aboutsummaryrefslogtreecommitdiffstats
diff options
context:
space:
mode:
authorKlement Sekera <ksekera@cisco.com>2017-09-20 08:26:30 +0200
committerFlorin Coras <florin.coras@gmail.com>2018-02-01 23:41:17 +0000
commit75e7d1301475d49311d14e202936c62df0c07d10 (patch)
tree1a7e1d211cc97352f093a0207e6c94952e39e611
parent2b9453c55a7294d8eedf126d616a608c32be0e36 (diff)
IPv4/6 reassembly
Change-Id: Ic5dcadd13c88b8a5e7896dab82404509c081614a Signed-off-by: Klement Sekera <ksekera@cisco.com>
-rw-r--r--src/vlib/buffer_funcs.h62
-rw-r--r--src/vnet.am2
-rw-r--r--src/vnet/buffer.h12
-rw-r--r--src/vnet/ip/icmp46_packet.h1
-rw-r--r--src/vnet/ip/icmp6.h2
-rw-r--r--src/vnet/ip/ip.api28
-rw-r--r--src/vnet/ip/ip4_error.h9
-rwxr-xr-xsrc/vnet/ip/ip4_forward.c13
-rw-r--r--src/vnet/ip/ip4_input.c2
-rw-r--r--src/vnet/ip/ip4_input.h1
-rw-r--r--src/vnet/ip/ip4_reassembly.c1277
-rw-r--r--src/vnet/ip/ip4_reassembly.h49
-rw-r--r--src/vnet/ip/ip6_error.h9
-rw-r--r--src/vnet/ip/ip6_forward.c2
-rw-r--r--src/vnet/ip/ip6_input.h1
-rw-r--r--src/vnet/ip/ip6_packet.h3
-rw-r--r--src/vnet/ip/ip6_reassembly.c1347
-rw-r--r--src/vnet/ip/ip6_reassembly.h49
-rw-r--r--src/vnet/ip/ip_api.c61
-rw-r--r--src/vnet/ip/lookup.c2
-rw-r--r--src/vnet/ip/lookup.h1
-rw-r--r--src/vnet/ip/protocols.def5
-rw-r--r--test/Makefile3
-rw-r--r--test/framework.py22
-rw-r--r--test/hook.py14
-rw-r--r--test/patches/scapy-2.3.3/inet6.py.patch25
-rwxr-xr-xtest/sys_req/set_system_parameters.sh57
-rw-r--r--test/sys_req/system_parameters18
-rw-r--r--test/test_reassembly.py966
-rw-r--r--test/util.py195
-rw-r--r--test/vpp_gre_interface.py28
-rw-r--r--test/vpp_interface.py12
-rw-r--r--test/vpp_lo_interface.py12
-rw-r--r--test/vpp_papi_provider.py23
-rw-r--r--test/vpp_punt_socket.py75
35 files changed, 4322 insertions, 66 deletions
diff --git a/src/vlib/buffer_funcs.h b/src/vlib/buffer_funcs.h
index 4831eb5ebac..cc56db7762c 100644
--- a/src/vlib/buffer_funcs.h
+++ b/src/vlib/buffer_funcs.h
@@ -1004,6 +1004,68 @@ vlib_validate_buffer_set_in_use (vlib_buffer_t * b, u32 expected)
#endif
}
+/** minimum data size of first buffer in a buffer chain */
+#define VLIB_BUFFER_CHAIN_MIN_FIRST_DATA_SIZE (256)
+
+/**
+ * @brief compress buffer chain in a way where the first buffer is at least
+ * VLIB_BUFFER_CHAIN_MIN_FIRST_DATA_SIZE long
+ *
+ * @param[in] vm - vlib_main
+ * @param[in,out] first - first buffer in chain
+ * @param[in,out] discard_vector - vector of buffer indexes which were removed
+ * from the chain
+ */
+always_inline void
+vlib_buffer_chain_compress (vlib_main_t * vm,
+ vlib_buffer_t * first, u32 ** discard_vector)
+{
+ if (first->current_length >= VLIB_BUFFER_CHAIN_MIN_FIRST_DATA_SIZE ||
+ !(first->flags & VLIB_BUFFER_NEXT_PRESENT))
+ {
+ /* this is already big enough or not a chain */
+ return;
+ }
+ /* probe free list to find allocated buffer size to avoid overfill */
+ u32 index;
+ vlib_buffer_free_list_t *free_list =
+ vlib_buffer_get_buffer_free_list (vm, first, &index);
+
+ u32 want_first_size = clib_min (VLIB_BUFFER_CHAIN_MIN_FIRST_DATA_SIZE,
+ free_list->n_data_bytes -
+ first->current_data);
+ do
+ {
+ vlib_buffer_t *second = vlib_get_buffer (vm, first->next_buffer);
+ u32 need = want_first_size - first->current_length;
+ u32 amount_to_copy = clib_min (need, second->current_length);
+ clib_memcpy (((u8 *) vlib_buffer_get_current (first)) +
+ first->current_length,
+ vlib_buffer_get_current (second), amount_to_copy);
+ first->current_length += amount_to_copy;
+ vlib_buffer_advance (second, amount_to_copy);
+ if (first->flags & VLIB_BUFFER_TOTAL_LENGTH_VALID)
+ {
+ first->total_length_not_including_first_buffer -= amount_to_copy;
+ }
+ if (!second->current_length)
+ {
+ vec_add1 (*discard_vector, first->next_buffer);
+ if (second->flags & VLIB_BUFFER_NEXT_PRESENT)
+ {
+ first->next_buffer = second->next_buffer;
+ }
+ else
+ {
+ first->flags &= ~VLIB_BUFFER_NEXT_PRESENT;
+ }
+ second->flags &= ~VLIB_BUFFER_NEXT_PRESENT;
+ }
+ }
+ while ((first->current_length < want_first_size) &&
+ (first->flags & VLIB_BUFFER_NEXT_PRESENT));
+}
+
#endif /* included_vlib_buffer_funcs_h */
/*
diff --git a/src/vnet.am b/src/vnet.am
index 7eb4159e75c..8f884748d0f 100644
--- a/src/vnet.am
+++ b/src/vnet.am
@@ -343,6 +343,7 @@ libvnet_la_SOURCES += \
vnet/ip/ip4_pg.c \
vnet/ip/ip4_source_and_port_range_check.c \
vnet/ip/ip4_source_check.c \
+ vnet/ip/ip4_reassembly.c \
vnet/ip/ip6_format.c \
vnet/ip/ip6_forward.c \
vnet/ip/ip6_punt_drop.c \
@@ -350,6 +351,7 @@ libvnet_la_SOURCES += \
vnet/ip/ip6_input.c \
vnet/ip/ip6_neighbor.c \
vnet/ip/ip6_pg.c \
+ vnet/ip/ip6_reassembly.c \
vnet/ip/ip_api.c \
vnet/ip/ip_checksum.c \
vnet/ip/ip_frag.c \
diff --git a/src/vnet/buffer.h b/src/vnet/buffer.h
index c9f1d6ef6a1..5d7273aa04f 100644
--- a/src/vnet/buffer.h
+++ b/src/vnet/buffer.h
@@ -177,6 +177,18 @@ typedef struct
u8 code;
u32 data;
} icmp;
+
+ /* reassembly */
+ struct
+ {
+ u16 fragment_first;
+ u16 fragment_last;
+ u16 range_first;
+ u16 range_last;
+ u32 next_range_bi;
+ u16 ip6_frag_hdr_offset;
+ u16 estimated_mtu;
+ } reass;
};
} ip;
diff --git a/src/vnet/ip/icmp46_packet.h b/src/vnet/ip/icmp46_packet.h
index a86cbd57bdb..e761af8bf16 100644
--- a/src/vnet/ip/icmp46_packet.h
+++ b/src/vnet/ip/icmp46_packet.h
@@ -149,6 +149,7 @@
_ (parameter_problem, 0, erroneous_header_field) \
_ (parameter_problem, 1, unrecognized_next_header) \
_ (parameter_problem, 2, unrecognized_option) \
+ _ (parameter_problem, 3, first_fragment_has_incomplete_header_chain) \
_ (router_renumbering, 0, command) \
_ (router_renumbering, 1, result) \
_ (node_information_request, 0, data_contains_ip6_address) \
diff --git a/src/vnet/ip/icmp6.h b/src/vnet/ip/icmp6.h
index 9a3487b1c5c..1cac8240466 100644
--- a/src/vnet/ip/icmp6.h
+++ b/src/vnet/ip/icmp6.h
@@ -45,7 +45,7 @@
_ (DEST_UNREACH_SENT, "destination unreachable response sent") \
_ (PACKET_TOO_BIG_SENT, "packet too big response sent") \
_ (TTL_EXPIRE_SENT, "hop limit exceeded response sent") \
- _ (PARAM_PROBLEM_SENT, "parameter Pproblem response sent") \
+ _ (PARAM_PROBLEM_SENT, "parameter problem response sent") \
_ (DROP, "error message dropped")
diff --git a/src/vnet/ip/ip.api b/src/vnet/ip/ip.api
index 0da0a61fb75..6ed5a9d6bdb 100644
--- a/src/vnet/ip/ip.api
+++ b/src/vnet/ip/ip.api
@@ -812,6 +812,34 @@ autoreply define ioam_disable
u16 id;
};
+autoreply define ip_reassembly_set
+{
+ u32 client_index;
+ u32 context;
+ u32 timeout_ms;
+ u32 max_reassemblies;
+ u32 expire_walk_interval_ms;
+ u8 is_ip6;
+};
+
+define ip_reassembly_get
+{
+ u32 client_index;
+ u32 context;
+ u8 is_ip6;
+};
+
+define ip_reassembly_get_reply
+{
+ u32 client_index;
+ u32 context;
+ i32 retval;
+ u32 timeout_ms;
+ u32 max_reassemblies;
+ u32 expire_walk_interval_ms;
+ u8 is_ip6;
+};
+
/*
* Local Variables:
* eval: (c-set-style "gnu")
diff --git a/src/vnet/ip/ip4_error.h b/src/vnet/ip/ip4_error.h
index 4e545aeda3c..00b1c6daad5 100644
--- a/src/vnet/ip/ip4_error.h
+++ b/src/vnet/ip/ip4_error.h
@@ -71,14 +71,19 @@
_ (UNICAST_SOURCE_CHECK_FAILS, "ip4 unicast source check fails") \
\
/* Spoofed packets in ip4-rewrite-local */ \
- _(SPOOFED_LOCAL_PACKETS, "ip4 spoofed local-address packet drops") \
+ _ (SPOOFED_LOCAL_PACKETS, "ip4 spoofed local-address packet drops") \
\
/* Errors singalled by ip4-inacl */ \
_ (INACL_TABLE_MISS, "input ACL table-miss drops") \
_ (INACL_SESSION_DENY, "input ACL session deny drops") \
\
/* Erros from mfib-forward */ \
- _ (RPF_FAILURE, "Multicast RPF check failed")
+ _ (RPF_FAILURE, "Multicast RPF check failed") \
+ \
+ /* Errors signalled by ip4-reassembly */ \
+ _ (REASS_DUPLICATE_FRAGMENT, "duplicate/overlapping fragments") \
+ _ (REASS_LIMIT_REACHED, "drops due to concurrent reassemblies limit") \
+ _ (REASS_TIMEOUT, "fragments dropped due to reassembly timeout")
typedef enum
{
diff --git a/src/vnet/ip/ip4_forward.c b/src/vnet/ip/ip4_forward.c
index 3ddf6dfa783..2a21135bb3e 100755
--- a/src/vnet/ip/ip4_forward.c
+++ b/src/vnet/ip/ip4_forward.c
@@ -1582,8 +1582,12 @@ ip4_local_inline (vlib_main_t * vm,
/* 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;
- proto1 = ip4_is_fragment (ip1) ? 0xfe : ip1->protocol;
+ proto0 =
+ ip4_is_fragment (ip0) ? IP_PROTOCOL_VPP_FRAGMENTATION :
+ ip0->protocol;
+ proto1 =
+ ip4_is_fragment (ip1) ? IP_PROTOCOL_VPP_FRAGMENTATION :
+ ip1->protocol;
if (head_of_feature_arc == 0)
goto skip_checks;
@@ -1748,7 +1752,9 @@ ip4_local_inline (vlib_main_t * vm,
/* 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;
+ proto0 =
+ ip4_is_fragment (ip0) ? IP_PROTOCOL_VPP_FRAGMENTATION :
+ ip0->protocol;
if (head_of_feature_arc == 0 || p0->flags & VNET_BUFFER_F_IS_NATED)
goto skip_check;
@@ -1839,6 +1845,7 @@ VLIB_REGISTER_NODE (ip4_local_node) =
[IP_LOCAL_NEXT_PUNT] = "ip4-punt",
[IP_LOCAL_NEXT_UDP_LOOKUP] = "ip4-udp-lookup",
[IP_LOCAL_NEXT_ICMP] = "ip4-icmp-input",
+ [IP_LOCAL_NEXT_REASSEMBLY] = "ip4-reassembly",
},
};
/* *INDENT-ON* */
diff --git a/src/vnet/ip/ip4_input.c b/src/vnet/ip/ip4_input.c
index 121f40f473d..61c86ef112d 100644
--- a/src/vnet/ip/ip4_input.c
+++ b/src/vnet/ip/ip4_input.c
@@ -289,6 +289,7 @@ VLIB_REGISTER_NODE (ip4_input_node) = {
[IP4_INPUT_NEXT_LOOKUP] = "ip4-lookup",
[IP4_INPUT_NEXT_LOOKUP_MULTICAST] = "ip4-mfib-forward-lookup",
[IP4_INPUT_NEXT_ICMP_ERROR] = "ip4-icmp-error",
+ [IP4_INPUT_NEXT_REASSEMBLY] = "ip4-reassembly",
},
.format_buffer = format_ip4_header,
@@ -311,6 +312,7 @@ VLIB_REGISTER_NODE (ip4_input_no_checksum_node,static) = {
[IP4_INPUT_NEXT_LOOKUP] = "ip4-lookup",
[IP4_INPUT_NEXT_LOOKUP_MULTICAST] = "ip4-mfib-forward-lookup",
[IP4_INPUT_NEXT_ICMP_ERROR] = "ip4-icmp-error",
+ [IP4_INPUT_NEXT_REASSEMBLY] = "ip4-reassembly",
},
.format_buffer = format_ip4_header,
diff --git a/src/vnet/ip/ip4_input.h b/src/vnet/ip/ip4_input.h
index 75306a34db3..2692848f8ad 100644
--- a/src/vnet/ip/ip4_input.h
+++ b/src/vnet/ip/ip4_input.h
@@ -52,6 +52,7 @@ typedef enum
IP4_INPUT_NEXT_LOOKUP,
IP4_INPUT_NEXT_LOOKUP_MULTICAST,
IP4_INPUT_NEXT_ICMP_ERROR,
+ IP4_INPUT_NEXT_REASSEMBLY,
IP4_INPUT_N_NEXT,
} ip4_input_next_t;
diff --git a/src/vnet/ip/ip4_reassembly.c b/src/vnet/ip/ip4_reassembly.c
new file mode 100644
index 00000000000..6b8d66532fa
--- /dev/null
+++ b/src/vnet/ip/ip4_reassembly.c
@@ -0,0 +1,1277 @@
+/*
+ * 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 IPv4 Reassembly.
+ *
+ * This file contains the source code for IPv4 reassembly.
+ */
+
+#include <vppinfra/vec.h>
+#include <vnet/vnet.h>
+#include <vnet/ip/ip.h>
+#include <vppinfra/bihash_24_8.h>
+#include <vnet/ip/ip4_reassembly.h>
+
+#define MSEC_PER_SEC 1000
+#define IP4_REASS_TIMEOUT_DEFAULT_MS 100
+#define IP4_REASS_EXPIRE_WALK_INTERVAL_DEFAULT_MS 10000 // 10 seconds default
+#define IP4_REASS_MAX_REASSEMBLIES_DEAFULT 1024
+#define IP4_REASS_HT_LOAD_FACTOR (0.75)
+
+#define IP4_REASS_DEBUG_BUFFERS 0
+#if IP4_REASS_DEBUG_BUFFERS
+#define IP4_REASS_DEBUG_BUFFER(bi, what) \
+ do \
+ { \
+ u32 _bi = bi; \
+ printf (#what "buffer %u", _bi); \
+ vlib_buffer_t *_b = vlib_get_buffer (vm, _bi); \
+ while (_b->flags & VLIB_BUFFER_NEXT_PRESENT) \
+ { \
+ _bi = _b->next_buffer; \
+ printf ("[%u]", _bi); \
+ _b = vlib_get_buffer (vm, _bi); \
+ } \
+ printf ("\n"); \
+ fflush (stdout); \
+ } \
+ while (0)
+#else
+#define IP4_REASS_DEBUG_BUFFER(...)
+#endif
+
+static vlib_node_registration_t ip4_reass_node;
+
+typedef struct
+{
+ union
+ {
+ struct
+ {
+ // align by making this 4 octets even though its a 2 octets field
+ u32 xx_id;
+ ip4_address_t src;
+ ip4_address_t dst;
+ // align by making this 4 octets even though its a 2 octets field
+ u32 frag_id;
+ // align by making this 4 octets even though its a 1 octet field
+ u32 proto;
+ u32 unused;
+ };
+ u64 as_u64[3];
+ };
+} ip4_reass_key_t;
+
+always_inline u32
+ip4_reass_buffer_get_data_offset_no_check (vlib_buffer_t * b)
+{
+ vnet_buffer_opaque_t *vnb = vnet_buffer (b);
+ return vnb->ip.reass.range_first - vnb->ip.reass.fragment_first;
+}
+
+always_inline u32
+ip4_reass_buffer_get_data_offset (vlib_buffer_t * b)
+{
+ vnet_buffer_opaque_t *vnb = vnet_buffer (b);
+ ASSERT (vnb->ip.reass.range_first >= vnb->ip.reass.fragment_first);
+ return ip4_reass_buffer_get_data_offset_no_check (b);
+}
+
+always_inline u16
+ip4_reass_buffer_get_data_len_no_check (vlib_buffer_t * b)
+{
+ vnet_buffer_opaque_t *vnb = vnet_buffer (b);
+ return clib_min (vnb->ip.reass.range_last, vnb->ip.reass.fragment_last) -
+ (vnb->ip.reass.fragment_first + ip4_reass_buffer_get_data_offset (b)) + 1;
+}
+
+always_inline u16
+ip4_reass_buffer_get_data_len (vlib_buffer_t * b)
+{
+ vnet_buffer_opaque_t *vnb = vnet_buffer (b);
+ ASSERT (vnb->ip.reass.range_last > vnb->ip.reass.fragment_first);
+ return ip4_reass_buffer_get_data_len_no_check (b);
+}
+
+typedef struct
+{
+ // hash table key
+ ip4_reass_key_t key;
+ f64 first_heard;
+ // time when last packet was received
+ f64 last_heard;
+ // internal id of this reassembly
+ u32 id;
+ // buffer index of first buffer in this reassembly context
+ u32 first_bi;
+ // last octet of packet, ~0 until fragment without more_fragments arrives
+ u32 last_packet_octet;
+ // length of data collected so far
+ u32 data_len;
+ // trace operation counter
+ u32 trace_op_counter;
+} ip4_reass_t;
+
+typedef struct
+{
+ // IPv4 config
+ u32 timeout_ms;
+ f64 timeout;
+ u32 expire_walk_interval_ms;
+ u32 max_reass_n;
+
+ // IPv4 runtime
+ ip4_reass_t *pool;
+ clib_bihash_24_8_t hash;
+ u32 reass_n;
+ u32 id_counter;
+ u32 buffers_n;
+
+ // convenience
+ vlib_main_t *vlib_main;
+ vnet_main_t *vnet_main;
+
+ // node index of ip4-drop node
+ u32 ip4_drop_idx;
+ u32 ip4_reass_expire_node_idx;
+
+} ip4_reass_main_t;
+
+ip4_reass_main_t ip4_reass_main;
+
+typedef enum
+{
+ IP4_REASSEMBLY_NEXT_INPUT,
+ IP4_REASSEMBLY_NEXT_DROP,
+ IP4_REASSEMBLY_N_NEXT,
+} ip4_reass_next_t;
+
+typedef enum
+{
+ RANGE_NEW,
+ RANGE_SHRINK,
+ RANGE_DISCARD,
+ RANGE_OVERLAP,
+ FINALIZE,
+} ip4_reass_trace_operation_e;
+
+typedef struct
+{
+ u16 range_first;
+ u16 range_last;
+ u32 range_bi;
+ i32 data_offset;
+ u32 data_len;
+ u32 first_bi;
+} ip4_reass_range_trace_t;
+
+typedef struct
+{
+ ip4_reass_trace_operation_e action;
+ u32 pool_index;
+ u32 reass_id;
+ ip4_reass_range_trace_t trace_range;
+ u32 size_diff;
+ u32 op_id;
+ u32 fragment_first;
+ u32 fragment_last;
+ u32 total_data_len;
+} ip4_reass_trace_t;
+
+void
+ip4_reass_trace_details (vlib_main_t * vm, u32 bi,
+ ip4_reass_range_trace_t * trace)
+{
+ vlib_buffer_t *b = vlib_get_buffer (vm, bi);
+ vnet_buffer_opaque_t *vnb = vnet_buffer (b);
+ trace->range_first = vnb->ip.reass.range_first;
+ trace->range_last = vnb->ip.reass.range_last;
+ trace->data_offset = ip4_reass_buffer_get_data_offset_no_check (b);
+ trace->data_len = ip4_reass_buffer_get_data_len_no_check (b);
+ trace->range_bi = bi;
+}
+
+u8 *
+format_ip4_reass_range_trace (u8 * s, va_list * args)
+{
+ ip4_reass_range_trace_t *trace = va_arg (*args, ip4_reass_range_trace_t *);
+ s = format (s, "range: [%u, %u], off %d, len %u, bi %u", trace->range_first,
+ trace->range_last, trace->data_offset, trace->data_len,
+ trace->range_bi);
+ return s;
+}
+
+u8 *
+format_ip4_reass_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 *);
+ ip4_reass_trace_t *t = va_arg (*args, ip4_reass_trace_t *);
+ s = format (s, "reass id: %u, op id: %u ", t->reass_id, t->op_id);
+ u32 indent = format_get_indent (s);
+ s = format (s, "first bi: %u, data len: %u, ip/fragment[%u, %u]",
+ t->trace_range.first_bi, t->total_data_len, t->fragment_first,
+ t->fragment_last);
+ switch (t->action)
+ {
+ case RANGE_SHRINK:
+ s = format (s, "\n%Ushrink %U by %u", format_white_space, indent,
+ format_ip4_reass_range_trace, &t->trace_range,
+ t->size_diff);
+ break;
+ case RANGE_DISCARD:
+ s = format (s, "\n%Udiscard %U", format_white_space, indent,
+ format_ip4_reass_range_trace, &t->trace_range);
+ break;
+ case RANGE_NEW:
+ s = format (s, "\n%Unew %U", format_white_space, indent,
+ format_ip4_reass_range_trace, &t->trace_range);
+ break;
+ case RANGE_OVERLAP:
+ s = format (s, "\n%Uoverlapping/ignored %U", format_white_space, indent,
+ format_ip4_reass_range_trace, &t->trace_range);
+ break;
+ case FINALIZE:
+ s = format (s, "\n%Ufinalize reassembly", format_white_space, indent);
+ break;
+ }
+ return s;
+}
+
+void
+ip4_reass_add_trace (vlib_main_t * vm, vlib_node_runtime_t * node,
+ ip4_reass_main_t * rm, ip4_reass_t * reass, u32 bi,
+ ip4_reass_trace_operation_e action, u32 size_diff)
+{
+ vlib_buffer_t *b = vlib_get_buffer (vm, bi);
+ vnet_buffer_opaque_t *vnb = vnet_buffer (b);
+ ip4_reass_trace_t *t = vlib_add_trace (vm, node, b, sizeof (t[0]));
+ t->pool_index = reass - rm->pool;
+ t->reass_id = reass->id;
+ t->action = action;
+ ip4_reass_trace_details (vm, bi, &t->trace_range);
+ t->size_diff = size_diff;
+ t->op_id = reass->trace_op_counter;
+ ++reass->trace_op_counter;
+ t->fragment_first = vnb->ip.reass.fragment_first;
+ t->fragment_last = vnb->ip.reass.fragment_last;
+ t->trace_range.first_bi = reass->first_bi;
+ t->total_data_len = reass->data_len;
+#if 0
+ static u8 *s = NULL;
+ s = format (s, "%U", format_ip4_reass_trace, NULL, NULL, t);
+ printf ("%.*s\n", vec_len (s), s);
+ fflush (stdout);
+ vec_reset_length (s);
+#endif
+}
+
+void
+ip4_reass_free (ip4_reass_main_t * rm, ip4_reass_t * reass)
+{
+ clib_bihash_kv_24_8_t kv;
+ kv.key[0] = reass->key.as_u64[0];
+ kv.key[1] = reass->key.as_u64[1];
+ kv.key[2] = reass->key.as_u64[2];
+ clib_bihash_add_del_24_8 (&rm->hash, &kv, 0);
+ pool_put (rm->pool, reass);
+ --rm->reass_n;
+}
+
+static void
+ip4_reass_on_timeout (vlib_main_t * vm, ip4_reass_main_t * rm,
+ ip4_reass_t * reass, u32 ** vec_drop_timeout)
+{
+ u32 range_bi = reass->first_bi;
+ vlib_buffer_t *range_b;
+ vnet_buffer_opaque_t *range_vnb;
+ while (~0 != range_bi)
+ {
+ range_b = vlib_get_buffer (vm, range_bi);
+ range_vnb = vnet_buffer (range_b);
+ u32 bi = range_bi;
+ while (~0 != bi)
+ {
+ vec_add1 (*vec_drop_timeout, bi);
+ vlib_buffer_t *b = vlib_get_buffer (vm, bi);
+ if (b->flags & VLIB_BUFFER_NEXT_PRESENT)
+ {
+ bi = b->next_buffer;
+ b->flags &= ~VLIB_BUFFER_NEXT_PRESENT;
+ }
+ else
+ {
+ bi = ~0;
+ }
+ }
+ range_bi = range_vnb->ip.reass.next_range_bi;
+ }
+}
+
+ip4_reass_t *
+ip4_reass_find_or_create (vlib_main_t * vm, ip4_reass_main_t * rm,
+ ip4_reass_key_t * k, u32 ** vec_drop_timeout)
+{
+ ip4_reass_t *reass = NULL;
+ f64 now = vlib_time_now (rm->vlib_main);
+ clib_bihash_kv_24_8_t kv, value;
+ kv.key[0] = k->as_u64[0];
+ kv.key[1] = k->as_u64[1];
+ kv.key[2] = k->as_u64[2];
+
+ if (!clib_bihash_search_24_8 (&rm->hash, &kv, &value))
+ {
+ reass = pool_elt_at_index (rm->pool, value.value);
+ if (now > reass->last_heard + rm->timeout)
+ {
+ ip4_reass_on_timeout (vm, rm, reass, vec_drop_timeout);
+ ip4_reass_free (rm, reass);
+ reass = NULL;
+ }
+ }
+
+ if (reass)
+ {
+ reass->last_heard = now;
+ return reass;
+ }
+
+ if (rm->reass_n >= rm->max_reass_n)
+ {
+ reass = NULL;
+ return reass;
+ }
+ else
+ {
+ pool_get (rm->pool, reass);
+ memset (reass, 0, sizeof (*reass));
+ reass->id = rm->id_counter;
+ ++rm->id_counter;
+ reass->first_bi = ~0;
+ reass->last_packet_octet = ~0;
+ reass->data_len = 0;
+ ++rm->reass_n;
+ }
+
+ reass->key.as_u64[0] = kv.key[0] = k->as_u64[0];
+ reass->key.as_u64[1] = kv.key[1] = k->as_u64[1];
+ reass->key.as_u64[2] = kv.key[2] = k->as_u64[2];
+ kv.value = reass - rm->pool;
+ reass->last_heard = now;
+
+ if (clib_bihash_add_del_24_8 (&rm->hash, &kv, 1))
+ {
+ ip4_reass_free (rm, reass);
+ reass = NULL;
+ }
+
+ return reass;
+}
+
+void
+ip4_reass_finalize (vlib_main_t * vm, vlib_node_runtime_t * node,
+ ip4_reass_main_t * rm, ip4_reass_t * reass, u32 * bi0,
+ u32 * next0, vlib_error_t * error0, u32 next_input,
+ u32 ** vec_drop_compress, u32 ** vec_drop_overlap)
+{
+ ASSERT (~0 != reass->first_bi);
+ vlib_buffer_t *first_b = vlib_get_buffer (vm, reass->first_bi);
+ vlib_buffer_t *last_b = NULL;
+ u32 sub_chain_bi = reass->first_bi;
+ u32 total_length = 0;
+ u32 buf_cnt = 0;
+ u32 dropped_cnt = 0;
+ do
+ {
+ u32 tmp_bi = sub_chain_bi;
+ vlib_buffer_t *tmp = vlib_get_buffer (vm, tmp_bi);
+ ip4_header_t *ip = vlib_buffer_get_current (tmp);
+ u32 data_len = ip4_reass_buffer_get_data_len (tmp);
+ u32 trim_front =
+ ip4_header_bytes (ip) + ip4_reass_buffer_get_data_offset (tmp);
+ u32 trim_end =
+ vlib_buffer_length_in_chain (vm, tmp) - trim_front - data_len;
+ if (tmp_bi == reass->first_bi)
+ {
+ /* first buffer - keep ip4 header */
+ ASSERT (0 == ip4_reass_buffer_get_data_offset (tmp));
+ trim_front = 0;
+ trim_end = vlib_buffer_length_in_chain (vm, tmp) - data_len -
+ ip4_header_bytes (ip);
+ }
+ u32 keep_data =
+ vlib_buffer_length_in_chain (vm, tmp) - trim_front - trim_end;
+ while (1)
+ {
+ ++buf_cnt;
+ if (trim_front)
+ {
+ if (trim_front > tmp->current_length)
+ {
+ /* drop whole buffer */
+ vec_add1 (*vec_drop_compress, tmp_bi);
+ ++dropped_cnt;
+ trim_front -= tmp->current_length;
+ ASSERT (tmp->flags & VLIB_BUFFER_NEXT_PRESENT);
+ tmp->flags &= ~VLIB_BUFFER_NEXT_PRESENT;
+ tmp_bi = tmp->next_buffer;
+ tmp = vlib_get_buffer (vm, tmp_bi);
+ continue;
+ }
+ else
+ {
+ vlib_buffer_advance (tmp, trim_front);
+ trim_front = 0;
+ }
+ }
+ if (keep_data)
+ {
+ if (last_b)
+ {
+ last_b->flags |= VLIB_BUFFER_NEXT_PRESENT;
+ last_b->next_buffer = tmp_bi;
+ }
+ last_b = tmp;
+ if (keep_data <= tmp->current_length)
+ {
+ tmp->current_length = keep_data;
+ keep_data = 0;
+ }
+ else
+ {
+ keep_data -= tmp->current_length;
+ ASSERT (tmp->flags & VLIB_BUFFER_NEXT_PRESENT);
+ }
+ total_length += tmp->current_length;
+ }
+ else
+ {
+ vec_add1 (*vec_drop_overlap, tmp_bi);
+ ASSERT (reass->first_bi != tmp_bi);
+ ++dropped_cnt;
+ }
+ if (tmp->flags & VLIB_BUFFER_NEXT_PRESENT)
+ {
+ tmp_bi = tmp->next_buffer;
+ tmp = vlib_get_buffer (vm, tmp->next_buffer);
+ }
+ else
+ {
+ break;
+ }
+ }
+ sub_chain_bi =
+ vnet_buffer (vlib_get_buffer (vm, sub_chain_bi))->ip.
+ reass.next_range_bi;
+ }
+ while (~0 != sub_chain_bi);
+ last_b->flags &= ~VLIB_BUFFER_NEXT_PRESENT;
+ ASSERT (rm->buffers_n >= (buf_cnt - dropped_cnt));
+ rm->buffers_n -= buf_cnt - dropped_cnt;
+ ASSERT (total_length >= first_b->current_length);
+ total_length -= first_b->current_length;
+ first_b->flags |= VLIB_BUFFER_TOTAL_LENGTH_VALID;
+ first_b->total_length_not_including_first_buffer = total_length;
+ ip4_header_t *ip = vlib_buffer_get_current (first_b);
+ ip->flags_and_fragment_offset = 0;
+ ip->length = clib_host_to_net_u16 (first_b->current_length + total_length);
+ ip->checksum = ip4_header_checksum (ip);
+ vlib_buffer_chain_compress (vm, first_b, vec_drop_compress);
+ if (PREDICT_FALSE (first_b->flags & VLIB_BUFFER_IS_TRACED))
+ {
+ ip4_reass_add_trace (vm, node, rm, reass, reass->first_bi, FINALIZE, 0);
+#if 0
+ // following code does a hexdump of packet fragments to stdout ...
+ do
+ {
+ u32 bi = reass->first_bi;
+ u8 *s = NULL;
+ while (~0 != bi)
+ {
+ vlib_buffer_t *b = vlib_get_buffer (vm, bi);
+ s = format (s, "%u: %U\n", bi, format_hexdump,
+ vlib_buffer_get_current (b), b->current_length);
+ if (b->flags & VLIB_BUFFER_NEXT_PRESENT)
+ {
+ bi = b->next_buffer;
+ }
+ else
+ {
+ break;
+ }
+ }
+ printf ("%.*s\n", vec_len (s), s);
+ fflush (stdout);
+ vec_free (s);
+ }
+ while (0);
+#endif
+ }
+ *bi0 = reass->first_bi;
+ *next0 = next_input;
+ *error0 = IP4_ERROR_NONE;
+ ip4_reass_free (rm, reass);
+ reass = NULL;
+}
+
+static u32
+ip4_reass_get_buffer_chain_length (vlib_main_t * vm, vlib_buffer_t * b)
+{
+ u32 len = 0;
+ while (b)
+ {
+ ++len;
+ if (PREDICT_FALSE (b->flags & VLIB_BUFFER_NEXT_PRESENT))
+ {
+ b = vlib_get_buffer (vm, b->next_buffer);
+ }
+ else
+ {
+ break;
+ }
+ }
+ return len;
+}
+
+static void
+ip4_reass_insert_range_in_chain (vlib_main_t * vm,
+ ip4_reass_main_t * rm,
+ ip4_reass_t * reass,
+ u32 prev_range_bi, u32 new_next_bi)
+{
+
+ vlib_buffer_t *new_next_b = vlib_get_buffer (vm, new_next_bi);
+ vnet_buffer_opaque_t *new_next_vnb = vnet_buffer (new_next_b);
+ if (~0 != prev_range_bi)
+ {
+ vlib_buffer_t *prev_b = vlib_get_buffer (vm, prev_range_bi);
+ vnet_buffer_opaque_t *prev_vnb = vnet_buffer (prev_b);
+ new_next_vnb->ip.reass.next_range_bi = prev_vnb->ip.reass.next_range_bi;
+ prev_vnb->ip.reass.next_range_bi = new_next_bi;
+ }
+ else
+ {
+ if (~0 != reass->first_bi)
+ {
+ new_next_vnb->ip.reass.next_range_bi = reass->first_bi;
+ }
+ reass->first_bi = new_next_bi;
+ }
+ reass->data_len += ip4_reass_buffer_get_data_len (new_next_b);
+ rm->buffers_n += ip4_reass_get_buffer_chain_length (vm, new_next_b);
+}
+
+static void
+ip4_reass_remove_range_from_chain (vlib_main_t * vm,
+ vlib_node_runtime_t * node,
+ ip4_reass_main_t * rm,
+ u32 ** vec_drop_overlap,
+ ip4_reass_t * reass, u32 prev_range_bi,
+ u32 discard_bi)
+{
+ vlib_buffer_t *discard_b = vlib_get_buffer (vm, discard_bi);
+ vnet_buffer_opaque_t *discard_vnb = vnet_buffer (discard_b);
+ if (~0 != prev_range_bi)
+ {
+ vlib_buffer_t *prev_b = vlib_get_buffer (vm, prev_range_bi);
+ vnet_buffer_opaque_t *prev_vnb = vnet_buffer (prev_b);
+ ASSERT (prev_vnb->ip.reass.next_range_bi == discard_bi);
+ prev_vnb->ip.reass.next_range_bi = discard_vnb->ip.reass.next_range_bi;
+ }
+ else
+ {
+ reass->first_bi = discard_vnb->ip.reass.next_range_bi;
+ }
+ reass->data_len -= ip4_reass_buffer_get_data_len (discard_b);
+ while (1)
+ {
+ vec_add1 (*vec_drop_overlap, discard_bi);
+ if (PREDICT_FALSE (discard_b->flags & VLIB_BUFFER_IS_TRACED))
+ {
+ ip4_reass_add_trace (vm, node, rm, reass, discard_bi, RANGE_DISCARD,
+ 0);
+ }
+ if (discard_b->flags & VLIB_BUFFER_NEXT_PRESENT)
+ {
+ discard_b->flags &= ~VLIB_BUFFER_NEXT_PRESENT;
+ discard_bi = discard_b->next_buffer;
+ discard_b = vlib_get_buffer (vm, discard_bi);
+ }
+ else
+ {
+ break;
+ }
+ }
+}
+
+void
+ip4_reass_update (vlib_main_t * vm, vlib_node_runtime_t * node,
+ ip4_reass_main_t * rm, ip4_reass_t * reass, u32 * bi0,
+ u32 * next0, vlib_error_t * error0,
+ u32 ** vec_drop_overlap, u32 ** vec_drop_compress,
+ u32 next_input, u32 next_drop)
+{
+ int consumed = 0;
+ vlib_buffer_t *fb = vlib_get_buffer (vm, *bi0);
+ ip4_header_t *fip = vlib_buffer_get_current (fb);
+ ASSERT (fb->current_length >= sizeof (*fip));
+ vnet_buffer_opaque_t *fvnb = vnet_buffer (fb);
+ u32 fragment_first = fvnb->ip.reass.fragment_first =
+ ip4_get_fragment_offset_bytes (fip);
+ u32 fragment_length =
+ clib_net_to_host_u16 (fip->length) - ip4_header_bytes (fip);
+ u32 fragment_last = fvnb->ip.reass.fragment_last =
+ fragment_first + fragment_length - 1;
+ int more_fragments = ip4_get_fragment_more (fip);
+ u32 candidate_range_bi = reass->first_bi;
+ u32 prev_range_bi = ~0;
+ fvnb->ip.reass.range_first = fragment_first;
+ fvnb->ip.reass.range_last = fragment_last;
+ fvnb->ip.reass.next_range_bi = ~0;
+ if (!more_fragments)
+ {
+ reass->last_packet_octet = fragment_last;
+ }
+ if (~0 == reass->first_bi)
+ {
+ // starting a new reassembly
+ ip4_reass_insert_range_in_chain (vm, rm, reass, prev_range_bi, *bi0);
+ if (PREDICT_FALSE (fb->flags & VLIB_BUFFER_IS_TRACED))
+ {
+ ip4_reass_add_trace (vm, node, rm, reass, *bi0, RANGE_NEW, 0);
+ }
+ *bi0 = ~0;
+ fvnb->ip.reass.estimated_mtu = clib_net_to_host_u16 (fip->length);
+ return;
+ }
+ fvnb->ip.reass.estimated_mtu = clib_min (clib_net_to_host_u16 (fip->length),
+ fvnb->ip.reass.estimated_mtu);
+ while (~0 != candidate_range_bi)
+ {
+ vlib_buffer_t *candidate_b = vlib_get_buffer (vm, candidate_range_bi);
+ vnet_buffer_opaque_t *candidate_vnb = vnet_buffer (candidate_b);
+ if (fragment_first > candidate_vnb->ip.reass.range_last)
+ {
+ // this fragments starts after candidate range
+ prev_range_bi = candidate_range_bi;
+ candidate_range_bi = candidate_vnb->ip.reass.next_range_bi;
+ if (candidate_vnb->ip.reass.range_last < fragment_last &&
+ ~0 == candidate_range_bi)
+ {
+ // special case - this fragment falls beyond all known ranges
+ ip4_reass_insert_range_in_chain (vm, rm, reass, prev_range_bi,
+ *bi0);
+ consumed = 1;
+ break;
+ }
+ continue;
+ }
+ if (fragment_last < candidate_vnb->ip.reass.range_first)
+ {
+ // this fragment ends before candidate range without any overlap
+ ip4_reass_insert_range_in_chain (vm, rm, reass, prev_range_bi,
+ *bi0);
+ consumed = 1;
+ }
+ else
+ {
+ if (fragment_first >= candidate_vnb->ip.reass.range_first &&
+ fragment_last <= candidate_vnb->ip.reass.range_last)
+ {
+ // this fragment is a (sub)part of existing range, ignore it
+ if (PREDICT_FALSE (fb->flags & VLIB_BUFFER_IS_TRACED))
+ {
+ ip4_reass_add_trace (vm, node, rm, reass, *bi0,
+ RANGE_OVERLAP, 0);
+ }
+ break;
+ }
+ int discard_candidate = 0;
+ if (fragment_first < candidate_vnb->ip.reass.range_first)
+ {
+ u32 overlap =
+ fragment_last - candidate_vnb->ip.reass.range_first + 1;
+ if (overlap < ip4_reass_buffer_get_data_len (candidate_b))
+ {
+ candidate_vnb->ip.reass.range_first += overlap;
+ ASSERT (reass->data_len >= overlap);
+ reass->data_len -= overlap;
+ if (PREDICT_FALSE (fb->flags & VLIB_BUFFER_IS_TRACED))
+ {
+ ip4_reass_add_trace (vm, node, rm, reass,
+ candidate_range_bi, RANGE_SHRINK,
+ overlap);
+ }
+ ip4_reass_insert_range_in_chain (vm, rm, reass,
+ prev_range_bi, *bi0);
+ consumed = 1;
+ }
+ else
+ {
+ discard_candidate = 1;
+ }
+ }
+ else if (fragment_last > candidate_vnb->ip.reass.range_last)
+ {
+ u32 overlap =
+ candidate_vnb->ip.reass.range_last - fragment_first + 1;
+ if (overlap < ip4_reass_buffer_get_data_len (candidate_b))
+ {
+ fvnb->ip.reass.range_first += overlap;
+ if (~0 != candidate_vnb->ip.reass.next_range_bi)
+ {
+ prev_range_bi = candidate_range_bi;
+ candidate_range_bi =
+ candidate_vnb->ip.reass.next_range_bi;
+ continue;
+ }
+ else
+ {
+ // special case - last range discarded
+ ip4_reass_insert_range_in_chain (vm, rm, reass,
+ candidate_range_bi,
+ *bi0);
+ consumed = 1;
+ }
+ }
+ else
+ {
+ discard_candidate = 1;
+ }
+ }
+ else
+ {
+ discard_candidate = 1;
+ }
+ if (discard_candidate)
+ {
+ u32 next_range_bi = candidate_vnb->ip.reass.next_range_bi;
+ // discard candidate range, probe next range
+ ip4_reass_remove_range_from_chain (vm, node, rm,
+ vec_drop_overlap, reass,
+ prev_range_bi,
+ candidate_range_bi);
+ if (~0 != next_range_bi)
+ {
+ candidate_range_bi = next_range_bi;
+ continue;
+ }
+ else
+ {
+ // special case - last range discarded
+ ip4_reass_insert_range_in_chain (vm, rm, reass,
+ prev_range_bi, *bi0);
+ consumed = 1;
+ }
+ }
+ }
+ break;
+ }
+ if (consumed)
+ {
+ if (PREDICT_FALSE (fb->flags & VLIB_BUFFER_IS_TRACED))
+ {
+ ip4_reass_add_trace (vm, node, rm, reass, *bi0, RANGE_NEW, 0);
+ }
+ }
+ if (~0 != reass->last_packet_octet &&
+ reass->data_len == reass->last_packet_octet + 1)
+ {
+ ip4_reass_finalize (vm, node, rm, reass, bi0, next0, error0, next_input,
+ vec_drop_compress, vec_drop_overlap);
+ }
+ else
+ {
+ if (consumed)
+ {
+ *bi0 = ~0;
+ }
+ else
+ {
+ *next0 = next_drop;
+ *error0 = IP4_ERROR_REASS_DUPLICATE_FRAGMENT;
+ }
+ }
+}
+
+always_inline uword
+ip4_reassembly (vlib_main_t * vm, vlib_node_runtime_t * node,
+ vlib_frame_t * frame)
+{
+ u32 *from = vlib_frame_vector_args (frame);
+ u32 n_left_from, n_left_to_next, *to_next, next_index;
+ ip4_reass_main_t *rm = &ip4_reass_main;
+
+ n_left_from = frame->n_vectors;
+ next_index = node->cached_next_index;
+ static u32 *vec_drop_timeout = NULL; // indexes of buffers which timed out
+ static u32 *vec_drop_overlap = NULL; // indexes of buffers which were discarded due to overlap
+ static u32 *vec_drop_compress = NULL; // indexes of buffers dicarded due to buffer compression
+ while (n_left_from > 0 || vec_len (vec_drop_timeout) > 0 ||
+ vec_len (vec_drop_overlap) > 0)
+ {
+ vlib_get_next_frame (vm, node, next_index, to_next, n_left_to_next);
+
+ while (vec_len (vec_drop_timeout) > 0 && n_left_to_next > 0)
+ {
+ u32 bi = vec_pop (vec_drop_timeout);
+ vlib_buffer_t *b = vlib_get_buffer (vm, bi);
+ b->error = node->errors[IP4_ERROR_REASS_TIMEOUT];
+ to_next[0] = bi;
+ to_next += 1;
+ n_left_to_next -= 1;
+ vlib_validate_buffer_enqueue_x1 (vm, node, next_index, to_next,
+ n_left_to_next, bi,
+ IP4_REASSEMBLY_NEXT_DROP);
+ IP4_REASS_DEBUG_BUFFER (bi, enqueue_drop_timeout);
+ ASSERT (rm->buffers_n > 0);
+ --rm->buffers_n;
+ }
+
+ while (vec_len (vec_drop_overlap) > 0 && n_left_to_next > 0)
+ {
+ u32 bi = vec_pop (vec_drop_overlap);
+ vlib_buffer_t *b = vlib_get_buffer (vm, bi);
+ b->error = node->errors[IP4_ERROR_REASS_DUPLICATE_FRAGMENT];
+ to_next[0] = bi;
+ to_next += 1;
+ n_left_to_next -= 1;
+ vlib_validate_buffer_enqueue_x1 (vm, node, next_index, to_next,
+ n_left_to_next, bi,
+ IP4_REASSEMBLY_NEXT_DROP);
+ IP4_REASS_DEBUG_BUFFER (bi, enqueue_drop_duplicate_fragment);
+ ASSERT (rm->buffers_n > 0);
+ --rm->buffers_n;
+ }
+
+ while (vec_len (vec_drop_compress) > 0 && n_left_to_next > 0)
+ {
+ u32 bi = vec_pop (vec_drop_compress);
+ vlib_buffer_t *b = vlib_get_buffer (vm, bi);
+ b->error = node->errors[IP4_ERROR_NONE];
+ to_next[0] = bi;
+ to_next += 1;
+ n_left_to_next -= 1;
+ vlib_validate_buffer_enqueue_x1 (vm, node, next_index, to_next,
+ n_left_to_next, bi,
+ IP4_REASSEMBLY_NEXT_DROP);
+ IP4_REASS_DEBUG_BUFFER (bi, enqueue_drop_compress);
+ ASSERT (rm->buffers_n > 0);
+ --rm->buffers_n;
+ }
+
+ while (n_left_from > 0 && n_left_to_next > 0)
+ {
+ u32 bi0;
+ vlib_buffer_t *b0;
+ u32 next0; //, error0;
+
+ bi0 = from[0];
+ b0 = vlib_get_buffer (vm, bi0);
+
+ ip4_header_t *ip0 = vlib_buffer_get_current (b0);
+ ip4_reass_key_t k;
+ k.src.as_u32 = ip0->src_address.as_u32;
+ k.dst.as_u32 = ip0->dst_address.as_u32;
+ k.xx_id = vnet_buffer (b0)->sw_if_index[VLIB_RX];
+ k.frag_id = ip0->fragment_id;
+ k.proto = ip0->protocol;
+ k.unused = 0;
+ ip4_reass_t *reass =
+ ip4_reass_find_or_create (vm, rm, &k, &vec_drop_timeout);
+
+ u32 error0 = IP4_ERROR_NONE;
+ if (reass)
+ {
+ ip4_reass_update (vm, node, rm, reass, &bi0, &next0, &error0,
+ &vec_drop_overlap, &vec_drop_compress,
+ IP4_REASSEMBLY_NEXT_INPUT,
+ IP4_REASSEMBLY_NEXT_DROP);
+ }
+ else
+ {
+ next0 = IP4_REASSEMBLY_NEXT_DROP;
+ error0 = IP4_ERROR_REASS_LIMIT_REACHED;
+ }
+
+ b0->error = node->errors[error0];
+
+ if (bi0 != ~0)
+ {
+ to_next[0] = bi0;
+ to_next += 1;
+ n_left_to_next -= 1;
+ vlib_validate_buffer_enqueue_x1 (vm, node, next_index, to_next,
+ n_left_to_next, bi0, next0);
+ IP4_REASS_DEBUG_BUFFER (bi0, enqueue_next);
+ }
+
+ from += 1;
+ n_left_from -= 1;
+ }
+
+ vlib_put_next_frame (vm, node, next_index, n_left_to_next);
+ }
+
+ return frame->n_vectors;
+}
+
+static char *ip4_reassembly_error_strings[] = {
+#define _(sym, string) string,
+ foreach_ip4_error
+#undef _
+};
+
+/* *INDENT-OFF* */
+VLIB_REGISTER_NODE (ip4_reass_node, static) = {
+ .function = ip4_reassembly,
+ .name = "ip4-reassembly",
+ .vector_size = sizeof (u32),
+ .format_trace = format_ip4_reass_trace,
+ .n_errors = ARRAY_LEN (ip4_reassembly_error_strings),
+ .error_strings = ip4_reassembly_error_strings,
+ .n_next_nodes = IP4_REASSEMBLY_N_NEXT,
+ .next_nodes =
+ {
+ [IP4_REASSEMBLY_NEXT_INPUT] = "ip4-input",
+ [IP4_REASSEMBLY_NEXT_DROP] = "ip4-drop",
+ },
+};
+/* *INDENT-ON* */
+
+VLIB_NODE_FUNCTION_MULTIARCH (ip4_reass_node, ip4_reassembly)
+ static u32 ip4_reass_get_nbuckets ()
+{
+ ip4_reass_main_t *rm = &ip4_reass_main;
+ u32 nbuckets;
+ u8 i;
+
+ nbuckets = (u32) (rm->max_reass_n / IP4_REASS_HT_LOAD_FACTOR);
+
+ for (i = 0; i < 31; i++)
+ if ((1 << i) >= nbuckets)
+ break;
+ nbuckets = 1 << i;
+
+ return nbuckets;
+}
+
+typedef enum
+{
+ IP4_EVENT_CONFIG_CHANGED = 1,
+} ip4_reass_event_t;
+
+typedef struct
+{
+ int failure;
+ clib_bihash_24_8_t *new_hash;
+} ip4_rehash_cb_ctx;
+
+void
+ip4_rehash_cb (clib_bihash_kv_24_8_t * kv, void *_ctx)
+{
+ ip4_rehash_cb_ctx *ctx = _ctx;
+ if (clib_bihash_add_del_24_8 (ctx->new_hash, kv, 1))
+ {
+ ctx->failure = 1;
+ }
+}
+
+vnet_api_error_t
+ip4_reass_set (u32 timeout_ms, u32 max_reassemblies,
+ u32 expire_walk_interval_ms)
+{
+ u32 old_nbuckets = ip4_reass_get_nbuckets ();
+ ip4_reass_main.timeout_ms = timeout_ms;
+ ip4_reass_main.timeout = (f64) timeout_ms / (f64) MSEC_PER_SEC;
+ ip4_reass_main.max_reass_n = max_reassemblies;
+ ip4_reass_main.expire_walk_interval_ms = expire_walk_interval_ms;
+ vlib_process_signal_event (ip4_reass_main.vlib_main,
+ ip4_reass_main.ip4_reass_expire_node_idx,
+ IP4_EVENT_CONFIG_CHANGED, 0);
+ u32 new_nbuckets = ip4_reass_get_nbuckets ();
+ if (ip4_reass_main.max_reass_n > 0 && new_nbuckets > 1 &&
+ new_nbuckets != old_nbuckets)
+ {
+ clib_bihash_24_8_t new_hash;
+ memset (&new_hash, 0, sizeof (new_hash));
+ ip4_rehash_cb_ctx ctx;
+ ctx.failure = 0;
+ ctx.new_hash = &new_hash;
+ clib_bihash_init_24_8 (&new_hash, "ip4-reass", new_nbuckets,
+ new_nbuckets * 1024);
+ clib_bihash_foreach_key_value_pair_24_8 (&ip4_reass_main.hash,
+ ip4_rehash_cb, &ctx);
+ if (ctx.failure)
+ {
+ clib_bihash_free_24_8 (&new_hash);
+ return -1;
+ }
+ else
+ {
+ clib_bihash_free_24_8 (&ip4_reass_main.hash);
+ clib_memcpy (&ip4_reass_main.hash, &new_hash,
+ sizeof (ip4_reass_main.hash));
+ }
+ }
+ return 0;
+}
+
+vnet_api_error_t
+ip4_reass_get (u32 * timeout_ms, u32 * max_reassemblies,
+ u32 * expire_walk_interval_ms)
+{
+ *timeout_ms = ip4_reass_main.timeout_ms;
+ *max_reassemblies = ip4_reass_main.max_reass_n;
+ *expire_walk_interval_ms = ip4_reass_main.expire_walk_interval_ms;
+ return 0;
+}
+
+clib_error_t *
+ip4_reass_init_function (vlib_main_t * vm)
+{
+ ip4_reass_main_t *rm = &ip4_reass_main;
+ clib_error_t *error = 0;
+ u32 nbuckets;
+
+ rm->vlib_main = vm;
+ rm->vnet_main = vnet_get_main ();
+
+ rm->reass_n = 0;
+ pool_alloc (rm->pool, rm->max_reass_n);
+ ip4_reass_set (IP4_REASS_TIMEOUT_DEFAULT_MS,
+ IP4_REASS_MAX_REASSEMBLIES_DEAFULT,
+ IP4_REASS_EXPIRE_WALK_INTERVAL_DEFAULT_MS);
+
+ nbuckets = ip4_reass_get_nbuckets ();
+ clib_bihash_init_24_8 (&rm->hash, "ip4-reass", nbuckets, nbuckets * 1024);
+
+ vlib_node_t *node = vlib_get_node_by_name (vm, (u8 *) "ip4-drop");
+ ASSERT (node);
+ rm->ip4_drop_idx = node->index;
+ node = vlib_get_node_by_name (vm, (u8 *) "ip4-reassembly-expire-walk");
+ ASSERT (node);
+ rm->ip4_reass_expire_node_idx = node->index;
+ return error;
+}
+
+VLIB_INIT_FUNCTION (ip4_reass_init_function);
+
+static uword
+ip4_reass_walk_expired (vlib_main_t * vm,
+ vlib_node_runtime_t * node, vlib_frame_t * f)
+{
+ ip4_reass_main_t *rm = &ip4_reass_main;
+ uword event_type, *event_data = 0;
+
+ while (true)
+ {
+ vlib_process_wait_for_event_or_clock (vm,
+ (f64) rm->expire_walk_interval_ms
+ / (f64) MSEC_PER_SEC);
+ event_type = vlib_process_get_events (vm, &event_data);
+
+ switch (event_type)
+ {
+ case ~0: /* no events => timeout */
+ /* nothing to do here */
+ break;
+ case IP4_EVENT_CONFIG_CHANGED:
+ break;
+ default:
+ clib_warning ("BUG: event type 0x%wx", event_type);
+ break;
+ }
+ f64 now = vlib_time_now (vm);
+
+ ip4_reass_t *reass;
+ u32 *vec_drop_timeout = NULL;
+ int *pool_indexes_to_free = NULL;
+
+ int index;
+ /* *INDENT-OFF* */
+ pool_foreach_index (index, rm->pool, ({
+ reass = pool_elt_at_index (rm->pool, index);
+ if (now > reass->last_heard + rm->timeout)
+ {
+ vec_add1 (pool_indexes_to_free, index);
+ }
+ }));
+ /* *INDENT-ON* */
+ int *i;
+ /* *INDENT-OFF* */
+ vec_foreach (i, pool_indexes_to_free)
+ {
+ ip4_reass_t *reass = pool_elt_at_index (rm->pool, i[0]);
+ ip4_reass_on_timeout (vm, rm, reass, &vec_drop_timeout);
+ ip4_reass_free (rm, reass);
+ }
+ /* *INDENT-ON* */
+
+ while (vec_len (vec_drop_timeout) > 0)
+ {
+ vlib_frame_t *f = vlib_get_frame_to_node (vm, rm->ip4_drop_idx);
+ u32 *to_next = vlib_frame_vector_args (f);
+ u32 n_left_to_next = VLIB_FRAME_SIZE - f->n_vectors;
+ u32 n_trace = 0;
+ while (vec_len (vec_drop_timeout) > 0 && n_left_to_next > 0)
+ {
+ u32 bi = vec_pop (vec_drop_timeout);
+ vlib_buffer_t *b = vlib_get_buffer (vm, bi);
+ if (PREDICT_FALSE (b->flags & VLIB_BUFFER_IS_TRACED))
+ {
+ if (pool_is_free_index (vm->trace_main.trace_buffer_pool,
+ b->trace_index))
+ {
+ /* the trace is gone, don't trace this buffer anymore */
+ b->flags &= ~VLIB_BUFFER_IS_TRACED;
+ }
+ else
+ {
+ ++n_trace;
+ }
+ }
+ b->error = node->errors[IP4_ERROR_REASS_TIMEOUT];
+ to_next[0] = bi;
+ ++f->n_vectors;
+ to_next += 1;
+ n_left_to_next -= 1;
+ IP4_REASS_DEBUG_BUFFER (bi, enqueue_drop_timeout_walk);
+ ASSERT (rm->buffers_n > 0);
+ --rm->buffers_n;
+ }
+ if (PREDICT_FALSE (n_trace > 0))
+ {
+ f->flags |= VLIB_FRAME_TRACE;
+ }
+ vlib_put_frame_to_node (vm, rm->ip4_drop_idx, f);
+ }
+
+ vec_free (pool_indexes_to_free);
+ vec_free (vec_drop_timeout);
+ if (event_data)
+ {
+ _vec_len (event_data) = 0;
+ }
+ }
+
+ return 0;
+}
+
+static vlib_node_registration_t ip4_reass_expire_node;
+
+/* *INDENT-OFF* */
+VLIB_REGISTER_NODE (ip4_reass_expire_node, static) = {
+ .function = ip4_reass_walk_expired,
+ .type = VLIB_NODE_TYPE_PROCESS,
+ .name = "ip4-reassembly-expire-walk",
+ .format_trace = format_ip4_reass_trace,
+ .n_errors = ARRAY_LEN (ip4_reassembly_error_strings),
+ .error_strings = ip4_reassembly_error_strings,
+
+};
+/* *INDENT-ON* */
+
+static u8 *
+format_ip4_reass_key (u8 * s, va_list * args)
+{
+ ip4_reass_key_t *key = va_arg (*args, ip4_reass_key_t *);
+ s = format (s, "xx_id: %u, src: %U, dst: %U, frag_id: %u, proto: %u",
+ key->xx_id, format_ip4_address, &key->src, format_ip4_address,
+ &key->dst, clib_net_to_host_u16 (key->frag_id), key->proto);
+ return s;
+}
+
+static u8 *
+format_ip4_reass (u8 * s, va_list * args)
+{
+ vlib_main_t *vm = va_arg (*args, vlib_main_t *);
+ ip4_reass_t *reass = va_arg (*args, ip4_reass_t *);
+
+ s = format (s, "ID: %u, key: %U\n first_bi: %u, data_len: %u, "
+ "last_packet_octet: %u, trace_op_counter: %u\n",
+ reass->id, format_ip4_reass_key, &reass->key, reass->first_bi,
+ reass->data_len, reass->last_packet_octet,
+ reass->trace_op_counter);
+ u32 bi = reass->first_bi;
+ u32 counter = 0;
+ while (~0 != bi)
+ {
+ vlib_buffer_t *b = vlib_get_buffer (vm, bi);
+ vnet_buffer_opaque_t *vnb = vnet_buffer (b);
+ s = format (s, " #%03u: range: [%u, %u], bi: %u, off: %d, len: %u, "
+ "fragment[%u, %u]\n",
+ counter, vnb->ip.reass.range_first,
+ vnb->ip.reass.range_last, bi,
+ ip4_reass_buffer_get_data_offset_no_check (b),
+ ip4_reass_buffer_get_data_len_no_check (b),
+ vnb->ip.reass.fragment_first, vnb->ip.reass.fragment_last);
+ if (b->flags & VLIB_BUFFER_NEXT_PRESENT)
+ {
+ bi = b->next_buffer;
+ }
+ else
+ {
+ bi = ~0;
+ }
+ }
+ return s;
+}
+
+static clib_error_t *
+show_ip4_reass (vlib_main_t * vm, unformat_input_t * input,
+ CLIB_UNUSED (vlib_cli_command_t * lmd))
+{
+ ip4_reass_main_t *rm = &ip4_reass_main;
+
+ vlib_cli_output (vm, "---------------------");
+ vlib_cli_output (vm, "IP4 reassembly status");
+ vlib_cli_output (vm, "---------------------");
+ if (unformat (input, "details"))
+ {
+ ip4_reass_t *reass;
+ /* *INDENT-OFF* */
+ pool_foreach (reass, rm->pool, {
+ vlib_cli_output (vm, "%U", format_ip4_reass, vm, reass);
+ });
+ /* *INDENT-ON* */
+ }
+ vlib_cli_output (vm, "---------------------");
+ vlib_cli_output (vm, "Current IP4 reassemblies count: %lu\n", rm->reass_n);
+ vlib_cli_output (vm,
+ "Maximum configured concurrent IP4 reassemblies: %lu\n",
+ (long unsigned) rm->max_reass_n);
+ vlib_cli_output (vm, "Buffers in use: %lu\n",
+ (long unsigned) rm->buffers_n);
+ return 0;
+}
+
+/* *INDENT-OFF* */
+VLIB_CLI_COMMAND (show_ip4_reassembly_cmd, static) = {
+ .path = "show ip4-reassembly",
+ .short_help = "show ip4-reassembly [details]",
+ .function = show_ip4_reass,
+};
+/* *INDENT-ON* */
+
+/*
+ * fd.io coding-style-patch-verification: ON
+ *
+ * Local Variables:
+ * eval: (c-set-style "gnu")
+ * End:
+ */
diff --git a/src/vnet/ip/ip4_reassembly.h b/src/vnet/ip/ip4_reassembly.h
new file mode 100644
index 00000000000..a48c07ef061
--- /dev/null
+++ b/src/vnet/ip/ip4_reassembly.h
@@ -0,0 +1,49 @@
+/*
+ * 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 IPv4 Reassembly.
+ *
+ * This file contains the source code for IPv4 reassembly.
+ */
+
+#ifndef __included_ip4_reassembly_h__
+#define __included_ip4_reassembly_h__
+
+#include <vnet/api_errno.h>
+#include <vnet/vnet.h>
+
+/**
+ * @brief set ip4 reassembly configuration
+ */
+vnet_api_error_t ip4_reass_set (u32 timeout_ms, u32 max_reassemblies,
+ u32 expire_walk_interval_ms);
+
+/**
+ * @brief get ip4 reassembly configuration
+ */
+vnet_api_error_t ip4_reass_get (u32 * timeout_ms, u32 * max_reassemblies,
+ u32 * expire_walk_interval_ms);
+
+#endif /* __included_ip4_reassembly_h */
+
+/*
+ * fd.io coding-style-patch-verification: ON
+ *
+ * Local Variables:
+ * eval: (c-set-style "gnu")
+ * End:
+ */
diff --git a/src/vnet/ip/ip6_error.h b/src/vnet/ip/ip6_error.h
index 7bb4b77b5a3..19f20625a10 100644
--- a/src/vnet/ip/ip6_error.h
+++ b/src/vnet/ip/ip6_error.h
@@ -71,7 +71,14 @@
\
/* Erros singalled by ip6-inacl */ \
_ (INACL_TABLE_MISS, "input ACL table-miss drops") \
- _ (INACL_SESSION_DENY, "input ACL session deny drops")
+ _ (INACL_SESSION_DENY, "input ACL session deny drops") \
+ \
+ /* Errors signalled by ip6-reassembly */ \
+ _ (REASS_MISSING_UPPER, "missing-upper layer drops") \
+ _ (REASS_DUPLICATE_FRAGMENT, "duplicate fragments") \
+ _ (REASS_OVERLAPPING_FRAGMENT, "overlapping fragments") \
+ _ (REASS_LIMIT_REACHED, "drops due to concurrent reassemblies limit") \
+ _ (REASS_TIMEOUT, "fragments dropped due to reassembly timeout")
typedef enum
{
diff --git a/src/vnet/ip/ip6_forward.c b/src/vnet/ip/ip6_forward.c
index c1c9ec0b586..5b06ca7609d 100644
--- a/src/vnet/ip/ip6_forward.c
+++ b/src/vnet/ip/ip6_forward.c
@@ -1578,6 +1578,7 @@ ip6_local_inline (vlib_main_t * vm, vlib_node_runtime_t * node,
next0 = lm->local_next_by_ip_protocol[ip0->protocol];
next0 =
error0 != IP6_ERROR_UNKNOWN_PROTOCOL ? IP_LOCAL_NEXT_DROP : next0;
+
p0->error = error_node->errors[error0];
if (head_of_feature_arc)
@@ -1619,6 +1620,7 @@ VLIB_REGISTER_NODE (ip6_local_node, static) =
[IP_LOCAL_NEXT_PUNT] = "ip6-punt",
[IP_LOCAL_NEXT_UDP_LOOKUP] = "ip6-udp-lookup",
[IP_LOCAL_NEXT_ICMP] = "ip6-icmp-input",
+ [IP_LOCAL_NEXT_REASSEMBLY] = "ip6-reassembly",
},
};
/* *INDENT-ON* */
diff --git a/src/vnet/ip/ip6_input.h b/src/vnet/ip/ip6_input.h
index 4c0d78459b1..fe993caa889 100644
--- a/src/vnet/ip/ip6_input.h
+++ b/src/vnet/ip/ip6_input.h
@@ -41,6 +41,7 @@
#define included_ip6_input_h
#include <vnet/ip/ip.h>
+#include <vnet/ip/icmp6.h>
extern char *ip6_error_strings[];
diff --git a/src/vnet/ip/ip6_packet.h b/src/vnet/ip/ip6_packet.h
index c0c745e27fe..6b724370e76 100644
--- a/src/vnet/ip/ip6_packet.h
+++ b/src/vnet/ip/ip6_packet.h
@@ -519,6 +519,9 @@ typedef CLIB_PACKED (struct {
#define ip6_frag_hdr_offset(hdr) \
(clib_net_to_host_u16((hdr)->fragment_offset_and_more) >> 3)
+#define ip6_frag_hdr_offset_bytes(hdr) \
+ (8 * ip6_frag_hdr_offset(hdr))
+
#define ip6_frag_hdr_more(hdr) \
(clib_net_to_host_u16((hdr)->fragment_offset_and_more) & 0x1)
diff --git a/src/vnet/ip/ip6_reassembly.c b/src/vnet/ip/ip6_reassembly.c
new file mode 100644
index 00000000000..d448032b913
--- /dev/null
+++ b/src/vnet/ip/ip6_reassembly.c
@@ -0,0 +1,1347 @@
+/*
+ * 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 IPv6 Reassembly.
+ *
+ * This file contains the source code for IPv6 reassembly.
+ */
+
+#include <vppinfra/vec.h>
+#include <vnet/vnet.h>
+#include <vnet/ip/ip.h>
+#include <vppinfra/bihash_48_8.h>
+#include <vnet/ip/ip6_reassembly.h>
+
+#define MSEC_PER_SEC 1000
+#define IP6_REASS_TIMEOUT_DEFAULT_MS 100
+#define IP6_REASS_EXPIRE_WALK_INTERVAL_DEFAULT_MS 10000 // 10 seconds default
+#define IP6_REASS_MAX_REASSEMBLIES_DEAFULT 1024
+#define IP6_REASS_HT_LOAD_FACTOR (0.75)
+
+static vlib_node_registration_t ip6_reass_node;
+
+typedef struct
+{
+ union
+ {
+ struct
+ {
+ ip6_address_t src;
+ ip6_address_t dst;
+ // align by making this 4 octets even though its a 2 octets field
+ u32 xx_id;
+ // align by making this 4 octets even though its a 2 octets field
+ u32 frag_id;
+ // align by making this 4 octets even though its a 1 octet field
+ u32 proto;
+ u32 unused;
+ };
+ u64 as_u64[6];
+ };
+} ip6_reass_key_t;
+
+always_inline u32
+ip6_reass_buffer_get_data_offset_no_check (vlib_buffer_t * b)
+{
+ vnet_buffer_opaque_t *vnb = vnet_buffer (b);
+ return vnb->ip.reass.range_first - vnb->ip.reass.fragment_first;
+}
+
+always_inline u32
+ip6_reass_buffer_get_data_offset (vlib_buffer_t * b)
+{
+ vnet_buffer_opaque_t *vnb = vnet_buffer (b);
+ ASSERT (vnb->ip.reass.range_first >= vnb->ip.reass.fragment_first);
+ return ip6_reass_buffer_get_data_offset_no_check (b);
+}
+
+always_inline u16
+ip6_reass_buffer_get_data_len_no_check (vlib_buffer_t * b)
+{
+ vnet_buffer_opaque_t *vnb = vnet_buffer (b);
+ return clib_min (vnb->ip.reass.range_last, vnb->ip.reass.fragment_last) -
+ (vnb->ip.reass.fragment_first + ip6_reass_buffer_get_data_offset (b)) + 1;
+}
+
+always_inline u16
+ip6_reass_buffer_get_data_len (vlib_buffer_t * b)
+{
+ vnet_buffer_opaque_t *vnb = vnet_buffer (b);
+ ASSERT (vnb->ip.reass.range_last > vnb->ip.reass.fragment_first);
+ return ip6_reass_buffer_get_data_len_no_check (b);
+}
+
+typedef struct
+{
+ // hash table key
+ ip6_reass_key_t key;
+ // time when first packet was received
+ f64 first_heard;
+ // time when last packet was received
+ f64 last_heard;
+ // internal id of this reassembly
+ u32 id;
+ // buffer index of first buffer in this reassembly context
+ u32 first_bi;
+ // last octet of packet, ~0 until fragment without more_fragments arrives
+ u32 last_packet_octet;
+ // length of data collected so far
+ u32 data_len;
+ // trace operation counter
+ u32 trace_op_counter;
+} ip6_reass_t;
+
+typedef struct
+{
+ // IPv6 config
+ u32 timeout_ms;
+ f64 timeout;
+ u32 expire_walk_interval_ms;
+ u32 max_reass_n;
+
+ // IPv6 runtime
+ ip6_reass_t *pool;
+ clib_bihash_48_8_t hash;
+ u32 reass_n;
+ u32 id_counter;
+ u32 buffers_n;
+
+ // convenience
+ vlib_main_t *vlib_main;
+ vnet_main_t *vnet_main;
+
+ // node index of ip6-drop node
+ u32 ip6_drop_idx;
+ u32 ip6_icmp_error_idx;
+ u32 ip6_reass_expire_node_idx;
+
+} ip6_reass_main_t;
+
+ip6_reass_main_t ip6_reass_main;
+
+typedef enum
+{
+ IP6_REASSEMBLY_NEXT_INPUT,
+ IP6_REASSEMBLY_NEXT_DROP,
+ IP6_REASSEMBLY_NEXT_ICMP_ERROR,
+ IP6_REASSEMBLY_N_NEXT,
+} ip6_reass_next_t;
+
+typedef enum
+{
+ RANGE_NEW,
+ RANGE_OVERLAP,
+ ICMP_ERROR_RT_EXCEEDED,
+ ICMP_ERROR_FL_TOO_BIG,
+ ICMP_ERROR_FL_NOT_MULT_8,
+ FINALIZE,
+} ip6_reass_trace_operation_e;
+
+typedef struct
+{
+ u16 range_first;
+ u16 range_last;
+ u32 range_bi;
+ i32 data_offset;
+ u32 data_len;
+ u32 first_bi;
+} ip6_reass_range_trace_t;
+
+typedef struct
+{
+ ip6_reass_trace_operation_e action;
+ u32 pool_index;
+ u32 reass_id;
+ ip6_reass_range_trace_t trace_range;
+ u32 size_diff;
+ u32 op_id;
+ u32 fragment_first;
+ u32 fragment_last;
+ u32 total_data_len;
+} ip6_reass_trace_t;
+
+static void
+ip6_reass_trace_details (vlib_main_t * vm, u32 bi,
+ ip6_reass_range_trace_t * trace)
+{
+ vlib_buffer_t *b = vlib_get_buffer (vm, bi);
+ vnet_buffer_opaque_t *vnb = vnet_buffer (b);
+ trace->range_first = vnb->ip.reass.range_first;
+ trace->range_last = vnb->ip.reass.range_last;
+ trace->data_offset = ip6_reass_buffer_get_data_offset_no_check (b);
+ trace->data_len = ip6_reass_buffer_get_data_len_no_check (b);
+ trace->range_bi = bi;
+}
+
+u8 *
+format_ip6_reass_range_trace (u8 * s, va_list * args)
+{
+ ip6_reass_range_trace_t *trace = va_arg (*args, ip6_reass_range_trace_t *);
+ s = format (s, "range: [%u, %u], off %d, len %u, bi %u", trace->range_first,
+ trace->range_last, trace->data_offset, trace->data_len,
+ trace->range_bi);
+ return s;
+}
+
+u8 *
+format_ip6_reass_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 *);
+ ip6_reass_trace_t *t = va_arg (*args, ip6_reass_trace_t *);
+ s = format (s, "reass id: %u, op id: %u ", t->reass_id, t->op_id);
+ u32 indent = format_get_indent (s);
+ s = format (s, "first bi: %u, data len: %u, ip/fragment[%u, %u]",
+ t->trace_range.first_bi, t->total_data_len, t->fragment_first,
+ t->fragment_last);
+ switch (t->action)
+ {
+ case RANGE_NEW:
+ s = format (s, "\n%Unew %U", format_white_space, indent,
+ format_ip6_reass_range_trace, &t->trace_range);
+ break;
+ case RANGE_OVERLAP:
+ s = format (s, "\n%Uoverlap %U", format_white_space, indent,
+ format_ip6_reass_range_trace, &t->trace_range);
+ break;
+ case ICMP_ERROR_FL_TOO_BIG:
+ s = format (s, "\n%Uicmp-error - frag_len > 65535 %U",
+ format_white_space, indent, format_ip6_reass_range_trace,
+ &t->trace_range);
+ break;
+ case ICMP_ERROR_FL_NOT_MULT_8:
+ s = format (s, "\n%Uicmp-error - frag_len mod 8 != 0 %U",
+ format_white_space, indent, format_ip6_reass_range_trace,
+ &t->trace_range);
+ break;
+ case ICMP_ERROR_RT_EXCEEDED:
+ s = format (s, "\n%Uicmp-error - reassembly time exceeded",
+ format_white_space, indent);
+ break;
+ case FINALIZE:
+ s = format (s, "\n%Ufinalize reassembly", format_white_space, indent);
+ break;
+ }
+ return s;
+}
+
+static void
+ip6_reass_add_trace (vlib_main_t * vm, vlib_node_runtime_t * node,
+ ip6_reass_main_t * rm, ip6_reass_t * reass,
+ u32 bi, ip6_reass_trace_operation_e action,
+ u32 size_diff)
+{
+ vlib_buffer_t *b = vlib_get_buffer (vm, bi);
+ vnet_buffer_opaque_t *vnb = vnet_buffer (b);
+ ip6_reass_trace_t *t = vlib_add_trace (vm, node, b, sizeof (t[0]));
+ t->pool_index = reass - rm->pool;
+ t->reass_id = reass->id;
+ t->action = action;
+ ip6_reass_trace_details (vm, bi, &t->trace_range);
+ t->size_diff = size_diff;
+ t->op_id = reass->trace_op_counter;
+ ++reass->trace_op_counter;
+ t->fragment_first = vnb->ip.reass.fragment_first;
+ t->fragment_last = vnb->ip.reass.fragment_last;
+ t->trace_range.first_bi = reass->first_bi;
+ t->total_data_len = reass->data_len;
+#if 0
+ static u8 *s = NULL;
+ s = format (s, "%U", format_ip6_reass_trace, NULL, NULL, t);
+ printf ("%.*s\n", vec_len (s), s);
+ fflush (stdout);
+ vec_reset_length (s);
+#endif
+}
+
+static void
+ip6_reass_free (ip6_reass_main_t * rm, ip6_reass_t * reass)
+{
+ clib_bihash_kv_48_8_t kv;
+ kv.key[0] = reass->key.as_u64[0];
+ kv.key[1] = reass->key.as_u64[1];
+ kv.key[2] = reass->key.as_u64[2];
+ kv.key[3] = reass->key.as_u64[3];
+ kv.key[4] = reass->key.as_u64[4];
+ kv.key[5] = reass->key.as_u64[5];
+ clib_bihash_add_del_48_8 (&rm->hash, &kv, 0);
+ pool_put (rm->pool, reass);
+ --rm->reass_n;
+}
+
+static void
+ip6_reass_drop_all (vlib_main_t * vm, ip6_reass_main_t * rm,
+ ip6_reass_t * reass, u32 ** vec_drop_bi)
+{
+ u32 range_bi = reass->first_bi;
+ vlib_buffer_t *range_b;
+ vnet_buffer_opaque_t *range_vnb;
+ while (~0 != range_bi)
+ {
+ range_b = vlib_get_buffer (vm, range_bi);
+ range_vnb = vnet_buffer (range_b);
+ u32 bi = range_bi;
+ while (~0 != bi)
+ {
+ vec_add1 (*vec_drop_bi, bi);
+ vlib_buffer_t *b = vlib_get_buffer (vm, bi);
+ if (b->flags & VLIB_BUFFER_NEXT_PRESENT)
+ {
+ bi = b->next_buffer;
+ b->flags &= ~VLIB_BUFFER_NEXT_PRESENT;
+ }
+ else
+ {
+ bi = ~0;
+ }
+ }
+ range_bi = range_vnb->ip.reass.next_range_bi;
+ }
+}
+
+static void
+ip6_reass_on_timeout (vlib_main_t * vm, vlib_node_runtime_t * node,
+ ip6_reass_main_t * rm, ip6_reass_t * reass,
+ u32 * icmp_bi, u32 ** vec_timeout)
+{
+ if (~0 == reass->first_bi)
+ {
+ return;
+ }
+ vlib_buffer_t *b = vlib_get_buffer (vm, reass->first_bi);
+ if (0 == vnet_buffer (b)->ip.reass.fragment_first)
+ {
+ *icmp_bi = reass->first_bi;
+ if (PREDICT_FALSE (b->flags & VLIB_BUFFER_IS_TRACED))
+ {
+ ip6_reass_add_trace (vm, node, rm, reass, reass->first_bi,
+ ICMP_ERROR_RT_EXCEEDED, 0);
+ }
+ // fragment with offset zero received - send icmp message back
+ if (b->flags & VLIB_BUFFER_NEXT_PRESENT)
+ {
+ // separate first buffer from chain and steer it towards icmp node
+ b->flags &= ~VLIB_BUFFER_NEXT_PRESENT;
+ reass->first_bi = b->next_buffer;
+ }
+ else
+ {
+ reass->first_bi = vnet_buffer (b)->ip.reass.next_range_bi;
+ }
+ icmp6_error_set_vnet_buffer (b, ICMP6_time_exceeded,
+ ICMP6_time_exceeded_fragment_reassembly_time_exceeded,
+ 0);
+ }
+ ip6_reass_drop_all (vm, rm, reass, vec_timeout);
+}
+
+static ip6_reass_t *
+ip6_reass_find_or_create (vlib_main_t * vm,
+ vlib_node_runtime_t * node,
+ ip6_reass_main_t * rm,
+ ip6_reass_key_t * k, u32 * icmp_bi,
+ u32 ** vec_timeout)
+{
+ ip6_reass_t *reass = NULL;
+ f64 now = vlib_time_now (rm->vlib_main);
+ clib_bihash_kv_48_8_t kv, value;
+ kv.key[0] = k->as_u64[0];
+ kv.key[1] = k->as_u64[1];
+ kv.key[2] = k->as_u64[2];
+ kv.key[3] = k->as_u64[3];
+ kv.key[4] = k->as_u64[4];
+ kv.key[5] = k->as_u64[5];
+
+ if (!clib_bihash_search_48_8 (&rm->hash, &kv, &value))
+ {
+ reass = pool_elt_at_index (rm->pool, value.value);
+ if (now > reass->last_heard + rm->timeout)
+ {
+ ip6_reass_on_timeout (vm, node, rm, reass, icmp_bi, vec_timeout);
+ ip6_reass_free (rm, reass);
+ reass = NULL;
+ }
+ }
+
+ if (reass)
+ {
+ reass->last_heard = now;
+ return reass;
+ }
+
+ if (rm->reass_n >= rm->max_reass_n)
+ {
+ reass = NULL;
+ return reass;
+ }
+ else
+ {
+ pool_get (rm->pool, reass);
+ memset (reass, 0, sizeof (*reass));
+ reass->id = rm->id_counter;
+ ++rm->id_counter;
+ reass->first_bi = ~0;
+ reass->last_packet_octet = ~0;
+ reass->data_len = 0;
+ ++rm->reass_n;
+ }
+
+ reass->key.as_u64[0] = kv.key[0] = k->as_u64[0];
+ reass->key.as_u64[1] = kv.key[1] = k->as_u64[1];
+ reass->key.as_u64[2] = kv.key[2] = k->as_u64[2];
+ reass->key.as_u64[3] = kv.key[3] = k->as_u64[3];
+ reass->key.as_u64[4] = kv.key[4] = k->as_u64[4];
+ reass->key.as_u64[5] = kv.key[5] = k->as_u64[5];
+ kv.value = reass - rm->pool;
+ reass->last_heard = now;
+
+ if (clib_bihash_add_del_48_8 (&rm->hash, &kv, 1))
+ {
+ ip6_reass_free (rm, reass);
+ reass = NULL;
+ }
+
+ return reass;
+}
+
+void
+ip6_reass_finalize (vlib_main_t * vm, vlib_node_runtime_t * node,
+ ip6_reass_main_t * rm, ip6_reass_t * reass, u32 * bi0,
+ u32 * next0, vlib_error_t * error0, u32 next_input,
+ u32 ** vec_drop_compress)
+{
+ ASSERT (~0 != reass->first_bi);
+ *bi0 = reass->first_bi;
+ *next0 = next_input;
+ *error0 = IP6_ERROR_NONE;
+ ip6_frag_hdr_t *frag_hdr;
+ vlib_buffer_t *last_b = NULL;
+ u32 sub_chain_bi = reass->first_bi;
+ u32 total_length = 0;
+ u32 buf_cnt = 0;
+ u32 dropped_cnt = 0;
+ do
+ {
+ u32 tmp_bi = sub_chain_bi;
+ vlib_buffer_t *tmp = vlib_get_buffer (vm, tmp_bi);
+ u32 data_len = ip6_reass_buffer_get_data_len (tmp);
+ u32 trim_front = vnet_buffer (tmp)->ip.reass.ip6_frag_hdr_offset +
+ sizeof (*frag_hdr) + ip6_reass_buffer_get_data_offset (tmp);
+ u32 trim_end =
+ vlib_buffer_length_in_chain (vm, tmp) - trim_front - data_len;
+ if (tmp_bi == reass->first_bi)
+ {
+ /* first buffer - keep ip6 header */
+ ASSERT (0 == ip6_reass_buffer_get_data_offset (tmp));
+ trim_front = 0;
+ trim_end = vlib_buffer_length_in_chain (vm, tmp) - data_len -
+ (vnet_buffer (tmp)->ip.reass.ip6_frag_hdr_offset +
+ sizeof (*frag_hdr));
+ }
+ u32 keep_data =
+ vlib_buffer_length_in_chain (vm, tmp) - trim_front - trim_end;
+ while (1)
+ {
+ ++buf_cnt;
+ if (trim_front)
+ {
+ if (trim_front > tmp->current_length)
+ {
+ /* drop whole buffer */
+ vec_add1 (*vec_drop_compress, tmp_bi);
+ ++dropped_cnt;
+ trim_front -= tmp->current_length;
+ ASSERT (tmp->flags & VLIB_BUFFER_NEXT_PRESENT);
+ tmp->flags &= ~VLIB_BUFFER_NEXT_PRESENT;
+ tmp_bi = tmp->next_buffer;
+ tmp = vlib_get_buffer (vm, tmp_bi);
+ continue;
+ }
+ else
+ {
+ vlib_buffer_advance (tmp, trim_front);
+ trim_front = 0;
+ }
+ }
+ if (keep_data)
+ {
+ if (last_b)
+ {
+ last_b->flags |= VLIB_BUFFER_NEXT_PRESENT;
+ last_b->next_buffer = tmp_bi;
+ }
+ last_b = tmp;
+ if (keep_data <= tmp->current_length)
+ {
+ tmp->current_length = keep_data;
+ keep_data = 0;
+ }
+ else
+ {
+ keep_data -= tmp->current_length;
+ ASSERT (tmp->flags & VLIB_BUFFER_NEXT_PRESENT);
+ }
+ total_length += tmp->current_length;
+ }
+ else
+ {
+ vec_add1 (*vec_drop_compress, tmp_bi);
+ ASSERT (reass->first_bi != tmp_bi);
+ ++dropped_cnt;
+ }
+ if (tmp->flags & VLIB_BUFFER_NEXT_PRESENT)
+ {
+ tmp_bi = tmp->next_buffer;
+ tmp = vlib_get_buffer (vm, tmp->next_buffer);
+ }
+ else
+ {
+ break;
+ }
+ }
+ sub_chain_bi =
+ vnet_buffer (vlib_get_buffer (vm, sub_chain_bi))->ip.
+ reass.next_range_bi;
+ }
+ while (~0 != sub_chain_bi);
+ last_b->flags &= ~VLIB_BUFFER_NEXT_PRESENT;
+ vlib_buffer_t *first_b = vlib_get_buffer (vm, reass->first_bi);
+ ASSERT (total_length >= first_b->current_length);
+ total_length -= first_b->current_length;
+ first_b->flags |= VLIB_BUFFER_TOTAL_LENGTH_VALID;
+ first_b->total_length_not_including_first_buffer = total_length;
+ // drop fragment header
+ vnet_buffer_opaque_t *first_b_vnb = vnet_buffer (first_b);
+ ip6_header_t *ip = vlib_buffer_get_current (first_b);
+ u16 ip6_frag_hdr_offset = first_b_vnb->ip.reass.ip6_frag_hdr_offset;
+ ip6_ext_header_t *prev_hdr;
+ ip6_ext_header_find_t (ip, prev_hdr, frag_hdr,
+ IP_PROTOCOL_IPV6_FRAGMENTATION);
+ if (prev_hdr)
+ {
+ prev_hdr->next_hdr = frag_hdr->next_hdr;
+ }
+ else
+ {
+ ip->protocol = frag_hdr->next_hdr;
+ }
+ ASSERT ((u8 *) frag_hdr - (u8 *) ip == ip6_frag_hdr_offset);
+ memmove (frag_hdr, (u8 *) frag_hdr + sizeof (*frag_hdr),
+ first_b->current_length - ip6_frag_hdr_offset -
+ sizeof (ip6_frag_hdr_t));
+ first_b->current_length -= sizeof (*frag_hdr);
+ ip->payload_length =
+ clib_host_to_net_u16 (total_length + first_b->current_length -
+ sizeof (*ip));
+ ip6_reass_free (rm, reass);
+ vlib_buffer_chain_compress (vm, first_b, vec_drop_compress);
+ if (PREDICT_FALSE (first_b->flags & VLIB_BUFFER_IS_TRACED))
+ {
+ ip6_reass_add_trace (vm, node, rm, reass, reass->first_bi, FINALIZE, 0);
+#if 0
+ // following code does a hexdump of packet fragments to stdout ...
+ do
+ {
+ u32 bi = reass->first_bi;
+ u8 *s = NULL;
+ while (~0 != bi)
+ {
+ vlib_buffer_t *b = vlib_get_buffer (vm, bi);
+ s = format (s, "%u: %U\n", bi, format_hexdump,
+ vlib_buffer_get_current (b), b->current_length);
+ if (b->flags & VLIB_BUFFER_NEXT_PRESENT)
+ {
+ bi = b->next_buffer;
+ }
+ else
+ {
+ break;
+ }
+ }
+ printf ("%.*s\n", vec_len (s), s);
+ fflush (stdout);
+ vec_free (s);
+ }
+ while (0);
+#endif
+ }
+ reass = NULL;
+}
+
+static u32
+ip6_reass_get_buffer_chain_length (vlib_main_t * vm, vlib_buffer_t * b)
+{
+ u32 len = 0;
+ while (b)
+ {
+ ++len;
+ if (PREDICT_FALSE (b->flags & VLIB_BUFFER_NEXT_PRESENT))
+ {
+ b = vlib_get_buffer (vm, b->next_buffer);
+ }
+ else
+ {
+ break;
+ }
+ }
+ return len;
+}
+
+static void
+ip6_reass_insert_range_in_chain (vlib_main_t * vm,
+ ip6_reass_main_t * rm,
+ ip6_reass_t * reass,
+ u32 prev_range_bi, u32 new_next_bi)
+{
+
+ vlib_buffer_t *new_next_b = vlib_get_buffer (vm, new_next_bi);
+ vnet_buffer_opaque_t *new_next_vnb = vnet_buffer (new_next_b);
+ if (~0 != prev_range_bi)
+ {
+ vlib_buffer_t *prev_b = vlib_get_buffer (vm, prev_range_bi);
+ vnet_buffer_opaque_t *prev_vnb = vnet_buffer (prev_b);
+ new_next_vnb->ip.reass.next_range_bi = prev_vnb->ip.reass.next_range_bi;
+ prev_vnb->ip.reass.next_range_bi = new_next_bi;
+ }
+ else
+ {
+ if (~0 != reass->first_bi)
+ {
+ new_next_vnb->ip.reass.next_range_bi = reass->first_bi;
+ }
+ reass->first_bi = new_next_bi;
+ }
+ reass->data_len += ip6_reass_buffer_get_data_len (new_next_b);
+ rm->buffers_n += ip6_reass_get_buffer_chain_length (vm, new_next_b);
+}
+
+void
+ip6_reass_update (vlib_main_t * vm, vlib_node_runtime_t * node,
+ ip6_reass_main_t * rm, ip6_reass_t * reass, u32 * bi0,
+ u32 * next0, vlib_error_t * error0,
+ ip6_frag_hdr_t * frag_hdr, u32 ** vec_drop_overlap,
+ u32 ** vec_drop_compress, u32 next_input, u32 next_drop,
+ u32 next_icmp_error)
+{
+ int consumed = 0;
+ vlib_buffer_t *fb = vlib_get_buffer (vm, *bi0);
+ vnet_buffer_opaque_t *fvnb = vnet_buffer (fb);
+ fvnb->ip.reass.ip6_frag_hdr_offset =
+ (u8 *) frag_hdr - (u8 *) vlib_buffer_get_current (fb);
+ ip6_header_t *fip = vlib_buffer_get_current (fb);
+ ASSERT (fb->current_length > sizeof (*fip));
+ ASSERT (fvnb->ip.reass.ip6_frag_hdr_offset > 0 &&
+ fvnb->ip.reass.ip6_frag_hdr_offset < fb->current_length);
+ u32 fragment_first = fvnb->ip.reass.fragment_first =
+ ip6_frag_hdr_offset_bytes (frag_hdr);
+ u32 fragment_length =
+ vlib_buffer_length_in_chain (vm, fb) -
+ (fvnb->ip.reass.ip6_frag_hdr_offset + sizeof (*frag_hdr));
+ u32 fragment_last = fvnb->ip.reass.fragment_last =
+ fragment_first + fragment_length - 1;
+ int more_fragments = ip6_frag_hdr_more (frag_hdr);
+ u32 candidate_range_bi = reass->first_bi;
+ u32 prev_range_bi = ~0;
+ fvnb->ip.reass.range_first = fragment_first;
+ fvnb->ip.reass.range_last = fragment_last;
+ fvnb->ip.reass.next_range_bi = ~0;
+ if (more_fragments && 0 != fragment_length % 8)
+ {
+ if (PREDICT_FALSE (fb->flags & VLIB_BUFFER_IS_TRACED))
+ {
+ ip6_reass_add_trace (vm, node, rm, reass, *bi0,
+ ICMP_ERROR_FL_NOT_MULT_8, 0);
+ }
+ *next0 = next_icmp_error;
+ icmp6_error_set_vnet_buffer (fb, ICMP6_parameter_problem,
+ ICMP6_parameter_problem_erroneous_header_field,
+ (u8 *) & fip->payload_length - (u8 *) fip);
+ return;
+ }
+ if (fragment_first + fragment_length > 65535)
+ {
+ if (PREDICT_FALSE (fb->flags & VLIB_BUFFER_IS_TRACED))
+ {
+ ip6_reass_add_trace (vm, node, rm, reass, *bi0,
+ ICMP_ERROR_FL_TOO_BIG, 0);
+ }
+ *next0 = next_icmp_error;
+ ip6_header_t *ip0 = vlib_buffer_get_current (fb);
+ icmp6_error_set_vnet_buffer (fb, ICMP6_parameter_problem,
+ ICMP6_parameter_problem_erroneous_header_field,
+ (u8 *) & frag_hdr->fragment_offset_and_more
+ - (u8 *) ip0);
+ return;
+ }
+ if (!more_fragments)
+ {
+ reass->last_packet_octet = fragment_last;
+ }
+ if (~0 == reass->first_bi)
+ {
+ // starting a new reassembly
+ ip6_reass_insert_range_in_chain (vm, rm, reass, prev_range_bi, *bi0);
+ if (PREDICT_FALSE (fb->flags & VLIB_BUFFER_IS_TRACED))
+ {
+ ip6_reass_add_trace (vm, node, rm, reass, *bi0, RANGE_NEW, 0);
+ }
+ *bi0 = ~0;
+ return;
+ }
+ fvnb->ip.reass.estimated_mtu =
+ clib_min (clib_net_to_host_u16 (fip->payload_length),
+ fvnb->ip.reass.estimated_mtu);
+ while (~0 != candidate_range_bi)
+ {
+ vlib_buffer_t *candidate_b = vlib_get_buffer (vm, candidate_range_bi);
+ vnet_buffer_opaque_t *candidate_vnb = vnet_buffer (candidate_b);
+ if (fragment_first > candidate_vnb->ip.reass.range_last)
+ {
+ // this fragments starts after candidate range
+ prev_range_bi = candidate_range_bi;
+ candidate_range_bi = candidate_vnb->ip.reass.next_range_bi;
+ if (candidate_vnb->ip.reass.range_last < fragment_last &&
+ ~0 == candidate_range_bi)
+ {
+ // special case - this fragment falls beyond all known ranges
+ ip6_reass_insert_range_in_chain (vm, rm, reass, prev_range_bi,
+ *bi0);
+ consumed = 1;
+ break;
+ }
+ continue;
+ }
+ if (fragment_last < candidate_vnb->ip.reass.range_first)
+ {
+ // this fragment ends before candidate range without any overlap
+ ip6_reass_insert_range_in_chain (vm, rm, reass, prev_range_bi,
+ *bi0);
+ consumed = 1;
+ }
+ else if (fragment_first == candidate_vnb->ip.reass.range_first &&
+ fragment_last == candidate_vnb->ip.reass.range_last)
+ {
+ // duplicate fragment - ignore
+ }
+ else
+ {
+ // overlapping fragment - not allowed by RFC 8200
+ ip6_reass_drop_all (vm, rm, reass, vec_drop_overlap);
+ ip6_reass_free (rm, reass);
+ if (PREDICT_FALSE (fb->flags & VLIB_BUFFER_IS_TRACED))
+ {
+ ip6_reass_add_trace (vm, node, rm, reass, *bi0, RANGE_OVERLAP,
+ 0);
+ }
+ *next0 = next_drop;
+ *error0 = IP6_ERROR_REASS_OVERLAPPING_FRAGMENT;
+ }
+ break;
+ }
+ if (consumed)
+ {
+ if (PREDICT_FALSE (fb->flags & VLIB_BUFFER_IS_TRACED))
+ {
+ ip6_reass_add_trace (vm, node, rm, reass, *bi0, RANGE_NEW, 0);
+ }
+ }
+ if (~0 != reass->last_packet_octet &&
+ reass->data_len == reass->last_packet_octet + 1)
+ {
+ ip6_reass_finalize (vm, node, rm, reass, bi0, next0, error0, next_input,
+ vec_drop_compress);
+ }
+ else
+ {
+ if (consumed)
+ {
+ *bi0 = ~0;
+ }
+ else
+ {
+ *next0 = next_drop;
+ *error0 = IP6_ERROR_REASS_DUPLICATE_FRAGMENT;
+ }
+ }
+}
+
+always_inline uword
+ip6_reassembly (vlib_main_t * vm, vlib_node_runtime_t * node,
+ vlib_frame_t * frame)
+{
+ u32 *from = vlib_frame_vector_args (frame);
+ u32 n_left_from, n_left_to_next, *to_next, next_index;
+ ip6_reass_main_t *rm = &ip6_reass_main;
+
+ n_left_from = frame->n_vectors;
+ next_index = node->cached_next_index;
+ static u32 *vec_timeout = NULL; // indexes of buffers which timed out
+ static u32 *vec_drop_overlap = NULL; // indexes of buffers dropped due to overlap
+ static u32 *vec_drop_compress = NULL; // indexes of buffers dropped due to buffer compression
+ while (n_left_from > 0 || vec_len (vec_timeout) > 0 ||
+ vec_len (vec_drop_overlap) > 0)
+ {
+ vlib_get_next_frame (vm, node, next_index, to_next, n_left_to_next);
+
+ while (vec_len (vec_timeout) > 0 && n_left_to_next > 0)
+ {
+ u32 bi = vec_pop (vec_timeout);
+ vlib_buffer_t *b = vlib_get_buffer (vm, bi);
+ b->error = node->errors[IP6_ERROR_REASS_TIMEOUT];
+ to_next[0] = bi;
+ to_next += 1;
+ n_left_to_next -= 1;
+ vlib_validate_buffer_enqueue_x1 (vm, node, next_index, to_next,
+ n_left_to_next, bi,
+ IP6_REASSEMBLY_NEXT_DROP);
+ ASSERT (rm->buffers_n > 0);
+ --rm->buffers_n;
+ }
+
+ while (vec_len (vec_drop_overlap) > 0 && n_left_to_next > 0)
+ {
+ u32 bi = vec_pop (vec_drop_overlap);
+ vlib_buffer_t *b = vlib_get_buffer (vm, bi);
+ b->error = node->errors[IP6_ERROR_REASS_OVERLAPPING_FRAGMENT];
+ to_next[0] = bi;
+ to_next += 1;
+ n_left_to_next -= 1;
+ vlib_validate_buffer_enqueue_x1 (vm, node, next_index, to_next,
+ n_left_to_next, bi,
+ IP6_REASSEMBLY_NEXT_DROP);
+ ASSERT (rm->buffers_n > 0);
+ --rm->buffers_n;
+ }
+
+ while (vec_len (vec_drop_compress) > 0 && n_left_to_next > 0)
+ {
+ u32 bi = vec_pop (vec_drop_compress);
+ vlib_buffer_t *b = vlib_get_buffer (vm, bi);
+ b->error = node->errors[IP6_ERROR_NONE];
+ to_next[0] = bi;
+ to_next += 1;
+ n_left_to_next -= 1;
+ vlib_validate_buffer_enqueue_x1 (vm, node, next_index, to_next,
+ n_left_to_next, bi,
+ IP6_REASSEMBLY_NEXT_DROP);
+ ASSERT (rm->buffers_n > 0);
+ --rm->buffers_n;
+ }
+
+ while (n_left_from > 0 && n_left_to_next > 0)
+ {
+ u32 bi0;
+ vlib_buffer_t *b0;
+ u32 next0; //, error0;
+
+ bi0 = from[0];
+ b0 = vlib_get_buffer (vm, bi0);
+
+ ip6_header_t *ip0 = vlib_buffer_get_current (b0);
+ ip6_frag_hdr_t *frag_hdr;
+ ip6_ext_header_t *prev_hdr;
+ ip6_ext_header_find_t (ip0, prev_hdr, frag_hdr,
+ IP_PROTOCOL_IPV6_FRAGMENTATION);
+ if (0 == ip6_frag_hdr_offset (frag_hdr))
+ {
+ // first fragment - verify upper-layer is present
+ ip6_ext_header_t *tmp = (ip6_ext_header_t *) frag_hdr;
+ while (ip6_ext_hdr (tmp->next_hdr))
+ {
+ tmp = ip6_ext_next_header (tmp);
+ }
+ if (IP_PROTOCOL_IP6_NONXT == tmp->next_hdr)
+ {
+ icmp6_error_set_vnet_buffer (b0, ICMP6_parameter_problem,
+ ICMP6_parameter_problem_first_fragment_has_incomplete_header_chain,
+ 0);
+ b0->error = node->errors[IP6_ERROR_REASS_MISSING_UPPER];
+
+ to_next[0] = bi0;
+ to_next += 1;
+ n_left_to_next -= 1;
+ vlib_validate_buffer_enqueue_x1 (vm, node, next_index,
+ to_next, n_left_to_next,
+ bi0,
+ IP6_REASSEMBLY_NEXT_ICMP_ERROR);
+ goto next;
+ }
+ }
+ vnet_buffer (b0)->ip.reass.ip6_frag_hdr_offset =
+ (u8 *) frag_hdr - (u8 *) ip0;
+
+ ip6_reass_key_t k;
+ k.src.as_u64[0] = ip0->src_address.as_u64[0];
+ k.src.as_u64[1] = ip0->src_address.as_u64[1];
+ k.dst.as_u64[0] = ip0->dst_address.as_u64[0];
+ k.dst.as_u64[1] = ip0->dst_address.as_u64[1];
+ k.xx_id = vnet_buffer (b0)->sw_if_index[VLIB_RX];
+ k.frag_id = frag_hdr->identification;
+ k.proto = ip0->protocol;
+ k.unused = 0;
+ u32 icmp_bi = ~0;
+ ip6_reass_t *reass =
+ ip6_reass_find_or_create (vm, node, rm, &k, &icmp_bi,
+ &vec_timeout);
+
+ u32 error0 = IP6_ERROR_NONE;
+ if (reass)
+ {
+ ip6_reass_update (vm, node, rm, reass, &bi0, &next0, &error0,
+ frag_hdr, &vec_drop_overlap,
+ &vec_drop_compress, IP6_REASSEMBLY_NEXT_INPUT,
+ IP6_REASSEMBLY_NEXT_DROP,
+ IP6_REASSEMBLY_NEXT_ICMP_ERROR);
+ }
+ else
+ {
+ next0 = IP6_REASSEMBLY_NEXT_DROP;
+ error0 = IP6_ERROR_REASS_LIMIT_REACHED;
+ }
+
+ b0->error = node->errors[error0];
+
+ if (~0 != bi0)
+ {
+ to_next[0] = bi0;
+ to_next += 1;
+ n_left_to_next -= 1;
+ vlib_validate_buffer_enqueue_x1 (vm, node, next_index, to_next,
+ n_left_to_next, bi0, next0);
+ }
+
+ if (~0 != icmp_bi)
+ {
+ next0 = IP6_REASSEMBLY_NEXT_ICMP_ERROR;
+ to_next[0] = icmp_bi;
+ to_next += 1;
+ n_left_to_next -= 1;
+ vlib_validate_buffer_enqueue_x1 (vm, node, next_index, to_next,
+ n_left_to_next, icmp_bi,
+ next0);
+ }
+ next:
+ from += 1;
+ n_left_from -= 1;
+ }
+
+ vlib_put_next_frame (vm, node, next_index, n_left_to_next);
+ }
+
+ return frame->n_vectors;
+}
+
+static char *ip6_reassembly_error_strings[] = {
+#define _(sym, string) string,
+ foreach_ip6_error
+#undef _
+};
+
+/* *INDENT-OFF* */
+VLIB_REGISTER_NODE (ip6_reass_node, static) = {
+ .function = ip6_reassembly,
+ .name = "ip6-reassembly",
+ .vector_size = sizeof (u32),
+ .format_trace = format_ip6_reass_trace,
+ .n_errors = ARRAY_LEN (ip6_reassembly_error_strings),
+ .error_strings = ip6_reassembly_error_strings,
+ .n_next_nodes = IP6_REASSEMBLY_N_NEXT,
+ .next_nodes =
+ {
+ [IP6_REASSEMBLY_NEXT_INPUT] = "ip6-input",
+ [IP6_REASSEMBLY_NEXT_DROP] = "ip6-drop",
+ [IP6_REASSEMBLY_NEXT_ICMP_ERROR] = "ip6-icmp-error",
+ },
+};
+/* *INDENT-ON* */
+
+VLIB_NODE_FUNCTION_MULTIARCH (ip6_reass_node, ip6_reassembly)
+ static u32 ip6_reass_get_nbuckets ()
+{
+ ip6_reass_main_t *rm = &ip6_reass_main;
+ u32 nbuckets;
+ u8 i;
+
+ nbuckets = (u32) (rm->max_reass_n / IP6_REASS_HT_LOAD_FACTOR);
+
+ for (i = 0; i < 31; i++)
+ if ((1 << i) >= nbuckets)
+ break;
+ nbuckets = 1 << i;
+
+ return nbuckets;
+}
+
+typedef enum
+{
+ IP6_EVENT_CONFIG_CHANGED = 1,
+} ip6_reass_event_t;
+
+typedef struct
+{
+ int failure;
+ clib_bihash_48_8_t *new_hash;
+} ip6_rehash_cb_ctx;
+
+void
+ip6_rehash_cb (clib_bihash_kv_48_8_t * kv, void *_ctx)
+{
+ ip6_rehash_cb_ctx *ctx = _ctx;
+ if (clib_bihash_add_del_48_8 (ctx->new_hash, kv, 1))
+ {
+ ctx->failure = 1;
+ }
+}
+
+vnet_api_error_t
+ip6_reass_set (u32 timeout_ms, u32 max_reassemblies,
+ u32 expire_walk_interval_ms)
+{
+ u32 old_nbuckets = ip6_reass_get_nbuckets ();
+ ip6_reass_main.timeout_ms = timeout_ms;
+ ip6_reass_main.timeout = (f64) timeout_ms / (f64) MSEC_PER_SEC;
+ ip6_reass_main.max_reass_n = max_reassemblies;
+ ip6_reass_main.expire_walk_interval_ms = expire_walk_interval_ms;
+ vlib_process_signal_event (ip6_reass_main.vlib_main,
+ ip6_reass_main.ip6_reass_expire_node_idx,
+ IP6_EVENT_CONFIG_CHANGED, 0);
+ u32 new_nbuckets = ip6_reass_get_nbuckets ();
+ if (ip6_reass_main.max_reass_n > 0 && new_nbuckets > 1 &&
+ new_nbuckets != old_nbuckets)
+ {
+ clib_bihash_48_8_t new_hash;
+ memset (&new_hash, 0, sizeof (new_hash));
+ ip6_rehash_cb_ctx ctx;
+ ctx.failure = 0;
+ ctx.new_hash = &new_hash;
+ clib_bihash_init_48_8 (&new_hash, "ip6-reass", new_nbuckets,
+ new_nbuckets * 1024);
+ clib_bihash_foreach_key_value_pair_48_8 (&ip6_reass_main.hash,
+ ip6_rehash_cb, &ctx);
+ if (ctx.failure)
+ {
+ clib_bihash_free_48_8 (&new_hash);
+ return -1;
+ }
+ else
+ {
+ clib_bihash_free_48_8 (&ip6_reass_main.hash);
+ clib_memcpy (&ip6_reass_main.hash, &new_hash,
+ sizeof (ip6_reass_main.hash));
+ }
+ }
+ return 0;
+}
+
+vnet_api_error_t
+ip6_reass_get (u32 * timeout_ms, u32 * max_reassemblies,
+ u32 * expire_walk_interval_ms)
+{
+ *timeout_ms = ip6_reass_main.timeout_ms;
+ *max_reassemblies = ip6_reass_main.max_reass_n;
+ *expire_walk_interval_ms = ip6_reass_main.expire_walk_interval_ms;
+ return 0;
+}
+
+clib_error_t *
+ip6_reass_init_function (vlib_main_t * vm)
+{
+ ip6_reass_main_t *rm = &ip6_reass_main;
+ clib_error_t *error = 0;
+ u32 nbuckets;
+
+ rm->vlib_main = vm;
+ rm->vnet_main = vnet_get_main ();
+
+ rm->reass_n = 0;
+ pool_alloc (rm->pool, rm->max_reass_n);
+ ip6_reass_set (IP6_REASS_TIMEOUT_DEFAULT_MS,
+ IP6_REASS_MAX_REASSEMBLIES_DEAFULT,
+ IP6_REASS_EXPIRE_WALK_INTERVAL_DEFAULT_MS);
+
+ nbuckets = ip6_reass_get_nbuckets ();
+ clib_bihash_init_48_8 (&rm->hash, "ip6-reass", nbuckets, nbuckets * 1024);
+
+ vlib_node_t *node = vlib_get_node_by_name (vm, (u8 *) "ip6-drop");
+ ASSERT (node);
+ rm->ip6_drop_idx = node->index;
+ node = vlib_get_node_by_name (vm, (u8 *) "ip6-reassembly-expire-walk");
+ ASSERT (node);
+ rm->ip6_reass_expire_node_idx = node->index;
+ node = vlib_get_node_by_name (vm, (u8 *) "ip6-icmp-error");
+ ASSERT (node);
+ rm->ip6_icmp_error_idx = node->index;
+
+ if ((error = vlib_call_init_function (vm, ip_main_init)))
+ return error;
+ ip6_register_protocol (IP_PROTOCOL_IPV6_FRAGMENTATION,
+ ip6_reass_node.index);
+ return error;
+}
+
+VLIB_INIT_FUNCTION (ip6_reass_init_function);
+
+static uword
+ip6_reass_walk_expired (vlib_main_t * vm,
+ vlib_node_runtime_t * node, vlib_frame_t * f)
+{
+ ip6_reass_main_t *rm = &ip6_reass_main;
+ uword event_type, *event_data = 0;
+
+ while (true)
+ {
+ vlib_process_wait_for_event_or_clock (vm,
+ (f64) rm->expire_walk_interval_ms
+ / (f64) MSEC_PER_SEC);
+ event_type = vlib_process_get_events (vm, &event_data);
+
+ switch (event_type)
+ {
+ case ~0: /* no events => timeout */
+ /* nothing to do here */
+ break;
+ case IP6_EVENT_CONFIG_CHANGED:
+ break;
+ default:
+ clib_warning ("BUG: event type 0x%wx", event_type);
+ break;
+ }
+ f64 now = vlib_time_now (vm);
+
+ ip6_reass_t *reass;
+ u32 *vec_timeout = NULL;
+ int *pool_indexes_to_free = NULL;
+
+ int index;
+ /* *INDENT-OFF* */
+ pool_foreach_index (index, rm->pool, ({
+ reass = pool_elt_at_index (rm->pool, index);
+ if (now > reass->last_heard + rm->timeout)
+ {
+ vec_add1 (pool_indexes_to_free, index);
+ }
+ }));
+ /* *INDENT-ON* */
+ int *i;
+ u32 *vec_icmp_bi = NULL;
+ /* *INDENT-OFF* */
+ vec_foreach (i, pool_indexes_to_free)
+ {
+ ip6_reass_t *reass = pool_elt_at_index (rm->pool, i[0]);
+ u32 icmp_bi = ~0;
+ ip6_reass_on_timeout (vm, node, rm, reass, &icmp_bi, &vec_timeout);
+ if (~0 != icmp_bi)
+ {
+ vec_add1 (vec_icmp_bi, icmp_bi);
+ }
+ ip6_reass_free (rm, reass);
+ }
+ /* *INDENT-ON* */
+
+ while (vec_len (vec_timeout) > 0)
+ {
+ vlib_frame_t *f = vlib_get_frame_to_node (vm, rm->ip6_drop_idx);
+ u32 *to_next = vlib_frame_vector_args (f);
+ u32 n_left_to_next = VLIB_FRAME_SIZE - f->n_vectors;
+ u32 n_trace = 0;
+ while (vec_len (vec_timeout) > 0 && n_left_to_next > 0)
+ {
+ u32 bi = vec_pop (vec_timeout);
+ vlib_buffer_t *b = vlib_get_buffer (vm, bi);
+ if (PREDICT_FALSE (b->flags & VLIB_BUFFER_IS_TRACED))
+ {
+ if (pool_is_free_index (vm->trace_main.trace_buffer_pool,
+ b->trace_index))
+ {
+ /* the trace is gone, don't trace this buffer anymore */
+ b->flags &= ~VLIB_BUFFER_IS_TRACED;
+ }
+ else
+ {
+ ++n_trace;
+ }
+ }
+ b->error = node->errors[IP6_ERROR_REASS_TIMEOUT];
+ to_next[0] = bi;
+ ++f->n_vectors;
+ to_next += 1;
+ n_left_to_next -= 1;
+ ASSERT (rm->buffers_n > 0);
+ --rm->buffers_n;
+ }
+ if (PREDICT_FALSE (n_trace > 0))
+ {
+ f->flags |= VLIB_FRAME_TRACE;
+ }
+ vlib_put_frame_to_node (vm, rm->ip6_drop_idx, f);
+ }
+
+ while (vec_len (vec_icmp_bi) > 0)
+ {
+ vlib_frame_t *f =
+ vlib_get_frame_to_node (vm, rm->ip6_icmp_error_idx);
+ u32 *to_next = vlib_frame_vector_args (f);
+ u32 n_left_to_next = VLIB_FRAME_SIZE - f->n_vectors;
+ u32 n_trace = 0;
+ while (vec_len (vec_icmp_bi) > 0 && n_left_to_next > 0)
+ {
+ u32 bi = vec_pop (vec_icmp_bi);
+ vlib_buffer_t *b = vlib_get_buffer (vm, bi);
+ if (PREDICT_FALSE (b->flags & VLIB_BUFFER_IS_TRACED))
+ {
+ if (pool_is_free_index (vm->trace_main.trace_buffer_pool,
+ b->trace_index))
+ {
+ /* the trace is gone, don't trace this buffer anymore */
+ b->flags &= ~VLIB_BUFFER_IS_TRACED;
+ }
+ else
+ {
+ ++n_trace;
+ }
+ }
+ b->error = node->errors[IP6_ERROR_REASS_TIMEOUT];
+ to_next[0] = bi;
+ ++f->n_vectors;
+ to_next += 1;
+ n_left_to_next -= 1;
+ ASSERT (rm->buffers_n > 0);
+ --rm->buffers_n;
+ }
+ if (PREDICT_FALSE (n_trace > 0))
+ {
+ f->flags |= VLIB_FRAME_TRACE;
+ }
+ vlib_put_frame_to_node (vm, rm->ip6_icmp_error_idx, f);
+ }
+
+ vec_free (pool_indexes_to_free);
+ vec_free (vec_timeout);
+ vec_free (vec_icmp_bi);
+ if (event_data)
+ {
+ _vec_len (event_data) = 0;
+ }
+ }
+
+ return 0;
+}
+
+static vlib_node_registration_t ip6_reass_expire_node;
+
+/* *INDENT-OFF* */
+VLIB_REGISTER_NODE (ip6_reass_expire_node, static) = {
+ .function = ip6_reass_walk_expired,
+ .format_trace = format_ip6_reass_trace,
+ .type = VLIB_NODE_TYPE_PROCESS,
+ .name = "ip6-reassembly-expire-walk",
+
+ .n_errors = ARRAY_LEN (ip6_reassembly_error_strings),
+ .error_strings = ip6_reassembly_error_strings,
+
+};
+/* *INDENT-ON* */
+
+static u8 *
+format_ip6_reass_key (u8 * s, va_list * args)
+{
+ ip6_reass_key_t *key = va_arg (*args, ip6_reass_key_t *);
+ s = format (s, "xx_id: %u, src: %U, dst: %U, frag_id: %u, proto: %u",
+ key->xx_id, format_ip6_address, &key->src, format_ip6_address,
+ &key->dst, clib_net_to_host_u16 (key->frag_id), key->proto);
+ return s;
+}
+
+static u8 *
+format_ip6_reass (u8 * s, va_list * args)
+{
+ vlib_main_t *vm = va_arg (*args, vlib_main_t *);
+ ip6_reass_t *reass = va_arg (*args, ip6_reass_t *);
+
+ s = format (s, "ID: %u, key: %U\n first_bi: %u, data_len: %u, "
+ "last_packet_octet: %u, trace_op_counter: %u\n",
+ reass->id, format_ip6_reass_key, &reass->key, reass->first_bi,
+ reass->data_len, reass->last_packet_octet,
+ reass->trace_op_counter);
+ u32 bi = reass->first_bi;
+ u32 counter = 0;
+ while (~0 != bi)
+ {
+ vlib_buffer_t *b = vlib_get_buffer (vm, bi);
+ vnet_buffer_opaque_t *vnb = vnet_buffer (b);
+ s = format (s, " #%03u: range: [%u, %u], bi: %u, off: %d, len: %u, "
+ "fragment[%u, %u]\n",
+ counter, vnb->ip.reass.range_first,
+ vnb->ip.reass.range_last, bi,
+ ip6_reass_buffer_get_data_offset_no_check (b),
+ ip6_reass_buffer_get_data_len_no_check (b),
+ vnb->ip.reass.fragment_first, vnb->ip.reass.fragment_last);
+ if (b->flags & VLIB_BUFFER_NEXT_PRESENT)
+ {
+ bi = b->next_buffer;
+ }
+ else
+ {
+ bi = ~0;
+ }
+ }
+ return s;
+}
+
+static clib_error_t *
+show_ip6_reass (vlib_main_t * vm, unformat_input_t * input,
+ CLIB_UNUSED (vlib_cli_command_t * lmd))
+{
+ ip6_reass_main_t *rm = &ip6_reass_main;
+
+ vlib_cli_output (vm, "---------------------");
+ vlib_cli_output (vm, "IP6 reassembly status");
+ vlib_cli_output (vm, "---------------------");
+ if (unformat (input, "details"))
+ {
+ ip6_reass_t *reass;
+ /* *INDENT-OFF* */
+ pool_foreach (reass, rm->pool, {
+ vlib_cli_output (vm, "%U", format_ip6_reass, vm, reass);
+ });
+ /* *INDENT-ON* */
+ }
+ vlib_cli_output (vm, "---------------------");
+ vlib_cli_output (vm, "Current IP6 reassemblies count: %lu\n", rm->reass_n);
+ vlib_cli_output (vm,
+ "Maximum configured concurrent IP6 reassemblies: %lu\n",
+ (long unsigned) rm->max_reass_n);
+ vlib_cli_output (vm, "Buffers in use: %lu\n",
+ (long unsigned) rm->buffers_n);
+ return 0;
+}
+
+/* *INDENT-OFF* */
+VLIB_CLI_COMMAND (show_ip6_reassembly_cmd, static) = {
+ .path = "show ip6-reassembly",
+ .short_help = "show ip6-reassembly [details]",
+ .function = show_ip6_reass,
+};
+/* *INDENT-ON* */
+
+/*
+ * fd.io coding-style-patch-verification: ON
+ *
+ * Local Variables:
+ * eval: (c-set-style "gnu")
+ * End:
+ */
diff --git a/src/vnet/ip/ip6_reassembly.h b/src/vnet/ip/ip6_reassembly.h
new file mode 100644
index 00000000000..0de4e04bb2b
--- /dev/null
+++ b/src/vnet/ip/ip6_reassembly.h
@@ -0,0 +1,49 @@
+/*
+ * 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 IPv6 Reassembly.
+ *
+ * This file contains the source code for IPv6 reassembly.
+ */
+
+#ifndef __included_ip6_reassembly_h__
+#define __included_ip6_reassembly_h__
+
+#include <vnet/api_errno.h>
+#include <vnet/vnet.h>
+
+/**
+ * @brief set ip6 reassembly configuration
+ */
+vnet_api_error_t ip6_reass_set (u32 timeout_ms, u32 max_reassemblies,
+ u32 expire_walk_interval_ms);
+
+/**
+ * @brief get ip6 reassembly configuration
+ */
+vnet_api_error_t ip6_reass_get (u32 * timeout_ms, u32 * max_reassemblies,
+ u32 * expire_walk_interval_ms);
+
+#endif /* __included_ip6_reassembly_h */
+
+/*
+ * fd.io coding-style-patch-verification: ON
+ *
+ * Local Variables:
+ * eval: (c-set-style "gnu")
+ * End:
+ */
diff --git a/src/vnet/ip/ip_api.c b/src/vnet/ip/ip_api.c
index c0889ebfa6b..60fa2faf926 100644
--- a/src/vnet/ip/ip_api.c
+++ b/src/vnet/ip/ip_api.c
@@ -42,6 +42,8 @@
#include <vnet/fib/ip4_fib.h>
#include <vnet/fib/ip6_fib.h>
#include <vnet/ip/ip6_hop_by_hop.h>
+#include <vnet/ip/ip4_reassembly.h>
+#include <vnet/ip/ip6_reassembly.h>
#include <vnet/vnet_msg_enum.h>
@@ -97,7 +99,9 @@ _(IOAM_DISABLE, ioam_disable) \
_(IP_SOURCE_AND_PORT_RANGE_CHECK_ADD_DEL, \
ip_source_and_port_range_check_add_del) \
_(IP_SOURCE_AND_PORT_RANGE_CHECK_INTERFACE_ADD_DEL, \
- ip_source_and_port_range_check_interface_add_del)
+ ip_source_and_port_range_check_interface_add_del) \
+_(IP_REASSEMBLY_SET, ip_reassembly_set) \
+_(IP_REASSEMBLY_GET, ip_reassembly_get)
extern void stats_dslock_with_hint (int hint, int tag);
extern void stats_dsunlock (void);
@@ -2786,6 +2790,61 @@ vl_api_set_arp_neighbor_limit_t_handler (vl_api_set_arp_neighbor_limit_t * mp)
REPLY_MACRO (VL_API_SET_ARP_NEIGHBOR_LIMIT_REPLY);
}
+void
+vl_api_ip_reassembly_set_t_handler (vl_api_ip_reassembly_set_t * mp)
+{
+ vl_api_ip_reassembly_set_reply_t *rmp;
+ int rv = 0;
+ if (mp->is_ip6)
+ {
+ rv = ip6_reass_set (clib_net_to_host_u32 (mp->timeout_ms),
+ clib_net_to_host_u32 (mp->max_reassemblies),
+ clib_net_to_host_u32 (mp->expire_walk_interval_ms));
+ }
+ else
+ {
+ rv = ip4_reass_set (clib_net_to_host_u32 (mp->timeout_ms),
+ clib_net_to_host_u32 (mp->max_reassemblies),
+ clib_net_to_host_u32 (mp->expire_walk_interval_ms));
+ }
+
+ REPLY_MACRO (VL_API_IP_REASSEMBLY_SET_REPLY);
+}
+
+void
+vl_api_ip_reassembly_get_t_handler (vl_api_ip_reassembly_get_t * mp)
+{
+ unix_shared_memory_queue_t *q;
+
+ q = vl_api_client_index_to_input_queue (mp->client_index);
+
+ if (q == 0)
+ return;
+
+ vl_api_ip_reassembly_get_reply_t *rmp = vl_msg_api_alloc (sizeof (*rmp));
+ memset (rmp, 0, sizeof (*rmp));
+ rmp->_vl_msg_id = ntohs (VL_API_IP_REASSEMBLY_GET_REPLY);
+ rmp->context = mp->context;
+ rmp->retval = 0;
+ if (mp->is_ip6)
+ {
+ rmp->is_ip6 = 1;
+ ip6_reass_get (&rmp->timeout_ms, &rmp->max_reassemblies,
+ &rmp->expire_walk_interval_ms);
+ }
+ else
+ {
+ rmp->is_ip6 = 0;
+ ip4_reass_get (&rmp->timeout_ms, &rmp->max_reassemblies,
+ &rmp->expire_walk_interval_ms);
+ }
+ rmp->timeout_ms = clib_host_to_net_u32 (rmp->timeout_ms);
+ rmp->max_reassemblies = clib_host_to_net_u32 (rmp->max_reassemblies);
+ rmp->expire_walk_interval_ms =
+ clib_host_to_net_u32 (rmp->expire_walk_interval_ms);
+ vl_msg_api_send_shmem (q, (u8 *) & rmp);
+}
+
#define vl_msg_name_crc_list
#include <vnet/ip/ip.api.h>
#undef vl_msg_name_crc_list
diff --git a/src/vnet/ip/lookup.c b/src/vnet/ip/lookup.c
index 67e46bd8cfa..f2880bf8cd9 100644
--- a/src/vnet/ip/lookup.c
+++ b/src/vnet/ip/lookup.c
@@ -230,6 +230,8 @@ ip_lookup_init (ip_lookup_main_t * lm, u32 is_ip6)
}
lm->local_next_by_ip_protocol[IP_PROTOCOL_UDP] = IP_LOCAL_NEXT_UDP_LOOKUP;
+ lm->local_next_by_ip_protocol[IP_PROTOCOL_VPP_FRAGMENTATION] =
+ IP_LOCAL_NEXT_REASSEMBLY;
lm->local_next_by_ip_protocol[is_ip6 ? IP_PROTOCOL_ICMP6 :
IP_PROTOCOL_ICMP] = IP_LOCAL_NEXT_ICMP;
lm->builtin_protocol_by_ip_protocol[IP_PROTOCOL_UDP] =
diff --git a/src/vnet/ip/lookup.h b/src/vnet/ip/lookup.h
index a89546f5d92..9fe7a1b0085 100644
--- a/src/vnet/ip/lookup.h
+++ b/src/vnet/ip/lookup.h
@@ -111,6 +111,7 @@ typedef enum
IP_LOCAL_NEXT_PUNT,
IP_LOCAL_NEXT_UDP_LOOKUP,
IP_LOCAL_NEXT_ICMP,
+ IP_LOCAL_NEXT_REASSEMBLY,
IP_LOCAL_N_NEXT,
} ip_local_next_t;
diff --git a/src/vnet/ip/protocols.def b/src/vnet/ip/protocols.def
index 77fab31da05..4ec0bbc9284 100644
--- a/src/vnet/ip/protocols.def
+++ b/src/vnet/ip/protocols.def
@@ -158,5 +158,10 @@ ip_protocol (134, RSVP_E2E_IGNORE)
ip_protocol (135, MOBILITY)
ip_protocol (136, UDP_LITE)
ip_protocol (137, MPLS_IN_IP)
+/*
+ * VPPs way of dealing with fragments is to mark them as 0xfe, to be picked
+ * by the corresponding nodes based on this protocol
+ */
+ip_protocol (0xfe, VPP_FRAGMENTATION)
ip_protocol (255, RESERVED)
diff --git a/test/Makefile b/test/Makefile
index f3b7f5c6d51..155325a9f17 100644
--- a/test/Makefile
+++ b/test/Makefile
@@ -88,7 +88,7 @@ $(PAPI_INSTALL_DONE): $(PIP_PATCH_DONE)
@touch $@
define retest-func
- @env VPP_TEST_FAILED_DIR=$(VPP_TEST_FAILED_DIR) scripts/setsid_wrapper.sh $(FORCE_FOREGROUND) $(PYTHON_VENV_PATH)/bin/activate python run_tests.py -d $(TEST_DIR) $(UNITTEST_EXTRA_OPTS) || env VPP_TEST_FAILED_DIR=$(VPP_TEST_FAILED_DIR) COMPRESS_FAILED_TEST_LOGS=$(COMPRESS_FAILED_TEST_LOGS) scripts/compress_failed.sh
+@env VPP_TEST_FAILED_DIR=$(VPP_TEST_FAILED_DIR) scripts/setsid_wrapper.sh $(FORCE_FOREGROUND) $(PYTHON_VENV_PATH)/bin/activate python run_tests.py -d $(TEST_DIR) $(UNITTEST_EXTRA_OPTS) || env VPP_TEST_FAILED_DIR=$(VPP_TEST_FAILED_DIR) COMPRESS_FAILED_TEST_LOGS=$(COMPRESS_FAILED_TEST_LOGS) scripts/compress_failed.sh
endef
.PHONY: sanity
@@ -103,6 +103,7 @@ endif
sanity: verify-no-running-vpp
@sys_req/dev_shm_size.sh
+ @sys_req/set_system_parameters.sh sys_req/system_parameters
@bash -c "$(SANITY_IMPORT_VPP_PAPI_CMD) ||\
(echo \"*******************************************************************\" &&\
echo \"* Sanity check failed, cannot import vpp_papi\" &&\
diff --git a/test/framework.py b/test/framework.py
index 2e3f9781e5d..fc6f550184f 100644
--- a/test/framework.py
+++ b/test/framework.py
@@ -24,6 +24,7 @@ from vpp_lo_interface import VppLoInterface
from vpp_papi_provider import VppPapiProvider
from log import *
from vpp_object import VppObjectRegistry
+from vpp_punt_socket import vpp_uds_socket_name
if os.name == 'posix' and sys.version_info[0] < 3:
# using subprocess32 is recommended by python official documentation
# @ https://docs.python.org/2/library/subprocess.html
@@ -255,7 +256,8 @@ class VppTestCase(unittest.TestCase):
coredump_size, "}", "api-trace", "{", "on", "}",
"api-segment", "{", "prefix", cls.shm_prefix, "}",
"plugins", "{", "plugin", "dpdk_plugin.so", "{",
- "disable", "}", "}"]
+ "disable", "}", "}",
+ "punt", "{", "socket", cls.punt_socket_path, "}"]
if plugin_path is not None:
cls.vpp_cmdline.extend(["plugin_path", plugin_path])
cls.logger.info("vpp_cmdline: %s" % cls.vpp_cmdline)
@@ -317,7 +319,7 @@ class VppTestCase(unittest.TestCase):
Remove shared memory files, start vpp and connect the vpp-api
"""
gc.collect() # run garbage collection first
- random.seed()
+ random.seed(1)
cls.logger = getLogger(cls.__name__)
cls.tempdir = tempfile.mkdtemp(
prefix='vpp-unittest-%s-' % cls.__name__)
@@ -328,6 +330,7 @@ class VppTestCase(unittest.TestCase):
cls.file_handler.setLevel(DEBUG)
cls.logger.addHandler(cls.file_handler)
cls.shm_prefix = cls.tempdir.split("/")[-1]
+ cls.punt_socket_path = '%s/%s' % (cls.tempdir, vpp_uds_socket_name)
os.chdir(cls.tempdir)
cls.logger.info("Temporary dir is %s, shm prefix is %s",
cls.tempdir, cls.shm_prefix)
@@ -499,13 +502,16 @@ class VppTestCase(unittest.TestCase):
type(self).test_instance = self
@classmethod
- def pg_enable_capture(cls, interfaces):
+ def pg_enable_capture(cls, interfaces=None):
"""
Enable capture on packet-generator interfaces
- :param interfaces: iterable interface indexes
+ :param interfaces: iterable interface indexes (if None,
+ use self.pg_interfaces)
"""
+ if interfaces is None:
+ interfaces = cls.pg_interfaces
for i in interfaces:
i.enable_capture()
@@ -573,19 +579,21 @@ class VppTestCase(unittest.TestCase):
return result
@staticmethod
- def extend_packet(packet, size):
+ def extend_packet(packet, size, padding=' '):
"""
- Extend packet to given size by padding with spaces
+ Extend packet to given size by padding with spaces or custom padding
NOTE: Currently works only when Raw layer is present.
:param packet: packet
:param size: target size
+ :param padding: padding used to extend the payload
"""
packet_len = len(packet) + 4
extend = size - packet_len
if extend > 0:
- packet[Raw].load += ' ' * extend
+ num = (extend / len(padding)) + 1
+ packet[Raw].load += (padding * num)[:extend]
@classmethod
def reset_packet_infos(cls):
diff --git a/test/hook.py b/test/hook.py
index 77a2fc5bb23..0ef1377ba99 100644
--- a/test/hook.py
+++ b/test/hook.py
@@ -2,7 +2,7 @@ import signal
import os
import traceback
from log import RED, single_line_delim, double_line_delim
-from debug import spawn_gdb, gdb_path
+from debug import spawn_gdb
class Hook(object):
@@ -62,14 +62,10 @@ class PollHook(Hook):
def on_crash(self, core_path):
if self.testcase.debug_core:
- if not spawn_gdb(self.testcase.vpp_bin, core_path, self.logger):
- self.logger.error(
- "Debugger '%s' does not exist or is not an executable.." %
- gdb_path)
- else:
- return
- self.logger.critical("Core file present, debug with: gdb %s %s" %
- (self.testcase.vpp_bin, core_path))
+ spawn_gdb(self.testcase.vpp_bin, core_path, self.logger)
+ else:
+ self.logger.critical("Core file present, debug with: gdb %s %s" %
+ (self.testcase.vpp_bin, core_path))
def poll_vpp(self):
"""
diff --git a/test/patches/scapy-2.3.3/inet6.py.patch b/test/patches/scapy-2.3.3/inet6.py.patch
index cf3217eb672..706e19738c4 100644
--- a/test/patches/scapy-2.3.3/inet6.py.patch
+++ b/test/patches/scapy-2.3.3/inet6.py.patch
@@ -1,7 +1,8 @@
diff --git a/scapy/layers/inet6.py b/scapy/layers/inet6.py
---- a/scapy/layers/inet6.py 2017-06-01 14:04:18.160881034 +0200
-+++ b/scapy/layers/inet6.py 2017-06-02 09:08:40.133800208 +0200
-@@ -369,6 +369,8 @@
+index 03b80ec..06ef27f 100644
+--- a/scapy/layers/inet6.py
++++ b/scapy/layers/inet6.py
+@@ -369,6 +369,8 @@ class _IPv6GuessPayload:
return Raw
elif self.nh == 135 and len(p) > 3: # Mobile IPv6
return _mip6_mhtype2cls.get(ord(p[2]), MIP6MH_Generic)
@@ -9,11 +10,11 @@ diff --git a/scapy/layers/inet6.py b/scapy/layers/inet6.py
+ return IPv6ExtHdrSegmentRouting
else:
return get_cls(ipv6nhcls.get(self.nh,"Raw"), "Raw")
-
-@@ -430,6 +432,14 @@
+
+@@ -430,6 +432,14 @@ class IPv6(_IPv6GuessPayload, Packet, IPTools):
sd = strxor(sd, a)
sd = inet_ntop(socket.AF_INET6, sd)
-
+
+ if self.nh == 43 and isinstance(self.payload, IPv6ExtHdrSegmentRouting):
+ # With segment routing header (rh == 4), the destination is
+ # the first address of the IPv6 addresses list
@@ -23,10 +24,10 @@ diff --git a/scapy/layers/inet6.py b/scapy/layers/inet6.py
+ sd = self.dst
+
if self.nh == 44 and isinstance(self.payload, IPv6ExtHdrFragment):
- nh = self.payload.nh
-
-@@ -489,6 +499,8 @@
- return self.payload.answers(other.payload.payload)
+ nh = self.payload.nh
+
+@@ -489,6 +499,8 @@ class IPv6(_IPv6GuessPayload, Packet, IPTools):
+ return self.payload.answers(other.payload.payload)
elif other.nh == 43 and isinstance(other.payload, IPv6ExtHdrRouting):
return self.payload.answers(other.payload.payload) # Buggy if self.payload is a IPv6ExtHdrRouting
+ elif other.nh == 43 and isinstance(other.payload, IPv6ExtHdrSegmentRouting):
@@ -34,10 +35,10 @@ diff --git a/scapy/layers/inet6.py b/scapy/layers/inet6.py
elif other.nh == 60 and isinstance(other.payload, IPv6ExtHdrDestOpt):
return self.payload.payload.answers(other.payload.payload)
elif self.nh == 60 and isinstance(self.payload, IPv6ExtHdrDestOpt): # BU in reply to BRR, for instance
-@@ -919,6 +931,148 @@
+@@ -919,6 +931,148 @@ class IPv6ExtHdrRouting(_IPv6ExtHdr):
pkt = pkt[:3]+struct.pack("B", len(self.addresses))+pkt[4:]
return _IPv6ExtHdr.post_build(self, pkt, pay)
-
+
+######################### Segment Routing Header ############################
+
+# This implementation is based on draft 06, available at:
diff --git a/test/sys_req/set_system_parameters.sh b/test/sys_req/set_system_parameters.sh
new file mode 100755
index 00000000000..533a02edbaa
--- /dev/null
+++ b/test/sys_req/set_system_parameters.sh
@@ -0,0 +1,57 @@
+#!/bin/bash
+
+file="$1"
+
+usage(){
+ echo "Usage: $0 <requirements file>"
+}
+
+if [ "$file" == "" ]
+then
+ echo "Invalid parameters specified."
+ usage
+ exit 1
+fi
+
+if [ ! -f $file ]
+then
+ echo "File '$file' does not exist."
+ usage
+ exit 1
+fi
+
+cat $file | grep -v -e '^#.*$' | grep -v -e '^ *$' | while read line
+do
+ value_file=`echo $line | awk '{print $1}'`
+ operator=`echo $line | awk '{print $2}'`
+ value=`echo $line | awk '{print $3}'`
+ set_value=`echo $line | awk '{print $4}'`
+ if [[ "$value_file" == "" || "$operator" == "" || "$value" == "" || "$set_value" == "" ]]
+ then
+ echo "Syntax error in requirements file."
+ exit 1
+ fi
+ current_value=`cat $value_file`
+ if test "$current_value" $operator "$value"
+ then
+ if test "$V" = "2"
+ then
+ echo "Requirement '$value_file $operator $value' satisfied."
+ fi
+ else
+ echo "Requirement '$value_file $operator $value' not satisfied."
+ echo "Writing '$set_value' to '$value_file'."
+ echo "$set_value" | tee "$value_file" > /dev/null
+ if ! test "`cat $value_file`" = "$set_value"
+ then
+ echo "Repeating the write using sudo..."
+ echo "$set_value" | sudo -n tee "$value_file" > /dev/null
+ if ! test "`cat $value_file`" = "$set_value"
+ then
+ echo "Couldn't set the required value. Is that value allowed? Is sudo working?"
+ exit 1
+ fi
+ fi
+ echo "Succesfully wrote '$set_value' to '$value_file'."
+ fi
+done
diff --git a/test/sys_req/system_parameters b/test/sys_req/system_parameters
new file mode 100644
index 00000000000..6373774dd94
--- /dev/null
+++ b/test/sys_req/system_parameters
@@ -0,0 +1,18 @@
+# test framework system requirements
+# format of this file is
+# <path> <operator> <comparison-value> <set-value>
+#
+# path - path to value e.g. in /proc which needs to be checked
+# operator - test operator (e.g. -gt)
+# comparison-value - value, against which the value read from <path> is compared
+# set-value - value, to which the path is set if the test fails
+#
+# the comparison is done using `test' command
+
+
+# test_reassembly.py
+# needed by test_reassembly which uses udp punt via unix domain sockets
+# to ensure that all data which vpp might produce in a burst fits into
+# the socket send buffer
+/proc/sys/net/core/wmem_max -ge 4636252 4636252
+/proc/sys/net/core/wmem_default -ge 4636252 4636252
diff --git a/test/test_reassembly.py b/test/test_reassembly.py
new file mode 100644
index 00000000000..a2d77d493de
--- /dev/null
+++ b/test/test_reassembly.py
@@ -0,0 +1,966 @@
+#!/usr/bin/env python
+import unittest
+from random import shuffle
+
+from framework import VppTestCase, VppTestRunner
+
+from scapy.packet import Raw
+from scapy.layers.l2 import Ether, GRE
+from scapy.layers.inet import IP, UDP
+from util import ppp, fragment_rfc791, fragment_rfc8200
+from vpp_punt_socket import VppUDSPuntSocket
+from scapy.layers.inet6 import IPv6, IPv6ExtHdrFragment, ICMPv6ParamProblem,\
+ ICMPv6TimeExceeded
+from vpp_gre_interface import VppGreInterface, VppGre6Interface
+from vpp_ip_route import VppIpRoute, VppRoutePath, DpoProto
+
+test_packet_count = 257
+
+
+class TestIPv4Reassembly(VppTestCase):
+ """ IPv4 Reassembly """
+
+ @classmethod
+ def setUpClass(cls):
+ super(TestIPv4Reassembly, cls).setUpClass()
+
+ cls.create_pg_interfaces([0])
+ cls.pg_if = cls.pg0
+
+ # setup all interfaces
+ for i in cls.pg_interfaces:
+ i.admin_up()
+ i.config_ip4()
+ i.resolve_arp()
+
+ cls.punt_port = 9999
+ cls.punt_socket = VppUDSPuntSocket(cls, cls.punt_port)
+
+ # packet sizes
+ cls.packet_sizes = [64, 512, 1518, 9018]
+ cls.padding = " abcdefghijklmn"
+ cls.create_stream(cls.packet_sizes)
+ cls.create_fragments()
+
+ def setUp(self):
+ """ Test setup - force timeout on existing reassemblies """
+ super(TestIPv4Reassembly, self).setUp()
+ self.vapi.ip_reassembly_set(timeout_ms=0, max_reassemblies=1000,
+ expire_walk_interval_ms=10)
+ self.sleep(.25)
+ self.vapi.ip_reassembly_set(timeout_ms=1000000, max_reassemblies=1000,
+ expire_walk_interval_ms=10000)
+
+ def tearDown(self):
+ super(TestIPv4Reassembly, self).tearDown()
+ self.logger.debug(self.vapi.ppcli("show ip4-reassembly details"))
+
+ @classmethod
+ def create_stream(cls, packet_sizes, packet_count=test_packet_count):
+ """Create input packet stream for defined interface.
+
+ :param list packet_sizes: Required packet sizes.
+ """
+ for i in range(0, packet_count):
+ info = cls.create_packet_info(cls.pg_if, cls.pg_if)
+ payload = cls.info_to_payload(info)
+ p = (Ether(dst=cls.pg_if.local_mac, src=cls.pg_if.remote_mac) /
+ IP(id=info.index, src=cls.pg_if.remote_ip4,
+ dst=cls.pg_if.local_ip4) /
+ UDP(sport=1234, dport=cls.punt_port) /
+ Raw(payload))
+ size = packet_sizes[(i // 2) % len(packet_sizes)]
+ cls.extend_packet(p, size, cls.padding)
+ info.data = p
+
+ @classmethod
+ def create_fragments(cls):
+ infos = cls._packet_infos
+ cls.pkt_infos = []
+ for index, info in infos.iteritems():
+ p = info.data
+ # self.logger.debug(ppp("Packet:", p.__class__(str(p))))
+ fragments_400 = fragment_rfc791(p, 400)
+ fragments_300 = fragment_rfc791(p, 300)
+ fragments_200 = [
+ x for f in fragments_400 for x in fragment_rfc791(f, 200)]
+ cls.pkt_infos.append(
+ (index, fragments_400, fragments_300, fragments_200))
+ cls.fragments_400 = [
+ x for (_, frags, _, _) in cls.pkt_infos for x in frags]
+ cls.fragments_300 = [
+ x for (_, _, frags, _) in cls.pkt_infos for x in frags]
+ cls.fragments_200 = [
+ x for (_, _, _, frags) in cls.pkt_infos for x in frags]
+ cls.logger.debug("Fragmented %s packets into %s 400-byte fragments, "
+ "%s 300-byte fragments and %s 200-byte fragments" %
+ (len(infos), len(cls.fragments_400),
+ len(cls.fragments_300), len(cls.fragments_200)))
+
+ def verify_capture(self, capture, dropped_packet_indexes=[]):
+ """Verify captured packet stream.
+
+ :param list capture: Captured packet stream.
+ """
+ info = None
+ seen = set()
+ for packet in capture:
+ try:
+ sw_if_index = packet['sw_if_index']
+ punt_action = packet['punt_action']
+ packet = Ether(packet['packet'])
+ self.logger.debug(ppp("Got packet from %s, action %s" %
+ (sw_if_index, punt_action), packet))
+ ip = packet[IP]
+ udp = packet[UDP]
+ payload_info = self.payload_to_info(str(packet[Raw]))
+ packet_index = payload_info.index
+ self.assertTrue(
+ packet_index not in dropped_packet_indexes,
+ ppp("Packet received, but should be dropped:", packet))
+ if packet_index in seen:
+ raise Exception(ppp("Duplicate packet received", packet))
+ seen.add(packet_index)
+ self.assertEqual(payload_info.dst, self.pg_if.sw_if_index)
+ info = self._packet_infos[packet_index]
+ self.assertTrue(info is not None)
+ self.assertEqual(packet_index, info.index)
+ saved_packet = info.data
+ self.assertEqual(ip.src, saved_packet[IP].src)
+ self.assertEqual(ip.dst, saved_packet[IP].dst)
+ self.assertEqual(udp.payload, saved_packet[UDP].payload)
+ except:
+ self.logger.error(ppp("Unexpected or invalid packet:", packet))
+ raise
+ for index in self._packet_infos:
+ self.assertTrue(index in seen or index in dropped_packet_indexes,
+ "Packet with packet_index %d not received" % index)
+
+ def test_reassembly(self):
+ """ basic reassembly """
+
+ self.pg_enable_capture()
+ self.pg_if.add_stream(self.fragments_200)
+ self.pg_start()
+
+ packets = self.punt_socket.wait_for_packets(len(self.pkt_infos))
+ self.verify_capture(packets)
+ self.pg_if.assert_nothing_captured()
+
+ # run it all again to verify correctness
+ self.pg_enable_capture()
+ self.pg_if.add_stream(self.fragments_200)
+ self.pg_start()
+
+ packets = self.punt_socket.wait_for_packets(len(self.pkt_infos))
+ self.verify_capture(packets)
+ self.pg_if.assert_nothing_captured()
+
+ def test_reversed(self):
+ """ reverse order reassembly """
+
+ fragments = list(self.fragments_200)
+ fragments.reverse()
+
+ self.pg_enable_capture()
+ self.pg_if.add_stream(fragments)
+ self.pg_start()
+
+ packets = self.punt_socket.wait_for_packets(len(self.packet_infos))
+ self.verify_capture(packets)
+ self.pg_if.assert_nothing_captured()
+
+ # run it all again to verify correctness
+ self.pg_enable_capture()
+ self.pg_if.add_stream(fragments)
+ self.pg_start()
+
+ packets = self.punt_socket.wait_for_packets(len(self.packet_infos))
+ self.verify_capture(packets)
+ self.pg_if.assert_nothing_captured()
+
+ def test_random(self):
+ """ random order reassembly """
+
+ fragments = list(self.fragments_200)
+ shuffle(fragments)
+
+ self.pg_enable_capture()
+ self.pg_if.add_stream(fragments)
+ self.pg_start()
+
+ packets = self.punt_socket.wait_for_packets(len(self.packet_infos))
+ self.verify_capture(packets)
+ self.pg_if.assert_nothing_captured()
+
+ # run it all again to verify correctness
+ self.pg_enable_capture()
+ self.pg_if.add_stream(fragments)
+ self.pg_start()
+
+ packets = self.punt_socket.wait_for_packets(len(self.packet_infos))
+ self.verify_capture(packets)
+ self.pg_if.assert_nothing_captured()
+
+ def test_duplicates(self):
+ """ duplicate fragments """
+
+ fragments = [
+ x for (_, frags, _, _) in self.pkt_infos
+ for x in frags
+ for _ in range(0, min(2, len(frags)))
+ ]
+
+ self.pg_enable_capture()
+ self.pg_if.add_stream(fragments)
+ self.pg_start()
+
+ packets = self.punt_socket.wait_for_packets(len(self.pkt_infos))
+ self.verify_capture(packets)
+ self.pg_if.assert_nothing_captured()
+
+ def test_overlap1(self):
+ """ overlapping fragments case #1 """
+
+ fragments = []
+ for _, _, frags_300, frags_200 in self.pkt_infos:
+ if len(frags_300) == 1:
+ fragments.extend(frags_300)
+ else:
+ for i, j in zip(frags_200, frags_300):
+ fragments.extend(i)
+ fragments.extend(j)
+
+ self.pg_enable_capture()
+ self.pg_if.add_stream(fragments)
+ self.pg_start()
+
+ packets = self.punt_socket.wait_for_packets(len(self.pkt_infos))
+ self.verify_capture(packets)
+ self.pg_if.assert_nothing_captured()
+
+ # run it all to verify correctness
+ self.pg_enable_capture()
+ self.pg_if.add_stream(fragments)
+ self.pg_start()
+
+ packets = self.punt_socket.wait_for_packets(len(self.pkt_infos))
+ self.verify_capture(packets)
+ self.pg_if.assert_nothing_captured()
+
+ def test_overlap2(self):
+ """ overlapping fragments case #2 """
+
+ fragments = []
+ for _, _, frags_300, frags_200 in self.pkt_infos:
+ if len(frags_300) == 1:
+ fragments.extend(frags_300)
+ else:
+ # care must be taken here so that there are no fragments
+ # received by vpp after reassembly is finished, otherwise
+ # new reassemblies will be started and packet generator will
+ # freak out when it detects unfreed buffers
+ zipped = zip(frags_300, frags_200)
+ for i, j in zipped[:-1]:
+ fragments.extend(i)
+ fragments.extend(j)
+ fragments.append(zipped[-1][0])
+
+ self.pg_enable_capture()
+ self.pg_if.add_stream(fragments)
+ self.pg_start()
+
+ packets = self.punt_socket.wait_for_packets(len(self.pkt_infos))
+ self.verify_capture(packets)
+ self.pg_if.assert_nothing_captured()
+
+ # run it all to verify correctness
+ self.pg_enable_capture()
+ self.pg_if.add_stream(fragments)
+ self.pg_start()
+
+ packets = self.punt_socket.wait_for_packets(len(self.pkt_infos))
+ self.verify_capture(packets)
+ self.pg_if.assert_nothing_captured()
+
+ def test_timeout_inline(self):
+ """ timeout (inline) """
+
+ dropped_packet_indexes = set(
+ index for (index, frags, _, _) in self.pkt_infos if len(frags) > 1
+ )
+
+ self.vapi.ip_reassembly_set(timeout_ms=0, max_reassemblies=1000,
+ expire_walk_interval_ms=10000)
+
+ self.pg_enable_capture()
+ self.pg_if.add_stream(self.fragments_400)
+ self.pg_start()
+
+ packets = self.punt_socket.wait_for_packets(
+ len(self.pkt_infos) - len(dropped_packet_indexes))
+ self.verify_capture(packets, dropped_packet_indexes)
+ self.pg_if.assert_nothing_captured()
+
+ def test_timeout_cleanup(self):
+ """ timeout (cleanup) """
+
+ # whole packets + fragmented packets sans last fragment
+ fragments = [
+ x for (_, frags_400, _, _) in self.pkt_infos
+ for x in frags_400[:-1 if len(frags_400) > 1 else None]
+ ]
+
+ # last fragments for fragmented packets
+ fragments2 = [frags_400[-1]
+ for (_, frags_400, _, _) in self.pkt_infos
+ if len(frags_400) > 1]
+
+ dropped_packet_indexes = set(
+ index for (index, frags_400, _, _) in self.pkt_infos
+ if len(frags_400) > 1)
+
+ self.vapi.ip_reassembly_set(timeout_ms=100, max_reassemblies=1000,
+ expire_walk_interval_ms=50)
+
+ self.pg_enable_capture()
+ self.pg_if.add_stream(fragments)
+ self.pg_start()
+
+ self.sleep(.25, "wait before sending rest of fragments")
+
+ self.pg_if.add_stream(fragments2)
+ self.pg_start()
+ self.sleep(.25, "wait for vpp to process packets")
+
+ packets = self.punt_socket.wait_for_packets(
+ len(self.pkt_infos) - len(dropped_packet_indexes))
+ self.verify_capture(packets, dropped_packet_indexes)
+ self.pg_if.assert_nothing_captured()
+
+ def test_disabled(self):
+ """ reassembly disabled """
+
+ dropped_packet_indexes = set(
+ index for (index, frags_400, _, _) in self.pkt_infos
+ if len(frags_400) > 1)
+
+ self.vapi.ip_reassembly_set(timeout_ms=1000, max_reassemblies=0,
+ expire_walk_interval_ms=10000)
+
+ self.pg_enable_capture()
+ self.pg_if.add_stream(self.fragments_400)
+ self.pg_start()
+
+ packets = self.punt_socket.wait_for_packets(
+ len(self.pkt_infos) - len(dropped_packet_indexes))
+ self.verify_capture(packets, dropped_packet_indexes)
+ self.pg_if.assert_nothing_captured()
+
+
+class TestIPv6Reassembly(VppTestCase):
+ """ IPv6 Reassembly """
+
+ @classmethod
+ def setUpClass(cls):
+ super(TestIPv6Reassembly, cls).setUpClass()
+
+ cls.create_pg_interfaces([0])
+ cls.pg_if = cls.pg0
+
+ # setup all interfaces
+ for i in cls.pg_interfaces:
+ i.admin_up()
+ i.config_ip6()
+ i.resolve_ndp()
+
+ cls.punt_port = 9999
+ cls.punt_socket = VppUDSPuntSocket(cls, cls.punt_port, is_ip4=0)
+
+ # packet sizes
+ cls.packet_sizes = [64, 512, 1518, 9018]
+ cls.padding = " abcdefghijklmn"
+ cls.create_stream(cls.packet_sizes)
+ cls.create_fragments()
+
+ def setUp(self):
+ """ Test setup - force timeout on existing reassemblies """
+ super(TestIPv6Reassembly, self).setUp()
+ self.vapi.ip_reassembly_set(timeout_ms=0, max_reassemblies=1000,
+ expire_walk_interval_ms=10, is_ip6=1)
+ self.sleep(.25)
+ self.vapi.ip_reassembly_set(timeout_ms=1000000, max_reassemblies=1000,
+ expire_walk_interval_ms=10000, is_ip6=1)
+
+ def tearDown(self):
+ super(TestIPv6Reassembly, self).tearDown()
+ self.logger.debug(self.vapi.ppcli("show ip6-reassembly details"))
+
+ @classmethod
+ def create_stream(cls, packet_sizes, packet_count=test_packet_count):
+ """Create input packet stream for defined interface.
+
+ :param list packet_sizes: Required packet sizes.
+ """
+ for i in range(0, packet_count):
+ info = cls.create_packet_info(cls.pg_if, cls.pg_if)
+ payload = cls.info_to_payload(info)
+ p = (Ether(dst=cls.pg_if.local_mac, src=cls.pg_if.remote_mac) /
+ IPv6(src=cls.pg_if.remote_ip6,
+ dst=cls.pg_if.local_ip6) /
+ UDP(sport=1234, dport=cls.punt_port) /
+ Raw(payload))
+ size = packet_sizes[(i // 2) % len(packet_sizes)]
+ cls.extend_packet(p, size, cls.padding)
+ info.data = p
+
+ @classmethod
+ def create_fragments(cls):
+ infos = cls._packet_infos
+ cls.pkt_infos = []
+ for index, info in infos.iteritems():
+ p = info.data
+ # self.logger.debug(ppp("Packet:", p.__class__(str(p))))
+ fragments_400 = fragment_rfc8200(p, info.index, 400)
+ fragments_300 = fragment_rfc8200(p, info.index, 300)
+ cls.pkt_infos.append((index, fragments_400, fragments_300))
+ cls.fragments_400 = [
+ x for _, frags, _ in cls.pkt_infos for x in frags]
+ cls.fragments_300 = [
+ x for _, _, frags in cls.pkt_infos for x in frags]
+ cls.logger.debug("Fragmented %s packets into %s 400-byte fragments, "
+ "and %s 300-byte fragments" %
+ (len(infos), len(cls.fragments_400),
+ len(cls.fragments_300)))
+
+ def verify_capture(self, capture, dropped_packet_indexes=[]):
+ """Verify captured packet strea .
+
+ :param list capture: Captured packet stream.
+ """
+ info = None
+ seen = set()
+ for packet in capture:
+ try:
+ sw_if_index = packet['sw_if_index']
+ punt_action = packet['punt_action']
+ packet = Ether(packet['packet'])
+ self.logger.debug(ppp("Got packet from %s, action %s" %
+ (sw_if_index, punt_action), packet))
+ ip = packet[IPv6]
+ udp = packet[UDP]
+ payload_info = self.payload_to_info(str(packet[Raw]))
+ packet_index = payload_info.index
+ self.assertTrue(
+ packet_index not in dropped_packet_indexes,
+ ppp("Packet received, but should be dropped:", packet))
+ if packet_index in seen:
+ raise Exception(ppp("Duplicate packet received", packet))
+ seen.add(packet_index)
+ self.assertEqual(payload_info.dst, self.pg_if.sw_if_index)
+ info = self._packet_infos[packet_index]
+ self.assertTrue(info is not None)
+ self.assertEqual(packet_index, info.index)
+ saved_packet = info.data
+ self.assertEqual(ip.src, saved_packet[IPv6].src)
+ self.assertEqual(ip.dst, saved_packet[IPv6].dst)
+ self.assertEqual(udp.payload, saved_packet[UDP].payload)
+ except:
+ self.logger.error(ppp("Unexpected or invalid packet:", packet))
+ raise
+ for index in self._packet_infos:
+ self.assertTrue(index in seen or index in dropped_packet_indexes,
+ "Packet with packet_index %d not received" % index)
+
+ def test_reassembly(self):
+ """ basic reassembly """
+
+ self.pg_enable_capture()
+ self.pg_if.add_stream(self.fragments_400)
+ self.pg_start()
+
+ packets = self.punt_socket.wait_for_packets(len(self.pkt_infos))
+ self.verify_capture(packets)
+ self.pg_if.assert_nothing_captured()
+
+ # run it all again to verify correctness
+ self.pg_enable_capture()
+ self.pg_if.add_stream(self.fragments_400)
+ self.pg_start()
+
+ packets = self.punt_socket.wait_for_packets(len(self.pkt_infos))
+ self.verify_capture(packets)
+ self.pg_if.assert_nothing_captured()
+
+ def test_reversed(self):
+ """ reverse order reassembly """
+
+ fragments = list(self.fragments_400)
+ fragments.reverse()
+
+ self.pg_enable_capture()
+ self.pg_if.add_stream(fragments)
+ self.pg_start()
+
+ packets = self.punt_socket.wait_for_packets(len(self.pkt_infos))
+ self.verify_capture(packets)
+ self.pg_if.assert_nothing_captured()
+
+ # run it all again to verify correctness
+ self.pg_enable_capture()
+ self.pg_if.add_stream(fragments)
+ self.pg_start()
+
+ packets = self.punt_socket.wait_for_packets(len(self.pkt_infos))
+ self.verify_capture(packets)
+ self.pg_if.assert_nothing_captured()
+
+ def test_random(self):
+ """ random order reassembly """
+
+ fragments = list(self.fragments_400)
+ shuffle(fragments)
+
+ self.pg_enable_capture()
+ self.pg_if.add_stream(fragments)
+ self.pg_start()
+
+ packets = self.punt_socket.wait_for_packets(len(self.pkt_infos))
+ self.verify_capture(packets)
+ self.pg_if.assert_nothing_captured()
+
+ # run it all again to verify correctness
+ self.pg_enable_capture()
+ self.pg_if.add_stream(fragments)
+ self.pg_start()
+
+ packets = self.punt_socket.wait_for_packets(len(self.pkt_infos))
+ self.verify_capture(packets)
+ self.pg_if.assert_nothing_captured()
+
+ def test_duplicates(self):
+ """ duplicate fragments """
+
+ fragments = [
+ x for (_, frags, _) in self.pkt_infos
+ for x in frags
+ for _ in range(0, min(2, len(frags)))
+ ]
+
+ self.pg_enable_capture()
+ self.pg_if.add_stream(fragments)
+ self.pg_start()
+
+ packets = self.punt_socket.wait_for_packets(len(self.pkt_infos))
+ self.verify_capture(packets)
+ self.pg_if.assert_nothing_captured()
+
+ def test_overlap1(self):
+ """ overlapping fragments case #1 """
+
+ fragments = []
+ for _, frags_400, frags_300 in self.pkt_infos:
+ if len(frags_300) == 1:
+ fragments.extend(frags_400)
+ else:
+ for i, j in zip(frags_300, frags_400):
+ fragments.extend(i)
+ fragments.extend(j)
+
+ dropped_packet_indexes = set(
+ index for (index, _, frags) in self.pkt_infos if len(frags) > 1
+ )
+
+ self.pg_enable_capture()
+ self.pg_if.add_stream(fragments)
+ self.pg_start()
+
+ self.sleep(.1, "wait for vpp to process packets")
+ packets = self.punt_socket.wait_for_packets(
+ len(self.pkt_infos) - len(dropped_packet_indexes))
+ self.verify_capture(packets, dropped_packet_indexes)
+ self.pg_if.assert_nothing_captured()
+
+ def test_overlap2(self):
+ """ overlapping fragments case #2 """
+
+ fragments = []
+ for _, frags_400, frags_30 in self.pkt_infos:
+ if len(frags_400) == 1:
+ fragments.extend(frags_400)
+ else:
+ # care must be taken here so that there are no fragments
+ # received by vpp after reassembly is finished, otherwise
+ # new reassemblies will be started and packet generator will
+ # freak out when it detects unfreed buffers
+ zipped = zip(frags_400, frags_30)
+ for i, j in zipped[:-1]:
+ fragments.extend(i)
+ fragments.extend(j)
+ fragments.append(zipped[-1][0])
+
+ dropped_packet_indexes = set(
+ index for (index, _, frags) in self.pkt_infos if len(frags) > 1
+ )
+
+ self.pg_enable_capture()
+ self.pg_if.add_stream(fragments)
+ self.pg_start()
+
+ self.sleep(.1, "wait for vpp to process packets")
+ packets = self.punt_socket.wait_for_packets(
+ len(self.pkt_infos) - len(dropped_packet_indexes))
+ self.verify_capture(packets, dropped_packet_indexes)
+ self.pg_if.assert_nothing_captured()
+
+ def test_timeout_inline(self):
+ """ timeout (inline) """
+
+ dropped_packet_indexes = set(
+ index for (index, frags, _) in self.pkt_infos if len(frags) > 1
+ )
+
+ self.vapi.ip_reassembly_set(timeout_ms=0, max_reassemblies=1000,
+ expire_walk_interval_ms=10000, is_ip6=1)
+
+ self.pg_enable_capture()
+ self.pg_if.add_stream(self.fragments_400)
+ self.pg_start()
+
+ packets = self.punt_socket.wait_for_packets(
+ len(self.pkt_infos) - len(dropped_packet_indexes))
+ self.verify_capture(packets, dropped_packet_indexes)
+ pkts = self.pg_if.get_capture(
+ expected_count=len(dropped_packet_indexes))
+ for icmp in pkts:
+ self.assertIn(ICMPv6TimeExceeded, icmp)
+ self.assertIn(IPv6ExtHdrFragment, icmp)
+ self.assertIn(icmp[IPv6ExtHdrFragment].id, dropped_packet_indexes)
+ dropped_packet_indexes.remove(icmp[IPv6ExtHdrFragment].id)
+
+ def test_timeout_cleanup(self):
+ """ timeout (cleanup) """
+
+ # whole packets + fragmented packets sans last fragment
+ fragments = [
+ x for (_, frags_400, _) in self.pkt_infos
+ for x in frags_400[:-1 if len(frags_400) > 1 else None]
+ ]
+
+ # last fragments for fragmented packets
+ fragments2 = [frags_400[-1]
+ for (_, frags_400, _) in self.pkt_infos
+ if len(frags_400) > 1]
+
+ dropped_packet_indexes = set(
+ index for (index, frags_400, _) in self.pkt_infos
+ if len(frags_400) > 1)
+
+ self.vapi.ip_reassembly_set(timeout_ms=100, max_reassemblies=1000,
+ expire_walk_interval_ms=50)
+
+ self.vapi.ip_reassembly_set(timeout_ms=100, max_reassemblies=1000,
+ expire_walk_interval_ms=50, is_ip6=1)
+
+ self.pg_enable_capture()
+ self.pg_if.add_stream(fragments)
+ self.pg_start()
+
+ self.sleep(.25, "wait before sending rest of fragments")
+
+ self.pg_if.add_stream(fragments2)
+ self.pg_start()
+ self.sleep(.25, "wait for vpp to process packets")
+
+ packets = self.punt_socket.wait_for_packets(
+ len(self.pkt_infos) - len(dropped_packet_indexes))
+ self.verify_capture(packets, dropped_packet_indexes)
+ pkts = self.pg_if.get_capture(
+ expected_count=len(dropped_packet_indexes))
+ for icmp in pkts:
+ self.assertIn(ICMPv6TimeExceeded, icmp)
+ self.assertIn(IPv6ExtHdrFragment, icmp)
+ self.assertIn(icmp[IPv6ExtHdrFragment].id, dropped_packet_indexes)
+ dropped_packet_indexes.remove(icmp[IPv6ExtHdrFragment].id)
+
+ def test_disabled(self):
+ """ reassembly disabled """
+
+ dropped_packet_indexes = set(
+ index for (index, frags_400, _) in self.pkt_infos
+ if len(frags_400) > 1)
+
+ self.vapi.ip_reassembly_set(timeout_ms=1000, max_reassemblies=0,
+ expire_walk_interval_ms=10000, is_ip6=1)
+
+ self.pg_enable_capture()
+ self.pg_if.add_stream(self.fragments_400)
+ self.pg_start()
+
+ packets = self.punt_socket.wait_for_packets(
+ len(self.pkt_infos) - len(dropped_packet_indexes))
+ self.verify_capture(packets, dropped_packet_indexes)
+ self.pg_if.assert_nothing_captured()
+
+ def test_missing_upper(self):
+ """ missing upper layer """
+ p = (Ether(dst=self.pg_if.local_mac, src=self.pg_if.remote_mac) /
+ IPv6(src=self.pg_if.remote_ip6,
+ dst=self.pg_if.local_ip6) /
+ UDP(sport=1234, dport=self.punt_port) /
+ Raw())
+ self.extend_packet(p, 1000, self.padding)
+ fragments = fragment_rfc8200(p, 1, 500)
+ bad_fragment = p.__class__(str(fragments[1]))
+ bad_fragment[IPv6ExtHdrFragment].nh = 59
+ bad_fragment[IPv6ExtHdrFragment].offset = 0
+ self.pg_enable_capture()
+ self.pg_if.add_stream([bad_fragment])
+ self.pg_start()
+ pkts = self.pg_if.get_capture(expected_count=1)
+ icmp = pkts[0]
+ self.assertIn(ICMPv6ParamProblem, icmp)
+ self.assert_equal(icmp[ICMPv6ParamProblem].code, 3, "ICMP code")
+
+ def test_invalid_frag_size(self):
+ """ fragment size not a multiple of 8 """
+ p = (Ether(dst=self.pg_if.local_mac, src=self.pg_if.remote_mac) /
+ IPv6(src=self.pg_if.remote_ip6,
+ dst=self.pg_if.local_ip6) /
+ UDP(sport=1234, dport=self.punt_port) /
+ Raw())
+ self.extend_packet(p, 1000, self.padding)
+ fragments = fragment_rfc8200(p, 1, 500)
+ bad_fragment = fragments[0]
+ self.extend_packet(bad_fragment, len(bad_fragment) + 5)
+ self.pg_enable_capture()
+ self.pg_if.add_stream([bad_fragment])
+ self.pg_start()
+ pkts = self.pg_if.get_capture(expected_count=1)
+ icmp = pkts[0]
+ self.assertIn(ICMPv6ParamProblem, icmp)
+ self.assert_equal(icmp[ICMPv6ParamProblem].code, 0, "ICMP code")
+
+ def test_invalid_packet_size(self):
+ """ total packet size > 65535 """
+ p = (Ether(dst=self.pg_if.local_mac, src=self.pg_if.remote_mac) /
+ IPv6(src=self.pg_if.remote_ip6,
+ dst=self.pg_if.local_ip6) /
+ UDP(sport=1234, dport=self.punt_port) /
+ Raw())
+ self.extend_packet(p, 1000, self.padding)
+ fragments = fragment_rfc8200(p, 1, 500)
+ bad_fragment = fragments[1]
+ bad_fragment[IPv6ExtHdrFragment].offset = 65500
+ self.pg_enable_capture()
+ self.pg_if.add_stream([bad_fragment])
+ self.pg_start()
+ pkts = self.pg_if.get_capture(expected_count=1)
+ icmp = pkts[0]
+ self.assertIn(ICMPv6ParamProblem, icmp)
+ self.assert_equal(icmp[ICMPv6ParamProblem].code, 0, "ICMP code")
+
+
+class TestFIFReassembly(VppTestCase):
+ """ Fragments in fragments reassembly """
+
+ @classmethod
+ def setUpClass(cls):
+ super(TestFIFReassembly, cls).setUpClass()
+
+ cls.create_pg_interfaces([0])
+ cls.pg_if = cls.pg0
+ cls.pg_if.admin_up()
+ cls.pg_if.config_ip4()
+ cls.pg_if.resolve_arp()
+ cls.pg_if.config_ip6()
+ cls.pg_if.resolve_ndp()
+
+ cls.punt_port = 9999
+ cls.punt4_socket = VppUDSPuntSocket(cls, cls.punt_port)
+ cls.punt6_socket = VppUDSPuntSocket(cls, cls.punt_port, is_ip4=0)
+ cls.packet_sizes = [64, 512, 1518, 9018]
+ cls.padding = " abcdefghijklmn"
+
+ def setUp(self):
+ """ Test setup - force timeout on existing reassemblies """
+ super(TestFIFReassembly, self).setUp()
+ self.vapi.ip_reassembly_set(timeout_ms=0, max_reassemblies=1000,
+ expire_walk_interval_ms=10)
+ self.vapi.ip_reassembly_set(timeout_ms=0, max_reassemblies=1000,
+ expire_walk_interval_ms=10, is_ip6=1)
+ self.sleep(.25)
+ self.vapi.ip_reassembly_set(timeout_ms=1000000, max_reassemblies=1000,
+ expire_walk_interval_ms=10000)
+ self.vapi.ip_reassembly_set(timeout_ms=1000000, max_reassemblies=1000,
+ expire_walk_interval_ms=10000, is_ip6=1)
+
+ def tearDown(self):
+ self.logger.debug(self.vapi.ppcli("show ip4-reassembly details"))
+ self.logger.debug(self.vapi.ppcli("show ip6-reassembly details"))
+ super(TestFIFReassembly, self).tearDown()
+
+ def verify_capture(self, capture, ip_class, dropped_packet_indexes=[]):
+ """Verify captured packet stream.
+
+ :param list capture: Captured packet stream.
+ """
+ info = None
+ seen = set()
+ for packet in capture:
+ try:
+ sw_if_index = packet['sw_if_index']
+ punt_action = packet['punt_action']
+ packet = Ether(packet['packet'])
+ self.logger.debug(ppp("Got packet from %s, action %s" %
+ (sw_if_index, punt_action), packet))
+ ip = packet[ip_class]
+ udp = packet[UDP]
+ payload_info = self.payload_to_info(str(packet[Raw]))
+ packet_index = payload_info.index
+ self.assertTrue(
+ packet_index not in dropped_packet_indexes,
+ ppp("Packet received, but should be dropped:", packet))
+ if packet_index in seen:
+ raise Exception(ppp("Duplicate packet received", packet))
+ seen.add(packet_index)
+ self.assertEqual(payload_info.dst, self.pg_if.sw_if_index)
+ info = self._packet_infos[packet_index]
+ self.assertTrue(info is not None)
+ self.assertEqual(packet_index, info.index)
+ saved_packet = info.data
+ self.assertEqual(ip.src, saved_packet[ip_class].src)
+ self.assertEqual(ip.dst, saved_packet[ip_class].dst)
+ self.assertEqual(udp.payload, saved_packet[UDP].payload)
+ except:
+ self.logger.error(ppp("Unexpected or invalid packet:", packet))
+ raise
+ for index in self._packet_infos:
+ self.assertTrue(index in seen or index in dropped_packet_indexes,
+ "Packet with packet_index %d not received" % index)
+
+ def test_fif4(self):
+ """ Fragments in fragments (4o4) """
+
+ # TODO this should be ideally in setUpClass, but then we hit a bug
+ # with VppIpRoute incorrectly reporting it's present when it's not
+ # so we need to manually remove the vpp config, thus we cannot have
+ # it shared for multiple test cases
+ self.tun_ip4 = "1.1.1.2"
+
+ self.gre4 = VppGreInterface(self, self.pg0.local_ip4, self.tun_ip4)
+ self.gre4.add_vpp_config()
+ self.gre4.admin_up()
+ self.gre4.config_ip4()
+
+ self.route4 = VppIpRoute(self, self.tun_ip4, 32,
+ [VppRoutePath(self.pg0.remote_ip4,
+ self.pg0.sw_if_index)])
+ self.route4.add_vpp_config()
+
+ self.reset_packet_infos()
+ for i in range(test_packet_count):
+ info = self.create_packet_info(self.pg0, self.pg0)
+ payload = self.info_to_payload(info)
+ p = (IP(id=i, src=self.pg0.remote_ip4, dst=self.pg0.local_ip4) /
+ UDP(sport=1234, dport=self.punt_port) /
+ Raw(payload))
+ size = self.packet_sizes[(i // 2) % len(self.packet_sizes)]
+ self.extend_packet(p, size, self.padding)
+ info.data = p
+
+ fragments = [x for _, p in self._packet_infos.iteritems()
+ for x in fragment_rfc791(p.data, 400)]
+
+ encapped_fragments = \
+ [Ether(dst=self.pg0.local_mac, src=self.pg0.remote_mac) /
+ IP(src=self.tun_ip4, dst=self.pg0.local_ip4) /
+ GRE() /
+ p
+ for p in fragments]
+
+ fragmented_encapped_fragments = \
+ [x for p in encapped_fragments
+ for x in fragment_rfc791(p, 200)]
+
+ self.pg0.add_stream(fragmented_encapped_fragments)
+
+ self.pg_enable_capture(self.pg_interfaces)
+ self.pg_start()
+
+ self.pg0.assert_nothing_captured()
+ packets = self.punt4_socket.wait_for_packets(len(self._packet_infos))
+ self.verify_capture(packets, IP)
+
+ # TODO remove gre vpp config by hand until VppIpRoute gets fixed
+ # so that it's query_vpp_config() works as it should
+ self.gre4.remove_vpp_config()
+
+ def test_fif6(self):
+ """ Fragments in fragments (6o6) """
+ # TODO this should be ideally in setUpClass, but then we hit a bug
+ # with VppIpRoute incorrectly reporting it's present when it's not
+ # so we need to manually remove the vpp config, thus we cannot have
+ # it shared for multiple test cases
+ self.tun_ip6 = "1002::1"
+
+ self.gre6 = VppGre6Interface(self, self.pg0.local_ip6, self.tun_ip6)
+ self.gre6.add_vpp_config()
+ self.gre6.admin_up()
+ self.gre6.config_ip6()
+
+ self.route6 = VppIpRoute(self, self.tun_ip6, 128,
+ [VppRoutePath(self.pg0.remote_ip6,
+ self.pg0.sw_if_index,
+ proto=DpoProto.DPO_PROTO_IP6)],
+ is_ip6=1)
+ self.route6.add_vpp_config()
+
+ self.reset_packet_infos()
+ for i in range(test_packet_count):
+ info = self.create_packet_info(self.pg0, self.pg0)
+ payload = self.info_to_payload(info)
+ p = (IPv6(src=self.pg0.remote_ip6, dst=self.pg0.local_ip6) /
+ UDP(sport=1234, dport=self.punt_port) /
+ Raw(payload))
+ size = self.packet_sizes[(i // 2) % len(self.packet_sizes)]
+ self.extend_packet(p, size, self.padding)
+ info.data = p
+
+ fragments = [x for _, i in self._packet_infos.iteritems()
+ for x in fragment_rfc8200(
+ i.data, i.index, 400)]
+
+ encapped_fragments = \
+ [Ether(dst=self.pg0.local_mac, src=self.pg0.remote_mac) /
+ IPv6(src=self.tun_ip6, dst=self.pg0.local_ip6) /
+ GRE() /
+ p
+ for p in fragments]
+
+ fragmented_encapped_fragments = \
+ [x for p in encapped_fragments for x in (
+ fragment_rfc8200(
+ p,
+ 2 * len(self._packet_infos) + p[IPv6ExtHdrFragment].id,
+ 200)
+ if IPv6ExtHdrFragment in p else [p]
+ )
+ ]
+
+ self.pg0.add_stream(fragmented_encapped_fragments)
+
+ self.pg_enable_capture(self.pg_interfaces)
+ self.pg_start()
+
+ self.pg0.assert_nothing_captured()
+ packets = self.punt6_socket.wait_for_packets(len(self._packet_infos))
+ self.verify_capture(packets, IPv6)
+
+ # TODO remove gre vpp config by hand until VppIpRoute gets fixed
+ # so that it's query_vpp_config() works as it should
+ self.gre6.remove_vpp_config()
+
+
+if __name__ == '__main__':
+ unittest.main(testRunner=VppTestRunner)
diff --git a/test/util.py b/test/util.py
index 3e0267a3f4d..512bf9e3b7e 100644
--- a/test/util.py
+++ b/test/util.py
@@ -8,10 +8,10 @@ from scapy.layers.inet6 import in6_mactoifaceid
from scapy.layers.l2 import Ether
from scapy.packet import Raw
-from scapy.layers.inet import IP, UDP, TCP
-from scapy.layers.inet6 import IPv6, ICMPv6Unknown, ICMPv6EchoRequest
-from scapy.packet import Packet
-from socket import inet_pton, AF_INET, AF_INET6
+from scapy.layers.inet import IP
+from scapy.layers.inet6 import IPv6, IPv6ExtHdrFragment, IPv6ExtHdrRouting,\
+ IPv6ExtHdrHopByHop
+from socket import AF_INET6
def ppp(headline, packet):
@@ -37,8 +37,7 @@ def ppc(headline, capture, limit=10):
tail = ""
if limit < len(capture):
tail = "\nPrint limit reached, %s out of %s packets printed" % (
- len(capture), limit)
- limit = len(capture)
+ limit, len(capture))
body = "".join([ppp("Packet #%s:" % count, p)
for count, p in zip(range(0, limit), capture)])
return "%s\n%s%s" % (headline, body, tail)
@@ -174,6 +173,7 @@ class ForeignAddressFactory(object):
class L4_Conn():
""" L4 'connection' tied to two VPP interfaces """
+
def __init__(self, testcase, if1, if2, af, l4proto, port1, port2):
self.testcase = testcase
self.ifs = [None, None]
@@ -189,7 +189,7 @@ class L4_Conn():
def pkt(self, side, l4args={}, payload="x"):
is_ip6 = 1 if self.address_family == AF_INET6 else 0
s0 = side
- s1 = 1-side
+ s1 = 1 - side
src_if = self.ifs[s0]
dst_if = self.ifs[s1]
layer_3 = [IP(src=src_if.remote_ip4, dst=dst_if.remote_ip4),
@@ -208,7 +208,7 @@ class L4_Conn():
l4args['flags'] = flags
self.ifs[side].add_stream(self.pkt(side,
l4args=l4args, payload=payload))
- self.ifs[1-side].enable_capture()
+ self.ifs[1 - side].enable_capture()
self.testcase.pg_start()
def recv(self, side):
@@ -217,15 +217,190 @@ class L4_Conn():
def send_through(self, side, flags=None, payload=""):
self.send(side, flags, payload)
- p = self.recv(1-side)
+ p = self.recv(1 - side)
return p
def send_pingpong(self, side, flags1=None, flags2=None):
p1 = self.send_through(side, flags1)
- p2 = self.send_through(1-side, flags2)
+ p2 = self.send_through(1 - side, flags2)
return [p1, p2]
class L4_CONN_SIDE:
L4_CONN_SIDE_ZERO = 0
L4_CONN_SIDE_ONE = 1
+
+
+class LoggerWrapper(object):
+ def __init__(self, logger=None):
+ self._logger = logger
+
+ def debug(self, *args, **kwargs):
+ if self._logger:
+ self._logger.debug(*args, **kwargs)
+
+ def error(self, *args, **kwargs):
+ if self._logger:
+ self._logger.error(*args, **kwargs)
+
+
+def fragment_rfc791(packet, fragsize, _logger=None):
+ """
+ Fragment an IPv4 packet per RFC 791
+ :param packet: packet to fragment
+ :param fragsize: size at which to fragment
+ :note: IP options are not supported
+ :returns: list of fragments
+ """
+ logger = LoggerWrapper(_logger)
+ logger.debug(ppp("Fragmenting packet:", packet))
+ packet = packet.__class__(str(packet)) # recalculate all values
+ if len(packet[IP].options) > 0:
+ raise Exception("Not implemented")
+ if len(packet) <= fragsize:
+ return [packet]
+
+ pre_ip_len = len(packet) - len(packet[IP])
+ ip_header_len = packet[IP].ihl * 4
+ hex_packet = str(packet)
+ hex_headers = hex_packet[:(pre_ip_len + ip_header_len)]
+ hex_payload = hex_packet[(pre_ip_len + ip_header_len):]
+
+ pkts = []
+ ihl = packet[IP].ihl
+ otl = len(packet[IP])
+ nfb = (fragsize - pre_ip_len - ihl * 4) / 8
+ fo = packet[IP].frag
+
+ p = packet.__class__(hex_headers + hex_payload[:nfb * 8])
+ p[IP].flags = "MF"
+ p[IP].frag = fo
+ p[IP].len = ihl * 4 + nfb * 8
+ del p[IP].chksum
+ pkts.append(p)
+
+ p = packet.__class__(hex_headers + hex_payload[nfb * 8:])
+ p[IP].len = otl - nfb * 8
+ p[IP].frag = fo + nfb
+ del p[IP].chksum
+
+ more_fragments = fragment_rfc791(p, fragsize, _logger)
+ pkts.extend(more_fragments)
+
+ return pkts
+
+
+def fragment_rfc8200(packet, identification, fragsize, _logger=None):
+ """
+ Fragment an IPv6 packet per RFC 8200
+ :param packet: packet to fragment
+ :param fragsize: size at which to fragment
+ :note: IP options are not supported
+ :returns: list of fragments
+ """
+ logger = LoggerWrapper(_logger)
+ packet = packet.__class__(str(packet)) # recalculate all values
+ if len(packet) <= fragsize:
+ return [packet]
+ logger.debug(ppp("Fragmenting packet:", packet))
+ pkts = []
+ counter = 0
+ routing_hdr = None
+ hop_by_hop_hdr = None
+ upper_layer = None
+ seen_ipv6 = False
+ ipv6_nr = -1
+ l = packet.getlayer(counter)
+ while l is not None:
+ if l.__class__ is IPv6:
+ if seen_ipv6:
+ # ignore 2nd IPv6 header and everything below..
+ break
+ ipv6_nr = counter
+ seen_ipv6 = True
+ elif l.__class__ is IPv6ExtHdrFragment:
+ raise Exception("Already fragmented")
+ elif l.__class__ is IPv6ExtHdrRouting:
+ routing_hdr = counter
+ elif l.__class__ is IPv6ExtHdrHopByHop:
+ hop_by_hop_hdr = counter
+ elif seen_ipv6 and not upper_layer and \
+ not l.__class__.__name__.startswith('IPv6ExtHdr'):
+ upper_layer = counter
+ counter = counter + 1
+ l = packet.getlayer(counter)
+
+ logger.debug(
+ "Layers seen: IPv6(#%s), Routing(#%s), HopByHop(#%s), upper(#%s)" %
+ (ipv6_nr, routing_hdr, hop_by_hop_hdr, upper_layer))
+
+ if upper_layer is None:
+ raise Exception("Upper layer header not found in IPv6 packet")
+
+ last_per_fragment_hdr = ipv6_nr
+ if routing_hdr is None:
+ if hop_by_hop_hdr is not None:
+ last_per_fragment_hdr = hop_by_hop_hdr
+ else:
+ last_per_fragment_hdr = routing_hdr
+ logger.debug("Last per-fragment hdr is #%s" % (last_per_fragment_hdr))
+
+ per_fragment_headers = packet.copy()
+ per_fragment_headers[last_per_fragment_hdr].remove_payload()
+ logger.debug(ppp("Per-fragment headers:", per_fragment_headers))
+
+ ext_and_upper_layer = packet.getlayer(last_per_fragment_hdr)[1]
+ hex_payload = str(ext_and_upper_layer)
+ logger.debug("Payload length is %s" % len(hex_payload))
+ logger.debug(ppp("Ext and upper layer:", ext_and_upper_layer))
+
+ fragment_ext_hdr = IPv6ExtHdrFragment()
+ logger.debug(ppp("Fragment header:", fragment_ext_hdr))
+
+ if len(per_fragment_headers) + len(fragment_ext_hdr) +\
+ len(ext_and_upper_layer) - len(ext_and_upper_layer.payload)\
+ > fragsize:
+ raise Exception("Cannot fragment this packet - MTU too small "
+ "(%s, %s, %s, %s, %s)" % (
+ len(per_fragment_headers), len(fragment_ext_hdr),
+ len(ext_and_upper_layer),
+ len(ext_and_upper_layer.payload), fragsize))
+
+ orig_nh = packet[IPv6].nh
+ p = per_fragment_headers
+ del p[IPv6].plen
+ del p[IPv6].nh
+ p = p / fragment_ext_hdr
+ del p[IPv6ExtHdrFragment].nh
+ first_payload_len_nfb = (fragsize - len(p)) / 8
+ p = p / Raw(hex_payload[:first_payload_len_nfb * 8])
+ del p[IPv6].plen
+ p[IPv6ExtHdrFragment].nh = orig_nh
+ p[IPv6ExtHdrFragment].id = identification
+ p[IPv6ExtHdrFragment].offset = 0
+ p[IPv6ExtHdrFragment].m = 1
+ p = p.__class__(str(p))
+ logger.debug(ppp("Fragment %s:" % len(pkts), p))
+ pkts.append(p)
+ offset = first_payload_len_nfb * 8
+ logger.debug("Offset after first fragment: %s" % offset)
+ while len(hex_payload) > offset:
+ p = per_fragment_headers
+ del p[IPv6].plen
+ del p[IPv6].nh
+ p = p / fragment_ext_hdr
+ del p[IPv6ExtHdrFragment].nh
+ l_nfb = (fragsize - len(p)) / 8
+ p = p / Raw(hex_payload[offset:offset + l_nfb * 8])
+ p[IPv6ExtHdrFragment].nh = orig_nh
+ p[IPv6ExtHdrFragment].id = identification
+ p[IPv6ExtHdrFragment].offset = offset / 8
+ p[IPv6ExtHdrFragment].m = 1
+ p = p.__class__(str(p))
+ logger.debug(ppp("Fragment %s:" % len(pkts), p))
+ pkts.append(p)
+ offset = offset + l_nfb * 8
+
+ pkts[-1][IPv6ExtHdrFragment].m = 0 # reset more-flags in last fragment
+
+ return pkts
diff --git a/test/vpp_gre_interface.py b/test/vpp_gre_interface.py
index acfd348d64f..998f6eab32b 100644
--- a/test/vpp_gre_interface.py
+++ b/test/vpp_gre_interface.py
@@ -26,14 +26,21 @@ class VppGreInterface(VppInterface):
is_teb=self.t_is_teb)
self._sw_if_index = r.sw_if_index
self.generate_remote_hosts()
+ self._test.registry.register(self, self._test.logger)
def remove_vpp_config(self):
s = socket.inet_pton(socket.AF_INET, self.t_src)
d = socket.inet_pton(socket.AF_INET, self.t_dst)
self.unconfig()
- r = self.test.vapi.gre_tunnel_add_del(s, d,
- outer_fib_id=self.t_outer_fib,
- is_add=0)
+ self.test.vapi.gre_tunnel_add_del(s, d,
+ outer_fib_id=self.t_outer_fib,
+ is_add=0)
+
+ def __str__(self):
+ return self.object_id()
+
+ def object_id(self):
+ return "gre-%d" % self._sw_if_index
class VppGre6Interface(VppInterface):
@@ -60,12 +67,19 @@ class VppGre6Interface(VppInterface):
is_ip6=1)
self._sw_if_index = r.sw_if_index
self.generate_remote_hosts()
+ self._test.registry.register(self, self._test.logger)
def remove_vpp_config(self):
s = socket.inet_pton(socket.AF_INET6, self.t_src)
d = socket.inet_pton(socket.AF_INET6, self.t_dst)
self.unconfig()
- r = self.test.vapi.gre_tunnel_add_del(s, d,
- outer_fib_id=self.t_outer_fib,
- is_add=0,
- is_ip6=1)
+ self.test.vapi.gre_tunnel_add_del(s, d,
+ outer_fib_id=self.t_outer_fib,
+ is_add=0,
+ is_ip6=1)
+
+ def __str__(self):
+ return self.object_id()
+
+ def object_id(self):
+ return "gre-%d" % self._sw_if_index
diff --git a/test/vpp_interface.py b/test/vpp_interface.py
index b8505ce39d7..8891dde6afe 100644
--- a/test/vpp_interface.py
+++ b/test/vpp_interface.py
@@ -372,3 +372,15 @@ class VppInterface(object):
self.test.vapi.proxy_arp_intfc_enable_disable(
self.sw_if_index,
enable)
+
+ def query_vpp_config(self):
+ dump = self.test.vapi.sw_interface_dump()
+ return self.is_interface_config_in_dump(dump)
+
+ def is_interface_config_in_dump(self, dump):
+ for i in dump:
+ if i.interface_name.rstrip(' \t\r\n\0') == self.name and \
+ i.sw_if_index == self.sw_if_index:
+ return True
+ else:
+ return False
diff --git a/test/vpp_lo_interface.py b/test/vpp_lo_interface.py
index 963123f6404..ae4111bec82 100644
--- a/test/vpp_lo_interface.py
+++ b/test/vpp_lo_interface.py
@@ -19,17 +19,5 @@ class VppLoInterface(VppInterface, VppObject):
def remove_vpp_config(self):
self.test.vapi.delete_loopback(self.sw_if_index)
- def query_vpp_config(self):
- dump = self.test.vapi.sw_interface_dump()
- return self.is_interface_config_in_dump(dump)
-
- def is_interface_config_in_dump(self, dump):
- for i in dump:
- if i.interface_name.rstrip(' \t\r\n\0') == self.name and \
- i.sw_if_index == self.sw_if_index:
- return True
- else:
- return False
-
def object_id(self):
return "loopback-%d" % self._sw_if_index
diff --git a/test/vpp_papi_provider.py b/test/vpp_papi_provider.py
index 0269736cd78..087a14b7fad 100644
--- a/test/vpp_papi_provider.py
+++ b/test/vpp_papi_provider.py
@@ -3102,3 +3102,26 @@ class VppPapiProvider(object):
'ip6_fib_id': ip6_fib_id,
'namespace_id': namespace_id,
'namespace_id_len': len(namespace_id)})
+
+ def punt_socket_register(self, l4_port, pathname, header_version=1,
+ is_ip4=1, l4_protocol=0x11):
+ """ Punt to socket """
+ return self.api(self.papi.punt_socket_register,
+ {'is_ip4': is_ip4,
+ 'l4_protocol': l4_protocol,
+ 'l4_port': l4_port,
+ 'pathname': pathname,
+ 'header_version': header_version})
+
+ def ip_reassembly_set(self, timeout_ms, max_reassemblies,
+ expire_walk_interval_ms, is_ip6=0):
+ """ Set IP reassembly parameters """
+ return self.api(self.papi.ip_reassembly_set,
+ {'is_ip6': is_ip6,
+ 'timeout_ms': timeout_ms,
+ 'expire_walk_interval_ms': expire_walk_interval_ms,
+ 'max_reassemblies': max_reassemblies})
+
+ def ip_reassembly_get(self, is_ip6=0):
+ """ Get IP reassembly parameters """
+ return self.api(self.papi.ip_reassembly_get, {'is_ip6': is_ip6})
diff --git a/test/vpp_punt_socket.py b/test/vpp_punt_socket.py
new file mode 100644
index 00000000000..5004b23396d
--- /dev/null
+++ b/test/vpp_punt_socket.py
@@ -0,0 +1,75 @@
+from socket import socket, AF_UNIX, SOCK_DGRAM
+from select import select
+from time import time
+from struct import unpack, calcsize
+from util import ppc
+from scapy.layers.l2 import Ether
+
+client_uds_socket_name = "client-uds-socket"
+vpp_uds_socket_name = "vpp-uds-socket"
+
+VPP_PUNT_HEADER_FMT = '=Ii'
+VPP_PUNT_HEADER_SIZE = calcsize(VPP_PUNT_HEADER_FMT)
+
+
+class VppPuntAction:
+ PUNT_L2 = 0
+ PUNT_IP4_ROUTED = 1
+ PUNT_IP6_ROUTED = 2
+
+
+class VppUDSPuntSocket(object):
+ def __init__(self, testcase, port, is_ip4=1, l4_protocol=0x11):
+ client_path = '%s/%s-%s-%s' % (testcase.tempdir,
+ client_uds_socket_name,
+ "4" if is_ip4 else "6", port)
+ testcase.vapi.punt_socket_register(
+ port, client_path, is_ip4=is_ip4, l4_protocol=l4_protocol)
+ self.testcase = testcase
+ self.uds = socket(AF_UNIX, SOCK_DGRAM)
+ self.uds.bind(client_path)
+ self.uds.connect(testcase.punt_socket_path)
+
+ def wait_for_packets(self, count, timeout=1):
+ packets = []
+ now = time()
+ deadline = now + timeout
+ while len(packets) < count and now < deadline:
+ r, w, e = select([self.uds], [], [self.uds], deadline - now)
+ if self.uds in r:
+ x = self.uds.recv(1024 * 1024)
+ sw_if_index, punt_action = unpack(
+ VPP_PUNT_HEADER_FMT, x[:VPP_PUNT_HEADER_SIZE])
+ packets.append({'sw_if_index': sw_if_index,
+ 'punt_action': punt_action,
+ 'packet': x[VPP_PUNT_HEADER_SIZE:]})
+
+ if self.uds in e:
+ raise Exception("select() indicates error on UDS socket")
+ now = time()
+
+ if len(packets) != count:
+ raise Exception("Unexpected packet count received, got %s packets,"
+ " expected %s packets" % (len(packets), count))
+ self.testcase.logger.debug(
+ "Got %s packets via punt socket" % len(packets))
+ return packets
+
+ def assert_nothing_captured(self, timeout=.25):
+ packets = []
+ now = time()
+ deadline = now + timeout
+ while now < deadline:
+ r, w, e = select([self.uds], [], [self.uds], deadline - now)
+ if self.uds in r:
+ x = self.uds.recv(1024 * 1024)
+ packets.append(Ether(x[VPP_PUNT_HEADER_SIZE:]))
+ if self.uds in e:
+ raise Exception("select() indicates error on UDS socket")
+ now = time()
+
+ if len(packets) > 0:
+ self.testcase.logger.error(
+ ppc("Unexpected packets captured:", packets))
+ raise Exception("Unexpected packet count received, got %s packets,"
+ " expected no packets" % len(packets))