, r = { { is_permit = 1, is_ipv6 = 1 }, { is_permit = 0, is_ipv6 = 1 }, { is_permit = 1, is_ipv6 = 0 } } })
+print(vpp.dump(reply))
+print("---")
+interface_acl_out = reply[1].acl_index
+
+
+reply = vpp:api_call("acl_interface_add_del", { context = 42, sw_if_index = 0, is_add = 1, is_input = 1, acl_index = interface_acl_in })
+print(vpp.dump(reply))
+print("---")
+
+reply = vpp:api_call("acl_interface_add_del", { context = 42, sw_if_index = 0, is_add = 1, is_input = 1, acl_index = interface_acl_in })
+print(vpp.dump(reply))
+print("---")
+
+reply = vpp:api_call("acl_interface_add_del", { context = 42, sw_if_index = 0, is_add = 1, is_input = 0, acl_index = interface_acl_out })
+print(vpp.dump(reply))
+print("---")
+
+reply = vpp:api_call("acl_interface_add_del", { context = 42, sw_if_index = 0, is_add = 1, is_input = 0, acl_index = interface_acl_out })
+print(vpp.dump(reply))
+print("---")
+
+reply = vpp:api_call("acl_add_replace", { context = 42, count = 0 })
+print(vpp.dump(reply))
+print("---")
+
+acl_index_to_delete = reply[1].acl_index
+print("Deleting " .. tostring(acl_index_to_delete))
+reply = vpp:api_call("acl_del", { context = 42, acl_index = acl_index_to_delete })
+print(vpp.dump(reply))
+print("---")
+
+reply = vpp:api_call("acl_dump", { context = 42, sw_if_index = 0})
+for ri, rv in ipairs(reply) do
+  print("Reply message #" .. tostring(ri))
+  print(vpp.dump(rv))
+  for ai, av in ipairs(rv.r) do
+    print("ACL rule #" .. tostring(ai) .. " : " .. vpp.dump(av))
+  end
+
+end
+print("---")
+
+reply = vpp:api_call("acl_del", { context = 42, acl_index = interface_acl_out })
+print(vpp.dump(reply))
+print("---")
+reply = vpp:api_call("acl_del", { context = 42, acl_index = interface_acl_in })
+print(vpp.dump(reply))
+print("---")
+
+reply = vpp:api_call("acl_dump", { context = 42, sw_if_index = 0})
+print(vpp.dump(reply))
+print("---")
+
+reply = vpp:api_call("acl_dump", { context = 42, sw_if_index = 4294967295 })
+print(vpp.dump(reply))
+print("---")
+
+
+]] -- end of comment out
+
+---- Should be nothing ^^
+r = {
+  { is_permit = 1, is_ipv6 = 1, dst_ip_addr = ip46("2001:db8:1::2"), dst_ip_prefix_len = 128 },
+  { is_permit = 0, is_ipv6 = 1, dst_ip_addr = ip46("2001:db8:1::3"), dst_ip_prefix_len = 128 },
+  { is_permit = 1, is_ipv6 = 1, dst_ip_addr = ip46("2001:db8::"), dst_ip_prefix_len = 32  },
+  { is_permit = 1, is_ipv6 = 0, dst_ip_addr = ip46("192.0.2.2"), dst_ip_prefix_len = 32},
+  { is_permit = 0, is_ipv6 = 0, dst_ip_addr = ip46("192.0.2.3"), dst_ip_prefix_len = 32 },
+}
+
+reply = vpp:api_call("acl_add_replace", { context = 42, count = 5, r = r })
+print(vpp.dump(reply))
+print("---")
+interface_acl_in = reply[1].acl_index
+
+reply = vpp:api_call("acl_add_replace", { context = 42, count = 3, r = { { is_permit = 1, is_ipv6 = 1 }, { is_permit = 0, is_ipv6 = 1 }, { is_permit = 1, is_ipv6 = 0 } } })
+print(vpp.dump(reply))
+print("---")
+interface_acl_out = reply[1].acl_in
+
+reply = vpp:api_call("acl_interface_add_del", { context = 42, sw_if_index = vpp_if_to_s1, is_add = 1, is_input = 1, acl_index = interface_acl_in })
+print(vpp.dump(reply))
+print("---")
+
+--reply = vpp:api_call("acl_interface_add_del", { context = 42, sw_if_index = vpp_if_to_s2, is_add = 1, is_input = 0, acl_index = interface_acl_out })
+-- print(vpp.dump(reply))
+--print("---")
+
+^D^D^D
+
+run VPP clear trace
+run VPP trace add af-packet-input 100
+run s1 ping6 -c 3 2001:db8:1::2
+expect s1 packet loss
+run VPP show trace
+expect VPP match: inacl 0 rule 0
+
+run VPP clear trace
+run VPP trace add af-packet-input 100
+run s1 ping6 -c 3 2001:db8:1::3
+expect s1 packet loss
+run VPP show trace 
+expect VPP match: inacl 0 rule 1
+
+run VPP clear trace
+run VPP trace add af-packet-input 100
+run s1 ping6 -c 3 2001:db8:1::4
+expect s1 packet loss
+run VPP show trace
+expect VPP match: inacl 0 rule 2
+
+run VPP clear trace
+run VPP trace add af-packet-input 100
+run s1 ping -c 3 192.0.2.2
+expect s1 packet loss
+run VPP show trace
+expect VPP match: inacl 0 rule 3
+
+run VPP clear trace
+run VPP trace add af-packet-input 100
+run s1 ping -c 3 192.0.2.3
+expect s1 packet loss
+run VPP show trace
+expect VPP match: inacl 0 rule 4
+
+
+cd lua
+
+--- TEST OUTBOUND ACL
+
+r1 = {
+  { is_permit = 1, is_ipv6 = 1, src_ip_addr = ip46("2001:db8:1::1"), src_ip_prefix_len = 128, dst_ip_addr = ip46("2001:db8:1::2"), dst_ip_prefix_len = 128 },
+  { is_permit = 0, is_ipv6 = 1, src_ip_addr = ip46("2001:db8:1::1"), src_ip_prefix_len = 128, dst_ip_addr = ip46("2001:db8:1::4"), dst_ip_prefix_len = 128 }
+}
+
+reply = vpp:api_call("acl_add_replace", { context = 42, count = 3, r = r1 })
+print(vpp.dump(reply))
+print("---")
+interface_acl_out = reply[1].acl_index
+
+reply = vpp:api_call("acl_interface_add_del", { context = 42, sw_if_index = vpp_if_to_s2, is_add = 1, is_input = 0, acl_index = interface_acl_out })
+print(vpp.dump(reply))
+print("---")
+
+
+^D^D^D
+
+run VPP clear trace
+run VPP trace add af-packet-input 100
+run s1 ping6 -c 3 2001:db8:1::2
+expect s1 packet loss
+run VPP show trace
+expect VPP match: outacl 2 rule 0
+
+run VPP clear trace
+run VPP trace add af-packet-input 100
+run s1 ping6 -c 3 2001:db8:1::3
+expect s1 packet loss
+run VPP show trace 
+expect VPP match: inacl 0 rule 1
+
+run VPP clear trace
+run VPP trace add af-packet-input 100
+run s1 ping6 -c 3 2001:db8:1::4
+expect s1 packet loss
+run VPP show trace
+expect VPP match: outacl 2 rule 1
+
+run lua print("ALL GOOD!")
+
diff --git a/src/vpp-api/lua/examples/lute/script-inout-acl-old.lute b/src/vpp-api/lua/examples/lute/script-inout-acl-old.lute
new file mode 100644
index 00000000..9edebf02
--- /dev/null
+++ b/src/vpp-api/lua/examples/lute/script-inout-acl-old.lute
@@ -0,0 +1,329 @@
+shell vppbuild
+run vppbuild stty -echo
+run vppbuild sudo -u ubuntu -i bash -c "(cd vpp && make plugins && echo ALLGOOD)"
+expect vppbuild ALLGOOD
+
+shell s0
+shell s1
+shell s2
+
+
+cd s1
+unshare -n /bin/bash
+/sbin/ifconfig -a
+^D^D^D
+
+cd s2
+unshare -n /bin/bash
+/sbin/ifconfig -a
+^D^D^D
+
+
+cd lua
+
+function session_get_bash_pid(s)
+  if not has_session(s) then
+    return nil
+  end
+  local fname = "/tmp/lute-"..s.."-pid.txt"
+
+  session_exec(s, "echo $$ >" .. fname)
+  -- it's a dirty hack but it's quick
+  sleep(0.5)
+  local pid = io.lines(fname)()
+  print("Got pid for " .. s .. " : " .. tostring(pid))
+  return(tonumber(pid))
+end
+
+function session_connect_with(s0, s1)
+  -- local pid0 = tostring(session_get_bash_pid(s0))
+  local pid1 = tostring(session_get_bash_pid(s1))
+  local eth_options = { "rx", "tx",  "sg", "tso", "ufo", "gso", "gro", "lro", "rxvlan", "txvlan", "rxhash" }
+  local this_end = s0 .. "_" .. s1
+  local other_end = s1 .. "_" .. s0
+  session_exec(s0, "ip link add name " .. this_end .. " type veth peer name " .. other_end)
+  session_exec(s0, "ip link set dev " .. this_end .. " up promisc on")
+  for i, option in ipairs(eth_options) do
+    session_exec(s0, "/sbin/ethtool --offload " .. this_end .. " " .. option .. " off")
+    session_exec(s0, "/sbin/ethtool --offload " .. other_end .. " " .. option .. " off")
+  end
+  session_exec(s0, "ip link set dev " .. other_end .. " up promisc on netns /proc/" .. pid1 .. "/ns/net")
+  sleep(0.5)
+end
+
+^D^D^D
+run lua session_connect_with("s0", "s1")
+run lua session_connect_with("s0", "s2")
+
+cd s1
+ip -6 addr add dev s1_s0 2001:db8:1::1/64
+ip -4 addr add dev s1_s0 192.0.2.1/24
+ip link set dev s1_s0 up promisc on
+^D^D^D
+
+cd s2
+ip -6 addr add dev s2_s0 2001:db8:1::2/64
+ip -6 addr add dev s2_s0 2001:db8:1::3/64
+ip -6 addr add dev s2_s0 2001:db8:1::4/64
+ip -4 addr add dev s2_s0 192.0.2.2/24
+ip -4 addr add dev s2_s0:1 192.0.2.3/24
+ip -4 addr add dev s2_s0:2 192.0.2.4/24
+ip link set dev s2_s0 up promisc on
+^D^D^D
+
+run s1 ip addr
+run s2 ip addr
+shell VPP
+cd VPP
+cd /home/ubuntu/vpp
+make debug 
+r
+^D^D^D
+expect VPP DBGvpp#
+
+cd lua
+-- Initialization of the Lua environment for talking to VPP
+vpp = require("vpp-lapi")
+root_dir = "/home/ubuntu/vpp"
+pneum_path = root_dir .. "/build-root/install-vpp_debug-native/vpp-api/lib64/libpneum.so"
+vpp:init({ pneum_path = pneum_path })
+vpp:consume_api(root_dir .. "/build-root/install-vpp_debug-native/vlib-api/vlibmemory/memclnt.api")
+vpp:consume_api(root_dir .. "/build-root/install-vpp_debug-native/vpp/vpp-api/vpe.api")
+vpp:connect("aytest")
+vpp:consume_api(root_dir .. "/plugins/acl-plugin/acl/acl.api", "acl")
+
+^D^D^D
+
+cd lua
+
+reply = vpp:api_call("af_packet_create", { host_if_name = "s0_s1", hw_addr = "AAAAAA" })
+vpp_if_to_s1 = reply[1].sw_if_index
+
+reply = vpp:api_call("af_packet_create", { host_if_name = "s0_s2", hw_addr = "AAAAAA" })
+vpp_if_to_s2 = reply[1].sw_if_index
+
+ifaces = { vpp_if_to_s1, vpp_if_to_s2 }
+
+reply = vpp:api_call("sw_interface_set_flags", { sw_if_index = vpp_if_to_s1, admin_up_down = 1, link_up_down = 1 })
+print(vpp.dump(reply))
+reply = vpp:api_call("sw_interface_set_flags", { sw_if_index = vpp_if_to_s2, admin_up_down = 1, link_up_down = 1 })
+print(vpp.dump(reply))
+
+bd_id = 42
+
+reply = vpp:api_call("bridge_domain_add_del", { bd_id = bd_id, flood = 1, uu_flood = 1, forward = 1, learn = 1, arp_term = 0, is_add = 1 })
+print(vpp.dump(reply))
+
+for i, v in ipairs(ifaces) do
+  reply = vpp:api_call("sw_interface_set_l2_bridge", { rx_sw_if_index = v, bd_id = bd_id, shg = 0, bvi = 0, enable = 1 } )
+  print(vpp.dump(reply))
+end
+
+^D^D^D
+
+run s1 ping -c 3 192.0.2.2
+expect s1 packet loss
+run s1 ping -c 3 192.0.2.3
+expect s1 packet loss
+run s1 ping -c 3 192.0.2.4
+expect s1 packet loss
+run s1 ping6 -c 3 2001:db8:1::2
+expect s1 packet loss
+run s1 ping6 -c 3 2001:db8:1::3
+expect s1 packet loss
+run s1 ping6 -c 3 2001:db8:1::4
+expect s1 packet loss
+
+
+cd lua
+--- ACL testing
+
+--[[ temporary comment out
+
+reply = vpp:api_call("acl_del", { context = 42, acl_index = 230 })
+print(vpp.dump(reply))
+print("---")
+
+reply = vpp:api_call("acl_del", { context = 42, acl_index = 8 })
+print(vpp.dump(reply))
+print("---")
+
+reply = vpp:api_call("acl_del", { context = 42, acl_index = 15 })
+print(vpp.dump(reply))
+print("---")
+
+reply = vpp:api_call("acl_add", { context = 42, count = 2, r = { { is_permit = 1, is_ipv6 = 1 }, { is_permit = 0, is_ipv6 = 1 } } })
+print(vpp.dump(reply))
+print("---")
+interface_acl_in = reply[1].acl_index
+
+reply = vpp:api_call("acl_add", { context = 42, count = 3, r = { { is_permit = 1, is_ipv6 = 1 }, { is_permit = 0, is_ipv6 = 1 }, { is_permit = 1, is_ipv6 = 0 } } })
+print(vpp.dump(reply))
+print("---")
+interface_acl_out = reply[1].acl_index
+
+
+reply = vpp:api_call("acl_interface_add_del", { context = 42, sw_if_index = 0, is_add = 1, is_input = 1, acl_index = interface_acl_in })
+print(vpp.dump(reply))
+print("---")
+
+reply = vpp:api_call("acl_interface_add_del", { context = 42, sw_if_index = 0, is_add = 1, is_input = 1, acl_index = interface_acl_in })
+print(vpp.dump(reply))
+print("---")
+
+reply = vpp:api_call("acl_interface_add_del", { context = 42, sw_if_index = 0, is_add = 1, is_input = 0, acl_index = interface_acl_out })
+print(vpp.dump(reply))
+print("---")
+
+reply = vpp:api_call("acl_interface_add_del", { context = 42, sw_if_index = 0, is_add = 1, is_input = 0, acl_index = interface_acl_out })
+print(vpp.dump(reply))
+print("---")
+
+reply = vpp:api_call("acl_add", { context = 42, count = 0 })
+print(vpp.dump(reply))
+print("---")
+
+acl_index_to_delete = reply[1].acl_index
+print("Deleting " .. tostring(acl_index_to_delete))
+reply = vpp:api_call("acl_del", { context = 42, acl_index = acl_index_to_delete })
+print(vpp.dump(reply))
+print("---")
+
+reply = vpp:api_call("acl_dump", { context = 42, sw_if_index = 0})
+for ri, rv in ipairs(reply) do
+  print("Reply message #" .. tostring(ri))
+  print(vpp.dump(rv))
+  for ai, av in ipairs(rv.r) do
+    print("ACL rule #" .. tostring(ai) .. " : " .. vpp.dump(av))
+  end
+
+end
+print("---")
+
+reply = vpp:api_call("acl_del", { context = 42, acl_index = interface_acl_out })
+print(vpp.dump(reply))
+print("---")
+reply = vpp:api_call("acl_del", { context = 42, acl_index = interface_acl_in })
+print(vpp.dump(reply))
+print("---")
+
+reply = vpp:api_call("acl_dump", { context = 42, sw_if_index = 0})
+print(vpp.dump(reply))
+print("---")
+
+reply = vpp:api_call("acl_dump", { context = 42, sw_if_index = 4294967295 })
+print(vpp.dump(reply))
+print("---")
+
+
+]] -- end of comment out
+
+---- Should be nothing ^^
+r = {
+  { is_permit = 1, is_ipv6 = 1, dst_ip_addr = ip46("2001:db8:1::2"), dst_ip_prefix_len = 128 },
+  { is_permit = 0, is_ipv6 = 1, dst_ip_addr = ip46("2001:db8:1::3"), dst_ip_prefix_len = 128 },
+  { is_permit = 1, is_ipv6 = 1, dst_ip_addr = ip46("2001:db8::"), dst_ip_prefix_len = 32  },
+  { is_permit = 1, is_ipv6 = 0, dst_ip_addr = ip46("192.0.2.2"), dst_ip_prefix_len = 32},
+  { is_permit = 0, is_ipv6 = 0, dst_ip_addr = ip46("192.0.2.3"), dst_ip_prefix_len = 32 },
+}
+
+reply = vpp:api_call("acl_add", { context = 42, count = 5, r = r })
+print(vpp.dump(reply))
+print("---")
+interface_acl_in = reply[1].acl_index
+
+reply = vpp:api_call("acl_add", { context = 42, count = 3, r = { { is_permit = 1, is_ipv6 = 1 }, { is_permit = 0, is_ipv6 = 1 }, { is_permit = 1, is_ipv6 = 0 } } })
+print(vpp.dump(reply))
+print("---")
+interface_acl_out = reply[1].acl_in
+
+reply = vpp:api_call("acl_interface_add_del", { context = 42, sw_if_index = vpp_if_to_s1, is_add = 1, is_input = 1, acl_index = interface_acl_in })
+print(vpp.dump(reply))
+print("---")
+
+--reply = vpp:api_call("acl_interface_add_del", { context = 42, sw_if_index = vpp_if_to_s2, is_add = 1, is_input = 0, acl_index = interface_acl_out })
+-- print(vpp.dump(reply))
+--print("---")
+
+^D^D^D
+
+run VPP clear trace
+run VPP trace add af-packet-input 100
+run s1 ping6 -c 3 2001:db8:1::2
+expect s1 packet loss
+run VPP show trace
+expect VPP match: inacl 0 rule 0
+
+run VPP clear trace
+run VPP trace add af-packet-input 100
+run s1 ping6 -c 3 2001:db8:1::3
+expect s1 packet loss
+run VPP show trace 
+expect VPP match: inacl 0 rule 1
+
+run VPP clear trace
+run VPP trace add af-packet-input 100
+run s1 ping6 -c 3 2001:db8:1::4
+expect s1 packet loss
+run VPP show trace
+expect VPP match: inacl 0 rule 2
+
+run VPP clear trace
+run VPP trace add af-packet-input 100
+run s1 ping -c 3 192.0.2.2
+expect s1 packet loss
+run VPP show trace
+expect VPP match: inacl 0 rule 3
+
+run VPP clear trace
+run VPP trace add af-packet-input 100
+run s1 ping -c 3 192.0.2.3
+expect s1 packet loss
+run VPP show trace
+expect VPP match: inacl 0 rule 4
+
+
+cd lua
+
+--- TEST OUTBOUND ACL
+
+r1 = {
+  { is_permit = 1, is_ipv6 = 1, src_ip_addr = ip46("2001:db8:1::1"), src_ip_prefix_len = 128, dst_ip_addr = ip46("2001:db8:1::2"), dst_ip_prefix_len = 128 },
+  { is_permit = 0, is_ipv6 = 1, src_ip_addr = ip46("2001:db8:1::1"), src_ip_prefix_len = 128, dst_ip_addr = ip46("2001:db8:1::4"), dst_ip_prefix_len = 128 }
+}
+
+reply = vpp:api_call("acl_add", { context = 42, count = 3, r = r1 })
+print(vpp.dump(reply))
+print("---")
+interface_acl_out = reply[1].acl_index
+
+reply = vpp:api_call("acl_interface_add_del", { context = 42, sw_if_index = vpp_if_to_s2, is_add = 1, is_input = 0, acl_index = interface_acl_out })
+print(vpp.dump(reply))
+print("---")
+
+
+^D^D^D
+
+run VPP clear trace
+run VPP trace add af-packet-input 100
+run s1 ping6 -c 3 2001:db8:1::2
+expect s1 packet loss
+run VPP show trace
+expect VPP match: outacl 2 rule 0
+
+run VPP clear trace
+run VPP trace add af-packet-input 100
+run s1 ping6 -c 3 2001:db8:1::3
+expect s1 packet loss
+run VPP show trace 
+expect VPP match: inacl 0 rule 1
+
+run VPP clear trace
+run VPP trace add af-packet-input 100
+run s1 ping6 -c 3 2001:db8:1::4
+expect s1 packet loss
+run VPP show trace
+expect VPP match: outacl 2 rule 1
+
+run lua print("ALL GOOD!")
+
diff --git a/src/vpp-api/lua/examples/lute/script-inout-acl.lute b/src/vpp-api/lua/examples/lute/script-inout-acl.lute
new file mode 100644
index 00000000..d7e7423c
--- /dev/null
+++ b/src/vpp-api/lua/examples/lute/script-inout-acl.lute
@@ -0,0 +1,329 @@
+shell vppbuild
+run vppbuild stty -echo
+run vppbuild sudo -u ubuntu -i bash -c "(cd vpp && make plugins && echo ALLGOOD)"
+expect vppbuild ALLGOOD
+
+shell s0
+shell s1
+shell s2
+
+
+cd s1
+unshare -n /bin/bash
+/sbin/ifconfig -a
+^D^D^D
+
+cd s2
+unshare -n /bin/bash
+/sbin/ifconfig -a
+^D^D^D
+
+
+cd lua
+
+function session_get_bash_pid(s)
+  if not has_session(s) then
+    return nil
+  end
+  local fname = "/tmp/lute-"..s.."-pid.txt"
+
+  session_exec(s, "echo $$ >" .. fname)
+  -- it's a dirty hack but it's quick
+  sleep(0.5)
+  local pid = io.lines(fname)()
+  print("Got pid for " .. s .. " : " .. tostring(pid))
+  return(tonumber(pid))
+end
+
+function session_connect_with(s0, s1)
+  -- local pid0 = tostring(session_get_bash_pid(s0))
+  local pid1 = tostring(session_get_bash_pid(s1))
+  local eth_options = { "rx", "tx",  "sg", "tso", "ufo", "gso", "gro", "lro", "rxvlan", "txvlan", "rxhash" }
+  local this_end = s0 .. "_" .. s1
+  local other_end = s1 .. "_" .. s0
+  session_exec(s0, "ip link add name " .. this_end .. " type veth peer name " .. other_end)
+  session_exec(s0, "ip link set dev " .. this_end .. " up promisc on")
+  for i, option in ipairs(eth_options) do
+    session_exec(s0, "/sbin/ethtool --offload " .. this_end .. " " .. option .. " off")
+    session_exec(s0, "/sbin/ethtool --offload " .. other_end .. " " .. option .. " off")
+  end
+  session_exec(s0, "ip link set dev " .. other_end .. " up promisc on netns /proc/" .. pid1 .. "/ns/net")
+  sleep(0.5)
+end
+
+^D^D^D
+run lua session_connect_with("s0", "s1")
+run lua session_connect_with("s0", "s2")
+
+cd s1
+ip -6 addr add dev s1_s0 2001:db8:1::1/64
+ip -4 addr add dev s1_s0 192.0.2.1/24
+ip link set dev s1_s0 up promisc on
+^D^D^D
+
+cd s2
+ip -6 addr add dev s2_s0 2001:db8:1::2/64
+ip -6 addr add dev s2_s0 2001:db8:1::3/64
+ip -6 addr add dev s2_s0 2001:db8:1::4/64
+ip -4 addr add dev s2_s0 192.0.2.2/24
+ip -4 addr add dev s2_s0:1 192.0.2.3/24
+ip -4 addr add dev s2_s0:2 192.0.2.4/24
+ip link set dev s2_s0 up promisc on
+^D^D^D
+
+run s1 ip addr
+run s2 ip addr
+shell VPP
+cd VPP
+cd /home/ubuntu/vpp
+make debug 
+r
+^D^D^D
+expect VPP DBGvpp#
+
+cd lua
+-- Initialization of the Lua environment for talking to VPP
+vpp = require("vpp-lapi")
+root_dir = "/home/ubuntu/vpp"
+pneum_path = root_dir .. "/build-root/install-vpp_debug-native/vpp-api/lib64/libpneum.so"
+vpp:init({ pneum_path = pneum_path })
+vpp:consume_api(root_dir .. "/build-root/install-vpp_debug-native/vlib-api/vlibmemory/memclnt.api")
+vpp:consume_api(root_dir .. "/build-root/install-vpp_debug-native/vpp/vpp-api/vpe.api")
+vpp:connect("aytest")
+vpp:consume_api(root_dir .. "/plugins/acl-plugin/acl/acl.api", "acl")
+
+^D^D^D
+
+cd lua
+
+reply = vpp:api_call("af_packet_create", { host_if_name = "s0_s1", hw_addr = "AAAAAA" })
+vpp_if_to_s1 = reply[1].sw_if_index
+
+reply = vpp:api_call("af_packet_create", { host_if_name = "s0_s2", hw_addr = "AAAAAA" })
+vpp_if_to_s2 = reply[1].sw_if_index
+
+ifaces = { vpp_if_to_s1, vpp_if_to_s2 }
+
+reply = vpp:api_call("sw_interface_set_flags", { sw_if_index = vpp_if_to_s1, admin_up_down = 1, link_up_down = 1 })
+print(vpp.dump(reply))
+reply = vpp:api_call("sw_interface_set_flags", { sw_if_index = vpp_if_to_s2, admin_up_down = 1, link_up_down = 1 })
+print(vpp.dump(reply))
+
+bd_id = 42
+
+reply = vpp:api_call("bridge_domain_add_del", { bd_id = bd_id, flood = 1, uu_flood = 1, forward = 1, learn = 1, arp_term = 0, is_add = 1 })
+print(vpp.dump(reply))
+
+for i, v in ipairs(ifaces) do
+  reply = vpp:api_call("sw_interface_set_l2_bridge", { rx_sw_if_index = v, bd_id = bd_id, shg = 0, bvi = 0, enable = 1 } )
+  print(vpp.dump(reply))
+end
+
+^D^D^D
+
+run s1 ping -c 3 192.0.2.2
+expect s1 packet loss
+run s1 ping -c 3 192.0.2.3
+expect s1 packet loss
+run s1 ping -c 3 192.0.2.4
+expect s1 packet loss
+run s1 ping6 -c 3 2001:db8:1::2
+expect s1 packet loss
+run s1 ping6 -c 3 2001:db8:1::3
+expect s1 packet loss
+run s1 ping6 -c 3 2001:db8:1::4
+expect s1 packet loss
+
+
+cd lua
+--- ACL testing
+
+--[[ temporary comment out
+
+reply = vpp:api_call("acl_del", { context = 42, acl_index = 230 })
+print(vpp.dump(reply))
+print("---")
+
+reply = vpp:api_call("acl_del", { context = 42, acl_index = 8 })
+print(vpp.dump(reply))
+print("---")
+
+reply = vpp:api_call("acl_del", { context = 42, acl_index = 15 })
+print(vpp.dump(reply))
+print("---")
+
+reply = vpp:api_call("acl_add_replace", { context = 42, acl_index = -1, count = 2, r = { { is_permit = 1, is_ipv6 = 1 }, { is_permit = 0, is_ipv6 = 1 } } })
+print(vpp.dump(reply))
+print("---")
+interface_acl_in = reply[1].acl_index
+
+reply = vpp:api_call("acl_add_replace", { context = 42, acl_index = -1, count = 3, r = { { is_permit = 1, is_ipv6 = 1 }, { is_permit = 0, is_ipv6 = 1 }, { is_permit = 1, is_ipv6 = 0 } } })
+print(vpp.dump(reply))
+print("---")
+interface_acl_out = reply[1].acl_index
+
+
+reply = vpp:api_call("acl_interface_add_del", { context = 42, sw_if_index = 0, is_add = 1, is_input = 1, acl_index = interface_acl_in })
+print(vpp.dump(reply))
+print("---")
+
+reply = vpp:api_call("acl_interface_add_del", { context = 42, sw_if_index = 0, is_add = 1, is_input = 1, acl_index = interface_acl_in })
+print(vpp.dump(reply))
+print("---")
+
+reply = vpp:api_call("acl_interface_add_del", { context = 42, sw_if_index = 0, is_add = 1, is_input = 0, acl_index = interface_acl_out })
+print(vpp.dump(reply))
+print("---")
+
+reply = vpp:api_call("acl_interface_add_del", { context = 42, sw_if_index = 0, is_add = 1, is_input = 0, acl_index = interface_acl_out })
+print(vpp.dump(reply))
+print("---")
+
+reply = vpp:api_call("acl_add_replace", { context = 42, acl_index = -1, count = 0 })
+print(vpp.dump(reply))
+print("---")
+
+acl_index_to_delete = reply[1].acl_index
+print("Deleting " .. tostring(acl_index_to_delete))
+reply = vpp:api_call("acl_del", { context = 42, acl_index = acl_index_to_delete })
+print(vpp.dump(reply))
+print("---")
+
+reply = vpp:api_call("acl_dump", { context = 42, sw_if_index = 0})
+for ri, rv in ipairs(reply) do
+  print("Reply message #" .. tostring(ri))
+  print(vpp.dump(rv))
+  for ai, av in ipairs(rv.r) do
+    print("ACL rule #" .. tostring(ai) .. " : " .. vpp.dump(av))
+  end
+
+end
+print("---")
+
+reply = vpp:api_call("acl_del", { context = 42, acl_index = interface_acl_out })
+print(vpp.dump(reply))
+print("---")
+reply = vpp:api_call("acl_del", { context = 42, acl_index = interface_acl_in })
+print(vpp.dump(reply))
+print("---")
+
+reply = vpp:api_call("acl_dump", { context = 42, sw_if_index = 0})
+print(vpp.dump(reply))
+print("---")
+
+reply = vpp:api_call("acl_dump", { context = 42, sw_if_index = 4294967295 })
+print(vpp.dump(reply))
+print("---")
+
+
+]] -- end of comment out
+
+---- Should be nothing ^^
+r = {
+  { is_permit = 1, is_ipv6 = 1, dst_ip_addr = ip46("2001:db8:1::2"), dst_ip_prefix_len = 128 },
+  { is_permit = 0, is_ipv6 = 1, dst_ip_addr = ip46("2001:db8:1::3"), dst_ip_prefix_len = 128 },
+  { is_permit = 1, is_ipv6 = 1, dst_ip_addr = ip46("2001:db8::"), dst_ip_prefix_len = 32  },
+  { is_permit = 1, is_ipv6 = 0, dst_ip_addr = ip46("192.0.2.2"), dst_ip_prefix_len = 32},
+  { is_permit = 0, is_ipv6 = 0, dst_ip_addr = ip46("192.0.2.3"), dst_ip_prefix_len = 32 },
+}
+
+reply = vpp:api_call("acl_add_replace", { context = 42, acl_index = -1, count = 5, r = r })
+print(vpp.dump(reply))
+print("---")
+interface_acl_in = reply[1].acl_index
+
+reply = vpp:api_call("acl_add_replace", { context = 42, acl_index = -1, count = 3, r = { { is_permit = 1, is_ipv6 = 1 }, { is_permit = 0, is_ipv6 = 1 }, { is_permit = 1, is_ipv6 = 0 } } })
+print(vpp.dump(reply))
+print("---")
+interface_acl_out = reply[1].acl_in
+
+reply = vpp:api_call("acl_interface_add_del", { context = 42, sw_if_index = vpp_if_to_s1, is_add = 1, is_input = 1, acl_index = interface_acl_in })
+print(vpp.dump(reply))
+print("---")
+
+--reply = vpp:api_call("acl_interface_add_del", { context = 42, sw_if_index = vpp_if_to_s2, is_add = 1, is_input = 0, acl_index = interface_acl_out })
+-- print(vpp.dump(reply))
+--print("---")
+
+^D^D^D
+
+run VPP clear trace
+run VPP trace add af-packet-input 100
+run s1 ping6 -c 3 2001:db8:1::2
+expect s1 packet loss
+run VPP show trace
+expect VPP match: inacl 0 rule 0
+
+run VPP clear trace
+run VPP trace add af-packet-input 100
+run s1 ping6 -c 3 2001:db8:1::3
+expect s1 packet loss
+run VPP show trace 
+expect VPP match: inacl 0 rule 1
+
+run VPP clear trace
+run VPP trace add af-packet-input 100
+run s1 ping6 -c 3 2001:db8:1::4
+expect s1 packet loss
+run VPP show trace
+expect VPP match: inacl 0 rule 2
+
+run VPP clear trace
+run VPP trace add af-packet-input 100
+run s1 ping -c 3 192.0.2.2
+expect s1 packet loss
+run VPP show trace
+expect VPP match: inacl 0 rule 3
+
+run VPP clear trace
+run VPP trace add af-packet-input 100
+run s1 ping -c 3 192.0.2.3
+expect s1 packet loss
+run VPP show trace
+expect VPP match: inacl 0 rule 4
+
+
+cd lua
+
+--- TEST OUTBOUND ACL
+
+r1 = {
+  { is_permit = 1, is_ipv6 = 1, src_ip_addr = ip46("2001:db8:1::1"), src_ip_prefix_len = 128, dst_ip_addr = ip46("2001:db8:1::2"), dst_ip_prefix_len = 128 },
+  { is_permit = 0, is_ipv6 = 1, src_ip_addr = ip46("2001:db8:1::1"), src_ip_prefix_len = 128, dst_ip_addr = ip46("2001:db8:1::4"), dst_ip_prefix_len = 128 }
+}
+
+reply = vpp:api_call("acl_add_replace", { context = 42, acl_index = -1, count = 3, r = r1 })
+print(vpp.dump(reply))
+print("---")
+interface_acl_out = reply[1].acl_index
+
+reply = vpp:api_call("acl_interface_add_del", { context = 42, sw_if_index = vpp_if_to_s2, is_add = 1, is_input = 0, acl_index = interface_acl_out })
+print(vpp.dump(reply))
+print("---")
+
+
+^D^D^D
+
+run VPP clear trace
+run VPP trace add af-packet-input 100
+run s1 ping6 -c 3 2001:db8:1::2
+expect s1 packet loss
+run VPP show trace
+expect VPP match: outacl 2 rule 0
+
+run VPP clear trace
+run VPP trace add af-packet-input 100
+run s1 ping6 -c 3 2001:db8:1::3
+expect s1 packet loss
+run VPP show trace 
+expect VPP match: inacl 0 rule 1
+
+run VPP clear trace
+run VPP trace add af-packet-input 100
+run s1 ping6 -c 3 2001:db8:1::4
+expect s1 packet loss
+run VPP show trace
+expect VPP match: outacl 2 rule 1
+
+run lua print("ALL GOOD!")
+
diff --git a/src/vpp-api/lua/examples/lute/script.lute b/src/vpp-api/lua/examples/lute/script.lute
new file mode 100644
index 00000000..c3dd90f2
--- /dev/null
+++ b/src/vpp-api/lua/examples/lute/script.lute
@@ -0,0 +1,7 @@
+shell s1
+expect s1 $
+run s1 echo testing123
+expect s1 $
+run s1 echo done
+quit
+
diff --git a/src/vpp-api/lua/examples/lute/sessions-acl.lute b/src/vpp-api/lua/examples/lute/sessions-acl.lute
new file mode 100644
index 00000000..ac237ef9
--- /dev/null
+++ b/src/vpp-api/lua/examples/lute/sessions-acl.lute
@@ -0,0 +1,308 @@
+run lua -- collectgarbage("stop")
+
+shell vppbuild
+run vppbuild stty -echo
+run vppbuild sudo -u ubuntu -i bash -c "(cd vpp && make plugins && echo ALLGOOD)"
+expect vppbuild ALLGOOD
+
+shell s0
+shell s1
+shell s2
+
+
+cd s1
+unshare -n /bin/bash
+/sbin/ifconfig -a
+^D^D^D
+
+cd s2
+unshare -n /bin/bash
+/sbin/ifconfig -a
+^D^D^D
+
+
+cd lua
+
+function session_get_bash_pid(s)
+  if not has_session(s) then
+    return nil
+  end
+  local fname = "/tmp/lute-"..s.."-pid.txt"
+
+  session_exec(s, "echo $$ >" .. fname)
+  -- it's a dirty hack but it's quick
+  sleep(0.5)
+  local pid = io.lines(fname)()
+  print("Got pid for " .. s .. " : " .. tostring(pid))
+  return(tonumber(pid))
+end
+
+function session_connect_with(s0, s1)
+  -- local pid0 = tostring(session_get_bash_pid(s0))
+  local pid1 = tostring(session_get_bash_pid(s1))
+  local eth_options = { "rx", "tx",  "sg", "tso", "ufo", "gso", "gro", "lro", "rxvlan", "txvlan", "rxhash" }
+  local this_end = s0 .. "_" .. s1
+  local other_end = s1 .. "_" .. s0
+  session_exec(s0, "ip link add name " .. this_end .. " type veth peer name " .. other_end)
+  session_exec(s0, "ip link set dev " .. this_end .. " up promisc on")
+  for i, option in ipairs(eth_options) do
+    session_exec(s0, "/sbin/ethtool --offload " .. this_end .. " " .. option .. " off")
+    session_exec(s0, "/sbin/ethtool --offload " .. other_end .. " " .. option .. " off")
+  end
+  session_exec(s0, "ip link set dev " .. other_end .. " up promisc on netns /proc/" .. pid1 .. "/ns/net")
+  sleep(0.5)
+end
+
+^D^D^D
+run lua session_connect_with("s0", "s1")
+run lua session_connect_with("s0", "s2")
+
+cd s1
+ip -6 addr add dev s1_s0 2001:db8:1::1/64
+ip -4 addr add dev s1_s0 192.0.2.1/24
+ip link set dev s1_s0 up promisc on
+^D^D^D
+
+cd s2
+ip -6 addr add dev s2_s0 2001:db8:1::2/64
+ip -6 addr add dev s2_s0 2001:db8:1::3/64
+ip -6 addr add dev s2_s0 2001:db8:1::4/64
+ip -4 addr add dev s2_s0 192.0.2.2/24
+ip -4 addr add dev s2_s0:1 192.0.2.3/24
+ip -4 addr add dev s2_s0:2 192.0.2.4/24
+ip link set dev s2_s0 up promisc on
+^D^D^D
+
+run s1 ip addr
+run s2 ip addr
+shell VPP
+cd VPP
+cd /home/ubuntu/vpp
+make debug 
+r
+^D^D^D
+expect VPP DBGvpp#
+
+cd lua
+-- Initialization of the Lua environment for talking to VPP
+vpp = require("vpp-lapi")
+root_dir = "/home/ubuntu/vpp"
+pneum_path = root_dir .. "/build-root/install-vpp_debug-native/vpp-api/lib64/libpneum.so"
+vpp:init({ pneum_path = pneum_path })
+vpp:consume_api(root_dir .. "/build-root/install-vpp_debug-native/vlib-api/vlibmemory/memclnt.api")
+vpp:consume_api(root_dir .. "/build-root/install-vpp_debug-native/vpp/vpp-api/vpe.api")
+vpp:connect("aytest")
+vpp:consume_api(root_dir .. "/plugins/acl-plugin/acl/acl.api", "acl")
+
+^D^D^D
+
+cd lua
+
+reply = vpp:api_call("af_packet_create", { host_if_name = "s0_s1", hw_addr = "AAAAAA" })
+vpp_if_to_s1 = reply[1].sw_if_index
+
+reply = vpp:api_call("af_packet_create", { host_if_name = "s0_s2", hw_addr = "AAAAAA" })
+vpp_if_to_s2 = reply[1].sw_if_index
+
+ifaces = { vpp_if_to_s1, vpp_if_to_s2 }
+
+reply = vpp:api_call("sw_interface_set_flags", { sw_if_index = vpp_if_to_s1, admin_up_down = 1, link_up_down = 1 })
+print(vpp.dump(reply))
+reply = vpp:api_call("sw_interface_set_flags", { sw_if_index = vpp_if_to_s2, admin_up_down = 1, link_up_down = 1 })
+print(vpp.dump(reply))
+
+bd_id = 42
+
+reply = vpp:api_call("bridge_domain_add_del", { bd_id = bd_id, flood = 1, uu_flood = 1, forward = 1, learn = 1, arp_term = 0, is_add = 1 })
+print(vpp.dump(reply))
+
+for i, v in ipairs(ifaces) do
+  reply = vpp:api_call("sw_interface_set_l2_bridge", { rx_sw_if_index = v, bd_id = bd_id, shg = 0, bvi = 0, enable = 1 } )
+  print(vpp.dump(reply))
+end
+
+^D^D^D
+
+run s1 ping -c 3 192.0.2.2
+expect s1 packet loss
+run s1 ping -c 3 192.0.2.3
+expect s1 packet loss
+run s1 ping -c 3 192.0.2.4
+expect s1 packet loss
+run s1 ping6 -c 3 2001:db8:1::2
+expect s1 packet loss
+run s1 ping6 -c 3 2001:db8:1::3
+expect s1 packet loss
+run s1 ping6 -c 3 2001:db8:1::4
+expect s1 packet loss
+
+
+cd lua
+--- ACL testing
+
+--[[ temporary comment out
+
+reply = vpp:api_call("acl_del", { context = 42, acl_index = 230 })
+print(vpp.dump(reply))
+print("---")
+
+reply = vpp:api_call("acl_del", { context = 42, acl_index = 8 })
+print(vpp.dump(reply))
+print("---")
+
+reply = vpp:api_call("acl_del", { context = 42, acl_index = 15 })
+print(vpp.dump(reply))
+print("---")
+
+reply = vpp:api_call("acl_add", { context = 42, count = 2, r = { { is_permit = 1, is_ipv6 = 1 }, { is_permit = 0, is_ipv6 = 1 } } })
+print(vpp.dump(reply))
+print("---")
+interface_acl_in = reply[1].acl_index
+
+reply = vpp:api_call("acl_add", { context = 42, count = 3, r = { { is_permit = 1, is_ipv6 = 1 }, { is_permit = 0, is_ipv6 = 1 }, { is_permit = 1, is_ipv6 = 0 } } })
+print(vpp.dump(reply))
+print("---")
+interface_acl_out = reply[1].acl_index
+
+
+reply = vpp:api_call("acl_interface_add_del", { context = 42, sw_if_index = 0, is_add = 1, is_input = 1, acl_index = interface_acl_in })
+print(vpp.dump(reply))
+print("---")
+
+reply = vpp:api_call("acl_interface_add_del", { context = 42, sw_if_index = 0, is_add = 1, is_input = 1, acl_index = interface_acl_in })
+print(vpp.dump(reply))
+print("---")
+
+reply = vpp:api_call("acl_interface_add_del", { context = 42, sw_if_index = 0, is_add = 1, is_input = 0, acl_index = interface_acl_out })
+print(vpp.dump(reply))
+print("---")
+
+reply = vpp:api_call("acl_interface_add_del", { context = 42, sw_if_index = 0, is_add = 1, is_input = 0, acl_index = interface_acl_out })
+print(vpp.dump(reply))
+print("---")
+
+reply = vpp:api_call("acl_add", { context = 42, count = 0 })
+print(vpp.dump(reply))
+print("---")
+
+acl_index_to_delete = reply[1].acl_index
+print("Deleting " .. tostring(acl_index_to_delete))
+reply = vpp:api_call("acl_del", { context = 42, acl_index = acl_index_to_delete })
+print(vpp.dump(reply))
+print("---")
+
+reply = vpp:api_call("acl_dump", { context = 42, sw_if_index = 0})
+for ri, rv in ipairs(reply) do
+  print("Reply message #" .. tostring(ri))
+  print(vpp.dump(rv))
+  for ai, av in ipairs(rv.r) do
+    print("ACL rule #" .. tostring(ai) .. " : " .. vpp.dump(av))
+  end
+
+end
+print("---")
+
+reply = vpp:api_call("acl_del", { context = 42, acl_index = interface_acl_out })
+print(vpp.dump(reply))
+print("---")
+reply = vpp:api_call("acl_del", { context = 42, acl_index = interface_acl_in })
+print(vpp.dump(reply))
+print("---")
+
+reply = vpp:api_call("acl_dump", { context = 42, sw_if_index = 0})
+print(vpp.dump(reply))
+print("---")
+
+reply = vpp:api_call("acl_dump", { context = 42, sw_if_index = 4294967295 })
+print(vpp.dump(reply))
+print("---")
+
+
+]] -- end of comment out
+
+---- Should be nothing ^^
+r = {
+  { is_permit = 1, is_ipv6 = 1, dst_ip_addr = ip46("2001:db8:1::2"), dst_ip_prefix_len = 128 },
+  { is_permit = 0, is_ipv6 = 1, dst_ip_addr = ip46("2001:db8:1::3"), dst_ip_prefix_len = 128 },
+  { is_permit = 1, is_ipv6 = 1, dst_ip_addr = ip46("2001:db8::"), dst_ip_prefix_len = 32  },
+  { is_permit = 1, is_ipv6 = 0, dst_ip_addr = ip46("192.0.2.2"), dst_ip_prefix_len = 32},
+  { is_permit = 0, is_ipv6 = 0, dst_ip_addr = ip46("192.0.2.3"), dst_ip_prefix_len = 32 },
+}
+
+reply = vpp:api_call("acl_add", { context = 42, count = 5, r = r })
+print(vpp.dump(reply))
+print("---")
+interface_acl_in = reply[1].acl_index
+
+reply = vpp:api_call("acl_add", { context = 42, count = 3, r = { { is_permit = 1, is_ipv6 = 1 }, { is_permit = 0, is_ipv6 = 1 }, { is_permit = 1, is_ipv6 = 0 } } })
+print(vpp.dump(reply))
+print("---")
+interface_acl_out = reply[1].acl_in
+
+reply = vpp:api_call("acl_interface_add_del", { context = 42, sw_if_index = vpp_if_to_s1, is_add = 1, is_input = 1, acl_index = interface_acl_in })
+print(vpp.dump(reply))
+print("---")
+
+--- TEST OUTBOUND ACL
+
+r1 = {
+  { is_permit = 1, is_ipv6 = 1, src_ip_addr = ip46("2001:db8:1::1"), src_ip_prefix_len = 128, dst_ip_addr = ip46("2001:db8:1::2"), dst_ip_prefix_len = 128 },
+  { is_permit = 0, is_ipv6 = 1, src_ip_addr = ip46("2001:db8:1::1"), src_ip_prefix_len = 128, dst_ip_addr = ip46("2001:db8:1::4"), dst_ip_prefix_len = 128 },
+  { is_permit = 2, is_ipv6 = 0 }
+}
+
+reply = vpp:api_call("acl_add", { context = 42, count = 3, r = r1 })
+print(vpp.dump(reply))
+print("---")
+interface_acl_out = reply[1].acl_index
+
+reply = vpp:api_call("acl_interface_add_del", { context = 42, sw_if_index = vpp_if_to_s2, is_add = 1, is_input = 0, acl_index = interface_acl_out })
+print(vpp.dump(reply))
+print("---")
+
+r2 = {
+  { is_permit = 1, is_ipv6 = 1 },
+  { is_permit = 0, is_ipv6 = 0 }
+}
+
+reply = vpp:api_call("acl_add", { context = 42, count = 2, r = r2 })
+print(vpp.dump(reply))
+print("---")
+second_interface_acl_in = reply[1].acl_index
+
+reply = vpp:api_call("acl_interface_add_del", { context = 42, sw_if_index = vpp_if_to_s2, is_add = 1, is_input = 1, acl_index = second_interface_acl_in })
+print(vpp.dump(reply))
+print("---")
+
+^D^D^D
+
+run VPP show classify tables
+run VPP clear trace
+run VPP trace add af-packet-input 100
+run s2 nc -v -l -p 22
+run s1 nc 192.0.2.2 22
+run s1 echo
+sleep 1
+run s1 break
+sleep 1
+run VPP show trace
+expect VPP match: outacl 2 rule 2
+run VPP show classify tables
+
+
+run VPP show classify tables
+run VPP clear trace
+run VPP trace add af-packet-input 100
+run s2 nc -v -l -p 22
+run s1 nc 192.0.2.2 22
+run s1 echo
+sleep 1
+run s1 break
+sleep 1
+run VPP show trace
+expect VPP match: outacl 2 rule 2
+run VPP show classify tables
+
+
+run lua print("ALL GOOD!")
+
diff --git a/src/vpp-api/lua/vpp-lapi.lua b/src/vpp-api/lua/vpp-lapi.lua
new file mode 100644
index 00000000..ebfd032b
--- /dev/null
+++ b/src/vpp-api/lua/vpp-lapi.lua
@@ -0,0 +1,989 @@
+--[[
+/*
+ * Copyright (c) 2016 Cisco and/or its affiliates.
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at:
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+]]
+
+-- json decode/encode from https://gist.github.com/tylerneylon/59f4bcf316be525b30ab
+-- licensed by the author tylerneylon into public domain. Thanks!
+
+local json = {}
+
+-- Internal functions.
+
+local function kind_of(obj)
+  if type(obj) ~= 'table' then return type(obj) end
+  local i = 1
+  for _ in pairs(obj) do
+    if obj[i] ~= nil then i = i + 1 else return 'table' end
+  end
+  if i == 1 then return 'table' else return 'array' end
+end
+
+local function escape_str(s)
+  local in_char  = {'\\', '"', '/', '\b', '\f', '\n', '\r', '\t'}
+  local out_char = {'\\', '"', '/',  'b',  'f',  'n',  'r',  't'}
+  for i, c in ipairs(in_char) do
+    s = s:gsub(c, '\\' .. out_char[i])
+  end
+  return s
+end
+
+-- Returns pos, did_find; there are two cases:
+-- 1. Delimiter found: pos = pos after leading space + delim; did_find = true.
+-- 2. Delimiter not found: pos = pos after leading space;     did_find = false.
+-- This throws an error if err_if_missing is true and the delim is not found.
+local function skip_delim(str, pos, delim, err_if_missing)
+  pos = pos + #str:match('^%s*', pos)
+  if str:sub(pos, pos) ~= delim then
+    if err_if_missing then
+      error('Expected ' .. delim .. ' near position ' .. pos)
+    end
+    return pos, false
+  end
+  return pos + 1, true
+end
+
+-- Expects the given pos to be the first character after the opening quote.
+-- Returns val, pos; the returned pos is after the closing quote character.
+local function parse_str_val(str, pos, val)
+  val = val or ''
+  local early_end_error = 'End of input found while parsing string.'
+  if pos > #str then error(early_end_error) end
+  local c = str:sub(pos, pos)
+  if c == '"'  then return val, pos + 1 end
+  if c ~= '\\' then return parse_str_val(str, pos + 1, val .. c) end
+  -- We must have a \ character.
+  local esc_map = {b = '\b', f = '\f', n = '\n', r = '\r', t = '\t'}
+  local nextc = str:sub(pos + 1, pos + 1)
+  if not nextc then error(early_end_error) end
+  return parse_str_val(str, pos + 2, val .. (esc_map[nextc] or nextc))
+end
+
+-- Returns val, pos; the returned pos is after the number's final character.
+local function parse_num_val(str, pos)
+  local num_str = str:match('^-?%d+%.?%d*[eE]?[+-]?%d*', pos)
+  local val = tonumber(num_str)
+  if not val then error('Error parsing number at position ' .. pos .. '.') end
+  return val, pos + #num_str
+end
+
+
+-- Public values and functions.
+
+function json.stringify(obj, as_key)
+  local s = {}  -- We'll build the string as an array of strings to be concatenated.
+  local kind = kind_of(obj)  -- This is 'array' if it's an array or type(obj) otherwise.
+  if kind == 'array' then
+    if as_key then error('Can\'t encode array as key.') end
+    s[#s + 1] = '['
+    for i, val in ipairs(obj) do
+      if i > 1 then s[#s + 1] = ', ' end
+      s[#s + 1] = json.stringify(val)
+    end
+    s[#s + 1] = ']'
+  elseif kind == 'table' then
+    if as_key then error('Can\'t encode table as key.') end
+    s[#s + 1] = '{'
+    for k, v in pairs(obj) do
+      if #s > 1 then s[#s + 1] = ', ' end
+      s[#s + 1] = json.stringify(k, true)
+      s[#s + 1] = ':'
+      s[#s + 1] = json.stringify(v)
+    end
+    s[#s + 1] = '}'
+  elseif kind == 'string' then
+    return '"' .. escape_str(obj) .. '"'
+  elseif kind == 'number' then
+    if as_key then return '"' .. tostring(obj) .. '"' end
+    return tostring(obj)
+  elseif kind == 'boolean' then
+    return tostring(obj)
+  elseif kind == 'nil' then
+    return 'null'
+  else
+    error('Unjsonifiable type: ' .. kind .. '.')
+  end
+  return table.concat(s)
+end
+
+json.null = {}  -- This is a one-off table to represent the null value.
+
+function json.parse(str, pos, end_delim)
+  pos = pos or 1
+  if pos > #str then error('Reached unexpected end of input.') end
+  local pos = pos + #str:match('^%s*', pos)  -- Skip whitespace.
+  local first = str:sub(pos, pos)
+  if first == '{' then  -- Parse an object.
+    local obj, key, delim_found = {}, true, true
+    pos = pos + 1
+    while true do
+      key, pos = json.parse(str, pos, '}')
+      if key == nil then return obj, pos end
+      if not delim_found then error('Comma missing between object items.') end
+      pos = skip_delim(str, pos, ':', true)  -- true -> error if missing.
+      obj[key], pos = json.parse(str, pos)
+      pos, delim_found = skip_delim(str, pos, ',')
+    end
+  elseif first == '[' then  -- Parse an array.
+    local arr, val, delim_found = {}, true, true
+    pos = pos + 1
+    while true do
+      val, pos = json.parse(str, pos, ']')
+      if val == nil then return arr, pos end
+      if not delim_found then error('Comma missing between array items.') end
+      arr[#arr + 1] = val
+      pos, delim_found = skip_delim(str, pos, ',')
+    end
+  elseif first == '"' then  -- Parse a string.
+    return parse_str_val(str, pos + 1)
+  elseif first == '-' or first:match('%d') then  -- Parse a number.
+    return parse_num_val(str, pos)
+  elseif first == end_delim then  -- End of an object or array.
+    return nil, pos + 1
+  else  -- Parse true, false, or null.
+    local literals = {['true'] = true, ['false'] = false, ['null'] = json.null}
+    for lit_str, lit_val in pairs(literals) do
+      local lit_end = pos + #lit_str - 1
+      if str:sub(pos, lit_end) == lit_str then return lit_val, lit_end + 1 end
+    end
+    local pos_info_str = 'position ' .. pos .. ': ' .. str:sub(pos, pos + 10)
+    error('Invalid json syntax starting at ' .. pos_info_str)
+  end
+end
+
+
+local vpp = {}
+
+local ffi = require("ffi")
+
+--[[
+
+The basic type definitions. A bit of weird gymnastic with
+unionization of the hton* and ntoh* functions results
+is to make handling of signed and unsigned types a bit cleaner,
+essentially building typecasting into a C union.
+
+The vl_api_opaque_message_t is a synthetic type assumed to have
+enough storage to hold the entire API message regardless of the type.
+During the operation it is casted to the specific message struct types.
+
+]]
+
+
+ffi.cdef([[
+
+typedef uint8_t u8;
+typedef int8_t i8;
+typedef uint16_t u16;
+typedef int16_t i16;
+typedef uint32_t u32;
+typedef int32_t i32;
+typedef uint64_t u64;
+typedef int64_t i64;
+typedef double f64;
+typedef float f32;
+
+#pragma pack(1)
+typedef union {
+  u16 u16;
+  i16 i16;
+} lua_ui16t;
+
+#pragma pack(1)
+typedef union {
+  u32 u32;
+  i32 i32;
+} lua_ui32t;
+
+u16 ntohs(uint16_t hostshort);
+u16 htons(uint16_t hostshort);
+u32 htonl(uint32_t along);
+u32 ntohl(uint32_t along);
+void *memset(void *s, int c, size_t n);
+void *memcpy(void *dest, void *src, size_t n);
+
+#pragma pack(1)
+typedef struct _vl_api_opaque_message {
+  u16 _vl_msg_id;
+  u8  data[65536];
+} vl_api_opaque_message_t;
+]])
+
+
+-- CRC-based version stuff
+
+local crc32c_table = ffi.new('const uint32_t[256]',
+  { 0x00000000, 0xF26B8303, 0xE13B70F7, 0x1350F3F4,
+  0xC79A971F, 0x35F1141C, 0x26A1E7E8, 0xD4CA64EB,
+  0x8AD958CF, 0x78B2DBCC, 0x6BE22838, 0x9989AB3B,
+  0x4D43CFD0, 0xBF284CD3, 0xAC78BF27, 0x5E133C24,
+  0x105EC76F, 0xE235446C, 0xF165B798, 0x030E349B,
+  0xD7C45070, 0x25AFD373, 0x36FF2087, 0xC494A384,
+  0x9A879FA0, 0x68EC1CA3, 0x7BBCEF57, 0x89D76C54,
+  0x5D1D08BF, 0xAF768BBC, 0xBC267848, 0x4E4DFB4B,
+  0x20BD8EDE, 0xD2D60DDD, 0xC186FE29, 0x33ED7D2A,
+  0xE72719C1, 0x154C9AC2, 0x061C6936, 0xF477EA35,
+  0xAA64D611, 0x580F5512, 0x4B5FA6E6, 0xB93425E5,
+  0x6DFE410E, 0x9F95C20D, 0x8CC531F9, 0x7EAEB2FA,
+  0x30E349B1, 0xC288CAB2, 0xD1D83946, 0x23B3BA45,
+  0xF779DEAE, 0x05125DAD, 0x1642AE59, 0xE4292D5A,
+  0xBA3A117E, 0x4851927D, 0x5B016189, 0xA96AE28A,
+  0x7DA08661, 0x8FCB0562, 0x9C9BF696, 0x6EF07595,
+  0x417B1DBC, 0xB3109EBF, 0xA0406D4B, 0x522BEE48,
+  0x86E18AA3, 0x748A09A0, 0x67DAFA54, 0x95B17957,
+  0xCBA24573, 0x39C9C670, 0x2A993584, 0xD8F2B687,
+  0x0C38D26C, 0xFE53516F, 0xED03A29B, 0x1F682198,
+  0x5125DAD3, 0xA34E59D0, 0xB01EAA24, 0x42752927,
+  0x96BF4DCC, 0x64D4CECF, 0x77843D3B, 0x85EFBE38,
+  0xDBFC821C, 0x2997011F, 0x3AC7F2EB, 0xC8AC71E8,
+  0x1C661503, 0xEE0D9600, 0xFD5D65F4, 0x0F36E6F7,
+  0x61C69362, 0x93AD1061, 0x80FDE395, 0x72966096,
+  0xA65C047D, 0x5437877E, 0x4767748A, 0xB50CF789,
+  0xEB1FCBAD, 0x197448AE, 0x0A24BB5A, 0xF84F3859,
+  0x2C855CB2, 0xDEEEDFB1, 0xCDBE2C45, 0x3FD5AF46,
+  0x7198540D, 0x83F3D70E, 0x90A324FA, 0x62C8A7F9,
+  0xB602C312, 0x44694011, 0x5739B3E5, 0xA55230E6,
+  0xFB410CC2, 0x092A8FC1, 0x1A7A7C35, 0xE811FF36,
+  0x3CDB9BDD, 0xCEB018DE, 0xDDE0EB2A, 0x2F8B6829,
+  0x82F63B78, 0x709DB87B, 0x63CD4B8F, 0x91A6C88C,
+  0x456CAC67, 0xB7072F64, 0xA457DC90, 0x563C5F93,
+  0x082F63B7, 0xFA44E0B4, 0xE9141340, 0x1B7F9043,
+  0xCFB5F4A8, 0x3DDE77AB, 0x2E8E845F, 0xDCE5075C,
+  0x92A8FC17, 0x60C37F14, 0x73938CE0, 0x81F80FE3,
+  0x55326B08, 0xA759E80B, 0xB4091BFF, 0x466298FC,
+  0x1871A4D8, 0xEA1A27DB, 0xF94AD42F, 0x0B21572C,
+  0xDFEB33C7, 0x2D80B0C4, 0x3ED04330, 0xCCBBC033,
+  0xA24BB5A6, 0x502036A5, 0x4370C551, 0xB11B4652,
+  0x65D122B9, 0x97BAA1BA, 0x84EA524E, 0x7681D14D,
+  0x2892ED69, 0xDAF96E6A, 0xC9A99D9E, 0x3BC21E9D,
+  0xEF087A76, 0x1D63F975, 0x0E330A81, 0xFC588982,
+  0xB21572C9, 0x407EF1CA, 0x532E023E, 0xA145813D,
+  0x758FE5D6, 0x87E466D5, 0x94B49521, 0x66DF1622,
+  0x38CC2A06, 0xCAA7A905, 0xD9F75AF1, 0x2B9CD9F2,
+  0xFF56BD19, 0x0D3D3E1A, 0x1E6DCDEE, 0xEC064EED,
+  0xC38D26C4, 0x31E6A5C7, 0x22B65633, 0xD0DDD530,
+  0x0417B1DB, 0xF67C32D8, 0xE52CC12C, 0x1747422F,
+  0x49547E0B, 0xBB3FFD08, 0xA86F0EFC, 0x5A048DFF,
+  0x8ECEE914, 0x7CA56A17, 0x6FF599E3, 0x9D9E1AE0,
+  0xD3D3E1AB, 0x21B862A8, 0x32E8915C, 0xC083125F,
+  0x144976B4, 0xE622F5B7, 0xF5720643, 0x07198540,
+  0x590AB964, 0xAB613A67, 0xB831C993, 0x4A5A4A90,
+  0x9E902E7B, 0x6CFBAD78, 0x7FAB5E8C, 0x8DC0DD8F,
+  0xE330A81A, 0x115B2B19, 0x020BD8ED, 0xF0605BEE,
+  0x24AA3F05, 0xD6C1BC06, 0xC5914FF2, 0x37FACCF1,
+  0x69E9F0D5, 0x9B8273D6, 0x88D28022, 0x7AB90321,
+  0xAE7367CA, 0x5C18E4C9, 0x4F48173D, 0xBD23943E,
+  0xF36E6F75, 0x0105EC76, 0x12551F82, 0xE03E9C81,
+  0x34F4F86A, 0xC69F7B69, 0xD5CF889D, 0x27A40B9E,
+  0x79B737BA, 0x8BDCB4B9, 0x988C474D, 0x6AE7C44E,
+  0xBE2DA0A5, 0x4C4623A6, 0x5F16D052, 0xAD7D5351 }
+);
+
+local function CRC8(crc, d)
+  return bit.bxor(bit.rshift(crc, 8), crc32c_table[bit.band(0xff, bit.bxor(crc, d))])
+end
+
+local function CRC16(crc, d)
+  crc = CRC8(crc, bit.band(d, 0xFF))
+  d = bit.rshift(d, 8)
+  crc = CRC8(crc, bit.band(d, 0xFF))
+  return crc
+end
+
+local function string_crc(str, crc)
+  for i=1,#str do
+    -- print("S", i, string.byte(str, i), string.char(string.byte(str, i)))
+    crc = CRC8(crc, string.byte(str, i))
+  end
+  return crc
+end
+
+local tokens = {
+  { ["match"] =' ', ["act"]             = { }  },
+  { ["match"] ='\n', ["act"]             = { }  },
+  { ["match"] ="manual_endian", ["act"]  = { "NODE_MANUAL_ENDIAN", "MANUAL_ENDIAN",    276 } },
+  { ["match"] ="define", ["act"]         = { "NODE_DEFINE",        "DEFINE",           267 } },
+  { ["match"] ="dont_trace", ["act"]     = { "NODE_DONT_TRACE",    "DONT_TRACE",       279 } },
+  { ["match"] ="f64", ["act"]            = { "NODE_F64",           "PRIMTYPE",         string_crc } },
+  { ["match"] ="i16", ["act"]            = { "NODE_I16",           "PRIMTYPE",         string_crc } },
+  { ["match"] ="i32", ["act"]            = { "NODE_I32",           "PRIMTYPE",         string_crc } },
+  { ["match"] ="i64", ["act"]            = { "NODE_I64",           "PRIMTYPE",         string_crc } },
+  { ["match"] ="i8", ["act"]             = { "NODE_I8",            "PRIMTYPE",         string_crc } },
+  { ["match"] ="manual_print", ["act"]   = { "NODE_MANUAL_PRINT",  "MANUAL_PRINT",     275 } },
+  { ["match"] ="noversion", ["act"]      = { "NODE_NOVERSION",     "NOVERSION",        274 } },
+  { ["match"] ="packed", ["act"]         = { "NODE_PACKED",        "TPACKED",          266 } },
+  { ["match"] ="typeonly", ["act"]       = { "NODE_TYPEONLY",      "TYPEONLY",         278 } },
+  { ["match"] ="u16", ["act"]            = { "NODE_U16",           "PRIMTYPE",         string_crc } },
+  { ["match"] ="u32", ["act"]            = { "NODE_U32",           "PRIMTYPE",         string_crc } },
+  { ["match"] ="u64", ["act"]            = { "NODE_U64",           "PRIMTYPE",         string_crc } },
+  { ["match"] ="u8", ["act"]             = { "NODE_U8",            "PRIMTYPE",         string_crc } },
+  { ["match"] ="union", ["act"]          = { "NODE_UNION",         "UNION",            271 } },
+  { ["match"] ="uword", ["act"]          = { "NODE_UWORD",         "PRIMTYPE",         string_crc } },
+  { ["match"] ="%(", ["act"]             = { "NODE_LPAR",          "LPAR",             259 } },
+  { ["match"] ="%)", ["act"]             = { "NODE_RPAR",          "RPAR",             258 } },
+  { ["match"] =";", ["act"]              = { "NODE_SEMI",          "SEMI",             260 } },
+  { ["match"] ="%[", ["act"]             = { "NODE_LBRACK",        "LBRACK",           261 } },
+  { ["match"] ="%]", ["act"]             = { "NODE_RBRACK",        "RBRACK",           262 } },
+  { ["match"] ="%{", ["act"]             = { "NODE_LCURLY",        "LCURLY",           268 } },
+  { ["match"] ="%}", ["act"]             = { "NODE_RCURLY",        "RCURLY",           269 } },
+  { ["match"] ='%b""', ["act"]           = { "NODE_STRING",        "STRING",           string_crc } },
+  { ["match"] ='%b@@', ["act"]           = { "NODE_HELPER",        "HELPER_STRING",    string_crc } },
+  -- TODO: \ must be consumed
+  { ["match"] ='[_a-zA-Z][_a-zA-Z0-9]*',
+                       ["act"]           = { "NODE_NAME",          "NAME",             string_crc } },
+  { ["match"] ='[0-9]+', ["act"]         = { "NODE_NUMBER",        "NUMBER",           string_crc } },
+  { ["match"] ='#[^\n]+', ["act"]            = { "NODE_PRAGMA",        "PRAGMA",           nil } },
+}
+
+
+function vpp.crc_version_string(data)
+  local input_crc = 0
+  -- Get rid of comments
+  data = data:gsub("/%*.-%*/", "")
+  data = data:gsub("//[^\n]+", "")
+  -- print(data)
+  idx = 1
+  while (true) do
+    local matched = nil
+    for k, v in ipairs(tokens) do
+      if not matched then
+        local x, y, cap = string.find(data, v["match"], idx)
+        if x == idx then
+          matched = { ["node"] = v["act"], ["x"] = x, ["y"] = y, ["cap"] = cap, ["chars"] = string.sub(data, x, y)  }
+          -- print(k, v, x, y, cap, matched.chars, matched.node[0] )
+        end
+      end
+    end
+    if matched then
+      idx = idx + (matched.y - matched.x + 1)
+      if matched.node[1] then
+        local act = matched.node[3]
+        if type(act) == "function" then
+          input_crc = act(matched.chars, input_crc)
+        elseif type(act) == "number" then
+          input_crc = CRC16(input_crc, act)
+        end
+        -- print(vpp.dump(matched))
+      end
+    else
+      -- print("NOT MATCHED!")
+      local crc = CRC16(input_crc, 0xFFFFFFFF)
+      return string.sub(string.format("%x", crc), -8)
+    end
+  end
+end
+
+
+function vpp.dump(o)
+   if type(o) == 'table' then
+      local s = '{ '
+      for k,v in pairs(o) do
+         if type(k) ~= 'number' then k = '"'..k..'"' end
+         s = s .. '['..k..'] = ' .. vpp.dump(v) .. ','
+      end
+      return s .. '} '
+   else
+      return tostring(o)
+   end
+end
+
+function vpp.hex_dump(buf)
+  local ret = {}
+  for i=1,math.ceil(#buf/16) * 16 do
+    if (i-1) % 16 == 0 then table.insert(ret, string.format('%08X  ', i-1)) end
+    table.insert(ret, ( i > #buf and '   ' or string.format('%02X ', buf:byte(i)) ))
+    if i %  8 == 0 then table.insert(ret, ' ') end
+    if i % 16 == 0 then table.insert(ret, buf:sub(i-16+1, i):gsub('%c','.')..'\n' ) end
+  end
+  return table.concat(ret)
+end
+
+
+function vpp.c_str(text_in)
+  local text = text_in -- \000 will be helpfully added by ffi.copy
+  local c_str = ffi.new("char[?]", #text+1)
+  ffi.copy(c_str, text)
+  return c_str
+end
+
+
+function vpp.init(vpp, args)
+  local pneum_api = args.pneum_api or [[
+ int cough_pneum_attach(char *pneum_path, char *cough_path);
+ int pneum_connect(char *name, char *chroot_prefix, void *cb);
+ int pneum_disconnect(void);
+ int pneum_read(char **data, int *l);
+ int pneum_write(char *data, int len);
+ void pneum_free(char *data);
+ uint32_t pneum_get_msg_index(unsigned char * name);
+]]
+
+  vpp.pneum_path = args.pneum_path
+  ffi.cdef(pneum_api)
+  local init_res = 0
+  vpp.pneum = ffi.load(vpp.pneum_path)
+  if (init_res < 0) then
+    return nil
+  end
+
+  vpp.next_msg_num = 1
+  vpp.msg_name_to_number = {}
+  vpp.msg_name_to_fields = {}
+  vpp.msg_number_to_name = {}
+  vpp.msg_number_to_type = {}
+  vpp.msg_number_to_pointer_type = {}
+  vpp.msg_name_to_crc = {}
+  vpp.c_type_to_fields = {}
+  vpp.events = {}
+  vpp.plugin_version = {}
+  vpp.is_connected = false
+
+
+  vpp.t_lua2c = {}
+  vpp.t_c2lua = {}
+  vpp.t_lua2c["u8"] = function(c_type, src, dst_c_ptr)
+    if type(src) == "string" then
+      -- ffi.copy adds a zero byte at the end. Grrr.
+      -- ffi.copy(dst_c_ptr, src)
+      ffi.C.memcpy(dst_c_ptr, vpp.c_str(src), #src)
+      return(#src)
+    elseif type(src) == "table" then
+      for i,v in ipairs(src) do
+        ffi.cast("u8 *", dst_c_ptr)[i-1] = v
+      end
+      return(#src)
+    else
+      return 1, src -- ffi.cast("u8", src)
+    end
+  end
+  vpp.t_c2lua["u8"] = function(c_type, src_ptr, src_len)
+    if src_len then
+      return ffi.string(src_ptr, src_len)
+    else
+      return (tonumber(src_ptr))
+    end
+  end
+
+  vpp.t_lua2c["u16"] = function(c_type, src, dst_c_ptr)
+    if type(src) == "table" then
+      for i,v in ipairs(src) do
+        ffi.cast("u16 *", dst_c_ptr)[i-1] = ffi.C.htons(v)
+      end
+      return(2 * #src)
+    else
+      return 2, (ffi.C.htons(src))
+    end
+  end
+  vpp.t_c2lua["u16"] = function(c_type, src_ptr, src_len)
+    if src_len then
+      local out = {}
+      for i = 0,src_len-1 do
+        out[i+1] = tonumber(ffi.C.ntohs(src_ptr[i]))
+      end
+      return out
+    else
+      return (tonumber(ffi.C.ntohs(src_ptr)))
+    end
+  end
+
+  vpp.t_lua2c["u32"] = function(c_type, src, dst_c_ptr)
+    if type(src) == "table" then
+      for i,v in ipairs(src) do
+        ffi.cast("u32 *", dst_c_ptr)[i-1] = ffi.C.htonl(v)
+      end
+      return(4 * #src)
+    else
+      return 4, (ffi.C.htonl(src))
+    end
+  end
+  vpp.t_c2lua["u32"] = function(c_type, src_ptr, src_len)
+    if src_len then
+      local out = {}
+      for i = 0,src_len-1 do
+        out[i+1] = tonumber(ffi.C.ntohl(src_ptr[i]))
+      end
+      return out
+    else
+      return (tonumber(ffi.C.ntohl(src_ptr)))
+    end
+  end
+  vpp.t_lua2c["i32"] = function(c_type, src, dst_c_ptr)
+    if type(src) == "table" then
+      for i,v in ipairs(src) do
+        ffi.cast("i32 *", dst_c_ptr)[i-1] = ffi.C.htonl(v)
+      end
+      return(4 * #src)
+    else
+      return 4, (ffi.C.htonl(src))
+    end
+  end
+  vpp.t_c2lua["i32"] = function(c_type, src_ptr, src_len)
+    local ntohl = function(src)
+      local u32val = ffi.cast("u32", src)
+      local ntohlval = (ffi.C.ntohl(u32val))
+      local out = tonumber(ffi.cast("i32", ntohlval + 0LL))
+      return out
+    end
+    if src_len then
+      local out = {}
+      for i = 0,src_len-1 do
+        out[i+1] = tonumber(ntohl(src_ptr[i]))
+      end
+    else
+      return (tonumber(ntohl(src_ptr)))
+    end
+  end
+
+  vpp.t_lua2c["u64"] = function(c_type, src, dst_c_ptr)
+    if type(src) == "table" then
+      for i,v in ipairs(src) do
+        ffi.cast("u64 *", dst_c_ptr)[i-1] = v --- FIXME ENDIAN
+      end
+      return(8 * #src)
+    else
+      return 8, ffi.cast("u64", src) --- FIXME ENDIAN
+    end
+  end
+  vpp.t_c2lua["u64"] = function(c_type, src_ptr, src_len)
+    if src_len then
+      local out = {}
+      for i = 0,src_len-1 do
+        out[i+1] = tonumber(src_ptr[i]) -- FIXME ENDIAN
+      end
+      return out
+    else
+      return (tonumber(src_ptr)) --FIXME ENDIAN
+    end
+  end
+
+
+
+
+  vpp.t_lua2c["__MSG__"] = function(c_type, src, dst_c_ptr)
+    local dst = ffi.cast(c_type .. " *", dst_c_ptr)
+    local additional_len = 0
+    local fields_info = vpp.c_type_to_fields[c_type]
+    -- print("__MSG__ type: " .. tostring(c_type))
+    ffi.C.memset(dst_c_ptr, 0, ffi.sizeof(dst[0]))
+    -- print(vpp.dump(fields_info))
+    -- print(vpp.dump(src))
+    for k,v in pairs(src) do
+      local field = fields_info[k]
+      if not field then
+        print("ERROR: field " .. tostring(k) .. " in message " .. tostring(c_type) .. " is unknown")
+      end
+      local lua2c = vpp.t_lua2c[field.c_type]
+      -- print("__MSG__ field " .. tostring(k) .. " : " .. vpp.dump(field))
+      -- if the field is not an array type, try to coerce the argument to a number
+      if not field.array and type(v) == "string" then
+        v = tonumber(v)
+      end
+      if not lua2c then
+        print("__MSG__ " .. tostring(c_type) .. " t_lua2c: can not store field " .. field.name ..
+              " type " .. field.c_type .. " dst " .. tostring(dst[k]))
+        return 0
+      end
+      local len = 0
+      local val = nil
+      if field.array and (type(v) == "table") then
+        -- print("NTFY: field " .. tostring(k) .. " in message " .. tostring(c_type) .. " is an array")
+        for field_i, field_v in ipairs(v) do
+          -- print("NTFY: setting member#" .. tostring(field_i) .. " to value " .. vpp.dump(field_v))
+          local field_len, field_val = lua2c(field.c_type, field_v, dst[k][field_i-1])
+          len = len + field_len
+        end
+      else
+        len, val = lua2c(field.c_type, v, dst[k])
+      end
+      if not field.array then
+        dst[k] = val
+      else
+        if 0 == field.array then
+          additional_len = additional_len + len
+          -- print("Adding " .. tostring(len) .. " bytes due to field " .. tostring(field.name))
+          -- If there is a variable storing the length
+          -- and the input table does not set it, do magic
+          if field.array_size and not src[field.array_size] then
+            local size_field = fields_info[field.array_size]
+            if size_field then
+              dst[field.array_size] = vpp.t_c2lua[size_field.c_type](size_field.c_type, len)
+            end
+          end
+        end
+      end
+      -- print("Full message:\n" .. vpp.hex_dump(ffi.string(ffi.cast('void *', req_store_cache), 64)))
+    end
+    return (ffi.sizeof(dst[0])+additional_len)
+  end
+
+  vpp.t_c2lua["__MSG__"] = function(c_type, src_ptr, src_len)
+    local out = {}
+    local reply_typed_ptr = ffi.cast(c_type .. " *", src_ptr)
+    local field_desc = vpp.c_type_to_fields[c_type]
+    if src_len then
+      for i = 0,src_len-1 do
+        out[i+1] = vpp.t_c2lua[c_type](c_type, src_ptr[i])
+      end
+      return out
+    end
+
+    for k, v in pairs(field_desc) do
+      local v_c2lua = vpp.t_c2lua[v.c_type]
+      if v_c2lua then
+        local len = v.array
+        -- print(dump(v))
+        if len then
+          local len_field_name = k .. "_length"
+          local len_field = field_desc[len_field_name]
+          if (len_field) then
+            local real_len = vpp.t_c2lua[len_field.c_type](len_field.c_type, reply_typed_ptr[len_field_name])
+            out[k] =  v_c2lua(v.c_type, reply_typed_ptr[k], real_len)
+          elseif len == 0 then
+            -- check if len = 0, then must be a field which contains the size
+            len_field =  field_desc[v.array_size]
+            local real_len = vpp.t_c2lua[len_field.c_type](len_field.c_type, reply_typed_ptr[v.array_size])
+            -- print("REAL length: " .. vpp.dump(v) .. " : " .. tostring(real_len))
+            out[k] = v_c2lua(v.c_type, reply_typed_ptr[k], real_len)
+          else
+            -- alas, just stuff the entire array
+            out[k] = v_c2lua(v.c_type, reply_typed_ptr[k], len)
+          end
+        else
+          out[k] =  v_c2lua(v.c_type, reply_typed_ptr[k])
+        end
+      else
+        out[k] = ""
+      end
+      -- print(k, out[k])
+    end
+    return out
+  end
+
+  return vpp
+end
+
+function vpp.resolve_message_number(msgname)
+  local name = msgname .. "_" .. vpp.msg_name_to_crc[msgname]
+  local idx = vpp.pneum.pneum_get_msg_index(vpp.c_str(name))
+  if vpp.debug_dump then
+    print("Index for " .. tostring(name) .. " is " .. tostring(idx))
+  end
+  vpp.msg_name_to_number[msgname] = idx
+  vpp.msg_number_to_name[idx] = msgname
+  vpp.msg_number_to_type[idx] = "vl_api_" .. msgname .. "_t"
+  vpp.msg_number_to_pointer_type[idx] = vpp.msg_number_to_type[idx] .. " *"
+  ffi.cdef("\n\n enum { vl_msg_" .. msgname .. " = " .. idx .. " };\n\n")
+end
+
+function vpp.connect(vpp, client_name)
+    local name = "lua_client"
+    if client_name then
+      name = client_name
+    end
+    local ret = vpp.pneum.pneum_connect(vpp.c_str(client_name), nil, nil)
+    if tonumber(ret) == 0 then
+      vpp.is_connected = true
+    end
+    for k, v in pairs(vpp.msg_name_to_number) do
+      vpp.resolve_message_number(k)
+    end
+  end
+
+function vpp.disconnect(vpp)
+    vpp.pneum.pneum_disconnect()
+  end
+
+function vpp.json_api(vpp, path, plugin_name)
+    -- print("Consuming the VPP api from "..path)
+    local ffii = {}
+    local f = io.open(path, "r")
+    if not f then
+      print("Could not open " .. path)
+      return nil
+    end
+    local data = f:read("*all")
+    local json = json.parse(data)
+    if not (json.types or json.messages) then
+      print("Can not parse " .. path)
+      return nil
+    end
+
+    local all_types = {}
+
+    for i, v in ipairs(json.types) do
+      table.insert(all_types, { typeonly = 1, desc = v })
+    end
+    for i, v in ipairs(json.messages) do
+      table.insert(all_types, { typeonly = 0, desc = v })
+    end
+    for i, v in ipairs(all_types) do
+      local typeonly = v.typeonly
+      local name = v.desc[1]
+      local c_type = "vl_api_" .. name .. "_t"
+
+      local fields = {}
+      -- vpp.msg_name_to_fields[name] = fields
+      -- print("CTYPE " .. c_type)
+      vpp.c_type_to_fields[c_type] = fields
+      vpp.t_lua2c[c_type] = vpp.t_lua2c["__MSG__"]
+      vpp.t_c2lua[c_type] = vpp.t_c2lua["__MSG__"]
+
+      local cdef = { "\n\n#pragma pack(1)\ntypedef struct _vl_api_", name, " {\n" }
+      for ii, vv in ipairs(v.desc) do
+        if type(vv) == "table" then
+          if vv.crc then
+            vpp.msg_name_to_crc[name] = string.sub(vv.crc, 3) -- strip the leading 0x
+          else
+            local fieldtype = vv[1]
+            local fieldname = vv[2]
+            local fieldcount = vv[3]
+            local fieldcountvar = vv[4]
+            local fieldrec = { name = fieldname, c_type = fieldtype, array = fieldcount, array_size = fieldcountvar }
+            if fieldcount then
+              table.insert(cdef, "  " .. fieldtype .. " " .. fieldname .. "[" .. fieldcount .. "];\n")
+              if fieldtype == "u8" then
+                -- any array of bytes is treated as a string
+              elseif vpp.t_lua2c[fieldtype] then
+                -- print("Array of " .. fieldtype .. " is ok!")
+              else
+                print("Unknown array type: ", name,  " : " , fieldname, " : ", fieldtype, ":", fieldcount, ":", fieldcountvar)
+              end
+            else
+              table.insert(cdef, "  " .. fieldtype .. " " .. fieldname .. ";\n")
+            end
+            fields[fieldname] = fieldrec
+          end
+        end
+      end
+
+      table.insert(cdef, "} vl_api_" .. name .. "_t;")
+      table.insert(ffii, table.concat(cdef))
+
+      if typeonly == 0 then
+        -- we will want to resolve this later
+        if vpp.debug_dump then
+          print("Remember to resolve " .. name)
+        end
+        vpp.msg_name_to_number[name] = -1
+        if vpp.is_connected then
+          vpp.resolve_message_number(name)
+        end
+      end
+
+    end
+    local cdef_full = table.concat(ffii)
+    ffi.cdef(cdef_full)
+end
+
+function vpp.consume_api(vpp, path, plugin_name)
+    -- print("Consuming the VPP api from "..path)
+    local ffii = {}
+    local f = io.open(path, "r")
+    if not f then
+      print("Could not open " .. path)
+      return nil
+    end
+    local data = f:read("*all")
+    -- Remove all C comments
+    data = data:gsub("/%*.-%*/", "")
+    if vpp.is_connected and not plugin_name then
+      print(path .. ": must specify plugin name!")
+      return
+    end
+    if plugin_name then
+      vpp.plugin_version[plugin_name] = vpp.crc_version_string(data)
+      local full_plugin_name = plugin_name .. "_" .. vpp.plugin_version[plugin_name]
+      local reply = vpp:api_call("get_first_msg_id", { name = full_plugin_name } )
+      vpp.next_msg_num = tonumber(reply[1].first_msg_id)
+      print("Plugin " .. full_plugin_name .. " first message is " .. tostring(vpp.next_msg_num))
+    end
+    -- print ("data len: ", #data)
+    data = data:gsub("\n(.-)(%S+)%s*{([^}]*)}", function (preamble, name, members)
+      local _, typeonly = preamble:gsub("typeonly", "")
+      local maybe_msg_id_field = { [0] = "u16 _vl_msg_id;", "" }
+      local onedef = "\n\n#pragma pack(1)\ntypedef struct _vl_api_"..name.. " {\n" ..
+	   -- "   u16 _vl_msg_id;" ..
+           maybe_msg_id_field[typeonly] ..
+	   members:gsub("%[[a-zA-Z_]+]", "[0]") ..
+	   "} vl_api_" .. name .. "_t;"
+
+      local c_type = "vl_api_" .. name .. "_t"
+
+      local fields = {}
+      -- vpp.msg_name_to_fields[name] = fields
+      -- print("CTYPE " .. c_type)
+      vpp.c_type_to_fields[c_type] = fields
+      vpp.t_lua2c[c_type] = vpp.t_lua2c["__MSG__"]
+      vpp.t_c2lua[c_type] = vpp.t_c2lua["__MSG__"]
+      local mirec = { name = "_vl_msg_id", c_type = "u16", array = nil, array_size = nil }
+      if typeonly == 0 then
+        fields[mirec.name] = mirec
+      end
+
+      -- populate the field reflection table for the message
+      -- sets the various type information as well as the accessors for lua<->C conversion
+      members:gsub("(%S+)%s+(%S+);", function (fieldtype, fieldname)
+          local fieldcount = nil
+          local fieldcountvar = nil
+          -- data = data:gsub("%[[a-zA-Z_]+]", "[0]")
+          fieldname = fieldname:gsub("(%b[])", function(cnt)
+              fieldcount = tonumber(cnt:sub(2, -2));
+              if not fieldcount then
+                fieldcount = 0
+                fieldcountvar = cnt:sub(2, -2)
+              end
+              return ""
+            end)
+	  local fieldrec = { name = fieldname, c_type = fieldtype, array = fieldcount, array_size = fieldcountvar }
+          if fieldcount then
+            if fieldtype == "u8" then
+              -- any array of bytes is treated as a string
+            elseif vpp.t_lua2c[fieldtype] then
+              -- print("Array of " .. fieldtype .. " is ok!")
+            else
+              print("Unknown array type: ", name,  " : " , fieldname, " : ", fieldtype, ":", fieldcount, ":", fieldcountvar)
+            end
+          end
+	  fields[fieldname] = fieldrec
+	end)
+
+      -- print(dump(fields))
+
+      if typeonly == 0 then
+	local this_message_number = vpp.next_msg_num
+	vpp.next_msg_num = vpp.next_msg_num + 1
+	vpp.msg_name_to_number[name] = this_message_number
+	vpp.msg_number_to_name[this_message_number] = name
+	vpp.msg_number_to_type[this_message_number] = "vl_api_" .. name .. "_t"
+	vpp.msg_number_to_pointer_type[this_message_number] = vpp.msg_number_to_type[this_message_number] .. " *"
+	onedef = onedef .. "\n\n enum { vl_msg_" .. name .. " = " .. this_message_number .. " };\n\n"
+      end
+      table.insert(ffii, onedef);
+      return "";
+      end)
+    local cdef = table.concat(ffii)
+    -- print(cdef)
+    ffi.cdef(cdef)
+  end
+
+
+function vpp.lua2c(vpp, c_type, src, dst_c_ptr)
+  -- returns the number of bytes written to memory pointed by dst
+  local lua2c = vpp.t_lua2c[c_type]
+  if lua2c then
+    return(lua2c(c_type, src, dst_c_ptr))
+  else
+    print("vpp.lua2c: do not know how to store type " .. tostring(c_type))
+    local x = "a" .. nil
+    return 0
+  end
+end
+
+function vpp.c2lua(vpp, c_type, src_ptr, src_len)
+  -- returns the lua data structure
+  local c2lua = vpp.t_c2lua[c_type]
+  if c2lua then
+    return(c2lua(c_type, src_ptr, src_len))
+  else
+    print("vpp.c2lua: do not know how to load type " .. c_type)
+    return nil
+  end
+end
+
+local req_store_cache = ffi.new("vl_api_opaque_message_t[1]")
+
+function vpp.api_write(vpp, api_name, req_table)
+    local msg_num = vpp.msg_name_to_number[api_name]
+    if not msg_num then
+      print ("API call "..api_name.." is not known")
+      return nil
+    end
+
+    if not req_table then
+      req_table = {}
+    end
+    req_table._vl_msg_id = msg_num
+
+    local packed_len = vpp:lua2c(vpp.msg_number_to_type[msg_num], req_table, req_store_cache)
+    if vpp.debug_dump then
+      print("Write Message length: " .. tostring(packed_len) .. "\n" .. vpp.hex_dump(ffi.string(ffi.cast('void *', req_store_cache), packed_len)))
+    end
+
+    res = vpp.pneum.pneum_write(ffi.cast('void *', req_store_cache), packed_len)
+    return res
+  end
+
+local rep_store_cache = ffi.new("vl_api_opaque_message_t *[1]")
+local rep_len_cache = ffi.new("int[1]")
+
+function vpp.api_read(vpp)
+    local rep_type = "vl_api_opaque_message_t"
+    local rep = rep_store_cache
+    local replen = rep_len_cache
+    res = vpp.pneum.pneum_read(ffi.cast("void *", rep), replen)
+    if vpp.debug_dump then
+      print("Read Message length: " .. tostring(replen[0]) .. "\n" .. vpp.hex_dump(ffi.string(ffi.cast('void *', rep[0]), replen[0])))
+    end
+
+    local reply_msg_num = ffi.C.ntohs(rep[0]._vl_msg_id)
+    local reply_msg_name = vpp.msg_number_to_name[reply_msg_num]
+
+    local reply_typed_ptr = ffi.cast(vpp.msg_number_to_pointer_type[reply_msg_num], rep[0])
+    local out = vpp:c2lua(vpp.msg_number_to_type[reply_msg_num], rep[0], nil, replen[0])
+    if type(out) == "table" then
+      out["luaapi_message_name"] = reply_msg_name
+    end
+
+    vpp.pneum.pneum_free(ffi.cast('void *',rep[0]))
+
+    return reply_msg_name, out
+  end
+
+function vpp.api_call(vpp, api_name, req_table, options_in)
+    local msg_num = vpp.msg_name_to_number[api_name]
+    local end_message_name = api_name .."_reply"
+    local replies = {}
+    local cstruct = ""
+    local options = options_in or {}
+    if msg_num then
+      if vpp.debug_dump then
+        print("Message #" .. tostring(msg_num) .. " for name " .. tostring(api_name))
+      end
+      vpp:api_write(api_name, req_table)
+      if not vpp.msg_name_to_number[end_message_name] or options.force_ping then
+        end_message_name = "control_ping_reply"
+        vpp:api_write("control_ping")
+      end
+      repeat
+        reply_message_name, reply = vpp:api_read()
+        if reply and not reply.context then
+          -- there may be async events inbetween
+          table.insert(vpp.events, reply)
+        else
+          if reply_message_name ~= "control_ping_reply" then
+            -- do not insert the control ping encapsulation
+            table.insert(replies, reply)
+          end
+        end
+        -- print(reply)
+      until reply_message_name == end_message_name
+    else
+      print(api_name .. " is an unknown API call")
+      return nil
+    end
+    return replies
+  end
+
+return vpp
diff --git a/src/vpp.am b/src/vpp.am
index a02206f8..1d07b90a 100644
--- a/src/vpp.am
+++ b/src/vpp.am
@@ -74,7 +74,7 @@ PLUGIN_DPDK_ARG=""
 endif
 
 vpp_plugin_configure:
-	@echo "PLUGIN CONFIGURE " $@ 
+	@echo "PLUGIN CFG" $@
 	@echo "#!/bin/bash" > $@
 	@echo " " >> $@
 	@echo "set +eu" >> $@
diff --git a/src/vppapigen.am b/src/vppapigen.am
index 083ee537..edde339d 100644
--- a/src/vppapigen.am
+++ b/src/vppapigen.am
@@ -16,9 +16,9 @@ bin_PROGRAMS += vppapigen
 BUILT_SOURCES += tools/vppapigen/gram.h
 
 tools/vppapigen/gram.h: tools/vppapigen/gram.y
-	$(YACC) -d @srcdir@/tools/vppapigen/gram.y
-	mv y.tab.h tools/vppapigen/gram.h
-	rm y.tab.c
+	@$(YACC) -d @srcdir@/tools/vppapigen/gram.y
+	@mv y.tab.h tools/vppapigen/gram.h
+	@rm y.tab.c
 
 vppapigen_SOURCES = tools/vppapigen/gram.y tools/vppapigen/lex.c tools/vppapigen/node.c
 vppapigen_LDADD = libvppinfra.la
diff --git a/vpp-api/Makefile.am b/vpp-api/Makefile.am
deleted file mode 100644
index b50522ef..00000000
--- a/vpp-api/Makefile.am
+++ /dev/null
@@ -1,4 +0,0 @@
-AUTOMAKE_OPTIONS = foreign
-SUBDIRS = java
-
-# vi:syntax=automake
diff --git a/vpp-api/configure.ac b/vpp-api/configure.ac
deleted file mode 100644
index 278041c7..00000000
--- a/vpp-api/configure.ac
+++ /dev/null
@@ -1,12 +0,0 @@
-AC_INIT(vpp-api, 1.1.0)
-LT_INIT
-AC_CONFIG_MACRO_DIR([m4])
-AC_CONFIG_SUBDIRS([java])
-AM_INIT_AUTOMAKE
-AM_SILENT_RULES([yes])
-
-AC_PROG_CC
-
-AC_CONFIG_FILES([Makefile])
-AC_OUTPUT
-
diff --git a/vpp-api/java/Makefile.am b/vpp-api/java/Makefile.am
deleted file mode 100644
index b0551965..00000000
--- a/vpp-api/java/Makefile.am
+++ /dev/null
@@ -1,116 +0,0 @@
-# Copyright (c) 2015 Cisco and/or its affiliates.
-# Licensed under the Apache License, Version 2.0 (the "License");
-# you may not use this file except in compliance with the License.
-# You may obtain a copy of the License at:
-#
-#     http://www.apache.org/licenses/LICENSE-2.0
-#
-# Unless required by applicable law or agreed to in writing, software
-# distributed under the License is distributed on an "AS IS" BASIS,
-# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-# See the License for the specific language governing permissions and
-# limitations under the License.
-
-AUTOMAKE_OPTIONS = foreign subdir-objects
-ACLOCAL_AMFLAGS = -I m4
-AM_CFLAGS = -Wall
-
-noinst_PROGRAMS =
-BUILT_SOURCES =
-bin_PROGRAMS =
-CLEANFILES =
-lib_LTLIBRARIES =
-
-#
-# jvpp-common
-#
-
-nobase_include_HEADERS =		\
-  jvpp-common/jvpp_common.h
-
-lib_LTLIBRARIES += libjvpp_common.la
-
-libjvpp_common_la_SOURCES = jvpp-common/jvpp_common.c
-libjvpp_common_la_LIBADD = -lvlibmemoryclient -lvlibapi -lsvm -lvppinfra \
-	              -lpthread -lm -lrt
-libjvpp_common_la_LDFLAGS = -module
-libjvpp_common_la_CPPFLAGS = -I$(JAVA_HOME)/include -I$(JAVA_HOME)/include/linux
-
-
-#
-# jvpp-registry (connection management + plugin registry)
-#
-lib_LTLIBRARIES += libjvpp_registry.la
-
-libjvpp_registry_la_SOURCES = jvpp-registry/jvpp_registry.c
-libjvpp_registry_la_LIBADD = -lvlibmemoryclient -lvlibapi -lsvm -lvppinfra \
-	              -lpthread -lm -lrt -ljvpp_common
-libjvpp_registry_la_LDFLAGS = -module
-libjvpp_registry_la_CPPFLAGS = -Ijvpp-registry -I$(JAVA_HOME)/include -I$(JAVA_HOME)/include/linux
-EXTRA_libjvpp_registry_la_DEPENDENCIES=libjvpp_common.la
-
-jarfile_jvpp_registry = jvpp-registry-$(PACKAGE_VERSION).jar
-packagedir_jvpp_registry = io/fd/vpp/jvpp
-
-BUILT_SOURCES += jvpp-registry/io_fd_vpp_jvpp_VppJNIConnection.h
-
-jvpp_registry_src_files := $(wildcard @srcdir@/jvpp-registry/$(packagedir_jvpp_registry)/*.java) $(wildcard @srcdir@/jvpp-registry/$(packagedir_jvpp_registry)/**/*.java)
-
-jvpp-registry/io_fd_vpp_jvpp_VppJNIConnection.h: $(jvpp_registry_src_files)
-	@echo " jvpp-registry.jar generation"
-	mkdir -p jvpp-registry/target
-	$(JAVAC) -d jvpp-registry/target $(jvpp_registry_src_files)
-	$(JAVAH) -force -classpath jvpp-registry/target -d jvpp-registry io.fd.vpp.jvpp.VppJNIConnection
-	$(JAVAH) -force -classpath jvpp-registry/target -d jvpp-registry io.fd.vpp.jvpp.JVppRegistryImpl
-
-$(jarfile_jvpp_registry): libjvpp_registry.la
-	cp .libs/libjvpp_registry.so.0.0.0 jvpp-registry/target;	\
-	$(JAR) cf $(JARFLAGS) $@ -C jvpp-registry/target .;
-
-#
-# jvpp-core (Java wrapper for vpe.api)
-#
-lib_LTLIBRARIES += libjvpp_core.la
-
-libjvpp_core_la_SOURCES = jvpp-core/jvpp_core.c jvpp-core/jvpp_core_gen.h
-libjvpp_core_la_LIBADD = -lvlibmemoryclient -lvlibapi -lsvm -lvppinfra \
-	              -lpthread -lm -lrt -ljvpp_common
-libjvpp_core_la_LDFLAGS = -module
-libjvpp_core_la_CPPFLAGS = -I$(JAVA_HOME)/include -I$(JAVA_HOME)/include/linux
-
-jarfile_jvpp_core = jvpp-core-$(PACKAGE_VERSION).jar
-packagedir_jvpp_core = io/fd/vpp/jvpp/core
-
-BUILT_SOURCES += jvpp-core/io_fd_vpp_jvpp_core_JVppCoreImpl.h
-
-JSON_FILES = \
-	$(wildcard $(prefix)/../vpp/share/vpp/api/core/*.api.json)
-
-jvpp-core/io_fd_vpp_jvpp_core_JVppCoreImpl.h:		\
-    jvpp-registry/io_fd_vpp_jvpp_VppJNIConnection.h	\
-    $(JSON_FILES)
-	cp -rf @srcdir@/jvpp-core/* -t jvpp-core/
-	mkdir -p jvpp-core/target
-	cd jvpp-core \
-		&& mkdir -p types dto future callfacade callback notification \
-		&& @srcdir@/jvpp/gen/jvpp_gen.py  --plugin_name core \
-		    -i $(JSON_FILES) \
-		&& cp -rf types dto future callfacade callback notification *.java -t $(packagedir_jvpp_core) \
-		&& rm -rf types dto future callfacade callback notification *.java
-
-	$(JAVAC) -classpath jvpp-registry/target -d jvpp-core/target jvpp-core/$(packagedir_jvpp_core)/*.java \
-		jvpp-core/$(packagedir_jvpp_core)/types/*.java \
-		jvpp-core/$(packagedir_jvpp_core)/dto/*.java \
-		jvpp-core/$(packagedir_jvpp_core)/callback/*.java \
-		jvpp-core/$(packagedir_jvpp_core)/notification/*.java \
-		jvpp-core/$(packagedir_jvpp_core)/future/*.java 	\
-		jvpp-core/$(packagedir_jvpp_core)/callfacade/*.java 	\
-		jvpp-core/$(packagedir_jvpp_core)/test/*.java 	\
-		|| (echo "jvpp-core compilation failed: $$?"; exit 1)
-	$(JAVAH) -force -classpath jvpp-registry/target:jvpp-core/target -d jvpp-core io.fd.vpp.jvpp.core.JVppCoreImpl
-
-$(jarfile_jvpp_core): libjvpp_core.la
-	cp .libs/libjvpp_core.so.0.0.0 jvpp-core/target
-	$(JAR) cf $(JARFLAGS) $@ -C jvpp-core/target .
-
-all-local: $(jarfile_jvpp_registry) $(jarfile_jvpp_core)
diff --git a/vpp-api/java/Readme.txt b/vpp-api/java/Readme.txt
deleted file mode 100644
index 689b9b37..00000000
--- a/vpp-api/java/Readme.txt
+++ /dev/null
@@ -1,236 +0,0 @@
-= JVpp
-
-JVpp is JNI based Java API for VPP.
-
-== Features
-It is:
-
-* Asynchronous
-* Fully generated
-* Lightweight
-
-== Architecture
-
-=== Plugin support
-
-  /-------------\        /--------------\          /---------------\
-  | JvppPlugin1 +<-------+ JVppRegistry +--------->+ VppConnection |
-  \-------------/  inits \--+-----------/   uses   \---------------/
-                            |
-  /-------------\           |
-  | JvppPlugin2 +<----------+ inits
-  \-------------/           |
-                            |
-  ...                       |
-                            |
-  /----------\              |
-  | JVppCore +<-------------+
-  \----------/
-
-
-VppRegistry opens connection to vpp (VppConnection) and manages jvpp plugins.
-Each plugin needs to be registered in the VppRegistry. Registration involves
-plugin initialization (providing JNI implementation with JVppCallback reference,
-vpp client identifier and vpp shared memory queue address).
-
-API user sends message by calling a method of appropriate plugin interface.
-The call is delegated to JNI implementation provided by the particular plugin.
-When JNI code receives reply, it invokes callback method of JVppCallback
-that corresponds to the received message reply.
-
-=== JVppCore as an example of JVpp plugin architecture
-
- JVpp Java
-
-  /--------------\             /----------\          /------------\    /------\
-  | JVppRegistry |             | JVppCore |          |  Callbacks |    | DTOs |
-  \----+---------/             \----+-----/          \------+-----/    \------/
-       ^                            ^                       ^
-       | implements                 | implements            | implements
-  /----+--------------\         /---+----------\      /-----+---------\
-  | JVppRegistryImpl* +-------->+ JVppCoreImpl |      |  JVppCallback |
-  \-------+-----------/  inits  \---+----------/      \-------+-------/
-          |                          |                       ^
-          |                          | uses                  | calls back
-          |                          |                       |
-----------|--------------------------|-----------------------|---------------------
-          |                          |                       |
- C JNI    |                          +-------------------+   |       /-----------------\
-          v                                              |   |   +-->+ jvpp_core_gen.h |
-  /--------+--------\                                    |   |   |   \-----------------/
-  | jpp_registry.c* +---+   /--------+----+----\         |   |   |
-  \-----------------/   |   | << shared lib >> |        /-+--+---+------\
-                        + ->+   jvpp_common*   <--------+  jvpp_core.c* |
-                      uses  \------------------/  uses  \---------------/
-
-
-* Components marked with an asterisk contain manually crafted code, which in addition
-to generated classes form jvpp. Exception applies to Callbacks and DTOs, since there are
-manually crafted marker interfaces in callback and dto package (dto/JVppRequest, dto/JVppReply,
-dto/JVppDump, dto/JVppReplyDump, callback/JVppCallback)
-
-Note: jvpp_core.c calls back the JVppCallback instance with every response. An instance of the
-JVppCallback is provided to jvpp_core.c by JVppRegistryImpl on JVppCoreImpl initialization.
-
-Part of the JVpp is also Future facade. It is asynchronous API returning Future objects
-on top of low level JVpp. It wraps dump reply messages in one DTO using control_ping message
-(provided by JVppRegistry).
-
-
-Future facade
-
-        /----------------\          /---------------\
-        | FutureJVppCore |      +-->+ JVppRegistry* |
-        \-----+----------/      |   \---------------/
-              ^                 |
-              | implements      | uses
-              |                 |
-     /--------+-------------\   |    /------------------------------\
-     | FutureJVppCoreFacade +---+--->+ FutureJVppCoreFacadeCallback |
-     \---------+------------/  uses  \-------+----------------------/
-               |                             |
----------------|-----------------------------|-------------------------------
-               | uses                        | implements
-JVpp Java      |                             |
-               |                             |
- /----------\  |                             |
- | JVppCore +<-+                             |
- \----+-----/                                |
-      ^                                      |
-      | implements                           v
- /----+---------\                   /--------+---------------\
- | JVppCoreImpl |                   | JVppCoreGlobalCallback |
- \--------------/                   \------------------------/
-
-
-
-Another useful utility of the JVpp is Callback facade. It is asynchronous API
-capable of calling specific callback instance (provided when performing a call)
-per call.
-
-
-Callback facade
-
-        /------------------\          /---------------\
-        | CallbackJVppCore |      +-->+ JVppRegistry* |
-        \-----+------------/      |   \---------------/
-              ^                   |
-              | implements        | uses
-              |                   |
-     /--------+---------------\   |    /--------------------------\
-     | CallbackJVppCoreFacade +---+--->+ CallbackJVppCoreCallback |
-     \---------+--------------/  uses  \-----+--------------------/
-               |                             |
----------------|-----------------------------|-------------------------------
-               | uses                        | implements
-JVpp Java      |                             |
-               |                             |
- /----------\  |                             |
- | JVppCore +<-+                             |
- \----+-----/                                |
-      ^                                      |
-      | implements                           v
- /----+---------\                   /----------+-------------\
- | JVppCoreImpl |                   | JVppCoreGlobalCallback |
- \--------------/                   \------------------------/
-
-
-== Package structure
-
-* *io.fd.vpp.jvpp* - top level package for generated JVpp interface+ implementation and hand-crafted
-VppConnection interface + implementation - packaged as jvpp-registry-version.jar
-
-* *io.fd.vpp.jvpp.[plugin]* - top level package for generated JVpp interface + implementation
-+ plugin's API tests - packaged as jvpp-[plugin]-version.jar
-
-** *dto* - package for DTOs generated from VPP API structures + base/marker hand-crafted interfaces
-(in case of jvpp-registry)
-** *callback* - package for low-level JVpp callbacks and a global callback interface implementing each of
-the low-level JVppcallbacks
-** *future* - package for future based facade on top of JVpp and callbacks
-** *callfacade* - package for callback based facade on top of JVpp and callbacks. Allowing
-users to provide callback per request
-** *test* - package for JVpp standalone tests. Can also serve as samples for JVpp.
-
-C code is structured into modules:
-
-* *jvpp_common* - shared library that provides jvpp_main_t reference used by jvpp_registry and plugins.
-
-* *jvpp_registry* - native library used by JVppRegistryImpl, responsible for:
-
-** VPP connection open/close
-** Rx thread to java thread attach
-** control ping message handling
-
-* *jvpp_core* - native library used by jvpp core plugin:
-** *jvpp_core.c* - contains hand crafted code for core plugin initialization
-** *jvpp_core_gen.h* - contains generated JNI compatible handlers for all requests and replies defined in vpe.api
-
-== Code generators
-All of the required code except the base/marker interfaces is generated using
-simple python2 code generators. The generators use __defs_vpp_papi.py__ input
-file produced by __vppapigen__ from vpe.api file.
-
-=== JNI compatible C code
-Produces __jvpp_[plugin]_gen.h__ file containing JNI compatible handlers for each VPP
-request and reply.
-
-[NOTE]
-====
-Source: jvpp_c_gen.py
-====
-
-=== Request/Reply DTOs
-For all the structures in __defs_vpp_papi.py__ a POJO DTO is produced. Logically,
-there are 4 types of DTOs:
-
-* Request - requests that can be sent to VPP and only a single response is expected
-* DumpRequest - requests that can be sent to VPP and a stream of responses is expected
-* Reply - reply to a simple request or a single response from dump triggered response stream
-* ReplyDump - collection of replies from a single dump request
-* Notifications/Events - Not implemented yet
-
-[NOTE]
-====
-Source: dto_gen.py
-====
-
-=== JVpp
-Produces __JVpp.java__ and __JVppImpl.java__. This is the layer right above JNI compatible C
-code.
-
-[NOTE]
-====
-Source: jvpp_impl_gen.py
-====
-
-=== Callbacks
-Produces callback interface for each VPP reply + a global callback interface called
-__JVpp[plugin]GlobalCallback.java__ aggregating all of the callback interfaces. The JNI
-compatible C code expects only a single instance of this global callback and calls
-it with every reply.
-
-[NOTE]
-====
-Source: callback_gen.py
-====
-
-=== Future facade
-Produces an asynchronous facade on top of JVpp and callbacks, which returns a Future that provides
-matching reply once VPP invocation finishes. Sources produced:
-__FutureJVpp[plugin].java, FutureJVpp[plugin]Facade.java and FutureJVpp[plugin]Callback.java__
-
-[NOTE]
-====
-Source: jvpp_future_facade_gen.py
-====
-
-=== Callback facade
-Similar to future facade, only this facade takes callback objects as part of the invocation
-and the callback is called with result once VPP invocation finishes. Sources produced:
-__CallbackJVpp[plugin].java, CallbackJVpp[plugin]Facade.java and CallbackJVpp[plugin]Callback.java__
-
-[NOTE]
-====
-Source: jvpp_callback_facade_gen.py
-====
diff --git a/vpp-api/java/configure.ac b/vpp-api/java/configure.ac
deleted file mode 100644
index 3082781e..00000000
--- a/vpp-api/java/configure.ac
+++ /dev/null
@@ -1,24 +0,0 @@
-AC_INIT(jvpp, 17.04)
-LT_INIT
-AC_CONFIG_MACRO_DIR([m4])
-AM_INIT_AUTOMAKE
-AM_SILENT_RULES([yes])
-
-AC_PROG_CC
-
-if test -f /usr/bin/lsb_release  && test `lsb_release -si` == "Ubuntu"  && test `lsb_release -sr` == "14.04"  && test -d /usr/lib/jvm/java-8-openjdk-amd64/ ; then
-    JAVA_HOME=/usr/lib/jvm/java-8-openjdk-amd64/
-    JAVAC=${JAVA_HOME}/bin/javac
-    PATH=${JAVA_HOME}/bin/:${PATH}
-    break
-fi
-
-AX_CHECK_JAVA_HOME
-AX_PROG_JAVAC
-AX_PROG_JAVAH
-AX_PROG_JAR
-AX_PROG_JAVADOC
-AX_PROG_JAVA
-
-AC_OUTPUT([Makefile])
-
diff --git a/vpp-api/java/jvpp-common/jvpp_common.c b/vpp-api/java/jvpp-common/jvpp_common.c
deleted file mode 100644
index a161c09c..00000000
--- a/vpp-api/java/jvpp-common/jvpp_common.c
+++ /dev/null
@@ -1,65 +0,0 @@
-/*
- * Copyright (c) 2016 Cisco and/or its affiliates.
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at:
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-#define _GNU_SOURCE /* for strcasestr(3) */
-
-#include "jvpp_common.h"
-
-#ifndef JVPP_DEBUG
-#define JVPP_DEBUG 0
-#endif
-
-#if JVPP_DEBUG == 1
-#define DEBUG_LOG(...) clib_warning(__VA_ARGS__)
-#else
-#define DEBUG_LOG(...)
-#endif
-
-/* shared jvpp main structure */
-jvpp_main_t jvpp_main __attribute__((aligned (64)));
-
-void call_on_error(const char* callName, int contextId, int retval,
-        jclass callbackClass, jobject callbackObject,
-        jclass callbackExceptionClass) {
-    DEBUG_LOG("\nCallOnError : callback=%s, retval=%d, context=%d\n", callName,
-            clib_net_to_host_u32(retval), clib_net_to_host_u32(context));
-    JNIEnv *env = jvpp_main.jenv;
-    if (!callbackClass) {
-        DEBUG_LOG("CallOnError : jm->callbackClass is null!\n");
-        return;
-    }
-    jmethodID excConstructor = (*env)->GetMethodID(env, callbackExceptionClass,
-            "", "(Ljava/lang/String;II)V");
-    if (!excConstructor) {
-        DEBUG_LOG("CallOnError : excConstructor is null!\n");
-        return;
-    }
-    jmethodID callbackExcMethod = (*env)->GetMethodID(env, callbackClass,
-            "onError", "(Lio/fd/vpp/jvpp/VppCallbackException;)V");
-    if (!callbackExcMethod) {
-        DEBUG_LOG("CallOnError : callbackExcMethod is null!\n");
-        return;
-    }
-
-    jobject excObject = (*env)->NewObject(env, callbackExceptionClass,
-            excConstructor, (*env)->NewStringUTF(env, callName),
-            clib_net_to_host_u32(contextId), clib_net_to_host_u32(retval));
-    if (!excObject) {
-        DEBUG_LOG("CallOnError : excObject is null!\n");
-        return;
-    }
-
-    (*env)->CallVoidMethod(env, callbackObject, callbackExcMethod, excObject);
-    DEBUG_LOG("CallOnError : Response sent\n");
-}
diff --git a/vpp-api/java/jvpp-common/jvpp_common.h b/vpp-api/java/jvpp-common/jvpp_common.h
deleted file mode 100644
index bbb203ed..00000000
--- a/vpp-api/java/jvpp-common/jvpp_common.h
+++ /dev/null
@@ -1,67 +0,0 @@
-/*
- * Copyright (c) 2016 Cisco and/or its affiliates.
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at:
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-#ifndef __included_jvpp_common_h__
-#define __included_jvpp_common_h__
-//
-#include 
-#include 
-#include 
-#include 
-
-typedef struct {
-    /* Unique identifier used for matching replays with requests  */
-    volatile u32 context_id;
-
-    /* Spinlock */
-    volatile u32 lock;
-    u32 tag;
-
-    /* JNI Native Method Interface pointer for message handlers */
-    JNIEnv *jenv;
-
-    /* JNI Invoke Interface pointer for attachment of rx thread to java thread */
-    JavaVM *jvm;
-
-    /* Convenience */
-    unix_shared_memory_queue_t * vl_input_queue;
-    u32 my_client_index;
-} jvpp_main_t;
-
-extern jvpp_main_t jvpp_main __attribute__((aligned (64)));
-
-static_always_inline u32 vppjni_get_context_id(jvpp_main_t * jm) {
-    return __sync_add_and_fetch(&jm->context_id, 1);
-}
-
-static_always_inline void vppjni_lock(jvpp_main_t * jm, u32 tag) {
-    while (__sync_lock_test_and_set(&jm->lock, 1))
-        ;
-    jm->tag = tag;
-}
-
-static_always_inline void vppjni_unlock(jvpp_main_t * jm) {
-    jm->tag = 0;
-    CLIB_MEMORY_BARRIER();
-    jm->lock = 0;
-}
-
-/**
- * Calls onError callback on callbackObject reference. Passes instance of callbackExceptionClass as parameter.
- */
-void call_on_error(const char* callName, int contextId, int retval,
-        jclass callbackClass, jobject callbackObject,
-        jclass callbackExceptionClass);
-
-#endif /* __included_jvpp_common_h__ */
diff --git a/vpp-api/java/jvpp-core/io/fd/vpp/jvpp/core/test/CallbackApiTest.java b/vpp-api/java/jvpp-core/io/fd/vpp/jvpp/core/test/CallbackApiTest.java
deleted file mode 100644
index 986993b8..00000000
--- a/vpp-api/java/jvpp-core/io/fd/vpp/jvpp/core/test/CallbackApiTest.java
+++ /dev/null
@@ -1,96 +0,0 @@
-/*
- * Copyright (c) 2016 Cisco and/or its affiliates.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at:
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-package io.fd.vpp.jvpp.core.test;
-
-import io.fd.vpp.jvpp.JVpp;
-import io.fd.vpp.jvpp.JVppRegistry;
-import io.fd.vpp.jvpp.JVppRegistryImpl;
-import io.fd.vpp.jvpp.VppCallbackException;
-import io.fd.vpp.jvpp.core.JVppCoreImpl;
-import io.fd.vpp.jvpp.core.callback.GetNodeIndexCallback;
-import io.fd.vpp.jvpp.core.callback.ShowVersionCallback;
-import io.fd.vpp.jvpp.core.callback.SwInterfaceCallback;
-import io.fd.vpp.jvpp.core.dto.GetNodeIndex;
-import io.fd.vpp.jvpp.core.dto.GetNodeIndexReply;
-import io.fd.vpp.jvpp.core.dto.ShowVersion;
-import io.fd.vpp.jvpp.core.dto.ShowVersionReply;
-import io.fd.vpp.jvpp.core.dto.SwInterfaceDetails;
-import io.fd.vpp.jvpp.core.dto.SwInterfaceDump;
-
-public class CallbackApiTest {
-
-    public static void main(String[] args) throws Exception {
-        testCallbackApi();
-    }
-
-    private static void testCallbackApi() throws Exception {
-        System.out.println("Testing Java callback API with JVppRegistry");
-        try (final JVppRegistry registry = new JVppRegistryImpl("CallbackApiTest");
-             final JVpp jvpp = new JVppCoreImpl()) {
-            registry.register(jvpp, new TestCallback());
-
-            System.out.println("Sending ShowVersion request...");
-            final int result = jvpp.send(new ShowVersion());
-            System.out.printf("ShowVersion send result = %d%n", result);
-
-            System.out.println("Sending GetNodeIndex request...");
-            GetNodeIndex getNodeIndexRequest = new GetNodeIndex();
-            getNodeIndexRequest.nodeName = "non-existing-node".getBytes();
-            jvpp.send(getNodeIndexRequest);
-
-            System.out.println("Sending SwInterfaceDump request...");
-            SwInterfaceDump swInterfaceDumpRequest = new SwInterfaceDump();
-            swInterfaceDumpRequest.nameFilterValid = 0;
-            swInterfaceDumpRequest.nameFilter = "".getBytes();
-            jvpp.send(swInterfaceDumpRequest);
-
-            Thread.sleep(1000);
-            System.out.println("Disconnecting...");
-        }
-        Thread.sleep(1000);
-    }
-
-    static class TestCallback implements GetNodeIndexCallback, ShowVersionCallback, SwInterfaceCallback {
-
-        @Override
-        public void onGetNodeIndexReply(final GetNodeIndexReply msg) {
-            System.out.printf("Received GetNodeIndexReply: %s%n", msg);
-        }
-
-        @Override
-        public void onShowVersionReply(final ShowVersionReply msg) {
-            System.out.printf("Received ShowVersionReply: context=%d, program=%s, version=%s, "
-                    + "buildDate=%s, buildDirectory=%s%n",
-                msg.context, new String(msg.program), new String(msg.version),
-                new String(msg.buildDate), new String(msg.buildDirectory));
-        }
-
-        @Override
-        public void onSwInterfaceDetails(final SwInterfaceDetails msg) {
-            System.out.printf("Received SwInterfaceDetails: interfaceName=%s, l2AddressLength=%d, adminUpDown=%d, "
-                    + "linkUpDown=%d, linkSpeed=%d, linkMtu=%d%n",
-                new String(msg.interfaceName), msg.l2AddressLength, msg.adminUpDown,
-                msg.linkUpDown, msg.linkSpeed, (int) msg.linkMtu);
-        }
-
-        @Override
-        public void onError(VppCallbackException ex) {
-            System.out.printf("Received onError exception: call=%s, context=%d, retval=%d%n", ex.getMethodName(),
-                ex.getCtxId(), ex.getErrorCode());
-        }
-    }
-}
diff --git a/vpp-api/java/jvpp-core/io/fd/vpp/jvpp/core/test/CallbackJVppFacadeNotificationTest.java b/vpp-api/java/jvpp-core/io/fd/vpp/jvpp/core/test/CallbackJVppFacadeNotificationTest.java
deleted file mode 100644
index d84cb034..00000000
--- a/vpp-api/java/jvpp-core/io/fd/vpp/jvpp/core/test/CallbackJVppFacadeNotificationTest.java
+++ /dev/null
@@ -1,87 +0,0 @@
-/*
- * Copyright (c) 2016 Cisco and/or its affiliates.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at:
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-package io.fd.vpp.jvpp.core.test;
-
-import io.fd.vpp.jvpp.JVppRegistry;
-import io.fd.vpp.jvpp.JVppRegistryImpl;
-import io.fd.vpp.jvpp.VppCallbackException;
-import io.fd.vpp.jvpp.core.JVppCore;
-import io.fd.vpp.jvpp.core.JVppCoreImpl;
-import io.fd.vpp.jvpp.core.callback.WantInterfaceEventsCallback;
-import io.fd.vpp.jvpp.core.callfacade.CallbackJVppCoreFacade;
-import io.fd.vpp.jvpp.core.dto.WantInterfaceEventsReply;
-
-public class CallbackJVppFacadeNotificationTest {
-
-    private static void testCallbackFacade() throws Exception {
-        System.out.println("Testing CallbackJVppFacade for notifications");
-
-        try (final JVppRegistry registry = new JVppRegistryImpl("CallbackFacadeTest");
-             final JVppCore jvpp = new JVppCoreImpl()) {
-            final CallbackJVppCoreFacade jvppCallbackFacade = new CallbackJVppCoreFacade(registry, jvpp);
-            System.out.println("Successfully connected to VPP");
-
-            final AutoCloseable notificationListenerReg =
-                jvppCallbackFacade.getNotificationRegistry().registerSwInterfaceSetFlagsNotificationCallback(
-                    NotificationUtils::printNotification
-                );
-
-            jvppCallbackFacade.wantInterfaceEvents(NotificationUtils.getEnableInterfaceNotificationsReq(),
-                new WantInterfaceEventsCallback() {
-                    @Override
-                    public void onWantInterfaceEventsReply(final WantInterfaceEventsReply reply) {
-                        System.out.println("Interface events started");
-                    }
-
-                    @Override
-                    public void onError(final VppCallbackException ex) {
-                        System.out.printf("Received onError exception: call=%s, context=%d, retval=%d%n",
-                            ex.getMethodName(), ex.getCtxId(), ex.getErrorCode());
-                    }
-                });
-
-            System.out.println("Changing interface configuration");
-            NotificationUtils.getChangeInterfaceState().send(jvpp);
-
-            Thread.sleep(1000);
-
-            jvppCallbackFacade.wantInterfaceEvents(NotificationUtils.getDisableInterfaceNotificationsReq(),
-                new WantInterfaceEventsCallback() {
-                    @Override
-                    public void onWantInterfaceEventsReply(final WantInterfaceEventsReply reply) {
-                        System.out.println("Interface events stopped");
-                    }
-
-                    @Override
-                    public void onError(final VppCallbackException ex) {
-                        System.out.printf("Received onError exception: call=%s, context=%d, retval=%d%n",
-                            ex.getMethodName(), ex.getCtxId(), ex.getErrorCode());
-                    }
-                });
-
-            notificationListenerReg.close();
-
-            Thread.sleep(2000);
-            System.out.println("Disconnecting...");
-        }
-        Thread.sleep(1000);
-    }
-
-    public static void main(String[] args) throws Exception {
-        testCallbackFacade();
-    }
-}
diff --git a/vpp-api/java/jvpp-core/io/fd/vpp/jvpp/core/test/CallbackJVppFacadeTest.java b/vpp-api/java/jvpp-core/io/fd/vpp/jvpp/core/test/CallbackJVppFacadeTest.java
deleted file mode 100644
index 9f7cb8de..00000000
--- a/vpp-api/java/jvpp-core/io/fd/vpp/jvpp/core/test/CallbackJVppFacadeTest.java
+++ /dev/null
@@ -1,103 +0,0 @@
-/*
- * Copyright (c) 2016 Cisco and/or its affiliates.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at:
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-package io.fd.vpp.jvpp.core.test;
-
-import io.fd.vpp.jvpp.JVppRegistry;
-import io.fd.vpp.jvpp.JVppRegistryImpl;
-import io.fd.vpp.jvpp.VppCallbackException;
-import io.fd.vpp.jvpp.core.JVppCoreImpl;
-import io.fd.vpp.jvpp.core.callback.GetNodeIndexCallback;
-import io.fd.vpp.jvpp.core.callback.ShowVersionCallback;
-import io.fd.vpp.jvpp.core.callfacade.CallbackJVppCoreFacade;
-import io.fd.vpp.jvpp.core.dto.GetNodeIndex;
-import io.fd.vpp.jvpp.core.dto.GetNodeIndexReply;
-import io.fd.vpp.jvpp.core.dto.ShowVersionReply;
-
-/**
- * CallbackJVppFacade together with CallbackJVppFacadeCallback allow for setting different callback for each request.
- * This is more convenient than the approach shown in CallbackApiTest.
- */
-public class CallbackJVppFacadeTest {
-
-    private static ShowVersionCallback showVersionCallback1 = new ShowVersionCallback() {
-        @Override
-        public void onShowVersionReply(final ShowVersionReply msg) {
-            System.out.printf("ShowVersionCallback1 received ShowVersionReply: context=%d, program=%s,"
-                    + "version=%s, buildDate=%s, buildDirectory=%s%n", msg.context, new String(msg.program),
-                new String(msg.version), new String(msg.buildDate), new String(msg.buildDirectory));
-        }
-
-        @Override
-        public void onError(VppCallbackException ex) {
-            System.out.printf("Received onError exception in showVersionCallback1: call=%s, reply=%d, context=%d%n",
-                ex.getMethodName(), ex.getErrorCode(), ex.getCtxId());
-        }
-    };
-
-    private static ShowVersionCallback showVersionCallback2 = new ShowVersionCallback() {
-        @Override
-        public void onShowVersionReply(final ShowVersionReply msg) {
-            System.out.printf("ShowVersionCallback2 received ShowVersionReply: context=%d, program=%s,"
-                    + "version=%s, buildDate=%s, buildDirectory=%s%n", msg.context, new String(msg.program),
-                new String(msg.version), new String(msg.buildDate), new String(msg.buildDirectory));
-        }
-
-        @Override
-        public void onError(VppCallbackException ex) {
-            System.out.printf("Received onError exception in showVersionCallback2: call=%s, reply=%d, context=%d%n",
-                ex.getMethodName(), ex.getErrorCode(), ex.getCtxId());
-        }
-
-    };
-
-    private static GetNodeIndexCallback getNodeIndexCallback = new GetNodeIndexCallback() {
-        @Override
-        public void onGetNodeIndexReply(final GetNodeIndexReply msg) {
-            System.out.printf("Received GetNodeIndexReply: %s%n", msg);
-        }
-
-        @Override
-        public void onError(VppCallbackException ex) {
-            System.out.printf("Received onError exception in getNodeIndexCallback: call=%s, reply=%d, context=%d%n",
-                ex.getMethodName(), ex.getErrorCode(), ex.getCtxId());
-        }
-    };
-
-    private static void testCallbackFacade() throws Exception {
-        System.out.println("Testing CallbackJVppFacade");
-
-        try (final JVppRegistry registry = new JVppRegistryImpl("CallbackFacadeTest");
-             final CallbackJVppCoreFacade callbackFacade = new CallbackJVppCoreFacade(registry, new JVppCoreImpl())) {
-            System.out.println("Successfully connected to VPP");
-
-            callbackFacade.showVersion(showVersionCallback1);
-            callbackFacade.showVersion(showVersionCallback2);
-
-            GetNodeIndex getNodeIndexRequest = new GetNodeIndex();
-            getNodeIndexRequest.nodeName = "dummyNode".getBytes();
-            callbackFacade.getNodeIndex(getNodeIndexRequest, getNodeIndexCallback);
-
-            Thread.sleep(2000);
-            System.out.println("Disconnecting...");
-        }
-        Thread.sleep(1000);
-    }
-
-    public static void main(String[] args) throws Exception {
-        testCallbackFacade();
-    }
-}
diff --git a/vpp-api/java/jvpp-core/io/fd/vpp/jvpp/core/test/CallbackNotificationApiTest.java b/vpp-api/java/jvpp-core/io/fd/vpp/jvpp/core/test/CallbackNotificationApiTest.java
deleted file mode 100644
index a9f71f11..00000000
--- a/vpp-api/java/jvpp-core/io/fd/vpp/jvpp/core/test/CallbackNotificationApiTest.java
+++ /dev/null
@@ -1,94 +0,0 @@
-/*
- * Copyright (c) 2016 Cisco and/or its affiliates.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at:
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-package io.fd.vpp.jvpp.core.test;
-
-import static io.fd.vpp.jvpp.core.test.NotificationUtils.getChangeInterfaceState;
-import static io.fd.vpp.jvpp.core.test.NotificationUtils.getDisableInterfaceNotificationsReq;
-import static io.fd.vpp.jvpp.core.test.NotificationUtils.getEnableInterfaceNotificationsReq;
-import static io.fd.vpp.jvpp.core.test.NotificationUtils.printNotification;
-
-import io.fd.vpp.jvpp.JVpp;
-import io.fd.vpp.jvpp.JVppRegistry;
-import io.fd.vpp.jvpp.JVppRegistryImpl;
-import io.fd.vpp.jvpp.VppCallbackException;
-import io.fd.vpp.jvpp.core.JVppCoreImpl;
-import io.fd.vpp.jvpp.core.callback.SwInterfaceSetFlagsCallback;
-import io.fd.vpp.jvpp.core.callback.SwInterfaceSetFlagsNotificationCallback;
-import io.fd.vpp.jvpp.core.callback.WantInterfaceEventsCallback;
-import io.fd.vpp.jvpp.core.dto.SwInterfaceSetFlagsNotification;
-import io.fd.vpp.jvpp.core.dto.SwInterfaceSetFlagsReply;
-import io.fd.vpp.jvpp.core.dto.WantInterfaceEventsReply;
-
-public class CallbackNotificationApiTest {
-
-    private static void testCallbackApi() throws Exception {
-        System.out.println("Testing Java callback API for notifications");
-        try (final JVppRegistry registry = new JVppRegistryImpl("CallbackNotificationTest");
-             final JVpp jvpp = new JVppCoreImpl()) {
-            registry.register(jvpp, new TestCallback());
-            System.out.println("Successfully connected to VPP");
-
-            getEnableInterfaceNotificationsReq().send(jvpp);
-            System.out.println("Interface notifications started");
-            // TODO test ifc dump which also triggers interface flags send
-
-            System.out.println("Changing interface configuration");
-            getChangeInterfaceState().send(jvpp);
-
-            // Notifications are received
-            Thread.sleep(500);
-
-            getDisableInterfaceNotificationsReq().send(jvpp);
-            System.out.println("Interface events stopped");
-
-            Thread.sleep(2000);
-            System.out.println("Disconnecting...");
-        }
-        Thread.sleep(1000);
-    }
-
-    public static void main(String[] args) throws Exception {
-        testCallbackApi();
-    }
-
-    private static class TestCallback implements SwInterfaceSetFlagsNotificationCallback,
-        WantInterfaceEventsCallback, SwInterfaceSetFlagsCallback {
-
-        @Override
-        public void onSwInterfaceSetFlagsNotification(
-            final SwInterfaceSetFlagsNotification msg) {
-            printNotification(msg);
-        }
-
-        @Override
-        public void onWantInterfaceEventsReply(final WantInterfaceEventsReply wantInterfaceEventsReply) {
-            System.out.println("Interface notification stream updated");
-        }
-
-        @Override
-        public void onSwInterfaceSetFlagsReply(final SwInterfaceSetFlagsReply swInterfaceSetFlagsReply) {
-            System.out.println("Interface flags set successfully");
-        }
-
-        @Override
-        public void onError(VppCallbackException ex) {
-            System.out.printf("Received onError exception in getNodeIndexCallback: call=%s, reply=%d, context=%d%n",
-                ex.getMethodName(), ex.getErrorCode(), ex.getCtxId());
-
-        }
-    }
-}
diff --git a/vpp-api/java/jvpp-core/io/fd/vpp/jvpp/core/test/ControlPingTest.java b/vpp-api/java/jvpp-core/io/fd/vpp/jvpp/core/test/ControlPingTest.java
deleted file mode 100644
index e97f4e3a..00000000
--- a/vpp-api/java/jvpp-core/io/fd/vpp/jvpp/core/test/ControlPingTest.java
+++ /dev/null
@@ -1,68 +0,0 @@
-/*
- * Copyright (c) 2016 Cisco and/or its affiliates.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at:
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-package io.fd.vpp.jvpp.core.test;
-
-import io.fd.vpp.jvpp.JVpp;
-import io.fd.vpp.jvpp.JVppRegistry;
-import io.fd.vpp.jvpp.JVppRegistryImpl;
-import io.fd.vpp.jvpp.VppCallbackException;
-import io.fd.vpp.jvpp.callback.ControlPingCallback;
-import io.fd.vpp.jvpp.core.JVppCoreImpl;
-import io.fd.vpp.jvpp.dto.ControlPing;
-import io.fd.vpp.jvpp.dto.ControlPingReply;
-
-public class ControlPingTest {
-
-    private static void testControlPing() throws Exception {
-        System.out.println("Testing ControlPing using Java callback API");
-        try (JVppRegistry registry = new JVppRegistryImpl("ControlPingTest");
-             JVpp jvpp = new JVppCoreImpl()) {
-
-            registry.register(jvpp, new ControlPingCallback() {
-                @Override
-                public void onControlPingReply(final ControlPingReply reply) {
-                    System.out.printf("Received ControlPingReply: %s%n", reply);
-                }
-
-                @Override
-                public void onError(VppCallbackException ex) {
-                    System.out.printf("Received onError exception: call=%s, reply=%d, context=%d ", ex.getMethodName(),
-                        ex.getErrorCode(), ex.getCtxId());
-                }
-
-            });
-            System.out.println("Successfully connected to VPP");
-            Thread.sleep(1000);
-
-            System.out.println("Sending control ping using JVppRegistry");
-            registry.controlPing(jvpp.getClass());
-
-            Thread.sleep(2000);
-
-            System.out.println("Sending control ping using JVpp plugin");
-            jvpp.send(new ControlPing());
-
-            Thread.sleep(2000);
-            System.out.println("Disconnecting...");
-        }
-        Thread.sleep(1000);
-    }
-
-    public static void main(String[] args) throws Exception {
-        testControlPing();
-    }
-}
diff --git a/vpp-api/java/jvpp-core/io/fd/vpp/jvpp/core/test/CreateSubInterfaceTest.java b/vpp-api/java/jvpp-core/io/fd/vpp/jvpp/core/test/CreateSubInterfaceTest.java
deleted file mode 100644
index a96258f4..00000000
--- a/vpp-api/java/jvpp-core/io/fd/vpp/jvpp/core/test/CreateSubInterfaceTest.java
+++ /dev/null
@@ -1,120 +0,0 @@
-/*
- * Copyright (c) 2016 Cisco and/or its affiliates.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at:
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-package io.fd.vpp.jvpp.core.test;
-
-import static java.util.Objects.requireNonNull;
-
-import io.fd.vpp.jvpp.JVppRegistry;
-import io.fd.vpp.jvpp.JVppRegistryImpl;
-import io.fd.vpp.jvpp.core.JVppCoreImpl;
-import io.fd.vpp.jvpp.core.dto.CreateSubif;
-import io.fd.vpp.jvpp.core.dto.CreateSubifReply;
-import io.fd.vpp.jvpp.core.dto.SwInterfaceDetailsReplyDump;
-import io.fd.vpp.jvpp.core.dto.SwInterfaceDump;
-import io.fd.vpp.jvpp.core.future.FutureJVppCoreFacade;
-
-/**
- * 
Tests sub-interface creation.
 Equivalent to:

- *
- * 
{@code
- * vppctl create sub GigabitEthernet0/9/0 1 dot1q 100 inner-dot1q any
- * }
- * 

- *
- * To verify invoke:

- * 
{@code
- * vpp_api_test json
- * vat# sw_interface_dump
- * }
- */
-public class CreateSubInterfaceTest {
-
-    private static SwInterfaceDump createSwInterfaceDumpRequest(final String ifaceName) {
-        SwInterfaceDump request = new SwInterfaceDump();
-        request.nameFilter = ifaceName.getBytes();
-        request.nameFilterValid = 1;
-        return request;
-    }
-
-    private static void requireSingleIface(final SwInterfaceDetailsReplyDump response, final String ifaceName) {
-        if (response.swInterfaceDetails.size() != 1) {
-            throw new IllegalStateException(
-                String.format("Expected one interface matching filter %s but was %d", ifaceName,
-                    response.swInterfaceDetails.size()));
-        }
-    }
-
-    private static CreateSubif createSubifRequest(final int swIfIndex, final int subId) {
-        CreateSubif request = new CreateSubif();
-        request.swIfIndex = swIfIndex; // super interface id
-        request.subId = subId;
-        request.noTags = 0;
-        request.oneTag = 0;
-        request.twoTags = 1;
-        request.dot1Ad = 0;
-        request.exactMatch = 1;
-        request.defaultSub = 0;
-        request.outerVlanIdAny = 0;
-        request.innerVlanIdAny = 1;
-        request.outerVlanId = 100;
-        request.innerVlanId = 0;
-        return request;
-    }
-
-    private static void print(CreateSubifReply reply) {
-        System.out.printf("CreateSubifReply: %s%n", reply);
-    }
-
-    private static void testCreateSubInterface() throws Exception {
-        System.out.println("Testing sub-interface creation using Java callback API");
-        try (final JVppRegistry registry = new JVppRegistryImpl("CreateSubInterface");
-             final FutureJVppCoreFacade jvppFacade = new FutureJVppCoreFacade(registry, new JVppCoreImpl())) {
-            System.out.println("Successfully connected to VPP");
-            Thread.sleep(1000);
-
-            final String ifaceName = "GigabitEthernet0/8/0";
-
-            final SwInterfaceDetailsReplyDump swInterfaceDetails =
-                jvppFacade.swInterfaceDump(createSwInterfaceDumpRequest(ifaceName)).toCompletableFuture().get();
-
-            requireNonNull(swInterfaceDetails, "swInterfaceDump returned null");
-            requireNonNull(swInterfaceDetails.swInterfaceDetails, "swInterfaceDetails is null");
-            requireSingleIface(swInterfaceDetails, ifaceName);
-
-            final int swIfIndex = swInterfaceDetails.swInterfaceDetails.get(0).swIfIndex;
-            final int subId = 1;
-
-            final CreateSubifReply createSubifReply =
-                jvppFacade.createSubif(createSubifRequest(swIfIndex, subId)).toCompletableFuture().get();
-            print(createSubifReply);
-
-            final String subIfaceName = "GigabitEthernet0/8/0." + subId;
-            final SwInterfaceDetailsReplyDump subIface =
-                jvppFacade.swInterfaceDump(createSwInterfaceDumpRequest(subIfaceName)).toCompletableFuture().get();
-            requireNonNull(swInterfaceDetails, "swInterfaceDump returned null");
-            requireNonNull(subIface.swInterfaceDetails, "swInterfaceDump returned null");
-            requireSingleIface(swInterfaceDetails, ifaceName);
-
-            System.out.println("Disconnecting...");
-        }
-        Thread.sleep(1000);
-    }
-
-    public static void main(String[] args) throws Exception {
-        testCreateSubInterface();
-    }
-}
diff --git a/vpp-api/java/jvpp-core/io/fd/vpp/jvpp/core/test/FutureApiNotificationTest.java b/vpp-api/java/jvpp-core/io/fd/vpp/jvpp/core/test/FutureApiNotificationTest.java
deleted file mode 100644
index 9efeae19..00000000
--- a/vpp-api/java/jvpp-core/io/fd/vpp/jvpp/core/test/FutureApiNotificationTest.java
+++ /dev/null
@@ -1,55 +0,0 @@
-/*
- * Copyright (c) 2016 Cisco and/or its affiliates.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at:
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-package io.fd.vpp.jvpp.core.test;
-
-import static io.fd.vpp.jvpp.core.test.NotificationUtils.getChangeInterfaceState;
-import static io.fd.vpp.jvpp.core.test.NotificationUtils.getDisableInterfaceNotificationsReq;
-import static io.fd.vpp.jvpp.core.test.NotificationUtils.getEnableInterfaceNotificationsReq;
-
-import io.fd.vpp.jvpp.JVppRegistry;
-import io.fd.vpp.jvpp.JVppRegistryImpl;
-import io.fd.vpp.jvpp.core.JVppCoreImpl;
-import io.fd.vpp.jvpp.core.future.FutureJVppCoreFacade;
-
-public class FutureApiNotificationTest {
-
-    private static void testFutureApi() throws Exception {
-        System.out.println("Testing Java future API for notifications");
-        try (final JVppRegistry registry = new JVppRegistryImpl("FutureApiNotificationTest");
-             final FutureJVppCoreFacade jvppFacade = new FutureJVppCoreFacade(registry, new JVppCoreImpl());
-             final AutoCloseable notificationListenerReg =
-                 jvppFacade.getNotificationRegistry()
-                     .registerSwInterfaceSetFlagsNotificationCallback(NotificationUtils::printNotification)) {
-            System.out.println("Successfully connected to VPP");
-            jvppFacade.wantInterfaceEvents(getEnableInterfaceNotificationsReq()).toCompletableFuture().get();
-            System.out.println("Interface events started");
-
-            System.out.println("Changing interface configuration");
-            jvppFacade.swInterfaceSetFlags(getChangeInterfaceState()).toCompletableFuture().get();
-
-            Thread.sleep(1000);
-
-            jvppFacade.wantInterfaceEvents(getDisableInterfaceNotificationsReq()).toCompletableFuture().get();
-            System.out.println("Interface events stopped");
-            System.out.println("Disconnecting...");
-        }
-    }
-
-    public static void main(String[] args) throws Exception {
-        testFutureApi();
-    }
-}
diff --git a/vpp-api/java/jvpp-core/io/fd/vpp/jvpp/core/test/FutureApiTest.java b/vpp-api/java/jvpp-core/io/fd/vpp/jvpp/core/test/FutureApiTest.java
deleted file mode 100644
index f478bab4..00000000
--- a/vpp-api/java/jvpp-core/io/fd/vpp/jvpp/core/test/FutureApiTest.java
+++ /dev/null
@@ -1,123 +0,0 @@
-/*
- * Copyright (c) 2016 Cisco and/or its affiliates.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at:
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-package io.fd.vpp.jvpp.core.test;
-
-import io.fd.vpp.jvpp.JVppRegistry;
-import io.fd.vpp.jvpp.JVppRegistryImpl;
-import io.fd.vpp.jvpp.core.JVppCoreImpl;
-import io.fd.vpp.jvpp.core.dto.BridgeDomainDetailsReplyDump;
-import io.fd.vpp.jvpp.core.dto.BridgeDomainDump;
-import io.fd.vpp.jvpp.core.dto.GetNodeIndex;
-import io.fd.vpp.jvpp.core.dto.GetNodeIndexReply;
-import io.fd.vpp.jvpp.core.dto.ShowVersion;
-import io.fd.vpp.jvpp.core.dto.ShowVersionReply;
-import io.fd.vpp.jvpp.core.dto.SwInterfaceDetails;
-import io.fd.vpp.jvpp.core.dto.SwInterfaceDetailsReplyDump;
-import io.fd.vpp.jvpp.core.dto.SwInterfaceDump;
-import io.fd.vpp.jvpp.core.future.FutureJVppCoreFacade;
-import java.util.Objects;
-import java.util.concurrent.CompletableFuture;
-import java.util.concurrent.Future;
-import java.util.logging.Level;
-import java.util.logging.Logger;
-
-public class FutureApiTest {
-
-    private static final Logger LOG = Logger.getLogger(FutureApiTest.class.getName());
-
-    private static void testShowVersion(final FutureJVppCoreFacade jvpp) throws Exception {
-        LOG.info("Sending ShowVersion request...");
-        final Future replyFuture = jvpp.showVersion(new ShowVersion()).toCompletableFuture();
-        final ShowVersionReply reply = replyFuture.get();
-        LOG.info(
-            String.format(
-                "Received ShowVersionReply: context=%d, program=%s, version=%s, buildDate=%s, buildDirectory=%s%n",
-                reply.context, new String(reply.program), new String(reply.version), new String(reply.buildDate),
-                new String(reply.buildDirectory)));
-    }
-
-    private static void testEmptyBridgeDomainDump(final FutureJVppCoreFacade jvpp) throws Exception {
-        LOG.info("Sending ShowVersion request...");
-        final BridgeDomainDump request = new BridgeDomainDump();
-        request.bdId = -1; // dump call
-
-        final CompletableFuture
-            replyFuture = jvpp.bridgeDomainDump(request).toCompletableFuture();
-        final BridgeDomainDetailsReplyDump reply = replyFuture.get();
-
-        if (reply == null || reply.bridgeDomainDetails == null) {
-            LOG.severe("Received null response for empty dump: " + reply);
-        } else {
-            LOG.info(
-                String.format(
-                    "Received empty bridge-domain dump reply with list of bridge-domains: %s, %s",
-                    reply.bridgeDomainDetails, reply.bridgeDomainSwIfDetails));
-        }
-    }
-
-    private static void testGetNodeIndex(final FutureJVppCoreFacade jvpp) {
-        LOG.info("Sending GetNodeIndex request...");
-        final GetNodeIndex request = new GetNodeIndex();
-        request.nodeName = "non-existing-node".getBytes();
-        final Future replyFuture = jvpp.getNodeIndex(request).toCompletableFuture();
-        try {
-            final GetNodeIndexReply reply = replyFuture.get();
-            LOG.info(
-                String.format(
-                    "Received GetNodeIndexReply: context=%d, nodeIndex=%d%n", reply.context, reply.nodeIndex));
-        } catch (Exception e) {
-            LOG.log(Level.SEVERE, "GetNodeIndex request failed", e);
-        }
-    }
-
-    private static void testSwInterfaceDump(final FutureJVppCoreFacade jvpp) throws Exception {
-        LOG.info("Sending SwInterfaceDump request...");
-        final SwInterfaceDump request = new SwInterfaceDump();
-        request.nameFilterValid = 0;
-        request.nameFilter = "".getBytes();
-
-        final Future replyFuture = jvpp.swInterfaceDump(request).toCompletableFuture();
-        final SwInterfaceDetailsReplyDump reply = replyFuture.get();
-        for (SwInterfaceDetails details : reply.swInterfaceDetails) {
-            Objects.requireNonNull(details, "reply.swInterfaceDetails contains null element!");
-            LOG.info(
-                String.format("Received SwInterfaceDetails: interfaceName=%s, l2AddressLength=%d, adminUpDown=%d, "
-                        + "linkUpDown=%d, linkSpeed=%d, linkMtu=%d%n",
-                    new String(details.interfaceName), details.l2AddressLength, details.adminUpDown,
-                    details.linkUpDown, details.linkSpeed, (int) details.linkMtu));
-        }
-    }
-
-    private static void testFutureApi() throws Exception {
-        LOG.info("Testing Java future API");
-        try (final JVppRegistry registry = new JVppRegistryImpl("FutureApiTest");
-             final FutureJVppCoreFacade jvppFacade = new FutureJVppCoreFacade(registry, new JVppCoreImpl())) {
-            LOG.info("Successfully connected to VPP");
-
-            testEmptyBridgeDomainDump(jvppFacade);
-            testShowVersion(jvppFacade);
-            testGetNodeIndex(jvppFacade);
-            testSwInterfaceDump(jvppFacade);
-
-            LOG.info("Disconnecting...");
-        }
-    }
-
-    public static void main(String[] args) throws Exception {
-        testFutureApi();
-    }
-}
diff --git a/vpp-api/java/jvpp-core/io/fd/vpp/jvpp/core/test/L2AclTest.java b/vpp-api/java/jvpp-core/io/fd/vpp/jvpp/core/test/L2AclTest.java
deleted file mode 100644
index 6b3fa993..00000000
--- a/vpp-api/java/jvpp-core/io/fd/vpp/jvpp/core/test/L2AclTest.java
+++ /dev/null
@@ -1,195 +0,0 @@
-/*
- * Copyright (c) 2016 Cisco and/or its affiliates.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at:
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-package io.fd.vpp.jvpp.core.test;
-
-import io.fd.vpp.jvpp.JVppRegistry;
-import io.fd.vpp.jvpp.JVppRegistryImpl;
-import io.fd.vpp.jvpp.core.JVppCoreImpl;
-import io.fd.vpp.jvpp.core.dto.ClassifyAddDelSession;
-import io.fd.vpp.jvpp.core.dto.ClassifyAddDelSessionReply;
-import io.fd.vpp.jvpp.core.dto.ClassifyAddDelTable;
-import io.fd.vpp.jvpp.core.dto.ClassifyAddDelTableReply;
-import io.fd.vpp.jvpp.core.dto.ClassifySessionDetailsReplyDump;
-import io.fd.vpp.jvpp.core.dto.ClassifySessionDump;
-import io.fd.vpp.jvpp.core.dto.ClassifyTableByInterface;
-import io.fd.vpp.jvpp.core.dto.ClassifyTableByInterfaceReply;
-import io.fd.vpp.jvpp.core.dto.ClassifyTableIds;
-import io.fd.vpp.jvpp.core.dto.ClassifyTableIdsReply;
-import io.fd.vpp.jvpp.core.dto.ClassifyTableInfo;
-import io.fd.vpp.jvpp.core.dto.ClassifyTableInfoReply;
-import io.fd.vpp.jvpp.core.dto.InputAclSetInterface;
-import io.fd.vpp.jvpp.core.dto.InputAclSetInterfaceReply;
-import io.fd.vpp.jvpp.core.future.FutureJVppCoreFacade;
-import javax.xml.bind.DatatypeConverter;
-
-/**
- * 
Tests L2 ACL creation and read.
 Equivalent to the following vppctl commands:

- *
- * 
{@code
- * vppctl classify table mask l2 src
- * vppctl classify session acl-hit-next deny opaque-index 0 table-index 0 match l2 src 01:02:03:04:05:06
- * vppctl set int input acl intfc local0 l2-table 0
- * vppctl sh class table verbose
- * }
- * 

- */
-public class L2AclTest {
-
-    private static final int LOCAL0_IFACE_ID = 0;
-
-    private static ClassifyAddDelTable createClassifyTable() {
-        ClassifyAddDelTable request = new ClassifyAddDelTable();
-        request.isAdd = 1;
-        request.tableIndex = ~0; // default
-        request.nbuckets = 2;
-        request.memorySize = 2 << 20;
-        request.nextTableIndex = ~0; // default
-        request.missNextIndex = ~0; // default
-        request.skipNVectors = 0;
-        request.matchNVectors = 1;
-        request.mask =
-            new byte[] {0x00, 0x00, 0x00, 0x00, 0x00, 0x00, (byte) 0xff, (byte) 0xff, (byte) 0xff, (byte) 0xff,
-                (byte) 0xff, (byte) 0xff, 0x00, 0x00, 0x00, 0x00};
-        return request;
-    }
-
-    private static ClassifyTableInfo createClassifyTableInfoRequest(final int tableId) {
-        ClassifyTableInfo request = new ClassifyTableInfo();
-        request.tableId = tableId;
-        return request;
-    }
-
-    private static ClassifyAddDelSession createClassifySession(final int tableIndex) {
-        ClassifyAddDelSession request = new ClassifyAddDelSession();
-        request.isAdd = 1;
-        request.tableIndex = tableIndex;
-        request.hitNextIndex = 0; // deny
-        request.opaqueIndex = 0;
-        request.advance = 0; // default
-        // match 01:02:03:04:05:06 mac address
-        request.match =
-            new byte[] {0x00, 0x00, 0x00, 0x00, 0x00, 0x00, (byte) 0x01, (byte) 0x02, (byte) 0x03, (byte) 0x04,
-                (byte) 0x05, (byte) 0x06, 0x00, 0x00, 0x00, 0x00};
-        return request;
-    }
-
-    private static ClassifySessionDump createClassifySessionDumpRequest(final int newTableIndex) {
-        ClassifySessionDump request = new ClassifySessionDump();
-        request.tableId = newTableIndex;
-        return request;
-    }
-
-    private static InputAclSetInterface aclSetInterface() {
-        InputAclSetInterface request = new InputAclSetInterface();
-        request.isAdd = 1;
-        request.swIfIndex = LOCAL0_IFACE_ID;
-        request.ip4TableIndex = ~0; // skip
-        request.ip6TableIndex = ~0; // skip
-        request.l2TableIndex = 0;
-        return request;
-    }
-
-    private static ClassifyTableByInterface createClassifyTableByInterfaceRequest() {
-        ClassifyTableByInterface request = new ClassifyTableByInterface();
-        request.swIfIndex = LOCAL0_IFACE_ID;
-        return request;
-    }
-
-    private static void print(ClassifyAddDelTableReply reply) {
-        System.out.printf("ClassifyAddDelTableReply: %s%n", reply);
-    }
-
-    private static void print(ClassifyTableIdsReply reply) {
-        System.out.printf("ClassifyTableIdsReply: %s%n", reply);
-    }
-
-    private static void print(final ClassifyTableInfoReply reply) {
-        System.out.println(reply);
-        if (reply != null) {
-            System.out.println("Mask hex: " + DatatypeConverter.printHexBinary(reply.mask));
-        }
-    }
-
-    private static void print(ClassifyAddDelSessionReply reply) {
-        System.out.printf("ClassifyAddDelSessionReply: context=%s%n", reply);
-    }
-
-    private static void print(final ClassifySessionDetailsReplyDump reply) {
-        System.out.println(reply);
-        reply.classifySessionDetails.forEach(detail -> {
-            System.out.println(detail);
-            System.out.println("Match hex: " + DatatypeConverter.printHexBinary(detail.match));
-        });
-    }
-
-    private static void print(final InputAclSetInterfaceReply reply) {
-        System.out.printf("InputAclSetInterfaceReply: context=%s%n", reply);
-    }
-
-    private static void print(final ClassifyTableByInterfaceReply reply) {
-        System.out.printf("ClassifyAddDelTableReply: %s%n", reply);
-    }
-
-    private static void testL2Acl() throws Exception {
-        System.out.println("Testing L2 ACLs using Java callback API");
-        try (final JVppRegistry registry = new JVppRegistryImpl("L2AclTest");
-             final FutureJVppCoreFacade jvppFacade = new FutureJVppCoreFacade(registry, new JVppCoreImpl())) {
-
-            System.out.println("Successfully connected to VPP");
-            Thread.sleep(1000);
-
-            final ClassifyAddDelTableReply classifyAddDelTableReply =
-                jvppFacade.classifyAddDelTable(createClassifyTable()).toCompletableFuture().get();
-            print(classifyAddDelTableReply);
-
-            final ClassifyTableIdsReply classifyTableIdsReply =
-                jvppFacade.classifyTableIds(new ClassifyTableIds()).toCompletableFuture().get();
-            print(classifyTableIdsReply);
-
-            final ClassifyTableInfoReply classifyTableInfoReply =
-                jvppFacade.classifyTableInfo(createClassifyTableInfoRequest(classifyAddDelTableReply.newTableIndex))
-                    .toCompletableFuture().get();
-            print(classifyTableInfoReply);
-
-            final ClassifyAddDelSessionReply classifyAddDelSessionReply =
-                jvppFacade.classifyAddDelSession(createClassifySession(classifyAddDelTableReply.newTableIndex))
-                    .toCompletableFuture().get();
-            print(classifyAddDelSessionReply);
-
-            final ClassifySessionDetailsReplyDump classifySessionDetailsReplyDump =
-                jvppFacade.classifySessionDump(createClassifySessionDumpRequest(classifyAddDelTableReply.newTableIndex))
-                    .toCompletableFuture().get();
-            print(classifySessionDetailsReplyDump);
-
-            final InputAclSetInterfaceReply inputAclSetInterfaceReply =
-                jvppFacade.inputAclSetInterface(aclSetInterface()).toCompletableFuture().get();
-            print(inputAclSetInterfaceReply);
-
-            final ClassifyTableByInterfaceReply classifyTableByInterfaceReply =
-                jvppFacade.classifyTableByInterface(createClassifyTableByInterfaceRequest()).toCompletableFuture()
-                    .get();
-            print(classifyTableByInterfaceReply);
-
-            System.out.println("Disconnecting...");
-        }
-        Thread.sleep(1000);
-    }
-
-    public static void main(String[] args) throws Exception {
-        testL2Acl();
-    }
-}
diff --git a/vpp-api/java/jvpp-core/io/fd/vpp/jvpp/core/test/LispAdjacencyTest.java b/vpp-api/java/jvpp-core/io/fd/vpp/jvpp/core/test/LispAdjacencyTest.java
deleted file mode 100644
index d7f5039b..00000000
--- a/vpp-api/java/jvpp-core/io/fd/vpp/jvpp/core/test/LispAdjacencyTest.java
+++ /dev/null
@@ -1,124 +0,0 @@
-/*
- * Copyright (c) 2016 Cisco and/or its affiliates.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at:
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-package io.fd.vpp.jvpp.core.test;
-
-import io.fd.vpp.jvpp.JVppRegistry;
-import io.fd.vpp.jvpp.JVppRegistryImpl;
-import io.fd.vpp.jvpp.core.JVppCoreImpl;
-import io.fd.vpp.jvpp.core.dto.LispAddDelAdjacency;
-import io.fd.vpp.jvpp.core.dto.LispAddDelLocalEid;
-import io.fd.vpp.jvpp.core.dto.LispAddDelLocatorSet;
-import io.fd.vpp.jvpp.core.dto.LispAddDelRemoteMapping;
-import io.fd.vpp.jvpp.core.dto.LispAdjacenciesGet;
-import io.fd.vpp.jvpp.core.dto.LispAdjacenciesGetReply;
-import io.fd.vpp.jvpp.core.dto.LispEnableDisable;
-import io.fd.vpp.jvpp.core.future.FutureJVppCoreFacade;
-import java.nio.charset.StandardCharsets;
-import java.util.concurrent.ExecutionException;
-import java.util.logging.Logger;
-
-/**
- * Tests lisp adjacency creation and read (custom vpe.api type support showcase).
- */
-public class LispAdjacencyTest {
-
-    private static final Logger LOG = Logger.getLogger(LispAdjacencyTest.class.getName());
-
-    private static void enableLisp(final FutureJVppCoreFacade jvpp) throws ExecutionException, InterruptedException {
-        final LispEnableDisable request = new LispEnableDisable();
-        request.isEn = 1;
-        jvpp.lispEnableDisable(request).toCompletableFuture().get();
-        LOG.info("Lisp enabled successfully");
-    }
-
-    private static void addLocatorSet(final FutureJVppCoreFacade jvpp) throws ExecutionException, InterruptedException {
-        final LispAddDelLocatorSet request = new LispAddDelLocatorSet();
-        request.isAdd = 1;
-        request.locatorSetName = "ls1".getBytes(StandardCharsets.UTF_8);
-        jvpp.lispAddDelLocatorSet(request).toCompletableFuture().get();
-        LOG.info("Locator set created successfully:" + request.toString());
-    }
-
-    private static void addLocalEid(final FutureJVppCoreFacade jvpp) throws ExecutionException, InterruptedException {
-        final LispAddDelLocalEid request = new LispAddDelLocalEid();
-        request.isAdd = 1;
-        request.locatorSetName = "ls1".getBytes(StandardCharsets.UTF_8);
-        request.eid = new byte[] {1, 2, 1, 10};
-        request.eidType = 0; // ip4
-        request.vni = 0;
-        request.prefixLen = 32;
-        jvpp.lispAddDelLocalEid(request).toCompletableFuture().get();
-        LOG.info("Local EID created successfully:" + request.toString());
-    }
-
-    private static void addRemoteMapping(final FutureJVppCoreFacade jvpp)
-        throws ExecutionException, InterruptedException {
-        final LispAddDelRemoteMapping request = new LispAddDelRemoteMapping();
-        request.isAdd = 1;
-        request.vni = 0;
-        request.eid = new byte[] {1, 2, 1, 20};
-        request.eidLen = 32;
-        request.rlocNum = 1;
-        request.rlocs = new byte[] {1, 1, 1, 1, 2, 1, 20, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0};
-        jvpp.lispAddDelRemoteMapping(request).toCompletableFuture().get();
-        LOG.info("Remote mapping created successfully:" + request.toString());
-    }
-
-    private static void addAdjacency(final FutureJVppCoreFacade jvpp) throws ExecutionException, InterruptedException {
-        final LispAddDelAdjacency request = new LispAddDelAdjacency();
-        request.isAdd = 1;
-        request.leid = new byte[] {1, 2, 1, 10};
-        request.leidLen = 32;
-        request.reid = new byte[] {1, 2, 1, 20};
-        request.reidLen = 32;
-        request.eidType = 0; // ip4
-        request.vni = 0;
-        jvpp.lispAddDelAdjacency(request).toCompletableFuture().get();
-        LOG.info("Lisp adjacency created successfully:" + request.toString());
-    }
-
-    private static void showAdjacencies(final FutureJVppCoreFacade jvpp)
-        throws ExecutionException, InterruptedException {
-        final LispAdjacenciesGetReply reply =
-            jvpp.lispAdjacenciesGet(new LispAdjacenciesGet()).toCompletableFuture().get();
-        LOG.info("Lisp adjacency received successfully:" + reply.toString());
-    }
-
-    private static void testAdjacency(final FutureJVppCoreFacade jvpp) throws Exception {
-        enableLisp(jvpp);
-        addLocatorSet(jvpp);
-        addLocalEid(jvpp);
-        addRemoteMapping(jvpp);
-        addAdjacency(jvpp);
-        showAdjacencies(jvpp);
-    }
-
-    private static void testFutureApi() throws Exception {
-        LOG.info("Create lisp adjacency test");
-        try (final JVppRegistry registry = new JVppRegistryImpl("LispAdjacencyTest");
-             final FutureJVppCoreFacade jvppFacade = new FutureJVppCoreFacade(registry, new JVppCoreImpl())) {
-            LOG.info("Successfully connected to VPP");
-
-            testAdjacency(jvppFacade);
-            LOG.info("Disconnecting...");
-        }
-    }
-
-    public static void main(String[] args) throws Exception {
-        testFutureApi();
-    }
-}
diff --git a/vpp-api/java/jvpp-core/io/fd/vpp/jvpp/core/test/NotificationUtils.java b/vpp-api/java/jvpp-core/io/fd/vpp/jvpp/core/test/NotificationUtils.java
deleted file mode 100644
index f82946c3..00000000
--- a/vpp-api/java/jvpp-core/io/fd/vpp/jvpp/core/test/NotificationUtils.java
+++ /dev/null
@@ -1,53 +0,0 @@
-/*
- * Copyright (c) 2016 Cisco and/or its affiliates.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at:
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-package io.fd.vpp.jvpp.core.test;
-
-import java.io.PrintStream;
-import io.fd.vpp.jvpp.core.dto.SwInterfaceSetFlags;
-import io.fd.vpp.jvpp.core.dto.SwInterfaceSetFlagsNotification;
-import io.fd.vpp.jvpp.core.dto.WantInterfaceEvents;
-
-final class NotificationUtils {
-
-    private NotificationUtils() {}
-
-    static PrintStream printNotification(final SwInterfaceSetFlagsNotification msg) {
-        return System.out.printf("Received interface notification: ifc: %s%n", msg);
-    }
-
-    static SwInterfaceSetFlags getChangeInterfaceState() {
-        final SwInterfaceSetFlags swInterfaceSetFlags = new SwInterfaceSetFlags();
-        swInterfaceSetFlags.swIfIndex = 0;
-        swInterfaceSetFlags.adminUpDown = 1;
-        swInterfaceSetFlags.deleted = 0;
-        return swInterfaceSetFlags;
-    }
-
-    static WantInterfaceEvents getEnableInterfaceNotificationsReq() {
-        WantInterfaceEvents wantInterfaceEvents = new WantInterfaceEvents();
-        wantInterfaceEvents.pid = 1;
-        wantInterfaceEvents.enableDisable = 1;
-        return wantInterfaceEvents;
-    }
-
-    static WantInterfaceEvents getDisableInterfaceNotificationsReq() {
-        WantInterfaceEvents wantInterfaceEvents = new WantInterfaceEvents();
-        wantInterfaceEvents.pid = 1;
-        wantInterfaceEvents.enableDisable = 0;
-        return wantInterfaceEvents;
-    }
-}
diff --git a/vpp-api/java/jvpp-core/io/fd/vpp/jvpp/core/test/Readme.txt b/vpp-api/java/jvpp-core/io/fd/vpp/jvpp/core/test/Readme.txt
deleted file mode 100644
index 1344dc9e..00000000
--- a/vpp-api/java/jvpp-core/io/fd/vpp/jvpp/core/test/Readme.txt
+++ /dev/null
@@ -1,17 +0,0 @@
-This package contains basic tests for jvpp. To run the tests:
-
-- Make sure VPP is running
-- From VPP's build-root/ folder execute:
-  - sudo java -cp build-vpp-native/vpp-api/java/jvpp-registry-17.01.jar:build-vpp-native/vpp-api/java/jvpp-core-17.01.jar io.fd.vpp.jvpp.core.test.[test name]
-
-Available tests:
-CallbackApiTest - Similar to ControlPingTest, invokes more complex calls (e.g. interface dump) using low level JVpp APIs
-CallbackJVppFacadeNotificationTest - Tests interface notifications using Callback based JVpp facade
-CallbackJVppFacadeTest - Execution of more complex calls using Callback based JVpp facade
-CallbackNotificationApiTest - Tests interface notifications using low level JVpp APIs
-ControlPingTest - Simple test executing a single control ping using low level JVpp APIs
-CreateSubInterfaceTest - Tests sub-interface creation
-FutureApiNotificationTest - Tests interface notifications using Future based JVpp facade
-FutureApiTest - Execution of more complex calls using Future based JVpp facade
-L2AclTest - Tests L2 ACL creation
-LispAdjacencyTest - Tests lisp adjacency creation and read (custom vpe.api type support showcase)
diff --git a/vpp-api/java/jvpp-core/jvpp_core.c b/vpp-api/java/jvpp-core/jvpp_core.c
deleted file mode 100644
index ef4cb8e3..00000000
--- a/vpp-api/java/jvpp-core/jvpp_core.c
+++ /dev/null
@@ -1,117 +0,0 @@
-/*
- * Copyright (c) 2016 Cisco and/or its affiliates.
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at:
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#include 
-
-#include 
-#define vl_typedefs             /* define message structures */
-#include 
-#undef vl_typedefs
-
-#define vl_endianfun
-#include 
-#undef vl_endianfun
-
-#define vl_print(handle, ...)
-#define vl_printfun
-#include 
-#undef vl_printfun
-
-#include 
-#include 
-#include 
-#include 
-
-#include 
-
-// TODO: generate jvpp_plugin_name.c files (or at least reuse plugin's main structure)
-typedef struct {
-    /* Base message index for the jvpp-core plugin */
-    u16 msg_id_base;
-
-    /* Pointer to shared memory queue */
-    unix_shared_memory_queue_t * vl_input_queue;
-
-    /* VPP api client index */
-    u32 my_client_index;
-
-    /* Callback object and class references enabling asynchronous Java calls */
-    jobject callbackObject;
-    jclass callbackClass;
-
-} core_main_t;
-
-core_main_t core_main __attribute__((aligned (64)));
-
-#include "io_fd_vpp_jvpp_core_JVppCoreImpl.h"
-#include "jvpp_core_gen.h"
-
-JNIEXPORT void JNICALL Java_io_fd_vpp_jvpp_core_JVppCoreImpl_init0
-(JNIEnv * env, jclass clazz, jobject callback, jlong queue_address, jint my_client_index) {
-    core_main_t * plugin_main = &core_main;
-    plugin_main->my_client_index = my_client_index;
-    plugin_main->vl_input_queue = (unix_shared_memory_queue_t *)queue_address;
-
-    plugin_main->callbackObject = (*env)->NewGlobalRef(env, callback);
-    plugin_main->callbackClass = (jclass)(*env)->NewGlobalRef(env, (*env)->GetObjectClass(env, callback));
-
-    #define _(N,n)                                  \
-        vl_msg_api_set_handlers(VL_API_##N, #n,     \
-                vl_api_##n##_t_handler,             \
-                vl_noop_handler,                    \
-                vl_noop_handler,              \
-                vl_noop_handler,               \
-                sizeof(vl_api_##n##_t), 1);
-        foreach_api_reply_handler;
-    #undef _
-}
-
-JNIEXPORT void JNICALL Java_io_fd_vpp_jvpp_core_JVppCoreImpl_close0
-(JNIEnv *env, jclass clazz) {
-    core_main_t * plugin_main = &core_main;
-
-    // cleanup:
-    (*env)->DeleteGlobalRef(env, plugin_main->callbackClass);
-    (*env)->DeleteGlobalRef(env, plugin_main->callbackObject);
-
-    plugin_main->callbackClass = NULL;
-    plugin_main->callbackObject = NULL;
-}
-
-jint JNI_OnLoad(JavaVM *vm, void *reserved) {
-    JNIEnv* env;
-
-    if ((*vm)->GetEnv(vm, (void**) &env, JNI_VERSION_1_8) != JNI_OK) {
-        return JNI_EVERSION;
-    }
-
-    if (cache_class_references(env) != 0) {
-        clib_warning ("Failed to cache class references\n");
-        return JNI_ERR;
-    }
-
-    return JNI_VERSION_1_8;
-}
-
-void JNI_OnUnload(JavaVM *vm, void *reserved) {
-    JNIEnv* env;
-    if ((*vm)->GetEnv(vm, (void**) &env, JNI_VERSION_1_8) != JNI_OK) {
-        return;
-    }
-    delete_class_references(env);
-}
-
-
-
diff --git a/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/JVpp.java b/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/JVpp.java
deleted file mode 100644
index 55f25a7b..00000000
--- a/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/JVpp.java
+++ /dev/null
@@ -1,56 +0,0 @@
-/*
- * Copyright (c) 2016 Cisco and/or its affiliates.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at:
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-package io.fd.vpp.jvpp;
-
-import io.fd.vpp.jvpp.callback.JVppCallback;
-import io.fd.vpp.jvpp.dto.ControlPing;
-import io.fd.vpp.jvpp.dto.JVppRequest;
-
-/**
- * Base interface for plugin's Java API.
- */
-public interface JVpp extends AutoCloseable {
-
-    /**
-     * Sends request to vpp.
-     *
-     * @param request request to be sent
-     * @return unique identifer of message in message queue
-     * @throws VppInvocationException when message could not be sent
-     */
-    int send(final JVppRequest request) throws VppInvocationException;
-
-    /**
-     * Initializes plugin's Java API.
-     *
-     * @param registry     plugin registry
-     * @param callback     called by vpe.api message handlers
-     * @param queueAddress address of vpp shared memory queue
-     * @param clientIndex  vpp client identifier
-     */
-    void init(final JVppRegistry registry, final JVppCallback callback, final long queueAddress,
-              final int clientIndex);
-
-    /**
-     * Sends control_ping message.
-     *
-     * @param controlPing request DTO
-     * @return unique identifer of message in message queue
-     * @throws VppInvocationException when message could not be sent
-     */
-    int controlPing(final ControlPing controlPing) throws VppInvocationException;
-}
diff --git a/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/JVppRegistry.java b/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/JVppRegistry.java
deleted file mode 100644
index 6535db02..00000000
--- a/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/JVppRegistry.java
+++ /dev/null
@@ -1,76 +0,0 @@
-/*
- * Copyright (c) 2016 Cisco and/or its affiliates.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at:
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-package io.fd.vpp.jvpp;
-
-import io.fd.vpp.jvpp.callback.JVppCallback;
-
-/**
- * Manages VPP connection and stores plugin callbacks.
- */
-public interface JVppRegistry extends AutoCloseable {
-
-    /**
-     * Vpp connection managed by the registry.
-     *
-     * @return representation of vpp connection
-     */
-    VppConnection getConnection();
-
-    /**
-     * Registers callback and initializes Java API for given plugin.
-     *
-     * @param jvpp     plugin name
-     * @param callback callback provided by the plugin
-     * @throws NullPointerException     if name or callback is null
-     * @throws IllegalArgumentException if plugin was already registered
-     */
-    void register(final JVpp jvpp, final JVppCallback callback);
-
-    /**
-     * Unregisters callback for the given plugin.
-     *
-     * @param name plugin name
-     * @throws NullPointerException     if name is null
-     * @throws IllegalArgumentException if plugin was not registered
-     */
-    void unregister(final String name);
-
-    /**
-     * Returns callback registered for the plugin.
-     *
-     * @param name plugin name
-     * @return callback provided by the plugin
-     * @throws NullPointerException     if name is null
-     * @throws IllegalArgumentException if plugin was not registered
-     */
-    JVppCallback get(final String name);
-
-    /**
-     * Sends control ping. Reply handler calls callback registered for give plugin.
-     *
-     * Control ping is used for initial RX thread to Java thread attachment
-     * that takes place in the plugin's JNI lib
-     * and to wrap dump message replies in one list.
-     *
-     * VPP plugins don't have to provide special control ping, therefore
-     * it is necessary to providing control ping support in JVppRegistry.
-
-     * @param clazz identifies plugin that should receive ping callback
-     * @return unique identifier of message in message queue
-     */
-    int controlPing(final Class clazz) throws VppInvocationException;
-}
diff --git a/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/JVppRegistryImpl.java b/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/JVppRegistryImpl.java
deleted file mode 100644
index 98ef1c15..00000000
--- a/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/JVppRegistryImpl.java
+++ /dev/null
@@ -1,147 +0,0 @@
-/*
- * Copyright (c) 2016 Cisco and/or its affiliates.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at:
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-package io.fd.vpp.jvpp;
-
-import static java.util.Objects.requireNonNull;
-
-import io.fd.vpp.jvpp.callback.ControlPingCallback;
-import io.fd.vpp.jvpp.callback.JVppCallback;
-import io.fd.vpp.jvpp.dto.ControlPingReply;
-import java.io.IOException;
-import java.util.HashMap;
-import java.util.Map;
-import java.util.concurrent.ConcurrentHashMap;
-import java.util.logging.Level;
-import java.util.logging.Logger;
-
-/**
- * Default implementation of JVppRegistry.
- */
-public final class JVppRegistryImpl implements JVppRegistry, ControlPingCallback {
-
-    private static final Logger LOG = Logger.getLogger(JVppRegistryImpl.class.getName());
-
-    private final VppJNIConnection connection;
-    // Unguarded concurrent map, no race conditions expected on top of that
-    private final Map pluginRegistry;
-    // Guarded by self
-    private final Map pingCalls;
-
-    public JVppRegistryImpl(final String clientName) throws IOException {
-        connection = new VppJNIConnection(clientName);
-        connection.connect();
-        pluginRegistry = new ConcurrentHashMap<>();
-        pingCalls = new HashMap<>();
-    }
-
-    @Override
-    public VppConnection getConnection() {
-        return connection;
-    }
-
-    @Override
-    public void register(final JVpp jvpp, final JVppCallback callback) {
-        requireNonNull(jvpp, "jvpp should not be null");
-        requireNonNull(callback, "Callback should not be null");
-        final String name = jvpp.getClass().getName();
-        if (pluginRegistry.containsKey(name)) {
-            throw new IllegalArgumentException(
-                String.format("Callback for plugin %s was already registered", name));
-        }
-        jvpp.init(this, callback, connection.getConnectionInfo().queueAddress,
-            connection.getConnectionInfo().clientIndex);
-        pluginRegistry.put(name, callback);
-    }
-
-    @Override
-    public void unregister(final String name) {
-        requireNonNull(name, "Plugin name should not be null");
-        final JVppCallback previous = pluginRegistry.remove(name);
-        assertPluginWasRegistered(name, previous);
-    }
-
-    @Override
-    public JVppCallback get(final String name) {
-        requireNonNull(name, "Plugin name should not be null");
-        JVppCallback value = pluginRegistry.get(name);
-        assertPluginWasRegistered(name, value);
-        return value;
-    }
-
-    private native int controlPing0() throws VppInvocationException;
-
-    @Override
-    public int controlPing(final Class clazz) throws VppInvocationException {
-        connection.checkActive();
-        final String name = clazz.getName();
-
-        final ControlPingCallback callback = (ControlPingCallback) pluginRegistry.get(clazz.getName());
-        assertPluginWasRegistered(name, callback);
-
-        synchronized (pingCalls) {
-            int context = controlPing0();
-            if (context < 0) {
-                throw new VppInvocationException("controlPing", context);
-            }
-
-            pingCalls.put(context, callback);
-            return context;
-        }
-    }
-
-    @Override
-    public void onControlPingReply(final ControlPingReply reply) {
-        final ControlPingCallback callback;
-        synchronized (pingCalls) {
-            callback = pingCalls.remove(reply.context);
-            if (callback == null) {
-                LOG.log(Level.WARNING, "No callback was registered for reply context=" + reply.context + " Contexts waiting="
-                    + pingCalls.keySet());
-                return;
-            }
-        }
-        // pass the reply to the callback registered by the ping caller
-        callback.onControlPingReply(reply);
-    }
-
-    @Override
-    public void onError(final VppCallbackException ex) {
-        final int ctxId = ex.getCtxId();
-        final ControlPingCallback callback;
-
-        synchronized (pingCalls) {
-            callback = pingCalls.get(ctxId);
-        }
-        if (callback == null) {
-            LOG.log(Level.WARNING, "No callback was registered for reply id={0} ", ctxId);
-            return;
-        }
-        // pass the error to the callback registered by the ping caller
-        callback.onError(ex);
-    }
-
-    private static void assertPluginWasRegistered(final String name, final JVppCallback value) {
-        if (value == null) {
-            throw new IllegalArgumentException(String.format("Callback for plugin %s is not registered", name));
-        }
-    }
-
-    @Override
-    public void close() throws Exception {
-        connection.close();
-    }
-}
diff --git a/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/NativeLibraryLoader.java b/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/NativeLibraryLoader.java
deleted file mode 100644
index ce6d1bfc..00000000
--- a/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/NativeLibraryLoader.java
+++ /dev/null
@@ -1,73 +0,0 @@
-/*
- * Copyright (c) 2016 Cisco and/or its affiliates.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at:
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-package io.fd.vpp.jvpp;
-
-import java.io.IOException;
-import java.io.InputStream;
-import java.nio.file.Files;
-import java.nio.file.Path;
-import java.nio.file.StandardCopyOption;
-import java.nio.file.attribute.PosixFilePermission;
-import java.nio.file.attribute.PosixFilePermissions;
-import java.util.Set;
-import java.util.logging.Level;
-import java.util.logging.Logger;
-
-/**
- * Utility class for loading JNI libraries.
- */
-public final class NativeLibraryLoader {
-
-    private static final Logger LOG = Logger.getLogger(NativeLibraryLoader.class.getName());
-
-    private NativeLibraryLoader() {
-        throw new UnsupportedOperationException("This utility class cannot be instantiated.");
-    }
-
-    /**
-     * Loads JNI library using class loader of the given class.
-     *
-     * @param libName name of the library to be loaded
-     */
-    public static void loadLibrary(final String libName, final Class clazz) throws IOException {
-        java.util.Objects.requireNonNull(libName, "libName should not be null");
-        java.util.Objects.requireNonNull(clazz, "clazz should not be null");
-        try (final InputStream is = clazz.getResourceAsStream('/' + libName)) {
-            if (is == null) {
-                throw new IOException("Failed to open library resource " + libName);
-            }
-            loadStream(libName, is);
-        }
-    }
-
-    private static void loadStream(final String libName, final InputStream is) throws IOException {
-        final Set perms = PosixFilePermissions.fromString("rwxr-x---");
-        final Path p = Files.createTempFile(libName, null, PosixFilePermissions.asFileAttribute(perms));
-        try {
-            Files.copy(is, p, StandardCopyOption.REPLACE_EXISTING);
-            Runtime.getRuntime().load(p.toString());
-        } catch (Exception e) {
-            throw new IOException("Failed to load library " + p, e);
-        } finally {
-            try {
-                Files.deleteIfExists(p);
-            } catch (IOException e) {
-                LOG.log(Level.WARNING, String.format("Failed to delete temporary file %s.", p), e);
-            }
-        }
-    }
-}
diff --git a/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/VppBaseCallException.java b/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/VppBaseCallException.java
deleted file mode 100644
index d71e3055..00000000
--- a/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/VppBaseCallException.java
+++ /dev/null
@@ -1,60 +0,0 @@
-/*
- * Copyright (c) 2016 Cisco and/or its affiliates.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at:
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-package io.fd.vpp.jvpp;
-
-/**
- * Base exception representing failed operation of JVpp request call
- */
-public abstract class VppBaseCallException extends Exception {
-    private final String methodName;
-    private final int errorCode;
-
-    /**
-     * Constructs an VppCallbackException with the specified api method name and error code.
-     *
-     * @param methodName name of a method, which invocation or execution failed
-     * @param errorCode  negative error code value associated with this failure
-     * @throws NullPointerException     if apiMethodName is null
-     */
-    public VppBaseCallException(final String methodName, final int errorCode) {
-        super(String.format("vppApi.%s failed with error code: %d", methodName, errorCode));
-        this.methodName = java.util.Objects.requireNonNull(methodName, "apiMethodName is null!");
-        this.errorCode = errorCode;
-        if(errorCode >= 0) {
-            throw new IllegalArgumentException("Error code must be < 0. Was " + errorCode +
-                    " for " + methodName );
-        }
-    }
-
-    /**
-     * Returns  name of a method, which invocation failed.
-     *
-     * @return method name
-     */
-    public String getMethodName() {
-        return methodName;
-    }
-
-    /**
-     * Returns the error code associated with this failure.
-     *
-     * @return a negative integer error code
-     */
-    public int getErrorCode() {
-        return errorCode;
-    }
-}
diff --git a/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/VppCallbackException.java b/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/VppCallbackException.java
deleted file mode 100644
index ccfcbd3c..00000000
--- a/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/VppCallbackException.java
+++ /dev/null
@@ -1,47 +0,0 @@
-/*
- * Copyright (c) 2016 Cisco and/or its affiliates.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at:
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-package io.fd.vpp.jvpp;
-
-/**
- * Callback Exception representing failed operation of JVpp request call
- */
-public class VppCallbackException extends VppBaseCallException {
-    private final int ctxId;
-
-    /**
-     * Constructs an VppCallbackException with the specified api method name and error code.
-     *
-     * @param methodName  name of a method, which invocation failed.
-     * @param ctxId      api request context identifier
-     * @param errorCode  negative error code value associated with this failure
-     * @throws NullPointerException     if apiMethodName is null
-     */
-    public VppCallbackException(final String methodName, final int ctxId, final int errorCode ){
-        super(methodName, errorCode);
-        this.ctxId = ctxId;
-    }
-
-    /**
-     * Returns api request context identifier.
-     *
-     * @return value of context identifier
-     */
-    public int getCtxId() {
-        return ctxId;
-    }
-
-}
diff --git a/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/VppConnection.java b/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/VppConnection.java
deleted file mode 100644
index e6fd3bdb..00000000
--- a/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/VppConnection.java
+++ /dev/null
@@ -1,45 +0,0 @@
-/*
- * Copyright (c) 2016 Cisco and/or its affiliates.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at:
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-package io.fd.vpp.jvpp;
-
-import java.io.IOException;
-
-/**
- * Representation of a management connection to VPP.
- */
-public interface VppConnection extends AutoCloseable {
-
-    /**
-     * Opens VppConnection for communication with VPP.
-     *
-     * @throws IOException if connection is not established
-     */
-    void connect() throws IOException;
-
-    /**
-     * Checks if this instance connection is active.
-     *
-     * @throws IllegalStateException if this instance was disconnected.
-     */
-    void checkActive();
-
-    /**
-     * Closes Vpp connection.
-     */
-    @Override
-    void close();
-}
diff --git a/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/VppInvocationException.java b/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/VppInvocationException.java
deleted file mode 100644
index a7ccb197..00000000
--- a/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/VppInvocationException.java
+++ /dev/null
@@ -1,33 +0,0 @@
-/*
- * Copyright (c) 2016 Cisco and/or its affiliates.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at:
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-package io.fd.vpp.jvpp;
-
-/**
- * Exception thrown when Vpp jAPI method invocation failed.
- */
-public class VppInvocationException extends VppBaseCallException {
-    /**
-     * Constructs an VppApiInvocationFailedException with the specified api method name and error code.
-     *
-     * @param methodName  name of a method, which invocation failed.
-     * @param errorCode  negative error code value associated with this failure
-     * @throws NullPointerException     if apiMethodName is null
-     */
-    public VppInvocationException(final String methodName, final int errorCode) {
-        super(methodName, errorCode);
-    }
-}
diff --git a/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/VppJNIConnection.java b/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/VppJNIConnection.java
deleted file mode 100644
index 7178bcf7..00000000
--- a/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/VppJNIConnection.java
+++ /dev/null
@@ -1,137 +0,0 @@
-/*
- * Copyright (c) 2016 Cisco and/or its affiliates.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at:
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-package io.fd.vpp.jvpp;
-
-import static io.fd.vpp.jvpp.NativeLibraryLoader.loadLibrary;
-
-import java.io.IOException;
-import java.util.HashMap;
-import java.util.Map;
-import java.util.Objects;
-import java.util.logging.Level;
-import java.util.logging.Logger;
-
-/**
- * JNI based representation of a management connection to VPP.
- */
-public final class VppJNIConnection implements VppConnection {
-    private static final Logger LOG = Logger.getLogger(VppJNIConnection.class.getName());
-
-    static {
-        final String libName = "libjvpp_registry.so.0.0.0";
-        try {
-            loadLibrary(libName, VppJNIConnection.class);
-        } catch (IOException e) {
-            LOG.log(Level.SEVERE, String.format("Can't find vpp jni library: %s", libName), e);
-            throw new ExceptionInInitializerError(e);
-        }
-    }
-
-    private ConnectionInfo connectionInfo;
-
-    private final String clientName;
-    private volatile boolean disconnected = false;
-
-    /**
-     * Create VPPJNIConnection instance for client connecting to VPP.
-     *
-     * @param clientName client name instance to be used for communication. Single connection per clientName is
-     *                   allowed.
-     */
-    public VppJNIConnection(final String clientName) {
-        this.clientName = Objects.requireNonNull(clientName, "Null clientName");
-    }
-
-    /**
-     * Guarded by VppJNIConnection.class
-     */
-    private static final Map connections = new HashMap<>();
-
-    /**
-     * Initiate VPP connection for current instance
-     *
-     * Multiple instances are allowed since this class is not a singleton (VPP allows multiple management connections).
-     *
-     * However only a single connection per clientName is allowed.
-     *
-     * @throws IOException in case the connection could not be established
-     */
-
-    @Override
-    public void connect() throws IOException {
-        _connect();
-    }
-
-    private void _connect() throws IOException {
-        synchronized (VppJNIConnection.class) {
-            if (connections.containsKey(clientName)) {
-                throw new IOException("Client " + clientName + " already connected");
-            }
-
-            connectionInfo = clientConnect(clientName);
-            if (connectionInfo.status != 0) {
-                throw new IOException("Connection returned error " + connectionInfo.status);
-            }
-            connections.put(clientName, this);
-        }
-    }
-
-    @Override
-    public final void checkActive() {
-        if (disconnected) {
-            throw new IllegalStateException("Disconnected client " + clientName);
-        }
-    }
-
-    @Override
-    public final synchronized void close() {
-        if (!disconnected) {
-            disconnected = true;
-            try {
-                clientDisconnect();
-            } finally {
-                synchronized (VppJNIConnection.class) {
-                    connections.remove(clientName);
-                }
-            }
-        }
-    }
-
-    public ConnectionInfo getConnectionInfo() {
-        return connectionInfo;
-    }
-
-    /**
-     * VPP connection information used by plugins to reuse the connection.
-     */
-    public static final class ConnectionInfo {
-        public final long queueAddress;
-        public final int clientIndex;
-        public final int status; // FIXME throw exception instead
-
-        public ConnectionInfo(long queueAddress, int clientIndex, int status) {
-            this.queueAddress = queueAddress;
-            this.clientIndex = clientIndex;
-            this.status = status;
-        }
-    }
-
-    private static native ConnectionInfo clientConnect(String clientName);
-
-    private static native void clientDisconnect();
-
-}
diff --git a/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/callback/ControlPingCallback.java b/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/callback/ControlPingCallback.java
deleted file mode 100644
index efddfdbb..00000000
--- a/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/callback/ControlPingCallback.java
+++ /dev/null
@@ -1,29 +0,0 @@
-/*
- * Copyright (c) 2016 Cisco and/or its affiliates.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at:
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-package io.fd.vpp.jvpp.callback;
-
-import io.fd.vpp.jvpp.dto.ControlPingReply;
-
-/**
- * Represents callback for control_ping message.
- */
-public interface ControlPingCallback extends JVppCallback {
-
-    void onControlPingReply(ControlPingReply reply);
-
-}
-
diff --git a/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/callback/JVppCallback.java b/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/callback/JVppCallback.java
deleted file mode 100644
index ae02063b..00000000
--- a/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/callback/JVppCallback.java
+++ /dev/null
@@ -1,29 +0,0 @@
-/*
- * Copyright (c) 2016 Cisco and/or its affiliates.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at:
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-package io.fd.vpp.jvpp.callback;
-import io.fd.vpp.jvpp.VppCallbackException;
-
-/**
- * Base JVppCallback interface
- */
-public interface JVppCallback {
-    /**
-     * onError callback handler used to report failing operation
-     * @param ex VppCallbackException object containing details about failing operation
-     */
-    void onError(VppCallbackException ex);
-}
diff --git a/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/callback/JVppNotificationCallback.java b/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/callback/JVppNotificationCallback.java
deleted file mode 100644
index 8ab0cb21..00000000
--- a/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/callback/JVppNotificationCallback.java
+++ /dev/null
@@ -1,24 +0,0 @@
-/*
- * Copyright (c) 2016 Cisco and/or its affiliates.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at:
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-package io.fd.vpp.jvpp.callback;
-
-/**
-* Notification callback
-*/
-public interface JVppNotificationCallback {
-
-}
diff --git a/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/dto/ControlPing.java b/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/dto/ControlPing.java
deleted file mode 100644
index 984e1674..00000000
--- a/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/dto/ControlPing.java
+++ /dev/null
@@ -1,34 +0,0 @@
-/*
- * Copyright (c) 2016 Cisco and/or its affiliates.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at:
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-package io.fd.vpp.jvpp.dto;
-
-import io.fd.vpp.jvpp.JVpp;
-import io.fd.vpp.jvpp.VppInvocationException;
-
-/**
- * Represents request DTO for control_ping message.
- */
-public final class ControlPing implements JVppRequest {
-
-    @Override
-    public int send(final JVpp jvpp) throws VppInvocationException {
-        return jvpp.controlPing(this);
-    }
-
-}
-
-
diff --git a/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/dto/ControlPingReply.java b/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/dto/ControlPingReply.java
deleted file mode 100644
index 61e4d0e4..00000000
--- a/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/dto/ControlPingReply.java
+++ /dev/null
@@ -1,58 +0,0 @@
-/*
- * Copyright (c) 2016 Cisco and/or its affiliates.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at:
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-package io.fd.vpp.jvpp.dto;
-
-import java.util.Objects;
-
-/**
- * Represents reply DTO for control_ping message.
- */
-public final class ControlPingReply implements JVppReply {
-
-    public int context;
-    public int clientIndex;
-    public int vpePid;
-
-    @Override
-    public boolean equals(final Object o) {
-        if (this == o) {
-            return true;
-        }
-        if (o == null || getClass() != o.getClass()) {
-            return false;
-        }
-        final ControlPingReply that = (ControlPingReply) o;
-        return context == that.context &&
-                clientIndex == that.clientIndex &&
-                vpePid == that.vpePid;
-    }
-
-    @Override
-    public int hashCode() {
-        return Objects.hash(context, clientIndex, vpePid);
-    }
-
-    @Override
-    public String toString() {
-        return "ControlPingReply{" +
-                "context=" + context +
-                ", clientIndex=" + clientIndex +
-                ", vpePid=" + vpePid +
-                '}';
-    }
-}
-
diff --git a/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/dto/JVppDump.java b/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/dto/JVppDump.java
deleted file mode 100644
index 60b98984..00000000
--- a/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/dto/JVppDump.java
+++ /dev/null
@@ -1,24 +0,0 @@
-/*
- * Copyright (c) 2016 Cisco and/or its affiliates.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at:
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-package io.fd.vpp.jvpp.dto;
-
-/**
-* Base interface for all dump requests
-*/
-public interface JVppDump extends JVppRequest {
-
-}
diff --git a/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/dto/JVppNotification.java b/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/dto/JVppNotification.java
deleted file mode 100644
index 5554f501..00000000
--- a/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/dto/JVppNotification.java
+++ /dev/null
@@ -1,23 +0,0 @@
-/*
- * Copyright (c) 2016 Cisco and/or its affiliates.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at:
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-package io.fd.vpp.jvpp.dto;
-
-/**
-* Base interface for all notification DTOs
-*/
-public interface JVppNotification {
-}
diff --git a/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/dto/JVppReply.java b/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/dto/JVppReply.java
deleted file mode 100644
index 73f512d4..00000000
--- a/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/dto/JVppReply.java
+++ /dev/null
@@ -1,24 +0,0 @@
-/*
- * Copyright (c) 2016 Cisco and/or its affiliates.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at:
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-package io.fd.vpp.jvpp.dto;
-
-/**
-* Base interface for all reply DTOs
-*/
-public interface JVppReply {
-
-}
diff --git a/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/dto/JVppReplyDump.java b/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/dto/JVppReplyDump.java
deleted file mode 100644
index 15111395..00000000
--- a/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/dto/JVppReplyDump.java
+++ /dev/null
@@ -1,25 +0,0 @@
-/*
- * Copyright (c) 2016 Cisco and/or its affiliates.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at:
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-package io.fd.vpp.jvpp.dto;
-
-/**
-* Base interface for all dump replies
-*/
-public interface JVppReplyDump>
-    extends JVppReply {
-
-}
diff --git a/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/dto/JVppRequest.java b/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/dto/JVppRequest.java
deleted file mode 100644
index 9b301da2..00000000
--- a/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/dto/JVppRequest.java
+++ /dev/null
@@ -1,34 +0,0 @@
-/*
- * Copyright (c) 2016 Cisco and/or its affiliates.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at:
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-package io.fd.vpp.jvpp.dto;
-
-import io.fd.vpp.jvpp.JVpp;
-import io.fd.vpp.jvpp.VppInvocationException;
-
-/**
-* Base interface for all request DTOs
-*/
-public interface JVppRequest {
-
-    /**
-     * Invoke current operation asynchronously on VPP
-     *
-     * @return context id of this request. Can be used to track incoming response
-     */
-    int send(JVpp jvpp) throws VppInvocationException;
-
-}
diff --git a/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/future/AbstractFutureJVppInvoker.java b/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/future/AbstractFutureJVppInvoker.java
deleted file mode 100644
index e7df528a..00000000
--- a/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/future/AbstractFutureJVppInvoker.java
+++ /dev/null
@@ -1,141 +0,0 @@
-/*
- * Copyright (c) 2016 Cisco and/or its affiliates.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at:
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-package io.fd.vpp.jvpp.future;
-
-
-import java.util.Map;
-import java.util.Objects;
-import java.util.concurrent.CompletableFuture;
-import java.util.concurrent.CompletionStage;
-import io.fd.vpp.jvpp.JVpp;
-import io.fd.vpp.jvpp.JVppRegistry;
-import io.fd.vpp.jvpp.VppInvocationException;
-import io.fd.vpp.jvpp.dto.JVppDump;
-import io.fd.vpp.jvpp.dto.JVppReply;
-import io.fd.vpp.jvpp.dto.JVppReplyDump;
-import io.fd.vpp.jvpp.dto.JVppRequest;
-
-/**
- * Future facade on top of JVpp
- */
-public abstract class AbstractFutureJVppInvoker implements FutureJVppInvoker {
-
-    private final JVpp jvpp;
-    private final JVppRegistry registry;
-
-    /**
-     * Guarded by self
-     */
-    private final Map>> requests;
-
-    protected AbstractFutureJVppInvoker(final JVpp jvpp, final JVppRegistry registry,
-                                        final Map>> requestMap) {
-        this.jvpp =  Objects.requireNonNull(jvpp, "jvpp should not be null");
-        this.registry =  Objects.requireNonNull(registry, "registry should not be null");
-        // Request map represents the shared state between this facade and it's callback
-        // where facade puts futures in and callback completes + removes them
-        this.requests = Objects.requireNonNull(requestMap, "Null requestMap");
-    }
-
-    protected final Map>> getRequests() {
-        synchronized (requests) {
-            return requests;
-        }
-    }
-
-    // TODO use Optional in Future, java8
-
-    @Override
-    @SuppressWarnings("unchecked")
-    public > CompletionStage send(REQ req) {
-        synchronized(requests) {
-            try {
-                final CompletableFuture replyCompletableFuture;
-                final int contextId = jvpp.send(req);
-
-                if(req instanceof JVppDump) {
-                    throw new IllegalArgumentException("Send with empty reply dump has to be used in case of dump calls");
-                }
-                replyCompletableFuture = new CompletableFuture<>();
-                requests.put(contextId, replyCompletableFuture);
-
-                // TODO in case of timeouts/missing replies, requests from the map are not removed
-                // consider adding cancel method, that would remove requests from the map and cancel
-                // associated replyCompletableFuture
-
-                return replyCompletableFuture;
-            } catch (VppInvocationException ex) {
-                final CompletableFuture replyCompletableFuture = new CompletableFuture<>();
-                replyCompletableFuture.completeExceptionally(ex);
-                return replyCompletableFuture;
-            }
-        }
-    }
-
-    @Override
-    @SuppressWarnings("unchecked")
-    public , DUMP extends JVppReplyDump> CompletionStage send(
-            REQ req, DUMP emptyReplyDump) {
-        synchronized(requests) {
-            try {
-                final CompletableDumpFuture replyCompletableFuture;
-                final int contextId = jvpp.send(req);
-
-                if(!(req instanceof JVppDump)) {
-                    throw new IllegalArgumentException("Send without empty reply dump has to be used in case of regular calls");
-                }
-                replyCompletableFuture = new CompletableDumpFuture<>(contextId, emptyReplyDump);
-
-                requests.put(contextId, replyCompletableFuture);
-                requests.put(registry.controlPing(jvpp.getClass()), replyCompletableFuture);
-
-                // TODO in case of timeouts/missing replies, requests from the map are not removed
-                // consider adding cancel method, that would remove requests from the map and cancel
-                // associated replyCompletableFuture
-
-                return replyCompletableFuture;
-            } catch (VppInvocationException ex) {
-                final CompletableFuture replyCompletableFuture = new CompletableFuture<>();
-                replyCompletableFuture.completeExceptionally(ex);
-                return replyCompletableFuture;
-            }
-        }
-    }
-
-    public static final class CompletableDumpFuture> extends CompletableFuture {
-        private final T replyDump;
-        private final int contextId;
-
-        public CompletableDumpFuture(final int contextId, final T emptyDump) {
-            this.contextId = contextId;
-            this.replyDump = emptyDump;
-        }
-
-        public int getContextId() {
-            return contextId;
-        }
-
-        public T getReplyDump() {
-            return replyDump;
-        }
-    }
-
-    @Override
-    public void close() throws Exception {
-        jvpp.close();
-    }
-}
diff --git a/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/future/FutureJVppInvoker.java b/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/future/FutureJVppInvoker.java
deleted file mode 100644
index 7a48e418..00000000
--- a/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/future/FutureJVppInvoker.java
+++ /dev/null
@@ -1,49 +0,0 @@
-/*
- * Copyright (c) 2016 Cisco and/or its affiliates.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at:
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-package io.fd.vpp.jvpp.future;
-
-
-import io.fd.vpp.jvpp.dto.JVppReply;
-import io.fd.vpp.jvpp.dto.JVppReplyDump;
-import io.fd.vpp.jvpp.dto.JVppRequest;
-
-import java.util.concurrent.CompletionStage;
-import io.fd.vpp.jvpp.notification.NotificationRegistryProvider;
-
-/**
-* Future facade on top of JVpp
-*/
-public interface FutureJVppInvoker extends NotificationRegistryProvider, AutoCloseable {
-
-    /**
-     * Invoke asynchronous operation on VPP
-     *
-     * @return CompletionStage with future result of an async VPP call
-     * @throws io.fd.vpp.jvpp.VppInvocationException when send request failed with details
-     */
-    > CompletionStage send(REQ req);
-
-
-    /**
-     * Invoke asynchronous dump operation on VPP
-     *
-     * @return CompletionStage with aggregated future result of an async VPP dump call
-     * @throws io.fd.vpp.jvpp.VppInvocationException when send request failed with details
-     */
-    , DUMP extends JVppReplyDump> CompletionStage send(
-            REQ req, DUMP emptyReplyDump);
-}
diff --git a/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/notification/NotificationRegistry.java b/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/notification/NotificationRegistry.java
deleted file mode 100644
index 3c72ff79..00000000
--- a/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/notification/NotificationRegistry.java
+++ /dev/null
@@ -1,25 +0,0 @@
-/*
- * Copyright (c) 2016 Cisco and/or its affiliates.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at:
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-package io.fd.vpp.jvpp.notification;
-
-/**
- * Base registry for notification callbacks.
- */
-public interface NotificationRegistry extends AutoCloseable {
-
-    void close();
-}
diff --git a/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/notification/NotificationRegistryProvider.java b/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/notification/NotificationRegistryProvider.java
deleted file mode 100644
index 4a6e06b7..00000000
--- a/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/notification/NotificationRegistryProvider.java
+++ /dev/null
@@ -1,28 +0,0 @@
-/*
- * Copyright (c) 2016 Cisco and/or its affiliates.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at:
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-package io.fd.vpp.jvpp.notification;
-
-/**
- * Provides notification registry
- */
-public interface NotificationRegistryProvider {
-
-    /**
-     * Get current notification registry instance
-     */
-    NotificationRegistry getNotificationRegistry();
-}
diff --git a/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/test/ConnectionTest.java b/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/test/ConnectionTest.java
deleted file mode 100644
index 27b4d29f..00000000
--- a/vpp-api/java/jvpp-registry/io/fd/vpp/jvpp/test/ConnectionTest.java
+++ /dev/null
@@ -1,44 +0,0 @@
-/*
- * Copyright (c) 2016 Cisco and/or its affiliates.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at:
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-package io.fd.vpp.jvpp.test;
-
-import io.fd.vpp.jvpp.JVppRegistry;
-import io.fd.vpp.jvpp.JVppRegistryImpl;
-
-/**
- * Run using:
- * sudo java -cp build-vpp-native/vpp-api/java/jvpp-registry-16.09.jar io.fd.vpp.jvpp.test.ConnectionTest
- */
-public class ConnectionTest {
-
-    private static void testConnect() throws Exception {
-        System.out.println("Testing JNI connection with JVppRegistry");
-        final JVppRegistry registry = new JVppRegistryImpl("ConnectionTest");
-        try {
-            System.out.println("Successfully connected to vpp");
-            Thread.sleep(5000);
-            System.out.println("Disconnecting...");
-            Thread.sleep(1000);
-        } finally {
-            registry.close();
-        }
-    }
-
-    public static void main(String[] args) throws Exception {
-        testConnect();
-    }
-}
diff --git a/vpp-api/java/jvpp-registry/jvpp_registry.c b/vpp-api/java/jvpp-registry/jvpp_registry.c
deleted file mode 100644
index cbd5e0ab..00000000
--- a/vpp-api/java/jvpp-registry/jvpp_registry.c
+++ /dev/null
@@ -1,352 +0,0 @@
-/*
- * Copyright (c) 2016 Cisco and/or its affiliates.
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at:
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-#define _GNU_SOURCE /* for strcasestr(3) */
-#include 
-
-#define vl_api_version(n,v) static u32 vpe_api_version = (v);
-#include 
-#undef vl_api_version
-
-
-#include 
-#include 
-#include "io_fd_vpp_jvpp_VppJNIConnection.h"
-#include "io_fd_vpp_jvpp_JVppRegistryImpl.h"
-
-#include 
-#define vl_typedefs             /* define message structures */
-#include 
-#undef vl_typedefs
-
-#define vl_endianfun
-#include 
-#undef vl_endianfun
-
-/* instantiate all the print functions we know about */
-#define vl_print(handle, ...)
-#define vl_printfun
-#include 
-#undef vl_printfun
-
-vlib_main_t vlib_global_main;
-vlib_main_t **vlib_mains;
-
-/*
- * The Java runtime isn't compile w/ -fstack-protector,
- * so we have to supply missing external references for the
- * regular vpp libraries.
- */
-void __stack_chk_guard(void) __attribute__((weak));
-void __stack_chk_guard(void) {
-}
-
-typedef struct {
-    /* UThread attachment */
-    volatile u32 control_ping_result_ready;
-    volatile i32 control_ping_retval;
-
-    /* Control poing callback */
-    jobject registryObject;
-    jclass registryClass;
-    jclass controlPingReplyClass;
-    jclass callbackExceptionClass;
-
-    /* Thread cleanup */
-    pthread_key_t cleanup_rx_thread_key;
-
-    /* Connected indication */
-    volatile u8 is_connected;
-} jvpp_registry_main_t;
-
-jvpp_registry_main_t jvpp_registry_main __attribute__((aligned (64)));
-
-void vl_client_add_api_signatures(vl_api_memclnt_create_t *mp) {
-    /*
-     * Send the main API signature in slot 0. This bit of code must
-     * match the checks in ../vpe/api/api.c: vl_msg_api_version_check().
-     */
-    mp->api_versions[0] = clib_host_to_net_u32(vpe_api_version);
-}
-
-/* cleanup handler for RX thread */
-static_always_inline void cleanup_rx_thread(void *arg) {
-    jvpp_main_t * jm = &jvpp_main;
-    jvpp_registry_main_t * rm = &jvpp_registry_main;
-
-    vppjni_lock(jm, 99);
-
-    int getEnvStat = (*jm->jvm)->GetEnv(jm->jvm, (void **) &(jm->jenv),
-    JNI_VERSION_1_8);
-    if (getEnvStat == JNI_EVERSION) {
-        clib_warning("Unsupported JNI version\n");
-        rm->control_ping_retval = VNET_API_ERROR_UNSUPPORTED_JNI_VERSION;
-        goto out;
-    } else if (getEnvStat != JNI_EDETACHED) {
-        (*jm->jvm)->DetachCurrentThread(jm->jvm);
-    }
-    out: vppjni_unlock(jm);
-}
-
-static void vl_api_control_ping_reply_t_handler(
-        vl_api_control_ping_reply_t * mp) {
-    jvpp_main_t * jm = &jvpp_main;
-    jvpp_registry_main_t * rm = &jvpp_registry_main;
-    char was_thread_connected = 0;
-
-    // attach to java thread if not attached
-    int getEnvStat = (*jm->jvm)->GetEnv(jm->jvm, (void **) &(jm->jenv),
-            JNI_VERSION_1_8);
-    if (getEnvStat == JNI_EDETACHED) {
-        if ((*jm->jvm)->AttachCurrentThread(jm->jvm, (void **) &(jm->jenv),
-                NULL) != 0) {
-            clib_warning("Failed to attach thread\n");
-            rm->control_ping_retval =
-                    VNET_API_ERROR_FAILED_TO_ATTACH_TO_JAVA_THREAD;
-            goto out;
-        }
-
-        // workaround as we can't use pthread_cleanup_push
-        pthread_key_create(&rm->cleanup_rx_thread_key, cleanup_rx_thread);
-        // destructor is only called if the value of key is non null
-        pthread_setspecific(rm->cleanup_rx_thread_key, (void *) 1);
-        was_thread_connected = 1;
-    } else if (getEnvStat == JNI_EVERSION) {
-        clib_warning("Unsupported JNI version\n");
-        rm->control_ping_retval = VNET_API_ERROR_UNSUPPORTED_JNI_VERSION;
-        goto out;
-    }
-
-    if (was_thread_connected == 0) {
-        JNIEnv *env = jm->jenv;
-        if (mp->retval < 0) {
-            call_on_error("controlPing", mp->context, mp->retval,
-                    rm->registryClass, rm->registryObject,
-                    rm->callbackExceptionClass);
-        } else {
-            jmethodID constructor = (*env)->GetMethodID(env,
-                    rm->controlPingReplyClass, "", "()V");
-            jmethodID callbackMethod = (*env)->GetMethodID(env,
-                    rm->registryClass, "onControlPingReply",
-                    "(Lio/fd/vpp/jvpp/dto/ControlPingReply;)V");
-
-            jobject dto = (*env)->NewObject(env, rm->controlPingReplyClass,
-                    constructor);
-
-            jfieldID contextFieldId = (*env)->GetFieldID(env,
-                    rm->controlPingReplyClass, "context", "I");
-            (*env)->SetIntField(env, dto, contextFieldId,
-                    clib_net_to_host_u32(mp->context));
-
-            jfieldID clientIndexFieldId = (*env)->GetFieldID(env,
-                    rm->controlPingReplyClass, "clientIndex", "I");
-            (*env)->SetIntField(env, dto, clientIndexFieldId,
-                    clib_net_to_host_u32(mp->client_index));
-
-            jfieldID vpePidFieldId = (*env)->GetFieldID(env,
-                    rm->controlPingReplyClass, "vpePid", "I");
-            (*env)->SetIntField(env, dto, vpePidFieldId,
-                    clib_net_to_host_u32(mp->vpe_pid));
-
-            (*env)->CallVoidMethod(env, rm->registryObject, callbackMethod,
-                    dto);
-            (*env)->DeleteLocalRef(env, dto);
-        }
-    }
-
-    out: rm->control_ping_result_ready = 1;
-}
-
-static int send_initial_control_ping() {
-    f64 timeout;
-    clib_time_t clib_time;
-    vl_api_control_ping_t * mp;
-    jvpp_main_t * jm = &jvpp_main;
-    jvpp_registry_main_t * rm = &jvpp_registry_main;
-
-    clib_time_init(&clib_time);
-
-    rm->control_ping_result_ready = 0;
-    mp = vl_msg_api_alloc(sizeof(*mp));
-    memset(mp, 0, sizeof(*mp));
-    mp->_vl_msg_id = ntohs(VL_API_CONTROL_PING);
-    mp->client_index = jm->my_client_index;
-
-    // send message:
-    vl_msg_api_send_shmem(jm->vl_input_queue, (u8 *) &mp);
-
-    // wait for results: Current time + 10 seconds is the timeout
-    timeout = clib_time_now(&clib_time) + 10.0;
-    int rv = VNET_API_ERROR_RESPONSE_NOT_READY;
-    while (clib_time_now(&clib_time) < timeout) {
-        if (rm->control_ping_result_ready == 1) {
-            rv = rm->control_ping_retval;
-            break;
-        }
-    }
-
-    if (rv != 0) {
-        clib_warning("common: first control ping failed: %d", rv);
-    }
-
-    return rv;
-}
-
-static int connect_to_vpe(char *name) {
-    jvpp_main_t * jm = &jvpp_main;
-    api_main_t * am = &api_main;
-
-    if (vl_client_connect_to_vlib("/vpe-api", name, 32) < 0)
-        return -1;
-
-    jm->my_client_index = am->my_client_index;
-
-    jm->vl_input_queue = am->shmem_hdr->vl_input_queue;
-
-    vl_msg_api_set_handlers(VL_API_CONTROL_PING_REPLY, "control_ping_reply",
-            vl_api_control_ping_reply_t_handler, vl_noop_handler,
-            vl_api_control_ping_reply_t_endian,
-            vl_api_control_ping_reply_t_print,
-            sizeof(vl_api_control_ping_reply_t), 1);
-
-    return send_initial_control_ping();
-}
-
-JNIEXPORT jobject JNICALL Java_io_fd_vpp_jvpp_VppJNIConnection_clientConnect(
-        JNIEnv *env, jclass obj, jstring clientName) {
-    int rv;
-    const char *client_name;
-    void vl_msg_reply_handler_hookup(void);
-    jvpp_main_t * jm = &jvpp_main;
-    jvpp_registry_main_t * rm = &jvpp_registry_main;
-
-    jclass connectionInfoClass = (*env)->FindClass(env,
-            "io/fd/vpp/jvpp/VppJNIConnection$ConnectionInfo");
-    jmethodID connectionInfoConstructor = (*env)->GetMethodID(env,
-            connectionInfoClass, "", "(JII)V");
-
-    /*
-     * Bail out now if we're not running as root
-     */
-    if (geteuid() != 0) {
-        return (*env)->NewObject(env, connectionInfoClass,
-                connectionInfoConstructor, 0, 0,
-                VNET_API_ERROR_NOT_RUNNING_AS_ROOT);
-    }
-
-    if (rm->is_connected) {
-        return (*env)->NewObject(env, connectionInfoClass,
-                connectionInfoConstructor, 0, 0,
-                VNET_API_ERROR_ALREADY_CONNECTED);
-    }
-
-    client_name = (*env)->GetStringUTFChars(env, clientName, 0);
-    if (!client_name) {
-        return (*env)->NewObject(env, connectionInfoClass,
-                connectionInfoConstructor, 0, 0, VNET_API_ERROR_INVALID_VALUE);
-    }
-
-    rv = connect_to_vpe((char *) client_name);
-
-    if (rv < 0)
-        clib_warning("connection failed, rv %d", rv);
-
-    (*env)->ReleaseStringUTFChars(env, clientName, client_name);
-
-    return (*env)->NewObject(env, connectionInfoClass,
-            connectionInfoConstructor, (jlong) jm->vl_input_queue,
-            (jint) jm->my_client_index, (jint) rv);
-}
-
-JNIEXPORT jint JNICALL Java_io_fd_vpp_jvpp_JVppRegistryImpl_controlPing0(
-        JNIEnv *env, jobject regstryObject) {
-    jvpp_main_t * jm = &jvpp_main;
-    vl_api_control_ping_t * mp;
-    u32 my_context_id = vppjni_get_context_id(&jvpp_main);
-    jvpp_registry_main_t * rm = &jvpp_registry_main;
-
-    if (rm->registryObject == 0) {
-        rm->registryObject = (*env)->NewGlobalRef(env, regstryObject);
-    }
-    if (rm->registryClass == 0) {
-        rm->registryClass = (jclass) (*env)->NewGlobalRef(env,
-                (*env)->GetObjectClass(env, regstryObject));
-    }
-
-    mp = vl_msg_api_alloc(sizeof(*mp));
-    memset(mp, 0, sizeof(*mp));
-    mp->_vl_msg_id = ntohs(VL_API_CONTROL_PING);
-    mp->client_index = jm->my_client_index;
-    mp->context = clib_host_to_net_u32(my_context_id);
-
-    // send message:
-    vl_msg_api_send_shmem(jm->vl_input_queue, (u8 *) &mp);
-    return my_context_id;
-}
-
-JNIEXPORT void JNICALL Java_io_fd_vpp_jvpp_VppJNIConnection_clientDisconnect(
-        JNIEnv *env, jclass clazz) {
-    jvpp_registry_main_t * rm = &jvpp_registry_main;
-    rm->is_connected = 0; // TODO make thread safe
-    vl_client_disconnect_from_vlib();
-
-    // cleanup:
-    if (rm->registryObject) {
-        (*env)->DeleteGlobalRef(env, rm->registryObject);
-        rm->registryObject = 0;
-    }
-    if (rm->registryClass) {
-        (*env)->DeleteGlobalRef(env, rm->registryClass);
-        rm->registryClass = 0;
-    }
-}
-
-jint JNI_OnLoad(JavaVM *vm, void *reserved) {
-    jvpp_main_t * jm = &jvpp_main;
-    jvpp_registry_main_t * rm = &jvpp_registry_main;
-    JNIEnv* env;
-
-    if ((*vm)->GetEnv(vm, (void**) &env, JNI_VERSION_1_8) != JNI_OK) {
-        return JNI_EVERSION;
-    }
-
-    rm->controlPingReplyClass = (jclass) (*env)->NewGlobalRef(env,
-            (*env)->FindClass(env, "io/fd/vpp/jvpp/dto/ControlPingReply"));
-    if ((*env)->ExceptionCheck(env)) {
-        (*env)->ExceptionDescribe(env);
-        clib_warning("Failed to cache class references\n");
-        return JNI_ERR;
-    }
-
-    rm->callbackExceptionClass = (jclass) (*env)->NewGlobalRef(env,
-            (*env)->FindClass(env, "io/fd/vpp/jvpp/VppCallbackException"));
-    if ((*env)->ExceptionCheck(env)) {
-        (*env)->ExceptionDescribe(env);
-        return JNI_ERR;
-    }
-
-    jm->jvm = vm;
-    return JNI_VERSION_1_8;
-}
-
-void JNI_OnUnload(JavaVM *vm, void *reserved) {
-    jvpp_main_t * jm = &jvpp_main;
-    JNIEnv* env;
-    if ((*vm)->GetEnv(vm, (void**) &env, JNI_VERSION_1_8) != JNI_OK) {
-        return;
-    }
-
-    jm->jenv = NULL;
-    jm->jvm = NULL;
-}
diff --git a/vpp-api/java/jvpp/gen/jvpp_gen.py b/vpp-api/java/jvpp/gen/jvpp_gen.py
deleted file mode 100755
index 5f8df2a9..00000000
--- a/vpp-api/java/jvpp/gen/jvpp_gen.py
+++ /dev/null
@@ -1,171 +0,0 @@
-#!/usr/bin/env python
-#
-# Copyright (c) 2016 Cisco and/or its affiliates.
-# Licensed under the Apache License, Version 2.0 (the "License");
-# you may not use this file except in compliance with the License.
-# You may obtain a copy of the License at:
-#
-#     http://www.apache.org/licenses/LICENSE-2.0
-# l
-# Unless required by applicable law or agreed to in writing, software
-# distributed under the License is distributed on an "AS IS" BASIS,
-# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-# See the License for the specific language governing permissions and
-# limitations under the License.
-#
-
-import argparse
-import importlib
-import sys
-import os
-import json
-
-from jvppgen import types_gen
-from jvppgen import callback_gen
-from jvppgen import notification_gen
-from jvppgen import dto_gen
-from jvppgen import jvpp_callback_facade_gen
-from jvppgen import jvpp_future_facade_gen
-from jvppgen import jvpp_impl_gen
-from jvppgen import jvpp_c_gen
-from jvppgen import util
-
-blacklist = [ "memclnt.api", "flowperpkt.api" ]
-
-# Invocation:
-# ~/Projects/vpp/vpp-api/jvpp/gen$ mkdir -p java/io/fd/vpp/jvpp && cd java/io/fd/vpp/jvpp
-# ~/Projects/vpp/vpp-api/jvpp/gen/java/io/fd/vpp/jvpp$ ../../../../jvpp_gen.py -idefs_api_vpp_papi.py
-#
-# Compilation:
-# ~/Projects/vpp/vpp-api/jvpp/gen/java/io/fd/vpp/jvpp$ javac *.java dto/*.java callback/*.java
-#
-# where
-# defs_api_vpp_papi.py - vpe.api in python format (generated by vppapigen)
-
-parser = argparse.ArgumentParser(description='VPP Java API generator')
-parser.add_argument('-i', action="store", dest="inputfiles", nargs='+')
-parser.add_argument('--plugin_name', action="store", dest="plugin_name")
-args = parser.parse_args()
-
-sys.path.append(".")
-
-print "Generating Java API for %s" % args.inputfiles
-print "inputfiles %s" % args.inputfiles
-plugin_name = args.plugin_name
-print "plugin_name %s" % plugin_name
-
-cfg = {}
-
-for inputfile in args.inputfiles:
-    if any(substring in inputfile for substring in blacklist):
-        print "WARNING: Imput file %s blacklisted" % inputfile
-        continue
-    _cfg = json.load(open(inputfile, 'r'))
-    if 'types' in cfg:
-        cfg['types'].extend(_cfg['types'])
-    else:
-        cfg['types'] = _cfg['types']
-    if 'messages' in cfg:
-        cfg['messages'].extend(_cfg['messages'])
-    else:
-        cfg['messages'] = _cfg['messages']
-
-
-def is_request_field(field_name):
-    return field_name not in {'_vl_msg_id', 'client_index', 'context'}
-
-
-def is_response_field(field_name):
-    return field_name not in {'_vl_msg_id'}
-
-
-def get_args(t, filter):
-    arg_list = []
-    for i in t:
-        if is_crc(i):
-            continue
-        if not filter(i[1]):
-            continue
-        arg_list.append(i[1])
-    return arg_list
-
-
-def get_types(t, filter):
-    types_list = []
-    lengths_list = []
-    crc = None
-    for i in t:
-        if is_crc(i):
-            crc = ('crc', i['crc'][2:])
-            continue
-        if not filter(i[1]):
-            continue
-        if len(i) is 3:  # array type
-            types_list.append(i[0] + '[]')
-            lengths_list.append((i[2], False))
-        elif len(i) is 4:  # variable length array type
-            types_list.append(i[0] + '[]')
-            lengths_list.append((i[3], True))
-        else:  # primitive type
-            types_list.append(i[0])
-            lengths_list.append((0, False))
-    return types_list, lengths_list, crc
-
-
-def is_crc(arg):
-    """ Check whether the argument inside message definition is just crc """
-    return 'crc' in arg
-
-
-def get_definitions(defs):
-    # Pass 1
-    func_list = []
-    func_name = {}
-    for a in defs:
-        java_name = util.underscore_to_camelcase(a[0])
-
-        # For replies include all the arguments except message_id
-        if util.is_reply(java_name):
-            types, lengths, crc = get_types(a[1:], is_response_field)
-            func_name[a[0]] = dict(
-                [('name', a[0]), ('java_name', java_name),
-                 ('args', get_args(a[1:], is_response_field)), ('full_args', get_args(a[1:], lambda x: True)),
-                 ('types', types), ('lengths', lengths), crc])
-        # For requests skip message_id, client_id and context
-        else:
-            types, lengths, crc = get_types(a[1:], is_request_field)
-            func_name[a[0]] = dict(
-                [('name', a[0]), ('java_name', java_name),
-                 ('args', get_args(a[1:], is_request_field)), ('full_args', get_args(a[1:], lambda x: True)),
-                 ('types', types), ('lengths', lengths), crc])
-
-        # Indexed by name
-        func_list.append(func_name[a[0]])
-    return func_list, func_name
-
-
-base_package = 'io.fd.vpp.jvpp'
-plugin_package = base_package + '.' + plugin_name
-types_package = 'types'
-dto_package = 'dto'
-callback_package = 'callback'
-notification_package = 'notification'
-future_package = 'future'
-# TODO find better package name
-callback_facade_package = 'callfacade'
-
-types_list, types_name = get_definitions(cfg['types'])
-
-types_gen.generate_types(types_list, plugin_package, types_package, args.inputfiles)
-
-func_list, func_name = get_definitions(cfg['messages'])
-
-dto_gen.generate_dtos(func_list, base_package, plugin_package, plugin_name.title(), dto_package, args.inputfiles)
-jvpp_impl_gen.generate_jvpp(func_list, base_package, plugin_package, plugin_name, dto_package, args.inputfiles)
-callback_gen.generate_callbacks(func_list, base_package, plugin_package, plugin_name.title(), callback_package, dto_package, args.inputfiles)
-notification_gen.generate_notification_registry(func_list, base_package, plugin_package, plugin_name.title(), notification_package, callback_package, dto_package, args.inputfiles)
-jvpp_c_gen.generate_jvpp(func_list, plugin_name, args.inputfiles)
-jvpp_future_facade_gen.generate_jvpp(func_list, base_package, plugin_package, plugin_name.title(), dto_package, callback_package, notification_package, future_package, args.inputfiles)
-jvpp_callback_facade_gen.generate_jvpp(func_list, base_package, plugin_package, plugin_name.title(), dto_package, callback_package, notification_package, callback_facade_package, args.inputfiles)
-
-print "Java API for %s generated successfully" % args.inputfiles
diff --git a/vpp-api/java/jvpp/gen/jvppgen/__init__.py b/vpp-api/java/jvpp/gen/jvppgen/__init__.py
deleted file mode 100644
index e69de29b..00000000
diff --git a/vpp-api/java/jvpp/gen/jvppgen/callback_gen.py b/vpp-api/java/jvpp/gen/jvppgen/callback_gen.py
deleted file mode 100644
index 68f70126..00000000
--- a/vpp-api/java/jvpp/gen/jvppgen/callback_gen.py
+++ /dev/null
@@ -1,105 +0,0 @@
-#!/usr/bin/env python
-#
-# Copyright (c) 2016 Cisco and/or its affiliates.
-# Licensed under the Apache License, Version 2.0 (the "License");
-# you may not use this file except in compliance with the License.
-# You may obtain a copy of the License at:
-#
-#     http://www.apache.org/licenses/LICENSE-2.0
-#
-# Unless required by applicable law or agreed to in writing, software
-# distributed under the License is distributed on an "AS IS" BASIS,
-# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-# See the License for the specific language governing permissions and
-# limitations under the License.
-
-import os
-import util
-from string import Template
-
-from util import remove_suffix
-
-callback_suffix = "Callback"
-
-callback_template = Template("""
-package $plugin_package.$callback_package;
-
-/**
- * 
Represents callback for plugin's api file message.
- * 
It was generated by callback_gen.py based on $inputfile preparsed data:
- * 
-$docs
- * 

- */
-public interface $cls_name extends $base_package.$callback_package.$callback_type {
-
-    $callback_method
-
-}
-""")
-
-global_callback_template = Template("""
-package $plugin_package.$callback_package;
-
-/**
- * 
Global aggregated callback interface.
- * 
It was generated by callback_gen.py based on $inputfile
- * 
(python representation of api file generated by vppapigen).
- */
-public interface JVpp${plugin_name}GlobalCallback extends $base_package.$callback_package.ControlPingCallback, $callbacks {
-}
-""")
-
-
-def generate_callbacks(func_list, base_package, plugin_package, plugin_name, callback_package, dto_package, inputfile):
-    """ Generates callback interfaces """
-    print "Generating Callback interfaces"
-
-    if not os.path.exists(callback_package):
-        raise Exception("%s folder is missing" % callback_package)
-
-    callbacks = []
-    for func in func_list:
-
-        camel_case_name_with_suffix = util.underscore_to_camelcase_upper(func['name'])
-
-        if util.is_ignored(func['name']) or util.is_control_ping(camel_case_name_with_suffix):
-            continue
-        if not util.is_reply(camel_case_name_with_suffix) and not util.is_notification(func['name']):
-            continue
-
-        if util.is_reply(camel_case_name_with_suffix):
-            camel_case_name = util.remove_reply_suffix(camel_case_name_with_suffix)
-            callback_type = "JVppCallback"
-        else:
-            camel_case_name_with_suffix = util.add_notification_suffix(camel_case_name_with_suffix)
-            camel_case_name = camel_case_name_with_suffix
-            callback_type = "JVppNotificationCallback"
-
-        callbacks.append("{0}.{1}.{2}".format(plugin_package, callback_package, camel_case_name + callback_suffix))
-        callback_path = os.path.join(callback_package, camel_case_name + callback_suffix + ".java")
-        callback_file = open(callback_path, 'w')
-
-        reply_type = "%s.%s.%s" % (plugin_package, dto_package, camel_case_name_with_suffix)
-        method = "void on{0}({1} reply);".format(camel_case_name_with_suffix, reply_type)
-        callback_file.write(
-            callback_template.substitute(inputfile=inputfile,
-                                         docs=util.api_message_to_javadoc(func),
-                                         cls_name=camel_case_name + callback_suffix,
-                                         callback_method=method,
-                                         base_package=base_package,
-                                         plugin_package=plugin_package,
-                                         callback_package=callback_package,
-                                         callback_type=callback_type))
-        callback_file.flush()
-        callback_file.close()
-
-    callback_file = open(os.path.join(callback_package, "JVpp%sGlobalCallback.java" % plugin_name), 'w')
-    callback_file.write(global_callback_template.substitute(inputfile=inputfile,
-                                                            callbacks=", ".join(callbacks),
-                                                            base_package=base_package,
-                                                            plugin_package=plugin_package,
-                                                            plugin_name=plugin_name,
-                                                            callback_package=callback_package))
-    callback_file.flush()
-    callback_file.close()
diff --git a/vpp-api/java/jvpp/gen/jvppgen/dto_gen.py b/vpp-api/java/jvpp/gen/jvppgen/dto_gen.py
deleted file mode 100644
index a043c945..00000000
--- a/vpp-api/java/jvpp/gen/jvppgen/dto_gen.py
+++ /dev/null
@@ -1,308 +0,0 @@
-#!/usr/bin/env python
-#
-# Copyright (c) 2016 Cisco and/or its affiliates.
-# Licensed under the Apache License, Version 2.0 (the "License");
-# you may not use this file except in compliance with the License.
-# You may obtain a copy of the License at:
-#
-#     http://www.apache.org/licenses/LICENSE-2.0
-#
-# Unless required by applicable law or agreed to in writing, software
-# distributed under the License is distributed on an "AS IS" BASIS,
-# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-# See the License for the specific language governing permissions and
-# limitations under the License.
-
-import os
-from string import Template
-
-import util
-
-dto_template = Template("""
-package $plugin_package.$dto_package;
-
-/**
- * 
This class represents $description.
- * 
It was generated by dto_gen.py based on $inputfile preparsed data:
- * 
-$docs
- * 

- */
-public final class $cls_name implements $base_package.$dto_package.$base_type {
-
-$fields
-$methods
-}
-""")
-
-field_template = Template("""    public $type $name;\n""")
-
-send_template = Template("""    @Override
-    public int send(final $base_package.JVpp jvpp) throws io.fd.vpp.jvpp.VppInvocationException {
-        return (($plugin_package.JVpp${plugin_name})jvpp).$method_name($args);
-    }""")
-
-
-def generate_dtos(func_list, base_package, plugin_package, plugin_name, dto_package, inputfile):
-    """ Generates dto objects in a dedicated package """
-    print "Generating DTOs"
-
-    if not os.path.exists(dto_package):
-        raise Exception("%s folder is missing" % dto_package)
-
-    for func in func_list:
-        camel_case_dto_name = util.underscore_to_camelcase_upper(func['name'])
-        camel_case_method_name = util.underscore_to_camelcase(func['name'])
-        dto_path = os.path.join(dto_package, camel_case_dto_name + ".java")
-
-        if util.is_ignored(func['name']) or util.is_control_ping(camel_case_dto_name):
-            continue
-
-        fields = generate_dto_fields(camel_case_dto_name, func)
-        methods = generate_dto_base_methods(camel_case_dto_name, func)
-        base_type = ""
-
-        # Generate request/reply or dump/dumpReply even if structure can be used as notification
-        if not util.is_just_notification(func["name"]):
-            if util.is_reply(camel_case_dto_name):
-                description = "reply DTO"
-                request_dto_name = get_request_name(camel_case_dto_name, func['name'])
-                if util.is_details(camel_case_dto_name):
-                    # FIXME assumption that dump calls end with "Dump" suffix. Not enforced in vpe.api
-                    base_type += "JVppReply<%s.%s.%s>" % (plugin_package, dto_package, request_dto_name + "Dump")
-                    generate_dump_reply_dto(request_dto_name, base_package, plugin_package, dto_package,
-                                            camel_case_dto_name, camel_case_method_name, func)
-                else:
-                    base_type += "JVppReply<%s.%s.%s>" % (plugin_package, dto_package, request_dto_name)
-            else:
-                args = "" if fields is "" else "this"
-                methods += send_template.substitute(method_name=camel_case_method_name,
-                                                    base_package=base_package,
-                                                    plugin_package=plugin_package,
-                                                    plugin_name=plugin_name,
-                                                    args=args)
-                if util.is_dump(camel_case_dto_name):
-                    base_type += "JVppDump"
-                    description = "dump request DTO"
-                else:
-                    base_type += "JVppRequest"
-                    description = "request DTO"
-
-            write_dto_file(base_package, plugin_package, base_type, camel_case_dto_name, description, dto_package,
-                           dto_path, fields, func, inputfile, methods)
-
-        # for structures that are also used as notifications, generate dedicated notification DTO
-        if util.is_notification(func["name"]):
-            base_type = "JVppNotification"
-            description = "notification DTO"
-            camel_case_dto_name = util.add_notification_suffix(camel_case_dto_name)
-            dto_path = os.path.join(dto_package, camel_case_dto_name + ".java")
-            methods = generate_dto_base_methods(camel_case_dto_name, func)
-            write_dto_file(base_package, plugin_package, base_type, camel_case_dto_name, description, dto_package,
-                           dto_path, fields, func, inputfile, methods)
-
-    flush_dump_reply_dtos(inputfile)
-
-
-def generate_dto_base_methods(camel_case_dto_name, func):
-    methods = generate_dto_hash(func)
-    methods += generate_dto_equals(camel_case_dto_name, func)
-    methods += generate_dto_tostring(camel_case_dto_name, func)
-    return methods
-
-
-def generate_dto_fields(camel_case_dto_name, func):
-    fields = ""
-    for t in zip(func['types'], func['args']):
-        # for retval don't generate dto field in Reply
-        field_name = util.underscore_to_camelcase(t[1])
-        if util.is_reply(camel_case_dto_name) and util.is_retval_field(field_name):
-            continue
-        fields += field_template.substitute(type=util.jni_2_java_type_mapping[t[0]],
-                                            name=field_name)
-    return fields
-
-
-tostring_field_template = Template("""                \"$field_name=\" + $field_name + ", " +\n""")
-tostring_array_field_template = Template("""                \"$field_name=\" + java.util.Arrays.toString($field_name) + ", " +\n""")
-tostring_template = Template("""    @Override
-    public String toString() {
-        return "$cls_name{" +
-$fields_tostring "}";
-    }\n\n""")
-
-
-def generate_dto_tostring(camel_case_dto_name, func):
-    tostring_fields = ""
-    for t in zip(func['types'], func['args']):
-
-        field_name = util.underscore_to_camelcase(t[1])
-        # for retval don't generate dto field in Reply
-        if util.is_retval_field(field_name):
-            continue
-
-        # handle array types
-        if util.is_array(util.jni_2_java_type_mapping[t[0]]):
-            tostring_fields += tostring_array_field_template.substitute(field_name=field_name)
-        else:
-            tostring_fields += tostring_field_template.substitute(field_name=field_name)
-
-    return tostring_template.substitute(cls_name=camel_case_dto_name,
-                                        fields_tostring=tostring_fields[:-8])
-
-equals_other_template = Template("""
-        final $cls_name other = ($cls_name) o;
-\n""")
-equals_field_template = Template("""        if (!java.util.Objects.equals(this.$field_name, other.$field_name)) {
-            return false;
-        }\n""")
-equals_array_field_template = Template("""        if (!java.util.Arrays.equals(this.$field_name, other.$field_name)) {
-            return false;
-        }\n""")
-equals_template = Template("""    @Override
-    public boolean equals(final Object o) {
-        if (this == o) {
-            return true;
-        }
-        if (o == null || getClass() != o.getClass()) {
-            return false;
-        }
-$comparisons
-        return true;
-    }\n\n""")
-
-
-def generate_dto_equals(camel_case_dto_name, func):
-    equals_fields = ""
-    for t in zip(func['types'], func['args']):
-        field_name = util.underscore_to_camelcase(t[1])
-        # for retval don't generate dto field in Reply
-        if util.is_retval_field(field_name):
-            continue
-
-        # handle array types
-        if util.is_array(util.jni_2_java_type_mapping[t[0]]):
-            equals_fields += equals_array_field_template.substitute(field_name=field_name)
-        else:
-            equals_fields += equals_field_template.substitute(field_name=field_name)
-
-    if equals_fields != "":
-        equals_fields = equals_other_template.substitute(cls_name=camel_case_dto_name) + equals_fields
-
-    return equals_template.substitute(comparisons=equals_fields)
-
-
-hash_template = Template("""    @Override
-    public int hashCode() {
-        return java.util.Objects.hash($fields);
-    }\n\n""")
-hash_single_array_type_template = Template("""    @Override
-    public int hashCode() {
-        return java.util.Arrays.hashCode($fields);
-    }\n\n""")
-
-
-def generate_dto_hash(func):
-    hash_fields = ""
-
-    # Special handling for hashCode in case just a single array field is present. Cannot use Objects.equals since the
-    # array is mistaken for a varargs parameter. Instead use Arrays.hashCode in such case.
-    if len(func['args']) == 1:
-        single_type = func['types'][0]
-        single_type_name = func['args'][0]
-        if util.is_array(util.jni_2_java_type_mapping[single_type]):
-            return hash_single_array_type_template.substitute(fields=util.underscore_to_camelcase(single_type_name))
-
-    for t in zip(func['types'], func['args']):
-        field_name = util.underscore_to_camelcase(t[1])
-        # for retval don't generate dto field in Reply
-        if util.is_retval_field(field_name):
-            continue
-
-        hash_fields += field_name + ", "
-
-    return hash_template.substitute(fields=hash_fields[:-2])
-
-
-def write_dto_file(base_package, plugin_package, base_type, camel_case_dto_name, description, dto_package, dto_path,
-                   fields, func, inputfile, methods):
-    dto_file = open(dto_path, 'w')
-    dto_file.write(dto_template.substitute(inputfile=inputfile,
-                                           description=description,
-                                           docs=util.api_message_to_javadoc(func),
-                                           cls_name=camel_case_dto_name,
-                                           fields=fields,
-                                           methods=methods,
-                                           base_package=base_package,
-                                           plugin_package=plugin_package,
-                                           base_type=base_type,
-                                           dto_package=dto_package))
-    dto_file.flush()
-    dto_file.close()
-
-
-dump_dto_suffix = "ReplyDump"
-dump_reply_artificial_dtos = {}
-
-
-# Returns request name or special one from unconventional_naming_rep_req map
-def get_request_name(camel_case_dto_name, func_name):
-    return util.underscore_to_camelcase_upper(
-        util.unconventional_naming_rep_req[func_name]) if func_name in util.unconventional_naming_rep_req \
-        else util.remove_reply_suffix(camel_case_dto_name)
-
-
-def flush_dump_reply_dtos(inputfile):
-    for dump_reply_artificial_dto in dump_reply_artificial_dtos.values():
-        dto_path = os.path.join(dump_reply_artificial_dto['dto_package'],
-                                dump_reply_artificial_dto['cls_name'] + ".java")
-        dto_file = open(dto_path, 'w')
-        dto_file.write(dto_template.substitute(inputfile=inputfile,
-                                               description="dump reply wrapper",
-                                               docs=dump_reply_artificial_dto['docs'],
-                                               cls_name=dump_reply_artificial_dto['cls_name'],
-                                               fields=dump_reply_artificial_dto['fields'],
-                                               methods=dump_reply_artificial_dto['methods'],
-                                               plugin_package=dump_reply_artificial_dto['plugin_package'],
-                                               base_package=dump_reply_artificial_dto['base_package'],
-                                               base_type=dump_reply_artificial_dto['base_type'],
-                                               dto_package=dump_reply_artificial_dto['dto_package']))
-        dto_file.flush()
-        dto_file.close()
-
-
-def generate_dump_reply_dto(request_dto_name, base_package, plugin_package, dto_package, camel_case_dto_name,
-                            camel_case_method_name, func):
-    base_type = "JVppReplyDump<%s.%s.%s, %s.%s.%s>" % (
-        plugin_package, dto_package, util.remove_reply_suffix(camel_case_dto_name) + "Dump",
-        plugin_package, dto_package, camel_case_dto_name)
-    fields = "    public java.util.List<%s> %s = new java.util.ArrayList<>();" % (camel_case_dto_name, camel_case_method_name)
-    cls_name = camel_case_dto_name + dump_dto_suffix
-    # using artificial type for fields, just to bypass the is_array check in base methods generators
-    # the type is not really used
-    artificial_type = 'u8'
-
-    # In case of already existing artificial reply dump DTO, just update it
-    # Used for sub-dump dtos
-    if request_dto_name in dump_reply_artificial_dtos.keys():
-        dump_reply_artificial_dtos[request_dto_name]['fields'] += '\n' + fields
-        dump_reply_artificial_dtos[request_dto_name]['field_names'].append(func['name'])
-        dump_reply_artificial_dtos[request_dto_name]['field_types'].append(artificial_type)
-        methods = '\n' + generate_dto_base_methods(dump_reply_artificial_dtos[request_dto_name]['cls_name'],
-                                            {'args': dump_reply_artificial_dtos[request_dto_name]['field_names'],
-                                             'types': dump_reply_artificial_dtos[request_dto_name]['field_types']})
-        dump_reply_artificial_dtos[request_dto_name]['methods'] = methods
-    else:
-        methods = '\n' + generate_dto_base_methods(cls_name, {'args': [func['name']],
-                                                              'types': [artificial_type]})
-        dump_reply_artificial_dtos[request_dto_name] = ({'docs': util.api_message_to_javadoc(func),
-                                                         'cls_name': cls_name,
-                                                         'fields': fields,
-                                                         'field_names': [func['name']],
-                                                         'field_types': [artificial_type],
-                                                         # strip too many newlines at the end of base method block
-                                                         'methods': methods,
-                                                         'plugin_package': plugin_package,
-                                                         'base_package': base_package,
-                                                         'base_type': base_type,
-                                                         'dto_package': dto_package})
diff --git a/vpp-api/java/jvpp/gen/jvppgen/jni_gen.py b/vpp-api/java/jvpp/gen/jvppgen/jni_gen.py
deleted file mode 100644
index 328cc8d3..00000000
--- a/vpp-api/java/jvpp/gen/jvppgen/jni_gen.py
+++ /dev/null
@@ -1,295 +0,0 @@
-#!/usr/bin/env python
-#
-# Copyright (c) 2016 Cisco and/or its affiliates.
-# Licensed under the Apache License, Version 2.0 (the "License");
-# you may not use this file except in compliance with the License.
-# You may obtain a copy of the License at:
-#
-#     http://www.apache.org/licenses/LICENSE-2.0
-#
-# Unless required by applicable law or agreed to in writing, software
-# distributed under the License is distributed on an "AS IS" BASIS,
-# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-# See the License for the specific language governing permissions and
-# limitations under the License.
-
-from string import Template
-
-import util
-
-variable_length_array_value_template = Template("""mp->${length_var_name}""")
-variable_length_array_template = Template("""clib_net_to_host_${length_field_type}(${value})""")
-
-dto_field_id_template = Template("""
-    jfieldID ${field_reference_name}FieldId = (*env)->GetFieldID(env, ${class_ref_name}Class, "${field_name}", "${jni_signature}");""")
-
-default_dto_field_setter_template = Template("""
-    (*env)->Set${jni_setter}(env, ${object_name}, ${field_reference_name}FieldId, mp->${c_name});
-""")
-
-variable_length_array_value_template = Template("""mp->${length_var_name}""")
-variable_length_array_template = Template("""clib_net_to_host_${length_field_type}(${value})""")
-
-u16_dto_field_setter_template = Template("""
-    (*env)->Set${jni_setter}(env, ${object_name}, ${field_reference_name}FieldId, clib_net_to_host_u16(mp->${c_name}));
-""")
-
-u32_dto_field_setter_template = Template("""
-    (*env)->Set${jni_setter}(env, ${object_name}, ${field_reference_name}FieldId, clib_net_to_host_u32(mp->${c_name}));
-""")
-
-u64_dto_field_setter_template = Template("""
-    (*env)->Set${jni_setter}(env, ${object_name}, ${field_reference_name}FieldId, clib_net_to_host_u64(mp->${c_name}));
-""")
-
-u8_array_dto_field_setter_template = Template("""
-    jbyteArray ${field_reference_name} = (*env)->NewByteArray(env, ${field_length});
-    (*env)->SetByteArrayRegion(env, ${field_reference_name}, 0, ${field_length}, (const jbyte*)mp->${c_name});
-    (*env)->SetObjectField(env, ${object_name}, ${field_reference_name}FieldId, ${field_reference_name});
-    (*env)->DeleteLocalRef(env, ${field_reference_name});
-""")
-
-u16_array_dto_field_setter_template = Template("""
-    {
-        jshortArray ${field_reference_name} = (*env)->NewShortArray(env, ${field_length});
-        jshort * ${field_reference_name}ArrayElements = (*env)->GetShortArrayElements(env, ${field_reference_name}, NULL);
-        unsigned int _i;
-        for (_i = 0; _i < ${field_length}; _i++) {
-            ${field_reference_name}ArrayElements[_i] = clib_net_to_host_u16(mp->${c_name}[_i]);
-        }
-
-        (*env)->ReleaseShortArrayElements(env,  ${field_reference_name}, ${field_reference_name}ArrayElements, 0);
-        (*env)->SetObjectField(env, ${object_name}, ${field_reference_name}FieldId, ${field_reference_name});
-        (*env)->DeleteLocalRef(env, ${field_reference_name});
-    }
-""")
-
-u32_array_dto_field_setter_template = Template("""
-    {
-        jintArray ${field_reference_name} = (*env)->NewIntArray(env, ${field_length});
-        jint * ${field_reference_name}ArrayElements = (*env)->GetIntArrayElements(env, ${field_reference_name}, NULL);
-        unsigned int _i;
-        for (_i = 0; _i < ${field_length}; _i++) {
-            ${field_reference_name}ArrayElements[_i] = clib_net_to_host_u32(mp->${c_name}[_i]);
-        }
-
-        (*env)->ReleaseIntArrayElements(env,  ${field_reference_name}, ${field_reference_name}ArrayElements, 0);
-        (*env)->SetObjectField(env, ${object_name}, ${field_reference_name}FieldId, ${field_reference_name});
-        (*env)->DeleteLocalRef(env, ${field_reference_name});
-    }
-""")
-
-# For each u64 array we get its elements. Then we convert values to host byte order.
-# All changes to  jlong* buffer are written to jlongArray (isCopy is set to NULL)
-u64_array_dto_field_setter_template = Template("""
-    {
-        jlongArray ${field_reference_name} = (*env)->NewLongArray(env, ${field_length});
-        jlong * ${field_reference_name}ArrayElements = (*env)->GetLongArrayElements(env, ${field_reference_name}, NULL);
-        unsigned int _i;
-        for (_i = 0; _i < ${field_length}; _i++) {
-            ${field_reference_name}ArrayElements[_i] = clib_net_to_host_u64(mp->${c_name}[_i]);
-        }
-
-        (*env)->ReleaseLongArrayElements(env,  ${field_reference_name}, ${field_reference_name}ArrayElements, 0);
-        (*env)->SetObjectField(env, ${object_name}, ${field_reference_name}FieldId, ${field_reference_name});
-        (*env)->DeleteLocalRef(env, ${field_reference_name});
-    }
-""")
-
-dto_field_setter_templates = {'u8': default_dto_field_setter_template,
-                              'u16': u16_dto_field_setter_template,
-                              'u32': u32_dto_field_setter_template,
-                              'i32': u32_dto_field_setter_template,
-                              'u64': u64_dto_field_setter_template,
-                              'f64': default_dto_field_setter_template,  # fixme
-                              'u8[]': u8_array_dto_field_setter_template,
-                              'u16[]': u16_array_dto_field_setter_template,
-                              'u32[]': u32_array_dto_field_setter_template,
-                              'u64[]': u64_array_dto_field_setter_template
-                              }
-
-
-def jni_reply_handler_for_type(handler_name, ref_name, field_type, c_name, field_reference_name,
-                               field_name, field_length, is_variable_len_array, length_field_type,
-                               object_name="dto"):
-    """
-    Generates jni code that initializes a field of java object (dto or custom type).
-    To be used in reply message handlers.
-    :param field_type: type of the field to be initialized (as defined in vpe.api)
-    :param c_name: name of the message struct member that stores initialization value
-    :param field_reference_name: name of the field reference in generated code
-    :param field_name: name of the field (camelcase)
-    :param field_length: integer or name of variable that stores field length
-    :param object_name: name of the object to be initialized
-    """
-
-    # todo move validation to vppapigen
-    if field_type.endswith('[]') and field_length == '0':
-        raise Exception('Variable array \'%s\' defined in \'%s\' '
-                        'should have defined length (e.g. \'%s[%s_length]\''
-                        % (c_name, handler_name, c_name, c_name))
-
-    if is_variable_len_array:
-        length_var_name = field_length
-        field_length = variable_length_array_value_template.substitute(length_var_name=length_var_name)
-        if length_field_type != 'u8':  # we need net to host conversion:
-            field_length = variable_length_array_template.substitute(
-                    length_field_type=length_field_type, value=field_length)
-
-    # for retval don't generate setters
-    if util.is_retval_field(c_name):
-        return ""
-
-    jni_signature = util.jni_2_signature_mapping[field_type]
-    jni_setter = util.jni_field_accessors[field_type]
-
-    result = dto_field_id_template.substitute(
-            field_reference_name=field_reference_name,
-            field_name=field_name,
-            class_ref_name=ref_name,
-            jni_signature=jni_signature)
-
-    dto_setter_template = dto_field_setter_templates[field_type]
-
-    result += dto_setter_template.substitute(
-            jni_signature=jni_signature,
-            object_name=object_name,
-            field_reference_name=field_reference_name,
-            c_name=c_name,
-            jni_setter=jni_setter,
-            field_length=field_length)
-    return result
-
-
-request_field_identifier_template = Template("""
-    jfieldID ${field_reference_name}FieldId = (*env)->GetFieldID(env, ${object_name}Class, "${field_name}", "${jni_signature}");
-    ${jni_type} ${field_reference_name} = (*env)->Get${jni_getter}(env, ${object_name}, ${field_reference_name}FieldId);
-    """)
-
-array_length_enforcement_template = Template("""
-        size_t max_size = ${field_length};
-        if (cnt > max_size) cnt = max_size;""")
-
-u8_struct_setter_template = Template("""
-    mp->${c_name} = ${field_reference_name};""")
-
-u16_struct_setter_template = Template("""
-    mp->${c_name} = clib_host_to_net_u16(${field_reference_name});""")
-
-u32_struct_setter_template = Template("""
-    mp->${c_name} = clib_host_to_net_u32(${field_reference_name});""")
-
-i32_struct_setter_template = Template("""
-    mp->${c_name} = clib_host_to_net_i32(${field_reference_name});!""")
-
-u64_struct_setter_template = Template("""
-    mp->${c_name} = clib_host_to_net_u64(${field_reference_name});""")
-
-array_length_enforcement_template = Template("""
-        size_t max_size = ${field_length};
-        if (cnt > max_size) cnt = max_size;""")
-
-u8_array_struct_setter_template = Template("""
-    if (${field_reference_name}) {
-        jsize cnt = (*env)->GetArrayLength (env, ${field_reference_name});
-        ${field_length_check}
-        (*env)->GetByteArrayRegion(env, ${field_reference_name}, 0, cnt, (jbyte *)mp->${c_name});
-    }
-""")
-
-u16_array_struct_setter_template = Template("""
-    if (${field_reference_name}) {
-        jshort * ${field_reference_name}ArrayElements = (*env)->GetShortArrayElements(env, ${field_reference_name}, NULL);
-        size_t _i;
-        jsize cnt = (*env)->GetArrayLength (env, ${field_reference_name});
-        ${field_length_check}
-        for (_i = 0; _i < cnt; _i++) {
-            mp->${c_name}[_i] = clib_host_to_net_u16(${field_reference_name}ArrayElements[_i]);
-        }
-        (*env)->ReleaseShortArrayElements (env, ${field_reference_name}, ${field_reference_name}ArrayElements, 0);
-    }
-    """)
-
-u32_array_struct_setter_template = Template("""
-    if (${field_reference_name}) {
-        jint * ${field_reference_name}ArrayElements = (*env)->GetIntArrayElements(env, ${field_reference_name}, NULL);
-        size_t _i;
-        jsize cnt = (*env)->GetArrayLength (env, ${field_reference_name});
-        ${field_length_check}
-        for (_i = 0; _i < cnt; _i++) {
-            mp->${c_name}[_i] = clib_host_to_net_u32(${field_reference_name}ArrayElements[_i]);
-        }
-        (*env)->ReleaseIntArrayElements (env, ${field_reference_name}, ${field_reference_name}ArrayElements, 0);
-    }
-    """)
-
-u64_array_struct_setter_template = Template("""
-    if (${field_reference_name}) {
-        jlong * ${field_reference_name}ArrayElements = (*env)->GetLongArrayElements(env, ${field_reference_name}, NULL);
-        size_t _i;
-        jsize cnt = (*env)->GetArrayLength (env, ${field_reference_name});
-        ${field_length_check}
-        for (_i = 0; _i < cnt; _i++) {
-            mp->${c_name}[_i] = clib_host_to_net_u64(${field_reference_name}ArrayElements[_i]);
-        }
-        (*env)->ReleaseLongArrayElements (env, ${field_reference_name}, ${field_reference_name}ArrayElements, 0);
-    }
-    """)
-
-struct_setter_templates = {'u8': u8_struct_setter_template,
-                           'u16': u16_struct_setter_template,
-                           'u32': u32_struct_setter_template,
-                           'i32': u32_struct_setter_template,
-                           'u64': u64_struct_setter_template,
-                           'u8[]': u8_array_struct_setter_template,
-                           'u16[]': u16_array_struct_setter_template,
-                           'u32[]': u32_array_struct_setter_template,
-                           'u64[]': u64_array_struct_setter_template
-                           }
-
-
-def jni_request_binding_for_type(field_type, c_name, field_reference_name, field_name, field_length,
-                                 is_variable_len_array, object_name="request"):
-    """
-    Generates jni code that initializes C structure that corresponds to a field of java object
-    (dto or custom type). To be used in request message handlers.
-    :param field_type: type of the field to be initialized (as defined in vpe.api)
-    :param c_name: name of the message struct member to be initialized
-    :param field_reference_name: name of the field reference in generated code
-    :param field_name: name of the field (camelcase)
-    :param field_length: integer or name of variable that stores field length
-    :param object_name: name of the object to be initialized
-    """
-    # field identifiers
-    jni_type = util.vpp_2_jni_type_mapping[field_type]
-    jni_signature = util.jni_2_signature_mapping[field_type]
-    jni_getter = util.jni_field_accessors[field_type]
-
-    # field identifier
-    msg_initialization = request_field_identifier_template.substitute(
-            jni_type=jni_type,
-            field_reference_name=field_reference_name,
-            field_name=field_name,
-            jni_signature=jni_signature,
-            jni_getter=jni_getter,
-            object_name=object_name)
-
-    # field setter
-    field_length_check = ""
-
-    # check if we are processing variable length array:
-    if is_variable_len_array:
-        field_length = util.underscore_to_camelcase(field_length)
-
-    # enforce max length if array has fixed length or uses variable length syntax
-    if str(field_length) != "0":
-        field_length_check = array_length_enforcement_template.substitute(field_length=field_length)
-
-    struct_setter_template = struct_setter_templates[field_type]
-
-    msg_initialization += struct_setter_template.substitute(
-            c_name=c_name,
-            field_reference_name=field_reference_name,
-            field_length_check=field_length_check)
-
-    return msg_initialization
diff --git a/vpp-api/java/jvpp/gen/jvppgen/jvpp_c_gen.py b/vpp-api/java/jvpp/gen/jvppgen/jvpp_c_gen.py
deleted file mode 100644
index 5fc84c7b..00000000
--- a/vpp-api/java/jvpp/gen/jvppgen/jvpp_c_gen.py
+++ /dev/null
@@ -1,343 +0,0 @@
-#!/usr/bin/env python
-#
-# Copyright (c) 2016 Cisco and/or its affiliates.
-# Licensed under the Apache License, Version 2.0 (the "License");
-# you may not use this file except in compliance with the License.
-# You may obtain a copy of the License at:
-#
-#     http://www.apache.org/licenses/LICENSE-2.0
-# l
-# Unless required by applicable law or agreed to in writing, software
-# distributed under the License is distributed on an "AS IS" BASIS,
-# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-# See the License for the specific language governing permissions and
-# limitations under the License.
-#
-
-import os, util
-from string import Template
-
-import jni_gen
-
-
-def is_manually_generated(f_name, plugin_name):
-    return f_name in {'control_ping_reply'}
-
-
-class_reference_template = Template("""jclass ${ref_name}Class;
-""")
-
-find_class_invocation_template = Template("""
-    ${ref_name}Class = (jclass)(*env)->NewGlobalRef(env, (*env)->FindClass(env, "io/fd/vpp/jvpp/${plugin_name}/dto/${class_name}"));
-    if ((*env)->ExceptionCheck(env)) {
-        (*env)->ExceptionDescribe(env);
-        return JNI_ERR;
-    }""")
-
-find_class_template = Template("""
-    ${ref_name}Class = (jclass)(*env)->NewGlobalRef(env, (*env)->FindClass(env, "${class_name}"));
-    if ((*env)->ExceptionCheck(env)) {
-        (*env)->ExceptionDescribe(env);
-        return JNI_ERR;
-    }""")
-
-delete_class_invocation_template = Template("""
-    if (${ref_name}Class) {
-        (*env)->DeleteGlobalRef(env, ${ref_name}Class);
-    }""")
-
-class_cache_template = Template("""
-$class_references
-static int cache_class_references(JNIEnv* env) {
-    $find_class_invocations
-    return 0;
-}
-
-static void delete_class_references(JNIEnv* env) {
-    $delete_class_invocations
-}""")
-
-
-def generate_class_cache(func_list, plugin_name):
-    class_references = []
-    find_class_invocations = []
-    delete_class_invocations = []
-    for f in func_list:
-        c_name = f['name']
-        class_name = util.underscore_to_camelcase_upper(c_name)
-        ref_name = util.underscore_to_camelcase(c_name)
-
-        if util.is_ignored(c_name) or util.is_control_ping(class_name):
-            continue
-
-        if util.is_reply(class_name):
-            class_references.append(class_reference_template.substitute(
-                ref_name=ref_name))
-            find_class_invocations.append(find_class_invocation_template.substitute(
-                plugin_name=plugin_name,
-                ref_name=ref_name,
-                class_name=class_name))
-            delete_class_invocations.append(delete_class_invocation_template.substitute(ref_name=ref_name))
-        elif util.is_notification(c_name):
-            class_references.append(class_reference_template.substitute(
-                ref_name=util.add_notification_suffix(ref_name)))
-            find_class_invocations.append(find_class_invocation_template.substitute(
-                plugin_name=plugin_name,
-                ref_name=util.add_notification_suffix(ref_name),
-                class_name=util.add_notification_suffix(class_name)))
-            delete_class_invocations.append(delete_class_invocation_template.substitute(
-                ref_name=util.add_notification_suffix(ref_name)))
-
-    # add exception class to class cache
-    ref_name = 'callbackException'
-    class_name = 'io/fd/vpp/jvpp/VppCallbackException'
-    class_references.append(class_reference_template.substitute(
-            ref_name=ref_name))
-    find_class_invocations.append(find_class_template.substitute(
-            ref_name=ref_name,
-            class_name=class_name))
-    delete_class_invocations.append(delete_class_invocation_template.substitute(ref_name=ref_name))
-
-    return class_cache_template.substitute(
-        class_references="".join(class_references), find_class_invocations="".join(find_class_invocations),
-        delete_class_invocations="".join(delete_class_invocations))
-
-
-# TODO: cache method and field identifiers to achieve better performance
-# https://jira.fd.io/browse/HONEYCOMB-42
-request_class_template = Template("""
-    jclass requestClass = (*env)->FindClass(env, "io/fd/vpp/jvpp/${plugin_name}/dto/${java_name_upper}");""")
-
-request_field_identifier_template = Template("""
-    jfieldID ${field_reference_name}FieldId = (*env)->GetFieldID(env, ${object_name}Class, "${field_name}", "${jni_signature}");
-    ${jni_type} ${field_reference_name} = (*env)->Get${jni_getter}(env, ${object_name}, ${field_reference_name}FieldId);
-    """)
-
-
-jni_impl_template = Template("""
-/**
- * JNI binding for sending ${c_name} message.
- * Generated based on $inputfile preparsed data:
-$api_data
- */
-JNIEXPORT jint JNICALL Java_io_fd_vpp_jvpp_${plugin_name}_JVpp${java_plugin_name}Impl_${field_name}0
-(JNIEnv * env, jclass clazz$args) {
-    ${plugin_name}_main_t *plugin_main = &${plugin_name}_main;
-    vl_api_${c_name}_t * mp;
-    u32 my_context_id = vppjni_get_context_id (&jvpp_main);
-    $request_class
-
-    // create message:
-    mp = vl_msg_api_alloc(sizeof(*mp));
-    memset (mp, 0, sizeof (*mp));
-    mp->_vl_msg_id = ntohs (VL_API_${c_name_uppercase} + plugin_main->msg_id_base);
-    mp->client_index = plugin_main->my_client_index;
-    mp->context = clib_host_to_net_u32 (my_context_id);
-
-    $msg_initialization
-
-    // send message:
-    vl_msg_api_send_shmem (plugin_main->vl_input_queue, (u8 *)&mp);
-    if ((*env)->ExceptionCheck(env)) {
-        return JNI_ERR;
-    }
-    return my_context_id;
-}""")
-
-def generate_jni_impl(func_list, plugin_name, inputfile):
-    jni_impl = []
-    for f in func_list:
-        f_name = f['name']
-        camel_case_function_name = util.underscore_to_camelcase(f_name)
-        if is_manually_generated(f_name, plugin_name) or util.is_reply(camel_case_function_name) \
-                or util.is_ignored(f_name) or util.is_just_notification(f_name):
-            continue
-
-        arguments = ''
-        request_class = ''
-        msg_initialization = ''
-        f_name_uppercase = f_name.upper()
-
-        if f['args']:
-            arguments = ', jobject request'
-            camel_case_function_name_upper = util.underscore_to_camelcase_upper(f_name)
-
-            request_class = request_class_template.substitute(
-                    java_name_upper=camel_case_function_name_upper,
-                    plugin_name=plugin_name)
-
-            for t in zip(f['types'], f['args'], f['lengths']):
-                field_name = util.underscore_to_camelcase(t[1])
-                msg_initialization += jni_gen.jni_request_binding_for_type(field_type=t[0], c_name=t[1],
-                                                                           field_reference_name=field_name,
-                                                                           field_name=field_name,
-                                                                           field_length=t[2][0],
-                                                                           is_variable_len_array=t[2][1])
-
-        jni_impl.append(jni_impl_template.substitute(
-                inputfile=inputfile,
-                api_data=util.api_message_to_javadoc(f),
-                field_reference_name=camel_case_function_name,
-                field_name=camel_case_function_name,
-                c_name_uppercase=f_name_uppercase,
-                c_name=f_name,
-                plugin_name=plugin_name,
-                java_plugin_name=plugin_name.title(),
-                request_class=request_class,
-                msg_initialization=msg_initialization,
-                args=arguments))
-
-    return "\n".join(jni_impl)
-
-# code fragment for checking result of the operation before sending request reply
-callback_err_handler_template = Template("""
-    // for negative result don't send callback message but send error callback
-    if (mp->retval<0) {
-        call_on_error("${handler_name}", mp->context, mp->retval, plugin_main->callbackClass, plugin_main->callbackObject, callbackExceptionClass);
-        return;
-    }
-    if (mp->retval == VNET_API_ERROR_IN_PROGRESS) {
-        clib_warning("Result in progress");
-        return;
-    }
-""")
-
-msg_handler_template = Template("""
-/**
- * Handler for ${handler_name} message.
- * Generated based on $inputfile preparsed data:
-$api_data
- */
-static void vl_api_${handler_name}_t_handler (vl_api_${handler_name}_t * mp)
-{
-    ${plugin_name}_main_t *plugin_main = &${plugin_name}_main;
-    JNIEnv *env = jvpp_main.jenv;
-
-    $err_handler
-
-    jmethodID constructor = (*env)->GetMethodID(env, ${class_ref_name}Class, "", "()V");
-    jmethodID callbackMethod = (*env)->GetMethodID(env, plugin_main->callbackClass, "on${dto_name}", "(Lio/fd/vpp/jvpp/${plugin_name}/dto/${dto_name};)V");
-
-    jobject dto = (*env)->NewObject(env, ${class_ref_name}Class, constructor);
-    $dto_setters
-
-    (*env)->CallVoidMethod(env, plugin_main->callbackObject, callbackMethod, dto);
-    // free DTO as per http://stackoverflow.com/questions/1340938/memory-leak-when-calling-java-code-from-c-using-jni
-    (*env)->DeleteLocalRef(env, dto);
-}""")
-
-
-def generate_msg_handlers(func_list, plugin_name, inputfile):
-    handlers = []
-    for f in func_list:
-        handler_name = f['name']
-        dto_name = util.underscore_to_camelcase_upper(handler_name)
-        ref_name = util.underscore_to_camelcase(handler_name)
-
-        if is_manually_generated(handler_name, plugin_name) or util.is_ignored(handler_name):
-            continue
-
-        if not util.is_reply(dto_name) and not util.is_notification(handler_name):
-            continue
-
-        if util.is_notification(handler_name):
-            dto_name = util.add_notification_suffix(dto_name)
-            ref_name = util.add_notification_suffix(ref_name)
-
-        dto_setters = ''
-        err_handler = ''
-        # dto setters
-        for t in zip(f['types'], f['args'], f['lengths']):
-            c_name = t[1]
-            java_name = util.underscore_to_camelcase(c_name)
-            field_length = t[2][0]
-            is_variable_len_array = t[2][1]
-            length_field_type = None
-            if is_variable_len_array:
-                length_field_type = f['types'][f['args'].index(field_length)]
-            dto_setters += jni_gen.jni_reply_handler_for_type(handler_name=handler_name, ref_name=ref_name,
-                                                              field_type=t[0], c_name=t[1],
-                                                              field_reference_name=java_name,
-                                                              field_name=java_name, field_length=field_length,
-                                                              is_variable_len_array=is_variable_len_array,
-                                                              length_field_type=length_field_type)
-
-            # for retval don't generate setters and generate retval check
-            if util.is_retval_field(c_name):
-                err_handler = callback_err_handler_template.substitute(
-                    handler_name=handler_name
-                )
-                continue
-
-        handlers.append(msg_handler_template.substitute(
-            inputfile=inputfile,
-            api_data=util.api_message_to_javadoc(f),
-            handler_name=handler_name,
-            plugin_name=plugin_name,
-            dto_name=dto_name,
-            class_ref_name=ref_name,
-            dto_setters=dto_setters,
-            err_handler=err_handler))
-
-    return "\n".join(handlers)
-
-
-handler_registration_template = Template("""_(${upercase_name}, ${name}) \\
-""")
-
-
-def generate_handler_registration(func_list):
-    handler_registration = ["#define foreach_api_reply_handler \\\n"]
-    for f in func_list:
-        name = f['name']
-        camelcase_name = util.underscore_to_camelcase(f['name'])
-
-        if (not util.is_reply(camelcase_name) and not util.is_notification(name)) or util.is_ignored(name) \
-                or util.is_control_ping(camelcase_name):
-            continue
-
-        handler_registration.append(handler_registration_template.substitute(
-            name=name,
-            upercase_name=name.upper()))
-
-    return "".join(handler_registration)
-
-
-jvpp_c_template = Template("""/**
- * This file contains JNI bindings for jvpp Java API.
- * It was generated by jvpp_c_gen.py based on $inputfile
- * (python representation of api file generated by vppapigen).
- */
-
-// JAVA class reference cache
-$class_cache
-
-// JNI bindings
-$jni_implementations
-
-// Message handlers
-$msg_handlers
-
-// Registration of message handlers in vlib
-$handler_registration
-""")
-
-def generate_jvpp(func_list, plugin_name, inputfile):
-    """ Generates jvpp C file """
-    print "Generating jvpp C"
-
-    class_cache = generate_class_cache(func_list, plugin_name)
-    jni_impl = generate_jni_impl(func_list, plugin_name, inputfile)
-    msg_handlers = generate_msg_handlers(func_list, plugin_name, inputfile)
-    handler_registration = generate_handler_registration(func_list)
-
-    jvpp_c_file = open("jvpp_%s_gen.h" % plugin_name, 'w')
-    jvpp_c_file.write(jvpp_c_template.substitute(
-            inputfile=inputfile,
-            class_cache=class_cache,
-            jni_implementations=jni_impl,
-            msg_handlers=msg_handlers,
-            handler_registration=handler_registration))
-    jvpp_c_file.flush()
-    jvpp_c_file.close()
-
diff --git a/vpp-api/java/jvpp/gen/jvppgen/jvpp_callback_facade_gen.py b/vpp-api/java/jvpp/gen/jvppgen/jvpp_callback_facade_gen.py
deleted file mode 100644
index ac096a71..00000000
--- a/vpp-api/java/jvpp/gen/jvppgen/jvpp_callback_facade_gen.py
+++ /dev/null
@@ -1,324 +0,0 @@
-#!/usr/bin/env python
-#
-# Copyright (c) 2016 Cisco and/or its affiliates.
-# Licensed under the Apache License, Version 2.0 (the "License");
-# you may not use this file except in compliance with the License.
-# You may obtain a copy of the License at:
-#
-#     http://www.apache.org/licenses/LICENSE-2.0
-#
-# Unless required by applicable law or agreed to in writing, software
-# distributed under the License is distributed on an "AS IS" BASIS,
-# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-# See the License for the specific language governing permissions and
-# limitations under the License.
-
-import os, util
-from string import Template
-
-import callback_gen
-import dto_gen
-
-jvpp_ifc_template = Template("""
-package $plugin_package.$callback_facade_package;
-
-/**
- * 
Callback Java API representation of $plugin_package plugin.
- * 
It was generated by jvpp_callback_facade_gen.py based on $inputfile
- * 
(python representation of api file generated by vppapigen).
- */
-public interface CallbackJVpp${plugin_name} extends $base_package.$notification_package.NotificationRegistryProvider, java.lang.AutoCloseable {
-
-    // TODO add send
-
-$methods
-}
-""")
-
-jvpp_impl_template = Template("""
-package $plugin_package.$callback_facade_package;
-
-/**
- * 
Default implementation of Callback${plugin_name}JVpp interface.
- * 
It was generated by jvpp_callback_facade_gen.py based on $inputfile
- * 
(python representation of api file generated by vppapigen).
- */
-public final class CallbackJVpp${plugin_name}Facade implements CallbackJVpp${plugin_name} {
-
-    private final $plugin_package.JVpp${plugin_name} jvpp;
-    private final java.util.Map callbacks;
-    private final $plugin_package.$notification_package.${plugin_name}NotificationRegistryImpl notificationRegistry = new $plugin_package.$notification_package.${plugin_name}NotificationRegistryImpl();
-    /**
-     * 
Create CallbackJVpp${plugin_name}Facade object for provided JVpp instance.
-     * Constructor internally creates CallbackJVppFacadeCallback class for processing callbacks
-     * and then connects to provided JVpp instance
-     *
-     * @param jvpp provided $base_package.JVpp instance
-     *
-     * @throws java.io.IOException in case instance cannot connect to JVPP
-     */
-    public CallbackJVpp${plugin_name}Facade(final $base_package.JVppRegistry registry, final $plugin_package.JVpp${plugin_name} jvpp) throws java.io.IOException {
-        this.jvpp = java.util.Objects.requireNonNull(jvpp,"jvpp is null");
-        this.callbacks = new java.util.HashMap<>();
-        java.util.Objects.requireNonNull(registry, "JVppRegistry should not be null");
-        registry.register(jvpp, new CallbackJVpp${plugin_name}FacadeCallback(this.callbacks, notificationRegistry));
-    }
-
-    @Override
-    public $plugin_package.$notification_package.${plugin_name}NotificationRegistry getNotificationRegistry() {
-        return notificationRegistry;
-    }
-
-    @Override
-    public void close() throws Exception {
-        jvpp.close();
-    }
-
-    // TODO add send()
-
-$methods
-}
-""")
-
-method_template = Template(
-    """    void $name($plugin_package.$dto_package.$request request, $plugin_package.$callback_package.$callback callback) throws $base_package.VppInvocationException;""")
-
-method_impl_template = Template("""    public final void $name($plugin_package.$dto_package.$request request, $plugin_package.$callback_package.$callback callback) throws $base_package.VppInvocationException {
-        synchronized (callbacks) {
-            callbacks.put(jvpp.$name(request), callback);
-        }
-    }
-""")
-
-no_arg_method_template = Template("""    void $name($plugin_package.$callback_package.$callback callback) throws $base_package.VppInvocationException;""")
-no_arg_method_impl_template = Template("""    public final void $name($plugin_package.$callback_package.$callback callback) throws $base_package.VppInvocationException {
-        synchronized (callbacks) {
-            callbacks.put(jvpp.$name(), callback);
-        }
-    }
-""")
-
-
-def generate_jvpp(func_list, base_package, plugin_package, plugin_name, dto_package, callback_package, notification_package, callback_facade_package, inputfile):
-    """ Generates callback facade """
-    print "Generating JVpp callback facade"
-
-    if os.path.exists(callback_facade_package):
-        util.remove_folder(callback_facade_package)
-
-    os.mkdir(callback_facade_package)
-
-    methods = []
-    methods_impl = []
-    for func in func_list:
-
-        if util.is_notification(func['name']) or util.is_ignored(func['name']):
-            continue
-
-        camel_case_name = util.underscore_to_camelcase(func['name'])
-        camel_case_name_upper = util.underscore_to_camelcase_upper(func['name'])
-        if util.is_reply(camel_case_name) or util.is_control_ping(camel_case_name):
-            continue
-
-        # Strip suffix for dump calls
-        callback_type = get_request_name(camel_case_name_upper, func['name']) + callback_gen.callback_suffix
-
-        if len(func['args']) == 0:
-            methods.append(no_arg_method_template.substitute(name=camel_case_name,
-                                                             base_package=base_package,
-                                                             plugin_package=plugin_package,
-                                                             dto_package=dto_package,
-                                                             callback_package=callback_package,
-                                                             callback=callback_type))
-            methods_impl.append(no_arg_method_impl_template.substitute(name=camel_case_name,
-                                                                       base_package=base_package,
-                                                                       plugin_package=plugin_package,
-                                                                       dto_package=dto_package,
-                                                                       callback_package=callback_package,
-                                                                       callback=callback_type))
-        else:
-            methods.append(method_template.substitute(name=camel_case_name,
-                                                      request=camel_case_name_upper,
-                                                      base_package=base_package,
-                                                      plugin_package=plugin_package,
-                                                      dto_package=dto_package,
-                                                      callback_package=callback_package,
-                                                      callback=callback_type))
-            methods_impl.append(method_impl_template.substitute(name=camel_case_name,
-                                                                request=camel_case_name_upper,
-                                                                base_package=base_package,
-                                                                plugin_package=plugin_package,
-                                                                dto_package=dto_package,
-                                                                callback_package=callback_package,
-                                                                callback=callback_type))
-
-    join = os.path.join(callback_facade_package, "CallbackJVpp%s.java" % plugin_name)
-    jvpp_file = open(join, 'w')
-    jvpp_file.write(
-        jvpp_ifc_template.substitute(inputfile=inputfile,
-                                     methods="\n".join(methods),
-                                     base_package=base_package,
-                                     plugin_package=plugin_package,
-                                     plugin_name=plugin_name,
-                                     dto_package=dto_package,
-                                     notification_package=notification_package,
-                                     callback_facade_package=callback_facade_package))
-    jvpp_file.flush()
-    jvpp_file.close()
-
-    jvpp_file = open(os.path.join(callback_facade_package, "CallbackJVpp%sFacade.java" % plugin_name), 'w')
-    jvpp_file.write(jvpp_impl_template.substitute(inputfile=inputfile,
-                                                  methods="\n".join(methods_impl),
-                                                  base_package=base_package,
-                                                  plugin_package=plugin_package,
-                                                  plugin_name=plugin_name,
-                                                  dto_package=dto_package,
-                                                  notification_package=notification_package,
-                                                  callback_package=callback_package,
-                                                  callback_facade_package=callback_facade_package))
-    jvpp_file.flush()
-    jvpp_file.close()
-
-    generate_callback(func_list, base_package, plugin_package, plugin_name, dto_package, callback_package, notification_package, callback_facade_package, inputfile)
-
-
-jvpp_facade_callback_template = Template("""
-package $plugin_package.$callback_facade_package;
-
-/**
- * 
Implementation of JVppGlobalCallback interface for Java Callback API.
- * 
It was generated by jvpp_callback_facade_gen.py based on $inputfile
- * 
(python representation of api file generated by vppapigen).
- */
-public final class CallbackJVpp${plugin_name}FacadeCallback implements $plugin_package.$callback_package.JVpp${plugin_name}GlobalCallback {
-
-    private final java.util.Map requests;
-    private final $plugin_package.$notification_package.Global${plugin_name}NotificationCallback notificationCallback;
-    private static final java.util.logging.Logger LOG = java.util.logging.Logger.getLogger(CallbackJVpp${plugin_name}FacadeCallback.class.getName());
-
-    public CallbackJVpp${plugin_name}FacadeCallback(final java.util.Map requestMap,
-                                      final $plugin_package.$notification_package.Global${plugin_name}NotificationCallback notificationCallback) {
-        this.requests = requestMap;
-        this.notificationCallback = notificationCallback;
-    }
-
-    @Override
-    public void onError($base_package.VppCallbackException reply) {
-
-        $base_package.$callback_package.JVppCallback failedCall;
-        synchronized(requests) {
-            failedCall = requests.remove(reply.getCtxId());
-        }
-
-        if(failedCall != null) {
-            try {
-                failedCall.onError(reply);
-            } catch(RuntimeException ex) {
-                ex.addSuppressed(reply);
-                LOG.log(java.util.logging.Level.WARNING, String.format("Callback: %s failed while handling exception: %s", failedCall, reply), ex);
-            }
-        }
-    }
-
-    @Override
-    @SuppressWarnings("unchecked")
-    public void onControlPingReply($base_package.$dto_package.ControlPingReply reply) {
-
-        $base_package.$callback_package.ControlPingCallback callback;
-        synchronized(requests) {
-            callback = ($base_package.$callback_package.ControlPingCallback) requests.remove(reply.context);
-        }
-
-        if(callback != null) {
-            callback.onControlPingReply(reply);
-        }
-    }
-
-$methods
-}
-""")
-
-jvpp_facade_callback_method_template = Template("""
-    @Override
-    @SuppressWarnings("unchecked")
-    public void on$callback_dto($plugin_package.$dto_package.$callback_dto reply) {
-
-        $plugin_package.$callback_package.$callback callback;
-        synchronized(requests) {
-            callback = ($plugin_package.$callback_package.$callback) requests.remove(reply.context);
-        }
-
-        if(callback != null) {
-            callback.on$callback_dto(reply);
-        }
-    }
-""")
-
-jvpp_facade_callback_notification_method_template = Template("""
-    @Override
-    @SuppressWarnings("unchecked")
-    public void on$callback_dto($plugin_package.$dto_package.$callback_dto notification) {
-        notificationCallback.on$callback_dto(notification);
-    }
-""")
-
-
-def generate_callback(func_list, base_package, plugin_package, plugin_name, dto_package, callback_package, notification_package, callback_facade_package, inputfile):
-    callbacks = []
-    for func in func_list:
-
-        camel_case_name_with_suffix = util.underscore_to_camelcase_upper(func['name'])
-
-        if util.is_ignored(func['name']) or util.is_control_ping(camel_case_name_with_suffix):
-            continue
-
-        if util.is_reply(camel_case_name_with_suffix):
-            callbacks.append(jvpp_facade_callback_method_template.substitute(plugin_package=plugin_package,
-                                                                             dto_package=dto_package,
-                                                                             callback_package=callback_package,
-                                                                             callback=util.remove_reply_suffix(camel_case_name_with_suffix) + callback_gen.callback_suffix,
-                                                                             callback_dto=camel_case_name_with_suffix))
-
-        if util.is_notification(func["name"]):
-            with_notification_suffix = util.add_notification_suffix(camel_case_name_with_suffix)
-            callbacks.append(jvpp_facade_callback_notification_method_template.substitute(plugin_package=plugin_package,
-                                                                             dto_package=dto_package,
-                                                                             callback_package=callback_package,
-                                                                             callback=with_notification_suffix + callback_gen.callback_suffix,
-                                                                             callback_dto=with_notification_suffix))
-
-    jvpp_file = open(os.path.join(callback_facade_package, "CallbackJVpp%sFacadeCallback.java" % plugin_name), 'w')
-    jvpp_file.write(jvpp_facade_callback_template.substitute(inputfile=inputfile,
-                                                             base_package=base_package,
-                                                             plugin_package=plugin_package,
-                                                             plugin_name=plugin_name,
-                                                             dto_package=dto_package,
-                                                             notification_package=notification_package,
-                                                             callback_package=callback_package,
-                                                             methods="".join(callbacks),
-                                                             callback_facade_package=callback_facade_package))
-    jvpp_file.flush()
-    jvpp_file.close()
-
-
-# Returns request name or special one from unconventional_naming_rep_req map
-def get_request_name(camel_case_dto_name, func_name):
-    if func_name in reverse_dict(util.unconventional_naming_rep_req):
-        request_name = util.underscore_to_camelcase_upper(reverse_dict(util.unconventional_naming_rep_req)[func_name])
-    else:
-        request_name = camel_case_dto_name
-    return remove_suffix(request_name)
-
-
-def reverse_dict(map):
-    return dict((v, k) for k, v in map.iteritems())
-
-
-def remove_suffix(name):
-    if util.is_reply(name):
-        return util.remove_reply_suffix(name)
-    else:
-        if util.is_dump(name):
-            return util.remove_suffix(name, util.dump_suffix)
-        else:
-            return name
diff --git a/vpp-api/java/jvpp/gen/jvppgen/jvpp_future_facade_gen.py b/vpp-api/java/jvpp/gen/jvppgen/jvpp_future_facade_gen.py
deleted file mode 100644
index ebb840f7..00000000
--- a/vpp-api/java/jvpp/gen/jvppgen/jvpp_future_facade_gen.py
+++ /dev/null
@@ -1,331 +0,0 @@
-#!/usr/bin/env python
-#
-# Copyright (c) 2016 Cisco and/or its affiliates.
-# Licensed under the Apache License, Version 2.0 (the "License");
-# you may not use this file except in compliance with the License.
-# You may obtain a copy of the License at:
-#
-#     http://www.apache.org/licenses/LICENSE-2.0
-#
-# Unless required by applicable law or agreed to in writing, software
-# distributed under the License is distributed on an "AS IS" BASIS,
-# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-# See the License for the specific language governing permissions and
-# limitations under the License.
-
-import os
-from string import Template
-
-import dto_gen
-import util
-
-jvpp_facade_callback_template = Template("""
-package $plugin_package.$future_package;
-
-/**
- * 
Async facade callback setting values to future objects
- * 
It was generated by jvpp_future_facade_gen.py based on $inputfile
- * 
(python representation of api file generated by vppapigen).
- */
-public final class FutureJVpp${plugin_name}FacadeCallback implements $plugin_package.$callback_package.JVpp${plugin_name}GlobalCallback {
-
-    private final java.util.Map>> requests;
-    private final $plugin_package.$notification_package.Global${plugin_name}NotificationCallback notificationCallback;
-
-    public FutureJVpp${plugin_name}FacadeCallback(
-        final java.util.Map>> requestMap,
-        final $plugin_package.$notification_package.Global${plugin_name}NotificationCallback notificationCallback) {
-        this.requests = requestMap;
-        this.notificationCallback = notificationCallback;
-    }
-
-    @Override
-    @SuppressWarnings("unchecked")
-    public void onError($base_package.VppCallbackException reply) {
-        final java.util.concurrent.CompletableFuture<$base_package.$dto_package.JVppReply> completableFuture;
-
-        synchronized(requests) {
-            completableFuture = (java.util.concurrent.CompletableFuture<$base_package.$dto_package.JVppReply>) requests.get(reply.getCtxId());
-        }
-
-        if(completableFuture != null) {
-            completableFuture.completeExceptionally(reply);
-
-            synchronized(requests) {
-                requests.remove(reply.getCtxId());
-            }
-        }
-    }
-
-    @Override
-    @SuppressWarnings("unchecked")
-    public void onControlPingReply($base_package.$dto_package.ControlPingReply reply) {
-        final java.util.concurrent.CompletableFuture<$base_package.$dto_package.JVppReply> completableFuture;
-
-        synchronized(requests) {
-            completableFuture = (java.util.concurrent.CompletableFuture<$base_package.$dto_package.JVppReply>) requests.get(reply.context);
-        }
-
-        if(completableFuture != null) {
-            // Finish dump call
-            if (completableFuture instanceof $base_package.$future_package.AbstractFutureJVppInvoker.CompletableDumpFuture) {
-                completableFuture.complete((($base_package.$future_package.AbstractFutureJVppInvoker.CompletableDumpFuture) completableFuture).getReplyDump());
-                // Remove future mapped to dump call context id
-                synchronized(requests) {
-                    requests.remove((($base_package.$future_package.AbstractFutureJVppInvoker.CompletableDumpFuture) completableFuture).getContextId());
-                }
-            } else {
-                completableFuture.complete(reply);
-            }
-
-            synchronized(requests) {
-                requests.remove(reply.context);
-            }
-        }
-    }
-
-$methods
-}
-""")
-
-jvpp_facade_callback_method_template = Template("""
-    @Override
-    @SuppressWarnings("unchecked")
-    public void on$callback_dto($plugin_package.$dto_package.$callback_dto reply) {
-        final java.util.concurrent.CompletableFuture<$base_package.$dto_package.JVppReply> completableFuture;
-
-        synchronized(requests) {
-            completableFuture = (java.util.concurrent.CompletableFuture<$base_package.$dto_package.JVppReply>) requests.get(reply.context);
-        }
-
-        if(completableFuture != null) {
-            completableFuture.complete(reply);
-
-            synchronized(requests) {
-                requests.remove(reply.context);
-            }
-        }
-    }
-""")
-
-jvpp_facade_callback_notification_method_template = Template("""
-    @Override
-    public void on$callback_dto($plugin_package.$dto_package.$callback_dto notification) {
-        notificationCallback.on$callback_dto(notification);
-    }
-""")
-
-jvpp_facade_details_callback_method_template = Template("""
-    @Override
-    @SuppressWarnings("unchecked")
-    public void on$callback_dto($plugin_package.$dto_package.$callback_dto reply) {
-        final $base_package.$future_package.AbstractFutureJVppInvoker.CompletableDumpFuture<$plugin_package.$dto_package.$callback_dto_reply_dump> completableFuture;
-
-        synchronized(requests) {
-            completableFuture = ($base_package.$future_package.AbstractFutureJVppInvoker.CompletableDumpFuture<$plugin_package.$dto_package.$callback_dto_reply_dump>) requests.get(reply.context);
-        }
-
-        if(completableFuture != null) {
-            completableFuture.getReplyDump().$callback_dto_field.add(reply);
-        }
-    }
-""")
-
-
-def generate_jvpp(func_list, base_package, plugin_package, plugin_name, dto_package, callback_package, notification_package, future_facade_package, inputfile):
-    """ Generates JVpp interface and JNI implementation """
-    print "Generating JVpp future facade"
-
-    if not os.path.exists(future_facade_package):
-        raise Exception("%s folder is missing" % future_facade_package)
-
-    methods = []
-    methods_impl = []
-    callbacks = []
-    for func in func_list:
-        camel_case_name_with_suffix = util.underscore_to_camelcase_upper(func['name'])
-
-        if util.is_ignored(func['name']) or util.is_control_ping(camel_case_name_with_suffix):
-            continue
-
-        if not util.is_reply(camel_case_name_with_suffix) and not util.is_notification(func['name']):
-            continue
-
-        camel_case_method_name = util.underscore_to_camelcase(func['name'])
-
-        if not util.is_notification(func["name"]):
-            camel_case_request_method_name = util.remove_reply_suffix(util.underscore_to_camelcase(func['name']))
-            if util.is_details(camel_case_name_with_suffix):
-                camel_case_reply_name = get_standard_dump_reply_name(util.underscore_to_camelcase_upper(func['name']),
-                                                                     func['name'])
-                callbacks.append(jvpp_facade_details_callback_method_template.substitute(base_package=base_package,
-                                                                                         plugin_package=plugin_package,
-                                                                                         dto_package=dto_package,
-                                                                                         callback_dto=camel_case_name_with_suffix,
-                                                                                         callback_dto_field=camel_case_method_name,
-                                                                                         callback_dto_reply_dump=camel_case_reply_name + dto_gen.dump_dto_suffix,
-                                                                                         future_package=future_facade_package))
-
-                methods.append(future_jvpp_method_template.substitute(plugin_package=plugin_package,
-                                                                      dto_package=dto_package,
-                                                                      method_name=camel_case_request_method_name +
-                                                                                  util.underscore_to_camelcase_upper(util.dump_suffix),
-                                                                      reply_name=camel_case_reply_name + dto_gen.dump_dto_suffix,
-                                                                      request_name=util.remove_reply_suffix(camel_case_reply_name) +
-                                                                                   util.underscore_to_camelcase_upper(util.dump_suffix)))
-                methods_impl.append(future_jvpp_dump_method_impl_template.substitute(plugin_package=plugin_package,
-                                                                                     dto_package=dto_package,
-                                                                                     method_name=camel_case_request_method_name +
-                                                                                                 util.underscore_to_camelcase_upper(util.dump_suffix),
-                                                                                     reply_name=camel_case_reply_name + dto_gen.dump_dto_suffix,
-                                                                                     request_name=util.remove_reply_suffix(camel_case_reply_name) +
-                                                                                                  util.underscore_to_camelcase_upper(util.dump_suffix)))
-            else:
-                request_name = util.underscore_to_camelcase_upper(util.unconventional_naming_rep_req[func['name']]) \
-                    if func['name'] in util.unconventional_naming_rep_req else util.remove_reply_suffix(camel_case_name_with_suffix)
-
-                methods.append(future_jvpp_method_template.substitute(plugin_package=plugin_package,
-                                                                      dto_package=dto_package,
-                                                                      method_name=camel_case_request_method_name,
-                                                                      reply_name=camel_case_name_with_suffix,
-                                                                      request_name=request_name))
-                methods_impl.append(future_jvpp_method_impl_template.substitute(plugin_package=plugin_package,
-                                                                                dto_package=dto_package,
-                                                                                method_name=camel_case_request_method_name,
-                                                                                reply_name=camel_case_name_with_suffix,
-                                                                                request_name=request_name))
-
-                callbacks.append(jvpp_facade_callback_method_template.substitute(base_package=base_package,
-                                                                                 plugin_package=plugin_package,
-                                                                                 dto_package=dto_package,
-                                                                                 callback_dto=camel_case_name_with_suffix))
-
-        if util.is_notification(func["name"]):
-            callbacks.append(jvpp_facade_callback_notification_method_template.substitute(plugin_package=plugin_package,
-                                                                                          dto_package=dto_package,
-                                                                                          callback_dto=util.add_notification_suffix(camel_case_name_with_suffix)))
-
-    jvpp_file = open(os.path.join(future_facade_package, "FutureJVpp%sFacadeCallback.java" % plugin_name), 'w')
-    jvpp_file.write(jvpp_facade_callback_template.substitute(inputfile=inputfile,
-                                                             base_package=base_package,
-                                                             plugin_package=plugin_package,
-                                                             plugin_name=plugin_name,
-                                                             dto_package=dto_package,
-                                                             notification_package=notification_package,
-                                                             callback_package=callback_package,
-                                                             methods="".join(callbacks),
-                                                             future_package=future_facade_package))
-    jvpp_file.flush()
-    jvpp_file.close()
-
-    jvpp_file = open(os.path.join(future_facade_package, "FutureJVpp%s.java" % plugin_name), 'w')
-    jvpp_file.write(future_jvpp_template.substitute(inputfile=inputfile,
-                                                    base_package=base_package,
-                                                    plugin_package=plugin_package,
-                                                    plugin_name=plugin_name,
-                                                    notification_package=notification_package,
-                                                    methods="".join(methods),
-                                                    future_package=future_facade_package))
-    jvpp_file.flush()
-    jvpp_file.close()
-
-    jvpp_file = open(os.path.join(future_facade_package, "FutureJVpp%sFacade.java" % plugin_name), 'w')
-    jvpp_file.write(future_jvpp_facade_template.substitute(inputfile=inputfile,
-                                                           base_package=base_package,
-                                                           plugin_package=plugin_package,
-                                                           plugin_name=plugin_name,
-                                                           dto_package=dto_package,
-                                                           notification_package=notification_package,
-                                                           methods="".join(methods_impl),
-                                                           future_package=future_facade_package))
-    jvpp_file.flush()
-    jvpp_file.close()
-
-
-future_jvpp_template = Template('''
-package $plugin_package.$future_package;
-
-/**
- * 
Async facade extension adding specific methods for each request invocation
- * 
It was generated by jvpp_future_facade_gen.py based on $inputfile
- * 
(python representation of api file generated by vppapigen).
- */
-public interface FutureJVpp${plugin_name} extends $base_package.$future_package.FutureJVppInvoker {
-$methods
-
-    @Override
-    public $plugin_package.$notification_package.${plugin_name}NotificationRegistry getNotificationRegistry();
-
-}
-''')
-
-future_jvpp_method_template = Template('''
-    java.util.concurrent.CompletionStage<$plugin_package.$dto_package.$reply_name> $method_name($plugin_package.$dto_package.$request_name request);
-''')
-
-
-future_jvpp_facade_template = Template('''
-package $plugin_package.$future_package;
-
-/**
- * 
Implementation of FutureJVpp based on AbstractFutureJVppInvoker
- * 
It was generated by jvpp_future_facade_gen.py based on $inputfile
- * 
(python representation of api file generated by vppapigen).
- */
-public class FutureJVpp${plugin_name}Facade extends $base_package.$future_package.AbstractFutureJVppInvoker implements FutureJVpp${plugin_name} {
-
-    private final $plugin_package.$notification_package.${plugin_name}NotificationRegistryImpl notificationRegistry = new $plugin_package.$notification_package.${plugin_name}NotificationRegistryImpl();
-
-    /**
-     * 
Create FutureJVpp${plugin_name}Facade object for provided JVpp instance.
-     * Constructor internally creates FutureJVppFacadeCallback class for processing callbacks
-     * and then connects to provided JVpp instance
-     *
-     * @param jvpp provided $base_package.JVpp instance
-     *
-     * @throws java.io.IOException in case instance cannot connect to JVPP
-     */
-    public FutureJVpp${plugin_name}Facade(final $base_package.JVppRegistry registry, final $base_package.JVpp jvpp) throws java.io.IOException {
-        super(jvpp, registry, new java.util.HashMap<>());
-        java.util.Objects.requireNonNull(registry, "JVppRegistry should not be null");
-        registry.register(jvpp, new FutureJVpp${plugin_name}FacadeCallback(getRequests(), notificationRegistry));
-    }
-
-    @Override
-    public $plugin_package.$notification_package.${plugin_name}NotificationRegistry getNotificationRegistry() {
-        return notificationRegistry;
-    }
-
-$methods
-}
-''')
-
-future_jvpp_method_impl_template = Template('''
-    @Override
-    public java.util.concurrent.CompletionStage<$plugin_package.$dto_package.$reply_name> $method_name($plugin_package.$dto_package.$request_name request) {
-        return send(request);
-    }
-''')
-
-future_jvpp_dump_method_impl_template = Template('''
-    @Override
-    public java.util.concurrent.CompletionStage<$plugin_package.$dto_package.$reply_name> $method_name($plugin_package.$dto_package.$request_name request) {
-        return send(request, new $plugin_package.$dto_package.$reply_name());
-    }
-''')
-
-
-# Returns request name or special one from unconventional_naming_rep_req map
-def get_standard_dump_reply_name(camel_case_dto_name, func_name):
-    # FIXME this is a hotfix for sub-details callbacks
-    # FIXME also for L2FibTableEntry
-    # It's all because unclear mapping between
-    #  request -> reply,
-    #  dump -> reply, details,
-    #  notification_start -> reply, notifications
-
-    # vpe.api needs to be "standardized" so we can parse the information and create maps before generating java code
-    suffix = func_name.split("_")[-1]
-    return util.underscore_to_camelcase_upper(
-        util.unconventional_naming_rep_req[func_name]) + util.underscore_to_camelcase_upper(suffix) if func_name in util.unconventional_naming_rep_req \
-        else camel_case_dto_name
diff --git a/vpp-api/java/jvpp/gen/jvppgen/jvpp_impl_gen.py b/vpp-api/java/jvpp/gen/jvppgen/jvpp_impl_gen.py
deleted file mode 100644
index 4146c141..00000000
--- a/vpp-api/java/jvpp/gen/jvppgen/jvpp_impl_gen.py
+++ /dev/null
@@ -1,219 +0,0 @@
-#!/usr/bin/env python
-#
-# Copyright (c) 2016 Cisco and/or its affiliates.
-# Licensed under the Apache License, Version 2.0 (the "License");
-# you may not use this file except in compliance with the License.
-# You may obtain a copy of the License at:
-#
-#     http://www.apache.org/licenses/LICENSE-2.0
-#
-# Unless required by applicable law or agreed to in writing, software
-# distributed under the License is distributed on an "AS IS" BASIS,
-# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-# See the License for the specific language governing permissions and
-# limitations under the License.
-
-import os, util
-from string import Template
-
-jvpp_ifc_template = Template("""
-package $plugin_package;
-
-/**
- * 
Java representation of plugin's api file.
- * 
It was generated by jvpp_impl_gen.py based on $inputfile
- * 
(python representation of api file generated by vppapigen).
- */
-public interface JVpp${plugin_name} extends $base_package.JVpp {
-
-    /**
-     * Generic dispatch method for sending requests to VPP
-     *
-     * @throws io.fd.vpp.jvpp.VppInvocationException if send request had failed
-     */
-    int send($base_package.$dto_package.JVppRequest request) throws io.fd.vpp.jvpp.VppInvocationException;
-
-$methods
-}
-""")
-
-jvpp_impl_template = Template("""
-package $plugin_package;
-
-import java.io.IOException;
-import java.io.InputStream;
-import java.nio.file.Files;
-import java.nio.file.Path;
-import java.nio.file.StandardCopyOption;
-import java.nio.file.attribute.PosixFilePermission;
-import java.nio.file.attribute.PosixFilePermissions;
-import java.util.Set;
-import java.util.logging.Logger;
-import $base_package.callback.JVppCallback;
-import $base_package.VppConnection;
-import $base_package.JVppRegistry;
-
-/**
- * 
Default implementation of JVpp interface.
- * 
It was generated by jvpp_impl_gen.py based on $inputfile
- * 
(python representation of api file generated by vppapigen).
- */
-public final class JVpp${plugin_name}Impl implements $plugin_package.JVpp${plugin_name} {
-
-    private final static Logger LOG = Logger.getLogger(JVpp${plugin_name}Impl.class.getName());
-    private static final String LIBNAME = "libjvpp_${plugin_name_underscore}.so.0.0.0";
-
-    // FIXME using NativeLibraryLoader makes load fail could not find (WantInterfaceEventsReply).
-    static {
-        try {
-            loadLibrary();
-        } catch (Exception e) {
-            LOG.severe("Can't find jvpp jni library: " + LIBNAME);
-            throw new ExceptionInInitializerError(e);
-        }
-    }
-
-    private static void loadStream(final InputStream is) throws IOException {
-        final Set perms = PosixFilePermissions.fromString("rwxr-x---");
-        final Path p = Files.createTempFile(LIBNAME, null, PosixFilePermissions.asFileAttribute(perms));
-        try {
-            Files.copy(is, p, StandardCopyOption.REPLACE_EXISTING);
-
-            try {
-                Runtime.getRuntime().load(p.toString());
-            } catch (UnsatisfiedLinkError e) {
-                throw new IOException("Failed to load library " + p, e);
-            }
-        } finally {
-            try {
-                Files.deleteIfExists(p);
-            } catch (IOException e) {
-            }
-        }
-    }
-
-    private static void loadLibrary() throws IOException {
-        try (final InputStream is = JVpp${plugin_name}Impl.class.getResourceAsStream('/' + LIBNAME)) {
-            if (is == null) {
-                throw new IOException("Failed to open library resource " + LIBNAME);
-            }
-            loadStream(is);
-        }
-    }
-
-    private VppConnection connection;
-    private JVppRegistry registry;
-
-    private static native void init0(final JVppCallback callback, final long queueAddress, final int clientIndex);
-    @Override
-    public void init(final JVppRegistry registry, final JVppCallback callback, final long queueAddress, final int clientIndex) {
-        this.registry = java.util.Objects.requireNonNull(registry, "registry should not be null");
-        this.connection = java.util.Objects.requireNonNull(registry.getConnection(), "connection should not be null");
-        connection.checkActive();
-        init0(callback, queueAddress, clientIndex);
-    }
-
-    private static native void close0();
-    @Override
-    public void close() {
-        close0();
-    }
-
-    @Override
-    public int send($base_package.$dto_package.JVppRequest request) throws io.fd.vpp.jvpp.VppInvocationException {
-        return request.send(this);
-    }
-
-    @Override
-    public final int controlPing(final io.fd.vpp.jvpp.dto.ControlPing controlPing) throws io.fd.vpp.jvpp.VppInvocationException {
-        return registry.controlPing(JVpp${plugin_name}Impl.class);
-    }
-
-$methods
-}
-""")
-
-method_template = Template("""    int $name($plugin_package.$dto_package.$request request) throws io.fd.vpp.jvpp.VppInvocationException;""")
-method_native_template = Template(
-    """    private static native int ${name}0($plugin_package.$dto_package.$request request);""")
-method_impl_template = Template("""    public final int $name($plugin_package.$dto_package.$request request) throws io.fd.vpp.jvpp.VppInvocationException {
-        java.util.Objects.requireNonNull(request,"Null request object");
-        connection.checkActive();
-        int result=${name}0(request);
-        if(result<0){
-            throw new io.fd.vpp.jvpp.VppInvocationException("${name}",result);
-        }
-        return result;
-    }
-""")
-
-no_arg_method_template = Template("""    int $name() throws io.fd.vpp.jvpp.VppInvocationException;""")
-no_arg_method_native_template = Template("""    private static native int ${name}0() throws io.fd.vpp.jvpp.VppInvocationException;""")
-no_arg_method_impl_template = Template("""    public final int $name() throws io.fd.vpp.jvpp.VppInvocationException {
-        connection.checkActive();
-        int result=${name}0();
-        if(result<0){
-            throw new io.fd.vpp.jvpp.VppInvocationException("${name}",result);
-        }
-        return result;
-    }
-""")
-
-
-def generate_jvpp(func_list, base_package, plugin_package, plugin_name_underscore, dto_package, inputfile):
-    """ Generates JVpp interface and JNI implementation """
-    print "Generating JVpp"
-    plugin_name = util.underscore_to_camelcase_upper(plugin_name_underscore)
-
-    methods = []
-    methods_impl = []
-    for func in func_list:
-
-        # Skip structures that are used only as notifications
-        if util.is_just_notification(func['name']) or util.is_ignored(func['name']):
-            continue
-
-        camel_case_name = util.underscore_to_camelcase(func['name'])
-        camel_case_name_upper = util.underscore_to_camelcase_upper(func['name'])
-        if util.is_reply(camel_case_name):
-            continue
-
-        if len(func['args']) == 0:
-            methods.append(no_arg_method_template.substitute(name=camel_case_name))
-            methods_impl.append(no_arg_method_native_template.substitute(name=camel_case_name))
-            methods_impl.append(no_arg_method_impl_template.substitute(name=camel_case_name))
-        else:
-            methods.append(method_template.substitute(name=camel_case_name,
-                                                      request=camel_case_name_upper,
-                                                      plugin_package=plugin_package,
-                                                      dto_package=dto_package))
-            methods_impl.append(method_native_template.substitute(name=camel_case_name,
-                                                                  request=camel_case_name_upper,
-                                                                  plugin_package=plugin_package,
-                                                                  dto_package=dto_package))
-            methods_impl.append(method_impl_template.substitute(name=camel_case_name,
-                                                                request=camel_case_name_upper,
-                                                                plugin_package=plugin_package,
-                                                                dto_package=dto_package))
-
-    jvpp_file = open("JVpp%s.java" % plugin_name, 'w')
-    jvpp_file.write(
-        jvpp_ifc_template.substitute(inputfile=inputfile,
-                                     methods="\n".join(methods),
-                                     base_package=base_package,
-                                     plugin_package=plugin_package,
-                                     plugin_name=plugin_name,
-                                     dto_package=dto_package))
-    jvpp_file.flush()
-    jvpp_file.close()
-
-    jvpp_file = open("JVpp%sImpl.java" % plugin_name, 'w')
-    jvpp_file.write(jvpp_impl_template.substitute(inputfile=inputfile,
-                                                  methods="\n".join(methods_impl),
-                                                  base_package=base_package,
-                                                  plugin_package=plugin_package,
-                                                  plugin_name=plugin_name,
-                                                  plugin_name_underscore=plugin_name_underscore,
-                                                  dto_package=dto_package))
-    jvpp_file.flush()
-    jvpp_file.close()
diff --git a/vpp-api/java/jvpp/gen/jvppgen/notification_gen.py b/vpp-api/java/jvpp/gen/jvppgen/notification_gen.py
deleted file mode 100644
index eb380fc8..00000000
--- a/vpp-api/java/jvpp/gen/jvppgen/notification_gen.py
+++ /dev/null
@@ -1,199 +0,0 @@
-#!/usr/bin/env python
-#
-# Copyright (c) 2016 Cisco and/or its affiliates.
-# Licensed under the Apache License, Version 2.0 (the "License");
-# you may not use this file except in compliance with the License.
-# You may obtain a copy of the License at:
-#
-#     http://www.apache.org/licenses/LICENSE-2.0
-#
-# Unless required by applicable law or agreed to in writing, software
-# distributed under the License is distributed on an "AS IS" BASIS,
-# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-# See the License for the specific language governing permissions and
-# limitations under the License.
-
-import os
-
-import callback_gen
-import util
-from string import Template
-
-notification_registry_template = Template("""
-package $plugin_package.$notification_package;
-
-/**
- * 
Registry for notification callbacks defined in ${plugin_name}.
- * 
It was generated by notification_gen.py based on $inputfile
- * 
(python representation of api file generated by vppapigen).
- */
-public interface ${plugin_name}NotificationRegistry extends $base_package.$notification_package.NotificationRegistry {
-
-    $register_callback_methods
-
-    @Override
-    void close();
-}
-""")
-
-global_notification_callback_template = Template("""
-package $plugin_package.$notification_package;
-
-/**
- * 
Aggregated callback interface for notifications only.
- * 
It was generated by notification_gen.py based on $inputfile
- * 
(python representation of api file generated by vppapigen).
- */
-public interface Global${plugin_name}NotificationCallback$callbacks {
-
-}
-""")
-
-notification_registry_impl_template = Template("""
-package $plugin_package.$notification_package;
-
-/**
- * 
Notification registry delegating notification processing to registered callbacks.
- * 
It was generated by notification_gen.py based on $inputfile
- * 
(python representation of api file generated by vppapigen).
- */
-public final class ${plugin_name}NotificationRegistryImpl implements ${plugin_name}NotificationRegistry, Global${plugin_name}NotificationCallback {
-
-    // TODO add a special NotificationCallback interface and only allow those to be registered
-    private final java.util.concurrent.ConcurrentMap, $base_package.$callback_package.JVppNotificationCallback> registeredCallbacks =
-        new java.util.concurrent.ConcurrentHashMap<>();
-
-    $register_callback_methods
-    $handler_methods
-
-    @Override
-    public void close() {
-        registeredCallbacks.clear();
-    }
-}
-""")
-
-register_callback_impl_template = Template("""
-    public java.lang.AutoCloseable register$callback(final $plugin_package.$callback_package.$callback callback){
-        if(null != registeredCallbacks.putIfAbsent($plugin_package.$dto_package.$notification.class, callback)){
-            throw new IllegalArgumentException("Callback for " + $plugin_package.$dto_package.$notification.class +
-                "notification already registered");
-        }
-        return () -> registeredCallbacks.remove($plugin_package.$dto_package.$notification.class);
-    }
-""")
-
-handler_impl_template = Template("""
-    @Override
-    public void on$notification(
-        final $plugin_package.$dto_package.$notification notification) {
-        final $base_package.$callback_package.JVppNotificationCallback jVppNotificationCallback = registeredCallbacks.get($plugin_package.$dto_package.$notification.class);
-        if (null != jVppNotificationCallback) {
-            (($plugin_package.$callback_package.$callback) registeredCallbacks
-                .get($plugin_package.$dto_package.$notification.class))
-                .on$notification(notification);
-        }
-    }
-""")
-
-notification_provider_template = Template("""
-package $plugin_package.$notification_package;
-
- /**
- * Provides ${plugin_name}NotificationRegistry.
- * 
The file was generated by notification_gen.py based on $inputfile
- * 
(python representation of api file generated by vppapigen).
- */
-public interface ${plugin_name}NotificationRegistryProvider extends $base_package.$notification_package.NotificationRegistryProvider {
-
-    @Override
-    public ${plugin_name}NotificationRegistry getNotificationRegistry();
-}
-""")
-
-
-def generate_notification_registry(func_list, base_package, plugin_package, plugin_name, notification_package, callback_package, dto_package, inputfile):
-    """ Generates notification registry interface and implementation """
-    print "Generating Notification interfaces and implementation"
-
-    if not os.path.exists(notification_package):
-        raise Exception("%s folder is missing" % notification_package)
-
-    callbacks = []
-    register_callback_methods = []
-    register_callback_methods_impl = []
-    handler_methods = []
-    for func in func_list:
-
-        if not util.is_notification(func['name']):
-            continue
-
-        camel_case_name_with_suffix = util.underscore_to_camelcase_upper(func['name'])
-        notification_dto = util.add_notification_suffix(camel_case_name_with_suffix)
-        callback_ifc = notification_dto + callback_gen.callback_suffix
-        fully_qualified_callback_ifc = "{0}.{1}.{2}".format(plugin_package, callback_package, callback_ifc)
-        callbacks.append(fully_qualified_callback_ifc)
-
-        # TODO create NotificationListenerRegistration and return that instead of AutoCloseable to better indicate
-        # that the registration should be closed
-        register_callback_methods.append("java.lang.AutoCloseable register{0}({1} callback);"
-                                         .format(callback_ifc, fully_qualified_callback_ifc))
-        register_callback_methods_impl.append(register_callback_impl_template.substitute(plugin_package=plugin_package,
-                                                                                         callback_package=callback_package,
-                                                                                         dto_package=dto_package,
-                                                                                         notification=notification_dto,
-                                                                                         callback=callback_ifc))
-        handler_methods.append(handler_impl_template.substitute(base_package=base_package,
-                                                                plugin_package=plugin_package,
-                                                                callback_package=callback_package,
-                                                                dto_package=dto_package,
-                                                                notification=notification_dto,
-                                                                callback=callback_ifc))
-
-
-    callback_file = open(os.path.join(notification_package, "%sNotificationRegistry.java" % plugin_name), 'w')
-    callback_file.write(notification_registry_template.substitute(inputfile=inputfile,
-                                                                register_callback_methods="\n    ".join(register_callback_methods),
-                                                                base_package=base_package,
-                                                                plugin_package=plugin_package,
-                                                                plugin_name=plugin_name,
-                                                                notification_package=notification_package))
-    callback_file.flush()
-    callback_file.close()
-
-    callback_file = open(os.path.join(notification_package, "Global%sNotificationCallback.java" % plugin_name), 'w')
-
-    global_notification_callback_callbacks = ""
-    if (callbacks):
-        global_notification_callback_callbacks = " extends " + ", ".join(callbacks)
-
-    callback_file.write(global_notification_callback_template.substitute(inputfile=inputfile,
-                                                                       callbacks=global_notification_callback_callbacks,
-                                                                       plugin_package=plugin_package,
-                                                                       plugin_name=plugin_name,
-                                                                       notification_package=notification_package))
-    callback_file.flush()
-    callback_file.close()
-
-    callback_file = open(os.path.join(notification_package, "%sNotificationRegistryImpl.java" % plugin_name), 'w')
-    callback_file.write(notification_registry_impl_template.substitute(inputfile=inputfile,
-                                                                     callback_package=callback_package,
-                                                                     dto_package=dto_package,
-                                                                     register_callback_methods="".join(register_callback_methods_impl),
-                                                                     handler_methods="".join(handler_methods),
-                                                                     base_package=base_package,
-                                                                     plugin_package=plugin_package,
-                                                                     plugin_name=plugin_name,
-                                                                     notification_package=notification_package))
-    callback_file.flush()
-    callback_file.close()
-
-    callback_file = open(os.path.join(notification_package, "%sNotificationRegistryProvider.java" % plugin_name), 'w')
-    callback_file.write(notification_provider_template.substitute(inputfile=inputfile,
-                                                                     base_package=base_package,
-                                                                     plugin_package=plugin_package,
-                                                                     plugin_name=plugin_name,
-                                                                     notification_package=notification_package))
-    callback_file.flush()
-    callback_file.close()
-
diff --git a/vpp-api/java/jvpp/gen/jvppgen/types_gen.py b/vpp-api/java/jvpp/gen/jvppgen/types_gen.py
deleted file mode 100644
index d12fb3d7..00000000
--- a/vpp-api/java/jvpp/gen/jvppgen/types_gen.py
+++ /dev/null
@@ -1,227 +0,0 @@
-#!/usr/bin/env python
-#
-# Copyright (c) 2016 Cisco and/or its affiliates.
-# Licensed under the Apache License, Version 2.0 (the "License");
-# you may not use this file except in compliance with the License.
-# You may obtain a copy of the License at:
-#
-#     http://www.apache.org/licenses/LICENSE-2.0
-#
-# Unless required by applicable law or agreed to in writing, software
-# distributed under the License is distributed on an "AS IS" BASIS,
-# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-# See the License for the specific language governing permissions and
-# limitations under the License.
-
-import os
-from string import Template
-
-import util
-import jni_gen
-import dto_gen
-
-type_template = Template("""
-package $plugin_package.$type_package;
-
-/**
- * 
This class represents $c_type_name type definition.
- * 
It was generated by types_gen.py based on $inputfile preparsed data:
- * 
-$docs
- * 

- */
-public final class $java_type_name {
-$fields
-$methods
-}
-""")
-
-field_template = Template("""    public $type $name;\n""")
-
-
-def generate_type_fields(type_definition):
-    """
-    Generates fields for class representing typeonly definition
-    :param type_definition: python representation of typeonly definition
-    :return: string representing class fields
-    """
-    fields = ""
-    for t in zip(type_definition['types'], type_definition['args']):
-        field_name = util.underscore_to_camelcase(t[1])
-        fields += field_template.substitute(type=util.jni_2_java_type_mapping[t[0]],
-                                            name=field_name)
-    return fields
-
-object_struct_setter_template = Template("""
-    {
-        jclass ${field_reference_name}Class = (*env)->FindClass(env, "${class_FQN}");
-        memset (&(mp->${c_name}), 0, sizeof (mp->${c_name}));
-        ${struct_initialization}
-    }
-""")
-
-object_array_struct_setter_template = Template("""
-    {
-        jclass ${field_reference_name}ArrayElementClass = (*env)->FindClass(env, "${class_FQN}");
-        if (${field_reference_name}) {
-            size_t _i;
-            jsize cnt = (*env)->GetArrayLength (env, ${field_reference_name});
-            ${field_length_check}
-            for (_i = 0; _i < cnt; _i++) {
-                jobject ${field_reference_name}ArrayElement = (*env)->GetObjectArrayElement(env, ${field_reference_name}, _i);
-                memset (&(mp->${c_name}[_i]), 0, sizeof (mp->${c_name}[_i]));
-                ${struct_initialization}
-            }
-        }
-    }
-""")
-
-object_dto_field_setter_template = Template("""
-    {
-        jclass ${field_reference_name}Class = (*env)->FindClass(env, "${class_FQN}");
-        jmethodID ${field_reference_name}Constructor = (*env)->GetMethodID(env, ${field_reference_name}Class, "", "()V");
-        jobject ${field_reference_name} = (*env)->NewObject(env, ${field_reference_name}Class,  ${field_reference_name}Constructor);
-        ${type_initialization}
-        (*env)->SetObjectField(env, dto, ${field_reference_name}FieldId, ${field_reference_name});
-    }
-""")
-
-object_array_dto_field_setter_template = Template("""
-    {
-        jclass ${field_reference_name}Class = (*env)->FindClass(env, "${class_FQN}");
-        jobjectArray ${field_reference_name} = (*env)->NewObjectArray(env, ${field_length}, ${field_reference_name}Class, 0);
-        unsigned int _i;
-        for (_i = 0; _i < ${field_length}; _i++) {
-            jmethodID ${field_reference_name}Constructor = (*env)->GetMethodID(env, ${field_reference_name}Class, "", "()V");
-            jobject ${field_reference_name}ArrayElement = (*env)->NewObject(env, ${field_reference_name}Class,  ${field_reference_name}Constructor);
-            ${type_initialization}
-            (*env)->SetObjectArrayElement(env, ${field_reference_name}, _i, ${field_reference_name}ArrayElement);
-        }
-        (*env)->SetObjectField(env, dto, ${field_reference_name}FieldId, ${field_reference_name});
-    }
-""")
-
-
-def generate_struct_initialization(type_def, c_name_prefix, object_name, indent):
-    struct_initialization = ""
-    # field identifiers
-    for t in zip(type_def['types'], type_def['args'], type_def['lengths']):
-        field_reference_name = "${c_name}" + util.underscore_to_camelcase_upper(t[1])
-        field_name = util.underscore_to_camelcase(t[1])
-        struct_initialization += jni_gen.jni_request_binding_for_type(field_type=t[0], c_name=c_name_prefix + t[1],
-                                                                     field_reference_name=field_reference_name,
-                                                                     field_name=field_name,
-                                                                     field_length=t[2][0],
-                                                                     is_variable_len_array=t[2][1],
-                                                                     object_name=object_name)
-    return indent + struct_initialization.replace('\n', '\n' + indent)
-
-
-def generate_type_setter(handler_name, type_def, c_name_prefix, object_name, indent):
-    type_initialization = ""
-    for t in zip(type_def['types'], type_def['args'], type_def['lengths']):
-        field_length = t[2][0]
-        is_variable_len_array = t[2][1]
-        length_field_type = None
-        if is_variable_len_array:
-            length_field_type = type_def['types'][type_def['args'].index(field_length)]
-        type_initialization += jni_gen.jni_reply_handler_for_type(handler_name=handler_name,
-                                                                  ref_name="${field_reference_name}",
-                                                                  field_type=t[0], c_name=c_name_prefix + t[1],
-                                                                  field_reference_name="${c_name}" + util.underscore_to_camelcase_upper(t[1]),
-                                                                  field_name=util.underscore_to_camelcase(t[1]),
-                                                                  field_length=field_length,
-                                                                  is_variable_len_array=is_variable_len_array,
-                                                                  length_field_type=length_field_type,
-                                                                  object_name=object_name)
-    return indent + type_initialization.replace('\n', '\n' + indent)
-
-
-def generate_types(types_list, plugin_package, types_package, inputfile):
-    """
-    Generates Java representation of custom types defined in api file.
-    """
-
-    #
-    if not types_list:
-        print "Skipping custom types generation (%s does not define custom types)." % inputfile
-        return
-
-    print "Generating custom types"
-
-    if not os.path.exists(types_package):
-        raise Exception("%s folder is missing" % types_package)
-
-    for type in types_list:
-        c_type_name = type['name']
-        java_type_name = util.underscore_to_camelcase_upper(type['name'])
-        dto_path = os.path.join(types_package, java_type_name + ".java")
-
-        fields = generate_type_fields(type)
-
-        dto_file = open(dto_path, 'w')
-        dto_file.write(type_template.substitute(plugin_package=plugin_package,
-                                                type_package=types_package,
-                                                c_type_name=c_type_name,
-                                                inputfile=inputfile,
-                                                docs=util.api_message_to_javadoc(type),
-                                                java_type_name=java_type_name,
-                                                fields=fields,
-                                                methods=dto_gen.generate_dto_base_methods(java_type_name, type)
-                                                ))
-
-        # update type mappings:
-        # todo fix vpe.api to use type_name instead of vl_api_type_name_t
-        type_name = "vl_api_" + c_type_name + "_t"
-        java_fqn = "%s.%s.%s" % (plugin_package, types_package, java_type_name)
-        util.vpp_2_jni_type_mapping[type_name] = "jobject"
-        util.vpp_2_jni_type_mapping[type_name + "[]"] = "jobjectArray"
-        util.jni_2_java_type_mapping[type_name] = java_fqn
-        util.jni_2_java_type_mapping[type_name + "[]"] = java_fqn + "[]"
-        jni_name = java_fqn.replace('.', "/")
-        jni_signature = "L" + jni_name + ";"
-        util.jni_2_signature_mapping[type_name] = "L" + jni_name + ";"
-        util.jni_2_signature_mapping[type_name + "[]"] = "[" + jni_signature
-        util.jni_field_accessors[type_name] = "ObjectField"
-        util.jni_field_accessors[type_name + "[]"] = "ObjectField"
-
-        jni_gen.struct_setter_templates[type_name] = Template(
-                object_struct_setter_template.substitute(
-                        c_name="${c_name}",
-                        field_reference_name="${field_reference_name}",
-                        class_FQN=jni_name,
-                        struct_initialization=generate_struct_initialization(type, "${c_name}.",
-                                                                           "${field_reference_name}", ' ' * 4))
-        )
-
-        jni_gen.struct_setter_templates[type_name+ "[]"] = Template(
-                object_array_struct_setter_template.substitute(
-                        c_name="${c_name}",
-                        field_reference_name="${field_reference_name}",
-                        field_length_check="${field_length_check}",
-                        class_FQN=jni_name,
-                        struct_initialization=generate_struct_initialization(type, "${c_name}[_i].",
-                                                                           "${field_reference_name}ArrayElement", ' ' * 8))
-        )
-
-        jni_gen.dto_field_setter_templates[type_name] = Template(
-                object_dto_field_setter_template.substitute(
-                        field_reference_name="${field_reference_name}",
-                        field_length="${field_length}",
-                        class_FQN=jni_name,
-                        type_initialization=generate_type_setter(c_type_name, type, "${c_name}.",
-                                                                 "${field_reference_name}", ' ' * 4))
-        )
-
-        jni_gen.dto_field_setter_templates[type_name + "[]"] = Template(
-                object_array_dto_field_setter_template.substitute(
-                        field_reference_name="${field_reference_name}",
-                        field_length="${field_length}",
-                        class_FQN=jni_name,
-                        type_initialization=generate_type_setter(c_type_name, type, "${c_name}[_i].",
-                                                                 "${field_reference_name}ArrayElement", ' ' * 8))
-        )
-
-        dto_file.flush()
-        dto_file.close()
-
diff --git a/vpp-api/java/jvpp/gen/jvppgen/util.py b/vpp-api/java/jvpp/gen/jvppgen/util.py
deleted file mode 100644
index fc971c17..00000000
--- a/vpp-api/java/jvpp/gen/jvppgen/util.py
+++ /dev/null
@@ -1,220 +0,0 @@
-#!/usr/bin/env python
-#
-# Copyright (c) 2016 Cisco and/or its affiliates.
-# Licensed under the Apache License, Version 2.0 (the "License");
-# you may not use this file except in compliance with the License.
-# You may obtain a copy of the License at:
-#
-#     http://www.apache.org/licenses/LICENSE-2.0
-#
-# Unless required by applicable law or agreed to in writing, software
-# distributed under the License is distributed on an "AS IS" BASIS,
-# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-# See the License for the specific language governing permissions and
-# limitations under the License.
-
-import os, pprint
-from os import removedirs
-
-
-def underscore_to_camelcase(name):
-    name = name.title().replace("_", "")
-    return name[0].lower() + name[1:]
-
-
-def underscore_to_camelcase_upper(name):
-    name = name.title().replace("_", "")
-    return name[0].upper() + name[1:]
-
-
-def remove_folder(folder):
-    """ Remove folder with all its files """
-    for root, dirs, files in os.walk(folder, topdown=False):
-        for name in files:
-            os.remove(os.path.join(root, name))
-        removedirs(folder)
-
-
-reply_suffixes = ("reply", "details", "l2fibtableentry")
-
-
-def is_reply(name):
-    return name.lower().endswith(reply_suffixes)
-
-
-def is_details(name):
-    return name.lower().endswith(reply_suffixes[1]) or name.lower().endswith(reply_suffixes[2])
-
-
-def is_retval_field(name):
-    return name == 'retval'
-
-dump_suffix = "dump"
-
-
-def is_dump(name):
-    return name.lower().endswith(dump_suffix)
-
-
-def get_reply_suffix(name):
-    for reply_suffix in reply_suffixes:
-        if name.lower().endswith(reply_suffix):
-            if reply_suffix == reply_suffixes[2]:
-                # FIXME workaround for l2_fib_table_entry
-                return 'entry'
-            else:
-                return reply_suffix
-
-# Mapping according to:
-# http://docs.oracle.com/javase/7/docs/technotes/guides/jni/spec/types.html
-#
-# Unsigned types are converted to signed java types that have the same size.
-# It is the API user responsibility to interpret them correctly.
-jni_2_java_type_mapping = {'u8': 'byte',
-                           'u8[]': 'byte[]',
-                           'i8': 'byte',
-                           'i8[]': 'byte[]',
-                           'u16': 'short',
-                           'u16[]': 'short[]',
-                           'i16': 'short',
-                           'i16[]': 'short[]',
-                           'u32': 'int',
-                           'u32[]': 'int[]',
-                           'i32': 'int',
-                           'i32[]': 'int[]',
-                           'u64': 'long',
-                           'u64[]': 'long[]',
-                           'i64': 'long',
-                           'i64[]': 'long[]',
-                           'f64': 'double',
-                           'f64[]': 'double[]'
-                           }
-
-vpp_2_jni_type_mapping = {'u8': 'jbyte',
-                          'u8[]': 'jbyteArray',
-                          'i8': 'jbyte',
-                          'u8[]': 'jbyteArray',
-                          'u16': 'jshort',
-                          'u16[]': 'jshortArray',
-                          'i16': 'jshort',
-                          'i16[]': 'jshortArray',
-                          'u32': 'jint',
-                          'u32[]': 'jintArray',
-                          'i32': 'jint',
-                          'i32[]': 'jintArray',
-                          'u64': 'jlong',
-                          'u64[]': 'jlongArray',
-                          'i64': 'jlong',
-                          'i64[]': 'jlongArray',
-                          'f64': 'jdouble',
-                          'f64[]': 'jdoubleArray'
-                          }
-
-# https://docs.oracle.com/javase/8/docs/technotes/guides/jni/spec/types.html#type_signatures
-jni_2_signature_mapping = {'u8': 'B',
-                           'u8[]': '[B',
-                           'i8': 'B',
-                           'i8[]': '[B',
-                           'u16': 'S',
-                           'u16[]': '[S',
-                           'i16': 'S',
-                           'i16[]': '[S',
-                           'u32': 'I',
-                           'u32[]': '[I',
-                           'i32': 'I',
-                           'i32[]': '[I',
-                           'u64': 'J',
-                           'u64[]': '[J',
-                           'i64': 'J',
-                           'i64[]': '[J',
-                           'f64': 'D',
-                           'f64[]': '[D'
-                           }
-
-# https://docs.oracle.com/javase/8/docs/technotes/guides/jni/spec/functions.html#Get_type_Field_routines
-jni_field_accessors =  {'u8': 'ByteField',
-                        'u8[]': 'ObjectField',
-                        'i8': 'ByteField',
-                        'i8[]': 'ObjectField',
-                        'u16': 'ShortField',
-                        'u16[]': 'ObjectField',
-                        'i16': 'ShortField',
-                        'i16[]': 'ObjectField',
-                        'u32': 'IntField',
-                        'u32[]': 'ObjectField',
-                        'i32': 'IntField',
-                        'i32[]': 'ObjectField',
-                        'u64': 'LongField',
-                        'u64[]': 'ObjectField',
-                        'i64': 'LongField',
-                        'i64[]': 'ObjectField',
-                        'f64': 'DoubleField',
-                        'f64[]': 'ObjectField'
-                        }
-
-
-# vpe.api calls that do not follow naming conventions and have to be handled exceptionally when finding reply -> request mapping
-# FIXME in vpe.api
-unconventional_naming_rep_req = {
-                                 'cli_reply': 'cli_request',
-                                 'vnet_summary_stats_reply': 'vnet_get_summary_stats',
-                                 # This below is actually a sub-details callback. We cannot derive the mapping of dump request
-                                 # belonging to this sub-details from naming conventions. We need special mapping
-                                 'bridge_domain_sw_if_details': 'bridge_domain',
-                                 # This is standard dump call + details reply. However it's not called details but entry
-                                 'l2_fib_table_entry': 'l2_fib_table'
-                                 }
-
-#
-# FIXME no convention in the naming of events (notifications) in vpe.api
-notifications_message_suffixes = ("event", "counters")
-notification_messages_reused = ["sw_interface_set_flags"]
-
-# messages that must be ignored. These messages are INSUFFICIENTLY marked as disabled in vpe.api
-# FIXME
-ignored_messages = ["is_address_reachable"]
-
-
-def is_notification(name):
-    """ Returns true if the structure is a notification regardless of its no other use """
-    return is_just_notification(name) or name.lower() in notification_messages_reused
-
-
-def is_just_notification(name):
-    """ Returns true if the structure is just a notification and has no other use """
-    return name.lower().endswith(notifications_message_suffixes)
-
-
-def is_ignored(param):
-    return param.lower() in ignored_messages
-
-
-def remove_reply_suffix(camel_case_name_with_suffix):
-    return remove_suffix(camel_case_name_with_suffix, get_reply_suffix(camel_case_name_with_suffix))
-
-
-def remove_suffix(camel_case_name_with_suffix, suffix):
-    suffix_length = len(suffix)
-    return camel_case_name_with_suffix[:-suffix_length] if suffix_length != 0 else camel_case_name_with_suffix
-
-
-def is_control_ping(camel_case_name_with_suffix):
-    return camel_case_name_with_suffix.lower().startswith("controlping");
-
-
-def api_message_to_javadoc(api_message):
-    """ Converts vpe.api message description to javadoc """
-    str = pprint.pformat(api_message, indent=4, width=120, depth=None)
-    return " * " + str.replace("\n", "\n * ")
-
-
-notification_dto_suffix = "Notification"
-
-
-def add_notification_suffix(camel_case_dto_name):
-    camel_case_dto_name += notification_dto_suffix
-    return camel_case_dto_name
-
-
-def is_array(java_type_as_string):
-    return java_type_as_string.endswith("[]")
diff --git a/vpp-api/java/m4/ax_check_java_home.m4 b/vpp-api/java/m4/ax_check_java_home.m4
deleted file mode 100644
index cfe8f589..00000000
--- a/vpp-api/java/m4/ax_check_java_home.m4
+++ /dev/null
@@ -1,80 +0,0 @@
-# ===========================================================================
-#    http://www.gnu.org/software/autoconf-archive/ax_check_java_home.html
-# ===========================================================================
-#
-# SYNOPSIS
-#
-#   AX_CHECK_JAVA_HOME
-#
-# DESCRIPTION
-#
-#   Check for Sun Java (JDK / JRE) installation, where the 'java' VM is in.
-#   If found, set environment variable JAVA_HOME = Java installation home,
-#   else left JAVA_HOME untouch, which in most case means JAVA_HOME is
-#   empty.
-#
-# LICENSE
-#
-#   Copyright (c) 2008 Gleen Salmon 
-#
-#   This program is free software; you can redistribute it and/or modify it
-#   under the terms of the GNU General Public License as published by the
-#   Free Software Foundation; either version 2 of the License, or (at your
-#   option) any later version.
-#
-#   This program is distributed in the hope that it will be useful, but
-#   WITHOUT ANY WARRANTY; without even the implied warranty of
-#   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General
-#   Public License for more details.
-#
-#   You should have received a copy of the GNU General Public License along
-#   with this program. If not, see .
-#
-#   As a special exception, the respective Autoconf Macro's copyright owner
-#   gives unlimited permission to copy, distribute and modify the configure
-#   scripts that are the output of Autoconf when processing the Macro. You
-#   need not follow the terms of the GNU General Public License when using
-#   or distributing such scripts, even though portions of the text of the
-#   Macro appear in them. The GNU General Public License (GPL) does govern
-#   all other use of the material that constitutes the Autoconf Macro.
-#
-#   This special exception to the GPL applies to versions of the Autoconf
-#   Macro released by the Autoconf Archive. When you make and distribute a
-#   modified version of the Autoconf Macro, you may extend this special
-#   exception to the GPL to apply to your modified version as well.
-
-#serial 6
-
-AU_ALIAS([AC_CHECK_JAVA_HOME], [AX_CHECK_JAVA_HOME])
-
-AC_DEFUN([AX_CHECK_JAVA_HOME],
-[AC_MSG_CHECKING([for JAVA_HOME])
-# We used a fake loop so that we can use "break" to exit when the result
-# is found.
-while true
-do
-  # If the user defined JAVA_HOME, don't touch it.
-  test "${JAVA_HOME+set}" = set && break
-
-  # On Mac OS X 10.5 and following, run /usr/libexec/java_home to get
-  # the value of JAVA_HOME to use.
-  # (http://developer.apple.com/library/mac/#qa/qa2001/qa1170.html).
-  JAVA_HOME=`/usr/libexec/java_home 2>/dev/null`
-  test x"$JAVA_HOME" != x && break
-
-  # See if we can find the java executable, and compute from there.
-  TRY_JAVA_HOME=`ls -dr /usr/java/* 2> /dev/null | head -n 1`
-  if test x$TRY_JAVA_HOME != x; then
-    PATH=$PATH:$TRY_JAVA_HOME/bin
-  fi
-  AC_PATH_PROG([JAVA_PATH_NAME], [java])
-  if test "x$JAVA_PATH_NAME" != x; then
-    JAVA_HOME=`echo $JAVA_PATH_NAME | sed "s/\(.*\)[[/]]bin[[/]]java.*/\1/"`
-    break
-  fi
-
-  AC_MSG_NOTICE([Could not compute JAVA_HOME])
-  break
-done
-AC_MSG_RESULT([$JAVA_HOME])
-])
diff --git a/vpp-api/java/m4/ax_check_java_plugin.m4 b/vpp-api/java/m4/ax_check_java_plugin.m4
deleted file mode 100644
index 920753e5..00000000
--- a/vpp-api/java/m4/ax_check_java_plugin.m4
+++ /dev/null
@@ -1,101 +0,0 @@
-# ===========================================================================
-#   http://www.gnu.org/software/autoconf-archive/ax_check_java_plugin.html
-# ===========================================================================
-#
-# SYNOPSIS
-#
-#   AX_CHECK_JAVA_PLUGIN()
-#
-# DESCRIPTION
-#
-#   This macro sets  to empty on failure and to a compatible
-#   version of plugin.jar otherwise. Directories searched are /usr/java/*
-#   and /usr/local/java/*, which are assumed to be j{dk,re} installations.
-#   Apply the shell variable as you see fit. If sun changes things so
-#   /lib/plugin.jar is not the magic file it will stop working.
-#
-#   This macro assumes that unzip, zipinfo or pkzipc is available (and can
-#   list the contents of the jar archive). The first two are assumed to work
-#   similarly enough to the infozip versisonms. The pkzipc version is
-#   assumed to work if I undertstand the documentation on pkware's site but
-#   YMMV. I do not have access to pwkware's version to test it.
-#
-# LICENSE
-#
-#   Copyright (c) 2008 Duncan Simpson 
-#
-#   This program is free software; you can redistribute it and/or modify it
-#   under the terms of the GNU General Public License as published by the
-#   Free Software Foundation; either version 2 of the License, or (at your
-#   option) any later version.
-#
-#   This program is distributed in the hope that it will be useful, but
-#   WITHOUT ANY WARRANTY; without even the implied warranty of
-#   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General
-#   Public License for more details.
-#
-#   You should have received a copy of the GNU General Public License along
-#   with this program. If not, see .
-#
-#   As a special exception, the respective Autoconf Macro's copyright owner
-#   gives unlimited permission to copy, distribute and modify the configure
-#   scripts that are the output of Autoconf when processing the Macro. You
-#   need not follow the terms of the GNU General Public License when using
-#   or distributing such scripts, even though portions of the text of the
-#   Macro appear in them. The GNU General Public License (GPL) does govern
-#   all other use of the material that constitutes the Autoconf Macro.
-#
-#   This special exception to the GPL applies to versions of the Autoconf
-#   Macro released by the Autoconf Archive. When you make and distribute a
-#   modified version of the Autoconf Macro, you may extend this special
-#   exception to the GPL to apply to your modified version as well.
-
-#serial 7
-
-AU_ALIAS([DPS_CHECK_PLUGIN], [AX_CHECK_JAVA_PLUGIN])
-AC_DEFUN([AX_CHECK_JAVA_PLUGIN],
-[AC_REQUIRE([AC_PROG_AWK])
-AC_REQUIRE([AC_PROG_FGREP])
-AC_CHECK_PROG(ZIPINFO,[zipinfo unzip pkzipc])
-AC_MSG_CHECKING([for the java plugin])
-case "x$ZIPINFO" in
-[*/zipinfo)]
-	zipinf="zipinfo -1" ;;
-[*/unzip)]
-	zipinf="unzip -l";;
-[*/pkzipc)]
-	ziping="unzipc -view";;
-[x*)]
-	AC_MSG_RESULT([skiped, none of zipinfo, unzip and pkzipc found])
-	AC_SUBST($1,[])
-	zipinf="";;
-esac
-if test "x$zipinf" != "x"; then
-jplugin=""
-for jhome in `ls -dr /usr/java/* /usr/local/java/* 2> /dev/null`; do
-for jfile in lib/plugin.jar jre/lib/plugin.jar; do
-if test "x$jplugin" = "x" && test -f "$jhome/$jfile"; then
-eval "$zipinf $jhome/$jfile | $AWK '{ print \$NF; }' | $FGREP netscape/javascript/JSObject" >/dev/null 2>/dev/null
-if test $? -eq 0; then
-dnl Some version of gcj (and javac) refuse to work with some files
-dnl that pass this test. To stop this problem make sure that the compiler
-dnl still works with this jar file in the classpath
-cat << \EOF > Test.java
-/* [#]line __oline__ "configure" */
-public class Test {
-}
-EOF
-if eval "$JAVAC -classpath $jhome/$jfile Test.java 2>/dev/null >/dev/null" && test -f Test.class; then
-jplugin="$jhome/$jfile"
-fi
-rm -f Test.java Test.class
-fi; fi; done; done
-if test "x$jplugin" != "x"; then
-AC_SUBST($1,$jplugin)
-AC_MSG_RESULT($jplugin)
-else
-AC_MSG_RESULT([java plugin not found])
-AC_SUBST($1,[])
-fi
-fi
-])
diff --git a/vpp-api/java/m4/ax_java_check_class.m4 b/vpp-api/java/m4/ax_java_check_class.m4
deleted file mode 100644
index 917638ae..00000000
--- a/vpp-api/java/m4/ax_java_check_class.m4
+++ /dev/null
@@ -1,85 +0,0 @@
-# ===========================================================================
-#    http://www.gnu.org/software/autoconf-archive/ax_java_check_class.html
-# ===========================================================================
-#
-# SYNOPSIS
-#
-#   AX_JAVA_CHECK_CLASS(,,)
-#
-# DESCRIPTION
-#
-#   Test if a Java class is available. Based on AX_PROG_JAVAC_WORKS. This
-#   version uses a cache variable which is both compiler, options and
-#   classpath dependent (so if you switch from javac to gcj it correctly
-#   notices and redoes the test).
-#
-#   The macro tries to compile a minimal program importing . Some
-#   newer compilers moan about the failure to use this but fail or produce a
-#   class file anyway. All moaing is sunk to /dev/null since I only wanted
-#   to know if the class could be imported. This is a recommended followup
-#   to AX_CHECK_JAVA_PLUGIN with classpath appropriately adjusted.
-#
-# LICENSE
-#
-#   Copyright (c) 2008 Duncan Simpson 
-#
-#   This program is free software; you can redistribute it and/or modify it
-#   under the terms of the GNU General Public License as published by the
-#   Free Software Foundation; either version 2 of the License, or (at your
-#   option) any later version.
-#
-#   This program is distributed in the hope that it will be useful, but
-#   WITHOUT ANY WARRANTY; without even the implied warranty of
-#   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General
-#   Public License for more details.
-#
-#   You should have received a copy of the GNU General Public License along
-#   with this program. If not, see .
-#
-#   As a special exception, the respective Autoconf Macro's copyright owner
-#   gives unlimited permission to copy, distribute and modify the configure
-#   scripts that are the output of Autoconf when processing the Macro. You
-#   need not follow the terms of the GNU General Public License when using
-#   or distributing such scripts, even though portions of the text of the
-#   Macro appear in them. The GNU General Public License (GPL) does govern
-#   all other use of the material that constitutes the Autoconf Macro.
-#
-#   This special exception to the GPL applies to versions of the Autoconf
-#   Macro released by the Autoconf Archive. When you make and distribute a
-#   modified version of the Autoconf Macro, you may extend this special
-#   exception to the GPL to apply to your modified version as well.
-
-#serial 9
-
-AU_ALIAS([DPS_JAVA_CHECK_CLASS], [AX_JAVA_CHECK_CLASS])
-AC_DEFUN([AX_JAVA_CHECK_CLASS],[
-m4_define([cache_val],[m4_translit(ax_cv_have_java_class_$1, " ." ,"__")])
-if test "x$CLASSPATH" != "x"; then
-xtra=" with classpath ${CLASSPATH}"
-xopts=`echo ${CLASSPATH} | ${SED} 's/^ *://'`
-xopts="-classpath $xopts"
-else xtra=""; xopts=""; fi
-cache_var="cache_val"AS_TR_SH([_Jc_${JAVAC}_Cp_${CLASSPATH}])
-AC_CACHE_CHECK([if the $1 class is available$xtra], [$cache_var], [
-JAVA_TEST=Test.java
-CLASS_TEST=Test.class
-cat << \EOF > $JAVA_TEST
-/* [#]xline __oline__ "configure" */
-import $1;
-public class Test {
-}
-EOF
-if AC_TRY_COMMAND($JAVAC $JAVACFLAGS $xopts $JAVA_TEST) >/dev/null 2>&1; then
-  eval "${cache_var}=yes"
-else
-  eval "${cache_var}=no"
-  echo "configure: failed program was:" >&AS_MESSAGE_LOG_FD
-  cat $JAVA_TEST >&AS_MESSAGE_LOG_FD
-fi
-rm -f $JAVA_TEST $CLASS_TEST
-])
-if eval 'test "x$'${cache_var}'" = "xyes"'; then
-$2
-true; else
-$3
-false; fi])
diff --git a/vpp-api/java/m4/ax_java_options.m4 b/vpp-api/java/m4/ax_java_options.m4
deleted file mode 100644
index 36c10d92..00000000
--- a/vpp-api/java/m4/ax_java_options.m4
+++ /dev/null
@@ -1,48 +0,0 @@
-# ===========================================================================
-#      http://www.gnu.org/software/autoconf-archive/ax_java_options.html
-# ===========================================================================
-#
-# SYNOPSIS
-#
-#   AX_JAVA_OPTIONS
-#
-# DESCRIPTION
-#
-#   AX_JAVA_OPTIONS adds configure command line options used for Java m4
-#   macros. This Macro is optional.
-#
-#   Note: This is part of the set of autoconf M4 macros for Java programs.
-#   It is VERY IMPORTANT that you download the whole set, some macros depend
-#   on other. Unfortunately, the autoconf archive does not support the
-#   concept of set of macros, so I had to break it for submission. The
-#   general documentation, as well as the sample configure.in, is included
-#   in the AX_PROG_JAVA macro.
-#
-# LICENSE
-#
-#   Copyright (c) 2008 Devin Weaver 
-#
-#   Copying and distribution of this file, with or without modification, are
-#   permitted in any medium without royalty provided the copyright notice
-#   and this notice are preserved. This file is offered as-is, without any
-#   warranty.
-
-#serial 6
-
-AU_ALIAS([AC_JAVA_OPTIONS], [AX_JAVA_OPTIONS])
-AC_DEFUN([AX_JAVA_OPTIONS],[
-AC_ARG_WITH(java-prefix,
-                        [  --with-java-prefix=PFX  prefix where Java runtime is installed (optional)])
-AC_ARG_WITH(javac-flags,
-                        [  --with-javac-flags=FLAGS flags to pass to the Java compiler (optional)])
-AC_ARG_WITH(java-flags,
-                        [  --with-java-flags=FLAGS flags to pass to the Java VM (optional)])
-JAVAPREFIX=$with_java_prefix
-JAVACFLAGS=$with_javac_flags
-JAVAFLAGS=$with_java_flags
-AC_SUBST(JAVAPREFIX)dnl
-AC_SUBST(JAVACFLAGS)dnl
-AC_SUBST(JAVAFLAGS)dnl
-AC_SUBST(JAVA)dnl
-AC_SUBST(JAVAC)dnl
-])
diff --git a/vpp-api/java/m4/ax_libgcj_jar.m4 b/vpp-api/java/m4/ax_libgcj_jar.m4
deleted file mode 100644
index 5e942857..00000000
--- a/vpp-api/java/m4/ax_libgcj_jar.m4
+++ /dev/null
@@ -1,83 +0,0 @@
-# ===========================================================================
-#       http://www.gnu.org/software/autoconf-archive/ax_libgcj_jar.html
-# ===========================================================================
-#
-# SYNOPSIS
-#
-#   AX_LIBGCJ_JAR
-#
-# DESCRIPTION
-#
-#   Locate libgcj.jar so you can place it before everything else when using
-#   gcj.
-#
-# LICENSE
-#
-#   Copyright (c) 2008 Duncan Simpson 
-#
-#   This program is free software; you can redistribute it and/or modify it
-#   under the terms of the GNU General Public License as published by the
-#   Free Software Foundation; either version 2 of the License, or (at your
-#   option) any later version.
-#
-#   This program is distributed in the hope that it will be useful, but
-#   WITHOUT ANY WARRANTY; without even the implied warranty of
-#   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General
-#   Public License for more details.
-#
-#   You should have received a copy of the GNU General Public License along
-#   with this program. If not, see .
-#
-#   As a special exception, the respective Autoconf Macro's copyright owner
-#   gives unlimited permission to copy, distribute and modify the configure
-#   scripts that are the output of Autoconf when processing the Macro. You
-#   need not follow the terms of the GNU General Public License when using
-#   or distributing such scripts, even though portions of the text of the
-#   Macro appear in them. The GNU General Public License (GPL) does govern
-#   all other use of the material that constitutes the Autoconf Macro.
-#
-#   This special exception to the GPL applies to versions of the Autoconf
-#   Macro released by the Autoconf Archive. When you make and distribute a
-#   modified version of the Autoconf Macro, you may extend this special
-#   exception to the GPL to apply to your modified version as well.
-
-#serial 9
-
-AU_ALIAS([DPS_LIBGCJ_JAR], [AX_LIBGCJ_JAR])
-AC_DEFUN([AX_LIBGCJ_JAR],
-[
-AC_REQUIRE([AC_EXEEXT])
-AC_REQUIRE([AX_PROG_JAVAC])
-AC_REQUIRE([AC_PROG_FGREP])
-AC_PROG_SED
-if test "x$SED" = "x"; then
-AC_MSG_WARN([sed not avaiable, so libgcj.jar test skipped])
-else
-AC_MSG_CHECKING([if $JAVAC is gcj]);
-jc=`eval "[echo x$JAVAC | $SED 's/^x.*\\/\\([^/]*\\)\$/x\\1/;s/^ *\\([^ ]*\\) .*$/\\1/;s/"$EXEEXT"$//']"`
-if test "x$jc" != "xxgcj"; then
-AC_MSG_RESULT(no)
-else
-AC_MSG_RESULT(yes)
-AC_MSG_CHECKING([libgcj.jar location])
-save_cp="$CLASSPATH";
-unset CLASSPATH;
-AC_MSG_CHECKING([gcj default classpath])
-cat << \EOF > Test.java
-/* [#]line __oline__ "configure" */
-public class Test {
-}
-EOF
-lgcj=`eval "[$JAVAC -v -C Test.java 2>&1 | $FGREP \\(system\\) | $SED 's/^ *\\([^ ]*\\) .*$/\\1/;s/\\.jar\\//.jar/']"`;
-if test -f Test.class && test "x$lgcj" != "x"; then
-AC_MSG_RESULT($lgcj)
-$1="$lgcj:"
-else
-AC_MSG_RESULT(failed)
-$1=""
-fi
-if test "x$save_cp" != "x"; then CLASSPATH="$save_cp"; fi
-rm -f Test.java Test.class
-fi
-fi
-])
diff --git a/vpp-api/java/m4/ax_prog_jar.m4 b/vpp-api/java/m4/ax_prog_jar.m4
deleted file mode 100644
index 3c60fcaf..00000000
--- a/vpp-api/java/m4/ax_prog_jar.m4
+++ /dev/null
@@ -1,49 +0,0 @@
-# ===========================================================================
-#        http://www.gnu.org/software/autoconf-archive/ax_prog_jar.html
-# ===========================================================================
-#
-# SYNOPSIS
-#
-#   AX_PROG_JAR
-#
-# DESCRIPTION
-#
-#   AX_PROG_JAR tests for an existing jar program. It uses the environment
-#   variable JAR then tests in sequence various common jar programs.
-#
-#   If you want to force a specific compiler:
-#
-#   - at the configure.in level, set JAR=yourcompiler before calling
-#   AX_PROG_JAR
-#
-#   - at the configure level, setenv JAR
-#
-#   You can use the JAR variable in your Makefile.in, with @JAR@.
-#
-#   Note: This macro depends on the autoconf M4 macros for Java programs. It
-#   is VERY IMPORTANT that you download that whole set, some macros depend
-#   on other. Unfortunately, the autoconf archive does not support the
-#   concept of set of macros, so I had to break it for submission.
-#
-#   The general documentation of those macros, as well as the sample
-#   configure.in, is included in the AX_PROG_JAVA macro.
-#
-# LICENSE
-#
-#   Copyright (c) 2008 Egon Willighagen 
-#
-#   Copying and distribution of this file, with or without modification, are
-#   permitted in any medium without royalty provided the copyright notice
-#   and this notice are preserved. This file is offered as-is, without any
-#   warranty.
-
-#serial 7
-
-AU_ALIAS([AC_PROG_JAR], [AX_PROG_JAR])
-AC_DEFUN([AX_PROG_JAR],[
-AS_IF([test "x$JAVAPREFIX" = x],
-      [test "x$JAR" = x && AC_CHECK_PROGS([JAR], [jar])],
-      [test "x$JAR" = x && AC_CHECK_PROGS([JAR], [jar], [], [$JAVAPREFIX/bin])])
-test "x$JAR" = x && AC_MSG_ERROR([no acceptable jar program found in \$PATH])
-AC_PROVIDE([$0])dnl
-])
diff --git a/vpp-api/java/m4/ax_prog_java.m4 b/vpp-api/java/m4/ax_prog_java.m4
deleted file mode 100644
index 03961db5..00000000
--- a/vpp-api/java/m4/ax_prog_java.m4
+++ /dev/null
@@ -1,115 +0,0 @@
-# ===========================================================================
-#       http://www.gnu.org/software/autoconf-archive/ax_prog_java.html
-# ===========================================================================
-#
-# SYNOPSIS
-#
-#   AX_PROG_JAVA
-#
-# DESCRIPTION
-#
-#   Here is a summary of the main macros:
-#
-#   AX_PROG_JAVAC: finds a Java compiler.
-#
-#   AX_PROG_JAVA: finds a Java virtual machine.
-#
-#   AX_CHECK_CLASS: finds if we have the given class (beware of CLASSPATH!).
-#
-#   AX_CHECK_RQRD_CLASS: finds if we have the given class and stops
-#   otherwise.
-#
-#   AX_TRY_COMPILE_JAVA: attempt to compile user given source.
-#
-#   AX_TRY_RUN_JAVA: attempt to compile and run user given source.
-#
-#   AX_JAVA_OPTIONS: adds Java configure options.
-#
-#   AX_PROG_JAVA tests an existing Java virtual machine. It uses the
-#   environment variable JAVA then tests in sequence various common Java
-#   virtual machines. For political reasons, it starts with the free ones.
-#   You *must* call [AX_PROG_JAVAC] before.
-#
-#   If you want to force a specific VM:
-#
-#   - at the configure.in level, set JAVA=yourvm before calling AX_PROG_JAVA
-#
-#     (but after AC_INIT)
-#
-#   - at the configure level, setenv JAVA
-#
-#   You can use the JAVA variable in your Makefile.in, with @JAVA@.
-#
-#   *Warning*: its success or failure can depend on a proper setting of the
-#   CLASSPATH env. variable.
-#
-#   TODO: allow to exclude virtual machines (rationale: most Java programs
-#   cannot run with some VM like kaffe).
-#
-#   Note: This is part of the set of autoconf M4 macros for Java programs.
-#   It is VERY IMPORTANT that you download the whole set, some macros depend
-#   on other. Unfortunately, the autoconf archive does not support the
-#   concept of set of macros, so I had to break it for submission.
-#
-#   A Web page, with a link to the latest CVS snapshot is at
-#   .
-#
-#   This is a sample configure.in Process this file with autoconf to produce
-#   a configure script.
-#
-#     AC_INIT(UnTag.java)
-#
-#     dnl Checks for programs.
-#     AC_CHECK_CLASSPATH
-#     AX_PROG_JAVAC
-#     AX_PROG_JAVA
-#
-#     dnl Checks for classes
-#     AX_CHECK_RQRD_CLASS(org.xml.sax.Parser)
-#     AX_CHECK_RQRD_CLASS(com.jclark.xml.sax.Driver)
-#
-#     AC_OUTPUT(Makefile)
-#
-# LICENSE
-#
-#   Copyright (c) 2008 Stephane Bortzmeyer 
-#
-#   This program is free software; you can redistribute it and/or modify it
-#   under the terms of the GNU General Public License as published by the
-#   Free Software Foundation; either version 2 of the License, or (at your
-#   option) any later version.
-#
-#   This program is distributed in the hope that it will be useful, but
-#   WITHOUT ANY WARRANTY; without even the implied warranty of
-#   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General
-#   Public License for more details.
-#
-#   You should have received a copy of the GNU General Public License along
-#   with this program. If not, see .
-#
-#   As a special exception, the respective Autoconf Macro's copyright owner
-#   gives unlimited permission to copy, distribute and modify the configure
-#   scripts that are the output of Autoconf when processing the Macro. You
-#   need not follow the terms of the GNU General Public License when using
-#   or distributing such scripts, even though portions of the text of the
-#   Macro appear in them. The GNU General Public License (GPL) does govern
-#   all other use of the material that constitutes the Autoconf Macro.
-#
-#   This special exception to the GPL applies to versions of the Autoconf
-#   Macro released by the Autoconf Archive. When you make and distribute a
-#   modified version of the Autoconf Macro, you may extend this special
-#   exception to the GPL to apply to your modified version as well.
-
-#serial 9
-
-AU_ALIAS([AC_PROG_JAVA], [AX_PROG_JAVA])
-AC_DEFUN([AX_PROG_JAVA],[
-m4_define([m4_ax_prog_java_list], [kaffe java])dnl
-AS_IF([test "x$JAVAPREFIX" = x],
-      [test x$JAVA = x && AC_CHECK_PROGS([JAVA], [m4_ax_prog_java_list])],
-      [test x$JAVA = x && AC_CHECK_PROGS([JAVA], [m4_ax_prog_java_list], [], [$JAVAPREFIX/bin])])
-test x$JAVA = x && AC_MSG_ERROR([no acceptable Java virtual machine found in \$PATH])
-m4_undefine([m4_ax_prog_java_list])dnl
-AX_PROG_JAVA_WORKS
-AC_PROVIDE([$0])dnl
-])
diff --git a/vpp-api/java/m4/ax_prog_java_cc.m4 b/vpp-api/java/m4/ax_prog_java_cc.m4
deleted file mode 100644
index 3df064ff..00000000
--- a/vpp-api/java/m4/ax_prog_java_cc.m4
+++ /dev/null
@@ -1,104 +0,0 @@
-# ===========================================================================
-#      http://www.gnu.org/software/autoconf-archive/ax_prog_java_cc.html
-# ===========================================================================
-#
-# SYNOPSIS
-#
-#   AX_PROG_JAVA_CC
-#
-# DESCRIPTION
-#
-#   Finds the appropriate java compiler on your path. By preference the java
-#   compiler is gcj, then jikes then javac.
-#
-#   The macro can take one argument specifying a space separated list of
-#   java compiler names.
-#
-#   For example:
-#
-#     AX_PROG_JAVA_CC(javac, gcj)
-#
-#   The macro also sets the compiler options variable: JAVA_CC_OPTS to
-#   something sensible:
-#
-#    - for GCJ it sets it to: @GCJ_OPTS@
-#      (if GCJ_OPTS is not yet defined then it is set to "-C")
-#
-#    - no other compiler has applicable options yet
-#
-#   Here's an example configure.in:
-#
-#     AC_INIT(Makefile.in)
-#     AX_PROG_JAVA_CC()
-#     AC_OUTPUT(Makefile)
-#     dnl End.
-#
-#   And here's the start of the Makefile.in:
-#
-#     PROJECT_ROOT      := @srcdir@
-#     # Tool definitions.
-#     JAVAC             := @JAVA_CC@
-#     JAVAC_OPTS        := @JAVA_CC_OPTS@
-#     JAR_TOOL          := @jar_tool@
-#
-# LICENSE
-#
-#   Copyright (c) 2008 Nic Ferrier 
-#
-#   This program is free software; you can redistribute it and/or modify it
-#   under the terms of the GNU General Public License as published by the
-#   Free Software Foundation; either version 2 of the License, or (at your
-#   option) any later version.
-#
-#   This program is distributed in the hope that it will be useful, but
-#   WITHOUT ANY WARRANTY; without even the implied warranty of
-#   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General
-#   Public License for more details.
-#
-#   You should have received a copy of the GNU General Public License along
-#   with this program. If not, see .
-#
-#   As a special exception, the respective Autoconf Macro's copyright owner
-#   gives unlimited permission to copy, distribute and modify the configure
-#   scripts that are the output of Autoconf when processing the Macro. You
-#   need not follow the terms of the GNU General Public License when using
-#   or distributing such scripts, even though portions of the text of the
-#   Macro appear in them. The GNU General Public License (GPL) does govern
-#   all other use of the material that constitutes the Autoconf Macro.
-#
-#   This special exception to the GPL applies to versions of the Autoconf
-#   Macro released by the Autoconf Archive. When you make and distribute a
-#   modified version of the Autoconf Macro, you may extend this special
-#   exception to the GPL to apply to your modified version as well.
-
-#serial 4
-
-# AX_PROG_JAVA_CC([COMPILER ...])
-# --------------------------
-# COMPILER ... is a space separated list of java compilers to search for.
-# This just gives the user an opportunity to specify an alternative
-# search list for the java compiler.
-AU_ALIAS([AC_PROG_JAVA_CC], [AX_PROG_JAVA_CC])
-AC_DEFUN([AX_PROG_JAVA_CC],
-[AC_ARG_VAR([JAVA_CC],                [java compiler command])dnl
-AC_ARG_VAR([JAVA_CC_FLAGS],           [java compiler flags])dnl
-m4_ifval([$1],
-      [AC_CHECK_TOOLS(JAVA_CC, [$1])],
-[AC_CHECK_TOOL(JAVA_CC, gcj)
-if test -z "$JAVA_CC"; then
-  AC_CHECK_TOOL(JAVA_CC, javac)
-fi
-if test -z "$JAVA_CC"; then
-  AC_CHECK_TOOL(JAVA_CC, jikes)
-fi
-])
-
-if test "$JAVA_CC" = "gcj"; then
-   if test "$GCJ_OPTS" = ""; then
-      AC_SUBST(GCJ_OPTS,-C)
-   fi
-   AC_SUBST(JAVA_CC_OPTS, @GCJ_OPTS@,
-        [Define the compilation options for GCJ])
-fi
-test -z "$JAVA_CC" && AC_MSG_ERROR([no acceptable java compiler found in \$PATH])
-])# AX_PROG_JAVA_CC
diff --git a/vpp-api/java/m4/ax_prog_java_works.m4 b/vpp-api/java/m4/ax_prog_java_works.m4
deleted file mode 100644
index 54e132af..00000000
--- a/vpp-api/java/m4/ax_prog_java_works.m4
+++ /dev/null
@@ -1,134 +0,0 @@
-# ===========================================================================
-#    http://www.gnu.org/software/autoconf-archive/ax_prog_java_works.html
-# ===========================================================================
-#
-# SYNOPSIS
-#
-#   AX_PROG_JAVA_WORKS
-#
-# DESCRIPTION
-#
-#   Internal use ONLY.
-#
-#   Note: This is part of the set of autoconf M4 macros for Java programs.
-#   It is VERY IMPORTANT that you download the whole set, some macros depend
-#   on other. Unfortunately, the autoconf archive does not support the
-#   concept of set of macros, so I had to break it for submission. The
-#   general documentation, as well as the sample configure.in, is included
-#   in the AX_PROG_JAVA macro.
-#
-# LICENSE
-#
-#   Copyright (c) 2008 Stephane Bortzmeyer 
-#
-#   This program is free software; you can redistribute it and/or modify it
-#   under the terms of the GNU General Public License as published by the
-#   Free Software Foundation; either version 2 of the License, or (at your
-#   option) any later version.
-#
-#   This program is distributed in the hope that it will be useful, but
-#   WITHOUT ANY WARRANTY; without even the implied warranty of
-#   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General
-#   Public License for more details.
-#
-#   You should have received a copy of the GNU General Public License along
-#   with this program. If not, see .
-#
-#   As a special exception, the respective Autoconf Macro's copyright owner
-#   gives unlimited permission to copy, distribute and modify the configure
-#   scripts that are the output of Autoconf when processing the Macro. You
-#   need not follow the terms of the GNU General Public License when using
-#   or distributing such scripts, even though portions of the text of the
-#   Macro appear in them. The GNU General Public License (GPL) does govern
-#   all other use of the material that constitutes the Autoconf Macro.
-#
-#   This special exception to the GPL applies to versions of the Autoconf
-#   Macro released by the Autoconf Archive. When you make and distribute a
-#   modified version of the Autoconf Macro, you may extend this special
-#   exception to the GPL to apply to your modified version as well.
-
-#serial 9
-
-AU_ALIAS([AC_PROG_JAVA_WORKS], [AX_PROG_JAVA_WORKS])
-AC_DEFUN([AX_PROG_JAVA_WORKS], [
-AC_PATH_PROG(UUDECODE, uudecode, [no])
-if test x$UUDECODE != xno; then
-AC_CACHE_CHECK([if uudecode can decode base 64 file], ac_cv_prog_uudecode_base64, [
-dnl /**
-dnl  * Test.java: used to test if java compiler works.
-dnl  */
-dnl public class Test
-dnl {
-dnl
-dnl public static void
-dnl main( String[] argv )
-dnl {
-dnl     System.exit (0);
-dnl }
-dnl
-dnl }
-cat << \EOF > Test.uue
-begin-base64 644 Test.class
-yv66vgADAC0AFQcAAgEABFRlc3QHAAQBABBqYXZhL2xhbmcvT2JqZWN0AQAE
-bWFpbgEAFihbTGphdmEvbGFuZy9TdHJpbmc7KVYBAARDb2RlAQAPTGluZU51
-bWJlclRhYmxlDAAKAAsBAARleGl0AQAEKEkpVgoADQAJBwAOAQAQamF2YS9s
-YW5nL1N5c3RlbQEABjxpbml0PgEAAygpVgwADwAQCgADABEBAApTb3VyY2VG
-aWxlAQAJVGVzdC5qYXZhACEAAQADAAAAAAACAAkABQAGAAEABwAAACEAAQAB
-AAAABQO4AAyxAAAAAQAIAAAACgACAAAACgAEAAsAAQAPABAAAQAHAAAAIQAB
-AAEAAAAFKrcAErEAAAABAAgAAAAKAAIAAAAEAAQABAABABMAAAACABQ=
-====
-EOF
-if $UUDECODE Test.uue; then
-        ac_cv_prog_uudecode_base64=yes
-else
-        echo "configure: __oline__: uudecode had trouble decoding base 64 file 'Test.uue'" >&AS_MESSAGE_LOG_FD
-        echo "configure: failed file was:" >&AS_MESSAGE_LOG_FD
-        cat Test.uue >&AS_MESSAGE_LOG_FD
-        ac_cv_prog_uudecode_base64=no
-fi
-rm -f Test.uue])
-fi
-if test x$ac_cv_prog_uudecode_base64 != xyes; then
-        rm -f Test.class
-        AC_MSG_WARN([I have to compile Test.class from scratch])
-        if test x$ac_cv_prog_javac_works = xno; then
-                AC_MSG_ERROR([Cannot compile java source. $JAVAC does not work properly])
-        fi
-        if test x$ac_cv_prog_javac_works = x; then
-                AX_PROG_JAVAC
-        fi
-fi
-AC_CACHE_CHECK(if $JAVA works, ac_cv_prog_java_works, [
-JAVA_TEST=Test.java
-CLASS_TEST=Test.class
-TEST=Test
-changequote(, )dnl
-cat << \EOF > $JAVA_TEST
-/* [#]line __oline__ "configure" */
-public class Test {
-public static void main (String args[]) {
-        System.exit (0);
-} }
-EOF
-changequote([, ])dnl
-if test x$ac_cv_prog_uudecode_base64 != xyes; then
-        if AC_TRY_COMMAND($JAVAC $JAVACFLAGS $JAVA_TEST) && test -s $CLASS_TEST; then
-                :
-        else
-          echo "configure: failed program was:" >&AS_MESSAGE_LOG_FD
-          cat $JAVA_TEST >&AS_MESSAGE_LOG_FD
-          AC_MSG_ERROR(The Java compiler $JAVAC failed (see config.log, check the CLASSPATH?))
-        fi
-fi
-if AC_TRY_COMMAND($JAVA -classpath . $JAVAFLAGS $TEST) >/dev/null 2>&1; then
-  ac_cv_prog_java_works=yes
-else
-  echo "configure: failed program was:" >&AS_MESSAGE_LOG_FD
-  cat $JAVA_TEST >&AS_MESSAGE_LOG_FD
-  AC_MSG_ERROR(The Java VM $JAVA failed (see config.log, check the CLASSPATH?))
-fi
-rm -fr $JAVA_TEST $CLASS_TEST Test.uue
-])
-AC_PROVIDE([$0])dnl
-]
-)
diff --git a/vpp-api/java/m4/ax_prog_javac.m4 b/vpp-api/java/m4/ax_prog_javac.m4
deleted file mode 100644
index d061243c..00000000
--- a/vpp-api/java/m4/ax_prog_javac.m4
+++ /dev/null
@@ -1,79 +0,0 @@
-# ===========================================================================
-#       http://www.gnu.org/software/autoconf-archive/ax_prog_javac.html
-# ===========================================================================
-#
-# SYNOPSIS
-#
-#   AX_PROG_JAVAC
-#
-# DESCRIPTION
-#
-#   AX_PROG_JAVAC tests an existing Java compiler. It uses the environment
-#   variable JAVAC then tests in sequence various common Java compilers. For
-#   political reasons, it starts with the free ones.
-#
-#   If you want to force a specific compiler:
-#
-#   - at the configure.in level, set JAVAC=yourcompiler before calling
-#   AX_PROG_JAVAC
-#
-#   - at the configure level, setenv JAVAC
-#
-#   You can use the JAVAC variable in your Makefile.in, with @JAVAC@.
-#
-#   *Warning*: its success or failure can depend on a proper setting of the
-#   CLASSPATH env. variable.
-#
-#   TODO: allow to exclude compilers (rationale: most Java programs cannot
-#   compile with some compilers like guavac).
-#
-#   Note: This is part of the set of autoconf M4 macros for Java programs.
-#   It is VERY IMPORTANT that you download the whole set, some macros depend
-#   on other. Unfortunately, the autoconf archive does not support the
-#   concept of set of macros, so I had to break it for submission. The
-#   general documentation, as well as the sample configure.in, is included
-#   in the AX_PROG_JAVA macro.
-#
-# LICENSE
-#
-#   Copyright (c) 2008 Stephane Bortzmeyer 
-#
-#   This program is free software; you can redistribute it and/or modify it
-#   under the terms of the GNU General Public License as published by the
-#   Free Software Foundation; either version 2 of the License, or (at your
-#   option) any later version.
-#
-#   This program is distributed in the hope that it will be useful, but
-#   WITHOUT ANY WARRANTY; without even the implied warranty of
-#   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General
-#   Public License for more details.
-#
-#   You should have received a copy of the GNU General Public License along
-#   with this program. If not, see .
-#
-#   As a special exception, the respective Autoconf Macro's copyright owner
-#   gives unlimited permission to copy, distribute and modify the configure
-#   scripts that are the output of Autoconf when processing the Macro. You
-#   need not follow the terms of the GNU General Public License when using
-#   or distributing such scripts, even though portions of the text of the
-#   Macro appear in them. The GNU General Public License (GPL) does govern
-#   all other use of the material that constitutes the Autoconf Macro.
-#
-#   This special exception to the GPL applies to versions of the Autoconf
-#   Macro released by the Autoconf Archive. When you make and distribute a
-#   modified version of the Autoconf Macro, you may extend this special
-#   exception to the GPL to apply to your modified version as well.
-
-#serial 7
-
-AU_ALIAS([AC_PROG_JAVAC], [AX_PROG_JAVAC])
-AC_DEFUN([AX_PROG_JAVAC],[
-m4_define([m4_ax_prog_javac_list],["gcj -C" guavac jikes javac])dnl
-AS_IF([test "x$JAVAPREFIX" = x],
-      [test "x$JAVAC" = x && AC_CHECK_PROGS([JAVAC], [m4_ax_prog_javac_list])],
-      [test "x$JAVAC" = x && AC_CHECK_PROGS([JAVAC], [m4_ax_prog_javac_list], [], [$JAVAPREFIX/bin])])
-m4_undefine([m4_ax_prog_javac_list])dnl
-test "x$JAVAC" = x && AC_MSG_ERROR([no acceptable Java compiler found in \$PATH])
-AX_PROG_JAVAC_WORKS
-AC_PROVIDE([$0])dnl
-])
diff --git a/vpp-api/java/m4/ax_prog_javac_works.m4 b/vpp-api/java/m4/ax_prog_javac_works.m4
deleted file mode 100644
index 7dfa1e37..00000000
--- a/vpp-api/java/m4/ax_prog_javac_works.m4
+++ /dev/null
@@ -1,72 +0,0 @@
-# ===========================================================================
-#    http://www.gnu.org/software/autoconf-archive/ax_prog_javac_works.html
-# ===========================================================================
-#
-# SYNOPSIS
-#
-#   AX_PROG_JAVAC_WORKS
-#
-# DESCRIPTION
-#
-#   Internal use ONLY.
-#
-#   Note: This is part of the set of autoconf M4 macros for Java programs.
-#   It is VERY IMPORTANT that you download the whole set, some macros depend
-#   on other. Unfortunately, the autoconf archive does not support the
-#   concept of set of macros, so I had to break it for submission. The
-#   general documentation, as well as the sample configure.in, is included
-#   in the AX_PROG_JAVA macro.
-#
-# LICENSE
-#
-#   Copyright (c) 2008 Stephane Bortzmeyer 
-#
-#   This program is free software; you can redistribute it and/or modify it
-#   under the terms of the GNU General Public License as published by the
-#   Free Software Foundation; either version 2 of the License, or (at your
-#   option) any later version.
-#
-#   This program is distributed in the hope that it will be useful, but
-#   WITHOUT ANY WARRANTY; without even the implied warranty of
-#   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General
-#   Public License for more details.
-#
-#   You should have received a copy of the GNU General Public License along
-#   with this program. If not, see .
-#
-#   As a special exception, the respective Autoconf Macro's copyright owner
-#   gives unlimited permission to copy, distribute and modify the configure
-#   scripts that are the output of Autoconf when processing the Macro. You
-#   need not follow the terms of the GNU General Public License when using
-#   or distributing such scripts, even though portions of the text of the
-#   Macro appear in them. The GNU General Public License (GPL) does govern
-#   all other use of the material that constitutes the Autoconf Macro.
-#
-#   This special exception to the GPL applies to versions of the Autoconf
-#   Macro released by the Autoconf Archive. When you make and distribute a
-#   modified version of the Autoconf Macro, you may extend this special
-#   exception to the GPL to apply to your modified version as well.
-
-#serial 6
-
-AU_ALIAS([AC_PROG_JAVAC_WORKS], [AX_PROG_JAVAC_WORKS])
-AC_DEFUN([AX_PROG_JAVAC_WORKS],[
-AC_CACHE_CHECK([if $JAVAC works], ac_cv_prog_javac_works, [
-JAVA_TEST=Test.java
-CLASS_TEST=Test.class
-cat << \EOF > $JAVA_TEST
-/* [#]line __oline__ "configure" */
-public class Test {
-}
-EOF
-if AC_TRY_COMMAND($JAVAC $JAVACFLAGS $JAVA_TEST) >/dev/null 2>&1; then
-  ac_cv_prog_javac_works=yes
-else
-  AC_MSG_ERROR([The Java compiler $JAVAC failed (see config.log, check the CLASSPATH?)])
-  echo "configure: failed program was:" >&AS_MESSAGE_LOG_FD
-  cat $JAVA_TEST >&AS_MESSAGE_LOG_FD
-fi
-rm -f $JAVA_TEST $CLASS_TEST
-])
-AC_PROVIDE([$0])dnl
-])
diff --git a/vpp-api/java/m4/ax_prog_javadoc.m4 b/vpp-api/java/m4/ax_prog_javadoc.m4
deleted file mode 100644
index bcb6045a..00000000
--- a/vpp-api/java/m4/ax_prog_javadoc.m4
+++ /dev/null
@@ -1,50 +0,0 @@
-# ===========================================================================
-#      http://www.gnu.org/software/autoconf-archive/ax_prog_javadoc.html
-# ===========================================================================
-#
-# SYNOPSIS
-#
-#   AX_PROG_JAVADOC
-#
-# DESCRIPTION
-#
-#   AX_PROG_JAVADOC tests for an existing javadoc generator. It uses the
-#   environment variable JAVADOC then tests in sequence various common
-#   javadoc generator.
-#
-#   If you want to force a specific compiler:
-#
-#   - at the configure.in level, set JAVADOC=yourgenerator before calling
-#   AX_PROG_JAVADOC
-#
-#   - at the configure level, setenv JAVADOC
-#
-#   You can use the JAVADOC variable in your Makefile.in, with @JAVADOC@.
-#
-#   Note: This macro depends on the autoconf M4 macros for Java programs. It
-#   is VERY IMPORTANT that you download that whole set, some macros depend
-#   on other. Unfortunately, the autoconf archive does not support the
-#   concept of set of macros, so I had to break it for submission.
-#
-#   The general documentation of those macros, as well as the sample
-#   configure.in, is included in the AX_PROG_JAVA macro.
-#
-# LICENSE
-#
-#   Copyright (c) 2008 Egon Willighagen 
-#
-#   Copying and distribution of this file, with or without modification, are
-#   permitted in any medium without royalty provided the copyright notice
-#   and this notice are preserved. This file is offered as-is, without any
-#   warranty.
-
-#serial 8
-
-AU_ALIAS([AC_PROG_JAVADOC], [AX_PROG_JAVADOC])
-AC_DEFUN([AX_PROG_JAVADOC],[
-AS_IF([test "x$JAVAPREFIX" = x],
-      [test "x$JAVADOC" = x && AC_CHECK_PROGS([JAVADOC], [javadoc])],
-      [test "x$JAVADOC" = x && AC_CHECK_PROGS([JAVADOC], [javadoc], [], [$JAVAPREFIX/bin])])
-test "x$JAVADOC" = x && AC_MSG_ERROR([no acceptable javadoc generator found in \$PATH])
-AC_PROVIDE([$0])dnl
-])
diff --git a/vpp-api/java/m4/ax_prog_javah.m4 b/vpp-api/java/m4/ax_prog_javah.m4
deleted file mode 100644
index cefc616d..00000000
--- a/vpp-api/java/m4/ax_prog_javah.m4
+++ /dev/null
@@ -1,64 +0,0 @@
-# ===========================================================================
-#       http://www.gnu.org/software/autoconf-archive/ax_prog_javah.html
-# ===========================================================================
-#
-# SYNOPSIS
-#
-#   AX_PROG_JAVAH
-#
-# DESCRIPTION
-#
-#   AX_PROG_JAVAH tests the availability of the javah header generator and
-#   looks for the jni.h header file. If available, JAVAH is set to the full
-#   path of javah and CPPFLAGS is updated accordingly.
-#
-# LICENSE
-#
-#   Copyright (c) 2008 Luc Maisonobe 
-#
-#   Copying and distribution of this file, with or without modification, are
-#   permitted in any medium without royalty provided the copyright notice
-#   and this notice are preserved. This file is offered as-is, without any
-#   warranty.
-
-#serial 8
-
-AU_ALIAS([AC_PROG_JAVAH], [AX_PROG_JAVAH])
-AC_DEFUN([AX_PROG_JAVAH],[
-AC_REQUIRE([AC_CANONICAL_BUILD])dnl
-AC_REQUIRE([AC_PROG_CPP])dnl
-AC_PATH_PROG(JAVAH,javah)
-AS_IF([test -n "$ac_cv_path_JAVAH"],
-      [
-        AC_TRY_CPP([#include ],,[
-        ac_save_CPPFLAGS="$CPPFLAGS"
-		_ACJAVAH_FOLLOW_SYMLINKS("$ac_cv_path_JAVAH")
-        ax_prog_javah_bin_dir=`AS_DIRNAME([$_ACJAVAH_FOLLOWED])`
-        ac_dir="`AS_DIRNAME([$ax_prog_javah_bin_dir])`/include"
-        AS_CASE([$build_os],
-                [cygwin*],
-                [ac_machdep=win32],
-                [ac_machdep=`AS_ECHO($build_os) | sed 's,[[-0-9]].*,,'`])
-        CPPFLAGS="$ac_save_CPPFLAGS -I$ac_dir -I$ac_dir/$ac_machdep"
-        AC_TRY_CPP([#include ],
-                   ac_save_CPPFLAGS="$CPPFLAGS",
-                   AC_MSG_WARN([unable to include ]))
-        CPPFLAGS="$ac_save_CPPFLAGS"])
-      ])
-])
-
-AC_DEFUN([_ACJAVAH_FOLLOW_SYMLINKS],[
-# find the include directory relative to the javac executable
-_cur="$1"
-while ls -ld "$_cur" 2>/dev/null | grep " -> " >/dev/null; do
-        AC_MSG_CHECKING([symlink for $_cur])
-        _slink=`ls -ld "$_cur" | sed 's/.* -> //'`
-        case "$_slink" in
-        /*) _cur="$_slink";;
-        # 'X' avoids triggering unwanted echo options.
-        *) _cur=`echo "X$_cur" | sed -e 's/^X//' -e 's:[[^/]]*$::'`"$_slink";;
-        esac
-        AC_MSG_RESULT([$_cur])
-done
-_ACJAVAH_FOLLOWED="$_cur"
-])
diff --git a/vpp-api/java/m4/ax_try_compile_java.m4 b/vpp-api/java/m4/ax_try_compile_java.m4
deleted file mode 100644
index a8ed6b2a..00000000
--- a/vpp-api/java/m4/ax_try_compile_java.m4
+++ /dev/null
@@ -1,55 +0,0 @@
-# ===========================================================================
-#    http://www.gnu.org/software/autoconf-archive/ax_try_compile_java.html
-# ===========================================================================
-#
-# SYNOPSIS
-#
-#   AX_TRY_COMPILE_JAVA
-#
-# DESCRIPTION
-#
-#   AX_TRY_COMPILE_JAVA attempt to compile user given source.
-#
-#   *Warning*: its success or failure can depend on a proper setting of the
-#   CLASSPATH env. variable.
-#
-#   Note: This is part of the set of autoconf M4 macros for Java programs.
-#   It is VERY IMPORTANT that you download the whole set, some macros depend
-#   on other. Unfortunately, the autoconf archive does not support the
-#   concept of set of macros, so I had to break it for submission. The
-#   general documentation, as well as the sample configure.in, is included
-#   in the AX_PROG_JAVA macro.
-#
-# LICENSE
-#
-#   Copyright (c) 2008 Devin Weaver 
-#
-#   Copying and distribution of this file, with or without modification, are
-#   permitted in any medium without royalty provided the copyright notice
-#   and this notice are preserved. This file is offered as-is, without any
-#   warranty.
-
-#serial 8
-
-AU_ALIAS([AC_TRY_COMPILE_JAVA], [AX_TRY_COMPILE_JAVA])
-AC_DEFUN([AX_TRY_COMPILE_JAVA],[
-AC_REQUIRE([AX_PROG_JAVAC])dnl
-cat << \EOF > Test.java
-/* [#]line __oline__ "configure" */
-ifelse([$1], , , [import $1;])
-public class Test {
-[$2]
-}
-EOF
-if AC_TRY_COMMAND($JAVAC $JAVACFLAGS Test.java) && test -s Test.class
-then
-dnl Don't remove the temporary files here, so they can be examined.
-  ifelse([$3], , :, [$3])
-else
-  echo "configure: failed program was:" >&AS_MESSAGE_LOG_FD
-  cat Test.java >&AS_MESSAGE_LOG_FD
-ifelse([$4], , , [  rm -fr Test.java Test.class
-  $4
-])dnl
-fi
-rm -fr Test.java Test.class])
diff --git a/vpp-api/java/m4/ax_try_run_java.m4 b/vpp-api/java/m4/ax_try_run_java.m4
deleted file mode 100644
index c680f03f..00000000
--- a/vpp-api/java/m4/ax_try_run_java.m4
+++ /dev/null
@@ -1,56 +0,0 @@
-# ===========================================================================
-#      http://www.gnu.org/software/autoconf-archive/ax_try_run_java.html
-# ===========================================================================
-#
-# SYNOPSIS
-#
-#   AX_TRY_RUN_JAVA
-#
-# DESCRIPTION
-#
-#   AX_TRY_RUN_JAVA attempt to compile and run user given source.
-#
-#   *Warning*: its success or failure can depend on a proper setting of the
-#   CLASSPATH env. variable.
-#
-#   Note: This is part of the set of autoconf M4 macros for Java programs.
-#   It is VERY IMPORTANT that you download the whole set, some macros depend
-#   on other. Unfortunately, the autoconf archive does not support the
-#   concept of set of macros, so I had to break it for submission. The
-#   general documentation, as well as the sample configure.in, is included
-#   in the AX_PROG_JAVA macro.
-#
-# LICENSE
-#
-#   Copyright (c) 2008 Devin Weaver 
-#
-#   Copying and distribution of this file, with or without modification, are
-#   permitted in any medium without royalty provided the copyright notice
-#   and this notice are preserved. This file is offered as-is, without any
-#   warranty.
-
-#serial 2
-
-AU_ALIAS([AC_TRY_RUN_JAVA], [AX_TRY_RUN_JAVA])
-AC_DEFUN([AX_TRY_RUN_JAVA],[
-AC_REQUIRE([AX_PROG_JAVAC])dnl
-AC_REQUIRE([AX_PROG_JAVA])dnl
-cat << \EOF > Test.java
-/* [#]line __oline__ "configure" */
-ifelse([$1], , , [include $1;])
-public class Test {
-[$2]
-}
-EOF
-if AC_TRY_COMMAND($JAVAC $JAVACFLAGS Test.java) && test -s Test.class && ($JAVA $JAVAFLAGS Test; exit) 2>/dev/null
-then
-dnl Don't remove the temporary files here, so they can be examined.
-  ifelse([$3], , :, [$3])
-else
-  echo "configure: failed program was:" >&AS_MESSAGE_LOG_FD
-  cat Test.java >&AS_MESSAGE_LOG_FD
-ifelse([$4], , , [  rm -fr Test.java Test.class
-  $4
-])dnl
-fi
-rm -fr Test.java Test.class])
diff --git a/vpp-api/lua/README.md b/vpp-api/lua/README.md
deleted file mode 100644
index 4ecdb34d..00000000
--- a/vpp-api/lua/README.md
+++ /dev/null
@@ -1,50 +0,0 @@
-This is the experimental version of Lua API, aimed for the luajit use.
-
-Please take a look and send the feedback to ayourtch@gmail.com.
-
-To run the examples here:
-
-1) install luajit - "sudo apt-get install luajit" on ubuntu
-
-2) "make build-vpp-api" in the top VPP directory
-
-3) "make run" in a separate terminal window
-   This ensures you have an instance of VPP running
-
-4) sudo luajit examples/example-cli.lua
-
-This will result in something like this:
-
-Version:
-00000000  00 00 00 00 00 00 00 00  00 00 00 00 00 00 00 00  ................
-00000010  00 00 00 00 00 00 00 00  00 00 00 00 00 00 00 00  ................
-
-{ [1] = { ["luaapi_message_name"] = show_version_reply,["program"] = vpe,["version"] = ,["build_date"] = Fri Nov 25 10:58:48 UTC 2016,["retval"] = 0,["build_directory"] = /home/ubuntu/vpp,["_vl_msg_id"] = 170,["context"] = 0,} ,}
----
-{ [1] = { ["luaapi_message_name"] = cli_inband_reply,["_vl_msg_id"] = 94,["length"] = 66,["reply"] = vpp v built by ubuntu on vpp-toys at Fri Nov 25 10:58:48 UTC 2016
-,["retval"] = 0,["context"] = 0,} ,}
----
-
-5) You can also run the performance test bench:
-
-$ sudo luajit bench.lua
-10001 iterations, average speed 5624LL per second
-10001 iterations, average speed 6650LL per second
-10001 iterations, average speed 6053LL per second
-10001 iterations, average speed 7056LL per second
-10001 iterations, average speed 6388LL per second
-10001 iterations, average speed 5849LL per second
-10001 iterations, average speed 6321LL per second
-10001 iterations, average speed 6368LL per second
-10001 iterations, average speed 5958LL per second
-10001 iterations, average speed 6482LL per second
-Average tps across the tests: 6274LL
-
-Note: the above is run in an lxd container running inside 2-core
-xhyve VM on a Macbook Pro, so I would not take the performance numbers for granted :)
-
-The "examples" directory contains a few naive examples, as well as a couple of more 
-advanced ones - a tab-completing CLI for VPP that can call both the APIs and CLI,
-and also a small test utility which I use for automating some small tests using
-VPP.
-
diff --git a/vpp-api/lua/bench.lua b/vpp-api/lua/bench.lua
deleted file mode 100644
index 8e5a0b4b..00000000
--- a/vpp-api/lua/bench.lua
+++ /dev/null
@@ -1,70 +0,0 @@
---[[
-/*
- * Copyright (c) 2016 Cisco and/or its affiliates.
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at:
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-]]
-
-local vpp = require "vpp-lapi"
-
-local ffi = require "ffi"
-
-ffi.cdef([[
-   struct timespec {
-               long tv_sec;        /* seconds */
-               long tv_nsec;       /* nanoseconds */
-           };
-
-   int clock_gettime(int clk_id, struct timespec *tp);
-]])
-
-
-local time_cache = ffi.new("struct timespec[1]")
-local time_cache_1 = time_cache[0]
-function get_ns()
-  ffi.C.clock_gettime(0, time_cache)
-  return time_cache_1.tv_nsec + 1000000000 * time_cache_1.tv_sec
-end
-
-function do_bench()
-  local cycle_start = get_ns()
-  local n_iterations =  10000
-  local count = 1
-  for i = 1,n_iterations do
-    -- print(i)
-    vpp:api_call("show_version")
-    count = count + 1
-    -- print(i, "done")
-  end
-  cycle_end = get_ns()
-  local tps = n_iterations*1000000000LL/(cycle_end - cycle_start)
-  print (tostring(count) .. " iterations, average speed " .. tostring(tps) .. " per second")
-  return tps
-end
-
-root_dir = "/home/ubuntu/vpp"
-pneum_path = root_dir .. "/build-root/install-vpp_lite_debug-native/vpp-api/lib64/libpneum.so"
-vpp:init({ pneum_path = pneum_path })
-vpp:json_api(root_dir .. "/build-root/install-vpp_lite_debug-native/vpp/vpp-api/vpe.api.json")
-
-vpp:connect("lua-bench")
-local n_tests = 10
-local tps_acc = 0LL
-for i=1,n_tests do
-  tps_acc = tps_acc + do_bench()
-end
-print("Average tps across the tests: " .. tostring(tps_acc/n_tests))
-
-vpp:disconnect()
-
-
diff --git a/vpp-api/lua/examples/cli/README.md b/vpp-api/lua/examples/cli/README.md
deleted file mode 100644
index 3a5f8ee9..00000000
--- a/vpp-api/lua/examples/cli/README.md
+++ /dev/null
@@ -1,5 +0,0 @@
-This is a small experiment to have a wrapper CLI which can call both API functions as well as debug CLI.
-
-To facilitate tab completion and help, the API call names are broken up with spaces replacing the underscores.
-
-
diff --git a/vpp-api/lua/examples/cli/lua-cli.lua b/vpp-api/lua/examples/cli/lua-cli.lua
deleted file mode 100644
index b3a24d7d..00000000
--- a/vpp-api/lua/examples/cli/lua-cli.lua
+++ /dev/null
@@ -1,747 +0,0 @@
---[[
-/*
- * Copyright (c) 2016 Cisco and/or its affiliates.
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at:
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-]]
-
--- Experimental prototype CLI using API to VPP, with tab completion
---
--- Written by Andrew Yourtchenko (ayourtch@cisco.com) 2010,2016
---
-
-vpp = require "vpp-lapi"
-
-
-local dotdotdot = "..."
-
--- First the "readline" routine
-
-readln = {
-split = function(str, pat)
-  local t = {}  -- NOTE: use {n = 0} in Lua-5.0
-  local fpat = "(.-)" .. pat
-  local last_end = 1
-  if str then
-    local s, e, cap = str:find(fpat, 1)
-    while s do
-      if s ~= 1 or cap ~= "" then
-        table.insert(t,cap)
-      end
-      last_end = e+1
-      s, e, cap = str:find(fpat, last_end)
-    end
-    if last_end <= #str then
-      cap = str:sub(last_end)
-      table.insert(t, cap)
-    end
-  end
-  return t
-end,
-
-reader = function()
-  local rl = {}
-
-  rl.init = function()
-    os.execute("stty -icanon min 1 -echo")
-    rl.rawmode = true
-  end
-
-  rl.done = function()
-    os.execute("stty icanon echo")
-    rl.rawmode = false
-  end
-
-  rl.prompt = ">"
-  rl.history = { "" }
-  rl.history_index = 1
-  rl.history_length = 1
-
-  rl.hide_cmd = function()
-    local bs = string.char(8) .. " " .. string.char(8)
-    for i = 1, #rl.command do
-      io.stdout:write(bs)
-    end
-  end
-
-  rl.show_cmd = function()
-    if rl.command then
-      io.stdout:write(rl.command)
-    end
-  end
-
-  rl.store_history = function(cmd)
-    if cmd == "" then
-      return
-    end
-    rl.history[rl.history_length] = cmd
-    rl.history_length = rl.history_length + 1
-    rl.history_index = rl.history_length
-    rl.history[rl.history_length] = ""
-  end
-
-  rl.readln = function()
-    local done = false
-    local need_prompt = true
-    rl.command = ""
-
-    if not rl.rawmode then
-      rl.init()
-    end
-
-    while not done do
-      if need_prompt then
-        io.stdout:write(rl.prompt)
-	io.stdout:write(rl.command)
-        need_prompt = false
-      end
-
-      local ch = io.stdin:read(1)
-      if ch:byte(1) == 27 then
-        -- CONTROL
-        local ch2 = io.stdin:read(1)
-        -- arrows
-        if ch2:byte(1) == 91 then
-          local ch3 = io.stdin:read(1)
-          local b = ch3:byte(1)
-          if b == 65 then
-            ch = "UP"
-          elseif b == 66 then
-            ch = "DOWN"
-          elseif b == 67 then
-            ch = "RIGHT"
-          elseif b == 68 then
-            ch = "LEFT"
-          end
-          -- print("Byte: " .. ch3:byte(1))
-          -- if ch3:byte(1)
-        end
-      end
-
-      if ch == "?" then
-        io.stdout:write(ch)
-        io.stdout:write("\n")
-        if rl.help then
-          rl.help(rl)
-        end
-        need_prompt = true
-      elseif ch == "\t" then
-        if rl.tab_complete then
-          rl.tab_complete(rl)
-        end
-        io.stdout:write("\n")
-        need_prompt = true
-      elseif ch == "\n" then
-        io.stdout:write(ch)
-        done = true
-      elseif ch == "\004" then
-        io.stdout:write("\n")
-        rl.command = nil
-	done = true
-      elseif ch == string.char(127) then
-        if rl.command ~= "" then
-          io.stdout:write(string.char(8) .. " " .. string.char(8))
-          rl.command = string.sub(rl.command, 1, -2)
-        end
-      elseif #ch > 1 then
-        -- control char
-        if ch == "UP" then
-          rl.hide_cmd()
-          if rl.history_index == #rl.history then
-            rl.history[rl.history_index] = rl.command
-          end
-          if rl.history_index > 1 then
-            rl.history_index = rl.history_index - 1
-            rl.command = rl.history[rl.history_index]
-          end
-          rl.show_cmd()
-        elseif ch == "DOWN" then
-          rl.hide_cmd()
-          if rl.history_index < rl.history_length then
-            rl.history_index = rl.history_index + 1
-            rl.command = rl.history[rl.history_index]
-          end
-          rl.show_cmd()
-        end
-      else
-        io.stdout:write(ch)
-        rl.command = rl.command .. ch
-      end
-    end
-    if rl.command then
-      rl.store_history(rl.command)
-    end
-    return rl.command
-  end
-  return rl
-end
-
-}
-
---[[
-
-r = reader()
-
-local done = false
-
-while not done do
-  local cmd = r.readln()
-  print("Command: " .. tostring(cmd))
-  if not cmd or cmd == "quit" then
-    done = true
-  end
-end
-
-r.done()
-
-]]
-
---------- MDS show tech parser
-
-local print_section = nil
-local list_sections = false
-
-local curr_section = "---"
-local curr_parser = nil
-
--- by default operate in batch mode
-local batch_mode = true
-
-local db = {}
-local device = {}
-device.output = {}
-local seen_section = {}
-
-function start_collection(name)
-  device = {}
-  seen_section = {}
-end
-
-function print_error(errmsg)
-  print("@#$:" .. errmsg)
-end
-
-function keys(tbl)
-  local t = {}
-  for k, v in pairs(tbl) do
-    table.insert(t, k)
-  end
-  return t
-end
-
-function tset (parent, ...)
-
-  -- print ('set', ...)
-
-  local len = select ('#', ...)
-  local key, value = select (len-1, ...)
-  local cutpoint, cutkey
-
-  for i=1,len-2 do
-
-    local key = select (i, ...)
-    local child = parent[key]
-
-    if value == nil then
-      if child == nil then  return
-      elseif next (child, next (child)) then  cutpoint = nil  cutkey = nil
-      elseif cutpoint == nil then  cutpoint = parent  cutkey = key  end
-
-    elseif child == nil then  child = {}  parent[key] = child  end
-
-    parent = child
-    end
-
-  if value == nil and cutpoint then  cutpoint[cutkey] = nil
-  else  parent[key] = value  return value  end
-  end
-
-
-function tget (parent, ...)
-  local len = select ('#', ...)
-  for i=1,len do
-    parent = parent[select (i, ...)]
-    if parent == nil then  break  end
-    end
-  return parent
-  end
-
-
-local pager_lines = 23
-local pager_printed = 0
-local pager_skipping = false
-local pager_filter_pipe = nil
-
-function pager_reset()
-  pager_printed = 0
-  pager_skipping = false
-  if pager_filter_pipe then
-    pager_filter_pipe:close()
-    pager_filter_pipe = nil
-  end
-end
-
-
-function print_more()
-  io.stdout:write(" --More-- ")
-end
-
-function print_nomore()
-  local bs = string.char(8)
-  local bs10 = bs ..  bs ..  bs ..  bs ..  bs ..  bs ..  bs ..  bs ..  bs ..  bs
-  io.stdout:write(bs10 .. "          " .. bs10)
-end
-
-function print_line(txt)
-  if pager_filter_pipe then
-    pager_filter_pipe:write(txt .. "\n")
-    return
-  end
-  if pager_printed >= pager_lines then
-    print_more()
-    local ch = io.stdin:read(1)
-    if ch == " " then
-      pager_printed = 0
-    elseif ch == "\n" then
-      pager_printed = pager_printed - 1
-    elseif ch == "q" then
-      pager_printed = 0
-      pager_skipping = true
-    end
-    print_nomore()
-  end
-  if not pager_skipping then
-    print(txt)
-    pager_printed = pager_printed + 1
-  else
-    -- skip printing
-  end
-end
-
-function paged_write(text)
-  local t = readln.split(text, "[\n]")
-  if string.sub(text, -1) == "\n" then
-    table.insert(t, "")
-  end
-  for i, v in ipairs(t) do
-    if i < #t then
-      print_line(v)
-    else
-      if pager_filter_pipe then
-        pager_filter_pipe:write(v)
-      else
-        io.stdout:write(v)
-      end
-    end
-  end
-end
-
-
-
-
-
-function get_choices(tbl, key)
-  local res = {}
-  for k, v in pairs(tbl) do
-    if string.sub(k, 1, #key) == key then
-      table.insert(res, k)
-    elseif 0 < #key and dotdotdot == k then
-      table.insert(res, k)
-    end
-  end
-  return res
-end
-
-function get_exact_choice(choices, val)
-  local exact_idx = nil
-  local substr_idx = nil
-  local substr_seen = false
-
-  if #choices == 1 then
-    if choices[1] == dotdotdot then
-      return 1
-    elseif string.sub(choices[1], 1, #val) == val then
-      return 1
-    else
-      return nil
-    end
-  else
-    for i, v in ipairs(choices) do
-      if v == val then
-        exact_idx = i
-        substr_seen = true
-      elseif choices[i] ~= dotdotdot and string.sub(choices[i], 1, #val) == val then
-        if substr_seen then
-          substr_idx = nil
-        else
-          substr_idx = i
-          substr_seen = true
-        end
-      elseif choices[i] == dotdotdot then
-        if substr_seen then
-          substr_idx = nil
-        else
-          substr_idx = i
-          substr_seen = true
-        end
-      end
-    end
-  end
-  return exact_idx or substr_idx
-end
-
-function device_cli_help(rl)
-  local key = readln.split(rl.command, "[ ]+")
-  local tree = rl.tree
-  local keylen = #key
-  local fullcmd = ""
-  local error = false
-  local terse = true
-
-  if ((#rl.command >= 1) and (string.sub(rl.command, -1) == " ")) or (#rl.command == 0) then
-    table.insert(key, "")
-    terse = false
-  end
-
-  for i, v in ipairs(key) do
-    local choices = get_choices(tree, v)
-    local idx = get_exact_choice(choices, v)
-    if idx then
-      local choice = choices[idx]
-      tree = tree[choice]
-      fullcmd = fullcmd .. choice .. " "
-    else
-      if i < #key then
-        error = true
-      end
-    end
-
-    if i == #key and not error then
-      for j, w in ipairs(choices) do
-        if terse then
-          paged_write(w .. "\t")
-        else
-          paged_write("  " .. w .. "\n")
-        end
-      end
-      paged_write("\n")
-      if terse then
-        paged_write(" \n")
-      end
-    end
-  end
-  pager_reset()
-end
-
-function device_cli_tab_complete(rl)
-  local key = readln.split(rl.command, "[ ]+")
-  local tree = rl.tree
-  local keylen = #key
-  local fullcmd = ""
-  local error = false
-
-  for i, v in ipairs(key) do
-    local choices = get_choices(tree, v)
-    local idx = get_exact_choice(choices, v)
-    if idx and choices[idx] ~= dotdotdot then
-      local choice = choices[idx]
-      tree = tree[choice]
-      -- print("level " .. i .. " '" .. choice .. "'")
-      fullcmd = fullcmd .. choice .. " "
-    else
-      -- print("level " .. i .. " : " .. table.concat(choices, " ") .. " ")
-      error = true
-    end
-  end
-  if not error then
-    rl.command = fullcmd
-  else
-    -- print("\n\nerror\n")
-  end
-  pager_reset()
-end
-
-function device_cli_exec(rl)
-
-  local cmd_nopipe = rl.command
-  local cmd_pipe = nil
-
-  local pipe1, pipe2 = string.find(rl.command, "[|]")
-  if pipe1 then
-    cmd_nopipe = string.sub(rl.command, 1, pipe1-1)
-    cmd_pipe = string.sub(rl.command, pipe2+1, -1)
-  end
-
-  local key = readln.split(cmd_nopipe .. " ", "[ ]+")
-  local tree = rl.tree
-  local keylen = #key
-  local fullcmd = ""
-  local error = false
-  local func = nil
-
-  if cmd_pipe then
-    pager_filter_pipe = io.popen(cmd_pipe, "w")
-  end
-
-
-  rl.choices = {}
-
-  for i, v in ipairs(key) do
-    local choices = get_choices(tree, v)
-    local idx = get_exact_choice(choices, v)
-    if idx then
-      local choice = choices[idx]
-      if i == #key then
-        func = tree[choice]
-      else
-        if choice == dotdotdot then
-          -- keep the tree the same, update the choice value to match the input string
-          choices[idx] = v
-          choice = v
-        else
-          tree = tree[choice]
-        end
-      end
-      -- print("level " .. i .. " '" .. choice .. "'")
-      table.insert(rl.choices, choice)
-    else
-      -- print("level " .. i .. " : " .. table.concat(choices, " ") .. " ")
-      error = true
-      return nil
-    end
-  end
-  return func
-end
-
-function populate_tree(commands)
-  local tree = {}
-
-  for k, v in pairs(commands) do
-    local key = readln.split(k .. " ", "[ ]+")
-    local xtree = tree
-    for i, kk in ipairs(key) do
-      if i == 1 and kk == "sh" then
-        kk = "show"
-      end
-      if i == #key then
-        if type(v) == "function" then
-          xtree[kk] = v
-        else
-          xtree[kk] = function(rl) paged_write(table.concat(v, "\n") .. "\n") end
-        end
-      else
-        if not xtree[kk] then
-          xtree[kk] = {}
-        end
-        xtree = xtree[kk]
-      end
-    end
-  end
-  return tree
-end
-
-function trim (s)
-  return (string.gsub(s, "^%s*(.-)%s*$", "%1"))
-end
-
-
-function init_vpp(vpp)
-  local root_dir = "/home/ubuntu/vpp"
-  local pneum_path = root_dir .. "/build-root/install-vpp_lite_debug-native/vpp-api/lib64/libpneum.so"
-
-  vpp:init({ pneum_path = pneum_path })
-
-  vpp:init({ pneum_path = pneum_path })
-  vpp:json_api(root_dir .. "/build-root/install-vpp_lite_debug-native/vpp/vpp-api/vpe.api.json")
-
-
-
-  vpp:connect("lua_cli")
-end
-
-function run_cli(vpp, cli)
-  local reply = vpp:api_call("cli_inband", { cmd = cli })
-  if reply and #reply == 1 then
-    local rep = reply[1]
-    if 0 == rep.retval then
-      return rep.reply
-    else
-      return "XXXXXLUACLI: API RETVAL ERROR : " .. tostring(rep.retval)
-    end
-  else
-    return "XXXXXLUACLI ERROR, RAW REPLY: " .. vpp.dump(reply)
-  end
-end
-
-
-function toprintablestring(s)
-  if type(s) == "string" then
-    return "\n"..vpp.hex_dump(s)
-  else
-    return tostring(s)
-  end
-end
-
-function interactive_cli(r)
-  while not done do
-    pager_reset()
-    local cmd = r.readln()
-    if not cmd then
-      done = true
-    elseif cmd == "quit" or cmd == "exit" then
-      done = true
-    else
-      local func = device_cli_exec(r)
-      if func then
-	func(r)
-      else
-	if trim(cmd) == "" then
-	else
-	  for i = 1, #r.prompt do
-	    paged_write(" ")
-	  end
-	  paged_write("^\n% Invalid input detected at '^' marker.\n\n")
-	end
-      end
-    end
-  end
-end
-
-device = {}
-device.output = {}
-
-init_vpp(vpp)
-cmds_str = run_cli(vpp, "?")
-vpp_cmds = readln.split(cmds_str, "\n")
-vpp_clis = {}
-
-for linenum, line in ipairs(vpp_cmds) do
-  local m,h = string.match(line, "^  (.-)  (.*)$")
-  if m and #m > 0 then
-    table.insert(vpp_clis, m)
-    device.output["vpp debug cli " .. m] = function(rl)
-      -- print("ARBITRARY CLI" .. vpp.dump(rl.choices))
-      print("LUACLI command: " .. table.concat(rl.choices, " "))
-      local sub = {}
-      --
-      for i=4, #rl.choices -1 do
-        table.insert(sub, rl.choices[i])
-      end
-      local cli = table.concat(sub, " ")
-      print("Running CLI: " .. tostring(cli))
-      paged_write(run_cli(vpp, cli))
-    end
-    device.output["vpp debug cli " .. m .. " " .. dotdotdot] = function(rl)
-      print("ARGH")
-    end
-
-    local ret = run_cli(vpp, "help " .. m)
-    device.output["help vpp debug cli " .. m] = { ret }
-  end
-end
-
-for linenum, line in ipairs(vpp_clis) do
-  -- print(line, ret)
-end
-
-for msgnum, msgname in pairs(vpp.msg_number_to_name) do
-  local cli, numspaces = string.gsub(msgname, "_", " ")
-  device.output["call " .. cli .. " " .. dotdotdot] = function(rl)
-    print("ARGH")
-  end
-  device.output["call " .. cli] = function(rl)
-    print("LUACLI command: " .. table.concat(rl.choices, " "))
-    print("Running API: " .. msgname) -- vpp.dump(rl.choices))
-    local out = {}
-    local args = {}
-    local ntaken = 0
-    local argname = ""
-    for i=(1+1+numspaces+1), #rl.choices-1 do
-      -- print(i, rl.choices[i])
-      if ntaken > 0 then
-        ntaken = ntaken -1
-      else
-        local fieldname = rl.choices[i]
-        local field = vpp.msg_name_to_fields[msgname][fieldname]
-        if field then
-          local s = rl.choices[i+1]
-          s=s:gsub("\\x(%x%x)",function (x) return string.char(tonumber(x,16)) end)
-          args[fieldname] = s
-          ntaken = 1
-        end
-      end
-    end
-    -- print("ARGS: ", vpp.dump(args))
-    local ret = vpp:api_call(msgname, args)
-    for i, reply in ipairs(ret) do
-      table.insert(out, "=================== Entry #" .. tostring(i))
-      for k, v in pairs(reply) do
-        table.insert(out, "   " .. tostring(k) .. " : " .. toprintablestring(v))
-      end
-    end
-    -- paged_write(vpp.dump(ret) .. "\n\n")
-    paged_write(table.concat(out, "\n").."\n\n")
-  end
-  device.output["call " .. cli .. " help"] = function(rl)
-    local out = {}
-    for k, v in pairs(vpp.msg_name_to_fields[msgname]) do
-      table.insert(out, tostring(k) .. " : " .. v["ctype"] .. " ; " .. tostring(vpp.dump(v)) )
-    end
-    -- paged_write(vpp.dump(vpp.msg_name_to_fields[msgname]) .. "\n\n")
-    paged_write(table.concat(out, "\n").."\n\n")
-  end
--- vpp.msg_name_to_number = {}
-end
-
-
-
-local r = readln.reader()
-local done = false
-
-r.prompt = "VPP(luaCLI)#"
-
-r.help = device_cli_help
-r.tab_complete = device_cli_tab_complete
-print("===== CLI view, use ^D to end =====")
-
-r.tree = populate_tree(device.output)
--- readln.pretty("xxxx", r.tree)
-
-
-for idx, an_arg in ipairs(arg) do
-  local fname = an_arg
-  if fname == "-i" then
-    pager_lines = 23
-    interactive_cli(r)
-  else
-    pager_lines = 100000000
-    for line in io.lines(fname) do
-      r.command = line
-      local func = device_cli_exec(r)
-      if func then
-	func(r)
-      end
-    end
-  end
-end
-
-if #arg == 0 then
-  print("You should specify '-i' as an argument for the interactive session,")
-  print("but with no other sources of commands, we start interactive session now anyway")
-   interactive_cli(r)
-end
-
-vpp:disconnect()
-r.done()
-
-
diff --git a/vpp-api/lua/examples/example-acl-plugin.lua b/vpp-api/lua/examples/example-acl-plugin.lua
deleted file mode 100644
index ca01f18d..00000000
--- a/vpp-api/lua/examples/example-acl-plugin.lua
+++ /dev/null
@@ -1,110 +0,0 @@
---[[
-/*
- * Copyright (c) 2016 Cisco and/or its affiliates.
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at:
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-]]
-
-
-vpp = require "vpp-lapi"
-
-root_dir = "/home/ubuntu/vpp"
-pneum_path = root_dir .. "/build-root/install-vpp_debug-native/vpp-api/lib64/libpneum.so"
-
-vpp:init({ pneum_path = pneum_path })
-
-vpp:consume_api(root_dir .. "/build-root/install-vpp_debug-native/vlib-api/vlibmemory/memclnt.api")
-vpp:consume_api(root_dir .. "/build-root/install-vpp_debug-native/vpp/vpp-api/vpe.api")
-vpp:connect("aytest")
-vpp:consume_api(root_dir .. "/plugins/acl-plugin/acl/acl.api", "acl")
-
--- api calls
-reply = vpp:api_call("show_version")
-print("Version: ", reply[1].version)
-print(vpp.hex_dump(reply[1].version))
-print(vpp.dump(reply))
-print("---")
-
-reply = vpp:api_call("acl_del", { context = 42, acl_index = 230 })
-print(vpp.dump(reply))
-print("---")
-
-reply = vpp:api_call("acl_del", { context = 42, acl_index = 8 })
-print(vpp.dump(reply))
-print("---")
-
-reply = vpp:api_call("acl_del", { context = 42, acl_index = 15 })
-print(vpp.dump(reply))
-print("---")
-
-reply = vpp:api_call("acl_add", { context = 42, count = 2, r = { { is_permit = 1, is_ipv6 = 1 }, { is_permit = 0, is_ipv6 = 1 } } })
-print(vpp.dump(reply))
-print("---")
-interface_acl_in = reply[1].acl_index
-
-reply = vpp:api_call("acl_add", { context = 42, count = 3, r = { { is_permit = 1, is_ipv6 = 1 }, { is_permit = 0, is_ipv6 = 1 }, { is_permit = 1, is_ipv6 = 0 } } })
-print(vpp.dump(reply))
-print("---")
-interface_acl_out = reply[1].acl_index
-
-
-reply = vpp:api_call("acl_interface_add_del", { context = 42, sw_if_index = 0, is_add = 1, is_input = 1, acl_index = interface_acl_in })
-print(vpp.dump(reply))
-print("---")
-reply = vpp:api_call("acl_interface_add_del", { context = 42, sw_if_index = 0, is_add = 1, is_input = 1, acl_index = interface_acl_in })
-print(vpp.dump(reply))
-print("---")
-
-reply = vpp:api_call("acl_interface_add_del", { context = 42, sw_if_index = 0, is_add = 1, is_input = 0, acl_index = interface_acl_out })
-print(vpp.dump(reply))
-print("---")
-reply = vpp:api_call("acl_interface_add_del", { context = 42, sw_if_index = 0, is_add = 1, is_input = 0, acl_index = interface_acl_out })
-print(vpp.dump(reply))
-print("---")
-
-reply = vpp:api_call("acl_add", { context = 42, count = 0 })
-print(vpp.dump(reply))
-print("---")
-
-acl_index_to_delete = reply[1].acl_index
-print("Deleting " .. tostring(acl_index_to_delete))
-reply = vpp:api_call("acl_del", { context = 42, acl_index = acl_index_to_delete })
-print(vpp.dump(reply))
-print("---")
-
-reply = vpp:api_call("acl_dump", { context = 42, sw_if_index = 0})
-for ri, rv in ipairs(reply) do 
-  print("Reply message #" .. tostring(ri))
-  print(vpp.dump(rv))
-  for ai, av in ipairs(rv.r) do
-    print("ACL rule #" .. tostring(ai) .. " : " .. vpp.dump(av))
-  end
-   
-end
-print("---")
-
-reply = vpp:api_call("acl_del", { context = 42, acl_index = interface_acl_out })
-print(vpp.dump(reply))
-print("---")
-reply = vpp:api_call("acl_del", { context = 42, acl_index = interface_acl_in })
-print(vpp.dump(reply))
-print("---")
-
-reply = vpp:api_call("acl_dump", { context = 42, sw_if_index = 0})
-print(vpp.dump(reply))
-print("---")
-
-
-vpp:disconnect()
-
-
diff --git a/vpp-api/lua/examples/example-classifier.lua b/vpp-api/lua/examples/example-classifier.lua
deleted file mode 100644
index ec9c3d3e..00000000
--- a/vpp-api/lua/examples/example-classifier.lua
+++ /dev/null
@@ -1,51 +0,0 @@
---[[
-/*
- * Copyright (c) 2016 Cisco and/or its affiliates.
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at:
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-]]
-
-
-local vpp = require "vpp-lapi"
-local bit = require("bit")
-
-root_dir = "/home/ubuntu/vpp"
-pneum_path = root_dir .. "/build-root/install-vpp_lite_debug-native/vpp-api/lib64/libpneum.so"
-
-
-vpp:init({ pneum_path = pneum_path })
-
-vpp:json_api(root_dir .. "/build-root/install-vpp_lite_debug-native/vpp/vpp-api/vpe.api.json")
-
-vpp:connect("aytest")
-
--- api calls
-
-print("Calling API to add a new classifier table")
-reply = vpp:api_call("classify_add_del_table", {
-  context = 43,
-  memory_size = bit.lshift(2, 20),
-  client_index = 42,
-  is_add = 1,
-  nbuckets = 32,
-  skip_n_vectors = 0,
-  match_n_vectors = 1,
-  mask = "\255\255\255\255\255\255\255\255" .. "\255\255\255\255\255\255\255\255"
-})
-print(vpp.dump(reply))
-print("---")
-
-
-vpp:disconnect()
-
-
diff --git a/vpp-api/lua/examples/example-cli.lua b/vpp-api/lua/examples/example-cli.lua
deleted file mode 100644
index 8b84989f..00000000
--- a/vpp-api/lua/examples/example-cli.lua
+++ /dev/null
@@ -1,44 +0,0 @@
---[[
-/*
- * Copyright (c) 2016 Cisco and/or its affiliates.
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at:
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-]]
-
-vpp = require "vpp-lapi"
-
-root_dir = "/home/ubuntu/vpp"
-pneum_path = root_dir .. "/build-root/install-vpp_lite_debug-native/vpp-api/lib64/libpneum.so"
-
-vpp:init({ pneum_path = pneum_path })
-
-vpp:json_api(root_dir .. "/build-root/install-vpp_lite_debug-native/vpp/vpp-api/vpe.api.json")
-
-vpp:connect("aytest")
-
--- api calls
-reply = vpp:api_call("show_version")
-print("Version: ", reply[1].version)
-print(vpp.hex_dump(reply[1].version))
-print(vpp.dump(reply))
-print("---")
-
-
-reply = vpp:api_call("cli_inband", { cmd = "show vers" }) 
-print(vpp.dump(reply))
-print("---")
-
-
-vpp:disconnect()
-
-
diff --git a/vpp-api/lua/examples/lute/README.md b/vpp-api/lua/examples/lute/README.md
deleted file mode 100644
index 8d37250a..00000000
--- a/vpp-api/lua/examples/lute/README.md
+++ /dev/null
@@ -1,66 +0,0 @@
-LUTE: Lua Unit Test Environment
-
-This is a small helper utility to automate some simple tests
-that one might need to do.
-
-Think of it as a hybrid of a screen and expect who
-also took some habits from HTML inline code.
-
-It is quite probably useless for building anything serious,
-but practice shows it is quite efficient at allowing
-convenient temporary quick tests, and for something
-that was written over a course of a couple of evenings it
-is quite a nice little helper tool.
-
-It allows do launch and drive multiple shell sessions,
-and by virtue of having been written in Lua, it of course
-also allows to add the business logic using the Lua code.
-
-If you launch the lute without parameters, it gives you
-the interactive shell to execute the commands in.
-
-If you launch it with an argument, it will attempt to
-read and execute the commands from the file.
-
-Commands:
-
-shell FOO
-
-  spawn a shell in a new PTY under the label FOO.
-
-run FOO bar
-
-  Send "bar" keystrokes followed by "ENTER" to the session FOO
-
-  Special case: "break" word on its own gets translated into ^C being sent.
-
-cd FOO
-
-  "change domain" into session FOO. All subsequent inputs will go,
-  line-buffered, into the session FOO. To jump back up, use ^D (Control-D),
-  or within the file, use ^D^D^D (caret D caret D caret D on its own line)
-
-expect FOO blablabla
-
-  Pause further interpretation of the batch mode until you see "blablabla"
-  in the output of session FOO, or until timeout happens.
-
-sleep N
-
-  Sleep an integer N seconds, if you are in batch mode.
-
-echo blabla
-
-  Echo the remainder of the line to standard output.
-
-For Lua code, there is a pre-existing pseudo-session called "lua",
-which accepts "run lua" command which does what you would expect
-(evaluate the rest of the string in Lua context - being the same
-as lute itself). Also you can do "cd lua" and get into a
-multiline-enabled interpreter shell.
-
-This way for the VPP case you can automate some of the things in your routine
-that you would have to have done manually, and test drive API as well
-as use the realistic native OS components to create the environment around it.
-
-
diff --git a/vpp-api/lua/examples/lute/lute.lua b/vpp-api/lua/examples/lute/lute.lua
deleted file mode 100644
index 89b9924b..00000000
--- a/vpp-api/lua/examples/lute/lute.lua
+++ /dev/null
@@ -1,777 +0,0 @@
---[[
-version = 1
-/*
- * Copyright (c) 2016 Cisco and/or its affiliates.
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at:
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-]]
-
--- LUTE: Lua Unit Test Environment
--- AKA what happens when screen tries to marry with lua and expect,
--- but escapes mid-ceremony.
---
--- comments: @ayourtch
-
-ffi = require("ffi")
-
-vpp = {}
-function vpp.dump(o)
-   if type(o) == 'table' then
-      local s = '{ '
-      for k,v in pairs(o) do
-         if type(k) ~= 'number' then k = '"'..k..'"' end
-         s = s .. '['..k..'] = ' .. vpp.dump(v) .. ','
-      end
-      return s .. '} '
-   else
-      return tostring(o)
-   end
-end
-
-
-ffi.cdef([[
-
-int posix_openpt(int flags);
-int grantpt(int fd);
-int unlockpt(int fd);
-char *ptsname(int fd);
-
-typedef long pid_t;
-typedef long ssize_t;
-typedef long size_t;
-typedef int nfds_t;
-typedef long time_t;
-typedef long suseconds_t;
-
-pid_t fork(void);
-pid_t setsid(void);
-
-int close(int fd);
-int open(char *pathname, int flags);
-
-int dup2(int oldfd, int newfd);
-
-ssize_t read(int fd, void *buf, size_t count);
-ssize_t write(int fd, const void *buf, size_t count);
-
-struct pollfd {
-               int   fd;         /* file descriptor */
-               short events;     /* requested events */
-               short revents;    /* returned events */
-           };
-
-int poll(struct pollfd *fds, nfds_t nfds, int timeout);
-
-struct timeval {
-               time_t      tv_sec;     /* seconds */
-               suseconds_t tv_usec;    /* microseconds */
-           };
-
-int gettimeofday(struct timeval *tv, struct timezone *tz);
-
-int inet_pton(int af, const char *src, void *dst);
-
-]])
-
-ffi.cdef([[
-void *memset(void *s, int c, size_t n);
-void *memcpy(void *dest, void *src, size_t n);
-void *memmove(void *dest, const void *src, size_t n);
-void *memmem(const void *haystack, size_t haystacklen,
-        const void *needle, size_t needlelen);
-]])
-
-
-
-local O_RDWR = 2
-
-
-function os_time()
-  local tv = ffi.new("struct timeval[1]")
-  local ret = ffi.C.gettimeofday(tv, nil)
-  return tonumber(tv[0].tv_sec) + (tonumber(tv[0].tv_usec)/1000000.0)
-end
-
-function sleep(n)
-  local when_wakeup = os_time() + n
-  while os_time() <= when_wakeup do
-    ffi.C.poll(nil, 0, 10)
-  end
-end
-
-
-function c_str(text_in)
-  local text = text_in 
-  local c_str = ffi.new("char[?]", #text+1)
-  ffi.copy(c_str, text)
-  return c_str
-end
-
-function ip46(addr_text)
-  local out = ffi.new("char [200]")
-  local AF_INET6 = 10
-  local AF_INET = 2
-  local is_ip6 = ffi.C.inet_pton(AF_INET6, c_str(addr_text), out)
-  if is_ip6 == 1 then
-    return ffi.string(out, 16), true
-  end
-  local is_ip4 = ffi.C.inet_pton(AF_INET, c_str(addr_text), out)
-  if is_ip4 then
-    return (string.rep("4", 12).. ffi.string(out, 4)), false
-  end
-end
-
-function pty_master_open()
-  local fd = ffi.C.posix_openpt(O_RDWR)
-  ffi.C.grantpt(fd)
-  ffi.C.unlockpt(fd)
-  local p = ffi.C.ptsname(fd)
-  print("PTS:" .. ffi.string(p))
-  return fd, ffi.string(p)
-end
-
-function pty_run(cmd)
-  local master_fd, pts_name = pty_master_open()
-  local child_pid = ffi.C.fork()
-  if (child_pid == -1) then
-    print("Error fork()ing")
-    return -1
-  end 
- 
-  if child_pid ~= 0 then
-    -- print("Parent")
-    return master_fd, child_pid
-  end
-
-  -- print("Child")
-  if (ffi.C.setsid() == -1) then
-    print("Child error setsid")
-    os.exit(-1)
-  end
-
-  ffi.C.close(master_fd)
-
-  local slave_fd = ffi.C.open(c_str(pts_name), O_RDWR)
-  if slave_fd == -1 then
-    print("Child can not open slave fd")
-    os.exit(-2)
-  end
-
-  ffi.C.dup2(slave_fd, 0)
-  ffi.C.dup2(slave_fd, 1)
-  ffi.C.dup2(slave_fd, 2)
-  os.execute(cmd)
-end
-
-function readch()
-  local buf = ffi.new("char[1]")
-  local nread= ffi.C.read(0, buf, 1)
-  -- print("\nREADCH : " .. string.char(buf[0]))
-  return string.char(buf[0])
-end
-
-function stdout_write(str)
-  ffi.C.write(1, c_str(str), #str)
-end
-
-
-readln = {
-split = function(str, pat)
-  local t = {}  -- NOTE: use {n = 0} in Lua-5.0
-  local fpat = "(.-)" .. pat
-  local last_end = 1
-  if str then
-    local s, e, cap = str:find(fpat, 1)
-    while s do
-      if s ~= 1 or cap ~= "" then
-        table.insert(t,cap)
-      end
-      last_end = e+1
-      s, e, cap = str:find(fpat, last_end)
-    end
-    if last_end <= #str then
-      cap = str:sub(last_end)
-      table.insert(t, cap)
-    end
-  end
-  return t
-end,
-
-reader = function()
-  local rl = {}
-
-  rl.init = function()
-    os.execute("stty -icanon min 1 -echo")
-    rl.rawmode = true
-  end
-
-  rl.done = function()
-    os.execute("stty icanon echo")
-    rl.rawmode = false
-  end
-
-  rl.prompt = ">"
-  rl.history = { "" }
-  rl.history_index = 1
-  rl.history_length = 1
-
-  rl.hide_cmd = function()
-    local bs = string.char(8) .. " " .. string.char(8)
-    for i = 1, #rl.command do
-      stdout_write(bs)
-    end
-  end
-
-  rl.show_cmd = function()
-    if rl.command then
-      stdout_write(rl.command)
-    end
-  end
-
-  rl.store_history = function(cmd)
-    if cmd == "" then
-      return
-    end
-    rl.history[rl.history_length] = cmd
-    rl.history_length = rl.history_length + 1
-    rl.history_index = rl.history_length
-    rl.history[rl.history_length] = ""
-  end
-
-  rl.readln = function(stdin_select_fn, batch_cmd, batch_when, batch_expect)
-    local done = false
-    local need_prompt = true
-    rl.command = ""
-
-    if not rl.rawmode then
-      rl.init()
-    end
-
-    while not done do
-      local indent_value = #rl.prompt + #rl.command
-      if need_prompt then
-        stdout_write(rl.prompt)
-        stdout_write(rl.command)
-        need_prompt = false
-      end
-      if type(stdin_select_fn) == "function" then
-        while not stdin_select_fn(indent_value, batch_cmd, batch_when, batch_expect) do
-          stdout_write(rl.prompt)
-          stdout_write(rl.command)
-          indent_value = #rl.prompt + #rl.command
-        end
-        if batch_cmd and ((os_time() > batch_when) or (batch_expect and expect_success(batch_expect, buf, 0))) then
-          stdout_write("\n" .. rl.prompt .. batch_cmd .. "\n")
-          if batch_expect then
-            expect_done(batch_expect)
-          end
-          return batch_cmd, batch_expect
-        end
-      end
-      local ch = readch()
-      if ch:byte(1) == 27 then
-        -- CONTROL
-        local ch2 = readch()
-        -- arrows
-        if ch2:byte(1) == 91 then
-          local ch3 = readch()
-          local b = ch3:byte(1)
-          if b == 65 then
-            ch = "UP"
-          elseif b == 66 then
-            ch = "DOWN"
-          elseif b == 67 then
-            ch = "RIGHT"
-          elseif b == 68 then
-            ch = "LEFT"
-          end
-          -- print("Byte: " .. ch3:byte(1))
-          -- if ch3:byte(1)
-        end
-      end
-
-      if ch == "?" then
-        stdout_write(ch)
-        stdout_write("\n")
-        if rl.help then
-          rl.help(rl)
-        end
-        need_prompt = true
-      elseif ch == "\t" then
-        if rl.tab_complete then
-          rl.tab_complete(rl)
-        end
-        stdout_write("\n")
-        need_prompt = true
-      elseif ch == "\n" then
-        stdout_write(ch)
-        done = true
-      elseif ch == "\004" then
-        stdout_write("\n")
-        rl.command = nil
-        done = true
-      elseif ch == string.char(127) then
-        if rl.command ~= "" then
-          stdout_write(string.char(8) .. " " .. string.char(8))
-          rl.command = string.sub(rl.command, 1, -2)
-        end
-      elseif #ch > 1 then
-        -- control char
-        if ch == "UP" then
-          rl.hide_cmd()
-          if rl.history_index == #rl.history then
-            rl.history[rl.history_index] = rl.command
-          end
-          if rl.history_index > 1 then
-            rl.history_index = rl.history_index - 1
-            rl.command = rl.history[rl.history_index]
-          end
-          rl.show_cmd()
-        elseif ch == "DOWN" then
-          rl.hide_cmd()
-          if rl.history_index < rl.history_length then
-            rl.history_index = rl.history_index + 1
-            rl.command = rl.history[rl.history_index]
-          end
-          rl.show_cmd()
-        end
-      else
-        stdout_write(ch)
-        rl.command = rl.command .. ch
-      end
-    end
-    if rl.command then
-      rl.store_history(rl.command)
-    end
-    return rl.command
-  end
-  return rl
-end
-
-}
-
-local select_fds = {}
-local sessions = {}
-
-local line_erased = false
-
-function erase_line(indent)
-  if not line_erased then
-    line_erased = true
-    stdout_write(string.rep(string.char(8), indent)..string.rep(" ", indent)..string.rep(string.char(8), indent))
-  end
-end
-
-function do_select_stdin(indent, batch_cmd, batch_when, batch_expect)
-  while true do
-    local nfds = 1+#select_fds
-    local pfds = ffi.new("struct pollfd[?]", nfds)
-    pfds[0].fd = 0;
-    pfds[0].events = 1;
-    pfds[0].revents = 0;
-    for i = 1,#select_fds do
-      pfds[i].fd = select_fds[i].fd
-      pfds[i].events = 1
-      pfds[i].revents = 0
-    end
-    if batch_cmd and ((os_time() > batch_when) or (batch_expect and expect_success(batch_expect, buf, 0))) then
-      return true
-    end
-    while ffi.C.poll(pfds, nfds, 10) == 0 do
-      if batch_cmd and ((os_time() > batch_when) or (batch_expect and expect_success(batch_expect, buf, 0))) then
-        return true
-      end
-      if line_erased then
-        line_erased = false
-        return false
-      end
-    end
-    if pfds[0].revents == 1 then
-      return true
-    end
-    for i = 1,#select_fds do
-      if(pfds[i].revents > 0) then
-        if pfds[i].fd ~= select_fds[i].fd then
-          print("File descriptors unequal", pfds[i].fd, select_fds[i].fd)
-        end
-        select_fds[i].cb(select_fds[i], pfds[i].revents, indent)
-      end
-    end
-  end
-end
-
-local buf = ffi.new("char [32768]")
-
-function session_stdout_write(prefix, data)
-  data = prefix .. data:gsub("\n", "\n"..prefix):gsub("\n"..prefix.."$", "\n")
-  
-  stdout_write(data)
-end
-
-function expect_success(sok, buf, nread)
-  local expect_buf_sz = ffi.sizeof(sok.expect_buf) - 128
-  local expect_buf_avail = expect_buf_sz - sok.expect_buf_idx
-  -- print("EXPECT_SUCCESS: nread ".. tostring(nread).. " expect_buf_idx: " .. tostring(sok.expect_buf_idx) .. " expect_buf_avail: " .. tostring(expect_buf_avail) )
-  if expect_buf_avail < 0 then
-    print "EXPECT BUFFER OVERRUN ALREADY"
-    os.exit(1)
-  end
-  if expect_buf_avail < nread then
-    if (nread >= ffi.sizeof(sok.expect_buf)) then
-      print("Read too large of a chunk to fit into expect buffer")
-      return nil
-    end
-    local delta = nread - expect_buf_avail
-
-    ffi.C.memmove(sok.expect_buf, sok.expect_buf + delta, expect_buf_sz - delta)
-    sok.expect_buf_idx = sok.expect_buf_idx - delta
-    expect_buf_avail = nread 
-  end
-  if sok.expect_buf_idx + nread > expect_buf_sz then
-    print("ERROR, I have just overrun the buffer !")
-    os.exit(1)
-  end
-  ffi.C.memcpy(sok.expect_buf + sok.expect_buf_idx, buf, nread)
-  sok.expect_buf_idx = sok.expect_buf_idx + nread
-  if sok.expect_str == nil then
-    return true
-  end
-  local match_p = ffi.C.memmem(sok.expect_buf, sok.expect_buf_idx, sok.expect_str, sok.expect_str_len)
-  if match_p ~= nil then
-    return true
-  end
-  return false
-end
-
-function expect_done(sok)
-  local expect_buf_sz = ffi.sizeof(sok.expect_buf) - 128
-  if not sok.expect_str then 
-    return false
-  end
-  local match_p = ffi.C.memmem(sok.expect_buf, sok.expect_buf_idx, sok.expect_str, sok.expect_str_len)
-  if match_p ~= nil then
-    if sok.expect_cb then
-      sok.expect_cb(sok)
-    end
-    local match_idx = ffi.cast("char *", match_p) - ffi.cast("char *", sok.expect_buf)
-    ffi.C.memmove(sok.expect_buf, ffi.cast("char *", match_p) + sok.expect_str_len, expect_buf_sz - match_idx - sok.expect_str_len)
-    sok.expect_buf_idx = match_idx + sok.expect_str_len
-    sok.expect_success = true
-
-    sok.expect_str = nil
-    sok.expect_str_len = 0
-    return true
-  end
-end
-
-function slave_events(sok, revents, indent)
-  local fd = sok.fd
-  local nread = ffi.C.read(fd, buf, ffi.sizeof(buf)-128)
-  local idx = nread - 1
-  while idx >= 0 and buf[idx] ~= 10 do
-    idx = idx - 1
-  end
-  if idx >= 0 then
-    erase_line(indent)
-    session_stdout_write(sok.prefix, sok.buf .. ffi.string(buf, idx+1))
-    sok.buf = ""
-  end
-  sok.buf = sok.buf .. ffi.string(buf+idx+1, nread-idx-1)
-  -- print("\nRead: " .. tostring(nread))
-  -- stdout_write(ffi.string(buf, nread))
-  if expect_success(sok, buf, nread) then
-    return true
-  end
-  return false
-end
-
-
-function start_session(name)
-  local mfd, cpid = pty_run("/bin/bash")
-  local sok =  { ["fd"] = mfd, ["cb"] = slave_events, ["buf"] = "", ["prefix"] = name .. ":", ["expect_buf"] = ffi.new("char [165536]"), ["expect_buf_idx"] = 0, ["expect_str"] = nil }
-  table.insert(select_fds, sok)
-  sessions[name] = sok
-end
-
-function command_transform(exe)
-  if exe == "break" then
-    exe = string.char(3)
-  end
-  return exe
-end
-
-function session_write(a_session, a_str)
-  if has_session(a_session) then
-    return tonumber(ffi.C.write(sessions[a_session].fd, c_str(a_str), #a_str))
-  else
-    return 0
-  end
-end
-
-function session_exec(a_session, a_cmd)
-  local exe = command_transform(a_cmd) .. "\n"
-  session_write(a_session, exe)
-end
-
-function session_cmd(ui, a_session, a_cmd)
-  if not has_session(a_session) then
-    stdout_write("ERR: No such session '" .. tostring(a_session) .. "'\n")
-    return nil
-  end
-  if a_session == "lua" then
-    local func, msg = loadstring(ui.lua_acc .. a_cmd)
-    -- stdout_write("LOADSTR: " .. vpp.dump({ ret, msg }) .. "\n")
-    if not func and string.match(msg, "") then
-      if a_session ~= ui.in_session then
-         stdout_write("ERR LOADSTR: " .. tostring(msg) .. "\n")
-         return nil
-      end
-      ui.lua_acc = ui.lua_acc .. a_cmd  .. "\n"
-      return true
-    end
-    ui.lua_acc = ""
-    local ret, msg = pcall(func)
-    if ret then
-      return true
-    else
-      stdout_write("ERR: " .. msg .. "\n") 
-      return nil
-    end
-  else
-    session_exec(a_session, a_cmd)
-    if ui.session_cmd_delay then
-      return { "delay", ui.session_cmd_delay }
-    end
-    return true
-  end
-end
-
-function has_session(a_session)
-  if a_session == "lua" then
-    return true
-  end
-  return (sessions[a_session] ~= nil)
-end
-
-function command_match(list, input, output)
-  for i, v in ipairs(list) do
-    local m = {}
-    m[1], m[2], m[3], m[4], m[5], m[6], m[7], m[8], m[9] = string.match(input, v[1])
-    -- print("MATCH: ", vpp.dump(m))
-    if m[1] then
-       output["result"] = m
-       output["result_index"] = i
-       return m
-    end 
-  end
-  return nil
-end
-
-function cmd_spawn_shell(ui, a_arg)
-  start_session(a_arg[1])
-  return true
-end
-
-function cmd_run_cmd(ui, a_arg)
-  local a_sess = a_arg[1]
-  local a_cmd = a_arg[2]
-  return session_cmd(ui, a_sess, a_cmd)
-end
-
-function cmd_cd(ui, a_arg)
-  local a_sess = a_arg[1]
-  if has_session(a_sess) then
-    ui.in_session = a_sess
-    return true
-  else
-    stdout_write("ERR: Unknown session '".. tostring(a_sess) .. "'\n")
-    return nil
-  end
-end
-
-function cmd_sleep(ui, a_arg)
-  return { "delay", tonumber(a_arg[1]) }
-end
-
-function cmd_expect(ui, a_arg)
-  local a_sess = a_arg[1]
-  local a_expect = a_arg[2]
-  local sok = sessions[a_sess]
-  if not sok then
-    stdout_write("ERR: unknown session '" .. tostring(a_sess) .. "'\n")
-    return nil
-  end
-  sok.expect_str = c_str(a_expect)
-  sok.expect_str_len = #a_expect
-  return { "expect", a_sess }
-end
-
-function cmd_info(ui, a_arg)
-  local a_sess = a_arg[1]
-  local sok = sessions[a_sess]
-  if not sok then
-    stdout_write("ERR: unknown session '" .. tostring(a_sess) .. "'\n")
-    return nil
-  end
-  print("Info for session " .. tostring(a_sess) .. "\n")
-  print("Expect buffer index: " .. tostring(sok.expect_buf_idx))
-  print("Expect buffer: '" .. tostring(ffi.string(sok.expect_buf, sok.expect_buf_idx)) .. "'\n")
-  if sok.expect_str then
-    print("Expect string: '" .. tostring(ffi.string(sok.expect_str, sok.expect_str_len)) .. "'\n")
-  else
-    print("Expect string not set\n")
-  end
-end
-
-function cmd_echo(ui, a_arg)
-  local a_data = a_arg[1]
-  print("ECHO: " .. tostring(a_data))
-end
-
-main_command_table = {
-  { "^shell ([a-zA-Z0-9_]+)$", cmd_spawn_shell },
-  { "^run ([a-zA-Z0-9_]+) (.+)$", cmd_run_cmd },
-  { "^cd ([a-zA-Z0-9_]+)$", cmd_cd  },
-  { "^sleep ([0-9]+)$", cmd_sleep },
-  { "^expect ([a-zA-Z0-9_]+) (.-)$", cmd_expect },
-  { "^info ([a-zA-Z0-9_]+)$", cmd_info },
-  { "^echo (.-)$", cmd_echo }
-}
-
-
-
-function ui_set_prompt(ui)
-  if ui.in_session then 
-    if ui.in_session == "lua" then
-      if #ui.lua_acc > 0 then
-        ui.r.prompt = ui.in_session .. ">>"
-      else
-        ui.r.prompt = ui.in_session .. ">"
-      end 
-    else
-      ui.r.prompt = ui.in_session .. "> "
-    end
-  else
-    ui.r.prompt = "> "
-  end
-  return ui.r.prompt
-end
-
-function ui_run_command(ui, cmd)
-  -- stdout_write("Command: " .. tostring(cmd) .. "\n")
-  local ret = false
-  if ui.in_session then
-    if cmd then
-      if cmd == "^D^D^D" then
-        ui.in_session = nil
-        ret = true
-      else
-        ret = session_cmd(ui, ui.in_session, cmd)
-      end
-    else
-      ui.in_session = nil
-      ret = true
-    end
-  else  
-    if cmd then
-      local out = {}
-      if cmd == "" then
-        ret = true
-      end
-      if command_match(main_command_table, cmd, out) then
-        local i = out.result_index
-        local m = out.result
-        if main_command_table[i][2] then
-          ret = main_command_table[i][2](ui, m)
-        end
-      end
-    end
-    if not cmd or cmd == "quit" then
-      return "quit"
-    end
-  end
-  return ret
-end
-
-local ui = {}
-ui.in_session = nil
-ui.r = readln.reader() 
-ui.lua_acc = ""
-ui.session_cmd_delay = 0.3
-
-local lines = ""
-
-local done = false
--- a helper function which always returns nil
-local no_next_line = function() return nil end
-
--- a function which returns the next batch line
-local next_line = no_next_line
-
-local batchfile = arg[1]
-
-if batchfile then
-  local f = io.lines(batchfile)
-  next_line = function() 
-    local line = f()
-    if line then 
-      return line
-    else 
-      next_line = no_next_line
-      session_stdout_write(batchfile .. ":", "End of batch\n")
-      return nil
-    end
-  end
-end
-
-
-local batch_when = 0
-local batch_expect = nil
-while not done do
-  local prompt = ui_set_prompt(ui)
-  local batch_cmd = next_line() 
-  local cmd, expect_sok = ui.r.readln(do_select_stdin, batch_cmd, batch_when, batch_expect)
-  if expect_sok and not expect_success(expect_sok, buf, 0) then
-    if not cmd_ret and next_line ~= no_next_line then
-      print("ERR: expect timeout\n")
-      next_line = no_next_line
-    end
-  else 
-    local cmd_ret = ui_run_command(ui, cmd)
-    if not cmd_ret and next_line ~= no_next_line then
-      print("ERR: Error during batch execution\n")
-      next_line = no_next_line
-    end
-
-    if cmd_ret  == "quit" then
-      done = true
-    end
-    batch_expect = nil
-    batch_when = 0
-    if type(cmd_ret) == "table" then
-      if cmd_ret[1] == "delay" then
-	batch_when = os_time() + tonumber(cmd_ret[2])
-      end
-      if cmd_ret[1] == "expect" then
-	batch_expect = sessions[cmd_ret[2]]
-	batch_when = os_time() + 15
-      end
-    end
-  end
-end
-ui.r.done()
-
-os.exit(1)
-
-
-
diff --git a/vpp-api/lua/examples/lute/script-inout-acl-noacl.lute b/vpp-api/lua/examples/lute/script-inout-acl-noacl.lute
deleted file mode 100644
index a24d04bf..00000000
--- a/vpp-api/lua/examples/lute/script-inout-acl-noacl.lute
+++ /dev/null
@@ -1,329 +0,0 @@
-shell vppbuild
-run vppbuild stty -echo
-run vppbuild sudo -u ubuntu -i bash -c "(cd vpp && make plugins && echo ALLGOOD)"
-expect vppbuild ALLGOOD
-
-shell s0
-shell s1
-shell s2
-
-
-cd s1
-unshare -n /bin/bash
-/sbin/ifconfig -a
-^D^D^D
-
-cd s2
-unshare -n /bin/bash
-/sbin/ifconfig -a
-^D^D^D
-
-
-cd lua
-
-function session_get_bash_pid(s)
-  if not has_session(s) then
-    return nil
-  end
-  local fname = "/tmp/lute-"..s.."-pid.txt"
-
-  session_exec(s, "echo $$ >" .. fname)
-  -- it's a dirty hack but it's quick
-  sleep(0.5)
-  local pid = io.lines(fname)()
-  print("Got pid for " .. s .. " : " .. tostring(pid))
-  return(tonumber(pid))
-end
-
-function session_connect_with(s0, s1)
-  -- local pid0 = tostring(session_get_bash_pid(s0))
-  local pid1 = tostring(session_get_bash_pid(s1))
-  local eth_options = { "rx", "tx",  "sg", "tso", "ufo", "gso", "gro", "lro", "rxvlan", "txvlan", "rxhash" }
-  local this_end = s0 .. "_" .. s1
-  local other_end = s1 .. "_" .. s0
-  session_exec(s0, "ip link add name " .. this_end .. " type veth peer name " .. other_end)
-  session_exec(s0, "ip link set dev " .. this_end .. " up promisc on")
-  for i, option in ipairs(eth_options) do
-    session_exec(s0, "/sbin/ethtool --offload " .. this_end .. " " .. option .. " off")
-    session_exec(s0, "/sbin/ethtool --offload " .. other_end .. " " .. option .. " off")
-  end
-  session_exec(s0, "ip link set dev " .. other_end .. " up promisc on netns /proc/" .. pid1 .. "/ns/net")
-  sleep(0.5)
-end
-
-^D^D^D
-run lua session_connect_with("s0", "s1")
-run lua session_connect_with("s0", "s2")
-
-cd s1
-ip -6 addr add dev s1_s0 2001:db8:1::1/64
-ip -4 addr add dev s1_s0 192.0.2.1/24
-ip link set dev s1_s0 up promisc on
-^D^D^D
-
-cd s2
-ip -6 addr add dev s2_s0 2001:db8:1::2/64
-ip -6 addr add dev s2_s0 2001:db8:1::3/64
-ip -6 addr add dev s2_s0 2001:db8:1::4/64
-ip -4 addr add dev s2_s0 192.0.2.2/24
-ip -4 addr add dev s2_s0:1 192.0.2.3/24
-ip -4 addr add dev s2_s0:2 192.0.2.4/24
-ip link set dev s2_s0 up promisc on
-^D^D^D
-
-run s1 ip addr
-run s2 ip addr
-shell VPP
-cd VPP
-cd /home/ubuntu/vpp
-make debug 
-r
-^D^D^D
-expect VPP DBGvpp#
-
-cd lua
--- Initialization of the Lua environment for talking to VPP
-vpp = require("vpp-lapi")
-root_dir = "/home/ubuntu/vpp"
-pneum_path = root_dir .. "/build-root/install-vpp_debug-native/vpp-api/lib64/libpneum.so"
-vpp:init({ pneum_path = pneum_path })
-vpp:consume_api(root_dir .. "/build-root/install-vpp_debug-native/vlib-api/vlibmemory/memclnt.api")
-vpp:consume_api(root_dir .. "/build-root/install-vpp_debug-native/vpp/vpp-api/vpe.api")
-vpp:connect("aytest")
-vpp:consume_api(root_dir .. "/plugins/acl-plugin/acl/acl.api", "acl")
-
-^D^D^D
-
-cd lua
-
-reply = vpp:api_call("af_packet_create", { host_if_name = "s0_s1", hw_addr = "AAAAAA" })
-vpp_if_to_s1 = reply[1].sw_if_index
-
-reply = vpp:api_call("af_packet_create", { host_if_name = "s0_s2", hw_addr = "AAAAAA" })
-vpp_if_to_s2 = reply[1].sw_if_index
-
-ifaces = { vpp_if_to_s1, vpp_if_to_s2 }
-
-reply = vpp:api_call("sw_interface_set_flags", { sw_if_index = vpp_if_to_s1, admin_up_down = 1, link_up_down = 1 })
-print(vpp.dump(reply))
-reply = vpp:api_call("sw_interface_set_flags", { sw_if_index = vpp_if_to_s2, admin_up_down = 1, link_up_down = 1 })
-print(vpp.dump(reply))
-
-bd_id = 42
-
-reply = vpp:api_call("bridge_domain_add_del", { bd_id = bd_id, flood = 1, uu_flood = 1, forward = 1, learn = 1, arp_term = 0, is_add = 1 })
-print(vpp.dump(reply))
-
-for i, v in ipairs(ifaces) do
-  reply = vpp:api_call("sw_interface_set_l2_bridge", { rx_sw_if_index = v, bd_id = bd_id, shg = 0, bvi = 0, enable = 1 } )
-  print(vpp.dump(reply))
-end
-
-^D^D^D
-
-run s1 ping -c 3 192.0.2.2
-expect s1 packet loss
-run s1 ping -c 3 192.0.2.3
-expect s1 packet loss
-run s1 ping -c 3 192.0.2.4
-expect s1 packet loss
-run s1 ping6 -c 3 2001:db8:1::2
-expect s1 packet loss
-run s1 ping6 -c 3 2001:db8:1::3
-expect s1 packet loss
-run s1 ping6 -c 3 2001:db8:1::4
-expect s1 packet loss
-
-
-cd lua
---- ACL testing
-
---[[ temporary comment out
-
-reply = vpp:api_call("acl_del", { context = 42, acl_index = 230 })
-print(vpp.dump(reply))
-print("---")
-
-reply = vpp:api_call("acl_del", { context = 42, acl_index = 8 })
-print(vpp.dump(reply))
-print("---")
-
-reply = vpp:api_call("acl_del", { context = 42, acl_index = 15 })
-print(vpp.dump(reply))
-print("---")
-
-reply = vpp:api_call("acl_add_replace", { context = 42, count = 2, r = { { is_permit = 1, is_ipv6 = 1 }, { is_permit = 0, is_ipv6 = 1 } } })
-print(vpp.dump(reply))
-print("---")
-interface_acl_in = reply[1].acl_index
-
-reply = vpp:api_call("acl_add_replace", { context = 42, count = 3, r = { { is_permit = 1, is_ipv6 = 1 }, { is_permit = 0, is_ipv6 = 1 }, { is_permit = 1, is_ipv6 = 0 } } })
-print(vpp.dump(reply))
-print("---")
-interface_acl_out = reply[1].acl_index
-
-
-reply = vpp:api_call("acl_interface_add_del", { context = 42, sw_if_index = 0, is_add = 1, is_input = 1, acl_index = interface_acl_in })
-print(vpp.dump(reply))
-print("---")
-
-reply = vpp:api_call("acl_interface_add_del", { context = 42, sw_if_index = 0, is_add = 1, is_input = 1, acl_index = interface_acl_in })
-print(vpp.dump(reply))
-print("---")
-
-reply = vpp:api_call("acl_interface_add_del", { context = 42, sw_if_index = 0, is_add = 1, is_input = 0, acl_index = interface_acl_out })
-print(vpp.dump(reply))
-print("---")
-
-reply = vpp:api_call("acl_interface_add_del", { context = 42, sw_if_index = 0, is_add = 1, is_input = 0, acl_index = interface_acl_out })
-print(vpp.dump(reply))
-print("---")
-
-reply = vpp:api_call("acl_add_replace", { context = 42, count = 0 })
-print(vpp.dump(reply))
-print("---")
-
-acl_index_to_delete = reply[1].acl_index
-print("Deleting " .. tostring(acl_index_to_delete))
-reply = vpp:api_call("acl_del", { context = 42, acl_index = acl_index_to_delete })
-print(vpp.dump(reply))
-print("---")
-
-reply = vpp:api_call("acl_dump", { context = 42, sw_if_index = 0})
-for ri, rv in ipairs(reply) do
-  print("Reply message #" .. tostring(ri))
-  print(vpp.dump(rv))
-  for ai, av in ipairs(rv.r) do
-    print("ACL rule #" .. tostring(ai) .. " : " .. vpp.dump(av))
-  end
-
-end
-print("---")
-
-reply = vpp:api_call("acl_del", { context = 42, acl_index = interface_acl_out })
-print(vpp.dump(reply))
-print("---")
-reply = vpp:api_call("acl_del", { context = 42, acl_index = interface_acl_in })
-print(vpp.dump(reply))
-print("---")
-
-reply = vpp:api_call("acl_dump", { context = 42, sw_if_index = 0})
-print(vpp.dump(reply))
-print("---")
-
-reply = vpp:api_call("acl_dump", { context = 42, sw_if_index = 4294967295 })
-print(vpp.dump(reply))
-print("---")
-
-
-]] -- end of comment out
-
----- Should be nothing ^^
-r = {
-  { is_permit = 1, is_ipv6 = 1, dst_ip_addr = ip46("2001:db8:1::2"), dst_ip_prefix_len = 128 },
-  { is_permit = 0, is_ipv6 = 1, dst_ip_addr = ip46("2001:db8:1::3"), dst_ip_prefix_len = 128 },
-  { is_permit = 1, is_ipv6 = 1, dst_ip_addr = ip46("2001:db8::"), dst_ip_prefix_len = 32  },
-  { is_permit = 1, is_ipv6 = 0, dst_ip_addr = ip46("192.0.2.2"), dst_ip_prefix_len = 32},
-  { is_permit = 0, is_ipv6 = 0, dst_ip_addr = ip46("192.0.2.3"), dst_ip_prefix_len = 32 },
-}
-
-reply = vpp:api_call("acl_add_replace", { context = 42, count = 5, r = r })
-print(vpp.dump(reply))
-print("---")
-interface_acl_in = reply[1].acl_index
-
-reply = vpp:api_call("acl_add_replace", { context = 42, count = 3, r = { { is_permit = 1, is_ipv6 = 1 }, { is_permit = 0, is_ipv6 = 1 }, { is_permit = 1, is_ipv6 = 0 } } })
-print(vpp.dump(reply))
-print("---")
-interface_acl_out = reply[1].acl_in
-
-reply = vpp:api_call("acl_interface_add_del", { context = 42, sw_if_index = vpp_if_to_s1, is_add = 1, is_input = 1, acl_index = interface_acl_in })
-print(vpp.dump(reply))
-print("---")
-
---reply = vpp:api_call("acl_interface_add_del", { context = 42, sw_if_index = vpp_if_to_s2, is_add = 1, is_input = 0, acl_index = interface_acl_out })
--- print(vpp.dump(reply))
---print("---")
-
-^D^D^D
-
-run VPP clear trace
-run VPP trace add af-packet-input 100
-run s1 ping6 -c 3 2001:db8:1::2
-expect s1 packet loss
-run VPP show trace
-expect VPP match: inacl 0 rule 0
-
-run VPP clear trace
-run VPP trace add af-packet-input 100
-run s1 ping6 -c 3 2001:db8:1::3
-expect s1 packet loss
-run VPP show trace 
-expect VPP match: inacl 0 rule 1
-
-run VPP clear trace
-run VPP trace add af-packet-input 100
-run s1 ping6 -c 3 2001:db8:1::4
-expect s1 packet loss
-run VPP show trace
-expect VPP match: inacl 0 rule 2
-
-run VPP clear trace
-run VPP trace add af-packet-input 100
-run s1 ping -c 3 192.0.2.2
-expect s1 packet loss
-run VPP show trace
-expect VPP match: inacl 0 rule 3
-
-run VPP clear trace
-run VPP trace add af-packet-input 100
-run s1 ping -c 3 192.0.2.3
-expect s1 packet loss
-run VPP show trace
-expect VPP match: inacl 0 rule 4
-
-
-cd lua
-
---- TEST OUTBOUND ACL
-
-r1 = {
-  { is_permit = 1, is_ipv6 = 1, src_ip_addr = ip46("2001:db8:1::1"), src_ip_prefix_len = 128, dst_ip_addr = ip46("2001:db8:1::2"), dst_ip_prefix_len = 128 },
-  { is_permit = 0, is_ipv6 = 1, src_ip_addr = ip46("2001:db8:1::1"), src_ip_prefix_len = 128, dst_ip_addr = ip46("2001:db8:1::4"), dst_ip_prefix_len = 128 }
-}
-
-reply = vpp:api_call("acl_add_replace", { context = 42, count = 3, r = r1 })
-print(vpp.dump(reply))
-print("---")
-interface_acl_out = reply[1].acl_index
-
-reply = vpp:api_call("acl_interface_add_del", { context = 42, sw_if_index = vpp_if_to_s2, is_add = 1, is_input = 0, acl_index = interface_acl_out })
-print(vpp.dump(reply))
-print("---")
-
-
-^D^D^D
-
-run VPP clear trace
-run VPP trace add af-packet-input 100
-run s1 ping6 -c 3 2001:db8:1::2
-expect s1 packet loss
-run VPP show trace
-expect VPP match: outacl 2 rule 0
-
-run VPP clear trace
-run VPP trace add af-packet-input 100
-run s1 ping6 -c 3 2001:db8:1::3
-expect s1 packet loss
-run VPP show trace 
-expect VPP match: inacl 0 rule 1
-
-run VPP clear trace
-run VPP trace add af-packet-input 100
-run s1 ping6 -c 3 2001:db8:1::4
-expect s1 packet loss
-run VPP show trace
-expect VPP match: outacl 2 rule 1
-
-run lua print("ALL GOOD!")
-
diff --git a/vpp-api/lua/examples/lute/script-inout-acl-old.lute b/vpp-api/lua/examples/lute/script-inout-acl-old.lute
deleted file mode 100644
index 9edebf02..00000000
--- a/vpp-api/lua/examples/lute/script-inout-acl-old.lute
+++ /dev/null
@@ -1,329 +0,0 @@
-shell vppbuild
-run vppbuild stty -echo
-run vppbuild sudo -u ubuntu -i bash -c "(cd vpp && make plugins && echo ALLGOOD)"
-expect vppbuild ALLGOOD
-
-shell s0
-shell s1
-shell s2
-
-
-cd s1
-unshare -n /bin/bash
-/sbin/ifconfig -a
-^D^D^D
-
-cd s2
-unshare -n /bin/bash
-/sbin/ifconfig -a
-^D^D^D
-
-
-cd lua
-
-function session_get_bash_pid(s)
-  if not has_session(s) then
-    return nil
-  end
-  local fname = "/tmp/lute-"..s.."-pid.txt"
-
-  session_exec(s, "echo $$ >" .. fname)
-  -- it's a dirty hack but it's quick
-  sleep(0.5)
-  local pid = io.lines(fname)()
-  print("Got pid for " .. s .. " : " .. tostring(pid))
-  return(tonumber(pid))
-end
-
-function session_connect_with(s0, s1)
-  -- local pid0 = tostring(session_get_bash_pid(s0))
-  local pid1 = tostring(session_get_bash_pid(s1))
-  local eth_options = { "rx", "tx",  "sg", "tso", "ufo", "gso", "gro", "lro", "rxvlan", "txvlan", "rxhash" }
-  local this_end = s0 .. "_" .. s1
-  local other_end = s1 .. "_" .. s0
-  session_exec(s0, "ip link add name " .. this_end .. " type veth peer name " .. other_end)
-  session_exec(s0, "ip link set dev " .. this_end .. " up promisc on")
-  for i, option in ipairs(eth_options) do
-    session_exec(s0, "/sbin/ethtool --offload " .. this_end .. " " .. option .. " off")
-    session_exec(s0, "/sbin/ethtool --offload " .. other_end .. " " .. option .. " off")
-  end
-  session_exec(s0, "ip link set dev " .. other_end .. " up promisc on netns /proc/" .. pid1 .. "/ns/net")
-  sleep(0.5)
-end
-
-^D^D^D
-run lua session_connect_with("s0", "s1")
-run lua session_connect_with("s0", "s2")
-
-cd s1
-ip -6 addr add dev s1_s0 2001:db8:1::1/64
-ip -4 addr add dev s1_s0 192.0.2.1/24
-ip link set dev s1_s0 up promisc on
-^D^D^D
-
-cd s2
-ip -6 addr add dev s2_s0 2001:db8:1::2/64
-ip -6 addr add dev s2_s0 2001:db8:1::3/64
-ip -6 addr add dev s2_s0 2001:db8:1::4/64
-ip -4 addr add dev s2_s0 192.0.2.2/24
-ip -4 addr add dev s2_s0:1 192.0.2.3/24
-ip -4 addr add dev s2_s0:2 192.0.2.4/24
-ip link set dev s2_s0 up promisc on
-^D^D^D
-
-run s1 ip addr
-run s2 ip addr
-shell VPP
-cd VPP
-cd /home/ubuntu/vpp
-make debug 
-r
-^D^D^D
-expect VPP DBGvpp#
-
-cd lua
--- Initialization of the Lua environment for talking to VPP
-vpp = require("vpp-lapi")
-root_dir = "/home/ubuntu/vpp"
-pneum_path = root_dir .. "/build-root/install-vpp_debug-native/vpp-api/lib64/libpneum.so"
-vpp:init({ pneum_path = pneum_path })
-vpp:consume_api(root_dir .. "/build-root/install-vpp_debug-native/vlib-api/vlibmemory/memclnt.api")
-vpp:consume_api(root_dir .. "/build-root/install-vpp_debug-native/vpp/vpp-api/vpe.api")
-vpp:connect("aytest")
-vpp:consume_api(root_dir .. "/plugins/acl-plugin/acl/acl.api", "acl")
-
-^D^D^D
-
-cd lua
-
-reply = vpp:api_call("af_packet_create", { host_if_name = "s0_s1", hw_addr = "AAAAAA" })
-vpp_if_to_s1 = reply[1].sw_if_index
-
-reply = vpp:api_call("af_packet_create", { host_if_name = "s0_s2", hw_addr = "AAAAAA" })
-vpp_if_to_s2 = reply[1].sw_if_index
-
-ifaces = { vpp_if_to_s1, vpp_if_to_s2 }
-
-reply = vpp:api_call("sw_interface_set_flags", { sw_if_index = vpp_if_to_s1, admin_up_down = 1, link_up_down = 1 })
-print(vpp.dump(reply))
-reply = vpp:api_call("sw_interface_set_flags", { sw_if_index = vpp_if_to_s2, admin_up_down = 1, link_up_down = 1 })
-print(vpp.dump(reply))
-
-bd_id = 42
-
-reply = vpp:api_call("bridge_domain_add_del", { bd_id = bd_id, flood = 1, uu_flood = 1, forward = 1, learn = 1, arp_term = 0, is_add = 1 })
-print(vpp.dump(reply))
-
-for i, v in ipairs(ifaces) do
-  reply = vpp:api_call("sw_interface_set_l2_bridge", { rx_sw_if_index = v, bd_id = bd_id, shg = 0, bvi = 0, enable = 1 } )
-  print(vpp.dump(reply))
-end
-
-^D^D^D
-
-run s1 ping -c 3 192.0.2.2
-expect s1 packet loss
-run s1 ping -c 3 192.0.2.3
-expect s1 packet loss
-run s1 ping -c 3 192.0.2.4
-expect s1 packet loss
-run s1 ping6 -c 3 2001:db8:1::2
-expect s1 packet loss
-run s1 ping6 -c 3 2001:db8:1::3
-expect s1 packet loss
-run s1 ping6 -c 3 2001:db8:1::4
-expect s1 packet loss
-
-
-cd lua
---- ACL testing
-
---[[ temporary comment out
-
-reply = vpp:api_call("acl_del", { context = 42, acl_index = 230 })
-print(vpp.dump(reply))
-print("---")
-
-reply = vpp:api_call("acl_del", { context = 42, acl_index = 8 })
-print(vpp.dump(reply))
-print("---")
-
-reply = vpp:api_call("acl_del", { context = 42, acl_index = 15 })
-print(vpp.dump(reply))
-print("---")
-
-reply = vpp:api_call("acl_add", { context = 42, count = 2, r = { { is_permit = 1, is_ipv6 = 1 }, { is_permit = 0, is_ipv6 = 1 } } })
-print(vpp.dump(reply))
-print("---")
-interface_acl_in = reply[1].acl_index
-
-reply = vpp:api_call("acl_add", { context = 42, count = 3, r = { { is_permit = 1, is_ipv6 = 1 }, { is_permit = 0, is_ipv6 = 1 }, { is_permit = 1, is_ipv6 = 0 } } })
-print(vpp.dump(reply))
-print("---")
-interface_acl_out = reply[1].acl_index
-
-
-reply = vpp:api_call("acl_interface_add_del", { context = 42, sw_if_index = 0, is_add = 1, is_input = 1, acl_index = interface_acl_in })
-print(vpp.dump(reply))
-print("---")
-
-reply = vpp:api_call("acl_interface_add_del", { context = 42, sw_if_index = 0, is_add = 1, is_input = 1, acl_index = interface_acl_in })
-print(vpp.dump(reply))
-print("---")
-
-reply = vpp:api_call("acl_interface_add_del", { context = 42, sw_if_index = 0, is_add = 1, is_input = 0, acl_index = interface_acl_out })
-print(vpp.dump(reply))
-print("---")
-
-reply = vpp:api_call("acl_interface_add_del", { context = 42, sw_if_index = 0, is_add = 1, is_input = 0, acl_index = interface_acl_out })
-print(vpp.dump(reply))
-print("---")
-
-reply = vpp:api_call("acl_add", { context = 42, count = 0 })
-print(vpp.dump(reply))
-print("---")
-
-acl_index_to_delete = reply[1].acl_index
-print("Deleting " .. tostring(acl_index_to_delete))
-reply = vpp:api_call("acl_del", { context = 42, acl_index = acl_index_to_delete })
-print(vpp.dump(reply))
-print("---")
-
-reply = vpp:api_call("acl_dump", { context = 42, sw_if_index = 0})
-for ri, rv in ipairs(reply) do
-  print("Reply message #" .. tostring(ri))
-  print(vpp.dump(rv))
-  for ai, av in ipairs(rv.r) do
-    print("ACL rule #" .. tostring(ai) .. " : " .. vpp.dump(av))
-  end
-
-end
-print("---")
-
-reply = vpp:api_call("acl_del", { context = 42, acl_index = interface_acl_out })
-print(vpp.dump(reply))
-print("---")
-reply = vpp:api_call("acl_del", { context = 42, acl_index = interface_acl_in })
-print(vpp.dump(reply))
-print("---")
-
-reply = vpp:api_call("acl_dump", { context = 42, sw_if_index = 0})
-print(vpp.dump(reply))
-print("---")
-
-reply = vpp:api_call("acl_dump", { context = 42, sw_if_index = 4294967295 })
-print(vpp.dump(reply))
-print("---")
-
-
-]] -- end of comment out
-
----- Should be nothing ^^
-r = {
-  { is_permit = 1, is_ipv6 = 1, dst_ip_addr = ip46("2001:db8:1::2"), dst_ip_prefix_len = 128 },
-  { is_permit = 0, is_ipv6 = 1, dst_ip_addr = ip46("2001:db8:1::3"), dst_ip_prefix_len = 128 },
-  { is_permit = 1, is_ipv6 = 1, dst_ip_addr = ip46("2001:db8::"), dst_ip_prefix_len = 32  },
-  { is_permit = 1, is_ipv6 = 0, dst_ip_addr = ip46("192.0.2.2"), dst_ip_prefix_len = 32},
-  { is_permit = 0, is_ipv6 = 0, dst_ip_addr = ip46("192.0.2.3"), dst_ip_prefix_len = 32 },
-}
-
-reply = vpp:api_call("acl_add", { context = 42, count = 5, r = r })
-print(vpp.dump(reply))
-print("---")
-interface_acl_in = reply[1].acl_index
-
-reply = vpp:api_call("acl_add", { context = 42, count = 3, r = { { is_permit = 1, is_ipv6 = 1 }, { is_permit = 0, is_ipv6 = 1 }, { is_permit = 1, is_ipv6 = 0 } } })
-print(vpp.dump(reply))
-print("---")
-interface_acl_out = reply[1].acl_in
-
-reply = vpp:api_call("acl_interface_add_del", { context = 42, sw_if_index = vpp_if_to_s1, is_add = 1, is_input = 1, acl_index = interface_acl_in })
-print(vpp.dump(reply))
-print("---")
-
---reply = vpp:api_call("acl_interface_add_del", { context = 42, sw_if_index = vpp_if_to_s2, is_add = 1, is_input = 0, acl_index = interface_acl_out })
--- print(vpp.dump(reply))
---print("---")
-
-^D^D^D
-
-run VPP clear trace
-run VPP trace add af-packet-input 100
-run s1 ping6 -c 3 2001:db8:1::2
-expect s1 packet loss
-run VPP show trace
-expect VPP match: inacl 0 rule 0
-
-run VPP clear trace
-run VPP trace add af-packet-input 100
-run s1 ping6 -c 3 2001:db8:1::3
-expect s1 packet loss
-run VPP show trace 
-expect VPP match: inacl 0 rule 1
-
-run VPP clear trace
-run VPP trace add af-packet-input 100
-run s1 ping6 -c 3 2001:db8:1::4
-expect s1 packet loss
-run VPP show trace
-expect VPP match: inacl 0 rule 2
-
-run VPP clear trace
-run VPP trace add af-packet-input 100
-run s1 ping -c 3 192.0.2.2
-expect s1 packet loss
-run VPP show trace
-expect VPP match: inacl 0 rule 3
-
-run VPP clear trace
-run VPP trace add af-packet-input 100
-run s1 ping -c 3 192.0.2.3
-expect s1 packet loss
-run VPP show trace
-expect VPP match: inacl 0 rule 4
-
-
-cd lua
-
---- TEST OUTBOUND ACL
-
-r1 = {
-  { is_permit = 1, is_ipv6 = 1, src_ip_addr = ip46("2001:db8:1::1"), src_ip_prefix_len = 128, dst_ip_addr = ip46("2001:db8:1::2"), dst_ip_prefix_len = 128 },
-  { is_permit = 0, is_ipv6 = 1, src_ip_addr = ip46("2001:db8:1::1"), src_ip_prefix_len = 128, dst_ip_addr = ip46("2001:db8:1::4"), dst_ip_prefix_len = 128 }
-}
-
-reply = vpp:api_call("acl_add", { context = 42, count = 3, r = r1 })
-print(vpp.dump(reply))
-print("---")
-interface_acl_out = reply[1].acl_index
-
-reply = vpp:api_call("acl_interface_add_del", { context = 42, sw_if_index = vpp_if_to_s2, is_add = 1, is_input = 0, acl_index = interface_acl_out })
-print(vpp.dump(reply))
-print("---")
-
-
-^D^D^D
-
-run VPP clear trace
-run VPP trace add af-packet-input 100
-run s1 ping6 -c 3 2001:db8:1::2
-expect s1 packet loss
-run VPP show trace
-expect VPP match: outacl 2 rule 0
-
-run VPP clear trace
-run VPP trace add af-packet-input 100
-run s1 ping6 -c 3 2001:db8:1::3
-expect s1 packet loss
-run VPP show trace 
-expect VPP match: inacl 0 rule 1
-
-run VPP clear trace
-run VPP trace add af-packet-input 100
-run s1 ping6 -c 3 2001:db8:1::4
-expect s1 packet loss
-run VPP show trace
-expect VPP match: outacl 2 rule 1
-
-run lua print("ALL GOOD!")
-
diff --git a/vpp-api/lua/examples/lute/script-inout-acl.lute b/vpp-api/lua/examples/lute/script-inout-acl.lute
deleted file mode 100644
index d7e7423c..00000000
--- a/vpp-api/lua/examples/lute/script-inout-acl.lute
+++ /dev/null
@@ -1,329 +0,0 @@
-shell vppbuild
-run vppbuild stty -echo
-run vppbuild sudo -u ubuntu -i bash -c "(cd vpp && make plugins && echo ALLGOOD)"
-expect vppbuild ALLGOOD
-
-shell s0
-shell s1
-shell s2
-
-
-cd s1
-unshare -n /bin/bash
-/sbin/ifconfig -a
-^D^D^D
-
-cd s2
-unshare -n /bin/bash
-/sbin/ifconfig -a
-^D^D^D
-
-
-cd lua
-
-function session_get_bash_pid(s)
-  if not has_session(s) then
-    return nil
-  end
-  local fname = "/tmp/lute-"..s.."-pid.txt"
-
-  session_exec(s, "echo $$ >" .. fname)
-  -- it's a dirty hack but it's quick
-  sleep(0.5)
-  local pid = io.lines(fname)()
-  print("Got pid for " .. s .. " : " .. tostring(pid))
-  return(tonumber(pid))
-end
-
-function session_connect_with(s0, s1)
-  -- local pid0 = tostring(session_get_bash_pid(s0))
-  local pid1 = tostring(session_get_bash_pid(s1))
-  local eth_options = { "rx", "tx",  "sg", "tso", "ufo", "gso", "gro", "lro", "rxvlan", "txvlan", "rxhash" }
-  local this_end = s0 .. "_" .. s1
-  local other_end = s1 .. "_" .. s0
-  session_exec(s0, "ip link add name " .. this_end .. " type veth peer name " .. other_end)
-  session_exec(s0, "ip link set dev " .. this_end .. " up promisc on")
-  for i, option in ipairs(eth_options) do
-    session_exec(s0, "/sbin/ethtool --offload " .. this_end .. " " .. option .. " off")
-    session_exec(s0, "/sbin/ethtool --offload " .. other_end .. " " .. option .. " off")
-  end
-  session_exec(s0, "ip link set dev " .. other_end .. " up promisc on netns /proc/" .. pid1 .. "/ns/net")
-  sleep(0.5)
-end
-
-^D^D^D
-run lua session_connect_with("s0", "s1")
-run lua session_connect_with("s0", "s2")
-
-cd s1
-ip -6 addr add dev s1_s0 2001:db8:1::1/64
-ip -4 addr add dev s1_s0 192.0.2.1/24
-ip link set dev s1_s0 up promisc on
-^D^D^D
-
-cd s2
-ip -6 addr add dev s2_s0 2001:db8:1::2/64
-ip -6 addr add dev s2_s0 2001:db8:1::3/64
-ip -6 addr add dev s2_s0 2001:db8:1::4/64
-ip -4 addr add dev s2_s0 192.0.2.2/24
-ip -4 addr add dev s2_s0:1 192.0.2.3/24
-ip -4 addr add dev s2_s0:2 192.0.2.4/24
-ip link set dev s2_s0 up promisc on
-^D^D^D
-
-run s1 ip addr
-run s2 ip addr
-shell VPP
-cd VPP
-cd /home/ubuntu/vpp
-make debug 
-r
-^D^D^D
-expect VPP DBGvpp#
-
-cd lua
--- Initialization of the Lua environment for talking to VPP
-vpp = require("vpp-lapi")
-root_dir = "/home/ubuntu/vpp"
-pneum_path = root_dir .. "/build-root/install-vpp_debug-native/vpp-api/lib64/libpneum.so"
-vpp:init({ pneum_path = pneum_path })
-vpp:consume_api(root_dir .. "/build-root/install-vpp_debug-native/vlib-api/vlibmemory/memclnt.api")
-vpp:consume_api(root_dir .. "/build-root/install-vpp_debug-native/vpp/vpp-api/vpe.api")
-vpp:connect("aytest")
-vpp:consume_api(root_dir .. "/plugins/acl-plugin/acl/acl.api", "acl")
-
-^D^D^D
-
-cd lua
-
-reply = vpp:api_call("af_packet_create", { host_if_name = "s0_s1", hw_addr = "AAAAAA" })
-vpp_if_to_s1 = reply[1].sw_if_index
-
-reply = vpp:api_call("af_packet_create", { host_if_name = "s0_s2", hw_addr = "AAAAAA" })
-vpp_if_to_s2 = reply[1].sw_if_index
-
-ifaces = { vpp_if_to_s1, vpp_if_to_s2 }
-
-reply = vpp:api_call("sw_interface_set_flags", { sw_if_index = vpp_if_to_s1, admin_up_down = 1, link_up_down = 1 })
-print(vpp.dump(reply))
-reply = vpp:api_call("sw_interface_set_flags", { sw_if_index = vpp_if_to_s2, admin_up_down = 1, link_up_down = 1 })
-print(vpp.dump(reply))
-
-bd_id = 42
-
-reply = vpp:api_call("bridge_domain_add_del", { bd_id = bd_id, flood = 1, uu_flood = 1, forward = 1, learn = 1, arp_term = 0, is_add = 1 })
-print(vpp.dump(reply))
-
-for i, v in ipairs(ifaces) do
-  reply = vpp:api_call("sw_interface_set_l2_bridge", { rx_sw_if_index = v, bd_id = bd_id, shg = 0, bvi = 0, enable = 1 } )
-  print(vpp.dump(reply))
-end
-
-^D^D^D
-
-run s1 ping -c 3 192.0.2.2
-expect s1 packet loss
-run s1 ping -c 3 192.0.2.3
-expect s1 packet loss
-run s1 ping -c 3 192.0.2.4
-expect s1 packet loss
-run s1 ping6 -c 3 2001:db8:1::2
-expect s1 packet loss
-run s1 ping6 -c 3 2001:db8:1::3
-expect s1 packet loss
-run s1 ping6 -c 3 2001:db8:1::4
-expect s1 packet loss
-
-
-cd lua
---- ACL testing
-
---[[ temporary comment out
-
-reply = vpp:api_call("acl_del", { context = 42, acl_index = 230 })
-print(vpp.dump(reply))
-print("---")
-
-reply = vpp:api_call("acl_del", { context = 42, acl_index = 8 })
-print(vpp.dump(reply))
-print("---")
-
-reply = vpp:api_call("acl_del", { context = 42, acl_index = 15 })
-print(vpp.dump(reply))
-print("---")
-
-reply = vpp:api_call("acl_add_replace", { context = 42, acl_index = -1, count = 2, r = { { is_permit = 1, is_ipv6 = 1 }, { is_permit = 0, is_ipv6 = 1 } } })
-print(vpp.dump(reply))
-print("---")
-interface_acl_in = reply[1].acl_index
-
-reply = vpp:api_call("acl_add_replace", { context = 42, acl_index = -1, count = 3, r = { { is_permit = 1, is_ipv6 = 1 }, { is_permit = 0, is_ipv6 = 1 }, { is_permit = 1, is_ipv6 = 0 } } })
-print(vpp.dump(reply))
-print("---")
-interface_acl_out = reply[1].acl_index
-
-
-reply = vpp:api_call("acl_interface_add_del", { context = 42, sw_if_index = 0, is_add = 1, is_input = 1, acl_index = interface_acl_in })
-print(vpp.dump(reply))
-print("---")
-
-reply = vpp:api_call("acl_interface_add_del", { context = 42, sw_if_index = 0, is_add = 1, is_input = 1, acl_index = interface_acl_in })
-print(vpp.dump(reply))
-print("---")
-
-reply = vpp:api_call("acl_interface_add_del", { context = 42, sw_if_index = 0, is_add = 1, is_input = 0, acl_index = interface_acl_out })
-print(vpp.dump(reply))
-print("---")
-
-reply = vpp:api_call("acl_interface_add_del", { context = 42, sw_if_index = 0, is_add = 1, is_input = 0, acl_index = interface_acl_out })
-print(vpp.dump(reply))
-print("---")
-
-reply = vpp:api_call("acl_add_replace", { context = 42, acl_index = -1, count = 0 })
-print(vpp.dump(reply))
-print("---")
-
-acl_index_to_delete = reply[1].acl_index
-print("Deleting " .. tostring(acl_index_to_delete))
-reply = vpp:api_call("acl_del", { context = 42, acl_index = acl_index_to_delete })
-print(vpp.dump(reply))
-print("---")
-
-reply = vpp:api_call("acl_dump", { context = 42, sw_if_index = 0})
-for ri, rv in ipairs(reply) do
-  print("Reply message #" .. tostring(ri))
-  print(vpp.dump(rv))
-  for ai, av in ipairs(rv.r) do
-    print("ACL rule #" .. tostring(ai) .. " : " .. vpp.dump(av))
-  end
-
-end
-print("---")
-
-reply = vpp:api_call("acl_del", { context = 42, acl_index = interface_acl_out })
-print(vpp.dump(reply))
-print("---")
-reply = vpp:api_call("acl_del", { context = 42, acl_index = interface_acl_in })
-print(vpp.dump(reply))
-print("---")
-
-reply = vpp:api_call("acl_dump", { context = 42, sw_if_index = 0})
-print(vpp.dump(reply))
-print("---")
-
-reply = vpp:api_call("acl_dump", { context = 42, sw_if_index = 4294967295 })
-print(vpp.dump(reply))
-print("---")
-
-
-]] -- end of comment out
-
----- Should be nothing ^^
-r = {
-  { is_permit = 1, is_ipv6 = 1, dst_ip_addr = ip46("2001:db8:1::2"), dst_ip_prefix_len = 128 },
-  { is_permit = 0, is_ipv6 = 1, dst_ip_addr = ip46("2001:db8:1::3"), dst_ip_prefix_len = 128 },
-  { is_permit = 1, is_ipv6 = 1, dst_ip_addr = ip46("2001:db8::"), dst_ip_prefix_len = 32  },
-  { is_permit = 1, is_ipv6 = 0, dst_ip_addr = ip46("192.0.2.2"), dst_ip_prefix_len = 32},
-  { is_permit = 0, is_ipv6 = 0, dst_ip_addr = ip46("192.0.2.3"), dst_ip_prefix_len = 32 },
-}
-
-reply = vpp:api_call("acl_add_replace", { context = 42, acl_index = -1, count = 5, r = r })
-print(vpp.dump(reply))
-print("---")
-interface_acl_in = reply[1].acl_index
-
-reply = vpp:api_call("acl_add_replace", { context = 42, acl_index = -1, count = 3, r = { { is_permit = 1, is_ipv6 = 1 }, { is_permit = 0, is_ipv6 = 1 }, { is_permit = 1, is_ipv6 = 0 } } })
-print(vpp.dump(reply))
-print("---")
-interface_acl_out = reply[1].acl_in
-
-reply = vpp:api_call("acl_interface_add_del", { context = 42, sw_if_index = vpp_if_to_s1, is_add = 1, is_input = 1, acl_index = interface_acl_in })
-print(vpp.dump(reply))
-print("---")
-
---reply = vpp:api_call("acl_interface_add_del", { context = 42, sw_if_index = vpp_if_to_s2, is_add = 1, is_input = 0, acl_index = interface_acl_out })
--- print(vpp.dump(reply))
---print("---")
-
-^D^D^D
-
-run VPP clear trace
-run VPP trace add af-packet-input 100
-run s1 ping6 -c 3 2001:db8:1::2
-expect s1 packet loss
-run VPP show trace
-expect VPP match: inacl 0 rule 0
-
-run VPP clear trace
-run VPP trace add af-packet-input 100
-run s1 ping6 -c 3 2001:db8:1::3
-expect s1 packet loss
-run VPP show trace 
-expect VPP match: inacl 0 rule 1
-
-run VPP clear trace
-run VPP trace add af-packet-input 100
-run s1 ping6 -c 3 2001:db8:1::4
-expect s1 packet loss
-run VPP show trace
-expect VPP match: inacl 0 rule 2
-
-run VPP clear trace
-run VPP trace add af-packet-input 100
-run s1 ping -c 3 192.0.2.2
-expect s1 packet loss
-run VPP show trace
-expect VPP match: inacl 0 rule 3
-
-run VPP clear trace
-run VPP trace add af-packet-input 100
-run s1 ping -c 3 192.0.2.3
-expect s1 packet loss
-run VPP show trace
-expect VPP match: inacl 0 rule 4
-
-
-cd lua
-
---- TEST OUTBOUND ACL
-
-r1 = {
-  { is_permit = 1, is_ipv6 = 1, src_ip_addr = ip46("2001:db8:1::1"), src_ip_prefix_len = 128, dst_ip_addr = ip46("2001:db8:1::2"), dst_ip_prefix_len = 128 },
-  { is_permit = 0, is_ipv6 = 1, src_ip_addr = ip46("2001:db8:1::1"), src_ip_prefix_len = 128, dst_ip_addr = ip46("2001:db8:1::4"), dst_ip_prefix_len = 128 }
-}
-
-reply = vpp:api_call("acl_add_replace", { context = 42, acl_index = -1, count = 3, r = r1 })
-print(vpp.dump(reply))
-print("---")
-interface_acl_out = reply[1].acl_index
-
-reply = vpp:api_call("acl_interface_add_del", { context = 42, sw_if_index = vpp_if_to_s2, is_add = 1, is_input = 0, acl_index = interface_acl_out })
-print(vpp.dump(reply))
-print("---")
-
-
-^D^D^D
-
-run VPP clear trace
-run VPP trace add af-packet-input 100
-run s1 ping6 -c 3 2001:db8:1::2
-expect s1 packet loss
-run VPP show trace
-expect VPP match: outacl 2 rule 0
-
-run VPP clear trace
-run VPP trace add af-packet-input 100
-run s1 ping6 -c 3 2001:db8:1::3
-expect s1 packet loss
-run VPP show trace 
-expect VPP match: inacl 0 rule 1
-
-run VPP clear trace
-run VPP trace add af-packet-input 100
-run s1 ping6 -c 3 2001:db8:1::4
-expect s1 packet loss
-run VPP show trace
-expect VPP match: outacl 2 rule 1
-
-run lua print("ALL GOOD!")
-
diff --git a/vpp-api/lua/examples/lute/script.lute b/vpp-api/lua/examples/lute/script.lute
deleted file mode 100644
index c3dd90f2..00000000
--- a/vpp-api/lua/examples/lute/script.lute
+++ /dev/null
@@ -1,7 +0,0 @@
-shell s1
-expect s1 $
-run s1 echo testing123
-expect s1 $
-run s1 echo done
-quit
-
diff --git a/vpp-api/lua/examples/lute/sessions-acl.lute b/vpp-api/lua/examples/lute/sessions-acl.lute
deleted file mode 100644
index ac237ef9..00000000
--- a/vpp-api/lua/examples/lute/sessions-acl.lute
+++ /dev/null
@@ -1,308 +0,0 @@
-run lua -- collectgarbage("stop")
-
-shell vppbuild
-run vppbuild stty -echo
-run vppbuild sudo -u ubuntu -i bash -c "(cd vpp && make plugins && echo ALLGOOD)"
-expect vppbuild ALLGOOD
-
-shell s0
-shell s1
-shell s2
-
-
-cd s1
-unshare -n /bin/bash
-/sbin/ifconfig -a
-^D^D^D
-
-cd s2
-unshare -n /bin/bash
-/sbin/ifconfig -a
-^D^D^D
-
-
-cd lua
-
-function session_get_bash_pid(s)
-  if not has_session(s) then
-    return nil
-  end
-  local fname = "/tmp/lute-"..s.."-pid.txt"
-
-  session_exec(s, "echo $$ >" .. fname)
-  -- it's a dirty hack but it's quick
-  sleep(0.5)
-  local pid = io.lines(fname)()
-  print("Got pid for " .. s .. " : " .. tostring(pid))
-  return(tonumber(pid))
-end
-
-function session_connect_with(s0, s1)
-  -- local pid0 = tostring(session_get_bash_pid(s0))
-  local pid1 = tostring(session_get_bash_pid(s1))
-  local eth_options = { "rx", "tx",  "sg", "tso", "ufo", "gso", "gro", "lro", "rxvlan", "txvlan", "rxhash" }
-  local this_end = s0 .. "_" .. s1
-  local other_end = s1 .. "_" .. s0
-  session_exec(s0, "ip link add name " .. this_end .. " type veth peer name " .. other_end)
-  session_exec(s0, "ip link set dev " .. this_end .. " up promisc on")
-  for i, option in ipairs(eth_options) do
-    session_exec(s0, "/sbin/ethtool --offload " .. this_end .. " " .. option .. " off")
-    session_exec(s0, "/sbin/ethtool --offload " .. other_end .. " " .. option .. " off")
-  end
-  session_exec(s0, "ip link set dev " .. other_end .. " up promisc on netns /proc/" .. pid1 .. "/ns/net")
-  sleep(0.5)
-end
-
-^D^D^D
-run lua session_connect_with("s0", "s1")
-run lua session_connect_with("s0", "s2")
-
-cd s1
-ip -6 addr add dev s1_s0 2001:db8:1::1/64
-ip -4 addr add dev s1_s0 192.0.2.1/24
-ip link set dev s1_s0 up promisc on
-^D^D^D
-
-cd s2
-ip -6 addr add dev s2_s0 2001:db8:1::2/64
-ip -6 addr add dev s2_s0 2001:db8:1::3/64
-ip -6 addr add dev s2_s0 2001:db8:1::4/64
-ip -4 addr add dev s2_s0 192.0.2.2/24
-ip -4 addr add dev s2_s0:1 192.0.2.3/24
-ip -4 addr add dev s2_s0:2 192.0.2.4/24
-ip link set dev s2_s0 up promisc on
-^D^D^D
-
-run s1 ip addr
-run s2 ip addr
-shell VPP
-cd VPP
-cd /home/ubuntu/vpp
-make debug 
-r
-^D^D^D
-expect VPP DBGvpp#
-
-cd lua
--- Initialization of the Lua environment for talking to VPP
-vpp = require("vpp-lapi")
-root_dir = "/home/ubuntu/vpp"
-pneum_path = root_dir .. "/build-root/install-vpp_debug-native/vpp-api/lib64/libpneum.so"
-vpp:init({ pneum_path = pneum_path })
-vpp:consume_api(root_dir .. "/build-root/install-vpp_debug-native/vlib-api/vlibmemory/memclnt.api")
-vpp:consume_api(root_dir .. "/build-root/install-vpp_debug-native/vpp/vpp-api/vpe.api")
-vpp:connect("aytest")
-vpp:consume_api(root_dir .. "/plugins/acl-plugin/acl/acl.api", "acl")
-
-^D^D^D
-
-cd lua
-
-reply = vpp:api_call("af_packet_create", { host_if_name = "s0_s1", hw_addr = "AAAAAA" })
-vpp_if_to_s1 = reply[1].sw_if_index
-
-reply = vpp:api_call("af_packet_create", { host_if_name = "s0_s2", hw_addr = "AAAAAA" })
-vpp_if_to_s2 = reply[1].sw_if_index
-
-ifaces = { vpp_if_to_s1, vpp_if_to_s2 }
-
-reply = vpp:api_call("sw_interface_set_flags", { sw_if_index = vpp_if_to_s1, admin_up_down = 1, link_up_down = 1 })
-print(vpp.dump(reply))
-reply = vpp:api_call("sw_interface_set_flags", { sw_if_index = vpp_if_to_s2, admin_up_down = 1, link_up_down = 1 })
-print(vpp.dump(reply))
-
-bd_id = 42
-
-reply = vpp:api_call("bridge_domain_add_del", { bd_id = bd_id, flood = 1, uu_flood = 1, forward = 1, learn = 1, arp_term = 0, is_add = 1 })
-print(vpp.dump(reply))
-
-for i, v in ipairs(ifaces) do
-  reply = vpp:api_call("sw_interface_set_l2_bridge", { rx_sw_if_index = v, bd_id = bd_id, shg = 0, bvi = 0, enable = 1 } )
-  print(vpp.dump(reply))
-end
-
-^D^D^D
-
-run s1 ping -c 3 192.0.2.2
-expect s1 packet loss
-run s1 ping -c 3 192.0.2.3
-expect s1 packet loss
-run s1 ping -c 3 192.0.2.4
-expect s1 packet loss
-run s1 ping6 -c 3 2001:db8:1::2
-expect s1 packet loss
-run s1 ping6 -c 3 2001:db8:1::3
-expect s1 packet loss
-run s1 ping6 -c 3 2001:db8:1::4
-expect s1 packet loss
-
-
-cd lua
---- ACL testing
-
---[[ temporary comment out
-
-reply = vpp:api_call("acl_del", { context = 42, acl_index = 230 })
-print(vpp.dump(reply))
-print("---")
-
-reply = vpp:api_call("acl_del", { context = 42, acl_index = 8 })
-print(vpp.dump(reply))
-print("---")
-
-reply = vpp:api_call("acl_del", { context = 42, acl_index = 15 })
-print(vpp.dump(reply))
-print("---")
-
-reply = vpp:api_call("acl_add", { context = 42, count = 2, r = { { is_permit = 1, is_ipv6 = 1 }, { is_permit = 0, is_ipv6 = 1 } } })
-print(vpp.dump(reply))
-print("---")
-interface_acl_in = reply[1].acl_index
-
-reply = vpp:api_call("acl_add", { context = 42, count = 3, r = { { is_permit = 1, is_ipv6 = 1 }, { is_permit = 0, is_ipv6 = 1 }, { is_permit = 1, is_ipv6 = 0 } } })
-print(vpp.dump(reply))
-print("---")
-interface_acl_out = reply[1].acl_index
-
-
-reply = vpp:api_call("acl_interface_add_del", { context = 42, sw_if_index = 0, is_add = 1, is_input = 1, acl_index = interface_acl_in })
-print(vpp.dump(reply))
-print("---")
-
-reply = vpp:api_call("acl_interface_add_del", { context = 42, sw_if_index = 0, is_add = 1, is_input = 1, acl_index = interface_acl_in })
-print(vpp.dump(reply))
-print("---")
-
-reply = vpp:api_call("acl_interface_add_del", { context = 42, sw_if_index = 0, is_add = 1, is_input = 0, acl_index = interface_acl_out })
-print(vpp.dump(reply))
-print("---")
-
-reply = vpp:api_call("acl_interface_add_del", { context = 42, sw_if_index = 0, is_add = 1, is_input = 0, acl_index = interface_acl_out })
-print(vpp.dump(reply))
-print("---")
-
-reply = vpp:api_call("acl_add", { context = 42, count = 0 })
-print(vpp.dump(reply))
-print("---")
-
-acl_index_to_delete = reply[1].acl_index
-print("Deleting " .. tostring(acl_index_to_delete))
-reply = vpp:api_call("acl_del", { context = 42, acl_index = acl_index_to_delete })
-print(vpp.dump(reply))
-print("---")
-
-reply = vpp:api_call("acl_dump", { context = 42, sw_if_index = 0})
-for ri, rv in ipairs(reply) do
-  print("Reply message #" .. tostring(ri))
-  print(vpp.dump(rv))
-  for ai, av in ipairs(rv.r) do
-    print("ACL rule #" .. tostring(ai) .. " : " .. vpp.dump(av))
-  end
-
-end
-print("---")
-
-reply = vpp:api_call("acl_del", { context = 42, acl_index = interface_acl_out })
-print(vpp.dump(reply))
-print("---")
-reply = vpp:api_call("acl_del", { context = 42, acl_index = interface_acl_in })
-print(vpp.dump(reply))
-print("---")
-
-reply = vpp:api_call("acl_dump", { context = 42, sw_if_index = 0})
-print(vpp.dump(reply))
-print("---")
-
-reply = vpp:api_call("acl_dump", { context = 42, sw_if_index = 4294967295 })
-print(vpp.dump(reply))
-print("---")
-
-
-]] -- end of comment out
-
----- Should be nothing ^^
-r = {
-  { is_permit = 1, is_ipv6 = 1, dst_ip_addr = ip46("2001:db8:1::2"), dst_ip_prefix_len = 128 },
-  { is_permit = 0, is_ipv6 = 1, dst_ip_addr = ip46("2001:db8:1::3"), dst_ip_prefix_len = 128 },
-  { is_permit = 1, is_ipv6 = 1, dst_ip_addr = ip46("2001:db8::"), dst_ip_prefix_len = 32  },
-  { is_permit = 1, is_ipv6 = 0, dst_ip_addr = ip46("192.0.2.2"), dst_ip_prefix_len = 32},
-  { is_permit = 0, is_ipv6 = 0, dst_ip_addr = ip46("192.0.2.3"), dst_ip_prefix_len = 32 },
-}
-
-reply = vpp:api_call("acl_add", { context = 42, count = 5, r = r })
-print(vpp.dump(reply))
-print("---")
-interface_acl_in = reply[1].acl_index
-
-reply = vpp:api_call("acl_add", { context = 42, count = 3, r = { { is_permit = 1, is_ipv6 = 1 }, { is_permit = 0, is_ipv6 = 1 }, { is_permit = 1, is_ipv6 = 0 } } })
-print(vpp.dump(reply))
-print("---")
-interface_acl_out = reply[1].acl_in
-
-reply = vpp:api_call("acl_interface_add_del", { context = 42, sw_if_index = vpp_if_to_s1, is_add = 1, is_input = 1, acl_index = interface_acl_in })
-print(vpp.dump(reply))
-print("---")
-
---- TEST OUTBOUND ACL
-
-r1 = {
-  { is_permit = 1, is_ipv6 = 1, src_ip_addr = ip46("2001:db8:1::1"), src_ip_prefix_len = 128, dst_ip_addr = ip46("2001:db8:1::2"), dst_ip_prefix_len = 128 },
-  { is_permit = 0, is_ipv6 = 1, src_ip_addr = ip46("2001:db8:1::1"), src_ip_prefix_len = 128, dst_ip_addr = ip46("2001:db8:1::4"), dst_ip_prefix_len = 128 },
-  { is_permit = 2, is_ipv6 = 0 }
-}
-
-reply = vpp:api_call("acl_add", { context = 42, count = 3, r = r1 })
-print(vpp.dump(reply))
-print("---")
-interface_acl_out = reply[1].acl_index
-
-reply = vpp:api_call("acl_interface_add_del", { context = 42, sw_if_index = vpp_if_to_s2, is_add = 1, is_input = 0, acl_index = interface_acl_out })
-print(vpp.dump(reply))
-print("---")
-
-r2 = {
-  { is_permit = 1, is_ipv6 = 1 },
-  { is_permit = 0, is_ipv6 = 0 }
-}
-
-reply = vpp:api_call("acl_add", { context = 42, count = 2, r = r2 })
-print(vpp.dump(reply))
-print("---")
-second_interface_acl_in = reply[1].acl_index
-
-reply = vpp:api_call("acl_interface_add_del", { context = 42, sw_if_index = vpp_if_to_s2, is_add = 1, is_input = 1, acl_index = second_interface_acl_in })
-print(vpp.dump(reply))
-print("---")
-
-^D^D^D
-
-run VPP show classify tables
-run VPP clear trace
-run VPP trace add af-packet-input 100
-run s2 nc -v -l -p 22
-run s1 nc 192.0.2.2 22
-run s1 echo
-sleep 1
-run s1 break
-sleep 1
-run VPP show trace
-expect VPP match: outacl 2 rule 2
-run VPP show classify tables
-
-
-run VPP show classify tables
-run VPP clear trace
-run VPP trace add af-packet-input 100
-run s2 nc -v -l -p 22
-run s1 nc 192.0.2.2 22
-run s1 echo
-sleep 1
-run s1 break
-sleep 1
-run VPP show trace
-expect VPP match: outacl 2 rule 2
-run VPP show classify tables
-
-
-run lua print("ALL GOOD!")
-
diff --git a/vpp-api/lua/vpp-lapi.lua b/vpp-api/lua/vpp-lapi.lua
deleted file mode 100644
index ebfd032b..00000000
--- a/vpp-api/lua/vpp-lapi.lua
+++ /dev/null
@@ -1,989 +0,0 @@
---[[
-/*
- * Copyright (c) 2016 Cisco and/or its affiliates.
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at:
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-]]
-
--- json decode/encode from https://gist.github.com/tylerneylon/59f4bcf316be525b30ab
--- licensed by the author tylerneylon into public domain. Thanks!
-
-local json = {}
-
--- Internal functions.
-
-local function kind_of(obj)
-  if type(obj) ~= 'table' then return type(obj) end
-  local i = 1
-  for _ in pairs(obj) do
-    if obj[i] ~= nil then i = i + 1 else return 'table' end
-  end
-  if i == 1 then return 'table' else return 'array' end
-end
-
-local function escape_str(s)
-  local in_char  = {'\\', '"', '/', '\b', '\f', '\n', '\r', '\t'}
-  local out_char = {'\\', '"', '/',  'b',  'f',  'n',  'r',  't'}
-  for i, c in ipairs(in_char) do
-    s = s:gsub(c, '\\' .. out_char[i])
-  end
-  return s
-end
-
--- Returns pos, did_find; there are two cases:
--- 1. Delimiter found: pos = pos after leading space + delim; did_find = true.
--- 2. Delimiter not found: pos = pos after leading space;     did_find = false.
--- This throws an error if err_if_missing is true and the delim is not found.
-local function skip_delim(str, pos, delim, err_if_missing)
-  pos = pos + #str:match('^%s*', pos)
-  if str:sub(pos, pos) ~= delim then
-    if err_if_missing then
-      error('Expected ' .. delim .. ' near position ' .. pos)
-    end
-    return pos, false
-  end
-  return pos + 1, true
-end
-
--- Expects the given pos to be the first character after the opening quote.
--- Returns val, pos; the returned pos is after the closing quote character.
-local function parse_str_val(str, pos, val)
-  val = val or ''
-  local early_end_error = 'End of input found while parsing string.'
-  if pos > #str then error(early_end_error) end
-  local c = str:sub(pos, pos)
-  if c == '"'  then return val, pos + 1 end
-  if c ~= '\\' then return parse_str_val(str, pos + 1, val .. c) end
-  -- We must have a \ character.
-  local esc_map = {b = '\b', f = '\f', n = '\n', r = '\r', t = '\t'}
-  local nextc = str:sub(pos + 1, pos + 1)
-  if not nextc then error(early_end_error) end
-  return parse_str_val(str, pos + 2, val .. (esc_map[nextc] or nextc))
-end
-
--- Returns val, pos; the returned pos is after the number's final character.
-local function parse_num_val(str, pos)
-  local num_str = str:match('^-?%d+%.?%d*[eE]?[+-]?%d*', pos)
-  local val = tonumber(num_str)
-  if not val then error('Error parsing number at position ' .. pos .. '.') end
-  return val, pos + #num_str
-end
-
-
--- Public values and functions.
-
-function json.stringify(obj, as_key)
-  local s = {}  -- We'll build the string as an array of strings to be concatenated.
-  local kind = kind_of(obj)  -- This is 'array' if it's an array or type(obj) otherwise.
-  if kind == 'array' then
-    if as_key then error('Can\'t encode array as key.') end
-    s[#s + 1] = '['
-    for i, val in ipairs(obj) do
-      if i > 1 then s[#s + 1] = ', ' end
-      s[#s + 1] = json.stringify(val)
-    end
-    s[#s + 1] = ']'
-  elseif kind == 'table' then
-    if as_key then error('Can\'t encode table as key.') end
-    s[#s + 1] = '{'
-    for k, v in pairs(obj) do
-      if #s > 1 then s[#s + 1] = ', ' end
-      s[#s + 1] = json.stringify(k, true)
-      s[#s + 1] = ':'
-      s[#s + 1] = json.stringify(v)
-    end
-    s[#s + 1] = '}'
-  elseif kind == 'string' then
-    return '"' .. escape_str(obj) .. '"'
-  elseif kind == 'number' then
-    if as_key then return '"' .. tostring(obj) .. '"' end
-    return tostring(obj)
-  elseif kind == 'boolean' then
-    return tostring(obj)
-  elseif kind == 'nil' then
-    return 'null'
-  else
-    error('Unjsonifiable type: ' .. kind .. '.')
-  end
-  return table.concat(s)
-end
-
-json.null = {}  -- This is a one-off table to represent the null value.
-
-function json.parse(str, pos, end_delim)
-  pos = pos or 1
-  if pos > #str then error('Reached unexpected end of input.') end
-  local pos = pos + #str:match('^%s*', pos)  -- Skip whitespace.
-  local first = str:sub(pos, pos)
-  if first == '{' then  -- Parse an object.
-    local obj, key, delim_found = {}, true, true
-    pos = pos + 1
-    while true do
-      key, pos = json.parse(str, pos, '}')
-      if key == nil then return obj, pos end
-      if not delim_found then error('Comma missing between object items.') end
-      pos = skip_delim(str, pos, ':', true)  -- true -> error if missing.
-      obj[key], pos = json.parse(str, pos)
-      pos, delim_found = skip_delim(str, pos, ',')
-    end
-  elseif first == '[' then  -- Parse an array.
-    local arr, val, delim_found = {}, true, true
-    pos = pos + 1
-    while true do
-      val, pos = json.parse(str, pos, ']')
-      if val == nil then return arr, pos end
-      if not delim_found then error('Comma missing between array items.') end
-      arr[#arr + 1] = val
-      pos, delim_found = skip_delim(str, pos, ',')
-    end
-  elseif first == '"' then  -- Parse a string.
-    return parse_str_val(str, pos + 1)
-  elseif first == '-' or first:match('%d') then  -- Parse a number.
-    return parse_num_val(str, pos)
-  elseif first == end_delim then  -- End of an object or array.
-    return nil, pos + 1
-  else  -- Parse true, false, or null.
-    local literals = {['true'] = true, ['false'] = false, ['null'] = json.null}
-    for lit_str, lit_val in pairs(literals) do
-      local lit_end = pos + #lit_str - 1
-      if str:sub(pos, lit_end) == lit_str then return lit_val, lit_end + 1 end
-    end
-    local pos_info_str = 'position ' .. pos .. ': ' .. str:sub(pos, pos + 10)
-    error('Invalid json syntax starting at ' .. pos_info_str)
-  end
-end
-
-
-local vpp = {}
-
-local ffi = require("ffi")
-
---[[
-
-The basic type definitions. A bit of weird gymnastic with
-unionization of the hton* and ntoh* functions results
-is to make handling of signed and unsigned types a bit cleaner,
-essentially building typecasting into a C union.
-
-The vl_api_opaque_message_t is a synthetic type assumed to have
-enough storage to hold the entire API message regardless of the type.
-During the operation it is casted to the specific message struct types.
-
-]]
-
-
-ffi.cdef([[
-
-typedef uint8_t u8;
-typedef int8_t i8;
-typedef uint16_t u16;
-typedef int16_t i16;
-typedef uint32_t u32;
-typedef int32_t i32;
-typedef uint64_t u64;
-typedef int64_t i64;
-typedef double f64;
-typedef float f32;
-
-#pragma pack(1)
-typedef union {
-  u16 u16;
-  i16 i16;
-} lua_ui16t;
-
-#pragma pack(1)
-typedef union {
-  u32 u32;
-  i32 i32;
-} lua_ui32t;
-
-u16 ntohs(uint16_t hostshort);
-u16 htons(uint16_t hostshort);
-u32 htonl(uint32_t along);
-u32 ntohl(uint32_t along);
-void *memset(void *s, int c, size_t n);
-void *memcpy(void *dest, void *src, size_t n);
-
-#pragma pack(1)
-typedef struct _vl_api_opaque_message {
-  u16 _vl_msg_id;
-  u8  data[65536];
-} vl_api_opaque_message_t;
-]])
-
-
--- CRC-based version stuff
-
-local crc32c_table = ffi.new('const uint32_t[256]',
-  { 0x00000000, 0xF26B8303, 0xE13B70F7, 0x1350F3F4,
-  0xC79A971F, 0x35F1141C, 0x26A1E7E8, 0xD4CA64EB,
-  0x8AD958CF, 0x78B2DBCC, 0x6BE22838, 0x9989AB3B,
-  0x4D43CFD0, 0xBF284CD3, 0xAC78BF27, 0x5E133C24,
-  0x105EC76F, 0xE235446C, 0xF165B798, 0x030E349B,
-  0xD7C45070, 0x25AFD373, 0x36FF2087, 0xC494A384,
-  0x9A879FA0, 0x68EC1CA3, 0x7BBCEF57, 0x89D76C54,
-  0x5D1D08BF, 0xAF768BBC, 0xBC267848, 0x4E4DFB4B,
-  0x20BD8EDE, 0xD2D60DDD, 0xC186FE29, 0x33ED7D2A,
-  0xE72719C1, 0x154C9AC2, 0x061C6936, 0xF477EA35,
-  0xAA64D611, 0x580F5512, 0x4B5FA6E6, 0xB93425E5,
-  0x6DFE410E, 0x9F95C20D, 0x8CC531F9, 0x7EAEB2FA,
-  0x30E349B1, 0xC288CAB2, 0xD1D83946, 0x23B3BA45,
-  0xF779DEAE, 0x05125DAD, 0x1642AE59, 0xE4292D5A,
-  0xBA3A117E, 0x4851927D, 0x5B016189, 0xA96AE28A,
-  0x7DA08661, 0x8FCB0562, 0x9C9BF696, 0x6EF07595,
-  0x417B1DBC, 0xB3109EBF, 0xA0406D4B, 0x522BEE48,
-  0x86E18AA3, 0x748A09A0, 0x67DAFA54, 0x95B17957,
-  0xCBA24573, 0x39C9C670, 0x2A993584, 0xD8F2B687,
-  0x0C38D26C, 0xFE53516F, 0xED03A29B, 0x1F682198,
-  0x5125DAD3, 0xA34E59D0, 0xB01EAA24, 0x42752927,
-  0x96BF4DCC, 0x64D4CECF, 0x77843D3B, 0x85EFBE38,
-  0xDBFC821C, 0x2997011F, 0x3AC7F2EB, 0xC8AC71E8,
-  0x1C661503, 0xEE0D9600, 0xFD5D65F4, 0x0F36E6F7,
-  0x61C69362, 0x93AD1061, 0x80FDE395, 0x72966096,
-  0xA65C047D, 0x5437877E, 0x4767748A, 0xB50CF789,
-  0xEB1FCBAD, 0x197448AE, 0x0A24BB5A, 0xF84F3859,
-  0x2C855CB2, 0xDEEEDFB1, 0xCDBE2C45, 0x3FD5AF46,
-  0x7198540D, 0x83F3D70E, 0x90A324FA, 0x62C8A7F9,
-  0xB602C312, 0x44694011, 0x5739B3E5, 0xA55230E6,
-  0xFB410CC2, 0x092A8FC1, 0x1A7A7C35, 0xE811FF36,
-  0x3CDB9BDD, 0xCEB018DE, 0xDDE0EB2A, 0x2F8B6829,
-  0x82F63B78, 0x709DB87B, 0x63CD4B8F, 0x91A6C88C,
-  0x456CAC67, 0xB7072F64, 0xA457DC90, 0x563C5F93,
-  0x082F63B7, 0xFA44E0B4, 0xE9141340, 0x1B7F9043,
-  0xCFB5F4A8, 0x3DDE77AB, 0x2E8E845F, 0xDCE5075C,
-  0x92A8FC17, 0x60C37F14, 0x73938CE0, 0x81F80FE3,
-  0x55326B08, 0xA759E80B, 0xB4091BFF, 0x466298FC,
-  0x1871A4D8, 0xEA1A27DB, 0xF94AD42F, 0x0B21572C,
-  0xDFEB33C7, 0x2D80B0C4, 0x3ED04330, 0xCCBBC033,
-  0xA24BB5A6, 0x502036A5, 0x4370C551, 0xB11B4652,
-  0x65D122B9, 0x97BAA1BA, 0x84EA524E, 0x7681D14D,
-  0x2892ED69, 0xDAF96E6A, 0xC9A99D9E, 0x3BC21E9D,
-  0xEF087A76, 0x1D63F975, 0x0E330A81, 0xFC588982,
-  0xB21572C9, 0x407EF1CA, 0x532E023E, 0xA145813D,
-  0x758FE5D6, 0x87E466D5, 0x94B49521, 0x66DF1622,
-  0x38CC2A06, 0xCAA7A905, 0xD9F75AF1, 0x2B9CD9F2,
-  0xFF56BD19, 0x0D3D3E1A, 0x1E6DCDEE, 0xEC064EED,
-  0xC38D26C4, 0x31E6A5C7, 0x22B65633, 0xD0DDD530,
-  0x0417B1DB, 0xF67C32D8, 0xE52CC12C, 0x1747422F,
-  0x49547E0B, 0xBB3FFD08, 0xA86F0EFC, 0x5A048DFF,
-  0x8ECEE914, 0x7CA56A17, 0x6FF599E3, 0x9D9E1AE0,
-  0xD3D3E1AB, 0x21B862A8, 0x32E8915C, 0xC083125F,
-  0x144976B4, 0xE622F5B7, 0xF5720643, 0x07198540,
-  0x590AB964, 0xAB613A67, 0xB831C993, 0x4A5A4A90,
-  0x9E902E7B, 0x6CFBAD78, 0x7FAB5E8C, 0x8DC0DD8F,
-  0xE330A81A, 0x115B2B19, 0x020BD8ED, 0xF0605BEE,
-  0x24AA3F05, 0xD6C1BC06, 0xC5914FF2, 0x37FACCF1,
-  0x69E9F0D5, 0x9B8273D6, 0x88D28022, 0x7AB90321,
-  0xAE7367CA, 0x5C18E4C9, 0x4F48173D, 0xBD23943E,
-  0xF36E6F75, 0x0105EC76, 0x12551F82, 0xE03E9C81,
-  0x34F4F86A, 0xC69F7B69, 0xD5CF889D, 0x27A40B9E,
-  0x79B737BA, 0x8BDCB4B9, 0x988C474D, 0x6AE7C44E,
-  0xBE2DA0A5, 0x4C4623A6, 0x5F16D052, 0xAD7D5351 }
-);
-
-local function CRC8(crc, d)
-  return bit.bxor(bit.rshift(crc, 8), crc32c_table[bit.band(0xff, bit.bxor(crc, d))])
-end
-
-local function CRC16(crc, d)
-  crc = CRC8(crc, bit.band(d, 0xFF))
-  d = bit.rshift(d, 8)
-  crc = CRC8(crc, bit.band(d, 0xFF))
-  return crc
-end
-
-local function string_crc(str, crc)
-  for i=1,#str do
-    -- print("S", i, string.byte(str, i), string.char(string.byte(str, i)))
-    crc = CRC8(crc, string.byte(str, i))
-  end
-  return crc
-end
-
-local tokens = {
-  { ["match"] =' ', ["act"]             = { }  },
-  { ["match"] ='\n', ["act"]             = { }  },
-  { ["match"] ="manual_endian", ["act"]  = { "NODE_MANUAL_ENDIAN", "MANUAL_ENDIAN",    276 } },
-  { ["match"] ="define", ["act"]         = { "NODE_DEFINE",        "DEFINE",           267 } },
-  { ["match"] ="dont_trace", ["act"]     = { "NODE_DONT_TRACE",    "DONT_TRACE",       279 } },
-  { ["match"] ="f64", ["act"]            = { "NODE_F64",           "PRIMTYPE",         string_crc } },
-  { ["match"] ="i16", ["act"]            = { "NODE_I16",           "PRIMTYPE",         string_crc } },
-  { ["match"] ="i32", ["act"]            = { "NODE_I32",           "PRIMTYPE",         string_crc } },
-  { ["match"] ="i64", ["act"]            = { "NODE_I64",           "PRIMTYPE",         string_crc } },
-  { ["match"] ="i8", ["act"]             = { "NODE_I8",            "PRIMTYPE",         string_crc } },
-  { ["match"] ="manual_print", ["act"]   = { "NODE_MANUAL_PRINT",  "MANUAL_PRINT",     275 } },
-  { ["match"] ="noversion", ["act"]      = { "NODE_NOVERSION",     "NOVERSION",        274 } },
-  { ["match"] ="packed", ["act"]         = { "NODE_PACKED",        "TPACKED",          266 } },
-  { ["match"] ="typeonly", ["act"]       = { "NODE_TYPEONLY",      "TYPEONLY",         278 } },
-  { ["match"] ="u16", ["act"]            = { "NODE_U16",           "PRIMTYPE",         string_crc } },
-  { ["match"] ="u32", ["act"]            = { "NODE_U32",           "PRIMTYPE",         string_crc } },
-  { ["match"] ="u64", ["act"]            = { "NODE_U64",           "PRIMTYPE",         string_crc } },
-  { ["match"] ="u8", ["act"]             = { "NODE_U8",            "PRIMTYPE",         string_crc } },
-  { ["match"] ="union", ["act"]          = { "NODE_UNION",         "UNION",            271 } },
-  { ["match"] ="uword", ["act"]          = { "NODE_UWORD",         "PRIMTYPE",         string_crc } },
-  { ["match"] ="%(", ["act"]             = { "NODE_LPAR",          "LPAR",             259 } },
-  { ["match"] ="%)", ["act"]             = { "NODE_RPAR",          "RPAR",             258 } },
-  { ["match"] =";", ["act"]              = { "NODE_SEMI",          "SEMI",             260 } },
-  { ["match"] ="%[", ["act"]             = { "NODE_LBRACK",        "LBRACK",           261 } },
-  { ["match"] ="%]", ["act"]             = { "NODE_RBRACK",        "RBRACK",           262 } },
-  { ["match"] ="%{", ["act"]             = { "NODE_LCURLY",        "LCURLY",           268 } },
-  { ["match"] ="%}", ["act"]             = { "NODE_RCURLY",        "RCURLY",           269 } },
-  { ["match"] ='%b""', ["act"]           = { "NODE_STRING",        "STRING",           string_crc } },
-  { ["match"] ='%b@@', ["act"]           = { "NODE_HELPER",        "HELPER_STRING",    string_crc } },
-  -- TODO: \ must be consumed
-  { ["match"] ='[_a-zA-Z][_a-zA-Z0-9]*',
-                       ["act"]           = { "NODE_NAME",          "NAME",             string_crc } },
-  { ["match"] ='[0-9]+', ["act"]         = { "NODE_NUMBER",        "NUMBER",           string_crc } },
-  { ["match"] ='#[^\n]+', ["act"]            = { "NODE_PRAGMA",        "PRAGMA",           nil } },
-}
-
-
-function vpp.crc_version_string(data)
-  local input_crc = 0
-  -- Get rid of comments
-  data = data:gsub("/%*.-%*/", "")
-  data = data:gsub("//[^\n]+", "")
-  -- print(data)
-  idx = 1
-  while (true) do
-    local matched = nil
-    for k, v in ipairs(tokens) do
-      if not matched then
-        local x, y, cap = string.find(data, v["match"], idx)
-        if x == idx then
-          matched = { ["node"] = v["act"], ["x"] = x, ["y"] = y, ["cap"] = cap, ["chars"] = string.sub(data, x, y)  }
-          -- print(k, v, x, y, cap, matched.chars, matched.node[0] )
-        end
-      end
-    end
-    if matched then
-      idx = idx + (matched.y - matched.x + 1)
-      if matched.node[1] then
-        local act = matched.node[3]
-        if type(act) == "function" then
-          input_crc = act(matched.chars, input_crc)
-        elseif type(act) == "number" then
-          input_crc = CRC16(input_crc, act)
-        end
-        -- print(vpp.dump(matched))
-      end
-    else
-      -- print("NOT MATCHED!")
-      local crc = CRC16(input_crc, 0xFFFFFFFF)
-      return string.sub(string.format("%x", crc), -8)
-    end
-  end
-end
-
-
-function vpp.dump(o)
-   if type(o) == 'table' then
-      local s = '{ '
-      for k,v in pairs(o) do
-         if type(k) ~= 'number' then k = '"'..k..'"' end
-         s = s .. '['..k..'] = ' .. vpp.dump(v) .. ','
-      end
-      return s .. '} '
-   else
-      return tostring(o)
-   end
-end
-
-function vpp.hex_dump(buf)
-  local ret = {}
-  for i=1,math.ceil(#buf/16) * 16 do
-    if (i-1) % 16 == 0 then table.insert(ret, string.format('%08X  ', i-1)) end
-    table.insert(ret, ( i > #buf and '   ' or string.format('%02X ', buf:byte(i)) ))
-    if i %  8 == 0 then table.insert(ret, ' ') end
-    if i % 16 == 0 then table.insert(ret, buf:sub(i-16+1, i):gsub('%c','.')..'\n' ) end
-  end
-  return table.concat(ret)
-end
-
-
-function vpp.c_str(text_in)
-  local text = text_in -- \000 will be helpfully added by ffi.copy
-  local c_str = ffi.new("char[?]", #text+1)
-  ffi.copy(c_str, text)
-  return c_str
-end
-
-
-function vpp.init(vpp, args)
-  local pneum_api = args.pneum_api or [[
- int cough_pneum_attach(char *pneum_path, char *cough_path);
- int pneum_connect(char *name, char *chroot_prefix, void *cb);
- int pneum_disconnect(void);
- int pneum_read(char **data, int *l);
- int pneum_write(char *data, int len);
- void pneum_free(char *data);
- uint32_t pneum_get_msg_index(unsigned char * name);
-]]
-
-  vpp.pneum_path = args.pneum_path
-  ffi.cdef(pneum_api)
-  local init_res = 0
-  vpp.pneum = ffi.load(vpp.pneum_path)
-  if (init_res < 0) then
-    return nil
-  end
-
-  vpp.next_msg_num = 1
-  vpp.msg_name_to_number = {}
-  vpp.msg_name_to_fields = {}
-  vpp.msg_number_to_name = {}
-  vpp.msg_number_to_type = {}
-  vpp.msg_number_to_pointer_type = {}
-  vpp.msg_name_to_crc = {}
-  vpp.c_type_to_fields = {}
-  vpp.events = {}
-  vpp.plugin_version = {}
-  vpp.is_connected = false
-
-
-  vpp.t_lua2c = {}
-  vpp.t_c2lua = {}
-  vpp.t_lua2c["u8"] = function(c_type, src, dst_c_ptr)
-    if type(src) == "string" then
-      -- ffi.copy adds a zero byte at the end. Grrr.
-      -- ffi.copy(dst_c_ptr, src)
-      ffi.C.memcpy(dst_c_ptr, vpp.c_str(src), #src)
-      return(#src)
-    elseif type(src) == "table" then
-      for i,v in ipairs(src) do
-        ffi.cast("u8 *", dst_c_ptr)[i-1] = v
-      end
-      return(#src)
-    else
-      return 1, src -- ffi.cast("u8", src)
-    end
-  end
-  vpp.t_c2lua["u8"] = function(c_type, src_ptr, src_len)
-    if src_len then
-      return ffi.string(src_ptr, src_len)
-    else
-      return (tonumber(src_ptr))
-    end
-  end
-
-  vpp.t_lua2c["u16"] = function(c_type, src, dst_c_ptr)
-    if type(src) == "table" then
-      for i,v in ipairs(src) do
-        ffi.cast("u16 *", dst_c_ptr)[i-1] = ffi.C.htons(v)
-      end
-      return(2 * #src)
-    else
-      return 2, (ffi.C.htons(src))
-    end
-  end
-  vpp.t_c2lua["u16"] = function(c_type, src_ptr, src_len)
-    if src_len then
-      local out = {}
-      for i = 0,src_len-1 do
-        out[i+1] = tonumber(ffi.C.ntohs(src_ptr[i]))
-      end
-      return out
-    else
-      return (tonumber(ffi.C.ntohs(src_ptr)))
-    end
-  end
-
-  vpp.t_lua2c["u32"] = function(c_type, src, dst_c_ptr)
-    if type(src) == "table" then
-      for i,v in ipairs(src) do
-        ffi.cast("u32 *", dst_c_ptr)[i-1] = ffi.C.htonl(v)
-      end
-      return(4 * #src)
-    else
-      return 4, (ffi.C.htonl(src))
-    end
-  end
-  vpp.t_c2lua["u32"] = function(c_type, src_ptr, src_len)
-    if src_len then
-      local out = {}
-      for i = 0,src_len-1 do
-        out[i+1] = tonumber(ffi.C.ntohl(src_ptr[i]))
-      end
-      return out
-    else
-      return (tonumber(ffi.C.ntohl(src_ptr)))
-    end
-  end
-  vpp.t_lua2c["i32"] = function(c_type, src, dst_c_ptr)
-    if type(src) == "table" then
-      for i,v in ipairs(src) do
-        ffi.cast("i32 *", dst_c_ptr)[i-1] = ffi.C.htonl(v)
-      end
-      return(4 * #src)
-    else
-      return 4, (ffi.C.htonl(src))
-    end
-  end
-  vpp.t_c2lua["i32"] = function(c_type, src_ptr, src_len)
-    local ntohl = function(src)
-      local u32val = ffi.cast("u32", src)
-      local ntohlval = (ffi.C.ntohl(u32val))
-      local out = tonumber(ffi.cast("i32", ntohlval + 0LL))
-      return out
-    end
-    if src_len then
-      local out = {}
-      for i = 0,src_len-1 do
-        out[i+1] = tonumber(ntohl(src_ptr[i]))
-      end
-    else
-      return (tonumber(ntohl(src_ptr)))
-    end
-  end
-
-  vpp.t_lua2c["u64"] = function(c_type, src, dst_c_ptr)
-    if type(src) == "table" then
-      for i,v in ipairs(src) do
-        ffi.cast("u64 *", dst_c_ptr)[i-1] = v --- FIXME ENDIAN
-      end
-      return(8 * #src)
-    else
-      return 8, ffi.cast("u64", src) --- FIXME ENDIAN
-    end
-  end
-  vpp.t_c2lua["u64"] = function(c_type, src_ptr, src_len)
-    if src_len then
-      local out = {}
-      for i = 0,src_len-1 do
-        out[i+1] = tonumber(src_ptr[i]) -- FIXME ENDIAN
-      end
-      return out
-    else
-      return (tonumber(src_ptr)) --FIXME ENDIAN
-    end
-  end
-
-
-
-
-  vpp.t_lua2c["__MSG__"] = function(c_type, src, dst_c_ptr)
-    local dst = ffi.cast(c_type .. " *", dst_c_ptr)
-    local additional_len = 0
-    local fields_info = vpp.c_type_to_fields[c_type]
-    -- print("__MSG__ type: " .. tostring(c_type))
-    ffi.C.memset(dst_c_ptr, 0, ffi.sizeof(dst[0]))
-    -- print(vpp.dump(fields_info))
-    -- print(vpp.dump(src))
-    for k,v in pairs(src) do
-      local field = fields_info[k]
-      if not field then
-        print("ERROR: field " .. tostring(k) .. " in message " .. tostring(c_type) .. " is unknown")
-      end
-      local lua2c = vpp.t_lua2c[field.c_type]
-      -- print("__MSG__ field " .. tostring(k) .. " : " .. vpp.dump(field))
-      -- if the field is not an array type, try to coerce the argument to a number
-      if not field.array and type(v) == "string" then
-        v = tonumber(v)
-      end
-      if not lua2c then
-        print("__MSG__ " .. tostring(c_type) .. " t_lua2c: can not store field " .. field.name ..
-              " type " .. field.c_type .. " dst " .. tostring(dst[k]))
-        return 0
-      end
-      local len = 0
-      local val = nil
-      if field.array and (type(v) == "table") then
-        -- print("NTFY: field " .. tostring(k) .. " in message " .. tostring(c_type) .. " is an array")
-        for field_i, field_v in ipairs(v) do
-          -- print("NTFY: setting member#" .. tostring(field_i) .. " to value " .. vpp.dump(field_v))
-          local field_len, field_val = lua2c(field.c_type, field_v, dst[k][field_i-1])
-          len = len + field_len
-        end
-      else
-        len, val = lua2c(field.c_type, v, dst[k])
-      end
-      if not field.array then
-        dst[k] = val
-      else
-        if 0 == field.array then
-          additional_len = additional_len + len
-          -- print("Adding " .. tostring(len) .. " bytes due to field " .. tostring(field.name))
-          -- If there is a variable storing the length
-          -- and the input table does not set it, do magic
-          if field.array_size and not src[field.array_size] then
-            local size_field = fields_info[field.array_size]
-            if size_field then
-              dst[field.array_size] = vpp.t_c2lua[size_field.c_type](size_field.c_type, len)
-            end
-          end
-        end
-      end
-      -- print("Full message:\n" .. vpp.hex_dump(ffi.string(ffi.cast('void *', req_store_cache), 64)))
-    end
-    return (ffi.sizeof(dst[0])+additional_len)
-  end
-
-  vpp.t_c2lua["__MSG__"] = function(c_type, src_ptr, src_len)
-    local out = {}
-    local reply_typed_ptr = ffi.cast(c_type .. " *", src_ptr)
-    local field_desc = vpp.c_type_to_fields[c_type]
-    if src_len then
-      for i = 0,src_len-1 do
-        out[i+1] = vpp.t_c2lua[c_type](c_type, src_ptr[i])
-      end
-      return out
-    end
-
-    for k, v in pairs(field_desc) do
-      local v_c2lua = vpp.t_c2lua[v.c_type]
-      if v_c2lua then
-        local len = v.array
-        -- print(dump(v))
-        if len then
-          local len_field_name = k .. "_length"
-          local len_field = field_desc[len_field_name]
-          if (len_field) then
-            local real_len = vpp.t_c2lua[len_field.c_type](len_field.c_type, reply_typed_ptr[len_field_name])
-            out[k] =  v_c2lua(v.c_type, reply_typed_ptr[k], real_len)
-          elseif len == 0 then
-            -- check if len = 0, then must be a field which contains the size
-            len_field =  field_desc[v.array_size]
-            local real_len = vpp.t_c2lua[len_field.c_type](len_field.c_type, reply_typed_ptr[v.array_size])
-            -- print("REAL length: " .. vpp.dump(v) .. " : " .. tostring(real_len))
-            out[k] = v_c2lua(v.c_type, reply_typed_ptr[k], real_len)
-          else
-            -- alas, just stuff the entire array
-            out[k] = v_c2lua(v.c_type, reply_typed_ptr[k], len)
-          end
-        else
-          out[k] =  v_c2lua(v.c_type, reply_typed_ptr[k])
-        end
-      else
-        out[k] = ""
-      end
-      -- print(k, out[k])
-    end
-    return out
-  end
-
-  return vpp
-end
-
-function vpp.resolve_message_number(msgname)
-  local name = msgname .. "_" .. vpp.msg_name_to_crc[msgname]
-  local idx = vpp.pneum.pneum_get_msg_index(vpp.c_str(name))
-  if vpp.debug_dump then
-    print("Index for " .. tostring(name) .. " is " .. tostring(idx))
-  end
-  vpp.msg_name_to_number[msgname] = idx
-  vpp.msg_number_to_name[idx] = msgname
-  vpp.msg_number_to_type[idx] = "vl_api_" .. msgname .. "_t"
-  vpp.msg_number_to_pointer_type[idx] = vpp.msg_number_to_type[idx] .. " *"
-  ffi.cdef("\n\n enum { vl_msg_" .. msgname .. " = " .. idx .. " };\n\n")
-end
-
-function vpp.connect(vpp, client_name)
-    local name = "lua_client"
-    if client_name then
-      name = client_name
-    end
-    local ret = vpp.pneum.pneum_connect(vpp.c_str(client_name), nil, nil)
-    if tonumber(ret) == 0 then
-      vpp.is_connected = true
-    end
-    for k, v in pairs(vpp.msg_name_to_number) do
-      vpp.resolve_message_number(k)
-    end
-  end
-
-function vpp.disconnect(vpp)
-    vpp.pneum.pneum_disconnect()
-  end
-
-function vpp.json_api(vpp, path, plugin_name)
-    -- print("Consuming the VPP api from "..path)
-    local ffii = {}
-    local f = io.open(path, "r")
-    if not f then
-      print("Could not open " .. path)
-      return nil
-    end
-    local data = f:read("*all")
-    local json = json.parse(data)
-    if not (json.types or json.messages) then
-      print("Can not parse " .. path)
-      return nil
-    end
-
-    local all_types = {}
-
-    for i, v in ipairs(json.types) do
-      table.insert(all_types, { typeonly = 1, desc = v })
-    end
-    for i, v in ipairs(json.messages) do
-      table.insert(all_types, { typeonly = 0, desc = v })
-    end
-    for i, v in ipairs(all_types) do
-      local typeonly = v.typeonly
-      local name = v.desc[1]
-      local c_type = "vl_api_" .. name .. "_t"
-
-      local fields = {}
-      -- vpp.msg_name_to_fields[name] = fields
-      -- print("CTYPE " .. c_type)
-      vpp.c_type_to_fields[c_type] = fields
-      vpp.t_lua2c[c_type] = vpp.t_lua2c["__MSG__"]
-      vpp.t_c2lua[c_type] = vpp.t_c2lua["__MSG__"]
-
-      local cdef = { "\n\n#pragma pack(1)\ntypedef struct _vl_api_", name, " {\n" }
-      for ii, vv in ipairs(v.desc) do
-        if type(vv) == "table" then
-          if vv.crc then
-            vpp.msg_name_to_crc[name] = string.sub(vv.crc, 3) -- strip the leading 0x
-          else
-            local fieldtype = vv[1]
-            local fieldname = vv[2]
-            local fieldcount = vv[3]
-            local fieldcountvar = vv[4]
-            local fieldrec = { name = fieldname, c_type = fieldtype, array = fieldcount, array_size = fieldcountvar }
-            if fieldcount then
-              table.insert(cdef, "  " .. fieldtype .. " " .. fieldname .. "[" .. fieldcount .. "];\n")
-              if fieldtype == "u8" then
-                -- any array of bytes is treated as a string
-              elseif vpp.t_lua2c[fieldtype] then
-                -- print("Array of " .. fieldtype .. " is ok!")
-              else
-                print("Unknown array type: ", name,  " : " , fieldname, " : ", fieldtype, ":", fieldcount, ":", fieldcountvar)
-              end
-            else
-              table.insert(cdef, "  " .. fieldtype .. " " .. fieldname .. ";\n")
-            end
-            fields[fieldname] = fieldrec
-          end
-        end
-      end
-
-      table.insert(cdef, "} vl_api_" .. name .. "_t;")
-      table.insert(ffii, table.concat(cdef))
-
-      if typeonly == 0 then
-        -- we will want to resolve this later
-        if vpp.debug_dump then
-          print("Remember to resolve " .. name)
-        end
-        vpp.msg_name_to_number[name] = -1
-        if vpp.is_connected then
-          vpp.resolve_message_number(name)
-        end
-      end
-
-    end
-    local cdef_full = table.concat(ffii)
-    ffi.cdef(cdef_full)
-end
-
-function vpp.consume_api(vpp, path, plugin_name)
-    -- print("Consuming the VPP api from "..path)
-    local ffii = {}
-    local f = io.open(path, "r")
-    if not f then
-      print("Could not open " .. path)
-      return nil
-    end
-    local data = f:read("*all")
-    -- Remove all C comments
-    data = data:gsub("/%*.-%*/", "")
-    if vpp.is_connected and not plugin_name then
-      print(path .. ": must specify plugin name!")
-      return
-    end
-    if plugin_name then
-      vpp.plugin_version[plugin_name] = vpp.crc_version_string(data)
-      local full_plugin_name = plugin_name .. "_" .. vpp.plugin_version[plugin_name]
-      local reply = vpp:api_call("get_first_msg_id", { name = full_plugin_name } )
-      vpp.next_msg_num = tonumber(reply[1].first_msg_id)
-      print("Plugin " .. full_plugin_name .. " first message is " .. tostring(vpp.next_msg_num))
-    end
-    -- print ("data len: ", #data)
-    data = data:gsub("\n(.-)(%S+)%s*{([^}]*)}", function (preamble, name, members)
-      local _, typeonly = preamble:gsub("typeonly", "")
-      local maybe_msg_id_field = { [0] = "u16 _vl_msg_id;", "" }
-      local onedef = "\n\n#pragma pack(1)\ntypedef struct _vl_api_"..name.. " {\n" ..
-	   -- "   u16 _vl_msg_id;" ..
-           maybe_msg_id_field[typeonly] ..
-	   members:gsub("%[[a-zA-Z_]+]", "[0]") ..
-	   "} vl_api_" .. name .. "_t;"
-
-      local c_type = "vl_api_" .. name .. "_t"
-
-      local fields = {}
-      -- vpp.msg_name_to_fields[name] = fields
-      -- print("CTYPE " .. c_type)
-      vpp.c_type_to_fields[c_type] = fields
-      vpp.t_lua2c[c_type] = vpp.t_lua2c["__MSG__"]
-      vpp.t_c2lua[c_type] = vpp.t_c2lua["__MSG__"]
-      local mirec = { name = "_vl_msg_id", c_type = "u16", array = nil, array_size = nil }
-      if typeonly == 0 then
-        fields[mirec.name] = mirec
-      end
-
-      -- populate the field reflection table for the message
-      -- sets the various type information as well as the accessors for lua<->C conversion
-      members:gsub("(%S+)%s+(%S+);", function (fieldtype, fieldname)
-          local fieldcount = nil
-          local fieldcountvar = nil
-          -- data = data:gsub("%[[a-zA-Z_]+]", "[0]")
-          fieldname = fieldname:gsub("(%b[])", function(cnt)
-              fieldcount = tonumber(cnt:sub(2, -2));
-              if not fieldcount then
-                fieldcount = 0
-                fieldcountvar = cnt:sub(2, -2)
-              end
-              return ""
-            end)
-	  local fieldrec = { name = fieldname, c_type = fieldtype, array = fieldcount, array_size = fieldcountvar }
-          if fieldcount then
-            if fieldtype == "u8" then
-              -- any array of bytes is treated as a string
-            elseif vpp.t_lua2c[fieldtype] then
-              -- print("Array of " .. fieldtype .. " is ok!")
-            else
-              print("Unknown array type: ", name,  " : " , fieldname, " : ", fieldtype, ":", fieldcount, ":", fieldcountvar)
-            end
-          end
-	  fields[fieldname] = fieldrec
-	end)
-
-      -- print(dump(fields))
-
-      if typeonly == 0 then
-	local this_message_number = vpp.next_msg_num
-	vpp.next_msg_num = vpp.next_msg_num + 1
-	vpp.msg_name_to_number[name] = this_message_number
-	vpp.msg_number_to_name[this_message_number] = name
-	vpp.msg_number_to_type[this_message_number] = "vl_api_" .. name .. "_t"
-	vpp.msg_number_to_pointer_type[this_message_number] = vpp.msg_number_to_type[this_message_number] .. " *"
-	onedef = onedef .. "\n\n enum { vl_msg_" .. name .. " = " .. this_message_number .. " };\n\n"
-      end
-      table.insert(ffii, onedef);
-      return "";
-      end)
-    local cdef = table.concat(ffii)
-    -- print(cdef)
-    ffi.cdef(cdef)
-  end
-
-
-function vpp.lua2c(vpp, c_type, src, dst_c_ptr)
-  -- returns the number of bytes written to memory pointed by dst
-  local lua2c = vpp.t_lua2c[c_type]
-  if lua2c then
-    return(lua2c(c_type, src, dst_c_ptr))
-  else
-    print("vpp.lua2c: do not know how to store type " .. tostring(c_type))
-    local x = "a" .. nil
-    return 0
-  end
-end
-
-function vpp.c2lua(vpp, c_type, src_ptr, src_len)
-  -- returns the lua data structure
-  local c2lua = vpp.t_c2lua[c_type]
-  if c2lua then
-    return(c2lua(c_type, src_ptr, src_len))
-  else
-    print("vpp.c2lua: do not know how to load type " .. c_type)
-    return nil
-  end
-end
-
-local req_store_cache = ffi.new("vl_api_opaque_message_t[1]")
-
-function vpp.api_write(vpp, api_name, req_table)
-    local msg_num = vpp.msg_name_to_number[api_name]
-    if not msg_num then
-      print ("API call "..api_name.." is not known")
-      return nil
-    end
-
-    if not req_table then
-      req_table = {}
-    end
-    req_table._vl_msg_id = msg_num
-
-    local packed_len = vpp:lua2c(vpp.msg_number_to_type[msg_num], req_table, req_store_cache)
-    if vpp.debug_dump then
-      print("Write Message length: " .. tostring(packed_len) .. "\n" .. vpp.hex_dump(ffi.string(ffi.cast('void *', req_store_cache), packed_len)))
-    end
-
-    res = vpp.pneum.pneum_write(ffi.cast('void *', req_store_cache), packed_len)
-    return res
-  end
-
-local rep_store_cache = ffi.new("vl_api_opaque_message_t *[1]")
-local rep_len_cache = ffi.new("int[1]")
-
-function vpp.api_read(vpp)
-    local rep_type = "vl_api_opaque_message_t"
-    local rep = rep_store_cache
-    local replen = rep_len_cache
-    res = vpp.pneum.pneum_read(ffi.cast("void *", rep), replen)
-    if vpp.debug_dump then
-      print("Read Message length: " .. tostring(replen[0]) .. "\n" .. vpp.hex_dump(ffi.string(ffi.cast('void *', rep[0]), replen[0])))
-    end
-
-    local reply_msg_num = ffi.C.ntohs(rep[0]._vl_msg_id)
-    local reply_msg_name = vpp.msg_number_to_name[reply_msg_num]
-
-    local reply_typed_ptr = ffi.cast(vpp.msg_number_to_pointer_type[reply_msg_num], rep[0])
-    local out = vpp:c2lua(vpp.msg_number_to_type[reply_msg_num], rep[0], nil, replen[0])
-    if type(out) == "table" then
-      out["luaapi_message_name"] = reply_msg_name
-    end
-
-    vpp.pneum.pneum_free(ffi.cast('void *',rep[0]))
-
-    return reply_msg_name, out
-  end
-
-function vpp.api_call(vpp, api_name, req_table, options_in)
-    local msg_num = vpp.msg_name_to_number[api_name]
-    local end_message_name = api_name .."_reply"
-    local replies = {}
-    local cstruct = ""
-    local options = options_in or {}
-    if msg_num then
-      if vpp.debug_dump then
-        print("Message #" .. tostring(msg_num) .. " for name " .. tostring(api_name))
-      end
-      vpp:api_write(api_name, req_table)
-      if not vpp.msg_name_to_number[end_message_name] or options.force_ping then
-        end_message_name = "control_ping_reply"
-        vpp:api_write("control_ping")
-      end
-      repeat
-        reply_message_name, reply = vpp:api_read()
-        if reply and not reply.context then
-          -- there may be async events inbetween
-          table.insert(vpp.events, reply)
-        else
-          if reply_message_name ~= "control_ping_reply" then
-            -- do not insert the control ping encapsulation
-            table.insert(replies, reply)
-          end
-        end
-        -- print(reply)
-      until reply_message_name == end_message_name
-    else
-      print(api_name .. " is an unknown API call")
-      return nil
-    end
-    return replies
-  end
-
-return vpp
-- 
cgit 1.2.3-korg


From 0be5ec304d2d4cfc6faecbb519ef165d9ab5d34e Mon Sep 17 00:00:00 2001
From: Damjan Marion 
Date: Wed, 28 Dec 2016 17:51:56 +0100
Subject: Do not require external vppapigen when not cross-compiling

Change-Id: I80b8348ed4efd53d292c37a1ff69c13ee4741986
Signed-off-by: Damjan Marion 
---
 build-data/platforms.mk |  2 +-
 build-root/rpm/vpp.spec |  1 -
 src/Makefile.am         |  1 -
 src/configure.ac        | 16 ++++++++++++++++
 src/plugins/Makefile.am |  1 -
 src/suffix-rules.mk     | 12 ++++++------
 src/vppapigen.am        |  9 ++++++---
 7 files changed, 29 insertions(+), 13 deletions(-)

(limited to 'src/Makefile.am')

diff --git a/build-data/platforms.mk b/build-data/platforms.mk
index 6f21b6dc..e192155b 100644
--- a/build-data/platforms.mk
+++ b/build-data/platforms.mk
@@ -76,7 +76,7 @@ install-deb: $(patsubst %,%-find-source,$(ROOT_PACKAGES))
 	   >> deb/debian/vpp.install ;					\
 									\
 	: dev package needs a couple of additions ;			\
-	echo ../build-tool-native/tools/vppapigen /usr/bin		\
+	echo ../$(INSTALL_PREFIX)$(ARCH)/vpp/bin/vppapigen /usr/bin	\
 	   >> deb/debian/vpp-dev.install ;				\
 	echo ../../src/vpp-api/java/jvpp/gen/jvpp_gen.py /usr/bin	\
 	   >> deb/debian/vpp-dev.install ;				\
diff --git a/build-root/rpm/vpp.spec b/build-root/rpm/vpp.spec
index 149ac51c..c3c0d927 100644
--- a/build-root/rpm/vpp.spec
+++ b/build-root/rpm/vpp.spec
@@ -113,7 +113,6 @@ groupadd -f -r vpp
 mkdir -p -m755 %{buildroot}%{_bindir}
 mkdir -p -m755 %{buildroot}%{_unitdir}
 install -p -m 755 %{_mu_build_dir}/%{_vpp_install_dir}/*/bin/* %{buildroot}%{_bindir}
-install -p -m 755 %{_mu_build_dir}/%{_vpp_build_dir}/tools/vppapigen %{buildroot}%{_bindir}
 
 # api
 mkdir -p -m755 %{buildroot}/usr/share/vpp/api
diff --git a/src/Makefile.am b/src/Makefile.am
index 0fc437a4..bba90eae 100644
--- a/src/Makefile.am
+++ b/src/Makefile.am
@@ -102,7 +102,6 @@ api_DATA = \
 	$(patsubst %.api,%.api.json,$(API_FILES))
 
 BUILT_SOURCES += \
-	$(patsubst %.api,%.api.json,$(API_FILES)) \
 	$(patsubst %.api,%.api.h,$(API_FILES))
 
 endif # if ENABLE_VLIB
diff --git a/src/configure.ac b/src/configure.ac
index 4ed55705..eb380d8b 100644
--- a/src/configure.ac
+++ b/src/configure.ac
@@ -11,6 +11,8 @@ AM_PROG_AS
 AM_PROG_LIBTOOL
 AC_PROG_YACC
 
+AM_CONDITIONAL([CROSSCOMPILE], [test "$cross_compiling" == "yes"])
+
 ###############################################################################
 # Macros
 ###############################################################################
@@ -172,6 +174,20 @@ AM_COND_IF([ENABLE_G2],
   PKG_CHECK_MODULES(g2, gtk+-2.0)
 ])
 
+# If cross-compiling, we need external vppapigen and we cannot continue without it
+# For native builds, we just set dependency on vpppaigen binary in top_builddir
+AM_COND_IF([CROSSCOMPILE],
+[
+  AC_PATH_PROG([VPPAPIGEN], [vppapigen], [no])
+  if test "$VPPAPIGEN" = "no"; then
+    AC_MSG_ERROR([Externaly built vppapigen is needed when cross-compiling...])
+  fi
+],[
+  VPPAPIGEN=\$\(top_builddir\)/vppapigen
+])
+AC_SUBST([VPPAPIGEN])
+
+
 ###############################################################################
 # JAVA
 ###############################################################################
diff --git a/src/plugins/Makefile.am b/src/plugins/Makefile.am
index 987310b7..3e9ab91c 100644
--- a/src/plugins/Makefile.am
+++ b/src/plugins/Makefile.am
@@ -73,6 +73,5 @@ api_DATA = \
 	$(patsubst %.api,%.api.json,$(API_FILES))
 
 BUILT_SOURCES += \
-	$(patsubst %.api,%.api.json,$(API_FILES)) \
 	$(patsubst %.api,%.api.h,$(API_FILES))
 
diff --git a/src/suffix-rules.mk b/src/suffix-rules.mk
index f97b2276..9115c55b 100644
--- a/src/suffix-rules.mk
+++ b/src/suffix-rules.mk
@@ -14,14 +14,14 @@
 # Shared suffix rules
 # Please do not set "SUFFIXES = .api.h .api" here
 
-%.api.h: %.api
+%.api.h: %.api @VPPAPIGEN@
 	@echo "  APIGEN  " $@ ;					\
 	mkdir -p `dirname $@` ;					\
-	$(CC) $(CPPFLAGS) -E -P -C -x c $^			\
-	| vppapigen --input - --output $@ --show-name $@ > /dev/null
+	$(CC) $(CPPFLAGS) -E -P -C -x c $<			\
+	| @VPPAPIGEN@ --input - --output $@ --show-name $@ > /dev/null
 
-%.api.json: %.api
+%.api.json: %.api @VPPAPIGEN@
 	@echo "  JSON API" $@ ;					\
 	mkdir -p `dirname $@` ;					\
-	$(CC) $(CPPFLAGS) -E -P -C -x c $^			\
-	| vppapigen --input - --json $@ > /dev/null
+	$(CC) $(CPPFLAGS) -E -P -C -x c $<			\
+	| @VPPAPIGEN@ --input - --json $@ > /dev/null
diff --git a/src/vppapigen.am b/src/vppapigen.am
index edde339d..3207c83a 100644
--- a/src/vppapigen.am
+++ b/src/vppapigen.am
@@ -13,14 +13,17 @@
 
 bin_PROGRAMS += vppapigen
 
-BUILT_SOURCES += tools/vppapigen/gram.h
+# We cannot rely on BUILT_SOURCES here as other built sources are relying
+# on vppapigen, so make can start compiling lex.c before gram.h is created.
+# This way we introduce new dependency by running C preprocessor.
 
-tools/vppapigen/gram.h: tools/vppapigen/gram.y
+tools/vppapigen/lex_e.c: tools/vppapigen/lex.c tools/vppapigen/gram.y
 	@$(YACC) -d @srcdir@/tools/vppapigen/gram.y
 	@mv y.tab.h tools/vppapigen/gram.h
 	@rm y.tab.c
+	@$(CC) -I. -E -o $@ $<
 
-vppapigen_SOURCES = tools/vppapigen/gram.y tools/vppapigen/lex.c tools/vppapigen/node.c
+vppapigen_SOURCES = tools/vppapigen/gram.y tools/vppapigen/lex_e.c tools/vppapigen/node.c
 vppapigen_LDADD = libvppinfra.la
 vppapigen_LDFLAGS = -static
 
-- 
cgit 1.2.3-korg


From 724f64ccf691e170973bdcdc6c09bfdc7e2ab5e4 Mon Sep 17 00:00:00 2001
From: Damjan Marion 
Date: Wed, 11 Jan 2017 11:11:00 +0100
Subject: Makefile.am cleanup

 - remove unused stuff
 - add --quiet flag to libtool
 - avoid building some tests programs when tests are not enabled

Change-Id: Ie34aeec1a598ad811256a00354f66cfddae9d0b9
Signed-off-by: Damjan Marion 
---
 build-root/emacs-lisp/README           |  2 +-
 build-root/scripts/make-plugin-toolkit | 40 --------------------------
 src/Makefile.am                        |  1 +
 src/plugins/Makefile.am                |  1 +
 src/vnet.am                            |  9 ++----
 src/vnet/plugin/p1.c                   | 52 ----------------------------------
 src/vpp-api/java/Makefile.am           |  1 +
 src/vpp-api/python/Makefile.am         |  5 ++--
 src/vpp.am                             | 36 ++---------------------
 src/vppinfra.am                        |  8 +-----
 10 files changed, 12 insertions(+), 143 deletions(-)
 delete mode 100755 build-root/scripts/make-plugin-toolkit
 delete mode 100644 src/vnet/plugin/p1.c

(limited to 'src/Makefile.am')

diff --git a/build-root/emacs-lisp/README b/build-root/emacs-lisp/README
index 483e1c39..1f09a3fa 100644
--- a/build-root/emacs-lisp/README
+++ b/build-root/emacs-lisp/README
@@ -66,7 +66,7 @@ Or, generate each file individually:
 
    $ mkdir build
    $ cd build
-   $ ../configure --with-plugin-toolkit
+   $ ../configure
    $ make
    $ sudo make install
 
diff --git a/build-root/scripts/make-plugin-toolkit b/build-root/scripts/make-plugin-toolkit
deleted file mode 100755
index e1d6fcfb..00000000
--- a/build-root/scripts/make-plugin-toolkit
+++ /dev/null
@@ -1,40 +0,0 @@
-#!/bin/bash
-
-set -eux
-
-build_tarball() {
-    for dir in vppinfra dpdk svm vlib-api vlib vnet vpp vpp-api-test
-    do
-        tar -C install-$1/$dir/include -cf - . | tar -C $tmp_dir/include -xf -
-    done
-    tar -C ../sample-plugin -cf - . \
-       | tar -C $tmp_dir/src/sample-plugin -xf -
-    cp tools/bin/vppapigen $tmp_dir/tools/bin
-    echo Created by `id -u -n` on `hostname` at `date` >		\
-		$tmp_dir/toolkit-version-stamp
-    cp scripts/vpp-plugin-toolkit-readme $tmp_dir/README
-    tar -C $tmp_dir -zcf $PWD/vpp-plugin-toolkit-$1.tar.gz .
-}
-
-if [ `basename $PWD` != "build-root" ] ; then
-   echo Please run this script from build-root
-   exit 1
-fi   
-
-echo Pull additional tookit repos
-make PLATFORM=vpp sample-plugin-find-source
-
-make PLATFORM=vpp TAG=vpp wipe-all
-echo Build vpp forwarder production package
-make PLATFORM=vpp TAG=vpp strip_sumbols=yes install-packages
-
-tmp_dir="`mktemp -d /tmp/plugin-XXXXXX`"
-trap "rm -rf $tmp_dir" err 
-
-echo Create vpp forwarder production plugin toolkit tarball
-mkdir -p $tmp_dir/tools/bin $tmp_dir/include $tmp_dir/lib64 \
-         $tmp_dir/src/sample-plugin
-build_tarball vpp-native
-rm -rf $tmp_dir
-
-exit 0
diff --git a/src/Makefile.am b/src/Makefile.am
index bba90eae..18a41a15 100644
--- a/src/Makefile.am
+++ b/src/Makefile.am
@@ -17,6 +17,7 @@
 
 AUTOMAKE_OPTIONS = foreign subdir-objects
 ACLOCAL_AMFLAGS = -I m4
+AM_LIBTOOLFLAGS = --quiet
 
 AM_CFLAGS = -Wall
 
diff --git a/src/plugins/Makefile.am b/src/plugins/Makefile.am
index 3e9ab91c..f0c455a5 100644
--- a/src/plugins/Makefile.am
+++ b/src/plugins/Makefile.am
@@ -16,6 +16,7 @@ AUTOMAKE_OPTIONS = foreign subdir-objects
 
 AM_CFLAGS = -Wall -I${top_srcdir} -I${top_builddir}
 AM_LDFLAGS = -module -shared -avoid-version
+AM_LIBTOOLFLAGS = --quiet
 SUFFIXES = .api.h .api .api.json
 API_FILES =
 BUILT_SOURCES =
diff --git a/src/vnet.am b/src/vnet.am
index bca56227..665a16ea 100644
--- a/src/vnet.am
+++ b/src/vnet.am
@@ -11,8 +11,9 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
+lib_LTLIBRARIES += libvnet.la
+
 libvnet_la_SOURCES =
-libvnetplugin_la_SOURCES =
 
 libvnet_la_LIBADD = libvppinfra.la libsvm.la
 libvnet_la_DEPENDENCIES = libvppinfra.la libvlib.la libsvmdb.la libsvm.la libvlibapi.la libvlibmemory.la libvlibmemoryclient.la
@@ -966,15 +967,9 @@ nobase_include_HEADERS +=			\
 # Plugin client library
 ########################################
 
-libvnetplugin_la_SOURCES +=                    \
-  vnet/plugin/p1.c
-
 nobase_include_HEADERS +=                      \
   vnet/plugin/plugin.h
 
-libvnetdir = ${libdir}
-libvnet_LTLIBRARIES = libvnet.la libvnetplugin.la
-
 pcap2pg_SOURCES = 				\
   vnet/unix/pcap2pg.c				\
   vnet/unix/pcap.h
diff --git a/src/vnet/plugin/p1.c b/src/vnet/plugin/p1.c
deleted file mode 100644
index 3102ecce..00000000
--- a/src/vnet/plugin/p1.c
+++ /dev/null
@@ -1,52 +0,0 @@
-/*
- * Copyright (c) 2015 Cisco and/or its affiliates.
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at:
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-/* 
- * This file and in fact the entire directory shouldn't even exist.
- *
- * Unfortunately, various things malfunction when we try to go there. 
- * Plugin DLL's end up with their own copies of critical
- * data structures. No one of these problems would be tough to fix, 
- * but there are quite a number of them.
- */
-
-/* 
- * Make certain that plugin .dll's which reference the following functions
- * can find them...
- */
-
-#if DPDK > 0
-#define foreach_dpdk_plugin_reference		\
-_(rte_calloc)                                   \
-_(rte_free)                                     \
-_(rte_malloc)                                   \
-_(rte_zmalloc)                                  \
-_(rte_malloc_virt2phy)                          \
-_(rte_eal_get_configuration)
-#else
-#define foreach_dpdk_plugin_reference
-#endif
-
-#define _(a) void a (void);
-foreach_dpdk_plugin_reference
-#undef _
-
-void *vnet_library_plugin_references[] =
-  {
-#define _(a) &a,
-    foreach_dpdk_plugin_reference
-#undef _
-  };
-
-void vnet_library_plugin_reference(void) { }
diff --git a/src/vpp-api/java/Makefile.am b/src/vpp-api/java/Makefile.am
index bfb5665a..00ee5946 100644
--- a/src/vpp-api/java/Makefile.am
+++ b/src/vpp-api/java/Makefile.am
@@ -13,6 +13,7 @@
 
 AUTOMAKE_OPTIONS = foreign
 ACLOCAL_AMFLAGS = -I m4
+AM_LIBTOOLFLAGS = --quiet
 
 AM_CFLAGS = -Wall -I${top_srcdir} -I${top_builddir} \
   -I$(JAVA_HOME)/include -I$(JAVA_HOME)/include/linux \
diff --git a/src/vpp-api/python/Makefile.am b/src/vpp-api/python/Makefile.am
index a5dabc3d..16c47627 100644
--- a/src/vpp-api/python/Makefile.am
+++ b/src/vpp-api/python/Makefile.am
@@ -13,6 +13,7 @@
 
 AUTOMAKE_OPTIONS = foreign
 ACLOCAL_AMFLAGS = -I m4
+AM_LIBTOOLFLAGS = --quiet
 AM_CFLAGS = -Wall -I${top_srcdir} -I${top_builddir}
 
 BUILT_SOURCES =
@@ -48,6 +49,7 @@ install-exec-local:
 #
 # Test client
 #
+if ENABLE_TESTS
 noinst_PROGRAMS += test_pneum
 test_pneum_SOURCES = pneum/pneum.c pneum/test_pneum.c
 test_pneum_LDADD = \
@@ -56,5 +58,4 @@ test_pneum_LDADD = \
   $(top_builddir)/libvlibapi.la \
   $(top_builddir)/libsvm.la \
   -lpthread -lm -lrt
-
-
+endif
diff --git a/src/vpp.am b/src/vpp.am
index 3bb2c1be..be55c400 100644
--- a/src/vpp.am
+++ b/src/vpp.am
@@ -68,39 +68,6 @@ VPP_VERSION = $(shell $(srcdir)/scripts/version)
 	  echo "#define VPP_BUILD_VER \"$(VPP_VERSION)\"" >> $$f			;\
 	fi
 
-
-# Generate a plugin configuration script. Misconfiguring a
-# plugin can cause very subtle problems.
-
-bin_SCRIPTS = vpp_plugin_configure
-
-BUILT_SOURCES += vpp_plugin_configure
-
-.PHONY: vpp_plugin_configure
-
-if WITH_DPDK
-PLUGIN_DPDK_ARG="--with-dpdk"
-else
-PLUGIN_DPDK_ARG=""
-endif
-
-vpp_plugin_configure:
-	@echo "PLUGIN CFG" $@
-	@echo "#!/bin/bash" > $@
-	@echo " " >> $@
-	@echo "set +eu" >> $@
-	@echo " " >> $@
-	@echo "if [ -f ./configure ] ; then" >> $@
-	@echo "    CFLAGS='$(CFLAGS) $(AM_CFLAGS) -I/usr/include/vpp-dpdk' ./configure --with-plugin-toolkit $(PLUGIN_DPDK_ARG)" >> $@
-	@echo "else" >> $@
-	@echo "    if [ -f ../configure ] ; then" >> $@
-	@echo "        CFLAGS='$(CFLAGS) $(AM_CFLAGS) -I/usr/include/vpp-dpdk' ../configure --with-plugin-toolkit $(PLUGIN_DPDK_ARG)" >> $@
-	@echo "    else" >> $@
-	@echo "        echo Couldnt find ./configure or ../configure " >> $@
-	@echo "        exit 1" >> $@
-	@echo "    fi" >> $@
-	@echo "fi" >> $@
-
 bin_vpp_LDADD = \
   libvlibapi.la \
   libvlibmemory.la \
@@ -108,12 +75,12 @@ bin_vpp_LDADD = \
   libvnet.la \
   libsvm.la \
   libsvmdb.la \
-  libvnetplugin.la \
   -lrt
 
 bin_vpp_LDFLAGS = $(DPDK_LD_FLAGS)
 bin_vpp_LDADD += libvppinfra.la -lm -lpthread -ldl $(DPDK_LD_ADD)
 
+if ENABLE_TESTS
 noinst_PROGRAMS += bin/test_client
 
 bin_test_client_SOURCES = \
@@ -138,6 +105,7 @@ bin_test_ha_LDADD = \
   libsvm.la \
   libvppinfra.la \
   -lpthread -lm -lrt
+endif
 
 noinst_PROGRAMS += bin/summary_stats_client
 
diff --git a/src/vppinfra.am b/src/vppinfra.am
index 836c7213..19485d2d 100644
--- a/src/vppinfra.am
+++ b/src/vppinfra.am
@@ -11,13 +11,7 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-libvppinfradir = ${libdir}
-
-if WITH_UNIX
-  libvppinfra_LTLIBRARIES = libvppinfra.la
-endif
-
-lib_LIBRARIES =
+lib_LTLIBRARIES += libvppinfra.la
 
 TESTS =
 
-- 
cgit 1.2.3-korg


From 69915cba2d6e169343c03c8eef48e6b2d2296e3a Mon Sep 17 00:00:00 2001
From: Padraig Connolly 
Date: Tue, 10 Jan 2017 17:10:39 +0000
Subject: vppctl: new bash completion for vppctl commands

Bash completion for vppctl
*vppctl_completion uses generated list vppctl-cmd-list to provide bash completion
*List of commands generated using grep, then moved by vpp packaging
*Once vpp package installed restart bash to use

Change-Id: I3b25e55a432c395af421231cf7c37c2e243fee61
Signed-off-by: Padraig Connolly 
---
 .gitignore                    |  3 +++
 build-data/platforms.mk       |  8 ++++++++
 build-root/rpm/vpp.spec       |  6 ++++++
 src/Makefile.am               |  8 ++++++++
 src/scripts/vppctl_completion | 30 ++++++++++++++++++++++++++++++
 5 files changed, 55 insertions(+)
 create mode 100644 src/scripts/vppctl_completion

(limited to 'src/Makefile.am')

diff --git a/.gitignore b/.gitignore
index 61c407ff..ab997faa 100644
--- a/.gitignore
+++ b/.gitignore
@@ -88,3 +88,6 @@ GTAGS
 
 # Build files in the test directory
 /test/*.ok
+
+# vppctl command list
+/src/scripts/vppctl-cmd-list
diff --git a/build-data/platforms.mk b/build-data/platforms.mk
index e192155b..41836750 100644
--- a/build-data/platforms.mk
+++ b/build-data/platforms.mk
@@ -75,6 +75,14 @@ install-deb: $(patsubst %,%-find-source,$(ROOT_PACKAGES))
 	echo ../../src/vpp/conf/80-vpp.conf /etc/sysctl.d 		\
 	   >> deb/debian/vpp.install ;					\
 									\
+	: bash completion for vppctl ;					\
+	echo ../../src/scripts/vppctl_completion /etc/bash_completion.d	\
+	   >> deb/debian/vpp.install ;					\
+									\
+	: move dictionary of vppctl commands ;				\
+	echo ../../src/scripts/vppctl-cmd-list /usr/share/vpp		\
+	   >> deb/debian/vpp.install ;					\
+									\
 	: dev package needs a couple of additions ;			\
 	echo ../$(INSTALL_PREFIX)$(ARCH)/vpp/bin/vppapigen /usr/bin	\
 	   >> deb/debian/vpp-dev.install ;				\
diff --git a/build-root/rpm/vpp.spec b/build-root/rpm/vpp.spec
index b63d3e17..8308dbf5 100644
--- a/build-root/rpm/vpp.spec
+++ b/build-root/rpm/vpp.spec
@@ -125,6 +125,8 @@ install -p -m 644 %{_mu_build_dir}/../src/vpp/conf/80-vpp.conf %{buildroot}/etc/
 # libraries
 #
 mkdir -p -m755 %{buildroot}%{_libdir}
+mkdir -p -m755 %{buildroot}/etc/bash_completion.d
+mkdir -p -m755 %{buildroot}/usr/share/vpp
 for file in $(find %{_mu_build_dir}/%{_vpp_install_dir}/*/lib* -type f -name '*.so.*.*.*' -print )
 do
 	install -p -m 755 $file %{buildroot}%{_libdir}
@@ -141,6 +143,8 @@ for file in $(find %{_mu_build_dir}/%{_vpp_install_dir}/vpp/share/vpp/api  -type
 do
 	install -p -m 644 $file %{buildroot}/usr/share/vpp/api
 done
+install -p -m 644 %{_mu_build_dir}/../src/scripts/vppctl_completion %{buildroot}/etc/bash_completion.d
+install -p -m 644 %{_mu_build_dir}/../src/scripts/vppctl-cmd-list %{buildroot}/usr/share/vpp
 
 # Lua bindings
 mkdir -p -m755 %{buildroot}/usr/share/doc/vpp/examples/lua/examples/cli
@@ -269,6 +273,8 @@ fi
 %exclude %{_libdir}/vpp_api_test_plugins
 %{_libdir}/*
 /usr/share/vpp/api/*
+/etc/bash_completion.d/vppctl_completion
+/usr/share/vpp/vppctl-cmd-list
 
 %files api-lua
 %defattr(644,root,root)
diff --git a/src/Makefile.am b/src/Makefile.am
index 18a41a15..227accc0 100644
--- a/src/Makefile.am
+++ b/src/Makefile.am
@@ -28,6 +28,14 @@ noinst_HEADERS =
 dist_bin_SCRIPTS =
 lib_LTLIBRARIES =
 BUILT_SOURCES =
+install-data-local:
+	@echo "Building vppctl command list..."
+	@DIR_SEARCH="$(srcdir)" ; \
+	DIR_EXCLUDE="examples" ; \
+	GREP_TIME=`time (grep -wIr "\.path = " $$DIR_SEARCH --exclude-dir=$$DIR_EXCLUDE \
+	| cut -d '"' -f2 | sort -u > $(srcdir)/scripts/vppctl-cmd-list) 2>&1` ; \
+	GREP_TIME=`echo $$GREP_TIME | awk '{print $$2}'` ; \
+	echo "Command list built, Time taken: $$GREP_TIME"
 
 ###############################################################################
 # DPDK
diff --git a/src/scripts/vppctl_completion b/src/scripts/vppctl_completion
new file mode 100644
index 00000000..2a64e9ab
--- /dev/null
+++ b/src/scripts/vppctl_completion
@@ -0,0 +1,30 @@
+#Copyright 2017 Intel Corporation
+#
+#Licensed under the Apache License, Version 2.0 (the "License");
+#you may not use this file except in compliance with the License.
+#You may obtain a copy of the License at
+#
+#    http://www.apache.org/licenses/LICENSE-2.0
+#
+#Unless required by applicable law or agreed to in writing, software
+#distributed under the License is distributed on an "AS IS" BASIS,
+#WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+#See the License for the specific language governing permissions and
+#limitations under the License.
+
+_vppctl()
+{
+    local cur prev num opts
+    COMPREPLY=( $(compgen -f ${cur}) )
+    cur="${COMP_WORDS[COMP_CWORD]}"
+    prev="${COMP_WORDS[@]:1}"
+    num="$((${#COMP_WORDS[@]}-1))"
+
+    VPP_CMD_LIST="$(cat /usr/share/vpp/vppctl-cmd-list)"
+
+    opts="$(awk -v prev="^$prev" -v num=$num '{if($0 ~ prev) print $num}' <<< "${VPP_CMD_LIST}")"
+
+    COMPREPLY=( $(compgen -W "${opts}" -- ${cur}) )
+
+}
+complete -F _vppctl vppctl
-- 
cgit 1.2.3-korg


From c74a86af08ae9cf312c47ad4dad54d70ae3093ed Mon Sep 17 00:00:00 2001
From: Damjan Marion 
Date: Mon, 16 Jan 2017 14:00:03 +0100
Subject: Define more dependencies between libs

Change-Id: Ia74699bad15c1c583571944458fc8a5c3ded574b
Signed-off-by: Damjan Marion 
---
 src/Makefile.am     |  2 +-
 src/vlib-api.am     | 32 ++++++++++++++++----------------
 src/vlib.am         |  2 --
 src/vnet.am         | 22 ++++++++++++++--------
 src/vpp-api-test.am |  4 +---
 src/vpp.am          |  6 ++----
 6 files changed, 34 insertions(+), 34 deletions(-)

(limited to 'src/Makefile.am')

diff --git a/src/Makefile.am b/src/Makefile.am
index 227accc0..5e248972 100644
--- a/src/Makefile.am
+++ b/src/Makefile.am
@@ -45,7 +45,7 @@ if WITH_DPDK
 if ENABLE_DPDK_SHARED
 DPDK_LD_FLAGS = -Wl,--whole-archive,-ldpdk,--no-whole-archive
 else
-DPDK_LD_FLAGS = -Wl,--whole-archive,-l:libdpdk.a,--no-whole-archive
+DPDK_LD_FLAGS = -Wl,--whole-archive,-l:libdpdk.a,--no-whole-archive,-lm,-ldl
 endif
 if WITH_DPDK_CRYPTO
 DPDK_LD_ADD = -L$(AESNI_MULTI_BUFFER_LIB_PATH) -lIPSec_MB
diff --git a/src/vlib-api.am b/src/vlib-api.am
index 128cc07a..6312f93d 100644
--- a/src/vlib-api.am
+++ b/src/vlib-api.am
@@ -11,21 +11,11 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-lib_LTLIBRARIES += libvlibapi.la libvlibmemory.la libvlibmemoryclient.la \
+lib_LTLIBRARIES += libvlibmemory.la libvlibapi.la libvlibmemoryclient.la \
 	           libvlibsocket.la
 
-libvlibapi_la_LIBADD = libvppinfra.la
-libvlibapi_la_DEPENDENCIES = libvppinfra.la libvlib.la
-libvlibapi_la_SOURCES = 			\
-	vlibapi/api.h				\
-        vlibapi/api_helper_macros.h		\
-	vlibapi/api_shared.c			\
-	vlibapi/node_serialize.c
-
-nobase_include_HEADERS += vlibapi/api.h vlibapi/api_helper_macros.h
-
-libvlibmemory_la_LIBADD = libvppinfra.la libsvm.la -lpthread
 libvlibmemory_la_DEPENDENCIES = libvppinfra.la libsvm.la libvlib.la
+libvlibmemory_la_LIBADD = $(libvlibmemory_la_DEPENDENCIES) -lpthread
 libvlibmemory_la_SOURCES =			\
 	vlibmemory/api.h			\
 	vlibmemory/memclnt.api			\
@@ -36,8 +26,18 @@ libvlibmemory_la_SOURCES =			\
 	vlibmemory/unix_shared_memory_queue.c	\
 	vlibmemory/unix_shared_memory_queue.h
 
-libvlibmemoryclient_la_LIBADD = libvppinfra.la libsvm.la -lpthread
-libvlibmemoryclient_la_DEPENDENCIES = libvppinfra.la libsvm.la libvlib.la
+libvlibapi_la_DEPENDENCIES = libvppinfra.la libvlib.la libvlibmemory.la
+libvlibapi_la_LIBADD = $(libvlibapi_la_DEPENDENCIES)
+libvlibapi_la_SOURCES = 			\
+	vlibapi/api.h				\
+        vlibapi/api_helper_macros.h		\
+	vlibapi/api_shared.c			\
+	vlibapi/node_serialize.c
+
+nobase_include_HEADERS += vlibapi/api.h vlibapi/api_helper_macros.h
+
+libvlibmemoryclient_la_DEPENDENCIES = libvppinfra.la libsvm.la libvlib.la libvlibmemory.la libvlibapi.la
+libvlibmemoryclient_la_LIBADD = $(libvlibmemoryclient_la_DEPENDENCIES) -lpthread
 libvlibmemoryclient_la_SOURCES = 		\
 	vlibmemory/api.h			\
 	vlibmemory/memclnt.api			\
@@ -55,8 +55,8 @@ nobase_include_HEADERS +=			\
 	vlibmemory/unix_shared_memory_queue.h 	\
 	vlibmemory/memclnt.api.h
 
-libvlibsocket_la_LIBADD = libvppinfra.la
-libvlibsocket_la_DEPENDENCIES = libvppinfra.la
+libvlibsocket_la_DEPENDENCIES = libvppinfra.la libvlib.la libvlibmemory.la libvlibapi.la
+libvlibsocket_la_LIBADD = $(libvlibsocket_la_DEPENDENCIES)
 libvlibsocket_la_SOURCES =			\
 	vlibsocket/api.h			\
 	vlibsocket/sockclnt.api			\
diff --git a/src/vlib.am b/src/vlib.am
index c21f88c4..111dcfa3 100644
--- a/src/vlib.am
+++ b/src/vlib.am
@@ -88,7 +88,6 @@ nobase_include_HEADERS +=			\
   vlib/unix/plugin.h				\
   vlib/unix/unix.h
 
-if !WITH_DPDK
 noinst_PROGRAMS += vlib_unix
 
 vlib_unix_SOURCES =				\
@@ -97,6 +96,5 @@ vlib_unix_SOURCES =				\
 
 vlib_unix_LDADD = libvlib.la \
 	          libvppinfra.la -lpthread -lm -ldl -lrt
-endif
 
 # vi:syntax=automake
diff --git a/src/vnet.am b/src/vnet.am
index 47c5eda7..76824fdb 100644
--- a/src/vnet.am
+++ b/src/vnet.am
@@ -14,9 +14,17 @@
 lib_LTLIBRARIES += libvnet.la
 
 libvnet_la_SOURCES =
-
-libvnet_la_LIBADD = libvppinfra.la libsvm.la
-libvnet_la_DEPENDENCIES = libvppinfra.la libvlib.la libsvmdb.la libsvm.la libvlibapi.la libvlibmemory.la libvlibmemoryclient.la
+libvnet_la_DEPENDENCIES = \
+	libvppinfra.la		\
+	libvlib.la		\
+	libsvmdb.la		\
+	libsvm.la 		\
+	libvlibapi.la		\
+	libvlibmemory.la	\
+	libvlibmemoryclient.la
+
+libvnet_la_LIBADD = $(libvnet_la_DEPENDENCIES) -lm -lpthread -ldl -lrt $(DPDK_LD_ADD)
+libvnet_la_LDFLAGS = $(DPDK_LD_FLAGS)
 
 if WITH_IPV6SR
 libvnet_la_LIBADD += -lcrypto
@@ -579,7 +587,7 @@ LDS = \
   libsvmdb.la \
   libvlibapi.la \
   libvlibmemory.la \
-  -lpthread -ldl -lrt -lm -lcrypto
+  -lpthread -ldl -lrt -lm
 
 TESTS += test_cp_serdes test_lisp_types
 
@@ -597,10 +605,8 @@ test_lisp_types_SOURCES =			\
 test_cp_serdes_CPPFLAGS = $(AM_CPPFLAGS) -DCLIB_DEBUG
 test_lisp_types_CPPFLAGS = $(AM_CPPFLAGS) -DCLIB_DEBUG
 
-test_cp_serdes_LDFLAGS = $(DPDK_LD_FLAGS)
-test_cp_serdes_LDADD = $(LDS) $(DPDK_LD_ADD)
-test_lisp_types_LDFLAGS = $(DPDK_LD_FLAGS)
-test_lisp_types_LDADD = $(LDS) $(DPDK_LD_ADD)
+test_cp_serdes_LDADD = $(LDS)
+test_lisp_types_LDADD = $(LDS)
 endif
 
 ########################################
diff --git a/src/vpp-api-test.am b/src/vpp-api-test.am
index c788666e..32610056 100644
--- a/src/vpp-api-test.am
+++ b/src/vpp-api-test.am
@@ -33,7 +33,6 @@ vpp_json_test_SOURCES = \
   vat/json_format.c	\
   vat/json_test.c
 
-vpp_api_test_LDFLAGS = $(DPDK_LD_FLAGS)
 vpp_api_test_LDADD = \
   libvlib.la				\
   libvlibmemoryclient.la		\
@@ -42,8 +41,7 @@ vpp_api_test_LDADD = \
   libvppinfra.la 			\
   libvlibapi.la				\
   libvlibmemory.la			\
-  libvnet.la \
-  $(DPDK_LD_ADD)			\
+  libvnet.la				\
   -lpthread -lm -lrt -ldl -lcrypto
 
 vpp_json_test_LDADD = libvppinfra.la -lm
diff --git a/src/vpp.am b/src/vpp.am
index be55c400..425f1e32 100644
--- a/src/vpp.am
+++ b/src/vpp.am
@@ -75,10 +75,8 @@ bin_vpp_LDADD = \
   libvnet.la \
   libsvm.la \
   libsvmdb.la \
-  -lrt
-
-bin_vpp_LDFLAGS = $(DPDK_LD_FLAGS)
-bin_vpp_LDADD += libvppinfra.la -lm -lpthread -ldl $(DPDK_LD_ADD)
+  libvppinfra.la \
+  -lrt -lm -lpthread -ldl
 
 if ENABLE_TESTS
 noinst_PROGRAMS += bin/test_client
-- 
cgit 1.2.3-korg


From d04b60bfa940e21ab4676a1cb3c15989748be40a Mon Sep 17 00:00:00 2001
From: Sergio Gonzalez Monroy 
Date: Fri, 20 Jan 2017 15:35:23 +0000
Subject: dpdk: rework cryptodev ipsec build and setup

Build Cryptodev IPsec support by default when DPDK is enabled but only build
hardware Cryptodev PMDs.

To enable Cryptodev support, a new startup.conf option for dpdk has been
introduced 'enable-cryptodev'.

During VPP init, if Cryptodev support is not enabled or not enough cryptodev
resources are available then default to OpenSSL ipsec implementation.

Change-Id: I5aa7e0d5c2676bdb41d775ef40364536a081956d
Signed-off-by: Sergio Gonzalez Monroy 
---
 build-data/packages/dpdk.mk                        |   6 +-
 build-data/packages/vpp.mk                         |   4 +-
 build-data/platforms/vpp.mk                        |   2 +-
 dpdk/Makefile                                      |  10 +-
 src/Makefile.am                                    |   2 +-
 src/configure.ac                                   |   4 +-
 src/vat/api_format.c                               |  31 ---
 src/vnet.am                                        |   4 +-
 src/vnet/devices/dpdk/dpdk.h                       |   1 +
 src/vnet/devices/dpdk/format.c                     |   3 +
 src/vnet/devices/dpdk/init.c                       |   3 +
 src/vnet/devices/dpdk/ipsec/cli.c                  |   8 +
 src/vnet/devices/dpdk/ipsec/crypto_node.c          |  25 +-
 .../devices/dpdk/ipsec/dpdk_crypto_ipsec_doc.md    |  37 ++-
 src/vnet/devices/dpdk/ipsec/esp.h                  |  95 ++------
 src/vnet/devices/dpdk/ipsec/esp_decrypt.c          |  11 +-
 src/vnet/devices/dpdk/ipsec/esp_encrypt.c          |  11 +-
 src/vnet/devices/dpdk/ipsec/ipsec.c                | 257 +++++++++++++++------
 src/vnet/devices/dpdk/ipsec/ipsec.h                |   2 +-
 src/vnet/ipsec-gre/interface.c                     |   8 +-
 src/vnet/ipsec/ipsec.c                             |  51 ++--
 src/vnet/ipsec/ipsec.h                             |  63 +++--
 src/vnet/ipsec/ipsec_api.c                         |  43 +---
 src/vnet/ipsec/ipsec_cli.c                         |  31 +--
 src/vnet/ipsec/ipsec_if.c                          |  73 ++++--
 src/vnet/ipsec/ipsec_if_in.c                       |  24 +-
 src/vnet/ipsec/ipsec_if_out.c                      |  22 +-
 src/vnet/ipsec/ipsec_input.c                       |  24 +-
 src/vnet/ipsec/ipsec_output.c                      |  20 --
 29 files changed, 452 insertions(+), 423 deletions(-)

(limited to 'src/Makefile.am')

diff --git a/build-data/packages/dpdk.mk b/build-data/packages/dpdk.mk
index 6c136824..6938392c 100644
--- a/build-data/packages/dpdk.mk
+++ b/build-data/packages/dpdk.mk
@@ -22,9 +22,9 @@ DPDK_MAKE_ARGS = -C $(call find_source_fn,$(PACKAGE_SOURCE)) \
 	DPDK_TUNE=$(DPDK_TUNE) \
 	DPDK_DEBUG=$(DPDK_DEBUG)
 
-DPDK_CRYPTO_PMD=$(strip $($(PLATFORM)_uses_dpdk_cryptodev))
-ifneq ($(DPDK_CRYPTO_PMD),)
-DPDK_MAKE_ARGS += DPDK_CRYPTO_PMD=y
+DPDK_CRYPTO_SW_PMD=$(strip $($(PLATFORM)_uses_dpdk_cryptodev_sw))
+ifneq ($(DPDK_CRYPTO_SW_PMD),)
+DPDK_MAKE_ARGS += DPDK_CRYPTO_SW_PMD=y
 endif
 
 DPDK_MLX5_PMD=$(strip $($(PLATFORM)_uses_dpdk_mlx5_pmd))
diff --git a/build-data/packages/vpp.mk b/build-data/packages/vpp.mk
index 81aeab69..64eb0d89 100644
--- a/build-data/packages/vpp.mk
+++ b/build-data/packages/vpp.mk
@@ -23,8 +23,8 @@ vpp_CPPFLAGS += $(call installed_includes_fn, dpdk)/dpdk
 vpp_LDFLAGS += $(call installed_libs_fn, dpdk)
 vpp_CPPFLAGS += -I/usr/include/dpdk
 endif
-ifeq ($($(PLATFORM)_uses_dpdk_cryptodev),yes)
-vpp_configure_args += --with-dpdk-crypto
+ifeq ($($(PLATFORM)_uses_dpdk_cryptodev_sw),yes)
+vpp_configure_args += --with-dpdk-crypto-sw
 endif
 ifeq ($($(PLATFORM)_uses_dpdk_mlx5_pmd),yes)
 vpp_configure_args += --with-dpdk-mlx5-pmd
diff --git a/build-data/platforms/vpp.mk b/build-data/platforms/vpp.mk
index dd6f9dc2..5b200587 100644
--- a/build-data/platforms/vpp.mk
+++ b/build-data/platforms/vpp.mk
@@ -44,7 +44,7 @@ vpp_configure_args_vpp = --with-dpdk
 vlib_configure_args_vpp = --with-pre-data=128
 
 # DPDK configuration parameters
-# vpp_uses_dpdk_cryptodev = yes
+# vpp_uses_dpdk_cryptodev_sw = yes
 # vpp_uses_dpdk_mlx5_pmd = yes
 # vpp_uses_external_dpdk = yes
 # vpp_dpdk_inc_dir = /usr/include/dpdk
diff --git a/dpdk/Makefile b/dpdk/Makefile
index 586d2425..22e97878 100644
--- a/dpdk/Makefile
+++ b/dpdk/Makefile
@@ -21,13 +21,13 @@ DPDK_DOWNLOAD_DIR     ?= $(HOME)/Downloads
 DPDK_MARCH            ?= native
 DPDK_TUNE             ?= generic
 DPDK_DEBUG            ?= n
-DPDK_CRYPTO_PMD       ?= n
+DPDK_CRYPTO_SW_PMD    ?= n
 DPDK_MLX5_PMD         ?= n
 
 B := $(DPDK_BUILD_DIR)
 I := $(DPDK_INSTALL_DIR)
 DPDK_VERSION ?= 16.11
-PKG_SUFFIX ?= vpp1
+PKG_SUFFIX ?= vpp2
 DPDK_BASE_URL ?= http://fast.dpdk.org/rel
 DPDK_TARBALL := dpdk-$(DPDK_VERSION).tar.xz
 DPDK_TAR_URL := $(DPDK_BASE_URL)/$(DPDK_TARBALL)
@@ -121,9 +121,9 @@ $(B)/custom-config: $(B)/.patch.ok Makefile
 	$(call set,RTE_LIBRTE_VMXNET3_DEBUG_INIT,$(DPDK_DEBUG))
 	$(call set,RTE_LIBRTE_PMD_BOND,y)
 	$(call set,RTE_LIBRTE_IP_FRAG,y)
-	$(call set,RTE_LIBRTE_PMD_AESNI_MB,$(DPDK_CRYPTO_PMD))
-	$(call set,RTE_LIBRTE_PMD_AESNI_GCM,$(DPDK_CRYPTO_PMD))
-	$(call set,RTE_LIBRTE_PMD_QAT,$(DPDK_CRYPTO_PMD))
+	$(call set,RTE_LIBRTE_PMD_QAT,y)
+	$(call set,RTE_LIBRTE_PMD_AESNI_MB,$(DPDK_CRYPTO_SW_PMD))
+	$(call set,RTE_LIBRTE_PMD_AESNI_GCM,$(DPDK_CRYPTO_SW_PMD))
 	$(call set,RTE_LIBRTE_MLX5_PMD,$(DPDK_MLX5_PMD))
 	@# not needed
 	$(call set,RTE_LIBRTE_TIMER,n)
diff --git a/src/Makefile.am b/src/Makefile.am
index 5e248972..239afeac 100644
--- a/src/Makefile.am
+++ b/src/Makefile.am
@@ -47,7 +47,7 @@ DPDK_LD_FLAGS = -Wl,--whole-archive,-ldpdk,--no-whole-archive
 else
 DPDK_LD_FLAGS = -Wl,--whole-archive,-l:libdpdk.a,--no-whole-archive,-lm,-ldl
 endif
-if WITH_DPDK_CRYPTO
+if WITH_DPDK_CRYPTO_SW
 DPDK_LD_ADD = -L$(AESNI_MULTI_BUFFER_LIB_PATH) -lIPSec_MB
 endif
 if WITH_DPDK_MLX5_PMD
diff --git a/src/configure.ac b/src/configure.ac
index fbedabf0..49da6248 100644
--- a/src/configure.ac
+++ b/src/configure.ac
@@ -98,7 +98,7 @@ DISABLE_ARG(japi,	[Disable Java API bindings])
 
 # --with-X
 WITH_ARG(dpdk,		[Use use DPDK])
-WITH_ARG(dpdk_crypto,	[Use DPDK cryptodev])
+WITH_ARG(dpdk_crypto_sw,[Use DPDK cryptodev SW PMDs])
 WITH_ARG(dpdk_mlx5_pmd,	[Use DPDK with mlx5 PMD])
 
 # --without-X
@@ -132,7 +132,7 @@ AC_SUBST(APICLI,		[-DVPP_API_TEST_BUILTIN=${n_with_apicli}])
 
 AC_DEFINE_UNQUOTED(DPDK, 		[${n_with_dpdk}])
 AC_DEFINE_UNQUOTED(DPDK_SHARED_LIB,	[${n_enable_dpdk_shared}])
-AC_DEFINE_UNQUOTED(DPDK_CRYPTO,		[${n_with_dpdk_crypto}])
+AC_DEFINE_UNQUOTED(DPDK_CRYPTO_SW,	[${n_with_dpdk_crypto_sw}])
 AC_DEFINE_UNQUOTED(WITH_LIBSSL,		[${n_with_libssl}])
 
 
diff --git a/src/vat/api_format.c b/src/vat/api_format.c
index 4cfe4a58..6b8c5fb9 100644
--- a/src/vat/api_format.c
+++ b/src/vat/api_format.c
@@ -12199,11 +12199,7 @@ api_ipsec_sad_add_del_entry (vat_main_t * vam)
 	if (unformat
 	    (i, "integ_alg %U", unformat_ipsec_integ_alg, &integ_alg))
 	{
-#if DPDK_CRYPTO==1
-	  if (integ_alg < IPSEC_INTEG_ALG_NONE ||
-#else
 	  if (integ_alg < IPSEC_INTEG_ALG_SHA1_96 ||
-#endif
 	      integ_alg >= IPSEC_INTEG_N_ALG)
 	    {
 	      clib_warning ("unsupported integ-alg: '%U'",
@@ -12221,33 +12217,6 @@ api_ipsec_sad_add_del_entry (vat_main_t * vam)
 
     }
 
-#if DPDK_CRYPTO==1
-  /*Special cases, aes-gcm-128 encryption */
-  if (crypto_alg == IPSEC_CRYPTO_ALG_AES_GCM_128)
-    {
-      if (integ_alg != IPSEC_INTEG_ALG_NONE
-	  && integ_alg != IPSEC_INTEG_ALG_AES_GCM_128)
-	{
-	  clib_warning
-	    ("unsupported: aes-gcm-128 crypto-alg needs none as integ-alg");
-	  return -99;
-	}
-      else			/*set integ-alg internally to aes-gcm-128 */
-	integ_alg = IPSEC_INTEG_ALG_AES_GCM_128;
-    }
-  else if (integ_alg == IPSEC_INTEG_ALG_AES_GCM_128)
-    {
-      clib_warning ("unsupported integ-alg: aes-gcm-128");
-      return -99;
-    }
-  else if (integ_alg == IPSEC_INTEG_ALG_NONE)
-    {
-      clib_warning ("unsupported integ-alg: none");
-      return -99;
-    }
-#endif
-
-
   M (IPSEC_SAD_ADD_DEL_ENTRY, ipsec_sad_add_del_entry);
 
   mp->sad_id = ntohl (sad_id);
diff --git a/src/vnet.am b/src/vnet.am
index 28a1b19a..96cfa557 100644
--- a/src/vnet.am
+++ b/src/vnet.am
@@ -400,7 +400,7 @@ libvnet_la_SOURCES +=				\
 
 API_FILES += vnet/ipsec/ipsec.api
 
-if WITH_DPDK_CRYPTO
+if WITH_DPDK
 libvnet_la_SOURCES +=				\
  vnet/devices/dpdk/ipsec/esp_encrypt.c		\
  vnet/devices/dpdk/ipsec/esp_decrypt.c		\
@@ -419,7 +419,7 @@ nobase_include_HEADERS +=			\
  vnet/ipsec/ikev2.h				\
  vnet/ipsec/ikev2_priv.h			\
  vnet/ipsec/ipsec.api.h
-if WITH_DPDK_CRYPTO
+if WITH_DPDK
 nobase_include_HEADERS +=			\
  vnet/devices/dpdk/ipsec/ipsec.h		\
  vnet/devices/dpdk/ipsec/esp.h
diff --git a/src/vnet/devices/dpdk/dpdk.h b/src/vnet/devices/dpdk/dpdk.h
index a91e87df..1b54460e 100644
--- a/src/vnet/devices/dpdk/dpdk.h
+++ b/src/vnet/devices/dpdk/dpdk.h
@@ -338,6 +338,7 @@ typedef struct
   u8 *uio_driver_name;
   u8 no_multi_seg;
   u8 enable_tcp_udp_checksum;
+  u8 cryptodev;
 
   /* Required config parameters */
   u8 coremask_set_manually;
diff --git a/src/vnet/devices/dpdk/format.c b/src/vnet/devices/dpdk/format.c
index ff7c7a5a..cc0d71af 100644
--- a/src/vnet/devices/dpdk/format.c
+++ b/src/vnet/devices/dpdk/format.c
@@ -684,6 +684,8 @@ format_dpdk_rte_mbuf (u8 * s, va_list * va)
   return s;
 }
 
+/* FIXME is this function used? */
+#if 0
 uword
 unformat_socket_mem (unformat_input_t * input, va_list * va)
 {
@@ -710,6 +712,7 @@ unformat_socket_mem (unformat_input_t * input, va_list * va)
 done:
   return 1;
 }
+#endif
 
 clib_error_t *
 unformat_rss_fn (unformat_input_t * input, uword * rss_fn)
diff --git a/src/vnet/devices/dpdk/init.c b/src/vnet/devices/dpdk/init.c
index 3fa656ea..01ef48cb 100755
--- a/src/vnet/devices/dpdk/init.c
+++ b/src/vnet/devices/dpdk/init.c
@@ -1054,6 +1054,9 @@ dpdk_config (vlib_main_t * vm, unformat_input_t * input)
       else if (unformat (input, "no-multi-seg"))
 	conf->no_multi_seg = 1;
 
+      else if (unformat (input, "enable-cryptodev"))
+	conf->cryptodev = 1;
+
       else if (unformat (input, "dev default %U", unformat_vlib_cli_sub_input,
 			 &sub_input))
 	{
diff --git a/src/vnet/devices/dpdk/ipsec/cli.c b/src/vnet/devices/dpdk/ipsec/cli.c
index 3b634e03..93df4a64 100644
--- a/src/vnet/devices/dpdk/ipsec/cli.c
+++ b/src/vnet/devices/dpdk/ipsec/cli.c
@@ -14,15 +14,23 @@
  */
 
 #include 
+#include 
 #include 
 
 static void
 dpdk_ipsec_show_mapping (vlib_main_t * vm, u16 detail_display)
 {
+  dpdk_config_main_t *conf = &dpdk_config_main;
   dpdk_crypto_main_t *dcm = &dpdk_crypto_main;
   vlib_thread_main_t *tm = vlib_get_thread_main ();
   u32 i, skip_master;
 
+  if (!conf->cryptodev)
+    {
+      vlib_cli_output (vm, "DPDK Cryptodev support is disabled\n");
+      return;
+    }
+
   if (detail_display)
     vlib_cli_output (vm, "worker\t%10s\t%15s\tdir\tdev\tqp\n",
 		     "cipher", "auth");
diff --git a/src/vnet/devices/dpdk/ipsec/crypto_node.c b/src/vnet/devices/dpdk/ipsec/crypto_node.c
index 7b32704e..e8fef235 100644
--- a/src/vnet/devices/dpdk/ipsec/crypto_node.c
+++ b/src/vnet/devices/dpdk/ipsec/crypto_node.c
@@ -22,6 +22,8 @@
 #include 
 #include 
 
+#include 
+#include 
 #include 
 
 #define foreach_dpdk_crypto_input_next		\
@@ -183,24 +185,27 @@ dpdk_crypto_input_fn (vlib_main_t * vm, vlib_node_runtime_t * node,
   return n_deq;
 }
 
+/* *INDENT-OFF* */
 VLIB_REGISTER_NODE (dpdk_crypto_input_node) =
 {
-  .function = dpdk_crypto_input_fn,.name = "dpdk-crypto-input",.format_trace =
-    format_dpdk_crypto_input_trace,.type = VLIB_NODE_TYPE_INPUT,.state =
-    VLIB_NODE_STATE_DISABLED,.n_errors =
-    DPDK_CRYPTO_INPUT_N_ERROR,.error_strings =
-    dpdk_crypto_input_error_strings,.n_next_nodes =
-    DPDK_CRYPTO_INPUT_N_NEXT,.next_nodes =
+  .function = dpdk_crypto_input_fn,
+  .name = "dpdk-crypto-input",
+  .format_trace = format_dpdk_crypto_input_trace,
+  .type = VLIB_NODE_TYPE_INPUT,
+  .state = VLIB_NODE_STATE_DISABLED,
+  .n_errors = DPDK_CRYPTO_INPUT_N_ERROR,
+  .error_strings = dpdk_crypto_input_error_strings,
+  .n_next_nodes = DPDK_CRYPTO_INPUT_N_NEXT,
+  .next_nodes =
   {
 #define _(s,n) [DPDK_CRYPTO_INPUT_NEXT_##s] = n,
     foreach_dpdk_crypto_input_next
 #undef _
-  }
-,};
+  },
+};
+/* *INDENT-ON* */
 
-#if DPDK_CRYPTO==1
 VLIB_NODE_FUNCTION_MULTIARCH (dpdk_crypto_input_node, dpdk_crypto_input_fn)
-#endif
 /*
  * fd.io coding-style-patch-verification: ON
  *
diff --git a/src/vnet/devices/dpdk/ipsec/dpdk_crypto_ipsec_doc.md b/src/vnet/devices/dpdk/ipsec/dpdk_crypto_ipsec_doc.md
index 8089696f..fed2fe0e 100644
--- a/src/vnet/devices/dpdk/ipsec/dpdk_crypto_ipsec_doc.md
+++ b/src/vnet/devices/dpdk/ipsec/dpdk_crypto_ipsec_doc.md
@@ -7,43 +7,55 @@ This document is meant to contain all related information about implementation a
 
 DPDK Cryptodev is an asynchronous crypto API that supports both Hardware and Software implementations (for more details refer to [DPDK Cryptography Device Library documentation](http://dpdk.org/doc/guides/prog_guide/cryptodev_lib.html)).
 
-When DPDK Cryptodev support is enabled, the node graph is modified by adding and replacing some of the nodes.
-
-The following nodes are replaced:
-* esp-encrypt -> dpdk-esp-encrypt
-* esp-decrypt -> dpdk-esp-decrypt
+When DPDK support is enabled and there are enough Cryptodev resources for all workers, the node graph is reconfigured by adding and changing default next nodes.
 
 The following nodes are added:
 * dpdk-crypto-input : polling input node, basically dequeuing from crypto devices.
+* dpdk-esp-encrypt : internal node.
+* dpdk-esp-decrypt : internal node.
 * dpdk-esp-encrypt-post : internal node.
 * dpdk-esp-decrypt-post : internal node.
 
+Set new default next nodes:
+* for esp encryption: esp-encrypt -> dpdk-esp-encrypt
+* for esp decryption: esp-decrypt -> dpdk-esp-decrypt
+
 
 ### How to enable VPP IPSec with DPDK Cryptodev support
 
-To enable DPDK Cryptodev support (disabled by default), we need the following env option:
+DPDK Cryptodev is supported in DPDK enabled VPP.
+By default, only HW Cryptodev is supported but needs to be explicetly enabled with the following config option:
+
+```
+dpdk {
+    enable-cryptodev
+}
+```
+
+To enable SW Cryptodev support (AESNI-MB-PMD and GCM-PMD), we need the following env option:
 
-    vpp_uses_dpdk_cryptodev=yes
+    vpp_uses_dpdk_cryptodev_sw=yes
 
 A couple of ways to achive this:
 * uncomment/add it in the platforms config (ie. build-data/platforms/vpp.mk)
-* set the option when building vpp (ie. make vpp_uses_dpdk_cryptodev=yes build-release)
+* set the option when building vpp (ie. make vpp_uses_dpdk_cryptodev_sw=yes build-release)
+
+When enabling SW Cryptodev support, it means that you need to pre-build the required crypto libraries needed by those SW Cryptodev PMDs.
 
 
 ### Crypto Resources allocation
 
 VPP allocates crypto resources based on a best effort approach:
 * first allocate Hardware crypto resources, then Software.
-* if there are not enough crypto resources for all workers, all packets will be dropped if they reach ESP encrypt/decrypt nodes, displaying the warning:
+* if there are not enough crypto resources for all workers, the graph node is not modifed, therefore the default VPP IPsec implementation based in OpenSSL is used. The following message is displayed:
 
       0: dpdk_ipsec_init: not enough cryptodevs for ipsec
 
 
 ### Configuration example
 
-No especial IPsec configuration is required.
-
-Once DPDK Cryptodev is enabled, the user just needs to provide cryptodevs in the startup.conf.
+To enable DPDK Cryptodev the user just need to provide the startup.conf option
+as mentioned previously.
 
 Example startup.conf:
 
@@ -53,6 +65,7 @@ dpdk {
     num-mbufs 131072
     dev 0000:81:00.0
     dev 0000:81:00.1
+    enable-cryptodev
     dev 0000:85:01.0
     dev 0000:85:01.1
     vdev cryptodev_aesni_mb_pmd,socket_id=1
diff --git a/src/vnet/devices/dpdk/ipsec/esp.h b/src/vnet/devices/dpdk/ipsec/esp.h
index 7ef90c49..d414d679 100644
--- a/src/vnet/devices/dpdk/ipsec/esp.h
+++ b/src/vnet/devices/dpdk/ipsec/esp.h
@@ -97,60 +97,11 @@ dpdk_esp_init ()
 }
 
 static_always_inline int
-add_del_sa_sess (u32 sa_index, u8 is_add)
-{
-  dpdk_crypto_main_t *dcm = &dpdk_crypto_main;
-  crypto_worker_main_t *cwm;
-  u8 skip_master = vlib_num_workers () > 0;
-
-  /* *INDENT-OFF* */
-  vec_foreach (cwm, dcm->workers_main)
-    {
-      crypto_sa_session_t *sa_sess;
-      u8 is_outbound;
-
-      if (skip_master)
-	{
-	  skip_master = 0;
-	  continue;
-	}
-
-      for (is_outbound = 0; is_outbound < 2; is_outbound++)
-	{
-	  if (is_add)
-	    {
-	      pool_get (cwm->sa_sess_d[is_outbound], sa_sess);
-	    }
-	  else
-	    {
-	      u8 dev_id;
-
-	      sa_sess = pool_elt_at_index (cwm->sa_sess_d[is_outbound], sa_index);
-	      dev_id = cwm->qp_data[sa_sess->qp_index].dev_id;
-
-	      if (!sa_sess->sess)
-		continue;
-
-	      if (rte_cryptodev_sym_session_free(dev_id, sa_sess->sess))
-		{
-		  clib_warning("failed to free session");
-		  return -1;
-		}
-	      memset(sa_sess, 0, sizeof(sa_sess[0]));
-	    }
-	}
-    }
-  /* *INDENT-OFF* */
-
-  return 0;
-}
-
-static_always_inline int
-translate_crypto_algo(ipsec_crypto_alg_t crypto_algo,
-		      struct rte_crypto_sym_xform *cipher_xform)
+translate_crypto_algo (ipsec_crypto_alg_t crypto_algo,
+		       struct rte_crypto_sym_xform *cipher_xform)
 {
   switch (crypto_algo)
-  {
+    {
     case IPSEC_CRYPTO_ALG_NONE:
       cipher_xform->cipher.algo = RTE_CRYPTO_CIPHER_NULL;
       break;
@@ -164,7 +115,7 @@ translate_crypto_algo(ipsec_crypto_alg_t crypto_algo,
       break;
     default:
       return -1;
-  }
+    }
 
   cipher_xform->type = RTE_CRYPTO_SYM_XFORM_CIPHER;
 
@@ -172,10 +123,11 @@ translate_crypto_algo(ipsec_crypto_alg_t crypto_algo,
 }
 
 static_always_inline int
-translate_integ_algo(ipsec_integ_alg_t integ_alg,
-		     struct rte_crypto_sym_xform *auth_xform, int use_esn)
+translate_integ_algo (ipsec_integ_alg_t integ_alg,
+		      struct rte_crypto_sym_xform *auth_xform, int use_esn)
 {
-  switch (integ_alg) {
+  switch (integ_alg)
+    {
     case IPSEC_INTEG_ALG_NONE:
       auth_xform->auth.algo = RTE_CRYPTO_AUTH_NULL;
       auth_xform->auth.digest_length = 0;
@@ -203,11 +155,11 @@ translate_integ_algo(ipsec_integ_alg_t integ_alg,
     case IPSEC_INTEG_ALG_AES_GCM_128:
       auth_xform->auth.algo = RTE_CRYPTO_AUTH_AES_GCM;
       auth_xform->auth.digest_length = 16;
-      auth_xform->auth.add_auth_data_length = use_esn? 12 : 8;
+      auth_xform->auth.add_auth_data_length = use_esn ? 12 : 8;
       break;
     default:
       return -1;
-  }
+    }
 
   auth_xform->type = RTE_CRYPTO_SYM_XFORM_AUTH;
 
@@ -215,25 +167,26 @@ translate_integ_algo(ipsec_integ_alg_t integ_alg,
 }
 
 static_always_inline int
-create_sym_sess(ipsec_sa_t *sa, crypto_sa_session_t *sa_sess, u8 is_outbound)
+create_sym_sess (ipsec_sa_t * sa, crypto_sa_session_t * sa_sess,
+		 u8 is_outbound)
 {
-  u32 cpu_index = os_get_cpu_number();
-  dpdk_crypto_main_t * dcm = &dpdk_crypto_main;
+  u32 cpu_index = os_get_cpu_number ();
+  dpdk_crypto_main_t *dcm = &dpdk_crypto_main;
   crypto_worker_main_t *cwm = &dcm->workers_main[cpu_index];
-  struct rte_crypto_sym_xform cipher_xform = {0};
-  struct rte_crypto_sym_xform auth_xform = {0};
+  struct rte_crypto_sym_xform cipher_xform = { 0 };
+  struct rte_crypto_sym_xform auth_xform = { 0 };
   struct rte_crypto_sym_xform *xfs;
   uword key = 0, *data;
-  crypto_worker_qp_key_t *p_key = (crypto_worker_qp_key_t *)&key;
+  crypto_worker_qp_key_t *p_key = (crypto_worker_qp_key_t *) & key;
 
   if (sa->crypto_alg == IPSEC_CRYPTO_ALG_AES_GCM_128)
     {
       sa->crypto_key_len -= 4;
-      clib_memcpy(&sa->salt, &sa->crypto_key[sa->crypto_key_len], 4);
+      clib_memcpy (&sa->salt, &sa->crypto_key[sa->crypto_key_len], 4);
     }
   else
     {
-      sa->salt = (u32) rand();
+      sa->salt = (u32) rand ();
     }
 
   cipher_xform.type = RTE_CRYPTO_SYM_XFORM_CIPHER;
@@ -244,11 +197,11 @@ create_sym_sess(ipsec_sa_t *sa, crypto_sa_session_t *sa_sess, u8 is_outbound)
   auth_xform.auth.key.data = sa->integ_key;
   auth_xform.auth.key.length = sa->integ_key_len;
 
-  if (translate_crypto_algo(sa->crypto_alg, &cipher_xform) < 0)
+  if (translate_crypto_algo (sa->crypto_alg, &cipher_xform) < 0)
     return -1;
   p_key->cipher_algo = cipher_xform.cipher.algo;
 
-  if (translate_integ_algo(sa->integ_alg, &auth_xform, sa->use_esn) < 0)
+  if (translate_integ_algo (sa->integ_alg, &auth_xform, sa->use_esn) < 0)
     return -1;
   p_key->auth_algo = auth_xform.auth.algo;
 
@@ -269,17 +222,17 @@ create_sym_sess(ipsec_sa_t *sa, crypto_sa_session_t *sa_sess, u8 is_outbound)
 
   p_key->is_outbound = is_outbound;
 
-  data = hash_get(cwm->algo_qp_map, key);
+  data = hash_get (cwm->algo_qp_map, key);
   if (!data)
     return -1;
 
   sa_sess->sess =
-    rte_cryptodev_sym_session_create(cwm->qp_data[*data].dev_id, xfs);
+    rte_cryptodev_sym_session_create (cwm->qp_data[*data].dev_id, xfs);
 
   if (!sa_sess->sess)
     return -1;
 
-  sa_sess->qp_index = (u8)*data;
+  sa_sess->qp_index = (u8) * data;
 
   return 0;
 }
diff --git a/src/vnet/devices/dpdk/ipsec/esp_decrypt.c b/src/vnet/devices/dpdk/ipsec/esp_decrypt.c
index 89ab9f9b..53b2d122 100644
--- a/src/vnet/devices/dpdk/ipsec/esp_decrypt.c
+++ b/src/vnet/devices/dpdk/ipsec/esp_decrypt.c
@@ -22,6 +22,8 @@
 #include 
 #include 
 #include 
+#include 
+#include 
 
 #define foreach_esp_decrypt_next	       \
 _(DROP, "error-drop")			       \
@@ -189,7 +191,14 @@ dpdk_esp_decrypt_node_fn (vlib_main_t * vm,
 	  if (PREDICT_FALSE(!sa_sess->sess))
 	    {
 	      int ret = create_sym_sess(sa0, sa_sess, 0);
-	      ASSERT(ret == 0);
+
+	      if (PREDICT_FALSE (ret))
+		{
+		  to_next[0] = bi0;
+		  to_next += 1;
+		  n_left_to_next -= 1;
+		  goto trace;
+		}
 	    }
 
 	  sess = sa_sess->sess;
diff --git a/src/vnet/devices/dpdk/ipsec/esp_encrypt.c b/src/vnet/devices/dpdk/ipsec/esp_encrypt.c
index 10bb4616..b6f00004 100644
--- a/src/vnet/devices/dpdk/ipsec/esp_encrypt.c
+++ b/src/vnet/devices/dpdk/ipsec/esp_encrypt.c
@@ -22,6 +22,8 @@
 #include 
 #include 
 #include 
+#include 
+#include 
 
 #define foreach_esp_encrypt_next                   \
 _(DROP, "error-drop")                              \
@@ -179,7 +181,14 @@ dpdk_esp_encrypt_node_fn (vlib_main_t * vm,
 	  if (PREDICT_FALSE (!sa_sess->sess))
 	    {
 	      int ret = create_sym_sess (sa0, sa_sess, 1);
-	      ASSERT (ret == 0);
+
+	      if (PREDICT_FALSE (ret))
+		{
+		  to_next[0] = bi0;
+		  to_next += 1;
+		  n_left_to_next -= 1;
+		  goto trace;
+		}
 	    }
 
 	  qp_index = sa_sess->qp_index;
diff --git a/src/vnet/devices/dpdk/ipsec/ipsec.c b/src/vnet/devices/dpdk/ipsec/ipsec.c
index de253f02..05c17c99 100644
--- a/src/vnet/devices/dpdk/ipsec/ipsec.c
+++ b/src/vnet/devices/dpdk/ipsec/ipsec.c
@@ -15,24 +15,69 @@
 #include 
 #include 
 #include 
+#include 
+#include 
+
 #include 
 #include 
 #include 
-#include 
 
-#define DPDK_CRYPTO_NB_OBJS	  2048
+#define DPDK_CRYPTO_NB_SESS_OBJS  20000
 #define DPDK_CRYPTO_CACHE_SIZE	  512
 #define DPDK_CRYPTO_PRIV_SIZE	  128
-#define DPDK_CRYPTO_N_QUEUE_DESC  512
+#define DPDK_CRYPTO_N_QUEUE_DESC  1024
 #define DPDK_CRYPTO_NB_COPS	  (1024 * 4)
 
-/*
- * return:
- * -1: 	update failed
- * 0:	already exist
- * 1:	mapped
- */
 static int
+add_del_sa_sess (u32 sa_index, u8 is_add)
+{
+  dpdk_crypto_main_t *dcm = &dpdk_crypto_main;
+  crypto_worker_main_t *cwm;
+  u8 skip_master = vlib_num_workers () > 0;
+
+  /* *INDENT-OFF* */
+  vec_foreach (cwm, dcm->workers_main)
+    {
+      crypto_sa_session_t *sa_sess;
+      u8 is_outbound;
+
+      if (skip_master)
+	{
+	  skip_master = 0;
+	  continue;
+	}
+
+      for (is_outbound = 0; is_outbound < 2; is_outbound++)
+	{
+	  if (is_add)
+	    {
+	      pool_get (cwm->sa_sess_d[is_outbound], sa_sess);
+	    }
+	  else
+	    {
+	      u8 dev_id;
+
+	      sa_sess = pool_elt_at_index (cwm->sa_sess_d[is_outbound], sa_index);
+	      dev_id = cwm->qp_data[sa_sess->qp_index].dev_id;
+
+	      if (!sa_sess->sess)
+		continue;
+
+	      if (rte_cryptodev_sym_session_free(dev_id, sa_sess->sess))
+		{
+		  clib_warning("failed to free session");
+		  return -1;
+		}
+	      memset(sa_sess, 0, sizeof(sa_sess[0]));
+	    }
+	}
+    }
+  /* *INDENT-OFF* */
+
+  return 0;
+}
+
+static void
 update_qp_data (crypto_worker_main_t * cwm,
 		u8 cdev_id, u16 qp_id, u8 is_outbound, u16 * idx)
 {
@@ -45,7 +90,7 @@ update_qp_data (crypto_worker_main_t * cwm,
 
       if (qpd->dev_id == cdev_id && qpd->qp_id == qp_id &&
 	  qpd->is_outbound == is_outbound)
-	  return 0;
+	  return;
     }
   /* *INDENT-ON* */
 
@@ -54,13 +99,10 @@ update_qp_data (crypto_worker_main_t * cwm,
   qpd->dev_id = cdev_id;
   qpd->qp_id = qp_id;
   qpd->is_outbound = is_outbound;
-
-  return 1;
 }
 
 /*
  * return:
- * 	-1: error
  * 	0: already exist
  * 	1: mapped
  */
@@ -70,7 +112,6 @@ add_mapping (crypto_worker_main_t * cwm,
 	     const struct rte_cryptodev_capabilities *cipher_cap,
 	     const struct rte_cryptodev_capabilities *auth_cap)
 {
-  int mapped;
   u16 qp_index;
   uword key = 0, data, *ret;
   crypto_worker_qp_key_t *p_key = (crypto_worker_qp_key_t *) & key;
@@ -83,17 +124,12 @@ add_mapping (crypto_worker_main_t * cwm,
   if (ret)
     return 0;
 
-  mapped = update_qp_data (cwm, cdev_id, qp, is_outbound, &qp_index);
-  if (mapped < 0)
-    return -1;
+  update_qp_data (cwm, cdev_id, qp, is_outbound, &qp_index);
 
   data = (uword) qp_index;
+  hash_set (cwm->algo_qp_map, key, data);
 
-  ret = hash_set (cwm->algo_qp_map, key, data);
-  if (!ret)
-    rte_panic ("Failed to insert hash table\n");
-
-  return mapped;
+  return 1;
 }
 
 /*
@@ -120,19 +156,13 @@ add_cdev_mapping (crypto_worker_main_t * cwm,
       for (j = dev_info->capabilities; j->op != RTE_CRYPTO_OP_TYPE_UNDEFINED;
 	   j++)
 	{
-	  int status = 0;
-
 	  if (j->sym.xform_type != RTE_CRYPTO_SYM_XFORM_AUTH)
 	    continue;
 
 	  if (check_algo_is_supported (j, NULL) != 0)
 	    continue;
 
-	  status = add_mapping (cwm, cdev_id, qp, is_outbound, i, j);
-	  if (status == 1)
-	    mapped += 1;
-	  if (status < 0)
-	    return status;
+	  mapped |= add_mapping (cwm, cdev_id, qp, is_outbound, i, j);
 	}
     }
 
@@ -169,8 +199,33 @@ check_cryptodev_queues ()
 }
 
 static clib_error_t *
-dpdk_ipsec_init (vlib_main_t * vm)
+dpdk_ipsec_check_support (ipsec_sa_t * sa)
+{
+  if (sa->crypto_alg == IPSEC_CRYPTO_ALG_AES_GCM_128)
+    {
+      if (sa->integ_alg != IPSEC_INTEG_ALG_NONE)
+	return clib_error_return (0, "unsupported integ-alg %U with "
+				  "crypto-algo aes-gcm-128",
+				  format_ipsec_integ_alg, sa->integ_alg);
+      sa->integ_alg = IPSEC_INTEG_ALG_AES_GCM_128;
+    }
+  else
+    {
+      if (sa->integ_alg == IPSEC_INTEG_ALG_NONE ||
+	  sa->integ_alg == IPSEC_INTEG_ALG_AES_GCM_128)
+	return clib_error_return (0, "unsupported integ-alg %U",
+				  format_ipsec_integ_alg, sa->integ_alg);
+    }
+
+  return 0;
+}
+
+static uword
+dpdk_ipsec_process (vlib_main_t * vm, vlib_node_runtime_t * rt,
+		    vlib_frame_t * f)
 {
+  dpdk_config_main_t *conf = &dpdk_config_main;
+  ipsec_main_t *im = &ipsec_main;
   dpdk_crypto_main_t *dcm = &dpdk_crypto_main;
   vlib_thread_main_t *tm = vlib_get_thread_main ();
   struct rte_cryptodev_config dev_conf;
@@ -180,8 +235,19 @@ dpdk_ipsec_init (vlib_main_t * vm)
   i32 dev_id, ret;
   u32 i, skip_master;
 
+  if (!conf->cryptodev)
+    {
+      clib_warning ("DPDK Cryptodev support is disabled, "
+		    "default to OpenSSL IPsec");
+      return 0;
+    }
+
   if (check_cryptodev_queues () < 0)
-    return clib_error_return (0, "not enough cryptodevs for ipsec");
+    {
+      conf->cryptodev = 0;
+      clib_warning ("not enough Cryptodevs, default to OpenSSL IPsec");
+      return 0;
+    }
 
   vec_alloc (dcm->workers_main, tm->n_vlib_mains);
   _vec_len (dcm->workers_main) = tm->n_vlib_mains;
@@ -221,24 +287,17 @@ dpdk_ipsec_init (vlib_main_t * vm)
 	    {
 	      map = hash_create (0, sizeof (crypto_worker_qp_key_t));
 	      if (!map)
-		return clib_error_return (0, "unable to create hash table "
-					  "for worker %u",
-					  vlib_mains[i]->cpu_index);
+		{
+		  clib_warning ("unable to create hash table for worker %u",
+				vlib_mains[i]->cpu_index);
+		  goto error;
+		}
 	      cwm->algo_qp_map = map;
 	    }
 
 	  for (is_outbound = 0; is_outbound < 2 && qp < max_nb_qp;
 	       is_outbound++)
-	    {
-	      int mapped = add_cdev_mapping (cwm, &cdev_info,
-					     dev_id, qp, is_outbound);
-	      if (mapped > 0)
-		qp++;
-
-	      if (mapped < 0)
-		return clib_error_return (0,
-					  "too many queues for one worker");
-	    }
+	    qp += add_cdev_mapping (cwm, &cdev_info, dev_id, qp, is_outbound);
 	}
 
       if (qp == 0)
@@ -246,12 +305,15 @@ dpdk_ipsec_init (vlib_main_t * vm)
 
       dev_conf.socket_id = rte_cryptodev_socket_id (dev_id);
       dev_conf.nb_queue_pairs = cdev_info.max_nb_queue_pairs;
-      dev_conf.session_mp.nb_objs = DPDK_CRYPTO_NB_OBJS;
+      dev_conf.session_mp.nb_objs = DPDK_CRYPTO_NB_SESS_OBJS;
       dev_conf.session_mp.cache_size = DPDK_CRYPTO_CACHE_SIZE;
 
       ret = rte_cryptodev_configure (dev_id, &dev_conf);
       if (ret < 0)
-	return clib_error_return (0, "cryptodev %u config error", dev_id);
+	{
+	  clib_warning ("cryptodev %u config error", dev_id);
+	  goto error;
+	}
 
       qp_conf.nb_descriptors = DPDK_CRYPTO_N_QUEUE_DESC;
       for (qp = 0; qp < dev_conf.nb_queue_pairs; qp++)
@@ -259,37 +321,64 @@ dpdk_ipsec_init (vlib_main_t * vm)
 	  ret = rte_cryptodev_queue_pair_setup (dev_id, qp, &qp_conf,
 						dev_conf.socket_id);
 	  if (ret < 0)
-	    return clib_error_return (0, "cryptodev %u qp %u setup error",
-				      dev_id, qp);
+	    {
+	      clib_warning ("cryptodev %u qp %u setup error", dev_id, qp);
+	      goto error;
+	    }
 	}
-      fprintf (stdout, "%u\t%u\t%u\t%u\n", dev_id, dev_conf.nb_queue_pairs,
-	       DPDK_CRYPTO_NB_OBJS, DPDK_CRYPTO_CACHE_SIZE);
-    }
+      vec_validate_aligned (dcm->cop_pools, dev_conf.socket_id,
+			    CLIB_CACHE_LINE_BYTES);
 
-  u32 socket_id = rte_socket_id ();
+      if (!vec_elt (dcm->cop_pools, dev_conf.socket_id))
+	{
+	  u8 *pool_name = format (0, "crypto_op_pool_socket%u%c",
+				  dev_conf.socket_id, 0);
+
+	  rmp = rte_crypto_op_pool_create ((char *) pool_name,
+					   RTE_CRYPTO_OP_TYPE_SYMMETRIC,
+					   DPDK_CRYPTO_NB_COPS *
+					   (1 + vlib_num_workers ()),
+					   DPDK_CRYPTO_CACHE_SIZE,
+					   DPDK_CRYPTO_PRIV_SIZE,
+					   dev_conf.socket_id);
+	  vec_free (pool_name);
+
+	  if (!rmp)
+	    {
+	      clib_warning ("failed to allocate mempool on socket %u",
+			    dev_conf.socket_id);
+	      goto error;
+	    }
+	  vec_elt (dcm->cop_pools, dev_conf.socket_id) = rmp;
+	}
 
-  vec_validate_aligned (dcm->cop_pools, socket_id, CLIB_CACHE_LINE_BYTES);
+      fprintf (stdout, "%u\t%u\t%u\t%u\n", dev_id, dev_conf.nb_queue_pairs,
+	       DPDK_CRYPTO_NB_SESS_OBJS, DPDK_CRYPTO_CACHE_SIZE);
+    }
 
-  /* pool already exists, nothing to do */
-  if (dcm->cop_pools[socket_id])
-    return 0;
+  dpdk_esp_init ();
 
-  u8 *pool_name = format (0, "crypto_op_pool_socket%u%c", socket_id, 0);
+  /* Add new next node and set as default */
+  vlib_node_t *node, *next_node;
 
-  rmp = rte_crypto_op_pool_create ((char *) pool_name,
-				   RTE_CRYPTO_OP_TYPE_SYMMETRIC,
-				   DPDK_CRYPTO_NB_COPS *
-				   (1 + vlib_num_workers ()),
-				   DPDK_CRYPTO_CACHE_SIZE,
-				   DPDK_CRYPTO_PRIV_SIZE, socket_id);
-  vec_free (pool_name);
+  next_node = vlib_get_node_by_name (vm, (u8 *) "dpdk-esp-encrypt");
+  ASSERT (next_node);
+  node = vlib_get_node_by_name (vm, (u8 *) "ipsec-output-ip4");
+  ASSERT (node);
+  im->esp_encrypt_node_index = next_node->index;
+  im->esp_encrypt_next_index =
+    vlib_node_add_next (vm, node->index, next_node->index);
 
-  if (!rmp)
-    return clib_error_return (0, "failed to allocate mempool on socket %u",
-			      socket_id);
-  dcm->cop_pools[socket_id] = rmp;
+  next_node = vlib_get_node_by_name (vm, (u8 *) "dpdk-esp-decrypt");
+  ASSERT (next_node);
+  node = vlib_get_node_by_name (vm, (u8 *) "ipsec-input-ip4");
+  ASSERT (node);
+  im->esp_decrypt_node_index = next_node->index;
+  im->esp_decrypt_next_index =
+    vlib_node_add_next (vm, node->index, next_node->index);
 
-  dpdk_esp_init ();
+  im->cb.check_support_cb = dpdk_ipsec_check_support;
+  im->cb.add_del_sa_sess_cb = add_del_sa_sess;
 
   if (vec_len (vlib_mains) == 0)
     vlib_node_set_state (&vlib_global_main, dpdk_crypto_input_node.index,
@@ -299,10 +388,38 @@ dpdk_ipsec_init (vlib_main_t * vm)
       vlib_node_set_state (vlib_mains[i], dpdk_crypto_input_node.index,
 			   VLIB_NODE_STATE_POLLING);
 
+  /* TODO cryptodev counters */
+
+  return 0;
+
+error:
+  ;
+  crypto_worker_main_t *cwm;
+  struct rte_mempool **mp;
+  /* *INDENT-OFF* */
+  vec_foreach (cwm, dcm->workers_main)
+    hash_free (cwm->algo_qp_map);
+
+  vec_foreach (mp, dcm->cop_pools)
+    {
+      if (mp)
+	rte_mempool_free (mp[0]);
+    }
+  /* *INDENT-ON* */
+  vec_free (dcm->workers_main);
+  vec_free (dcm->cop_pools);
+
   return 0;
 }
 
-VLIB_MAIN_LOOP_ENTER_FUNCTION (dpdk_ipsec_init);
+/* *INDENT-OFF* */
+VLIB_REGISTER_NODE (dpdk_ipsec_process_node,static) = {
+    .function = dpdk_ipsec_process,
+    .type = VLIB_NODE_TYPE_PROCESS,
+    .name = "dpdk-ipsec-process",
+    .process_log2_n_stack_bytes = 17,
+};
+/* *INDENT-ON* */
 
 /*
  * fd.io coding-style-patch-verification: ON
diff --git a/src/vnet/devices/dpdk/ipsec/ipsec.h b/src/vnet/devices/dpdk/ipsec/ipsec.h
index e6c7498c..3465b361 100644
--- a/src/vnet/devices/dpdk/ipsec/ipsec.h
+++ b/src/vnet/devices/dpdk/ipsec/ipsec.h
@@ -167,7 +167,7 @@ check_algo_is_supported (const struct rte_cryptodev_capabilities *cap,
     .type = RTE_CRYPTO_SYM_XFORM_CIPHER,.cipher =
 	RTE_CRYPTO_CIPHER_3DES_CBC,.name = "3DES-CBC"},
     {
-    .type = RTE_CRYPTO_SYM_XFORM_CIPHER,.auth =
+    .type = RTE_CRYPTO_SYM_XFORM_CIPHER,.cipher =
 	RTE_CRYPTO_CIPHER_AES_GCM,.name = "AES-GCM"},
     {
     .type = RTE_CRYPTO_SYM_XFORM_AUTH,.auth =
diff --git a/src/vnet/ipsec-gre/interface.c b/src/vnet/ipsec-gre/interface.c
index 56832ee1..3b6e4ac2 100644
--- a/src/vnet/ipsec-gre/interface.c
+++ b/src/vnet/ipsec-gre/interface.c
@@ -28,13 +28,7 @@
 #include 
 #include 
 
-#if DPDK_CRYPTO==1
-#include 
-#define ESP_NODE "dpdk-esp-encrypt"
-#else
 #include 
-#define ESP_NODE "esp-encrypt"
-#endif
 
 u8 *
 format_ipsec_gre_tunnel (u8 * s, va_list * args)
@@ -193,7 +187,7 @@ vnet_ipsec_gre_add_del_tunnel (vnet_ipsec_gre_add_del_tunnel_args_t * a,
       hash_set (igm->tunnel_by_key, key, t - igm->tunnels);
 
       slot = vlib_node_add_named_next_with_slot
-	(vnm->vlib_main, hi->tx_node_index, ESP_NODE,
+	(vnm->vlib_main, hi->tx_node_index, "esp-encrypt",
 	 IPSEC_GRE_OUTPUT_NEXT_ESP_ENCRYPT);
 
       ASSERT (slot == IPSEC_GRE_OUTPUT_NEXT_ESP_ENCRYPT);
diff --git a/src/vnet/ipsec/ipsec.c b/src/vnet/ipsec/ipsec.c
index ee85c402..cfe434ab 100644
--- a/src/vnet/ipsec/ipsec.c
+++ b/src/vnet/ipsec/ipsec.c
@@ -22,23 +22,7 @@
 
 #include 
 #include 
-
-#if DPDK_CRYPTO==1
-#include 
-#define ESP_NODE "dpdk-esp-encrypt"
-#else
 #include 
-#define ESP_NODE "esp-encrypt"
-#endif
-
-#if DPDK_CRYPTO==0
-/* dummy function */
-static int
-add_del_sa_sess (u32 sa_index, u8 is_add)
-{
-  return 0;
-}
-#endif
 
 u32
 ipsec_get_sa_index_by_sa_id (u32 sa_id)
@@ -449,7 +433,9 @@ ipsec_add_del_sa (vlib_main_t * vm, ipsec_sa_t * new_sa, int is_add)
 	  return VNET_API_ERROR_SYSCALL_ERROR_1;	/* sa used in policy */
 	}
       hash_unset (im->sa_index_by_sa_id, sa->id);
-      add_del_sa_sess (sa_index, is_add);
+      if (im->cb.add_del_sa_sess_cb &&
+	  im->cb.add_del_sa_sess_cb (sa_index, is_add) < 0)
+	return VNET_API_ERROR_SYSCALL_ERROR_1;
       pool_put (im->sad, sa);
     }
   else				/* create new SA */
@@ -458,7 +444,8 @@ ipsec_add_del_sa (vlib_main_t * vm, ipsec_sa_t * new_sa, int is_add)
       clib_memcpy (sa, new_sa, sizeof (*sa));
       sa_index = sa - im->sad;
       hash_set (im->sa_index_by_sa_id, sa->id, sa_index);
-      if (add_del_sa_sess (sa_index, is_add) < 0)
+      if (im->cb.add_del_sa_sess_cb &&
+	  im->cb.add_del_sa_sess_cb (sa_index, is_add) < 0)
 	return VNET_API_ERROR_SYSCALL_ERROR_1;
     }
   return 0;
@@ -497,7 +484,8 @@ ipsec_set_sa_key (vlib_main_t * vm, ipsec_sa_t * sa_update)
 
   if (sa->crypto_key_len + sa->integ_key_len > 0)
     {
-      if (add_del_sa_sess (sa_index, 0) < 0)
+      if (im->cb.add_del_sa_sess_cb &&
+	  im->cb.add_del_sa_sess_cb (sa_index, 0) < 0)
 	return VNET_API_ERROR_SYSCALL_ERROR_1;
     }
 
@@ -521,6 +509,19 @@ ipsec_rand_seed (void)
   RAND_seed ((const void *) &seed_data, sizeof (seed_data));
 }
 
+static clib_error_t *
+ipsec_check_support (ipsec_sa_t * sa)
+{
+  if (sa->crypto_alg == IPSEC_CRYPTO_ALG_AES_GCM_128)
+    return clib_error_return (0, "unsupported aes-gcm-128 crypto-alg");
+  if (sa->integ_alg == IPSEC_INTEG_ALG_NONE)
+    return clib_error_return (0, "unsupported none integ-alg");
+  if (sa->integ_alg == IPSEC_INTEG_ALG_AES_GCM_128)
+    return clib_error_return (0, "unsupported aes-gcm-128 integ-alg");
+
+  return 0;
+}
+
 static clib_error_t *
 ipsec_init (vlib_main_t * vm)
 {
@@ -547,14 +548,18 @@ ipsec_init (vlib_main_t * vm)
   ASSERT (node);
   im->error_drop_node_index = node->index;
 
-  node = vlib_get_node_by_name (vm, (u8 *) ESP_NODE);
-
+  node = vlib_get_node_by_name (vm, (u8 *) "esp-encrypt");
   ASSERT (node);
   im->esp_encrypt_node_index = node->index;
 
-  node = vlib_get_node_by_name (vm, (u8 *) "ip4-lookup");
+  node = vlib_get_node_by_name (vm, (u8 *) "esp-decrypt");
   ASSERT (node);
-  im->ip4_lookup_node_index = node->index;
+  im->esp_decrypt_node_index = node->index;
+
+  im->esp_encrypt_next_index = IPSEC_OUTPUT_NEXT_ESP_ENCRYPT;
+  im->esp_decrypt_next_index = IPSEC_INPUT_NEXT_ESP_DECRYPT;
+
+  im->cb.check_support_cb = ipsec_check_support;
 
   if ((error = vlib_call_init_function (vm, ipsec_cli_init)))
     return error;
diff --git a/src/vnet/ipsec/ipsec.h b/src/vnet/ipsec/ipsec.h
index 32c7edfc..6726dba0 100644
--- a/src/vnet/ipsec/ipsec.h
+++ b/src/vnet/ipsec/ipsec.h
@@ -17,6 +17,33 @@
 
 #define IPSEC_FLAG_IPSEC_GRE_TUNNEL (1 << 0)
 
+
+#define foreach_ipsec_output_next                \
+_(DROP, "error-drop")                            \
+_(ESP_ENCRYPT, "esp-encrypt")
+
+#define _(v, s) IPSEC_OUTPUT_NEXT_##v,
+typedef enum
+{
+  foreach_ipsec_output_next
+#undef _
+    IPSEC_OUTPUT_N_NEXT,
+} ipsec_output_next_t;
+
+
+#define foreach_ipsec_input_next                \
+_(DROP, "error-drop")                           \
+_(ESP_DECRYPT, "esp-decrypt")
+
+#define _(v, s) IPSEC_INPUT_NEXT_##v,
+typedef enum
+{
+  foreach_ipsec_input_next
+#undef _
+    IPSEC_INPUT_N_NEXT,
+} ipsec_input_next_t;
+
+
 #define foreach_ipsec_policy_action \
   _(0, BYPASS,  "bypass")          \
   _(1, DISCARD, "discard")         \
@@ -31,20 +58,12 @@ typedef enum
     IPSEC_POLICY_N_ACTION,
 } ipsec_policy_action_t;
 
-#if DPDK_CRYPTO==1
 #define foreach_ipsec_crypto_alg \
   _(0, NONE,  "none")               \
   _(1, AES_CBC_128, "aes-cbc-128")  \
   _(2, AES_CBC_192, "aes-cbc-192")  \
   _(3, AES_CBC_256, "aes-cbc-256")  \
   _(4, AES_GCM_128, "aes-gcm-128")
-#else
-#define foreach_ipsec_crypto_alg \
-  _(0, NONE,  "none")               \
-  _(1, AES_CBC_128, "aes-cbc-128")  \
-  _(2, AES_CBC_192, "aes-cbc-192")  \
-  _(3, AES_CBC_256, "aes-cbc-256")
-#endif
 
 typedef enum
 {
@@ -54,7 +73,6 @@ typedef enum
     IPSEC_CRYPTO_N_ALG,
 } ipsec_crypto_alg_t;
 
-#if DPDK_CRYPTO==1
 #define foreach_ipsec_integ_alg \
   _(0, NONE,  "none")                                                     \
   _(1, MD5_96, "md5-96")           /* RFC2403 */                          \
@@ -63,17 +81,7 @@ typedef enum
   _(4, SHA_256_128, "sha-256-128") /* RFC4868 */                          \
   _(5, SHA_384_192, "sha-384-192") /* RFC4868 */                          \
   _(6, SHA_512_256, "sha-512-256") /* RFC4868 */                          \
-  _(7, AES_GCM_128, "aes-gcm-128")
-#else
-#define foreach_ipsec_integ_alg \
-  _(0, NONE,  "none")                                                     \
-  _(1, MD5_96, "md5-96")           /* RFC2403 */                          \
-  _(2, SHA1_96, "sha1-96")         /* RFC2404 */                          \
-  _(3, SHA_256_96, "sha-256-96")   /* draft-ietf-ipsec-ciph-sha-256-00 */ \
-  _(4, SHA_256_128, "sha-256-128") /* RFC4868 */                          \
-  _(5, SHA_384_192, "sha-384-192") /* RFC4868 */                          \
-  _(6, SHA_512_256, "sha-512-256")	/* RFC4868 */
-#endif
+  _(7, AES_GCM_128, "aes-gcm-128")	/* RFC4106 */
 
 typedef enum
 {
@@ -223,6 +231,12 @@ typedef struct
   u32 hw_if_index;
 } ipsec_tunnel_if_t;
 
+typedef struct
+{
+  i32 (*add_del_sa_sess_cb) (u32 sa_index, u8 is_add);
+  clib_error_t *(*check_support_cb) (ipsec_sa_t * sa);
+} ipsec_main_callbacks_t;
+
 typedef struct
 {
   /* pool of tunnel instances */
@@ -250,11 +264,16 @@ typedef struct
   uword *sa_index_by_sa_id;
   uword *ipsec_if_pool_index_by_key;
 
-  /* node indexes */
+  /* node indeces */
   u32 error_drop_node_index;
-  u32 ip4_lookup_node_index;
   u32 esp_encrypt_node_index;
+  u32 esp_decrypt_node_index;
+  /* next node indeces */
+  u32 esp_encrypt_next_index;
+  u32 esp_decrypt_next_index;
 
+  /* callbacks */
+  ipsec_main_callbacks_t cb;
 } ipsec_main_t;
 
 ipsec_main_t ipsec_main;
diff --git a/src/vnet/ipsec/ipsec_api.c b/src/vnet/ipsec/ipsec_api.c
index 9bcf63b4..30732266 100644
--- a/src/vnet/ipsec/ipsec_api.c
+++ b/src/vnet/ipsec/ipsec_api.c
@@ -177,6 +177,7 @@ static void vl_api_ipsec_sad_add_del_entry_t_handler
   vl_api_ipsec_sad_add_del_entry_reply_t *rmp;
   int rv;
 #if WITH_LIBSSL > 0
+  ipsec_main_t *im = &ipsec_main;
   ipsec_sa_t sa;
 
   memset (&sa, 0, sizeof (sa));
@@ -204,11 +205,7 @@ static void vl_api_ipsec_sad_add_del_entry_t_handler
   sa.crypto_key_len = mp->crypto_key_length;
   clib_memcpy (&sa.crypto_key, mp->crypto_key, sizeof (sa.crypto_key));
   /* check for unsupported integ-alg */
-#if DPDK_CRYPTO==1
   if (mp->integrity_algorithm < IPSEC_INTEG_ALG_NONE ||
-#else
-  if (mp->integrity_algorithm < IPSEC_INTEG_ALG_SHA1_96 ||
-#endif
       mp->integrity_algorithm >= IPSEC_INTEG_N_ALG)
     {
       clib_warning ("unsupported integ-alg: '%U'", format_ipsec_integ_alg,
@@ -217,35 +214,6 @@ static void vl_api_ipsec_sad_add_del_entry_t_handler
       goto out;
     }
 
-#if DPDK_CRYPTO==1
-  /*Special cases, aes-gcm-128 encryption */
-  if (mp->crypto_algorithm == IPSEC_CRYPTO_ALG_AES_GCM_128)
-    {
-      if (mp->integrity_algorithm != IPSEC_INTEG_ALG_NONE
-	  && mp->integrity_algorithm != IPSEC_INTEG_ALG_AES_GCM_128)
-	{
-	  clib_warning
-	    ("unsupported: aes-gcm-128 crypto-alg needs none as integ-alg");
-	  rv = VNET_API_ERROR_UNIMPLEMENTED;
-	  goto out;
-	}
-      else			/*set integ-alg internally to aes-gcm-128 */
-	mp->integrity_algorithm = IPSEC_INTEG_ALG_AES_GCM_128;
-    }
-  else if (mp->integrity_algorithm == IPSEC_INTEG_ALG_AES_GCM_128)
-    {
-      clib_warning ("unsupported integ-alg: aes-gcm-128");
-      rv = VNET_API_ERROR_UNIMPLEMENTED;
-      goto out;
-    }
-  else if (mp->integrity_algorithm == IPSEC_INTEG_ALG_NONE)
-    {
-      clib_warning ("unsupported integ-alg: none");
-      rv = VNET_API_ERROR_UNIMPLEMENTED;
-      goto out;
-    }
-#endif
-
   sa.integ_alg = mp->integrity_algorithm;
   sa.integ_key_len = mp->integrity_key_length;
   clib_memcpy (&sa.integ_key, mp->integrity_key, sizeof (sa.integ_key));
@@ -263,6 +231,15 @@ static void vl_api_ipsec_sad_add_del_entry_t_handler
       clib_memcpy (&sa.tunnel_dst_addr.ip4.data, mp->tunnel_dst_address, 4);
     }
 
+  ASSERT (im->cb.check_support_cb);
+  clib_error_t *err = im->cb.check_support_cb (&sa);
+  if (err)
+    {
+      clib_warning ("%s", err->what);
+      rv = VNET_API_ERROR_UNIMPLEMENTED;
+      goto out;
+    }
+
   rv = ipsec_add_del_sa (vm, &sa, mp->is_add);
 #else
   rv = VNET_API_ERROR_UNIMPLEMENTED;
diff --git a/src/vnet/ipsec/ipsec_cli.c b/src/vnet/ipsec/ipsec_cli.c
index 7ab85d4a..3c1e26f2 100644
--- a/src/vnet/ipsec/ipsec_cli.c
+++ b/src/vnet/ipsec/ipsec_cli.c
@@ -67,10 +67,12 @@ ipsec_sa_add_del_command_fn (vlib_main_t * vm,
 			     unformat_input_t * input,
 			     vlib_cli_command_t * cmd)
 {
+  ipsec_main_t *im = &ipsec_main;
   unformat_input_t _line_input, *line_input = &_line_input;
   ipsec_sa_t sa;
   int is_add = ~0;
   u8 *ck = 0, *ik = 0;
+  clib_error_t *err = 0;
 
   memset (&sa, 0, sizeof (sa));
 
@@ -109,11 +111,7 @@ ipsec_sa_add_del_command_fn (vlib_main_t * vm,
       else if (unformat (line_input, "integ-alg %U", unformat_ipsec_integ_alg,
 			 &sa.integ_alg))
 	{
-#if DPDK_CRYPTO==1
-	  if (sa.integ_alg < IPSEC_INTEG_ALG_NONE ||
-#else
 	  if (sa.integ_alg < IPSEC_INTEG_ALG_SHA1_96 ||
-#endif
 	      sa.integ_alg >= IPSEC_INTEG_N_ALG)
 	    return clib_error_return (0, "unsupported integ-alg: '%U'",
 				      format_ipsec_integ_alg, sa.integ_alg);
@@ -141,23 +139,6 @@ ipsec_sa_add_del_command_fn (vlib_main_t * vm,
 				  format_unformat_error, line_input);
     }
 
-#if DPDK_CRYPTO==1
-  /*Special cases, aes-gcm-128 encryption */
-  if (sa.crypto_alg == IPSEC_CRYPTO_ALG_AES_GCM_128)
-    {
-      if (sa.integ_alg != IPSEC_INTEG_ALG_NONE
-	  && sa.integ_alg != IPSEC_INTEG_ALG_AES_GCM_128)
-	return clib_error_return (0,
-				  "unsupported: aes-gcm-128 crypto-alg needs none as integ-alg");
-      else			/*set integ-alg internally to aes-gcm-128 */
-	sa.integ_alg = IPSEC_INTEG_ALG_AES_GCM_128;
-    }
-  else if (sa.integ_alg == IPSEC_INTEG_ALG_AES_GCM_128)
-    return clib_error_return (0, "unsupported integ-alg: aes-gcm-128");
-  else if (sa.integ_alg == IPSEC_INTEG_ALG_NONE)
-    return clib_error_return (0, "unsupported integ-alg: none");
-#endif
-
   unformat_free (line_input);
 
   if (sa.crypto_key_len > sizeof (sa.crypto_key))
@@ -172,6 +153,14 @@ ipsec_sa_add_del_command_fn (vlib_main_t * vm,
   if (ik)
     strncpy ((char *) sa.integ_key, (char *) ik, sa.integ_key_len);
 
+  if (is_add)
+    {
+      ASSERT (im->cb.check_support_cb);
+      err = im->cb.check_support_cb (&sa);
+      if (err)
+	return err;
+    }
+
   ipsec_add_del_sa (vm, &sa, is_add);
 
   return 0;
diff --git a/src/vnet/ipsec/ipsec_if.c b/src/vnet/ipsec/ipsec_if.c
index a8da046f..ca6b0092 100644
--- a/src/vnet/ipsec/ipsec_if.c
+++ b/src/vnet/ipsec/ipsec_if.c
@@ -20,20 +20,7 @@
 #include 
 
 #include 
-#if DPDK_CRYPTO==1
-#include 
-#else
 #include 
-#endif
-
-#if DPDK_CRYPTO==0
-/* dummy function */
-static int
-add_del_sa_sess (u32 sa_index, u8 is_add)
-{
-  return 0;
-}
-#endif
 
 void vl_api_rpc_call_main_thread (void *fp, u8 * data, u32 data_length);
 
@@ -52,6 +39,39 @@ dummy_interface_tx (vlib_main_t * vm,
   return frame->n_vectors;
 }
 
+static clib_error_t *
+ipsec_admin_up_down_function (vnet_main_t * vnm, u32 hw_if_index, u32 flags)
+{
+  ipsec_main_t *im = &ipsec_main;
+  clib_error_t *err = 0;
+  ipsec_tunnel_if_t *t;
+  vnet_hw_interface_t *hi;
+  ipsec_sa_t *sa;
+
+  hi = vnet_get_hw_interface (vnm, hw_if_index);
+  if (flags & VNET_SW_INTERFACE_FLAG_ADMIN_UP)
+    {
+      t = pool_elt_at_index (im->tunnel_interfaces, hi->hw_instance);
+      ASSERT (im->cb.check_support_cb);
+      sa = pool_elt_at_index (im->sad, t->input_sa_index);
+      err = im->cb.check_support_cb (sa);
+      if (err)
+	return err;
+
+      sa = pool_elt_at_index (im->sad, t->output_sa_index);
+      err = im->cb.check_support_cb (sa);
+      if (err)
+	return err;
+
+      vnet_sw_interface_set_flags (vnm, hi->sw_if_index,
+				   VNET_HW_INTERFACE_FLAG_LINK_UP);
+    }
+  else
+    vnet_sw_interface_set_flags (vnm, hi->sw_if_index, 0 /* down */ );
+
+  return /* no error */ 0;
+}
+
 /* *INDENT-OFF* */
 VNET_DEVICE_CLASS (ipsec_device_class, static) =
 {
@@ -59,6 +79,7 @@ VNET_DEVICE_CLASS (ipsec_device_class, static) =
   .format_device_name = format_ipsec_name,
   .format_tx_trace = format_ipsec_if_output_trace,
   .tx_function = dummy_interface_tx,
+  .admin_up_down_function = ipsec_admin_up_down_function,
 };
 /* *INDENT-ON* */
 
@@ -138,7 +159,9 @@ ipsec_add_del_tunnel_if_internal (vnet_main_t * vnm,
 		       args->remote_crypto_key_len);
 	}
 
-      add_del_sa_sess (t->input_sa_index, args->is_add);
+      if (im->cb.add_del_sa_sess_cb &&
+	  im->cb.add_del_sa_sess_cb (t->input_sa_index, args->is_add) < 0)
+	return VNET_API_ERROR_SYSCALL_ERROR_1;
 
       pool_get (im->sad, sa);
       memset (sa, 0, sizeof (*sa));
@@ -165,7 +188,9 @@ ipsec_add_del_tunnel_if_internal (vnet_main_t * vnm,
 		       args->local_crypto_key_len);
 	}
 
-      add_del_sa_sess (t->output_sa_index, args->is_add);
+      if (im->cb.add_del_sa_sess_cb &&
+	  im->cb.add_del_sa_sess_cb (t->output_sa_index, args->is_add) < 0)
+	return VNET_API_ERROR_SYSCALL_ERROR_1;
 
       hash_set (im->ipsec_if_pool_index_by_key, key,
 		t - im->tunnel_interfaces);
@@ -211,14 +236,16 @@ ipsec_add_del_tunnel_if_internal (vnet_main_t * vnm,
       /* delete input and output SA */
       sa = pool_elt_at_index (im->sad, t->input_sa_index);
 
-      if (add_del_sa_sess (t->input_sa_index, args->is_add) < 0)
+      if (im->cb.add_del_sa_sess_cb &&
+	  im->cb.add_del_sa_sess_cb (t->input_sa_index, args->is_add) < 0)
 	return VNET_API_ERROR_SYSCALL_ERROR_1;
 
       pool_put (im->sad, sa);
 
       sa = pool_elt_at_index (im->sad, t->output_sa_index);
 
-      if (add_del_sa_sess (t->output_sa_index, args->is_add) < 0)
+      if (im->cb.add_del_sa_sess_cb &&
+	  im->cb.add_del_sa_sess_cb (t->output_sa_index, args->is_add) < 0)
 	return VNET_API_ERROR_SYSCALL_ERROR_1;
 
       pool_put (im->sad, sa);
@@ -310,7 +337,8 @@ ipsec_set_interface_key (vnet_main_t * vnm, u32 hw_if_index,
       sa->crypto_key_len = vec_len (key);
       clib_memcpy (sa->crypto_key, key, vec_len (key));
 
-      if (add_del_sa_sess (t->input_sa_index, 0) < 0)
+      if (im->cb.add_del_sa_sess_cb &&
+	  im->cb.add_del_sa_sess_cb (t->output_sa_index, 0) < 0)
 	return VNET_API_ERROR_SYSCALL_ERROR_1;
     }
   else if (type == IPSEC_IF_SET_KEY_TYPE_LOCAL_INTEG)
@@ -320,7 +348,8 @@ ipsec_set_interface_key (vnet_main_t * vnm, u32 hw_if_index,
       sa->integ_key_len = vec_len (key);
       clib_memcpy (sa->integ_key, key, vec_len (key));
 
-      if (add_del_sa_sess (t->output_sa_index, 0) < 0)
+      if (im->cb.add_del_sa_sess_cb &&
+	  im->cb.add_del_sa_sess_cb (t->output_sa_index, 0) < 0)
 	return VNET_API_ERROR_SYSCALL_ERROR_1;
     }
   else if (type == IPSEC_IF_SET_KEY_TYPE_REMOTE_CRYPTO)
@@ -330,7 +359,8 @@ ipsec_set_interface_key (vnet_main_t * vnm, u32 hw_if_index,
       sa->crypto_key_len = vec_len (key);
       clib_memcpy (sa->crypto_key, key, vec_len (key));
 
-      if (add_del_sa_sess (t->input_sa_index, 0) < 0)
+      if (im->cb.add_del_sa_sess_cb &&
+	  im->cb.add_del_sa_sess_cb (t->input_sa_index, 0) < 0)
 	return VNET_API_ERROR_SYSCALL_ERROR_1;
     }
   else if (type == IPSEC_IF_SET_KEY_TYPE_REMOTE_INTEG)
@@ -340,7 +370,8 @@ ipsec_set_interface_key (vnet_main_t * vnm, u32 hw_if_index,
       sa->integ_key_len = vec_len (key);
       clib_memcpy (sa->integ_key, key, vec_len (key));
 
-      if (add_del_sa_sess (t->output_sa_index, 0) < 0)
+      if (im->cb.add_del_sa_sess_cb &&
+	  im->cb.add_del_sa_sess_cb (t->input_sa_index, 0) < 0)
 	return VNET_API_ERROR_SYSCALL_ERROR_1;
     }
   else
diff --git a/src/vnet/ipsec/ipsec_if_in.c b/src/vnet/ipsec/ipsec_if_in.c
index db75ab92..bd2a9f78 100644
--- a/src/vnet/ipsec/ipsec_if_in.c
+++ b/src/vnet/ipsec/ipsec_if_in.c
@@ -22,12 +22,6 @@
 #include 
 #include 
 
-#if DPDK_CRYPTO==1
-#define ESP_NODE "dpdk-esp-decrypt"
-#else
-#define ESP_NODE "esp-decrypt"
-#endif
-
 /* Statistics (not really errors) */
 #define foreach_ipsec_if_input_error    \
 _(RX, "good packets received")
@@ -46,12 +40,6 @@ typedef enum
     IPSEC_IF_INPUT_N_ERROR,
 } ipsec_if_input_error_t;
 
-typedef enum
-{
-  IPSEC_IF_INPUT_NEXT_ESP_DECRYPT,
-  IPSEC_IF_INPUT_NEXT_DROP,
-  IPSEC_IF_INPUT_N_NEXT,
-} ipsec_if_input_next_t;
 
 typedef struct
 {
@@ -59,7 +47,6 @@ typedef struct
   u32 seq;
 } ipsec_if_input_trace_t;
 
-
 u8 *
 format_ipsec_if_input_trace (u8 * s, va_list * args)
 {
@@ -106,7 +93,7 @@ ipsec_if_input_node_fn (vlib_main_t * vm, vlib_node_runtime_t * node,
 	  ip0 = vlib_buffer_get_current (b0);
 	  esp0 = (esp_header_t *) ((u8 *) ip0 + ip4_header_bytes (ip0));
 
-	  next0 = IPSEC_IF_INPUT_NEXT_DROP;
+	  next0 = IPSEC_INPUT_NEXT_DROP;
 
 	  u64 key = (u64) ip0->src_address.as_u32 << 32 |
 	    (u64) clib_net_to_host_u32 (esp0->spi);
@@ -121,7 +108,7 @@ ipsec_if_input_node_fn (vlib_main_t * vm, vlib_node_runtime_t * node,
 	      vnet_buffer (b0)->ipsec.flags =
 		t->hw_if_index == ~0 ? IPSEC_FLAG_IPSEC_GRE_TUNNEL : 0;
 	      vlib_buffer_advance (b0, ip4_header_bytes (ip0));
-	      next0 = IPSEC_IF_INPUT_NEXT_ESP_DECRYPT;
+	      next0 = im->esp_decrypt_next_index;
 	    }
 
 	  if (PREDICT_FALSE (b0->flags & VLIB_BUFFER_IS_TRACED))
@@ -156,12 +143,7 @@ VLIB_REGISTER_NODE (ipsec_if_input_node) = {
   .n_errors = ARRAY_LEN(ipsec_if_input_error_strings),
   .error_strings = ipsec_if_input_error_strings,
 
-  .n_next_nodes = IPSEC_IF_INPUT_N_NEXT,
-
-  .next_nodes = {
-	[IPSEC_IF_INPUT_NEXT_ESP_DECRYPT] = ESP_NODE,
-	[IPSEC_IF_INPUT_NEXT_DROP] = "error-drop",
-  },
+  .sibling_of = "ipsec-input-ip4",
 };
 /* *INDENT-ON* */
 
diff --git a/src/vnet/ipsec/ipsec_if_out.c b/src/vnet/ipsec/ipsec_if_out.c
index 8f062828..62ff67ac 100644
--- a/src/vnet/ipsec/ipsec_if_out.c
+++ b/src/vnet/ipsec/ipsec_if_out.c
@@ -21,12 +21,6 @@
 
 #include 
 
-#if DPDK_CRYPTO==1
-#define ESP_NODE "dpdk-esp-encrypt"
-#else
-#define ESP_NODE "esp-encrypt"
-#endif
-
 /* Statistics (not really errors) */
 #define foreach_ipsec_if_output_error    \
 _(TX, "good packets transmitted")
@@ -45,12 +39,6 @@ typedef enum
     IPSEC_IF_OUTPUT_N_ERROR,
 } ipsec_if_output_error_t;
 
-typedef enum
-{
-  IPSEC_IF_OUTPUT_NEXT_ESP_ENCRYPT,
-  IPSEC_IF_OUTPUT_NEXT_DROP,
-  IPSEC_IF_OUTPUT_N_NEXT,
-} ipsec_if_output_next_t;
 
 typedef struct
 {
@@ -58,7 +46,6 @@ typedef struct
   u32 seq;
 } ipsec_if_output_trace_t;
 
-
 u8 *
 format_ipsec_if_output_trace (u8 * s, va_list * args)
 {
@@ -106,7 +93,7 @@ ipsec_if_output_node_fn (vlib_main_t * vm, vlib_node_runtime_t * node,
 	  hi0 = vnet_get_sup_hw_interface (vnm, sw_if_index0);
 	  t0 = pool_elt_at_index (im->tunnel_interfaces, hi0->dev_instance);
 	  vnet_buffer (b0)->ipsec.sad_index = t0->output_sa_index;
-	  next0 = IPSEC_IF_OUTPUT_NEXT_ESP_ENCRYPT;
+	  next0 = im->esp_encrypt_next_index;
 
 	  if (PREDICT_FALSE (b0->flags & VLIB_BUFFER_IS_TRACED))
 	    {
@@ -142,12 +129,7 @@ VLIB_REGISTER_NODE (ipsec_if_output_node) = {
   .n_errors = ARRAY_LEN(ipsec_if_output_error_strings),
   .error_strings = ipsec_if_output_error_strings,
 
-  .n_next_nodes = IPSEC_IF_OUTPUT_N_NEXT,
-
-  .next_nodes = {
-        [IPSEC_IF_OUTPUT_NEXT_ESP_ENCRYPT] = ESP_NODE,
-        [IPSEC_IF_OUTPUT_NEXT_DROP] = "error-drop",
-  },
+  .sibling_of = "ipsec-output-ip4",
 };
 /* *INDENT-ON* */
 
diff --git a/src/vnet/ipsec/ipsec_input.c b/src/vnet/ipsec/ipsec_input.c
index 4662c1a1..deaa7b7b 100644
--- a/src/vnet/ipsec/ipsec_input.c
+++ b/src/vnet/ipsec/ipsec_input.c
@@ -23,30 +23,10 @@
 #include 
 #include 
 
-#if DPDK_CRYPTO==1
-#define ESP_NODE "dpdk-esp-decrypt"
-#else
-#define ESP_NODE "esp-decrypt"
-#endif
-
-#define foreach_ipsec_input_next                \
-_(DROP, "error-drop")                           \
-_(ESP_DECRYPT, ESP_NODE)
-
-#define _(v, s) IPSEC_INPUT_NEXT_##v,
-typedef enum
-{
-  foreach_ipsec_input_next
-#undef _
-    IPSEC_INPUT_N_NEXT,
-} ipsec_input_next_t;
-
-
 #define foreach_ipsec_input_error                   \
  _(RX_PKTS, "IPSEC pkts received")                  \
  _(DECRYPTION_FAILED, "IPSEC decryption failed")
 
-
 typedef enum
 {
 #define _(sym,str) IPSEC_INPUT_ERROR_##sym,
@@ -262,7 +242,7 @@ ipsec_input_ip4_node_fn (vlib_main_t * vm,
 		  p0->counter.bytes += clib_net_to_host_u16 (ip0->length);
 		  vnet_buffer (b0)->ipsec.sad_index = p0->sa_index;
 		  vnet_buffer (b0)->ipsec.flags = 0;
-		  next0 = IPSEC_INPUT_NEXT_ESP_DECRYPT;
+		  next0 = im->esp_decrypt_next_index;
 		  vlib_buffer_advance (b0, ip4_header_bytes (ip0));
 		  goto trace0;
 		}
@@ -392,7 +372,7 @@ VLIB_NODE_FUNCTION_MULTIARCH (ipsec_input_ip4_node, ipsec_input_ip4_node_fn)
 		  p0->counter.bytes += header_size;
 		  vnet_buffer (b0)->ipsec.sad_index = p0->sa_index;
 		  vnet_buffer (b0)->ipsec.flags = 0;
-		  next0 = IPSEC_INPUT_NEXT_ESP_DECRYPT;
+		  next0 = im->esp_decrypt_next_index;
 		  vlib_buffer_advance (b0, header_size);
 		  goto trace0;
 		}
diff --git a/src/vnet/ipsec/ipsec_output.c b/src/vnet/ipsec/ipsec_output.c
index df93b5e4..1b8070d6 100644
--- a/src/vnet/ipsec/ipsec_output.c
+++ b/src/vnet/ipsec/ipsec_output.c
@@ -21,27 +21,8 @@
 
 #include 
 
-#if DPDK_CRYPTO==1
-#define ESP_NODE "dpdk-esp-encrypt"
-#else
-#define ESP_NODE "esp-encrypt"
-#endif
-
 #if WITH_LIBSSL > 0
 
-#define foreach_ipsec_output_next                \
-_(DROP, "error-drop")                            \
-_(ESP_ENCRYPT, ESP_NODE)
-
-#define _(v, s) IPSEC_OUTPUT_NEXT_##v,
-typedef enum
-{
-  foreach_ipsec_output_next
-#undef _
-    IPSEC_OUTPUT_N_NEXT,
-} ipsec_output_next_t;
-
-
 #define foreach_ipsec_output_error                   \
  _(RX_PKTS, "IPSec pkts received")                   \
  _(POLICY_DISCARD, "IPSec policy discard")           \
@@ -50,7 +31,6 @@ typedef enum
  _(POLICY_BYPASS, "IPSec policy bypass")             \
  _(ENCAPS_FAILED, "IPSec encapsulation failed")
 
-
 typedef enum
 {
 #define _(sym,str) IPSEC_OUTPUT_ERROR_##sym,
-- 
cgit 1.2.3-korg


From 006eb478bb7a14ba7ba4199fecce29ebc495fb9c Mon Sep 17 00:00:00 2001
From: Burt Silverman 
Date: Fri, 27 Jan 2017 15:29:54 -0500
Subject: Add files to CLEANFILES for robust make clean.

At the least, $(BUILT_SOURCES) should be added to CLEANFILES. Also
beneficial is $(api_DATA), and in the case of Java, *.files and *.h.
Also there is a vpp/app/version.h, and some grammar and lex files in
vppapigen.

Change-Id: Ic6d3f2d40ce65e1d9a8b88217fa1f36de393ebb4
Signed-off-by: Burt Silverman 
---
 src/Makefile.am                        | 3 +++
 src/examples/sample-plugin/Makefile.am | 2 ++
 src/plugins/Makefile.am                | 1 +
 src/vpp-api/java/Makefile.am           | 2 +-
 src/vpp.am                             | 2 ++
 src/vppapigen.am                       | 2 ++
 6 files changed, 11 insertions(+), 1 deletion(-)

(limited to 'src/Makefile.am')

diff --git a/src/Makefile.am b/src/Makefile.am
index 239afeac..7da86fcb 100644
--- a/src/Makefile.am
+++ b/src/Makefile.am
@@ -28,6 +28,7 @@ noinst_HEADERS =
 dist_bin_SCRIPTS =
 lib_LTLIBRARIES =
 BUILT_SOURCES =
+CLEANFILES =
 install-data-local:
 	@echo "Building vppctl command list..."
 	@DIR_SEARCH="$(srcdir)" ; \
@@ -115,3 +116,5 @@ BUILT_SOURCES += \
 
 endif # if ENABLE_VLIB
 endif # if ENABLE_SVM
+
+CLEANFILES += $(BUILT_SOURCES) $(api_DATA)
diff --git a/src/examples/sample-plugin/Makefile.am b/src/examples/sample-plugin/Makefile.am
index 47642055..a105afdd 100644
--- a/src/examples/sample-plugin/Makefile.am
+++ b/src/examples/sample-plugin/Makefile.am
@@ -54,3 +54,5 @@ sample_test_plugin_la_SOURCES = sample/sample_test.c sample/sample_plugin.api.h
 install-data-hook:
 	@(cd $(vpppluginsdir) && $(RM) $(vppplugins_LTLIBRARIES))
 	@(cd $(vppapitestpluginsdir) && $(RM) $(vppapitestplugins_LTLIBRARIES))
+
+CLEANFILES = $(BUILT_SOURCES)
diff --git a/src/plugins/Makefile.am b/src/plugins/Makefile.am
index f0c455a5..06b575d1 100644
--- a/src/plugins/Makefile.am
+++ b/src/plugins/Makefile.am
@@ -76,3 +76,4 @@ api_DATA = \
 BUILT_SOURCES += \
 	$(patsubst %.api,%.api.h,$(API_FILES))
 
+CLEANFILES = $(BUILT_SOURCES) $(api_DATA)
diff --git a/src/vpp-api/java/Makefile.am b/src/vpp-api/java/Makefile.am
index 3696e5bf..6eb50084 100644
--- a/src/vpp-api/java/Makefile.am
+++ b/src/vpp-api/java/Makefile.am
@@ -218,7 +218,7 @@ all-local: $(JAR_FILES)
 #
 # Cleanup
 #
-CLEANFILES = jvpp-registry.ok $(JAR_FILES) $(BUILT_SOURCES)
+CLEANFILES = jvpp-registry.ok $(JAR_FILES) $(BUILT_SOURCES) *.files */*.h
 
 clean-local:
 	rm -rf $(CLEANDIRS)
diff --git a/src/vpp.am b/src/vpp.am
index 0b605ec5..37466c60 100644
--- a/src/vpp.am
+++ b/src/vpp.am
@@ -132,4 +132,6 @@ bin_vpp_get_metrics_LDADD = \
   libvppinfra.la \
   -lpthread -lm -lrt
 
+CLEANFILES += vpp/app/version.h
+
 # vi:syntax=automake
diff --git a/src/vppapigen.am b/src/vppapigen.am
index 3207c83a..5c25e1ec 100644
--- a/src/vppapigen.am
+++ b/src/vppapigen.am
@@ -27,4 +27,6 @@ vppapigen_SOURCES = tools/vppapigen/gram.y tools/vppapigen/lex_e.c tools/vppapig
 vppapigen_LDADD = libvppinfra.la
 vppapigen_LDFLAGS = -static
 
+CLEANFILES += tools/vppapigen/gram.c tools/vppapigen/gram.h
+CLEANFILES += tools/vppapigen/lex_e.c
 # vi:syntax=automake
-- 
cgit 1.2.3-korg


From 6ca42d333b247eaee4995a7f779a43759a81909e Mon Sep 17 00:00:00 2001
From: Radu Nicolau 
Date: Thu, 16 Feb 2017 13:54:42 +0000
Subject: dpdk: updated build to automatically download Intel(R) Multi-Buffer
 Crypto for IPsec Library

Change-Id: I58182edb7b0d314bb6dfa1daf7b00012196fd3e1
Signed-off-by: Radu Nicolau 
---
 Makefile        |  2 +-
 dpdk/Makefile   | 22 ++++++++++++++++++++++
 src/Makefile.am |  2 +-
 3 files changed, 24 insertions(+), 2 deletions(-)

(limited to 'src/Makefile.am')

diff --git a/Makefile b/Makefile
index bbbb2acf..8122c6a4 100644
--- a/Makefile
+++ b/Makefile
@@ -39,7 +39,7 @@ endif
 DEB_DEPENDS  = curl build-essential autoconf automake bison libssl-dev ccache
 DEB_DEPENDS += debhelper dkms git libtool libganglia1-dev libapr1-dev dh-systemd
 DEB_DEPENDS += libconfuse-dev git-review exuberant-ctags cscope pkg-config
-DEB_DEPENDS += python-dev python-virtualenv python-pip lcov chrpath autoconf
+DEB_DEPENDS += python-dev python-virtualenv python-pip lcov chrpath autoconf nasm
 ifeq ($(OS_VERSION_ID),14.04)
 	DEB_DEPENDS += openjdk-8-jdk-headless
 else
diff --git a/dpdk/Makefile b/dpdk/Makefile
index bdd7993a..f8f1ca13 100644
--- a/dpdk/Makefile
+++ b/dpdk/Makefile
@@ -33,6 +33,12 @@ DPDK_16.07_TARBALL_MD5_CKSUM := 690a2bb570103e58d12f9806e8bf21be
 DPDK_16.11_TARBALL_MD5_CKSUM := 06c1c577795360719d0b4fafaeee21e9
 DPDK_17.02_TARBALL_MD5_CKSUM := 6b9f7387c35641f4e8dbba3e528f2376
 DPDK_SOURCE := $(B)/dpdk-$(DPDK_VERSION)
+ifeq ($(DPDK_CRYPTO_SW_PMD),y)
+AESNIMB_LIB_TARBALL := v0.44.tar.gz
+AESNIMB_LIB_TARBALL_URL := http://github.com/01org/intel-ipsec-mb/archive/$(AESNIMB_LIB_TARBALL)
+AESNIMB_LIB_SOURCE := $(B)/intel-ipsec-mb-0.44
+export AESNI_MULTI_BUFFER_LIB_PATH=$(AESNIMB_LIB_SOURCE)
+endif
 
 ifneq (,$(findstring clang,$(CC)))
 DPDK_CC=clang
@@ -144,6 +150,12 @@ $(CURDIR)/$(DPDK_TARBALL):
 		then cp $(DPDK_DOWNLOAD_DIR)/$(DPDK_TARBALL) $(CURDIR) ; \
 		else curl -o $(CURDIR)/$(DPDK_TARBALL) -LO $(DPDK_TAR_URL) ; \
 	fi
+ifeq ($(DPDK_CRYPTO_SW_PMD),y)
+	@if [ -e $(DPDK_DOWNLOAD_DIR)/$(AESNIMB_LIB_TARBALL) ] ; \
+		then cp $(DPDK_DOWNLOAD_DIR)/$(AESNIMB_LIB_TARBALL) $(CURDIR) ; \
+		else curl -o $(CURDIR)/$(AESNIMB_LIB_TARBALL) -LO $(AESNIMB_LIB_TARBALL_URL) ; \
+	fi
+endif
 	@rm -f $(B)/.download.ok
 
 $(B)/.download.ok: $(CURDIR)/$(DPDK_TARBALL)
@@ -160,6 +172,10 @@ download: $(B)/.download.ok
 $(B)/.extract.ok: $(B)/.download.ok
 	@echo --- extracting $(DPDK_TARBALL) ---
 	@tar --directory $(B) --extract --file $(CURDIR)/$(DPDK_TARBALL)
+ifeq ($(DPDK_CRYPTO_SW_PMD),y)
+	@echo --- extracting $(AESNIMB_LIB_TARBALL) ---
+	@tar --directory $(B) --extract --file $(CURDIR)/$(AESNIMB_LIB_TARBALL)
+endif
 	@touch $@
 
 .PHONY: extract
@@ -179,6 +195,9 @@ endif
 patch: $(B)/.patch.ok
 
 $(B)/.config.ok: $(B)/.patch.ok $(B)/custom-config
+ifeq ($(DPDK_CRYPTO_SW_PMD),y)
+	@make -C $(AESNIMB_LIB_SOURCE)
+endif
 	@make $(DPDK_MAKE_ARGS) config
 	@touch $@
 
@@ -188,6 +207,9 @@ config: $(B)/.config.ok
 $(B)/.build.ok: $(DPDK_SOURCE_FILES)
 	@if [ ! -e $(B)/.config.ok ] ; then echo 'Please run "make config" first' && false ; fi
 	@make $(DPDK_MAKE_ARGS) install
+ifeq ($(DPDK_CRYPTO_SW_PMD),y)
+	@cp $(AESNIMB_LIB_SOURCE)/libIPSec_MB.a $(I)/lib/
+endif
 	@touch $@
 
 .PHONY: build
diff --git a/src/Makefile.am b/src/Makefile.am
index 7da86fcb..a0e62dbd 100644
--- a/src/Makefile.am
+++ b/src/Makefile.am
@@ -49,7 +49,7 @@ else
 DPDK_LD_FLAGS = -Wl,--whole-archive,-l:libdpdk.a,--no-whole-archive,-lm,-ldl
 endif
 if WITH_DPDK_CRYPTO_SW
-DPDK_LD_ADD = -L$(AESNI_MULTI_BUFFER_LIB_PATH) -lIPSec_MB
+DPDK_LD_ADD = -lIPSec_MB
 endif
 if WITH_DPDK_MLX5_PMD
 DPDK_LD_FLAGS += -libverbs -lmlx5 -lnuma
-- 
cgit 1.2.3-korg


From 2e3677bb2085d4992f74156bdff8fe050ac9de24 Mon Sep 17 00:00:00 2001
From: Radu Nicolau 
Date: Mon, 20 Feb 2017 12:27:02 +0000
Subject: cryptodev:  Automatically download and build ISA-L Crypto library

Change-Id: I5454053461e6fb98e7f58f9562efde3590bb7cb5
Signed-off-by: Radu Nicolau 
---
 dpdk/Makefile   | 21 ++++++++++++++++++++-
 src/Makefile.am |  2 +-
 2 files changed, 21 insertions(+), 2 deletions(-)

(limited to 'src/Makefile.am')

diff --git a/dpdk/Makefile b/dpdk/Makefile
index f8f1ca13..f8c85c96 100644
--- a/dpdk/Makefile
+++ b/dpdk/Makefile
@@ -33,11 +33,14 @@ DPDK_16.07_TARBALL_MD5_CKSUM := 690a2bb570103e58d12f9806e8bf21be
 DPDK_16.11_TARBALL_MD5_CKSUM := 06c1c577795360719d0b4fafaeee21e9
 DPDK_17.02_TARBALL_MD5_CKSUM := 6b9f7387c35641f4e8dbba3e528f2376
 DPDK_SOURCE := $(B)/dpdk-$(DPDK_VERSION)
+
 ifeq ($(DPDK_CRYPTO_SW_PMD),y)
 AESNIMB_LIB_TARBALL := v0.44.tar.gz
 AESNIMB_LIB_TARBALL_URL := http://github.com/01org/intel-ipsec-mb/archive/$(AESNIMB_LIB_TARBALL)
 AESNIMB_LIB_SOURCE := $(B)/intel-ipsec-mb-0.44
-export AESNI_MULTI_BUFFER_LIB_PATH=$(AESNIMB_LIB_SOURCE)
+ISA_L_CRYPTO_LIB_TARBALL := isa_l_crypto.tar.gz
+ISA_L_CRYPTO_LIB_TARBALL_URL := http://github.com/01org/isa-l_crypto/archive/master.tar.gz
+ISA_L_CRYPTO_LIB_SOURCE := $(B)/isa-l_crypto-master
 endif
 
 ifneq (,$(findstring clang,$(CC)))
@@ -84,6 +87,12 @@ else
 DPDK_EXTRA_CFLAGS := -g -O0
 endif
 
+ifeq ($(DPDK_CRYPTO_SW_PMD),y)
+DPDK_EXTRA_CFLAGS += -I$(ISA_L_CRYPTO_LIB_SOURCE)
+DPDK_EXTRA_LDFLAGS += -L$(ISA_L_CRYPTO_LIB_SOURCE)/.libs
+DPDK_MAKE_EXTRA_ARGS += AESNI_MULTI_BUFFER_LIB_PATH=$(AESNIMB_LIB_SOURCE)
+endif
+
 # assemble DPDK make arguments
 DPDK_MAKE_ARGS := -C $(DPDK_SOURCE) -j $(JOBS) \
 	T=$(DPDK_TARGET) \
@@ -155,6 +164,10 @@ ifeq ($(DPDK_CRYPTO_SW_PMD),y)
 		then cp $(DPDK_DOWNLOAD_DIR)/$(AESNIMB_LIB_TARBALL) $(CURDIR) ; \
 		else curl -o $(CURDIR)/$(AESNIMB_LIB_TARBALL) -LO $(AESNIMB_LIB_TARBALL_URL) ; \
 	fi
+	@if [ -e $(DPDK_DOWNLOAD_DIR)/$(ISA_L_CRYPTO_LIB_TARBALL) ] ; \
+		then cp $(DPDK_DOWNLOAD_DIR)/$(ISA_L_CRYPTO_LIB_TARBALL) $(CURDIR) ; \
+		else curl -o $(CURDIR)/$(ISA_L_CRYPTO_LIB_TARBALL) -LO $(ISA_L_CRYPTO_LIB_TARBALL_URL) ; \
+	fi
 endif
 	@rm -f $(B)/.download.ok
 
@@ -175,6 +188,8 @@ $(B)/.extract.ok: $(B)/.download.ok
 ifeq ($(DPDK_CRYPTO_SW_PMD),y)
 	@echo --- extracting $(AESNIMB_LIB_TARBALL) ---
 	@tar --directory $(B) --extract --file $(CURDIR)/$(AESNIMB_LIB_TARBALL)
+	@echo --- extracting $(ISA_L_CRYPTO_LIB_TARBALL) ---
+	@tar --directory $(B) --extract --file $(CURDIR)/$(ISA_L_CRYPTO_LIB_TARBALL)
 endif
 	@touch $@
 
@@ -197,6 +212,9 @@ patch: $(B)/.patch.ok
 $(B)/.config.ok: $(B)/.patch.ok $(B)/custom-config
 ifeq ($(DPDK_CRYPTO_SW_PMD),y)
 	@make -C $(AESNIMB_LIB_SOURCE)
+	@cd $(ISA_L_CRYPTO_LIB_SOURCE) && ./autogen.sh && ./configure
+	@make -C $(ISA_L_CRYPTO_LIB_SOURCE)
+	@cp $(ISA_L_CRYPTO_LIB_SOURCE)/include $(ISA_L_CRYPTO_LIB_SOURCE)/isa-l_crypto -r
 endif
 	@make $(DPDK_MAKE_ARGS) config
 	@touch $@
@@ -209,6 +227,7 @@ $(B)/.build.ok: $(DPDK_SOURCE_FILES)
 	@make $(DPDK_MAKE_ARGS) install
 ifeq ($(DPDK_CRYPTO_SW_PMD),y)
 	@cp $(AESNIMB_LIB_SOURCE)/libIPSec_MB.a $(I)/lib/
+	@cp $(ISA_L_CRYPTO_LIB_SOURCE)/.libs/libisal_crypto.a $(I)/lib/
 endif
 	@touch $@
 
diff --git a/src/Makefile.am b/src/Makefile.am
index a0e62dbd..08feb29a 100644
--- a/src/Makefile.am
+++ b/src/Makefile.am
@@ -49,7 +49,7 @@ else
 DPDK_LD_FLAGS = -Wl,--whole-archive,-l:libdpdk.a,--no-whole-archive,-lm,-ldl
 endif
 if WITH_DPDK_CRYPTO_SW
-DPDK_LD_ADD = -lIPSec_MB
+DPDK_LD_ADD = -lIPSec_MB -lisal_crypto
 endif
 if WITH_DPDK_MLX5_PMD
 DPDK_LD_FLAGS += -libverbs -lmlx5 -lnuma
-- 
cgit 1.2.3-korg


From 68b0fb0c620c7451ef1a6380c43c39de6614db51 Mon Sep 17 00:00:00 2001
From: Dave Barach 
Date: Tue, 28 Feb 2017 15:15:56 -0500
Subject: VPP-598: tcp stack initial commit

Change-Id: I49e5ce0aae6e4ff634024387ceaf7dbc432a0351
Signed-off-by: Dave Barach 
Signed-off-by: Florin Coras 
---
 src/Makefile.am                                  |    1 +
 src/plugins/ioam/export-common/ioam_export.h     |    2 +-
 src/plugins/ioam/ipfixcollector/ipfixcollector.c |    2 +-
 src/plugins/ioam/lib-vxlan-gpe/ioam_transit.c    |    2 +-
 src/plugins/snat/in2out.c                        |   26 +-
 src/plugins/snat/out2in.c                        |   24 +-
 src/scripts/vnet/tcp                             |   18 +-
 src/scripts/vnet/udp                             |   19 +
 src/scripts/vnet/uri/tcp-setup.sh                |   39 +
 src/scripts/vnet/uri/tcp_server                  |    4 +
 src/scripts/vnet/uri/udp                         |   19 +
 src/svm.am                                       |   10 +-
 src/svm/ssvm.c                                   |   16 +
 src/svm/ssvm.h                                   |   18 +-
 src/svm/svm_fifo.c                               |  568 ++++++
 src/svm/svm_fifo.h                               |  157 ++
 src/svm/svm_fifo_segment.c                       |  193 ++
 src/svm/svm_fifo_segment.h                       |   89 +
 src/svm/test_svm_fifo1.c                         |  361 ++++
 src/uri.am                                       |   22 +
 src/uri/uri_tcp_test.c                           |  916 +++++++++
 src/uri/uri_udp_test.c                           |  553 ++++++
 src/uri/uri_udp_test2.c                          |  954 +++++++++
 src/uri/uritest.c                                |  484 +++++
 src/vlib/buffer.c                                |    2 +-
 src/vlib/buffer.h                                |   68 +
 src/vlibmemory/unix_shared_memory_queue.c        |   12 +-
 src/vlibmemory/unix_shared_memory_queue.h        |    2 +-
 src/vnet.am                                      |   66 +-
 src/vnet/api_errno.h                             |   21 +-
 src/vnet/bfd/bfd_udp.c                           |    4 +-
 src/vnet/buffer.h                                |   10 +
 src/vnet/classify/vnet_classify.c                |    4 +-
 src/vnet/dhcp/dhcp_proxy.h                       |    2 +-
 src/vnet/flow/flow_report.h                      |    2 +-
 src/vnet/ip/ip.h                                 |    4 +-
 src/vnet/ip/ip4.h                                |   42 +-
 src/vnet/ip/ip4_forward.c                        |  173 +-
 src/vnet/ip/ip4_packet.h                         |   26 +-
 src/vnet/ip/ip6.h                                |   44 +-
 src/vnet/ip/ip6_packet.h                         |   26 +-
 src/vnet/ip/punt.c                               |    2 +-
 src/vnet/ip/tcp_packet.h                         |  141 --
 src/vnet/ip/udp.h                                |  315 ---
 src/vnet/ip/udp_error.def                        |   21 -
 src/vnet/ip/udp_format.c                         |   91 -
 src/vnet/ip/udp_init.c                           |   71 -
 src/vnet/ip/udp_local.c                          |  645 ------
 src/vnet/ip/udp_packet.h                         |   65 -
 src/vnet/ip/udp_pg.c                             |  237 ---
 src/vnet/ipsec/ikev2.c                           |    2 +-
 src/vnet/ipsec/ikev2_cli.c                       |    2 +-
 src/vnet/ipsec/ikev2_crypto.c                    |    2 +-
 src/vnet/lisp-cp/packets.c                       |   65 +-
 src/vnet/lisp-cp/packets.h                       |   45 -
 src/vnet/lisp-gpe/interface.c                    |    2 +-
 src/vnet/lisp-gpe/lisp_gpe.h                     |    4 +-
 src/vnet/lisp-gpe/lisp_gpe_adjacency.c           |    2 +
 src/vnet/session/application.c                   |  343 ++++
 src/vnet/session/application.h                   |  120 ++
 src/vnet/session/application_interface.c         |  459 +++++
 src/vnet/session/application_interface.h         |  136 ++
 src/vnet/session/hashes.c                        |   28 +
 src/vnet/session/node.c                          |  435 ++++
 src/vnet/session/session.api                     |  429 ++++
 src/vnet/session/session.c                       | 1286 ++++++++++++
 src/vnet/session/session.h                       |  380 ++++
 src/vnet/session/session_api.c                   |  821 ++++++++
 src/vnet/session/session_cli.c                   |  189 ++
 src/vnet/session/transport.c                     |   64 +
 src/vnet/session/transport.h                     |  250 +++
 src/vnet/tcp/tcp.c                               |  708 +++++++
 src/vnet/tcp/tcp.h                               |  624 ++++++
 src/vnet/tcp/tcp_error.def                       |   35 +
 src/vnet/tcp/tcp_format.c                        |  136 ++
 src/vnet/tcp/tcp_input.c                         | 2316 ++++++++++++++++++++++
 src/vnet/tcp/tcp_newreno.c                       |   93 +
 src/vnet/tcp/tcp_output.c                        | 1412 +++++++++++++
 src/vnet/tcp/tcp_packet.h                        |  184 ++
 src/vnet/tcp/tcp_pg.c                            |  236 +++
 src/vnet/tcp/tcp_syn_filter4.c                   |  542 +++++
 src/vnet/tcp/tcp_timer.h                         |   29 +
 src/vnet/udp/builtin_server.c                    |  239 +++
 src/vnet/udp/udp.c                               |  342 ++++
 src/vnet/udp/udp.h                               |  362 ++++
 src/vnet/udp/udp_error.def                       |   21 +
 src/vnet/udp/udp_format.c                        |   91 +
 src/vnet/udp/udp_input.c                         |  314 +++
 src/vnet/udp/udp_local.c                         |  666 +++++++
 src/vnet/udp/udp_packet.h                        |   65 +
 src/vnet/udp/udp_pg.c                            |  237 +++
 src/vnet/vnet_all_api_h.h                        |    1 +
 src/vnet/vxlan-gpe/vxlan_gpe.h                   |    2 +-
 src/vnet/vxlan/vxlan.h                           |    2 +-
 src/vpp/api/vpe.api                              |    1 +
 src/vppinfra.am                                  |    5 +
 src/vppinfra/bihash_16_8.h                       |  103 +
 src/vppinfra/bihash_48_8.h                       |  116 ++
 src/vppinfra/tw_timer_16t_1w_2048sl.c            |   26 +
 src/vppinfra/tw_timer_16t_1w_2048sl.h            |   46 +
 100 files changed, 18737 insertions(+), 1874 deletions(-)
 create mode 100644 src/scripts/vnet/udp
 create mode 100755 src/scripts/vnet/uri/tcp-setup.sh
 create mode 100644 src/scripts/vnet/uri/tcp_server
 create mode 100644 src/scripts/vnet/uri/udp
 create mode 100644 src/svm/svm_fifo.c
 create mode 100644 src/svm/svm_fifo.h
 create mode 100644 src/svm/svm_fifo_segment.c
 create mode 100644 src/svm/svm_fifo_segment.h
 create mode 100644 src/svm/test_svm_fifo1.c
 create mode 100644 src/uri.am
 create mode 100644 src/uri/uri_tcp_test.c
 create mode 100644 src/uri/uri_udp_test.c
 create mode 100644 src/uri/uri_udp_test2.c
 create mode 100644 src/uri/uritest.c
 delete mode 100644 src/vnet/ip/tcp_packet.h
 delete mode 100644 src/vnet/ip/udp.h
 delete mode 100644 src/vnet/ip/udp_error.def
 delete mode 100644 src/vnet/ip/udp_format.c
 delete mode 100644 src/vnet/ip/udp_init.c
 delete mode 100644 src/vnet/ip/udp_local.c
 delete mode 100644 src/vnet/ip/udp_packet.h
 delete mode 100644 src/vnet/ip/udp_pg.c
 create mode 100644 src/vnet/session/application.c
 create mode 100644 src/vnet/session/application.h
 create mode 100644 src/vnet/session/application_interface.c
 create mode 100644 src/vnet/session/application_interface.h
 create mode 100644 src/vnet/session/hashes.c
 create mode 100644 src/vnet/session/node.c
 create mode 100644 src/vnet/session/session.api
 create mode 100644 src/vnet/session/session.c
 create mode 100644 src/vnet/session/session.h
 create mode 100644 src/vnet/session/session_api.c
 create mode 100644 src/vnet/session/session_cli.c
 create mode 100644 src/vnet/session/transport.c
 create mode 100644 src/vnet/session/transport.h
 create mode 100644 src/vnet/tcp/tcp.c
 create mode 100644 src/vnet/tcp/tcp.h
 create mode 100644 src/vnet/tcp/tcp_error.def
 create mode 100644 src/vnet/tcp/tcp_format.c
 create mode 100644 src/vnet/tcp/tcp_input.c
 create mode 100644 src/vnet/tcp/tcp_newreno.c
 create mode 100644 src/vnet/tcp/tcp_output.c
 create mode 100644 src/vnet/tcp/tcp_packet.h
 create mode 100644 src/vnet/tcp/tcp_pg.c
 create mode 100644 src/vnet/tcp/tcp_syn_filter4.c
 create mode 100644 src/vnet/tcp/tcp_timer.h
 create mode 100644 src/vnet/udp/builtin_server.c
 create mode 100644 src/vnet/udp/udp.c
 create mode 100644 src/vnet/udp/udp.h
 create mode 100644 src/vnet/udp/udp_error.def
 create mode 100644 src/vnet/udp/udp_format.c
 create mode 100644 src/vnet/udp/udp_input.c
 create mode 100644 src/vnet/udp/udp_local.c
 create mode 100644 src/vnet/udp/udp_packet.h
 create mode 100644 src/vnet/udp/udp_pg.c
 create mode 100644 src/vppinfra/bihash_16_8.h
 create mode 100644 src/vppinfra/bihash_48_8.h
 create mode 100644 src/vppinfra/tw_timer_16t_1w_2048sl.c
 create mode 100644 src/vppinfra/tw_timer_16t_1w_2048sl.h

(limited to 'src/Makefile.am')

diff --git a/src/Makefile.am b/src/Makefile.am
index 08feb29a..641707ed 100644
--- a/src/Makefile.am
+++ b/src/Makefile.am
@@ -88,6 +88,7 @@ include vlib-api.am
 include vnet.am
 include vpp.am
 include vpp-api-test.am
+include uri.am
 
 SUBDIRS += plugins
 
diff --git a/src/plugins/ioam/export-common/ioam_export.h b/src/plugins/ioam/export-common/ioam_export.h
index e84dab0b..dd48a93b 100644
--- a/src/plugins/ioam/export-common/ioam_export.h
+++ b/src/plugins/ioam/export-common/ioam_export.h
@@ -21,7 +21,7 @@
 #include 
 #include 
 #include 
-#include 
+#include 
 #include 
 
 #include 
diff --git a/src/plugins/ioam/ipfixcollector/ipfixcollector.c b/src/plugins/ioam/ipfixcollector/ipfixcollector.c
index 4ae47edc..71b934ec 100644
--- a/src/plugins/ioam/ipfixcollector/ipfixcollector.c
+++ b/src/plugins/ioam/ipfixcollector/ipfixcollector.c
@@ -15,7 +15,7 @@
 
 #include 
 #include 
-#include 
+#include 
 #include 
 
 ipfix_collector_main_t ipfix_collector_main;
diff --git a/src/plugins/ioam/lib-vxlan-gpe/ioam_transit.c b/src/plugins/ioam/lib-vxlan-gpe/ioam_transit.c
index b42c357c..f334c983 100644
--- a/src/plugins/ioam/lib-vxlan-gpe/ioam_transit.c
+++ b/src/plugins/ioam/lib-vxlan-gpe/ioam_transit.c
@@ -16,7 +16,7 @@
 #include 
 #include 
 #include 
-#include 
+#include 
 #include 
 #include 
 #include 
diff --git a/src/plugins/snat/in2out.c b/src/plugins/snat/in2out.c
index e30c913c..b4b7793d 100644
--- a/src/plugins/snat/in2out.c
+++ b/src/plugins/snat/in2out.c
@@ -689,12 +689,12 @@ snat_hairpinning (snat_main_t *sm,
                              ip4_header_t, dst_address);
       ip0->checksum = ip_csum_fold (sum0);
 
-      old_dst_port0 = tcp0->ports.dst;
+      old_dst_port0 = tcp0->dst;
       if (PREDICT_TRUE(new_dst_port0 != old_dst_port0))
         {
           if (PREDICT_TRUE(proto0 == SNAT_PROTOCOL_TCP))
             {
-              tcp0->ports.dst = new_dst_port0;
+              tcp0->dst = new_dst_port0;
               sum0 = tcp0->checksum;
               sum0 = ip_csum_update (sum0, old_dst_addr0, new_dst_addr0,
                                      ip4_header_t, dst_address);
@@ -872,9 +872,9 @@ snat_in2out_node_fn_inline (vlib_main_t * vm,
 
           if (PREDICT_TRUE(proto0 == SNAT_PROTOCOL_TCP))
             {
-              old_port0 = tcp0->ports.src;
-              tcp0->ports.src = s0->out2in.port;
-              new_port0 = tcp0->ports.src;
+              old_port0 = tcp0->src_port;
+              tcp0->src_port = s0->out2in.port;
+              new_port0 = tcp0->src_port;
 
               sum0 = tcp0->checksum;
               sum0 = ip_csum_update (sum0, old_addr0, new_addr0,
@@ -1012,9 +1012,9 @@ snat_in2out_node_fn_inline (vlib_main_t * vm,
 
           if (PREDICT_TRUE(proto1 == SNAT_PROTOCOL_TCP))
             {
-              old_port1 = tcp1->ports.src;
-              tcp1->ports.src = s1->out2in.port;
-              new_port1 = tcp1->ports.src;
+              old_port1 = tcp1->src_port;
+              tcp1->src_port = s1->out2in.port;
+              new_port1 = tcp1->src_port;
 
               sum1 = tcp1->checksum;
               sum1 = ip_csum_update (sum1, old_addr1, new_addr1,
@@ -1188,9 +1188,9 @@ snat_in2out_node_fn_inline (vlib_main_t * vm,
 
           if (PREDICT_TRUE(proto0 == SNAT_PROTOCOL_TCP))
             {
-              old_port0 = tcp0->ports.src;
-              tcp0->ports.src = s0->out2in.port;
-              new_port0 = tcp0->ports.src;
+              old_port0 = tcp0->src_port;
+              tcp0->src_port = s0->out2in.port;
+              new_port0 = tcp0->src_port;
 
               sum0 = tcp0->checksum;
               sum0 = ip_csum_update (sum0, old_addr0, new_addr0,
@@ -1667,8 +1667,8 @@ snat_in2out_fast_static_map_fn (vlib_main_t * vm,
             {
               if (PREDICT_TRUE(proto0 == SNAT_PROTOCOL_TCP))
                 {
-                  old_port0 = tcp0->ports.src;
-                  tcp0->ports.src = new_port0;
+                  old_port0 = tcp0->src_port;
+                  tcp0->src_port = new_port0;
 
                   sum0 = tcp0->checksum;
                   sum0 = ip_csum_update (sum0, old_addr0, new_addr0,
diff --git a/src/plugins/snat/out2in.c b/src/plugins/snat/out2in.c
index 328f5ba4..3bfc0aa3 100644
--- a/src/plugins/snat/out2in.c
+++ b/src/plugins/snat/out2in.c
@@ -19,7 +19,7 @@
 #include 
 
 #include 
-#include 
+#include 
 #include 
 #include 
 #include 
@@ -602,9 +602,9 @@ snat_out2in_node_fn (vlib_main_t * vm,
 
           if (PREDICT_TRUE(proto0 == SNAT_PROTOCOL_TCP))
             {
-              old_port0 = tcp0->ports.dst;
-              tcp0->ports.dst = s0->in2out.port;
-              new_port0 = tcp0->ports.dst;
+              old_port0 = tcp0->dst_port;
+              tcp0->dst_port = s0->in2out.port;
+              new_port0 = tcp0->dst_port;
 
               sum0 = tcp0->checksum;
               sum0 = ip_csum_update (sum0, old_addr0, new_addr0,
@@ -737,9 +737,9 @@ snat_out2in_node_fn (vlib_main_t * vm,
 
           if (PREDICT_TRUE(proto1 == SNAT_PROTOCOL_TCP))
             {
-              old_port1 = tcp1->ports.dst;
-              tcp1->ports.dst = s1->in2out.port;
-              new_port1 = tcp1->ports.dst;
+              old_port1 = tcp1->dst_port;
+              tcp1->dst_port = s1->in2out.port;
+              new_port1 = tcp1->dst_port;
 
               sum1 = tcp1->checksum;
               sum1 = ip_csum_update (sum1, old_addr1, new_addr1,
@@ -907,9 +907,9 @@ snat_out2in_node_fn (vlib_main_t * vm,
 
           if (PREDICT_TRUE(proto0 == SNAT_PROTOCOL_TCP))
             {
-              old_port0 = tcp0->ports.dst;
-              tcp0->ports.dst = s0->in2out.port;
-              new_port0 = tcp0->ports.dst;
+              old_port0 = tcp0->dst_port;
+              tcp0->dst_port = s0->in2out.port;
+              new_port0 = tcp0->dst_port;
 
               sum0 = tcp0->checksum;
               sum0 = ip_csum_update (sum0, old_addr0, new_addr0,
@@ -1369,8 +1369,8 @@ snat_out2in_fast_node_fn (vlib_main_t * vm,
             {
                if (PREDICT_TRUE(proto0 == SNAT_PROTOCOL_TCP))
                 {
-                  old_port0 = tcp0->ports.dst;
-                  tcp0->ports.dst = new_port0;
+                  old_port0 = tcp0->dst_port;
+                  tcp0->dst_port = new_port0;
 
                   sum0 = tcp0->checksum;
                   sum0 = ip_csum_update (sum0, old_addr0, new_addr0,
diff --git a/src/scripts/vnet/tcp b/src/scripts/vnet/tcp
index a2ee8b2d..b9c23c3a 100644
--- a/src/scripts/vnet/tcp
+++ b/src/scripts/vnet/tcp
@@ -1,16 +1,18 @@
+loop create
+set int ip address loop0 192.168.1.1/8
+set int state loop0 up
+
 packet-generator new {
   name x
-  limit 1
+  limit 2048
   node ip4-input
-  size 64-64
+  size 100-100
+  interface loop0
   no-recycle
   data {
-    TCP: 1.2.3.4 -> 5.6.7.8
-    TCP: 1234 -> 5678
+    TCP: 192.168.1.2 -> 192.168.1.1
+    TCP: 32415 -> 80
+    SYN
     incrementing 100
   }
 }
-
-tr add pg-input 100
-ip route 5.6.7.8/32 via local
-ip route 1.2.3.4/32 via local
diff --git a/src/scripts/vnet/udp b/src/scripts/vnet/udp
new file mode 100644
index 00000000..7dda1eec
--- /dev/null
+++ b/src/scripts/vnet/udp
@@ -0,0 +1,19 @@
+loop create
+set int ip address loop0 192.168.1.1/8
+set int state loop0 up
+
+packet-generator new {
+  name udp
+  limit 512
+  rate 1e4
+  node ip4-input
+  size 100-100
+  interface loop0
+  no-recycle
+  data {
+   UDP: 192.168.1.2 - 192.168.2.255 -> 192.168.1.1
+   UDP: 4321 -> 1234
+    length 72
+    incrementing 100
+  }
+}
diff --git a/src/scripts/vnet/uri/tcp-setup.sh b/src/scripts/vnet/uri/tcp-setup.sh
new file mode 100755
index 00000000..e0b01588
--- /dev/null
+++ b/src/scripts/vnet/uri/tcp-setup.sh
@@ -0,0 +1,39 @@
+#!/usr/bin/env bash
+
+function topo_setup
+{
+  ip netns add vppns1
+  ip link add veth_vpp1 type veth peer name vpp1
+  ip link set dev vpp1 up
+  ip link set dev veth_vpp1 up netns vppns1
+
+  ip netns exec vppns1                          \
+  bash -c "
+    ip link set dev lo up
+    ip addr add 6.0.1.2/24 dev veth_vpp1
+  "
+
+  ethtool --offload  vpp1 rx off tx off
+  ip netns exec vppns1 ethtool --offload veth_vpp1 rx off tx off
+
+}
+
+function topo_clean
+{
+  ip link del dev veth_vpp1 &> /dev/null
+  ip netns del vppns1 &> /dev/null
+}
+
+if [ "$1" == "clean" ] ; then
+  topo_clean
+    exit 0
+else
+  topo_setup
+fi
+
+# to test connectivity do:
+# sudo ip netns exec vppns1 telnet 6.0.1.1 1234
+# to push traffic to the server
+# dd if=/dev/zero bs=1024K count=512 | nc 6.0.1.1
+# to listen for incoming connection from vpp
+# nc -l 1234
diff --git a/src/scripts/vnet/uri/tcp_server b/src/scripts/vnet/uri/tcp_server
new file mode 100644
index 00000000..7f5a86de
--- /dev/null
+++ b/src/scripts/vnet/uri/tcp_server
@@ -0,0 +1,4 @@
+create host-interface name vpp1
+set int state host-vpp1 up
+set int ip address host-vpp1 6.0.1.1/24
+trace add af-packet-input 10
diff --git a/src/scripts/vnet/uri/udp b/src/scripts/vnet/uri/udp
new file mode 100644
index 00000000..ca13b83c
--- /dev/null
+++ b/src/scripts/vnet/uri/udp
@@ -0,0 +1,19 @@
+loop create
+set int ip address loop0 10.0.0.1/32
+set int state loop0 up
+
+packet-generator new {
+  name udp
+  limit 512
+  rate 1e4
+  node ip4-input
+  size 100-100
+  interface loop0
+  no-recycle
+  data {
+   UDP: 192.168.1.2 - 192.168.2.255 -> 192.168.1.1
+   UDP: 4321 -> 1234
+    length 72
+    incrementing 100
+  }
+}
diff --git a/src/svm.am b/src/svm.am
index 2cd385bd..442eba8e 100644
--- a/src/svm.am
+++ b/src/svm.am
@@ -13,13 +13,14 @@
 
 bin_PROGRAMS += svmtool svmdbtool
 
-nobase_include_HEADERS += svm/svm.h svm/ssvm.h svm/svmdb.h
+nobase_include_HEADERS += svm/svm.h svm/ssvm.h svm/svmdb.h \
+	svm/svm_fifo.h svm/svm_fifo_segment.h
 
 lib_LTLIBRARIES += libsvm.la libsvmdb.la
 
+libsvm_la_SOURCES = svm/svm.c svm/ssvm.c svm/svm_fifo.c svm/svm_fifo_segment.c
 libsvm_la_LIBADD = libvppinfra.la -lrt -lpthread
 libsvm_la_DEPENDENCIES = libvppinfra.la
-libsvm_la_SOURCES = svm/svm.c svm/ssvm.c
 
 svmtool_SOURCES = svm/svmtool.c
 svmtool_LDADD = libsvm.la libvppinfra.la -lpthread -lrt
@@ -31,4 +32,9 @@ libsvmdb_la_SOURCES = svm/svmdb.c
 svmdbtool_SOURCES = svm/svmdbtool.c
 svmdbtool_LDADD = libsvmdb.la libsvm.la libvppinfra.la -lpthread -lrt
 
+noinst_PROGRAMS += test_svm_fifo1
+test_svm_fifo1_SOURCES = svm/test_svm_fifo1.c
+test_svm_fifo1_LDADD = libsvm.la libvppinfra.la -lpthread -lrt
+test_svm_fifo1_LDFLAGS = -static
+
 # vi:syntax=automake
diff --git a/src/svm/ssvm.c b/src/svm/ssvm.c
index 6f409eb6..6cda1f27 100644
--- a/src/svm/ssvm.c
+++ b/src/svm/ssvm.c
@@ -169,6 +169,22 @@ re_map_it:
   return 0;
 }
 
+void
+ssvm_delete (ssvm_private_t * ssvm)
+{
+  u8 *fn;
+
+  fn = format (0, "/dev/shm/%s%c", ssvm->name, 0);
+
+  /* Throw away the backing file */
+  if (unlink ((char *) fn) < 0)
+    clib_unix_warning ("unlink segment '%s'", ssvm->name);
+
+  munmap ((void *) ssvm->requested_va, ssvm->ssvm_size);
+  vec_free (fn);
+}
+
+
 /*
  * fd.io coding-style-patch-verification: ON
  *
diff --git a/src/svm/ssvm.h b/src/svm/ssvm.h
index 9e61b9a0..bccfc164 100644
--- a/src/svm/ssvm.h
+++ b/src/svm/ssvm.h
@@ -38,7 +38,10 @@
 #include 
 #include 
 
-#define MMAP_PAGESIZE (4<<10)
+#ifndef MMAP_PAGESIZE
+#define MMAP_PAGESIZE (clib_mem_get_page_size())
+#endif
+
 #define SSVM_N_OPAQUE 7
 
 typedef struct
@@ -125,12 +128,12 @@ ssvm_pop_heap (void *oldheap)
 }
 
 #define foreach_ssvm_api_error                  \
-_(NO_NAME, "No shared segment name", -10)       \
-_(NO_SIZE, "Size not set (master)", -11)        \
-_(CREATE_FAILURE, "Create failed", -12)		\
-_(SET_SIZE, "Set size failed", -13)		\
-_(MMAP, "mmap failed", -14)			\
-_(SLAVE_TIMEOUT, "Slave map timeout", -15)
+_(NO_NAME, "No shared segment name", -100)      \
+_(NO_SIZE, "Size not set (master)", -101)       \
+_(CREATE_FAILURE, "Create failed", -102)        \
+_(SET_SIZE, "Set size failed", -103)		\
+_(MMAP, "mmap failed", -104)			\
+_(SLAVE_TIMEOUT, "Slave map timeout", -105)
 
 typedef enum
 {
@@ -143,6 +146,7 @@ typedef enum
 
 int ssvm_master_init (ssvm_private_t * ssvm, u32 master_index);
 int ssvm_slave_init (ssvm_private_t * ssvm, int timeout_in_seconds);
+void ssvm_delete (ssvm_private_t * ssvm);
 
 #endif /* __included_ssvm_h__ */
 
diff --git a/src/svm/svm_fifo.c b/src/svm/svm_fifo.c
new file mode 100644
index 00000000..11f90193
--- /dev/null
+++ b/src/svm/svm_fifo.c
@@ -0,0 +1,568 @@
+/*
+ * Copyright (c) 2016 Cisco and/or its affiliates.
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at:
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#include "svm_fifo.h"
+
+/** create an svm fifo, in the current heap. Fails vs blow up the process */
+svm_fifo_t *
+svm_fifo_create (u32 data_size_in_bytes)
+{
+  svm_fifo_t *f;
+  pthread_mutexattr_t attr;
+  pthread_condattr_t cattr;
+
+  f = clib_mem_alloc_aligned_or_null (sizeof (*f) + data_size_in_bytes,
+				      CLIB_CACHE_LINE_BYTES);
+  if (f == 0)
+    return 0;
+
+  memset (f, 0, sizeof (*f) + data_size_in_bytes);
+  f->nitems = data_size_in_bytes;
+  f->ooos_list_head = OOO_SEGMENT_INVALID_INDEX;
+
+  memset (&attr, 0, sizeof (attr));
+  memset (&cattr, 0, sizeof (cattr));
+
+  if (pthread_mutexattr_init (&attr))
+    clib_unix_warning ("mutexattr_init");
+  if (pthread_mutexattr_setpshared (&attr, PTHREAD_PROCESS_SHARED))
+    clib_unix_warning ("pthread_mutexattr_setpshared");
+  if (pthread_mutex_init (&f->mutex, &attr))
+    clib_unix_warning ("mutex_init");
+  if (pthread_mutexattr_destroy (&attr))
+    clib_unix_warning ("mutexattr_destroy");
+  if (pthread_condattr_init (&cattr))
+    clib_unix_warning ("condattr_init");
+  if (pthread_condattr_setpshared (&cattr, PTHREAD_PROCESS_SHARED))
+    clib_unix_warning ("condattr_setpshared");
+  if (pthread_cond_init (&f->condvar, &cattr))
+    clib_unix_warning ("cond_init1");
+  if (pthread_condattr_destroy (&cattr))
+    clib_unix_warning ("cond_init2");
+
+  return (f);
+}
+
+always_inline ooo_segment_t *
+ooo_segment_new (svm_fifo_t * f, u32 start, u32 length)
+{
+  ooo_segment_t *s;
+
+  pool_get (f->ooo_segments, s);
+
+  s->fifo_position = start;
+  s->length = length;
+
+  s->prev = s->next = OOO_SEGMENT_INVALID_INDEX;
+
+  return s;
+}
+
+always_inline void
+ooo_segment_del (svm_fifo_t * f, u32 index)
+{
+  ooo_segment_t *cur, *prev = 0, *next = 0;
+  cur = pool_elt_at_index (f->ooo_segments, index);
+
+  if (cur->next != OOO_SEGMENT_INVALID_INDEX)
+    {
+      next = pool_elt_at_index (f->ooo_segments, cur->next);
+      next->prev = cur->prev;
+    }
+
+  if (cur->prev != OOO_SEGMENT_INVALID_INDEX)
+    {
+      prev = pool_elt_at_index (f->ooo_segments, cur->prev);
+      prev->next = cur->next;
+    }
+  else
+    {
+      f->ooos_list_head = cur->next;
+    }
+
+  pool_put (f->ooo_segments, cur);
+}
+
+/**
+ * Add segment to fifo's out-of-order segment list. Takes care of merging
+ * adjacent segments and removing overlapping ones.
+ */
+static void
+ooo_segment_add (svm_fifo_t * f, u32 offset, u32 length)
+{
+  ooo_segment_t *s, *new_s, *prev, *next, *it;
+  u32 new_index, position, end_offset, s_sof, s_eof, s_index;
+
+  position = (f->tail + offset) % f->nitems;
+  end_offset = offset + length;
+
+  if (f->ooos_list_head == OOO_SEGMENT_INVALID_INDEX)
+    {
+      s = ooo_segment_new (f, position, length);
+      f->ooos_list_head = s - f->ooo_segments;
+      f->ooos_newest = f->ooos_list_head;
+      return;
+    }
+
+  /* Find first segment that starts after new segment */
+  s = pool_elt_at_index (f->ooo_segments, f->ooos_list_head);
+  while (s->next != OOO_SEGMENT_INVALID_INDEX
+	 && ooo_segment_offset (f, s) <= offset)
+    s = pool_elt_at_index (f->ooo_segments, s->next);
+
+  s_index = s - f->ooo_segments;
+  s_sof = ooo_segment_offset (f, s);
+  s_eof = ooo_segment_end_offset (f, s);
+
+  /* No overlap, add before current segment */
+  if (end_offset < s_sof)
+    {
+      new_s = ooo_segment_new (f, position, length);
+      new_index = new_s - f->ooo_segments;
+
+      /* Pool might've moved, get segment again */
+      s = pool_elt_at_index (f->ooo_segments, s_index);
+
+      if (s->prev != OOO_SEGMENT_INVALID_INDEX)
+	{
+	  new_s->prev = s->prev;
+
+	  prev = pool_elt_at_index (f->ooo_segments, new_s->prev);
+	  prev->next = new_index;
+	}
+      else
+	{
+	  /* New head */
+	  f->ooos_list_head = new_index;
+	}
+
+      new_s->next = s - f->ooo_segments;
+      s->prev = new_index;
+      f->ooos_newest = new_index;
+      return;
+    }
+  /* No overlap, add after current segment */
+  else if (s_eof < offset)
+    {
+      new_s = ooo_segment_new (f, position, length);
+      new_index = new_s - f->ooo_segments;
+
+      /* Pool might've moved, get segment again */
+      s = pool_elt_at_index (f->ooo_segments, s_index);
+
+      if (s->next != OOO_SEGMENT_INVALID_INDEX)
+	{
+	  new_s->next = s->next;
+
+	  next = pool_elt_at_index (f->ooo_segments, new_s->next);
+	  next->prev = new_index;
+	}
+
+      new_s->prev = s - f->ooo_segments;
+      s->next = new_index;
+      f->ooos_newest = new_index;
+
+      return;
+    }
+
+  /*
+   * Merge needed
+   */
+
+  /* Merge at head */
+  if (offset <= s_sof)
+    {
+      /* If we have a previous, check if we overlap */
+      if (s->prev != OOO_SEGMENT_INVALID_INDEX)
+	{
+	  prev = pool_elt_at_index (f->ooo_segments, s->prev);
+
+	  /* New segment merges prev and current. Remove previous and
+	   * update position of current. */
+	  if (ooo_segment_end_offset (f, prev) >= offset)
+	    {
+	      s->fifo_position = prev->fifo_position;
+	      s->length = s_eof - ooo_segment_offset (f, prev);
+	      ooo_segment_del (f, s->prev);
+	    }
+	}
+      else
+	{
+	  s->fifo_position = position;
+	  s->length = s_eof - ooo_segment_offset (f, s);
+	}
+
+      /* The new segment's tail may cover multiple smaller ones */
+      if (s_eof < end_offset)
+	{
+	  /* Remove segments completely covered */
+	  it = (s->next != OOO_SEGMENT_INVALID_INDEX) ?
+	    pool_elt_at_index (f->ooo_segments, s->next) : 0;
+	  while (it && ooo_segment_end_offset (f, it) < end_offset)
+	    {
+	      next = (it->next != OOO_SEGMENT_INVALID_INDEX) ?
+		pool_elt_at_index (f->ooo_segments, it->next) : 0;
+	      ooo_segment_del (f, it - f->ooo_segments);
+	      it = next;
+	    }
+
+	  /* Update length. Segment's start might have changed. */
+	  s->length = end_offset - ooo_segment_offset (f, s);
+
+	  /* If partial overlap with last, merge */
+	  if (it && ooo_segment_offset (f, it) < end_offset)
+	    {
+	      s->length +=
+		it->length - (ooo_segment_offset (f, it) - end_offset);
+	      ooo_segment_del (f, it - f->ooo_segments);
+	    }
+	}
+    }
+  /* Last but overlapping previous */
+  else if (s_eof <= end_offset)
+    {
+      s->length = end_offset - ooo_segment_offset (f, s);
+    }
+  /* New segment completely covered by current one */
+  else
+    {
+      /* Do Nothing */
+    }
+
+  /* Most recently updated segment */
+  f->ooos_newest = s - f->ooo_segments;
+}
+
+/**
+ * Removes segments that can now be enqueued because the fifo's tail has
+ * advanced. Returns the number of bytes added to tail.
+ */
+static int
+ooo_segment_try_collect (svm_fifo_t * f, u32 n_bytes_enqueued)
+{
+  ooo_segment_t *s;
+  u32 index, bytes = 0, diff;
+
+  s = pool_elt_at_index (f->ooo_segments, f->ooos_list_head);
+
+  /* If last tail update overlaps one/multiple ooo segments, remove them */
+  diff = (f->nitems + f->tail - s->fifo_position) % f->nitems;
+  while (0 < diff && diff < n_bytes_enqueued)
+    {
+      /* Segment end is beyond the tail. Advance tail and be done */
+      if (diff < s->length)
+	{
+	  f->tail += s->length - diff;
+	  f->tail %= f->nitems;
+	  break;
+	}
+      /* If we have next go on */
+      else if (s->next != OOO_SEGMENT_INVALID_INDEX)
+	{
+	  index = s - f->ooo_segments;
+	  s = pool_elt_at_index (f->ooo_segments, s->next);
+	  diff = (f->nitems + f->tail - s->fifo_position) % f->nitems;
+	  ooo_segment_del (f, index);
+	}
+      /* End of search */
+      else
+	{
+	  break;
+	}
+    }
+
+  /* If tail is adjacent to an ooo segment, 'consume' it */
+  if (diff == 0)
+    {
+      bytes = ((f->nitems - f->cursize) >= s->length) ? s->length :
+	f->nitems - f->cursize;
+
+      f->tail += bytes;
+      f->tail %= f->nitems;
+
+      ooo_segment_del (f, s - f->ooo_segments);
+    }
+
+  return bytes;
+}
+
+static int
+svm_fifo_enqueue_internal (svm_fifo_t * f,
+			   int pid, u32 max_bytes, u8 * copy_from_here)
+{
+  u32 total_copy_bytes, first_copy_bytes, second_copy_bytes;
+  u32 cursize, nitems;
+
+  if (PREDICT_FALSE (f->cursize == f->nitems))
+    return -2;			/* fifo stuffed */
+
+  /* read cursize, which can only decrease while we're working */
+  cursize = f->cursize;
+  nitems = f->nitems;
+
+  /* Number of bytes we're going to copy */
+  total_copy_bytes = (nitems - cursize) < max_bytes ?
+    (nitems - cursize) : max_bytes;
+
+  if (PREDICT_TRUE (copy_from_here != 0))
+    {
+      /* Number of bytes in first copy segment */
+      first_copy_bytes = ((nitems - f->tail) < total_copy_bytes)
+	? (nitems - f->tail) : total_copy_bytes;
+
+      clib_memcpy (&f->data[f->tail], copy_from_here, first_copy_bytes);
+      f->tail += first_copy_bytes;
+      f->tail = (f->tail == nitems) ? 0 : f->tail;
+
+      /* Number of bytes in second copy segment, if any */
+      second_copy_bytes = total_copy_bytes - first_copy_bytes;
+      if (second_copy_bytes)
+	{
+	  clib_memcpy (&f->data[f->tail], copy_from_here + first_copy_bytes,
+		       second_copy_bytes);
+	  f->tail += second_copy_bytes;
+	  f->tail = (f->tail == nitems) ? 0 : f->tail;
+	}
+    }
+  else
+    {
+      /* Account for a zero-copy enqueue done elsewhere */
+      ASSERT (max_bytes <= (nitems - cursize));
+      f->tail += max_bytes;
+      f->tail = f->tail % nitems;
+      total_copy_bytes = max_bytes;
+    }
+
+  /* Any out-of-order segments to collect? */
+  if (PREDICT_FALSE (f->ooos_list_head != OOO_SEGMENT_INVALID_INDEX))
+    total_copy_bytes += ooo_segment_try_collect (f, total_copy_bytes);
+
+  /* Atomically increase the queue length */
+  __sync_fetch_and_add (&f->cursize, total_copy_bytes);
+
+  return (total_copy_bytes);
+}
+
+int
+svm_fifo_enqueue_nowait (svm_fifo_t * f,
+			 int pid, u32 max_bytes, u8 * copy_from_here)
+{
+  return svm_fifo_enqueue_internal (f, pid, max_bytes, copy_from_here);
+}
+
+/** Enqueue a future segment.
+ * Two choices: either copies the entire segment, or copies nothing
+ * Returns 0 of the entire segment was copied
+ * Returns -1 if none of the segment was copied due to lack of space
+ */
+
+static int
+svm_fifo_enqueue_with_offset_internal2 (svm_fifo_t * f,
+					int pid,
+					u32 offset,
+					u32 required_bytes,
+					u8 * copy_from_here)
+{
+  u32 total_copy_bytes, first_copy_bytes, second_copy_bytes;
+  u32 cursize, nitems;
+  u32 tail_plus_offset;
+
+  ASSERT (offset > 0);
+
+  /* read cursize, which can only decrease while we're working */
+  cursize = f->cursize;
+  nitems = f->nitems;
+
+  /* Will this request fit? */
+  if ((required_bytes + offset) > (nitems - cursize))
+    return -1;
+
+  ooo_segment_add (f, offset, required_bytes);
+
+  /* Number of bytes we're going to copy */
+  total_copy_bytes = required_bytes;
+  tail_plus_offset = (f->tail + offset) % nitems;
+
+  /* Number of bytes in first copy segment */
+  first_copy_bytes = ((nitems - tail_plus_offset) < total_copy_bytes)
+    ? (nitems - tail_plus_offset) : total_copy_bytes;
+
+  clib_memcpy (&f->data[tail_plus_offset], copy_from_here, first_copy_bytes);
+
+  /* Number of bytes in second copy segment, if any */
+  second_copy_bytes = total_copy_bytes - first_copy_bytes;
+  if (second_copy_bytes)
+    {
+      tail_plus_offset += first_copy_bytes;
+      tail_plus_offset %= nitems;
+
+      ASSERT (tail_plus_offset == 0);
+
+      clib_memcpy (&f->data[tail_plus_offset],
+		   copy_from_here + first_copy_bytes, second_copy_bytes);
+    }
+
+  return (0);
+}
+
+
+int
+svm_fifo_enqueue_with_offset (svm_fifo_t * f,
+			      int pid,
+			      u32 offset,
+			      u32 required_bytes, u8 * copy_from_here)
+{
+  return svm_fifo_enqueue_with_offset_internal2
+    (f, pid, offset, required_bytes, copy_from_here);
+}
+
+
+static int
+svm_fifo_dequeue_internal2 (svm_fifo_t * f,
+			    int pid, u32 max_bytes, u8 * copy_here)
+{
+  u32 total_copy_bytes, first_copy_bytes, second_copy_bytes;
+  u32 cursize, nitems;
+
+  if (PREDICT_FALSE (f->cursize == 0))
+    return -2;			/* nothing in the fifo */
+
+  /* read cursize, which can only increase while we're working */
+  cursize = f->cursize;
+  nitems = f->nitems;
+
+  /* Number of bytes we're going to copy */
+  total_copy_bytes = (cursize < max_bytes) ? cursize : max_bytes;
+
+  if (PREDICT_TRUE (copy_here != 0))
+    {
+      /* Number of bytes in first copy segment */
+      first_copy_bytes = ((nitems - f->head) < total_copy_bytes)
+	? (nitems - f->head) : total_copy_bytes;
+      clib_memcpy (copy_here, &f->data[f->head], first_copy_bytes);
+      f->head += first_copy_bytes;
+      f->head = (f->head == nitems) ? 0 : f->head;
+
+      /* Number of bytes in second copy segment, if any */
+      second_copy_bytes = total_copy_bytes - first_copy_bytes;
+      if (second_copy_bytes)
+	{
+	  clib_memcpy (copy_here + first_copy_bytes,
+		       &f->data[f->head], second_copy_bytes);
+	  f->head += second_copy_bytes;
+	  f->head = (f->head == nitems) ? 0 : f->head;
+	}
+    }
+  else
+    {
+      /* Account for a zero-copy dequeue done elsewhere */
+      ASSERT (max_bytes <= cursize);
+      f->head += max_bytes;
+      f->head = f->head % nitems;
+      cursize -= max_bytes;
+      total_copy_bytes = max_bytes;
+    }
+
+  __sync_fetch_and_sub (&f->cursize, total_copy_bytes);
+
+  return (total_copy_bytes);
+}
+
+int
+svm_fifo_dequeue_nowait (svm_fifo_t * f,
+			 int pid, u32 max_bytes, u8 * copy_here)
+{
+  return svm_fifo_dequeue_internal2 (f, pid, max_bytes, copy_here);
+}
+
+int
+svm_fifo_peek (svm_fifo_t * f, int pid, u32 offset, u32 max_bytes,
+	       u8 * copy_here)
+{
+  u32 total_copy_bytes, first_copy_bytes, second_copy_bytes;
+  u32 cursize, nitems;
+
+  if (PREDICT_FALSE (f->cursize == 0))
+    return -2;			/* nothing in the fifo */
+
+  /* read cursize, which can only increase while we're working */
+  cursize = f->cursize;
+  nitems = f->nitems;
+
+  /* Number of bytes we're going to copy */
+  total_copy_bytes = (cursize < max_bytes) ? cursize : max_bytes;
+
+  if (PREDICT_TRUE (copy_here != 0))
+    {
+      /* Number of bytes in first copy segment */
+      first_copy_bytes =
+	((nitems - f->head) < total_copy_bytes) ?
+	(nitems - f->head) : total_copy_bytes;
+      clib_memcpy (copy_here, &f->data[f->head], first_copy_bytes);
+
+      /* Number of bytes in second copy segment, if any */
+      second_copy_bytes = total_copy_bytes - first_copy_bytes;
+      if (second_copy_bytes)
+	{
+	  clib_memcpy (copy_here + first_copy_bytes, &f->data[0],
+		       second_copy_bytes);
+	}
+    }
+  return total_copy_bytes;
+}
+
+int
+svm_fifo_dequeue_drop (svm_fifo_t * f, int pid, u32 max_bytes)
+{
+  u32 total_drop_bytes, first_drop_bytes, second_drop_bytes;
+  u32 cursize, nitems;
+
+  if (PREDICT_FALSE (f->cursize == 0))
+    return -2;			/* nothing in the fifo */
+
+  /* read cursize, which can only increase while we're working */
+  cursize = f->cursize;
+  nitems = f->nitems;
+
+  /* Number of bytes we're going to drop */
+  total_drop_bytes = (cursize < max_bytes) ? cursize : max_bytes;
+
+  /* Number of bytes in first copy segment */
+  first_drop_bytes =
+    ((nitems - f->head) < total_drop_bytes) ?
+    (nitems - f->head) : total_drop_bytes;
+  f->head += first_drop_bytes;
+  f->head = (f->head == nitems) ? 0 : f->head;
+
+  /* Number of bytes in second drop segment, if any */
+  second_drop_bytes = total_drop_bytes - first_drop_bytes;
+  if (second_drop_bytes)
+    {
+      f->head += second_drop_bytes;
+      f->head = (f->head == nitems) ? 0 : f->head;
+    }
+
+  __sync_fetch_and_sub (&f->cursize, total_drop_bytes);
+
+  return total_drop_bytes;
+}
+
+/*
+ * fd.io coding-style-patch-verification: ON
+ *
+ * Local Variables:
+ * eval: (c-set-style "gnu")
+ * End:
+ */
diff --git a/src/svm/svm_fifo.h b/src/svm/svm_fifo.h
new file mode 100644
index 00000000..70624b74
--- /dev/null
+++ b/src/svm/svm_fifo.h
@@ -0,0 +1,157 @@
+/*
+ * Copyright (c) 2016 Cisco and/or its affiliates.
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at:
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+#ifndef __included_ssvm_fifo_h__
+#define __included_ssvm_fifo_h__
+
+#include 
+#include 
+#include 
+#include 
+#include 
+#include 
+#include 
+
+typedef enum
+{
+  SVM_FIFO_TAG_NOT_HELD = 0,
+  SVM_FIFO_TAG_DEQUEUE,
+  SVM_FIFO_TAG_ENQUEUE,
+} svm_lock_tag_t;
+
+/** Out-of-order segment */
+typedef struct
+{
+  u32 next;	/**< Next linked-list element pool index */
+  u32 prev;	/**< Previous linked-list element pool index */
+
+  u32 fifo_position;	/**< Start of segment, normalized*/
+  u32 length;		/**< Length of segment */
+} ooo_segment_t;
+
+#define OOO_SEGMENT_INVALID_INDEX ((u32)~0)
+
+typedef struct
+{
+  pthread_mutex_t mutex;	/* 8 bytes */
+  pthread_cond_t condvar;	/* 8 bytes */
+  u32 owner_pid;
+  svm_lock_tag_t tag;
+  volatile u32 cursize;
+  u32 nitems;
+
+  /* Backpointers */
+  u32 server_session_index;
+  u32 client_session_index;
+  u8 server_thread_index;
+  u8 client_thread_index;
+    CLIB_CACHE_LINE_ALIGN_MARK (end_shared);
+  u32 head;
+    CLIB_CACHE_LINE_ALIGN_MARK (end_consumer);
+
+  /* producer */
+  u32 tail;
+
+  ooo_segment_t *ooo_segments;	/**< Pool of ooo segments */
+  u32 ooos_list_head;		/**< Head of out-of-order linked-list */
+  u32 ooos_newest;		/**< Last segment to have been updated */
+
+    CLIB_CACHE_LINE_ALIGN_MARK (data);
+} svm_fifo_t;
+
+static inline int
+svm_fifo_lock (svm_fifo_t * f, u32 pid, u32 tag, int nowait)
+{
+  if (PREDICT_TRUE (nowait == 0))
+    pthread_mutex_lock (&f->mutex);
+  else
+    {
+      if (pthread_mutex_trylock (&f->mutex))
+	return -1;
+    }
+  f->owner_pid = pid;
+  f->tag = tag;
+  return 0;
+}
+
+static inline void
+svm_fifo_unlock (svm_fifo_t * f)
+{
+  f->owner_pid = 0;
+  f->tag = 0;
+  CLIB_MEMORY_BARRIER ();
+  pthread_mutex_unlock (&f->mutex);
+}
+
+static inline u32
+svm_fifo_max_dequeue (svm_fifo_t * f)
+{
+  return f->cursize;
+}
+
+static inline u32
+svm_fifo_max_enqueue (svm_fifo_t * f)
+{
+  return f->nitems - f->cursize;
+}
+
+static inline u8
+svm_fifo_has_ooo_data (svm_fifo_t * f)
+{
+  return f->ooos_list_head != OOO_SEGMENT_INVALID_INDEX;
+}
+
+svm_fifo_t *svm_fifo_create (u32 data_size_in_bytes);
+
+int svm_fifo_enqueue_nowait (svm_fifo_t * f, int pid, u32 max_bytes,
+			     u8 * copy_from_here);
+
+int svm_fifo_enqueue_with_offset (svm_fifo_t * f, int pid,
+				  u32 offset, u32 required_bytes,
+				  u8 * copy_from_here);
+
+int svm_fifo_dequeue_nowait (svm_fifo_t * f, int pid, u32 max_bytes,
+			     u8 * copy_here);
+
+int svm_fifo_peek (svm_fifo_t * f, int pid, u32 offset, u32 max_bytes,
+		   u8 * copy_here);
+int svm_fifo_dequeue_drop (svm_fifo_t * f, int pid, u32 max_bytes);
+
+always_inline ooo_segment_t *
+svm_fifo_newest_ooo_segment (svm_fifo_t * f)
+{
+  return f->ooo_segments + f->ooos_newest;
+}
+
+always_inline u32
+ooo_segment_offset (svm_fifo_t * f, ooo_segment_t * s)
+{
+  return ((f->nitems + s->fifo_position - f->tail) % f->nitems);
+}
+
+always_inline u32
+ooo_segment_end_offset (svm_fifo_t * f, ooo_segment_t * s)
+{
+  return ((f->nitems + s->fifo_position + s->length - f->tail) % f->nitems);
+}
+
+#endif /* __included_ssvm_fifo_h__ */
+
+/*
+ * fd.io coding-style-patch-verification: ON
+ *
+ * Local Variables:
+ * eval: (c-set-style "gnu")
+ * End:
+ */
diff --git a/src/svm/svm_fifo_segment.c b/src/svm/svm_fifo_segment.c
new file mode 100644
index 00000000..acabb3bd
--- /dev/null
+++ b/src/svm/svm_fifo_segment.c
@@ -0,0 +1,193 @@
+/*
+ * Copyright (c) 2016 Cisco and/or its affiliates.
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at:
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#include 
+
+svm_fifo_segment_main_t svm_fifo_segment_main;
+
+/** (master) create an svm fifo segment */
+int
+svm_fifo_segment_create (svm_fifo_segment_create_args_t * a)
+{
+  int rv;
+  svm_fifo_segment_private_t *s;
+  svm_fifo_segment_main_t *sm = &svm_fifo_segment_main;
+  ssvm_shared_header_t *sh;
+  svm_fifo_segment_header_t *fsh;
+  void *oldheap;
+
+  /* Allocate a fresh segment */
+  pool_get (sm->segments, s);
+  memset (s, 0, sizeof (*s));
+
+  s->ssvm.ssvm_size = a->segment_size;
+  s->ssvm.i_am_master = 1;
+  s->ssvm.my_pid = getpid ();
+  s->ssvm.name = (u8 *) a->segment_name;
+  s->ssvm.requested_va = sm->next_baseva;
+
+  rv = ssvm_master_init (&s->ssvm, s - sm->segments);
+
+  if (rv)
+    {
+      _vec_len (s) = vec_len (s) - 1;
+      return (rv);
+    }
+
+  /* Note; requested_va updated due to seg base addr randomization */
+  sm->next_baseva = s->ssvm.requested_va + a->segment_size;
+
+  sh = s->ssvm.sh;
+  oldheap = ssvm_push_heap (sh);
+
+  /* Set up svm_fifo_segment shared header */
+  fsh = clib_mem_alloc (sizeof (*fsh));
+  memset (fsh, 0, sizeof (*fsh));
+  sh->opaque[0] = fsh;
+  s->h = fsh;
+  fsh->segment_name = format (0, "%s%c", a->segment_name, 0);
+
+  /* Avoid vec_add1(...) failure when adding a fifo, etc. */
+  vec_validate (fsh->fifos, 64);
+  _vec_len (fsh->fifos) = 0;
+
+  ssvm_pop_heap (oldheap);
+
+  sh->ready = 1;
+  a->new_segment_index = s - sm->segments;
+  return (0);
+}
+
+/** (slave) attach to an svm fifo segment */
+int
+svm_fifo_segment_attach (svm_fifo_segment_create_args_t * a)
+{
+  int rv;
+  svm_fifo_segment_private_t *s;
+  svm_fifo_segment_main_t *sm = &svm_fifo_segment_main;
+  ssvm_shared_header_t *sh;
+  svm_fifo_segment_header_t *fsh;
+
+  /* Allocate a fresh segment */
+  pool_get (sm->segments, s);
+
+  memset (s, 0, sizeof (*s));
+
+  s->ssvm.ssvm_size = a->segment_size;
+  s->ssvm.my_pid = getpid ();
+  s->ssvm.name = (u8 *) a->segment_name;
+  s->ssvm.requested_va = sm->next_baseva;
+
+  rv = ssvm_slave_init (&s->ssvm, sm->timeout_in_seconds);
+
+  if (rv)
+    {
+      _vec_len (s) = vec_len (s) - 1;
+      return (rv);
+    }
+
+  /* Fish the segment header */
+  sh = s->ssvm.sh;
+  fsh = (svm_fifo_segment_header_t *) sh->opaque[0];
+  s->h = fsh;
+
+  a->new_segment_index = s - sm->segments;
+  return (0);
+}
+
+void
+svm_fifo_segment_delete (svm_fifo_segment_private_t * s)
+{
+  svm_fifo_segment_main_t *sm = &svm_fifo_segment_main;
+  ssvm_delete (&s->ssvm);
+  pool_put (sm->segments, s);
+}
+
+svm_fifo_t *
+svm_fifo_segment_alloc_fifo (svm_fifo_segment_private_t * s,
+			     u32 data_size_in_bytes)
+{
+  ssvm_shared_header_t *sh;
+  svm_fifo_segment_header_t *fsh;
+  svm_fifo_t *f;
+  void *oldheap;
+
+  sh = s->ssvm.sh;
+  fsh = (svm_fifo_segment_header_t *) sh->opaque[0];
+  oldheap = ssvm_push_heap (sh);
+
+  /* Note: this can fail, in which case: create another segment */
+  f = svm_fifo_create (data_size_in_bytes);
+  if (f == 0)
+    {
+      ssvm_pop_heap (oldheap);
+      return (0);
+    }
+
+  vec_add1 (fsh->fifos, f);
+
+  ssvm_pop_heap (oldheap);
+  return (f);
+}
+
+void
+svm_fifo_segment_free_fifo (svm_fifo_segment_private_t * s, svm_fifo_t * f)
+{
+  ssvm_shared_header_t *sh;
+  svm_fifo_segment_header_t *fsh;
+  void *oldheap;
+  int i;
+
+  sh = s->ssvm.sh;
+  fsh = (svm_fifo_segment_header_t *) sh->opaque[0];
+  oldheap = ssvm_push_heap (sh);
+
+  for (i = 0; i < vec_len (fsh->fifos); i++)
+    {
+      if (fsh->fifos[i] == f)
+	{
+	  vec_delete (fsh->fifos, 1, i);
+	  goto found;
+	}
+    }
+  clib_warning ("fifo 0x%llx not found in fifo table...", f);
+
+found:
+  clib_mem_free (f);
+  ssvm_pop_heap (oldheap);
+}
+
+void
+svm_fifo_segment_init (u64 baseva, u32 timeout_in_seconds)
+{
+  svm_fifo_segment_main_t *sm = &svm_fifo_segment_main;
+
+  sm->next_baseva = baseva;
+  sm->timeout_in_seconds = timeout_in_seconds;
+}
+
+u32
+svm_fifo_segment_index (svm_fifo_segment_private_t * s)
+{
+  return s - svm_fifo_segment_main.segments;
+}
+
+/*
+ * fd.io coding-style-patch-verification: ON
+ *
+ * Local Variables:
+ * eval: (c-set-style "gnu")
+ * End:
+ */
diff --git a/src/svm/svm_fifo_segment.h b/src/svm/svm_fifo_segment.h
new file mode 100644
index 00000000..793fa7c8
--- /dev/null
+++ b/src/svm/svm_fifo_segment.h
@@ -0,0 +1,89 @@
+/*
+ * Copyright (c) 2016 Cisco and/or its affiliates.
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at:
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+#ifndef __included_ssvm_fifo_segment_h__
+#define __included_ssvm_fifo_segment_h__
+
+#include "svm_fifo.h"
+#include "ssvm.h"
+
+typedef struct
+{
+  volatile svm_fifo_t **fifos;
+  u8 *segment_name;
+} svm_fifo_segment_header_t;
+
+typedef struct
+{
+  ssvm_private_t ssvm;
+  svm_fifo_segment_header_t *h;
+} svm_fifo_segment_private_t;
+
+typedef struct
+{
+  /** pool of segments */
+  svm_fifo_segment_private_t *segments;
+  /* Where to put the next one */
+  u64 next_baseva;
+  u32 timeout_in_seconds;
+} svm_fifo_segment_main_t;
+
+extern svm_fifo_segment_main_t svm_fifo_segment_main;
+
+typedef struct
+{
+  char *segment_name;
+  u32 segment_size;
+  u32 new_segment_index;
+} svm_fifo_segment_create_args_t;
+
+static inline svm_fifo_segment_private_t *
+svm_fifo_get_segment (u32 segment_index)
+{
+  svm_fifo_segment_main_t *ssm = &svm_fifo_segment_main;
+  return vec_elt_at_index (ssm->segments, segment_index);
+}
+
+#define foreach_ssvm_fifo_segment_api_error             \
+_(OUT_OF_SPACE, "Out of space in segment", -200)
+
+typedef enum
+{
+#define _(n,s,c) SSVM_FIFO_SEGMENT_API_ERROR_##n = c,
+  foreach_ssvm_fifo_segment_api_error
+#undef _
+} ssvm_fifo_segment_api_error_enum_t;
+
+int svm_fifo_segment_create (svm_fifo_segment_create_args_t * a);
+int svm_fifo_segment_attach (svm_fifo_segment_create_args_t * a);
+void svm_fifo_segment_delete (svm_fifo_segment_private_t * s);
+
+svm_fifo_t *svm_fifo_segment_alloc_fifo (svm_fifo_segment_private_t * s,
+					 u32 data_size_in_bytes);
+void svm_fifo_segment_free_fifo (svm_fifo_segment_private_t * s,
+				 svm_fifo_t * f);
+
+void svm_fifo_segment_init (u64 baseva, u32 timeout_in_seconds);
+
+u32 svm_fifo_segment_index (svm_fifo_segment_private_t * s);
+
+#endif /* __included_ssvm_fifo_segment_h__ */
+
+/*
+ * fd.io coding-style-patch-verification: ON
+ *
+ * Local Variables:
+ * eval: (c-set-style "gnu")
+ * End:
+ */
diff --git a/src/svm/test_svm_fifo1.c b/src/svm/test_svm_fifo1.c
new file mode 100644
index 00000000..355653df
--- /dev/null
+++ b/src/svm/test_svm_fifo1.c
@@ -0,0 +1,361 @@
+/*
+ * Copyright (c) 2016 Cisco and/or its affiliates.
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at:
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#include "svm_fifo_segment.h"
+
+clib_error_t *
+hello_world (int verbose)
+{
+  svm_fifo_segment_create_args_t _a, *a = &_a;
+  svm_fifo_segment_private_t *sp;
+  svm_fifo_t *f;
+  int rv;
+  u8 *test_data;
+  u8 *retrieved_data = 0;
+  clib_error_t *error = 0;
+  int pid = getpid ();
+
+  memset (a, 0, sizeof (*a));
+
+  a->segment_name = "fifo-test1";
+  a->segment_size = 256 << 10;
+
+  rv = svm_fifo_segment_create (a);
+
+  if (rv)
+    return clib_error_return (0, "svm_fifo_segment_create returned %d", rv);
+
+  sp = svm_fifo_get_segment (a->new_segment_index);
+
+  f = svm_fifo_segment_alloc_fifo (sp, 4096);
+
+  if (f == 0)
+    return clib_error_return (0, "svm_fifo_segment_alloc_fifo failed");
+
+  test_data = format (0, "Hello world%c", 0);
+  vec_validate (retrieved_data, vec_len (test_data) - 1);
+
+  while (svm_fifo_max_enqueue (f) >= vec_len (test_data))
+    svm_fifo_enqueue_nowait (f, pid, vec_len (test_data), test_data);
+
+  while (svm_fifo_max_dequeue (f) >= vec_len (test_data))
+    svm_fifo_dequeue_nowait (f, pid, vec_len (retrieved_data),
+			     retrieved_data);
+
+  while (svm_fifo_max_enqueue (f) >= vec_len (test_data))
+    svm_fifo_enqueue_nowait (f, pid, vec_len (test_data), test_data);
+
+  while (svm_fifo_max_dequeue (f) >= vec_len (test_data))
+    svm_fifo_dequeue_nowait (f, pid, vec_len (retrieved_data),
+			     retrieved_data);
+
+  if (!memcmp (retrieved_data, test_data, vec_len (test_data)))
+    error = clib_error_return (0, "data test OK, got '%s'", retrieved_data);
+  else
+    error = clib_error_return (0, "data test FAIL!");
+
+  svm_fifo_segment_free_fifo (sp, f);
+
+  return error;
+}
+
+clib_error_t *
+master (int verbose)
+{
+  svm_fifo_segment_create_args_t _a, *a = &_a;
+  svm_fifo_segment_private_t *sp;
+  svm_fifo_t *f;
+  int rv;
+  u8 *test_data;
+  u8 *retrieved_data = 0;
+  int i;
+  int pid = getpid ();
+
+  memset (a, 0, sizeof (*a));
+
+  a->segment_name = "fifo-test1";
+  a->segment_size = 256 << 10;
+
+  rv = svm_fifo_segment_create (a);
+
+  if (rv)
+    return clib_error_return (0, "svm_fifo_segment_create returned %d", rv);
+
+  sp = svm_fifo_get_segment (a->new_segment_index);
+
+  f = svm_fifo_segment_alloc_fifo (sp, 4096);
+
+  if (f == 0)
+    return clib_error_return (0, "svm_fifo_segment_alloc_fifo failed");
+
+  test_data = format (0, "Hello world%c", 0);
+  vec_validate (retrieved_data, vec_len (test_data) - 1);
+
+  for (i = 0; i < 1000; i++)
+    svm_fifo_enqueue_nowait (f, pid, vec_len (test_data), test_data);
+
+  return clib_error_return (0, "master (enqueue) done");
+}
+
+clib_error_t *
+mempig (int verbose)
+{
+  svm_fifo_segment_create_args_t _a, *a = &_a;
+  svm_fifo_segment_private_t *sp;
+  svm_fifo_t *f;
+  svm_fifo_t **flist = 0;
+  int rv;
+  int i;
+
+  memset (a, 0, sizeof (*a));
+
+  a->segment_name = "fifo-test1";
+  a->segment_size = 256 << 10;
+
+  rv = svm_fifo_segment_create (a);
+
+  if (rv)
+    return clib_error_return (0, "svm_fifo_segment_create returned %d", rv);
+
+  sp = svm_fifo_get_segment (a->new_segment_index);
+
+  for (i = 0; i < 1000; i++)
+    {
+      f = svm_fifo_segment_alloc_fifo (sp, 4096);
+      if (f == 0)
+	break;
+      vec_add1 (flist, f);
+    }
+
+  fformat (stdout, "Try #1: created %d fifos...\n", vec_len (flist));
+  for (i = 0; i < vec_len (flist); i++)
+    {
+      f = flist[i];
+      svm_fifo_segment_free_fifo (sp, f);
+    }
+
+  _vec_len (flist) = 0;
+
+  for (i = 0; i < 1000; i++)
+    {
+      f = svm_fifo_segment_alloc_fifo (sp, 4096);
+      if (f == 0)
+	break;
+      vec_add1 (flist, f);
+    }
+
+  fformat (stdout, "Try #2: created %d fifos...\n", vec_len (flist));
+  for (i = 0; i < vec_len (flist); i++)
+    {
+      f = flist[i];
+      svm_fifo_segment_free_fifo (sp, f);
+    }
+
+  return 0;
+}
+
+clib_error_t *
+offset (int verbose)
+{
+  svm_fifo_segment_create_args_t _a, *a = &_a;
+  svm_fifo_segment_private_t *sp;
+  svm_fifo_t *f;
+  int rv;
+  u32 *test_data = 0;
+  u32 *recovered_data = 0;
+  int i;
+  int pid = getpid ();
+
+  memset (a, 0, sizeof (*a));
+
+  a->segment_name = "fifo-test1";
+  a->segment_size = 256 << 10;
+
+  rv = svm_fifo_segment_create (a);
+
+  if (rv)
+    return clib_error_return (0, "svm_fifo_segment_create returned %d", rv);
+
+  sp = svm_fifo_get_segment (a->new_segment_index);
+
+  f = svm_fifo_segment_alloc_fifo (sp, 200 << 10);
+
+  if (f == 0)
+    return clib_error_return (0, "svm_fifo_segment_alloc_fifo failed");
+
+  for (i = 0; i < (3 * 1024); i++)
+    vec_add1 (test_data, i);
+
+  /* Enqueue the first 1024 u32's */
+  svm_fifo_enqueue_nowait (f, pid, 4096 /* bytes to enqueue */ ,
+			   (u8 *) test_data);
+
+  /* Enqueue the third 1024 u32's 2048 ahead of the current tail */
+  svm_fifo_enqueue_with_offset (f, pid, 4096, 4096, (u8 *) & test_data[2048]);
+
+  /* Enqueue the second 1024 u32's at the current tail */
+  svm_fifo_enqueue_nowait (f, pid, 4096 /* bytes to enqueue */ ,
+			   (u8 *) & test_data[1024]);
+
+  vec_validate (recovered_data, (3 * 1024) - 1);
+
+  svm_fifo_dequeue_nowait (f, pid, 3 * 4096, (u8 *) recovered_data);
+
+  for (i = 0; i < (3 * 1024); i++)
+    {
+      if (recovered_data[i] != test_data[i])
+	{
+	  clib_warning ("[%d] expected %d recovered %d", i,
+			test_data[i], recovered_data[i]);
+	  return clib_error_return (0, "offset test FAILED");
+	}
+    }
+
+  return clib_error_return (0, "offset test OK");
+}
+
+clib_error_t *
+slave (int verbose)
+{
+  svm_fifo_segment_create_args_t _a, *a = &_a;
+  svm_fifo_segment_private_t *sp;
+  svm_fifo_segment_header_t *fsh;
+  svm_fifo_t *f;
+  ssvm_shared_header_t *sh;
+  int rv;
+  u8 *test_data;
+  u8 *retrieved_data = 0;
+  int pid = getpid ();
+  int i;
+
+  memset (a, 0, sizeof (*a));
+
+  a->segment_name = "fifo-test1";
+
+  rv = svm_fifo_segment_attach (a);
+
+  if (rv)
+    return clib_error_return (0, "svm_fifo_segment_attach returned %d", rv);
+
+  sp = svm_fifo_get_segment (a->new_segment_index);
+  sh = sp->ssvm.sh;
+  fsh = (svm_fifo_segment_header_t *) sh->opaque[0];
+
+  /* might wanna wait.. */
+  f = (svm_fifo_t *) fsh->fifos[0];
+
+  /* Lazy bastards united */
+  test_data = format (0, "Hello world%c", 0);
+  vec_validate (retrieved_data, vec_len (test_data) - 1);
+
+  for (i = 0; i < 1000; i++)
+    {
+      svm_fifo_dequeue_nowait (f, pid, vec_len (retrieved_data),
+			       retrieved_data);
+      if (memcmp (retrieved_data, test_data, vec_len (retrieved_data)))
+	return clib_error_return (0, "retrieved data incorrect, '%s'",
+				  retrieved_data);
+    }
+
+  return clib_error_return (0, "slave (dequeue) done");
+}
+
+
+int
+test_ssvm_fifo1 (unformat_input_t * input)
+{
+  clib_error_t *error = 0;
+  int verbose = 0;
+  int test_id = 0;
+
+  svm_fifo_segment_init (0x200000000ULL, 20);
+
+  while (unformat_check_input (input) != UNFORMAT_END_OF_INPUT)
+    {
+      if (unformat (input, "verbose %d", &verbose))
+	;
+      else if (unformat (input, "verbose"))
+	verbose = 1;
+      else if (unformat (input, "master"))
+	test_id = 1;
+      else if (unformat (input, "slave"))
+	test_id = 2;
+      else if (unformat (input, "mempig"))
+	test_id = 3;
+      else if (unformat (input, "offset"))
+	test_id = 4;
+      else
+	{
+	  error = clib_error_create ("unknown input `%U'\n",
+				     format_unformat_error, input);
+	  goto out;
+	}
+    }
+
+  switch (test_id)
+    {
+    case 0:
+      error = hello_world (verbose);
+      break;
+
+    case 1:
+      error = master (verbose);
+      break;
+
+    case 2:
+      error = slave (verbose);
+      break;
+
+    case 3:
+      error = mempig (verbose);
+      break;
+
+    case 4:
+      error = offset (verbose);
+      break;
+
+    default:
+      error = clib_error_return (0, "test id %d unknown", test_id);
+      break;
+    }
+
+out:
+  if (error)
+    clib_error_report (error);
+
+  return 0;
+}
+
+
+
+int
+main (int argc, char *argv[])
+{
+  unformat_input_t i;
+  int r;
+
+  unformat_init_command_line (&i, argv);
+  r = test_ssvm_fifo1 (&i);
+  unformat_free (&i);
+  return r;
+}
+
+/*
+ * fd.io coding-style-patch-verification: ON
+ *
+ * Local Variables:
+ * eval: (c-set-style "gnu")
+ * End:
+ */
diff --git a/src/uri.am b/src/uri.am
new file mode 100644
index 00000000..8cdd77c6
--- /dev/null
+++ b/src/uri.am
@@ -0,0 +1,22 @@
+# Copyright (c) 2016 Cisco and/or its affiliates.
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at:
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+noinst_PROGRAMS += uri_udp_test2 uri_tcp_test
+
+uri_udp_test2_SOURCES = uri/uri_udp_test2.c			
+uri_udp_test2_LDADD = libvlibmemoryclient.la libvlibapi.la libsvm.la \
+	libvppinfra.la -lpthread -lm -lrt 
+
+uri_tcp_test_SOURCES = uri/uri_tcp_test.c
+uri_tcp_test_LDADD = libvlibmemoryclient.la libvlibapi.la libsvm.la \
+	libvppinfra.la -lpthread -lm -lrt 
diff --git a/src/uri/uri_tcp_test.c b/src/uri/uri_tcp_test.c
new file mode 100644
index 00000000..ed5a37d8
--- /dev/null
+++ b/src/uri/uri_tcp_test.c
@@ -0,0 +1,916 @@
+/*
+ * Copyright (c) 2016 Cisco and/or its affiliates.
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at:
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#include 
+#include 
+#include 
+#include 
+#include 
+#include 
+#include 
+
+#include "../vnet/session/application_interface.h"
+
+#define vl_typedefs             /* define message structures */
+#include 
+#undef vl_typedefs
+
+/* declare message handlers for each api */
+
+#define vl_endianfun            /* define message structures */
+#include 
+#undef vl_endianfun
+
+/* instantiate all the print functions we know about */
+#define vl_print(handle, ...)
+#define vl_printfun
+#include 
+#undef vl_printfun
+
+/* Satisfy external references when not linking with -lvlib */
+vlib_main_t vlib_global_main;
+vlib_main_t **vlib_mains;
+
+typedef struct
+{
+  svm_fifo_t * server_rx_fifo;
+  svm_fifo_t * server_tx_fifo;
+
+  u32 vpp_session_index;
+  u32 vpp_session_thread;
+} session_t;
+
+typedef enum
+{
+  STATE_START,
+  STATE_READY,
+  STATE_DISCONNECTING,
+  STATE_FAILED
+} connection_state_t;
+
+typedef struct
+{
+  /* vpe input queue */
+  unix_shared_memory_queue_t *vl_input_queue;
+
+  /* API client handle */
+  u32 my_client_index;
+
+  /* The URI we're playing with */
+  u8 * uri;
+
+  /* Session pool */
+  session_t * sessions;
+
+  /* Hash table for disconnect processing */
+  uword * session_index_by_vpp_handles;
+
+  /* intermediate rx buffer */
+  u8 * rx_buf;
+
+  /* URI for slave's connect */
+  u8 * connect_uri;
+
+  u32 connected_session_index;
+
+  int i_am_master;
+
+  /* drop all packets */
+  int drop_packets;
+
+  /* Our event queue */
+  unix_shared_memory_queue_t * our_event_queue;
+
+  /* $$$ single thread only for the moment */
+  unix_shared_memory_queue_t * vpp_event_queue;
+
+  pid_t my_pid;
+
+  /* For deadman timers */
+  clib_time_t clib_time;
+
+  /* State of the connection, shared between msg RX thread and main thread */
+  volatile connection_state_t state;
+
+  /* Signal variables */
+  volatile int time_to_stop;
+  volatile int time_to_print_stats;
+
+  u32 configured_segment_size;
+
+  /* VNET_API_ERROR_FOO -> "Foo" hash table */
+  uword * error_string_by_error_number;
+
+  /* convenience */
+  svm_fifo_segment_main_t * segment_main;
+
+  u8 *connect_test_data;
+} uri_tcp_test_main_t;
+
+uri_tcp_test_main_t uri_tcp_test_main;
+
+#if CLIB_DEBUG > 0
+#define NITER 10000
+#else
+#define NITER 4000000
+#endif
+
+int
+wait_for_state_change (uri_tcp_test_main_t * utm, connection_state_t state)
+{
+#if CLIB_DEBUG > 0
+#define TIMEOUT 600.0
+#else
+#define TIMEOUT 600.0
+#endif
+
+  f64 timeout = clib_time_now (&utm->clib_time) + TIMEOUT;
+
+  while (clib_time_now (&utm->clib_time) < timeout)
+    {
+      if (utm->state == state)
+        return 0;
+      if (utm->state == STATE_FAILED)
+	return -1;
+    }
+  clib_warning ("timeout waiting for STATE_READY");
+  return -1;
+}
+
+static void
+init_error_string_table (uri_tcp_test_main_t * utm)
+{
+  utm->error_string_by_error_number = hash_create (0, sizeof (uword));
+
+#define _(n,v,s) hash_set (utm->error_string_by_error_number, -v, s);
+  foreach_vnet_api_error;
+#undef _
+
+  hash_set (utm->error_string_by_error_number, 99, "Misc");
+}
+
+static void
+stop_signal (int signum)
+{
+  uri_tcp_test_main_t *um = &uri_tcp_test_main;
+
+  um->time_to_stop = 1;
+}
+
+static void
+stats_signal (int signum)
+{
+  uri_tcp_test_main_t *um = &uri_tcp_test_main;
+
+  um->time_to_print_stats = 1;
+}
+
+static clib_error_t *
+setup_signal_handlers (void)
+{
+  signal (SIGINT, stats_signal);
+  signal (SIGQUIT, stop_signal);
+  signal (SIGTERM, stop_signal);
+
+  return 0;
+}
+
+void
+vlib_cli_output (struct vlib_main_t *vm, char *fmt, ...)
+{
+  clib_warning ("BUG");
+}
+
+int
+connect_to_vpp (char *name)
+{
+  uri_tcp_test_main_t *utm = &uri_tcp_test_main;
+  api_main_t *am = &api_main;
+
+  if (vl_client_connect_to_vlib ("/vpe-api", name, 32) < 0)
+    return -1;
+
+  utm->vl_input_queue = am->shmem_hdr->vl_input_queue;
+  utm->my_client_index = am->my_client_index;
+
+  return 0;
+}
+
+static void
+vl_api_map_another_segment_t_handler (vl_api_map_another_segment_t *mp)
+{
+  svm_fifo_segment_create_args_t _a, *a = &_a;
+  int rv;
+
+  a->segment_name = (char *) mp->segment_name;
+  a->segment_size = mp->segment_size;
+  /* Attach to the segment vpp created */
+  rv = svm_fifo_segment_attach (a);
+  if (rv)
+    {
+      clib_warning ("svm_fifo_segment_attach ('%s') failed",
+                    mp->segment_name);
+      return;
+    }
+  clib_warning ("Mapped new segment '%s' size %d", mp->segment_name,
+                mp->segment_size);
+}
+
+static void
+vl_api_disconnect_session_t_handler (vl_api_disconnect_session_t * mp)
+{
+  uri_tcp_test_main_t *utm = &uri_tcp_test_main;
+  session_t * session;
+  vl_api_disconnect_session_reply_t * rmp;
+  uword * p;
+  int rv = 0;
+  u64 key;
+
+  key = (((u64)mp->session_thread_index) << 32) | (u64)mp->session_index;
+
+  p = hash_get (utm->session_index_by_vpp_handles, key);
+
+  if (p)
+    {
+      session = pool_elt_at_index (utm->sessions, p[0]);
+      hash_unset (utm->session_index_by_vpp_handles, key);
+      pool_put (utm->sessions, session);
+    }
+  else
+    {
+      clib_warning ("couldn't find session key %llx", key);
+      rv = -11;
+    }
+
+  rmp = vl_msg_api_alloc (sizeof (*rmp));
+  memset (rmp, 0, sizeof (*rmp));
+
+  rmp->_vl_msg_id = ntohs (VL_API_DISCONNECT_SESSION_REPLY);
+  rmp->retval = rv;
+  rmp->session_index = mp->session_index;
+  rmp->session_thread_index = mp->session_thread_index;
+  vl_msg_api_send_shmem (utm->vl_input_queue, (u8 *)&rmp);
+}
+
+static void
+vl_api_reset_session_t_handler (vl_api_reset_session_t * mp)
+{
+  uri_tcp_test_main_t *utm = &uri_tcp_test_main;
+  session_t * session;
+  vl_api_reset_session_reply_t * rmp;
+  uword * p;
+  int rv = 0;
+  u64 key;
+
+  key = (((u64)mp->session_thread_index) << 32) | (u64)mp->session_index;
+
+  p = hash_get(utm->session_index_by_vpp_handles, key);
+
+  if (p)
+    {
+      session = pool_elt_at_index(utm->sessions, p[0]);
+      hash_unset(utm->session_index_by_vpp_handles, key);
+      pool_put(utm->sessions, session);
+    }
+  else
+    {
+      clib_warning("couldn't find session key %llx", key);
+      rv = -11;
+    }
+
+  rmp = vl_msg_api_alloc (sizeof (*rmp));
+  memset (rmp, 0, sizeof (*rmp));
+  rmp->_vl_msg_id = ntohs (VL_API_DISCONNECT_SESSION_REPLY);
+  rmp->retval = rv;
+  rmp->session_index = mp->session_index;
+  rmp->session_thread_index = mp->session_thread_index;
+  vl_msg_api_send_shmem (utm->vl_input_queue, (u8 *)&rmp);
+}
+
+void
+handle_fifo_event_connect_rx (uri_tcp_test_main_t *utm, session_fifo_event_t * e)
+{
+  svm_fifo_t * rx_fifo;
+  int n_read, bytes;
+
+  rx_fifo = e->fifo;
+
+  bytes = e->enqueue_length;
+  do
+    {
+      n_read = svm_fifo_dequeue_nowait (rx_fifo, 0, vec_len(utm->rx_buf),
+                                         utm->rx_buf);
+      if (n_read > 0)
+        bytes -= n_read;
+    }
+  while (n_read < 0 || bytes > 0);
+
+  //      bytes_to_read = svm_fifo_max_dequeue (rx_fifo);
+  //
+  //      bytes_to_read = vec_len(utm->rx_buf) > bytes_to_read ?
+  //        bytes_to_read : vec_len(utm->rx_buf);
+  //
+  //      buffer_offset = 0;
+  //      while (bytes_to_read > 0)
+  //        {
+  //          rv = svm_fifo_dequeue_nowait2 (rx_fifo, mypid,
+  //                                         bytes_to_read,
+  //                                         utm->rx_buf + buffer_offset);
+  //          if (rv > 0)
+  //            {
+  //              bytes_to_read -= rv;
+  //              buffer_offset += rv;
+  //              bytes_received += rv;
+  //            }
+  //        }
+
+
+  //  while (bytes_received < bytes_sent)
+  //    {
+  //      rv = svm_fifo_dequeue_nowait2 (rx_fifo, mypid,
+  //                                     vec_len (utm->rx_buf),
+  //                                     utm->rx_buf);
+  //      if (rv > 0)
+  //        {
+  //#if CLIB_DEBUG > 0
+  //          int j;
+  //          for (j = 0; j < rv; j++)
+  //            {
+  //              if (utm->rx_buf[j] != ((bytes_received + j) & 0xff))
+  //                {
+  //                  clib_warning ("error at byte %lld, 0x%x not 0x%x",
+  //                                bytes_received + j,
+  //                                utm->rx_buf[j],
+  //                                ((bytes_received + j )&0xff));
+  //                }
+  //            }
+  //#endif
+  //          bytes_received += (u64) rv;
+  //        }
+  //    }
+}
+
+void
+handle_connect_event_queue (uri_tcp_test_main_t * utm)
+{
+  session_fifo_event_t _e, *e = &_e;;
+
+  unix_shared_memory_queue_sub (utm->our_event_queue, (u8 *) e, 0 /* nowait */);
+  switch (e->event_type)
+    {
+    case FIFO_EVENT_SERVER_RX:
+      handle_fifo_event_connect_rx (utm, e);
+      break;
+
+    case FIFO_EVENT_SERVER_EXIT:
+      return;
+
+    default:
+      clib_warning("unknown event type %d", e->event_type);
+      break;
+    }
+}
+
+void
+uri_tcp_connect_send (uri_tcp_test_main_t *utm)
+{
+  u8 *test_data = utm->connect_test_data;
+  u64 bytes_sent = 0;
+  int rv;
+  int mypid = getpid();
+  session_t * session;
+  svm_fifo_t *tx_fifo;
+  int buffer_offset, bytes_to_send = 0;
+  session_fifo_event_t evt;
+  static int serial_number = 0;
+  int i;
+  u32 max_chunk = 64 << 10, write;
+
+  session = pool_elt_at_index (utm->sessions, utm->connected_session_index);
+  tx_fifo = session->server_tx_fifo;
+
+  vec_validate (utm->rx_buf, vec_len (test_data) - 1);
+
+  for (i = 0; i < 10; i++)
+    {
+      bytes_to_send = vec_len (test_data);
+      buffer_offset = 0;
+      while (bytes_to_send > 0)
+        {
+          write = bytes_to_send > max_chunk ? max_chunk : bytes_to_send;
+          rv = svm_fifo_enqueue_nowait (tx_fifo, mypid, write,
+                                         test_data + buffer_offset);
+
+          if (rv > 0)
+            {
+              bytes_to_send -= rv;
+              buffer_offset += rv;
+              bytes_sent += rv;
+
+              /* Fabricate TX event, send to vpp */
+              evt.fifo = tx_fifo;
+              evt.event_type = FIFO_EVENT_SERVER_TX;
+              /* $$$$ for event logging */
+              evt.enqueue_length = rv;
+              evt.event_id = serial_number++;
+
+              unix_shared_memory_queue_add (utm->vpp_event_queue, (u8 *) &evt,
+                                            0 /* do wait for mutex */);
+            }
+        }
+    }
+}
+
+static void
+uri_tcp_client_test (uri_tcp_test_main_t * utm)
+{
+  vl_api_connect_uri_t * cmp;
+  vl_api_disconnect_session_t *dmp;
+  session_t *connected_session;
+  int i;
+
+  cmp = vl_msg_api_alloc (sizeof (*cmp));
+  memset (cmp, 0, sizeof (*cmp));
+
+  cmp->_vl_msg_id = ntohs (VL_API_CONNECT_URI);
+  cmp->client_index = utm->my_client_index;
+  cmp->context = ntohl(0xfeedface);
+  memcpy (cmp->uri, utm->connect_uri, vec_len (utm->connect_uri));
+  vl_msg_api_send_shmem (utm->vl_input_queue, (u8 *)&cmp);
+
+  if (wait_for_state_change (utm, STATE_READY))
+    {
+      return;
+    }
+
+  /* Init test data */
+  vec_validate (utm->connect_test_data, 64 * 1024 - 1);
+  for (i = 0; i < vec_len (utm->connect_test_data); i++)
+    utm->connect_test_data[i] = i & 0xff;
+
+  /* Start reader thread */
+  /* handle_connect_event_queue (utm); */
+
+  /* Start send */
+  uri_tcp_connect_send (utm);
+
+  /* Disconnect */
+  connected_session = pool_elt_at_index(utm->sessions,
+					utm->connected_session_index);
+  dmp = vl_msg_api_alloc (sizeof (*dmp));
+  memset (dmp, 0, sizeof (*dmp));
+  dmp->_vl_msg_id = ntohs (VL_API_DISCONNECT_SESSION);
+  dmp->client_index = utm->my_client_index;
+  dmp->session_index = connected_session->vpp_session_index;
+  dmp->session_thread_index = connected_session->vpp_session_thread;
+  vl_msg_api_send_shmem (utm->vl_input_queue, (u8 *)&dmp);
+}
+
+void
+handle_fifo_event_server_rx (uri_tcp_test_main_t *utm, session_fifo_event_t * e)
+{
+  svm_fifo_t * rx_fifo, * tx_fifo;
+  int n_read;
+
+  session_fifo_event_t evt;
+  unix_shared_memory_queue_t *q;
+  int rv, bytes;
+
+  rx_fifo = e->fifo;
+  tx_fifo = utm->sessions[rx_fifo->client_session_index].server_tx_fifo;
+
+  bytes = e->enqueue_length;
+  do
+    {
+      n_read = svm_fifo_dequeue_nowait (rx_fifo, 0, vec_len(utm->rx_buf),
+                                         utm->rx_buf);
+
+      /* Reflect if a non-drop session */
+      if (!utm->drop_packets && n_read > 0)
+        {
+          do
+            {
+              rv = svm_fifo_enqueue_nowait (tx_fifo, 0, n_read, utm->rx_buf);
+            }
+          while (rv == -2);
+
+          /* Fabricate TX event, send to vpp */
+          evt.fifo = tx_fifo;
+          evt.event_type = FIFO_EVENT_SERVER_TX;
+          /* $$$$ for event logging */
+          evt.enqueue_length = n_read;
+          evt.event_id = e->event_id;
+          q = utm->vpp_event_queue;
+          unix_shared_memory_queue_add (q, (u8 *) &evt, 0 /* do wait for mutex */);
+        }
+
+      if (n_read > 0)
+        bytes -= n_read;
+    }
+  while (n_read < 0 || bytes > 0);
+}
+
+void
+handle_event_queue (uri_tcp_test_main_t * utm)
+{
+  session_fifo_event_t _e, *e = &_e;;
+
+  while (1)
+    {
+      unix_shared_memory_queue_sub (utm->our_event_queue, (u8 *)e,
+                                    0 /* nowait */);
+      switch (e->event_type)
+        {
+        case FIFO_EVENT_SERVER_RX:
+          handle_fifo_event_server_rx (utm, e);
+          break;
+
+        case FIFO_EVENT_SERVER_EXIT:
+          return;
+
+        default:
+          clib_warning ("unknown event type %d", e->event_type);
+          break;
+        }
+      if (PREDICT_FALSE(utm->time_to_stop == 1))
+        break;
+      if (PREDICT_FALSE(utm->time_to_print_stats == 1))
+        {
+          utm->time_to_print_stats = 0;
+          fformat(stdout, "%d connections\n", pool_elts (utm->sessions));
+        }
+    }
+}
+
+static void
+vl_api_bind_uri_reply_t_handler (vl_api_bind_uri_reply_t * mp)
+{
+  uri_tcp_test_main_t *utm = &uri_tcp_test_main;
+  svm_fifo_segment_create_args_t _a, *a = &_a;
+  int rv;
+
+  if (mp->retval)
+    {
+      clib_warning("bind failed: %d", mp->retval);
+      return;
+    }
+
+  if (mp->segment_name_length == 0)
+    {
+      clib_warning("segment_name_length zero");
+      return;
+    }
+
+  a->segment_name = (char *) mp->segment_name;
+  a->segment_size = mp->segment_size;
+
+  ASSERT(mp->server_event_queue_address);
+
+  /* Attach to the segment vpp created */
+  rv = svm_fifo_segment_attach (a);
+  if (rv)
+    {
+      clib_warning("svm_fifo_segment_attach ('%s') failed", mp->segment_name);
+      return;
+    }
+
+  utm->our_event_queue =
+      (unix_shared_memory_queue_t *) mp->server_event_queue_address;
+
+  utm->state = STATE_READY;
+}
+
+static void
+vl_api_connect_uri_reply_t_handler (vl_api_connect_uri_reply_t * mp)
+{
+  uri_tcp_test_main_t *utm = &uri_tcp_test_main;
+  svm_fifo_segment_create_args_t _a, *a = &_a;
+  session_t *session;
+  u32 session_index;
+  svm_fifo_t *rx_fifo, *tx_fifo;
+  int rv;
+
+  if (mp->retval)
+    {
+      clib_warning ("connection failed with code: %d", mp->retval);
+      utm->state = STATE_FAILED;
+      return;
+    }
+  /*
+   * Attatch to segment
+   */
+
+  if (mp->segment_name_length == 0)
+    {
+      clib_warning ("segment_name_length zero");
+      utm->state = STATE_FAILED;
+      return;
+    }
+
+  a->segment_name = (char *) mp->segment_name;
+  a->segment_size = mp->segment_size;
+
+  ASSERT(mp->client_event_queue_address);
+
+  /* Attach to the segment vpp created */
+  rv = svm_fifo_segment_attach (a);
+  if (rv)
+    {
+      clib_warning ("svm_fifo_segment_attach ('%s') failed",
+                    mp->segment_name);
+      return;
+    }
+
+  /*
+   * Save the queues
+   */
+
+  utm->our_event_queue = (unix_shared_memory_queue_t *)
+    mp->client_event_queue_address;
+
+  utm->vpp_event_queue = (unix_shared_memory_queue_t *)
+    mp->vpp_event_queue_address;
+
+  /*
+   * Setup session
+   */
+
+  pool_get (utm->sessions, session);
+  session_index = session - utm->sessions;
+
+  rx_fifo = (svm_fifo_t *)mp->server_rx_fifo;
+  rx_fifo->client_session_index = session_index;
+  tx_fifo = (svm_fifo_t *)mp->server_tx_fifo;
+  tx_fifo->client_session_index = session_index;
+
+  session->server_rx_fifo = rx_fifo;
+  session->server_tx_fifo = tx_fifo;
+  session->vpp_session_index = mp->session_index;
+  session->vpp_session_thread = mp->session_thread_index;
+
+  /* Save handle */
+  utm->connected_session_index = session_index;
+
+  utm->state = STATE_READY;
+}
+
+void
+uri_tcp_bind (uri_tcp_test_main_t *utm)
+{
+  vl_api_bind_uri_t * bmp;
+  u32 fifo_size = 3 << 20;
+  bmp = vl_msg_api_alloc (sizeof (*bmp));
+  memset (bmp, 0, sizeof (*bmp));
+
+  bmp->_vl_msg_id = ntohs (VL_API_BIND_URI);
+  bmp->client_index = utm->my_client_index;
+  bmp->context = ntohl(0xfeedface);
+  bmp->initial_segment_size = 256<<20;    /* size of initial segment */
+  bmp->options[SESSION_OPTIONS_FLAGS] =
+    SESSION_OPTIONS_FLAGS_USE_FIFO | SESSION_OPTIONS_FLAGS_ADD_SEGMENT;
+  bmp->options[SESSION_OPTIONS_RX_FIFO_SIZE] = fifo_size;
+  bmp->options[SESSION_OPTIONS_TX_FIFO_SIZE] = fifo_size;
+  bmp->options[SESSION_OPTIONS_ADD_SEGMENT_SIZE] = 128<<20;
+  memcpy (bmp->uri, utm->uri, vec_len (utm->uri));
+  vl_msg_api_send_shmem (utm->vl_input_queue, (u8 *)&bmp);
+}
+
+static void
+vl_api_unbind_uri_reply_t_handler (vl_api_unbind_uri_reply_t *mp)
+{
+  uri_tcp_test_main_t *utm = &uri_tcp_test_main;
+
+  if (mp->retval != 0)
+    clib_warning ("returned %d", ntohl(mp->retval));
+
+  utm->state = STATE_START;
+}
+
+void
+uri_tcp_unbind (uri_tcp_test_main_t *utm)
+{
+  vl_api_unbind_uri_t * ump;
+
+  ump = vl_msg_api_alloc (sizeof (*ump));
+  memset (ump, 0, sizeof (*ump));
+
+  ump->_vl_msg_id = ntohs (VL_API_UNBIND_URI);
+  ump->client_index = utm->my_client_index;
+  memcpy (ump->uri, utm->uri, vec_len (utm->uri));
+  vl_msg_api_send_shmem (utm->vl_input_queue, (u8 *)&ump);
+}
+
+static void
+vl_api_accept_session_t_handler (vl_api_accept_session_t * mp)
+{
+  uri_tcp_test_main_t *utm = &uri_tcp_test_main;
+  vl_api_accept_session_reply_t *rmp;
+  svm_fifo_t * rx_fifo, * tx_fifo;
+  session_t * session;
+  static f64 start_time;
+  u64 key;
+  u32 session_index;
+
+  if (start_time == 0.0)
+      start_time = clib_time_now (&utm->clib_time);
+
+  utm->vpp_event_queue = (unix_shared_memory_queue_t *)
+    mp->vpp_event_queue_address;
+
+  /* Allocate local session and set it up */
+  pool_get (utm->sessions, session);
+  session_index = session - utm->sessions;
+
+  rx_fifo = (svm_fifo_t *)mp->server_rx_fifo;
+  rx_fifo->client_session_index = session_index;
+  tx_fifo = (svm_fifo_t *)mp->server_tx_fifo;
+  tx_fifo->client_session_index = session_index;
+
+  session->server_rx_fifo = rx_fifo;
+  session->server_tx_fifo = tx_fifo;
+
+  /* Add it to lookup table */
+  key = (((u64)mp->session_thread_index) << 32) | (u64)mp->session_index;
+  hash_set (utm->session_index_by_vpp_handles, key, session_index);
+
+  utm->state = STATE_READY;
+
+  /* Stats printing */
+  if (pool_elts (utm->sessions) && (pool_elts(utm->sessions) % 20000) == 0)
+    {
+      f64 now = clib_time_now (&utm->clib_time);
+      fformat (stdout, "%d active sessions in %.2f seconds, %.2f/sec...\n",
+               pool_elts(utm->sessions), now - start_time,
+               (f64)pool_elts(utm->sessions) / (now - start_time));
+    }
+
+  /* Send accept reply to vpp */
+  rmp = vl_msg_api_alloc (sizeof (*rmp));
+  memset (rmp, 0, sizeof (*rmp));
+  rmp->_vl_msg_id = ntohs (VL_API_ACCEPT_SESSION_REPLY);
+  rmp->session_type = mp->session_type;
+  rmp->session_index = mp->session_index;
+  rmp->session_thread_index = mp->session_thread_index;
+  vl_msg_api_send_shmem (utm->vl_input_queue, (u8 *)&rmp);
+}
+
+void
+uri_tcp_server_test (uri_tcp_test_main_t * utm)
+{
+
+  /* Bind to uri */
+  uri_tcp_bind (utm);
+
+  if (wait_for_state_change (utm, STATE_READY))
+    {
+      clib_warning ("timeout waiting for STATE_READY");
+      return;
+    }
+
+  /* Enter handle event loop */
+  handle_event_queue (utm);
+
+  /* Cleanup */
+  uri_tcp_unbind (utm);
+
+  if (wait_for_state_change (utm, STATE_START))
+    {
+      clib_warning ("timeout waiting for STATE_START");
+      return;
+    }
+
+  fformat (stdout, "Test complete...\n");
+}
+
+#define foreach_uri_msg                         \
+_(BIND_URI_REPLY, bind_uri_reply)               \
+_(UNBIND_URI_REPLY, unbind_uri_reply)           \
+_(ACCEPT_SESSION, accept_session)               \
+_(CONNECT_URI_REPLY, connect_uri_reply)         \
+_(DISCONNECT_SESSION, disconnect_session)       \
+_(RESET_SESSION, reset_session)       		\
+_(MAP_ANOTHER_SEGMENT, map_another_segment)
+
+void
+uri_api_hookup (uri_tcp_test_main_t * utm)
+{
+#define _(N,n)                                                  \
+    vl_msg_api_set_handlers(VL_API_##N, #n,                     \
+                           vl_api_##n##_t_handler,              \
+                           vl_noop_handler,                     \
+                           vl_api_##n##_t_endian,               \
+                           vl_api_##n##_t_print,                \
+                           sizeof(vl_api_##n##_t), 1);
+  foreach_uri_msg;
+#undef _
+}
+
+int
+main (int argc, char **argv)
+{
+  uri_tcp_test_main_t *utm = &uri_tcp_test_main;
+  unformat_input_t _argv, *a = &_argv;
+  u8 *chroot_prefix;
+  u8 *heap;
+  u8 * bind_name = (u8 *) "tcp://0.0.0.0/1234";
+  u32 tmp;
+  mheap_t *h;
+  session_t * session;
+  int i;
+  int i_am_master = 1, drop_packets = 0;
+
+  clib_mem_init (0, 256 << 20);
+
+  heap = clib_mem_get_per_cpu_heap ();
+  h = mheap_header (heap);
+
+  /* make the main heap thread-safe */
+  h->flags |= MHEAP_FLAG_THREAD_SAFE;
+
+  vec_validate (utm->rx_buf, 65536);
+
+  utm->session_index_by_vpp_handles =
+    hash_create (0, sizeof(uword));
+
+  utm->my_pid = getpid();
+  utm->configured_segment_size = 1<<20;
+
+  clib_time_init (&utm->clib_time);
+  init_error_string_table (utm);
+  svm_fifo_segment_init(0x200000000ULL, 20);
+  unformat_init_command_line (a, argv);
+
+  while (unformat_check_input (a) != UNFORMAT_END_OF_INPUT)
+    {
+      if (unformat (a, "chroot prefix %s", &chroot_prefix))
+        {
+          vl_set_memory_root_path ((char *) chroot_prefix);
+        }
+      else if (unformat (a, "uri %s", &bind_name))
+        ;
+      else if (unformat (a, "segment-size %dM", &tmp))
+        utm->configured_segment_size = tmp<<20;
+      else if (unformat (a, "segment-size %dG", &tmp))
+        utm->configured_segment_size = tmp<<30;
+      else if (unformat (a, "master"))
+        i_am_master = 1;
+      else if (unformat (a, "slave"))
+        i_am_master = 0;
+      else if (unformat (a, "drop"))
+        drop_packets = 1;
+      else
+        {
+          fformat (stderr, "%s: usage [master|slave]\n");
+          exit (1);
+        }
+    }
+
+  utm->uri = format (0, "%s%c", bind_name, 0);
+  utm->i_am_master = i_am_master;
+  utm->segment_main = &svm_fifo_segment_main;
+  utm->drop_packets = drop_packets;
+
+  utm->connect_uri = format (0, "tcp://6.0.1.2/1234%c", 0);
+
+  setup_signal_handlers();
+  uri_api_hookup (utm);
+
+  if (connect_to_vpp (i_am_master? "uri_tcp_server":"uri_tcp_client") < 0)
+    {
+      svm_region_exit ();
+      fformat (stderr, "Couldn't connect to vpe, exiting...\n");
+      exit (1);
+    }
+
+  if (i_am_master == 0)
+    {
+      uri_tcp_client_test (utm);
+      exit (0);
+    }
+
+  /* $$$$ hack preallocation */
+  for (i = 0; i < 200000; i++)
+    {
+      pool_get (utm->sessions, session);
+      memset (session, 0, sizeof (*session));
+    }
+  for (i = 0; i < 200000; i++)
+    pool_put_index (utm->sessions, i);
+
+  uri_tcp_server_test (utm);
+
+  vl_client_disconnect_from_vlib ();
+  exit (0);
+}
diff --git a/src/uri/uri_udp_test.c b/src/uri/uri_udp_test.c
new file mode 100644
index 00000000..6f5284c9
--- /dev/null
+++ b/src/uri/uri_udp_test.c
@@ -0,0 +1,553 @@
+/*
+ * Copyright (c) 2016 Cisco and/or its affiliates.
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at:
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#include 
+#include 
+#include 
+#include 
+#include 
+#include 
+#include 
+#include 
+#include 
+#include 
+#include 
+#include 
+#include 
+#include 
+#include 
+
+#include 
+
+#define vl_typedefs		/* define message structures */
+#include 
+#undef vl_typedefs
+
+/* declare message handlers for each api */
+
+#define vl_endianfun		/* define message structures */
+#include 
+#undef vl_endianfun
+
+/* instantiate all the print functions we know about */
+#define vl_print(handle, ...)
+#define vl_printfun
+#include 
+#undef vl_printfun
+
+/* Satisfy external references when not linking with -lvlib */
+vlib_main_t vlib_global_main;
+vlib_main_t **vlib_mains;
+
+typedef enum
+{
+  STATE_START,
+  STATE_READY,
+  STATE_DISCONNECTING,
+} connection_state_t;
+
+typedef struct
+{
+  svm_fifo_t *server_rx_fifo;
+  svm_fifo_t *server_tx_fifo;
+} session_t;
+
+typedef struct
+{
+  /* vpe input queue */
+  unix_shared_memory_queue_t *vl_input_queue;
+
+  /* API client handle */
+  u32 my_client_index;
+
+  /* The URI we're playing with */
+  u8 *uri;
+
+  /* Session pool */
+  session_t *sessions;
+
+  /* Hash table for disconnect processing */
+  uword *session_index_by_vpp_handles;
+
+  /* fifo segment */
+  svm_fifo_segment_private_t *seg;
+
+  /* intermediate rx buffer */
+  u8 *rx_buf;
+
+  /* Our event queue */
+  unix_shared_memory_queue_t *our_event_queue;
+
+  /* $$$ single thread only for the moment */
+  unix_shared_memory_queue_t *vpp_event_queue;
+
+  /* For deadman timers */
+  clib_time_t clib_time;
+
+  /* State of the connection, shared between msg RX thread and main thread */
+  volatile connection_state_t state;
+
+  volatile int time_to_stop;
+  volatile int time_to_print_stats;
+
+  /* VNET_API_ERROR_FOO -> "Foo" hash table */
+  uword *error_string_by_error_number;
+} uri_udp_test_main_t;
+
+#if CLIB_DEBUG > 0
+#define NITER 1000
+#else
+#define NITER 1000000
+#endif
+
+uri_udp_test_main_t uri_udp_test_main;
+
+static void
+stop_signal (int signum)
+{
+  uri_udp_test_main_t *um = &uri_udp_test_main;
+
+  um->time_to_stop = 1;
+}
+
+static void
+stats_signal (int signum)
+{
+  uri_udp_test_main_t *um = &uri_udp_test_main;
+
+  um->time_to_print_stats = 1;
+}
+
+static clib_error_t *
+setup_signal_handlers (void)
+{
+  signal (SIGINT, stats_signal);
+  signal (SIGQUIT, stop_signal);
+  signal (SIGTERM, stop_signal);
+
+  return 0;
+}
+
+u8 *
+format_api_error (u8 * s, va_list * args)
+{
+  uri_udp_test_main_t *utm = va_arg (*args, uri_udp_test_main_t *);
+  i32 error = va_arg (*args, u32);
+  uword *p;
+
+  p = hash_get (utm->error_string_by_error_number, -error);
+
+  if (p)
+    s = format (s, "%s", p[0]);
+  else
+    s = format (s, "%d", error);
+  return s;
+}
+
+int
+wait_for_state_change (uri_udp_test_main_t * utm, connection_state_t state)
+{
+  f64 timeout = clib_time_now (&utm->clib_time) + 5.0;
+
+  while (clib_time_now (&utm->clib_time) < timeout)
+    {
+      if (utm->state == state)
+	return 0;
+    }
+  return -1;
+}
+
+static void
+vl_api_bind_uri_reply_t_handler (vl_api_bind_uri_reply_t * mp)
+{
+  uri_udp_test_main_t *utm = &uri_udp_test_main;
+  svm_fifo_segment_create_args_t _a, *a = &_a;
+  int rv;
+
+  if (mp->segment_name_length == 0)
+    {
+      clib_warning ("segment_name_length zero");
+      return;
+    }
+
+  a->segment_name = (char *) mp->segment_name;
+
+  /* Attach to the segment vpp created */
+  rv = svm_fifo_segment_attach (a);
+  if (rv)
+    {
+      clib_warning ("sm_fifo_segment_create ('%s') failed", mp->segment_name);
+      return;
+    }
+
+  utm->our_event_queue = (unix_shared_memory_queue_t *)
+    mp->server_event_queue_address;
+
+  utm->state = STATE_READY;
+}
+
+static void
+vl_api_unbind_uri_reply_t_handler (vl_api_unbind_uri_reply_t * mp)
+{
+  uri_udp_test_main_t *utm = &uri_udp_test_main;
+
+  if (mp->retval != 0)
+    clib_warning ("returned %d", ntohl (mp->retval));
+
+  utm->state = STATE_START;
+}
+
+static void
+vl_api_accept_session_t_handler (vl_api_accept_session_t * mp)
+{
+  uri_udp_test_main_t *utm = &uri_udp_test_main;
+  vl_api_accept_session_reply_t *rmp;
+  svm_fifo_t *rx_fifo, *tx_fifo;
+  session_t *session;
+  static f64 start_time;
+  u64 key;
+
+  if (start_time == 0.0)
+    start_time = clib_time_now (&utm->clib_time);
+
+  utm->vpp_event_queue = (unix_shared_memory_queue_t *)
+    mp->vpp_event_queue_address;
+
+  pool_get (utm->sessions, session);
+
+  rx_fifo = (svm_fifo_t *) mp->server_rx_fifo;
+  rx_fifo->client_session_index = session - utm->sessions;
+  tx_fifo = (svm_fifo_t *) mp->server_tx_fifo;
+  tx_fifo->client_session_index = session - utm->sessions;
+
+  session->server_rx_fifo = rx_fifo;
+  session->server_tx_fifo = tx_fifo;
+
+  key = (((u64) mp->session_thread_index) << 32) | (u64) mp->session_index;
+
+  hash_set (utm->session_index_by_vpp_handles, key, session - utm->sessions);
+
+  utm->state = STATE_READY;
+
+  if (pool_elts (utm->sessions) && (pool_elts (utm->sessions) % 20000) == 0)
+    {
+      f64 now = clib_time_now (&utm->clib_time);
+      fformat (stdout, "%d active sessions in %.2f seconds, %.2f/sec...\n",
+	       pool_elts (utm->sessions), now - start_time,
+	       (f64) pool_elts (utm->sessions) / (now - start_time));
+    }
+
+  rmp = vl_msg_api_alloc (sizeof (*rmp));
+  memset (rmp, 0, sizeof (*rmp));
+  rmp->_vl_msg_id = ntohs (VL_API_ACCEPT_SESSION_REPLY);
+  rmp->session_type = mp->session_type;
+  rmp->session_index = mp->session_index;
+  rmp->session_thread_index = mp->session_thread_index;
+  vl_msg_api_send_shmem (utm->vl_input_queue, (u8 *) & rmp);
+}
+
+static void
+vl_api_disconnect_session_t_handler (vl_api_disconnect_session_t * mp)
+{
+  uri_udp_test_main_t *utm = &uri_udp_test_main;
+  session_t *session;
+  vl_api_disconnect_session_reply_t *rmp;
+  uword *p;
+  int rv = 0;
+  u64 key;
+
+  key = (((u64) mp->session_thread_index) << 32) | (u64) mp->session_index;
+
+  p = hash_get (utm->session_index_by_vpp_handles, key);
+
+  if (p)
+    {
+      session = pool_elt_at_index (utm->sessions, p[0]);
+      hash_unset (utm->session_index_by_vpp_handles, key);
+      pool_put (utm->sessions, session);
+    }
+  else
+    {
+      clib_warning ("couldn't find session key %llx", key);
+      rv = -11;
+    }
+
+  rmp = vl_msg_api_alloc (sizeof (*rmp));
+  memset (rmp, 0, sizeof (*rmp));
+  rmp->_vl_msg_id = ntohs (VL_API_DISCONNECT_SESSION_REPLY);
+  rmp->retval = rv;
+  rmp->session_index = mp->session_index;
+  rmp->session_thread_index = mp->session_thread_index;
+  vl_msg_api_send_shmem (utm->vl_input_queue, (u8 *) & rmp);
+}
+
+#define foreach_uri_msg                         \
+_(BIND_URI_REPLY, bind_uri_reply)               \
+_(UNBIND_URI_REPLY, unbind_uri_reply)           \
+_(ACCEPT_SESSION, accept_session)		\
+_(DISCONNECT_SESSION, disconnect_session)
+
+void
+uri_api_hookup (uri_udp_test_main_t * utm)
+{
+#define _(N,n)                                                  \
+    vl_msg_api_set_handlers(VL_API_##N, #n,                     \
+                           vl_api_##n##_t_handler,	        \
+                           vl_noop_handler,                     \
+                           vl_api_##n##_t_endian,               \
+                           vl_api_##n##_t_print,                \
+                           sizeof(vl_api_##n##_t), 1);
+  foreach_uri_msg;
+#undef _
+
+}
+
+
+int
+connect_to_vpp (char *name)
+{
+  uri_udp_test_main_t *utm = &uri_udp_test_main;
+  api_main_t *am = &api_main;
+
+  if (vl_client_connect_to_vlib ("/vpe-api", name, 32) < 0)
+    return -1;
+
+  utm->vl_input_queue = am->shmem_hdr->vl_input_queue;
+  utm->my_client_index = am->my_client_index;
+
+  return 0;
+}
+
+void
+vlib_cli_output (struct vlib_main_t *vm, char *fmt, ...)
+{
+  clib_warning ("BUG");
+}
+
+static void
+init_error_string_table (uri_udp_test_main_t * utm)
+{
+  utm->error_string_by_error_number = hash_create (0, sizeof (uword));
+
+#define _(n,v,s) hash_set (utm->error_string_by_error_number, -v, s);
+  foreach_vnet_api_error;
+#undef _
+
+  hash_set (utm->error_string_by_error_number, 99, "Misc");
+}
+
+void
+handle_fifo_event_server_rx (uri_udp_test_main_t * utm,
+			     session_fifo_event_t * e)
+{
+  svm_fifo_t *rx_fifo, *tx_fifo;
+  int nbytes;
+
+  session_fifo_event_t evt;
+  unix_shared_memory_queue_t *q;
+  int rv;
+
+  rx_fifo = e->fifo;
+  tx_fifo = utm->sessions[rx_fifo->client_session_index].server_tx_fifo;
+
+  do
+    {
+      nbytes = svm_fifo_dequeue_nowait (rx_fifo, 0,
+					vec_len (utm->rx_buf), utm->rx_buf);
+    }
+  while (nbytes <= 0);
+  do
+    {
+      rv = svm_fifo_enqueue_nowait (tx_fifo, 0, nbytes, utm->rx_buf);
+    }
+  while (rv == -2);
+
+  /* Fabricate TX event, send to vpp */
+  evt.fifo = tx_fifo;
+  evt.event_type = FIFO_EVENT_SERVER_TX;
+  /* $$$$ for event logging */
+  evt.enqueue_length = nbytes;
+  evt.event_id = e->event_id;
+  q = utm->vpp_event_queue;
+  unix_shared_memory_queue_add (q, (u8 *) & evt, 0 /* do wait for mutex */ );
+}
+
+void
+handle_event_queue (uri_udp_test_main_t * utm)
+{
+  session_fifo_event_t _e, *e = &_e;;
+
+  while (1)
+    {
+      unix_shared_memory_queue_sub (utm->our_event_queue, (u8 *) e,
+				    0 /* nowait */ );
+      switch (e->event_type)
+	{
+	case FIFO_EVENT_SERVER_RX:
+	  handle_fifo_event_server_rx (utm, e);
+	  break;
+
+	case FIFO_EVENT_SERVER_EXIT:
+	  return;
+
+	default:
+	  clib_warning ("unknown event type %d", e->event_type);
+	  break;
+	}
+      if (PREDICT_FALSE (utm->time_to_stop == 1))
+	break;
+      if (PREDICT_FALSE (utm->time_to_print_stats == 1))
+	{
+	  utm->time_to_print_stats = 0;
+	  fformat (stdout, "%d connections\n", pool_elts (utm->sessions));
+	}
+    }
+}
+
+void
+uri_udp_test (uri_udp_test_main_t * utm)
+{
+  vl_api_bind_uri_t *bmp;
+  vl_api_unbind_uri_t *ump;
+
+  bmp = vl_msg_api_alloc (sizeof (*bmp));
+  memset (bmp, 0, sizeof (*bmp));
+
+  bmp->_vl_msg_id = ntohs (VL_API_BIND_URI);
+  bmp->client_index = utm->my_client_index;
+  bmp->context = ntohl (0xfeedface);
+  bmp->segment_size = 2 << 30;
+  memcpy (bmp->uri, utm->uri, vec_len (utm->uri));
+  vl_msg_api_send_shmem (utm->vl_input_queue, (u8 *) & bmp);
+
+  if (wait_for_state_change (utm, STATE_READY))
+    {
+      clib_warning ("timeout waiting for STATE_READY");
+      return;
+    }
+
+  handle_event_queue (utm);
+
+  ump = vl_msg_api_alloc (sizeof (*ump));
+  memset (ump, 0, sizeof (*ump));
+
+  ump->_vl_msg_id = ntohs (VL_API_UNBIND_URI);
+  ump->client_index = utm->my_client_index;
+  memcpy (ump->uri, utm->uri, vec_len (utm->uri));
+  vl_msg_api_send_shmem (utm->vl_input_queue, (u8 *) & ump);
+
+  if (wait_for_state_change (utm, STATE_START))
+    {
+      clib_warning ("timeout waiting for STATE_START");
+      return;
+    }
+
+  fformat (stdout, "Test complete...\n");
+}
+
+int
+main (int argc, char **argv)
+{
+  uri_udp_test_main_t *utm = &uri_udp_test_main;
+  unformat_input_t _argv, *a = &_argv;
+  u8 *chroot_prefix;
+  u8 *heap;
+  u8 *bind_name = (u8 *) "udp4:1234";
+  mheap_t *h;
+  session_t *session;
+  int i;
+
+  clib_mem_init (0, 256 << 20);
+
+  heap = clib_mem_get_per_cpu_heap ();
+  h = mheap_header (heap);
+
+  /* make the main heap thread-safe */
+  h->flags |= MHEAP_FLAG_THREAD_SAFE;
+
+  vec_validate (utm->rx_buf, 8192);
+
+  utm->session_index_by_vpp_handles = hash_create (0, sizeof (uword));
+
+  clib_time_init (&utm->clib_time);
+  init_error_string_table (utm);
+  svm_fifo_segment_init (0x200000000ULL, 20);
+  unformat_init_command_line (a, argv);
+
+  while (unformat_check_input (a) != UNFORMAT_END_OF_INPUT)
+    {
+      if (unformat (a, "chroot prefix %s", &chroot_prefix))
+	{
+	  vl_set_memory_root_path ((char *) chroot_prefix);
+	}
+      else if (unformat (a, "uri %s", &bind_name))
+	;
+      else
+	{
+	  fformat (stderr, "%s: usage [master|slave]\n");
+	  exit (1);
+	}
+    }
+
+  utm->uri = format (0, "%s%c", bind_name, 0);
+
+  setup_signal_handlers ();
+
+  uri_api_hookup (utm);
+
+  if (connect_to_vpp ("uri_udp_test") < 0)
+    {
+      svm_region_exit ();
+      fformat (stderr, "Couldn't connect to vpe, exiting...\n");
+      exit (1);
+    }
+
+  /* $$$$ hack preallocation */
+  for (i = 0; i < 200000; i++)
+    {
+      pool_get (utm->sessions, session);
+      memset (session, 0, sizeof (*session));
+    }
+  for (i = 0; i < 200000; i++)
+    pool_put_index (utm->sessions, i);
+
+  uri_udp_test (utm);
+
+  vl_client_disconnect_from_vlib ();
+  exit (0);
+}
+
+#undef vl_api_version
+#define vl_api_version(n,v) static u32 vpe_api_version = v;
+#include 
+#undef vl_api_version
+
+void
+vl_client_add_api_signatures (vl_api_memclnt_create_t * mp)
+{
+  /*
+   * Send the main API signature in slot 0. This bit of code must
+   * match the checks in ../vpe/api/api.c: vl_msg_api_version_check().
+   */
+  mp->api_versions[0] = clib_host_to_net_u32 (vpe_api_version);
+}
+
+/*
+ * fd.io coding-style-patch-verification: ON
+ *
+ * Local Variables:
+ * eval: (c-set-style "gnu")
+ * End:
+ */
diff --git a/src/uri/uri_udp_test2.c b/src/uri/uri_udp_test2.c
new file mode 100644
index 00000000..ddfffaa6
--- /dev/null
+++ b/src/uri/uri_udp_test2.c
@@ -0,0 +1,954 @@
+/*
+ * Copyright (c) 2016 Cisco and/or its affiliates.
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at:
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#include 
+#include 
+#include 
+#include 
+#include 
+#include 
+#include 
+#include 
+#include 
+#include 
+#include 
+#include 
+#include 
+#include 
+#include 
+#include 
+
+#include "../vnet/session/application_interface.h"
+
+#define vl_typedefs		/* define message structures */
+#include 
+#undef vl_typedefs
+
+/* declare message handlers for each api */
+
+#define vl_endianfun		/* define message structures */
+#include 
+#undef vl_endianfun
+
+/* instantiate all the print functions we know about */
+#define vl_print(handle, ...)
+#define vl_printfun
+#include 
+#undef vl_printfun
+
+/* Satisfy external references when not linking with -lvlib */
+vlib_main_t vlib_global_main;
+vlib_main_t **vlib_mains;
+
+typedef enum
+{
+  STATE_START,
+  STATE_READY,
+  STATE_DISCONNECTING,
+} connection_state_t;
+
+typedef struct
+{
+  svm_fifo_t *server_rx_fifo;
+  svm_fifo_t *server_tx_fifo;
+} session_t;
+
+typedef struct
+{
+  /* vpe input queue */
+  unix_shared_memory_queue_t *vl_input_queue;
+
+  /* API client handle */
+  u32 my_client_index;
+
+  /* The URI we're playing with */
+  u8 *uri;
+
+  /* Session pool */
+  session_t *sessions;
+
+  /* Hash table for disconnect processing */
+  uword *session_index_by_vpp_handles;
+
+  /* fifo segment */
+  svm_fifo_segment_private_t *seg;
+
+  /* intermediate rx buffer */
+  u8 *rx_buf;
+
+  /* URI for connect */
+  u8 *connect_uri;
+
+  int i_am_master;
+
+  /* Our event queue */
+  unix_shared_memory_queue_t *our_event_queue;
+
+  /* $$$ single thread only for the moment */
+  unix_shared_memory_queue_t *vpp_event_queue;
+
+  /* $$$$ hack: cut-through session index */
+  volatile u32 cut_through_session_index;
+
+  /* unique segment name counter */
+  u32 unique_segment_index;
+
+  pid_t my_pid;
+
+  /* pthread handle */
+  pthread_t cut_through_thread_handle;
+
+  /* For deadman timers */
+  clib_time_t clib_time;
+
+  /* State of the connection, shared between msg RX thread and main thread */
+  volatile connection_state_t state;
+
+  volatile int time_to_stop;
+  volatile int time_to_print_stats;
+
+  u32 configured_segment_size;
+
+  /* VNET_API_ERROR_FOO -> "Foo" hash table */
+  uword *error_string_by_error_number;
+
+  /* convenience */
+  svm_fifo_segment_main_t *segment_main;
+
+} uri_udp_test_main_t;
+
+#if CLIB_DEBUG > 0
+#define NITER 10000
+#else
+#define NITER 4000000
+#endif
+
+uri_udp_test_main_t uri_udp_test_main;
+
+static void
+stop_signal (int signum)
+{
+  uri_udp_test_main_t *um = &uri_udp_test_main;
+
+  um->time_to_stop = 1;
+}
+
+static void
+stats_signal (int signum)
+{
+  uri_udp_test_main_t *um = &uri_udp_test_main;
+
+  um->time_to_print_stats = 1;
+}
+
+static clib_error_t *
+setup_signal_handlers (void)
+{
+  signal (SIGINT, stats_signal);
+  signal (SIGQUIT, stop_signal);
+  signal (SIGTERM, stop_signal);
+
+  return 0;
+}
+
+u8 *
+format_api_error (u8 * s, va_list * args)
+{
+  uri_udp_test_main_t *utm = va_arg (*args, uri_udp_test_main_t *);
+  i32 error = va_arg (*args, u32);
+  uword *p;
+
+  p = hash_get (utm->error_string_by_error_number, -error);
+
+  if (p)
+    s = format (s, "%s", p[0]);
+  else
+    s = format (s, "%d", error);
+  return s;
+}
+
+int
+wait_for_state_change (uri_udp_test_main_t * utm, connection_state_t state)
+{
+#if CLIB_DEBUG > 0
+#define TIMEOUT 600.0
+#else
+#define TIMEOUT 600.0
+#endif
+
+  f64 timeout = clib_time_now (&utm->clib_time) + TIMEOUT;
+
+  while (clib_time_now (&utm->clib_time) < timeout)
+    {
+      if (utm->state == state)
+	return 0;
+    }
+  return -1;
+}
+
+u64 server_bytes_received, server_bytes_sent;
+
+static void *
+cut_through_thread_fn (void *arg)
+{
+  session_t *s;
+  svm_fifo_t *rx_fifo;
+  svm_fifo_t *tx_fifo;
+  u8 *my_copy_buffer = 0;
+  uri_udp_test_main_t *utm = &uri_udp_test_main;
+  i32 actual_transfer;
+  int rv;
+  u32 buffer_offset;
+
+  while (utm->cut_through_session_index == ~0)
+    ;
+
+  s = pool_elt_at_index (utm->sessions, utm->cut_through_session_index);
+
+  rx_fifo = s->server_rx_fifo;
+  tx_fifo = s->server_tx_fifo;
+
+  vec_validate (my_copy_buffer, 64 * 1024 - 1);
+
+  while (true)
+    {
+      /* We read from the tx fifo and write to the rx fifo */
+      do
+	{
+	  actual_transfer = svm_fifo_dequeue_nowait (tx_fifo, 0,
+						     vec_len (my_copy_buffer),
+						     my_copy_buffer);
+	}
+      while (actual_transfer <= 0);
+
+      server_bytes_received += actual_transfer;
+
+      buffer_offset = 0;
+      while (actual_transfer > 0)
+	{
+	  rv = svm_fifo_enqueue_nowait (rx_fifo, 0, actual_transfer,
+					my_copy_buffer + buffer_offset);
+	  if (rv > 0)
+	    {
+	      actual_transfer -= rv;
+	      buffer_offset += rv;
+	      server_bytes_sent += rv;
+	    }
+
+	}
+      if (PREDICT_FALSE (utm->time_to_stop))
+	break;
+    }
+
+  pthread_exit (0);
+}
+
+static void
+uri_udp_slave_test (uri_udp_test_main_t * utm)
+{
+  vl_api_connect_uri_t *cmp;
+  int i;
+  u8 *test_data = 0;
+  u64 bytes_received = 0, bytes_sent = 0;
+  i32 bytes_to_read;
+  int rv;
+  int mypid = getpid ();
+  f64 before, after, delta, bytes_per_second;
+  session_t *session;
+  svm_fifo_t *rx_fifo, *tx_fifo;
+  int buffer_offset, bytes_to_send = 0;
+
+  vec_validate (test_data, 64 * 1024 - 1);
+  for (i = 0; i < vec_len (test_data); i++)
+    test_data[i] = i & 0xff;
+
+  cmp = vl_msg_api_alloc (sizeof (*cmp));
+  memset (cmp, 0, sizeof (*cmp));
+
+  cmp->_vl_msg_id = ntohs (VL_API_CONNECT_URI);
+  cmp->client_index = utm->my_client_index;
+  cmp->context = ntohl (0xfeedface);
+  memcpy (cmp->uri, utm->connect_uri, vec_len (utm->connect_uri));
+  vl_msg_api_send_shmem (utm->vl_input_queue, (u8 *) & cmp);
+
+  if (wait_for_state_change (utm, STATE_READY))
+    {
+      clib_warning ("timeout waiting for STATE_READY");
+      return;
+    }
+
+  session = pool_elt_at_index (utm->sessions, utm->cut_through_session_index);
+  rx_fifo = session->server_rx_fifo;
+  tx_fifo = session->server_tx_fifo;
+
+  before = clib_time_now (&utm->clib_time);
+
+  vec_validate (utm->rx_buf, vec_len (test_data) - 1);
+
+  for (i = 0; i < NITER; i++)
+    {
+      bytes_to_send = vec_len (test_data);
+      buffer_offset = 0;
+      while (bytes_to_send > 0)
+	{
+	  rv = svm_fifo_enqueue_nowait (tx_fifo, mypid,
+					bytes_to_send,
+					test_data + buffer_offset);
+
+	  if (rv > 0)
+	    {
+	      bytes_to_send -= rv;
+	      buffer_offset += rv;
+	      bytes_sent += rv;
+	    }
+	}
+
+      bytes_to_read = svm_fifo_max_dequeue (rx_fifo);
+
+      bytes_to_read = vec_len (utm->rx_buf) > bytes_to_read ?
+	bytes_to_read : vec_len (utm->rx_buf);
+
+      buffer_offset = 0;
+      while (bytes_to_read > 0)
+	{
+	  rv = svm_fifo_dequeue_nowait (rx_fifo, mypid,
+					bytes_to_read,
+					utm->rx_buf + buffer_offset);
+	  if (rv > 0)
+	    {
+	      bytes_to_read -= rv;
+	      buffer_offset += rv;
+	      bytes_received += rv;
+	    }
+	}
+    }
+  while (bytes_received < bytes_sent)
+    {
+      rv = svm_fifo_dequeue_nowait (rx_fifo, mypid,
+				    vec_len (utm->rx_buf), utm->rx_buf);
+      if (rv > 0)
+	{
+#if CLIB_DEBUG > 0
+	  int j;
+	  for (j = 0; j < rv; j++)
+	    {
+	      if (utm->rx_buf[j] != ((bytes_received + j) & 0xff))
+		{
+		  clib_warning ("error at byte %lld, 0x%x not 0x%x",
+				bytes_received + j,
+				utm->rx_buf[j],
+				((bytes_received + j) & 0xff));
+		}
+	    }
+#endif
+	  bytes_received += (u64) rv;
+	}
+    }
+
+  after = clib_time_now (&utm->clib_time);
+  delta = after - before;
+  bytes_per_second = 0.0;
+
+  if (delta > 0.0)
+    bytes_per_second = (f64) bytes_received / delta;
+
+  fformat (stdout,
+	   "Done: %lld recv bytes in %.2f seconds, %.2f bytes/sec...\n\n",
+	   bytes_received, delta, bytes_per_second);
+  fformat (stdout,
+	   "Done: %lld sent bytes in %.2f seconds, %.2f bytes/sec...\n\n",
+	   bytes_sent, delta, bytes_per_second);
+  fformat (stdout,
+	   "client -> server -> client round trip: %.2f Gbit/sec \n\n",
+	   (bytes_per_second * 8.0) / 1e9);
+}
+
+static void
+vl_api_bind_uri_reply_t_handler (vl_api_bind_uri_reply_t * mp)
+{
+  uri_udp_test_main_t *utm = &uri_udp_test_main;
+  svm_fifo_segment_create_args_t _a, *a = &_a;
+  int rv;
+
+  if (mp->segment_name_length == 0)
+    {
+      clib_warning ("segment_name_length zero");
+      return;
+    }
+
+  a->segment_name = (char *) mp->segment_name;
+  a->segment_size = mp->segment_size;
+
+  ASSERT (mp->server_event_queue_address);
+
+  /* Attach to the segment vpp created */
+  rv = svm_fifo_segment_attach (a);
+  if (rv)
+    {
+      clib_warning ("svm_fifo_segment_attach ('%s') failed",
+		    mp->segment_name);
+      return;
+    }
+
+  utm->our_event_queue = (unix_shared_memory_queue_t *)
+    mp->server_event_queue_address;
+
+  utm->state = STATE_READY;
+}
+
+static void
+vl_api_map_another_segment_t_handler (vl_api_map_another_segment_t * mp)
+{
+  svm_fifo_segment_create_args_t _a, *a = &_a;
+  int rv;
+
+  a->segment_name = (char *) mp->segment_name;
+  a->segment_size = mp->segment_size;
+  /* Attach to the segment vpp created */
+  rv = svm_fifo_segment_attach (a);
+  if (rv)
+    {
+      clib_warning ("svm_fifo_segment_attach ('%s') failed",
+		    mp->segment_name);
+      return;
+    }
+  clib_warning ("Mapped new segment '%s' size %d", mp->segment_name,
+		mp->segment_size);
+}
+
+static void
+vl_api_connect_uri_t_handler (vl_api_connect_uri_t * mp)
+{
+  u32 segment_index;
+  uri_udp_test_main_t *utm = &uri_udp_test_main;
+  svm_fifo_segment_main_t *sm = &svm_fifo_segment_main;
+  svm_fifo_segment_create_args_t _a, *a = &_a;
+  svm_fifo_segment_private_t *seg;
+  unix_shared_memory_queue_t *client_q;
+  vl_api_connect_uri_reply_t *rmp;
+  session_t *session;
+  int rv = 0;
+
+  /* Create the segment */
+  a->segment_name = (char *) format (0, "%d:segment%d%c", utm->my_pid,
+				     utm->unique_segment_index++, 0);
+  a->segment_size = utm->configured_segment_size;
+
+  rv = svm_fifo_segment_create (a);
+  if (rv)
+    {
+      clib_warning ("sm_fifo_segment_create ('%s') failed", a->segment_name);
+      rv = VNET_API_ERROR_URI_FIFO_CREATE_FAILED;
+      goto send_reply;
+    }
+
+  vec_add2 (utm->seg, seg, 1);
+
+  segment_index = vec_len (sm->segments) - 1;
+
+  memcpy (seg, sm->segments + segment_index, sizeof (utm->seg[0]));
+
+  pool_get (utm->sessions, session);
+
+  /*
+   * By construction the master's idea of the rx fifo ends up in
+   * fsh->fifos[0], and the master's idea of the tx fifo ends up in
+   * fsh->fifos[1].
+   */
+  session->server_rx_fifo = svm_fifo_segment_alloc_fifo (utm->seg,
+							 128 * 1024);
+  ASSERT (session->server_rx_fifo);
+
+  session->server_tx_fifo = svm_fifo_segment_alloc_fifo (utm->seg,
+							 128 * 1024);
+  ASSERT (session->server_tx_fifo);
+
+  session->server_rx_fifo->server_session_index = session - utm->sessions;
+  session->server_tx_fifo->server_session_index = session - utm->sessions;
+  utm->cut_through_session_index = session - utm->sessions;
+
+  rv = pthread_create (&utm->cut_through_thread_handle,
+		       NULL /*attr */ , cut_through_thread_fn, 0);
+  if (rv)
+    {
+      clib_warning ("pthread_create returned %d", rv);
+      rv = VNET_API_ERROR_SYSCALL_ERROR_1;
+    }
+
+send_reply:
+  rmp = vl_msg_api_alloc (sizeof (*rmp));
+  memset (rmp, 0, sizeof (*rmp));
+
+  rmp->_vl_msg_id = ntohs (VL_API_CONNECT_URI_REPLY);
+  rmp->context = mp->context;
+  rmp->retval = ntohl (rv);
+  rmp->segment_name_length = vec_len (a->segment_name);
+  memcpy (rmp->segment_name, a->segment_name, vec_len (a->segment_name));
+
+  vec_free (a->segment_name);
+
+  client_q = (unix_shared_memory_queue_t *) mp->client_queue_address;
+  vl_msg_api_send_shmem (client_q, (u8 *) & rmp);
+}
+
+static void
+vl_api_unbind_uri_reply_t_handler (vl_api_unbind_uri_reply_t * mp)
+{
+  uri_udp_test_main_t *utm = &uri_udp_test_main;
+
+  if (mp->retval != 0)
+    clib_warning ("returned %d", ntohl (mp->retval));
+
+  utm->state = STATE_START;
+}
+
+static void
+vl_api_accept_session_t_handler (vl_api_accept_session_t * mp)
+{
+  uri_udp_test_main_t *utm = &uri_udp_test_main;
+  vl_api_accept_session_reply_t *rmp;
+  svm_fifo_t *rx_fifo, *tx_fifo;
+  session_t *session;
+  static f64 start_time;
+  u64 key;
+
+  if (start_time == 0.0)
+    start_time = clib_time_now (&utm->clib_time);
+
+  utm->vpp_event_queue = (unix_shared_memory_queue_t *)
+    mp->vpp_event_queue_address;
+
+  pool_get (utm->sessions, session);
+
+  rx_fifo = (svm_fifo_t *) mp->server_rx_fifo;
+  rx_fifo->client_session_index = session - utm->sessions;
+  tx_fifo = (svm_fifo_t *) mp->server_tx_fifo;
+  tx_fifo->client_session_index = session - utm->sessions;
+
+  session->server_rx_fifo = rx_fifo;
+  session->server_tx_fifo = tx_fifo;
+
+  key = (((u64) mp->session_thread_index) << 32) | (u64) mp->session_index;
+
+  hash_set (utm->session_index_by_vpp_handles, key, session - utm->sessions);
+
+  utm->state = STATE_READY;
+
+  if (pool_elts (utm->sessions) && (pool_elts (utm->sessions) % 20000) == 0)
+    {
+      f64 now = clib_time_now (&utm->clib_time);
+      fformat (stdout, "%d active sessions in %.2f seconds, %.2f/sec...\n",
+	       pool_elts (utm->sessions), now - start_time,
+	       (f64) pool_elts (utm->sessions) / (now - start_time));
+    }
+
+  rmp = vl_msg_api_alloc (sizeof (*rmp));
+  memset (rmp, 0, sizeof (*rmp));
+  rmp->_vl_msg_id = ntohs (VL_API_ACCEPT_SESSION_REPLY);
+  rmp->session_type = mp->session_type;
+  rmp->session_index = mp->session_index;
+  rmp->session_thread_index = mp->session_thread_index;
+  vl_msg_api_send_shmem (utm->vl_input_queue, (u8 *) & rmp);
+}
+
+static void
+vl_api_disconnect_session_t_handler (vl_api_disconnect_session_t * mp)
+{
+  uri_udp_test_main_t *utm = &uri_udp_test_main;
+  session_t *session;
+  vl_api_disconnect_session_reply_t *rmp;
+  uword *p;
+  int rv = 0;
+  u64 key;
+
+  key = (((u64) mp->session_thread_index) << 32) | (u64) mp->session_index;
+
+  p = hash_get (utm->session_index_by_vpp_handles, key);
+
+  if (p)
+    {
+      session = pool_elt_at_index (utm->sessions, p[0]);
+      hash_unset (utm->session_index_by_vpp_handles, key);
+      pool_put (utm->sessions, session);
+    }
+  else
+    {
+      clib_warning ("couldn't find session key %llx", key);
+      rv = -11;
+    }
+
+  rmp = vl_msg_api_alloc (sizeof (*rmp));
+  memset (rmp, 0, sizeof (*rmp));
+  rmp->_vl_msg_id = ntohs (VL_API_DISCONNECT_SESSION_REPLY);
+  rmp->retval = rv;
+  rmp->session_index = mp->session_index;
+  rmp->session_thread_index = mp->session_thread_index;
+  vl_msg_api_send_shmem (utm->vl_input_queue, (u8 *) & rmp);
+}
+
+static void
+vl_api_connect_uri_reply_t_handler (vl_api_connect_uri_reply_t * mp)
+{
+  svm_fifo_segment_main_t *sm = &svm_fifo_segment_main;
+  uri_udp_test_main_t *utm = &uri_udp_test_main;
+  svm_fifo_segment_create_args_t _a, *a = &_a;
+  ssvm_shared_header_t *sh;
+  svm_fifo_segment_private_t *seg;
+  svm_fifo_segment_header_t *fsh;
+  session_t *session;
+  u32 segment_index;
+  int rv;
+
+  ASSERT (utm->i_am_master == 0);
+
+  if (mp->segment_name_length == 0)
+    {
+      clib_warning ("segment_name_length zero");
+      return;
+    }
+
+  memset (a, 0, sizeof (*a));
+
+  a->segment_name = (char *) mp->segment_name;
+
+  sleep (1);
+
+  rv = svm_fifo_segment_attach (a);
+  if (rv)
+    {
+      clib_warning ("sm_fifo_segment_create ('%v') failed", mp->segment_name);
+      return;
+    }
+
+  segment_index = vec_len (sm->segments) - 1;
+
+  vec_add2 (utm->seg, seg, 1);
+
+  memcpy (seg, sm->segments + segment_index, sizeof (*seg));
+  sh = seg->ssvm.sh;
+  fsh = (svm_fifo_segment_header_t *) sh->opaque[0];
+
+  while (vec_len (fsh->fifos) < 2)
+    sleep (1);
+
+  pool_get (utm->sessions, session);
+  utm->cut_through_session_index = session - utm->sessions;
+
+  session->server_rx_fifo = (svm_fifo_t *) fsh->fifos[0];
+  ASSERT (session->server_rx_fifo);
+  session->server_tx_fifo = (svm_fifo_t *) fsh->fifos[1];
+  ASSERT (session->server_tx_fifo);
+
+  /* security: could unlink /dev/shm/segment_name> here, maybe */
+
+  utm->state = STATE_READY;
+}
+
+#define foreach_uri_msg                         \
+_(BIND_URI_REPLY, bind_uri_reply)               \
+_(CONNECT_URI, connect_uri)                     \
+_(CONNECT_URI_REPLY, connect_uri_reply)         \
+_(UNBIND_URI_REPLY, unbind_uri_reply)           \
+_(ACCEPT_SESSION, accept_session)		\
+_(DISCONNECT_SESSION, disconnect_session)	\
+_(MAP_ANOTHER_SEGMENT, map_another_segment)
+
+void
+uri_api_hookup (uri_udp_test_main_t * utm)
+{
+#define _(N,n)                                                  \
+    vl_msg_api_set_handlers(VL_API_##N, #n,                     \
+                           vl_api_##n##_t_handler,              \
+                           vl_noop_handler,                     \
+                           vl_api_##n##_t_endian,               \
+                           vl_api_##n##_t_print,                \
+                           sizeof(vl_api_##n##_t), 1);
+  foreach_uri_msg;
+#undef _
+
+}
+
+
+int
+connect_to_vpp (char *name)
+{
+  uri_udp_test_main_t *utm = &uri_udp_test_main;
+  api_main_t *am = &api_main;
+
+  if (vl_client_connect_to_vlib ("/vpe-api", name, 32) < 0)
+    return -1;
+
+  utm->vl_input_queue = am->shmem_hdr->vl_input_queue;
+  utm->my_client_index = am->my_client_index;
+
+  return 0;
+}
+
+void
+vlib_cli_output (struct vlib_main_t *vm, char *fmt, ...)
+{
+  clib_warning ("BUG");
+}
+
+static void
+init_error_string_table (uri_udp_test_main_t * utm)
+{
+  utm->error_string_by_error_number = hash_create (0, sizeof (uword));
+
+#define _(n,v,s) hash_set (utm->error_string_by_error_number, -v, s);
+  foreach_vnet_api_error;
+#undef _
+
+  hash_set (utm->error_string_by_error_number, 99, "Misc");
+}
+
+void
+handle_fifo_event_server_rx (uri_udp_test_main_t * utm,
+			     session_fifo_event_t * e)
+{
+  svm_fifo_t *rx_fifo, *tx_fifo;
+  int nbytes;
+
+  session_fifo_event_t evt;
+  unix_shared_memory_queue_t *q;
+  int rv;
+
+  rx_fifo = e->fifo;
+  tx_fifo = utm->sessions[rx_fifo->client_session_index].server_tx_fifo;
+
+  do
+    {
+      nbytes = svm_fifo_dequeue_nowait (rx_fifo, 0,
+					vec_len (utm->rx_buf), utm->rx_buf);
+    }
+  while (nbytes <= 0);
+  do
+    {
+      rv = svm_fifo_enqueue_nowait (tx_fifo, 0, nbytes, utm->rx_buf);
+    }
+  while (rv == -2);
+
+  /* Fabricate TX event, send to vpp */
+  evt.fifo = tx_fifo;
+  evt.event_type = FIFO_EVENT_SERVER_TX;
+  /* $$$$ for event logging */
+  evt.enqueue_length = nbytes;
+  evt.event_id = e->event_id;
+  q = utm->vpp_event_queue;
+  unix_shared_memory_queue_add (q, (u8 *) & evt, 0 /* do wait for mutex */ );
+}
+
+void
+handle_event_queue (uri_udp_test_main_t * utm)
+{
+  session_fifo_event_t _e, *e = &_e;;
+
+  while (1)
+    {
+      unix_shared_memory_queue_sub (utm->our_event_queue, (u8 *) e,
+				    0 /* nowait */ );
+      switch (e->event_type)
+	{
+	case FIFO_EVENT_SERVER_RX:
+	  handle_fifo_event_server_rx (utm, e);
+	  break;
+
+	case FIFO_EVENT_SERVER_EXIT:
+	  return;
+
+	default:
+	  clib_warning ("unknown event type %d", e->event_type);
+	  break;
+	}
+      if (PREDICT_FALSE (utm->time_to_stop == 1))
+	break;
+      if (PREDICT_FALSE (utm->time_to_print_stats == 1))
+	{
+	  utm->time_to_print_stats = 0;
+	  fformat (stdout, "%d connections\n", pool_elts (utm->sessions));
+	}
+    }
+}
+
+void
+uri_udp_test (uri_udp_test_main_t * utm)
+{
+  vl_api_bind_uri_t *bmp;
+  vl_api_unbind_uri_t *ump;
+
+  bmp = vl_msg_api_alloc (sizeof (*bmp));
+  memset (bmp, 0, sizeof (*bmp));
+
+  bmp->_vl_msg_id = ntohs (VL_API_BIND_URI);
+  bmp->client_index = utm->my_client_index;
+  bmp->context = ntohl (0xfeedface);
+  bmp->initial_segment_size = 256 << 20;	/* size of initial segment */
+  bmp->options[SESSION_OPTIONS_FLAGS] =
+    SESSION_OPTIONS_FLAGS_USE_FIFO | SESSION_OPTIONS_FLAGS_ADD_SEGMENT;
+  bmp->options[SESSION_OPTIONS_RX_FIFO_SIZE] = 16 << 10;
+  bmp->options[SESSION_OPTIONS_TX_FIFO_SIZE] = 16 << 10;
+  bmp->options[SESSION_OPTIONS_ADD_SEGMENT_SIZE] = 128 << 20;
+  memcpy (bmp->uri, utm->uri, vec_len (utm->uri));
+  vl_msg_api_send_shmem (utm->vl_input_queue, (u8 *) & bmp);
+
+  if (wait_for_state_change (utm, STATE_READY))
+    {
+      clib_warning ("timeout waiting for STATE_READY");
+      return;
+    }
+
+  handle_event_queue (utm);
+
+  ump = vl_msg_api_alloc (sizeof (*ump));
+  memset (ump, 0, sizeof (*ump));
+
+  ump->_vl_msg_id = ntohs (VL_API_UNBIND_URI);
+  ump->client_index = utm->my_client_index;
+  memcpy (ump->uri, utm->uri, vec_len (utm->uri));
+  vl_msg_api_send_shmem (utm->vl_input_queue, (u8 *) & ump);
+
+  if (wait_for_state_change (utm, STATE_START))
+    {
+      clib_warning ("timeout waiting for STATE_START");
+      return;
+    }
+
+  fformat (stdout, "Test complete...\n");
+}
+
+int
+main (int argc, char **argv)
+{
+  uri_udp_test_main_t *utm = &uri_udp_test_main;
+  unformat_input_t _argv, *a = &_argv;
+  u8 *chroot_prefix;
+  u8 *heap;
+  u8 *bind_name = (u8 *) "udp://0.0.0.0/1234";
+  u32 tmp;
+  mheap_t *h;
+  session_t *session;
+  int i;
+  int i_am_master = 1;
+
+  clib_mem_init (0, 256 << 20);
+
+  heap = clib_mem_get_per_cpu_heap ();
+  h = mheap_header (heap);
+
+  /* make the main heap thread-safe */
+  h->flags |= MHEAP_FLAG_THREAD_SAFE;
+
+  vec_validate (utm->rx_buf, 8192);
+
+  utm->session_index_by_vpp_handles = hash_create (0, sizeof (uword));
+
+  utm->my_pid = getpid ();
+  utm->configured_segment_size = 1 << 20;
+
+  clib_time_init (&utm->clib_time);
+  init_error_string_table (utm);
+  svm_fifo_segment_init (0x200000000ULL, 20);
+  unformat_init_command_line (a, argv);
+
+  while (unformat_check_input (a) != UNFORMAT_END_OF_INPUT)
+    {
+      if (unformat (a, "chroot prefix %s", &chroot_prefix))
+	{
+	  vl_set_memory_root_path ((char *) chroot_prefix);
+	}
+      else if (unformat (a, "uri %s", &bind_name))
+	;
+      else if (unformat (a, "segment-size %dM", &tmp))
+	utm->configured_segment_size = tmp << 20;
+      else if (unformat (a, "segment-size %dG", &tmp))
+	utm->configured_segment_size = tmp << 30;
+      else if (unformat (a, "master"))
+	i_am_master = 1;
+      else if (unformat (a, "slave"))
+	i_am_master = 0;
+      else
+	{
+	  fformat (stderr, "%s: usage [master|slave]\n");
+	  exit (1);
+	}
+    }
+
+  utm->cut_through_session_index = ~0;
+  utm->uri = format (0, "%s%c", bind_name, 0);
+  utm->i_am_master = i_am_master;
+  utm->segment_main = &svm_fifo_segment_main;
+
+  utm->connect_uri = format (0, "udp://10.0.0.1/1234%c", 0);
+
+  setup_signal_handlers ();
+
+  uri_api_hookup (utm);
+
+  if (connect_to_vpp (i_am_master ? "uri_udp_master" : "uri_udp_slave") < 0)
+    {
+      svm_region_exit ();
+      fformat (stderr, "Couldn't connect to vpe, exiting...\n");
+      exit (1);
+    }
+
+  if (i_am_master == 0)
+    {
+      uri_udp_slave_test (utm);
+      exit (0);
+    }
+
+  /* $$$$ hack preallocation */
+  for (i = 0; i < 200000; i++)
+    {
+      pool_get (utm->sessions, session);
+      memset (session, 0, sizeof (*session));
+    }
+  for (i = 0; i < 200000; i++)
+    pool_put_index (utm->sessions, i);
+
+  uri_udp_test (utm);
+
+  vl_client_disconnect_from_vlib ();
+  exit (0);
+}
+
+#undef vl_api_version
+#define vl_api_version(n,v) static u32 vpe_api_version = v;
+#include 
+#undef vl_api_version
+
+void
+vl_client_add_api_signatures (vl_api_memclnt_create_t * mp)
+{
+  /*
+   * Send the main API signature in slot 0. This bit of code must
+   * match the checks in ../vpe/api/api.c: vl_msg_api_version_check().
+   */
+  mp->api_versions[0] = clib_host_to_net_u32 (vpe_api_version);
+}
+
+u32
+vl (void *p)
+{
+  return vec_len (p);
+}
+
+/*
+ * fd.io coding-style-patch-verification: ON
+ *
+ * Local Variables:
+ * eval: (c-set-style "gnu")
+ * End:
+ */
diff --git a/src/uri/uritest.c b/src/uri/uritest.c
new file mode 100644
index 00000000..edcdb3ad
--- /dev/null
+++ b/src/uri/uritest.c
@@ -0,0 +1,484 @@
+/*
+ * Copyright (c) 2016 Cisco and/or its affiliates.
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at:
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#include 
+#include 
+#include 
+#include 
+#include 
+#include 
+#include 
+#include 
+#include 
+#include 
+#include 
+#include 
+#include 
+#include 
+
+#define vl_typedefs		/* define message structures */
+#include 
+#undef vl_typedefs
+
+/* declare message handlers for each api */
+
+#define vl_endianfun		/* define message structures */
+#include 
+#undef vl_endianfun
+
+/* instantiate all the print functions we know about */
+#define vl_print(handle, ...)
+#define vl_printfun
+#include 
+#undef vl_printfun
+
+typedef enum
+{
+  STATE_START,
+  STATE_READY,
+  STATE_DISCONNECTING,
+} connection_state_t;
+
+typedef struct
+{
+  /* vpe input queue */
+  unix_shared_memory_queue_t *vl_input_queue;
+
+  /* API client handle */
+  u32 my_client_index;
+
+  /* role */
+  int i_am_master;
+
+  /* The URI we're playing with */
+  u8 *uri;
+
+  /* fifo segment */
+  svm_fifo_segment_private_t *seg;
+
+  svm_fifo_t *rx_fifo;
+  svm_fifo_t *tx_fifo;
+
+  /* For deadman timers */
+  clib_time_t clib_time;
+
+  /* State of the connection, shared between msg RX thread and main thread */
+  volatile connection_state_t state;
+
+  /* VNET_API_ERROR_FOO -> "Foo" hash table */
+  uword *error_string_by_error_number;
+} uritest_main_t;
+
+#if CLIB_DEBUG > 0
+#define NITER 1000
+#else
+#define NITER 1000000
+#endif
+
+uritest_main_t uritest_main;
+
+u8 *
+format_api_error (u8 * s, va_list * args)
+{
+  uritest_main_t *utm = va_arg (*args, uritest_main_t *);
+  i32 error = va_arg (*args, u32);
+  uword *p;
+
+  p = hash_get (utm->error_string_by_error_number, -error);
+
+  if (p)
+    s = format (s, "%s", p[0]);
+  else
+    s = format (s, "%d", error);
+  return s;
+}
+
+int
+wait_for_state_change (uritest_main_t * utm, connection_state_t state)
+{
+  f64 timeout = clib_time_now (&utm->clib_time) + 1.0;
+
+  while (clib_time_now (&utm->clib_time) < timeout)
+    {
+      if (utm->state == state)
+	return 0;
+    }
+  return -1;
+}
+
+static void
+vl_api_bind_uri_reply_t_handler (vl_api_bind_uri_reply_t * mp)
+{
+  uritest_main_t *utm = &uritest_main;
+  svm_fifo_segment_create_args_t _a, *a = &_a;
+  int rv;
+
+  ASSERT (utm->i_am_master);
+
+  if (mp->segment_name_length == 0)
+    {
+      clib_warning ("segment_name_length zero");
+      return;
+    }
+
+  a->segment_name = (char *) mp->segment_name;
+  a->segment_size = mp->segment_size;
+
+  /* Create the segment */
+  rv = svm_fifo_segment_create (a);
+  if (rv)
+    {
+      clib_warning ("sm_fifo_segment_create ('%s') failed", mp->segment_name);
+      return;
+    }
+
+  vec_validate (utm->seg, 0);
+
+  memcpy (utm->seg, a->rv, sizeof (*utm->seg));
+
+  /*
+   * By construction the master's idea of the rx fifo ends up in
+   * fsh->fifos[0], and the master's idea of the tx fifo ends up in
+   * fsh->fifos[1].
+   */
+  utm->rx_fifo = svm_fifo_segment_alloc_fifo (utm->seg, 10240);
+  ASSERT (utm->rx_fifo);
+
+  utm->tx_fifo = svm_fifo_segment_alloc_fifo (utm->seg, 10240);
+  ASSERT (utm->tx_fifo);
+
+  utm->state = STATE_READY;
+}
+
+static void
+vl_api_connect_uri_reply_t_handler (vl_api_connect_uri_reply_t * mp)
+{
+  uritest_main_t *utm = &uritest_main;
+  svm_fifo_segment_create_args_t _a, *a = &_a;
+  ssvm_shared_header_t *sh;
+  svm_fifo_segment_header_t *fsh;
+  int rv;
+
+  ASSERT (utm->i_am_master == 0);
+
+  if (mp->segment_name_length == 0)
+    {
+      clib_warning ("segment_name_length zero");
+      return;
+    }
+
+  memset (a, 0, sizeof (*a));
+
+  a->segment_name = (char *) mp->segment_name;
+
+  rv = svm_fifo_segment_attach (a);
+  if (rv)
+    {
+      clib_warning ("sm_fifo_segment_create ('%s') failed", mp->segment_name);
+      return;
+    }
+
+  vec_validate (utm->seg, 0);
+
+  memcpy (utm->seg, a->rv, sizeof (*utm->seg));
+  sh = utm->seg->ssvm.sh;
+  fsh = (svm_fifo_segment_header_t *) sh->opaque[0];
+
+  while (vec_len (fsh->fifos) < 2)
+    sleep (1);
+
+  utm->rx_fifo = (svm_fifo_t *) fsh->fifos[1];
+  ASSERT (utm->rx_fifo);
+  utm->tx_fifo = (svm_fifo_t *) fsh->fifos[0];
+  ASSERT (utm->tx_fifo);
+
+  /* security: could unlink /dev/shm/segment_name> here, maybe */
+
+  utm->state = STATE_READY;
+}
+
+static void
+vl_api_unbind_uri_reply_t_handler (vl_api_unbind_uri_reply_t * mp)
+{
+  uritest_main_t *utm = &uritest_main;
+
+  if (mp->retval != 0)
+    clib_warning ("returned %d", ntohl (mp->retval));
+
+  utm->state = STATE_START;
+}
+
+#define foreach_uri_msg                         \
+_(BIND_URI_REPLY, bind_uri_reply)               \
+_(CONNECT_URI_REPLY, connect_uri_reply)         \
+_(UNBIND_URI_REPLY, unbind_uri_reply)
+
+void
+uri_api_hookup (uritest_main_t * utm)
+{
+#define _(N,n)                                                  \
+    vl_msg_api_set_handlers(VL_API_##N, #n,                     \
+                           vl_api_##n##_t_handler,	        \
+                           vl_noop_handler,                     \
+                           vl_api_##n##_t_endian,               \
+                           vl_api_##n##_t_print,                \
+                           sizeof(vl_api_##n##_t), 1);
+  foreach_uri_msg;
+#undef _
+
+}
+
+
+int
+connect_to_vpp (char *name)
+{
+  uritest_main_t *utm = &uritest_main;
+  api_main_t *am = &api_main;
+
+  if (vl_client_connect_to_vlib ("/vpe-api", name, 32) < 0)
+    return -1;
+
+  utm->vl_input_queue = am->shmem_hdr->vl_input_queue;
+  utm->my_client_index = am->my_client_index;
+
+  return 0;
+}
+
+void
+vlib_cli_output (struct vlib_main_t *vm, char *fmt, ...)
+{
+  clib_warning ("BUG");
+}
+
+static void
+init_error_string_table (uritest_main_t * utm)
+{
+  utm->error_string_by_error_number = hash_create (0, sizeof (uword));
+
+#define _(n,v,s) hash_set (utm->error_string_by_error_number, -v, s);
+  foreach_vnet_api_error;
+#undef _
+
+  hash_set (utm->error_string_by_error_number, 99, "Misc");
+}
+
+void
+uritest_master (uritest_main_t * utm)
+{
+  vl_api_bind_uri_t *bmp;
+  vl_api_unbind_uri_t *ump;
+  int i;
+  u8 *test_data = 0;
+  u8 *reply = 0;
+  u32 reply_len;
+  int mypid = getpid ();
+
+  for (i = 0; i < 2048; i++)
+    vec_add1 (test_data, 'a' + (i % 32));
+
+  bmp = vl_msg_api_alloc (sizeof (*bmp));
+  memset (bmp, 0, sizeof (*bmp));
+
+  bmp->_vl_msg_id = ntohs (VL_API_BIND_URI);
+  bmp->client_index = utm->my_client_index;
+  bmp->context = ntohl (0xfeedface);
+  bmp->segment_size = 256 << 10;
+  memcpy (bmp->uri, utm->uri, vec_len (utm->uri));
+  vl_msg_api_send_shmem (utm->vl_input_queue, (u8 *) & bmp);
+
+  if (wait_for_state_change (utm, STATE_READY))
+    {
+      clib_warning ("timeout waiting for STATE_READY");
+      return;
+    }
+
+  for (i = 0; i < NITER; i++)
+    svm_fifo_enqueue (utm->tx_fifo, mypid, vec_len (test_data), test_data);
+
+  vec_validate (reply, 0);
+
+  reply_len = svm_fifo_dequeue (utm->rx_fifo, mypid, vec_len (reply), reply);
+
+  if (reply_len != 1)
+    clib_warning ("reply length %d", reply_len);
+
+  if (reply[0] == 1)
+    fformat (stdout, "Test OK...");
+
+  ump = vl_msg_api_alloc (sizeof (*ump));
+  memset (ump, 0, sizeof (*ump));
+
+  ump->_vl_msg_id = ntohs (VL_API_UNBIND_URI);
+  ump->client_index = utm->my_client_index;
+  memcpy (ump->uri, utm->uri, vec_len (utm->uri));
+  vl_msg_api_send_shmem (utm->vl_input_queue, (u8 *) & ump);
+
+  if (wait_for_state_change (utm, STATE_START))
+    {
+      clib_warning ("timeout waiting for STATE_READY");
+      return;
+    }
+
+  fformat (stdout, "Master done...\n");
+}
+
+void
+uritest_slave (uritest_main_t * utm)
+{
+  vl_api_connect_uri_t *cmp;
+  int i, j;
+  u8 *test_data = 0;
+  u8 *reply = 0;
+  u32 bytes_received = 0;
+  u32 actual_bytes;
+  int mypid = getpid ();
+  u8 ok;
+  f64 before, after, delta, bytes_per_second;
+
+  vec_validate (test_data, 4095);
+
+  cmp = vl_msg_api_alloc (sizeof (*cmp));
+  memset (cmp, 0, sizeof (*cmp));
+
+  cmp->_vl_msg_id = ntohs (VL_API_CONNECT_URI);
+  cmp->client_index = utm->my_client_index;
+  cmp->context = ntohl (0xfeedface);
+  memcpy (cmp->uri, utm->uri, vec_len (utm->uri));
+  vl_msg_api_send_shmem (utm->vl_input_queue, (u8 *) & cmp);
+
+  if (wait_for_state_change (utm, STATE_READY))
+    {
+      clib_warning ("timeout waiting for STATE_READY");
+      return;
+    }
+
+  ok = 1;
+  before = clib_time_now (&utm->clib_time);
+  for (i = 0; i < NITER; i++)
+    {
+      actual_bytes = svm_fifo_dequeue (utm->rx_fifo, mypid,
+				       vec_len (test_data), test_data);
+      j = 0;
+      while (j < actual_bytes)
+	{
+	  if (test_data[j] != ('a' + (bytes_received % 32)))
+	    ok = 0;
+	  bytes_received++;
+	  j++;
+	}
+      if (bytes_received == NITER * 2048)
+	break;
+    }
+
+  vec_add1 (reply, ok);
+
+  svm_fifo_enqueue (utm->tx_fifo, mypid, vec_len (reply), reply);
+  after = clib_time_now (&utm->clib_time);
+  delta = after - before;
+  bytes_per_second = 0.0;
+
+  if (delta > 0.0)
+    bytes_per_second = (f64) bytes_received / delta;
+
+  fformat (stdout,
+	   "Slave done, %d bytes in %.2f seconds, %.2f bytes/sec...\n",
+	   bytes_received, delta, bytes_per_second);
+}
+
+int
+main (int argc, char **argv)
+{
+  uritest_main_t *utm = &uritest_main;
+  unformat_input_t _argv, *a = &_argv;
+  u8 *chroot_prefix;
+  u8 *heap;
+  char *bind_name = "fifo:uritest";
+  mheap_t *h;
+  int i_am_master = 0;
+
+  clib_mem_init (0, 128 << 20);
+
+  heap = clib_mem_get_per_cpu_heap ();
+  h = mheap_header (heap);
+
+  /* make the main heap thread-safe */
+  h->flags |= MHEAP_FLAG_THREAD_SAFE;
+
+  clib_time_init (&utm->clib_time);
+  init_error_string_table (utm);
+  svm_fifo_segment_init (0x200000000ULL, 20);
+  unformat_init_command_line (a, argv);
+
+  utm->uri = format (0, "%s%c", bind_name, 0);
+
+  while (unformat_check_input (a) != UNFORMAT_END_OF_INPUT)
+    {
+      if (unformat (a, "master"))
+	i_am_master = 1;
+      else if (unformat (a, "slave"))
+	i_am_master = 0;
+      else if (unformat (a, "chroot prefix %s", &chroot_prefix))
+	{
+	  vl_set_memory_root_path ((char *) chroot_prefix);
+	}
+      else
+	{
+	  fformat (stderr, "%s: usage [master|slave]\n");
+	  exit (1);
+	}
+    }
+
+  uri_api_hookup (utm);
+
+  if (connect_to_vpp (i_am_master ? "uritest_master" : "uritest_slave") < 0)
+    {
+      svm_region_exit ();
+      fformat (stderr, "Couldn't connect to vpe, exiting...\n");
+      exit (1);
+    }
+
+  utm->i_am_master = i_am_master;
+
+  if (i_am_master)
+    uritest_master (utm);
+  else
+    uritest_slave (utm);
+
+  vl_client_disconnect_from_vlib ();
+  exit (0);
+}
+
+#undef vl_api_version
+#define vl_api_version(n,v) static u32 vpe_api_version = v;
+#include 
+#undef vl_api_version
+
+void
+vl_client_add_api_signatures (vl_api_memclnt_create_t * mp)
+{
+  /*
+   * Send the main API signature in slot 0. This bit of code must
+   * match the checks in ../vpe/api/api.c: vl_msg_api_version_check().
+   */
+  mp->api_versions[0] = clib_host_to_net_u32 (vpe_api_version);
+}
+
+/*
+ * fd.io coding-style-patch-verification: ON
+ *
+ * Local Variables:
+ * eval: (c-set-style "gnu")
+ * End:
+ */
diff --git a/src/vlib/buffer.c b/src/vlib/buffer.c
index 4f5eb09d..9f26bec7 100644
--- a/src/vlib/buffer.c
+++ b/src/vlib/buffer.c
@@ -360,7 +360,7 @@ vlib_buffer_create_free_list_helper (vlib_main_t * vm,
   memset (f, 0, sizeof (f[0]));
   f->index = f - bm->buffer_free_list_pool;
   f->n_data_bytes = vlib_buffer_round_size (n_data_bytes);
-  f->min_n_buffers_each_physmem_alloc = 16;
+  f->min_n_buffers_each_physmem_alloc = VLIB_FRAME_SIZE;
   f->name = clib_mem_is_heap_object (name) ? name : format (0, "%s", name);
 
   /* Setup free buffer template. */
diff --git a/src/vlib/buffer.h b/src/vlib/buffer.h
index 1f723f3b..69c8c7cc 100644
--- a/src/vlib/buffer.h
+++ b/src/vlib/buffer.h
@@ -240,6 +240,74 @@ vlib_get_buffer_opaque2 (vlib_buffer_t * b)
   return (void *) b->opaque2;
 }
 
+/** \brief Get pointer to the end of buffer's data
+ * @param b     pointer to the buffer
+ * @return      pointer to tail of packet's data
+ */
+always_inline u8 *
+vlib_buffer_get_tail (vlib_buffer_t * b)
+{
+  return b->data + b->current_data + b->current_length;
+}
+
+/** \brief Append uninitialized data to buffer
+ * @param b     pointer to the buffer
+ * @param size  number of uninitialized bytes
+ * @return      pointer to beginning of uninitialized data
+ */
+always_inline void *
+vlib_buffer_put_uninit (vlib_buffer_t * b, u8 size)
+{
+  void *p = vlib_buffer_get_tail (b);
+  /* XXX make sure there's enough space */
+  b->current_length += size;
+  return p;
+}
+
+/** \brief Prepend uninitialized data to buffer
+ * @param b     pointer to the buffer
+ * @param size  number of uninitialized bytes
+ * @return      pointer to beginning of uninitialized data
+ */
+always_inline void *
+vlib_buffer_push_uninit (vlib_buffer_t * b, u8 size)
+{
+  ASSERT (b->current_data + VLIB_BUFFER_PRE_DATA_SIZE >= size);
+  b->current_data -= size;
+  b->current_length += size;
+
+  return vlib_buffer_get_current (b);
+}
+
+/** \brief Make head room, typically for packet headers
+ * @param b     pointer to the buffer
+ * @param size  number of head room bytes
+ * @return      pointer to start of buffer (current data)
+ */
+always_inline void *
+vlib_buffer_make_headroom (vlib_buffer_t * b, u8 size)
+{
+  ASSERT (b->current_data + VLIB_BUFFER_PRE_DATA_SIZE >= size);
+  b->current_data += size;
+  return vlib_buffer_get_current (b);
+}
+
+/** \brief Retrieve bytes from buffer head
+ * @param b     pointer to the buffer
+ * @param size  number of bytes to pull
+ * @return      pointer to start of buffer (current data)
+ */
+always_inline void *
+vlib_buffer_pull (vlib_buffer_t * b, u8 size)
+{
+  if (b->current_length + VLIB_BUFFER_PRE_DATA_SIZE < size)
+    return 0;
+
+  void *data = vlib_buffer_get_current (b);
+  vlib_buffer_advance (b, size);
+  return data;
+}
+
 /* Forward declaration. */
 struct vlib_main_t;
 
diff --git a/src/vlibmemory/unix_shared_memory_queue.c b/src/vlibmemory/unix_shared_memory_queue.c
index 25d28910..e86edec3 100644
--- a/src/vlibmemory/unix_shared_memory_queue.c
+++ b/src/vlibmemory/unix_shared_memory_queue.c
@@ -33,18 +33,13 @@
  * nels = number of elements on the queue
  * elsize = element size, presumably 4 and cacheline-size will
  *          be popular choices.
- * coid  = consumer coid, from ChannelCreate
  * pid   = consumer pid
- * pulse_code  = pulse code consumer expects
- * pulse_value = pulse value consumer expects
- * consumer_prio = consumer's priority, so pulses won't change
- *                 the consumer's priority.
  *
  * The idea is to call this function in the queue consumer,
  * and e-mail the queue pointer to the producer(s).
  *
- * The spp process / main thread allocates one of these
- * at startup; its main input queue. The spp main input queue
+ * The vpp process / main thread allocates one of these
+ * at startup; its main input queue. The vpp main input queue
  * has a pointer to it in the shared memory segment header.
  *
  * You probably want to be on an svm data heap before calling this
@@ -70,7 +65,7 @@ unix_shared_memory_queue_init (int nels,
   q->signal_when_queue_non_empty = signal_when_queue_non_empty;
 
   memset (&attr, 0, sizeof (attr));
-  memset (&cattr, 0, sizeof (attr));
+  memset (&cattr, 0, sizeof (cattr));
 
   if (pthread_mutexattr_init (&attr))
     clib_unix_warning ("mutexattr_init");
@@ -277,6 +272,7 @@ unix_shared_memory_queue_sub (unix_shared_memory_queue_t * q,
   clib_memcpy (elem, headp, q->elsize);
 
   q->head++;
+  /* $$$$ JFC shouldn't this be == 0? */
   if (q->cursize == q->maxsize)
     need_broadcast = 1;
 
diff --git a/src/vlibmemory/unix_shared_memory_queue.h b/src/vlibmemory/unix_shared_memory_queue.h
index f758f17c..13800065 100644
--- a/src/vlibmemory/unix_shared_memory_queue.h
+++ b/src/vlibmemory/unix_shared_memory_queue.h
@@ -29,7 +29,7 @@ typedef struct _unix_shared_memory_queue
   pthread_cond_t condvar;	/* 8 bytes */
   int head;
   int tail;
-  int cursize;
+  volatile int cursize;
   int maxsize;
   int elsize;
   int consumer_pid;
diff --git a/src/vnet.am b/src/vnet.am
index 64484e18..923f61d8 100644
--- a/src/vnet.am
+++ b/src/vnet.am
@@ -324,11 +324,7 @@ libvnet_la_SOURCES +=				\
  vnet/ip/ip_input_acl.c				\
  vnet/ip/lookup.c				\
  vnet/ip/ping.c					\
- vnet/ip/punt.c					\
- vnet/ip/udp_format.c				\
- vnet/ip/udp_init.c				\
- vnet/ip/udp_local.c				\
- vnet/ip/udp_pg.c
+ vnet/ip/punt.c
 
 nobase_include_HEADERS +=			\
  vnet/ip/format.h				\
@@ -354,11 +350,7 @@ nobase_include_HEADERS +=			\
  vnet/ip/ports.def				\
  vnet/ip/protocols.def				\
  vnet/ip/punt_error.def				\
- vnet/ip/punt.h					\
- vnet/ip/tcp_packet.h				\
- vnet/ip/udp_error.def				\
- vnet/ip/udp.h					\
- vnet/ip/udp_packet.h
+ vnet/ip/punt.h
 
 API_FILES += vnet/ip/ip.api
 
@@ -473,6 +465,38 @@ test_map_LDADD = libvnet.la libvppinfra.la libvlib.la	\
 test_map_LDFLAGS = -static
 endif
 
+########################################
+# Layer 4 protocol: tcp
+########################################
+libvnet_la_SOURCES +=				\
+ vnet/tcp/tcp_format.c				\
+ vnet/tcp/tcp_pg.c				\
+ vnet/tcp/tcp_syn_filter4.c			\
+ vnet/tcp/tcp_output.c				\
+ vnet/tcp/tcp_input.c				\
+ vnet/tcp/tcp_newreno.c				\
+ vnet/tcp/tcp.c
+
+nobase_include_HEADERS +=			\
+ vnet/tcp/tcp_packet.h				\
+ vnet/tcp/tcp_timer.h				\
+ vnet/tcp/tcp.h
+
+########################################
+# Layer 4 protocol: udp
+########################################
+libvnet_la_SOURCES +=				\
+ vnet/udp/udp.c					\
+ vnet/udp/udp_input.c				\
+ vnet/udp/builtin_server.c			\
+ vnet/udp/udp_format.c				\
+ vnet/udp/udp_local.c				\
+ vnet/udp/udp_pg.c
+
+nobase_include_HEADERS +=			\
+  vnet/udp/udp_error.def                       	\
+  vnet/udp/udp.h                               	\
+  vnet/udp/udp_packet.h
 
 ########################################
 # Tunnel protocol: gre
@@ -833,6 +857,28 @@ libvnet_la_SOURCES +=				\
 nobase_include_HEADERS +=			\
   vnet/devices/ssvm/ssvm_eth.h
 
+########################################
+# session managmeent
+########################################
+
+libvnet_la_SOURCES +=				\
+  vnet/session/session.c			\
+  vnet/session/node.c				\
+  vnet/session/transport.c			\
+  vnet/session/application.c			\
+  vnet/session/session_cli.c			\
+  vnet/session/hashes.c				\
+  vnet/session/application_interface.c		\
+  vnet/session/session_api.c
+
+nobase_include_HEADERS +=			\
+  vnet/session/session.h			\
+  vnet/session/application.h			\
+  vnet/session/transport.h			\
+  vnet/session/application_interface.h
+
+API_FILES += vnet/session/session.api
+
 ########################################
 # Linux packet interface
 ########################################
diff --git a/src/vnet/api_errno.h b/src/vnet/api_errno.h
index 8680ef7c..861a5767 100644
--- a/src/vnet/api_errno.h
+++ b/src/vnet/api_errno.h
@@ -91,14 +91,19 @@ _(INVALID_ADDRESS_FAMILY, -97, "Invalid address family")                \
 _(INVALID_SUB_SW_IF_INDEX, -98, "Invalid sub-interface sw_if_index")    \
 _(TABLE_TOO_BIG, -99, "Table too big")                                  \
 _(CANNOT_ENABLE_DISABLE_FEATURE, -100, "Cannot enable/disable feature") \
-_(BFD_EEXIST, -101, "Duplicate BFD object") \
-_(BFD_ENOENT, -102, "No such BFD object") \
-_(BFD_EINUSE, -103, "BFD object in use") \
-_(BFD_NOTSUPP, -104, "BFD feature not supported") \
-_(LISP_RLOC_LOCAL, -105, "RLOC address is local") \
-_(BFD_EAGAIN, -106, "BFD object cannot be manipulated at this time")    \
-_(INVALID_GPE_MODE, -107, "Invalid GPE mode")                           \
-_(LISP_GPE_ENTRIES_PRESENT, -108, "LISP GPE entries are present")
+_(BFD_EEXIST, -101, "Duplicate BFD object")                             \
+_(BFD_ENOENT, -102, "No such BFD object")                               \
+_(BFD_EINUSE, -103, "BFD object in use")                                \
+_(BFD_NOTSUPP, -104, "BFD feature not supported")                       \
+_(ADDRESS_IN_USE, -105, "Address in use")				\
+_(ADDRESS_NOT_IN_USE, -106, "Address not in use")			\
+_(QUEUE_FULL, -107, "Queue full")                                       \
+_(UNKNOWN_URI_TYPE, -108, "Unknown URI type")				\
+_(URI_FIFO_CREATE_FAILED, -109, "URI FIFO segment create failed")       \
+_(LISP_RLOC_LOCAL, -110, "RLOC address is local")                       \
+_(BFD_EAGAIN, -111, "BFD object cannot be manipulated at this time")	\
+_(INVALID_GPE_MODE, -112, "Invalid GPE mode")                           \
+_(LISP_GPE_ENTRIES_PRESENT, -113, "LISP GPE entries are present")
 
 typedef enum
 {
diff --git a/src/vnet/bfd/bfd_udp.c b/src/vnet/bfd/bfd_udp.c
index 146faad6..cf05089b 100644
--- a/src/vnet/bfd/bfd_udp.c
+++ b/src/vnet/bfd/bfd_udp.c
@@ -18,12 +18,12 @@
 #include 
 #include 
 #include 
-#include 
+#include 
+#include 
 #include 
 #include 
 #include 
 #include 
-#include 
 #include 
 #include 
 #include 
diff --git a/src/vnet/buffer.h b/src/vnet/buffer.h
index f1cc6371..3de01f2a 100644
--- a/src/vnet/buffer.h
+++ b/src/vnet/buffer.h
@@ -277,6 +277,16 @@ typedef struct
       u16 buffer_advance;
     } device_input_feat;
 
+    /* TCP */
+    struct
+    {
+      u32 connection_index;
+      u32 seq_number;
+      u32 seq_end;
+      u32 ack_number;
+      u8 flags;
+    } tcp;
+
     u32 unused[6];
   };
 } vnet_buffer_opaque_t;
diff --git a/src/vnet/classify/vnet_classify.c b/src/vnet/classify/vnet_classify.c
index 6093e2ac..b651a1f1 100644
--- a/src/vnet/classify/vnet_classify.c
+++ b/src/vnet/classify/vnet_classify.c
@@ -695,8 +695,8 @@ int vnet_classify_add_del_table (vnet_classify_main_t * cm,
 }
 
 #define foreach_tcp_proto_field                 \
-_(src_port)                                     \
-_(dst_port)
+_(src)                                          \
+_(dst)
 
 #define foreach_udp_proto_field                 \
 _(src_port)                                     \
diff --git a/src/vnet/dhcp/dhcp_proxy.h b/src/vnet/dhcp/dhcp_proxy.h
index c0d79c41..4586d883 100644
--- a/src/vnet/dhcp/dhcp_proxy.h
+++ b/src/vnet/dhcp/dhcp_proxy.h
@@ -26,7 +26,7 @@
 #include 
 #include 
 #include 
-#include 
+#include 
 
 typedef enum {
 #define dhcp_proxy_error(n,s) DHCP_PROXY_ERROR_##n,
diff --git a/src/vnet/flow/flow_report.h b/src/vnet/flow/flow_report.h
index 4e764377..e8ed3818 100644
--- a/src/vnet/flow/flow_report.h
+++ b/src/vnet/flow/flow_report.h
@@ -23,7 +23,7 @@
 #include 
 #include 
 #include 
-#include 
+#include 
 #include 
 #include 
 #include 
diff --git a/src/vnet/ip/ip.h b/src/vnet/ip/ip.h
index 02a1a963..70b4ccd8 100644
--- a/src/vnet/ip/ip.h
+++ b/src/vnet/ip/ip.h
@@ -50,8 +50,8 @@
 #include 
 #include 
 
-#include 
-#include 
+#include 
+#include 
 #include 
 
 #include 
diff --git a/src/vnet/ip/ip4.h b/src/vnet/ip/ip4.h
index b184fbae..4e075d0f 100644
--- a/src/vnet/ip/ip4.h
+++ b/src/vnet/ip/ip4.h
@@ -309,8 +309,8 @@ ip4_compute_flow_hash (const ip4_header_t * ip,
   b = (flow_hash_config & IP_FLOW_HASH_REVERSE_SRC_DST) ? t1 : t2;
   b ^= (flow_hash_config & IP_FLOW_HASH_PROTO) ? ip->protocol : 0;
 
-  t1 = is_tcp_udp ? tcp->ports.src : 0;
-  t2 = is_tcp_udp ? tcp->ports.dst : 0;
+  t1 = is_tcp_udp ? tcp->src : 0;
+  t2 = is_tcp_udp ? tcp->dst : 0;
 
   t1 = (flow_hash_config & IP_FLOW_HASH_SRC_PORT) ? t1 : 0;
   t2 = (flow_hash_config & IP_FLOW_HASH_DST_PORT) ? t2 : 0;
@@ -334,6 +334,44 @@ u8 *format_ip4_forward_next_trace (u8 * s, va_list * args);
 
 u32 ip4_tcp_udp_validate_checksum (vlib_main_t * vm, vlib_buffer_t * p0);
 
+#define IP_DF 0x4000		/* don't fragment */
+
+/**
+ * Push IPv4 header to buffer
+ *
+ * This does not support fragmentation.
+ *
+ * @param vm - vlib_main
+ * @param b - buffer to write the header to
+ * @param src - source IP
+ * @param dst - destination IP
+ * @param prot - payload proto
+ *
+ * @return - pointer to start of IP header
+ */
+always_inline void *
+vlib_buffer_push_ip4 (vlib_main_t * vm, vlib_buffer_t * b,
+		      ip4_address_t * src, ip4_address_t * dst, int proto)
+{
+  ip4_header_t *ih;
+
+  /* make some room */
+  ih = vlib_buffer_push_uninit (b, sizeof (ip4_header_t));
+
+  ih->ip_version_and_header_length = 0x45;
+  ih->tos = 0;
+  ih->length = clib_host_to_net_u16 (vlib_buffer_length_in_chain (vm, b));
+
+  /* No fragments */
+  ih->flags_and_fragment_offset = clib_host_to_net_u16 (IP_DF);
+  ih->ttl = 255;
+  ih->protocol = proto;
+  ih->src_address.as_u32 = src->as_u32;
+  ih->dst_address.as_u32 = dst->as_u32;
+
+  ih->checksum = ip4_header_checksum (ih);
+  return ih;
+}
 #endif /* included_ip_ip4_h */
 
 /*
diff --git a/src/vnet/ip/ip4_forward.c b/src/vnet/ip/ip4_forward.c
index 8081b34b..66d91ab6 100644
--- a/src/vnet/ip/ip4_forward.c
+++ b/src/vnet/ip/ip4_forward.c
@@ -1478,8 +1478,18 @@ ip4_tcp_udp_validate_checksum (vlib_main_t * vm, vlib_buffer_t * p0)
   return p0->flags;
 }
 
-static uword
-ip4_local (vlib_main_t * vm, vlib_node_runtime_t * node, vlib_frame_t * frame)
+/* *INDENT-OFF* */
+VNET_FEATURE_ARC_INIT (ip4_local) =
+{
+  .arc_name  = "ip4-local",
+  .start_nodes = VNET_FEATURES ("ip4-local"),
+};
+/* *INDENT-ON* */
+
+static inline uword
+ip4_local_inline (vlib_main_t * vm,
+		  vlib_node_runtime_t * node,
+		  vlib_frame_t * frame, int head_of_feature_arc)
 {
   ip4_main_t *im = &ip4_main;
   ip_lookup_main_t *lm = &im->lookup_main;
@@ -1487,6 +1497,7 @@ ip4_local (vlib_main_t * vm, vlib_node_runtime_t * node, vlib_frame_t * frame)
   u32 *from, *to_next, n_left_from, n_left_to_next;
   vlib_node_runtime_t *error_node =
     vlib_node_get_runtime (vm, ip4_input_node.index);
+  u8 arc_index = vnet_feat_arc_ip4_local.feature_arc_index;
 
   from = vlib_frame_vector_args (frame);
   n_left_from = frame->n_vectors;
@@ -1513,7 +1524,7 @@ ip4_local (vlib_main_t * vm, vlib_node_runtime_t * node, vlib_frame_t * frame)
 	  i32 len_diff0, len_diff1;
 	  u8 error0, is_udp0, is_tcp_udp0, good_tcp_udp0, proto0;
 	  u8 error1, is_udp1, is_tcp_udp1, good_tcp_udp1, proto1;
-	  u8 enqueue_code;
+	  u32 sw_if_index0, sw_if_index1;
 
 	  pi0 = to_next[0] = from[0];
 	  pi1 = to_next[1] = from[1];
@@ -1522,6 +1533,8 @@ ip4_local (vlib_main_t * vm, vlib_node_runtime_t * node, vlib_frame_t * frame)
 	  to_next += 2;
 	  n_left_to_next -= 2;
 
+	  next0 = next1 = IP_LOCAL_NEXT_DROP;
+
 	  p0 = vlib_get_buffer (vm, pi0);
 	  p1 = vlib_get_buffer (vm, pi1);
 
@@ -1531,14 +1544,18 @@ ip4_local (vlib_main_t * vm, vlib_node_runtime_t * node, vlib_frame_t * frame)
 	  vnet_buffer (p0)->ip.start_of_ip_header = p0->current_data;
 	  vnet_buffer (p1)->ip.start_of_ip_header = p1->current_data;
 
-	  fib_index0 = vec_elt (im->fib_index_by_sw_if_index,
-				vnet_buffer (p0)->sw_if_index[VLIB_RX]);
+	  sw_if_index0 = vnet_buffer (p0)->sw_if_index[VLIB_RX];
+	  sw_if_index1 = vnet_buffer (p1)->sw_if_index[VLIB_RX];
+
+	  fib_index0 = vec_elt (im->fib_index_by_sw_if_index, sw_if_index0);
+	  fib_index1 = vec_elt (im->fib_index_by_sw_if_index, sw_if_index1);
+
+	  fib_index0 = vec_elt (im->fib_index_by_sw_if_index, sw_if_index0);
 	  fib_index0 =
 	    (vnet_buffer (p0)->sw_if_index[VLIB_TX] ==
 	     (u32) ~ 0) ? fib_index0 : vnet_buffer (p0)->sw_if_index[VLIB_TX];
 
-	  fib_index1 = vec_elt (im->fib_index_by_sw_if_index,
-				vnet_buffer (p1)->sw_if_index[VLIB_RX]);
+	  fib_index1 = vec_elt (im->fib_index_by_sw_if_index, sw_if_index1);
 	  fib_index1 =
 	    (vnet_buffer (p1)->sw_if_index[VLIB_TX] ==
 	     (u32) ~ 0) ? fib_index1 : vnet_buffer (p1)->sw_if_index[VLIB_TX];
@@ -1557,6 +1574,13 @@ ip4_local (vlib_main_t * vm, vlib_node_runtime_t * node, vlib_frame_t * frame)
 	     until support of IP frag reassembly is implemented */
 	  proto0 = ip4_is_fragment (ip0) ? 0xfe : ip0->protocol;
 	  proto1 = ip4_is_fragment (ip1) ? 0xfe : ip1->protocol;
+
+	  if (head_of_feature_arc == 0)
+	    {
+	      error0 = error1 = IP4_ERROR_UNKNOWN_PROTOCOL;
+	      goto skip_checks;
+	    }
+
 	  is_udp0 = proto0 == IP_PROTOCOL_UDP;
 	  is_udp1 = proto1 == IP_PROTOCOL_UDP;
 	  is_tcp_udp0 = is_udp0 || proto0 == IP_PROTOCOL_TCP;
@@ -1686,6 +1710,7 @@ ip4_local (vlib_main_t * vm, vlib_node_runtime_t * node, vlib_frame_t * frame)
 	  next0 = lm->local_next_by_ip_protocol[proto0];
 	  next1 = lm->local_next_by_ip_protocol[proto1];
 
+	skip_checks:
 	  next0 =
 	    error0 != IP4_ERROR_UNKNOWN_PROTOCOL ? IP_LOCAL_NEXT_DROP : next0;
 	  next1 =
@@ -1694,44 +1719,17 @@ ip4_local (vlib_main_t * vm, vlib_node_runtime_t * node, vlib_frame_t * frame)
 	  p0->error = error0 ? error_node->errors[error0] : 0;
 	  p1->error = error1 ? error_node->errors[error1] : 0;
 
-	  enqueue_code = (next0 != next_index) + 2 * (next1 != next_index);
-
-	  if (PREDICT_FALSE (enqueue_code != 0))
+	  if (head_of_feature_arc)
 	    {
-	      switch (enqueue_code)
-		{
-		case 1:
-		  /* A B A */
-		  to_next[-2] = pi1;
-		  to_next -= 1;
-		  n_left_to_next += 1;
-		  vlib_set_next_frame_buffer (vm, node, next0, pi0);
-		  break;
-
-		case 2:
-		  /* A A B */
-		  to_next -= 1;
-		  n_left_to_next += 1;
-		  vlib_set_next_frame_buffer (vm, node, next1, pi1);
-		  break;
-
-		case 3:
-		  /* A B B or A B C */
-		  to_next -= 2;
-		  n_left_to_next += 2;
-		  vlib_set_next_frame_buffer (vm, node, next0, pi0);
-		  vlib_set_next_frame_buffer (vm, node, next1, pi1);
-		  if (next0 == next1)
-		    {
-		      vlib_put_next_frame (vm, node, next_index,
-					   n_left_to_next);
-		      next_index = next1;
-		      vlib_get_next_frame (vm, node, next_index, to_next,
-					   n_left_to_next);
-		    }
-		  break;
-		}
+	      if (PREDICT_TRUE (error0 == (u8) IP4_ERROR_UNKNOWN_PROTOCOL))
+		vnet_feature_arc_start (arc_index, sw_if_index0, &next0, p0);
+	      if (PREDICT_TRUE (error1 == (u8) IP4_ERROR_UNKNOWN_PROTOCOL))
+		vnet_feature_arc_start (arc_index, sw_if_index1, &next1, p1);
 	    }
+
+	  vlib_validate_buffer_enqueue_x2 (vm, node, next_index, to_next,
+					   n_left_to_next, pi0, pi1,
+					   next0, next1);
 	}
 
       while (n_left_from > 0 && n_left_to_next > 0)
@@ -1746,6 +1744,7 @@ ip4_local (vlib_main_t * vm, vlib_node_runtime_t * node, vlib_frame_t * frame)
 	  u8 error0, is_udp0, is_tcp_udp0, good_tcp_udp0, proto0;
 	  load_balance_t *lb0;
 	  const dpo_id_t *dpo0;
+	  u32 sw_if_index0;
 
 	  pi0 = to_next[0] = from[0];
 	  from += 1;
@@ -1753,14 +1752,18 @@ ip4_local (vlib_main_t * vm, vlib_node_runtime_t * node, vlib_frame_t * frame)
 	  to_next += 1;
 	  n_left_to_next -= 1;
 
+	  next0 = IP_LOCAL_NEXT_DROP;
+
 	  p0 = vlib_get_buffer (vm, pi0);
 
 	  ip0 = vlib_buffer_get_current (p0);
 
 	  vnet_buffer (p0)->ip.start_of_ip_header = p0->current_data;
 
-	  fib_index0 = vec_elt (im->fib_index_by_sw_if_index,
-				vnet_buffer (p0)->sw_if_index[VLIB_RX]);
+	  sw_if_index0 = vnet_buffer (p0)->sw_if_index[VLIB_RX];
+
+	  fib_index0 = vec_elt (im->fib_index_by_sw_if_index, sw_if_index0);
+
 	  fib_index0 =
 	    (vnet_buffer (p0)->sw_if_index[VLIB_TX] ==
 	     (u32) ~ 0) ? fib_index0 : vnet_buffer (p0)->sw_if_index[VLIB_TX];
@@ -1775,6 +1778,13 @@ ip4_local (vlib_main_t * vm, vlib_node_runtime_t * node, vlib_frame_t * frame)
 	  /* Treat IP frag packets as "experimental" protocol for now
 	     until support of IP frag reassembly is implemented */
 	  proto0 = ip4_is_fragment (ip0) ? 0xfe : ip0->protocol;
+
+	  if (head_of_feature_arc == 0)
+	    {
+	      error0 = IP4_ERROR_UNKNOWN_PROTOCOL;
+	      goto skip_check;
+	    }
+
 	  is_udp0 = proto0 == IP_PROTOCOL_UDP;
 	  is_tcp_udp0 = is_udp0 || proto0 == IP_PROTOCOL_TCP;
 
@@ -1847,6 +1857,8 @@ ip4_local (vlib_main_t * vm, vlib_node_runtime_t * node, vlib_frame_t * frame)
 		     ip0->dst_address.as_u32 != 0xFFFFFFFF)
 		    ? IP4_ERROR_SRC_LOOKUP_MISS : error0);
 
+	skip_check:
+
 	  next0 = lm->local_next_by_ip_protocol[proto0];
 
 	  next0 =
@@ -1854,18 +1866,15 @@ ip4_local (vlib_main_t * vm, vlib_node_runtime_t * node, vlib_frame_t * frame)
 
 	  p0->error = error0 ? error_node->errors[error0] : 0;
 
-	  if (PREDICT_FALSE (next0 != next_index))
+	  if (head_of_feature_arc)
 	    {
-	      n_left_to_next += 1;
-	      vlib_put_next_frame (vm, node, next_index, n_left_to_next);
-
-	      next_index = next0;
-	      vlib_get_next_frame (vm, node, next_index, to_next,
-				   n_left_to_next);
-	      to_next[0] = pi0;
-	      to_next += 1;
-	      n_left_to_next -= 1;
+	      if (PREDICT_TRUE (error0 == (u8) IP4_ERROR_UNKNOWN_PROTOCOL))
+		vnet_feature_arc_start (arc_index, sw_if_index0, &next0, p0);
 	    }
+
+	  vlib_validate_buffer_enqueue_x1 (vm, node, next_index, to_next,
+					   n_left_to_next, pi0, next0);
+
 	}
 
       vlib_put_next_frame (vm, node, next_index, n_left_to_next);
@@ -1874,21 +1883,57 @@ ip4_local (vlib_main_t * vm, vlib_node_runtime_t * node, vlib_frame_t * frame)
   return frame->n_vectors;
 }
 
+static uword
+ip4_local (vlib_main_t * vm, vlib_node_runtime_t * node, vlib_frame_t * frame)
+{
+  return ip4_local_inline (vm, node, frame, 1 /* head of feature arc */ );
+}
+
+/* *INDENT-OFF* */
 VLIB_REGISTER_NODE (ip4_local_node) =
 {
-  .function = ip4_local,.name = "ip4-local",.vector_size =
-    sizeof (u32),.format_trace =
-    format_ip4_forward_next_trace,.n_next_nodes =
-    IP_LOCAL_N_NEXT,.next_nodes =
+  .function = ip4_local,
+  .name = "ip4-local",
+  .vector_size = sizeof (u32),
+  .format_trace = format_ip4_forward_next_trace,
+  .n_next_nodes = IP_LOCAL_N_NEXT,
+  .next_nodes =
   {
-  [IP_LOCAL_NEXT_DROP] = "error-drop",
-      [IP_LOCAL_NEXT_PUNT] = "error-punt",
-      [IP_LOCAL_NEXT_UDP_LOOKUP] = "ip4-udp-lookup",
-      [IP_LOCAL_NEXT_ICMP] = "ip4-icmp-input",}
-,};
+    [IP_LOCAL_NEXT_DROP] = "error-drop",
+    [IP_LOCAL_NEXT_PUNT] = "error-punt",
+    [IP_LOCAL_NEXT_UDP_LOOKUP] = "ip4-udp-lookup",
+    [IP_LOCAL_NEXT_ICMP] = "ip4-icmp-input",},
+};
+/* *INDENT-ON* */
 
 VLIB_NODE_FUNCTION_MULTIARCH (ip4_local_node, ip4_local);
 
+static uword
+ip4_local_end_of_arc (vlib_main_t * vm,
+		      vlib_node_runtime_t * node, vlib_frame_t * frame)
+{
+  return ip4_local_inline (vm, node, frame, 0 /* head of feature arc */ );
+}
+
+/* *INDENT-OFF* */
+VLIB_REGISTER_NODE (ip4_local_end_of_arc_node,static) = {
+  .function = ip4_local_end_of_arc,
+  .name = "ip4-local-end-of-arc",
+  .vector_size = sizeof (u32),
+
+  .format_trace = format_ip4_forward_next_trace,
+  .sibling_of = "ip4-local",
+};
+
+VLIB_NODE_FUNCTION_MULTIARCH (ip4_local_end_of_arc_node, ip4_local_end_of_arc)
+
+VNET_FEATURE_INIT (ip4_local_end_of_arc, static) = {
+  .arc_name = "ip4-local",
+  .node_name = "ip4-local-end-of-arc",
+  .runs_before = 0, /* not before any other features */
+};
+/* *INDENT-ON* */
+
 void
 ip4_register_protocol (u32 protocol, u32 node_index)
 {
diff --git a/src/vnet/ip/ip4_packet.h b/src/vnet/ip/ip4_packet.h
index 8da788b4..b2c1fcd4 100644
--- a/src/vnet/ip/ip4_packet.h
+++ b/src/vnet/ip/ip4_packet.h
@@ -41,7 +41,7 @@
 #define included_ip4_packet_h
 
 #include 	/* for ip_csum_t */
-#include 	/* for tcp_header_t */
+#include 	/* for tcp_header_t */
 #include 	/* for clib_net_to_host_u16 */
 
 /* IP4 address which can be accessed either as 4 bytes
@@ -342,10 +342,10 @@ ip4_tcp_reply_x1 (ip4_header_t * ip0, tcp_header_t * tcp0)
   ip0->src_address.data_u32 = dst0;
   ip0->dst_address.data_u32 = src0;
 
-  src0 = tcp0->ports.src;
-  dst0 = tcp0->ports.dst;
-  tcp0->ports.src = dst0;
-  tcp0->ports.dst = src0;
+  src0 = tcp0->src;
+  dst0 = tcp0->dst;
+  tcp0->src = dst0;
+  tcp0->dst = src0;
 }
 
 always_inline void
@@ -363,14 +363,14 @@ ip4_tcp_reply_x2 (ip4_header_t * ip0, ip4_header_t * ip1,
   ip0->dst_address.data_u32 = src0;
   ip1->dst_address.data_u32 = src1;
 
-  src0 = tcp0->ports.src;
-  src1 = tcp1->ports.src;
-  dst0 = tcp0->ports.dst;
-  dst1 = tcp1->ports.dst;
-  tcp0->ports.src = dst0;
-  tcp1->ports.src = dst1;
-  tcp0->ports.dst = src0;
-  tcp1->ports.dst = src1;
+  src0 = tcp0->src;
+  src1 = tcp1->src;
+  dst0 = tcp0->dst;
+  dst1 = tcp1->dst;
+  tcp0->src = dst0;
+  tcp1->src = dst1;
+  tcp0->dst = src0;
+  tcp1->dst = src1;
 }
 
 #endif /* included_ip4_packet_h */
diff --git a/src/vnet/ip/ip6.h b/src/vnet/ip/ip6.h
index 5456f0f2..2615fbfa 100644
--- a/src/vnet/ip/ip6.h
+++ b/src/vnet/ip/ip6.h
@@ -461,8 +461,8 @@ ip6_compute_flow_hash (const ip6_header_t * ip,
   b = (flow_hash_config & IP_FLOW_HASH_REVERSE_SRC_DST) ? t1 : t2;
   b ^= (flow_hash_config & IP_FLOW_HASH_PROTO) ? ip->protocol : 0;
 
-  t1 = is_tcp_udp ? tcp->ports.src : 0;
-  t2 = is_tcp_udp ? tcp->ports.dst : 0;
+  t1 = is_tcp_udp ? tcp->src : 0;
+  t2 = is_tcp_udp ? tcp->dst : 0;
 
   t1 = (flow_hash_config & IP_FLOW_HASH_SRC_PORT) ? t1 : 0;
   t2 = (flow_hash_config & IP_FLOW_HASH_DST_PORT) ? t2 : 0;
@@ -497,6 +497,46 @@ int ip6_hbh_register_option (u8 option,
 int ip6_hbh_unregister_option (u8 option);
 void ip6_hbh_set_next_override (uword next);
 
+/**
+ * Push IPv6 header to buffer
+ *
+ * @param vm - vlib_main
+ * @param b - buffer to write the header to
+ * @param src - source IP
+ * @param dst - destination IP
+ * @param prot - payload proto
+ *
+ * @return - pointer to start of IP header
+ */
+always_inline void *
+vlib_buffer_push_ip6 (vlib_main_t * vm, vlib_buffer_t * b,
+		      ip6_address_t * src, ip6_address_t * dst, int proto)
+{
+  ip6_header_t *ip6h;
+  u16 payload_length;
+
+  /* make some room */
+  ip6h = vlib_buffer_push_uninit (b, sizeof (ip6_header_t));
+
+  ip6h->ip_version_traffic_class_and_flow_label =
+    clib_host_to_net_u32 (0x6 << 28);
+
+  /* calculate ip6 payload length */
+  payload_length = vlib_buffer_length_in_chain (vm, b);
+  payload_length -= sizeof (*ip6h);
+
+  ip6h->payload_length = clib_host_to_net_u16 (payload_length);
+
+  ip6h->hop_limit = 0xff;
+  ip6h->protocol = proto;
+  clib_memcpy (ip6h->src_address.as_u8, src->as_u8,
+	       sizeof (ip6h->src_address));
+  clib_memcpy (ip6h->dst_address.as_u8, dst->as_u8,
+	       sizeof (ip6h->src_address));
+
+  return ip6h;
+}
+
 #endif /* included_ip_ip6_h */
 
 /*
diff --git a/src/vnet/ip/ip6_packet.h b/src/vnet/ip/ip6_packet.h
index 1e551c8b..4fd14b96 100644
--- a/src/vnet/ip/ip6_packet.h
+++ b/src/vnet/ip/ip6_packet.h
@@ -40,7 +40,7 @@
 #ifndef included_ip6_packet_h
 #define included_ip6_packet_h
 
-#include 
+#include 
 #include 
 
 typedef union
@@ -373,10 +373,10 @@ ip6_tcp_reply_x1 (ip6_header_t * ip0, tcp_header_t * tcp0)
   {
     u16 src0, dst0;
 
-    src0 = tcp0->ports.src;
-    dst0 = tcp0->ports.dst;
-    tcp0->ports.src = dst0;
-    tcp0->ports.dst = src0;
+    src0 = tcp0->src;
+    dst0 = tcp0->dst;
+    tcp0->src = dst0;
+    tcp0->dst = src0;
   }
 }
 
@@ -400,14 +400,14 @@ ip6_tcp_reply_x2 (ip6_header_t * ip0, ip6_header_t * ip1,
   {
     u16 src0, dst0, src1, dst1;
 
-    src0 = tcp0->ports.src;
-    src1 = tcp1->ports.src;
-    dst0 = tcp0->ports.dst;
-    dst1 = tcp1->ports.dst;
-    tcp0->ports.src = dst0;
-    tcp1->ports.src = dst1;
-    tcp0->ports.dst = src0;
-    tcp1->ports.dst = src1;
+    src0 = tcp0->src;
+    src1 = tcp1->src;
+    dst0 = tcp0->dst;
+    dst1 = tcp1->dst;
+    tcp0->src = dst0;
+    tcp1->src = dst1;
+    tcp0->dst = src0;
+    tcp1->dst = src1;
   }
 }
 
diff --git a/src/vnet/ip/punt.c b/src/vnet/ip/punt.c
index 9c735128..48558401 100644
--- a/src/vnet/ip/punt.c
+++ b/src/vnet/ip/punt.c
@@ -23,7 +23,7 @@
  */
 #include 
 #include 
-#include 
+#include 
 #include 
 
 #define foreach_punt_next \
diff --git a/src/vnet/ip/tcp_packet.h b/src/vnet/ip/tcp_packet.h
deleted file mode 100644
index 93f73e01..00000000
--- a/src/vnet/ip/tcp_packet.h
+++ /dev/null
@@ -1,141 +0,0 @@
-/*
- * Copyright (c) 2015 Cisco and/or its affiliates.
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at:
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-/*
- * ip4/tcp_packet.h: TCP packet format (see RFC 793)
- *
- * Copyright (c) 2008 Eliot Dresselhaus
- *
- * Permission is hereby granted, free of charge, to any person obtaining
- * a copy of this software and associated documentation files (the
- * "Software"), to deal in the Software without restriction, including
- * without limitation the rights to use, copy, modify, merge, publish,
- * distribute, sublicense, and/or sell copies of the Software, and to
- * permit persons to whom the Software is furnished to do so, subject to
- * the following conditions:
- *
- * The above copyright notice and this permission notice shall be
- * included in all copies or substantial portions of the Software.
- *
- *  THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
- *  EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
- *  MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
- *  NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE
- *  LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
- *  OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
- *  WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
- */
-
-#ifndef included_tcp_packet_h
-#define included_tcp_packet_h
-
-/* TCP flags bit 0 first. */
-#define foreach_tcp_flag			\
-  _ (FIN)					\
-  _ (SYN)					\
-  _ (RST)					\
-  _ (PSH)					\
-  _ (ACK)					\
-  _ (URG)					\
-  _ (ECE)					\
-  _ (CWR)
-
-enum
-{
-#define _(f) TCP_FLAG_BIT_##f,
-  foreach_tcp_flag
-#undef _
-    TCP_N_FLAG_BITS,
-
-#define _(f) TCP_FLAG_##f = 1 << TCP_FLAG_BIT_##f,
-  foreach_tcp_flag
-#undef _
-};
-
-typedef struct
-{
-  /* Source and destination port. */
-  union
-  {
-    union
-    {
-      struct
-      {
-	u16 src, dst;
-      };
-      u32 src_and_dst;
-    } ports;
-    struct
-    {
-      u16 src_port, dst_port;
-    };
-  };
-
-  /* Sequence and acknowledgment number. */
-  u32 seq_number, ack_number;
-
-  /* Size of TCP header in 32-bit units plus 4 reserved bits. */
-  u8 tcp_header_u32s_and_reserved;
-
-  /* see foreach_tcp_flag for enumation of tcp flags. */
-  u8 flags;
-
-  /* Current window advertised by sender.
-     This is the number of bytes sender is willing to receive
-     right now. */
-  u16 window;
-
-  /* Checksum of TCP pseudo header and data. */
-  u16 checksum;
-
-  u16 urgent_pointer;
-} tcp_header_t;
-
-always_inline int
-tcp_header_bytes (tcp_header_t * t)
-{
-  return (t->tcp_header_u32s_and_reserved >> 4) * sizeof (u32);
-}
-
-/* TCP options. */
-typedef enum tcp_option_type
-{
-  TCP_OPTION_END = 0,
-  TCP_OPTION_NOP = 1,
-  TCP_OPTION_MSS = 2,
-  TCP_OPTION_WINDOW_SCALE = 3,
-  TCP_OPTION_SACK_PERMITTED = 4,
-  TCP_OPTION_SACK_BLOCK = 5,
-  TCP_OPTION_TIME_STAMP = 8,
-} tcp_option_type_t;
-
-/* All except NOP and END have 1 byte length field. */
-typedef struct
-{
-  tcp_option_type_t type:8;
-
-  /* Length of this option in bytes. */
-  u8 length;
-} tcp_option_with_length_t;
-
-#endif /* included_tcp_packet_h */
-
-
-/*
- * fd.io coding-style-patch-verification: ON
- *
- * Local Variables:
- * eval: (c-set-style "gnu")
- * End:
- */
diff --git a/src/vnet/ip/udp.h b/src/vnet/ip/udp.h
deleted file mode 100644
index bad58b5d..00000000
--- a/src/vnet/ip/udp.h
+++ /dev/null
@@ -1,315 +0,0 @@
-/*
- * ip/udp.h: udp protocol
- *
- * Copyright (c) 2013 Cisco and/or its affiliates.
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at:
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#ifndef included_udp_h
-#define included_udp_h
-
-#include 
-#include 
-#include 
-#include 
-#include 
-#include 
-#include 
-
-typedef enum
-{
-#define udp_error(n,s) UDP_ERROR_##n,
-#include 
-#undef udp_error
-  UDP_N_ERROR,
-} udp_error_t;
-
-#define foreach_udp4_dst_port			\
-_ (67, dhcp_to_server)                          \
-_ (68, dhcp_to_client)                          \
-_ (500, ikev2)                                  \
-_ (3784, bfd4)                                  \
-_ (3785, bfd_echo4)                             \
-_ (4341, lisp_gpe)                              \
-_ (4342, lisp_cp)                          	\
-_ (4739, ipfix)                                 \
-_ (4789, vxlan)					\
-_ (4789, vxlan6)				\
-_ (4790, vxlan_gpe)				\
-_ (6633, vpath_3)
-
-
-#define foreach_udp6_dst_port                   \
-_ (547, dhcpv6_to_server)                       \
-_ (546, dhcpv6_to_client)			\
-_ (3784, bfd6)                                  \
-_ (3785, bfd_echo6)                             \
-_ (4341, lisp_gpe6)                             \
-_ (4342, lisp_cp6)                          	\
-_ (4790, vxlan6_gpe)      \
-_ (6633, vpath6_3)
-
-typedef enum
-{
-#define _(n,f) UDP_DST_PORT_##f = n,
-  foreach_udp4_dst_port foreach_udp6_dst_port
-#undef _
-} udp_dst_port_t;
-
-typedef enum
-{
-#define _(n,f) UDP6_DST_PORT_##f = n,
-  foreach_udp6_dst_port
-#undef _
-} udp6_dst_port_t;
-
-typedef struct
-{
-  /* Name (a c string). */
-  char *name;
-
-  /* GRE protocol type in host byte order. */
-  udp_dst_port_t dst_port;
-
-  /* Node which handles this type. */
-  u32 node_index;
-
-  /* Next index for this type. */
-  u32 next_index;
-} udp_dst_port_info_t;
-
-typedef enum
-{
-  UDP_IP6 = 0,
-  UDP_IP4,			/* the code is full of is_ip4... */
-  N_UDP_AF,
-} udp_af_t;
-
-typedef struct
-{
-  udp_dst_port_info_t *dst_port_infos[N_UDP_AF];
-
-  /* Hash tables mapping name/protocol to protocol info index. */
-  uword *dst_port_info_by_name[N_UDP_AF];
-  uword *dst_port_info_by_dst_port[N_UDP_AF];
-
-  /* convenience */
-  vlib_main_t *vlib_main;
-} udp_main_t;
-
-always_inline udp_dst_port_info_t *
-udp_get_dst_port_info (udp_main_t * um, udp_dst_port_t dst_port, u8 is_ip4)
-{
-  uword *p = hash_get (um->dst_port_info_by_dst_port[is_ip4], dst_port);
-  return p ? vec_elt_at_index (um->dst_port_infos[is_ip4], p[0]) : 0;
-}
-
-format_function_t format_udp_header;
-format_function_t format_udp_rx_trace;
-
-unformat_function_t unformat_udp_header;
-
-void udp_register_dst_port (vlib_main_t * vm,
-			    udp_dst_port_t dst_port,
-			    u32 node_index, u8 is_ip4);
-
-void udp_punt_unknown (vlib_main_t * vm, u8 is_ip4, u8 is_add);
-
-always_inline void
-ip_udp_fixup_one (vlib_main_t * vm, vlib_buffer_t * b0, u8 is_ip4)
-{
-  u16 new_l0;
-  udp_header_t *udp0;
-
-  if (is_ip4)
-    {
-      ip4_header_t *ip0;
-      ip_csum_t sum0;
-      u16 old_l0 = 0;
-
-      ip0 = vlib_buffer_get_current (b0);
-
-      /* fix the ing outer-IP checksum */
-      sum0 = ip0->checksum;
-      /* old_l0 always 0, see the rewrite setup */
-      new_l0 = clib_host_to_net_u16 (vlib_buffer_length_in_chain (vm, b0));
-
-      sum0 = ip_csum_update (sum0, old_l0, new_l0, ip4_header_t,
-			     length /* changed member */ );
-      ip0->checksum = ip_csum_fold (sum0);
-      ip0->length = new_l0;
-
-      /* Fix UDP length */
-      udp0 = (udp_header_t *) (ip0 + 1);
-      new_l0 = clib_host_to_net_u16 (vlib_buffer_length_in_chain (vm, b0)
-				     - sizeof (*ip0));
-      udp0->length = new_l0;
-    }
-  else
-    {
-      ip6_header_t *ip0;
-      int bogus0;
-
-      ip0 = vlib_buffer_get_current (b0);
-
-      new_l0 = clib_host_to_net_u16 (vlib_buffer_length_in_chain (vm, b0)
-				     - sizeof (*ip0));
-      ip0->payload_length = new_l0;
-
-      /* Fix UDP length */
-      udp0 = (udp_header_t *) (ip0 + 1);
-      udp0->length = new_l0;
-
-      udp0->checksum =
-	ip6_tcp_udp_icmp_compute_checksum (vm, b0, ip0, &bogus0);
-      ASSERT (bogus0 == 0);
-
-      if (udp0->checksum == 0)
-	udp0->checksum = 0xffff;
-    }
-}
-
-always_inline void
-ip_udp_encap_one (vlib_main_t * vm, vlib_buffer_t * b0, u8 * ec0, word ec_len,
-		  u8 is_ip4)
-{
-  vlib_buffer_advance (b0, -ec_len);
-
-  if (is_ip4)
-    {
-      ip4_header_t *ip0;
-
-      ip0 = vlib_buffer_get_current (b0);
-
-      /* Apply the encap string. */
-      clib_memcpy (ip0, ec0, ec_len);
-      ip_udp_fixup_one (vm, b0, 1);
-    }
-  else
-    {
-      ip6_header_t *ip0;
-
-      ip0 = vlib_buffer_get_current (b0);
-
-      /* Apply the encap string. */
-      clib_memcpy (ip0, ec0, ec_len);
-      ip_udp_fixup_one (vm, b0, 0);
-    }
-}
-
-always_inline void
-ip_udp_encap_two (vlib_main_t * vm, vlib_buffer_t * b0, vlib_buffer_t * b1,
-		  u8 * ec0, u8 * ec1, word ec_len, u8 is_v4)
-{
-  u16 new_l0, new_l1;
-  udp_header_t *udp0, *udp1;
-
-  ASSERT (_vec_len (ec0) == _vec_len (ec1));
-
-  vlib_buffer_advance (b0, -ec_len);
-  vlib_buffer_advance (b1, -ec_len);
-
-  if (is_v4)
-    {
-      ip4_header_t *ip0, *ip1;
-      ip_csum_t sum0, sum1;
-      u16 old_l0 = 0, old_l1 = 0;
-
-      ip0 = vlib_buffer_get_current (b0);
-      ip1 = vlib_buffer_get_current (b1);
-
-      /* Apply the encap string */
-      clib_memcpy (ip0, ec0, ec_len);
-      clib_memcpy (ip1, ec1, ec_len);
-
-      /* fix the ing outer-IP checksum */
-      sum0 = ip0->checksum;
-      sum1 = ip1->checksum;
-
-      /* old_l0 always 0, see the rewrite setup */
-      new_l0 = clib_host_to_net_u16 (vlib_buffer_length_in_chain (vm, b0));
-      new_l1 = clib_host_to_net_u16 (vlib_buffer_length_in_chain (vm, b1));
-
-      sum0 = ip_csum_update (sum0, old_l0, new_l0, ip4_header_t,
-			     length /* changed member */ );
-      sum1 = ip_csum_update (sum1, old_l1, new_l1, ip4_header_t,
-			     length /* changed member */ );
-
-      ip0->checksum = ip_csum_fold (sum0);
-      ip1->checksum = ip_csum_fold (sum1);
-
-      ip0->length = new_l0;
-      ip1->length = new_l1;
-
-      /* Fix UDP length */
-      udp0 = (udp_header_t *) (ip0 + 1);
-      udp1 = (udp_header_t *) (ip1 + 1);
-
-      new_l0 =
-	clib_host_to_net_u16 (vlib_buffer_length_in_chain (vm, b0) -
-			      sizeof (*ip0));
-      new_l1 =
-	clib_host_to_net_u16 (vlib_buffer_length_in_chain (vm, b1) -
-			      sizeof (*ip1));
-      udp0->length = new_l0;
-      udp1->length = new_l1;
-    }
-  else
-    {
-      ip6_header_t *ip0, *ip1;
-      int bogus0, bogus1;
-
-      ip0 = vlib_buffer_get_current (b0);
-      ip1 = vlib_buffer_get_current (b1);
-
-      /* Apply the encap string. */
-      clib_memcpy (ip0, ec0, ec_len);
-      clib_memcpy (ip1, ec1, ec_len);
-
-      new_l0 = clib_host_to_net_u16 (vlib_buffer_length_in_chain (vm, b0)
-				     - sizeof (*ip0));
-      new_l1 = clib_host_to_net_u16 (vlib_buffer_length_in_chain (vm, b1)
-				     - sizeof (*ip1));
-      ip0->payload_length = new_l0;
-      ip1->payload_length = new_l1;
-
-      /* Fix UDP length */
-      udp0 = (udp_header_t *) (ip0 + 1);
-      udp1 = (udp_header_t *) (ip1 + 1);
-
-      udp0->length = new_l0;
-      udp1->length = new_l1;
-
-      udp0->checksum =
-	ip6_tcp_udp_icmp_compute_checksum (vm, b0, ip0, &bogus0);
-      udp1->checksum =
-	ip6_tcp_udp_icmp_compute_checksum (vm, b1, ip1, &bogus1);
-      ASSERT (bogus0 == 0);
-      ASSERT (bogus1 == 0);
-
-      if (udp0->checksum == 0)
-	udp0->checksum = 0xffff;
-      if (udp1->checksum == 0)
-	udp1->checksum = 0xffff;
-    }
-}
-
-#endif /* included_udp_h */
-
-/*
- * fd.io coding-style-patch-verification: ON
- *
- * Local Variables:
- * eval: (c-set-style "gnu")
- * End:
- */
diff --git a/src/vnet/ip/udp_error.def b/src/vnet/ip/udp_error.def
deleted file mode 100644
index bfdae0ac..00000000
--- a/src/vnet/ip/udp_error.def
+++ /dev/null
@@ -1,21 +0,0 @@
-/*
- * udp_error.def: udp errors
- *
- * Copyright (c) 2013-2016 Cisco and/or its affiliates.
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at:
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-udp_error (NONE, "no error")
-udp_error (NO_LISTENER, "no listener for dst port")
-udp_error (LENGTH_ERROR, "UDP packets with length errors")
-udp_error (PUNT, "no listener punt")
diff --git a/src/vnet/ip/udp_format.c b/src/vnet/ip/udp_format.c
deleted file mode 100644
index abdf561e..00000000
--- a/src/vnet/ip/udp_format.c
+++ /dev/null
@@ -1,91 +0,0 @@
-/*
- * Copyright (c) 2015 Cisco and/or its affiliates.
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at:
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-/*
- * ip/udp_format.c: udp formatting
- *
- * Copyright (c) 2008 Eliot Dresselhaus
- *
- * Permission is hereby granted, free of charge, to any person obtaining
- * a copy of this software and associated documentation files (the
- * "Software"), to deal in the Software without restriction, including
- * without limitation the rights to use, copy, modify, merge, publish,
- * distribute, sublicense, and/or sell copies of the Software, and to
- * permit persons to whom the Software is furnished to do so, subject to
- * the following conditions:
- *
- * The above copyright notice and this permission notice shall be
- * included in all copies or substantial portions of the Software.
- *
- *  THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
- *  EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
- *  MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
- *  NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE
- *  LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
- *  OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
- *  WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
- */
-
-#include 
-
-/* Format UDP header. */
-u8 *
-format_udp_header (u8 * s, va_list * args)
-{
-  udp_header_t *udp = va_arg (*args, udp_header_t *);
-  u32 max_header_bytes = va_arg (*args, u32);
-  uword indent;
-  u32 header_bytes = sizeof (udp[0]);
-
-  /* Nothing to do. */
-  if (max_header_bytes < sizeof (udp[0]))
-    return format (s, "UDP header truncated");
-
-  indent = format_get_indent (s);
-  indent += 2;
-
-  s = format (s, "UDP: %d -> %d",
-	      clib_net_to_host_u16 (udp->src_port),
-	      clib_net_to_host_u16 (udp->dst_port));
-
-  s = format (s, "\n%Ulength %d, checksum 0x%04x",
-	      format_white_space, indent,
-	      clib_net_to_host_u16 (udp->length),
-	      clib_net_to_host_u16 (udp->checksum));
-
-  /* Recurse into next protocol layer. */
-  if (max_header_bytes != 0 && header_bytes < max_header_bytes)
-    {
-      ip_main_t *im = &ip_main;
-      tcp_udp_port_info_t *pi;
-
-      pi = ip_get_tcp_udp_port_info (im, udp->dst_port);
-
-      if (pi && pi->format_header)
-	s = format (s, "\n%U%U",
-		    format_white_space, indent - 2, pi->format_header,
-		    /* next protocol header */ (udp + 1),
-		    max_header_bytes - sizeof (udp[0]));
-    }
-
-  return s;
-}
-
-/*
- * fd.io coding-style-patch-verification: ON
- *
- * Local Variables:
- * eval: (c-set-style "gnu")
- * End:
- */
diff --git a/src/vnet/ip/udp_init.c b/src/vnet/ip/udp_init.c
deleted file mode 100644
index 1241ca4a..00000000
--- a/src/vnet/ip/udp_init.c
+++ /dev/null
@@ -1,71 +0,0 @@
-/*
- * Copyright (c) 2015 Cisco and/or its affiliates.
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at:
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-/*
- * ip/udp_init.c: udp initialization
- *
- * Copyright (c) 2008 Eliot Dresselhaus
- *
- * Permission is hereby granted, free of charge, to any person obtaining
- * a copy of this software and associated documentation files (the
- * "Software"), to deal in the Software without restriction, including
- * without limitation the rights to use, copy, modify, merge, publish,
- * distribute, sublicense, and/or sell copies of the Software, and to
- * permit persons to whom the Software is furnished to do so, subject to
- * the following conditions:
- *
- * The above copyright notice and this permission notice shall be
- * included in all copies or substantial portions of the Software.
- *
- *  THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
- *  EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
- *  MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
- *  NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE
- *  LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
- *  OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
- *  WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
- */
-
-#include 
-
-clib_error_t *
-udp_init (vlib_main_t * vm)
-{
-  ip_main_t *im = &ip_main;
-  ip_protocol_info_t *pi;
-  clib_error_t *error;
-
-  error = vlib_call_init_function (vm, ip_main_init);
-
-  if (!error)
-    {
-      pi = ip_get_protocol_info (im, IP_PROTOCOL_UDP);
-      if (pi == 0)
-	return clib_error_return (0, "UDP protocol info AWOL");
-      pi->format_header = format_udp_header;
-      pi->unformat_pg_edit = unformat_pg_udp_header;
-    }
-
-  return 0;
-}
-
-VLIB_INIT_FUNCTION (udp_init);
-
-/*
- * fd.io coding-style-patch-verification: ON
- *
- * Local Variables:
- * eval: (c-set-style "gnu")
- * End:
- */
diff --git a/src/vnet/ip/udp_local.c b/src/vnet/ip/udp_local.c
deleted file mode 100644
index 13ab6e4f..00000000
--- a/src/vnet/ip/udp_local.c
+++ /dev/null
@@ -1,645 +0,0 @@
-/*
- * node.c: udp packet processing
- *
- * Copyright (c) 2013 Cisco and/or its affiliates.
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at:
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#include 
-#include 
-#include 
-#include 
-#include 
-
-udp_main_t udp_main;
-
-#define foreach_udp_input_next                  \
-  _ (PUNT, "error-punt")                        \
-  _ (DROP, "error-drop")                        \
-  _ (ICMP4_ERROR, "ip4-icmp-error")             \
-  _ (ICMP6_ERROR, "ip6-icmp-error")
-
-typedef enum
-{
-#define _(s,n) UDP_INPUT_NEXT_##s,
-  foreach_udp_input_next
-#undef _
-    UDP_INPUT_N_NEXT,
-} udp_input_next_t;
-
-typedef struct
-{
-  u16 src_port;
-  u16 dst_port;
-  u8 bound;
-} udp_rx_trace_t;
-
-u8 *
-format_udp_rx_trace (u8 * s, va_list * args)
-{
-  CLIB_UNUSED (vlib_main_t * vm) = va_arg (*args, vlib_main_t *);
-  CLIB_UNUSED (vlib_node_t * node) = va_arg (*args, vlib_node_t *);
-  udp_rx_trace_t *t = va_arg (*args, udp_rx_trace_t *);
-
-  s = format (s, "UDP: src-port %d dst-port %d%s",
-	      clib_net_to_host_u16 (t->src_port),
-	      clib_net_to_host_u16 (t->dst_port),
-	      t->bound ? "" : " (no listener)");
-  return s;
-}
-
-typedef struct
-{
-  /* Sparse vector mapping udp dst_port in network byte order
-     to next index. */
-  u16 *next_by_dst_port;
-  u8 punt_unknown;
-} udp_input_runtime_t;
-
-vlib_node_registration_t udp4_input_node;
-vlib_node_registration_t udp6_input_node;
-
-always_inline uword
-udp46_input_inline (vlib_main_t * vm,
-		    vlib_node_runtime_t * node,
-		    vlib_frame_t * from_frame, int is_ip4)
-{
-  udp_input_runtime_t *rt = is_ip4 ?
-    (void *) vlib_node_get_runtime_data (vm, udp4_input_node.index)
-    : (void *) vlib_node_get_runtime_data (vm, udp6_input_node.index);
-  __attribute__ ((unused)) u32 n_left_from, next_index, *from, *to_next;
-  word n_no_listener = 0;
-  u8 punt_unknown = rt->punt_unknown;
-
-  from = vlib_frame_vector_args (from_frame);
-  n_left_from = from_frame->n_vectors;
-
-  next_index = node->cached_next_index;
-
-  while (n_left_from > 0)
-    {
-      u32 n_left_to_next;
-
-      vlib_get_next_frame (vm, node, next_index, to_next, n_left_to_next);
-
-      while (n_left_from >= 4 && n_left_to_next >= 2)
-	{
-	  u32 bi0, bi1;
-	  vlib_buffer_t *b0, *b1;
-	  udp_header_t *h0 = 0, *h1 = 0;
-	  u32 i0, i1, dst_port0, dst_port1;
-	  u32 advance0, advance1;
-	  u32 error0, next0, error1, next1;
-
-	  /* Prefetch next iteration. */
-	  {
-	    vlib_buffer_t *p2, *p3;
-
-	    p2 = vlib_get_buffer (vm, from[2]);
-	    p3 = vlib_get_buffer (vm, from[3]);
-
-	    vlib_prefetch_buffer_header (p2, LOAD);
-	    vlib_prefetch_buffer_header (p3, LOAD);
-
-	    CLIB_PREFETCH (p2->data, sizeof (h0[0]), LOAD);
-	    CLIB_PREFETCH (p3->data, sizeof (h1[0]), LOAD);
-	  }
-
-	  bi0 = from[0];
-	  bi1 = from[1];
-	  to_next[0] = bi0;
-	  to_next[1] = bi1;
-	  from += 2;
-	  to_next += 2;
-	  n_left_to_next -= 2;
-	  n_left_from -= 2;
-
-	  b0 = vlib_get_buffer (vm, bi0);
-	  b1 = vlib_get_buffer (vm, bi1);
-
-	  /* ip4/6_local hands us the ip header, not the udp header */
-	  if (is_ip4)
-	    {
-	      advance0 = sizeof (ip4_header_t);
-	      advance1 = sizeof (ip4_header_t);
-	    }
-	  else
-	    {
-	      advance0 = sizeof (ip6_header_t);
-	      advance1 = sizeof (ip6_header_t);
-	    }
-
-	  if (PREDICT_FALSE (b0->current_length < advance0 + sizeof (*h0)))
-	    {
-	      error0 = UDP_ERROR_LENGTH_ERROR;
-	      next0 = UDP_INPUT_NEXT_DROP;
-	    }
-	  else
-	    {
-	      vlib_buffer_advance (b0, advance0);
-	      h0 = vlib_buffer_get_current (b0);
-	      error0 = next0 = 0;
-	      if (PREDICT_FALSE (clib_net_to_host_u16 (h0->length) >
-				 vlib_buffer_length_in_chain (vm, b0)))
-		{
-		  error0 = UDP_ERROR_LENGTH_ERROR;
-		  next0 = UDP_INPUT_NEXT_DROP;
-		}
-	    }
-
-	  if (PREDICT_FALSE (b1->current_length < advance1 + sizeof (*h1)))
-	    {
-	      error1 = UDP_ERROR_LENGTH_ERROR;
-	      next1 = UDP_INPUT_NEXT_DROP;
-	    }
-	  else
-	    {
-	      vlib_buffer_advance (b1, advance1);
-	      h1 = vlib_buffer_get_current (b1);
-	      error1 = next1 = 0;
-	      if (PREDICT_FALSE (clib_net_to_host_u16 (h1->length) >
-				 vlib_buffer_length_in_chain (vm, b1)))
-		{
-		  error1 = UDP_ERROR_LENGTH_ERROR;
-		  next1 = UDP_INPUT_NEXT_DROP;
-		}
-	    }
-
-	  /* Index sparse array with network byte order. */
-	  dst_port0 = (error0 == 0) ? h0->dst_port : 0;
-	  dst_port1 = (error1 == 0) ? h1->dst_port : 0;
-	  sparse_vec_index2 (rt->next_by_dst_port, dst_port0, dst_port1,
-			     &i0, &i1);
-	  next0 = (error0 == 0) ? vec_elt (rt->next_by_dst_port, i0) : next0;
-	  next1 = (error1 == 0) ? vec_elt (rt->next_by_dst_port, i1) : next1;
-
-	  if (PREDICT_FALSE (i0 == SPARSE_VEC_INVALID_INDEX))
-	    {
-	      // move the pointer back so icmp-error can find the
-	      // ip packet header
-	      vlib_buffer_advance (b0, -(word) advance0);
-
-	      if (PREDICT_FALSE (punt_unknown))
-		{
-		  b0->error = node->errors[UDP_ERROR_PUNT];
-		  next0 = UDP_INPUT_NEXT_PUNT;
-		}
-	      else if (is_ip4)
-		{
-		  icmp4_error_set_vnet_buffer (b0,
-					       ICMP4_destination_unreachable,
-					       ICMP4_destination_unreachable_port_unreachable,
-					       0);
-		  next0 = UDP_INPUT_NEXT_ICMP4_ERROR;
-		  n_no_listener++;
-		}
-	      else
-		{
-		  icmp6_error_set_vnet_buffer (b0,
-					       ICMP6_destination_unreachable,
-					       ICMP6_destination_unreachable_port_unreachable,
-					       0);
-		  next0 = UDP_INPUT_NEXT_ICMP6_ERROR;
-		  n_no_listener++;
-		}
-	    }
-	  else
-	    {
-	      b0->error = node->errors[UDP_ERROR_NONE];
-	      // advance to the payload
-	      vlib_buffer_advance (b0, sizeof (*h0));
-	    }
-
-	  if (PREDICT_FALSE (i1 == SPARSE_VEC_INVALID_INDEX))
-	    {
-	      // move the pointer back so icmp-error can find the
-	      // ip packet header
-	      vlib_buffer_advance (b1, -(word) advance1);
-
-	      if (PREDICT_FALSE (punt_unknown))
-		{
-		  b1->error = node->errors[UDP_ERROR_PUNT];
-		  next1 = UDP_INPUT_NEXT_PUNT;
-		}
-	      else if (is_ip4)
-		{
-		  icmp4_error_set_vnet_buffer (b1,
-					       ICMP4_destination_unreachable,
-					       ICMP4_destination_unreachable_port_unreachable,
-					       0);
-		  next1 = UDP_INPUT_NEXT_ICMP4_ERROR;
-		  n_no_listener++;
-		}
-	      else
-		{
-		  icmp6_error_set_vnet_buffer (b1,
-					       ICMP6_destination_unreachable,
-					       ICMP6_destination_unreachable_port_unreachable,
-					       0);
-		  next1 = UDP_INPUT_NEXT_ICMP6_ERROR;
-		  n_no_listener++;
-		}
-	    }
-	  else
-	    {
-	      b1->error = node->errors[UDP_ERROR_NONE];
-	      // advance to the payload
-	      vlib_buffer_advance (b1, sizeof (*h1));
-	    }
-
-	  if (PREDICT_FALSE (b0->flags & VLIB_BUFFER_IS_TRACED))
-	    {
-	      udp_rx_trace_t *tr = vlib_add_trace (vm, node,
-						   b0, sizeof (*tr));
-	      if (b0->error != node->errors[UDP_ERROR_LENGTH_ERROR])
-		{
-		  tr->src_port = h0 ? h0->src_port : 0;
-		  tr->dst_port = h0 ? h0->dst_port : 0;
-		  tr->bound = (next0 != UDP_INPUT_NEXT_ICMP4_ERROR &&
-			       next0 != UDP_INPUT_NEXT_ICMP6_ERROR);
-		}
-	    }
-	  if (PREDICT_FALSE (b1->flags & VLIB_BUFFER_IS_TRACED))
-	    {
-	      udp_rx_trace_t *tr = vlib_add_trace (vm, node,
-						   b1, sizeof (*tr));
-	      if (b1->error != node->errors[UDP_ERROR_LENGTH_ERROR])
-		{
-		  tr->src_port = h1 ? h1->src_port : 0;
-		  tr->dst_port = h1 ? h1->dst_port : 0;
-		  tr->bound = (next1 != UDP_INPUT_NEXT_ICMP4_ERROR &&
-			       next1 != UDP_INPUT_NEXT_ICMP6_ERROR);
-		}
-	    }
-
-	  vlib_validate_buffer_enqueue_x2 (vm, node, next_index,
-					   to_next, n_left_to_next,
-					   bi0, bi1, next0, next1);
-	}
-
-      while (n_left_from > 0 && n_left_to_next > 0)
-	{
-	  u32 bi0;
-	  vlib_buffer_t *b0;
-	  udp_header_t *h0 = 0;
-	  u32 i0, next0;
-	  u32 advance0;
-
-	  bi0 = from[0];
-	  to_next[0] = bi0;
-	  from += 1;
-	  to_next += 1;
-	  n_left_from -= 1;
-	  n_left_to_next -= 1;
-
-	  b0 = vlib_get_buffer (vm, bi0);
-
-	  /* ip4/6_local hands us the ip header, not the udp header */
-	  if (is_ip4)
-	    advance0 = sizeof (ip4_header_t);
-	  else
-	    advance0 = sizeof (ip6_header_t);
-
-	  if (PREDICT_FALSE (b0->current_length < advance0 + sizeof (*h0)))
-	    {
-	      b0->error = node->errors[UDP_ERROR_LENGTH_ERROR];
-	      next0 = UDP_INPUT_NEXT_DROP;
-	      goto trace_x1;
-	    }
-
-	  vlib_buffer_advance (b0, advance0);
-
-	  h0 = vlib_buffer_get_current (b0);
-
-	  if (PREDICT_TRUE (clib_net_to_host_u16 (h0->length) <=
-			    vlib_buffer_length_in_chain (vm, b0)))
-	    {
-	      i0 = sparse_vec_index (rt->next_by_dst_port, h0->dst_port);
-	      next0 = vec_elt (rt->next_by_dst_port, i0);
-
-	      if (PREDICT_FALSE (i0 == SPARSE_VEC_INVALID_INDEX))
-		{
-		  // move the pointer back so icmp-error can find the
-		  // ip packet header
-		  vlib_buffer_advance (b0, -(word) advance0);
-
-		  if (PREDICT_FALSE (punt_unknown))
-		    {
-		      b0->error = node->errors[UDP_ERROR_PUNT];
-		      next0 = UDP_INPUT_NEXT_PUNT;
-		    }
-		  else if (is_ip4)
-		    {
-		      icmp4_error_set_vnet_buffer (b0,
-						   ICMP4_destination_unreachable,
-						   ICMP4_destination_unreachable_port_unreachable,
-						   0);
-		      next0 = UDP_INPUT_NEXT_ICMP4_ERROR;
-		      n_no_listener++;
-		    }
-		  else
-		    {
-		      icmp6_error_set_vnet_buffer (b0,
-						   ICMP6_destination_unreachable,
-						   ICMP6_destination_unreachable_port_unreachable,
-						   0);
-		      next0 = UDP_INPUT_NEXT_ICMP6_ERROR;
-		      n_no_listener++;
-		    }
-		}
-	      else
-		{
-		  b0->error = node->errors[UDP_ERROR_NONE];
-		  // advance to the payload
-		  vlib_buffer_advance (b0, sizeof (*h0));
-		}
-	    }
-	  else
-	    {
-	      b0->error = node->errors[UDP_ERROR_LENGTH_ERROR];
-	      next0 = UDP_INPUT_NEXT_DROP;
-	    }
-
-	trace_x1:
-	  if (PREDICT_FALSE (b0->flags & VLIB_BUFFER_IS_TRACED))
-	    {
-	      udp_rx_trace_t *tr = vlib_add_trace (vm, node,
-						   b0, sizeof (*tr));
-	      if (b0->error != node->errors[UDP_ERROR_LENGTH_ERROR])
-		{
-		  tr->src_port = h0->src_port;
-		  tr->dst_port = h0->dst_port;
-		  tr->bound = (next0 != UDP_INPUT_NEXT_ICMP4_ERROR &&
-			       next0 != UDP_INPUT_NEXT_ICMP6_ERROR);
-		}
-	    }
-
-	  vlib_validate_buffer_enqueue_x1 (vm, node, next_index,
-					   to_next, n_left_to_next,
-					   bi0, next0);
-	}
-
-      vlib_put_next_frame (vm, node, next_index, n_left_to_next);
-    }
-  vlib_error_count (vm, node->node_index, UDP_ERROR_NO_LISTENER,
-		    n_no_listener);
-  return from_frame->n_vectors;
-}
-
-static char *udp_error_strings[] = {
-#define udp_error(n,s) s,
-#include "udp_error.def"
-#undef udp_error
-};
-
-static uword
-udp4_input (vlib_main_t * vm,
-	    vlib_node_runtime_t * node, vlib_frame_t * from_frame)
-{
-  return udp46_input_inline (vm, node, from_frame, 1 /* is_ip4 */ );
-}
-
-static uword
-udp6_input (vlib_main_t * vm,
-	    vlib_node_runtime_t * node, vlib_frame_t * from_frame)
-{
-  return udp46_input_inline (vm, node, from_frame, 0 /* is_ip4 */ );
-}
-
-
-/* *INDENT-OFF* */
-VLIB_REGISTER_NODE (udp4_input_node) = {
-  .function = udp4_input,
-  .name = "ip4-udp-lookup",
-  /* Takes a vector of packets. */
-  .vector_size = sizeof (u32),
-
-  .runtime_data_bytes = sizeof (udp_input_runtime_t),
-
-  .n_errors = UDP_N_ERROR,
-  .error_strings = udp_error_strings,
-
-  .n_next_nodes = UDP_INPUT_N_NEXT,
-  .next_nodes = {
-#define _(s,n) [UDP_INPUT_NEXT_##s] = n,
-    foreach_udp_input_next
-#undef _
-  },
-
-  .format_buffer = format_udp_header,
-  .format_trace = format_udp_rx_trace,
-  .unformat_buffer = unformat_udp_header,
-};
-/* *INDENT-ON* */
-
-VLIB_NODE_FUNCTION_MULTIARCH (udp4_input_node, udp4_input);
-
-/* *INDENT-OFF* */
-VLIB_REGISTER_NODE (udp6_input_node) = {
-  .function = udp6_input,
-  .name = "ip6-udp-lookup",
-  /* Takes a vector of packets. */
-  .vector_size = sizeof (u32),
-
-  .runtime_data_bytes = sizeof (udp_input_runtime_t),
-
-  .n_errors = UDP_N_ERROR,
-  .error_strings = udp_error_strings,
-
-  .n_next_nodes = UDP_INPUT_N_NEXT,
-  .next_nodes = {
-#define _(s,n) [UDP_INPUT_NEXT_##s] = n,
-    foreach_udp_input_next
-#undef _
-  },
-
-  .format_buffer = format_udp_header,
-  .format_trace = format_udp_rx_trace,
-  .unformat_buffer = unformat_udp_header,
-};
-/* *INDENT-ON* */
-
-VLIB_NODE_FUNCTION_MULTIARCH (udp6_input_node, udp6_input);
-
-static void
-add_dst_port (udp_main_t * um,
-	      udp_dst_port_t dst_port, char *dst_port_name, u8 is_ip4)
-{
-  udp_dst_port_info_t *pi;
-  u32 i;
-
-  vec_add2 (um->dst_port_infos[is_ip4], pi, 1);
-  i = pi - um->dst_port_infos[is_ip4];
-
-  pi->name = dst_port_name;
-  pi->dst_port = dst_port;
-  pi->next_index = pi->node_index = ~0;
-
-  hash_set (um->dst_port_info_by_dst_port[is_ip4], dst_port, i);
-
-  if (pi->name)
-    hash_set_mem (um->dst_port_info_by_name[is_ip4], pi->name, i);
-}
-
-void
-udp_register_dst_port (vlib_main_t * vm,
-		       udp_dst_port_t dst_port, u32 node_index, u8 is_ip4)
-{
-  udp_main_t *um = &udp_main;
-  udp_dst_port_info_t *pi;
-  udp_input_runtime_t *rt;
-  u16 *n;
-
-  {
-    clib_error_t *error = vlib_call_init_function (vm, udp_local_init);
-    if (error)
-      clib_error_report (error);
-  }
-
-  pi = udp_get_dst_port_info (um, dst_port, is_ip4);
-  if (!pi)
-    {
-      add_dst_port (um, dst_port, 0, is_ip4);
-      pi = udp_get_dst_port_info (um, dst_port, is_ip4);
-      ASSERT (pi);
-    }
-
-  pi->node_index = node_index;
-  pi->next_index = vlib_node_add_next (vm,
-				       is_ip4 ? udp4_input_node.index
-				       : udp6_input_node.index, node_index);
-
-  /* Setup udp protocol -> next index sparse vector mapping. */
-  rt = vlib_node_get_runtime_data
-    (vm, is_ip4 ? udp4_input_node.index : udp6_input_node.index);
-  n = sparse_vec_validate (rt->next_by_dst_port,
-			   clib_host_to_net_u16 (dst_port));
-  n[0] = pi->next_index;
-}
-
-void
-udp_punt_unknown (vlib_main_t * vm, u8 is_ip4, u8 is_add)
-{
-  udp_input_runtime_t *rt;
-
-  {
-    clib_error_t *error = vlib_call_init_function (vm, udp_local_init);
-    if (error)
-      clib_error_report (error);
-  }
-
-  rt = vlib_node_get_runtime_data
-    (vm, is_ip4 ? udp4_input_node.index : udp6_input_node.index);
-
-  rt->punt_unknown = is_add;
-}
-
-/* Parse a UDP header. */
-uword
-unformat_udp_header (unformat_input_t * input, va_list * args)
-{
-  u8 **result = va_arg (*args, u8 **);
-  udp_header_t *udp;
-  __attribute__ ((unused)) int old_length;
-  u16 src_port, dst_port;
-
-  /* Allocate space for IP header. */
-  {
-    void *p;
-
-    old_length = vec_len (*result);
-    vec_add2 (*result, p, sizeof (ip4_header_t));
-    udp = p;
-  }
-
-  memset (udp, 0, sizeof (udp[0]));
-  if (unformat (input, "src-port %d dst-port %d", &src_port, &dst_port))
-    {
-      udp->src_port = clib_host_to_net_u16 (src_port);
-      udp->dst_port = clib_host_to_net_u16 (dst_port);
-      return 1;
-    }
-  return 0;
-}
-
-static void
-udp_setup_node (vlib_main_t * vm, u32 node_index)
-{
-  vlib_node_t *n = vlib_get_node (vm, node_index);
-  pg_node_t *pn = pg_get_node (node_index);
-
-  n->format_buffer = format_udp_header;
-  n->unformat_buffer = unformat_udp_header;
-  pn->unformat_edit = unformat_pg_udp_header;
-}
-
-clib_error_t *
-udp_local_init (vlib_main_t * vm)
-{
-  udp_input_runtime_t *rt;
-  udp_main_t *um = &udp_main;
-  int i;
-
-  {
-    clib_error_t *error;
-    error = vlib_call_init_function (vm, udp_init);
-    if (error)
-      clib_error_report (error);
-  }
-
-
-  for (i = 0; i < 2; i++)
-    {
-      um->dst_port_info_by_name[i] = hash_create_string (0, sizeof (uword));
-      um->dst_port_info_by_dst_port[i] = hash_create (0, sizeof (uword));
-    }
-
-  udp_setup_node (vm, udp4_input_node.index);
-  udp_setup_node (vm, udp6_input_node.index);
-
-  rt = vlib_node_get_runtime_data (vm, udp4_input_node.index);
-
-  rt->next_by_dst_port = sparse_vec_new
-    ( /* elt bytes */ sizeof (rt->next_by_dst_port[0]),
-     /* bits in index */ BITS (((udp_header_t *) 0)->dst_port));
-
-  rt->punt_unknown = 0;
-
-#define _(n,s) add_dst_port (um, UDP_DST_PORT_##s, #s, 1 /* is_ip4 */);
-  foreach_udp4_dst_port
-#undef _
-    rt = vlib_node_get_runtime_data (vm, udp6_input_node.index);
-
-  rt->next_by_dst_port = sparse_vec_new
-    ( /* elt bytes */ sizeof (rt->next_by_dst_port[0]),
-     /* bits in index */ BITS (((udp_header_t *) 0)->dst_port));
-
-  rt->punt_unknown = 0;
-
-#define _(n,s) add_dst_port (um, UDP_DST_PORT_##s, #s, 0 /* is_ip4 */);
-  foreach_udp6_dst_port
-#undef _
-    ip4_register_protocol (IP_PROTOCOL_UDP, udp4_input_node.index);
-  /* Note: ip6 differs from ip4, UDP is hotwired to ip6-udp-lookup */
-  return 0;
-}
-
-VLIB_INIT_FUNCTION (udp_local_init);
-
-/*
- * fd.io coding-style-patch-verification: ON
- *
- * Local Variables:
- * eval: (c-set-style "gnu")
- * End:
- */
diff --git a/src/vnet/ip/udp_packet.h b/src/vnet/ip/udp_packet.h
deleted file mode 100644
index beea3059..00000000
--- a/src/vnet/ip/udp_packet.h
+++ /dev/null
@@ -1,65 +0,0 @@
-/*
- * Copyright (c) 2015 Cisco and/or its affiliates.
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at:
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-/*
- * ip4/udp_packet.h: UDP packet format
- *
- * Copyright (c) 2008 Eliot Dresselhaus
- *
- * Permission is hereby granted, free of charge, to any person obtaining
- * a copy of this software and associated documentation files (the
- * "Software"), to deal in the Software without restriction, including
- * without limitation the rights to use, copy, modify, merge, publish,
- * distribute, sublicense, and/or sell copies of the Software, and to
- * permit persons to whom the Software is furnished to do so, subject to
- * the following conditions:
- *
- * The above copyright notice and this permission notice shall be
- * included in all copies or substantial portions of the Software.
- *
- *  THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
- *  EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
- *  MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
- *  NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE
- *  LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
- *  OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
- *  WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
- */
-
-#ifndef included_udp_packet_h
-#define included_udp_packet_h
-
-typedef struct
-{
-  /* Source and destination port. */
-  u16 src_port, dst_port;
-
-  /* Length of UDP header plus payload. */
-  u16 length;
-
-  /* Checksum of UDP pseudo-header and data or
-     zero if checksum is disabled. */
-  u16 checksum;
-} udp_header_t;
-
-#endif /* included_udp_packet_h */
-
-
-/*
- * fd.io coding-style-patch-verification: ON
- *
- * Local Variables:
- * eval: (c-set-style "gnu")
- * End:
- */
diff --git a/src/vnet/ip/udp_pg.c b/src/vnet/ip/udp_pg.c
deleted file mode 100644
index c9d8d38c..00000000
--- a/src/vnet/ip/udp_pg.c
+++ /dev/null
@@ -1,237 +0,0 @@
-/*
- * Copyright (c) 2015 Cisco and/or its affiliates.
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at:
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-/*
- * ip/udp_pg: UDP packet-generator interface
- *
- * Copyright (c) 2008 Eliot Dresselhaus
- *
- * Permission is hereby granted, free of charge, to any person obtaining
- * a copy of this software and associated documentation files (the
- * "Software"), to deal in the Software without restriction, including
- * without limitation the rights to use, copy, modify, merge, publish,
- * distribute, sublicense, and/or sell copies of the Software, and to
- * permit persons to whom the Software is furnished to do so, subject to
- * the following conditions:
- *
- * The above copyright notice and this permission notice shall be
- * included in all copies or substantial portions of the Software.
- *
- *  THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
- *  EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
- *  MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
- *  NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE
- *  LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
- *  OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
- *  WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
- */
-
-#include 
-#include 		/* for unformat_udp_udp_port */
-
-#define UDP_PG_EDIT_LENGTH (1 << 0)
-#define UDP_PG_EDIT_CHECKSUM (1 << 1)
-
-always_inline void
-udp_pg_edit_function_inline (pg_main_t * pg,
-			     pg_stream_t * s,
-			     pg_edit_group_t * g,
-			     u32 * packets, u32 n_packets, u32 flags)
-{
-  vlib_main_t *vm = vlib_get_main ();
-  u32 ip_offset, udp_offset;
-
-  udp_offset = g->start_byte_offset;
-  ip_offset = (g - 1)->start_byte_offset;
-
-  while (n_packets >= 1)
-    {
-      vlib_buffer_t *p0;
-      ip4_header_t *ip0;
-      udp_header_t *udp0;
-      u32 udp_len0;
-
-      p0 = vlib_get_buffer (vm, packets[0]);
-      n_packets -= 1;
-      packets += 1;
-
-      ip0 = (void *) (p0->data + ip_offset);
-      udp0 = (void *) (p0->data + udp_offset);
-      udp_len0 = clib_net_to_host_u16 (ip0->length) - sizeof (ip0[0]);
-
-      if (flags & UDP_PG_EDIT_LENGTH)
-	udp0->length =
-	  clib_net_to_host_u16 (vlib_buffer_length_in_chain (vm, p0)
-				- ip_offset);
-
-      /* Initialize checksum with header. */
-      if (flags & UDP_PG_EDIT_CHECKSUM)
-	{
-	  ip_csum_t sum0;
-
-	  sum0 = clib_mem_unaligned (&ip0->src_address, u64);
-
-	  sum0 = ip_csum_with_carry
-	    (sum0, clib_host_to_net_u32 (udp_len0 + (ip0->protocol << 16)));
-
-	  /* Invalidate possibly old checksum. */
-	  udp0->checksum = 0;
-
-	  sum0 =
-	    ip_incremental_checksum_buffer (vm, p0, udp_offset, udp_len0,
-					    sum0);
-
-	  sum0 = ~ip_csum_fold (sum0);
-
-	  /* Zero checksum means checksumming disabled. */
-	  sum0 = sum0 != 0 ? sum0 : 0xffff;
-
-	  udp0->checksum = sum0;
-	}
-    }
-}
-
-static void
-udp_pg_edit_function (pg_main_t * pg,
-		      pg_stream_t * s,
-		      pg_edit_group_t * g, u32 * packets, u32 n_packets)
-{
-  switch (g->edit_function_opaque)
-    {
-    case UDP_PG_EDIT_LENGTH:
-      udp_pg_edit_function_inline (pg, s, g, packets, n_packets,
-				   UDP_PG_EDIT_LENGTH);
-      break;
-
-    case UDP_PG_EDIT_CHECKSUM:
-      udp_pg_edit_function_inline (pg, s, g, packets, n_packets,
-				   UDP_PG_EDIT_CHECKSUM);
-      break;
-
-    case UDP_PG_EDIT_CHECKSUM | UDP_PG_EDIT_LENGTH:
-      udp_pg_edit_function_inline (pg, s, g, packets, n_packets,
-				   UDP_PG_EDIT_CHECKSUM | UDP_PG_EDIT_LENGTH);
-      break;
-
-    default:
-      ASSERT (0);
-      break;
-    }
-}
-
-typedef struct
-{
-  pg_edit_t src_port, dst_port;
-  pg_edit_t length;
-  pg_edit_t checksum;
-} pg_udp_header_t;
-
-static inline void
-pg_udp_header_init (pg_udp_header_t * p)
-{
-  /* Initialize fields that are not bit fields in the IP header. */
-#define _(f) pg_edit_init (&p->f, udp_header_t, f);
-  _(src_port);
-  _(dst_port);
-  _(length);
-  _(checksum);
-#undef _
-}
-
-uword
-unformat_pg_udp_header (unformat_input_t * input, va_list * args)
-{
-  pg_stream_t *s = va_arg (*args, pg_stream_t *);
-  pg_udp_header_t *p;
-  u32 group_index;
-
-  p = pg_create_edit_group (s, sizeof (p[0]), sizeof (udp_header_t),
-			    &group_index);
-  pg_udp_header_init (p);
-
-  /* Defaults. */
-  p->checksum.type = PG_EDIT_UNSPECIFIED;
-  p->length.type = PG_EDIT_UNSPECIFIED;
-
-  if (!unformat (input, "UDP: %U -> %U",
-		 unformat_pg_edit,
-		 unformat_tcp_udp_port, &p->src_port,
-		 unformat_pg_edit, unformat_tcp_udp_port, &p->dst_port))
-    goto error;
-
-  /* Parse options. */
-  while (1)
-    {
-      if (unformat (input, "length %U",
-		    unformat_pg_edit, unformat_pg_number, &p->length))
-	;
-
-      else if (unformat (input, "checksum %U",
-			 unformat_pg_edit, unformat_pg_number, &p->checksum))
-	;
-
-      /* Can't parse input: try next protocol level. */
-      else
-	break;
-    }
-
-  {
-    ip_main_t *im = &ip_main;
-    u16 dst_port;
-    tcp_udp_port_info_t *pi;
-
-    pi = 0;
-    if (p->dst_port.type == PG_EDIT_FIXED)
-      {
-	dst_port = pg_edit_get_value (&p->dst_port, PG_EDIT_LO);
-	pi = ip_get_tcp_udp_port_info (im, dst_port);
-      }
-
-    if (pi && pi->unformat_pg_edit
-	&& unformat_user (input, pi->unformat_pg_edit, s))
-      ;
-
-    else if (!unformat_user (input, unformat_pg_payload, s))
-      goto error;
-
-    p = pg_get_edit_group (s, group_index);
-    if (p->checksum.type == PG_EDIT_UNSPECIFIED
-	|| p->length.type == PG_EDIT_UNSPECIFIED)
-      {
-	pg_edit_group_t *g = pg_stream_get_group (s, group_index);
-	g->edit_function = udp_pg_edit_function;
-	g->edit_function_opaque = 0;
-	if (p->checksum.type == PG_EDIT_UNSPECIFIED)
-	  g->edit_function_opaque |= UDP_PG_EDIT_CHECKSUM;
-	if (p->length.type == PG_EDIT_UNSPECIFIED)
-	  g->edit_function_opaque |= UDP_PG_EDIT_LENGTH;
-      }
-
-    return 1;
-  }
-
-error:
-  /* Free up any edits we may have added. */
-  pg_free_edit_group (s);
-  return 0;
-}
-
-
-/*
- * fd.io coding-style-patch-verification: ON
- *
- * Local Variables:
- * eval: (c-set-style "gnu")
- * End:
- */
diff --git a/src/vnet/ipsec/ikev2.c b/src/vnet/ipsec/ikev2.c
index 09209334..2c1074d8 100644
--- a/src/vnet/ipsec/ikev2.c
+++ b/src/vnet/ipsec/ikev2.c
@@ -17,7 +17,7 @@
 #include 
 #include 
 #include 
-#include 
+#include 
 #include 
 #include 
 #include 
diff --git a/src/vnet/ipsec/ikev2_cli.c b/src/vnet/ipsec/ikev2_cli.c
index 5c88d8d4..05ed4e60 100644
--- a/src/vnet/ipsec/ikev2_cli.c
+++ b/src/vnet/ipsec/ikev2_cli.c
@@ -16,7 +16,7 @@
 #include 
 #include 
 #include 
-#include 
+#include 
 #include 
 #include 
 
diff --git a/src/vnet/ipsec/ikev2_crypto.c b/src/vnet/ipsec/ikev2_crypto.c
index c201d3eb..ca56158f 100644
--- a/src/vnet/ipsec/ikev2_crypto.c
+++ b/src/vnet/ipsec/ikev2_crypto.c
@@ -17,7 +17,7 @@
 #include 
 #include 
 #include 
-#include 
+#include 
 #include 
 #include 
 #include 
diff --git a/src/vnet/lisp-cp/packets.c b/src/vnet/lisp-cp/packets.c
index 3a4f421b..f24024f1 100644
--- a/src/vnet/lisp-cp/packets.c
+++ b/src/vnet/lisp-cp/packets.c
@@ -15,7 +15,7 @@
 
 #include 
 #include 
-#include 
+#include 
 
 /* Returns IP ID for the packet */
 /* static u16 ip_id = 0;
@@ -141,61 +141,6 @@ pkt_push_udp (vlib_main_t * vm, vlib_buffer_t * b, u16 sp, u16 dp)
   return uh;
 }
 
-void *
-pkt_push_ipv4 (vlib_main_t * vm, vlib_buffer_t * b, ip4_address_t * src,
-	       ip4_address_t * dst, int proto)
-{
-  ip4_header_t *ih;
-
-  /* make some room */
-  ih = vlib_buffer_push_uninit (b, sizeof (ip4_header_t));
-
-  ih->ip_version_and_header_length = 0x45;
-  ih->tos = 0;
-  ih->length = clib_host_to_net_u16 (vlib_buffer_length_in_chain (vm, b));
-
-  /* iph->fragment_id = clib_host_to_net_u16(get_IP_ID ()); */
-
-  /* TODO: decide if we allow fragments in case of control */
-  ih->flags_and_fragment_offset = clib_host_to_net_u16 (IP_DF);
-  ih->ttl = 255;
-  ih->protocol = proto;
-  ih->src_address.as_u32 = src->as_u32;
-  ih->dst_address.as_u32 = dst->as_u32;
-
-  ih->checksum = ip4_header_checksum (ih);
-  return ih;
-}
-
-void *
-pkt_push_ipv6 (vlib_main_t * vm, vlib_buffer_t * b, ip6_address_t * src,
-	       ip6_address_t * dst, int proto)
-{
-  ip6_header_t *ip6h;
-  u16 payload_length;
-
-  /* make some room */
-  ip6h = vlib_buffer_push_uninit (b, sizeof (ip6_header_t));
-
-  ip6h->ip_version_traffic_class_and_flow_label =
-    clib_host_to_net_u32 (0x6 << 28);
-
-  /* calculate ip6 payload length */
-  payload_length = vlib_buffer_length_in_chain (vm, b);
-  payload_length -= sizeof (*ip6h);
-
-  ip6h->payload_length = clib_host_to_net_u16 (payload_length);
-
-  ip6h->hop_limit = 0xff;
-  ip6h->protocol = proto;
-  clib_memcpy (ip6h->src_address.as_u8, src->as_u8,
-	       sizeof (ip6h->src_address));
-  clib_memcpy (ip6h->dst_address.as_u8, dst->as_u8,
-	       sizeof (ip6h->src_address));
-
-  return ip6h;
-}
-
 void *
 pkt_push_ip (vlib_main_t * vm, vlib_buffer_t * b, ip_address_t * src,
 	     ip_address_t * dst, u32 proto)
@@ -210,12 +155,12 @@ pkt_push_ip (vlib_main_t * vm, vlib_buffer_t * b, ip_address_t * src,
   switch (ip_addr_version (src))
     {
     case IP4:
-      return pkt_push_ipv4 (vm, b, &ip_addr_v4 (src), &ip_addr_v4 (dst),
-			    proto);
+      return vlib_buffer_push_ip4 (vm, b, &ip_addr_v4 (src),
+				   &ip_addr_v4 (dst), proto);
       break;
     case IP6:
-      return pkt_push_ipv6 (vm, b, &ip_addr_v6 (src), &ip_addr_v6 (dst),
-			    proto);
+      return vlib_buffer_push_ip6 (vm, b, &ip_addr_v6 (src),
+				   &ip_addr_v6 (dst), proto);
       break;
     }
 
diff --git a/src/vnet/lisp-cp/packets.h b/src/vnet/lisp-cp/packets.h
index 212a1d78..f6da3bf4 100644
--- a/src/vnet/lisp-cp/packets.h
+++ b/src/vnet/lisp-cp/packets.h
@@ -26,51 +26,6 @@ void *pkt_push_udp_and_ip (vlib_main_t * vm, vlib_buffer_t * b, u16 sp,
 
 void *pkt_push_ecm_hdr (vlib_buffer_t * b);
 
-always_inline u8 *
-vlib_buffer_get_tail (vlib_buffer_t * b)
-{
-  return b->data + b->current_data + b->current_length;
-}
-
-always_inline void *
-vlib_buffer_put_uninit (vlib_buffer_t * b, u8 size)
-{
-  /* XXX should make sure there's enough space! */
-  void *p = vlib_buffer_get_tail (b);
-  b->current_length += size;
-  return p;
-}
-
-always_inline void *
-vlib_buffer_push_uninit (vlib_buffer_t * b, u8 size)
-{
-  /* XXX should make sure there's enough space! */
-  ASSERT (b->current_data >= size);
-  b->current_data -= size;
-  b->current_length += size;
-
-  return vlib_buffer_get_current (b);
-}
-
-always_inline void *
-vlib_buffer_make_headroom (vlib_buffer_t * b, u8 size)
-{
-  /* XXX should make sure there's enough space! */
-  b->current_data += size;
-  return vlib_buffer_get_current (b);
-}
-
-always_inline void *
-vlib_buffer_pull (vlib_buffer_t * b, u8 size)
-{
-  if (b->current_length < size)
-    return 0;
-
-  void *data = vlib_buffer_get_current (b);
-  vlib_buffer_advance (b, size);
-  return data;
-}
-
 /* *INDENT-ON* */
 
 /*
diff --git a/src/vnet/lisp-gpe/interface.c b/src/vnet/lisp-gpe/interface.c
index 13359277..292c7e6a 100644
--- a/src/vnet/lisp-gpe/interface.c
+++ b/src/vnet/lisp-gpe/interface.c
@@ -23,7 +23,7 @@
 #include 
 #include 
 #include 
-#include 
+#include 
 #include 
 #include 
 #include 
diff --git a/src/vnet/lisp-gpe/lisp_gpe.h b/src/vnet/lisp-gpe/lisp_gpe.h
index c898a7da..b5a50ec6 100644
--- a/src/vnet/lisp-gpe/lisp_gpe.h
+++ b/src/vnet/lisp-gpe/lisp_gpe.h
@@ -27,10 +27,12 @@
 #include 
 #include 
 #include 
-#include 
+#include 
 #include 
 #include 
 #include 
+#include 
+#include 
 
 /** IP4-UDP-LISP encap header */
 /* *INDENT-OFF* */
diff --git a/src/vnet/lisp-gpe/lisp_gpe_adjacency.c b/src/vnet/lisp-gpe/lisp_gpe_adjacency.c
index 65006b81..dbcf7134 100644
--- a/src/vnet/lisp-gpe/lisp_gpe_adjacency.c
+++ b/src/vnet/lisp-gpe/lisp_gpe_adjacency.c
@@ -25,6 +25,8 @@
 #include 
 #include 
 #include 
+#include 
+#include 
 
 /**
  * Memory pool of all adjacencies
diff --git a/src/vnet/session/application.c b/src/vnet/session/application.c
new file mode 100644
index 00000000..a561e7d1
--- /dev/null
+++ b/src/vnet/session/application.c
@@ -0,0 +1,343 @@
+/*
+ * Copyright (c) 2017 Cisco and/or its affiliates.
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at:
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#include 
+#include 
+
+/*
+ * Pool from which we allocate all applications
+ */
+static application_t *app_pool;
+
+/*
+ * Hash table of apps by api client index
+ */
+static uword *app_by_api_client_index;
+
+int
+application_api_queue_is_full (application_t * app)
+{
+  unix_shared_memory_queue_t *q;
+
+  /* builtin servers are always OK */
+  if (app->api_client_index == ~0)
+    return 0;
+
+  q = vl_api_client_index_to_input_queue (app->api_client_index);
+  if (!q)
+    return 1;
+
+  if (q->cursize == q->maxsize)
+    return 1;
+  return 0;
+}
+
+static void
+application_table_add (application_t * app)
+{
+  hash_set (app_by_api_client_index, app->api_client_index, app->index);
+}
+
+static void
+application_table_del (application_t * app)
+{
+  hash_unset (app_by_api_client_index, app->api_client_index);
+}
+
+application_t *
+application_lookup (u32 api_client_index)
+{
+  uword *p;
+  p = hash_get (app_by_api_client_index, api_client_index);
+  if (p)
+    return application_get (p[0]);
+
+  return 0;
+}
+
+void
+application_del (application_t * app)
+{
+  session_manager_main_t *smm = vnet_get_session_manager_main ();
+  api_main_t *am = &api_main;
+  void *oldheap;
+  session_manager_t *sm;
+
+  if (app->mode == APP_SERVER)
+    {
+      sm = session_manager_get (app->session_manager_index);
+      session_manager_del (smm, sm);
+    }
+
+  /* Free the event fifo in the /vpe-api shared-memory segment */
+  oldheap = svm_push_data_heap (am->vlib_rp);
+  if (app->event_queue)
+    unix_shared_memory_queue_free (app->event_queue);
+  svm_pop_heap (oldheap);
+
+  application_table_del (app);
+
+  pool_put (app_pool, app);
+}
+
+application_t *
+application_new (application_type_t type, session_type_t sst,
+		 u32 api_client_index, u32 flags, session_cb_vft_t * cb_fns)
+{
+  session_manager_main_t *smm = vnet_get_session_manager_main ();
+  api_main_t *am = &api_main;
+  application_t *app;
+  void *oldheap;
+  session_manager_t *sm;
+
+  pool_get (app_pool, app);
+  memset (app, 0, sizeof (*app));
+
+  /* Allocate event fifo in the /vpe-api shared-memory segment */
+  oldheap = svm_push_data_heap (am->vlib_rp);
+
+  /* Allocate server event queue */
+  app->event_queue =
+    unix_shared_memory_queue_init (128 /* nels $$$$ config */ ,
+				   sizeof (session_fifo_event_t),
+				   0 /* consumer pid */ ,
+				   0
+				   /* (do not) signal when queue non-empty */
+    );
+
+  svm_pop_heap (oldheap);
+
+  /* If a server, allocate session manager */
+  if (type == APP_SERVER)
+    {
+      pool_get (smm->session_managers, sm);
+      memset (sm, 0, sizeof (*sm));
+
+      app->session_manager_index = sm - smm->session_managers;
+    }
+  else if (type == APP_CLIENT)
+    {
+      /* Allocate connect session manager if needed */
+      if (smm->connect_manager_index[sst] == INVALID_INDEX)
+	connects_session_manager_init (smm, sst);
+      app->session_manager_index = smm->connect_manager_index[sst];
+    }
+
+  app->mode = type;
+  app->index = application_get_index (app);
+  app->session_type = sst;
+  app->api_client_index = api_client_index;
+  app->flags = flags;
+  app->cb_fns = *cb_fns;
+
+  /* Add app to lookup by api_client_index table */
+  application_table_add (app);
+
+  return app;
+}
+
+application_t *
+application_get (u32 index)
+{
+  return pool_elt_at_index (app_pool, index);
+}
+
+u32
+application_get_index (application_t * app)
+{
+  return app - app_pool;
+}
+
+int
+application_server_init (application_t * server, u32 segment_size,
+			 u32 add_segment_size, u32 rx_fifo_size,
+			 u32 tx_fifo_size, u8 ** segment_name)
+{
+  session_manager_main_t *smm = vnet_get_session_manager_main ();
+  session_manager_t *sm;
+  int rv;
+
+  sm = session_manager_get (server->session_manager_index);
+
+  /* Add first segment */
+  if ((rv = session_manager_add_first_segment (smm, sm, segment_size,
+					       segment_name)))
+    {
+      return rv;
+    }
+
+  /* Setup session manager */
+  sm->add_segment_size = add_segment_size;
+  sm->rx_fifo_size = rx_fifo_size;
+  sm->tx_fifo_size = tx_fifo_size;
+  sm->add_segment = sm->add_segment_size != 0;
+  return 0;
+}
+
+u8 *
+format_application_server (u8 * s, va_list * args)
+{
+  application_t *srv = va_arg (*args, application_t *);
+  int verbose = va_arg (*args, int);
+  vl_api_registration_t *regp;
+  stream_session_t *listener;
+  u8 *server_name, *str, *seg_name;
+  u32 segment_size;
+
+  if (srv == 0)
+    {
+      if (verbose)
+	s = format (s, "%-40s%-20s%-15s%-15s%-10s", "Connection", "Server",
+		    "Segment", "API Client", "Cookie");
+      else
+	s = format (s, "%-40s%-20s", "Connection", "Server");
+
+      return s;
+    }
+
+  regp = vl_api_client_index_to_registration (srv->api_client_index);
+  if (!regp)
+    server_name = format (0, "%s%c", regp->name, 0);
+  else
+    server_name = regp->name;
+
+  listener = stream_session_listener_get (srv->session_type,
+					  srv->session_index);
+  str = format (0, "%U", format_stream_session, listener, verbose);
+
+  session_manager_get_segment_info (listener->server_segment_index, &seg_name,
+				    &segment_size);
+  if (verbose)
+    {
+      s = format (s, "%-40s%-20s%-20s%-10d%-10d", str, server_name,
+		  seg_name, srv->api_client_index, srv->accept_cookie);
+    }
+  else
+    s = format (s, "%-40s%-20s", str, server_name);
+  return s;
+}
+
+u8 *
+format_application_client (u8 * s, va_list * args)
+{
+  application_t *client = va_arg (*args, application_t *);
+  int verbose = va_arg (*args, int);
+  stream_session_t *session;
+  u8 *str, *seg_name;
+  u32 segment_size;
+
+  if (client == 0)
+    {
+      if (verbose)
+	s =
+	  format (s, "%-40s%-20s%-10s", "Connection", "Segment",
+		  "API Client");
+      else
+	s = format (s, "%-40s", "Connection");
+
+      return s;
+    }
+
+  session = stream_session_get (client->session_index, client->thread_index);
+  str = format (0, "%U", format_stream_session, session, verbose);
+
+  session_manager_get_segment_info (session->server_segment_index, &seg_name,
+				    &segment_size);
+  if (verbose)
+    {
+      s = format (s, "%-40s%-20s%-10d%", str, seg_name,
+		  client->api_client_index);
+    }
+  else
+    s = format (s, "%-40s", str);
+  return s;
+}
+
+static clib_error_t *
+show_app_command_fn (vlib_main_t * vm, unformat_input_t * input,
+		     vlib_cli_command_t * cmd)
+{
+  application_t *app;
+  int do_server = 0;
+  int do_client = 0;
+  int verbose = 0;
+
+  while (unformat_check_input (input) != UNFORMAT_END_OF_INPUT)
+    {
+      if (unformat (input, "server"))
+	do_server = 1;
+      else if (unformat (input, "client"))
+	do_client = 1;
+      else if (unformat (input, "verbose"))
+	verbose = 1;
+      else
+	break;
+    }
+
+  if (do_server)
+    {
+      if (pool_elts (app_pool))
+	{
+	  vlib_cli_output (vm, "%U", format_application_server,
+			   0 /* header */ ,
+			   verbose);
+          /* *INDENT-OFF* */
+          pool_foreach (app, app_pool,
+          ({
+            if (app->mode == APP_SERVER)
+              vlib_cli_output (vm, "%U", format_application_server, app,
+                               verbose);
+          }));
+          /* *INDENT-ON* */
+	}
+      else
+	vlib_cli_output (vm, "No active server bindings");
+    }
+
+  if (do_client)
+    {
+      if (pool_elts (app_pool))
+	{
+	  vlib_cli_output (vm, "%U", format_application_client,
+			   0 /* header */ ,
+			   verbose);
+          /* *INDENT-OFF* */
+          pool_foreach (app, app_pool,
+          ({
+            if (app->mode == APP_CLIENT)
+              vlib_cli_output (vm, "%U", format_application_client, app,
+                               verbose);
+          }));
+          /* *INDENT-ON* */
+	}
+      else
+	vlib_cli_output (vm, "No active server bindings");
+    }
+
+  return 0;
+}
+
+VLIB_CLI_COMMAND (show_app_command, static) =
+{
+.path = "show app",.short_help =
+    "show app [server|client] [verbose]",.function = show_app_command_fn,};
+
+/*
+ * fd.io coding-style-patch-verification: ON
+ *
+ * Local Variables:
+ * eval: (c-set-style "gnu")
+ * End:
+ */
diff --git a/src/vnet/session/application.h b/src/vnet/session/application.h
new file mode 100644
index 00000000..027d6967
--- /dev/null
+++ b/src/vnet/session/application.h
@@ -0,0 +1,120 @@
+/*
+ * Copyright (c) 2017 Cisco and/or its affiliates.
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at:
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#ifndef SRC_VNET_SESSION_APPLICATION_H_
+#define SRC_VNET_SESSION_APPLICATION_H_
+
+#include 
+#include 
+
+typedef enum
+{
+  APP_SERVER,
+  APP_CLIENT
+} application_type_t;
+
+typedef struct _stream_session_cb_vft
+{
+  /** Notify server of new segment */
+  int (*add_segment_callback) (u32 api_client_index, const u8 * seg_name,
+			       u32 seg_size);
+
+  /** Notify server of newly accepted session */
+  int (*session_accept_callback) (stream_session_t * new_session);
+
+  /* Connection request callback */
+  int (*session_connected_callback) (u32 api_client_index,
+				     stream_session_t * s, u8 code);
+
+  /** Notify app that session is closing */
+  void (*session_disconnect_callback) (stream_session_t * s);
+
+  /** Notify app that session was reset */
+  void (*session_reset_callback) (stream_session_t * s);
+
+  /* Direct RX callback, for built-in servers */
+  int (*builtin_server_rx_callback) (stream_session_t * session);
+
+  /* Redirect connection to local server */
+  int (*redirect_connect_callback) (u32 api_client_index, void *mp);
+} session_cb_vft_t;
+
+typedef struct _application
+{
+  /** Index in server pool */
+  u32 index;
+
+  /** Flags */
+  u32 flags;
+
+  /** Binary API connection index, ~0 if internal */
+  u32 api_client_index;
+
+  /* */
+  u32 api_context;
+
+  /** Application listens for events on this svm queue */
+  unix_shared_memory_queue_t *event_queue;
+
+  /** Stream session type */
+  u8 session_type;
+
+  /* Stream server mode: accept or connect */
+  u8 mode;
+
+  u32 session_manager_index;
+
+  /*
+   * Bind/Listen specific
+   */
+
+  /** Accept cookie, for multiple session flavors ($$$ maybe) */
+  u32 accept_cookie;
+
+  /** Index of the listen session or connect session */
+  u32 session_index;
+
+  /** Session thread index for client connect sessions */
+  u32 thread_index;
+
+  /*
+   * Callbacks: shoulder-taps for the server/client
+   */
+  session_cb_vft_t cb_fns;
+} application_t;
+
+application_t *application_new (application_type_t type, session_type_t sst,
+				u32 api_client_index, u32 flags,
+				session_cb_vft_t * cb_fns);
+void application_del (application_t * app);
+application_t *application_get (u32 index);
+application_t *application_lookup (u32 api_client_index);
+u32 application_get_index (application_t * app);
+
+int
+application_server_init (application_t * server, u32 segment_size,
+			 u32 add_segment_size, u32 rx_fifo_size,
+			 u32 tx_fifo_size, u8 ** segment_name);
+int application_api_queue_is_full (application_t * app);
+
+#endif /* SRC_VNET_SESSION_APPLICATION_H_ */
+
+/*
+ * fd.io coding-style-patch-verification: ON
+ *
+ * Local Variables:
+ * eval: (c-set-style "gnu")
+ * End:
+ */
diff --git a/src/vnet/session/application_interface.c b/src/vnet/session/application_interface.c
new file mode 100644
index 00000000..0ea77fd8
--- /dev/null
+++ b/src/vnet/session/application_interface.c
@@ -0,0 +1,459 @@
+/*
+ * Copyright (c) 2016 Cisco and/or its affiliates.
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at:
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+#include 
+
+#include 
+#include 
+#include 
+#include 
+
+/** @file
+    VPP's application/session API bind/unbind/connect/disconnect calls
+*/
+
+static u8
+ip_is_zero (ip46_address_t * ip46_address, u8 is_ip4)
+{
+  if (is_ip4)
+    return (ip46_address->ip4.as_u32 == 0);
+  else
+    return (ip46_address->as_u64[0] == 0 && ip46_address->as_u64[1] == 0);
+}
+
+static u8
+ip_is_local (ip46_address_t * ip46_address, u8 is_ip4)
+{
+  fib_node_index_t fei;
+  fib_entry_flag_t flags;
+  fib_prefix_t prefix;
+
+  /* Check if requester is local */
+  if (is_ip4)
+    {
+      prefix.fp_len = 32;
+      prefix.fp_proto = FIB_PROTOCOL_IP4;
+    }
+  else
+    {
+      prefix.fp_len = 128;
+      prefix.fp_proto = FIB_PROTOCOL_IP6;
+    }
+
+  clib_memcpy (&prefix.fp_addr, ip46_address, sizeof (ip46_address));
+  fei = fib_table_lookup (0, &prefix);
+  flags = fib_entry_get_flags (fei);
+
+  return (flags & FIB_ENTRY_FLAG_LOCAL);
+}
+
+int
+api_parse_session_handle (u64 handle, u32 * session_index, u32 * thread_index)
+{
+  session_manager_main_t *smm = vnet_get_session_manager_main ();
+  stream_session_t *pool;
+
+  *thread_index = handle & 0xFFFFFFFF;
+  *session_index = handle >> 32;
+
+  if (*thread_index >= vec_len (smm->sessions))
+    return VNET_API_ERROR_INVALID_VALUE;
+
+  pool = smm->sessions[*thread_index];
+
+  if (pool_is_free_index (pool, *session_index))
+    return VNET_API_ERROR_INVALID_VALUE_2;
+
+  return 0;
+}
+
+int
+vnet_bind_i (u32 api_client_index, ip46_address_t * ip46, u16 port_host_order,
+	     session_type_t sst, u64 * options, session_cb_vft_t * cb_fns,
+	     application_t ** app, u32 * len_seg_name, char *seg_name)
+{
+  u8 *segment_name = 0;
+  application_t *server = 0;
+  stream_session_t *listener;
+  u8 is_ip4;
+
+  listener =
+    stream_session_lookup_listener (ip46,
+				    clib_host_to_net_u16 (port_host_order),
+				    sst);
+
+  if (listener)
+    return VNET_API_ERROR_ADDRESS_IN_USE;
+
+  if (application_lookup (api_client_index))
+    {
+      clib_warning ("Only one bind supported for now");
+      return VNET_API_ERROR_ADDRESS_IN_USE;
+    }
+
+  is_ip4 = SESSION_TYPE_IP4_UDP == sst || SESSION_TYPE_IP4_TCP == sst;
+  if (!ip_is_zero (ip46, is_ip4) && !ip_is_local (ip46, is_ip4))
+    return VNET_API_ERROR_INVALID_VALUE;
+
+  /* Allocate and initialize stream server */
+  server = application_new (APP_SERVER, sst, api_client_index,
+			    options[SESSION_OPTIONS_FLAGS], cb_fns);
+
+  application_server_init (server, options[SESSION_OPTIONS_SEGMENT_SIZE],
+			   options[SESSION_OPTIONS_ADD_SEGMENT_SIZE],
+			   options[SESSION_OPTIONS_RX_FIFO_SIZE],
+			   options[SESSION_OPTIONS_TX_FIFO_SIZE],
+			   &segment_name);
+
+  /* Setup listen path down to transport */
+  stream_session_start_listen (server->index, ip46, port_host_order);
+
+  /*
+   * Return values
+   */
+
+  ASSERT (vec_len (segment_name) <= 128);
+  *len_seg_name = vec_len (segment_name);
+  memcpy (seg_name, segment_name, *len_seg_name);
+  *app = server;
+
+  return 0;
+}
+
+int
+vnet_unbind_i (u32 api_client_index)
+{
+  application_t *server;
+
+  /*
+   * Find the stream_server_t corresponding to the api client
+   */
+  server = application_lookup (api_client_index);
+  if (!server)
+    return VNET_API_ERROR_INVALID_VALUE_2;
+
+  /* Clear the listener */
+  stream_session_stop_listen (server->index);
+  application_del (server);
+
+  return 0;
+}
+
+int
+vnet_connect_i (u32 api_client_index, u32 api_context, session_type_t sst,
+		ip46_address_t * ip46, u16 port, u64 * options, void *mp,
+		session_cb_vft_t * cb_fns)
+{
+  stream_session_t *listener;
+  application_t *server, *app;
+
+  /*
+   * Figure out if connecting to a local server
+   */
+  listener = stream_session_lookup_listener (ip46,
+					     clib_host_to_net_u16 (port),
+					     sst);
+  if (listener)
+    {
+      server = application_get (listener->app_index);
+
+      /*
+       * Server is willing to have a direct fifo connection created
+       * instead of going through the state machine, etc.
+       */
+      if (server->flags & SESSION_OPTIONS_FLAGS_USE_FIFO)
+	return server->cb_fns.
+	  redirect_connect_callback (server->api_client_index, mp);
+    }
+
+  /* Create client app */
+  app = application_new (APP_CLIENT, sst, api_client_index,
+			 options[SESSION_OPTIONS_FLAGS], cb_fns);
+
+  app->api_context = api_context;
+
+  /*
+   * Not connecting to a local server. Create regular session
+   */
+  stream_session_open (sst, ip46, port, app->index);
+
+  return 0;
+}
+
+/**
+ * unformat a vnet URI
+ *
+ * fifo://name
+ * tcp://ip46-addr:port
+ * udp://ip46-addr:port
+ *
+ * u8 ip46_address[16];
+ * u16  port_in_host_byte_order;
+ * stream_session_type_t sst;
+ * u8 *fifo_name;
+ *
+ * if (unformat (input, "%U", unformat_vnet_uri, &ip46_address,
+ *              &sst, &port, &fifo_name))
+ *  etc...
+ *
+ */
+uword
+unformat_vnet_uri (unformat_input_t * input, va_list * args)
+{
+  ip46_address_t *address = va_arg (*args, ip46_address_t *);
+  session_type_t *sst = va_arg (*args, session_type_t *);
+  u16 *port = va_arg (*args, u16 *);
+
+  if (unformat (input, "tcp://%U/%d", unformat_ip4_address, &address->ip4,
+		port))
+    {
+      *sst = SESSION_TYPE_IP4_TCP;
+      return 1;
+    }
+  if (unformat (input, "udp://%U/%d", unformat_ip4_address, &address->ip4,
+		port))
+    {
+      *sst = SESSION_TYPE_IP4_UDP;
+      return 1;
+    }
+  if (unformat (input, "udp://%U/%d", unformat_ip6_address, &address->ip6,
+		port))
+    {
+      *sst = SESSION_TYPE_IP6_UDP;
+      return 1;
+    }
+  if (unformat (input, "tcp://%U/%d", unformat_ip6_address, &address->ip6,
+		port))
+    {
+      *sst = SESSION_TYPE_IP6_TCP;
+      return 1;
+    }
+
+  return 0;
+}
+
+int
+parse_uri (char *uri, session_type_t * sst, ip46_address_t * addr,
+	   u16 * port_number_host_byte_order)
+{
+  unformat_input_t _input, *input = &_input;
+
+  /* Make sure */
+  uri = (char *) format (0, "%s%c", uri, 0);
+
+  /* Parse uri */
+  unformat_init_string (input, uri, strlen (uri));
+  if (!unformat (input, "%U", unformat_vnet_uri, addr, sst,
+		 port_number_host_byte_order))
+    {
+      unformat_free (input);
+      return VNET_API_ERROR_INVALID_VALUE;
+    }
+  unformat_free (input);
+
+  return 0;
+}
+
+int
+vnet_bind_uri (vnet_bind_args_t * a)
+{
+  application_t *server = 0;
+  u16 port_host_order;
+  session_type_t sst = SESSION_N_TYPES;
+  ip46_address_t ip46;
+  int rv;
+
+  memset (&ip46, 0, sizeof (ip46));
+  rv = parse_uri (a->uri, &sst, &ip46, &port_host_order);
+  if (rv)
+    return rv;
+
+  if ((rv = vnet_bind_i (a->api_client_index, &ip46, port_host_order, sst,
+			 a->options, a->session_cb_vft, &server,
+			 &a->segment_name_length, a->segment_name)))
+    return rv;
+
+  a->server_event_queue_address = (u64) server->event_queue;
+  return 0;
+}
+
+session_type_t
+session_type_from_proto_and_ip (session_api_proto_t proto, u8 is_ip4)
+{
+  if (proto == SESSION_PROTO_TCP)
+    {
+      if (is_ip4)
+	return SESSION_TYPE_IP4_TCP;
+      else
+	return SESSION_TYPE_IP6_TCP;
+    }
+  else
+    {
+      if (is_ip4)
+	return SESSION_TYPE_IP4_UDP;
+      else
+	return SESSION_TYPE_IP6_UDP;
+    }
+
+  return SESSION_N_TYPES;
+}
+
+int
+vnet_unbind_uri (char *uri, u32 api_client_index)
+{
+  u16 port_number_host_byte_order;
+  session_type_t sst = SESSION_N_TYPES;
+  ip46_address_t ip46_address;
+  stream_session_t *listener;
+  int rv;
+
+  rv = parse_uri (uri, &sst, &ip46_address, &port_number_host_byte_order);
+  if (rv)
+    return rv;
+
+  listener =
+    stream_session_lookup_listener (&ip46_address,
+				    clib_host_to_net_u16
+				    (port_number_host_byte_order), sst);
+
+  if (!listener)
+    return VNET_API_ERROR_ADDRESS_NOT_IN_USE;
+
+  /* External client? */
+  if (api_client_index != ~0)
+    {
+      ASSERT (vl_api_client_index_to_registration (api_client_index));
+    }
+
+  return vnet_unbind_i (api_client_index);
+}
+
+int
+vnet_connect_uri (vnet_connect_args_t * a)
+{
+  ip46_address_t ip46_address;
+  u16 port;
+  session_type_t sst;
+  application_t *app;
+  int rv;
+
+  app = application_lookup (a->api_client_index);
+  if (app)
+    {
+      clib_warning ("Already have a connect from this app");
+      return VNET_API_ERROR_INVALID_VALUE_2;
+    }
+
+  /* Parse uri */
+  rv = parse_uri (a->uri, &sst, &ip46_address, &port);
+  if (rv)
+    return rv;
+
+  return vnet_connect_i (a->api_client_index, a->api_context, sst,
+			 &ip46_address, port, a->options, a->mp,
+			 a->session_cb_vft);
+}
+
+int
+vnet_disconnect_session (u32 client_index, u32 session_index,
+			 u32 thread_index)
+{
+  stream_session_t *session;
+
+  session = stream_session_get (session_index, thread_index);
+  stream_session_disconnect (session);
+
+  return 0;
+}
+
+
+int
+vnet_bind (vnet_bind_args_t * a)
+{
+  application_t *server = 0;
+  session_type_t sst = SESSION_N_TYPES;
+  int rv;
+
+  sst = session_type_from_proto_and_ip (a->proto, a->tep.is_ip4);
+  if ((rv = vnet_bind_i (a->api_client_index, &a->tep.ip, a->tep.port, sst,
+			 a->options, a->session_cb_vft, &server,
+			 &a->segment_name_length, a->segment_name)))
+    return rv;
+
+  a->server_event_queue_address = (u64) server->event_queue;
+  a->handle = (u64) a->tep.vrf << 32 | (u64) server->session_index;
+  return 0;
+}
+
+int
+vnet_unbind (vnet_unbind_args_t * a)
+{
+  application_t *server;
+
+  if (a->api_client_index != ~0)
+    {
+      ASSERT (vl_api_client_index_to_registration (a->api_client_index));
+    }
+
+  /* Make sure this is the right one */
+  server = application_lookup (a->api_client_index);
+  ASSERT (server->session_index == (0xFFFFFFFF & a->handle));
+
+  /* TODO use handle to disambiguate namespaces/vrfs */
+  return vnet_unbind_i (a->api_client_index);
+}
+
+int
+vnet_connect (vnet_connect_args_t * a)
+{
+  session_type_t sst;
+  application_t *app;
+
+  app = application_lookup (a->api_client_index);
+  if (app)
+    {
+      clib_warning ("Already have a connect from this app");
+      return VNET_API_ERROR_INVALID_VALUE_2;
+    }
+
+  sst = session_type_from_proto_and_ip (a->proto, a->tep.is_ip4);
+  return vnet_connect_i (a->api_client_index, a->api_context, sst, &a->tep.ip,
+			 a->tep.port, a->options, a->mp, a->session_cb_vft);
+}
+
+int
+vnet_disconnect (vnet_disconnect_args_t * a)
+{
+  stream_session_t *session;
+  u32 session_index, thread_index;
+
+  if (api_parse_session_handle (a->handle, &session_index, &thread_index))
+    {
+      clib_warning ("Invalid handle");
+      return -1;
+    }
+
+  session = stream_session_get (session_index, thread_index);
+  stream_session_disconnect (session);
+
+  return 0;
+}
+
+/*
+ * fd.io coding-style-patch-verification: ON
+ *
+ * Local Variables:
+ * eval: (c-set-style "gnu")
+ * End:
+ */
diff --git a/src/vnet/session/application_interface.h b/src/vnet/session/application_interface.h
new file mode 100644
index 00000000..8d87c067
--- /dev/null
+++ b/src/vnet/session/application_interface.h
@@ -0,0 +1,136 @@
+/*
+ * Copyright (c) 2016 Cisco and/or its affiliates.
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at:
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+#ifndef __included_uri_h__
+#define __included_uri_h__
+
+#include 
+#include 
+#include 
+#include 
+#include 
+#include 
+
+typedef enum _session_api_proto
+{
+  SESSION_PROTO_TCP,
+  SESSION_PROTO_UDP
+} session_api_proto_t;
+
+typedef struct _vnet_bind_args_t
+{
+  union
+  {
+    char *uri;
+    struct
+    {
+      transport_endpoint_t tep;
+      session_api_proto_t proto;
+    };
+  };
+
+  u32 api_client_index;
+  u64 *options;
+  session_cb_vft_t *session_cb_vft;
+
+  /*
+   * Results
+   */
+  char *segment_name;
+  u32 segment_name_length;
+  u64 server_event_queue_address;
+  u64 handle;
+} vnet_bind_args_t;
+
+typedef struct _vnet_unbind_args_t
+{
+  union
+  {
+    char *uri;
+    u64 handle;
+  };
+  u32 api_client_index;
+} vnet_unbind_args_t;
+
+typedef struct _vnet_connect_args
+{
+  union
+  {
+    char *uri;
+    struct
+    {
+      transport_endpoint_t tep;
+      session_api_proto_t proto;
+    };
+  };
+  u32 api_client_index;
+  u32 api_context;
+  u64 *options;
+  session_cb_vft_t *session_cb_vft;
+
+  /* Used for redirects */
+  void *mp;
+} vnet_connect_args_t;
+
+typedef struct _vnet_disconnect_args_t
+{
+  u64 handle;
+  u32 api_client_index;
+} vnet_disconnect_args_t;
+
+/* Bind / connect options */
+typedef enum
+{
+  SESSION_OPTIONS_FLAGS,
+  SESSION_OPTIONS_SEGMENT_SIZE,
+  SESSION_OPTIONS_ADD_SEGMENT_SIZE,
+  SESSION_OPTIONS_RX_FIFO_SIZE,
+  SESSION_OPTIONS_TX_FIFO_SIZE,
+  SESSION_OPTIONS_ACCEPT_COOKIE,
+  SESSION_OPTIONS_N_OPTIONS
+} session_options_index_t;
+
+/** Server can handle delegated connect requests from local clients */
+#define SESSION_OPTIONS_FLAGS_USE_FIFO	(1<<0)
+
+/** Server wants vpp to add segments when out of memory for fifos */
+#define SESSION_OPTIONS_FLAGS_ADD_SEGMENT   (1<<1)
+
+#define VNET_CONNECT_REDIRECTED	123
+
+int vnet_bind_uri (vnet_bind_args_t *);
+int vnet_unbind_uri (char *uri, u32 api_client_index);
+int vnet_connect_uri (vnet_connect_args_t * a);
+int
+vnet_disconnect_session (u32 client_index, u32 session_index,
+			 u32 thread_index);
+
+int vnet_bind (vnet_bind_args_t * a);
+int vnet_connect (vnet_connect_args_t * a);
+int vnet_unbind (vnet_unbind_args_t * a);
+int vnet_disconnect (vnet_disconnect_args_t * a);
+
+int
+api_parse_session_handle (u64 handle, u32 * session_index,
+			  u32 * thread_index);
+
+#endif /* __included_uri_h__ */
+
+/*
+ * fd.io coding-style-patch-verification: ON
+ *
+ * Local Variables:
+ * eval: (c-set-style "gnu")
+ * End:
+ */
diff --git a/src/vnet/session/hashes.c b/src/vnet/session/hashes.c
new file mode 100644
index 00000000..1808dd73
--- /dev/null
+++ b/src/vnet/session/hashes.c
@@ -0,0 +1,28 @@
+/*
+ * Copyright (c) 2016 Cisco and/or its affiliates.
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at:
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+/** Generate typed init functions for multiple hash table styles... */
+
+#include 
+#include 
+
+#include 
+
+#undef __included_bihash_template_h__
+
+#include 
+#include 
+
+#include 
diff --git a/src/vnet/session/node.c b/src/vnet/session/node.c
new file mode 100644
index 00000000..e467f4e9
--- /dev/null
+++ b/src/vnet/session/node.c
@@ -0,0 +1,435 @@
+/*
+ * Copyright (c) 2017 Cisco and/or its affiliates.
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at:
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#include 
+#include 
+#include 
+#include 
+
+#include 
+
+#include 
+#include 
+#include 
+#include 
+
+#include 
+#include 
+#include 
+
+vlib_node_registration_t session_queue_node;
+
+typedef struct
+{
+  u32 session_index;
+  u32 server_thread_index;
+} session_queue_trace_t;
+
+/* packet trace format function */
+static u8 *
+format_session_queue_trace (u8 * s, va_list * args)
+{
+  CLIB_UNUSED (vlib_main_t * vm) = va_arg (*args, vlib_main_t *);
+  CLIB_UNUSED (vlib_node_t * node) = va_arg (*args, vlib_node_t *);
+  session_queue_trace_t *t = va_arg (*args, session_queue_trace_t *);
+
+  s = format (s, "SESSION_QUEUE: session index %d, server thread index %d",
+	      t->session_index, t->server_thread_index);
+  return s;
+}
+
+vlib_node_registration_t session_queue_node;
+
+#define foreach_session_queue_error                 \
+_(TX, "Packets transmitted")                    \
+_(TIMER, "Timer events")
+
+typedef enum
+{
+#define _(sym,str) SESSION_QUEUE_ERROR_##sym,
+  foreach_session_queue_error
+#undef _
+    SESSION_QUEUE_N_ERROR,
+} session_queue_error_t;
+
+static char *session_queue_error_strings[] = {
+#define _(sym,string) string,
+  foreach_session_queue_error
+#undef _
+};
+
+static u32 session_type_to_next[] = {
+  SESSION_QUEUE_NEXT_TCP_IP4_OUTPUT,
+  SESSION_QUEUE_NEXT_IP4_LOOKUP,
+  SESSION_QUEUE_NEXT_TCP_IP6_OUTPUT,
+  SESSION_QUEUE_NEXT_IP6_LOOKUP,
+};
+
+always_inline int
+session_fifo_rx_i (vlib_main_t * vm, vlib_node_runtime_t * node,
+		   session_manager_main_t * smm, session_fifo_event_t * e0,
+		   stream_session_t * s0, u32 thread_index, int *n_tx_packets,
+		   u8 peek_data)
+{
+  u32 n_trace = vlib_get_trace_count (vm, node);
+  u32 left_to_snd0, max_len_to_snd0, len_to_deq0, n_bufs, snd_space0;
+  u32 n_frame_bytes, n_frames_per_evt;
+  transport_connection_t *tc0;
+  transport_proto_vft_t *transport_vft;
+  u32 next_index, next0, *to_next, n_left_to_next, bi0;
+  vlib_buffer_t *b0;
+  u32 rx_offset;
+  u16 snd_mss0;
+  u8 *data0;
+  int i;
+
+  next_index = next0 = session_type_to_next[s0->session_type];
+
+  transport_vft = session_get_transport_vft (s0->session_type);
+  tc0 = transport_vft->get_connection (s0->connection_index, thread_index);
+
+  /* Make sure we have space to send and there's something to dequeue */
+  snd_space0 = transport_vft->send_space (tc0);
+  snd_mss0 = transport_vft->send_mss (tc0);
+
+  if (snd_space0 == 0 || svm_fifo_max_dequeue (s0->server_tx_fifo) == 0
+      || snd_mss0 == 0)
+    return 0;
+
+  ASSERT (e0->enqueue_length > 0);
+
+  /* Ensure we're not writing more than transport window allows */
+  max_len_to_snd0 = clib_min (e0->enqueue_length, snd_space0);
+
+  if (peek_data)
+    {
+      /* Offset in rx fifo from where to peek data  */
+      rx_offset = transport_vft->rx_fifo_offset (tc0);
+    }
+
+  /* TODO check if transport is willing to send len_to_snd0
+   * bytes (Nagle) */
+
+  n_frame_bytes = snd_mss0 * VLIB_FRAME_SIZE;
+  n_frames_per_evt = ceil ((double) max_len_to_snd0 / n_frame_bytes);
+
+  n_bufs = vec_len (smm->tx_buffers[thread_index]);
+  left_to_snd0 = max_len_to_snd0;
+  for (i = 0; i < n_frames_per_evt; i++)
+    {
+      /* Make sure we have at least one full frame of buffers ready */
+      if (PREDICT_FALSE (n_bufs < VLIB_FRAME_SIZE))
+	{
+	  vec_validate (smm->tx_buffers[thread_index],
+			n_bufs + VLIB_FRAME_SIZE - 1);
+	  n_bufs +=
+	    vlib_buffer_alloc (vm, &smm->tx_buffers[thread_index][n_bufs],
+			       VLIB_FRAME_SIZE);
+
+	  /* buffer shortage
+	   * XXX 0.9 because when debugging we might not get a full frame */
+	  if (PREDICT_FALSE (n_bufs < 0.9 * VLIB_FRAME_SIZE))
+	    {
+	      /* Keep track of how much we've dequeued and exit */
+	      e0->enqueue_length -= max_len_to_snd0 - left_to_snd0;
+	      return -1;
+	    }
+
+	  _vec_len (smm->tx_buffers[thread_index]) = n_bufs;
+	}
+
+      vlib_get_next_frame (vm, node, next_index, to_next, n_left_to_next);
+      while (left_to_snd0 && n_left_to_next)
+	{
+	  /* Get free buffer */
+	  n_bufs--;
+	  bi0 = smm->tx_buffers[thread_index][n_bufs];
+	  _vec_len (smm->tx_buffers[thread_index]) = n_bufs;
+
+	  b0 = vlib_get_buffer (vm, bi0);
+	  b0->error = 0;
+	  b0->flags = VLIB_BUFFER_TOTAL_LENGTH_VALID
+	    | VNET_BUFFER_LOCALLY_ORIGINATED;
+	  b0->current_data = 0;
+
+	  /* RX on the local interface. tx in default fib */
+	  vnet_buffer (b0)->sw_if_index[VLIB_RX] = 0;
+	  vnet_buffer (b0)->sw_if_index[VLIB_TX] = (u32) ~ 0;
+
+	  /* usual speculation, or the enqueue_x1 macro will barf */
+	  to_next[0] = bi0;
+	  to_next += 1;
+	  n_left_to_next -= 1;
+
+	  VLIB_BUFFER_TRACE_TRAJECTORY_INIT (b0);
+	  if (PREDICT_FALSE (n_trace > 0))
+	    {
+	      session_queue_trace_t *t0;
+	      vlib_trace_buffer (vm, node, next_index, b0,
+				 1 /* follow_chain */ );
+	      vlib_set_trace_count (vm, node, --n_trace);
+	      t0 = vlib_add_trace (vm, node, b0, sizeof (*t0));
+	      t0->session_index = s0->session_index;
+	      t0->server_thread_index = s0->thread_index;
+	    }
+
+	  if (1)
+	    {
+	      ELOG_TYPE_DECLARE (e) =
+	      {
+	      .format = "evt-dequeue: id %d length %d",.format_args =
+		  "i4i4",};
+	      struct
+	      {
+		u32 data[2];
+	      } *ed;
+	      ed = ELOG_DATA (&vm->elog_main, e);
+	      ed->data[0] = e0->event_id;
+	      ed->data[1] = e0->enqueue_length;
+	    }
+
+	  len_to_deq0 = (left_to_snd0 < snd_mss0) ? left_to_snd0 : snd_mss0;
+
+	  /* Make room for headers */
+	  data0 = vlib_buffer_make_headroom (b0, MAX_HDRS_LEN);
+
+	  /* Dequeue the data
+	   * TODO 1) peek instead of dequeue
+	   *      2) buffer chains */
+	  if (peek_data)
+	    {
+	      int n_bytes_read;
+	      n_bytes_read = svm_fifo_peek (s0->server_tx_fifo, s0->pid,
+					    rx_offset, len_to_deq0, data0);
+	      if (n_bytes_read < 0)
+		goto dequeue_fail;
+
+	      /* Keep track of progress locally, transport is also supposed to
+	       * increment it independently when pushing header */
+	      rx_offset += n_bytes_read;
+	    }
+	  else
+	    {
+	      if (svm_fifo_dequeue_nowait (s0->server_tx_fifo, s0->pid,
+					   len_to_deq0, data0) < 0)
+		goto dequeue_fail;
+	    }
+
+	  b0->current_length = len_to_deq0;
+
+	  /* Ask transport to push header */
+	  transport_vft->push_header (tc0, b0);
+
+	  left_to_snd0 -= len_to_deq0;
+	  *n_tx_packets = *n_tx_packets + 1;
+
+	  vlib_validate_buffer_enqueue_x1 (vm, node, next_index,
+					   to_next, n_left_to_next,
+					   bi0, next0);
+	}
+      vlib_put_next_frame (vm, node, next_index, n_left_to_next);
+    }
+
+  /* If we couldn't dequeue all bytes store progress */
+  if (max_len_to_snd0 < e0->enqueue_length)
+    {
+      e0->enqueue_length -= max_len_to_snd0;
+      vec_add1 (smm->evts_partially_read[thread_index], *e0);
+    }
+  return 0;
+
+dequeue_fail:
+  /* Can't read from fifo. Store event rx progress, save as partially read,
+   * return buff to free list and return  */
+  e0->enqueue_length -= max_len_to_snd0 - left_to_snd0;
+  vec_add1 (smm->evts_partially_read[thread_index], *e0);
+
+  to_next -= 1;
+  n_left_to_next += 1;
+  _vec_len (smm->tx_buffers[thread_index]) += 1;
+
+  clib_warning ("dequeue fail");
+  return 0;
+}
+
+int
+session_fifo_rx_peek (vlib_main_t * vm, vlib_node_runtime_t * node,
+		      session_manager_main_t * smm, session_fifo_event_t * e0,
+		      stream_session_t * s0, u32 thread_index, int *n_tx_pkts)
+{
+  return session_fifo_rx_i (vm, node, smm, e0, s0, thread_index, n_tx_pkts,
+			    1);
+}
+
+int
+session_fifo_rx_dequeue (vlib_main_t * vm, vlib_node_runtime_t * node,
+			 session_manager_main_t * smm,
+			 session_fifo_event_t * e0, stream_session_t * s0,
+			 u32 thread_index, int *n_tx_pkts)
+{
+  return session_fifo_rx_i (vm, node, smm, e0, s0, thread_index, n_tx_pkts,
+			    0);
+}
+
+static uword
+session_queue_node_fn (vlib_main_t * vm, vlib_node_runtime_t * node,
+		       vlib_frame_t * frame)
+{
+  session_manager_main_t *smm = vnet_get_session_manager_main ();
+  session_fifo_event_t *my_fifo_events, *e;
+  u32 n_to_dequeue;
+  unix_shared_memory_queue_t *q;
+  int n_tx_packets = 0;
+  u32 my_thread_index = vm->cpu_index;
+  int i, rv;
+
+  /*
+   *  Update TCP time
+   */
+  tcp_update_time (vlib_time_now (vm), my_thread_index);
+
+  /*
+   * Get vpp queue events
+   */
+  q = smm->vpp_event_queues[my_thread_index];
+  if (PREDICT_FALSE (q == 0))
+    return 0;
+
+  /* min number of events we can dequeue without blocking */
+  n_to_dequeue = q->cursize;
+  if (n_to_dequeue == 0)
+    return 0;
+
+  my_fifo_events = smm->fifo_events[my_thread_index];
+
+  /* If we didn't manage to process previous events try going
+   * over them again without dequeuing new ones.
+   * XXX: Block senders to sessions that can't keep up */
+  if (vec_len (my_fifo_events) >= 100)
+    goto skip_dequeue;
+
+  /* See you in the next life, don't be late */
+  if (pthread_mutex_trylock (&q->mutex))
+    return 0;
+
+  for (i = 0; i < n_to_dequeue; i++)
+    {
+      vec_add2 (my_fifo_events, e, 1);
+      unix_shared_memory_queue_sub_raw (q, (u8 *) e);
+    }
+
+  /* The other side of the connection is not polling */
+  if (q->cursize < (q->maxsize / 8))
+    (void) pthread_cond_broadcast (&q->condvar);
+  pthread_mutex_unlock (&q->mutex);
+
+  smm->fifo_events[my_thread_index] = my_fifo_events;
+
+skip_dequeue:
+
+  for (i = 0; i < n_to_dequeue; i++)
+    {
+      svm_fifo_t *f0;		/* $$$ prefetch 1 ahead maybe */
+      stream_session_t *s0;
+      u32 server_session_index0, server_thread_index0;
+      session_fifo_event_t *e0;
+
+      e0 = &my_fifo_events[i];
+      f0 = e0->fifo;
+      server_session_index0 = f0->server_session_index;
+      server_thread_index0 = f0->server_thread_index;
+
+      /* $$$ add multiple event queues, per vpp worker thread */
+      ASSERT (server_thread_index0 == my_thread_index);
+
+      s0 = pool_elt_at_index (smm->sessions[my_thread_index],
+			      server_session_index0);
+
+      ASSERT (s0->thread_index == my_thread_index);
+
+      switch (e0->event_type)
+	{
+	case FIFO_EVENT_SERVER_TX:
+	  /* Spray packets in per session type frames, since they go to
+	   * different nodes */
+	  rv = (smm->session_rx_fns[s0->session_type]) (vm, node, smm, e0, s0,
+							my_thread_index,
+							&n_tx_packets);
+	  if (rv < 0)
+	    goto done;
+
+	  break;
+
+	default:
+	  clib_warning ("unhandled event type %d", e0->event_type);
+	}
+    }
+
+done:
+
+  /* Couldn't process all events. Probably out of buffers */
+  if (PREDICT_FALSE (i < n_to_dequeue))
+    {
+      session_fifo_event_t *partially_read =
+	smm->evts_partially_read[my_thread_index];
+      vec_add (partially_read, &my_fifo_events[i], n_to_dequeue - i);
+      vec_free (my_fifo_events);
+      smm->fifo_events[my_thread_index] = partially_read;
+      smm->evts_partially_read[my_thread_index] = 0;
+    }
+  else
+    {
+      vec_free (smm->fifo_events[my_thread_index]);
+      smm->fifo_events[my_thread_index] =
+	smm->evts_partially_read[my_thread_index];
+      smm->evts_partially_read[my_thread_index] = 0;
+    }
+
+  vlib_node_increment_counter (vm, session_queue_node.index,
+			       SESSION_QUEUE_ERROR_TX, n_tx_packets);
+
+  return n_tx_packets;
+}
+
+/* *INDENT-OFF* */
+VLIB_REGISTER_NODE (session_queue_node) =
+{
+  .function = session_queue_node_fn,
+  .name = "session-queue",
+  .format_trace = format_session_queue_trace,
+  .type = VLIB_NODE_TYPE_INPUT,
+  .n_errors = ARRAY_LEN (session_queue_error_strings),
+  .error_strings = session_queue_error_strings,
+  .n_next_nodes = SESSION_QUEUE_N_NEXT,
+  /* .state = VLIB_NODE_STATE_DISABLED, enable on-demand? */
+  /* edit / add dispositions here */
+  .next_nodes =
+  {
+      [SESSION_QUEUE_NEXT_DROP] = "error-drop",
+      [SESSION_QUEUE_NEXT_IP4_LOOKUP] = "ip4-lookup",
+      [SESSION_QUEUE_NEXT_IP6_LOOKUP] = "ip6-lookup",
+      [SESSION_QUEUE_NEXT_TCP_IP4_OUTPUT] = "tcp4-output",
+      [SESSION_QUEUE_NEXT_TCP_IP6_OUTPUT] = "tcp6-output",
+  },
+};
+/* *INDENT-ON* */
+
+/*
+ * fd.io coding-style-patch-verification: ON
+ *
+ * Local Variables:
+ * eval: (c-set-style "gnu")
+ * End:
+ */
diff --git a/src/vnet/session/session.api b/src/vnet/session/session.api
new file mode 100644
index 00000000..a7b28c1d
--- /dev/null
+++ b/src/vnet/session/session.api
@@ -0,0 +1,429 @@
+/*
+ * Copyright (c) 2015-2016 Cisco and/or its affiliates.
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at:
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+ 
+ /** \brief Bind to a given URI
+    @param client_index - opaque cookie to identify the sender
+    @param context - sender context, to match reply w/ request
+    @param accept_cookie - sender accept cookie, to identify this bind flavor
+    @param uri - a URI, e.g. "tcp://0.0.0.0/0/80" [ipv4]
+                 "tcp://::/0/80" [ipv6] etc.
+    @param options - socket options, fifo sizes, etc.
+*/
+define bind_uri {
+  u32 client_index;
+  u32 context;
+  u32 accept_cookie;
+  u32 initial_segment_size;
+  u8 uri[128];
+  u64 options[16];
+};
+
+/** \brief Unbind a given URI
+    @param client_index - opaque cookie to identify the sender
+    @param context - sender context, to match reply w/ request
+    @param uri - a URI, e.g. "tcp://0.0.0.0/0/80" [ipv4]
+                 "tcp://::/0/80" [ipv6], etc.
+    @param options - socket options, fifo sizes, etc.
+*/
+define unbind_uri {
+  u32 client_index;
+  u32 context;
+  u8 uri[128];
+};
+
+/** \brief Connect to a given URI
+    @param client_index - opaque cookie to identify the sender
+    @param context - sender context, to match reply w/ request
+    @param accept_cookie - sender accept cookie, to identify this bind flavor
+    @param uri - a URI, e.g. "tcp4://0.0.0.0/0/80"
+                 "tcp6://::/0/80" [ipv6], etc.
+    @param options - socket options, fifo sizes, etc.
+*/
+define connect_uri {
+  u32 client_index;
+  u32 context;
+  u8 uri[128];
+  u64 client_queue_address;
+  u64 options[16];
+};
+
+/** \brief Bind reply
+    @param context - sender context, to match reply w/ request
+    @param retval - return code for the request
+    @param event_queue_address - vpp event queue address or 0 if this 
+                                 connection shouldn't send events
+    @param segment_name_length - length of segment name 
+    @param segment_name - name of segment client needs to attach to
+*/
+define bind_uri_reply {
+    u32 context;
+    i32 retval;
+    u64 server_event_queue_address;
+    u8 segment_name_length;
+    u32 segment_size;
+    u8 segment_name[128];
+};
+
+/** \brief unbind reply
+    @param context - sender context, to match reply w/ request
+    @param retval - return code for the request
+*/
+define unbind_uri_reply {
+    u32 context;
+    i32 retval;
+};
+
+/** \brief vpp->client, connect reply
+    @param context - sender context, to match reply w/ request
+    @param retval - return code for the request
+    @param server_rx_fifo - rx (vpp -> vpp-client) fifo address 
+    @param server_tx_fifo - tx (vpp-client -> vpp) fifo address 
+    @param session_index - session index;
+    @param session_thread_index - session thread index
+    @param session_type - session thread type
+    @param vpp_event_queue_address - vpp's event queue address
+    @param client_event_queue_address - client's event queue address
+    @param segment_name_length - non-zero if the client needs to attach to 
+                                 the fifo segment
+    @param segment_name - set if the client needs to attach to the segment
+*/
+define connect_uri_reply {
+  u32 context;
+  i32 retval;
+  u64 server_rx_fifo;
+  u64 server_tx_fifo;
+  u32 session_index;
+  u32 session_thread_index;
+  u8 session_type;
+  u64 client_event_queue_address;
+  u64 vpp_event_queue_address;
+  u32 segment_size;
+  u8 segment_name_length;
+  u8 segment_name[128];
+};
+
+/** \brief vpp->client, please map an additional shared memory segment
+    @param context - sender context, to match reply w/ request
+    @param segment_name - 
+*/
+define map_another_segment {
+    u32 client_index;
+    u32 context;
+    u32 segment_size;
+    u8 segment_name[128];
+};
+
+/** \brief client->vpp
+    @param context - sender context, to match reply w/ request
+    @param retval - return code for the request
+*/
+define map_another_segment_reply {
+    u32 context;
+    i32 retval;
+};
+
+/** \brief vpp->client, accept this session
+    @param context - sender context, to match reply w/ request
+    @param accept_cookie - tells client which bind flavor just occurred
+    @param rx_fifo_address - rx (vpp -> vpp-client) fifo address 
+    @param tx_fifo_address - tx (vpp-client -> vpp) fifo address 
+    @param session_index - index of new session
+    @param session_thread_index - thread index of new session
+    @param vpp_event_queue_address - vpp's event queue address
+    @param session_type - type of session
+    
+*/
+define accept_session {
+  u32 client_index;
+  u32 context;
+  u32 accept_cookie;
+  u64 server_rx_fifo;
+  u64 server_tx_fifo;
+  u32 session_index;
+  u32 session_thread_index;
+  u64 vpp_event_queue_address;
+  u8 session_type;
+};
+
+/** \brief client->vpp, reply to an accept message
+    @param context - sender context, to match reply w/ request
+    @param retval - return code for the request
+    @param session_index - session index from accept_session / connect_reply
+    @param session_thread_index - thread index from accept_session /
+                                  connect_reply
+*/
+define accept_session_reply {
+  u32 context;
+  i32 retval;
+  u8 session_type;
+  u8 session_thread_index;
+  u32 session_index;
+};
+
+/** \brief bidirectional disconnect API
+    @param client_index - opaque cookie to identify the sender
+                          client to vpp direction only
+    @param context - sender context, to match reply w/ request
+    @param session_index - cookie #1 from accept_session / connect_reply
+    @param session_thread_index - cookie #2
+*/
+define disconnect_session {
+  u32 client_index;
+  u32 context;
+  u32 session_index;
+  u32 session_thread_index;
+};
+
+/** \brief bidirectional disconnect reply API
+    @param client_index - opaque cookie to identify the sender
+                          client to vpp direction only
+    @param context - sender context, to match reply w/ request
+    @param retval - return code for the request
+    @param session_index - session index from accept_session / connect_reply
+    @param session_thread_index - thread index from accept_session /
+                                  connect_reply
+*/
+define disconnect_session_reply {
+  u32 client_index;
+  u32 context;
+  i32 retval;
+  u32 session_index;
+  u32 session_thread_index;
+};
+
+/** \brief vpp->client reset session API
+    @param client_index - opaque cookie to identify the sender
+                          client to vpp direction only
+    @param context - sender context, to match reply w/ request
+    @param session_index - session index from accept_session / connect_reply
+    @param session_thread_index - thread index from accept_session /
+                                  connect_reply
+*/
+define reset_session {
+  u32 client_index;
+  u32 context;
+  u32 session_index;
+  u32 session_thread_index;
+};
+
+/** \brief client->vpp reset session reply
+    @param client_index - opaque cookie to identify the sender
+                          client to vpp direction only
+    @param context - sender context, to match reply w/ request
+    @param retval - return code for the request
+    @param session_index - session index from accept_session / connect_reply
+    @param session_thread_index - thread index from accept_session /
+                                  connect_reply
+*/
+define reset_session_reply {
+  u32 client_index;
+  u32 context;
+  i32 retval;
+  u32 session_index;
+  u32 session_thread_index;
+};
+
+/** \brief Bind to an ip:port pair for a given transport protocol
+    @param client_index - opaque cookie to identify the sender
+    @param context - sender context, to match reply w/ request
+    @param vrf - bind namespace
+    @param is_ip4 - flag that is 1 if ip address family is IPv4
+    @param ip - ip address
+    @param port - port 
+    @param proto - protocol 0 - TCP 1 - UDP
+    @param options - socket options, fifo sizes, etc.
+*/
+define bind_sock {
+  u32 client_index;
+  u32 context;
+  u32 vrf;
+  u8 is_ip4;
+  u8 ip[16];
+  u16 port;
+  u8 proto;
+  u64 options[16];
+};
+
+/** \brief Unbind 
+    @param client_index - opaque cookie to identify the sender
+    @param context - sender context, to match reply w/ request
+    @param handle - bind handle obtained from bind reply
+*/
+define unbind_sock {
+  u32 client_index;
+  u32 context;
+  u64 handle;
+};
+
+/** \brief Connect to a remote peer
+    @param client_index - opaque cookie to identify the sender
+    @param context - sender context, to match reply w/ request
+    @param vrf - connection namespace
+    @param is_ip4 - flag that is 1 if ip address family is IPv4
+    @param ip - ip address
+    @param port - port 
+    @param proto - protocol 0 - TCP 1 - UDP
+    @param client_queue_address - client's API queue address. Non-zero when 
+                                  used to perform redirects
+    @param options - socket options, fifo sizes, etc.
+*/
+define connect_sock {
+  u32 client_index;
+  u32 context;
+  u32 vrf;
+  u8 is_ip4;
+  u8 ip[16];
+  u16 port;
+  u8 proto;
+  u64 client_queue_address;
+  u64 options[16];
+};
+
+/** \brief Bind reply
+    @param context - sender context, to match reply w/ request
+    @param handle - bind handle
+    @param retval - return code for the request
+    @param event_queue_address - vpp event queue address or 0 if this 
+                                 connection shouldn't send events
+    @param segment_name_length - length of segment name 
+    @param segment_name - name of segment client needs to attach to
+*/
+define bind_sock_reply {
+  u32 context;
+  u64 handle;
+  i32 retval;
+  u64 server_event_queue_address;
+  u32 segment_size;
+  u8 segment_name_length;
+  u8 segment_name[128];
+};
+
+/** \brief unbind reply
+    @param context - sender context, to match reply w/ request
+    @param retval - return code for the request
+*/
+define unbind_sock_reply {
+  u32 context;
+  i32 retval;
+};
+
+/** \brief vpp/server->client, connect reply
+    @param context - sender context, to match reply w/ request
+    @param retval - return code for the request
+    @param handle - connection handle
+    @param server_rx_fifo - rx (vpp -> vpp-client) fifo address 
+    @param server_tx_fifo - tx (vpp-client -> vpp) fifo address 
+    @param vpp_event_queue_address - vpp's event queue address
+    @param client_event_queue_address - client's event queue address
+    @param segment_name_length - non-zero if the client needs to attach to 
+                                 the fifo segment
+    @param segment_name - set if the client needs to attach to the segment
+*/
+define connect_sock_reply {
+  u32 context;
+  i32 retval;
+  u64 handle;
+  u64 server_rx_fifo;
+  u64 server_tx_fifo;
+  u64 client_event_queue_address;
+  u64 vpp_event_queue_address;
+  u32 segment_size;
+  u8 segment_name_length;
+  u8 segment_name[128];
+};
+
+/** \brief bidirectional disconnect API
+    @param client_index - opaque cookie to identify the sender
+                          client to vpp direction only
+    @param context - sender context, to match reply w/ request
+    @param handle - session handle obtained through accept/connect
+*/
+define disconnect_sock {
+  u32 client_index;
+  u32 context;
+  u64 handle;
+};
+
+/** \brief bidirectional disconnect reply API
+    @param client_index - opaque cookie to identify the sender
+                          client to vpp direction only
+    @param client_context - sender context, to match reply w/ request
+    @param handle - session handle obtained through accept/connect
+*/
+define disconnect_sock_reply {
+  u32 client_index;
+  u32 context;
+  i32 retval;
+  u64 handle;
+};
+
+/** \brief vpp->client, accept this session
+    @param context - sender context, to match reply w/ request
+    @param accept_cookie - tells client which bind flavor just occurred
+    @param handle - session handle obtained through accept/connect
+    @param rx_fifo_address - rx (vpp -> vpp-client) fifo address 
+    @param tx_fifo_address - tx (vpp-client -> vpp) fifo address 
+    @param vpp_event_queue_address - vpp's event queue address
+*/
+define accept_sock {
+  u32 client_index;
+  u32 context;
+  u32 accept_cookie;
+  u64 handle;
+  u64 server_rx_fifo;
+  u64 server_tx_fifo;
+  u64 vpp_event_queue_address;
+};
+
+/** \brief client->vpp, reply to an accept message
+    @param context - sender context, to match reply w/ request
+    @param retval - return code for the request
+    @param handle - session handle obtained through accept/connect
+*/
+define accept_sock_reply {
+  u32 context;
+  i32 retval;
+  u64 handle;
+};
+
+/** \brief vpp->client reset session API
+    @param client_index - opaque cookie to identify the sender
+                          client to vpp direction only
+    @param context - sender context, to match reply w/ request
+    @param handle - session handle obtained through accept/connect
+*/
+define reset_sock {
+  u32 client_index;
+  u32 context;
+  u64 handle;
+};
+
+/** \brief client->vpp reset session reply
+    @param client_index - opaque cookie to identify the sender
+                          client to vpp direction only
+    @param context - sender context, to match reply w/ request
+    @param handle - session handle obtained through accept/connect
+*/
+define reset_sock_reply {
+  u32 client_index;
+  u32 context;
+  i32 retval;
+  u64 handle;
+};
+/*
+ * Local Variables:
+ * eval: (c-set-style "gnu")
+ * End:
+ */
\ No newline at end of file
diff --git a/src/vnet/session/session.c b/src/vnet/session/session.c
new file mode 100644
index 00000000..539da613
--- /dev/null
+++ b/src/vnet/session/session.c
@@ -0,0 +1,1286 @@
+/*
+ * Copyright (c) 2017 Cisco and/or its affiliates.
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at:
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+/**
+ * @file
+ * @brief Session and session manager
+ */
+
+#include 
+#include 
+#include 
+#include 
+#include 
+
+/**
+ * Per-type vector of transport protocol virtual function tables
+ */
+static transport_proto_vft_t *tp_vfts;
+
+session_manager_main_t session_manager_main;
+
+/*
+ * Session lookup key; (src-ip, dst-ip, src-port, dst-port, session-type)
+ * Value: (owner thread index << 32 | session_index);
+ */
+static void
+stream_session_table_add_for_tc (u8 sst, transport_connection_t * tc,
+				 u64 value)
+{
+  session_manager_main_t *smm = &session_manager_main;
+  session_kv4_t kv4;
+  session_kv6_t kv6;
+
+  switch (sst)
+    {
+    case SESSION_TYPE_IP4_UDP:
+    case SESSION_TYPE_IP4_TCP:
+      make_v4_ss_kv_from_tc (&kv4, tc);
+      kv4.value = value;
+      clib_bihash_add_del_16_8 (&smm->v4_session_hash, &kv4, 1 /* is_add */ );
+      break;
+    case SESSION_TYPE_IP6_UDP:
+    case SESSION_TYPE_IP6_TCP:
+      make_v6_ss_kv_from_tc (&kv6, tc);
+      kv6.value = value;
+      clib_bihash_add_del_48_8 (&smm->v6_session_hash, &kv6, 1 /* is_add */ );
+      break;
+    default:
+      clib_warning ("Session type not supported");
+      ASSERT (0);
+    }
+}
+
+void
+stream_session_table_add (session_manager_main_t * smm, stream_session_t * s,
+			  u64 value)
+{
+  transport_connection_t *tc;
+
+  tc = tp_vfts[s->session_type].get_connection (s->connection_index,
+						s->thread_index);
+  stream_session_table_add_for_tc (s->session_type, tc, value);
+}
+
+static void
+stream_session_half_open_table_add (u8 sst, transport_connection_t * tc,
+				    u64 value)
+{
+  session_manager_main_t *smm = &session_manager_main;
+  session_kv4_t kv4;
+  session_kv6_t kv6;
+
+  switch (sst)
+    {
+    case SESSION_TYPE_IP4_UDP:
+    case SESSION_TYPE_IP4_TCP:
+      make_v4_ss_kv_from_tc (&kv4, tc);
+      kv4.value = value;
+      clib_bihash_add_del_16_8 (&smm->v4_half_open_hash, &kv4,
+				1 /* is_add */ );
+      break;
+    case SESSION_TYPE_IP6_UDP:
+    case SESSION_TYPE_IP6_TCP:
+      make_v6_ss_kv_from_tc (&kv6, tc);
+      kv6.value = value;
+      clib_bihash_add_del_48_8 (&smm->v6_half_open_hash, &kv6,
+				1 /* is_add */ );
+      break;
+    default:
+      clib_warning ("Session type not supported");
+      ASSERT (0);
+    }
+}
+
+static int
+stream_session_table_del_for_tc (session_manager_main_t * smm, u8 sst,
+				 transport_connection_t * tc)
+{
+  session_kv4_t kv4;
+  session_kv6_t kv6;
+
+  switch (sst)
+    {
+    case SESSION_TYPE_IP4_UDP:
+    case SESSION_TYPE_IP4_TCP:
+      make_v4_ss_kv_from_tc (&kv4, tc);
+      return clib_bihash_add_del_16_8 (&smm->v4_session_hash, &kv4,
+				       0 /* is_add */ );
+      break;
+    case SESSION_TYPE_IP6_UDP:
+    case SESSION_TYPE_IP6_TCP:
+      make_v6_ss_kv_from_tc (&kv6, tc);
+      return clib_bihash_add_del_48_8 (&smm->v6_session_hash, &kv6,
+				       0 /* is_add */ );
+      break;
+    default:
+      clib_warning ("Session type not supported");
+      ASSERT (0);
+    }
+
+  return 0;
+}
+
+static int
+stream_session_table_del (session_manager_main_t * smm, stream_session_t * s)
+{
+  transport_connection_t *ts;
+
+  ts = tp_vfts[s->session_type].get_connection (s->connection_index,
+						s->thread_index);
+  return stream_session_table_del_for_tc (smm, s->session_type, ts);
+}
+
+static void
+stream_session_half_open_table_del (session_manager_main_t * smm, u8 sst,
+				    transport_connection_t * tc)
+{
+  session_kv4_t kv4;
+  session_kv6_t kv6;
+
+  switch (sst)
+    {
+    case SESSION_TYPE_IP4_UDP:
+    case SESSION_TYPE_IP4_TCP:
+      make_v4_ss_kv_from_tc (&kv4, tc);
+      clib_bihash_add_del_16_8 (&smm->v4_half_open_hash, &kv4,
+				0 /* is_add */ );
+      break;
+    case SESSION_TYPE_IP6_UDP:
+    case SESSION_TYPE_IP6_TCP:
+      make_v6_ss_kv_from_tc (&kv6, tc);
+      clib_bihash_add_del_48_8 (&smm->v6_half_open_hash, &kv6,
+				0 /* is_add */ );
+      break;
+    default:
+      clib_warning ("Session type not supported");
+      ASSERT (0);
+    }
+}
+
+stream_session_t *
+stream_session_lookup_listener4 (ip4_address_t * lcl, u16 lcl_port, u8 proto)
+{
+  session_manager_main_t *smm = &session_manager_main;
+  session_kv4_t kv4;
+  int rv;
+
+  make_v4_listener_kv (&kv4, lcl, lcl_port, proto);
+  rv = clib_bihash_search_inline_16_8 (&smm->v4_session_hash, &kv4);
+  if (rv == 0)
+    return pool_elt_at_index (smm->listen_sessions[proto], (u32) kv4.value);
+
+  /* Zero out the lcl ip */
+  kv4.key[0] = 0;
+  rv = clib_bihash_search_inline_16_8 (&smm->v4_session_hash, &kv4);
+  if (rv == 0)
+    return pool_elt_at_index (smm->listen_sessions[proto], kv4.value);
+
+  return 0;
+}
+
+/** Looks up a session based on the 5-tuple passed as argument.
+ *
+ * First it tries to find an established session, if this fails, it tries
+ * finding a listener session if this fails, it tries a lookup with a
+ * wildcarded local source (listener bound to all interfaces)
+ */
+stream_session_t *
+stream_session_lookup4 (ip4_address_t * lcl, ip4_address_t * rmt,
+			u16 lcl_port, u16 rmt_port, u8 proto,
+			u32 my_thread_index)
+{
+  session_manager_main_t *smm = &session_manager_main;
+  session_kv4_t kv4;
+  int rv;
+
+  /* Lookup session amongst established ones */
+  make_v4_ss_kv (&kv4, lcl, rmt, lcl_port, rmt_port, proto);
+  rv = clib_bihash_search_inline_16_8 (&smm->v4_session_hash, &kv4);
+  if (rv == 0)
+    return stream_session_get_tsi (kv4.value, my_thread_index);
+
+  /* If nothing is found, check if any listener is available */
+  return stream_session_lookup_listener4 (lcl, lcl_port, proto);
+}
+
+stream_session_t *
+stream_session_lookup_listener6 (ip6_address_t * lcl, u16 lcl_port, u8 proto)
+{
+  session_manager_main_t *smm = &session_manager_main;
+  session_kv6_t kv6;
+  int rv;
+
+  make_v6_listener_kv (&kv6, lcl, lcl_port, proto);
+  rv = clib_bihash_search_inline_48_8 (&smm->v6_session_hash, &kv6);
+  if (rv == 0)
+    return pool_elt_at_index (smm->listen_sessions[proto], kv6.value);
+
+  /* Zero out the lcl ip */
+  kv6.key[0] = kv6.key[1] = 0;
+  rv = clib_bihash_search_inline_48_8 (&smm->v6_session_hash, &kv6);
+  if (rv == 0)
+    return pool_elt_at_index (smm->listen_sessions[proto], kv6.value);
+
+  return 0;
+}
+
+/* Looks up a session based on the 5-tuple passed as argument.
+ * First it tries to find an established session, if this fails, it tries
+ * finding a listener session if this fails, it tries a lookup with a
+ * wildcarded local source (listener bound to all interfaces) */
+stream_session_t *
+stream_session_lookup6 (ip6_address_t * lcl, ip6_address_t * rmt,
+			u16 lcl_port, u16 rmt_port, u8 proto,
+			u32 my_thread_index)
+{
+  session_manager_main_t *smm = vnet_get_session_manager_main ();
+  session_kv6_t kv6;
+  int rv;
+
+  make_v6_ss_kv (&kv6, lcl, rmt, lcl_port, rmt_port, proto);
+  rv = clib_bihash_search_inline_48_8 (&smm->v6_session_hash, &kv6);
+  if (rv == 0)
+    return stream_session_get_tsi (kv6.value, my_thread_index);
+
+  /* If nothing is found, check if any listener is available */
+  return stream_session_lookup_listener6 (lcl, lcl_port, proto);
+}
+
+stream_session_t *
+stream_session_lookup_listener (ip46_address_t * lcl, u16 lcl_port, u8 proto)
+{
+  switch (proto)
+    {
+    case SESSION_TYPE_IP4_UDP:
+    case SESSION_TYPE_IP4_TCP:
+      return stream_session_lookup_listener4 (&lcl->ip4, lcl_port, proto);
+      break;
+    case SESSION_TYPE_IP6_UDP:
+    case SESSION_TYPE_IP6_TCP:
+      return stream_session_lookup_listener6 (&lcl->ip6, lcl_port, proto);
+      break;
+    }
+  return 0;
+}
+
+static u64
+stream_session_half_open_lookup (session_manager_main_t * smm,
+				 ip46_address_t * lcl, ip46_address_t * rmt,
+				 u16 lcl_port, u16 rmt_port, u8 proto)
+{
+  session_kv4_t kv4;
+  session_kv6_t kv6;
+  int rv;
+
+  switch (proto)
+    {
+    case SESSION_TYPE_IP4_UDP:
+    case SESSION_TYPE_IP4_TCP:
+      make_v4_ss_kv (&kv4, &lcl->ip4, &rmt->ip4, lcl_port, rmt_port, proto);
+      rv = clib_bihash_search_inline_16_8 (&smm->v4_half_open_hash, &kv4);
+
+      if (rv == 0)
+	return kv4.value;
+
+      return (u64) ~ 0;
+      break;
+    case SESSION_TYPE_IP6_UDP:
+    case SESSION_TYPE_IP6_TCP:
+      make_v6_ss_kv (&kv6, &lcl->ip6, &rmt->ip6, lcl_port, rmt_port, proto);
+      rv = clib_bihash_search_inline_48_8 (&smm->v6_half_open_hash, &kv6);
+
+      if (rv == 0)
+	return kv6.value;
+
+      return (u64) ~ 0;
+      break;
+    }
+  return 0;
+}
+
+transport_connection_t *
+stream_session_lookup_transport4 (session_manager_main_t * smm,
+				  ip4_address_t * lcl, ip4_address_t * rmt,
+				  u16 lcl_port, u16 rmt_port, u8 proto,
+				  u32 my_thread_index)
+{
+  session_kv4_t kv4;
+  stream_session_t *s;
+  int rv;
+
+  /* Lookup session amongst established ones */
+  make_v4_ss_kv (&kv4, lcl, rmt, lcl_port, rmt_port, proto);
+  rv = clib_bihash_search_inline_16_8 (&smm->v4_session_hash, &kv4);
+  if (rv == 0)
+    {
+      s = stream_session_get_tsi (kv4.value, my_thread_index);
+
+      return tp_vfts[s->session_type].get_connection (s->connection_index,
+						      my_thread_index);
+    }
+
+  /* If nothing is found, check if any listener is available */
+  s = stream_session_lookup_listener4 (lcl, lcl_port, proto);
+  if (s)
+    return tp_vfts[s->session_type].get_listener (s->connection_index);
+
+  /* Finally, try half-open connections */
+  rv = clib_bihash_search_inline_16_8 (&smm->v4_half_open_hash, &kv4);
+  if (rv == 0)
+    return tp_vfts[proto].get_half_open (kv4.value & 0xFFFFFFFF);
+
+  return 0;
+}
+
+transport_connection_t *
+stream_session_lookup_transport6 (session_manager_main_t * smm,
+				  ip6_address_t * lcl, ip6_address_t * rmt,
+				  u16 lcl_port, u16 rmt_port, u8 proto,
+				  u32 my_thread_index)
+{
+  stream_session_t *s;
+  session_kv6_t kv6;
+  int rv;
+
+  make_v6_ss_kv (&kv6, lcl, rmt, lcl_port, rmt_port, proto);
+  rv = clib_bihash_search_inline_48_8 (&smm->v6_session_hash, &kv6);
+  if (rv == 0)
+    {
+      s = stream_session_get_tsi (kv6.value, my_thread_index);
+
+      return tp_vfts[s->session_type].get_connection (s->connection_index,
+						      my_thread_index);
+    }
+
+  /* If nothing is found, check if any listener is available */
+  s = stream_session_lookup_listener6 (lcl, lcl_port, proto);
+  if (s)
+    return tp_vfts[s->session_type].get_listener (s->connection_index);
+
+  /* Finally, try half-open connections */
+  rv = clib_bihash_search_inline_48_8 (&smm->v6_half_open_hash, &kv6);
+  if (rv == 0)
+    return tp_vfts[s->session_type].get_half_open (kv6.value & 0xFFFFFFFF);
+
+  return 0;
+}
+
+/**
+ * Allocate vpp event queue (once) per worker thread
+ */
+void
+vpp_session_event_queue_allocate (session_manager_main_t * smm,
+				  u32 thread_index)
+{
+  api_main_t *am = &api_main;
+  void *oldheap;
+
+  if (smm->vpp_event_queues[thread_index] == 0)
+    {
+      /* Allocate event fifo in the /vpe-api shared-memory segment */
+      oldheap = svm_push_data_heap (am->vlib_rp);
+
+      smm->vpp_event_queues[thread_index] =
+	unix_shared_memory_queue_init (2048 /* nels $$$$ config */ ,
+				       sizeof (session_fifo_event_t),
+				       0 /* consumer pid */ ,
+				       0
+				       /* (do not) send signal when queue non-empty */
+	);
+
+      svm_pop_heap (oldheap);
+    }
+}
+
+void
+session_manager_get_segment_info (u32 index, u8 ** name, u32 * size)
+{
+  svm_fifo_segment_private_t *s;
+  s = svm_fifo_get_segment (index);
+  *name = s->h->segment_name;
+  *size = s->ssvm.ssvm_size;
+}
+
+always_inline int
+session_manager_add_segment_i (session_manager_main_t * smm,
+			       session_manager_t * sm,
+			       u32 segment_size, u8 * segment_name)
+{
+  svm_fifo_segment_create_args_t _ca, *ca = &_ca;
+  int rv;
+
+  memset (ca, 0, sizeof (*ca));
+
+  ca->segment_name = (char *) segment_name;
+  ca->segment_size = segment_size;
+
+  rv = svm_fifo_segment_create (ca);
+  if (rv)
+    {
+      clib_warning ("svm_fifo_segment_create ('%s', %d) failed",
+		    ca->segment_name, ca->segment_size);
+      vec_free (segment_name);
+      return -1;
+    }
+
+  vec_add1 (sm->segment_indices, ca->new_segment_index);
+
+  return 0;
+}
+
+static int
+session_manager_add_segment (session_manager_main_t * smm,
+			     session_manager_t * sm)
+{
+  u8 *segment_name;
+  svm_fifo_segment_create_args_t _ca, *ca = &_ca;
+  u32 add_segment_size;
+  u32 default_segment_size = 128 << 10;
+
+  memset (ca, 0, sizeof (*ca));
+  segment_name = format (0, "%d-%d%c", getpid (),
+			 smm->unique_segment_name_counter++, 0);
+  add_segment_size =
+    sm->add_segment_size ? sm->add_segment_size : default_segment_size;
+
+  return session_manager_add_segment_i (smm, sm, add_segment_size,
+					segment_name);
+}
+
+int
+session_manager_add_first_segment (session_manager_main_t * smm,
+				   session_manager_t * sm, u32 segment_size,
+				   u8 ** segment_name)
+{
+  svm_fifo_segment_create_args_t _ca, *ca = &_ca;
+  memset (ca, 0, sizeof (*ca));
+  *segment_name = format (0, "%d-%d%c", getpid (),
+			  smm->unique_segment_name_counter++, 0);
+  return session_manager_add_segment_i (smm, sm, segment_size, *segment_name);
+}
+
+void
+session_manager_del (session_manager_main_t * smm, session_manager_t * sm)
+{
+  u32 *deleted_sessions = 0;
+  u32 *deleted_thread_indices = 0;
+  int i, j;
+
+  /* Across all fifo segments used by the server */
+  for (j = 0; j < vec_len (sm->segment_indices); j++)
+    {
+      svm_fifo_segment_private_t *fifo_segment;
+      svm_fifo_t **fifos;
+      /* Vector of fifos allocated in the segment */
+      fifo_segment = svm_fifo_get_segment (sm->segment_indices[j]);
+      fifos = (svm_fifo_t **) fifo_segment->h->fifos;
+
+      /*
+       * Remove any residual sessions from the session lookup table
+       * Don't bother deleting the individual fifos, we're going to
+       * throw away the fifo segment in a minute.
+       */
+      for (i = 0; i < vec_len (fifos); i++)
+	{
+	  svm_fifo_t *fifo;
+	  u32 session_index, thread_index;
+	  stream_session_t *session;
+
+	  fifo = fifos[i];
+	  session_index = fifo->server_session_index;
+	  thread_index = fifo->server_thread_index;
+
+	  session = pool_elt_at_index (smm->sessions[thread_index],
+				       session_index);
+
+	  /* Add to the deleted_sessions vector (once!) */
+	  if (!session->is_deleted)
+	    {
+	      session->is_deleted = 1;
+	      vec_add1 (deleted_sessions,
+			session - smm->sessions[thread_index]);
+	      vec_add1 (deleted_thread_indices, thread_index);
+	    }
+	}
+
+      for (i = 0; i < vec_len (deleted_sessions); i++)
+	{
+	  stream_session_t *session;
+
+	  session =
+	    pool_elt_at_index (smm->sessions[deleted_thread_indices[i]],
+			       deleted_sessions[i]);
+
+	  /* Instead of directly removing the session call disconnect */
+	  stream_session_disconnect (session);
+
+	  /*
+	     stream_session_table_del (smm, session);
+	     pool_put(smm->sessions[deleted_thread_indices[i]], session);
+	   */
+	}
+
+      vec_reset_length (deleted_sessions);
+      vec_reset_length (deleted_thread_indices);
+
+      /* Instead of removing the segment, test when removing the session if
+       * the segment can be removed
+       */
+      /* svm_fifo_segment_delete (fifo_segment); */
+    }
+
+  vec_free (deleted_sessions);
+  vec_free (deleted_thread_indices);
+}
+
+int
+session_manager_allocate_session_fifos (session_manager_main_t * smm,
+					session_manager_t * sm,
+					svm_fifo_t ** server_rx_fifo,
+					svm_fifo_t ** server_tx_fifo,
+					u32 * fifo_segment_index,
+					u8 * added_a_segment)
+{
+  svm_fifo_segment_private_t *fifo_segment;
+  u32 fifo_size, default_fifo_size = 8192 /* TODO config */ ;
+  int i;
+
+  *added_a_segment = 0;
+
+  /* Allocate svm fifos */
+  ASSERT (vec_len (sm->segment_indices));
+
+again:
+  for (i = 0; i < vec_len (sm->segment_indices); i++)
+    {
+      *fifo_segment_index = sm->segment_indices[i];
+      fifo_segment = svm_fifo_get_segment (*fifo_segment_index);
+
+      fifo_size = sm->rx_fifo_size;
+      fifo_size = (fifo_size == 0) ? default_fifo_size : fifo_size;
+      *server_rx_fifo = svm_fifo_segment_alloc_fifo (fifo_segment, fifo_size);
+
+      fifo_size = sm->tx_fifo_size;
+      fifo_size = (fifo_size == 0) ? default_fifo_size : fifo_size;
+      *server_tx_fifo = svm_fifo_segment_alloc_fifo (fifo_segment, fifo_size);
+
+      if (*server_rx_fifo == 0)
+	{
+	  /* This would be very odd, but handle it... */
+	  if (*server_tx_fifo != 0)
+	    {
+	      svm_fifo_segment_free_fifo (fifo_segment, *server_tx_fifo);
+	      *server_tx_fifo = 0;
+	    }
+	  continue;
+	}
+      if (*server_tx_fifo == 0)
+	{
+	  if (*server_rx_fifo != 0)
+	    {
+	      svm_fifo_segment_free_fifo (fifo_segment, *server_rx_fifo);
+	      *server_rx_fifo = 0;
+	    }
+	  continue;
+	}
+      break;
+    }
+
+  /* See if we're supposed to create another segment */
+  if (*server_rx_fifo == 0)
+    {
+      if (sm->add_segment)
+	{
+	  if (*added_a_segment)
+	    {
+	      clib_warning ("added a segment, still cant allocate a fifo");
+	      return SESSION_ERROR_NEW_SEG_NO_SPACE;
+	    }
+
+	  if (session_manager_add_segment (smm, sm))
+	    return VNET_API_ERROR_URI_FIFO_CREATE_FAILED;
+
+	  *added_a_segment = 1;
+	  goto again;
+	}
+      else
+	return SESSION_ERROR_NO_SPACE;
+    }
+  return 0;
+}
+
+int
+stream_session_create_i (session_manager_main_t * smm, application_t * app,
+			 transport_connection_t * tc,
+			 stream_session_t ** ret_s)
+{
+  int rv;
+  svm_fifo_t *server_rx_fifo = 0, *server_tx_fifo = 0;
+  u32 fifo_segment_index;
+  u32 pool_index, seg_size;
+  stream_session_t *s;
+  u64 value;
+  u32 thread_index = tc->thread_index;
+  session_manager_t *sm;
+  u8 segment_added;
+  u8 *seg_name;
+
+  sm = session_manager_get (app->session_manager_index);
+
+  /* Check the API queue */
+  if (app->mode == APP_SERVER && application_api_queue_is_full (app))
+    return SESSION_ERROR_API_QUEUE_FULL;
+
+  if ((rv = session_manager_allocate_session_fifos (smm, sm, &server_rx_fifo,
+						    &server_tx_fifo,
+						    &fifo_segment_index,
+						    &segment_added)))
+    return rv;
+
+  if (segment_added && app->mode == APP_SERVER)
+    {
+      /* Send an API message to the external server, to map new segment */
+      ASSERT (app->cb_fns.add_segment_callback);
+
+      session_manager_get_segment_info (fifo_segment_index, &seg_name,
+					&seg_size);
+      if (app->cb_fns.add_segment_callback (app->api_client_index, seg_name,
+					    seg_size))
+	return VNET_API_ERROR_URI_FIFO_CREATE_FAILED;
+    }
+
+  /* Create the session */
+  pool_get (smm->sessions[thread_index], s);
+  memset (s, 0, sizeof (*s));
+
+  /* Initialize backpointers */
+  pool_index = s - smm->sessions[thread_index];
+  server_rx_fifo->server_session_index = pool_index;
+  server_rx_fifo->server_thread_index = thread_index;
+
+  server_tx_fifo->server_session_index = pool_index;
+  server_tx_fifo->server_thread_index = thread_index;
+
+  s->server_rx_fifo = server_rx_fifo;
+  s->server_tx_fifo = server_tx_fifo;
+
+  /* Initialize state machine, such as it is... */
+  s->session_type = app->session_type;
+  s->session_state = SESSION_STATE_CONNECTING;
+  s->app_index = application_get_index (app);
+  s->server_segment_index = fifo_segment_index;
+  s->thread_index = thread_index;
+  s->session_index = pool_index;
+
+  /* Attach transport to session */
+  s->connection_index = tc->c_index;
+
+  /* Attach session to transport */
+  tc->s_index = s->session_index;
+
+  /* Add to the main lookup table */
+  value = (((u64) thread_index) << 32) | (u64) s->session_index;
+  stream_session_table_add_for_tc (app->session_type, tc, value);
+
+  *ret_s = s;
+
+  return 0;
+}
+
+/*
+ * Enqueue data for delivery to session peer. Does not notify peer of enqueue
+ * event but on request can queue notification events for later delivery by
+ * calling stream_server_flush_enqueue_events().
+ *
+ * @param tc Transport connection which is to be enqueued data
+ * @param data Data to be enqueued
+ * @param len Length of data to be enqueued
+ * @param queue_event Flag to indicate if peer is to be notified or if event
+ *                    is to be queued. The former is useful when more data is
+ *                    enqueued and only one event is to be generated.
+ * @return Number of bytes enqueued or a negative value if enqueueing failed.
+ */
+int
+stream_session_enqueue_data (transport_connection_t * tc, u8 * data, u16 len,
+			     u8 queue_event)
+{
+  stream_session_t *s;
+  int enqueued;
+
+  s = stream_session_get (tc->s_index, tc->thread_index);
+
+  /* Make sure there's enough space left. We might've filled the pipes */
+  if (PREDICT_FALSE (len > svm_fifo_max_enqueue (s->server_rx_fifo)))
+    return -1;
+
+  enqueued = svm_fifo_enqueue_nowait (s->server_rx_fifo, s->pid, len, data);
+
+  if (queue_event)
+    {
+      /* Queue RX event on this fifo. Eventually these will need to be flushed
+       * by calling stream_server_flush_enqueue_events () */
+      session_manager_main_t *smm = vnet_get_session_manager_main ();
+      u32 thread_index = s->thread_index;
+      u32 my_enqueue_epoch = smm->current_enqueue_epoch[thread_index];
+
+      if (s->enqueue_epoch != my_enqueue_epoch)
+	{
+	  s->enqueue_epoch = my_enqueue_epoch;
+	  vec_add1 (smm->session_indices_to_enqueue_by_thread[thread_index],
+		    s - smm->sessions[thread_index]);
+	}
+    }
+
+  return enqueued;
+}
+
+/** Check if we have space in rx fifo to push more bytes */
+u8
+stream_session_no_space (transport_connection_t * tc, u32 thread_index,
+			 u16 data_len)
+{
+  stream_session_t *s = stream_session_get (tc->c_index, thread_index);
+
+  if (PREDICT_FALSE (s->session_state != SESSION_STATE_READY))
+    return 1;
+
+  if (data_len > svm_fifo_max_enqueue (s->server_rx_fifo))
+    return 1;
+
+  return 0;
+}
+
+u32
+stream_session_peek_bytes (transport_connection_t * tc, u8 * buffer,
+			   u32 offset, u32 max_bytes)
+{
+  stream_session_t *s = stream_session_get (tc->s_index, tc->thread_index);
+  return svm_fifo_peek (s->server_tx_fifo, s->pid, offset, max_bytes, buffer);
+}
+
+u32
+stream_session_dequeue_drop (transport_connection_t * tc, u32 max_bytes)
+{
+  stream_session_t *s = stream_session_get (tc->s_index, tc->thread_index);
+  return svm_fifo_dequeue_drop (s->server_tx_fifo, s->pid, max_bytes);
+}
+
+/**
+ * Notify session peer that new data has been enqueued.
+ *
+ * @param s Stream session for which the event is to be generated.
+ * @param block Flag to indicate if call should block if event queue is full.
+ *
+ * @return 0 on succes or negative number if failed to send notification.
+ */
+static int
+stream_session_enqueue_notify (stream_session_t * s, u8 block)
+{
+  application_t *app;
+  session_fifo_event_t evt;
+  unix_shared_memory_queue_t *q;
+  static u32 serial_number;
+
+  if (PREDICT_FALSE (s->session_state == SESSION_STATE_CLOSED))
+    return 0;
+
+  /* Get session's server */
+  app = application_get (s->app_index);
+
+  /* Fabricate event */
+  evt.fifo = s->server_rx_fifo;
+  evt.event_type = FIFO_EVENT_SERVER_RX;
+  evt.event_id = serial_number++;
+  evt.enqueue_length = svm_fifo_max_dequeue (s->server_rx_fifo);
+
+  /* Add event to server's event queue */
+  q = app->event_queue;
+
+  /* Based on request block (or not) for lack of space */
+  if (block || PREDICT_TRUE (q->cursize < q->maxsize))
+    unix_shared_memory_queue_add (app->event_queue, (u8 *) & evt,
+				  0 /* do wait for mutex */ );
+  else
+    return -1;
+
+  if (1)
+    {
+      ELOG_TYPE_DECLARE (e) =
+      {
+      .format = "evt-enqueue: id %d length %d",.format_args = "i4i4",};
+      struct
+      {
+	u32 data[2];
+      } *ed;
+      ed = ELOG_DATA (&vlib_global_main.elog_main, e);
+      ed->data[0] = evt.event_id;
+      ed->data[1] = evt.enqueue_length;
+    }
+
+  return 0;
+}
+
+/**
+ * Flushes queue of sessions that are to be notified of new data
+ * enqueued events.
+ *
+ * @param thread_index Thread index for which the flush is to be performed.
+ * @return 0 on success or a positive number indicating the number of
+ *         failures due to API queue being full.
+ */
+int
+session_manager_flush_enqueue_events (u32 thread_index)
+{
+  session_manager_main_t *smm = &session_manager_main;
+  u32 *session_indices_to_enqueue;
+  int i, errors = 0;
+
+  session_indices_to_enqueue =
+    smm->session_indices_to_enqueue_by_thread[thread_index];
+
+  for (i = 0; i < vec_len (session_indices_to_enqueue); i++)
+    {
+      stream_session_t *s0;
+
+      /* Get session */
+      s0 = stream_session_get (session_indices_to_enqueue[i], thread_index);
+      if (stream_session_enqueue_notify (s0, 0 /* don't block */ ))
+	{
+	  errors++;
+	}
+    }
+
+  vec_reset_length (session_indices_to_enqueue);
+
+  smm->session_indices_to_enqueue_by_thread[thread_index] =
+    session_indices_to_enqueue;
+
+  /* Increment enqueue epoch for next round */
+  smm->current_enqueue_epoch[thread_index]++;
+
+  return errors;
+}
+
+/*
+ * Start listening on server's ip/port pair for requested transport.
+ *
+ * Creates a 'dummy' stream session with state LISTENING to be used in session
+ * lookups, prior to establishing connection. Requests transport to build
+ * it's own specific listening connection.
+ */
+int
+stream_session_start_listen (u32 server_index, ip46_address_t * ip, u16 port)
+{
+  session_manager_main_t *smm = &session_manager_main;
+  stream_session_t *s;
+  transport_connection_t *tc;
+  application_t *srv;
+  u32 tci;
+
+  srv = application_get (server_index);
+
+  pool_get (smm->listen_sessions[srv->session_type], s);
+  memset (s, 0, sizeof (*s));
+
+  s->session_type = srv->session_type;
+  s->session_state = SESSION_STATE_LISTENING;
+  s->session_index = s - smm->listen_sessions[srv->session_type];
+  s->app_index = srv->index;
+
+  /* Transport bind/listen  */
+  tci = tp_vfts[srv->session_type].bind (smm->vlib_main, s->session_index, ip,
+					 port);
+
+  /* Attach transport to session */
+  s->connection_index = tci;
+  tc = tp_vfts[srv->session_type].get_listener (tci);
+
+  srv->session_index = s->session_index;
+
+  /* Add to the main lookup table */
+  stream_session_table_add_for_tc (s->session_type, tc, s->session_index);
+
+  return 0;
+}
+
+void
+stream_session_stop_listen (u32 server_index)
+{
+  session_manager_main_t *smm = &session_manager_main;
+  stream_session_t *listener;
+  transport_connection_t *tc;
+  application_t *srv;
+
+  srv = application_get (server_index);
+  listener = pool_elt_at_index (smm->listen_sessions[srv->session_type],
+				srv->session_index);
+
+  tc = tp_vfts[srv->session_type].get_listener (listener->connection_index);
+  stream_session_table_del_for_tc (smm, listener->session_type, tc);
+
+  tp_vfts[srv->session_type].unbind (smm->vlib_main,
+				     listener->connection_index);
+  pool_put (smm->listen_sessions[srv->session_type], listener);
+}
+
+int
+connect_server_add_segment_cb (application_t * ss, char *segment_name,
+			       u32 segment_size)
+{
+  /* Does exactly nothing, but die */
+  ASSERT (0);
+  return 0;
+}
+
+void
+connects_session_manager_init (session_manager_main_t * smm, u8 session_type)
+{
+  session_manager_t *sm;
+  u32 connect_fifo_size = 8 << 10;	/* Config? */
+  u32 default_segment_size = 1 << 20;
+
+  pool_get (smm->session_managers, sm);
+  memset (sm, 0, sizeof (*sm));
+
+  sm->add_segment_size = default_segment_size;
+  sm->rx_fifo_size = connect_fifo_size;
+  sm->tx_fifo_size = connect_fifo_size;
+  sm->add_segment = 1;
+
+  session_manager_add_segment (smm, sm);
+  smm->connect_manager_index[session_type] = sm - smm->session_managers;
+}
+
+void
+stream_session_connect_notify (transport_connection_t * tc, u8 sst,
+			       u8 is_fail)
+{
+  session_manager_main_t *smm = &session_manager_main;
+  application_t *app;
+  stream_session_t *new_s = 0;
+  u64 value;
+
+  value = stream_session_half_open_lookup (smm, &tc->lcl_ip, &tc->rmt_ip,
+					   tc->lcl_port, tc->rmt_port,
+					   tc->proto);
+  if (value == HALF_OPEN_LOOKUP_INVALID_VALUE)
+    {
+      clib_warning ("This can't be good!");
+      return;
+    }
+
+  app = application_get (value >> 32);
+
+  if (!is_fail)
+    {
+      /* Create new session (server segments are allocated if needed) */
+      if (stream_session_create_i (smm, app, tc, &new_s))
+	return;
+
+      app->session_index = stream_session_get_index (new_s);
+      app->thread_index = new_s->thread_index;
+
+      /* Allocate vpp event queue for this thread if needed */
+      vpp_session_event_queue_allocate (smm, tc->thread_index);
+    }
+
+  /* Notify client */
+  app->cb_fns.session_connected_callback (app->api_client_index, new_s,
+					  is_fail);
+
+  /* Cleanup session lookup */
+  stream_session_half_open_table_del (smm, sst, tc);
+}
+
+void
+stream_session_accept_notify (transport_connection_t * tc)
+{
+  application_t *server;
+  stream_session_t *s;
+
+  s = stream_session_get (tc->s_index, tc->thread_index);
+  server = application_get (s->app_index);
+  server->cb_fns.session_accept_callback (s);
+}
+
+/**
+ * Notification from transport that connection is being closed.
+ *
+ * A disconnect is sent to application but state is not removed. Once
+ * disconnect is acknowledged by application, session disconnect is called.
+ * Ultimately this leads to close being called on transport (passive close).
+ */
+void
+stream_session_disconnect_notify (transport_connection_t * tc)
+{
+  application_t *server;
+  stream_session_t *s;
+
+  s = stream_session_get (tc->s_index, tc->thread_index);
+  server = application_get (s->app_index);
+  server->cb_fns.session_disconnect_callback (s);
+}
+
+/**
+ * Cleans up session and associated app if needed.
+ */
+void
+stream_session_delete (stream_session_t * s)
+{
+  session_manager_main_t *smm = vnet_get_session_manager_main ();
+  svm_fifo_segment_private_t *fifo_segment;
+  application_t *app;
+  int rv;
+
+  /* delete from the main lookup table */
+  rv = stream_session_table_del (smm, s);
+
+  if (rv)
+    clib_warning ("hash delete error, rv %d", rv);
+
+  /* Cleanup fifo segments */
+  fifo_segment = svm_fifo_get_segment (s->server_segment_index);
+  svm_fifo_segment_free_fifo (fifo_segment, s->server_rx_fifo);
+  svm_fifo_segment_free_fifo (fifo_segment, s->server_tx_fifo);
+
+  /* Cleanup app if client */
+  app = application_get (s->app_index);
+  if (app->mode == APP_CLIENT)
+    {
+      application_del (app);
+    }
+  else if (app->mode == APP_SERVER)
+    {
+      session_manager_t *sm;
+      svm_fifo_segment_private_t *fifo_segment;
+      svm_fifo_t **fifos;
+      u32 fifo_index;
+
+      sm = session_manager_get (app->session_manager_index);
+
+      /* Delete fifo */
+      fifo_segment = svm_fifo_get_segment (s->server_segment_index);
+      fifos = (svm_fifo_t **) fifo_segment->h->fifos;
+
+      fifo_index = svm_fifo_segment_index (fifo_segment);
+
+      /* Remove segment only if it holds no fifos and not the first */
+      if (sm->segment_indices[0] != fifo_index && vec_len (fifos) == 0)
+	svm_fifo_segment_delete (fifo_segment);
+    }
+
+  pool_put (smm->sessions[s->thread_index], s);
+}
+
+/**
+ * Notification from transport that connection is being deleted
+ *
+ * This should be called only on previously fully established sessions. For
+ * instance failed connects should call stream_session_connect_notify and
+ * indicate that the connect has failed.
+ */
+void
+stream_session_delete_notify (transport_connection_t * tc)
+{
+  stream_session_t *s;
+
+  s = stream_session_get_if_valid (tc->s_index, tc->thread_index);
+  if (!s)
+    {
+      clib_warning ("Surprised!");
+      return;
+    }
+  stream_session_delete (s);
+}
+
+/**
+ * Notify application that connection has been reset.
+ */
+void
+stream_session_reset_notify (transport_connection_t * tc)
+{
+  stream_session_t *s;
+  application_t *app;
+  s = stream_session_get (tc->s_index, tc->thread_index);
+
+  app = application_get (s->app_index);
+  app->cb_fns.session_reset_callback (s);
+}
+
+/**
+ * Accept a stream session. Optionally ping the server by callback.
+ */
+int
+stream_session_accept (transport_connection_t * tc, u32 listener_index,
+		       u8 sst, u8 notify)
+{
+  session_manager_main_t *smm = &session_manager_main;
+  application_t *server;
+  stream_session_t *s, *listener;
+
+  int rv;
+
+  /* Find the server */
+  listener = pool_elt_at_index (smm->listen_sessions[sst], listener_index);
+  server = application_get (listener->app_index);
+
+  if ((rv = stream_session_create_i (smm, server, tc, &s)))
+    return rv;
+
+  /* Allocate vpp event queue for this thread if needed */
+  vpp_session_event_queue_allocate (smm, tc->thread_index);
+
+  /* Shoulder-tap the server */
+  if (notify)
+    {
+      server->cb_fns.session_accept_callback (s);
+    }
+
+  return 0;
+}
+
+void
+stream_session_open (u8 sst, ip46_address_t * addr, u16 port_host_byte_order,
+		     u32 app_index)
+{
+  transport_connection_t *tc;
+  u32 tci;
+  u64 value;
+
+  /* Ask transport to open connection */
+  tci = tp_vfts[sst].open (addr, port_host_byte_order);
+
+  /* Get transport connection */
+  tc = tp_vfts[sst].get_half_open (tci);
+
+  /* Store api_client_index and transport connection index */
+  value = (((u64) app_index) << 32) | (u64) tc->c_index;
+
+  /* Add to the half-open lookup table */
+  stream_session_half_open_table_add (sst, tc, value);
+}
+
+/**
+ * Disconnect session and propagate to transport. This should eventually
+ * result in a delete notification that allows us to cleanup session state.
+ * Called for both active/passive disconnects.
+ */
+void
+stream_session_disconnect (stream_session_t * s)
+{
+  tp_vfts[s->session_type].close (s->connection_index, s->thread_index);
+  s->session_state = SESSION_STATE_CLOSED;
+}
+
+/**
+ * Cleanup transport and session state.
+ */
+void
+stream_session_cleanup (stream_session_t * s)
+{
+  tp_vfts[s->session_type].cleanup (s->connection_index, s->thread_index);
+  stream_session_delete (s);
+}
+
+void
+session_register_transport (u8 type, const transport_proto_vft_t * vft)
+{
+  session_manager_main_t *smm = vnet_get_session_manager_main ();
+
+  vec_validate (tp_vfts, type);
+  tp_vfts[type] = *vft;
+
+  /* If an offset function is provided, then peek instead of dequeue */
+  smm->session_rx_fns[type] =
+    (vft->rx_fifo_offset) ? session_fifo_rx_peek : session_fifo_rx_dequeue;
+}
+
+transport_proto_vft_t *
+session_get_transport_vft (u8 type)
+{
+  if (type >= vec_len (tp_vfts))
+    return 0;
+  return &tp_vfts[type];
+}
+
+static clib_error_t *
+session_manager_main_init (vlib_main_t * vm)
+{
+  u32 num_threads;
+  vlib_thread_main_t *vtm = vlib_get_thread_main ();
+  session_manager_main_t *smm = &session_manager_main;
+  int i;
+
+  smm->vlib_main = vm;
+  smm->vnet_main = vnet_get_main ();
+
+  num_threads = 1 /* main thread */  + vtm->n_threads;
+
+  if (num_threads < 1)
+    return clib_error_return (0, "n_thread_stacks not set");
+
+  /* $$$ config parameters */
+  svm_fifo_segment_init (0x200000000ULL /* first segment base VA */ ,
+			 20 /* timeout in seconds */ );
+
+  /* configure per-thread ** vectors */
+  vec_validate (smm->sessions, num_threads - 1);
+  vec_validate (smm->session_indices_to_enqueue_by_thread, num_threads - 1);
+  vec_validate (smm->tx_buffers, num_threads - 1);
+  vec_validate (smm->fifo_events, num_threads - 1);
+  vec_validate (smm->evts_partially_read, num_threads - 1);
+  vec_validate (smm->current_enqueue_epoch, num_threads - 1);
+  vec_validate (smm->vpp_event_queues, num_threads - 1);
+
+  /* $$$$ preallocate hack config parameter */
+  for (i = 0; i < 200000; i++)
+    {
+      stream_session_t *ss;
+      pool_get (smm->sessions[0], ss);
+      memset (ss, 0, sizeof (*ss));
+    }
+
+  for (i = 0; i < 200000; i++)
+    pool_put_index (smm->sessions[0], i);
+
+  clib_bihash_init_16_8 (&smm->v4_session_hash, "v4 session table",
+			 200000 /* $$$$ config parameter nbuckets */ ,
+			 (64 << 20) /*$$$ config parameter table size */ );
+  clib_bihash_init_48_8 (&smm->v6_session_hash, "v6 session table",
+			 200000 /* $$$$ config parameter nbuckets */ ,
+			 (64 << 20) /*$$$ config parameter table size */ );
+
+  clib_bihash_init_16_8 (&smm->v4_half_open_hash, "v4 half-open table",
+			 200000 /* $$$$ config parameter nbuckets */ ,
+			 (64 << 20) /*$$$ config parameter table size */ );
+  clib_bihash_init_48_8 (&smm->v6_half_open_hash, "v6 half-open table",
+			 200000 /* $$$$ config parameter nbuckets */ ,
+			 (64 << 20) /*$$$ config parameter table size */ );
+
+  for (i = 0; i < SESSION_N_TYPES; i++)
+    smm->connect_manager_index[i] = INVALID_INDEX;
+
+  return 0;
+}
+
+VLIB_INIT_FUNCTION (session_manager_main_init);
+
+/*
+ * fd.io coding-style-patch-verification: ON
+ *
+ * Local Variables:
+ * eval: (c-set-style "gnu")
+ * End:
+ */
diff --git a/src/vnet/session/session.h b/src/vnet/session/session.h
new file mode 100644
index 00000000..cf14cca9
--- /dev/null
+++ b/src/vnet/session/session.h
@@ -0,0 +1,380 @@
+/*
+ * Copyright (c) 2017 Cisco and/or its affiliates.
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at:
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+#ifndef __included_session_h__
+#define __included_session_h__
+
+#include 
+#include 
+#include 
+#include 
+#include 
+
+#define HALF_OPEN_LOOKUP_INVALID_VALUE ((u64)~0)
+#define INVALID_INDEX ((u32)~0)
+
+/* TODO decide how much since we have pre-data as well */
+#define MAX_HDRS_LEN    100	/* Max number of bytes for headers */
+
+typedef enum
+{
+  FIFO_EVENT_SERVER_RX,
+  FIFO_EVENT_SERVER_TX,
+  FIFO_EVENT_TIMEOUT,
+  FIFO_EVENT_SERVER_EXIT,
+} fifo_event_type_t;
+
+#define foreach_session_input_error                                         \
+_(NO_SESSION, "No session drops")                                       \
+_(NO_LISTENER, "No listener for dst port drops")                        \
+_(ENQUEUED, "Packets pushed into rx fifo")                              \
+_(NOT_READY, "Session not ready packets")                               \
+_(FIFO_FULL, "Packets dropped for lack of rx fifo space")               \
+_(EVENT_FIFO_FULL, "Events not sent for lack of event fifo space")      \
+_(API_QUEUE_FULL, "Sessions not created for lack of API queue space")   \
+_(NEW_SEG_NO_SPACE, "Created segment, couldn't allocate a fifo pair")   \
+_(NO_SPACE, "Couldn't allocate a fifo pair")
+
+typedef enum
+{
+#define _(sym,str) SESSION_ERROR_##sym,
+  foreach_session_input_error
+#undef _
+    SESSION_N_ERROR,
+} session_error_t;
+
+/* Event queue input node static next indices */
+typedef enum
+{
+  SESSION_QUEUE_NEXT_DROP,
+  SESSION_QUEUE_NEXT_TCP_IP4_OUTPUT,
+  SESSION_QUEUE_NEXT_IP4_LOOKUP,
+  SESSION_QUEUE_NEXT_TCP_IP6_OUTPUT,
+  SESSION_QUEUE_NEXT_IP6_LOOKUP,
+  SESSION_QUEUE_N_NEXT,
+} session_queue_next_t;
+
+#define foreach_session_type                    \
+  _(IP4_TCP, ip4_tcp)                           \
+  _(IP4_UDP, ip4_udp)                           \
+  _(IP6_TCP, ip6_tcp)                           \
+  _(IP6_UDP, ip6_udp)
+
+typedef enum
+{
+#define _(A, a) SESSION_TYPE_##A,
+  foreach_session_type
+#undef _
+    SESSION_N_TYPES,
+} session_type_t;
+
+/*
+ * Application session state
+ */
+typedef enum
+{
+  SESSION_STATE_LISTENING,
+  SESSION_STATE_CONNECTING,
+  SESSION_STATE_READY,
+  SESSION_STATE_CLOSED,
+  SESSION_STATE_N_STATES,
+} stream_session_state_t;
+
+typedef CLIB_PACKED (struct
+		     {
+		     svm_fifo_t * fifo;
+		     u8 event_type;
+		     /* $$$$ for event logging */
+		     u16 event_id;
+		     u32 enqueue_length;
+		     }) session_fifo_event_t;
+
+typedef struct _stream_session_t
+{
+  /** Type */
+  u8 session_type;
+
+  /** State */
+  u8 session_state;
+
+  /** Session index in per_thread pool */
+  u32 session_index;
+
+  /** Transport specific */
+  u32 connection_index;
+
+  u8 thread_index;
+
+  /** Application specific */
+  u32 pid;
+
+  /** fifo pointers. Once allocated, these do not move */
+  svm_fifo_t *server_rx_fifo;
+  svm_fifo_t *server_tx_fifo;
+
+  /** To avoid n**2 "one event per frame" check */
+  u8 enqueue_epoch;
+
+  /** used during unbind processing */
+  u8 is_deleted;
+
+  /** stream server pool index */
+  u32 app_index;
+
+  /** svm segment index */
+  u32 server_segment_index;
+} stream_session_t;
+
+typedef struct _session_manager
+{
+  /** segments mapped by this server */
+  u32 *segment_indices;
+
+  /** Session fifo sizes. They are provided for binds and take default
+   * values for connects */
+  u32 rx_fifo_size;
+  u32 tx_fifo_size;
+
+  /** Configured additional segment size */
+  u32 add_segment_size;
+
+  /** Flag that indicates if additional segments should be created */
+  u8 add_segment;
+} session_manager_t;
+
+/* Forward definition */
+typedef struct _session_manager_main session_manager_main_t;
+
+typedef int
+  (session_fifo_rx_fn) (vlib_main_t * vm, vlib_node_runtime_t * node,
+			session_manager_main_t * smm,
+			session_fifo_event_t * e0, stream_session_t * s0,
+			u32 thread_index, int *n_tx_pkts);
+
+extern session_fifo_rx_fn session_fifo_rx_peek;
+extern session_fifo_rx_fn session_fifo_rx_dequeue;
+
+struct _session_manager_main
+{
+  /** Lookup tables for established sessions and listeners */
+  clib_bihash_16_8_t v4_session_hash;
+  clib_bihash_48_8_t v6_session_hash;
+
+  /** Lookup tables for half-open sessions */
+  clib_bihash_16_8_t v4_half_open_hash;
+  clib_bihash_48_8_t v6_half_open_hash;
+
+  /** Per worker thread session pools */
+  stream_session_t **sessions;
+
+  /** Pool of listen sessions. Same type as stream sessions to ease lookups */
+  stream_session_t *listen_sessions[SESSION_N_TYPES];
+
+  /** Sparse vector to map dst port to stream server  */
+  u16 *stream_server_by_dst_port[SESSION_N_TYPES];
+
+  /** per-worker enqueue epoch counters */
+  u8 *current_enqueue_epoch;
+
+  /** Per-worker thread vector of sessions to enqueue */
+  u32 **session_indices_to_enqueue_by_thread;
+
+  /** per-worker tx buffer free lists */
+  u32 **tx_buffers;
+
+  /** Per worker-thread vector of partially read events */
+  session_fifo_event_t **evts_partially_read;
+
+  /** per-worker active event vectors */
+  session_fifo_event_t **fifo_events;
+
+  /** vpp fifo event queue */
+  unix_shared_memory_queue_t **vpp_event_queues;
+
+  /** Unique segment name counter */
+  u32 unique_segment_name_counter;
+
+  /* Connection manager used by incoming connects */
+  u32 connect_manager_index[SESSION_N_TYPES];
+
+  session_manager_t *session_managers;
+
+  /** Per transport rx function that can either dequeue or peek */
+  session_fifo_rx_fn *session_rx_fns[SESSION_N_TYPES];
+
+  /* Convenience */
+  vlib_main_t *vlib_main;
+  vnet_main_t *vnet_main;
+};
+
+extern session_manager_main_t session_manager_main;
+
+/*
+ * Session manager function
+ */
+always_inline session_manager_main_t *
+vnet_get_session_manager_main ()
+{
+  return &session_manager_main;
+}
+
+always_inline session_manager_t *
+session_manager_get (u32 index)
+{
+  return pool_elt_at_index (session_manager_main.session_managers, index);
+}
+
+always_inline unix_shared_memory_queue_t *
+session_manager_get_vpp_event_queue (u32 thread_index)
+{
+  return session_manager_main.vpp_event_queues[thread_index];
+}
+
+always_inline session_manager_t *
+connects_session_manager_get (session_manager_main_t * smm,
+			      session_type_t session_type)
+{
+  return pool_elt_at_index (smm->session_managers,
+			    smm->connect_manager_index[session_type]);
+}
+
+void session_manager_get_segment_info (u32 index, u8 ** name, u32 * size);
+int session_manager_flush_enqueue_events (u32 thread_index);
+int
+session_manager_add_first_segment (session_manager_main_t * smm,
+				   session_manager_t * sm, u32 segment_size,
+				   u8 ** segment_name);
+void
+session_manager_del (session_manager_main_t * smm, session_manager_t * sm);
+void
+connects_session_manager_init (session_manager_main_t * smm, u8 session_type);
+
+/*
+ * Stream session functions
+ */
+
+stream_session_t *stream_session_lookup_listener4 (ip4_address_t * lcl,
+						   u16 lcl_port, u8 proto);
+stream_session_t *stream_session_lookup4 (ip4_address_t * lcl,
+					  ip4_address_t * rmt, u16 lcl_port,
+					  u16 rmt_port, u8 proto,
+					  u32 thread_index);
+stream_session_t *stream_session_lookup_listener6 (ip6_address_t * lcl,
+						   u16 lcl_port, u8 proto);
+stream_session_t *stream_session_lookup6 (ip6_address_t * lcl,
+					  ip6_address_t * rmt, u16 lcl_port,
+					  u16 rmt_port, u8, u32 thread_index);
+transport_connection_t
+  * stream_session_lookup_transport4 (session_manager_main_t * smm,
+				      ip4_address_t * lcl,
+				      ip4_address_t * rmt, u16 lcl_port,
+				      u16 rmt_port, u8 proto,
+				      u32 thread_index);
+transport_connection_t
+  * stream_session_lookup_transport6 (session_manager_main_t * smm,
+				      ip6_address_t * lcl,
+				      ip6_address_t * rmt, u16 lcl_port,
+				      u16 rmt_port, u8 proto,
+				      u32 thread_index);
+stream_session_t *stream_session_lookup_listener (ip46_address_t * lcl,
+						  u16 lcl_port, u8 proto);
+
+always_inline stream_session_t *
+stream_session_get_tsi (u64 ti_and_si, u32 thread_index)
+{
+  ASSERT ((u32) (ti_and_si >> 32) == thread_index);
+  return pool_elt_at_index (session_manager_main.sessions[thread_index],
+			    ti_and_si & 0xFFFFFFFFULL);
+}
+
+always_inline stream_session_t *
+stream_session_get (u64 si, u32 thread_index)
+{
+  return pool_elt_at_index (session_manager_main.sessions[thread_index], si);
+}
+
+always_inline stream_session_t *
+stream_session_get_if_valid (u64 si, u32 thread_index)
+{
+  if (thread_index >= vec_len (session_manager_main.sessions))
+    return 0;
+
+  if (pool_is_free_index (session_manager_main.sessions[thread_index], si))
+    return 0;
+
+  return pool_elt_at_index (session_manager_main.sessions[thread_index], si);
+}
+
+always_inline stream_session_t *
+stream_session_listener_get (u8 sst, u64 si)
+{
+  return pool_elt_at_index (session_manager_main.listen_sessions[sst], si);
+}
+
+always_inline u32
+stream_session_get_index (stream_session_t * s)
+{
+  if (s->session_state == SESSION_STATE_LISTENING)
+    return s - session_manager_main.listen_sessions[s->session_type];
+
+  return s - session_manager_main.sessions[s->thread_index];
+}
+
+always_inline u32
+stream_session_max_enqueue (transport_connection_t * tc)
+{
+  stream_session_t *s = stream_session_get (tc->s_index, tc->thread_index);
+  return svm_fifo_max_enqueue (s->server_rx_fifo);
+}
+
+int
+stream_session_enqueue_data (transport_connection_t * tc, u8 * data, u16 len,
+			     u8 queue_event);
+u32
+stream_session_peek_bytes (transport_connection_t * tc, u8 * buffer,
+			   u32 offset, u32 max_bytes);
+u32 stream_session_dequeue_drop (transport_connection_t * tc, u32 max_bytes);
+
+void
+stream_session_connect_notify (transport_connection_t * tc, u8 sst,
+			       u8 is_fail);
+void stream_session_accept_notify (transport_connection_t * tc);
+void stream_session_disconnect_notify (transport_connection_t * tc);
+void stream_session_delete_notify (transport_connection_t * tc);
+void stream_session_reset_notify (transport_connection_t * tc);
+int
+stream_session_accept (transport_connection_t * tc, u32 listener_index,
+		       u8 sst, u8 notify);
+void stream_session_open (u8 sst, ip46_address_t * addr,
+			  u16 port_host_byte_order, u32 api_client_index);
+void stream_session_disconnect (stream_session_t * s);
+void stream_session_cleanup (stream_session_t * s);
+int
+stream_session_start_listen (u32 server_index, ip46_address_t * ip, u16 port);
+void stream_session_stop_listen (u32 server_index);
+
+u8 *format_stream_session (u8 * s, va_list * args);
+
+void session_register_transport (u8 type, const transport_proto_vft_t * vft);
+transport_proto_vft_t *session_get_transport_vft (u8 type);
+
+#endif /* __included_session_h__ */
+
+/*
+ * fd.io coding-style-patch-verification: ON
+ *
+ * Local Variables:
+ * eval: (c-set-style "gnu")
+ * End:
+ */
diff --git a/src/vnet/session/session_api.c b/src/vnet/session/session_api.c
new file mode 100644
index 00000000..9d068684
--- /dev/null
+++ b/src/vnet/session/session_api.c
@@ -0,0 +1,821 @@
+/*
+ * Copyright (c) 2015-2016 Cisco and/or its affiliates.
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at:
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#include 
+#include 
+#include 
+
+#include 
+#include "application_interface.h"
+
+#define vl_typedefs		/* define message structures */
+#include 
+#undef vl_typedefs
+
+#define vl_endianfun		/* define message structures */
+#include 
+#undef vl_endianfun
+
+/* instantiate all the print functions we know about */
+#define vl_print(handle, ...) vlib_cli_output (handle, __VA_ARGS__)
+#define vl_printfun
+#include 
+#undef vl_printfun
+
+#include 
+
+#define foreach_session_api_msg                                         \
+_(MAP_ANOTHER_SEGMENT_REPLY, map_another_segment_reply)                 \
+_(BIND_URI, bind_uri)                                                   \
+_(UNBIND_URI, unbind_uri)                                               \
+_(CONNECT_URI, connect_uri)                                             \
+_(DISCONNECT_SESSION, disconnect_session)                               \
+_(DISCONNECT_SESSION_REPLY, disconnect_session_reply)                   \
+_(ACCEPT_SESSION_REPLY, accept_session_reply)                           \
+_(RESET_SESSION_REPLY, reset_session_reply)                   		\
+_(BIND_SOCK, bind_sock) 		                                \
+_(UNBIND_SOCK, unbind_sock)                                             \
+_(CONNECT_SOCK, connect_sock)                                          	\
+_(DISCONNECT_SOCK, disconnect_sock)                               	\
+_(DISCONNECT_SOCK_REPLY, disconnect_sock_reply)                        	\
+_(ACCEPT_SOCK_REPLY, accept_sock_reply)                           	\
+_(RESET_SOCK_REPLY, reset_sock_reply)                   		\
+
+static int
+send_add_segment_callback (u32 api_client_index, const u8 * segment_name,
+			   u32 segment_size)
+{
+  vl_api_map_another_segment_t *mp;
+  unix_shared_memory_queue_t *q;
+
+  q = vl_api_client_index_to_input_queue (api_client_index);
+
+  if (!q)
+    return -1;
+
+  mp = vl_msg_api_alloc (sizeof (*mp));
+  memset (mp, 0, sizeof (*mp));
+  mp->_vl_msg_id = clib_host_to_net_u16 (VL_API_MAP_ANOTHER_SEGMENT);
+  mp->segment_size = segment_size;
+  strncpy ((char *) mp->segment_name, (char *) segment_name,
+	   sizeof (mp->segment_name) - 1);
+
+  vl_msg_api_send_shmem (q, (u8 *) & mp);
+
+  return 0;
+}
+
+static int
+send_session_accept_uri_callback (stream_session_t * s)
+{
+  vl_api_accept_session_t *mp;
+  unix_shared_memory_queue_t *q, *vpp_queue;
+  application_t *server = application_get (s->app_index);
+
+  q = vl_api_client_index_to_input_queue (server->api_client_index);
+  vpp_queue = session_manager_get_vpp_event_queue (s->thread_index);
+
+  if (!q)
+    return -1;
+
+  mp = vl_msg_api_alloc (sizeof (*mp));
+  mp->_vl_msg_id = clib_host_to_net_u16 (VL_API_ACCEPT_SESSION);
+
+  /* Note: session_type is the first octet in all types of sessions */
+
+  mp->accept_cookie = server->accept_cookie;
+  mp->server_rx_fifo = (u64) s->server_rx_fifo;
+  mp->server_tx_fifo = (u64) s->server_tx_fifo;
+  mp->session_thread_index = s->thread_index;
+  mp->session_index = s->session_index;
+  mp->session_type = s->session_type;
+  mp->vpp_event_queue_address = (u64) vpp_queue;
+  vl_msg_api_send_shmem (q, (u8 *) & mp);
+
+  return 0;
+}
+
+static void
+send_session_disconnect_uri_callback (stream_session_t * s)
+{
+  vl_api_disconnect_session_t *mp;
+  unix_shared_memory_queue_t *q;
+  application_t *app = application_get (s->app_index);
+
+  q = vl_api_client_index_to_input_queue (app->api_client_index);
+
+  if (!q)
+    return;
+
+  mp = vl_msg_api_alloc (sizeof (*mp));
+  memset (mp, 0, sizeof (*mp));
+  mp->_vl_msg_id = clib_host_to_net_u16 (VL_API_DISCONNECT_SESSION);
+
+  mp->session_thread_index = s->thread_index;
+  mp->session_index = s->session_index;
+  vl_msg_api_send_shmem (q, (u8 *) & mp);
+}
+
+static int
+send_session_connected_uri_callback (u32 api_client_index,
+				     stream_session_t * s, u8 is_fail)
+{
+  vl_api_connect_uri_reply_t *mp;
+  unix_shared_memory_queue_t *q;
+  application_t *app = application_lookup (api_client_index);
+  u8 *seg_name;
+  unix_shared_memory_queue_t *vpp_queue;
+
+  q = vl_api_client_index_to_input_queue (app->api_client_index);
+
+  if (!q)
+    return -1;
+
+  mp = vl_msg_api_alloc (sizeof (*mp));
+  mp->_vl_msg_id = clib_host_to_net_u16 (VL_API_CONNECT_URI_REPLY);
+  mp->context = app->api_context;
+  mp->retval = is_fail;
+  if (!is_fail)
+    {
+      vpp_queue = session_manager_get_vpp_event_queue (s->thread_index);
+      mp->server_rx_fifo = (u64) s->server_rx_fifo;
+      mp->server_tx_fifo = (u64) s->server_tx_fifo;
+      mp->session_thread_index = s->thread_index;
+      mp->session_index = s->session_index;
+      mp->session_type = s->session_type;
+      mp->vpp_event_queue_address = (u64) vpp_queue;
+      mp->client_event_queue_address = (u64) app->event_queue;
+
+      session_manager_get_segment_info (s->server_segment_index, &seg_name,
+					&mp->segment_size);
+      mp->segment_name_length = vec_len (seg_name);
+      if (mp->segment_name_length)
+	clib_memcpy (mp->segment_name, seg_name, mp->segment_name_length);
+    }
+
+  vl_msg_api_send_shmem (q, (u8 *) & mp);
+
+  /* Remove client if connect failed */
+  if (is_fail)
+    application_del (app);
+
+  return 0;
+}
+
+/**
+ * Redirect a connect_uri message to the indicated server.
+ * Only sent if the server has bound the related port with
+ * URI_OPTIONS_FLAGS_USE_FIFO
+ */
+static int
+redirect_connect_uri_callback (u32 server_api_client_index, void *mp_arg)
+{
+  vl_api_connect_uri_t *mp = mp_arg;
+  unix_shared_memory_queue_t *server_q, *client_q;
+  vlib_main_t *vm = vlib_get_main ();
+  f64 timeout = vlib_time_now (vm) + 0.5;
+  int rv = 0;
+
+  server_q = vl_api_client_index_to_input_queue (server_api_client_index);
+
+  if (!server_q)
+    {
+      rv = VNET_API_ERROR_INVALID_VALUE;
+      goto out;
+    }
+
+  client_q = vl_api_client_index_to_input_queue (mp->client_index);
+  if (!client_q)
+    {
+      rv = VNET_API_ERROR_INVALID_VALUE_2;
+      goto out;
+    }
+
+  /* Tell the server the client's API queue address, so it can reply */
+  mp->client_queue_address = (u64) client_q;
+
+  /*
+   * Bounce message handlers MUST NOT block the data-plane.
+   * Spin waiting for the queue lock, but
+   */
+
+  while (vlib_time_now (vm) < timeout)
+    {
+      rv =
+	unix_shared_memory_queue_add (server_q, (u8 *) & mp, 1 /*nowait */ );
+      switch (rv)
+	{
+	  /* correctly enqueued */
+	case 0:
+	  return VNET_CONNECT_REDIRECTED;
+
+	  /* continue spinning, wait for pthread_mutex_trylock to work */
+	case -1:
+	  continue;
+
+	  /* queue stuffed, drop the msg */
+	case -2:
+	  rv = VNET_API_ERROR_QUEUE_FULL;
+	  goto out;
+	}
+    }
+out:
+  /* Dispose of the message */
+  vl_msg_api_free (mp);
+  return rv;
+}
+
+static u64
+make_session_handle (stream_session_t * s)
+{
+  return (u64) s->session_index << 32 | (u64) s->thread_index;
+}
+
+static int
+send_session_accept_callback (stream_session_t * s)
+{
+  vl_api_accept_sock_t *mp;
+  unix_shared_memory_queue_t *q, *vpp_queue;
+  application_t *server = application_get (s->app_index);
+
+  q = vl_api_client_index_to_input_queue (server->api_client_index);
+  vpp_queue = session_manager_get_vpp_event_queue (s->thread_index);
+
+  if (!q)
+    return -1;
+
+  mp = vl_msg_api_alloc (sizeof (*mp));
+  mp->_vl_msg_id = clib_host_to_net_u16 (VL_API_ACCEPT_SOCK);
+
+  /* Note: session_type is the first octet in all types of sessions */
+
+  mp->accept_cookie = server->accept_cookie;
+  mp->server_rx_fifo = (u64) s->server_rx_fifo;
+  mp->server_tx_fifo = (u64) s->server_tx_fifo;
+  mp->handle = make_session_handle (s);
+  mp->vpp_event_queue_address = (u64) vpp_queue;
+  vl_msg_api_send_shmem (q, (u8 *) & mp);
+
+  return 0;
+}
+
+static int
+send_session_connected_callback (u32 api_client_index, stream_session_t * s,
+				 u8 is_fail)
+{
+  vl_api_connect_sock_reply_t *mp;
+  unix_shared_memory_queue_t *q;
+  application_t *app = application_lookup (api_client_index);
+  u8 *seg_name;
+  unix_shared_memory_queue_t *vpp_queue;
+
+  q = vl_api_client_index_to_input_queue (app->api_client_index);
+
+  if (!q)
+    return -1;
+
+  mp = vl_msg_api_alloc (sizeof (*mp));
+  mp->_vl_msg_id = clib_host_to_net_u16 (VL_API_CONNECT_SOCK_REPLY);
+  mp->context = app->api_context;
+  mp->retval = is_fail;
+  if (!is_fail)
+    {
+      vpp_queue = session_manager_get_vpp_event_queue (s->thread_index);
+      mp->server_rx_fifo = (u64) s->server_rx_fifo;
+      mp->server_tx_fifo = (u64) s->server_tx_fifo;
+      mp->handle = make_session_handle (s);
+      mp->vpp_event_queue_address = (u64) vpp_queue;
+      mp->client_event_queue_address = (u64) app->event_queue;
+
+      session_manager_get_segment_info (s->server_segment_index, &seg_name,
+					&mp->segment_size);
+      mp->segment_name_length = vec_len (seg_name);
+      if (mp->segment_name_length)
+	clib_memcpy (mp->segment_name, seg_name, mp->segment_name_length);
+    }
+
+  vl_msg_api_send_shmem (q, (u8 *) & mp);
+
+  /* Remove client if connect failed */
+  if (is_fail)
+    application_del (app);
+
+  return 0;
+}
+
+static void
+send_session_disconnect_callback (stream_session_t * s)
+{
+  vl_api_disconnect_sock_t *mp;
+  unix_shared_memory_queue_t *q;
+  application_t *app = application_get (s->app_index);
+
+  q = vl_api_client_index_to_input_queue (app->api_client_index);
+
+  if (!q)
+    return;
+
+  mp = vl_msg_api_alloc (sizeof (*mp));
+  memset (mp, 0, sizeof (*mp));
+  mp->_vl_msg_id = clib_host_to_net_u16 (VL_API_DISCONNECT_SOCK);
+
+  mp->handle = make_session_handle (s);
+  vl_msg_api_send_shmem (q, (u8 *) & mp);
+}
+
+/**
+ * Redirect a connect_uri message to the indicated server.
+ * Only sent if the server has bound the related port with
+ * URI_OPTIONS_FLAGS_USE_FIFO
+ */
+static int
+redirect_connect_callback (u32 server_api_client_index, void *mp_arg)
+{
+  vl_api_connect_sock_t *mp = mp_arg;
+  unix_shared_memory_queue_t *server_q, *client_q;
+  vlib_main_t *vm = vlib_get_main ();
+  f64 timeout = vlib_time_now (vm) + 0.5;
+  int rv = 0;
+
+  server_q = vl_api_client_index_to_input_queue (server_api_client_index);
+
+  if (!server_q)
+    {
+      rv = VNET_API_ERROR_INVALID_VALUE;
+      goto out;
+    }
+
+  client_q = vl_api_client_index_to_input_queue (mp->client_index);
+  if (!client_q)
+    {
+      rv = VNET_API_ERROR_INVALID_VALUE_2;
+      goto out;
+    }
+
+  /* Tell the server the client's API queue address, so it can reply */
+  mp->client_queue_address = (u64) client_q;
+
+  /*
+   * Bounce message handlers MUST NOT block the data-plane.
+   * Spin waiting for the queue lock, but
+   */
+
+  while (vlib_time_now (vm) < timeout)
+    {
+      rv =
+	unix_shared_memory_queue_add (server_q, (u8 *) & mp, 1 /*nowait */ );
+      switch (rv)
+	{
+	  /* correctly enqueued */
+	case 0:
+	  return VNET_CONNECT_REDIRECTED;
+
+	  /* continue spinning, wait for pthread_mutex_trylock to work */
+	case -1:
+	  continue;
+
+	  /* queue stuffed, drop the msg */
+	case -2:
+	  rv = VNET_API_ERROR_QUEUE_FULL;
+	  goto out;
+	}
+    }
+out:
+  /* Dispose of the message */
+  vl_msg_api_free (mp);
+  return rv;
+}
+
+static session_cb_vft_t uri_session_cb_vft = {
+  .session_accept_callback = send_session_accept_uri_callback,
+  .session_disconnect_callback = send_session_disconnect_uri_callback,
+  .session_connected_callback = send_session_connected_uri_callback,
+  .add_segment_callback = send_add_segment_callback,
+  .redirect_connect_callback = redirect_connect_uri_callback
+};
+
+static session_cb_vft_t session_cb_vft = {
+  .session_accept_callback = send_session_accept_callback,
+  .session_disconnect_callback = send_session_disconnect_callback,
+  .session_connected_callback = send_session_connected_callback,
+  .add_segment_callback = send_add_segment_callback,
+  .redirect_connect_callback = redirect_connect_callback
+};
+
+static int
+api_session_not_valid (u32 session_index, u32 thread_index)
+{
+  session_manager_main_t *smm = vnet_get_session_manager_main ();
+  stream_session_t *pool;
+
+  if (thread_index >= vec_len (smm->sessions))
+    return VNET_API_ERROR_INVALID_VALUE;
+
+  pool = smm->sessions[thread_index];
+
+  if (pool_is_free_index (pool, session_index))
+    return VNET_API_ERROR_INVALID_VALUE_2;
+
+  return 0;
+}
+
+static void
+vl_api_bind_uri_t_handler (vl_api_bind_uri_t * mp)
+{
+  vl_api_bind_uri_reply_t *rmp;
+  vnet_bind_args_t _a, *a = &_a;
+  char segment_name[128];
+  u32 segment_name_length;
+  int rv;
+
+  _Static_assert (sizeof (u64) * SESSION_OPTIONS_N_OPTIONS <=
+		  sizeof (mp->options),
+		  "Out of options, fix api message definition");
+
+  segment_name_length = ARRAY_LEN (segment_name);
+
+  memset (a, 0, sizeof (*a));
+
+  a->uri = (char *) mp->uri;
+  a->api_client_index = mp->client_index;
+  a->options = mp->options;
+  a->segment_name = segment_name;
+  a->segment_name_length = segment_name_length;
+  a->session_cb_vft = &uri_session_cb_vft;
+
+  a->options[SESSION_OPTIONS_SEGMENT_SIZE] = mp->initial_segment_size;
+  a->options[SESSION_OPTIONS_ACCEPT_COOKIE] = mp->accept_cookie;
+  rv = vnet_bind_uri (a);
+
+  /* *INDENT-OFF* */
+  REPLY_MACRO2 (VL_API_BIND_URI_REPLY, ({
+    rmp->retval = rv;
+    if (!rv)
+      {
+	rmp->segment_name_length = 0;
+	/* $$$$ policy? */
+	rmp->segment_size = mp->initial_segment_size;
+	if (segment_name_length)
+	  {
+	    memcpy (rmp->segment_name, segment_name, segment_name_length);
+	    rmp->segment_name_length = segment_name_length;
+	  }
+	rmp->server_event_queue_address = a->server_event_queue_address;
+      }
+  }));
+  /* *INDENT-ON* */
+
+}
+
+static void
+vl_api_unbind_uri_t_handler (vl_api_unbind_uri_t * mp)
+{
+  vl_api_unbind_uri_reply_t *rmp;
+  int rv;
+
+  rv = vnet_unbind_uri ((char *) mp->uri, mp->client_index);
+
+  REPLY_MACRO (VL_API_UNBIND_URI_REPLY);
+}
+
+static void
+vl_api_connect_uri_t_handler (vl_api_connect_uri_t * mp)
+{
+  vnet_connect_args_t _a, *a = &_a;
+
+  a->uri = (char *) mp->uri;
+  a->api_client_index = mp->client_index;
+  a->api_context = mp->context;
+  a->options = mp->options;
+  a->session_cb_vft = &uri_session_cb_vft;
+  a->mp = mp;
+  vnet_connect_uri (a);
+}
+
+static void
+vl_api_disconnect_session_t_handler (vl_api_disconnect_session_t * mp)
+{
+  vl_api_disconnect_session_reply_t *rmp;
+  int rv;
+
+  rv = api_session_not_valid (mp->session_index, mp->session_thread_index);
+  if (!rv)
+    rv = vnet_disconnect_session (mp->client_index, mp->session_index,
+				  mp->session_thread_index);
+
+  REPLY_MACRO (VL_API_DISCONNECT_SESSION_REPLY);
+}
+
+static void
+vl_api_disconnect_session_reply_t_handler (vl_api_disconnect_session_reply_t *
+					   mp)
+{
+  if (api_session_not_valid (mp->session_index, mp->session_thread_index))
+    {
+      clib_warning ("Invalid session!");
+      return;
+    }
+
+  /* Client objected to disconnecting the session, log and continue */
+  if (mp->retval)
+    {
+      clib_warning ("client retval %d", mp->retval);
+      return;
+    }
+
+  /* Disconnect has been confirmed. Confirm close to transport */
+  vnet_disconnect_session (mp->client_index, mp->session_index,
+			   mp->session_thread_index);
+}
+
+static void
+vl_api_reset_session_reply_t_handler (vl_api_reset_session_reply_t * mp)
+{
+  stream_session_t *s;
+
+  if (api_session_not_valid (mp->session_index, mp->session_thread_index))
+    {
+      clib_warning ("Invalid session!");
+      return;
+    }
+
+  /* Client objected to resetting the session, log and continue */
+  if (mp->retval)
+    {
+      clib_warning ("client retval %d", mp->retval);
+      return;
+    }
+
+  s = stream_session_get (mp->session_index, mp->session_thread_index);
+
+  /* This comes as a response to a reset, transport only waiting for
+   * confirmation to remove connection state, no need to disconnect */
+  stream_session_cleanup (s);
+}
+
+static void
+vl_api_accept_session_reply_t_handler (vl_api_accept_session_reply_t * mp)
+{
+  stream_session_t *s;
+  int rv;
+
+  if (api_session_not_valid (mp->session_index, mp->session_thread_index))
+    return;
+
+  s = stream_session_get (mp->session_index, mp->session_thread_index);
+  rv = mp->retval;
+
+  if (rv)
+    {
+      /* Server isn't interested, kill the session */
+      stream_session_disconnect (s);
+      return;
+    }
+
+  s->session_state = SESSION_STATE_READY;
+}
+
+static void
+vl_api_map_another_segment_reply_t_handler (vl_api_map_another_segment_reply_t
+					    * mp)
+{
+  clib_warning ("not implemented");
+}
+
+static void
+vl_api_bind_sock_t_handler (vl_api_bind_sock_t * mp)
+{
+  vl_api_bind_sock_reply_t *rmp;
+  vnet_bind_args_t _a, *a = &_a;
+  char segment_name[128];
+  u32 segment_name_length;
+  int rv;
+
+  STATIC_ASSERT (sizeof (u64) * SESSION_OPTIONS_N_OPTIONS <=
+		 sizeof (mp->options),
+		 "Out of options, fix api message definition");
+
+  segment_name_length = ARRAY_LEN (segment_name);
+
+  memset (a, 0, sizeof (*a));
+
+  clib_memcpy (&a->tep.ip, mp->ip,
+	       (mp->is_ip4 ? sizeof (ip4_address_t) :
+		sizeof (ip6_address_t)));
+  a->tep.is_ip4 = mp->is_ip4;
+  a->tep.port = mp->port;
+  a->tep.vrf = mp->vrf;
+
+  a->api_client_index = mp->client_index;
+  a->options = mp->options;
+  a->segment_name = segment_name;
+  a->segment_name_length = segment_name_length;
+  a->session_cb_vft = &session_cb_vft;
+
+  rv = vnet_bind_uri (a);
+
+  /* *INDENT-OFF* */
+  REPLY_MACRO2 (VL_API_BIND_SOCK_REPLY, ({
+    rmp->retval = rv;
+    if (!rv)
+      {
+	rmp->segment_name_length = 0;
+	rmp->segment_size = mp->options[SESSION_OPTIONS_SEGMENT_SIZE];
+	if (segment_name_length)
+	  {
+	    memcpy(rmp->segment_name, segment_name, segment_name_length);
+	    rmp->segment_name_length = segment_name_length;
+	  }
+	rmp->server_event_queue_address = a->server_event_queue_address;
+      }
+  }));
+  /* *INDENT-ON* */
+}
+
+static void
+vl_api_unbind_sock_t_handler (vl_api_unbind_sock_t * mp)
+{
+  vl_api_unbind_sock_reply_t *rmp;
+  vnet_unbind_args_t _a, *a = &_a;
+  int rv;
+
+  a->api_client_index = mp->client_index;
+  a->handle = mp->handle;
+
+  rv = vnet_unbind (a);
+
+  REPLY_MACRO (VL_API_UNBIND_SOCK_REPLY);
+}
+
+static void
+vl_api_connect_sock_t_handler (vl_api_connect_sock_t * mp)
+{
+  vnet_connect_args_t _a, *a = &_a;
+
+  clib_memcpy (&a->tep.ip, mp->ip,
+	       (mp->is_ip4 ? sizeof (ip4_address_t) :
+		sizeof (ip6_address_t)));
+  a->tep.is_ip4 = mp->is_ip4;
+  a->tep.port = mp->port;
+  a->tep.vrf = mp->vrf;
+  a->options = mp->options;
+  a->session_cb_vft = &session_cb_vft;
+  a->api_context = mp->context;
+  a->mp = mp;
+
+  vnet_connect (a);
+}
+
+static void
+vl_api_disconnect_sock_t_handler (vl_api_disconnect_sock_t * mp)
+{
+  vnet_disconnect_args_t _a, *a = &_a;
+  vl_api_disconnect_sock_reply_t *rmp;
+  int rv;
+
+  a->api_client_index = mp->client_index;
+  a->handle = mp->handle;
+  rv = vnet_disconnect (a);
+
+  REPLY_MACRO (VL_API_DISCONNECT_SOCK_REPLY);
+}
+
+static void
+vl_api_disconnect_sock_reply_t_handler (vl_api_disconnect_sock_reply_t * mp)
+{
+  vnet_disconnect_args_t _a, *a = &_a;
+
+  /* Client objected to disconnecting the session, log and continue */
+  if (mp->retval)
+    {
+      clib_warning ("client retval %d", mp->retval);
+      return;
+    }
+
+  a->api_client_index = mp->client_index;
+  a->handle = mp->handle;
+
+  vnet_disconnect (a);
+}
+
+static void
+vl_api_reset_sock_reply_t_handler (vl_api_reset_sock_reply_t * mp)
+{
+  stream_session_t *s;
+  u32 session_index, thread_index;
+
+  /* Client objected to resetting the session, log and continue */
+  if (mp->retval)
+    {
+      clib_warning ("client retval %d", mp->retval);
+      return;
+    }
+
+  if (api_parse_session_handle (mp->handle, &session_index, &thread_index))
+    {
+      clib_warning ("Invalid handle");
+      return;
+    }
+
+  s = stream_session_get (session_index, thread_index);
+
+  /* This comes as a response to a reset, transport only waiting for
+   * confirmation to remove connection state, no need to disconnect */
+  stream_session_cleanup (s);
+}
+
+static void
+vl_api_accept_sock_reply_t_handler (vl_api_accept_sock_reply_t * mp)
+{
+  stream_session_t *s;
+  u32 session_index, thread_index;
+
+  if (api_parse_session_handle (mp->handle, &session_index, &thread_index))
+    {
+      clib_warning ("Invalid handle");
+      return;
+    }
+  s = stream_session_get (session_index, thread_index);
+
+  if (mp->retval)
+    {
+      /* Server isn't interested, kill the session */
+      stream_session_disconnect (s);
+      return;
+    }
+
+  s->session_state = SESSION_STATE_READY;
+}
+
+#define vl_msg_name_crc_list
+#include 
+#undef vl_msg_name_crc_list
+
+static void
+setup_message_id_table (api_main_t * am)
+{
+#define _(id,n,crc) vl_msg_api_add_msg_name_crc (am, #n "_" #crc, id);
+  foreach_vl_msg_name_crc_session;
+#undef _
+}
+
+/*
+ * session_api_hookup
+ * Add uri's API message handlers to the table.
+ * vlib has alread mapped shared memory and
+ * added the client registration handlers.
+ * See .../open-repo/vlib/memclnt_vlib.c:memclnt_process()
+ */
+static clib_error_t *
+session_api_hookup (vlib_main_t * vm)
+{
+  api_main_t *am = &api_main;
+
+#define _(N,n)                                                  \
+    vl_msg_api_set_handlers(VL_API_##N, #n,                     \
+                           vl_api_##n##_t_handler,              \
+                           vl_noop_handler,                     \
+                           vl_api_##n##_t_endian,               \
+                           vl_api_##n##_t_print,                \
+                           sizeof(vl_api_##n##_t), 1);
+  foreach_session_api_msg;
+#undef _
+
+  /*
+   * Messages which bounce off the data-plane to
+   * an API client. Simply tells the message handling infra not
+   * to free the message.
+   *
+   * Bounced message handlers MUST NOT block the data plane
+   */
+  am->message_bounce[VL_API_CONNECT_URI] = 1;
+  am->message_bounce[VL_API_CONNECT_SOCK] = 1;
+
+  /*
+   * Set up the (msg_name, crc, message-id) table
+   */
+  setup_message_id_table (am);
+
+  return 0;
+}
+
+VLIB_API_INIT_FUNCTION (session_api_hookup);
+/*
+ * fd.io coding-style-patch-verification: ON
+ *
+ * Local Variables:
+ * eval: (c-set-style "gnu")
+ * End:
+ */
diff --git a/src/vnet/session/session_cli.c b/src/vnet/session/session_cli.c
new file mode 100644
index 00000000..b2943a1c
--- /dev/null
+++ b/src/vnet/session/session_cli.c
@@ -0,0 +1,189 @@
+/*
+ * Copyright (c) 2017 Cisco and/or its affiliates.
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at:
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+#include 
+#include 
+
+/**
+ * Format stream session as per the following format
+ *
+ * verbose:
+ *   "Connection", "Rx fifo", "Tx fifo", "Session Index"
+ * non-verbose:
+ *   "Connection"
+ */
+u8 *
+format_stream_session (u8 * s, va_list * args)
+{
+  stream_session_t *ss = va_arg (*args, stream_session_t *);
+  int verbose = va_arg (*args, int);
+  transport_proto_vft_t *tp_vft;
+  u8 *str = 0;
+
+  tp_vft = session_get_transport_vft (ss->session_type);
+
+  if (verbose)
+    str = format (0, "%-20llp%-20llp%-15lld", ss->server_rx_fifo,
+		  ss->server_tx_fifo, stream_session_get_index (ss));
+
+  if (ss->session_state == SESSION_STATE_READY)
+    {
+      s = format (s, "%-40U%v", tp_vft->format_connection,
+		  ss->connection_index, ss->thread_index, str);
+    }
+  else if (ss->session_state == SESSION_STATE_LISTENING)
+    {
+      s = format (s, "%-40U%v", tp_vft->format_listener, ss->connection_index,
+		  str);
+    }
+  else if (ss->session_state == SESSION_STATE_READY)
+    {
+      s =
+	format (s, "%-40U%v", tp_vft->format_half_open, ss->connection_index,
+		str);
+    }
+  else if (ss->session_state == SESSION_STATE_CLOSED)
+    {
+      s = format (s, "[CL] %-40U%v", tp_vft->format_connection,
+		  ss->connection_index, ss->thread_index, str);
+    }
+  else
+    {
+      clib_warning ("Session in unknown state!");
+    }
+
+  vec_free (str);
+
+  return s;
+}
+
+static clib_error_t *
+show_session_command_fn (vlib_main_t * vm, unformat_input_t * input,
+			 vlib_cli_command_t * cmd)
+{
+  session_manager_main_t *smm = &session_manager_main;
+  int verbose = 0, i;
+  stream_session_t *pool;
+  stream_session_t *s;
+  u8 *str = 0;
+
+  while (unformat_check_input (input) != UNFORMAT_END_OF_INPUT)
+    {
+      if (unformat (input, "verbose"))
+	verbose = 1;
+      else
+	break;
+    }
+
+  for (i = 0; i < vec_len (smm->sessions); i++)
+    {
+      u32 once_per_pool;
+      pool = smm->sessions[i];
+
+      once_per_pool = 1;
+
+      if (pool_elts (pool))
+	{
+
+	  vlib_cli_output (vm, "Thread %d: %d active sessions",
+			   i, pool_elts (pool));
+	  if (verbose)
+	    {
+	      if (once_per_pool)
+		{
+		  str = format (str, "%-40s%-20s%-20s%-15s",
+				"Connection", "Rx fifo", "Tx fifo",
+				"Session Index");
+		  vlib_cli_output (vm, "%v", str);
+		  vec_reset_length (str);
+		  once_per_pool = 0;
+		}
+
+              /* *INDENT-OFF* */
+              pool_foreach (s, pool,
+              ({
+                vlib_cli_output (vm, "%U", format_stream_session, s, verbose);
+              }));
+              /* *INDENT-ON* */
+	    }
+	}
+      else
+	vlib_cli_output (vm, "Thread %d: no active sessions", i);
+    }
+  vec_free (str);
+
+  return 0;
+}
+
+VLIB_CLI_COMMAND (show_uri_command, static) =
+{
+.path = "show session",.short_help = "show session [verbose]",.function =
+    show_session_command_fn,};
+
+
+static clib_error_t *
+clear_session_command_fn (vlib_main_t * vm, unformat_input_t * input,
+			  vlib_cli_command_t * cmd)
+{
+  session_manager_main_t *smm = &session_manager_main;
+  u32 thread_index = 0;
+  u32 session_index = ~0;
+  stream_session_t *pool, *session;
+  application_t *server;
+
+  while (unformat_check_input (input) != UNFORMAT_END_OF_INPUT)
+    {
+      if (unformat (input, "thread %d", &thread_index))
+	;
+      else if (unformat (input, "session %d", &session_index))
+	;
+      else
+	return clib_error_return (0, "unknown input `%U'",
+				  format_unformat_error, input);
+    }
+
+  if (session_index == ~0)
+    return clib_error_return (0, "session  required, but not set.");
+
+  if (thread_index > vec_len (smm->sessions))
+    return clib_error_return (0, "thread %d out of range [0-%d]",
+			      thread_index, vec_len (smm->sessions));
+
+  pool = smm->sessions[thread_index];
+
+  if (pool_is_free_index (pool, session_index))
+    return clib_error_return (0, "session %d not active", session_index);
+
+  session = pool_elt_at_index (pool, session_index);
+  server = application_get (session->app_index);
+
+  /* Disconnect both app and transport */
+  server->cb_fns.session_disconnect_callback (session);
+
+  return 0;
+}
+
+VLIB_CLI_COMMAND (clear_uri_session_command, static) =
+{
+.path = "clear session",.short_help =
+    "clear session thread  session ",.function =
+    clear_session_command_fn,};
+
+/*
+ * fd.io coding-style-patch-verification: ON
+ *
+ * Local Variables:
+ * eval: (c-set-style "gnu")
+ * End:
+ */
diff --git a/src/vnet/session/transport.c b/src/vnet/session/transport.c
new file mode 100644
index 00000000..abd94ba4
--- /dev/null
+++ b/src/vnet/session/transport.c
@@ -0,0 +1,64 @@
+/*
+ * Copyright (c) 2017 Cisco and/or its affiliates.
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at:
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#include 
+
+u32
+transport_endpoint_lookup (transport_endpoint_table_t *ht, ip46_address_t *ip,
+                           u16 port)
+{
+  clib_bihash_kv_24_8_t kv;
+  int rv;
+
+  kv.key[0] = ip->as_u64[0];
+  kv.key[1] = ip->as_u64[1];
+  kv.key[2] = port;
+
+  rv = clib_bihash_search_inline_24_8 (ht, &kv);
+  if (rv == 0)
+    return kv.value;
+
+  return TRANSPORT_ENDPOINT_INVALID_INDEX;
+}
+
+void
+transport_endpoint_table_add (transport_endpoint_table_t *ht,
+                              transport_endpoint_t *te, u32 value)
+{
+  clib_bihash_kv_24_8_t kv;
+
+  kv.key[0] = te->ip.as_u64[0];
+  kv.key[1] = te->ip.as_u64[1];
+  kv.key[2] = te->port;
+  kv.value = value;
+
+  clib_bihash_add_del_24_8 (ht, &kv, 1);
+}
+
+void
+transport_endpoint_table_del (transport_endpoint_table_t *ht,
+                              transport_endpoint_t *te)
+{
+  clib_bihash_kv_24_8_t kv;
+
+  kv.key[0] = te->ip.as_u64[0];
+  kv.key[1] = te->ip.as_u64[1];
+  kv.key[2] = te->port;
+
+  clib_bihash_add_del_24_8 (ht, &kv, 0);
+}
+
+
+
diff --git a/src/vnet/session/transport.h b/src/vnet/session/transport.h
new file mode 100644
index 00000000..2d4415ba
--- /dev/null
+++ b/src/vnet/session/transport.h
@@ -0,0 +1,250 @@
+/*
+ * Copyright (c) 2016 Cisco and/or its affiliates.
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at:
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#ifndef VNET_VNET_URI_TRANSPORT_H_
+#define VNET_VNET_URI_TRANSPORT_H_
+
+#include 
+#include 
+#include 
+#include 
+
+/*
+ * Protocol independent transport properties associated to a session
+ */
+typedef struct _transport_connection
+{
+  ip46_address_t rmt_ip;	/**< Remote IP */
+  ip46_address_t lcl_ip;	/**< Local IP */
+  u16 lcl_port;			/**< Local port */
+  u16 rmt_port;			/**< Remote port */
+  u8 proto;			/**< Transport protocol id */
+
+  u32 s_index;			/**< Parent session index */
+  u32 c_index;			/**< Connection index in transport pool */
+  u8 is_ip4;			/**< Flag if IP4 connection */
+  u32 thread_index;		/**< Worker-thread index */
+
+  /** Macros for 'derived classes' where base is named "connection" */
+#define c_lcl_ip connection.lcl_ip
+#define c_rmt_ip connection.rmt_ip
+#define c_lcl_ip4 connection.lcl_ip.ip4
+#define c_rmt_ip4 connection.rmt_ip.ip4
+#define c_lcl_ip6 connection.lcl_ip.ip6
+#define c_rmt_ip6 connection.rmt_ip.ip6
+#define c_lcl_port connection.lcl_port
+#define c_rmt_port connection.rmt_port
+#define c_proto connection.proto
+#define c_state connection.state
+#define c_s_index connection.s_index
+#define c_c_index connection.c_index
+#define c_is_ip4 connection.is_ip4
+#define c_thread_index connection.thread_index
+} transport_connection_t;
+
+/*
+ * Transport protocol virtual function table
+ */
+typedef struct _transport_proto_vft
+{
+  /*
+   * Setup
+   */
+  u32 (*bind) (vlib_main_t *, u32, ip46_address_t *, u16);
+  u32 (*unbind) (vlib_main_t *, u32);
+  int (*open) (ip46_address_t * addr, u16 port_host_byte_order);
+  void (*close) (u32 conn_index, u32 thread_index);
+  void (*cleanup) (u32 conn_index, u32 thread_index);
+
+  /*
+   * Transmission
+   */
+    u32 (*push_header) (transport_connection_t * tconn, vlib_buffer_t * b);
+    u16 (*send_mss) (transport_connection_t * tc);
+    u32 (*send_space) (transport_connection_t * tc);
+    u32 (*rx_fifo_offset) (transport_connection_t * tc);
+
+  /*
+   * Connection retrieval
+   */
+  transport_connection_t *(*get_connection) (u32 conn_idx, u32 thread_idx);
+  transport_connection_t *(*get_listener) (u32 conn_index);
+  transport_connection_t *(*get_half_open) (u32 conn_index);
+
+  /*
+   * Format
+   */
+  u8 *(*format_connection) (u8 * s, va_list * args);
+  u8 *(*format_listener) (u8 * s, va_list * args);
+  u8 *(*format_half_open) (u8 * s, va_list * args);
+
+} transport_proto_vft_t;
+
+/* 16 octets */
+typedef CLIB_PACKED (struct
+		     {
+		     union
+		     {
+		     struct
+		     {
+		     ip4_address_t src; ip4_address_t dst;
+		     u16 src_port;
+		     u16 dst_port;
+		     /* align by making this 4 octets even though its a 1-bit field
+		      * NOTE: avoid key overlap with other transports that use 5 tuples for
+		      * session identification.
+		      */
+		     u32 proto;
+		     };
+		     u64 as_u64[2];
+		     };
+		     }) v4_connection_key_t;
+
+typedef CLIB_PACKED (struct
+		     {
+		     union
+		     {
+		     struct
+		     {
+		     /* 48 octets */
+		     ip6_address_t src; ip6_address_t dst;
+		     u16 src_port;
+		     u16 dst_port; u32 proto; u8 unused_for_now[8];
+		     }; u64 as_u64[6];
+		     };
+		     }) v6_connection_key_t;
+
+typedef clib_bihash_kv_16_8_t session_kv4_t;
+typedef clib_bihash_kv_48_8_t session_kv6_t;
+
+always_inline void
+make_v4_ss_kv (session_kv4_t * kv, ip4_address_t * lcl, ip4_address_t * rmt,
+	       u16 lcl_port, u16 rmt_port, u8 proto)
+{
+  v4_connection_key_t key;
+  memset (&key, 0, sizeof (v4_connection_key_t));
+
+  key.src.as_u32 = lcl->as_u32;
+  key.dst.as_u32 = rmt->as_u32;
+  key.src_port = lcl_port;
+  key.dst_port = rmt_port;
+  key.proto = proto;
+
+  kv->key[0] = key.as_u64[0];
+  kv->key[1] = key.as_u64[1];
+  kv->value = ~0ULL;
+}
+
+always_inline void
+make_v4_listener_kv (session_kv4_t * kv, ip4_address_t * lcl, u16 lcl_port,
+		     u8 proto)
+{
+  v4_connection_key_t key;
+  memset (&key, 0, sizeof (v4_connection_key_t));
+
+  key.src.as_u32 = lcl->as_u32;
+  key.dst.as_u32 = 0;
+  key.src_port = lcl_port;
+  key.dst_port = 0;
+  key.proto = proto;
+
+  kv->key[0] = key.as_u64[0];
+  kv->key[1] = key.as_u64[1];
+  kv->value = ~0ULL;
+}
+
+always_inline void
+make_v4_ss_kv_from_tc (session_kv4_t * kv, transport_connection_t * t)
+{
+  return make_v4_ss_kv (kv, &t->lcl_ip.ip4, &t->rmt_ip.ip4, t->lcl_port,
+			t->rmt_port, t->proto);
+}
+
+always_inline void
+make_v6_ss_kv (session_kv6_t * kv, ip6_address_t * lcl, ip6_address_t * rmt,
+	       u16 lcl_port, u16 rmt_port, u8 proto)
+{
+  v6_connection_key_t key;
+  memset (&key, 0, sizeof (v6_connection_key_t));
+
+  key.src.as_u64[0] = lcl->as_u64[0];
+  key.src.as_u64[1] = lcl->as_u64[1];
+  key.dst.as_u64[0] = rmt->as_u64[0];
+  key.dst.as_u64[1] = rmt->as_u64[1];
+  key.src_port = lcl_port;
+  key.dst_port = rmt_port;
+  key.proto = proto;
+
+  kv->key[0] = key.as_u64[0];
+  kv->key[1] = key.as_u64[1];
+  kv->value = ~0ULL;
+}
+
+always_inline void
+make_v6_listener_kv (session_kv6_t * kv, ip6_address_t * lcl, u16 lcl_port,
+		     u8 proto)
+{
+  v6_connection_key_t key;
+  memset (&key, 0, sizeof (v6_connection_key_t));
+
+  key.src.as_u64[0] = lcl->as_u64[0];
+  key.src.as_u64[1] = lcl->as_u64[1];
+  key.dst.as_u64[0] = 0;
+  key.dst.as_u64[1] = 0;
+  key.src_port = lcl_port;
+  key.dst_port = 0;
+  key.proto = proto;
+
+  kv->key[0] = key.as_u64[0];
+  kv->key[1] = key.as_u64[1];
+  kv->value = ~0ULL;
+}
+
+always_inline void
+make_v6_ss_kv_from_tc (session_kv6_t * kv, transport_connection_t * t)
+{
+  make_v6_ss_kv (kv, &t->lcl_ip.ip6, &t->rmt_ip.ip6, t->lcl_port,
+		 t->rmt_port, t->proto);
+}
+
+typedef struct _transport_endpoint
+{
+  ip46_address_t ip;
+  u16 port;
+  u8 is_ip4;
+  u32 vrf;
+} transport_endpoint_t;
+
+typedef clib_bihash_24_8_t transport_endpoint_table_t;
+
+#define TRANSPORT_ENDPOINT_INVALID_INDEX ((u32)~0)
+
+u32
+transport_endpoint_lookup (transport_endpoint_table_t * ht,
+			   ip46_address_t * ip, u16 port);
+void transport_endpoint_table_add (transport_endpoint_table_t * ht,
+				   transport_endpoint_t * te, u32 value);
+void transport_endpoint_table_del (transport_endpoint_table_t * ht,
+				   transport_endpoint_t * te);
+
+#endif /* VNET_VNET_URI_TRANSPORT_H_ */
+
+/*
+ * fd.io coding-style-patch-verification: ON
+ *
+ * Local Variables:
+ * eval: (c-set-style "gnu")
+ * End:
+ */
diff --git a/src/vnet/tcp/tcp.c b/src/vnet/tcp/tcp.c
new file mode 100644
index 00000000..0f9b7097
--- /dev/null
+++ b/src/vnet/tcp/tcp.c
@@ -0,0 +1,708 @@
+/*
+ * Copyright (c) 2016 Cisco and/or its affiliates.
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at:
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#include 
+#include 
+#include 
+#include 
+
+tcp_main_t tcp_main;
+
+static u32
+tcp_connection_bind (vlib_main_t * vm, u32 session_index, ip46_address_t * ip,
+		     u16 port_host_byte_order, u8 is_ip4)
+{
+  tcp_main_t *tm = &tcp_main;
+  tcp_connection_t *listener;
+
+  pool_get (tm->listener_pool, listener);
+  memset (listener, 0, sizeof (*listener));
+
+  listener->c_c_index = listener - tm->listener_pool;
+  listener->c_lcl_port = clib_host_to_net_u16 (port_host_byte_order);
+
+  if (is_ip4)
+    listener->c_lcl_ip4.as_u32 = ip->ip4.as_u32;
+  else
+    clib_memcpy (&listener->c_lcl_ip6, &ip->ip6, sizeof (ip6_address_t));
+
+  listener->c_s_index = session_index;
+  listener->c_proto = SESSION_TYPE_IP4_TCP;
+  listener->state = TCP_STATE_LISTEN;
+  listener->c_is_ip4 = 1;
+
+  return listener->c_c_index;
+}
+
+u32
+tcp_session_bind_ip4 (vlib_main_t * vm, u32 session_index,
+		      ip46_address_t * ip, u16 port_host_byte_order)
+{
+  return tcp_connection_bind (vm, session_index, ip, port_host_byte_order, 1);
+}
+
+u32
+tcp_session_bind_ip6 (vlib_main_t * vm, u32 session_index,
+		      ip46_address_t * ip, u16 port_host_byte_order)
+{
+  return tcp_connection_bind (vm, session_index, ip, port_host_byte_order, 0);
+
+}
+
+static void
+tcp_session_unbind (u32 listener_index)
+{
+  tcp_main_t *tm = vnet_get_tcp_main ();
+  pool_put_index (tm->listener_pool, listener_index);
+}
+
+u32
+tcp_session_unbind_ip4 (vlib_main_t * vm, u32 listener_index)
+{
+  tcp_session_unbind (listener_index);
+  return 0;
+}
+
+u32
+tcp_session_unbind_ip6 (vlib_main_t * vm, u32 listener_index)
+{
+  tcp_session_unbind (listener_index);
+  return 0;
+}
+
+transport_connection_t *
+tcp_session_get_listener (u32 listener_index)
+{
+  tcp_main_t *tm = vnet_get_tcp_main ();
+  tcp_connection_t *tc;
+  tc = pool_elt_at_index (tm->listener_pool, listener_index);
+  return &tc->connection;
+}
+
+/**
+ * Cleans up connection state.
+ *
+ * No notifications.
+ */
+void
+tcp_connection_cleanup (tcp_connection_t * tc)
+{
+  tcp_main_t *tm = &tcp_main;
+  u32 tepi;
+  transport_endpoint_t *tep;
+
+  /* Cleanup local endpoint if this was an active connect */
+  tepi = transport_endpoint_lookup (&tm->local_endpoints_table, &tc->c_lcl_ip,
+				    tc->c_lcl_port);
+
+  /*XXX lock */
+  if (tepi != TRANSPORT_ENDPOINT_INVALID_INDEX)
+    {
+      tep = pool_elt_at_index (tm->local_endpoints, tepi);
+      transport_endpoint_table_del (&tm->local_endpoints_table, tep);
+      pool_put (tm->local_endpoints, tep);
+    }
+
+  /* Make sure all timers are cleared */
+  tcp_connection_timers_reset (tc);
+
+  /* Check if half-open */
+  if (tc->state == TCP_STATE_SYN_SENT)
+    pool_put (tm->half_open_connections, tc);
+  else
+    pool_put (tm->connections[tc->c_thread_index], tc);
+}
+
+/**
+ * Connection removal.
+ *
+ * This should be called only once connection enters CLOSED state. Note
+ * that it notifies the session of the removal event, so if the goal is to
+ * just remove the connection, call tcp_connection_cleanup instead.
+ */
+void
+tcp_connection_del (tcp_connection_t * tc)
+{
+  stream_session_delete_notify (&tc->connection);
+  tcp_connection_cleanup (tc);
+}
+
+/**
+ * Begin connection closing procedure.
+ *
+ * If at the end the connection is not in CLOSED state, it is not removed.
+ * Instead, we rely on on TCP to advance through state machine to either
+ * 1) LAST_ACK (passive close) whereby when the last ACK is received
+ * tcp_connection_del is called. This notifies session of the delete and
+ * calls cleanup.
+ * 2) TIME_WAIT (active close) whereby after 2MSL the 2MSL timer triggers
+ * and cleanup is called.
+ */
+void
+tcp_connection_close (tcp_connection_t * tc)
+{
+  /* Send FIN if needed */
+  if (tc->state == TCP_STATE_ESTABLISHED || tc->state == TCP_STATE_SYN_RCVD
+      || tc->state == TCP_STATE_CLOSE_WAIT)
+    tcp_send_fin (tc);
+
+  /* Switch state */
+  if (tc->state == TCP_STATE_ESTABLISHED || tc->state == TCP_STATE_SYN_RCVD)
+    tc->state = TCP_STATE_FIN_WAIT_1;
+  else if (tc->state == TCP_STATE_SYN_SENT)
+    tc->state = TCP_STATE_CLOSED;
+  else if (tc->state == TCP_STATE_CLOSE_WAIT)
+    tc->state = TCP_STATE_LAST_ACK;
+
+  /* Half-close connections are not supported XXX */
+
+  if (tc->state == TCP_STATE_CLOSED)
+    tcp_connection_del (tc);
+}
+
+void
+tcp_session_close (u32 conn_index, u32 thread_index)
+{
+  tcp_connection_t *tc;
+  tc = tcp_connection_get (conn_index, thread_index);
+  tcp_connection_close (tc);
+}
+
+void
+tcp_session_cleanup (u32 conn_index, u32 thread_index)
+{
+  tcp_connection_t *tc;
+  tc = tcp_connection_get (conn_index, thread_index);
+  tcp_connection_cleanup (tc);
+}
+
+void *
+ip_interface_get_first_ip (u32 sw_if_index, u8 is_ip4)
+{
+  ip_lookup_main_t *lm4 = &ip4_main.lookup_main;
+  ip_lookup_main_t *lm6 = &ip6_main.lookup_main;
+  ip_interface_address_t *ia = 0;
+
+  if (is_ip4)
+    {
+      /* *INDENT-OFF* */
+      foreach_ip_interface_address (lm4, ia, sw_if_index, 1 /* unnumbered */ ,
+      ({
+        return ip_interface_address_get_address (lm4, ia);
+      }));
+      /* *INDENT-ON* */
+    }
+  else
+    {
+      /* *INDENT-OFF* */
+      foreach_ip_interface_address (lm6, ia, sw_if_index, 1 /* unnumbered */ ,
+      ({
+        return ip_interface_address_get_address (lm6, ia);
+      }));
+      /* *INDENT-ON* */
+    }
+
+  return 0;
+}
+
+/**
+ * Allocate local port and add if successful add entry to local endpoint
+ * table to mark the pair as used.
+ */
+u16
+tcp_allocate_local_port (tcp_main_t * tm, ip46_address_t * ip)
+{
+  u8 unique = 0;
+  transport_endpoint_t *tep;
+  u32 time_now, tei;
+  u16 min = 1024, max = 65535, tries;	/* XXX configurable ? */
+
+  tries = max - min;
+  time_now = tcp_time_now ();
+
+  /* Start at random point or max */
+  pool_get (tm->local_endpoints, tep);
+  clib_memcpy (&tep->ip, ip, sizeof (*ip));
+  tep->port = random_u32 (&time_now) << 16;
+  tep->port = tep->port < min ? max : tep->port;
+
+  /* Search for first free slot */
+  while (tries)
+    {
+      tei = transport_endpoint_lookup (&tm->local_endpoints_table, &tep->ip,
+				       tep->port);
+      if (tei == TRANSPORT_ENDPOINT_INVALID_INDEX)
+	{
+	  unique = 1;
+	  break;
+	}
+
+      tep->port--;
+
+      if (tep->port < min)
+	tep->port = max;
+
+      tries--;
+    }
+
+  if (unique)
+    {
+      transport_endpoint_table_add (&tm->local_endpoints_table, tep,
+				    tep - tm->local_endpoints);
+
+      return tep->port;
+    }
+
+  /* Failed */
+  pool_put (tm->local_endpoints, tep);
+  return -1;
+}
+
+/**
+ * Initialize all connection timers as invalid
+ */
+void
+tcp_connection_timers_init (tcp_connection_t * tc)
+{
+  int i;
+
+  /* Set all to invalid */
+  for (i = 0; i < TCP_N_TIMERS; i++)
+    {
+      tc->timers[i] = TCP_TIMER_HANDLE_INVALID;
+    }
+
+  tc->rto = TCP_RTO_INIT;
+}
+
+/**
+ * Stop all connection timers
+ */
+void
+tcp_connection_timers_reset (tcp_connection_t * tc)
+{
+  int i;
+  for (i = 0; i < TCP_N_TIMERS; i++)
+    {
+      tcp_timer_reset (tc, i);
+    }
+}
+
+/** Initialize tcp connection variables
+ *
+ * Should be called after having received a msg from the peer, i.e., a SYN or
+ * a SYNACK, such that connection options have already been exchanged. */
+void
+tcp_connection_init_vars (tcp_connection_t * tc)
+{
+  tcp_connection_timers_init (tc);
+  tcp_set_snd_mss (tc);
+  tc->sack_sb.head = TCP_INVALID_SACK_HOLE_INDEX;
+  tcp_cc_init (tc);
+}
+
+int
+tcp_connection_open (ip46_address_t * rmt_addr, u16 rmt_port, u8 is_ip4)
+{
+  tcp_main_t *tm = vnet_get_tcp_main ();
+  tcp_connection_t *tc;
+  fib_prefix_t prefix;
+  u32 fei, sw_if_index;
+  ip46_address_t lcl_addr;
+  u16 lcl_port;
+
+  /*
+   * Find the local address and allocate port
+   */
+  memset (&lcl_addr, 0, sizeof (lcl_addr));
+
+  /* Find a FIB path to the destination */
+  clib_memcpy (&prefix.fp_addr, rmt_addr, sizeof (*rmt_addr));
+  prefix.fp_proto = is_ip4 ? FIB_PROTOCOL_IP4 : FIB_PROTOCOL_IP6;
+  prefix.fp_len = is_ip4 ? 32 : 128;
+
+  fei = fib_table_lookup (0, &prefix);
+
+  /* Couldn't find route to destination. Bail out. */
+  if (fei == FIB_NODE_INDEX_INVALID)
+    return -1;
+
+  sw_if_index = fib_entry_get_resolving_interface (fei);
+
+  if (sw_if_index == (u32) ~ 0)
+    return -1;
+
+  if (is_ip4)
+    {
+      ip4_address_t *ip4;
+      ip4 = ip_interface_get_first_ip (sw_if_index, 1);
+      lcl_addr.ip4.as_u32 = ip4->as_u32;
+    }
+  else
+    {
+      ip6_address_t *ip6;
+      ip6 = ip_interface_get_first_ip (sw_if_index, 0);
+      clib_memcpy (&lcl_addr.ip6, ip6, sizeof (*ip6));
+    }
+
+  /* Allocate source port */
+  lcl_port = tcp_allocate_local_port (tm, &lcl_addr);
+  if (lcl_port < 1)
+    return -1;
+
+  /*
+   * Create connection and send SYN
+   */
+
+  pool_get (tm->half_open_connections, tc);
+  memset (tc, 0, sizeof (*tc));
+
+  clib_memcpy (&tc->c_rmt_ip, rmt_addr, sizeof (ip46_address_t));
+  clib_memcpy (&tc->c_lcl_ip, &lcl_addr, sizeof (ip46_address_t));
+  tc->c_rmt_port = clib_host_to_net_u16 (rmt_port);
+  tc->c_lcl_port = clib_host_to_net_u16 (lcl_port);
+  tc->c_c_index = tc - tm->half_open_connections;
+  tc->c_is_ip4 = is_ip4;
+
+  /* The other connection vars will be initialized after SYN ACK */
+  tcp_connection_timers_init (tc);
+
+  tcp_send_syn (tc);
+
+  tc->state = TCP_STATE_SYN_SENT;
+
+  return tc->c_c_index;
+}
+
+int
+tcp_session_open_ip4 (ip46_address_t * addr, u16 port)
+{
+  return tcp_connection_open (addr, port, 1);
+}
+
+int
+tcp_session_open_ip6 (ip46_address_t * addr, u16 port)
+{
+  return tcp_connection_open (addr, port, 0);
+}
+
+u8 *
+format_tcp_session_ip4 (u8 * s, va_list * args)
+{
+  u32 tci = va_arg (*args, u32);
+  u32 thread_index = va_arg (*args, u32);
+  tcp_connection_t *tc;
+
+  tc = tcp_connection_get (tci, thread_index);
+
+  s = format (s, "[%s] %U:%d->%U:%d", "tcp", format_ip4_address,
+	      &tc->c_lcl_ip4, clib_net_to_host_u16 (tc->c_lcl_port),
+	      format_ip4_address, &tc->c_rmt_ip4,
+	      clib_net_to_host_u16 (tc->c_rmt_port));
+
+  return s;
+}
+
+u8 *
+format_tcp_session_ip6 (u8 * s, va_list * args)
+{
+  u32 tci = va_arg (*args, u32);
+  u32 thread_index = va_arg (*args, u32);
+  tcp_connection_t *tc = tcp_connection_get (tci, thread_index);
+  s = format (s, "[%s] %U:%d->%U:%d", "tcp", format_ip6_address,
+	      &tc->c_lcl_ip6, clib_net_to_host_u16 (tc->c_lcl_port),
+	      format_ip6_address, &tc->c_rmt_ip6,
+	      clib_net_to_host_u16 (tc->c_rmt_port));
+  return s;
+}
+
+u8 *
+format_tcp_listener_session_ip4 (u8 * s, va_list * args)
+{
+  u32 tci = va_arg (*args, u32);
+  tcp_connection_t *tc = tcp_listener_get (tci);
+  s = format (s, "[%s] %U:%d->%U:%d", "tcp", format_ip4_address,
+	      &tc->c_lcl_ip4, clib_net_to_host_u16 (tc->c_lcl_port),
+	      format_ip4_address, &tc->c_rmt_ip4,
+	      clib_net_to_host_u16 (tc->c_rmt_port));
+  return s;
+}
+
+u8 *
+format_tcp_listener_session_ip6 (u8 * s, va_list * args)
+{
+  u32 tci = va_arg (*args, u32);
+  tcp_connection_t *tc = tcp_listener_get (tci);
+  s = format (s, "[%s] %U:%d->%U:%d", "tcp", format_ip6_address,
+	      &tc->c_lcl_ip6, clib_net_to_host_u16 (tc->c_lcl_port),
+	      format_ip6_address, &tc->c_rmt_ip6,
+	      clib_net_to_host_u16 (tc->c_rmt_port));
+  return s;
+}
+
+u8 *
+format_tcp_half_open_session_ip4 (u8 * s, va_list * args)
+{
+  u32 tci = va_arg (*args, u32);
+  tcp_connection_t *tc = tcp_half_open_connection_get (tci);
+  s = format (s, "[%s] %U:%d->%U:%d", "tcp", format_ip4_address,
+	      &tc->c_lcl_ip4, clib_net_to_host_u16 (tc->c_lcl_port),
+	      format_ip4_address, &tc->c_rmt_ip4,
+	      clib_net_to_host_u16 (tc->c_rmt_port));
+  return s;
+}
+
+u8 *
+format_tcp_half_open_session_ip6 (u8 * s, va_list * args)
+{
+  u32 tci = va_arg (*args, u32);
+  tcp_connection_t *tc = tcp_half_open_connection_get (tci);
+  s = format (s, "[%s] %U:%d->%U:%d", "tcp", format_ip6_address,
+	      &tc->c_lcl_ip6, clib_net_to_host_u16 (tc->c_lcl_port),
+	      format_ip6_address, &tc->c_rmt_ip6,
+	      clib_net_to_host_u16 (tc->c_rmt_port));
+  return s;
+}
+
+transport_connection_t *
+tcp_session_get_transport (u32 conn_index, u32 thread_index)
+{
+  tcp_connection_t *tc = tcp_connection_get (conn_index, thread_index);
+  return &tc->connection;
+}
+
+transport_connection_t *
+tcp_half_open_session_get_transport (u32 conn_index)
+{
+  tcp_connection_t *tc = tcp_half_open_connection_get (conn_index);
+  return &tc->connection;
+}
+
+u16
+tcp_session_send_mss (transport_connection_t * trans_conn)
+{
+  tcp_connection_t *tc = (tcp_connection_t *) trans_conn;
+  return tc->snd_mss;
+}
+
+u32
+tcp_session_send_space (transport_connection_t * trans_conn)
+{
+  tcp_connection_t *tc = (tcp_connection_t *) trans_conn;
+  return tcp_available_snd_space (tc);
+}
+
+u32
+tcp_session_rx_fifo_offset (transport_connection_t * trans_conn)
+{
+  tcp_connection_t *tc = (tcp_connection_t *) trans_conn;
+  return (tc->snd_una_max - tc->snd_una);
+}
+
+/* *INDENT-OFF* */
+const static transport_proto_vft_t tcp4_proto = {
+  .bind = tcp_session_bind_ip4,
+  .unbind = tcp_session_unbind_ip4,
+  .push_header = tcp_push_header,
+  .get_connection = tcp_session_get_transport,
+  .get_listener = tcp_session_get_listener,
+  .get_half_open = tcp_half_open_session_get_transport,
+  .open = tcp_session_open_ip4,
+  .close = tcp_session_close,
+  .cleanup = tcp_session_cleanup,
+  .send_mss = tcp_session_send_mss,
+  .send_space = tcp_session_send_space,
+  .rx_fifo_offset = tcp_session_rx_fifo_offset,
+  .format_connection = format_tcp_session_ip4,
+  .format_listener = format_tcp_listener_session_ip4,
+  .format_half_open = format_tcp_half_open_session_ip4
+};
+
+const static transport_proto_vft_t tcp6_proto = {
+  .bind = tcp_session_bind_ip6,
+  .unbind = tcp_session_unbind_ip6,
+  .push_header = tcp_push_header,
+  .get_connection = tcp_session_get_transport,
+  .get_listener = tcp_session_get_listener,
+  .get_half_open = tcp_half_open_session_get_transport,
+  .open = tcp_session_open_ip6,
+  .close = tcp_session_close,
+  .cleanup = tcp_session_cleanup,
+  .send_mss = tcp_session_send_mss,
+  .send_space = tcp_session_send_space,
+  .rx_fifo_offset = tcp_session_rx_fifo_offset,
+  .format_connection = format_tcp_session_ip6,
+  .format_listener = format_tcp_listener_session_ip6,
+  .format_half_open = format_tcp_half_open_session_ip6
+};
+/* *INDENT-ON* */
+
+void
+tcp_timer_keep_handler (u32 conn_index)
+{
+  u32 cpu_index = os_get_cpu_number ();
+  tcp_connection_t *tc;
+
+  tc = tcp_connection_get (conn_index, cpu_index);
+  tc->timers[TCP_TIMER_KEEP] = TCP_TIMER_HANDLE_INVALID;
+
+  tcp_connection_close (tc);
+}
+
+void
+tcp_timer_establish_handler (u32 conn_index)
+{
+  tcp_connection_t *tc;
+  u8 sst;
+
+  tc = tcp_half_open_connection_get (conn_index);
+  tc->timers[TCP_TIMER_ESTABLISH] = TCP_TIMER_HANDLE_INVALID;
+
+  ASSERT (tc->state == TCP_STATE_SYN_SENT);
+
+  sst = tc->c_is_ip4 ? SESSION_TYPE_IP4_TCP : SESSION_TYPE_IP6_TCP;
+  stream_session_connect_notify (&tc->connection, sst, 1 /* fail */ );
+
+  tcp_connection_cleanup (tc);
+}
+
+void
+tcp_timer_2msl_handler (u32 conn_index)
+{
+  u32 cpu_index = os_get_cpu_number ();
+  tcp_connection_t *tc;
+
+  tc = tcp_connection_get (conn_index, cpu_index);
+  tc->timers[TCP_TIMER_2MSL] = TCP_TIMER_HANDLE_INVALID;
+
+  tcp_connection_del (tc);
+}
+
+/* *INDENT-OFF* */
+static timer_expiration_handler *timer_expiration_handlers[TCP_N_TIMERS] =
+{
+    tcp_timer_retransmit_handler,
+    tcp_timer_delack_handler,
+    0,
+    tcp_timer_keep_handler,
+    tcp_timer_2msl_handler,
+    tcp_timer_retransmit_syn_handler,
+    tcp_timer_establish_handler
+};
+/* *INDENT-ON* */
+
+static void
+tcp_expired_timers_dispatch (u32 * expired_timers)
+{
+  int i;
+  u32 connection_index, timer_id;
+
+  for (i = 0; i < vec_len (expired_timers); i++)
+    {
+      /* Get session index and timer id */
+      connection_index = expired_timers[i] & 0x0FFFFFFF;
+      timer_id = expired_timers[i] >> 28;
+
+      /* Handle expiration */
+      (*timer_expiration_handlers[timer_id]) (connection_index);
+    }
+}
+
+void
+tcp_initialize_timer_wheels (tcp_main_t * tm)
+{
+  tw_timer_wheel_16t_2w_512sl_t *tw;
+  vec_foreach (tw, tm->timer_wheels)
+  {
+    tw_timer_wheel_init_16t_2w_512sl (tw, tcp_expired_timers_dispatch,
+				      100e-3 /* timer period 100ms */ , ~0);
+    tw->last_run_time = vlib_time_now (tm->vlib_main);
+  }
+}
+
+clib_error_t *
+tcp_init (vlib_main_t * vm)
+{
+  ip_main_t *im = &ip_main;
+  ip_protocol_info_t *pi;
+  tcp_main_t *tm = vnet_get_tcp_main ();
+  vlib_thread_main_t *vtm = vlib_get_thread_main ();
+  clib_error_t *error = 0;
+  u32 num_threads;
+
+  tm->vlib_main = vm;
+  tm->vnet_main = vnet_get_main ();
+
+  if ((error = vlib_call_init_function (vm, ip_main_init)))
+    return error;
+  if ((error = vlib_call_init_function (vm, ip4_lookup_init)))
+    return error;
+  if ((error = vlib_call_init_function (vm, ip6_lookup_init)))
+    return error;
+
+  /*
+   * Registrations
+   */
+
+  /* Register with IP */
+  pi = ip_get_protocol_info (im, IP_PROTOCOL_TCP);
+  if (pi == 0)
+    return clib_error_return (0, "TCP protocol info AWOL");
+  pi->format_header = format_tcp_header;
+  pi->unformat_pg_edit = unformat_pg_tcp_header;
+
+  ip4_register_protocol (IP_PROTOCOL_TCP, tcp4_input_node.index);
+
+  /* Register as transport with URI */
+  session_register_transport (SESSION_TYPE_IP4_TCP, &tcp4_proto);
+  session_register_transport (SESSION_TYPE_IP6_TCP, &tcp6_proto);
+
+  /*
+   * Initialize data structures
+   */
+
+  num_threads = 1 /* main thread */  + vtm->n_threads;
+  vec_validate (tm->connections, num_threads - 1);
+
+  /* Initialize per worker thread tx buffers (used for control messages) */
+  vec_validate (tm->tx_buffers, num_threads - 1);
+
+  /* Initialize timer wheels */
+  vec_validate (tm->timer_wheels, num_threads - 1);
+  tcp_initialize_timer_wheels (tm);
+
+  vec_validate (tm->delack_connections, num_threads - 1);
+
+  /* Initialize clocks per tick for TCP timestamp. Used to compute
+   * monotonically increasing timestamps. */
+  tm->tstamp_ticks_per_clock = vm->clib_time.seconds_per_clock
+    / TCP_TSTAMP_RESOLUTION;
+
+  clib_bihash_init_24_8 (&tm->local_endpoints_table, "local endpoint table",
+			 200000 /* $$$$ config parameter nbuckets */ ,
+			 (64 << 20) /*$$$ config parameter table size */ );
+
+  return error;
+}
+
+VLIB_INIT_FUNCTION (tcp_init);
+
+/*
+ * fd.io coding-style-patch-verification: ON
+ *
+ * Local Variables:
+ * eval: (c-set-style "gnu")
+ * End:
+ */
diff --git a/src/vnet/tcp/tcp.h b/src/vnet/tcp/tcp.h
new file mode 100644
index 00000000..22f00a63
--- /dev/null
+++ b/src/vnet/tcp/tcp.h
@@ -0,0 +1,624 @@
+/*
+ * Copyright (c) 2016 Cisco and/or its affiliates.
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at:
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#ifndef _vnet_tcp_h_
+#define _vnet_tcp_h_
+
+#include 
+#include 
+#include 
+#include 
+#include 
+#include 
+
+#define TCP_TICK 10e-3			/**< TCP tick period (s) */
+#define THZ 1/TCP_TICK			/**< TCP tick frequency */
+#define TCP_TSTAMP_RESOLUTION TCP_TICK	/**< Time stamp resolution */
+#define TCP_PAWS_IDLE 24 * 24 * 60 * 60 * THZ /**< 24 days */
+#define TCP_MAX_OPTION_SPACE 40
+
+#define TCP_DUPACK_THRESHOLD 3
+#define TCP_DEFAULT_RX_FIFO_SIZE 64 << 10
+
+/** TCP FSM state definitions as per RFC793. */
+#define foreach_tcp_fsm_state   \
+  _(CLOSED, "CLOSED")           \
+  _(LISTEN, "LISTEN")           \
+  _(SYN_SENT, "SYN_SENT")       \
+  _(SYN_RCVD, "SYN_RCVD")       \
+  _(ESTABLISHED, "ESTABLISHED") \
+  _(CLOSE_WAIT, "CLOSE_WAIT")   \
+  _(FIN_WAIT_1, "FIN_WAIT_1")   \
+  _(LAST_ACK, "LAST_ACK")       \
+  _(CLOSING, "CLOSING")         \
+  _(FIN_WAIT_2, "FIN_WAIT_2")   \
+  _(TIME_WAIT, "TIME_WAIT")
+
+typedef enum _tcp_state
+{
+#define _(sym, str) TCP_STATE_##sym,
+  foreach_tcp_fsm_state
+#undef _
+  TCP_N_STATES
+} tcp_state_t;
+
+format_function_t format_tcp_state;
+
+/** TCP timers */
+#define foreach_tcp_timer               \
+  _(RETRANSMIT, "RETRANSMIT")           \
+  _(DELACK, "DELAYED ACK")              \
+  _(PERSIST, "PERSIST")                 \
+  _(KEEP, "KEEP")                       \
+  _(2MSL, "2MSL")                       \
+  _(RETRANSMIT_SYN, "RETRANSMIT_SYN")   \
+  _(ESTABLISH, "ESTABLISH")
+
+typedef enum _tcp_timers
+{
+#define _(sym, str) TCP_TIMER_##sym,
+  foreach_tcp_timer
+#undef _
+  TCP_N_TIMERS
+} tcp_timers_e;
+
+typedef void (timer_expiration_handler) (u32 index);
+
+extern timer_expiration_handler tcp_timer_delack_handler;
+extern timer_expiration_handler tcp_timer_retransmit_handler;
+extern timer_expiration_handler tcp_timer_retransmit_syn_handler;
+
+#define TCP_TIMER_HANDLE_INVALID ((u32) ~0)
+
+/* Timer delays as multiples of 100ms */
+#define TCP_TO_TIMER_TICK       TCP_TICK*10	/* Period for converting from TCP
+						 * ticks to timer units */
+#define TCP_DELACK_TIME         1	/* 0.1s */
+#define TCP_ESTABLISH_TIME      750	/* 75s */
+#define TCP_2MSL_TIME           300	/* 30s */
+
+#define TCP_RTO_MAX 60 * THZ	/* Min max RTO (60s) as per RFC6298 */
+#define TCP_RTT_MAX 30 * THZ	/* 30s (probably too much) */
+#define TCP_RTO_SYN_RETRIES 3	/* SYN retries without doubling RTO */
+#define TCP_RTO_INIT 1 * THZ	/* Initial retransmit timer */
+
+void tcp_update_time (f64 now, u32 thread_index);
+
+/** TCP connection flags */
+#define foreach_tcp_connection_flag             \
+  _(DELACK, "Delay ACK")                        \
+  _(SNDACK, "Send ACK")                         \
+  _(BURSTACK, "Burst ACK set")                  \
+  _(SENT_RCV_WND0, "Sent 0 receive window")     \
+  _(RECOVERY, "Recovery on")                    \
+  _(FAST_RECOVERY, "Fast Recovery on")
+
+typedef enum _tcp_connection_flag_bits
+{
+#define _(sym, str) TCP_CONN_##sym##_BIT,
+  foreach_tcp_connection_flag
+#undef _
+  TCP_CONN_N_FLAG_BITS
+} tcp_connection_flag_bits_e;
+
+typedef enum _tcp_connection_flag
+{
+#define _(sym, str) TCP_CONN_##sym = 1 << TCP_CONN_##sym##_BIT,
+  foreach_tcp_connection_flag
+#undef _
+  TCP_CONN_N_FLAGS
+} tcp_connection_flags_e;
+
+/** TCP buffer flags */
+#define foreach_tcp_buf_flag                            \
+  _ (ACK)       /**< Sending ACK. */                    \
+  _ (DUPACK)    /**< Sending DUPACK. */                 \
+
+enum
+{
+#define _(f) TCP_BUF_BIT_##f,
+  foreach_tcp_buf_flag
+#undef _
+    TCP_N_BUF_BITS,
+};
+
+enum
+{
+#define _(f) TCP_BUF_FLAG_##f = 1 << TCP_BUF_BIT_##f,
+  foreach_tcp_buf_flag
+#undef _
+};
+
+#define TCP_MAX_SACK_BLOCKS 5	/**< Max number of SACK blocks stored */
+#define TCP_INVALID_SACK_HOLE_INDEX ((u32)~0)
+
+typedef struct _sack_scoreboard_hole
+{
+  u32 next;		/**< Index for next entry in linked list */
+  u32 prev;		/**< Index for previous entry in linked list */
+  u32 start;		/**< Start sequence number */
+  u32 end;		/**< End sequence number */
+} sack_scoreboard_hole_t;
+
+typedef struct _sack_scoreboard
+{
+  sack_scoreboard_hole_t *holes;	/**< Pool of holes */
+  u32 head;				/**< Index to first entry */
+  u32 sacked_bytes;			/**< Number of bytes sacked in sb */
+} sack_scoreboard_t;
+
+typedef enum _tcp_cc_algorithm_type
+{
+  TCP_CC_NEWRENO,
+} tcp_cc_algorithm_type_e;
+
+typedef struct _tcp_cc_algorithm tcp_cc_algorithm_t;
+
+typedef enum _tcp_cc_ack_t
+{
+  TCP_CC_ACK,
+  TCP_CC_DUPACK,
+  TCP_CC_PARTIALACK
+} tcp_cc_ack_t;
+
+typedef struct _tcp_connection
+{
+  transport_connection_t connection;  /**< Common transport data. First! */
+
+  u8 state;			/**< TCP state as per tcp_state_t */
+  u16 flags;			/**< Connection flags (see tcp_conn_flags_e) */
+  u32 timers[TCP_N_TIMERS];	/**< Timer handles into timer wheel */
+
+  /* TODO RFC4898 */
+
+  /** Send sequence variables RFC793 */
+  u32 snd_una;		/**< oldest unacknowledged sequence number */
+  u32 snd_una_max;	/**< newest unacknowledged sequence number + 1*/
+  u32 snd_wnd;		/**< send window */
+  u32 snd_wl1;		/**< seq number used for last snd.wnd update */
+  u32 snd_wl2;		/**< ack number used for last snd.wnd update */
+  u32 snd_nxt;		/**< next seq number to be sent */
+
+  /** Receive sequence variables RFC793 */
+  u32 rcv_nxt;		/**< next sequence number expected */
+  u32 rcv_wnd;		/**< receive window we expect */
+
+  u32 rcv_las;		/**< rcv_nxt at last ack sent/rcv_wnd update */
+  u32 iss;		/**< initial sent sequence */
+  u32 irs;		/**< initial remote sequence */
+
+  /* Options */
+  tcp_options_t opt;	/**< TCP connection options parsed */
+  u8 rcv_wscale;	/**< Window scale to advertise to peer */
+  u8 snd_wscale;	/**< Window scale to use when sending */
+  u32 tsval_recent;	/**< Last timestamp received */
+  u32 tsval_recent_age;	/**< When last updated tstamp_recent*/
+
+  sack_block_t *snd_sacks;	/**< Vector of SACKs to send. XXX Fixed size? */
+  sack_scoreboard_t sack_sb;	/**< SACK "scoreboard" that tracks holes */
+
+  u8 rcv_dupacks;	/**< Number of DUPACKs received */
+  u8 snt_dupacks;	/**< Number of DUPACKs sent in a burst */
+
+  /* Congestion control */
+  u32 cwnd;		/**< Congestion window */
+  u32 ssthresh;		/**< Slow-start threshold */
+  u32 prev_ssthresh;	/**< ssthresh before congestion */
+  u32 bytes_acked;	/**< Bytes acknowledged by current segment */
+  u32 rtx_bytes;	/**< Retransmitted bytes */
+  u32 tsecr_last_ack;	/**< Timestamp echoed to us in last health ACK */
+  tcp_cc_algorithm_t *cc_algo;	/**< Congestion control algorithm */
+
+  /* RTT and RTO */
+  u32 rto;		/**< Retransmission timeout */
+  u32 rto_boff;		/**< Index for RTO backoff */
+  u32 srtt;		/**< Smoothed RTT */
+  u32 rttvar;		/**< Smoothed mean RTT difference. Approximates variance */
+  u32 rtt_ts;		/**< Timestamp for tracked ACK */
+  u32 rtt_seq;		/**< Sequence number for tracked ACK */
+
+  u16 snd_mss;		/**< Send MSS */
+} tcp_connection_t;
+
+struct _tcp_cc_algorithm
+{
+  void (*rcv_ack) (tcp_connection_t * tc);
+  void (*rcv_cong_ack) (tcp_connection_t * tc, tcp_cc_ack_t ack);
+  void (*congestion) (tcp_connection_t * tc);
+  void (*recovered) (tcp_connection_t * tc);
+  void (*init) (tcp_connection_t * tc);
+};
+
+#define tcp_fastrecovery_on(tc) (tc)->flags |= TCP_CONN_FAST_RECOVERY
+#define tcp_fastrecovery_off(tc) (tc)->flags &= ~TCP_CONN_FAST_RECOVERY
+#define tcp_in_fastrecovery(tc) ((tc)->flags & TCP_CONN_FAST_RECOVERY)
+#define tcp_in_recovery(tc) ((tc)->flags & (TCP_CONN_FAST_RECOVERY | TCP_CONN_RECOVERY))
+#define tcp_recovery_off(tc) ((tc)->flags &= ~(TCP_CONN_FAST_RECOVERY | TCP_CONN_RECOVERY))
+#define tcp_in_slowstart(tc) (tc->cwnd < tc->ssthresh)
+
+typedef enum
+{
+  TCP_IP4,
+  TCP_IP6,
+  TCP_N_AF,
+} tcp_af_t;
+
+typedef enum _tcp_error
+{
+#define tcp_error(n,s) TCP_ERROR_##n,
+#include 
+#undef tcp_error
+  TCP_N_ERROR,
+} tcp_error_t;
+
+typedef struct _tcp_lookup_dispatch
+{
+  u8 next, error;
+} tcp_lookup_dispatch_t;
+
+typedef struct _tcp_main
+{
+  /* Per-worker thread tcp connection pools */
+  tcp_connection_t **connections;
+
+  /* Pool of listeners. */
+  tcp_connection_t *listener_pool;
+
+  /** Dispatch table by state and flags */
+  tcp_lookup_dispatch_t dispatch_table[TCP_N_STATES][64];
+
+  u8 log2_tstamp_clocks_per_tick;
+  f64 tstamp_ticks_per_clock;
+
+  /** per-worker tx buffer free lists */
+  u32 **tx_buffers;
+
+  /* Per worker-thread timer wheel for connections timers */
+  tw_timer_wheel_16t_2w_512sl_t *timer_wheels;
+
+  /* Convenience per worker-thread vector of connections to DELACK */
+  u32 **delack_connections;
+
+  /* Pool of half-open connections on which we've sent a SYN */
+  tcp_connection_t *half_open_connections;
+
+  /* Pool of local TCP endpoints */
+  transport_endpoint_t *local_endpoints;
+
+  /* Local endpoints lookup table */
+  transport_endpoint_table_t local_endpoints_table;
+
+  /* Congestion control algorithms registered */
+  tcp_cc_algorithm_t *cc_algos;
+
+  /* convenience */
+  vlib_main_t *vlib_main;
+  vnet_main_t *vnet_main;
+  ip4_main_t *ip4_main;
+  ip6_main_t *ip6_main;
+} tcp_main_t;
+
+extern tcp_main_t tcp_main;
+extern vlib_node_registration_t tcp4_input_node;
+extern vlib_node_registration_t tcp6_input_node;
+extern vlib_node_registration_t tcp4_output_node;
+extern vlib_node_registration_t tcp6_output_node;
+
+always_inline tcp_main_t *
+vnet_get_tcp_main ()
+{
+  return &tcp_main;
+}
+
+always_inline tcp_connection_t *
+tcp_connection_get (u32 conn_index, u32 thread_index)
+{
+  return pool_elt_at_index (tcp_main.connections[thread_index], conn_index);
+}
+
+always_inline tcp_connection_t *
+tcp_connection_get_if_valid (u32 conn_index, u32 thread_index)
+{
+  if (tcp_main.connections[thread_index] == 0)
+    return 0;
+  if (pool_is_free_index (tcp_main.connections[thread_index], conn_index))
+    return 0;
+  return pool_elt_at_index (tcp_main.connections[thread_index], conn_index);
+}
+
+void tcp_connection_close (tcp_connection_t * tc);
+void tcp_connection_cleanup (tcp_connection_t * tc);
+void tcp_connection_del (tcp_connection_t * tc);
+
+always_inline tcp_connection_t *
+tcp_listener_get (u32 tli)
+{
+  return pool_elt_at_index (tcp_main.listener_pool, tli);
+}
+
+always_inline tcp_connection_t *
+tcp_half_open_connection_get (u32 conn_index)
+{
+  return pool_elt_at_index (tcp_main.half_open_connections, conn_index);
+}
+
+void tcp_make_ack (tcp_connection_t * ts, vlib_buffer_t * b);
+void tcp_make_finack (tcp_connection_t * tc, vlib_buffer_t * b);
+void tcp_make_synack (tcp_connection_t * ts, vlib_buffer_t * b);
+void tcp_send_reset (vlib_buffer_t * pkt, u8 is_ip4);
+void tcp_send_syn (tcp_connection_t * tc);
+void tcp_send_fin (tcp_connection_t * tc);
+void tcp_set_snd_mss (tcp_connection_t * tc);
+
+always_inline u32
+tcp_end_seq (tcp_header_t * th, u32 len)
+{
+  return th->seq_number + tcp_is_syn (th) + tcp_is_fin (th) + len;
+}
+
+/* Modulo arithmetic for TCP sequence numbers */
+#define seq_lt(_s1, _s2) ((i32)((_s1)-(_s2)) < 0)
+#define seq_leq(_s1, _s2) ((i32)((_s1)-(_s2)) <= 0)
+#define seq_gt(_s1, _s2) ((i32)((_s1)-(_s2)) > 0)
+#define seq_geq(_s1, _s2) ((i32)((_s1)-(_s2)) >= 0)
+
+/* Modulo arithmetic for timestamps */
+#define timestamp_lt(_t1, _t2) ((i32)((_t1)-(_t2)) < 0)
+#define timestamp_leq(_t1, _t2) ((i32)((_t1)-(_t2)) <= 0)
+
+always_inline u32
+tcp_flight_size (const tcp_connection_t * tc)
+{
+  return tc->snd_una_max - tc->snd_una - tc->sack_sb.sacked_bytes
+    + tc->rtx_bytes;
+}
+
+/**
+ * Initial cwnd as per RFC5681
+ */
+always_inline u32
+tcp_initial_cwnd (const tcp_connection_t * tc)
+{
+  if (tc->snd_mss > 2190)
+    return 2 * tc->snd_mss;
+  else if (tc->snd_mss > 1095)
+    return 3 * tc->snd_mss;
+  else
+    return 4 * tc->snd_mss;
+}
+
+always_inline u32
+tcp_loss_wnd (const tcp_connection_t * tc)
+{
+  return tc->snd_mss;
+}
+
+always_inline u32
+tcp_available_wnd (const tcp_connection_t * tc)
+{
+  return clib_min (tc->cwnd, tc->snd_wnd);
+}
+
+always_inline u32
+tcp_available_snd_space (const tcp_connection_t * tc)
+{
+  u32 available_wnd = tcp_available_wnd (tc);
+  u32 flight_size = tcp_flight_size (tc);
+
+  if (available_wnd <= flight_size)
+    return 0;
+
+  return available_wnd - flight_size;
+}
+
+void tcp_retransmit_first_unacked (tcp_connection_t * tc);
+
+void tcp_fast_retransmit (tcp_connection_t * tc);
+
+always_inline u32
+tcp_time_now (void)
+{
+  return clib_cpu_time_now () * tcp_main.tstamp_ticks_per_clock;
+}
+
+u32 tcp_push_header (transport_connection_t * tconn, vlib_buffer_t * b);
+
+u32
+tcp_prepare_retransmit_segment (tcp_connection_t * tc, vlib_buffer_t * b,
+				u32 max_bytes);
+
+void tcp_connection_timers_init (tcp_connection_t * tc);
+void tcp_connection_timers_reset (tcp_connection_t * tc);
+
+void tcp_connection_init_vars (tcp_connection_t * tc);
+
+always_inline void
+tcp_connection_force_ack (tcp_connection_t * tc, vlib_buffer_t * b)
+{
+  /* Reset flags, make sure ack is sent */
+  tc->flags = TCP_CONN_SNDACK;
+  vnet_buffer (b)->tcp.flags &= ~TCP_BUF_FLAG_DUPACK;
+}
+
+always_inline void
+tcp_timer_set (tcp_connection_t * tc, u8 timer_id, u32 interval)
+{
+  tc->timers[timer_id]
+    = tw_timer_start_16t_2w_512sl (&tcp_main.timer_wheels[tc->c_thread_index],
+				   tc->c_c_index, timer_id, interval);
+}
+
+always_inline void
+tcp_retransmit_timer_set (tcp_main_t * tm, tcp_connection_t * tc)
+{
+  /* XXX Switch to faster TW */
+  tcp_timer_set (tc, TCP_TIMER_RETRANSMIT,
+		 clib_max (tc->rto * TCP_TO_TIMER_TICK, 1));
+}
+
+always_inline void
+tcp_timer_reset (tcp_connection_t * tc, u8 timer_id)
+{
+  if (tc->timers[timer_id] == TCP_TIMER_HANDLE_INVALID)
+    return;
+
+  tw_timer_stop_16t_2w_512sl (&tcp_main.timer_wheels[tc->c_thread_index],
+			      tc->timers[timer_id]);
+  tc->timers[timer_id] = TCP_TIMER_HANDLE_INVALID;
+}
+
+always_inline void
+tcp_timer_update (tcp_connection_t * tc, u8 timer_id, u32 interval)
+{
+  if (tc->timers[timer_id] != TCP_TIMER_HANDLE_INVALID)
+    tw_timer_stop_16t_2w_512sl (&tcp_main.timer_wheels[tc->c_thread_index],
+				tc->timers[timer_id]);
+  tc->timers[timer_id] =
+    tw_timer_start_16t_2w_512sl (&tcp_main.timer_wheels[tc->c_thread_index],
+				 tc->c_c_index, timer_id, interval);
+}
+
+always_inline u8
+tcp_timer_is_active (tcp_connection_t * tc, tcp_timers_e timer)
+{
+  return tc->timers[timer] != TCP_TIMER_HANDLE_INVALID;
+}
+
+void
+scoreboard_remove_hole (sack_scoreboard_t * sb,
+			sack_scoreboard_hole_t * hole);
+
+always_inline sack_scoreboard_hole_t *
+scoreboard_next_hole (sack_scoreboard_t * sb, sack_scoreboard_hole_t * hole)
+{
+  if (hole->next != TCP_INVALID_SACK_HOLE_INDEX)
+    return pool_elt_at_index (sb->holes, hole->next);
+  return 0;
+}
+
+always_inline sack_scoreboard_hole_t *
+scoreboard_first_hole (sack_scoreboard_t * sb)
+{
+  if (sb->head != TCP_INVALID_SACK_HOLE_INDEX)
+    return pool_elt_at_index (sb->holes, sb->head);
+  return 0;
+}
+
+always_inline void
+scoreboard_clear (sack_scoreboard_t * sb)
+{
+  sack_scoreboard_hole_t *hole = scoreboard_first_hole (sb);
+  while ((hole = scoreboard_first_hole (sb)))
+    {
+      scoreboard_remove_hole (sb, hole);
+    }
+}
+
+always_inline u32
+scoreboard_hole_bytes (sack_scoreboard_hole_t * hole)
+{
+  return hole->end - hole->start;
+}
+
+always_inline void
+tcp_cc_algo_register (tcp_cc_algorithm_type_e type,
+		      const tcp_cc_algorithm_t * vft)
+{
+  tcp_main_t *tm = vnet_get_tcp_main ();
+  vec_validate (tm->cc_algos, type);
+
+  tm->cc_algos[type] = *vft;
+}
+
+always_inline tcp_cc_algorithm_t *
+tcp_cc_algo_get (tcp_cc_algorithm_type_e type)
+{
+  tcp_main_t *tm = vnet_get_tcp_main ();
+  return &tm->cc_algos[type];
+}
+
+void tcp_cc_init (tcp_connection_t * tc);
+
+/**
+ * Push TCP header to buffer
+ *
+ * @param vm - vlib_main
+ * @param b - buffer to write the header to
+ * @param sp_net - source port net order
+ * @param dp_net - destination port net order
+ * @param seq - sequence number net order
+ * @param ack - ack number net order
+ * @param tcp_hdr_opts_len - header and options length in bytes
+ * @param flags - header flags
+ * @param wnd - window size
+ *
+ * @return - pointer to start of TCP header
+ */
+always_inline void *
+vlib_buffer_push_tcp_net_order (vlib_buffer_t * b, u16 sp, u16 dp, u32 seq,
+				u32 ack, u8 tcp_hdr_opts_len, u8 flags,
+				u16 wnd)
+{
+  tcp_header_t *th;
+
+  th = vlib_buffer_push_uninit (b, tcp_hdr_opts_len);
+
+  th->src_port = sp;
+  th->dst_port = dp;
+  th->seq_number = seq;
+  th->ack_number = ack;
+  th->data_offset_and_reserved = (tcp_hdr_opts_len >> 2) << 4;
+  th->flags = flags;
+  th->window = wnd;
+  th->checksum = 0;
+  th->urgent_pointer = 0;
+  return th;
+}
+
+/**
+ * Push TCP header to buffer
+ *
+ * @param vm - vlib_main
+ * @param b - buffer to write the header to
+ * @param sp_net - source port net order
+ * @param dp_net - destination port net order
+ * @param seq - sequence number host order
+ * @param ack - ack number host order
+ * @param tcp_hdr_opts_len - header and options length in bytes
+ * @param flags - header flags
+ * @param wnd - window size
+ *
+ * @return - pointer to start of TCP header
+ */
+always_inline void *
+vlib_buffer_push_tcp (vlib_buffer_t * b, u16 sp_net, u16 dp_net, u32 seq,
+		      u32 ack, u8 tcp_hdr_opts_len, u8 flags, u16 wnd)
+{
+  return vlib_buffer_push_tcp_net_order (b, sp_net, dp_net,
+					 clib_host_to_net_u32 (seq),
+					 clib_host_to_net_u32 (ack),
+					 tcp_hdr_opts_len, flags,
+					 clib_host_to_net_u16 (wnd));
+}
+
+#endif /* _vnet_tcp_h_ */
+
+/*
+ * fd.io coding-style-patch-verification: ON
+ *
+ * Local Variables:
+ * eval: (c-set-style "gnu")
+ * End:
+ */
diff --git a/src/vnet/tcp/tcp_error.def b/src/vnet/tcp/tcp_error.def
new file mode 100644
index 00000000..cff5ec13
--- /dev/null
+++ b/src/vnet/tcp/tcp_error.def
@@ -0,0 +1,35 @@
+/*
+ * Copyright (c) 2016 Cisco and/or its affiliates.
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at:
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+ 
+tcp_error (NONE, "no error")
+tcp_error (NO_LISTENER, "no listener for dst port")
+tcp_error (LOOKUP_DROPS, "lookup drops")
+tcp_error (DISPATCH, "Dispatch error")
+tcp_error (ENQUEUED, "Packets pushed into rx fifo")                              
+tcp_error (PURE_ACK, "Pure acks")
+tcp_error (SYNS_RCVD, "SYNs received")
+tcp_error (SYN_ACKS_RCVD, "SYN-ACKs received")
+tcp_error (NOT_READY, "Session not ready for packets")                               
+tcp_error (FIFO_FULL, "Packets dropped for lack of rx fifo space")               
+tcp_error (EVENT_FIFO_FULL, "Events not sent for lack of event fifo space")      
+tcp_error (API_QUEUE_FULL, "Sessions not created for lack of API queue space")
+tcp_error (CREATE_SESSION_FAIL, "Sessions couldn't be allocated")
+tcp_error (SEGMENT_INVALID, "Invalid segment")
+tcp_error (ACK_INVALID, "Invalid ACK")
+tcp_error (ACK_DUP, "Duplicate ACK")
+tcp_error (ACK_OLD, "Old ACK")
+tcp_error (PKTS_SENT, "Packets sent")
+tcp_error (FILTERED_DUPACKS, "Filtered duplicate ACKs")
+tcp_error (RST_SENT, "Resets sent")
\ No newline at end of file
diff --git a/src/vnet/tcp/tcp_format.c b/src/vnet/tcp/tcp_format.c
new file mode 100644
index 00000000..7136741d
--- /dev/null
+++ b/src/vnet/tcp/tcp_format.c
@@ -0,0 +1,136 @@
+/*
+ * Copyright (c) 2015 Cisco and/or its affiliates.
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at:
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+/*
+ * tcp/tcp_format.c: tcp formatting
+ *
+ * Copyright (c) 2008 Eliot Dresselhaus
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining
+ * a copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sublicense, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be
+ * included in all copies or substantial portions of the Software.
+ *
+ *  THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ *  EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ *  MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ *  NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE
+ *  LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
+ *  OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
+ *  WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+#include 
+#include 
+
+static u8 *
+format_tcp_flags (u8 * s, va_list * args)
+{
+  int flags = va_arg (*args, int);
+
+#define _(f) if (flags & TCP_FLAG_##f) s = format (s, "%s, ", #f);
+  foreach_tcp_flag
+#undef _
+    return s;
+}
+
+/* Format TCP header. */
+u8 *
+format_tcp_header (u8 * s, va_list * args)
+{
+  tcp_header_t *tcp = va_arg (*args, tcp_header_t *);
+  u32 max_header_bytes = va_arg (*args, u32);
+  u32 header_bytes;
+  uword indent;
+
+  /* Nothing to do. */
+  if (max_header_bytes < sizeof (tcp[0]))
+    return format (s, "TCP header truncated");
+
+  indent = format_get_indent (s);
+  indent += 2;
+  header_bytes = tcp_header_bytes (tcp);
+
+  s = format (s, "TCP: %d -> %d", clib_net_to_host_u16 (tcp->src),
+	      clib_net_to_host_u16 (tcp->dst));
+
+  s = format (s, "\n%Useq. 0x%08x ack 0x%08x", format_white_space, indent,
+	      clib_net_to_host_u32 (tcp->seq_number),
+	      clib_net_to_host_u32 (tcp->ack_number));
+
+  s = format (s, "\n%Uflags %U, tcp header: %d bytes", format_white_space,
+	      indent, format_tcp_flags, tcp->flags, header_bytes);
+
+  s = format (s, "\n%Uwindow %d, checksum 0x%04x", format_white_space, indent,
+	      clib_net_to_host_u16 (tcp->window),
+	      clib_net_to_host_u16 (tcp->checksum));
+
+
+#if 0
+  /* Format TCP options. */
+  {
+    u8 *o;
+    u8 *option_start = (void *) (tcp + 1);
+    u8 *option_end = (void *) tcp + header_bytes;
+
+    for (o = option_start; o < option_end;)
+      {
+	u32 length = o[1];
+	switch (o[0])
+	  {
+	  case TCP_OPTION_END:
+	    length = 1;
+	    o = option_end;
+	    break;
+
+	  case TCP_OPTION_NOOP:
+	    length = 1;
+	    break;
+
+	  }
+      }
+  }
+#endif
+
+  /* Recurse into next protocol layer. */
+  if (max_header_bytes != 0 && header_bytes < max_header_bytes)
+    {
+      ip_main_t *im = &ip_main;
+      tcp_udp_port_info_t *pi;
+
+      pi = ip_get_tcp_udp_port_info (im, tcp->dst);
+
+      if (pi && pi->format_header)
+	s = format (s, "\n%U%U", format_white_space, indent - 2,
+		    pi->format_header,
+		    /* next protocol header */ (void *) tcp + header_bytes,
+		    max_header_bytes - header_bytes);
+    }
+
+  return s;
+}
+
+/*
+ * fd.io coding-style-patch-verification: ON
+ *
+ * Local Variables:
+ * eval: (c-set-style "gnu")
+ * End:
+ */
diff --git a/src/vnet/tcp/tcp_input.c b/src/vnet/tcp/tcp_input.c
new file mode 100644
index 00000000..daa0683b
--- /dev/null
+++ b/src/vnet/tcp/tcp_input.c
@@ -0,0 +1,2316 @@
+/*
+ * Copyright (c) 2016 Cisco and/or its affiliates.
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at:
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#include 
+#include 
+#include 
+#include 
+#include 
+
+static char *tcp_error_strings[] = {
+#define tcp_error(n,s) s,
+#include 
+#undef tcp_error
+};
+
+/* All TCP nodes have the same outgoing arcs */
+#define foreach_tcp_state_next                  \
+  _ (DROP, "error-drop")                        \
+  _ (TCP4_OUTPUT, "tcp4-output")                \
+  _ (TCP6_OUTPUT, "tcp6-output")
+
+typedef enum _tcp_established_next
+{
+#define _(s,n) TCP_ESTABLISHED_NEXT_##s,
+  foreach_tcp_state_next
+#undef _
+    TCP_ESTABLISHED_N_NEXT,
+} tcp_established_next_t;
+
+typedef enum _tcp_rcv_process_next
+{
+#define _(s,n) TCP_RCV_PROCESS_NEXT_##s,
+  foreach_tcp_state_next
+#undef _
+    TCP_RCV_PROCESS_N_NEXT,
+} tcp_rcv_process_next_t;
+
+typedef enum _tcp_syn_sent_next
+{
+#define _(s,n) TCP_SYN_SENT_NEXT_##s,
+  foreach_tcp_state_next
+#undef _
+    TCP_SYN_SENT_N_NEXT,
+} tcp_syn_sent_next_t;
+
+typedef enum _tcp_listen_next
+{
+#define _(s,n) TCP_LISTEN_NEXT_##s,
+  foreach_tcp_state_next
+#undef _
+    TCP_LISTEN_N_NEXT,
+} tcp_listen_next_t;
+
+/* Generic, state independent indices */
+typedef enum _tcp_state_next
+{
+#define _(s,n) TCP_NEXT_##s,
+  foreach_tcp_state_next
+#undef _
+    TCP_STATE_N_NEXT,
+} tcp_state_next_t;
+
+#define tcp_next_output(is_ip4) (is_ip4 ? TCP_NEXT_TCP4_OUTPUT          \
+                                        : TCP_NEXT_TCP6_OUTPUT)
+
+vlib_node_registration_t tcp4_established_node;
+vlib_node_registration_t tcp6_established_node;
+
+/**
+ * Validate segment sequence number. As per RFC793:
+ *
+ * Segment Receive Test
+ *      Length  Window
+ *      ------- -------  -------------------------------------------
+ *      0       0       SEG.SEQ = RCV.NXT
+ *      0       >0      RCV.NXT =< SEG.SEQ < RCV.NXT+RCV.WND
+ *      >0      0       not acceptable
+ *      >0      >0      RCV.NXT =< SEG.SEQ < RCV.NXT+RCV.WND
+ *                      or RCV.NXT =< SEG.SEQ+SEG.LEN-1 < RCV.NXT+RCV.WND
+ *
+ * This ultimately consists in checking if segment falls within the window.
+ * The one important difference compared to RFC793 is that we use rcv_las,
+ * or the rcv_nxt at last ack sent instead of rcv_nxt since that's the
+ * peer's reference when computing our receive window.
+ *
+ * This accepts only segments within the window.
+ */
+always_inline u8
+tcp_segment_in_rcv_wnd (tcp_connection_t * tc, u32 seq, u32 end_seq)
+{
+  return seq_leq (end_seq, tc->rcv_las + tc->rcv_wnd)
+    && seq_geq (seq, tc->rcv_nxt);
+}
+
+void
+tcp_options_parse (tcp_header_t * th, tcp_options_t * to)
+{
+  const u8 *data;
+  u8 opt_len, opts_len, kind;
+  int j;
+  sack_block_t b;
+
+  opts_len = (tcp_doff (th) << 2) - sizeof (tcp_header_t);
+  data = (const u8 *) (th + 1);
+
+  /* Zero out all flags but those set in SYN */
+  to->flags &= (TCP_OPTS_FLAG_SACK_PERMITTED | TCP_OPTS_FLAG_WSCALE);
+
+  for (; opts_len > 0; opts_len -= opt_len, data += opt_len)
+    {
+      kind = data[0];
+
+      /* Get options length */
+      if (kind == TCP_OPTION_EOL)
+	break;
+      else if (kind == TCP_OPTION_NOOP)
+	opt_len = 1;
+      else
+	{
+	  /* broken options */
+	  if (opts_len < 2)
+	    break;
+	  opt_len = data[1];
+
+	  /* weird option length */
+	  if (opt_len < 2 || opt_len > opts_len)
+	    break;
+	}
+
+      /* Parse options */
+      switch (kind)
+	{
+	case TCP_OPTION_MSS:
+	  if ((opt_len == TCP_OPTION_LEN_MSS) && tcp_syn (th))
+	    {
+	      to->flags |= TCP_OPTS_FLAG_MSS;
+	      to->mss = clib_net_to_host_u16 (*(u16 *) (data + 2));
+	    }
+	  break;
+	case TCP_OPTION_WINDOW_SCALE:
+	  if ((opt_len == TCP_OPTION_LEN_WINDOW_SCALE) && tcp_syn (th))
+	    {
+	      to->flags |= TCP_OPTS_FLAG_WSCALE;
+	      to->wscale = data[2];
+	      if (to->wscale > TCP_MAX_WND_SCALE)
+		{
+		  clib_warning ("Illegal window scaling value: %d",
+				to->wscale);
+		  to->wscale = TCP_MAX_WND_SCALE;
+		}
+	    }
+	  break;
+	case TCP_OPTION_TIMESTAMP:
+	  if (opt_len == TCP_OPTION_LEN_TIMESTAMP)
+	    {
+	      to->flags |= TCP_OPTS_FLAG_TSTAMP;
+	      to->tsval = clib_net_to_host_u32 (*(u32 *) (data + 2));
+	      to->tsecr = clib_net_to_host_u32 (*(u32 *) (data + 6));
+	    }
+	  break;
+	case TCP_OPTION_SACK_PERMITTED:
+	  if (opt_len == TCP_OPTION_LEN_SACK_PERMITTED && tcp_syn (th))
+	    to->flags |= TCP_OPTS_FLAG_SACK_PERMITTED;
+	  break;
+	case TCP_OPTION_SACK_BLOCK:
+	  /* If SACK permitted was not advertised or a SYN, break */
+	  if ((to->flags & TCP_OPTS_FLAG_SACK_PERMITTED) == 0 || tcp_syn (th))
+	    break;
+
+	  /* If too short or not correctly formatted, break */
+	  if (opt_len < 10 || ((opt_len - 2) % TCP_OPTION_LEN_SACK_BLOCK))
+	    break;
+
+	  to->flags |= TCP_OPTS_FLAG_SACK;
+	  to->n_sack_blocks = (opt_len - 2) / TCP_OPTION_LEN_SACK_BLOCK;
+	  vec_reset_length (to->sacks);
+	  for (j = 0; j < to->n_sack_blocks; j++)
+	    {
+	      b.start = clib_net_to_host_u32 (*(u32 *) (data + 2 + 4 * j));
+	      b.end = clib_net_to_host_u32 (*(u32 *) (data + 6 + 4 * j));
+	      vec_add1 (to->sacks, b);
+	    }
+	  break;
+	default:
+	  /* Nothing to see here */
+	  continue;
+	}
+    }
+}
+
+always_inline int
+tcp_segment_check_paws (tcp_connection_t * tc)
+{
+  /* XXX normally test for timestamp should be lt instead of leq, but for
+   * local testing this is not enough */
+  return tcp_opts_tstamp (&tc->opt) && tc->tsval_recent
+    && timestamp_lt (tc->opt.tsval, tc->tsval_recent);
+}
+
+/**
+ * Validate incoming segment as per RFC793 p. 69 and RFC1323 p. 19
+ *
+ * It first verifies if segment has a wrapped sequence number (PAWS) and then
+ * does the processing associated to the first four steps (ignoring security
+ * and precedence): sequence number, rst bit and syn bit checks.
+ *
+ * @return 0 if segments passes validation.
+ */
+static int
+tcp_segment_validate (vlib_main_t * vm, tcp_connection_t * tc0,
+		      vlib_buffer_t * b0, tcp_header_t * th0, u32 * next0)
+{
+  u8 paws_failed;
+
+  if (PREDICT_FALSE (!tcp_ack (th0) && !tcp_rst (th0) && !tcp_syn (th0)))
+    return -1;
+
+  tcp_options_parse (th0, &tc0->opt);
+
+  /* RFC1323: Check against wrapped sequence numbers (PAWS). If we have
+   * timestamp to echo and it's less than tsval_recent, drop segment
+   * but still send an ACK in order to retain TCP's mechanism for detecting
+   * and recovering from half-open connections */
+  paws_failed = tcp_segment_check_paws (tc0);
+  if (paws_failed)
+    {
+      clib_warning ("paws failed");
+
+      /* If it just so happens that a segment updates tsval_recent for a
+       * segment over 24 days old, invalidate tsval_recent. */
+      if (timestamp_lt (tc0->tsval_recent_age + TCP_PAWS_IDLE,
+			tcp_time_now ()))
+	{
+	  /* Age isn't reset until we get a valid tsval (bsd inspired) */
+	  tc0->tsval_recent = 0;
+	}
+      else
+	{
+	  /* Drop after ack if not rst */
+	  if (!tcp_rst (th0))
+	    {
+	      tcp_make_ack (tc0, b0);
+	      *next0 = tcp_next_output (tc0->c_is_ip4);
+	      return -1;
+	    }
+	}
+    }
+
+  /* 1st: check sequence number */
+  if (!tcp_segment_in_rcv_wnd (tc0, vnet_buffer (b0)->tcp.seq_number,
+			       vnet_buffer (b0)->tcp.seq_end))
+    {
+      if (!tcp_rst (th0))
+	{
+	  /* Send dup ack */
+	  tcp_make_ack (tc0, b0);
+	  *next0 = tcp_next_output (tc0->c_is_ip4);
+	}
+      return -1;
+    }
+
+  /* 2nd: check the RST bit */
+  if (tcp_rst (th0))
+    {
+      /* Notify session that connection has been reset. Switch
+       * state to closed and await for session to do the cleanup. */
+      stream_session_reset_notify (&tc0->connection);
+      tc0->state = TCP_STATE_CLOSED;
+      return -1;
+    }
+
+  /* 3rd: check security and precedence (skip) */
+
+  /* 4th: check the SYN bit */
+  if (tcp_syn (th0))
+    {
+      tcp_send_reset (b0, tc0->c_is_ip4);
+      return -1;
+    }
+
+  /* If PAWS passed and segment in window, save timestamp */
+  if (!paws_failed)
+    {
+      tc0->tsval_recent = tc0->opt.tsval;
+      tc0->tsval_recent_age = tcp_time_now ();
+    }
+
+  return 0;
+}
+
+always_inline int
+tcp_rcv_ack_is_acceptable (tcp_connection_t * tc0, vlib_buffer_t * tb0)
+{
+  /* SND.UNA =< SEG.ACK =< SND.NXT */
+  return (seq_leq (tc0->snd_una, vnet_buffer (tb0)->tcp.ack_number)
+	  && seq_leq (vnet_buffer (tb0)->tcp.ack_number, tc0->snd_nxt));
+}
+
+/**
+ * Compute smoothed RTT as per VJ's '88 SIGCOMM and RFC6298
+ *
+ * Note that although the original article, srtt and rttvar are scaled
+ * to minimize round-off errors, here we don't. Instead, we rely on
+ * better precision time measurements.
+ *
+ * TODO support us rtt resolution
+ */
+static void
+tcp_estimate_rtt (tcp_connection_t * tc, u32 mrtt)
+{
+  int err;
+
+  if (tc->srtt != 0)
+    {
+      err = mrtt - tc->srtt;
+      tc->srtt += err >> 3;
+
+      /* XXX Drop in RTT results in RTTVAR increase and bigger RTO.
+       * The increase should be bound */
+      tc->rttvar += (clib_abs (err) - tc->rttvar) >> 2;
+    }
+  else
+    {
+      /* First measurement. */
+      tc->srtt = mrtt;
+      tc->rttvar = mrtt << 1;
+    }
+}
+
+/** Update RTT estimate and RTO timer
+ *
+ * Measure RTT: We have two sources of RTT measurements: TSOPT and ACK
+ * timing. Middle boxes are known to fiddle with TCP options so we
+ * should give higher priority to ACK timing.
+ *
+ * return 1 if valid rtt 0 otherwise
+ */
+static int
+tcp_update_rtt (tcp_connection_t * tc, u32 ack)
+{
+  u32 mrtt = 0;
+
+  /* Karn's rule, part 1. Don't use retransmitted segments to estimate
+   * RTT because they're ambiguous. */
+  if (tc->rtt_seq && seq_gt (ack, tc->rtt_seq) && !tc->rto_boff)
+    {
+      mrtt = tcp_time_now () - tc->rtt_ts;
+      tc->rtt_seq = 0;
+    }
+
+  /* As per RFC7323 TSecr can be used for RTTM only if the segment advances
+   * snd_una, i.e., the left side of the send window:
+   * seq_lt (tc->snd_una, ack). Note: last condition could be dropped, we don't
+   * try to update rtt for dupacks */
+  else if (tcp_opts_tstamp (&tc->opt) && tc->opt.tsecr && tc->bytes_acked)
+    {
+      mrtt = tcp_time_now () - tc->opt.tsecr;
+    }
+
+  /* Ignore dubious measurements */
+  if (mrtt == 0 || mrtt > TCP_RTT_MAX)
+    return 0;
+
+  tcp_estimate_rtt (tc, mrtt);
+
+  tc->rto = clib_min (tc->srtt + (tc->rttvar << 2), TCP_RTO_MAX);
+
+  return 1;
+}
+
+/**
+ * Dequeue bytes that have been acked and while at it update RTT estimates.
+ */
+static void
+tcp_dequeue_acked (tcp_connection_t * tc, u32 ack)
+{
+  /* Dequeue the newly ACKed bytes */
+  stream_session_dequeue_drop (&tc->connection, tc->bytes_acked);
+
+  /* Update rtt and rto */
+  if (tcp_update_rtt (tc, ack))
+    {
+      /* Good ACK received and valid RTT, make sure retransmit backoff is 0 */
+      tc->rto_boff = 0;
+    }
+}
+
+/** Check if dupack as per RFC5681 Sec. 2 */
+always_inline u8
+tcp_ack_is_dupack (tcp_connection_t * tc, vlib_buffer_t * b, u32 new_snd_wnd)
+{
+  return ((vnet_buffer (b)->tcp.ack_number == tc->snd_una)
+	  && seq_gt (tc->snd_una_max, tc->snd_una)
+	  && (vnet_buffer (b)->tcp.seq_end == vnet_buffer (b)->tcp.seq_number)
+	  && (new_snd_wnd == tc->snd_wnd));
+}
+
+void
+scoreboard_remove_hole (sack_scoreboard_t * sb, sack_scoreboard_hole_t * hole)
+{
+  sack_scoreboard_hole_t *next, *prev;
+
+  if (hole->next != TCP_INVALID_SACK_HOLE_INDEX)
+    {
+      next = pool_elt_at_index (sb->holes, hole->next);
+      next->prev = hole->prev;
+    }
+
+  if (hole->prev != TCP_INVALID_SACK_HOLE_INDEX)
+    {
+      prev = pool_elt_at_index (sb->holes, hole->prev);
+      prev->next = hole->next;
+    }
+  else
+    {
+      sb->head = hole->next;
+    }
+
+  pool_put (sb->holes, hole);
+}
+
+sack_scoreboard_hole_t *
+scoreboard_insert_hole (sack_scoreboard_t * sb, sack_scoreboard_hole_t * prev,
+			u32 start, u32 end)
+{
+  sack_scoreboard_hole_t *hole, *next;
+  u32 hole_index;
+
+  pool_get (sb->holes, hole);
+  memset (hole, 0, sizeof (*hole));
+
+  hole->start = start;
+  hole->end = end;
+  hole_index = hole - sb->holes;
+
+  if (prev)
+    {
+      hole->prev = prev - sb->holes;
+      hole->next = prev->next;
+
+      if ((next = scoreboard_next_hole (sb, hole)))
+	next->prev = hole_index;
+
+      prev->next = hole_index;
+    }
+  else
+    {
+      sb->head = hole_index;
+      hole->prev = TCP_INVALID_SACK_HOLE_INDEX;
+      hole->next = TCP_INVALID_SACK_HOLE_INDEX;
+    }
+
+  return hole;
+}
+
+static void
+tcp_rcv_sacks (tcp_connection_t * tc, u32 ack)
+{
+  sack_scoreboard_t *sb = &tc->sack_sb;
+  sack_block_t *blk, tmp;
+  sack_scoreboard_hole_t *hole, *next_hole;
+  u32 blk_index = 0;
+  int i, j;
+
+  if (!tcp_opts_sack (tc) && sb->head == TCP_INVALID_SACK_HOLE_INDEX)
+    return;
+
+  /* Remove invalid blocks */
+  vec_foreach (blk, tc->opt.sacks)
+  {
+    if (seq_lt (blk->start, blk->end)
+	&& seq_gt (blk->start, tc->snd_una)
+	&& seq_gt (blk->start, ack) && seq_lt (blk->end, tc->snd_nxt))
+      continue;
+
+    vec_del1 (tc->opt.sacks, blk - tc->opt.sacks);
+  }
+
+  /* Add block for cumulative ack */
+  if (seq_gt (ack, tc->snd_una))
+    {
+      tmp.start = tc->snd_una;
+      tmp.end = ack;
+      vec_add1 (tc->opt.sacks, tmp);
+    }
+
+  if (vec_len (tc->opt.sacks) == 0)
+    return;
+
+  /* Make sure blocks are ordered */
+  for (i = 0; i < vec_len (tc->opt.sacks); i++)
+    for (j = i; j < vec_len (tc->opt.sacks); j++)
+      if (seq_lt (tc->opt.sacks[j].start, tc->opt.sacks[i].start))
+	{
+	  tmp = tc->opt.sacks[i];
+	  tc->opt.sacks[i] = tc->opt.sacks[j];
+	  tc->opt.sacks[j] = tmp;
+	}
+
+  /* If no holes, insert the first that covers all outstanding bytes */
+  if (sb->head == TCP_INVALID_SACK_HOLE_INDEX)
+    {
+      scoreboard_insert_hole (sb, 0, tc->snd_una, tc->snd_una_max);
+    }
+
+  /* Walk the holes with the SACK blocks */
+  hole = pool_elt_at_index (sb->holes, sb->head);
+  while (hole && blk_index < vec_len (tc->opt.sacks))
+    {
+      blk = &tc->opt.sacks[blk_index];
+
+      if (seq_leq (blk->start, hole->start))
+	{
+	  /* Block covers hole. Remove hole */
+	  if (seq_geq (blk->end, hole->end))
+	    {
+	      next_hole = scoreboard_next_hole (sb, hole);
+
+	      /* Byte accounting */
+	      if (seq_lt (hole->end, ack))
+		{
+		  /* Bytes lost because snd wnd left edge advances */
+		  if (seq_lt (next_hole->start, ack))
+		    sb->sacked_bytes -= next_hole->start - hole->end;
+		  else
+		    sb->sacked_bytes -= ack - hole->end;
+		}
+	      else
+		{
+		  sb->sacked_bytes += scoreboard_hole_bytes (hole);
+		}
+
+	      scoreboard_remove_hole (sb, hole);
+	      hole = next_hole;
+	    }
+	  /* Partial overlap */
+	  else
+	    {
+	      sb->sacked_bytes += blk->end - hole->start;
+	      hole->start = blk->end;
+	      blk_index++;
+	    }
+	}
+      else
+	{
+	  /* Hole must be split */
+	  if (seq_leq (blk->end, hole->end))
+	    {
+	      sb->sacked_bytes += blk->end - blk->start;
+	      scoreboard_insert_hole (sb, hole, blk->end, hole->end);
+	      hole->end = blk->start - 1;
+	      blk_index++;
+	    }
+	  else
+	    {
+	      sb->sacked_bytes += hole->end - blk->start + 1;
+	      hole->end = blk->start - 1;
+	      hole = scoreboard_next_hole (sb, hole);
+	    }
+	}
+    }
+}
+
+/** Update snd_wnd
+ *
+ * If (SND.WL1 < SEG.SEQ or (SND.WL1 = SEG.SEQ and SND.WL2 =< SEG.ACK)), set
+ * SND.WND <- SEG.WND, set SND.WL1 <- SEG.SEQ, and set SND.WL2 <- SEG.ACK */
+static void
+tcp_update_snd_wnd (tcp_connection_t * tc, u32 seq, u32 ack, u32 snd_wnd)
+{
+  if (tc->snd_wl1 < seq || (tc->snd_wl1 == seq && tc->snd_wl2 <= ack))
+    {
+      tc->snd_wnd = snd_wnd;
+      tc->snd_wl1 = seq;
+      tc->snd_wl2 = ack;
+    }
+}
+
+static void
+tcp_cc_congestion (tcp_connection_t * tc)
+{
+  tc->cc_algo->congestion (tc);
+}
+
+static void
+tcp_cc_recover (tcp_connection_t * tc)
+{
+  if (tcp_in_fastrecovery (tc))
+    {
+      tc->cc_algo->recovered (tc);
+      tcp_recovery_off (tc);
+    }
+  else if (tcp_in_recovery (tc))
+    {
+      tcp_recovery_off (tc);
+      tc->cwnd = tcp_loss_wnd (tc);
+    }
+}
+
+static void
+tcp_cc_rcv_ack (tcp_connection_t * tc)
+{
+  u8 partial_ack;
+
+  if (tcp_in_recovery (tc))
+    {
+      partial_ack = seq_lt (tc->snd_una, tc->snd_una_max);
+      if (!partial_ack)
+	{
+	  /* Clear retransmitted bytes. */
+	  tc->rtx_bytes = 0;
+	  tcp_cc_recover (tc);
+	}
+      else
+	{
+	  /* Clear retransmitted bytes. XXX should we clear all? */
+	  tc->rtx_bytes = 0;
+	  tc->cc_algo->rcv_cong_ack (tc, TCP_CC_PARTIALACK);
+
+	  /* Retransmit first unacked segment */
+	  tcp_retransmit_first_unacked (tc);
+	}
+    }
+  else
+    {
+      tc->cc_algo->rcv_ack (tc);
+    }
+
+  tc->rcv_dupacks = 0;
+  tc->tsecr_last_ack = tc->opt.tsecr;
+}
+
+static void
+tcp_cc_rcv_dupack (tcp_connection_t * tc, u32 ack)
+{
+  ASSERT (tc->snd_una == ack);
+
+  tc->rcv_dupacks++;
+  if (tc->rcv_dupacks == TCP_DUPACK_THRESHOLD)
+    {
+      /* RFC6582 NewReno heuristic to avoid multiple fast retransmits */
+      if (tc->opt.tsecr != tc->tsecr_last_ack)
+	{
+	  tc->rcv_dupacks = 0;
+	  return;
+	}
+
+      tcp_fastrecovery_on (tc);
+
+      /* Handle congestion and dupack */
+      tcp_cc_congestion (tc);
+      tc->cc_algo->rcv_cong_ack (tc, TCP_CC_DUPACK);
+
+      tcp_fast_retransmit (tc);
+
+      /* Post retransmit update cwnd to ssthresh and account for the
+       * three segments that have left the network and should've been
+       * buffered at the receiver */
+      tc->cwnd = tc->ssthresh + TCP_DUPACK_THRESHOLD * tc->snd_mss;
+    }
+  else if (tc->rcv_dupacks > TCP_DUPACK_THRESHOLD)
+    {
+      ASSERT (tcp_in_fastrecovery (tc));
+
+      tc->cc_algo->rcv_cong_ack (tc, TCP_CC_DUPACK);
+    }
+}
+
+void
+tcp_cc_init (tcp_connection_t * tc)
+{
+  tc->cc_algo = tcp_cc_algo_get (TCP_CC_NEWRENO);
+  tc->cc_algo->init (tc);
+}
+
+static int
+tcp_rcv_ack (tcp_connection_t * tc, vlib_buffer_t * b,
+	     tcp_header_t * th, u32 * next, u32 * error)
+{
+  u32 new_snd_wnd;
+
+  /* If the ACK acks something not yet sent (SEG.ACK > SND.NXT) then send an
+   * ACK, drop the segment, and return  */
+  if (seq_gt (vnet_buffer (b)->tcp.ack_number, tc->snd_nxt))
+    {
+      tcp_make_ack (tc, b);
+      *next = tcp_next_output (tc->c_is_ip4);
+      *error = TCP_ERROR_ACK_INVALID;
+      return -1;
+    }
+
+  /* If old ACK, discard */
+  if (seq_lt (vnet_buffer (b)->tcp.ack_number, tc->snd_una))
+    {
+      *error = TCP_ERROR_ACK_OLD;
+      return -1;
+    }
+
+  if (tcp_opts_sack_permitted (&tc->opt))
+    tcp_rcv_sacks (tc, vnet_buffer (b)->tcp.ack_number);
+
+  new_snd_wnd = clib_net_to_host_u32 (th->window) << tc->snd_wscale;
+
+  if (tcp_ack_is_dupack (tc, b, new_snd_wnd))
+    {
+      tcp_cc_rcv_dupack (tc, vnet_buffer (b)->tcp.ack_number);
+      *error = TCP_ERROR_ACK_DUP;
+      return -1;
+    }
+
+  /* Valid ACK */
+  tc->bytes_acked = vnet_buffer (b)->tcp.ack_number - tc->snd_una;
+  tc->snd_una = vnet_buffer (b)->tcp.ack_number;
+
+  /* Dequeue ACKed packet and update RTT */
+  tcp_dequeue_acked (tc, vnet_buffer (b)->tcp.ack_number);
+
+  tcp_update_snd_wnd (tc, vnet_buffer (b)->tcp.seq_number,
+		      vnet_buffer (b)->tcp.ack_number, new_snd_wnd);
+
+  /* Updates congestion control (slow start/congestion avoidance) */
+  tcp_cc_rcv_ack (tc);
+
+  /* If everything has been acked, stop retransmit timer
+   * otherwise update */
+  if (tc->snd_una == tc->snd_una_max)
+    tcp_timer_reset (tc, TCP_TIMER_RETRANSMIT);
+  else
+    tcp_timer_update (tc, TCP_TIMER_RETRANSMIT, tc->rto);
+
+  return 0;
+}
+
+/**
+ * Build SACK list as per RFC2018.
+ *
+ * Makes sure the first block contains the segment that generated the current
+ * ACK and the following ones are the ones most recently reported in SACK
+ * blocks.
+ *
+ * @param tc TCP connection for which the SACK list is updated
+ * @param start Start sequence number of the newest SACK block
+ * @param end End sequence of the newest SACK block
+ */
+static void
+tcp_update_sack_list (tcp_connection_t * tc, u32 start, u32 end)
+{
+  sack_block_t *new_list = 0, block;
+  u32 n_elts;
+  int i;
+  u8 new_head = 0;
+
+  /* If the first segment is ooo add it to the list. Last write might've moved
+   * rcv_nxt over the first segment. */
+  if (seq_lt (tc->rcv_nxt, start))
+    {
+      block.start = start;
+      block.end = end;
+      vec_add1 (new_list, block);
+      new_head = 1;
+    }
+
+  /* Find the blocks still worth keeping. */
+  for (i = 0; i < vec_len (tc->snd_sacks); i++)
+    {
+      /* Discard if:
+       * 1) rcv_nxt advanced beyond current block OR
+       * 2) Segment overlapped by the first segment, i.e., it has been merged
+       *    into it.*/
+      if (seq_leq (tc->snd_sacks[i].start, tc->rcv_nxt)
+	  || seq_leq (tc->snd_sacks[i].start, end))
+	continue;
+
+      /* Save subsequent segments to new SACK list. */
+      n_elts = clib_min (vec_len (tc->snd_sacks) - i,
+			 TCP_MAX_SACK_BLOCKS - new_head);
+      vec_insert_elts (new_list, &tc->snd_sacks[i], n_elts, new_head);
+      break;
+    }
+
+  /* Replace old vector with new one */
+  vec_free (tc->snd_sacks);
+  tc->snd_sacks = new_list;
+}
+
+/** Enqueue data for delivery to application */
+always_inline u32
+tcp_session_enqueue_data (tcp_connection_t * tc, vlib_buffer_t * b,
+			  u16 data_len)
+{
+  int written;
+
+  /* Pure ACK. Update rcv_nxt and be done. */
+  if (PREDICT_FALSE (data_len == 0))
+    {
+      tc->rcv_nxt = vnet_buffer (b)->tcp.seq_end;
+      return TCP_ERROR_PURE_ACK;
+    }
+
+  written = stream_session_enqueue_data (&tc->connection,
+					 vlib_buffer_get_current (b),
+					 data_len, 1 /* queue event */ );
+
+  /* Update rcv_nxt */
+  if (PREDICT_TRUE (written == data_len))
+    {
+      tc->rcv_nxt = vnet_buffer (b)->tcp.seq_end;
+    }
+  /* If more data written than expected, account for out-of-order bytes. */
+  else if (written > data_len)
+    {
+      tc->rcv_nxt = vnet_buffer (b)->tcp.seq_end + written - data_len;
+
+      /* Send ACK confirming the update */
+      tc->flags |= TCP_CONN_SNDACK;
+
+      /* Update SACK list if need be */
+      if (tcp_opts_sack_permitted (&tc->opt))
+	{
+	  /* Remove SACK blocks that have been delivered */
+	  tcp_update_sack_list (tc, tc->rcv_nxt, tc->rcv_nxt);
+	}
+    }
+  else
+    {
+      ASSERT (0);
+      return TCP_ERROR_FIFO_FULL;
+    }
+
+  return TCP_ERROR_ENQUEUED;
+}
+
+/** Enqueue out-of-order data */
+always_inline u32
+tcp_session_enqueue_ooo (tcp_connection_t * tc, vlib_buffer_t * b,
+			 u16 data_len)
+{
+  stream_session_t *s0;
+  u32 offset, seq;
+
+  s0 = stream_session_get (tc->c_s_index, tc->c_thread_index);
+  seq = vnet_buffer (b)->tcp.seq_number;
+  offset = seq - tc->rcv_nxt;
+
+  if (svm_fifo_enqueue_with_offset (s0->server_rx_fifo, s0->pid, offset,
+				    data_len, vlib_buffer_get_current (b)))
+    return TCP_ERROR_FIFO_FULL;
+
+  /* Update SACK list if in use */
+  if (tcp_opts_sack_permitted (&tc->opt))
+    {
+      ooo_segment_t *newest;
+      u32 start, end;
+
+      /* Get the newest segment from the fifo */
+      newest = svm_fifo_newest_ooo_segment (s0->server_rx_fifo);
+      start = tc->rcv_nxt + ooo_segment_offset (s0->server_rx_fifo, newest);
+      end = tc->rcv_nxt + ooo_segment_end_offset (s0->server_rx_fifo, newest);
+
+      tcp_update_sack_list (tc, start, end);
+    }
+
+  return TCP_ERROR_ENQUEUED;
+}
+
+/**
+ * Check if ACK could be delayed. DELACK timer is set only after frame is
+ * processed so this can return true for a full bursts of packets.
+ */
+always_inline int
+tcp_can_delack (tcp_connection_t * tc)
+{
+  /* If there's no DELACK timer set and the last window sent wasn't 0 we
+   * can safely delay. */
+  if (!tcp_timer_is_active (tc, TCP_TIMER_DELACK)
+      && (tc->flags & TCP_CONN_SENT_RCV_WND0) == 0
+      && (tc->flags & TCP_CONN_SNDACK) == 0)
+    return 1;
+
+  return 0;
+}
+
+static int
+tcp_segment_rcv (tcp_main_t * tm, tcp_connection_t * tc, vlib_buffer_t * b,
+		 u16 n_data_bytes, u32 * next0)
+{
+  u32 error = 0;
+
+  /* Handle out-of-order data */
+  if (PREDICT_FALSE (vnet_buffer (b)->tcp.seq_number != tc->rcv_nxt))
+    {
+      error = tcp_session_enqueue_ooo (tc, b, n_data_bytes);
+
+      /* Don't send more than 3 dupacks per burst
+       * XXX decide if this is good */
+      if (tc->snt_dupacks < 3)
+	{
+	  /* RFC2581: Send DUPACK for fast retransmit */
+	  tcp_make_ack (tc, b);
+	  *next0 = tcp_next_output (tc->c_is_ip4);
+
+	  /* Mark as DUPACK. We may filter these in output if
+	   * the burst fills the holes. */
+	  vnet_buffer (b)->tcp.flags = TCP_BUF_FLAG_DUPACK;
+
+	  tc->snt_dupacks++;
+	}
+
+      goto done;
+    }
+
+  /* In order data, enqueue. Fifo figures out by itself if any out-of-order
+   * segments can be enqueued after fifo tail offset changes. */
+  error = tcp_session_enqueue_data (tc, b, n_data_bytes);
+
+  /* Check if ACK can be delayed */
+  if (tcp_can_delack (tc))
+    {
+      /* Nothing to do for pure ACKs */
+      if (n_data_bytes == 0)
+	goto done;
+
+      /* If connection has not been previously marked for delay ack
+       * add it to the list and flag it */
+      if (!tc->flags & TCP_CONN_DELACK)
+	{
+	  vec_add1 (tm->delack_connections[tc->c_thread_index],
+		    tc->c_c_index);
+	  tc->flags |= TCP_CONN_DELACK;
+	}
+    }
+  else
+    {
+      /* Check if a packet has already been enqueued to output for burst.
+       * If yes, then drop this one, otherwise, let it pass through to
+       * output */
+      if ((tc->flags & TCP_CONN_BURSTACK) == 0)
+	{
+	  *next0 = tcp_next_output (tc->c_is_ip4);
+	  tcp_make_ack (tc, b);
+	  error = TCP_ERROR_ENQUEUED;
+
+	  /* TODO: maybe add counter to ensure N acks will be sent/burst */
+	  tc->flags |= TCP_CONN_BURSTACK;
+	}
+    }
+
+done:
+  return error;
+}
+
+void
+delack_timers_init (tcp_main_t * tm, u32 thread_index)
+{
+  tcp_connection_t *tc;
+  u32 i, *conns;
+  tw_timer_wheel_16t_2w_512sl_t *tw;
+
+  tw = &tm->timer_wheels[thread_index];
+  conns = tm->delack_connections[thread_index];
+  for (i = 0; i < vec_len (conns); i++)
+    {
+      tc = pool_elt_at_index (tm->connections[thread_index], conns[i]);
+      ASSERT (0 != tc);
+
+      tc->timers[TCP_TIMER_DELACK]
+	= tw_timer_start_16t_2w_512sl (tw, conns[i],
+				       TCP_TIMER_DELACK, TCP_DELACK_TIME);
+    }
+  vec_reset_length (tm->delack_connections[thread_index]);
+}
+
+always_inline uword
+tcp46_established_inline (vlib_main_t * vm, vlib_node_runtime_t * node,
+			  vlib_frame_t * from_frame, int is_ip4)
+{
+  u32 n_left_from, next_index, *from, *to_next;
+  u32 my_thread_index = vm->cpu_index, errors = 0;
+  tcp_main_t *tm = vnet_get_tcp_main ();
+
+  from = vlib_frame_vector_args (from_frame);
+  n_left_from = from_frame->n_vectors;
+
+  next_index = node->cached_next_index;
+
+  while (n_left_from > 0)
+    {
+      u32 n_left_to_next;
+
+      vlib_get_next_frame (vm, node, next_index, to_next, n_left_to_next);
+
+      while (n_left_from > 0 && n_left_to_next > 0)
+	{
+	  u32 bi0;
+	  vlib_buffer_t *b0;
+	  tcp_header_t *th0 = 0;
+	  tcp_connection_t *tc0;
+	  ip4_header_t *ip40;
+	  ip6_header_t *ip60;
+	  u32 n_advance_bytes0, n_data_bytes0;
+	  u32 next0 = TCP_ESTABLISHED_NEXT_DROP, error0 = TCP_ERROR_ENQUEUED;
+
+	  bi0 = from[0];
+	  to_next[0] = bi0;
+	  from += 1;
+	  to_next += 1;
+	  n_left_from -= 1;
+	  n_left_to_next -= 1;
+
+	  b0 = vlib_get_buffer (vm, bi0);
+	  tc0 = tcp_connection_get (vnet_buffer (b0)->tcp.connection_index,
+				    my_thread_index);
+
+	  /* Checksum computed by ipx_local no need to compute again */
+
+	  if (is_ip4)
+	    {
+	      ip40 = vlib_buffer_get_current (b0);
+	      th0 = ip4_next_header (ip40);
+	      n_advance_bytes0 = (ip4_header_bytes (ip40)
+				  + tcp_header_bytes (th0));
+	      n_data_bytes0 = clib_net_to_host_u16 (ip40->length)
+		- n_advance_bytes0;
+	    }
+	  else
+	    {
+	      ip60 = vlib_buffer_get_current (b0);
+	      th0 = ip6_next_header (ip60);
+	      n_advance_bytes0 = tcp_header_bytes (th0);
+	      n_data_bytes0 = clib_net_to_host_u16 (ip60->payload_length)
+		- n_advance_bytes0;
+	      n_advance_bytes0 += sizeof (ip60[0]);
+	    }
+
+	  /* SYNs, FINs and data consume sequence numbers */
+	  vnet_buffer (b0)->tcp.seq_end = vnet_buffer (b0)->tcp.seq_number
+	    + tcp_is_syn (th0) + tcp_is_fin (th0) + n_data_bytes0;
+
+	  /* TODO header prediction fast path */
+
+	  /* 1-4: check SEQ, RST, SYN */
+	  if (PREDICT_FALSE (tcp_segment_validate (vm, tc0, b0, th0, &next0)))
+	    {
+	      error0 = TCP_ERROR_SEGMENT_INVALID;
+	      goto drop;
+	    }
+
+	  /* 5: check the ACK field  */
+	  if (tcp_rcv_ack (tc0, b0, th0, &next0, &error0))
+	    {
+	      goto drop;
+	    }
+
+	  /* 6: check the URG bit TODO */
+
+	  /* 7: process the segment text */
+	  vlib_buffer_advance (b0, n_advance_bytes0);
+	  error0 = tcp_segment_rcv (tm, tc0, b0, n_data_bytes0, &next0);
+
+	  /* 8: check the FIN bit */
+	  if (tcp_fin (th0))
+	    {
+	      /* Send ACK and enter CLOSE-WAIT */
+	      tcp_make_ack (tc0, b0);
+	      tcp_connection_force_ack (tc0, b0);
+	      next0 = tcp_next_output (tc0->c_is_ip4);
+	      tc0->state = TCP_STATE_CLOSE_WAIT;
+	      stream_session_disconnect_notify (&tc0->connection);
+	    }
+
+	drop:
+	  b0->error = node->errors[error0];
+	  if (PREDICT_FALSE (b0->flags & VLIB_BUFFER_IS_TRACED))
+	    {
+
+	    }
+
+	  vlib_validate_buffer_enqueue_x1 (vm, node, next_index, to_next,
+					   n_left_to_next, bi0, next0);
+	}
+
+      vlib_put_next_frame (vm, node, next_index, n_left_to_next);
+    }
+
+  errors = session_manager_flush_enqueue_events (my_thread_index);
+  if (errors)
+    {
+      if (is_ip4)
+	vlib_node_increment_counter (vm, tcp4_established_node.index,
+				     TCP_ERROR_EVENT_FIFO_FULL, errors);
+      else
+	vlib_node_increment_counter (vm, tcp6_established_node.index,
+				     TCP_ERROR_EVENT_FIFO_FULL, errors);
+    }
+
+  delack_timers_init (tm, my_thread_index);
+
+  return from_frame->n_vectors;
+}
+
+static uword
+tcp4_established (vlib_main_t * vm, vlib_node_runtime_t * node,
+		  vlib_frame_t * from_frame)
+{
+  return tcp46_established_inline (vm, node, from_frame, 1 /* is_ip4 */ );
+}
+
+static uword
+tcp6_established (vlib_main_t * vm, vlib_node_runtime_t * node,
+		  vlib_frame_t * from_frame)
+{
+  return tcp46_established_inline (vm, node, from_frame, 0 /* is_ip4 */ );
+}
+
+/* *INDENT-OFF* */
+VLIB_REGISTER_NODE (tcp4_established_node) =
+{
+  .function = tcp4_established,
+  .name = "tcp4-established",
+  /* Takes a vector of packets. */
+  .vector_size = sizeof (u32),
+  .n_errors = TCP_N_ERROR,.error_strings = tcp_error_strings,
+  .n_next_nodes = TCP_ESTABLISHED_N_NEXT,
+  .next_nodes =
+  {
+#define _(s,n) [TCP_ESTABLISHED_NEXT_##s] = n,
+    foreach_tcp_state_next
+#undef _
+  },
+};
+/* *INDENT-ON* */
+
+VLIB_NODE_FUNCTION_MULTIARCH (tcp4_established_node, tcp4_established);
+
+/* *INDENT-OFF* */
+VLIB_REGISTER_NODE (tcp6_established_node) =
+{
+  .function = tcp6_established,
+  .name = "tcp6-established",
+  /* Takes a vector of packets. */
+  .vector_size = sizeof (u32),
+  .n_errors = TCP_N_ERROR,
+  .error_strings = tcp_error_strings,
+  .n_next_nodes = TCP_ESTABLISHED_N_NEXT,
+  .next_nodes =
+  {
+#define _(s,n) [TCP_ESTABLISHED_NEXT_##s] = n,
+    foreach_tcp_state_next
+#undef _
+  },
+};
+/* *INDENT-ON* */
+
+
+VLIB_NODE_FUNCTION_MULTIARCH (tcp6_established_node, tcp6_established);
+
+vlib_node_registration_t tcp4_syn_sent_node;
+vlib_node_registration_t tcp6_syn_sent_node;
+
+always_inline uword
+tcp46_syn_sent_inline (vlib_main_t * vm, vlib_node_runtime_t * node,
+		       vlib_frame_t * from_frame, int is_ip4)
+{
+  tcp_main_t *tm = vnet_get_tcp_main ();
+  u32 n_left_from, next_index, *from, *to_next;
+  u32 my_thread_index = vm->cpu_index, errors = 0;
+  u8 sst = is_ip4 ? SESSION_TYPE_IP4_TCP : SESSION_TYPE_IP6_TCP;
+
+  from = vlib_frame_vector_args (from_frame);
+  n_left_from = from_frame->n_vectors;
+
+  next_index = node->cached_next_index;
+
+  while (n_left_from > 0)
+    {
+      u32 n_left_to_next;
+
+      vlib_get_next_frame (vm, node, next_index, to_next, n_left_to_next);
+
+      while (n_left_from > 0 && n_left_to_next > 0)
+	{
+	  u32 bi0, ack0, seq0;
+	  vlib_buffer_t *b0;
+	  tcp_header_t *tcp0 = 0;
+	  tcp_connection_t *tc0;
+	  ip4_header_t *ip40;
+	  ip6_header_t *ip60;
+	  u32 n_advance_bytes0, n_data_bytes0;
+	  tcp_connection_t *new_tc0;
+	  u32 next0 = TCP_SYN_SENT_NEXT_DROP, error0 = TCP_ERROR_ENQUEUED;
+
+	  bi0 = from[0];
+	  to_next[0] = bi0;
+	  from += 1;
+	  to_next += 1;
+	  n_left_from -= 1;
+	  n_left_to_next -= 1;
+
+	  b0 = vlib_get_buffer (vm, bi0);
+	  tc0 =
+	    tcp_half_open_connection_get (vnet_buffer (b0)->
+					  tcp.connection_index);
+
+	  ack0 = vnet_buffer (b0)->tcp.ack_number;
+	  seq0 = vnet_buffer (b0)->tcp.seq_number;
+
+	  /* Checksum computed by ipx_local no need to compute again */
+
+	  if (is_ip4)
+	    {
+	      ip40 = vlib_buffer_get_current (b0);
+	      tcp0 = ip4_next_header (ip40);
+	      n_advance_bytes0 = (ip4_header_bytes (ip40)
+				  + tcp_header_bytes (tcp0));
+	      n_data_bytes0 = clib_net_to_host_u16 (ip40->length)
+		- n_advance_bytes0;
+	    }
+	  else
+	    {
+	      ip60 = vlib_buffer_get_current (b0);
+	      tcp0 = ip6_next_header (ip60);
+	      n_advance_bytes0 = tcp_header_bytes (tcp0);
+	      n_data_bytes0 = clib_net_to_host_u16 (ip60->payload_length)
+		- n_advance_bytes0;
+	      n_advance_bytes0 += sizeof (ip60[0]);
+	    }
+
+	  if (PREDICT_FALSE
+	      (!tcp_ack (tcp0) && !tcp_rst (tcp0) && !tcp_syn (tcp0)))
+	    goto drop;
+
+	  /* SYNs, FINs and data consume sequence numbers */
+	  vnet_buffer (b0)->tcp.seq_end = seq0 + tcp_is_syn (tcp0)
+	    + tcp_is_fin (tcp0) + n_data_bytes0;
+
+	  /*
+	   *  1. check the ACK bit
+	   */
+
+	  /*
+	   *   If the ACK bit is set
+	   *     If SEG.ACK =< ISS, or SEG.ACK > SND.NXT, send a reset (unless
+	   *     the RST bit is set, if so drop the segment and return)
+	   *       
+	   *     and discard the segment.  Return.
+	   *     If SND.UNA =< SEG.ACK =< SND.NXT then the ACK is acceptable.
+	   */
+	  if (tcp_ack (tcp0))
+	    {
+	      if (ack0 <= tc0->iss || ack0 > tc0->snd_nxt)
+		{
+		  if (!tcp_rst (tcp0))
+		    tcp_send_reset (b0, is_ip4);
+
+		  goto drop;
+		}
+
+	      /* Make sure ACK is valid */
+	      if (tc0->snd_una > ack0)
+		goto drop;
+	    }
+
+	  /*
+	   * 2. check the RST bit
+	   */
+
+	  if (tcp_rst (tcp0))
+	    {
+	      /* If ACK is acceptable, signal client that peer is not
+	       * willing to accept connection and drop connection*/
+	      if (tcp_ack (tcp0))
+		{
+		  stream_session_connect_notify (&tc0->connection, sst,
+						 1 /* fail */ );
+		  tcp_connection_cleanup (tc0);
+		}
+	      goto drop;
+	    }
+
+	  /*
+	   * 3. check the security and precedence (skipped)
+	   */
+
+	  /*
+	   * 4. check the SYN bit
+	   */
+
+	  /* No SYN flag. Drop. */
+	  if (!tcp_syn (tcp0))
+	    goto drop;
+
+	  /* Stop connection establishment and retransmit timers */
+	  tcp_timer_reset (tc0, TCP_TIMER_ESTABLISH);
+	  tcp_timer_reset (tc0, TCP_TIMER_RETRANSMIT_SYN);
+
+	  /* Valid SYN or SYN-ACK. Move connection from half-open pool to
+	   * current thread pool. */
+	  pool_get (tm->connections[my_thread_index], new_tc0);
+	  clib_memcpy (new_tc0, tc0, sizeof (*new_tc0));
+
+	  new_tc0->c_thread_index = my_thread_index;
+
+	  /* Cleanup half-open connection XXX lock */
+	  pool_put (tm->half_open_connections, tc0);
+
+	  new_tc0->rcv_nxt = vnet_buffer (b0)->tcp.seq_end;
+	  new_tc0->irs = seq0;
+
+	  /* Parse options */
+	  tcp_options_parse (tcp0, &new_tc0->opt);
+	  tcp_connection_init_vars (new_tc0);
+
+	  if (tcp_opts_tstamp (&new_tc0->opt))
+	    {
+	      new_tc0->tsval_recent = new_tc0->opt.tsval;
+	      new_tc0->tsval_recent_age = tcp_time_now ();
+	    }
+
+	  if (tcp_opts_wscale (&new_tc0->opt))
+	    new_tc0->snd_wscale = new_tc0->opt.wscale;
+
+	  new_tc0->snd_wnd = clib_net_to_host_u32 (tcp0->window)
+	    << new_tc0->snd_wscale;
+	  new_tc0->snd_wl1 = seq0;
+	  new_tc0->snd_wl2 = ack0;
+
+	  /* SYN-ACK: See if we can switch to ESTABLISHED state */
+	  if (tcp_ack (tcp0))
+	    {
+	      /* Our SYN is ACKed: we have iss < ack = snd_una */
+
+	      /* TODO Dequeue acknowledged segments if we support Fast Open */
+	      new_tc0->snd_una = ack0;
+	      new_tc0->state = TCP_STATE_ESTABLISHED;
+
+	      /* Notify app that we have connection */
+	      stream_session_connect_notify (&new_tc0->connection, sst, 0);
+
+	      /* Make sure after data segment processing ACK is sent */
+	      new_tc0->flags |= TCP_CONN_SNDACK;
+	    }
+	  /* SYN: Simultaneous open. Change state to SYN-RCVD and send SYN-ACK */
+	  else
+	    {
+	      new_tc0->state = TCP_STATE_SYN_RCVD;
+
+	      /* Notify app that we have connection XXX */
+	      stream_session_connect_notify (&new_tc0->connection, sst, 0);
+
+	      tcp_make_synack (new_tc0, b0);
+	      next0 = tcp_next_output (is_ip4);
+
+	      goto drop;
+	    }
+
+	  /* Read data, if any */
+	  if (n_data_bytes0)
+	    {
+	      error0 =
+		tcp_segment_rcv (tm, new_tc0, b0, n_data_bytes0, &next0);
+	      if (error0 == TCP_ERROR_PURE_ACK)
+		error0 = TCP_ERROR_SYN_ACKS_RCVD;
+	    }
+	  else
+	    {
+	      tcp_make_ack (new_tc0, b0);
+	      next0 = tcp_next_output (new_tc0->c_is_ip4);
+	    }
+
+	drop:
+
+	  b0->error = error0 ? node->errors[error0] : 0;
+	  if (PREDICT_FALSE (b0->flags & VLIB_BUFFER_IS_TRACED))
+	    {
+
+	    }
+
+	  vlib_validate_buffer_enqueue_x1 (vm, node, next_index, to_next,
+					   n_left_to_next, bi0, next0);
+	}
+
+      vlib_put_next_frame (vm, node, next_index, n_left_to_next);
+    }
+
+  errors = session_manager_flush_enqueue_events (my_thread_index);
+  if (errors)
+    {
+      if (is_ip4)
+	vlib_node_increment_counter (vm, tcp4_established_node.index,
+				     TCP_ERROR_EVENT_FIFO_FULL, errors);
+      else
+	vlib_node_increment_counter (vm, tcp6_established_node.index,
+				     TCP_ERROR_EVENT_FIFO_FULL, errors);
+    }
+
+  return from_frame->n_vectors;
+}
+
+static uword
+tcp4_syn_sent (vlib_main_t * vm, vlib_node_runtime_t * node,
+	       vlib_frame_t * from_frame)
+{
+  return tcp46_syn_sent_inline (vm, node, from_frame, 1 /* is_ip4 */ );
+}
+
+static uword
+tcp6_syn_sent_rcv (vlib_main_t * vm, vlib_node_runtime_t * node,
+		   vlib_frame_t * from_frame)
+{
+  return tcp46_syn_sent_inline (vm, node, from_frame, 0 /* is_ip4 */ );
+}
+
+/* *INDENT-OFF* */
+VLIB_REGISTER_NODE (tcp4_syn_sent_node) =
+{
+  .function = tcp4_syn_sent,
+  .name = "tcp4-syn-sent",
+  /* Takes a vector of packets. */
+  .vector_size = sizeof (u32),
+  .n_errors = TCP_N_ERROR,
+  .error_strings = tcp_error_strings,
+  .n_next_nodes = TCP_SYN_SENT_N_NEXT,
+  .next_nodes =
+  {
+#define _(s,n) [TCP_SYN_SENT_NEXT_##s] = n,
+    foreach_tcp_state_next
+#undef _
+  },
+};
+/* *INDENT-ON* */
+
+VLIB_NODE_FUNCTION_MULTIARCH (tcp4_syn_sent_node, tcp4_syn_sent);
+
+/* *INDENT-OFF* */
+VLIB_REGISTER_NODE (tcp6_syn_sent_node) =
+{
+  .function = tcp6_syn_sent_rcv,
+  .name = "tcp6-syn-sent",
+  /* Takes a vector of packets. */
+  .vector_size = sizeof (u32),
+  .n_errors = TCP_N_ERROR,
+  .error_strings = tcp_error_strings,
+  .n_next_nodes = TCP_SYN_SENT_N_NEXT,
+  .next_nodes =
+  {
+#define _(s,n) [TCP_SYN_SENT_NEXT_##s] = n,
+    foreach_tcp_state_next
+#undef _
+  }
+,};
+/* *INDENT-ON* */
+
+VLIB_NODE_FUNCTION_MULTIARCH (tcp6_syn_sent_node, tcp6_syn_sent_rcv);
+/**
+ * Handles reception for all states except LISTEN, SYN-SEND and ESTABLISHED
+ * as per RFC793 p. 64
+ */
+always_inline uword
+tcp46_rcv_process_inline (vlib_main_t * vm, vlib_node_runtime_t * node,
+			  vlib_frame_t * from_frame, int is_ip4)
+{
+  tcp_main_t *tm = vnet_get_tcp_main ();
+  u32 n_left_from, next_index, *from, *to_next;
+  u32 my_thread_index = vm->cpu_index, errors = 0;
+
+  from = vlib_frame_vector_args (from_frame);
+  n_left_from = from_frame->n_vectors;
+
+  next_index = node->cached_next_index;
+
+  while (n_left_from > 0)
+    {
+      u32 n_left_to_next;
+
+      vlib_get_next_frame (vm, node, next_index, to_next, n_left_to_next);
+
+      while (n_left_from > 0 && n_left_to_next > 0)
+	{
+	  u32 bi0;
+	  vlib_buffer_t *b0;
+	  tcp_header_t *tcp0 = 0;
+	  tcp_connection_t *tc0;
+	  ip4_header_t *ip40;
+	  ip6_header_t *ip60;
+	  u32 n_advance_bytes0, n_data_bytes0;
+	  u32 next0 = TCP_RCV_PROCESS_NEXT_DROP, error0 = TCP_ERROR_ENQUEUED;
+
+	  bi0 = from[0];
+	  to_next[0] = bi0;
+	  from += 1;
+	  to_next += 1;
+	  n_left_from -= 1;
+	  n_left_to_next -= 1;
+
+	  b0 = vlib_get_buffer (vm, bi0);
+	  tc0 = tcp_connection_get (vnet_buffer (b0)->tcp.connection_index,
+				    my_thread_index);
+
+	  /* Checksum computed by ipx_local no need to compute again */
+
+	  if (is_ip4)
+	    {
+	      ip40 = vlib_buffer_get_current (b0);
+	      tcp0 = ip4_next_header (ip40);
+	      n_advance_bytes0 = (ip4_header_bytes (ip40)
+				  + tcp_header_bytes (tcp0));
+	      n_data_bytes0 = clib_net_to_host_u16 (ip40->length)
+		- n_advance_bytes0;
+	    }
+	  else
+	    {
+	      ip60 = vlib_buffer_get_current (b0);
+	      tcp0 = ip6_next_header (ip60);
+	      n_advance_bytes0 = tcp_header_bytes (tcp0);
+	      n_data_bytes0 = clib_net_to_host_u16 (ip60->payload_length)
+		- n_advance_bytes0;
+	      n_advance_bytes0 += sizeof (ip60[0]);
+	    }
+
+	  /* SYNs, FINs and data consume sequence numbers */
+	  vnet_buffer (b0)->tcp.seq_end = vnet_buffer (b0)->tcp.seq_number
+	    + tcp_is_syn (tcp0) + tcp_is_fin (tcp0) + n_data_bytes0;
+
+	  /*
+	   * Special treatment for CLOSED
+	   */
+	  switch (tc0->state)
+	    {
+	    case TCP_STATE_CLOSED:
+	      goto drop;
+	      break;
+	    }
+
+	  /*
+	   * For all other states (except LISTEN)
+	   */
+
+	  /* 1-4: check SEQ, RST, SYN */
+	  if (PREDICT_FALSE
+	      (tcp_segment_validate (vm, tc0, b0, tcp0, &next0)))
+	    {
+	      error0 = TCP_ERROR_SEGMENT_INVALID;
+	      goto drop;
+	    }
+
+	  /* 5: check the ACK field  */
+	  switch (tc0->state)
+	    {
+	    case TCP_STATE_SYN_RCVD:
+	      /*
+	       * If the segment acknowledgment is not acceptable, form a
+	       * reset segment,
+	       *  
+	       * and send it.
+	       */
+	      if (!tcp_rcv_ack_is_acceptable (tc0, b0))
+		{
+		  tcp_send_reset (b0, is_ip4);
+		  goto drop;
+		}
+	      /* Switch state to ESTABLISHED */
+	      tc0->state = TCP_STATE_ESTABLISHED;
+
+	      /* Initialize session variables */
+	      tc0->snd_una = vnet_buffer (b0)->tcp.ack_number;
+	      tc0->snd_wnd = clib_net_to_host_u32 (tcp0->window)
+		<< tc0->opt.wscale;
+	      tc0->snd_wl1 = vnet_buffer (b0)->tcp.seq_number;
+	      tc0->snd_wl2 = vnet_buffer (b0)->tcp.ack_number;
+
+	      /* Shoulder tap the server */
+	      stream_session_accept_notify (&tc0->connection);
+
+	      tcp_timer_reset (tc0, TCP_TIMER_RETRANSMIT_SYN);
+	      break;
+	    case TCP_STATE_ESTABLISHED:
+	      /* We can get packets in established state here because they
+	       * were enqueued before state change */
+	      if (tcp_rcv_ack (tc0, b0, tcp0, &next0, &error0))
+		goto drop;
+
+	      break;
+	    case TCP_STATE_FIN_WAIT_1:
+	      /* In addition to the processing for the ESTABLISHED state, if
+	       * our FIN is now acknowledged then enter FIN-WAIT-2 and
+	       * continue processing in that state. */
+	      if (tcp_rcv_ack (tc0, b0, tcp0, &next0, &error0))
+		goto drop;
+	      tc0->state = TCP_STATE_FIN_WAIT_2;
+	      /* Stop all timers, 2MSL will be set lower */
+	      tcp_connection_timers_reset (tc0);
+	      break;
+	    case TCP_STATE_FIN_WAIT_2:
+	      /* In addition to the processing for the ESTABLISHED state, if
+	       * the retransmission queue is empty, the user's CLOSE can be
+	       * acknowledged ("ok") but do not delete the TCB. */
+	      if (tcp_rcv_ack (tc0, b0, tcp0, &next0, &error0))
+		goto drop;
+	      /* check if rtx queue is empty and ack CLOSE TODO */
+	      break;
+	    case TCP_STATE_CLOSE_WAIT:
+	      /* Do the same processing as for the ESTABLISHED state. */
+	      if (tcp_rcv_ack (tc0, b0, tcp0, &next0, &error0))
+		goto drop;
+	      break;
+	    case TCP_STATE_CLOSING:
+	      /* In addition to the processing for the ESTABLISHED state, if
+	       * the ACK acknowledges our FIN then enter the TIME-WAIT state,
+	       * otherwise ignore the segment. */
+	      if (tcp_rcv_ack (tc0, b0, tcp0, &next0, &error0))
+		goto drop;
+
+	      /* XXX test that send queue empty */
+	      tc0->state = TCP_STATE_TIME_WAIT;
+	      goto drop;
+
+	      break;
+	    case TCP_STATE_LAST_ACK:
+	      /* The only thing that can arrive in this state is an
+	       * acknowledgment of our FIN. If our FIN is now acknowledged,
+	       * delete the TCB, enter the CLOSED state, and return. */
+
+	      if (!tcp_rcv_ack_is_acceptable (tc0, b0))
+		goto drop;
+
+	      tcp_connection_del (tc0);
+	      goto drop;
+
+	      break;
+	    case TCP_STATE_TIME_WAIT:
+	      /* The only thing that can arrive in this state is a
+	       * retransmission of the remote FIN. Acknowledge it, and restart
+	       * the 2 MSL timeout. */
+
+	      /* TODO */
+	      goto drop;
+	      break;
+	    default:
+	      ASSERT (0);
+	    }
+
+	  /* 6: check the URG bit TODO */
+
+	  /* 7: process the segment text */
+	  switch (tc0->state)
+	    {
+	    case TCP_STATE_ESTABLISHED:
+	    case TCP_STATE_FIN_WAIT_1:
+	    case TCP_STATE_FIN_WAIT_2:
+	      error0 = tcp_segment_rcv (tm, tc0, b0, n_data_bytes0, &next0);
+	      break;
+	    case TCP_STATE_CLOSE_WAIT:
+	    case TCP_STATE_CLOSING:
+	    case TCP_STATE_LAST_ACK:
+	    case TCP_STATE_TIME_WAIT:
+	      /* This should not occur, since a FIN has been received from the
+	       * remote side.  Ignore the segment text. */
+	      break;
+	    }
+
+	  /* 8: check the FIN bit */
+	  if (!tcp_fin (tcp0))
+	    goto drop;
+
+	  switch (tc0->state)
+	    {
+	    case TCP_STATE_ESTABLISHED:
+	    case TCP_STATE_SYN_RCVD:
+	      /* Send FIN-ACK notify app and enter CLOSE-WAIT */
+	      tcp_connection_timers_reset (tc0);
+	      tcp_make_finack (tc0, b0);
+	      next0 = tcp_next_output (tc0->c_is_ip4);
+	      stream_session_disconnect_notify (&tc0->connection);
+	      tc0->state = TCP_STATE_CLOSE_WAIT;
+	      break;
+	    case TCP_STATE_CLOSE_WAIT:
+	    case TCP_STATE_CLOSING:
+	    case TCP_STATE_LAST_ACK:
+	      /* move along .. */
+	      break;
+	    case TCP_STATE_FIN_WAIT_1:
+	      tc0->state = TCP_STATE_TIME_WAIT;
+	      tcp_connection_timers_reset (tc0);
+	      tcp_timer_set (tc0, TCP_TIMER_2MSL, TCP_2MSL_TIME);
+	      break;
+	    case TCP_STATE_FIN_WAIT_2:
+	      /* Got FIN, send ACK! */
+	      tc0->state = TCP_STATE_TIME_WAIT;
+	      tcp_timer_set (tc0, TCP_TIMER_2MSL, TCP_2MSL_TIME);
+	      tcp_make_ack (tc0, b0);
+	      next0 = tcp_next_output (is_ip4);
+	      break;
+	    case TCP_STATE_TIME_WAIT:
+	      /* Remain in the TIME-WAIT state. Restart the 2 MSL time-wait
+	       * timeout.
+	       */
+	      tcp_timer_update (tc0, TCP_TIMER_2MSL, TCP_2MSL_TIME);
+	      break;
+	    }
+
+	  b0->error = error0 ? node->errors[error0] : 0;
+
+	drop:
+	  if (PREDICT_FALSE (b0->flags & VLIB_BUFFER_IS_TRACED))
+	    {
+
+	    }
+
+	  vlib_validate_buffer_enqueue_x1 (vm, node, next_index, to_next,
+					   n_left_to_next, bi0, next0);
+	}
+
+      vlib_put_next_frame (vm, node, next_index, n_left_to_next);
+    }
+
+  errors = session_manager_flush_enqueue_events (my_thread_index);
+  if (errors)
+    {
+      if (is_ip4)
+	vlib_node_increment_counter (vm, tcp4_established_node.index,
+				     TCP_ERROR_EVENT_FIFO_FULL, errors);
+      else
+	vlib_node_increment_counter (vm, tcp6_established_node.index,
+				     TCP_ERROR_EVENT_FIFO_FULL, errors);
+    }
+
+  return from_frame->n_vectors;
+}
+
+static uword
+tcp4_rcv_process (vlib_main_t * vm, vlib_node_runtime_t * node,
+		  vlib_frame_t * from_frame)
+{
+  return tcp46_rcv_process_inline (vm, node, from_frame, 1 /* is_ip4 */ );
+}
+
+static uword
+tcp6_rcv_process (vlib_main_t * vm, vlib_node_runtime_t * node,
+		  vlib_frame_t * from_frame)
+{
+  return tcp46_rcv_process_inline (vm, node, from_frame, 0 /* is_ip4 */ );
+}
+
+/* *INDENT-OFF* */
+VLIB_REGISTER_NODE (tcp4_rcv_process_node) =
+{
+  .function = tcp4_rcv_process,
+  .name = "tcp4-rcv-process",
+  /* Takes a vector of packets. */
+  .vector_size = sizeof (u32),
+  .n_errors = TCP_N_ERROR,
+  .error_strings = tcp_error_strings,
+  .n_next_nodes = TCP_RCV_PROCESS_N_NEXT,
+  .next_nodes =
+  {
+#define _(s,n) [TCP_RCV_PROCESS_NEXT_##s] = n,
+    foreach_tcp_state_next
+#undef _
+  },
+};
+/* *INDENT-ON* */
+
+VLIB_NODE_FUNCTION_MULTIARCH (tcp4_rcv_process_node, tcp4_rcv_process);
+
+/* *INDENT-OFF* */
+VLIB_REGISTER_NODE (tcp6_rcv_process_node) =
+{
+  .function = tcp6_rcv_process,
+  .name = "tcp6-rcv-process",
+  /* Takes a vector of packets. */
+  .vector_size = sizeof (u32),
+  .n_errors = TCP_N_ERROR,
+  .error_strings = tcp_error_strings,
+  .n_next_nodes = TCP_RCV_PROCESS_N_NEXT,
+  .next_nodes =
+  {
+#define _(s,n) [TCP_RCV_PROCESS_NEXT_##s] = n,
+    foreach_tcp_state_next
+#undef _
+  },
+};
+/* *INDENT-ON* */
+
+VLIB_NODE_FUNCTION_MULTIARCH (tcp6_rcv_process_node, tcp6_rcv_process);
+
+vlib_node_registration_t tcp4_listen_node;
+vlib_node_registration_t tcp6_listen_node;
+
+/**
+ * LISTEN state processing as per RFC 793 p. 65
+ */
+always_inline uword
+tcp46_listen_inline (vlib_main_t * vm, vlib_node_runtime_t * node,
+		     vlib_frame_t * from_frame, int is_ip4)
+{
+  u32 n_left_from, next_index, *from, *to_next;
+  u32 my_thread_index = vm->cpu_index;
+  tcp_main_t *tm = vnet_get_tcp_main ();
+  u8 sst = is_ip4 ? SESSION_TYPE_IP4_TCP : SESSION_TYPE_IP6_TCP;
+
+  from = vlib_frame_vector_args (from_frame);
+  n_left_from = from_frame->n_vectors;
+
+  next_index = node->cached_next_index;
+
+  while (n_left_from > 0)
+    {
+      u32 n_left_to_next;
+
+      vlib_get_next_frame (vm, node, next_index, to_next, n_left_to_next);
+
+      while (n_left_from > 0 && n_left_to_next > 0)
+	{
+	  u32 bi0;
+	  vlib_buffer_t *b0;
+	  tcp_header_t *th0 = 0;
+	  tcp_connection_t *lc0;
+	  ip4_header_t *ip40;
+	  ip6_header_t *ip60;
+	  tcp_connection_t *child0;
+	  u32 error0 = TCP_ERROR_SYNS_RCVD, next0 = TCP_LISTEN_NEXT_DROP;
+
+	  bi0 = from[0];
+	  to_next[0] = bi0;
+	  from += 1;
+	  to_next += 1;
+	  n_left_from -= 1;
+	  n_left_to_next -= 1;
+
+	  b0 = vlib_get_buffer (vm, bi0);
+	  lc0 = tcp_listener_get (vnet_buffer (b0)->tcp.connection_index);
+
+	  if (is_ip4)
+	    {
+	      ip40 = vlib_buffer_get_current (b0);
+	      th0 = ip4_next_header (ip40);
+	    }
+	  else
+	    {
+	      ip60 = vlib_buffer_get_current (b0);
+	      th0 = ip6_next_header (ip60);
+	    }
+
+	  /* Create child session. For syn-flood protection use filter */
+
+	  /* 1. first check for an RST */
+	  if (tcp_rst (th0))
+	    goto drop;
+
+	  /* 2. second check for an ACK */
+	  if (tcp_ack (th0))
+	    {
+	      tcp_send_reset (b0, is_ip4);
+	      goto drop;
+	    }
+
+	  /* 3. check for a SYN (did that already) */
+
+	  /* Create child session and send SYN-ACK */
+	  pool_get (tm->connections[my_thread_index], child0);
+	  memset (child0, 0, sizeof (*child0));
+
+	  child0->c_c_index = child0 - tm->connections[my_thread_index];
+	  child0->c_lcl_port = lc0->c_lcl_port;
+	  child0->c_rmt_port = th0->src_port;
+	  child0->c_is_ip4 = is_ip4;
+	  child0->c_thread_index = my_thread_index;
+
+	  if (is_ip4)
+	    {
+	      child0->c_lcl_ip4.as_u32 = ip40->dst_address.as_u32;
+	      child0->c_rmt_ip4.as_u32 = ip40->src_address.as_u32;
+	    }
+	  else
+	    {
+	      clib_memcpy (&child0->c_lcl_ip6, &ip60->dst_address,
+			   sizeof (ip6_address_t));
+	      clib_memcpy (&child0->c_rmt_ip6, &ip60->src_address,
+			   sizeof (ip6_address_t));
+	    }
+
+	  if (stream_session_accept (&child0->connection, lc0->c_s_index, sst,
+				     0 /* notify */ ))
+	    {
+	      error0 = TCP_ERROR_CREATE_SESSION_FAIL;
+	      goto drop;
+	    }
+
+	  tcp_options_parse (th0, &child0->opt);
+	  tcp_connection_init_vars (child0);
+
+	  child0->irs = vnet_buffer (b0)->tcp.seq_number;
+	  child0->rcv_nxt = vnet_buffer (b0)->tcp.seq_number + 1;
+	  child0->state = TCP_STATE_SYN_RCVD;
+
+	  /* RFC1323: TSval timestamps sent on {SYN} and {SYN,ACK}
+	   * segments are used to initialize PAWS. */
+	  if (tcp_opts_tstamp (&child0->opt))
+	    {
+	      child0->tsval_recent = child0->opt.tsval;
+	      child0->tsval_recent_age = tcp_time_now ();
+	    }
+
+	  /* Reuse buffer to make syn-ack and send */
+	  tcp_make_synack (child0, b0);
+	  next0 = tcp_next_output (is_ip4);
+
+	drop:
+	  if (PREDICT_FALSE (b0->flags & VLIB_BUFFER_IS_TRACED))
+	    {
+
+	    }
+
+	  b0->error = error0 ? node->errors[error0] : 0;
+
+	  vlib_validate_buffer_enqueue_x1 (vm, node, next_index, to_next,
+					   n_left_to_next, bi0, next0);
+	}
+
+      vlib_put_next_frame (vm, node, next_index, n_left_to_next);
+    }
+  return from_frame->n_vectors;
+}
+
+static uword
+tcp4_listen (vlib_main_t * vm, vlib_node_runtime_t * node,
+	     vlib_frame_t * from_frame)
+{
+  return tcp46_listen_inline (vm, node, from_frame, 1 /* is_ip4 */ );
+}
+
+static uword
+tcp6_listen (vlib_main_t * vm, vlib_node_runtime_t * node,
+	     vlib_frame_t * from_frame)
+{
+  return tcp46_listen_inline (vm, node, from_frame, 0 /* is_ip4 */ );
+}
+
+/* *INDENT-OFF* */
+VLIB_REGISTER_NODE (tcp4_listen_node) =
+{
+  .function = tcp4_listen,
+  .name = "tcp4-listen",
+  /* Takes a vector of packets. */
+  .vector_size = sizeof (u32),
+  .n_errors = TCP_N_ERROR,
+  .error_strings = tcp_error_strings,
+  .n_next_nodes = TCP_LISTEN_N_NEXT,
+  .next_nodes =
+  {
+#define _(s,n) [TCP_LISTEN_NEXT_##s] = n,
+    foreach_tcp_state_next
+#undef _
+  },
+};
+/* *INDENT-ON* */
+
+VLIB_NODE_FUNCTION_MULTIARCH (tcp4_listen_node, tcp4_listen);
+
+/* *INDENT-OFF* */
+VLIB_REGISTER_NODE (tcp6_listen_node) =
+{
+  .function = tcp6_listen,
+  .name = "tcp6-listen",
+  /* Takes a vector of packets. */
+  .vector_size = sizeof (u32),
+  .n_errors = TCP_N_ERROR,
+  .error_strings = tcp_error_strings,
+  .n_next_nodes = TCP_LISTEN_N_NEXT,
+  .next_nodes =
+  {
+#define _(s,n) [TCP_LISTEN_NEXT_##s] = n,
+    foreach_tcp_state_next
+#undef _
+  },
+};
+/* *INDENT-ON* */
+
+VLIB_NODE_FUNCTION_MULTIARCH (tcp6_listen_node, tcp6_listen);
+
+vlib_node_registration_t tcp4_input_node;
+vlib_node_registration_t tcp6_input_node;
+
+typedef enum _tcp_input_next
+{
+  TCP_INPUT_NEXT_DROP,
+  TCP_INPUT_NEXT_LISTEN,
+  TCP_INPUT_NEXT_RCV_PROCESS,
+  TCP_INPUT_NEXT_SYN_SENT,
+  TCP_INPUT_NEXT_ESTABLISHED,
+  TCP_INPUT_NEXT_RESET,
+  TCP_INPUT_N_NEXT
+} tcp_input_next_t;
+
+#define foreach_tcp4_input_next                 \
+  _ (DROP, "error-drop")                        \
+  _ (LISTEN, "tcp4-listen")                     \
+  _ (RCV_PROCESS, "tcp4-rcv-process")           \
+  _ (SYN_SENT, "tcp4-syn-sent")                 \
+  _ (ESTABLISHED, "tcp4-established")		\
+  _ (RESET, "tcp4-reset")
+
+#define foreach_tcp6_input_next                 \
+  _ (DROP, "error-drop")                        \
+  _ (LISTEN, "tcp6-listen")                     \
+  _ (RCV_PROCESS, "tcp6-rcv-process")           \
+  _ (SYN_SENT, "tcp6-syn-sent")                 \
+  _ (ESTABLISHED, "tcp6-established")		\
+  _ (RESET, "tcp6-reset")
+
+typedef struct
+{
+  u16 src_port;
+  u16 dst_port;
+  u8 state;
+} tcp_rx_trace_t;
+
+const char *tcp_fsm_states[] = {
+#define _(sym, str) str,
+  foreach_tcp_fsm_state
+#undef _
+};
+
+u8 *
+format_tcp_state (u8 * s, va_list * args)
+{
+  tcp_state_t *state = va_arg (*args, tcp_state_t *);
+
+  if (state[0] < TCP_N_STATES)
+    s = format (s, "%s", tcp_fsm_states[state[0]]);
+  else
+    s = format (s, "UNKNOWN");
+
+  return s;
+}
+
+u8 *
+format_tcp_rx_trace (u8 * s, va_list * args)
+{
+  CLIB_UNUSED (vlib_main_t * vm) = va_arg (*args, vlib_main_t *);
+  CLIB_UNUSED (vlib_node_t * node) = va_arg (*args, vlib_node_t *);
+  tcp_rx_trace_t *t = va_arg (*args, tcp_rx_trace_t *);
+
+  s = format (s, "TCP: src-port %d dst-port %U%s\n",
+	      clib_net_to_host_u16 (t->src_port),
+	      clib_net_to_host_u16 (t->dst_port), format_tcp_state, t->state);
+
+  return s;
+}
+
+#define filter_flags (TCP_FLAG_SYN|TCP_FLAG_ACK|TCP_FLAG_RST|TCP_FLAG_FIN)
+
+always_inline uword
+tcp46_input_inline (vlib_main_t * vm, vlib_node_runtime_t * node,
+		    vlib_frame_t * from_frame, int is_ip4)
+{
+  u32 n_left_from, next_index, *from, *to_next;
+  u32 my_thread_index = vm->cpu_index;
+  tcp_main_t *tm = vnet_get_tcp_main ();
+  session_manager_main_t *ssm = vnet_get_session_manager_main ();
+
+  from = vlib_frame_vector_args (from_frame);
+  n_left_from = from_frame->n_vectors;
+
+  next_index = node->cached_next_index;
+
+  while (n_left_from > 0)
+    {
+      u32 n_left_to_next;
+
+      vlib_get_next_frame (vm, node, next_index, to_next, n_left_to_next);
+
+      while (n_left_from > 0 && n_left_to_next > 0)
+	{
+	  u32 bi0;
+	  vlib_buffer_t *b0;
+	  tcp_header_t *tcp0 = 0;
+	  tcp_connection_t *tc0;
+	  ip4_header_t *ip40;
+	  ip6_header_t *ip60;
+	  u32 error0 = TCP_ERROR_NO_LISTENER, next0 = TCP_INPUT_NEXT_DROP;
+	  u8 flags0;
+
+	  bi0 = from[0];
+	  to_next[0] = bi0;
+	  from += 1;
+	  to_next += 1;
+	  n_left_from -= 1;
+	  n_left_to_next -= 1;
+
+	  b0 = vlib_get_buffer (vm, bi0);
+
+	  if (is_ip4)
+	    {
+	      ip40 = vlib_buffer_get_current (b0);
+	      tcp0 = ip4_next_header (ip40);
+
+	      /* lookup session */
+	      tc0 =
+		(tcp_connection_t *) stream_session_lookup_transport4 (ssm,
+								       &ip40->dst_address,
+								       &ip40->src_address,
+								       tcp0->dst_port,
+								       tcp0->src_port,
+								       SESSION_TYPE_IP4_TCP,
+								       my_thread_index);
+	    }
+	  else
+	    {
+	      ip60 = vlib_buffer_get_current (b0);
+	      tcp0 = ip6_next_header (ip60);
+	      tc0 =
+		(tcp_connection_t *) stream_session_lookup_transport6 (ssm,
+								       &ip60->src_address,
+								       &ip60->dst_address,
+								       tcp0->src_port,
+								       tcp0->dst_port,
+								       SESSION_TYPE_IP6_TCP,
+								       my_thread_index);
+	    }
+
+	  /* Session exists */
+	  if (PREDICT_TRUE (0 != tc0))
+	    {
+	      /* Save connection index */
+	      vnet_buffer (b0)->tcp.connection_index = tc0->c_c_index;
+	      vnet_buffer (b0)->tcp.seq_number =
+		clib_net_to_host_u32 (tcp0->seq_number);
+	      vnet_buffer (b0)->tcp.ack_number =
+		clib_net_to_host_u32 (tcp0->ack_number);
+
+	      flags0 = tcp0->flags & filter_flags;
+	      next0 = tm->dispatch_table[tc0->state][flags0].next;
+	      error0 = tm->dispatch_table[tc0->state][flags0].error;
+
+	      if (PREDICT_FALSE (error0 == TCP_ERROR_DISPATCH))
+		{
+		  /* Overload tcp flags to store state */
+		  vnet_buffer (b0)->tcp.flags = tc0->state;
+		}
+	    }
+	  else
+	    {
+	      /* Send reset */
+	      next0 = TCP_INPUT_NEXT_RESET;
+	      error0 = TCP_ERROR_NO_LISTENER;
+	      vnet_buffer (b0)->tcp.flags = 0;
+	    }
+
+	  b0->error = error0 ? node->errors[error0] : 0;
+
+	  if (PREDICT_FALSE (b0->flags & VLIB_BUFFER_IS_TRACED))
+	    {
+
+	    }
+
+	  vlib_validate_buffer_enqueue_x1 (vm, node, next_index, to_next,
+					   n_left_to_next, bi0, next0);
+	}
+
+      vlib_put_next_frame (vm, node, next_index, n_left_to_next);
+    }
+
+  return from_frame->n_vectors;
+}
+
+static uword
+tcp4_input (vlib_main_t * vm, vlib_node_runtime_t * node,
+	    vlib_frame_t * from_frame)
+{
+  return tcp46_input_inline (vm, node, from_frame, 1 /* is_ip4 */ );
+}
+
+static uword
+tcp6_input (vlib_main_t * vm, vlib_node_runtime_t * node,
+	    vlib_frame_t * from_frame)
+{
+  return tcp46_input_inline (vm, node, from_frame, 0 /* is_ip4 */ );
+}
+
+/* *INDENT-OFF* */
+VLIB_REGISTER_NODE (tcp4_input_node) =
+{
+  .function = tcp4_input,
+  .name = "tcp4-input",
+  /* Takes a vector of packets. */
+  .vector_size = sizeof (u32),
+  .n_errors = TCP_N_ERROR,
+  .error_strings = tcp_error_strings,
+  .n_next_nodes = TCP_INPUT_N_NEXT,
+  .next_nodes =
+  {
+#define _(s,n) [TCP_INPUT_NEXT_##s] = n,
+    foreach_tcp4_input_next
+#undef _
+  },
+  .format_buffer = format_tcp_header,
+  .format_trace = format_tcp_rx_trace,
+};
+/* *INDENT-ON* */
+
+VLIB_NODE_FUNCTION_MULTIARCH (tcp4_input_node, tcp4_input);
+
+/* *INDENT-OFF* */
+VLIB_REGISTER_NODE (tcp6_input_node) =
+{
+  .function = tcp6_input,
+  .name = "tcp6-input",
+  /* Takes a vector of packets. */
+  .vector_size = sizeof (u32),
+  .n_errors = TCP_N_ERROR,
+  .error_strings = tcp_error_strings,
+  .n_next_nodes = TCP_INPUT_N_NEXT,
+  .next_nodes =
+  {
+#define _(s,n) [TCP_INPUT_NEXT_##s] = n,
+    foreach_tcp6_input_next
+#undef _
+  },
+  .format_buffer = format_tcp_header,
+  .format_trace = format_tcp_rx_trace,
+};
+/* *INDENT-ON* */
+
+VLIB_NODE_FUNCTION_MULTIARCH (tcp6_input_node, tcp6_input);
+void
+tcp_update_time (f64 now, u32 thread_index)
+{
+  tcp_main_t *tm = vnet_get_tcp_main ();
+  tw_timer_expire_timers_16t_2w_512sl (&tm->timer_wheels[thread_index], now);
+}
+
+static void
+tcp_dispatch_table_init (tcp_main_t * tm)
+{
+  int i, j;
+  for (i = 0; i < ARRAY_LEN (tm->dispatch_table); i++)
+    for (j = 0; j < ARRAY_LEN (tm->dispatch_table[i]); j++)
+      {
+	tm->dispatch_table[i][j].next = TCP_INPUT_NEXT_DROP;
+	tm->dispatch_table[i][j].error = TCP_ERROR_DISPATCH;
+      }
+
+#define _(t,f,n,e)                                           	\
+do {                                                       	\
+    tm->dispatch_table[TCP_STATE_##t][f].next = (n);         	\
+    tm->dispatch_table[TCP_STATE_##t][f].error = (e);        	\
+} while (0)
+
+  /* SYNs for new connections -> tcp-listen. */
+  _(LISTEN, TCP_FLAG_SYN, TCP_INPUT_NEXT_LISTEN, TCP_ERROR_NONE);
+  /* ACK for for a SYN-ACK -> tcp-rcv-process. */
+  _(SYN_RCVD, TCP_FLAG_ACK, TCP_INPUT_NEXT_RCV_PROCESS, TCP_ERROR_NONE);
+  /* SYN-ACK for a SYN */
+  _(SYN_SENT, TCP_FLAG_SYN | TCP_FLAG_ACK, TCP_INPUT_NEXT_SYN_SENT,
+    TCP_ERROR_NONE);
+  _(SYN_SENT, TCP_FLAG_ACK, TCP_INPUT_NEXT_SYN_SENT, TCP_ERROR_NONE);
+  _(SYN_SENT, TCP_FLAG_RST, TCP_INPUT_NEXT_SYN_SENT, TCP_ERROR_NONE);
+  _(SYN_SENT, TCP_FLAG_RST | TCP_FLAG_ACK, TCP_INPUT_NEXT_SYN_SENT,
+    TCP_ERROR_NONE);
+  /* ACK for for established connection -> tcp-established. */
+  _(ESTABLISHED, TCP_FLAG_ACK, TCP_INPUT_NEXT_ESTABLISHED, TCP_ERROR_NONE);
+  /* FIN for for established connection -> tcp-established. */
+  _(ESTABLISHED, TCP_FLAG_FIN, TCP_INPUT_NEXT_ESTABLISHED, TCP_ERROR_NONE);
+  _(ESTABLISHED, TCP_FLAG_FIN | TCP_FLAG_ACK, TCP_INPUT_NEXT_ESTABLISHED,
+    TCP_ERROR_NONE);
+  /* ACK or FIN-ACK to our FIN */
+  _(FIN_WAIT_1, TCP_FLAG_ACK, TCP_INPUT_NEXT_RCV_PROCESS, TCP_ERROR_NONE);
+  _(FIN_WAIT_1, TCP_FLAG_ACK | TCP_FLAG_FIN, TCP_INPUT_NEXT_RCV_PROCESS,
+    TCP_ERROR_NONE);
+  /* FIN in reply to our FIN from the other side */
+  _(FIN_WAIT_1, TCP_FLAG_FIN, TCP_INPUT_NEXT_RCV_PROCESS, TCP_ERROR_NONE);
+  /* FIN confirming that the peer (app) has closed */
+  _(FIN_WAIT_2, TCP_FLAG_FIN, TCP_INPUT_NEXT_RCV_PROCESS, TCP_ERROR_NONE);
+  _(FIN_WAIT_2, TCP_FLAG_FIN | TCP_FLAG_ACK, TCP_INPUT_NEXT_RCV_PROCESS,
+    TCP_ERROR_NONE);
+  _(LAST_ACK, TCP_FLAG_ACK, TCP_INPUT_NEXT_RCV_PROCESS, TCP_ERROR_NONE);
+#undef _
+}
+
+clib_error_t *
+tcp_input_init (vlib_main_t * vm)
+{
+  clib_error_t *error = 0;
+  tcp_main_t *tm = vnet_get_tcp_main ();
+
+  if ((error = vlib_call_init_function (vm, tcp_init)))
+    return error;
+
+  /* Initialize dispatch table. */
+  tcp_dispatch_table_init (tm);
+
+  return error;
+}
+
+VLIB_INIT_FUNCTION (tcp_input_init);
+
+/*
+ * fd.io coding-style-patch-verification: ON
+ *
+ * Local Variables:
+ * eval: (c-set-style "gnu")
+ * End:
+ */
diff --git a/src/vnet/tcp/tcp_newreno.c b/src/vnet/tcp/tcp_newreno.c
new file mode 100644
index 00000000..856dffe4
--- /dev/null
+++ b/src/vnet/tcp/tcp_newreno.c
@@ -0,0 +1,93 @@
+/*
+ * Copyright (c) 2017 Cisco and/or its affiliates.
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at:
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#include 
+
+void
+newreno_congestion (tcp_connection_t * tc)
+{
+  tc->prev_ssthresh = tc->ssthresh;
+  tc->ssthresh = clib_max (tcp_flight_size (tc) / 2, 2 * tc->snd_mss);
+}
+
+void
+newreno_recovered (tcp_connection_t * tc)
+{
+  tc->cwnd = tc->ssthresh;
+}
+
+void
+newreno_rcv_ack (tcp_connection_t * tc)
+{
+  if (tcp_in_slowstart (tc))
+    {
+      tc->cwnd += clib_min (tc->snd_mss, tc->bytes_acked);
+    }
+  else
+    {
+      /* Round up to 1 if needed */
+      tc->cwnd += clib_max (tc->snd_mss * tc->snd_mss / tc->cwnd, 1);
+    }
+}
+
+void
+newreno_rcv_cong_ack (tcp_connection_t * tc, tcp_cc_ack_t ack_type)
+{
+  if (ack_type == TCP_CC_DUPACK)
+    {
+      tc->cwnd += tc->snd_mss;
+    }
+  else if (ack_type == TCP_CC_PARTIALACK)
+    {
+      tc->cwnd -= tc->bytes_acked;
+      if (tc->bytes_acked > tc->snd_mss)
+	tc->bytes_acked += tc->snd_mss;
+    }
+}
+
+void
+newreno_conn_init (tcp_connection_t * tc)
+{
+  tc->ssthresh = tc->snd_wnd;
+  tc->cwnd = tcp_initial_cwnd (tc);
+}
+
+const static tcp_cc_algorithm_t tcp_newreno = {
+  .congestion = newreno_congestion,
+  .recovered = newreno_recovered,
+  .rcv_ack = newreno_rcv_ack,
+  .rcv_cong_ack = newreno_rcv_cong_ack,
+  .init = newreno_conn_init
+};
+
+clib_error_t *
+newreno_init (vlib_main_t * vm)
+{
+  clib_error_t *error = 0;
+
+  tcp_cc_algo_register (TCP_CC_NEWRENO, &tcp_newreno);
+
+  return error;
+}
+
+VLIB_INIT_FUNCTION (newreno_init);
+
+/*
+ * fd.io coding-style-patch-verification: ON
+ *
+ * Local Variables:
+ * eval: (c-set-style "gnu")
+ * End:
+ */
diff --git a/src/vnet/tcp/tcp_output.c b/src/vnet/tcp/tcp_output.c
new file mode 100644
index 00000000..dbcf1f74
--- /dev/null
+++ b/src/vnet/tcp/tcp_output.c
@@ -0,0 +1,1412 @@
+/*
+ * Copyright (c) 2016 Cisco and/or its affiliates.
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at:
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#include 
+#include 
+
+vlib_node_registration_t tcp4_output_node;
+vlib_node_registration_t tcp6_output_node;
+
+typedef enum _tcp_output_nect
+{
+  TCP_OUTPUT_NEXT_DROP,
+  TCP_OUTPUT_NEXT_IP_LOOKUP,
+  TCP_OUTPUT_N_NEXT
+} tcp_output_next_t;
+
+#define foreach_tcp4_output_next              	\
+  _ (DROP, "error-drop")                        \
+  _ (IP_LOOKUP, "ip4-lookup")
+
+#define foreach_tcp6_output_next              	\
+  _ (DROP, "error-drop")                        \
+  _ (IP_LOOKUP, "ip6-lookup")
+
+static char *tcp_error_strings[] = {
+#define tcp_error(n,s) s,
+#include 
+#undef tcp_error
+};
+
+typedef struct
+{
+  u16 src_port;
+  u16 dst_port;
+  u8 state;
+} tcp_tx_trace_t;
+
+u16 dummy_mtu = 400;
+
+u8 *
+format_tcp_tx_trace (u8 * s, va_list * args)
+{
+  CLIB_UNUSED (vlib_main_t * vm) = va_arg (*args, vlib_main_t *);
+  CLIB_UNUSED (vlib_node_t * node) = va_arg (*args, vlib_node_t *);
+
+  s = format (s, "TBD\n");
+
+  return s;
+}
+
+void
+tcp_set_snd_mss (tcp_connection_t * tc)
+{
+  u16 snd_mss;
+
+  /* TODO find our iface MTU */
+  snd_mss = dummy_mtu;
+
+  /* TODO cache mss and consider PMTU discovery */
+  snd_mss = tc->opt.mss < snd_mss ? tc->opt.mss : snd_mss;
+
+  tc->snd_mss = snd_mss;
+
+  if (tc->snd_mss == 0)
+    {
+      clib_warning ("snd mss is 0");
+      tc->snd_mss = dummy_mtu;
+    }
+}
+
+static u8
+tcp_window_compute_scale (u32 available_space)
+{
+  u8 wnd_scale = 0;
+  while (wnd_scale < TCP_MAX_WND_SCALE
+	 && (available_space >> wnd_scale) > TCP_WND_MAX)
+    wnd_scale++;
+  return wnd_scale;
+}
+
+/**
+ * Compute initial window and scale factor. As per RFC1323, window field in
+ * SYN and SYN-ACK segments is never scaled.
+ */
+u32
+tcp_initial_window_to_advertise (tcp_connection_t * tc)
+{
+  u32 available_space;
+
+  /* Initial wnd for SYN. Fifos are not allocated yet.
+   * Use some predefined value */
+  if (tc->state != TCP_STATE_SYN_RCVD)
+    {
+      return TCP_DEFAULT_RX_FIFO_SIZE;
+    }
+
+  available_space = stream_session_max_enqueue (&tc->connection);
+  tc->rcv_wscale = tcp_window_compute_scale (available_space);
+  tc->rcv_wnd = clib_min (available_space, TCP_WND_MAX << tc->rcv_wscale);
+
+  return clib_min (tc->rcv_wnd, TCP_WND_MAX);
+}
+
+/**
+ * Compute and return window to advertise, scaled as per RFC1323
+ */
+u32
+tcp_window_to_advertise (tcp_connection_t * tc, tcp_state_t state)
+{
+  u32 available_space, wnd, scaled_space;
+
+  if (state != TCP_STATE_ESTABLISHED)
+    return tcp_initial_window_to_advertise (tc);
+
+  available_space = stream_session_max_enqueue (&tc->connection);
+  scaled_space = available_space >> tc->rcv_wscale;
+
+  /* Need to update scale */
+  if (PREDICT_FALSE ((scaled_space == 0 && available_space != 0))
+      || (scaled_space >= TCP_WND_MAX))
+    tc->rcv_wscale = tcp_window_compute_scale (available_space);
+
+  wnd = clib_min (available_space, TCP_WND_MAX << tc->rcv_wscale);
+  tc->rcv_wnd = wnd;
+
+  return wnd >> tc->rcv_wscale;
+}
+
+/**
+ * Write TCP options to segment.
+ */
+u32
+tcp_options_write (u8 * data, tcp_options_t * opts)
+{
+  u32 opts_len = 0;
+  u32 buf, seq_len = 4;
+
+  if (tcp_opts_mss (opts))
+    {
+      *data++ = TCP_OPTION_MSS;
+      *data++ = TCP_OPTION_LEN_MSS;
+      buf = clib_host_to_net_u16 (opts->mss);
+      clib_memcpy (data, &buf, sizeof (opts->mss));
+      data += sizeof (opts->mss);
+      opts_len += TCP_OPTION_LEN_MSS;
+    }
+
+  if (tcp_opts_wscale (opts))
+    {
+      *data++ = TCP_OPTION_WINDOW_SCALE;
+      *data++ = TCP_OPTION_LEN_WINDOW_SCALE;
+      *data++ = opts->wscale;
+      opts_len += TCP_OPTION_LEN_WINDOW_SCALE;
+    }
+
+  if (tcp_opts_sack_permitted (opts))
+    {
+      *data++ = TCP_OPTION_SACK_PERMITTED;
+      *data++ = TCP_OPTION_LEN_SACK_PERMITTED;
+      opts_len += TCP_OPTION_LEN_SACK_PERMITTED;
+    }
+
+  if (tcp_opts_tstamp (opts))
+    {
+      *data++ = TCP_OPTION_TIMESTAMP;
+      *data++ = TCP_OPTION_LEN_TIMESTAMP;
+      buf = clib_host_to_net_u32 (opts->tsval);
+      clib_memcpy (data, &buf, sizeof (opts->tsval));
+      data += sizeof (opts->tsval);
+      buf = clib_host_to_net_u32 (opts->tsecr);
+      clib_memcpy (data, &buf, sizeof (opts->tsecr));
+      data += sizeof (opts->tsecr);
+      opts_len += TCP_OPTION_LEN_TIMESTAMP;
+    }
+
+  if (tcp_opts_sack (opts))
+    {
+      int i;
+      u32 n_sack_blocks = clib_min (vec_len (opts->sacks),
+				    TCP_OPTS_MAX_SACK_BLOCKS);
+
+      if (n_sack_blocks != 0)
+	{
+	  *data++ = TCP_OPTION_SACK_BLOCK;
+	  *data++ = 2 + n_sack_blocks * TCP_OPTION_LEN_SACK_BLOCK;
+	  for (i = 0; i < n_sack_blocks; i++)
+	    {
+	      buf = clib_host_to_net_u32 (opts->sacks[i].start);
+	      clib_memcpy (data, &buf, seq_len);
+	      data += seq_len;
+	      buf = clib_host_to_net_u32 (opts->sacks[i].end);
+	      clib_memcpy (data, &buf, seq_len);
+	      data += seq_len;
+	    }
+	  opts_len += 2 + n_sack_blocks * TCP_OPTION_LEN_SACK_BLOCK;
+	}
+    }
+
+  /* Terminate TCP options */
+  if (opts_len % 4)
+    {
+      *data++ = TCP_OPTION_EOL;
+      opts_len += TCP_OPTION_LEN_EOL;
+    }
+
+  /* Pad with zeroes to a u32 boundary */
+  while (opts_len % 4)
+    {
+      *data++ = TCP_OPTION_NOOP;
+      opts_len += TCP_OPTION_LEN_NOOP;
+    }
+  return opts_len;
+}
+
+always_inline int
+tcp_make_syn_options (tcp_options_t * opts, u32 initial_wnd)
+{
+  u8 len = 0;
+
+  opts->flags |= TCP_OPTS_FLAG_MSS;
+  opts->mss = dummy_mtu;	/*XXX discover that */
+  len += TCP_OPTION_LEN_MSS;
+
+  opts->flags |= TCP_OPTS_FLAG_WSCALE;
+  opts->wscale = tcp_window_compute_scale (initial_wnd);
+  len += TCP_OPTION_LEN_WINDOW_SCALE;
+
+  opts->flags |= TCP_OPTS_FLAG_TSTAMP;
+  opts->tsval = tcp_time_now ();
+  opts->tsecr = 0;
+  len += TCP_OPTION_LEN_TIMESTAMP;
+
+  opts->flags |= TCP_OPTS_FLAG_SACK_PERMITTED;
+  len += TCP_OPTION_LEN_SACK_PERMITTED;
+
+  /* Align to needed boundary */
+  len += (TCP_OPTS_ALIGN - len % TCP_OPTS_ALIGN) % TCP_OPTS_ALIGN;
+  return len;
+}
+
+always_inline int
+tcp_make_synack_options (tcp_connection_t * tc, tcp_options_t * opts)
+{
+  u8 len = 0;
+
+  opts->flags |= TCP_OPTS_FLAG_MSS;
+  opts->mss = dummy_mtu;	/*XXX discover that */
+  len += TCP_OPTION_LEN_MSS;
+
+  if (tcp_opts_wscale (&tc->opt))
+    {
+      opts->flags |= TCP_OPTS_FLAG_WSCALE;
+      opts->wscale = tc->rcv_wscale;
+      len += TCP_OPTION_LEN_WINDOW_SCALE;
+    }
+
+  if (tcp_opts_tstamp (&tc->opt))
+    {
+      opts->flags |= TCP_OPTS_FLAG_TSTAMP;
+      opts->tsval = tcp_time_now ();
+      opts->tsecr = tc->tsval_recent;
+      len += TCP_OPTION_LEN_TIMESTAMP;
+    }
+
+  if (tcp_opts_sack_permitted (&tc->opt))
+    {
+      opts->flags |= TCP_OPTS_FLAG_SACK_PERMITTED;
+      len += TCP_OPTION_LEN_SACK_PERMITTED;
+    }
+
+  /* Align to needed boundary */
+  len += (TCP_OPTS_ALIGN - len % TCP_OPTS_ALIGN) % TCP_OPTS_ALIGN;
+  return len;
+}
+
+always_inline int
+tcp_make_established_options (tcp_connection_t * tc, tcp_options_t * opts)
+{
+  u8 len = 0;
+
+  opts->flags = 0;
+
+  if (tcp_opts_tstamp (&tc->opt))
+    {
+      opts->flags |= TCP_OPTS_FLAG_TSTAMP;
+      opts->tsval = tcp_time_now ();
+      opts->tsecr = tc->tsval_recent;
+      len += TCP_OPTION_LEN_TIMESTAMP;
+    }
+  if (tcp_opts_sack_permitted (&tc->opt))
+    {
+      if (vec_len (tc->snd_sacks))
+	{
+	  opts->flags |= TCP_OPTS_FLAG_SACK;
+	  opts->sacks = tc->snd_sacks;
+	  opts->n_sack_blocks = vec_len (tc->snd_sacks);
+	  len += 2 + TCP_OPTION_LEN_SACK_BLOCK * opts->n_sack_blocks;
+	}
+    }
+
+  /* Align to needed boundary */
+  len += (TCP_OPTS_ALIGN - len % TCP_OPTS_ALIGN) % TCP_OPTS_ALIGN;
+  return len;
+}
+
+always_inline int
+tcp_make_options (tcp_connection_t * tc, tcp_options_t * opts,
+		  tcp_state_t state)
+{
+  switch (state)
+    {
+    case TCP_STATE_ESTABLISHED:
+    case TCP_STATE_FIN_WAIT_1:
+      return tcp_make_established_options (tc, opts);
+    case TCP_STATE_SYN_RCVD:
+      return tcp_make_synack_options (tc, opts);
+    case TCP_STATE_SYN_SENT:
+      return tcp_make_syn_options (opts,
+				   tcp_initial_window_to_advertise (tc));
+    default:
+      clib_warning ("Not handled!");
+      return 0;
+    }
+}
+
+#define tcp_get_free_buffer_index(tm, bidx)                             \
+do {                                                                    \
+  u32 *my_tx_buffers, n_free_buffers;                                   \
+  u32 cpu_index = tm->vlib_main->cpu_index;                             \
+  my_tx_buffers = tm->tx_buffers[cpu_index];                            \
+  if (PREDICT_FALSE(vec_len (my_tx_buffers) == 0))                      \
+    {                                                                   \
+      n_free_buffers = 32;      /* TODO config or macro */              \
+      vec_validate (my_tx_buffers, n_free_buffers - 1);                 \
+      _vec_len(my_tx_buffers) = vlib_buffer_alloc_from_free_list (      \
+          tm->vlib_main, my_tx_buffers, n_free_buffers,                 \
+          VLIB_BUFFER_DEFAULT_FREE_LIST_INDEX);                         \
+      tm->tx_buffers[cpu_index] = my_tx_buffers;                        \
+    }                                                                   \
+  /* buffer shortage */                                                 \
+  if (PREDICT_FALSE (vec_len (my_tx_buffers) == 0))                     \
+    return;                                                             \
+  *bidx = my_tx_buffers[_vec_len (my_tx_buffers)-1];                    \
+  _vec_len (my_tx_buffers) -= 1;                                        \
+} while (0)
+
+always_inline void
+tcp_reuse_buffer (vlib_main_t * vm, vlib_buffer_t * b)
+{
+  vlib_buffer_t *it = b;
+  do
+    {
+      it->current_data = 0;
+      it->current_length = 0;
+      it->total_length_not_including_first_buffer = 0;
+    }
+  while ((it->flags & VLIB_BUFFER_NEXT_PRESENT)
+	 && (it = vlib_get_buffer (vm, it->next_buffer)));
+
+  /* Leave enough space for headers */
+  vlib_buffer_make_headroom (b, MAX_HDRS_LEN);
+}
+
+/**
+ * Prepare ACK
+ */
+void
+tcp_make_ack_i (tcp_connection_t * tc, vlib_buffer_t * b, tcp_state_t state,
+		u8 flags)
+{
+  tcp_options_t _snd_opts, *snd_opts = &_snd_opts;
+  u8 tcp_opts_len, tcp_hdr_opts_len;
+  tcp_header_t *th;
+  u16 wnd;
+
+  wnd = tcp_window_to_advertise (tc, state);
+
+  /* Make and write options */
+  tcp_opts_len = tcp_make_established_options (tc, snd_opts);
+  tcp_hdr_opts_len = tcp_opts_len + sizeof (tcp_header_t);
+
+  th = vlib_buffer_push_tcp (b, tc->c_lcl_port, tc->c_rmt_port, tc->snd_nxt,
+			     tc->rcv_nxt, tcp_hdr_opts_len, flags, wnd);
+
+  tcp_options_write ((u8 *) (th + 1), snd_opts);
+
+  /* Mark as ACK */
+  vnet_buffer (b)->tcp.connection_index = tc->c_c_index;
+}
+
+/**
+ * Convert buffer to ACK
+ */
+void
+tcp_make_ack (tcp_connection_t * tc, vlib_buffer_t * b)
+{
+  tcp_main_t *tm = vnet_get_tcp_main ();
+  vlib_main_t *vm = tm->vlib_main;
+
+  tcp_reuse_buffer (vm, b);
+  tcp_make_ack_i (tc, b, TCP_STATE_ESTABLISHED, TCP_FLAG_ACK);
+  vnet_buffer (b)->tcp.flags = TCP_BUF_FLAG_ACK;
+}
+
+/**
+ * Convert buffer to FIN-ACK
+ */
+void
+tcp_make_finack (tcp_connection_t * tc, vlib_buffer_t * b)
+{
+  tcp_main_t *tm = vnet_get_tcp_main ();
+  vlib_main_t *vm = tm->vlib_main;
+
+  tcp_reuse_buffer (vm, b);
+  tcp_make_ack_i (tc, b, TCP_STATE_ESTABLISHED, TCP_FLAG_ACK | TCP_FLAG_FIN);
+
+  /* Reset flags, make sure ack is sent */
+  tc->flags = TCP_CONN_SNDACK;
+  vnet_buffer (b)->tcp.flags &= ~TCP_BUF_FLAG_DUPACK;
+
+  tc->snd_nxt += 1;
+}
+
+/**
+ * Convert buffer to SYN-ACK
+ */
+void
+tcp_make_synack (tcp_connection_t * tc, vlib_buffer_t * b)
+{
+  tcp_main_t *tm = vnet_get_tcp_main ();
+  vlib_main_t *vm = tm->vlib_main;
+  tcp_options_t _snd_opts, *snd_opts = &_snd_opts;
+  u8 tcp_opts_len, tcp_hdr_opts_len;
+  tcp_header_t *th;
+  u16 initial_wnd;
+  u32 time_now;
+
+  memset (snd_opts, 0, sizeof (*snd_opts));
+
+  tcp_reuse_buffer (vm, b);
+
+  /* Set random initial sequence */
+  time_now = tcp_time_now ();
+
+  tc->iss = random_u32 (&time_now);
+  tc->snd_una = tc->iss;
+  tc->snd_nxt = tc->iss + 1;
+  tc->snd_una_max = tc->snd_nxt;
+
+  initial_wnd = tcp_initial_window_to_advertise (tc);
+
+  /* Make and write options */
+  tcp_opts_len = tcp_make_synack_options (tc, snd_opts);
+  tcp_hdr_opts_len = tcp_opts_len + sizeof (tcp_header_t);
+
+  th = vlib_buffer_push_tcp (b, tc->c_lcl_port, tc->c_rmt_port, tc->iss,
+			     tc->rcv_nxt, tcp_hdr_opts_len,
+			     TCP_FLAG_SYN | TCP_FLAG_ACK, initial_wnd);
+
+  tcp_options_write ((u8 *) (th + 1), snd_opts);
+
+  vnet_buffer (b)->tcp.connection_index = tc->c_c_index;
+  vnet_buffer (b)->tcp.flags = TCP_BUF_FLAG_ACK;
+
+  /* Init retransmit timer */
+  tcp_retransmit_timer_set (tm, tc);
+}
+
+always_inline void
+tcp_enqueue_to_ip_lookup (vlib_main_t * vm, vlib_buffer_t * b, u32 bi,
+			  u8 is_ip4)
+{
+  u32 *to_next, next_index;
+  vlib_frame_t *f;
+
+  b->flags |= VNET_BUFFER_LOCALLY_ORIGINATED;
+  b->error = 0;
+
+  /* Default FIB for now */
+  vnet_buffer (b)->sw_if_index[VLIB_TX] = 0;
+
+  /* Send to IP lookup */
+  next_index = is_ip4 ? ip4_lookup_node.index : ip6_lookup_node.index;
+  f = vlib_get_frame_to_node (vm, next_index);
+
+  /* Enqueue the packet */
+  to_next = vlib_frame_vector_args (f);
+  to_next[0] = bi;
+  f->n_vectors = 1;
+  vlib_put_frame_to_node (vm, next_index, f);
+}
+
+int
+tcp_make_reset_in_place (vlib_main_t * vm, vlib_buffer_t * b0,
+			 tcp_state_t state, u32 my_thread_index, u8 is_ip4)
+{
+  u8 tcp_hdr_len = sizeof (tcp_header_t);
+  ip4_header_t *ih4;
+  ip6_header_t *ih6;
+  tcp_header_t *th0;
+  ip4_address_t src_ip40;
+  ip6_address_t src_ip60;
+  u16 src_port0;
+  u32 tmp;
+
+  /* Find IP and TCP headers */
+  if (is_ip4)
+    {
+      ih4 = vlib_buffer_get_current (b0);
+      th0 = ip4_next_header (ih4);
+    }
+  else
+    {
+      ih6 = vlib_buffer_get_current (b0);
+      th0 = ip6_next_header (ih6);
+    }
+
+  /* Swap src and dst ip */
+  if (is_ip4)
+    {
+      ASSERT ((ih4->ip_version_and_header_length & 0xF0) == 0x40);
+      src_ip40.as_u32 = ih4->src_address.as_u32;
+      ih4->src_address.as_u32 = ih4->dst_address.as_u32;
+      ih4->dst_address.as_u32 = src_ip40.as_u32;
+
+      /* Chop the end of the pkt */
+      b0->current_length += ip4_header_bytes (ih4) + tcp_hdr_len;
+    }
+  else
+    {
+      ASSERT ((ih6->ip_version_traffic_class_and_flow_label & 0xF0) == 0x60);
+      clib_memcpy (&src_ip60, &ih6->src_address, sizeof (ip6_address_t));
+      clib_memcpy (&ih6->src_address, &ih6->dst_address,
+		   sizeof (ip6_address_t));
+      clib_memcpy (&ih6->dst_address, &src_ip60, sizeof (ip6_address_t));
+
+      /* Chop the end of the pkt */
+      b0->current_length += sizeof (ip6_header_t) + tcp_hdr_len;
+    }
+
+  /* Try to determine what/why we're actually resetting and swap
+   * src and dst ports */
+  if (state == TCP_STATE_CLOSED)
+    {
+      if (!tcp_syn (th0))
+	return -1;
+
+      tmp = clib_net_to_host_u32 (th0->seq_number);
+
+      /* Got a SYN for no listener. */
+      th0->flags = TCP_FLAG_RST | TCP_FLAG_ACK;
+      th0->ack_number = clib_host_to_net_u32 (tmp + 1);
+      th0->seq_number = 0;
+
+    }
+  else if (state >= TCP_STATE_SYN_SENT)
+    {
+      th0->flags = TCP_FLAG_RST | TCP_FLAG_ACK;
+      th0->seq_number = th0->ack_number;
+      th0->ack_number = 0;
+    }
+
+  src_port0 = th0->src_port;
+  th0->src_port = th0->dst_port;
+  th0->dst_port = src_port0;
+  th0->window = 0;
+  th0->data_offset_and_reserved = (tcp_hdr_len >> 2) << 4;
+  th0->urgent_pointer = 0;
+
+  /* Compute checksum */
+  if (is_ip4)
+    {
+      th0->checksum = ip4_tcp_udp_compute_checksum (vm, b0, ih4);
+    }
+  else
+    {
+      int bogus = ~0;
+      th0->checksum = ip6_tcp_udp_icmp_compute_checksum (vm, b0, ih6, &bogus);
+      ASSERT (!bogus);
+    }
+
+  return 0;
+}
+
+/**
+ *  Send reset without reusing existing buffer
+ */
+void
+tcp_send_reset (vlib_buffer_t * pkt, u8 is_ip4)
+{
+  vlib_buffer_t *b;
+  u32 bi;
+  tcp_main_t *tm = vnet_get_tcp_main ();
+  vlib_main_t *vm = tm->vlib_main;
+  u8 tcp_hdr_len, flags = 0;
+  tcp_header_t *th, *pkt_th;
+  u32 seq, ack;
+  ip4_header_t *ih4, *pkt_ih4;
+  ip6_header_t *ih6, *pkt_ih6;
+
+  tcp_get_free_buffer_index (tm, &bi);
+  b = vlib_get_buffer (vm, bi);
+
+  /* Leave enough space for headers */
+  vlib_buffer_make_headroom (b, MAX_HDRS_LEN);
+
+  /* Make and write options */
+  tcp_hdr_len = sizeof (tcp_header_t);
+
+  if (is_ip4)
+    {
+      pkt_ih4 = vlib_buffer_get_current (pkt);
+      pkt_th = ip4_next_header (pkt_ih4);
+    }
+  else
+    {
+      pkt_ih6 = vlib_buffer_get_current (pkt);
+      pkt_th = ip6_next_header (pkt_ih6);
+    }
+
+  if (tcp_ack (pkt_th))
+    {
+      flags = TCP_FLAG_RST;
+      seq = pkt_th->ack_number;
+      ack = 0;
+    }
+  else
+    {
+      flags = TCP_FLAG_RST | TCP_FLAG_ACK;
+      seq = 0;
+      ack = clib_host_to_net_u32 (vnet_buffer (pkt)->tcp.seq_end);
+    }
+
+  th = vlib_buffer_push_tcp_net_order (b, pkt_th->dst_port, pkt_th->src_port,
+				       seq, ack, tcp_hdr_len, flags, 0);
+
+  /* Swap src and dst ip */
+  if (is_ip4)
+    {
+      ASSERT ((pkt_ih4->ip_version_and_header_length & 0xF0) == 0x40);
+      ih4 = vlib_buffer_push_ip4 (vm, b, &pkt_ih4->dst_address,
+				  &pkt_ih4->src_address, IP_PROTOCOL_TCP);
+      th->checksum = ip4_tcp_udp_compute_checksum (vm, b, ih4);
+    }
+  else
+    {
+      int bogus = ~0;
+      pkt_ih6 = (ip6_header_t *) (pkt_th - 1);
+      ASSERT ((pkt_ih6->ip_version_traffic_class_and_flow_label & 0xF0) ==
+	      0x60);
+      ih6 =
+	vlib_buffer_push_ip6 (vm, b, &pkt_ih6->dst_address,
+			      &pkt_ih6->src_address, IP_PROTOCOL_TCP);
+      th->checksum = ip6_tcp_udp_icmp_compute_checksum (vm, b, ih6, &bogus);
+      ASSERT (!bogus);
+    }
+
+  tcp_enqueue_to_ip_lookup (vm, b, bi, is_ip4);
+}
+
+void
+tcp_push_ip_hdr (tcp_main_t * tm, tcp_connection_t * tc, vlib_buffer_t * b)
+{
+  tcp_header_t *th = vlib_buffer_get_current (b);
+
+  if (tc->c_is_ip4)
+    {
+      ip4_header_t *ih;
+      ih = vlib_buffer_push_ip4 (tm->vlib_main, b, &tc->c_lcl_ip4,
+				 &tc->c_rmt_ip4, IP_PROTOCOL_TCP);
+      th->checksum = ip4_tcp_udp_compute_checksum (tm->vlib_main, b, ih);
+    }
+  else
+    {
+      ip6_header_t *ih;
+      int bogus = ~0;
+
+      ih = vlib_buffer_push_ip6 (tm->vlib_main, b, &tc->c_lcl_ip6,
+				 &tc->c_rmt_ip6, IP_PROTOCOL_TCP);
+      th->checksum = ip6_tcp_udp_icmp_compute_checksum (tm->vlib_main, b, ih,
+							&bogus);
+      ASSERT (!bogus);
+    }
+}
+
+/**
+ *  Send SYN
+ *
+ *  Builds a SYN packet for a half-open connection and sends it to ipx_lookup.
+ *  The packet is not forwarded through tcpx_output to avoid doing lookups
+ *  in the half_open pool.
+ */
+void
+tcp_send_syn (tcp_connection_t * tc)
+{
+  vlib_buffer_t *b;
+  u32 bi;
+  tcp_main_t *tm = vnet_get_tcp_main ();
+  vlib_main_t *vm = tm->vlib_main;
+  u8 tcp_hdr_opts_len, tcp_opts_len;
+  tcp_header_t *th;
+  u32 time_now;
+  u16 initial_wnd;
+  tcp_options_t snd_opts;
+
+  tcp_get_free_buffer_index (tm, &bi);
+  b = vlib_get_buffer (vm, bi);
+
+  /* Leave enough space for headers */
+  vlib_buffer_make_headroom (b, MAX_HDRS_LEN);
+
+  /* Set random initial sequence */
+  time_now = tcp_time_now ();
+
+  tc->iss = random_u32 (&time_now);
+  tc->snd_una = tc->iss;
+  tc->snd_una_max = tc->snd_nxt = tc->iss + 1;
+
+  initial_wnd = tcp_initial_window_to_advertise (tc);
+
+  /* Make and write options */
+  memset (&snd_opts, 0, sizeof (snd_opts));
+  tcp_opts_len = tcp_make_syn_options (&snd_opts, initial_wnd);
+  tcp_hdr_opts_len = tcp_opts_len + sizeof (tcp_header_t);
+
+  th = vlib_buffer_push_tcp (b, tc->c_lcl_port, tc->c_rmt_port, tc->iss,
+			     tc->rcv_nxt, tcp_hdr_opts_len, TCP_FLAG_SYN,
+			     initial_wnd);
+
+  tcp_options_write ((u8 *) (th + 1), &snd_opts);
+
+  /* Measure RTT with this */
+  tc->rtt_ts = tcp_time_now ();
+  tc->rtt_seq = tc->snd_nxt;
+
+  /* Start retransmit trimer  */
+  tcp_timer_set (tc, TCP_TIMER_RETRANSMIT_SYN, tc->rto * TCP_TO_TIMER_TICK);
+  tc->rto_boff = 0;
+
+  /* Set the connection establishment timer */
+  tcp_timer_set (tc, TCP_TIMER_ESTABLISH, TCP_ESTABLISH_TIME);
+
+  tcp_push_ip_hdr (tm, tc, b);
+  tcp_enqueue_to_ip_lookup (vm, b, bi, tc->c_is_ip4);
+}
+
+always_inline void
+tcp_enqueue_to_output (vlib_main_t * vm, vlib_buffer_t * b, u32 bi, u8 is_ip4)
+{
+  u32 *to_next, next_index;
+  vlib_frame_t *f;
+
+  b->flags |= VNET_BUFFER_LOCALLY_ORIGINATED;
+  b->error = 0;
+
+  /* Decide where to send the packet */
+  next_index = is_ip4 ? tcp4_output_node.index : tcp6_output_node.index;
+  f = vlib_get_frame_to_node (vm, next_index);
+
+  /* Enqueue the packet */
+  to_next = vlib_frame_vector_args (f);
+  to_next[0] = bi;
+  f->n_vectors = 1;
+  vlib_put_frame_to_node (vm, next_index, f);
+}
+
+/**
+ *  Send FIN
+ */
+void
+tcp_send_fin (tcp_connection_t * tc)
+{
+  vlib_buffer_t *b;
+  u32 bi;
+  tcp_main_t *tm = vnet_get_tcp_main ();
+  vlib_main_t *vm = tm->vlib_main;
+
+  tcp_get_free_buffer_index (tm, &bi);
+  b = vlib_get_buffer (vm, bi);
+
+  /* Leave enough space for headers */
+  vlib_buffer_make_headroom (b, MAX_HDRS_LEN);
+
+  tcp_make_finack (tc, b);
+
+  tcp_enqueue_to_output (vm, b, bi, tc->c_is_ip4);
+}
+
+always_inline u8
+tcp_make_state_flags (tcp_state_t next_state)
+{
+  switch (next_state)
+    {
+    case TCP_STATE_ESTABLISHED:
+      return TCP_FLAG_ACK;
+    case TCP_STATE_SYN_RCVD:
+      return TCP_FLAG_SYN | TCP_FLAG_ACK;
+    case TCP_STATE_SYN_SENT:
+      return TCP_FLAG_SYN;
+    case TCP_STATE_LAST_ACK:
+    case TCP_STATE_FIN_WAIT_1:
+      return TCP_FLAG_FIN;
+    default:
+      clib_warning ("Shouldn't be here!");
+    }
+  return 0;
+}
+
+/**
+ * Push TCP header and update connection variables
+ */
+static void
+tcp_push_hdr_i (tcp_connection_t * tc, vlib_buffer_t * b,
+		tcp_state_t next_state)
+{
+  u32 advertise_wnd, data_len;
+  u8 tcp_opts_len, tcp_hdr_opts_len, opts_write_len, flags;
+  tcp_options_t _snd_opts, *snd_opts = &_snd_opts;
+  tcp_header_t *th;
+
+  data_len = b->current_length;
+  vnet_buffer (b)->tcp.flags = 0;
+
+  /* Make and write options */
+  memset (snd_opts, 0, sizeof (*snd_opts));
+  tcp_opts_len = tcp_make_options (tc, snd_opts, next_state);
+  tcp_hdr_opts_len = tcp_opts_len + sizeof (tcp_header_t);
+
+  /* Get rcv window to advertise */
+  advertise_wnd = tcp_window_to_advertise (tc, next_state);
+  flags = tcp_make_state_flags (next_state);
+
+  /* Push header and options */
+  th = vlib_buffer_push_tcp (b, tc->c_lcl_port, tc->c_rmt_port, tc->snd_nxt,
+			     tc->rcv_nxt, tcp_hdr_opts_len, flags,
+			     advertise_wnd);
+
+  opts_write_len = tcp_options_write ((u8 *) (th + 1), snd_opts);
+
+  ASSERT (opts_write_len == tcp_opts_len);
+
+  /* Tag the buffer with the connection index  */
+  vnet_buffer (b)->tcp.connection_index = tc->c_c_index;
+
+  tc->snd_nxt += data_len;
+}
+
+/* Send delayed ACK when timer expires */
+void
+tcp_timer_delack_handler (u32 index)
+{
+  tcp_main_t *tm = vnet_get_tcp_main ();
+  vlib_main_t *vm = tm->vlib_main;
+  u32 thread_index = os_get_cpu_number ();
+  tcp_connection_t *tc;
+  vlib_buffer_t *b;
+  u32 bi;
+
+  tc = tcp_connection_get (index, thread_index);
+
+  /* Get buffer */
+  tcp_get_free_buffer_index (tm, &bi);
+  b = vlib_get_buffer (vm, bi);
+
+  /* Fill in the ACK */
+  tcp_make_ack (tc, b);
+
+  tc->timers[TCP_TIMER_DELACK] = TCP_TIMER_HANDLE_INVALID;
+  tc->flags &= ~TCP_CONN_DELACK;
+
+  tcp_enqueue_to_output (vm, b, bi, tc->c_is_ip4);
+}
+
+/** Build a retransmit segment
+ *
+ * @return the number of bytes in the segment or 0 if there's nothing to
+ *         retransmit
+ * */
+u32
+tcp_prepare_retransmit_segment (tcp_connection_t * tc, vlib_buffer_t * b,
+				u32 max_bytes)
+{
+  tcp_main_t *tm = vnet_get_tcp_main ();
+  vlib_main_t *vm = tm->vlib_main;
+  u32 n_bytes, offset = 0;
+  sack_scoreboard_hole_t *hole;
+  u32 hole_size;
+
+  tcp_reuse_buffer (vm, b);
+
+  ASSERT (tc->state == TCP_STATE_ESTABLISHED);
+  ASSERT (max_bytes != 0);
+
+  if (tcp_opts_sack_permitted (&tc->opt))
+    {
+      /* XXX get first hole not retransmitted yet  */
+      hole = scoreboard_first_hole (&tc->sack_sb);
+      if (!hole)
+	return 0;
+
+      offset = hole->start - tc->snd_una;
+      hole_size = hole->end - hole->start;
+
+      ASSERT (hole_size);
+
+      if (hole_size < max_bytes)
+	max_bytes = hole_size;
+    }
+  else
+    {
+      if (seq_geq (tc->snd_nxt, tc->snd_una_max))
+	return 0;
+    }
+
+  n_bytes = stream_session_peek_bytes (&tc->connection,
+				       vlib_buffer_get_current (b), offset,
+				       max_bytes);
+  ASSERT (n_bytes != 0);
+
+  tc->snd_nxt += n_bytes;
+  tcp_push_hdr_i (tc, b, tc->state);
+
+  return n_bytes;
+}
+
+static void
+tcp_timer_retransmit_handler_i (u32 index, u8 is_syn)
+{
+  tcp_main_t *tm = vnet_get_tcp_main ();
+  vlib_main_t *vm = tm->vlib_main;
+  u32 thread_index = os_get_cpu_number ();
+  tcp_connection_t *tc;
+  vlib_buffer_t *b;
+  u32 bi, max_bytes, snd_space;
+
+  if (is_syn)
+    {
+      tc = tcp_half_open_connection_get (index);
+    }
+  else
+    {
+      tc = tcp_connection_get (index, thread_index);
+    }
+
+  /* Make sure timer handle is set to invalid */
+  tc->timers[TCP_TIMER_RETRANSMIT] = TCP_TIMER_HANDLE_INVALID;
+
+  /* Increment RTO backoff (also equal to number of retries) */
+  tc->rto_boff += 1;
+
+  /* Go back to first un-acked byte */
+  tc->snd_nxt = tc->snd_una;
+
+  /* Get buffer */
+  tcp_get_free_buffer_index (tm, &bi);
+  b = vlib_get_buffer (vm, bi);
+
+  if (tc->state == TCP_STATE_ESTABLISHED)
+    {
+      tcp_fastrecovery_off (tc);
+
+      /* Exponential backoff */
+      tc->rto = clib_min (tc->rto << 1, TCP_RTO_MAX);
+
+      /* Figure out what and how many bytes we can send */
+      snd_space = tcp_available_snd_space (tc);
+      max_bytes = clib_min (tc->snd_mss, snd_space);
+      tcp_prepare_retransmit_segment (tc, b, max_bytes);
+
+      tc->rtx_bytes += max_bytes;
+
+      /* No fancy recovery for now! */
+      scoreboard_clear (&tc->sack_sb);
+    }
+  else
+    {
+      /* Retransmit for SYN/SYNACK */
+      ASSERT (tc->state == TCP_STATE_SYN_RCVD
+	      || tc->state == TCP_STATE_SYN_SENT);
+
+      /* Try without increasing RTO a number of times. If this fails,
+       * start growing RTO exponentially */
+      if (tc->rto_boff > TCP_RTO_SYN_RETRIES)
+	tc->rto = clib_min (tc->rto << 1, TCP_RTO_MAX);
+
+      vlib_buffer_make_headroom (b, MAX_HDRS_LEN);
+      tcp_push_hdr_i (tc, b, tc->state);
+    }
+
+  if (!is_syn)
+    {
+      tcp_enqueue_to_output (vm, b, bi, tc->c_is_ip4);
+
+      /* Re-enable retransmit timer */
+      tcp_retransmit_timer_set (tm, tc);
+    }
+  else
+    {
+      ASSERT (tc->state == TCP_STATE_SYN_SENT);
+
+      /* This goes straight to ipx_lookup */
+      tcp_push_ip_hdr (tm, tc, b);
+      tcp_enqueue_to_ip_lookup (vm, b, bi, tc->c_is_ip4);
+
+      /* Re-enable retransmit timer */
+      tcp_timer_set (tc, TCP_TIMER_RETRANSMIT_SYN,
+		     tc->rto * TCP_TO_TIMER_TICK);
+    }
+}
+
+void
+tcp_timer_retransmit_handler (u32 index)
+{
+  tcp_timer_retransmit_handler_i (index, 0);
+}
+
+void
+tcp_timer_retransmit_syn_handler (u32 index)
+{
+  tcp_timer_retransmit_handler_i (index, 1);
+}
+
+/**
+ * Retansmit first unacked segment */
+void
+tcp_retransmit_first_unacked (tcp_connection_t * tc)
+{
+  tcp_main_t *tm = vnet_get_tcp_main ();
+  u32 snd_nxt = tc->snd_nxt;
+  vlib_buffer_t *b;
+  u32 bi;
+
+  tc->snd_nxt = tc->snd_una;
+
+  /* Get buffer */
+  tcp_get_free_buffer_index (tm, &bi);
+  b = vlib_get_buffer (tm->vlib_main, bi);
+
+  tcp_prepare_retransmit_segment (tc, b, tc->snd_mss);
+  tcp_enqueue_to_output (tm->vlib_main, b, bi, tc->c_is_ip4);
+
+  tc->snd_nxt = snd_nxt;
+  tc->rtx_bytes += tc->snd_mss;
+}
+
+void
+tcp_fast_retransmit (tcp_connection_t * tc)
+{
+  tcp_main_t *tm = vnet_get_tcp_main ();
+  u32 snd_space, max_bytes, n_bytes, bi;
+  vlib_buffer_t *b;
+
+  ASSERT (tcp_in_fastrecovery (tc));
+
+  clib_warning ("fast retransmit!");
+
+  /* Start resending from first un-acked segment */
+  tc->snd_nxt = tc->snd_una;
+
+  snd_space = tcp_available_snd_space (tc);
+
+  while (snd_space)
+    {
+      tcp_get_free_buffer_index (tm, &bi);
+      b = vlib_get_buffer (tm->vlib_main, bi);
+
+      max_bytes = clib_min (tc->snd_mss, snd_space);
+      n_bytes = tcp_prepare_retransmit_segment (tc, b, max_bytes);
+
+      /* Nothing left to retransmit */
+      if (n_bytes == 0)
+	return;
+
+      tcp_enqueue_to_output (tm->vlib_main, b, bi, tc->c_is_ip4);
+
+      snd_space -= n_bytes;
+    }
+
+  /* If window allows, send new data */
+  tc->snd_nxt = tc->snd_una_max;
+}
+
+always_inline u32
+tcp_session_has_ooo_data (tcp_connection_t * tc)
+{
+  stream_session_t *s =
+    stream_session_get (tc->c_s_index, tc->c_thread_index);
+  return svm_fifo_has_ooo_data (s->server_rx_fifo);
+}
+
+always_inline uword
+tcp46_output_inline (vlib_main_t * vm,
+		     vlib_node_runtime_t * node,
+		     vlib_frame_t * from_frame, int is_ip4)
+{
+  tcp_main_t *tm = vnet_get_tcp_main ();
+  u32 n_left_from, next_index, *from, *to_next;
+  u32 my_thread_index = vm->cpu_index;
+
+  from = vlib_frame_vector_args (from_frame);
+  n_left_from = from_frame->n_vectors;
+
+  next_index = node->cached_next_index;
+
+  while (n_left_from > 0)
+    {
+      u32 n_left_to_next;
+
+      vlib_get_next_frame (vm, node, next_index, to_next, n_left_to_next);
+
+      while (n_left_from > 0 && n_left_to_next > 0)
+	{
+	  u32 bi0;
+	  vlib_buffer_t *b0;
+	  tcp_connection_t *tc0;
+	  tcp_header_t *th0;
+	  u32 error0 = TCP_ERROR_PKTS_SENT, next0 = TCP_OUTPUT_NEXT_IP_LOOKUP;
+
+	  bi0 = from[0];
+	  to_next[0] = bi0;
+	  from += 1;
+	  to_next += 1;
+	  n_left_from -= 1;
+	  n_left_to_next -= 1;
+
+	  b0 = vlib_get_buffer (vm, bi0);
+	  tc0 = tcp_connection_get (vnet_buffer (b0)->tcp.connection_index,
+				    my_thread_index);
+	  th0 = vlib_buffer_get_current (b0);
+
+	  if (is_ip4)
+	    {
+	      ip4_header_t *ih0;
+	      ih0 = vlib_buffer_push_ip4 (vm, b0, &tc0->c_lcl_ip4,
+					  &tc0->c_rmt_ip4, IP_PROTOCOL_TCP);
+	      th0->checksum = ip4_tcp_udp_compute_checksum (vm, b0, ih0);
+	    }
+	  else
+	    {
+	      ip6_header_t *ih0;
+	      int bogus = ~0;
+
+	      ih0 = vlib_buffer_push_ip6 (vm, b0, &tc0->c_lcl_ip6,
+					  &tc0->c_rmt_ip6, IP_PROTOCOL_TCP);
+	      th0->checksum = ip6_tcp_udp_icmp_compute_checksum (vm, b0, ih0,
+								 &bogus);
+	      ASSERT (!bogus);
+	    }
+
+	  /* Filter out DUPACKs if there are no OOO segments left */
+	  if (PREDICT_FALSE
+	      (vnet_buffer (b0)->tcp.flags & TCP_BUF_FLAG_DUPACK))
+	    {
+	      tc0->snt_dupacks--;
+	      ASSERT (tc0->snt_dupacks >= 0);
+	      if (!tcp_session_has_ooo_data (tc0))
+		{
+		  error0 = TCP_ERROR_FILTERED_DUPACKS;
+		  next0 = TCP_OUTPUT_NEXT_DROP;
+		  goto done;
+		}
+	    }
+
+	  /* Retransmitted SYNs do reach this but it should be harmless */
+	  tc0->rcv_las = tc0->rcv_nxt;
+
+	  /* Stop DELACK timer and fix flags */
+	  tc0->flags &=
+	    ~(TCP_CONN_SNDACK | TCP_CONN_DELACK | TCP_CONN_BURSTACK);
+	  if (tcp_timer_is_active (tc0, TCP_TIMER_DELACK))
+	    {
+	      tcp_timer_reset (tc0, TCP_TIMER_DELACK);
+	    }
+
+	  /* If not retransmitting
+	   * 1) update snd_una_max (SYN, SYNACK, new data, FIN)
+	   * 2) If we're not tracking an ACK, start tracking */
+	  if (seq_lt (tc0->snd_una_max, tc0->snd_nxt))
+	    {
+	      tc0->snd_una_max = tc0->snd_nxt;
+	      if (tc0->rtt_ts == 0)
+		{
+		  tc0->rtt_ts = tcp_time_now ();
+		  tc0->rtt_seq = tc0->snd_nxt;
+		}
+	    }
+
+	  /* Set the retransmit timer if not set already and not
+	   * doing a pure ACK */
+	  if (!tcp_timer_is_active (tc0, TCP_TIMER_RETRANSMIT)
+	      && tc0->snd_nxt != tc0->snd_una)
+	    {
+	      tcp_retransmit_timer_set (tm, tc0);
+	      tc0->rto_boff = 0;
+	    }
+
+	  /* set fib index to default and lookup node */
+	  /* XXX network virtualization (vrf/vni) */
+	  vnet_buffer (b0)->sw_if_index[VLIB_RX] = 0;
+	  vnet_buffer (b0)->sw_if_index[VLIB_TX] = (u32) ~ 0;
+
+	  b0->flags |= VNET_BUFFER_LOCALLY_ORIGINATED;
+
+	done:
+	  b0->error = error0 != 0 ? node->errors[error0] : 0;
+	  if (PREDICT_FALSE (b0->flags & VLIB_BUFFER_IS_TRACED))
+	    {
+
+	    }
+
+	  vlib_validate_buffer_enqueue_x1 (vm, node, next_index, to_next,
+					   n_left_to_next, bi0, next0);
+	}
+
+      vlib_put_next_frame (vm, node, next_index, n_left_to_next);
+    }
+
+  return from_frame->n_vectors;
+}
+
+static uword
+tcp4_output (vlib_main_t * vm, vlib_node_runtime_t * node,
+	     vlib_frame_t * from_frame)
+{
+  return tcp46_output_inline (vm, node, from_frame, 1 /* is_ip4 */ );
+}
+
+static uword
+tcp6_output (vlib_main_t * vm, vlib_node_runtime_t * node,
+	     vlib_frame_t * from_frame)
+{
+  return tcp46_output_inline (vm, node, from_frame, 0 /* is_ip4 */ );
+}
+
+VLIB_REGISTER_NODE (tcp4_output_node) =
+{
+  .function = tcp4_output,.name = "tcp4-output",
+    /* Takes a vector of packets. */
+    .vector_size = sizeof (u32),.n_errors = TCP_N_ERROR,.error_strings =
+    tcp_error_strings,.n_next_nodes = TCP_OUTPUT_N_NEXT,.next_nodes =
+  {
+#define _(s,n) [TCP_OUTPUT_NEXT_##s] = n,
+    foreach_tcp4_output_next
+#undef _
+  }
+,.format_buffer = format_tcp_header,.format_trace = format_tcp_tx_trace,};
+
+VLIB_NODE_FUNCTION_MULTIARCH (tcp4_output_node, tcp4_output)
+VLIB_REGISTER_NODE (tcp6_output_node) =
+{
+  .function = tcp6_output,.name = "tcp6-output",
+    /* Takes a vector of packets. */
+    .vector_size = sizeof (u32),.n_errors = TCP_N_ERROR,.error_strings =
+    tcp_error_strings,.n_next_nodes = TCP_OUTPUT_N_NEXT,.next_nodes =
+  {
+#define _(s,n) [TCP_OUTPUT_NEXT_##s] = n,
+    foreach_tcp6_output_next
+#undef _
+  }
+,.format_buffer = format_tcp_header,.format_trace = format_tcp_tx_trace,};
+
+VLIB_NODE_FUNCTION_MULTIARCH (tcp6_output_node, tcp6_output) u32
+tcp_push_header (transport_connection_t * tconn, vlib_buffer_t * b)
+{
+  tcp_connection_t *tc;
+
+  tc = (tcp_connection_t *) tconn;
+  tcp_push_hdr_i (tc, b, TCP_STATE_ESTABLISHED);
+  return 0;
+}
+
+typedef enum _tcp_reset_next
+{
+  TCP_RESET_NEXT_DROP,
+  TCP_RESET_NEXT_IP_LOOKUP,
+  TCP_RESET_N_NEXT
+} tcp_reset_next_t;
+
+#define foreach_tcp4_reset_next        	\
+  _(DROP, "error-drop")                 \
+  _(IP_LOOKUP, "ip4-lookup")
+
+#define foreach_tcp6_reset_next        	\
+  _(DROP, "error-drop")                 \
+  _(IP_LOOKUP, "ip6-lookup")
+
+static uword
+tcp46_send_reset_inline (vlib_main_t * vm, vlib_node_runtime_t * node,
+			 vlib_frame_t * from_frame, u8 is_ip4)
+{
+  u32 n_left_from, next_index, *from, *to_next;
+  u32 my_thread_index = vm->cpu_index;
+
+  from = vlib_frame_vector_args (from_frame);
+  n_left_from = from_frame->n_vectors;
+
+  next_index = node->cached_next_index;
+
+  while (n_left_from > 0)
+    {
+      u32 n_left_to_next;
+
+      vlib_get_next_frame (vm, node, next_index, to_next, n_left_to_next);
+
+      while (n_left_from > 0 && n_left_to_next > 0)
+	{
+	  u32 bi0;
+	  vlib_buffer_t *b0;
+	  u32 error0 = TCP_ERROR_RST_SENT, next0 = TCP_RESET_NEXT_IP_LOOKUP;
+
+	  bi0 = from[0];
+	  to_next[0] = bi0;
+	  from += 1;
+	  to_next += 1;
+	  n_left_from -= 1;
+	  n_left_to_next -= 1;
+
+	  b0 = vlib_get_buffer (vm, bi0);
+
+	  if (tcp_make_reset_in_place (vm, b0, vnet_buffer (b0)->tcp.flags,
+				       my_thread_index, is_ip4))
+	    {
+	      error0 = TCP_ERROR_LOOKUP_DROPS;
+	      next0 = TCP_RESET_NEXT_DROP;
+	      goto done;
+	    }
+
+	  /* Prepare to send to IP lookup */
+	  vnet_buffer (b0)->sw_if_index[VLIB_TX] = 0;
+	  next0 = TCP_RESET_NEXT_IP_LOOKUP;
+
+	done:
+	  b0->error = error0 != 0 ? node->errors[error0] : 0;
+	  b0->flags |= VNET_BUFFER_LOCALLY_ORIGINATED;
+	  if (PREDICT_FALSE (b0->flags & VLIB_BUFFER_IS_TRACED))
+	    {
+
+	    }
+
+	  vlib_validate_buffer_enqueue_x1 (vm, node, next_index, to_next,
+					   n_left_to_next, bi0, next0);
+	}
+      vlib_put_next_frame (vm, node, next_index, n_left_to_next);
+    }
+  return from_frame->n_vectors;
+}
+
+static uword
+tcp4_send_reset (vlib_main_t * vm, vlib_node_runtime_t * node,
+		 vlib_frame_t * from_frame)
+{
+  return tcp46_send_reset_inline (vm, node, from_frame, 1);
+}
+
+static uword
+tcp6_send_reset (vlib_main_t * vm, vlib_node_runtime_t * node,
+		 vlib_frame_t * from_frame)
+{
+  return tcp46_send_reset_inline (vm, node, from_frame, 0);
+}
+
+/* *INDENT-OFF* */
+VLIB_REGISTER_NODE (tcp4_reset_node) = {
+  .function = tcp4_send_reset,
+  .name = "tcp4-reset",
+  .vector_size = sizeof (u32),
+  .n_errors = TCP_N_ERROR,
+  .error_strings = tcp_error_strings,
+  .n_next_nodes = TCP_RESET_N_NEXT,
+  .next_nodes = {
+#define _(s,n) [TCP_RESET_NEXT_##s] = n,
+    foreach_tcp4_reset_next
+#undef _
+  },
+};
+/* *INDENT-ON* */
+
+/* *INDENT-OFF* */
+VLIB_REGISTER_NODE (tcp6_reset_node) = {
+  .function = tcp6_send_reset,
+  .name = "tcp6-reset",
+  .vector_size = sizeof (u32),
+  .n_errors = TCP_N_ERROR,
+  .error_strings = tcp_error_strings,
+  .n_next_nodes = TCP_RESET_N_NEXT,
+  .next_nodes = {
+#define _(s,n) [TCP_RESET_NEXT_##s] = n,
+    foreach_tcp6_reset_next
+#undef _
+  },
+};
+/* *INDENT-ON* */
+
+/*
+ * fd.io coding-style-patch-verification: ON
+ *
+ * Local Variables:
+ * eval: (c-set-style "gnu")
+ * End:
+ */
diff --git a/src/vnet/tcp/tcp_packet.h b/src/vnet/tcp/tcp_packet.h
new file mode 100644
index 00000000..866c5fd6
--- /dev/null
+++ b/src/vnet/tcp/tcp_packet.h
@@ -0,0 +1,184 @@
+/*
+ * Copyright (c) 2016 Cisco and/or its affiliates.
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at:
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#ifndef included_tcp_packet_h
+#define included_tcp_packet_h
+
+#include 
+
+/* TCP flags bit 0 first. */
+#define foreach_tcp_flag                                \
+  _ (FIN) /**< No more data from sender. */             \
+  _ (SYN) /**< Synchronize sequence numbers. */         \
+  _ (RST) /**< Reset the connection. */                 \
+  _ (PSH) /**< Push function. */                        \
+  _ (ACK) /**< Ack field significant. */                \
+  _ (URG) /**< Urgent pointer field significant. */     \
+  _ (ECE) /**< ECN-echo. Receiver got CE packet */      \
+  _ (CWR) /**< Sender reduced congestion window */
+
+enum
+{
+#define _(f) TCP_FLAG_BIT_##f,
+  foreach_tcp_flag
+#undef _
+    TCP_N_FLAG_BITS,
+};
+
+enum
+{
+#define _(f) TCP_FLAG_##f = 1 << TCP_FLAG_BIT_##f,
+  foreach_tcp_flag
+#undef _
+};
+
+typedef struct _tcp_header
+{
+  union
+  {
+    struct
+    {
+      u16 src_port; /**< Source port. */
+      u16 dst_port; /**< Destination port. */
+    };
+    struct
+    {
+      u16 src, dst;
+    };
+  };
+
+  u32 seq_number;	/**< Sequence number of the first data octet in this
+                         *   segment, except when SYN is present. If SYN
+                         *   is present the seq number is is the ISN and the
+                         *   first data octet is ISN+1 */
+  u32 ack_number;	/**< Acknowledgement number if ACK is set. It contains
+                         *   the value of the next sequence number the sender
+                         *   of the segment is expecting to receive. */
+  u8 data_offset_and_reserved;
+  u8 flags;		/**< Flags: see the macro above */
+  u16 window;		/**< Number of bytes sender is willing to receive. */
+
+  u16 checksum;		/**< Checksum of TCP pseudo header and data. */
+  u16 urgent_pointer;	/**< Seq number of the byte after the urgent data. */
+} __attribute__ ((packed)) tcp_header_t;
+
+/* Flag tests that return 0 or !0 */
+#define tcp_doff(_th) ((_th)->data_offset_and_reserved >> 4)
+#define tcp_fin(_th) ((_th)->flags & TCP_FLAG_FIN)
+#define tcp_syn(_th) ((_th)->flags & TCP_FLAG_SYN)
+#define tcp_rst(_th) ((_th)->flags & TCP_FLAG_RST)
+#define tcp_psh(_th) ((_th)->flags & TCP_FLAG_PSH)
+#define tcp_ack(_th) ((_th)->flags & TCP_FLAG_ACK)
+#define tcp_urg(_th) ((_th)->flags & TCP_FLAG_URG)
+#define tcp_ece(_th) ((_th)->flags & TCP_FLAG_ECE)
+#define tcp_cwr(_th) ((_th)->flags & TCP_FLAG_CWR)
+
+/* Flag tests that return 0 or 1 */
+#define tcp_is_syn(_th) !!((_th)->flags & TCP_FLAG_SYN)
+#define tcp_is_fin(_th) !!((_th)->flags & TCP_FLAG_FIN)
+
+always_inline int
+tcp_header_bytes (tcp_header_t * t)
+{
+  return tcp_doff (t) * sizeof (u32);
+}
+
+/*
+ * TCP options.
+ */
+
+typedef enum tcp_option_type
+{
+  TCP_OPTION_EOL = 0,			/**< End of options. */
+  TCP_OPTION_NOOP = 1,			/**< No operation. */
+  TCP_OPTION_MSS = 2,			/**< Limit MSS. */
+  TCP_OPTION_WINDOW_SCALE = 3,		/**< Window scale. */
+  TCP_OPTION_SACK_PERMITTED = 4,	/**< Selective Ack permitted. */
+  TCP_OPTION_SACK_BLOCK = 5,		/**< Selective Ack block. */
+  TCP_OPTION_TIMESTAMP = 8,		/**< Timestamps. */
+  TCP_OPTION_UTO = 28,			/**< User timeout. */
+  TCP_OPTION_AO = 29,			/**< Authentication Option. */
+} tcp_option_type_t;
+
+#define foreach_tcp_options_flag                                        \
+  _ (MSS)               /**< MSS advertised in SYN */                   \
+  _ (TSTAMP)            /**< Timestamp capability advertised in SYN */  \
+  _ (WSCALE)            /**< Wnd scale capability advertised in SYN */  \
+  _ (SACK_PERMITTED)    /**< SACK capability advertised in SYN */       \
+  _ (SACK)		/**< SACK present */
+
+enum
+{
+#define _(f) TCP_OPTS_FLAG_BIT_##f,
+  foreach_tcp_options_flag
+#undef _
+    TCP_OPTIONS_N_FLAG_BITS,
+};
+
+enum
+{
+#define _(f) TCP_OPTS_FLAG_##f = 1 << TCP_OPTS_FLAG_BIT_##f,
+  foreach_tcp_options_flag
+#undef _
+};
+
+typedef struct _sack_block
+{
+  u32 start;		/**< Start sequence number */
+  u32 end;		/**< End sequence number */
+} sack_block_t;
+
+typedef struct
+{
+  u8 flags;		/** Option flags, see above */
+
+  /* Received options */
+  u16 mss;		/**< Maximum segment size advertised by peer */
+  u8 wscale;		/**< Window scale advertised by peer */
+  u32 tsval;		/**< Peer's timestamp value */
+  u32 tsecr;		/**< Echoed/reflected time stamp */
+  sack_block_t *sacks;	/**< SACK blocks received */
+  u8 n_sack_blocks;	/**< Number of SACKs blocks */
+} tcp_options_t;
+
+/* Flag tests that return 0 or !0 */
+#define tcp_opts_mss(_to) ((_to)->flags & TCP_OPTS_FLAG_MSS)
+#define tcp_opts_tstamp(_to) ((_to)->flags & TCP_OPTS_FLAG_TSTAMP)
+#define tcp_opts_wscale(_to) ((_to)->flags & TCP_OPTS_FLAG_WSCALE)
+#define tcp_opts_sack(_to) ((_to)->flags & TCP_OPTS_FLAG_SACK)
+#define tcp_opts_sack_permitted(_to) ((_to)->flags & TCP_OPTS_FLAG_SACK_PERMITTED)
+
+/* TCP option lengths */
+#define TCP_OPTION_LEN_EOL              1
+#define TCP_OPTION_LEN_NOOP             1
+#define TCP_OPTION_LEN_MSS              4
+#define TCP_OPTION_LEN_WINDOW_SCALE     3
+#define TCP_OPTION_LEN_SACK_PERMITTED   2
+#define TCP_OPTION_LEN_TIMESTAMP        10
+#define TCP_OPTION_LEN_SACK_BLOCK        8
+
+#define TCP_WND_MAX                     65535U
+#define TCP_MAX_WND_SCALE               14	/* See RFC 1323 */
+#define TCP_OPTS_ALIGN                  4
+#define TCP_OPTS_MAX_SACK_BLOCKS        3
+#endif /* included_tcp_packet_h */
+
+/*
+ * fd.io coding-style-patch-verification: ON
+ *
+ * Local Variables:
+ * eval: (c-set-style "gnu")
+ * End:
+ */
diff --git a/src/vnet/tcp/tcp_pg.c b/src/vnet/tcp/tcp_pg.c
new file mode 100644
index 00000000..dc324049
--- /dev/null
+++ b/src/vnet/tcp/tcp_pg.c
@@ -0,0 +1,236 @@
+/*
+ * Copyright (c) 2016 Cisco and/or its affiliates.
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at:
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+/*
+ * ip/tcp_pg: TCP packet-generator interface
+ *
+ * Copyright (c) 2008 Eliot Dresselhaus
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining
+ * a copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sublicense, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be
+ * included in all copies or substantial portions of the Software.
+ *
+ *  THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ *  EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ *  MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ *  NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE
+ *  LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
+ *  OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
+ *  WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+#include 
+#include 
+
+/* TCP flags bit 0 first. */
+#define foreach_tcp_flag			\
+  _ (FIN)					\
+  _ (SYN)					\
+  _ (RST)					\
+  _ (PSH)					\
+  _ (ACK)					\
+  _ (URG)					\
+  _ (ECE)					\
+  _ (CWR)
+
+static void
+tcp_pg_edit_function (pg_main_t * pg,
+		      pg_stream_t * s,
+		      pg_edit_group_t * g,
+		      u32 * packets,
+		      u32 n_packets)
+{
+  vlib_main_t * vm = vlib_get_main();
+  u32 ip_offset, tcp_offset;
+
+  tcp_offset = g->start_byte_offset;
+  ip_offset = (g-1)->start_byte_offset;
+
+  while (n_packets >= 1)
+    {
+      vlib_buffer_t * p0;
+      ip4_header_t * ip0;
+      tcp_header_t * tcp0;
+      ip_csum_t sum0;
+      u32 tcp_len0;
+
+      p0 = vlib_get_buffer (vm, packets[0]);
+      n_packets -= 1;
+      packets += 1;
+
+      ASSERT (p0->current_data == 0);
+      ip0 = (void *) (p0->data + ip_offset);
+      tcp0 = (void *) (p0->data + tcp_offset);
+      tcp_len0 = clib_net_to_host_u16 (ip0->length) - sizeof (ip0[0]);
+
+      /* Initialize checksum with header. */
+      if (BITS (sum0) == 32)
+	{
+	  sum0 = clib_mem_unaligned (&ip0->src_address, u32);
+	  sum0 = ip_csum_with_carry (sum0, clib_mem_unaligned (&ip0->dst_address, u32));
+	}
+      else
+	sum0 = clib_mem_unaligned (&ip0->src_address, u64);
+
+      sum0 = ip_csum_with_carry
+	(sum0, clib_host_to_net_u32 (tcp_len0 + (ip0->protocol << 16)));
+
+      /* Invalidate possibly old checksum. */
+      tcp0->checksum = 0;
+
+      sum0 = ip_incremental_checksum_buffer (vm, p0, tcp_offset, tcp_len0, sum0);
+
+      tcp0->checksum = ~ ip_csum_fold (sum0);
+    }
+}
+
+typedef struct {
+  pg_edit_t src, dst;
+  pg_edit_t seq_number, ack_number;
+  pg_edit_t data_offset_and_reserved;
+#define _(f) pg_edit_t f##_flag;
+  foreach_tcp_flag
+#undef _
+  pg_edit_t window;
+  pg_edit_t checksum;
+  pg_edit_t urgent_pointer;
+} pg_tcp_header_t;
+
+static inline void
+pg_tcp_header_init (pg_tcp_header_t * p)
+{
+  /* Initialize fields that are not bit fields in the IP header. */
+#define _(f) pg_edit_init (&p->f, tcp_header_t, f);
+  _ (src);
+  _ (dst);
+  _ (seq_number);
+  _ (ack_number);
+  _ (window);
+  _ (checksum);
+  _ (urgent_pointer);
+#undef _
+
+  /* Initialize bit fields. */
+#define _(f)						\
+  pg_edit_init_bitfield (&p->f##_flag, tcp_header_t,	\
+			 flags,				\
+			 TCP_FLAG_BIT_##f, 1);
+
+  foreach_tcp_flag
+#undef _
+
+  pg_edit_init_bitfield (&p->data_offset_and_reserved, tcp_header_t,
+			 data_offset_and_reserved,
+			 4, 4);
+}
+
+uword
+unformat_pg_tcp_header (unformat_input_t * input, va_list * args)
+{
+  pg_stream_t * s = va_arg (*args, pg_stream_t *);
+  pg_tcp_header_t * p;
+  u32 group_index;
+  
+  p = pg_create_edit_group (s, sizeof (p[0]), sizeof (tcp_header_t),
+			    &group_index);
+  pg_tcp_header_init (p);
+
+  /* Defaults. */
+  pg_edit_set_fixed (&p->seq_number, 0);
+  pg_edit_set_fixed (&p->ack_number, 0);
+
+  pg_edit_set_fixed (&p->data_offset_and_reserved, 
+                     sizeof (tcp_header_t) / sizeof (u32));
+
+  pg_edit_set_fixed (&p->window, 4096);
+  pg_edit_set_fixed (&p->urgent_pointer, 0);
+
+#define _(f) pg_edit_set_fixed (&p->f##_flag, 0);
+  foreach_tcp_flag
+#undef _
+
+  p->checksum.type = PG_EDIT_UNSPECIFIED;
+
+  if (! unformat (input, "TCP: %U -> %U",
+		  unformat_pg_edit,
+		    unformat_tcp_udp_port, &p->src,
+		  unformat_pg_edit,
+		    unformat_tcp_udp_port, &p->dst))
+    goto error;
+
+  /* Parse options. */
+  while (1)
+    {
+      if (unformat (input, "window %U",
+		    unformat_pg_edit,
+		    unformat_pg_number, &p->window))
+	;
+
+      else if (unformat (input, "checksum %U",
+			 unformat_pg_edit,
+			 unformat_pg_number, &p->checksum))
+	;
+
+      /* Flags. */
+#define _(f) else if (unformat (input, #f)) pg_edit_set_fixed (&p->f##_flag, 1);
+  foreach_tcp_flag
+#undef _
+
+      /* Can't parse input: try next protocol level. */
+      else
+	break;
+    }
+
+  {
+    ip_main_t * im = &ip_main;
+    u16 dst_port;
+    tcp_udp_port_info_t * pi;
+
+    pi = 0;
+    if (p->dst.type == PG_EDIT_FIXED)
+      {
+	dst_port = pg_edit_get_value (&p->dst, PG_EDIT_LO);
+	pi = ip_get_tcp_udp_port_info (im, dst_port);
+      }
+
+    if (pi && pi->unformat_pg_edit
+	&& unformat_user (input, pi->unformat_pg_edit, s))
+      ;
+
+    else if (! unformat_user (input, unformat_pg_payload, s))
+      goto error;
+
+    if (p->checksum.type == PG_EDIT_UNSPECIFIED)
+      {
+	pg_edit_group_t * g = pg_stream_get_group (s, group_index);
+	g->edit_function = tcp_pg_edit_function;
+	g->edit_function_opaque = 0;
+      }
+
+    return 1;
+  }
+
+ error:
+  /* Free up any edits we may have added. */
+  pg_free_edit_group (s);
+  return 0;
+}
+
diff --git a/src/vnet/tcp/tcp_syn_filter4.c b/src/vnet/tcp/tcp_syn_filter4.c
new file mode 100644
index 00000000..c7605a30
--- /dev/null
+++ b/src/vnet/tcp/tcp_syn_filter4.c
@@ -0,0 +1,542 @@
+/*
+ * Copyright (c) 2016 Cisco and/or its affiliates.
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at:
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#include 
+#include 
+#include 
+#include 
+#include 
+#include 
+#include 
+
+typedef struct
+{
+  f64 next_reset;
+  f64 reset_interval;
+  u8 *syn_counts;
+} syn_filter4_runtime_t;
+
+typedef struct
+{
+  u32 next_index;
+  int not_a_syn;
+  u8 filter_value;
+} syn_filter4_trace_t;
+
+/* packet trace format function */
+static u8 *
+format_syn_filter4_trace (u8 * s, va_list * args)
+{
+  CLIB_UNUSED (vlib_main_t * vm) = va_arg (*args, vlib_main_t *);
+  CLIB_UNUSED (vlib_node_t * node) = va_arg (*args, vlib_node_t *);
+  syn_filter4_trace_t *t = va_arg (*args, syn_filter4_trace_t *);
+
+  s = format (s, "SYN_FILTER4: next index %d, %s",
+	      t->next_index, t->not_a_syn ? "not a syn" : "syn");
+  if (t->not_a_syn == 0)
+    s = format (s, ", filter value %d\n", t->filter_value);
+  else
+    s = format (s, "\n");
+  return s;
+}
+
+static vlib_node_registration_t syn_filter4_node;
+
+#define foreach_syn_filter_error                \
+_(THROTTLED, "TCP SYN packet throttle drops")   \
+_(OK, "TCP SYN packets passed")
+
+typedef enum
+{
+#define _(sym,str) SYN_FILTER_ERROR_##sym,
+  foreach_syn_filter_error
+#undef _
+    SYN_FILTER_N_ERROR,
+} syn_filter_error_t;
+
+static char *syn_filter4_error_strings[] = {
+#define _(sym,string) string,
+  foreach_syn_filter_error
+#undef _
+};
+
+typedef enum
+{
+  SYN_FILTER_NEXT_DROP,
+  SYN_FILTER_N_NEXT,
+} syn_filter_next_t;
+
+extern vnet_feature_arc_registration_t vnet_feat_arc_ip4_local;
+
+static uword
+syn_filter4_node_fn (vlib_main_t * vm,
+		     vlib_node_runtime_t * node, vlib_frame_t * frame)
+{
+  u32 n_left_from, *from, *to_next;
+  syn_filter_next_t next_index;
+  u32 ok_syn_packets = 0;
+  vnet_feature_main_t *fm = &feature_main;
+  u8 arc_index = vnet_feat_arc_ip4_local.feature_arc_index;
+  vnet_feature_config_main_t *cm = &fm->feature_config_mains[arc_index];
+  syn_filter4_runtime_t *rt = (syn_filter4_runtime_t *) node->runtime_data;
+  f64 now = vlib_time_now (vm);
+  /* Shut up spurious gcc warnings. */
+  u8 *c0 = 0, *c1 = 0, *c2 = 0, *c3 = 0;
+
+  from = vlib_frame_vector_args (frame);
+  n_left_from = frame->n_vectors;
+  next_index = node->cached_next_index;
+
+  if (now > rt->next_reset)
+    {
+      memset (rt->syn_counts, 0, vec_len (rt->syn_counts));
+      rt->next_reset = now + rt->reset_interval;
+    }
+
+  while (n_left_from > 0)
+    {
+      u32 n_left_to_next;
+
+      vlib_get_next_frame (vm, node, next_index, to_next, n_left_to_next);
+
+      while (n_left_from >= 8 && n_left_to_next >= 4)
+	{
+	  u32 bi0, bi1, bi2, bi3;
+	  vlib_buffer_t *b0, *b1, *b2, *b3;
+	  u32 next0, next1, next2, next3;
+	  ip4_header_t *ip0, *ip1, *ip2, *ip3;
+	  tcp_header_t *tcp0, *tcp1, *tcp2, *tcp3;
+	  u32 not_a_syn0 = 1, not_a_syn1 = 1, not_a_syn2 = 1, not_a_syn3 = 1;
+	  u64 hash0, hash1, hash2, hash3;
+
+	  /* Prefetch next iteration. */
+	  {
+	    vlib_buffer_t *p4, *p5, *p6, *p7;
+
+	    p4 = vlib_get_buffer (vm, from[4]);
+	    p5 = vlib_get_buffer (vm, from[5]);
+	    p6 = vlib_get_buffer (vm, from[6]);
+	    p7 = vlib_get_buffer (vm, from[7]);
+
+	    vlib_prefetch_buffer_header (p4, LOAD);
+	    vlib_prefetch_buffer_header (p5, LOAD);
+	    vlib_prefetch_buffer_header (p6, LOAD);
+	    vlib_prefetch_buffer_header (p7, LOAD);
+
+	    CLIB_PREFETCH (p4->data, CLIB_CACHE_LINE_BYTES, STORE);
+	    CLIB_PREFETCH (p5->data, CLIB_CACHE_LINE_BYTES, STORE);
+	    CLIB_PREFETCH (p6->data, CLIB_CACHE_LINE_BYTES, STORE);
+	    CLIB_PREFETCH (p7->data, CLIB_CACHE_LINE_BYTES, STORE);
+	  }
+
+	  /* speculatively enqueue b0 and b1 to the current next frame */
+	  to_next[0] = bi0 = from[0];
+	  to_next[1] = bi1 = from[1];
+	  to_next[2] = bi2 = from[2];
+	  to_next[3] = bi3 = from[3];
+	  from += 4;
+	  to_next += 4;
+	  n_left_from -= 4;
+	  n_left_to_next -= 4;
+
+	  b0 = vlib_get_buffer (vm, bi0);
+	  b1 = vlib_get_buffer (vm, bi1);
+	  b2 = vlib_get_buffer (vm, bi2);
+	  b3 = vlib_get_buffer (vm, bi3);
+
+	  vnet_get_config_data
+	    (&cm->config_main, &b0->current_config_index,
+	     &next0, 0 /* sizeof (c0[0]) */ );
+	  vnet_get_config_data
+	    (&cm->config_main, &b1->current_config_index,
+	     &next1, 0 /* sizeof (c0[0]) */ );
+	  vnet_get_config_data
+	    (&cm->config_main, &b2->current_config_index,
+	     &next2, 0 /* sizeof (c0[0]) */ );
+	  vnet_get_config_data
+	    (&cm->config_main, &b3->current_config_index,
+	     &next3, 0 /* sizeof (c0[0]) */ );
+
+	  /* Not TCP? */
+	  ip0 = vlib_buffer_get_current (b0);
+	  if (ip0->protocol != IP_PROTOCOL_TCP)
+	    goto trace00;
+
+	  tcp0 = ip4_next_header (ip0);
+	  /*
+	   * Not a SYN?
+	   * $$$$ hack: the TCP bitfield flags seem not to compile
+	   * correct code.
+	   */
+	  if (PREDICT_TRUE (!(tcp0->flags & 0x2)))
+	    goto trace00;
+
+	  not_a_syn0 = 0;
+	  hash0 = clib_xxhash ((u64) ip0->src_address.as_u32);
+	  c0 = &rt->syn_counts[hash0 & (_vec_len (rt->syn_counts) - 1)];
+	  if (PREDICT_FALSE (*c0 >= 0x80))
+	    {
+	      next0 = SYN_FILTER_NEXT_DROP;
+	      b0->error = node->errors[SYN_FILTER_ERROR_THROTTLED];
+	      goto trace00;
+	    }
+	  *c0 += 1;
+	  ok_syn_packets++;
+
+	trace00:
+	  if (PREDICT_FALSE ((node->flags & VLIB_NODE_FLAG_TRACE)
+			     && (b0->flags & VLIB_BUFFER_IS_TRACED)))
+	    {
+	      syn_filter4_trace_t *t =
+		vlib_add_trace (vm, node, b0, sizeof (*t));
+	      t->not_a_syn = not_a_syn0;
+	      t->next_index = next0;
+	      t->filter_value = not_a_syn0 ? 0 : *c0;
+	    }
+
+	  /* Not TCP? */
+	  ip1 = vlib_buffer_get_current (b1);
+	  if (ip1->protocol != IP_PROTOCOL_TCP)
+	    goto trace01;
+
+	  tcp1 = ip4_next_header (ip1);
+	  /*
+	   * Not a SYN?
+	   * $$$$ hack: the TCP bitfield flags seem not to compile
+	   * correct code.
+	   */
+	  if (PREDICT_TRUE (!(tcp1->flags & 0x2)))
+	    goto trace01;
+
+	  not_a_syn1 = 0;
+	  hash1 = clib_xxhash ((u64) ip1->src_address.as_u32);
+	  c1 = &rt->syn_counts[hash1 & (_vec_len (rt->syn_counts) - 1)];
+	  if (PREDICT_FALSE (*c1 >= 0x80))
+	    {
+	      next1 = SYN_FILTER_NEXT_DROP;
+	      b1->error = node->errors[SYN_FILTER_ERROR_THROTTLED];
+	      goto trace01;
+	    }
+	  *c1 += 1;
+	  ok_syn_packets++;
+
+	trace01:
+	  if (PREDICT_FALSE ((node->flags & VLIB_NODE_FLAG_TRACE)
+			     && (b1->flags & VLIB_BUFFER_IS_TRACED)))
+	    {
+	      syn_filter4_trace_t *t =
+		vlib_add_trace (vm, node, b1, sizeof (*t));
+	      t->not_a_syn = not_a_syn1;
+	      t->next_index = next1;
+	      t->filter_value = not_a_syn1 ? 0 : *c1;
+	    }
+
+	  /* Not TCP? */
+	  ip2 = vlib_buffer_get_current (b2);
+	  if (ip2->protocol != IP_PROTOCOL_TCP)
+	    goto trace02;
+
+	  tcp2 = ip4_next_header (ip2);
+	  /*
+	   * Not a SYN?
+	   * $$$$ hack: the TCP bitfield flags seem not to compile
+	   * correct code.
+	   */
+	  if (PREDICT_TRUE (!(tcp2->flags & 0x2)))
+	    goto trace02;
+
+	  not_a_syn2 = 0;
+	  hash2 = clib_xxhash ((u64) ip2->src_address.as_u32);
+	  c2 = &rt->syn_counts[hash2 & (_vec_len (rt->syn_counts) - 1)];
+	  if (PREDICT_FALSE (*c2 >= 0x80))
+	    {
+	      next2 = SYN_FILTER_NEXT_DROP;
+	      b2->error = node->errors[SYN_FILTER_ERROR_THROTTLED];
+	      goto trace02;
+	    }
+	  *c2 += 1;
+	  ok_syn_packets++;
+
+	trace02:
+	  if (PREDICT_FALSE ((node->flags & VLIB_NODE_FLAG_TRACE)
+			     && (b2->flags & VLIB_BUFFER_IS_TRACED)))
+	    {
+	      syn_filter4_trace_t *t =
+		vlib_add_trace (vm, node, b2, sizeof (*t));
+	      t->not_a_syn = not_a_syn2;
+	      t->next_index = next2;
+	      t->filter_value = not_a_syn2 ? 0 : *c2;
+	    }
+
+	  /* Not TCP? */
+	  ip3 = vlib_buffer_get_current (b3);
+	  if (ip3->protocol != IP_PROTOCOL_TCP)
+	    goto trace03;
+
+	  tcp3 = ip4_next_header (ip3);
+	  /*
+	   * Not a SYN?
+	   * $$$$ hack: the TCP bitfield flags seem not to compile
+	   * correct code.
+	   */
+	  if (PREDICT_TRUE (!(tcp3->flags & 0x2)))
+	    goto trace03;
+
+	  not_a_syn3 = 0;
+	  hash3 = clib_xxhash ((u64) ip3->src_address.as_u32);
+	  c3 = &rt->syn_counts[hash3 & (_vec_len (rt->syn_counts) - 1)];
+	  if (PREDICT_FALSE (*c3 >= 0x80))
+	    {
+	      next3 = SYN_FILTER_NEXT_DROP;
+	      b3->error = node->errors[SYN_FILTER_ERROR_THROTTLED];
+	      goto trace03;
+	    }
+	  *c3 += 1;
+	  ok_syn_packets++;
+
+	trace03:
+	  if (PREDICT_FALSE ((node->flags & VLIB_NODE_FLAG_TRACE)
+			     && (b3->flags & VLIB_BUFFER_IS_TRACED)))
+	    {
+	      syn_filter4_trace_t *t =
+		vlib_add_trace (vm, node, b3, sizeof (*t));
+	      t->not_a_syn = not_a_syn3;
+	      t->next_index = next3;
+	      t->filter_value = not_a_syn3 ? 0 : *c3;
+	    }
+	  vlib_validate_buffer_enqueue_x4 (vm, node, next_index,
+					   to_next, n_left_to_next,
+					   bi0, bi1, bi2, bi3,
+					   next0, next1, next2, next3);
+	}
+
+      while (n_left_from > 0 && n_left_to_next > 0)
+	{
+	  u32 bi0;
+	  vlib_buffer_t *b0;
+	  u32 next0;
+	  ip4_header_t *ip0;
+	  tcp_header_t *tcp0;
+	  u32 not_a_syn0 = 1;
+	  u32 hash0;
+	  u8 *c0;
+
+	  /* speculatively enqueue b0 to the current next frame */
+	  bi0 = from[0];
+	  to_next[0] = bi0;
+	  from += 1;
+	  to_next += 1;
+	  n_left_from -= 1;
+	  n_left_to_next -= 1;
+
+	  b0 = vlib_get_buffer (vm, bi0);
+
+	  vnet_get_config_data
+	    (&cm->config_main, &b0->current_config_index,
+	     &next0, 0 /* sizeof (c0[0]) */ );
+
+	  /* Not TCP? */
+	  ip0 = vlib_buffer_get_current (b0);
+	  if (ip0->protocol != IP_PROTOCOL_TCP)
+	    goto trace0;
+
+	  tcp0 = ip4_next_header (ip0);
+	  /*
+	   * Not a SYN?
+	   * $$$$ hack: the TCP bitfield flags seem not to compile
+	   * correct code.
+	   */
+	  if (PREDICT_TRUE (!(tcp0->flags & 0x2)))
+	    goto trace0;
+
+	  not_a_syn0 = 0;
+	  hash0 = clib_xxhash ((u64) ip0->src_address.as_u32);
+	  c0 = &rt->syn_counts[hash0 & (_vec_len (rt->syn_counts) - 1)];
+	  if (PREDICT_FALSE (*c0 >= 0x80))
+	    {
+	      next0 = SYN_FILTER_NEXT_DROP;
+	      b0->error = node->errors[SYN_FILTER_ERROR_THROTTLED];
+	      goto trace0;
+	    }
+	  *c0 += 1;
+	  ok_syn_packets++;
+
+	trace0:
+
+	  if (PREDICT_FALSE ((node->flags & VLIB_NODE_FLAG_TRACE)
+			     && (b0->flags & VLIB_BUFFER_IS_TRACED)))
+	    {
+	      syn_filter4_trace_t *t =
+		vlib_add_trace (vm, node, b0, sizeof (*t));
+	      t->not_a_syn = not_a_syn0;
+	      t->next_index = next0;
+	      t->filter_value = not_a_syn0 ? 0 : *c0;
+	    }
+
+	  /* verify speculative enqueue, maybe switch current next frame */
+	  vlib_validate_buffer_enqueue_x1 (vm, node, next_index,
+					   to_next, n_left_to_next,
+					   bi0, next0);
+	}
+
+      vlib_put_next_frame (vm, node, next_index, n_left_to_next);
+    }
+
+  vlib_node_increment_counter (vm, syn_filter4_node.index,
+			       SYN_FILTER_ERROR_OK, ok_syn_packets);
+  return frame->n_vectors;
+}
+
+/* *INDENT-OFF* */
+VLIB_REGISTER_NODE (syn_filter4_node, static) =
+{
+  .function = syn_filter4_node_fn,
+  .name = "syn-filter-4",
+  .vector_size = sizeof (u32),
+  .format_trace = format_syn_filter4_trace,
+  .type = VLIB_NODE_TYPE_INTERNAL,
+
+  .runtime_data_bytes = sizeof (syn_filter4_runtime_t),
+  .n_errors = ARRAY_LEN(syn_filter4_error_strings),
+  .error_strings = syn_filter4_error_strings,
+
+  .n_next_nodes = SYN_FILTER_N_NEXT,
+
+  /* edit / add dispositions here */
+  .next_nodes = {
+    [SYN_FILTER_NEXT_DROP] = "error-drop",
+  },
+};
+/* *INDENT-ON* */
+
+VLIB_NODE_FUNCTION_MULTIARCH (syn_filter4_node, syn_filter4_node_fn);
+
+/* *INDENT-OFF* */
+VNET_FEATURE_INIT (syn_filter_4, static) =
+{
+  .arc_name = "ip4-local",
+  .node_name = "syn-filter-4",
+  .runs_before = VNET_FEATURES("ip4-local-end-of-arc"),
+};
+/* *INDENT-ON* */
+
+int
+syn_filter_enable_disable (u32 sw_if_index, int enable_disable)
+{
+  vnet_main_t *vnm = vnet_get_main ();
+  vnet_sw_interface_t *sw;
+  int rv = 0;
+
+  /* Utterly wrong? */
+  if (pool_is_free_index (vnm->interface_main.sw_interfaces, sw_if_index))
+    return VNET_API_ERROR_INVALID_SW_IF_INDEX;
+
+  /* Not a physical port? */
+  sw = vnet_get_sw_interface (vnm, sw_if_index);
+  if (sw->type != VNET_SW_INTERFACE_TYPE_HARDWARE)
+    return VNET_API_ERROR_INVALID_SW_IF_INDEX;
+
+  if (enable_disable)
+    {
+      vlib_main_t *vm = vlib_get_main ();
+      syn_filter4_runtime_t *rt;
+
+      rt = vlib_node_get_runtime_data (vm, syn_filter4_node.index);
+      vec_validate (rt->syn_counts, 1023);
+      /*
+       * Given perfect disperson / optimal hashing results:
+       * Allow 128k (successful) syns/sec. 1024, buckets each of which
+       * absorb 128 syns before filtering. Reset table once a second.
+       * Reality bites, lets try resetting once every 100ms.
+       */
+      rt->reset_interval = 0.1;	/* reset interval in seconds */
+    }
+
+  rv = vnet_feature_enable_disable ("ip4-local", "syn-filter-4",
+				    sw_if_index, enable_disable, 0, 0);
+
+  return rv;
+}
+
+static clib_error_t *
+syn_filter_enable_disable_command_fn (vlib_main_t * vm,
+				      unformat_input_t * input,
+				      vlib_cli_command_t * cmd)
+{
+  vnet_main_t *vnm = vnet_get_main ();
+  u32 sw_if_index = ~0;
+  int enable_disable = 1;
+  int rv;
+
+  while (unformat_check_input (input) != UNFORMAT_END_OF_INPUT)
+    {
+      if (unformat (input, "disable"))
+	enable_disable = 0;
+      else if (unformat (input, "%U", unformat_vnet_sw_interface,
+			 vnm, &sw_if_index))
+	;
+      else
+	break;
+    }
+
+  if (sw_if_index == ~0)
+    return clib_error_return (0, "Please specify an interface...");
+
+  rv = syn_filter_enable_disable (sw_if_index, enable_disable);
+
+  switch (rv)
+    {
+    case 0:
+      break;
+
+    case VNET_API_ERROR_INVALID_SW_IF_INDEX:
+      return clib_error_return
+	(0, "Invalid interface, only works on physical ports");
+      break;
+
+    case VNET_API_ERROR_UNIMPLEMENTED:
+      return clib_error_return (0,
+				"Device driver doesn't support redirection");
+      break;
+
+    case VNET_API_ERROR_INVALID_VALUE:
+      return clib_error_return (0, "feature arc not found");
+
+    case VNET_API_ERROR_INVALID_VALUE_2:
+      return clib_error_return (0, "feature node not found");
+
+    default:
+      return clib_error_return (0, "syn_filter_enable_disable returned %d",
+				rv);
+    }
+  return 0;
+}
+
+/* *INDENT-OFF* */
+VLIB_CLI_COMMAND (sr_content_command, static) =
+{
+  .path = "ip syn filter",
+  .short_help = "ip syn filter  [disable]",
+  .function = syn_filter_enable_disable_command_fn,
+};
+/* *INDENT-ON* */
+
+/*
+ * fd.io coding-style-patch-verification: ON
+ *
+ * Local Variables:
+ * eval: (c-set-style "gnu")
+ * End:
+ */
diff --git a/src/vnet/tcp/tcp_timer.h b/src/vnet/tcp/tcp_timer.h
new file mode 100644
index 00000000..fa25268c
--- /dev/null
+++ b/src/vnet/tcp/tcp_timer.h
@@ -0,0 +1,29 @@
+/*
+ * Copyright (c) 2016 Cisco and/or its affiliates.
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at:
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+#ifndef __included_tcp_timer_h__
+#define __included_tcp_timer_h__
+
+#include 
+#include 
+
+#endif /* __included_tcp_timer_h__ */
+
+/*
+ * fd.io coding-style-patch-verification: ON
+ *
+ * Local Variables:
+ * eval: (c-set-style "gnu")
+ * End:
+ */
diff --git a/src/vnet/udp/builtin_server.c b/src/vnet/udp/builtin_server.c
new file mode 100644
index 00000000..afa66ba4
--- /dev/null
+++ b/src/vnet/udp/builtin_server.c
@@ -0,0 +1,239 @@
+/*
+ * Copyright (c) 2016 Cisco and/or its affiliates.
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at:
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+/** @file
+    udp builtin server
+*/
+
+#include 
+#include 
+#include 
+
+/** per-worker built-in server copy buffers */
+u8 **copy_buffers;
+
+static int
+builtin_session_create_callback (stream_session_t * s)
+{
+  /* Simple version: declare session ready-to-go... */
+  s->session_state = SESSION_STATE_READY;
+  return 0;
+}
+
+static void
+builtin_session_disconnect_callback (stream_session_t * s)
+{
+  stream_session_disconnect (s);
+}
+
+static int
+builtin_server_rx_callback (stream_session_t * s)
+{
+  svm_fifo_t *rx_fifo, *tx_fifo;
+  u32 this_transfer;
+  int actual_transfer;
+  u8 *my_copy_buffer;
+  session_fifo_event_t evt;
+  unix_shared_memory_queue_t *q;
+
+  my_copy_buffer = copy_buffers[s->thread_index];
+  rx_fifo = s->server_rx_fifo;
+  tx_fifo = s->server_tx_fifo;
+
+  this_transfer = svm_fifo_max_enqueue (tx_fifo)
+    < svm_fifo_max_dequeue (rx_fifo) ?
+    svm_fifo_max_enqueue (tx_fifo) : svm_fifo_max_dequeue (rx_fifo);
+
+  vec_validate (my_copy_buffer, this_transfer - 1);
+  _vec_len (my_copy_buffer) = this_transfer;
+
+  actual_transfer = svm_fifo_dequeue_nowait (rx_fifo, 0, this_transfer,
+					     my_copy_buffer);
+  ASSERT (actual_transfer == this_transfer);
+  actual_transfer = svm_fifo_enqueue_nowait (tx_fifo, 0, this_transfer,
+					     my_copy_buffer);
+
+  copy_buffers[s->thread_index] = my_copy_buffer;
+
+  /* Fabricate TX event, send to ourselves */
+  evt.fifo = tx_fifo;
+  evt.event_type = FIFO_EVENT_SERVER_TX;
+  /* $$$$ for event logging */
+  evt.enqueue_length = actual_transfer;
+  evt.event_id = 0;
+  q = session_manager_get_vpp_event_queue (s->thread_index);
+  unix_shared_memory_queue_add (q, (u8 *) & evt, 0 /* do wait for mutex */ );
+
+  return 0;
+}
+
+/* *INDENT-OFF* */
+static session_cb_vft_t builtin_server = {
+    .session_accept_callback = builtin_session_create_callback,
+    .session_disconnect_callback = builtin_session_disconnect_callback,
+    .builtin_server_rx_callback = builtin_server_rx_callback
+};
+/* *INDENT-ON* */
+
+static int
+bind_builtin_uri_server (u8 * uri)
+{
+  vnet_bind_args_t _a, *a = &_a;
+  char segment_name[128];
+  u32 segment_name_length;
+  int rv;
+  u64 options[16];
+
+  segment_name_length = ARRAY_LEN (segment_name);
+
+  memset (a, 0, sizeof (*a));
+  memset (options, 0, sizeof (options));
+
+  a->uri = (char *) uri;
+  a->api_client_index = ~0;	/* built-in server */
+  a->segment_name = segment_name;
+  a->segment_name_length = segment_name_length;
+  a->session_cb_vft = &builtin_server;
+
+  options[SESSION_OPTIONS_ACCEPT_COOKIE] = 0x12345678;
+  options[SESSION_OPTIONS_SEGMENT_SIZE] = (2 << 30);	/*$$$$ config / arg */
+  a->options = options;
+
+  rv = vnet_bind_uri (a);
+
+  return rv;
+}
+
+static int
+unbind_builtin_uri_server (u8 * uri)
+{
+  int rv;
+
+  rv = vnet_unbind_uri ((char *) uri, ~0 /* client_index */ );
+
+  return rv;
+}
+
+static clib_error_t *
+builtin_server_init (vlib_main_t * vm)
+{
+  vlib_thread_main_t *vtm = vlib_get_thread_main ();
+  u32 num_threads;
+
+  num_threads = 1 /* main thread */  + vtm->n_threads;
+
+  vec_validate (copy_buffers, num_threads - 1);
+  return 0;
+}
+
+VLIB_INIT_FUNCTION (builtin_server_init);
+
+static clib_error_t *
+builtin_uri_bind_command_fn (vlib_main_t * vm,
+			     unformat_input_t * input,
+			     vlib_cli_command_t * cmd)
+{
+  u8 *uri = 0;
+  int rv;
+
+  while (unformat_check_input (input) != UNFORMAT_END_OF_INPUT)
+    {
+      if (unformat (input, "uri %s", &uri))
+	;
+      else
+	break;
+    }
+
+  if (uri == 0)
+    return clib_error_return (0, "uri to bind not specified...");
+
+  rv = bind_builtin_uri_server (uri);
+
+  vec_free (uri);
+
+  switch (rv)
+    {
+    case 0:
+      break;
+
+    default:
+      return clib_error_return (0, "bind_uri_server returned %d", rv);
+      break;
+    }
+
+  return 0;
+}
+
+/* *INDENT-OFF* */
+VLIB_CLI_COMMAND (builtin_uri_bind_command, static) =
+{
+  .path = "builtin uri bind",
+  .short_help = "builtin uri bind",
+  .function = builtin_uri_bind_command_fn,
+};
+/* *INDENT-ON* */
+
+static clib_error_t *
+builtin_uri_unbind_command_fn (vlib_main_t * vm,
+			       unformat_input_t * input,
+			       vlib_cli_command_t * cmd)
+{
+  u8 *uri = 0;
+  int rv;
+
+  while (unformat_check_input (input) != UNFORMAT_END_OF_INPUT)
+    {
+      if (unformat (input, "uri %s", &uri))
+	;
+      else
+	break;
+    }
+
+  if (uri == 0)
+    return clib_error_return (0, "uri to unbind not specified...");
+
+  rv = unbind_builtin_uri_server (uri);
+
+  vec_free (uri);
+
+  switch (rv)
+    {
+    case 0:
+      break;
+
+    default:
+      return clib_error_return (0, "unbind_uri_server returned %d", rv);
+      break;
+    }
+
+  return 0;
+}
+
+/* *INDENT-OFF* */
+VLIB_CLI_COMMAND (builtin_uri_unbind_command, static) =
+{
+  .path = "builtin uri unbind",
+  .short_help = "builtin uri unbind",
+  .function = builtin_uri_unbind_command_fn,
+};
+/* *INDENT-ON* */
+
+/*
+ * fd.io coding-style-patch-verification: ON
+ *
+ * Local Variables:
+ * eval: (c-set-style "gnu")
+ * End:
+ */
diff --git a/src/vnet/udp/udp.c b/src/vnet/udp/udp.c
new file mode 100644
index 00000000..9e740466
--- /dev/null
+++ b/src/vnet/udp/udp.c
@@ -0,0 +1,342 @@
+/*
+ * Copyright (c) 2016 Cisco and/or its affiliates.
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at:
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+/** @file
+    udp state machine, etc.
+*/
+
+#include 
+#include 
+#include 
+#include 
+
+udp_uri_main_t udp_uri_main;
+
+u32
+udp_session_bind_ip4 (vlib_main_t * vm, u32 session_index,
+		      ip46_address_t * ip, u16 port_number_host_byte_order)
+{
+  udp_uri_main_t *um = vnet_get_udp_main ();
+  udp_connection_t *listener;
+
+  pool_get (um->udp_listeners, listener);
+  memset (listener, 0, sizeof (udp_connection_t));
+  listener->c_lcl_port = clib_host_to_net_u16 (port_number_host_byte_order);
+  listener->c_lcl_ip4.as_u32 = ip->ip4.as_u32;
+  listener->c_proto = SESSION_TYPE_IP4_UDP;
+  udp_register_dst_port (um->vlib_main, port_number_host_byte_order,
+			 udp4_uri_input_node.index, 1 /* is_ipv4 */ );
+  return 0;
+}
+
+u32
+udp_session_bind_ip6 (vlib_main_t * vm, u32 session_index,
+		      ip46_address_t * ip, u16 port_number_host_byte_order)
+{
+  udp_uri_main_t *um = vnet_get_udp_main ();
+  udp_connection_t *listener;
+
+  pool_get (um->udp_listeners, listener);
+  listener->c_lcl_port = clib_host_to_net_u16 (port_number_host_byte_order);
+  clib_memcpy (&listener->c_lcl_ip6, &ip->ip6, sizeof (ip6_address_t));
+  listener->c_proto = SESSION_TYPE_IP6_UDP;
+  udp_register_dst_port (um->vlib_main, port_number_host_byte_order,
+			 udp4_uri_input_node.index, 0 /* is_ipv4 */ );
+  return 0;
+}
+
+u32
+udp_session_unbind_ip4 (vlib_main_t * vm, u32 listener_index)
+{
+  udp_connection_t *listener;
+  listener = udp_listener_get (listener_index);
+
+  /* deregister the udp_local mapping */
+  udp_unregister_dst_port (vm, listener->c_lcl_port, 1 /* is_ipv4 */ );
+  return 0;
+}
+
+u32
+udp_session_unbind_ip6 (vlib_main_t * vm, u32 listener_index)
+{
+  udp_connection_t *listener;
+
+  listener = udp_listener_get (listener_index);
+
+  /* deregister the udp_local mapping */
+  udp_unregister_dst_port (vm, listener->c_lcl_port, 0 /* is_ipv4 */ );
+  return 0;
+}
+
+transport_connection_t *
+udp_session_get_listener (u32 listener_index)
+{
+  udp_connection_t *us;
+
+  us = udp_listener_get (listener_index);
+  return &us->connection;
+}
+
+u32
+udp_push_header (transport_connection_t * tconn, vlib_buffer_t * b)
+{
+  udp_connection_t *us;
+  u8 *data;
+  udp_header_t *udp;
+
+  us = (udp_connection_t *) tconn;
+
+  if (tconn->is_ip4)
+    {
+      ip4_header_t *ip;
+
+      data = vlib_buffer_get_current (b);
+      udp = (udp_header_t *) (data - sizeof (udp_header_t));
+      ip = (ip4_header_t *) ((u8 *) udp - sizeof (ip4_header_t));
+
+      /* Build packet header, swap rx key src + dst fields */
+      ip->src_address.as_u32 = us->c_lcl_ip4.as_u32;
+      ip->dst_address.as_u32 = us->c_rmt_ip4.as_u32;
+      ip->ip_version_and_header_length = 0x45;
+      ip->ttl = 254;
+      ip->protocol = IP_PROTOCOL_UDP;
+      ip->length = clib_host_to_net_u16 (b->current_length + sizeof (*udp));
+      ip->checksum = ip4_header_checksum (ip);
+
+      udp->src_port = us->c_lcl_port;
+      udp->dst_port = us->c_rmt_port;
+      udp->length = clib_host_to_net_u16 (b->current_length);
+      udp->checksum = 0;
+
+      b->current_length = sizeof (*ip) + sizeof (*udp);
+      return SESSION_QUEUE_NEXT_IP4_LOOKUP;
+    }
+  else
+    {
+      vlib_main_t *vm = vlib_get_main ();
+      ip6_header_t *ip;
+      u16 payload_length;
+      int bogus = ~0;
+
+      data = vlib_buffer_get_current (b);
+      udp = (udp_header_t *) (data - sizeof (udp_header_t));
+      ip = (ip6_header_t *) ((u8 *) udp - sizeof (ip6_header_t));
+
+      /* Build packet header, swap rx key src + dst fields */
+      clib_memcpy (&ip->src_address, &us->c_lcl_ip6, sizeof (ip6_address_t));
+      clib_memcpy (&ip->dst_address, &us->c_rmt_ip6, sizeof (ip6_address_t));
+
+      ip->ip_version_traffic_class_and_flow_label =
+	clib_host_to_net_u32 (0x6 << 28);
+
+      ip->hop_limit = 0xff;
+      ip->protocol = IP_PROTOCOL_UDP;
+
+      payload_length = vlib_buffer_length_in_chain (vm, b);
+      payload_length -= sizeof (*ip);
+
+      ip->payload_length = clib_host_to_net_u16 (payload_length);
+
+      udp->checksum = ip6_tcp_udp_icmp_compute_checksum (vm, b, ip, &bogus);
+      ASSERT (!bogus);
+
+      udp->src_port = us->c_lcl_port;
+      udp->dst_port = us->c_rmt_port;
+      udp->length = clib_host_to_net_u16 (b->current_length);
+      udp->checksum = 0;
+
+      b->current_length = sizeof (*ip) + sizeof (*udp);
+
+      return SESSION_QUEUE_NEXT_IP6_LOOKUP;
+    }
+}
+
+transport_connection_t *
+udp_session_get (u32 connection_index, u32 my_thread_index)
+{
+  udp_uri_main_t *um = vnet_get_udp_main ();
+
+  udp_connection_t *us;
+  us =
+    pool_elt_at_index (um->udp_sessions[my_thread_index], connection_index);
+  return &us->connection;
+}
+
+void
+udp_session_close (u32 connection_index, u32 my_thread_index)
+{
+  udp_uri_main_t *um = vnet_get_udp_main ();
+  pool_put_index (um->udp_sessions[my_thread_index], connection_index);
+}
+
+u8 *
+format_udp_session_ip4 (u8 * s, va_list * args)
+{
+  u32 uci = va_arg (*args, u32);
+  u32 thread_index = va_arg (*args, u32);
+  udp_connection_t *u4;
+
+  u4 = udp_connection_get (uci, thread_index);
+
+  s = format (s, "[%s] %U:%d->%U:%d", "udp", format_ip4_address,
+	      &u4->c_lcl_ip4, clib_net_to_host_u16 (u4->c_lcl_port),
+	      format_ip4_address, &u4->c_rmt_ip4,
+	      clib_net_to_host_u16 (u4->c_rmt_port));
+  return s;
+}
+
+u8 *
+format_udp_session_ip6 (u8 * s, va_list * args)
+{
+  u32 uci = va_arg (*args, u32);
+  u32 thread_index = va_arg (*args, u32);
+  udp_connection_t *tc = udp_connection_get (uci, thread_index);
+  s = format (s, "[%s] %U:%d->%U:%d", "udp", format_ip6_address,
+	      &tc->c_lcl_ip6, clib_net_to_host_u16 (tc->c_lcl_port),
+	      format_ip6_address, &tc->c_rmt_ip6,
+	      clib_net_to_host_u16 (tc->c_rmt_port));
+  return s;
+}
+
+u8 *
+format_udp_listener_session_ip4 (u8 * s, va_list * args)
+{
+  u32 tci = va_arg (*args, u32);
+  udp_connection_t *tc = udp_listener_get (tci);
+  s = format (s, "[%s] %U:%d->%U:%d", "udp", format_ip4_address,
+	      &tc->c_lcl_ip4, clib_net_to_host_u16 (tc->c_lcl_port),
+	      format_ip4_address, &tc->c_rmt_ip4,
+	      clib_net_to_host_u16 (tc->c_rmt_port));
+  return s;
+}
+
+u8 *
+format_udp_listener_session_ip6 (u8 * s, va_list * args)
+{
+  u32 tci = va_arg (*args, u32);
+  udp_connection_t *tc = udp_listener_get (tci);
+  s = format (s, "[%s] %U:%d->%U:%d", "udp", format_ip6_address,
+	      &tc->c_lcl_ip6, clib_net_to_host_u16 (tc->c_lcl_port),
+	      format_ip6_address, &tc->c_rmt_ip6,
+	      clib_net_to_host_u16 (tc->c_rmt_port));
+  return s;
+}
+
+u16
+udp_send_mss_uri (transport_connection_t * t)
+{
+  /* TODO figure out MTU of output interface */
+  return 400;
+}
+
+u32
+udp_send_space_uri (transport_connection_t * t)
+{
+  /* No constraint on TX window */
+  return ~0;
+}
+
+int
+udp_open_connection (ip46_address_t * addr, u16 port)
+{
+  clib_warning ("Not implemented");
+  return 0;
+}
+
+/* *INDENT-OFF* */
+const static transport_proto_vft_t udp4_proto = {
+  .bind = udp_session_bind_ip4,
+  .open = udp_open_connection,
+  .unbind = udp_session_unbind_ip4,
+  .push_header = udp_push_header,
+  .get_connection = udp_session_get,
+  .get_listener = udp_session_get_listener,
+  .close = udp_session_close,
+  .send_mss = udp_send_mss_uri,
+  .send_space = udp_send_space_uri,
+  .format_connection = format_udp_session_ip4,
+  .format_listener = format_udp_listener_session_ip4
+};
+
+const static transport_proto_vft_t udp6_proto = {
+  .bind = udp_session_bind_ip6,
+  .open = udp_open_connection,
+  .unbind = udp_session_unbind_ip6,
+  .push_header = udp_push_header,
+  .get_connection = udp_session_get,
+  .get_listener = udp_session_get_listener,
+  .close = udp_session_close,
+  .send_mss = udp_send_mss_uri,
+  .send_space = udp_send_space_uri,
+  .format_connection = format_udp_session_ip6,
+  .format_listener = format_udp_listener_session_ip6
+};
+/* *INDENT-ON* */
+
+static clib_error_t *
+udp_init (vlib_main_t * vm)
+{
+  udp_uri_main_t *um = vnet_get_udp_main ();
+  ip_main_t *im = &ip_main;
+  vlib_thread_main_t *tm = vlib_get_thread_main ();
+  u32 num_threads;
+  clib_error_t *error = 0;
+  ip_protocol_info_t *pi;
+
+  um->vlib_main = vm;
+  um->vnet_main = vnet_get_main ();
+
+  if ((error = vlib_call_init_function (vm, ip_main_init)))
+    return error;
+  if ((error = vlib_call_init_function (vm, ip4_lookup_init)))
+    return error;
+  if ((error = vlib_call_init_function (vm, ip6_lookup_init)))
+    return error;
+
+  /*
+   * Registrations
+   */
+
+  /* IP registration */
+  pi = ip_get_protocol_info (im, IP_PROTOCOL_UDP);
+  if (pi == 0)
+    return clib_error_return (0, "UDP protocol info AWOL");
+  pi->format_header = format_udp_header;
+  pi->unformat_pg_edit = unformat_pg_udp_header;
+
+
+  /* Register as transport with URI */
+  session_register_transport (SESSION_TYPE_IP4_UDP, &udp4_proto);
+  session_register_transport (SESSION_TYPE_IP6_UDP, &udp6_proto);
+
+  /*
+   * Initialize data structures
+   */
+
+  num_threads = 1 /* main thread */  + tm->n_threads;
+  vec_validate (um->udp_sessions, num_threads - 1);
+
+  return error;
+}
+
+VLIB_INIT_FUNCTION (udp_init);
+
+/*
+ * fd.io coding-style-patch-verification: ON
+ *
+ * Local Variables:
+ * eval: (c-set-style "gnu")
+ * End:
+ */
diff --git a/src/vnet/udp/udp.h b/src/vnet/udp/udp.h
new file mode 100644
index 00000000..7ab26ce9
--- /dev/null
+++ b/src/vnet/udp/udp.h
@@ -0,0 +1,362 @@
+/*
+ * Copyright (c) 2017 Cisco and/or its affiliates.
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at:
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+#ifndef __included_udp_h__
+#define __included_udp_h__
+
+#include 
+#include 
+#include 
+#include 
+#include 
+#include 
+#include 
+
+#include 
+#include 
+
+typedef struct
+{
+  transport_connection_t connection;	      /** must be first */
+
+  /** ersatz MTU to limit fifo pushes to test data size */
+  u32 mtu;
+} udp_connection_t;
+
+typedef struct _udp_uri_main
+{
+  /* Per-worker thread udp connection pools */
+  udp_connection_t **udp_sessions;
+  udp_connection_t *udp_listeners;
+
+  /* convenience */
+  vlib_main_t *vlib_main;
+  vnet_main_t *vnet_main;
+  ip4_main_t *ip4_main;
+  ip6_main_t *ip6_main;
+} udp_uri_main_t;
+
+extern udp_uri_main_t udp_uri_main;
+extern vlib_node_registration_t udp4_uri_input_node;
+
+always_inline udp_uri_main_t *
+vnet_get_udp_main ()
+{
+  return &udp_uri_main;
+}
+
+always_inline udp_connection_t *
+udp_connection_get (u32 conn_index, u32 thread_index)
+{
+  return pool_elt_at_index (udp_uri_main.udp_sessions[thread_index],
+			    conn_index);
+}
+
+always_inline udp_connection_t *
+udp_listener_get (u32 conn_index)
+{
+  return pool_elt_at_index (udp_uri_main.udp_listeners, conn_index);
+}
+
+typedef enum
+{
+#define udp_error(n,s) UDP_ERROR_##n,
+#include 
+#undef udp_error
+  UDP_N_ERROR,
+} udp_error_t;
+
+#define foreach_udp4_dst_port			\
+_ (67, dhcp_to_server)                          \
+_ (68, dhcp_to_client)                          \
+_ (500, ikev2)                                  \
+_ (3784, bfd4)                                  \
+_ (3785, bfd_echo4)                             \
+_ (4341, lisp_gpe)                              \
+_ (4342, lisp_cp)                          	\
+_ (4739, ipfix)                                 \
+_ (4789, vxlan)					\
+_ (4789, vxlan6)				\
+_ (4790, vxlan_gpe)				\
+_ (6633, vpath_3)
+
+
+#define foreach_udp6_dst_port                   \
+_ (547, dhcpv6_to_server)                       \
+_ (546, dhcpv6_to_client)			\
+_ (3784, bfd6)                                  \
+_ (3785, bfd_echo6)                             \
+_ (4341, lisp_gpe6)                             \
+_ (4342, lisp_cp6)                          	\
+_ (4790, vxlan6_gpe)      \
+_ (6633, vpath6_3)
+
+typedef enum
+{
+#define _(n,f) UDP_DST_PORT_##f = n,
+  foreach_udp4_dst_port foreach_udp6_dst_port
+#undef _
+} udp_dst_port_t;
+
+typedef enum
+{
+#define _(n,f) UDP6_DST_PORT_##f = n,
+  foreach_udp6_dst_port
+#undef _
+} udp6_dst_port_t;
+
+typedef struct
+{
+  /* Name (a c string). */
+  char *name;
+
+  /* GRE protocol type in host byte order. */
+  udp_dst_port_t dst_port;
+
+  /* Node which handles this type. */
+  u32 node_index;
+
+  /* Next index for this type. */
+  u32 next_index;
+} udp_dst_port_info_t;
+
+typedef enum
+{
+  UDP_IP6 = 0,
+  UDP_IP4,			/* the code is full of is_ip4... */
+  N_UDP_AF,
+} udp_af_t;
+
+typedef struct
+{
+  udp_dst_port_info_t *dst_port_infos[N_UDP_AF];
+
+  /* Hash tables mapping name/protocol to protocol info index. */
+  uword *dst_port_info_by_name[N_UDP_AF];
+  uword *dst_port_info_by_dst_port[N_UDP_AF];
+
+  /* convenience */
+  vlib_main_t *vlib_main;
+} udp_main_t;
+
+always_inline udp_dst_port_info_t *
+udp_get_dst_port_info (udp_main_t * um, udp_dst_port_t dst_port, u8 is_ip4)
+{
+  uword *p = hash_get (um->dst_port_info_by_dst_port[is_ip4], dst_port);
+  return p ? vec_elt_at_index (um->dst_port_infos[is_ip4], p[0]) : 0;
+}
+
+format_function_t format_udp_header;
+format_function_t format_udp_rx_trace;
+
+unformat_function_t unformat_udp_header;
+
+void udp_register_dst_port (vlib_main_t * vm,
+			    udp_dst_port_t dst_port,
+			    u32 node_index, u8 is_ip4);
+
+void
+udp_unregister_dst_port (vlib_main_t * vm,
+			 udp_dst_port_t dst_port, u8 is_ip4);
+
+void udp_punt_unknown (vlib_main_t * vm, u8 is_ip4, u8 is_add);
+
+always_inline void
+ip_udp_fixup_one (vlib_main_t * vm, vlib_buffer_t * b0, u8 is_ip4)
+{
+  u16 new_l0;
+  udp_header_t *udp0;
+
+  if (is_ip4)
+    {
+      ip4_header_t *ip0;
+      ip_csum_t sum0;
+      u16 old_l0 = 0;
+
+      ip0 = vlib_buffer_get_current (b0);
+
+      /* fix the ing outer-IP checksum */
+      sum0 = ip0->checksum;
+      /* old_l0 always 0, see the rewrite setup */
+      new_l0 = clib_host_to_net_u16 (vlib_buffer_length_in_chain (vm, b0));
+
+      sum0 = ip_csum_update (sum0, old_l0, new_l0, ip4_header_t,
+			     length /* changed member */ );
+      ip0->checksum = ip_csum_fold (sum0);
+      ip0->length = new_l0;
+
+      /* Fix UDP length */
+      udp0 = (udp_header_t *) (ip0 + 1);
+      new_l0 = clib_host_to_net_u16 (vlib_buffer_length_in_chain (vm, b0)
+				     - sizeof (*ip0));
+      udp0->length = new_l0;
+    }
+  else
+    {
+      ip6_header_t *ip0;
+      int bogus0;
+
+      ip0 = vlib_buffer_get_current (b0);
+
+      new_l0 = clib_host_to_net_u16 (vlib_buffer_length_in_chain (vm, b0)
+				     - sizeof (*ip0));
+      ip0->payload_length = new_l0;
+
+      /* Fix UDP length */
+      udp0 = (udp_header_t *) (ip0 + 1);
+      udp0->length = new_l0;
+
+      udp0->checksum =
+	ip6_tcp_udp_icmp_compute_checksum (vm, b0, ip0, &bogus0);
+      ASSERT (bogus0 == 0);
+
+      if (udp0->checksum == 0)
+	udp0->checksum = 0xffff;
+    }
+}
+
+always_inline void
+ip_udp_encap_one (vlib_main_t * vm, vlib_buffer_t * b0, u8 * ec0, word ec_len,
+		  u8 is_ip4)
+{
+  vlib_buffer_advance (b0, -ec_len);
+
+  if (is_ip4)
+    {
+      ip4_header_t *ip0;
+
+      ip0 = vlib_buffer_get_current (b0);
+
+      /* Apply the encap string. */
+      clib_memcpy (ip0, ec0, ec_len);
+      ip_udp_fixup_one (vm, b0, 1);
+    }
+  else
+    {
+      ip6_header_t *ip0;
+
+      ip0 = vlib_buffer_get_current (b0);
+
+      /* Apply the encap string. */
+      clib_memcpy (ip0, ec0, ec_len);
+      ip_udp_fixup_one (vm, b0, 0);
+    }
+}
+
+always_inline void
+ip_udp_encap_two (vlib_main_t * vm, vlib_buffer_t * b0, vlib_buffer_t * b1,
+		  u8 * ec0, u8 * ec1, word ec_len, u8 is_v4)
+{
+  u16 new_l0, new_l1;
+  udp_header_t *udp0, *udp1;
+
+  ASSERT (_vec_len (ec0) == _vec_len (ec1));
+
+  vlib_buffer_advance (b0, -ec_len);
+  vlib_buffer_advance (b1, -ec_len);
+
+  if (is_v4)
+    {
+      ip4_header_t *ip0, *ip1;
+      ip_csum_t sum0, sum1;
+      u16 old_l0 = 0, old_l1 = 0;
+
+      ip0 = vlib_buffer_get_current (b0);
+      ip1 = vlib_buffer_get_current (b1);
+
+      /* Apply the encap string */
+      clib_memcpy (ip0, ec0, ec_len);
+      clib_memcpy (ip1, ec1, ec_len);
+
+      /* fix the ing outer-IP checksum */
+      sum0 = ip0->checksum;
+      sum1 = ip1->checksum;
+
+      /* old_l0 always 0, see the rewrite setup */
+      new_l0 = clib_host_to_net_u16 (vlib_buffer_length_in_chain (vm, b0));
+      new_l1 = clib_host_to_net_u16 (vlib_buffer_length_in_chain (vm, b1));
+
+      sum0 = ip_csum_update (sum0, old_l0, new_l0, ip4_header_t,
+			     length /* changed member */ );
+      sum1 = ip_csum_update (sum1, old_l1, new_l1, ip4_header_t,
+			     length /* changed member */ );
+
+      ip0->checksum = ip_csum_fold (sum0);
+      ip1->checksum = ip_csum_fold (sum1);
+
+      ip0->length = new_l0;
+      ip1->length = new_l1;
+
+      /* Fix UDP length */
+      udp0 = (udp_header_t *) (ip0 + 1);
+      udp1 = (udp_header_t *) (ip1 + 1);
+
+      new_l0 =
+	clib_host_to_net_u16 (vlib_buffer_length_in_chain (vm, b0) -
+			      sizeof (*ip0));
+      new_l1 =
+	clib_host_to_net_u16 (vlib_buffer_length_in_chain (vm, b1) -
+			      sizeof (*ip1));
+      udp0->length = new_l0;
+      udp1->length = new_l1;
+    }
+  else
+    {
+      ip6_header_t *ip0, *ip1;
+      int bogus0, bogus1;
+
+      ip0 = vlib_buffer_get_current (b0);
+      ip1 = vlib_buffer_get_current (b1);
+
+      /* Apply the encap string. */
+      clib_memcpy (ip0, ec0, ec_len);
+      clib_memcpy (ip1, ec1, ec_len);
+
+      new_l0 = clib_host_to_net_u16 (vlib_buffer_length_in_chain (vm, b0)
+				     - sizeof (*ip0));
+      new_l1 = clib_host_to_net_u16 (vlib_buffer_length_in_chain (vm, b1)
+				     - sizeof (*ip1));
+      ip0->payload_length = new_l0;
+      ip1->payload_length = new_l1;
+
+      /* Fix UDP length */
+      udp0 = (udp_header_t *) (ip0 + 1);
+      udp1 = (udp_header_t *) (ip1 + 1);
+
+      udp0->length = new_l0;
+      udp1->length = new_l1;
+
+      udp0->checksum =
+	ip6_tcp_udp_icmp_compute_checksum (vm, b0, ip0, &bogus0);
+      udp1->checksum =
+	ip6_tcp_udp_icmp_compute_checksum (vm, b1, ip1, &bogus1);
+      ASSERT (bogus0 == 0);
+      ASSERT (bogus1 == 0);
+
+      if (udp0->checksum == 0)
+	udp0->checksum = 0xffff;
+      if (udp1->checksum == 0)
+	udp1->checksum = 0xffff;
+    }
+}
+
+/*
+ * fd.io coding-style-patch-verification: ON
+ *
+ * Local Variables:
+ * eval: (c-set-style "gnu")
+ * End:
+ */
+
+#endif /* __included_udp_h__ */
diff --git a/src/vnet/udp/udp_error.def b/src/vnet/udp/udp_error.def
new file mode 100644
index 00000000..bfdae0ac
--- /dev/null
+++ b/src/vnet/udp/udp_error.def
@@ -0,0 +1,21 @@
+/*
+ * udp_error.def: udp errors
+ *
+ * Copyright (c) 2013-2016 Cisco and/or its affiliates.
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at:
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+udp_error (NONE, "no error")
+udp_error (NO_LISTENER, "no listener for dst port")
+udp_error (LENGTH_ERROR, "UDP packets with length errors")
+udp_error (PUNT, "no listener punt")
diff --git a/src/vnet/udp/udp_format.c b/src/vnet/udp/udp_format.c
new file mode 100644
index 00000000..abdf561e
--- /dev/null
+++ b/src/vnet/udp/udp_format.c
@@ -0,0 +1,91 @@
+/*
+ * Copyright (c) 2015 Cisco and/or its affiliates.
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at:
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+/*
+ * ip/udp_format.c: udp formatting
+ *
+ * Copyright (c) 2008 Eliot Dresselhaus
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining
+ * a copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sublicense, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be
+ * included in all copies or substantial portions of the Software.
+ *
+ *  THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ *  EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ *  MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ *  NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE
+ *  LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
+ *  OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
+ *  WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+#include 
+
+/* Format UDP header. */
+u8 *
+format_udp_header (u8 * s, va_list * args)
+{
+  udp_header_t *udp = va_arg (*args, udp_header_t *);
+  u32 max_header_bytes = va_arg (*args, u32);
+  uword indent;
+  u32 header_bytes = sizeof (udp[0]);
+
+  /* Nothing to do. */
+  if (max_header_bytes < sizeof (udp[0]))
+    return format (s, "UDP header truncated");
+
+  indent = format_get_indent (s);
+  indent += 2;
+
+  s = format (s, "UDP: %d -> %d",
+	      clib_net_to_host_u16 (udp->src_port),
+	      clib_net_to_host_u16 (udp->dst_port));
+
+  s = format (s, "\n%Ulength %d, checksum 0x%04x",
+	      format_white_space, indent,
+	      clib_net_to_host_u16 (udp->length),
+	      clib_net_to_host_u16 (udp->checksum));
+
+  /* Recurse into next protocol layer. */
+  if (max_header_bytes != 0 && header_bytes < max_header_bytes)
+    {
+      ip_main_t *im = &ip_main;
+      tcp_udp_port_info_t *pi;
+
+      pi = ip_get_tcp_udp_port_info (im, udp->dst_port);
+
+      if (pi && pi->format_header)
+	s = format (s, "\n%U%U",
+		    format_white_space, indent - 2, pi->format_header,
+		    /* next protocol header */ (udp + 1),
+		    max_header_bytes - sizeof (udp[0]));
+    }
+
+  return s;
+}
+
+/*
+ * fd.io coding-style-patch-verification: ON
+ *
+ * Local Variables:
+ * eval: (c-set-style "gnu")
+ * End:
+ */
diff --git a/src/vnet/udp/udp_input.c b/src/vnet/udp/udp_input.c
new file mode 100644
index 00000000..4d509335
--- /dev/null
+++ b/src/vnet/udp/udp_input.c
@@ -0,0 +1,314 @@
+/*
+ * Copyright (c) 2016 Cisco and/or its affiliates.
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at:
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#include 
+#include 
+#include 
+#include 
+
+#include 
+#include 
+#include 
+#include 
+
+#include 
+
+#include 
+#include "../session/application_interface.h"
+
+vlib_node_registration_t udp4_uri_input_node;
+
+typedef struct
+{
+  u32 session;
+  u32 disposition;
+  u32 thread_index;
+} udp4_uri_input_trace_t;
+
+/* packet trace format function */
+static u8 *
+format_udp4_uri_input_trace (u8 * s, va_list * args)
+{
+  CLIB_UNUSED (vlib_main_t * vm) = va_arg (*args, vlib_main_t *);
+  CLIB_UNUSED (vlib_node_t * node) = va_arg (*args, vlib_node_t *);
+  udp4_uri_input_trace_t *t = va_arg (*args, udp4_uri_input_trace_t *);
+
+  s = format (s, "UDP4_URI_INPUT: session %d, disposition %d, thread %d",
+	      t->session, t->disposition, t->thread_index);
+  return s;
+}
+
+typedef enum
+{
+  UDP4_URI_INPUT_NEXT_DROP,
+  UDP4_URI_INPUT_N_NEXT,
+} udp4_uri_input_next_t;
+
+static char *udp4_uri_input_error_strings[] = {
+#define _(sym,string) string,
+  foreach_session_input_error
+#undef _
+};
+
+static uword
+udp4_uri_input_node_fn (vlib_main_t * vm,
+			vlib_node_runtime_t * node, vlib_frame_t * frame)
+{
+  u32 n_left_from, *from, *to_next;
+  udp4_uri_input_next_t next_index;
+  udp_uri_main_t *um = vnet_get_udp_main ();
+  session_manager_main_t *smm = vnet_get_session_manager_main ();
+  u32 my_thread_index = vm->cpu_index;
+  u8 my_enqueue_epoch;
+  u32 *session_indices_to_enqueue;
+  static u32 serial_number;
+  int i;
+
+  my_enqueue_epoch = ++smm->current_enqueue_epoch[my_thread_index];
+
+  from = vlib_frame_vector_args (frame);
+  n_left_from = frame->n_vectors;
+  next_index = node->cached_next_index;
+
+  while (n_left_from > 0)
+    {
+      u32 n_left_to_next;
+
+      vlib_get_next_frame (vm, node, next_index, to_next, n_left_to_next);
+
+      while (n_left_from > 0 && n_left_to_next > 0)
+	{
+	  u32 bi0;
+	  vlib_buffer_t *b0;
+	  u32 next0 = UDP4_URI_INPUT_NEXT_DROP;
+	  u32 error0 = SESSION_ERROR_ENQUEUED;
+	  udp_header_t *udp0;
+	  ip4_header_t *ip0;
+	  stream_session_t *s0;
+	  svm_fifo_t *f0;
+	  u16 udp_len0;
+	  u8 *data0;
+
+	  /* speculatively enqueue b0 to the current next frame */
+	  bi0 = from[0];
+	  to_next[0] = bi0;
+	  from += 1;
+	  to_next += 1;
+	  n_left_from -= 1;
+	  n_left_to_next -= 1;
+
+	  b0 = vlib_get_buffer (vm, bi0);
+
+	  /* udp_local hands us a pointer to the udp data */
+
+	  data0 = vlib_buffer_get_current (b0);
+	  udp0 = (udp_header_t *) (data0 - sizeof (*udp0));
+
+	  /* $$$$ fixme: udp_local doesn't do ip options correctly anyhow */
+	  ip0 = (ip4_header_t *) (((u8 *) udp0) - sizeof (*ip0));
+	  s0 = 0;
+
+	  /* lookup session */
+	  s0 = stream_session_lookup4 (&ip0->dst_address, &ip0->src_address,
+				       udp0->dst_port, udp0->src_port,
+				       SESSION_TYPE_IP4_UDP, my_thread_index);
+
+	  /* no listener */
+	  if (PREDICT_FALSE (s0 == 0))
+	    {
+	      error0 = SESSION_ERROR_NO_LISTENER;
+	      goto trace0;
+	    }
+
+	  f0 = s0->server_rx_fifo;
+
+	  /* established hit */
+	  if (PREDICT_TRUE (s0->session_state == SESSION_STATE_READY))
+	    {
+	      udp_len0 = clib_net_to_host_u16 (udp0->length);
+
+	      if (PREDICT_FALSE (udp_len0 > svm_fifo_max_enqueue (f0)))
+		{
+		  error0 = SESSION_ERROR_FIFO_FULL;
+		  goto trace0;
+		}
+
+	      svm_fifo_enqueue_nowait (f0, 0 /* pid */ ,
+				       udp_len0 - sizeof (*udp0),
+				       (u8 *) (udp0 + 1));
+
+	      b0->error = node->errors[SESSION_ERROR_ENQUEUED];
+
+	      /* We need to send an RX event on this fifo */
+	      if (s0->enqueue_epoch != my_enqueue_epoch)
+		{
+		  s0->enqueue_epoch = my_enqueue_epoch;
+
+		  vec_add1 (smm->session_indices_to_enqueue_by_thread
+			    [my_thread_index],
+			    s0 - smm->sessions[my_thread_index]);
+		}
+	    }
+	  /* listener hit */
+	  else if (s0->session_state == SESSION_STATE_LISTENING)
+	    {
+	      udp_connection_t *us;
+	      int rv;
+
+	      error0 = SESSION_ERROR_NOT_READY;
+
+	      /*
+	       * create udp transport session
+	       */
+	      pool_get (um->udp_sessions[my_thread_index], us);
+
+	      us->mtu = 1024;	/* $$$$ policy */
+
+	      us->c_lcl_ip4.as_u32 = ip0->dst_address.as_u32;
+	      us->c_rmt_ip4.as_u32 = ip0->src_address.as_u32;
+	      us->c_lcl_port = udp0->dst_port;
+	      us->c_rmt_port = udp0->src_port;
+	      us->c_proto = SESSION_TYPE_IP4_UDP;
+	      us->c_c_index = us - um->udp_sessions[my_thread_index];
+
+	      /*
+	       * create stream session and attach the udp session to it
+	       */
+	      rv = stream_session_accept (&us->connection, s0->session_index,
+					  SESSION_TYPE_IP4_UDP,
+					  1 /*notify */ );
+	      if (rv)
+		error0 = rv;
+
+	    }
+	  else
+	    {
+
+	      error0 = SESSION_ERROR_NOT_READY;
+	      goto trace0;
+	    }
+
+	trace0:
+	  b0->error = node->errors[error0];
+
+	  if (PREDICT_FALSE ((node->flags & VLIB_NODE_FLAG_TRACE)
+			     && (b0->flags & VLIB_BUFFER_IS_TRACED)))
+	    {
+	      udp4_uri_input_trace_t *t =
+		vlib_add_trace (vm, node, b0, sizeof (*t));
+
+	      t->session = ~0;
+	      if (s0)
+		t->session = s0 - smm->sessions[my_thread_index];
+	      t->disposition = error0;
+	      t->thread_index = my_thread_index;
+	    }
+
+	  /* verify speculative enqueue, maybe switch current next frame */
+	  vlib_validate_buffer_enqueue_x1 (vm, node, next_index,
+					   to_next, n_left_to_next,
+					   bi0, next0);
+	}
+
+      vlib_put_next_frame (vm, node, next_index, n_left_to_next);
+    }
+
+  /* Send enqueue events */
+
+  session_indices_to_enqueue =
+    smm->session_indices_to_enqueue_by_thread[my_thread_index];
+
+  for (i = 0; i < vec_len (session_indices_to_enqueue); i++)
+    {
+      session_fifo_event_t evt;
+      unix_shared_memory_queue_t *q;
+      stream_session_t *s0;
+      application_t *server0;
+
+      /* Get session */
+      s0 = pool_elt_at_index (smm->sessions[my_thread_index],
+			      session_indices_to_enqueue[i]);
+
+      /* Get session's server */
+      server0 = application_get (s0->app_index);
+
+      /* Built-in server? Deliver the goods... */
+      if (server0->cb_fns.builtin_server_rx_callback)
+	{
+	  server0->cb_fns.builtin_server_rx_callback (s0);
+	  continue;
+	}
+
+      /* Fabricate event */
+      evt.fifo = s0->server_rx_fifo;
+      evt.event_type = FIFO_EVENT_SERVER_RX;
+      evt.event_id = serial_number++;
+      evt.enqueue_length = svm_fifo_max_dequeue (s0->server_rx_fifo);
+
+      /* Add event to server's event queue */
+      q = server0->event_queue;
+
+      /* Don't block for lack of space */
+      if (PREDICT_TRUE (q->cursize < q->maxsize))
+	unix_shared_memory_queue_add (server0->event_queue, (u8 *) & evt,
+				      0 /* do wait for mutex */ );
+      else
+	{
+	  vlib_node_increment_counter (vm, udp4_uri_input_node.index,
+				       SESSION_ERROR_FIFO_FULL, 1);
+	}
+      if (1)
+	{
+	  ELOG_TYPE_DECLARE (e) =
+	  {
+	  .format = "evt-enqueue: id %d length %d",.format_args = "i4i4",};
+	  struct
+	  {
+	    u32 data[2];
+	  } *ed;
+	  ed = ELOG_DATA (&vlib_global_main.elog_main, e);
+	  ed->data[0] = evt.event_id;
+	  ed->data[1] = evt.enqueue_length;
+	}
+    }
+
+  vec_reset_length (session_indices_to_enqueue);
+
+  smm->session_indices_to_enqueue_by_thread[my_thread_index] =
+    session_indices_to_enqueue;
+
+  return frame->n_vectors;
+}
+
+VLIB_REGISTER_NODE (udp4_uri_input_node) =
+{
+  .function = udp4_uri_input_node_fn,.name = "udp4-uri-input",.vector_size =
+    sizeof (u32),.format_trace = format_udp4_uri_input_trace,.type =
+    VLIB_NODE_TYPE_INTERNAL,.n_errors =
+    ARRAY_LEN (udp4_uri_input_error_strings),.error_strings =
+    udp4_uri_input_error_strings,.n_next_nodes = UDP4_URI_INPUT_N_NEXT,
+    /* edit / add dispositions here */
+    .next_nodes =
+  {
+  [UDP4_URI_INPUT_NEXT_DROP] = "error-drop",}
+,};
+
+/*
+ * fd.io coding-style-patch-verification: ON
+ *
+ * Local Variables:
+ * eval: (c-set-style "gnu")
+ * End:
+ */
diff --git a/src/vnet/udp/udp_local.c b/src/vnet/udp/udp_local.c
new file mode 100644
index 00000000..6b239f73
--- /dev/null
+++ b/src/vnet/udp/udp_local.c
@@ -0,0 +1,666 @@
+/*
+ * node.c: udp packet processing
+ *
+ * Copyright (c) 2013 Cisco and/or its affiliates.
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at:
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#include 
+#include 
+#include 
+#include 
+#include 
+
+udp_main_t udp_main;
+
+#define foreach_udp_input_next                  \
+  _ (PUNT, "error-punt")                        \
+  _ (DROP, "error-drop")                        \
+  _ (ICMP4_ERROR, "ip4-icmp-error")             \
+  _ (ICMP6_ERROR, "ip6-icmp-error")
+
+typedef enum
+{
+#define _(s,n) UDP_INPUT_NEXT_##s,
+  foreach_udp_input_next
+#undef _
+    UDP_INPUT_N_NEXT,
+} udp_input_next_t;
+
+typedef struct
+{
+  u16 src_port;
+  u16 dst_port;
+  u8 bound;
+} udp_rx_trace_t;
+
+u8 *
+format_udp_rx_trace (u8 * s, va_list * args)
+{
+  CLIB_UNUSED (vlib_main_t * vm) = va_arg (*args, vlib_main_t *);
+  CLIB_UNUSED (vlib_node_t * node) = va_arg (*args, vlib_node_t *);
+  udp_rx_trace_t *t = va_arg (*args, udp_rx_trace_t *);
+
+  s = format (s, "UDP: src-port %d dst-port %d%s",
+	      clib_net_to_host_u16 (t->src_port),
+	      clib_net_to_host_u16 (t->dst_port),
+	      t->bound ? "" : " (no listener)");
+  return s;
+}
+
+typedef struct
+{
+  /* Sparse vector mapping udp dst_port in network byte order
+     to next index. */
+  u16 *next_by_dst_port;
+  u8 punt_unknown;
+} udp_input_runtime_t;
+
+vlib_node_registration_t udp4_input_node;
+vlib_node_registration_t udp6_input_node;
+
+always_inline uword
+udp46_input_inline (vlib_main_t * vm,
+		    vlib_node_runtime_t * node,
+		    vlib_frame_t * from_frame, int is_ip4)
+{
+  udp_input_runtime_t *rt = is_ip4 ?
+    (void *) vlib_node_get_runtime_data (vm, udp4_input_node.index)
+    : (void *) vlib_node_get_runtime_data (vm, udp6_input_node.index);
+  __attribute__ ((unused)) u32 n_left_from, next_index, *from, *to_next;
+  word n_no_listener = 0;
+  u8 punt_unknown = rt->punt_unknown;
+
+  from = vlib_frame_vector_args (from_frame);
+  n_left_from = from_frame->n_vectors;
+
+  next_index = node->cached_next_index;
+
+  while (n_left_from > 0)
+    {
+      u32 n_left_to_next;
+
+      vlib_get_next_frame (vm, node, next_index, to_next, n_left_to_next);
+
+      while (n_left_from >= 4 && n_left_to_next >= 2)
+	{
+	  u32 bi0, bi1;
+	  vlib_buffer_t *b0, *b1;
+	  udp_header_t *h0 = 0, *h1 = 0;
+	  u32 i0, i1, dst_port0, dst_port1;
+	  u32 advance0, advance1;
+	  u32 error0, next0, error1, next1;
+
+	  /* Prefetch next iteration. */
+	  {
+	    vlib_buffer_t *p2, *p3;
+
+	    p2 = vlib_get_buffer (vm, from[2]);
+	    p3 = vlib_get_buffer (vm, from[3]);
+
+	    vlib_prefetch_buffer_header (p2, LOAD);
+	    vlib_prefetch_buffer_header (p3, LOAD);
+
+	    CLIB_PREFETCH (p2->data, sizeof (h0[0]), LOAD);
+	    CLIB_PREFETCH (p3->data, sizeof (h1[0]), LOAD);
+	  }
+
+	  bi0 = from[0];
+	  bi1 = from[1];
+	  to_next[0] = bi0;
+	  to_next[1] = bi1;
+	  from += 2;
+	  to_next += 2;
+	  n_left_to_next -= 2;
+	  n_left_from -= 2;
+
+	  b0 = vlib_get_buffer (vm, bi0);
+	  b1 = vlib_get_buffer (vm, bi1);
+
+	  /* ip4/6_local hands us the ip header, not the udp header */
+	  if (is_ip4)
+	    {
+	      advance0 = sizeof (ip4_header_t);
+	      advance1 = sizeof (ip4_header_t);
+	    }
+	  else
+	    {
+	      advance0 = sizeof (ip6_header_t);
+	      advance1 = sizeof (ip6_header_t);
+	    }
+
+	  if (PREDICT_FALSE (b0->current_length < advance0 + sizeof (*h0)))
+	    {
+	      error0 = UDP_ERROR_LENGTH_ERROR;
+	      next0 = UDP_INPUT_NEXT_DROP;
+	    }
+	  else
+	    {
+	      vlib_buffer_advance (b0, advance0);
+	      h0 = vlib_buffer_get_current (b0);
+	      error0 = next0 = 0;
+	      if (PREDICT_FALSE (clib_net_to_host_u16 (h0->length) >
+				 vlib_buffer_length_in_chain (vm, b0)))
+		{
+		  error0 = UDP_ERROR_LENGTH_ERROR;
+		  next0 = UDP_INPUT_NEXT_DROP;
+		}
+	    }
+
+	  if (PREDICT_FALSE (b1->current_length < advance1 + sizeof (*h1)))
+	    {
+	      error1 = UDP_ERROR_LENGTH_ERROR;
+	      next1 = UDP_INPUT_NEXT_DROP;
+	    }
+	  else
+	    {
+	      vlib_buffer_advance (b1, advance1);
+	      h1 = vlib_buffer_get_current (b1);
+	      error1 = next1 = 0;
+	      if (PREDICT_FALSE (clib_net_to_host_u16 (h1->length) >
+				 vlib_buffer_length_in_chain (vm, b1)))
+		{
+		  error1 = UDP_ERROR_LENGTH_ERROR;
+		  next1 = UDP_INPUT_NEXT_DROP;
+		}
+	    }
+
+	  /* Index sparse array with network byte order. */
+	  dst_port0 = (error0 == 0) ? h0->dst_port : 0;
+	  dst_port1 = (error1 == 0) ? h1->dst_port : 0;
+	  sparse_vec_index2 (rt->next_by_dst_port, dst_port0, dst_port1,
+			     &i0, &i1);
+	  next0 = (error0 == 0) ? vec_elt (rt->next_by_dst_port, i0) : next0;
+	  next1 = (error1 == 0) ? vec_elt (rt->next_by_dst_port, i1) : next1;
+
+	  if (PREDICT_FALSE (i0 == SPARSE_VEC_INVALID_INDEX))
+	    {
+	      // move the pointer back so icmp-error can find the
+	      // ip packet header
+	      vlib_buffer_advance (b0, -(word) advance0);
+
+	      if (PREDICT_FALSE (punt_unknown))
+		{
+		  b0->error = node->errors[UDP_ERROR_PUNT];
+		  next0 = UDP_INPUT_NEXT_PUNT;
+		}
+	      else if (is_ip4)
+		{
+		  icmp4_error_set_vnet_buffer (b0,
+					       ICMP4_destination_unreachable,
+					       ICMP4_destination_unreachable_port_unreachable,
+					       0);
+		  next0 = UDP_INPUT_NEXT_ICMP4_ERROR;
+		  n_no_listener++;
+		}
+	      else
+		{
+		  icmp6_error_set_vnet_buffer (b0,
+					       ICMP6_destination_unreachable,
+					       ICMP6_destination_unreachable_port_unreachable,
+					       0);
+		  next0 = UDP_INPUT_NEXT_ICMP6_ERROR;
+		  n_no_listener++;
+		}
+	    }
+	  else
+	    {
+	      b0->error = node->errors[UDP_ERROR_NONE];
+	      // advance to the payload
+	      vlib_buffer_advance (b0, sizeof (*h0));
+	    }
+
+	  if (PREDICT_FALSE (i1 == SPARSE_VEC_INVALID_INDEX))
+	    {
+	      // move the pointer back so icmp-error can find the
+	      // ip packet header
+	      vlib_buffer_advance (b1, -(word) advance1);
+
+	      if (PREDICT_FALSE (punt_unknown))
+		{
+		  b1->error = node->errors[UDP_ERROR_PUNT];
+		  next1 = UDP_INPUT_NEXT_PUNT;
+		}
+	      else if (is_ip4)
+		{
+		  icmp4_error_set_vnet_buffer (b1,
+					       ICMP4_destination_unreachable,
+					       ICMP4_destination_unreachable_port_unreachable,
+					       0);
+		  next1 = UDP_INPUT_NEXT_ICMP4_ERROR;
+		  n_no_listener++;
+		}
+	      else
+		{
+		  icmp6_error_set_vnet_buffer (b1,
+					       ICMP6_destination_unreachable,
+					       ICMP6_destination_unreachable_port_unreachable,
+					       0);
+		  next1 = UDP_INPUT_NEXT_ICMP6_ERROR;
+		  n_no_listener++;
+		}
+	    }
+	  else
+	    {
+	      b1->error = node->errors[UDP_ERROR_NONE];
+	      // advance to the payload
+	      vlib_buffer_advance (b1, sizeof (*h1));
+	    }
+
+	  if (PREDICT_FALSE (b0->flags & VLIB_BUFFER_IS_TRACED))
+	    {
+	      udp_rx_trace_t *tr = vlib_add_trace (vm, node,
+						   b0, sizeof (*tr));
+	      if (b0->error != node->errors[UDP_ERROR_LENGTH_ERROR])
+		{
+		  tr->src_port = h0 ? h0->src_port : 0;
+		  tr->dst_port = h0 ? h0->dst_port : 0;
+		  tr->bound = (next0 != UDP_INPUT_NEXT_ICMP4_ERROR &&
+			       next0 != UDP_INPUT_NEXT_ICMP6_ERROR);
+		}
+	    }
+	  if (PREDICT_FALSE (b1->flags & VLIB_BUFFER_IS_TRACED))
+	    {
+	      udp_rx_trace_t *tr = vlib_add_trace (vm, node,
+						   b1, sizeof (*tr));
+	      if (b1->error != node->errors[UDP_ERROR_LENGTH_ERROR])
+		{
+		  tr->src_port = h1 ? h1->src_port : 0;
+		  tr->dst_port = h1 ? h1->dst_port : 0;
+		  tr->bound = (next1 != UDP_INPUT_NEXT_ICMP4_ERROR &&
+			       next1 != UDP_INPUT_NEXT_ICMP6_ERROR);
+		}
+	    }
+
+	  vlib_validate_buffer_enqueue_x2 (vm, node, next_index,
+					   to_next, n_left_to_next,
+					   bi0, bi1, next0, next1);
+	}
+
+      while (n_left_from > 0 && n_left_to_next > 0)
+	{
+	  u32 bi0;
+	  vlib_buffer_t *b0;
+	  udp_header_t *h0 = 0;
+	  u32 i0, next0;
+	  u32 advance0;
+
+	  bi0 = from[0];
+	  to_next[0] = bi0;
+	  from += 1;
+	  to_next += 1;
+	  n_left_from -= 1;
+	  n_left_to_next -= 1;
+
+	  b0 = vlib_get_buffer (vm, bi0);
+
+	  /* ip4/6_local hands us the ip header, not the udp header */
+	  if (is_ip4)
+	    advance0 = sizeof (ip4_header_t);
+	  else
+	    advance0 = sizeof (ip6_header_t);
+
+	  if (PREDICT_FALSE (b0->current_length < advance0 + sizeof (*h0)))
+	    {
+	      b0->error = node->errors[UDP_ERROR_LENGTH_ERROR];
+	      next0 = UDP_INPUT_NEXT_DROP;
+	      goto trace_x1;
+	    }
+
+	  vlib_buffer_advance (b0, advance0);
+
+	  h0 = vlib_buffer_get_current (b0);
+
+	  if (PREDICT_TRUE (clib_net_to_host_u16 (h0->length) <=
+			    vlib_buffer_length_in_chain (vm, b0)))
+	    {
+	      i0 = sparse_vec_index (rt->next_by_dst_port, h0->dst_port);
+	      next0 = vec_elt (rt->next_by_dst_port, i0);
+
+	      if (PREDICT_FALSE (i0 == SPARSE_VEC_INVALID_INDEX))
+		{
+		  // move the pointer back so icmp-error can find the
+		  // ip packet header
+		  vlib_buffer_advance (b0, -(word) advance0);
+
+		  if (PREDICT_FALSE (punt_unknown))
+		    {
+		      b0->error = node->errors[UDP_ERROR_PUNT];
+		      next0 = UDP_INPUT_NEXT_PUNT;
+		    }
+		  else if (is_ip4)
+		    {
+		      icmp4_error_set_vnet_buffer (b0,
+						   ICMP4_destination_unreachable,
+						   ICMP4_destination_unreachable_port_unreachable,
+						   0);
+		      next0 = UDP_INPUT_NEXT_ICMP4_ERROR;
+		      n_no_listener++;
+		    }
+		  else
+		    {
+		      icmp6_error_set_vnet_buffer (b0,
+						   ICMP6_destination_unreachable,
+						   ICMP6_destination_unreachable_port_unreachable,
+						   0);
+		      next0 = UDP_INPUT_NEXT_ICMP6_ERROR;
+		      n_no_listener++;
+		    }
+		}
+	      else
+		{
+		  b0->error = node->errors[UDP_ERROR_NONE];
+		  // advance to the payload
+		  vlib_buffer_advance (b0, sizeof (*h0));
+		}
+	    }
+	  else
+	    {
+	      b0->error = node->errors[UDP_ERROR_LENGTH_ERROR];
+	      next0 = UDP_INPUT_NEXT_DROP;
+	    }
+
+	trace_x1:
+	  if (PREDICT_FALSE (b0->flags & VLIB_BUFFER_IS_TRACED))
+	    {
+	      udp_rx_trace_t *tr = vlib_add_trace (vm, node,
+						   b0, sizeof (*tr));
+	      if (b0->error != node->errors[UDP_ERROR_LENGTH_ERROR])
+		{
+		  tr->src_port = h0->src_port;
+		  tr->dst_port = h0->dst_port;
+		  tr->bound = (next0 != UDP_INPUT_NEXT_ICMP4_ERROR &&
+			       next0 != UDP_INPUT_NEXT_ICMP6_ERROR);
+		}
+	    }
+
+	  vlib_validate_buffer_enqueue_x1 (vm, node, next_index,
+					   to_next, n_left_to_next,
+					   bi0, next0);
+	}
+
+      vlib_put_next_frame (vm, node, next_index, n_left_to_next);
+    }
+  vlib_error_count (vm, node->node_index, UDP_ERROR_NO_LISTENER,
+		    n_no_listener);
+  return from_frame->n_vectors;
+}
+
+static char *udp_error_strings[] = {
+#define udp_error(n,s) s,
+#include "udp_error.def"
+#undef udp_error
+};
+
+static uword
+udp4_input (vlib_main_t * vm,
+	    vlib_node_runtime_t * node, vlib_frame_t * from_frame)
+{
+  return udp46_input_inline (vm, node, from_frame, 1 /* is_ip4 */ );
+}
+
+static uword
+udp6_input (vlib_main_t * vm,
+	    vlib_node_runtime_t * node, vlib_frame_t * from_frame)
+{
+  return udp46_input_inline (vm, node, from_frame, 0 /* is_ip4 */ );
+}
+
+
+/* *INDENT-OFF* */
+VLIB_REGISTER_NODE (udp4_input_node) = {
+  .function = udp4_input,
+  .name = "ip4-udp-lookup",
+  /* Takes a vector of packets. */
+  .vector_size = sizeof (u32),
+
+  .runtime_data_bytes = sizeof (udp_input_runtime_t),
+
+  .n_errors = UDP_N_ERROR,
+  .error_strings = udp_error_strings,
+
+  .n_next_nodes = UDP_INPUT_N_NEXT,
+  .next_nodes = {
+#define _(s,n) [UDP_INPUT_NEXT_##s] = n,
+    foreach_udp_input_next
+#undef _
+  },
+
+  .format_buffer = format_udp_header,
+  .format_trace = format_udp_rx_trace,
+  .unformat_buffer = unformat_udp_header,
+};
+/* *INDENT-ON* */
+
+VLIB_NODE_FUNCTION_MULTIARCH (udp4_input_node, udp4_input);
+
+/* *INDENT-OFF* */
+VLIB_REGISTER_NODE (udp6_input_node) = {
+  .function = udp6_input,
+  .name = "ip6-udp-lookup",
+  /* Takes a vector of packets. */
+  .vector_size = sizeof (u32),
+
+  .runtime_data_bytes = sizeof (udp_input_runtime_t),
+
+  .n_errors = UDP_N_ERROR,
+  .error_strings = udp_error_strings,
+
+  .n_next_nodes = UDP_INPUT_N_NEXT,
+  .next_nodes = {
+#define _(s,n) [UDP_INPUT_NEXT_##s] = n,
+    foreach_udp_input_next
+#undef _
+  },
+
+  .format_buffer = format_udp_header,
+  .format_trace = format_udp_rx_trace,
+  .unformat_buffer = unformat_udp_header,
+};
+/* *INDENT-ON* */
+
+VLIB_NODE_FUNCTION_MULTIARCH (udp6_input_node, udp6_input);
+
+static void
+add_dst_port (udp_main_t * um,
+	      udp_dst_port_t dst_port, char *dst_port_name, u8 is_ip4)
+{
+  udp_dst_port_info_t *pi;
+  u32 i;
+
+  vec_add2 (um->dst_port_infos[is_ip4], pi, 1);
+  i = pi - um->dst_port_infos[is_ip4];
+
+  pi->name = dst_port_name;
+  pi->dst_port = dst_port;
+  pi->next_index = pi->node_index = ~0;
+
+  hash_set (um->dst_port_info_by_dst_port[is_ip4], dst_port, i);
+
+  if (pi->name)
+    hash_set_mem (um->dst_port_info_by_name[is_ip4], pi->name, i);
+}
+
+void
+udp_register_dst_port (vlib_main_t * vm,
+		       udp_dst_port_t dst_port, u32 node_index, u8 is_ip4)
+{
+  udp_main_t *um = &udp_main;
+  udp_dst_port_info_t *pi;
+  udp_input_runtime_t *rt;
+  u16 *n;
+
+  {
+    clib_error_t *error = vlib_call_init_function (vm, udp_local_init);
+    if (error)
+      clib_error_report (error);
+  }
+
+  pi = udp_get_dst_port_info (um, dst_port, is_ip4);
+  if (!pi)
+    {
+      add_dst_port (um, dst_port, 0, is_ip4);
+      pi = udp_get_dst_port_info (um, dst_port, is_ip4);
+      ASSERT (pi);
+    }
+
+  pi->node_index = node_index;
+  pi->next_index = vlib_node_add_next (vm,
+				       is_ip4 ? udp4_input_node.index
+				       : udp6_input_node.index, node_index);
+
+  /* Setup udp protocol -> next index sparse vector mapping. */
+  rt = vlib_node_get_runtime_data
+    (vm, is_ip4 ? udp4_input_node.index : udp6_input_node.index);
+  n = sparse_vec_validate (rt->next_by_dst_port,
+			   clib_host_to_net_u16 (dst_port));
+  n[0] = pi->next_index;
+}
+
+void
+udp_unregister_dst_port (vlib_main_t * vm, udp_dst_port_t dst_port, u8 is_ip4)
+{
+  udp_main_t *um = &udp_main;
+  udp_dst_port_info_t *pi;
+  udp_input_runtime_t *rt;
+  u16 *n;
+
+  pi = udp_get_dst_port_info (um, dst_port, is_ip4);
+  /* Not registered? Fagedaboudit */
+  if (!pi)
+    return;
+
+  /* Kill the mapping. Don't bother killing the pi, it may be back. */
+  rt = vlib_node_get_runtime_data
+    (vm, is_ip4 ? udp4_input_node.index : udp6_input_node.index);
+  n = sparse_vec_validate (rt->next_by_dst_port,
+			   clib_host_to_net_u16 (dst_port));
+  n[0] = SPARSE_VEC_INVALID_INDEX;
+}
+
+void
+udp_punt_unknown (vlib_main_t * vm, u8 is_ip4, u8 is_add)
+{
+  udp_input_runtime_t *rt;
+
+  {
+    clib_error_t *error = vlib_call_init_function (vm, udp_local_init);
+    if (error)
+      clib_error_report (error);
+  }
+
+  rt = vlib_node_get_runtime_data
+    (vm, is_ip4 ? udp4_input_node.index : udp6_input_node.index);
+
+  rt->punt_unknown = is_add;
+}
+
+/* Parse a UDP header. */
+uword
+unformat_udp_header (unformat_input_t * input, va_list * args)
+{
+  u8 **result = va_arg (*args, u8 **);
+  udp_header_t *udp;
+  __attribute__ ((unused)) int old_length;
+  u16 src_port, dst_port;
+
+  /* Allocate space for IP header. */
+  {
+    void *p;
+
+    old_length = vec_len (*result);
+    vec_add2 (*result, p, sizeof (ip4_header_t));
+    udp = p;
+  }
+
+  memset (udp, 0, sizeof (udp[0]));
+  if (unformat (input, "src-port %d dst-port %d", &src_port, &dst_port))
+    {
+      udp->src_port = clib_host_to_net_u16 (src_port);
+      udp->dst_port = clib_host_to_net_u16 (dst_port);
+      return 1;
+    }
+  return 0;
+}
+
+static void
+udp_setup_node (vlib_main_t * vm, u32 node_index)
+{
+  vlib_node_t *n = vlib_get_node (vm, node_index);
+  pg_node_t *pn = pg_get_node (node_index);
+
+  n->format_buffer = format_udp_header;
+  n->unformat_buffer = unformat_udp_header;
+  pn->unformat_edit = unformat_pg_udp_header;
+}
+
+clib_error_t *
+udp_local_init (vlib_main_t * vm)
+{
+  udp_input_runtime_t *rt;
+  udp_main_t *um = &udp_main;
+  int i;
+
+  {
+    clib_error_t *error;
+    error = vlib_call_init_function (vm, udp_init);
+    if (error)
+      clib_error_report (error);
+  }
+
+
+  for (i = 0; i < 2; i++)
+    {
+      um->dst_port_info_by_name[i] = hash_create_string (0, sizeof (uword));
+      um->dst_port_info_by_dst_port[i] = hash_create (0, sizeof (uword));
+    }
+
+  udp_setup_node (vm, udp4_input_node.index);
+  udp_setup_node (vm, udp6_input_node.index);
+
+  rt = vlib_node_get_runtime_data (vm, udp4_input_node.index);
+
+  rt->next_by_dst_port = sparse_vec_new
+    ( /* elt bytes */ sizeof (rt->next_by_dst_port[0]),
+     /* bits in index */ BITS (((udp_header_t *) 0)->dst_port));
+
+  rt->punt_unknown = 0;
+
+#define _(n,s) add_dst_port (um, UDP_DST_PORT_##s, #s, 1 /* is_ip4 */);
+  foreach_udp4_dst_port
+#undef _
+    rt = vlib_node_get_runtime_data (vm, udp6_input_node.index);
+
+  rt->next_by_dst_port = sparse_vec_new
+    ( /* elt bytes */ sizeof (rt->next_by_dst_port[0]),
+     /* bits in index */ BITS (((udp_header_t *) 0)->dst_port));
+
+  rt->punt_unknown = 0;
+
+#define _(n,s) add_dst_port (um, UDP_DST_PORT_##s, #s, 0 /* is_ip4 */);
+  foreach_udp6_dst_port
+#undef _
+    ip4_register_protocol (IP_PROTOCOL_UDP, udp4_input_node.index);
+  /* Note: ip6 differs from ip4, UDP is hotwired to ip6-udp-lookup */
+  return 0;
+}
+
+VLIB_INIT_FUNCTION (udp_local_init);
+
+/*
+ * fd.io coding-style-patch-verification: ON
+ *
+ * Local Variables:
+ * eval: (c-set-style "gnu")
+ * End:
+ */
diff --git a/src/vnet/udp/udp_packet.h b/src/vnet/udp/udp_packet.h
new file mode 100644
index 00000000..beea3059
--- /dev/null
+++ b/src/vnet/udp/udp_packet.h
@@ -0,0 +1,65 @@
+/*
+ * Copyright (c) 2015 Cisco and/or its affiliates.
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at:
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+/*
+ * ip4/udp_packet.h: UDP packet format
+ *
+ * Copyright (c) 2008 Eliot Dresselhaus
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining
+ * a copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sublicense, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be
+ * included in all copies or substantial portions of the Software.
+ *
+ *  THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ *  EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ *  MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ *  NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE
+ *  LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
+ *  OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
+ *  WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+#ifndef included_udp_packet_h
+#define included_udp_packet_h
+
+typedef struct
+{
+  /* Source and destination port. */
+  u16 src_port, dst_port;
+
+  /* Length of UDP header plus payload. */
+  u16 length;
+
+  /* Checksum of UDP pseudo-header and data or
+     zero if checksum is disabled. */
+  u16 checksum;
+} udp_header_t;
+
+#endif /* included_udp_packet_h */
+
+
+/*
+ * fd.io coding-style-patch-verification: ON
+ *
+ * Local Variables:
+ * eval: (c-set-style "gnu")
+ * End:
+ */
diff --git a/src/vnet/udp/udp_pg.c b/src/vnet/udp/udp_pg.c
new file mode 100644
index 00000000..c9d8d38c
--- /dev/null
+++ b/src/vnet/udp/udp_pg.c
@@ -0,0 +1,237 @@
+/*
+ * Copyright (c) 2015 Cisco and/or its affiliates.
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at:
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+/*
+ * ip/udp_pg: UDP packet-generator interface
+ *
+ * Copyright (c) 2008 Eliot Dresselhaus
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining
+ * a copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sublicense, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be
+ * included in all copies or substantial portions of the Software.
+ *
+ *  THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ *  EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ *  MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ *  NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE
+ *  LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
+ *  OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
+ *  WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+#include 
+#include 		/* for unformat_udp_udp_port */
+
+#define UDP_PG_EDIT_LENGTH (1 << 0)
+#define UDP_PG_EDIT_CHECKSUM (1 << 1)
+
+always_inline void
+udp_pg_edit_function_inline (pg_main_t * pg,
+			     pg_stream_t * s,
+			     pg_edit_group_t * g,
+			     u32 * packets, u32 n_packets, u32 flags)
+{
+  vlib_main_t *vm = vlib_get_main ();
+  u32 ip_offset, udp_offset;
+
+  udp_offset = g->start_byte_offset;
+  ip_offset = (g - 1)->start_byte_offset;
+
+  while (n_packets >= 1)
+    {
+      vlib_buffer_t *p0;
+      ip4_header_t *ip0;
+      udp_header_t *udp0;
+      u32 udp_len0;
+
+      p0 = vlib_get_buffer (vm, packets[0]);
+      n_packets -= 1;
+      packets += 1;
+
+      ip0 = (void *) (p0->data + ip_offset);
+      udp0 = (void *) (p0->data + udp_offset);
+      udp_len0 = clib_net_to_host_u16 (ip0->length) - sizeof (ip0[0]);
+
+      if (flags & UDP_PG_EDIT_LENGTH)
+	udp0->length =
+	  clib_net_to_host_u16 (vlib_buffer_length_in_chain (vm, p0)
+				- ip_offset);
+
+      /* Initialize checksum with header. */
+      if (flags & UDP_PG_EDIT_CHECKSUM)
+	{
+	  ip_csum_t sum0;
+
+	  sum0 = clib_mem_unaligned (&ip0->src_address, u64);
+
+	  sum0 = ip_csum_with_carry
+	    (sum0, clib_host_to_net_u32 (udp_len0 + (ip0->protocol << 16)));
+
+	  /* Invalidate possibly old checksum. */
+	  udp0->checksum = 0;
+
+	  sum0 =
+	    ip_incremental_checksum_buffer (vm, p0, udp_offset, udp_len0,
+					    sum0);
+
+	  sum0 = ~ip_csum_fold (sum0);
+
+	  /* Zero checksum means checksumming disabled. */
+	  sum0 = sum0 != 0 ? sum0 : 0xffff;
+
+	  udp0->checksum = sum0;
+	}
+    }
+}
+
+static void
+udp_pg_edit_function (pg_main_t * pg,
+		      pg_stream_t * s,
+		      pg_edit_group_t * g, u32 * packets, u32 n_packets)
+{
+  switch (g->edit_function_opaque)
+    {
+    case UDP_PG_EDIT_LENGTH:
+      udp_pg_edit_function_inline (pg, s, g, packets, n_packets,
+				   UDP_PG_EDIT_LENGTH);
+      break;
+
+    case UDP_PG_EDIT_CHECKSUM:
+      udp_pg_edit_function_inline (pg, s, g, packets, n_packets,
+				   UDP_PG_EDIT_CHECKSUM);
+      break;
+
+    case UDP_PG_EDIT_CHECKSUM | UDP_PG_EDIT_LENGTH:
+      udp_pg_edit_function_inline (pg, s, g, packets, n_packets,
+				   UDP_PG_EDIT_CHECKSUM | UDP_PG_EDIT_LENGTH);
+      break;
+
+    default:
+      ASSERT (0);
+      break;
+    }
+}
+
+typedef struct
+{
+  pg_edit_t src_port, dst_port;
+  pg_edit_t length;
+  pg_edit_t checksum;
+} pg_udp_header_t;
+
+static inline void
+pg_udp_header_init (pg_udp_header_t * p)
+{
+  /* Initialize fields that are not bit fields in the IP header. */
+#define _(f) pg_edit_init (&p->f, udp_header_t, f);
+  _(src_port);
+  _(dst_port);
+  _(length);
+  _(checksum);
+#undef _
+}
+
+uword
+unformat_pg_udp_header (unformat_input_t * input, va_list * args)
+{
+  pg_stream_t *s = va_arg (*args, pg_stream_t *);
+  pg_udp_header_t *p;
+  u32 group_index;
+
+  p = pg_create_edit_group (s, sizeof (p[0]), sizeof (udp_header_t),
+			    &group_index);
+  pg_udp_header_init (p);
+
+  /* Defaults. */
+  p->checksum.type = PG_EDIT_UNSPECIFIED;
+  p->length.type = PG_EDIT_UNSPECIFIED;
+
+  if (!unformat (input, "UDP: %U -> %U",
+		 unformat_pg_edit,
+		 unformat_tcp_udp_port, &p->src_port,
+		 unformat_pg_edit, unformat_tcp_udp_port, &p->dst_port))
+    goto error;
+
+  /* Parse options. */
+  while (1)
+    {
+      if (unformat (input, "length %U",
+		    unformat_pg_edit, unformat_pg_number, &p->length))
+	;
+
+      else if (unformat (input, "checksum %U",
+			 unformat_pg_edit, unformat_pg_number, &p->checksum))
+	;
+
+      /* Can't parse input: try next protocol level. */
+      else
+	break;
+    }
+
+  {
+    ip_main_t *im = &ip_main;
+    u16 dst_port;
+    tcp_udp_port_info_t *pi;
+
+    pi = 0;
+    if (p->dst_port.type == PG_EDIT_FIXED)
+      {
+	dst_port = pg_edit_get_value (&p->dst_port, PG_EDIT_LO);
+	pi = ip_get_tcp_udp_port_info (im, dst_port);
+      }
+
+    if (pi && pi->unformat_pg_edit
+	&& unformat_user (input, pi->unformat_pg_edit, s))
+      ;
+
+    else if (!unformat_user (input, unformat_pg_payload, s))
+      goto error;
+
+    p = pg_get_edit_group (s, group_index);
+    if (p->checksum.type == PG_EDIT_UNSPECIFIED
+	|| p->length.type == PG_EDIT_UNSPECIFIED)
+      {
+	pg_edit_group_t *g = pg_stream_get_group (s, group_index);
+	g->edit_function = udp_pg_edit_function;
+	g->edit_function_opaque = 0;
+	if (p->checksum.type == PG_EDIT_UNSPECIFIED)
+	  g->edit_function_opaque |= UDP_PG_EDIT_CHECKSUM;
+	if (p->length.type == PG_EDIT_UNSPECIFIED)
+	  g->edit_function_opaque |= UDP_PG_EDIT_LENGTH;
+      }
+
+    return 1;
+  }
+
+error:
+  /* Free up any edits we may have added. */
+  pg_free_edit_group (s);
+  return 0;
+}
+
+
+/*
+ * fd.io coding-style-patch-verification: ON
+ *
+ * Local Variables:
+ * eval: (c-set-style "gnu")
+ * End:
+ */
diff --git a/src/vnet/vnet_all_api_h.h b/src/vnet/vnet_all_api_h.h
index 142acedc..c4075db6 100644
--- a/src/vnet/vnet_all_api_h.h
+++ b/src/vnet/vnet_all_api_h.h
@@ -51,6 +51,7 @@
 #include 
 #include 
 #include 
+#include 
 #include 
 #include 
 #include 
diff --git a/src/vnet/vxlan-gpe/vxlan_gpe.h b/src/vnet/vxlan-gpe/vxlan_gpe.h
index 1b4bc44e..e768d230 100644
--- a/src/vnet/vxlan-gpe/vxlan_gpe.h
+++ b/src/vnet/vxlan-gpe/vxlan_gpe.h
@@ -29,7 +29,7 @@
 #include 
 #include 
 #include 
-#include 
+#include 
 
 /**
  * @brief VXLAN GPE header struct
diff --git a/src/vnet/vxlan/vxlan.h b/src/vnet/vxlan/vxlan.h
index adfa3a8e..dca1cd12 100644
--- a/src/vnet/vxlan/vxlan.h
+++ b/src/vnet/vxlan/vxlan.h
@@ -26,7 +26,7 @@
 #include 
 #include 
 #include 
-#include 
+#include 
 #include 
 #include 
 
diff --git a/src/vpp/api/vpe.api b/src/vpp/api/vpe.api
index 24f48293..2d6e4f37 100644
--- a/src/vpp/api/vpe.api
+++ b/src/vpp/api/vpe.api
@@ -38,6 +38,7 @@
  * IPSEC-GRE APIs: see .../src/vnet/ipsec-gre/{ipsec_gre.api, ipsec_gre_api.c}
  * LISP APIs: see .../src/vnet/lisp/{lisp.api, lisp_api.c}
  * LISP-GPE APIs: see .../src/vnet/lisp-gpe/{lisp_gpe.api, lisp_gpe_api.c}
+ * SESSION APIs: .../vnet/session/{session.api session_api.c}
  * MPLS APIs: see .../src/vnet/mpls/{mpls.api, mpls_api.c}
  * SR APIs: see .../src/vnet/sr/{sr.api, sr_api.c}
  * DPDK APIs: see ... /src/vnet/devices/dpdk/{dpdk.api, dpdk_api.c}
diff --git a/src/vppinfra.am b/src/vppinfra.am
index 8d375958..4b9f0c29 100644
--- a/src/vppinfra.am
+++ b/src/vppinfra.am
@@ -157,7 +157,9 @@ nobase_include_HEADERS = \
   vppinfra/asm_mips.h \
   vppinfra/asm_x86.h \
   vppinfra/bihash_8_8.h \
+  vppinfra/bihash_16_8.h \
   vppinfra/bihash_24_8.h \
+  vppinfra/bihash_48_8.h \
   vppinfra/bihash_template.h \
   vppinfra/bihash_template.c \
   vppinfra/bitmap.h \
@@ -206,6 +208,7 @@ nobase_include_HEADERS = \
   vppinfra/timer.h \
   vppinfra/tw_timer_2t_1w_2048sl.h \
   vppinfra/tw_timer_16t_2w_512sl.h \
+  vppinfra/tw_timer_16t_1w_2048sl.h \
   vppinfra/tw_timer_template.h \
   vppinfra/tw_timer_template.c \
   vppinfra/types.h \
@@ -261,6 +264,8 @@ CLIB_CORE = \
   vppinfra/tw_timer_2t_1w_2048sl.c \
   vppinfra/tw_timer_16t_2w_512sl.h \
   vppinfra/tw_timer_16t_2w_512sl.c \
+  vppinfra/tw_timer_16t_1w_2048sl.h \
+  vppinfra/tw_timer_16t_1w_2048sl.c \
   vppinfra/unformat.c \
   vppinfra/vec.c \
   vppinfra/vector.c \
diff --git a/src/vppinfra/bihash_16_8.h b/src/vppinfra/bihash_16_8.h
new file mode 100644
index 00000000..ce80f70e
--- /dev/null
+++ b/src/vppinfra/bihash_16_8.h
@@ -0,0 +1,103 @@
+/*
+ * Copyright (c) 2016 Cisco and/or its affiliates.
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at:
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+#undef BIHASH_TYPE
+
+#define BIHASH_TYPE _16_8
+#define BIHASH_KVP_PER_PAGE 4
+
+#ifndef __included_bihash_16_8_h__
+#define __included_bihash_16_8_h__
+
+#include 
+#include 
+#include 
+#include 
+
+typedef struct
+{
+  u64 key[2];
+  u64 value;
+} clib_bihash_kv_16_8_t;
+
+static inline int
+clib_bihash_is_free_16_8 (clib_bihash_kv_16_8_t * v)
+{
+  /* Free values are memset to 0xff, check a bit... */
+  if (v->key[0] == ~0ULL && v->value == ~0ULL)
+    return 1;
+  return 0;
+}
+
+#if __SSE4_2__
+#ifndef __defined_crc_u32__
+#define __defined_crc_u32__
+static inline u32
+crc_u32 (u32 data, u32 value)
+{
+  __asm__ volatile ("crc32l %[data], %[value];":[value] "+r" (value):[data]
+		    "rm" (data));
+  return value;
+}
+#endif /* __defined_crc_u32__ */
+
+static inline u64
+clib_bihash_hash_16_8 (clib_bihash_kv_16_8_t * v)
+{
+  u32 *dp = (u32 *) & v->key[0];
+  u32 value = 0;
+
+  value = crc_u32 (dp[0], value);
+  value = crc_u32 (dp[1], value);
+  value = crc_u32 (dp[2], value);
+  value = crc_u32 (dp[3], value);
+
+  return value;
+}
+#else
+static inline u64
+clib_bihash_hash_16_8 (clib_bihash_kv_16_8_t * v)
+{
+  u64 tmp = v->key[0] ^ v->key[1];
+  return clib_xxhash (tmp);
+}
+#endif
+
+static inline u8 *
+format_bihash_kvp_16_8 (u8 * s, va_list * args)
+{
+  clib_bihash_kv_16_8_t *v = va_arg (*args, clib_bihash_kv_16_8_t *);
+
+  s = format (s, "key %llu %llu value %llu", v->key[0], v->key[1], v->value);
+  return s;
+}
+
+static inline int
+clib_bihash_key_compare_16_8 (u64 * a, u64 * b)
+{
+  return ((a[0] ^ b[0]) | (a[1] ^ b[1])) == 0;
+}
+
+#undef __included_bihash_template_h__
+#include 
+
+#endif /* __included_bihash_16_8_h__ */
+
+/*
+ * fd.io coding-style-patch-verification: ON
+ *
+ * Local Variables:
+ * eval: (c-set-style "gnu")
+ * End:
+ */
diff --git a/src/vppinfra/bihash_48_8.h b/src/vppinfra/bihash_48_8.h
new file mode 100644
index 00000000..1a6e7691
--- /dev/null
+++ b/src/vppinfra/bihash_48_8.h
@@ -0,0 +1,116 @@
+/*
+ * Copyright (c) 2016 Cisco and/or its affiliates.
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at:
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#undef BIHASH_TYPE
+
+#define BIHASH_TYPE _48_8
+#define BIHASH_KVP_PER_PAGE 4
+
+#ifndef __included_bihash_48_8_h__
+#define __included_bihash_48_8_h__
+
+#include 
+#include 
+#include 
+#include 
+
+typedef struct
+{
+  u64 key[6];
+  u64 value;
+} clib_bihash_kv_48_8_t;
+
+static inline int
+clib_bihash_is_free_48_8 (const clib_bihash_kv_48_8_t * v)
+{
+  /* Free values are memset to 0xff, check a bit... */
+  if (v->key[0] == ~0ULL && v->value == ~0ULL)
+    return 1;
+  return 0;
+}
+
+#if __SSE4_2__
+#ifndef __defined_crc_u32__
+#define __defined_crc_u32__
+static inline u32
+crc_u32 (u32 data, u32 value)
+{
+  __asm__ volatile ("crc32l %[data], %[value];":[value] "+r" (value):[data]
+		    "rm" (data));
+  return value;
+}
+#endif /* __defined_crc_u32__ */
+
+static inline u64
+clib_bihash_hash_48_8 (const clib_bihash_kv_48_8_t * v)
+{
+  const u32 *dp = (const u32 *) &v->key[0];
+  u32 value = 0;
+
+  value = crc_u32 (dp[0], value);
+  value = crc_u32 (dp[1], value);
+  value = crc_u32 (dp[2], value);
+  value = crc_u32 (dp[3], value);
+  value = crc_u32 (dp[4], value);
+  value = crc_u32 (dp[5], value);
+  value = crc_u32 (dp[6], value);
+  value = crc_u32 (dp[7], value);
+  value = crc_u32 (dp[8], value);
+  value = crc_u32 (dp[9], value);
+  value = crc_u32 (dp[10], value);
+  value = crc_u32 (dp[11], value);
+
+  return value;
+}
+#else
+static inline u64
+clib_bihash_hash_48_8 (const clib_bihash_kv_48_8_t * v)
+{
+  u64 tmp = v->key[0] ^ v->key[1] ^ v->key[2] ^ v->key[3] ^ v->key[4]
+    ^ v->key[5];
+  return clib_xxhash (tmp);
+}
+#endif
+
+static inline u8 *
+format_bihash_kvp_48_8 (u8 * s, va_list * args)
+{
+  clib_bihash_kv_48_8_t *v = va_arg (*args, clib_bihash_kv_48_8_t *);
+
+  s = format (s, "key %llu %llu %llu %llu %llu %llu value %llu", v->key[0],
+	      v->key[1], v->key[2], v->key[3], v->key[4], v->key[5],
+	      v->value);
+  return s;
+}
+
+static inline int
+clib_bihash_key_compare_48_8 (const u64 * a, const u64 * b)
+{
+  return ((a[0] ^ b[0]) | (a[1] ^ b[1]) | (a[2] ^ b[2]) | (a[3] ^ b[3])
+	  | (a[4] ^ b[4]) | (a[5] ^ b[5])) == 0;
+}
+
+#undef __included_bihash_template_h__
+#include 
+
+#endif /* __included_bihash_48_8_h__ */
+
+/*
+ * fd.io coding-style-patch-verification: ON
+ *
+ * Local Variables:
+ * eval: (c-set-style "gnu")
+ * End:
+ */
diff --git a/src/vppinfra/tw_timer_16t_1w_2048sl.c b/src/vppinfra/tw_timer_16t_1w_2048sl.c
new file mode 100644
index 00000000..3f342045
--- /dev/null
+++ b/src/vppinfra/tw_timer_16t_1w_2048sl.c
@@ -0,0 +1,26 @@
+/*
+ * Copyright (c) 2017 Cisco and/or its affiliates.
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at:
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#include 
+#include "tw_timer_16t_1w_2048sl.h"
+#include "tw_timer_template.c"
+
+/*
+ * fd.io coding-style-patch-verification: ON
+ *
+ * Local Variables:
+ * eval: (c-set-style "gnu")
+ * End:
+ */
diff --git a/src/vppinfra/tw_timer_16t_1w_2048sl.h b/src/vppinfra/tw_timer_16t_1w_2048sl.h
new file mode 100644
index 00000000..685ac31e
--- /dev/null
+++ b/src/vppinfra/tw_timer_16t_1w_2048sl.h
@@ -0,0 +1,46 @@
+/*
+ * Copyright (c) 2017 Cisco and/or its affiliates.
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at:
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#ifndef __included_tw_timer_16t_2w_512sl_h__
+#define __included_tw_timer_16t_2w_512sl_h__
+
+/* ... So that a client app can create multiple wheel geometries */
+#undef TW_TIMER_WHEELS
+#undef TW_SLOTS_PER_RING
+#undef TW_RING_SHIFT
+#undef TW_RING_MASK
+#undef TW_TIMERS_PER_OBJECT
+#undef LOG2_TW_TIMERS_PER_OBJECT
+#undef TW_SUFFIX
+
+#define TW_TIMER_WHEELS 1
+#define TW_SLOTS_PER_RING 2048
+#define TW_RING_SHIFT 11
+#define TW_RING_MASK (TW_SLOTS_PER_RING -1)
+#define TW_TIMERS_PER_OBJECT 16
+#define LOG2_TW_TIMERS_PER_OBJECT 4
+#define TW_SUFFIX _16t_1w_2048sl
+
+#include 
+
+#endif /* __included_tw_timer_16t_2w_512sl_h__ */
+
+/*
+ * fd.io coding-style-patch-verification: ON
+ *
+ * Local Variables:
+ * eval: (c-set-style "gnu")
+ * End:
+ */
-- 
cgit 1.2.3-korg


From c3a814be9dc769be942ff8029c7b6eccd4b3af05 Mon Sep 17 00:00:00 2001
From: Damjan Marion 
Date: Tue, 28 Feb 2017 19:22:22 +0100
Subject: dpdk: be a plugin

Change-Id: I238258cdeb77035adc5e88903d824593d0a1da90
Signed-off-by: Damjan Marion 
---
 src/Makefile.am                                    |   21 -
 src/plugins/Makefile.am                            |    5 +
 src/plugins/dpdk.am                                |   50 +
 src/plugins/dpdk/api/dpdk.api                      |  103 +
 src/plugins/dpdk/api/dpdk_all_api_h.h              |   19 +
 src/plugins/dpdk/api/dpdk_msg_enum.h               |   31 +
 src/plugins/dpdk/api/dpdk_test.c                   |  397 ++++
 src/plugins/dpdk/buffer.c                          |  588 ++++++
 src/plugins/dpdk/device/cli.c                      | 2079 ++++++++++++++++++++
 src/plugins/dpdk/device/device.c                   |  852 ++++++++
 src/plugins/dpdk/device/dpdk.h                     |  490 +++++
 src/plugins/dpdk/device/dpdk_priv.h                |  135 ++
 src/plugins/dpdk/device/format.c                   |  754 +++++++
 src/plugins/dpdk/device/node.c                     |  674 +++++++
 src/plugins/dpdk/dir.dox                           |   27 +
 src/plugins/dpdk/hqos/hqos.c                       |  775 ++++++++
 src/plugins/dpdk/hqos/qos_doc.md                   |  411 ++++
 src/plugins/dpdk/init.c                            | 2074 +++++++++++++++++++
 src/plugins/dpdk/ipsec/cli.c                       |  154 ++
 src/plugins/dpdk/ipsec/crypto_node.c               |  215 ++
 src/plugins/dpdk/ipsec/dir.dox                     |   18 +
 src/plugins/dpdk/ipsec/dpdk_crypto_ipsec_doc.md    |   86 +
 src/plugins/dpdk/ipsec/esp.h                       |  249 +++
 src/plugins/dpdk/ipsec/esp_decrypt.c               |  594 ++++++
 src/plugins/dpdk/ipsec/esp_encrypt.c               |  609 ++++++
 src/plugins/dpdk/ipsec/ipsec.c                     |  430 ++++
 src/plugins/dpdk/ipsec/ipsec.h                     |  227 +++
 src/plugins/dpdk/main.c                            |   95 +
 src/plugins/dpdk/thread.c                          |   85 +
 src/vat/api_format.c                               |  320 ---
 src/vnet.am                                        |   41 +-
 src/vnet/devices/dpdk/buffer.c                     |  588 ------
 src/vnet/devices/dpdk/cli.c                        | 2079 --------------------
 src/vnet/devices/dpdk/device.c                     |  852 --------
 src/vnet/devices/dpdk/dir.dox                      |   27 -
 src/vnet/devices/dpdk/dpdk.api                     |  103 -
 src/vnet/devices/dpdk/dpdk.h                       |  487 -----
 src/vnet/devices/dpdk/dpdk_api.c                   |  246 ---
 src/vnet/devices/dpdk/dpdk_priv.h                  |  135 --
 src/vnet/devices/dpdk/format.c                     |  754 -------
 src/vnet/devices/dpdk/hqos.c                       |  775 --------
 src/vnet/devices/dpdk/init.c                       | 1801 -----------------
 src/vnet/devices/dpdk/ipsec/cli.c                  |  154 --
 src/vnet/devices/dpdk/ipsec/crypto_node.c          |  215 --
 src/vnet/devices/dpdk/ipsec/dir.dox                |   18 -
 .../devices/dpdk/ipsec/dpdk_crypto_ipsec_doc.md    |   86 -
 src/vnet/devices/dpdk/ipsec/esp.h                  |  249 ---
 src/vnet/devices/dpdk/ipsec/esp_decrypt.c          |  594 ------
 src/vnet/devices/dpdk/ipsec/esp_encrypt.c          |  609 ------
 src/vnet/devices/dpdk/ipsec/ipsec.c                |  430 ----
 src/vnet/devices/dpdk/ipsec/ipsec.h                |  227 ---
 src/vnet/devices/dpdk/main.c                       |   85 -
 src/vnet/devices/dpdk/node.c                       |  674 -------
 src/vnet/devices/dpdk/qos_doc.md                   |  411 ----
 src/vnet/devices/dpdk/thread.c                     |   85 -
 src/vnet/devices/virtio/vhost-user.h               |   11 -
 src/vnet/ipsec/ipsec_api.c                         |    4 -
 src/vnet/pg/input.c                                |   11 +-
 src/vnet/pg/stream.c                               |    5 +-
 src/vnet/replication.c                             |    6 +-
 src/vnet/vnet_all_api_h.h                          |    3 -
 src/vpp/api/custom_dump.c                          |   64 -
 src/vpp/api/gmon.c                                 |    3 +-
 src/vpp/api/vpe.api                                |    1 -
 src/vpp/app/l2t.c                                  |  562 ------
 src/vpp/app/l2t_l2.c                               |  267 ---
 66 files changed, 12241 insertions(+), 12988 deletions(-)
 create mode 100644 src/plugins/dpdk.am
 create mode 100644 src/plugins/dpdk/api/dpdk.api
 create mode 100644 src/plugins/dpdk/api/dpdk_all_api_h.h
 create mode 100644 src/plugins/dpdk/api/dpdk_msg_enum.h
 create mode 100644 src/plugins/dpdk/api/dpdk_test.c
 create mode 100644 src/plugins/dpdk/buffer.c
 create mode 100644 src/plugins/dpdk/device/cli.c
 create mode 100644 src/plugins/dpdk/device/device.c
 create mode 100644 src/plugins/dpdk/device/dpdk.h
 create mode 100644 src/plugins/dpdk/device/dpdk_priv.h
 create mode 100644 src/plugins/dpdk/device/format.c
 create mode 100644 src/plugins/dpdk/device/node.c
 create mode 100644 src/plugins/dpdk/dir.dox
 create mode 100644 src/plugins/dpdk/hqos/hqos.c
 create mode 100644 src/plugins/dpdk/hqos/qos_doc.md
 create mode 100755 src/plugins/dpdk/init.c
 create mode 100644 src/plugins/dpdk/ipsec/cli.c
 create mode 100644 src/plugins/dpdk/ipsec/crypto_node.c
 create mode 100644 src/plugins/dpdk/ipsec/dir.dox
 create mode 100644 src/plugins/dpdk/ipsec/dpdk_crypto_ipsec_doc.md
 create mode 100644 src/plugins/dpdk/ipsec/esp.h
 create mode 100644 src/plugins/dpdk/ipsec/esp_decrypt.c
 create mode 100644 src/plugins/dpdk/ipsec/esp_encrypt.c
 create mode 100644 src/plugins/dpdk/ipsec/ipsec.c
 create mode 100644 src/plugins/dpdk/ipsec/ipsec.h
 create mode 100644 src/plugins/dpdk/main.c
 create mode 100644 src/plugins/dpdk/thread.c
 delete mode 100644 src/vnet/devices/dpdk/buffer.c
 delete mode 100644 src/vnet/devices/dpdk/cli.c
 delete mode 100644 src/vnet/devices/dpdk/device.c
 delete mode 100644 src/vnet/devices/dpdk/dir.dox
 delete mode 100644 src/vnet/devices/dpdk/dpdk.api
 delete mode 100644 src/vnet/devices/dpdk/dpdk.h
 delete mode 100644 src/vnet/devices/dpdk/dpdk_api.c
 delete mode 100644 src/vnet/devices/dpdk/dpdk_priv.h
 delete mode 100644 src/vnet/devices/dpdk/format.c
 delete mode 100644 src/vnet/devices/dpdk/hqos.c
 delete mode 100755 src/vnet/devices/dpdk/init.c
 delete mode 100644 src/vnet/devices/dpdk/ipsec/cli.c
 delete mode 100644 src/vnet/devices/dpdk/ipsec/crypto_node.c
 delete mode 100644 src/vnet/devices/dpdk/ipsec/dir.dox
 delete mode 100644 src/vnet/devices/dpdk/ipsec/dpdk_crypto_ipsec_doc.md
 delete mode 100644 src/vnet/devices/dpdk/ipsec/esp.h
 delete mode 100644 src/vnet/devices/dpdk/ipsec/esp_decrypt.c
 delete mode 100644 src/vnet/devices/dpdk/ipsec/esp_encrypt.c
 delete mode 100644 src/vnet/devices/dpdk/ipsec/ipsec.c
 delete mode 100644 src/vnet/devices/dpdk/ipsec/ipsec.h
 delete mode 100644 src/vnet/devices/dpdk/main.c
 delete mode 100644 src/vnet/devices/dpdk/node.c
 delete mode 100644 src/vnet/devices/dpdk/qos_doc.md
 delete mode 100644 src/vnet/devices/dpdk/thread.c
 delete mode 100644 src/vpp/app/l2t.c
 delete mode 100644 src/vpp/app/l2t_l2.c

(limited to 'src/Makefile.am')

diff --git a/src/Makefile.am b/src/Makefile.am
index 641707ed..5daaa48e 100644
--- a/src/Makefile.am
+++ b/src/Makefile.am
@@ -38,27 +38,6 @@ install-data-local:
 	GREP_TIME=`echo $$GREP_TIME | awk '{print $$2}'` ; \
 	echo "Command list built, Time taken: $$GREP_TIME"
 
-###############################################################################
-# DPDK
-###############################################################################
-
-if WITH_DPDK
-if ENABLE_DPDK_SHARED
-DPDK_LD_FLAGS = -Wl,--whole-archive,-ldpdk,--no-whole-archive
-else
-DPDK_LD_FLAGS = -Wl,--whole-archive,-l:libdpdk.a,--no-whole-archive,-lm,-ldl
-endif
-if WITH_DPDK_CRYPTO_SW
-DPDK_LD_ADD = -lIPSec_MB -lisal_crypto
-endif
-if WITH_DPDK_MLX5_PMD
-DPDK_LD_FLAGS += -libverbs -lmlx5 -lnuma
-endif
-else
-DPDK_LD_FLAGS =
-DPDK_LD_ADD =
-endif
-
 ###############################################################################
 # Components
 ###############################################################################
diff --git a/src/plugins/Makefile.am b/src/plugins/Makefile.am
index 06b575d1..c8877899 100644
--- a/src/plugins/Makefile.am
+++ b/src/plugins/Makefile.am
@@ -24,6 +24,7 @@ vppplugins_LTLIBRARIES =
 vppapitestplugins_LTLIBRARIES =
 noinst_HEADERS =
 nobase_apiinclude_HEADERS =
+nobase_include_HEADERS =
 
 vppapitestpluginsdir = ${libdir}/vpp_api_test_plugins
 vpppluginsdir = ${libdir}/vpp_plugins
@@ -32,6 +33,10 @@ if ENABLE_ACL_PLUGIN
 include acl.am
 endif
 
+if WITH_DPDK
+include dpdk.am
+endif
+
 if ENABLE_FLOWPERPKT_PLUGIN
 include flowperpkt.am
 endif
diff --git a/src/plugins/dpdk.am b/src/plugins/dpdk.am
new file mode 100644
index 00000000..212bbb73
--- /dev/null
+++ b/src/plugins/dpdk.am
@@ -0,0 +1,50 @@
+# Copyright (c) 2016 Cisco Systems, Inc.
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at:
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+vppapitestplugins_LTLIBRARIES += dpdk_test_plugin.la
+vppplugins_LTLIBRARIES += dpdk_plugin.la
+
+dpdk_plugin_la_LDFLAGS = $(AM_LDFLAGS) -Wl,--whole-archive,-l:libdpdk.a,--no-whole-archive,-lm,-ldl
+
+dpdk_plugin_la_SOURCES =					\
+  dpdk/init.c							\
+  dpdk/main.c							\
+  dpdk/buffer.c							\
+  dpdk/thread.c							\
+  dpdk/device/cli.c						\
+  dpdk/device/dpdk_priv.h					\
+  dpdk/device/device.c						\
+  dpdk/device/format.c						\
+  dpdk/device/node.c						\
+  dpdk/hqos/hqos.c						\
+  dpdk/ipsec/esp_encrypt.c					\
+  dpdk/ipsec/esp_decrypt.c					\
+  dpdk/ipsec/crypto_node.c					\
+  dpdk/ipsec/cli.c						\
+  dpdk/ipsec/ipsec.c						\
+  dpdk/api/dpdk_plugin.api.h
+
+API_FILES += dpdk/api/dpdk.api
+
+nobase_include_HEADERS +=					\
+  dpdk/device/dpdk.h						\
+  dpdk/api/dpdk_all_api_h.h
+
+nobase_include_HEADERS +=			\
+ dpdk/ipsec/ipsec.h		\
+ dpdk/ipsec/esp.h
+
+dpdk_test_plugin_la_SOURCES = 					\
+  dpdk/api/dpdk_test.c dpdk/api/dpdk_plugin.api.h
+
+# vi:syntax=automake
diff --git a/src/plugins/dpdk/api/dpdk.api b/src/plugins/dpdk/api/dpdk.api
new file mode 100644
index 00000000..21215d45
--- /dev/null
+++ b/src/plugins/dpdk/api/dpdk.api
@@ -0,0 +1,103 @@
+/*
+ * Copyright (c) 2015-2016 Cisco and/or its affiliates.
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at:
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+/** \brief DPDK interface HQoS pipe profile set request
+    @param client_index - opaque cookie to identify the sender
+    @param context - sender context, to match reply w/ request
+    @param sw_if_index - the interface
+    @param subport - subport ID
+    @param pipe - pipe ID within its subport
+    @param profile - pipe profile ID
+*/
+define sw_interface_set_dpdk_hqos_pipe {
+    u32 client_index;
+    u32 context;
+    u32 sw_if_index;
+    u32 subport;
+    u32 pipe;
+    u32 profile;
+};
+
+/** \brief DPDK interface HQoS pipe profile set reply
+    @param context - sender context, to match reply w/ request
+    @param retval - request return code
+*/
+define sw_interface_set_dpdk_hqos_pipe_reply {
+    u32 context;
+    i32 retval;
+};
+
+/** \brief DPDK interface HQoS subport parameters set request
+    @param client_index - opaque cookie to identify the sender
+    @param context - sender context, to match reply w/ request
+    @param sw_if_index - the interface
+    @param subport - subport ID
+    @param tb_rate - subport token bucket rate (measured in bytes/second)
+    @param tb_size - subport token bucket size (measured in credits)
+    @param tc_rate - subport traffic class 0 .. 3 rates (measured in bytes/second)
+    @param tc_period - enforcement period for rates (measured in milliseconds)
+*/
+define sw_interface_set_dpdk_hqos_subport {
+    u32 client_index;
+    u32 context;
+    u32 sw_if_index;
+    u32 subport;
+    u32 tb_rate;
+    u32 tb_size;
+    u32 tc_rate[4];
+    u32 tc_period;
+};
+
+/** \brief DPDK interface HQoS subport parameters set reply
+    @param context - sender context, to match reply w/ request
+    @param retval - request return code
+*/
+define sw_interface_set_dpdk_hqos_subport_reply {
+    u32 context;
+    i32 retval;
+};
+
+/** \brief DPDK interface HQoS tctbl entry set request
+    @param client_index - opaque cookie to identify the sender
+    @param context - sender context, to match reply w/ request
+    @param sw_if_index - the interface
+    @param entry - entry index ID
+    @param tc - traffic class (0 .. 3)
+    @param queue - traffic class queue (0 .. 3)
+*/
+define sw_interface_set_dpdk_hqos_tctbl {
+    u32 client_index;
+    u32 context;
+    u32 sw_if_index;
+    u32 entry;
+    u32 tc;
+    u32 queue;
+};
+
+/** \brief DPDK interface HQoS tctbl entry set reply
+    @param context - sender context, to match reply w/ request
+    @param retval - request return code
+*/
+define sw_interface_set_dpdk_hqos_tctbl_reply {
+    u32 context;
+    i32 retval;
+};
+
+/*
+ * Local Variables:
+ * eval: (c-set-style "gnu")
+ * End:
+ */
+ 
\ No newline at end of file
diff --git a/src/plugins/dpdk/api/dpdk_all_api_h.h b/src/plugins/dpdk/api/dpdk_all_api_h.h
new file mode 100644
index 00000000..15eb98d6
--- /dev/null
+++ b/src/plugins/dpdk/api/dpdk_all_api_h.h
@@ -0,0 +1,19 @@
+
+/*
+ * dpdk_all_api_h.h - skeleton vpp engine plug-in api #include file
+ *
+ * Copyright (c)  
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at:
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+/* Include the generated file, see BUILT_SOURCES in Makefile.am */
+#include 
diff --git a/src/plugins/dpdk/api/dpdk_msg_enum.h b/src/plugins/dpdk/api/dpdk_msg_enum.h
new file mode 100644
index 00000000..952ce6ad
--- /dev/null
+++ b/src/plugins/dpdk/api/dpdk_msg_enum.h
@@ -0,0 +1,31 @@
+
+/*
+ * dpdk_msg_enum.h - skeleton vpp engine plug-in message enumeration
+ *
+ * Copyright (c)  
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at:
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+#ifndef included_dpdk_msg_enum_h
+#define included_dpdk_msg_enum_h
+
+#include 
+
+#define vl_msg_id(n,h) n,
+typedef enum {
+#include 
+    /* We'll want to know how many messages IDs we need... */
+    VL_MSG_FIRST_AVAILABLE,
+} vl_msg_id_t;
+#undef vl_msg_id
+
+#endif /* included_dpdk_msg_enum_h */
diff --git a/src/plugins/dpdk/api/dpdk_test.c b/src/plugins/dpdk/api/dpdk_test.c
new file mode 100644
index 00000000..9fe0f934
--- /dev/null
+++ b/src/plugins/dpdk/api/dpdk_test.c
@@ -0,0 +1,397 @@
+
+/*
+ * dpdk_test.c - skeleton vpp-api-test plug-in
+ *
+ * Copyright (c)  
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at:
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+#include 
+#include 
+#include 
+#include 
+#include 
+#include 
+
+uword unformat_sw_if_index (unformat_input_t * input, va_list * args);
+
+/* Declare message IDs */
+#include 
+
+/* define message structures */
+#define vl_typedefs
+#include 
+#undef vl_typedefs
+
+/* declare message handlers for each api */
+
+#define vl_endianfun             /* define message structures */
+#include 
+#undef vl_endianfun
+
+/* instantiate all the print functions we know about */
+#define vl_print(handle, ...)
+#define vl_printfun
+#include 
+#undef vl_printfun
+
+/* Get the API version number. */
+#define vl_api_version(n,v) static u32 api_version=(v);
+#include 
+#undef vl_api_version
+
+typedef struct {
+    /* API message ID base */
+    u16 msg_id_base;
+    vat_main_t *vat_main;
+} dpdk_test_main_t;
+
+dpdk_test_main_t dpdk_test_main;
+
+#define foreach_standard_reply_retval_handler         \
+_(sw_interface_set_dpdk_hqos_pipe_reply)              \
+_(sw_interface_set_dpdk_hqos_subport_reply)           \
+_(sw_interface_set_dpdk_hqos_tctbl_reply)
+
+#define _(n)                                          \
+    static void vl_api_##n##_t_handler                \
+    (vl_api_##n##_t * mp)                             \
+    {                                                 \
+        vat_main_t * vam = dpdk_test_main.vat_main;  \
+        i32 retval = ntohl(mp->retval);               \
+        if (vam->async_mode) {                        \
+            vam->async_errors += (retval < 0);        \
+        } else {                                      \
+            vam->retval = retval;                     \
+            vam->result_ready = 1;                    \
+        }                                             \
+    }
+foreach_standard_reply_retval_handler;
+#undef _
+
+/* 
+ * Table of message reply handlers, must include boilerplate handlers
+ * we just generated
+ */
+#define foreach_vpe_api_reply_msg                               \
+_(SW_INTERFACE_SET_DPDK_HQOS_PIPE_REPLY,                        \
+  sw_interface_set_dpdk_hqos_pipe_reply)                        \
+_(SW_INTERFACE_SET_DPDK_HQOS_SUBPORT_REPLY,                     \
+  sw_interface_set_dpdk_hqos_subport_reply)                     \
+_(SW_INTERFACE_SET_DPDK_HQOS_TCTBL_REPLY,                       \
+  sw_interface_set_dpdk_hqos_tctbl_reply)
+
+/* M: construct, but don't yet send a message */
+#define M(T,t)                                                  \
+do {                                                            \
+    vam->result_ready = 0;                                      \
+    mp = vl_msg_api_alloc(sizeof(*mp));                         \
+    memset (mp, 0, sizeof (*mp));                               \
+    mp->_vl_msg_id = ntohs (VL_API_##T + dm->msg_id_base);      \
+    mp->client_index = vam->my_client_index;                    \
+} while(0);
+
+#define M2(T,t,n)                                               \
+do {                                                            \
+    vam->result_ready = 0;                                      \
+    mp = vl_msg_api_alloc(sizeof(*mp)+(n));                     \
+    memset (mp, 0, sizeof (*mp));                               \
+    mp->_vl_msg_id = ntohs (VL_API_##T + dm->msg_id_base);      \
+    mp->client_index = vam->my_client_index;                    \
+} while(0);
+
+/* S: send a message */
+#define S (vl_msg_api_send_shmem (vam->vl_input_queue, (u8 *)&mp))
+
+/* W: wait for results, with timeout */
+#define W                                       \
+do {                                            \
+    timeout = vat_time_now (vam) + 1.0;         \
+                                                \
+    while (vat_time_now (vam) < timeout) {      \
+        if (vam->result_ready == 1) {           \
+            return (vam->retval);               \
+        }                                       \
+    }                                           \
+    return -99;                                 \
+} while(0);
+
+static int
+api_sw_interface_set_dpdk_hqos_pipe (vat_main_t * vam)
+{
+  dpdk_test_main_t * dm = &dpdk_test_main;
+  unformat_input_t *i = vam->input;
+  vl_api_sw_interface_set_dpdk_hqos_pipe_t *mp;
+  f64 timeout;
+  u32 sw_if_index;
+  u8 sw_if_index_set = 0;
+  u32 subport;
+  u8 subport_set = 0;
+  u32 pipe;
+  u8 pipe_set = 0;
+  u32 profile;
+  u8 profile_set = 0;
+
+  /* Parse args required to build the message */
+  while (unformat_check_input (i) != UNFORMAT_END_OF_INPUT)
+    {
+        if (unformat (i, "rx sw_if_index %u", &sw_if_index))
+  sw_if_index_set = 1;
+      else if (unformat (i, "subport %u", &subport))
+  subport_set = 1;
+      else if (unformat (i, "pipe %u", &pipe))
+  pipe_set = 1;
+      else if (unformat (i, "profile %u", &profile))
+  profile_set = 1;
+      else
+  break;
+    }
+
+  if (sw_if_index_set == 0)
+    {
+      errmsg ("missing interface name or sw_if_index");
+      return -99;
+    }
+
+  if (subport_set == 0)
+    {
+      errmsg ("missing subport ");
+      return -99;
+    }
+
+  if (pipe_set == 0)
+    {
+      errmsg ("missing pipe");
+      return -99;
+    }
+
+  if (profile_set == 0)
+    {
+      errmsg ("missing profile");
+      return -99;
+    }
+
+  M (SW_INTERFACE_SET_DPDK_HQOS_PIPE, sw_interface_set_dpdk_hqos_pipe);
+
+  mp->sw_if_index = ntohl (sw_if_index);
+  mp->subport = ntohl (subport);
+  mp->pipe = ntohl (pipe);
+  mp->profile = ntohl (profile);
+
+
+  S;
+  W;
+  /* NOTREACHED */
+  return 0;
+}
+
+static int
+api_sw_interface_set_dpdk_hqos_subport (vat_main_t * vam)
+{
+  dpdk_test_main_t * dm = &dpdk_test_main;
+  unformat_input_t *i = vam->input;
+  vl_api_sw_interface_set_dpdk_hqos_subport_t *mp;
+  f64 timeout;
+  u32 sw_if_index;
+  u8 sw_if_index_set = 0;
+  u32 subport;
+  u8 subport_set = 0;
+  u32 tb_rate = 1250000000; /* 10GbE */
+  u32 tb_size = 1000000;
+  u32 tc_rate[] = { 1250000000, 1250000000, 1250000000, 1250000000 };
+  u32 tc_period = 10;
+
+  /* Parse args required to build the message */
+  while (unformat_check_input (i) != UNFORMAT_END_OF_INPUT)
+    {
+      if (unformat (i, "rx sw_if_index %u", &sw_if_index))
+  sw_if_index_set = 1;
+      else if (unformat (i, "subport %u", &subport))
+  subport_set = 1;
+      else if (unformat (i, "rate %u", &tb_rate))
+  {
+    u32 tc_id;
+
+    for (tc_id = 0; tc_id < (sizeof (tc_rate) / sizeof (tc_rate[0]));
+         tc_id++)
+      tc_rate[tc_id] = tb_rate;
+  }
+      else if (unformat (i, "bktsize %u", &tb_size))
+  ;
+      else if (unformat (i, "tc0 %u", &tc_rate[0]))
+  ;
+      else if (unformat (i, "tc1 %u", &tc_rate[1]))
+  ;
+      else if (unformat (i, "tc2 %u", &tc_rate[2]))
+  ;
+      else if (unformat (i, "tc3 %u", &tc_rate[3]))
+  ;
+      else if (unformat (i, "period %u", &tc_period))
+  ;
+      else
+  break;
+    }
+
+  if (sw_if_index_set == 0)
+    {
+      errmsg ("missing interface name or sw_if_index");
+      return -99;
+    }
+
+  if (subport_set == 0)
+    {
+      errmsg ("missing subport ");
+      return -99;
+    }
+
+  M (SW_INTERFACE_SET_DPDK_HQOS_SUBPORT, sw_interface_set_dpdk_hqos_subport);
+
+  mp->sw_if_index = ntohl (sw_if_index);
+  mp->subport = ntohl (subport);
+  mp->tb_rate = ntohl (tb_rate);
+  mp->tb_size = ntohl (tb_size);
+  mp->tc_rate[0] = ntohl (tc_rate[0]);
+  mp->tc_rate[1] = ntohl (tc_rate[1]);
+  mp->tc_rate[2] = ntohl (tc_rate[2]);
+  mp->tc_rate[3] = ntohl (tc_rate[3]);
+  mp->tc_period = ntohl (tc_period);
+
+  S;
+  W;
+  /* NOTREACHED */
+  return 0;
+}
+
+static int
+api_sw_interface_set_dpdk_hqos_tctbl (vat_main_t * vam)
+{
+  dpdk_test_main_t * dm = &dpdk_test_main;
+  unformat_input_t *i = vam->input;
+  vl_api_sw_interface_set_dpdk_hqos_tctbl_t *mp;
+  f64 timeout;
+  u32 sw_if_index;
+  u8 sw_if_index_set = 0;
+  u8 entry_set = 0;
+  u8 tc_set = 0;
+  u8 queue_set = 0;
+  u32 entry, tc, queue;
+
+  /* Parse args required to build the message */
+  while (unformat_check_input (i) != UNFORMAT_END_OF_INPUT)
+    {
+      if (unformat (i, "rx sw_if_index %u", &sw_if_index))
+  sw_if_index_set = 1;
+      else if (unformat (i, "entry %d", &entry))
+  entry_set = 1;
+      else if (unformat (i, "tc %d", &tc))
+  tc_set = 1;
+      else if (unformat (i, "queue %d", &queue))
+  queue_set = 1;
+      else
+  break;
+    }
+
+  if (sw_if_index_set == 0)
+    {
+      errmsg ("missing interface name or sw_if_index");
+      return -99;
+    }
+
+  if (entry_set == 0)
+    {
+      errmsg ("missing entry ");
+      return -99;
+    }
+
+  if (tc_set == 0)
+    {
+      errmsg ("missing traffic class ");
+      return -99;
+    }
+
+  if (queue_set == 0)
+    {
+      errmsg ("missing queue ");
+      return -99;
+    }
+
+  M (SW_INTERFACE_SET_DPDK_HQOS_TCTBL, sw_interface_set_dpdk_hqos_tctbl);
+
+  mp->sw_if_index = ntohl (sw_if_index);
+  mp->entry = ntohl (entry);
+  mp->tc = ntohl (tc);
+  mp->queue = ntohl (queue);
+
+  S;
+  W;
+  /* NOTREACHED */
+  return 0;
+}
+
+/* 
+ * List of messages that the api test plugin sends,
+ * and that the data plane plugin processes
+ */
+#define foreach_vpe_api_msg                                               \
+_(sw_interface_set_dpdk_hqos_pipe,                                        \
+  "rx sw_if_index  subport  pipe \n"   \
+  "profile \n")                                               \
+_(sw_interface_set_dpdk_hqos_subport,                                     \
+  "rx sw_if_index  subport  [rate ]\n"       \
+  "[bktsize ] [tc0 ] [tc1 ] [tc2 ] [tc3 ] [period ]\n") \
+_(sw_interface_set_dpdk_hqos_tctbl,                                       \
+  "rx sw_if_index  entry  tc  queue \n")
+
+void vat_api_hookup (vat_main_t *vam)
+{
+  dpdk_test_main_t * dm __attribute__((unused)) = &dpdk_test_main;
+  /* Hook up handlers for replies from the data plane plug-in */
+#define _(N,n)                                                  \
+  vl_msg_api_set_handlers((VL_API_##N + dm->msg_id_base),       \
+                          #n,                                   \
+                          vl_api_##n##_t_handler,               \
+                          vl_noop_handler,                      \
+                          vl_api_##n##_t_endian,                \
+                          vl_api_##n##_t_print,                 \
+                          sizeof(vl_api_##n##_t), 1); 
+  foreach_vpe_api_reply_msg;
+#undef _
+
+  /* API messages we can send */
+#define _(n,h) hash_set_mem (vam->function_by_name, #n, api_##n);
+  foreach_vpe_api_msg;
+#undef _    
+    
+  /* Help strings */
+#define _(n,h) hash_set_mem (vam->help_by_name, #n, h);
+  foreach_vpe_api_msg;
+#undef _
+}
+
+clib_error_t * vat_plugin_register (vat_main_t *vam)
+{
+  dpdk_test_main_t * dm = &dpdk_test_main;
+  u8 * name;
+
+  dm->vat_main = vam;
+
+  /* Ask the vpp engine for the first assigned message-id */
+  name = format (0, "dpdk_%08x%c", api_version, 0);
+  dm->msg_id_base = vl_client_get_first_plugin_msg_id ((char *) name);
+
+  if (dm->msg_id_base != (u16) ~0)
+    vat_api_hookup (vam);
+  
+  vec_free(name);
+  
+  return 0;
+}
diff --git a/src/plugins/dpdk/buffer.c b/src/plugins/dpdk/buffer.c
new file mode 100644
index 00000000..2765c292
--- /dev/null
+++ b/src/plugins/dpdk/buffer.c
@@ -0,0 +1,588 @@
+/*
+ * Copyright (c) 2017 Cisco and/or its affiliates.
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at:
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+/*
+ * buffer.c: allocate/free network buffers.
+ *
+ * Copyright (c) 2008 Eliot Dresselhaus
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining
+ * a copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sublicense, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be
+ * included in all copies or substantial portions of the Software.
+ *
+ *  THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ *  EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ *  MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ *  NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE
+ *  LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
+ *  OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
+ *  WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+/**
+ * @file
+ *
+ * Allocate/free network buffers.
+ */
+
+#include 
+
+#include 
+#include 
+#include 
+#include 
+#include 
+#include 
+#include 
+#include 
+#include 
+#include 
+#include 
+#include 
+#include 
+#include 
+#include 
+#include 
+#include 
+#include 
+#include 
+#include 
+#include 
+#include 
+#include 
+#include 
+
+#include 
+#include 
+#include 
+#include 
+
+
+STATIC_ASSERT (VLIB_BUFFER_PRE_DATA_SIZE == RTE_PKTMBUF_HEADROOM,
+	       "VLIB_BUFFER_PRE_DATA_SIZE must be equal to RTE_PKTMBUF_HEADROOM");
+
+static_always_inline void
+dpdk_rte_pktmbuf_free (vlib_main_t * vm, vlib_buffer_t * b)
+{
+  vlib_buffer_t *hb = b;
+  struct rte_mbuf *mb;
+  u32 next, flags;
+  mb = rte_mbuf_from_vlib_buffer (hb);
+
+next:
+  flags = b->flags;
+  next = b->next_buffer;
+  mb = rte_mbuf_from_vlib_buffer (b);
+
+  if (PREDICT_FALSE (b->n_add_refs))
+    {
+      rte_mbuf_refcnt_update (mb, b->n_add_refs);
+      b->n_add_refs = 0;
+    }
+
+  rte_pktmbuf_free_seg (mb);
+
+  if (flags & VLIB_BUFFER_NEXT_PRESENT)
+    {
+      b = vlib_get_buffer (vm, next);
+      goto next;
+    }
+}
+
+static void
+del_free_list (vlib_main_t * vm, vlib_buffer_free_list_t * f)
+{
+  u32 i;
+  vlib_buffer_t *b;
+
+  for (i = 0; i < vec_len (f->buffers); i++)
+    {
+      b = vlib_get_buffer (vm, f->buffers[i]);
+      dpdk_rte_pktmbuf_free (vm, b);
+    }
+
+  vec_free (f->name);
+  vec_free (f->buffers);
+}
+
+/* Add buffer free list. */
+static void
+dpdk_buffer_delete_free_list (vlib_main_t * vm, u32 free_list_index)
+{
+  vlib_buffer_main_t *bm = vm->buffer_main;
+  vlib_buffer_free_list_t *f;
+  u32 merge_index;
+  int i;
+
+  ASSERT (os_get_cpu_number () == 0);
+
+  f = vlib_buffer_get_free_list (vm, free_list_index);
+
+  merge_index = vlib_buffer_get_free_list_with_size (vm, f->n_data_bytes);
+  if (merge_index != ~0 && merge_index != free_list_index)
+    {
+      vlib_buffer_merge_free_lists (pool_elt_at_index
+				    (bm->buffer_free_list_pool, merge_index),
+				    f);
+    }
+
+  del_free_list (vm, f);
+
+  /* Poison it. */
+  memset (f, 0xab, sizeof (f[0]));
+
+  pool_put (bm->buffer_free_list_pool, f);
+
+  for (i = 1; i < vec_len (vlib_mains); i++)
+    {
+      bm = vlib_mains[i]->buffer_main;
+      f = vlib_buffer_get_free_list (vlib_mains[i], free_list_index);;
+      memset (f, 0xab, sizeof (f[0]));
+      pool_put (bm->buffer_free_list_pool, f);
+    }
+}
+
+/* Make sure free list has at least given number of free buffers. */
+static uword
+fill_free_list (vlib_main_t * vm,
+		vlib_buffer_free_list_t * fl, uword min_free_buffers)
+{
+  dpdk_main_t *dm = &dpdk_main;
+  vlib_buffer_t *b0, *b1, *b2, *b3;
+  int n, i;
+  u32 bi0, bi1, bi2, bi3;
+  unsigned socket_id = rte_socket_id ();
+  struct rte_mempool *rmp = dm->pktmbuf_pools[socket_id];
+  struct rte_mbuf *mb0, *mb1, *mb2, *mb3;
+
+  /* Too early? */
+  if (PREDICT_FALSE (rmp == 0))
+    return 0;
+
+  /* Already have enough free buffers on free list? */
+  n = min_free_buffers - vec_len (fl->buffers);
+  if (n <= 0)
+    return min_free_buffers;
+
+  /* Always allocate round number of buffers. */
+  n = round_pow2 (n, CLIB_CACHE_LINE_BYTES / sizeof (u32));
+
+  /* Always allocate new buffers in reasonably large sized chunks. */
+  n = clib_max (n, fl->min_n_buffers_each_physmem_alloc);
+
+  vec_validate (vm->mbuf_alloc_list, n - 1);
+
+  if (rte_mempool_get_bulk (rmp, vm->mbuf_alloc_list, n) < 0)
+    return 0;
+
+  _vec_len (vm->mbuf_alloc_list) = n;
+
+  i = 0;
+
+  while (i < (n - 7))
+    {
+      vlib_prefetch_buffer_header (vlib_buffer_from_rte_mbuf
+				   (vm->mbuf_alloc_list[i + 4]), STORE);
+      vlib_prefetch_buffer_header (vlib_buffer_from_rte_mbuf
+				   (vm->mbuf_alloc_list[i + 5]), STORE);
+      vlib_prefetch_buffer_header (vlib_buffer_from_rte_mbuf
+				   (vm->mbuf_alloc_list[i + 6]), STORE);
+      vlib_prefetch_buffer_header (vlib_buffer_from_rte_mbuf
+				   (vm->mbuf_alloc_list[i + 7]), STORE);
+
+      mb0 = vm->mbuf_alloc_list[i];
+      mb1 = vm->mbuf_alloc_list[i + 1];
+      mb2 = vm->mbuf_alloc_list[i + 2];
+      mb3 = vm->mbuf_alloc_list[i + 3];
+
+      ASSERT (rte_mbuf_refcnt_read (mb0) == 0);
+      ASSERT (rte_mbuf_refcnt_read (mb1) == 0);
+      ASSERT (rte_mbuf_refcnt_read (mb2) == 0);
+      ASSERT (rte_mbuf_refcnt_read (mb3) == 0);
+
+      rte_mbuf_refcnt_set (mb0, 1);
+      rte_mbuf_refcnt_set (mb1, 1);
+      rte_mbuf_refcnt_set (mb2, 1);
+      rte_mbuf_refcnt_set (mb3, 1);
+
+      b0 = vlib_buffer_from_rte_mbuf (mb0);
+      b1 = vlib_buffer_from_rte_mbuf (mb1);
+      b2 = vlib_buffer_from_rte_mbuf (mb2);
+      b3 = vlib_buffer_from_rte_mbuf (mb3);
+
+      bi0 = vlib_get_buffer_index (vm, b0);
+      bi1 = vlib_get_buffer_index (vm, b1);
+      bi2 = vlib_get_buffer_index (vm, b2);
+      bi3 = vlib_get_buffer_index (vm, b3);
+
+      vec_add1_aligned (fl->buffers, bi0, CLIB_CACHE_LINE_BYTES);
+      vec_add1_aligned (fl->buffers, bi1, CLIB_CACHE_LINE_BYTES);
+      vec_add1_aligned (fl->buffers, bi2, CLIB_CACHE_LINE_BYTES);
+      vec_add1_aligned (fl->buffers, bi3, CLIB_CACHE_LINE_BYTES);
+
+      vlib_buffer_init_for_free_list (b0, fl);
+      vlib_buffer_init_for_free_list (b1, fl);
+      vlib_buffer_init_for_free_list (b2, fl);
+      vlib_buffer_init_for_free_list (b3, fl);
+
+      if (fl->buffer_init_function)
+	{
+	  fl->buffer_init_function (vm, fl, &bi0, 1);
+	  fl->buffer_init_function (vm, fl, &bi1, 1);
+	  fl->buffer_init_function (vm, fl, &bi2, 1);
+	  fl->buffer_init_function (vm, fl, &bi3, 1);
+	}
+      i += 4;
+    }
+
+  while (i < n)
+    {
+      mb0 = vm->mbuf_alloc_list[i];
+
+      ASSERT (rte_mbuf_refcnt_read (mb0) == 0);
+      rte_mbuf_refcnt_set (mb0, 1);
+
+      b0 = vlib_buffer_from_rte_mbuf (mb0);
+      bi0 = vlib_get_buffer_index (vm, b0);
+
+      vec_add1_aligned (fl->buffers, bi0, CLIB_CACHE_LINE_BYTES);
+
+      vlib_buffer_init_for_free_list (b0, fl);
+
+      if (fl->buffer_init_function)
+	fl->buffer_init_function (vm, fl, &bi0, 1);
+      i++;
+    }
+
+  fl->n_alloc += n;
+
+  return n;
+}
+
+static u32
+alloc_from_free_list (vlib_main_t * vm,
+		      vlib_buffer_free_list_t * free_list,
+		      u32 * alloc_buffers, u32 n_alloc_buffers)
+{
+  u32 *dst, *src;
+  uword len, n_filled;
+
+  dst = alloc_buffers;
+
+  n_filled = fill_free_list (vm, free_list, n_alloc_buffers);
+  if (n_filled == 0)
+    return 0;
+
+  len = vec_len (free_list->buffers);
+  ASSERT (len >= n_alloc_buffers);
+
+  src = free_list->buffers + len - n_alloc_buffers;
+  clib_memcpy (dst, src, n_alloc_buffers * sizeof (u32));
+
+  _vec_len (free_list->buffers) -= n_alloc_buffers;
+
+  return n_alloc_buffers;
+}
+
+/* Allocate a given number of buffers into given array.
+   Returns number actually allocated which will be either zero or
+   number requested. */
+u32
+dpdk_buffer_alloc (vlib_main_t * vm, u32 * buffers, u32 n_buffers)
+{
+  vlib_buffer_main_t *bm = vm->buffer_main;
+
+  return alloc_from_free_list
+    (vm,
+     pool_elt_at_index (bm->buffer_free_list_pool,
+			VLIB_BUFFER_DEFAULT_FREE_LIST_INDEX),
+     buffers, n_buffers);
+}
+
+
+u32
+dpdk_buffer_alloc_from_free_list (vlib_main_t * vm,
+				  u32 * buffers,
+				  u32 n_buffers, u32 free_list_index)
+{
+  vlib_buffer_main_t *bm = vm->buffer_main;
+  vlib_buffer_free_list_t *f;
+  f = pool_elt_at_index (bm->buffer_free_list_pool, free_list_index);
+  return alloc_from_free_list (vm, f, buffers, n_buffers);
+}
+
+static_always_inline void
+vlib_buffer_free_inline (vlib_main_t * vm,
+			 u32 * buffers, u32 n_buffers, u32 follow_buffer_next)
+{
+  vlib_buffer_main_t *bm = vm->buffer_main;
+  vlib_buffer_free_list_t *fl;
+  u32 fi;
+  int i;
+  u32 (*cb) (vlib_main_t * vm, u32 * buffers, u32 n_buffers,
+	     u32 follow_buffer_next);
+
+  cb = bm->buffer_free_callback;
+
+  if (PREDICT_FALSE (cb != 0))
+    n_buffers = (*cb) (vm, buffers, n_buffers, follow_buffer_next);
+
+  if (!n_buffers)
+    return;
+
+  for (i = 0; i < n_buffers; i++)
+    {
+      vlib_buffer_t *b;
+
+      b = vlib_get_buffer (vm, buffers[i]);
+
+      fl = vlib_buffer_get_buffer_free_list (vm, b, &fi);
+
+      /* The only current use of this callback: multicast recycle */
+      if (PREDICT_FALSE (fl->buffers_added_to_freelist_function != 0))
+	{
+	  int j;
+
+	  vlib_buffer_add_to_free_list
+	    (vm, fl, buffers[i], (b->flags & VLIB_BUFFER_RECYCLE) == 0);
+
+	  for (j = 0; j < vec_len (bm->announce_list); j++)
+	    {
+	      if (fl == bm->announce_list[j])
+		goto already_announced;
+	    }
+	  vec_add1 (bm->announce_list, fl);
+	already_announced:
+	  ;
+	}
+      else
+	{
+	  if (PREDICT_TRUE ((b->flags & VLIB_BUFFER_RECYCLE) == 0))
+	    dpdk_rte_pktmbuf_free (vm, b);
+	}
+    }
+  if (vec_len (bm->announce_list))
+    {
+      vlib_buffer_free_list_t *fl;
+      for (i = 0; i < vec_len (bm->announce_list); i++)
+	{
+	  fl = bm->announce_list[i];
+	  fl->buffers_added_to_freelist_function (vm, fl);
+	}
+      _vec_len (bm->announce_list) = 0;
+    }
+}
+
+static void
+dpdk_buffer_free (vlib_main_t * vm, u32 * buffers, u32 n_buffers)
+{
+  vlib_buffer_free_inline (vm, buffers, n_buffers,	/* follow_buffer_next */
+			   1);
+}
+
+static void
+dpdk_buffer_free_no_next (vlib_main_t * vm, u32 * buffers, u32 n_buffers)
+{
+  vlib_buffer_free_inline (vm, buffers, n_buffers,	/* follow_buffer_next */
+			   0);
+}
+
+static void
+dpdk_packet_template_init (vlib_main_t * vm,
+			   void *vt,
+			   void *packet_data,
+			   uword n_packet_data_bytes,
+			   uword min_n_buffers_each_physmem_alloc, u8 * name)
+{
+  vlib_packet_template_t *t = (vlib_packet_template_t *) vt;
+
+  vlib_worker_thread_barrier_sync (vm);
+  memset (t, 0, sizeof (t[0]));
+
+  vec_add (t->packet_data, packet_data, n_packet_data_bytes);
+
+  vlib_worker_thread_barrier_release (vm);
+}
+
+clib_error_t *
+vlib_buffer_pool_create (vlib_main_t * vm, unsigned num_mbufs,
+			 unsigned socket_id)
+{
+  dpdk_main_t *dm = &dpdk_main;
+  vlib_physmem_main_t *vpm = &vm->physmem_main;
+  struct rte_mempool *rmp;
+  int i;
+
+  vec_validate_aligned (dm->pktmbuf_pools, socket_id, CLIB_CACHE_LINE_BYTES);
+
+  /* pool already exists, nothing to do */
+  if (dm->pktmbuf_pools[socket_id])
+    return 0;
+
+  u8 *pool_name = format (0, "mbuf_pool_socket%u%c", socket_id, 0);
+
+  rmp = rte_pktmbuf_pool_create ((char *) pool_name,	/* pool name */
+				 num_mbufs,	/* number of mbufs */
+				 512,	/* cache size */
+				 VLIB_BUFFER_HDR_SIZE,	/* priv size */
+				 VLIB_BUFFER_PRE_DATA_SIZE + VLIB_BUFFER_DATA_SIZE,	/* dataroom size */
+				 socket_id);	/* cpu socket */
+
+  if (rmp)
+    {
+      {
+	uword this_pool_end;
+	uword this_pool_start;
+	uword this_pool_size;
+	uword save_vpm_start, save_vpm_end, save_vpm_size;
+	struct rte_mempool_memhdr *memhdr;
+
+	this_pool_start = ~0ULL;
+	this_pool_end = 0LL;
+
+	STAILQ_FOREACH (memhdr, &rmp->mem_list, next)
+	{
+	  if (((uword) (memhdr->addr + memhdr->len)) > this_pool_end)
+	    this_pool_end = (uword) (memhdr->addr + memhdr->len);
+	  if (((uword) memhdr->addr) < this_pool_start)
+	    this_pool_start = (uword) (memhdr->addr);
+	}
+	ASSERT (this_pool_start < ~0ULL && this_pool_end > 0);
+	this_pool_size = this_pool_end - this_pool_start;
+
+	if (CLIB_DEBUG > 1)
+	  {
+	    clib_warning ("%s: pool start %llx pool end %llx pool size %lld",
+			  pool_name, this_pool_start, this_pool_end,
+			  this_pool_size);
+	    clib_warning
+	      ("before: virtual.start %llx virtual.end %llx virtual.size %lld",
+	       vpm->virtual.start, vpm->virtual.end, vpm->virtual.size);
+	  }
+
+	save_vpm_start = vpm->virtual.start;
+	save_vpm_end = vpm->virtual.end;
+	save_vpm_size = vpm->virtual.size;
+
+	if ((this_pool_start < vpm->virtual.start) || vpm->virtual.start == 0)
+	  vpm->virtual.start = this_pool_start;
+	if (this_pool_end > vpm->virtual.end)
+	  vpm->virtual.end = this_pool_end;
+
+	vpm->virtual.size = vpm->virtual.end - vpm->virtual.start;
+
+	if (CLIB_DEBUG > 1)
+	  {
+	    clib_warning
+	      ("after: virtual.start %llx virtual.end %llx virtual.size %lld",
+	       vpm->virtual.start, vpm->virtual.end, vpm->virtual.size);
+	  }
+
+	/* check if fits into buffer index range */
+	if ((u64) vpm->virtual.size >
+	    ((u64) 1 << (32 + CLIB_LOG2_CACHE_LINE_BYTES)))
+	  {
+	    clib_warning ("physmem: virtual size out of range!");
+	    vpm->virtual.start = save_vpm_start;
+	    vpm->virtual.end = save_vpm_end;
+	    vpm->virtual.size = save_vpm_size;
+	    rmp = 0;
+	  }
+      }
+      if (rmp)
+	{
+	  dm->pktmbuf_pools[socket_id] = rmp;
+	  vec_free (pool_name);
+	  return 0;
+	}
+    }
+
+  vec_free (pool_name);
+
+  /* no usable pool for this socket, try to use pool from another one */
+  for (i = 0; i < vec_len (dm->pktmbuf_pools); i++)
+    {
+      if (dm->pktmbuf_pools[i])
+	{
+	  clib_warning
+	    ("WARNING: Failed to allocate mempool for CPU socket %u. "
+	     "Threads running on socket %u will use socket %u mempool.",
+	     socket_id, socket_id, i);
+	  dm->pktmbuf_pools[socket_id] = dm->pktmbuf_pools[i];
+	  return 0;
+	}
+    }
+
+  return clib_error_return (0, "failed to allocate mempool on socket %u",
+			    socket_id);
+}
+
+#if CLIB_DEBUG > 0
+
+u32 *vlib_buffer_state_validation_lock;
+uword *vlib_buffer_state_validation_hash;
+void *vlib_buffer_state_heap;
+
+static clib_error_t *
+buffer_state_validation_init (vlib_main_t * vm)
+{
+  void *oldheap;
+
+  vlib_buffer_state_heap = mheap_alloc (0, 10 << 20);
+
+  oldheap = clib_mem_set_heap (vlib_buffer_state_heap);
+
+  vlib_buffer_state_validation_hash = hash_create (0, sizeof (uword));
+  vec_validate_aligned (vlib_buffer_state_validation_lock, 0,
+			CLIB_CACHE_LINE_BYTES);
+  clib_mem_set_heap (oldheap);
+  return 0;
+}
+
+VLIB_INIT_FUNCTION (buffer_state_validation_init);
+#endif
+
+static vlib_buffer_callbacks_t callbacks = {
+  .vlib_buffer_alloc_cb = &dpdk_buffer_alloc,
+  .vlib_buffer_alloc_from_free_list_cb = &dpdk_buffer_alloc_from_free_list,
+  .vlib_buffer_free_cb = &dpdk_buffer_free,
+  .vlib_buffer_free_no_next_cb = &dpdk_buffer_free_no_next,
+  .vlib_packet_template_init_cb = &dpdk_packet_template_init,
+  .vlib_buffer_delete_free_list_cb = &dpdk_buffer_delete_free_list,
+};
+
+static clib_error_t *
+dpdk_buffer_init (vlib_main_t * vm)
+{
+  vlib_buffer_cb_register (vm, &callbacks);
+  return 0;
+}
+
+VLIB_INIT_FUNCTION (dpdk_buffer_init);
+
+/** @endcond */
+/*
+ * fd.io coding-style-patch-verification: ON
+ *
+ * Local Variables:
+ * eval: (c-set-style "gnu")
+ * End:
+ */
diff --git a/src/plugins/dpdk/device/cli.c b/src/plugins/dpdk/device/cli.c
new file mode 100644
index 00000000..d2def2fc
--- /dev/null
+++ b/src/plugins/dpdk/device/cli.c
@@ -0,0 +1,2079 @@
+/*
+ * Copyright (c) 2015 Cisco and/or its affiliates.
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at:
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+#include 
+#include 
+#include 
+#include 
+#include 
+
+#include 
+#include 
+#include 
+#include 
+
+#include 
+
+/**
+ * @file
+ * @brief CLI for DPDK Abstraction Layer and pcap Tx Trace.
+ *
+ * This file contains the source code for CLI for DPDK
+ * Abstraction Layer and pcap Tx Trace.
+ */
+
+
+static clib_error_t *
+get_hqos (u32 hw_if_index, u32 subport_id, dpdk_device_t ** xd,
+	  dpdk_device_config_t ** devconf)
+{
+  dpdk_main_t *dm = &dpdk_main;
+  vnet_hw_interface_t *hw;
+  struct rte_eth_dev_info dev_info;
+  uword *p = 0;
+  clib_error_t *error = NULL;
+
+
+  if (hw_if_index == (u32) ~ 0)
+    {
+      error = clib_error_return (0, "please specify valid interface name");
+      goto done;
+    }
+
+  if (subport_id != 0)
+    {
+      error = clib_error_return (0, "Invalid subport");
+      goto done;
+    }
+
+  hw = vnet_get_hw_interface (dm->vnet_main, hw_if_index);
+  *xd = vec_elt_at_index (dm->devices, hw->dev_instance);
+
+  rte_eth_dev_info_get ((*xd)->device_index, &dev_info);
+  if (dev_info.pci_dev)
+    {				/* bonded interface has no pci info */
+      vlib_pci_addr_t pci_addr;
+
+      pci_addr.domain = dev_info.pci_dev->addr.domain;
+      pci_addr.bus = dev_info.pci_dev->addr.bus;
+      pci_addr.slot = dev_info.pci_dev->addr.devid;
+      pci_addr.function = dev_info.pci_dev->addr.function;
+
+      p =
+	hash_get (dm->conf->device_config_index_by_pci_addr, pci_addr.as_u32);
+    }
+
+  if (p)
+    (*devconf) = pool_elt_at_index (dm->conf->dev_confs, p[0]);
+  else
+    (*devconf) = &dm->conf->default_devconf;
+
+done:
+  return error;
+}
+
+static clib_error_t *
+pcap_trace_command_fn (vlib_main_t * vm,
+		       unformat_input_t * input, vlib_cli_command_t * cmd)
+{
+#define PCAP_DEF_PKT_TO_CAPTURE (100)
+
+  unformat_input_t _line_input, *line_input = &_line_input;
+  dpdk_main_t *dm = &dpdk_main;
+  u8 *filename;
+  u8 *chroot_filename = 0;
+  u32 max = 0;
+  int enabled = 0;
+  int errorFlag = 0;
+  clib_error_t *error = 0;
+
+  /* Get a line of input. */
+  if (!unformat_user (input, unformat_line_input, line_input))
+    return 0;
+
+  while (unformat_check_input (line_input) != UNFORMAT_END_OF_INPUT)
+    {
+      if (unformat (line_input, "on"))
+	{
+	  if (dm->tx_pcap_enable == 0)
+	    {
+	      enabled = 1;
+	    }
+	  else
+	    {
+	      vlib_cli_output (vm, "pcap tx capture already on...");
+	      errorFlag = 1;
+	      break;
+	    }
+	}
+      else if (unformat (line_input, "off"))
+	{
+	  if (dm->tx_pcap_enable)
+	    {
+	      vlib_cli_output (vm, "captured %d pkts...",
+			       dm->pcap_main.n_packets_captured + 1);
+	      if (dm->pcap_main.n_packets_captured)
+		{
+		  dm->pcap_main.n_packets_to_capture =
+		    dm->pcap_main.n_packets_captured;
+		  error = pcap_write (&dm->pcap_main);
+		  if (error)
+		    clib_error_report (error);
+		  else
+		    vlib_cli_output (vm, "saved to %s...", dm->pcap_filename);
+		}
+
+	      dm->tx_pcap_enable = 0;
+	    }
+	  else
+	    {
+	      vlib_cli_output (vm, "pcap tx capture already off...");
+	      errorFlag = 1;
+	      break;
+	    }
+	}
+      else if (unformat (line_input, "max %d", &max))
+	{
+	  if (dm->tx_pcap_enable)
+	    {
+	      vlib_cli_output (vm,
+			       "can't change max value while pcap tx capture active...");
+	      errorFlag = 1;
+	      break;
+	    }
+	}
+      else if (unformat (line_input, "intfc %U",
+			 unformat_vnet_sw_interface, dm->vnet_main,
+			 &dm->pcap_sw_if_index))
+	;
+
+      else if (unformat (line_input, "intfc any"))
+	{
+	  dm->pcap_sw_if_index = 0;
+	}
+      else if (unformat (line_input, "file %s", &filename))
+	{
+	  if (dm->tx_pcap_enable)
+	    {
+	      vlib_cli_output (vm,
+			       "can't change file while pcap tx capture active...");
+	      errorFlag = 1;
+	      break;
+	    }
+
+	  /* Brain-police user path input */
+	  if (strstr ((char *) filename, "..")
+	      || index ((char *) filename, '/'))
+	    {
+	      vlib_cli_output (vm, "illegal characters in filename '%s'",
+			       filename);
+	      vlib_cli_output (vm,
+			       "Hint: Only filename, do not enter directory structure.");
+	      vec_free (filename);
+	      errorFlag = 1;
+	      break;
+	    }
+
+	  chroot_filename = format (0, "/tmp/%s%c", filename, 0);
+	  vec_free (filename);
+	}
+      else if (unformat (line_input, "status"))
+	{
+	  if (dm->pcap_sw_if_index == 0)
+	    {
+	      vlib_cli_output (vm, "max is %d for any interface to file %s",
+			       dm->
+			       pcap_pkts_to_capture ? dm->pcap_pkts_to_capture
+			       : PCAP_DEF_PKT_TO_CAPTURE,
+			       dm->
+			       pcap_filename ? dm->pcap_filename : (u8 *)
+			       "/tmp/vpe.pcap");
+	    }
+	  else
+	    {
+	      vlib_cli_output (vm, "max is %d for interface %U to file %s",
+			       dm->
+			       pcap_pkts_to_capture ? dm->pcap_pkts_to_capture
+			       : PCAP_DEF_PKT_TO_CAPTURE,
+			       format_vnet_sw_if_index_name, dm->vnet_main,
+			       dm->pcap_sw_if_index,
+			       dm->
+			       pcap_filename ? dm->pcap_filename : (u8 *)
+			       "/tmp/vpe.pcap");
+	    }
+
+	  if (dm->tx_pcap_enable == 0)
+	    {
+	      vlib_cli_output (vm, "pcap tx capture is off...");
+	    }
+	  else
+	    {
+	      vlib_cli_output (vm, "pcap tx capture is on: %d of %d pkts...",
+			       dm->pcap_main.n_packets_captured,
+			       dm->pcap_main.n_packets_to_capture);
+	    }
+	  break;
+	}
+
+      else
+	{
+	  error = clib_error_return (0, "unknown input `%U'",
+				     format_unformat_error, line_input);
+	  errorFlag = 1;
+	  break;
+	}
+    }
+  unformat_free (line_input);
+
+
+  if (errorFlag == 0)
+    {
+      /* Since no error, save configured values. */
+      if (chroot_filename)
+	{
+	  if (dm->pcap_filename)
+	    vec_free (dm->pcap_filename);
+	  vec_add1 (chroot_filename, 0);
+	  dm->pcap_filename = chroot_filename;
+	}
+
+      if (max)
+	dm->pcap_pkts_to_capture = max;
+
+
+      if (enabled)
+	{
+	  if (dm->pcap_filename == 0)
+	    dm->pcap_filename = format (0, "/tmp/vpe.pcap%c", 0);
+
+	  memset (&dm->pcap_main, 0, sizeof (dm->pcap_main));
+	  dm->pcap_main.file_name = (char *) dm->pcap_filename;
+	  dm->pcap_main.n_packets_to_capture = PCAP_DEF_PKT_TO_CAPTURE;
+	  if (dm->pcap_pkts_to_capture)
+	    dm->pcap_main.n_packets_to_capture = dm->pcap_pkts_to_capture;
+
+	  dm->pcap_main.packet_type = PCAP_PACKET_TYPE_ethernet;
+	  dm->tx_pcap_enable = 1;
+	  vlib_cli_output (vm, "pcap tx capture on...");
+	}
+    }
+  else if (chroot_filename)
+    vec_free (chroot_filename);
+
+
+  return error;
+}
+
+/*?
+ * This command is used to start or stop a packet capture, or show
+ * the status of packet capture.
+ *
+ * This command has the following optional parameters:
+ *
+ * - on|off - Used to start or stop a packet capture.
+ *
+ * - max  - Depth of local buffer. Once 'nn' number
+ *   of packets have been received, buffer is flushed to file. Once another
+ *   'nn' number of packets have been received, buffer is flushed
+ *   to file, overwriting previous write. If not entered, value defaults
+ *   to 100. Can only be updated if packet capture is off.
+ *
+ * - intfc |any - Used to specify a given interface,
+ *   or use 'any' to run packet capture on all interfaces.
+ *   'any' is the default if not provided. Settings from a previous
+ *   packet capture are preserved, so 'any' can be used to reset
+ *   the interface setting.
+ *
+ * - file  - Used to specify the output filename. The file will
+ *   be placed in the '/tmp' directory, so only the filename is
+ *   supported. Directory should not be entered. If file already exists, file
+ *   will be overwritten. If no filename is provided, '/tmp/vpe.pcap'
+ *   will be used. Can only be updated if packet capture is off.
+ *
+ * - status - Displays the current status and configured attributes
+ *   associated with a packet capture. If packet capture is in progress,
+ *   'status' also will return the number of packets currently in
+ *   the local buffer. All additional attributes entered on command line
+ *   with 'status' will be ingnored and not applied.
+ *
+ * @cliexpar
+ * Example of how to display the status of a tx packet capture when off:
+ * @cliexstart{pcap tx trace status}
+ * max is 100, for any interface to file /tmp/vpe.pcap
+ * pcap tx capture is off...
+ * @cliexend
+ * Example of how to start a tx packet capture:
+ * @cliexstart{pcap tx trace on max 35 intfc GigabitEthernet0/8/0 file vppTest.pcap}
+ * pcap tx capture on...
+ * @cliexend
+ * Example of how to display the status of a tx packet capture in progress:
+ * @cliexstart{pcap tx trace status}
+ * max is 35, for interface GigabitEthernet0/8/0 to file /tmp/vppTest.pcap
+ * pcap tx capture is on: 20 of 35 pkts...
+ * @cliexend
+ * Example of how to stop a tx packet capture:
+ * @cliexstart{vppctl pcap tx trace off}
+ * captured 21 pkts...
+ * saved to /tmp/vppTest.pcap...
+ * @cliexend
+?*/
+/* *INDENT-OFF* */
+VLIB_CLI_COMMAND (pcap_trace_command, static) = {
+    .path = "pcap tx trace",
+    .short_help =
+    "pcap tx trace [on|off] [max ] [intfc |any] [file ] [status]",
+    .function = pcap_trace_command_fn,
+};
+/* *INDENT-ON* */
+
+
+static clib_error_t *
+show_dpdk_buffer (vlib_main_t * vm, unformat_input_t * input,
+		  vlib_cli_command_t * cmd)
+{
+  struct rte_mempool *rmp;
+  int i;
+
+  for (i = 0; i < vec_len (dpdk_main.pktmbuf_pools); i++)
+    {
+      rmp = dpdk_main.pktmbuf_pools[i];
+      if (rmp)
+	{
+	  unsigned count = rte_mempool_avail_count (rmp);
+	  unsigned free_count = rte_mempool_in_use_count (rmp);
+
+	  vlib_cli_output (vm,
+			   "name=\"%s\"  available = %7d allocated = %7d total = %7d\n",
+			   rmp->name, (u32) count, (u32) free_count,
+			   (u32) (count + free_count));
+	}
+      else
+	{
+	  vlib_cli_output (vm, "rte_mempool is NULL (!)\n");
+	}
+    }
+  return 0;
+}
+
+/*?
+ * This command displays statistics of each DPDK mempool.
+ *
+ * @cliexpar
+ * Example of how to display DPDK buffer data:
+ * @cliexstart{show dpdk buffer}
+ * name="mbuf_pool_socket0"  available =   15104 allocated =    1280 total =   16384
+ * @cliexend
+?*/
+/* *INDENT-OFF* */
+VLIB_CLI_COMMAND (cmd_show_dpdk_bufferr,static) = {
+    .path = "show dpdk buffer",
+    .short_help = "show dpdk buffer",
+    .function = show_dpdk_buffer,
+    .is_mp_safe = 1,
+};
+/* *INDENT-ON* */
+
+static clib_error_t *
+test_dpdk_buffer (vlib_main_t * vm, unformat_input_t * input,
+		  vlib_cli_command_t * cmd)
+{
+  static u32 *allocated_buffers;
+  u32 n_alloc = 0;
+  u32 n_free = 0;
+  u32 first, actual_alloc;
+
+  while (unformat_check_input (input) != UNFORMAT_END_OF_INPUT)
+    {
+      if (unformat (input, "allocate %d", &n_alloc))
+	;
+      else if (unformat (input, "free %d", &n_free))
+	;
+      else
+	break;
+    }
+
+  if (n_free)
+    {
+      if (vec_len (allocated_buffers) < n_free)
+	return clib_error_return (0, "Can't free %d, only %d allocated",
+				  n_free, vec_len (allocated_buffers));
+
+      first = vec_len (allocated_buffers) - n_free;
+      vlib_buffer_free (vm, allocated_buffers + first, n_free);
+      _vec_len (allocated_buffers) = first;
+    }
+  if (n_alloc)
+    {
+      first = vec_len (allocated_buffers);
+      vec_validate (allocated_buffers,
+		    vec_len (allocated_buffers) + n_alloc - 1);
+
+      actual_alloc = vlib_buffer_alloc (vm, allocated_buffers + first,
+					n_alloc);
+      _vec_len (allocated_buffers) = first + actual_alloc;
+
+      if (actual_alloc < n_alloc)
+	vlib_cli_output (vm, "WARNING: only allocated %d buffers",
+			 actual_alloc);
+    }
+
+  vlib_cli_output (vm, "Currently %d buffers allocated",
+		   vec_len (allocated_buffers));
+
+  if (allocated_buffers && vec_len (allocated_buffers) == 0)
+    vec_free (allocated_buffers);
+
+  return 0;
+}
+
+/*?
+ * This command tests the allocation and freeing of DPDK buffers.
+ * If both 'allocate' and 'free' are entered on the
+ * same command, the 'free' is executed first. If no
+ * parameters are provided, this command display how many DPDK buffers
+ * the test command has allocated.
+ *
+ * @cliexpar
+ * @parblock
+ *
+ * Example of how to display how many DPDK buffer test command has allcoated:
+ * @cliexstart{test dpdk buffer}
+ * Currently 0 buffers allocated
+ * @cliexend
+ *
+ * Example of how to allocate DPDK buffers using the test command:
+ * @cliexstart{test dpdk buffer allocate 10}
+ * Currently 10 buffers allocated
+ * @cliexend
+ *
+ * Example of how to free DPDK buffers allocated by the test command:
+ * @cliexstart{test dpdk buffer free 10}
+ * Currently 0 buffers allocated
+ * @cliexend
+ * @endparblock
+?*/
+/* *INDENT-OFF* */
+VLIB_CLI_COMMAND (cmd_test_dpdk_buffer,static) = {
+    .path = "test dpdk buffer",
+    .short_help = "test dpdk buffer [allocate ] [free ]",
+    .function = test_dpdk_buffer,
+    .is_mp_safe = 1,
+};
+/* *INDENT-ON* */
+
+static clib_error_t *
+set_dpdk_if_desc (vlib_main_t * vm, unformat_input_t * input,
+		  vlib_cli_command_t * cmd)
+{
+  unformat_input_t _line_input, *line_input = &_line_input;
+  dpdk_main_t *dm = &dpdk_main;
+  vnet_hw_interface_t *hw;
+  dpdk_device_t *xd;
+  u32 hw_if_index = (u32) ~ 0;
+  u32 nb_rx_desc = (u32) ~ 0;
+  u32 nb_tx_desc = (u32) ~ 0;
+  clib_error_t *error = NULL;
+
+  if (!unformat_user (input, unformat_line_input, line_input))
+    return 0;
+
+  while (unformat_check_input (line_input) != UNFORMAT_END_OF_INPUT)
+    {
+      if (unformat
+	  (line_input, "%U", unformat_vnet_hw_interface, dm->vnet_main,
+	   &hw_if_index))
+	;
+      else if (unformat (line_input, "tx %d", &nb_tx_desc))
+	;
+      else if (unformat (line_input, "rx %d", &nb_rx_desc))
+	;
+      else
+	{
+	  error = clib_error_return (0, "parse error: '%U'",
+				     format_unformat_error, line_input);
+	  goto done;
+	}
+    }
+
+  if (hw_if_index == (u32) ~ 0)
+    {
+      error = clib_error_return (0, "please specify valid interface name");
+      goto done;
+    }
+
+  hw = vnet_get_hw_interface (dm->vnet_main, hw_if_index);
+  xd = vec_elt_at_index (dm->devices, hw->dev_instance);
+
+  if ((xd->flags & DPDK_DEVICE_FLAG_PMD) == 0)
+    {
+      error =
+	clib_error_return (0,
+			   "number of descriptors can be set only for "
+			   "physical devices");
+      goto done;
+    }
+
+  if ((nb_rx_desc == (u32) ~ 0 || nb_rx_desc == xd->nb_rx_desc) &&
+      (nb_tx_desc == (u32) ~ 0 || nb_tx_desc == xd->nb_tx_desc))
+    {
+      error = clib_error_return (0, "nothing changed");
+      goto done;
+    }
+
+  if (nb_rx_desc != (u32) ~ 0)
+    xd->nb_rx_desc = nb_rx_desc;
+
+  if (nb_tx_desc != (u32) ~ 0)
+    xd->nb_tx_desc = nb_tx_desc;
+
+  error = dpdk_port_setup (dm, xd);
+
+done:
+  unformat_free (line_input);
+
+  return error;
+}
+
+/*?
+ * This command sets the number of DPDK 'rx' and
+ * 'tx' descriptors for the given physical interface. Use
+ * the command 'show hardware-interface' to display the
+ * current descriptor allocation.
+ *
+ * @cliexpar
+ * Example of how to set the DPDK interface descriptors:
+ * @cliexcmd{set dpdk interface descriptors GigabitEthernet0/8/0 rx 512 tx 512}
+?*/
+/* *INDENT-OFF* */
+VLIB_CLI_COMMAND (cmd_set_dpdk_if_desc,static) = {
+    .path = "set dpdk interface descriptors",
+    .short_help = "set dpdk interface descriptors  [rx ] [tx ]",
+    .function = set_dpdk_if_desc,
+};
+/* *INDENT-ON* */
+
+static clib_error_t *
+show_dpdk_if_placement (vlib_main_t * vm, unformat_input_t * input,
+			vlib_cli_command_t * cmd)
+{
+  vlib_thread_main_t *tm = vlib_get_thread_main ();
+  dpdk_main_t *dm = &dpdk_main;
+  dpdk_device_and_queue_t *dq;
+  int cpu;
+
+  if (tm->n_vlib_mains == 1)
+    vlib_cli_output (vm, "All interfaces are handled by main thread");
+
+  for (cpu = 0; cpu < vec_len (dm->devices_by_cpu); cpu++)
+    {
+      if (cpu >= dm->input_cpu_first_index &&
+	  cpu < (dm->input_cpu_first_index + dm->input_cpu_count))
+	vlib_cli_output (vm, "Thread %u (%s at lcore %u):", cpu,
+			 vlib_worker_threads[cpu].name,
+			 vlib_worker_threads[cpu].lcore_id);
+
+      /* *INDENT-OFF* */
+      vec_foreach(dq, dm->devices_by_cpu[cpu])
+        {
+          u32 hw_if_index = dm->devices[dq->device].vlib_hw_if_index;
+          vnet_hw_interface_t * hi =  vnet_get_hw_interface(dm->vnet_main, hw_if_index);
+          vlib_cli_output(vm, "  %v queue %u", hi->name, dq->queue_id);
+        }
+      /* *INDENT-ON* */
+    }
+  return 0;
+}
+
+/*?
+ * This command is used to display the thread and core each
+ * DPDK interface and queue is assigned too.
+ *
+ * @cliexpar
+ * Example of how to display the DPDK interface placement:
+ * @cliexstart{show dpdk interface placement}
+ * Thread 1 (vpp_wk_0 at lcore 1):
+ *   GigabitEthernet0/8/0 queue 0
+ *   GigabitEthernet0/9/0 queue 0
+ * Thread 2 (vpp_wk_1 at lcore 2):
+ *   GigabitEthernet0/8/0 queue 1
+ *   GigabitEthernet0/9/0 queue 1
+ * @cliexend
+?*/
+/* *INDENT-OFF* */
+VLIB_CLI_COMMAND (cmd_show_dpdk_if_placement,static) = {
+    .path = "show dpdk interface placement",
+    .short_help = "show dpdk interface placement",
+    .function = show_dpdk_if_placement,
+};
+/* *INDENT-ON* */
+
+static int
+dpdk_device_queue_sort (void *a1, void *a2)
+{
+  dpdk_device_and_queue_t *dq1 = a1;
+  dpdk_device_and_queue_t *dq2 = a2;
+
+  if (dq1->device > dq2->device)
+    return 1;
+  else if (dq1->device < dq2->device)
+    return -1;
+  else if (dq1->queue_id > dq2->queue_id)
+    return 1;
+  else if (dq1->queue_id < dq2->queue_id)
+    return -1;
+  else
+    return 0;
+}
+
+static clib_error_t *
+set_dpdk_if_placement (vlib_main_t * vm, unformat_input_t * input,
+		       vlib_cli_command_t * cmd)
+{
+  unformat_input_t _line_input, *line_input = &_line_input;
+  dpdk_main_t *dm = &dpdk_main;
+  dpdk_device_and_queue_t *dq;
+  vnet_hw_interface_t *hw;
+  dpdk_device_t *xd;
+  u32 hw_if_index = (u32) ~ 0;
+  u32 queue = (u32) 0;
+  u32 cpu = (u32) ~ 0;
+  int i;
+  clib_error_t *error = NULL;
+
+  if (!unformat_user (input, unformat_line_input, line_input))
+    return 0;
+
+  while (unformat_check_input (line_input) != UNFORMAT_END_OF_INPUT)
+    {
+      if (unformat
+	  (line_input, "%U", unformat_vnet_hw_interface, dm->vnet_main,
+	   &hw_if_index))
+	;
+      else if (unformat (line_input, "queue %d", &queue))
+	;
+      else if (unformat (line_input, "thread %d", &cpu))
+	;
+      else
+	{
+	  error = clib_error_return (0, "parse error: '%U'",
+				     format_unformat_error, line_input);
+	  goto done;
+	}
+    }
+
+  if (hw_if_index == (u32) ~ 0)
+    {
+      error = clib_error_return (0, "please specify valid interface name");
+      goto done;
+    }
+
+  if (cpu < dm->input_cpu_first_index ||
+      cpu >= (dm->input_cpu_first_index + dm->input_cpu_count))
+    {
+      error = clib_error_return (0, "please specify valid thread id");
+      goto done;
+    }
+
+  hw = vnet_get_hw_interface (dm->vnet_main, hw_if_index);
+  xd = vec_elt_at_index (dm->devices, hw->dev_instance);
+
+  for (i = 0; i < vec_len (dm->devices_by_cpu); i++)
+    {
+      /* *INDENT-OFF* */
+      vec_foreach(dq, dm->devices_by_cpu[i])
+        {
+          if (hw_if_index == dm->devices[dq->device].vlib_hw_if_index &&
+              queue == dq->queue_id)
+            {
+              if (cpu == i) /* nothing to do */
+                goto done;
+
+              vec_del1(dm->devices_by_cpu[i], dq - dm->devices_by_cpu[i]);
+              vec_add2(dm->devices_by_cpu[cpu], dq, 1);
+              dq->queue_id = queue;
+              dq->device = xd->device_index;
+              xd->cpu_socket_id_by_queue[queue] =
+                rte_lcore_to_socket_id(vlib_worker_threads[cpu].lcore_id);
+
+              vec_sort_with_function(dm->devices_by_cpu[i],
+		                     dpdk_device_queue_sort);
+
+              vec_sort_with_function(dm->devices_by_cpu[cpu],
+		                     dpdk_device_queue_sort);
+
+              if (vec_len(dm->devices_by_cpu[i]) == 0)
+                vlib_node_set_state (vlib_mains[i], dpdk_input_node.index,
+		                     VLIB_NODE_STATE_DISABLED);
+
+              if (vec_len(dm->devices_by_cpu[cpu]) == 1)
+                vlib_node_set_state (vlib_mains[cpu], dpdk_input_node.index,
+		                     VLIB_NODE_STATE_POLLING);
+
+              goto done;
+            }
+        }
+      /* *INDENT-ON* */
+    }
+
+  error = clib_error_return (0, "not found");
+
+done:
+  unformat_free (line_input);
+
+  return error;
+}
+
+/*?
+ * This command is used to assign a given interface, and optionally a
+ * given queue, to a different thread. This will not create a thread,
+ * so the thread must already exist. Use '/etc/vpp/startup.conf'
+ * for the initial thread creation. If the 'queue' is not provided,
+ * it defaults to 0.
+ *
+ * @cliexpar
+ * Example of how to display the DPDK interface placement:
+ * @cliexstart{show dpdk interface placement}
+ * Thread 1 (vpp_wk_0 at lcore 1):
+ *   GigabitEthernet0/8/0 queue 0
+ *   GigabitEthernet0/9/0 queue 0
+ * Thread 2 (vpp_wk_1 at lcore 2):
+ *   GigabitEthernet0/8/0 queue 1
+ *   GigabitEthernet0/9/0 queue 1
+ * @cliexend
+ * Example of how to assign a DPDK interface and queue to a thread:
+ * @cliexcmd{set dpdk interface placement GigabitEthernet0/8/0 queue 1 thread 1}
+?*/
+/* *INDENT-OFF* */
+VLIB_CLI_COMMAND (cmd_set_dpdk_if_placement,static) = {
+    .path = "set dpdk interface placement",
+    .short_help = "set dpdk interface placement  [queue ] thread ",
+    .function = set_dpdk_if_placement,
+};
+/* *INDENT-ON* */
+
+static clib_error_t *
+show_dpdk_if_hqos_placement (vlib_main_t * vm, unformat_input_t * input,
+			     vlib_cli_command_t * cmd)
+{
+  vlib_thread_main_t *tm = vlib_get_thread_main ();
+  dpdk_main_t *dm = &dpdk_main;
+  dpdk_device_and_queue_t *dq;
+  int cpu;
+
+  if (tm->n_vlib_mains == 1)
+    vlib_cli_output (vm, "All interfaces are handled by main thread");
+
+  for (cpu = 0; cpu < vec_len (dm->devices_by_hqos_cpu); cpu++)
+    {
+      if (cpu >= dm->hqos_cpu_first_index &&
+	  cpu < (dm->hqos_cpu_first_index + dm->hqos_cpu_count))
+	vlib_cli_output (vm, "Thread %u (%s at lcore %u):", cpu,
+			 vlib_worker_threads[cpu].name,
+			 vlib_worker_threads[cpu].lcore_id);
+
+      vec_foreach (dq, dm->devices_by_hqos_cpu[cpu])
+      {
+	u32 hw_if_index = dm->devices[dq->device].vlib_hw_if_index;
+	vnet_hw_interface_t *hi =
+	  vnet_get_hw_interface (dm->vnet_main, hw_if_index);
+	vlib_cli_output (vm, "  %v queue %u", hi->name, dq->queue_id);
+      }
+    }
+  return 0;
+}
+
+/*?
+ * This command is used to display the thread and core each
+ * DPDK output interface and HQoS queue is assigned too.
+ *
+ * @cliexpar
+ * Example of how to display the DPDK output interface and HQoS queue placement:
+ * @cliexstart{show dpdk interface hqos placement}
+ * Thread 1 (vpp_hqos-threads_0 at lcore 3):
+ *   GigabitEthernet0/8/0 queue 0
+ * Thread 2 (vpp_hqos-threads_1 at lcore 4):
+ *   GigabitEthernet0/9/0 queue 0
+ * @cliexend
+?*/
+/* *INDENT-OFF* */
+VLIB_CLI_COMMAND (cmd_show_dpdk_if_hqos_placement, static) = {
+  .path = "show dpdk interface hqos placement",
+  .short_help = "show dpdk interface hqos placement",
+  .function = show_dpdk_if_hqos_placement,
+};
+/* *INDENT-ON* */
+
+static clib_error_t *
+set_dpdk_if_hqos_placement (vlib_main_t * vm, unformat_input_t * input,
+			    vlib_cli_command_t * cmd)
+{
+  unformat_input_t _line_input, *line_input = &_line_input;
+  dpdk_main_t *dm = &dpdk_main;
+  dpdk_device_and_queue_t *dq;
+  vnet_hw_interface_t *hw;
+  dpdk_device_t *xd;
+  u32 hw_if_index = (u32) ~ 0;
+  u32 cpu = (u32) ~ 0;
+  int i;
+  clib_error_t *error = NULL;
+
+  if (!unformat_user (input, unformat_line_input, line_input))
+    return 0;
+
+  while (unformat_check_input (line_input) != UNFORMAT_END_OF_INPUT)
+    {
+      if (unformat
+	  (line_input, "%U", unformat_vnet_hw_interface, dm->vnet_main,
+	   &hw_if_index))
+	;
+      else if (unformat (line_input, "thread %d", &cpu))
+	;
+      else
+	{
+	  error = clib_error_return (0, "parse error: '%U'",
+				     format_unformat_error, line_input);
+	  goto done;
+	}
+    }
+
+  if (hw_if_index == (u32) ~ 0)
+    return clib_error_return (0, "please specify valid interface name");
+
+  if (cpu < dm->hqos_cpu_first_index ||
+      cpu >= (dm->hqos_cpu_first_index + dm->hqos_cpu_count))
+    {
+      error = clib_error_return (0, "please specify valid thread id");
+      goto done;
+    }
+
+  hw = vnet_get_hw_interface (dm->vnet_main, hw_if_index);
+  xd = vec_elt_at_index (dm->devices, hw->dev_instance);
+
+  for (i = 0; i < vec_len (dm->devices_by_hqos_cpu); i++)
+    {
+      vec_foreach (dq, dm->devices_by_hqos_cpu[i])
+      {
+	if (hw_if_index == dm->devices[dq->device].vlib_hw_if_index)
+	  {
+	    if (cpu == i)	/* nothing to do */
+	      goto done;
+
+	    vec_del1 (dm->devices_by_hqos_cpu[i],
+		      dq - dm->devices_by_hqos_cpu[i]);
+	    vec_add2 (dm->devices_by_hqos_cpu[cpu], dq, 1);
+	    dq->queue_id = 0;
+	    dq->device = xd->device_index;
+
+	    vec_sort_with_function (dm->devices_by_hqos_cpu[i],
+				    dpdk_device_queue_sort);
+
+	    vec_sort_with_function (dm->devices_by_hqos_cpu[cpu],
+				    dpdk_device_queue_sort);
+
+	    goto done;
+	  }
+      }
+    }
+
+  error = clib_error_return (0, "not found");
+
+done:
+  unformat_free (line_input);
+
+  return error;
+}
+
+/*?
+ * This command is used to assign a given DPDK output interface and
+ * HQoS queue to a different thread. This will not create a thread,
+ * so the thread must already exist. Use '/etc/vpp/startup.conf'
+ * for the initial thread creation. See @ref qos_doc for more details.
+ *
+ * @cliexpar
+ * Example of how to display the DPDK output interface and HQoS queue placement:
+ * @cliexstart{show dpdk interface hqos placement}
+ * Thread 1 (vpp_hqos-threads_0 at lcore 3):
+ *   GigabitEthernet0/8/0 queue 0
+ * Thread 2 (vpp_hqos-threads_1 at lcore 4):
+ *   GigabitEthernet0/9/0 queue 0
+ * @cliexend
+ * Example of how to assign a DPDK output interface and HQoS queue to a thread:
+ * @cliexcmd{set dpdk interface hqos placement GigabitEthernet0/8/0 thread 2}
+?*/
+/* *INDENT-OFF* */
+VLIB_CLI_COMMAND (cmd_set_dpdk_if_hqos_placement, static) = {
+  .path = "set dpdk interface hqos placement",
+  .short_help = "set dpdk interface hqos placement  thread ",
+  .function = set_dpdk_if_hqos_placement,
+};
+/* *INDENT-ON* */
+
+static clib_error_t *
+set_dpdk_if_hqos_pipe (vlib_main_t * vm, unformat_input_t * input,
+		       vlib_cli_command_t * cmd)
+{
+  unformat_input_t _line_input, *line_input = &_line_input;
+  dpdk_main_t *dm = &dpdk_main;
+  vnet_hw_interface_t *hw;
+  dpdk_device_t *xd;
+  u32 hw_if_index = (u32) ~ 0;
+  u32 subport_id = (u32) ~ 0;
+  u32 pipe_id = (u32) ~ 0;
+  u32 profile_id = (u32) ~ 0;
+  int rv;
+  clib_error_t *error = NULL;
+
+  if (!unformat_user (input, unformat_line_input, line_input))
+    return 0;
+
+  while (unformat_check_input (line_input) != UNFORMAT_END_OF_INPUT)
+    {
+      if (unformat
+	  (line_input, "%U", unformat_vnet_hw_interface, dm->vnet_main,
+	   &hw_if_index))
+	;
+      else if (unformat (line_input, "subport %d", &subport_id))
+	;
+      else if (unformat (line_input, "pipe %d", &pipe_id))
+	;
+      else if (unformat (line_input, "profile %d", &profile_id))
+	;
+      else
+	{
+	  error = clib_error_return (0, "parse error: '%U'",
+				     format_unformat_error, line_input);
+	  goto done;
+	}
+    }
+
+  if (hw_if_index == (u32) ~ 0)
+    {
+      error = clib_error_return (0, "please specify valid interface name");
+      goto done;
+    }
+
+  hw = vnet_get_hw_interface (dm->vnet_main, hw_if_index);
+  xd = vec_elt_at_index (dm->devices, hw->dev_instance);
+
+  rv =
+    rte_sched_pipe_config (xd->hqos_ht->hqos, subport_id, pipe_id,
+			   profile_id);
+  if (rv)
+    {
+      error = clib_error_return (0, "pipe configuration failed");
+      goto done;
+    }
+
+done:
+  unformat_free (line_input);
+
+  return error;
+}
+
+/*?
+ * This command is used to change the profile associate with a HQoS pipe. The
+ * '' is zero based. Use the command
+ * 'show dpdk interface hqos' to display the content of each profile.
+ * See @ref qos_doc for more details.
+ *
+ * @note
+ * Currently there is not an API to create a new HQoS pipe profile. One is
+ * created by default in the code (search for 'hqos_pipe_params_default'').
+ * Additional profiles can be created in code and code recompiled. Then use this
+ * command to assign it.
+ *
+ * @cliexpar
+ * Example of how to assign a new profile to a HQoS pipe:
+ * @cliexcmd{set dpdk interface hqos pipe GigabitEthernet0/8/0 subport 0 pipe 2 profile 1}
+?*/
+/* *INDENT-OFF* */
+VLIB_CLI_COMMAND (cmd_set_dpdk_if_hqos_pipe, static) =
+{
+  .path = "set dpdk interface hqos pipe",
+  .short_help = "set dpdk interface hqos pipe  subport  pipe  "
+                  "profile ",
+  .function = set_dpdk_if_hqos_pipe,
+};
+/* *INDENT-ON* */
+
+static clib_error_t *
+set_dpdk_if_hqos_subport (vlib_main_t * vm, unformat_input_t * input,
+			  vlib_cli_command_t * cmd)
+{
+  unformat_input_t _line_input, *line_input = &_line_input;
+  dpdk_main_t *dm = &dpdk_main;
+  dpdk_device_t *xd = NULL;
+  u32 hw_if_index = (u32) ~ 0;
+  u32 subport_id = (u32) ~ 0;
+  struct rte_sched_subport_params p;
+  int rv;
+  clib_error_t *error = NULL;
+  u32 tb_rate = (u32) ~ 0;
+  u32 tb_size = (u32) ~ 0;
+  u32 tc_rate[RTE_SCHED_TRAFFIC_CLASSES_PER_PIPE] =
+    { (u32) ~ 0, (u32) ~ 0, (u32) ~ 0, (u32) ~ 0 };
+  u32 tc_period = (u32) ~ 0;
+  dpdk_device_config_t *devconf = NULL;
+
+  if (!unformat_user (input, unformat_line_input, line_input))
+    return 0;
+
+  while (unformat_check_input (line_input) != UNFORMAT_END_OF_INPUT)
+    {
+      if (unformat
+	  (line_input, "%U", unformat_vnet_hw_interface, dm->vnet_main,
+	   &hw_if_index))
+	;
+      else if (unformat (line_input, "subport %d", &subport_id))
+	;
+      else if (unformat (line_input, "rate %d", &tb_rate))
+	;
+      else if (unformat (line_input, "bktsize %d", &tb_size))
+	;
+      else if (unformat (line_input, "tc0 %d", &tc_rate[0]))
+	;
+      else if (unformat (line_input, "tc1 %d", &tc_rate[1]))
+	;
+      else if (unformat (line_input, "tc2 %d", &tc_rate[2]))
+	;
+      else if (unformat (line_input, "tc3 %d", &tc_rate[3]))
+	;
+      else if (unformat (line_input, "period %d", &tc_period))
+	;
+      else
+	{
+	  error = clib_error_return (0, "parse error: '%U'",
+				     format_unformat_error, line_input);
+	  goto done;
+	}
+    }
+
+  error = get_hqos (hw_if_index, subport_id, &xd, &devconf);
+
+  if (error == NULL)
+    {
+      /* Copy the current values over to local structure. */
+      memcpy (&p, &devconf->hqos.subport[subport_id], sizeof (p));
+
+      /* Update local structure with input values. */
+      if (tb_rate != (u32) ~ 0)
+	{
+	  p.tb_rate = tb_rate;
+	  p.tc_rate[0] = tb_rate;
+	  p.tc_rate[1] = tb_rate;
+	  p.tc_rate[2] = tb_rate;
+	  p.tc_rate[3] = tb_rate;
+	}
+      if (tb_size != (u32) ~ 0)
+	{
+	  p.tb_size = tb_size;
+	}
+      if (tc_rate[0] != (u32) ~ 0)
+	{
+	  p.tc_rate[0] = tc_rate[0];
+	}
+      if (tc_rate[1] != (u32) ~ 0)
+	{
+	  p.tc_rate[1] = tc_rate[1];
+	}
+      if (tc_rate[2] != (u32) ~ 0)
+	{
+	  p.tc_rate[2] = tc_rate[2];
+	}
+      if (tc_rate[3] != (u32) ~ 0)
+	{
+	  p.tc_rate[3] = tc_rate[3];
+	}
+      if (tc_period != (u32) ~ 0)
+	{
+	  p.tc_period = tc_period;
+	}
+
+      /* Apply changes. */
+      rv = rte_sched_subport_config (xd->hqos_ht->hqos, subport_id, &p);
+      if (rv)
+	{
+	  error = clib_error_return (0, "subport configuration failed");
+	  goto done;
+	}
+      else
+	{
+	  /* Successfully applied, so save of the input values. */
+	  memcpy (&devconf->hqos.subport[subport_id], &p, sizeof (p));
+	}
+    }
+
+done:
+  unformat_free (line_input);
+
+  return error;
+}
+
+/*?
+ * This command is used to set the subport level parameters such as token
+ * bucket rate (bytes per seconds), token bucket size (bytes), traffic class
+ * rates (bytes per seconds) and token update period (Milliseconds).
+ *
+ * By default, the 'rate' is set to 1250000000 bytes/second (10GbE
+ * rate) and each of the four traffic classes is set to 100% of the port rate.
+ * If the 'rate' is updated by this command, all four traffic classes
+ * are assigned the same value. Each of the four traffic classes can be updated
+ * individually.
+ *
+ * @cliexpar
+ * Example of how modify the subport attributes for a 1GbE link:
+ * @cliexcmd{set dpdk interface hqos subport GigabitEthernet0/8/0 subport 0 rate 125000000}
+?*/
+/* *INDENT-OFF* */
+VLIB_CLI_COMMAND (cmd_set_dpdk_if_hqos_subport, static) = {
+  .path = "set dpdk interface hqos subport",
+  .short_help = "set dpdk interface hqos subport  subport  "
+                 "[rate ] [bktsize ] [tc0 ] [tc1 ] [tc2 ] [tc3 ] "
+                 "[period ]",
+  .function = set_dpdk_if_hqos_subport,
+};
+/* *INDENT-ON* */
+
+static clib_error_t *
+set_dpdk_if_hqos_tctbl (vlib_main_t * vm, unformat_input_t * input,
+			vlib_cli_command_t * cmd)
+{
+  unformat_input_t _line_input, *line_input = &_line_input;
+  vlib_thread_main_t *tm = vlib_get_thread_main ();
+  dpdk_main_t *dm = &dpdk_main;
+  vnet_hw_interface_t *hw;
+  dpdk_device_t *xd;
+  u32 hw_if_index = (u32) ~ 0;
+  u32 tc = (u32) ~ 0;
+  u32 queue = (u32) ~ 0;
+  u32 entry = (u32) ~ 0;
+  u32 val, i;
+  clib_error_t *error = NULL;
+
+  if (!unformat_user (input, unformat_line_input, line_input))
+    return 0;
+
+  while (unformat_check_input (line_input) != UNFORMAT_END_OF_INPUT)
+    {
+      if (unformat
+	  (line_input, "%U", unformat_vnet_hw_interface, dm->vnet_main,
+	   &hw_if_index))
+	;
+      else if (unformat (line_input, "entry %d", &entry))
+	;
+      else if (unformat (line_input, "tc %d", &tc))
+	;
+      else if (unformat (line_input, "queue %d", &queue))
+	;
+      else
+	{
+	  error = clib_error_return (0, "parse error: '%U'",
+				     format_unformat_error, line_input);
+	  goto done;
+	}
+    }
+
+  if (hw_if_index == (u32) ~ 0)
+    {
+      error = clib_error_return (0, "please specify valid interface name");
+      goto done;
+    }
+  if (entry >= 64)
+    {
+      error = clib_error_return (0, "invalid entry");
+      goto done;
+    }
+  if (tc >= RTE_SCHED_TRAFFIC_CLASSES_PER_PIPE)
+    {
+      error = clib_error_return (0, "invalid traffic class");
+      goto done;
+    }
+  if (queue >= RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS)
+    {
+      error = clib_error_return (0, "invalid traffic class queue");
+      goto done;
+    }
+
+  hw = vnet_get_hw_interface (dm->vnet_main, hw_if_index);
+  xd = vec_elt_at_index (dm->devices, hw->dev_instance);
+
+  /* Detect the set of worker threads */
+  uword *p = hash_get_mem (tm->thread_registrations_by_name, "workers");
+  /* Should never happen, shut up Coverity warning */
+  if (p == 0)
+    {
+      error = clib_error_return (0, "no worker registrations?");
+      goto done;
+    }
+
+  vlib_thread_registration_t *tr = (vlib_thread_registration_t *) p[0];
+  int worker_thread_first = tr->first_index;
+  int worker_thread_count = tr->count;
+
+  val = tc * RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS + queue;
+  for (i = 0; i < worker_thread_count; i++)
+    xd->hqos_wt[worker_thread_first + i].hqos_tc_table[entry] = val;
+
+done:
+  unformat_free (line_input);
+
+  return error;
+}
+
+/*?
+ * This command is used to set the traffic class translation table. The
+ * traffic class translation table is used to map 64 values (0-63) to one of
+ * four traffic class and one of four HQoS input queue. Use the 'show
+ * dpdk interface hqos' command to display the traffic class translation
+ * table. See @ref qos_doc for more details.
+ *
+ * This command has the following parameters:
+ *
+ * -  - Used to specify the output interface.
+ *
+ * - entry  - Mapped value (0-63) to assign traffic class and queue to.
+ *
+ * - tc  - Traffic class (0-3) to be used by the provided mapped value.
+ *
+ * - queue  - HQoS input queue (0-3) to be used by the provided mapped value.
+ *
+ * @cliexpar
+ * Example of how modify the traffic class translation table:
+ * @cliexcmd{set dpdk interface hqos tctbl GigabitEthernet0/8/0 entry 16 tc 2 queue 2}
+?*/
+/* *INDENT-OFF* */
+VLIB_CLI_COMMAND (cmd_set_dpdk_if_hqos_tctbl, static) = {
+  .path = "set dpdk interface hqos tctbl",
+  .short_help = "set dpdk interface hqos tctbl  entry  tc  queue ",
+  .function = set_dpdk_if_hqos_tctbl,
+};
+/* *INDENT-ON* */
+
+static clib_error_t *
+set_dpdk_if_hqos_pktfield (vlib_main_t * vm, unformat_input_t * input,
+			   vlib_cli_command_t * cmd)
+{
+  unformat_input_t _line_input, *line_input = &_line_input;
+  vlib_thread_main_t *tm = vlib_get_thread_main ();
+  dpdk_main_t *dm = &dpdk_main;
+  clib_error_t *error = NULL;
+
+  /* Device specific data */
+  struct rte_eth_dev_info dev_info;
+  dpdk_device_config_t *devconf = 0;
+  vnet_hw_interface_t *hw;
+  dpdk_device_t *xd;
+  u32 hw_if_index = (u32) ~ 0;
+
+  /* Detect the set of worker threads */
+  uword *p = hash_get_mem (tm->thread_registrations_by_name, "workers");
+  /* Should never happen, shut up Coverity warning */
+  if (p == 0)
+    return clib_error_return (0, "no worker registrations?");
+
+  vlib_thread_registration_t *tr = (vlib_thread_registration_t *) p[0];
+  int worker_thread_first = tr->first_index;
+  int worker_thread_count = tr->count;
+
+  /* Packet field configuration */
+  u64 mask = (u64) ~ 0;
+  u32 id = (u32) ~ 0;
+  u32 offset = (u32) ~ 0;
+
+  /* HQoS params */
+  u32 n_subports_per_port, n_pipes_per_subport, tctbl_size;
+
+  u32 i;
+
+  /* Parse input arguments */
+  if (!unformat_user (input, unformat_line_input, line_input))
+    return 0;
+
+  while (unformat_check_input (line_input) != UNFORMAT_END_OF_INPUT)
+    {
+      if (unformat
+	  (line_input, "%U", unformat_vnet_hw_interface, dm->vnet_main,
+	   &hw_if_index))
+	;
+      else if (unformat (line_input, "id subport"))
+	id = 0;
+      else if (unformat (line_input, "id pipe"))
+	id = 1;
+      else if (unformat (line_input, "id tc"))
+	id = 2;
+      else if (unformat (line_input, "id %d", &id))
+	;
+      else if (unformat (line_input, "offset %d", &offset))
+	;
+      else if (unformat (line_input, "mask %llx", &mask))
+	;
+      else
+	{
+	  error = clib_error_return (0, "parse error: '%U'",
+				     format_unformat_error, line_input);
+	  goto done;
+	}
+    }
+
+  /* Get interface */
+  if (hw_if_index == (u32) ~ 0)
+    {
+      error = clib_error_return (0, "please specify valid interface name");
+      goto done;
+    }
+
+  hw = vnet_get_hw_interface (dm->vnet_main, hw_if_index);
+  xd = vec_elt_at_index (dm->devices, hw->dev_instance);
+
+  rte_eth_dev_info_get (xd->device_index, &dev_info);
+  if (dev_info.pci_dev)
+    {				/* bonded interface has no pci info */
+      vlib_pci_addr_t pci_addr;
+
+      pci_addr.domain = dev_info.pci_dev->addr.domain;
+      pci_addr.bus = dev_info.pci_dev->addr.bus;
+      pci_addr.slot = dev_info.pci_dev->addr.devid;
+      pci_addr.function = dev_info.pci_dev->addr.function;
+
+      p =
+	hash_get (dm->conf->device_config_index_by_pci_addr, pci_addr.as_u32);
+    }
+
+  if (p)
+    devconf = pool_elt_at_index (dm->conf->dev_confs, p[0]);
+  else
+    devconf = &dm->conf->default_devconf;
+
+  if (devconf->hqos_enabled == 0)
+    {
+      vlib_cli_output (vm, "HQoS disabled for this interface");
+      goto done;
+    }
+
+  n_subports_per_port = devconf->hqos.port.n_subports_per_port;
+  n_pipes_per_subport = devconf->hqos.port.n_pipes_per_subport;
+  tctbl_size = RTE_DIM (devconf->hqos.tc_table);
+
+  /* Validate packet field configuration: id, offset and mask */
+  if (id >= 3)
+    {
+      error = clib_error_return (0, "invalid packet field id");
+      goto done;
+    }
+
+  switch (id)
+    {
+    case 0:
+      if (dpdk_hqos_validate_mask (mask, n_subports_per_port) != 0)
+	{
+	  error = clib_error_return (0, "invalid subport ID mask "
+				     "(n_subports_per_port = %u)",
+				     n_subports_per_port);
+	  goto done;
+	}
+      break;
+    case 1:
+      if (dpdk_hqos_validate_mask (mask, n_pipes_per_subport) != 0)
+	{
+	  error = clib_error_return (0, "invalid pipe ID mask "
+				     "(n_pipes_per_subport = %u)",
+				     n_pipes_per_subport);
+	  goto done;
+	}
+      break;
+    case 2:
+    default:
+      if (dpdk_hqos_validate_mask (mask, tctbl_size) != 0)
+	{
+	  error = clib_error_return (0, "invalid TC table index mask "
+				     "(TC table size = %u)", tctbl_size);
+	  goto done;
+	}
+    }
+
+  /* Propagate packet field configuration to all workers */
+  for (i = 0; i < worker_thread_count; i++)
+    switch (id)
+      {
+      case 0:
+	xd->hqos_wt[worker_thread_first + i].hqos_field0_slabpos = offset;
+	xd->hqos_wt[worker_thread_first + i].hqos_field0_slabmask = mask;
+	xd->hqos_wt[worker_thread_first + i].hqos_field0_slabshr =
+	  __builtin_ctzll (mask);
+	break;
+      case 1:
+	xd->hqos_wt[worker_thread_first + i].hqos_field1_slabpos = offset;
+	xd->hqos_wt[worker_thread_first + i].hqos_field1_slabmask = mask;
+	xd->hqos_wt[worker_thread_first + i].hqos_field1_slabshr =
+	  __builtin_ctzll (mask);
+	break;
+      case 2:
+      default:
+	xd->hqos_wt[worker_thread_first + i].hqos_field2_slabpos = offset;
+	xd->hqos_wt[worker_thread_first + i].hqos_field2_slabmask = mask;
+	xd->hqos_wt[worker_thread_first + i].hqos_field2_slabshr =
+	  __builtin_ctzll (mask);
+      }
+
+done:
+  unformat_free (line_input);
+
+  return error;
+}
+
+/*?
+ * This command is used to set the packet fields required for classifiying the
+ * incoming packet. As a result of classification process, packet field
+ * information will be mapped to 5 tuples (subport, pipe, traffic class, pipe,
+ * color) and stored in packet mbuf.
+ *
+ * This command has the following parameters:
+ *
+ * -  - Used to specify the output interface.
+ *
+ * - id subport|pipe|tc - Classification occurs across three fields.
+ * This parameter indicates which of the three masks are being configured. Legacy
+ * code used 0-2 to represent these three fields, so 0-2 is still accepted.
+ *   - subport|0 - Currently only one subport is supported, so only
+ * an empty mask is supported for the subport classification.
+ *   - pipe|1 - Currently, 4096 pipes per subport are supported, so a
+ * 12-bit mask should be configure to map to the 0-4095 pipes.
+ *   - tc|2 - The translation table (see 'set dpdk interface hqos
+ * tctbl' command) maps each value (0-63) into one of the 4 traffic classes
+ * per pipe. A 6-bit mask should be configure to map this field to a traffic class.
+ *
+ * - offset  - Offset in the packet to apply the 64-bit mask for classification.
+ * The offset should be on an 8-byte boundary (0,8,16,24..).
+ *
+ * - mask  - 64-bit mask to apply to packet at the given 'offset'.
+ * Bits must be contiguous and should not include '0x'.
+ *
+ * The default values for the 'pktfield' assumes Ethernet/IPv4/UDP packets with
+ * no VLAN. Adjust based on expected packet format and desired classification field.
+ * - 'subport' is always empty (offset 0 mask 0000000000000000)
+ * - By default, 'pipe' maps to the UDP payload bits 12 .. 23 (offset 40
+ * mask 0000000fff000000)
+ * - By default, 'tc' maps to the DSCP field in IP header (offset 48 mask
+ * 00000000000000fc)
+ *
+ * @cliexpar
+ * Example of how modify the 'pipe' classification filter to match VLAN:
+ * @cliexcmd{set dpdk interface hqos pktfield GigabitEthernet0/8/0 id pipe offset 8 mask 0000000000000FFF}
+?*/
+/* *INDENT-OFF* */
+VLIB_CLI_COMMAND (cmd_set_dpdk_if_hqos_pktfield, static) = {
+  .path = "set dpdk interface hqos pktfield",
+  .short_help = "set dpdk interface hqos pktfield  id subport|pipe|tc offset  "
+                 "mask ",
+  .function = set_dpdk_if_hqos_pktfield,
+};
+/* *INDENT-ON* */
+
+static clib_error_t *
+show_dpdk_if_hqos (vlib_main_t * vm, unformat_input_t * input,
+		   vlib_cli_command_t * cmd)
+{
+  unformat_input_t _line_input, *line_input = &_line_input;
+  vlib_thread_main_t *tm = vlib_get_thread_main ();
+  dpdk_main_t *dm = &dpdk_main;
+  vnet_hw_interface_t *hw;
+  dpdk_device_t *xd;
+  dpdk_device_config_hqos_t *cfg;
+  dpdk_device_hqos_per_hqos_thread_t *ht;
+  dpdk_device_hqos_per_worker_thread_t *wk;
+  u32 *tctbl;
+  u32 hw_if_index = (u32) ~ 0;
+  u32 profile_id, subport_id, i;
+  struct rte_eth_dev_info dev_info;
+  dpdk_device_config_t *devconf = 0;
+  vlib_thread_registration_t *tr;
+  uword *p = 0;
+  clib_error_t *error = NULL;
+
+  if (!unformat_user (input, unformat_line_input, line_input))
+    return 0;
+
+  while (unformat_check_input (line_input) != UNFORMAT_END_OF_INPUT)
+    {
+      if (unformat
+	  (line_input, "%U", unformat_vnet_hw_interface, dm->vnet_main,
+	   &hw_if_index))
+	;
+      else
+	{
+	  error = clib_error_return (0, "parse error: '%U'",
+				     format_unformat_error, line_input);
+	  goto done;
+	}
+    }
+
+  if (hw_if_index == (u32) ~ 0)
+    {
+      error = clib_error_return (0, "please specify interface name!!");
+      goto done;
+    }
+
+  hw = vnet_get_hw_interface (dm->vnet_main, hw_if_index);
+  xd = vec_elt_at_index (dm->devices, hw->dev_instance);
+
+  rte_eth_dev_info_get (xd->device_index, &dev_info);
+  if (dev_info.pci_dev)
+    {				/* bonded interface has no pci info */
+      vlib_pci_addr_t pci_addr;
+
+      pci_addr.domain = dev_info.pci_dev->addr.domain;
+      pci_addr.bus = dev_info.pci_dev->addr.bus;
+      pci_addr.slot = dev_info.pci_dev->addr.devid;
+      pci_addr.function = dev_info.pci_dev->addr.function;
+
+      p =
+	hash_get (dm->conf->device_config_index_by_pci_addr, pci_addr.as_u32);
+    }
+
+  if (p)
+    devconf = pool_elt_at_index (dm->conf->dev_confs, p[0]);
+  else
+    devconf = &dm->conf->default_devconf;
+
+  if (devconf->hqos_enabled == 0)
+    {
+      vlib_cli_output (vm, "HQoS disabled for this interface");
+      goto done;
+    }
+
+  /* Detect the set of worker threads */
+  p = hash_get_mem (tm->thread_registrations_by_name, "workers");
+
+  /* Should never happen, shut up Coverity warning */
+  if (p == 0)
+    {
+      error = clib_error_return (0, "no worker registrations?");
+      goto done;
+    }
+
+  tr = (vlib_thread_registration_t *) p[0];
+
+  cfg = &devconf->hqos;
+  ht = xd->hqos_ht;
+  wk = &xd->hqos_wt[tr->first_index];
+  tctbl = wk->hqos_tc_table;
+
+  vlib_cli_output (vm, " Thread:");
+  vlib_cli_output (vm, "   Input SWQ size = %u packets", cfg->swq_size);
+  vlib_cli_output (vm, "   Enqueue burst size = %u packets",
+		   ht->hqos_burst_enq);
+  vlib_cli_output (vm, "   Dequeue burst size = %u packets",
+		   ht->hqos_burst_deq);
+
+  vlib_cli_output (vm,
+		   "   Packet field 0: slab position = %4u, slab bitmask = 0x%016llx   (subport)",
+		   wk->hqos_field0_slabpos, wk->hqos_field0_slabmask);
+  vlib_cli_output (vm,
+		   "   Packet field 1: slab position = %4u, slab bitmask = 0x%016llx   (pipe)",
+		   wk->hqos_field1_slabpos, wk->hqos_field1_slabmask);
+  vlib_cli_output (vm,
+		   "   Packet field 2: slab position = %4u, slab bitmask = 0x%016llx   (tc)",
+		   wk->hqos_field2_slabpos, wk->hqos_field2_slabmask);
+  vlib_cli_output (vm,
+		   "   Packet field 2  tc translation table: ([Mapped Value Range]: tc/queue tc/queue ...)");
+  vlib_cli_output (vm,
+		   "     [ 0 .. 15]: "
+		   "%u/%u %u/%u %u/%u %u/%u %u/%u %u/%u %u/%u %u/%u %u/%u %u/%u %u/%u %u/%u %u/%u %u/%u %u/%u %u/%u",
+		   tctbl[0] / RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
+		   tctbl[0] % RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
+		   tctbl[1] / RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
+		   tctbl[1] % RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
+		   tctbl[2] / RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
+		   tctbl[2] % RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
+		   tctbl[3] / RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
+		   tctbl[3] % RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
+		   tctbl[4] / RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
+		   tctbl[4] % RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
+		   tctbl[5] / RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
+		   tctbl[5] % RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
+		   tctbl[6] / RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
+		   tctbl[6] % RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
+		   tctbl[7] / RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
+		   tctbl[7] % RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
+		   tctbl[8] / RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
+		   tctbl[8] % RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
+		   tctbl[9] / RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
+		   tctbl[9] % RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
+		   tctbl[10] / RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
+		   tctbl[10] % RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
+		   tctbl[11] / RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
+		   tctbl[11] % RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
+		   tctbl[12] / RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
+		   tctbl[12] % RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
+		   tctbl[13] / RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
+		   tctbl[13] % RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
+		   tctbl[14] / RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
+		   tctbl[14] % RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
+		   tctbl[15] / RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
+		   tctbl[15] % RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS);
+  vlib_cli_output (vm,
+		   "     [16 .. 31]: "
+		   "%u/%u %u/%u %u/%u %u/%u %u/%u %u/%u %u/%u %u/%u %u/%u %u/%u %u/%u %u/%u %u/%u %u/%u %u/%u %u/%u",
+		   tctbl[16] / RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
+		   tctbl[16] % RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
+		   tctbl[17] / RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
+		   tctbl[17] % RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
+		   tctbl[18] / RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
+		   tctbl[18] % RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
+		   tctbl[19] / RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
+		   tctbl[19] % RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
+		   tctbl[20] / RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
+		   tctbl[20] % RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
+		   tctbl[21] / RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
+		   tctbl[21] % RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
+		   tctbl[22] / RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
+		   tctbl[22] % RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
+		   tctbl[23] / RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
+		   tctbl[23] % RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
+		   tctbl[24] / RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
+		   tctbl[24] % RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
+		   tctbl[25] / RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
+		   tctbl[25] % RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
+		   tctbl[26] / RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
+		   tctbl[26] % RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
+		   tctbl[27] / RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
+		   tctbl[27] % RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
+		   tctbl[28] / RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
+		   tctbl[28] % RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
+		   tctbl[29] / RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
+		   tctbl[29] % RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
+		   tctbl[30] / RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
+		   tctbl[30] % RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
+		   tctbl[31] / RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
+		   tctbl[31] % RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS);
+  vlib_cli_output (vm,
+		   "     [32 .. 47]: "
+		   "%u/%u %u/%u %u/%u %u/%u %u/%u %u/%u %u/%u %u/%u %u/%u %u/%u %u/%u %u/%u %u/%u %u/%u %u/%u %u/%u",
+		   tctbl[32] / RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
+		   tctbl[32] % RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
+		   tctbl[33] / RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
+		   tctbl[33] % RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
+		   tctbl[34] / RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
+		   tctbl[34] % RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
+		   tctbl[35] / RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
+		   tctbl[35] % RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
+		   tctbl[36] / RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
+		   tctbl[36] % RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
+		   tctbl[37] / RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
+		   tctbl[37] % RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
+		   tctbl[38] / RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
+		   tctbl[38] % RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
+		   tctbl[39] / RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
+		   tctbl[39] % RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
+		   tctbl[40] / RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
+		   tctbl[40] % RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
+		   tctbl[41] / RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
+		   tctbl[41] % RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
+		   tctbl[42] / RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
+		   tctbl[42] % RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
+		   tctbl[43] / RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
+		   tctbl[43] % RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
+		   tctbl[44] / RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
+		   tctbl[44] % RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
+		   tctbl[45] / RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
+		   tctbl[45] % RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
+		   tctbl[46] / RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
+		   tctbl[46] % RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
+		   tctbl[47] / RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
+		   tctbl[47] % RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS);
+  vlib_cli_output (vm,
+		   "     [48 .. 63]: "
+		   "%u/%u %u/%u %u/%u %u/%u %u/%u %u/%u %u/%u %u/%u %u/%u %u/%u %u/%u %u/%u %u/%u %u/%u %u/%u %u/%u",
+		   tctbl[48] / RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
+		   tctbl[48] % RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
+		   tctbl[49] / RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
+		   tctbl[49] % RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
+		   tctbl[50] / RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
+		   tctbl[50] % RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
+		   tctbl[51] / RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
+		   tctbl[51] % RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
+		   tctbl[52] / RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
+		   tctbl[52] % RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
+		   tctbl[53] / RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
+		   tctbl[53] % RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
+		   tctbl[54] / RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
+		   tctbl[54] % RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
+		   tctbl[55] / RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
+		   tctbl[55] % RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
+		   tctbl[56] / RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
+		   tctbl[56] % RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
+		   tctbl[57] / RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
+		   tctbl[57] % RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
+		   tctbl[58] / RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
+		   tctbl[58] % RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
+		   tctbl[59] / RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
+		   tctbl[59] % RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
+		   tctbl[60] / RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
+		   tctbl[60] % RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
+		   tctbl[61] / RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
+		   tctbl[61] % RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
+		   tctbl[62] / RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
+		   tctbl[62] % RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
+		   tctbl[63] / RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
+		   tctbl[63] % RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS);
+  vlib_cli_output (vm, " Port:");
+  vlib_cli_output (vm, "   Rate = %u bytes/second", cfg->port.rate);
+  vlib_cli_output (vm, "   MTU = %u bytes", cfg->port.mtu);
+  vlib_cli_output (vm, "   Frame overhead = %u bytes",
+		   cfg->port.frame_overhead);
+  vlib_cli_output (vm, "   Number of subports = %u",
+		   cfg->port.n_subports_per_port);
+  vlib_cli_output (vm, "   Number of pipes per subport = %u",
+		   cfg->port.n_pipes_per_subport);
+  vlib_cli_output (vm,
+		   "   Packet queue size: TC0 = %u, TC1 = %u, TC2 = %u, TC3 = %u packets",
+		   cfg->port.qsize[0], cfg->port.qsize[1], cfg->port.qsize[2],
+		   cfg->port.qsize[3]);
+  vlib_cli_output (vm, "   Number of pipe profiles = %u",
+		   cfg->port.n_pipe_profiles);
+
+  for (subport_id = 0; subport_id < vec_len (cfg->subport); subport_id++)
+    {
+      vlib_cli_output (vm, " Subport %u:", subport_id);
+      vlib_cli_output (vm, "   Rate = %u bytes/second",
+		       cfg->subport[subport_id].tb_rate);
+      vlib_cli_output (vm, "   Token bucket size = %u bytes",
+		       cfg->subport[subport_id].tb_size);
+      vlib_cli_output (vm,
+		       "   Traffic class rate: TC0 = %u, TC1 = %u, TC2 = %u, TC3 = %u bytes/second",
+		       cfg->subport[subport_id].tc_rate[0],
+		       cfg->subport[subport_id].tc_rate[1],
+		       cfg->subport[subport_id].tc_rate[2],
+		       cfg->subport[subport_id].tc_rate[3]);
+      vlib_cli_output (vm, "   TC period = %u milliseconds",
+		       cfg->subport[subport_id].tc_period);
+    }
+
+  for (profile_id = 0; profile_id < vec_len (cfg->pipe); profile_id++)
+    {
+      vlib_cli_output (vm, " Pipe profile %u:", profile_id);
+      vlib_cli_output (vm, "   Rate = %u bytes/second",
+		       cfg->pipe[profile_id].tb_rate);
+      vlib_cli_output (vm, "   Token bucket size = %u bytes",
+		       cfg->pipe[profile_id].tb_size);
+      vlib_cli_output (vm,
+		       "   Traffic class rate: TC0 = %u, TC1 = %u, TC2 = %u, TC3 = %u bytes/second",
+		       cfg->pipe[profile_id].tc_rate[0],
+		       cfg->pipe[profile_id].tc_rate[1],
+		       cfg->pipe[profile_id].tc_rate[2],
+		       cfg->pipe[profile_id].tc_rate[3]);
+      vlib_cli_output (vm, "   TC period = %u milliseconds",
+		       cfg->pipe[profile_id].tc_period);
+#ifdef RTE_SCHED_SUBPORT_TC_OV
+      vlib_cli_output (vm, "   TC3 oversubscription_weight = %u",
+		       cfg->pipe[profile_id].tc_ov_weight);
+#endif
+
+      for (i = 0; i < RTE_SCHED_TRAFFIC_CLASSES_PER_PIPE; i++)
+	{
+	  vlib_cli_output (vm,
+			   "   TC%u WRR weights: Q0 = %u, Q1 = %u, Q2 = %u, Q3 = %u",
+			   i, cfg->pipe[profile_id].wrr_weights[i * 4],
+			   cfg->pipe[profile_id].wrr_weights[i * 4 + 1],
+			   cfg->pipe[profile_id].wrr_weights[i * 4 + 2],
+			   cfg->pipe[profile_id].wrr_weights[i * 4 + 3]);
+	}
+    }
+
+#ifdef RTE_SCHED_RED
+  vlib_cli_output (vm, " Weighted Random Early Detection (WRED):");
+  for (i = 0; i < RTE_SCHED_TRAFFIC_CLASSES_PER_PIPE; i++)
+    {
+      vlib_cli_output (vm, "   TC%u min: G = %u, Y = %u, R = %u", i,
+		       cfg->port.red_params[i][e_RTE_METER_GREEN].min_th,
+		       cfg->port.red_params[i][e_RTE_METER_YELLOW].min_th,
+		       cfg->port.red_params[i][e_RTE_METER_RED].min_th);
+
+      vlib_cli_output (vm, "   TC%u max: G = %u, Y = %u, R = %u", i,
+		       cfg->port.red_params[i][e_RTE_METER_GREEN].max_th,
+		       cfg->port.red_params[i][e_RTE_METER_YELLOW].max_th,
+		       cfg->port.red_params[i][e_RTE_METER_RED].max_th);
+
+      vlib_cli_output (vm,
+		       "   TC%u inverted probability: G = %u, Y = %u, R = %u",
+		       i, cfg->port.red_params[i][e_RTE_METER_GREEN].maxp_inv,
+		       cfg->port.red_params[i][e_RTE_METER_YELLOW].maxp_inv,
+		       cfg->port.red_params[i][e_RTE_METER_RED].maxp_inv);
+
+      vlib_cli_output (vm, "   TC%u weight: R = %u, Y = %u, R = %u", i,
+		       cfg->port.red_params[i][e_RTE_METER_GREEN].wq_log2,
+		       cfg->port.red_params[i][e_RTE_METER_YELLOW].wq_log2,
+		       cfg->port.red_params[i][e_RTE_METER_RED].wq_log2);
+    }
+#endif
+
+done:
+  unformat_free (line_input);
+
+  return error;
+}
+
+/*?
+ * This command is used to display details of an output interface's HQoS
+ * settings.
+ *
+ * @cliexpar
+ * Example of how to display HQoS settings for an interfaces:
+ * @cliexstart{show dpdk interface hqos GigabitEthernet0/8/0}
+ *  Thread:
+ *    Input SWQ size = 4096 packets
+ *    Enqueue burst size = 256 packets
+ *    Dequeue burst size = 220 packets
+ *    Packet field 0: slab position =    0, slab bitmask = 0x0000000000000000   (subport)
+ *    Packet field 1: slab position =   40, slab bitmask = 0x0000000fff000000   (pipe)
+ *    Packet field 2: slab position =    8, slab bitmask = 0x00000000000000fc   (tc)
+ *    Packet field 2  tc translation table: ([Mapped Value Range]: tc/queue tc/queue ...)
+ *      [ 0 .. 15]: 0/0 0/1 0/2 0/3 1/0 1/1 1/2 1/3 2/0 2/1 2/2 2/3 3/0 3/1 3/2 3/3
+ *      [16 .. 31]: 0/0 0/1 0/2 0/3 1/0 1/1 1/2 1/3 2/0 2/1 2/2 2/3 3/0 3/1 3/2 3/3
+ *      [32 .. 47]: 0/0 0/1 0/2 0/3 1/0 1/1 1/2 1/3 2/0 2/1 2/2 2/3 3/0 3/1 3/2 3/3
+ *      [48 .. 63]: 0/0 0/1 0/2 0/3 1/0 1/1 1/2 1/3 2/0 2/1 2/2 2/3 3/0 3/1 3/2 3/3
+ *  Port:
+ *    Rate = 1250000000 bytes/second
+ *    MTU = 1514 bytes
+ *    Frame overhead = 24 bytes
+ *    Number of subports = 1
+ *    Number of pipes per subport = 4096
+ *    Packet queue size: TC0 = 64, TC1 = 64, TC2 = 64, TC3 = 64 packets
+ *    Number of pipe profiles = 2
+ *  Subport 0:
+ *    Rate = 1250000000 bytes/second
+ *    Token bucket size = 1000000 bytes
+ *    Traffic class rate: TC0 = 1250000000, TC1 = 1250000000, TC2 = 1250000000, TC3 = 1250000000 bytes/second
+ *    TC period = 10 milliseconds
+ *  Pipe profile 0:
+ *    Rate = 305175 bytes/second
+ *    Token bucket size = 1000000 bytes
+ *    Traffic class rate: TC0 = 305175, TC1 = 305175, TC2 = 305175, TC3 = 305175 bytes/second
+ *    TC period = 40 milliseconds
+ *    TC0 WRR weights: Q0 = 1, Q1 = 1, Q2 = 1, Q3 = 1
+ *    TC1 WRR weights: Q0 = 1, Q1 = 1, Q2 = 1, Q3 = 1
+ *    TC2 WRR weights: Q0 = 1, Q1 = 1, Q2 = 1, Q3 = 1
+ *    TC3 WRR weights: Q0 = 1, Q1 = 1, Q2 = 1, Q3 = 1
+ * @cliexend
+?*/
+/* *INDENT-OFF* */
+VLIB_CLI_COMMAND (cmd_show_dpdk_if_hqos, static) = {
+  .path = "show dpdk interface hqos",
+  .short_help = "show dpdk interface hqos ",
+  .function = show_dpdk_if_hqos,
+};
+
+/* *INDENT-ON* */
+
+static clib_error_t *
+show_dpdk_hqos_queue_stats (vlib_main_t * vm, unformat_input_t * input,
+			    vlib_cli_command_t * cmd)
+{
+  unformat_input_t _line_input, *line_input = &_line_input;
+  clib_error_t *error = NULL;
+#ifdef RTE_SCHED_COLLECT_STATS
+  dpdk_main_t *dm = &dpdk_main;
+  u32 hw_if_index = (u32) ~ 0;
+  u32 subport = (u32) ~ 0;
+  u32 pipe = (u32) ~ 0;
+  u32 tc = (u32) ~ 0;
+  u32 tc_q = (u32) ~ 0;
+  vnet_hw_interface_t *hw;
+  dpdk_device_t *xd;
+  uword *p = 0;
+  struct rte_eth_dev_info dev_info;
+  dpdk_device_config_t *devconf = 0;
+  u32 qindex;
+  struct rte_sched_queue_stats stats;
+  u16 qlen;
+
+  if (!unformat_user (input, unformat_line_input, line_input))
+    return 0;
+
+  while (unformat_check_input (line_input) != UNFORMAT_END_OF_INPUT)
+    {
+      if (unformat
+	  (line_input, "%U", unformat_vnet_hw_interface, dm->vnet_main,
+	   &hw_if_index))
+	;
+
+      else if (unformat (line_input, "subport %d", &subport))
+	;
+
+      else if (unformat (line_input, "pipe %d", &pipe))
+	;
+
+      else if (unformat (line_input, "tc %d", &tc))
+	;
+
+      else if (unformat (line_input, "tc_q %d", &tc_q))
+	;
+
+      else
+	{
+	  error = clib_error_return (0, "parse error: '%U'",
+				     format_unformat_error, line_input);
+	  goto done;
+	}
+    }
+
+  if (hw_if_index == (u32) ~ 0)
+    {
+      error = clib_error_return (0, "please specify interface name!!");
+      goto done;
+    }
+
+  hw = vnet_get_hw_interface (dm->vnet_main, hw_if_index);
+  xd = vec_elt_at_index (dm->devices, hw->dev_instance);
+
+  rte_eth_dev_info_get (xd->device_index, &dev_info);
+  if (dev_info.pci_dev)
+    {				/* bonded interface has no pci info */
+      vlib_pci_addr_t pci_addr;
+
+      pci_addr.domain = dev_info.pci_dev->addr.domain;
+      pci_addr.bus = dev_info.pci_dev->addr.bus;
+      pci_addr.slot = dev_info.pci_dev->addr.devid;
+      pci_addr.function = dev_info.pci_dev->addr.function;
+
+      p =
+	hash_get (dm->conf->device_config_index_by_pci_addr, pci_addr.as_u32);
+    }
+
+  if (p)
+    devconf = pool_elt_at_index (dm->conf->dev_confs, p[0]);
+  else
+    devconf = &dm->conf->default_devconf;
+
+  if (devconf->hqos_enabled == 0)
+    {
+      vlib_cli_output (vm, "HQoS disabled for this interface");
+      goto done;
+    }
+
+  /*
+   * Figure out which queue to query.  cf rte_sched_port_qindex.  (Not sure why
+   * that method isn't made public by DPDK - how _should_ we get the queue ID?)
+   */
+  qindex = subport * devconf->hqos.port.n_pipes_per_subport + pipe;
+  qindex = qindex * RTE_SCHED_TRAFFIC_CLASSES_PER_PIPE + tc;
+  qindex = qindex * RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS + tc_q;
+
+  if (rte_sched_queue_read_stats (xd->hqos_ht->hqos, qindex, &stats, &qlen) !=
+      0)
+    {
+      error = clib_error_return (0, "failed to read stats");
+      goto done;
+    }
+
+  vlib_cli_output (vm, "%=24s%=16s", "Stats Parameter", "Value");
+  vlib_cli_output (vm, "%=24s%=16d", "Packets", stats.n_pkts);
+  vlib_cli_output (vm, "%=24s%=16d", "Packets dropped", stats.n_pkts_dropped);
+#ifdef RTE_SCHED_RED
+  vlib_cli_output (vm, "%=24s%=16d", "Packets dropped (RED)",
+		   stats.n_pkts_red_dropped);
+#endif
+  vlib_cli_output (vm, "%=24s%=16d", "Bytes", stats.n_bytes);
+  vlib_cli_output (vm, "%=24s%=16d", "Bytes dropped", stats.n_bytes_dropped);
+
+#else
+
+  /* Get a line of input */
+  if (!unformat_user (input, unformat_line_input, line_input))
+    return 0;
+
+  vlib_cli_output (vm, "RTE_SCHED_COLLECT_STATS disabled in DPDK");
+  goto done;
+
+#endif
+
+done:
+  unformat_free (line_input);
+
+  return error;
+}
+
+/*?
+ * This command is used to display statistics associated with a HQoS traffic class
+ * queue.
+ *
+ * @note
+ * Statistic collection by the scheduler is disabled by default in DPDK. In order to
+ * turn it on, add the following line to '../vpp/dpdk/Makefile':
+ * - $(call set,RTE_SCHED_COLLECT_STATS,y)
+ *
+ * @cliexpar
+ * Example of how to display statistics of HQoS a HQoS traffic class queue:
+ * @cliexstart{show dpdk hqos queue GigabitEthernet0/9/0 subport 0 pipe 3181 tc 0 tc_q 0}
+ *      Stats Parameter          Value
+ *          Packets               140
+ *      Packets dropped            0
+ *           Bytes               8400
+ *       Bytes dropped             0
+ * @cliexend
+?*/
+/* *INDENT-OFF* */
+VLIB_CLI_COMMAND (cmd_show_dpdk_hqos_queue_stats, static) = {
+  .path = "show dpdk hqos queue",
+  .short_help = "show dpdk hqos queue  subport  pipe  tc  tc_q ",
+  .function = show_dpdk_hqos_queue_stats,
+};
+/* *INDENT-ON* */
+
+static clib_error_t *
+show_dpdk_version_command_fn (vlib_main_t * vm,
+			      unformat_input_t * input,
+			      vlib_cli_command_t * cmd)
+{
+#define _(a,b,c) vlib_cli_output (vm, "%-25s " b, a ":", c);
+  _("DPDK Version", "%s", rte_version ());
+  _("DPDK EAL init args", "%s", dpdk_config_main.eal_init_args_str);
+#undef _
+  return 0;
+}
+
+/*?
+ * This command is used to display the current DPDK version and
+ * the list of arguments passed to DPDK when started.
+ *
+ * @cliexpar
+ * Example of how to display how many DPDK buffer test command has allcoated:
+ * @cliexstart{show dpdk version}
+ * DPDK Version:        DPDK 16.11.0
+ * DPDK EAL init args:  -c 1 -n 4 --huge-dir /run/vpp/hugepages --file-prefix vpp -w 0000:00:08.0 -w 0000:00:09.0 --master-lcore 0 --socket-mem 256
+ * @cliexend
+?*/
+/* *INDENT-OFF* */
+VLIB_CLI_COMMAND (show_vpe_version_command, static) = {
+  .path = "show dpdk version",
+  .short_help = "show dpdk version",
+  .function = show_dpdk_version_command_fn,
+};
+/* *INDENT-ON* */
+
+clib_error_t *
+dpdk_cli_init (vlib_main_t * vm)
+{
+  return 0;
+}
+
+VLIB_INIT_FUNCTION (dpdk_cli_init);
+
+/*
+ * fd.io coding-style-patch-verification: ON
+ *
+ * Local Variables:
+ * eval: (c-set-style "gnu")
+ * End:
+ */
diff --git a/src/plugins/dpdk/device/device.c b/src/plugins/dpdk/device/device.c
new file mode 100644
index 00000000..50b26689
--- /dev/null
+++ b/src/plugins/dpdk/device/device.c
@@ -0,0 +1,852 @@
+/*
+ * Copyright (c) 2015 Cisco and/or its affiliates.
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at:
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+#include 
+#include 
+#include 
+#include 
+#include 
+
+#include 
+#include 
+
+#include 
+#include 
+
+#define foreach_dpdk_tx_func_error			\
+  _(BAD_RETVAL, "DPDK tx function returned an error")	\
+  _(RING_FULL, "Tx packet drops (ring full)")	        \
+  _(PKT_DROP, "Tx packet drops (dpdk tx failure)")	\
+  _(REPL_FAIL, "Tx packet drops (replication failure)")
+
+typedef enum
+{
+#define _(f,s) DPDK_TX_FUNC_ERROR_##f,
+  foreach_dpdk_tx_func_error
+#undef _
+    DPDK_TX_FUNC_N_ERROR,
+} dpdk_tx_func_error_t;
+
+static char *dpdk_tx_func_error_strings[] = {
+#define _(n,s) s,
+  foreach_dpdk_tx_func_error
+#undef _
+};
+
+clib_error_t *
+dpdk_set_mac_address (vnet_hw_interface_t * hi, char *address)
+{
+  int error;
+  dpdk_main_t *dm = &dpdk_main;
+  dpdk_device_t *xd = vec_elt_at_index (dm->devices, hi->dev_instance);
+
+  error = rte_eth_dev_default_mac_addr_set (xd->device_index,
+					    (struct ether_addr *) address);
+
+  if (error)
+    {
+      return clib_error_return (0, "mac address set failed: %d", error);
+    }
+  else
+    {
+      vec_reset_length (xd->default_mac_address);
+      vec_add (xd->default_mac_address, address, sizeof (address));
+      return NULL;
+    }
+}
+
+clib_error_t *
+dpdk_set_mc_filter (vnet_hw_interface_t * hi,
+		    struct ether_addr mc_addr_vec[], int naddr)
+{
+  int error;
+  dpdk_main_t *dm = &dpdk_main;
+  dpdk_device_t *xd = vec_elt_at_index (dm->devices, hi->dev_instance);
+
+  error = rte_eth_dev_set_mc_addr_list (xd->device_index, mc_addr_vec, naddr);
+
+  if (error)
+    {
+      return clib_error_return (0, "mc addr list failed: %d", error);
+    }
+  else
+    {
+      return NULL;
+    }
+}
+
+struct rte_mbuf *
+dpdk_replicate_packet_mb (vlib_buffer_t * b)
+{
+  dpdk_main_t *dm = &dpdk_main;
+  struct rte_mbuf **mbufs = 0, *s, *d;
+  u8 nb_segs;
+  unsigned socket_id = rte_socket_id ();
+  int i;
+
+  ASSERT (dm->pktmbuf_pools[socket_id]);
+  s = rte_mbuf_from_vlib_buffer (b);
+  nb_segs = s->nb_segs;
+  vec_validate (mbufs, nb_segs - 1);
+
+  if (rte_pktmbuf_alloc_bulk (dm->pktmbuf_pools[socket_id], mbufs, nb_segs))
+    {
+      vec_free (mbufs);
+      return 0;
+    }
+
+  d = mbufs[0];
+  d->nb_segs = s->nb_segs;
+  d->data_len = s->data_len;
+  d->pkt_len = s->pkt_len;
+  d->data_off = s->data_off;
+  clib_memcpy (d->buf_addr, s->buf_addr, RTE_PKTMBUF_HEADROOM + s->data_len);
+
+  for (i = 1; i < nb_segs; i++)
+    {
+      d->next = mbufs[i];
+      d = mbufs[i];
+      s = s->next;
+      d->data_len = s->data_len;
+      clib_memcpy (d->buf_addr, s->buf_addr,
+		   RTE_PKTMBUF_HEADROOM + s->data_len);
+    }
+
+  d = mbufs[0];
+  vec_free (mbufs);
+  return d;
+}
+
+static void
+dpdk_tx_trace_buffer (dpdk_main_t * dm,
+		      vlib_node_runtime_t * node,
+		      dpdk_device_t * xd,
+		      u16 queue_id, u32 buffer_index, vlib_buffer_t * buffer)
+{
+  vlib_main_t *vm = vlib_get_main ();
+  dpdk_tx_dma_trace_t *t0;
+  struct rte_mbuf *mb;
+
+  mb = rte_mbuf_from_vlib_buffer (buffer);
+
+  t0 = vlib_add_trace (vm, node, buffer, sizeof (t0[0]));
+  t0->queue_index = queue_id;
+  t0->device_index = xd->device_index;
+  t0->buffer_index = buffer_index;
+  clib_memcpy (&t0->mb, mb, sizeof (t0->mb));
+  clib_memcpy (&t0->buffer, buffer,
+	       sizeof (buffer[0]) - sizeof (buffer->pre_data));
+  clib_memcpy (t0->buffer.pre_data, buffer->data + buffer->current_data,
+	       sizeof (t0->buffer.pre_data));
+}
+
+static_always_inline void
+dpdk_validate_rte_mbuf (vlib_main_t * vm, vlib_buffer_t * b,
+			int maybe_multiseg)
+{
+  struct rte_mbuf *mb, *first_mb, *last_mb;
+
+  /* buffer is coming from non-dpdk source so we need to init
+     rte_mbuf header */
+  if (PREDICT_FALSE ((b->flags & VLIB_BUFFER_EXT_HDR_VALID) == 0))
+    {
+      vlib_buffer_t *b2 = b;
+      last_mb = mb = rte_mbuf_from_vlib_buffer (b2);
+      rte_pktmbuf_reset (mb);
+      while (maybe_multiseg && (b2->flags & VLIB_BUFFER_NEXT_PRESENT))
+	{
+	  b2 = vlib_get_buffer (vm, b2->next_buffer);
+	  mb = rte_mbuf_from_vlib_buffer (b2);
+	  rte_pktmbuf_reset (mb);
+	}
+    }
+
+  last_mb = first_mb = mb = rte_mbuf_from_vlib_buffer (b);
+  first_mb->nb_segs = 1;
+  mb->data_len = b->current_length;
+  mb->pkt_len = maybe_multiseg ? vlib_buffer_length_in_chain (vm, b) :
+    b->current_length;
+  mb->data_off = VLIB_BUFFER_PRE_DATA_SIZE + b->current_data;
+
+  while (maybe_multiseg && (b->flags & VLIB_BUFFER_NEXT_PRESENT))
+    {
+      b = vlib_get_buffer (vm, b->next_buffer);
+      mb = rte_mbuf_from_vlib_buffer (b);
+      last_mb->next = mb;
+      last_mb = mb;
+      mb->data_len = b->current_length;
+      mb->pkt_len = b->current_length;
+      mb->data_off = VLIB_BUFFER_PRE_DATA_SIZE + b->current_data;
+      first_mb->nb_segs++;
+      if (PREDICT_FALSE (b->n_add_refs))
+	{
+	  rte_mbuf_refcnt_update (mb, b->n_add_refs);
+	  b->n_add_refs = 0;
+	}
+    }
+}
+
+/*
+ * This function calls the dpdk's tx_burst function to transmit the packets
+ * on the tx_vector. It manages a lock per-device if the device does not
+ * support multiple queues. It returns the number of packets untransmitted
+ * on the tx_vector. If all packets are transmitted (the normal case), the
+ * function returns 0.
+ *
+ * The function assumes there is at least one packet on the tx_vector.
+ */
+static_always_inline
+  u32 tx_burst_vector_internal (vlib_main_t * vm,
+				dpdk_device_t * xd,
+				struct rte_mbuf **tx_vector)
+{
+  dpdk_main_t *dm = &dpdk_main;
+  u32 n_packets;
+  u32 tx_head;
+  u32 tx_tail;
+  u32 n_retry;
+  int rv;
+  int queue_id;
+  tx_ring_hdr_t *ring;
+
+  ring = vec_header (tx_vector, sizeof (*ring));
+
+  n_packets = ring->tx_head - ring->tx_tail;
+
+  tx_head = ring->tx_head % xd->nb_tx_desc;
+
+  /*
+   * Ensure rte_eth_tx_burst is not called with 0 packets, which can lead to
+   * unpredictable results.
+   */
+  ASSERT (n_packets > 0);
+
+  /*
+   * Check for tx_vector overflow. If this fails it is a system configuration
+   * error. The ring should be sized big enough to handle the largest un-flowed
+   * off burst from a traffic manager. A larger size also helps performance
+   * a bit because it decreases the probability of having to issue two tx_burst
+   * calls due to a ring wrap.
+   */
+  ASSERT (n_packets < xd->nb_tx_desc);
+  ASSERT (ring->tx_tail == 0);
+
+  n_retry = 16;
+  queue_id = vm->cpu_index;
+
+  do
+    {
+      /* start the burst at the tail */
+      tx_tail = ring->tx_tail % xd->nb_tx_desc;
+
+      /*
+       * This device only supports one TX queue,
+       * and we're running multi-threaded...
+       */
+      if (PREDICT_FALSE (xd->lockp != 0))
+	{
+	  queue_id = queue_id % xd->tx_q_used;
+	  while (__sync_lock_test_and_set (xd->lockp[queue_id], 1))
+	    /* zzzz */
+	    queue_id = (queue_id + 1) % xd->tx_q_used;
+	}
+
+      if (PREDICT_FALSE (xd->flags & DPDK_DEVICE_FLAG_HQOS))	/* HQoS ON */
+	{
+	  /* no wrap, transmit in one burst */
+	  dpdk_device_hqos_per_worker_thread_t *hqos =
+	    &xd->hqos_wt[vm->cpu_index];
+
+	  ASSERT (hqos->swq != NULL);
+
+	  dpdk_hqos_metadata_set (hqos,
+				  &tx_vector[tx_tail], tx_head - tx_tail);
+	  rv = rte_ring_sp_enqueue_burst (hqos->swq,
+					  (void **) &tx_vector[tx_tail],
+					  (uint16_t) (tx_head - tx_tail));
+	}
+      else if (PREDICT_TRUE (xd->flags & DPDK_DEVICE_FLAG_PMD))
+	{
+	  /* no wrap, transmit in one burst */
+	  rv = rte_eth_tx_burst (xd->device_index,
+				 (uint16_t) queue_id,
+				 &tx_vector[tx_tail],
+				 (uint16_t) (tx_head - tx_tail));
+	}
+      else
+	{
+	  ASSERT (0);
+	  rv = 0;
+	}
+
+      if (PREDICT_FALSE (xd->lockp != 0))
+	*xd->lockp[queue_id] = 0;
+
+      if (PREDICT_FALSE (rv < 0))
+	{
+	  // emit non-fatal message, bump counter
+	  vnet_main_t *vnm = dm->vnet_main;
+	  vnet_interface_main_t *im = &vnm->interface_main;
+	  u32 node_index;
+
+	  node_index = vec_elt_at_index (im->hw_interfaces,
+					 xd->vlib_hw_if_index)->tx_node_index;
+
+	  vlib_error_count (vm, node_index, DPDK_TX_FUNC_ERROR_BAD_RETVAL, 1);
+	  clib_warning ("rte_eth_tx_burst[%d]: error %d", xd->device_index,
+			rv);
+	  return n_packets;	// untransmitted packets
+	}
+      ring->tx_tail += (u16) rv;
+      n_packets -= (uint16_t) rv;
+    }
+  while (rv && n_packets && (n_retry > 0));
+
+  return n_packets;
+}
+
+static_always_inline void
+dpdk_prefetch_buffer_by_index (vlib_main_t * vm, u32 bi)
+{
+  vlib_buffer_t *b;
+  struct rte_mbuf *mb;
+  b = vlib_get_buffer (vm, bi);
+  mb = rte_mbuf_from_vlib_buffer (b);
+  CLIB_PREFETCH (mb, CLIB_CACHE_LINE_BYTES, LOAD);
+  CLIB_PREFETCH (b, CLIB_CACHE_LINE_BYTES, LOAD);
+}
+
+static_always_inline void
+dpdk_buffer_recycle (vlib_main_t * vm, vlib_node_runtime_t * node,
+		     vlib_buffer_t * b, u32 bi, struct rte_mbuf **mbp)
+{
+  dpdk_main_t *dm = &dpdk_main;
+  u32 my_cpu = vm->cpu_index;
+  struct rte_mbuf *mb_new;
+
+  if (PREDICT_FALSE (b->flags & VLIB_BUFFER_RECYCLE) == 0)
+    return;
+
+  mb_new = dpdk_replicate_packet_mb (b);
+  if (PREDICT_FALSE (mb_new == 0))
+    {
+      vlib_error_count (vm, node->node_index,
+			DPDK_TX_FUNC_ERROR_REPL_FAIL, 1);
+      b->flags |= VLIB_BUFFER_REPL_FAIL;
+    }
+  else
+    *mbp = mb_new;
+
+  vec_add1 (dm->recycle[my_cpu], bi);
+}
+
+/*
+ * Transmits the packets on the frame to the interface associated with the
+ * node. It first copies packets on the frame to a tx_vector containing the
+ * rte_mbuf pointers. It then passes this vector to tx_burst_vector_internal
+ * which calls the dpdk tx_burst function.
+ */
+static uword
+dpdk_interface_tx (vlib_main_t * vm,
+		   vlib_node_runtime_t * node, vlib_frame_t * f)
+{
+  dpdk_main_t *dm = &dpdk_main;
+  vnet_interface_output_runtime_t *rd = (void *) node->runtime_data;
+  dpdk_device_t *xd = vec_elt_at_index (dm->devices, rd->dev_instance);
+  u32 n_packets = f->n_vectors;
+  u32 n_left;
+  u32 *from;
+  struct rte_mbuf **tx_vector;
+  u16 i;
+  u16 nb_tx_desc = xd->nb_tx_desc;
+  int queue_id;
+  u32 my_cpu;
+  u32 tx_pkts = 0;
+  tx_ring_hdr_t *ring;
+  u32 n_on_ring;
+
+  my_cpu = vm->cpu_index;
+
+  queue_id = my_cpu;
+
+  tx_vector = xd->tx_vectors[queue_id];
+  ring = vec_header (tx_vector, sizeof (*ring));
+
+  n_on_ring = ring->tx_head - ring->tx_tail;
+  from = vlib_frame_vector_args (f);
+
+  ASSERT (n_packets <= VLIB_FRAME_SIZE);
+
+  if (PREDICT_FALSE (n_on_ring + n_packets > nb_tx_desc))
+    {
+      /*
+       * Overflowing the ring should never happen.
+       * If it does then drop the whole frame.
+       */
+      vlib_error_count (vm, node->node_index, DPDK_TX_FUNC_ERROR_RING_FULL,
+			n_packets);
+
+      while (n_packets--)
+	{
+	  u32 bi0 = from[n_packets];
+	  vlib_buffer_t *b0 = vlib_get_buffer (vm, bi0);
+	  struct rte_mbuf *mb0 = rte_mbuf_from_vlib_buffer (b0);
+	  rte_pktmbuf_free (mb0);
+	}
+      return n_on_ring;
+    }
+
+  if (PREDICT_FALSE (dm->tx_pcap_enable))
+    {
+      n_left = n_packets;
+      while (n_left > 0)
+	{
+	  u32 bi0 = from[0];
+	  vlib_buffer_t *b0 = vlib_get_buffer (vm, bi0);
+	  if (dm->pcap_sw_if_index == 0 ||
+	      dm->pcap_sw_if_index == vnet_buffer (b0)->sw_if_index[VLIB_TX])
+	    pcap_add_buffer (&dm->pcap_main, vm, bi0, 512);
+	  from++;
+	  n_left--;
+	}
+    }
+
+  from = vlib_frame_vector_args (f);
+  n_left = n_packets;
+  i = ring->tx_head % nb_tx_desc;
+
+  while (n_left >= 8)
+    {
+      u32 bi0, bi1, bi2, bi3;
+      struct rte_mbuf *mb0, *mb1, *mb2, *mb3;
+      vlib_buffer_t *b0, *b1, *b2, *b3;
+      u32 or_flags;
+
+      dpdk_prefetch_buffer_by_index (vm, from[4]);
+      dpdk_prefetch_buffer_by_index (vm, from[5]);
+      dpdk_prefetch_buffer_by_index (vm, from[6]);
+      dpdk_prefetch_buffer_by_index (vm, from[7]);
+
+      bi0 = from[0];
+      bi1 = from[1];
+      bi2 = from[2];
+      bi3 = from[3];
+      from += 4;
+
+      b0 = vlib_get_buffer (vm, bi0);
+      b1 = vlib_get_buffer (vm, bi1);
+      b2 = vlib_get_buffer (vm, bi2);
+      b3 = vlib_get_buffer (vm, bi3);
+
+      or_flags = b0->flags | b1->flags | b2->flags | b3->flags;
+
+      if (or_flags & VLIB_BUFFER_NEXT_PRESENT)
+	{
+	  dpdk_validate_rte_mbuf (vm, b0, 1);
+	  dpdk_validate_rte_mbuf (vm, b1, 1);
+	  dpdk_validate_rte_mbuf (vm, b2, 1);
+	  dpdk_validate_rte_mbuf (vm, b3, 1);
+	}
+      else
+	{
+	  dpdk_validate_rte_mbuf (vm, b0, 0);
+	  dpdk_validate_rte_mbuf (vm, b1, 0);
+	  dpdk_validate_rte_mbuf (vm, b2, 0);
+	  dpdk_validate_rte_mbuf (vm, b3, 0);
+	}
+
+      mb0 = rte_mbuf_from_vlib_buffer (b0);
+      mb1 = rte_mbuf_from_vlib_buffer (b1);
+      mb2 = rte_mbuf_from_vlib_buffer (b2);
+      mb3 = rte_mbuf_from_vlib_buffer (b3);
+
+      if (PREDICT_FALSE (or_flags & VLIB_BUFFER_RECYCLE))
+	{
+	  dpdk_buffer_recycle (vm, node, b0, bi0, &mb0);
+	  dpdk_buffer_recycle (vm, node, b1, bi1, &mb1);
+	  dpdk_buffer_recycle (vm, node, b2, bi2, &mb2);
+	  dpdk_buffer_recycle (vm, node, b3, bi3, &mb3);
+
+	  /* dont enqueue packets if replication failed as they must
+	     be sent back to recycle */
+	  if (PREDICT_TRUE ((b0->flags & VLIB_BUFFER_REPL_FAIL) == 0))
+	    tx_vector[i++ % nb_tx_desc] = mb0;
+	  if (PREDICT_TRUE ((b1->flags & VLIB_BUFFER_REPL_FAIL) == 0))
+	    tx_vector[i++ % nb_tx_desc] = mb1;
+	  if (PREDICT_TRUE ((b2->flags & VLIB_BUFFER_REPL_FAIL) == 0))
+	    tx_vector[i++ % nb_tx_desc] = mb2;
+	  if (PREDICT_TRUE ((b3->flags & VLIB_BUFFER_REPL_FAIL) == 0))
+	    tx_vector[i++ % nb_tx_desc] = mb3;
+	}
+      else
+	{
+	  if (PREDICT_FALSE (i + 3 >= nb_tx_desc))
+	    {
+	      tx_vector[i++ % nb_tx_desc] = mb0;
+	      tx_vector[i++ % nb_tx_desc] = mb1;
+	      tx_vector[i++ % nb_tx_desc] = mb2;
+	      tx_vector[i++ % nb_tx_desc] = mb3;
+	      i %= nb_tx_desc;
+	    }
+	  else
+	    {
+	      tx_vector[i++] = mb0;
+	      tx_vector[i++] = mb1;
+	      tx_vector[i++] = mb2;
+	      tx_vector[i++] = mb3;
+	    }
+	}
+
+
+      if (PREDICT_FALSE (node->flags & VLIB_NODE_FLAG_TRACE))
+	{
+	  if (b0->flags & VLIB_BUFFER_IS_TRACED)
+	    dpdk_tx_trace_buffer (dm, node, xd, queue_id, bi0, b0);
+	  if (b1->flags & VLIB_BUFFER_IS_TRACED)
+	    dpdk_tx_trace_buffer (dm, node, xd, queue_id, bi1, b1);
+	  if (b2->flags & VLIB_BUFFER_IS_TRACED)
+	    dpdk_tx_trace_buffer (dm, node, xd, queue_id, bi2, b2);
+	  if (b3->flags & VLIB_BUFFER_IS_TRACED)
+	    dpdk_tx_trace_buffer (dm, node, xd, queue_id, bi3, b3);
+	}
+
+      n_left -= 4;
+    }
+  while (n_left > 0)
+    {
+      u32 bi0;
+      struct rte_mbuf *mb0;
+      vlib_buffer_t *b0;
+
+      bi0 = from[0];
+      from++;
+
+      b0 = vlib_get_buffer (vm, bi0);
+
+      dpdk_validate_rte_mbuf (vm, b0, 1);
+
+      mb0 = rte_mbuf_from_vlib_buffer (b0);
+      dpdk_buffer_recycle (vm, node, b0, bi0, &mb0);
+
+      if (PREDICT_FALSE (node->flags & VLIB_NODE_FLAG_TRACE))
+	if (b0->flags & VLIB_BUFFER_IS_TRACED)
+	  dpdk_tx_trace_buffer (dm, node, xd, queue_id, bi0, b0);
+
+      if (PREDICT_TRUE ((b0->flags & VLIB_BUFFER_REPL_FAIL) == 0))
+	{
+	  tx_vector[i % nb_tx_desc] = mb0;
+	  i++;
+	}
+      n_left--;
+    }
+
+  /* account for additional packets in the ring */
+  ring->tx_head += n_packets;
+  n_on_ring = ring->tx_head - ring->tx_tail;
+
+  /* transmit as many packets as possible */
+  n_packets = tx_burst_vector_internal (vm, xd, tx_vector);
+
+  /*
+   * tx_pkts is the number of packets successfully transmitted
+   * This is the number originally on ring minus the number remaining on ring
+   */
+  tx_pkts = n_on_ring - n_packets;
+
+  {
+    /* If there is no callback then drop any non-transmitted packets */
+    if (PREDICT_FALSE (n_packets))
+      {
+	vlib_simple_counter_main_t *cm;
+	vnet_main_t *vnm = vnet_get_main ();
+
+	cm = vec_elt_at_index (vnm->interface_main.sw_if_counters,
+			       VNET_INTERFACE_COUNTER_TX_ERROR);
+
+	vlib_increment_simple_counter (cm, my_cpu, xd->vlib_sw_if_index,
+				       n_packets);
+
+	vlib_error_count (vm, node->node_index, DPDK_TX_FUNC_ERROR_PKT_DROP,
+			  n_packets);
+
+	while (n_packets--)
+	  rte_pktmbuf_free (tx_vector[ring->tx_tail + n_packets]);
+      }
+
+    /* Reset head/tail to avoid unnecessary wrap */
+    ring->tx_head = 0;
+    ring->tx_tail = 0;
+  }
+
+  /* Recycle replicated buffers */
+  if (PREDICT_FALSE (vec_len (dm->recycle[my_cpu])))
+    {
+      vlib_buffer_free (vm, dm->recycle[my_cpu],
+			vec_len (dm->recycle[my_cpu]));
+      _vec_len (dm->recycle[my_cpu]) = 0;
+    }
+
+  ASSERT (ring->tx_head >= ring->tx_tail);
+
+  return tx_pkts;
+}
+
+static void
+dpdk_clear_hw_interface_counters (u32 instance)
+{
+  dpdk_main_t *dm = &dpdk_main;
+  dpdk_device_t *xd = vec_elt_at_index (dm->devices, instance);
+
+  /*
+   * Set the "last_cleared_stats" to the current stats, so that
+   * things appear to clear from a display perspective.
+   */
+  dpdk_update_counters (xd, vlib_time_now (dm->vlib_main));
+
+  clib_memcpy (&xd->last_cleared_stats, &xd->stats, sizeof (xd->stats));
+  clib_memcpy (xd->last_cleared_xstats, xd->xstats,
+	       vec_len (xd->last_cleared_xstats) *
+	       sizeof (xd->last_cleared_xstats[0]));
+
+}
+
+static clib_error_t *
+dpdk_interface_admin_up_down (vnet_main_t * vnm, u32 hw_if_index, u32 flags)
+{
+  vnet_hw_interface_t *hif = vnet_get_hw_interface (vnm, hw_if_index);
+  uword is_up = (flags & VNET_SW_INTERFACE_FLAG_ADMIN_UP) != 0;
+  dpdk_main_t *dm = &dpdk_main;
+  dpdk_device_t *xd = vec_elt_at_index (dm->devices, hif->dev_instance);
+  int rv = 0;
+
+  if (is_up)
+    {
+      f64 now = vlib_time_now (dm->vlib_main);
+
+      if ((xd->flags & DPDK_DEVICE_FLAG_ADMIN_UP) == 0)
+	{
+	  rv = rte_eth_dev_start (xd->device_index);
+	  if (!rv && xd->default_mac_address)
+	    rv = rte_eth_dev_default_mac_addr_set (xd->device_index,
+						   (struct ether_addr *)
+						   xd->default_mac_address);
+	}
+
+      if (xd->flags & DPDK_DEVICE_FLAG_PROMISC)
+	rte_eth_promiscuous_enable (xd->device_index);
+      else
+	rte_eth_promiscuous_disable (xd->device_index);
+
+      rte_eth_allmulticast_enable (xd->device_index);
+      xd->flags |= DPDK_DEVICE_FLAG_ADMIN_UP;
+      dpdk_update_counters (xd, now);
+      dpdk_update_link_state (xd, now);
+    }
+  else
+    {
+      xd->flags &= ~DPDK_DEVICE_FLAG_ADMIN_UP;
+
+      rte_eth_allmulticast_disable (xd->device_index);
+      vnet_hw_interface_set_flags (vnm, xd->vlib_hw_if_index, 0);
+      rte_eth_dev_stop (xd->device_index);
+
+      /* For bonded interface, stop slave links */
+      if (xd->pmd == VNET_DPDK_PMD_BOND)
+	{
+	  u8 slink[16];
+	  int nlink = rte_eth_bond_slaves_get (xd->device_index, slink, 16);
+	  while (nlink >= 1)
+	    {
+	      u8 dpdk_port = slink[--nlink];
+	      rte_eth_dev_stop (dpdk_port);
+	    }
+	}
+    }
+
+  if (rv < 0)
+    clib_warning ("rte_eth_dev_%s error: %d", is_up ? "start" : "stop", rv);
+
+  return /* no error */ 0;
+}
+
+/*
+ * Dynamically redirect all pkts from a specific interface
+ * to the specified node
+ */
+static void
+dpdk_set_interface_next_node (vnet_main_t * vnm, u32 hw_if_index,
+			      u32 node_index)
+{
+  dpdk_main_t *xm = &dpdk_main;
+  vnet_hw_interface_t *hw = vnet_get_hw_interface (vnm, hw_if_index);
+  dpdk_device_t *xd = vec_elt_at_index (xm->devices, hw->dev_instance);
+
+  /* Shut off redirection */
+  if (node_index == ~0)
+    {
+      xd->per_interface_next_index = node_index;
+      return;
+    }
+
+  xd->per_interface_next_index =
+    vlib_node_add_next (xm->vlib_main, dpdk_input_node.index, node_index);
+}
+
+
+static clib_error_t *
+dpdk_subif_add_del_function (vnet_main_t * vnm,
+			     u32 hw_if_index,
+			     struct vnet_sw_interface_t *st, int is_add)
+{
+  dpdk_main_t *xm = &dpdk_main;
+  vnet_hw_interface_t *hw = vnet_get_hw_interface (vnm, hw_if_index);
+  dpdk_device_t *xd = vec_elt_at_index (xm->devices, hw->dev_instance);
+  vnet_sw_interface_t *t = (vnet_sw_interface_t *) st;
+  int r, vlan_offload;
+  u32 prev_subifs = xd->num_subifs;
+  clib_error_t *err = 0;
+
+  if (is_add)
+    xd->num_subifs++;
+  else if (xd->num_subifs)
+    xd->num_subifs--;
+
+  if ((xd->flags & DPDK_DEVICE_FLAG_PMD) == 0)
+    goto done;
+
+  /* currently we program VLANS only for IXGBE VF and I40E VF */
+  if ((xd->pmd != VNET_DPDK_PMD_IXGBEVF) && (xd->pmd != VNET_DPDK_PMD_I40EVF))
+    goto done;
+
+  if (t->sub.eth.flags.no_tags == 1)
+    goto done;
+
+  if ((t->sub.eth.flags.one_tag != 1) || (t->sub.eth.flags.exact_match != 1))
+    {
+      xd->num_subifs = prev_subifs;
+      err = clib_error_return (0, "unsupported VLAN setup");
+      goto done;
+    }
+
+  vlan_offload = rte_eth_dev_get_vlan_offload (xd->device_index);
+  vlan_offload |= ETH_VLAN_FILTER_OFFLOAD;
+
+  if ((r = rte_eth_dev_set_vlan_offload (xd->device_index, vlan_offload)))
+    {
+      xd->num_subifs = prev_subifs;
+      err = clib_error_return (0, "rte_eth_dev_set_vlan_offload[%d]: err %d",
+			       xd->device_index, r);
+      goto done;
+    }
+
+
+  if ((r =
+       rte_eth_dev_vlan_filter (xd->device_index, t->sub.eth.outer_vlan_id,
+				is_add)))
+    {
+      xd->num_subifs = prev_subifs;
+      err = clib_error_return (0, "rte_eth_dev_vlan_filter[%d]: err %d",
+			       xd->device_index, r);
+      goto done;
+    }
+
+done:
+  if (xd->num_subifs)
+    xd->flags |= DPDK_DEVICE_FLAG_HAVE_SUBIF;
+  else
+    xd->flags &= ~DPDK_DEVICE_FLAG_HAVE_SUBIF;
+
+  return err;
+}
+
+/* *INDENT-OFF* */
+VNET_DEVICE_CLASS (dpdk_device_class) = {
+  .name = "dpdk",
+  .tx_function = dpdk_interface_tx,
+  .tx_function_n_errors = DPDK_TX_FUNC_N_ERROR,
+  .tx_function_error_strings = dpdk_tx_func_error_strings,
+  .format_device_name = format_dpdk_device_name,
+  .format_device = format_dpdk_device,
+  .format_tx_trace = format_dpdk_tx_dma_trace,
+  .clear_counters = dpdk_clear_hw_interface_counters,
+  .admin_up_down_function = dpdk_interface_admin_up_down,
+  .subif_add_del_function = dpdk_subif_add_del_function,
+  .rx_redirect_to_node = dpdk_set_interface_next_node,
+  .mac_addr_change_function = dpdk_set_mac_address,
+};
+
+VLIB_DEVICE_TX_FUNCTION_MULTIARCH (dpdk_device_class, dpdk_interface_tx)
+/* *INDENT-ON* */
+
+#define UP_DOWN_FLAG_EVENT 1
+
+uword
+admin_up_down_process (vlib_main_t * vm,
+		       vlib_node_runtime_t * rt, vlib_frame_t * f)
+{
+  clib_error_t *error = 0;
+  uword event_type;
+  uword *event_data = 0;
+  u32 sw_if_index;
+  u32 flags;
+
+  while (1)
+    {
+      vlib_process_wait_for_event (vm);
+
+      event_type = vlib_process_get_events (vm, &event_data);
+
+      dpdk_main.admin_up_down_in_progress = 1;
+
+      switch (event_type)
+	{
+	case UP_DOWN_FLAG_EVENT:
+	  {
+	    if (vec_len (event_data) == 2)
+	      {
+		sw_if_index = event_data[0];
+		flags = event_data[1];
+		error =
+		  vnet_sw_interface_set_flags (vnet_get_main (), sw_if_index,
+					       flags);
+		clib_error_report (error);
+	      }
+	  }
+	  break;
+	}
+
+      vec_reset_length (event_data);
+
+      dpdk_main.admin_up_down_in_progress = 0;
+
+    }
+  return 0;			/* or not */
+}
+
+/* *INDENT-OFF* */
+VLIB_REGISTER_NODE (admin_up_down_process_node,static) = {
+    .function = admin_up_down_process,
+    .type = VLIB_NODE_TYPE_PROCESS,
+    .name = "admin-up-down-process",
+    .process_log2_n_stack_bytes = 17,  // 256KB
+};
+/* *INDENT-ON* */
+
+/*
+ * fd.io coding-style-patch-verification: ON
+ *
+ * Local Variables:
+ * eval: (c-set-style "gnu")
+ * End:
+ */
diff --git a/src/plugins/dpdk/device/dpdk.h b/src/plugins/dpdk/device/dpdk.h
new file mode 100644
index 00000000..2a1a6205
--- /dev/null
+++ b/src/plugins/dpdk/device/dpdk.h
@@ -0,0 +1,490 @@
+/*
+ * Copyright (c) 2015 Cisco and/or its affiliates.
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at:
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+#ifndef __included_dpdk_h__
+#define __included_dpdk_h__
+
+/* $$$$ We should rename always_inline -> clib_always_inline */
+#undef always_inline
+
+#include 
+
+#include 
+#include 
+#include 
+#include 
+#include 
+#include 
+#include 
+#include 
+#include 
+#include 
+#include 
+#include 
+#include 
+#include 
+#include 
+#include 
+#include 
+#include 
+#include 
+#include 
+#include 
+#include 
+#include 
+#include 
+#include 
+#include 
+#include 
+#include 
+
+#include 
+#include 
+
+#if CLIB_DEBUG > 0
+#define always_inline static inline
+#else
+#define always_inline static inline __attribute__ ((__always_inline__))
+#endif
+
+#include 
+
+#define NB_MBUF   (16<<10)
+
+extern vnet_device_class_t dpdk_device_class;
+extern vlib_node_registration_t dpdk_input_node;
+extern vlib_node_registration_t handoff_dispatch_node;
+
+#define foreach_dpdk_pmd          \
+  _ ("net_thunderx", THUNDERX)    \
+  _ ("net_e1000_em", E1000EM)     \
+  _ ("net_e1000_igb", IGB)        \
+  _ ("net_e1000_igb_vf", IGBVF)   \
+  _ ("net_ixgbe", IXGBE)          \
+  _ ("net_ixgbe_vf", IXGBEVF)     \
+  _ ("net_i40e", I40E)            \
+  _ ("net_i40e_vf", I40EVF)       \
+  _ ("net_virtio", VIRTIO)        \
+  _ ("net_enic", ENIC)            \
+  _ ("net_vmxnet3", VMXNET3)      \
+  _ ("AF_PACKET PMD", AF_PACKET)  \
+  _ ("rte_bond_pmd", BOND)        \
+  _ ("net_fm10k", FM10K)          \
+  _ ("net_cxgbe", CXGBE)          \
+  _ ("net_mlx5", MLX5)            \
+  _ ("net_dpaa2", DPAA2)
+
+typedef enum
+{
+  VNET_DPDK_PMD_NONE,
+#define _(s,f) VNET_DPDK_PMD_##f,
+  foreach_dpdk_pmd
+#undef _
+    VNET_DPDK_PMD_UNKNOWN,	/* must be last */
+} dpdk_pmd_t;
+
+typedef enum
+{
+  VNET_DPDK_PORT_TYPE_ETH_1G,
+  VNET_DPDK_PORT_TYPE_ETH_10G,
+  VNET_DPDK_PORT_TYPE_ETH_40G,
+  VNET_DPDK_PORT_TYPE_ETH_100G,
+  VNET_DPDK_PORT_TYPE_ETH_BOND,
+  VNET_DPDK_PORT_TYPE_ETH_SWITCH,
+  VNET_DPDK_PORT_TYPE_AF_PACKET,
+  VNET_DPDK_PORT_TYPE_UNKNOWN,
+} dpdk_port_type_t;
+
+/*
+ * The header for the tx_vector in dpdk_device_t.
+ * Head and tail are indexes into the tx_vector and are of type
+ * u64 so they never overflow.
+ */
+typedef struct
+{
+  u64 tx_head;
+  u64 tx_tail;
+} tx_ring_hdr_t;
+
+typedef struct
+{
+  struct rte_ring *swq;
+
+  u64 hqos_field0_slabmask;
+  u32 hqos_field0_slabpos;
+  u32 hqos_field0_slabshr;
+  u64 hqos_field1_slabmask;
+  u32 hqos_field1_slabpos;
+  u32 hqos_field1_slabshr;
+  u64 hqos_field2_slabmask;
+  u32 hqos_field2_slabpos;
+  u32 hqos_field2_slabshr;
+  u32 hqos_tc_table[64];
+} dpdk_device_hqos_per_worker_thread_t;
+
+typedef struct
+{
+  struct rte_ring **swq;
+  struct rte_mbuf **pkts_enq;
+  struct rte_mbuf **pkts_deq;
+  struct rte_sched_port *hqos;
+  u32 hqos_burst_enq;
+  u32 hqos_burst_deq;
+  u32 pkts_enq_len;
+  u32 swq_pos;
+  u32 flush_count;
+} dpdk_device_hqos_per_hqos_thread_t;
+
+typedef struct
+{
+  CLIB_CACHE_LINE_ALIGN_MARK (cacheline0);
+  volatile u32 **lockp;
+
+  /* Instance ID */
+  u32 device_index;
+
+  u32 vlib_hw_if_index;
+  u32 vlib_sw_if_index;
+
+  /* next node index if we decide to steal the rx graph arc */
+  u32 per_interface_next_index;
+
+  /* dpdk rte_mbuf rx and tx vectors, VLIB_FRAME_SIZE */
+  struct rte_mbuf ***tx_vectors;	/* one per worker thread */
+  struct rte_mbuf ***rx_vectors;
+
+  /* vector of traced contexts, per device */
+  u32 **d_trace_buffers;
+
+  dpdk_pmd_t pmd:8;
+  i8 cpu_socket;
+
+  u16 flags;
+#define DPDK_DEVICE_FLAG_ADMIN_UP           (1 << 0)
+#define DPDK_DEVICE_FLAG_PROMISC            (1 << 1)
+#define DPDK_DEVICE_FLAG_PMD                (1 << 2)
+#define DPDK_DEVICE_FLAG_PMD_SUPPORTS_PTYPE (1 << 3)
+#define DPDK_DEVICE_FLAG_MAYBE_MULTISEG     (1 << 4)
+#define DPDK_DEVICE_FLAG_HAVE_SUBIF         (1 << 5)
+#define DPDK_DEVICE_FLAG_HQOS               (1 << 6)
+
+  u16 nb_tx_desc;
+    CLIB_CACHE_LINE_ALIGN_MARK (cacheline1);
+
+  u8 *interface_name_suffix;
+
+  /* number of sub-interfaces */
+  u16 num_subifs;
+
+  /* PMD related */
+  u16 tx_q_used;
+  u16 rx_q_used;
+  u16 nb_rx_desc;
+  u16 *cpu_socket_id_by_queue;
+  struct rte_eth_conf port_conf;
+  struct rte_eth_txconf tx_conf;
+
+  /* HQoS related */
+  dpdk_device_hqos_per_worker_thread_t *hqos_wt;
+  dpdk_device_hqos_per_hqos_thread_t *hqos_ht;
+
+  /* af_packet */
+  u8 af_packet_port_id;
+
+  struct rte_eth_link link;
+  f64 time_last_link_update;
+
+  struct rte_eth_stats stats;
+  struct rte_eth_stats last_stats;
+  struct rte_eth_stats last_cleared_stats;
+  struct rte_eth_xstat *xstats;
+  struct rte_eth_xstat *last_cleared_xstats;
+  f64 time_last_stats_update;
+  dpdk_port_type_t port_type;
+
+  /* mac address */
+  u8 *default_mac_address;
+} dpdk_device_t;
+
+#define DPDK_STATS_POLL_INTERVAL      (10.0)
+#define DPDK_MIN_STATS_POLL_INTERVAL  (0.001)	/* 1msec */
+
+#define DPDK_LINK_POLL_INTERVAL       (3.0)
+#define DPDK_MIN_LINK_POLL_INTERVAL   (0.001)	/* 1msec */
+
+typedef struct
+{
+  u32 device;
+  u16 queue_id;
+} dpdk_device_and_queue_t;
+
+#ifndef DPDK_HQOS_DBG_BYPASS
+#define DPDK_HQOS_DBG_BYPASS 0
+#endif
+
+#ifndef HQOS_FLUSH_COUNT_THRESHOLD
+#define HQOS_FLUSH_COUNT_THRESHOLD              100000
+#endif
+
+typedef struct dpdk_device_config_hqos_t
+{
+  u32 hqos_thread;
+  u32 hqos_thread_valid;
+
+  u32 swq_size;
+  u32 burst_enq;
+  u32 burst_deq;
+
+  u32 pktfield0_slabpos;
+  u32 pktfield1_slabpos;
+  u32 pktfield2_slabpos;
+  u64 pktfield0_slabmask;
+  u64 pktfield1_slabmask;
+  u64 pktfield2_slabmask;
+  u32 tc_table[64];
+
+  struct rte_sched_port_params port;
+  struct rte_sched_subport_params *subport;
+  struct rte_sched_pipe_params *pipe;
+  uint32_t *pipe_map;
+} dpdk_device_config_hqos_t;
+
+int dpdk_hqos_validate_mask (u64 mask, u32 n);
+void dpdk_device_config_hqos_pipe_profile_default (dpdk_device_config_hqos_t *
+						   hqos, u32 pipe_profile_id);
+void dpdk_device_config_hqos_default (dpdk_device_config_hqos_t * hqos);
+clib_error_t *dpdk_port_setup_hqos (dpdk_device_t * xd,
+				    dpdk_device_config_hqos_t * hqos);
+void dpdk_hqos_metadata_set (dpdk_device_hqos_per_worker_thread_t * hqos,
+			     struct rte_mbuf **pkts, u32 n_pkts);
+
+#define foreach_dpdk_device_config_item \
+  _ (num_rx_queues) \
+  _ (num_tx_queues) \
+  _ (num_rx_desc) \
+  _ (num_tx_desc) \
+  _ (rss_fn)
+
+typedef struct
+{
+  vlib_pci_addr_t pci_addr;
+  u8 is_blacklisted;
+  u8 vlan_strip_offload;
+#define DPDK_DEVICE_VLAN_STRIP_DEFAULT 0
+#define DPDK_DEVICE_VLAN_STRIP_OFF 1
+#define DPDK_DEVICE_VLAN_STRIP_ON  2
+
+#define _(x) uword x;
+    foreach_dpdk_device_config_item
+#undef _
+    clib_bitmap_t * workers;
+  u32 hqos_enabled;
+  dpdk_device_config_hqos_t hqos;
+} dpdk_device_config_t;
+
+typedef struct
+{
+
+  /* Config stuff */
+  u8 **eal_init_args;
+  u8 *eal_init_args_str;
+  u8 *uio_driver_name;
+  u8 no_multi_seg;
+  u8 enable_tcp_udp_checksum;
+  u8 cryptodev;
+
+  /* Required config parameters */
+  u8 coremask_set_manually;
+  u8 nchannels_set_manually;
+  u32 coremask;
+  u32 nchannels;
+  u32 num_mbufs;
+  u8 num_kni;			/* while kni_init allows u32, port_id in callback fn is only u8 */
+
+  /*
+   * format interface names ala xxxEthernet%d/%d/%d instead of
+   * xxxEthernet%x/%x/%x.
+   */
+  u8 interface_name_format_decimal;
+
+  /* per-device config */
+  dpdk_device_config_t default_devconf;
+  dpdk_device_config_t *dev_confs;
+  uword *device_config_index_by_pci_addr;
+
+} dpdk_config_main_t;
+
+dpdk_config_main_t dpdk_config_main;
+
+typedef struct
+{
+
+  /* Devices */
+  dpdk_device_t *devices;
+  dpdk_device_and_queue_t **devices_by_cpu;
+  dpdk_device_and_queue_t **devices_by_hqos_cpu;
+
+  /* per-thread recycle lists */
+  u32 **recycle;
+
+  /* buffer flags template, configurable to enable/disable tcp / udp cksum */
+  u32 buffer_flags_template;
+
+  /* vlib buffer free list, must be same size as an rte_mbuf */
+  u32 vlib_buffer_free_list_index;
+
+  /* Ethernet input node index */
+  u32 ethernet_input_node_index;
+
+  /* pcap tracing [only works if (CLIB_DEBUG > 0)] */
+  int tx_pcap_enable;
+  pcap_main_t pcap_main;
+  u8 *pcap_filename;
+  u32 pcap_sw_if_index;
+  u32 pcap_pkts_to_capture;
+
+  /* hashes */
+  uword *dpdk_device_by_kni_port_id;
+  uword *vu_sw_if_index_by_listener_fd;
+  uword *vu_sw_if_index_by_sock_fd;
+  u32 *vu_inactive_interfaces_device_index;
+
+  /*
+   * flag indicating that a posted admin up/down
+   * (via post_sw_interface_set_flags) is in progress
+   */
+  u8 admin_up_down_in_progress;
+
+  u8 use_rss;
+
+  /* which cpus are running dpdk-input */
+  int input_cpu_first_index;
+  int input_cpu_count;
+
+  /* which cpus are running I/O TX */
+  int hqos_cpu_first_index;
+  int hqos_cpu_count;
+
+  /* control interval of dpdk link state and stat polling */
+  f64 link_state_poll_interval;
+  f64 stat_poll_interval;
+
+  /* Sleep for this many MS after each device poll */
+  u32 poll_sleep;
+
+  /* convenience */
+  vlib_main_t *vlib_main;
+  vnet_main_t *vnet_main;
+  dpdk_config_main_t *conf;
+
+  /* mempool */
+  struct rte_mempool **pktmbuf_pools;
+
+  /* API message ID base */
+  u16 msg_id_base;
+} dpdk_main_t;
+
+dpdk_main_t dpdk_main;
+
+typedef struct
+{
+  u32 buffer_index;
+  u16 device_index;
+  u8 queue_index;
+  struct rte_mbuf mb;
+  /* Copy of VLIB buffer; packet data stored in pre_data. */
+  vlib_buffer_t buffer;
+} dpdk_tx_dma_trace_t;
+
+typedef struct
+{
+  u32 buffer_index;
+  u16 device_index;
+  u16 queue_index;
+  struct rte_mbuf mb;
+  vlib_buffer_t buffer;		/* Copy of VLIB buffer; pkt data stored in pre_data. */
+  u8 data[256];			/* First 256 data bytes, used for hexdump */
+} dpdk_rx_dma_trace_t;
+
+void vnet_buffer_needs_dpdk_mb (vlib_buffer_t * b);
+
+clib_error_t *dpdk_set_mac_address (vnet_hw_interface_t * hi, char *address);
+
+clib_error_t *dpdk_set_mc_filter (vnet_hw_interface_t * hi,
+				  struct ether_addr mc_addr_vec[], int naddr);
+
+void dpdk_thread_input (dpdk_main_t * dm, dpdk_device_t * xd);
+
+clib_error_t *dpdk_port_setup (dpdk_main_t * dm, dpdk_device_t * xd);
+
+u32 dpdk_interface_tx_vector (vlib_main_t * vm, u32 dev_instance);
+
+struct rte_mbuf *dpdk_replicate_packet_mb (vlib_buffer_t * b);
+struct rte_mbuf *dpdk_zerocopy_replicate_packet_mb (vlib_buffer_t * b);
+
+#define foreach_dpdk_error						\
+  _(NONE, "no error")							\
+  _(RX_PACKET_ERROR, "Rx packet errors")				\
+  _(RX_BAD_FCS, "Rx bad fcs")						\
+  _(IP_CHECKSUM_ERROR, "Rx ip checksum errors")				\
+  _(RX_ALLOC_FAIL, "rx buf alloc from free list failed")		\
+  _(RX_ALLOC_NO_PHYSMEM, "rx buf alloc failed no physmem")		\
+  _(RX_ALLOC_DROP_PKTS, "rx packets dropped due to alloc error")
+
+typedef enum
+{
+#define _(f,s) DPDK_ERROR_##f,
+  foreach_dpdk_error
+#undef _
+    DPDK_N_ERROR,
+} dpdk_error_t;
+
+int dpdk_set_stat_poll_interval (f64 interval);
+int dpdk_set_link_state_poll_interval (f64 interval);
+void dpdk_update_link_state (dpdk_device_t * xd, f64 now);
+void dpdk_device_lock_init (dpdk_device_t * xd);
+void dpdk_device_lock_free (dpdk_device_t * xd);
+
+void dpdk_rx_trace (dpdk_main_t * dm,
+		    vlib_node_runtime_t * node,
+		    dpdk_device_t * xd,
+		    u16 queue_id, u32 * buffers, uword n_buffers);
+
+#define EFD_OPERATION_LESS_THAN          0
+#define EFD_OPERATION_GREATER_OR_EQUAL   1
+
+format_function_t format_dpdk_device_name;
+format_function_t format_dpdk_device;
+format_function_t format_dpdk_tx_dma_trace;
+format_function_t format_dpdk_rx_dma_trace;
+format_function_t format_dpdk_rte_mbuf;
+format_function_t format_dpdk_rx_rte_mbuf;
+unformat_function_t unformat_socket_mem;
+clib_error_t *unformat_rss_fn (unformat_input_t * input, uword * rss_fn);
+clib_error_t *unformat_hqos (unformat_input_t * input,
+			     dpdk_device_config_hqos_t * hqos);
+
+uword
+admin_up_down_process (vlib_main_t * vm,
+		       vlib_node_runtime_t * rt, vlib_frame_t * f);
+
+#endif /* __included_dpdk_h__ */
+
+/*
+ * fd.io coding-style-patch-verification: ON
+ *
+ * Local Variables:
+ * eval: (c-set-style "gnu")
+ * End:
+ */
diff --git a/src/plugins/dpdk/device/dpdk_priv.h b/src/plugins/dpdk/device/dpdk_priv.h
new file mode 100644
index 00000000..dd40ff48
--- /dev/null
+++ b/src/plugins/dpdk/device/dpdk_priv.h
@@ -0,0 +1,135 @@
+/*
+ * Copyright (c) 2015 Cisco and/or its affiliates.
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at:
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#define rte_mbuf_from_vlib_buffer(x) (((struct rte_mbuf *)x) - 1)
+#define vlib_buffer_from_rte_mbuf(x) ((vlib_buffer_t *)(x+1))
+
+#define DPDK_NB_RX_DESC_DEFAULT   1024
+#define DPDK_NB_TX_DESC_DEFAULT   1024
+#define DPDK_NB_RX_DESC_VIRTIO    256
+#define DPDK_NB_TX_DESC_VIRTIO    256
+
+#define I40E_DEV_ID_SFP_XL710           0x1572
+#define I40E_DEV_ID_QSFP_A              0x1583
+#define I40E_DEV_ID_QSFP_B              0x1584
+#define I40E_DEV_ID_QSFP_C              0x1585
+#define I40E_DEV_ID_10G_BASE_T          0x1586
+#define I40E_DEV_ID_VF                  0x154C
+
+/* These args appear by themselves */
+#define foreach_eal_double_hyphen_predicate_arg \
+_(no-shconf)                                    \
+_(no-hpet)                                      \
+_(no-huge)                                      \
+_(vmware-tsc-map)
+
+#define foreach_eal_single_hyphen_mandatory_arg \
+_(coremask, c)                                  \
+_(nchannels, n)                                 \
+
+#define foreach_eal_single_hyphen_arg           \
+_(blacklist, b)                                 \
+_(mem-alloc-request, m)                         \
+_(force-ranks, r)
+
+/* These args are preceeded by "--" and followed by a single string */
+#define foreach_eal_double_hyphen_arg           \
+_(huge-dir)                                     \
+_(proc-type)                                    \
+_(file-prefix)                                  \
+_(vdev)
+
+static inline void
+dpdk_get_xstats (dpdk_device_t * xd)
+{
+  int len;
+  if ((len = rte_eth_xstats_get (xd->device_index, NULL, 0)) > 0)
+    {
+      vec_validate (xd->xstats, len - 1);
+      vec_validate (xd->last_cleared_xstats, len - 1);
+
+      len =
+	rte_eth_xstats_get (xd->device_index, xd->xstats,
+			    vec_len (xd->xstats));
+
+      ASSERT (vec_len (xd->xstats) == len);
+      ASSERT (vec_len (xd->last_cleared_xstats) == len);
+
+      _vec_len (xd->xstats) = len;
+      _vec_len (xd->last_cleared_xstats) = len;
+
+    }
+}
+
+
+static inline void
+dpdk_update_counters (dpdk_device_t * xd, f64 now)
+{
+  vlib_simple_counter_main_t *cm;
+  vnet_main_t *vnm = vnet_get_main ();
+  u32 my_cpu = os_get_cpu_number ();
+  u64 rxerrors, last_rxerrors;
+
+  /* only update counters for PMD interfaces */
+  if ((xd->flags & DPDK_DEVICE_FLAG_PMD) == 0)
+    return;
+
+  xd->time_last_stats_update = now ? now : xd->time_last_stats_update;
+  clib_memcpy (&xd->last_stats, &xd->stats, sizeof (xd->last_stats));
+  rte_eth_stats_get (xd->device_index, &xd->stats);
+
+  /* maybe bump interface rx no buffer counter */
+  if (PREDICT_FALSE (xd->stats.rx_nombuf != xd->last_stats.rx_nombuf))
+    {
+      cm = vec_elt_at_index (vnm->interface_main.sw_if_counters,
+			     VNET_INTERFACE_COUNTER_RX_NO_BUF);
+
+      vlib_increment_simple_counter (cm, my_cpu, xd->vlib_sw_if_index,
+				     xd->stats.rx_nombuf -
+				     xd->last_stats.rx_nombuf);
+    }
+
+  /* missed pkt counter */
+  if (PREDICT_FALSE (xd->stats.imissed != xd->last_stats.imissed))
+    {
+      cm = vec_elt_at_index (vnm->interface_main.sw_if_counters,
+			     VNET_INTERFACE_COUNTER_RX_MISS);
+
+      vlib_increment_simple_counter (cm, my_cpu, xd->vlib_sw_if_index,
+				     xd->stats.imissed -
+				     xd->last_stats.imissed);
+    }
+  rxerrors = xd->stats.ierrors;
+  last_rxerrors = xd->last_stats.ierrors;
+
+  if (PREDICT_FALSE (rxerrors != last_rxerrors))
+    {
+      cm = vec_elt_at_index (vnm->interface_main.sw_if_counters,
+			     VNET_INTERFACE_COUNTER_RX_ERROR);
+
+      vlib_increment_simple_counter (cm, my_cpu, xd->vlib_sw_if_index,
+				     rxerrors - last_rxerrors);
+    }
+
+  dpdk_get_xstats (xd);
+}
+
+/*
+ * fd.io coding-style-patch-verification: ON
+ *
+ * Local Variables:
+ * eval: (c-set-style "gnu")
+ * End:
+ */
diff --git a/src/plugins/dpdk/device/format.c b/src/plugins/dpdk/device/format.c
new file mode 100644
index 00000000..25a8c5cb
--- /dev/null
+++ b/src/plugins/dpdk/device/format.c
@@ -0,0 +1,754 @@
+/*
+ * Copyright (c) 2015 Cisco and/or its affiliates.
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at:
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+#include 
+#include 
+#include 
+#include 
+#include 
+
+#include 
+#include 
+
+#include 
+#include 
+
+#define foreach_dpdk_counter                    \
+  _ (tx_frames_ok, opackets)                    \
+  _ (tx_bytes_ok, obytes)                       \
+  _ (tx_errors, oerrors)                        \
+  _ (rx_frames_ok, ipackets)                    \
+  _ (rx_bytes_ok, ibytes)                       \
+  _ (rx_errors, ierrors)                        \
+  _ (rx_missed, imissed)                        \
+  _ (rx_no_bufs, rx_nombuf)
+
+#define foreach_dpdk_q_counter                  \
+  _ (rx_frames_ok, q_ipackets)                  \
+  _ (tx_frames_ok, q_opackets)                  \
+  _ (rx_bytes_ok, q_ibytes)                     \
+  _ (tx_bytes_ok, q_obytes)                     \
+  _ (rx_errors, q_errors)
+
+#define foreach_dpdk_rss_hf                    \
+  _(ETH_RSS_FRAG_IPV4,          "ipv4-frag")   \
+  _(ETH_RSS_NONFRAG_IPV4_TCP,   "ipv4-tcp")    \
+  _(ETH_RSS_NONFRAG_IPV4_UDP,   "ipv4-udp")    \
+  _(ETH_RSS_NONFRAG_IPV4_SCTP,  "ipv4-sctp")   \
+  _(ETH_RSS_NONFRAG_IPV4_OTHER, "ipv4-other")  \
+  _(ETH_RSS_IPV4,               "ipv4")        \
+  _(ETH_RSS_IPV6_TCP_EX,        "ipv6-tcp-ex") \
+  _(ETH_RSS_IPV6_UDP_EX,        "ipv6-udp-ex") \
+  _(ETH_RSS_FRAG_IPV6,          "ipv6-frag")   \
+  _(ETH_RSS_NONFRAG_IPV6_TCP,   "ipv6-tcp")    \
+  _(ETH_RSS_NONFRAG_IPV6_UDP,   "ipv6-udp")    \
+  _(ETH_RSS_NONFRAG_IPV6_SCTP,  "ipv6-sctp")   \
+  _(ETH_RSS_NONFRAG_IPV6_OTHER, "ipv6-other")  \
+  _(ETH_RSS_L2_PAYLOAD,         "l2-payload")  \
+  _(ETH_RSS_IPV6_EX,            "ipv6-ex")     \
+  _(ETH_RSS_IPV6,               "ipv6")
+
+
+#define foreach_dpdk_rx_offload_caps            \
+  _(DEV_RX_OFFLOAD_VLAN_STRIP, "vlan-strip")    \
+  _(DEV_RX_OFFLOAD_IPV4_CKSUM, "ipv4-cksum")    \
+  _(DEV_RX_OFFLOAD_UDP_CKSUM , "udp-cksum")     \
+  _(DEV_RX_OFFLOAD_TCP_CKSUM , "tcp-cksum")     \
+  _(DEV_RX_OFFLOAD_TCP_LRO   , "rcp-lro")       \
+  _(DEV_RX_OFFLOAD_QINQ_STRIP, "qinq-strip")
+
+#define foreach_dpdk_tx_offload_caps           \
+  _(DEV_TX_OFFLOAD_VLAN_INSERT, "vlan-insert") \
+  _(DEV_TX_OFFLOAD_IPV4_CKSUM,  "ipv4-cksum")  \
+  _(DEV_TX_OFFLOAD_UDP_CKSUM  , "udp-cksum")   \
+  _(DEV_TX_OFFLOAD_TCP_CKSUM  , "tcp-cksum")   \
+  _(DEV_TX_OFFLOAD_SCTP_CKSUM , "sctp-cksum")  \
+  _(DEV_TX_OFFLOAD_TCP_TSO    , "tcp-tso")     \
+  _(DEV_TX_OFFLOAD_UDP_TSO    , "udp-tso")     \
+  _(DEV_TX_OFFLOAD_OUTER_IPV4_CKSUM, "outer-ipv4-cksum") \
+  _(DEV_TX_OFFLOAD_QINQ_INSERT, "qinq-insert")
+
+#define foreach_dpdk_pkt_rx_offload_flag                                \
+  _ (PKT_RX_VLAN_PKT, "RX packet is a 802.1q VLAN packet")              \
+  _ (PKT_RX_RSS_HASH, "RX packet with RSS hash result")                 \
+  _ (PKT_RX_FDIR, "RX packet with FDIR infos")                          \
+  _ (PKT_RX_L4_CKSUM_BAD, "L4 cksum of RX pkt. is not OK")              \
+  _ (PKT_RX_IP_CKSUM_BAD, "IP cksum of RX pkt. is not OK")              \
+  _ (PKT_RX_VLAN_STRIPPED, "RX packet VLAN tag stripped")               \
+  _ (PKT_RX_IP_CKSUM_GOOD, "IP cksum of RX pkt. is valid")              \
+  _ (PKT_RX_L4_CKSUM_GOOD, "L4 cksum of RX pkt. is valid")              \
+  _ (PKT_RX_IEEE1588_PTP, "RX IEEE1588 L2 Ethernet PT Packet")          \
+  _ (PKT_RX_IEEE1588_TMST, "RX IEEE1588 L2/L4 timestamped packet")      \
+  _ (PKT_RX_QINQ_STRIPPED, "RX packet QinQ tags stripped")
+
+#define foreach_dpdk_pkt_type                                           \
+  _ (L2, ETHER, "Ethernet packet")                                      \
+  _ (L2, ETHER_TIMESYNC, "Ethernet packet for time sync")               \
+  _ (L2, ETHER_ARP, "ARP packet")                                       \
+  _ (L2, ETHER_LLDP, "LLDP (Link Layer Discovery Protocol) packet")     \
+  _ (L2, ETHER_NSH, "NSH (Network Service Header) packet")              \
+  _ (L2, ETHER_VLAN, "VLAN packet")                                     \
+  _ (L2, ETHER_QINQ, "QinQ packet")                                     \
+  _ (L3, IPV4, "IPv4 packet without extension headers")                 \
+  _ (L3, IPV4_EXT, "IPv4 packet with extension headers")                \
+  _ (L3, IPV4_EXT_UNKNOWN, "IPv4 packet with or without extension headers") \
+  _ (L3, IPV6, "IPv6 packet without extension headers")                 \
+  _ (L3, IPV6_EXT, "IPv6 packet with extension headers")                \
+  _ (L3, IPV6_EXT_UNKNOWN, "IPv6 packet with or without extension headers") \
+  _ (L4, TCP, "TCP packet")                                             \
+  _ (L4, UDP, "UDP packet")                                             \
+  _ (L4, FRAG, "Fragmented IP packet")                                  \
+  _ (L4, SCTP, "SCTP (Stream Control Transmission Protocol) packet")    \
+  _ (L4, ICMP, "ICMP packet")                                           \
+  _ (L4, NONFRAG, "Non-fragmented IP packet")                           \
+  _ (TUNNEL, GRE, "GRE tunneling packet")                               \
+  _ (TUNNEL, VXLAN, "VXLAN tunneling packet")                           \
+  _ (TUNNEL, NVGRE, "NVGRE Tunneling packet")                           \
+  _ (TUNNEL, GENEVE, "GENEVE Tunneling packet")                         \
+  _ (TUNNEL, GRENAT, "Teredo, VXLAN or GRE Tunneling packet")           \
+  _ (INNER_L2, ETHER, "Inner Ethernet packet")                          \
+  _ (INNER_L2, ETHER_VLAN, "Inner Ethernet packet with VLAN")           \
+  _ (INNER_L3, IPV4, "Inner IPv4 packet without extension headers")     \
+  _ (INNER_L3, IPV4_EXT, "Inner IPv4 packet with extension headers")    \
+  _ (INNER_L3, IPV4_EXT_UNKNOWN, "Inner IPv4 packet with or without extension headers") \
+  _ (INNER_L3, IPV6, "Inner IPv6 packet without extension headers")     \
+  _ (INNER_L3, IPV6_EXT, "Inner IPv6 packet with extension headers")    \
+  _ (INNER_L3, IPV6_EXT_UNKNOWN, "Inner IPv6 packet with or without extension headers") \
+  _ (INNER_L4, TCP, "Inner TCP packet")                                 \
+  _ (INNER_L4, UDP, "Inner UDP packet")                                 \
+  _ (INNER_L4, FRAG, "Inner fagmented IP packet")                       \
+  _ (INNER_L4, SCTP, "Inner SCTP (Stream Control Transmission Protocol) packet") \
+  _ (INNER_L4, ICMP, "Inner ICMP packet")                               \
+  _ (INNER_L4, NONFRAG, "Inner non-fragmented IP packet")
+
+#define foreach_dpdk_pkt_tx_offload_flag                                \
+  _ (PKT_TX_VLAN_PKT, "TX packet is a 802.1q VLAN packet")              \
+  _ (PKT_TX_IP_CKSUM, "IP cksum of TX pkt. computed by NIC")            \
+  _ (PKT_TX_TCP_CKSUM, "TCP cksum of TX pkt. computed by NIC")          \
+  _ (PKT_TX_SCTP_CKSUM, "SCTP cksum of TX pkt. computed by NIC")        \
+  _ (PKT_TX_IEEE1588_TMST, "TX IEEE1588 packet to timestamp")
+
+#define foreach_dpdk_pkt_offload_flag           \
+  foreach_dpdk_pkt_rx_offload_flag              \
+  foreach_dpdk_pkt_tx_offload_flag
+
+u8 *
+format_dpdk_device_name (u8 * s, va_list * args)
+{
+  dpdk_main_t *dm = &dpdk_main;
+  char *devname_format;
+  char *device_name;
+  u32 i = va_arg (*args, u32);
+  struct rte_eth_dev_info dev_info;
+  u8 *ret;
+
+  if (dm->conf->interface_name_format_decimal)
+    devname_format = "%s%d/%d/%d";
+  else
+    devname_format = "%s%x/%x/%x";
+
+  switch (dm->devices[i].port_type)
+    {
+    case VNET_DPDK_PORT_TYPE_ETH_1G:
+      device_name = "GigabitEthernet";
+      break;
+
+    case VNET_DPDK_PORT_TYPE_ETH_10G:
+      device_name = "TenGigabitEthernet";
+      break;
+
+    case VNET_DPDK_PORT_TYPE_ETH_40G:
+      device_name = "FortyGigabitEthernet";
+      break;
+
+    case VNET_DPDK_PORT_TYPE_ETH_100G:
+      device_name = "HundredGigabitEthernet";
+      break;
+
+    case VNET_DPDK_PORT_TYPE_ETH_BOND:
+      return format (s, "BondEthernet%d", dm->devices[i].device_index);
+
+    case VNET_DPDK_PORT_TYPE_ETH_SWITCH:
+      device_name = "EthernetSwitch";
+      break;
+
+    case VNET_DPDK_PORT_TYPE_AF_PACKET:
+      rte_eth_dev_info_get (i, &dev_info);
+      return format (s, "af_packet%d", dm->devices[i].af_packet_port_id);
+
+    default:
+    case VNET_DPDK_PORT_TYPE_UNKNOWN:
+      device_name = "UnknownEthernet";
+      break;
+    }
+
+  rte_eth_dev_info_get (i, &dev_info);
+
+  if (dev_info.pci_dev)
+    ret = format (s, devname_format, device_name, dev_info.pci_dev->addr.bus,
+		  dev_info.pci_dev->addr.devid,
+		  dev_info.pci_dev->addr.function);
+  else
+    ret = format (s, "%s%d", device_name, dm->devices[i].device_index);
+
+  if (dm->devices[i].interface_name_suffix)
+    return format (ret, "/%s", dm->devices[i].interface_name_suffix);
+  return ret;
+}
+
+static u8 *
+format_dpdk_device_type (u8 * s, va_list * args)
+{
+  dpdk_main_t *dm = &dpdk_main;
+  char *dev_type;
+  u32 i = va_arg (*args, u32);
+
+  switch (dm->devices[i].pmd)
+    {
+    case VNET_DPDK_PMD_E1000EM:
+      dev_type = "Intel 82540EM (e1000)";
+      break;
+
+    case VNET_DPDK_PMD_IGB:
+      dev_type = "Intel e1000";
+      break;
+
+    case VNET_DPDK_PMD_I40E:
+      dev_type = "Intel X710/XL710 Family";
+      break;
+
+    case VNET_DPDK_PMD_I40EVF:
+      dev_type = "Intel X710/XL710 Family VF";
+      break;
+
+    case VNET_DPDK_PMD_FM10K:
+      dev_type = "Intel FM10000 Family Ethernet Switch";
+      break;
+
+    case VNET_DPDK_PMD_IGBVF:
+      dev_type = "Intel e1000 VF";
+      break;
+
+    case VNET_DPDK_PMD_VIRTIO:
+      dev_type = "Red Hat Virtio";
+      break;
+
+    case VNET_DPDK_PMD_IXGBEVF:
+      dev_type = "Intel 82599 VF";
+      break;
+
+    case VNET_DPDK_PMD_IXGBE:
+      dev_type = "Intel 82599";
+      break;
+
+    case VNET_DPDK_PMD_ENIC:
+      dev_type = "Cisco VIC";
+      break;
+
+    case VNET_DPDK_PMD_CXGBE:
+      dev_type = "Chelsio T4/T5";
+      break;
+
+    case VNET_DPDK_PMD_MLX5:
+      dev_type = "Mellanox ConnectX-4 Family";
+      break;
+
+    case VNET_DPDK_PMD_VMXNET3:
+      dev_type = "VMware VMXNET3";
+      break;
+
+    case VNET_DPDK_PMD_AF_PACKET:
+      dev_type = "af_packet";
+      break;
+
+    case VNET_DPDK_PMD_BOND:
+      dev_type = "Ethernet Bonding";
+      break;
+
+    case VNET_DPDK_PMD_DPAA2:
+      dev_type = "NXP DPAA2 Mac";
+      break;
+
+    default:
+    case VNET_DPDK_PMD_UNKNOWN:
+      dev_type = "### UNKNOWN ###";
+      break;
+    }
+
+  return format (s, dev_type);
+}
+
+static u8 *
+format_dpdk_link_status (u8 * s, va_list * args)
+{
+  dpdk_device_t *xd = va_arg (*args, dpdk_device_t *);
+  struct rte_eth_link *l = &xd->link;
+  vnet_main_t *vnm = vnet_get_main ();
+  vnet_hw_interface_t *hi = vnet_get_hw_interface (vnm, xd->vlib_hw_if_index);
+
+  s = format (s, "%s ", l->link_status ? "up" : "down");
+  if (l->link_status)
+    {
+      u32 promisc = rte_eth_promiscuous_get (xd->device_index);
+
+      s = format (s, "%s duplex ", (l->link_duplex == ETH_LINK_FULL_DUPLEX) ?
+		  "full" : "half");
+      s = format (s, "speed %u mtu %d %s\n", l->link_speed,
+		  hi->max_packet_bytes, promisc ? " promisc" : "");
+    }
+  else
+    s = format (s, "\n");
+
+  return s;
+}
+
+#define _line_len 72
+#define _(v, str)                                            \
+if (bitmap & v) {                                            \
+  if (format_get_indent (s) > next_split ) {                 \
+    next_split += _line_len;                                 \
+    s = format(s,"\n%U", format_white_space, indent);        \
+  }                                                          \
+  s = format(s, "%s ", str);                                 \
+}
+
+static u8 *
+format_dpdk_rss_hf_name (u8 * s, va_list * args)
+{
+  u64 bitmap = va_arg (*args, u64);
+  int next_split = _line_len;
+  int indent = format_get_indent (s);
+
+  if (!bitmap)
+    return format (s, "none");
+
+  foreach_dpdk_rss_hf return s;
+}
+
+static u8 *
+format_dpdk_rx_offload_caps (u8 * s, va_list * args)
+{
+  u32 bitmap = va_arg (*args, u32);
+  int next_split = _line_len;
+  int indent = format_get_indent (s);
+
+  if (!bitmap)
+    return format (s, "none");
+
+  foreach_dpdk_rx_offload_caps return s;
+}
+
+static u8 *
+format_dpdk_tx_offload_caps (u8 * s, va_list * args)
+{
+  u32 bitmap = va_arg (*args, u32);
+  int next_split = _line_len;
+  int indent = format_get_indent (s);
+  if (!bitmap)
+    return format (s, "none");
+
+  foreach_dpdk_tx_offload_caps return s;
+}
+
+#undef _line_len
+#undef _
+
+u8 *
+format_dpdk_device (u8 * s, va_list * args)
+{
+  u32 dev_instance = va_arg (*args, u32);
+  int verbose = va_arg (*args, int);
+  dpdk_main_t *dm = &dpdk_main;
+  dpdk_device_t *xd = vec_elt_at_index (dm->devices, dev_instance);
+  uword indent = format_get_indent (s);
+  f64 now = vlib_time_now (dm->vlib_main);
+  struct rte_eth_dev_info di;
+
+  dpdk_update_counters (xd, now);
+  dpdk_update_link_state (xd, now);
+
+  s = format (s, "%U\n%Ucarrier %U",
+	      format_dpdk_device_type, xd->device_index,
+	      format_white_space, indent + 2, format_dpdk_link_status, xd);
+
+  rte_eth_dev_info_get (xd->device_index, &di);
+
+  if (verbose > 1 && xd->flags & DPDK_DEVICE_FLAG_PMD)
+    {
+      struct rte_pci_device *pci;
+      struct rte_eth_rss_conf rss_conf;
+      int vlan_off;
+      int retval;
+
+      rss_conf.rss_key = 0;
+      retval = rte_eth_dev_rss_hash_conf_get (xd->device_index, &rss_conf);
+      if (retval < 0)
+	clib_warning ("rte_eth_dev_rss_hash_conf_get returned %d", retval);
+      pci = di.pci_dev;
+
+      if (pci)
+	s =
+	  format (s,
+		  "%Upci id:            device %04x:%04x subsystem %04x:%04x\n"
+		  "%Upci address:       %04x:%02x:%02x.%02x\n",
+		  format_white_space, indent + 2, pci->id.vendor_id,
+		  pci->id.device_id, pci->id.subsystem_vendor_id,
+		  pci->id.subsystem_device_id, format_white_space, indent + 2,
+		  pci->addr.domain, pci->addr.bus, pci->addr.devid,
+		  pci->addr.function);
+      s =
+	format (s, "%Umax rx packet len: %d\n", format_white_space,
+		indent + 2, di.max_rx_pktlen);
+      s =
+	format (s, "%Umax num of queues: rx %d tx %d\n", format_white_space,
+		indent + 2, di.max_rx_queues, di.max_tx_queues);
+      s =
+	format (s, "%Upromiscuous:       unicast %s all-multicast %s\n",
+		format_white_space, indent + 2,
+		rte_eth_promiscuous_get (xd->device_index) ? "on" : "off",
+		rte_eth_promiscuous_get (xd->device_index) ? "on" : "off");
+      vlan_off = rte_eth_dev_get_vlan_offload (xd->device_index);
+      s = format (s, "%Uvlan offload:      strip %s filter %s qinq %s\n",
+		  format_white_space, indent + 2,
+		  vlan_off & ETH_VLAN_STRIP_OFFLOAD ? "on" : "off",
+		  vlan_off & ETH_VLAN_FILTER_OFFLOAD ? "on" : "off",
+		  vlan_off & ETH_VLAN_EXTEND_OFFLOAD ? "on" : "off");
+      s = format (s, "%Urx offload caps:   %U\n",
+		  format_white_space, indent + 2,
+		  format_dpdk_rx_offload_caps, di.rx_offload_capa);
+      s = format (s, "%Utx offload caps:   %U\n",
+		  format_white_space, indent + 2,
+		  format_dpdk_tx_offload_caps, di.tx_offload_capa);
+      s = format (s, "%Urss active:        %U\n"
+		  "%Urss supported:     %U\n",
+		  format_white_space, indent + 2,
+		  format_dpdk_rss_hf_name, rss_conf.rss_hf,
+		  format_white_space, indent + 2,
+		  format_dpdk_rss_hf_name, di.flow_type_rss_offloads);
+    }
+
+  s = format (s, "%Urx queues %d, rx desc %d, tx queues %d, tx desc %d\n",
+	      format_white_space, indent + 2,
+	      xd->rx_q_used, xd->nb_rx_desc, xd->tx_q_used, xd->nb_tx_desc);
+
+  if (xd->cpu_socket > -1)
+    s = format (s, "%Ucpu socket %d\n",
+		format_white_space, indent + 2, xd->cpu_socket);
+
+  /* $$$ MIB counters  */
+  {
+#define _(N, V)							\
+    if ((xd->stats.V - xd->last_cleared_stats.V) != 0) {       \
+      s = format (s, "\n%U%-40U%16Ld",                         \
+                  format_white_space, indent + 2,              \
+                  format_c_identifier, #N,                     \
+                  xd->stats.V - xd->last_cleared_stats.V);     \
+    }                                                          \
+
+    foreach_dpdk_counter
+#undef _
+  }
+
+  u8 *xs = 0;
+  u32 i = 0;
+  struct rte_eth_xstat *xstat, *last_xstat;
+  struct rte_eth_xstat_name *xstat_names = 0;
+  int len = rte_eth_xstats_get_names (xd->device_index, NULL, 0);
+  vec_validate (xstat_names, len - 1);
+  rte_eth_xstats_get_names (xd->device_index, xstat_names, len);
+
+  ASSERT (vec_len (xd->xstats) == vec_len (xd->last_cleared_xstats));
+
+  /* *INDENT-OFF* */
+  vec_foreach_index(i, xd->xstats)
+    {
+      u64 delta = 0;
+      xstat = vec_elt_at_index(xd->xstats, i);
+      last_xstat = vec_elt_at_index(xd->last_cleared_xstats, i);
+
+      delta = xstat->value - last_xstat->value;
+      if (verbose == 2 || (verbose && delta))
+        {
+          /* format_c_identifier doesn't like c strings inside vector */
+          u8 * name = format(0,"%s", xstat_names[i].name);
+          xs = format(xs, "\n%U%-38U%16Ld",
+                      format_white_space, indent + 4,
+                      format_c_identifier, name, delta);
+          vec_free(name);
+        }
+    }
+  /* *INDENT-ON* */
+
+  vec_free (xstat_names);
+
+  if (xs)
+    {
+      s = format (s, "\n%Uextended stats:%v",
+		  format_white_space, indent + 2, xs);
+      vec_free (xs);
+    }
+
+  return s;
+}
+
+u8 *
+format_dpdk_tx_dma_trace (u8 * s, va_list * va)
+{
+  CLIB_UNUSED (vlib_main_t * vm) = va_arg (*va, vlib_main_t *);
+  CLIB_UNUSED (vlib_node_t * node) = va_arg (*va, vlib_node_t *);
+  CLIB_UNUSED (vnet_main_t * vnm) = vnet_get_main ();
+  dpdk_tx_dma_trace_t *t = va_arg (*va, dpdk_tx_dma_trace_t *);
+  dpdk_main_t *dm = &dpdk_main;
+  dpdk_device_t *xd = vec_elt_at_index (dm->devices, t->device_index);
+  uword indent = format_get_indent (s);
+  vnet_sw_interface_t *sw = vnet_get_sw_interface (vnm, xd->vlib_sw_if_index);
+
+  s = format (s, "%U tx queue %d",
+	      format_vnet_sw_interface_name, vnm, sw, t->queue_index);
+
+  s = format (s, "\n%Ubuffer 0x%x: %U",
+	      format_white_space, indent,
+	      t->buffer_index, format_vlib_buffer, &t->buffer);
+
+  s = format (s, "\n%U%U", format_white_space, indent,
+	      format_ethernet_header_with_length, t->buffer.pre_data,
+	      sizeof (t->buffer.pre_data));
+
+  return s;
+}
+
+u8 *
+format_dpdk_rx_dma_trace (u8 * s, va_list * va)
+{
+  CLIB_UNUSED (vlib_main_t * vm) = va_arg (*va, vlib_main_t *);
+  CLIB_UNUSED (vlib_node_t * node) = va_arg (*va, vlib_node_t *);
+  CLIB_UNUSED (vnet_main_t * vnm) = vnet_get_main ();
+  dpdk_rx_dma_trace_t *t = va_arg (*va, dpdk_rx_dma_trace_t *);
+  dpdk_main_t *dm = &dpdk_main;
+  dpdk_device_t *xd = vec_elt_at_index (dm->devices, t->device_index);
+  format_function_t *f;
+  uword indent = format_get_indent (s);
+  vnet_sw_interface_t *sw = vnet_get_sw_interface (vnm, xd->vlib_sw_if_index);
+
+  s = format (s, "%U rx queue %d",
+	      format_vnet_sw_interface_name, vnm, sw, t->queue_index);
+
+  s = format (s, "\n%Ubuffer 0x%x: %U",
+	      format_white_space, indent,
+	      t->buffer_index, format_vlib_buffer, &t->buffer);
+
+  s = format (s, "\n%U%U",
+	      format_white_space, indent,
+	      format_dpdk_rte_mbuf, &t->mb, &t->data);
+
+  if (vm->trace_main.verbose)
+    {
+      s = format (s, "\n%UPacket Dump%s", format_white_space, indent + 2,
+		  t->mb.data_len > sizeof (t->data) ? " (truncated)" : "");
+      s = format (s, "\n%U%U", format_white_space, indent + 4,
+		  format_hexdump, &t->data,
+		  t->mb.data_len >
+		  sizeof (t->data) ? sizeof (t->data) : t->mb.data_len);
+    }
+  f = node->format_buffer;
+  if (!f)
+    f = format_hex_bytes;
+  s = format (s, "\n%U%U", format_white_space, indent,
+	      f, t->buffer.pre_data, sizeof (t->buffer.pre_data));
+
+  return s;
+}
+
+
+static inline u8 *
+format_dpdk_pkt_types (u8 * s, va_list * va)
+{
+  u32 *pkt_types = va_arg (*va, u32 *);
+  uword indent __attribute__ ((unused)) = format_get_indent (s) + 2;
+
+  if (!*pkt_types)
+    return s;
+
+  s = format (s, "Packet Types");
+
+#define _(L, F, S)             \
+  if ((*pkt_types & RTE_PTYPE_##L##_MASK) == RTE_PTYPE_##L##_##F)           \
+    {                                                                       \
+      s = format (s, "\n%U%s (0x%04x) %s", format_white_space, indent,      \
+                     "RTE_PTYPE_" #L "_" #F, RTE_PTYPE_##L##_##F, S);       \
+    }
+
+  foreach_dpdk_pkt_type
+#undef _
+    return s;
+}
+
+static inline u8 *
+format_dpdk_pkt_offload_flags (u8 * s, va_list * va)
+{
+  u64 *ol_flags = va_arg (*va, u64 *);
+  uword indent = format_get_indent (s) + 2;
+
+  if (!*ol_flags)
+    return s;
+
+  s = format (s, "Packet Offload Flags");
+
+#define _(F, S)             \
+  if (*ol_flags & F)            \
+    {               \
+      s = format (s, "\n%U%s (0x%04x) %s",      \
+      format_white_space, indent, #F, F, S);  \
+    }
+
+  foreach_dpdk_pkt_offload_flag
+#undef _
+    return s;
+}
+
+u8 *
+format_dpdk_rte_mbuf_vlan (u8 * s, va_list * va)
+{
+  ethernet_vlan_header_tv_t *vlan_hdr =
+    va_arg (*va, ethernet_vlan_header_tv_t *);
+
+  if (clib_net_to_host_u16 (vlan_hdr->type) == ETHERNET_TYPE_DOT1AD)
+    {
+      s = format (s, "%U 802.1q vlan ",
+		  format_ethernet_vlan_tci,
+		  clib_net_to_host_u16 (vlan_hdr->priority_cfi_and_id));
+      vlan_hdr++;
+    }
+
+  s = format (s, "%U",
+	      format_ethernet_vlan_tci,
+	      clib_net_to_host_u16 (vlan_hdr->priority_cfi_and_id));
+
+  return s;
+}
+
+u8 *
+format_dpdk_rte_mbuf (u8 * s, va_list * va)
+{
+  struct rte_mbuf *mb = va_arg (*va, struct rte_mbuf *);
+  ethernet_header_t *eth_hdr = va_arg (*va, ethernet_header_t *);
+  uword indent = format_get_indent (s) + 2;
+
+  s = format (s, "PKT MBUF: port %d, nb_segs %d, pkt_len %d"
+	      "\n%Ubuf_len %d, data_len %d, ol_flags 0x%x, data_off %d, phys_addr 0x%x"
+	      "\n%Upacket_type 0x%x",
+	      mb->port, mb->nb_segs, mb->pkt_len,
+	      format_white_space, indent,
+	      mb->buf_len, mb->data_len, mb->ol_flags, mb->data_off,
+	      mb->buf_physaddr, format_white_space, indent, mb->packet_type);
+
+  if (mb->ol_flags)
+    s = format (s, "\n%U%U", format_white_space, indent,
+		format_dpdk_pkt_offload_flags, &mb->ol_flags);
+
+  if ((mb->ol_flags & PKT_RX_VLAN_PKT) &&
+      ((mb->ol_flags & (PKT_RX_VLAN_STRIPPED | PKT_RX_QINQ_STRIPPED)) == 0))
+    {
+      ethernet_vlan_header_tv_t *vlan_hdr =
+	((ethernet_vlan_header_tv_t *) & (eth_hdr->type));
+      s = format (s, " %U", format_dpdk_rte_mbuf_vlan, vlan_hdr);
+    }
+
+  if (mb->packet_type)
+    s = format (s, "\n%U%U", format_white_space, indent,
+		format_dpdk_pkt_types, &mb->packet_type);
+
+  return s;
+}
+
+/* FIXME is this function used? */
+#if 0
+uword
+unformat_socket_mem (unformat_input_t * input, va_list * va)
+{
+  uword **r = va_arg (*va, uword **);
+  int i = 0;
+  u32 mem;
+
+  while (unformat_check_input (input) != UNFORMAT_END_OF_INPUT)
+    {
+      if (unformat (input, ","))
+	hash_set (*r, i, 1024);
+      else if (unformat (input, "%u,", &mem))
+	hash_set (*r, i, mem);
+      else if (unformat (input, "%u", &mem))
+	hash_set (*r, i, mem);
+      else
+	{
+	  unformat_put_input (input);
+	  goto done;
+	}
+      i++;
+    }
+
+done:
+  return 1;
+}
+#endif
+
+clib_error_t *
+unformat_rss_fn (unformat_input_t * input, uword * rss_fn)
+{
+  while (unformat_check_input (input) != UNFORMAT_END_OF_INPUT)
+    {
+      if (0)
+	;
+#undef _
+#define _(f, s)                                 \
+      else if (unformat (input, s))             \
+        *rss_fn |= f;
+
+      foreach_dpdk_rss_hf
+#undef _
+	else
+	{
+	  return clib_error_return (0, "unknown input `%U'",
+				    format_unformat_error, input);
+	}
+    }
+  return 0;
+}
+
+clib_error_t *
+unformat_hqos (unformat_input_t * input, dpdk_device_config_hqos_t * hqos)
+{
+  clib_error_t *error = 0;
+
+  while (unformat_check_input (input) != UNFORMAT_END_OF_INPUT)
+    {
+      if (unformat (input, "hqos-thread %u", &hqos->hqos_thread))
+	hqos->hqos_thread_valid = 1;
+      else
+	{
+	  error = clib_error_return (0, "unknown input `%U'",
+				     format_unformat_error, input);
+	  break;
+	}
+    }
+
+  return error;
+}
+
+/*
+ * fd.io coding-style-patch-verification: ON
+ *
+ * Local Variables:
+ * eval: (c-set-style "gnu")
+ * End:
+ */
diff --git a/src/plugins/dpdk/device/node.c b/src/plugins/dpdk/device/node.c
new file mode 100644
index 00000000..8824d789
--- /dev/null
+++ b/src/plugins/dpdk/device/node.c
@@ -0,0 +1,674 @@
+/*
+ * Copyright (c) 2015 Cisco and/or its affiliates.
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at:
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+#include 
+#include 
+#include 
+#include 
+#include 
+
+#include 
+#include 
+#include 
+#include 
+#include 
+#include 
+#include 
+
+#include 
+
+static char *dpdk_error_strings[] = {
+#define _(n,s) s,
+  foreach_dpdk_error
+#undef _
+};
+
+always_inline int
+vlib_buffer_is_ip4 (vlib_buffer_t * b)
+{
+  ethernet_header_t *h = (ethernet_header_t *) b->data;
+  return (h->type == clib_host_to_net_u16 (ETHERNET_TYPE_IP4));
+}
+
+always_inline int
+vlib_buffer_is_ip6 (vlib_buffer_t * b)
+{
+  ethernet_header_t *h = (ethernet_header_t *) b->data;
+  return (h->type == clib_host_to_net_u16 (ETHERNET_TYPE_IP6));
+}
+
+always_inline int
+vlib_buffer_is_mpls (vlib_buffer_t * b)
+{
+  ethernet_header_t *h = (ethernet_header_t *) b->data;
+  return (h->type == clib_host_to_net_u16 (ETHERNET_TYPE_MPLS_UNICAST));
+}
+
+always_inline u32
+dpdk_rx_next_from_etype (struct rte_mbuf * mb, vlib_buffer_t * b0)
+{
+  if (PREDICT_TRUE (vlib_buffer_is_ip4 (b0)))
+    if (PREDICT_TRUE ((mb->ol_flags & PKT_RX_IP_CKSUM_GOOD) != 0))
+      return VNET_DEVICE_INPUT_NEXT_IP4_NCS_INPUT;
+    else
+      return VNET_DEVICE_INPUT_NEXT_IP4_INPUT;
+  else if (PREDICT_TRUE (vlib_buffer_is_ip6 (b0)))
+    return VNET_DEVICE_INPUT_NEXT_IP6_INPUT;
+  else if (PREDICT_TRUE (vlib_buffer_is_mpls (b0)))
+    return VNET_DEVICE_INPUT_NEXT_MPLS_INPUT;
+  else
+    return VNET_DEVICE_INPUT_NEXT_ETHERNET_INPUT;
+}
+
+always_inline int
+dpdk_mbuf_is_vlan (struct rte_mbuf *mb)
+{
+  return (mb->packet_type & RTE_PTYPE_L2_ETHER_VLAN) ==
+    RTE_PTYPE_L2_ETHER_VLAN;
+}
+
+always_inline int
+dpdk_mbuf_is_ip4 (struct rte_mbuf *mb)
+{
+  return RTE_ETH_IS_IPV4_HDR (mb->packet_type) != 0;
+}
+
+always_inline int
+dpdk_mbuf_is_ip6 (struct rte_mbuf *mb)
+{
+  return RTE_ETH_IS_IPV6_HDR (mb->packet_type) != 0;
+}
+
+always_inline u32
+dpdk_rx_next_from_mb (struct rte_mbuf * mb, vlib_buffer_t * b0)
+{
+  if (PREDICT_FALSE (dpdk_mbuf_is_vlan (mb)))
+    return VNET_DEVICE_INPUT_NEXT_ETHERNET_INPUT;
+  else if (PREDICT_TRUE (dpdk_mbuf_is_ip4 (mb)))
+    return VNET_DEVICE_INPUT_NEXT_IP4_NCS_INPUT;
+  else if (PREDICT_TRUE (dpdk_mbuf_is_ip6 (mb)))
+    return VNET_DEVICE_INPUT_NEXT_IP6_INPUT;
+  else if (PREDICT_TRUE (vlib_buffer_is_mpls (b0)))
+    return VNET_DEVICE_INPUT_NEXT_MPLS_INPUT;
+  else
+    return dpdk_rx_next_from_etype (mb, b0);
+}
+
+always_inline void
+dpdk_rx_error_from_mb (struct rte_mbuf *mb, u32 * next, u8 * error)
+{
+  if (mb->ol_flags & PKT_RX_IP_CKSUM_BAD)
+    {
+      *error = DPDK_ERROR_IP_CHECKSUM_ERROR;
+      *next = VNET_DEVICE_INPUT_NEXT_DROP;
+    }
+  else
+    *error = DPDK_ERROR_NONE;
+}
+
+void
+dpdk_rx_trace (dpdk_main_t * dm,
+	       vlib_node_runtime_t * node,
+	       dpdk_device_t * xd,
+	       u16 queue_id, u32 * buffers, uword n_buffers)
+{
+  vlib_main_t *vm = vlib_get_main ();
+  u32 *b, n_left;
+  u32 next0;
+
+  n_left = n_buffers;
+  b = buffers;
+
+  while (n_left >= 1)
+    {
+      u32 bi0;
+      vlib_buffer_t *b0;
+      dpdk_rx_dma_trace_t *t0;
+      struct rte_mbuf *mb;
+      u8 error0;
+
+      bi0 = b[0];
+      n_left -= 1;
+
+      b0 = vlib_get_buffer (vm, bi0);
+      mb = rte_mbuf_from_vlib_buffer (b0);
+
+      if (PREDICT_FALSE (xd->per_interface_next_index != ~0))
+	next0 = xd->per_interface_next_index;
+      else if (PREDICT_TRUE
+	       ((xd->flags & DPDK_DEVICE_FLAG_PMD_SUPPORTS_PTYPE) != 0))
+	next0 = dpdk_rx_next_from_mb (mb, b0);
+      else
+	next0 = dpdk_rx_next_from_etype (mb, b0);
+
+      dpdk_rx_error_from_mb (mb, &next0, &error0);
+
+      vlib_trace_buffer (vm, node, next0, b0, /* follow_chain */ 0);
+      t0 = vlib_add_trace (vm, node, b0, sizeof (t0[0]));
+      t0->queue_index = queue_id;
+      t0->device_index = xd->device_index;
+      t0->buffer_index = bi0;
+
+      clib_memcpy (&t0->mb, mb, sizeof (t0->mb));
+      clib_memcpy (&t0->buffer, b0, sizeof (b0[0]) - sizeof (b0->pre_data));
+      clib_memcpy (t0->buffer.pre_data, b0->data,
+		   sizeof (t0->buffer.pre_data));
+      clib_memcpy (&t0->data, mb->buf_addr + mb->data_off, sizeof (t0->data));
+
+      b += 1;
+    }
+}
+
+static inline u32
+dpdk_rx_burst (dpdk_main_t * dm, dpdk_device_t * xd, u16 queue_id)
+{
+  u32 n_buffers;
+  u32 n_left;
+  u32 n_this_chunk;
+
+  n_left = VLIB_FRAME_SIZE;
+  n_buffers = 0;
+
+  if (PREDICT_TRUE (xd->flags & DPDK_DEVICE_FLAG_PMD))
+    {
+      while (n_left)
+	{
+	  n_this_chunk = rte_eth_rx_burst (xd->device_index, queue_id,
+					   xd->rx_vectors[queue_id] +
+					   n_buffers, n_left);
+	  n_buffers += n_this_chunk;
+	  n_left -= n_this_chunk;
+
+	  /* Empirically, DPDK r1.8 produces vectors w/ 32 or fewer elts */
+	  if (n_this_chunk < 32)
+	    break;
+	}
+    }
+  else
+    {
+      ASSERT (0);
+    }
+
+  return n_buffers;
+}
+
+
+static_always_inline void
+dpdk_process_subseq_segs (vlib_main_t * vm, vlib_buffer_t * b,
+			  struct rte_mbuf *mb, vlib_buffer_free_list_t * fl)
+{
+  u8 nb_seg = 1;
+  struct rte_mbuf *mb_seg = 0;
+  vlib_buffer_t *b_seg, *b_chain = 0;
+  mb_seg = mb->next;
+  b_chain = b;
+
+  while ((mb->nb_segs > 1) && (nb_seg < mb->nb_segs))
+    {
+      ASSERT (mb_seg != 0);
+
+      b_seg = vlib_buffer_from_rte_mbuf (mb_seg);
+      vlib_buffer_init_for_free_list (b_seg, fl);
+
+      ASSERT ((b_seg->flags & VLIB_BUFFER_NEXT_PRESENT) == 0);
+      ASSERT (b_seg->current_data == 0);
+
+      /*
+       * The driver (e.g. virtio) may not put the packet data at the start
+       * of the segment, so don't assume b_seg->current_data == 0 is correct.
+       */
+      b_seg->current_data =
+	(mb_seg->buf_addr + mb_seg->data_off) - (void *) b_seg->data;
+
+      b_seg->current_length = mb_seg->data_len;
+      b->total_length_not_including_first_buffer += mb_seg->data_len;
+
+      b_chain->flags |= VLIB_BUFFER_NEXT_PRESENT;
+      b_chain->next_buffer = vlib_get_buffer_index (vm, b_seg);
+
+      b_chain = b_seg;
+      mb_seg = mb_seg->next;
+      nb_seg++;
+    }
+}
+
+static_always_inline void
+dpdk_prefetch_buffer (struct rte_mbuf *mb)
+{
+  vlib_buffer_t *b = vlib_buffer_from_rte_mbuf (mb);
+  CLIB_PREFETCH (mb, CLIB_CACHE_LINE_BYTES, LOAD);
+  CLIB_PREFETCH (b, CLIB_CACHE_LINE_BYTES, STORE);
+}
+
+/*
+ * This function is used when there are no worker threads.
+ * The main thread performs IO and forwards the packets.
+ */
+static_always_inline u32
+dpdk_device_input (dpdk_main_t * dm, dpdk_device_t * xd,
+		   vlib_node_runtime_t * node, u32 cpu_index, u16 queue_id)
+{
+  u32 n_buffers;
+  u32 next_index = VNET_DEVICE_INPUT_NEXT_ETHERNET_INPUT;
+  u32 n_left_to_next, *to_next;
+  u32 mb_index;
+  vlib_main_t *vm = vlib_get_main ();
+  uword n_rx_bytes = 0;
+  u32 n_trace, trace_cnt __attribute__ ((unused));
+  vlib_buffer_free_list_t *fl;
+  u32 buffer_flags_template;
+
+  if ((xd->flags & DPDK_DEVICE_FLAG_ADMIN_UP) == 0)
+    return 0;
+
+  n_buffers = dpdk_rx_burst (dm, xd, queue_id);
+
+  if (n_buffers == 0)
+    {
+      return 0;
+    }
+
+  buffer_flags_template = dm->buffer_flags_template;
+
+  vec_reset_length (xd->d_trace_buffers[cpu_index]);
+  trace_cnt = n_trace = vlib_get_trace_count (vm, node);
+
+  if (n_trace > 0)
+    {
+      u32 n = clib_min (n_trace, n_buffers);
+      mb_index = 0;
+
+      while (n--)
+	{
+	  struct rte_mbuf *mb = xd->rx_vectors[queue_id][mb_index++];
+	  vlib_buffer_t *b = vlib_buffer_from_rte_mbuf (mb);
+	  vec_add1 (xd->d_trace_buffers[cpu_index],
+		    vlib_get_buffer_index (vm, b));
+	}
+    }
+
+  fl = vlib_buffer_get_free_list (vm, VLIB_BUFFER_DEFAULT_FREE_LIST_INDEX);
+
+  mb_index = 0;
+
+  while (n_buffers > 0)
+    {
+      vlib_buffer_t *b0, *b1, *b2, *b3;
+      u32 bi0, next0, l3_offset0;
+      u32 bi1, next1, l3_offset1;
+      u32 bi2, next2, l3_offset2;
+      u32 bi3, next3, l3_offset3;
+      u8 error0, error1, error2, error3;
+      u64 or_ol_flags;
+
+      vlib_get_next_frame (vm, node, next_index, to_next, n_left_to_next);
+
+      while (n_buffers > 8 && n_left_to_next > 4)
+	{
+	  struct rte_mbuf *mb0 = xd->rx_vectors[queue_id][mb_index];
+	  struct rte_mbuf *mb1 = xd->rx_vectors[queue_id][mb_index + 1];
+	  struct rte_mbuf *mb2 = xd->rx_vectors[queue_id][mb_index + 2];
+	  struct rte_mbuf *mb3 = xd->rx_vectors[queue_id][mb_index + 3];
+
+	  dpdk_prefetch_buffer (xd->rx_vectors[queue_id][mb_index + 4]);
+	  dpdk_prefetch_buffer (xd->rx_vectors[queue_id][mb_index + 5]);
+	  dpdk_prefetch_buffer (xd->rx_vectors[queue_id][mb_index + 6]);
+	  dpdk_prefetch_buffer (xd->rx_vectors[queue_id][mb_index + 7]);
+
+	  if (xd->flags & DPDK_DEVICE_FLAG_MAYBE_MULTISEG)
+	    {
+	      if (PREDICT_FALSE (mb0->nb_segs > 1))
+		dpdk_prefetch_buffer (mb0->next);
+	      if (PREDICT_FALSE (mb1->nb_segs > 1))
+		dpdk_prefetch_buffer (mb1->next);
+	      if (PREDICT_FALSE (mb2->nb_segs > 1))
+		dpdk_prefetch_buffer (mb2->next);
+	      if (PREDICT_FALSE (mb3->nb_segs > 1))
+		dpdk_prefetch_buffer (mb3->next);
+	    }
+
+	  ASSERT (mb0);
+	  ASSERT (mb1);
+	  ASSERT (mb2);
+	  ASSERT (mb3);
+
+	  or_ol_flags = (mb0->ol_flags | mb1->ol_flags |
+			 mb2->ol_flags | mb3->ol_flags);
+	  b0 = vlib_buffer_from_rte_mbuf (mb0);
+	  b1 = vlib_buffer_from_rte_mbuf (mb1);
+	  b2 = vlib_buffer_from_rte_mbuf (mb2);
+	  b3 = vlib_buffer_from_rte_mbuf (mb3);
+
+	  vlib_buffer_init_for_free_list (b0, fl);
+	  vlib_buffer_init_for_free_list (b1, fl);
+	  vlib_buffer_init_for_free_list (b2, fl);
+	  vlib_buffer_init_for_free_list (b3, fl);
+
+	  bi0 = vlib_get_buffer_index (vm, b0);
+	  bi1 = vlib_get_buffer_index (vm, b1);
+	  bi2 = vlib_get_buffer_index (vm, b2);
+	  bi3 = vlib_get_buffer_index (vm, b3);
+
+	  to_next[0] = bi0;
+	  to_next[1] = bi1;
+	  to_next[2] = bi2;
+	  to_next[3] = bi3;
+	  to_next += 4;
+	  n_left_to_next -= 4;
+
+	  if (PREDICT_FALSE (xd->per_interface_next_index != ~0))
+	    {
+	      next0 = next1 = next2 = next3 = xd->per_interface_next_index;
+	    }
+	  else if (PREDICT_TRUE
+		   ((xd->flags & DPDK_DEVICE_FLAG_PMD_SUPPORTS_PTYPE) != 0))
+	    {
+	      next0 = dpdk_rx_next_from_mb (mb0, b0);
+	      next1 = dpdk_rx_next_from_mb (mb1, b1);
+	      next2 = dpdk_rx_next_from_mb (mb2, b2);
+	      next3 = dpdk_rx_next_from_mb (mb3, b3);
+	    }
+	  else
+	    {
+	      next0 = dpdk_rx_next_from_etype (mb0, b0);
+	      next1 = dpdk_rx_next_from_etype (mb1, b1);
+	      next2 = dpdk_rx_next_from_etype (mb2, b2);
+	      next3 = dpdk_rx_next_from_etype (mb3, b3);
+	    }
+
+	  if (PREDICT_FALSE (or_ol_flags & PKT_RX_IP_CKSUM_BAD))
+	    {
+	      dpdk_rx_error_from_mb (mb0, &next0, &error0);
+	      dpdk_rx_error_from_mb (mb1, &next1, &error1);
+	      dpdk_rx_error_from_mb (mb2, &next2, &error2);
+	      dpdk_rx_error_from_mb (mb3, &next3, &error3);
+	      b0->error = node->errors[error0];
+	      b1->error = node->errors[error1];
+	      b2->error = node->errors[error2];
+	      b3->error = node->errors[error3];
+	    }
+	  else
+	    {
+	      b0->error = b1->error = node->errors[DPDK_ERROR_NONE];
+	      b2->error = b3->error = node->errors[DPDK_ERROR_NONE];
+	    }
+
+	  l3_offset0 = device_input_next_node_advance[next0];
+	  l3_offset1 = device_input_next_node_advance[next1];
+	  l3_offset2 = device_input_next_node_advance[next2];
+	  l3_offset3 = device_input_next_node_advance[next3];
+
+	  b0->current_data = l3_offset0 + mb0->data_off;
+	  b1->current_data = l3_offset1 + mb1->data_off;
+	  b2->current_data = l3_offset2 + mb2->data_off;
+	  b3->current_data = l3_offset3 + mb3->data_off;
+
+	  b0->current_data -= RTE_PKTMBUF_HEADROOM;
+	  b1->current_data -= RTE_PKTMBUF_HEADROOM;
+	  b2->current_data -= RTE_PKTMBUF_HEADROOM;
+	  b3->current_data -= RTE_PKTMBUF_HEADROOM;
+
+	  b0->current_length = mb0->data_len - l3_offset0;
+	  b1->current_length = mb1->data_len - l3_offset1;
+	  b2->current_length = mb2->data_len - l3_offset2;
+	  b3->current_length = mb3->data_len - l3_offset3;
+
+	  b0->flags = buffer_flags_template;
+	  b1->flags = buffer_flags_template;
+	  b2->flags = buffer_flags_template;
+	  b3->flags = buffer_flags_template;
+
+	  vnet_buffer (b0)->sw_if_index[VLIB_RX] = xd->vlib_sw_if_index;
+	  vnet_buffer (b1)->sw_if_index[VLIB_RX] = xd->vlib_sw_if_index;
+	  vnet_buffer (b2)->sw_if_index[VLIB_RX] = xd->vlib_sw_if_index;
+	  vnet_buffer (b3)->sw_if_index[VLIB_RX] = xd->vlib_sw_if_index;
+
+	  vnet_buffer (b0)->sw_if_index[VLIB_TX] = (u32) ~ 0;
+	  vnet_buffer (b1)->sw_if_index[VLIB_TX] = (u32) ~ 0;
+	  vnet_buffer (b2)->sw_if_index[VLIB_TX] = (u32) ~ 0;
+	  vnet_buffer (b3)->sw_if_index[VLIB_TX] = (u32) ~ 0;
+
+	  n_rx_bytes += mb0->pkt_len;
+	  n_rx_bytes += mb1->pkt_len;
+	  n_rx_bytes += mb2->pkt_len;
+	  n_rx_bytes += mb3->pkt_len;
+
+	  /* Process subsequent segments of multi-segment packets */
+	  if (xd->flags & DPDK_DEVICE_FLAG_MAYBE_MULTISEG)
+	    {
+	      dpdk_process_subseq_segs (vm, b0, mb0, fl);
+	      dpdk_process_subseq_segs (vm, b1, mb1, fl);
+	      dpdk_process_subseq_segs (vm, b2, mb2, fl);
+	      dpdk_process_subseq_segs (vm, b3, mb3, fl);
+	    }
+
+	  /*
+	   * Turn this on if you run into
+	   * "bad monkey" contexts, and you want to know exactly
+	   * which nodes they've visited... See main.c...
+	   */
+	  VLIB_BUFFER_TRACE_TRAJECTORY_INIT (b0);
+	  VLIB_BUFFER_TRACE_TRAJECTORY_INIT (b1);
+	  VLIB_BUFFER_TRACE_TRAJECTORY_INIT (b2);
+	  VLIB_BUFFER_TRACE_TRAJECTORY_INIT (b3);
+
+	  /* Do we have any driver RX features configured on the interface? */
+	  vnet_feature_start_device_input_x4 (xd->vlib_sw_if_index,
+					      &next0, &next1, &next2, &next3,
+					      b0, b1, b2, b3,
+					      l3_offset0, l3_offset1,
+					      l3_offset2, l3_offset3);
+
+	  vlib_validate_buffer_enqueue_x4 (vm, node, next_index,
+					   to_next, n_left_to_next,
+					   bi0, bi1, bi2, bi3,
+					   next0, next1, next2, next3);
+	  n_buffers -= 4;
+	  mb_index += 4;
+	}
+      while (n_buffers > 0 && n_left_to_next > 0)
+	{
+	  struct rte_mbuf *mb0 = xd->rx_vectors[queue_id][mb_index];
+
+	  ASSERT (mb0);
+
+	  b0 = vlib_buffer_from_rte_mbuf (mb0);
+
+	  /* Prefetch one next segment if it exists. */
+	  if (PREDICT_FALSE (mb0->nb_segs > 1))
+	    dpdk_prefetch_buffer (mb0->next);
+
+	  vlib_buffer_init_for_free_list (b0, fl);
+
+	  bi0 = vlib_get_buffer_index (vm, b0);
+
+	  to_next[0] = bi0;
+	  to_next++;
+	  n_left_to_next--;
+
+	  if (PREDICT_FALSE (xd->per_interface_next_index != ~0))
+	    next0 = xd->per_interface_next_index;
+	  else if (PREDICT_TRUE
+		   ((xd->flags & DPDK_DEVICE_FLAG_PMD_SUPPORTS_PTYPE) != 0))
+	    next0 = dpdk_rx_next_from_mb (mb0, b0);
+	  else
+	    next0 = dpdk_rx_next_from_etype (mb0, b0);
+
+	  dpdk_rx_error_from_mb (mb0, &next0, &error0);
+	  b0->error = node->errors[error0];
+
+	  l3_offset0 = device_input_next_node_advance[next0];
+
+	  b0->current_data = l3_offset0;
+	  b0->current_data += mb0->data_off - RTE_PKTMBUF_HEADROOM;
+	  b0->current_length = mb0->data_len - l3_offset0;
+
+	  b0->flags = buffer_flags_template;
+
+	  vnet_buffer (b0)->sw_if_index[VLIB_RX] = xd->vlib_sw_if_index;
+	  vnet_buffer (b0)->sw_if_index[VLIB_TX] = (u32) ~ 0;
+	  n_rx_bytes += mb0->pkt_len;
+
+	  /* Process subsequent segments of multi-segment packets */
+	  dpdk_process_subseq_segs (vm, b0, mb0, fl);
+
+	  /*
+	   * Turn this on if you run into
+	   * "bad monkey" contexts, and you want to know exactly
+	   * which nodes they've visited... See main.c...
+	   */
+	  VLIB_BUFFER_TRACE_TRAJECTORY_INIT (b0);
+
+	  /* Do we have any driver RX features configured on the interface? */
+	  vnet_feature_start_device_input_x1 (xd->vlib_sw_if_index, &next0,
+					      b0, l3_offset0);
+
+	  vlib_validate_buffer_enqueue_x1 (vm, node, next_index,
+					   to_next, n_left_to_next,
+					   bi0, next0);
+	  n_buffers--;
+	  mb_index++;
+	}
+      vlib_put_next_frame (vm, node, next_index, n_left_to_next);
+    }
+
+  if (PREDICT_FALSE (vec_len (xd->d_trace_buffers[cpu_index]) > 0))
+    {
+      dpdk_rx_trace (dm, node, xd, queue_id, xd->d_trace_buffers[cpu_index],
+		     vec_len (xd->d_trace_buffers[cpu_index]));
+      vlib_set_trace_count (vm, node, n_trace -
+			    vec_len (xd->d_trace_buffers[cpu_index]));
+    }
+
+  vlib_increment_combined_counter
+    (vnet_get_main ()->interface_main.combined_sw_if_counters
+     + VNET_INTERFACE_COUNTER_RX,
+     cpu_index, xd->vlib_sw_if_index, mb_index, n_rx_bytes);
+
+  vnet_device_increment_rx_packets (cpu_index, mb_index);
+
+  return mb_index;
+}
+
+static inline void
+poll_rate_limit (dpdk_main_t * dm)
+{
+  /* Limit the poll rate by sleeping for N msec between polls */
+  if (PREDICT_FALSE (dm->poll_sleep != 0))
+    {
+      struct timespec ts, tsrem;
+
+      ts.tv_sec = 0;
+      ts.tv_nsec = 1000 * 1000 * dm->poll_sleep;	/* 1ms */
+
+      while (nanosleep (&ts, &tsrem) < 0)
+	{
+	  ts = tsrem;
+	}
+    }
+}
+
+/** \brief Main DPDK input node
+    @node dpdk-input
+
+    This is the main DPDK input node: across each assigned interface,
+    call rte_eth_rx_burst(...) or similar to obtain a vector of
+    packets to process. Handle early packet discard. Derive @c
+    vlib_buffer_t metadata from struct rte_mbuf metadata,
+    Depending on the resulting metadata: adjust b->current_data,
+    b->current_length  and dispatch directly to
+    ip4-input-no-checksum, or ip6-input. Trace the packet if required.
+
+    @param vm   vlib_main_t corresponding to the current thread
+    @param node vlib_node_runtime_t
+    @param f    vlib_frame_t input-node, not used.
+
+    @par Graph mechanics: buffer metadata, next index usage
+
+    @em Uses:
+    - struct rte_mbuf mb->ol_flags
+        - PKT_RX_IP_CKSUM_BAD
+    -  RTE_ETH_IS_xxx_HDR(mb->packet_type) 
+        - packet classification result
+
+    @em Sets:
+    - b->error if the packet is to be dropped immediately
+    - b->current_data, b->current_length
+        - adjusted as needed to skip the L2 header in  direct-dispatch cases
+    - vnet_buffer(b)->sw_if_index[VLIB_RX]
+        - rx interface sw_if_index
+    - vnet_buffer(b)->sw_if_index[VLIB_TX] = ~0
+        - required by ipX-lookup
+    - b->flags
+        - to indicate multi-segment pkts (VLIB_BUFFER_NEXT_PRESENT), etc.
+
+    Next Nodes:
+    - Static arcs to: error-drop, ethernet-input,
+      ip4-input-no-checksum, ip6-input, mpls-input
+    - per-interface redirection, controlled by
+      xd->per_interface_next_index
+*/
+
+static uword
+dpdk_input (vlib_main_t * vm, vlib_node_runtime_t * node, vlib_frame_t * f)
+{
+  dpdk_main_t *dm = &dpdk_main;
+  dpdk_device_t *xd;
+  uword n_rx_packets = 0;
+  dpdk_device_and_queue_t *dq;
+  u32 cpu_index = os_get_cpu_number ();
+
+  /*
+   * Poll all devices on this cpu for input/interrupts.
+   */
+  /* *INDENT-OFF* */
+  vec_foreach (dq, dm->devices_by_cpu[cpu_index])
+    {
+      xd = vec_elt_at_index(dm->devices, dq->device);
+      n_rx_packets += dpdk_device_input (dm, xd, node, cpu_index, dq->queue_id);
+    }
+  /* *INDENT-ON* */
+
+  poll_rate_limit (dm);
+
+  return n_rx_packets;
+}
+
+/* *INDENT-OFF* */
+VLIB_REGISTER_NODE (dpdk_input_node) = {
+  .function = dpdk_input,
+  .type = VLIB_NODE_TYPE_INPUT,
+  .name = "dpdk-input",
+  .sibling_of = "device-input",
+
+  /* Will be enabled if/when hardware is detected. */
+  .state = VLIB_NODE_STATE_DISABLED,
+
+  .format_buffer = format_ethernet_header_with_length,
+  .format_trace = format_dpdk_rx_dma_trace,
+
+  .n_errors = DPDK_N_ERROR,
+  .error_strings = dpdk_error_strings,
+};
+
+VLIB_NODE_FUNCTION_MULTIARCH (dpdk_input_node, dpdk_input);
+/* *INDENT-ON* */
+
+/*
+ * fd.io coding-style-patch-verification: ON
+ *
+ * Local Variables:
+ * eval: (c-set-style "gnu")
+ * End:
+ */
diff --git a/src/plugins/dpdk/dir.dox b/src/plugins/dpdk/dir.dox
new file mode 100644
index 00000000..43e36753
--- /dev/null
+++ b/src/plugins/dpdk/dir.dox
@@ -0,0 +1,27 @@
+/*
+ * Copyright (c) 2017 Cisco and/or its affiliates.
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at:
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+/* Doxygen directory documentation */
+
+/**
+@dir
+@brief DPDK Abstraction Layer.
+
+This directory contains the source code for the DPDK abstraction layer.
+
+*/
+/*? %%clicmd:group_label DPDK and pcap tx %% ?*/
+/*? %%syscfg:group_label DPDK and pcap tx %% ?*/
diff --git a/src/plugins/dpdk/hqos/hqos.c b/src/plugins/dpdk/hqos/hqos.c
new file mode 100644
index 00000000..a288fca7
--- /dev/null
+++ b/src/plugins/dpdk/hqos/hqos.c
@@ -0,0 +1,775 @@
+/*
+ * Copyright(c) 2016 Intel Corporation. All rights reserved.
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at:
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+#include 
+#include 
+#include 
+#include 
+#include 
+#include 
+#include 
+
+#include 
+#include 
+#include 
+#include 
+
+#include 
+#include 
+#include 
+
+#include 
+#include 
+#include 
+#include 	/* enumerate all vlib messages */
+
+#define vl_typedefs		/* define message structures */
+#include 
+#undef vl_typedefs
+
+/* instantiate all the print functions we know about */
+#define vl_print(handle, ...) vlib_cli_output (handle, __VA_ARGS__)
+#define vl_printfun
+#include 
+#undef vl_printfun
+
+#include 
+
+dpdk_main_t dpdk_main;
+
+/***
+ *
+ * HQoS default configuration values
+ *
+ ***/
+
+static dpdk_device_config_hqos_t hqos_params_default = {
+  .hqos_thread_valid = 0,
+
+  .swq_size = 4096,
+  .burst_enq = 256,
+  .burst_deq = 220,
+
+  /*
+   * Packet field to identify the subport.
+   *
+   * Default value: Since only one subport is defined by default (see below:
+   *     n_subports_per_port = 1), the subport ID is hardcoded to 0.
+   */
+  .pktfield0_slabpos = 0,
+  .pktfield0_slabmask = 0,
+
+  /*
+   * Packet field to identify the pipe.
+   *
+   * Default value: Assuming Ethernet/IPv4/UDP packets, UDP payload bits 12 .. 23
+   */
+  .pktfield1_slabpos = 40,
+  .pktfield1_slabmask = 0x0000000FFF000000LLU,
+
+  /* Packet field used as index into TC translation table to identify the traffic
+   *     class and queue.
+   *
+   * Default value: Assuming Ethernet/IPv4 packets, IPv4 DSCP field
+   */
+  .pktfield2_slabpos = 8,
+  .pktfield2_slabmask = 0x00000000000000FCLLU,
+  .tc_table = {
+	       0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15,
+	       0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15,
+	       0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15,
+	       0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15,
+	       },
+
+  /* port */
+  .port = {
+	   .name = NULL,	/* Set at init */
+	   .socket = 0,		/* Set at init */
+	   .rate = 1250000000,	/* Assuming 10GbE port */
+	   .mtu = 14 + 1500,	/* Assuming Ethernet/IPv4 pkt (Ethernet FCS not included) */
+	   .frame_overhead = RTE_SCHED_FRAME_OVERHEAD_DEFAULT,
+	   .n_subports_per_port = 1,
+	   .n_pipes_per_subport = 4096,
+	   .qsize = {64, 64, 64, 64},
+	   .pipe_profiles = NULL,	/* Set at config */
+	   .n_pipe_profiles = 1,
+
+#ifdef RTE_SCHED_RED
+	   .red_params = {
+			  /* Traffic Class 0 Colors Green / Yellow / Red */
+			  [0][0] = {.min_th = 48,.max_th = 64,.maxp_inv =
+				    10,.wq_log2 = 9},
+			  [0][1] = {.min_th = 40,.max_th = 64,.maxp_inv =
+				    10,.wq_log2 = 9},
+			  [0][2] = {.min_th = 32,.max_th = 64,.maxp_inv =
+				    10,.wq_log2 = 9},
+
+			  /* Traffic Class 1 - Colors Green / Yellow / Red */
+			  [1][0] = {.min_th = 48,.max_th = 64,.maxp_inv =
+				    10,.wq_log2 = 9},
+			  [1][1] = {.min_th = 40,.max_th = 64,.maxp_inv =
+				    10,.wq_log2 = 9},
+			  [1][2] = {.min_th = 32,.max_th = 64,.maxp_inv =
+				    10,.wq_log2 = 9},
+
+			  /* Traffic Class 2 - Colors Green / Yellow / Red */
+			  [2][0] = {.min_th = 48,.max_th = 64,.maxp_inv =
+				    10,.wq_log2 = 9},
+			  [2][1] = {.min_th = 40,.max_th = 64,.maxp_inv =
+				    10,.wq_log2 = 9},
+			  [2][2] = {.min_th = 32,.max_th = 64,.maxp_inv =
+				    10,.wq_log2 = 9},
+
+			  /* Traffic Class 3 - Colors Green / Yellow / Red */
+			  [3][0] = {.min_th = 48,.max_th = 64,.maxp_inv =
+				    10,.wq_log2 = 9},
+			  [3][1] = {.min_th = 40,.max_th = 64,.maxp_inv =
+				    10,.wq_log2 = 9},
+			  [3][2] = {.min_th = 32,.max_th = 64,.maxp_inv =
+				    10,.wq_log2 = 9}
+			  },
+#endif /* RTE_SCHED_RED */
+	   },
+};
+
+static struct rte_sched_subport_params hqos_subport_params_default = {
+  .tb_rate = 1250000000,	/* 10GbE line rate (measured in bytes/second) */
+  .tb_size = 1000000,
+  .tc_rate = {1250000000, 1250000000, 1250000000, 1250000000},
+  .tc_period = 10,
+};
+
+static struct rte_sched_pipe_params hqos_pipe_params_default = {
+  .tb_rate = 305175,		/* 10GbE line rate divided by 4K pipes */
+  .tb_size = 1000000,
+  .tc_rate = {305175, 305175, 305175, 305175},
+  .tc_period = 40,
+#ifdef RTE_SCHED_SUBPORT_TC_OV
+  .tc_ov_weight = 1,
+#endif
+  .wrr_weights = {1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1},
+};
+
+/***
+ *
+ * HQoS configuration
+ *
+ ***/
+
+int
+dpdk_hqos_validate_mask (u64 mask, u32 n)
+{
+  int count = __builtin_popcountll (mask);
+  int pos_lead = sizeof (u64) * 8 - __builtin_clzll (mask);
+  int pos_trail = __builtin_ctzll (mask);
+  int count_expected = __builtin_popcount (n - 1);
+
+  /* Handle the exceptions */
+  if (n == 0)
+    return -1;			/* Error */
+
+  if ((mask == 0) && (n == 1))
+    return 0;			/* OK */
+
+  if (((mask == 0) && (n != 1)) || ((mask != 0) && (n == 1)))
+    return -2;			/* Error */
+
+  /* Check that mask is contiguous */
+  if ((pos_lead - pos_trail) != count)
+    return -3;			/* Error */
+
+  /* Check that mask contains the expected number of bits set */
+  if (count != count_expected)
+    return -4;			/* Error */
+
+  return 0;			/* OK */
+}
+
+void
+dpdk_device_config_hqos_pipe_profile_default (dpdk_device_config_hqos_t *
+					      hqos, u32 pipe_profile_id)
+{
+  memcpy (&hqos->pipe[pipe_profile_id], &hqos_pipe_params_default,
+	  sizeof (hqos_pipe_params_default));
+}
+
+void
+dpdk_device_config_hqos_default (dpdk_device_config_hqos_t * hqos)
+{
+  struct rte_sched_subport_params *subport_params;
+  struct rte_sched_pipe_params *pipe_params;
+  u32 *pipe_map;
+  u32 i;
+
+  memcpy (hqos, &hqos_params_default, sizeof (hqos_params_default));
+
+  /* pipe */
+  vec_add2 (hqos->pipe, pipe_params, hqos->port.n_pipe_profiles);
+
+  for (i = 0; i < vec_len (hqos->pipe); i++)
+    memcpy (&pipe_params[i],
+	    &hqos_pipe_params_default, sizeof (hqos_pipe_params_default));
+
+  hqos->port.pipe_profiles = hqos->pipe;
+
+  /* subport */
+  vec_add2 (hqos->subport, subport_params, hqos->port.n_subports_per_port);
+
+  for (i = 0; i < vec_len (hqos->subport); i++)
+    memcpy (&subport_params[i],
+	    &hqos_subport_params_default,
+	    sizeof (hqos_subport_params_default));
+
+  /* pipe profile */
+  vec_add2 (hqos->pipe_map,
+	    pipe_map,
+	    hqos->port.n_subports_per_port * hqos->port.n_pipes_per_subport);
+
+  for (i = 0; i < vec_len (hqos->pipe_map); i++)
+    pipe_map[i] = 0;
+}
+
+/***
+ *
+ * HQoS init
+ *
+ ***/
+
+clib_error_t *
+dpdk_port_setup_hqos (dpdk_device_t * xd, dpdk_device_config_hqos_t * hqos)
+{
+  vlib_thread_main_t *tm = vlib_get_thread_main ();
+  char name[32];
+  u32 subport_id, i;
+  int rv;
+
+  /* Detect the set of worker threads */
+  int worker_thread_first = 0;
+  int worker_thread_count = 0;
+
+  uword *p = hash_get_mem (tm->thread_registrations_by_name, "workers");
+  vlib_thread_registration_t *tr =
+    p ? (vlib_thread_registration_t *) p[0] : 0;
+
+  if (tr && tr->count > 0)
+    {
+      worker_thread_first = tr->first_index;
+      worker_thread_count = tr->count;
+    }
+
+  /* Allocate the per-thread device data array */
+  vec_validate_aligned (xd->hqos_wt, tm->n_vlib_mains - 1,
+			CLIB_CACHE_LINE_BYTES);
+  memset (xd->hqos_wt, 0, tm->n_vlib_mains * sizeof (xd->hqos_wt[0]));
+
+  vec_validate_aligned (xd->hqos_ht, 0, CLIB_CACHE_LINE_BYTES);
+  memset (xd->hqos_ht, 0, sizeof (xd->hqos_ht[0]));
+
+  /* Allocate space for one SWQ per worker thread in the I/O TX thread data structure */
+  vec_validate (xd->hqos_ht->swq, worker_thread_count);
+
+  /* SWQ */
+  for (i = 0; i < worker_thread_count + 1; i++)
+    {
+      u32 swq_flags = RING_F_SP_ENQ | RING_F_SC_DEQ;
+
+      snprintf (name, sizeof (name), "SWQ-worker%u-to-device%u", i,
+		xd->device_index);
+      xd->hqos_ht->swq[i] =
+	rte_ring_create (name, hqos->swq_size, xd->cpu_socket, swq_flags);
+      if (xd->hqos_ht->swq[i] == NULL)
+	return clib_error_return (0,
+				  "SWQ-worker%u-to-device%u: rte_ring_create err",
+				  i, xd->device_index);
+    }
+
+  /*
+   * HQoS
+   */
+
+  /* HQoS port */
+  snprintf (name, sizeof (name), "HQoS%u", xd->device_index);
+  hqos->port.name = strdup (name);
+  if (hqos->port.name == NULL)
+    return clib_error_return (0, "HQoS%u: strdup err", xd->device_index);
+
+  hqos->port.socket = rte_eth_dev_socket_id (xd->device_index);
+  if (hqos->port.socket == SOCKET_ID_ANY)
+    hqos->port.socket = 0;
+
+  xd->hqos_ht->hqos = rte_sched_port_config (&hqos->port);
+  if (xd->hqos_ht->hqos == NULL)
+    return clib_error_return (0, "HQoS%u: rte_sched_port_config err",
+			      xd->device_index);
+
+  /* HQoS subport */
+  for (subport_id = 0; subport_id < hqos->port.n_subports_per_port;
+       subport_id++)
+    {
+      u32 pipe_id;
+
+      rv =
+	rte_sched_subport_config (xd->hqos_ht->hqos, subport_id,
+				  &hqos->subport[subport_id]);
+      if (rv)
+	return clib_error_return (0,
+				  "HQoS%u subport %u: rte_sched_subport_config err (%d)",
+				  xd->device_index, subport_id, rv);
+
+      /* HQoS pipe */
+      for (pipe_id = 0; pipe_id < hqos->port.n_pipes_per_subport; pipe_id++)
+	{
+	  u32 pos = subport_id * hqos->port.n_pipes_per_subport + pipe_id;
+	  u32 profile_id = hqos->pipe_map[pos];
+
+	  rv =
+	    rte_sched_pipe_config (xd->hqos_ht->hqos, subport_id, pipe_id,
+				   profile_id);
+	  if (rv)
+	    return clib_error_return (0,
+				      "HQoS%u subport %u pipe %u: rte_sched_pipe_config err (%d)",
+				      xd->device_index, subport_id, pipe_id,
+				      rv);
+	}
+    }
+
+  /* Set up per-thread device data for the I/O TX thread */
+  xd->hqos_ht->hqos_burst_enq = hqos->burst_enq;
+  xd->hqos_ht->hqos_burst_deq = hqos->burst_deq;
+  vec_validate (xd->hqos_ht->pkts_enq, 2 * hqos->burst_enq - 1);
+  vec_validate (xd->hqos_ht->pkts_deq, hqos->burst_deq - 1);
+  xd->hqos_ht->pkts_enq_len = 0;
+  xd->hqos_ht->swq_pos = 0;
+  xd->hqos_ht->flush_count = 0;
+
+  /* Set up per-thread device data for each worker thread */
+  for (i = 0; i < worker_thread_count + 1; i++)
+    {
+      u32 tid;
+      if (i)
+	tid = worker_thread_first + (i - 1);
+      else
+	tid = i;
+
+      xd->hqos_wt[tid].swq = xd->hqos_ht->swq[i];
+      xd->hqos_wt[tid].hqos_field0_slabpos = hqos->pktfield0_slabpos;
+      xd->hqos_wt[tid].hqos_field0_slabmask = hqos->pktfield0_slabmask;
+      xd->hqos_wt[tid].hqos_field0_slabshr =
+	__builtin_ctzll (hqos->pktfield0_slabmask);
+      xd->hqos_wt[tid].hqos_field1_slabpos = hqos->pktfield1_slabpos;
+      xd->hqos_wt[tid].hqos_field1_slabmask = hqos->pktfield1_slabmask;
+      xd->hqos_wt[tid].hqos_field1_slabshr =
+	__builtin_ctzll (hqos->pktfield1_slabmask);
+      xd->hqos_wt[tid].hqos_field2_slabpos = hqos->pktfield2_slabpos;
+      xd->hqos_wt[tid].hqos_field2_slabmask = hqos->pktfield2_slabmask;
+      xd->hqos_wt[tid].hqos_field2_slabshr =
+	__builtin_ctzll (hqos->pktfield2_slabmask);
+      memcpy (xd->hqos_wt[tid].hqos_tc_table, hqos->tc_table,
+	      sizeof (hqos->tc_table));
+    }
+
+  return 0;
+}
+
+/***
+ *
+ * HQoS run-time
+ *
+ ***/
+/*
+ * dpdk_hqos_thread - Contains the main loop of an HQoS thread.
+ *
+ * w
+ *     Information for the current thread
+ */
+static_always_inline void
+dpdk_hqos_thread_internal_hqos_dbg_bypass (vlib_main_t * vm)
+{
+  dpdk_main_t *dm = &dpdk_main;
+  u32 cpu_index = vm->cpu_index;
+  u32 dev_pos;
+
+  dev_pos = 0;
+  while (1)
+    {
+      vlib_worker_thread_barrier_check ();
+
+      u32 n_devs = vec_len (dm->devices_by_hqos_cpu[cpu_index]);
+      if (dev_pos >= n_devs)
+	dev_pos = 0;
+
+      dpdk_device_and_queue_t *dq =
+	vec_elt_at_index (dm->devices_by_hqos_cpu[cpu_index], dev_pos);
+      dpdk_device_t *xd = vec_elt_at_index (dm->devices, dq->device);
+
+      dpdk_device_hqos_per_hqos_thread_t *hqos = xd->hqos_ht;
+      u32 device_index = xd->device_index;
+      u16 queue_id = dq->queue_id;
+
+      struct rte_mbuf **pkts_enq = hqos->pkts_enq;
+      u32 pkts_enq_len = hqos->pkts_enq_len;
+      u32 swq_pos = hqos->swq_pos;
+      u32 n_swq = vec_len (hqos->swq), i;
+      u32 flush_count = hqos->flush_count;
+
+      for (i = 0; i < n_swq; i++)
+	{
+	  /* Get current SWQ for this device */
+	  struct rte_ring *swq = hqos->swq[swq_pos];
+
+	  /* Read SWQ burst to packet buffer of this device */
+	  pkts_enq_len += rte_ring_sc_dequeue_burst (swq,
+						     (void **)
+						     &pkts_enq[pkts_enq_len],
+						     hqos->hqos_burst_enq);
+
+	  /* Get next SWQ for this device */
+	  swq_pos++;
+	  if (swq_pos >= n_swq)
+	    swq_pos = 0;
+	  hqos->swq_pos = swq_pos;
+
+	  /* HWQ TX enqueue when burst available */
+	  if (pkts_enq_len >= hqos->hqos_burst_enq)
+	    {
+	      u32 n_pkts = rte_eth_tx_burst (device_index,
+					     (uint16_t) queue_id,
+					     pkts_enq,
+					     (uint16_t) pkts_enq_len);
+
+	      for (; n_pkts < pkts_enq_len; n_pkts++)
+		rte_pktmbuf_free (pkts_enq[n_pkts]);
+
+	      pkts_enq_len = 0;
+	      flush_count = 0;
+	      break;
+	    }
+	}
+      if (pkts_enq_len)
+	{
+	  flush_count++;
+	  if (PREDICT_FALSE (flush_count == HQOS_FLUSH_COUNT_THRESHOLD))
+	    {
+	      rte_sched_port_enqueue (hqos->hqos, pkts_enq, pkts_enq_len);
+
+	      pkts_enq_len = 0;
+	      flush_count = 0;
+	    }
+	}
+      hqos->pkts_enq_len = pkts_enq_len;
+      hqos->flush_count = flush_count;
+
+      /* Advance to next device */
+      dev_pos++;
+    }
+}
+
+static_always_inline void
+dpdk_hqos_thread_internal (vlib_main_t * vm)
+{
+  dpdk_main_t *dm = &dpdk_main;
+  u32 cpu_index = vm->cpu_index;
+  u32 dev_pos;
+
+  dev_pos = 0;
+  while (1)
+    {
+      vlib_worker_thread_barrier_check ();
+
+      u32 n_devs = vec_len (dm->devices_by_hqos_cpu[cpu_index]);
+      if (PREDICT_FALSE (n_devs == 0))
+	{
+	  dev_pos = 0;
+	  continue;
+	}
+      if (dev_pos >= n_devs)
+	dev_pos = 0;
+
+      dpdk_device_and_queue_t *dq =
+	vec_elt_at_index (dm->devices_by_hqos_cpu[cpu_index], dev_pos);
+      dpdk_device_t *xd = vec_elt_at_index (dm->devices, dq->device);
+
+      dpdk_device_hqos_per_hqos_thread_t *hqos = xd->hqos_ht;
+      u32 device_index = xd->device_index;
+      u16 queue_id = dq->queue_id;
+
+      struct rte_mbuf **pkts_enq = hqos->pkts_enq;
+      struct rte_mbuf **pkts_deq = hqos->pkts_deq;
+      u32 pkts_enq_len = hqos->pkts_enq_len;
+      u32 swq_pos = hqos->swq_pos;
+      u32 n_swq = vec_len (hqos->swq), i;
+      u32 flush_count = hqos->flush_count;
+
+      /*
+       * SWQ dequeue and HQoS enqueue for current device
+       */
+      for (i = 0; i < n_swq; i++)
+	{
+	  /* Get current SWQ for this device */
+	  struct rte_ring *swq = hqos->swq[swq_pos];
+
+	  /* Read SWQ burst to packet buffer of this device */
+	  pkts_enq_len += rte_ring_sc_dequeue_burst (swq,
+						     (void **)
+						     &pkts_enq[pkts_enq_len],
+						     hqos->hqos_burst_enq);
+
+	  /* Get next SWQ for this device */
+	  swq_pos++;
+	  if (swq_pos >= n_swq)
+	    swq_pos = 0;
+	  hqos->swq_pos = swq_pos;
+
+	  /* HQoS enqueue when burst available */
+	  if (pkts_enq_len >= hqos->hqos_burst_enq)
+	    {
+	      rte_sched_port_enqueue (hqos->hqos, pkts_enq, pkts_enq_len);
+
+	      pkts_enq_len = 0;
+	      flush_count = 0;
+	      break;
+	    }
+	}
+      if (pkts_enq_len)
+	{
+	  flush_count++;
+	  if (PREDICT_FALSE (flush_count == HQOS_FLUSH_COUNT_THRESHOLD))
+	    {
+	      rte_sched_port_enqueue (hqos->hqos, pkts_enq, pkts_enq_len);
+
+	      pkts_enq_len = 0;
+	      flush_count = 0;
+	    }
+	}
+      hqos->pkts_enq_len = pkts_enq_len;
+      hqos->flush_count = flush_count;
+
+      /*
+       * HQoS dequeue and HWQ TX enqueue for current device
+       */
+      {
+	u32 pkts_deq_len, n_pkts;
+
+	pkts_deq_len = rte_sched_port_dequeue (hqos->hqos,
+					       pkts_deq,
+					       hqos->hqos_burst_deq);
+
+	for (n_pkts = 0; n_pkts < pkts_deq_len;)
+	  n_pkts += rte_eth_tx_burst (device_index,
+				      (uint16_t) queue_id,
+				      &pkts_deq[n_pkts],
+				      (uint16_t) (pkts_deq_len - n_pkts));
+      }
+
+      /* Advance to next device */
+      dev_pos++;
+    }
+}
+
+void
+dpdk_hqos_thread (vlib_worker_thread_t * w)
+{
+  vlib_main_t *vm;
+  vlib_thread_main_t *tm = vlib_get_thread_main ();
+  dpdk_main_t *dm = &dpdk_main;
+
+  vm = vlib_get_main ();
+
+  ASSERT (vm->cpu_index == os_get_cpu_number ());
+
+  clib_time_init (&vm->clib_time);
+  clib_mem_set_heap (w->thread_mheap);
+
+  /* Wait until the dpdk init sequence is complete */
+  while (tm->worker_thread_release == 0)
+    vlib_worker_thread_barrier_check ();
+
+  if (vec_len (dm->devices_by_hqos_cpu[vm->cpu_index]) == 0)
+    return
+      clib_error
+      ("current I/O TX thread does not have any devices assigned to it");
+
+  if (DPDK_HQOS_DBG_BYPASS)
+    dpdk_hqos_thread_internal_hqos_dbg_bypass (vm);
+  else
+    dpdk_hqos_thread_internal (vm);
+}
+
+void
+dpdk_hqos_thread_fn (void *arg)
+{
+  vlib_worker_thread_t *w = (vlib_worker_thread_t *) arg;
+  vlib_worker_thread_init (w);
+  dpdk_hqos_thread (w);
+}
+
+/* *INDENT-OFF* */
+VLIB_REGISTER_THREAD (hqos_thread_reg, static) =
+{
+  .name = "hqos-threads",
+  .short_name = "hqos-threads",
+  .function = dpdk_hqos_thread_fn,
+};
+/* *INDENT-ON* */
+
+/*
+ * HQoS run-time code to be called by the worker threads
+ */
+#define BITFIELD(byte_array, slab_pos, slab_mask, slab_shr)     \
+({                                                              \
+  u64 slab = *((u64 *) &byte_array[slab_pos]);                  \
+  u64 val = (rte_be_to_cpu_64(slab) & slab_mask) >> slab_shr;   \
+  val;                                                          \
+})
+
+#define RTE_SCHED_PORT_HIERARCHY(subport, pipe, traffic_class, queue, color) \
+  ((((u64) (queue)) & 0x3) |                               \
+  ((((u64) (traffic_class)) & 0x3) << 2) |                 \
+  ((((u64) (color)) & 0x3) << 4) |                         \
+  ((((u64) (subport)) & 0xFFFF) << 16) |                   \
+  ((((u64) (pipe)) & 0xFFFFFFFF) << 32))
+
+void
+dpdk_hqos_metadata_set (dpdk_device_hqos_per_worker_thread_t * hqos,
+			struct rte_mbuf **pkts, u32 n_pkts)
+{
+  u32 i;
+
+  for (i = 0; i < (n_pkts & (~0x3)); i += 4)
+    {
+      struct rte_mbuf *pkt0 = pkts[i];
+      struct rte_mbuf *pkt1 = pkts[i + 1];
+      struct rte_mbuf *pkt2 = pkts[i + 2];
+      struct rte_mbuf *pkt3 = pkts[i + 3];
+
+      u8 *pkt0_data = rte_pktmbuf_mtod (pkt0, u8 *);
+      u8 *pkt1_data = rte_pktmbuf_mtod (pkt1, u8 *);
+      u8 *pkt2_data = rte_pktmbuf_mtod (pkt2, u8 *);
+      u8 *pkt3_data = rte_pktmbuf_mtod (pkt3, u8 *);
+
+      u64 pkt0_subport = BITFIELD (pkt0_data, hqos->hqos_field0_slabpos,
+				   hqos->hqos_field0_slabmask,
+				   hqos->hqos_field0_slabshr);
+      u64 pkt0_pipe = BITFIELD (pkt0_data, hqos->hqos_field1_slabpos,
+				hqos->hqos_field1_slabmask,
+				hqos->hqos_field1_slabshr);
+      u64 pkt0_dscp = BITFIELD (pkt0_data, hqos->hqos_field2_slabpos,
+				hqos->hqos_field2_slabmask,
+				hqos->hqos_field2_slabshr);
+      u32 pkt0_tc = hqos->hqos_tc_table[pkt0_dscp & 0x3F] >> 2;
+      u32 pkt0_tc_q = hqos->hqos_tc_table[pkt0_dscp & 0x3F] & 0x3;
+
+      u64 pkt1_subport = BITFIELD (pkt1_data, hqos->hqos_field0_slabpos,
+				   hqos->hqos_field0_slabmask,
+				   hqos->hqos_field0_slabshr);
+      u64 pkt1_pipe = BITFIELD (pkt1_data, hqos->hqos_field1_slabpos,
+				hqos->hqos_field1_slabmask,
+				hqos->hqos_field1_slabshr);
+      u64 pkt1_dscp = BITFIELD (pkt1_data, hqos->hqos_field2_slabpos,
+				hqos->hqos_field2_slabmask,
+				hqos->hqos_field2_slabshr);
+      u32 pkt1_tc = hqos->hqos_tc_table[pkt1_dscp & 0x3F] >> 2;
+      u32 pkt1_tc_q = hqos->hqos_tc_table[pkt1_dscp & 0x3F] & 0x3;
+
+      u64 pkt2_subport = BITFIELD (pkt2_data, hqos->hqos_field0_slabpos,
+				   hqos->hqos_field0_slabmask,
+				   hqos->hqos_field0_slabshr);
+      u64 pkt2_pipe = BITFIELD (pkt2_data, hqos->hqos_field1_slabpos,
+				hqos->hqos_field1_slabmask,
+				hqos->hqos_field1_slabshr);
+      u64 pkt2_dscp = BITFIELD (pkt2_data, hqos->hqos_field2_slabpos,
+				hqos->hqos_field2_slabmask,
+				hqos->hqos_field2_slabshr);
+      u32 pkt2_tc = hqos->hqos_tc_table[pkt2_dscp & 0x3F] >> 2;
+      u32 pkt2_tc_q = hqos->hqos_tc_table[pkt2_dscp & 0x3F] & 0x3;
+
+      u64 pkt3_subport = BITFIELD (pkt3_data, hqos->hqos_field0_slabpos,
+				   hqos->hqos_field0_slabmask,
+				   hqos->hqos_field0_slabshr);
+      u64 pkt3_pipe = BITFIELD (pkt3_data, hqos->hqos_field1_slabpos,
+				hqos->hqos_field1_slabmask,
+				hqos->hqos_field1_slabshr);
+      u64 pkt3_dscp = BITFIELD (pkt3_data, hqos->hqos_field2_slabpos,
+				hqos->hqos_field2_slabmask,
+				hqos->hqos_field2_slabshr);
+      u32 pkt3_tc = hqos->hqos_tc_table[pkt3_dscp & 0x3F] >> 2;
+      u32 pkt3_tc_q = hqos->hqos_tc_table[pkt3_dscp & 0x3F] & 0x3;
+
+      u64 pkt0_sched = RTE_SCHED_PORT_HIERARCHY (pkt0_subport,
+						 pkt0_pipe,
+						 pkt0_tc,
+						 pkt0_tc_q,
+						 0);
+      u64 pkt1_sched = RTE_SCHED_PORT_HIERARCHY (pkt1_subport,
+						 pkt1_pipe,
+						 pkt1_tc,
+						 pkt1_tc_q,
+						 0);
+      u64 pkt2_sched = RTE_SCHED_PORT_HIERARCHY (pkt2_subport,
+						 pkt2_pipe,
+						 pkt2_tc,
+						 pkt2_tc_q,
+						 0);
+      u64 pkt3_sched = RTE_SCHED_PORT_HIERARCHY (pkt3_subport,
+						 pkt3_pipe,
+						 pkt3_tc,
+						 pkt3_tc_q,
+						 0);
+
+      pkt0->hash.sched.lo = pkt0_sched & 0xFFFFFFFF;
+      pkt0->hash.sched.hi = pkt0_sched >> 32;
+      pkt1->hash.sched.lo = pkt1_sched & 0xFFFFFFFF;
+      pkt1->hash.sched.hi = pkt1_sched >> 32;
+      pkt2->hash.sched.lo = pkt2_sched & 0xFFFFFFFF;
+      pkt2->hash.sched.hi = pkt2_sched >> 32;
+      pkt3->hash.sched.lo = pkt3_sched & 0xFFFFFFFF;
+      pkt3->hash.sched.hi = pkt3_sched >> 32;
+    }
+
+  for (; i < n_pkts; i++)
+    {
+      struct rte_mbuf *pkt = pkts[i];
+
+      u8 *pkt_data = rte_pktmbuf_mtod (pkt, u8 *);
+
+      u64 pkt_subport = BITFIELD (pkt_data, hqos->hqos_field0_slabpos,
+				  hqos->hqos_field0_slabmask,
+				  hqos->hqos_field0_slabshr);
+      u64 pkt_pipe = BITFIELD (pkt_data, hqos->hqos_field1_slabpos,
+			       hqos->hqos_field1_slabmask,
+			       hqos->hqos_field1_slabshr);
+      u64 pkt_dscp = BITFIELD (pkt_data, hqos->hqos_field2_slabpos,
+			       hqos->hqos_field2_slabmask,
+			       hqos->hqos_field2_slabshr);
+      u32 pkt_tc = hqos->hqos_tc_table[pkt_dscp & 0x3F] >> 2;
+      u32 pkt_tc_q = hqos->hqos_tc_table[pkt_dscp & 0x3F] & 0x3;
+
+      u64 pkt_sched = RTE_SCHED_PORT_HIERARCHY (pkt_subport,
+						pkt_pipe,
+						pkt_tc,
+						pkt_tc_q,
+						0);
+
+      pkt->hash.sched.lo = pkt_sched & 0xFFFFFFFF;
+      pkt->hash.sched.hi = pkt_sched >> 32;
+    }
+}
+
+/*
+ * fd.io coding-style-patch-verification: ON
+ *
+ * Local Variables:
+ * eval: (c-set-style "gnu")
+ * End:
+ */
diff --git a/src/plugins/dpdk/hqos/qos_doc.md b/src/plugins/dpdk/hqos/qos_doc.md
new file mode 100644
index 00000000..7c064246
--- /dev/null
+++ b/src/plugins/dpdk/hqos/qos_doc.md
@@ -0,0 +1,411 @@
+# QoS Hierarchical Scheduler    {#qos_doc}
+
+The Quality-of-Service (QoS) scheduler performs egress-traffic management by
+prioritizing the transmission of the packets of different type services and
+subcribers based on the Service Level Agreements (SLAs). The QoS scheduler can
+be enabled on one or more NIC output interfaces depending upon the
+requirement.
+
+
+## Overview
+
+The QoS schdeuler supports a number of scheduling and shaping levels which
+construct hierarchical-tree. The first level in the hierarchy is port (i.e.
+the physical interface) that constitutes the root node of the tree. The
+subsequent level is subport which represents the group of the
+users/subscribers. The individual user/subscriber is represented by the pipe
+at the next level. Each user can have different traffic type based on the
+criteria of specific loss rate, jitter, and latency. These traffic types are
+represented at the traffic-class level in the form of different traffic-
+classes. The last level contains number of queues which are grouped together
+to host the packets of the specific class type traffic.
+
+The QoS scheduler implementation requires flow classification, enqueue  and
+dequeue operations. The flow classification is mandatory stage for HQoS where
+incoming packets are classified by mapping the packet fields information to
+5-tuple (HQoS subport, pipe, traffic class, queue within traffic class, and
+color) and storing that information in mbuf sched field. The enqueue operation
+uses this information to determine the queue for storing the packet, and at
+this stage, if the specific queue is full, QoS drops the packet. The dequeue
+operation consists of scheduling the packet based on its length and available
+credits, and handing over the scheduled packet to the output interface.
+
+For more information on QoS Scheduler, please refer DPDK Programmer's Guide-
+http://dpdk.org/doc/guides/prog_guide/qos_framework.html
+
+
+### QoS Schdeuler Parameters
+
+Following illustrates the default HQoS configuration for each 10GbE output
+port:
+
+Single subport (subport 0):
+  - Subport rate set to 100% of port rate
+  - Each of the 4 traffic classes has rate set to 100% of port rate
+
+4K pipes per subport 0 (pipes 0 .. 4095) with identical configuration:
+  - Pipe rate set to 1/4K of port rate
+  - Each of the 4 traffic classes has rate set to 100% of pipe rate
+  - Within each traffic class, the byte-level WRR weights for the 4 queues are set to 1:1:1:1
+
+
+#### Port configuration
+
+```
+port {
+  rate 1250000000           /* Assuming 10GbE port */
+  frame_overhead 24         /* Overhead fields per Ethernet frame:
+                             * 7B (Preamble) +
+                             * 1B (Start of Frame Delimiter (SFD)) + 
+                             * 4B (Frame Check Sequence (FCS)) +
+                             * 12B (Inter Frame Gap (IFG))
+                             */
+  mtu 1522                  /* Assuming Ethernet/IPv4 pkt (FCS not included) */
+  n_subports_per_port 1     /* Number of subports per output interface */
+  n_pipes_per_subport 4096  /* Number of pipes (users/subscribers) */
+  queue_sizes 64 64 64 64   /* Packet queue size for each traffic class.
+                             * All queues within the same pipe traffic class
+                             * have the same size. Queues from different
+                             * pipes serving the same traffic class have
+                             * the same size. */
+}
+```
+
+
+#### Subport configuration
+
+```
+subport 0 {
+  tb_rate 1250000000        /* Subport level token bucket rate (bytes per second) */
+  tb_size 1000000           /* Subport level token bucket size (bytes) */
+  tc0_rate 1250000000       /* Subport level token bucket rate for traffic class 0 (bytes per second) */
+  tc1_rate 1250000000       /* Subport level token bucket rate for traffic class 1 (bytes per second) */
+  tc2_rate 1250000000       /* Subport level token bucket rate for traffic class 2 (bytes per second) */
+  tc3_rate 1250000000       /* Subport level token bucket rate for traffic class 3 (bytes per second) */
+  tc_period 10              /* Time interval for refilling the token bucket associated with traffic class (Milliseconds) */
+  pipe 0 4095 profile 0     /* pipes (users/subscribers) configured with pipe profile 0 */
+}
+```
+
+
+#### Pipe configuration
+
+```
+pipe_profile 0 {
+  tb_rate 305175          /* Pipe level token bucket rate (bytes per second) */
+  tb_size 1000000         /* Pipe level token bucket size (bytes) */
+  tc0_rate 305175         /* Pipe level token bucket rate for traffic class 0 (bytes per second) */
+  tc1_rate 305175         /* Pipe level token bucket rate for traffic class 1 (bytes per second) */
+  tc2_rate 305175         /* Pipe level token bucket rate for traffic class 2 (bytes per second) */
+  tc3_rate 305175         /* Pipe level token bucket rate for traffic class 3 (bytes per second) */
+  tc_period 40            /* Time interval for refilling the token bucket associated with traffic class at pipe level (Milliseconds) */
+  tc3_oversubscription_weight 1 /* Weight traffic class 3 oversubscription */
+  tc0_wrr_weights 1 1 1 1     /* Pipe queues WRR weights for traffic class 0 */
+  tc1_wrr_weights 1 1 1 1     /* Pipe queues WRR weights for traffic class 1 */
+  tc2_wrr_weights 1 1 1 1     /* Pipe queues WRR weights for traffic class 2 */
+  tc3_wrr_weights 1 1 1 1     /* Pipe queues WRR weights for traffic class 3 */
+}
+```
+
+
+#### Random Early Detection (RED) parameters per traffic class and color (Green / Yellow / Red)
+
+```
+red {
+  tc0_wred_min 48 40 32     /* Minimum threshold for traffic class 0 queue (min_th) in number of packets */
+  tc0_wred_max 64 64 64     /* Maximum threshold for traffic class 0 queue (max_th) in number of packets */
+  tc0_wred_inv_prob 10 10 10    /* Inverse of packet marking probability for traffic class 0 queue (maxp = 1 / maxp_inv) */
+  tc0_wred_weight 9 9 9     /* Traffic Class 0 queue weight */
+  tc1_wred_min 48 40 32     /* Minimum threshold for traffic class 1 queue (min_th) in number of packets */
+  tc1_wred_max 64 64 64     /* Maximum threshold for traffic class 1 queue (max_th) in number of packets */
+  tc1_wred_inv_prob 10 10 10    /* Inverse of packet marking probability for traffic class 1 queue (maxp = 1 / maxp_inv) */
+  tc1_wred_weight 9 9 9     /* Traffic Class 1 queue weight */
+  tc2_wred_min 48 40 32     /* Minimum threshold for traffic class 2 queue (min_th) in number of packets */
+  tc2_wred_max 64 64 64     /* Maximum threshold for traffic class 2 queue (max_th) in number of packets */
+  tc2_wred_inv_prob 10 10 10    /* Inverse of packet marking probability for traffic class 2 queue (maxp = 1 / maxp_inv) */
+  tc2_wred_weight 9 9 9     /* Traffic Class 2 queue weight */
+  tc3_wred_min 48 40 32     /* Minimum threshold for traffic class 3 queue (min_th) in number of packets */
+  tc3_wred_max 64 64 64     /* Maximum threshold for traffic class 3 queue (max_th) in number of packets */
+  tc3_wred_inv_prob 10 10 10    /* Inverse of packet marking probability for traffic class 3 queue (maxp = 1 / maxp_inv) */
+  tc3_wred_weight 9 9 9     /* Traffic Class 3 queue weight */
+}
+```
+
+
+### DPDK QoS Scheduler Integration in VPP
+
+The Hierarchical Quaity-of-Service (HQoS) scheduler object could be seen as
+part of the logical NIC output interface. To enable HQoS on specific output
+interface, vpp startup.conf file has to be configured accordingly. The output
+interface that requires HQoS, should have "hqos" parameter specified in dpdk
+section. Another optional parameter "hqos-thread"  has been defined which can
+be used to associate the output interface with specific hqos thread. In cpu
+section of the config file, "corelist-hqos-threads" is introduced to assign
+logical cpu cores to run the HQoS threads. A HQoS thread can run multiple HQoS
+objects each associated with different output interfaces. All worker threads
+instead of writing packets to NIC TX queue directly, write the packets to a
+software queues. The hqos_threads read the software queues, and enqueue the
+packets to HQoS objects, as well as dequeue packets from HQOS objects and
+write them to NIC output interfaces. The worker threads need to be able to
+send the packets to any output interface, therefore, each HQoS object
+associated with NIC output interface should have software queues equal to
+worker threads count.
+
+Following illustrates the sample startup configuration file with 4x worker
+threads feeding 2x hqos threads that handle each QoS scheduler for 1x output
+interface.
+
+```
+dpdk {
+  socket-mem 16384,16384
+
+  dev 0000:02:00.0 {
+    num-rx-queues 2
+    hqos
+  }
+  dev 0000:06:00.0 {
+    num-rx-queues 2
+    hqos
+  }
+
+  num-mbufs 1000000
+}
+
+cpu {
+  main-core 0
+  corelist-workers  1, 2, 3, 4
+  corelist-hqos-threads  5, 6
+}
+```
+
+
+### QoS scheduler CLI Commands
+
+Each QoS scheduler instance is initialised with default parameters required to
+configure hqos port, subport, pipe and queues. Some of the parameters can be
+re-configured in run-time through CLI commands.
+
+
+#### Configuration
+
+Following commands can be used to configure QoS scheduler parameters.
+
+The command below can be used to set the subport level parameters such as
+token bucket rate (bytes per seconds), token bucket size (bytes), traffic
+class rates (bytes per seconds) and token update period (Milliseconds).
+
+```
+set dpdk interface hqos subport  subport  [rate ]
+    [bktsize ] [tc0 ] [tc1 ] [tc2 ] [tc3 ] [period ]
+```
+
+For setting the pipe profile, following command can be used.    
+
+```
+set dpdk interface hqos pipe  subport  pipe 
+    profile 
+```
+
+To assign QoS scheduler instance to the specific thread, following command can
+be used.
+
+```
+set dpdk interface hqos placement  thread 
+```
+
+The command below is used to set the packet fields required for classifiying
+the incoming packet. As a result of classification process,     packet field
+information will be mapped to 5 tuples (subport, pipe, traffic class, pipe,
+color) and stored in packet mbuf.
+
+```
+set dpdk interface hqos pktfield  id subport|pipe|tc offset 
+    mask 
+```
+
+The DSCP table entries used for idenfiying the traffic class and queue can be set using the command below;   
+
+```  
+set dpdk interface hqos tctbl  entry  tc  queue 
+```
+
+
+#### Show Command
+
+The QoS Scheduler configuration can displayed using the command below.
+
+```
+   vpp# show dpdk interface hqos TenGigabitEthernet2/0/0
+   Thread:
+     Input SWQ size = 4096 packets
+     Enqueue burst size = 256 packets
+     Dequeue burst size = 220 packets
+     Packet field 0: slab position =    0, slab bitmask = 0x0000000000000000   (subport)
+     Packet field 1: slab position =   40, slab bitmask = 0x0000000fff000000   (pipe)
+     Packet field 2: slab position =    8, slab bitmask = 0x00000000000000fc   (tc)
+     Packet field 2  tc translation table: ([Mapped Value Range]: tc/queue tc/queue ...)
+     [ 0 .. 15]: 0/0 0/1 0/2 0/3 1/0 1/1 1/2 1/3 2/0 2/1 2/2 2/3 3/0 3/1 3/2 3/3
+     [16 .. 31]: 0/0 0/1 0/2 0/3 1/0 1/1 1/2 1/3 2/0 2/1 2/2 2/3 3/0 3/1 3/2 3/3
+     [32 .. 47]: 0/0 0/1 0/2 0/3 1/0 1/1 1/2 1/3 2/0 2/1 2/2 2/3 3/0 3/1 3/2 3/3
+     [48 .. 63]: 0/0 0/1 0/2 0/3 1/0 1/1 1/2 1/3 2/0 2/1 2/2 2/3 3/0 3/1 3/2 3/3
+   Port:
+     Rate = 1250000000 bytes/second
+     MTU = 1514 bytes
+     Frame overhead = 24 bytes
+     Number of subports = 1
+     Number of pipes per subport = 4096
+     Packet queue size: TC0 = 64, TC1 = 64, TC2 = 64, TC3 = 64 packets
+     Number of pipe profiles = 1
+   Subport 0:
+     Rate = 120000000 bytes/second
+     Token bucket size = 1000000 bytes
+     Traffic class rate: TC0 = 120000000, TC1 = 120000000, TC2 = 120000000, TC3 = 120000000 bytes/second
+     TC period = 10 milliseconds
+   Pipe profile 0:
+     Rate = 305175 bytes/second
+     Token bucket size = 1000000 bytes
+     Traffic class rate: TC0 = 305175, TC1 = 305175, TC2 = 305175, TC3 = 305175 bytes/second
+     TC period = 40 milliseconds
+     TC0 WRR weights: Q0 = 1, Q1 = 1, Q2 = 1, Q3 = 1
+     TC1 WRR weights: Q0 = 1, Q1 = 1, Q2 = 1, Q3 = 1
+     TC2 WRR weights: Q0 = 1, Q1 = 1, Q2 = 1, Q3 = 1
+     TC3 WRR weights: Q0 = 1, Q1 = 1, Q2 = 1, Q3 = 1
+```
+
+The QoS Scheduler placement over the logical cpu cores can be displayed using
+below command.
+
+```
+    vpp# show dpdk interface hqos placement
+    Thread 5 (vpp_hqos-threads_0 at lcore 5):
+      TenGigabitEthernet2/0/0 queue 0
+    Thread 6 (vpp_hqos-threads_1 at lcore 6):
+      TenGigabitEthernet4/0/1 queue 0
+```
+
+
+### QoS Scheduler Binary APIs
+
+This section explans the available binary APIs for configuring QoS scheduler
+parameters in run-time.
+
+The following API can be used to set the pipe profile of a pipe that belongs
+to a given subport:
+
+```
+sw_interface_set_dpdk_hqos_pipe rx  | sw_if_index 
+    subport  pipe  profile 
+```
+
+The data structures used for set the pipe profile parameter are as follows;
+
+```
+  /** \\brief DPDK interface HQoS pipe profile set request
+    @param client_index - opaque cookie to identify the sender
+    @param context - sender context, to match reply w/ request
+    @param sw_if_index - the interface
+    @param subport - subport ID
+    @param pipe - pipe ID within its subport
+    @param profile - pipe profile ID
+  */
+  define sw_interface_set_dpdk_hqos_pipe {
+    u32 client_index;
+    u32 context;
+    u32 sw_if_index;
+    u32 subport;
+    u32 pipe;
+    u32 profile;
+  };
+
+  /** \\brief DPDK interface HQoS pipe profile set reply
+    @param context - sender context, to match reply w/ request
+    @param retval - request return code
+  */
+  define sw_interface_set_dpdk_hqos_pipe_reply {
+    u32 context;
+    i32 retval;
+  };
+```
+
+The following API can be used to set the subport level parameters, for
+example- token bucket rate (bytes per seconds), token bucket size (bytes),
+traffic class rate (bytes per seconds) and tokens update period.
+
+```
+sw_interface_set_dpdk_hqos_subport rx  | sw_if_index 
+    subport  [rate ] [bktsize ]
+    [tc0 ] [tc1 ] [tc2 ] [tc3 ] [period ]
+```
+
+The data structures used for set the subport level parameter are as follows;
+
+```
+  /** \\brief DPDK interface HQoS subport parameters set request
+    @param client_index - opaque cookie to identify the sender
+    @param context - sender context, to match reply w/ request
+    @param sw_if_index - the interface
+    @param subport - subport ID
+    @param tb_rate - subport token bucket rate (measured in bytes/second)
+    @param tb_size - subport token bucket size (measured in credits)
+    @param tc_rate - subport traffic class 0 .. 3 rates (measured in bytes/second)
+    @param tc_period - enforcement period for rates (measured in milliseconds)
+  */
+  define sw_interface_set_dpdk_hqos_subport {
+    u32 client_index;
+    u32 context;
+    u32 sw_if_index;
+    u32 subport;
+    u32 tb_rate;
+    u32 tb_size;
+    u32 tc_rate[4];
+    u32 tc_period;
+  };
+
+  /** \\brief DPDK interface HQoS subport parameters set reply
+    @param context - sender context, to match reply w/ request
+    @param retval - request return code
+  */
+  define sw_interface_set_dpdk_hqos_subport_reply {
+    u32 context;
+    i32 retval;
+  };
+```
+
+The following API can be used set the DSCP table entry. The DSCP table have
+64 entries to map the packet DSCP field onto traffic class and hqos input
+queue.
+
+```
+sw_interface_set_dpdk_hqos_tctbl rx  | sw_if_index  
+    entry  tc  queue 
+```
+
+The data structures used for setting DSCP table entries are given below.
+
+```
+  /** \\brief DPDK interface HQoS tctbl entry set request
+    @param client_index - opaque cookie to identify the sender
+    @param context - sender context, to match reply w/ request
+    @param sw_if_index - the interface
+    @param entry - entry index ID
+    @param tc - traffic class (0 .. 3)
+    @param queue - traffic class queue (0 .. 3)
+  */
+  define sw_interface_set_dpdk_hqos_tctbl {
+    u32 client_index;
+    u32 context;
+    u32 sw_if_index;
+    u32 entry;
+    u32 tc;
+    u32 queue;
+  };
+
+  /** \\brief DPDK interface HQoS tctbl entry set reply
+    @param context - sender context, to match reply w/ request
+    @param retval - request return code
+  */
+  define sw_interface_set_dpdk_hqos_tctbl_reply {
+    u32 context;
+    i32 retval;
+  };
+```
diff --git a/src/plugins/dpdk/init.c b/src/plugins/dpdk/init.c
new file mode 100755
index 00000000..e009ef3e
--- /dev/null
+++ b/src/plugins/dpdk/init.c
@@ -0,0 +1,2074 @@
+/*
+ * Copyright (c) 2015 Cisco and/or its affiliates.
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at:
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+#include 
+#include 
+#include 
+#include 
+#include 
+
+#include 
+#include 
+#include 
+#include 
+
+#include 
+#include 
+#include 
+#include 
+#include 
+#include 
+#include 
+
+#include 
+
+dpdk_main_t dpdk_main;
+
+#include 
+#include 
+
+/* define message IDs */
+#include 
+
+#define vl_typedefs		/* define message structures */
+#include 
+#undef vl_typedefs
+
+#define vl_endianfun		/* define message structures */
+#include 
+#undef vl_endianfun
+
+#define vl_print(handle, ...) vlib_cli_output (handle, __VA_ARGS__)
+
+/* Get the API version number. */
+#define vl_api_version(n,v) static u32 api_version=(v);
+#include 
+#undef vl_api_version
+
+/* Macro to finish up custom dump fns */
+#define FINISH                                  \
+    vec_add1 (s, 0);                            \
+    vl_print (handle, (char *)s);               \
+    vec_free (s);                               \
+    return handle;
+
+#include 
+
+static void
+  vl_api_sw_interface_set_dpdk_hqos_pipe_t_handler
+  (vl_api_sw_interface_set_dpdk_hqos_pipe_t * mp)
+{
+  vl_api_sw_interface_set_dpdk_hqos_pipe_reply_t *rmp;
+  int rv = 0;
+
+  dpdk_main_t *dm = &dpdk_main;
+  dpdk_device_t *xd;
+
+  u32 sw_if_index = ntohl (mp->sw_if_index);
+  u32 subport = ntohl (mp->subport);
+  u32 pipe = ntohl (mp->pipe);
+  u32 profile = ntohl (mp->profile);
+  vnet_hw_interface_t *hw;
+
+  VALIDATE_SW_IF_INDEX (mp);
+
+  /* hw_if & dpdk device */
+  hw = vnet_get_sup_hw_interface (dm->vnet_main, sw_if_index);
+
+  xd = vec_elt_at_index (dm->devices, hw->dev_instance);
+
+  rv = rte_sched_pipe_config (xd->hqos_ht->hqos, subport, pipe, profile);
+
+  BAD_SW_IF_INDEX_LABEL;
+
+  REPLY_MACRO (VL_API_SW_INTERFACE_SET_DPDK_HQOS_PIPE_REPLY);
+}
+
+static void *vl_api_sw_interface_set_dpdk_hqos_pipe_t_print
+  (vl_api_sw_interface_set_dpdk_hqos_pipe_t * mp, void *handle)
+{
+  u8 *s;
+
+  s = format (0, "SCRIPT: sw_interface_set_dpdk_hqos_pipe ");
+
+  s = format (s, "sw_if_index %u ", ntohl (mp->sw_if_index));
+
+  s = format (s, "subport %u  pipe %u  profile %u ",
+	      ntohl (mp->subport), ntohl (mp->pipe), ntohl (mp->profile));
+
+  FINISH;
+}
+
+static void
+  vl_api_sw_interface_set_dpdk_hqos_subport_t_handler
+  (vl_api_sw_interface_set_dpdk_hqos_subport_t * mp)
+{
+  vl_api_sw_interface_set_dpdk_hqos_subport_reply_t *rmp;
+  int rv = 0;
+
+  dpdk_main_t *dm = &dpdk_main;
+  dpdk_device_t *xd;
+  struct rte_sched_subport_params p;
+
+  u32 sw_if_index = ntohl (mp->sw_if_index);
+  u32 subport = ntohl (mp->subport);
+  p.tb_rate = ntohl (mp->tb_rate);
+  p.tb_size = ntohl (mp->tb_size);
+  p.tc_rate[0] = ntohl (mp->tc_rate[0]);
+  p.tc_rate[1] = ntohl (mp->tc_rate[1]);
+  p.tc_rate[2] = ntohl (mp->tc_rate[2]);
+  p.tc_rate[3] = ntohl (mp->tc_rate[3]);
+  p.tc_period = ntohl (mp->tc_period);
+
+  vnet_hw_interface_t *hw;
+
+  VALIDATE_SW_IF_INDEX (mp);
+
+  /* hw_if & dpdk device */
+  hw = vnet_get_sup_hw_interface (dm->vnet_main, sw_if_index);
+
+  xd = vec_elt_at_index (dm->devices, hw->dev_instance);
+
+  rv = rte_sched_subport_config (xd->hqos_ht->hqos, subport, &p);
+
+  BAD_SW_IF_INDEX_LABEL;
+
+  REPLY_MACRO (VL_API_SW_INTERFACE_SET_DPDK_HQOS_SUBPORT_REPLY);
+}
+
+static void *vl_api_sw_interface_set_dpdk_hqos_subport_t_print
+  (vl_api_sw_interface_set_dpdk_hqos_subport_t * mp, void *handle)
+{
+  u8 *s;
+
+  s = format (0, "SCRIPT: sw_interface_set_dpdk_hqos_subport ");
+
+  s = format (s, "sw_if_index %u ", ntohl (mp->sw_if_index));
+
+  s =
+    format (s,
+	    "subport %u  rate %u  bkt_size %u  tc0 %u tc1 %u tc2 %u tc3 %u period %u",
+	    ntohl (mp->subport), ntohl (mp->tb_rate), ntohl (mp->tb_size),
+	    ntohl (mp->tc_rate[0]), ntohl (mp->tc_rate[1]),
+	    ntohl (mp->tc_rate[2]), ntohl (mp->tc_rate[3]),
+	    ntohl (mp->tc_period));
+
+  FINISH;
+}
+
+static void
+  vl_api_sw_interface_set_dpdk_hqos_tctbl_t_handler
+  (vl_api_sw_interface_set_dpdk_hqos_tctbl_t * mp)
+{
+  vl_api_sw_interface_set_dpdk_hqos_tctbl_reply_t *rmp;
+  int rv = 0;
+
+  dpdk_main_t *dm = &dpdk_main;
+  vlib_thread_main_t *tm = vlib_get_thread_main ();
+  dpdk_device_t *xd;
+
+  u32 sw_if_index = ntohl (mp->sw_if_index);
+  u32 entry = ntohl (mp->entry);
+  u32 tc = ntohl (mp->tc);
+  u32 queue = ntohl (mp->queue);
+  u32 val, i;
+
+  vnet_hw_interface_t *hw;
+
+  VALIDATE_SW_IF_INDEX (mp);
+
+  /* hw_if & dpdk device */
+  hw = vnet_get_sup_hw_interface (dm->vnet_main, sw_if_index);
+
+  xd = vec_elt_at_index (dm->devices, hw->dev_instance);
+
+  if (tc >= RTE_SCHED_TRAFFIC_CLASSES_PER_PIPE)
+    {
+      clib_warning ("invalid traffic class !!");
+      rv = VNET_API_ERROR_INVALID_VALUE;
+      goto done;
+    }
+  if (queue >= RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS)
+    {
+      clib_warning ("invalid queue !!");
+      rv = VNET_API_ERROR_INVALID_VALUE;
+      goto done;
+    }
+
+  /* Detect the set of worker threads */
+  uword *p = hash_get_mem (tm->thread_registrations_by_name, "workers");
+
+  if (p == 0)
+    {
+      clib_warning ("worker thread registration AWOL !!");
+      rv = VNET_API_ERROR_INVALID_VALUE_2;
+      goto done;
+    }
+
+  vlib_thread_registration_t *tr = (vlib_thread_registration_t *) p[0];
+  int worker_thread_first = tr->first_index;
+  int worker_thread_count = tr->count;
+
+  val = tc * RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS + queue;
+  for (i = 0; i < worker_thread_count; i++)
+    xd->hqos_wt[worker_thread_first + i].hqos_tc_table[entry] = val;
+
+  BAD_SW_IF_INDEX_LABEL;
+done:
+
+  REPLY_MACRO (VL_API_SW_INTERFACE_SET_DPDK_HQOS_TCTBL_REPLY);
+}
+
+static void *vl_api_sw_interface_set_dpdk_hqos_tctbl_t_print
+  (vl_api_sw_interface_set_dpdk_hqos_tctbl_t * mp, void *handle)
+{
+  u8 *s;
+
+  s = format (0, "SCRIPT: sw_interface_set_dpdk_hqos_tctbl ");
+
+  s = format (s, "sw_if_index %u ", ntohl (mp->sw_if_index));
+
+  s = format (s, "entry %u  tc %u  queue %u",
+	      ntohl (mp->entry), ntohl (mp->tc), ntohl (mp->queue));
+
+  FINISH;
+}
+
+#define foreach_dpdk_plugin_api_msg                                       \
+_(SW_INTERFACE_SET_DPDK_HQOS_PIPE, sw_interface_set_dpdk_hqos_pipe)       \
+_(SW_INTERFACE_SET_DPDK_HQOS_SUBPORT, sw_interface_set_dpdk_hqos_subport) \
+_(SW_INTERFACE_SET_DPDK_HQOS_TCTBL, sw_interface_set_dpdk_hqos_tctbl)
+
+/* Set up the API message handling tables */
+static clib_error_t *
+dpdk_plugin_api_hookup (vlib_main_t * vm)
+{
+  dpdk_main_t *dm __attribute__ ((unused)) = &dpdk_main;
+#define _(N,n)                                                  \
+    vl_msg_api_set_handlers((VL_API_##N + dm->msg_id_base),     \
+                           #n,          \
+                           vl_api_##n##_t_handler,              \
+                           vl_noop_handler,                     \
+                           vl_api_##n##_t_endian,               \
+                           vl_api_##n##_t_print,                \
+                           sizeof(vl_api_##n##_t), 1);
+  foreach_dpdk_plugin_api_msg;
+#undef _
+  return 0;
+}
+
+#define vl_msg_name_crc_list
+#include 
+#undef vl_msg_name_crc_list
+
+static void
+setup_message_id_table (dpdk_main_t * dm, api_main_t * am)
+{
+#define _(id,n,crc) \
+  vl_msg_api_add_msg_name_crc (am, #n "_" #crc, id + dm->msg_id_base);
+  foreach_vl_msg_name_crc_dpdk;
+#undef _
+}
+
+//  TODO
+/*
+static void plugin_custom_dump_configure (dpdk_main_t * dm)
+{
+#define _(n,f) dm->api_main->msg_print_handlers \
+  [VL_API_##n + dm->msg_id_base]                \
+    = (void *) vl_api_##f##_t_print;
+  foreach_dpdk_plugin_api_msg;
+#undef _
+}
+*/
+/* force linker to link functions used by vlib and declared weak */
+void *vlib_weakly_linked_functions[] = {
+  &rte_pktmbuf_init,
+  &rte_pktmbuf_pool_init,
+};
+
+#define LINK_STATE_ELOGS	0
+
+#define DEFAULT_HUGE_DIR "/run/vpp/hugepages"
+#define VPP_RUN_DIR "/run/vpp"
+
+/* Port configuration, mildly modified Intel app values */
+
+static struct rte_eth_conf port_conf_template = {
+  .rxmode = {
+	     .split_hdr_size = 0,
+	     .header_split = 0,		/**< Header Split disabled */
+	     .hw_ip_checksum = 0,	/**< IP checksum offload disabled */
+	     .hw_vlan_filter = 0,	/**< VLAN filtering disabled */
+	     .hw_strip_crc = 0,		/**< CRC stripped by hardware */
+	     },
+  .txmode = {
+	     .mq_mode = ETH_MQ_TX_NONE,
+	     },
+};
+
+clib_error_t *
+dpdk_port_setup (dpdk_main_t * dm, dpdk_device_t * xd)
+{
+  int rv;
+  int j;
+
+  ASSERT (os_get_cpu_number () == 0);
+
+  if (xd->flags & DPDK_DEVICE_FLAG_ADMIN_UP)
+    {
+      vnet_hw_interface_set_flags (dm->vnet_main, xd->vlib_hw_if_index, 0);
+      rte_eth_dev_stop (xd->device_index);
+    }
+
+  rv = rte_eth_dev_configure (xd->device_index, xd->rx_q_used,
+			      xd->tx_q_used, &xd->port_conf);
+
+  if (rv < 0)
+    return clib_error_return (0, "rte_eth_dev_configure[%d]: err %d",
+			      xd->device_index, rv);
+
+  /* Set up one TX-queue per worker thread */
+  for (j = 0; j < xd->tx_q_used; j++)
+    {
+      rv = rte_eth_tx_queue_setup (xd->device_index, j, xd->nb_tx_desc,
+				   xd->cpu_socket, &xd->tx_conf);
+
+      /* retry with any other CPU socket */
+      if (rv < 0)
+	rv = rte_eth_tx_queue_setup (xd->device_index, j, xd->nb_tx_desc,
+				     SOCKET_ID_ANY, &xd->tx_conf);
+      if (rv < 0)
+	break;
+    }
+
+  if (rv < 0)
+    return clib_error_return (0, "rte_eth_tx_queue_setup[%d]: err %d",
+			      xd->device_index, rv);
+
+  for (j = 0; j < xd->rx_q_used; j++)
+    {
+
+      rv = rte_eth_rx_queue_setup (xd->device_index, j, xd->nb_rx_desc,
+				   xd->cpu_socket, 0,
+				   dm->
+				   pktmbuf_pools[xd->cpu_socket_id_by_queue
+						 [j]]);
+
+      /* retry with any other CPU socket */
+      if (rv < 0)
+	rv = rte_eth_rx_queue_setup (xd->device_index, j, xd->nb_rx_desc,
+				     SOCKET_ID_ANY, 0,
+				     dm->
+				     pktmbuf_pools[xd->cpu_socket_id_by_queue
+						   [j]]);
+      if (rv < 0)
+	return clib_error_return (0, "rte_eth_rx_queue_setup[%d]: err %d",
+				  xd->device_index, rv);
+    }
+
+  if (xd->flags & DPDK_DEVICE_FLAG_ADMIN_UP)
+    {
+      int rv;
+      rv = rte_eth_dev_start (xd->device_index);
+      if (!rv && xd->default_mac_address)
+	rv = rte_eth_dev_default_mac_addr_set (xd->device_index,
+					       (struct ether_addr *)
+					       xd->default_mac_address);
+      if (rv < 0)
+	clib_warning ("rte_eth_dev_start %d returned %d",
+		      xd->device_index, rv);
+    }
+  return 0;
+}
+
+static u32
+dpdk_flag_change (vnet_main_t * vnm, vnet_hw_interface_t * hi, u32 flags)
+{
+  dpdk_main_t *dm = &dpdk_main;
+  dpdk_device_t *xd = vec_elt_at_index (dm->devices, hi->dev_instance);
+  u32 old = 0;
+
+  if (ETHERNET_INTERFACE_FLAG_CONFIG_PROMISC (flags))
+    {
+      old = (xd->flags & DPDK_DEVICE_FLAG_PROMISC) != 0;
+
+      if (flags & ETHERNET_INTERFACE_FLAG_ACCEPT_ALL)
+	xd->flags |= DPDK_DEVICE_FLAG_PROMISC;
+      else
+	xd->flags &= ~DPDK_DEVICE_FLAG_PROMISC;
+
+      if (xd->flags & DPDK_DEVICE_FLAG_ADMIN_UP)
+	{
+	  if (xd->flags & DPDK_DEVICE_FLAG_PROMISC)
+	    rte_eth_promiscuous_enable (xd->device_index);
+	  else
+	    rte_eth_promiscuous_disable (xd->device_index);
+	}
+    }
+  else if (ETHERNET_INTERFACE_FLAG_CONFIG_MTU (flags))
+    {
+      /*
+       * DAW-FIXME: The Cisco VIC firmware does not provide an api for a
+       *            driver to dynamically change the mtu.  If/when the
+       *            VIC firmware gets fixed, then this should be removed.
+       */
+      if (xd->pmd == VNET_DPDK_PMD_ENIC)
+	{
+	  struct rte_eth_dev_info dev_info;
+
+	  /*
+	   * Restore mtu to what has been set by CIMC in the firmware cfg.
+	   */
+	  rte_eth_dev_info_get (xd->device_index, &dev_info);
+	  hi->max_packet_bytes = dev_info.max_rx_pktlen;
+
+	  vlib_cli_output (vlib_get_main (),
+			   "Cisco VIC mtu can only be changed "
+			   "using CIMC then rebooting the server!");
+	}
+      else
+	{
+	  int rv;
+
+	  xd->port_conf.rxmode.max_rx_pkt_len = hi->max_packet_bytes;
+
+	  if (xd->flags & DPDK_DEVICE_FLAG_ADMIN_UP)
+	    rte_eth_dev_stop (xd->device_index);
+
+	  rv = rte_eth_dev_configure
+	    (xd->device_index, xd->rx_q_used, xd->tx_q_used, &xd->port_conf);
+
+	  if (rv < 0)
+	    vlib_cli_output (vlib_get_main (),
+			     "rte_eth_dev_configure[%d]: err %d",
+			     xd->device_index, rv);
+
+	  rte_eth_dev_set_mtu (xd->device_index, hi->max_packet_bytes);
+
+	  if (xd->flags & DPDK_DEVICE_FLAG_ADMIN_UP)
+	    {
+	      int rv = rte_eth_dev_start (xd->device_index);
+	      if (!rv && xd->default_mac_address)
+		rv = rte_eth_dev_default_mac_addr_set (xd->device_index,
+						       (struct ether_addr *)
+						       xd->default_mac_address);
+	      if (rv < 0)
+		clib_warning ("rte_eth_dev_start %d returned %d",
+			      xd->device_index, rv);
+	    }
+	}
+    }
+  return old;
+}
+
+void
+dpdk_device_lock_init (dpdk_device_t * xd)
+{
+  int q;
+  vec_validate (xd->lockp, xd->tx_q_used - 1);
+  for (q = 0; q < xd->tx_q_used; q++)
+    {
+      xd->lockp[q] = clib_mem_alloc_aligned (CLIB_CACHE_LINE_BYTES,
+					     CLIB_CACHE_LINE_BYTES);
+      memset ((void *) xd->lockp[q], 0, CLIB_CACHE_LINE_BYTES);
+    }
+}
+
+void
+dpdk_device_lock_free (dpdk_device_t * xd)
+{
+  int q;
+
+  for (q = 0; q < vec_len (xd->lockp); q++)
+    clib_mem_free ((void *) xd->lockp[q]);
+  vec_free (xd->lockp);
+  xd->lockp = 0;
+}
+
+static clib_error_t *
+dpdk_lib_init (dpdk_main_t * dm)
+{
+  u32 nports;
+  u32 nb_desc = 0;
+  int i;
+  clib_error_t *error;
+  vlib_main_t *vm = vlib_get_main ();
+  vlib_thread_main_t *tm = vlib_get_thread_main ();
+  vnet_sw_interface_t *sw;
+  vnet_hw_interface_t *hi;
+  dpdk_device_t *xd;
+  vlib_pci_addr_t last_pci_addr;
+  u32 last_pci_addr_port = 0;
+  vlib_thread_registration_t *tr, *tr_hqos;
+  uword *p, *p_hqos;
+
+  u32 next_cpu = 0, next_hqos_cpu = 0;
+  u8 af_packet_port_id = 0;
+  last_pci_addr.as_u32 = ~0;
+
+  dm->input_cpu_first_index = 0;
+  dm->input_cpu_count = 1;
+
+  /* find out which cpus will be used for input */
+  p = hash_get_mem (tm->thread_registrations_by_name, "workers");
+  tr = p ? (vlib_thread_registration_t *) p[0] : 0;
+
+  if (tr && tr->count > 0)
+    {
+      dm->input_cpu_first_index = tr->first_index;
+      dm->input_cpu_count = tr->count;
+    }
+
+  vec_validate_aligned (dm->devices_by_cpu, tm->n_vlib_mains - 1,
+			CLIB_CACHE_LINE_BYTES);
+
+  dm->hqos_cpu_first_index = 0;
+  dm->hqos_cpu_count = 0;
+
+  /* find out which cpus will be used for I/O TX */
+  p_hqos = hash_get_mem (tm->thread_registrations_by_name, "hqos-threads");
+  tr_hqos = p_hqos ? (vlib_thread_registration_t *) p_hqos[0] : 0;
+
+  if (tr_hqos && tr_hqos->count > 0)
+    {
+      dm->hqos_cpu_first_index = tr_hqos->first_index;
+      dm->hqos_cpu_count = tr_hqos->count;
+    }
+
+  vec_validate_aligned (dm->devices_by_hqos_cpu, tm->n_vlib_mains - 1,
+			CLIB_CACHE_LINE_BYTES);
+
+  nports = rte_eth_dev_count ();
+  if (nports < 1)
+    {
+      clib_warning ("DPDK drivers found no ports...");
+    }
+
+  if (CLIB_DEBUG > 0)
+    clib_warning ("DPDK drivers found %d ports...", nports);
+
+  /*
+   * All buffers are all allocated from the same rte_mempool.
+   * Thus they all have the same number of data bytes.
+   */
+  dm->vlib_buffer_free_list_index =
+    vlib_buffer_get_or_create_free_list (vm,
+					 VLIB_BUFFER_DEFAULT_FREE_LIST_BYTES,
+					 "dpdk rx");
+
+  if (dm->conf->enable_tcp_udp_checksum)
+    dm->buffer_flags_template &= ~(IP_BUFFER_L4_CHECKSUM_CORRECT
+				   | IP_BUFFER_L4_CHECKSUM_COMPUTED);
+
+  for (i = 0; i < nports; i++)
+    {
+      u8 addr[6];
+      u8 vlan_strip = 0;
+      int j;
+      struct rte_eth_dev_info dev_info;
+      clib_error_t *rv;
+      struct rte_eth_link l;
+      dpdk_device_config_t *devconf = 0;
+      vlib_pci_addr_t pci_addr;
+      uword *p = 0;
+
+      rte_eth_dev_info_get (i, &dev_info);
+      if (dev_info.pci_dev)	/* bonded interface has no pci info */
+	{
+	  pci_addr.domain = dev_info.pci_dev->addr.domain;
+	  pci_addr.bus = dev_info.pci_dev->addr.bus;
+	  pci_addr.slot = dev_info.pci_dev->addr.devid;
+	  pci_addr.function = dev_info.pci_dev->addr.function;
+	  p =
+	    hash_get (dm->conf->device_config_index_by_pci_addr,
+		      pci_addr.as_u32);
+	}
+
+      if (p)
+	devconf = pool_elt_at_index (dm->conf->dev_confs, p[0]);
+      else
+	devconf = &dm->conf->default_devconf;
+
+      /* Create vnet interface */
+      vec_add2_aligned (dm->devices, xd, 1, CLIB_CACHE_LINE_BYTES);
+      xd->nb_rx_desc = DPDK_NB_RX_DESC_DEFAULT;
+      xd->nb_tx_desc = DPDK_NB_TX_DESC_DEFAULT;
+      xd->cpu_socket = (i8) rte_eth_dev_socket_id (i);
+
+      /* Handle interface naming for devices with multiple ports sharing same PCI ID */
+      if (dev_info.pci_dev)
+	{
+	  struct rte_eth_dev_info di = { 0 };
+	  rte_eth_dev_info_get (i + 1, &di);
+	  if (di.pci_dev && pci_addr.as_u32 != last_pci_addr.as_u32 &&
+	      memcmp (&dev_info.pci_dev->addr, &di.pci_dev->addr,
+		      sizeof (struct rte_pci_addr)) == 0)
+	    {
+	      xd->interface_name_suffix = format (0, "0");
+	      last_pci_addr.as_u32 = pci_addr.as_u32;
+	      last_pci_addr_port = i;
+	    }
+	  else if (pci_addr.as_u32 == last_pci_addr.as_u32)
+	    {
+	      xd->interface_name_suffix =
+		format (0, "%u", i - last_pci_addr_port);
+	    }
+	  else
+	    {
+	      last_pci_addr.as_u32 = ~0;
+	    }
+	}
+      else
+	last_pci_addr.as_u32 = ~0;
+
+      clib_memcpy (&xd->tx_conf, &dev_info.default_txconf,
+		   sizeof (struct rte_eth_txconf));
+      if (dm->conf->no_multi_seg)
+	{
+	  xd->tx_conf.txq_flags |= ETH_TXQ_FLAGS_NOMULTSEGS;
+	  port_conf_template.rxmode.jumbo_frame = 0;
+	}
+      else
+	{
+	  xd->tx_conf.txq_flags &= ~ETH_TXQ_FLAGS_NOMULTSEGS;
+	  port_conf_template.rxmode.jumbo_frame = 1;
+	  xd->flags |= DPDK_DEVICE_FLAG_MAYBE_MULTISEG;
+	}
+
+      clib_memcpy (&xd->port_conf, &port_conf_template,
+		   sizeof (struct rte_eth_conf));
+
+      xd->tx_q_used = clib_min (dev_info.max_tx_queues, tm->n_vlib_mains);
+
+      if (devconf->num_tx_queues > 0
+	  && devconf->num_tx_queues < xd->tx_q_used)
+	xd->tx_q_used = clib_min (xd->tx_q_used, devconf->num_tx_queues);
+
+      if (devconf->num_rx_queues > 1 && dm->use_rss == 0)
+	{
+	  dm->use_rss = 1;
+	}
+
+      if (devconf->num_rx_queues > 1
+	  && dev_info.max_rx_queues >= devconf->num_rx_queues)
+	{
+	  xd->rx_q_used = devconf->num_rx_queues;
+	  xd->port_conf.rxmode.mq_mode = ETH_MQ_RX_RSS;
+	  if (devconf->rss_fn == 0)
+	    xd->port_conf.rx_adv_conf.rss_conf.rss_hf =
+	      ETH_RSS_IP | ETH_RSS_UDP | ETH_RSS_TCP;
+	  else
+	    xd->port_conf.rx_adv_conf.rss_conf.rss_hf = devconf->rss_fn;
+	}
+      else
+	xd->rx_q_used = 1;
+
+      xd->flags |= DPDK_DEVICE_FLAG_PMD;
+
+      /* workaround for drivers not setting driver_name */
+      if ((!dev_info.driver_name) && (dev_info.pci_dev))
+	dev_info.driver_name = dev_info.pci_dev->driver->driver.name;
+
+      ASSERT (dev_info.driver_name);
+
+      if (!xd->pmd)
+	{
+
+
+#define _(s,f) else if (dev_info.driver_name &&                 \
+                        !strcmp(dev_info.driver_name, s))       \
+                 xd->pmd = VNET_DPDK_PMD_##f;
+	  if (0)
+	    ;
+	  foreach_dpdk_pmd
+#undef _
+	    else
+	    xd->pmd = VNET_DPDK_PMD_UNKNOWN;
+
+	  xd->port_type = VNET_DPDK_PORT_TYPE_UNKNOWN;
+	  xd->nb_rx_desc = DPDK_NB_RX_DESC_DEFAULT;
+	  xd->nb_tx_desc = DPDK_NB_TX_DESC_DEFAULT;
+
+	  switch (xd->pmd)
+	    {
+	      /* 1G adapters */
+	    case VNET_DPDK_PMD_E1000EM:
+	    case VNET_DPDK_PMD_IGB:
+	    case VNET_DPDK_PMD_IGBVF:
+	      xd->port_type = VNET_DPDK_PORT_TYPE_ETH_1G;
+	      break;
+
+	      /* 10G adapters */
+	    case VNET_DPDK_PMD_IXGBE:
+	    case VNET_DPDK_PMD_IXGBEVF:
+	    case VNET_DPDK_PMD_THUNDERX:
+	      xd->port_type = VNET_DPDK_PORT_TYPE_ETH_10G;
+	      break;
+	    case VNET_DPDK_PMD_DPAA2:
+	      xd->port_type = VNET_DPDK_PORT_TYPE_ETH_10G;
+	      break;
+
+	      /* Cisco VIC */
+	    case VNET_DPDK_PMD_ENIC:
+	      rte_eth_link_get_nowait (i, &l);
+	      xd->flags |= DPDK_DEVICE_FLAG_PMD_SUPPORTS_PTYPE;
+	      if (l.link_speed == 40000)
+		xd->port_type = VNET_DPDK_PORT_TYPE_ETH_40G;
+	      else
+		xd->port_type = VNET_DPDK_PORT_TYPE_ETH_10G;
+	      break;
+
+	      /* Intel Fortville */
+	    case VNET_DPDK_PMD_I40E:
+	    case VNET_DPDK_PMD_I40EVF:
+	      xd->flags |= DPDK_DEVICE_FLAG_PMD_SUPPORTS_PTYPE;
+	      xd->port_type = VNET_DPDK_PORT_TYPE_ETH_40G;
+
+	      switch (dev_info.pci_dev->id.device_id)
+		{
+		case I40E_DEV_ID_10G_BASE_T:
+		case I40E_DEV_ID_SFP_XL710:
+		  xd->port_type = VNET_DPDK_PORT_TYPE_ETH_10G;
+		  break;
+		case I40E_DEV_ID_QSFP_A:
+		case I40E_DEV_ID_QSFP_B:
+		case I40E_DEV_ID_QSFP_C:
+		  xd->port_type = VNET_DPDK_PORT_TYPE_ETH_40G;
+		  break;
+		case I40E_DEV_ID_VF:
+		  rte_eth_link_get_nowait (i, &l);
+		  xd->port_type = l.link_speed == 10000 ?
+		    VNET_DPDK_PORT_TYPE_ETH_10G : VNET_DPDK_PORT_TYPE_ETH_40G;
+		  break;
+		default:
+		  xd->port_type = VNET_DPDK_PORT_TYPE_UNKNOWN;
+		}
+	      break;
+
+	    case VNET_DPDK_PMD_CXGBE:
+	      switch (dev_info.pci_dev->id.device_id)
+		{
+		case 0x540d:	/* T580-CR */
+		case 0x5410:	/* T580-LP-cr */
+		  xd->port_type = VNET_DPDK_PORT_TYPE_ETH_40G;
+		  break;
+		case 0x5403:	/* T540-CR */
+		  xd->port_type = VNET_DPDK_PORT_TYPE_ETH_10G;
+		  break;
+		default:
+		  xd->port_type = VNET_DPDK_PORT_TYPE_UNKNOWN;
+		}
+	      break;
+
+	    case VNET_DPDK_PMD_MLX5:
+	      {
+		char *pn_100g[] = { "MCX415A-CCAT", "MCX416A-CCAT", 0 };
+		char *pn_40g[] = { "MCX413A-BCAT", "MCX414A-BCAT",
+		  "MCX415A-BCAT", "MCX416A-BCAT", "MCX4131A-BCAT", 0
+		};
+		char *pn_10g[] = { "MCX4111A-XCAT", "MCX4121A-XCAT", 0 };
+
+		vlib_pci_device_t *pd = vlib_get_pci_device (&pci_addr);
+		u8 *pn = 0;
+		char **c;
+		int found = 0;
+		pn = format (0, "%U%c",
+			     format_vlib_pci_vpd, pd->vpd_r, "PN", 0);
+
+		if (!pn)
+		  break;
+
+		c = pn_100g;
+		while (!found && c[0])
+		  {
+		    if (strncmp ((char *) pn, c[0], strlen (c[0])) == 0)
+		      {
+			xd->port_type = VNET_DPDK_PORT_TYPE_ETH_100G;
+			break;
+		      }
+		    c++;
+		  }
+
+		c = pn_40g;
+		while (!found && c[0])
+		  {
+		    if (strncmp ((char *) pn, c[0], strlen (c[0])) == 0)
+		      {
+			xd->port_type = VNET_DPDK_PORT_TYPE_ETH_40G;
+			break;
+		      }
+		    c++;
+		  }
+
+		c = pn_10g;
+		while (!found && c[0])
+		  {
+		    if (strncmp ((char *) pn, c[0], strlen (c[0])) == 0)
+		      {
+			xd->port_type = VNET_DPDK_PORT_TYPE_ETH_10G;
+			break;
+		      }
+		    c++;
+		  }
+
+		vec_free (pn);
+	      }
+
+	      break;
+	      /* Intel Red Rock Canyon */
+	    case VNET_DPDK_PMD_FM10K:
+	      xd->port_type = VNET_DPDK_PORT_TYPE_ETH_SWITCH;
+	      break;
+
+	      /* virtio */
+	    case VNET_DPDK_PMD_VIRTIO:
+	      xd->port_type = VNET_DPDK_PORT_TYPE_ETH_1G;
+	      xd->nb_rx_desc = DPDK_NB_RX_DESC_VIRTIO;
+	      xd->nb_tx_desc = DPDK_NB_TX_DESC_VIRTIO;
+	      break;
+
+	      /* vmxnet3 */
+	    case VNET_DPDK_PMD_VMXNET3:
+	      xd->port_type = VNET_DPDK_PORT_TYPE_ETH_1G;
+	      xd->tx_conf.txq_flags |= ETH_TXQ_FLAGS_NOMULTSEGS;
+	      break;
+
+	    case VNET_DPDK_PMD_AF_PACKET:
+	      xd->port_type = VNET_DPDK_PORT_TYPE_AF_PACKET;
+	      xd->af_packet_port_id = af_packet_port_id++;
+	      break;
+
+	    case VNET_DPDK_PMD_BOND:
+	      xd->flags |= DPDK_DEVICE_FLAG_PMD_SUPPORTS_PTYPE;
+	      xd->port_type = VNET_DPDK_PORT_TYPE_ETH_BOND;
+	      break;
+
+	    default:
+	      xd->port_type = VNET_DPDK_PORT_TYPE_UNKNOWN;
+	    }
+
+	  if (devconf->num_rx_desc)
+	    xd->nb_rx_desc = devconf->num_rx_desc;
+
+	  if (devconf->num_tx_desc)
+	    xd->nb_tx_desc = devconf->num_tx_desc;
+	}
+
+      /*
+       * Ensure default mtu is not > the mtu read from the hardware.
+       * Otherwise rte_eth_dev_configure() will fail and the port will
+       * not be available.
+       */
+      if (ETHERNET_MAX_PACKET_BYTES > dev_info.max_rx_pktlen)
+	{
+	  /*
+	   * This device does not support the platforms's max frame
+	   * size. Use it's advertised mru instead.
+	   */
+	  xd->port_conf.rxmode.max_rx_pkt_len = dev_info.max_rx_pktlen;
+	}
+      else
+	{
+	  xd->port_conf.rxmode.max_rx_pkt_len = ETHERNET_MAX_PACKET_BYTES;
+
+	  /*
+	   * Some platforms do not account for Ethernet FCS (4 bytes) in
+	   * MTU calculations. To interop with them increase mru but only
+	   * if the device's settings can support it.
+	   */
+	  if ((dev_info.max_rx_pktlen >= (ETHERNET_MAX_PACKET_BYTES + 4)) &&
+	      xd->port_conf.rxmode.hw_strip_crc)
+	    {
+	      /*
+	       * Allow additional 4 bytes (for Ethernet FCS). These bytes are
+	       * stripped by h/w and so will not consume any buffer memory.
+	       */
+	      xd->port_conf.rxmode.max_rx_pkt_len += 4;
+	    }
+	}
+
+      if (xd->pmd == VNET_DPDK_PMD_AF_PACKET)
+	{
+	  f64 now = vlib_time_now (vm);
+	  u32 rnd;
+	  rnd = (u32) (now * 1e6);
+	  rnd = random_u32 (&rnd);
+	  clib_memcpy (addr + 2, &rnd, sizeof (rnd));
+	  addr[0] = 2;
+	  addr[1] = 0xfe;
+	}
+      else
+	rte_eth_macaddr_get (i, (struct ether_addr *) addr);
+
+      if (xd->tx_q_used < tm->n_vlib_mains)
+	dpdk_device_lock_init (xd);
+
+      xd->device_index = xd - dm->devices;
+      ASSERT (i == xd->device_index);
+      xd->per_interface_next_index = ~0;
+
+      /* assign interface to input thread */
+      dpdk_device_and_queue_t *dq;
+      int q;
+
+      if (devconf->workers)
+	{
+	  int i;
+	  q = 0;
+	  /* *INDENT-OFF* */
+	  clib_bitmap_foreach (i, devconf->workers, ({
+	    int cpu = dm->input_cpu_first_index + i;
+	    unsigned lcore = vlib_worker_threads[cpu].lcore_id;
+	    vec_validate(xd->cpu_socket_id_by_queue, q);
+	    xd->cpu_socket_id_by_queue[q] = rte_lcore_to_socket_id(lcore);
+	    vec_add2(dm->devices_by_cpu[cpu], dq, 1);
+	    dq->device = xd->device_index;
+	    dq->queue_id = q++;
+	  }));
+	  /* *INDENT-ON* */
+	}
+      else
+	for (q = 0; q < xd->rx_q_used; q++)
+	  {
+	    int cpu = dm->input_cpu_first_index + next_cpu;
+	    unsigned lcore = vlib_worker_threads[cpu].lcore_id;
+
+	    /*
+	     * numa node for worker thread handling this queue
+	     * needed for taking buffers from the right mempool
+	     */
+	    vec_validate (xd->cpu_socket_id_by_queue, q);
+	    xd->cpu_socket_id_by_queue[q] = rte_lcore_to_socket_id (lcore);
+
+	    /*
+	     * construct vector of (device,queue) pairs for each worker thread
+	     */
+	    vec_add2 (dm->devices_by_cpu[cpu], dq, 1);
+	    dq->device = xd->device_index;
+	    dq->queue_id = q;
+
+	    next_cpu++;
+	    if (next_cpu == dm->input_cpu_count)
+	      next_cpu = 0;
+	  }
+
+
+      if (devconf->hqos_enabled)
+	{
+	  xd->flags |= DPDK_DEVICE_FLAG_HQOS;
+
+	  if (devconf->hqos.hqos_thread_valid)
+	    {
+	      int cpu = dm->hqos_cpu_first_index + devconf->hqos.hqos_thread;
+
+	      if (devconf->hqos.hqos_thread >= dm->hqos_cpu_count)
+		return clib_error_return (0, "invalid HQoS thread index");
+
+	      vec_add2 (dm->devices_by_hqos_cpu[cpu], dq, 1);
+	      dq->device = xd->device_index;
+	      dq->queue_id = 0;
+	    }
+	  else
+	    {
+	      int cpu = dm->hqos_cpu_first_index + next_hqos_cpu;
+
+	      if (dm->hqos_cpu_count == 0)
+		return clib_error_return (0, "no HQoS threads available");
+
+	      vec_add2 (dm->devices_by_hqos_cpu[cpu], dq, 1);
+	      dq->device = xd->device_index;
+	      dq->queue_id = 0;
+
+	      next_hqos_cpu++;
+	      if (next_hqos_cpu == dm->hqos_cpu_count)
+		next_hqos_cpu = 0;
+
+	      devconf->hqos.hqos_thread_valid = 1;
+	      devconf->hqos.hqos_thread = cpu;
+	    }
+	}
+
+      vec_validate_aligned (xd->tx_vectors, tm->n_vlib_mains,
+			    CLIB_CACHE_LINE_BYTES);
+      for (j = 0; j < tm->n_vlib_mains; j++)
+	{
+	  vec_validate_ha (xd->tx_vectors[j], xd->nb_tx_desc,
+			   sizeof (tx_ring_hdr_t), CLIB_CACHE_LINE_BYTES);
+	  vec_reset_length (xd->tx_vectors[j]);
+	}
+
+      vec_validate_aligned (xd->rx_vectors, xd->rx_q_used,
+			    CLIB_CACHE_LINE_BYTES);
+      for (j = 0; j < xd->rx_q_used; j++)
+	{
+	  vec_validate_aligned (xd->rx_vectors[j], VLIB_FRAME_SIZE - 1,
+				CLIB_CACHE_LINE_BYTES);
+	  vec_reset_length (xd->rx_vectors[j]);
+	}
+
+      vec_validate_aligned (xd->d_trace_buffers, tm->n_vlib_mains,
+			    CLIB_CACHE_LINE_BYTES);
+
+      rv = dpdk_port_setup (dm, xd);
+
+      if (rv)
+	return rv;
+
+      if (devconf->hqos_enabled)
+	{
+	  rv = dpdk_port_setup_hqos (xd, &devconf->hqos);
+	  if (rv)
+	    return rv;
+	}
+
+      /* count the number of descriptors used for this device */
+      nb_desc += xd->nb_rx_desc + xd->nb_tx_desc * xd->tx_q_used;
+
+      error = ethernet_register_interface
+	(dm->vnet_main, dpdk_device_class.index, xd->device_index,
+	 /* ethernet address */ addr,
+	 &xd->vlib_hw_if_index, dpdk_flag_change);
+      if (error)
+	return error;
+
+      sw = vnet_get_hw_sw_interface (dm->vnet_main, xd->vlib_hw_if_index);
+      xd->vlib_sw_if_index = sw->sw_if_index;
+      hi = vnet_get_hw_interface (dm->vnet_main, xd->vlib_hw_if_index);
+
+      /*
+       * DAW-FIXME: The Cisco VIC firmware does not provide an api for a
+       *            driver to dynamically change the mtu.  If/when the
+       *            VIC firmware gets fixed, then this should be removed.
+       */
+      if (xd->pmd == VNET_DPDK_PMD_ENIC)
+	{
+	  /*
+	   * Initialize mtu to what has been set by CIMC in the firmware cfg.
+	   */
+	  hi->max_packet_bytes = dev_info.max_rx_pktlen;
+	  if (devconf->vlan_strip_offload != DPDK_DEVICE_VLAN_STRIP_OFF)
+	    vlan_strip = 1;	/* remove vlan tag from VIC port by default */
+	  else
+	    clib_warning ("VLAN strip disabled for interface\n");
+	}
+      else if (devconf->vlan_strip_offload == DPDK_DEVICE_VLAN_STRIP_ON)
+	vlan_strip = 1;
+
+      if (vlan_strip)
+	{
+	  int vlan_off;
+	  vlan_off = rte_eth_dev_get_vlan_offload (xd->device_index);
+	  vlan_off |= ETH_VLAN_STRIP_OFFLOAD;
+	  xd->port_conf.rxmode.hw_vlan_strip = vlan_off;
+	  if (rte_eth_dev_set_vlan_offload (xd->device_index, vlan_off) == 0)
+	    clib_warning ("VLAN strip enabled for interface\n");
+	  else
+	    clib_warning ("VLAN strip cannot be supported by interface\n");
+	}
+
+      hi->max_l3_packet_bytes[VLIB_RX] = hi->max_l3_packet_bytes[VLIB_TX] =
+	xd->port_conf.rxmode.max_rx_pkt_len - sizeof (ethernet_header_t);
+
+      rte_eth_dev_set_mtu (xd->device_index, hi->max_packet_bytes);
+    }
+
+  if (nb_desc > dm->conf->num_mbufs)
+    clib_warning ("%d mbufs allocated but total rx/tx ring size is %d\n",
+		  dm->conf->num_mbufs, nb_desc);
+
+  return 0;
+}
+
+static void
+dpdk_bind_devices_to_uio (dpdk_config_main_t * conf)
+{
+  vlib_pci_main_t *pm = &pci_main;
+  clib_error_t *error;
+  vlib_pci_device_t *d;
+  u8 *pci_addr = 0;
+  int num_whitelisted = vec_len (conf->dev_confs);
+
+  /* *INDENT-OFF* */
+  pool_foreach (d, pm->pci_devs, ({
+    dpdk_device_config_t * devconf = 0;
+    vec_reset_length (pci_addr);
+    pci_addr = format (pci_addr, "%U%c", format_vlib_pci_addr, &d->bus_address, 0);
+
+    if (d->device_class != PCI_CLASS_NETWORK_ETHERNET && d->device_class != PCI_CLASS_PROCESSOR_CO)
+      continue;
+
+    if (num_whitelisted)
+      {
+	uword * p = hash_get (conf->device_config_index_by_pci_addr, d->bus_address.as_u32);
+
+	if (!p)
+	  continue;
+
+	devconf = pool_elt_at_index (conf->dev_confs, p[0]);
+      }
+
+    /* virtio */
+    if (d->vendor_id == 0x1af4 && d->device_id == 0x1000)
+      ;
+    /* vmxnet3 */
+    else if (d->vendor_id == 0x15ad && d->device_id == 0x07b0)
+      ;
+    /* all Intel devices */
+    else if (d->vendor_id == 0x8086)
+      ;
+    /* Cisco VIC */
+    else if (d->vendor_id == 0x1137 && d->device_id == 0x0043)
+      ;
+    /* Chelsio T4/T5 */
+    else if (d->vendor_id == 0x1425 && (d->device_id & 0xe000) == 0x4000)
+      ;
+    else
+      {
+        clib_warning ("Unsupported Ethernet PCI device 0x%04x:0x%04x found "
+		      "at PCI address %s\n", (u16) d->vendor_id, (u16) d->device_id,
+		      pci_addr);
+        continue;
+      }
+
+    error = vlib_pci_bind_to_uio (d, (char *) conf->uio_driver_name);
+
+    if (error)
+      {
+	if (devconf == 0)
+	  {
+	    pool_get (conf->dev_confs, devconf);
+	    hash_set (conf->device_config_index_by_pci_addr, d->bus_address.as_u32,
+		      devconf - conf->dev_confs);
+	    devconf->pci_addr.as_u32 = d->bus_address.as_u32;
+	  }
+	devconf->is_blacklisted = 1;
+	clib_error_report (error);
+      }
+  }));
+  /* *INDENT-ON* */
+  vec_free (pci_addr);
+}
+
+static clib_error_t *
+dpdk_device_config (dpdk_config_main_t * conf, vlib_pci_addr_t pci_addr,
+		    unformat_input_t * input, u8 is_default)
+{
+  clib_error_t *error = 0;
+  uword *p;
+  dpdk_device_config_t *devconf;
+  unformat_input_t sub_input;
+
+  if (is_default)
+    {
+      devconf = &conf->default_devconf;
+    }
+  else
+    {
+      p = hash_get (conf->device_config_index_by_pci_addr, pci_addr.as_u32);
+
+      if (!p)
+	{
+	  pool_get (conf->dev_confs, devconf);
+	  hash_set (conf->device_config_index_by_pci_addr, pci_addr.as_u32,
+		    devconf - conf->dev_confs);
+	}
+      else
+	return clib_error_return (0,
+				  "duplicate configuration for PCI address %U",
+				  format_vlib_pci_addr, &pci_addr);
+    }
+
+  devconf->pci_addr.as_u32 = pci_addr.as_u32;
+  devconf->hqos_enabled = 0;
+  dpdk_device_config_hqos_default (&devconf->hqos);
+
+  if (!input)
+    return 0;
+
+  unformat_skip_white_space (input);
+  while (unformat_check_input (input) != UNFORMAT_END_OF_INPUT)
+    {
+      if (unformat (input, "num-rx-queues %u", &devconf->num_rx_queues))
+	;
+      else if (unformat (input, "num-tx-queues %u", &devconf->num_tx_queues))
+	;
+      else if (unformat (input, "num-rx-desc %u", &devconf->num_rx_desc))
+	;
+      else if (unformat (input, "num-tx-desc %u", &devconf->num_tx_desc))
+	;
+      else if (unformat (input, "workers %U", unformat_bitmap_list,
+			 &devconf->workers))
+	;
+      else
+	if (unformat
+	    (input, "rss %U", unformat_vlib_cli_sub_input, &sub_input))
+	{
+	  error = unformat_rss_fn (&sub_input, &devconf->rss_fn);
+	  if (error)
+	    break;
+	}
+      else if (unformat (input, "vlan-strip-offload off"))
+	devconf->vlan_strip_offload = DPDK_DEVICE_VLAN_STRIP_OFF;
+      else if (unformat (input, "vlan-strip-offload on"))
+	devconf->vlan_strip_offload = DPDK_DEVICE_VLAN_STRIP_ON;
+      else
+	if (unformat
+	    (input, "hqos %U", unformat_vlib_cli_sub_input, &sub_input))
+	{
+	  devconf->hqos_enabled = 1;
+	  error = unformat_hqos (&sub_input, &devconf->hqos);
+	  if (error)
+	    break;
+	}
+      else if (unformat (input, "hqos"))
+	{
+	  devconf->hqos_enabled = 1;
+	}
+      else
+	{
+	  error = clib_error_return (0, "unknown input `%U'",
+				     format_unformat_error, input);
+	  break;
+	}
+    }
+
+  if (error)
+    return error;
+
+  if (devconf->workers && devconf->num_rx_queues == 0)
+    devconf->num_rx_queues = clib_bitmap_count_set_bits (devconf->workers);
+  else if (devconf->workers &&
+	   clib_bitmap_count_set_bits (devconf->workers) !=
+	   devconf->num_rx_queues)
+    error =
+      clib_error_return (0,
+			 "%U: number of worker threadds must be "
+			 "equal to number of rx queues", format_vlib_pci_addr,
+			 &pci_addr);
+
+  return error;
+}
+
+static clib_error_t *
+dpdk_config (vlib_main_t * vm, unformat_input_t * input)
+{
+  clib_error_t *error = 0;
+  dpdk_main_t *dm = &dpdk_main;
+  dpdk_config_main_t *conf = &dpdk_config_main;
+  vlib_thread_main_t *tm = vlib_get_thread_main ();
+  dpdk_device_config_t *devconf;
+  vlib_pci_addr_t pci_addr;
+  unformat_input_t sub_input;
+  u8 *s, *tmp = 0;
+  u8 *rte_cmd = 0, *ethname = 0;
+  u32 log_level;
+  int ret, i;
+  int num_whitelisted = 0;
+  u8 no_pci = 0;
+  u8 no_huge = 0;
+  u8 huge_dir = 0;
+  u8 file_prefix = 0;
+  u8 *socket_mem = 0;
+
+  conf->device_config_index_by_pci_addr = hash_create (0, sizeof (uword));
+
+  while (unformat_check_input (input) != UNFORMAT_END_OF_INPUT)
+    {
+      /* Prime the pump */
+      if (unformat (input, "no-hugetlb"))
+	{
+	  vec_add1 (conf->eal_init_args, (u8 *) "no-huge");
+	  no_huge = 1;
+	}
+
+      else if (unformat (input, "enable-tcp-udp-checksum"))
+	conf->enable_tcp_udp_checksum = 1;
+
+      else if (unformat (input, "decimal-interface-names"))
+	conf->interface_name_format_decimal = 1;
+
+      else if (unformat (input, "no-multi-seg"))
+	conf->no_multi_seg = 1;
+
+      else if (unformat (input, "enable-cryptodev"))
+	conf->cryptodev = 1;
+
+      else if (unformat (input, "dev default %U", unformat_vlib_cli_sub_input,
+			 &sub_input))
+	{
+	  error =
+	    dpdk_device_config (conf, (vlib_pci_addr_t) (u32) ~ 1, &sub_input,
+				1);
+
+	  if (error)
+	    return error;
+	}
+      else
+	if (unformat
+	    (input, "dev %U %U", unformat_vlib_pci_addr, &pci_addr,
+	     unformat_vlib_cli_sub_input, &sub_input))
+	{
+	  error = dpdk_device_config (conf, pci_addr, &sub_input, 0);
+
+	  if (error)
+	    return error;
+
+	  num_whitelisted++;
+	}
+      else if (unformat (input, "dev %U", unformat_vlib_pci_addr, &pci_addr))
+	{
+	  error = dpdk_device_config (conf, pci_addr, 0, 0);
+
+	  if (error)
+	    return error;
+
+	  num_whitelisted++;
+	}
+      else if (unformat (input, "num-mbufs %d", &conf->num_mbufs))
+	;
+      else if (unformat (input, "kni %d", &conf->num_kni))
+	;
+      else if (unformat (input, "uio-driver %s", &conf->uio_driver_name))
+	;
+      else if (unformat (input, "socket-mem %s", &socket_mem))
+	;
+      else if (unformat (input, "no-pci"))
+	{
+	  no_pci = 1;
+	  tmp = format (0, "--no-pci%c", 0);
+	  vec_add1 (conf->eal_init_args, tmp);
+	}
+      else if (unformat (input, "poll-sleep %d", &dm->poll_sleep))
+	;
+
+#define _(a)                                    \
+      else if (unformat(input, #a))             \
+        {                                       \
+          tmp = format (0, "--%s%c", #a, 0);    \
+          vec_add1 (conf->eal_init_args, tmp);    \
+        }
+      foreach_eal_double_hyphen_predicate_arg
+#undef _
+#define _(a)                                          \
+	else if (unformat(input, #a " %s", &s))	      \
+	  {					      \
+            if (!strncmp(#a, "huge-dir", 8))          \
+              huge_dir = 1;                           \
+            else if (!strncmp(#a, "file-prefix", 11)) \
+              file_prefix = 1;                        \
+	    tmp = format (0, "--%s%c", #a, 0);	      \
+	    vec_add1 (conf->eal_init_args, tmp);      \
+	    vec_add1 (s, 0);			      \
+            if (!strncmp(#a, "vdev", 4))              \
+              if (strstr((char*)s, "af_packet"))      \
+                clib_warning ("af_packet obsoleted. Use CLI 'create host-interface'."); \
+	    vec_add1 (conf->eal_init_args, s);	      \
+	  }
+	foreach_eal_double_hyphen_arg
+#undef _
+#define _(a,b)						\
+	  else if (unformat(input, #a " %s", &s))	\
+	    {						\
+	      tmp = format (0, "-%s%c", #b, 0);		\
+	      vec_add1 (conf->eal_init_args, tmp);	\
+	      vec_add1 (s, 0);				\
+	      vec_add1 (conf->eal_init_args, s);	\
+	    }
+	foreach_eal_single_hyphen_arg
+#undef _
+#define _(a,b)						\
+	    else if (unformat(input, #a " %s", &s))	\
+	      {						\
+		tmp = format (0, "-%s%c", #b, 0);	\
+		vec_add1 (conf->eal_init_args, tmp);	\
+		vec_add1 (s, 0);			\
+		vec_add1 (conf->eal_init_args, s);	\
+		conf->a##_set_manually = 1;		\
+	      }
+	foreach_eal_single_hyphen_mandatory_arg
+#undef _
+	else if (unformat (input, "default"))
+	;
+
+      else if (unformat_skip_white_space (input))
+	;
+      else
+	{
+	  error = clib_error_return (0, "unknown input `%U'",
+				     format_unformat_error, input);
+	  goto done;
+	}
+    }
+
+  if (!conf->uio_driver_name)
+    conf->uio_driver_name = format (0, "uio_pci_generic%c", 0);
+
+  /*
+   * Use 1G huge pages if available.
+   */
+  if (!no_huge && !huge_dir)
+    {
+      u32 x, *mem_by_socket = 0;
+      uword c = 0;
+      u8 use_1g = 1;
+      u8 use_2m = 1;
+      u8 less_than_1g = 1;
+      int rv;
+
+      umount (DEFAULT_HUGE_DIR);
+
+      /* Process "socket-mem" parameter value */
+      if (vec_len (socket_mem))
+	{
+	  unformat_input_t in;
+	  unformat_init_vector (&in, socket_mem);
+	  while (unformat_check_input (&in) != UNFORMAT_END_OF_INPUT)
+	    {
+	      if (unformat (&in, "%u,", &x))
+		;
+	      else if (unformat (&in, "%u", &x))
+		;
+	      else if (unformat (&in, ","))
+		x = 0;
+	      else
+		break;
+
+	      vec_add1 (mem_by_socket, x);
+
+	      if (x > 1023)
+		less_than_1g = 0;
+	    }
+	  /* Note: unformat_free vec_frees(in.buffer), aka socket_mem... */
+	  unformat_free (&in);
+	  socket_mem = 0;
+	}
+      else
+	{
+	  /* *INDENT-OFF* */
+	  clib_bitmap_foreach (c, tm->cpu_socket_bitmap, (
+	    {
+	      vec_validate(mem_by_socket, c);
+	      mem_by_socket[c] = 256; /* default per-socket mem */
+	    }
+	  ));
+	  /* *INDENT-ON* */
+	}
+
+      /* check if available enough 1GB pages for each socket */
+      /* *INDENT-OFF* */
+      clib_bitmap_foreach (c, tm->cpu_socket_bitmap, (
+        {
+	  int pages_avail, page_size, mem;
+
+	  vec_validate(mem_by_socket, c);
+	  mem = mem_by_socket[c];
+
+	  page_size = 1024;
+	  pages_avail = vlib_sysfs_get_free_hugepages(c, page_size * 1024);
+
+	  if (pages_avail < 0 || page_size * pages_avail < mem)
+	    use_1g = 0;
+
+	  page_size = 2;
+	  pages_avail = vlib_sysfs_get_free_hugepages(c, page_size * 1024);
+
+	  if (pages_avail < 0 || page_size * pages_avail < mem)
+	    use_2m = 0;
+      }));
+      /* *INDENT-ON* */
+
+      if (mem_by_socket == 0)
+	{
+	  error = clib_error_return (0, "mem_by_socket NULL");
+	  goto done;
+	}
+      _vec_len (mem_by_socket) = c + 1;
+
+      /* regenerate socket_mem string */
+      vec_foreach_index (x, mem_by_socket)
+	socket_mem = format (socket_mem, "%s%u",
+			     socket_mem ? "," : "", mem_by_socket[x]);
+      socket_mem = format (socket_mem, "%c", 0);
+
+      vec_free (mem_by_socket);
+
+      rv = mkdir (VPP_RUN_DIR, 0755);
+      if (rv && errno != EEXIST)
+	{
+	  error = clib_error_return (0, "mkdir '%s' failed errno %d",
+				     VPP_RUN_DIR, errno);
+	  goto done;
+	}
+
+      rv = mkdir (DEFAULT_HUGE_DIR, 0755);
+      if (rv && errno != EEXIST)
+	{
+	  error = clib_error_return (0, "mkdir '%s' failed errno %d",
+				     DEFAULT_HUGE_DIR, errno);
+	  goto done;
+	}
+
+      if (use_1g && !(less_than_1g && use_2m))
+	{
+	  rv =
+	    mount ("none", DEFAULT_HUGE_DIR, "hugetlbfs", 0, "pagesize=1G");
+	}
+      else if (use_2m)
+	{
+	  rv = mount ("none", DEFAULT_HUGE_DIR, "hugetlbfs", 0, NULL);
+	}
+      else
+	{
+	  return clib_error_return (0, "not enough free huge pages");
+	}
+
+      if (rv)
+	{
+	  error = clib_error_return (0, "mount failed %d", errno);
+	  goto done;
+	}
+
+      tmp = format (0, "--huge-dir%c", 0);
+      vec_add1 (conf->eal_init_args, tmp);
+      tmp = format (0, "%s%c", DEFAULT_HUGE_DIR, 0);
+      vec_add1 (conf->eal_init_args, tmp);
+      if (!file_prefix)
+	{
+	  tmp = format (0, "--file-prefix%c", 0);
+	  vec_add1 (conf->eal_init_args, tmp);
+	  tmp = format (0, "vpp%c", 0);
+	  vec_add1 (conf->eal_init_args, tmp);
+	}
+    }
+
+  vec_free (rte_cmd);
+  vec_free (ethname);
+
+  if (error)
+    return error;
+
+  /* I'll bet that -c and -n must be the first and second args... */
+  if (!conf->coremask_set_manually)
+    {
+      vlib_thread_registration_t *tr;
+      uword *coremask = 0;
+      int i;
+
+      /* main thread core */
+      coremask = clib_bitmap_set (coremask, tm->main_lcore, 1);
+
+      for (i = 0; i < vec_len (tm->registrations); i++)
+	{
+	  tr = tm->registrations[i];
+	  coremask = clib_bitmap_or (coremask, tr->coremask);
+	}
+
+      vec_insert (conf->eal_init_args, 2, 1);
+      conf->eal_init_args[1] = (u8 *) "-c";
+      tmp = format (0, "%U%c", format_bitmap_hex, coremask, 0);
+      conf->eal_init_args[2] = tmp;
+      clib_bitmap_free (coremask);
+    }
+
+  if (!conf->nchannels_set_manually)
+    {
+      vec_insert (conf->eal_init_args, 2, 3);
+      conf->eal_init_args[3] = (u8 *) "-n";
+      tmp = format (0, "%d", conf->nchannels);
+      conf->eal_init_args[4] = tmp;
+    }
+
+  if (no_pci == 0 && geteuid () == 0)
+    dpdk_bind_devices_to_uio (conf);
+
+#define _(x) \
+    if (devconf->x == 0 && conf->default_devconf.x > 0) \
+      devconf->x = conf->default_devconf.x ;
+
+  /* *INDENT-OFF* */
+  pool_foreach (devconf, conf->dev_confs, ({
+
+    /* default per-device config items */
+    foreach_dpdk_device_config_item
+
+    /* add DPDK EAL whitelist/blacklist entry */
+    if (num_whitelisted > 0 && devconf->is_blacklisted == 0)
+      {
+	tmp = format (0, "-w%c", 0);
+	vec_add1 (conf->eal_init_args, tmp);
+	tmp = format (0, "%U%c", format_vlib_pci_addr, &devconf->pci_addr, 0);
+	vec_add1 (conf->eal_init_args, tmp);
+      }
+    else if (num_whitelisted == 0 && devconf->is_blacklisted != 0)
+      {
+	tmp = format (0, "-b%c", 0);
+	vec_add1 (conf->eal_init_args, tmp);
+	tmp = format (0, "%U%c", format_vlib_pci_addr, &devconf->pci_addr, 0);
+	vec_add1 (conf->eal_init_args, tmp);
+      }
+  }));
+  /* *INDENT-ON* */
+
+#undef _
+
+  /* set master-lcore */
+  tmp = format (0, "--master-lcore%c", 0);
+  vec_add1 (conf->eal_init_args, tmp);
+  tmp = format (0, "%u%c", tm->main_lcore, 0);
+  vec_add1 (conf->eal_init_args, tmp);
+
+  /* set socket-mem */
+  tmp = format (0, "--socket-mem%c", 0);
+  vec_add1 (conf->eal_init_args, tmp);
+  tmp = format (0, "%s%c", socket_mem, 0);
+  vec_add1 (conf->eal_init_args, tmp);
+
+  /* NULL terminate the "argv" vector, in case of stupidity */
+  vec_add1 (conf->eal_init_args, 0);
+  _vec_len (conf->eal_init_args) -= 1;
+
+  /* Set up DPDK eal and packet mbuf pool early. */
+
+  log_level = (CLIB_DEBUG > 0) ? RTE_LOG_DEBUG : RTE_LOG_NOTICE;
+
+  rte_set_log_level (log_level);
+
+  vm = vlib_get_main ();
+
+  /* make copy of args as rte_eal_init tends to mess up with arg array */
+  for (i = 1; i < vec_len (conf->eal_init_args); i++)
+    conf->eal_init_args_str = format (conf->eal_init_args_str, "%s ",
+				      conf->eal_init_args[i]);
+
+  ret =
+    rte_eal_init (vec_len (conf->eal_init_args),
+		  (char **) conf->eal_init_args);
+
+  /* lazy umount hugepages */
+  umount2 (DEFAULT_HUGE_DIR, MNT_DETACH);
+
+  if (ret < 0)
+    return clib_error_return (0, "rte_eal_init returned %d", ret);
+
+  /* Dump the physical memory layout prior to creating the mbuf_pool */
+  fprintf (stdout, "DPDK physical memory layout:\n");
+  rte_dump_physmem_layout (stdout);
+
+  /* main thread 1st */
+  error = vlib_buffer_pool_create (vm, conf->num_mbufs, rte_socket_id ());
+  if (error)
+    return error;
+
+  for (i = 0; i < RTE_MAX_LCORE; i++)
+    {
+      error = vlib_buffer_pool_create (vm, conf->num_mbufs,
+				       rte_lcore_to_socket_id (i));
+      if (error)
+	return error;
+    }
+
+done:
+  return error;
+}
+
+VLIB_CONFIG_FUNCTION (dpdk_config, "dpdk");
+
+void
+dpdk_update_link_state (dpdk_device_t * xd, f64 now)
+{
+  vnet_main_t *vnm = vnet_get_main ();
+  struct rte_eth_link prev_link = xd->link;
+  u32 hw_flags = 0;
+  u8 hw_flags_chg = 0;
+
+  /* only update link state for PMD interfaces */
+  if ((xd->flags & DPDK_DEVICE_FLAG_PMD) == 0)
+    return;
+
+  xd->time_last_link_update = now ? now : xd->time_last_link_update;
+  memset (&xd->link, 0, sizeof (xd->link));
+  rte_eth_link_get_nowait (xd->device_index, &xd->link);
+
+  if (LINK_STATE_ELOGS)
+    {
+      vlib_main_t *vm = vlib_get_main ();
+      ELOG_TYPE_DECLARE (e) =
+      {
+      .format =
+	  "update-link-state: sw_if_index %d, admin_up %d,"
+	  "old link_state %d new link_state %d",.format_args = "i4i1i1i1",};
+
+      struct
+      {
+	u32 sw_if_index;
+	u8 admin_up;
+	u8 old_link_state;
+	u8 new_link_state;
+      } *ed;
+      ed = ELOG_DATA (&vm->elog_main, e);
+      ed->sw_if_index = xd->vlib_sw_if_index;
+      ed->admin_up = (xd->flags & DPDK_DEVICE_FLAG_ADMIN_UP) != 0;
+      ed->old_link_state = (u8)
+	vnet_hw_interface_is_link_up (vnm, xd->vlib_hw_if_index);
+      ed->new_link_state = (u8) xd->link.link_status;
+    }
+
+  if ((xd->flags & DPDK_DEVICE_FLAG_ADMIN_UP) &&
+      ((xd->link.link_status != 0) ^
+       vnet_hw_interface_is_link_up (vnm, xd->vlib_hw_if_index)))
+    {
+      hw_flags_chg = 1;
+      hw_flags |= (xd->link.link_status ? VNET_HW_INTERFACE_FLAG_LINK_UP : 0);
+    }
+
+  if (hw_flags_chg || (xd->link.link_duplex != prev_link.link_duplex))
+    {
+      hw_flags_chg = 1;
+      switch (xd->link.link_duplex)
+	{
+	case ETH_LINK_HALF_DUPLEX:
+	  hw_flags |= VNET_HW_INTERFACE_FLAG_HALF_DUPLEX;
+	  break;
+	case ETH_LINK_FULL_DUPLEX:
+	  hw_flags |= VNET_HW_INTERFACE_FLAG_FULL_DUPLEX;
+	  break;
+	default:
+	  break;
+	}
+    }
+  if (hw_flags_chg || (xd->link.link_speed != prev_link.link_speed))
+    {
+      hw_flags_chg = 1;
+      switch (xd->link.link_speed)
+	{
+	case ETH_SPEED_NUM_10M:
+	  hw_flags |= VNET_HW_INTERFACE_FLAG_SPEED_10M;
+	  break;
+	case ETH_SPEED_NUM_100M:
+	  hw_flags |= VNET_HW_INTERFACE_FLAG_SPEED_100M;
+	  break;
+	case ETH_SPEED_NUM_1G:
+	  hw_flags |= VNET_HW_INTERFACE_FLAG_SPEED_1G;
+	  break;
+	case ETH_SPEED_NUM_10G:
+	  hw_flags |= VNET_HW_INTERFACE_FLAG_SPEED_10G;
+	  break;
+	case ETH_SPEED_NUM_40G:
+	  hw_flags |= VNET_HW_INTERFACE_FLAG_SPEED_40G;
+	  break;
+	case 0:
+	  break;
+	default:
+	  clib_warning ("unknown link speed %d", xd->link.link_speed);
+	  break;
+	}
+    }
+  if (hw_flags_chg)
+    {
+      if (LINK_STATE_ELOGS)
+	{
+	  vlib_main_t *vm = vlib_get_main ();
+
+	  ELOG_TYPE_DECLARE (e) =
+	  {
+	  .format =
+	      "update-link-state: sw_if_index %d, new flags %d",.format_args
+	      = "i4i4",};
+
+	  struct
+	  {
+	    u32 sw_if_index;
+	    u32 flags;
+	  } *ed;
+	  ed = ELOG_DATA (&vm->elog_main, e);
+	  ed->sw_if_index = xd->vlib_sw_if_index;
+	  ed->flags = hw_flags;
+	}
+      vnet_hw_interface_set_flags (vnm, xd->vlib_hw_if_index, hw_flags);
+    }
+}
+
+static uword
+dpdk_process (vlib_main_t * vm, vlib_node_runtime_t * rt, vlib_frame_t * f)
+{
+  clib_error_t *error;
+  vnet_main_t *vnm = vnet_get_main ();
+  dpdk_main_t *dm = &dpdk_main;
+  ethernet_main_t *em = ðernet_main;
+  dpdk_device_t *xd;
+  vlib_thread_main_t *tm = vlib_get_thread_main ();
+  int i;
+
+  error = dpdk_lib_init (dm);
+
+  /*
+   * Turn on the input node if we found some devices to drive
+   * and we're not running worker threads or i/o threads
+   */
+
+  if (error == 0 && vec_len (dm->devices) > 0)
+    {
+      if (tm->n_vlib_mains == 1)
+	vlib_node_set_state (vm, dpdk_input_node.index,
+			     VLIB_NODE_STATE_POLLING);
+      else
+	for (i = 0; i < tm->n_vlib_mains; i++)
+	  if (vec_len (dm->devices_by_cpu[i]) > 0)
+	    vlib_node_set_state (vlib_mains[i], dpdk_input_node.index,
+				 VLIB_NODE_STATE_POLLING);
+    }
+
+  if (error)
+    clib_error_report (error);
+
+  tm->worker_thread_release = 1;
+
+  f64 now = vlib_time_now (vm);
+  vec_foreach (xd, dm->devices)
+  {
+    dpdk_update_link_state (xd, now);
+  }
+
+  {
+    /*
+     * Extra set up for bond interfaces:
+     *  1. Setup MACs for bond interfaces and their slave links which was set
+     *     in dpdk_port_setup() but needs to be done again here to take effect.
+     *  2. Set up info for bond interface related CLI support.
+     */
+    int nports = rte_eth_dev_count ();
+    if (nports > 0)
+      {
+	for (i = 0; i < nports; i++)
+	  {
+	    struct rte_eth_dev_info dev_info;
+	    rte_eth_dev_info_get (i, &dev_info);
+	    if (!dev_info.driver_name)
+	      dev_info.driver_name = dev_info.pci_dev->driver->driver.name;
+
+	    ASSERT (dev_info.driver_name);
+	    if (strncmp (dev_info.driver_name, "rte_bond_pmd", 12) == 0)
+	      {
+		u8 addr[6];
+		u8 slink[16];
+		int nlink = rte_eth_bond_slaves_get (i, slink, 16);
+		if (nlink > 0)
+		  {
+		    vnet_hw_interface_t *bhi;
+		    ethernet_interface_t *bei;
+		    int rv;
+
+		    /* Get MAC of 1st slave link */
+		    rte_eth_macaddr_get (slink[0],
+					 (struct ether_addr *) addr);
+		    /* Set MAC of bounded interface to that of 1st slave link */
+		    rv =
+		      rte_eth_bond_mac_address_set (i,
+						    (struct ether_addr *)
+						    addr);
+		    if (rv < 0)
+		      clib_warning ("Failed to set MAC address");
+
+		    /* Populate MAC of bonded interface in VPP hw tables */
+		    bhi =
+		      vnet_get_hw_interface (vnm,
+					     dm->devices[i].vlib_hw_if_index);
+		    bei =
+		      pool_elt_at_index (em->interfaces, bhi->hw_instance);
+		    clib_memcpy (bhi->hw_address, addr, 6);
+		    clib_memcpy (bei->address, addr, 6);
+		    /* Init l3 packet size allowed on bonded interface */
+		    bhi->max_packet_bytes = ETHERNET_MAX_PACKET_BYTES;
+		    bhi->max_l3_packet_bytes[VLIB_RX] =
+		      bhi->max_l3_packet_bytes[VLIB_TX] =
+		      ETHERNET_MAX_PACKET_BYTES - sizeof (ethernet_header_t);
+		    while (nlink >= 1)
+		      {		/* for all slave links */
+			int slave = slink[--nlink];
+			dpdk_device_t *sdev = &dm->devices[slave];
+			vnet_hw_interface_t *shi;
+			vnet_sw_interface_t *ssi;
+			/* Add MAC to all slave links except the first one */
+			if (nlink)
+			  rte_eth_dev_mac_addr_add (slave,
+						    (struct ether_addr *)
+						    addr, 0);
+			/* Set slaves bitmap for bonded interface */
+			bhi->bond_info =
+			  clib_bitmap_set (bhi->bond_info,
+					   sdev->vlib_hw_if_index, 1);
+			/* Set slave link flags on slave interface */
+			shi =
+			  vnet_get_hw_interface (vnm, sdev->vlib_hw_if_index);
+			ssi =
+			  vnet_get_sw_interface (vnm, sdev->vlib_sw_if_index);
+			shi->bond_info = VNET_HW_INTERFACE_BOND_INFO_SLAVE;
+			ssi->flags |= VNET_SW_INTERFACE_FLAG_BOND_SLAVE;
+
+			/* Set l3 packet size allowed as the lowest of slave */
+			if (bhi->max_l3_packet_bytes[VLIB_RX] >
+			    shi->max_l3_packet_bytes[VLIB_RX])
+			  bhi->max_l3_packet_bytes[VLIB_RX] =
+			    bhi->max_l3_packet_bytes[VLIB_TX] =
+			    shi->max_l3_packet_bytes[VLIB_RX];
+
+			/* Set max packet size allowed as the lowest of slave */
+			if (bhi->max_packet_bytes > shi->max_packet_bytes)
+			  bhi->max_packet_bytes = shi->max_packet_bytes;
+		      }
+		  }
+	      }
+	  }
+      }
+  }
+
+  while (1)
+    {
+      /*
+       * check each time through the loop in case intervals are changed
+       */
+      f64 min_wait = dm->link_state_poll_interval < dm->stat_poll_interval ?
+	dm->link_state_poll_interval : dm->stat_poll_interval;
+
+      vlib_process_wait_for_event_or_clock (vm, min_wait);
+
+      if (dm->admin_up_down_in_progress)
+	/* skip the poll if an admin up down is in progress (on any interface) */
+	continue;
+
+      vec_foreach (xd, dm->devices)
+      {
+	f64 now = vlib_time_now (vm);
+	if ((now - xd->time_last_stats_update) >= dm->stat_poll_interval)
+	  dpdk_update_counters (xd, now);
+	if ((now - xd->time_last_link_update) >= dm->link_state_poll_interval)
+	  dpdk_update_link_state (xd, now);
+
+      }
+    }
+
+  return 0;
+}
+
+/* *INDENT-OFF* */
+VLIB_REGISTER_NODE (dpdk_process_node,static) = {
+    .function = dpdk_process,
+    .type = VLIB_NODE_TYPE_PROCESS,
+    .name = "dpdk-process",
+    .process_log2_n_stack_bytes = 17,
+};
+/* *INDENT-ON* */
+
+int
+dpdk_set_stat_poll_interval (f64 interval)
+{
+  if (interval < DPDK_MIN_STATS_POLL_INTERVAL)
+    return (VNET_API_ERROR_INVALID_VALUE);
+
+  dpdk_main.stat_poll_interval = interval;
+
+  return 0;
+}
+
+int
+dpdk_set_link_state_poll_interval (f64 interval)
+{
+  if (interval < DPDK_MIN_LINK_POLL_INTERVAL)
+    return (VNET_API_ERROR_INVALID_VALUE);
+
+  dpdk_main.link_state_poll_interval = interval;
+
+  return 0;
+}
+
+clib_error_t *
+dpdk_init (vlib_main_t * vm)
+{
+  dpdk_main_t *dm = &dpdk_main;
+  vlib_node_t *ei;
+  clib_error_t *error = 0;
+  vlib_thread_main_t *tm = vlib_get_thread_main ();
+
+  /* verify that structs are cacheline aligned */
+  STATIC_ASSERT (offsetof (dpdk_device_t, cacheline0) == 0,
+		 "Cache line marker must be 1st element in dpdk_device_t");
+  STATIC_ASSERT (offsetof (dpdk_device_t, cacheline1) ==
+		 CLIB_CACHE_LINE_BYTES,
+		 "Data in cache line 0 is bigger than cache line size");
+  STATIC_ASSERT (offsetof (frame_queue_trace_t, cacheline0) == 0,
+		 "Cache line marker must be 1st element in frame_queue_trace_t");
+
+  u8 *name;
+  name = format (0, "dpdk_%08x%c", api_version, 0);
+
+  /* Ask for a correctly-sized block of API message decode slots */
+  dm->msg_id_base = vl_msg_api_get_msg_ids
+    ((char *) name, VL_MSG_FIRST_AVAILABLE);
+  vec_free (name);
+
+  dm->vlib_main = vm;
+  dm->vnet_main = vnet_get_main ();
+  dm->conf = &dpdk_config_main;
+
+  error = dpdk_plugin_api_hookup (vm);
+
+  /* Add our API messages to the global name_crc hash table */
+  setup_message_id_table (dm, &api_main);
+
+//  TODO
+//  plugin_custom_dump_configure (dm);
+
+  ei = vlib_get_node_by_name (vm, (u8 *) "ethernet-input");
+  if (ei == 0)
+    return clib_error_return (0, "ethernet-input node AWOL");
+
+  dm->ethernet_input_node_index = ei->index;
+
+  dm->conf->nchannels = 4;
+  dm->conf->num_mbufs = dm->conf->num_mbufs ? dm->conf->num_mbufs : NB_MBUF;
+  vec_add1 (dm->conf->eal_init_args, (u8 *) "vnet");
+
+  dm->dpdk_device_by_kni_port_id = hash_create (0, sizeof (uword));
+  dm->vu_sw_if_index_by_listener_fd = hash_create (0, sizeof (uword));
+  dm->vu_sw_if_index_by_sock_fd = hash_create (0, sizeof (uword));
+
+  /* $$$ use n_thread_stacks since it's known-good at this point */
+  vec_validate (dm->recycle, tm->n_thread_stacks - 1);
+
+  /* Default vlib_buffer_t flags, DISABLES tcp/udp checksumming... */
+  dm->buffer_flags_template =
+    (VLIB_BUFFER_TOTAL_LENGTH_VALID | VLIB_BUFFER_EXT_HDR_VALID
+     | IP_BUFFER_L4_CHECKSUM_COMPUTED | IP_BUFFER_L4_CHECKSUM_CORRECT);
+
+  dm->stat_poll_interval = DPDK_STATS_POLL_INTERVAL;
+  dm->link_state_poll_interval = DPDK_LINK_POLL_INTERVAL;
+
+  /* init CLI */
+  if ((error = vlib_call_init_function (vm, dpdk_cli_init)))
+    return error;
+
+  return error;
+}
+
+VLIB_INIT_FUNCTION (dpdk_init);
+
+
+/*
+ * fd.io coding-style-patch-verification: ON
+ *
+ * Local Variables:
+ * eval: (c-set-style "gnu")
+ * End:
+ */
diff --git a/src/plugins/dpdk/ipsec/cli.c b/src/plugins/dpdk/ipsec/cli.c
new file mode 100644
index 00000000..40cee39b
--- /dev/null
+++ b/src/plugins/dpdk/ipsec/cli.c
@@ -0,0 +1,154 @@
+/*
+ * Copyright (c) 2016 Intel and/or its affiliates.
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at:
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#include 
+#include 
+#include 
+
+static void
+dpdk_ipsec_show_mapping (vlib_main_t * vm, u16 detail_display)
+{
+  dpdk_config_main_t *conf = &dpdk_config_main;
+  dpdk_crypto_main_t *dcm = &dpdk_crypto_main;
+  vlib_thread_main_t *tm = vlib_get_thread_main ();
+  u32 i, skip_master;
+
+  if (!conf->cryptodev)
+    {
+      vlib_cli_output (vm, "DPDK Cryptodev support is disabled\n");
+      return;
+    }
+
+  if (detail_display)
+    vlib_cli_output (vm, "worker\t%10s\t%15s\tdir\tdev\tqp\n",
+		     "cipher", "auth");
+  else
+    vlib_cli_output (vm, "worker\tcrypto device id(type)\n");
+
+  skip_master = vlib_num_workers () > 0;
+
+  for (i = 0; i < tm->n_vlib_mains; i++)
+    {
+      uword key, data;
+      u32 cpu_index = vlib_mains[i]->cpu_index;
+      crypto_worker_main_t *cwm = &dcm->workers_main[cpu_index];
+      u8 *s = 0;
+
+      if (skip_master)
+	{
+	  skip_master = 0;
+	  continue;
+	}
+
+      if (!detail_display)
+	{
+	  i32 last_cdev = -1;
+	  crypto_qp_data_t *qpd;
+
+	  s = format (s, "%u\t", cpu_index);
+
+	  /* *INDENT-OFF* */
+	  vec_foreach (qpd, cwm->qp_data)
+	    {
+	      u32 dev_id = qpd->dev_id;
+
+	      if ((u16) last_cdev != dev_id)
+		{
+		  struct rte_cryptodev_info cdev_info;
+
+		  rte_cryptodev_info_get (dev_id, &cdev_info);
+
+		  s = format(s, "%u(%s)\t", dev_id, cdev_info.feature_flags &
+			     RTE_CRYPTODEV_FF_HW_ACCELERATED ? "HW" : "SW");
+		}
+	      last_cdev = dev_id;
+	    }
+	  /* *INDENT-ON* */
+	  vlib_cli_output (vm, "%s", s);
+	}
+      else
+	{
+	  char cipher_str[15], auth_str[15];
+	  struct rte_cryptodev_capabilities cap;
+	  crypto_worker_qp_key_t *p_key = (crypto_worker_qp_key_t *) & key;
+	  /* *INDENT-OFF* */
+	  hash_foreach (key, data, cwm->algo_qp_map,
+	  ({
+	    cap.op = RTE_CRYPTO_OP_TYPE_SYMMETRIC;
+	    cap.sym.xform_type = RTE_CRYPTO_SYM_XFORM_CIPHER;
+	    cap.sym.cipher.algo = p_key->cipher_algo;
+	    check_algo_is_supported (&cap, cipher_str);
+	    cap.op = RTE_CRYPTO_OP_TYPE_SYMMETRIC;
+	    cap.sym.xform_type = RTE_CRYPTO_SYM_XFORM_AUTH;
+	    cap.sym.auth.algo = p_key->auth_algo;
+	    check_algo_is_supported (&cap, auth_str);
+	    vlib_cli_output (vm, "%u\t%10s\t%15s\t%3s\t%u\t%u\n",
+			     vlib_mains[i]->cpu_index, cipher_str, auth_str,
+			     p_key->is_outbound ? "out" : "in",
+			     cwm->qp_data[data].dev_id,
+			     cwm->qp_data[data].qp_id);
+	  }));
+	  /* *INDENT-ON* */
+	}
+    }
+}
+
+static clib_error_t *
+lcore_cryptodev_map_fn (vlib_main_t * vm, unformat_input_t * input,
+			vlib_cli_command_t * cmd)
+{
+  unformat_input_t _line_input, *line_input = &_line_input;
+  u16 detail = 0;
+  clib_error_t *error = NULL;
+
+  if (!unformat_user (input, unformat_line_input, line_input))
+    return 0;
+
+  while (unformat_check_input (line_input) != UNFORMAT_END_OF_INPUT)
+    {
+      if (unformat (line_input, "verbose"))
+	detail = 1;
+      else
+	{
+	  error = clib_error_return (0, "parse error: '%U'",
+				     format_unformat_error, line_input);
+	  goto done;
+	}
+    }
+
+  dpdk_ipsec_show_mapping (vm, detail);
+
+done:
+  unformat_free (line_input);
+
+  return error;
+}
+
+/* *INDENT-OFF* */
+VLIB_CLI_COMMAND (lcore_cryptodev_map, static) = {
+    .path = "show crypto device mapping",
+    .short_help =
+    "show cryptodev device mapping ",
+    .function = lcore_cryptodev_map_fn,
+};
+/* *INDENT-ON* */
+
+/*
+ * fd.io coding-style-patch-verification: ON
+ *
+ * Local Variables:
+ * eval: (c-set-style "gnu")
+ * End:
+ */
diff --git a/src/plugins/dpdk/ipsec/crypto_node.c b/src/plugins/dpdk/ipsec/crypto_node.c
new file mode 100644
index 00000000..dc3452b2
--- /dev/null
+++ b/src/plugins/dpdk/ipsec/crypto_node.c
@@ -0,0 +1,215 @@
+/*
+ *------------------------------------------------------------------
+ * crypto_node.c - DPDK Cryptodev input node
+ *
+ * Copyright (c) 2016 Intel and/or its affiliates.
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at:
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ *------------------------------------------------------------------
+ */
+
+#include 
+#include 
+#include 
+#include 
+
+#include 
+#include 
+#include 
+
+#define foreach_dpdk_crypto_input_next		\
+  _(DROP, "error-drop")				\
+  _(ENCRYPT_POST, "dpdk-esp-encrypt-post")	\
+  _(DECRYPT_POST, "dpdk-esp-decrypt-post")
+
+typedef enum
+{
+#define _(f,s) DPDK_CRYPTO_INPUT_NEXT_##f,
+  foreach_dpdk_crypto_input_next
+#undef _
+    DPDK_CRYPTO_INPUT_N_NEXT,
+} dpdk_crypto_input_next_t;
+
+#define foreach_dpdk_crypto_input_error		\
+  _(DQ_COPS, "Crypto ops dequeued")		\
+  _(COP_FAILED, "Crypto op failed")
+
+typedef enum
+{
+#define _(f,s) DPDK_CRYPTO_INPUT_ERROR_##f,
+  foreach_dpdk_crypto_input_error
+#undef _
+    DPDK_CRYPTO_INPUT_N_ERROR,
+} dpdk_crypto_input_error_t;
+
+static char *dpdk_crypto_input_error_strings[] = {
+#define _(n, s) s,
+  foreach_dpdk_crypto_input_error
+#undef _
+};
+
+vlib_node_registration_t dpdk_crypto_input_node;
+
+typedef struct
+{
+  u32 cdev;
+  u32 qp;
+  u32 status;
+  u32 sa_idx;
+  u32 next_index;
+} dpdk_crypto_input_trace_t;
+
+static u8 *
+format_dpdk_crypto_input_trace (u8 * s, va_list * args)
+{
+  CLIB_UNUSED (vlib_main_t * vm) = va_arg (*args, vlib_main_t *);
+  CLIB_UNUSED (vlib_node_t * node) = va_arg (*args, vlib_node_t *);
+  dpdk_crypto_input_trace_t *t = va_arg (*args, dpdk_crypto_input_trace_t *);
+
+  s = format (s, "dpdk_crypto: cryptodev-id %u queue-pair %u next-index %d",
+	      t->cdev, t->qp, t->next_index);
+
+  s = format (s, " status %u sa-idx %u\n", t->status, t->sa_idx);
+
+  return s;
+}
+
+static_always_inline u32
+dpdk_crypto_dequeue (vlib_main_t * vm, vlib_node_runtime_t * node,
+		     crypto_qp_data_t * qpd)
+{
+  u32 n_deq, *to_next = 0, next_index, n_cops, def_next_index;
+  struct rte_crypto_op **cops = qpd->cops;
+
+  if (qpd->inflights == 0)
+    return 0;
+
+  if (qpd->is_outbound)
+    def_next_index = DPDK_CRYPTO_INPUT_NEXT_ENCRYPT_POST;
+  else
+    def_next_index = DPDK_CRYPTO_INPUT_NEXT_DECRYPT_POST;
+
+  n_cops = rte_cryptodev_dequeue_burst (qpd->dev_id, qpd->qp_id,
+					cops, VLIB_FRAME_SIZE);
+  n_deq = n_cops;
+  next_index = def_next_index;
+
+  qpd->inflights -= n_cops;
+  ASSERT (qpd->inflights >= 0);
+
+  while (n_cops > 0)
+    {
+      u32 n_left_to_next;
+
+      vlib_get_next_frame (vm, node, next_index, to_next, n_left_to_next);
+
+      while (n_cops > 0 && n_left_to_next > 0)
+	{
+	  u32 bi0, next0;
+	  vlib_buffer_t *b0 = 0;
+	  struct rte_crypto_op *cop;
+	  struct rte_crypto_sym_op *sym_cop;
+
+	  cop = cops[0];
+	  cops += 1;
+	  n_cops -= 1;
+	  n_left_to_next -= 1;
+
+	  next0 = def_next_index;
+
+	  if (PREDICT_FALSE (cop->status != RTE_CRYPTO_OP_STATUS_SUCCESS))
+	    {
+	      next0 = DPDK_CRYPTO_INPUT_NEXT_DROP;
+	      vlib_node_increment_counter (vm, dpdk_crypto_input_node.index,
+					   DPDK_CRYPTO_INPUT_ERROR_COP_FAILED,
+					   1);
+	    }
+	  cop->status = RTE_CRYPTO_OP_STATUS_NOT_PROCESSED;
+
+	  sym_cop = (struct rte_crypto_sym_op *) (cop + 1);
+	  b0 = vlib_buffer_from_rte_mbuf (sym_cop->m_src);
+	  bi0 = vlib_get_buffer_index (vm, b0);
+
+	  to_next[0] = bi0;
+	  to_next += 1;
+
+	  if (PREDICT_FALSE (b0->flags & VLIB_BUFFER_IS_TRACED))
+	    {
+	      vlib_trace_next_frame (vm, node, next0);
+	      dpdk_crypto_input_trace_t *tr =
+		vlib_add_trace (vm, node, b0, sizeof (*tr));
+	      tr->cdev = qpd->dev_id;
+	      tr->qp = qpd->qp_id;
+	      tr->status = cop->status;
+	      tr->next_index = next0;
+	      tr->sa_idx = vnet_buffer (b0)->ipsec.sad_index;
+	    }
+
+	  vlib_validate_buffer_enqueue_x1 (vm, node, next_index, to_next,
+					   n_left_to_next, bi0, next0);
+	}
+      vlib_put_next_frame (vm, node, next_index, n_left_to_next);
+    }
+
+  crypto_free_cop (qpd, qpd->cops, n_deq);
+
+  vlib_node_increment_counter (vm, dpdk_crypto_input_node.index,
+			       DPDK_CRYPTO_INPUT_ERROR_DQ_COPS, n_deq);
+  return n_deq;
+}
+
+static uword
+dpdk_crypto_input_fn (vlib_main_t * vm, vlib_node_runtime_t * node,
+		      vlib_frame_t * frame)
+{
+  u32 cpu_index = os_get_cpu_number ();
+  dpdk_crypto_main_t *dcm = &dpdk_crypto_main;
+  crypto_worker_main_t *cwm = &dcm->workers_main[cpu_index];
+  crypto_qp_data_t *qpd;
+  u32 n_deq = 0;
+
+  /* *INDENT-OFF* */
+  vec_foreach (qpd, cwm->qp_data)
+      n_deq += dpdk_crypto_dequeue(vm, node, qpd);
+  /* *INDENT-ON* */
+
+  return n_deq;
+}
+
+/* *INDENT-OFF* */
+VLIB_REGISTER_NODE (dpdk_crypto_input_node) =
+{
+  .function = dpdk_crypto_input_fn,
+  .name = "dpdk-crypto-input",
+  .format_trace = format_dpdk_crypto_input_trace,
+  .type = VLIB_NODE_TYPE_INPUT,
+  .state = VLIB_NODE_STATE_DISABLED,
+  .n_errors = DPDK_CRYPTO_INPUT_N_ERROR,
+  .error_strings = dpdk_crypto_input_error_strings,
+  .n_next_nodes = DPDK_CRYPTO_INPUT_N_NEXT,
+  .next_nodes =
+  {
+#define _(s,n) [DPDK_CRYPTO_INPUT_NEXT_##s] = n,
+    foreach_dpdk_crypto_input_next
+#undef _
+  },
+};
+/* *INDENT-ON* */
+
+VLIB_NODE_FUNCTION_MULTIARCH (dpdk_crypto_input_node, dpdk_crypto_input_fn)
+/*
+ * fd.io coding-style-patch-verification: ON
+ *
+ * Local Variables:
+ * eval: (c-set-style "gnu")
+ * End:
+ */
diff --git a/src/plugins/dpdk/ipsec/dir.dox b/src/plugins/dpdk/ipsec/dir.dox
new file mode 100644
index 00000000..ffebfc4d
--- /dev/null
+++ b/src/plugins/dpdk/ipsec/dir.dox
@@ -0,0 +1,18 @@
+/*
+ * Copyright (c) 2016 Intel and/or its affiliates.
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at:
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+/**
+ @dir vnet/vnet/devices/dpdk/ipsec
+ @brief IPSec ESP encrypt/decrypt using DPDK Cryptodev API
+*/
diff --git a/src/plugins/dpdk/ipsec/dpdk_crypto_ipsec_doc.md b/src/plugins/dpdk/ipsec/dpdk_crypto_ipsec_doc.md
new file mode 100644
index 00000000..fed2fe0e
--- /dev/null
+++ b/src/plugins/dpdk/ipsec/dpdk_crypto_ipsec_doc.md
@@ -0,0 +1,86 @@
+# VPP IPSec implementation using DPDK Cryptodev API    {#dpdk_crypto_ipsec_doc}
+
+This document is meant to contain all related information about implementation and usability.
+
+
+## VPP IPsec with DPDK Cryptodev
+
+DPDK Cryptodev is an asynchronous crypto API that supports both Hardware and Software implementations (for more details refer to [DPDK Cryptography Device Library documentation](http://dpdk.org/doc/guides/prog_guide/cryptodev_lib.html)).
+
+When DPDK support is enabled and there are enough Cryptodev resources for all workers, the node graph is reconfigured by adding and changing default next nodes.
+
+The following nodes are added:
+* dpdk-crypto-input : polling input node, basically dequeuing from crypto devices.
+* dpdk-esp-encrypt : internal node.
+* dpdk-esp-decrypt : internal node.
+* dpdk-esp-encrypt-post : internal node.
+* dpdk-esp-decrypt-post : internal node.
+
+Set new default next nodes:
+* for esp encryption: esp-encrypt -> dpdk-esp-encrypt
+* for esp decryption: esp-decrypt -> dpdk-esp-decrypt
+
+
+### How to enable VPP IPSec with DPDK Cryptodev support
+
+DPDK Cryptodev is supported in DPDK enabled VPP.
+By default, only HW Cryptodev is supported but needs to be explicetly enabled with the following config option:
+
+```
+dpdk {
+    enable-cryptodev
+}
+```
+
+To enable SW Cryptodev support (AESNI-MB-PMD and GCM-PMD), we need the following env option:
+
+    vpp_uses_dpdk_cryptodev_sw=yes
+
+A couple of ways to achive this:
+* uncomment/add it in the platforms config (ie. build-data/platforms/vpp.mk)
+* set the option when building vpp (ie. make vpp_uses_dpdk_cryptodev_sw=yes build-release)
+
+When enabling SW Cryptodev support, it means that you need to pre-build the required crypto libraries needed by those SW Cryptodev PMDs.
+
+
+### Crypto Resources allocation
+
+VPP allocates crypto resources based on a best effort approach:
+* first allocate Hardware crypto resources, then Software.
+* if there are not enough crypto resources for all workers, the graph node is not modifed, therefore the default VPP IPsec implementation based in OpenSSL is used. The following message is displayed:
+
+      0: dpdk_ipsec_init: not enough cryptodevs for ipsec
+
+
+### Configuration example
+
+To enable DPDK Cryptodev the user just need to provide the startup.conf option
+as mentioned previously.
+
+Example startup.conf:
+
+```
+dpdk {
+    socket-mem 1024,1024
+    num-mbufs 131072
+    dev 0000:81:00.0
+    dev 0000:81:00.1
+    enable-cryptodev
+    dev 0000:85:01.0
+    dev 0000:85:01.1
+    vdev cryptodev_aesni_mb_pmd,socket_id=1
+    vdev cryptodev_aesni_mb_pmd,socket_id=1
+}
+```
+
+In the above configuration:
+* 0000:85:01.0 and 0000:85:01.1 are crypto BDFs and they require the same driver binding as DPDK Ethernet devices but they do not support any extra configuration options.
+* Two AESNI-MB Software Cryptodev PMDs are created in NUMA node 1.
+
+For further details refer to [DPDK Crypto Device Driver documentation](http://dpdk.org/doc/guides/cryptodevs/index.html)
+
+### Operational data
+
+The following CLI command displays the Cryptodev/Worker mapping:
+
+    show crypto device mapping [verbose]
diff --git a/src/plugins/dpdk/ipsec/esp.h b/src/plugins/dpdk/ipsec/esp.h
new file mode 100644
index 00000000..320295b1
--- /dev/null
+++ b/src/plugins/dpdk/ipsec/esp.h
@@ -0,0 +1,249 @@
+/*
+ * Copyright (c) 2016 Intel and/or its affiliates.
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at:
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+#ifndef __DPDK_ESP_H__
+#define __DPDK_ESP_H__
+
+#include 
+#include 
+#include 
+
+typedef struct
+{
+  enum rte_crypto_cipher_algorithm algo;
+  u8 key_len;
+  u8 iv_len;
+} dpdk_esp_crypto_alg_t;
+
+typedef struct
+{
+  enum rte_crypto_auth_algorithm algo;
+  u8 trunc_size;
+} dpdk_esp_integ_alg_t;
+
+typedef struct
+{
+  dpdk_esp_crypto_alg_t *esp_crypto_algs;
+  dpdk_esp_integ_alg_t *esp_integ_algs;
+} dpdk_esp_main_t;
+
+dpdk_esp_main_t dpdk_esp_main;
+
+static_always_inline void
+dpdk_esp_init ()
+{
+  dpdk_esp_main_t *em = &dpdk_esp_main;
+  dpdk_esp_integ_alg_t *i;
+  dpdk_esp_crypto_alg_t *c;
+
+  vec_validate (em->esp_crypto_algs, IPSEC_CRYPTO_N_ALG - 1);
+
+  c = &em->esp_crypto_algs[IPSEC_CRYPTO_ALG_AES_CBC_128];
+  c->algo = RTE_CRYPTO_CIPHER_AES_CBC;
+  c->key_len = 16;
+  c->iv_len = 16;
+
+  c = &em->esp_crypto_algs[IPSEC_CRYPTO_ALG_AES_CBC_192];
+  c->algo = RTE_CRYPTO_CIPHER_AES_CBC;
+  c->key_len = 24;
+  c->iv_len = 16;
+
+  c = &em->esp_crypto_algs[IPSEC_CRYPTO_ALG_AES_CBC_256];
+  c->algo = RTE_CRYPTO_CIPHER_AES_CBC;
+  c->key_len = 32;
+  c->iv_len = 16;
+
+  c = &em->esp_crypto_algs[IPSEC_CRYPTO_ALG_AES_GCM_128];
+  c->algo = RTE_CRYPTO_CIPHER_AES_GCM;
+  c->key_len = 16;
+  c->iv_len = 8;
+
+  vec_validate (em->esp_integ_algs, IPSEC_INTEG_N_ALG - 1);
+
+  i = &em->esp_integ_algs[IPSEC_INTEG_ALG_SHA1_96];
+  i->algo = RTE_CRYPTO_AUTH_SHA1_HMAC;
+  i->trunc_size = 12;
+
+  i = &em->esp_integ_algs[IPSEC_INTEG_ALG_SHA_256_96];
+  i->algo = RTE_CRYPTO_AUTH_SHA256_HMAC;
+  i->trunc_size = 12;
+
+  i = &em->esp_integ_algs[IPSEC_INTEG_ALG_SHA_256_128];
+  i->algo = RTE_CRYPTO_AUTH_SHA256_HMAC;
+  i->trunc_size = 16;
+
+  i = &em->esp_integ_algs[IPSEC_INTEG_ALG_SHA_384_192];
+  i->algo = RTE_CRYPTO_AUTH_SHA384_HMAC;
+  i->trunc_size = 24;
+
+  i = &em->esp_integ_algs[IPSEC_INTEG_ALG_SHA_512_256];
+  i->algo = RTE_CRYPTO_AUTH_SHA512_HMAC;
+  i->trunc_size = 32;
+
+  i = &em->esp_integ_algs[IPSEC_INTEG_ALG_AES_GCM_128];
+  i->algo = RTE_CRYPTO_AUTH_AES_GCM;
+  i->trunc_size = 16;
+}
+
+static_always_inline int
+translate_crypto_algo (ipsec_crypto_alg_t crypto_algo,
+		       struct rte_crypto_sym_xform *cipher_xform)
+{
+  switch (crypto_algo)
+    {
+    case IPSEC_CRYPTO_ALG_NONE:
+      cipher_xform->cipher.algo = RTE_CRYPTO_CIPHER_NULL;
+      break;
+    case IPSEC_CRYPTO_ALG_AES_CBC_128:
+    case IPSEC_CRYPTO_ALG_AES_CBC_192:
+    case IPSEC_CRYPTO_ALG_AES_CBC_256:
+      cipher_xform->cipher.algo = RTE_CRYPTO_CIPHER_AES_CBC;
+      break;
+    case IPSEC_CRYPTO_ALG_AES_GCM_128:
+      cipher_xform->cipher.algo = RTE_CRYPTO_CIPHER_AES_GCM;
+      break;
+    default:
+      return -1;
+    }
+
+  cipher_xform->type = RTE_CRYPTO_SYM_XFORM_CIPHER;
+
+  return 0;
+}
+
+static_always_inline int
+translate_integ_algo (ipsec_integ_alg_t integ_alg,
+		      struct rte_crypto_sym_xform *auth_xform, int use_esn)
+{
+  switch (integ_alg)
+    {
+    case IPSEC_INTEG_ALG_NONE:
+      auth_xform->auth.algo = RTE_CRYPTO_AUTH_NULL;
+      auth_xform->auth.digest_length = 0;
+      break;
+    case IPSEC_INTEG_ALG_SHA1_96:
+      auth_xform->auth.algo = RTE_CRYPTO_AUTH_SHA1_HMAC;
+      auth_xform->auth.digest_length = 12;
+      break;
+    case IPSEC_INTEG_ALG_SHA_256_96:
+      auth_xform->auth.algo = RTE_CRYPTO_AUTH_SHA256_HMAC;
+      auth_xform->auth.digest_length = 12;
+      break;
+    case IPSEC_INTEG_ALG_SHA_256_128:
+      auth_xform->auth.algo = RTE_CRYPTO_AUTH_SHA256_HMAC;
+      auth_xform->auth.digest_length = 16;
+      break;
+    case IPSEC_INTEG_ALG_SHA_384_192:
+      auth_xform->auth.algo = RTE_CRYPTO_AUTH_SHA384_HMAC;
+      auth_xform->auth.digest_length = 24;
+      break;
+    case IPSEC_INTEG_ALG_SHA_512_256:
+      auth_xform->auth.algo = RTE_CRYPTO_AUTH_SHA512_HMAC;
+      auth_xform->auth.digest_length = 32;
+      break;
+    case IPSEC_INTEG_ALG_AES_GCM_128:
+      auth_xform->auth.algo = RTE_CRYPTO_AUTH_AES_GCM;
+      auth_xform->auth.digest_length = 16;
+      auth_xform->auth.add_auth_data_length = use_esn ? 12 : 8;
+      break;
+    default:
+      return -1;
+    }
+
+  auth_xform->type = RTE_CRYPTO_SYM_XFORM_AUTH;
+
+  return 0;
+}
+
+static_always_inline int
+create_sym_sess (ipsec_sa_t * sa, crypto_sa_session_t * sa_sess,
+		 u8 is_outbound)
+{
+  u32 cpu_index = os_get_cpu_number ();
+  dpdk_crypto_main_t *dcm = &dpdk_crypto_main;
+  crypto_worker_main_t *cwm = &dcm->workers_main[cpu_index];
+  struct rte_crypto_sym_xform cipher_xform = { 0 };
+  struct rte_crypto_sym_xform auth_xform = { 0 };
+  struct rte_crypto_sym_xform *xfs;
+  uword key = 0, *data;
+  crypto_worker_qp_key_t *p_key = (crypto_worker_qp_key_t *) & key;
+
+  if (sa->crypto_alg == IPSEC_CRYPTO_ALG_AES_GCM_128)
+    {
+      sa->crypto_key_len -= 4;
+      clib_memcpy (&sa->salt, &sa->crypto_key[sa->crypto_key_len], 4);
+    }
+  else
+    {
+      u32 seed = (u32) clib_cpu_time_now ();
+      sa->salt = random_u32 (&seed);
+    }
+
+  cipher_xform.type = RTE_CRYPTO_SYM_XFORM_CIPHER;
+  cipher_xform.cipher.key.data = sa->crypto_key;
+  cipher_xform.cipher.key.length = sa->crypto_key_len;
+
+  auth_xform.type = RTE_CRYPTO_SYM_XFORM_AUTH;
+  auth_xform.auth.key.data = sa->integ_key;
+  auth_xform.auth.key.length = sa->integ_key_len;
+
+  if (translate_crypto_algo (sa->crypto_alg, &cipher_xform) < 0)
+    return -1;
+  p_key->cipher_algo = cipher_xform.cipher.algo;
+
+  if (translate_integ_algo (sa->integ_alg, &auth_xform, sa->use_esn) < 0)
+    return -1;
+  p_key->auth_algo = auth_xform.auth.algo;
+
+  if (is_outbound)
+    {
+      cipher_xform.cipher.op = RTE_CRYPTO_CIPHER_OP_ENCRYPT;
+      auth_xform.auth.op = RTE_CRYPTO_AUTH_OP_GENERATE;
+      cipher_xform.next = &auth_xform;
+      xfs = &cipher_xform;
+    }
+  else
+    {
+      cipher_xform.cipher.op = RTE_CRYPTO_CIPHER_OP_DECRYPT;
+      auth_xform.auth.op = RTE_CRYPTO_AUTH_OP_VERIFY;
+      auth_xform.next = &cipher_xform;
+      xfs = &auth_xform;
+    }
+
+  p_key->is_outbound = is_outbound;
+
+  data = hash_get (cwm->algo_qp_map, key);
+  if (!data)
+    return -1;
+
+  sa_sess->sess =
+    rte_cryptodev_sym_session_create (cwm->qp_data[*data].dev_id, xfs);
+
+  if (!sa_sess->sess)
+    return -1;
+
+  sa_sess->qp_index = (u8) * data;
+
+  return 0;
+}
+
+#endif /* __DPDK_ESP_H__ */
+
+/*
+ * fd.io coding-style-patch-verification: ON
+ *
+ * Local Variables:
+ * eval: (c-set-style "gnu")
+ * End:
+ */
diff --git a/src/plugins/dpdk/ipsec/esp_decrypt.c b/src/plugins/dpdk/ipsec/esp_decrypt.c
new file mode 100644
index 00000000..286e03f8
--- /dev/null
+++ b/src/plugins/dpdk/ipsec/esp_decrypt.c
@@ -0,0 +1,594 @@
+/*
+ * esp_decrypt.c : IPSec ESP Decrypt node using DPDK Cryptodev
+ *
+ * Copyright (c) 2016 Intel and/or its affiliates.
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at:
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#include 
+#include 
+#include 
+
+#include 
+#include 
+#include 
+#include 
+#include 
+
+#define foreach_esp_decrypt_next	       \
+_(DROP, "error-drop")			       \
+_(IP4_INPUT, "ip4-input")		       \
+_(IP6_INPUT, "ip6-input")
+
+#define _(v, s) ESP_DECRYPT_NEXT_##v,
+typedef enum {
+  foreach_esp_decrypt_next
+#undef _
+  ESP_DECRYPT_N_NEXT,
+} esp_decrypt_next_t;
+
+#define foreach_esp_decrypt_error		 \
+ _(RX_PKTS, "ESP pkts received")		 \
+ _(DECRYPTION_FAILED, "ESP decryption failed")   \
+ _(REPLAY, "SA replayed packet")	         \
+ _(NOT_IP, "Not IP packet (dropped)")	         \
+ _(ENQ_FAIL, "Enqueue failed (buffer full)")     \
+ _(NO_CRYPTODEV, "Cryptodev not configured")     \
+ _(BAD_LEN, "Invalid ciphertext length")         \
+ _(UNSUPPORTED, "Cipher/Auth not supported")
+
+
+typedef enum {
+#define _(sym,str) ESP_DECRYPT_ERROR_##sym,
+  foreach_esp_decrypt_error
+#undef _
+  ESP_DECRYPT_N_ERROR,
+} esp_decrypt_error_t;
+
+static char * esp_decrypt_error_strings[] = {
+#define _(sym,string) string,
+  foreach_esp_decrypt_error
+#undef _
+};
+
+vlib_node_registration_t dpdk_esp_decrypt_node;
+
+typedef struct {
+  ipsec_crypto_alg_t crypto_alg;
+  ipsec_integ_alg_t integ_alg;
+} esp_decrypt_trace_t;
+
+/* packet trace format function */
+static u8 * format_esp_decrypt_trace (u8 * s, va_list * args)
+{
+  CLIB_UNUSED (vlib_main_t * vm) = va_arg (*args, vlib_main_t *);
+  CLIB_UNUSED (vlib_node_t * node) = va_arg (*args, vlib_node_t *);
+  esp_decrypt_trace_t * t = va_arg (*args, esp_decrypt_trace_t *);
+
+  s = format (s, "esp: crypto %U integrity %U",
+	      format_ipsec_crypto_alg, t->crypto_alg,
+	      format_ipsec_integ_alg, t->integ_alg);
+  return s;
+}
+
+static uword
+dpdk_esp_decrypt_node_fn (vlib_main_t * vm,
+	     vlib_node_runtime_t * node,
+	     vlib_frame_t * from_frame)
+{
+  u32 n_left_from, *from, *to_next, next_index;
+  ipsec_main_t *im = &ipsec_main;
+  u32 cpu_index = os_get_cpu_number();
+  dpdk_crypto_main_t * dcm = &dpdk_crypto_main;
+  dpdk_esp_main_t * em = &dpdk_esp_main;
+  u32 i;
+
+  from = vlib_frame_vector_args (from_frame);
+  n_left_from = from_frame->n_vectors;
+
+  if (PREDICT_FALSE(!dcm->workers_main))
+    {
+      vlib_node_increment_counter (vm, dpdk_esp_decrypt_node.index,
+	      ESP_DECRYPT_ERROR_NO_CRYPTODEV, n_left_from);
+      vlib_buffer_free(vm, from, n_left_from);
+      return n_left_from;
+    }
+
+  crypto_worker_main_t *cwm = vec_elt_at_index(dcm->workers_main, cpu_index);
+  u32 n_qps = vec_len(cwm->qp_data);
+  struct rte_crypto_op ** cops_to_enq[n_qps];
+  u32 n_cop_qp[n_qps], * bi_to_enq[n_qps];
+
+  for (i = 0; i < n_qps; i++)
+    {
+      bi_to_enq[i] = cwm->qp_data[i].bi;
+      cops_to_enq[i] = cwm->qp_data[i].cops;
+    }
+
+  memset(n_cop_qp, 0, n_qps * sizeof(u32));
+
+  crypto_alloc_cops();
+
+  next_index = ESP_DECRYPT_NEXT_DROP;
+
+  while (n_left_from > 0)
+    {
+      u32 n_left_to_next;
+
+      vlib_get_next_frame (vm, node, next_index, to_next, n_left_to_next);
+
+      while (n_left_from > 0 && n_left_to_next > 0)
+	{
+	  u32 bi0, sa_index0 = ~0, seq, icv_size, iv_size;
+	  vlib_buffer_t * b0;
+	  esp_header_t * esp0;
+	  ipsec_sa_t * sa0;
+	  struct rte_mbuf * mb0 = 0;
+	  const int BLOCK_SIZE = 16;
+	  crypto_sa_session_t * sa_sess;
+	  void * sess;
+	  u16 qp_index;
+	  struct rte_crypto_op * cop = 0;
+
+	  bi0 = from[0];
+	  from += 1;
+	  n_left_from -= 1;
+
+	  b0 = vlib_get_buffer (vm, bi0);
+	  esp0 = vlib_buffer_get_current (b0);
+
+	  sa_index0 = vnet_buffer(b0)->ipsec.sad_index;
+	  sa0 = pool_elt_at_index (im->sad, sa_index0);
+
+	  seq = clib_host_to_net_u32(esp0->seq);
+
+	  /* anti-replay check */
+	  if (sa0->use_anti_replay)
+	    {
+	      int rv = 0;
+
+	      if (PREDICT_TRUE(sa0->use_esn))
+		rv = esp_replay_check_esn(sa0, seq);
+	      else
+		rv = esp_replay_check(sa0, seq);
+
+	      if (PREDICT_FALSE(rv))
+		{
+		  clib_warning ("anti-replay SPI %u seq %u", sa0->spi, seq);
+		  vlib_node_increment_counter (vm, dpdk_esp_decrypt_node.index,
+					       ESP_DECRYPT_ERROR_REPLAY, 1);
+		  to_next[0] = bi0;
+		  to_next += 1;
+		  n_left_to_next -= 1;
+		  goto trace;
+		}
+	    }
+
+	  sa0->total_data_size += b0->current_length;
+
+	  if (PREDICT_FALSE(sa0->integ_alg == IPSEC_INTEG_ALG_NONE) ||
+		  PREDICT_FALSE(sa0->crypto_alg == IPSEC_CRYPTO_ALG_NONE))
+	    {
+	      clib_warning ("SPI %u : only cipher + auth supported", sa0->spi);
+	      vlib_node_increment_counter (vm, dpdk_esp_decrypt_node.index,
+					   ESP_DECRYPT_ERROR_UNSUPPORTED, 1);
+	      to_next[0] = bi0;
+	      to_next += 1;
+	      n_left_to_next -= 1;
+	      goto trace;
+	    }
+
+	  sa_sess = pool_elt_at_index(cwm->sa_sess_d[0], sa_index0);
+
+	  if (PREDICT_FALSE(!sa_sess->sess))
+	    {
+	      int ret = create_sym_sess(sa0, sa_sess, 0);
+
+	      if (PREDICT_FALSE (ret))
+		{
+		  to_next[0] = bi0;
+		  to_next += 1;
+		  n_left_to_next -= 1;
+		  goto trace;
+		}
+	    }
+
+	  sess = sa_sess->sess;
+	  qp_index = sa_sess->qp_index;
+
+	  ASSERT (vec_len (vec_elt (cwm->qp_data, qp_index).free_cops) > 0);
+	  cop = vec_pop (vec_elt (cwm->qp_data, qp_index).free_cops);
+	  ASSERT (cop->status == RTE_CRYPTO_OP_STATUS_NOT_PROCESSED);
+
+	  cops_to_enq[qp_index][0] = cop;
+	  cops_to_enq[qp_index] += 1;
+	  n_cop_qp[qp_index] += 1;
+	  bi_to_enq[qp_index][0] = bi0;
+	  bi_to_enq[qp_index] += 1;
+
+	  rte_crypto_op_attach_sym_session(cop, sess);
+
+	  icv_size = em->esp_integ_algs[sa0->integ_alg].trunc_size;
+	  iv_size = em->esp_crypto_algs[sa0->crypto_alg].iv_len;
+
+	  /* Convert vlib buffer to mbuf */
+	  mb0 = rte_mbuf_from_vlib_buffer(b0);
+	  mb0->data_len = b0->current_length;
+	  mb0->pkt_len = b0->current_length;
+	  mb0->data_off = RTE_PKTMBUF_HEADROOM + b0->current_data;
+
+	  /* Outer IP header has already been stripped */
+	  u16 payload_len = rte_pktmbuf_pkt_len(mb0) - sizeof (esp_header_t) -
+	      iv_size - icv_size;
+
+	  if ((payload_len & (BLOCK_SIZE - 1)) || (payload_len <= 0))
+	    {
+	      clib_warning ("payload %u not multiple of %d\n",
+			    payload_len, BLOCK_SIZE);
+	      vlib_node_increment_counter (vm, dpdk_esp_decrypt_node.index,
+					   ESP_DECRYPT_ERROR_BAD_LEN, 1);
+	      vec_add (vec_elt (cwm->qp_data, qp_index).free_cops, &cop, 1);
+	      bi_to_enq[qp_index] -= 1;
+	      cops_to_enq[qp_index] -= 1;
+	      n_cop_qp[qp_index] -= 1;
+	      to_next[0] = bi0;
+	      to_next += 1;
+	      n_left_to_next -= 1;
+	      goto trace;
+	    }
+
+	  struct rte_crypto_sym_op *sym_cop = (struct rte_crypto_sym_op *)(cop + 1);
+
+	  sym_cop->m_src = mb0;
+	  sym_cop->cipher.data.offset = sizeof (esp_header_t) + iv_size;
+	  sym_cop->cipher.data.length = payload_len;
+
+          u8 *iv = rte_pktmbuf_mtod_offset(mb0, void*, sizeof (esp_header_t));
+          dpdk_cop_priv_t * priv = (dpdk_cop_priv_t *)(sym_cop + 1);
+
+          if (sa0->crypto_alg == IPSEC_CRYPTO_ALG_AES_GCM_128)
+            {
+              dpdk_gcm_cnt_blk *icb = &priv->cb;
+              icb->salt = sa0->salt;
+              clib_memcpy(icb->iv, iv, 8);
+              icb->cnt = clib_host_to_net_u32(1);
+              sym_cop->cipher.iv.data = (u8 *)icb;
+              sym_cop->cipher.iv.phys_addr = cop->phys_addr +
+		(uintptr_t)icb - (uintptr_t)cop;
+              sym_cop->cipher.iv.length = 16;
+
+              u8 *aad = priv->aad;
+              clib_memcpy(aad, iv - sizeof(esp_header_t), 8);
+              sym_cop->auth.aad.data = aad;
+              sym_cop->auth.aad.phys_addr = cop->phys_addr +
+                  (uintptr_t)aad - (uintptr_t)cop;
+              if (sa0->use_esn)
+                {
+                  *((u32*)&aad[8]) = sa0->seq_hi;
+                  sym_cop->auth.aad.length = 12;
+                }
+              else
+                {
+                  sym_cop->auth.aad.length = 8;
+                }
+
+              sym_cop->auth.digest.data = rte_pktmbuf_mtod_offset(mb0, void*,
+                       rte_pktmbuf_pkt_len(mb0) - icv_size);
+              sym_cop->auth.digest.phys_addr = rte_pktmbuf_mtophys_offset(mb0,
+                       rte_pktmbuf_pkt_len(mb0) - icv_size);
+              sym_cop->auth.digest.length = icv_size;
+
+            }
+          else
+            {
+              sym_cop->cipher.iv.data = rte_pktmbuf_mtod_offset(mb0, void*,
+                       sizeof (esp_header_t));
+              sym_cop->cipher.iv.phys_addr = rte_pktmbuf_mtophys_offset(mb0,
+                       sizeof (esp_header_t));
+              sym_cop->cipher.iv.length = iv_size;
+
+              if (sa0->use_esn)
+                {
+                  dpdk_cop_priv_t* priv = (dpdk_cop_priv_t*) (sym_cop + 1);
+                  u8* payload_end = rte_pktmbuf_mtod_offset(
+                      mb0, u8*, sizeof(esp_header_t) + iv_size + payload_len);
+
+                  clib_memcpy (priv->icv, payload_end, icv_size);
+                  *((u32*) payload_end) = sa0->seq_hi;
+                  sym_cop->auth.data.offset = 0;
+                  sym_cop->auth.data.length = sizeof(esp_header_t) + iv_size
+                      + payload_len + sizeof(sa0->seq_hi);
+                  sym_cop->auth.digest.data = priv->icv;
+                  sym_cop->auth.digest.phys_addr = cop->phys_addr
+                      + (uintptr_t) priv->icv - (uintptr_t) cop;
+                  sym_cop->auth.digest.length = icv_size;
+                }
+              else
+                {
+                  sym_cop->auth.data.offset = 0;
+                  sym_cop->auth.data.length = sizeof(esp_header_t) +
+                           iv_size + payload_len;
+
+                  sym_cop->auth.digest.data = rte_pktmbuf_mtod_offset(mb0, void*,
+                           rte_pktmbuf_pkt_len(mb0) - icv_size);
+                  sym_cop->auth.digest.phys_addr = rte_pktmbuf_mtophys_offset(mb0,
+                           rte_pktmbuf_pkt_len(mb0) - icv_size);
+                  sym_cop->auth.digest.length = icv_size;
+                }
+            }
+
+trace:
+	  if (PREDICT_FALSE(b0->flags & VLIB_BUFFER_IS_TRACED))
+	    {
+	      esp_decrypt_trace_t *tr = vlib_add_trace (vm, node, b0, sizeof (*tr));
+	      tr->crypto_alg = sa0->crypto_alg;
+	      tr->integ_alg = sa0->integ_alg;
+	    }
+	}
+      vlib_put_next_frame (vm, node, next_index, n_left_to_next);
+    }
+  vlib_node_increment_counter (vm, dpdk_esp_decrypt_node.index,
+			       ESP_DECRYPT_ERROR_RX_PKTS,
+			       from_frame->n_vectors);
+  crypto_qp_data_t *qpd;
+  /* *INDENT-OFF* */
+  vec_foreach_index (i, cwm->qp_data)
+    {
+      u32 enq;
+
+      qpd = vec_elt_at_index(cwm->qp_data, i);
+      enq = rte_cryptodev_enqueue_burst(qpd->dev_id, qpd->qp_id,
+					qpd->cops, n_cop_qp[i]);
+      qpd->inflights += enq;
+
+      if (PREDICT_FALSE(enq < n_cop_qp[i]))
+	{
+	  crypto_free_cop (qpd, &qpd->cops[enq], n_cop_qp[i] - enq);
+	  vlib_buffer_free (vm, &qpd->bi[enq], n_cop_qp[i] - enq);
+
+	  vlib_node_increment_counter (vm, dpdk_esp_decrypt_node.index,
+				       ESP_DECRYPT_ERROR_ENQ_FAIL,
+				       n_cop_qp[i] - enq);
+	}
+    }
+  /* *INDENT-ON* */
+
+  return from_frame->n_vectors;
+}
+
+VLIB_REGISTER_NODE (dpdk_esp_decrypt_node) = {
+  .function = dpdk_esp_decrypt_node_fn,
+  .name = "dpdk-esp-decrypt",
+  .vector_size = sizeof (u32),
+  .format_trace = format_esp_decrypt_trace,
+  .type = VLIB_NODE_TYPE_INTERNAL,
+
+  .n_errors = ARRAY_LEN(esp_decrypt_error_strings),
+  .error_strings = esp_decrypt_error_strings,
+
+  .n_next_nodes = ESP_DECRYPT_N_NEXT,
+  .next_nodes = {
+#define _(s,n) [ESP_DECRYPT_NEXT_##s] = n,
+    foreach_esp_decrypt_next
+#undef _
+  },
+};
+
+VLIB_NODE_FUNCTION_MULTIARCH (dpdk_esp_decrypt_node, dpdk_esp_decrypt_node_fn)
+
+/*
+ * Decrypt Post Node
+ */
+
+#define foreach_esp_decrypt_post_error	      \
+ _(PKTS, "ESP post pkts")
+
+typedef enum {
+#define _(sym,str) ESP_DECRYPT_POST_ERROR_##sym,
+  foreach_esp_decrypt_post_error
+#undef _
+  ESP_DECRYPT_POST_N_ERROR,
+} esp_decrypt_post_error_t;
+
+static char * esp_decrypt_post_error_strings[] = {
+#define _(sym,string) string,
+  foreach_esp_decrypt_post_error
+#undef _
+};
+
+vlib_node_registration_t dpdk_esp_decrypt_post_node;
+
+static u8 * format_esp_decrypt_post_trace (u8 * s, va_list * args)
+{
+  return s;
+}
+
+static uword
+dpdk_esp_decrypt_post_node_fn (vlib_main_t * vm,
+	     vlib_node_runtime_t * node,
+	     vlib_frame_t * from_frame)
+{
+  u32 n_left_from, *from, *to_next = 0, next_index;
+  ipsec_sa_t * sa0;
+  u32 sa_index0 = ~0;
+  ipsec_main_t *im = &ipsec_main;
+  dpdk_esp_main_t *em = &dpdk_esp_main;
+
+  from = vlib_frame_vector_args (from_frame);
+  n_left_from = from_frame->n_vectors;
+
+  next_index = node->cached_next_index;
+
+  while (n_left_from > 0)
+    {
+      u32 n_left_to_next;
+
+      vlib_get_next_frame (vm, node, next_index, to_next, n_left_to_next);
+
+      while (n_left_from > 0 && n_left_to_next > 0)
+	{
+	  esp_footer_t * f0;
+	  u32 bi0, next0, icv_size, iv_size;
+	  vlib_buffer_t * b0 = 0;
+	  ip4_header_t *ih4 = 0, *oh4 = 0;
+	  ip6_header_t *ih6 = 0, *oh6 = 0;
+	  u8 tunnel_mode = 1;
+	  u8 transport_ip6 = 0;
+
+	  next0 = ESP_DECRYPT_NEXT_DROP;
+
+	  bi0 = from[0];
+	  from += 1;
+	  n_left_from -= 1;
+	  n_left_to_next -= 1;
+
+	  b0 = vlib_get_buffer (vm, bi0);
+
+	  sa_index0 = vnet_buffer(b0)->ipsec.sad_index;
+	  sa0 = pool_elt_at_index (im->sad, sa_index0);
+
+	  to_next[0] = bi0;
+	  to_next += 1;
+
+	  icv_size = em->esp_integ_algs[sa0->integ_alg].trunc_size;
+	  iv_size = em->esp_crypto_algs[sa0->crypto_alg].iv_len;
+
+	  if (sa0->use_anti_replay)
+	    {
+	      esp_header_t * esp0 = vlib_buffer_get_current (b0);
+	      u32 seq;
+	      seq = clib_host_to_net_u32(esp0->seq);
+	      if (PREDICT_TRUE(sa0->use_esn))
+		esp_replay_advance_esn(sa0, seq);
+	      else
+		esp_replay_advance(sa0, seq);
+	    }
+
+	  ih4 = (ip4_header_t *) (b0->data + sizeof(ethernet_header_t));
+	  vlib_buffer_advance (b0, sizeof (esp_header_t) + iv_size);
+
+	  b0->current_length -= (icv_size + 2);
+	  b0->flags = VLIB_BUFFER_TOTAL_LENGTH_VALID;
+	  f0 = (esp_footer_t *) ((u8 *) vlib_buffer_get_current (b0) +
+				 b0->current_length);
+	  b0->current_length -= f0->pad_length;
+
+	  /* transport mode */
+	  if (PREDICT_FALSE(!sa0->is_tunnel && !sa0->is_tunnel_ip6))
+	    {
+	      tunnel_mode = 0;
+
+	      if (PREDICT_TRUE((ih4->ip_version_and_header_length & 0xF0) != 0x40))
+		{
+		  if (PREDICT_TRUE((ih4->ip_version_and_header_length & 0xF0) == 0x60))
+		    transport_ip6 = 1;
+		  else
+		    {
+		      clib_warning("next header: 0x%x", f0->next_header);
+		      vlib_node_increment_counter (vm, dpdk_esp_decrypt_node.index,
+						   ESP_DECRYPT_ERROR_NOT_IP, 1);
+		      goto trace;
+		    }
+		}
+	    }
+
+	  if (PREDICT_TRUE (tunnel_mode))
+	    {
+	      if (PREDICT_TRUE(f0->next_header == IP_PROTOCOL_IP_IN_IP))
+		next0 = ESP_DECRYPT_NEXT_IP4_INPUT;
+	      else if (f0->next_header == IP_PROTOCOL_IPV6)
+		next0 = ESP_DECRYPT_NEXT_IP6_INPUT;
+	      else
+		{
+		  clib_warning("next header: 0x%x", f0->next_header);
+		  vlib_node_increment_counter (vm, dpdk_esp_decrypt_node.index,
+					       ESP_DECRYPT_ERROR_DECRYPTION_FAILED,
+					       1);
+		  goto trace;
+		}
+	    }
+	  /* transport mode */
+	  else
+	    {
+	      if (PREDICT_FALSE(transport_ip6))
+		{
+		  ih6 = (ip6_header_t *) (b0->data + sizeof(ethernet_header_t));
+		  vlib_buffer_advance (b0, -sizeof(ip6_header_t));
+		  oh6 = vlib_buffer_get_current (b0);
+		  memmove(oh6, ih6, sizeof(ip6_header_t));
+
+		  next0 = ESP_DECRYPT_NEXT_IP6_INPUT;
+		  oh6->protocol = f0->next_header;
+		  oh6->payload_length =
+		      clib_host_to_net_u16 (
+			  vlib_buffer_length_in_chain(vm, b0) -
+			  sizeof (ip6_header_t));
+		}
+	      else
+		{
+		  vlib_buffer_advance (b0, -sizeof(ip4_header_t));
+		  oh4 = vlib_buffer_get_current (b0);
+		  memmove(oh4, ih4, sizeof(ip4_header_t));
+
+		  next0 = ESP_DECRYPT_NEXT_IP4_INPUT;
+		  oh4->ip_version_and_header_length = 0x45;
+		  oh4->fragment_id = 0;
+		  oh4->flags_and_fragment_offset = 0;
+		  oh4->protocol = f0->next_header;
+		  oh4->length = clib_host_to_net_u16 (
+		      vlib_buffer_length_in_chain (vm, b0));
+		  oh4->checksum = ip4_header_checksum (oh4);
+		}
+	    }
+
+	  vnet_buffer (b0)->sw_if_index[VLIB_TX] = (u32)~0;
+
+trace:
+	  if (PREDICT_FALSE(b0->flags & VLIB_BUFFER_IS_TRACED))
+	    {
+	      esp_decrypt_trace_t *tr = vlib_add_trace (vm, node, b0, sizeof (*tr));
+	      tr->crypto_alg = sa0->crypto_alg;
+	      tr->integ_alg = sa0->integ_alg;
+	    }
+
+	  vlib_validate_buffer_enqueue_x1 (vm, node, next_index,
+					   to_next, n_left_to_next, bi0, next0);
+	}
+      vlib_put_next_frame (vm, node, next_index, n_left_to_next);
+    }
+  vlib_node_increment_counter (vm, dpdk_esp_decrypt_post_node.index,
+			       ESP_DECRYPT_POST_ERROR_PKTS,
+			       from_frame->n_vectors);
+
+  return from_frame->n_vectors;
+}
+
+VLIB_REGISTER_NODE (dpdk_esp_decrypt_post_node) = {
+  .function = dpdk_esp_decrypt_post_node_fn,
+  .name = "dpdk-esp-decrypt-post",
+  .vector_size = sizeof (u32),
+  .format_trace = format_esp_decrypt_post_trace,
+  .type = VLIB_NODE_TYPE_INTERNAL,
+
+  .n_errors = ARRAY_LEN(esp_decrypt_post_error_strings),
+  .error_strings = esp_decrypt_post_error_strings,
+
+  .n_next_nodes = ESP_DECRYPT_N_NEXT,
+  .next_nodes = {
+#define _(s,n) [ESP_DECRYPT_NEXT_##s] = n,
+    foreach_esp_decrypt_next
+#undef _
+  },
+};
+
+VLIB_NODE_FUNCTION_MULTIARCH (dpdk_esp_decrypt_post_node, dpdk_esp_decrypt_post_node_fn)
diff --git a/src/plugins/dpdk/ipsec/esp_encrypt.c b/src/plugins/dpdk/ipsec/esp_encrypt.c
new file mode 100644
index 00000000..5b03de73
--- /dev/null
+++ b/src/plugins/dpdk/ipsec/esp_encrypt.c
@@ -0,0 +1,609 @@
+/*
+ * esp_encrypt.c : IPSec ESP encrypt node using DPDK Cryptodev
+ *
+ * Copyright (c) 2016 Intel and/or its affiliates.
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at:
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#include 
+#include 
+#include 
+
+#include 
+#include 
+#include 
+#include 
+#include 
+
+#define foreach_esp_encrypt_next                   \
+_(DROP, "error-drop")                              \
+_(IP4_LOOKUP, "ip4-lookup")                        \
+_(IP6_LOOKUP, "ip6-lookup")                        \
+_(INTERFACE_OUTPUT, "interface-output")
+
+#define _(v, s) ESP_ENCRYPT_NEXT_##v,
+typedef enum
+{
+  foreach_esp_encrypt_next
+#undef _
+    ESP_ENCRYPT_N_NEXT,
+} esp_encrypt_next_t;
+
+#define foreach_esp_encrypt_error                   \
+ _(RX_PKTS, "ESP pkts received")                    \
+ _(SEQ_CYCLED, "sequence number cycled")            \
+ _(ENQ_FAIL, "Enqueue failed (buffer full)")        \
+ _(NO_CRYPTODEV, "Cryptodev not configured")        \
+ _(UNSUPPORTED, "Cipher/Auth not supported")
+
+
+typedef enum
+{
+#define _(sym,str) ESP_ENCRYPT_ERROR_##sym,
+  foreach_esp_encrypt_error
+#undef _
+    ESP_ENCRYPT_N_ERROR,
+} esp_encrypt_error_t;
+
+static char *esp_encrypt_error_strings[] = {
+#define _(sym,string) string,
+  foreach_esp_encrypt_error
+#undef _
+};
+
+vlib_node_registration_t dpdk_esp_encrypt_node;
+
+typedef struct
+{
+  u32 spi;
+  u32 seq;
+  ipsec_crypto_alg_t crypto_alg;
+  ipsec_integ_alg_t integ_alg;
+} esp_encrypt_trace_t;
+
+/* packet trace format function */
+static u8 *
+format_esp_encrypt_trace (u8 * s, va_list * args)
+{
+  CLIB_UNUSED (vlib_main_t * vm) = va_arg (*args, vlib_main_t *);
+  CLIB_UNUSED (vlib_node_t * node) = va_arg (*args, vlib_node_t *);
+  esp_encrypt_trace_t *t = va_arg (*args, esp_encrypt_trace_t *);
+
+  s = format (s, "esp: spi %u seq %u crypto %U integrity %U",
+	      t->spi, t->seq,
+	      format_ipsec_crypto_alg, t->crypto_alg,
+	      format_ipsec_integ_alg, t->integ_alg);
+  return s;
+}
+
+static uword
+dpdk_esp_encrypt_node_fn (vlib_main_t * vm,
+			  vlib_node_runtime_t * node,
+			  vlib_frame_t * from_frame)
+{
+  u32 n_left_from, *from, *to_next, next_index;
+  ipsec_main_t *im = &ipsec_main;
+  u32 cpu_index = os_get_cpu_number ();
+  dpdk_crypto_main_t *dcm = &dpdk_crypto_main;
+  dpdk_esp_main_t *em = &dpdk_esp_main;
+  u32 i;
+
+  from = vlib_frame_vector_args (from_frame);
+  n_left_from = from_frame->n_vectors;
+
+  if (PREDICT_FALSE (!dcm->workers_main))
+    {
+      /* Likely there are not enough cryptodevs, so drop frame */
+      vlib_node_increment_counter (vm, dpdk_esp_encrypt_node.index,
+				   ESP_ENCRYPT_ERROR_NO_CRYPTODEV,
+				   n_left_from);
+      vlib_buffer_free (vm, from, n_left_from);
+      return n_left_from;
+    }
+
+  crypto_worker_main_t *cwm = vec_elt_at_index (dcm->workers_main, cpu_index);
+  u32 n_qps = vec_len (cwm->qp_data);
+  struct rte_crypto_op **cops_to_enq[n_qps];
+  u32 n_cop_qp[n_qps], *bi_to_enq[n_qps];
+
+  for (i = 0; i < n_qps; i++)
+    {
+      bi_to_enq[i] = cwm->qp_data[i].bi;
+      cops_to_enq[i] = cwm->qp_data[i].cops;
+    }
+
+  memset (n_cop_qp, 0, n_qps * sizeof (u32));
+
+  crypto_alloc_cops ();
+
+  next_index = ESP_ENCRYPT_NEXT_DROP;
+
+  while (n_left_from > 0)
+    {
+      u32 n_left_to_next;
+
+      vlib_get_next_frame (vm, node, next_index, to_next, n_left_to_next);
+
+      while (n_left_from > 0 && n_left_to_next > 0)
+	{
+	  u32 bi0, next0;
+	  vlib_buffer_t *b0 = 0;
+	  u32 sa_index0;
+	  ipsec_sa_t *sa0;
+	  ip4_and_esp_header_t *ih0, *oh0 = 0;
+	  ip6_and_esp_header_t *ih6_0, *oh6_0 = 0;
+	  struct rte_mbuf *mb0 = 0;
+	  esp_footer_t *f0;
+	  u8 is_ipv6;
+	  u8 ip_hdr_size;
+	  u8 next_hdr_type;
+	  u8 transport_mode = 0;
+	  const int BLOCK_SIZE = 16;
+	  u32 iv_size;
+	  u16 orig_sz;
+	  crypto_sa_session_t *sa_sess;
+	  void *sess;
+	  struct rte_crypto_op *cop = 0;
+	  u16 qp_index;
+
+	  bi0 = from[0];
+	  from += 1;
+	  n_left_from -= 1;
+
+	  b0 = vlib_get_buffer (vm, bi0);
+	  sa_index0 = vnet_buffer (b0)->ipsec.sad_index;
+	  sa0 = pool_elt_at_index (im->sad, sa_index0);
+
+	  if (PREDICT_FALSE (esp_seq_advance (sa0)))
+	    {
+	      clib_warning ("sequence number counter has cycled SPI %u",
+			    sa0->spi);
+	      vlib_node_increment_counter (vm, dpdk_esp_encrypt_node.index,
+					   ESP_ENCRYPT_ERROR_SEQ_CYCLED, 1);
+	      //TODO: rekey SA
+	      to_next[0] = bi0;
+	      to_next += 1;
+	      n_left_to_next -= 1;
+	      goto trace;
+	    }
+
+	  sa0->total_data_size += b0->current_length;
+
+	  sa_sess = pool_elt_at_index (cwm->sa_sess_d[1], sa_index0);
+	  if (PREDICT_FALSE (!sa_sess->sess))
+	    {
+	      int ret = create_sym_sess (sa0, sa_sess, 1);
+
+	      if (PREDICT_FALSE (ret))
+		{
+		  to_next[0] = bi0;
+		  to_next += 1;
+		  n_left_to_next -= 1;
+		  goto trace;
+		}
+	    }
+
+	  qp_index = sa_sess->qp_index;
+	  sess = sa_sess->sess;
+
+	  ASSERT (vec_len (vec_elt (cwm->qp_data, qp_index).free_cops) > 0);
+	  cop = vec_pop (vec_elt (cwm->qp_data, qp_index).free_cops);
+	  ASSERT (cop->status == RTE_CRYPTO_OP_STATUS_NOT_PROCESSED);
+
+	  cops_to_enq[qp_index][0] = cop;
+	  cops_to_enq[qp_index] += 1;
+	  n_cop_qp[qp_index] += 1;
+	  bi_to_enq[qp_index][0] = bi0;
+	  bi_to_enq[qp_index] += 1;
+
+	  ssize_t adv;
+	  iv_size = em->esp_crypto_algs[sa0->crypto_alg].iv_len;
+	  ih0 = vlib_buffer_get_current (b0);
+	  orig_sz = b0->current_length;
+	  is_ipv6 = (ih0->ip4.ip_version_and_header_length & 0xF0) == 0x60;
+	  /* is ipv6 */
+	  if (PREDICT_TRUE (sa0->is_tunnel))
+	    {
+	      if (PREDICT_TRUE (!is_ipv6))
+		adv = -sizeof (ip4_and_esp_header_t);
+	      else
+		adv = -sizeof (ip6_and_esp_header_t);
+	    }
+	  else
+	    {
+	      adv = -sizeof (esp_header_t);
+	      if (PREDICT_TRUE (!is_ipv6))
+		orig_sz -= sizeof (ip4_header_t);
+	      else
+		orig_sz -= sizeof (ip6_header_t);
+	    }
+
+	  /*transport mode save the eth header before it is overwritten */
+	  if (PREDICT_FALSE (!sa0->is_tunnel))
+	    {
+	      ethernet_header_t *ieh0 = (ethernet_header_t *)
+		((u8 *) vlib_buffer_get_current (b0) -
+		 sizeof (ethernet_header_t));
+	      ethernet_header_t *oeh0 =
+		(ethernet_header_t *) ((u8 *) ieh0 + (adv - iv_size));
+	      clib_memcpy (oeh0, ieh0, sizeof (ethernet_header_t));
+	    }
+
+	  vlib_buffer_advance (b0, adv - iv_size);
+
+	  /* XXX IP6/ip4 and IP4/IP6 not supported, only IP4/IP4 and IP6/IP6 */
+
+	  /* is ipv6 */
+	  if (PREDICT_FALSE (is_ipv6))
+	    {
+	      ih6_0 = (ip6_and_esp_header_t *) ih0;
+	      ip_hdr_size = sizeof (ip6_header_t);
+	      oh6_0 = vlib_buffer_get_current (b0);
+
+	      if (PREDICT_TRUE (sa0->is_tunnel))
+		{
+		  next_hdr_type = IP_PROTOCOL_IPV6;
+		  oh6_0->ip6.ip_version_traffic_class_and_flow_label =
+		    ih6_0->ip6.ip_version_traffic_class_and_flow_label;
+		}
+	      else
+		{
+		  next_hdr_type = ih6_0->ip6.protocol;
+		  memmove (oh6_0, ih6_0, sizeof (ip6_header_t));
+		}
+
+	      oh6_0->ip6.protocol = IP_PROTOCOL_IPSEC_ESP;
+	      oh6_0->ip6.hop_limit = 254;
+	      oh6_0->esp.spi = clib_net_to_host_u32 (sa0->spi);
+	      oh6_0->esp.seq = clib_net_to_host_u32 (sa0->seq);
+	    }
+	  else
+	    {
+	      ip_hdr_size = sizeof (ip4_header_t);
+	      oh0 = vlib_buffer_get_current (b0);
+
+	      if (PREDICT_TRUE (sa0->is_tunnel))
+		{
+		  next_hdr_type = IP_PROTOCOL_IP_IN_IP;
+		  oh0->ip4.tos = ih0->ip4.tos;
+		}
+	      else
+		{
+		  next_hdr_type = ih0->ip4.protocol;
+		  memmove (oh0, ih0, sizeof (ip4_header_t));
+		}
+
+	      oh0->ip4.ip_version_and_header_length = 0x45;
+	      oh0->ip4.fragment_id = 0;
+	      oh0->ip4.flags_and_fragment_offset = 0;
+	      oh0->ip4.ttl = 254;
+	      oh0->ip4.protocol = IP_PROTOCOL_IPSEC_ESP;
+	      oh0->esp.spi = clib_net_to_host_u32 (sa0->spi);
+	      oh0->esp.seq = clib_net_to_host_u32 (sa0->seq);
+	    }
+
+	  if (PREDICT_TRUE (sa0->is_tunnel && !sa0->is_tunnel_ip6))
+	    {
+	      oh0->ip4.src_address.as_u32 = sa0->tunnel_src_addr.ip4.as_u32;
+	      oh0->ip4.dst_address.as_u32 = sa0->tunnel_dst_addr.ip4.as_u32;
+
+	      /* in tunnel mode send it back to FIB */
+	      next0 = ESP_ENCRYPT_NEXT_IP4_LOOKUP;
+	      vnet_buffer (b0)->sw_if_index[VLIB_TX] = (u32) ~ 0;
+	    }
+	  else if (sa0->is_tunnel && sa0->is_tunnel_ip6)
+	    {
+	      oh6_0->ip6.src_address.as_u64[0] =
+		sa0->tunnel_src_addr.ip6.as_u64[0];
+	      oh6_0->ip6.src_address.as_u64[1] =
+		sa0->tunnel_src_addr.ip6.as_u64[1];
+	      oh6_0->ip6.dst_address.as_u64[0] =
+		sa0->tunnel_dst_addr.ip6.as_u64[0];
+	      oh6_0->ip6.dst_address.as_u64[1] =
+		sa0->tunnel_dst_addr.ip6.as_u64[1];
+
+	      /* in tunnel mode send it back to FIB */
+	      next0 = ESP_ENCRYPT_NEXT_IP6_LOOKUP;
+	      vnet_buffer (b0)->sw_if_index[VLIB_TX] = (u32) ~ 0;
+	    }
+	  else
+	    {
+	      next0 = ESP_ENCRYPT_NEXT_INTERFACE_OUTPUT;
+	      transport_mode = 1;
+	    }
+
+	  ASSERT (sa0->crypto_alg < IPSEC_CRYPTO_N_ALG);
+	  ASSERT (sa0->crypto_alg != IPSEC_CRYPTO_ALG_NONE);
+
+	  int blocks = 1 + (orig_sz + 1) / BLOCK_SIZE;
+
+	  /* pad packet in input buffer */
+	  u8 pad_bytes = BLOCK_SIZE * blocks - 2 - orig_sz;
+	  u8 i;
+	  u8 *padding = vlib_buffer_get_current (b0) + b0->current_length;
+
+	  for (i = 0; i < pad_bytes; ++i)
+	    padding[i] = i + 1;
+
+	  f0 = vlib_buffer_get_current (b0) + b0->current_length + pad_bytes;
+	  f0->pad_length = pad_bytes;
+	  f0->next_header = next_hdr_type;
+	  b0->current_length += pad_bytes + 2 +
+	    em->esp_integ_algs[sa0->integ_alg].trunc_size;
+
+	  vnet_buffer (b0)->sw_if_index[VLIB_RX] =
+	    vnet_buffer (b0)->sw_if_index[VLIB_RX];
+	  b0->flags |= VLIB_BUFFER_TOTAL_LENGTH_VALID;
+
+	  struct rte_crypto_sym_op *sym_cop;
+	  sym_cop = (struct rte_crypto_sym_op *) (cop + 1);
+
+	  dpdk_cop_priv_t *priv = (dpdk_cop_priv_t *) (sym_cop + 1);
+
+	  vnet_buffer (b0)->unused[0] = next0;
+
+	  mb0 = rte_mbuf_from_vlib_buffer (b0);
+	  mb0->data_len = b0->current_length;
+	  mb0->pkt_len = b0->current_length;
+	  mb0->data_off = RTE_PKTMBUF_HEADROOM + b0->current_data;
+
+	  rte_crypto_op_attach_sym_session (cop, sess);
+
+	  sym_cop->m_src = mb0;
+
+	  dpdk_gcm_cnt_blk *icb = &priv->cb;
+	  icb->salt = sa0->salt;
+	  icb->iv[0] = sa0->seq;
+	  icb->iv[1] = sa0->seq_hi;
+
+	  if (sa0->crypto_alg == IPSEC_CRYPTO_ALG_AES_GCM_128)
+	    {
+	      icb->cnt = clib_host_to_net_u32 (1);
+	      clib_memcpy (vlib_buffer_get_current (b0) + ip_hdr_size +
+			   sizeof (esp_header_t), icb->iv, 8);
+	      sym_cop->cipher.data.offset =
+		ip_hdr_size + sizeof (esp_header_t) + iv_size;
+	      sym_cop->cipher.data.length = BLOCK_SIZE * blocks;
+	      sym_cop->cipher.iv.length = 16;
+	    }
+	  else
+	    {
+	      sym_cop->cipher.data.offset =
+		ip_hdr_size + sizeof (esp_header_t);
+	      sym_cop->cipher.data.length = BLOCK_SIZE * blocks + iv_size;
+	      sym_cop->cipher.iv.length = iv_size;
+	    }
+
+	  sym_cop->cipher.iv.data = (u8 *) icb;
+	  sym_cop->cipher.iv.phys_addr = cop->phys_addr + (uintptr_t) icb
+	    - (uintptr_t) cop;
+
+
+	  ASSERT (sa0->integ_alg < IPSEC_INTEG_N_ALG);
+	  ASSERT (sa0->integ_alg != IPSEC_INTEG_ALG_NONE);
+
+	  if (PREDICT_FALSE (sa0->integ_alg == IPSEC_INTEG_ALG_AES_GCM_128))
+	    {
+	      u8 *aad = priv->aad;
+	      clib_memcpy (aad, vlib_buffer_get_current (b0) + ip_hdr_size,
+			   8);
+	      sym_cop->auth.aad.data = aad;
+	      sym_cop->auth.aad.phys_addr = cop->phys_addr +
+		(uintptr_t) aad - (uintptr_t) cop;
+
+	      if (PREDICT_FALSE (sa0->use_esn))
+		{
+		  *((u32 *) & aad[8]) = sa0->seq_hi;
+		  sym_cop->auth.aad.length = 12;
+		}
+	      else
+		{
+		  sym_cop->auth.aad.length = 8;
+		}
+	    }
+	  else
+	    {
+	      sym_cop->auth.data.offset = ip_hdr_size;
+	      sym_cop->auth.data.length = b0->current_length - ip_hdr_size
+		- em->esp_integ_algs[sa0->integ_alg].trunc_size;
+
+	      if (PREDICT_FALSE (sa0->use_esn))
+		{
+		  u8 *payload_end =
+		    vlib_buffer_get_current (b0) + b0->current_length;
+		  *((u32 *) payload_end) = sa0->seq_hi;
+		  sym_cop->auth.data.length += sizeof (sa0->seq_hi);
+		}
+	    }
+	  sym_cop->auth.digest.data = vlib_buffer_get_current (b0) +
+	    b0->current_length -
+	    em->esp_integ_algs[sa0->integ_alg].trunc_size;
+	  sym_cop->auth.digest.phys_addr = rte_pktmbuf_mtophys_offset (mb0,
+								       b0->current_length
+								       -
+								       em->esp_integ_algs
+								       [sa0->integ_alg].trunc_size);
+	  sym_cop->auth.digest.length =
+	    em->esp_integ_algs[sa0->integ_alg].trunc_size;
+
+
+	  if (PREDICT_FALSE (is_ipv6))
+	    {
+	      oh6_0->ip6.payload_length =
+		clib_host_to_net_u16 (vlib_buffer_length_in_chain (vm, b0) -
+				      sizeof (ip6_header_t));
+	    }
+	  else
+	    {
+	      oh0->ip4.length =
+		clib_host_to_net_u16 (vlib_buffer_length_in_chain (vm, b0));
+	      oh0->ip4.checksum = ip4_header_checksum (&oh0->ip4);
+	    }
+
+	  if (transport_mode)
+	    vlib_buffer_advance (b0, -sizeof (ethernet_header_t));
+
+	trace:
+	  if (PREDICT_FALSE (b0->flags & VLIB_BUFFER_IS_TRACED))
+	    {
+	      esp_encrypt_trace_t *tr =
+		vlib_add_trace (vm, node, b0, sizeof (*tr));
+	      tr->spi = sa0->spi;
+	      tr->seq = sa0->seq - 1;
+	      tr->crypto_alg = sa0->crypto_alg;
+	      tr->integ_alg = sa0->integ_alg;
+	    }
+	}
+      vlib_put_next_frame (vm, node, next_index, n_left_to_next);
+    }
+  vlib_node_increment_counter (vm, dpdk_esp_encrypt_node.index,
+			       ESP_ENCRYPT_ERROR_RX_PKTS,
+			       from_frame->n_vectors);
+  crypto_qp_data_t *qpd;
+  /* *INDENT-OFF* */
+  vec_foreach_index (i, cwm->qp_data)
+    {
+      u32 enq;
+
+      qpd = vec_elt_at_index(cwm->qp_data, i);
+      enq = rte_cryptodev_enqueue_burst(qpd->dev_id, qpd->qp_id,
+					qpd->cops, n_cop_qp[i]);
+      qpd->inflights += enq;
+
+      if (PREDICT_FALSE(enq < n_cop_qp[i]))
+	{
+	  crypto_free_cop (qpd, &qpd->cops[enq], n_cop_qp[i] - enq);
+	  vlib_buffer_free (vm, &qpd->bi[enq], n_cop_qp[i] - enq);
+
+          vlib_node_increment_counter (vm, dpdk_esp_encrypt_node.index,
+				       ESP_ENCRYPT_ERROR_ENQ_FAIL,
+				       n_cop_qp[i] - enq);
+        }
+    }
+  /* *INDENT-ON* */
+
+  return from_frame->n_vectors;
+}
+
+VLIB_REGISTER_NODE (dpdk_esp_encrypt_node) =
+{
+  .function = dpdk_esp_encrypt_node_fn,.name = "dpdk-esp-encrypt",.flags =
+    VLIB_NODE_FLAG_IS_OUTPUT,.vector_size = sizeof (u32),.format_trace =
+    format_esp_encrypt_trace,.n_errors =
+    ARRAY_LEN (esp_encrypt_error_strings),.error_strings =
+    esp_encrypt_error_strings,.n_next_nodes = 1,.next_nodes =
+  {
+  [ESP_ENCRYPT_NEXT_DROP] = "error-drop",}
+};
+
+VLIB_NODE_FUNCTION_MULTIARCH (dpdk_esp_encrypt_node, dpdk_esp_encrypt_node_fn)
+/*
+ * ESP Encrypt Post Node
+ */
+#define foreach_esp_encrypt_post_error              \
+ _(PKTS, "ESP post pkts")
+     typedef enum
+     {
+#define _(sym,str) ESP_ENCRYPT_POST_ERROR_##sym,
+       foreach_esp_encrypt_post_error
+#undef _
+	 ESP_ENCRYPT_POST_N_ERROR,
+     } esp_encrypt_post_error_t;
+
+     static char *esp_encrypt_post_error_strings[] = {
+#define _(sym,string) string,
+       foreach_esp_encrypt_post_error
+#undef _
+     };
+
+vlib_node_registration_t dpdk_esp_encrypt_post_node;
+
+static u8 *
+format_esp_encrypt_post_trace (u8 * s, va_list * args)
+{
+  return s;
+}
+
+static uword
+dpdk_esp_encrypt_post_node_fn (vlib_main_t * vm,
+			       vlib_node_runtime_t * node,
+			       vlib_frame_t * from_frame)
+{
+  u32 n_left_from, *from, *to_next = 0, next_index;
+
+  from = vlib_frame_vector_args (from_frame);
+  n_left_from = from_frame->n_vectors;
+
+  next_index = node->cached_next_index;
+
+  while (n_left_from > 0)
+    {
+      u32 n_left_to_next;
+
+      vlib_get_next_frame (vm, node, next_index, to_next, n_left_to_next);
+
+      while (n_left_from > 0 && n_left_to_next > 0)
+	{
+	  u32 bi0, next0;
+	  vlib_buffer_t *b0 = 0;
+
+	  bi0 = from[0];
+	  from += 1;
+	  n_left_from -= 1;
+	  n_left_to_next -= 1;
+
+	  b0 = vlib_get_buffer (vm, bi0);
+
+	  to_next[0] = bi0;
+	  to_next += 1;
+
+	  next0 = vnet_buffer (b0)->unused[0];
+
+	  vlib_validate_buffer_enqueue_x1 (vm, node, next_index,
+					   to_next, n_left_to_next, bi0,
+					   next0);
+	}
+      vlib_put_next_frame (vm, node, next_index, n_left_to_next);
+    }
+
+  vlib_node_increment_counter (vm, dpdk_esp_encrypt_post_node.index,
+			       ESP_ENCRYPT_POST_ERROR_PKTS,
+			       from_frame->n_vectors);
+
+  return from_frame->n_vectors;
+}
+
+VLIB_REGISTER_NODE (dpdk_esp_encrypt_post_node) =
+{
+  .function = dpdk_esp_encrypt_post_node_fn,.name =
+    "dpdk-esp-encrypt-post",.vector_size = sizeof (u32),.format_trace =
+    format_esp_encrypt_post_trace,.type = VLIB_NODE_TYPE_INTERNAL,.n_errors =
+    ARRAY_LEN (esp_encrypt_post_error_strings),.error_strings =
+    esp_encrypt_post_error_strings,.n_next_nodes =
+    ESP_ENCRYPT_N_NEXT,.next_nodes =
+  {
+#define _(s,n) [ESP_ENCRYPT_NEXT_##s] = n,
+    foreach_esp_encrypt_next
+#undef _
+  }
+};
+
+VLIB_NODE_FUNCTION_MULTIARCH (dpdk_esp_encrypt_post_node,
+			      dpdk_esp_encrypt_post_node_fn)
+/*
+ * fd.io coding-style-patch-verification: ON
+ *
+ * Local Variables:
+ * eval: (c-set-style "gnu")
+ * End:
+ */
diff --git a/src/plugins/dpdk/ipsec/ipsec.c b/src/plugins/dpdk/ipsec/ipsec.c
new file mode 100644
index 00000000..16bec20a
--- /dev/null
+++ b/src/plugins/dpdk/ipsec/ipsec.c
@@ -0,0 +1,430 @@
+/*
+ * Copyright (c) 2016 Intel and/or its affiliates.
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at:
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+#include 
+#include 
+#include 
+#include 
+#include 
+
+#include 
+#include 
+#include 
+
+#define DPDK_CRYPTO_NB_SESS_OBJS  20000
+#define DPDK_CRYPTO_CACHE_SIZE	  512
+#define DPDK_CRYPTO_PRIV_SIZE	  128
+#define DPDK_CRYPTO_N_QUEUE_DESC  1024
+#define DPDK_CRYPTO_NB_COPS	  (1024 * 4)
+
+static int
+add_del_sa_sess (u32 sa_index, u8 is_add)
+{
+  dpdk_crypto_main_t *dcm = &dpdk_crypto_main;
+  crypto_worker_main_t *cwm;
+  u8 skip_master = vlib_num_workers () > 0;
+
+  /* *INDENT-OFF* */
+  vec_foreach (cwm, dcm->workers_main)
+    {
+      crypto_sa_session_t *sa_sess;
+      u8 is_outbound;
+
+      if (skip_master)
+	{
+	  skip_master = 0;
+	  continue;
+	}
+
+      for (is_outbound = 0; is_outbound < 2; is_outbound++)
+	{
+	  if (is_add)
+	    {
+	      pool_get (cwm->sa_sess_d[is_outbound], sa_sess);
+	    }
+	  else
+	    {
+	      u8 dev_id;
+
+	      sa_sess = pool_elt_at_index (cwm->sa_sess_d[is_outbound], sa_index);
+	      dev_id = cwm->qp_data[sa_sess->qp_index].dev_id;
+
+	      if (!sa_sess->sess)
+		continue;
+
+	      if (rte_cryptodev_sym_session_free(dev_id, sa_sess->sess))
+		{
+		  clib_warning("failed to free session");
+		  return -1;
+		}
+	      memset(sa_sess, 0, sizeof(sa_sess[0]));
+	    }
+	}
+    }
+  /* *INDENT-OFF* */
+
+  return 0;
+}
+
+static void
+update_qp_data (crypto_worker_main_t * cwm,
+		u8 cdev_id, u16 qp_id, u8 is_outbound, u16 * idx)
+{
+  crypto_qp_data_t *qpd;
+
+  /* *INDENT-OFF* */
+  vec_foreach_index (*idx, cwm->qp_data)
+    {
+      qpd = vec_elt_at_index(cwm->qp_data, *idx);
+
+      if (qpd->dev_id == cdev_id && qpd->qp_id == qp_id &&
+	  qpd->is_outbound == is_outbound)
+	  return;
+    }
+  /* *INDENT-ON* */
+
+  vec_add2 (cwm->qp_data, qpd, 1);
+
+  qpd->dev_id = cdev_id;
+  qpd->qp_id = qp_id;
+  qpd->is_outbound = is_outbound;
+}
+
+/*
+ * return:
+ * 	0: already exist
+ * 	1: mapped
+ */
+static int
+add_mapping (crypto_worker_main_t * cwm,
+	     u8 cdev_id, u16 qp, u8 is_outbound,
+	     const struct rte_cryptodev_capabilities *cipher_cap,
+	     const struct rte_cryptodev_capabilities *auth_cap)
+{
+  u16 qp_index;
+  uword key = 0, data, *ret;
+  crypto_worker_qp_key_t *p_key = (crypto_worker_qp_key_t *) & key;
+
+  p_key->cipher_algo = (u8) cipher_cap->sym.cipher.algo;
+  p_key->auth_algo = (u8) auth_cap->sym.auth.algo;
+  p_key->is_outbound = is_outbound;
+
+  ret = hash_get (cwm->algo_qp_map, key);
+  if (ret)
+    return 0;
+
+  update_qp_data (cwm, cdev_id, qp, is_outbound, &qp_index);
+
+  data = (uword) qp_index;
+  hash_set (cwm->algo_qp_map, key, data);
+
+  return 1;
+}
+
+/*
+ * return:
+ * 	0: already exist
+ * 	1: mapped
+ */
+static int
+add_cdev_mapping (crypto_worker_main_t * cwm,
+		  struct rte_cryptodev_info *dev_info, u8 cdev_id,
+		  u16 qp, u8 is_outbound)
+{
+  const struct rte_cryptodev_capabilities *i, *j;
+  u32 mapped = 0;
+
+  for (i = dev_info->capabilities; i->op != RTE_CRYPTO_OP_TYPE_UNDEFINED; i++)
+    {
+      if (i->sym.xform_type != RTE_CRYPTO_SYM_XFORM_CIPHER)
+	continue;
+
+      if (check_algo_is_supported (i, NULL) != 0)
+	continue;
+
+      for (j = dev_info->capabilities; j->op != RTE_CRYPTO_OP_TYPE_UNDEFINED;
+	   j++)
+	{
+	  if (j->sym.xform_type != RTE_CRYPTO_SYM_XFORM_AUTH)
+	    continue;
+
+	  if (check_algo_is_supported (j, NULL) != 0)
+	    continue;
+
+	  mapped |= add_mapping (cwm, cdev_id, qp, is_outbound, i, j);
+	}
+    }
+
+  return mapped;
+}
+
+static int
+check_cryptodev_queues ()
+{
+  u32 n_qs = 0;
+  u8 cdev_id;
+  u32 n_req_qs = 2;
+
+  if (vlib_num_workers () > 0)
+    n_req_qs = vlib_num_workers () * 2;
+
+  for (cdev_id = 0; cdev_id < rte_cryptodev_count (); cdev_id++)
+    {
+      struct rte_cryptodev_info cdev_info;
+
+      rte_cryptodev_info_get (cdev_id, &cdev_info);
+
+      if (!
+	  (cdev_info.feature_flags & RTE_CRYPTODEV_FF_SYM_OPERATION_CHAINING))
+	continue;
+
+      n_qs += cdev_info.max_nb_queue_pairs;
+    }
+
+  if (n_qs >= n_req_qs)
+    return 0;
+  else
+    return -1;
+}
+
+static clib_error_t *
+dpdk_ipsec_check_support (ipsec_sa_t * sa)
+{
+  if (sa->crypto_alg == IPSEC_CRYPTO_ALG_AES_GCM_128)
+    {
+      if (sa->integ_alg != IPSEC_INTEG_ALG_NONE)
+	return clib_error_return (0, "unsupported integ-alg %U with "
+				  "crypto-algo aes-gcm-128",
+				  format_ipsec_integ_alg, sa->integ_alg);
+      sa->integ_alg = IPSEC_INTEG_ALG_AES_GCM_128;
+    }
+  else
+    {
+      if (sa->integ_alg == IPSEC_INTEG_ALG_NONE ||
+	  sa->integ_alg == IPSEC_INTEG_ALG_AES_GCM_128)
+	return clib_error_return (0, "unsupported integ-alg %U",
+				  format_ipsec_integ_alg, sa->integ_alg);
+    }
+
+  return 0;
+}
+
+static uword
+dpdk_ipsec_process (vlib_main_t * vm, vlib_node_runtime_t * rt,
+		    vlib_frame_t * f)
+{
+  dpdk_config_main_t *conf = &dpdk_config_main;
+  ipsec_main_t *im = &ipsec_main;
+  dpdk_crypto_main_t *dcm = &dpdk_crypto_main;
+  vlib_thread_main_t *tm = vlib_get_thread_main ();
+  struct rte_cryptodev_config dev_conf;
+  struct rte_cryptodev_qp_conf qp_conf;
+  struct rte_cryptodev_info cdev_info;
+  struct rte_mempool *rmp;
+  i32 dev_id, ret;
+  u32 i, skip_master;
+
+  if (!conf->cryptodev)
+    {
+      clib_warning ("DPDK Cryptodev support is disabled, "
+		    "default to OpenSSL IPsec");
+      return 0;
+    }
+
+  if (check_cryptodev_queues () < 0)
+    {
+      conf->cryptodev = 0;
+      clib_warning ("not enough Cryptodevs, default to OpenSSL IPsec");
+      return 0;
+    }
+
+  vec_alloc (dcm->workers_main, tm->n_vlib_mains);
+  _vec_len (dcm->workers_main) = tm->n_vlib_mains;
+
+  fprintf (stdout, "DPDK Cryptodevs info:\n");
+  fprintf (stdout, "dev_id\tn_qp\tnb_obj\tcache_size\n");
+  /* HW cryptodevs have higher dev_id, use HW first */
+  for (dev_id = rte_cryptodev_count () - 1; dev_id >= 0; dev_id--)
+    {
+      u16 max_nb_qp, qp = 0;
+      skip_master = vlib_num_workers () > 0;
+
+      rte_cryptodev_info_get (dev_id, &cdev_info);
+
+      if (!
+	  (cdev_info.feature_flags & RTE_CRYPTODEV_FF_SYM_OPERATION_CHAINING))
+	continue;
+
+      max_nb_qp = cdev_info.max_nb_queue_pairs;
+
+      for (i = 0; i < tm->n_vlib_mains; i++)
+	{
+	  u8 is_outbound;
+	  crypto_worker_main_t *cwm;
+	  uword *map;
+
+	  if (skip_master)
+	    {
+	      skip_master = 0;
+	      continue;
+	    }
+
+	  cwm = vec_elt_at_index (dcm->workers_main, i);
+	  map = cwm->algo_qp_map;
+
+	  if (!map)
+	    {
+	      map = hash_create (0, sizeof (crypto_worker_qp_key_t));
+	      if (!map)
+		{
+		  clib_warning ("unable to create hash table for worker %u",
+				vlib_mains[i]->cpu_index);
+		  goto error;
+		}
+	      cwm->algo_qp_map = map;
+	    }
+
+	  for (is_outbound = 0; is_outbound < 2 && qp < max_nb_qp;
+	       is_outbound++)
+	    qp += add_cdev_mapping (cwm, &cdev_info, dev_id, qp, is_outbound);
+	}
+
+      if (qp == 0)
+	continue;
+
+      dev_conf.socket_id = rte_cryptodev_socket_id (dev_id);
+      dev_conf.nb_queue_pairs = cdev_info.max_nb_queue_pairs;
+      dev_conf.session_mp.nb_objs = DPDK_CRYPTO_NB_SESS_OBJS;
+      dev_conf.session_mp.cache_size = DPDK_CRYPTO_CACHE_SIZE;
+
+      ret = rte_cryptodev_configure (dev_id, &dev_conf);
+      if (ret < 0)
+	{
+	  clib_warning ("cryptodev %u config error", dev_id);
+	  goto error;
+	}
+
+      qp_conf.nb_descriptors = DPDK_CRYPTO_N_QUEUE_DESC;
+      for (qp = 0; qp < dev_conf.nb_queue_pairs; qp++)
+	{
+	  ret = rte_cryptodev_queue_pair_setup (dev_id, qp, &qp_conf,
+						dev_conf.socket_id);
+	  if (ret < 0)
+	    {
+	      clib_warning ("cryptodev %u qp %u setup error", dev_id, qp);
+	      goto error;
+	    }
+	}
+      vec_validate_aligned (dcm->cop_pools, dev_conf.socket_id,
+			    CLIB_CACHE_LINE_BYTES);
+
+      if (!vec_elt (dcm->cop_pools, dev_conf.socket_id))
+	{
+	  u8 *pool_name = format (0, "crypto_op_pool_socket%u%c",
+				  dev_conf.socket_id, 0);
+
+	  rmp = rte_crypto_op_pool_create ((char *) pool_name,
+					   RTE_CRYPTO_OP_TYPE_SYMMETRIC,
+					   DPDK_CRYPTO_NB_COPS *
+					   (1 + vlib_num_workers ()),
+					   DPDK_CRYPTO_CACHE_SIZE,
+					   DPDK_CRYPTO_PRIV_SIZE,
+					   dev_conf.socket_id);
+	  vec_free (pool_name);
+
+	  if (!rmp)
+	    {
+	      clib_warning ("failed to allocate mempool on socket %u",
+			    dev_conf.socket_id);
+	      goto error;
+	    }
+	  vec_elt (dcm->cop_pools, dev_conf.socket_id) = rmp;
+	}
+
+      fprintf (stdout, "%u\t%u\t%u\t%u\n", dev_id, dev_conf.nb_queue_pairs,
+	       DPDK_CRYPTO_NB_SESS_OBJS, DPDK_CRYPTO_CACHE_SIZE);
+    }
+
+  dpdk_esp_init ();
+
+  /* Add new next node and set as default */
+  vlib_node_t *node, *next_node;
+
+  next_node = vlib_get_node_by_name (vm, (u8 *) "dpdk-esp-encrypt");
+  ASSERT (next_node);
+  node = vlib_get_node_by_name (vm, (u8 *) "ipsec-output-ip4");
+  ASSERT (node);
+  im->esp_encrypt_node_index = next_node->index;
+  im->esp_encrypt_next_index =
+    vlib_node_add_next (vm, node->index, next_node->index);
+
+  next_node = vlib_get_node_by_name (vm, (u8 *) "dpdk-esp-decrypt");
+  ASSERT (next_node);
+  node = vlib_get_node_by_name (vm, (u8 *) "ipsec-input-ip4");
+  ASSERT (node);
+  im->esp_decrypt_node_index = next_node->index;
+  im->esp_decrypt_next_index =
+    vlib_node_add_next (vm, node->index, next_node->index);
+
+  im->cb.check_support_cb = dpdk_ipsec_check_support;
+  im->cb.add_del_sa_sess_cb = add_del_sa_sess;
+
+  if (vec_len (vlib_mains) == 0)
+    vlib_node_set_state (&vlib_global_main, dpdk_crypto_input_node.index,
+			 VLIB_NODE_STATE_POLLING);
+  else
+    for (i = 1; i < tm->n_vlib_mains; i++)
+      vlib_node_set_state (vlib_mains[i], dpdk_crypto_input_node.index,
+			   VLIB_NODE_STATE_POLLING);
+
+  /* TODO cryptodev counters */
+
+  return 0;
+
+error:
+  ;
+  crypto_worker_main_t *cwm;
+  struct rte_mempool **mp;
+  /* *INDENT-OFF* */
+  vec_foreach (cwm, dcm->workers_main)
+    hash_free (cwm->algo_qp_map);
+
+  vec_foreach (mp, dcm->cop_pools)
+    {
+      if (mp)
+	rte_mempool_free (mp[0]);
+    }
+  /* *INDENT-ON* */
+  vec_free (dcm->workers_main);
+  vec_free (dcm->cop_pools);
+
+  return 0;
+}
+
+/* *INDENT-OFF* */
+VLIB_REGISTER_NODE (dpdk_ipsec_process_node,static) = {
+    .function = dpdk_ipsec_process,
+    .type = VLIB_NODE_TYPE_PROCESS,
+    .name = "dpdk-ipsec-process",
+    .process_log2_n_stack_bytes = 17,
+};
+/* *INDENT-ON* */
+
+/*
+ * fd.io coding-style-patch-verification: ON
+ *
+ * Local Variables:
+ * eval: (c-set-style "gnu")
+ * End:
+ */
diff --git a/src/plugins/dpdk/ipsec/ipsec.h b/src/plugins/dpdk/ipsec/ipsec.h
new file mode 100644
index 00000000..3465b361
--- /dev/null
+++ b/src/plugins/dpdk/ipsec/ipsec.h
@@ -0,0 +1,227 @@
+/*
+ * Copyright (c) 2016 Intel and/or its affiliates.
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at:
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+#ifndef __DPDK_IPSEC_H__
+#define __DPDK_IPSEC_H__
+
+#include 
+
+#undef always_inline
+#include 
+#include 
+
+#if CLIB_DEBUG > 0
+#define always_inline static inline
+#else
+#define always_inline static inline __attribute__ ((__always_inline__))
+#endif
+
+
+#define MAX_QP_PER_LCORE 16
+
+typedef struct
+{
+  u32 salt;
+  u32 iv[2];
+  u32 cnt;
+} dpdk_gcm_cnt_blk;
+
+typedef struct
+{
+  dpdk_gcm_cnt_blk cb;
+  union
+  {
+    u8 aad[12];
+    u8 icv[64];
+  };
+} dpdk_cop_priv_t;
+
+typedef struct
+{
+  u8 cipher_algo;
+  u8 auth_algo;
+  u8 is_outbound;
+} crypto_worker_qp_key_t;
+
+typedef struct
+{
+  u16 dev_id;
+  u16 qp_id;
+  u16 is_outbound;
+  i16 inflights;
+  u32 bi[VLIB_FRAME_SIZE];
+  struct rte_crypto_op *cops[VLIB_FRAME_SIZE];
+  struct rte_crypto_op **free_cops;
+} crypto_qp_data_t;
+
+typedef struct
+{
+  u8 qp_index;
+  void *sess;
+} crypto_sa_session_t;
+
+typedef struct
+{
+  crypto_sa_session_t *sa_sess_d[2];
+  crypto_qp_data_t *qp_data;
+  uword *algo_qp_map;
+} crypto_worker_main_t;
+
+typedef struct
+{
+  struct rte_mempool **cop_pools;
+  crypto_worker_main_t *workers_main;
+} dpdk_crypto_main_t;
+
+dpdk_crypto_main_t dpdk_crypto_main;
+
+extern vlib_node_registration_t dpdk_crypto_input_node;
+
+#define CRYPTO_N_FREE_COPS (VLIB_FRAME_SIZE * 3)
+
+static_always_inline void
+crypto_alloc_cops ()
+{
+  dpdk_crypto_main_t *dcm = &dpdk_crypto_main;
+  u32 cpu_index = os_get_cpu_number ();
+  crypto_worker_main_t *cwm = &dcm->workers_main[cpu_index];
+  unsigned socket_id = rte_socket_id ();
+  crypto_qp_data_t *qpd;
+
+  /* *INDENT-OFF* */
+  vec_foreach (qpd, cwm->qp_data)
+    {
+      u32 l = vec_len (qpd->free_cops);
+
+      if (PREDICT_FALSE (l < VLIB_FRAME_SIZE))
+	{
+	  u32 n_alloc;
+
+	  if (PREDICT_FALSE (!qpd->free_cops))
+	    vec_alloc (qpd->free_cops, CRYPTO_N_FREE_COPS);
+
+	  n_alloc = rte_crypto_op_bulk_alloc (dcm->cop_pools[socket_id],
+					      RTE_CRYPTO_OP_TYPE_SYMMETRIC,
+					      &qpd->free_cops[l],
+					      CRYPTO_N_FREE_COPS - l - 1);
+
+	  _vec_len (qpd->free_cops) = l + n_alloc;
+	}
+    }
+  /* *INDENT-ON* */
+}
+
+static_always_inline void
+crypto_free_cop (crypto_qp_data_t * qpd, struct rte_crypto_op **cops, u32 n)
+{
+  u32 l = vec_len (qpd->free_cops);
+
+  if (l + n >= CRYPTO_N_FREE_COPS)
+    {
+      l -= VLIB_FRAME_SIZE;
+      rte_mempool_put_bulk (cops[0]->mempool,
+			    (void **) &qpd->free_cops[l], VLIB_FRAME_SIZE);
+    }
+  clib_memcpy (&qpd->free_cops[l], cops, sizeof (*cops) * n);
+
+  _vec_len (qpd->free_cops) = l + n;
+}
+
+static_always_inline int
+check_algo_is_supported (const struct rte_cryptodev_capabilities *cap,
+			 char *name)
+{
+  struct
+  {
+    uint8_t cipher_algo;
+    enum rte_crypto_sym_xform_type type;
+    union
+    {
+      enum rte_crypto_auth_algorithm auth;
+      enum rte_crypto_cipher_algorithm cipher;
+    };
+    char *name;
+  } supported_algo[] =
+  {
+    {
+    .type = RTE_CRYPTO_SYM_XFORM_CIPHER,.cipher =
+	RTE_CRYPTO_CIPHER_NULL,.name = "NULL"},
+    {
+    .type = RTE_CRYPTO_SYM_XFORM_CIPHER,.cipher =
+	RTE_CRYPTO_CIPHER_AES_CBC,.name = "AES_CBC"},
+    {
+    .type = RTE_CRYPTO_SYM_XFORM_CIPHER,.cipher =
+	RTE_CRYPTO_CIPHER_AES_CTR,.name = "AES_CTR"},
+    {
+    .type = RTE_CRYPTO_SYM_XFORM_CIPHER,.cipher =
+	RTE_CRYPTO_CIPHER_3DES_CBC,.name = "3DES-CBC"},
+    {
+    .type = RTE_CRYPTO_SYM_XFORM_CIPHER,.cipher =
+	RTE_CRYPTO_CIPHER_AES_GCM,.name = "AES-GCM"},
+    {
+    .type = RTE_CRYPTO_SYM_XFORM_AUTH,.auth =
+	RTE_CRYPTO_AUTH_SHA1_HMAC,.name = "HMAC-SHA1"},
+    {
+    .type = RTE_CRYPTO_SYM_XFORM_AUTH,.auth =
+	RTE_CRYPTO_AUTH_SHA256_HMAC,.name = "HMAC-SHA256"},
+    {
+    .type = RTE_CRYPTO_SYM_XFORM_AUTH,.auth =
+	RTE_CRYPTO_AUTH_SHA384_HMAC,.name = "HMAC-SHA384"},
+    {
+    .type = RTE_CRYPTO_SYM_XFORM_AUTH,.auth =
+	RTE_CRYPTO_AUTH_SHA512_HMAC,.name = "HMAC-SHA512"},
+    {
+    .type = RTE_CRYPTO_SYM_XFORM_AUTH,.auth =
+	RTE_CRYPTO_AUTH_AES_XCBC_MAC,.name = "AES-XCBC-MAC"},
+    {
+    .type = RTE_CRYPTO_SYM_XFORM_AUTH,.auth =
+	RTE_CRYPTO_AUTH_AES_GCM,.name = "AES-GCM"},
+    {
+      /* tail */
+  .type = RTE_CRYPTO_SYM_XFORM_NOT_SPECIFIED},};
+  uint32_t i = 0;
+
+  if (cap->op != RTE_CRYPTO_OP_TYPE_SYMMETRIC)
+    return -1;
+
+  while (supported_algo[i].type != RTE_CRYPTO_SYM_XFORM_NOT_SPECIFIED)
+    {
+      if (cap->sym.xform_type == supported_algo[i].type)
+	{
+	  if ((cap->sym.xform_type == RTE_CRYPTO_SYM_XFORM_CIPHER &&
+	       cap->sym.cipher.algo == supported_algo[i].cipher) ||
+	      (cap->sym.xform_type == RTE_CRYPTO_SYM_XFORM_AUTH &&
+	       cap->sym.auth.algo == supported_algo[i].auth))
+	    {
+	      if (name)
+		strcpy (name, supported_algo[i].name);
+	      return 0;
+	    }
+	}
+
+      i++;
+    }
+
+  return -1;
+}
+
+#endif /* __DPDK_IPSEC_H__ */
+
+/*
+ * fd.io coding-style-patch-verification: ON
+ *
+ * Local Variables:
+ * eval: (c-set-style "gnu")
+ * End:
+ */
diff --git a/src/plugins/dpdk/main.c b/src/plugins/dpdk/main.c
new file mode 100644
index 00000000..8073a50a
--- /dev/null
+++ b/src/plugins/dpdk/main.c
@@ -0,0 +1,95 @@
+/*
+ * Copyright (c) 2017 Cisco and/or its affiliates.
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at:
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#include 
+#include 
+#include 
+#include 
+
+/*
+ * Called by the dpdk driver's rte_delay_us() function.
+ * Return 0 to have the dpdk do a regular delay loop.
+ * Return 1 if to skip the delay loop because we are suspending
+ * the calling vlib process instead.
+ */
+int
+rte_delay_us_override (unsigned us)
+{
+  vlib_main_t *vm;
+
+  /* Don't bother intercepting for short delays */
+  if (us < 10)
+    return 0;
+
+  /*
+   * Only intercept if we are in a vlib process.
+   * If we are called from a vlib worker thread or the vlib main
+   * thread then do not intercept. (Must not be called from an
+   * independent pthread).
+   */
+  if (os_get_cpu_number () == 0)
+    {
+      /*
+       * We're in the vlib main thread or a vlib process. Make sure
+       * the process is running and we're not still initializing.
+       */
+      vm = vlib_get_main ();
+      if (vlib_in_process_context (vm))
+	{
+	  /* Only suspend for the admin_down_process */
+	  vlib_process_t *proc = vlib_get_current_process (vm);
+	  if (!(proc->flags & VLIB_PROCESS_IS_RUNNING) ||
+	      (proc->node_runtime.function != admin_up_down_process))
+	    return 0;
+
+	  f64 delay = 1e-6 * us;
+	  vlib_process_suspend (vm, delay);
+	  return 1;
+	}
+    }
+  return 0;			// no override
+}
+
+static void
+rte_delay_us_override_cb (unsigned us)
+{
+  if (rte_delay_us_override (us) == 0)
+    rte_delay_us_block (us);
+}
+
+static clib_error_t * dpdk_main_init (vlib_main_t * vm)
+{
+  dpdk_main_t * dm = &dpdk_main;
+  clib_error_t * error = 0;
+
+  dm->vlib_main = vm;
+  dm->vnet_main = vnet_get_main ();
+
+  if ((error = vlib_call_init_function (vm, dpdk_init)))
+    return error;
+
+  /* register custom delay function */
+  rte_delay_us_callback_register (rte_delay_us_override_cb);
+
+  return error;
+}
+
+VLIB_INIT_FUNCTION (dpdk_main_init);
+
+/* *INDENT-OFF* */
+VLIB_PLUGIN_REGISTER () = {
+    .version = VPP_BUILD_VER,
+};
+/* *INDENT-ON* */
diff --git a/src/plugins/dpdk/thread.c b/src/plugins/dpdk/thread.c
new file mode 100644
index 00000000..3a3fcc6c
--- /dev/null
+++ b/src/plugins/dpdk/thread.c
@@ -0,0 +1,85 @@
+/*
+ * Copyright (c) 2017 Cisco and/or its affiliates.
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at:
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#include 
+
+#include 
+#include 
+#include 
+#include 
+#include 
+#include 
+#include 
+#include 
+#include 
+#include 
+#include 
+#include 
+#include 
+#include 
+#include 
+#include 
+#include 
+#include 
+#include 
+#include 
+#include 
+#include 
+#include 
+#include 
+
+#include 
+#include 
+#include 
+#include 
+
+static clib_error_t *
+dpdk_launch_thread (void *fp, vlib_worker_thread_t * w, unsigned lcore_id)
+{
+  int r;
+  r = rte_eal_remote_launch (fp, (void *) w, lcore_id);
+  if (r)
+    return clib_error_return (0, "Failed to launch thread %u", lcore_id);
+  return 0;
+}
+
+static clib_error_t *
+dpdk_thread_set_lcore (u32 thread, u16 lcore)
+{
+  return 0;
+}
+
+static vlib_thread_callbacks_t callbacks = {
+  .vlib_launch_thread_cb = &dpdk_launch_thread,
+  .vlib_thread_set_lcore_cb = &dpdk_thread_set_lcore,
+};
+
+static clib_error_t *
+dpdk_thread_init (vlib_main_t * vm)
+{
+  vlib_thread_cb_register (vm, &callbacks);
+  return 0;
+}
+
+VLIB_INIT_FUNCTION (dpdk_thread_init);
+
+/** @endcond */
+/*
+ * fd.io coding-style-patch-verification: ON
+ *
+ * Local Variables:
+ * eval: (c-set-style "gnu")
+ * End:
+ */
diff --git a/src/vat/api_format.c b/src/vat/api_format.c
index 999f9869..14e78817 100644
--- a/src/vat/api_format.c
+++ b/src/vat/api_format.c
@@ -3970,13 +3970,6 @@ _(feature_enable_disable_reply)				\
 _(sw_interface_tag_add_del_reply)			\
 _(sw_interface_set_mtu_reply)
 
-#if DPDK > 0
-#define foreach_standard_dpdk_reply_retval_handler      \
-_(sw_interface_set_dpdk_hqos_pipe_reply)                \
-_(sw_interface_set_dpdk_hqos_subport_reply)             \
-_(sw_interface_set_dpdk_hqos_tctbl_reply)
-#endif
-
 #define _(n)                                    \
     static void vl_api_##n##_t_handler          \
     (vl_api_##n##_t * mp)                       \
@@ -4008,39 +4001,6 @@ foreach_standard_reply_retval_handler;
 foreach_standard_reply_retval_handler;
 #undef _
 
-#if DPDK > 0
-#define _(n)                                    \
-    static void vl_api_##n##_t_handler          \
-    (vl_api_##n##_t * mp)                       \
-    {                                           \
-        vat_main_t * vam = &vat_main;           \
-        i32 retval = ntohl(mp->retval);         \
-        if (vam->async_mode) {                  \
-            vam->async_errors += (retval < 0);  \
-        } else {                                \
-            vam->retval = retval;               \
-            vam->result_ready = 1;              \
-        }                                       \
-    }
-foreach_standard_dpdk_reply_retval_handler;
-#undef _
-
-#define _(n)                                    \
-    static void vl_api_##n##_t_handler_json     \
-    (vl_api_##n##_t * mp)                       \
-    {                                           \
-        vat_main_t * vam = &vat_main;           \
-        vat_json_node_t node;                   \
-        vat_json_init_object(&node);            \
-        vat_json_object_add_int(&node, "retval", ntohl(mp->retval));    \
-        vat_json_print(vam->ofp, &node);        \
-        vam->retval = ntohl(mp->retval);        \
-        vam->result_ready = 1;                  \
-    }
-foreach_standard_dpdk_reply_retval_handler;
-#undef _
-#endif
-
 /*
  * Table of message reply handlers, must include boilerplate handlers
  * we just generated
@@ -4272,16 +4232,6 @@ _(SW_INTERFACE_SET_MTU_REPLY, sw_interface_set_mtu_reply)               \
 _(IP_NEIGHBOR_DETAILS, ip_neighbor_details)                             \
 _(SW_INTERFACE_GET_TABLE_REPLY, sw_interface_get_table_reply)
 
-#if DPDK > 0
-#define foreach_vpe_dpdk_api_reply_msg                                  \
-_(SW_INTERFACE_SET_DPDK_HQOS_PIPE_REPLY,                                \
-  sw_interface_set_dpdk_hqos_pipe_reply)                                \
-_(SW_INTERFACE_SET_DPDK_HQOS_SUBPORT_REPLY,                             \
-  sw_interface_set_dpdk_hqos_subport_reply)                             \
-_(SW_INTERFACE_SET_DPDK_HQOS_TCTBL_REPLY,                               \
-  sw_interface_set_dpdk_hqos_tctbl_reply)
-#endif
-
 typedef struct
 {
   u8 *name;
@@ -5081,226 +5031,6 @@ api_sw_interface_clear_stats (vat_main_t * vam)
   return ret;
 }
 
-#if DPDK >0
-static int
-api_sw_interface_set_dpdk_hqos_pipe (vat_main_t * vam)
-{
-  unformat_input_t *i = vam->input;
-  vl_api_sw_interface_set_dpdk_hqos_pipe_t *mp;
-  u32 sw_if_index;
-  u8 sw_if_index_set = 0;
-  u32 subport;
-  u8 subport_set = 0;
-  u32 pipe;
-  u8 pipe_set = 0;
-  u32 profile;
-  u8 profile_set = 0;
-  int ret;
-
-  /* Parse args required to build the message */
-  while (unformat_check_input (i) != UNFORMAT_END_OF_INPUT)
-    {
-      if (unformat (i, "rx %U", api_unformat_sw_if_index, vam, &sw_if_index))
-	sw_if_index_set = 1;
-      else if (unformat (i, "sw_if_index %u", &sw_if_index))
-	sw_if_index_set = 1;
-      else if (unformat (i, "subport %u", &subport))
-	subport_set = 1;
-      else
-	if (unformat (i, "%U", api_unformat_sw_if_index, vam, &sw_if_index))
-	sw_if_index_set = 1;
-      else if (unformat (i, "pipe %u", &pipe))
-	pipe_set = 1;
-      else if (unformat (i, "profile %u", &profile))
-	profile_set = 1;
-      else
-	break;
-    }
-
-  if (sw_if_index_set == 0)
-    {
-      errmsg ("missing interface name or sw_if_index");
-      return -99;
-    }
-
-  if (subport_set == 0)
-    {
-      errmsg ("missing subport ");
-      return -99;
-    }
-
-  if (pipe_set == 0)
-    {
-      errmsg ("missing pipe");
-      return -99;
-    }
-
-  if (profile_set == 0)
-    {
-      errmsg ("missing profile");
-      return -99;
-    }
-
-  M (SW_INTERFACE_SET_DPDK_HQOS_PIPE, mp);
-
-  mp->sw_if_index = ntohl (sw_if_index);
-  mp->subport = ntohl (subport);
-  mp->pipe = ntohl (pipe);
-  mp->profile = ntohl (profile);
-
-
-  S (mp);
-  W (ret);
-  return ret;
-}
-
-static int
-api_sw_interface_set_dpdk_hqos_subport (vat_main_t * vam)
-{
-  unformat_input_t *i = vam->input;
-  vl_api_sw_interface_set_dpdk_hqos_subport_t *mp;
-  u32 sw_if_index;
-  u8 sw_if_index_set = 0;
-  u32 subport;
-  u8 subport_set = 0;
-  u32 tb_rate = 1250000000;	/* 10GbE */
-  u32 tb_size = 1000000;
-  u32 tc_rate[] = { 1250000000, 1250000000, 1250000000, 1250000000 };
-  u32 tc_period = 10;
-  int ret;
-
-  /* Parse args required to build the message */
-  while (unformat_check_input (i) != UNFORMAT_END_OF_INPUT)
-    {
-      if (unformat (i, "rx %U", api_unformat_sw_if_index, vam, &sw_if_index))
-	sw_if_index_set = 1;
-      else if (unformat (i, "sw_if_index %u", &sw_if_index))
-	sw_if_index_set = 1;
-      else if (unformat (i, "subport %u", &subport))
-	subport_set = 1;
-      else
-	if (unformat (i, "%U", api_unformat_sw_if_index, vam, &sw_if_index))
-	sw_if_index_set = 1;
-      else if (unformat (i, "rate %u", &tb_rate))
-	{
-	  u32 tc_id;
-
-	  for (tc_id = 0; tc_id < (sizeof (tc_rate) / sizeof (tc_rate[0]));
-	       tc_id++)
-	    tc_rate[tc_id] = tb_rate;
-	}
-      else if (unformat (i, "bktsize %u", &tb_size))
-	;
-      else if (unformat (i, "tc0 %u", &tc_rate[0]))
-	;
-      else if (unformat (i, "tc1 %u", &tc_rate[1]))
-	;
-      else if (unformat (i, "tc2 %u", &tc_rate[2]))
-	;
-      else if (unformat (i, "tc3 %u", &tc_rate[3]))
-	;
-      else if (unformat (i, "period %u", &tc_period))
-	;
-      else
-	break;
-    }
-
-  if (sw_if_index_set == 0)
-    {
-      errmsg ("missing interface name or sw_if_index");
-      return -99;
-    }
-
-  if (subport_set == 0)
-    {
-      errmsg ("missing subport ");
-      return -99;
-    }
-
-  M (SW_INTERFACE_SET_DPDK_HQOS_SUBPORT, mp);
-
-  mp->sw_if_index = ntohl (sw_if_index);
-  mp->subport = ntohl (subport);
-  mp->tb_rate = ntohl (tb_rate);
-  mp->tb_size = ntohl (tb_size);
-  mp->tc_rate[0] = ntohl (tc_rate[0]);
-  mp->tc_rate[1] = ntohl (tc_rate[1]);
-  mp->tc_rate[2] = ntohl (tc_rate[2]);
-  mp->tc_rate[3] = ntohl (tc_rate[3]);
-  mp->tc_period = ntohl (tc_period);
-
-  S (mp);
-  W (ret);
-  return ret;
-}
-
-static int
-api_sw_interface_set_dpdk_hqos_tctbl (vat_main_t * vam)
-{
-  unformat_input_t *i = vam->input;
-  vl_api_sw_interface_set_dpdk_hqos_tctbl_t *mp;
-  u32 sw_if_index;
-  u8 sw_if_index_set = 0;
-  u8 entry_set = 0;
-  u8 tc_set = 0;
-  u8 queue_set = 0;
-  u32 entry, tc, queue;
-  int ret;
-
-  /* Parse args required to build the message */
-  while (unformat_check_input (i) != UNFORMAT_END_OF_INPUT)
-    {
-      if (unformat (i, "rx %U", api_unformat_sw_if_index, vam, &sw_if_index))
-	sw_if_index_set = 1;
-      else if (unformat (i, "sw_if_index %u", &sw_if_index))
-	sw_if_index_set = 1;
-      else if (unformat (i, "entry %d", &entry))
-	entry_set = 1;
-      else if (unformat (i, "tc %d", &tc))
-	tc_set = 1;
-      else if (unformat (i, "queue %d", &queue))
-	queue_set = 1;
-      else
-	break;
-    }
-
-  if (sw_if_index_set == 0)
-    {
-      errmsg ("missing interface name or sw_if_index");
-      return -99;
-    }
-
-  if (entry_set == 0)
-    {
-      errmsg ("missing entry ");
-      return -99;
-    }
-
-  if (tc_set == 0)
-    {
-      errmsg ("missing traffic class ");
-      return -99;
-    }
-
-  if (queue_set == 0)
-    {
-      errmsg ("missing queue ");
-      return -99;
-    }
-
-  M (SW_INTERFACE_SET_DPDK_HQOS_TCTBL, mp);
-
-  mp->sw_if_index = ntohl (sw_if_index);
-  mp->entry = ntohl (entry);
-  mp->tc = ntohl (tc);
-  mp->queue = ntohl (queue);
-
-  S (mp);
-  W (ret);
-  return ret;
-}
-#endif
-
 static int
 api_sw_interface_add_del_address (vat_main_t * vam)
 {
@@ -18656,18 +18386,6 @@ _(sw_interface_set_mtu, " | sw_if_index  mtu ")        \
 _(ip_neighbor_dump, "[ip6]  | sw_if_index ")                 \
 _(sw_interface_get_table, " | sw_if_index  [ipv6]")
 
-#if DPDK > 0
-#define foreach_vpe_dpdk_api_msg                                        \
-_(sw_interface_set_dpdk_hqos_pipe,                                      \
-  "rx  | sw_if_index  subport  pipe \n" \
-  "profile \n")                                             \
-_(sw_interface_set_dpdk_hqos_subport,                                   \
-  "rx  | sw_if_index  subport  [rate ]\n"     \
-  "[bktsize ] [tc0 ] [tc1 ] [tc2 ] [tc3 ] [period ]\n") \
-_(sw_interface_set_dpdk_hqos_tctbl,                                     \
-  "rx  | sw_if_index  entry  tc  queue \n")
-#endif
-
 /* List of command functions, CLI names map directly to functions */
 #define foreach_cli_function                                    \
 _(comment, "usage: comment ")		\
@@ -18705,22 +18423,6 @@ _(unset, "usage: unset ")
 foreach_vpe_api_reply_msg;
 #undef _
 
-#if DPDK > 0
-#define _(N,n)                                  \
-    static void vl_api_##n##_t_handler_uni      \
-    (vl_api_##n##_t * mp)                       \
-    {                                           \
-        vat_main_t * vam = &vat_main;           \
-        if (vam->json_output) {                 \
-            vl_api_##n##_t_handler_json(mp);    \
-        } else {                                \
-            vl_api_##n##_t_handler(mp);         \
-        }                                       \
-    }
-foreach_vpe_dpdk_api_reply_msg;
-#undef _
-#endif
-
 void
 vat_api_hookup (vat_main_t * vam)
 {
@@ -18734,18 +18436,6 @@ vat_api_hookup (vat_main_t * vam)
   foreach_vpe_api_reply_msg;
 #undef _
 
-#if DPDK > 0
-#define _(N,n)                                                  \
-    vl_msg_api_set_handlers(VL_API_##N, #n,                     \
-                           vl_api_##n##_t_handler_uni,          \
-                           vl_noop_handler,                     \
-                           vl_api_##n##_t_endian,               \
-                           vl_api_##n##_t_print,                \
-                           sizeof(vl_api_##n##_t), 1);
-  foreach_vpe_dpdk_api_reply_msg;
-#undef _
-#endif
-
 #if (VPP_API_TEST_BUILTIN==0)
   vl_msg_api_set_first_available_msg_id (VL_MSG_FIRST_AVAILABLE);
 #endif
@@ -18760,21 +18450,11 @@ vat_api_hookup (vat_main_t * vam)
 #define _(n,h) hash_set_mem (vam->function_by_name, #n, api_##n);
   foreach_vpe_api_msg;
 #undef _
-#if DPDK >0
-#define _(n,h) hash_set_mem (vam->function_by_name, #n, api_##n);
-  foreach_vpe_dpdk_api_msg;
-#undef _
-#endif
 
   /* Help strings */
 #define _(n,h) hash_set_mem (vam->help_by_name, #n, h);
   foreach_vpe_api_msg;
 #undef _
-#if DPDK >0
-#define _(n,h) hash_set_mem (vam->help_by_name, #n, h);
-  foreach_vpe_dpdk_api_msg;
-#undef _
-#endif
 
   /* CLI functions */
 #define _(n,h) hash_set_mem (vam->function_by_name, #n, n);
diff --git a/src/vnet.am b/src/vnet.am
index 923f61d8..84930f05 100644
--- a/src/vnet.am
+++ b/src/vnet.am
@@ -23,8 +23,7 @@ libvnet_la_DEPENDENCIES = \
 	libvlibmemory.la	\
 	libvlibmemoryclient.la
 
-libvnet_la_LIBADD = $(libvnet_la_DEPENDENCIES) -lm -lpthread -ldl -lrt $(DPDK_LD_ADD)
-libvnet_la_LDFLAGS = $(DPDK_LD_FLAGS)
+libvnet_la_LIBADD = $(libvnet_la_DEPENDENCIES) -lm -lpthread -ldl -lrt
 
 if WITH_LIBSSL
 libvnet_la_LIBADD += -lcrypto
@@ -396,15 +395,6 @@ libvnet_la_SOURCES +=				\
  vnet/ipsec/ipsec_api.c
 
 API_FILES += vnet/ipsec/ipsec.api
-
-if WITH_DPDK
-libvnet_la_SOURCES +=				\
- vnet/devices/dpdk/ipsec/esp_encrypt.c		\
- vnet/devices/dpdk/ipsec/esp_decrypt.c		\
- vnet/devices/dpdk/ipsec/crypto_node.c		\
- vnet/devices/dpdk/ipsec/cli.c			\
- vnet/devices/dpdk/ipsec/ipsec.c
-endif
 endif
 
 libvnet_la_SOURCES +=				\
@@ -416,11 +406,6 @@ nobase_include_HEADERS +=			\
  vnet/ipsec/ikev2.h				\
  vnet/ipsec/ikev2_priv.h			\
  vnet/ipsec/ipsec.api.h
-if WITH_DPDK
-nobase_include_HEADERS +=			\
- vnet/devices/dpdk/ipsec/ipsec.h		\
- vnet/devices/dpdk/ipsec/esp.h
-endif
 
 ########################################
 # Layer 3 protocol: osi
@@ -803,29 +788,7 @@ nobase_include_HEADERS +=			\
   vnet/pg/pg.h					\
   vnet/pg/edit.h
 
-########################################
-# DPDK
-########################################
-if WITH_DPDK
-libvnet_la_SOURCES +=				\
-  vnet/devices/dpdk/buffer.c			\
-  vnet/devices/dpdk/dpdk_priv.h		\
-  vnet/devices/dpdk/device.c		\
-  vnet/devices/dpdk/format.c		\
-  vnet/devices/dpdk/init.c			\
-  vnet/devices/dpdk/main.c			\
-  vnet/devices/dpdk/node.c			\
-  vnet/devices/dpdk/thread.c			\
-  vnet/devices/dpdk/hqos.c			\
-  vnet/devices/dpdk/cli.c			\
-  vnet/devices/dpdk/dpdk_api.c
-
-nobase_include_HEADERS +=			\
-  vnet/devices/dpdk/dpdk.h			\
-  vnet/devices/dpdk/dpdk.api.h
-
-API_FILES += vnet/devices/dpdk/dpdk.api
-else
+if !WITH_DPDK
 libvnet_la_SOURCES +=				\
   vnet/devices/nic/ixge.c                       \
   vnet/devices/nic/ixge.h                       \
diff --git a/src/vnet/devices/dpdk/buffer.c b/src/vnet/devices/dpdk/buffer.c
deleted file mode 100644
index f95d4cb5..00000000
--- a/src/vnet/devices/dpdk/buffer.c
+++ /dev/null
@@ -1,588 +0,0 @@
-/*
- * Copyright (c) 2017 Cisco and/or its affiliates.
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at:
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-/*
- * buffer.c: allocate/free network buffers.
- *
- * Copyright (c) 2008 Eliot Dresselhaus
- *
- * Permission is hereby granted, free of charge, to any person obtaining
- * a copy of this software and associated documentation files (the
- * "Software"), to deal in the Software without restriction, including
- * without limitation the rights to use, copy, modify, merge, publish,
- * distribute, sublicense, and/or sell copies of the Software, and to
- * permit persons to whom the Software is furnished to do so, subject to
- * the following conditions:
- *
- * The above copyright notice and this permission notice shall be
- * included in all copies or substantial portions of the Software.
- *
- *  THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
- *  EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
- *  MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
- *  NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE
- *  LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
- *  OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
- *  WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
- */
-
-/**
- * @file
- *
- * Allocate/free network buffers.
- */
-
-#include 
-
-#include 
-#include 
-#include 
-#include 
-#include 
-#include 
-#include 
-#include 
-#include 
-#include 
-#include 
-#include 
-#include 
-#include 
-#include 
-#include 
-#include 
-#include 
-#include 
-#include 
-#include 
-#include 
-#include 
-#include 
-
-#include 
-#include 
-#include 
-#include 
-
-
-STATIC_ASSERT (VLIB_BUFFER_PRE_DATA_SIZE == RTE_PKTMBUF_HEADROOM,
-	       "VLIB_BUFFER_PRE_DATA_SIZE must be equal to RTE_PKTMBUF_HEADROOM");
-
-static_always_inline void
-dpdk_rte_pktmbuf_free (vlib_main_t * vm, vlib_buffer_t * b)
-{
-  vlib_buffer_t *hb = b;
-  struct rte_mbuf *mb;
-  u32 next, flags;
-  mb = rte_mbuf_from_vlib_buffer (hb);
-
-next:
-  flags = b->flags;
-  next = b->next_buffer;
-  mb = rte_mbuf_from_vlib_buffer (b);
-
-  if (PREDICT_FALSE (b->n_add_refs))
-    {
-      rte_mbuf_refcnt_update (mb, b->n_add_refs);
-      b->n_add_refs = 0;
-    }
-
-  rte_pktmbuf_free_seg (mb);
-
-  if (flags & VLIB_BUFFER_NEXT_PRESENT)
-    {
-      b = vlib_get_buffer (vm, next);
-      goto next;
-    }
-}
-
-static void
-del_free_list (vlib_main_t * vm, vlib_buffer_free_list_t * f)
-{
-  u32 i;
-  vlib_buffer_t *b;
-
-  for (i = 0; i < vec_len (f->buffers); i++)
-    {
-      b = vlib_get_buffer (vm, f->buffers[i]);
-      dpdk_rte_pktmbuf_free (vm, b);
-    }
-
-  vec_free (f->name);
-  vec_free (f->buffers);
-}
-
-/* Add buffer free list. */
-static void
-dpdk_buffer_delete_free_list (vlib_main_t * vm, u32 free_list_index)
-{
-  vlib_buffer_main_t *bm = vm->buffer_main;
-  vlib_buffer_free_list_t *f;
-  u32 merge_index;
-  int i;
-
-  ASSERT (os_get_cpu_number () == 0);
-
-  f = vlib_buffer_get_free_list (vm, free_list_index);
-
-  merge_index = vlib_buffer_get_free_list_with_size (vm, f->n_data_bytes);
-  if (merge_index != ~0 && merge_index != free_list_index)
-    {
-      vlib_buffer_merge_free_lists (pool_elt_at_index
-				    (bm->buffer_free_list_pool, merge_index),
-				    f);
-    }
-
-  del_free_list (vm, f);
-
-  /* Poison it. */
-  memset (f, 0xab, sizeof (f[0]));
-
-  pool_put (bm->buffer_free_list_pool, f);
-
-  for (i = 1; i < vec_len (vlib_mains); i++)
-    {
-      bm = vlib_mains[i]->buffer_main;
-      f = vlib_buffer_get_free_list (vlib_mains[i], free_list_index);;
-      memset (f, 0xab, sizeof (f[0]));
-      pool_put (bm->buffer_free_list_pool, f);
-    }
-}
-
-/* Make sure free list has at least given number of free buffers. */
-static uword
-fill_free_list (vlib_main_t * vm,
-		vlib_buffer_free_list_t * fl, uword min_free_buffers)
-{
-  dpdk_main_t *dm = &dpdk_main;
-  vlib_buffer_t *b0, *b1, *b2, *b3;
-  int n, i;
-  u32 bi0, bi1, bi2, bi3;
-  unsigned socket_id = rte_socket_id ();
-  struct rte_mempool *rmp = dm->pktmbuf_pools[socket_id];
-  struct rte_mbuf *mb0, *mb1, *mb2, *mb3;
-
-  /* Too early? */
-  if (PREDICT_FALSE (rmp == 0))
-    return 0;
-
-  /* Already have enough free buffers on free list? */
-  n = min_free_buffers - vec_len (fl->buffers);
-  if (n <= 0)
-    return min_free_buffers;
-
-  /* Always allocate round number of buffers. */
-  n = round_pow2 (n, CLIB_CACHE_LINE_BYTES / sizeof (u32));
-
-  /* Always allocate new buffers in reasonably large sized chunks. */
-  n = clib_max (n, fl->min_n_buffers_each_physmem_alloc);
-
-  vec_validate (vm->mbuf_alloc_list, n - 1);
-
-  if (rte_mempool_get_bulk (rmp, vm->mbuf_alloc_list, n) < 0)
-    return 0;
-
-  _vec_len (vm->mbuf_alloc_list) = n;
-
-  i = 0;
-
-  while (i < (n - 7))
-    {
-      vlib_prefetch_buffer_header (vlib_buffer_from_rte_mbuf
-				   (vm->mbuf_alloc_list[i + 4]), STORE);
-      vlib_prefetch_buffer_header (vlib_buffer_from_rte_mbuf
-				   (vm->mbuf_alloc_list[i + 5]), STORE);
-      vlib_prefetch_buffer_header (vlib_buffer_from_rte_mbuf
-				   (vm->mbuf_alloc_list[i + 6]), STORE);
-      vlib_prefetch_buffer_header (vlib_buffer_from_rte_mbuf
-				   (vm->mbuf_alloc_list[i + 7]), STORE);
-
-      mb0 = vm->mbuf_alloc_list[i];
-      mb1 = vm->mbuf_alloc_list[i + 1];
-      mb2 = vm->mbuf_alloc_list[i + 2];
-      mb3 = vm->mbuf_alloc_list[i + 3];
-
-      ASSERT (rte_mbuf_refcnt_read (mb0) == 0);
-      ASSERT (rte_mbuf_refcnt_read (mb1) == 0);
-      ASSERT (rte_mbuf_refcnt_read (mb2) == 0);
-      ASSERT (rte_mbuf_refcnt_read (mb3) == 0);
-
-      rte_mbuf_refcnt_set (mb0, 1);
-      rte_mbuf_refcnt_set (mb1, 1);
-      rte_mbuf_refcnt_set (mb2, 1);
-      rte_mbuf_refcnt_set (mb3, 1);
-
-      b0 = vlib_buffer_from_rte_mbuf (mb0);
-      b1 = vlib_buffer_from_rte_mbuf (mb1);
-      b2 = vlib_buffer_from_rte_mbuf (mb2);
-      b3 = vlib_buffer_from_rte_mbuf (mb3);
-
-      bi0 = vlib_get_buffer_index (vm, b0);
-      bi1 = vlib_get_buffer_index (vm, b1);
-      bi2 = vlib_get_buffer_index (vm, b2);
-      bi3 = vlib_get_buffer_index (vm, b3);
-
-      vec_add1_aligned (fl->buffers, bi0, CLIB_CACHE_LINE_BYTES);
-      vec_add1_aligned (fl->buffers, bi1, CLIB_CACHE_LINE_BYTES);
-      vec_add1_aligned (fl->buffers, bi2, CLIB_CACHE_LINE_BYTES);
-      vec_add1_aligned (fl->buffers, bi3, CLIB_CACHE_LINE_BYTES);
-
-      vlib_buffer_init_for_free_list (b0, fl);
-      vlib_buffer_init_for_free_list (b1, fl);
-      vlib_buffer_init_for_free_list (b2, fl);
-      vlib_buffer_init_for_free_list (b3, fl);
-
-      if (fl->buffer_init_function)
-	{
-	  fl->buffer_init_function (vm, fl, &bi0, 1);
-	  fl->buffer_init_function (vm, fl, &bi1, 1);
-	  fl->buffer_init_function (vm, fl, &bi2, 1);
-	  fl->buffer_init_function (vm, fl, &bi3, 1);
-	}
-      i += 4;
-    }
-
-  while (i < n)
-    {
-      mb0 = vm->mbuf_alloc_list[i];
-
-      ASSERT (rte_mbuf_refcnt_read (mb0) == 0);
-      rte_mbuf_refcnt_set (mb0, 1);
-
-      b0 = vlib_buffer_from_rte_mbuf (mb0);
-      bi0 = vlib_get_buffer_index (vm, b0);
-
-      vec_add1_aligned (fl->buffers, bi0, CLIB_CACHE_LINE_BYTES);
-
-      vlib_buffer_init_for_free_list (b0, fl);
-
-      if (fl->buffer_init_function)
-	fl->buffer_init_function (vm, fl, &bi0, 1);
-      i++;
-    }
-
-  fl->n_alloc += n;
-
-  return n;
-}
-
-static u32
-alloc_from_free_list (vlib_main_t * vm,
-		      vlib_buffer_free_list_t * free_list,
-		      u32 * alloc_buffers, u32 n_alloc_buffers)
-{
-  u32 *dst, *src;
-  uword len, n_filled;
-
-  dst = alloc_buffers;
-
-  n_filled = fill_free_list (vm, free_list, n_alloc_buffers);
-  if (n_filled == 0)
-    return 0;
-
-  len = vec_len (free_list->buffers);
-  ASSERT (len >= n_alloc_buffers);
-
-  src = free_list->buffers + len - n_alloc_buffers;
-  clib_memcpy (dst, src, n_alloc_buffers * sizeof (u32));
-
-  _vec_len (free_list->buffers) -= n_alloc_buffers;
-
-  return n_alloc_buffers;
-}
-
-/* Allocate a given number of buffers into given array.
-   Returns number actually allocated which will be either zero or
-   number requested. */
-u32
-dpdk_buffer_alloc (vlib_main_t * vm, u32 * buffers, u32 n_buffers)
-{
-  vlib_buffer_main_t *bm = vm->buffer_main;
-
-  return alloc_from_free_list
-    (vm,
-     pool_elt_at_index (bm->buffer_free_list_pool,
-			VLIB_BUFFER_DEFAULT_FREE_LIST_INDEX),
-     buffers, n_buffers);
-}
-
-
-u32
-dpdk_buffer_alloc_from_free_list (vlib_main_t * vm,
-				  u32 * buffers,
-				  u32 n_buffers, u32 free_list_index)
-{
-  vlib_buffer_main_t *bm = vm->buffer_main;
-  vlib_buffer_free_list_t *f;
-  f = pool_elt_at_index (bm->buffer_free_list_pool, free_list_index);
-  return alloc_from_free_list (vm, f, buffers, n_buffers);
-}
-
-static_always_inline void
-vlib_buffer_free_inline (vlib_main_t * vm,
-			 u32 * buffers, u32 n_buffers, u32 follow_buffer_next)
-{
-  vlib_buffer_main_t *bm = vm->buffer_main;
-  vlib_buffer_free_list_t *fl;
-  u32 fi;
-  int i;
-  u32 (*cb) (vlib_main_t * vm, u32 * buffers, u32 n_buffers,
-	     u32 follow_buffer_next);
-
-  cb = bm->buffer_free_callback;
-
-  if (PREDICT_FALSE (cb != 0))
-    n_buffers = (*cb) (vm, buffers, n_buffers, follow_buffer_next);
-
-  if (!n_buffers)
-    return;
-
-  for (i = 0; i < n_buffers; i++)
-    {
-      vlib_buffer_t *b;
-
-      b = vlib_get_buffer (vm, buffers[i]);
-
-      fl = vlib_buffer_get_buffer_free_list (vm, b, &fi);
-
-      /* The only current use of this callback: multicast recycle */
-      if (PREDICT_FALSE (fl->buffers_added_to_freelist_function != 0))
-	{
-	  int j;
-
-	  vlib_buffer_add_to_free_list
-	    (vm, fl, buffers[i], (b->flags & VLIB_BUFFER_RECYCLE) == 0);
-
-	  for (j = 0; j < vec_len (bm->announce_list); j++)
-	    {
-	      if (fl == bm->announce_list[j])
-		goto already_announced;
-	    }
-	  vec_add1 (bm->announce_list, fl);
-	already_announced:
-	  ;
-	}
-      else
-	{
-	  if (PREDICT_TRUE ((b->flags & VLIB_BUFFER_RECYCLE) == 0))
-	    dpdk_rte_pktmbuf_free (vm, b);
-	}
-    }
-  if (vec_len (bm->announce_list))
-    {
-      vlib_buffer_free_list_t *fl;
-      for (i = 0; i < vec_len (bm->announce_list); i++)
-	{
-	  fl = bm->announce_list[i];
-	  fl->buffers_added_to_freelist_function (vm, fl);
-	}
-      _vec_len (bm->announce_list) = 0;
-    }
-}
-
-static void
-dpdk_buffer_free (vlib_main_t * vm, u32 * buffers, u32 n_buffers)
-{
-  vlib_buffer_free_inline (vm, buffers, n_buffers,	/* follow_buffer_next */
-			   1);
-}
-
-static void
-dpdk_buffer_free_no_next (vlib_main_t * vm, u32 * buffers, u32 n_buffers)
-{
-  vlib_buffer_free_inline (vm, buffers, n_buffers,	/* follow_buffer_next */
-			   0);
-}
-
-static void
-dpdk_packet_template_init (vlib_main_t * vm,
-			   void *vt,
-			   void *packet_data,
-			   uword n_packet_data_bytes,
-			   uword min_n_buffers_each_physmem_alloc, u8 * name)
-{
-  vlib_packet_template_t *t = (vlib_packet_template_t *) vt;
-
-  vlib_worker_thread_barrier_sync (vm);
-  memset (t, 0, sizeof (t[0]));
-
-  vec_add (t->packet_data, packet_data, n_packet_data_bytes);
-
-  vlib_worker_thread_barrier_release (vm);
-}
-
-clib_error_t *
-vlib_buffer_pool_create (vlib_main_t * vm, unsigned num_mbufs,
-			 unsigned socket_id)
-{
-  dpdk_main_t *dm = &dpdk_main;
-  vlib_physmem_main_t *vpm = &vm->physmem_main;
-  struct rte_mempool *rmp;
-  int i;
-
-  vec_validate_aligned (dm->pktmbuf_pools, socket_id, CLIB_CACHE_LINE_BYTES);
-
-  /* pool already exists, nothing to do */
-  if (dm->pktmbuf_pools[socket_id])
-    return 0;
-
-  u8 *pool_name = format (0, "mbuf_pool_socket%u%c", socket_id, 0);
-
-  rmp = rte_pktmbuf_pool_create ((char *) pool_name,	/* pool name */
-				 num_mbufs,	/* number of mbufs */
-				 512,	/* cache size */
-				 VLIB_BUFFER_HDR_SIZE,	/* priv size */
-				 VLIB_BUFFER_PRE_DATA_SIZE + VLIB_BUFFER_DATA_SIZE,	/* dataroom size */
-				 socket_id);	/* cpu socket */
-
-  if (rmp)
-    {
-      {
-	uword this_pool_end;
-	uword this_pool_start;
-	uword this_pool_size;
-	uword save_vpm_start, save_vpm_end, save_vpm_size;
-	struct rte_mempool_memhdr *memhdr;
-
-	this_pool_start = ~0ULL;
-	this_pool_end = 0LL;
-
-	STAILQ_FOREACH (memhdr, &rmp->mem_list, next)
-	{
-	  if (((uword) (memhdr->addr + memhdr->len)) > this_pool_end)
-	    this_pool_end = (uword) (memhdr->addr + memhdr->len);
-	  if (((uword) memhdr->addr) < this_pool_start)
-	    this_pool_start = (uword) (memhdr->addr);
-	}
-	ASSERT (this_pool_start < ~0ULL && this_pool_end > 0);
-	this_pool_size = this_pool_end - this_pool_start;
-
-	if (CLIB_DEBUG > 1)
-	  {
-	    clib_warning ("%s: pool start %llx pool end %llx pool size %lld",
-			  pool_name, this_pool_start, this_pool_end,
-			  this_pool_size);
-	    clib_warning
-	      ("before: virtual.start %llx virtual.end %llx virtual.size %lld",
-	       vpm->virtual.start, vpm->virtual.end, vpm->virtual.size);
-	  }
-
-	save_vpm_start = vpm->virtual.start;
-	save_vpm_end = vpm->virtual.end;
-	save_vpm_size = vpm->virtual.size;
-
-	if ((this_pool_start < vpm->virtual.start) || vpm->virtual.start == 0)
-	  vpm->virtual.start = this_pool_start;
-	if (this_pool_end > vpm->virtual.end)
-	  vpm->virtual.end = this_pool_end;
-
-	vpm->virtual.size = vpm->virtual.end - vpm->virtual.start;
-
-	if (CLIB_DEBUG > 1)
-	  {
-	    clib_warning
-	      ("after: virtual.start %llx virtual.end %llx virtual.size %lld",
-	       vpm->virtual.start, vpm->virtual.end, vpm->virtual.size);
-	  }
-
-	/* check if fits into buffer index range */
-	if ((u64) vpm->virtual.size >
-	    ((u64) 1 << (32 + CLIB_LOG2_CACHE_LINE_BYTES)))
-	  {
-	    clib_warning ("physmem: virtual size out of range!");
-	    vpm->virtual.start = save_vpm_start;
-	    vpm->virtual.end = save_vpm_end;
-	    vpm->virtual.size = save_vpm_size;
-	    rmp = 0;
-	  }
-      }
-      if (rmp)
-	{
-	  dm->pktmbuf_pools[socket_id] = rmp;
-	  vec_free (pool_name);
-	  return 0;
-	}
-    }
-
-  vec_free (pool_name);
-
-  /* no usable pool for this socket, try to use pool from another one */
-  for (i = 0; i < vec_len (dm->pktmbuf_pools); i++)
-    {
-      if (dm->pktmbuf_pools[i])
-	{
-	  clib_warning
-	    ("WARNING: Failed to allocate mempool for CPU socket %u. "
-	     "Threads running on socket %u will use socket %u mempool.",
-	     socket_id, socket_id, i);
-	  dm->pktmbuf_pools[socket_id] = dm->pktmbuf_pools[i];
-	  return 0;
-	}
-    }
-
-  return clib_error_return (0, "failed to allocate mempool on socket %u",
-			    socket_id);
-}
-
-#if CLIB_DEBUG > 0
-
-u32 *vlib_buffer_state_validation_lock;
-uword *vlib_buffer_state_validation_hash;
-void *vlib_buffer_state_heap;
-
-static clib_error_t *
-buffer_state_validation_init (vlib_main_t * vm)
-{
-  void *oldheap;
-
-  vlib_buffer_state_heap = mheap_alloc (0, 10 << 20);
-
-  oldheap = clib_mem_set_heap (vlib_buffer_state_heap);
-
-  vlib_buffer_state_validation_hash = hash_create (0, sizeof (uword));
-  vec_validate_aligned (vlib_buffer_state_validation_lock, 0,
-			CLIB_CACHE_LINE_BYTES);
-  clib_mem_set_heap (oldheap);
-  return 0;
-}
-
-VLIB_INIT_FUNCTION (buffer_state_validation_init);
-#endif
-
-static vlib_buffer_callbacks_t callbacks = {
-  .vlib_buffer_alloc_cb = &dpdk_buffer_alloc,
-  .vlib_buffer_alloc_from_free_list_cb = &dpdk_buffer_alloc_from_free_list,
-  .vlib_buffer_free_cb = &dpdk_buffer_free,
-  .vlib_buffer_free_no_next_cb = &dpdk_buffer_free_no_next,
-  .vlib_packet_template_init_cb = &dpdk_packet_template_init,
-  .vlib_buffer_delete_free_list_cb = &dpdk_buffer_delete_free_list,
-};
-
-static clib_error_t *
-dpdk_buffer_init (vlib_main_t * vm)
-{
-  vlib_buffer_cb_register (vm, &callbacks);
-  return 0;
-}
-
-VLIB_INIT_FUNCTION (dpdk_buffer_init);
-
-/** @endcond */
-/*
- * fd.io coding-style-patch-verification: ON
- *
- * Local Variables:
- * eval: (c-set-style "gnu")
- * End:
- */
diff --git a/src/vnet/devices/dpdk/cli.c b/src/vnet/devices/dpdk/cli.c
deleted file mode 100644
index 99998862..00000000
--- a/src/vnet/devices/dpdk/cli.c
+++ /dev/null
@@ -1,2079 +0,0 @@
-/*
- * Copyright (c) 2015 Cisco and/or its affiliates.
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at:
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-#include 
-#include 
-#include 
-#include 
-#include 
-
-#include 
-#include 
-#include 
-#include 
-
-#include "dpdk_priv.h"
-
-/**
- * @file
- * @brief CLI for DPDK Abstraction Layer and pcap Tx Trace.
- *
- * This file contains the source code for CLI for DPDK
- * Abstraction Layer and pcap Tx Trace.
- */
-
-
-static clib_error_t *
-get_hqos (u32 hw_if_index, u32 subport_id, dpdk_device_t ** xd,
-	  dpdk_device_config_t ** devconf)
-{
-  dpdk_main_t *dm = &dpdk_main;
-  vnet_hw_interface_t *hw;
-  struct rte_eth_dev_info dev_info;
-  uword *p = 0;
-  clib_error_t *error = NULL;
-
-
-  if (hw_if_index == (u32) ~ 0)
-    {
-      error = clib_error_return (0, "please specify valid interface name");
-      goto done;
-    }
-
-  if (subport_id != 0)
-    {
-      error = clib_error_return (0, "Invalid subport");
-      goto done;
-    }
-
-  hw = vnet_get_hw_interface (dm->vnet_main, hw_if_index);
-  *xd = vec_elt_at_index (dm->devices, hw->dev_instance);
-
-  rte_eth_dev_info_get ((*xd)->device_index, &dev_info);
-  if (dev_info.pci_dev)
-    {				/* bonded interface has no pci info */
-      vlib_pci_addr_t pci_addr;
-
-      pci_addr.domain = dev_info.pci_dev->addr.domain;
-      pci_addr.bus = dev_info.pci_dev->addr.bus;
-      pci_addr.slot = dev_info.pci_dev->addr.devid;
-      pci_addr.function = dev_info.pci_dev->addr.function;
-
-      p =
-	hash_get (dm->conf->device_config_index_by_pci_addr, pci_addr.as_u32);
-    }
-
-  if (p)
-    (*devconf) = pool_elt_at_index (dm->conf->dev_confs, p[0]);
-  else
-    (*devconf) = &dm->conf->default_devconf;
-
-done:
-  return error;
-}
-
-static clib_error_t *
-pcap_trace_command_fn (vlib_main_t * vm,
-		       unformat_input_t * input, vlib_cli_command_t * cmd)
-{
-#define PCAP_DEF_PKT_TO_CAPTURE (100)
-
-  unformat_input_t _line_input, *line_input = &_line_input;
-  dpdk_main_t *dm = &dpdk_main;
-  u8 *filename;
-  u8 *chroot_filename = 0;
-  u32 max = 0;
-  int enabled = 0;
-  int errorFlag = 0;
-  clib_error_t *error = 0;
-
-  /* Get a line of input. */
-  if (!unformat_user (input, unformat_line_input, line_input))
-    return 0;
-
-  while (unformat_check_input (line_input) != UNFORMAT_END_OF_INPUT)
-    {
-      if (unformat (line_input, "on"))
-	{
-	  if (dm->tx_pcap_enable == 0)
-	    {
-	      enabled = 1;
-	    }
-	  else
-	    {
-	      vlib_cli_output (vm, "pcap tx capture already on...");
-	      errorFlag = 1;
-	      break;
-	    }
-	}
-      else if (unformat (line_input, "off"))
-	{
-	  if (dm->tx_pcap_enable)
-	    {
-	      vlib_cli_output (vm, "captured %d pkts...",
-			       dm->pcap_main.n_packets_captured + 1);
-	      if (dm->pcap_main.n_packets_captured)
-		{
-		  dm->pcap_main.n_packets_to_capture =
-		    dm->pcap_main.n_packets_captured;
-		  error = pcap_write (&dm->pcap_main);
-		  if (error)
-		    clib_error_report (error);
-		  else
-		    vlib_cli_output (vm, "saved to %s...", dm->pcap_filename);
-		}
-
-	      dm->tx_pcap_enable = 0;
-	    }
-	  else
-	    {
-	      vlib_cli_output (vm, "pcap tx capture already off...");
-	      errorFlag = 1;
-	      break;
-	    }
-	}
-      else if (unformat (line_input, "max %d", &max))
-	{
-	  if (dm->tx_pcap_enable)
-	    {
-	      vlib_cli_output (vm,
-			       "can't change max value while pcap tx capture active...");
-	      errorFlag = 1;
-	      break;
-	    }
-	}
-      else if (unformat (line_input, "intfc %U",
-			 unformat_vnet_sw_interface, dm->vnet_main,
-			 &dm->pcap_sw_if_index))
-	;
-
-      else if (unformat (line_input, "intfc any"))
-	{
-	  dm->pcap_sw_if_index = 0;
-	}
-      else if (unformat (line_input, "file %s", &filename))
-	{
-	  if (dm->tx_pcap_enable)
-	    {
-	      vlib_cli_output (vm,
-			       "can't change file while pcap tx capture active...");
-	      errorFlag = 1;
-	      break;
-	    }
-
-	  /* Brain-police user path input */
-	  if (strstr ((char *) filename, "..")
-	      || index ((char *) filename, '/'))
-	    {
-	      vlib_cli_output (vm, "illegal characters in filename '%s'",
-			       filename);
-	      vlib_cli_output (vm,
-			       "Hint: Only filename, do not enter directory structure.");
-	      vec_free (filename);
-	      errorFlag = 1;
-	      break;
-	    }
-
-	  chroot_filename = format (0, "/tmp/%s%c", filename, 0);
-	  vec_free (filename);
-	}
-      else if (unformat (line_input, "status"))
-	{
-	  if (dm->pcap_sw_if_index == 0)
-	    {
-	      vlib_cli_output (vm, "max is %d for any interface to file %s",
-			       dm->
-			       pcap_pkts_to_capture ? dm->pcap_pkts_to_capture
-			       : PCAP_DEF_PKT_TO_CAPTURE,
-			       dm->
-			       pcap_filename ? dm->pcap_filename : (u8 *)
-			       "/tmp/vpe.pcap");
-	    }
-	  else
-	    {
-	      vlib_cli_output (vm, "max is %d for interface %U to file %s",
-			       dm->
-			       pcap_pkts_to_capture ? dm->pcap_pkts_to_capture
-			       : PCAP_DEF_PKT_TO_CAPTURE,
-			       format_vnet_sw_if_index_name, dm->vnet_main,
-			       dm->pcap_sw_if_index,
-			       dm->
-			       pcap_filename ? dm->pcap_filename : (u8 *)
-			       "/tmp/vpe.pcap");
-	    }
-
-	  if (dm->tx_pcap_enable == 0)
-	    {
-	      vlib_cli_output (vm, "pcap tx capture is off...");
-	    }
-	  else
-	    {
-	      vlib_cli_output (vm, "pcap tx capture is on: %d of %d pkts...",
-			       dm->pcap_main.n_packets_captured,
-			       dm->pcap_main.n_packets_to_capture);
-	    }
-	  break;
-	}
-
-      else
-	{
-	  error = clib_error_return (0, "unknown input `%U'",
-				     format_unformat_error, line_input);
-	  errorFlag = 1;
-	  break;
-	}
-    }
-  unformat_free (line_input);
-
-
-  if (errorFlag == 0)
-    {
-      /* Since no error, save configured values. */
-      if (chroot_filename)
-	{
-	  if (dm->pcap_filename)
-	    vec_free (dm->pcap_filename);
-	  vec_add1 (chroot_filename, 0);
-	  dm->pcap_filename = chroot_filename;
-	}
-
-      if (max)
-	dm->pcap_pkts_to_capture = max;
-
-
-      if (enabled)
-	{
-	  if (dm->pcap_filename == 0)
-	    dm->pcap_filename = format (0, "/tmp/vpe.pcap%c", 0);
-
-	  memset (&dm->pcap_main, 0, sizeof (dm->pcap_main));
-	  dm->pcap_main.file_name = (char *) dm->pcap_filename;
-	  dm->pcap_main.n_packets_to_capture = PCAP_DEF_PKT_TO_CAPTURE;
-	  if (dm->pcap_pkts_to_capture)
-	    dm->pcap_main.n_packets_to_capture = dm->pcap_pkts_to_capture;
-
-	  dm->pcap_main.packet_type = PCAP_PACKET_TYPE_ethernet;
-	  dm->tx_pcap_enable = 1;
-	  vlib_cli_output (vm, "pcap tx capture on...");
-	}
-    }
-  else if (chroot_filename)
-    vec_free (chroot_filename);
-
-
-  return error;
-}
-
-/*?
- * This command is used to start or stop a packet capture, or show
- * the status of packet capture.
- *
- * This command has the following optional parameters:
- *
- * - on|off - Used to start or stop a packet capture.
- *
- * - max  - Depth of local buffer. Once 'nn' number
- *   of packets have been received, buffer is flushed to file. Once another
- *   'nn' number of packets have been received, buffer is flushed
- *   to file, overwriting previous write. If not entered, value defaults
- *   to 100. Can only be updated if packet capture is off.
- *
- * - intfc |any - Used to specify a given interface,
- *   or use 'any' to run packet capture on all interfaces.
- *   'any' is the default if not provided. Settings from a previous
- *   packet capture are preserved, so 'any' can be used to reset
- *   the interface setting.
- *
- * - file  - Used to specify the output filename. The file will
- *   be placed in the '/tmp' directory, so only the filename is
- *   supported. Directory should not be entered. If file already exists, file
- *   will be overwritten. If no filename is provided, '/tmp/vpe.pcap'
- *   will be used. Can only be updated if packet capture is off.
- *
- * - status - Displays the current status and configured attributes
- *   associated with a packet capture. If packet capture is in progress,
- *   'status' also will return the number of packets currently in
- *   the local buffer. All additional attributes entered on command line
- *   with 'status' will be ingnored and not applied.
- *
- * @cliexpar
- * Example of how to display the status of a tx packet capture when off:
- * @cliexstart{pcap tx trace status}
- * max is 100, for any interface to file /tmp/vpe.pcap
- * pcap tx capture is off...
- * @cliexend
- * Example of how to start a tx packet capture:
- * @cliexstart{pcap tx trace on max 35 intfc GigabitEthernet0/8/0 file vppTest.pcap}
- * pcap tx capture on...
- * @cliexend
- * Example of how to display the status of a tx packet capture in progress:
- * @cliexstart{pcap tx trace status}
- * max is 35, for interface GigabitEthernet0/8/0 to file /tmp/vppTest.pcap
- * pcap tx capture is on: 20 of 35 pkts...
- * @cliexend
- * Example of how to stop a tx packet capture:
- * @cliexstart{vppctl pcap tx trace off}
- * captured 21 pkts...
- * saved to /tmp/vppTest.pcap...
- * @cliexend
-?*/
-/* *INDENT-OFF* */
-VLIB_CLI_COMMAND (pcap_trace_command, static) = {
-    .path = "pcap tx trace",
-    .short_help =
-    "pcap tx trace [on|off] [max ] [intfc |any] [file ] [status]",
-    .function = pcap_trace_command_fn,
-};
-/* *INDENT-ON* */
-
-
-static clib_error_t *
-show_dpdk_buffer (vlib_main_t * vm, unformat_input_t * input,
-		  vlib_cli_command_t * cmd)
-{
-  struct rte_mempool *rmp;
-  int i;
-
-  for (i = 0; i < vec_len (dpdk_main.pktmbuf_pools); i++)
-    {
-      rmp = dpdk_main.pktmbuf_pools[i];
-      if (rmp)
-	{
-	  unsigned count = rte_mempool_avail_count (rmp);
-	  unsigned free_count = rte_mempool_in_use_count (rmp);
-
-	  vlib_cli_output (vm,
-			   "name=\"%s\"  available = %7d allocated = %7d total = %7d\n",
-			   rmp->name, (u32) count, (u32) free_count,
-			   (u32) (count + free_count));
-	}
-      else
-	{
-	  vlib_cli_output (vm, "rte_mempool is NULL (!)\n");
-	}
-    }
-  return 0;
-}
-
-/*?
- * This command displays statistics of each DPDK mempool.
- *
- * @cliexpar
- * Example of how to display DPDK buffer data:
- * @cliexstart{show dpdk buffer}
- * name="mbuf_pool_socket0"  available =   15104 allocated =    1280 total =   16384
- * @cliexend
-?*/
-/* *INDENT-OFF* */
-VLIB_CLI_COMMAND (cmd_show_dpdk_bufferr,static) = {
-    .path = "show dpdk buffer",
-    .short_help = "show dpdk buffer",
-    .function = show_dpdk_buffer,
-    .is_mp_safe = 1,
-};
-/* *INDENT-ON* */
-
-static clib_error_t *
-test_dpdk_buffer (vlib_main_t * vm, unformat_input_t * input,
-		  vlib_cli_command_t * cmd)
-{
-  static u32 *allocated_buffers;
-  u32 n_alloc = 0;
-  u32 n_free = 0;
-  u32 first, actual_alloc;
-
-  while (unformat_check_input (input) != UNFORMAT_END_OF_INPUT)
-    {
-      if (unformat (input, "allocate %d", &n_alloc))
-	;
-      else if (unformat (input, "free %d", &n_free))
-	;
-      else
-	break;
-    }
-
-  if (n_free)
-    {
-      if (vec_len (allocated_buffers) < n_free)
-	return clib_error_return (0, "Can't free %d, only %d allocated",
-				  n_free, vec_len (allocated_buffers));
-
-      first = vec_len (allocated_buffers) - n_free;
-      vlib_buffer_free (vm, allocated_buffers + first, n_free);
-      _vec_len (allocated_buffers) = first;
-    }
-  if (n_alloc)
-    {
-      first = vec_len (allocated_buffers);
-      vec_validate (allocated_buffers,
-		    vec_len (allocated_buffers) + n_alloc - 1);
-
-      actual_alloc = vlib_buffer_alloc (vm, allocated_buffers + first,
-					n_alloc);
-      _vec_len (allocated_buffers) = first + actual_alloc;
-
-      if (actual_alloc < n_alloc)
-	vlib_cli_output (vm, "WARNING: only allocated %d buffers",
-			 actual_alloc);
-    }
-
-  vlib_cli_output (vm, "Currently %d buffers allocated",
-		   vec_len (allocated_buffers));
-
-  if (allocated_buffers && vec_len (allocated_buffers) == 0)
-    vec_free (allocated_buffers);
-
-  return 0;
-}
-
-/*?
- * This command tests the allocation and freeing of DPDK buffers.
- * If both 'allocate' and 'free' are entered on the
- * same command, the 'free' is executed first. If no
- * parameters are provided, this command display how many DPDK buffers
- * the test command has allocated.
- *
- * @cliexpar
- * @parblock
- *
- * Example of how to display how many DPDK buffer test command has allcoated:
- * @cliexstart{test dpdk buffer}
- * Currently 0 buffers allocated
- * @cliexend
- *
- * Example of how to allocate DPDK buffers using the test command:
- * @cliexstart{test dpdk buffer allocate 10}
- * Currently 10 buffers allocated
- * @cliexend
- *
- * Example of how to free DPDK buffers allocated by the test command:
- * @cliexstart{test dpdk buffer free 10}
- * Currently 0 buffers allocated
- * @cliexend
- * @endparblock
-?*/
-/* *INDENT-OFF* */
-VLIB_CLI_COMMAND (cmd_test_dpdk_buffer,static) = {
-    .path = "test dpdk buffer",
-    .short_help = "test dpdk buffer [allocate ] [free ]",
-    .function = test_dpdk_buffer,
-    .is_mp_safe = 1,
-};
-/* *INDENT-ON* */
-
-static clib_error_t *
-set_dpdk_if_desc (vlib_main_t * vm, unformat_input_t * input,
-		  vlib_cli_command_t * cmd)
-{
-  unformat_input_t _line_input, *line_input = &_line_input;
-  dpdk_main_t *dm = &dpdk_main;
-  vnet_hw_interface_t *hw;
-  dpdk_device_t *xd;
-  u32 hw_if_index = (u32) ~ 0;
-  u32 nb_rx_desc = (u32) ~ 0;
-  u32 nb_tx_desc = (u32) ~ 0;
-  clib_error_t *error = NULL;
-
-  if (!unformat_user (input, unformat_line_input, line_input))
-    return 0;
-
-  while (unformat_check_input (line_input) != UNFORMAT_END_OF_INPUT)
-    {
-      if (unformat
-	  (line_input, "%U", unformat_vnet_hw_interface, dm->vnet_main,
-	   &hw_if_index))
-	;
-      else if (unformat (line_input, "tx %d", &nb_tx_desc))
-	;
-      else if (unformat (line_input, "rx %d", &nb_rx_desc))
-	;
-      else
-	{
-	  error = clib_error_return (0, "parse error: '%U'",
-				     format_unformat_error, line_input);
-	  goto done;
-	}
-    }
-
-  if (hw_if_index == (u32) ~ 0)
-    {
-      error = clib_error_return (0, "please specify valid interface name");
-      goto done;
-    }
-
-  hw = vnet_get_hw_interface (dm->vnet_main, hw_if_index);
-  xd = vec_elt_at_index (dm->devices, hw->dev_instance);
-
-  if ((xd->flags & DPDK_DEVICE_FLAG_PMD) == 0)
-    {
-      error =
-	clib_error_return (0,
-			   "number of descriptors can be set only for "
-			   "physical devices");
-      goto done;
-    }
-
-  if ((nb_rx_desc == (u32) ~ 0 || nb_rx_desc == xd->nb_rx_desc) &&
-      (nb_tx_desc == (u32) ~ 0 || nb_tx_desc == xd->nb_tx_desc))
-    {
-      error = clib_error_return (0, "nothing changed");
-      goto done;
-    }
-
-  if (nb_rx_desc != (u32) ~ 0)
-    xd->nb_rx_desc = nb_rx_desc;
-
-  if (nb_tx_desc != (u32) ~ 0)
-    xd->nb_tx_desc = nb_tx_desc;
-
-  error = dpdk_port_setup (dm, xd);
-
-done:
-  unformat_free (line_input);
-
-  return error;
-}
-
-/*?
- * This command sets the number of DPDK 'rx' and
- * 'tx' descriptors for the given physical interface. Use
- * the command 'show hardware-interface' to display the
- * current descriptor allocation.
- *
- * @cliexpar
- * Example of how to set the DPDK interface descriptors:
- * @cliexcmd{set dpdk interface descriptors GigabitEthernet0/8/0 rx 512 tx 512}
-?*/
-/* *INDENT-OFF* */
-VLIB_CLI_COMMAND (cmd_set_dpdk_if_desc,static) = {
-    .path = "set dpdk interface descriptors",
-    .short_help = "set dpdk interface descriptors  [rx ] [tx ]",
-    .function = set_dpdk_if_desc,
-};
-/* *INDENT-ON* */
-
-static clib_error_t *
-show_dpdk_if_placement (vlib_main_t * vm, unformat_input_t * input,
-			vlib_cli_command_t * cmd)
-{
-  vlib_thread_main_t *tm = vlib_get_thread_main ();
-  dpdk_main_t *dm = &dpdk_main;
-  dpdk_device_and_queue_t *dq;
-  int cpu;
-
-  if (tm->n_vlib_mains == 1)
-    vlib_cli_output (vm, "All interfaces are handled by main thread");
-
-  for (cpu = 0; cpu < vec_len (dm->devices_by_cpu); cpu++)
-    {
-      if (cpu >= dm->input_cpu_first_index &&
-	  cpu < (dm->input_cpu_first_index + dm->input_cpu_count))
-	vlib_cli_output (vm, "Thread %u (%s at lcore %u):", cpu,
-			 vlib_worker_threads[cpu].name,
-			 vlib_worker_threads[cpu].lcore_id);
-
-      /* *INDENT-OFF* */
-      vec_foreach(dq, dm->devices_by_cpu[cpu])
-        {
-          u32 hw_if_index = dm->devices[dq->device].vlib_hw_if_index;
-          vnet_hw_interface_t * hi =  vnet_get_hw_interface(dm->vnet_main, hw_if_index);
-          vlib_cli_output(vm, "  %v queue %u", hi->name, dq->queue_id);
-        }
-      /* *INDENT-ON* */
-    }
-  return 0;
-}
-
-/*?
- * This command is used to display the thread and core each
- * DPDK interface and queue is assigned too.
- *
- * @cliexpar
- * Example of how to display the DPDK interface placement:
- * @cliexstart{show dpdk interface placement}
- * Thread 1 (vpp_wk_0 at lcore 1):
- *   GigabitEthernet0/8/0 queue 0
- *   GigabitEthernet0/9/0 queue 0
- * Thread 2 (vpp_wk_1 at lcore 2):
- *   GigabitEthernet0/8/0 queue 1
- *   GigabitEthernet0/9/0 queue 1
- * @cliexend
-?*/
-/* *INDENT-OFF* */
-VLIB_CLI_COMMAND (cmd_show_dpdk_if_placement,static) = {
-    .path = "show dpdk interface placement",
-    .short_help = "show dpdk interface placement",
-    .function = show_dpdk_if_placement,
-};
-/* *INDENT-ON* */
-
-static int
-dpdk_device_queue_sort (void *a1, void *a2)
-{
-  dpdk_device_and_queue_t *dq1 = a1;
-  dpdk_device_and_queue_t *dq2 = a2;
-
-  if (dq1->device > dq2->device)
-    return 1;
-  else if (dq1->device < dq2->device)
-    return -1;
-  else if (dq1->queue_id > dq2->queue_id)
-    return 1;
-  else if (dq1->queue_id < dq2->queue_id)
-    return -1;
-  else
-    return 0;
-}
-
-static clib_error_t *
-set_dpdk_if_placement (vlib_main_t * vm, unformat_input_t * input,
-		       vlib_cli_command_t * cmd)
-{
-  unformat_input_t _line_input, *line_input = &_line_input;
-  dpdk_main_t *dm = &dpdk_main;
-  dpdk_device_and_queue_t *dq;
-  vnet_hw_interface_t *hw;
-  dpdk_device_t *xd;
-  u32 hw_if_index = (u32) ~ 0;
-  u32 queue = (u32) 0;
-  u32 cpu = (u32) ~ 0;
-  int i;
-  clib_error_t *error = NULL;
-
-  if (!unformat_user (input, unformat_line_input, line_input))
-    return 0;
-
-  while (unformat_check_input (line_input) != UNFORMAT_END_OF_INPUT)
-    {
-      if (unformat
-	  (line_input, "%U", unformat_vnet_hw_interface, dm->vnet_main,
-	   &hw_if_index))
-	;
-      else if (unformat (line_input, "queue %d", &queue))
-	;
-      else if (unformat (line_input, "thread %d", &cpu))
-	;
-      else
-	{
-	  error = clib_error_return (0, "parse error: '%U'",
-				     format_unformat_error, line_input);
-	  goto done;
-	}
-    }
-
-  if (hw_if_index == (u32) ~ 0)
-    {
-      error = clib_error_return (0, "please specify valid interface name");
-      goto done;
-    }
-
-  if (cpu < dm->input_cpu_first_index ||
-      cpu >= (dm->input_cpu_first_index + dm->input_cpu_count))
-    {
-      error = clib_error_return (0, "please specify valid thread id");
-      goto done;
-    }
-
-  hw = vnet_get_hw_interface (dm->vnet_main, hw_if_index);
-  xd = vec_elt_at_index (dm->devices, hw->dev_instance);
-
-  for (i = 0; i < vec_len (dm->devices_by_cpu); i++)
-    {
-      /* *INDENT-OFF* */
-      vec_foreach(dq, dm->devices_by_cpu[i])
-        {
-          if (hw_if_index == dm->devices[dq->device].vlib_hw_if_index &&
-              queue == dq->queue_id)
-            {
-              if (cpu == i) /* nothing to do */
-                goto done;
-
-              vec_del1(dm->devices_by_cpu[i], dq - dm->devices_by_cpu[i]);
-              vec_add2(dm->devices_by_cpu[cpu], dq, 1);
-              dq->queue_id = queue;
-              dq->device = xd->device_index;
-              xd->cpu_socket_id_by_queue[queue] =
-                rte_lcore_to_socket_id(vlib_worker_threads[cpu].lcore_id);
-
-              vec_sort_with_function(dm->devices_by_cpu[i],
-		                     dpdk_device_queue_sort);
-
-              vec_sort_with_function(dm->devices_by_cpu[cpu],
-		                     dpdk_device_queue_sort);
-
-              if (vec_len(dm->devices_by_cpu[i]) == 0)
-                vlib_node_set_state (vlib_mains[i], dpdk_input_node.index,
-		                     VLIB_NODE_STATE_DISABLED);
-
-              if (vec_len(dm->devices_by_cpu[cpu]) == 1)
-                vlib_node_set_state (vlib_mains[cpu], dpdk_input_node.index,
-		                     VLIB_NODE_STATE_POLLING);
-
-              goto done;
-            }
-        }
-      /* *INDENT-ON* */
-    }
-
-  error = clib_error_return (0, "not found");
-
-done:
-  unformat_free (line_input);
-
-  return error;
-}
-
-/*?
- * This command is used to assign a given interface, and optionally a
- * given queue, to a different thread. This will not create a thread,
- * so the thread must already exist. Use '/etc/vpp/startup.conf'
- * for the initial thread creation. If the 'queue' is not provided,
- * it defaults to 0.
- *
- * @cliexpar
- * Example of how to display the DPDK interface placement:
- * @cliexstart{show dpdk interface placement}
- * Thread 1 (vpp_wk_0 at lcore 1):
- *   GigabitEthernet0/8/0 queue 0
- *   GigabitEthernet0/9/0 queue 0
- * Thread 2 (vpp_wk_1 at lcore 2):
- *   GigabitEthernet0/8/0 queue 1
- *   GigabitEthernet0/9/0 queue 1
- * @cliexend
- * Example of how to assign a DPDK interface and queue to a thread:
- * @cliexcmd{set dpdk interface placement GigabitEthernet0/8/0 queue 1 thread 1}
-?*/
-/* *INDENT-OFF* */
-VLIB_CLI_COMMAND (cmd_set_dpdk_if_placement,static) = {
-    .path = "set dpdk interface placement",
-    .short_help = "set dpdk interface placement  [queue ] thread ",
-    .function = set_dpdk_if_placement,
-};
-/* *INDENT-ON* */
-
-static clib_error_t *
-show_dpdk_if_hqos_placement (vlib_main_t * vm, unformat_input_t * input,
-			     vlib_cli_command_t * cmd)
-{
-  vlib_thread_main_t *tm = vlib_get_thread_main ();
-  dpdk_main_t *dm = &dpdk_main;
-  dpdk_device_and_queue_t *dq;
-  int cpu;
-
-  if (tm->n_vlib_mains == 1)
-    vlib_cli_output (vm, "All interfaces are handled by main thread");
-
-  for (cpu = 0; cpu < vec_len (dm->devices_by_hqos_cpu); cpu++)
-    {
-      if (cpu >= dm->hqos_cpu_first_index &&
-	  cpu < (dm->hqos_cpu_first_index + dm->hqos_cpu_count))
-	vlib_cli_output (vm, "Thread %u (%s at lcore %u):", cpu,
-			 vlib_worker_threads[cpu].name,
-			 vlib_worker_threads[cpu].lcore_id);
-
-      vec_foreach (dq, dm->devices_by_hqos_cpu[cpu])
-      {
-	u32 hw_if_index = dm->devices[dq->device].vlib_hw_if_index;
-	vnet_hw_interface_t *hi =
-	  vnet_get_hw_interface (dm->vnet_main, hw_if_index);
-	vlib_cli_output (vm, "  %v queue %u", hi->name, dq->queue_id);
-      }
-    }
-  return 0;
-}
-
-/*?
- * This command is used to display the thread and core each
- * DPDK output interface and HQoS queue is assigned too.
- *
- * @cliexpar
- * Example of how to display the DPDK output interface and HQoS queue placement:
- * @cliexstart{show dpdk interface hqos placement}
- * Thread 1 (vpp_hqos-threads_0 at lcore 3):
- *   GigabitEthernet0/8/0 queue 0
- * Thread 2 (vpp_hqos-threads_1 at lcore 4):
- *   GigabitEthernet0/9/0 queue 0
- * @cliexend
-?*/
-/* *INDENT-OFF* */
-VLIB_CLI_COMMAND (cmd_show_dpdk_if_hqos_placement, static) = {
-  .path = "show dpdk interface hqos placement",
-  .short_help = "show dpdk interface hqos placement",
-  .function = show_dpdk_if_hqos_placement,
-};
-/* *INDENT-ON* */
-
-static clib_error_t *
-set_dpdk_if_hqos_placement (vlib_main_t * vm, unformat_input_t * input,
-			    vlib_cli_command_t * cmd)
-{
-  unformat_input_t _line_input, *line_input = &_line_input;
-  dpdk_main_t *dm = &dpdk_main;
-  dpdk_device_and_queue_t *dq;
-  vnet_hw_interface_t *hw;
-  dpdk_device_t *xd;
-  u32 hw_if_index = (u32) ~ 0;
-  u32 cpu = (u32) ~ 0;
-  int i;
-  clib_error_t *error = NULL;
-
-  if (!unformat_user (input, unformat_line_input, line_input))
-    return 0;
-
-  while (unformat_check_input (line_input) != UNFORMAT_END_OF_INPUT)
-    {
-      if (unformat
-	  (line_input, "%U", unformat_vnet_hw_interface, dm->vnet_main,
-	   &hw_if_index))
-	;
-      else if (unformat (line_input, "thread %d", &cpu))
-	;
-      else
-	{
-	  error = clib_error_return (0, "parse error: '%U'",
-				     format_unformat_error, line_input);
-	  goto done;
-	}
-    }
-
-  if (hw_if_index == (u32) ~ 0)
-    return clib_error_return (0, "please specify valid interface name");
-
-  if (cpu < dm->hqos_cpu_first_index ||
-      cpu >= (dm->hqos_cpu_first_index + dm->hqos_cpu_count))
-    {
-      error = clib_error_return (0, "please specify valid thread id");
-      goto done;
-    }
-
-  hw = vnet_get_hw_interface (dm->vnet_main, hw_if_index);
-  xd = vec_elt_at_index (dm->devices, hw->dev_instance);
-
-  for (i = 0; i < vec_len (dm->devices_by_hqos_cpu); i++)
-    {
-      vec_foreach (dq, dm->devices_by_hqos_cpu[i])
-      {
-	if (hw_if_index == dm->devices[dq->device].vlib_hw_if_index)
-	  {
-	    if (cpu == i)	/* nothing to do */
-	      goto done;
-
-	    vec_del1 (dm->devices_by_hqos_cpu[i],
-		      dq - dm->devices_by_hqos_cpu[i]);
-	    vec_add2 (dm->devices_by_hqos_cpu[cpu], dq, 1);
-	    dq->queue_id = 0;
-	    dq->device = xd->device_index;
-
-	    vec_sort_with_function (dm->devices_by_hqos_cpu[i],
-				    dpdk_device_queue_sort);
-
-	    vec_sort_with_function (dm->devices_by_hqos_cpu[cpu],
-				    dpdk_device_queue_sort);
-
-	    goto done;
-	  }
-      }
-    }
-
-  error = clib_error_return (0, "not found");
-
-done:
-  unformat_free (line_input);
-
-  return error;
-}
-
-/*?
- * This command is used to assign a given DPDK output interface and
- * HQoS queue to a different thread. This will not create a thread,
- * so the thread must already exist. Use '/etc/vpp/startup.conf'
- * for the initial thread creation. See @ref qos_doc for more details.
- *
- * @cliexpar
- * Example of how to display the DPDK output interface and HQoS queue placement:
- * @cliexstart{show dpdk interface hqos placement}
- * Thread 1 (vpp_hqos-threads_0 at lcore 3):
- *   GigabitEthernet0/8/0 queue 0
- * Thread 2 (vpp_hqos-threads_1 at lcore 4):
- *   GigabitEthernet0/9/0 queue 0
- * @cliexend
- * Example of how to assign a DPDK output interface and HQoS queue to a thread:
- * @cliexcmd{set dpdk interface hqos placement GigabitEthernet0/8/0 thread 2}
-?*/
-/* *INDENT-OFF* */
-VLIB_CLI_COMMAND (cmd_set_dpdk_if_hqos_placement, static) = {
-  .path = "set dpdk interface hqos placement",
-  .short_help = "set dpdk interface hqos placement  thread ",
-  .function = set_dpdk_if_hqos_placement,
-};
-/* *INDENT-ON* */
-
-static clib_error_t *
-set_dpdk_if_hqos_pipe (vlib_main_t * vm, unformat_input_t * input,
-		       vlib_cli_command_t * cmd)
-{
-  unformat_input_t _line_input, *line_input = &_line_input;
-  dpdk_main_t *dm = &dpdk_main;
-  vnet_hw_interface_t *hw;
-  dpdk_device_t *xd;
-  u32 hw_if_index = (u32) ~ 0;
-  u32 subport_id = (u32) ~ 0;
-  u32 pipe_id = (u32) ~ 0;
-  u32 profile_id = (u32) ~ 0;
-  int rv;
-  clib_error_t *error = NULL;
-
-  if (!unformat_user (input, unformat_line_input, line_input))
-    return 0;
-
-  while (unformat_check_input (line_input) != UNFORMAT_END_OF_INPUT)
-    {
-      if (unformat
-	  (line_input, "%U", unformat_vnet_hw_interface, dm->vnet_main,
-	   &hw_if_index))
-	;
-      else if (unformat (line_input, "subport %d", &subport_id))
-	;
-      else if (unformat (line_input, "pipe %d", &pipe_id))
-	;
-      else if (unformat (line_input, "profile %d", &profile_id))
-	;
-      else
-	{
-	  error = clib_error_return (0, "parse error: '%U'",
-				     format_unformat_error, line_input);
-	  goto done;
-	}
-    }
-
-  if (hw_if_index == (u32) ~ 0)
-    {
-      error = clib_error_return (0, "please specify valid interface name");
-      goto done;
-    }
-
-  hw = vnet_get_hw_interface (dm->vnet_main, hw_if_index);
-  xd = vec_elt_at_index (dm->devices, hw->dev_instance);
-
-  rv =
-    rte_sched_pipe_config (xd->hqos_ht->hqos, subport_id, pipe_id,
-			   profile_id);
-  if (rv)
-    {
-      error = clib_error_return (0, "pipe configuration failed");
-      goto done;
-    }
-
-done:
-  unformat_free (line_input);
-
-  return error;
-}
-
-/*?
- * This command is used to change the profile associate with a HQoS pipe. The
- * '' is zero based. Use the command
- * 'show dpdk interface hqos' to display the content of each profile.
- * See @ref qos_doc for more details.
- *
- * @note
- * Currently there is not an API to create a new HQoS pipe profile. One is
- * created by default in the code (search for 'hqos_pipe_params_default'').
- * Additional profiles can be created in code and code recompiled. Then use this
- * command to assign it.
- *
- * @cliexpar
- * Example of how to assign a new profile to a HQoS pipe:
- * @cliexcmd{set dpdk interface hqos pipe GigabitEthernet0/8/0 subport 0 pipe 2 profile 1}
-?*/
-/* *INDENT-OFF* */
-VLIB_CLI_COMMAND (cmd_set_dpdk_if_hqos_pipe, static) =
-{
-  .path = "set dpdk interface hqos pipe",
-  .short_help = "set dpdk interface hqos pipe  subport  pipe  "
-                  "profile ",
-  .function = set_dpdk_if_hqos_pipe,
-};
-/* *INDENT-ON* */
-
-static clib_error_t *
-set_dpdk_if_hqos_subport (vlib_main_t * vm, unformat_input_t * input,
-			  vlib_cli_command_t * cmd)
-{
-  unformat_input_t _line_input, *line_input = &_line_input;
-  dpdk_main_t *dm = &dpdk_main;
-  dpdk_device_t *xd = NULL;
-  u32 hw_if_index = (u32) ~ 0;
-  u32 subport_id = (u32) ~ 0;
-  struct rte_sched_subport_params p;
-  int rv;
-  clib_error_t *error = NULL;
-  u32 tb_rate = (u32) ~ 0;
-  u32 tb_size = (u32) ~ 0;
-  u32 tc_rate[RTE_SCHED_TRAFFIC_CLASSES_PER_PIPE] =
-    { (u32) ~ 0, (u32) ~ 0, (u32) ~ 0, (u32) ~ 0 };
-  u32 tc_period = (u32) ~ 0;
-  dpdk_device_config_t *devconf = NULL;
-
-  if (!unformat_user (input, unformat_line_input, line_input))
-    return 0;
-
-  while (unformat_check_input (line_input) != UNFORMAT_END_OF_INPUT)
-    {
-      if (unformat
-	  (line_input, "%U", unformat_vnet_hw_interface, dm->vnet_main,
-	   &hw_if_index))
-	;
-      else if (unformat (line_input, "subport %d", &subport_id))
-	;
-      else if (unformat (line_input, "rate %d", &tb_rate))
-	;
-      else if (unformat (line_input, "bktsize %d", &tb_size))
-	;
-      else if (unformat (line_input, "tc0 %d", &tc_rate[0]))
-	;
-      else if (unformat (line_input, "tc1 %d", &tc_rate[1]))
-	;
-      else if (unformat (line_input, "tc2 %d", &tc_rate[2]))
-	;
-      else if (unformat (line_input, "tc3 %d", &tc_rate[3]))
-	;
-      else if (unformat (line_input, "period %d", &tc_period))
-	;
-      else
-	{
-	  error = clib_error_return (0, "parse error: '%U'",
-				     format_unformat_error, line_input);
-	  goto done;
-	}
-    }
-
-  error = get_hqos (hw_if_index, subport_id, &xd, &devconf);
-
-  if (error == NULL)
-    {
-      /* Copy the current values over to local structure. */
-      memcpy (&p, &devconf->hqos.subport[subport_id], sizeof (p));
-
-      /* Update local structure with input values. */
-      if (tb_rate != (u32) ~ 0)
-	{
-	  p.tb_rate = tb_rate;
-	  p.tc_rate[0] = tb_rate;
-	  p.tc_rate[1] = tb_rate;
-	  p.tc_rate[2] = tb_rate;
-	  p.tc_rate[3] = tb_rate;
-	}
-      if (tb_size != (u32) ~ 0)
-	{
-	  p.tb_size = tb_size;
-	}
-      if (tc_rate[0] != (u32) ~ 0)
-	{
-	  p.tc_rate[0] = tc_rate[0];
-	}
-      if (tc_rate[1] != (u32) ~ 0)
-	{
-	  p.tc_rate[1] = tc_rate[1];
-	}
-      if (tc_rate[2] != (u32) ~ 0)
-	{
-	  p.tc_rate[2] = tc_rate[2];
-	}
-      if (tc_rate[3] != (u32) ~ 0)
-	{
-	  p.tc_rate[3] = tc_rate[3];
-	}
-      if (tc_period != (u32) ~ 0)
-	{
-	  p.tc_period = tc_period;
-	}
-
-      /* Apply changes. */
-      rv = rte_sched_subport_config (xd->hqos_ht->hqos, subport_id, &p);
-      if (rv)
-	{
-	  error = clib_error_return (0, "subport configuration failed");
-	  goto done;
-	}
-      else
-	{
-	  /* Successfully applied, so save of the input values. */
-	  memcpy (&devconf->hqos.subport[subport_id], &p, sizeof (p));
-	}
-    }
-
-done:
-  unformat_free (line_input);
-
-  return error;
-}
-
-/*?
- * This command is used to set the subport level parameters such as token
- * bucket rate (bytes per seconds), token bucket size (bytes), traffic class
- * rates (bytes per seconds) and token update period (Milliseconds).
- *
- * By default, the 'rate' is set to 1250000000 bytes/second (10GbE
- * rate) and each of the four traffic classes is set to 100% of the port rate.
- * If the 'rate' is updated by this command, all four traffic classes
- * are assigned the same value. Each of the four traffic classes can be updated
- * individually.
- *
- * @cliexpar
- * Example of how modify the subport attributes for a 1GbE link:
- * @cliexcmd{set dpdk interface hqos subport GigabitEthernet0/8/0 subport 0 rate 125000000}
-?*/
-/* *INDENT-OFF* */
-VLIB_CLI_COMMAND (cmd_set_dpdk_if_hqos_subport, static) = {
-  .path = "set dpdk interface hqos subport",
-  .short_help = "set dpdk interface hqos subport  subport  "
-                 "[rate ] [bktsize ] [tc0 ] [tc1 ] [tc2 ] [tc3 ] "
-                 "[period ]",
-  .function = set_dpdk_if_hqos_subport,
-};
-/* *INDENT-ON* */
-
-static clib_error_t *
-set_dpdk_if_hqos_tctbl (vlib_main_t * vm, unformat_input_t * input,
-			vlib_cli_command_t * cmd)
-{
-  unformat_input_t _line_input, *line_input = &_line_input;
-  vlib_thread_main_t *tm = vlib_get_thread_main ();
-  dpdk_main_t *dm = &dpdk_main;
-  vnet_hw_interface_t *hw;
-  dpdk_device_t *xd;
-  u32 hw_if_index = (u32) ~ 0;
-  u32 tc = (u32) ~ 0;
-  u32 queue = (u32) ~ 0;
-  u32 entry = (u32) ~ 0;
-  u32 val, i;
-  clib_error_t *error = NULL;
-
-  if (!unformat_user (input, unformat_line_input, line_input))
-    return 0;
-
-  while (unformat_check_input (line_input) != UNFORMAT_END_OF_INPUT)
-    {
-      if (unformat
-	  (line_input, "%U", unformat_vnet_hw_interface, dm->vnet_main,
-	   &hw_if_index))
-	;
-      else if (unformat (line_input, "entry %d", &entry))
-	;
-      else if (unformat (line_input, "tc %d", &tc))
-	;
-      else if (unformat (line_input, "queue %d", &queue))
-	;
-      else
-	{
-	  error = clib_error_return (0, "parse error: '%U'",
-				     format_unformat_error, line_input);
-	  goto done;
-	}
-    }
-
-  if (hw_if_index == (u32) ~ 0)
-    {
-      error = clib_error_return (0, "please specify valid interface name");
-      goto done;
-    }
-  if (entry >= 64)
-    {
-      error = clib_error_return (0, "invalid entry");
-      goto done;
-    }
-  if (tc >= RTE_SCHED_TRAFFIC_CLASSES_PER_PIPE)
-    {
-      error = clib_error_return (0, "invalid traffic class");
-      goto done;
-    }
-  if (queue >= RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS)
-    {
-      error = clib_error_return (0, "invalid traffic class queue");
-      goto done;
-    }
-
-  hw = vnet_get_hw_interface (dm->vnet_main, hw_if_index);
-  xd = vec_elt_at_index (dm->devices, hw->dev_instance);
-
-  /* Detect the set of worker threads */
-  uword *p = hash_get_mem (tm->thread_registrations_by_name, "workers");
-  /* Should never happen, shut up Coverity warning */
-  if (p == 0)
-    {
-      error = clib_error_return (0, "no worker registrations?");
-      goto done;
-    }
-
-  vlib_thread_registration_t *tr = (vlib_thread_registration_t *) p[0];
-  int worker_thread_first = tr->first_index;
-  int worker_thread_count = tr->count;
-
-  val = tc * RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS + queue;
-  for (i = 0; i < worker_thread_count; i++)
-    xd->hqos_wt[worker_thread_first + i].hqos_tc_table[entry] = val;
-
-done:
-  unformat_free (line_input);
-
-  return error;
-}
-
-/*?
- * This command is used to set the traffic class translation table. The
- * traffic class translation table is used to map 64 values (0-63) to one of
- * four traffic class and one of four HQoS input queue. Use the 'show
- * dpdk interface hqos' command to display the traffic class translation
- * table. See @ref qos_doc for more details.
- *
- * This command has the following parameters:
- *
- * -  - Used to specify the output interface.
- *
- * - entry  - Mapped value (0-63) to assign traffic class and queue to.
- *
- * - tc  - Traffic class (0-3) to be used by the provided mapped value.
- *
- * - queue  - HQoS input queue (0-3) to be used by the provided mapped value.
- *
- * @cliexpar
- * Example of how modify the traffic class translation table:
- * @cliexcmd{set dpdk interface hqos tctbl GigabitEthernet0/8/0 entry 16 tc 2 queue 2}
-?*/
-/* *INDENT-OFF* */
-VLIB_CLI_COMMAND (cmd_set_dpdk_if_hqos_tctbl, static) = {
-  .path = "set dpdk interface hqos tctbl",
-  .short_help = "set dpdk interface hqos tctbl  entry  tc  queue ",
-  .function = set_dpdk_if_hqos_tctbl,
-};
-/* *INDENT-ON* */
-
-static clib_error_t *
-set_dpdk_if_hqos_pktfield (vlib_main_t * vm, unformat_input_t * input,
-			   vlib_cli_command_t * cmd)
-{
-  unformat_input_t _line_input, *line_input = &_line_input;
-  vlib_thread_main_t *tm = vlib_get_thread_main ();
-  dpdk_main_t *dm = &dpdk_main;
-  clib_error_t *error = NULL;
-
-  /* Device specific data */
-  struct rte_eth_dev_info dev_info;
-  dpdk_device_config_t *devconf = 0;
-  vnet_hw_interface_t *hw;
-  dpdk_device_t *xd;
-  u32 hw_if_index = (u32) ~ 0;
-
-  /* Detect the set of worker threads */
-  uword *p = hash_get_mem (tm->thread_registrations_by_name, "workers");
-  /* Should never happen, shut up Coverity warning */
-  if (p == 0)
-    return clib_error_return (0, "no worker registrations?");
-
-  vlib_thread_registration_t *tr = (vlib_thread_registration_t *) p[0];
-  int worker_thread_first = tr->first_index;
-  int worker_thread_count = tr->count;
-
-  /* Packet field configuration */
-  u64 mask = (u64) ~ 0;
-  u32 id = (u32) ~ 0;
-  u32 offset = (u32) ~ 0;
-
-  /* HQoS params */
-  u32 n_subports_per_port, n_pipes_per_subport, tctbl_size;
-
-  u32 i;
-
-  /* Parse input arguments */
-  if (!unformat_user (input, unformat_line_input, line_input))
-    return 0;
-
-  while (unformat_check_input (line_input) != UNFORMAT_END_OF_INPUT)
-    {
-      if (unformat
-	  (line_input, "%U", unformat_vnet_hw_interface, dm->vnet_main,
-	   &hw_if_index))
-	;
-      else if (unformat (line_input, "id subport"))
-	id = 0;
-      else if (unformat (line_input, "id pipe"))
-	id = 1;
-      else if (unformat (line_input, "id tc"))
-	id = 2;
-      else if (unformat (line_input, "id %d", &id))
-	;
-      else if (unformat (line_input, "offset %d", &offset))
-	;
-      else if (unformat (line_input, "mask %llx", &mask))
-	;
-      else
-	{
-	  error = clib_error_return (0, "parse error: '%U'",
-				     format_unformat_error, line_input);
-	  goto done;
-	}
-    }
-
-  /* Get interface */
-  if (hw_if_index == (u32) ~ 0)
-    {
-      error = clib_error_return (0, "please specify valid interface name");
-      goto done;
-    }
-
-  hw = vnet_get_hw_interface (dm->vnet_main, hw_if_index);
-  xd = vec_elt_at_index (dm->devices, hw->dev_instance);
-
-  rte_eth_dev_info_get (xd->device_index, &dev_info);
-  if (dev_info.pci_dev)
-    {				/* bonded interface has no pci info */
-      vlib_pci_addr_t pci_addr;
-
-      pci_addr.domain = dev_info.pci_dev->addr.domain;
-      pci_addr.bus = dev_info.pci_dev->addr.bus;
-      pci_addr.slot = dev_info.pci_dev->addr.devid;
-      pci_addr.function = dev_info.pci_dev->addr.function;
-
-      p =
-	hash_get (dm->conf->device_config_index_by_pci_addr, pci_addr.as_u32);
-    }
-
-  if (p)
-    devconf = pool_elt_at_index (dm->conf->dev_confs, p[0]);
-  else
-    devconf = &dm->conf->default_devconf;
-
-  if (devconf->hqos_enabled == 0)
-    {
-      vlib_cli_output (vm, "HQoS disabled for this interface");
-      goto done;
-    }
-
-  n_subports_per_port = devconf->hqos.port.n_subports_per_port;
-  n_pipes_per_subport = devconf->hqos.port.n_pipes_per_subport;
-  tctbl_size = RTE_DIM (devconf->hqos.tc_table);
-
-  /* Validate packet field configuration: id, offset and mask */
-  if (id >= 3)
-    {
-      error = clib_error_return (0, "invalid packet field id");
-      goto done;
-    }
-
-  switch (id)
-    {
-    case 0:
-      if (dpdk_hqos_validate_mask (mask, n_subports_per_port) != 0)
-	{
-	  error = clib_error_return (0, "invalid subport ID mask "
-				     "(n_subports_per_port = %u)",
-				     n_subports_per_port);
-	  goto done;
-	}
-      break;
-    case 1:
-      if (dpdk_hqos_validate_mask (mask, n_pipes_per_subport) != 0)
-	{
-	  error = clib_error_return (0, "invalid pipe ID mask "
-				     "(n_pipes_per_subport = %u)",
-				     n_pipes_per_subport);
-	  goto done;
-	}
-      break;
-    case 2:
-    default:
-      if (dpdk_hqos_validate_mask (mask, tctbl_size) != 0)
-	{
-	  error = clib_error_return (0, "invalid TC table index mask "
-				     "(TC table size = %u)", tctbl_size);
-	  goto done;
-	}
-    }
-
-  /* Propagate packet field configuration to all workers */
-  for (i = 0; i < worker_thread_count; i++)
-    switch (id)
-      {
-      case 0:
-	xd->hqos_wt[worker_thread_first + i].hqos_field0_slabpos = offset;
-	xd->hqos_wt[worker_thread_first + i].hqos_field0_slabmask = mask;
-	xd->hqos_wt[worker_thread_first + i].hqos_field0_slabshr =
-	  __builtin_ctzll (mask);
-	break;
-      case 1:
-	xd->hqos_wt[worker_thread_first + i].hqos_field1_slabpos = offset;
-	xd->hqos_wt[worker_thread_first + i].hqos_field1_slabmask = mask;
-	xd->hqos_wt[worker_thread_first + i].hqos_field1_slabshr =
-	  __builtin_ctzll (mask);
-	break;
-      case 2:
-      default:
-	xd->hqos_wt[worker_thread_first + i].hqos_field2_slabpos = offset;
-	xd->hqos_wt[worker_thread_first + i].hqos_field2_slabmask = mask;
-	xd->hqos_wt[worker_thread_first + i].hqos_field2_slabshr =
-	  __builtin_ctzll (mask);
-      }
-
-done:
-  unformat_free (line_input);
-
-  return error;
-}
-
-/*?
- * This command is used to set the packet fields required for classifiying the
- * incoming packet. As a result of classification process, packet field
- * information will be mapped to 5 tuples (subport, pipe, traffic class, pipe,
- * color) and stored in packet mbuf.
- *
- * This command has the following parameters:
- *
- * -  - Used to specify the output interface.
- *
- * - id subport|pipe|tc - Classification occurs across three fields.
- * This parameter indicates which of the three masks are being configured. Legacy
- * code used 0-2 to represent these three fields, so 0-2 is still accepted.
- *   - subport|0 - Currently only one subport is supported, so only
- * an empty mask is supported for the subport classification.
- *   - pipe|1 - Currently, 4096 pipes per subport are supported, so a
- * 12-bit mask should be configure to map to the 0-4095 pipes.
- *   - tc|2 - The translation table (see 'set dpdk interface hqos
- * tctbl' command) maps each value (0-63) into one of the 4 traffic classes
- * per pipe. A 6-bit mask should be configure to map this field to a traffic class.
- *
- * - offset  - Offset in the packet to apply the 64-bit mask for classification.
- * The offset should be on an 8-byte boundary (0,8,16,24..).
- *
- * - mask  - 64-bit mask to apply to packet at the given 'offset'.
- * Bits must be contiguous and should not include '0x'.
- *
- * The default values for the 'pktfield' assumes Ethernet/IPv4/UDP packets with
- * no VLAN. Adjust based on expected packet format and desired classification field.
- * - 'subport' is always empty (offset 0 mask 0000000000000000)
- * - By default, 'pipe' maps to the UDP payload bits 12 .. 23 (offset 40
- * mask 0000000fff000000)
- * - By default, 'tc' maps to the DSCP field in IP header (offset 48 mask
- * 00000000000000fc)
- *
- * @cliexpar
- * Example of how modify the 'pipe' classification filter to match VLAN:
- * @cliexcmd{set dpdk interface hqos pktfield GigabitEthernet0/8/0 id pipe offset 8 mask 0000000000000FFF}
-?*/
-/* *INDENT-OFF* */
-VLIB_CLI_COMMAND (cmd_set_dpdk_if_hqos_pktfield, static) = {
-  .path = "set dpdk interface hqos pktfield",
-  .short_help = "set dpdk interface hqos pktfield  id subport|pipe|tc offset  "
-                 "mask ",
-  .function = set_dpdk_if_hqos_pktfield,
-};
-/* *INDENT-ON* */
-
-static clib_error_t *
-show_dpdk_if_hqos (vlib_main_t * vm, unformat_input_t * input,
-		   vlib_cli_command_t * cmd)
-{
-  unformat_input_t _line_input, *line_input = &_line_input;
-  vlib_thread_main_t *tm = vlib_get_thread_main ();
-  dpdk_main_t *dm = &dpdk_main;
-  vnet_hw_interface_t *hw;
-  dpdk_device_t *xd;
-  dpdk_device_config_hqos_t *cfg;
-  dpdk_device_hqos_per_hqos_thread_t *ht;
-  dpdk_device_hqos_per_worker_thread_t *wk;
-  u32 *tctbl;
-  u32 hw_if_index = (u32) ~ 0;
-  u32 profile_id, subport_id, i;
-  struct rte_eth_dev_info dev_info;
-  dpdk_device_config_t *devconf = 0;
-  vlib_thread_registration_t *tr;
-  uword *p = 0;
-  clib_error_t *error = NULL;
-
-  if (!unformat_user (input, unformat_line_input, line_input))
-    return 0;
-
-  while (unformat_check_input (line_input) != UNFORMAT_END_OF_INPUT)
-    {
-      if (unformat
-	  (line_input, "%U", unformat_vnet_hw_interface, dm->vnet_main,
-	   &hw_if_index))
-	;
-      else
-	{
-	  error = clib_error_return (0, "parse error: '%U'",
-				     format_unformat_error, line_input);
-	  goto done;
-	}
-    }
-
-  if (hw_if_index == (u32) ~ 0)
-    {
-      error = clib_error_return (0, "please specify interface name!!");
-      goto done;
-    }
-
-  hw = vnet_get_hw_interface (dm->vnet_main, hw_if_index);
-  xd = vec_elt_at_index (dm->devices, hw->dev_instance);
-
-  rte_eth_dev_info_get (xd->device_index, &dev_info);
-  if (dev_info.pci_dev)
-    {				/* bonded interface has no pci info */
-      vlib_pci_addr_t pci_addr;
-
-      pci_addr.domain = dev_info.pci_dev->addr.domain;
-      pci_addr.bus = dev_info.pci_dev->addr.bus;
-      pci_addr.slot = dev_info.pci_dev->addr.devid;
-      pci_addr.function = dev_info.pci_dev->addr.function;
-
-      p =
-	hash_get (dm->conf->device_config_index_by_pci_addr, pci_addr.as_u32);
-    }
-
-  if (p)
-    devconf = pool_elt_at_index (dm->conf->dev_confs, p[0]);
-  else
-    devconf = &dm->conf->default_devconf;
-
-  if (devconf->hqos_enabled == 0)
-    {
-      vlib_cli_output (vm, "HQoS disabled for this interface");
-      goto done;
-    }
-
-  /* Detect the set of worker threads */
-  p = hash_get_mem (tm->thread_registrations_by_name, "workers");
-
-  /* Should never happen, shut up Coverity warning */
-  if (p == 0)
-    {
-      error = clib_error_return (0, "no worker registrations?");
-      goto done;
-    }
-
-  tr = (vlib_thread_registration_t *) p[0];
-
-  cfg = &devconf->hqos;
-  ht = xd->hqos_ht;
-  wk = &xd->hqos_wt[tr->first_index];
-  tctbl = wk->hqos_tc_table;
-
-  vlib_cli_output (vm, " Thread:");
-  vlib_cli_output (vm, "   Input SWQ size = %u packets", cfg->swq_size);
-  vlib_cli_output (vm, "   Enqueue burst size = %u packets",
-		   ht->hqos_burst_enq);
-  vlib_cli_output (vm, "   Dequeue burst size = %u packets",
-		   ht->hqos_burst_deq);
-
-  vlib_cli_output (vm,
-		   "   Packet field 0: slab position = %4u, slab bitmask = 0x%016llx   (subport)",
-		   wk->hqos_field0_slabpos, wk->hqos_field0_slabmask);
-  vlib_cli_output (vm,
-		   "   Packet field 1: slab position = %4u, slab bitmask = 0x%016llx   (pipe)",
-		   wk->hqos_field1_slabpos, wk->hqos_field1_slabmask);
-  vlib_cli_output (vm,
-		   "   Packet field 2: slab position = %4u, slab bitmask = 0x%016llx   (tc)",
-		   wk->hqos_field2_slabpos, wk->hqos_field2_slabmask);
-  vlib_cli_output (vm,
-		   "   Packet field 2  tc translation table: ([Mapped Value Range]: tc/queue tc/queue ...)");
-  vlib_cli_output (vm,
-		   "     [ 0 .. 15]: "
-		   "%u/%u %u/%u %u/%u %u/%u %u/%u %u/%u %u/%u %u/%u %u/%u %u/%u %u/%u %u/%u %u/%u %u/%u %u/%u %u/%u",
-		   tctbl[0] / RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
-		   tctbl[0] % RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
-		   tctbl[1] / RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
-		   tctbl[1] % RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
-		   tctbl[2] / RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
-		   tctbl[2] % RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
-		   tctbl[3] / RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
-		   tctbl[3] % RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
-		   tctbl[4] / RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
-		   tctbl[4] % RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
-		   tctbl[5] / RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
-		   tctbl[5] % RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
-		   tctbl[6] / RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
-		   tctbl[6] % RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
-		   tctbl[7] / RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
-		   tctbl[7] % RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
-		   tctbl[8] / RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
-		   tctbl[8] % RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
-		   tctbl[9] / RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
-		   tctbl[9] % RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
-		   tctbl[10] / RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
-		   tctbl[10] % RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
-		   tctbl[11] / RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
-		   tctbl[11] % RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
-		   tctbl[12] / RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
-		   tctbl[12] % RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
-		   tctbl[13] / RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
-		   tctbl[13] % RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
-		   tctbl[14] / RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
-		   tctbl[14] % RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
-		   tctbl[15] / RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
-		   tctbl[15] % RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS);
-  vlib_cli_output (vm,
-		   "     [16 .. 31]: "
-		   "%u/%u %u/%u %u/%u %u/%u %u/%u %u/%u %u/%u %u/%u %u/%u %u/%u %u/%u %u/%u %u/%u %u/%u %u/%u %u/%u",
-		   tctbl[16] / RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
-		   tctbl[16] % RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
-		   tctbl[17] / RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
-		   tctbl[17] % RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
-		   tctbl[18] / RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
-		   tctbl[18] % RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
-		   tctbl[19] / RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
-		   tctbl[19] % RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
-		   tctbl[20] / RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
-		   tctbl[20] % RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
-		   tctbl[21] / RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
-		   tctbl[21] % RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
-		   tctbl[22] / RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
-		   tctbl[22] % RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
-		   tctbl[23] / RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
-		   tctbl[23] % RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
-		   tctbl[24] / RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
-		   tctbl[24] % RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
-		   tctbl[25] / RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
-		   tctbl[25] % RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
-		   tctbl[26] / RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
-		   tctbl[26] % RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
-		   tctbl[27] / RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
-		   tctbl[27] % RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
-		   tctbl[28] / RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
-		   tctbl[28] % RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
-		   tctbl[29] / RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
-		   tctbl[29] % RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
-		   tctbl[30] / RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
-		   tctbl[30] % RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
-		   tctbl[31] / RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
-		   tctbl[31] % RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS);
-  vlib_cli_output (vm,
-		   "     [32 .. 47]: "
-		   "%u/%u %u/%u %u/%u %u/%u %u/%u %u/%u %u/%u %u/%u %u/%u %u/%u %u/%u %u/%u %u/%u %u/%u %u/%u %u/%u",
-		   tctbl[32] / RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
-		   tctbl[32] % RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
-		   tctbl[33] / RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
-		   tctbl[33] % RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
-		   tctbl[34] / RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
-		   tctbl[34] % RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
-		   tctbl[35] / RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
-		   tctbl[35] % RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
-		   tctbl[36] / RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
-		   tctbl[36] % RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
-		   tctbl[37] / RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
-		   tctbl[37] % RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
-		   tctbl[38] / RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
-		   tctbl[38] % RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
-		   tctbl[39] / RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
-		   tctbl[39] % RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
-		   tctbl[40] / RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
-		   tctbl[40] % RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
-		   tctbl[41] / RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
-		   tctbl[41] % RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
-		   tctbl[42] / RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
-		   tctbl[42] % RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
-		   tctbl[43] / RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
-		   tctbl[43] % RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
-		   tctbl[44] / RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
-		   tctbl[44] % RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
-		   tctbl[45] / RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
-		   tctbl[45] % RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
-		   tctbl[46] / RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
-		   tctbl[46] % RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
-		   tctbl[47] / RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
-		   tctbl[47] % RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS);
-  vlib_cli_output (vm,
-		   "     [48 .. 63]: "
-		   "%u/%u %u/%u %u/%u %u/%u %u/%u %u/%u %u/%u %u/%u %u/%u %u/%u %u/%u %u/%u %u/%u %u/%u %u/%u %u/%u",
-		   tctbl[48] / RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
-		   tctbl[48] % RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
-		   tctbl[49] / RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
-		   tctbl[49] % RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
-		   tctbl[50] / RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
-		   tctbl[50] % RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
-		   tctbl[51] / RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
-		   tctbl[51] % RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
-		   tctbl[52] / RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
-		   tctbl[52] % RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
-		   tctbl[53] / RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
-		   tctbl[53] % RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
-		   tctbl[54] / RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
-		   tctbl[54] % RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
-		   tctbl[55] / RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
-		   tctbl[55] % RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
-		   tctbl[56] / RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
-		   tctbl[56] % RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
-		   tctbl[57] / RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
-		   tctbl[57] % RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
-		   tctbl[58] / RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
-		   tctbl[58] % RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
-		   tctbl[59] / RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
-		   tctbl[59] % RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
-		   tctbl[60] / RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
-		   tctbl[60] % RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
-		   tctbl[61] / RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
-		   tctbl[61] % RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
-		   tctbl[62] / RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
-		   tctbl[62] % RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
-		   tctbl[63] / RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS,
-		   tctbl[63] % RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS);
-  vlib_cli_output (vm, " Port:");
-  vlib_cli_output (vm, "   Rate = %u bytes/second", cfg->port.rate);
-  vlib_cli_output (vm, "   MTU = %u bytes", cfg->port.mtu);
-  vlib_cli_output (vm, "   Frame overhead = %u bytes",
-		   cfg->port.frame_overhead);
-  vlib_cli_output (vm, "   Number of subports = %u",
-		   cfg->port.n_subports_per_port);
-  vlib_cli_output (vm, "   Number of pipes per subport = %u",
-		   cfg->port.n_pipes_per_subport);
-  vlib_cli_output (vm,
-		   "   Packet queue size: TC0 = %u, TC1 = %u, TC2 = %u, TC3 = %u packets",
-		   cfg->port.qsize[0], cfg->port.qsize[1], cfg->port.qsize[2],
-		   cfg->port.qsize[3]);
-  vlib_cli_output (vm, "   Number of pipe profiles = %u",
-		   cfg->port.n_pipe_profiles);
-
-  for (subport_id = 0; subport_id < vec_len (cfg->subport); subport_id++)
-    {
-      vlib_cli_output (vm, " Subport %u:", subport_id);
-      vlib_cli_output (vm, "   Rate = %u bytes/second",
-		       cfg->subport[subport_id].tb_rate);
-      vlib_cli_output (vm, "   Token bucket size = %u bytes",
-		       cfg->subport[subport_id].tb_size);
-      vlib_cli_output (vm,
-		       "   Traffic class rate: TC0 = %u, TC1 = %u, TC2 = %u, TC3 = %u bytes/second",
-		       cfg->subport[subport_id].tc_rate[0],
-		       cfg->subport[subport_id].tc_rate[1],
-		       cfg->subport[subport_id].tc_rate[2],
-		       cfg->subport[subport_id].tc_rate[3]);
-      vlib_cli_output (vm, "   TC period = %u milliseconds",
-		       cfg->subport[subport_id].tc_period);
-    }
-
-  for (profile_id = 0; profile_id < vec_len (cfg->pipe); profile_id++)
-    {
-      vlib_cli_output (vm, " Pipe profile %u:", profile_id);
-      vlib_cli_output (vm, "   Rate = %u bytes/second",
-		       cfg->pipe[profile_id].tb_rate);
-      vlib_cli_output (vm, "   Token bucket size = %u bytes",
-		       cfg->pipe[profile_id].tb_size);
-      vlib_cli_output (vm,
-		       "   Traffic class rate: TC0 = %u, TC1 = %u, TC2 = %u, TC3 = %u bytes/second",
-		       cfg->pipe[profile_id].tc_rate[0],
-		       cfg->pipe[profile_id].tc_rate[1],
-		       cfg->pipe[profile_id].tc_rate[2],
-		       cfg->pipe[profile_id].tc_rate[3]);
-      vlib_cli_output (vm, "   TC period = %u milliseconds",
-		       cfg->pipe[profile_id].tc_period);
-#ifdef RTE_SCHED_SUBPORT_TC_OV
-      vlib_cli_output (vm, "   TC3 oversubscription_weight = %u",
-		       cfg->pipe[profile_id].tc_ov_weight);
-#endif
-
-      for (i = 0; i < RTE_SCHED_TRAFFIC_CLASSES_PER_PIPE; i++)
-	{
-	  vlib_cli_output (vm,
-			   "   TC%u WRR weights: Q0 = %u, Q1 = %u, Q2 = %u, Q3 = %u",
-			   i, cfg->pipe[profile_id].wrr_weights[i * 4],
-			   cfg->pipe[profile_id].wrr_weights[i * 4 + 1],
-			   cfg->pipe[profile_id].wrr_weights[i * 4 + 2],
-			   cfg->pipe[profile_id].wrr_weights[i * 4 + 3]);
-	}
-    }
-
-#ifdef RTE_SCHED_RED
-  vlib_cli_output (vm, " Weighted Random Early Detection (WRED):");
-  for (i = 0; i < RTE_SCHED_TRAFFIC_CLASSES_PER_PIPE; i++)
-    {
-      vlib_cli_output (vm, "   TC%u min: G = %u, Y = %u, R = %u", i,
-		       cfg->port.red_params[i][e_RTE_METER_GREEN].min_th,
-		       cfg->port.red_params[i][e_RTE_METER_YELLOW].min_th,
-		       cfg->port.red_params[i][e_RTE_METER_RED].min_th);
-
-      vlib_cli_output (vm, "   TC%u max: G = %u, Y = %u, R = %u", i,
-		       cfg->port.red_params[i][e_RTE_METER_GREEN].max_th,
-		       cfg->port.red_params[i][e_RTE_METER_YELLOW].max_th,
-		       cfg->port.red_params[i][e_RTE_METER_RED].max_th);
-
-      vlib_cli_output (vm,
-		       "   TC%u inverted probability: G = %u, Y = %u, R = %u",
-		       i, cfg->port.red_params[i][e_RTE_METER_GREEN].maxp_inv,
-		       cfg->port.red_params[i][e_RTE_METER_YELLOW].maxp_inv,
-		       cfg->port.red_params[i][e_RTE_METER_RED].maxp_inv);
-
-      vlib_cli_output (vm, "   TC%u weight: R = %u, Y = %u, R = %u", i,
-		       cfg->port.red_params[i][e_RTE_METER_GREEN].wq_log2,
-		       cfg->port.red_params[i][e_RTE_METER_YELLOW].wq_log2,
-		       cfg->port.red_params[i][e_RTE_METER_RED].wq_log2);
-    }
-#endif
-
-done:
-  unformat_free (line_input);
-
-  return error;
-}
-
-/*?
- * This command is used to display details of an output interface's HQoS
- * settings.
- *
- * @cliexpar
- * Example of how to display HQoS settings for an interfaces:
- * @cliexstart{show dpdk interface hqos GigabitEthernet0/8/0}
- *  Thread:
- *    Input SWQ size = 4096 packets
- *    Enqueue burst size = 256 packets
- *    Dequeue burst size = 220 packets
- *    Packet field 0: slab position =    0, slab bitmask = 0x0000000000000000   (subport)
- *    Packet field 1: slab position =   40, slab bitmask = 0x0000000fff000000   (pipe)
- *    Packet field 2: slab position =    8, slab bitmask = 0x00000000000000fc   (tc)
- *    Packet field 2  tc translation table: ([Mapped Value Range]: tc/queue tc/queue ...)
- *      [ 0 .. 15]: 0/0 0/1 0/2 0/3 1/0 1/1 1/2 1/3 2/0 2/1 2/2 2/3 3/0 3/1 3/2 3/3
- *      [16 .. 31]: 0/0 0/1 0/2 0/3 1/0 1/1 1/2 1/3 2/0 2/1 2/2 2/3 3/0 3/1 3/2 3/3
- *      [32 .. 47]: 0/0 0/1 0/2 0/3 1/0 1/1 1/2 1/3 2/0 2/1 2/2 2/3 3/0 3/1 3/2 3/3
- *      [48 .. 63]: 0/0 0/1 0/2 0/3 1/0 1/1 1/2 1/3 2/0 2/1 2/2 2/3 3/0 3/1 3/2 3/3
- *  Port:
- *    Rate = 1250000000 bytes/second
- *    MTU = 1514 bytes
- *    Frame overhead = 24 bytes
- *    Number of subports = 1
- *    Number of pipes per subport = 4096
- *    Packet queue size: TC0 = 64, TC1 = 64, TC2 = 64, TC3 = 64 packets
- *    Number of pipe profiles = 2
- *  Subport 0:
- *    Rate = 1250000000 bytes/second
- *    Token bucket size = 1000000 bytes
- *    Traffic class rate: TC0 = 1250000000, TC1 = 1250000000, TC2 = 1250000000, TC3 = 1250000000 bytes/second
- *    TC period = 10 milliseconds
- *  Pipe profile 0:
- *    Rate = 305175 bytes/second
- *    Token bucket size = 1000000 bytes
- *    Traffic class rate: TC0 = 305175, TC1 = 305175, TC2 = 305175, TC3 = 305175 bytes/second
- *    TC period = 40 milliseconds
- *    TC0 WRR weights: Q0 = 1, Q1 = 1, Q2 = 1, Q3 = 1
- *    TC1 WRR weights: Q0 = 1, Q1 = 1, Q2 = 1, Q3 = 1
- *    TC2 WRR weights: Q0 = 1, Q1 = 1, Q2 = 1, Q3 = 1
- *    TC3 WRR weights: Q0 = 1, Q1 = 1, Q2 = 1, Q3 = 1
- * @cliexend
-?*/
-/* *INDENT-OFF* */
-VLIB_CLI_COMMAND (cmd_show_dpdk_if_hqos, static) = {
-  .path = "show dpdk interface hqos",
-  .short_help = "show dpdk interface hqos ",
-  .function = show_dpdk_if_hqos,
-};
-
-/* *INDENT-ON* */
-
-static clib_error_t *
-show_dpdk_hqos_queue_stats (vlib_main_t * vm, unformat_input_t * input,
-			    vlib_cli_command_t * cmd)
-{
-  unformat_input_t _line_input, *line_input = &_line_input;
-  clib_error_t *error = NULL;
-#ifdef RTE_SCHED_COLLECT_STATS
-  dpdk_main_t *dm = &dpdk_main;
-  u32 hw_if_index = (u32) ~ 0;
-  u32 subport = (u32) ~ 0;
-  u32 pipe = (u32) ~ 0;
-  u32 tc = (u32) ~ 0;
-  u32 tc_q = (u32) ~ 0;
-  vnet_hw_interface_t *hw;
-  dpdk_device_t *xd;
-  uword *p = 0;
-  struct rte_eth_dev_info dev_info;
-  dpdk_device_config_t *devconf = 0;
-  u32 qindex;
-  struct rte_sched_queue_stats stats;
-  u16 qlen;
-
-  if (!unformat_user (input, unformat_line_input, line_input))
-    return 0;
-
-  while (unformat_check_input (line_input) != UNFORMAT_END_OF_INPUT)
-    {
-      if (unformat
-	  (line_input, "%U", unformat_vnet_hw_interface, dm->vnet_main,
-	   &hw_if_index))
-	;
-
-      else if (unformat (line_input, "subport %d", &subport))
-	;
-
-      else if (unformat (line_input, "pipe %d", &pipe))
-	;
-
-      else if (unformat (line_input, "tc %d", &tc))
-	;
-
-      else if (unformat (line_input, "tc_q %d", &tc_q))
-	;
-
-      else
-	{
-	  error = clib_error_return (0, "parse error: '%U'",
-				     format_unformat_error, line_input);
-	  goto done;
-	}
-    }
-
-  if (hw_if_index == (u32) ~ 0)
-    {
-      error = clib_error_return (0, "please specify interface name!!");
-      goto done;
-    }
-
-  hw = vnet_get_hw_interface (dm->vnet_main, hw_if_index);
-  xd = vec_elt_at_index (dm->devices, hw->dev_instance);
-
-  rte_eth_dev_info_get (xd->device_index, &dev_info);
-  if (dev_info.pci_dev)
-    {				/* bonded interface has no pci info */
-      vlib_pci_addr_t pci_addr;
-
-      pci_addr.domain = dev_info.pci_dev->addr.domain;
-      pci_addr.bus = dev_info.pci_dev->addr.bus;
-      pci_addr.slot = dev_info.pci_dev->addr.devid;
-      pci_addr.function = dev_info.pci_dev->addr.function;
-
-      p =
-	hash_get (dm->conf->device_config_index_by_pci_addr, pci_addr.as_u32);
-    }
-
-  if (p)
-    devconf = pool_elt_at_index (dm->conf->dev_confs, p[0]);
-  else
-    devconf = &dm->conf->default_devconf;
-
-  if (devconf->hqos_enabled == 0)
-    {
-      vlib_cli_output (vm, "HQoS disabled for this interface");
-      goto done;
-    }
-
-  /*
-   * Figure out which queue to query.  cf rte_sched_port_qindex.  (Not sure why
-   * that method isn't made public by DPDK - how _should_ we get the queue ID?)
-   */
-  qindex = subport * devconf->hqos.port.n_pipes_per_subport + pipe;
-  qindex = qindex * RTE_SCHED_TRAFFIC_CLASSES_PER_PIPE + tc;
-  qindex = qindex * RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS + tc_q;
-
-  if (rte_sched_queue_read_stats (xd->hqos_ht->hqos, qindex, &stats, &qlen) !=
-      0)
-    {
-      error = clib_error_return (0, "failed to read stats");
-      goto done;
-    }
-
-  vlib_cli_output (vm, "%=24s%=16s", "Stats Parameter", "Value");
-  vlib_cli_output (vm, "%=24s%=16d", "Packets", stats.n_pkts);
-  vlib_cli_output (vm, "%=24s%=16d", "Packets dropped", stats.n_pkts_dropped);
-#ifdef RTE_SCHED_RED
-  vlib_cli_output (vm, "%=24s%=16d", "Packets dropped (RED)",
-		   stats.n_pkts_red_dropped);
-#endif
-  vlib_cli_output (vm, "%=24s%=16d", "Bytes", stats.n_bytes);
-  vlib_cli_output (vm, "%=24s%=16d", "Bytes dropped", stats.n_bytes_dropped);
-
-#else
-
-  /* Get a line of input */
-  if (!unformat_user (input, unformat_line_input, line_input))
-    return 0;
-
-  vlib_cli_output (vm, "RTE_SCHED_COLLECT_STATS disabled in DPDK");
-  goto done;
-
-#endif
-
-done:
-  unformat_free (line_input);
-
-  return error;
-}
-
-/*?
- * This command is used to display statistics associated with a HQoS traffic class
- * queue.
- *
- * @note
- * Statistic collection by the scheduler is disabled by default in DPDK. In order to
- * turn it on, add the following line to '../vpp/dpdk/Makefile':
- * - $(call set,RTE_SCHED_COLLECT_STATS,y)
- *
- * @cliexpar
- * Example of how to display statistics of HQoS a HQoS traffic class queue:
- * @cliexstart{show dpdk hqos queue GigabitEthernet0/9/0 subport 0 pipe 3181 tc 0 tc_q 0}
- *      Stats Parameter          Value
- *          Packets               140
- *      Packets dropped            0
- *           Bytes               8400
- *       Bytes dropped             0
- * @cliexend
-?*/
-/* *INDENT-OFF* */
-VLIB_CLI_COMMAND (cmd_show_dpdk_hqos_queue_stats, static) = {
-  .path = "show dpdk hqos queue",
-  .short_help = "show dpdk hqos queue  subport  pipe  tc  tc_q ",
-  .function = show_dpdk_hqos_queue_stats,
-};
-/* *INDENT-ON* */
-
-static clib_error_t *
-show_dpdk_version_command_fn (vlib_main_t * vm,
-			      unformat_input_t * input,
-			      vlib_cli_command_t * cmd)
-{
-#define _(a,b,c) vlib_cli_output (vm, "%-25s " b, a ":", c);
-  _("DPDK Version", "%s", rte_version ());
-  _("DPDK EAL init args", "%s", dpdk_config_main.eal_init_args_str);
-#undef _
-  return 0;
-}
-
-/*?
- * This command is used to display the current DPDK version and
- * the list of arguments passed to DPDK when started.
- *
- * @cliexpar
- * Example of how to display how many DPDK buffer test command has allcoated:
- * @cliexstart{show dpdk version}
- * DPDK Version:        DPDK 16.11.0
- * DPDK EAL init args:  -c 1 -n 4 --huge-dir /run/vpp/hugepages --file-prefix vpp -w 0000:00:08.0 -w 0000:00:09.0 --master-lcore 0 --socket-mem 256
- * @cliexend
-?*/
-/* *INDENT-OFF* */
-VLIB_CLI_COMMAND (show_vpe_version_command, static) = {
-  .path = "show dpdk version",
-  .short_help = "show dpdk version",
-  .function = show_dpdk_version_command_fn,
-};
-/* *INDENT-ON* */
-
-clib_error_t *
-dpdk_cli_init (vlib_main_t * vm)
-{
-  return 0;
-}
-
-VLIB_INIT_FUNCTION (dpdk_cli_init);
-
-/*
- * fd.io coding-style-patch-verification: ON
- *
- * Local Variables:
- * eval: (c-set-style "gnu")
- * End:
- */
diff --git a/src/vnet/devices/dpdk/device.c b/src/vnet/devices/dpdk/device.c
deleted file mode 100644
index 17397900..00000000
--- a/src/vnet/devices/dpdk/device.c
+++ /dev/null
@@ -1,852 +0,0 @@
-/*
- * Copyright (c) 2015 Cisco and/or its affiliates.
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at:
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-#include 
-#include 
-#include 
-#include 
-#include 
-
-#include 
-#include 
-
-#include "dpdk_priv.h"
-#include 
-
-#define foreach_dpdk_tx_func_error			\
-  _(BAD_RETVAL, "DPDK tx function returned an error")	\
-  _(RING_FULL, "Tx packet drops (ring full)")	        \
-  _(PKT_DROP, "Tx packet drops (dpdk tx failure)")	\
-  _(REPL_FAIL, "Tx packet drops (replication failure)")
-
-typedef enum
-{
-#define _(f,s) DPDK_TX_FUNC_ERROR_##f,
-  foreach_dpdk_tx_func_error
-#undef _
-    DPDK_TX_FUNC_N_ERROR,
-} dpdk_tx_func_error_t;
-
-static char *dpdk_tx_func_error_strings[] = {
-#define _(n,s) s,
-  foreach_dpdk_tx_func_error
-#undef _
-};
-
-clib_error_t *
-dpdk_set_mac_address (vnet_hw_interface_t * hi, char *address)
-{
-  int error;
-  dpdk_main_t *dm = &dpdk_main;
-  dpdk_device_t *xd = vec_elt_at_index (dm->devices, hi->dev_instance);
-
-  error = rte_eth_dev_default_mac_addr_set (xd->device_index,
-					    (struct ether_addr *) address);
-
-  if (error)
-    {
-      return clib_error_return (0, "mac address set failed: %d", error);
-    }
-  else
-    {
-      vec_reset_length (xd->default_mac_address);
-      vec_add (xd->default_mac_address, address, sizeof (address));
-      return NULL;
-    }
-}
-
-clib_error_t *
-dpdk_set_mc_filter (vnet_hw_interface_t * hi,
-		    struct ether_addr mc_addr_vec[], int naddr)
-{
-  int error;
-  dpdk_main_t *dm = &dpdk_main;
-  dpdk_device_t *xd = vec_elt_at_index (dm->devices, hi->dev_instance);
-
-  error = rte_eth_dev_set_mc_addr_list (xd->device_index, mc_addr_vec, naddr);
-
-  if (error)
-    {
-      return clib_error_return (0, "mc addr list failed: %d", error);
-    }
-  else
-    {
-      return NULL;
-    }
-}
-
-struct rte_mbuf *
-dpdk_replicate_packet_mb (vlib_buffer_t * b)
-{
-  dpdk_main_t *dm = &dpdk_main;
-  struct rte_mbuf **mbufs = 0, *s, *d;
-  u8 nb_segs;
-  unsigned socket_id = rte_socket_id ();
-  int i;
-
-  ASSERT (dm->pktmbuf_pools[socket_id]);
-  s = rte_mbuf_from_vlib_buffer (b);
-  nb_segs = s->nb_segs;
-  vec_validate (mbufs, nb_segs - 1);
-
-  if (rte_pktmbuf_alloc_bulk (dm->pktmbuf_pools[socket_id], mbufs, nb_segs))
-    {
-      vec_free (mbufs);
-      return 0;
-    }
-
-  d = mbufs[0];
-  d->nb_segs = s->nb_segs;
-  d->data_len = s->data_len;
-  d->pkt_len = s->pkt_len;
-  d->data_off = s->data_off;
-  clib_memcpy (d->buf_addr, s->buf_addr, RTE_PKTMBUF_HEADROOM + s->data_len);
-
-  for (i = 1; i < nb_segs; i++)
-    {
-      d->next = mbufs[i];
-      d = mbufs[i];
-      s = s->next;
-      d->data_len = s->data_len;
-      clib_memcpy (d->buf_addr, s->buf_addr,
-		   RTE_PKTMBUF_HEADROOM + s->data_len);
-    }
-
-  d = mbufs[0];
-  vec_free (mbufs);
-  return d;
-}
-
-static void
-dpdk_tx_trace_buffer (dpdk_main_t * dm,
-		      vlib_node_runtime_t * node,
-		      dpdk_device_t * xd,
-		      u16 queue_id, u32 buffer_index, vlib_buffer_t * buffer)
-{
-  vlib_main_t *vm = vlib_get_main ();
-  dpdk_tx_dma_trace_t *t0;
-  struct rte_mbuf *mb;
-
-  mb = rte_mbuf_from_vlib_buffer (buffer);
-
-  t0 = vlib_add_trace (vm, node, buffer, sizeof (t0[0]));
-  t0->queue_index = queue_id;
-  t0->device_index = xd->device_index;
-  t0->buffer_index = buffer_index;
-  clib_memcpy (&t0->mb, mb, sizeof (t0->mb));
-  clib_memcpy (&t0->buffer, buffer,
-	       sizeof (buffer[0]) - sizeof (buffer->pre_data));
-  clib_memcpy (t0->buffer.pre_data, buffer->data + buffer->current_data,
-	       sizeof (t0->buffer.pre_data));
-}
-
-static_always_inline void
-dpdk_validate_rte_mbuf (vlib_main_t * vm, vlib_buffer_t * b,
-			int maybe_multiseg)
-{
-  struct rte_mbuf *mb, *first_mb, *last_mb;
-
-  /* buffer is coming from non-dpdk source so we need to init
-     rte_mbuf header */
-  if (PREDICT_FALSE ((b->flags & VLIB_BUFFER_EXT_HDR_VALID) == 0))
-    {
-      vlib_buffer_t *b2 = b;
-      last_mb = mb = rte_mbuf_from_vlib_buffer (b2);
-      rte_pktmbuf_reset (mb);
-      while (maybe_multiseg && (b2->flags & VLIB_BUFFER_NEXT_PRESENT))
-	{
-	  b2 = vlib_get_buffer (vm, b2->next_buffer);
-	  mb = rte_mbuf_from_vlib_buffer (b2);
-	  rte_pktmbuf_reset (mb);
-	}
-    }
-
-  last_mb = first_mb = mb = rte_mbuf_from_vlib_buffer (b);
-  first_mb->nb_segs = 1;
-  mb->data_len = b->current_length;
-  mb->pkt_len = maybe_multiseg ? vlib_buffer_length_in_chain (vm, b) :
-    b->current_length;
-  mb->data_off = VLIB_BUFFER_PRE_DATA_SIZE + b->current_data;
-
-  while (maybe_multiseg && (b->flags & VLIB_BUFFER_NEXT_PRESENT))
-    {
-      b = vlib_get_buffer (vm, b->next_buffer);
-      mb = rte_mbuf_from_vlib_buffer (b);
-      last_mb->next = mb;
-      last_mb = mb;
-      mb->data_len = b->current_length;
-      mb->pkt_len = b->current_length;
-      mb->data_off = VLIB_BUFFER_PRE_DATA_SIZE + b->current_data;
-      first_mb->nb_segs++;
-      if (PREDICT_FALSE (b->n_add_refs))
-	{
-	  rte_mbuf_refcnt_update (mb, b->n_add_refs);
-	  b->n_add_refs = 0;
-	}
-    }
-}
-
-/*
- * This function calls the dpdk's tx_burst function to transmit the packets
- * on the tx_vector. It manages a lock per-device if the device does not
- * support multiple queues. It returns the number of packets untransmitted
- * on the tx_vector. If all packets are transmitted (the normal case), the
- * function returns 0.
- *
- * The function assumes there is at least one packet on the tx_vector.
- */
-static_always_inline
-  u32 tx_burst_vector_internal (vlib_main_t * vm,
-				dpdk_device_t * xd,
-				struct rte_mbuf **tx_vector)
-{
-  dpdk_main_t *dm = &dpdk_main;
-  u32 n_packets;
-  u32 tx_head;
-  u32 tx_tail;
-  u32 n_retry;
-  int rv;
-  int queue_id;
-  tx_ring_hdr_t *ring;
-
-  ring = vec_header (tx_vector, sizeof (*ring));
-
-  n_packets = ring->tx_head - ring->tx_tail;
-
-  tx_head = ring->tx_head % xd->nb_tx_desc;
-
-  /*
-   * Ensure rte_eth_tx_burst is not called with 0 packets, which can lead to
-   * unpredictable results.
-   */
-  ASSERT (n_packets > 0);
-
-  /*
-   * Check for tx_vector overflow. If this fails it is a system configuration
-   * error. The ring should be sized big enough to handle the largest un-flowed
-   * off burst from a traffic manager. A larger size also helps performance
-   * a bit because it decreases the probability of having to issue two tx_burst
-   * calls due to a ring wrap.
-   */
-  ASSERT (n_packets < xd->nb_tx_desc);
-  ASSERT (ring->tx_tail == 0);
-
-  n_retry = 16;
-  queue_id = vm->cpu_index;
-
-  do
-    {
-      /* start the burst at the tail */
-      tx_tail = ring->tx_tail % xd->nb_tx_desc;
-
-      /*
-       * This device only supports one TX queue,
-       * and we're running multi-threaded...
-       */
-      if (PREDICT_FALSE (xd->lockp != 0))
-	{
-	  queue_id = queue_id % xd->tx_q_used;
-	  while (__sync_lock_test_and_set (xd->lockp[queue_id], 1))
-	    /* zzzz */
-	    queue_id = (queue_id + 1) % xd->tx_q_used;
-	}
-
-      if (PREDICT_FALSE (xd->flags & DPDK_DEVICE_FLAG_HQOS))	/* HQoS ON */
-	{
-	  /* no wrap, transmit in one burst */
-	  dpdk_device_hqos_per_worker_thread_t *hqos =
-	    &xd->hqos_wt[vm->cpu_index];
-
-	  ASSERT (hqos->swq != NULL);
-
-	  dpdk_hqos_metadata_set (hqos,
-				  &tx_vector[tx_tail], tx_head - tx_tail);
-	  rv = rte_ring_sp_enqueue_burst (hqos->swq,
-					  (void **) &tx_vector[tx_tail],
-					  (uint16_t) (tx_head - tx_tail));
-	}
-      else if (PREDICT_TRUE (xd->flags & DPDK_DEVICE_FLAG_PMD))
-	{
-	  /* no wrap, transmit in one burst */
-	  rv = rte_eth_tx_burst (xd->device_index,
-				 (uint16_t) queue_id,
-				 &tx_vector[tx_tail],
-				 (uint16_t) (tx_head - tx_tail));
-	}
-      else
-	{
-	  ASSERT (0);
-	  rv = 0;
-	}
-
-      if (PREDICT_FALSE (xd->lockp != 0))
-	*xd->lockp[queue_id] = 0;
-
-      if (PREDICT_FALSE (rv < 0))
-	{
-	  // emit non-fatal message, bump counter
-	  vnet_main_t *vnm = dm->vnet_main;
-	  vnet_interface_main_t *im = &vnm->interface_main;
-	  u32 node_index;
-
-	  node_index = vec_elt_at_index (im->hw_interfaces,
-					 xd->vlib_hw_if_index)->tx_node_index;
-
-	  vlib_error_count (vm, node_index, DPDK_TX_FUNC_ERROR_BAD_RETVAL, 1);
-	  clib_warning ("rte_eth_tx_burst[%d]: error %d", xd->device_index,
-			rv);
-	  return n_packets;	// untransmitted packets
-	}
-      ring->tx_tail += (u16) rv;
-      n_packets -= (uint16_t) rv;
-    }
-  while (rv && n_packets && (n_retry > 0));
-
-  return n_packets;
-}
-
-static_always_inline void
-dpdk_prefetch_buffer_by_index (vlib_main_t * vm, u32 bi)
-{
-  vlib_buffer_t *b;
-  struct rte_mbuf *mb;
-  b = vlib_get_buffer (vm, bi);
-  mb = rte_mbuf_from_vlib_buffer (b);
-  CLIB_PREFETCH (mb, CLIB_CACHE_LINE_BYTES, LOAD);
-  CLIB_PREFETCH (b, CLIB_CACHE_LINE_BYTES, LOAD);
-}
-
-static_always_inline void
-dpdk_buffer_recycle (vlib_main_t * vm, vlib_node_runtime_t * node,
-		     vlib_buffer_t * b, u32 bi, struct rte_mbuf **mbp)
-{
-  dpdk_main_t *dm = &dpdk_main;
-  u32 my_cpu = vm->cpu_index;
-  struct rte_mbuf *mb_new;
-
-  if (PREDICT_FALSE (b->flags & VLIB_BUFFER_RECYCLE) == 0)
-    return;
-
-  mb_new = dpdk_replicate_packet_mb (b);
-  if (PREDICT_FALSE (mb_new == 0))
-    {
-      vlib_error_count (vm, node->node_index,
-			DPDK_TX_FUNC_ERROR_REPL_FAIL, 1);
-      b->flags |= VLIB_BUFFER_REPL_FAIL;
-    }
-  else
-    *mbp = mb_new;
-
-  vec_add1 (dm->recycle[my_cpu], bi);
-}
-
-/*
- * Transmits the packets on the frame to the interface associated with the
- * node. It first copies packets on the frame to a tx_vector containing the
- * rte_mbuf pointers. It then passes this vector to tx_burst_vector_internal
- * which calls the dpdk tx_burst function.
- */
-static uword
-dpdk_interface_tx (vlib_main_t * vm,
-		   vlib_node_runtime_t * node, vlib_frame_t * f)
-{
-  dpdk_main_t *dm = &dpdk_main;
-  vnet_interface_output_runtime_t *rd = (void *) node->runtime_data;
-  dpdk_device_t *xd = vec_elt_at_index (dm->devices, rd->dev_instance);
-  u32 n_packets = f->n_vectors;
-  u32 n_left;
-  u32 *from;
-  struct rte_mbuf **tx_vector;
-  u16 i;
-  u16 nb_tx_desc = xd->nb_tx_desc;
-  int queue_id;
-  u32 my_cpu;
-  u32 tx_pkts = 0;
-  tx_ring_hdr_t *ring;
-  u32 n_on_ring;
-
-  my_cpu = vm->cpu_index;
-
-  queue_id = my_cpu;
-
-  tx_vector = xd->tx_vectors[queue_id];
-  ring = vec_header (tx_vector, sizeof (*ring));
-
-  n_on_ring = ring->tx_head - ring->tx_tail;
-  from = vlib_frame_vector_args (f);
-
-  ASSERT (n_packets <= VLIB_FRAME_SIZE);
-
-  if (PREDICT_FALSE (n_on_ring + n_packets > nb_tx_desc))
-    {
-      /*
-       * Overflowing the ring should never happen.
-       * If it does then drop the whole frame.
-       */
-      vlib_error_count (vm, node->node_index, DPDK_TX_FUNC_ERROR_RING_FULL,
-			n_packets);
-
-      while (n_packets--)
-	{
-	  u32 bi0 = from[n_packets];
-	  vlib_buffer_t *b0 = vlib_get_buffer (vm, bi0);
-	  struct rte_mbuf *mb0 = rte_mbuf_from_vlib_buffer (b0);
-	  rte_pktmbuf_free (mb0);
-	}
-      return n_on_ring;
-    }
-
-  if (PREDICT_FALSE (dm->tx_pcap_enable))
-    {
-      n_left = n_packets;
-      while (n_left > 0)
-	{
-	  u32 bi0 = from[0];
-	  vlib_buffer_t *b0 = vlib_get_buffer (vm, bi0);
-	  if (dm->pcap_sw_if_index == 0 ||
-	      dm->pcap_sw_if_index == vnet_buffer (b0)->sw_if_index[VLIB_TX])
-	    pcap_add_buffer (&dm->pcap_main, vm, bi0, 512);
-	  from++;
-	  n_left--;
-	}
-    }
-
-  from = vlib_frame_vector_args (f);
-  n_left = n_packets;
-  i = ring->tx_head % nb_tx_desc;
-
-  while (n_left >= 8)
-    {
-      u32 bi0, bi1, bi2, bi3;
-      struct rte_mbuf *mb0, *mb1, *mb2, *mb3;
-      vlib_buffer_t *b0, *b1, *b2, *b3;
-      u32 or_flags;
-
-      dpdk_prefetch_buffer_by_index (vm, from[4]);
-      dpdk_prefetch_buffer_by_index (vm, from[5]);
-      dpdk_prefetch_buffer_by_index (vm, from[6]);
-      dpdk_prefetch_buffer_by_index (vm, from[7]);
-
-      bi0 = from[0];
-      bi1 = from[1];
-      bi2 = from[2];
-      bi3 = from[3];
-      from += 4;
-
-      b0 = vlib_get_buffer (vm, bi0);
-      b1 = vlib_get_buffer (vm, bi1);
-      b2 = vlib_get_buffer (vm, bi2);
-      b3 = vlib_get_buffer (vm, bi3);
-
-      or_flags = b0->flags | b1->flags | b2->flags | b3->flags;
-
-      if (or_flags & VLIB_BUFFER_NEXT_PRESENT)
-	{
-	  dpdk_validate_rte_mbuf (vm, b0, 1);
-	  dpdk_validate_rte_mbuf (vm, b1, 1);
-	  dpdk_validate_rte_mbuf (vm, b2, 1);
-	  dpdk_validate_rte_mbuf (vm, b3, 1);
-	}
-      else
-	{
-	  dpdk_validate_rte_mbuf (vm, b0, 0);
-	  dpdk_validate_rte_mbuf (vm, b1, 0);
-	  dpdk_validate_rte_mbuf (vm, b2, 0);
-	  dpdk_validate_rte_mbuf (vm, b3, 0);
-	}
-
-      mb0 = rte_mbuf_from_vlib_buffer (b0);
-      mb1 = rte_mbuf_from_vlib_buffer (b1);
-      mb2 = rte_mbuf_from_vlib_buffer (b2);
-      mb3 = rte_mbuf_from_vlib_buffer (b3);
-
-      if (PREDICT_FALSE (or_flags & VLIB_BUFFER_RECYCLE))
-	{
-	  dpdk_buffer_recycle (vm, node, b0, bi0, &mb0);
-	  dpdk_buffer_recycle (vm, node, b1, bi1, &mb1);
-	  dpdk_buffer_recycle (vm, node, b2, bi2, &mb2);
-	  dpdk_buffer_recycle (vm, node, b3, bi3, &mb3);
-
-	  /* dont enqueue packets if replication failed as they must
-	     be sent back to recycle */
-	  if (PREDICT_TRUE ((b0->flags & VLIB_BUFFER_REPL_FAIL) == 0))
-	    tx_vector[i++ % nb_tx_desc] = mb0;
-	  if (PREDICT_TRUE ((b1->flags & VLIB_BUFFER_REPL_FAIL) == 0))
-	    tx_vector[i++ % nb_tx_desc] = mb1;
-	  if (PREDICT_TRUE ((b2->flags & VLIB_BUFFER_REPL_FAIL) == 0))
-	    tx_vector[i++ % nb_tx_desc] = mb2;
-	  if (PREDICT_TRUE ((b3->flags & VLIB_BUFFER_REPL_FAIL) == 0))
-	    tx_vector[i++ % nb_tx_desc] = mb3;
-	}
-      else
-	{
-	  if (PREDICT_FALSE (i + 3 >= nb_tx_desc))
-	    {
-	      tx_vector[i++ % nb_tx_desc] = mb0;
-	      tx_vector[i++ % nb_tx_desc] = mb1;
-	      tx_vector[i++ % nb_tx_desc] = mb2;
-	      tx_vector[i++ % nb_tx_desc] = mb3;
-	      i %= nb_tx_desc;
-	    }
-	  else
-	    {
-	      tx_vector[i++] = mb0;
-	      tx_vector[i++] = mb1;
-	      tx_vector[i++] = mb2;
-	      tx_vector[i++] = mb3;
-	    }
-	}
-
-
-      if (PREDICT_FALSE (node->flags & VLIB_NODE_FLAG_TRACE))
-	{
-	  if (b0->flags & VLIB_BUFFER_IS_TRACED)
-	    dpdk_tx_trace_buffer (dm, node, xd, queue_id, bi0, b0);
-	  if (b1->flags & VLIB_BUFFER_IS_TRACED)
-	    dpdk_tx_trace_buffer (dm, node, xd, queue_id, bi1, b1);
-	  if (b2->flags & VLIB_BUFFER_IS_TRACED)
-	    dpdk_tx_trace_buffer (dm, node, xd, queue_id, bi2, b2);
-	  if (b3->flags & VLIB_BUFFER_IS_TRACED)
-	    dpdk_tx_trace_buffer (dm, node, xd, queue_id, bi3, b3);
-	}
-
-      n_left -= 4;
-    }
-  while (n_left > 0)
-    {
-      u32 bi0;
-      struct rte_mbuf *mb0;
-      vlib_buffer_t *b0;
-
-      bi0 = from[0];
-      from++;
-
-      b0 = vlib_get_buffer (vm, bi0);
-
-      dpdk_validate_rte_mbuf (vm, b0, 1);
-
-      mb0 = rte_mbuf_from_vlib_buffer (b0);
-      dpdk_buffer_recycle (vm, node, b0, bi0, &mb0);
-
-      if (PREDICT_FALSE (node->flags & VLIB_NODE_FLAG_TRACE))
-	if (b0->flags & VLIB_BUFFER_IS_TRACED)
-	  dpdk_tx_trace_buffer (dm, node, xd, queue_id, bi0, b0);
-
-      if (PREDICT_TRUE ((b0->flags & VLIB_BUFFER_REPL_FAIL) == 0))
-	{
-	  tx_vector[i % nb_tx_desc] = mb0;
-	  i++;
-	}
-      n_left--;
-    }
-
-  /* account for additional packets in the ring */
-  ring->tx_head += n_packets;
-  n_on_ring = ring->tx_head - ring->tx_tail;
-
-  /* transmit as many packets as possible */
-  n_packets = tx_burst_vector_internal (vm, xd, tx_vector);
-
-  /*
-   * tx_pkts is the number of packets successfully transmitted
-   * This is the number originally on ring minus the number remaining on ring
-   */
-  tx_pkts = n_on_ring - n_packets;
-
-  {
-    /* If there is no callback then drop any non-transmitted packets */
-    if (PREDICT_FALSE (n_packets))
-      {
-	vlib_simple_counter_main_t *cm;
-	vnet_main_t *vnm = vnet_get_main ();
-
-	cm = vec_elt_at_index (vnm->interface_main.sw_if_counters,
-			       VNET_INTERFACE_COUNTER_TX_ERROR);
-
-	vlib_increment_simple_counter (cm, my_cpu, xd->vlib_sw_if_index,
-				       n_packets);
-
-	vlib_error_count (vm, node->node_index, DPDK_TX_FUNC_ERROR_PKT_DROP,
-			  n_packets);
-
-	while (n_packets--)
-	  rte_pktmbuf_free (tx_vector[ring->tx_tail + n_packets]);
-      }
-
-    /* Reset head/tail to avoid unnecessary wrap */
-    ring->tx_head = 0;
-    ring->tx_tail = 0;
-  }
-
-  /* Recycle replicated buffers */
-  if (PREDICT_FALSE (vec_len (dm->recycle[my_cpu])))
-    {
-      vlib_buffer_free (vm, dm->recycle[my_cpu],
-			vec_len (dm->recycle[my_cpu]));
-      _vec_len (dm->recycle[my_cpu]) = 0;
-    }
-
-  ASSERT (ring->tx_head >= ring->tx_tail);
-
-  return tx_pkts;
-}
-
-static void
-dpdk_clear_hw_interface_counters (u32 instance)
-{
-  dpdk_main_t *dm = &dpdk_main;
-  dpdk_device_t *xd = vec_elt_at_index (dm->devices, instance);
-
-  /*
-   * Set the "last_cleared_stats" to the current stats, so that
-   * things appear to clear from a display perspective.
-   */
-  dpdk_update_counters (xd, vlib_time_now (dm->vlib_main));
-
-  clib_memcpy (&xd->last_cleared_stats, &xd->stats, sizeof (xd->stats));
-  clib_memcpy (xd->last_cleared_xstats, xd->xstats,
-	       vec_len (xd->last_cleared_xstats) *
-	       sizeof (xd->last_cleared_xstats[0]));
-
-}
-
-static clib_error_t *
-dpdk_interface_admin_up_down (vnet_main_t * vnm, u32 hw_if_index, u32 flags)
-{
-  vnet_hw_interface_t *hif = vnet_get_hw_interface (vnm, hw_if_index);
-  uword is_up = (flags & VNET_SW_INTERFACE_FLAG_ADMIN_UP) != 0;
-  dpdk_main_t *dm = &dpdk_main;
-  dpdk_device_t *xd = vec_elt_at_index (dm->devices, hif->dev_instance);
-  int rv = 0;
-
-  if (is_up)
-    {
-      f64 now = vlib_time_now (dm->vlib_main);
-
-      if ((xd->flags & DPDK_DEVICE_FLAG_ADMIN_UP) == 0)
-	{
-	  rv = rte_eth_dev_start (xd->device_index);
-	  if (!rv && xd->default_mac_address)
-	    rv = rte_eth_dev_default_mac_addr_set (xd->device_index,
-						   (struct ether_addr *)
-						   xd->default_mac_address);
-	}
-
-      if (xd->flags & DPDK_DEVICE_FLAG_PROMISC)
-	rte_eth_promiscuous_enable (xd->device_index);
-      else
-	rte_eth_promiscuous_disable (xd->device_index);
-
-      rte_eth_allmulticast_enable (xd->device_index);
-      xd->flags |= DPDK_DEVICE_FLAG_ADMIN_UP;
-      dpdk_update_counters (xd, now);
-      dpdk_update_link_state (xd, now);
-    }
-  else
-    {
-      xd->flags &= ~DPDK_DEVICE_FLAG_ADMIN_UP;
-
-      rte_eth_allmulticast_disable (xd->device_index);
-      vnet_hw_interface_set_flags (vnm, xd->vlib_hw_if_index, 0);
-      rte_eth_dev_stop (xd->device_index);
-
-      /* For bonded interface, stop slave links */
-      if (xd->pmd == VNET_DPDK_PMD_BOND)
-	{
-	  u8 slink[16];
-	  int nlink = rte_eth_bond_slaves_get (xd->device_index, slink, 16);
-	  while (nlink >= 1)
-	    {
-	      u8 dpdk_port = slink[--nlink];
-	      rte_eth_dev_stop (dpdk_port);
-	    }
-	}
-    }
-
-  if (rv < 0)
-    clib_warning ("rte_eth_dev_%s error: %d", is_up ? "start" : "stop", rv);
-
-  return /* no error */ 0;
-}
-
-/*
- * Dynamically redirect all pkts from a specific interface
- * to the specified node
- */
-static void
-dpdk_set_interface_next_node (vnet_main_t * vnm, u32 hw_if_index,
-			      u32 node_index)
-{
-  dpdk_main_t *xm = &dpdk_main;
-  vnet_hw_interface_t *hw = vnet_get_hw_interface (vnm, hw_if_index);
-  dpdk_device_t *xd = vec_elt_at_index (xm->devices, hw->dev_instance);
-
-  /* Shut off redirection */
-  if (node_index == ~0)
-    {
-      xd->per_interface_next_index = node_index;
-      return;
-    }
-
-  xd->per_interface_next_index =
-    vlib_node_add_next (xm->vlib_main, dpdk_input_node.index, node_index);
-}
-
-
-static clib_error_t *
-dpdk_subif_add_del_function (vnet_main_t * vnm,
-			     u32 hw_if_index,
-			     struct vnet_sw_interface_t *st, int is_add)
-{
-  dpdk_main_t *xm = &dpdk_main;
-  vnet_hw_interface_t *hw = vnet_get_hw_interface (vnm, hw_if_index);
-  dpdk_device_t *xd = vec_elt_at_index (xm->devices, hw->dev_instance);
-  vnet_sw_interface_t *t = (vnet_sw_interface_t *) st;
-  int r, vlan_offload;
-  u32 prev_subifs = xd->num_subifs;
-  clib_error_t *err = 0;
-
-  if (is_add)
-    xd->num_subifs++;
-  else if (xd->num_subifs)
-    xd->num_subifs--;
-
-  if ((xd->flags & DPDK_DEVICE_FLAG_PMD) == 0)
-    goto done;
-
-  /* currently we program VLANS only for IXGBE VF and I40E VF */
-  if ((xd->pmd != VNET_DPDK_PMD_IXGBEVF) && (xd->pmd != VNET_DPDK_PMD_I40EVF))
-    goto done;
-
-  if (t->sub.eth.flags.no_tags == 1)
-    goto done;
-
-  if ((t->sub.eth.flags.one_tag != 1) || (t->sub.eth.flags.exact_match != 1))
-    {
-      xd->num_subifs = prev_subifs;
-      err = clib_error_return (0, "unsupported VLAN setup");
-      goto done;
-    }
-
-  vlan_offload = rte_eth_dev_get_vlan_offload (xd->device_index);
-  vlan_offload |= ETH_VLAN_FILTER_OFFLOAD;
-
-  if ((r = rte_eth_dev_set_vlan_offload (xd->device_index, vlan_offload)))
-    {
-      xd->num_subifs = prev_subifs;
-      err = clib_error_return (0, "rte_eth_dev_set_vlan_offload[%d]: err %d",
-			       xd->device_index, r);
-      goto done;
-    }
-
-
-  if ((r =
-       rte_eth_dev_vlan_filter (xd->device_index, t->sub.eth.outer_vlan_id,
-				is_add)))
-    {
-      xd->num_subifs = prev_subifs;
-      err = clib_error_return (0, "rte_eth_dev_vlan_filter[%d]: err %d",
-			       xd->device_index, r);
-      goto done;
-    }
-
-done:
-  if (xd->num_subifs)
-    xd->flags |= DPDK_DEVICE_FLAG_HAVE_SUBIF;
-  else
-    xd->flags &= ~DPDK_DEVICE_FLAG_HAVE_SUBIF;
-
-  return err;
-}
-
-/* *INDENT-OFF* */
-VNET_DEVICE_CLASS (dpdk_device_class) = {
-  .name = "dpdk",
-  .tx_function = dpdk_interface_tx,
-  .tx_function_n_errors = DPDK_TX_FUNC_N_ERROR,
-  .tx_function_error_strings = dpdk_tx_func_error_strings,
-  .format_device_name = format_dpdk_device_name,
-  .format_device = format_dpdk_device,
-  .format_tx_trace = format_dpdk_tx_dma_trace,
-  .clear_counters = dpdk_clear_hw_interface_counters,
-  .admin_up_down_function = dpdk_interface_admin_up_down,
-  .subif_add_del_function = dpdk_subif_add_del_function,
-  .rx_redirect_to_node = dpdk_set_interface_next_node,
-  .mac_addr_change_function = dpdk_set_mac_address,
-};
-
-VLIB_DEVICE_TX_FUNCTION_MULTIARCH (dpdk_device_class, dpdk_interface_tx)
-/* *INDENT-ON* */
-
-#define UP_DOWN_FLAG_EVENT 1
-
-uword
-admin_up_down_process (vlib_main_t * vm,
-		       vlib_node_runtime_t * rt, vlib_frame_t * f)
-{
-  clib_error_t *error = 0;
-  uword event_type;
-  uword *event_data = 0;
-  u32 sw_if_index;
-  u32 flags;
-
-  while (1)
-    {
-      vlib_process_wait_for_event (vm);
-
-      event_type = vlib_process_get_events (vm, &event_data);
-
-      dpdk_main.admin_up_down_in_progress = 1;
-
-      switch (event_type)
-	{
-	case UP_DOWN_FLAG_EVENT:
-	  {
-	    if (vec_len (event_data) == 2)
-	      {
-		sw_if_index = event_data[0];
-		flags = event_data[1];
-		error =
-		  vnet_sw_interface_set_flags (vnet_get_main (), sw_if_index,
-					       flags);
-		clib_error_report (error);
-	      }
-	  }
-	  break;
-	}
-
-      vec_reset_length (event_data);
-
-      dpdk_main.admin_up_down_in_progress = 0;
-
-    }
-  return 0;			/* or not */
-}
-
-/* *INDENT-OFF* */
-VLIB_REGISTER_NODE (admin_up_down_process_node,static) = {
-    .function = admin_up_down_process,
-    .type = VLIB_NODE_TYPE_PROCESS,
-    .name = "admin-up-down-process",
-    .process_log2_n_stack_bytes = 17,  // 256KB
-};
-/* *INDENT-ON* */
-
-/*
- * fd.io coding-style-patch-verification: ON
- *
- * Local Variables:
- * eval: (c-set-style "gnu")
- * End:
- */
diff --git a/src/vnet/devices/dpdk/dir.dox b/src/vnet/devices/dpdk/dir.dox
deleted file mode 100644
index 43e36753..00000000
--- a/src/vnet/devices/dpdk/dir.dox
+++ /dev/null
@@ -1,27 +0,0 @@
-/*
- * Copyright (c) 2017 Cisco and/or its affiliates.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at:
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-/* Doxygen directory documentation */
-
-/**
-@dir
-@brief DPDK Abstraction Layer.
-
-This directory contains the source code for the DPDK abstraction layer.
-
-*/
-/*? %%clicmd:group_label DPDK and pcap tx %% ?*/
-/*? %%syscfg:group_label DPDK and pcap tx %% ?*/
diff --git a/src/vnet/devices/dpdk/dpdk.api b/src/vnet/devices/dpdk/dpdk.api
deleted file mode 100644
index 21215d45..00000000
--- a/src/vnet/devices/dpdk/dpdk.api
+++ /dev/null
@@ -1,103 +0,0 @@
-/*
- * Copyright (c) 2015-2016 Cisco and/or its affiliates.
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at:
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-/** \brief DPDK interface HQoS pipe profile set request
-    @param client_index - opaque cookie to identify the sender
-    @param context - sender context, to match reply w/ request
-    @param sw_if_index - the interface
-    @param subport - subport ID
-    @param pipe - pipe ID within its subport
-    @param profile - pipe profile ID
-*/
-define sw_interface_set_dpdk_hqos_pipe {
-    u32 client_index;
-    u32 context;
-    u32 sw_if_index;
-    u32 subport;
-    u32 pipe;
-    u32 profile;
-};
-
-/** \brief DPDK interface HQoS pipe profile set reply
-    @param context - sender context, to match reply w/ request
-    @param retval - request return code
-*/
-define sw_interface_set_dpdk_hqos_pipe_reply {
-    u32 context;
-    i32 retval;
-};
-
-/** \brief DPDK interface HQoS subport parameters set request
-    @param client_index - opaque cookie to identify the sender
-    @param context - sender context, to match reply w/ request
-    @param sw_if_index - the interface
-    @param subport - subport ID
-    @param tb_rate - subport token bucket rate (measured in bytes/second)
-    @param tb_size - subport token bucket size (measured in credits)
-    @param tc_rate - subport traffic class 0 .. 3 rates (measured in bytes/second)
-    @param tc_period - enforcement period for rates (measured in milliseconds)
-*/
-define sw_interface_set_dpdk_hqos_subport {
-    u32 client_index;
-    u32 context;
-    u32 sw_if_index;
-    u32 subport;
-    u32 tb_rate;
-    u32 tb_size;
-    u32 tc_rate[4];
-    u32 tc_period;
-};
-
-/** \brief DPDK interface HQoS subport parameters set reply
-    @param context - sender context, to match reply w/ request
-    @param retval - request return code
-*/
-define sw_interface_set_dpdk_hqos_subport_reply {
-    u32 context;
-    i32 retval;
-};
-
-/** \brief DPDK interface HQoS tctbl entry set request
-    @param client_index - opaque cookie to identify the sender
-    @param context - sender context, to match reply w/ request
-    @param sw_if_index - the interface
-    @param entry - entry index ID
-    @param tc - traffic class (0 .. 3)
-    @param queue - traffic class queue (0 .. 3)
-*/
-define sw_interface_set_dpdk_hqos_tctbl {
-    u32 client_index;
-    u32 context;
-    u32 sw_if_index;
-    u32 entry;
-    u32 tc;
-    u32 queue;
-};
-
-/** \brief DPDK interface HQoS tctbl entry set reply
-    @param context - sender context, to match reply w/ request
-    @param retval - request return code
-*/
-define sw_interface_set_dpdk_hqos_tctbl_reply {
-    u32 context;
-    i32 retval;
-};
-
-/*
- * Local Variables:
- * eval: (c-set-style "gnu")
- * End:
- */
- 
\ No newline at end of file
diff --git a/src/vnet/devices/dpdk/dpdk.h b/src/vnet/devices/dpdk/dpdk.h
deleted file mode 100644
index bf9f2768..00000000
--- a/src/vnet/devices/dpdk/dpdk.h
+++ /dev/null
@@ -1,487 +0,0 @@
-/*
- * Copyright (c) 2015 Cisco and/or its affiliates.
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at:
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-#ifndef __included_dpdk_h__
-#define __included_dpdk_h__
-
-/* $$$$ We should rename always_inline -> clib_always_inline */
-#undef always_inline
-
-#include 
-
-#include 
-#include 
-#include 
-#include 
-#include 
-#include 
-#include 
-#include 
-#include 
-#include 
-#include 
-#include 
-#include 
-#include 
-#include 
-#include 
-#include 
-#include 
-#include 
-#include 
-#include 
-#include 
-#include 
-#include 
-#include 
-#include 
-#include 
-#include 
-
-#include 
-#include 
-
-#if CLIB_DEBUG > 0
-#define always_inline static inline
-#else
-#define always_inline static inline __attribute__ ((__always_inline__))
-#endif
-
-#include 
-
-#define NB_MBUF   (16<<10)
-
-extern vnet_device_class_t dpdk_device_class;
-extern vlib_node_registration_t dpdk_input_node;
-extern vlib_node_registration_t handoff_dispatch_node;
-
-#define foreach_dpdk_pmd          \
-  _ ("net_thunderx", THUNDERX)    \
-  _ ("net_e1000_em", E1000EM)     \
-  _ ("net_e1000_igb", IGB)        \
-  _ ("net_e1000_igb_vf", IGBVF)   \
-  _ ("net_ixgbe", IXGBE)          \
-  _ ("net_ixgbe_vf", IXGBEVF)     \
-  _ ("net_i40e", I40E)            \
-  _ ("net_i40e_vf", I40EVF)       \
-  _ ("net_virtio", VIRTIO)        \
-  _ ("net_enic", ENIC)            \
-  _ ("net_vmxnet3", VMXNET3)      \
-  _ ("AF_PACKET PMD", AF_PACKET)  \
-  _ ("rte_bond_pmd", BOND)        \
-  _ ("net_fm10k", FM10K)          \
-  _ ("net_cxgbe", CXGBE)          \
-  _ ("net_mlx5", MLX5)            \
-  _ ("net_dpaa2", DPAA2)
-
-typedef enum
-{
-  VNET_DPDK_PMD_NONE,
-#define _(s,f) VNET_DPDK_PMD_##f,
-  foreach_dpdk_pmd
-#undef _
-    VNET_DPDK_PMD_UNKNOWN,	/* must be last */
-} dpdk_pmd_t;
-
-typedef enum
-{
-  VNET_DPDK_PORT_TYPE_ETH_1G,
-  VNET_DPDK_PORT_TYPE_ETH_10G,
-  VNET_DPDK_PORT_TYPE_ETH_40G,
-  VNET_DPDK_PORT_TYPE_ETH_100G,
-  VNET_DPDK_PORT_TYPE_ETH_BOND,
-  VNET_DPDK_PORT_TYPE_ETH_SWITCH,
-  VNET_DPDK_PORT_TYPE_AF_PACKET,
-  VNET_DPDK_PORT_TYPE_UNKNOWN,
-} dpdk_port_type_t;
-
-/*
- * The header for the tx_vector in dpdk_device_t.
- * Head and tail are indexes into the tx_vector and are of type
- * u64 so they never overflow.
- */
-typedef struct
-{
-  u64 tx_head;
-  u64 tx_tail;
-} tx_ring_hdr_t;
-
-typedef struct
-{
-  struct rte_ring *swq;
-
-  u64 hqos_field0_slabmask;
-  u32 hqos_field0_slabpos;
-  u32 hqos_field0_slabshr;
-  u64 hqos_field1_slabmask;
-  u32 hqos_field1_slabpos;
-  u32 hqos_field1_slabshr;
-  u64 hqos_field2_slabmask;
-  u32 hqos_field2_slabpos;
-  u32 hqos_field2_slabshr;
-  u32 hqos_tc_table[64];
-} dpdk_device_hqos_per_worker_thread_t;
-
-typedef struct
-{
-  struct rte_ring **swq;
-  struct rte_mbuf **pkts_enq;
-  struct rte_mbuf **pkts_deq;
-  struct rte_sched_port *hqos;
-  u32 hqos_burst_enq;
-  u32 hqos_burst_deq;
-  u32 pkts_enq_len;
-  u32 swq_pos;
-  u32 flush_count;
-} dpdk_device_hqos_per_hqos_thread_t;
-
-typedef struct
-{
-  CLIB_CACHE_LINE_ALIGN_MARK (cacheline0);
-  volatile u32 **lockp;
-
-  /* Instance ID */
-  u32 device_index;
-
-  u32 vlib_hw_if_index;
-  u32 vlib_sw_if_index;
-
-  /* next node index if we decide to steal the rx graph arc */
-  u32 per_interface_next_index;
-
-  /* dpdk rte_mbuf rx and tx vectors, VLIB_FRAME_SIZE */
-  struct rte_mbuf ***tx_vectors;	/* one per worker thread */
-  struct rte_mbuf ***rx_vectors;
-
-  /* vector of traced contexts, per device */
-  u32 **d_trace_buffers;
-
-  dpdk_pmd_t pmd:8;
-  i8 cpu_socket;
-
-  u16 flags;
-#define DPDK_DEVICE_FLAG_ADMIN_UP           (1 << 0)
-#define DPDK_DEVICE_FLAG_PROMISC            (1 << 1)
-#define DPDK_DEVICE_FLAG_PMD                (1 << 2)
-#define DPDK_DEVICE_FLAG_PMD_SUPPORTS_PTYPE (1 << 3)
-#define DPDK_DEVICE_FLAG_MAYBE_MULTISEG     (1 << 4)
-#define DPDK_DEVICE_FLAG_HAVE_SUBIF         (1 << 5)
-#define DPDK_DEVICE_FLAG_HQOS               (1 << 6)
-
-  u16 nb_tx_desc;
-    CLIB_CACHE_LINE_ALIGN_MARK (cacheline1);
-
-  u8 *interface_name_suffix;
-
-  /* number of sub-interfaces */
-  u16 num_subifs;
-
-  /* PMD related */
-  u16 tx_q_used;
-  u16 rx_q_used;
-  u16 nb_rx_desc;
-  u16 *cpu_socket_id_by_queue;
-  struct rte_eth_conf port_conf;
-  struct rte_eth_txconf tx_conf;
-
-  /* HQoS related */
-  dpdk_device_hqos_per_worker_thread_t *hqos_wt;
-  dpdk_device_hqos_per_hqos_thread_t *hqos_ht;
-
-  /* af_packet */
-  u8 af_packet_port_id;
-
-  struct rte_eth_link link;
-  f64 time_last_link_update;
-
-  struct rte_eth_stats stats;
-  struct rte_eth_stats last_stats;
-  struct rte_eth_stats last_cleared_stats;
-  struct rte_eth_xstat *xstats;
-  struct rte_eth_xstat *last_cleared_xstats;
-  f64 time_last_stats_update;
-  dpdk_port_type_t port_type;
-
-  /* mac address */
-  u8 *default_mac_address;
-} dpdk_device_t;
-
-#define DPDK_STATS_POLL_INTERVAL      (10.0)
-#define DPDK_MIN_STATS_POLL_INTERVAL  (0.001)	/* 1msec */
-
-#define DPDK_LINK_POLL_INTERVAL       (3.0)
-#define DPDK_MIN_LINK_POLL_INTERVAL   (0.001)	/* 1msec */
-
-typedef struct
-{
-  u32 device;
-  u16 queue_id;
-} dpdk_device_and_queue_t;
-
-#ifndef DPDK_HQOS_DBG_BYPASS
-#define DPDK_HQOS_DBG_BYPASS 0
-#endif
-
-#ifndef HQOS_FLUSH_COUNT_THRESHOLD
-#define HQOS_FLUSH_COUNT_THRESHOLD              100000
-#endif
-
-typedef struct dpdk_device_config_hqos_t
-{
-  u32 hqos_thread;
-  u32 hqos_thread_valid;
-
-  u32 swq_size;
-  u32 burst_enq;
-  u32 burst_deq;
-
-  u32 pktfield0_slabpos;
-  u32 pktfield1_slabpos;
-  u32 pktfield2_slabpos;
-  u64 pktfield0_slabmask;
-  u64 pktfield1_slabmask;
-  u64 pktfield2_slabmask;
-  u32 tc_table[64];
-
-  struct rte_sched_port_params port;
-  struct rte_sched_subport_params *subport;
-  struct rte_sched_pipe_params *pipe;
-  uint32_t *pipe_map;
-} dpdk_device_config_hqos_t;
-
-int dpdk_hqos_validate_mask (u64 mask, u32 n);
-void dpdk_device_config_hqos_pipe_profile_default (dpdk_device_config_hqos_t *
-						   hqos, u32 pipe_profile_id);
-void dpdk_device_config_hqos_default (dpdk_device_config_hqos_t * hqos);
-clib_error_t *dpdk_port_setup_hqos (dpdk_device_t * xd,
-				    dpdk_device_config_hqos_t * hqos);
-void dpdk_hqos_metadata_set (dpdk_device_hqos_per_worker_thread_t * hqos,
-			     struct rte_mbuf **pkts, u32 n_pkts);
-
-#define foreach_dpdk_device_config_item \
-  _ (num_rx_queues) \
-  _ (num_tx_queues) \
-  _ (num_rx_desc) \
-  _ (num_tx_desc) \
-  _ (rss_fn)
-
-typedef struct
-{
-  vlib_pci_addr_t pci_addr;
-  u8 is_blacklisted;
-  u8 vlan_strip_offload;
-#define DPDK_DEVICE_VLAN_STRIP_DEFAULT 0
-#define DPDK_DEVICE_VLAN_STRIP_OFF 1
-#define DPDK_DEVICE_VLAN_STRIP_ON  2
-
-#define _(x) uword x;
-    foreach_dpdk_device_config_item
-#undef _
-    clib_bitmap_t * workers;
-  u32 hqos_enabled;
-  dpdk_device_config_hqos_t hqos;
-} dpdk_device_config_t;
-
-typedef struct
-{
-
-  /* Config stuff */
-  u8 **eal_init_args;
-  u8 *eal_init_args_str;
-  u8 *uio_driver_name;
-  u8 no_multi_seg;
-  u8 enable_tcp_udp_checksum;
-  u8 cryptodev;
-
-  /* Required config parameters */
-  u8 coremask_set_manually;
-  u8 nchannels_set_manually;
-  u32 coremask;
-  u32 nchannels;
-  u32 num_mbufs;
-  u8 num_kni;			/* while kni_init allows u32, port_id in callback fn is only u8 */
-
-  /*
-   * format interface names ala xxxEthernet%d/%d/%d instead of
-   * xxxEthernet%x/%x/%x.
-   */
-  u8 interface_name_format_decimal;
-
-  /* per-device config */
-  dpdk_device_config_t default_devconf;
-  dpdk_device_config_t *dev_confs;
-  uword *device_config_index_by_pci_addr;
-
-} dpdk_config_main_t;
-
-dpdk_config_main_t dpdk_config_main;
-
-typedef struct
-{
-
-  /* Devices */
-  dpdk_device_t *devices;
-  dpdk_device_and_queue_t **devices_by_cpu;
-  dpdk_device_and_queue_t **devices_by_hqos_cpu;
-
-  /* per-thread recycle lists */
-  u32 **recycle;
-
-  /* buffer flags template, configurable to enable/disable tcp / udp cksum */
-  u32 buffer_flags_template;
-
-  /* vlib buffer free list, must be same size as an rte_mbuf */
-  u32 vlib_buffer_free_list_index;
-
-  /* Ethernet input node index */
-  u32 ethernet_input_node_index;
-
-  /* pcap tracing [only works if (CLIB_DEBUG > 0)] */
-  int tx_pcap_enable;
-  pcap_main_t pcap_main;
-  u8 *pcap_filename;
-  u32 pcap_sw_if_index;
-  u32 pcap_pkts_to_capture;
-
-  /* hashes */
-  uword *dpdk_device_by_kni_port_id;
-  uword *vu_sw_if_index_by_listener_fd;
-  uword *vu_sw_if_index_by_sock_fd;
-  u32 *vu_inactive_interfaces_device_index;
-
-  /*
-   * flag indicating that a posted admin up/down
-   * (via post_sw_interface_set_flags) is in progress
-   */
-  u8 admin_up_down_in_progress;
-
-  u8 use_rss;
-
-  /* which cpus are running dpdk-input */
-  int input_cpu_first_index;
-  int input_cpu_count;
-
-  /* which cpus are running I/O TX */
-  int hqos_cpu_first_index;
-  int hqos_cpu_count;
-
-  /* control interval of dpdk link state and stat polling */
-  f64 link_state_poll_interval;
-  f64 stat_poll_interval;
-
-  /* Sleep for this many MS after each device poll */
-  u32 poll_sleep;
-
-  /* convenience */
-  vlib_main_t *vlib_main;
-  vnet_main_t *vnet_main;
-  dpdk_config_main_t *conf;
-
-  /* mempool */
-  struct rte_mempool **pktmbuf_pools;
-} dpdk_main_t;
-
-dpdk_main_t dpdk_main;
-
-typedef struct
-{
-  u32 buffer_index;
-  u16 device_index;
-  u8 queue_index;
-  struct rte_mbuf mb;
-  /* Copy of VLIB buffer; packet data stored in pre_data. */
-  vlib_buffer_t buffer;
-} dpdk_tx_dma_trace_t;
-
-typedef struct
-{
-  u32 buffer_index;
-  u16 device_index;
-  u16 queue_index;
-  struct rte_mbuf mb;
-  vlib_buffer_t buffer;		/* Copy of VLIB buffer; pkt data stored in pre_data. */
-  u8 data[256];			/* First 256 data bytes, used for hexdump */
-} dpdk_rx_dma_trace_t;
-
-void vnet_buffer_needs_dpdk_mb (vlib_buffer_t * b);
-
-clib_error_t *dpdk_set_mac_address (vnet_hw_interface_t * hi, char *address);
-
-clib_error_t *dpdk_set_mc_filter (vnet_hw_interface_t * hi,
-				  struct ether_addr mc_addr_vec[], int naddr);
-
-void dpdk_thread_input (dpdk_main_t * dm, dpdk_device_t * xd);
-
-clib_error_t *dpdk_port_setup (dpdk_main_t * dm, dpdk_device_t * xd);
-
-u32 dpdk_interface_tx_vector (vlib_main_t * vm, u32 dev_instance);
-
-struct rte_mbuf *dpdk_replicate_packet_mb (vlib_buffer_t * b);
-struct rte_mbuf *dpdk_zerocopy_replicate_packet_mb (vlib_buffer_t * b);
-
-#define foreach_dpdk_error						\
-  _(NONE, "no error")							\
-  _(RX_PACKET_ERROR, "Rx packet errors")				\
-  _(RX_BAD_FCS, "Rx bad fcs")						\
-  _(IP_CHECKSUM_ERROR, "Rx ip checksum errors")				\
-  _(RX_ALLOC_FAIL, "rx buf alloc from free list failed")		\
-  _(RX_ALLOC_NO_PHYSMEM, "rx buf alloc failed no physmem")		\
-  _(RX_ALLOC_DROP_PKTS, "rx packets dropped due to alloc error")
-
-typedef enum
-{
-#define _(f,s) DPDK_ERROR_##f,
-  foreach_dpdk_error
-#undef _
-    DPDK_N_ERROR,
-} dpdk_error_t;
-
-int dpdk_set_stat_poll_interval (f64 interval);
-int dpdk_set_link_state_poll_interval (f64 interval);
-void dpdk_update_link_state (dpdk_device_t * xd, f64 now);
-void dpdk_device_lock_init (dpdk_device_t * xd);
-void dpdk_device_lock_free (dpdk_device_t * xd);
-
-void dpdk_rx_trace (dpdk_main_t * dm,
-		    vlib_node_runtime_t * node,
-		    dpdk_device_t * xd,
-		    u16 queue_id, u32 * buffers, uword n_buffers);
-
-#define EFD_OPERATION_LESS_THAN          0
-#define EFD_OPERATION_GREATER_OR_EQUAL   1
-
-format_function_t format_dpdk_device_name;
-format_function_t format_dpdk_device;
-format_function_t format_dpdk_tx_dma_trace;
-format_function_t format_dpdk_rx_dma_trace;
-format_function_t format_dpdk_rte_mbuf;
-format_function_t format_dpdk_rx_rte_mbuf;
-unformat_function_t unformat_socket_mem;
-clib_error_t *unformat_rss_fn (unformat_input_t * input, uword * rss_fn);
-clib_error_t *unformat_hqos (unformat_input_t * input,
-			     dpdk_device_config_hqos_t * hqos);
-
-uword
-admin_up_down_process (vlib_main_t * vm,
-		       vlib_node_runtime_t * rt, vlib_frame_t * f);
-
-#endif /* __included_dpdk_h__ */
-
-/*
- * fd.io coding-style-patch-verification: ON
- *
- * Local Variables:
- * eval: (c-set-style "gnu")
- * End:
- */
diff --git a/src/vnet/devices/dpdk/dpdk_api.c b/src/vnet/devices/dpdk/dpdk_api.c
deleted file mode 100644
index 8faf5c2c..00000000
--- a/src/vnet/devices/dpdk/dpdk_api.c
+++ /dev/null
@@ -1,246 +0,0 @@
-/*
- *------------------------------------------------------------------
- * dpdk_api.c - dpdk interface api
- *
- * Copyright (c) 2016 Cisco and/or its affiliates.
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at:
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- *------------------------------------------------------------------
- */
-
-#include 
-#include 
-
-#if DPDK > 0
-#include 
-#endif
-
-#include 
-
-#define vl_typedefs		/* define message structures */
-#include 
-#undef vl_typedefs
-
-#define vl_endianfun		/* define message structures */
-#include 
-#undef vl_endianfun
-
-/* instantiate all the print functions we know about */
-#define vl_print(handle, ...) vlib_cli_output (handle, __VA_ARGS__)
-#define vl_printfun
-#include 
-#undef vl_printfun
-
-#include 
-
-#define foreach_vpe_api_msg                                               \
-_(SW_INTERFACE_SET_DPDK_HQOS_PIPE, sw_interface_set_dpdk_hqos_pipe)       \
-_(SW_INTERFACE_SET_DPDK_HQOS_SUBPORT, sw_interface_set_dpdk_hqos_subport) \
-_(SW_INTERFACE_SET_DPDK_HQOS_TCTBL, sw_interface_set_dpdk_hqos_tctbl)
-
-static void
-  vl_api_sw_interface_set_dpdk_hqos_pipe_t_handler
-  (vl_api_sw_interface_set_dpdk_hqos_pipe_t * mp)
-{
-  vl_api_sw_interface_set_dpdk_hqos_pipe_reply_t *rmp;
-  int rv = 0;
-
-#if DPDK > 0
-  dpdk_main_t *dm = &dpdk_main;
-  dpdk_device_t *xd;
-
-  u32 sw_if_index = ntohl (mp->sw_if_index);
-  u32 subport = ntohl (mp->subport);
-  u32 pipe = ntohl (mp->pipe);
-  u32 profile = ntohl (mp->profile);
-  vnet_hw_interface_t *hw;
-
-  VALIDATE_SW_IF_INDEX (mp);
-
-  /* hw_if & dpdk device */
-  hw = vnet_get_sup_hw_interface (dm->vnet_main, sw_if_index);
-
-  xd = vec_elt_at_index (dm->devices, hw->dev_instance);
-
-  rv = rte_sched_pipe_config (xd->hqos_ht->hqos, subport, pipe, profile);
-
-  BAD_SW_IF_INDEX_LABEL;
-#else
-  clib_warning ("setting HQoS pipe parameters without DPDK not implemented");
-  rv = VNET_API_ERROR_UNIMPLEMENTED;
-#endif /* DPDK */
-
-  REPLY_MACRO (VL_API_SW_INTERFACE_SET_DPDK_HQOS_PIPE_REPLY);
-}
-
-static void
-  vl_api_sw_interface_set_dpdk_hqos_subport_t_handler
-  (vl_api_sw_interface_set_dpdk_hqos_subport_t * mp)
-{
-  vl_api_sw_interface_set_dpdk_hqos_subport_reply_t *rmp;
-  int rv = 0;
-
-#if DPDK > 0
-  dpdk_main_t *dm = &dpdk_main;
-  dpdk_device_t *xd;
-  struct rte_sched_subport_params p;
-
-  u32 sw_if_index = ntohl (mp->sw_if_index);
-  u32 subport = ntohl (mp->subport);
-  p.tb_rate = ntohl (mp->tb_rate);
-  p.tb_size = ntohl (mp->tb_size);
-  p.tc_rate[0] = ntohl (mp->tc_rate[0]);
-  p.tc_rate[1] = ntohl (mp->tc_rate[1]);
-  p.tc_rate[2] = ntohl (mp->tc_rate[2]);
-  p.tc_rate[3] = ntohl (mp->tc_rate[3]);
-  p.tc_period = ntohl (mp->tc_period);
-
-  vnet_hw_interface_t *hw;
-
-  VALIDATE_SW_IF_INDEX (mp);
-
-  /* hw_if & dpdk device */
-  hw = vnet_get_sup_hw_interface (dm->vnet_main, sw_if_index);
-
-  xd = vec_elt_at_index (dm->devices, hw->dev_instance);
-
-  rv = rte_sched_subport_config (xd->hqos_ht->hqos, subport, &p);
-
-  BAD_SW_IF_INDEX_LABEL;
-#else
-  clib_warning
-    ("setting HQoS subport parameters without DPDK not implemented");
-  rv = VNET_API_ERROR_UNIMPLEMENTED;
-#endif /* DPDK */
-
-  REPLY_MACRO (VL_API_SW_INTERFACE_SET_DPDK_HQOS_SUBPORT_REPLY);
-}
-
-static void
-  vl_api_sw_interface_set_dpdk_hqos_tctbl_t_handler
-  (vl_api_sw_interface_set_dpdk_hqos_tctbl_t * mp)
-{
-  vl_api_sw_interface_set_dpdk_hqos_tctbl_reply_t *rmp;
-  int rv = 0;
-
-#if DPDK > 0
-  dpdk_main_t *dm = &dpdk_main;
-  vlib_thread_main_t *tm = vlib_get_thread_main ();
-  dpdk_device_t *xd;
-
-  u32 sw_if_index = ntohl (mp->sw_if_index);
-  u32 entry = ntohl (mp->entry);
-  u32 tc = ntohl (mp->tc);
-  u32 queue = ntohl (mp->queue);
-  u32 val, i;
-
-  vnet_hw_interface_t *hw;
-
-  VALIDATE_SW_IF_INDEX (mp);
-
-  /* hw_if & dpdk device */
-  hw = vnet_get_sup_hw_interface (dm->vnet_main, sw_if_index);
-
-  xd = vec_elt_at_index (dm->devices, hw->dev_instance);
-
-  if (tc >= RTE_SCHED_TRAFFIC_CLASSES_PER_PIPE)
-    {
-      clib_warning ("invalid traffic class !!");
-      rv = VNET_API_ERROR_INVALID_VALUE;
-      goto done;
-    }
-  if (queue >= RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS)
-    {
-      clib_warning ("invalid queue !!");
-      rv = VNET_API_ERROR_INVALID_VALUE;
-      goto done;
-    }
-
-  /* Detect the set of worker threads */
-  uword *p = hash_get_mem (tm->thread_registrations_by_name, "workers");
-
-  if (p == 0)
-    {
-      clib_warning ("worker thread registration AWOL !!");
-      rv = VNET_API_ERROR_INVALID_VALUE_2;
-      goto done;
-    }
-
-  vlib_thread_registration_t *tr = (vlib_thread_registration_t *) p[0];
-  int worker_thread_first = tr->first_index;
-  int worker_thread_count = tr->count;
-
-  val = tc * RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS + queue;
-  for (i = 0; i < worker_thread_count; i++)
-    xd->hqos_wt[worker_thread_first + i].hqos_tc_table[entry] = val;
-
-  BAD_SW_IF_INDEX_LABEL;
-done:
-#else
-  clib_warning ("setting HQoS DSCP table entry without DPDK not implemented");
-  rv = VNET_API_ERROR_UNIMPLEMENTED;
-#endif /* DPDK */
-
-  REPLY_MACRO (VL_API_SW_INTERFACE_SET_DPDK_HQOS_TCTBL_REPLY);
-}
-
-/*
- * dpdk_api_hookup
- * Add vpe's API message handlers to the table.
- * vlib has alread mapped shared memory and
- * added the client registration handlers.
- * See .../vlib-api/vlibmemory/memclnt_vlib.c:memclnt_process()
- */
-#define vl_msg_name_crc_list
-#include 
-#undef vl_msg_name_crc_list
-
-static void
-setup_message_id_table (api_main_t * am)
-{
-#define _(id,n,crc) vl_msg_api_add_msg_name_crc (am, #n "_" #crc, id);
-  foreach_vl_msg_name_crc_dpdk;
-#undef _
-}
-
-static clib_error_t *
-dpdk_api_hookup (vlib_main_t * vm)
-{
-  api_main_t *am = &api_main;
-
-#define _(N,n)                                                  \
-    vl_msg_api_set_handlers(VL_API_##N, #n,                     \
-                           vl_api_##n##_t_handler,              \
-                           vl_noop_handler,                     \
-                           vl_api_##n##_t_endian,               \
-                           vl_api_##n##_t_print,                \
-                           sizeof(vl_api_##n##_t), 1);
-  foreach_vpe_api_msg;
-#undef _
-
-  /*
-   * Set up the (msg_name, crc, message-id) table
-   */
-  setup_message_id_table (am);
-
-  return 0;
-}
-
-VLIB_API_INIT_FUNCTION (dpdk_api_hookup);
-
-/*
- * fd.io coding-style-patch-verification: ON
- *
- * Local Variables:
- * eval: (c-set-style "gnu")
- * End:
- */
diff --git a/src/vnet/devices/dpdk/dpdk_priv.h b/src/vnet/devices/dpdk/dpdk_priv.h
deleted file mode 100644
index dd40ff48..00000000
--- a/src/vnet/devices/dpdk/dpdk_priv.h
+++ /dev/null
@@ -1,135 +0,0 @@
-/*
- * Copyright (c) 2015 Cisco and/or its affiliates.
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at:
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#define rte_mbuf_from_vlib_buffer(x) (((struct rte_mbuf *)x) - 1)
-#define vlib_buffer_from_rte_mbuf(x) ((vlib_buffer_t *)(x+1))
-
-#define DPDK_NB_RX_DESC_DEFAULT   1024
-#define DPDK_NB_TX_DESC_DEFAULT   1024
-#define DPDK_NB_RX_DESC_VIRTIO    256
-#define DPDK_NB_TX_DESC_VIRTIO    256
-
-#define I40E_DEV_ID_SFP_XL710           0x1572
-#define I40E_DEV_ID_QSFP_A              0x1583
-#define I40E_DEV_ID_QSFP_B              0x1584
-#define I40E_DEV_ID_QSFP_C              0x1585
-#define I40E_DEV_ID_10G_BASE_T          0x1586
-#define I40E_DEV_ID_VF                  0x154C
-
-/* These args appear by themselves */
-#define foreach_eal_double_hyphen_predicate_arg \
-_(no-shconf)                                    \
-_(no-hpet)                                      \
-_(no-huge)                                      \
-_(vmware-tsc-map)
-
-#define foreach_eal_single_hyphen_mandatory_arg \
-_(coremask, c)                                  \
-_(nchannels, n)                                 \
-
-#define foreach_eal_single_hyphen_arg           \
-_(blacklist, b)                                 \
-_(mem-alloc-request, m)                         \
-_(force-ranks, r)
-
-/* These args are preceeded by "--" and followed by a single string */
-#define foreach_eal_double_hyphen_arg           \
-_(huge-dir)                                     \
-_(proc-type)                                    \
-_(file-prefix)                                  \
-_(vdev)
-
-static inline void
-dpdk_get_xstats (dpdk_device_t * xd)
-{
-  int len;
-  if ((len = rte_eth_xstats_get (xd->device_index, NULL, 0)) > 0)
-    {
-      vec_validate (xd->xstats, len - 1);
-      vec_validate (xd->last_cleared_xstats, len - 1);
-
-      len =
-	rte_eth_xstats_get (xd->device_index, xd->xstats,
-			    vec_len (xd->xstats));
-
-      ASSERT (vec_len (xd->xstats) == len);
-      ASSERT (vec_len (xd->last_cleared_xstats) == len);
-
-      _vec_len (xd->xstats) = len;
-      _vec_len (xd->last_cleared_xstats) = len;
-
-    }
-}
-
-
-static inline void
-dpdk_update_counters (dpdk_device_t * xd, f64 now)
-{
-  vlib_simple_counter_main_t *cm;
-  vnet_main_t *vnm = vnet_get_main ();
-  u32 my_cpu = os_get_cpu_number ();
-  u64 rxerrors, last_rxerrors;
-
-  /* only update counters for PMD interfaces */
-  if ((xd->flags & DPDK_DEVICE_FLAG_PMD) == 0)
-    return;
-
-  xd->time_last_stats_update = now ? now : xd->time_last_stats_update;
-  clib_memcpy (&xd->last_stats, &xd->stats, sizeof (xd->last_stats));
-  rte_eth_stats_get (xd->device_index, &xd->stats);
-
-  /* maybe bump interface rx no buffer counter */
-  if (PREDICT_FALSE (xd->stats.rx_nombuf != xd->last_stats.rx_nombuf))
-    {
-      cm = vec_elt_at_index (vnm->interface_main.sw_if_counters,
-			     VNET_INTERFACE_COUNTER_RX_NO_BUF);
-
-      vlib_increment_simple_counter (cm, my_cpu, xd->vlib_sw_if_index,
-				     xd->stats.rx_nombuf -
-				     xd->last_stats.rx_nombuf);
-    }
-
-  /* missed pkt counter */
-  if (PREDICT_FALSE (xd->stats.imissed != xd->last_stats.imissed))
-    {
-      cm = vec_elt_at_index (vnm->interface_main.sw_if_counters,
-			     VNET_INTERFACE_COUNTER_RX_MISS);
-
-      vlib_increment_simple_counter (cm, my_cpu, xd->vlib_sw_if_index,
-				     xd->stats.imissed -
-				     xd->last_stats.imissed);
-    }
-  rxerrors = xd->stats.ierrors;
-  last_rxerrors = xd->last_stats.ierrors;
-
-  if (PREDICT_FALSE (rxerrors != last_rxerrors))
-    {
-      cm = vec_elt_at_index (vnm->interface_main.sw_if_counters,
-			     VNET_INTERFACE_COUNTER_RX_ERROR);
-
-      vlib_increment_simple_counter (cm, my_cpu, xd->vlib_sw_if_index,
-				     rxerrors - last_rxerrors);
-    }
-
-  dpdk_get_xstats (xd);
-}
-
-/*
- * fd.io coding-style-patch-verification: ON
- *
- * Local Variables:
- * eval: (c-set-style "gnu")
- * End:
- */
diff --git a/src/vnet/devices/dpdk/format.c b/src/vnet/devices/dpdk/format.c
deleted file mode 100644
index 1558630c..00000000
--- a/src/vnet/devices/dpdk/format.c
+++ /dev/null
@@ -1,754 +0,0 @@
-/*
- * Copyright (c) 2015 Cisco and/or its affiliates.
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at:
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-#include 
-#include 
-#include 
-#include 
-#include 
-
-#include 
-#include 
-
-#include "dpdk_priv.h"
-#include 
-
-#define foreach_dpdk_counter                    \
-  _ (tx_frames_ok, opackets)                    \
-  _ (tx_bytes_ok, obytes)                       \
-  _ (tx_errors, oerrors)                        \
-  _ (rx_frames_ok, ipackets)                    \
-  _ (rx_bytes_ok, ibytes)                       \
-  _ (rx_errors, ierrors)                        \
-  _ (rx_missed, imissed)                        \
-  _ (rx_no_bufs, rx_nombuf)
-
-#define foreach_dpdk_q_counter                  \
-  _ (rx_frames_ok, q_ipackets)                  \
-  _ (tx_frames_ok, q_opackets)                  \
-  _ (rx_bytes_ok, q_ibytes)                     \
-  _ (tx_bytes_ok, q_obytes)                     \
-  _ (rx_errors, q_errors)
-
-#define foreach_dpdk_rss_hf                    \
-  _(ETH_RSS_FRAG_IPV4,          "ipv4-frag")   \
-  _(ETH_RSS_NONFRAG_IPV4_TCP,   "ipv4-tcp")    \
-  _(ETH_RSS_NONFRAG_IPV4_UDP,   "ipv4-udp")    \
-  _(ETH_RSS_NONFRAG_IPV4_SCTP,  "ipv4-sctp")   \
-  _(ETH_RSS_NONFRAG_IPV4_OTHER, "ipv4-other")  \
-  _(ETH_RSS_IPV4,               "ipv4")        \
-  _(ETH_RSS_IPV6_TCP_EX,        "ipv6-tcp-ex") \
-  _(ETH_RSS_IPV6_UDP_EX,        "ipv6-udp-ex") \
-  _(ETH_RSS_FRAG_IPV6,          "ipv6-frag")   \
-  _(ETH_RSS_NONFRAG_IPV6_TCP,   "ipv6-tcp")    \
-  _(ETH_RSS_NONFRAG_IPV6_UDP,   "ipv6-udp")    \
-  _(ETH_RSS_NONFRAG_IPV6_SCTP,  "ipv6-sctp")   \
-  _(ETH_RSS_NONFRAG_IPV6_OTHER, "ipv6-other")  \
-  _(ETH_RSS_L2_PAYLOAD,         "l2-payload")  \
-  _(ETH_RSS_IPV6_EX,            "ipv6-ex")     \
-  _(ETH_RSS_IPV6,               "ipv6")
-
-
-#define foreach_dpdk_rx_offload_caps            \
-  _(DEV_RX_OFFLOAD_VLAN_STRIP, "vlan-strip")    \
-  _(DEV_RX_OFFLOAD_IPV4_CKSUM, "ipv4-cksum")    \
-  _(DEV_RX_OFFLOAD_UDP_CKSUM , "udp-cksum")     \
-  _(DEV_RX_OFFLOAD_TCP_CKSUM , "tcp-cksum")     \
-  _(DEV_RX_OFFLOAD_TCP_LRO   , "rcp-lro")       \
-  _(DEV_RX_OFFLOAD_QINQ_STRIP, "qinq-strip")
-
-#define foreach_dpdk_tx_offload_caps           \
-  _(DEV_TX_OFFLOAD_VLAN_INSERT, "vlan-insert") \
-  _(DEV_TX_OFFLOAD_IPV4_CKSUM,  "ipv4-cksum")  \
-  _(DEV_TX_OFFLOAD_UDP_CKSUM  , "udp-cksum")   \
-  _(DEV_TX_OFFLOAD_TCP_CKSUM  , "tcp-cksum")   \
-  _(DEV_TX_OFFLOAD_SCTP_CKSUM , "sctp-cksum")  \
-  _(DEV_TX_OFFLOAD_TCP_TSO    , "tcp-tso")     \
-  _(DEV_TX_OFFLOAD_UDP_TSO    , "udp-tso")     \
-  _(DEV_TX_OFFLOAD_OUTER_IPV4_CKSUM, "outer-ipv4-cksum") \
-  _(DEV_TX_OFFLOAD_QINQ_INSERT, "qinq-insert")
-
-#define foreach_dpdk_pkt_rx_offload_flag                                \
-  _ (PKT_RX_VLAN_PKT, "RX packet is a 802.1q VLAN packet")              \
-  _ (PKT_RX_RSS_HASH, "RX packet with RSS hash result")                 \
-  _ (PKT_RX_FDIR, "RX packet with FDIR infos")                          \
-  _ (PKT_RX_L4_CKSUM_BAD, "L4 cksum of RX pkt. is not OK")              \
-  _ (PKT_RX_IP_CKSUM_BAD, "IP cksum of RX pkt. is not OK")              \
-  _ (PKT_RX_VLAN_STRIPPED, "RX packet VLAN tag stripped")               \
-  _ (PKT_RX_IP_CKSUM_GOOD, "IP cksum of RX pkt. is valid")              \
-  _ (PKT_RX_L4_CKSUM_GOOD, "L4 cksum of RX pkt. is valid")              \
-  _ (PKT_RX_IEEE1588_PTP, "RX IEEE1588 L2 Ethernet PT Packet")          \
-  _ (PKT_RX_IEEE1588_TMST, "RX IEEE1588 L2/L4 timestamped packet")      \
-  _ (PKT_RX_QINQ_STRIPPED, "RX packet QinQ tags stripped")
-
-#define foreach_dpdk_pkt_type                                           \
-  _ (L2, ETHER, "Ethernet packet")                                      \
-  _ (L2, ETHER_TIMESYNC, "Ethernet packet for time sync")               \
-  _ (L2, ETHER_ARP, "ARP packet")                                       \
-  _ (L2, ETHER_LLDP, "LLDP (Link Layer Discovery Protocol) packet")     \
-  _ (L2, ETHER_NSH, "NSH (Network Service Header) packet")              \
-  _ (L2, ETHER_VLAN, "VLAN packet")                                     \
-  _ (L2, ETHER_QINQ, "QinQ packet")                                     \
-  _ (L3, IPV4, "IPv4 packet without extension headers")                 \
-  _ (L3, IPV4_EXT, "IPv4 packet with extension headers")                \
-  _ (L3, IPV4_EXT_UNKNOWN, "IPv4 packet with or without extension headers") \
-  _ (L3, IPV6, "IPv6 packet without extension headers")                 \
-  _ (L3, IPV6_EXT, "IPv6 packet with extension headers")                \
-  _ (L3, IPV6_EXT_UNKNOWN, "IPv6 packet with or without extension headers") \
-  _ (L4, TCP, "TCP packet")                                             \
-  _ (L4, UDP, "UDP packet")                                             \
-  _ (L4, FRAG, "Fragmented IP packet")                                  \
-  _ (L4, SCTP, "SCTP (Stream Control Transmission Protocol) packet")    \
-  _ (L4, ICMP, "ICMP packet")                                           \
-  _ (L4, NONFRAG, "Non-fragmented IP packet")                           \
-  _ (TUNNEL, GRE, "GRE tunneling packet")                               \
-  _ (TUNNEL, VXLAN, "VXLAN tunneling packet")                           \
-  _ (TUNNEL, NVGRE, "NVGRE Tunneling packet")                           \
-  _ (TUNNEL, GENEVE, "GENEVE Tunneling packet")                         \
-  _ (TUNNEL, GRENAT, "Teredo, VXLAN or GRE Tunneling packet")           \
-  _ (INNER_L2, ETHER, "Inner Ethernet packet")                          \
-  _ (INNER_L2, ETHER_VLAN, "Inner Ethernet packet with VLAN")           \
-  _ (INNER_L3, IPV4, "Inner IPv4 packet without extension headers")     \
-  _ (INNER_L3, IPV4_EXT, "Inner IPv4 packet with extension headers")    \
-  _ (INNER_L3, IPV4_EXT_UNKNOWN, "Inner IPv4 packet with or without extension headers") \
-  _ (INNER_L3, IPV6, "Inner IPv6 packet without extension headers")     \
-  _ (INNER_L3, IPV6_EXT, "Inner IPv6 packet with extension headers")    \
-  _ (INNER_L3, IPV6_EXT_UNKNOWN, "Inner IPv6 packet with or without extension headers") \
-  _ (INNER_L4, TCP, "Inner TCP packet")                                 \
-  _ (INNER_L4, UDP, "Inner UDP packet")                                 \
-  _ (INNER_L4, FRAG, "Inner fagmented IP packet")                       \
-  _ (INNER_L4, SCTP, "Inner SCTP (Stream Control Transmission Protocol) packet") \
-  _ (INNER_L4, ICMP, "Inner ICMP packet")                               \
-  _ (INNER_L4, NONFRAG, "Inner non-fragmented IP packet")
-
-#define foreach_dpdk_pkt_tx_offload_flag                                \
-  _ (PKT_TX_VLAN_PKT, "TX packet is a 802.1q VLAN packet")              \
-  _ (PKT_TX_IP_CKSUM, "IP cksum of TX pkt. computed by NIC")            \
-  _ (PKT_TX_TCP_CKSUM, "TCP cksum of TX pkt. computed by NIC")          \
-  _ (PKT_TX_SCTP_CKSUM, "SCTP cksum of TX pkt. computed by NIC")        \
-  _ (PKT_TX_IEEE1588_TMST, "TX IEEE1588 packet to timestamp")
-
-#define foreach_dpdk_pkt_offload_flag           \
-  foreach_dpdk_pkt_rx_offload_flag              \
-  foreach_dpdk_pkt_tx_offload_flag
-
-u8 *
-format_dpdk_device_name (u8 * s, va_list * args)
-{
-  dpdk_main_t *dm = &dpdk_main;
-  char *devname_format;
-  char *device_name;
-  u32 i = va_arg (*args, u32);
-  struct rte_eth_dev_info dev_info;
-  u8 *ret;
-
-  if (dm->conf->interface_name_format_decimal)
-    devname_format = "%s%d/%d/%d";
-  else
-    devname_format = "%s%x/%x/%x";
-
-  switch (dm->devices[i].port_type)
-    {
-    case VNET_DPDK_PORT_TYPE_ETH_1G:
-      device_name = "GigabitEthernet";
-      break;
-
-    case VNET_DPDK_PORT_TYPE_ETH_10G:
-      device_name = "TenGigabitEthernet";
-      break;
-
-    case VNET_DPDK_PORT_TYPE_ETH_40G:
-      device_name = "FortyGigabitEthernet";
-      break;
-
-    case VNET_DPDK_PORT_TYPE_ETH_100G:
-      device_name = "HundredGigabitEthernet";
-      break;
-
-    case VNET_DPDK_PORT_TYPE_ETH_BOND:
-      return format (s, "BondEthernet%d", dm->devices[i].device_index);
-
-    case VNET_DPDK_PORT_TYPE_ETH_SWITCH:
-      device_name = "EthernetSwitch";
-      break;
-
-    case VNET_DPDK_PORT_TYPE_AF_PACKET:
-      rte_eth_dev_info_get (i, &dev_info);
-      return format (s, "af_packet%d", dm->devices[i].af_packet_port_id);
-
-    default:
-    case VNET_DPDK_PORT_TYPE_UNKNOWN:
-      device_name = "UnknownEthernet";
-      break;
-    }
-
-  rte_eth_dev_info_get (i, &dev_info);
-
-  if (dev_info.pci_dev)
-    ret = format (s, devname_format, device_name, dev_info.pci_dev->addr.bus,
-		  dev_info.pci_dev->addr.devid,
-		  dev_info.pci_dev->addr.function);
-  else
-    ret = format (s, "%s%d", device_name, dm->devices[i].device_index);
-
-  if (dm->devices[i].interface_name_suffix)
-    return format (ret, "/%s", dm->devices[i].interface_name_suffix);
-  return ret;
-}
-
-static u8 *
-format_dpdk_device_type (u8 * s, va_list * args)
-{
-  dpdk_main_t *dm = &dpdk_main;
-  char *dev_type;
-  u32 i = va_arg (*args, u32);
-
-  switch (dm->devices[i].pmd)
-    {
-    case VNET_DPDK_PMD_E1000EM:
-      dev_type = "Intel 82540EM (e1000)";
-      break;
-
-    case VNET_DPDK_PMD_IGB:
-      dev_type = "Intel e1000";
-      break;
-
-    case VNET_DPDK_PMD_I40E:
-      dev_type = "Intel X710/XL710 Family";
-      break;
-
-    case VNET_DPDK_PMD_I40EVF:
-      dev_type = "Intel X710/XL710 Family VF";
-      break;
-
-    case VNET_DPDK_PMD_FM10K:
-      dev_type = "Intel FM10000 Family Ethernet Switch";
-      break;
-
-    case VNET_DPDK_PMD_IGBVF:
-      dev_type = "Intel e1000 VF";
-      break;
-
-    case VNET_DPDK_PMD_VIRTIO:
-      dev_type = "Red Hat Virtio";
-      break;
-
-    case VNET_DPDK_PMD_IXGBEVF:
-      dev_type = "Intel 82599 VF";
-      break;
-
-    case VNET_DPDK_PMD_IXGBE:
-      dev_type = "Intel 82599";
-      break;
-
-    case VNET_DPDK_PMD_ENIC:
-      dev_type = "Cisco VIC";
-      break;
-
-    case VNET_DPDK_PMD_CXGBE:
-      dev_type = "Chelsio T4/T5";
-      break;
-
-    case VNET_DPDK_PMD_MLX5:
-      dev_type = "Mellanox ConnectX-4 Family";
-      break;
-
-    case VNET_DPDK_PMD_VMXNET3:
-      dev_type = "VMware VMXNET3";
-      break;
-
-    case VNET_DPDK_PMD_AF_PACKET:
-      dev_type = "af_packet";
-      break;
-
-    case VNET_DPDK_PMD_BOND:
-      dev_type = "Ethernet Bonding";
-      break;
-
-    case VNET_DPDK_PMD_DPAA2:
-      dev_type = "NXP DPAA2 Mac";
-      break;
-
-    default:
-    case VNET_DPDK_PMD_UNKNOWN:
-      dev_type = "### UNKNOWN ###";
-      break;
-    }
-
-  return format (s, dev_type);
-}
-
-static u8 *
-format_dpdk_link_status (u8 * s, va_list * args)
-{
-  dpdk_device_t *xd = va_arg (*args, dpdk_device_t *);
-  struct rte_eth_link *l = &xd->link;
-  vnet_main_t *vnm = vnet_get_main ();
-  vnet_hw_interface_t *hi = vnet_get_hw_interface (vnm, xd->vlib_hw_if_index);
-
-  s = format (s, "%s ", l->link_status ? "up" : "down");
-  if (l->link_status)
-    {
-      u32 promisc = rte_eth_promiscuous_get (xd->device_index);
-
-      s = format (s, "%s duplex ", (l->link_duplex == ETH_LINK_FULL_DUPLEX) ?
-		  "full" : "half");
-      s = format (s, "speed %u mtu %d %s\n", l->link_speed,
-		  hi->max_packet_bytes, promisc ? " promisc" : "");
-    }
-  else
-    s = format (s, "\n");
-
-  return s;
-}
-
-#define _line_len 72
-#define _(v, str)                                            \
-if (bitmap & v) {                                            \
-  if (format_get_indent (s) > next_split ) {                 \
-    next_split += _line_len;                                 \
-    s = format(s,"\n%U", format_white_space, indent);        \
-  }                                                          \
-  s = format(s, "%s ", str);                                 \
-}
-
-static u8 *
-format_dpdk_rss_hf_name (u8 * s, va_list * args)
-{
-  u64 bitmap = va_arg (*args, u64);
-  int next_split = _line_len;
-  int indent = format_get_indent (s);
-
-  if (!bitmap)
-    return format (s, "none");
-
-  foreach_dpdk_rss_hf return s;
-}
-
-static u8 *
-format_dpdk_rx_offload_caps (u8 * s, va_list * args)
-{
-  u32 bitmap = va_arg (*args, u32);
-  int next_split = _line_len;
-  int indent = format_get_indent (s);
-
-  if (!bitmap)
-    return format (s, "none");
-
-  foreach_dpdk_rx_offload_caps return s;
-}
-
-static u8 *
-format_dpdk_tx_offload_caps (u8 * s, va_list * args)
-{
-  u32 bitmap = va_arg (*args, u32);
-  int next_split = _line_len;
-  int indent = format_get_indent (s);
-  if (!bitmap)
-    return format (s, "none");
-
-  foreach_dpdk_tx_offload_caps return s;
-}
-
-#undef _line_len
-#undef _
-
-u8 *
-format_dpdk_device (u8 * s, va_list * args)
-{
-  u32 dev_instance = va_arg (*args, u32);
-  int verbose = va_arg (*args, int);
-  dpdk_main_t *dm = &dpdk_main;
-  dpdk_device_t *xd = vec_elt_at_index (dm->devices, dev_instance);
-  uword indent = format_get_indent (s);
-  f64 now = vlib_time_now (dm->vlib_main);
-  struct rte_eth_dev_info di;
-
-  dpdk_update_counters (xd, now);
-  dpdk_update_link_state (xd, now);
-
-  s = format (s, "%U\n%Ucarrier %U",
-	      format_dpdk_device_type, xd->device_index,
-	      format_white_space, indent + 2, format_dpdk_link_status, xd);
-
-  rte_eth_dev_info_get (xd->device_index, &di);
-
-  if (verbose > 1 && xd->flags & DPDK_DEVICE_FLAG_PMD)
-    {
-      struct rte_pci_device *pci;
-      struct rte_eth_rss_conf rss_conf;
-      int vlan_off;
-      int retval;
-
-      rss_conf.rss_key = 0;
-      retval = rte_eth_dev_rss_hash_conf_get (xd->device_index, &rss_conf);
-      if (retval < 0)
-	clib_warning ("rte_eth_dev_rss_hash_conf_get returned %d", retval);
-      pci = di.pci_dev;
-
-      if (pci)
-	s =
-	  format (s,
-		  "%Upci id:            device %04x:%04x subsystem %04x:%04x\n"
-		  "%Upci address:       %04x:%02x:%02x.%02x\n",
-		  format_white_space, indent + 2, pci->id.vendor_id,
-		  pci->id.device_id, pci->id.subsystem_vendor_id,
-		  pci->id.subsystem_device_id, format_white_space, indent + 2,
-		  pci->addr.domain, pci->addr.bus, pci->addr.devid,
-		  pci->addr.function);
-      s =
-	format (s, "%Umax rx packet len: %d\n", format_white_space,
-		indent + 2, di.max_rx_pktlen);
-      s =
-	format (s, "%Umax num of queues: rx %d tx %d\n", format_white_space,
-		indent + 2, di.max_rx_queues, di.max_tx_queues);
-      s =
-	format (s, "%Upromiscuous:       unicast %s all-multicast %s\n",
-		format_white_space, indent + 2,
-		rte_eth_promiscuous_get (xd->device_index) ? "on" : "off",
-		rte_eth_promiscuous_get (xd->device_index) ? "on" : "off");
-      vlan_off = rte_eth_dev_get_vlan_offload (xd->device_index);
-      s = format (s, "%Uvlan offload:      strip %s filter %s qinq %s\n",
-		  format_white_space, indent + 2,
-		  vlan_off & ETH_VLAN_STRIP_OFFLOAD ? "on" : "off",
-		  vlan_off & ETH_VLAN_FILTER_OFFLOAD ? "on" : "off",
-		  vlan_off & ETH_VLAN_EXTEND_OFFLOAD ? "on" : "off");
-      s = format (s, "%Urx offload caps:   %U\n",
-		  format_white_space, indent + 2,
-		  format_dpdk_rx_offload_caps, di.rx_offload_capa);
-      s = format (s, "%Utx offload caps:   %U\n",
-		  format_white_space, indent + 2,
-		  format_dpdk_tx_offload_caps, di.tx_offload_capa);
-      s = format (s, "%Urss active:        %U\n"
-		  "%Urss supported:     %U\n",
-		  format_white_space, indent + 2,
-		  format_dpdk_rss_hf_name, rss_conf.rss_hf,
-		  format_white_space, indent + 2,
-		  format_dpdk_rss_hf_name, di.flow_type_rss_offloads);
-    }
-
-  s = format (s, "%Urx queues %d, rx desc %d, tx queues %d, tx desc %d\n",
-	      format_white_space, indent + 2,
-	      xd->rx_q_used, xd->nb_rx_desc, xd->tx_q_used, xd->nb_tx_desc);
-
-  if (xd->cpu_socket > -1)
-    s = format (s, "%Ucpu socket %d\n",
-		format_white_space, indent + 2, xd->cpu_socket);
-
-  /* $$$ MIB counters  */
-  {
-#define _(N, V)							\
-    if ((xd->stats.V - xd->last_cleared_stats.V) != 0) {       \
-      s = format (s, "\n%U%-40U%16Ld",                         \
-                  format_white_space, indent + 2,              \
-                  format_c_identifier, #N,                     \
-                  xd->stats.V - xd->last_cleared_stats.V);     \
-    }                                                          \
-
-    foreach_dpdk_counter
-#undef _
-  }
-
-  u8 *xs = 0;
-  u32 i = 0;
-  struct rte_eth_xstat *xstat, *last_xstat;
-  struct rte_eth_xstat_name *xstat_names = 0;
-  int len = rte_eth_xstats_get_names (xd->device_index, NULL, 0);
-  vec_validate (xstat_names, len - 1);
-  rte_eth_xstats_get_names (xd->device_index, xstat_names, len);
-
-  ASSERT (vec_len (xd->xstats) == vec_len (xd->last_cleared_xstats));
-
-  /* *INDENT-OFF* */
-  vec_foreach_index(i, xd->xstats)
-    {
-      u64 delta = 0;
-      xstat = vec_elt_at_index(xd->xstats, i);
-      last_xstat = vec_elt_at_index(xd->last_cleared_xstats, i);
-
-      delta = xstat->value - last_xstat->value;
-      if (verbose == 2 || (verbose && delta))
-        {
-          /* format_c_identifier doesn't like c strings inside vector */
-          u8 * name = format(0,"%s", xstat_names[i].name);
-          xs = format(xs, "\n%U%-38U%16Ld",
-                      format_white_space, indent + 4,
-                      format_c_identifier, name, delta);
-          vec_free(name);
-        }
-    }
-  /* *INDENT-ON* */
-
-  vec_free (xstat_names);
-
-  if (xs)
-    {
-      s = format (s, "\n%Uextended stats:%v",
-		  format_white_space, indent + 2, xs);
-      vec_free (xs);
-    }
-
-  return s;
-}
-
-u8 *
-format_dpdk_tx_dma_trace (u8 * s, va_list * va)
-{
-  CLIB_UNUSED (vlib_main_t * vm) = va_arg (*va, vlib_main_t *);
-  CLIB_UNUSED (vlib_node_t * node) = va_arg (*va, vlib_node_t *);
-  CLIB_UNUSED (vnet_main_t * vnm) = vnet_get_main ();
-  dpdk_tx_dma_trace_t *t = va_arg (*va, dpdk_tx_dma_trace_t *);
-  dpdk_main_t *dm = &dpdk_main;
-  dpdk_device_t *xd = vec_elt_at_index (dm->devices, t->device_index);
-  uword indent = format_get_indent (s);
-  vnet_sw_interface_t *sw = vnet_get_sw_interface (vnm, xd->vlib_sw_if_index);
-
-  s = format (s, "%U tx queue %d",
-	      format_vnet_sw_interface_name, vnm, sw, t->queue_index);
-
-  s = format (s, "\n%Ubuffer 0x%x: %U",
-	      format_white_space, indent,
-	      t->buffer_index, format_vlib_buffer, &t->buffer);
-
-  s = format (s, "\n%U%U", format_white_space, indent,
-	      format_ethernet_header_with_length, t->buffer.pre_data,
-	      sizeof (t->buffer.pre_data));
-
-  return s;
-}
-
-u8 *
-format_dpdk_rx_dma_trace (u8 * s, va_list * va)
-{
-  CLIB_UNUSED (vlib_main_t * vm) = va_arg (*va, vlib_main_t *);
-  CLIB_UNUSED (vlib_node_t * node) = va_arg (*va, vlib_node_t *);
-  CLIB_UNUSED (vnet_main_t * vnm) = vnet_get_main ();
-  dpdk_rx_dma_trace_t *t = va_arg (*va, dpdk_rx_dma_trace_t *);
-  dpdk_main_t *dm = &dpdk_main;
-  dpdk_device_t *xd = vec_elt_at_index (dm->devices, t->device_index);
-  format_function_t *f;
-  uword indent = format_get_indent (s);
-  vnet_sw_interface_t *sw = vnet_get_sw_interface (vnm, xd->vlib_sw_if_index);
-
-  s = format (s, "%U rx queue %d",
-	      format_vnet_sw_interface_name, vnm, sw, t->queue_index);
-
-  s = format (s, "\n%Ubuffer 0x%x: %U",
-	      format_white_space, indent,
-	      t->buffer_index, format_vlib_buffer, &t->buffer);
-
-  s = format (s, "\n%U%U",
-	      format_white_space, indent,
-	      format_dpdk_rte_mbuf, &t->mb, &t->data);
-
-  if (vm->trace_main.verbose)
-    {
-      s = format (s, "\n%UPacket Dump%s", format_white_space, indent + 2,
-		  t->mb.data_len > sizeof (t->data) ? " (truncated)" : "");
-      s = format (s, "\n%U%U", format_white_space, indent + 4,
-		  format_hexdump, &t->data,
-		  t->mb.data_len >
-		  sizeof (t->data) ? sizeof (t->data) : t->mb.data_len);
-    }
-  f = node->format_buffer;
-  if (!f)
-    f = format_hex_bytes;
-  s = format (s, "\n%U%U", format_white_space, indent,
-	      f, t->buffer.pre_data, sizeof (t->buffer.pre_data));
-
-  return s;
-}
-
-
-static inline u8 *
-format_dpdk_pkt_types (u8 * s, va_list * va)
-{
-  u32 *pkt_types = va_arg (*va, u32 *);
-  uword indent __attribute__ ((unused)) = format_get_indent (s) + 2;
-
-  if (!*pkt_types)
-    return s;
-
-  s = format (s, "Packet Types");
-
-#define _(L, F, S)             \
-  if ((*pkt_types & RTE_PTYPE_##L##_MASK) == RTE_PTYPE_##L##_##F)           \
-    {                                                                       \
-      s = format (s, "\n%U%s (0x%04x) %s", format_white_space, indent,      \
-                     "RTE_PTYPE_" #L "_" #F, RTE_PTYPE_##L##_##F, S);       \
-    }
-
-  foreach_dpdk_pkt_type
-#undef _
-    return s;
-}
-
-static inline u8 *
-format_dpdk_pkt_offload_flags (u8 * s, va_list * va)
-{
-  u64 *ol_flags = va_arg (*va, u64 *);
-  uword indent = format_get_indent (s) + 2;
-
-  if (!*ol_flags)
-    return s;
-
-  s = format (s, "Packet Offload Flags");
-
-#define _(F, S)             \
-  if (*ol_flags & F)            \
-    {               \
-      s = format (s, "\n%U%s (0x%04x) %s",      \
-      format_white_space, indent, #F, F, S);  \
-    }
-
-  foreach_dpdk_pkt_offload_flag
-#undef _
-    return s;
-}
-
-u8 *
-format_dpdk_rte_mbuf_vlan (u8 * s, va_list * va)
-{
-  ethernet_vlan_header_tv_t *vlan_hdr =
-    va_arg (*va, ethernet_vlan_header_tv_t *);
-
-  if (clib_net_to_host_u16 (vlan_hdr->type) == ETHERNET_TYPE_DOT1AD)
-    {
-      s = format (s, "%U 802.1q vlan ",
-		  format_ethernet_vlan_tci,
-		  clib_net_to_host_u16 (vlan_hdr->priority_cfi_and_id));
-      vlan_hdr++;
-    }
-
-  s = format (s, "%U",
-	      format_ethernet_vlan_tci,
-	      clib_net_to_host_u16 (vlan_hdr->priority_cfi_and_id));
-
-  return s;
-}
-
-u8 *
-format_dpdk_rte_mbuf (u8 * s, va_list * va)
-{
-  struct rte_mbuf *mb = va_arg (*va, struct rte_mbuf *);
-  ethernet_header_t *eth_hdr = va_arg (*va, ethernet_header_t *);
-  uword indent = format_get_indent (s) + 2;
-
-  s = format (s, "PKT MBUF: port %d, nb_segs %d, pkt_len %d"
-	      "\n%Ubuf_len %d, data_len %d, ol_flags 0x%x, data_off %d, phys_addr 0x%x"
-	      "\n%Upacket_type 0x%x",
-	      mb->port, mb->nb_segs, mb->pkt_len,
-	      format_white_space, indent,
-	      mb->buf_len, mb->data_len, mb->ol_flags, mb->data_off,
-	      mb->buf_physaddr, format_white_space, indent, mb->packet_type);
-
-  if (mb->ol_flags)
-    s = format (s, "\n%U%U", format_white_space, indent,
-		format_dpdk_pkt_offload_flags, &mb->ol_flags);
-
-  if ((mb->ol_flags & PKT_RX_VLAN_PKT) &&
-      ((mb->ol_flags & (PKT_RX_VLAN_STRIPPED | PKT_RX_QINQ_STRIPPED)) == 0))
-    {
-      ethernet_vlan_header_tv_t *vlan_hdr =
-	((ethernet_vlan_header_tv_t *) & (eth_hdr->type));
-      s = format (s, " %U", format_dpdk_rte_mbuf_vlan, vlan_hdr);
-    }
-
-  if (mb->packet_type)
-    s = format (s, "\n%U%U", format_white_space, indent,
-		format_dpdk_pkt_types, &mb->packet_type);
-
-  return s;
-}
-
-/* FIXME is this function used? */
-#if 0
-uword
-unformat_socket_mem (unformat_input_t * input, va_list * va)
-{
-  uword **r = va_arg (*va, uword **);
-  int i = 0;
-  u32 mem;
-
-  while (unformat_check_input (input) != UNFORMAT_END_OF_INPUT)
-    {
-      if (unformat (input, ","))
-	hash_set (*r, i, 1024);
-      else if (unformat (input, "%u,", &mem))
-	hash_set (*r, i, mem);
-      else if (unformat (input, "%u", &mem))
-	hash_set (*r, i, mem);
-      else
-	{
-	  unformat_put_input (input);
-	  goto done;
-	}
-      i++;
-    }
-
-done:
-  return 1;
-}
-#endif
-
-clib_error_t *
-unformat_rss_fn (unformat_input_t * input, uword * rss_fn)
-{
-  while (unformat_check_input (input) != UNFORMAT_END_OF_INPUT)
-    {
-      if (0)
-	;
-#undef _
-#define _(f, s)                                 \
-      else if (unformat (input, s))             \
-        *rss_fn |= f;
-
-      foreach_dpdk_rss_hf
-#undef _
-	else
-	{
-	  return clib_error_return (0, "unknown input `%U'",
-				    format_unformat_error, input);
-	}
-    }
-  return 0;
-}
-
-clib_error_t *
-unformat_hqos (unformat_input_t * input, dpdk_device_config_hqos_t * hqos)
-{
-  clib_error_t *error = 0;
-
-  while (unformat_check_input (input) != UNFORMAT_END_OF_INPUT)
-    {
-      if (unformat (input, "hqos-thread %u", &hqos->hqos_thread))
-	hqos->hqos_thread_valid = 1;
-      else
-	{
-	  error = clib_error_return (0, "unknown input `%U'",
-				     format_unformat_error, input);
-	  break;
-	}
-    }
-
-  return error;
-}
-
-/*
- * fd.io coding-style-patch-verification: ON
- *
- * Local Variables:
- * eval: (c-set-style "gnu")
- * End:
- */
diff --git a/src/vnet/devices/dpdk/hqos.c b/src/vnet/devices/dpdk/hqos.c
deleted file mode 100644
index d68bc48f..00000000
--- a/src/vnet/devices/dpdk/hqos.c
+++ /dev/null
@@ -1,775 +0,0 @@
-/*
- * Copyright(c) 2016 Intel Corporation. All rights reserved.
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at:
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-#include 
-#include 
-#include 
-#include 
-#include 
-#include 
-#include 
-
-#include 
-#include 
-#include 
-#include 
-
-#include 
-#include 
-#include 
-
-#include 
-#include 
-#include 
-#include 	/* enumerate all vlib messages */
-
-#define vl_typedefs		/* define message structures */
-#include 
-#undef vl_typedefs
-
-/* instantiate all the print functions we know about */
-#define vl_print(handle, ...) vlib_cli_output (handle, __VA_ARGS__)
-#define vl_printfun
-#include 
-#undef vl_printfun
-
-#include "dpdk_priv.h"
-
-dpdk_main_t dpdk_main;
-
-/***
- *
- * HQoS default configuration values
- *
- ***/
-
-static dpdk_device_config_hqos_t hqos_params_default = {
-  .hqos_thread_valid = 0,
-
-  .swq_size = 4096,
-  .burst_enq = 256,
-  .burst_deq = 220,
-
-  /*
-   * Packet field to identify the subport.
-   *
-   * Default value: Since only one subport is defined by default (see below:
-   *     n_subports_per_port = 1), the subport ID is hardcoded to 0.
-   */
-  .pktfield0_slabpos = 0,
-  .pktfield0_slabmask = 0,
-
-  /*
-   * Packet field to identify the pipe.
-   *
-   * Default value: Assuming Ethernet/IPv4/UDP packets, UDP payload bits 12 .. 23
-   */
-  .pktfield1_slabpos = 40,
-  .pktfield1_slabmask = 0x0000000FFF000000LLU,
-
-  /* Packet field used as index into TC translation table to identify the traffic
-   *     class and queue.
-   *
-   * Default value: Assuming Ethernet/IPv4 packets, IPv4 DSCP field
-   */
-  .pktfield2_slabpos = 8,
-  .pktfield2_slabmask = 0x00000000000000FCLLU,
-  .tc_table = {
-	       0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15,
-	       0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15,
-	       0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15,
-	       0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15,
-	       },
-
-  /* port */
-  .port = {
-	   .name = NULL,	/* Set at init */
-	   .socket = 0,		/* Set at init */
-	   .rate = 1250000000,	/* Assuming 10GbE port */
-	   .mtu = 14 + 1500,	/* Assuming Ethernet/IPv4 pkt (Ethernet FCS not included) */
-	   .frame_overhead = RTE_SCHED_FRAME_OVERHEAD_DEFAULT,
-	   .n_subports_per_port = 1,
-	   .n_pipes_per_subport = 4096,
-	   .qsize = {64, 64, 64, 64},
-	   .pipe_profiles = NULL,	/* Set at config */
-	   .n_pipe_profiles = 1,
-
-#ifdef RTE_SCHED_RED
-	   .red_params = {
-			  /* Traffic Class 0 Colors Green / Yellow / Red */
-			  [0][0] = {.min_th = 48,.max_th = 64,.maxp_inv =
-				    10,.wq_log2 = 9},
-			  [0][1] = {.min_th = 40,.max_th = 64,.maxp_inv =
-				    10,.wq_log2 = 9},
-			  [0][2] = {.min_th = 32,.max_th = 64,.maxp_inv =
-				    10,.wq_log2 = 9},
-
-			  /* Traffic Class 1 - Colors Green / Yellow / Red */
-			  [1][0] = {.min_th = 48,.max_th = 64,.maxp_inv =
-				    10,.wq_log2 = 9},
-			  [1][1] = {.min_th = 40,.max_th = 64,.maxp_inv =
-				    10,.wq_log2 = 9},
-			  [1][2] = {.min_th = 32,.max_th = 64,.maxp_inv =
-				    10,.wq_log2 = 9},
-
-			  /* Traffic Class 2 - Colors Green / Yellow / Red */
-			  [2][0] = {.min_th = 48,.max_th = 64,.maxp_inv =
-				    10,.wq_log2 = 9},
-			  [2][1] = {.min_th = 40,.max_th = 64,.maxp_inv =
-				    10,.wq_log2 = 9},
-			  [2][2] = {.min_th = 32,.max_th = 64,.maxp_inv =
-				    10,.wq_log2 = 9},
-
-			  /* Traffic Class 3 - Colors Green / Yellow / Red */
-			  [3][0] = {.min_th = 48,.max_th = 64,.maxp_inv =
-				    10,.wq_log2 = 9},
-			  [3][1] = {.min_th = 40,.max_th = 64,.maxp_inv =
-				    10,.wq_log2 = 9},
-			  [3][2] = {.min_th = 32,.max_th = 64,.maxp_inv =
-				    10,.wq_log2 = 9}
-			  },
-#endif /* RTE_SCHED_RED */
-	   },
-};
-
-static struct rte_sched_subport_params hqos_subport_params_default = {
-  .tb_rate = 1250000000,	/* 10GbE line rate (measured in bytes/second) */
-  .tb_size = 1000000,
-  .tc_rate = {1250000000, 1250000000, 1250000000, 1250000000},
-  .tc_period = 10,
-};
-
-static struct rte_sched_pipe_params hqos_pipe_params_default = {
-  .tb_rate = 305175,		/* 10GbE line rate divided by 4K pipes */
-  .tb_size = 1000000,
-  .tc_rate = {305175, 305175, 305175, 305175},
-  .tc_period = 40,
-#ifdef RTE_SCHED_SUBPORT_TC_OV
-  .tc_ov_weight = 1,
-#endif
-  .wrr_weights = {1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1},
-};
-
-/***
- *
- * HQoS configuration
- *
- ***/
-
-int
-dpdk_hqos_validate_mask (u64 mask, u32 n)
-{
-  int count = __builtin_popcountll (mask);
-  int pos_lead = sizeof (u64) * 8 - __builtin_clzll (mask);
-  int pos_trail = __builtin_ctzll (mask);
-  int count_expected = __builtin_popcount (n - 1);
-
-  /* Handle the exceptions */
-  if (n == 0)
-    return -1;			/* Error */
-
-  if ((mask == 0) && (n == 1))
-    return 0;			/* OK */
-
-  if (((mask == 0) && (n != 1)) || ((mask != 0) && (n == 1)))
-    return -2;			/* Error */
-
-  /* Check that mask is contiguous */
-  if ((pos_lead - pos_trail) != count)
-    return -3;			/* Error */
-
-  /* Check that mask contains the expected number of bits set */
-  if (count != count_expected)
-    return -4;			/* Error */
-
-  return 0;			/* OK */
-}
-
-void
-dpdk_device_config_hqos_pipe_profile_default (dpdk_device_config_hqos_t *
-					      hqos, u32 pipe_profile_id)
-{
-  memcpy (&hqos->pipe[pipe_profile_id], &hqos_pipe_params_default,
-	  sizeof (hqos_pipe_params_default));
-}
-
-void
-dpdk_device_config_hqos_default (dpdk_device_config_hqos_t * hqos)
-{
-  struct rte_sched_subport_params *subport_params;
-  struct rte_sched_pipe_params *pipe_params;
-  u32 *pipe_map;
-  u32 i;
-
-  memcpy (hqos, &hqos_params_default, sizeof (hqos_params_default));
-
-  /* pipe */
-  vec_add2 (hqos->pipe, pipe_params, hqos->port.n_pipe_profiles);
-
-  for (i = 0; i < vec_len (hqos->pipe); i++)
-    memcpy (&pipe_params[i],
-	    &hqos_pipe_params_default, sizeof (hqos_pipe_params_default));
-
-  hqos->port.pipe_profiles = hqos->pipe;
-
-  /* subport */
-  vec_add2 (hqos->subport, subport_params, hqos->port.n_subports_per_port);
-
-  for (i = 0; i < vec_len (hqos->subport); i++)
-    memcpy (&subport_params[i],
-	    &hqos_subport_params_default,
-	    sizeof (hqos_subport_params_default));
-
-  /* pipe profile */
-  vec_add2 (hqos->pipe_map,
-	    pipe_map,
-	    hqos->port.n_subports_per_port * hqos->port.n_pipes_per_subport);
-
-  for (i = 0; i < vec_len (hqos->pipe_map); i++)
-    pipe_map[i] = 0;
-}
-
-/***
- *
- * HQoS init
- *
- ***/
-
-clib_error_t *
-dpdk_port_setup_hqos (dpdk_device_t * xd, dpdk_device_config_hqos_t * hqos)
-{
-  vlib_thread_main_t *tm = vlib_get_thread_main ();
-  char name[32];
-  u32 subport_id, i;
-  int rv;
-
-  /* Detect the set of worker threads */
-  int worker_thread_first = 0;
-  int worker_thread_count = 0;
-
-  uword *p = hash_get_mem (tm->thread_registrations_by_name, "workers");
-  vlib_thread_registration_t *tr =
-    p ? (vlib_thread_registration_t *) p[0] : 0;
-
-  if (tr && tr->count > 0)
-    {
-      worker_thread_first = tr->first_index;
-      worker_thread_count = tr->count;
-    }
-
-  /* Allocate the per-thread device data array */
-  vec_validate_aligned (xd->hqos_wt, tm->n_vlib_mains - 1,
-			CLIB_CACHE_LINE_BYTES);
-  memset (xd->hqos_wt, 0, tm->n_vlib_mains * sizeof (xd->hqos_wt[0]));
-
-  vec_validate_aligned (xd->hqos_ht, 0, CLIB_CACHE_LINE_BYTES);
-  memset (xd->hqos_ht, 0, sizeof (xd->hqos_ht[0]));
-
-  /* Allocate space for one SWQ per worker thread in the I/O TX thread data structure */
-  vec_validate (xd->hqos_ht->swq, worker_thread_count);
-
-  /* SWQ */
-  for (i = 0; i < worker_thread_count + 1; i++)
-    {
-      u32 swq_flags = RING_F_SP_ENQ | RING_F_SC_DEQ;
-
-      snprintf (name, sizeof (name), "SWQ-worker%u-to-device%u", i,
-		xd->device_index);
-      xd->hqos_ht->swq[i] =
-	rte_ring_create (name, hqos->swq_size, xd->cpu_socket, swq_flags);
-      if (xd->hqos_ht->swq[i] == NULL)
-	return clib_error_return (0,
-				  "SWQ-worker%u-to-device%u: rte_ring_create err",
-				  i, xd->device_index);
-    }
-
-  /*
-   * HQoS
-   */
-
-  /* HQoS port */
-  snprintf (name, sizeof (name), "HQoS%u", xd->device_index);
-  hqos->port.name = strdup (name);
-  if (hqos->port.name == NULL)
-    return clib_error_return (0, "HQoS%u: strdup err", xd->device_index);
-
-  hqos->port.socket = rte_eth_dev_socket_id (xd->device_index);
-  if (hqos->port.socket == SOCKET_ID_ANY)
-    hqos->port.socket = 0;
-
-  xd->hqos_ht->hqos = rte_sched_port_config (&hqos->port);
-  if (xd->hqos_ht->hqos == NULL)
-    return clib_error_return (0, "HQoS%u: rte_sched_port_config err",
-			      xd->device_index);
-
-  /* HQoS subport */
-  for (subport_id = 0; subport_id < hqos->port.n_subports_per_port;
-       subport_id++)
-    {
-      u32 pipe_id;
-
-      rv =
-	rte_sched_subport_config (xd->hqos_ht->hqos, subport_id,
-				  &hqos->subport[subport_id]);
-      if (rv)
-	return clib_error_return (0,
-				  "HQoS%u subport %u: rte_sched_subport_config err (%d)",
-				  xd->device_index, subport_id, rv);
-
-      /* HQoS pipe */
-      for (pipe_id = 0; pipe_id < hqos->port.n_pipes_per_subport; pipe_id++)
-	{
-	  u32 pos = subport_id * hqos->port.n_pipes_per_subport + pipe_id;
-	  u32 profile_id = hqos->pipe_map[pos];
-
-	  rv =
-	    rte_sched_pipe_config (xd->hqos_ht->hqos, subport_id, pipe_id,
-				   profile_id);
-	  if (rv)
-	    return clib_error_return (0,
-				      "HQoS%u subport %u pipe %u: rte_sched_pipe_config err (%d)",
-				      xd->device_index, subport_id, pipe_id,
-				      rv);
-	}
-    }
-
-  /* Set up per-thread device data for the I/O TX thread */
-  xd->hqos_ht->hqos_burst_enq = hqos->burst_enq;
-  xd->hqos_ht->hqos_burst_deq = hqos->burst_deq;
-  vec_validate (xd->hqos_ht->pkts_enq, 2 * hqos->burst_enq - 1);
-  vec_validate (xd->hqos_ht->pkts_deq, hqos->burst_deq - 1);
-  xd->hqos_ht->pkts_enq_len = 0;
-  xd->hqos_ht->swq_pos = 0;
-  xd->hqos_ht->flush_count = 0;
-
-  /* Set up per-thread device data for each worker thread */
-  for (i = 0; i < worker_thread_count + 1; i++)
-    {
-      u32 tid;
-      if (i)
-	tid = worker_thread_first + (i - 1);
-      else
-	tid = i;
-
-      xd->hqos_wt[tid].swq = xd->hqos_ht->swq[i];
-      xd->hqos_wt[tid].hqos_field0_slabpos = hqos->pktfield0_slabpos;
-      xd->hqos_wt[tid].hqos_field0_slabmask = hqos->pktfield0_slabmask;
-      xd->hqos_wt[tid].hqos_field0_slabshr =
-	__builtin_ctzll (hqos->pktfield0_slabmask);
-      xd->hqos_wt[tid].hqos_field1_slabpos = hqos->pktfield1_slabpos;
-      xd->hqos_wt[tid].hqos_field1_slabmask = hqos->pktfield1_slabmask;
-      xd->hqos_wt[tid].hqos_field1_slabshr =
-	__builtin_ctzll (hqos->pktfield1_slabmask);
-      xd->hqos_wt[tid].hqos_field2_slabpos = hqos->pktfield2_slabpos;
-      xd->hqos_wt[tid].hqos_field2_slabmask = hqos->pktfield2_slabmask;
-      xd->hqos_wt[tid].hqos_field2_slabshr =
-	__builtin_ctzll (hqos->pktfield2_slabmask);
-      memcpy (xd->hqos_wt[tid].hqos_tc_table, hqos->tc_table,
-	      sizeof (hqos->tc_table));
-    }
-
-  return 0;
-}
-
-/***
- *
- * HQoS run-time
- *
- ***/
-/*
- * dpdk_hqos_thread - Contains the main loop of an HQoS thread.
- *
- * w
- *     Information for the current thread
- */
-static_always_inline void
-dpdk_hqos_thread_internal_hqos_dbg_bypass (vlib_main_t * vm)
-{
-  dpdk_main_t *dm = &dpdk_main;
-  u32 cpu_index = vm->cpu_index;
-  u32 dev_pos;
-
-  dev_pos = 0;
-  while (1)
-    {
-      vlib_worker_thread_barrier_check ();
-
-      u32 n_devs = vec_len (dm->devices_by_hqos_cpu[cpu_index]);
-      if (dev_pos >= n_devs)
-	dev_pos = 0;
-
-      dpdk_device_and_queue_t *dq =
-	vec_elt_at_index (dm->devices_by_hqos_cpu[cpu_index], dev_pos);
-      dpdk_device_t *xd = vec_elt_at_index (dm->devices, dq->device);
-
-      dpdk_device_hqos_per_hqos_thread_t *hqos = xd->hqos_ht;
-      u32 device_index = xd->device_index;
-      u16 queue_id = dq->queue_id;
-
-      struct rte_mbuf **pkts_enq = hqos->pkts_enq;
-      u32 pkts_enq_len = hqos->pkts_enq_len;
-      u32 swq_pos = hqos->swq_pos;
-      u32 n_swq = vec_len (hqos->swq), i;
-      u32 flush_count = hqos->flush_count;
-
-      for (i = 0; i < n_swq; i++)
-	{
-	  /* Get current SWQ for this device */
-	  struct rte_ring *swq = hqos->swq[swq_pos];
-
-	  /* Read SWQ burst to packet buffer of this device */
-	  pkts_enq_len += rte_ring_sc_dequeue_burst (swq,
-						     (void **)
-						     &pkts_enq[pkts_enq_len],
-						     hqos->hqos_burst_enq);
-
-	  /* Get next SWQ for this device */
-	  swq_pos++;
-	  if (swq_pos >= n_swq)
-	    swq_pos = 0;
-	  hqos->swq_pos = swq_pos;
-
-	  /* HWQ TX enqueue when burst available */
-	  if (pkts_enq_len >= hqos->hqos_burst_enq)
-	    {
-	      u32 n_pkts = rte_eth_tx_burst (device_index,
-					     (uint16_t) queue_id,
-					     pkts_enq,
-					     (uint16_t) pkts_enq_len);
-
-	      for (; n_pkts < pkts_enq_len; n_pkts++)
-		rte_pktmbuf_free (pkts_enq[n_pkts]);
-
-	      pkts_enq_len = 0;
-	      flush_count = 0;
-	      break;
-	    }
-	}
-      if (pkts_enq_len)
-	{
-	  flush_count++;
-	  if (PREDICT_FALSE (flush_count == HQOS_FLUSH_COUNT_THRESHOLD))
-	    {
-	      rte_sched_port_enqueue (hqos->hqos, pkts_enq, pkts_enq_len);
-
-	      pkts_enq_len = 0;
-	      flush_count = 0;
-	    }
-	}
-      hqos->pkts_enq_len = pkts_enq_len;
-      hqos->flush_count = flush_count;
-
-      /* Advance to next device */
-      dev_pos++;
-    }
-}
-
-static_always_inline void
-dpdk_hqos_thread_internal (vlib_main_t * vm)
-{
-  dpdk_main_t *dm = &dpdk_main;
-  u32 cpu_index = vm->cpu_index;
-  u32 dev_pos;
-
-  dev_pos = 0;
-  while (1)
-    {
-      vlib_worker_thread_barrier_check ();
-
-      u32 n_devs = vec_len (dm->devices_by_hqos_cpu[cpu_index]);
-      if (PREDICT_FALSE (n_devs == 0))
-	{
-	  dev_pos = 0;
-	  continue;
-	}
-      if (dev_pos >= n_devs)
-	dev_pos = 0;
-
-      dpdk_device_and_queue_t *dq =
-	vec_elt_at_index (dm->devices_by_hqos_cpu[cpu_index], dev_pos);
-      dpdk_device_t *xd = vec_elt_at_index (dm->devices, dq->device);
-
-      dpdk_device_hqos_per_hqos_thread_t *hqos = xd->hqos_ht;
-      u32 device_index = xd->device_index;
-      u16 queue_id = dq->queue_id;
-
-      struct rte_mbuf **pkts_enq = hqos->pkts_enq;
-      struct rte_mbuf **pkts_deq = hqos->pkts_deq;
-      u32 pkts_enq_len = hqos->pkts_enq_len;
-      u32 swq_pos = hqos->swq_pos;
-      u32 n_swq = vec_len (hqos->swq), i;
-      u32 flush_count = hqos->flush_count;
-
-      /*
-       * SWQ dequeue and HQoS enqueue for current device
-       */
-      for (i = 0; i < n_swq; i++)
-	{
-	  /* Get current SWQ for this device */
-	  struct rte_ring *swq = hqos->swq[swq_pos];
-
-	  /* Read SWQ burst to packet buffer of this device */
-	  pkts_enq_len += rte_ring_sc_dequeue_burst (swq,
-						     (void **)
-						     &pkts_enq[pkts_enq_len],
-						     hqos->hqos_burst_enq);
-
-	  /* Get next SWQ for this device */
-	  swq_pos++;
-	  if (swq_pos >= n_swq)
-	    swq_pos = 0;
-	  hqos->swq_pos = swq_pos;
-
-	  /* HQoS enqueue when burst available */
-	  if (pkts_enq_len >= hqos->hqos_burst_enq)
-	    {
-	      rte_sched_port_enqueue (hqos->hqos, pkts_enq, pkts_enq_len);
-
-	      pkts_enq_len = 0;
-	      flush_count = 0;
-	      break;
-	    }
-	}
-      if (pkts_enq_len)
-	{
-	  flush_count++;
-	  if (PREDICT_FALSE (flush_count == HQOS_FLUSH_COUNT_THRESHOLD))
-	    {
-	      rte_sched_port_enqueue (hqos->hqos, pkts_enq, pkts_enq_len);
-
-	      pkts_enq_len = 0;
-	      flush_count = 0;
-	    }
-	}
-      hqos->pkts_enq_len = pkts_enq_len;
-      hqos->flush_count = flush_count;
-
-      /*
-       * HQoS dequeue and HWQ TX enqueue for current device
-       */
-      {
-	u32 pkts_deq_len, n_pkts;
-
-	pkts_deq_len = rte_sched_port_dequeue (hqos->hqos,
-					       pkts_deq,
-					       hqos->hqos_burst_deq);
-
-	for (n_pkts = 0; n_pkts < pkts_deq_len;)
-	  n_pkts += rte_eth_tx_burst (device_index,
-				      (uint16_t) queue_id,
-				      &pkts_deq[n_pkts],
-				      (uint16_t) (pkts_deq_len - n_pkts));
-      }
-
-      /* Advance to next device */
-      dev_pos++;
-    }
-}
-
-void
-dpdk_hqos_thread (vlib_worker_thread_t * w)
-{
-  vlib_main_t *vm;
-  vlib_thread_main_t *tm = vlib_get_thread_main ();
-  dpdk_main_t *dm = &dpdk_main;
-
-  vm = vlib_get_main ();
-
-  ASSERT (vm->cpu_index == os_get_cpu_number ());
-
-  clib_time_init (&vm->clib_time);
-  clib_mem_set_heap (w->thread_mheap);
-
-  /* Wait until the dpdk init sequence is complete */
-  while (tm->worker_thread_release == 0)
-    vlib_worker_thread_barrier_check ();
-
-  if (vec_len (dm->devices_by_hqos_cpu[vm->cpu_index]) == 0)
-    return
-      clib_error
-      ("current I/O TX thread does not have any devices assigned to it");
-
-  if (DPDK_HQOS_DBG_BYPASS)
-    dpdk_hqos_thread_internal_hqos_dbg_bypass (vm);
-  else
-    dpdk_hqos_thread_internal (vm);
-}
-
-void
-dpdk_hqos_thread_fn (void *arg)
-{
-  vlib_worker_thread_t *w = (vlib_worker_thread_t *) arg;
-  vlib_worker_thread_init (w);
-  dpdk_hqos_thread (w);
-}
-
-/* *INDENT-OFF* */
-VLIB_REGISTER_THREAD (hqos_thread_reg, static) =
-{
-  .name = "hqos-threads",
-  .short_name = "hqos-threads",
-  .function = dpdk_hqos_thread_fn,
-};
-/* *INDENT-ON* */
-
-/*
- * HQoS run-time code to be called by the worker threads
- */
-#define BITFIELD(byte_array, slab_pos, slab_mask, slab_shr)     \
-({                                                              \
-  u64 slab = *((u64 *) &byte_array[slab_pos]);                  \
-  u64 val = (rte_be_to_cpu_64(slab) & slab_mask) >> slab_shr;   \
-  val;                                                          \
-})
-
-#define RTE_SCHED_PORT_HIERARCHY(subport, pipe, traffic_class, queue, color) \
-  ((((u64) (queue)) & 0x3) |                               \
-  ((((u64) (traffic_class)) & 0x3) << 2) |                 \
-  ((((u64) (color)) & 0x3) << 4) |                         \
-  ((((u64) (subport)) & 0xFFFF) << 16) |                   \
-  ((((u64) (pipe)) & 0xFFFFFFFF) << 32))
-
-void
-dpdk_hqos_metadata_set (dpdk_device_hqos_per_worker_thread_t * hqos,
-			struct rte_mbuf **pkts, u32 n_pkts)
-{
-  u32 i;
-
-  for (i = 0; i < (n_pkts & (~0x3)); i += 4)
-    {
-      struct rte_mbuf *pkt0 = pkts[i];
-      struct rte_mbuf *pkt1 = pkts[i + 1];
-      struct rte_mbuf *pkt2 = pkts[i + 2];
-      struct rte_mbuf *pkt3 = pkts[i + 3];
-
-      u8 *pkt0_data = rte_pktmbuf_mtod (pkt0, u8 *);
-      u8 *pkt1_data = rte_pktmbuf_mtod (pkt1, u8 *);
-      u8 *pkt2_data = rte_pktmbuf_mtod (pkt2, u8 *);
-      u8 *pkt3_data = rte_pktmbuf_mtod (pkt3, u8 *);
-
-      u64 pkt0_subport = BITFIELD (pkt0_data, hqos->hqos_field0_slabpos,
-				   hqos->hqos_field0_slabmask,
-				   hqos->hqos_field0_slabshr);
-      u64 pkt0_pipe = BITFIELD (pkt0_data, hqos->hqos_field1_slabpos,
-				hqos->hqos_field1_slabmask,
-				hqos->hqos_field1_slabshr);
-      u64 pkt0_dscp = BITFIELD (pkt0_data, hqos->hqos_field2_slabpos,
-				hqos->hqos_field2_slabmask,
-				hqos->hqos_field2_slabshr);
-      u32 pkt0_tc = hqos->hqos_tc_table[pkt0_dscp & 0x3F] >> 2;
-      u32 pkt0_tc_q = hqos->hqos_tc_table[pkt0_dscp & 0x3F] & 0x3;
-
-      u64 pkt1_subport = BITFIELD (pkt1_data, hqos->hqos_field0_slabpos,
-				   hqos->hqos_field0_slabmask,
-				   hqos->hqos_field0_slabshr);
-      u64 pkt1_pipe = BITFIELD (pkt1_data, hqos->hqos_field1_slabpos,
-				hqos->hqos_field1_slabmask,
-				hqos->hqos_field1_slabshr);
-      u64 pkt1_dscp = BITFIELD (pkt1_data, hqos->hqos_field2_slabpos,
-				hqos->hqos_field2_slabmask,
-				hqos->hqos_field2_slabshr);
-      u32 pkt1_tc = hqos->hqos_tc_table[pkt1_dscp & 0x3F] >> 2;
-      u32 pkt1_tc_q = hqos->hqos_tc_table[pkt1_dscp & 0x3F] & 0x3;
-
-      u64 pkt2_subport = BITFIELD (pkt2_data, hqos->hqos_field0_slabpos,
-				   hqos->hqos_field0_slabmask,
-				   hqos->hqos_field0_slabshr);
-      u64 pkt2_pipe = BITFIELD (pkt2_data, hqos->hqos_field1_slabpos,
-				hqos->hqos_field1_slabmask,
-				hqos->hqos_field1_slabshr);
-      u64 pkt2_dscp = BITFIELD (pkt2_data, hqos->hqos_field2_slabpos,
-				hqos->hqos_field2_slabmask,
-				hqos->hqos_field2_slabshr);
-      u32 pkt2_tc = hqos->hqos_tc_table[pkt2_dscp & 0x3F] >> 2;
-      u32 pkt2_tc_q = hqos->hqos_tc_table[pkt2_dscp & 0x3F] & 0x3;
-
-      u64 pkt3_subport = BITFIELD (pkt3_data, hqos->hqos_field0_slabpos,
-				   hqos->hqos_field0_slabmask,
-				   hqos->hqos_field0_slabshr);
-      u64 pkt3_pipe = BITFIELD (pkt3_data, hqos->hqos_field1_slabpos,
-				hqos->hqos_field1_slabmask,
-				hqos->hqos_field1_slabshr);
-      u64 pkt3_dscp = BITFIELD (pkt3_data, hqos->hqos_field2_slabpos,
-				hqos->hqos_field2_slabmask,
-				hqos->hqos_field2_slabshr);
-      u32 pkt3_tc = hqos->hqos_tc_table[pkt3_dscp & 0x3F] >> 2;
-      u32 pkt3_tc_q = hqos->hqos_tc_table[pkt3_dscp & 0x3F] & 0x3;
-
-      u64 pkt0_sched = RTE_SCHED_PORT_HIERARCHY (pkt0_subport,
-						 pkt0_pipe,
-						 pkt0_tc,
-						 pkt0_tc_q,
-						 0);
-      u64 pkt1_sched = RTE_SCHED_PORT_HIERARCHY (pkt1_subport,
-						 pkt1_pipe,
-						 pkt1_tc,
-						 pkt1_tc_q,
-						 0);
-      u64 pkt2_sched = RTE_SCHED_PORT_HIERARCHY (pkt2_subport,
-						 pkt2_pipe,
-						 pkt2_tc,
-						 pkt2_tc_q,
-						 0);
-      u64 pkt3_sched = RTE_SCHED_PORT_HIERARCHY (pkt3_subport,
-						 pkt3_pipe,
-						 pkt3_tc,
-						 pkt3_tc_q,
-						 0);
-
-      pkt0->hash.sched.lo = pkt0_sched & 0xFFFFFFFF;
-      pkt0->hash.sched.hi = pkt0_sched >> 32;
-      pkt1->hash.sched.lo = pkt1_sched & 0xFFFFFFFF;
-      pkt1->hash.sched.hi = pkt1_sched >> 32;
-      pkt2->hash.sched.lo = pkt2_sched & 0xFFFFFFFF;
-      pkt2->hash.sched.hi = pkt2_sched >> 32;
-      pkt3->hash.sched.lo = pkt3_sched & 0xFFFFFFFF;
-      pkt3->hash.sched.hi = pkt3_sched >> 32;
-    }
-
-  for (; i < n_pkts; i++)
-    {
-      struct rte_mbuf *pkt = pkts[i];
-
-      u8 *pkt_data = rte_pktmbuf_mtod (pkt, u8 *);
-
-      u64 pkt_subport = BITFIELD (pkt_data, hqos->hqos_field0_slabpos,
-				  hqos->hqos_field0_slabmask,
-				  hqos->hqos_field0_slabshr);
-      u64 pkt_pipe = BITFIELD (pkt_data, hqos->hqos_field1_slabpos,
-			       hqos->hqos_field1_slabmask,
-			       hqos->hqos_field1_slabshr);
-      u64 pkt_dscp = BITFIELD (pkt_data, hqos->hqos_field2_slabpos,
-			       hqos->hqos_field2_slabmask,
-			       hqos->hqos_field2_slabshr);
-      u32 pkt_tc = hqos->hqos_tc_table[pkt_dscp & 0x3F] >> 2;
-      u32 pkt_tc_q = hqos->hqos_tc_table[pkt_dscp & 0x3F] & 0x3;
-
-      u64 pkt_sched = RTE_SCHED_PORT_HIERARCHY (pkt_subport,
-						pkt_pipe,
-						pkt_tc,
-						pkt_tc_q,
-						0);
-
-      pkt->hash.sched.lo = pkt_sched & 0xFFFFFFFF;
-      pkt->hash.sched.hi = pkt_sched >> 32;
-    }
-}
-
-/*
- * fd.io coding-style-patch-verification: ON
- *
- * Local Variables:
- * eval: (c-set-style "gnu")
- * End:
- */
diff --git a/src/vnet/devices/dpdk/init.c b/src/vnet/devices/dpdk/init.c
deleted file mode 100755
index 29423e15..00000000
--- a/src/vnet/devices/dpdk/init.c
+++ /dev/null
@@ -1,1801 +0,0 @@
-/*
- * Copyright (c) 2015 Cisco and/or its affiliates.
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at:
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-#include 
-#include 
-#include 
-#include 
-#include 
-
-#include 
-#include 
-#include 
-#include 
-
-#include 
-#include 
-#include 
-#include 
-#include 
-#include 
-#include 
-
-#include "dpdk_priv.h"
-
-dpdk_main_t dpdk_main;
-
-/* force linker to link functions used by vlib and declared weak */
-void *vlib_weakly_linked_functions[] = {
-  &rte_pktmbuf_init,
-  &rte_pktmbuf_pool_init,
-};
-
-#define LINK_STATE_ELOGS	0
-
-#define DEFAULT_HUGE_DIR "/run/vpp/hugepages"
-#define VPP_RUN_DIR "/run/vpp"
-
-/* Port configuration, mildly modified Intel app values */
-
-static struct rte_eth_conf port_conf_template = {
-  .rxmode = {
-	     .split_hdr_size = 0,
-	     .header_split = 0,		/**< Header Split disabled */
-	     .hw_ip_checksum = 0,	/**< IP checksum offload disabled */
-	     .hw_vlan_filter = 0,	/**< VLAN filtering disabled */
-	     .hw_strip_crc = 0,		/**< CRC stripped by hardware */
-	     },
-  .txmode = {
-	     .mq_mode = ETH_MQ_TX_NONE,
-	     },
-};
-
-clib_error_t *
-dpdk_port_setup (dpdk_main_t * dm, dpdk_device_t * xd)
-{
-  int rv;
-  int j;
-
-  ASSERT (os_get_cpu_number () == 0);
-
-  if (xd->flags & DPDK_DEVICE_FLAG_ADMIN_UP)
-    {
-      vnet_hw_interface_set_flags (dm->vnet_main, xd->vlib_hw_if_index, 0);
-      rte_eth_dev_stop (xd->device_index);
-    }
-
-  rv = rte_eth_dev_configure (xd->device_index, xd->rx_q_used,
-			      xd->tx_q_used, &xd->port_conf);
-
-  if (rv < 0)
-    return clib_error_return (0, "rte_eth_dev_configure[%d]: err %d",
-			      xd->device_index, rv);
-
-  /* Set up one TX-queue per worker thread */
-  for (j = 0; j < xd->tx_q_used; j++)
-    {
-      rv = rte_eth_tx_queue_setup (xd->device_index, j, xd->nb_tx_desc,
-				   xd->cpu_socket, &xd->tx_conf);
-
-      /* retry with any other CPU socket */
-      if (rv < 0)
-	rv = rte_eth_tx_queue_setup (xd->device_index, j, xd->nb_tx_desc,
-				     SOCKET_ID_ANY, &xd->tx_conf);
-      if (rv < 0)
-	break;
-    }
-
-  if (rv < 0)
-    return clib_error_return (0, "rte_eth_tx_queue_setup[%d]: err %d",
-			      xd->device_index, rv);
-
-  for (j = 0; j < xd->rx_q_used; j++)
-    {
-
-      rv = rte_eth_rx_queue_setup (xd->device_index, j, xd->nb_rx_desc,
-				   xd->cpu_socket, 0,
-				   dm->
-				   pktmbuf_pools[xd->cpu_socket_id_by_queue
-						 [j]]);
-
-      /* retry with any other CPU socket */
-      if (rv < 0)
-	rv = rte_eth_rx_queue_setup (xd->device_index, j, xd->nb_rx_desc,
-				     SOCKET_ID_ANY, 0,
-				     dm->
-				     pktmbuf_pools[xd->cpu_socket_id_by_queue
-						   [j]]);
-      if (rv < 0)
-	return clib_error_return (0, "rte_eth_rx_queue_setup[%d]: err %d",
-				  xd->device_index, rv);
-    }
-
-  if (xd->flags & DPDK_DEVICE_FLAG_ADMIN_UP)
-    {
-      int rv;
-      rv = rte_eth_dev_start (xd->device_index);
-      if (!rv && xd->default_mac_address)
-	rv = rte_eth_dev_default_mac_addr_set (xd->device_index,
-					       (struct ether_addr *)
-					       xd->default_mac_address);
-      if (rv < 0)
-	clib_warning ("rte_eth_dev_start %d returned %d",
-		      xd->device_index, rv);
-    }
-  return 0;
-}
-
-static u32
-dpdk_flag_change (vnet_main_t * vnm, vnet_hw_interface_t * hi, u32 flags)
-{
-  dpdk_main_t *dm = &dpdk_main;
-  dpdk_device_t *xd = vec_elt_at_index (dm->devices, hi->dev_instance);
-  u32 old = 0;
-
-  if (ETHERNET_INTERFACE_FLAG_CONFIG_PROMISC (flags))
-    {
-      old = (xd->flags & DPDK_DEVICE_FLAG_PROMISC) != 0;
-
-      if (flags & ETHERNET_INTERFACE_FLAG_ACCEPT_ALL)
-	xd->flags |= DPDK_DEVICE_FLAG_PROMISC;
-      else
-	xd->flags &= ~DPDK_DEVICE_FLAG_PROMISC;
-
-      if (xd->flags & DPDK_DEVICE_FLAG_ADMIN_UP)
-	{
-	  if (xd->flags & DPDK_DEVICE_FLAG_PROMISC)
-	    rte_eth_promiscuous_enable (xd->device_index);
-	  else
-	    rte_eth_promiscuous_disable (xd->device_index);
-	}
-    }
-  else if (ETHERNET_INTERFACE_FLAG_CONFIG_MTU (flags))
-    {
-      /*
-       * DAW-FIXME: The Cisco VIC firmware does not provide an api for a
-       *            driver to dynamically change the mtu.  If/when the
-       *            VIC firmware gets fixed, then this should be removed.
-       */
-      if (xd->pmd == VNET_DPDK_PMD_ENIC)
-	{
-	  struct rte_eth_dev_info dev_info;
-
-	  /*
-	   * Restore mtu to what has been set by CIMC in the firmware cfg.
-	   */
-	  rte_eth_dev_info_get (xd->device_index, &dev_info);
-	  hi->max_packet_bytes = dev_info.max_rx_pktlen;
-
-	  vlib_cli_output (vlib_get_main (),
-			   "Cisco VIC mtu can only be changed "
-			   "using CIMC then rebooting the server!");
-	}
-      else
-	{
-	  int rv;
-
-	  xd->port_conf.rxmode.max_rx_pkt_len = hi->max_packet_bytes;
-
-	  if (xd->flags & DPDK_DEVICE_FLAG_ADMIN_UP)
-	    rte_eth_dev_stop (xd->device_index);
-
-	  rv = rte_eth_dev_configure
-	    (xd->device_index, xd->rx_q_used, xd->tx_q_used, &xd->port_conf);
-
-	  if (rv < 0)
-	    vlib_cli_output (vlib_get_main (),
-			     "rte_eth_dev_configure[%d]: err %d",
-			     xd->device_index, rv);
-
-	  rte_eth_dev_set_mtu (xd->device_index, hi->max_packet_bytes);
-
-	  if (xd->flags & DPDK_DEVICE_FLAG_ADMIN_UP)
-	    {
-	      int rv = rte_eth_dev_start (xd->device_index);
-	      if (!rv && xd->default_mac_address)
-		rv = rte_eth_dev_default_mac_addr_set (xd->device_index,
-						       (struct ether_addr *)
-						       xd->default_mac_address);
-	      if (rv < 0)
-		clib_warning ("rte_eth_dev_start %d returned %d",
-			      xd->device_index, rv);
-	    }
-	}
-    }
-  return old;
-}
-
-void
-dpdk_device_lock_init (dpdk_device_t * xd)
-{
-  int q;
-  vec_validate (xd->lockp, xd->tx_q_used - 1);
-  for (q = 0; q < xd->tx_q_used; q++)
-    {
-      xd->lockp[q] = clib_mem_alloc_aligned (CLIB_CACHE_LINE_BYTES,
-					     CLIB_CACHE_LINE_BYTES);
-      memset ((void *) xd->lockp[q], 0, CLIB_CACHE_LINE_BYTES);
-    }
-}
-
-void
-dpdk_device_lock_free (dpdk_device_t * xd)
-{
-  int q;
-
-  for (q = 0; q < vec_len (xd->lockp); q++)
-    clib_mem_free ((void *) xd->lockp[q]);
-  vec_free (xd->lockp);
-  xd->lockp = 0;
-}
-
-static clib_error_t *
-dpdk_lib_init (dpdk_main_t * dm)
-{
-  u32 nports;
-  u32 nb_desc = 0;
-  int i;
-  clib_error_t *error;
-  vlib_main_t *vm = vlib_get_main ();
-  vlib_thread_main_t *tm = vlib_get_thread_main ();
-  vnet_sw_interface_t *sw;
-  vnet_hw_interface_t *hi;
-  dpdk_device_t *xd;
-  vlib_pci_addr_t last_pci_addr;
-  u32 last_pci_addr_port = 0;
-  vlib_thread_registration_t *tr, *tr_hqos;
-  uword *p, *p_hqos;
-
-  u32 next_cpu = 0, next_hqos_cpu = 0;
-  u8 af_packet_port_id = 0;
-  last_pci_addr.as_u32 = ~0;
-
-  dm->input_cpu_first_index = 0;
-  dm->input_cpu_count = 1;
-
-  /* find out which cpus will be used for input */
-  p = hash_get_mem (tm->thread_registrations_by_name, "workers");
-  tr = p ? (vlib_thread_registration_t *) p[0] : 0;
-
-  if (tr && tr->count > 0)
-    {
-      dm->input_cpu_first_index = tr->first_index;
-      dm->input_cpu_count = tr->count;
-    }
-
-  vec_validate_aligned (dm->devices_by_cpu, tm->n_vlib_mains - 1,
-			CLIB_CACHE_LINE_BYTES);
-
-  dm->hqos_cpu_first_index = 0;
-  dm->hqos_cpu_count = 0;
-
-  /* find out which cpus will be used for I/O TX */
-  p_hqos = hash_get_mem (tm->thread_registrations_by_name, "hqos-threads");
-  tr_hqos = p_hqos ? (vlib_thread_registration_t *) p_hqos[0] : 0;
-
-  if (tr_hqos && tr_hqos->count > 0)
-    {
-      dm->hqos_cpu_first_index = tr_hqos->first_index;
-      dm->hqos_cpu_count = tr_hqos->count;
-    }
-
-  vec_validate_aligned (dm->devices_by_hqos_cpu, tm->n_vlib_mains - 1,
-			CLIB_CACHE_LINE_BYTES);
-
-  nports = rte_eth_dev_count ();
-  if (nports < 1)
-    {
-      clib_warning ("DPDK drivers found no ports...");
-    }
-
-  if (CLIB_DEBUG > 0)
-    clib_warning ("DPDK drivers found %d ports...", nports);
-
-  /*
-   * All buffers are all allocated from the same rte_mempool.
-   * Thus they all have the same number of data bytes.
-   */
-  dm->vlib_buffer_free_list_index =
-    vlib_buffer_get_or_create_free_list (vm,
-					 VLIB_BUFFER_DEFAULT_FREE_LIST_BYTES,
-					 "dpdk rx");
-
-  if (dm->conf->enable_tcp_udp_checksum)
-    dm->buffer_flags_template &= ~(IP_BUFFER_L4_CHECKSUM_CORRECT
-				   | IP_BUFFER_L4_CHECKSUM_COMPUTED);
-
-  for (i = 0; i < nports; i++)
-    {
-      u8 addr[6];
-      u8 vlan_strip = 0;
-      int j;
-      struct rte_eth_dev_info dev_info;
-      clib_error_t *rv;
-      struct rte_eth_link l;
-      dpdk_device_config_t *devconf = 0;
-      vlib_pci_addr_t pci_addr;
-      uword *p = 0;
-
-      rte_eth_dev_info_get (i, &dev_info);
-      if (dev_info.pci_dev)	/* bonded interface has no pci info */
-	{
-	  pci_addr.domain = dev_info.pci_dev->addr.domain;
-	  pci_addr.bus = dev_info.pci_dev->addr.bus;
-	  pci_addr.slot = dev_info.pci_dev->addr.devid;
-	  pci_addr.function = dev_info.pci_dev->addr.function;
-	  p =
-	    hash_get (dm->conf->device_config_index_by_pci_addr,
-		      pci_addr.as_u32);
-	}
-
-      if (p)
-	devconf = pool_elt_at_index (dm->conf->dev_confs, p[0]);
-      else
-	devconf = &dm->conf->default_devconf;
-
-      /* Create vnet interface */
-      vec_add2_aligned (dm->devices, xd, 1, CLIB_CACHE_LINE_BYTES);
-      xd->nb_rx_desc = DPDK_NB_RX_DESC_DEFAULT;
-      xd->nb_tx_desc = DPDK_NB_TX_DESC_DEFAULT;
-      xd->cpu_socket = (i8) rte_eth_dev_socket_id (i);
-
-      /* Handle interface naming for devices with multiple ports sharing same PCI ID */
-      if (dev_info.pci_dev)
-	{
-	  struct rte_eth_dev_info di = { 0 };
-	  rte_eth_dev_info_get (i + 1, &di);
-	  if (di.pci_dev && pci_addr.as_u32 != last_pci_addr.as_u32 &&
-	      memcmp (&dev_info.pci_dev->addr, &di.pci_dev->addr,
-		      sizeof (struct rte_pci_addr)) == 0)
-	    {
-	      xd->interface_name_suffix = format (0, "0");
-	      last_pci_addr.as_u32 = pci_addr.as_u32;
-	      last_pci_addr_port = i;
-	    }
-	  else if (pci_addr.as_u32 == last_pci_addr.as_u32)
-	    {
-	      xd->interface_name_suffix =
-		format (0, "%u", i - last_pci_addr_port);
-	    }
-	  else
-	    {
-	      last_pci_addr.as_u32 = ~0;
-	    }
-	}
-      else
-	last_pci_addr.as_u32 = ~0;
-
-      clib_memcpy (&xd->tx_conf, &dev_info.default_txconf,
-		   sizeof (struct rte_eth_txconf));
-      if (dm->conf->no_multi_seg)
-	{
-	  xd->tx_conf.txq_flags |= ETH_TXQ_FLAGS_NOMULTSEGS;
-	  port_conf_template.rxmode.jumbo_frame = 0;
-	}
-      else
-	{
-	  xd->tx_conf.txq_flags &= ~ETH_TXQ_FLAGS_NOMULTSEGS;
-	  port_conf_template.rxmode.jumbo_frame = 1;
-	  xd->flags |= DPDK_DEVICE_FLAG_MAYBE_MULTISEG;
-	}
-
-      clib_memcpy (&xd->port_conf, &port_conf_template,
-		   sizeof (struct rte_eth_conf));
-
-      xd->tx_q_used = clib_min (dev_info.max_tx_queues, tm->n_vlib_mains);
-
-      if (devconf->num_tx_queues > 0
-	  && devconf->num_tx_queues < xd->tx_q_used)
-	xd->tx_q_used = clib_min (xd->tx_q_used, devconf->num_tx_queues);
-
-      if (devconf->num_rx_queues > 1 && dm->use_rss == 0)
-	{
-	  dm->use_rss = 1;
-	}
-
-      if (devconf->num_rx_queues > 1
-	  && dev_info.max_rx_queues >= devconf->num_rx_queues)
-	{
-	  xd->rx_q_used = devconf->num_rx_queues;
-	  xd->port_conf.rxmode.mq_mode = ETH_MQ_RX_RSS;
-	  if (devconf->rss_fn == 0)
-	    xd->port_conf.rx_adv_conf.rss_conf.rss_hf =
-	      ETH_RSS_IP | ETH_RSS_UDP | ETH_RSS_TCP;
-	  else
-	    xd->port_conf.rx_adv_conf.rss_conf.rss_hf = devconf->rss_fn;
-	}
-      else
-	xd->rx_q_used = 1;
-
-      xd->flags |= DPDK_DEVICE_FLAG_PMD;
-
-      /* workaround for drivers not setting driver_name */
-      if ((!dev_info.driver_name) && (dev_info.pci_dev))
-	dev_info.driver_name = dev_info.pci_dev->driver->driver.name;
-
-      ASSERT (dev_info.driver_name);
-
-      if (!xd->pmd)
-	{
-
-
-#define _(s,f) else if (dev_info.driver_name &&                 \
-                        !strcmp(dev_info.driver_name, s))       \
-                 xd->pmd = VNET_DPDK_PMD_##f;
-	  if (0)
-	    ;
-	  foreach_dpdk_pmd
-#undef _
-	    else
-	    xd->pmd = VNET_DPDK_PMD_UNKNOWN;
-
-	  xd->port_type = VNET_DPDK_PORT_TYPE_UNKNOWN;
-	  xd->nb_rx_desc = DPDK_NB_RX_DESC_DEFAULT;
-	  xd->nb_tx_desc = DPDK_NB_TX_DESC_DEFAULT;
-
-	  switch (xd->pmd)
-	    {
-	      /* 1G adapters */
-	    case VNET_DPDK_PMD_E1000EM:
-	    case VNET_DPDK_PMD_IGB:
-	    case VNET_DPDK_PMD_IGBVF:
-	      xd->port_type = VNET_DPDK_PORT_TYPE_ETH_1G;
-	      break;
-
-	      /* 10G adapters */
-	    case VNET_DPDK_PMD_IXGBE:
-	    case VNET_DPDK_PMD_IXGBEVF:
-	    case VNET_DPDK_PMD_THUNDERX:
-	      xd->port_type = VNET_DPDK_PORT_TYPE_ETH_10G;
-	      break;
-	    case VNET_DPDK_PMD_DPAA2:
-	      xd->port_type = VNET_DPDK_PORT_TYPE_ETH_10G;
-	      break;
-
-	      /* Cisco VIC */
-	    case VNET_DPDK_PMD_ENIC:
-	      rte_eth_link_get_nowait (i, &l);
-	      xd->flags |= DPDK_DEVICE_FLAG_PMD_SUPPORTS_PTYPE;
-	      if (l.link_speed == 40000)
-		xd->port_type = VNET_DPDK_PORT_TYPE_ETH_40G;
-	      else
-		xd->port_type = VNET_DPDK_PORT_TYPE_ETH_10G;
-	      break;
-
-	      /* Intel Fortville */
-	    case VNET_DPDK_PMD_I40E:
-	    case VNET_DPDK_PMD_I40EVF:
-	      xd->flags |= DPDK_DEVICE_FLAG_PMD_SUPPORTS_PTYPE;
-	      xd->port_type = VNET_DPDK_PORT_TYPE_ETH_40G;
-
-	      switch (dev_info.pci_dev->id.device_id)
-		{
-		case I40E_DEV_ID_10G_BASE_T:
-		case I40E_DEV_ID_SFP_XL710:
-		  xd->port_type = VNET_DPDK_PORT_TYPE_ETH_10G;
-		  break;
-		case I40E_DEV_ID_QSFP_A:
-		case I40E_DEV_ID_QSFP_B:
-		case I40E_DEV_ID_QSFP_C:
-		  xd->port_type = VNET_DPDK_PORT_TYPE_ETH_40G;
-		  break;
-		case I40E_DEV_ID_VF:
-		  rte_eth_link_get_nowait (i, &l);
-		  xd->port_type = l.link_speed == 10000 ?
-		    VNET_DPDK_PORT_TYPE_ETH_10G : VNET_DPDK_PORT_TYPE_ETH_40G;
-		  break;
-		default:
-		  xd->port_type = VNET_DPDK_PORT_TYPE_UNKNOWN;
-		}
-	      break;
-
-	    case VNET_DPDK_PMD_CXGBE:
-	      switch (dev_info.pci_dev->id.device_id)
-		{
-		case 0x540d:	/* T580-CR */
-		case 0x5410:	/* T580-LP-cr */
-		  xd->port_type = VNET_DPDK_PORT_TYPE_ETH_40G;
-		  break;
-		case 0x5403:	/* T540-CR */
-		  xd->port_type = VNET_DPDK_PORT_TYPE_ETH_10G;
-		  break;
-		default:
-		  xd->port_type = VNET_DPDK_PORT_TYPE_UNKNOWN;
-		}
-	      break;
-
-	    case VNET_DPDK_PMD_MLX5:
-	      {
-		char *pn_100g[] = { "MCX415A-CCAT", "MCX416A-CCAT", 0 };
-		char *pn_40g[] = { "MCX413A-BCAT", "MCX414A-BCAT",
-		  "MCX415A-BCAT", "MCX416A-BCAT", "MCX4131A-BCAT", 0
-		};
-		char *pn_10g[] = { "MCX4111A-XCAT", "MCX4121A-XCAT", 0 };
-
-		vlib_pci_device_t *pd = vlib_get_pci_device (&pci_addr);
-		u8 *pn = 0;
-		char **c;
-		int found = 0;
-		pn = format (0, "%U%c",
-			     format_vlib_pci_vpd, pd->vpd_r, "PN", 0);
-
-		if (!pn)
-		  break;
-
-		c = pn_100g;
-		while (!found && c[0])
-		  {
-		    if (strncmp ((char *) pn, c[0], strlen (c[0])) == 0)
-		      {
-			xd->port_type = VNET_DPDK_PORT_TYPE_ETH_100G;
-			break;
-		      }
-		    c++;
-		  }
-
-		c = pn_40g;
-		while (!found && c[0])
-		  {
-		    if (strncmp ((char *) pn, c[0], strlen (c[0])) == 0)
-		      {
-			xd->port_type = VNET_DPDK_PORT_TYPE_ETH_40G;
-			break;
-		      }
-		    c++;
-		  }
-
-		c = pn_10g;
-		while (!found && c[0])
-		  {
-		    if (strncmp ((char *) pn, c[0], strlen (c[0])) == 0)
-		      {
-			xd->port_type = VNET_DPDK_PORT_TYPE_ETH_10G;
-			break;
-		      }
-		    c++;
-		  }
-
-		vec_free (pn);
-	      }
-
-	      break;
-	      /* Intel Red Rock Canyon */
-	    case VNET_DPDK_PMD_FM10K:
-	      xd->port_type = VNET_DPDK_PORT_TYPE_ETH_SWITCH;
-	      break;
-
-	      /* virtio */
-	    case VNET_DPDK_PMD_VIRTIO:
-	      xd->port_type = VNET_DPDK_PORT_TYPE_ETH_1G;
-	      xd->nb_rx_desc = DPDK_NB_RX_DESC_VIRTIO;
-	      xd->nb_tx_desc = DPDK_NB_TX_DESC_VIRTIO;
-	      break;
-
-	      /* vmxnet3 */
-	    case VNET_DPDK_PMD_VMXNET3:
-	      xd->port_type = VNET_DPDK_PORT_TYPE_ETH_1G;
-	      xd->tx_conf.txq_flags |= ETH_TXQ_FLAGS_NOMULTSEGS;
-	      break;
-
-	    case VNET_DPDK_PMD_AF_PACKET:
-	      xd->port_type = VNET_DPDK_PORT_TYPE_AF_PACKET;
-	      xd->af_packet_port_id = af_packet_port_id++;
-	      break;
-
-	    case VNET_DPDK_PMD_BOND:
-	      xd->flags |= DPDK_DEVICE_FLAG_PMD_SUPPORTS_PTYPE;
-	      xd->port_type = VNET_DPDK_PORT_TYPE_ETH_BOND;
-	      break;
-
-	    default:
-	      xd->port_type = VNET_DPDK_PORT_TYPE_UNKNOWN;
-	    }
-
-	  if (devconf->num_rx_desc)
-	    xd->nb_rx_desc = devconf->num_rx_desc;
-
-	  if (devconf->num_tx_desc)
-	    xd->nb_tx_desc = devconf->num_tx_desc;
-	}
-
-      /*
-       * Ensure default mtu is not > the mtu read from the hardware.
-       * Otherwise rte_eth_dev_configure() will fail and the port will
-       * not be available.
-       */
-      if (ETHERNET_MAX_PACKET_BYTES > dev_info.max_rx_pktlen)
-	{
-	  /*
-	   * This device does not support the platforms's max frame
-	   * size. Use it's advertised mru instead.
-	   */
-	  xd->port_conf.rxmode.max_rx_pkt_len = dev_info.max_rx_pktlen;
-	}
-      else
-	{
-	  xd->port_conf.rxmode.max_rx_pkt_len = ETHERNET_MAX_PACKET_BYTES;
-
-	  /*
-	   * Some platforms do not account for Ethernet FCS (4 bytes) in
-	   * MTU calculations. To interop with them increase mru but only
-	   * if the device's settings can support it.
-	   */
-	  if ((dev_info.max_rx_pktlen >= (ETHERNET_MAX_PACKET_BYTES + 4)) &&
-	      xd->port_conf.rxmode.hw_strip_crc)
-	    {
-	      /*
-	       * Allow additional 4 bytes (for Ethernet FCS). These bytes are
-	       * stripped by h/w and so will not consume any buffer memory.
-	       */
-	      xd->port_conf.rxmode.max_rx_pkt_len += 4;
-	    }
-	}
-
-      if (xd->pmd == VNET_DPDK_PMD_AF_PACKET)
-	{
-	  f64 now = vlib_time_now (vm);
-	  u32 rnd;
-	  rnd = (u32) (now * 1e6);
-	  rnd = random_u32 (&rnd);
-	  clib_memcpy (addr + 2, &rnd, sizeof (rnd));
-	  addr[0] = 2;
-	  addr[1] = 0xfe;
-	}
-      else
-	rte_eth_macaddr_get (i, (struct ether_addr *) addr);
-
-      if (xd->tx_q_used < tm->n_vlib_mains)
-	dpdk_device_lock_init (xd);
-
-      xd->device_index = xd - dm->devices;
-      ASSERT (i == xd->device_index);
-      xd->per_interface_next_index = ~0;
-
-      /* assign interface to input thread */
-      dpdk_device_and_queue_t *dq;
-      int q;
-
-      if (devconf->workers)
-	{
-	  int i;
-	  q = 0;
-	  /* *INDENT-OFF* */
-	  clib_bitmap_foreach (i, devconf->workers, ({
-	    int cpu = dm->input_cpu_first_index + i;
-	    unsigned lcore = vlib_worker_threads[cpu].lcore_id;
-	    vec_validate(xd->cpu_socket_id_by_queue, q);
-	    xd->cpu_socket_id_by_queue[q] = rte_lcore_to_socket_id(lcore);
-	    vec_add2(dm->devices_by_cpu[cpu], dq, 1);
-	    dq->device = xd->device_index;
-	    dq->queue_id = q++;
-	  }));
-	  /* *INDENT-ON* */
-	}
-      else
-	for (q = 0; q < xd->rx_q_used; q++)
-	  {
-	    int cpu = dm->input_cpu_first_index + next_cpu;
-	    unsigned lcore = vlib_worker_threads[cpu].lcore_id;
-
-	    /*
-	     * numa node for worker thread handling this queue
-	     * needed for taking buffers from the right mempool
-	     */
-	    vec_validate (xd->cpu_socket_id_by_queue, q);
-	    xd->cpu_socket_id_by_queue[q] = rte_lcore_to_socket_id (lcore);
-
-	    /*
-	     * construct vector of (device,queue) pairs for each worker thread
-	     */
-	    vec_add2 (dm->devices_by_cpu[cpu], dq, 1);
-	    dq->device = xd->device_index;
-	    dq->queue_id = q;
-
-	    next_cpu++;
-	    if (next_cpu == dm->input_cpu_count)
-	      next_cpu = 0;
-	  }
-
-
-      if (devconf->hqos_enabled)
-	{
-	  xd->flags |= DPDK_DEVICE_FLAG_HQOS;
-
-	  if (devconf->hqos.hqos_thread_valid)
-	    {
-	      int cpu = dm->hqos_cpu_first_index + devconf->hqos.hqos_thread;
-
-	      if (devconf->hqos.hqos_thread >= dm->hqos_cpu_count)
-		return clib_error_return (0, "invalid HQoS thread index");
-
-	      vec_add2 (dm->devices_by_hqos_cpu[cpu], dq, 1);
-	      dq->device = xd->device_index;
-	      dq->queue_id = 0;
-	    }
-	  else
-	    {
-	      int cpu = dm->hqos_cpu_first_index + next_hqos_cpu;
-
-	      if (dm->hqos_cpu_count == 0)
-		return clib_error_return (0, "no HQoS threads available");
-
-	      vec_add2 (dm->devices_by_hqos_cpu[cpu], dq, 1);
-	      dq->device = xd->device_index;
-	      dq->queue_id = 0;
-
-	      next_hqos_cpu++;
-	      if (next_hqos_cpu == dm->hqos_cpu_count)
-		next_hqos_cpu = 0;
-
-	      devconf->hqos.hqos_thread_valid = 1;
-	      devconf->hqos.hqos_thread = cpu;
-	    }
-	}
-
-      vec_validate_aligned (xd->tx_vectors, tm->n_vlib_mains,
-			    CLIB_CACHE_LINE_BYTES);
-      for (j = 0; j < tm->n_vlib_mains; j++)
-	{
-	  vec_validate_ha (xd->tx_vectors[j], xd->nb_tx_desc,
-			   sizeof (tx_ring_hdr_t), CLIB_CACHE_LINE_BYTES);
-	  vec_reset_length (xd->tx_vectors[j]);
-	}
-
-      vec_validate_aligned (xd->rx_vectors, xd->rx_q_used,
-			    CLIB_CACHE_LINE_BYTES);
-      for (j = 0; j < xd->rx_q_used; j++)
-	{
-	  vec_validate_aligned (xd->rx_vectors[j], VLIB_FRAME_SIZE - 1,
-				CLIB_CACHE_LINE_BYTES);
-	  vec_reset_length (xd->rx_vectors[j]);
-	}
-
-      vec_validate_aligned (xd->d_trace_buffers, tm->n_vlib_mains,
-			    CLIB_CACHE_LINE_BYTES);
-
-      rv = dpdk_port_setup (dm, xd);
-
-      if (rv)
-	return rv;
-
-      if (devconf->hqos_enabled)
-	{
-	  rv = dpdk_port_setup_hqos (xd, &devconf->hqos);
-	  if (rv)
-	    return rv;
-	}
-
-      /* count the number of descriptors used for this device */
-      nb_desc += xd->nb_rx_desc + xd->nb_tx_desc * xd->tx_q_used;
-
-      error = ethernet_register_interface
-	(dm->vnet_main, dpdk_device_class.index, xd->device_index,
-	 /* ethernet address */ addr,
-	 &xd->vlib_hw_if_index, dpdk_flag_change);
-      if (error)
-	return error;
-
-      sw = vnet_get_hw_sw_interface (dm->vnet_main, xd->vlib_hw_if_index);
-      xd->vlib_sw_if_index = sw->sw_if_index;
-      hi = vnet_get_hw_interface (dm->vnet_main, xd->vlib_hw_if_index);
-
-      /*
-       * DAW-FIXME: The Cisco VIC firmware does not provide an api for a
-       *            driver to dynamically change the mtu.  If/when the
-       *            VIC firmware gets fixed, then this should be removed.
-       */
-      if (xd->pmd == VNET_DPDK_PMD_ENIC)
-	{
-	  /*
-	   * Initialize mtu to what has been set by CIMC in the firmware cfg.
-	   */
-	  hi->max_packet_bytes = dev_info.max_rx_pktlen;
-	  if (devconf->vlan_strip_offload != DPDK_DEVICE_VLAN_STRIP_OFF)
-	    vlan_strip = 1;	/* remove vlan tag from VIC port by default */
-	  else
-	    clib_warning ("VLAN strip disabled for interface\n");
-	}
-      else if (devconf->vlan_strip_offload == DPDK_DEVICE_VLAN_STRIP_ON)
-	vlan_strip = 1;
-
-      if (vlan_strip)
-	{
-	  int vlan_off;
-	  vlan_off = rte_eth_dev_get_vlan_offload (xd->device_index);
-	  vlan_off |= ETH_VLAN_STRIP_OFFLOAD;
-	  xd->port_conf.rxmode.hw_vlan_strip = vlan_off;
-	  if (rte_eth_dev_set_vlan_offload (xd->device_index, vlan_off) == 0)
-	    clib_warning ("VLAN strip enabled for interface\n");
-	  else
-	    clib_warning ("VLAN strip cannot be supported by interface\n");
-	}
-
-      hi->max_l3_packet_bytes[VLIB_RX] = hi->max_l3_packet_bytes[VLIB_TX] =
-	xd->port_conf.rxmode.max_rx_pkt_len - sizeof (ethernet_header_t);
-
-      rte_eth_dev_set_mtu (xd->device_index, hi->max_packet_bytes);
-    }
-
-  if (nb_desc > dm->conf->num_mbufs)
-    clib_warning ("%d mbufs allocated but total rx/tx ring size is %d\n",
-		  dm->conf->num_mbufs, nb_desc);
-
-  return 0;
-}
-
-static void
-dpdk_bind_devices_to_uio (dpdk_config_main_t * conf)
-{
-  vlib_pci_main_t *pm = &pci_main;
-  clib_error_t *error;
-  vlib_pci_device_t *d;
-  u8 *pci_addr = 0;
-  int num_whitelisted = vec_len (conf->dev_confs);
-
-  /* *INDENT-OFF* */
-  pool_foreach (d, pm->pci_devs, ({
-    dpdk_device_config_t * devconf = 0;
-    vec_reset_length (pci_addr);
-    pci_addr = format (pci_addr, "%U%c", format_vlib_pci_addr, &d->bus_address, 0);
-
-    if (d->device_class != PCI_CLASS_NETWORK_ETHERNET && d->device_class != PCI_CLASS_PROCESSOR_CO)
-      continue;
-
-    if (num_whitelisted)
-      {
-	uword * p = hash_get (conf->device_config_index_by_pci_addr, d->bus_address.as_u32);
-
-	if (!p)
-	  continue;
-
-	devconf = pool_elt_at_index (conf->dev_confs, p[0]);
-      }
-
-    /* virtio */
-    if (d->vendor_id == 0x1af4 && d->device_id == 0x1000)
-      ;
-    /* vmxnet3 */
-    else if (d->vendor_id == 0x15ad && d->device_id == 0x07b0)
-      ;
-    /* all Intel devices */
-    else if (d->vendor_id == 0x8086)
-      ;
-    /* Cisco VIC */
-    else if (d->vendor_id == 0x1137 && d->device_id == 0x0043)
-      ;
-    /* Chelsio T4/T5 */
-    else if (d->vendor_id == 0x1425 && (d->device_id & 0xe000) == 0x4000)
-      ;
-    else
-      {
-        clib_warning ("Unsupported Ethernet PCI device 0x%04x:0x%04x found "
-		      "at PCI address %s\n", (u16) d->vendor_id, (u16) d->device_id,
-		      pci_addr);
-        continue;
-      }
-
-    error = vlib_pci_bind_to_uio (d, (char *) conf->uio_driver_name);
-
-    if (error)
-      {
-	if (devconf == 0)
-	  {
-	    pool_get (conf->dev_confs, devconf);
-	    hash_set (conf->device_config_index_by_pci_addr, d->bus_address.as_u32,
-		      devconf - conf->dev_confs);
-	    devconf->pci_addr.as_u32 = d->bus_address.as_u32;
-	  }
-	devconf->is_blacklisted = 1;
-	clib_error_report (error);
-      }
-  }));
-  /* *INDENT-ON* */
-  vec_free (pci_addr);
-}
-
-static clib_error_t *
-dpdk_device_config (dpdk_config_main_t * conf, vlib_pci_addr_t pci_addr,
-		    unformat_input_t * input, u8 is_default)
-{
-  clib_error_t *error = 0;
-  uword *p;
-  dpdk_device_config_t *devconf;
-  unformat_input_t sub_input;
-
-  if (is_default)
-    {
-      devconf = &conf->default_devconf;
-    }
-  else
-    {
-      p = hash_get (conf->device_config_index_by_pci_addr, pci_addr.as_u32);
-
-      if (!p)
-	{
-	  pool_get (conf->dev_confs, devconf);
-	  hash_set (conf->device_config_index_by_pci_addr, pci_addr.as_u32,
-		    devconf - conf->dev_confs);
-	}
-      else
-	return clib_error_return (0,
-				  "duplicate configuration for PCI address %U",
-				  format_vlib_pci_addr, &pci_addr);
-    }
-
-  devconf->pci_addr.as_u32 = pci_addr.as_u32;
-  devconf->hqos_enabled = 0;
-  dpdk_device_config_hqos_default (&devconf->hqos);
-
-  if (!input)
-    return 0;
-
-  unformat_skip_white_space (input);
-  while (unformat_check_input (input) != UNFORMAT_END_OF_INPUT)
-    {
-      if (unformat (input, "num-rx-queues %u", &devconf->num_rx_queues))
-	;
-      else if (unformat (input, "num-tx-queues %u", &devconf->num_tx_queues))
-	;
-      else if (unformat (input, "num-rx-desc %u", &devconf->num_rx_desc))
-	;
-      else if (unformat (input, "num-tx-desc %u", &devconf->num_tx_desc))
-	;
-      else if (unformat (input, "workers %U", unformat_bitmap_list,
-			 &devconf->workers))
-	;
-      else
-	if (unformat
-	    (input, "rss %U", unformat_vlib_cli_sub_input, &sub_input))
-	{
-	  error = unformat_rss_fn (&sub_input, &devconf->rss_fn);
-	  if (error)
-	    break;
-	}
-      else if (unformat (input, "vlan-strip-offload off"))
-	devconf->vlan_strip_offload = DPDK_DEVICE_VLAN_STRIP_OFF;
-      else if (unformat (input, "vlan-strip-offload on"))
-	devconf->vlan_strip_offload = DPDK_DEVICE_VLAN_STRIP_ON;
-      else
-	if (unformat
-	    (input, "hqos %U", unformat_vlib_cli_sub_input, &sub_input))
-	{
-	  devconf->hqos_enabled = 1;
-	  error = unformat_hqos (&sub_input, &devconf->hqos);
-	  if (error)
-	    break;
-	}
-      else if (unformat (input, "hqos"))
-	{
-	  devconf->hqos_enabled = 1;
-	}
-      else
-	{
-	  error = clib_error_return (0, "unknown input `%U'",
-				     format_unformat_error, input);
-	  break;
-	}
-    }
-
-  if (error)
-    return error;
-
-  if (devconf->workers && devconf->num_rx_queues == 0)
-    devconf->num_rx_queues = clib_bitmap_count_set_bits (devconf->workers);
-  else if (devconf->workers &&
-	   clib_bitmap_count_set_bits (devconf->workers) !=
-	   devconf->num_rx_queues)
-    error =
-      clib_error_return (0,
-			 "%U: number of worker threadds must be "
-			 "equal to number of rx queues", format_vlib_pci_addr,
-			 &pci_addr);
-
-  return error;
-}
-
-static clib_error_t *
-dpdk_config (vlib_main_t * vm, unformat_input_t * input)
-{
-  clib_error_t *error = 0;
-  dpdk_main_t *dm = &dpdk_main;
-  dpdk_config_main_t *conf = &dpdk_config_main;
-  vlib_thread_main_t *tm = vlib_get_thread_main ();
-  dpdk_device_config_t *devconf;
-  vlib_pci_addr_t pci_addr;
-  unformat_input_t sub_input;
-  u8 *s, *tmp = 0;
-  u8 *rte_cmd = 0, *ethname = 0;
-  u32 log_level;
-  int ret, i;
-  int num_whitelisted = 0;
-  u8 no_pci = 0;
-  u8 no_huge = 0;
-  u8 huge_dir = 0;
-  u8 file_prefix = 0;
-  u8 *socket_mem = 0;
-
-  conf->device_config_index_by_pci_addr = hash_create (0, sizeof (uword));
-
-  while (unformat_check_input (input) != UNFORMAT_END_OF_INPUT)
-    {
-      /* Prime the pump */
-      if (unformat (input, "no-hugetlb"))
-	{
-	  vec_add1 (conf->eal_init_args, (u8 *) "no-huge");
-	  no_huge = 1;
-	}
-
-      else if (unformat (input, "enable-tcp-udp-checksum"))
-	conf->enable_tcp_udp_checksum = 1;
-
-      else if (unformat (input, "decimal-interface-names"))
-	conf->interface_name_format_decimal = 1;
-
-      else if (unformat (input, "no-multi-seg"))
-	conf->no_multi_seg = 1;
-
-      else if (unformat (input, "enable-cryptodev"))
-	conf->cryptodev = 1;
-
-      else if (unformat (input, "dev default %U", unformat_vlib_cli_sub_input,
-			 &sub_input))
-	{
-	  error =
-	    dpdk_device_config (conf, (vlib_pci_addr_t) (u32) ~ 1, &sub_input,
-				1);
-
-	  if (error)
-	    return error;
-	}
-      else
-	if (unformat
-	    (input, "dev %U %U", unformat_vlib_pci_addr, &pci_addr,
-	     unformat_vlib_cli_sub_input, &sub_input))
-	{
-	  error = dpdk_device_config (conf, pci_addr, &sub_input, 0);
-
-	  if (error)
-	    return error;
-
-	  num_whitelisted++;
-	}
-      else if (unformat (input, "dev %U", unformat_vlib_pci_addr, &pci_addr))
-	{
-	  error = dpdk_device_config (conf, pci_addr, 0, 0);
-
-	  if (error)
-	    return error;
-
-	  num_whitelisted++;
-	}
-      else if (unformat (input, "num-mbufs %d", &conf->num_mbufs))
-	;
-      else if (unformat (input, "kni %d", &conf->num_kni))
-	;
-      else if (unformat (input, "uio-driver %s", &conf->uio_driver_name))
-	;
-      else if (unformat (input, "socket-mem %s", &socket_mem))
-	;
-      else if (unformat (input, "no-pci"))
-	{
-	  no_pci = 1;
-	  tmp = format (0, "--no-pci%c", 0);
-	  vec_add1 (conf->eal_init_args, tmp);
-	}
-      else if (unformat (input, "poll-sleep %d", &dm->poll_sleep))
-	;
-
-#define _(a)                                    \
-      else if (unformat(input, #a))             \
-        {                                       \
-          tmp = format (0, "--%s%c", #a, 0);    \
-          vec_add1 (conf->eal_init_args, tmp);    \
-        }
-      foreach_eal_double_hyphen_predicate_arg
-#undef _
-#define _(a)                                          \
-	else if (unformat(input, #a " %s", &s))	      \
-	  {					      \
-            if (!strncmp(#a, "huge-dir", 8))          \
-              huge_dir = 1;                           \
-            else if (!strncmp(#a, "file-prefix", 11)) \
-              file_prefix = 1;                        \
-	    tmp = format (0, "--%s%c", #a, 0);	      \
-	    vec_add1 (conf->eal_init_args, tmp);      \
-	    vec_add1 (s, 0);			      \
-            if (!strncmp(#a, "vdev", 4))              \
-              if (strstr((char*)s, "af_packet"))      \
-                clib_warning ("af_packet obsoleted. Use CLI 'create host-interface'."); \
-	    vec_add1 (conf->eal_init_args, s);	      \
-	  }
-	foreach_eal_double_hyphen_arg
-#undef _
-#define _(a,b)						\
-	  else if (unformat(input, #a " %s", &s))	\
-	    {						\
-	      tmp = format (0, "-%s%c", #b, 0);		\
-	      vec_add1 (conf->eal_init_args, tmp);	\
-	      vec_add1 (s, 0);				\
-	      vec_add1 (conf->eal_init_args, s);	\
-	    }
-	foreach_eal_single_hyphen_arg
-#undef _
-#define _(a,b)						\
-	    else if (unformat(input, #a " %s", &s))	\
-	      {						\
-		tmp = format (0, "-%s%c", #b, 0);	\
-		vec_add1 (conf->eal_init_args, tmp);	\
-		vec_add1 (s, 0);			\
-		vec_add1 (conf->eal_init_args, s);	\
-		conf->a##_set_manually = 1;		\
-	      }
-	foreach_eal_single_hyphen_mandatory_arg
-#undef _
-	else if (unformat (input, "default"))
-	;
-
-      else if (unformat_skip_white_space (input))
-	;
-      else
-	{
-	  error = clib_error_return (0, "unknown input `%U'",
-				     format_unformat_error, input);
-	  goto done;
-	}
-    }
-
-  if (!conf->uio_driver_name)
-    conf->uio_driver_name = format (0, "uio_pci_generic%c", 0);
-
-  /*
-   * Use 1G huge pages if available.
-   */
-  if (!no_huge && !huge_dir)
-    {
-      u32 x, *mem_by_socket = 0;
-      uword c = 0;
-      u8 use_1g = 1;
-      u8 use_2m = 1;
-      u8 less_than_1g = 1;
-      int rv;
-
-      umount (DEFAULT_HUGE_DIR);
-
-      /* Process "socket-mem" parameter value */
-      if (vec_len (socket_mem))
-	{
-	  unformat_input_t in;
-	  unformat_init_vector (&in, socket_mem);
-	  while (unformat_check_input (&in) != UNFORMAT_END_OF_INPUT)
-	    {
-	      if (unformat (&in, "%u,", &x))
-		;
-	      else if (unformat (&in, "%u", &x))
-		;
-	      else if (unformat (&in, ","))
-		x = 0;
-	      else
-		break;
-
-	      vec_add1 (mem_by_socket, x);
-
-	      if (x > 1023)
-		less_than_1g = 0;
-	    }
-	  /* Note: unformat_free vec_frees(in.buffer), aka socket_mem... */
-	  unformat_free (&in);
-	  socket_mem = 0;
-	}
-      else
-	{
-	  /* *INDENT-OFF* */
-	  clib_bitmap_foreach (c, tm->cpu_socket_bitmap, (
-	    {
-	      vec_validate(mem_by_socket, c);
-	      mem_by_socket[c] = 256; /* default per-socket mem */
-	    }
-	  ));
-	  /* *INDENT-ON* */
-	}
-
-      /* check if available enough 1GB pages for each socket */
-      /* *INDENT-OFF* */
-      clib_bitmap_foreach (c, tm->cpu_socket_bitmap, (
-        {
-	  int pages_avail, page_size, mem;
-
-	  vec_validate(mem_by_socket, c);
-	  mem = mem_by_socket[c];
-
-	  page_size = 1024;
-	  pages_avail = vlib_sysfs_get_free_hugepages(c, page_size * 1024);
-
-	  if (pages_avail < 0 || page_size * pages_avail < mem)
-	    use_1g = 0;
-
-	  page_size = 2;
-	  pages_avail = vlib_sysfs_get_free_hugepages(c, page_size * 1024);
-
-	  if (pages_avail < 0 || page_size * pages_avail < mem)
-	    use_2m = 0;
-      }));
-      /* *INDENT-ON* */
-
-      if (mem_by_socket == 0)
-	{
-	  error = clib_error_return (0, "mem_by_socket NULL");
-	  goto done;
-	}
-      _vec_len (mem_by_socket) = c + 1;
-
-      /* regenerate socket_mem string */
-      vec_foreach_index (x, mem_by_socket)
-	socket_mem = format (socket_mem, "%s%u",
-			     socket_mem ? "," : "", mem_by_socket[x]);
-      socket_mem = format (socket_mem, "%c", 0);
-
-      vec_free (mem_by_socket);
-
-      rv = mkdir (VPP_RUN_DIR, 0755);
-      if (rv && errno != EEXIST)
-	{
-	  error = clib_error_return (0, "mkdir '%s' failed errno %d",
-				     VPP_RUN_DIR, errno);
-	  goto done;
-	}
-
-      rv = mkdir (DEFAULT_HUGE_DIR, 0755);
-      if (rv && errno != EEXIST)
-	{
-	  error = clib_error_return (0, "mkdir '%s' failed errno %d",
-				     DEFAULT_HUGE_DIR, errno);
-	  goto done;
-	}
-
-      if (use_1g && !(less_than_1g && use_2m))
-	{
-	  rv =
-	    mount ("none", DEFAULT_HUGE_DIR, "hugetlbfs", 0, "pagesize=1G");
-	}
-      else if (use_2m)
-	{
-	  rv = mount ("none", DEFAULT_HUGE_DIR, "hugetlbfs", 0, NULL);
-	}
-      else
-	{
-	  return clib_error_return (0, "not enough free huge pages");
-	}
-
-      if (rv)
-	{
-	  error = clib_error_return (0, "mount failed %d", errno);
-	  goto done;
-	}
-
-      tmp = format (0, "--huge-dir%c", 0);
-      vec_add1 (conf->eal_init_args, tmp);
-      tmp = format (0, "%s%c", DEFAULT_HUGE_DIR, 0);
-      vec_add1 (conf->eal_init_args, tmp);
-      if (!file_prefix)
-	{
-	  tmp = format (0, "--file-prefix%c", 0);
-	  vec_add1 (conf->eal_init_args, tmp);
-	  tmp = format (0, "vpp%c", 0);
-	  vec_add1 (conf->eal_init_args, tmp);
-	}
-    }
-
-  vec_free (rte_cmd);
-  vec_free (ethname);
-
-  if (error)
-    return error;
-
-  /* I'll bet that -c and -n must be the first and second args... */
-  if (!conf->coremask_set_manually)
-    {
-      vlib_thread_registration_t *tr;
-      uword *coremask = 0;
-      int i;
-
-      /* main thread core */
-      coremask = clib_bitmap_set (coremask, tm->main_lcore, 1);
-
-      for (i = 0; i < vec_len (tm->registrations); i++)
-	{
-	  tr = tm->registrations[i];
-	  coremask = clib_bitmap_or (coremask, tr->coremask);
-	}
-
-      vec_insert (conf->eal_init_args, 2, 1);
-      conf->eal_init_args[1] = (u8 *) "-c";
-      tmp = format (0, "%U%c", format_bitmap_hex, coremask, 0);
-      conf->eal_init_args[2] = tmp;
-      clib_bitmap_free (coremask);
-    }
-
-  if (!conf->nchannels_set_manually)
-    {
-      vec_insert (conf->eal_init_args, 2, 3);
-      conf->eal_init_args[3] = (u8 *) "-n";
-      tmp = format (0, "%d", conf->nchannels);
-      conf->eal_init_args[4] = tmp;
-    }
-
-  if (no_pci == 0 && geteuid () == 0)
-    dpdk_bind_devices_to_uio (conf);
-
-#define _(x) \
-    if (devconf->x == 0 && conf->default_devconf.x > 0) \
-      devconf->x = conf->default_devconf.x ;
-
-  /* *INDENT-OFF* */
-  pool_foreach (devconf, conf->dev_confs, ({
-
-    /* default per-device config items */
-    foreach_dpdk_device_config_item
-
-    /* add DPDK EAL whitelist/blacklist entry */
-    if (num_whitelisted > 0 && devconf->is_blacklisted == 0)
-      {
-	tmp = format (0, "-w%c", 0);
-	vec_add1 (conf->eal_init_args, tmp);
-	tmp = format (0, "%U%c", format_vlib_pci_addr, &devconf->pci_addr, 0);
-	vec_add1 (conf->eal_init_args, tmp);
-      }
-    else if (num_whitelisted == 0 && devconf->is_blacklisted != 0)
-      {
-	tmp = format (0, "-b%c", 0);
-	vec_add1 (conf->eal_init_args, tmp);
-	tmp = format (0, "%U%c", format_vlib_pci_addr, &devconf->pci_addr, 0);
-	vec_add1 (conf->eal_init_args, tmp);
-      }
-  }));
-  /* *INDENT-ON* */
-
-#undef _
-
-  /* set master-lcore */
-  tmp = format (0, "--master-lcore%c", 0);
-  vec_add1 (conf->eal_init_args, tmp);
-  tmp = format (0, "%u%c", tm->main_lcore, 0);
-  vec_add1 (conf->eal_init_args, tmp);
-
-  /* set socket-mem */
-  tmp = format (0, "--socket-mem%c", 0);
-  vec_add1 (conf->eal_init_args, tmp);
-  tmp = format (0, "%s%c", socket_mem, 0);
-  vec_add1 (conf->eal_init_args, tmp);
-
-  /* NULL terminate the "argv" vector, in case of stupidity */
-  vec_add1 (conf->eal_init_args, 0);
-  _vec_len (conf->eal_init_args) -= 1;
-
-  /* Set up DPDK eal and packet mbuf pool early. */
-
-  log_level = (CLIB_DEBUG > 0) ? RTE_LOG_DEBUG : RTE_LOG_NOTICE;
-
-  rte_set_log_level (log_level);
-
-  vm = vlib_get_main ();
-
-  /* make copy of args as rte_eal_init tends to mess up with arg array */
-  for (i = 1; i < vec_len (conf->eal_init_args); i++)
-    conf->eal_init_args_str = format (conf->eal_init_args_str, "%s ",
-				      conf->eal_init_args[i]);
-
-  ret =
-    rte_eal_init (vec_len (conf->eal_init_args),
-		  (char **) conf->eal_init_args);
-
-  /* lazy umount hugepages */
-  umount2 (DEFAULT_HUGE_DIR, MNT_DETACH);
-
-  if (ret < 0)
-    return clib_error_return (0, "rte_eal_init returned %d", ret);
-
-  /* Dump the physical memory layout prior to creating the mbuf_pool */
-  fprintf (stdout, "DPDK physical memory layout:\n");
-  rte_dump_physmem_layout (stdout);
-
-  /* main thread 1st */
-  error = vlib_buffer_pool_create (vm, conf->num_mbufs, rte_socket_id ());
-  if (error)
-    return error;
-
-  for (i = 0; i < RTE_MAX_LCORE; i++)
-    {
-      error = vlib_buffer_pool_create (vm, conf->num_mbufs,
-				       rte_lcore_to_socket_id (i));
-      if (error)
-	return error;
-    }
-
-done:
-  return error;
-}
-
-VLIB_CONFIG_FUNCTION (dpdk_config, "dpdk");
-
-void
-dpdk_update_link_state (dpdk_device_t * xd, f64 now)
-{
-  vnet_main_t *vnm = vnet_get_main ();
-  struct rte_eth_link prev_link = xd->link;
-  u32 hw_flags = 0;
-  u8 hw_flags_chg = 0;
-
-  /* only update link state for PMD interfaces */
-  if ((xd->flags & DPDK_DEVICE_FLAG_PMD) == 0)
-    return;
-
-  xd->time_last_link_update = now ? now : xd->time_last_link_update;
-  memset (&xd->link, 0, sizeof (xd->link));
-  rte_eth_link_get_nowait (xd->device_index, &xd->link);
-
-  if (LINK_STATE_ELOGS)
-    {
-      vlib_main_t *vm = vlib_get_main ();
-      ELOG_TYPE_DECLARE (e) =
-      {
-      .format =
-	  "update-link-state: sw_if_index %d, admin_up %d,"
-	  "old link_state %d new link_state %d",.format_args = "i4i1i1i1",};
-
-      struct
-      {
-	u32 sw_if_index;
-	u8 admin_up;
-	u8 old_link_state;
-	u8 new_link_state;
-      } *ed;
-      ed = ELOG_DATA (&vm->elog_main, e);
-      ed->sw_if_index = xd->vlib_sw_if_index;
-      ed->admin_up = (xd->flags & DPDK_DEVICE_FLAG_ADMIN_UP) != 0;
-      ed->old_link_state = (u8)
-	vnet_hw_interface_is_link_up (vnm, xd->vlib_hw_if_index);
-      ed->new_link_state = (u8) xd->link.link_status;
-    }
-
-  if ((xd->flags & DPDK_DEVICE_FLAG_ADMIN_UP) &&
-      ((xd->link.link_status != 0) ^
-       vnet_hw_interface_is_link_up (vnm, xd->vlib_hw_if_index)))
-    {
-      hw_flags_chg = 1;
-      hw_flags |= (xd->link.link_status ? VNET_HW_INTERFACE_FLAG_LINK_UP : 0);
-    }
-
-  if (hw_flags_chg || (xd->link.link_duplex != prev_link.link_duplex))
-    {
-      hw_flags_chg = 1;
-      switch (xd->link.link_duplex)
-	{
-	case ETH_LINK_HALF_DUPLEX:
-	  hw_flags |= VNET_HW_INTERFACE_FLAG_HALF_DUPLEX;
-	  break;
-	case ETH_LINK_FULL_DUPLEX:
-	  hw_flags |= VNET_HW_INTERFACE_FLAG_FULL_DUPLEX;
-	  break;
-	default:
-	  break;
-	}
-    }
-  if (hw_flags_chg || (xd->link.link_speed != prev_link.link_speed))
-    {
-      hw_flags_chg = 1;
-      switch (xd->link.link_speed)
-	{
-	case ETH_SPEED_NUM_10M:
-	  hw_flags |= VNET_HW_INTERFACE_FLAG_SPEED_10M;
-	  break;
-	case ETH_SPEED_NUM_100M:
-	  hw_flags |= VNET_HW_INTERFACE_FLAG_SPEED_100M;
-	  break;
-	case ETH_SPEED_NUM_1G:
-	  hw_flags |= VNET_HW_INTERFACE_FLAG_SPEED_1G;
-	  break;
-	case ETH_SPEED_NUM_10G:
-	  hw_flags |= VNET_HW_INTERFACE_FLAG_SPEED_10G;
-	  break;
-	case ETH_SPEED_NUM_40G:
-	  hw_flags |= VNET_HW_INTERFACE_FLAG_SPEED_40G;
-	  break;
-	case 0:
-	  break;
-	default:
-	  clib_warning ("unknown link speed %d", xd->link.link_speed);
-	  break;
-	}
-    }
-  if (hw_flags_chg)
-    {
-      if (LINK_STATE_ELOGS)
-	{
-	  vlib_main_t *vm = vlib_get_main ();
-
-	  ELOG_TYPE_DECLARE (e) =
-	  {
-	  .format =
-	      "update-link-state: sw_if_index %d, new flags %d",.format_args
-	      = "i4i4",};
-
-	  struct
-	  {
-	    u32 sw_if_index;
-	    u32 flags;
-	  } *ed;
-	  ed = ELOG_DATA (&vm->elog_main, e);
-	  ed->sw_if_index = xd->vlib_sw_if_index;
-	  ed->flags = hw_flags;
-	}
-      vnet_hw_interface_set_flags (vnm, xd->vlib_hw_if_index, hw_flags);
-    }
-}
-
-static uword
-dpdk_process (vlib_main_t * vm, vlib_node_runtime_t * rt, vlib_frame_t * f)
-{
-  clib_error_t *error;
-  vnet_main_t *vnm = vnet_get_main ();
-  dpdk_main_t *dm = &dpdk_main;
-  ethernet_main_t *em = ðernet_main;
-  dpdk_device_t *xd;
-  vlib_thread_main_t *tm = vlib_get_thread_main ();
-  int i;
-
-  error = dpdk_lib_init (dm);
-
-  /*
-   * Turn on the input node if we found some devices to drive
-   * and we're not running worker threads or i/o threads
-   */
-
-  if (error == 0 && vec_len (dm->devices) > 0)
-    {
-      if (tm->n_vlib_mains == 1)
-	vlib_node_set_state (vm, dpdk_input_node.index,
-			     VLIB_NODE_STATE_POLLING);
-      else
-	for (i = 0; i < tm->n_vlib_mains; i++)
-	  if (vec_len (dm->devices_by_cpu[i]) > 0)
-	    vlib_node_set_state (vlib_mains[i], dpdk_input_node.index,
-				 VLIB_NODE_STATE_POLLING);
-    }
-
-  if (error)
-    clib_error_report (error);
-
-  tm->worker_thread_release = 1;
-
-  f64 now = vlib_time_now (vm);
-  vec_foreach (xd, dm->devices)
-  {
-    dpdk_update_link_state (xd, now);
-  }
-
-  {
-    /*
-     * Extra set up for bond interfaces:
-     *  1. Setup MACs for bond interfaces and their slave links which was set
-     *     in dpdk_port_setup() but needs to be done again here to take effect.
-     *  2. Set up info for bond interface related CLI support.
-     */
-    int nports = rte_eth_dev_count ();
-    if (nports > 0)
-      {
-	for (i = 0; i < nports; i++)
-	  {
-	    struct rte_eth_dev_info dev_info;
-	    rte_eth_dev_info_get (i, &dev_info);
-	    if (!dev_info.driver_name)
-	      dev_info.driver_name = dev_info.pci_dev->driver->driver.name;
-
-	    ASSERT (dev_info.driver_name);
-	    if (strncmp (dev_info.driver_name, "rte_bond_pmd", 12) == 0)
-	      {
-		u8 addr[6];
-		u8 slink[16];
-		int nlink = rte_eth_bond_slaves_get (i, slink, 16);
-		if (nlink > 0)
-		  {
-		    vnet_hw_interface_t *bhi;
-		    ethernet_interface_t *bei;
-		    int rv;
-
-		    /* Get MAC of 1st slave link */
-		    rte_eth_macaddr_get (slink[0],
-					 (struct ether_addr *) addr);
-		    /* Set MAC of bounded interface to that of 1st slave link */
-		    rv =
-		      rte_eth_bond_mac_address_set (i,
-						    (struct ether_addr *)
-						    addr);
-		    if (rv < 0)
-		      clib_warning ("Failed to set MAC address");
-
-		    /* Populate MAC of bonded interface in VPP hw tables */
-		    bhi =
-		      vnet_get_hw_interface (vnm,
-					     dm->devices[i].vlib_hw_if_index);
-		    bei =
-		      pool_elt_at_index (em->interfaces, bhi->hw_instance);
-		    clib_memcpy (bhi->hw_address, addr, 6);
-		    clib_memcpy (bei->address, addr, 6);
-		    /* Init l3 packet size allowed on bonded interface */
-		    bhi->max_packet_bytes = ETHERNET_MAX_PACKET_BYTES;
-		    bhi->max_l3_packet_bytes[VLIB_RX] =
-		      bhi->max_l3_packet_bytes[VLIB_TX] =
-		      ETHERNET_MAX_PACKET_BYTES - sizeof (ethernet_header_t);
-		    while (nlink >= 1)
-		      {		/* for all slave links */
-			int slave = slink[--nlink];
-			dpdk_device_t *sdev = &dm->devices[slave];
-			vnet_hw_interface_t *shi;
-			vnet_sw_interface_t *ssi;
-			/* Add MAC to all slave links except the first one */
-			if (nlink)
-			  rte_eth_dev_mac_addr_add (slave,
-						    (struct ether_addr *)
-						    addr, 0);
-			/* Set slaves bitmap for bonded interface */
-			bhi->bond_info =
-			  clib_bitmap_set (bhi->bond_info,
-					   sdev->vlib_hw_if_index, 1);
-			/* Set slave link flags on slave interface */
-			shi =
-			  vnet_get_hw_interface (vnm, sdev->vlib_hw_if_index);
-			ssi =
-			  vnet_get_sw_interface (vnm, sdev->vlib_sw_if_index);
-			shi->bond_info = VNET_HW_INTERFACE_BOND_INFO_SLAVE;
-			ssi->flags |= VNET_SW_INTERFACE_FLAG_BOND_SLAVE;
-
-			/* Set l3 packet size allowed as the lowest of slave */
-			if (bhi->max_l3_packet_bytes[VLIB_RX] >
-			    shi->max_l3_packet_bytes[VLIB_RX])
-			  bhi->max_l3_packet_bytes[VLIB_RX] =
-			    bhi->max_l3_packet_bytes[VLIB_TX] =
-			    shi->max_l3_packet_bytes[VLIB_RX];
-
-			/* Set max packet size allowed as the lowest of slave */
-			if (bhi->max_packet_bytes > shi->max_packet_bytes)
-			  bhi->max_packet_bytes = shi->max_packet_bytes;
-		      }
-		  }
-	      }
-	  }
-      }
-  }
-
-  while (1)
-    {
-      /*
-       * check each time through the loop in case intervals are changed
-       */
-      f64 min_wait = dm->link_state_poll_interval < dm->stat_poll_interval ?
-	dm->link_state_poll_interval : dm->stat_poll_interval;
-
-      vlib_process_wait_for_event_or_clock (vm, min_wait);
-
-      if (dm->admin_up_down_in_progress)
-	/* skip the poll if an admin up down is in progress (on any interface) */
-	continue;
-
-      vec_foreach (xd, dm->devices)
-      {
-	f64 now = vlib_time_now (vm);
-	if ((now - xd->time_last_stats_update) >= dm->stat_poll_interval)
-	  dpdk_update_counters (xd, now);
-	if ((now - xd->time_last_link_update) >= dm->link_state_poll_interval)
-	  dpdk_update_link_state (xd, now);
-
-      }
-    }
-
-  return 0;
-}
-
-/* *INDENT-OFF* */
-VLIB_REGISTER_NODE (dpdk_process_node,static) = {
-    .function = dpdk_process,
-    .type = VLIB_NODE_TYPE_PROCESS,
-    .name = "dpdk-process",
-    .process_log2_n_stack_bytes = 17,
-};
-/* *INDENT-ON* */
-
-int
-dpdk_set_stat_poll_interval (f64 interval)
-{
-  if (interval < DPDK_MIN_STATS_POLL_INTERVAL)
-    return (VNET_API_ERROR_INVALID_VALUE);
-
-  dpdk_main.stat_poll_interval = interval;
-
-  return 0;
-}
-
-int
-dpdk_set_link_state_poll_interval (f64 interval)
-{
-  if (interval < DPDK_MIN_LINK_POLL_INTERVAL)
-    return (VNET_API_ERROR_INVALID_VALUE);
-
-  dpdk_main.link_state_poll_interval = interval;
-
-  return 0;
-}
-
-clib_error_t *
-dpdk_init (vlib_main_t * vm)
-{
-  dpdk_main_t *dm = &dpdk_main;
-  vlib_node_t *ei;
-  clib_error_t *error = 0;
-  vlib_thread_main_t *tm = vlib_get_thread_main ();
-
-  /* verify that structs are cacheline aligned */
-  STATIC_ASSERT (offsetof (dpdk_device_t, cacheline0) == 0,
-		 "Cache line marker must be 1st element in dpdk_device_t");
-  STATIC_ASSERT (offsetof (dpdk_device_t, cacheline1) ==
-		 CLIB_CACHE_LINE_BYTES,
-		 "Data in cache line 0 is bigger than cache line size");
-  STATIC_ASSERT (offsetof (frame_queue_trace_t, cacheline0) == 0,
-		 "Cache line marker must be 1st element in frame_queue_trace_t");
-
-  dm->vlib_main = vm;
-  dm->vnet_main = vnet_get_main ();
-  dm->conf = &dpdk_config_main;
-
-  ei = vlib_get_node_by_name (vm, (u8 *) "ethernet-input");
-  if (ei == 0)
-    return clib_error_return (0, "ethernet-input node AWOL");
-
-  dm->ethernet_input_node_index = ei->index;
-
-  dm->conf->nchannels = 4;
-  dm->conf->num_mbufs = dm->conf->num_mbufs ? dm->conf->num_mbufs : NB_MBUF;
-  vec_add1 (dm->conf->eal_init_args, (u8 *) "vnet");
-
-  dm->dpdk_device_by_kni_port_id = hash_create (0, sizeof (uword));
-  dm->vu_sw_if_index_by_listener_fd = hash_create (0, sizeof (uword));
-  dm->vu_sw_if_index_by_sock_fd = hash_create (0, sizeof (uword));
-
-  /* $$$ use n_thread_stacks since it's known-good at this point */
-  vec_validate (dm->recycle, tm->n_thread_stacks - 1);
-
-  /* Default vlib_buffer_t flags, DISABLES tcp/udp checksumming... */
-  dm->buffer_flags_template =
-    (VLIB_BUFFER_TOTAL_LENGTH_VALID | VLIB_BUFFER_EXT_HDR_VALID
-     | IP_BUFFER_L4_CHECKSUM_COMPUTED | IP_BUFFER_L4_CHECKSUM_CORRECT);
-
-  dm->stat_poll_interval = DPDK_STATS_POLL_INTERVAL;
-  dm->link_state_poll_interval = DPDK_LINK_POLL_INTERVAL;
-
-  /* init CLI */
-  if ((error = vlib_call_init_function (vm, dpdk_cli_init)))
-    return error;
-
-  return error;
-}
-
-VLIB_INIT_FUNCTION (dpdk_init);
-
-
-/*
- * fd.io coding-style-patch-verification: ON
- *
- * Local Variables:
- * eval: (c-set-style "gnu")
- * End:
- */
diff --git a/src/vnet/devices/dpdk/ipsec/cli.c b/src/vnet/devices/dpdk/ipsec/cli.c
deleted file mode 100644
index f9d3a5d0..00000000
--- a/src/vnet/devices/dpdk/ipsec/cli.c
+++ /dev/null
@@ -1,154 +0,0 @@
-/*
- * Copyright (c) 2016 Intel and/or its affiliates.
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at:
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#include 
-#include 
-#include 
-
-static void
-dpdk_ipsec_show_mapping (vlib_main_t * vm, u16 detail_display)
-{
-  dpdk_config_main_t *conf = &dpdk_config_main;
-  dpdk_crypto_main_t *dcm = &dpdk_crypto_main;
-  vlib_thread_main_t *tm = vlib_get_thread_main ();
-  u32 i, skip_master;
-
-  if (!conf->cryptodev)
-    {
-      vlib_cli_output (vm, "DPDK Cryptodev support is disabled\n");
-      return;
-    }
-
-  if (detail_display)
-    vlib_cli_output (vm, "worker\t%10s\t%15s\tdir\tdev\tqp\n",
-		     "cipher", "auth");
-  else
-    vlib_cli_output (vm, "worker\tcrypto device id(type)\n");
-
-  skip_master = vlib_num_workers () > 0;
-
-  for (i = 0; i < tm->n_vlib_mains; i++)
-    {
-      uword key, data;
-      u32 cpu_index = vlib_mains[i]->cpu_index;
-      crypto_worker_main_t *cwm = &dcm->workers_main[cpu_index];
-      u8 *s = 0;
-
-      if (skip_master)
-	{
-	  skip_master = 0;
-	  continue;
-	}
-
-      if (!detail_display)
-	{
-	  i32 last_cdev = -1;
-	  crypto_qp_data_t *qpd;
-
-	  s = format (s, "%u\t", cpu_index);
-
-	  /* *INDENT-OFF* */
-	  vec_foreach (qpd, cwm->qp_data)
-	    {
-	      u32 dev_id = qpd->dev_id;
-
-	      if ((u16) last_cdev != dev_id)
-		{
-		  struct rte_cryptodev_info cdev_info;
-
-		  rte_cryptodev_info_get (dev_id, &cdev_info);
-
-		  s = format(s, "%u(%s)\t", dev_id, cdev_info.feature_flags &
-			     RTE_CRYPTODEV_FF_HW_ACCELERATED ? "HW" : "SW");
-		}
-	      last_cdev = dev_id;
-	    }
-	  /* *INDENT-ON* */
-	  vlib_cli_output (vm, "%s", s);
-	}
-      else
-	{
-	  char cipher_str[15], auth_str[15];
-	  struct rte_cryptodev_capabilities cap;
-	  crypto_worker_qp_key_t *p_key = (crypto_worker_qp_key_t *) & key;
-	  /* *INDENT-OFF* */
-	  hash_foreach (key, data, cwm->algo_qp_map,
-	  ({
-	    cap.op = RTE_CRYPTO_OP_TYPE_SYMMETRIC;
-	    cap.sym.xform_type = RTE_CRYPTO_SYM_XFORM_CIPHER;
-	    cap.sym.cipher.algo = p_key->cipher_algo;
-	    check_algo_is_supported (&cap, cipher_str);
-	    cap.op = RTE_CRYPTO_OP_TYPE_SYMMETRIC;
-	    cap.sym.xform_type = RTE_CRYPTO_SYM_XFORM_AUTH;
-	    cap.sym.auth.algo = p_key->auth_algo;
-	    check_algo_is_supported (&cap, auth_str);
-	    vlib_cli_output (vm, "%u\t%10s\t%15s\t%3s\t%u\t%u\n",
-			     vlib_mains[i]->cpu_index, cipher_str, auth_str,
-			     p_key->is_outbound ? "out" : "in",
-			     cwm->qp_data[data].dev_id,
-			     cwm->qp_data[data].qp_id);
-	  }));
-	  /* *INDENT-ON* */
-	}
-    }
-}
-
-static clib_error_t *
-lcore_cryptodev_map_fn (vlib_main_t * vm, unformat_input_t * input,
-			vlib_cli_command_t * cmd)
-{
-  unformat_input_t _line_input, *line_input = &_line_input;
-  u16 detail = 0;
-  clib_error_t *error = NULL;
-
-  if (!unformat_user (input, unformat_line_input, line_input))
-    return 0;
-
-  while (unformat_check_input (line_input) != UNFORMAT_END_OF_INPUT)
-    {
-      if (unformat (line_input, "verbose"))
-	detail = 1;
-      else
-	{
-	  error = clib_error_return (0, "parse error: '%U'",
-				     format_unformat_error, line_input);
-	  goto done;
-	}
-    }
-
-  dpdk_ipsec_show_mapping (vm, detail);
-
-done:
-  unformat_free (line_input);
-
-  return error;
-}
-
-/* *INDENT-OFF* */
-VLIB_CLI_COMMAND (lcore_cryptodev_map, static) = {
-    .path = "show crypto device mapping",
-    .short_help =
-    "show cryptodev device mapping ",
-    .function = lcore_cryptodev_map_fn,
-};
-/* *INDENT-ON* */
-
-/*
- * fd.io coding-style-patch-verification: ON
- *
- * Local Variables:
- * eval: (c-set-style "gnu")
- * End:
- */
diff --git a/src/vnet/devices/dpdk/ipsec/crypto_node.c b/src/vnet/devices/dpdk/ipsec/crypto_node.c
deleted file mode 100644
index e8fef235..00000000
--- a/src/vnet/devices/dpdk/ipsec/crypto_node.c
+++ /dev/null
@@ -1,215 +0,0 @@
-/*
- *------------------------------------------------------------------
- * crypto_node.c - DPDK Cryptodev input node
- *
- * Copyright (c) 2016 Intel and/or its affiliates.
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at:
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- *------------------------------------------------------------------
- */
-
-#include 
-#include 
-#include 
-#include 
-
-#include 
-#include 
-#include 
-
-#define foreach_dpdk_crypto_input_next		\
-  _(DROP, "error-drop")				\
-  _(ENCRYPT_POST, "dpdk-esp-encrypt-post")	\
-  _(DECRYPT_POST, "dpdk-esp-decrypt-post")
-
-typedef enum
-{
-#define _(f,s) DPDK_CRYPTO_INPUT_NEXT_##f,
-  foreach_dpdk_crypto_input_next
-#undef _
-    DPDK_CRYPTO_INPUT_N_NEXT,
-} dpdk_crypto_input_next_t;
-
-#define foreach_dpdk_crypto_input_error		\
-  _(DQ_COPS, "Crypto ops dequeued")		\
-  _(COP_FAILED, "Crypto op failed")
-
-typedef enum
-{
-#define _(f,s) DPDK_CRYPTO_INPUT_ERROR_##f,
-  foreach_dpdk_crypto_input_error
-#undef _
-    DPDK_CRYPTO_INPUT_N_ERROR,
-} dpdk_crypto_input_error_t;
-
-static char *dpdk_crypto_input_error_strings[] = {
-#define _(n, s) s,
-  foreach_dpdk_crypto_input_error
-#undef _
-};
-
-vlib_node_registration_t dpdk_crypto_input_node;
-
-typedef struct
-{
-  u32 cdev;
-  u32 qp;
-  u32 status;
-  u32 sa_idx;
-  u32 next_index;
-} dpdk_crypto_input_trace_t;
-
-static u8 *
-format_dpdk_crypto_input_trace (u8 * s, va_list * args)
-{
-  CLIB_UNUSED (vlib_main_t * vm) = va_arg (*args, vlib_main_t *);
-  CLIB_UNUSED (vlib_node_t * node) = va_arg (*args, vlib_node_t *);
-  dpdk_crypto_input_trace_t *t = va_arg (*args, dpdk_crypto_input_trace_t *);
-
-  s = format (s, "dpdk_crypto: cryptodev-id %u queue-pair %u next-index %d",
-	      t->cdev, t->qp, t->next_index);
-
-  s = format (s, " status %u sa-idx %u\n", t->status, t->sa_idx);
-
-  return s;
-}
-
-static_always_inline u32
-dpdk_crypto_dequeue (vlib_main_t * vm, vlib_node_runtime_t * node,
-		     crypto_qp_data_t * qpd)
-{
-  u32 n_deq, *to_next = 0, next_index, n_cops, def_next_index;
-  struct rte_crypto_op **cops = qpd->cops;
-
-  if (qpd->inflights == 0)
-    return 0;
-
-  if (qpd->is_outbound)
-    def_next_index = DPDK_CRYPTO_INPUT_NEXT_ENCRYPT_POST;
-  else
-    def_next_index = DPDK_CRYPTO_INPUT_NEXT_DECRYPT_POST;
-
-  n_cops = rte_cryptodev_dequeue_burst (qpd->dev_id, qpd->qp_id,
-					cops, VLIB_FRAME_SIZE);
-  n_deq = n_cops;
-  next_index = def_next_index;
-
-  qpd->inflights -= n_cops;
-  ASSERT (qpd->inflights >= 0);
-
-  while (n_cops > 0)
-    {
-      u32 n_left_to_next;
-
-      vlib_get_next_frame (vm, node, next_index, to_next, n_left_to_next);
-
-      while (n_cops > 0 && n_left_to_next > 0)
-	{
-	  u32 bi0, next0;
-	  vlib_buffer_t *b0 = 0;
-	  struct rte_crypto_op *cop;
-	  struct rte_crypto_sym_op *sym_cop;
-
-	  cop = cops[0];
-	  cops += 1;
-	  n_cops -= 1;
-	  n_left_to_next -= 1;
-
-	  next0 = def_next_index;
-
-	  if (PREDICT_FALSE (cop->status != RTE_CRYPTO_OP_STATUS_SUCCESS))
-	    {
-	      next0 = DPDK_CRYPTO_INPUT_NEXT_DROP;
-	      vlib_node_increment_counter (vm, dpdk_crypto_input_node.index,
-					   DPDK_CRYPTO_INPUT_ERROR_COP_FAILED,
-					   1);
-	    }
-	  cop->status = RTE_CRYPTO_OP_STATUS_NOT_PROCESSED;
-
-	  sym_cop = (struct rte_crypto_sym_op *) (cop + 1);
-	  b0 = vlib_buffer_from_rte_mbuf (sym_cop->m_src);
-	  bi0 = vlib_get_buffer_index (vm, b0);
-
-	  to_next[0] = bi0;
-	  to_next += 1;
-
-	  if (PREDICT_FALSE (b0->flags & VLIB_BUFFER_IS_TRACED))
-	    {
-	      vlib_trace_next_frame (vm, node, next0);
-	      dpdk_crypto_input_trace_t *tr =
-		vlib_add_trace (vm, node, b0, sizeof (*tr));
-	      tr->cdev = qpd->dev_id;
-	      tr->qp = qpd->qp_id;
-	      tr->status = cop->status;
-	      tr->next_index = next0;
-	      tr->sa_idx = vnet_buffer (b0)->ipsec.sad_index;
-	    }
-
-	  vlib_validate_buffer_enqueue_x1 (vm, node, next_index, to_next,
-					   n_left_to_next, bi0, next0);
-	}
-      vlib_put_next_frame (vm, node, next_index, n_left_to_next);
-    }
-
-  crypto_free_cop (qpd, qpd->cops, n_deq);
-
-  vlib_node_increment_counter (vm, dpdk_crypto_input_node.index,
-			       DPDK_CRYPTO_INPUT_ERROR_DQ_COPS, n_deq);
-  return n_deq;
-}
-
-static uword
-dpdk_crypto_input_fn (vlib_main_t * vm, vlib_node_runtime_t * node,
-		      vlib_frame_t * frame)
-{
-  u32 cpu_index = os_get_cpu_number ();
-  dpdk_crypto_main_t *dcm = &dpdk_crypto_main;
-  crypto_worker_main_t *cwm = &dcm->workers_main[cpu_index];
-  crypto_qp_data_t *qpd;
-  u32 n_deq = 0;
-
-  /* *INDENT-OFF* */
-  vec_foreach (qpd, cwm->qp_data)
-      n_deq += dpdk_crypto_dequeue(vm, node, qpd);
-  /* *INDENT-ON* */
-
-  return n_deq;
-}
-
-/* *INDENT-OFF* */
-VLIB_REGISTER_NODE (dpdk_crypto_input_node) =
-{
-  .function = dpdk_crypto_input_fn,
-  .name = "dpdk-crypto-input",
-  .format_trace = format_dpdk_crypto_input_trace,
-  .type = VLIB_NODE_TYPE_INPUT,
-  .state = VLIB_NODE_STATE_DISABLED,
-  .n_errors = DPDK_CRYPTO_INPUT_N_ERROR,
-  .error_strings = dpdk_crypto_input_error_strings,
-  .n_next_nodes = DPDK_CRYPTO_INPUT_N_NEXT,
-  .next_nodes =
-  {
-#define _(s,n) [DPDK_CRYPTO_INPUT_NEXT_##s] = n,
-    foreach_dpdk_crypto_input_next
-#undef _
-  },
-};
-/* *INDENT-ON* */
-
-VLIB_NODE_FUNCTION_MULTIARCH (dpdk_crypto_input_node, dpdk_crypto_input_fn)
-/*
- * fd.io coding-style-patch-verification: ON
- *
- * Local Variables:
- * eval: (c-set-style "gnu")
- * End:
- */
diff --git a/src/vnet/devices/dpdk/ipsec/dir.dox b/src/vnet/devices/dpdk/ipsec/dir.dox
deleted file mode 100644
index ffebfc4d..00000000
--- a/src/vnet/devices/dpdk/ipsec/dir.dox
+++ /dev/null
@@ -1,18 +0,0 @@
-/*
- * Copyright (c) 2016 Intel and/or its affiliates.
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at:
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-/**
- @dir vnet/vnet/devices/dpdk/ipsec
- @brief IPSec ESP encrypt/decrypt using DPDK Cryptodev API
-*/
diff --git a/src/vnet/devices/dpdk/ipsec/dpdk_crypto_ipsec_doc.md b/src/vnet/devices/dpdk/ipsec/dpdk_crypto_ipsec_doc.md
deleted file mode 100644
index fed2fe0e..00000000
--- a/src/vnet/devices/dpdk/ipsec/dpdk_crypto_ipsec_doc.md
+++ /dev/null
@@ -1,86 +0,0 @@
-# VPP IPSec implementation using DPDK Cryptodev API    {#dpdk_crypto_ipsec_doc}
-
-This document is meant to contain all related information about implementation and usability.
-
-
-## VPP IPsec with DPDK Cryptodev
-
-DPDK Cryptodev is an asynchronous crypto API that supports both Hardware and Software implementations (for more details refer to [DPDK Cryptography Device Library documentation](http://dpdk.org/doc/guides/prog_guide/cryptodev_lib.html)).
-
-When DPDK support is enabled and there are enough Cryptodev resources for all workers, the node graph is reconfigured by adding and changing default next nodes.
-
-The following nodes are added:
-* dpdk-crypto-input : polling input node, basically dequeuing from crypto devices.
-* dpdk-esp-encrypt : internal node.
-* dpdk-esp-decrypt : internal node.
-* dpdk-esp-encrypt-post : internal node.
-* dpdk-esp-decrypt-post : internal node.
-
-Set new default next nodes:
-* for esp encryption: esp-encrypt -> dpdk-esp-encrypt
-* for esp decryption: esp-decrypt -> dpdk-esp-decrypt
-
-
-### How to enable VPP IPSec with DPDK Cryptodev support
-
-DPDK Cryptodev is supported in DPDK enabled VPP.
-By default, only HW Cryptodev is supported but needs to be explicetly enabled with the following config option:
-
-```
-dpdk {
-    enable-cryptodev
-}
-```
-
-To enable SW Cryptodev support (AESNI-MB-PMD and GCM-PMD), we need the following env option:
-
-    vpp_uses_dpdk_cryptodev_sw=yes
-
-A couple of ways to achive this:
-* uncomment/add it in the platforms config (ie. build-data/platforms/vpp.mk)
-* set the option when building vpp (ie. make vpp_uses_dpdk_cryptodev_sw=yes build-release)
-
-When enabling SW Cryptodev support, it means that you need to pre-build the required crypto libraries needed by those SW Cryptodev PMDs.
-
-
-### Crypto Resources allocation
-
-VPP allocates crypto resources based on a best effort approach:
-* first allocate Hardware crypto resources, then Software.
-* if there are not enough crypto resources for all workers, the graph node is not modifed, therefore the default VPP IPsec implementation based in OpenSSL is used. The following message is displayed:
-
-      0: dpdk_ipsec_init: not enough cryptodevs for ipsec
-
-
-### Configuration example
-
-To enable DPDK Cryptodev the user just need to provide the startup.conf option
-as mentioned previously.
-
-Example startup.conf:
-
-```
-dpdk {
-    socket-mem 1024,1024
-    num-mbufs 131072
-    dev 0000:81:00.0
-    dev 0000:81:00.1
-    enable-cryptodev
-    dev 0000:85:01.0
-    dev 0000:85:01.1
-    vdev cryptodev_aesni_mb_pmd,socket_id=1
-    vdev cryptodev_aesni_mb_pmd,socket_id=1
-}
-```
-
-In the above configuration:
-* 0000:85:01.0 and 0000:85:01.1 are crypto BDFs and they require the same driver binding as DPDK Ethernet devices but they do not support any extra configuration options.
-* Two AESNI-MB Software Cryptodev PMDs are created in NUMA node 1.
-
-For further details refer to [DPDK Crypto Device Driver documentation](http://dpdk.org/doc/guides/cryptodevs/index.html)
-
-### Operational data
-
-The following CLI command displays the Cryptodev/Worker mapping:
-
-    show crypto device mapping [verbose]
diff --git a/src/vnet/devices/dpdk/ipsec/esp.h b/src/vnet/devices/dpdk/ipsec/esp.h
deleted file mode 100644
index d0b27618..00000000
--- a/src/vnet/devices/dpdk/ipsec/esp.h
+++ /dev/null
@@ -1,249 +0,0 @@
-/*
- * Copyright (c) 2016 Intel and/or its affiliates.
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at:
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-#ifndef __DPDK_ESP_H__
-#define __DPDK_ESP_H__
-
-#include 
-#include 
-#include 
-
-typedef struct
-{
-  enum rte_crypto_cipher_algorithm algo;
-  u8 key_len;
-  u8 iv_len;
-} dpdk_esp_crypto_alg_t;
-
-typedef struct
-{
-  enum rte_crypto_auth_algorithm algo;
-  u8 trunc_size;
-} dpdk_esp_integ_alg_t;
-
-typedef struct
-{
-  dpdk_esp_crypto_alg_t *esp_crypto_algs;
-  dpdk_esp_integ_alg_t *esp_integ_algs;
-} dpdk_esp_main_t;
-
-dpdk_esp_main_t dpdk_esp_main;
-
-static_always_inline void
-dpdk_esp_init ()
-{
-  dpdk_esp_main_t *em = &dpdk_esp_main;
-  dpdk_esp_integ_alg_t *i;
-  dpdk_esp_crypto_alg_t *c;
-
-  vec_validate (em->esp_crypto_algs, IPSEC_CRYPTO_N_ALG - 1);
-
-  c = &em->esp_crypto_algs[IPSEC_CRYPTO_ALG_AES_CBC_128];
-  c->algo = RTE_CRYPTO_CIPHER_AES_CBC;
-  c->key_len = 16;
-  c->iv_len = 16;
-
-  c = &em->esp_crypto_algs[IPSEC_CRYPTO_ALG_AES_CBC_192];
-  c->algo = RTE_CRYPTO_CIPHER_AES_CBC;
-  c->key_len = 24;
-  c->iv_len = 16;
-
-  c = &em->esp_crypto_algs[IPSEC_CRYPTO_ALG_AES_CBC_256];
-  c->algo = RTE_CRYPTO_CIPHER_AES_CBC;
-  c->key_len = 32;
-  c->iv_len = 16;
-
-  c = &em->esp_crypto_algs[IPSEC_CRYPTO_ALG_AES_GCM_128];
-  c->algo = RTE_CRYPTO_CIPHER_AES_GCM;
-  c->key_len = 16;
-  c->iv_len = 8;
-
-  vec_validate (em->esp_integ_algs, IPSEC_INTEG_N_ALG - 1);
-
-  i = &em->esp_integ_algs[IPSEC_INTEG_ALG_SHA1_96];
-  i->algo = RTE_CRYPTO_AUTH_SHA1_HMAC;
-  i->trunc_size = 12;
-
-  i = &em->esp_integ_algs[IPSEC_INTEG_ALG_SHA_256_96];
-  i->algo = RTE_CRYPTO_AUTH_SHA256_HMAC;
-  i->trunc_size = 12;
-
-  i = &em->esp_integ_algs[IPSEC_INTEG_ALG_SHA_256_128];
-  i->algo = RTE_CRYPTO_AUTH_SHA256_HMAC;
-  i->trunc_size = 16;
-
-  i = &em->esp_integ_algs[IPSEC_INTEG_ALG_SHA_384_192];
-  i->algo = RTE_CRYPTO_AUTH_SHA384_HMAC;
-  i->trunc_size = 24;
-
-  i = &em->esp_integ_algs[IPSEC_INTEG_ALG_SHA_512_256];
-  i->algo = RTE_CRYPTO_AUTH_SHA512_HMAC;
-  i->trunc_size = 32;
-
-  i = &em->esp_integ_algs[IPSEC_INTEG_ALG_AES_GCM_128];
-  i->algo = RTE_CRYPTO_AUTH_AES_GCM;
-  i->trunc_size = 16;
-}
-
-static_always_inline int
-translate_crypto_algo (ipsec_crypto_alg_t crypto_algo,
-		       struct rte_crypto_sym_xform *cipher_xform)
-{
-  switch (crypto_algo)
-    {
-    case IPSEC_CRYPTO_ALG_NONE:
-      cipher_xform->cipher.algo = RTE_CRYPTO_CIPHER_NULL;
-      break;
-    case IPSEC_CRYPTO_ALG_AES_CBC_128:
-    case IPSEC_CRYPTO_ALG_AES_CBC_192:
-    case IPSEC_CRYPTO_ALG_AES_CBC_256:
-      cipher_xform->cipher.algo = RTE_CRYPTO_CIPHER_AES_CBC;
-      break;
-    case IPSEC_CRYPTO_ALG_AES_GCM_128:
-      cipher_xform->cipher.algo = RTE_CRYPTO_CIPHER_AES_GCM;
-      break;
-    default:
-      return -1;
-    }
-
-  cipher_xform->type = RTE_CRYPTO_SYM_XFORM_CIPHER;
-
-  return 0;
-}
-
-static_always_inline int
-translate_integ_algo (ipsec_integ_alg_t integ_alg,
-		      struct rte_crypto_sym_xform *auth_xform, int use_esn)
-{
-  switch (integ_alg)
-    {
-    case IPSEC_INTEG_ALG_NONE:
-      auth_xform->auth.algo = RTE_CRYPTO_AUTH_NULL;
-      auth_xform->auth.digest_length = 0;
-      break;
-    case IPSEC_INTEG_ALG_SHA1_96:
-      auth_xform->auth.algo = RTE_CRYPTO_AUTH_SHA1_HMAC;
-      auth_xform->auth.digest_length = 12;
-      break;
-    case IPSEC_INTEG_ALG_SHA_256_96:
-      auth_xform->auth.algo = RTE_CRYPTO_AUTH_SHA256_HMAC;
-      auth_xform->auth.digest_length = 12;
-      break;
-    case IPSEC_INTEG_ALG_SHA_256_128:
-      auth_xform->auth.algo = RTE_CRYPTO_AUTH_SHA256_HMAC;
-      auth_xform->auth.digest_length = 16;
-      break;
-    case IPSEC_INTEG_ALG_SHA_384_192:
-      auth_xform->auth.algo = RTE_CRYPTO_AUTH_SHA384_HMAC;
-      auth_xform->auth.digest_length = 24;
-      break;
-    case IPSEC_INTEG_ALG_SHA_512_256:
-      auth_xform->auth.algo = RTE_CRYPTO_AUTH_SHA512_HMAC;
-      auth_xform->auth.digest_length = 32;
-      break;
-    case IPSEC_INTEG_ALG_AES_GCM_128:
-      auth_xform->auth.algo = RTE_CRYPTO_AUTH_AES_GCM;
-      auth_xform->auth.digest_length = 16;
-      auth_xform->auth.add_auth_data_length = use_esn ? 12 : 8;
-      break;
-    default:
-      return -1;
-    }
-
-  auth_xform->type = RTE_CRYPTO_SYM_XFORM_AUTH;
-
-  return 0;
-}
-
-static_always_inline int
-create_sym_sess (ipsec_sa_t * sa, crypto_sa_session_t * sa_sess,
-		 u8 is_outbound)
-{
-  u32 cpu_index = os_get_cpu_number ();
-  dpdk_crypto_main_t *dcm = &dpdk_crypto_main;
-  crypto_worker_main_t *cwm = &dcm->workers_main[cpu_index];
-  struct rte_crypto_sym_xform cipher_xform = { 0 };
-  struct rte_crypto_sym_xform auth_xform = { 0 };
-  struct rte_crypto_sym_xform *xfs;
-  uword key = 0, *data;
-  crypto_worker_qp_key_t *p_key = (crypto_worker_qp_key_t *) & key;
-
-  if (sa->crypto_alg == IPSEC_CRYPTO_ALG_AES_GCM_128)
-    {
-      sa->crypto_key_len -= 4;
-      clib_memcpy (&sa->salt, &sa->crypto_key[sa->crypto_key_len], 4);
-    }
-  else
-    {
-      u32 seed = (u32) clib_cpu_time_now ();
-      sa->salt = random_u32 (&seed);
-    }
-
-  cipher_xform.type = RTE_CRYPTO_SYM_XFORM_CIPHER;
-  cipher_xform.cipher.key.data = sa->crypto_key;
-  cipher_xform.cipher.key.length = sa->crypto_key_len;
-
-  auth_xform.type = RTE_CRYPTO_SYM_XFORM_AUTH;
-  auth_xform.auth.key.data = sa->integ_key;
-  auth_xform.auth.key.length = sa->integ_key_len;
-
-  if (translate_crypto_algo (sa->crypto_alg, &cipher_xform) < 0)
-    return -1;
-  p_key->cipher_algo = cipher_xform.cipher.algo;
-
-  if (translate_integ_algo (sa->integ_alg, &auth_xform, sa->use_esn) < 0)
-    return -1;
-  p_key->auth_algo = auth_xform.auth.algo;
-
-  if (is_outbound)
-    {
-      cipher_xform.cipher.op = RTE_CRYPTO_CIPHER_OP_ENCRYPT;
-      auth_xform.auth.op = RTE_CRYPTO_AUTH_OP_GENERATE;
-      cipher_xform.next = &auth_xform;
-      xfs = &cipher_xform;
-    }
-  else
-    {
-      cipher_xform.cipher.op = RTE_CRYPTO_CIPHER_OP_DECRYPT;
-      auth_xform.auth.op = RTE_CRYPTO_AUTH_OP_VERIFY;
-      auth_xform.next = &cipher_xform;
-      xfs = &auth_xform;
-    }
-
-  p_key->is_outbound = is_outbound;
-
-  data = hash_get (cwm->algo_qp_map, key);
-  if (!data)
-    return -1;
-
-  sa_sess->sess =
-    rte_cryptodev_sym_session_create (cwm->qp_data[*data].dev_id, xfs);
-
-  if (!sa_sess->sess)
-    return -1;
-
-  sa_sess->qp_index = (u8) * data;
-
-  return 0;
-}
-
-#endif /* __DPDK_ESP_H__ */
-
-/*
- * fd.io coding-style-patch-verification: ON
- *
- * Local Variables:
- * eval: (c-set-style "gnu")
- * End:
- */
diff --git a/src/vnet/devices/dpdk/ipsec/esp_decrypt.c b/src/vnet/devices/dpdk/ipsec/esp_decrypt.c
deleted file mode 100644
index 76007609..00000000
--- a/src/vnet/devices/dpdk/ipsec/esp_decrypt.c
+++ /dev/null
@@ -1,594 +0,0 @@
-/*
- * esp_decrypt.c : IPSec ESP Decrypt node using DPDK Cryptodev
- *
- * Copyright (c) 2016 Intel and/or its affiliates.
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at:
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#include 
-#include 
-#include 
-
-#include 
-#include 
-#include 
-#include 
-#include 
-
-#define foreach_esp_decrypt_next	       \
-_(DROP, "error-drop")			       \
-_(IP4_INPUT, "ip4-input")		       \
-_(IP6_INPUT, "ip6-input")
-
-#define _(v, s) ESP_DECRYPT_NEXT_##v,
-typedef enum {
-  foreach_esp_decrypt_next
-#undef _
-  ESP_DECRYPT_N_NEXT,
-} esp_decrypt_next_t;
-
-#define foreach_esp_decrypt_error		 \
- _(RX_PKTS, "ESP pkts received")		 \
- _(DECRYPTION_FAILED, "ESP decryption failed")   \
- _(REPLAY, "SA replayed packet")	         \
- _(NOT_IP, "Not IP packet (dropped)")	         \
- _(ENQ_FAIL, "Enqueue failed (buffer full)")     \
- _(NO_CRYPTODEV, "Cryptodev not configured")     \
- _(BAD_LEN, "Invalid ciphertext length")         \
- _(UNSUPPORTED, "Cipher/Auth not supported")
-
-
-typedef enum {
-#define _(sym,str) ESP_DECRYPT_ERROR_##sym,
-  foreach_esp_decrypt_error
-#undef _
-  ESP_DECRYPT_N_ERROR,
-} esp_decrypt_error_t;
-
-static char * esp_decrypt_error_strings[] = {
-#define _(sym,string) string,
-  foreach_esp_decrypt_error
-#undef _
-};
-
-vlib_node_registration_t dpdk_esp_decrypt_node;
-
-typedef struct {
-  ipsec_crypto_alg_t crypto_alg;
-  ipsec_integ_alg_t integ_alg;
-} esp_decrypt_trace_t;
-
-/* packet trace format function */
-static u8 * format_esp_decrypt_trace (u8 * s, va_list * args)
-{
-  CLIB_UNUSED (vlib_main_t * vm) = va_arg (*args, vlib_main_t *);
-  CLIB_UNUSED (vlib_node_t * node) = va_arg (*args, vlib_node_t *);
-  esp_decrypt_trace_t * t = va_arg (*args, esp_decrypt_trace_t *);
-
-  s = format (s, "esp: crypto %U integrity %U",
-	      format_ipsec_crypto_alg, t->crypto_alg,
-	      format_ipsec_integ_alg, t->integ_alg);
-  return s;
-}
-
-static uword
-dpdk_esp_decrypt_node_fn (vlib_main_t * vm,
-	     vlib_node_runtime_t * node,
-	     vlib_frame_t * from_frame)
-{
-  u32 n_left_from, *from, *to_next, next_index;
-  ipsec_main_t *im = &ipsec_main;
-  u32 cpu_index = os_get_cpu_number();
-  dpdk_crypto_main_t * dcm = &dpdk_crypto_main;
-  dpdk_esp_main_t * em = &dpdk_esp_main;
-  u32 i;
-
-  from = vlib_frame_vector_args (from_frame);
-  n_left_from = from_frame->n_vectors;
-
-  if (PREDICT_FALSE(!dcm->workers_main))
-    {
-      vlib_node_increment_counter (vm, dpdk_esp_decrypt_node.index,
-	      ESP_DECRYPT_ERROR_NO_CRYPTODEV, n_left_from);
-      vlib_buffer_free(vm, from, n_left_from);
-      return n_left_from;
-    }
-
-  crypto_worker_main_t *cwm = vec_elt_at_index(dcm->workers_main, cpu_index);
-  u32 n_qps = vec_len(cwm->qp_data);
-  struct rte_crypto_op ** cops_to_enq[n_qps];
-  u32 n_cop_qp[n_qps], * bi_to_enq[n_qps];
-
-  for (i = 0; i < n_qps; i++)
-    {
-      bi_to_enq[i] = cwm->qp_data[i].bi;
-      cops_to_enq[i] = cwm->qp_data[i].cops;
-    }
-
-  memset(n_cop_qp, 0, n_qps * sizeof(u32));
-
-  crypto_alloc_cops();
-
-  next_index = ESP_DECRYPT_NEXT_DROP;
-
-  while (n_left_from > 0)
-    {
-      u32 n_left_to_next;
-
-      vlib_get_next_frame (vm, node, next_index, to_next, n_left_to_next);
-
-      while (n_left_from > 0 && n_left_to_next > 0)
-	{
-	  u32 bi0, sa_index0 = ~0, seq, icv_size, iv_size;
-	  vlib_buffer_t * b0;
-	  esp_header_t * esp0;
-	  ipsec_sa_t * sa0;
-	  struct rte_mbuf * mb0 = 0;
-	  const int BLOCK_SIZE = 16;
-	  crypto_sa_session_t * sa_sess;
-	  void * sess;
-	  u16 qp_index;
-	  struct rte_crypto_op * cop = 0;
-
-	  bi0 = from[0];
-	  from += 1;
-	  n_left_from -= 1;
-
-	  b0 = vlib_get_buffer (vm, bi0);
-	  esp0 = vlib_buffer_get_current (b0);
-
-	  sa_index0 = vnet_buffer(b0)->ipsec.sad_index;
-	  sa0 = pool_elt_at_index (im->sad, sa_index0);
-
-	  seq = clib_host_to_net_u32(esp0->seq);
-
-	  /* anti-replay check */
-	  if (sa0->use_anti_replay)
-	    {
-	      int rv = 0;
-
-	      if (PREDICT_TRUE(sa0->use_esn))
-		rv = esp_replay_check_esn(sa0, seq);
-	      else
-		rv = esp_replay_check(sa0, seq);
-
-	      if (PREDICT_FALSE(rv))
-		{
-		  clib_warning ("anti-replay SPI %u seq %u", sa0->spi, seq);
-		  vlib_node_increment_counter (vm, dpdk_esp_decrypt_node.index,
-					       ESP_DECRYPT_ERROR_REPLAY, 1);
-		  to_next[0] = bi0;
-		  to_next += 1;
-		  n_left_to_next -= 1;
-		  goto trace;
-		}
-	    }
-
-	  sa0->total_data_size += b0->current_length;
-
-	  if (PREDICT_FALSE(sa0->integ_alg == IPSEC_INTEG_ALG_NONE) ||
-		  PREDICT_FALSE(sa0->crypto_alg == IPSEC_CRYPTO_ALG_NONE))
-	    {
-	      clib_warning ("SPI %u : only cipher + auth supported", sa0->spi);
-	      vlib_node_increment_counter (vm, dpdk_esp_decrypt_node.index,
-					   ESP_DECRYPT_ERROR_UNSUPPORTED, 1);
-	      to_next[0] = bi0;
-	      to_next += 1;
-	      n_left_to_next -= 1;
-	      goto trace;
-	    }
-
-	  sa_sess = pool_elt_at_index(cwm->sa_sess_d[0], sa_index0);
-
-	  if (PREDICT_FALSE(!sa_sess->sess))
-	    {
-	      int ret = create_sym_sess(sa0, sa_sess, 0);
-
-	      if (PREDICT_FALSE (ret))
-		{
-		  to_next[0] = bi0;
-		  to_next += 1;
-		  n_left_to_next -= 1;
-		  goto trace;
-		}
-	    }
-
-	  sess = sa_sess->sess;
-	  qp_index = sa_sess->qp_index;
-
-	  ASSERT (vec_len (vec_elt (cwm->qp_data, qp_index).free_cops) > 0);
-	  cop = vec_pop (vec_elt (cwm->qp_data, qp_index).free_cops);
-	  ASSERT (cop->status == RTE_CRYPTO_OP_STATUS_NOT_PROCESSED);
-
-	  cops_to_enq[qp_index][0] = cop;
-	  cops_to_enq[qp_index] += 1;
-	  n_cop_qp[qp_index] += 1;
-	  bi_to_enq[qp_index][0] = bi0;
-	  bi_to_enq[qp_index] += 1;
-
-	  rte_crypto_op_attach_sym_session(cop, sess);
-
-	  icv_size = em->esp_integ_algs[sa0->integ_alg].trunc_size;
-	  iv_size = em->esp_crypto_algs[sa0->crypto_alg].iv_len;
-
-	  /* Convert vlib buffer to mbuf */
-	  mb0 = rte_mbuf_from_vlib_buffer(b0);
-	  mb0->data_len = b0->current_length;
-	  mb0->pkt_len = b0->current_length;
-	  mb0->data_off = RTE_PKTMBUF_HEADROOM + b0->current_data;
-
-	  /* Outer IP header has already been stripped */
-	  u16 payload_len = rte_pktmbuf_pkt_len(mb0) - sizeof (esp_header_t) -
-	      iv_size - icv_size;
-
-	  if ((payload_len & (BLOCK_SIZE - 1)) || (payload_len <= 0))
-	    {
-	      clib_warning ("payload %u not multiple of %d\n",
-			    payload_len, BLOCK_SIZE);
-	      vlib_node_increment_counter (vm, dpdk_esp_decrypt_node.index,
-					   ESP_DECRYPT_ERROR_BAD_LEN, 1);
-	      vec_add (vec_elt (cwm->qp_data, qp_index).free_cops, &cop, 1);
-	      bi_to_enq[qp_index] -= 1;
-	      cops_to_enq[qp_index] -= 1;
-	      n_cop_qp[qp_index] -= 1;
-	      to_next[0] = bi0;
-	      to_next += 1;
-	      n_left_to_next -= 1;
-	      goto trace;
-	    }
-
-	  struct rte_crypto_sym_op *sym_cop = (struct rte_crypto_sym_op *)(cop + 1);
-
-	  sym_cop->m_src = mb0;
-	  sym_cop->cipher.data.offset = sizeof (esp_header_t) + iv_size;
-	  sym_cop->cipher.data.length = payload_len;
-
-          u8 *iv = rte_pktmbuf_mtod_offset(mb0, void*, sizeof (esp_header_t));
-          dpdk_cop_priv_t * priv = (dpdk_cop_priv_t *)(sym_cop + 1);
-
-          if (sa0->crypto_alg == IPSEC_CRYPTO_ALG_AES_GCM_128)
-            {
-              dpdk_gcm_cnt_blk *icb = &priv->cb;
-              icb->salt = sa0->salt;
-              clib_memcpy(icb->iv, iv, 8);
-              icb->cnt = clib_host_to_net_u32(1);
-              sym_cop->cipher.iv.data = (u8 *)icb;
-              sym_cop->cipher.iv.phys_addr = cop->phys_addr +
-		(uintptr_t)icb - (uintptr_t)cop;
-              sym_cop->cipher.iv.length = 16;
-
-              u8 *aad = priv->aad;
-              clib_memcpy(aad, iv - sizeof(esp_header_t), 8);
-              sym_cop->auth.aad.data = aad;
-              sym_cop->auth.aad.phys_addr = cop->phys_addr +
-                  (uintptr_t)aad - (uintptr_t)cop;
-              if (sa0->use_esn)
-                {
-                  *((u32*)&aad[8]) = sa0->seq_hi;
-                  sym_cop->auth.aad.length = 12;
-                }
-              else
-                {
-                  sym_cop->auth.aad.length = 8;
-                }
-
-              sym_cop->auth.digest.data = rte_pktmbuf_mtod_offset(mb0, void*,
-                       rte_pktmbuf_pkt_len(mb0) - icv_size);
-              sym_cop->auth.digest.phys_addr = rte_pktmbuf_mtophys_offset(mb0,
-                       rte_pktmbuf_pkt_len(mb0) - icv_size);
-              sym_cop->auth.digest.length = icv_size;
-
-            }
-          else
-            {
-              sym_cop->cipher.iv.data = rte_pktmbuf_mtod_offset(mb0, void*,
-                       sizeof (esp_header_t));
-              sym_cop->cipher.iv.phys_addr = rte_pktmbuf_mtophys_offset(mb0,
-                       sizeof (esp_header_t));
-              sym_cop->cipher.iv.length = iv_size;
-
-              if (sa0->use_esn)
-                {
-                  dpdk_cop_priv_t* priv = (dpdk_cop_priv_t*) (sym_cop + 1);
-                  u8* payload_end = rte_pktmbuf_mtod_offset(
-                      mb0, u8*, sizeof(esp_header_t) + iv_size + payload_len);
-
-                  clib_memcpy (priv->icv, payload_end, icv_size);
-                  *((u32*) payload_end) = sa0->seq_hi;
-                  sym_cop->auth.data.offset = 0;
-                  sym_cop->auth.data.length = sizeof(esp_header_t) + iv_size
-                      + payload_len + sizeof(sa0->seq_hi);
-                  sym_cop->auth.digest.data = priv->icv;
-                  sym_cop->auth.digest.phys_addr = cop->phys_addr
-                      + (uintptr_t) priv->icv - (uintptr_t) cop;
-                  sym_cop->auth.digest.length = icv_size;
-                }
-              else
-                {
-                  sym_cop->auth.data.offset = 0;
-                  sym_cop->auth.data.length = sizeof(esp_header_t) +
-                           iv_size + payload_len;
-
-                  sym_cop->auth.digest.data = rte_pktmbuf_mtod_offset(mb0, void*,
-                           rte_pktmbuf_pkt_len(mb0) - icv_size);
-                  sym_cop->auth.digest.phys_addr = rte_pktmbuf_mtophys_offset(mb0,
-                           rte_pktmbuf_pkt_len(mb0) - icv_size);
-                  sym_cop->auth.digest.length = icv_size;
-                }
-            }
-
-trace:
-	  if (PREDICT_FALSE(b0->flags & VLIB_BUFFER_IS_TRACED))
-	    {
-	      esp_decrypt_trace_t *tr = vlib_add_trace (vm, node, b0, sizeof (*tr));
-	      tr->crypto_alg = sa0->crypto_alg;
-	      tr->integ_alg = sa0->integ_alg;
-	    }
-	}
-      vlib_put_next_frame (vm, node, next_index, n_left_to_next);
-    }
-  vlib_node_increment_counter (vm, dpdk_esp_decrypt_node.index,
-			       ESP_DECRYPT_ERROR_RX_PKTS,
-			       from_frame->n_vectors);
-  crypto_qp_data_t *qpd;
-  /* *INDENT-OFF* */
-  vec_foreach_index (i, cwm->qp_data)
-    {
-      u32 enq;
-
-      qpd = vec_elt_at_index(cwm->qp_data, i);
-      enq = rte_cryptodev_enqueue_burst(qpd->dev_id, qpd->qp_id,
-					qpd->cops, n_cop_qp[i]);
-      qpd->inflights += enq;
-
-      if (PREDICT_FALSE(enq < n_cop_qp[i]))
-	{
-	  crypto_free_cop (qpd, &qpd->cops[enq], n_cop_qp[i] - enq);
-	  vlib_buffer_free (vm, &qpd->bi[enq], n_cop_qp[i] - enq);
-
-	  vlib_node_increment_counter (vm, dpdk_esp_decrypt_node.index,
-				       ESP_DECRYPT_ERROR_ENQ_FAIL,
-				       n_cop_qp[i] - enq);
-	}
-    }
-  /* *INDENT-ON* */
-
-  return from_frame->n_vectors;
-}
-
-VLIB_REGISTER_NODE (dpdk_esp_decrypt_node) = {
-  .function = dpdk_esp_decrypt_node_fn,
-  .name = "dpdk-esp-decrypt",
-  .vector_size = sizeof (u32),
-  .format_trace = format_esp_decrypt_trace,
-  .type = VLIB_NODE_TYPE_INTERNAL,
-
-  .n_errors = ARRAY_LEN(esp_decrypt_error_strings),
-  .error_strings = esp_decrypt_error_strings,
-
-  .n_next_nodes = ESP_DECRYPT_N_NEXT,
-  .next_nodes = {
-#define _(s,n) [ESP_DECRYPT_NEXT_##s] = n,
-    foreach_esp_decrypt_next
-#undef _
-  },
-};
-
-VLIB_NODE_FUNCTION_MULTIARCH (dpdk_esp_decrypt_node, dpdk_esp_decrypt_node_fn)
-
-/*
- * Decrypt Post Node
- */
-
-#define foreach_esp_decrypt_post_error	      \
- _(PKTS, "ESP post pkts")
-
-typedef enum {
-#define _(sym,str) ESP_DECRYPT_POST_ERROR_##sym,
-  foreach_esp_decrypt_post_error
-#undef _
-  ESP_DECRYPT_POST_N_ERROR,
-} esp_decrypt_post_error_t;
-
-static char * esp_decrypt_post_error_strings[] = {
-#define _(sym,string) string,
-  foreach_esp_decrypt_post_error
-#undef _
-};
-
-vlib_node_registration_t dpdk_esp_decrypt_post_node;
-
-static u8 * format_esp_decrypt_post_trace (u8 * s, va_list * args)
-{
-  return s;
-}
-
-static uword
-dpdk_esp_decrypt_post_node_fn (vlib_main_t * vm,
-	     vlib_node_runtime_t * node,
-	     vlib_frame_t * from_frame)
-{
-  u32 n_left_from, *from, *to_next = 0, next_index;
-  ipsec_sa_t * sa0;
-  u32 sa_index0 = ~0;
-  ipsec_main_t *im = &ipsec_main;
-  dpdk_esp_main_t *em = &dpdk_esp_main;
-
-  from = vlib_frame_vector_args (from_frame);
-  n_left_from = from_frame->n_vectors;
-
-  next_index = node->cached_next_index;
-
-  while (n_left_from > 0)
-    {
-      u32 n_left_to_next;
-
-      vlib_get_next_frame (vm, node, next_index, to_next, n_left_to_next);
-
-      while (n_left_from > 0 && n_left_to_next > 0)
-	{
-	  esp_footer_t * f0;
-	  u32 bi0, next0, icv_size, iv_size;
-	  vlib_buffer_t * b0 = 0;
-	  ip4_header_t *ih4 = 0, *oh4 = 0;
-	  ip6_header_t *ih6 = 0, *oh6 = 0;
-	  u8 tunnel_mode = 1;
-	  u8 transport_ip6 = 0;
-
-	  next0 = ESP_DECRYPT_NEXT_DROP;
-
-	  bi0 = from[0];
-	  from += 1;
-	  n_left_from -= 1;
-	  n_left_to_next -= 1;
-
-	  b0 = vlib_get_buffer (vm, bi0);
-
-	  sa_index0 = vnet_buffer(b0)->ipsec.sad_index;
-	  sa0 = pool_elt_at_index (im->sad, sa_index0);
-
-	  to_next[0] = bi0;
-	  to_next += 1;
-
-	  icv_size = em->esp_integ_algs[sa0->integ_alg].trunc_size;
-	  iv_size = em->esp_crypto_algs[sa0->crypto_alg].iv_len;
-
-	  if (sa0->use_anti_replay)
-	    {
-	      esp_header_t * esp0 = vlib_buffer_get_current (b0);
-	      u32 seq;
-	      seq = clib_host_to_net_u32(esp0->seq);
-	      if (PREDICT_TRUE(sa0->use_esn))
-		esp_replay_advance_esn(sa0, seq);
-	      else
-		esp_replay_advance(sa0, seq);
-	    }
-
-	  ih4 = (ip4_header_t *) (b0->data + sizeof(ethernet_header_t));
-	  vlib_buffer_advance (b0, sizeof (esp_header_t) + iv_size);
-
-	  b0->current_length -= (icv_size + 2);
-	  b0->flags = VLIB_BUFFER_TOTAL_LENGTH_VALID;
-	  f0 = (esp_footer_t *) ((u8 *) vlib_buffer_get_current (b0) +
-				 b0->current_length);
-	  b0->current_length -= f0->pad_length;
-
-	  /* transport mode */
-	  if (PREDICT_FALSE(!sa0->is_tunnel && !sa0->is_tunnel_ip6))
-	    {
-	      tunnel_mode = 0;
-
-	      if (PREDICT_TRUE((ih4->ip_version_and_header_length & 0xF0) != 0x40))
-		{
-		  if (PREDICT_TRUE((ih4->ip_version_and_header_length & 0xF0) == 0x60))
-		    transport_ip6 = 1;
-		  else
-		    {
-		      clib_warning("next header: 0x%x", f0->next_header);
-		      vlib_node_increment_counter (vm, dpdk_esp_decrypt_node.index,
-						   ESP_DECRYPT_ERROR_NOT_IP, 1);
-		      goto trace;
-		    }
-		}
-	    }
-
-	  if (PREDICT_TRUE (tunnel_mode))
-	    {
-	      if (PREDICT_TRUE(f0->next_header == IP_PROTOCOL_IP_IN_IP))
-		next0 = ESP_DECRYPT_NEXT_IP4_INPUT;
-	      else if (f0->next_header == IP_PROTOCOL_IPV6)
-		next0 = ESP_DECRYPT_NEXT_IP6_INPUT;
-	      else
-		{
-		  clib_warning("next header: 0x%x", f0->next_header);
-		  vlib_node_increment_counter (vm, dpdk_esp_decrypt_node.index,
-					       ESP_DECRYPT_ERROR_DECRYPTION_FAILED,
-					       1);
-		  goto trace;
-		}
-	    }
-	  /* transport mode */
-	  else
-	    {
-	      if (PREDICT_FALSE(transport_ip6))
-		{
-		  ih6 = (ip6_header_t *) (b0->data + sizeof(ethernet_header_t));
-		  vlib_buffer_advance (b0, -sizeof(ip6_header_t));
-		  oh6 = vlib_buffer_get_current (b0);
-		  memmove(oh6, ih6, sizeof(ip6_header_t));
-
-		  next0 = ESP_DECRYPT_NEXT_IP6_INPUT;
-		  oh6->protocol = f0->next_header;
-		  oh6->payload_length =
-		      clib_host_to_net_u16 (
-			  vlib_buffer_length_in_chain(vm, b0) -
-			  sizeof (ip6_header_t));
-		}
-	      else
-		{
-		  vlib_buffer_advance (b0, -sizeof(ip4_header_t));
-		  oh4 = vlib_buffer_get_current (b0);
-		  memmove(oh4, ih4, sizeof(ip4_header_t));
-
-		  next0 = ESP_DECRYPT_NEXT_IP4_INPUT;
-		  oh4->ip_version_and_header_length = 0x45;
-		  oh4->fragment_id = 0;
-		  oh4->flags_and_fragment_offset = 0;
-		  oh4->protocol = f0->next_header;
-		  oh4->length = clib_host_to_net_u16 (
-		      vlib_buffer_length_in_chain (vm, b0));
-		  oh4->checksum = ip4_header_checksum (oh4);
-		}
-	    }
-
-	  vnet_buffer (b0)->sw_if_index[VLIB_TX] = (u32)~0;
-
-trace:
-	  if (PREDICT_FALSE(b0->flags & VLIB_BUFFER_IS_TRACED))
-	    {
-	      esp_decrypt_trace_t *tr = vlib_add_trace (vm, node, b0, sizeof (*tr));
-	      tr->crypto_alg = sa0->crypto_alg;
-	      tr->integ_alg = sa0->integ_alg;
-	    }
-
-	  vlib_validate_buffer_enqueue_x1 (vm, node, next_index,
-					   to_next, n_left_to_next, bi0, next0);
-	}
-      vlib_put_next_frame (vm, node, next_index, n_left_to_next);
-    }
-  vlib_node_increment_counter (vm, dpdk_esp_decrypt_post_node.index,
-			       ESP_DECRYPT_POST_ERROR_PKTS,
-			       from_frame->n_vectors);
-
-  return from_frame->n_vectors;
-}
-
-VLIB_REGISTER_NODE (dpdk_esp_decrypt_post_node) = {
-  .function = dpdk_esp_decrypt_post_node_fn,
-  .name = "dpdk-esp-decrypt-post",
-  .vector_size = sizeof (u32),
-  .format_trace = format_esp_decrypt_post_trace,
-  .type = VLIB_NODE_TYPE_INTERNAL,
-
-  .n_errors = ARRAY_LEN(esp_decrypt_post_error_strings),
-  .error_strings = esp_decrypt_post_error_strings,
-
-  .n_next_nodes = ESP_DECRYPT_N_NEXT,
-  .next_nodes = {
-#define _(s,n) [ESP_DECRYPT_NEXT_##s] = n,
-    foreach_esp_decrypt_next
-#undef _
-  },
-};
-
-VLIB_NODE_FUNCTION_MULTIARCH (dpdk_esp_decrypt_post_node, dpdk_esp_decrypt_post_node_fn)
diff --git a/src/vnet/devices/dpdk/ipsec/esp_encrypt.c b/src/vnet/devices/dpdk/ipsec/esp_encrypt.c
deleted file mode 100644
index 6eb1afc9..00000000
--- a/src/vnet/devices/dpdk/ipsec/esp_encrypt.c
+++ /dev/null
@@ -1,609 +0,0 @@
-/*
- * esp_encrypt.c : IPSec ESP encrypt node using DPDK Cryptodev
- *
- * Copyright (c) 2016 Intel and/or its affiliates.
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at:
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#include 
-#include 
-#include 
-
-#include 
-#include 
-#include 
-#include 
-#include 
-
-#define foreach_esp_encrypt_next                   \
-_(DROP, "error-drop")                              \
-_(IP4_LOOKUP, "ip4-lookup")                        \
-_(IP6_LOOKUP, "ip6-lookup")                        \
-_(INTERFACE_OUTPUT, "interface-output")
-
-#define _(v, s) ESP_ENCRYPT_NEXT_##v,
-typedef enum
-{
-  foreach_esp_encrypt_next
-#undef _
-    ESP_ENCRYPT_N_NEXT,
-} esp_encrypt_next_t;
-
-#define foreach_esp_encrypt_error                   \
- _(RX_PKTS, "ESP pkts received")                    \
- _(SEQ_CYCLED, "sequence number cycled")            \
- _(ENQ_FAIL, "Enqueue failed (buffer full)")        \
- _(NO_CRYPTODEV, "Cryptodev not configured")        \
- _(UNSUPPORTED, "Cipher/Auth not supported")
-
-
-typedef enum
-{
-#define _(sym,str) ESP_ENCRYPT_ERROR_##sym,
-  foreach_esp_encrypt_error
-#undef _
-    ESP_ENCRYPT_N_ERROR,
-} esp_encrypt_error_t;
-
-static char *esp_encrypt_error_strings[] = {
-#define _(sym,string) string,
-  foreach_esp_encrypt_error
-#undef _
-};
-
-vlib_node_registration_t dpdk_esp_encrypt_node;
-
-typedef struct
-{
-  u32 spi;
-  u32 seq;
-  ipsec_crypto_alg_t crypto_alg;
-  ipsec_integ_alg_t integ_alg;
-} esp_encrypt_trace_t;
-
-/* packet trace format function */
-static u8 *
-format_esp_encrypt_trace (u8 * s, va_list * args)
-{
-  CLIB_UNUSED (vlib_main_t * vm) = va_arg (*args, vlib_main_t *);
-  CLIB_UNUSED (vlib_node_t * node) = va_arg (*args, vlib_node_t *);
-  esp_encrypt_trace_t *t = va_arg (*args, esp_encrypt_trace_t *);
-
-  s = format (s, "esp: spi %u seq %u crypto %U integrity %U",
-	      t->spi, t->seq,
-	      format_ipsec_crypto_alg, t->crypto_alg,
-	      format_ipsec_integ_alg, t->integ_alg);
-  return s;
-}
-
-static uword
-dpdk_esp_encrypt_node_fn (vlib_main_t * vm,
-			  vlib_node_runtime_t * node,
-			  vlib_frame_t * from_frame)
-{
-  u32 n_left_from, *from, *to_next, next_index;
-  ipsec_main_t *im = &ipsec_main;
-  u32 cpu_index = os_get_cpu_number ();
-  dpdk_crypto_main_t *dcm = &dpdk_crypto_main;
-  dpdk_esp_main_t *em = &dpdk_esp_main;
-  u32 i;
-
-  from = vlib_frame_vector_args (from_frame);
-  n_left_from = from_frame->n_vectors;
-
-  if (PREDICT_FALSE (!dcm->workers_main))
-    {
-      /* Likely there are not enough cryptodevs, so drop frame */
-      vlib_node_increment_counter (vm, dpdk_esp_encrypt_node.index,
-				   ESP_ENCRYPT_ERROR_NO_CRYPTODEV,
-				   n_left_from);
-      vlib_buffer_free (vm, from, n_left_from);
-      return n_left_from;
-    }
-
-  crypto_worker_main_t *cwm = vec_elt_at_index (dcm->workers_main, cpu_index);
-  u32 n_qps = vec_len (cwm->qp_data);
-  struct rte_crypto_op **cops_to_enq[n_qps];
-  u32 n_cop_qp[n_qps], *bi_to_enq[n_qps];
-
-  for (i = 0; i < n_qps; i++)
-    {
-      bi_to_enq[i] = cwm->qp_data[i].bi;
-      cops_to_enq[i] = cwm->qp_data[i].cops;
-    }
-
-  memset (n_cop_qp, 0, n_qps * sizeof (u32));
-
-  crypto_alloc_cops ();
-
-  next_index = ESP_ENCRYPT_NEXT_DROP;
-
-  while (n_left_from > 0)
-    {
-      u32 n_left_to_next;
-
-      vlib_get_next_frame (vm, node, next_index, to_next, n_left_to_next);
-
-      while (n_left_from > 0 && n_left_to_next > 0)
-	{
-	  u32 bi0, next0;
-	  vlib_buffer_t *b0 = 0;
-	  u32 sa_index0;
-	  ipsec_sa_t *sa0;
-	  ip4_and_esp_header_t *ih0, *oh0 = 0;
-	  ip6_and_esp_header_t *ih6_0, *oh6_0 = 0;
-	  struct rte_mbuf *mb0 = 0;
-	  esp_footer_t *f0;
-	  u8 is_ipv6;
-	  u8 ip_hdr_size;
-	  u8 next_hdr_type;
-	  u8 transport_mode = 0;
-	  const int BLOCK_SIZE = 16;
-	  u32 iv_size;
-	  u16 orig_sz;
-	  crypto_sa_session_t *sa_sess;
-	  void *sess;
-	  struct rte_crypto_op *cop = 0;
-	  u16 qp_index;
-
-	  bi0 = from[0];
-	  from += 1;
-	  n_left_from -= 1;
-
-	  b0 = vlib_get_buffer (vm, bi0);
-	  sa_index0 = vnet_buffer (b0)->ipsec.sad_index;
-	  sa0 = pool_elt_at_index (im->sad, sa_index0);
-
-	  if (PREDICT_FALSE (esp_seq_advance (sa0)))
-	    {
-	      clib_warning ("sequence number counter has cycled SPI %u",
-			    sa0->spi);
-	      vlib_node_increment_counter (vm, dpdk_esp_encrypt_node.index,
-					   ESP_ENCRYPT_ERROR_SEQ_CYCLED, 1);
-	      //TODO: rekey SA
-	      to_next[0] = bi0;
-	      to_next += 1;
-	      n_left_to_next -= 1;
-	      goto trace;
-	    }
-
-	  sa0->total_data_size += b0->current_length;
-
-	  sa_sess = pool_elt_at_index (cwm->sa_sess_d[1], sa_index0);
-	  if (PREDICT_FALSE (!sa_sess->sess))
-	    {
-	      int ret = create_sym_sess (sa0, sa_sess, 1);
-
-	      if (PREDICT_FALSE (ret))
-		{
-		  to_next[0] = bi0;
-		  to_next += 1;
-		  n_left_to_next -= 1;
-		  goto trace;
-		}
-	    }
-
-	  qp_index = sa_sess->qp_index;
-	  sess = sa_sess->sess;
-
-	  ASSERT (vec_len (vec_elt (cwm->qp_data, qp_index).free_cops) > 0);
-	  cop = vec_pop (vec_elt (cwm->qp_data, qp_index).free_cops);
-	  ASSERT (cop->status == RTE_CRYPTO_OP_STATUS_NOT_PROCESSED);
-
-	  cops_to_enq[qp_index][0] = cop;
-	  cops_to_enq[qp_index] += 1;
-	  n_cop_qp[qp_index] += 1;
-	  bi_to_enq[qp_index][0] = bi0;
-	  bi_to_enq[qp_index] += 1;
-
-	  ssize_t adv;
-	  iv_size = em->esp_crypto_algs[sa0->crypto_alg].iv_len;
-	  ih0 = vlib_buffer_get_current (b0);
-	  orig_sz = b0->current_length;
-	  is_ipv6 = (ih0->ip4.ip_version_and_header_length & 0xF0) == 0x60;
-	  /* is ipv6 */
-	  if (PREDICT_TRUE (sa0->is_tunnel))
-	    {
-	      if (PREDICT_TRUE (!is_ipv6))
-		adv = -sizeof (ip4_and_esp_header_t);
-	      else
-		adv = -sizeof (ip6_and_esp_header_t);
-	    }
-	  else
-	    {
-	      adv = -sizeof (esp_header_t);
-	      if (PREDICT_TRUE (!is_ipv6))
-		orig_sz -= sizeof (ip4_header_t);
-	      else
-		orig_sz -= sizeof (ip6_header_t);
-	    }
-
-	  /*transport mode save the eth header before it is overwritten */
-	  if (PREDICT_FALSE (!sa0->is_tunnel))
-	    {
-	      ethernet_header_t *ieh0 = (ethernet_header_t *)
-		((u8 *) vlib_buffer_get_current (b0) -
-		 sizeof (ethernet_header_t));
-	      ethernet_header_t *oeh0 =
-		(ethernet_header_t *) ((u8 *) ieh0 + (adv - iv_size));
-	      clib_memcpy (oeh0, ieh0, sizeof (ethernet_header_t));
-	    }
-
-	  vlib_buffer_advance (b0, adv - iv_size);
-
-	  /* XXX IP6/ip4 and IP4/IP6 not supported, only IP4/IP4 and IP6/IP6 */
-
-	  /* is ipv6 */
-	  if (PREDICT_FALSE (is_ipv6))
-	    {
-	      ih6_0 = (ip6_and_esp_header_t *) ih0;
-	      ip_hdr_size = sizeof (ip6_header_t);
-	      oh6_0 = vlib_buffer_get_current (b0);
-
-	      if (PREDICT_TRUE (sa0->is_tunnel))
-		{
-		  next_hdr_type = IP_PROTOCOL_IPV6;
-		  oh6_0->ip6.ip_version_traffic_class_and_flow_label =
-		    ih6_0->ip6.ip_version_traffic_class_and_flow_label;
-		}
-	      else
-		{
-		  next_hdr_type = ih6_0->ip6.protocol;
-		  memmove (oh6_0, ih6_0, sizeof (ip6_header_t));
-		}
-
-	      oh6_0->ip6.protocol = IP_PROTOCOL_IPSEC_ESP;
-	      oh6_0->ip6.hop_limit = 254;
-	      oh6_0->esp.spi = clib_net_to_host_u32 (sa0->spi);
-	      oh6_0->esp.seq = clib_net_to_host_u32 (sa0->seq);
-	    }
-	  else
-	    {
-	      ip_hdr_size = sizeof (ip4_header_t);
-	      oh0 = vlib_buffer_get_current (b0);
-
-	      if (PREDICT_TRUE (sa0->is_tunnel))
-		{
-		  next_hdr_type = IP_PROTOCOL_IP_IN_IP;
-		  oh0->ip4.tos = ih0->ip4.tos;
-		}
-	      else
-		{
-		  next_hdr_type = ih0->ip4.protocol;
-		  memmove (oh0, ih0, sizeof (ip4_header_t));
-		}
-
-	      oh0->ip4.ip_version_and_header_length = 0x45;
-	      oh0->ip4.fragment_id = 0;
-	      oh0->ip4.flags_and_fragment_offset = 0;
-	      oh0->ip4.ttl = 254;
-	      oh0->ip4.protocol = IP_PROTOCOL_IPSEC_ESP;
-	      oh0->esp.spi = clib_net_to_host_u32 (sa0->spi);
-	      oh0->esp.seq = clib_net_to_host_u32 (sa0->seq);
-	    }
-
-	  if (PREDICT_TRUE (sa0->is_tunnel && !sa0->is_tunnel_ip6))
-	    {
-	      oh0->ip4.src_address.as_u32 = sa0->tunnel_src_addr.ip4.as_u32;
-	      oh0->ip4.dst_address.as_u32 = sa0->tunnel_dst_addr.ip4.as_u32;
-
-	      /* in tunnel mode send it back to FIB */
-	      next0 = ESP_ENCRYPT_NEXT_IP4_LOOKUP;
-	      vnet_buffer (b0)->sw_if_index[VLIB_TX] = (u32) ~ 0;
-	    }
-	  else if (sa0->is_tunnel && sa0->is_tunnel_ip6)
-	    {
-	      oh6_0->ip6.src_address.as_u64[0] =
-		sa0->tunnel_src_addr.ip6.as_u64[0];
-	      oh6_0->ip6.src_address.as_u64[1] =
-		sa0->tunnel_src_addr.ip6.as_u64[1];
-	      oh6_0->ip6.dst_address.as_u64[0] =
-		sa0->tunnel_dst_addr.ip6.as_u64[0];
-	      oh6_0->ip6.dst_address.as_u64[1] =
-		sa0->tunnel_dst_addr.ip6.as_u64[1];
-
-	      /* in tunnel mode send it back to FIB */
-	      next0 = ESP_ENCRYPT_NEXT_IP6_LOOKUP;
-	      vnet_buffer (b0)->sw_if_index[VLIB_TX] = (u32) ~ 0;
-	    }
-	  else
-	    {
-	      next0 = ESP_ENCRYPT_NEXT_INTERFACE_OUTPUT;
-	      transport_mode = 1;
-	    }
-
-	  ASSERT (sa0->crypto_alg < IPSEC_CRYPTO_N_ALG);
-	  ASSERT (sa0->crypto_alg != IPSEC_CRYPTO_ALG_NONE);
-
-	  int blocks = 1 + (orig_sz + 1) / BLOCK_SIZE;
-
-	  /* pad packet in input buffer */
-	  u8 pad_bytes = BLOCK_SIZE * blocks - 2 - orig_sz;
-	  u8 i;
-	  u8 *padding = vlib_buffer_get_current (b0) + b0->current_length;
-
-	  for (i = 0; i < pad_bytes; ++i)
-	    padding[i] = i + 1;
-
-	  f0 = vlib_buffer_get_current (b0) + b0->current_length + pad_bytes;
-	  f0->pad_length = pad_bytes;
-	  f0->next_header = next_hdr_type;
-	  b0->current_length += pad_bytes + 2 +
-	    em->esp_integ_algs[sa0->integ_alg].trunc_size;
-
-	  vnet_buffer (b0)->sw_if_index[VLIB_RX] =
-	    vnet_buffer (b0)->sw_if_index[VLIB_RX];
-	  b0->flags |= VLIB_BUFFER_TOTAL_LENGTH_VALID;
-
-	  struct rte_crypto_sym_op *sym_cop;
-	  sym_cop = (struct rte_crypto_sym_op *) (cop + 1);
-
-	  dpdk_cop_priv_t *priv = (dpdk_cop_priv_t *) (sym_cop + 1);
-
-	  vnet_buffer (b0)->unused[0] = next0;
-
-	  mb0 = rte_mbuf_from_vlib_buffer (b0);
-	  mb0->data_len = b0->current_length;
-	  mb0->pkt_len = b0->current_length;
-	  mb0->data_off = RTE_PKTMBUF_HEADROOM + b0->current_data;
-
-	  rte_crypto_op_attach_sym_session (cop, sess);
-
-	  sym_cop->m_src = mb0;
-
-	  dpdk_gcm_cnt_blk *icb = &priv->cb;
-	  icb->salt = sa0->salt;
-	  icb->iv[0] = sa0->seq;
-	  icb->iv[1] = sa0->seq_hi;
-
-	  if (sa0->crypto_alg == IPSEC_CRYPTO_ALG_AES_GCM_128)
-	    {
-	      icb->cnt = clib_host_to_net_u32 (1);
-	      clib_memcpy (vlib_buffer_get_current (b0) + ip_hdr_size +
-			   sizeof (esp_header_t), icb->iv, 8);
-	      sym_cop->cipher.data.offset =
-		ip_hdr_size + sizeof (esp_header_t) + iv_size;
-	      sym_cop->cipher.data.length = BLOCK_SIZE * blocks;
-	      sym_cop->cipher.iv.length = 16;
-	    }
-	  else
-	    {
-	      sym_cop->cipher.data.offset =
-		ip_hdr_size + sizeof (esp_header_t);
-	      sym_cop->cipher.data.length = BLOCK_SIZE * blocks + iv_size;
-	      sym_cop->cipher.iv.length = iv_size;
-	    }
-
-	  sym_cop->cipher.iv.data = (u8 *) icb;
-	  sym_cop->cipher.iv.phys_addr = cop->phys_addr + (uintptr_t) icb
-	    - (uintptr_t) cop;
-
-
-	  ASSERT (sa0->integ_alg < IPSEC_INTEG_N_ALG);
-	  ASSERT (sa0->integ_alg != IPSEC_INTEG_ALG_NONE);
-
-	  if (PREDICT_FALSE (sa0->integ_alg == IPSEC_INTEG_ALG_AES_GCM_128))
-	    {
-	      u8 *aad = priv->aad;
-	      clib_memcpy (aad, vlib_buffer_get_current (b0) + ip_hdr_size,
-			   8);
-	      sym_cop->auth.aad.data = aad;
-	      sym_cop->auth.aad.phys_addr = cop->phys_addr +
-		(uintptr_t) aad - (uintptr_t) cop;
-
-	      if (PREDICT_FALSE (sa0->use_esn))
-		{
-		  *((u32 *) & aad[8]) = sa0->seq_hi;
-		  sym_cop->auth.aad.length = 12;
-		}
-	      else
-		{
-		  sym_cop->auth.aad.length = 8;
-		}
-	    }
-	  else
-	    {
-	      sym_cop->auth.data.offset = ip_hdr_size;
-	      sym_cop->auth.data.length = b0->current_length - ip_hdr_size
-		- em->esp_integ_algs[sa0->integ_alg].trunc_size;
-
-	      if (PREDICT_FALSE (sa0->use_esn))
-		{
-		  u8 *payload_end =
-		    vlib_buffer_get_current (b0) + b0->current_length;
-		  *((u32 *) payload_end) = sa0->seq_hi;
-		  sym_cop->auth.data.length += sizeof (sa0->seq_hi);
-		}
-	    }
-	  sym_cop->auth.digest.data = vlib_buffer_get_current (b0) +
-	    b0->current_length -
-	    em->esp_integ_algs[sa0->integ_alg].trunc_size;
-	  sym_cop->auth.digest.phys_addr = rte_pktmbuf_mtophys_offset (mb0,
-								       b0->current_length
-								       -
-								       em->esp_integ_algs
-								       [sa0->integ_alg].trunc_size);
-	  sym_cop->auth.digest.length =
-	    em->esp_integ_algs[sa0->integ_alg].trunc_size;
-
-
-	  if (PREDICT_FALSE (is_ipv6))
-	    {
-	      oh6_0->ip6.payload_length =
-		clib_host_to_net_u16 (vlib_buffer_length_in_chain (vm, b0) -
-				      sizeof (ip6_header_t));
-	    }
-	  else
-	    {
-	      oh0->ip4.length =
-		clib_host_to_net_u16 (vlib_buffer_length_in_chain (vm, b0));
-	      oh0->ip4.checksum = ip4_header_checksum (&oh0->ip4);
-	    }
-
-	  if (transport_mode)
-	    vlib_buffer_advance (b0, -sizeof (ethernet_header_t));
-
-	trace:
-	  if (PREDICT_FALSE (b0->flags & VLIB_BUFFER_IS_TRACED))
-	    {
-	      esp_encrypt_trace_t *tr =
-		vlib_add_trace (vm, node, b0, sizeof (*tr));
-	      tr->spi = sa0->spi;
-	      tr->seq = sa0->seq - 1;
-	      tr->crypto_alg = sa0->crypto_alg;
-	      tr->integ_alg = sa0->integ_alg;
-	    }
-	}
-      vlib_put_next_frame (vm, node, next_index, n_left_to_next);
-    }
-  vlib_node_increment_counter (vm, dpdk_esp_encrypt_node.index,
-			       ESP_ENCRYPT_ERROR_RX_PKTS,
-			       from_frame->n_vectors);
-  crypto_qp_data_t *qpd;
-  /* *INDENT-OFF* */
-  vec_foreach_index (i, cwm->qp_data)
-    {
-      u32 enq;
-
-      qpd = vec_elt_at_index(cwm->qp_data, i);
-      enq = rte_cryptodev_enqueue_burst(qpd->dev_id, qpd->qp_id,
-					qpd->cops, n_cop_qp[i]);
-      qpd->inflights += enq;
-
-      if (PREDICT_FALSE(enq < n_cop_qp[i]))
-	{
-	  crypto_free_cop (qpd, &qpd->cops[enq], n_cop_qp[i] - enq);
-	  vlib_buffer_free (vm, &qpd->bi[enq], n_cop_qp[i] - enq);
-
-          vlib_node_increment_counter (vm, dpdk_esp_encrypt_node.index,
-				       ESP_ENCRYPT_ERROR_ENQ_FAIL,
-				       n_cop_qp[i] - enq);
-        }
-    }
-  /* *INDENT-ON* */
-
-  return from_frame->n_vectors;
-}
-
-VLIB_REGISTER_NODE (dpdk_esp_encrypt_node) =
-{
-  .function = dpdk_esp_encrypt_node_fn,.name = "dpdk-esp-encrypt",.flags =
-    VLIB_NODE_FLAG_IS_OUTPUT,.vector_size = sizeof (u32),.format_trace =
-    format_esp_encrypt_trace,.n_errors =
-    ARRAY_LEN (esp_encrypt_error_strings),.error_strings =
-    esp_encrypt_error_strings,.n_next_nodes = 1,.next_nodes =
-  {
-  [ESP_ENCRYPT_NEXT_DROP] = "error-drop",}
-};
-
-VLIB_NODE_FUNCTION_MULTIARCH (dpdk_esp_encrypt_node, dpdk_esp_encrypt_node_fn)
-/*
- * ESP Encrypt Post Node
- */
-#define foreach_esp_encrypt_post_error              \
- _(PKTS, "ESP post pkts")
-     typedef enum
-     {
-#define _(sym,str) ESP_ENCRYPT_POST_ERROR_##sym,
-       foreach_esp_encrypt_post_error
-#undef _
-	 ESP_ENCRYPT_POST_N_ERROR,
-     } esp_encrypt_post_error_t;
-
-     static char *esp_encrypt_post_error_strings[] = {
-#define _(sym,string) string,
-       foreach_esp_encrypt_post_error
-#undef _
-     };
-
-vlib_node_registration_t dpdk_esp_encrypt_post_node;
-
-static u8 *
-format_esp_encrypt_post_trace (u8 * s, va_list * args)
-{
-  return s;
-}
-
-static uword
-dpdk_esp_encrypt_post_node_fn (vlib_main_t * vm,
-			       vlib_node_runtime_t * node,
-			       vlib_frame_t * from_frame)
-{
-  u32 n_left_from, *from, *to_next = 0, next_index;
-
-  from = vlib_frame_vector_args (from_frame);
-  n_left_from = from_frame->n_vectors;
-
-  next_index = node->cached_next_index;
-
-  while (n_left_from > 0)
-    {
-      u32 n_left_to_next;
-
-      vlib_get_next_frame (vm, node, next_index, to_next, n_left_to_next);
-
-      while (n_left_from > 0 && n_left_to_next > 0)
-	{
-	  u32 bi0, next0;
-	  vlib_buffer_t *b0 = 0;
-
-	  bi0 = from[0];
-	  from += 1;
-	  n_left_from -= 1;
-	  n_left_to_next -= 1;
-
-	  b0 = vlib_get_buffer (vm, bi0);
-
-	  to_next[0] = bi0;
-	  to_next += 1;
-
-	  next0 = vnet_buffer (b0)->unused[0];
-
-	  vlib_validate_buffer_enqueue_x1 (vm, node, next_index,
-					   to_next, n_left_to_next, bi0,
-					   next0);
-	}
-      vlib_put_next_frame (vm, node, next_index, n_left_to_next);
-    }
-
-  vlib_node_increment_counter (vm, dpdk_esp_encrypt_post_node.index,
-			       ESP_ENCRYPT_POST_ERROR_PKTS,
-			       from_frame->n_vectors);
-
-  return from_frame->n_vectors;
-}
-
-VLIB_REGISTER_NODE (dpdk_esp_encrypt_post_node) =
-{
-  .function = dpdk_esp_encrypt_post_node_fn,.name =
-    "dpdk-esp-encrypt-post",.vector_size = sizeof (u32),.format_trace =
-    format_esp_encrypt_post_trace,.type = VLIB_NODE_TYPE_INTERNAL,.n_errors =
-    ARRAY_LEN (esp_encrypt_post_error_strings),.error_strings =
-    esp_encrypt_post_error_strings,.n_next_nodes =
-    ESP_ENCRYPT_N_NEXT,.next_nodes =
-  {
-#define _(s,n) [ESP_ENCRYPT_NEXT_##s] = n,
-    foreach_esp_encrypt_next
-#undef _
-  }
-};
-
-VLIB_NODE_FUNCTION_MULTIARCH (dpdk_esp_encrypt_post_node,
-			      dpdk_esp_encrypt_post_node_fn)
-/*
- * fd.io coding-style-patch-verification: ON
- *
- * Local Variables:
- * eval: (c-set-style "gnu")
- * End:
- */
diff --git a/src/vnet/devices/dpdk/ipsec/ipsec.c b/src/vnet/devices/dpdk/ipsec/ipsec.c
deleted file mode 100644
index 05c17c99..00000000
--- a/src/vnet/devices/dpdk/ipsec/ipsec.c
+++ /dev/null
@@ -1,430 +0,0 @@
-/*
- * Copyright (c) 2016 Intel and/or its affiliates.
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at:
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-#include 
-#include 
-#include 
-#include 
-#include 
-
-#include 
-#include 
-#include 
-
-#define DPDK_CRYPTO_NB_SESS_OBJS  20000
-#define DPDK_CRYPTO_CACHE_SIZE	  512
-#define DPDK_CRYPTO_PRIV_SIZE	  128
-#define DPDK_CRYPTO_N_QUEUE_DESC  1024
-#define DPDK_CRYPTO_NB_COPS	  (1024 * 4)
-
-static int
-add_del_sa_sess (u32 sa_index, u8 is_add)
-{
-  dpdk_crypto_main_t *dcm = &dpdk_crypto_main;
-  crypto_worker_main_t *cwm;
-  u8 skip_master = vlib_num_workers () > 0;
-
-  /* *INDENT-OFF* */
-  vec_foreach (cwm, dcm->workers_main)
-    {
-      crypto_sa_session_t *sa_sess;
-      u8 is_outbound;
-
-      if (skip_master)
-	{
-	  skip_master = 0;
-	  continue;
-	}
-
-      for (is_outbound = 0; is_outbound < 2; is_outbound++)
-	{
-	  if (is_add)
-	    {
-	      pool_get (cwm->sa_sess_d[is_outbound], sa_sess);
-	    }
-	  else
-	    {
-	      u8 dev_id;
-
-	      sa_sess = pool_elt_at_index (cwm->sa_sess_d[is_outbound], sa_index);
-	      dev_id = cwm->qp_data[sa_sess->qp_index].dev_id;
-
-	      if (!sa_sess->sess)
-		continue;
-
-	      if (rte_cryptodev_sym_session_free(dev_id, sa_sess->sess))
-		{
-		  clib_warning("failed to free session");
-		  return -1;
-		}
-	      memset(sa_sess, 0, sizeof(sa_sess[0]));
-	    }
-	}
-    }
-  /* *INDENT-OFF* */
-
-  return 0;
-}
-
-static void
-update_qp_data (crypto_worker_main_t * cwm,
-		u8 cdev_id, u16 qp_id, u8 is_outbound, u16 * idx)
-{
-  crypto_qp_data_t *qpd;
-
-  /* *INDENT-OFF* */
-  vec_foreach_index (*idx, cwm->qp_data)
-    {
-      qpd = vec_elt_at_index(cwm->qp_data, *idx);
-
-      if (qpd->dev_id == cdev_id && qpd->qp_id == qp_id &&
-	  qpd->is_outbound == is_outbound)
-	  return;
-    }
-  /* *INDENT-ON* */
-
-  vec_add2 (cwm->qp_data, qpd, 1);
-
-  qpd->dev_id = cdev_id;
-  qpd->qp_id = qp_id;
-  qpd->is_outbound = is_outbound;
-}
-
-/*
- * return:
- * 	0: already exist
- * 	1: mapped
- */
-static int
-add_mapping (crypto_worker_main_t * cwm,
-	     u8 cdev_id, u16 qp, u8 is_outbound,
-	     const struct rte_cryptodev_capabilities *cipher_cap,
-	     const struct rte_cryptodev_capabilities *auth_cap)
-{
-  u16 qp_index;
-  uword key = 0, data, *ret;
-  crypto_worker_qp_key_t *p_key = (crypto_worker_qp_key_t *) & key;
-
-  p_key->cipher_algo = (u8) cipher_cap->sym.cipher.algo;
-  p_key->auth_algo = (u8) auth_cap->sym.auth.algo;
-  p_key->is_outbound = is_outbound;
-
-  ret = hash_get (cwm->algo_qp_map, key);
-  if (ret)
-    return 0;
-
-  update_qp_data (cwm, cdev_id, qp, is_outbound, &qp_index);
-
-  data = (uword) qp_index;
-  hash_set (cwm->algo_qp_map, key, data);
-
-  return 1;
-}
-
-/*
- * return:
- * 	0: already exist
- * 	1: mapped
- */
-static int
-add_cdev_mapping (crypto_worker_main_t * cwm,
-		  struct rte_cryptodev_info *dev_info, u8 cdev_id,
-		  u16 qp, u8 is_outbound)
-{
-  const struct rte_cryptodev_capabilities *i, *j;
-  u32 mapped = 0;
-
-  for (i = dev_info->capabilities; i->op != RTE_CRYPTO_OP_TYPE_UNDEFINED; i++)
-    {
-      if (i->sym.xform_type != RTE_CRYPTO_SYM_XFORM_CIPHER)
-	continue;
-
-      if (check_algo_is_supported (i, NULL) != 0)
-	continue;
-
-      for (j = dev_info->capabilities; j->op != RTE_CRYPTO_OP_TYPE_UNDEFINED;
-	   j++)
-	{
-	  if (j->sym.xform_type != RTE_CRYPTO_SYM_XFORM_AUTH)
-	    continue;
-
-	  if (check_algo_is_supported (j, NULL) != 0)
-	    continue;
-
-	  mapped |= add_mapping (cwm, cdev_id, qp, is_outbound, i, j);
-	}
-    }
-
-  return mapped;
-}
-
-static int
-check_cryptodev_queues ()
-{
-  u32 n_qs = 0;
-  u8 cdev_id;
-  u32 n_req_qs = 2;
-
-  if (vlib_num_workers () > 0)
-    n_req_qs = vlib_num_workers () * 2;
-
-  for (cdev_id = 0; cdev_id < rte_cryptodev_count (); cdev_id++)
-    {
-      struct rte_cryptodev_info cdev_info;
-
-      rte_cryptodev_info_get (cdev_id, &cdev_info);
-
-      if (!
-	  (cdev_info.feature_flags & RTE_CRYPTODEV_FF_SYM_OPERATION_CHAINING))
-	continue;
-
-      n_qs += cdev_info.max_nb_queue_pairs;
-    }
-
-  if (n_qs >= n_req_qs)
-    return 0;
-  else
-    return -1;
-}
-
-static clib_error_t *
-dpdk_ipsec_check_support (ipsec_sa_t * sa)
-{
-  if (sa->crypto_alg == IPSEC_CRYPTO_ALG_AES_GCM_128)
-    {
-      if (sa->integ_alg != IPSEC_INTEG_ALG_NONE)
-	return clib_error_return (0, "unsupported integ-alg %U with "
-				  "crypto-algo aes-gcm-128",
-				  format_ipsec_integ_alg, sa->integ_alg);
-      sa->integ_alg = IPSEC_INTEG_ALG_AES_GCM_128;
-    }
-  else
-    {
-      if (sa->integ_alg == IPSEC_INTEG_ALG_NONE ||
-	  sa->integ_alg == IPSEC_INTEG_ALG_AES_GCM_128)
-	return clib_error_return (0, "unsupported integ-alg %U",
-				  format_ipsec_integ_alg, sa->integ_alg);
-    }
-
-  return 0;
-}
-
-static uword
-dpdk_ipsec_process (vlib_main_t * vm, vlib_node_runtime_t * rt,
-		    vlib_frame_t * f)
-{
-  dpdk_config_main_t *conf = &dpdk_config_main;
-  ipsec_main_t *im = &ipsec_main;
-  dpdk_crypto_main_t *dcm = &dpdk_crypto_main;
-  vlib_thread_main_t *tm = vlib_get_thread_main ();
-  struct rte_cryptodev_config dev_conf;
-  struct rte_cryptodev_qp_conf qp_conf;
-  struct rte_cryptodev_info cdev_info;
-  struct rte_mempool *rmp;
-  i32 dev_id, ret;
-  u32 i, skip_master;
-
-  if (!conf->cryptodev)
-    {
-      clib_warning ("DPDK Cryptodev support is disabled, "
-		    "default to OpenSSL IPsec");
-      return 0;
-    }
-
-  if (check_cryptodev_queues () < 0)
-    {
-      conf->cryptodev = 0;
-      clib_warning ("not enough Cryptodevs, default to OpenSSL IPsec");
-      return 0;
-    }
-
-  vec_alloc (dcm->workers_main, tm->n_vlib_mains);
-  _vec_len (dcm->workers_main) = tm->n_vlib_mains;
-
-  fprintf (stdout, "DPDK Cryptodevs info:\n");
-  fprintf (stdout, "dev_id\tn_qp\tnb_obj\tcache_size\n");
-  /* HW cryptodevs have higher dev_id, use HW first */
-  for (dev_id = rte_cryptodev_count () - 1; dev_id >= 0; dev_id--)
-    {
-      u16 max_nb_qp, qp = 0;
-      skip_master = vlib_num_workers () > 0;
-
-      rte_cryptodev_info_get (dev_id, &cdev_info);
-
-      if (!
-	  (cdev_info.feature_flags & RTE_CRYPTODEV_FF_SYM_OPERATION_CHAINING))
-	continue;
-
-      max_nb_qp = cdev_info.max_nb_queue_pairs;
-
-      for (i = 0; i < tm->n_vlib_mains; i++)
-	{
-	  u8 is_outbound;
-	  crypto_worker_main_t *cwm;
-	  uword *map;
-
-	  if (skip_master)
-	    {
-	      skip_master = 0;
-	      continue;
-	    }
-
-	  cwm = vec_elt_at_index (dcm->workers_main, i);
-	  map = cwm->algo_qp_map;
-
-	  if (!map)
-	    {
-	      map = hash_create (0, sizeof (crypto_worker_qp_key_t));
-	      if (!map)
-		{
-		  clib_warning ("unable to create hash table for worker %u",
-				vlib_mains[i]->cpu_index);
-		  goto error;
-		}
-	      cwm->algo_qp_map = map;
-	    }
-
-	  for (is_outbound = 0; is_outbound < 2 && qp < max_nb_qp;
-	       is_outbound++)
-	    qp += add_cdev_mapping (cwm, &cdev_info, dev_id, qp, is_outbound);
-	}
-
-      if (qp == 0)
-	continue;
-
-      dev_conf.socket_id = rte_cryptodev_socket_id (dev_id);
-      dev_conf.nb_queue_pairs = cdev_info.max_nb_queue_pairs;
-      dev_conf.session_mp.nb_objs = DPDK_CRYPTO_NB_SESS_OBJS;
-      dev_conf.session_mp.cache_size = DPDK_CRYPTO_CACHE_SIZE;
-
-      ret = rte_cryptodev_configure (dev_id, &dev_conf);
-      if (ret < 0)
-	{
-	  clib_warning ("cryptodev %u config error", dev_id);
-	  goto error;
-	}
-
-      qp_conf.nb_descriptors = DPDK_CRYPTO_N_QUEUE_DESC;
-      for (qp = 0; qp < dev_conf.nb_queue_pairs; qp++)
-	{
-	  ret = rte_cryptodev_queue_pair_setup (dev_id, qp, &qp_conf,
-						dev_conf.socket_id);
-	  if (ret < 0)
-	    {
-	      clib_warning ("cryptodev %u qp %u setup error", dev_id, qp);
-	      goto error;
-	    }
-	}
-      vec_validate_aligned (dcm->cop_pools, dev_conf.socket_id,
-			    CLIB_CACHE_LINE_BYTES);
-
-      if (!vec_elt (dcm->cop_pools, dev_conf.socket_id))
-	{
-	  u8 *pool_name = format (0, "crypto_op_pool_socket%u%c",
-				  dev_conf.socket_id, 0);
-
-	  rmp = rte_crypto_op_pool_create ((char *) pool_name,
-					   RTE_CRYPTO_OP_TYPE_SYMMETRIC,
-					   DPDK_CRYPTO_NB_COPS *
-					   (1 + vlib_num_workers ()),
-					   DPDK_CRYPTO_CACHE_SIZE,
-					   DPDK_CRYPTO_PRIV_SIZE,
-					   dev_conf.socket_id);
-	  vec_free (pool_name);
-
-	  if (!rmp)
-	    {
-	      clib_warning ("failed to allocate mempool on socket %u",
-			    dev_conf.socket_id);
-	      goto error;
-	    }
-	  vec_elt (dcm->cop_pools, dev_conf.socket_id) = rmp;
-	}
-
-      fprintf (stdout, "%u\t%u\t%u\t%u\n", dev_id, dev_conf.nb_queue_pairs,
-	       DPDK_CRYPTO_NB_SESS_OBJS, DPDK_CRYPTO_CACHE_SIZE);
-    }
-
-  dpdk_esp_init ();
-
-  /* Add new next node and set as default */
-  vlib_node_t *node, *next_node;
-
-  next_node = vlib_get_node_by_name (vm, (u8 *) "dpdk-esp-encrypt");
-  ASSERT (next_node);
-  node = vlib_get_node_by_name (vm, (u8 *) "ipsec-output-ip4");
-  ASSERT (node);
-  im->esp_encrypt_node_index = next_node->index;
-  im->esp_encrypt_next_index =
-    vlib_node_add_next (vm, node->index, next_node->index);
-
-  next_node = vlib_get_node_by_name (vm, (u8 *) "dpdk-esp-decrypt");
-  ASSERT (next_node);
-  node = vlib_get_node_by_name (vm, (u8 *) "ipsec-input-ip4");
-  ASSERT (node);
-  im->esp_decrypt_node_index = next_node->index;
-  im->esp_decrypt_next_index =
-    vlib_node_add_next (vm, node->index, next_node->index);
-
-  im->cb.check_support_cb = dpdk_ipsec_check_support;
-  im->cb.add_del_sa_sess_cb = add_del_sa_sess;
-
-  if (vec_len (vlib_mains) == 0)
-    vlib_node_set_state (&vlib_global_main, dpdk_crypto_input_node.index,
-			 VLIB_NODE_STATE_POLLING);
-  else
-    for (i = 1; i < tm->n_vlib_mains; i++)
-      vlib_node_set_state (vlib_mains[i], dpdk_crypto_input_node.index,
-			   VLIB_NODE_STATE_POLLING);
-
-  /* TODO cryptodev counters */
-
-  return 0;
-
-error:
-  ;
-  crypto_worker_main_t *cwm;
-  struct rte_mempool **mp;
-  /* *INDENT-OFF* */
-  vec_foreach (cwm, dcm->workers_main)
-    hash_free (cwm->algo_qp_map);
-
-  vec_foreach (mp, dcm->cop_pools)
-    {
-      if (mp)
-	rte_mempool_free (mp[0]);
-    }
-  /* *INDENT-ON* */
-  vec_free (dcm->workers_main);
-  vec_free (dcm->cop_pools);
-
-  return 0;
-}
-
-/* *INDENT-OFF* */
-VLIB_REGISTER_NODE (dpdk_ipsec_process_node,static) = {
-    .function = dpdk_ipsec_process,
-    .type = VLIB_NODE_TYPE_PROCESS,
-    .name = "dpdk-ipsec-process",
-    .process_log2_n_stack_bytes = 17,
-};
-/* *INDENT-ON* */
-
-/*
- * fd.io coding-style-patch-verification: ON
- *
- * Local Variables:
- * eval: (c-set-style "gnu")
- * End:
- */
diff --git a/src/vnet/devices/dpdk/ipsec/ipsec.h b/src/vnet/devices/dpdk/ipsec/ipsec.h
deleted file mode 100644
index 3465b361..00000000
--- a/src/vnet/devices/dpdk/ipsec/ipsec.h
+++ /dev/null
@@ -1,227 +0,0 @@
-/*
- * Copyright (c) 2016 Intel and/or its affiliates.
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at:
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-#ifndef __DPDK_IPSEC_H__
-#define __DPDK_IPSEC_H__
-
-#include 
-
-#undef always_inline
-#include 
-#include 
-
-#if CLIB_DEBUG > 0
-#define always_inline static inline
-#else
-#define always_inline static inline __attribute__ ((__always_inline__))
-#endif
-
-
-#define MAX_QP_PER_LCORE 16
-
-typedef struct
-{
-  u32 salt;
-  u32 iv[2];
-  u32 cnt;
-} dpdk_gcm_cnt_blk;
-
-typedef struct
-{
-  dpdk_gcm_cnt_blk cb;
-  union
-  {
-    u8 aad[12];
-    u8 icv[64];
-  };
-} dpdk_cop_priv_t;
-
-typedef struct
-{
-  u8 cipher_algo;
-  u8 auth_algo;
-  u8 is_outbound;
-} crypto_worker_qp_key_t;
-
-typedef struct
-{
-  u16 dev_id;
-  u16 qp_id;
-  u16 is_outbound;
-  i16 inflights;
-  u32 bi[VLIB_FRAME_SIZE];
-  struct rte_crypto_op *cops[VLIB_FRAME_SIZE];
-  struct rte_crypto_op **free_cops;
-} crypto_qp_data_t;
-
-typedef struct
-{
-  u8 qp_index;
-  void *sess;
-} crypto_sa_session_t;
-
-typedef struct
-{
-  crypto_sa_session_t *sa_sess_d[2];
-  crypto_qp_data_t *qp_data;
-  uword *algo_qp_map;
-} crypto_worker_main_t;
-
-typedef struct
-{
-  struct rte_mempool **cop_pools;
-  crypto_worker_main_t *workers_main;
-} dpdk_crypto_main_t;
-
-dpdk_crypto_main_t dpdk_crypto_main;
-
-extern vlib_node_registration_t dpdk_crypto_input_node;
-
-#define CRYPTO_N_FREE_COPS (VLIB_FRAME_SIZE * 3)
-
-static_always_inline void
-crypto_alloc_cops ()
-{
-  dpdk_crypto_main_t *dcm = &dpdk_crypto_main;
-  u32 cpu_index = os_get_cpu_number ();
-  crypto_worker_main_t *cwm = &dcm->workers_main[cpu_index];
-  unsigned socket_id = rte_socket_id ();
-  crypto_qp_data_t *qpd;
-
-  /* *INDENT-OFF* */
-  vec_foreach (qpd, cwm->qp_data)
-    {
-      u32 l = vec_len (qpd->free_cops);
-
-      if (PREDICT_FALSE (l < VLIB_FRAME_SIZE))
-	{
-	  u32 n_alloc;
-
-	  if (PREDICT_FALSE (!qpd->free_cops))
-	    vec_alloc (qpd->free_cops, CRYPTO_N_FREE_COPS);
-
-	  n_alloc = rte_crypto_op_bulk_alloc (dcm->cop_pools[socket_id],
-					      RTE_CRYPTO_OP_TYPE_SYMMETRIC,
-					      &qpd->free_cops[l],
-					      CRYPTO_N_FREE_COPS - l - 1);
-
-	  _vec_len (qpd->free_cops) = l + n_alloc;
-	}
-    }
-  /* *INDENT-ON* */
-}
-
-static_always_inline void
-crypto_free_cop (crypto_qp_data_t * qpd, struct rte_crypto_op **cops, u32 n)
-{
-  u32 l = vec_len (qpd->free_cops);
-
-  if (l + n >= CRYPTO_N_FREE_COPS)
-    {
-      l -= VLIB_FRAME_SIZE;
-      rte_mempool_put_bulk (cops[0]->mempool,
-			    (void **) &qpd->free_cops[l], VLIB_FRAME_SIZE);
-    }
-  clib_memcpy (&qpd->free_cops[l], cops, sizeof (*cops) * n);
-
-  _vec_len (qpd->free_cops) = l + n;
-}
-
-static_always_inline int
-check_algo_is_supported (const struct rte_cryptodev_capabilities *cap,
-			 char *name)
-{
-  struct
-  {
-    uint8_t cipher_algo;
-    enum rte_crypto_sym_xform_type type;
-    union
-    {
-      enum rte_crypto_auth_algorithm auth;
-      enum rte_crypto_cipher_algorithm cipher;
-    };
-    char *name;
-  } supported_algo[] =
-  {
-    {
-    .type = RTE_CRYPTO_SYM_XFORM_CIPHER,.cipher =
-	RTE_CRYPTO_CIPHER_NULL,.name = "NULL"},
-    {
-    .type = RTE_CRYPTO_SYM_XFORM_CIPHER,.cipher =
-	RTE_CRYPTO_CIPHER_AES_CBC,.name = "AES_CBC"},
-    {
-    .type = RTE_CRYPTO_SYM_XFORM_CIPHER,.cipher =
-	RTE_CRYPTO_CIPHER_AES_CTR,.name = "AES_CTR"},
-    {
-    .type = RTE_CRYPTO_SYM_XFORM_CIPHER,.cipher =
-	RTE_CRYPTO_CIPHER_3DES_CBC,.name = "3DES-CBC"},
-    {
-    .type = RTE_CRYPTO_SYM_XFORM_CIPHER,.cipher =
-	RTE_CRYPTO_CIPHER_AES_GCM,.name = "AES-GCM"},
-    {
-    .type = RTE_CRYPTO_SYM_XFORM_AUTH,.auth =
-	RTE_CRYPTO_AUTH_SHA1_HMAC,.name = "HMAC-SHA1"},
-    {
-    .type = RTE_CRYPTO_SYM_XFORM_AUTH,.auth =
-	RTE_CRYPTO_AUTH_SHA256_HMAC,.name = "HMAC-SHA256"},
-    {
-    .type = RTE_CRYPTO_SYM_XFORM_AUTH,.auth =
-	RTE_CRYPTO_AUTH_SHA384_HMAC,.name = "HMAC-SHA384"},
-    {
-    .type = RTE_CRYPTO_SYM_XFORM_AUTH,.auth =
-	RTE_CRYPTO_AUTH_SHA512_HMAC,.name = "HMAC-SHA512"},
-    {
-    .type = RTE_CRYPTO_SYM_XFORM_AUTH,.auth =
-	RTE_CRYPTO_AUTH_AES_XCBC_MAC,.name = "AES-XCBC-MAC"},
-    {
-    .type = RTE_CRYPTO_SYM_XFORM_AUTH,.auth =
-	RTE_CRYPTO_AUTH_AES_GCM,.name = "AES-GCM"},
-    {
-      /* tail */
-  .type = RTE_CRYPTO_SYM_XFORM_NOT_SPECIFIED},};
-  uint32_t i = 0;
-
-  if (cap->op != RTE_CRYPTO_OP_TYPE_SYMMETRIC)
-    return -1;
-
-  while (supported_algo[i].type != RTE_CRYPTO_SYM_XFORM_NOT_SPECIFIED)
-    {
-      if (cap->sym.xform_type == supported_algo[i].type)
-	{
-	  if ((cap->sym.xform_type == RTE_CRYPTO_SYM_XFORM_CIPHER &&
-	       cap->sym.cipher.algo == supported_algo[i].cipher) ||
-	      (cap->sym.xform_type == RTE_CRYPTO_SYM_XFORM_AUTH &&
-	       cap->sym.auth.algo == supported_algo[i].auth))
-	    {
-	      if (name)
-		strcpy (name, supported_algo[i].name);
-	      return 0;
-	    }
-	}
-
-      i++;
-    }
-
-  return -1;
-}
-
-#endif /* __DPDK_IPSEC_H__ */
-
-/*
- * fd.io coding-style-patch-verification: ON
- *
- * Local Variables:
- * eval: (c-set-style "gnu")
- * End:
- */
diff --git a/src/vnet/devices/dpdk/main.c b/src/vnet/devices/dpdk/main.c
deleted file mode 100644
index 9ea3aa04..00000000
--- a/src/vnet/devices/dpdk/main.c
+++ /dev/null
@@ -1,85 +0,0 @@
-/*
- * Copyright (c) 2017 Cisco and/or its affiliates.
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at:
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#include 
-#include 
-
-
-/*
- * Called by the dpdk driver's rte_delay_us() function.
- * Return 0 to have the dpdk do a regular delay loop.
- * Return 1 if to skip the delay loop because we are suspending
- * the calling vlib process instead.
- */
-int
-rte_delay_us_override (unsigned us)
-{
-  vlib_main_t *vm;
-
-  /* Don't bother intercepting for short delays */
-  if (us < 10)
-    return 0;
-
-  /*
-   * Only intercept if we are in a vlib process.
-   * If we are called from a vlib worker thread or the vlib main
-   * thread then do not intercept. (Must not be called from an
-   * independent pthread).
-   */
-  if (os_get_cpu_number () == 0)
-    {
-      /*
-       * We're in the vlib main thread or a vlib process. Make sure
-       * the process is running and we're not still initializing.
-       */
-      vm = vlib_get_main ();
-      if (vlib_in_process_context (vm))
-	{
-	  /* Only suspend for the admin_down_process */
-	  vlib_process_t *proc = vlib_get_current_process (vm);
-	  if (!(proc->flags & VLIB_PROCESS_IS_RUNNING) ||
-	      (proc->node_runtime.function != admin_up_down_process))
-	    return 0;
-
-	  f64 delay = 1e-6 * us;
-	  vlib_process_suspend (vm, delay);
-	  return 1;
-	}
-    }
-  return 0;			// no override
-}
-
-static void
-rte_delay_us_override_cb (unsigned us)
-{
-  if (rte_delay_us_override (us) == 0)
-    rte_delay_us_block (us);
-}
-
-static clib_error_t * dpdk_main_init (vlib_main_t * vm)
-{
-  clib_error_t * error = 0;
-
-  if ((error = vlib_call_init_function (vm, dpdk_init)))
-    return error;
-
-  /* register custom delay function */
-  rte_delay_us_callback_register (rte_delay_us_override_cb);
-
-  return error;
-}
-
-VLIB_INIT_FUNCTION (dpdk_main_init);
-
diff --git a/src/vnet/devices/dpdk/node.c b/src/vnet/devices/dpdk/node.c
deleted file mode 100644
index 0d64ae08..00000000
--- a/src/vnet/devices/dpdk/node.c
+++ /dev/null
@@ -1,674 +0,0 @@
-/*
- * Copyright (c) 2015 Cisco and/or its affiliates.
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at:
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-#include 
-#include 
-#include 
-#include 
-#include 
-
-#include 
-#include 
-#include 
-#include 
-#include 
-#include 
-#include 
-
-#include "dpdk_priv.h"
-
-static char *dpdk_error_strings[] = {
-#define _(n,s) s,
-  foreach_dpdk_error
-#undef _
-};
-
-always_inline int
-vlib_buffer_is_ip4 (vlib_buffer_t * b)
-{
-  ethernet_header_t *h = (ethernet_header_t *) b->data;
-  return (h->type == clib_host_to_net_u16 (ETHERNET_TYPE_IP4));
-}
-
-always_inline int
-vlib_buffer_is_ip6 (vlib_buffer_t * b)
-{
-  ethernet_header_t *h = (ethernet_header_t *) b->data;
-  return (h->type == clib_host_to_net_u16 (ETHERNET_TYPE_IP6));
-}
-
-always_inline int
-vlib_buffer_is_mpls (vlib_buffer_t * b)
-{
-  ethernet_header_t *h = (ethernet_header_t *) b->data;
-  return (h->type == clib_host_to_net_u16 (ETHERNET_TYPE_MPLS_UNICAST));
-}
-
-always_inline u32
-dpdk_rx_next_from_etype (struct rte_mbuf * mb, vlib_buffer_t * b0)
-{
-  if (PREDICT_TRUE (vlib_buffer_is_ip4 (b0)))
-    if (PREDICT_TRUE ((mb->ol_flags & PKT_RX_IP_CKSUM_GOOD) != 0))
-      return VNET_DEVICE_INPUT_NEXT_IP4_NCS_INPUT;
-    else
-      return VNET_DEVICE_INPUT_NEXT_IP4_INPUT;
-  else if (PREDICT_TRUE (vlib_buffer_is_ip6 (b0)))
-    return VNET_DEVICE_INPUT_NEXT_IP6_INPUT;
-  else if (PREDICT_TRUE (vlib_buffer_is_mpls (b0)))
-    return VNET_DEVICE_INPUT_NEXT_MPLS_INPUT;
-  else
-    return VNET_DEVICE_INPUT_NEXT_ETHERNET_INPUT;
-}
-
-always_inline int
-dpdk_mbuf_is_vlan (struct rte_mbuf *mb)
-{
-  return (mb->packet_type & RTE_PTYPE_L2_ETHER_VLAN) ==
-    RTE_PTYPE_L2_ETHER_VLAN;
-}
-
-always_inline int
-dpdk_mbuf_is_ip4 (struct rte_mbuf *mb)
-{
-  return RTE_ETH_IS_IPV4_HDR (mb->packet_type) != 0;
-}
-
-always_inline int
-dpdk_mbuf_is_ip6 (struct rte_mbuf *mb)
-{
-  return RTE_ETH_IS_IPV6_HDR (mb->packet_type) != 0;
-}
-
-always_inline u32
-dpdk_rx_next_from_mb (struct rte_mbuf * mb, vlib_buffer_t * b0)
-{
-  if (PREDICT_FALSE (dpdk_mbuf_is_vlan (mb)))
-    return VNET_DEVICE_INPUT_NEXT_ETHERNET_INPUT;
-  else if (PREDICT_TRUE (dpdk_mbuf_is_ip4 (mb)))
-    return VNET_DEVICE_INPUT_NEXT_IP4_NCS_INPUT;
-  else if (PREDICT_TRUE (dpdk_mbuf_is_ip6 (mb)))
-    return VNET_DEVICE_INPUT_NEXT_IP6_INPUT;
-  else if (PREDICT_TRUE (vlib_buffer_is_mpls (b0)))
-    return VNET_DEVICE_INPUT_NEXT_MPLS_INPUT;
-  else
-    return dpdk_rx_next_from_etype (mb, b0);
-}
-
-always_inline void
-dpdk_rx_error_from_mb (struct rte_mbuf *mb, u32 * next, u8 * error)
-{
-  if (mb->ol_flags & PKT_RX_IP_CKSUM_BAD)
-    {
-      *error = DPDK_ERROR_IP_CHECKSUM_ERROR;
-      *next = VNET_DEVICE_INPUT_NEXT_DROP;
-    }
-  else
-    *error = DPDK_ERROR_NONE;
-}
-
-void
-dpdk_rx_trace (dpdk_main_t * dm,
-	       vlib_node_runtime_t * node,
-	       dpdk_device_t * xd,
-	       u16 queue_id, u32 * buffers, uword n_buffers)
-{
-  vlib_main_t *vm = vlib_get_main ();
-  u32 *b, n_left;
-  u32 next0;
-
-  n_left = n_buffers;
-  b = buffers;
-
-  while (n_left >= 1)
-    {
-      u32 bi0;
-      vlib_buffer_t *b0;
-      dpdk_rx_dma_trace_t *t0;
-      struct rte_mbuf *mb;
-      u8 error0;
-
-      bi0 = b[0];
-      n_left -= 1;
-
-      b0 = vlib_get_buffer (vm, bi0);
-      mb = rte_mbuf_from_vlib_buffer (b0);
-
-      if (PREDICT_FALSE (xd->per_interface_next_index != ~0))
-	next0 = xd->per_interface_next_index;
-      else if (PREDICT_TRUE
-	       ((xd->flags & DPDK_DEVICE_FLAG_PMD_SUPPORTS_PTYPE) != 0))
-	next0 = dpdk_rx_next_from_mb (mb, b0);
-      else
-	next0 = dpdk_rx_next_from_etype (mb, b0);
-
-      dpdk_rx_error_from_mb (mb, &next0, &error0);
-
-      vlib_trace_buffer (vm, node, next0, b0, /* follow_chain */ 0);
-      t0 = vlib_add_trace (vm, node, b0, sizeof (t0[0]));
-      t0->queue_index = queue_id;
-      t0->device_index = xd->device_index;
-      t0->buffer_index = bi0;
-
-      clib_memcpy (&t0->mb, mb, sizeof (t0->mb));
-      clib_memcpy (&t0->buffer, b0, sizeof (b0[0]) - sizeof (b0->pre_data));
-      clib_memcpy (t0->buffer.pre_data, b0->data,
-		   sizeof (t0->buffer.pre_data));
-      clib_memcpy (&t0->data, mb->buf_addr + mb->data_off, sizeof (t0->data));
-
-      b += 1;
-    }
-}
-
-static inline u32
-dpdk_rx_burst (dpdk_main_t * dm, dpdk_device_t * xd, u16 queue_id)
-{
-  u32 n_buffers;
-  u32 n_left;
-  u32 n_this_chunk;
-
-  n_left = VLIB_FRAME_SIZE;
-  n_buffers = 0;
-
-  if (PREDICT_TRUE (xd->flags & DPDK_DEVICE_FLAG_PMD))
-    {
-      while (n_left)
-	{
-	  n_this_chunk = rte_eth_rx_burst (xd->device_index, queue_id,
-					   xd->rx_vectors[queue_id] +
-					   n_buffers, n_left);
-	  n_buffers += n_this_chunk;
-	  n_left -= n_this_chunk;
-
-	  /* Empirically, DPDK r1.8 produces vectors w/ 32 or fewer elts */
-	  if (n_this_chunk < 32)
-	    break;
-	}
-    }
-  else
-    {
-      ASSERT (0);
-    }
-
-  return n_buffers;
-}
-
-
-static_always_inline void
-dpdk_process_subseq_segs (vlib_main_t * vm, vlib_buffer_t * b,
-			  struct rte_mbuf *mb, vlib_buffer_free_list_t * fl)
-{
-  u8 nb_seg = 1;
-  struct rte_mbuf *mb_seg = 0;
-  vlib_buffer_t *b_seg, *b_chain = 0;
-  mb_seg = mb->next;
-  b_chain = b;
-
-  while ((mb->nb_segs > 1) && (nb_seg < mb->nb_segs))
-    {
-      ASSERT (mb_seg != 0);
-
-      b_seg = vlib_buffer_from_rte_mbuf (mb_seg);
-      vlib_buffer_init_for_free_list (b_seg, fl);
-
-      ASSERT ((b_seg->flags & VLIB_BUFFER_NEXT_PRESENT) == 0);
-      ASSERT (b_seg->current_data == 0);
-
-      /*
-       * The driver (e.g. virtio) may not put the packet data at the start
-       * of the segment, so don't assume b_seg->current_data == 0 is correct.
-       */
-      b_seg->current_data =
-	(mb_seg->buf_addr + mb_seg->data_off) - (void *) b_seg->data;
-
-      b_seg->current_length = mb_seg->data_len;
-      b->total_length_not_including_first_buffer += mb_seg->data_len;
-
-      b_chain->flags |= VLIB_BUFFER_NEXT_PRESENT;
-      b_chain->next_buffer = vlib_get_buffer_index (vm, b_seg);
-
-      b_chain = b_seg;
-      mb_seg = mb_seg->next;
-      nb_seg++;
-    }
-}
-
-static_always_inline void
-dpdk_prefetch_buffer (struct rte_mbuf *mb)
-{
-  vlib_buffer_t *b = vlib_buffer_from_rte_mbuf (mb);
-  CLIB_PREFETCH (mb, CLIB_CACHE_LINE_BYTES, LOAD);
-  CLIB_PREFETCH (b, CLIB_CACHE_LINE_BYTES, STORE);
-}
-
-/*
- * This function is used when there are no worker threads.
- * The main thread performs IO and forwards the packets.
- */
-static_always_inline u32
-dpdk_device_input (dpdk_main_t * dm, dpdk_device_t * xd,
-		   vlib_node_runtime_t * node, u32 cpu_index, u16 queue_id)
-{
-  u32 n_buffers;
-  u32 next_index = VNET_DEVICE_INPUT_NEXT_ETHERNET_INPUT;
-  u32 n_left_to_next, *to_next;
-  u32 mb_index;
-  vlib_main_t *vm = vlib_get_main ();
-  uword n_rx_bytes = 0;
-  u32 n_trace, trace_cnt __attribute__ ((unused));
-  vlib_buffer_free_list_t *fl;
-  u32 buffer_flags_template;
-
-  if ((xd->flags & DPDK_DEVICE_FLAG_ADMIN_UP) == 0)
-    return 0;
-
-  n_buffers = dpdk_rx_burst (dm, xd, queue_id);
-
-  if (n_buffers == 0)
-    {
-      return 0;
-    }
-
-  buffer_flags_template = dm->buffer_flags_template;
-
-  vec_reset_length (xd->d_trace_buffers[cpu_index]);
-  trace_cnt = n_trace = vlib_get_trace_count (vm, node);
-
-  if (n_trace > 0)
-    {
-      u32 n = clib_min (n_trace, n_buffers);
-      mb_index = 0;
-
-      while (n--)
-	{
-	  struct rte_mbuf *mb = xd->rx_vectors[queue_id][mb_index++];
-	  vlib_buffer_t *b = vlib_buffer_from_rte_mbuf (mb);
-	  vec_add1 (xd->d_trace_buffers[cpu_index],
-		    vlib_get_buffer_index (vm, b));
-	}
-    }
-
-  fl = vlib_buffer_get_free_list (vm, VLIB_BUFFER_DEFAULT_FREE_LIST_INDEX);
-
-  mb_index = 0;
-
-  while (n_buffers > 0)
-    {
-      vlib_buffer_t *b0, *b1, *b2, *b3;
-      u32 bi0, next0, l3_offset0;
-      u32 bi1, next1, l3_offset1;
-      u32 bi2, next2, l3_offset2;
-      u32 bi3, next3, l3_offset3;
-      u8 error0, error1, error2, error3;
-      u64 or_ol_flags;
-
-      vlib_get_next_frame (vm, node, next_index, to_next, n_left_to_next);
-
-      while (n_buffers > 8 && n_left_to_next > 4)
-	{
-	  struct rte_mbuf *mb0 = xd->rx_vectors[queue_id][mb_index];
-	  struct rte_mbuf *mb1 = xd->rx_vectors[queue_id][mb_index + 1];
-	  struct rte_mbuf *mb2 = xd->rx_vectors[queue_id][mb_index + 2];
-	  struct rte_mbuf *mb3 = xd->rx_vectors[queue_id][mb_index + 3];
-
-	  dpdk_prefetch_buffer (xd->rx_vectors[queue_id][mb_index + 4]);
-	  dpdk_prefetch_buffer (xd->rx_vectors[queue_id][mb_index + 5]);
-	  dpdk_prefetch_buffer (xd->rx_vectors[queue_id][mb_index + 6]);
-	  dpdk_prefetch_buffer (xd->rx_vectors[queue_id][mb_index + 7]);
-
-	  if (xd->flags & DPDK_DEVICE_FLAG_MAYBE_MULTISEG)
-	    {
-	      if (PREDICT_FALSE (mb0->nb_segs > 1))
-		dpdk_prefetch_buffer (mb0->next);
-	      if (PREDICT_FALSE (mb1->nb_segs > 1))
-		dpdk_prefetch_buffer (mb1->next);
-	      if (PREDICT_FALSE (mb2->nb_segs > 1))
-		dpdk_prefetch_buffer (mb2->next);
-	      if (PREDICT_FALSE (mb3->nb_segs > 1))
-		dpdk_prefetch_buffer (mb3->next);
-	    }
-
-	  ASSERT (mb0);
-	  ASSERT (mb1);
-	  ASSERT (mb2);
-	  ASSERT (mb3);
-
-	  or_ol_flags = (mb0->ol_flags | mb1->ol_flags |
-			 mb2->ol_flags | mb3->ol_flags);
-	  b0 = vlib_buffer_from_rte_mbuf (mb0);
-	  b1 = vlib_buffer_from_rte_mbuf (mb1);
-	  b2 = vlib_buffer_from_rte_mbuf (mb2);
-	  b3 = vlib_buffer_from_rte_mbuf (mb3);
-
-	  vlib_buffer_init_for_free_list (b0, fl);
-	  vlib_buffer_init_for_free_list (b1, fl);
-	  vlib_buffer_init_for_free_list (b2, fl);
-	  vlib_buffer_init_for_free_list (b3, fl);
-
-	  bi0 = vlib_get_buffer_index (vm, b0);
-	  bi1 = vlib_get_buffer_index (vm, b1);
-	  bi2 = vlib_get_buffer_index (vm, b2);
-	  bi3 = vlib_get_buffer_index (vm, b3);
-
-	  to_next[0] = bi0;
-	  to_next[1] = bi1;
-	  to_next[2] = bi2;
-	  to_next[3] = bi3;
-	  to_next += 4;
-	  n_left_to_next -= 4;
-
-	  if (PREDICT_FALSE (xd->per_interface_next_index != ~0))
-	    {
-	      next0 = next1 = next2 = next3 = xd->per_interface_next_index;
-	    }
-	  else if (PREDICT_TRUE
-		   ((xd->flags & DPDK_DEVICE_FLAG_PMD_SUPPORTS_PTYPE) != 0))
-	    {
-	      next0 = dpdk_rx_next_from_mb (mb0, b0);
-	      next1 = dpdk_rx_next_from_mb (mb1, b1);
-	      next2 = dpdk_rx_next_from_mb (mb2, b2);
-	      next3 = dpdk_rx_next_from_mb (mb3, b3);
-	    }
-	  else
-	    {
-	      next0 = dpdk_rx_next_from_etype (mb0, b0);
-	      next1 = dpdk_rx_next_from_etype (mb1, b1);
-	      next2 = dpdk_rx_next_from_etype (mb2, b2);
-	      next3 = dpdk_rx_next_from_etype (mb3, b3);
-	    }
-
-	  if (PREDICT_FALSE (or_ol_flags & PKT_RX_IP_CKSUM_BAD))
-	    {
-	      dpdk_rx_error_from_mb (mb0, &next0, &error0);
-	      dpdk_rx_error_from_mb (mb1, &next1, &error1);
-	      dpdk_rx_error_from_mb (mb2, &next2, &error2);
-	      dpdk_rx_error_from_mb (mb3, &next3, &error3);
-	      b0->error = node->errors[error0];
-	      b1->error = node->errors[error1];
-	      b2->error = node->errors[error2];
-	      b3->error = node->errors[error3];
-	    }
-	  else
-	    {
-	      b0->error = b1->error = node->errors[DPDK_ERROR_NONE];
-	      b2->error = b3->error = node->errors[DPDK_ERROR_NONE];
-	    }
-
-	  l3_offset0 = device_input_next_node_advance[next0];
-	  l3_offset1 = device_input_next_node_advance[next1];
-	  l3_offset2 = device_input_next_node_advance[next2];
-	  l3_offset3 = device_input_next_node_advance[next3];
-
-	  b0->current_data = l3_offset0 + mb0->data_off;
-	  b1->current_data = l3_offset1 + mb1->data_off;
-	  b2->current_data = l3_offset2 + mb2->data_off;
-	  b3->current_data = l3_offset3 + mb3->data_off;
-
-	  b0->current_data -= RTE_PKTMBUF_HEADROOM;
-	  b1->current_data -= RTE_PKTMBUF_HEADROOM;
-	  b2->current_data -= RTE_PKTMBUF_HEADROOM;
-	  b3->current_data -= RTE_PKTMBUF_HEADROOM;
-
-	  b0->current_length = mb0->data_len - l3_offset0;
-	  b1->current_length = mb1->data_len - l3_offset1;
-	  b2->current_length = mb2->data_len - l3_offset2;
-	  b3->current_length = mb3->data_len - l3_offset3;
-
-	  b0->flags = buffer_flags_template;
-	  b1->flags = buffer_flags_template;
-	  b2->flags = buffer_flags_template;
-	  b3->flags = buffer_flags_template;
-
-	  vnet_buffer (b0)->sw_if_index[VLIB_RX] = xd->vlib_sw_if_index;
-	  vnet_buffer (b1)->sw_if_index[VLIB_RX] = xd->vlib_sw_if_index;
-	  vnet_buffer (b2)->sw_if_index[VLIB_RX] = xd->vlib_sw_if_index;
-	  vnet_buffer (b3)->sw_if_index[VLIB_RX] = xd->vlib_sw_if_index;
-
-	  vnet_buffer (b0)->sw_if_index[VLIB_TX] = (u32) ~ 0;
-	  vnet_buffer (b1)->sw_if_index[VLIB_TX] = (u32) ~ 0;
-	  vnet_buffer (b2)->sw_if_index[VLIB_TX] = (u32) ~ 0;
-	  vnet_buffer (b3)->sw_if_index[VLIB_TX] = (u32) ~ 0;
-
-	  n_rx_bytes += mb0->pkt_len;
-	  n_rx_bytes += mb1->pkt_len;
-	  n_rx_bytes += mb2->pkt_len;
-	  n_rx_bytes += mb3->pkt_len;
-
-	  /* Process subsequent segments of multi-segment packets */
-	  if (xd->flags & DPDK_DEVICE_FLAG_MAYBE_MULTISEG)
-	    {
-	      dpdk_process_subseq_segs (vm, b0, mb0, fl);
-	      dpdk_process_subseq_segs (vm, b1, mb1, fl);
-	      dpdk_process_subseq_segs (vm, b2, mb2, fl);
-	      dpdk_process_subseq_segs (vm, b3, mb3, fl);
-	    }
-
-	  /*
-	   * Turn this on if you run into
-	   * "bad monkey" contexts, and you want to know exactly
-	   * which nodes they've visited... See main.c...
-	   */
-	  VLIB_BUFFER_TRACE_TRAJECTORY_INIT (b0);
-	  VLIB_BUFFER_TRACE_TRAJECTORY_INIT (b1);
-	  VLIB_BUFFER_TRACE_TRAJECTORY_INIT (b2);
-	  VLIB_BUFFER_TRACE_TRAJECTORY_INIT (b3);
-
-	  /* Do we have any driver RX features configured on the interface? */
-	  vnet_feature_start_device_input_x4 (xd->vlib_sw_if_index,
-					      &next0, &next1, &next2, &next3,
-					      b0, b1, b2, b3,
-					      l3_offset0, l3_offset1,
-					      l3_offset2, l3_offset3);
-
-	  vlib_validate_buffer_enqueue_x4 (vm, node, next_index,
-					   to_next, n_left_to_next,
-					   bi0, bi1, bi2, bi3,
-					   next0, next1, next2, next3);
-	  n_buffers -= 4;
-	  mb_index += 4;
-	}
-      while (n_buffers > 0 && n_left_to_next > 0)
-	{
-	  struct rte_mbuf *mb0 = xd->rx_vectors[queue_id][mb_index];
-
-	  ASSERT (mb0);
-
-	  b0 = vlib_buffer_from_rte_mbuf (mb0);
-
-	  /* Prefetch one next segment if it exists. */
-	  if (PREDICT_FALSE (mb0->nb_segs > 1))
-	    dpdk_prefetch_buffer (mb0->next);
-
-	  vlib_buffer_init_for_free_list (b0, fl);
-
-	  bi0 = vlib_get_buffer_index (vm, b0);
-
-	  to_next[0] = bi0;
-	  to_next++;
-	  n_left_to_next--;
-
-	  if (PREDICT_FALSE (xd->per_interface_next_index != ~0))
-	    next0 = xd->per_interface_next_index;
-	  else if (PREDICT_TRUE
-		   ((xd->flags & DPDK_DEVICE_FLAG_PMD_SUPPORTS_PTYPE) != 0))
-	    next0 = dpdk_rx_next_from_mb (mb0, b0);
-	  else
-	    next0 = dpdk_rx_next_from_etype (mb0, b0);
-
-	  dpdk_rx_error_from_mb (mb0, &next0, &error0);
-	  b0->error = node->errors[error0];
-
-	  l3_offset0 = device_input_next_node_advance[next0];
-
-	  b0->current_data = l3_offset0;
-	  b0->current_data += mb0->data_off - RTE_PKTMBUF_HEADROOM;
-	  b0->current_length = mb0->data_len - l3_offset0;
-
-	  b0->flags = buffer_flags_template;
-
-	  vnet_buffer (b0)->sw_if_index[VLIB_RX] = xd->vlib_sw_if_index;
-	  vnet_buffer (b0)->sw_if_index[VLIB_TX] = (u32) ~ 0;
-	  n_rx_bytes += mb0->pkt_len;
-
-	  /* Process subsequent segments of multi-segment packets */
-	  dpdk_process_subseq_segs (vm, b0, mb0, fl);
-
-	  /*
-	   * Turn this on if you run into
-	   * "bad monkey" contexts, and you want to know exactly
-	   * which nodes they've visited... See main.c...
-	   */
-	  VLIB_BUFFER_TRACE_TRAJECTORY_INIT (b0);
-
-	  /* Do we have any driver RX features configured on the interface? */
-	  vnet_feature_start_device_input_x1 (xd->vlib_sw_if_index, &next0,
-					      b0, l3_offset0);
-
-	  vlib_validate_buffer_enqueue_x1 (vm, node, next_index,
-					   to_next, n_left_to_next,
-					   bi0, next0);
-	  n_buffers--;
-	  mb_index++;
-	}
-      vlib_put_next_frame (vm, node, next_index, n_left_to_next);
-    }
-
-  if (PREDICT_FALSE (vec_len (xd->d_trace_buffers[cpu_index]) > 0))
-    {
-      dpdk_rx_trace (dm, node, xd, queue_id, xd->d_trace_buffers[cpu_index],
-		     vec_len (xd->d_trace_buffers[cpu_index]));
-      vlib_set_trace_count (vm, node, n_trace -
-			    vec_len (xd->d_trace_buffers[cpu_index]));
-    }
-
-  vlib_increment_combined_counter
-    (vnet_get_main ()->interface_main.combined_sw_if_counters
-     + VNET_INTERFACE_COUNTER_RX,
-     cpu_index, xd->vlib_sw_if_index, mb_index, n_rx_bytes);
-
-  vnet_device_increment_rx_packets (cpu_index, mb_index);
-
-  return mb_index;
-}
-
-static inline void
-poll_rate_limit (dpdk_main_t * dm)
-{
-  /* Limit the poll rate by sleeping for N msec between polls */
-  if (PREDICT_FALSE (dm->poll_sleep != 0))
-    {
-      struct timespec ts, tsrem;
-
-      ts.tv_sec = 0;
-      ts.tv_nsec = 1000 * 1000 * dm->poll_sleep;	/* 1ms */
-
-      while (nanosleep (&ts, &tsrem) < 0)
-	{
-	  ts = tsrem;
-	}
-    }
-}
-
-/** \brief Main DPDK input node
-    @node dpdk-input
-
-    This is the main DPDK input node: across each assigned interface,
-    call rte_eth_rx_burst(...) or similar to obtain a vector of
-    packets to process. Handle early packet discard. Derive @c
-    vlib_buffer_t metadata from struct rte_mbuf metadata,
-    Depending on the resulting metadata: adjust b->current_data,
-    b->current_length  and dispatch directly to
-    ip4-input-no-checksum, or ip6-input. Trace the packet if required.
-
-    @param vm   vlib_main_t corresponding to the current thread
-    @param node vlib_node_runtime_t
-    @param f    vlib_frame_t input-node, not used.
-
-    @par Graph mechanics: buffer metadata, next index usage
-
-    @em Uses:
-    - struct rte_mbuf mb->ol_flags
-        - PKT_RX_IP_CKSUM_BAD
-    -  RTE_ETH_IS_xxx_HDR(mb->packet_type) 
-        - packet classification result
-
-    @em Sets:
-    - b->error if the packet is to be dropped immediately
-    - b->current_data, b->current_length
-        - adjusted as needed to skip the L2 header in  direct-dispatch cases
-    - vnet_buffer(b)->sw_if_index[VLIB_RX]
-        - rx interface sw_if_index
-    - vnet_buffer(b)->sw_if_index[VLIB_TX] = ~0
-        - required by ipX-lookup
-    - b->flags
-        - to indicate multi-segment pkts (VLIB_BUFFER_NEXT_PRESENT), etc.
-
-    Next Nodes:
-    - Static arcs to: error-drop, ethernet-input,
-      ip4-input-no-checksum, ip6-input, mpls-input
-    - per-interface redirection, controlled by
-      xd->per_interface_next_index
-*/
-
-static uword
-dpdk_input (vlib_main_t * vm, vlib_node_runtime_t * node, vlib_frame_t * f)
-{
-  dpdk_main_t *dm = &dpdk_main;
-  dpdk_device_t *xd;
-  uword n_rx_packets = 0;
-  dpdk_device_and_queue_t *dq;
-  u32 cpu_index = os_get_cpu_number ();
-
-  /*
-   * Poll all devices on this cpu for input/interrupts.
-   */
-  /* *INDENT-OFF* */
-  vec_foreach (dq, dm->devices_by_cpu[cpu_index])
-    {
-      xd = vec_elt_at_index(dm->devices, dq->device);
-      n_rx_packets += dpdk_device_input (dm, xd, node, cpu_index, dq->queue_id);
-    }
-  /* *INDENT-ON* */
-
-  poll_rate_limit (dm);
-
-  return n_rx_packets;
-}
-
-/* *INDENT-OFF* */
-VLIB_REGISTER_NODE (dpdk_input_node) = {
-  .function = dpdk_input,
-  .type = VLIB_NODE_TYPE_INPUT,
-  .name = "dpdk-input",
-  .sibling_of = "device-input",
-
-  /* Will be enabled if/when hardware is detected. */
-  .state = VLIB_NODE_STATE_DISABLED,
-
-  .format_buffer = format_ethernet_header_with_length,
-  .format_trace = format_dpdk_rx_dma_trace,
-
-  .n_errors = DPDK_N_ERROR,
-  .error_strings = dpdk_error_strings,
-};
-
-VLIB_NODE_FUNCTION_MULTIARCH (dpdk_input_node, dpdk_input);
-/* *INDENT-ON* */
-
-/*
- * fd.io coding-style-patch-verification: ON
- *
- * Local Variables:
- * eval: (c-set-style "gnu")
- * End:
- */
diff --git a/src/vnet/devices/dpdk/qos_doc.md b/src/vnet/devices/dpdk/qos_doc.md
deleted file mode 100644
index 7c064246..00000000
--- a/src/vnet/devices/dpdk/qos_doc.md
+++ /dev/null
@@ -1,411 +0,0 @@
-# QoS Hierarchical Scheduler    {#qos_doc}
-
-The Quality-of-Service (QoS) scheduler performs egress-traffic management by
-prioritizing the transmission of the packets of different type services and
-subcribers based on the Service Level Agreements (SLAs). The QoS scheduler can
-be enabled on one or more NIC output interfaces depending upon the
-requirement.
-
-
-## Overview
-
-The QoS schdeuler supports a number of scheduling and shaping levels which
-construct hierarchical-tree. The first level in the hierarchy is port (i.e.
-the physical interface) that constitutes the root node of the tree. The
-subsequent level is subport which represents the group of the
-users/subscribers. The individual user/subscriber is represented by the pipe
-at the next level. Each user can have different traffic type based on the
-criteria of specific loss rate, jitter, and latency. These traffic types are
-represented at the traffic-class level in the form of different traffic-
-classes. The last level contains number of queues which are grouped together
-to host the packets of the specific class type traffic.
-
-The QoS scheduler implementation requires flow classification, enqueue  and
-dequeue operations. The flow classification is mandatory stage for HQoS where
-incoming packets are classified by mapping the packet fields information to
-5-tuple (HQoS subport, pipe, traffic class, queue within traffic class, and
-color) and storing that information in mbuf sched field. The enqueue operation
-uses this information to determine the queue for storing the packet, and at
-this stage, if the specific queue is full, QoS drops the packet. The dequeue
-operation consists of scheduling the packet based on its length and available
-credits, and handing over the scheduled packet to the output interface.
-
-For more information on QoS Scheduler, please refer DPDK Programmer's Guide-
-http://dpdk.org/doc/guides/prog_guide/qos_framework.html
-
-
-### QoS Schdeuler Parameters
-
-Following illustrates the default HQoS configuration for each 10GbE output
-port:
-
-Single subport (subport 0):
-  - Subport rate set to 100% of port rate
-  - Each of the 4 traffic classes has rate set to 100% of port rate
-
-4K pipes per subport 0 (pipes 0 .. 4095) with identical configuration:
-  - Pipe rate set to 1/4K of port rate
-  - Each of the 4 traffic classes has rate set to 100% of pipe rate
-  - Within each traffic class, the byte-level WRR weights for the 4 queues are set to 1:1:1:1
-
-
-#### Port configuration
-
-```
-port {
-  rate 1250000000           /* Assuming 10GbE port */
-  frame_overhead 24         /* Overhead fields per Ethernet frame:
-                             * 7B (Preamble) +
-                             * 1B (Start of Frame Delimiter (SFD)) + 
-                             * 4B (Frame Check Sequence (FCS)) +
-                             * 12B (Inter Frame Gap (IFG))
-                             */
-  mtu 1522                  /* Assuming Ethernet/IPv4 pkt (FCS not included) */
-  n_subports_per_port 1     /* Number of subports per output interface */
-  n_pipes_per_subport 4096  /* Number of pipes (users/subscribers) */
-  queue_sizes 64 64 64 64   /* Packet queue size for each traffic class.
-                             * All queues within the same pipe traffic class
-                             * have the same size. Queues from different
-                             * pipes serving the same traffic class have
-                             * the same size. */
-}
-```
-
-
-#### Subport configuration
-
-```
-subport 0 {
-  tb_rate 1250000000        /* Subport level token bucket rate (bytes per second) */
-  tb_size 1000000           /* Subport level token bucket size (bytes) */
-  tc0_rate 1250000000       /* Subport level token bucket rate for traffic class 0 (bytes per second) */
-  tc1_rate 1250000000       /* Subport level token bucket rate for traffic class 1 (bytes per second) */
-  tc2_rate 1250000000       /* Subport level token bucket rate for traffic class 2 (bytes per second) */
-  tc3_rate 1250000000       /* Subport level token bucket rate for traffic class 3 (bytes per second) */
-  tc_period 10              /* Time interval for refilling the token bucket associated with traffic class (Milliseconds) */
-  pipe 0 4095 profile 0     /* pipes (users/subscribers) configured with pipe profile 0 */
-}
-```
-
-
-#### Pipe configuration
-
-```
-pipe_profile 0 {
-  tb_rate 305175          /* Pipe level token bucket rate (bytes per second) */
-  tb_size 1000000         /* Pipe level token bucket size (bytes) */
-  tc0_rate 305175         /* Pipe level token bucket rate for traffic class 0 (bytes per second) */
-  tc1_rate 305175         /* Pipe level token bucket rate for traffic class 1 (bytes per second) */
-  tc2_rate 305175         /* Pipe level token bucket rate for traffic class 2 (bytes per second) */
-  tc3_rate 305175         /* Pipe level token bucket rate for traffic class 3 (bytes per second) */
-  tc_period 40            /* Time interval for refilling the token bucket associated with traffic class at pipe level (Milliseconds) */
-  tc3_oversubscription_weight 1 /* Weight traffic class 3 oversubscription */
-  tc0_wrr_weights 1 1 1 1     /* Pipe queues WRR weights for traffic class 0 */
-  tc1_wrr_weights 1 1 1 1     /* Pipe queues WRR weights for traffic class 1 */
-  tc2_wrr_weights 1 1 1 1     /* Pipe queues WRR weights for traffic class 2 */
-  tc3_wrr_weights 1 1 1 1     /* Pipe queues WRR weights for traffic class 3 */
-}
-```
-
-
-#### Random Early Detection (RED) parameters per traffic class and color (Green / Yellow / Red)
-
-```
-red {
-  tc0_wred_min 48 40 32     /* Minimum threshold for traffic class 0 queue (min_th) in number of packets */
-  tc0_wred_max 64 64 64     /* Maximum threshold for traffic class 0 queue (max_th) in number of packets */
-  tc0_wred_inv_prob 10 10 10    /* Inverse of packet marking probability for traffic class 0 queue (maxp = 1 / maxp_inv) */
-  tc0_wred_weight 9 9 9     /* Traffic Class 0 queue weight */
-  tc1_wred_min 48 40 32     /* Minimum threshold for traffic class 1 queue (min_th) in number of packets */
-  tc1_wred_max 64 64 64     /* Maximum threshold for traffic class 1 queue (max_th) in number of packets */
-  tc1_wred_inv_prob 10 10 10    /* Inverse of packet marking probability for traffic class 1 queue (maxp = 1 / maxp_inv) */
-  tc1_wred_weight 9 9 9     /* Traffic Class 1 queue weight */
-  tc2_wred_min 48 40 32     /* Minimum threshold for traffic class 2 queue (min_th) in number of packets */
-  tc2_wred_max 64 64 64     /* Maximum threshold for traffic class 2 queue (max_th) in number of packets */
-  tc2_wred_inv_prob 10 10 10    /* Inverse of packet marking probability for traffic class 2 queue (maxp = 1 / maxp_inv) */
-  tc2_wred_weight 9 9 9     /* Traffic Class 2 queue weight */
-  tc3_wred_min 48 40 32     /* Minimum threshold for traffic class 3 queue (min_th) in number of packets */
-  tc3_wred_max 64 64 64     /* Maximum threshold for traffic class 3 queue (max_th) in number of packets */
-  tc3_wred_inv_prob 10 10 10    /* Inverse of packet marking probability for traffic class 3 queue (maxp = 1 / maxp_inv) */
-  tc3_wred_weight 9 9 9     /* Traffic Class 3 queue weight */
-}
-```
-
-
-### DPDK QoS Scheduler Integration in VPP
-
-The Hierarchical Quaity-of-Service (HQoS) scheduler object could be seen as
-part of the logical NIC output interface. To enable HQoS on specific output
-interface, vpp startup.conf file has to be configured accordingly. The output
-interface that requires HQoS, should have "hqos" parameter specified in dpdk
-section. Another optional parameter "hqos-thread"  has been defined which can
-be used to associate the output interface with specific hqos thread. In cpu
-section of the config file, "corelist-hqos-threads" is introduced to assign
-logical cpu cores to run the HQoS threads. A HQoS thread can run multiple HQoS
-objects each associated with different output interfaces. All worker threads
-instead of writing packets to NIC TX queue directly, write the packets to a
-software queues. The hqos_threads read the software queues, and enqueue the
-packets to HQoS objects, as well as dequeue packets from HQOS objects and
-write them to NIC output interfaces. The worker threads need to be able to
-send the packets to any output interface, therefore, each HQoS object
-associated with NIC output interface should have software queues equal to
-worker threads count.
-
-Following illustrates the sample startup configuration file with 4x worker
-threads feeding 2x hqos threads that handle each QoS scheduler for 1x output
-interface.
-
-```
-dpdk {
-  socket-mem 16384,16384
-
-  dev 0000:02:00.0 {
-    num-rx-queues 2
-    hqos
-  }
-  dev 0000:06:00.0 {
-    num-rx-queues 2
-    hqos
-  }
-
-  num-mbufs 1000000
-}
-
-cpu {
-  main-core 0
-  corelist-workers  1, 2, 3, 4
-  corelist-hqos-threads  5, 6
-}
-```
-
-
-### QoS scheduler CLI Commands
-
-Each QoS scheduler instance is initialised with default parameters required to
-configure hqos port, subport, pipe and queues. Some of the parameters can be
-re-configured in run-time through CLI commands.
-
-
-#### Configuration
-
-Following commands can be used to configure QoS scheduler parameters.
-
-The command below can be used to set the subport level parameters such as
-token bucket rate (bytes per seconds), token bucket size (bytes), traffic
-class rates (bytes per seconds) and token update period (Milliseconds).
-
-```
-set dpdk interface hqos subport  subport  [rate ]
-    [bktsize ] [tc0 ] [tc1 ] [tc2 ] [tc3 ] [period ]
-```
-
-For setting the pipe profile, following command can be used.    
-
-```
-set dpdk interface hqos pipe  subport  pipe 
-    profile 
-```
-
-To assign QoS scheduler instance to the specific thread, following command can
-be used.
-
-```
-set dpdk interface hqos placement  thread 
-```
-
-The command below is used to set the packet fields required for classifiying
-the incoming packet. As a result of classification process,     packet field
-information will be mapped to 5 tuples (subport, pipe, traffic class, pipe,
-color) and stored in packet mbuf.
-
-```
-set dpdk interface hqos pktfield  id subport|pipe|tc offset 
-    mask 
-```
-
-The DSCP table entries used for idenfiying the traffic class and queue can be set using the command below;   
-
-```  
-set dpdk interface hqos tctbl  entry  tc  queue 
-```
-
-
-#### Show Command
-
-The QoS Scheduler configuration can displayed using the command below.
-
-```
-   vpp# show dpdk interface hqos TenGigabitEthernet2/0/0
-   Thread:
-     Input SWQ size = 4096 packets
-     Enqueue burst size = 256 packets
-     Dequeue burst size = 220 packets
-     Packet field 0: slab position =    0, slab bitmask = 0x0000000000000000   (subport)
-     Packet field 1: slab position =   40, slab bitmask = 0x0000000fff000000   (pipe)
-     Packet field 2: slab position =    8, slab bitmask = 0x00000000000000fc   (tc)
-     Packet field 2  tc translation table: ([Mapped Value Range]: tc/queue tc/queue ...)
-     [ 0 .. 15]: 0/0 0/1 0/2 0/3 1/0 1/1 1/2 1/3 2/0 2/1 2/2 2/3 3/0 3/1 3/2 3/3
-     [16 .. 31]: 0/0 0/1 0/2 0/3 1/0 1/1 1/2 1/3 2/0 2/1 2/2 2/3 3/0 3/1 3/2 3/3
-     [32 .. 47]: 0/0 0/1 0/2 0/3 1/0 1/1 1/2 1/3 2/0 2/1 2/2 2/3 3/0 3/1 3/2 3/3
-     [48 .. 63]: 0/0 0/1 0/2 0/3 1/0 1/1 1/2 1/3 2/0 2/1 2/2 2/3 3/0 3/1 3/2 3/3
-   Port:
-     Rate = 1250000000 bytes/second
-     MTU = 1514 bytes
-     Frame overhead = 24 bytes
-     Number of subports = 1
-     Number of pipes per subport = 4096
-     Packet queue size: TC0 = 64, TC1 = 64, TC2 = 64, TC3 = 64 packets
-     Number of pipe profiles = 1
-   Subport 0:
-     Rate = 120000000 bytes/second
-     Token bucket size = 1000000 bytes
-     Traffic class rate: TC0 = 120000000, TC1 = 120000000, TC2 = 120000000, TC3 = 120000000 bytes/second
-     TC period = 10 milliseconds
-   Pipe profile 0:
-     Rate = 305175 bytes/second
-     Token bucket size = 1000000 bytes
-     Traffic class rate: TC0 = 305175, TC1 = 305175, TC2 = 305175, TC3 = 305175 bytes/second
-     TC period = 40 milliseconds
-     TC0 WRR weights: Q0 = 1, Q1 = 1, Q2 = 1, Q3 = 1
-     TC1 WRR weights: Q0 = 1, Q1 = 1, Q2 = 1, Q3 = 1
-     TC2 WRR weights: Q0 = 1, Q1 = 1, Q2 = 1, Q3 = 1
-     TC3 WRR weights: Q0 = 1, Q1 = 1, Q2 = 1, Q3 = 1
-```
-
-The QoS Scheduler placement over the logical cpu cores can be displayed using
-below command.
-
-```
-    vpp# show dpdk interface hqos placement
-    Thread 5 (vpp_hqos-threads_0 at lcore 5):
-      TenGigabitEthernet2/0/0 queue 0
-    Thread 6 (vpp_hqos-threads_1 at lcore 6):
-      TenGigabitEthernet4/0/1 queue 0
-```
-
-
-### QoS Scheduler Binary APIs
-
-This section explans the available binary APIs for configuring QoS scheduler
-parameters in run-time.
-
-The following API can be used to set the pipe profile of a pipe that belongs
-to a given subport:
-
-```
-sw_interface_set_dpdk_hqos_pipe rx  | sw_if_index 
-    subport  pipe  profile 
-```
-
-The data structures used for set the pipe profile parameter are as follows;
-
-```
-  /** \\brief DPDK interface HQoS pipe profile set request
-    @param client_index - opaque cookie to identify the sender
-    @param context - sender context, to match reply w/ request
-    @param sw_if_index - the interface
-    @param subport - subport ID
-    @param pipe - pipe ID within its subport
-    @param profile - pipe profile ID
-  */
-  define sw_interface_set_dpdk_hqos_pipe {
-    u32 client_index;
-    u32 context;
-    u32 sw_if_index;
-    u32 subport;
-    u32 pipe;
-    u32 profile;
-  };
-
-  /** \\brief DPDK interface HQoS pipe profile set reply
-    @param context - sender context, to match reply w/ request
-    @param retval - request return code
-  */
-  define sw_interface_set_dpdk_hqos_pipe_reply {
-    u32 context;
-    i32 retval;
-  };
-```
-
-The following API can be used to set the subport level parameters, for
-example- token bucket rate (bytes per seconds), token bucket size (bytes),
-traffic class rate (bytes per seconds) and tokens update period.
-
-```
-sw_interface_set_dpdk_hqos_subport rx  | sw_if_index 
-    subport  [rate ] [bktsize ]
-    [tc0 ] [tc1 ] [tc2 ] [tc3 ] [period ]
-```
-
-The data structures used for set the subport level parameter are as follows;
-
-```
-  /** \\brief DPDK interface HQoS subport parameters set request
-    @param client_index - opaque cookie to identify the sender
-    @param context - sender context, to match reply w/ request
-    @param sw_if_index - the interface
-    @param subport - subport ID
-    @param tb_rate - subport token bucket rate (measured in bytes/second)
-    @param tb_size - subport token bucket size (measured in credits)
-    @param tc_rate - subport traffic class 0 .. 3 rates (measured in bytes/second)
-    @param tc_period - enforcement period for rates (measured in milliseconds)
-  */
-  define sw_interface_set_dpdk_hqos_subport {
-    u32 client_index;
-    u32 context;
-    u32 sw_if_index;
-    u32 subport;
-    u32 tb_rate;
-    u32 tb_size;
-    u32 tc_rate[4];
-    u32 tc_period;
-  };
-
-  /** \\brief DPDK interface HQoS subport parameters set reply
-    @param context - sender context, to match reply w/ request
-    @param retval - request return code
-  */
-  define sw_interface_set_dpdk_hqos_subport_reply {
-    u32 context;
-    i32 retval;
-  };
-```
-
-The following API can be used set the DSCP table entry. The DSCP table have
-64 entries to map the packet DSCP field onto traffic class and hqos input
-queue.
-
-```
-sw_interface_set_dpdk_hqos_tctbl rx  | sw_if_index  
-    entry  tc  queue 
-```
-
-The data structures used for setting DSCP table entries are given below.
-
-```
-  /** \\brief DPDK interface HQoS tctbl entry set request
-    @param client_index - opaque cookie to identify the sender
-    @param context - sender context, to match reply w/ request
-    @param sw_if_index - the interface
-    @param entry - entry index ID
-    @param tc - traffic class (0 .. 3)
-    @param queue - traffic class queue (0 .. 3)
-  */
-  define sw_interface_set_dpdk_hqos_tctbl {
-    u32 client_index;
-    u32 context;
-    u32 sw_if_index;
-    u32 entry;
-    u32 tc;
-    u32 queue;
-  };
-
-  /** \\brief DPDK interface HQoS tctbl entry set reply
-    @param context - sender context, to match reply w/ request
-    @param retval - request return code
-  */
-  define sw_interface_set_dpdk_hqos_tctbl_reply {
-    u32 context;
-    i32 retval;
-  };
-```
diff --git a/src/vnet/devices/dpdk/thread.c b/src/vnet/devices/dpdk/thread.c
deleted file mode 100644
index 475dd142..00000000
--- a/src/vnet/devices/dpdk/thread.c
+++ /dev/null
@@ -1,85 +0,0 @@
-/*
- * Copyright (c) 2017 Cisco and/or its affiliates.
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at:
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#include 
-
-#include 
-#include 
-#include 
-#include 
-#include 
-#include 
-#include 
-#include 
-#include 
-#include 
-#include 
-#include 
-#include 
-#include 
-#include 
-#include 
-#include 
-#include 
-#include 
-#include 
-#include 
-#include 
-#include 
-#include 
-
-#include 
-#include 
-#include 
-#include 
-
-static clib_error_t *
-dpdk_launch_thread (void *fp, vlib_worker_thread_t * w, unsigned lcore_id)
-{
-  int r;
-  r = rte_eal_remote_launch (fp, (void *) w, lcore_id);
-  if (r)
-    return clib_error_return (0, "Failed to launch thread %u", lcore_id);
-  return 0;
-}
-
-static clib_error_t *
-dpdk_thread_set_lcore (u32 thread, u16 lcore)
-{
-  return 0;
-}
-
-static vlib_thread_callbacks_t callbacks = {
-  .vlib_launch_thread_cb = &dpdk_launch_thread,
-  .vlib_thread_set_lcore_cb = &dpdk_thread_set_lcore,
-};
-
-static clib_error_t *
-dpdk_thread_init (vlib_main_t * vm)
-{
-  vlib_thread_cb_register (vm, &callbacks);
-  return 0;
-}
-
-VLIB_INIT_FUNCTION (dpdk_thread_init);
-
-/** @endcond */
-/*
- * fd.io coding-style-patch-verification: ON
- *
- * Local Variables:
- * eval: (c-set-style "gnu")
- * End:
- */
diff --git a/src/vnet/devices/virtio/vhost-user.h b/src/vnet/devices/virtio/vhost-user.h
index 3083b614..dd23a909 100644
--- a/src/vnet/devices/virtio/vhost-user.h
+++ b/src/vnet/devices/virtio/vhost-user.h
@@ -328,17 +328,6 @@ typedef struct
 int vhost_user_dump_ifs (vnet_main_t * vnm, vlib_main_t * vm,
 			 vhost_user_intf_details_t ** out_vuids);
 
-// CLI commands to be used from dpdk
-clib_error_t *vhost_user_connect_command_fn (vlib_main_t * vm,
-					     unformat_input_t * input,
-					     vlib_cli_command_t * cmd);
-clib_error_t *vhost_user_delete_command_fn (vlib_main_t * vm,
-					    unformat_input_t * input,
-					    vlib_cli_command_t * cmd);
-clib_error_t *show_vhost_user_command_fn (vlib_main_t * vm,
-					  unformat_input_t * input,
-					  vlib_cli_command_t * cmd);
-
 #endif
 
 /*
diff --git a/src/vnet/ipsec/ipsec_api.c b/src/vnet/ipsec/ipsec_api.c
index 49b475cf..e37bccee 100644
--- a/src/vnet/ipsec/ipsec_api.c
+++ b/src/vnet/ipsec/ipsec_api.c
@@ -79,11 +79,7 @@ static void vl_api_ipsec_spd_add_del_t_handler
   vl_api_ipsec_spd_add_del_reply_t *rmp;
   int rv;
 
-#if DPDK > 0
   rv = ipsec_add_del_spd (vm, ntohl (mp->spd_id), mp->is_add);
-#else
-  rv = VNET_API_ERROR_UNIMPLEMENTED;
-#endif
 
   REPLY_MACRO (VL_API_IPSEC_SPD_ADD_DEL_REPLY);
 #endif
diff --git a/src/vnet/pg/input.c b/src/vnet/pg/input.c
index e15faeb8..4a65b024 100644
--- a/src/vnet/pg/input.c
+++ b/src/vnet/pg/input.c
@@ -1212,10 +1212,10 @@ pg_stream_fill_helper (pg_main_t * pg,
 
   /*
    * Historically, the pg maintained its own free lists and
-   * device drivers tx paths would return pkts. With the DPDK,
-   * that doesn't happen.
+   * device drivers tx paths would return pkts.
    */
-  if (DPDK == 0 && !(s->flags & PG_STREAM_FLAGS_DISABLE_BUFFER_RECYCLE))
+  if (vm->buffer_main->extern_buffer_mgmt == 0 &&
+      !(s->flags & PG_STREAM_FLAGS_DISABLE_BUFFER_RECYCLE))
     f->buffer_init_function = pg_buffer_init;
   f->buffer_init_function_opaque =
     (s - pg->streams) | ((bi - s->buffer_indices) << 24);
@@ -1238,7 +1238,7 @@ pg_stream_fill_helper (pg_main_t * pg,
   n_alloc = n_allocated;
 
   /* Reinitialize buffers */
-  if (DPDK == 0 || CLIB_DEBUG > 0
+  if (vm->buffer_main->extern_buffer_mgmt == 0 || CLIB_DEBUG > 0
       || (s->flags & PG_STREAM_FLAGS_DISABLE_BUFFER_RECYCLE))
     init_buffers_inline
       (vm, s,
@@ -1246,7 +1246,8 @@ pg_stream_fill_helper (pg_main_t * pg,
        n_alloc, (bi - s->buffer_indices) * s->buffer_bytes /* data offset */ ,
        s->buffer_bytes,
        /* set_data */
-       DPDK == 1 || (s->flags & PG_STREAM_FLAGS_DISABLE_BUFFER_RECYCLE) != 0);
+       vm->buffer_main->extern_buffer_mgmt != 0
+       || (s->flags & PG_STREAM_FLAGS_DISABLE_BUFFER_RECYCLE) != 0);
 
   if (next_buffers)
     pg_set_next_buffer_pointers (pg, s, buffers, next_buffers, n_alloc);
diff --git a/src/vnet/pg/stream.c b/src/vnet/pg/stream.c
index c46875e1..560c4b07 100644
--- a/src/vnet/pg/stream.c
+++ b/src/vnet/pg/stream.c
@@ -442,9 +442,8 @@ pg_stream_add (pg_main_t * pg, pg_stream_t * s_init)
     pg_buffer_index_t *bi;
     int n;
 
-#if DPDK > 0
-    s->buffer_bytes = VLIB_BUFFER_DATA_SIZE;
-#endif
+    if (vm->buffer_main->extern_buffer_mgmt)
+      s->buffer_bytes = VLIB_BUFFER_DATA_SIZE;
 
     if (!s->buffer_bytes)
       s->buffer_bytes = s->max_packet_bytes;
diff --git a/src/vnet/replication.c b/src/vnet/replication.c
index 02755195..86d922b5 100644
--- a/src/vnet/replication.c
+++ b/src/vnet/replication.c
@@ -214,9 +214,9 @@ replication_recycle_callback (vlib_main_t * vm, vlib_buffer_free_list_t * fl)
 	  b0->flags |= VLIB_BUFFER_IS_RECYCLED;
 
 #if (CLIB_DEBUG > 0)
-#if DPDK == 0
-	  vlib_buffer_set_known_state (vm, bi0, VLIB_BUFFER_KNOWN_ALLOCATED);
-#endif
+	  if (vm->buffer_main->extern_buffer_mgmt == 0)
+	    vlib_buffer_set_known_state (vm, bi0,
+					 VLIB_BUFFER_KNOWN_ALLOCATED);
 #endif
 
 	  /* If buffer is traced, mark frame as traced */
diff --git a/src/vnet/vnet_all_api_h.h b/src/vnet/vnet_all_api_h.h
index c4075db6..9d3abae5 100644
--- a/src/vnet/vnet_all_api_h.h
+++ b/src/vnet/vnet_all_api_h.h
@@ -30,9 +30,6 @@
 #endif /* included_from_layer_3 */
 
 #include 
-#if DPDK > 0
-#include 
-#endif
 #include 
 #include 
 #include 
diff --git a/src/vpp/api/custom_dump.c b/src/vpp/api/custom_dump.c
index 4cc6aa73..3871601b 100644
--- a/src/vpp/api/custom_dump.c
+++ b/src/vpp/api/custom_dump.c
@@ -237,58 +237,6 @@ static void *vl_api_sw_interface_set_l2_bridge_t_print
   FINISH;
 }
 
-#if DPDK > 0
-static void *vl_api_sw_interface_set_dpdk_hqos_pipe_t_print
-  (vl_api_sw_interface_set_dpdk_hqos_pipe_t * mp, void *handle)
-{
-  u8 *s;
-
-  s = format (0, "SCRIPT: sw_interface_set_dpdk_hqos_pipe ");
-
-  s = format (s, "sw_if_index %u ", ntohl (mp->sw_if_index));
-
-  s = format (s, "subport %u  pipe %u  profile %u ",
-	      ntohl (mp->subport), ntohl (mp->pipe), ntohl (mp->profile));
-
-  FINISH;
-}
-
-static void *vl_api_sw_interface_set_dpdk_hqos_subport_t_print
-  (vl_api_sw_interface_set_dpdk_hqos_subport_t * mp, void *handle)
-{
-  u8 *s;
-
-  s = format (0, "SCRIPT: sw_interface_set_dpdk_hqos_subport ");
-
-  s = format (s, "sw_if_index %u ", ntohl (mp->sw_if_index));
-
-  s =
-    format (s,
-	    "subport %u  rate %u  bkt_size %u  tc0 %u tc1 %u tc2 %u tc3 %u period %u",
-	    ntohl (mp->subport), ntohl (mp->tb_rate), ntohl (mp->tb_size),
-	    ntohl (mp->tc_rate[0]), ntohl (mp->tc_rate[1]),
-	    ntohl (mp->tc_rate[2]), ntohl (mp->tc_rate[3]),
-	    ntohl (mp->tc_period));
-
-  FINISH;
-}
-
-static void *vl_api_sw_interface_set_dpdk_hqos_tctbl_t_print
-  (vl_api_sw_interface_set_dpdk_hqos_tctbl_t * mp, void *handle)
-{
-  u8 *s;
-
-  s = format (0, "SCRIPT: sw_interface_set_dpdk_hqos_tctbl ");
-
-  s = format (s, "sw_if_index %u ", ntohl (mp->sw_if_index));
-
-  s = format (s, "entry %u  tc %u  queue %u",
-	      ntohl (mp->entry), ntohl (mp->tc), ntohl (mp->queue));
-
-  FINISH;
-}
-#endif
-
 static void *vl_api_bridge_domain_add_del_t_print
   (vl_api_bridge_domain_add_del_t * mp, void *handle)
 {
@@ -3036,18 +2984,6 @@ vl_msg_api_custom_dump_configure (api_main_t * am)
     = (void *) vl_api_##f##_t_print;
   foreach_custom_print_function;
 #undef _
-
-#if DPDK > 0
-  /*
-   * manually add DPDK hqos print handlers
-   */
-  am->msg_print_handlers[VL_API_SW_INTERFACE_SET_DPDK_HQOS_PIPE] =
-    (void *) vl_api_sw_interface_set_dpdk_hqos_pipe_t_print;
-  am->msg_print_handlers[VL_API_SW_INTERFACE_SET_DPDK_HQOS_SUBPORT] =
-    (void *) vl_api_sw_interface_set_dpdk_hqos_subport_t_print;
-  am->msg_print_handlers[VL_API_SW_INTERFACE_SET_DPDK_HQOS_TCTBL] =
-    (void *) vl_api_sw_interface_set_dpdk_hqos_tctbl_t_print;
-#endif
 }
 
 /*
diff --git a/src/vpp/api/gmon.c b/src/vpp/api/gmon.c
index b28608f0..610f40ed 100644
--- a/src/vpp/api/gmon.c
+++ b/src/vpp/api/gmon.c
@@ -137,7 +137,8 @@ gmon_process (vlib_main_t * vm, vlib_node_runtime_t * rt, vlib_frame_t * f)
       *gm->vector_rate_ptr = vector_rate;
       now = vlib_time_now (vm);
       dt = now - last_runtime;
-      input_packets = vnet_get_aggregate_rx_packets ();
+      // TODO
+      //input_packets = vnet_get_aggregate_rx_packets ();
       *gm->input_rate_ptr = (f64) (input_packets - last_input_packets) / dt;
       last_runtime = now;
       last_input_packets = input_packets;
diff --git a/src/vpp/api/vpe.api b/src/vpp/api/vpe.api
index 2d6e4f37..7f9c2038 100644
--- a/src/vpp/api/vpe.api
+++ b/src/vpp/api/vpe.api
@@ -41,7 +41,6 @@
  * SESSION APIs: .../vnet/session/{session.api session_api.c}
  * MPLS APIs: see .../src/vnet/mpls/{mpls.api, mpls_api.c}
  * SR APIs: see .../src/vnet/sr/{sr.api, sr_api.c}
- * DPDK APIs: see ... /src/vnet/devices/dpdk/{dpdk.api, dpdk_api.c}
  * CLASSIFY APIs: see ... /src/vnet/classify/{classify.api, classify_api.c}
  * FLOW APIs: see ... /src/vnet/flow/{flow.api, flow_api.c}
  * DHCP APIs: see ... /src/vnet/dhcp/{dhcpk.api, dhcp_api.c}
diff --git a/src/vpp/app/l2t.c b/src/vpp/app/l2t.c
deleted file mode 100644
index e1eda155..00000000
--- a/src/vpp/app/l2t.c
+++ /dev/null
@@ -1,562 +0,0 @@
-/*
- * Copyright (c) 2015 Cisco and/or its affiliates.
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at:
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-#include 
-#include 
-#include 
-
-#if DPDK == 0
-#include 
-#else
-#include 
-#endif
-
-#include 
-#include 
-#include 
-
-l2t_main_t l2t_main;
-
-/* $$$$ unused?
- * get_interface_ethernet_address
- * paints the ethernet address for a given interface
- * into the supplied destination
- */
-void
-get_interface_ethernet_address (l2t_main_t * lm, u8 * dst, u32 sw_if_index)
-{
-  ethernet_main_t *em = ethernet_get_main (lm->vlib_main);
-  ethernet_interface_t *ei;
-  vnet_hw_interface_t *hi;
-
-  hi = vnet_get_sup_hw_interface (lm->vnet_main, sw_if_index);
-  ei = pool_elt_at_index (em->interfaces, hi->hw_instance);
-  clib_memcpy (dst, ei->address, sizeof (ei->address));
-}
-
-/* packet trace format function */
-u8 *
-format_l2t_trace (u8 * s, va_list * args)
-{
-  CLIB_UNUSED (vlib_main_t * vm) = va_arg (*args, vlib_main_t *);
-  CLIB_UNUSED (vlib_node_t * node) = va_arg (*args, vlib_node_t *);
-  l2t_trace_t *t = va_arg (*args, l2t_trace_t *);
-
-  if (t->is_user_to_network)
-    s = format (s, "L2T: %U (client) -> %U (our) session %d",
-		format_ip6_address, &t->client_address,
-		format_ip6_address, &t->our_address, t->session_index);
-  else
-    s = format (s, "L2T: %U (our) -> %U (client) session %d)",
-		format_ip6_address, &t->our_address,
-		format_ip6_address, &t->client_address, t->session_index);
-  return s;
-}
-
-u8 *
-format_l2t_session (u8 * s, va_list * args)
-{
-  l2t_session_t *session = va_arg (*args, l2t_session_t *);
-  l2t_main_t *lm = &l2t_main;
-  u32 counter_index;
-  vlib_counter_t v;
-
-  s = format (s, "[%d] %U (our) %U (client) vlan-id %d rx_sw_if_index %d\n",
-	      session - lm->sessions,
-	      format_ip6_address, &session->our_address,
-	      format_ip6_address, &session->client_address,
-	      clib_net_to_host_u16 (session->vlan_id), session->sw_if_index);
-
-  s = format (s, "   local cookie %llx remote cookie %llx\n",
-	      clib_net_to_host_u64 (session->local_cookie),
-	      clib_net_to_host_u64 (session->remote_cookie));
-
-  if (session->cookie_flags & L2TP_COOKIE_ROLLOVER_LOCAL)
-    {
-      s = format (s, "   local rollover cookie %llx\n",
-		  clib_net_to_host_u64 (session->lcl_ro_cookie));
-    }
-
-  s = format (s, "   local session-id %d remote session-id %d\n",
-	      clib_net_to_host_u32 (session->local_session_id),
-	      clib_net_to_host_u32 (session->remote_session_id));
-
-  s = format (s, "   l2 specific sublayer %s\n",
-	      session->l2_sublayer_present ? "preset" : "absent");
-
-  counter_index =
-    session_index_to_counter_index (session - lm->sessions,
-				    SESSION_COUNTER_USER_TO_NETWORK);
-
-  vlib_get_combined_counter (&lm->counter_main, counter_index, &v);
-  if (v.packets != 0)
-    s = format (s, "   user-to-net: %llu pkts %llu bytes\n",
-		v.packets, v.bytes);
-
-  vlib_get_combined_counter (&lm->counter_main, counter_index + 1, &v);
-
-  if (v.packets != 0)
-    s = format (s, "   net-to-user: %llu pkts %llu bytes\n",
-		v.packets, v.bytes);
-  return s;
-}
-
-static clib_error_t *
-show_session_summary_command_fn (vlib_main_t * vm,
-				 unformat_input_t * input,
-				 vlib_cli_command_t * cmd)
-{
-  l2t_main_t *lm = &l2t_main;
-
-  vlib_cli_output (vm, "%d active sessions\n", pool_elts (lm->sessions));
-
-  return 0;
-}
-
-/* *INDENT-OFF* */
-static VLIB_CLI_COMMAND (show_session_summary_command) = {
-  .path = "show session",
-  .short_help = "show session summary",
-  .function = show_session_summary_command_fn,
-};
-/* *INDENT-ON* */
-
-static clib_error_t *
-show_session_detail_command_fn (vlib_main_t * vm,
-				unformat_input_t * input,
-				vlib_cli_command_t * cmd)
-{
-  l2t_session_t *session;
-  l2t_main_t *lm = &l2t_main;
-
-  /* *INDENT-OFF* */
-  pool_foreach (session, lm->sessions,
-  ({
-    vlib_cli_output (vm, "%U", format_l2t_session, session);
-  }));
-  /* *INDENT-ON* */
-
-  return 0;
-}
-
-/* *INDENT-OFF* */
-static VLIB_CLI_COMMAND (show_session_detail_command) = {
-  .path = "show session detail",
-  .short_help = "show session table detail",
-  .function = show_session_detail_command_fn,
-};
-/* *INDENT-ON* */
-
-static clib_error_t *
-test_counters_command_fn (vlib_main_t * vm,
-			  unformat_input_t * input, vlib_cli_command_t * cmd)
-{
-  l2t_session_t *session;
-  l2t_main_t *lm = &l2t_main;
-  u32 session_index;
-  u32 counter_index;
-  u32 nincr = 0;
-
-  /* *INDENT-OFF* */
-  pool_foreach (session, lm->sessions,
-  ({
-    session_index = session - lm->sessions;
-    counter_index =
-      session_index_to_counter_index (session_index,
-                                      SESSION_COUNTER_USER_TO_NETWORK);
-    vlib_increment_combined_counter (&lm->counter_main,
-                                     counter_index,
-                                     1/*pkt*/, 1111 /*bytes*/);
-    vlib_increment_combined_counter (&lm->counter_main,
-                                     counter_index+1,
-                                     1/*pkt*/, 2222 /*bytes*/);
-    nincr++;
-  }));
-  /* *INDENT-ON* */
-  vlib_cli_output (vm, "Incremented %d active counters\n", nincr);
-
-  return 0;
-}
-
-/* *INDENT-OFF* */
-static VLIB_CLI_COMMAND (test_counters_command) = {
-  .path = "test counters",
-  .short_help = "increment all active counters",
-  .function = test_counters_command_fn,
-};
-/* *INDENT-ON* */
-
-static clib_error_t *
-clear_counters_command_fn (vlib_main_t * vm,
-			   unformat_input_t * input, vlib_cli_command_t * cmd)
-{
-  l2t_session_t *session;
-  l2t_main_t *lm = &l2t_main;
-  u32 session_index;
-  u32 counter_index;
-  u32 nincr = 0;
-
-  /* *INDENT-OFF* */
-  pool_foreach (session, lm->sessions,
-  ({
-    session_index = session - lm->sessions;
-    counter_index =
-      session_index_to_counter_index (session_index,
-                                      SESSION_COUNTER_USER_TO_NETWORK);
-    vlib_zero_combined_counter (&lm->counter_main, counter_index);
-    vlib_zero_combined_counter (&lm->counter_main, counter_index+1);
-    nincr++;
-  }));
-  /* *INDENT-ON* */
-  vlib_cli_output (vm, "Cleared %d active counters\n", nincr);
-
-  return 0;
-}
-
-/* *INDENT-OFF* */
-static VLIB_CLI_COMMAND (clear_counters_command) = {
-  .path = "clear counters",
-  .short_help = "clear all active counters",
-  .function = clear_counters_command_fn,
-};
-/* *INDENT-ON* */
-
-static clib_error_t *
-l2tp_session_add_command_fn (vlib_main_t * vm,
-			     unformat_input_t * input,
-			     vlib_cli_command_t * cmd)
-{
-  ip6_address_t client_address, our_address;
-  ip6_address_t *dst_address_copy, *src_address_copy;
-  unformat_input_t _line_input, *line_input = &_line_input;
-  u32 vlan_id;
-  u32 sw_if_index = (u32) ~ 0;
-  l2t_main_t *lm = &l2t_main;
-  l2t_session_t *s;
-  uword *p;
-  vnet_hw_interface_t *hi;
-  vnet_sw_interface_t *si;
-  u32 next_index;
-  uword vlan_and_sw_if_index_key;
-  u32 counter_index;
-  u64 local_cookie = (u64) ~ 0, remote_cookie = (u64) ~ 0;
-  u32 local_session_id = 1, remote_session_id = 1;
-  int our_address_set = 0, client_address_set = 0;
-  int l2_sublayer_present = 0;
-  clib_error_t *error = NULL;
-
-  /* Get a line of input. */
-  if (!unformat_user (input, unformat_line_input, line_input))
-    return 0;
-
-  while (unformat_check_input (line_input) != UNFORMAT_END_OF_INPUT)
-    {
-      if (unformat (line_input, "client %U",
-		    unformat_ip6_address, &client_address))
-	client_address_set = 1;
-      else if (unformat (line_input, "our %U",
-			 unformat_ip6_address, &our_address))
-	our_address_set = 1;
-      else if (unformat (line_input, "vlan %d", &vlan_id))
-	;
-      else if (unformat (line_input, "l2-interface %U",
-			 unformat_vnet_sw_interface,
-			 vnet_get_main (), &sw_if_index))
-	;
-      else if (unformat (line_input, "interface %U",
-			 unformat_vnet_sw_interface,
-			 vnet_get_main (), &sw_if_index))
-	;
-      else if (unformat (line_input, "local-cookie %llx", &local_cookie))
-	;
-      else if (unformat (line_input, "remote-cookie %llx", &remote_cookie))
-	;
-      else if (unformat (line_input, "local-session-id %d",
-			 &local_session_id))
-	;
-      else if (unformat (line_input, "remote-session-id %d",
-			 &remote_session_id))
-	;
-      else if (unformat (line_input, "l2-sublayer-present"))
-	l2_sublayer_present = 1;
-      else
-	{
-	  error = clib_error_return (0, "parse error: '%U'",
-				     format_unformat_error, line_input);
-	  unformat_free (line_input);
-	  return error;
-	}
-    }
-
-  unformat_free (line_input);
-
-  if (sw_if_index == (u32) ~ 0)
-    return clib_error_return (0, "l2-interface not specified");
-  if (our_address_set == 0)
-    return clib_error_return (0, "our address not specified");
-  if (client_address_set == 0)
-    return clib_error_return (0, "client address not specified");
-
-  remote_session_id = clib_host_to_net_u32 (remote_session_id);
-  local_session_id = clib_host_to_net_u32 (local_session_id);
-
-  switch (lm->lookup_type)
-    {
-    case L2T_LOOKUP_SRC_ADDRESS:
-      p = hash_get_mem (lm->session_by_src_address, &client_address);
-      if (p)
-	return clib_error_return
-	  (0, "Session w/ client address %U already exists",
-	   format_ip6_address, &client_address);
-      break;
-
-    case L2T_LOOKUP_DST_ADDRESS:
-      p = hash_get_mem (lm->session_by_dst_address, &our_address);
-      if (p)
-	return clib_error_return
-	  (0, "Session w/ our address %U already exists",
-	   format_ip6_address, &our_address);
-      break;
-
-    case L2T_LOOKUP_SESSION_ID:
-      p = hash_get (lm->session_by_session_id, local_session_id);
-      if (p)
-	return clib_error_return
-	  (0,
-	   "Session w/ local session id %d already exists",
-	   clib_net_to_host_u32 (local_session_id));
-      break;
-
-    default:
-      ASSERT (0);
-    }
-
-  pool_get (lm->sessions, s);
-  memset (s, 0, sizeof (*s));
-  clib_memcpy (&s->our_address, &our_address, sizeof (s->our_address));
-  clib_memcpy (&s->client_address, &client_address,
-	       sizeof (s->client_address));
-  s->sw_if_index = sw_if_index;
-  s->vlan_id = clib_host_to_net_u16 (vlan_id);
-  s->local_cookie = clib_host_to_net_u64 (local_cookie);
-  l2tp_session_set_remote_cookie (s, remote_cookie);
-  s->local_session_id = local_session_id;
-  s->remote_session_id = remote_session_id;
-  s->l2_sublayer_present = l2_sublayer_present;
-
-  hi = vnet_get_sup_hw_interface (lm->vnet_main, sw_if_index);
-  si = vnet_get_sup_sw_interface (lm->vnet_main, sw_if_index);
-
-  next_index = vlib_node_add_next (vm, l2t_ip6_node.index,
-				   hi->output_node_index);
-  s->l2_output_next_index = next_index;
-  s->l2_output_sw_if_index = si->sw_if_index;
-
-  /* Setup hash table entries */
-  switch (lm->lookup_type)
-    {
-    case L2T_LOOKUP_SRC_ADDRESS:
-      src_address_copy = clib_mem_alloc (sizeof (*src_address_copy));
-      clib_memcpy (src_address_copy, &client_address,
-		   sizeof (*src_address_copy));
-      hash_set_mem (lm->session_by_src_address, src_address_copy,
-		    s - lm->sessions);
-      break;
-    case L2T_LOOKUP_DST_ADDRESS:
-      dst_address_copy = clib_mem_alloc (sizeof (*dst_address_copy));
-      clib_memcpy (dst_address_copy, &our_address,
-		   sizeof (*dst_address_copy));
-      hash_set_mem (lm->session_by_dst_address, dst_address_copy,
-		    s - lm->sessions);
-      break;
-    case L2T_LOOKUP_SESSION_ID:
-      hash_set (lm->session_by_session_id, local_session_id,
-		s - lm->sessions);
-      break;
-
-    default:
-      ASSERT (0);
-    }
-
-  vlan_and_sw_if_index_key = ((uword) (s->vlan_id) << 32) | sw_if_index;
-  hash_set (lm->session_by_vlan_and_rx_sw_if_index,
-	    vlan_and_sw_if_index_key, s - lm->sessions);
-
-  /* validate counters */
-  counter_index =
-    session_index_to_counter_index (s - lm->sessions,
-				    SESSION_COUNTER_USER_TO_NETWORK);
-  vlib_validate_counter (&lm->counter_main, counter_index);
-  vlib_validate_counter (&lm->counter_main, counter_index + 1);
-
-  /* Set promiscuous mode on the l2 interface */
-  ethernet_set_flags (lm->vnet_main, hi->hw_if_index,
-		      ETHERNET_INTERFACE_FLAG_ACCEPT_ALL);
-  vnet_hw_interface_rx_redirect_to_node (lm->vnet_main, hi->hw_if_index,
-					 l2t_l2_node.index);
-  return 0;
-}
-
-/* *INDENT-OFF* */
-static VLIB_CLI_COMMAND (l2tp_session_add_command) = {
-  .path = "l2tp session add",
-  .short_help =
-  "l2tp session add client  our  vlan  local-cookie  remote-cookie  local-session  remote-session  l2-interface ",
-  .function = l2tp_session_add_command_fn,
-};
-/* *INDENT-ON* */
-
-static clib_error_t *
-l2tp_session_del_command_fn (vlib_main_t * vm,
-			     unformat_input_t * input,
-			     vlib_cli_command_t * cmd)
-{
-  l2t_main_t *lm = &l2t_main;
-  u32 session_index;
-  l2t_session_t *s;
-  hash_pair_t *hp;
-  void *key;
-  uword vlan_and_sw_if_index_key;
-
-  if (!unformat (input, "%d", &session_index))
-    return clib_error_return (0, "missing session index: '%U'",
-			      format_unformat_error, input);
-
-  if (pool_is_free_index (lm->sessions, session_index))
-    return clib_error_return (0, "session %d not in use", session_index);
-
-  s = pool_elt_at_index (lm->sessions, session_index);
-
-  switch (lm->lookup_type)
-    {
-    case L2T_LOOKUP_SRC_ADDRESS:
-      hp = hash_get_pair_mem (lm->session_by_src_address, &s->client_address);
-      if (hp)
-	{
-	  key = (void *) (hp->key);
-	  hash_unset_mem (lm->session_by_src_address, &s->client_address);
-	  clib_mem_free (key);
-	}
-      else
-	clib_warning ("session %d src address key %U AWOL",
-		      s - lm->sessions,
-		      format_ip6_address, &s->client_address);
-      break;
-
-    case L2T_LOOKUP_DST_ADDRESS:
-      hp = hash_get_pair_mem (lm->session_by_dst_address, &s->our_address);
-      if (hp)
-	{
-	  key = (void *) (hp->key);
-	  hash_unset_mem (lm->session_by_dst_address, &s->our_address);
-	  clib_mem_free (key);
-	}
-      else
-	clib_warning ("session %d dst address key %U AWOL",
-		      s - lm->sessions, format_ip6_address, &s->our_address);
-      break;
-
-    case L2T_LOOKUP_SESSION_ID:
-      hash_unset (lm->session_by_session_id, s->local_session_id);
-      break;
-
-    default:
-      ASSERT (0);
-    }
-
-  vlan_and_sw_if_index_key = ((uword) (s->vlan_id) << 32) | s->sw_if_index;
-
-  hash_unset (lm->session_by_vlan_and_rx_sw_if_index,
-	      vlan_and_sw_if_index_key);
-
-  pool_put (lm->sessions, s);
-  return 0;
-}
-
-/* *INDENT-OFF* */
-static VLIB_CLI_COMMAND (l2tp_session_del_command) = {
-  .path = "l2tp session delete",
-  .short_help =
-  "l2tp session delete ",
-  .function = l2tp_session_del_command_fn,
-};
-/* *INDENT-ON* */
-
-static clib_error_t *
-l2tp_session_cookie_command_fn (vlib_main_t * vm,
-				unformat_input_t * input,
-				vlib_cli_command_t * cmd)
-{
-  l2t_main_t *lm = &l2t_main;
-  u32 session_index;
-  l2t_session_t *s;
-  u64 lcl_ro_cookie = (u64) ~ 0, rem_ro_cookie = (u64) ~ 0;
-  u8 cookie_flags = 0;
-
-  if (!unformat (input, "%d", &session_index))
-    return clib_error_return (0, "missing session index: '%U'",
-			      format_unformat_error, input);
-
-  if (pool_is_free_index (lm->sessions, session_index))
-    return clib_error_return (0, "session %d not in use", session_index);
-
-  s = pool_elt_at_index (lm->sessions, session_index);
-
-  if (unformat (input, "commit"))
-    {
-      if (!s->cookie_flags)
-	{
-	  return clib_error_return (0, "no rollover cookie ready to commit");
-	}
-      else
-	{
-	  l2tp_session_cookie_commit (s);
-	  return 0;
-	}
-    }
-  if (!unformat (input, "rollover"))
-    return clib_error_return (0, "missing 'commit|rollover': '%U'",
-			      format_unformat_error, input);
-  if (unformat (input, "local %llx", &lcl_ro_cookie))
-    {
-      cookie_flags |= L2TP_COOKIE_ROLLOVER_LOCAL;
-      l2tp_session_set_local_rollover_cookie (s, lcl_ro_cookie);
-    }
-  if (unformat (input, "remote %llx", &rem_ro_cookie))
-    {
-      cookie_flags |= L2TP_COOKIE_ROLLOVER_REMOTE;
-      l2tp_session_set_remote_cookie (s, rem_ro_cookie);
-    }
-  if (!cookie_flags)
-    return clib_error_return (0, "no rollover cookie specified");
-
-  return 0;
-}
-
-/* *INDENT-OFF* */
-static VLIB_CLI_COMMAND (l2tp_session_cookie_command) = {
-  .path = "l2tp session cookie",
-  .short_help =
-  "l2tp session cookie  commit|rollover [local ] [remote ]",
-  .function = l2tp_session_cookie_command_fn,
-};
-/* *INDENT-ON* */
-
-/*
- * fd.io coding-style-patch-verification: ON
- *
- * Local Variables:
- * eval: (c-set-style "gnu")
- * End:
- */
diff --git a/src/vpp/app/l2t_l2.c b/src/vpp/app/l2t_l2.c
deleted file mode 100644
index 07d30d9a..00000000
--- a/src/vpp/app/l2t_l2.c
+++ /dev/null
@@ -1,267 +0,0 @@
-/*
- * Copyright (c) 2015 Cisco and/or its affiliates.
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at:
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-#include 
-#include 
-#include 
-
-#if DPDK == 0
-#include 
-#include 
-#include 
-#else
-#include 
-#endif
-
-#include 
-#include 
-#include 
-
-l2t_main_t l2t_main;
-
-/* Statistics (not really errors) */
-#define foreach_l2t_l2_error                       \
-_(NETWORK_TO_USER, "L2 network to user (ip6) pkts")
-
-static char *l2t_l2_error_strings[] = {
-#define _(sym,string) string,
-  foreach_l2t_l2_error
-#undef _
-};
-
-typedef enum
-{
-#define _(sym,str) L2T_L2_ERROR_##sym,
-  foreach_l2t_l2_error
-#undef _
-    L2T_L2_N_ERROR,
-} l2t_l2_error_t;
-
-/*
- * Packets go to ethernet-input when they don't match a mapping
- */
-typedef enum
-{
-  L2T_L2_NEXT_DROP,
-  L2T_L2_NEXT_ETHERNET_INPUT,
-  L2T_L2_NEXT_IP6_LOOKUP,
-  L2T_L2_N_NEXT,
-} l2t_l2_next_t;
-
-vlib_node_registration_t l2t_l2_node;
-
-#define NSTAGES 3
-
-static inline void
-stage0 (vlib_main_t * vm, vlib_node_runtime_t * node, u32 buffer_index)
-{
-  vlib_buffer_t *b = vlib_get_buffer (vm, buffer_index);
-  vlib_prefetch_buffer_header (b, STORE);
-  CLIB_PREFETCH (b->data, 2 * CLIB_CACHE_LINE_BYTES, STORE);
-}
-
-static inline void
-stage1 (vlib_main_t * vm, vlib_node_runtime_t * node, u32 bi)
-{
-  vlib_buffer_t *b = vlib_get_buffer (vm, bi);
-  l2t_main_t *lm = &l2t_main;
-  ethernet_header_t *eh;
-  ethernet_vlan_header_t *vh;
-  u32 session_index;
-  uword *p;
-  uword vlan_and_sw_if_index_key;
-
-  /* just in case, needed to test with the tun/tap device */
-  vlib_buffer_reset (b);
-
-  eh = vlib_buffer_get_current (b);
-
-  /* Not a VLAN pkt? send to ethernet-input... */
-  if (PREDICT_FALSE (eh->type != clib_host_to_net_u16 (0x8100)))
-    {
-      vnet_buffer (b)->l2t.next_index = L2T_L2_NEXT_ETHERNET_INPUT;
-      return;
-    }
-  vh = (ethernet_vlan_header_t *) (eh + 1);
-
-  /* look up session */
-  vlan_and_sw_if_index_key = ((uword) (vh->priority_cfi_and_id) << 32)
-    | vnet_buffer (b)->sw_if_index[VLIB_RX];
-
-  p = hash_get (lm->session_by_vlan_and_rx_sw_if_index,
-		vlan_and_sw_if_index_key);
-
-  if (PREDICT_FALSE (p == 0))
-    {
-      /* $$$ drop here if not for our MAC? */
-      vnet_buffer (b)->l2t.next_index = L2T_L2_NEXT_ETHERNET_INPUT;
-      return;
-    }
-  else
-    {
-      session_index = p[0];
-    }
-
-  /* Remember mapping index, prefetch the mini counter */
-  vnet_buffer (b)->l2t.next_index = L2T_L2_NEXT_IP6_LOOKUP;
-  vnet_buffer (b)->l2t.session_index = session_index;
-
-  /* Each mapping has 2 x (pkt, byte) counters, hence the shift */
-  CLIB_PREFETCH (lm->counter_main.mini + (p[0] << 1), CLIB_CACHE_LINE_BYTES,
-		 STORE);
-}
-
-static inline u32
-last_stage (vlib_main_t * vm, vlib_node_runtime_t * node, u32 bi)
-{
-  vlib_buffer_t *b = vlib_get_buffer (vm, bi);
-  l2t_main_t *lm = &l2t_main;
-  ethernet_header_t *eh = vlib_buffer_get_current (b);
-  vlib_node_t *n = vlib_get_node (vm, l2t_l2_node.index);
-  u32 node_counter_base_index = n->error_heap_index;
-  vlib_error_main_t *em = &vm->error_main;
-  l2tpv3_header_t *l2t;		/* l2 header */
-  ethernet_vlan_header_t *vh;	/* 802.1q vlan header */
-  u32 counter_index;
-  l2t_session_t *s;
-  ip6_header_t *ip6;
-  u16 payload_ethertype;
-  u8 dst_mac_address[6];
-  u8 src_mac_address[6];
-  u16 payload_length;
-  i32 backup;
-
-  /* Other-than-output pkt? We're done... */
-  if (vnet_buffer (b)->l2t.next_index != L2T_L2_NEXT_IP6_LOOKUP)
-    return vnet_buffer (b)->l2t.next_index;
-
-  vh = (ethernet_vlan_header_t *) (eh + 1);
-
-  em->counters[node_counter_base_index + L2T_L2_ERROR_NETWORK_TO_USER] += 1;
-
-  counter_index =
-    session_index_to_counter_index (vnet_buffer (b)->l2t.session_index,
-				    SESSION_COUNTER_NETWORK_TO_USER);
-
-  /* per-mapping byte stats include the ethernet header */
-  vlib_increment_combined_counter (&lm->counter_main, counter_index,
-				   1 /* packet_increment */ ,
-				   vlib_buffer_length_in_chain (vm, b) +
-				   sizeof (ethernet_header_t));
-
-  s = pool_elt_at_index (lm->sessions, vnet_buffer (b)->l2t.session_index);
-
-  /* Save src/dst MAC addresses */
-#define _(i)  dst_mac_address[i] = eh->dst_address[i];
-  _(0) _(1) _(2) _(3) _(4) _(5);
-#undef _
-#define _(i)  src_mac_address[i] = eh->src_address[i];
-  _(0) _(1) _(2) _(3) _(4) _(5);
-#undef _
-
-  payload_ethertype = vh->type;
-
-  /* Splice out the 802.1q vlan tag */
-  vlib_buffer_advance (b, 4);
-  eh = vlib_buffer_get_current (b);
-
-  /* restore src/dst MAC addresses */
-#define _(i)   eh->dst_address[i] = dst_mac_address[i];
-  _(0) _(1) _(2) _(3) _(4) _(5);
-#undef _
-#define _(i)  eh->src_address[i] = src_mac_address[i];
-  _(0) _(1) _(2) _(3) _(4) _(5);
-#undef _
-  eh->type = payload_ethertype;
-
-  /* Paint on an l2tpv3 hdr */
-  backup = sizeof (*l2t);
-#if 0
-  /* back up 4 bytes less if no l2 sublayer */
-  backup -= s->l2_sublayer_present ? 0 : 4;
-#endif
-
-  vlib_buffer_advance (b, -backup);
-  l2t = vlib_buffer_get_current (b);
-
-  l2t->session_id = s->remote_session_id;
-  l2t->cookie = s->remote_cookie;
-
-#if 0
-  if (s->l2_sublayer_present)
-    l2t->l2_specific_sublayer = 0;
-#endif
-
-  /* Paint on an ip6 header */
-  vlib_buffer_advance (b, -(sizeof (*ip6)));
-  ip6 = vlib_buffer_get_current (b);
-
-  ip6->ip_version_traffic_class_and_flow_label =
-    clib_host_to_net_u32 (0x6 << 28);
-
-  /* calculate ip6 payload length */
-  payload_length = vlib_buffer_length_in_chain (vm, b);
-  payload_length -= sizeof (*ip6);
-
-  ip6->payload_length = clib_host_to_net_u16 (payload_length);
-  ip6->protocol = 0x73;		/* l2tpv3 */
-  ip6->hop_limit = 0xff;
-  ip6->src_address.as_u64[0] = s->our_address.as_u64[0];
-  ip6->src_address.as_u64[1] = s->our_address.as_u64[1];
-  ip6->dst_address.as_u64[0] = s->client_address.as_u64[0];
-  ip6->dst_address.as_u64[1] = s->client_address.as_u64[1];
-
-  return L2T_L2_NEXT_IP6_LOOKUP;
-}
-
-#include 
-
-static uword
-l2t_l2_node_fn (vlib_main_t * vm,
-		vlib_node_runtime_t * node, vlib_frame_t * frame)
-{
-  return dispatch_pipeline (vm, node, frame);
-}
-
-/* *INDENT-OFF* */
-VLIB_REGISTER_NODE (l2t_l2_node) = {
-  .function = l2t_l2_node_fn,
-  .name = "l2t-l2-input",
-  .vector_size = sizeof (u32),
-  .format_trace = format_l2t_trace,
-  .type = VLIB_NODE_TYPE_INTERNAL,
-
-  .n_errors = ARRAY_LEN(l2t_l2_error_strings),
-  .error_strings = l2t_l2_error_strings,
-
-  .n_next_nodes = L2T_L2_N_NEXT,
-
-  /* edit / add dispositions here */
-  .next_nodes = {
-    [L2T_L2_NEXT_IP6_LOOKUP] = "ip6-lookup",
-    [L2T_L2_NEXT_ETHERNET_INPUT] = "ethernet-input",
-    [L2T_L2_NEXT_DROP] = "error-drop",
-  },
-};
-/* *INDENT-ON* */
-
-VLIB_NODE_FUNCTION_MULTIARCH (l2t_l2_node, l2t_l2_node_fn);
-
-/*
- * fd.io coding-style-patch-verification: ON
- *
- * Local Variables:
- * eval: (c-set-style "gnu")
- * End:
- */
-- 
cgit 1.2.3-korg


From 3cc4971882235a539bc6177e8e4b4d92129b3a12 Mon Sep 17 00:00:00 2001
From: Ole Troan 
Date: Wed, 8 Mar 2017 12:02:24 +0100
Subject: Python API: Change from cPython to CFFI.

Change-Id: I03e52466fb3f909ae52b8fba601168f3eadbd972
Signed-off-by: Ole Troan 
---
 Makefile                                 |   5 +-
 build-root/rpm/vpp.spec                  |  12 +-
 src/Makefile.am                          |   1 +
 src/configure.ac                         |   9 +
 src/vpp-api.am                           |  45 +++
 src/vpp-api/pneum/pneum.c                | 488 ++++++++++++++++++++++++
 src/vpp-api/pneum/pneum.h                |  37 ++
 src/vpp-api/pneum/test_pneum.c           | 143 +++++++
 src/vpp-api/python/Makefile.am           |  57 +--
 src/vpp-api/python/pneum/pneum.c         | 467 -----------------------
 src/vpp-api/python/pneum/pneum.h         |  34 --
 src/vpp-api/python/pneum/test_pneum.c    | 143 -------
 src/vpp-api/python/setup.cfg             |   2 +-
 src/vpp-api/python/setup.py              |  15 +-
 src/vpp-api/python/vpp_papi.py           | 635 +++++++++++++++++++++++++++++++
 src/vpp-api/python/vpp_papi/__init__.py  |   3 -
 src/vpp-api/python/vpp_papi/pneum_wrap.c | 226 -----------
 src/vpp-api/python/vpp_papi/vpp_papi.py  | 599 -----------------------------
 test/Makefile                            |   2 +-
 test/vpp_papi_provider.py                |   2 +-
 20 files changed, 1380 insertions(+), 1545 deletions(-)
 create mode 100644 src/vpp-api.am
 create mode 100644 src/vpp-api/pneum/pneum.c
 create mode 100644 src/vpp-api/pneum/pneum.h
 create mode 100644 src/vpp-api/pneum/test_pneum.c
 delete mode 100644 src/vpp-api/python/pneum/pneum.c
 delete mode 100644 src/vpp-api/python/pneum/pneum.h
 delete mode 100644 src/vpp-api/python/pneum/test_pneum.c
 create mode 100644 src/vpp-api/python/vpp_papi.py
 delete mode 100644 src/vpp-api/python/vpp_papi/__init__.py
 delete mode 100644 src/vpp-api/python/vpp_papi/pneum_wrap.c
 delete mode 100644 src/vpp-api/python/vpp_papi/vpp_papi.py

(limited to 'src/Makefile.am')

diff --git a/Makefile b/Makefile
index ebf94db7..f47c98a5 100644
--- a/Makefile
+++ b/Makefile
@@ -39,7 +39,8 @@ endif
 DEB_DEPENDS  = curl build-essential autoconf automake bison libssl-dev ccache
 DEB_DEPENDS += debhelper dkms git libtool libganglia1-dev libapr1-dev dh-systemd
 DEB_DEPENDS += libconfuse-dev git-review exuberant-ctags cscope pkg-config
-DEB_DEPENDS += python-dev python-virtualenv python-pip lcov chrpath autoconf nasm
+DEB_DEPENDS += lcov chrpath autoconf nasm
+DEB_DEPENDS += python-dev python-virtualenv python-pip libffi6
 ifeq ($(OS_VERSION_ID),14.04)
 	DEB_DEPENDS += openjdk-8-jdk-headless
 else
@@ -49,7 +50,7 @@ endif
 RPM_DEPENDS_GROUPS = 'Development Tools'
 RPM_DEPENDS  = redhat-lsb glibc-static java-1.8.0-openjdk-devel yum-utils
 RPM_DEPENDS += openssl-devel https://dl.fedoraproject.org/pub/epel/epel-release-latest-7.noarch.rpm apr-devel
-RPM_DEPENDS += python-devel python-virtualenv lcov chrpath
+RPM_DEPENDS += python-devel python-virtualenv lcov chrpath libffi-devel
 RPM_DEPENDS += https://kojipkgs.fedoraproject.org//packages/nasm/2.12.02/2.fc26/x86_64/nasm-2.12.02-2.fc26.x86_64.rpm
 EPEL_DEPENDS = libconfuse-devel ganglia-devel
 
diff --git a/build-root/rpm/vpp.spec b/build-root/rpm/vpp.spec
index 7bc18ca4..7fa97888 100644
--- a/build-root/rpm/vpp.spec
+++ b/build-root/rpm/vpp.spec
@@ -162,8 +162,8 @@ do
 done
 
 # Python bindings
-mkdir -p -m755 %{buildroot}%{python2_sitelib}
-install -p -m 666 %{_mu_build_dir}/%{_vpp_install_dir}/*/lib/python2.7/site-packages/vpp_papi-*.egg %{buildroot}%{python2_sitelib}
+cd %{_mu_build_dir}/../src/vpp-api/python
+%py2_install
 
 #
 # devel
@@ -226,15 +226,9 @@ done
 sysctl --system
 %systemd_post vpp.service
 
-%post api-python
-easy_install -z %{python2_sitelib}/vpp_papi-*.egg
-
 %preun
 %systemd_preun vpp.service
 
-%preun api-python
-easy_install -mxNq vpp_papi
-
 %postun
 %systemd_postun
 
@@ -285,7 +279,7 @@ fi
 
 %files api-python
 %defattr(644,root,root)
-%{python2_sitelib}/vpp_papi-*.egg
+%{python2_sitelib}/*
 
 %files devel
 %defattr(-,bin,bin)
diff --git a/src/Makefile.am b/src/Makefile.am
index 5daaa48e..41076e0e 100644
--- a/src/Makefile.am
+++ b/src/Makefile.am
@@ -72,6 +72,7 @@ include uri.am
 SUBDIRS += plugins
 
 if ENABLE_PAPI
+include vpp-api.am
 SUBDIRS += vpp-api/python
 endif
 
diff --git a/src/configure.ac b/src/configure.ac
index d90740d9..5e02adc1 100644
--- a/src/configure.ac
+++ b/src/configure.ac
@@ -201,6 +201,15 @@ AM_COND_IF([ENABLE_JAPI],
   AC_SUBST(JAR)
 ])
 
+###############################################################################
+# PYTHON
+###############################################################################
+
+AM_COND_IF([ENABLE_PAPI],
+[
+  AM_PATH_PYTHON
+])
+
 ###############################################################################
 # Output
 ###############################################################################
diff --git a/src/vpp-api.am b/src/vpp-api.am
new file mode 100644
index 00000000..0e05d60d
--- /dev/null
+++ b/src/vpp-api.am
@@ -0,0 +1,45 @@
+# Copyright (c) 2017 Cisco and/or its affiliates.
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at:
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+#
+# VPP API C wrapper extension
+#
+lib_LTLIBRARIES += libpneum.la
+libpneum_la_SOURCES = vpp-api/pneum/pneum.c
+libpneum_la_LIBADD = \
+  $(top_builddir)/libvppinfra.la \
+  $(top_builddir)/libvlibmemoryclient.la \
+  $(top_builddir)/libvlibapi.la \
+  $(top_builddir)/libsvm.la \
+  -lpthread -lm -lrt
+
+libpneum_la_LDFLAGS = -module
+libpneum_la_CPPFLAGS =
+
+nobase_include_HEADERS += vpp-api/pneum/pneum.h
+
+#
+# Test client
+#
+if ENABLE_TESTS
+noinst_PROGRAMS += test_pneum
+test_pneum_SOURCES = vpp-api/pneum/pneum.c vpp-api/pneum/test_pneum.c
+test_pneum_LDADD = \
+  $(top_builddir)/libvppinfra.la \
+  $(top_builddir)/libvlibmemoryclient.la \
+  $(top_builddir)/libvlibapi.la \
+  $(top_builddir)/libsvm.la \
+  -lpthread -lm -lrt
+endif
+
+# vi:syntax=automake
diff --git a/src/vpp-api/pneum/pneum.c b/src/vpp-api/pneum/pneum.c
new file mode 100644
index 00000000..cbae5cff
--- /dev/null
+++ b/src/vpp-api/pneum/pneum.c
@@ -0,0 +1,488 @@
+/*
+ * Copyright (c) 2016 Cisco and/or its affiliates.
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at:
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+#include 
+#include 
+#include 
+#include 
+#include 
+#include 
+#include 
+#include 
+#include 
+#include 
+#include 
+#include 
+#include 
+#include 
+#include 
+#include 
+
+#include 
+
+#include "pneum.h"
+
+/*
+ * Asynchronous mode:
+ *  Client registers a callback. All messages are sent to the callback.
+ * Synchronous mode:
+ *  Client calls blocking read().
+ *  Clients are expected to collate events on a queue.
+ *  pneum_write() -> suspends RX thread
+ *  pneum_read() -> resumes RX thread
+ */
+
+#define vl_typedefs             /* define message structures */
+#include 
+#undef vl_typedefs
+
+#define vl_endianfun             /* define message structures */
+#include 
+#undef vl_endianfun
+
+vlib_main_t vlib_global_main;
+vlib_main_t **vlib_mains;
+
+typedef struct {
+  u8 connected_to_vlib;
+  pthread_t rx_thread_handle;
+  pthread_t timeout_thread_handle;
+  pthread_mutex_t queue_lock;
+  pthread_cond_t suspend_cv;
+  pthread_cond_t resume_cv;
+  pthread_mutex_t timeout_lock;
+  pthread_cond_t timeout_cv;
+  pthread_cond_t timeout_cancel_cv;
+  pthread_cond_t terminate_cv;
+} pneum_main_t;
+
+pneum_main_t pneum_main;
+pneum_callback_t pneum_callback;
+u16 read_timeout = 0;
+bool rx_is_running = false;
+
+static void
+init (void)
+{
+  pneum_main_t *pm = &pneum_main;
+  memset(pm, 0, sizeof(*pm));
+  pthread_mutex_init(&pm->queue_lock, NULL);
+  pthread_cond_init(&pm->suspend_cv, NULL);
+  pthread_cond_init(&pm->resume_cv, NULL);
+  pthread_mutex_init(&pm->timeout_lock, NULL);
+  pthread_cond_init(&pm->timeout_cv, NULL);
+  pthread_cond_init(&pm->timeout_cancel_cv, NULL);
+  pthread_cond_init(&pm->terminate_cv, NULL);
+}
+
+static void
+cleanup (void)
+{
+  pneum_main_t *pm = &pneum_main;
+  pthread_cond_destroy(&pm->suspend_cv);
+  pthread_cond_destroy(&pm->resume_cv);
+  pthread_cond_destroy(&pm->timeout_cv);
+  pthread_cond_destroy(&pm->timeout_cancel_cv);
+  pthread_cond_destroy(&pm->terminate_cv);
+  pthread_mutex_destroy(&pm->queue_lock);
+  pthread_mutex_destroy(&pm->timeout_lock);
+  memset (pm, 0, sizeof (*pm));
+}
+
+/*
+ * Satisfy external references when -lvlib is not available.
+ */
+void vlib_cli_output (struct vlib_main_t * vm, char * fmt, ...)
+{
+  clib_warning ("vlib_cli_output called...");
+}
+
+void
+pneum_free (void * msg)
+{
+  vl_msg_api_free (msg);
+}
+
+static void
+pneum_api_handler (void *msg)
+{
+  u16 id = ntohs(*((u16 *)msg));
+  msgbuf_t *msgbuf = (msgbuf_t *)(((u8 *)msg) - offsetof(msgbuf_t, data));
+  int l = ntohl(msgbuf->data_len);
+  if (l == 0)
+    clib_warning("Message ID %d has wrong length: %d\n", id, l);
+
+  /* Call Python callback */
+  ASSERT(pneum_callback);
+  (pneum_callback)(msg, l);
+  pneum_free(msg);
+}
+
+static void *
+pneum_rx_thread_fn (void *arg)
+{
+  unix_shared_memory_queue_t *q;
+  pneum_main_t *pm = &pneum_main;
+  api_main_t *am = &api_main;
+  uword msg;
+
+  q = am->vl_input_queue;
+
+  while (1)
+    while (!unix_shared_memory_queue_sub(q, (u8 *)&msg, 0))
+      {
+	u16 id = ntohs(*((u16 *)msg));
+	switch (id) {
+	case VL_API_RX_THREAD_EXIT:
+	  vl_msg_api_free((void *) msg);
+	  /* signal waiting threads that this thread is about to terminate */
+	  pthread_mutex_lock(&pm->queue_lock);
+	  pthread_cond_signal(&pm->terminate_cv);
+	  pthread_mutex_unlock(&pm->queue_lock);
+	  pthread_exit(0);
+	  return 0;
+	  break;
+
+	case VL_API_MEMCLNT_RX_THREAD_SUSPEND:
+	  vl_msg_api_free((void * )msg);
+	  /* Suspend thread and signal reader */
+	  pthread_mutex_lock(&pm->queue_lock);
+	  pthread_cond_signal(&pm->suspend_cv);
+	  /* Wait for the resume signal */
+	  pthread_cond_wait (&pm->resume_cv, &pm->queue_lock);
+	  pthread_mutex_unlock(&pm->queue_lock);
+	  break;
+
+	case VL_API_MEMCLNT_READ_TIMEOUT:
+	  clib_warning("Received read timeout in async thread\n");
+	  vl_msg_api_free((void *) msg);
+	  break;
+
+	default:
+	  pneum_api_handler((void *)msg);
+	}
+      }
+}
+
+static void *
+pneum_timeout_thread_fn (void *arg)
+{
+  vl_api_memclnt_read_timeout_t *ep;
+  pneum_main_t *pm = &pneum_main;
+  api_main_t *am = &api_main;
+  struct timespec ts;
+  struct timeval tv;
+  u16 timeout;
+  int rv;
+
+  while (1)
+    {
+      /* Wait for poke */
+      pthread_mutex_lock(&pm->timeout_lock);
+      pthread_cond_wait (&pm->timeout_cv, &pm->timeout_lock);
+      timeout = read_timeout;
+      gettimeofday(&tv, NULL);
+      ts.tv_sec = tv.tv_sec + timeout;
+      ts.tv_nsec = 0;
+      rv = pthread_cond_timedwait (&pm->timeout_cancel_cv,
+				   &pm->timeout_lock, &ts);
+      pthread_mutex_unlock(&pm->timeout_lock);
+      if (rv == ETIMEDOUT)
+	{
+	  ep = vl_msg_api_alloc (sizeof (*ep));
+	  ep->_vl_msg_id = ntohs(VL_API_MEMCLNT_READ_TIMEOUT);
+	  vl_msg_api_send_shmem(am->vl_input_queue, (u8 *)&ep);
+	}
+    }
+  pthread_exit(0);
+}
+
+void
+pneum_rx_suspend (void)
+{
+  api_main_t *am = &api_main;
+  pneum_main_t *pm = &pneum_main;
+  vl_api_memclnt_rx_thread_suspend_t *ep;
+
+  if (!pm->rx_thread_handle) return;
+  pthread_mutex_lock(&pm->queue_lock);
+  if (rx_is_running)
+    {
+      ep = vl_msg_api_alloc (sizeof (*ep));
+      ep->_vl_msg_id = ntohs(VL_API_MEMCLNT_RX_THREAD_SUSPEND);
+      vl_msg_api_send_shmem(am->vl_input_queue, (u8 *)&ep);
+      /* Wait for RX thread to tell us it has suspendend */
+      pthread_cond_wait(&pm->suspend_cv, &pm->queue_lock);
+      rx_is_running = false;
+    }
+  pthread_mutex_unlock(&pm->queue_lock);
+}
+
+void
+pneum_rx_resume (void)
+{
+  pneum_main_t *pm = &pneum_main;
+  if (!pm->rx_thread_handle) return;
+  pthread_mutex_lock(&pm->queue_lock);
+  if (rx_is_running) return;
+  pthread_cond_signal(&pm->resume_cv);
+  rx_is_running = true;
+  pthread_mutex_unlock(&pm->queue_lock);
+}
+
+static uword *
+pneum_msg_table_get_hash (void)
+{
+  api_main_t *am = &api_main;
+  return (am->msg_index_by_name_and_crc);
+}
+
+int
+pneum_msg_table_size(void)
+{
+  api_main_t *am = &api_main;
+  return hash_elts(am->msg_index_by_name_and_crc);
+}
+
+int
+pneum_connect (char * name, char * chroot_prefix, pneum_callback_t cb,
+               int rx_qlen)
+{
+  int rv = 0;
+  pneum_main_t *pm = &pneum_main;
+
+  init();
+  if (chroot_prefix != NULL)
+    vl_set_memory_root_path (chroot_prefix);
+
+  if ((rv = vl_client_api_map("/vpe-api"))) {
+    clib_warning ("vl_client_api map rv %d", rv);
+    return rv;
+  }
+
+  if (vl_client_connect(name, 0, rx_qlen) < 0) {
+    vl_client_api_unmap();
+    return (-1);
+  }
+
+  if (cb) {
+    /* Start the rx queue thread */
+    rv = pthread_create(&pm->rx_thread_handle, NULL, pneum_rx_thread_fn, 0);
+    if (rv) {
+      clib_warning("pthread_create returned %d", rv);
+      vl_client_api_unmap();
+      return (-1);
+    }
+    pneum_callback = cb;
+    rx_is_running = true;
+  }
+
+  /* Start read timeout thread */
+  rv = pthread_create(&pm->timeout_thread_handle, NULL,
+		      pneum_timeout_thread_fn, 0);
+  if (rv) {
+    clib_warning("pthread_create returned %d", rv);
+    vl_client_api_unmap();
+    return (-1);
+  }
+
+  pm->connected_to_vlib = 1;
+
+  return (0);
+}
+
+int
+pneum_disconnect (void)
+{
+  api_main_t *am = &api_main;
+  pneum_main_t *pm = &pneum_main;
+
+  if (!pm->connected_to_vlib) return 0;
+
+  if (pm->rx_thread_handle) {
+    vl_api_rx_thread_exit_t *ep;
+    uword junk;
+    ep = vl_msg_api_alloc (sizeof (*ep));
+    ep->_vl_msg_id = ntohs(VL_API_RX_THREAD_EXIT);
+    vl_msg_api_send_shmem(am->vl_input_queue, (u8 *)&ep);
+
+    /* wait (with timeout) until RX thread has finished */
+    struct timespec ts;
+    struct timeval tv;
+    gettimeofday(&tv, NULL);
+    ts.tv_sec = tv.tv_sec + 5;
+    ts.tv_nsec = 0;
+    pthread_mutex_lock(&pm->queue_lock);
+    int rv = pthread_cond_timedwait(&pm->terminate_cv, &pm->queue_lock, &ts);
+    pthread_mutex_unlock(&pm->queue_lock);
+    /* now join so we wait until thread has -really- finished */
+    if (rv == ETIMEDOUT)
+      pthread_cancel(pm->rx_thread_handle);
+    else
+      pthread_join(pm->rx_thread_handle, (void **) &junk);
+  }
+  if (pm->timeout_thread_handle)
+    pthread_cancel(pm->timeout_thread_handle);
+
+  vl_client_disconnect();
+  vl_client_api_unmap();
+  pneum_callback = 0;
+
+  cleanup();
+
+  return (0);
+}
+
+static void
+set_timeout (unsigned short timeout)
+{
+  pneum_main_t *pm = &pneum_main;
+  pthread_mutex_lock(&pm->timeout_lock);
+  read_timeout = timeout;
+  pthread_cond_signal(&pm->timeout_cv);
+  pthread_mutex_unlock(&pm->timeout_lock);
+}
+
+static void
+unset_timeout (void)
+{
+  pneum_main_t *pm = &pneum_main;
+  pthread_mutex_lock(&pm->timeout_lock);
+  pthread_cond_signal(&pm->timeout_cancel_cv);
+  pthread_mutex_unlock(&pm->timeout_lock);
+}
+
+int
+pneum_read (char **p, int *l, u16 timeout)
+{
+  unix_shared_memory_queue_t *q;
+  api_main_t *am = &api_main;
+  pneum_main_t *pm = &pneum_main;
+  uword msg;
+  msgbuf_t *msgbuf;
+
+  if (!pm->connected_to_vlib) return -1;
+
+  *l = 0;
+
+  if (am->our_pid == 0) return (-1);
+
+  /* Poke timeout thread */
+  if (timeout)
+    set_timeout(timeout);
+
+  q = am->vl_input_queue;
+  int rv = unix_shared_memory_queue_sub(q, (u8 *)&msg, 0);
+  if (rv == 0) {
+    u16 msg_id = ntohs(*((u16 *)msg));
+    switch (msg_id) {
+    case VL_API_RX_THREAD_EXIT:
+      printf("Received thread exit\n");
+      return -1;
+    case VL_API_MEMCLNT_RX_THREAD_SUSPEND:
+      printf("Received thread suspend\n");
+      goto error;
+    case VL_API_MEMCLNT_READ_TIMEOUT:
+      printf("Received read timeout %ds\n", timeout);
+      goto error;
+
+    default:
+      msgbuf = (msgbuf_t *)(((u8 *)msg) - offsetof(msgbuf_t, data));
+      *l = ntohl(msgbuf->data_len);
+      if (*l == 0) {
+	printf("Unregistered API message: %d\n", msg_id);
+	goto error;
+      }
+    }
+    *p = (char *)msg;
+
+    /* Let timeout notification thread know we're done */
+    unset_timeout();
+
+  } else {
+    printf("Read failed with %d\n", rv);
+  }
+  return (rv);
+
+ error:
+  vl_msg_api_free((void *) msg);
+  /* Client might forget to resume RX thread on failure */
+  pneum_rx_resume ();
+  return -1;
+}
+
+/*
+ * XXX: Makes the assumption that client_index is the first member
+ */
+typedef VL_API_PACKED(struct _vl_api_header {
+  u16 _vl_msg_id;
+  u32 client_index;
+}) vl_api_header_t;
+
+static unsigned int
+pneum_client_index (void)
+{
+  return (api_main.my_client_index);
+}
+
+int
+pneum_write (char *p, int l)
+{
+  int rv = -1;
+  api_main_t *am = &api_main;
+  vl_api_header_t *mp = vl_msg_api_alloc(l);
+  unix_shared_memory_queue_t *q;
+  pneum_main_t *pm = &pneum_main;
+
+  if (!pm->connected_to_vlib) return -1;
+  if (!mp) return (-1);
+
+  memcpy(mp, p, l);
+  mp->client_index = pneum_client_index();
+  q = am->shmem_hdr->vl_input_queue;
+  rv = unix_shared_memory_queue_add(q, (u8 *)&mp, 0);
+  if (rv != 0) {
+    clib_warning("vpe_api_write fails: %d\n", rv);
+    /* Clear message */
+    pneum_free(mp);
+  }
+  return (rv);
+}
+
+int
+pneum_get_msg_index (unsigned char * name)
+{
+  return vl_api_get_msg_index (name);
+}
+
+int
+pneum_msg_table_max_index(void)
+{
+  int max = 0;
+  hash_pair_t *hp;
+  uword *h = pneum_msg_table_get_hash();
+  hash_foreach_pair (hp, h,
+  ({
+    if (hp->value[0] > max)
+      max = hp->value[0];
+  }));
+
+  return max;
+}
+
+void
+pneum_set_error_handler (pneum_error_callback_t cb)
+{
+  if (cb) clib_error_register_handler (cb, 0);
+}
diff --git a/src/vpp-api/pneum/pneum.h b/src/vpp-api/pneum/pneum.h
new file mode 100644
index 00000000..669298df
--- /dev/null
+++ b/src/vpp-api/pneum/pneum.h
@@ -0,0 +1,37 @@
+/*
+ * Copyright (c) 2016 Cisco and/or its affiliates.
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at:
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+#ifndef included_pneum_h
+#define included_pneum_h
+
+#include 
+#include 
+
+typedef void (*pneum_callback_t)(unsigned char * data, int len);
+typedef void (*pneum_error_callback_t)(void *, unsigned char *, int);
+int pneum_connect(char * name, char * chroot_prefix, pneum_callback_t cb,
+    int rx_qlen);
+int pneum_disconnect(void);
+int pneum_read(char **data, int *l, unsigned short timeout);
+int pneum_write(char *data, int len);
+void pneum_free(void * msg);
+
+int pneum_get_msg_index(unsigned char * name);
+int pneum_msg_table_size(void);
+int pneum_msg_table_max_index(void);
+
+void pneum_rx_suspend (void);
+void pneum_rx_resume (void);
+void pneum_set_error_handler(pneum_error_callback_t);
+#endif
diff --git a/src/vpp-api/pneum/test_pneum.c b/src/vpp-api/pneum/test_pneum.c
new file mode 100644
index 00000000..334e58e9
--- /dev/null
+++ b/src/vpp-api/pneum/test_pneum.c
@@ -0,0 +1,143 @@
+/*
+ *------------------------------------------------------------------
+ * test_pneum.c
+ * 
+ * Copyright (c) 2016 Cisco and/or its affiliates.
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at:
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License. 
+ *------------------------------------------------------------------
+ */
+#include 
+#include 
+#include 
+#include 
+#include 
+#include 
+#include 
+#include 
+
+#include        /* time_t, time (for timestamp in second) */
+#include   /* ftime, timeb (for timestamp in millisecond) */
+#include    /* gettimeofday, timeval (for timestamp in microsecond) */
+
+#include 
+#include 
+#include 
+#include 
+#include 
+#include 
+
+#include 
+#include 
+#include 
+#include "pneum.h"
+
+#define vl_typedefs             /* define message structures */
+#include  
+#undef vl_typedefs
+
+/* we are not linking with vlib */
+vlib_main_t vlib_global_main;
+vlib_main_t **vlib_mains;
+
+volatile int sigterm_received = 0;
+volatile u32 result_ready;
+volatile u16 result_msg_id;
+
+/* M_NOALLOC: construct, but don't yet send a message */
+
+#define M_NOALLOC(T,t)                          \
+  do {						\
+    result_ready = 0;                           \
+    memset (mp, 0, sizeof (*mp));		\
+    mp->_vl_msg_id = ntohs (VL_API_##T);	\
+    mp->client_index = am->my_client_index;	\
+  } while(0);
+
+
+
+int
+wrap_pneum_callback (char *data, int len)
+{
+  //printf("Callback %d\n", len);
+  result_ready = 1;
+  result_msg_id = ntohs(*((u16 *)data));
+  return (0);
+}
+
+int main (int argc, char ** argv)
+{
+  api_main_t * am = &api_main;
+  vl_api_show_version_t message;
+  vl_api_show_version_t *mp;
+  int async = 1;
+  int rv = pneum_connect("pneum_client", NULL, NULL, 32 /* rx queue-length*/);
+
+  if (rv != 0) {
+    printf("Connect failed: %d\n", rv);
+    exit(rv);
+  }
+ 
+ struct timeb timer_msec;
+  long long int timestamp_msec_start; /* timestamp in millisecond. */
+  if (!ftime(&timer_msec)) {
+    timestamp_msec_start = ((long long int) timer_msec.time) * 1000ll + 
+      (long long int) timer_msec.millitm;
+  }
+  else {
+    timestamp_msec_start = -1;
+  }
+
+ 
+  /*
+   * Test vpe_api_write and vpe_api_read to send and recv message for an
+   * API 
+   */
+  int i;
+  long int no_msgs = 10000;
+  mp = &message;
+
+  for (i = 0; i < no_msgs; i++) {
+    /* Construct the API message */
+    M_NOALLOC(SHOW_VERSION, show_version);
+    pneum_write((char *)mp, sizeof(*mp));
+#ifndef __COVERITY__
+    /* As given, async is always 1. Shut up Coverity about it */
+    if (!async)
+      while (result_ready == 0);
+#endif
+  }
+  if (async) {
+    vl_api_control_ping_t control;
+    vl_api_control_ping_t *mp;
+    mp = &control;
+    M_NOALLOC(CONTROL_PING, control_ping);
+    pneum_write((char *)mp, sizeof(*mp));
+
+    while (result_msg_id != VL_API_CONTROL_PING_REPLY);
+  }
+
+  long long int timestamp_msec_end; /* timestamp in millisecond. */
+  if (!ftime(&timer_msec)) {
+    timestamp_msec_end = ((long long int) timer_msec.time) * 1000ll + 
+      (long long int) timer_msec.millitm;
+  }
+  else {
+    timestamp_msec_end = -1;
+  }
+  
+  printf("Took %lld msec, %lld msgs/msec \n", (timestamp_msec_end - timestamp_msec_start),
+	 no_msgs/(timestamp_msec_end - timestamp_msec_start));
+  fformat(stdout, "Exiting...\n");
+  pneum_disconnect();
+  exit (0);
+}
diff --git a/src/vpp-api/python/Makefile.am b/src/vpp-api/python/Makefile.am
index 54076822..6f5beb69 100644
--- a/src/vpp-api/python/Makefile.am
+++ b/src/vpp-api/python/Makefile.am
@@ -11,52 +11,11 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-AUTOMAKE_OPTIONS = foreign
-ACLOCAL_AMFLAGS = -I m4
-AM_LIBTOOLFLAGS = --quiet
-AM_CFLAGS = -Wall -I${top_srcdir} -I${top_builddir}
-
-BUILT_SOURCES =
-bin_PROGRAMS =
-CLEANFILES =
-lib_LTLIBRARIES =
-noinst_PROGRAMS =
-nobase_include_HEADERS = pneum/pneum.h
-
-#
-# Python / C extension
-#
-lib_LTLIBRARIES += libpneum.la
-libpneum_la_SOURCES = pneum/pneum.c
-libpneum_la_LIBADD = \
-  $(top_builddir)/libvppinfra.la \
-  $(top_builddir)/libvlibmemoryclient.la \
-  $(top_builddir)/libvlibapi.la \
-  $(top_builddir)/libsvm.la \
-  -lpthread -lm -lrt
-
-libpneum_la_LDFLAGS = -module
-libpneum_la_CPPFLAGS =
-
-# TODO: Support both Python 2 and 3.
-install-exec-local: $(lib_LTLIBRARIES)
-	cd $(srcdir);							\
-	mkdir -p $(pythondir);						\
-	mkdir -p $(pyexecdir); 						\
-	PYTHONUSERBASE=$(prefix)					\
-	python setup.py build_ext -L $(libdir)				\
-	-I $(prefix)/include/ install --user
-
-#
-# Test client
-#
-if ENABLE_TESTS
-noinst_PROGRAMS += test_pneum
-test_pneum_SOURCES = pneum/pneum.c pneum/test_pneum.c
-test_pneum_LDADD = \
-  $(top_builddir)/libvppinfra.la \
-  $(top_builddir)/libvlibmemoryclient.la \
-  $(top_builddir)/libvlibapi.la \
-  $(top_builddir)/libsvm.la \
-  -lpthread -lm -lrt
-endif
+install-exec-local:
+	(cd $(srcdir) ; $(PYTHON) $(srcdir)/setup.py build              \
+	--build-base $(shell readlink -f $(builddir))/build             \
+	install                                                         \
+	--root /                                                        \
+	--prefix $(DESTDIR)$(prefix)                                    \
+	--single-version-externally-managed                             \
+	--verbose)
diff --git a/src/vpp-api/python/pneum/pneum.c b/src/vpp-api/python/pneum/pneum.c
deleted file mode 100644
index da9d69df..00000000
--- a/src/vpp-api/python/pneum/pneum.c
+++ /dev/null
@@ -1,467 +0,0 @@
-/*
- * Copyright (c) 2016 Cisco and/or its affiliates.
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at:
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-#include 
-#include 
-#include 
-#include 
-#include 
-#include 
-#include 
-#include 
-#include 
-#include 
-#include 
-#include 
-#include 
-#include 
-#include 
-#include 
-
-#include 
-
-#include "pneum.h"
-
-/*
- * Asynchronous mode:
- *  Client registers a callback. All messages are sent to the callback.
- * Synchronous mode:
- *  Client calls blocking read().
- *  Clients are expected to collate events on a queue.
- *  pneum_write() -> suspends RX thread
- *  pneum_read() -> resumes RX thread
- */
-
-#define vl_typedefs             /* define message structures */
-#include 
-#undef vl_typedefs
-
-#define vl_endianfun             /* define message structures */
-#include 
-#undef vl_endianfun
-
-vlib_main_t vlib_global_main;
-vlib_main_t **vlib_mains;
-
-typedef struct {
-  u8 connected_to_vlib;
-  pthread_t rx_thread_handle;
-  pthread_t timeout_thread_handle;
-  pthread_mutex_t queue_lock;
-  pthread_cond_t suspend_cv;
-  pthread_cond_t resume_cv;
-  pthread_mutex_t timeout_lock;
-  pthread_cond_t timeout_cv;
-  pthread_cond_t timeout_cancel_cv;
-  pthread_cond_t terminate_cv;
-} pneum_main_t;
-
-pneum_main_t pneum_main;
-pneum_callback_t pneum_callback;
-u16 read_timeout = 0;
-bool rx_is_running = false;
-
-static void
-init (void)
-{
-  pneum_main_t *pm = &pneum_main;
-  memset(pm, 0, sizeof(*pm));
-  pthread_mutex_init(&pm->queue_lock, NULL);
-  pthread_cond_init(&pm->suspend_cv, NULL);
-  pthread_cond_init(&pm->resume_cv, NULL);
-  pthread_mutex_init(&pm->timeout_lock, NULL);
-  pthread_cond_init(&pm->timeout_cv, NULL);
-  pthread_cond_init(&pm->timeout_cancel_cv, NULL);
-  pthread_cond_init(&pm->terminate_cv, NULL);
-}
-
-static void
-cleanup (void)
-{
-  pneum_main_t *pm = &pneum_main;
-  pthread_cond_destroy(&pm->suspend_cv);
-  pthread_cond_destroy(&pm->resume_cv);
-  pthread_cond_destroy(&pm->timeout_cv);
-  pthread_cond_destroy(&pm->timeout_cancel_cv);
-  pthread_cond_destroy(&pm->terminate_cv);
-  pthread_mutex_destroy(&pm->queue_lock);
-  pthread_mutex_destroy(&pm->timeout_lock);
-  memset (pm, 0, sizeof (*pm));
-}
-
-/*
- * Satisfy external references when -lvlib is not available.
- */
-void vlib_cli_output (struct vlib_main_t * vm, char * fmt, ...)
-{
-  clib_warning ("vlib_cli_output called...");
-}
-
-void
-pneum_free (void * msg)
-{
-  vl_msg_api_free (msg);
-}
-
-static void
-pneum_api_handler (void *msg)
-{
-  u16 id = ntohs(*((u16 *)msg));
-  msgbuf_t *msgbuf = (msgbuf_t *)(((u8 *)msg) - offsetof(msgbuf_t, data));
-  int l = ntohl(msgbuf->data_len);
-  if (l == 0)
-    clib_warning("Message ID %d has wrong length: %d\n", id, l);
-
-  /* Call Python callback */
-  ASSERT(pneum_callback);
-  (pneum_callback)(msg, l);
-  pneum_free(msg);
-}
-
-static void *
-pneum_rx_thread_fn (void *arg)
-{
-  unix_shared_memory_queue_t *q;
-  pneum_main_t *pm = &pneum_main;
-  api_main_t *am = &api_main;
-  uword msg;
-
-  q = am->vl_input_queue;
-
-  while (1)
-    while (!unix_shared_memory_queue_sub(q, (u8 *)&msg, 0))
-      {
-	u16 id = ntohs(*((u16 *)msg));
-	switch (id) {
-	case VL_API_RX_THREAD_EXIT:
-	  vl_msg_api_free((void *) msg);
-	  /* signal waiting threads that this thread is about to terminate */
-	  pthread_mutex_lock(&pm->queue_lock);
-	  pthread_cond_signal(&pm->terminate_cv);
-	  pthread_mutex_unlock(&pm->queue_lock);
-	  pthread_exit(0);
-	  return 0;
-	  break;
-
-	case VL_API_MEMCLNT_RX_THREAD_SUSPEND:
-	  vl_msg_api_free((void * )msg);
-	  /* Suspend thread and signal reader */
-	  pthread_mutex_lock(&pm->queue_lock);
-	  pthread_cond_signal(&pm->suspend_cv);
-	  /* Wait for the resume signal */
-	  pthread_cond_wait (&pm->resume_cv, &pm->queue_lock);
-	  pthread_mutex_unlock(&pm->queue_lock);
-	  break;
-
-	case VL_API_MEMCLNT_READ_TIMEOUT:
-	  clib_warning("Received read timeout in async thread\n");
-	  vl_msg_api_free((void *) msg);
-	  break;
-
-	default:
-	  pneum_api_handler((void *)msg);
-	}
-      }
-}
-
-static void *
-pneum_timeout_thread_fn (void *arg)
-{
-  vl_api_memclnt_read_timeout_t *ep;
-  pneum_main_t *pm = &pneum_main;
-  api_main_t *am = &api_main;
-  struct timespec ts;
-  struct timeval tv;
-  u16 timeout;
-  int rv;
-
-  while (1)
-    {
-      /* Wait for poke */
-      pthread_mutex_lock(&pm->timeout_lock);
-      pthread_cond_wait (&pm->timeout_cv, &pm->timeout_lock);
-      timeout = read_timeout;
-      gettimeofday(&tv, NULL);
-      ts.tv_sec = tv.tv_sec + timeout;
-      ts.tv_nsec = 0;
-      rv = pthread_cond_timedwait (&pm->timeout_cancel_cv,
-				   &pm->timeout_lock, &ts);
-      pthread_mutex_unlock(&pm->timeout_lock);
-      if (rv == ETIMEDOUT)
-	{
-	  ep = vl_msg_api_alloc (sizeof (*ep));
-	  ep->_vl_msg_id = ntohs(VL_API_MEMCLNT_READ_TIMEOUT);
-	  vl_msg_api_send_shmem(am->vl_input_queue, (u8 *)&ep);
-	}
-    }
-  pthread_exit(0);
-}
-
-void
-pneum_rx_suspend (void)
-{
-  api_main_t *am = &api_main;
-  pneum_main_t *pm = &pneum_main;
-  vl_api_memclnt_rx_thread_suspend_t *ep;
-
-  if (!pm->rx_thread_handle) return;
-  pthread_mutex_lock(&pm->queue_lock);
-  if (rx_is_running)
-    {
-      ep = vl_msg_api_alloc (sizeof (*ep));
-      ep->_vl_msg_id = ntohs(VL_API_MEMCLNT_RX_THREAD_SUSPEND);
-      vl_msg_api_send_shmem(am->vl_input_queue, (u8 *)&ep);
-      /* Wait for RX thread to tell us it has suspendend */
-      pthread_cond_wait(&pm->suspend_cv, &pm->queue_lock);
-      rx_is_running = false;
-    }
-  pthread_mutex_unlock(&pm->queue_lock);
-}
-
-void
-pneum_rx_resume (void)
-{
-  pneum_main_t *pm = &pneum_main;
-  if (!pm->rx_thread_handle) return;
-  pthread_mutex_lock(&pm->queue_lock);
-  if (rx_is_running) return;
-  pthread_cond_signal(&pm->resume_cv);
-  rx_is_running = true;
-  pthread_mutex_unlock(&pm->queue_lock);
-}
-
-uword *
-pneum_msg_table_get_hash (void)
-{
-  api_main_t *am = &api_main;
-  return (am->msg_index_by_name_and_crc);
-}
-
-int
-pneum_msg_table_size(void)
-{
-  api_main_t *am = &api_main;
-  return hash_elts(am->msg_index_by_name_and_crc);
-}
-
-int
-pneum_connect (char * name, char * chroot_prefix, pneum_callback_t cb,
-               int rx_qlen)
-{
-  int rv = 0;
-  pneum_main_t *pm = &pneum_main;
-
-  init();
-  if (chroot_prefix != NULL)
-    vl_set_memory_root_path (chroot_prefix);
-
-  if ((rv = vl_client_api_map("/vpe-api"))) {
-    clib_warning ("vl_client_api map rv %d", rv);
-    return rv;
-  }
-
-  if (vl_client_connect(name, 0, rx_qlen) < 0) {
-    vl_client_api_unmap();
-    return (-1);
-  }
-
-  if (cb) {
-    /* Start the rx queue thread */
-    rv = pthread_create(&pm->rx_thread_handle, NULL, pneum_rx_thread_fn, 0);
-    if (rv) {
-      clib_warning("pthread_create returned %d", rv);
-      vl_client_api_unmap();
-      return (-1);
-    }
-    pneum_callback = cb;
-    rx_is_running = true;
-  }
-
-  /* Start read timeout thread */
-  rv = pthread_create(&pm->timeout_thread_handle, NULL,
-		      pneum_timeout_thread_fn, 0);
-  if (rv) {
-    clib_warning("pthread_create returned %d", rv);
-    vl_client_api_unmap();
-    return (-1);
-  }
-
-  pm->connected_to_vlib = 1;
-
-  return (0);
-}
-
-int
-pneum_disconnect (void)
-{
-  api_main_t *am = &api_main;
-  pneum_main_t *pm = &pneum_main;
-
-  if (!pm->connected_to_vlib) return 0;
-
-  if (pm->rx_thread_handle) {
-    vl_api_rx_thread_exit_t *ep;
-    uword junk;
-    ep = vl_msg_api_alloc (sizeof (*ep));
-    ep->_vl_msg_id = ntohs(VL_API_RX_THREAD_EXIT);
-    vl_msg_api_send_shmem(am->vl_input_queue, (u8 *)&ep);
-
-    /* wait (with timeout) until RX thread has finished */
-    struct timespec ts;
-    struct timeval tv;
-    gettimeofday(&tv, NULL);
-    ts.tv_sec = tv.tv_sec + 5;
-    ts.tv_nsec = 0;
-    pthread_mutex_lock(&pm->queue_lock);
-    int rv = pthread_cond_timedwait(&pm->terminate_cv, &pm->queue_lock, &ts);
-    pthread_mutex_unlock(&pm->queue_lock);
-    /* now join so we wait until thread has -really- finished */
-    if (rv == ETIMEDOUT)
-      pthread_cancel(pm->rx_thread_handle);
-    else
-      pthread_join(pm->rx_thread_handle, (void **) &junk);
-  }
-  if (pm->timeout_thread_handle)
-    pthread_cancel(pm->timeout_thread_handle);
-
-  vl_client_disconnect();
-  vl_client_api_unmap();
-  pneum_callback = 0;
-
-  cleanup();
-
-  return (0);
-}
-
-static void
-set_timeout (unsigned short timeout)
-{
-  pneum_main_t *pm = &pneum_main;
-  pthread_mutex_lock(&pm->timeout_lock);
-  read_timeout = timeout;
-  pthread_cond_signal(&pm->timeout_cv);
-  pthread_mutex_unlock(&pm->timeout_lock);
-}
-
-static void
-unset_timeout (void)
-{
-  pneum_main_t *pm = &pneum_main;
-  pthread_mutex_lock(&pm->timeout_lock);
-  pthread_cond_signal(&pm->timeout_cancel_cv);
-  pthread_mutex_unlock(&pm->timeout_lock);
-}
-
-int
-pneum_read (char **p, int *l, u16 timeout)
-{
-  unix_shared_memory_queue_t *q;
-  api_main_t *am = &api_main;
-  pneum_main_t *pm = &pneum_main;
-  uword msg;
-  msgbuf_t *msgbuf;
-
-  if (!pm->connected_to_vlib) return -1;
-
-  *l = 0;
-
-  if (am->our_pid == 0) return (-1);
-
-  /* Poke timeout thread */
-  if (timeout)
-    set_timeout(timeout);
-
-  q = am->vl_input_queue;
-  int rv = unix_shared_memory_queue_sub(q, (u8 *)&msg, 0);
-  if (rv == 0) {
-    u16 msg_id = ntohs(*((u16 *)msg));
-    switch (msg_id) {
-    case VL_API_RX_THREAD_EXIT:
-      printf("Received thread exit\n");
-      return -1;
-    case VL_API_MEMCLNT_RX_THREAD_SUSPEND:
-      printf("Received thread suspend\n");
-      goto error;
-    case VL_API_MEMCLNT_READ_TIMEOUT:
-      printf("Received read timeout %ds\n", timeout);
-      goto error;
-
-    default:
-      msgbuf = (msgbuf_t *)(((u8 *)msg) - offsetof(msgbuf_t, data));
-      *l = ntohl(msgbuf->data_len);
-      if (*l == 0) {
-	printf("Unregistered API message: %d\n", msg_id);
-	goto error;
-      }
-    }
-    *p = (char *)msg;
-
-    /* Let timeout notification thread know we're done */
-    unset_timeout();
-
-  } else {
-    printf("Read failed with %d\n", rv);
-  }
-  return (rv);
-
- error:
-  vl_msg_api_free((void *) msg);
-  /* Client might forget to resume RX thread on failure */
-  pneum_rx_resume ();
-  return -1;
-}
-
-/*
- * XXX: Makes the assumption that client_index is the first member
- */
-typedef VL_API_PACKED(struct _vl_api_header {
-  u16 _vl_msg_id;
-  u32 client_index;
-}) vl_api_header_t;
-
-static unsigned int
-pneum_client_index (void)
-{
-  return (api_main.my_client_index);
-}
-
-int
-pneum_write (char *p, int l)
-{
-  int rv = -1;
-  api_main_t *am = &api_main;
-  vl_api_header_t *mp = vl_msg_api_alloc(l);
-  unix_shared_memory_queue_t *q;
-  pneum_main_t *pm = &pneum_main;
-
-  if (!pm->connected_to_vlib) return -1;
-  if (!mp) return (-1);
-
-  memcpy(mp, p, l);
-  mp->client_index = pneum_client_index();
-  q = am->shmem_hdr->vl_input_queue;
-  rv = unix_shared_memory_queue_add(q, (u8 *)&mp, 0);
-  if (rv != 0) {
-    clib_warning("vpe_api_write fails: %d\n", rv);
-    /* Clear message */
-    pneum_free(mp);
-  }
-  return (rv);
-}
-
-uint32_t
-pneum_get_msg_index (unsigned char * name)
-{
-  return vl_api_get_msg_index (name);
-}
diff --git a/src/vpp-api/python/pneum/pneum.h b/src/vpp-api/python/pneum/pneum.h
deleted file mode 100644
index c4b55ae0..00000000
--- a/src/vpp-api/python/pneum/pneum.h
+++ /dev/null
@@ -1,34 +0,0 @@
-/*
- * Copyright (c) 2016 Cisco and/or its affiliates.
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at:
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-#ifndef included_pneum_h
-#define included_pneum_h
-
-#include 
-#include 
-
-typedef void (*pneum_callback_t)(unsigned char * data, int len);
-int pneum_connect(char * name, char * chroot_prefix, pneum_callback_t cb,
-    int rx_qlen);
-int pneum_disconnect(void);
-int pneum_read(char **data, int *l, unsigned short timeout);
-int pneum_write(char *data, int len);
-void pneum_free(void * msg);
-uword * pneum_msg_table_get_hash (void);
-int pneum_msg_table_size(void);
-uint32_t pneum_get_msg_index(unsigned char * name);
-void pneum_rx_suspend (void);
-void pneum_rx_resume (void);
-
-#endif
diff --git a/src/vpp-api/python/pneum/test_pneum.c b/src/vpp-api/python/pneum/test_pneum.c
deleted file mode 100644
index 334e58e9..00000000
--- a/src/vpp-api/python/pneum/test_pneum.c
+++ /dev/null
@@ -1,143 +0,0 @@
-/*
- *------------------------------------------------------------------
- * test_pneum.c
- * 
- * Copyright (c) 2016 Cisco and/or its affiliates.
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at:
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License. 
- *------------------------------------------------------------------
- */
-#include 
-#include 
-#include 
-#include 
-#include 
-#include 
-#include 
-#include 
-
-#include        /* time_t, time (for timestamp in second) */
-#include   /* ftime, timeb (for timestamp in millisecond) */
-#include    /* gettimeofday, timeval (for timestamp in microsecond) */
-
-#include 
-#include 
-#include 
-#include 
-#include 
-#include 
-
-#include 
-#include 
-#include 
-#include "pneum.h"
-
-#define vl_typedefs             /* define message structures */
-#include  
-#undef vl_typedefs
-
-/* we are not linking with vlib */
-vlib_main_t vlib_global_main;
-vlib_main_t **vlib_mains;
-
-volatile int sigterm_received = 0;
-volatile u32 result_ready;
-volatile u16 result_msg_id;
-
-/* M_NOALLOC: construct, but don't yet send a message */
-
-#define M_NOALLOC(T,t)                          \
-  do {						\
-    result_ready = 0;                           \
-    memset (mp, 0, sizeof (*mp));		\
-    mp->_vl_msg_id = ntohs (VL_API_##T);	\
-    mp->client_index = am->my_client_index;	\
-  } while(0);
-
-
-
-int
-wrap_pneum_callback (char *data, int len)
-{
-  //printf("Callback %d\n", len);
-  result_ready = 1;
-  result_msg_id = ntohs(*((u16 *)data));
-  return (0);
-}
-
-int main (int argc, char ** argv)
-{
-  api_main_t * am = &api_main;
-  vl_api_show_version_t message;
-  vl_api_show_version_t *mp;
-  int async = 1;
-  int rv = pneum_connect("pneum_client", NULL, NULL, 32 /* rx queue-length*/);
-
-  if (rv != 0) {
-    printf("Connect failed: %d\n", rv);
-    exit(rv);
-  }
- 
- struct timeb timer_msec;
-  long long int timestamp_msec_start; /* timestamp in millisecond. */
-  if (!ftime(&timer_msec)) {
-    timestamp_msec_start = ((long long int) timer_msec.time) * 1000ll + 
-      (long long int) timer_msec.millitm;
-  }
-  else {
-    timestamp_msec_start = -1;
-  }
-
- 
-  /*
-   * Test vpe_api_write and vpe_api_read to send and recv message for an
-   * API 
-   */
-  int i;
-  long int no_msgs = 10000;
-  mp = &message;
-
-  for (i = 0; i < no_msgs; i++) {
-    /* Construct the API message */
-    M_NOALLOC(SHOW_VERSION, show_version);
-    pneum_write((char *)mp, sizeof(*mp));
-#ifndef __COVERITY__
-    /* As given, async is always 1. Shut up Coverity about it */
-    if (!async)
-      while (result_ready == 0);
-#endif
-  }
-  if (async) {
-    vl_api_control_ping_t control;
-    vl_api_control_ping_t *mp;
-    mp = &control;
-    M_NOALLOC(CONTROL_PING, control_ping);
-    pneum_write((char *)mp, sizeof(*mp));
-
-    while (result_msg_id != VL_API_CONTROL_PING_REPLY);
-  }
-
-  long long int timestamp_msec_end; /* timestamp in millisecond. */
-  if (!ftime(&timer_msec)) {
-    timestamp_msec_end = ((long long int) timer_msec.time) * 1000ll + 
-      (long long int) timer_msec.millitm;
-  }
-  else {
-    timestamp_msec_end = -1;
-  }
-  
-  printf("Took %lld msec, %lld msgs/msec \n", (timestamp_msec_end - timestamp_msec_start),
-	 no_msgs/(timestamp_msec_end - timestamp_msec_start));
-  fformat(stdout, "Exiting...\n");
-  pneum_disconnect();
-  exit (0);
-}
diff --git a/src/vpp-api/python/setup.cfg b/src/vpp-api/python/setup.cfg
index d645be77..79bc6784 100644
--- a/src/vpp-api/python/setup.cfg
+++ b/src/vpp-api/python/setup.cfg
@@ -2,4 +2,4 @@
 # This flag says that the code is written to work on both Python 2 and Python
 # 3. If at all possible, it is good practice to do this. If you cannot, you
 # will need to generate wheels for each Python version that you support.
-universal=0
+universal=1
diff --git a/src/vpp-api/python/setup.py b/src/vpp-api/python/setup.py
index 8a34d501..28c2ecce 100644
--- a/src/vpp-api/python/setup.py
+++ b/src/vpp-api/python/setup.py
@@ -13,23 +13,18 @@
 # limitations under the License.
 
 try:
-    from setuptools import setup, Extension
+    from setuptools import setup
 except ImportError:
-    from distutils.core import setup, Extension
+    from distutils.core import setup
 
 setup (name = 'vpp_papi',
-       version = '1.3',
+       version = '1.4',
        description = 'VPP Python binding',
        author = 'Ole Troan',
        author_email = 'ot@cisco.com',
        test_suite = 'tests',
-       packages=['vpp_papi'],
-       ext_modules = [
-           Extension(
-               'vpp_api',
-               sources = ['vpp_papi/pneum_wrap.c'],
-               libraries = ['pneum'],
-           )],
+       install_requires=['cffi'],
+       py_modules=['vpp_papi'],
        long_description = '''VPP Python language binding.''',
        zip_safe = True,
 )
diff --git a/src/vpp-api/python/vpp_papi.py b/src/vpp-api/python/vpp_papi.py
new file mode 100644
index 00000000..81f6903b
--- /dev/null
+++ b/src/vpp-api/python/vpp_papi.py
@@ -0,0 +1,635 @@
+#!/usr/bin/env python
+#
+# Copyright (c) 2016 Cisco and/or its affiliates.
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at:
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+#
+
+from __future__ import print_function
+import sys, os, logging, collections, struct, json, threading, glob
+import atexit, Queue
+
+from cffi import FFI
+ffi = FFI()
+ffi.cdef("""
+typedef void (*pneum_callback_t)(unsigned char * data, int len);
+typedef void (*pneum_error_callback_t)(void *, unsigned char *, int);
+int pneum_connect(char * name, char * chroot_prefix, pneum_callback_t cb,
+    int rx_qlen);
+int pneum_disconnect(void);
+int pneum_read(char **data, int *l, unsigned short timeout);
+int pneum_write(char *data, int len);
+void pneum_free(void * msg);
+
+int pneum_get_msg_index(unsigned char * name);
+int pneum_msg_table_size(void);
+int pneum_msg_table_max_index(void);
+
+void pneum_rx_suspend (void);
+void pneum_rx_resume (void);
+void pneum_set_error_handler(pneum_error_callback_t);
+ """)
+
+# Barfs on failure, no need to check success.
+vpp_api = ffi.dlopen('libpneum.so')
+
+def vpp_atexit(self):
+    """Clean up VPP connection on shutdown."""
+    if self.connected:
+        self.logger.debug('Cleaning up VPP on exit')
+        self.disconnect()
+
+vpp_object = None
+
+@ffi.callback("void(unsigned char *, int)")
+def pneum_callback_sync(data, len):
+    vpp_object.msg_handler_sync(ffi.buffer(data, len))
+@ffi.callback("void(unsigned char *, int)")
+def pneum_callback_async(data, len):
+    vpp_object.msg_handler_async(ffi.buffer(data, len))
+@ffi.callback("void(void *, unsigned char *, int)")
+def pneum_error_handler(arg, msg, msg_len):
+    vpp_object.logger.warning("PNEUM: %s", ffi.string(msg, msg_len))
+
+class Empty(object):
+    pass
+
+
+class FuncWrapper(object):
+    def __init__(self, func):
+        self._func = func
+        self.__name__ = func.__name__
+
+    def __call__(self, **kwargs):
+        return self._func(**kwargs)
+
+
+class VPP():
+    """VPP interface.
+
+    This class provides the APIs to VPP.  The APIs are loaded
+    from provided .api.json files and makes functions accordingly.
+    These functions are documented in the VPP .api files, as they
+    are dynamically created.
+
+    Additionally, VPP can send callback messages; this class
+    provides a means to register a callback function to receive
+    these messages in a background thread.
+    """
+    def __init__(self, apifiles = None, testmode = False, async_thread = True,
+                 logger = logging.getLogger('vpp_papi'), loglevel = 'debug'):
+        """Create a VPP API object.
+
+        apifiles is a list of files containing API
+        descriptions that will be loaded - methods will be
+        dynamically created reflecting these APIs.  If not
+        provided this will load the API files from VPP's
+        default install location.
+        """
+        global vpp_object
+        vpp_object = self
+        self.logger = logger
+        logging.basicConfig(level=getattr(logging, loglevel.upper()))
+
+        self.messages = {}
+        self.id_names = []
+        self.id_msgdef = []
+        self.buffersize = 10000
+        self.connected = False
+        self.header = struct.Struct('>HI')
+        self.apifiles = []
+        self.event_callback = None
+        self.message_queue = Queue.Queue()
+        self.read_timeout = 0
+        self.vpp_api = vpp_api
+        if async_thread:
+            self.event_thread = threading.Thread(target=self.thread_msg_handler)
+            self.event_thread.daemon = True
+            self.event_thread.start()
+
+        if not apifiles:
+            # Pick up API definitions from default directory
+            apifiles = glob.glob('/usr/share/vpp/api/*.api.json')
+
+        for file in apifiles:
+            with open(file) as apidef_file:
+                api = json.load(apidef_file)
+                for t in api['types']:
+                    self.add_type(t[0], t[1:])
+
+                for m in api['messages']:
+                    self.add_message(m[0], m[1:])
+	self.apifiles = apifiles
+
+        # Basic sanity check
+        if len(self.messages) == 0 and not testmode:
+            raise ValueError(1, 'Missing JSON message definitions')
+
+        # Make sure we allow VPP to clean up the message rings.
+        atexit.register(vpp_atexit, self)
+
+        # Register error handler
+        vpp_api.pneum_set_error_handler(pneum_error_handler)
+
+    class ContextId(object):
+        """Thread-safe provider of unique context IDs."""
+        def __init__(self):
+            self.context = 0
+	    self.lock = threading.Lock()
+        def __call__(self):
+            """Get a new unique (or, at least, not recently used) context."""
+	    with self.lock:
+		self.context += 1
+		return self.context
+    get_context = ContextId()
+
+    def status(self):
+        """Debug function: report current VPP API status to stdout."""
+        print('Connected') if self.connected else print('Not Connected')
+        print('Read API definitions from', ', '.join(self.apifiles))
+
+    def __struct (self, t, n = None, e = -1, vl = None):
+        """Create a packing structure for a message."""
+        base_types = { 'u8' : 'B',
+                       'u16' : 'H',
+                       'u32' : 'I',
+                       'i32' : 'i',
+                       'u64' : 'Q',
+                       'f64' : 'd',
+                       }
+        pack = None
+        if t in base_types:
+            pack = base_types[t]
+            if not vl:
+                if e > 0 and t == 'u8':
+                    # Fixed byte array
+                    return struct.Struct('>' + str(e) + 's')
+                if e > 0:
+                    # Fixed array of base type
+                    return [e, struct.Struct('>' + base_types[t])]
+                elif e == 0:
+                    # Old style variable array
+                    return [-1, struct.Struct('>' + base_types[t])]
+            else:
+                # Variable length array
+                return [vl, struct.Struct('>s')] if t == 'u8' else \
+                    [vl, struct.Struct('>' + base_types[t])]
+
+            return struct.Struct('>' + base_types[t])
+
+        if t in self.messages:
+            ### Return a list in case of array ###
+            if e > 0 and not vl:
+                return [e, lambda self, encode, buf, offset, args: (
+                    self.__struct_type(encode, self.messages[t], buf, offset,
+                                       args))]
+            if vl:
+                return [vl, lambda self, encode, buf, offset, args: (
+                    self.__struct_type(encode, self.messages[t], buf, offset,
+                                       args))]
+            elif e == 0:
+                # Old style VLA
+                raise NotImplementedError(1, 'No support for compound types ' + t)
+            return lambda self, encode, buf, offset, args: (
+                self.__struct_type(encode, self.messages[t], buf, offset, args)
+            )
+
+        raise ValueError(1, 'Invalid message type: ' + t)
+
+    def __struct_type(self, encode, msgdef, buf, offset, kwargs):
+        """Get a message packer or unpacker."""
+        if encode:
+            return self.__struct_type_encode(msgdef, buf, offset, kwargs)
+        else:
+            return self.__struct_type_decode(msgdef, buf, offset)
+
+    def __struct_type_encode(self, msgdef, buf, offset, kwargs):
+        off = offset
+        size = 0
+
+        for k in kwargs:
+            if k not in msgdef['args']:
+                raise ValueError(1, 'Invalid field-name in message call ' + k)
+
+        for k,v in msgdef['args'].iteritems():
+            off += size
+            if k in kwargs:
+                if type(v) is list:
+                    if callable(v[1]):
+                        e = kwargs[v[0]] if v[0] in kwargs else v[0]
+                        size = 0
+                        for i in range(e):
+                            size += v[1](self, True, buf, off + size,
+                                         kwargs[k][i])
+                    else:
+                        if v[0] in kwargs:
+                            l = kwargs[v[0]]
+                        else:
+                            l = len(kwargs[k])
+                        if v[1].size == 1:
+                            buf[off:off + l] = bytearray(kwargs[k])
+                            size = l
+                        else:
+                            size = 0
+                            for i in kwargs[k]:
+                                v[1].pack_into(buf, off + size, i)
+                                size += v[1].size
+                else:
+                    if callable(v):
+                        size = v(self, True, buf, off, kwargs[k])
+                    else:
+                        v.pack_into(buf, off, kwargs[k])
+                        size = v.size
+            else:
+                size = v.size if not type(v) is list else 0
+
+        return off + size - offset
+
+
+    def __getitem__(self, name):
+        if name in self.messages:
+            return self.messages[name]
+        return None
+
+    def encode(self, msgdef, kwargs):
+        # Make suitably large buffer
+        buf = bytearray(self.buffersize)
+        offset = 0
+        size = self.__struct_type(True, msgdef, buf, offset, kwargs)
+        return buf[:offset + size]
+
+    def decode(self, msgdef, buf):
+        return self.__struct_type(False, msgdef, buf, 0, None)[1]
+
+    def __struct_type_decode(self, msgdef, buf, offset):
+        res = []
+        off = offset
+        size = 0
+        for k,v in msgdef['args'].iteritems():
+            off += size
+            if type(v) is list:
+                lst = []
+                if callable(v[1]): # compound type
+                    size = 0
+                    if v[0] in msgdef['args']: # vla
+                        e = res[v[2]]
+                    else: # fixed array
+                        e = v[0]
+                    res.append(lst)
+                    for i in range(e):
+                        (s,l) = v[1](self, False, buf, off + size, None)
+                        lst.append(l)
+                        size += s
+                    continue
+                if v[1].size == 1:
+                    if type(v[0]) is int:
+                        size = len(buf) - off
+                    else:
+                        size = res[v[2]]
+                    res.append(buf[off:off + size])
+                else:
+                    e = v[0] if type(v[0]) is int else res[v[2]]
+                    if e == -1:
+                        e = (len(buf) - off) / v[1].size
+                    lst = []
+                    res.append(lst)
+                    size = 0
+                    for i in range(e):
+                        lst.append(v[1].unpack_from(buf, off + size)[0])
+                        size += v[1].size
+            else:
+                if callable(v):
+                    (s,l) = v(self, False, buf, off, None)
+                    res.append(l)
+                    size += s
+                else:
+                    res.append(v.unpack_from(buf, off)[0])
+                    size = v.size
+
+        return off + size - offset, msgdef['return_tuple']._make(res)
+
+    def ret_tup(self, name):
+        if name in self.messages and 'return_tuple' in self.messages[name]:
+            return self.messages[name]['return_tuple']
+        return None
+
+    def add_message(self, name, msgdef, typeonly = False):
+        if name in self.messages:
+            raise ValueError('Duplicate message name: ' + name)
+
+        args = collections.OrderedDict()
+        argtypes = collections.OrderedDict()
+        fields = []
+        msg = {}
+        for i, f in enumerate(msgdef):
+            if type(f) is dict and 'crc' in f:
+                msg['crc'] = f['crc']
+                continue
+            field_type = f[0]
+            field_name = f[1]
+            if len(f) == 3 and f[2] == 0 and i != len(msgdef) - 2:
+                raise ValueError('Variable Length Array must be last: ' + name)
+            args[field_name] = self.__struct(*f)
+            argtypes[field_name] = field_type
+            if len(f) == 4: # Find offset to # elements field
+                args[field_name].append(args.keys().index(f[3]) - i)
+            fields.append(field_name)
+        msg['return_tuple'] = collections.namedtuple(name, fields,
+                                                     rename = True)
+        self.messages[name] = msg
+        self.messages[name]['args'] = args
+        self.messages[name]['argtypes'] = argtypes
+        self.messages[name]['typeonly'] = typeonly
+        return self.messages[name]
+
+    def add_type(self, name, typedef):
+        return self.add_message('vl_api_' + name + '_t', typedef, typeonly=True)
+
+    def make_function(self, name, i, msgdef, multipart, async):
+        if (async):
+            f = lambda **kwargs: (self._call_vpp_async(i, msgdef, **kwargs))
+        else:
+            f = lambda **kwargs: (self._call_vpp(i, msgdef, multipart, **kwargs))
+        args = self.messages[name]['args']
+        argtypes = self.messages[name]['argtypes']
+        f.__name__ = str(name)
+        f.__doc__ = ", ".join(["%s %s" % (argtypes[k], k) for k in args.keys()])
+        return f
+
+    @property
+    def api(self):
+        if not hasattr(self, "_api"):
+            raise Exception("Not connected, api definitions not available")
+        return self._api
+
+    def _register_functions(self, async=False):
+        self.id_names = [None] * (self.vpp_dictionary_maxid + 1)
+        self.id_msgdef = [None] * (self.vpp_dictionary_maxid + 1)
+        self._api = Empty()
+        for name, msgdef in self.messages.iteritems():
+            if self.messages[name]['typeonly']: continue
+            crc = self.messages[name]['crc']
+            n = name + '_' + crc[2:]
+            i = vpp_api.pneum_get_msg_index(bytes(n))
+            if i > 0:
+                self.id_msgdef[i] = msgdef
+                self.id_names[i] = name
+                multipart = True if name.find('_dump') > 0 else False
+                f = self.make_function(name, i, msgdef, multipart, async)
+                setattr(self._api, name, FuncWrapper(f))
+
+                # old API stuff starts here - will be removed in 17.07
+                if hasattr(self, name):
+                    raise NameError(
+                        3, "Conflicting name in JSON definition: `%s'" % name)
+                setattr(self, name, f)
+                # old API stuff ends here
+            else:
+                self.logger.debug('No such message type or failed CRC checksum: %s', n)
+
+    def _write (self, buf):
+        """Send a binary-packed message to VPP."""
+        if not self.connected:
+            raise IOError(1, 'Not connected')
+        return vpp_api.pneum_write(str(buf), len(buf))
+
+    def _read (self):
+        if not self.connected:
+            raise IOError(1, 'Not connected')
+        mem = ffi.new("char **")
+        size = ffi.new("int *")
+        rv = vpp_api.pneum_read(mem, size, self.read_timeout)
+        if rv:
+            raise IOError(rv, 'pneum_read filed')
+        msg = bytes(ffi.buffer(mem[0], size[0]))
+        vpp_api.pneum_free(mem[0])
+        return msg
+
+    def connect_internal(self, name, msg_handler, chroot_prefix, rx_qlen, async):
+	rv = vpp_api.pneum_connect(name, chroot_prefix, msg_handler, rx_qlen)
+        if rv != 0:
+            raise IOError(2, 'Connect failed')
+        self.connected = True
+
+        self.vpp_dictionary_maxid = vpp_api.pneum_msg_table_max_index()
+        self._register_functions(async=async)
+
+        # Initialise control ping
+        crc = self.messages['control_ping']['crc']
+        self.control_ping_index = \
+                                  vpp_api.pneum_get_msg_index(
+                                      bytes('control_ping' + '_' + crc[2:]))
+        self.control_ping_msgdef = self.messages['control_ping']
+
+    def connect(self, name, chroot_prefix = ffi.NULL,
+                async = False, rx_qlen = 32):
+        """Attach to VPP.
+
+        name - the name of the client.
+        chroot_prefix - if VPP is chroot'ed, the prefix of the jail
+        async - if true, messages are sent without waiting for a reply
+        rx_qlen - the length of the VPP message receive queue between
+        client and server.
+        """
+        msg_handler = pneum_callback_sync if not async \
+                      else pneum_callback_async
+        return self.connect_internal(name, msg_handler, chroot_prefix, rx_qlen,
+                                     async)
+
+    def connect_sync (self, name, chroot_prefix = ffi.NULL, rx_qlen = 32):
+        """Attach to VPP in synchronous mode. Application must poll for events.
+
+        name - the name of the client.
+        chroot_prefix - if VPP is chroot'ed, the prefix of the jail
+        rx_qlen - the length of the VPP message receive queue between
+        client and server.
+        """
+
+        return self.connect_internal(name, ffi.NULL, chroot_prefix, rx_qlen,
+                                     async=False)
+
+    def disconnect(self):
+        """Detach from VPP."""
+        rv = vpp_api.pneum_disconnect()
+        self.connected = False
+        return rv
+
+    def msg_handler_sync(self, msg):
+        """Process an incoming message from VPP in sync mode.
+
+        The message may be a reply or it may be an async notification.
+        """
+        r = self.decode_incoming_msg(msg)
+        if r is None:
+            return
+
+        # If we have a context, then use the context to find any
+        # request waiting for a reply
+        context = 0
+        if hasattr(r, 'context') and r.context > 0:
+            context = r.context
+
+        msgname = type(r).__name__
+
+        if context == 0:
+            # No context -> async notification that we feed to the callback
+            self.message_queue.put_nowait(r)
+        else:
+            raise IOError(2, 'RPC reply message received in event handler')
+
+    def decode_incoming_msg(self, msg):
+        if not msg:
+            self.logger.warning('vpp_api.read failed')
+            return
+
+        i, ci = self.header.unpack_from(msg, 0)
+        if self.id_names[i] == 'rx_thread_exit':
+            return
+
+        #
+        # Decode message and returns a tuple.
+        #
+        msgdef = self.id_msgdef[i]
+        if not msgdef:
+            raise IOError(2, 'Reply message undefined')
+
+        r = self.decode(msgdef, msg)
+
+        return r
+
+    def msg_handler_async(self, msg):
+        """Process a message from VPP in async mode.
+
+        In async mode, all messages are returned to the callback.
+        """
+        r = self.decode_incoming_msg(msg)
+        if r is None:
+            return
+
+        msgname = type(r).__name__
+
+	if self.event_callback:
+	    self.event_callback(msgname, r)
+
+    def _control_ping(self, context):
+        """Send a ping command."""
+        self._call_vpp_async(self.control_ping_index,
+			     self.control_ping_msgdef,
+                             context=context)
+
+    def _call_vpp(self, i, msgdef, multipart, **kwargs):
+        """Given a message, send the message and await a reply.
+
+        msgdef - the message packing definition
+        i - the message type index
+        multipart - True if the message returns multiple
+        messages in return.
+        context - context number - chosen at random if not
+        supplied.
+        The remainder of the kwargs are the arguments to the API call.
+
+        The return value is the message or message array containing
+        the response.  It will raise an IOError exception if there was
+        no response within the timeout window.
+        """
+
+        if not 'context' in kwargs:
+            context = self.get_context()
+            kwargs['context'] = context
+        else:
+            context = kwargs['context']
+        kwargs['_vl_msg_id'] = i
+        b = self.encode(msgdef, kwargs)
+
+        vpp_api.pneum_rx_suspend()
+        self._write(b)
+
+        if multipart:
+            # Send a ping after the request - we use its response
+            # to detect that we have seen all results.
+            self._control_ping(context)
+
+        # Block until we get a reply.
+        rl = []
+        while (True):
+            msg = self._read()
+            if not msg:
+                print('PNEUM ERROR: OH MY GOD')
+                raise IOError(2, 'PNEUM read failed')
+
+            r = self.decode_incoming_msg(msg)
+            msgname = type(r).__name__
+            if not context in r or r.context == 0 or context != r.context:
+                self.message_queue.put_nowait(r)
+                continue
+
+            if not multipart:
+                rl = r
+                break
+            if msgname == 'control_ping_reply':
+                break
+
+            rl.append(r)
+
+        vpp_api.pneum_rx_resume()
+
+        return rl
+
+    def _call_vpp_async(self, i, msgdef, **kwargs):
+        """Given a message, send the message and await a reply.
+
+        msgdef - the message packing definition
+        i - the message type index
+        context - context number - chosen at random if not
+        supplied.
+        The remainder of the kwargs are the arguments to the API call.
+        """
+        if not 'context' in kwargs:
+            context = self.get_context()
+            kwargs['context'] = context
+        else:
+            context = kwargs['context']
+        kwargs['_vl_msg_id'] = i
+        b = self.encode(msgdef, kwargs)
+
+        self._write(b)
+
+    def register_event_callback(self, callback):
+        """Register a callback for async messages.
+
+        This will be called for async notifications in sync mode,
+        and all messages in async mode.  In sync mode, replies to
+        requests will not come here.
+
+        callback is a fn(msg_type_name, msg_type) that will be
+        called when a message comes in.  While this function is
+        executing, note that (a) you are in a background thread and
+        may wish to use threading.Lock to protect your datastructures,
+        and (b) message processing from VPP will stop (so if you take
+        a long while about it you may provoke reply timeouts or cause
+        VPP to fill the RX buffer).  Passing None will disable the
+        callback.
+        """
+        self.event_callback = callback
+
+    def thread_msg_handler(self):
+        """Python thread calling the user registerd message handler.
+
+        This is to emulate the old style event callback scheme. Modern
+        clients should provide their own thread to poll the event
+        queue.
+        """
+        while True:
+            r = self.message_queue.get()
+            msgname = type(r).__name__
+	    if self.event_callback:
+	        self.event_callback(msgname, r)
diff --git a/src/vpp-api/python/vpp_papi/__init__.py b/src/vpp-api/python/vpp_papi/__init__.py
deleted file mode 100644
index 6688ffb8..00000000
--- a/src/vpp-api/python/vpp_papi/__init__.py
+++ /dev/null
@@ -1,3 +0,0 @@
-__import__('pkg_resources').declare_namespace(__name__)
-from . vpp_papi import *
-
diff --git a/src/vpp-api/python/vpp_papi/pneum_wrap.c b/src/vpp-api/python/vpp_papi/pneum_wrap.c
deleted file mode 100644
index c5a7eea1..00000000
--- a/src/vpp-api/python/vpp_papi/pneum_wrap.c
+++ /dev/null
@@ -1,226 +0,0 @@
-/*
- * Copyright (c) 2016 Cisco and/or its affiliates.
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at:
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#include 
-#include "../pneum/pneum.h"
-#include 
-
-static PyObject *pneum_callback = NULL;
-
-static void
-wrap_pneum_callback (unsigned char * data, int len)
-{
-  PyGILState_STATE gstate;
-  PyObject *result;//, *arglist;
-
-  gstate = PyGILState_Ensure();
-
-  /* Time to call the callback */
-#if PY_VERSION_HEX >= 0x03000000
-  result = PyObject_CallFunction(pneum_callback, "y#", data, len);
-#else
-  result = PyObject_CallFunction(pneum_callback, "s#", data, len);
-#endif
-  if (result)
-    Py_DECREF(result);
-  else
-    PyErr_Print();
-
-  PyGILState_Release(gstate);
-}
-
-static PyObject *
-wrap_connect (PyObject *self, PyObject *args, PyObject *kw)
-{
-  char * name, * chroot_prefix = NULL;
-  int rx_qlen = 32; /* default rx queue length */
-  int rv;
-  PyObject * temp = NULL;
-  pneum_callback_t cb = NULL;
-
-  if (!PyArg_ParseTuple(args, "sOzi:wrap_connect",
-			&name, &temp, &chroot_prefix, &rx_qlen))
-    return (NULL);
-
-  if (temp != Py_None)
-    {
-      if (!PyCallable_Check(temp))
-	{
-	  PyErr_SetString(PyExc_TypeError, "parameter must be callable");
-	  return NULL;
-	}
-
-      Py_XINCREF(temp);         /* Add a reference to new callback */
-      Py_XDECREF(pneum_callback);  /* Dispose of previous callback */
-      pneum_callback = temp;       /* Remember new callback */
-      cb = wrap_pneum_callback;
-    }
-  Py_BEGIN_ALLOW_THREADS
-  rv = pneum_connect(name, chroot_prefix, cb, rx_qlen);
-  Py_END_ALLOW_THREADS
-  return PyLong_FromLong(rv);
-}
-
-static PyObject *
-wrap_disconnect (PyObject *self, PyObject *args)
-{
-  int rv;
-  Py_BEGIN_ALLOW_THREADS
-  rv = pneum_disconnect();
-  Py_END_ALLOW_THREADS
-  return PyLong_FromLong(rv);
-}
-
-static PyObject *
-wrap_write (PyObject *self, PyObject *args)
-{
-  char *data;
-  int len, rv;
-
-  if (!PyArg_ParseTuple(args, "s#", &data, &len))
-    return NULL;
-
-  Py_BEGIN_ALLOW_THREADS
-  rv = pneum_write(data, len);
-  Py_END_ALLOW_THREADS
-
-  return PyLong_FromLong(rv);
-}
-
-static PyObject *
-wrap_read (PyObject *self, PyObject *args)
-{
-  char *data;
-  int len, rv;
-  unsigned short timeout;
-
-  if (!PyArg_ParseTuple(args, "H", &timeout))
-    return (NULL);
-  Py_BEGIN_ALLOW_THREADS
-  rv = pneum_read(&data, &len, timeout);
-  Py_END_ALLOW_THREADS
-
-  if (rv != 0) { Py_RETURN_NONE; }
-#if PY_VERSION_HEX >= 0x03000000
-  PyObject *ret = Py_BuildValue("y#", data, len);
-#else
-  PyObject *ret = Py_BuildValue("s#", data, len);
-#endif
-  pneum_free(data);
-  if (!ret) { Py_RETURN_NONE; }
-
-  return ret;
-}
-
-static PyObject *
-wrap_msg_table (PyObject *self, PyObject *args)
-{
-  int i = 0, rv = 0;
-  hash_pair_t *hp;
-  uword *h = pneum_msg_table_get_hash();
-  PyObject *ret = PyList_New(pneum_msg_table_size());
-  if (!ret) goto error;
-  hash_foreach_pair (hp, h,
-  ({
-    PyObject *item = PyTuple_New(2);
-    if (!item) goto error;
-    rv = PyTuple_SetItem(item, 0, PyLong_FromLong((u32)hp->value[0]));
-    if (rv) goto error;
-    rv = PyTuple_SetItem(item, 1, PyString_FromString((char *)hp->key));
-    if (rv) goto error;
-    PyList_SetItem(ret, i, item);
-    i++;
-  }));
-
-  return ret;
-
- error:
-  /* TODO: Raise exception */
-  printf("msg_table failed");
-  Py_RETURN_NONE;
-}
-
-static PyObject *
-wrap_suspend (PyObject *self, PyObject *args)
-{
-  Py_BEGIN_ALLOW_THREADS
-  pneum_rx_suspend();
-  Py_END_ALLOW_THREADS
-  Py_RETURN_NONE;
-}
-
-static PyObject *
-wrap_resume (PyObject *self, PyObject *args)
-{
-  Py_BEGIN_ALLOW_THREADS
-  pneum_rx_resume();
-  Py_END_ALLOW_THREADS
-  Py_RETURN_NONE;
-}
-
-static PyMethodDef vpp_api_Methods[] = {
-  {"connect", wrap_connect, METH_VARARGS, "Connect to the VPP API."},
-  {"disconnect", wrap_disconnect, METH_VARARGS, "Disconnect from the VPP API."},
-  {"write", wrap_write, METH_VARARGS, "Write data to the VPP API."},
-  {"read", wrap_read, METH_VARARGS, "Read data from the VPP API."},
-  {"msg_table", wrap_msg_table, METH_VARARGS, "Get API dictionary."},
-  {"suspend", wrap_suspend, METH_VARARGS, "Suspend RX thread."},
-  {"resume", wrap_resume, METH_VARARGS, "Resume RX thread."},
-  {NULL, NULL, 0, NULL}        /* Sentinel */
-};
-
-#if PY_VERSION_HEX >= 0x03000000
-PyMODINIT_FUNC
-PyInit_vpp_api (void)
-#else
-void
-initvpp_api (void)
-#endif
-{
-#if PY_VERSION_HEX >= 0x03000000
-  static struct PyModuleDef vpp_api_module = {
-#if PY_VERSION_HEX >= 0x03020000
-    PyModuleDef_HEAD_INIT,
-#else
-    {
-      PyObject_HEAD_INIT(NULL)
-      NULL, /* m_init */
-      0,    /* m_index */
-      NULL, /* m_copy */
-    },
-#endif
-    (char *) "vpp_api",
-    NULL,
-    -1,
-    vpp_api_Methods,
-    NULL,
-    NULL,
-    NULL,
-    NULL
-  };
-#endif
-
-  /* Ensure threading is initialised */
-  if (!PyEval_ThreadsInitialized()) {
-    PyEval_InitThreads();
-  }
-
-#if PY_VERSION_HEX >= 0x03000000
-  return PyModule_Create(&vpp_api_module);
-#else
-  Py_InitModule((char *) "vpp_api", vpp_api_Methods);
-  return;
-#endif
-}
diff --git a/src/vpp-api/python/vpp_papi/vpp_papi.py b/src/vpp-api/python/vpp_papi/vpp_papi.py
deleted file mode 100644
index 0c40f171..00000000
--- a/src/vpp-api/python/vpp_papi/vpp_papi.py
+++ /dev/null
@@ -1,599 +0,0 @@
-#!/usr/bin/env python
-#
-# Copyright (c) 2016 Cisco and/or its affiliates.
-# Licensed under the Apache License, Version 2.0 (the "License");
-# you may not use this file except in compliance with the License.
-# You may obtain a copy of the License at:
-#
-#     http://www.apache.org/licenses/LICENSE-2.0
-#
-# Unless required by applicable law or agreed to in writing, software
-# distributed under the License is distributed on an "AS IS" BASIS,
-# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-# See the License for the specific language governing permissions and
-# limitations under the License.
-#
-
-from __future__ import print_function
-import sys, os, logging, collections, struct, json, threading, glob
-import atexit, Queue
-
-logging.basicConfig(level=logging.DEBUG)
-import vpp_api
-
-def eprint(*args, **kwargs):
-    """Print critical diagnostics to stderr."""
-    print(*args, file=sys.stderr, **kwargs)
-
-def vpp_atexit(self):
-    """Clean up VPP connection on shutdown."""
-    if self.connected:
-        eprint ('Cleaning up VPP on exit')
-        self.disconnect()
-
-
-class Empty(object):
-    pass
-
-
-class FuncWrapper(object):
-    def __init__(self, func):
-        self._func = func
-        self.__name__ = func.__name__
-
-    def __call__(self, **kwargs):
-        return self._func(**kwargs)
-
-
-class VPP():
-    """VPP interface.
-
-    This class provides the APIs to VPP.  The APIs are loaded
-    from provided .api.json files and makes functions accordingly.
-    These functions are documented in the VPP .api files, as they
-    are dynamically created.
-
-    Additionally, VPP can send callback messages; this class
-    provides a means to register a callback function to receive
-    these messages in a background thread.
-    """
-    def __init__(self, apifiles = None, testmode = False, async_thread = True):
-        """Create a VPP API object.
-
-        apifiles is a list of files containing API
-        descriptions that will be loaded - methods will be
-        dynamically created reflecting these APIs.  If not
-        provided this will load the API files from VPP's
-        default install location.
-        """
-        self.messages = {}
-        self.id_names = []
-        self.id_msgdef = []
-        self.buffersize = 10000
-        self.connected = False
-        self.header = struct.Struct('>HI')
-        self.apifiles = []
-        self.event_callback = None
-        self.message_queue = Queue.Queue()
-        self.read_timeout = 0
-        self.vpp_api = vpp_api
-        if async_thread:
-            self.event_thread = threading.Thread(target=self.thread_msg_handler)
-            self.event_thread.daemon = True
-            self.event_thread.start()
-
-        if not apifiles:
-            # Pick up API definitions from default directory
-            apifiles = glob.glob('/usr/share/vpp/api/*.api.json')
-
-        for file in apifiles:
-            with open(file) as apidef_file:
-                api = json.load(apidef_file)
-                for t in api['types']:
-                    self.add_type(t[0], t[1:])
-
-                for m in api['messages']:
-                    self.add_message(m[0], m[1:])
-	self.apifiles = apifiles
-
-        # Basic sanity check
-        if len(self.messages) == 0 and not testmode:
-            raise ValueError(1, 'Missing JSON message definitions')
-
-        # Make sure we allow VPP to clean up the message rings.
-        atexit.register(vpp_atexit, self)
-
-    class ContextId(object):
-        """Thread-safe provider of unique context IDs."""
-        def __init__(self):
-            self.context = 0
-	    self.lock = threading.Lock()
-        def __call__(self):
-            """Get a new unique (or, at least, not recently used) context."""
-	    with self.lock:
-		self.context += 1
-		return self.context
-    get_context = ContextId()
-
-    def status(self):
-        """Debug function: report current VPP API status to stdout."""
-        print('Connected') if self.connected else print('Not Connected')
-        print('Read API definitions from', ', '.join(self.apifiles))
-
-    def __struct (self, t, n = None, e = -1, vl = None):
-        """Create a packing structure for a message."""
-        base_types = { 'u8' : 'B',
-                       'u16' : 'H',
-                       'u32' : 'I',
-                       'i32' : 'i',
-                       'u64' : 'Q',
-                       'f64' : 'd',
-                       }
-        pack = None
-        if t in base_types:
-            pack = base_types[t]
-            if not vl:
-                if e > 0 and t == 'u8':
-                    # Fixed byte array
-                    return struct.Struct('>' + str(e) + 's')
-                if e > 0:
-                    # Fixed array of base type
-                    return [e, struct.Struct('>' + base_types[t])]
-                elif e == 0:
-                    # Old style variable array
-                    return [-1, struct.Struct('>' + base_types[t])]
-            else:
-                # Variable length array
-                return [vl, struct.Struct('>s')] if t == 'u8' else \
-                    [vl, struct.Struct('>' + base_types[t])]
-
-            return struct.Struct('>' + base_types[t])
-
-        if t in self.messages:
-            ### Return a list in case of array ###
-            if e > 0 and not vl:
-                return [e, lambda self, encode, buf, offset, args: (
-                    self.__struct_type(encode, self.messages[t], buf, offset,
-                                       args))]
-            if vl:
-                return [vl, lambda self, encode, buf, offset, args: (
-                    self.__struct_type(encode, self.messages[t], buf, offset,
-                                       args))]
-            elif e == 0:
-                # Old style VLA
-                raise NotImplementedError(1, 'No support for compound types ' + t)
-            return lambda self, encode, buf, offset, args: (
-                self.__struct_type(encode, self.messages[t], buf, offset, args)
-            )
-
-        raise ValueError(1, 'Invalid message type: ' + t)
-
-    def __struct_type(self, encode, msgdef, buf, offset, kwargs):
-        """Get a message packer or unpacker."""
-        if encode:
-            return self.__struct_type_encode(msgdef, buf, offset, kwargs)
-        else:
-            return self.__struct_type_decode(msgdef, buf, offset)
-
-    def __struct_type_encode(self, msgdef, buf, offset, kwargs):
-        off = offset
-        size = 0
-
-        for k in kwargs:
-            if k not in msgdef['args']:
-                raise ValueError(1, 'Invalid field-name in message call ' + k)
-
-        for k,v in msgdef['args'].iteritems():
-            off += size
-            if k in kwargs:
-                if type(v) is list:
-                    if callable(v[1]):
-                        e = kwargs[v[0]] if v[0] in kwargs else v[0]
-                        size = 0
-                        for i in range(e):
-                            size += v[1](self, True, buf, off + size,
-                                         kwargs[k][i])
-                    else:
-                        if v[0] in kwargs:
-                            l = kwargs[v[0]]
-                        else:
-                            l = len(kwargs[k])
-                        if v[1].size == 1:
-                            buf[off:off + l] = bytearray(kwargs[k])
-                            size = l
-                        else:
-                            size = 0
-                            for i in kwargs[k]:
-                                v[1].pack_into(buf, off + size, i)
-                                size += v[1].size
-                else:
-                    if callable(v):
-                        size = v(self, True, buf, off, kwargs[k])
-                    else:
-                        v.pack_into(buf, off, kwargs[k])
-                        size = v.size
-            else:
-                size = v.size if not type(v) is list else 0
-
-        return off + size - offset
-
-
-    def __getitem__(self, name):
-        if name in self.messages:
-            return self.messages[name]
-        return None
-
-    def encode(self, msgdef, kwargs):
-        # Make suitably large buffer
-        buf = bytearray(self.buffersize)
-        offset = 0
-        size = self.__struct_type(True, msgdef, buf, offset, kwargs)
-        return buf[:offset + size]
-
-    def decode(self, msgdef, buf):
-        return self.__struct_type(False, msgdef, buf, 0, None)[1]
-
-    def __struct_type_decode(self, msgdef, buf, offset):
-        res = []
-        off = offset
-        size = 0
-        for k,v in msgdef['args'].iteritems():
-            off += size
-            if type(v) is list:
-                lst = []
-                if callable(v[1]): # compound type
-                    size = 0
-                    if v[0] in msgdef['args']: # vla
-                        e = res[v[2]]
-                    else: # fixed array
-                        e = v[0]
-                    res.append(lst)
-                    for i in range(e):
-                        (s,l) = v[1](self, False, buf, off + size, None)
-                        lst.append(l)
-                        size += s
-                    continue
-                if v[1].size == 1:
-                    if type(v[0]) is int:
-                        size = len(buf) - off
-                    else:
-                        size = res[v[2]]
-                    res.append(buf[off:off + size])
-                else:
-                    e = v[0] if type(v[0]) is int else res[v[2]]
-                    if e == -1:
-                        e = (len(buf) - off) / v[1].size
-                    lst = []
-                    res.append(lst)
-                    size = 0
-                    for i in range(e):
-                        lst.append(v[1].unpack_from(buf, off + size)[0])
-                        size += v[1].size
-            else:
-                if callable(v):
-                    (s,l) = v(self, False, buf, off, None)
-                    res.append(l)
-                    size += s
-                else:
-                    res.append(v.unpack_from(buf, off)[0])
-                    size = v.size
-
-        return off + size - offset, msgdef['return_tuple']._make(res)
-
-    def ret_tup(self, name):
-        if name in self.messages and 'return_tuple' in self.messages[name]:
-            return self.messages[name]['return_tuple']
-        return None
-
-    def add_message(self, name, msgdef):
-        if name in self.messages:
-            raise ValueError('Duplicate message name: ' + name)
-
-        args = collections.OrderedDict()
-        argtypes = collections.OrderedDict()
-        fields = []
-        msg = {}
-        for i, f in enumerate(msgdef):
-            if type(f) is dict and 'crc' in f:
-                msg['crc'] = f['crc']
-                continue
-            field_type = f[0]
-            field_name = f[1]
-            if len(f) == 3 and f[2] == 0 and i != len(msgdef) - 2:
-                raise ValueError('Variable Length Array must be last: ' + name)
-            args[field_name] = self.__struct(*f)
-            argtypes[field_name] = field_type
-            if len(f) == 4: # Find offset to # elements field
-                args[field_name].append(args.keys().index(f[3]) - i)
-            fields.append(field_name)
-        msg['return_tuple'] = collections.namedtuple(name, fields,
-                                                     rename = True)
-        self.messages[name] = msg
-        self.messages[name]['args'] = args
-        self.messages[name]['argtypes'] = argtypes
-        return self.messages[name]
-
-    def add_type(self, name, typedef):
-        return self.add_message('vl_api_' + name + '_t', typedef)
-
-    def make_function(self, name, i, msgdef, multipart, async):
-        if (async):
-            f = lambda **kwargs: (self._call_vpp_async(i, msgdef, **kwargs))
-        else:
-            f = lambda **kwargs: (self._call_vpp(i, msgdef, multipart, **kwargs))
-        args = self.messages[name]['args']
-        argtypes = self.messages[name]['argtypes']
-        f.__name__ = str(name)
-        f.__doc__ = ", ".join(["%s %s" % (argtypes[k], k) for k in args.keys()])
-        return f
-
-    @property
-    def api(self):
-        if not hasattr(self, "_api"):
-            raise Exception("Not connected, api definitions not available")
-        return self._api
-
-    def _register_functions(self, async=False):
-        self.id_names = [None] * (self.vpp_dictionary_maxid + 1)
-        self.id_msgdef = [None] * (self.vpp_dictionary_maxid + 1)
-        self._api = Empty()
-        for name, msgdef in self.messages.iteritems():
-            if name in self.vpp_dictionary:
-                if self.messages[name]['crc'] != self.vpp_dictionary[name]['crc']:
-                    raise ValueError(3, 'Failed CRC checksum ' + name +
-                                     ' ' + self.messages[name]['crc'] +
-                                     ' ' + self.vpp_dictionary[name]['crc'])
-                i = self.vpp_dictionary[name]['id']
-                self.id_msgdef[i] = msgdef
-                self.id_names[i] = name
-                multipart = True if name.find('_dump') > 0 else False
-                f = self.make_function(name, i, msgdef, multipart, async)
-                setattr(self._api, name, FuncWrapper(f))
-
-                # old API stuff starts here - will be removed in 17.07
-                if hasattr(self, name):
-                    raise NameError(
-                        3, "Conflicting name in JSON definition: `%s'" % name)
-                setattr(self, name, f)
-                # old API stuff ends here
-
-    def _write (self, buf):
-        """Send a binary-packed message to VPP."""
-        if not self.connected:
-            raise IOError(1, 'Not connected')
-        return vpp_api.write(str(buf))
-
-    def _read (self):
-        if not self.connected:
-            raise IOError(1, 'Not connected')
-
-        return vpp_api.read(self.read_timeout)
-
-    def _load_dictionary(self):
-        self.vpp_dictionary = {}
-        self.vpp_dictionary_maxid = 0
-        d = vpp_api.msg_table()
-
-        if not d:
-            raise IOError(3, 'Cannot get VPP API dictionary')
-        for i,n in d:
-            name, crc =  n.rsplit('_', 1)
-            crc = '0x' + crc
-            self.vpp_dictionary[name] = { 'id' : i, 'crc' : crc }
-            self.vpp_dictionary_maxid = max(self.vpp_dictionary_maxid, i)
-
-    def connect_internal(self, name, msg_handler, chroot_prefix, rx_qlen, async):
-	rv = vpp_api.connect(name, msg_handler, chroot_prefix, rx_qlen)
-        if rv != 0:
-            raise IOError(2, 'Connect failed')
-        self.connected = True
-
-        self._load_dictionary()
-        self._register_functions(async=async)
-
-        # Initialise control ping
-        self.control_ping_index = self.vpp_dictionary['control_ping']['id']
-        self.control_ping_msgdef = self.messages['control_ping']
-
-    def connect(self, name, chroot_prefix = None, async = False, rx_qlen = 32):
-        """Attach to VPP.
-
-        name - the name of the client.
-        chroot_prefix - if VPP is chroot'ed, the prefix of the jail
-        async - if true, messages are sent without waiting for a reply
-        rx_qlen - the length of the VPP message receive queue between
-        client and server.
-        """
-        msg_handler = self.msg_handler_sync if not async \
-                      else self.msg_handler_async
-        return self.connect_internal(name, msg_handler, chroot_prefix, rx_qlen,
-                                     async)
-
-    def connect_sync (self, name, chroot_prefix = None, rx_qlen = 32):
-        """Attach to VPP in synchronous mode. Application must poll for events.
-
-        name - the name of the client.
-        chroot_prefix - if VPP is chroot'ed, the prefix of the jail
-        rx_qlen - the length of the VPP message receive queue between
-        client and server.
-        """
-
-        return self.connect_internal(name, None, chroot_prefix, rx_qlen,
-                                     async=False)
-
-    def disconnect(self):
-        """Detach from VPP."""
-        rv = vpp_api.disconnect()
-        self.connected = False
-        return rv
-
-    def msg_handler_sync(self, msg):
-        """Process an incoming message from VPP in sync mode.
-
-        The message may be a reply or it may be an async notification.
-        """
-        r = self.decode_incoming_msg(msg)
-        if r is None:
-            return
-
-        # If we have a context, then use the context to find any
-        # request waiting for a reply
-        context = 0
-        if hasattr(r, 'context') and r.context > 0:
-            context = r.context
-
-        msgname = type(r).__name__
-
-        if context == 0:
-            # No context -> async notification that we feed to the callback
-            self.message_queue.put_nowait(r)
-        else:
-            raise IOError(2, 'RPC reply message received in event handler')
-
-    def decode_incoming_msg(self, msg):
-        if not msg:
-            eprint('vpp_api.read failed')
-            return
-
-        i, ci = self.header.unpack_from(msg, 0)
-        if self.id_names[i] == 'rx_thread_exit':
-            return
-
-        #
-        # Decode message and returns a tuple.
-        #
-        msgdef = self.id_msgdef[i]
-        if not msgdef:
-            raise IOError(2, 'Reply message undefined')
-
-        r = self.decode(msgdef, msg)
-
-        return r
-
-    def msg_handler_async(self, msg):
-        """Process a message from VPP in async mode.
-
-        In async mode, all messages are returned to the callback.
-        """
-        r = self.decode_incoming_msg(msg)
-        if r is None:
-            return
-
-        msgname = type(r).__name__
-
-	if self.event_callback:
-	    self.event_callback(msgname, r)
-
-    def _control_ping(self, context):
-        """Send a ping command."""
-        self._call_vpp_async(self.control_ping_index,
-			     self.control_ping_msgdef,
-                             context=context)
-
-    def _call_vpp(self, i, msgdef, multipart, **kwargs):
-        """Given a message, send the message and await a reply.
-
-        msgdef - the message packing definition
-        i - the message type index
-        multipart - True if the message returns multiple
-        messages in return.
-        context - context number - chosen at random if not
-        supplied.
-        The remainder of the kwargs are the arguments to the API call.
-
-        The return value is the message or message array containing
-        the response.  It will raise an IOError exception if there was
-        no response within the timeout window.
-        """
-
-        if not 'context' in kwargs:
-            context = self.get_context()
-            kwargs['context'] = context
-        else:
-            context = kwargs['context']
-        kwargs['_vl_msg_id'] = i
-        b = self.encode(msgdef, kwargs)
-
-        vpp_api.suspend()
-        self._write(b)
-
-        if multipart:
-            # Send a ping after the request - we use its response
-            # to detect that we have seen all results.
-            self._control_ping(context)
-
-        # Block until we get a reply.
-        rl = []
-        while (True):
-            msg = self._read()
-            if not msg:
-                print('PNEUM ERROR: OH MY GOD')
-                raise IOError(2, 'PNEUM read failed')
-
-            r = self.decode_incoming_msg(msg)
-            msgname = type(r).__name__
-            if not context in r or r.context == 0 or context != r.context:
-                self.message_queue.put_nowait(r)
-                continue
-
-            if not multipart:
-                rl = r
-                break
-            if msgname == 'control_ping_reply':
-                break
-
-            rl.append(r)
-
-        vpp_api.resume()
-
-        return rl
-
-    def _call_vpp_async(self, i, msgdef, **kwargs):
-        """Given a message, send the message and await a reply.
-
-        msgdef - the message packing definition
-        i - the message type index
-        context - context number - chosen at random if not
-        supplied.
-        The remainder of the kwargs are the arguments to the API call.
-        """
-        if not 'context' in kwargs:
-            context = self.get_context()
-            kwargs['context'] = context
-        else:
-            context = kwargs['context']
-        kwargs['_vl_msg_id'] = i
-        b = self.encode(msgdef, kwargs)
-
-        self._write(b)
-
-    def register_event_callback(self, callback):
-        """Register a callback for async messages.
-
-        This will be called for async notifications in sync mode,
-        and all messages in async mode.  In sync mode, replies to
-        requests will not come here.
-
-        callback is a fn(msg_type_name, msg_type) that will be
-        called when a message comes in.  While this function is
-        executing, note that (a) you are in a background thread and
-        may wish to use threading.Lock to protect your datastructures,
-        and (b) message processing from VPP will stop (so if you take
-        a long while about it you may provoke reply timeouts or cause
-        VPP to fill the RX buffer).  Passing None will disable the
-        callback.
-        """
-        self.event_callback = callback
-
-    def thread_msg_handler(self):
-        """Python thread calling the user registerd message handler.
-
-        This is to emulate the old style event callback scheme. Modern
-        clients should provide their own thread to poll the event
-        queue.
-        """
-        while True:
-            r = self.message_queue.get()
-            msgname = type(r).__name__
-	    if self.event_callback:
-	        self.event_callback(msgname, r)
diff --git a/test/Makefile b/test/Makefile
index 8594633b..4338e096 100644
--- a/test/Makefile
+++ b/test/Makefile
@@ -30,7 +30,7 @@ UNITTEST_EXTRA_OPTS="-f"
 endif
 
 PYTHON_VENV_PATH=$(VPP_PYTHON_PREFIX)/virtualenv
-PYTHON_DEPENDS=scapy==2.3.3 pexpect subprocess32 git+https://github.com/klement/py-lispnetworking@setup
+PYTHON_DEPENDS=scapy==2.3.3 pexpect subprocess32 cffi git+https://github.com/klement/py-lispnetworking@setup
 SCAPY_SOURCE=$(PYTHON_VENV_PATH)/lib/python2.7/site-packages/
 BUILD_COV_DIR = $(BR)/test-cov
 
diff --git a/test/vpp_papi_provider.py b/test/vpp_papi_provider.py
index 2d683dc2..7f9e2ae1 100644
--- a/test/vpp_papi_provider.py
+++ b/test/vpp_papi_provider.py
@@ -55,7 +55,7 @@ class VppPapiProvider(object):
             for filename in fnmatch.filter(filenames, '*.api.json'):
                 jsonfiles.append(os.path.join(root, filename))
 
-        self.vpp = VPP(jsonfiles)
+        self.vpp = VPP(jsonfiles, logger=test_class.logger)
         self._events = deque()
 
     def __enter__(self):
-- 
cgit 1.2.3-korg


From 8f2a4eafeaa439432107563033728e09665c16d9 Mon Sep 17 00:00:00 2001
From: Klement Sekera 
Date: Thu, 4 May 2017 06:15:18 +0200
Subject: Add new C API

Change-Id: I717ce3cd7c867c155de149ec56623269d26d0ff7
Signed-off-by: Klement Sekera 
---
 .gitignore                                |    1 +
 Makefile                                  |   11 +-
 src/Makefile.am                           |    2 +
 src/configure.ac                          |    2 +-
 src/vlibmemory/unix_shared_memory_queue.c |   65 ++
 src/vlibmemory/unix_shared_memory_queue.h |   11 +-
 src/vpp-api/vapi/Makefile.am              |   63 ++
 src/vpp-api/vapi/libvapiclient.map        |   41 +
 src/vpp-api/vapi/vapi.c                   |  895 ++++++++++++++++++++++
 src/vpp-api/vapi/vapi.h                   |  285 +++++++
 src/vpp-api/vapi/vapi_c_gen.py            |  809 ++++++++++++++++++++
 src/vpp-api/vapi/vapi_dbg.h               |   76 ++
 src/vpp-api/vapi/vapi_internal.h          |  126 ++++
 src/vpp-api/vapi/vapi_json_parser.py      |  303 ++++++++
 test/Makefile                             |    6 +-
 test/ext/Makefile                         |   17 +
 test/ext/vapi_test.c                      | 1152 +++++++++++++++++++++++++++++
 test/scripts/test-loop.sh                 |   14 +-
 test/test_vapi.py                         |   78 ++
 19 files changed, 3943 insertions(+), 14 deletions(-)
 create mode 100644 src/vpp-api/vapi/Makefile.am
 create mode 100644 src/vpp-api/vapi/libvapiclient.map
 create mode 100644 src/vpp-api/vapi/vapi.c
 create mode 100644 src/vpp-api/vapi/vapi.h
 create mode 100755 src/vpp-api/vapi/vapi_c_gen.py
 create mode 100644 src/vpp-api/vapi/vapi_dbg.h
 create mode 100644 src/vpp-api/vapi/vapi_internal.h
 create mode 100644 src/vpp-api/vapi/vapi_json_parser.py
 create mode 100644 test/ext/Makefile
 create mode 100644 test/ext/vapi_test.c
 create mode 100644 test/test_vapi.py

(limited to 'src/Makefile.am')

diff --git a/.gitignore b/.gitignore
index ba4e104a..5a6266d7 100644
--- a/.gitignore
+++ b/.gitignore
@@ -17,6 +17,7 @@
 /build-root/test-doc/
 /build-root/test-cov/
 /build-root/python/
+/build-root/vapi_test/
 /build-config.mk
 /dpdk/*.tar.gz
 /dpdk/*.tar.xz
diff --git a/Makefile b/Makefile
index c08115d4..c46fa6bb 100644
--- a/Makefile
+++ b/Makefile
@@ -62,7 +62,7 @@ DEB_DEPENDS  = curl build-essential autoconf automake bison libssl-dev ccache
 DEB_DEPENDS += debhelper dkms git libtool libapr1-dev dh-systemd
 DEB_DEPENDS += libconfuse-dev git-review exuberant-ctags cscope pkg-config
 DEB_DEPENDS += lcov chrpath autoconf nasm indent libnuma-dev
-DEB_DEPENDS += python-all python-dev python-virtualenv python-pip libffi6
+DEB_DEPENDS += python-all python-dev python-virtualenv python-pip libffi6 check
 ifeq ($(OS_VERSION_ID),14.04)
 	DEB_DEPENDS += openjdk-8-jdk-headless
 else ifeq ($(OS_ID)-$(OS_VERSION_ID),debian-8)
@@ -76,6 +76,7 @@ RPM_DEPENDS  = redhat-lsb glibc-static java-1.8.0-openjdk-devel yum-utils
 RPM_DEPENDS += apr-devel
 RPM_DEPENDS += openssl-devel
 RPM_DEPENDS += numactl-devel
+RPM_DEPENDS += check
 ifeq ($(OS_ID)-$(OS_VERSION_ID),fedora-25)
 	RPM_DEPENDS += python-devel
 	RPM_DEPENDS += python2-virtualenv
@@ -101,7 +102,13 @@ endif
 
 RPM_SUSE_DEPENDS = autoconf automake bison ccache chrpath distribution-release gcc6 glibc-devel-static
 RPM_SUSE_DEPENDS += java-1_8_0-openjdk-devel libopenssl-devel libtool lsb-release make openssl-devel
-RPM_SUSE_DEPENDS += python-devel python-pip python-rpm-macros shadow nasm libnuma-devel
+RPM_SUSE_DEPENDS += python-devel python-pip python-rpm-macros shadow nasm libnuma-devel python3
+
+ifeq ($(filter rhel centos,$(OS_ID)),$(OS_ID))
+	RPM_DEPENDS += python34
+else
+	RPM_DEPENDS += python3
+endif
 
 ifneq ($(wildcard $(STARTUP_DIR)/startup.conf),)
         STARTUP_CONF ?= $(STARTUP_DIR)/startup.conf
diff --git a/src/Makefile.am b/src/Makefile.am
index 41076e0e..7b35e50c 100644
--- a/src/Makefile.am
+++ b/src/Makefile.am
@@ -80,6 +80,8 @@ if ENABLE_JAPI
 SUBDIRS += vpp-api/java
 endif
 
+SUBDIRS += vpp-api/vapi
+
 ###############################################################################
 # API
 ###############################################################################
diff --git a/src/configure.ac b/src/configure.ac
index 6b6d9636..2efb23ad 100644
--- a/src/configure.ac
+++ b/src/configure.ac
@@ -3,7 +3,7 @@ LT_INIT
 AC_CONFIG_AUX_DIR([.])
 AM_INIT_AUTOMAKE([subdir-objects])
 AM_SILENT_RULES([yes])
-AC_CONFIG_FILES([Makefile plugins/Makefile vpp-api/python/Makefile vpp-api/java/Makefile])
+AC_CONFIG_FILES([Makefile plugins/Makefile vpp-api/python/Makefile vpp-api/java/Makefile vpp-api/vapi/Makefile])
 AC_CONFIG_MACRO_DIR([m4])
 
 AC_PROG_CC
diff --git a/src/vlibmemory/unix_shared_memory_queue.c b/src/vlibmemory/unix_shared_memory_queue.c
index e86edec3..4db4851c 100644
--- a/src/vlibmemory/unix_shared_memory_queue.c
+++ b/src/vlibmemory/unix_shared_memory_queue.c
@@ -234,6 +234,71 @@ unix_shared_memory_queue_add (unix_shared_memory_queue_t * q,
   return 0;
 }
 
+/*
+ * unix_shared_memory_queue_add2
+ */
+int
+unix_shared_memory_queue_add2 (unix_shared_memory_queue_t * q, u8 * elem,
+			       u8 * elem2, int nowait)
+{
+  i8 *tailp;
+  int need_broadcast = 0;
+
+  if (nowait)
+    {
+      /* zero on success */
+      if (pthread_mutex_trylock (&q->mutex))
+	{
+	  return (-1);
+	}
+    }
+  else
+    pthread_mutex_lock (&q->mutex);
+
+  if (PREDICT_FALSE (q->cursize + 1 == q->maxsize))
+    {
+      if (nowait)
+	{
+	  pthread_mutex_unlock (&q->mutex);
+	  return (-2);
+	}
+      while (q->cursize + 1 == q->maxsize)
+	{
+	  (void) pthread_cond_wait (&q->condvar, &q->mutex);
+	}
+    }
+
+  tailp = (i8 *) (&q->data[0] + q->elsize * q->tail);
+  clib_memcpy (tailp, elem, q->elsize);
+
+  q->tail++;
+  q->cursize++;
+
+  if (q->tail == q->maxsize)
+    q->tail = 0;
+
+  need_broadcast = (q->cursize == 1);
+
+  tailp = (i8 *) (&q->data[0] + q->elsize * q->tail);
+  clib_memcpy (tailp, elem2, q->elsize);
+
+  q->tail++;
+  q->cursize++;
+
+  if (q->tail == q->maxsize)
+    q->tail = 0;
+
+  if (need_broadcast)
+    {
+      (void) pthread_cond_broadcast (&q->condvar);
+      if (q->signal_when_queue_non_empty)
+	kill (q->consumer_pid, q->signal_when_queue_non_empty);
+    }
+  pthread_mutex_unlock (&q->mutex);
+
+  return 0;
+}
+
 /*
  * unix_shared_memory_queue_sub
  */
diff --git a/src/vlibmemory/unix_shared_memory_queue.h b/src/vlibmemory/unix_shared_memory_queue.h
index 13800065..27de3218 100644
--- a/src/vlibmemory/unix_shared_memory_queue.h
+++ b/src/vlibmemory/unix_shared_memory_queue.h
@@ -21,7 +21,6 @@
 #define included_unix_shared_memory_queue_h
 
 #include 
-#include 
 
 typedef struct _unix_shared_memory_queue
 {
@@ -43,10 +42,12 @@ unix_shared_memory_queue_t *unix_shared_memory_queue_init (int nels,
 							   int
 							   signal_when_queue_non_empty);
 void unix_shared_memory_queue_free (unix_shared_memory_queue_t * q);
-int unix_shared_memory_queue_add (unix_shared_memory_queue_t * q,
-				  u8 * elem, int nowait);
-int unix_shared_memory_queue_sub (unix_shared_memory_queue_t * q,
-				  u8 * elem, int nowait);
+int unix_shared_memory_queue_add (unix_shared_memory_queue_t * q, u8 * elem,
+				  int nowait);
+int unix_shared_memory_queue_add2 (unix_shared_memory_queue_t * q, u8 * elem,
+				   u8 * elem2, int nowait);
+int unix_shared_memory_queue_sub (unix_shared_memory_queue_t * q, u8 * elem,
+				  int nowait);
 void unix_shared_memory_queue_lock (unix_shared_memory_queue_t * q);
 void unix_shared_memory_queue_unlock (unix_shared_memory_queue_t * q);
 int unix_shared_memory_queue_is_full (unix_shared_memory_queue_t * q);
diff --git a/src/vpp-api/vapi/Makefile.am b/src/vpp-api/vapi/Makefile.am
new file mode 100644
index 00000000..ce681c38
--- /dev/null
+++ b/src/vpp-api/vapi/Makefile.am
@@ -0,0 +1,63 @@
+# Copyright (c) 2017 Cisco and/or its affiliates.
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at:
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+AUTOMAKE_OPTIONS = foreign
+ACLOCAL_AMFLAGS = -I m4
+AM_LIBTOOLFLAGS = --quiet
+
+AM_CFLAGS = -Wall -I${top_srcdir} -I${top_builddir} -I. -I$(top_srcdir)/vpp-api/vapi
+
+AM_LDFLAGS = -shared -avoid-version -rpath /none -no-undefined
+
+bin_PROGRAMS =
+noinst_LTLIBRARIES =
+CLEANDIRS =
+
+%.api.vapi.h: %.api.json vapi_c_gen.py
+	@echo "  VAPI C GEN $< " $@ ;			\
+	mkdir -p `dirname $@` ;				\
+        $(top_srcdir)/vpp-api/vapi/vapi_c_gen.py $<
+
+%.api.json:
+	find $(top_builddir) -name '$@' | xargs ln -s
+
+BUILT_SOURCES = $(shell find $(top_builddir) -name '*.api.json' | xargs -n1 basename) \
+  $(patsubst %.api.json,%.api.vapi.h,$(JSON_FILES))
+
+vapi.c: $(BUILT_SOURCES)
+
+JSON_FILES = $(wildcard *.api.json)
+
+
+lib_LTLIBRARIES = libvapiclient.la
+
+libvapiclient_la_SOURCES = vapi.c
+
+libvapiclient_la_LIBADD = -lpthread -lm -lrt \
+  $(top_builddir)/libvppinfra.la \
+  $(top_builddir)/libvlibmemoryclient.la \
+  $(top_builddir)/libsvm.la
+
+libvapiclient_la_LDFLAGS = \
+  -Wl,-L$(top_builddir)/.libs,--whole-archive,--no-whole-archive \
+  -Wl,--version-script=$(srcdir)/libvapiclient.map,-lrt
+
+libvapiclient_la_CPPFLAGS = -I. -I$(top_builddir)/vpp-api/vapi
+
+nobase_include_HEADERS = ${top_srcdir}/vpp-api/client/vppapiclient.h \
+  vapi.h \
+  vapi_dbg.h \
+  vapi_internal.h \
+  $(patsubst %.api.json,%.api.vapi.h,$(JSON_FILES))
+
+# vi:syntax=automake
diff --git a/src/vpp-api/vapi/libvapiclient.map b/src/vpp-api/vapi/libvapiclient.map
new file mode 100644
index 00000000..53733002
--- /dev/null
+++ b/src/vpp-api/vapi/libvapiclient.map
@@ -0,0 +1,41 @@
+
+VAPICLIENT_17.07 {
+	global:
+	vapi_msg_alloc;
+	vapi_msg_free;
+	vapi_ctx_alloc;
+	vapi_ctx_free;
+	vapi_is_msg_available;
+	vapi_connect;
+	vapi_disconnect;
+	vapi_get_fd;
+	vapi_send;
+	vapi_send2;
+	vapi_recv;
+	vapi_wait;
+	vapi_dispatch_one;
+	vapi_dispatch;
+	vapi_set_event_cb;
+	vapi_clear_event_cb;
+	vapi_set_generic_event_cb;
+	vapi_clear_generic_event_cb;
+	vapi_get_client_index;
+	vapi_register_msg;
+	vapi_get_client_index;
+	vapi_is_nonblocking;
+	vapi_requests_full;
+	vapi_gen_req_context;
+	vapi_producer_lock;
+	vapi_send_with_control_ping;
+	vapi_store_request;
+	vapi_is_nonblocking;
+	vapi_producer_unlock;
+	vapi_lookup_vl_msg_id;
+	vapi_lookup_vapi_msg_id_t;
+	vapi_msg_is_with_context;
+	vapi_get_context_offset;
+	vapi_msg_id_control_ping;
+	vapi_msg_id_control_ping_reply;
+
+	local: *;
+};
diff --git a/src/vpp-api/vapi/vapi.c b/src/vpp-api/vapi/vapi.c
new file mode 100644
index 00000000..b9c81a13
--- /dev/null
+++ b/src/vpp-api/vapi/vapi.c
@@ -0,0 +1,895 @@
+/*
+ *------------------------------------------------------------------
+ * Copyright (c) 2017 Cisco and/or its affiliates.
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at:
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ *------------------------------------------------------------------
+ */
+
+#include 
+#include 
+#include 
+#include 
+#include 
+#include 
+
+#include 
+#include 
+#include 
+#include 
+#include 
+#include 
+
+/* we need to use control pings for some stuff and because we're forced to put
+ * the code in headers, we need a way to be able to grab the ids of these
+ * messages - so declare them here as extern */
+vapi_msg_id_t vapi_msg_id_control_ping = 0;
+vapi_msg_id_t vapi_msg_id_control_ping_reply = 0;
+
+struct
+{
+  size_t count;
+  vapi_message_desc_t **msgs;
+  size_t max_len_name_with_crc;
+} __vapi_metadata;
+
+typedef struct
+{
+  u32 context;
+  vapi_cb_t callback;
+  void *callback_ctx;
+  bool is_dump;
+} vapi_req_t;
+
+static const u32 context_counter_mask = (1 << 31);
+
+typedef struct
+{
+  vapi_error_e (*cb) (vapi_ctx_t ctx, void *callback_ctx, vapi_msg_id_t id,
+		      void *payload);
+  void *ctx;
+} vapi_generic_cb_with_ctx;
+
+typedef struct
+{
+  vapi_error_e (*cb) (vapi_ctx_t ctx, void *callback_ctx, void *payload);
+  void *ctx;
+} vapi_event_cb_with_ctx;
+
+struct vapi_ctx_s
+{
+  vapi_mode_e mode;
+  int requests_size;		/* size of the requests array (circular queue) */
+  int requests_start;		/* index of first request */
+  int requests_count;		/* number of used slots */
+  vapi_req_t *requests;
+  u32 context_counter;
+  vapi_generic_cb_with_ctx generic_cb;
+  vapi_event_cb_with_ctx *event_cbs;
+  u16 *vapi_msg_id_t_to_vl_msg_id;
+  u16 vl_msg_id_max;
+  vapi_msg_id_t *vl_msg_id_to_vapi_msg_t;
+  bool connected;
+  pthread_mutex_t requests_mutex;
+};
+
+u32
+vapi_gen_req_context (vapi_ctx_t ctx)
+{
+  ++ctx->context_counter;
+  ctx->context_counter %= context_counter_mask;
+  return ctx->context_counter | context_counter_mask;
+}
+
+size_t
+vapi_get_request_count (vapi_ctx_t ctx)
+{
+  return ctx->requests_count;
+}
+
+bool
+vapi_requests_full (vapi_ctx_t ctx)
+{
+  return (ctx->requests_count == ctx->requests_size);
+}
+
+static bool
+vapi_requests_empty (vapi_ctx_t ctx)
+{
+  return (0 == ctx->requests_count);
+}
+
+static int
+vapi_requests_end (vapi_ctx_t ctx)
+{
+  return (ctx->requests_start + ctx->requests_count) % ctx->requests_size;
+}
+
+void
+vapi_store_request (vapi_ctx_t ctx, u32 context, bool is_dump,
+		    vapi_cb_t callback, void *callback_ctx)
+{
+  assert (!vapi_requests_full (ctx));
+  /* if the mutex is not held, bad things will happen */
+  assert (0 != pthread_mutex_trylock (&ctx->requests_mutex));
+  const int requests_end = vapi_requests_end (ctx);
+  vapi_req_t *slot = &ctx->requests[requests_end];
+  slot->is_dump = is_dump;
+  slot->context = context;
+  slot->callback = callback;
+  slot->callback_ctx = callback_ctx;
+  VAPI_DBG ("stored@%d: context:%x (start is @%d)", requests_end, context,
+	    ctx->requests_start);
+  ++ctx->requests_count;
+  assert (!vapi_requests_empty (ctx));
+}
+
+#if VAPI_DEBUG_ALLOC
+struct to_be_freed_s;
+struct to_be_freed_s
+{
+  void *v;
+  struct to_be_freed_s *next;
+};
+
+static struct to_be_freed_s *to_be_freed = NULL;
+
+void
+vapi_add_to_be_freed (void *v)
+{
+  struct to_be_freed_s *prev = NULL;
+  struct to_be_freed_s *tmp;
+  tmp = to_be_freed;
+  while (tmp && tmp->v)
+    {
+      prev = tmp;
+      tmp = tmp->next;
+    }
+  if (!tmp)
+    {
+      if (!prev)
+	{
+	  tmp = to_be_freed = calloc (1, sizeof (*to_be_freed));
+	}
+      else
+	{
+	  tmp = prev->next = calloc (1, sizeof (*to_be_freed));
+	}
+    }
+  VAPI_DBG ("To be freed %p", v);
+  tmp->v = v;
+}
+
+void
+vapi_trace_free (void *v)
+{
+  struct to_be_freed_s *tmp = to_be_freed;
+  while (tmp && tmp->v != v)
+    {
+      tmp = tmp->next;
+    }
+  if (tmp && tmp->v == v)
+    {
+      VAPI_DBG ("Freed %p", v);
+      tmp->v = NULL;
+    }
+  else
+    {
+      VAPI_ERR ("Trying to free untracked pointer %p", v);
+      abort ();
+    }
+}
+
+void
+vapi_to_be_freed_validate ()
+{
+  struct to_be_freed_s *tmp = to_be_freed;
+  while (tmp)
+    {
+      if (tmp->v)
+	{
+	  VAPI_ERR ("Unfreed msg %p!", tmp->v);
+	}
+      tmp = tmp->next;
+    }
+}
+
+#endif
+
+void *
+vapi_msg_alloc (vapi_ctx_t ctx, size_t size)
+{
+  if (!ctx->connected)
+    {
+      return NULL;
+    }
+  void *rv = vl_msg_api_alloc_or_null (size);
+  return rv;
+}
+
+void
+vapi_msg_free (vapi_ctx_t ctx, void *msg)
+{
+  if (!ctx->connected)
+    {
+      return;
+    }
+#if VAPI_DEBUG_ALLOC
+  vapi_trace_free (msg);
+#endif
+  vl_msg_api_free (msg);
+}
+
+vapi_error_e
+vapi_ctx_alloc (vapi_ctx_t * result)
+{
+  vapi_ctx_t ctx = calloc (1, sizeof (struct vapi_ctx_s));
+  if (!ctx)
+    {
+      return VAPI_ENOMEM;
+    }
+  ctx->context_counter = 0;
+  ctx->vapi_msg_id_t_to_vl_msg_id =
+    malloc (__vapi_metadata.count *
+	    sizeof (*ctx->vapi_msg_id_t_to_vl_msg_id));
+  if (!ctx->vapi_msg_id_t_to_vl_msg_id)
+    {
+      goto fail;
+    }
+  ctx->event_cbs = calloc (__vapi_metadata.count, sizeof (*ctx->event_cbs));
+  if (!ctx->event_cbs)
+    {
+      goto fail;
+    }
+  pthread_mutex_init (&ctx->requests_mutex, NULL);
+  *result = ctx;
+  return VAPI_OK;
+fail:
+  vapi_ctx_free (ctx);
+  return VAPI_ENOMEM;
+}
+
+void
+vapi_ctx_free (vapi_ctx_t ctx)
+{
+  assert (!ctx->connected);
+  free (ctx->requests);
+  free (ctx->vapi_msg_id_t_to_vl_msg_id);
+  free (ctx->event_cbs);
+  free (ctx->vl_msg_id_to_vapi_msg_t);
+  pthread_mutex_destroy (&ctx->requests_mutex);
+  free (ctx);
+}
+
+bool
+vapi_is_msg_available (vapi_ctx_t ctx, vapi_msg_id_t id)
+{
+  return vapi_lookup_vl_msg_id (ctx, id) != UINT16_MAX;
+}
+
+vapi_error_e
+vapi_connect (vapi_ctx_t ctx, const char *name,
+	      const char *chroot_prefix,
+	      int max_outstanding_requests,
+	      int response_queue_size, vapi_mode_e mode)
+{
+  if (response_queue_size <= 0 || max_outstanding_requests <= 0)
+    {
+      return VAPI_EINVAL;
+    }
+  ctx->requests_size = max_outstanding_requests;
+  const size_t size = ctx->requests_size * sizeof (*ctx->requests);
+  void *tmp = realloc (ctx->requests, size);
+  if (!tmp)
+    {
+      return VAPI_ENOMEM;
+    }
+  ctx->requests = tmp;
+  memset (ctx->requests, 0, size);
+  ctx->requests_start = ctx->requests_count = 0;
+  if (chroot_prefix)
+    {
+      VAPI_DBG ("set memory root path `%s'", chroot_prefix);
+      vl_set_memory_root_path ((char *) chroot_prefix);
+    }
+  static char api_map[] = "/vpe-api";
+  VAPI_DBG ("client api map `%s'", api_map);
+  if ((vl_client_api_map (api_map)) < 0)
+    {
+      return VAPI_EMAP_FAIL;
+    }
+  VAPI_DBG ("connect client `%s'", name);
+  if (vl_client_connect ((char *) name, 0, response_queue_size) < 0)
+    {
+      vl_client_api_unmap ();
+      return VAPI_ECON_FAIL;
+    }
+#if VAPI_DEBUG_CONNECT
+  VAPI_DBG ("start probing messages");
+#endif
+  int rv;
+  int i;
+  for (i = 0; i < __vapi_metadata.count; ++i)
+    {
+      vapi_message_desc_t *m = __vapi_metadata.msgs[i];
+      u8 scratch[m->name_with_crc_len + 1];
+      memcpy (scratch, m->name_with_crc, m->name_with_crc_len + 1);
+      u32 id = vl_api_get_msg_index (scratch);
+      if (~0 != id)
+	{
+	  if (id > UINT16_MAX)
+	    {
+	      VAPI_ERR ("Returned vl_msg_id `%u' > UINT16MAX `%u'!", id,
+			UINT16_MAX);
+	      rv = VAPI_EINVAL;
+	      goto fail;
+	    }
+	  if (id > ctx->vl_msg_id_max)
+	    {
+	      vapi_msg_id_t *tmp = realloc (ctx->vl_msg_id_to_vapi_msg_t,
+					    sizeof
+					    (*ctx->vl_msg_id_to_vapi_msg_t) *
+					    (id + 1));
+	      if (!tmp)
+		{
+		  rv = VAPI_ENOMEM;
+		  goto fail;
+		}
+	      ctx->vl_msg_id_to_vapi_msg_t = tmp;
+	      ctx->vl_msg_id_max = id;
+	    }
+	  ctx->vl_msg_id_to_vapi_msg_t[id] = m->id;
+	  ctx->vapi_msg_id_t_to_vl_msg_id[m->id] = id;
+#if VAPI_DEBUG_CONNECT
+	  VAPI_DBG ("Message `%s' has vl_msg_id `%u'", m->name_with_crc,
+		    (unsigned) id);
+#endif
+	}
+      else
+	{
+	  ctx->vapi_msg_id_t_to_vl_msg_id[m->id] = UINT16_MAX;
+	  VAPI_DBG ("Message `%s' not available", m->name_with_crc);
+	}
+    }
+#if VAPI_DEBUG_CONNECT
+  VAPI_DBG ("finished probing messages");
+#endif
+  if (!vapi_is_msg_available (ctx, vapi_msg_id_control_ping) ||
+      !vapi_is_msg_available (ctx, vapi_msg_id_control_ping_reply))
+    {
+      VAPI_ERR
+	("control ping or control ping reply not available, cannot connect");
+      rv = VAPI_EINCOMPATIBLE;
+      goto fail;
+    }
+  ctx->mode = mode;
+  ctx->connected = true;
+  return VAPI_OK;
+fail:
+  vl_client_disconnect ();
+  vl_client_api_unmap ();
+  return rv;
+}
+
+vapi_error_e
+vapi_disconnect (vapi_ctx_t ctx)
+{
+  if (!ctx->connected)
+    {
+      return VAPI_EINVAL;
+    }
+  vl_client_disconnect ();
+  vl_client_api_unmap ();
+#if VAPI_DEBUG_ALLOC
+  vapi_to_be_freed_validate ();
+#endif
+  ctx->connected = false;
+  return VAPI_OK;
+}
+
+vapi_error_e
+vapi_get_fd (vapi_ctx_t ctx, int *fd)
+{
+  return VAPI_ENOTSUP;
+}
+
+vapi_error_e
+vapi_send (vapi_ctx_t ctx, void *msg)
+{
+  vapi_error_e rv = VAPI_OK;
+  if (!ctx || !msg || !ctx->connected)
+    {
+      rv = VAPI_EINVAL;
+      goto out;
+    }
+  int tmp;
+  unix_shared_memory_queue_t *q = api_main.shmem_hdr->vl_input_queue;
+#if VAPI_DEBUG
+  unsigned msgid = be16toh (*(u16 *) msg);
+  if (msgid <= ctx->vl_msg_id_max)
+    {
+      vapi_msg_id_t id = ctx->vl_msg_id_to_vapi_msg_t[msgid];
+      if (id < __vapi_metadata.count)
+	{
+	  VAPI_DBG ("send msg %u[%s]", msgid, __vapi_metadata.msgs[id]->name);
+	}
+      else
+	{
+	  VAPI_DBG ("send msg %u[UNKNOWN]", msgid);
+	}
+    }
+  else
+    {
+      VAPI_DBG ("send msg %u[UNKNOWN]", msgid);
+    }
+#endif
+  tmp = unix_shared_memory_queue_add (q, (u8 *) & msg,
+				      VAPI_MODE_BLOCKING ==
+				      ctx->mode ? 0 : 1);
+  if (tmp < 0)
+    {
+      rv = VAPI_EAGAIN;
+    }
+out:
+  VAPI_DBG ("vapi_send() rv = %d", rv);
+  return rv;
+}
+
+vapi_error_e
+vapi_send2 (vapi_ctx_t ctx, void *msg1, void *msg2)
+{
+  vapi_error_e rv = VAPI_OK;
+  if (!ctx || !msg1 || !msg2 || !ctx->connected)
+    {
+      rv = VAPI_EINVAL;
+      goto out;
+    }
+  unix_shared_memory_queue_t *q = api_main.shmem_hdr->vl_input_queue;
+#if VAPI_DEBUG
+  unsigned msgid1 = be16toh (*(u16 *) msg1);
+  unsigned msgid2 = be16toh (*(u16 *) msg2);
+  const char *name1 = "UNKNOWN";
+  const char *name2 = "UNKNOWN";
+  if (msgid1 <= ctx->vl_msg_id_max)
+    {
+      vapi_msg_id_t id = ctx->vl_msg_id_to_vapi_msg_t[msgid1];
+      if (id < __vapi_metadata.count)
+	{
+	  name1 = __vapi_metadata.msgs[id]->name;
+	}
+    }
+  if (msgid2 <= ctx->vl_msg_id_max)
+    {
+      vapi_msg_id_t id = ctx->vl_msg_id_to_vapi_msg_t[msgid2];
+      if (id < __vapi_metadata.count)
+	{
+	  name2 = __vapi_metadata.msgs[id]->name;
+	}
+    }
+  VAPI_DBG ("send two: %u[%s], %u[%s]", msgid1, name1, msgid2, name2);
+#endif
+  int tmp = unix_shared_memory_queue_add2 (q, (u8 *) & msg1, (u8 *) & msg2,
+					   VAPI_MODE_BLOCKING ==
+					   ctx->mode ? 0 : 1);
+  if (tmp < 0)
+    {
+      rv = VAPI_EAGAIN;
+    }
+out:
+  VAPI_DBG ("vapi_send() rv = %d", rv);
+  return rv;
+}
+
+vapi_error_e
+vapi_recv (vapi_ctx_t ctx, void **msg, size_t * msg_size)
+{
+  if (!ctx || !ctx->connected || !msg || !msg_size)
+    {
+      return VAPI_EINVAL;
+    }
+  vapi_error_e rv = VAPI_OK;
+  api_main_t *am = &api_main;
+  uword data;
+
+  if (am->our_pid == 0)
+    {
+      return VAPI_EINVAL;
+    }
+
+  unix_shared_memory_queue_t *q = am->vl_input_queue;
+  VAPI_DBG ("doing shm queue sub");
+  int tmp = unix_shared_memory_queue_sub (q, (u8 *) & data, 0);
+  if (tmp == 0)
+    {
+#if VAPI_DEBUG_ALLOC
+      vapi_add_to_be_freed ((void *) data);
+#endif
+      msgbuf_t *msgbuf =
+	(msgbuf_t *) ((u8 *) data - offsetof (msgbuf_t, data));
+      if (!msgbuf->data_len)
+	{
+	  vapi_msg_free (ctx, (u8 *) data);
+	  return VAPI_EAGAIN;
+	}
+      *msg = (u8 *) data;
+      *msg_size = ntohl (msgbuf->data_len);
+      VAPI_DBG ("recv msg %p", *msg);
+    }
+  else
+    {
+      rv = VAPI_EAGAIN;
+    }
+  return rv;
+}
+
+vapi_error_e
+vapi_wait (vapi_ctx_t ctx, vapi_wait_mode_e mode)
+{
+  /* FIXME */
+  return VAPI_ENOTSUP;
+}
+
+static vapi_error_e
+vapi_dispatch_response (vapi_ctx_t ctx, vapi_msg_id_t id,
+			u32 context, void *msg)
+{
+  int mrv;
+  if (0 != (mrv = pthread_mutex_lock (&ctx->requests_mutex)))
+    {
+      VAPI_DBG ("pthread_mutex_lock() failed, rv=%d:%s", mrv, strerror (mrv));
+      return VAPI_MUTEX_FAILURE;
+    }
+  int tmp = ctx->requests_start;
+  const int requests_end = vapi_requests_end (ctx);
+  while (ctx->requests[tmp].context != context && tmp != requests_end)
+    {
+      ++tmp;
+      if (tmp == ctx->requests_size)
+	{
+	  tmp = 0;
+	}
+    }
+  VAPI_DBG ("dispatch, search from %d, %s at %d", ctx->requests_start,
+	    ctx->requests[tmp].context == context ? "matched" : "stopped",
+	    tmp);
+  vapi_error_e rv = VAPI_OK;
+  if (ctx->requests[tmp].context == context)
+    {
+      while (ctx->requests_start != tmp)
+	{
+	  VAPI_ERR ("No response to req with context=%u",
+		    (unsigned) ctx->requests[tmp].context);
+	  ctx->requests[ctx->requests_start].callback (ctx,
+						       ctx->requests
+						       [ctx->
+							requests_start].callback_ctx,
+						       VAPI_ENORESP, true,
+						       NULL);
+	  memset (&ctx->requests[ctx->requests_start], 0,
+		  sizeof (ctx->requests[ctx->requests_start]));
+	  ++ctx->requests_start;
+	  --ctx->requests_count;
+	  if (ctx->requests_start == ctx->requests_size)
+	    {
+	      ctx->requests_start = 0;
+	    }
+	}
+      // now ctx->requests_start == tmp
+      int payload_offset = vapi_get_payload_offset (id);
+      void *payload = ((u8 *) msg) + payload_offset;
+      bool is_last = true;
+      if (ctx->requests[tmp].is_dump)
+	{
+	  if (vapi_msg_id_control_ping_reply == id)
+	    {
+	      payload = NULL;
+	    }
+	  else
+	    {
+	      is_last = false;
+	    }
+	}
+      if (payload_offset != -1)
+	{
+	  rv =
+	    ctx->requests[tmp].callback (ctx, ctx->requests[tmp].callback_ctx,
+					 VAPI_OK, is_last, payload);
+	}
+      else
+	{
+	  /* this is a message without payload, so bend the callback a little
+	   */
+	  rv =
+	    ((vapi_error_e (*)(vapi_ctx_t, void *, vapi_error_e, bool))
+	     ctx->requests[tmp].callback) (ctx,
+					   ctx->requests[tmp].callback_ctx,
+					   VAPI_OK, is_last);
+	}
+      if (is_last)
+	{
+	  memset (&ctx->requests[ctx->requests_start], 0,
+		  sizeof (ctx->requests[ctx->requests_start]));
+	  ++ctx->requests_start;
+	  --ctx->requests_count;
+	  if (ctx->requests_start == ctx->requests_size)
+	    {
+	      ctx->requests_start = 0;
+	    }
+	}
+      VAPI_DBG ("after dispatch, req start = %d, end = %d, count = %d",
+		ctx->requests_start, requests_end, ctx->requests_count);
+    }
+  if (0 != (mrv = pthread_mutex_unlock (&ctx->requests_mutex)))
+    {
+      VAPI_DBG ("pthread_mutex_unlock() failed, rv=%d:%s", mrv,
+		strerror (mrv));
+      abort ();			/* this really shouldn't happen */
+    }
+  return rv;
+}
+
+static vapi_error_e
+vapi_dispatch_event (vapi_ctx_t ctx, vapi_msg_id_t id, void *msg)
+{
+  if (ctx->event_cbs[id].cb)
+    {
+      return ctx->event_cbs[id].cb (ctx, ctx->event_cbs[id].ctx, msg);
+    }
+  else if (ctx->generic_cb.cb)
+    {
+      return ctx->generic_cb.cb (ctx, ctx->generic_cb.ctx, id, msg);
+    }
+  else
+    {
+      VAPI_DBG
+	("No handler/generic handler for msg id %u[%s], message ignored",
+	 (unsigned) id, __vapi_metadata.msgs[id]->name);
+    }
+  return VAPI_OK;
+}
+
+static bool
+vapi_msg_is_with_context (vapi_msg_id_t id)
+{
+  assert (id <= __vapi_metadata.count);
+  return __vapi_metadata.msgs[id]->has_context;
+}
+
+vapi_error_e
+vapi_dispatch_one (vapi_ctx_t ctx)
+{
+  VAPI_DBG ("vapi_dispatch_one()");
+  void *msg;
+  size_t size;
+  vapi_error_e rv = vapi_recv (ctx, &msg, &size);
+  if (VAPI_OK != rv)
+    {
+      VAPI_DBG ("vapi_recv failed with rv=%d", rv);
+      return rv;
+    }
+  u16 vpp_id = be16toh (*(u16 *) msg);
+  if (vpp_id > ctx->vl_msg_id_max)
+    {
+      VAPI_ERR ("Unknown msg ID received, id `%u', out of range <0,%u>",
+		(unsigned) vpp_id, (unsigned) ctx->vl_msg_id_max);
+      vapi_msg_free (ctx, msg);
+      return VAPI_EINVAL;
+    }
+  if (~0 == (unsigned) ctx->vl_msg_id_to_vapi_msg_t[vpp_id])
+    {
+      VAPI_ERR ("Unknown msg ID received, id `%u' marked as not supported",
+		(unsigned) vpp_id);
+      vapi_msg_free (ctx, msg);
+      return VAPI_EINVAL;
+    }
+  const vapi_msg_id_t id = ctx->vl_msg_id_to_vapi_msg_t[vpp_id];
+  const size_t expect_size = vapi_get_message_size (id);
+  if (size < expect_size)
+    {
+      VAPI_ERR
+	("Invalid msg received, unexpected size `%zu' < expected min `%zu'",
+	 size, expect_size);
+      vapi_msg_free (ctx, msg);
+      return VAPI_EINVAL;
+    }
+  u32 context;
+  vapi_get_swap_to_host_func (id) (msg);
+  if (vapi_msg_is_with_context (id))
+    {
+      context = *(u32 *) (((u8 *) msg) + vapi_get_context_offset (id));
+      /* is this a message originating from VAPI? */
+      VAPI_DBG ("dispatch, context is %x", context);
+      if (context & context_counter_mask)
+	{
+	  rv = vapi_dispatch_response (ctx, id, context, msg);
+	  goto done;
+	}
+    }
+  rv = vapi_dispatch_event (ctx, id, msg);
+
+done:
+  vapi_msg_free (ctx, msg);
+  return rv;
+}
+
+vapi_error_e
+vapi_dispatch (vapi_ctx_t ctx)
+{
+  vapi_error_e rv = VAPI_OK;
+  while (!vapi_requests_empty (ctx))
+    {
+      rv = vapi_dispatch_one (ctx);
+      if (VAPI_OK != rv)
+	{
+	  return rv;
+	}
+    }
+  return rv;
+}
+
+void
+vapi_set_event_cb (vapi_ctx_t ctx, vapi_msg_id_t id,
+		   vapi_event_cb callback, void *callback_ctx)
+{
+  vapi_event_cb_with_ctx *c = &ctx->event_cbs[id];
+  c->cb = callback;
+  c->ctx = callback_ctx;
+}
+
+void
+vapi_clear_event_cb (vapi_ctx_t ctx, vapi_msg_id_t id)
+{
+  vapi_set_event_cb (ctx, id, NULL, NULL);
+}
+
+void
+vapi_set_generic_event_cb (vapi_ctx_t ctx, vapi_generic_event_cb callback,
+			   void *callback_ctx)
+{
+  ctx->generic_cb.cb = callback;
+  ctx->generic_cb.ctx = callback_ctx;
+}
+
+void
+vapi_clear_generic_event_cb (vapi_ctx_t ctx)
+{
+  ctx->generic_cb.cb = NULL;
+  ctx->generic_cb.ctx = NULL;
+}
+
+u16
+vapi_lookup_vl_msg_id (vapi_ctx_t ctx, vapi_msg_id_t id)
+{
+  assert (id < __vapi_metadata.count);
+  return ctx->vapi_msg_id_t_to_vl_msg_id[id];
+}
+
+int
+vapi_get_client_index (vapi_ctx_t ctx)
+{
+  return api_main.my_client_index;
+}
+
+bool
+vapi_is_nonblocking (vapi_ctx_t ctx)
+{
+  return (VAPI_MODE_NONBLOCKING == ctx->mode);
+}
+
+bool vapi_requests_full (vapi_ctx_t ctx);
+
+size_t vapi_get_request_count (vapi_ctx_t ctx);
+
+size_t
+vapi_get_max_request_count (vapi_ctx_t ctx)
+{
+  return ctx->requests_size - 1;
+}
+
+int
+vapi_get_payload_offset (vapi_msg_id_t id)
+{
+  assert (id < __vapi_metadata.count);
+  return __vapi_metadata.msgs[id]->payload_offset;
+}
+
+void (*vapi_get_swap_to_host_func (vapi_msg_id_t id)) (void *msg)
+{
+  assert (id < __vapi_metadata.count);
+  return __vapi_metadata.msgs[id]->swap_to_host;
+}
+
+void (*vapi_get_swap_to_be_func (vapi_msg_id_t id)) (void *msg)
+{
+  assert (id < __vapi_metadata.count);
+  return __vapi_metadata.msgs[id]->swap_to_be;
+}
+
+size_t
+vapi_get_message_size (vapi_msg_id_t id)
+{
+  assert (id < __vapi_metadata.count);
+  return __vapi_metadata.msgs[id]->size;
+}
+
+size_t
+vapi_get_context_offset (vapi_msg_id_t id)
+{
+  assert (id < __vapi_metadata.count);
+  return __vapi_metadata.msgs[id]->context_offset;
+}
+
+vapi_msg_id_t
+vapi_register_msg (vapi_message_desc_t * msg)
+{
+  int i = 0;
+  for (i = 0; i < __vapi_metadata.count; ++i)
+    {
+      if (!strcmp
+	  (msg->name_with_crc, __vapi_metadata.msgs[i]->name_with_crc))
+	{
+	  /* this happens if somebody is linking together several objects while
+	   * using the static inline headers, just fill in the already
+	   * assigned id here so that all the objects are in sync */
+	  msg->id = __vapi_metadata.msgs[i]->id;
+	  return msg->id;
+	}
+    }
+  vapi_msg_id_t id = __vapi_metadata.count;
+  ++__vapi_metadata.count;
+  __vapi_metadata.msgs =
+    realloc (__vapi_metadata.msgs,
+	     sizeof (*__vapi_metadata.msgs) * __vapi_metadata.count);
+  __vapi_metadata.msgs[id] = msg;
+  size_t s = strlen (msg->name_with_crc);
+  if (s > __vapi_metadata.max_len_name_with_crc)
+    {
+      __vapi_metadata.max_len_name_with_crc = s;
+    }
+  msg->id = id;
+  return id;
+}
+
+vapi_error_e
+vapi_producer_lock (vapi_ctx_t ctx)
+{
+  int mrv;
+  if (0 != (mrv = pthread_mutex_lock (&ctx->requests_mutex)))
+    {
+      VAPI_DBG ("pthread_mutex_lock() failed, rv=%d:%s", mrv, strerror (mrv));
+      (void) mrv;		/* avoid warning if the above debug is not enabled */
+      return VAPI_MUTEX_FAILURE;
+    }
+  return VAPI_OK;
+}
+
+vapi_error_e
+vapi_producer_unlock (vapi_ctx_t ctx)
+{
+  int mrv;
+  if (0 != (mrv = pthread_mutex_unlock (&ctx->requests_mutex)))
+    {
+      VAPI_DBG ("pthread_mutex_unlock() failed, rv=%d:%s", mrv,
+		strerror (mrv));
+      (void) mrv;		/* avoid warning if the above debug is not enabled */
+      return VAPI_MUTEX_FAILURE;
+    }
+  return VAPI_OK;
+}
+
+/*
+ * fd.io coding-style-patch-verification: ON
+ *
+ * Local Variables:
+ * eval: (c-set-style "gnu")
+ * End:
+ */
diff --git a/src/vpp-api/vapi/vapi.h b/src/vpp-api/vapi/vapi.h
new file mode 100644
index 00000000..1e1d567a
--- /dev/null
+++ b/src/vpp-api/vapi/vapi.h
@@ -0,0 +1,285 @@
+/*
+ *------------------------------------------------------------------
+ * Copyright (c) 2017 Cisco and/or its affiliates.
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at:
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ *------------------------------------------------------------------
+ */
+
+#ifndef vpp_api_h_included
+#define vpp_api_h_included
+
+#include 
+#include 
+#include 
+
+/**
+ * @file vapi.h
+ *
+ * common vpp api C declarations
+ *
+ * This file declares the common C API functions. These include connect,
+ * disconnect and utility functions as well as the low-level vapi_send and
+ * vapi_recv API. This is only the transport layer.
+ *
+ * Message formats and higher-level APIs are generated by running the
+ * vapi_c_gen.py script (which is run for in-tree APIs as part of the build
+ * process). It's not recommended to mix the higher and lower level APIs. Due
+ * to version issues, the higher-level APIs are not part of the shared library.
+ */
+
+typedef enum
+{
+  VAPI_OK = 0,	      /**< success */
+  VAPI_EINVAL,	      /**< invalid value encountered */
+  VAPI_EAGAIN,	      /**< operation would block */
+  VAPI_ENOTSUP,	      /**< operation not supported */
+  VAPI_ENOMEM,	      /**< out of memory */
+  VAPI_ENORESP,	      /**< no response to request */
+  VAPI_EMAP_FAIL,     /**< failure while mapping api */
+  VAPI_ECON_FAIL,     /**< failure while connecting to vpp */
+  VAPI_EINCOMPATIBLE, /**< fundamental incompatibility while connecting to vpp
+                           (control ping/control ping reply mismatch) */
+  VAPI_MUTEX_FAILURE, /**< failure manipulating internal mutex(es) */
+  VAPI_EUSER,	      /**< user error used for breaking dispatch,
+                           never used by VAPI */
+} vapi_error_e;
+
+typedef enum
+{
+  VAPI_MODE_BLOCKING = 1,    /**< operations block until response received */
+  VAPI_MODE_NONBLOCKING = 2, /**< operations never block */
+} vapi_mode_e;
+
+typedef enum
+{
+  VAPI_WAIT_FOR_READ,	    /**< wait until a message can be read */
+  VAPI_WAIT_FOR_WRITE,	    /**< wait until a message can be written */
+  VAPI_WAIT_FOR_READ_WRITE, /**< wait until a read or write can be done */
+} vapi_wait_mode_e;
+
+typedef int vapi_msg_id_t;
+typedef struct vapi_ctx_s *vapi_ctx_t;
+
+/**
+ * @brief allocate vapi message of given size
+ *
+ * @note message must be freed by vapi_msg_free if not consumed by vapi_send
+ * call
+ *
+ * @param ctx opaque vapi context
+ *
+ * @return pointer to message or NULL if out of memory
+ */
+void *vapi_msg_alloc (vapi_ctx_t ctx, size_t size);
+
+/**
+ * @brief free a vapi message
+ *
+ * @note messages received by vapi_recv must be freed when no longer needed
+ *
+ * @param ctx opaque vapi context
+ * @param msg message to be freed
+ */
+void vapi_msg_free (vapi_ctx_t ctx, void *msg);
+
+/**
+ * @brief allocate vapi context
+ *
+ * @param[out] pointer to result variable
+ *
+ * @return VAPI_OK on success, other error code on error
+ */
+vapi_error_e vapi_ctx_alloc (vapi_ctx_t * result);
+
+/**
+ * @brief free vapi context
+ */
+void vapi_ctx_free (vapi_ctx_t ctx);
+
+/**
+ * @brief check if message identified by it's message id is known by the vpp to
+ * which the connection is open
+ */
+bool vapi_is_msg_available (vapi_ctx_t ctx, vapi_msg_id_t type);
+
+/**
+ * @brief connect to vpp
+ *
+ * @param ctx opaque vapi context, must be allocated using vapi_ctx_alloc first
+ * @param name application name
+ * @param chroot_prefix shared memory prefix
+ * @param max_outstanding_requests max number of outstanding requests queued
+ * @param response_queue_size size of the response queue
+ * @param mode mode of operation - blocking or nonblocking
+ *
+ * @return VAPI_OK on success, other error code on error
+ */
+vapi_error_e vapi_connect (vapi_ctx_t ctx, const char *name,
+			   const char *chroot_prefix,
+			   int max_outstanding_requests,
+			   int response_queue_size, vapi_mode_e mode);
+
+/**
+ * @brief disconnect from vpp
+ *
+ * @param ctx opaque vapi context
+ *
+ * @return VAPI_OK on success, other error code on error
+ */
+vapi_error_e vapi_disconnect (vapi_ctx_t ctx);
+
+/**
+ * @brief get event file descriptor
+ *
+ * @note this file descriptor becomes readable when messages (from vpp)
+ * are waiting in queue
+ *
+ * @param ctx opaque vapi context
+ * @param[out] fd pointer to result variable
+ *
+ * @return VAPI_OK on success, other error code on error
+ */
+vapi_error_e vapi_get_fd (vapi_ctx_t ctx, int *fd);
+
+/**
+ * @brief low-level api for sending messages to vpp
+ *
+ * @note it is not recommended to use this api directly, use generated api
+ * instead
+ *
+ * @param ctx opaque vapi context
+ * @param msg message to send
+ *
+ * @return VAPI_OK on success, other error code on error
+ */
+vapi_error_e vapi_send (vapi_ctx_t ctx, void *msg);
+
+/**
+ * @brief low-level api for atomically sending two messages to vpp - either
+ * both messages are sent or neither one is
+ *
+ * @note it is not recommended to use this api directly, use generated api
+ * instead
+ *
+ * @param ctx opaque vapi context
+ * @param msg1 first message to send
+ * @param msg2 second message to send
+ *
+ * @return VAPI_OK on success, other error code on error
+ */
+vapi_error_e vapi_send2 (vapi_ctx_t ctx, void *msg1, void *msg2);
+
+/**
+ * @brief low-level api for reading messages from vpp
+ *
+ * @note it is not recommended to use this api directly, use generated api
+ * instead
+ *
+ * @param ctx opaque vapi context
+ * @param[out] msg pointer to result variable containing message
+ * @param[out] msg_size pointer to result variable containing message size
+ *
+ * @return VAPI_OK on success, other error code on error
+ */
+vapi_error_e vapi_recv (vapi_ctx_t ctx, void **msg, size_t * msg_size);
+
+/**
+ * @brief wait for connection to become readable or writable
+ *
+ * @param ctx opaque vapi context
+ * @param mode type of property to wait for - readability, writability or both
+ *
+ * @return VAPI_OK on success, other error code on error
+ */
+vapi_error_e vapi_wait (vapi_ctx_t ctx, vapi_wait_mode_e mode);
+
+/**
+ * @brief pick next message sent by vpp and call the appropriate callback
+ *
+ * @return VAPI_OK on success, other error code on error
+ */
+vapi_error_e vapi_dispatch_one (vapi_ctx_t ctx);
+
+/**
+ * @brief loop vapi_dispatch_one until responses to all currently outstanding
+ * requests have been received and their callbacks called
+ *
+ * @note the dispatch loop is interrupted if any error is encountered or
+ * returned from the callback, in which case this error is returned as the
+ * result of vapi_dispatch. In this case it might be necessary to call dispatch
+ * again to process the remaining messages. Returning VAPI_EUSER from
+ * a callback allows the user to break the dispatch loop (and distinguish
+ * this case in the calling code from other failures). VAPI never returns
+ * VAPI_EUSER on its own.
+ *
+ * @return VAPI_OK on success, other error code on error
+ */
+vapi_error_e vapi_dispatch (vapi_ctx_t ctx);
+
+/** generic vapi event callback */
+typedef vapi_error_e (*vapi_event_cb) (vapi_ctx_t ctx, void *callback_ctx,
+				       void *payload);
+
+/**
+ * @brief set event callback to call when message with given id is dispatched
+ *
+ * @param ctx opaque vapi context
+ * @param id message id
+ * @param callback callback
+ * @param callback_ctx context pointer stored and passed to callback
+ */
+void vapi_set_event_cb (vapi_ctx_t ctx, vapi_msg_id_t id,
+			vapi_event_cb callback, void *callback_ctx);
+
+/**
+ * @brief clear event callback for given message id
+ *
+ * @param ctx opaque vapi context
+ * @param id message id
+ */
+void vapi_clear_event_cb (vapi_ctx_t ctx, vapi_msg_id_t id);
+
+/** generic vapi event callback */
+typedef vapi_error_e (*vapi_generic_event_cb) (vapi_ctx_t ctx,
+					       void *callback_ctx,
+					       vapi_msg_id_t id, void *msg);
+/**
+ * @brief set generic event callback
+ *
+ * @note this callback is called by dispatch if no message-type specific
+ * callback is set (so it's a fallback callback)
+ *
+ * @param ctx opaque vapi context
+ * @param callback callback
+ * @param callback_ctx context pointer stored and passed to callback
+ */
+void vapi_set_generic_event_cb (vapi_ctx_t ctx,
+				vapi_generic_event_cb callback,
+				void *callback_ctx);
+
+/**
+ * @brief clear generic event callback
+ *
+ * @param ctx opaque vapi context
+ */
+void vapi_clear_generic_event_cb (vapi_ctx_t ctx);
+
+#endif
+
+/*
+ * fd.io coding-style-patch-verification: ON
+ *
+ * Local Variables:
+ * eval: (c-set-style "gnu")
+ * End:
+ */
diff --git a/src/vpp-api/vapi/vapi_c_gen.py b/src/vpp-api/vapi/vapi_c_gen.py
new file mode 100755
index 00000000..2bc1eef8
--- /dev/null
+++ b/src/vpp-api/vapi/vapi_c_gen.py
@@ -0,0 +1,809 @@
+#!/usr/bin/env python3
+
+import argparse
+import os
+import sys
+import logging
+from vapi_json_parser import Field, Struct, Message, JsonParser,\
+    SimpleType, StructType
+
+
+class CField(Field):
+    def __init__(
+            self,
+            field_name,
+            field_type,
+            array_len=None,
+            nelem_field=None):
+        super().__init__(field_name, field_type, array_len, nelem_field)
+
+    def get_c_def(self):
+        if self.len is not None:
+            return "%s %s[%d]" % (self.type.get_c_name(), self.name, self.len)
+        else:
+            return "%s %s" % (self.type.get_c_name(), self.name)
+
+    def get_swap_to_be_code(self, struct, var):
+        if self.len is not None:
+            if self.len > 0:
+                return "do { int i; for (i = 0; i < %d; ++i) { %s } }"\
+                    " while(0);" % (
+                        self.len,
+                        self.type.get_swap_to_be_code(struct, "%s[i]" % var))
+            else:
+                if self.nelem_field.needs_byte_swap():
+                    nelem_field = "%s(%s%s)" % (
+                        self.nelem_field.type.get_swap_to_host_func_name(),
+                        struct, self.nelem_field.name)
+                else:
+                    nelem_field = "%s%s" % (struct, self.nelem_field.name)
+                return (
+                    "do { int i; for (i = 0; i < %s; ++i) { %s } }"
+                    " while(0);" %
+                    (nelem_field, self.type.get_swap_to_be_code(
+                        struct, "%s[i]" % var)))
+        return self.type.get_swap_to_be_code(struct, "%s" % var)
+
+    def get_swap_to_host_code(self, struct, var):
+        if self.len is not None:
+            if self.len > 0:
+                return "do { int i; for (i = 0; i < %d; ++i) { %s } }"\
+                    " while(0);" % (
+                        self.len,
+                        self.type.get_swap_to_host_code(struct, "%s[i]" % var))
+            else:
+                # nelem_field already swapped to host here...
+                return (
+                    "do { int i; for (i = 0; i < %s%s; ++i) { %s } }"
+                    " while(0);" %
+                    (struct, self.nelem_field.name,
+                     self.type.get_swap_to_host_code(
+                         struct, "%s[i]" % var)))
+        return self.type.get_swap_to_host_code(struct, "%s" % var)
+
+    def needs_byte_swap(self):
+        return self.type.needs_byte_swap()
+
+
+class CStruct(Struct):
+    def __init__(self, name, fields):
+        super().__init__(name, fields)
+
+    def get_c_def(self):
+        return "\n".join([
+            "typedef struct __attribute__((__packed__)) {",
+            "%s;" % ";\n".join(["  %s" % x.get_c_def()
+                                for x in self.fields]),
+            "} %s;" % self.get_c_name()])
+
+
+class CSimpleType (SimpleType):
+
+    swap_to_be_dict = {
+        'i16': 'htobe16', 'u16': 'htobe16',
+        'i32': 'htobe32', 'u32': 'htobe32',
+        'i64': 'htobe64', 'u64': 'htobe64',
+    }
+
+    swap_to_host_dict = {
+        'i16': 'be16toh', 'u16': 'be16toh',
+        'i32': 'be32toh', 'u32': 'be32toh',
+        'i64': 'be64toh', 'u64': 'be64toh',
+    }
+
+    def __init__(self, name):
+        super().__init__(name)
+
+    def get_c_name(self):
+        return self.name
+
+    def get_swap_to_be_func_name(self):
+        return self.swap_to_be_dict[self.name]
+
+    def get_swap_to_host_func_name(self):
+        return self.swap_to_host_dict[self.name]
+
+    def get_swap_to_be_code(self, struct, var):
+        x = "%s%s" % (struct, var)
+        return "%s = %s(%s);" % (x, self.get_swap_to_be_func_name(), x)
+
+    def get_swap_to_host_code(self, struct, var):
+        x = "%s%s" % (struct, var)
+        return "%s = %s(%s);" % (x, self.get_swap_to_host_func_name(), x)
+
+    def needs_byte_swap(self):
+        try:
+            self.get_swap_to_host_func_name()
+            return True
+        except:
+            pass
+        return False
+
+
+class CStructType (StructType, CStruct):
+    def __init__(self, definition, typedict, field_class):
+        super().__init__(definition, typedict, field_class)
+
+    def get_c_name(self):
+        return "vapi_type_%s" % self.name
+
+    def get_swap_to_be_func_name(self):
+        return "%s_hton" % self.get_c_name()
+
+    def get_swap_to_host_func_name(self):
+        return "%s_ntoh" % self.get_c_name()
+
+    def get_swap_to_be_func_decl(self):
+        return "void %s(%s *msg)" % (
+            self.get_swap_to_be_func_name(), self.get_c_name())
+
+    def get_swap_to_be_func_def(self):
+        return "%s\n{\n%s\n}" % (
+            self.get_swap_to_be_func_decl(),
+            "\n".join([
+                "  %s" % p.get_swap_to_be_code("msg->", "%s" % p.name)
+                for p in self.fields if p.needs_byte_swap()]),
+        )
+
+    def get_swap_to_host_func_decl(self):
+        return "void %s(%s *msg)" % (
+            self.get_swap_to_host_func_name(), self.get_c_name())
+
+    def get_swap_to_host_func_def(self):
+        return "%s\n{\n%s\n}" % (
+            self.get_swap_to_host_func_decl(),
+            "\n".join([
+                "  %s" % p.get_swap_to_host_code("msg->", "%s" % p.name)
+                for p in self.fields if p.needs_byte_swap()]),
+        )
+
+    def get_swap_to_be_code(self, struct, var):
+        return "%s(&%s%s);" % (self.get_swap_to_be_func_name(), struct, var)
+
+    def get_swap_to_host_code(self, struct, var):
+        return "%s(&%s%s);" % (self.get_swap_to_host_func_name(), struct, var)
+
+    def needs_byte_swap(self):
+        for f in self.fields:
+            if f.needs_byte_swap():
+                return True
+        return False
+
+
+class CMessage (Message):
+    def __init__(self, logger, definition, typedict,
+                 struct_type_class, simple_type_class, field_class):
+        super().__init__(logger, definition, typedict, struct_type_class,
+                         simple_type_class, field_class)
+        self.payload_members = [
+            "  %s" % p.get_c_def()
+            for p in self.fields
+            if p.type != self.header
+        ]
+
+    def has_payload(self):
+        return len(self.payload_members) > 0
+
+    def get_msg_id_name(self):
+        return "vapi_msg_id_%s" % self.name
+
+    def get_c_name(self):
+        return "vapi_msg_%s" % self.name
+
+    def get_payload_struct_name(self):
+        return "vapi_payload_%s" % self.name
+
+    def get_alloc_func_vla_field_length_name(self, field):
+        return "%s_array_size" % field.name
+
+    def get_alloc_func_name(self):
+        return "vapi_alloc_%s" % self.name
+
+    def get_alloc_func_decl(self):
+        return "%s* %s(struct vapi_ctx_s *ctx%s)" % (
+            self.get_c_name(),
+            self.get_alloc_func_name(),
+            "".join([", size_t %s" %
+                     self.get_alloc_func_vla_field_length_name(f)
+                     for f in self.fields
+                     if f.nelem_field is not None]))
+
+    def get_alloc_func_def(self):
+        extra = []
+        if self.header.has_field('client_index'):
+            extra.append(
+                "  msg->header.client_index = vapi_get_client_index(ctx);")
+        if self.header.has_field('context'):
+            extra.append("  msg->header.context = 0;")
+        return "\n".join([
+            "%s" % self.get_alloc_func_decl(),
+            "{",
+            "  %s *msg = NULL;" % self.get_c_name(),
+            "  const size_t size = sizeof(%s)%s;" % (
+                self.get_c_name(),
+                "".join([
+                    " + sizeof(msg->payload.%s[0]) * %s" % (
+                        f.name,
+                        self.get_alloc_func_vla_field_length_name(f))
+                    for f in self.fields
+                    if f.nelem_field is not None
+                ])),
+            "  msg = vapi_msg_alloc(ctx, size);",
+            "  if (!msg) {",
+            "    return NULL;",
+            "  }",
+        ] + extra + [
+            "  msg->header._vl_msg_id = vapi_lookup_vl_msg_id(ctx, %s);" %
+            self.get_msg_id_name(),
+            "\n".join(["  msg->payload.%s = %s;" % (
+                f.nelem_field.name,
+                self.get_alloc_func_vla_field_length_name(f))
+                for f in self.fields
+                if f.nelem_field is not None]),
+            "  return msg;",
+            "}"])
+
+    def get_calc_msg_size_func_name(self):
+        return "vapi_calc_%s_msg_size" % self.name
+
+    def get_calc_msg_size_func_decl(self):
+        return "uword %s(%s *msg)" % (
+            self.get_calc_msg_size_func_name(),
+            self.get_c_name())
+
+    def get_calc_msg_size_func_def(self):
+        return "\n".join([
+            "%s" % self.get_calc_msg_size_func_decl(),
+            "{",
+            "  return sizeof(*msg)%s;" %
+            "".join(["+ msg->payload.%s * sizeof(msg->payload.%s[0])" % (
+                    f.nelem_field.name,
+                    f.name)
+                for f in self.fields
+                if f.nelem_field is not None
+            ]),
+            "}",
+        ])
+
+    def get_c_def(self):
+        if self.has_payload():
+            return "\n".join([
+                "typedef struct __attribute__ ((__packed__)) {",
+                "%s; " %
+                ";\n".join(self.payload_members),
+                "} %s;" % self.get_payload_struct_name(),
+                "",
+                "typedef struct __attribute__ ((__packed__)) {",
+                ("  %s %s;" % (self.header.get_c_name(),
+                               self.fields[0].name)
+                    if self.header is not None else ""),
+                "  %s payload;" % self.get_payload_struct_name(),
+                "} %s;" % self.get_c_name(), ])
+        else:
+            return "\n".join([
+                "typedef struct __attribute__ ((__packed__)) {",
+                ("  %s %s;" % (self.header.get_c_name(),
+                               self.fields[0].name)
+                    if self.header is not None else ""),
+                "} %s;" % self.get_c_name(), ])
+
+    def get_swap_payload_to_host_func_name(self):
+        return "%s_payload_ntoh" % self.get_c_name()
+
+    def get_swap_payload_to_be_func_name(self):
+        return "%s_payload_hton" % self.get_c_name()
+
+    def get_swap_payload_to_host_func_decl(self):
+        return "void %s(%s *payload)" % (
+            self.get_swap_payload_to_host_func_name(),
+            self.get_payload_struct_name())
+
+    def get_swap_payload_to_be_func_decl(self):
+        return "void %s(%s *payload)" % (
+            self.get_swap_payload_to_be_func_name(),
+            self.get_payload_struct_name())
+
+    def get_swap_payload_to_be_func_def(self):
+        return "%s\n{\n%s\n}" % (
+            self.get_swap_payload_to_be_func_decl(),
+            "\n".join([
+                "  %s" % p.get_swap_to_be_code("payload->", "%s" % p.name)
+                for p in self.fields
+                if p.needs_byte_swap() and p.type != self.header]),
+        )
+
+    def get_swap_payload_to_host_func_def(self):
+        return "%s\n{\n%s\n}" % (
+            self.get_swap_payload_to_host_func_decl(),
+            "\n".join([
+                "  %s" % p.get_swap_to_host_code("payload->", "%s" % p.name)
+                for p in self.fields
+                if p.needs_byte_swap() and p.type != self.header]),
+        )
+
+    def get_swap_to_host_func_name(self):
+        return "%s_ntoh" % self.get_c_name()
+
+    def get_swap_to_be_func_name(self):
+        return "%s_hton" % self.get_c_name()
+
+    def get_swap_to_host_func_decl(self):
+        return "void %s(%s *msg)" % (
+            self.get_swap_to_host_func_name(), self.get_c_name())
+
+    def get_swap_to_be_func_decl(self):
+        return "void %s(%s *msg)" % (
+            self.get_swap_to_be_func_name(), self.get_c_name())
+
+    def get_swap_to_be_func_def(self):
+        return "\n".join([
+            "%s" % self.get_swap_to_be_func_decl(),
+            "{",
+            ("  VAPI_DBG(\"Swapping `%s'@%%p to big endian\", msg);" %
+                self.get_c_name()),
+            "  %s(&msg->header);" % self.header.get_swap_to_be_func_name()
+            if self.header is not None else "",
+            "  %s(&msg->payload);" % self.get_swap_payload_to_be_func_name()
+            if self.has_payload() else "",
+            "}",
+        ])
+
+    def get_swap_to_host_func_def(self):
+        return "\n".join([
+            "%s" % self.get_swap_to_host_func_decl(),
+            "{",
+            ("  VAPI_DBG(\"Swapping `%s'@%%p to host byte order\", msg);" %
+                self.get_c_name()),
+            "  %s(&msg->header);" % self.header.get_swap_to_host_func_name()
+            if self.header is not None else "",
+            "  %s(&msg->payload);" % self.get_swap_payload_to_host_func_name()
+            if self.has_payload() else "",
+            "}",
+        ])
+
+    def get_op_func_name(self):
+        return "vapi_%s" % self.name
+
+    def get_op_func_decl(self):
+        if self.reply.has_payload():
+            return "vapi_error_e %s(%s)" % (
+                self.get_op_func_name(),
+                ",\n  ".join([
+                    'struct vapi_ctx_s *ctx',
+                    '%s *msg' % self.get_c_name(),
+                    'vapi_error_e (*callback)(struct vapi_ctx_s *ctx',
+                    '                         void *callback_ctx',
+                    '                         vapi_error_e rv',
+                    '                         bool is_last',
+                    '                         %s *reply)' %
+                    self.reply.get_payload_struct_name(),
+                    'void *callback_ctx',
+                ])
+            )
+        else:
+            return "vapi_error_e %s(%s)" % (
+                self.get_op_func_name(),
+                ",\n  ".join([
+                    'struct vapi_ctx_s *ctx',
+                    '%s *msg' % self.get_c_name(),
+                    'vapi_error_e (*callback)(struct vapi_ctx_s *ctx',
+                    '                         void *callback_ctx',
+                    '                         vapi_error_e rv',
+                    '                         bool is_last)',
+                    'void *callback_ctx',
+                ])
+            )
+
+    def get_op_func_def(self):
+        return "\n".join([
+            "%s" % self.get_op_func_decl(),
+            "{",
+            "  if (!msg || !callback) {",
+            "    return VAPI_EINVAL;",
+            "  }",
+            "  if (vapi_is_nonblocking(ctx) && vapi_requests_full(ctx)) {",
+            "    return VAPI_EAGAIN;",
+            "  }",
+            "  vapi_error_e rv;",
+            "  if (VAPI_OK != (rv = vapi_producer_lock (ctx))) {",
+            "    return rv;",
+            "  }",
+            "  u32 req_context = vapi_gen_req_context(ctx);",
+            "  msg->header.context = req_context;",
+            "  %s(msg);" % self.get_swap_to_be_func_name(),
+            ("  if (VAPI_OK == (rv = vapi_send_with_control_ping "
+                "(ctx, msg, req_context))) {"
+                if self.is_dump() else
+                "  if (VAPI_OK == (rv = vapi_send (ctx, msg))) {"
+             ),
+            ("    vapi_store_request(ctx, req_context, %s, "
+             "(vapi_cb_t)callback, callback_ctx);" %
+             ("true" if self.is_dump() else "false")),
+            "    if (VAPI_OK != vapi_producer_unlock (ctx)) {",
+            "      abort (); /* this really shouldn't happen */",
+            "    }",
+            "    if (vapi_is_nonblocking(ctx)) {",
+            "      rv = VAPI_OK;",
+            "    } else {",
+            "      rv = vapi_dispatch(ctx);",
+            "    }",
+            "  } else {",
+            "    %s(msg);" % self.get_swap_to_host_func_name(),
+            "    if (VAPI_OK != vapi_producer_unlock (ctx)) {",
+            "      abort (); /* this really shouldn't happen */",
+            "    }",
+            "  }",
+            "  return rv;",
+            "}",
+            "",
+        ])
+
+    def get_event_cb_func_decl(self):
+        if not self.is_reply():
+            raise Exception(
+                "Cannot register event callback for non-reply function")
+        if self.has_payload():
+            return "\n".join([
+                "void vapi_set_%s_event_cb (" %
+                self.get_c_name(),
+                "  struct vapi_ctx_s *ctx, ",
+                ("  vapi_error_e (*callback)(struct vapi_ctx_s *ctx, "
+                 "void *callback_ctx, %s *payload)," %
+                 self.get_payload_struct_name()),
+                "  void *callback_ctx)",
+            ])
+        else:
+            return "\n".join([
+                "void vapi_set_%s_event_cb (" %
+                self.get_c_name(),
+                "  struct vapi_ctx_s *ctx, ",
+                "  vapi_error_e (*callback)(struct vapi_ctx_s *ctx, "
+                "void *callback_ctx),",
+                "  void *callback_ctx)",
+            ])
+
+    def get_event_cb_func_def(self):
+        if not self.is_reply():
+            raise Exception(
+                "Cannot register event callback for non-reply function")
+        return "\n".join([
+            "%s" % self.get_event_cb_func_decl(),
+            "{",
+            ("  vapi_set_event_cb(ctx, %s, (vapi_event_cb)callback, "
+             "callback_ctx);" %
+             self.get_msg_id_name()),
+            "}"])
+
+    def get_c_metadata_struct_name(self):
+        return "__vapi_metadata_%s" % self.name
+
+    def get_c_constructor(self):
+        has_context = False
+        if self.header is not None:
+            has_context = self.header.has_field('context')
+        return '\n'.join([
+            'static void __attribute__((constructor)) __vapi_constructor_%s()'
+            % self.name,
+            '{',
+            '  static const char name[] = "%s";' % self.name,
+            '  static const char name_with_crc[] = "%s_%s";'
+            % (self.name, self.crc[2:]),
+            '  static vapi_message_desc_t %s = {' %
+            self.get_c_metadata_struct_name(),
+            '    name,',
+            '    sizeof(name) - 1,',
+            '    name_with_crc,',
+            '    sizeof(name_with_crc) - 1,',
+            '    true,' if has_context else '    false,',
+            '    offsetof(%s, context),' % self.header.get_c_name()
+            if has_context else '    0,',
+            ('    offsetof(%s, payload),' % self.get_c_name())
+            if self.has_payload() else '-1,',
+            '    sizeof(%s),' % self.get_c_name(),
+            '    (generic_swap_fn_t)%s,' % self.get_swap_to_be_func_name(),
+            '    (generic_swap_fn_t)%s,' % self.get_swap_to_host_func_name(),
+            '    ~0,',
+            '  };',
+            '',
+            '  %s = vapi_register_msg(&%s);' %
+            (self.get_msg_id_name(), self.get_c_metadata_struct_name()),
+            '  VAPI_DBG("Assigned msg id %%d to %s", %s);' %
+            (self.name, self.get_msg_id_name()),
+            '}',
+        ])
+
+
+vapi_send_with_control_ping = """
+static inline vapi_error_e
+vapi_send_with_control_ping (vapi_ctx_t ctx, void *msg, u32 context)
+{
+  vapi_msg_control_ping *ping = vapi_alloc_control_ping (ctx);
+  if (!ping)
+    {
+      return VAPI_ENOMEM;
+    }
+  ping->header.context = context;
+  vapi_msg_control_ping_hton (ping);
+  return vapi_send2 (ctx, msg, ping);
+}
+"""
+
+
+def gen_json_header(parser, logger, j, io):
+    logger.info("Generating header `%s'" % io.name)
+    orig_stdout = sys.stdout
+    sys.stdout = io
+    include_guard = "__included_%s" % (
+        j.replace(".", "_").replace("/", "_").replace("-", "_"))
+    print("#ifndef %s" % include_guard)
+    print("#define %s" % include_guard)
+    print("")
+    print("#include ")
+    print("")
+    if io.name == "vpe.api.vapi.h":
+        print("static inline vapi_error_e vapi_send_with_control_ping "
+              "(vapi_ctx_t ctx, void * msg, u32 context);")
+        print("")
+    for m in parser.messages_by_json[j].values():
+        print("extern vapi_msg_id_t %s;" % m.get_msg_id_name())
+    print("")
+    for t in parser.types_by_json[j].values():
+        try:
+            print("%s" % t.get_c_def())
+            print("")
+        except:
+            pass
+    for t in parser.types_by_json[j].values():
+        print("%s;" % t.get_swap_to_be_func_decl())
+        print("")
+        print("%s;" % t.get_swap_to_host_func_decl())
+        print("")
+    for m in parser.messages_by_json[j].values():
+        print("%s" % m.get_c_def())
+        print("")
+    for m in parser.messages_by_json[j].values():
+        if not m.is_reply():
+            print("%s;" % m.get_alloc_func_decl())
+            print("")
+            print("%s;" % m.get_op_func_decl())
+        if m.has_payload():
+            print("%s;" % m.get_swap_payload_to_be_func_decl())
+            print("")
+            print("%s;" % m.get_swap_payload_to_host_func_decl())
+            print("")
+        print("%s;" % m.get_calc_msg_size_func_decl())
+        print("")
+        print("%s;" % m.get_swap_to_host_func_decl())
+        print("")
+        print("%s;" % m.get_swap_to_be_func_decl())
+        print("")
+    for m in parser.messages_by_json[j].values():
+        if not m.is_reply():
+            continue
+        print("%s;" % m.get_event_cb_func_decl())
+        print("")
+
+    if io.name == "vpe.api.vapi.h":
+        print("%s" % vapi_send_with_control_ping)
+        print("")
+
+    print("#endif")
+    sys.stdout = orig_stdout
+
+
+def gen_json_code(parser, logger, j, io):
+    logger.info("Generating code `%s'" % io.name)
+    orig_stdout = sys.stdout
+    sys.stdout = io
+    print("#include <%s>" % json_to_header_name(j))
+    print("#include ")
+    print("#include ")
+    print("#include ")
+    print("#include ")
+    print("#include ")
+    print("")
+    for t in parser.types_by_json[j].values():
+        print("%s" % t.get_swap_to_be_func_def())
+        print("")
+        print("%s" % t.get_swap_to_host_func_def())
+        print("")
+    for m in parser.messages_by_json[j].values():
+        if m.has_payload():
+            print("%s" % m.get_swap_payload_to_be_func_def())
+            print("")
+            print("%s" % m.get_swap_payload_to_host_func_def())
+            print("")
+        print("%s" % m.get_calc_msg_size_func_def())
+        print("")
+        print("%s" % m.get_swap_to_be_func_def())
+        print("")
+        print("%s" % m.get_swap_to_host_func_def())
+        print("")
+    for m in parser.messages_by_json[j].values():
+        if m.is_reply():
+            continue
+        print("%s" % m.get_alloc_func_def())
+        print("")
+        print("%s" % m.get_op_func_def())
+        print("")
+    print("")
+    for m in parser.messages_by_json[j].values():
+        print("%s" % m.get_c_constructor())
+        print("")
+    print("")
+    for m in parser.messages_by_json[j].values():
+        if not m.is_reply():
+            continue
+        print("%s;" % m.get_event_cb_func_def())
+        print("")
+    print("")
+    for m in parser.messages_by_json[j].values():
+        print("vapi_msg_id_t %s;" % m.get_msg_id_name())
+    sys.stdout = orig_stdout
+
+
+def gen_json_unified_header(parser, logger, j, io):
+    logger.info("Generating header `%s'" % io.name)
+    orig_stdout = sys.stdout
+    sys.stdout = io
+    include_guard = "__included_%s" % (
+        j.replace(".", "_").replace("/", "_").replace("-", "_"))
+    print("#ifndef %s" % include_guard)
+    print("#define %s" % include_guard)
+    print("")
+    print("#include ")
+    print("#include ")
+    print("#include ")
+    print("#include ")
+    print("#include ")
+    print("#include ")
+    if io.name == "vpe.api.vapi.h":
+        print("")
+        print("static inline vapi_error_e vapi_send_with_control_ping "
+              "(vapi_ctx_t ctx, void * msg, u32 context);")
+    else:
+        print("#include ")
+    print("")
+    for m in parser.messages_by_json[j].values():
+        print("extern vapi_msg_id_t %s;" % m.get_msg_id_name())
+    print("")
+    print("#define DEFINE_VAPI_MSG_IDS_%s\\" %
+          j.replace(".", "_").replace("/", "_").replace("-", "_").upper())
+    print("\\\n".join([
+        "  vapi_msg_id_t %s;" % m.get_msg_id_name()
+        for m in parser.messages_by_json[j].values()
+    ]))
+    print("")
+    print("")
+    for t in parser.types_by_json[j].values():
+        try:
+            print("%s" % t.get_c_def())
+            print("")
+        except:
+            pass
+    for m in parser.messages_by_json[j].values():
+        print("%s" % m.get_c_def())
+        print("")
+
+    print("")
+    function_attrs = "static inline "
+    for t in parser.types_by_json[j].values():
+        print("%s%s" % (function_attrs, t.get_swap_to_be_func_def()))
+        print("")
+        print("%s%s" % (function_attrs, t.get_swap_to_host_func_def()))
+        print("")
+    for m in parser.messages_by_json[j].values():
+        if m.has_payload():
+            print("%s%s" % (function_attrs,
+                            m.get_swap_payload_to_be_func_def()))
+            print("")
+            print("%s%s" % (function_attrs,
+                            m.get_swap_payload_to_host_func_def()))
+            print("")
+        print("%s%s" % (function_attrs, m.get_calc_msg_size_func_def()))
+        print("")
+        print("%s%s" % (function_attrs, m.get_swap_to_be_func_def()))
+        print("")
+        print("%s%s" % (function_attrs, m.get_swap_to_host_func_def()))
+        print("")
+    for m in parser.messages_by_json[j].values():
+        if m.is_reply():
+            continue
+        print("%s%s" % (function_attrs, m.get_alloc_func_def()))
+        print("")
+        print("%s%s" % (function_attrs, m.get_op_func_def()))
+        print("")
+    print("")
+    for m in parser.messages_by_json[j].values():
+        print("%s" % m.get_c_constructor())
+        print("")
+    print("")
+    for m in parser.messages_by_json[j].values():
+        if not m.is_reply():
+            continue
+        print("%s%s;" % (function_attrs, m.get_event_cb_func_def()))
+        print("")
+    print("")
+
+    if io.name == "vpe.api.vapi.h":
+        print("%s" % vapi_send_with_control_ping)
+        print("")
+
+    print("#endif")
+    sys.stdout = orig_stdout
+
+
+def json_to_header_name(json_name):
+    if json_name.endswith(".json"):
+        return "%s.vapi.h" % os.path.splitext(json_name)[0]
+    raise Exception("Unexpected json name `%s'!" % json_name)
+
+
+def json_to_code_name(json_name):
+    if json_name.endswith(".json"):
+        return "%s.vapi.c" % os.path.splitext(json_name)[0]
+    raise Exception("Unexpected json name `%s'!" % json_name)
+
+
+def gen_c_headers_and_code(parser, logger, prefix):
+    if prefix == "" or prefix is None:
+        prefix = ""
+    else:
+        prefix = "%s/" % prefix
+    for j in parser.json_files:
+        with open('%s%s' % (prefix, json_to_header_name(j)), "w") as io:
+            gen_json_header(parser, logger, j, io)
+        with open('%s%s' % (prefix, json_to_code_name(j)), "w") as io:
+            gen_json_code(parser, logger, j, io)
+
+
+def gen_c_unified_headers(parser, logger, prefix):
+    if prefix == "" or prefix is None:
+        prefix = ""
+    else:
+        prefix = "%s/" % prefix
+    for j in parser.json_files:
+        with open('%s%s' % (prefix, json_to_header_name(j)), "w") as io:
+            gen_json_unified_header(parser, logger, j, io)
+
+
+if __name__ == '__main__':
+    try:
+        verbose = int(os.getenv("V", 0))
+    except:
+        verbose = 0
+
+    if verbose >= 2:
+        log_level = 10
+    elif verbose == 1:
+        log_level = 20
+    else:
+        log_level = 40
+
+    logging.basicConfig(stream=sys.stdout, level=log_level)
+    logger = logging.getLogger("VAPI C GEN")
+    logger.setLevel(log_level)
+
+    argparser = argparse.ArgumentParser(description="VPP JSON API parser")
+    argparser.add_argument('files', metavar='api-file', action='append',
+                           type=str, help='json api file'
+                           '(may be specified multiple times)')
+    argparser.add_argument('--prefix', action='store', default=None,
+                           help='path prefix')
+    args = argparser.parse_args()
+
+    jsonparser = JsonParser(logger, args.files,
+                            simple_type_class=CSimpleType,
+                            struct_type_class=CStructType,
+                            field_class=CField,
+                            message_class=CMessage)
+
+    # not using the model of having separate generated header and code files
+    # with generated symbols present in shared library (per discussion with
+    # Damjan), to avoid symbol version issues in .so
+    # gen_c_headers_and_code(jsonparser, logger, args.prefix)
+
+    gen_c_unified_headers(jsonparser, logger, args.prefix)
+
+    for e in jsonparser.exceptions:
+        logger.error(e)
diff --git a/src/vpp-api/vapi/vapi_dbg.h b/src/vpp-api/vapi/vapi_dbg.h
new file mode 100644
index 00000000..95a80089
--- /dev/null
+++ b/src/vpp-api/vapi/vapi_dbg.h
@@ -0,0 +1,76 @@
+/*
+ *------------------------------------------------------------------
+ * Copyright (c) 2017 Cisco and/or its affiliates.
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at:
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ *------------------------------------------------------------------
+ */
+
+#ifndef __included_vapi_debug_h__
+#define __included_vapi_debug_h__
+
+/* controls debug prints */
+#define VAPI_DEBUG (0)
+#define VAPI_DEBUG_CONNECT (0)
+#define VAPI_DEBUG_ALLOC (0)
+
+#if VAPI_DEBUG
+#include 
+#define VAPI_DEBUG_FILE_DEF           \
+  static const char *__file = NULL;   \
+  {                                   \
+    __file = strrchr (__FILE__, '/'); \
+    if (__file)                       \
+      {                               \
+        ++__file;                     \
+      }                               \
+    else                              \
+      {                               \
+        __file = __FILE__;            \
+      }                               \
+  }
+
+#define VAPI_DBG(fmt, ...)                                       \
+  do                                                             \
+    {                                                            \
+      VAPI_DEBUG_FILE_DEF                                        \
+      printf ("DBG:%s:%d:%s():" fmt, __file, __LINE__, __func__, \
+              ##__VA_ARGS__);                                    \
+      printf ("\n");                                             \
+      fflush (stdout);                                           \
+    }                                                            \
+  while (0);
+
+#define VAPI_ERR(fmt, ...)                                       \
+  do                                                             \
+    {                                                            \
+      VAPI_DEBUG_FILE_DEF                                        \
+      printf ("ERR:%s:%d:%s():" fmt, __file, __LINE__, __func__, \
+              ##__VA_ARGS__);                                    \
+      printf ("\n");                                             \
+      fflush (stdout);                                           \
+    }                                                            \
+  while (0);
+#else
+#define VAPI_DBG(...)
+#define VAPI_ERR(...)
+#endif
+
+#endif /* __included_vapi_debug_h__ */
+
+/*
+ * fd.io coding-style-patch-verification: ON
+ *
+ * Local Variables:
+ * eval: (c-set-style "gnu")
+ * End:
+ */
diff --git a/src/vpp-api/vapi/vapi_internal.h b/src/vpp-api/vapi/vapi_internal.h
new file mode 100644
index 00000000..5b85788d
--- /dev/null
+++ b/src/vpp-api/vapi/vapi_internal.h
@@ -0,0 +1,126 @@
+/*
+ *------------------------------------------------------------------
+ * Copyright (c) 2017 Cisco and/or its affiliates.
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at:
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ *------------------------------------------------------------------
+ */
+
+#ifndef VAPI_INTERNAL_H
+#define VAPI_INTERNAL_H
+
+#include 
+#include 
+
+/**
+ * @file vapi_internal.h
+ *
+ * internal vpp api C declarations
+ *
+ * This file contains internal vpp api C declarations. It's not intended to be
+ * used by the client programmer and the API defined here might change at any
+ * time..
+ */
+
+struct vapi_ctx_s;
+
+typedef struct __attribute__ ((__packed__))
+{
+  u16 _vl_msg_id;
+  u32 context;
+} vapi_type_msg_header1_t;
+
+typedef struct __attribute__ ((__packed__))
+{
+  u16 _vl_msg_id;
+  u32 client_index;
+  u32 context;
+} vapi_type_msg_header2_t;
+
+static inline void
+vapi_type_msg_header1_t_hton (vapi_type_msg_header1_t * h)
+{
+  h->_vl_msg_id = htobe16 (h->_vl_msg_id);
+}
+
+static inline void
+vapi_type_msg_header1_t_ntoh (vapi_type_msg_header1_t * h)
+{
+  h->_vl_msg_id = be16toh (h->_vl_msg_id);
+}
+
+static inline void
+vapi_type_msg_header2_t_hton (vapi_type_msg_header2_t * h)
+{
+  h->_vl_msg_id = htobe16 (h->_vl_msg_id);
+}
+
+static inline void
+vapi_type_msg_header2_t_ntoh (vapi_type_msg_header2_t * h)
+{
+  h->_vl_msg_id = be16toh (h->_vl_msg_id);
+}
+
+
+#include 
+
+typedef vapi_error_e (*vapi_cb_t) (struct vapi_ctx_s *, void *, vapi_error_e,
+				   bool, void *);
+
+typedef void (*generic_swap_fn_t) (void *payload);
+
+typedef struct
+{
+  const char *name;
+  size_t name_len;
+  const char *name_with_crc;
+  size_t name_with_crc_len;
+  bool has_context;
+  size_t context_offset;
+  size_t payload_offset;
+  size_t size;
+  generic_swap_fn_t swap_to_be;
+  generic_swap_fn_t swap_to_host;
+  vapi_msg_id_t id;		/* assigned at run-time */
+} vapi_message_desc_t;
+
+typedef struct
+{
+  const char *name;
+  int payload_offset;
+  size_t size;
+  void (*swap_to_be) (void *payload);
+  void (*swap_to_host) (void *payload);
+} vapi_event_desc_t;
+
+extern bool *__vapi_msg_is_with_context;
+
+vapi_msg_id_t vapi_register_msg (vapi_message_desc_t * msg);
+u16 vapi_lookup_vl_msg_id (vapi_ctx_t ctx, vapi_msg_id_t id);
+int vapi_get_client_index (vapi_ctx_t ctx);
+bool vapi_is_nonblocking (vapi_ctx_t ctx);
+bool vapi_requests_full (vapi_ctx_t ctx);
+size_t vapi_get_request_count (vapi_ctx_t ctx);
+size_t vapi_get_max_request_count (vapi_ctx_t ctx);
+u32 vapi_gen_req_context (vapi_ctx_t ctx);
+void vapi_store_request (vapi_ctx_t ctx, u32 context, bool is_dump,
+			 vapi_cb_t callback, void *callback_ctx);
+int vapi_get_payload_offset (vapi_msg_id_t id);
+void (*vapi_get_swap_to_host_func (vapi_msg_id_t id)) (void *payload);
+void (*vapi_get_swap_to_be_func (vapi_msg_id_t id)) (void *payload);
+size_t vapi_get_message_size (vapi_msg_id_t id);
+size_t vapi_get_context_offset (vapi_msg_id_t id);
+
+vapi_error_e vapi_producer_lock (vapi_ctx_t ctx);
+vapi_error_e vapi_producer_unlock (vapi_ctx_t ctx);
+
+#endif
diff --git a/src/vpp-api/vapi/vapi_json_parser.py b/src/vpp-api/vapi/vapi_json_parser.py
new file mode 100644
index 00000000..57a22383
--- /dev/null
+++ b/src/vpp-api/vapi/vapi_json_parser.py
@@ -0,0 +1,303 @@
+#!/usr/bin/env python3
+
+import json
+
+
+def msg_is_reply(name):
+    return name.endswith('_reply') or name.endswith('_details') \
+        or name.endswith('_event') or name.endswith('_counters')
+
+
+class ParseError (Exception):
+    pass
+
+
+magic_prefix = "vl_api_"
+magic_suffix = "_t"
+
+
+def remove_magic(what):
+    if what.startswith(magic_prefix) and what.endswith(magic_suffix):
+        return what[len(magic_prefix): - len(magic_suffix)]
+    return what
+
+
+class Field:
+
+    def __init__(
+            self,
+            field_name,
+            field_type,
+            array_len=None,
+            nelem_field=None):
+        self.name = field_name
+        self.type = field_type
+        self.len = array_len
+        self.nelem_field = nelem_field
+
+    def __str__(self):
+        if self.len is None:
+            return "name: %s, type: %s" % (self.name, self.type)
+        elif self.len > 0:
+            return "name: %s, type: %s, length: %s" % (self.name, self.type,
+                                                       self.len)
+        else:
+            return ("name: %s, type: %s, variable length stored in: %s" %
+                    (self.name, self.type, self.nelem_field))
+
+
+class Type:
+    def __init__(self, name):
+        self.name = name
+
+
+class SimpleType (Type):
+
+    def __init__(self, name):
+        super().__init__(name)
+
+    def __str__(self):
+        return self.name
+
+
+def get_msg_header_defs(struct_type_class, field_class, typedict):
+    return [
+        struct_type_class(['msg_header1_t',
+                           ['u16', '_vl_msg_id'],
+                           ['u32', 'context'],
+                           ],
+                          typedict, field_class
+                          ),
+        struct_type_class(['msg_header2_t',
+                           ['u16', '_vl_msg_id'],
+                           ['u32', 'client_index'],
+                           ['u32', 'context'],
+                           ],
+                          typedict, field_class
+                          ),
+    ]
+
+
+class Struct:
+
+    def __init__(self, name, fields):
+        self.name = name
+        self.fields = fields
+        self.field_names = [n.name for n in self.fields]
+
+
+class Message:
+
+    def __init__(self, logger, definition, typedict,
+                 struct_type_class, simple_type_class, field_class):
+        self.logger = logger
+        m = definition
+        logger.debug("Parsing message definition `%s'" % m)
+        name = m[0]
+        self.name = name
+        logger.debug("Message name is `%s'" % name)
+        ignore = True
+        self.header = None
+        fields = []
+        for header in get_msg_header_defs(struct_type_class, field_class,
+                                          typedict):
+            logger.debug("Probing header `%s'" % header.name)
+            if header.is_part_of_def(m[1:]):
+                self.header = header
+                logger.debug("Found header `%s'" % header.name)
+                fields.append(field_class(field_name='header',
+                                          field_type=self.header))
+                ignore = False
+                break
+        if ignore and not msg_is_reply(name):
+            raise ParseError("While parsing message `%s': could not find all "
+                             "common header fields" % name)
+        for field in m[1:]:
+            if len(field) == 1 and 'crc' in field:
+                self.crc = field['crc']
+                logger.debug("Found CRC `%s'" % self.crc)
+                continue
+            else:
+                field_type = field[0]
+                if field_type in typedict:
+                    field_type = typedict[field_type]
+                else:
+                    field_type = typedict[remove_magic(field_type)]
+                if len(field) == 2:
+                    if self.header is not None and\
+                            self.header.has_field(field[1]):
+                        continue
+                    p = field_class(field_name=field[1],
+                                    field_type=field_type)
+                elif len(field) == 3:
+                    if field[2] == 0:
+                        raise ParseError(
+                            "While parsing message `%s': variable length "
+                            "array `%s' doesn't have reference to member "
+                            "containing the actual length" % (
+                                name, field[1]))
+                    p = field_class(
+                        field_name=field[1],
+                        field_type=field_type,
+                        array_len=field[2])
+                elif len(field) == 4:
+                    nelem_field = None
+                    for f in fields:
+                        if f.name == field[3]:
+                            nelem_field = f
+                    if nelem_field is None:
+                        raise ParseError(
+                            "While parsing message `%s': couldn't find "
+                            "variable length array `%s' member containing "
+                            "the actual length `%s'" % (
+                                name, field[1], field[3]))
+                    p = field_class(
+                        field_name=field[1],
+                        field_type=field_type,
+                        array_len=field[2],
+                        nelem_field=nelem_field)
+                else:
+                    raise Exception("Don't know how to parse message "
+                                    "definition for message `%s': `%s'" %
+                                    (m, m[1:]))
+                logger.debug("Parsed field `%s'" % p)
+                fields.append(p)
+        self.fields = fields
+
+    def is_dump(self):
+        return self.name.endswith('_dump')
+
+    def is_reply(self):
+        return msg_is_reply(self.name)
+
+
+class StructType (Type, Struct):
+
+    def __init__(self, definition, typedict, field_class):
+        t = definition
+        name = t[0]
+        fields = []
+        for field in t[1:]:
+            if len(field) == 1 and 'crc' in field:
+                self.crc = field['crc']
+                continue
+            elif len(field) == 2:
+                p = field_class(field_name=field[1],
+                                field_type=typedict[field[0]])
+            elif len(field) == 3:
+                if field[2] == 0:
+                    raise ParseError("While parsing type `%s': array `%s' has "
+                                     "variable length" % (name, field[1]))
+                p = field_class(field_name=field[1],
+                                field_type=typedict[field[0]],
+                                array_len=field[2])
+            else:
+                raise ParseError(
+                    "Don't know how to parse type definition for "
+                    "type `%s': `%s'" % (t, t[1:]))
+            fields.append(p)
+        Type.__init__(self, name)
+        Struct.__init__(self, name, fields)
+
+    def has_field(self, name):
+        return name in self.field_names
+
+    def is_part_of_def(self, definition):
+        for idx in range(len(self.fields)):
+            field = definition[idx]
+            p = self.fields[idx]
+            if field[1] != p.name:
+                return False
+            if field[0] != p.type.name:
+                raise ParseError(
+                    "Unexpected field type `%s' (should be `%s'), "
+                    "while parsing msg/def/field `%s/%s/%s'" %
+                    (field[0], p.type, p.name, definition, field))
+        return True
+
+
+class JsonParser:
+    def __init__(self, logger, files, simple_type_class=SimpleType,
+                 struct_type_class=StructType, field_class=Field,
+                 message_class=Message):
+        self.messages = {}
+        self.types = {
+            x: simple_type_class(x) for x in [
+                'i8', 'i16', 'i32', 'i64',
+                'u8', 'u16', 'u32', 'u64',
+                'f64'
+            ]
+        }
+
+        self.simple_type_class = simple_type_class
+        self.struct_type_class = struct_type_class
+        self.field_class = field_class
+        self.message_class = message_class
+
+        self.exceptions = []
+        self.json_files = []
+        self.types_by_json = {}
+        self.messages_by_json = {}
+        self.logger = logger
+        for f in files:
+            self.parse_json_file(f)
+        self.finalize_parsing()
+
+    def parse_json_file(self, path):
+        self.logger.info("Parsing json api file: `%s'" % path)
+        self.json_files.append(path)
+        self.types_by_json[path] = {}
+        self.messages_by_json[path] = {}
+        with open(path) as f:
+            j = json.load(f)
+            for t in j['types']:
+                try:
+                    type_ = self.struct_type_class(t, self.types,
+                                                   self.field_class)
+                    if type_.name in self.types:
+                        raise ParseError("Duplicate type `%s'" % type_.name)
+                except ParseError as e:
+                    self.exceptions.append(e)
+                    continue
+                self.types[type_.name] = type_
+                self.types_by_json[path][type_.name] = type_
+            for m in j['messages']:
+                try:
+                    msg = self.message_class(self.logger, m, self.types,
+                                             self.struct_type_class,
+                                             self.simple_type_class,
+                                             self.field_class)
+                    if msg.name in self.messages:
+                        raise ParseError("Duplicate message `%s'" % msg.name)
+                except ParseError as e:
+                    self.exceptions.append(e)
+                    continue
+                self.messages[msg.name] = msg
+                self.messages_by_json[path][msg.name] = msg
+
+    def get_reply(self, message):
+        if self.messages[message].is_dump():
+            return self.messages["%s_details" % message[:-len("_dump")]]
+        return self.messages["%s_reply" % message]
+
+    def finalize_parsing(self):
+        if len(self.messages) == 0:
+            for e in self.exceptions:
+                self.logger.error(e)
+            raise Exception("No messages parsed.")
+        for jn, j in self.messages_by_json.items():
+            remove = []
+            for n, m in j.items():
+                try:
+                    if not m.is_reply():
+                        try:
+                            m.reply = self.get_reply(n)
+                        except:
+                            raise ParseError(
+                                "Cannot find reply to message `%s'" % n)
+                except ParseError as e:
+                    self.exceptions.append(e)
+                    remove.append(n)
+
+            self.messages_by_json[jn] = {
+                k: v for k, v in j.items() if k not in remove}
diff --git a/test/Makefile b/test/Makefile
index 72b4dac7..132ebee6 100644
--- a/test/Makefile
+++ b/test/Makefile
@@ -107,7 +107,11 @@ sanity: verify-no-running-vpp
 	         echo \"*******************************************************************\" &&\
 		 false)"
 
-test: verify-python-path $(PAPI_INSTALL_DONE) sanity reset 
+.PHONY: ext
+ext:
+	make -C ext
+
+test: verify-python-path $(PAPI_INSTALL_DONE) ext sanity reset
 	$(call retest-func)
 
 retest: verify-python-path sanity reset
diff --git a/test/ext/Makefile b/test/ext/Makefile
new file mode 100644
index 00000000..4a45fef6
--- /dev/null
+++ b/test/ext/Makefile
@@ -0,0 +1,17 @@
+BINDIR = $(BR)/vapi_test/
+BIN = $(addprefix $(BINDIR), vapi_test)
+LIBS = -L$(VPP_TEST_BUILD_DIR)/vpp/.libs/ -L$(VPP_TEST_BUILD_DIR)/vpp/vpp-api/vapi/.libs/ -lvppinfra -lvlibmemoryclient -lsvm -lpthread -lcheck -lsubunit -lrt -lm -lvapiclient
+CFLAGS = -ggdb -O0 -Wall -pthread -I$(WS_ROOT)/src -I$(VPP_TEST_BUILD_DIR)/vpp/vpp-api/vapi -I$(WS_ROOT)/src/vpp-api/vapi/
+
+all: $(BIN)
+
+$(BINDIR):
+	mkdir -p $(BINDIR)
+
+SRC = vapi_test.c
+
+$(BIN): $(SRC) $(BINDIR) $(VPP_TEST_BUILD_DIR)/vpp/vpp-api/vapi/.libs/libvapiclient.so $(VPP_TEST_BUILD_DIR)/vpp/.libs/libvppinfra.so $(VPP_TEST_BUILD_DIR)/vpp/.libs/libvlibmemoryclient.so $(VPP_TEST_BUILD_DIR)/vpp/.libs/libsvm.so
+	gcc -ggdb -o $@ $(SRC) $(CFLAGS) $(LIBS)
+
+clean:
+	rm -rf $(BINDIR)
diff --git a/test/ext/vapi_test.c b/test/ext/vapi_test.c
new file mode 100644
index 00000000..eca6be7d
--- /dev/null
+++ b/test/ext/vapi_test.c
@@ -0,0 +1,1152 @@
+/*
+ *------------------------------------------------------------------
+ * Copyright (c) 2017 Cisco and/or its affiliates.
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at:
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ *------------------------------------------------------------------
+ */
+
+#include 
+#include 
+#include 
+#include 
+#include 
+#include 
+#include 
+#include 
+#include 
+#include 
+#include 
+#include 
+
+DEFINE_VAPI_MSG_IDS_VPE_API_JSON;
+DEFINE_VAPI_MSG_IDS_INTERFACE_API_JSON;
+DEFINE_VAPI_MSG_IDS_L2_API_JSON;
+DEFINE_VAPI_MSG_IDS_STATS_API_JSON;
+
+static char *app_name = NULL;
+static char *api_prefix = NULL;
+static const int max_outstanding_requests = 64;
+static const int response_queue_size = 32;
+
+START_TEST (test_invalid_values)
+{
+  vapi_ctx_t ctx;
+  vapi_error_e rv = vapi_ctx_alloc (&ctx);
+  ck_assert_int_eq (VAPI_OK, rv);
+  vapi_msg_show_version *sv = vapi_alloc_show_version (ctx);
+  ck_assert_ptr_eq (NULL, sv);
+  rv = vapi_send (ctx, sv);
+  ck_assert_int_eq (VAPI_EINVAL, rv);
+  rv = vapi_connect (ctx, app_name, api_prefix, max_outstanding_requests,
+		     response_queue_size, VAPI_MODE_BLOCKING);
+  ck_assert_int_eq (VAPI_OK, rv);
+  rv = vapi_send (ctx, NULL);
+  ck_assert_int_eq (VAPI_EINVAL, rv);
+  rv = vapi_send (NULL, NULL);
+  ck_assert_int_eq (VAPI_EINVAL, rv);
+  rv = vapi_recv (NULL, NULL, NULL);
+  ck_assert_int_eq (VAPI_EINVAL, rv);
+  rv = vapi_recv (ctx, NULL, NULL);
+  ck_assert_int_eq (VAPI_EINVAL, rv);
+  vapi_msg_show_version_reply *reply;
+  rv = vapi_recv (ctx, (void **) &reply, NULL);
+  ck_assert_int_eq (VAPI_EINVAL, rv);
+  rv = vapi_disconnect (ctx);
+  ck_assert_int_eq (VAPI_OK, rv);
+  vapi_ctx_free (ctx);
+}
+
+END_TEST;
+
+START_TEST (test_hton_1)
+{
+  const u16 _vl_msg_id = 1;
+  vapi_type_msg_header1_t h;
+  h._vl_msg_id = _vl_msg_id;
+  vapi_type_msg_header1_t_hton (&h);
+  ck_assert_int_eq (be16toh (h._vl_msg_id), _vl_msg_id);
+}
+
+END_TEST;
+
+START_TEST (test_hton_2)
+{
+  const u16 _vl_msg_id = 1;
+  const u32 client_index = 3;
+  vapi_type_msg_header2_t h;
+  h._vl_msg_id = _vl_msg_id;
+  h.client_index = client_index;
+  vapi_type_msg_header2_t_hton (&h);
+  ck_assert_int_eq (be16toh (h._vl_msg_id), _vl_msg_id);
+  ck_assert_int_eq (h.client_index, client_index);
+}
+
+END_TEST;
+
+START_TEST (test_hton_3)
+{
+  const size_t data_size = 10;
+  vapi_msg_vnet_interface_combined_counters *m =
+    malloc (sizeof (vapi_msg_vnet_interface_combined_counters) +
+	    data_size * sizeof (vapi_type_vlib_counter));
+  ck_assert_ptr_ne (NULL, m);
+  vapi_payload_vnet_interface_combined_counters *p = &m->payload;
+  const u16 _vl_msg_id = 1;
+  p->_vl_msg_id = _vl_msg_id;
+  const u32 first_sw_if_index = 2;
+  p->first_sw_if_index = first_sw_if_index;
+  p->count = data_size;
+  const u64 packets = 1234;
+  const u64 bytes = 2345;
+  int i;
+  for (i = 0; i < data_size; ++i)
+    {
+      p->data[i].packets = packets;
+      p->data[i].bytes = bytes;
+    }
+  vapi_msg_vnet_interface_combined_counters_hton (m);
+  ck_assert_int_eq (_vl_msg_id, be16toh (p->_vl_msg_id));
+  ck_assert_int_eq (first_sw_if_index, be32toh (p->first_sw_if_index));
+  ck_assert_int_eq (data_size, be32toh (p->count));
+  for (i = 0; i < data_size; ++i)
+    {
+      ck_assert_int_eq (packets, be64toh (p->data[i].packets));
+      ck_assert_int_eq (bytes, be64toh (p->data[i].bytes));
+    }
+  free (p);
+}
+
+END_TEST;
+
+#define verify_hton_swap(expr, value)           \
+  if (4 == sizeof (expr))                       \
+    {                                           \
+      ck_assert_int_eq (expr, htobe32 (value)); \
+    }                                           \
+  else if (2 == sizeof (expr))                  \
+    {                                           \
+      ck_assert_int_eq (expr, htobe16 (value)); \
+    }                                           \
+  else                                          \
+    {                                           \
+      ck_assert_int_eq (expr, value);           \
+    }
+
+START_TEST (test_hton_4)
+{
+  const int vla_count = 3;
+  char x[sizeof (vapi_msg_bridge_domain_details) +
+	 vla_count * sizeof (vapi_type_bridge_domain_sw_if)];
+  vapi_msg_bridge_domain_details *d = (void *) x;
+  int cnt = 1;
+  d->header._vl_msg_id = cnt++;
+  d->header.context = cnt++;
+  d->payload.bd_id = cnt++;
+  d->payload.flood = cnt++;
+  d->payload.uu_flood = cnt++;
+  d->payload.forward = cnt++;
+  d->payload.learn = cnt++;
+  d->payload.arp_term = cnt++;
+  d->payload.mac_age = cnt++;
+  d->payload.bvi_sw_if_index = cnt++;
+  d->payload.n_sw_ifs = vla_count;
+  int i;
+  for (i = 0; i < vla_count; ++i)
+    {
+      vapi_type_bridge_domain_sw_if *det = &d->payload.sw_if_details[i];
+      det->context = cnt++;
+      det->sw_if_index = cnt++;
+      det->shg = cnt++;
+    }
+  ck_assert_int_eq (sizeof (x), vapi_calc_bridge_domain_details_msg_size (d));
+  vapi_msg_bridge_domain_details_hton (d);
+  int tmp = 1;
+  verify_hton_swap (d->header._vl_msg_id, tmp);
+  ++tmp;
+  ck_assert_int_eq (d->header.context, tmp);
+  ++tmp;
+  verify_hton_swap (d->payload.bd_id, tmp);
+  ++tmp;
+  verify_hton_swap (d->payload.flood, tmp);
+  ++tmp;
+  verify_hton_swap (d->payload.uu_flood, tmp);
+  ++tmp;
+  verify_hton_swap (d->payload.forward, tmp);
+  ++tmp;
+  verify_hton_swap (d->payload.learn, tmp);
+  ++tmp;
+  verify_hton_swap (d->payload.arp_term, tmp);
+  ++tmp;
+  verify_hton_swap (d->payload.mac_age, tmp);
+  ++tmp;
+  verify_hton_swap (d->payload.bvi_sw_if_index, tmp);
+  ++tmp;
+  ck_assert_int_eq (d->payload.n_sw_ifs, htobe32 (vla_count));
+  for (i = 0; i < vla_count; ++i)
+    {
+      vapi_type_bridge_domain_sw_if *det = &d->payload.sw_if_details[i];
+      verify_hton_swap (det->context, tmp);
+      ++tmp;
+      verify_hton_swap (det->sw_if_index, tmp);
+      ++tmp;
+      verify_hton_swap (det->shg, tmp);
+      ++tmp;
+    }
+  vapi_msg_bridge_domain_details_ntoh (d);
+  tmp = 1;
+  ck_assert_int_eq (d->header._vl_msg_id, tmp);
+  ++tmp;
+  ck_assert_int_eq (d->header.context, tmp);
+  ++tmp;
+  ck_assert_int_eq (d->payload.bd_id, tmp);
+  ++tmp;
+  ck_assert_int_eq (d->payload.flood, tmp);
+  ++tmp;
+  ck_assert_int_eq (d->payload.uu_flood, tmp);
+  ++tmp;
+  ck_assert_int_eq (d->payload.forward, tmp);
+  ++tmp;
+  ck_assert_int_eq (d->payload.learn, tmp);
+  ++tmp;
+  ck_assert_int_eq (d->payload.arp_term, tmp);
+  ++tmp;
+  ck_assert_int_eq (d->payload.mac_age, tmp);
+  ++tmp;
+  ck_assert_int_eq (d->payload.bvi_sw_if_index, tmp);
+  ++tmp;
+  ck_assert_int_eq (d->payload.n_sw_ifs, vla_count);
+  for (i = 0; i < vla_count; ++i)
+    {
+      vapi_type_bridge_domain_sw_if *det = &d->payload.sw_if_details[i];
+      ck_assert_int_eq (det->context, tmp);
+      ++tmp;
+      ck_assert_int_eq (det->sw_if_index, tmp);
+      ++tmp;
+      ck_assert_int_eq (det->shg, tmp);
+      ++tmp;
+    }
+  ck_assert_int_eq (sizeof (x), vapi_calc_bridge_domain_details_msg_size (d));
+}
+
+END_TEST;
+
+START_TEST (test_ntoh_1)
+{
+  const u16 _vl_msg_id = 1;
+  vapi_type_msg_header1_t h;
+  h._vl_msg_id = _vl_msg_id;
+  vapi_type_msg_header1_t_ntoh (&h);
+  ck_assert_int_eq (htobe16 (h._vl_msg_id), _vl_msg_id);
+}
+
+END_TEST;
+
+START_TEST (test_ntoh_2)
+{
+  const u16 _vl_msg_id = 1;
+  const u32 client_index = 3;
+  vapi_type_msg_header2_t h;
+  h._vl_msg_id = _vl_msg_id;
+  h.client_index = client_index;
+  vapi_type_msg_header2_t_ntoh (&h);
+  ck_assert_int_eq (htobe16 (h._vl_msg_id), _vl_msg_id);
+  ck_assert_int_eq (h.client_index, client_index);
+}
+
+END_TEST;
+
+START_TEST (test_ntoh_3)
+{
+  const size_t data_size = 10;
+  vapi_msg_vnet_interface_combined_counters *m =
+    malloc (sizeof (vapi_msg_vnet_interface_combined_counters) +
+	    data_size * sizeof (vapi_type_vlib_counter));
+  ck_assert_ptr_ne (NULL, m);
+  vapi_payload_vnet_interface_combined_counters *p = &m->payload;
+  const u16 _vl_msg_id = 1;
+  p->_vl_msg_id = _vl_msg_id;
+  const u32 first_sw_if_index = 2;
+  p->first_sw_if_index = first_sw_if_index;
+  const size_t be_data_size = htobe32 (data_size);
+  p->count = be_data_size;
+  const u64 packets = 1234;
+  const u64 bytes = 2345;
+  int i;
+  for (i = 0; i < data_size; ++i)
+    {
+      p->data[i].packets = packets;
+      p->data[i].bytes = bytes;
+    }
+  vapi_msg_vnet_interface_combined_counters_ntoh (m);
+  ck_assert_int_eq (_vl_msg_id, be16toh (p->_vl_msg_id));
+  ck_assert_int_eq (first_sw_if_index, be32toh (p->first_sw_if_index));
+  ck_assert_int_eq (be_data_size, be32toh (p->count));
+  for (i = 0; i < data_size; ++i)
+    {
+      ck_assert_int_eq (packets, htobe64 (p->data[i].packets));
+      ck_assert_int_eq (bytes, htobe64 (p->data[i].bytes));
+    }
+  free (p);
+}
+
+END_TEST;
+
+#define verify_ntoh_swap(expr, value)           \
+  if (4 == sizeof (expr))                       \
+    {                                           \
+      ck_assert_int_eq (expr, be32toh (value)); \
+    }                                           \
+  else if (2 == sizeof (expr))                  \
+    {                                           \
+      ck_assert_int_eq (expr, be16toh (value)); \
+    }                                           \
+  else                                          \
+    {                                           \
+      ck_assert_int_eq (expr, value);           \
+    }
+
+START_TEST (test_ntoh_4)
+{
+  const int vla_count = 3;
+  char x[sizeof (vapi_msg_bridge_domain_details) +
+	 vla_count * sizeof (vapi_type_bridge_domain_sw_if)];
+  vapi_msg_bridge_domain_details *d = (void *) x;
+  int cnt = 1;
+  d->header._vl_msg_id = cnt++;
+  d->header.context = cnt++;
+  d->payload.bd_id = cnt++;
+  d->payload.flood = cnt++;
+  d->payload.uu_flood = cnt++;
+  d->payload.forward = cnt++;
+  d->payload.learn = cnt++;
+  d->payload.arp_term = cnt++;
+  d->payload.mac_age = cnt++;
+  d->payload.bvi_sw_if_index = cnt++;
+  d->payload.n_sw_ifs = htobe32 (vla_count);
+  int i;
+  for (i = 0; i < vla_count; ++i)
+    {
+      vapi_type_bridge_domain_sw_if *det = &d->payload.sw_if_details[i];
+      det->context = cnt++;
+      det->sw_if_index = cnt++;
+      det->shg = cnt++;
+    }
+  vapi_msg_bridge_domain_details_ntoh (d);
+  ck_assert_int_eq (sizeof (x), vapi_calc_bridge_domain_details_msg_size (d));
+  int tmp = 1;
+  verify_ntoh_swap (d->header._vl_msg_id, tmp);
+  ++tmp;
+  ck_assert_int_eq (d->header.context, tmp);
+  ++tmp;
+  verify_ntoh_swap (d->payload.bd_id, tmp);
+  ++tmp;
+  verify_ntoh_swap (d->payload.flood, tmp);
+  ++tmp;
+  verify_ntoh_swap (d->payload.uu_flood, tmp);
+  ++tmp;
+  verify_ntoh_swap (d->payload.forward, tmp);
+  ++tmp;
+  verify_ntoh_swap (d->payload.learn, tmp);
+  ++tmp;
+  verify_ntoh_swap (d->payload.arp_term, tmp);
+  ++tmp;
+  verify_ntoh_swap (d->payload.mac_age, tmp);
+  ++tmp;
+  verify_ntoh_swap (d->payload.bvi_sw_if_index, tmp);
+  ++tmp;
+  ck_assert_int_eq (d->payload.n_sw_ifs, vla_count);
+  for (i = 0; i < vla_count; ++i)
+    {
+      vapi_type_bridge_domain_sw_if *det = &d->payload.sw_if_details[i];
+      verify_ntoh_swap (det->context, tmp);
+      ++tmp;
+      verify_ntoh_swap (det->sw_if_index, tmp);
+      ++tmp;
+      verify_ntoh_swap (det->shg, tmp);
+      ++tmp;
+    }
+  vapi_msg_bridge_domain_details_hton (d);
+  tmp = 1;
+  ck_assert_int_eq (d->header._vl_msg_id, tmp);
+  ++tmp;
+  ck_assert_int_eq (d->header.context, tmp);
+  ++tmp;
+  ck_assert_int_eq (d->payload.bd_id, tmp);
+  ++tmp;
+  ck_assert_int_eq (d->payload.flood, tmp);
+  ++tmp;
+  ck_assert_int_eq (d->payload.uu_flood, tmp);
+  ++tmp;
+  ck_assert_int_eq (d->payload.forward, tmp);
+  ++tmp;
+  ck_assert_int_eq (d->payload.learn, tmp);
+  ++tmp;
+  ck_assert_int_eq (d->payload.arp_term, tmp);
+  ++tmp;
+  ck_assert_int_eq (d->payload.mac_age, tmp);
+  ++tmp;
+  ck_assert_int_eq (d->payload.bvi_sw_if_index, tmp);
+  ++tmp;
+  ck_assert_int_eq (d->payload.n_sw_ifs, htobe32 (vla_count));
+  for (i = 0; i < vla_count; ++i)
+    {
+      vapi_type_bridge_domain_sw_if *det = &d->payload.sw_if_details[i];
+      ck_assert_int_eq (det->context, tmp);
+      ++tmp;
+      ck_assert_int_eq (det->sw_if_index, tmp);
+      ++tmp;
+      ck_assert_int_eq (det->shg, tmp);
+      ++tmp;
+    }
+}
+
+END_TEST;
+
+vapi_error_e
+show_version_cb (vapi_ctx_t ctx, void *caller_ctx,
+		 vapi_error_e rv, bool is_last,
+		 vapi_payload_show_version_reply * p)
+{
+  ck_assert_int_eq (VAPI_OK, rv);
+  ck_assert_int_eq (true, is_last);
+  ck_assert_str_eq ("vpe", (char *) p->program);
+  printf
+    ("show_version_reply: program: `%s', version: `%s', build directory: "
+     "`%s', build date: `%s'\n", p->program, p->version, p->build_directory,
+     p->build_date);
+  ++*(int *) caller_ctx;
+  return VAPI_OK;
+}
+
+typedef struct
+{
+  int called;
+  int expected_retval;
+  u32 *sw_if_index_storage;
+} test_create_loopback_ctx_t;
+
+vapi_error_e
+loopback_create_cb (vapi_ctx_t ctx, void *caller_ctx,
+		    vapi_error_e rv, bool is_last,
+		    vapi_payload_create_loopback_reply * p)
+{
+  test_create_loopback_ctx_t *clc = caller_ctx;
+  ck_assert_int_eq (clc->expected_retval, p->retval);
+  *clc->sw_if_index_storage = p->sw_if_index;
+  ++clc->called;
+  return VAPI_OK;
+}
+
+typedef struct
+{
+  int called;
+  int expected_retval;
+  u32 *sw_if_index_storage;
+} test_delete_loopback_ctx_t;
+
+vapi_error_e
+loopback_delete_cb (vapi_ctx_t ctx, void *caller_ctx,
+		    vapi_error_e rv, bool is_last,
+		    vapi_payload_delete_loopback_reply * p)
+{
+  test_delete_loopback_ctx_t *dlc = caller_ctx;
+  ck_assert_int_eq (dlc->expected_retval, p->retval);
+  ++dlc->called;
+  return VAPI_OK;
+}
+
+START_TEST (test_connect)
+{
+  vapi_ctx_t ctx;
+  vapi_error_e rv = vapi_ctx_alloc (&ctx);
+  ck_assert_int_eq (VAPI_OK, rv);
+  rv = vapi_connect (ctx, app_name, api_prefix, max_outstanding_requests,
+		     response_queue_size, VAPI_MODE_BLOCKING);
+  ck_assert_int_eq (VAPI_OK, rv);
+  rv = vapi_disconnect (ctx);
+  ck_assert_int_eq (VAPI_OK, rv);
+  vapi_ctx_free (ctx);
+}
+
+END_TEST;
+
+vapi_ctx_t ctx;
+
+void
+setup_blocking (void)
+{
+  vapi_error_e rv = vapi_ctx_alloc (&ctx);
+  ck_assert_int_eq (VAPI_OK, rv);
+  rv = vapi_connect (ctx, app_name, api_prefix, max_outstanding_requests,
+		     response_queue_size, VAPI_MODE_BLOCKING);
+  ck_assert_int_eq (VAPI_OK, rv);
+}
+
+void
+setup_nonblocking (void)
+{
+  vapi_error_e rv = vapi_ctx_alloc (&ctx);
+  ck_assert_int_eq (VAPI_OK, rv);
+  rv = vapi_connect (ctx, app_name, api_prefix, max_outstanding_requests,
+		     response_queue_size, VAPI_MODE_NONBLOCKING);
+  ck_assert_int_eq (VAPI_OK, rv);
+}
+
+void
+teardown (void)
+{
+  vapi_disconnect (ctx);
+  vapi_ctx_free (ctx);
+}
+
+START_TEST (test_show_version_1)
+{
+  printf ("--- Basic show version message - reply test ---\n");
+  vapi_msg_show_version *sv = vapi_alloc_show_version (ctx);
+  ck_assert_ptr_ne (NULL, sv);
+  vapi_msg_show_version_hton (sv);
+  vapi_error_e rv = vapi_send (ctx, sv);
+  ck_assert_int_eq (VAPI_OK, rv);
+  vapi_msg_show_version_reply *resp;
+  size_t size;
+  rv = vapi_recv (ctx, (void *) &resp, &size);
+  ck_assert_int_eq (VAPI_OK, rv);
+  vapi_payload_show_version_reply *payload = &resp->payload;
+  int dummy;
+  show_version_cb (NULL, &dummy, VAPI_OK, true, payload);
+  vapi_msg_free (ctx, resp);
+}
+
+END_TEST;
+
+START_TEST (test_show_version_2)
+{
+  int called = 0;
+  printf ("--- Show version via blocking callback API ---\n");
+  const int attempts = response_queue_size * 4;
+  int i = 0;
+  for (i = 0; i < attempts; ++i)
+    {
+      vapi_msg_show_version *sv = vapi_alloc_show_version (ctx);
+      ck_assert_ptr_ne (NULL, sv);
+      vapi_error_e rv = vapi_show_version (ctx, sv, show_version_cb, &called);
+      ck_assert_int_eq (VAPI_OK, rv);
+    }
+  ck_assert_int_eq (attempts, called);
+}
+
+END_TEST;
+
+typedef struct
+{
+  bool last_called;
+  size_t num_ifs;
+  u32 *sw_if_indexes;
+  bool *seen;
+  int called;
+} sw_interface_dump_ctx;
+
+vapi_error_e
+sw_interface_dump_cb (struct vapi_ctx_s *ctx, void *callback_ctx,
+		      vapi_error_e rv, bool is_last,
+		      vapi_payload_sw_interface_details * reply)
+{
+  sw_interface_dump_ctx *dctx = callback_ctx;
+  ck_assert_int_eq (false, dctx->last_called);
+  if (is_last)
+    {
+      ck_assert (NULL == reply);
+      dctx->last_called = true;
+    }
+  else
+    {
+      ck_assert (reply);
+      printf ("Interface dump entry: [%u]: %s\n", reply->sw_if_index,
+	      reply->interface_name);
+      size_t i = 0;
+      for (i = 0; i < dctx->num_ifs; ++i)
+	{
+	  if (dctx->sw_if_indexes[i] == reply->sw_if_index)
+	    {
+	      ck_assert_int_eq (false, dctx->seen[i]);
+	      dctx->seen[i] = true;
+	    }
+	}
+    }
+  ++dctx->called;
+  return VAPI_OK;
+}
+
+START_TEST (test_loopbacks_1)
+{
+  printf ("--- Create/delete loopbacks using blocking API ---\n");
+  const size_t num_ifs = 5;
+  u8 mac_addresses[num_ifs][6];
+  memset (&mac_addresses, 0, sizeof (mac_addresses));
+  u32 sw_if_indexes[num_ifs];
+  memset (&sw_if_indexes, 0xff, sizeof (sw_if_indexes));
+  test_create_loopback_ctx_t clcs[num_ifs];
+  memset (&clcs, 0, sizeof (clcs));
+  test_delete_loopback_ctx_t dlcs[num_ifs];
+  memset (&dlcs, 0, sizeof (dlcs));
+  int i;
+  for (i = 0; i < num_ifs; ++i)
+    {
+      memcpy (&mac_addresses[i], "\1\2\3\4\5\6", 6);
+      mac_addresses[i][5] = i;
+      clcs[i].sw_if_index_storage = &sw_if_indexes[i];
+    }
+  for (i = 0; i < num_ifs; ++i)
+    {
+      vapi_msg_create_loopback *cl = vapi_alloc_create_loopback (ctx);
+      memcpy (cl->payload.mac_address, mac_addresses[i],
+	      sizeof (cl->payload.mac_address));
+      vapi_error_e rv =
+	vapi_create_loopback (ctx, cl, loopback_create_cb, &clcs[i]);
+      ck_assert_int_eq (VAPI_OK, rv);
+    }
+  for (i = 0; i < num_ifs; ++i)
+    {
+      ck_assert_int_eq (1, clcs[i].called);
+      printf ("Created loopback with MAC %02x:%02x:%02x:%02x:%02x:%02x --> "
+	      "sw_if_index %u\n",
+	      mac_addresses[i][0], mac_addresses[i][1], mac_addresses[i][2],
+	      mac_addresses[i][3], mac_addresses[i][4], mac_addresses[i][5],
+	      sw_if_indexes[i]);
+    }
+  bool seen[num_ifs];
+  sw_interface_dump_ctx dctx = { false, num_ifs, sw_if_indexes, seen, 0 };
+  vapi_msg_sw_interface_dump *dump;
+  vapi_error_e rv;
+  const int attempts = response_queue_size * 4;
+  for (i = 0; i < attempts; ++i)
+    {
+      dctx.last_called = false;
+      memset (&seen, 0, sizeof (seen));
+      dump = vapi_alloc_sw_interface_dump (ctx);
+      dump->payload.name_filter_valid = 0;
+      memset (dump->payload.name_filter, 0,
+	      sizeof (dump->payload.name_filter));
+      while (VAPI_EAGAIN ==
+	     (rv =
+	      vapi_sw_interface_dump (ctx, dump, sw_interface_dump_cb,
+				      &dctx)))
+	;
+      ck_assert_int_eq (true, dctx.last_called);
+      int j = 0;
+      for (j = 0; j < num_ifs; ++j)
+	{
+	  ck_assert_int_eq (true, seen[j]);
+	}
+    }
+  memset (&seen, 0, sizeof (seen));
+  for (i = 0; i < num_ifs; ++i)
+    {
+      vapi_msg_delete_loopback *dl = vapi_alloc_delete_loopback (ctx);
+      dl->payload.sw_if_index = sw_if_indexes[i];
+      vapi_error_e rv =
+	vapi_delete_loopback (ctx, dl, loopback_delete_cb, &dlcs[i]);
+      ck_assert_int_eq (VAPI_OK, rv);
+    }
+  for (i = 0; i < num_ifs; ++i)
+    {
+      ck_assert_int_eq (1, dlcs[i].called);
+      printf ("Deleted loopback with sw_if_index %u\n", sw_if_indexes[i]);
+    }
+  dctx.last_called = false;
+  memset (&seen, 0, sizeof (seen));
+  dump = vapi_alloc_sw_interface_dump (ctx);
+  dump->payload.name_filter_valid = 0;
+  memset (dump->payload.name_filter, 0, sizeof (dump->payload.name_filter));
+  while (VAPI_EAGAIN ==
+	 (rv =
+	  vapi_sw_interface_dump (ctx, dump, sw_interface_dump_cb, &dctx)))
+    ;
+  ck_assert_int_eq (true, dctx.last_called);
+  for (i = 0; i < num_ifs; ++i)
+    {
+      ck_assert_int_eq (false, seen[i]);
+    }
+}
+
+END_TEST;
+
+START_TEST (test_show_version_3)
+{
+  printf ("--- Show version via async callback ---\n");
+  int called = 0;
+  vapi_error_e rv;
+  vapi_msg_show_version *sv = vapi_alloc_show_version (ctx);
+  ck_assert_ptr_ne (NULL, sv);
+  while (VAPI_EAGAIN ==
+	 (rv = vapi_show_version (ctx, sv, show_version_cb, &called)))
+    ;
+  ck_assert_int_eq (VAPI_OK, rv);
+  ck_assert_int_eq (0, called);
+  rv = vapi_dispatch (ctx);
+  ck_assert_int_eq (VAPI_OK, rv);
+  ck_assert_int_eq (1, called);
+  called = 0;
+  rv = vapi_dispatch (ctx);
+  ck_assert_int_eq (VAPI_OK, rv);
+  ck_assert_int_eq (0, called);
+}
+
+END_TEST;
+
+START_TEST (test_show_version_4)
+{
+  printf ("--- Show version via async callback - multiple messages ---\n");
+  vapi_error_e rv;
+  const size_t num_req = 5;
+  int contexts[num_req];
+  memset (contexts, 0, sizeof (contexts));
+  int i;
+  for (i = 0; i < num_req; ++i)
+    {
+      vapi_msg_show_version *sv = vapi_alloc_show_version (ctx);
+      ck_assert_ptr_ne (NULL, sv);
+      while (VAPI_EAGAIN ==
+	     (rv =
+	      vapi_show_version (ctx, sv, show_version_cb, &contexts[i])))
+	;
+      ck_assert_int_eq (VAPI_OK, rv);
+      int j;
+      for (j = 0; j < num_req; ++j)
+	{
+	  ck_assert_int_eq (0, contexts[j]);
+	}
+    }
+  rv = vapi_dispatch (ctx);
+  ck_assert_int_eq (VAPI_OK, rv);
+  for (i = 0; i < num_req; ++i)
+    {
+      ck_assert_int_eq (1, contexts[i]);
+    }
+  memset (contexts, 0, sizeof (contexts));
+  rv = vapi_dispatch (ctx);
+  ck_assert_int_eq (VAPI_OK, rv);
+  for (i = 0; i < num_req; ++i)
+    {
+      ck_assert_int_eq (0, contexts[i]);
+    }
+}
+
+END_TEST;
+
+START_TEST (test_loopbacks_2)
+{
+  printf ("--- Create/delete loopbacks using non-blocking API ---\n");
+  vapi_error_e rv;
+  const size_t num_ifs = 5;
+  u8 mac_addresses[num_ifs][6];
+  memset (&mac_addresses, 0, sizeof (mac_addresses));
+  u32 sw_if_indexes[num_ifs];
+  memset (&sw_if_indexes, 0xff, sizeof (sw_if_indexes));
+  test_create_loopback_ctx_t clcs[num_ifs];
+  memset (&clcs, 0, sizeof (clcs));
+  test_delete_loopback_ctx_t dlcs[num_ifs];
+  memset (&dlcs, 0, sizeof (dlcs));
+  int i;
+  for (i = 0; i < num_ifs; ++i)
+    {
+      memcpy (&mac_addresses[i], "\1\2\3\4\5\6", 6);
+      mac_addresses[i][5] = i;
+      clcs[i].sw_if_index_storage = &sw_if_indexes[i];
+    }
+  for (i = 0; i < num_ifs; ++i)
+    {
+      vapi_msg_create_loopback *cl = vapi_alloc_create_loopback (ctx);
+      memcpy (cl->payload.mac_address, mac_addresses[i],
+	      sizeof (cl->payload.mac_address));
+      while (VAPI_EAGAIN ==
+	     (rv =
+	      vapi_create_loopback (ctx, cl, loopback_create_cb, &clcs[i])))
+	;
+      ck_assert_int_eq (VAPI_OK, rv);
+    }
+  rv = vapi_dispatch (ctx);
+  ck_assert_int_eq (VAPI_OK, rv);
+  for (i = 0; i < num_ifs; ++i)
+    {
+      ck_assert_int_eq (1, clcs[i].called);
+      printf ("Loopback with MAC %02x:%02x:%02x:%02x:%02x:%02x --> "
+	      "sw_if_index %u\n",
+	      mac_addresses[i][0], mac_addresses[i][1], mac_addresses[i][2],
+	      mac_addresses[i][3], mac_addresses[i][4], mac_addresses[i][5],
+	      sw_if_indexes[i]);
+    }
+  bool seen[num_ifs];
+  memset (&seen, 0, sizeof (seen));
+  sw_interface_dump_ctx dctx = { false, num_ifs, sw_if_indexes, seen, 0 };
+  vapi_msg_sw_interface_dump *dump = vapi_alloc_sw_interface_dump (ctx);
+  dump->payload.name_filter_valid = 0;
+  memset (dump->payload.name_filter, 0, sizeof (dump->payload.name_filter));
+  while (VAPI_EAGAIN ==
+	 (rv =
+	  vapi_sw_interface_dump (ctx, dump, sw_interface_dump_cb, &dctx)))
+    ;
+  for (i = 0; i < num_ifs; ++i)
+    {
+      ck_assert_int_eq (false, seen[i]);
+    }
+  memset (&seen, 0, sizeof (seen));
+  ck_assert_int_eq (false, dctx.last_called);
+  rv = vapi_dispatch (ctx);
+  ck_assert_int_eq (VAPI_OK, rv);
+  for (i = 0; i < num_ifs; ++i)
+    {
+      ck_assert_int_eq (true, seen[i]);
+    }
+  memset (&seen, 0, sizeof (seen));
+  ck_assert_int_eq (true, dctx.last_called);
+  for (i = 0; i < num_ifs; ++i)
+    {
+      vapi_msg_delete_loopback *dl = vapi_alloc_delete_loopback (ctx);
+      dl->payload.sw_if_index = sw_if_indexes[i];
+      while (VAPI_EAGAIN ==
+	     (rv =
+	      vapi_delete_loopback (ctx, dl, loopback_delete_cb, &dlcs[i])))
+	;
+      ck_assert_int_eq (VAPI_OK, rv);
+    }
+  rv = vapi_dispatch (ctx);
+  ck_assert_int_eq (VAPI_OK, rv);
+  for (i = 0; i < num_ifs; ++i)
+    {
+      ck_assert_int_eq (1, dlcs[i].called);
+      printf ("Deleted loopback with sw_if_index %u\n", sw_if_indexes[i]);
+    }
+  memset (&seen, 0, sizeof (seen));
+  dctx.last_called = false;
+  dump = vapi_alloc_sw_interface_dump (ctx);
+  dump->payload.name_filter_valid = 0;
+  memset (dump->payload.name_filter, 0, sizeof (dump->payload.name_filter));
+  while (VAPI_EAGAIN ==
+	 (rv =
+	  vapi_sw_interface_dump (ctx, dump, sw_interface_dump_cb, &dctx)))
+    ;
+  rv = vapi_dispatch (ctx);
+  ck_assert_int_eq (VAPI_OK, rv);
+  for (i = 0; i < num_ifs; ++i)
+    {
+      ck_assert_int_eq (false, seen[i]);
+    }
+  memset (&seen, 0, sizeof (seen));
+  ck_assert_int_eq (true, dctx.last_called);
+}
+
+END_TEST;
+
+vapi_error_e
+interface_simple_stats_cb (vapi_ctx_t ctx, void *callback_ctx,
+			   vapi_error_e rv, bool is_last,
+			   vapi_payload_want_interface_simple_stats_reply *
+			   payload)
+{
+  return VAPI_OK;
+}
+
+vapi_error_e
+simple_counters_cb (vapi_ctx_t ctx, void *callback_ctx,
+		    vapi_payload_vnet_interface_simple_counters * payload)
+{
+  int *called = callback_ctx;
+  ++*called;
+  printf ("simple counters: first_sw_if_index=%u\n",
+	  payload->first_sw_if_index);
+  return VAPI_OK;
+}
+
+START_TEST (test_stats_1)
+{
+  printf ("--- Receive stats using generic blocking API ---\n");
+  vapi_msg_want_interface_simple_stats *ws =
+    vapi_alloc_want_interface_simple_stats (ctx);
+  ws->payload.enable_disable = 1;
+  ws->payload.pid = getpid ();
+  vapi_error_e rv;
+  rv = vapi_want_interface_simple_stats (ctx, ws, interface_simple_stats_cb,
+					 NULL);
+  ck_assert_int_eq (VAPI_OK, rv);
+  int called = 0;
+  vapi_set_event_cb (ctx, vapi_msg_id_vnet_interface_simple_counters,
+		     (vapi_event_cb) simple_counters_cb, &called);
+  rv = vapi_dispatch_one (ctx);
+  ck_assert_int_eq (VAPI_OK, rv);
+  ck_assert_int_eq (1, called);
+}
+
+END_TEST;
+
+START_TEST (test_stats_2)
+{
+  printf ("--- Receive stats using stat-specific blocking API ---\n");
+  vapi_msg_want_interface_simple_stats *ws =
+    vapi_alloc_want_interface_simple_stats (ctx);
+  ws->payload.enable_disable = 1;
+  ws->payload.pid = getpid ();
+  vapi_error_e rv;
+  rv = vapi_want_interface_simple_stats (ctx, ws, interface_simple_stats_cb,
+					 NULL);
+  ck_assert_int_eq (VAPI_OK, rv);
+  int called = 0;
+  vapi_set_vapi_msg_vnet_interface_simple_counters_event_cb (ctx,
+							     simple_counters_cb,
+							     &called);
+  rv = vapi_dispatch_one (ctx);
+  ck_assert_int_eq (VAPI_OK, rv);
+  ck_assert_int_eq (1, called);
+}
+
+END_TEST;
+
+vapi_error_e
+generic_cb (vapi_ctx_t ctx, void *callback_ctx, vapi_msg_id_t id, void *msg)
+{
+  int *called = callback_ctx;
+  ck_assert_int_eq (0, *called);
+  ++*called;
+  ck_assert_int_eq (id, vapi_msg_id_show_version_reply);
+  ck_assert_ptr_ne (NULL, msg);
+  vapi_msg_show_version_reply *reply = msg;
+  ck_assert_str_eq ("vpe", (char *) reply->payload.program);
+  return VAPI_OK;
+}
+
+START_TEST (test_show_version_5)
+{
+  printf ("--- Receive show version using generic callback - nonblocking "
+	  "API ---\n");
+  vapi_error_e rv;
+  vapi_msg_show_version *sv = vapi_alloc_show_version (ctx);
+  ck_assert_ptr_ne (NULL, sv);
+  vapi_msg_show_version_hton (sv);
+  while (VAPI_EAGAIN == (rv = vapi_send (ctx, sv)))
+    ;
+  ck_assert_int_eq (VAPI_OK, rv);
+  int called = 0;
+  vapi_set_generic_event_cb (ctx, generic_cb, &called);
+  ck_assert_int_eq (VAPI_OK, rv);
+  rv = vapi_dispatch_one (ctx);
+  ck_assert_int_eq (VAPI_OK, rv);
+  ck_assert_int_eq (1, called);
+  sv = vapi_alloc_show_version (ctx);
+  ck_assert_ptr_ne (NULL, sv);
+  vapi_msg_show_version_hton (sv);
+  while (VAPI_EAGAIN == (rv = vapi_send (ctx, sv)))
+    ;
+  ck_assert_int_eq (VAPI_OK, rv);
+  vapi_clear_generic_event_cb (ctx);
+  rv = vapi_dispatch_one (ctx);
+  ck_assert_int_eq (VAPI_OK, rv);
+  ck_assert_int_eq (1, called);	/* needs to remain unchanged */
+}
+
+END_TEST;
+
+vapi_error_e
+combined_counters_cb (struct vapi_ctx_s *ctx, void *callback_ctx,
+		      vapi_payload_vnet_interface_combined_counters * payload)
+{
+  int *called = callback_ctx;
+  ++*called;
+  printf ("combined counters: first_sw_if_index=%u\n",
+	  payload->first_sw_if_index);
+  return VAPI_OK;
+}
+
+vapi_error_e
+stats_cb (vapi_ctx_t ctx, void *callback_ctx, vapi_error_e rv,
+	  bool is_last, vapi_payload_want_stats_reply * payload)
+{
+  return VAPI_OK;
+}
+
+START_TEST (test_stats_3)
+{
+  printf ("--- Receive multiple stats using stat-specific non-blocking API "
+	  "---\n");
+  vapi_msg_want_stats *ws = vapi_alloc_want_stats (ctx);
+  ws->payload.enable_disable = 1;
+  ws->payload.pid = getpid ();
+  vapi_error_e rv;
+  rv = vapi_want_stats (ctx, ws, stats_cb, NULL);
+  ck_assert_int_eq (VAPI_OK, rv);
+  int called = 0;
+  int called2 = 0;
+  vapi_set_vapi_msg_vnet_interface_simple_counters_event_cb (ctx,
+							     simple_counters_cb,
+							     &called);
+  vapi_set_vapi_msg_vnet_interface_combined_counters_event_cb (ctx,
+							       combined_counters_cb,
+							       &called2);
+  while (!called || !called2)
+    {
+      if (VAPI_EAGAIN != (rv = vapi_dispatch_one (ctx)))
+	{
+	  ck_assert_int_eq (VAPI_OK, rv);
+	}
+    }
+}
+
+END_TEST;
+
+vapi_error_e
+show_version_no_cb (vapi_ctx_t ctx, void *caller_ctx,
+		    vapi_error_e rv, bool is_last,
+		    vapi_payload_show_version_reply * p)
+{
+  ck_assert_int_eq (VAPI_ENORESP, rv);
+  ck_assert_int_eq (true, is_last);
+  ck_assert_ptr_eq (NULL, p);
+  ++*(int *) caller_ctx;
+  return VAPI_OK;
+}
+
+START_TEST (test_no_response_1)
+{
+  printf ("--- Simulate no response to regular message ---\n");
+  vapi_error_e rv;
+  vapi_msg_show_version *sv = vapi_alloc_show_version (ctx);
+  ck_assert_ptr_ne (NULL, sv);
+  sv->header._vl_msg_id = ~0;	/* malformed ID causes vpp to drop the msg */
+  int called = 0;
+  while (VAPI_EAGAIN ==
+	 (rv = vapi_show_version (ctx, sv, show_version_no_cb, &called)))
+    ;
+  ck_assert_int_eq (VAPI_OK, rv);
+  sv = vapi_alloc_show_version (ctx);
+  ck_assert_ptr_ne (NULL, sv);
+  while (VAPI_EAGAIN ==
+	 (rv = vapi_show_version (ctx, sv, show_version_cb, &called)))
+    ;
+  ck_assert_int_eq (VAPI_OK, rv);
+  rv = vapi_dispatch (ctx);
+  ck_assert_int_eq (VAPI_OK, rv);
+  ck_assert_int_eq (2, called);
+}
+
+END_TEST;
+
+vapi_error_e
+no_msg_cb (struct vapi_ctx_s *ctx, void *callback_ctx,
+	   vapi_error_e rv, bool is_last,
+	   vapi_payload_sw_interface_details * reply)
+{
+  int *called = callback_ctx;
+  ++*called;
+  ck_assert_int_eq (VAPI_OK, rv);
+  ck_assert_int_eq (true, is_last);
+  ck_assert_ptr_eq (NULL, reply);
+  return VAPI_OK;
+}
+
+START_TEST (test_no_response_2)
+{
+  printf ("--- Simulate no response to dump message ---\n");
+  vapi_error_e rv;
+  vapi_msg_sw_interface_dump *dump = vapi_alloc_sw_interface_dump (ctx);
+  dump->header._vl_msg_id = ~0;	/* malformed ID causes vpp to drop the msg */
+  int no_called = 0;
+  while (VAPI_EAGAIN ==
+	 (rv = vapi_sw_interface_dump (ctx, dump, no_msg_cb, &no_called)))
+    ;
+  ck_assert_int_eq (VAPI_OK, rv);
+  rv = vapi_dispatch (ctx);
+  ck_assert_int_eq (VAPI_OK, rv);
+  ck_assert_int_eq (1, no_called);
+}
+
+END_TEST;
+Suite *
+test_suite (void)
+{
+  Suite *s = suite_create ("VAPI test");
+
+  TCase *tc_negative = tcase_create ("Negative tests");
+  tcase_add_test (tc_negative, test_invalid_values);
+  suite_add_tcase (s, tc_negative);
+
+  TCase *tc_swap = tcase_create ("Byteswap tests");
+  tcase_add_test (tc_swap, test_hton_1);
+  tcase_add_test (tc_swap, test_hton_2);
+  tcase_add_test (tc_swap, test_hton_3);
+  tcase_add_test (tc_swap, test_hton_4);
+  tcase_add_test (tc_swap, test_ntoh_1);
+  tcase_add_test (tc_swap, test_ntoh_2);
+  tcase_add_test (tc_swap, test_ntoh_3);
+  tcase_add_test (tc_swap, test_ntoh_4);
+  suite_add_tcase (s, tc_swap);
+
+  TCase *tc_connect = tcase_create ("Connect");
+  tcase_add_test (tc_connect, test_connect);
+  suite_add_tcase (s, tc_connect);
+
+  TCase *tc_block = tcase_create ("Blocking API");
+  tcase_set_timeout (tc_block, 25);
+  tcase_add_checked_fixture (tc_block, setup_blocking, teardown);
+  tcase_add_test (tc_block, test_show_version_1);
+  tcase_add_test (tc_block, test_show_version_2);
+  tcase_add_test (tc_block, test_loopbacks_1);
+  tcase_add_test (tc_block, test_stats_1);
+  tcase_add_test (tc_block, test_stats_2);
+  suite_add_tcase (s, tc_block);
+
+  TCase *tc_nonblock = tcase_create ("Nonblocking API");
+  tcase_set_timeout (tc_nonblock, 25);
+  tcase_add_checked_fixture (tc_nonblock, setup_nonblocking, teardown);
+  tcase_add_test (tc_nonblock, test_show_version_3);
+  tcase_add_test (tc_nonblock, test_show_version_4);
+  tcase_add_test (tc_nonblock, test_show_version_5);
+  tcase_add_test (tc_nonblock, test_loopbacks_2);
+  tcase_add_test (tc_nonblock, test_stats_3);
+  tcase_add_test (tc_nonblock, test_no_response_1);
+  tcase_add_test (tc_nonblock, test_no_response_2);
+  suite_add_tcase (s, tc_nonblock);
+
+  return s;
+}
+
+int
+main (int argc, char *argv[])
+{
+  if (3 != argc)
+    {
+      printf ("Invalid argc==`%d'\n", argc);
+      return EXIT_FAILURE;
+    }
+  app_name = argv[1];
+  api_prefix = argv[2];
+  printf ("App name: `%s', API prefix: `%s'\n", app_name, api_prefix);
+
+  int number_failed;
+  Suite *s;
+  SRunner *sr;
+
+  s = test_suite ();
+  sr = srunner_create (s);
+
+  srunner_run_all (sr, CK_NORMAL);
+  number_failed = srunner_ntests_failed (sr);
+  srunner_free (sr);
+  return (number_failed == 0) ? EXIT_SUCCESS : EXIT_FAILURE;
+}
+
+/*
+ * fd.io coding-style-patch-verification: ON
+ *
+ * Local Variables:
+ * eval: (c-set-style "gnu")
+ * End:
+ */
diff --git a/test/scripts/test-loop.sh b/test/scripts/test-loop.sh
index 17dc7c39..51f5d5ce 100755
--- a/test/scripts/test-loop.sh
+++ b/test/scripts/test-loop.sh
@@ -3,14 +3,15 @@
 function usage() {
  echo "$0" 1>&2
  echo "" 1>&2
- echo "Usage: $0 [-p ] [-m ] -- " 1>&2
+ echo "Usage: $0 [-p ] [-m ] -- " 1>&2
  echo "" 1>&2
  echo "Parameters:" 1>&2
  echo "    -p  - run a command before each test loop (e.g. 'git pull')" 1>&2
  echo "    -m         - if set, email is sent to this address on failure" 1>&2
  echo "" 1>&2
- echo "Example:" 1>&2
- echo "    $0 -m  -- test-debug TEST=l2bd"
+ echo "Examples:" 1>&2
+ echo "    $0 -m  -- test-debug TEST=l2bd" 1>&2
+ echo "    $0 -m  -- verify" 1>&2
  exit 1;
 }
 
@@ -44,8 +45,11 @@ shift $((OPTIND-1))
 
 if ! echo $* | grep test >/dev/null
 then
-        echo "Error: command line doesn't look right - should contain \`test' token..." >&2
-	usage
+	if ! echo $* | grep verify >/dev/null
+	then
+		echo "Error: command line doesn't look right - should contain \`test' or \`verify' token..." >&2
+		usage
+	fi
 fi
 
 function finish {
diff --git a/test/test_vapi.py b/test/test_vapi.py
new file mode 100644
index 00000000..86c1ee06
--- /dev/null
+++ b/test/test_vapi.py
@@ -0,0 +1,78 @@
+#!/usr/bin/env python
+""" VAPI test """
+
+from __future__ import division
+import unittest
+import os
+import signal
+import subprocess
+from threading import Thread
+from log import single_line_delim
+from framework import VppTestCase, running_extended_tests, VppTestRunner
+
+
+class Worker(Thread):
+    def __init__(self, args, logger):
+        self.logger = logger
+        self.args = args
+        self.result = None
+        super(Worker, self).__init__()
+
+    def run(self):
+        executable = self.args[0]
+        self.logger.debug("Running executable w/args `%s'" % self.args)
+        env = os.environ.copy()
+        env["CK_LOG_FILE_NAME"] = "-"
+        self.process = subprocess.Popen(
+            self.args, shell=False, env=env, preexec_fn=os.setpgrp,
+            stdout=subprocess.PIPE, stderr=subprocess.PIPE)
+        out, err = self.process.communicate()
+        self.logger.debug("Finished running `%s'" % executable)
+        self.logger.info("Return code is `%s'" % self.process.returncode)
+        self.logger.info(single_line_delim)
+        self.logger.info("Executable `%s' wrote to stdout:" % executable)
+        self.logger.info(single_line_delim)
+        self.logger.info(out)
+        self.logger.info(single_line_delim)
+        self.logger.info("Executable `%s' wrote to stderr:" % executable)
+        self.logger.info(single_line_delim)
+        self.logger.error(err)
+        self.logger.info(single_line_delim)
+        self.result = self.process.returncode
+
+
+@unittest.skipUnless(running_extended_tests(), "part of extended tests")
+class VAPITestCase(VppTestCase):
+    """ VAPI test """
+
+    def test_vapi(self):
+        """ run VAPI tests """
+        var = "BR"
+        built_root = os.getenv(var, None)
+        self.assertIsNotNone(built_root,
+                             "Environment variable `%s' not set" % var)
+        executable = "%s/vapi_test/vapi_test" % built_root
+        worker = Worker(
+            [executable, "vapi client", self.shm_prefix], self.logger)
+        worker.start()
+        timeout = 45
+        worker.join(timeout)
+        self.logger.info("Worker result is `%s'" % worker.result)
+        error = False
+        if worker.result is None:
+            try:
+                error = True
+                self.logger.error(
+                    "Timeout! Worker did not finish in %ss" % timeout)
+                os.killpg(os.getpgid(worker.process.pid), signal.SIGTERM)
+                worker.join()
+            except:
+                raise Exception("Couldn't kill worker-spawned process")
+        if error:
+            raise Exception(
+                "Timeout! Worker did not finish in %ss" % timeout)
+        self.assert_equal(worker.result, 0, "Binary test return code")
+
+
+if __name__ == '__main__':
+    unittest.main(testRunner=VppTestRunner)
-- 
cgit 1.2.3-korg