From 7cd468a3d7dee7d6c92f69a0bb7061ae208ec727 Mon Sep 17 00:00:00 2001 From: Damjan Marion Date: Mon, 19 Dec 2016 23:05:39 +0100 Subject: Reorganize source tree to use single autotools instance Change-Id: I7b51f88292e057c6443b12224486f2d0c9f8ae23 Signed-off-by: Damjan Marion --- src/vnet/classify/README | 180 +++ src/vnet/classify/flow_classify.c | 212 +++ src/vnet/classify/flow_classify.h | 51 + src/vnet/classify/flow_classify_node.c | 338 +++++ src/vnet/classify/input_acl.c | 283 ++++ src/vnet/classify/input_acl.h | 54 + src/vnet/classify/ip_classify.c | 365 +++++ src/vnet/classify/policer_classify.c | 227 +++ src/vnet/classify/policer_classify.h | 55 + src/vnet/classify/vnet_classify.c | 2436 ++++++++++++++++++++++++++++++++ src/vnet/classify/vnet_classify.h | 523 +++++++ 11 files changed, 4724 insertions(+) create mode 100644 src/vnet/classify/README create mode 100644 src/vnet/classify/flow_classify.c create mode 100644 src/vnet/classify/flow_classify.h create mode 100644 src/vnet/classify/flow_classify_node.c create mode 100644 src/vnet/classify/input_acl.c create mode 100644 src/vnet/classify/input_acl.h create mode 100644 src/vnet/classify/ip_classify.c create mode 100644 src/vnet/classify/policer_classify.c create mode 100644 src/vnet/classify/policer_classify.h create mode 100644 src/vnet/classify/vnet_classify.c create mode 100644 src/vnet/classify/vnet_classify.h (limited to 'src/vnet/classify') diff --git a/src/vnet/classify/README b/src/vnet/classify/README new file mode 100644 index 00000000..1ef5ab5a --- /dev/null +++ b/src/vnet/classify/README @@ -0,0 +1,180 @@ +=== vnet classifier theory of operation === + +The vnet classifier trades off simplicity and perf / scale +characteristics. At a certain level, it's a dumb robot. Given an +incoming packet, search an ordered list of (mask, match) tables. If +the classifier finds a matching entry, take the indicated action. If +not, take a last-resort action. + +We use the MMX-unit to match or hash 16 octets at a time. For hardware +backward compatibility, the code does not [currently] use 256-bit +(32-octet) vector instructions. + +Effective use of the classifier centers around building table lists +which "hit" as soon as practicable. In many cases, established +sessions hit in the first table. In this mode of operation, the +classifier easily processes multiple MPPS / core - even with millions +of sessions in the data base. Searching 357 tables on a regular basis +will neatly solve the halting problem. + +==== Basic operation ==== + +The classifier mask-and-match operation proceeds as follows. Given a +starting classifier table index, lay hands on the indicated mask +vector. When building tables, we arrange for the mask to obey +mmx-unit (16-octet) alignment. + +We know that the first octet of packet data starts on a cache-line +boundary. Further, it's reasonably likely that folks won't want to use +the generalized classifier on the L2 header; preferring to decode the +Ethertype manually. That scheme makes it easy to select among ip4 / +ip6 / MPLS, etc. classifier table sets. + +A no-vlan-tag L2 header is 14 octets long. A typical ipv4 header +begins with the octets 0x4500: version=4, header_length=5, DSCP=0, +ECN=0. If one doesn't intend to classify on (DSCP, ECN) - the typical +case - we program the classifier to skip the first 16-octet vector. + +To classify untagged ipv4 packets on source address, we program the +classifier to skip one vector, and mask-and-match one vector. + +The basic match-and-match operation looks like this: + + switch (t->match_n_vectors) + { + case 1: + result = (data[0 + t->skip_n_vectors] & mask[0]) ^ key[0]; + break; + + case 2: + result = (data[0 + t->skip_n_vectors] & mask[0]) ^ key[0]; + result |= (data[1 + t->skip_n_vectors] & mask[1]) ^ key[1]; + break; + + + } + + result_mask = u32x4_zero_byte_mask (result); + if (result_mask == 0xffff) + return (v); + +Net of setup, it costs a couple of clock cycles to mask-and-match 16 +octets. + +At the risk of belaboring an obvious point, the control-plane +'''must''' pay attention to detail. When skipping one (or more) +vectors, masks and matches must reflect that decision. See +.../vnet/vnet/classify/vnet_classify.c:unformat_classify_[mask|match]. Note +that vec_validate (xxx, 13) creates a 14-element vector. + +==== Creating a classifier table ==== + +To create a new classifier table via the control-plane API, send a +"classify_add_del_table" message. The underlying action routine, +vnet_classify_add_del_table(...), is located in +.../vnet/vnet/classify/vnet_classify.c, and has the following +prototype: + + int vnet_classify_add_del_table (vnet_classify_main_t * cm, + u8 * mask, + u32 nbuckets, + u32 memory_size, + u32 skip, + u32 match, + u32 next_table_index, + u32 miss_next_index, + u32 * table_index, + int is_add) + +Pass cm = &vnet_classify_main if calling this routine directly. Mask, +skip(_n_vectors) and match(_n_vectors) are as described above. Mask +need not be aligned, but it must be match*16 octets in length. To +avoid having your head explode, be absolutely certain that '''only''' +the bits you intend to match on are set. + +The classifier uses thread-safe, no-reader-locking-required +bounded-index extensible hashing. Nbuckets is the [fixed] size of the +hash bucket vector. The algorithm works in constant time regardless of +hash collisions, but wastes space when the bucket array is too +small. A good rule of thumb: let nbuckets = approximate number of +entries expected. + +At a signficant cost in complexity, it would be possible to resize the +bucket array dynamically. We have no plans to implement that function. + +Each classifier table has its own clib mheap memory allocation +arena. To pick the memory_size parameter, note that each classifier +table entry needs 16*(1 + match_n_vectors) bytes. Within reason, aim a +bit high. Clib mheap memory uses o/s level virtual memory - not wired +or hugetlb memory - so it's best not to scrimp on size. + +The "next_table_index" parameter is as described: the pool index in +vnet_classify_main.tables of the next table to search. Code ~0 to +indicate the end of the table list. 0 is a valid table index! + +We often create classification tables in reverse order - +last-table-searched to first-table-searched - so we can easily set +this parameter. Of course, one can manually adjust the data structure +after-the-fact. + +Specific classifier client nodes - for example, +.../vnet/vnet/classify/ip_classify.c - interpret the "miss_next_index" +parameter as a vpp graph-node next index. When packet classification +fails to produce a match, ip_classify_inline sends packets to the +indicated disposition. A classifier application might program this +parameter to send packets which don't match an existing session to a +"first-sign-of-life, create-new-session" node. + +Finally, the is_add parameter indicates whether to add or delete the +indicated table. The delete case implicitly terminates all sessions +with extreme prejudice, by freeing the specified clib mheap. + +==== Creating a classifier session ==== + +To create a new classifier session via the control-plane API, send a +"classify_add_del_session" message. The underlying action routine, +vnet_classify_add_del_session(...), is located in +.../vnet/vnet/classify/vnet_classify.c, and has the following +prototype: + +int vnet_classify_add_del_session (vnet_classify_main_t * cm, + u32 table_index, + u8 * match, + u32 hit_next_index, + u32 opaque_index, + i32 advance, + int is_add) + +Pass cm = &vnet_classify_main if calling this routine directly. Table +index specifies the table which receives the new session / contains +the session to delete depending on is_add. + +Match is the key for the indicated session. It need not be aligned, +but it must be table->match_n_vectors*16 octets in length. As a +courtesy, vnet_classify_add_del_session applies the table's mask to +the stored key-value. In this way, one can create a session by passing +unmasked (packet_data + offset) as the "match" parameter, and end up +with unconfusing session keys. + +Specific classifier client nodes - for example, +.../vnet/vnet/classify/ip_classify.c - interpret the per-session +hit_next_index parameter as a vpp graph-node next index. When packet +classification produces a match, ip_classify_inline sends packets to +the indicated disposition. + +ip4/6_classify place the per-session opaque_index parameter into +vnet_buffer(b)->l2_classify.opaque_index; a slight misnomer, but +anyhow classifier applications can send session-hit packets to +specific graph nodes, with useful values in buffer metadata. Depending +on the required semantics, we send known-session traffic to a certain +node, with e.g. a session pool index in buffer metadata. It's totally +up to the control-plane and the specific use-case. + +Finally, nodes such as ip4/6-classify apply the advance parameter as a +[signed!] argument to vlib_buffer_advance(...); to "consume" a +networking layer. Example: if we classify incoming tunneled IP packets +by (inner) source/dest address and source/dest port, we might choose +to decapsulate and reencapsulate the inner packet. In such a case, +program the advance parameter to perform the tunnel decapsulation, and +program next_index to send traffic to a node which uses +e.g. opaque_index to output traffic on a specific tunnel interface. diff --git a/src/vnet/classify/flow_classify.c b/src/vnet/classify/flow_classify.c new file mode 100644 index 00000000..0a624204 --- /dev/null +++ b/src/vnet/classify/flow_classify.c @@ -0,0 +1,212 @@ +/* + * Copyright (c) 2016 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ +#include + +static void +vnet_flow_classify_feature_enable (vlib_main_t * vnm, + flow_classify_main_t * fcm, + u32 sw_if_index, + flow_classify_table_id_t tid, + int feature_enable) +{ + vnet_feature_config_main_t *vfcm; + u8 arc; + + if (tid == FLOW_CLASSIFY_TABLE_IP4) + { + vnet_feature_enable_disable ("ip4-unicast", "ip4-flow-classify", + sw_if_index, feature_enable, 0, 0); + arc = vnet_get_feature_arc_index ("ip4-unicast"); + } + else + { + vnet_feature_enable_disable ("ip6-unicast", "ip6-flow-classify", + sw_if_index, feature_enable, 0, 0); + arc = vnet_get_feature_arc_index ("ip6-unicast"); + } + + vfcm = vnet_get_feature_arc_config_main (arc); + fcm->vnet_config_main[tid] = &vfcm->config_main; +} + +int vnet_set_flow_classify_intfc (vlib_main_t * vm, u32 sw_if_index, + u32 ip4_table_index, u32 ip6_table_index, + u32 is_add) +{ + flow_classify_main_t * fcm = &flow_classify_main; + vnet_classify_main_t * vcm = fcm->vnet_classify_main; + u32 pct[FLOW_CLASSIFY_N_TABLES] = {ip4_table_index, ip6_table_index}; + u32 ti; + + /* Assume that we've validated sw_if_index in the API layer */ + + for (ti = 0; ti < FLOW_CLASSIFY_N_TABLES; ti++) + { + if (pct[ti] == ~0) + continue; + + if (pool_is_free_index (vcm->tables, pct[ti])) + return VNET_API_ERROR_NO_SUCH_TABLE; + + vec_validate_init_empty + (fcm->classify_table_index_by_sw_if_index[ti], sw_if_index, ~0); + + /* Reject any DEL operation with wrong sw_if_index */ + if (!is_add && + (pct[ti] != fcm->classify_table_index_by_sw_if_index[ti][sw_if_index])) + { + clib_warning ("Non-existent intf_idx=%d with table_index=%d for delete", + sw_if_index, pct[ti]); + return VNET_API_ERROR_NO_SUCH_TABLE; + } + + /* Return ok on ADD operaton if feature is already enabled */ + if (is_add && + fcm->classify_table_index_by_sw_if_index[ti][sw_if_index] != ~0) + return 0; + + vnet_flow_classify_feature_enable (vm, fcm, sw_if_index, ti, is_add); + + if (is_add) + fcm->classify_table_index_by_sw_if_index[ti][sw_if_index] = pct[ti]; + else + fcm->classify_table_index_by_sw_if_index[ti][sw_if_index] = ~0; + } + + + return 0; +} + +static clib_error_t * +set_flow_classify_command_fn (vlib_main_t * vm, + unformat_input_t * input, + vlib_cli_command_t * cmd) +{ + vnet_main_t * vnm = vnet_get_main(); + u32 sw_if_index = ~0; + u32 ip4_table_index = ~0; + u32 ip6_table_index = ~0; + u32 is_add = 1; + u32 idx_cnt = 0; + int rv; + + while (unformat_check_input (input) != UNFORMAT_END_OF_INPUT) + { + if (unformat (input, "interface %U", unformat_vnet_sw_interface, + vnm, &sw_if_index)) + ; + else if (unformat (input, "ip4-table %d", &ip4_table_index)) + idx_cnt++; + else if (unformat (input, "ip6-table %d", &ip6_table_index)) + idx_cnt++; + else if (unformat (input, "del")) + is_add = 0; + else + break; + } + + if (sw_if_index == ~0) + return clib_error_return (0, "Interface must be specified."); + + if (!idx_cnt) + return clib_error_return (0, "Table index should be specified."); + + if (idx_cnt > 1) + return clib_error_return (0, "Only one table index per API is allowed."); + + rv = vnet_set_flow_classify_intfc(vm, sw_if_index, ip4_table_index, + ip6_table_index, is_add); + + switch (rv) + { + case 0: + break; + + case VNET_API_ERROR_NO_MATCHING_INTERFACE: + return clib_error_return (0, "No such interface"); + + case VNET_API_ERROR_NO_SUCH_ENTRY: + return clib_error_return (0, "No such classifier table"); + } + return 0; +} + +VLIB_CLI_COMMAND (set_input_acl_command, static) = { + .path = "set flow classify", + .short_help = + "set flow classify interface [ip4-table ]\n" + " [ip6-table ] [del]", + .function = set_flow_classify_command_fn, +}; + +static uword +unformat_table_type (unformat_input_t * input, va_list * va) +{ + u32 * r = va_arg (*va, u32 *); + u32 tid; + + if (unformat (input, "ip4")) + tid = FLOW_CLASSIFY_TABLE_IP4; + else if (unformat (input, "ip6")) + tid = FLOW_CLASSIFY_TABLE_IP6; + else + return 0; + + *r = tid; + return 1; +} +static clib_error_t * +show_flow_classify_command_fn (vlib_main_t * vm, + unformat_input_t * input, + vlib_cli_command_t * cmd) +{ + flow_classify_main_t * fcm = &flow_classify_main; + u32 type = FLOW_CLASSIFY_N_TABLES; + u32 * vec_tbl; + int i; + + if (unformat (input, "type %U", unformat_table_type, &type)) + ; + else + return clib_error_return (0, "Type must be specified.");; + + if (type == FLOW_CLASSIFY_N_TABLES) + return clib_error_return (0, "Invalid table type."); + + vec_tbl = fcm->classify_table_index_by_sw_if_index[type]; + + if (vec_len(vec_tbl)) + vlib_cli_output (vm, "%10s%20s\t\t%s", "Intfc idx", "Classify table", + "Interface name"); + else + vlib_cli_output (vm, "No tables configured."); + + for (i = 0; i < vec_len (vec_tbl); i++) + { + if (vec_elt(vec_tbl, i) == ~0) + continue; + + vlib_cli_output (vm, "%10d%20d\t\t%U", i, vec_elt(vec_tbl, i), + format_vnet_sw_if_index_name, fcm->vnet_main, i); + } + + return 0; +} + +VLIB_CLI_COMMAND (show_flow_classify_command, static) = { + .path = "show classify flow", + .short_help = "show classify flow type [ip4|ip6]", + .function = show_flow_classify_command_fn, +}; diff --git a/src/vnet/classify/flow_classify.h b/src/vnet/classify/flow_classify.h new file mode 100644 index 00000000..3ae04cd7 --- /dev/null +++ b/src/vnet/classify/flow_classify.h @@ -0,0 +1,51 @@ +/* + * Copyright (c) 2016 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +#ifndef __included_vnet_flow_classify_h__ +#define __included_vnet_flow_classify_h__ + +#include +#include +#include + +typedef enum { + FLOW_CLASSIFY_TABLE_IP4, + FLOW_CLASSIFY_TABLE_IP6, + FLOW_CLASSIFY_N_TABLES, +} flow_classify_table_id_t; + +typedef enum { + FLOW_CLASSIFY_NEXT_INDEX_DROP, + FLOW_CLASSIFY_NEXT_INDEX_N_NEXT, +} flow_classify_next_index_t; + +typedef struct { + /* Classifier table vectors */ + u32 * classify_table_index_by_sw_if_index [FLOW_CLASSIFY_N_TABLES]; + + /* Convenience variables */ + vlib_main_t * vlib_main; + vnet_main_t * vnet_main; + vnet_classify_main_t * vnet_classify_main; + vnet_config_main_t * vnet_config_main [FLOW_CLASSIFY_N_TABLES]; +} flow_classify_main_t; + +flow_classify_main_t flow_classify_main; + +int vnet_set_flow_classify_intfc (vlib_main_t * vm, u32 sw_if_index, + u32 ip4_table_index, u32 ip6_table_index, + u32 is_add); + +#endif /* __included_vnet_flow_classify_h__ */ diff --git a/src/vnet/classify/flow_classify_node.c b/src/vnet/classify/flow_classify_node.c new file mode 100644 index 00000000..d3261d33 --- /dev/null +++ b/src/vnet/classify/flow_classify_node.c @@ -0,0 +1,338 @@ +/* + * Copyright (c) 2016 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +#include + +#include +#include +#include +#include +#include + +typedef struct { + u32 sw_if_index; + u32 next_index; + u32 table_index; + u32 offset; +} flow_classify_trace_t; + +static u8 * +format_flow_classify_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 *); + flow_classify_trace_t * t = va_arg (*args, flow_classify_trace_t *); + + s = format (s, "FLOW_CLASSIFY: sw_if_index %d next %d table %d offset %d", + t->sw_if_index, t->next_index, t->table_index, t->offset); + return s; +} + +#define foreach_flow_classify_error \ +_(MISS, "Flow classify misses") \ +_(HIT, "Flow classify hits") \ +_(CHAIN_HIT, "Flow classify hits after chain walk") \ +_(DROP, "Flow classify action drop") + +typedef enum { +#define _(sym,str) FLOW_CLASSIFY_ERROR_##sym, + foreach_flow_classify_error +#undef _ + FLOW_CLASSIFY_N_ERROR, +} flow_classify_error_t; + +static char * flow_classify_error_strings[] = { +#define _(sym,string) string, + foreach_flow_classify_error +#undef _ +}; + +static inline uword +flow_classify_inline (vlib_main_t * vm, + vlib_node_runtime_t * node, + vlib_frame_t * frame, + flow_classify_table_id_t tid) +{ + u32 n_left_from, * from, * to_next; + flow_classify_next_index_t next_index; + flow_classify_main_t * fcm = &flow_classify_main; + vnet_classify_main_t * vcm = fcm->vnet_classify_main; + f64 now = vlib_time_now (vm); + u32 hits = 0; + u32 misses = 0; + u32 chain_hits = 0; + u32 drop = 0; + + from = vlib_frame_vector_args (frame); + n_left_from = frame->n_vectors; + + /* First pass: compute hashes */ + while (n_left_from > 2) + { + vlib_buffer_t * b0, * b1; + u32 bi0, bi1; + u8 * h0, * h1; + u32 sw_if_index0, sw_if_index1; + u32 table_index0, table_index1; + vnet_classify_table_t * t0, * t1; + + /* Prefetch next iteration */ + { + vlib_buffer_t * p1, * p2; + + p1 = vlib_get_buffer (vm, from[1]); + p2 = vlib_get_buffer (vm, from[2]); + + vlib_prefetch_buffer_header (p1, STORE); + CLIB_PREFETCH (p1->data, CLIB_CACHE_LINE_BYTES, STORE); + vlib_prefetch_buffer_header (p2, STORE); + CLIB_PREFETCH (p2->data, CLIB_CACHE_LINE_BYTES, STORE); + } + + bi0 = from[0]; + b0 = vlib_get_buffer (vm, bi0); + h0 = b0->data; + + bi1 = from[1]; + b1 = vlib_get_buffer (vm, bi1); + h1 = b1->data; + + sw_if_index0 = vnet_buffer (b0)->sw_if_index[VLIB_RX]; + table_index0 = fcm->classify_table_index_by_sw_if_index[tid][sw_if_index0]; + + sw_if_index1 = vnet_buffer (b1)->sw_if_index[VLIB_RX]; + table_index1 = fcm->classify_table_index_by_sw_if_index[tid][sw_if_index1]; + + t0 = pool_elt_at_index (vcm->tables, table_index0); + + t1 = pool_elt_at_index (vcm->tables, table_index1); + + vnet_buffer(b0)->l2_classify.hash = + vnet_classify_hash_packet (t0, (u8 *) h0); + + vnet_classify_prefetch_bucket (t0, vnet_buffer(b0)->l2_classify.hash); + + vnet_buffer(b1)->l2_classify.hash = + vnet_classify_hash_packet (t1, (u8 *) h1); + + vnet_classify_prefetch_bucket (t1, vnet_buffer(b1)->l2_classify.hash); + + vnet_buffer(b0)->l2_classify.table_index = table_index0; + + vnet_buffer(b1)->l2_classify.table_index = table_index1; + + from += 2; + n_left_from -= 2; + } + + while (n_left_from > 0) + { + vlib_buffer_t * b0; + u32 bi0; + u8 * h0; + u32 sw_if_index0; + u32 table_index0; + vnet_classify_table_t * t0; + + bi0 = from[0]; + b0 = vlib_get_buffer (vm, bi0); + h0 = b0->data; + + sw_if_index0 = vnet_buffer (b0)->sw_if_index[VLIB_RX]; + table_index0 = fcm->classify_table_index_by_sw_if_index[tid][sw_if_index0]; + + t0 = pool_elt_at_index (vcm->tables, table_index0); + vnet_buffer(b0)->l2_classify.hash = + vnet_classify_hash_packet (t0, (u8 *) h0); + + vnet_buffer(b0)->l2_classify.table_index = table_index0; + vnet_classify_prefetch_bucket (t0, vnet_buffer(b0)->l2_classify.hash); + + from++; + n_left_from--; + } + + next_index = node->cached_next_index; + from = vlib_frame_vector_args (frame); + n_left_from = frame->n_vectors; + + while (n_left_from > 0) + { + u32 n_left_to_next; + + vlib_get_next_frame (vm, node, next_index, to_next, n_left_to_next); + + /* Not enough load/store slots to dual loop... */ + while (n_left_from > 0 && n_left_to_next > 0) + { + u32 bi0; + vlib_buffer_t * b0; + u32 next0 = FLOW_CLASSIFY_NEXT_INDEX_DROP; + u32 table_index0; + vnet_classify_table_t * t0; + vnet_classify_entry_t * e0; + u64 hash0; + u8 * h0; + + /* Stride 3 seems to work best */ + if (PREDICT_TRUE (n_left_from > 3)) + { + vlib_buffer_t * p1 = vlib_get_buffer(vm, from[3]); + vnet_classify_table_t * tp1; + u32 table_index1; + u64 phash1; + + table_index1 = vnet_buffer(p1)->l2_classify.table_index; + + if (PREDICT_TRUE (table_index1 != ~0)) + { + tp1 = pool_elt_at_index (vcm->tables, table_index1); + phash1 = vnet_buffer(p1)->l2_classify.hash; + vnet_classify_prefetch_entry (tp1, phash1); + } + } + + /* Speculatively enqueue b0 to the current next frame */ + bi0 = from[0]; + to_next[0] = bi0; + from += 1; + to_next += 1; + n_left_from -= 1; + n_left_to_next -= 1; + + b0 = vlib_get_buffer (vm, bi0); + h0 = b0->data; + table_index0 = vnet_buffer(b0)->l2_classify.table_index; + e0 = 0; + t0 = 0; + + vnet_get_config_data (fcm->vnet_config_main[tid], + &b0->current_config_index, + &next0, + /* # bytes of config data */ 0); + + if (PREDICT_TRUE(table_index0 != ~0)) + { + hash0 = vnet_buffer(b0)->l2_classify.hash; + t0 = pool_elt_at_index (vcm->tables, table_index0); + e0 = vnet_classify_find_entry (t0, (u8 *) h0, hash0, now); + if (e0) + { + hits++; + } + else + { + misses++; + vnet_classify_add_del_session (vcm, table_index0, + h0, ~0, 0, 0, 0, 0, 1); + /* increment counter */ + vnet_classify_find_entry (t0, (u8 *) h0, hash0, now); + } + } + if (PREDICT_FALSE((node->flags & VLIB_NODE_FLAG_TRACE) + && (b0->flags & VLIB_BUFFER_IS_TRACED))) + { + flow_classify_trace_t * t = + vlib_add_trace (vm, node, b0, sizeof (*t)); + t->sw_if_index = vnet_buffer(b0)->sw_if_index[VLIB_RX]; + t->next_index = next0; + t->table_index = t0 ? t0 - vcm->tables : ~0; + t->offset = (t0 && e0) ? vnet_classify_get_offset (t0, e0): ~0; + } + + /* Verify speculative enqueue, maybe switch current next frame */ + vlib_validate_buffer_enqueue_x1 (vm, node, next_index, to_next, + n_left_to_next, bi0, next0); + } + + vlib_put_next_frame (vm, node, next_index, n_left_to_next); + } + + vlib_node_increment_counter (vm, node->node_index, + FLOW_CLASSIFY_ERROR_MISS, + misses); + vlib_node_increment_counter (vm, node->node_index, + FLOW_CLASSIFY_ERROR_HIT, + hits); + vlib_node_increment_counter (vm, node->node_index, + FLOW_CLASSIFY_ERROR_CHAIN_HIT, + chain_hits); + vlib_node_increment_counter (vm, node->node_index, + FLOW_CLASSIFY_ERROR_DROP, + drop); + + return frame->n_vectors; +} + +static uword +ip4_flow_classify (vlib_main_t * vm, + vlib_node_runtime_t * node, + vlib_frame_t * frame) +{ + return flow_classify_inline(vm, node, frame, FLOW_CLASSIFY_TABLE_IP4); +} + +VLIB_REGISTER_NODE (ip4_flow_classify_node) = { + .function = ip4_flow_classify, + .name = "ip4-flow-classify", + .vector_size = sizeof (u32), + .format_trace = format_flow_classify_trace, + .n_errors = ARRAY_LEN(flow_classify_error_strings), + .error_strings = flow_classify_error_strings, + .n_next_nodes = FLOW_CLASSIFY_NEXT_INDEX_N_NEXT, + .next_nodes = { + [FLOW_CLASSIFY_NEXT_INDEX_DROP] = "error-drop", + }, +}; + +VLIB_NODE_FUNCTION_MULTIARCH (ip4_flow_classify_node, ip4_flow_classify); + +static uword +ip6_flow_classify (vlib_main_t * vm, + vlib_node_runtime_t * node, + vlib_frame_t * frame) +{ + return flow_classify_inline(vm, node, frame, FLOW_CLASSIFY_TABLE_IP6); +} + +VLIB_REGISTER_NODE (ip6_flow_classify_node) = { + .function = ip6_flow_classify, + .name = "ip6-flow-classify", + .vector_size = sizeof (u32), + .format_trace = format_flow_classify_trace, + .n_errors = ARRAY_LEN(flow_classify_error_strings), + .error_strings = flow_classify_error_strings, + .n_next_nodes = FLOW_CLASSIFY_NEXT_INDEX_N_NEXT, + .next_nodes = { + [FLOW_CLASSIFY_NEXT_INDEX_DROP] = "error-drop", + }, +}; + +VLIB_NODE_FUNCTION_MULTIARCH (ip6_flow_classify_node, ip6_flow_classify); + + +static clib_error_t * +flow_classify_init (vlib_main_t *vm) +{ + flow_classify_main_t * fcm = &flow_classify_main; + + fcm->vlib_main = vm; + fcm->vnet_main = vnet_get_main(); + fcm->vnet_classify_main = &vnet_classify_main; + + return 0; +} + +VLIB_INIT_FUNCTION (flow_classify_init); diff --git a/src/vnet/classify/input_acl.c b/src/vnet/classify/input_acl.c new file mode 100644 index 00000000..c446f2d6 --- /dev/null +++ b/src/vnet/classify/input_acl.c @@ -0,0 +1,283 @@ +/* + * Copyright (c) 2015 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ +#include +#include +#include + +input_acl_main_t input_acl_main; + +static int +vnet_inacl_ip_feature_enable (vlib_main_t * vnm, + input_acl_main_t *am, + u32 sw_if_index, + input_acl_table_id_t tid, + int feature_enable) +{ + + if (tid == INPUT_ACL_TABLE_L2) + { + l2input_intf_bitmap_enable (sw_if_index, L2INPUT_FEAT_ACL, + feature_enable); + } + else + { /* IP[46] */ + vnet_feature_config_main_t *fcm; + u8 arc; + + if (tid == INPUT_ACL_TABLE_IP4) + { + vnet_feature_enable_disable ("ip4-unicast", "ip4-inacl", + sw_if_index, feature_enable, 0, 0); + arc = vnet_get_feature_arc_index ("ip4-unicast"); + } + else + { + vnet_feature_enable_disable ("ip6-unicast", "ip6-inacl", + sw_if_index, feature_enable, 0, 0); + arc = vnet_get_feature_arc_index ("ip6-unicast"); + } + + fcm = vnet_get_feature_arc_config_main (arc); + am->vnet_config_main[tid] = &fcm->config_main; + } + + return 0; +} + +int vnet_set_input_acl_intfc (vlib_main_t * vm, u32 sw_if_index, + u32 ip4_table_index, + u32 ip6_table_index, + u32 l2_table_index, u32 is_add) +{ + input_acl_main_t * am = &input_acl_main; + vnet_classify_main_t * vcm = am->vnet_classify_main; + u32 acl[INPUT_ACL_N_TABLES] = {ip4_table_index, ip6_table_index, + l2_table_index}; + u32 ti; + + /* Assume that we've validated sw_if_index in the API layer */ + + for (ti = 0; ti < INPUT_ACL_N_TABLES; ti++) + { + if (acl[ti] == ~0) + continue; + + if (pool_is_free_index (vcm->tables, acl[ti])) + return VNET_API_ERROR_NO_SUCH_TABLE; + + vec_validate_init_empty + (am->classify_table_index_by_sw_if_index[ti], sw_if_index, ~0); + + /* Reject any DEL operation with wrong sw_if_index */ + if (!is_add && + (acl[ti] != am->classify_table_index_by_sw_if_index[ti][sw_if_index])) + { + clib_warning ("Non-existent intf_idx=%d with table_index=%d for delete", + sw_if_index, acl[ti]); + return VNET_API_ERROR_NO_SUCH_TABLE; + } + + /* Return ok on ADD operaton if feature is already enabled */ + if (is_add && + am->classify_table_index_by_sw_if_index[ti][sw_if_index] != ~0) + return 0; + + vnet_inacl_ip_feature_enable (vm, am, sw_if_index, ti, is_add); + + if (is_add) + am->classify_table_index_by_sw_if_index[ti][sw_if_index] = acl[ti]; + else + am->classify_table_index_by_sw_if_index[ti][sw_if_index] = ~0; + } + + return 0; +} + +static clib_error_t * +set_input_acl_command_fn (vlib_main_t * vm, + unformat_input_t * input, + vlib_cli_command_t * cmd) +{ + vnet_main_t * vnm = vnet_get_main(); + u32 sw_if_index = ~0; + u32 ip4_table_index = ~0; + u32 ip6_table_index = ~0; + u32 l2_table_index = ~0; + u32 is_add = 1; + u32 idx_cnt = 0; + int rv; + + while (unformat_check_input (input) != UNFORMAT_END_OF_INPUT) + { + if (unformat (input, "intfc %U", unformat_vnet_sw_interface, + vnm, &sw_if_index)) + ; + else if (unformat (input, "ip4-table %d", &ip4_table_index)) + idx_cnt++; + else if (unformat (input, "ip6-table %d", &ip6_table_index)) + idx_cnt++; + else if (unformat (input, "l2-table %d", &l2_table_index)) + idx_cnt++; + else if (unformat (input, "del")) + is_add = 0; + else + break; + } + + if (sw_if_index == ~0) + return clib_error_return (0, "Interface must be specified."); + + if (!idx_cnt) + return clib_error_return (0, "Table index should be specified."); + + if (idx_cnt > 1) + return clib_error_return (0, "Only one table index per API is allowed."); + + rv = vnet_set_input_acl_intfc (vm, sw_if_index, ip4_table_index, + ip6_table_index, l2_table_index, is_add); + + switch (rv) + { + case 0: + break; + + case VNET_API_ERROR_NO_MATCHING_INTERFACE: + return clib_error_return (0, "No such interface"); + + case VNET_API_ERROR_NO_SUCH_ENTRY: + return clib_error_return (0, "No such classifier table"); + } + return 0; +} + +/* + * Configure interface to enable/disble input ACL feature: + * intfc - interface name to be configured as input ACL + * Ip4-table [del] - enable/disable IP4 input ACL + * Ip6-table [del] - enable/disable IP6 input ACL + * l2-table [del] - enable/disable Layer2 input ACL + * + * Note: Only one table index per API call is allowed. + * + */ +VLIB_CLI_COMMAND (set_input_acl_command, static) = { + .path = "set interface input acl", + .short_help = + "set interface input acl intfc [ip4-table ]\n" + " [ip6-table ] [l2-table ] [del]", + .function = set_input_acl_command_fn, +}; + +clib_error_t *input_acl_init (vlib_main_t *vm) +{ + input_acl_main_t * am = &input_acl_main; + clib_error_t * error = 0; + + if ((error = vlib_call_init_function (vm, ip_inacl_init))) + return error; + + am->vlib_main = vm; + am->vnet_main = vnet_get_main(); + am->vnet_classify_main = &vnet_classify_main; + + return 0; +} + +VLIB_INIT_FUNCTION (input_acl_init); + +uword unformat_acl_type (unformat_input_t * input, va_list * args) +{ + u32 * acl_type = va_arg (*args, u32 *); + u32 tid = INPUT_ACL_N_TABLES; + + while (unformat_check_input (input) != UNFORMAT_END_OF_INPUT) { + if (unformat (input, "ip4")) + tid = INPUT_ACL_TABLE_IP4; + else if (unformat (input, "ip6")) + tid = INPUT_ACL_TABLE_IP6; + else if (unformat (input, "l2")) + tid = INPUT_ACL_TABLE_L2; + else + break; + } + + *acl_type = tid; + return 1; +} + +u8 * format_vnet_inacl_info (u8 * s, va_list * va) +{ + input_acl_main_t * am = va_arg (*va, input_acl_main_t *); + int sw_if_idx = va_arg (*va, int); + u32 tid = va_arg (*va, u32); + + if (tid == ~0) + { + s = format (s, "%10s%20s\t\t%s", "Intfc idx", "Classify table", + "Interface name"); + return s; + } + + s = format (s, "%10d%20d\t\t%U", sw_if_idx, tid, + format_vnet_sw_if_index_name, am->vnet_main, sw_if_idx); + + return s; +} + +static clib_error_t * +show_inacl_command_fn (vlib_main_t * vm, + unformat_input_t * input, + vlib_cli_command_t * cmd) +{ + input_acl_main_t * am = &input_acl_main; + u32 type = INPUT_ACL_N_TABLES; + int i; + u32 * vec_tbl; + + while (unformat_check_input (input) != UNFORMAT_END_OF_INPUT) + { + if (unformat (input, "type %U", unformat_acl_type, &type)) + ; + else + break; + } + + if (type == INPUT_ACL_N_TABLES) + return clib_error_return (0, "Invalid input ACL table type."); + + vec_tbl = am->classify_table_index_by_sw_if_index[type]; + + if (vec_len(vec_tbl)) + vlib_cli_output (vm, "%U", format_vnet_inacl_info, am, ~0 /* hdr */, ~0); + else + vlib_cli_output (vm, "No input ACL tables configured"); + + for (i = 0; i < vec_len (vec_tbl); i++) + { + if (vec_elt(vec_tbl, i) == ~0) + continue; + + vlib_cli_output (vm, "%U", format_vnet_inacl_info, + am, i, vec_elt(vec_tbl, i)); + } + + return 0; +} + +VLIB_CLI_COMMAND (show_inacl_command, static) = { + .path = "show inacl", + .short_help = "show inacl type [ip4|ip6|l2]", + .function = show_inacl_command_fn, +}; diff --git a/src/vnet/classify/input_acl.h b/src/vnet/classify/input_acl.h new file mode 100644 index 00000000..7ffc189f --- /dev/null +++ b/src/vnet/classify/input_acl.h @@ -0,0 +1,54 @@ +/* + * Copyright (c) 2015 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +#ifndef __included_vnet_input_acl_h__ +#define __included_vnet_input_acl_h__ + +#include +#include +#include + +typedef enum { + INPUT_ACL_TABLE_IP4, + INPUT_ACL_TABLE_IP6, + INPUT_ACL_TABLE_L2, + INPUT_ACL_N_TABLES, +} input_acl_table_id_t; + +typedef enum { + ACL_NEXT_INDEX_DENY, + ACL_NEXT_INDEX_N_NEXT, +} acl_next_index_t; + +typedef struct { + + /* classifier table vectors */ + u32 * classify_table_index_by_sw_if_index [INPUT_ACL_N_TABLES]; + + /* convenience variables */ + vlib_main_t * vlib_main; + vnet_main_t * vnet_main; + vnet_classify_main_t * vnet_classify_main; + vnet_config_main_t * vnet_config_main [INPUT_ACL_N_TABLES]; +} input_acl_main_t; + +extern input_acl_main_t input_acl_main; + +int vnet_set_input_acl_intfc (vlib_main_t * vm, u32 sw_if_index, + u32 ip4_table_index, + u32 ip6_table_index, + u32 l2_table_index, u32 is_add); + +#endif /* __included_vnet_input_acl_h__ */ diff --git a/src/vnet/classify/ip_classify.c b/src/vnet/classify/ip_classify.c new file mode 100644 index 00000000..44973ae5 --- /dev/null +++ b/src/vnet/classify/ip_classify.c @@ -0,0 +1,365 @@ +/* + * Copyright (c) 2015 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ +#include +#include /* for ethernet_header_t */ +#include +#include + +typedef struct { + u32 next_index; + u32 table_index; + u32 entry_index; +} ip_classify_trace_t; + +/* packet trace format function */ +static u8 * format_ip_classify_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 *); + ip_classify_trace_t * t = va_arg (*args, ip_classify_trace_t *); + + s = format (s, "IP_CLASSIFY: next_index %d, table %d, entry %d", + t->next_index, t->table_index, t->entry_index); + return s; +} + +vlib_node_registration_t ip4_classify_node; +vlib_node_registration_t ip6_classify_node; + +#define foreach_ip_classify_error \ +_(MISS, "Classify misses") \ +_(HIT, "Classify hits") \ +_(CHAIN_HIT, "Classify hits after chain walk") + +typedef enum { +#define _(sym,str) IP_CLASSIFY_ERROR_##sym, + foreach_ip_classify_error +#undef _ + IP_CLASSIFY_N_ERROR, +} ip_classify_error_t; + +static char * ip_classify_error_strings[] = { +#define _(sym,string) string, + foreach_ip_classify_error +#undef _ +}; + +static inline uword +ip_classify_inline (vlib_main_t * vm, + vlib_node_runtime_t * node, + vlib_frame_t * frame, int is_ip4) +{ + u32 n_left_from, * from, * to_next; + ip_lookup_next_t next_index; + vnet_classify_main_t * vcm = &vnet_classify_main; + f64 now = vlib_time_now (vm); + u32 hits = 0; + u32 misses = 0; + u32 chain_hits = 0; + u32 n_next; + + if (is_ip4) { + n_next = IP4_LOOKUP_N_NEXT; + } else { + n_next = IP6_LOOKUP_N_NEXT; + } + + from = vlib_frame_vector_args (frame); + n_left_from = frame->n_vectors; + + /* First pass: compute hashes */ + + while (n_left_from > 2) + { + vlib_buffer_t * b0, * b1; + u32 bi0, bi1; + u8 * h0, * h1; + u32 cd_index0, cd_index1; + classify_dpo_t *cd0, * cd1; + u32 table_index0, table_index1; + vnet_classify_table_t * t0, * t1; + + /* prefetch next iteration */ + { + vlib_buffer_t * p1, * p2; + + p1 = vlib_get_buffer (vm, from[1]); + p2 = vlib_get_buffer (vm, from[2]); + + vlib_prefetch_buffer_header (p1, STORE); + CLIB_PREFETCH (p1->data, CLIB_CACHE_LINE_BYTES, STORE); + vlib_prefetch_buffer_header (p2, STORE); + CLIB_PREFETCH (p2->data, CLIB_CACHE_LINE_BYTES, STORE); + } + + bi0 = from[0]; + b0 = vlib_get_buffer (vm, bi0); + h0 = (void *)vlib_buffer_get_current(b0) - + ethernet_buffer_header_size(b0); + + bi1 = from[1]; + b1 = vlib_get_buffer (vm, bi1); + h1 = (void *)vlib_buffer_get_current(b1) - + ethernet_buffer_header_size(b1); + + cd_index0 = vnet_buffer (b0)->ip.adj_index[VLIB_TX]; + cd0 = classify_dpo_get(cd_index0); + table_index0 = cd0->cd_table_index; + + cd_index1 = vnet_buffer (b1)->ip.adj_index[VLIB_TX]; + cd1 = classify_dpo_get(cd_index1); + table_index1 = cd1->cd_table_index; + + t0 = pool_elt_at_index (vcm->tables, table_index0); + + t1 = pool_elt_at_index (vcm->tables, table_index1); + + vnet_buffer(b0)->l2_classify.hash = + vnet_classify_hash_packet (t0, (u8 *) h0); + + vnet_classify_prefetch_bucket (t0, vnet_buffer(b0)->l2_classify.hash); + + vnet_buffer(b1)->l2_classify.hash = + vnet_classify_hash_packet (t1, (u8 *) h1); + + vnet_classify_prefetch_bucket (t1, vnet_buffer(b1)->l2_classify.hash); + + vnet_buffer(b0)->l2_classify.table_index = table_index0; + + vnet_buffer(b1)->l2_classify.table_index = table_index1; + + from += 2; + n_left_from -= 2; + } + + while (n_left_from > 0) + { + vlib_buffer_t * b0; + u32 bi0; + u8 * h0; + u32 cd_index0; + classify_dpo_t *cd0; + u32 table_index0; + vnet_classify_table_t * t0; + + bi0 = from[0]; + b0 = vlib_get_buffer (vm, bi0); + h0 = (void *)vlib_buffer_get_current(b0) - + ethernet_buffer_header_size(b0); + + cd_index0 = vnet_buffer (b0)->ip.adj_index[VLIB_TX]; + cd0 = classify_dpo_get(cd_index0); + table_index0 = cd0->cd_table_index; + + t0 = pool_elt_at_index (vcm->tables, table_index0); + vnet_buffer(b0)->l2_classify.hash = + vnet_classify_hash_packet (t0, (u8 *) h0); + + vnet_buffer(b0)->l2_classify.table_index = table_index0; + vnet_classify_prefetch_bucket (t0, vnet_buffer(b0)->l2_classify.hash); + + from++; + n_left_from--; + } + + next_index = node->cached_next_index; + from = vlib_frame_vector_args (frame); + n_left_from = frame->n_vectors; + + while (n_left_from > 0) + { + u32 n_left_to_next; + + vlib_get_next_frame (vm, node, next_index, + to_next, n_left_to_next); + + /* Not enough load/store slots to dual loop... */ + while (n_left_from > 0 && n_left_to_next > 0) + { + u32 bi0; + vlib_buffer_t * b0; + u32 next0 = IP_LOOKUP_NEXT_DROP; + u32 table_index0; + vnet_classify_table_t * t0; + vnet_classify_entry_t * e0; + u64 hash0; + u8 * h0; + + /* Stride 3 seems to work best */ + if (PREDICT_TRUE (n_left_from > 3)) + { + vlib_buffer_t * p1 = vlib_get_buffer(vm, from[3]); + vnet_classify_table_t * tp1; + u32 table_index1; + u64 phash1; + + table_index1 = vnet_buffer(p1)->l2_classify.table_index; + + if (PREDICT_TRUE (table_index1 != ~0)) + { + tp1 = pool_elt_at_index (vcm->tables, table_index1); + phash1 = vnet_buffer(p1)->l2_classify.hash; + vnet_classify_prefetch_entry (tp1, phash1); + } + } + + /* speculatively enqueue b0 to the current next frame */ + bi0 = from[0]; + to_next[0] = bi0; + from += 1; + to_next += 1; + n_left_from -= 1; + n_left_to_next -= 1; + + b0 = vlib_get_buffer (vm, bi0); + h0 = b0->data; + table_index0 = vnet_buffer(b0)->l2_classify.table_index; + e0 = 0; + t0 = 0; + vnet_buffer(b0)->l2_classify.opaque_index = ~0; + + if (PREDICT_TRUE(table_index0 != ~0)) + { + hash0 = vnet_buffer(b0)->l2_classify.hash; + t0 = pool_elt_at_index (vcm->tables, table_index0); + + e0 = vnet_classify_find_entry (t0, (u8 *) h0, hash0, + now); + if (e0) + { + vnet_buffer(b0)->l2_classify.opaque_index + = e0->opaque_index; + vlib_buffer_advance (b0, e0->advance); + next0 = (e0->next_index < node->n_next_nodes)? + e0->next_index:next0; + hits++; + } + else + { + while (1) + { + if (t0->next_table_index != ~0) + t0 = pool_elt_at_index (vcm->tables, + t0->next_table_index); + else + { + next0 = (t0->miss_next_index < n_next) ? + t0->miss_next_index : next0; + misses++; + break; + } + + hash0 = vnet_classify_hash_packet (t0, (u8 *) h0); + e0 = vnet_classify_find_entry + (t0, (u8 *) h0, hash0, now); + if (e0) + { + vnet_buffer(b0)->l2_classify.opaque_index + = e0->opaque_index; + vlib_buffer_advance (b0, e0->advance); + next0 = (e0->next_index < node->n_next_nodes)? + e0->next_index:next0; + hits++; + chain_hits++; + break; + } + } + } + } + + if (PREDICT_FALSE((node->flags & VLIB_NODE_FLAG_TRACE) + && (b0->flags & VLIB_BUFFER_IS_TRACED))) + { + ip_classify_trace_t *t = + vlib_add_trace (vm, node, b0, sizeof (*t)); + t->next_index = next0; + t->table_index = t0 ? t0 - vcm->tables : ~0; + t->entry_index = e0 ? e0 - t0->entries : ~0; + } + + /* verify speculative enqueue, maybe switch current next frame */ + vlib_validate_buffer_enqueue_x1 (vm, node, next_index, + to_next, n_left_to_next, + bi0, next0); + } + + vlib_put_next_frame (vm, node, next_index, n_left_to_next); + } + + vlib_node_increment_counter (vm, node->node_index, + IP_CLASSIFY_ERROR_MISS, + misses); + vlib_node_increment_counter (vm, node->node_index, + IP_CLASSIFY_ERROR_HIT, + hits); + vlib_node_increment_counter (vm, node->node_index, + IP_CLASSIFY_ERROR_CHAIN_HIT, + chain_hits); + return frame->n_vectors; +} + +static uword +ip4_classify (vlib_main_t * vm, + vlib_node_runtime_t * node, + vlib_frame_t * frame) +{ + return ip_classify_inline (vm, node, frame, 1 /* is_ip4 */); +} + + +VLIB_REGISTER_NODE (ip4_classify_node) = { + .function = ip4_classify, + .name = "ip4-classify", + .vector_size = sizeof (u32), + .sibling_of = "ip4-lookup", + .format_trace = format_ip_classify_trace, + .n_errors = ARRAY_LEN(ip_classify_error_strings), + .error_strings = ip_classify_error_strings, + + .n_next_nodes = 0, +}; + +VLIB_NODE_FUNCTION_MULTIARCH (ip4_classify_node, ip4_classify) + +static uword +ip6_classify (vlib_main_t * vm, + vlib_node_runtime_t * node, + vlib_frame_t * frame) +{ + return ip_classify_inline (vm, node, frame, 0 /* is_ip4 */); +} + + +VLIB_REGISTER_NODE (ip6_classify_node) = { + .function = ip6_classify, + .name = "ip6-classify", + .vector_size = sizeof (u32), + .sibling_of = "ip6-lookup", + .format_trace = format_ip_classify_trace, + .n_errors = ARRAY_LEN(ip_classify_error_strings), + .error_strings = ip_classify_error_strings, + + .n_next_nodes = 0, +}; + +VLIB_NODE_FUNCTION_MULTIARCH (ip6_classify_node, ip6_classify) + +static clib_error_t * +ip_classify_init (vlib_main_t * vm) +{ + return 0; +} + +VLIB_INIT_FUNCTION (ip_classify_init); diff --git a/src/vnet/classify/policer_classify.c b/src/vnet/classify/policer_classify.c new file mode 100644 index 00000000..569234fb --- /dev/null +++ b/src/vnet/classify/policer_classify.c @@ -0,0 +1,227 @@ +/* + * Copyright (c) 2016 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ +#include + +static void +vnet_policer_classify_feature_enable (vlib_main_t * vnm, + policer_classify_main_t * pcm, + u32 sw_if_index, + policer_classify_table_id_t tid, + int feature_enable) +{ + if (tid == POLICER_CLASSIFY_TABLE_L2) + { + l2input_intf_bitmap_enable (sw_if_index, L2INPUT_FEAT_POLICER_CLAS, + feature_enable); + } + else + { + vnet_feature_config_main_t * fcm; + u8 arc; + + if (tid == POLICER_CLASSIFY_TABLE_IP4) + { + vnet_feature_enable_disable ("ip4-unicast", "ip4-policer-classify", + sw_if_index, feature_enable, 0, 0); + arc = vnet_get_feature_arc_index ("ip4-unicast"); + } + + else + { + vnet_feature_enable_disable ("ip6-unicast", "ip6-policer-classify", + sw_if_index, feature_enable, 0, 0); + arc = vnet_get_feature_arc_index ("ip6-unicast"); + } + + fcm = vnet_get_feature_arc_config_main (arc); + pcm->vnet_config_main[tid] = &fcm->config_main; + } +} + +int vnet_set_policer_classify_intfc (vlib_main_t * vm, u32 sw_if_index, + u32 ip4_table_index, u32 ip6_table_index, + u32 l2_table_index, u32 is_add) +{ + policer_classify_main_t * pcm = &policer_classify_main; + vnet_classify_main_t * vcm = pcm->vnet_classify_main; + u32 pct[POLICER_CLASSIFY_N_TABLES] = {ip4_table_index, ip6_table_index, + l2_table_index}; + u32 ti; + + /* Assume that we've validated sw_if_index in the API layer */ + + for (ti = 0; ti < POLICER_CLASSIFY_N_TABLES; ti++) + { + if (pct[ti] == ~0) + continue; + + if (pool_is_free_index (vcm->tables, pct[ti])) + return VNET_API_ERROR_NO_SUCH_TABLE; + + vec_validate_init_empty + (pcm->classify_table_index_by_sw_if_index[ti], sw_if_index, ~0); + + /* Reject any DEL operation with wrong sw_if_index */ + if (!is_add && + (pct[ti] != pcm->classify_table_index_by_sw_if_index[ti][sw_if_index])) + { + clib_warning ("Non-existent intf_idx=%d with table_index=%d for delete", + sw_if_index, pct[ti]); + return VNET_API_ERROR_NO_SUCH_TABLE; + } + + /* Return ok on ADD operaton if feature is already enabled */ + if (is_add && + pcm->classify_table_index_by_sw_if_index[ti][sw_if_index] != ~0) + return 0; + + vnet_policer_classify_feature_enable (vm, pcm, sw_if_index, ti, is_add); + + if (is_add) + pcm->classify_table_index_by_sw_if_index[ti][sw_if_index] = pct[ti]; + else + pcm->classify_table_index_by_sw_if_index[ti][sw_if_index] = ~0; + } + + + return 0; +} + +static clib_error_t * +set_policer_classify_command_fn (vlib_main_t * vm, + unformat_input_t * input, + vlib_cli_command_t * cmd) +{ + vnet_main_t * vnm = vnet_get_main(); + u32 sw_if_index = ~0; + u32 ip4_table_index = ~0; + u32 ip6_table_index = ~0; + u32 l2_table_index = ~0; + u32 is_add = 1; + u32 idx_cnt = 0; + int rv; + + while (unformat_check_input (input) != UNFORMAT_END_OF_INPUT) + { + if (unformat (input, "interface %U", unformat_vnet_sw_interface, + vnm, &sw_if_index)) + ; + else if (unformat (input, "ip4-table %d", &ip4_table_index)) + idx_cnt++; + else if (unformat (input, "ip6-table %d", &ip6_table_index)) + idx_cnt++; + else if (unformat (input, "l2-table %d", &l2_table_index)) + idx_cnt++; + else if (unformat (input, "del")) + is_add = 0; + else + break; + } + + if (sw_if_index == ~0) + return clib_error_return (0, "Interface must be specified."); + + if (!idx_cnt) + return clib_error_return (0, "Table index should be specified."); + + if (idx_cnt > 1) + return clib_error_return (0, "Only one table index per API is allowed."); + + rv = vnet_set_policer_classify_intfc(vm, sw_if_index, ip4_table_index, + ip6_table_index, l2_table_index, is_add); + + switch (rv) + { + case 0: + break; + + case VNET_API_ERROR_NO_MATCHING_INTERFACE: + return clib_error_return (0, "No such interface"); + + case VNET_API_ERROR_NO_SUCH_ENTRY: + return clib_error_return (0, "No such classifier table"); + } + return 0; +} + +VLIB_CLI_COMMAND (set_policer_classify_command, static) = { + .path = "set policer classify", + .short_help = + "set policer classify interface [ip4-table ]\n" + " [ip6-table ] [l2-table ] [del]", + .function = set_policer_classify_command_fn, +}; + +static uword +unformat_table_type (unformat_input_t * input, va_list * va) +{ + u32 * r = va_arg (*va, u32 *); + u32 tid; + + if (unformat (input, "ip4")) + tid = POLICER_CLASSIFY_TABLE_IP4; + else if (unformat (input, "ip6")) + tid = POLICER_CLASSIFY_TABLE_IP6; + else if (unformat (input, "l2")) + tid = POLICER_CLASSIFY_TABLE_L2; + else + return 0; + + *r = tid; + return 1; +} +static clib_error_t * +show_policer_classify_command_fn (vlib_main_t * vm, + unformat_input_t * input, + vlib_cli_command_t * cmd) +{ + policer_classify_main_t * pcm = &policer_classify_main; + u32 type = POLICER_CLASSIFY_N_TABLES; + u32 * vec_tbl; + int i; + + if (unformat (input, "type %U", unformat_table_type, &type)) + ; + else + return clib_error_return (0, "Type must be specified.");; + + if (type == POLICER_CLASSIFY_N_TABLES) + return clib_error_return (0, "Invalid table type."); + + vec_tbl = pcm->classify_table_index_by_sw_if_index[type]; + + if (vec_len(vec_tbl)) + vlib_cli_output (vm, "%10s%20s\t\t%s", "Intfc idx", "Classify table", + "Interface name"); + else + vlib_cli_output (vm, "No tables configured."); + + for (i = 0; i < vec_len (vec_tbl); i++) + { + if (vec_elt(vec_tbl, i) == ~0) + continue; + + vlib_cli_output (vm, "%10d%20d\t\t%U", i, vec_elt(vec_tbl, i), + format_vnet_sw_if_index_name, pcm->vnet_main, i); + } + + return 0; +} + +VLIB_CLI_COMMAND (show_policer_classify_command, static) = { + .path = "show classify policer", + .short_help = "show classify policer type [ip4|ip6|l2]", + .function = show_policer_classify_command_fn, +}; diff --git a/src/vnet/classify/policer_classify.h b/src/vnet/classify/policer_classify.h new file mode 100644 index 00000000..30656444 --- /dev/null +++ b/src/vnet/classify/policer_classify.h @@ -0,0 +1,55 @@ +/* + * Copyright (c) 2016 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +#ifndef __included_vnet_policer_classify_h__ +#define __included_vnet_policer_classify_h__ + +#include +#include +#include + +typedef enum { + POLICER_CLASSIFY_TABLE_IP4, + POLICER_CLASSIFY_TABLE_IP6, + POLICER_CLASSIFY_TABLE_L2, + POLICER_CLASSIFY_N_TABLES, +} policer_classify_table_id_t; + +typedef enum { + POLICER_CLASSIFY_NEXT_INDEX_DROP, + POLICER_CLASSIFY_NEXT_INDEX_N_NEXT, +} policer_classify_next_index_t; + +typedef struct { + /* Classifier table vectors */ + u32 * classify_table_index_by_sw_if_index [POLICER_CLASSIFY_N_TABLES]; + + /* L2 next nodes for each feature */ + u32 feat_next_node_index[32]; + + /* Convenience variables */ + vlib_main_t * vlib_main; + vnet_main_t * vnet_main; + vnet_classify_main_t * vnet_classify_main; + vnet_config_main_t * vnet_config_main [POLICER_CLASSIFY_N_TABLES]; +} policer_classify_main_t; + +policer_classify_main_t policer_classify_main; + +int vnet_set_policer_classify_intfc (vlib_main_t * vm, u32 sw_if_index, + u32 ip4_table_index, u32 ip6_table_index, + u32 l2_table_index, u32 is_add); + +#endif /* __included_vnet_policer_classify_h__ */ diff --git a/src/vnet/classify/vnet_classify.c b/src/vnet/classify/vnet_classify.c new file mode 100644 index 00000000..ce38f9f1 --- /dev/null +++ b/src/vnet/classify/vnet_classify.c @@ -0,0 +1,2436 @@ +/* + * Copyright (c) 2015 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ +#include +#include +#include +#include /* for API error numbers */ +#include /* for L2_INPUT_CLASSIFY_NEXT_xxx */ +#include + +vnet_classify_main_t vnet_classify_main; + +#if VALIDATION_SCAFFOLDING +/* Validation scaffolding */ +void mv (vnet_classify_table_t * t) +{ + void * oldheap; + + oldheap = clib_mem_set_heap (t->mheap); + clib_mem_validate(); + clib_mem_set_heap (oldheap); +} + +void rogue (vnet_classify_table_t * t) +{ + int i, j, k; + vnet_classify_entry_t * v, * save_v; + u32 active_elements = 0; + vnet_classify_bucket_t * b; + + for (i = 0; i < t->nbuckets; i++) + { + b = &t->buckets [i]; + if (b->offset == 0) + continue; + save_v = vnet_classify_get_entry (t, b->offset); + for (j = 0; j < (1<log2_pages); j++) + { + for (k = 0; k < t->entries_per_page; k++) + { + v = vnet_classify_entry_at_index + (t, save_v, j*t->entries_per_page + k); + + if (vnet_classify_entry_is_busy (v)) + active_elements++; + } + } + } + + if (active_elements != t->active_elements) + clib_warning ("found %u expected %u elts", active_elements, + t->active_elements); +} +#else +void mv (vnet_classify_table_t * t) { } +void rogue (vnet_classify_table_t * t) { } +#endif + +void vnet_classify_register_unformat_l2_next_index_fn (unformat_function_t * fn) +{ + vnet_classify_main_t * cm = &vnet_classify_main; + + vec_add1 (cm->unformat_l2_next_index_fns, fn); +} + +void vnet_classify_register_unformat_ip_next_index_fn (unformat_function_t * fn) +{ + vnet_classify_main_t * cm = &vnet_classify_main; + + vec_add1 (cm->unformat_ip_next_index_fns, fn); +} + +void +vnet_classify_register_unformat_acl_next_index_fn (unformat_function_t * fn) +{ + vnet_classify_main_t * cm = &vnet_classify_main; + + vec_add1 (cm->unformat_acl_next_index_fns, fn); +} + +void +vnet_classify_register_unformat_policer_next_index_fn (unformat_function_t * fn) +{ + vnet_classify_main_t * cm = &vnet_classify_main; + + vec_add1 (cm->unformat_policer_next_index_fns, fn); +} + +void vnet_classify_register_unformat_opaque_index_fn (unformat_function_t * fn) +{ + vnet_classify_main_t * cm = &vnet_classify_main; + + vec_add1 (cm->unformat_opaque_index_fns, fn); +} + +vnet_classify_table_t * +vnet_classify_new_table (vnet_classify_main_t *cm, + u8 * mask, u32 nbuckets, u32 memory_size, + u32 skip_n_vectors, + u32 match_n_vectors) +{ + vnet_classify_table_t * t; + void * oldheap; + + nbuckets = 1 << (max_log2 (nbuckets)); + + pool_get_aligned (cm->tables, t, CLIB_CACHE_LINE_BYTES); + memset(t, 0, sizeof (*t)); + + vec_validate_aligned (t->mask, match_n_vectors - 1, sizeof(u32x4)); + clib_memcpy (t->mask, mask, match_n_vectors * sizeof (u32x4)); + + t->next_table_index = ~0; + t->nbuckets = nbuckets; + t->log2_nbuckets = max_log2 (nbuckets); + t->match_n_vectors = match_n_vectors; + t->skip_n_vectors = skip_n_vectors; + t->entries_per_page = 2; + + t->mheap = mheap_alloc (0 /* use VM */, memory_size); + + vec_validate_aligned (t->buckets, nbuckets - 1, CLIB_CACHE_LINE_BYTES); + oldheap = clib_mem_set_heap (t->mheap); + + t->writer_lock = clib_mem_alloc_aligned (CLIB_CACHE_LINE_BYTES, + CLIB_CACHE_LINE_BYTES); + t->writer_lock[0] = 0; + + clib_mem_set_heap (oldheap); + return (t); +} + +void vnet_classify_delete_table_index (vnet_classify_main_t *cm, + u32 table_index, int del_chain) +{ + vnet_classify_table_t * t; + + /* Tolerate multiple frees, up to a point */ + if (pool_is_free_index (cm->tables, table_index)) + return; + + t = pool_elt_at_index (cm->tables, table_index); + if (del_chain && t->next_table_index != ~0) + /* Recursively delete the entire chain */ + vnet_classify_delete_table_index (cm, t->next_table_index, del_chain); + + vec_free (t->mask); + vec_free (t->buckets); + mheap_free (t->mheap); + + pool_put (cm->tables, t); +} + +static vnet_classify_entry_t * +vnet_classify_entry_alloc (vnet_classify_table_t * t, u32 log2_pages) +{ + vnet_classify_entry_t * rv = 0; +#define _(size) \ + vnet_classify_entry_##size##_t * rv##size = 0; + foreach_size_in_u32x4; +#undef _ + + void * oldheap; + + ASSERT (t->writer_lock[0]); + if (log2_pages >= vec_len (t->freelists) || t->freelists [log2_pages] == 0) + { + oldheap = clib_mem_set_heap (t->mheap); + + vec_validate (t->freelists, log2_pages); + + switch(t->match_n_vectors) + { + /* Euchre the vector allocator into allocating the right sizes */ +#define _(size) \ + case size: \ + vec_validate_aligned \ + (rv##size, ((1<entries_per_page) - 1, \ + CLIB_CACHE_LINE_BYTES); \ + rv = (vnet_classify_entry_t *) rv##size; \ + break; + foreach_size_in_u32x4; +#undef _ + + default: + abort(); + } + + clib_mem_set_heap (oldheap); + goto initialize; + } + rv = t->freelists[log2_pages]; + t->freelists[log2_pages] = rv->next_free; + +initialize: + ASSERT(rv); + ASSERT (vec_len(rv) == (1<entries_per_page); + + switch (t->match_n_vectors) + { +#define _(size) \ + case size: \ + if(vec_len(rv)) \ + memset (rv, 0xff, sizeof (*rv##size) * vec_len(rv)); \ + break; + foreach_size_in_u32x4; +#undef _ + + default: + abort(); + } + + return rv; +} + +static void +vnet_classify_entry_free (vnet_classify_table_t * t, + vnet_classify_entry_t * v) +{ + u32 free_list_index; + + ASSERT (t->writer_lock[0]); + + free_list_index = min_log2(vec_len(v)/t->entries_per_page); + + ASSERT(vec_len (t->freelists) > free_list_index); + + v->next_free = t->freelists[free_list_index]; + t->freelists[free_list_index] = v; +} + +static inline void make_working_copy +(vnet_classify_table_t * t, vnet_classify_bucket_t * b) +{ + vnet_classify_entry_t * v; + vnet_classify_bucket_t working_bucket __attribute__((aligned (8))); + void * oldheap; + vnet_classify_entry_t * working_copy; +#define _(size) \ + vnet_classify_entry_##size##_t * working_copy##size = 0; + foreach_size_in_u32x4; +#undef _ + u32 cpu_number = os_get_cpu_number(); + + if (cpu_number >= vec_len (t->working_copies)) + { + oldheap = clib_mem_set_heap (t->mheap); + vec_validate (t->working_copies, cpu_number); + clib_mem_set_heap (oldheap); + } + + /* + * working_copies are per-cpu so that near-simultaneous + * updates from multiple threads will not result in sporadic, spurious + * lookup failures. + */ + working_copy = t->working_copies[cpu_number]; + + t->saved_bucket.as_u64 = b->as_u64; + oldheap = clib_mem_set_heap (t->mheap); + + if ((1<log2_pages)*t->entries_per_page > vec_len (working_copy)) + { + switch(t->match_n_vectors) + { + /* Euchre the vector allocator into allocating the right sizes */ +#define _(size) \ + case size: \ + working_copy##size = (void *) working_copy; \ + vec_validate_aligned \ + (working_copy##size, \ + ((1<log2_pages)*t->entries_per_page) - 1, \ + CLIB_CACHE_LINE_BYTES); \ + working_copy = (void *) working_copy##size; \ + break; + foreach_size_in_u32x4; +#undef _ + + default: + abort(); + } + t->working_copies[cpu_number] = working_copy; + } + + _vec_len(working_copy) = (1<log2_pages)*t->entries_per_page; + clib_mem_set_heap (oldheap); + + v = vnet_classify_get_entry (t, b->offset); + + switch(t->match_n_vectors) + { +#define _(size) \ + case size: \ + clib_memcpy (working_copy, v, \ + sizeof (vnet_classify_entry_##size##_t) \ + * (1<log2_pages) \ + * (t->entries_per_page)); \ + break; + foreach_size_in_u32x4 ; +#undef _ + + default: + abort(); + } + + working_bucket.as_u64 = b->as_u64; + working_bucket.offset = vnet_classify_get_offset (t, working_copy); + CLIB_MEMORY_BARRIER(); + b->as_u64 = working_bucket.as_u64; + t->working_copies[cpu_number] = working_copy; +} + +static vnet_classify_entry_t * +split_and_rehash (vnet_classify_table_t * t, + vnet_classify_entry_t * old_values, + u32 new_log2_pages) +{ + vnet_classify_entry_t * new_values, * v, * new_v; + int i, j, k; + + new_values = vnet_classify_entry_alloc (t, new_log2_pages); + + for (i = 0; i < (vec_len (old_values)/t->entries_per_page); i++) + { + u64 new_hash; + + for (j = 0; j < t->entries_per_page; j++) + { + v = vnet_classify_entry_at_index + (t, old_values, i * t->entries_per_page + j); + + if (vnet_classify_entry_is_busy (v)) + { + /* Hack so we can use the packet hash routine */ + u8 * key_minus_skip; + key_minus_skip = (u8 *) v->key; + key_minus_skip -= t->skip_n_vectors * sizeof (u32x4); + + new_hash = vnet_classify_hash_packet (t, key_minus_skip); + new_hash >>= t->log2_nbuckets; + new_hash &= (1<entries_per_page; k++) + { + new_v = vnet_classify_entry_at_index (t, new_values, + new_hash + k); + + if (vnet_classify_entry_is_free (new_v)) + { + clib_memcpy (new_v, v, sizeof (vnet_classify_entry_t) + + (t->match_n_vectors * sizeof (u32x4))); + new_v->flags &= ~(VNET_CLASSIFY_ENTRY_FREE); + goto doublebreak; + } + } + /* Crap. Tell caller to try again */ + vnet_classify_entry_free (t, new_values); + return 0; + } + doublebreak: + ; + } + } + return new_values; +} + +int vnet_classify_add_del (vnet_classify_table_t * t, + vnet_classify_entry_t * add_v, + int is_add) +{ + u32 bucket_index; + vnet_classify_bucket_t * b, tmp_b; + vnet_classify_entry_t * v, * new_v, * save_new_v, * working_copy, * save_v; + u32 value_index; + int rv = 0; + int i; + u64 hash, new_hash; + u32 new_log2_pages; + u32 cpu_number = os_get_cpu_number(); + u8 * key_minus_skip; + + ASSERT ((add_v->flags & VNET_CLASSIFY_ENTRY_FREE) == 0); + + key_minus_skip = (u8 *) add_v->key; + key_minus_skip -= t->skip_n_vectors * sizeof (u32x4); + + hash = vnet_classify_hash_packet (t, key_minus_skip); + + bucket_index = hash & (t->nbuckets-1); + b = &t->buckets[bucket_index]; + + hash >>= t->log2_nbuckets; + + while (__sync_lock_test_and_set (t->writer_lock, 1)) + ; + + /* First elt in the bucket? */ + if (b->offset == 0) + { + if (is_add == 0) + { + rv = -1; + goto unlock; + } + + v = vnet_classify_entry_alloc (t, 0 /* new_log2_pages */); + clib_memcpy (v, add_v, sizeof (vnet_classify_entry_t) + + t->match_n_vectors * sizeof (u32x4)); + v->flags &= ~(VNET_CLASSIFY_ENTRY_FREE); + + tmp_b.as_u64 = 0; + tmp_b.offset = vnet_classify_get_offset (t, v); + + b->as_u64 = tmp_b.as_u64; + t->active_elements ++; + + goto unlock; + } + + make_working_copy (t, b); + + save_v = vnet_classify_get_entry (t, t->saved_bucket.offset); + value_index = hash & ((1<saved_bucket.log2_pages)-1); + + if (is_add) + { + /* + * For obvious (in hindsight) reasons, see if we're supposed to + * replace an existing key, then look for an empty slot. + */ + + for (i = 0; i < t->entries_per_page; i++) + { + v = vnet_classify_entry_at_index (t, save_v, value_index + i); + + if (!memcmp (v->key, add_v->key, t->match_n_vectors * sizeof (u32x4))) + { + clib_memcpy (v, add_v, sizeof (vnet_classify_entry_t) + + t->match_n_vectors * sizeof(u32x4)); + v->flags &= ~(VNET_CLASSIFY_ENTRY_FREE); + + CLIB_MEMORY_BARRIER(); + /* Restore the previous (k,v) pairs */ + b->as_u64 = t->saved_bucket.as_u64; + goto unlock; + } + } + for (i = 0; i < t->entries_per_page; i++) + { + v = vnet_classify_entry_at_index (t, save_v, value_index + i); + + if (vnet_classify_entry_is_free (v)) + { + clib_memcpy (v, add_v, sizeof (vnet_classify_entry_t) + + t->match_n_vectors * sizeof(u32x4)); + v->flags &= ~(VNET_CLASSIFY_ENTRY_FREE); + CLIB_MEMORY_BARRIER(); + b->as_u64 = t->saved_bucket.as_u64; + t->active_elements ++; + goto unlock; + } + } + /* no room at the inn... split case... */ + } + else + { + for (i = 0; i < t->entries_per_page; i++) + { + v = vnet_classify_entry_at_index (t, save_v, value_index + i); + + if (!memcmp (v->key, add_v->key, t->match_n_vectors * sizeof (u32x4))) + { + memset (v, 0xff, sizeof (vnet_classify_entry_t) + + t->match_n_vectors * sizeof(u32x4)); + v->flags |= VNET_CLASSIFY_ENTRY_FREE; + CLIB_MEMORY_BARRIER(); + b->as_u64 = t->saved_bucket.as_u64; + t->active_elements --; + goto unlock; + } + } + rv = -3; + b->as_u64 = t->saved_bucket.as_u64; + goto unlock; + } + + new_log2_pages = t->saved_bucket.log2_pages + 1; + + expand_again: + working_copy = t->working_copies[cpu_number]; + new_v = split_and_rehash (t, working_copy, new_log2_pages); + + if (new_v == 0) + { + new_log2_pages++; + goto expand_again; + } + + /* Try to add the new entry */ + save_new_v = new_v; + + key_minus_skip = (u8 *) add_v->key; + key_minus_skip -= t->skip_n_vectors * sizeof (u32x4); + + new_hash = vnet_classify_hash_packet_inline (t, key_minus_skip); + new_hash >>= t->log2_nbuckets; + new_hash &= (1<entries_per_page))) - 1; + + for (i = 0; i < t->entries_per_page; i++) + { + new_v = vnet_classify_entry_at_index (t, save_new_v, new_hash + i); + + if (vnet_classify_entry_is_free (new_v)) + { + clib_memcpy (new_v, add_v, sizeof (vnet_classify_entry_t) + + t->match_n_vectors * sizeof(u32x4)); + new_v->flags &= ~(VNET_CLASSIFY_ENTRY_FREE); + goto expand_ok; + } + } + /* Crap. Try again */ + new_log2_pages++; + vnet_classify_entry_free (t, save_new_v); + goto expand_again; + + expand_ok: + tmp_b.log2_pages = min_log2 (vec_len (save_new_v)/t->entries_per_page); + tmp_b.offset = vnet_classify_get_offset (t, save_new_v); + CLIB_MEMORY_BARRIER(); + b->as_u64 = tmp_b.as_u64; + t->active_elements ++; + v = vnet_classify_get_entry (t, t->saved_bucket.offset); + vnet_classify_entry_free (t, v); + + unlock: + CLIB_MEMORY_BARRIER(); + t->writer_lock[0] = 0; + + return rv; +} + +typedef CLIB_PACKED(struct { + ethernet_header_t eh; + ip4_header_t ip; +}) classify_data_or_mask_t; + +u64 vnet_classify_hash_packet (vnet_classify_table_t * t, u8 * h) +{ + return vnet_classify_hash_packet_inline (t, h); +} + +vnet_classify_entry_t * +vnet_classify_find_entry (vnet_classify_table_t * t, + u8 * h, u64 hash, f64 now) +{ + return vnet_classify_find_entry_inline (t, h, hash, now); +} + +static u8 * format_classify_entry (u8 * s, va_list * args) + { + vnet_classify_table_t * t = va_arg (*args, vnet_classify_table_t *); + vnet_classify_entry_t * e = va_arg (*args, vnet_classify_entry_t *); + + s = format + (s, "[%u]: next_index %d advance %d opaque %d action %d metadata %d\n", + vnet_classify_get_offset (t, e), e->next_index, e->advance, + e->opaque_index, e->action, e->metadata); + + + s = format (s, " k: %U\n", format_hex_bytes, e->key, + t->match_n_vectors * sizeof(u32x4)); + + if (vnet_classify_entry_is_busy (e)) + s = format (s, " hits %lld, last_heard %.2f\n", + e->hits, e->last_heard); + else + s = format (s, " entry is free\n"); + return s; + } + +u8 * format_classify_table (u8 * s, va_list * args) +{ + vnet_classify_table_t * t = va_arg (*args, vnet_classify_table_t *); + int verbose = va_arg (*args, int); + vnet_classify_bucket_t * b; + vnet_classify_entry_t * v, * save_v; + int i, j, k; + u64 active_elements = 0; + + for (i = 0; i < t->nbuckets; i++) + { + b = &t->buckets [i]; + if (b->offset == 0) + { + if (verbose > 1) + s = format (s, "[%d]: empty\n", i); + continue; + } + + if (verbose) + { + s = format (s, "[%d]: heap offset %d, len %d\n", i, + b->offset, (1<log2_pages)); + } + + save_v = vnet_classify_get_entry (t, b->offset); + for (j = 0; j < (1<log2_pages); j++) + { + for (k = 0; k < t->entries_per_page; k++) + { + + v = vnet_classify_entry_at_index (t, save_v, + j*t->entries_per_page + k); + + if (vnet_classify_entry_is_free (v)) + { + if (verbose > 1) + s = format (s, " %d: empty\n", + j * t->entries_per_page + k); + continue; + } + if (verbose) + { + s = format (s, " %d: %U\n", + j * t->entries_per_page + k, + format_classify_entry, t, v); + } + active_elements++; + } + } + } + + s = format (s, " %lld active elements\n", active_elements); + s = format (s, " %d free lists\n", vec_len (t->freelists)); + return s; +} + +int vnet_classify_add_del_table (vnet_classify_main_t * cm, + u8 * mask, + u32 nbuckets, + u32 memory_size, + u32 skip, + u32 match, + u32 next_table_index, + u32 miss_next_index, + u32 * table_index, + u8 current_data_flag, + i16 current_data_offset, + int is_add, + int del_chain) +{ + vnet_classify_table_t * t; + + if (is_add) + { + if (*table_index == ~0) /* add */ + { + if (memory_size == 0) + return VNET_API_ERROR_INVALID_MEMORY_SIZE; + + if (nbuckets == 0) + return VNET_API_ERROR_INVALID_VALUE; + + t = vnet_classify_new_table (cm, mask, nbuckets, memory_size, + skip, match); + t->next_table_index = next_table_index; + t->miss_next_index = miss_next_index; + t->current_data_flag = current_data_flag; + t->current_data_offset = current_data_offset; + *table_index = t - cm->tables; + } + else /* update */ + { + vnet_classify_main_t *cm = &vnet_classify_main; + t = pool_elt_at_index (cm->tables, *table_index); + + t->next_table_index = next_table_index; + } + return 0; + } + + vnet_classify_delete_table_index (cm, *table_index, del_chain); + return 0; +} + +#define foreach_tcp_proto_field \ +_(src_port) \ +_(dst_port) + +#define foreach_udp_proto_field \ +_(src_port) \ +_(dst_port) + +#define foreach_ip4_proto_field \ +_(src_address) \ +_(dst_address) \ +_(tos) \ +_(length) \ +_(fragment_id) \ +_(ttl) \ +_(protocol) \ +_(checksum) + +uword unformat_tcp_mask (unformat_input_t * input, va_list * args) +{ + u8 ** maskp = va_arg (*args, u8 **); + u8 * mask = 0; + u8 found_something = 0; + tcp_header_t * tcp; + +#define _(a) u8 a=0; + foreach_tcp_proto_field; +#undef _ + + while (unformat_check_input (input) != UNFORMAT_END_OF_INPUT) + { + if (0) ; +#define _(a) else if (unformat (input, #a)) a=1; + foreach_tcp_proto_field +#undef _ + else + break; + } + +#define _(a) found_something += a; + foreach_tcp_proto_field; +#undef _ + + if (found_something == 0) + return 0; + + vec_validate (mask, sizeof (*tcp) - 1); + + tcp = (tcp_header_t *) mask; + +#define _(a) if (a) memset (&tcp->a, 0xff, sizeof (tcp->a)); + foreach_tcp_proto_field; +#undef _ + + *maskp = mask; + return 1; +} + +uword unformat_udp_mask (unformat_input_t * input, va_list * args) +{ + u8 ** maskp = va_arg (*args, u8 **); + u8 * mask = 0; + u8 found_something = 0; + udp_header_t * udp; + +#define _(a) u8 a=0; + foreach_udp_proto_field; +#undef _ + + while (unformat_check_input (input) != UNFORMAT_END_OF_INPUT) + { + if (0) ; +#define _(a) else if (unformat (input, #a)) a=1; + foreach_udp_proto_field +#undef _ + else + break; + } + +#define _(a) found_something += a; + foreach_udp_proto_field; +#undef _ + + if (found_something == 0) + return 0; + + vec_validate (mask, sizeof (*udp) - 1); + + udp = (udp_header_t *) mask; + +#define _(a) if (a) memset (&udp->a, 0xff, sizeof (udp->a)); + foreach_udp_proto_field; +#undef _ + + *maskp = mask; + return 1; +} + +typedef struct { + u16 src_port, dst_port; +} tcpudp_header_t; + +uword unformat_l4_mask (unformat_input_t * input, va_list * args) +{ + u8 ** maskp = va_arg (*args, u8 **); + u16 src_port = 0, dst_port = 0; + tcpudp_header_t * tcpudp; + + while (unformat_check_input (input) != UNFORMAT_END_OF_INPUT) + { + if (unformat (input, "tcp %U", unformat_tcp_mask, maskp)) + return 1; + else if (unformat (input, "udp %U", unformat_udp_mask, maskp)) + return 1; + else if (unformat (input, "src_port")) + src_port = 0xFFFF; + else if (unformat (input, "dst_port")) + dst_port = 0xFFFF; + else + return 0; + } + + if (!src_port && !dst_port) + return 0; + + u8 * mask = 0; + vec_validate (mask, sizeof (tcpudp_header_t) - 1); + + tcpudp = (tcpudp_header_t *) mask; + tcpudp->src_port = src_port; + tcpudp->dst_port = dst_port; + + *maskp = mask; + + return 1; +} + +uword unformat_ip4_mask (unformat_input_t * input, va_list * args) +{ + u8 ** maskp = va_arg (*args, u8 **); + u8 * mask = 0; + u8 found_something = 0; + ip4_header_t * ip; + +#define _(a) u8 a=0; + foreach_ip4_proto_field; +#undef _ + u8 version = 0; + u8 hdr_length = 0; + + + while (unformat_check_input (input) != UNFORMAT_END_OF_INPUT) + { + if (unformat (input, "version")) + version = 1; + else if (unformat (input, "hdr_length")) + hdr_length = 1; + else if (unformat (input, "src")) + src_address = 1; + else if (unformat (input, "dst")) + dst_address = 1; + else if (unformat (input, "proto")) + protocol = 1; + +#define _(a) else if (unformat (input, #a)) a=1; + foreach_ip4_proto_field +#undef _ + else + break; + } + +#define _(a) found_something += a; + foreach_ip4_proto_field; +#undef _ + + if (found_something == 0) + return 0; + + vec_validate (mask, sizeof (*ip) - 1); + + ip = (ip4_header_t *) mask; + +#define _(a) if (a) memset (&ip->a, 0xff, sizeof (ip->a)); + foreach_ip4_proto_field; +#undef _ + + ip->ip_version_and_header_length = 0; + + if (version) + ip->ip_version_and_header_length |= 0xF0; + + if (hdr_length) + ip->ip_version_and_header_length |= 0x0F; + + *maskp = mask; + return 1; +} + +#define foreach_ip6_proto_field \ +_(src_address) \ +_(dst_address) \ +_(payload_length) \ +_(hop_limit) \ +_(protocol) + +uword unformat_ip6_mask (unformat_input_t * input, va_list * args) +{ + u8 ** maskp = va_arg (*args, u8 **); + u8 * mask = 0; + u8 found_something = 0; + ip6_header_t * ip; + u32 ip_version_traffic_class_and_flow_label; + +#define _(a) u8 a=0; + foreach_ip6_proto_field; +#undef _ + u8 version = 0; + u8 traffic_class = 0; + u8 flow_label = 0; + + while (unformat_check_input (input) != UNFORMAT_END_OF_INPUT) + { + if (unformat (input, "version")) + version = 1; + else if (unformat (input, "traffic-class")) + traffic_class = 1; + else if (unformat (input, "flow-label")) + flow_label = 1; + else if (unformat (input, "src")) + src_address = 1; + else if (unformat (input, "dst")) + dst_address = 1; + else if (unformat (input, "proto")) + protocol = 1; + +#define _(a) else if (unformat (input, #a)) a=1; + foreach_ip6_proto_field +#undef _ + else + break; + } + +#define _(a) found_something += a; + foreach_ip6_proto_field; +#undef _ + + if (found_something == 0) + return 0; + + vec_validate (mask, sizeof (*ip) - 1); + + ip = (ip6_header_t *) mask; + +#define _(a) if (a) memset (&ip->a, 0xff, sizeof (ip->a)); + foreach_ip6_proto_field; +#undef _ + + ip_version_traffic_class_and_flow_label = 0; + + if (version) + ip_version_traffic_class_and_flow_label |= 0xF0000000; + + if (traffic_class) + ip_version_traffic_class_and_flow_label |= 0x0FF00000; + + if (flow_label) + ip_version_traffic_class_and_flow_label |= 0x000FFFFF; + + ip->ip_version_traffic_class_and_flow_label = + clib_host_to_net_u32 (ip_version_traffic_class_and_flow_label); + + *maskp = mask; + return 1; +} + +uword unformat_l3_mask (unformat_input_t * input, va_list * args) +{ + u8 ** maskp = va_arg (*args, u8 **); + + while (unformat_check_input (input) != UNFORMAT_END_OF_INPUT) { + if (unformat (input, "ip4 %U", unformat_ip4_mask, maskp)) + return 1; + else if (unformat (input, "ip6 %U", unformat_ip6_mask, maskp)) + return 1; + else + break; + } + return 0; +} + +uword unformat_l2_mask (unformat_input_t * input, va_list * args) +{ + u8 ** maskp = va_arg (*args, u8 **); + u8 * mask = 0; + u8 src = 0; + u8 dst = 0; + u8 proto = 0; + u8 tag1 = 0; + u8 tag2 = 0; + u8 ignore_tag1 = 0; + u8 ignore_tag2 = 0; + u8 cos1 = 0; + u8 cos2 = 0; + u8 dot1q = 0; + u8 dot1ad = 0; + int len = 14; + + while (unformat_check_input (input) != UNFORMAT_END_OF_INPUT) { + if (unformat (input, "src")) + src = 1; + else if (unformat (input, "dst")) + dst = 1; + else if (unformat (input, "proto")) + proto = 1; + else if (unformat (input, "tag1")) + tag1 = 1; + else if (unformat (input, "tag2")) + tag2 = 1; + else if (unformat (input, "ignore-tag1")) + ignore_tag1 = 1; + else if (unformat (input, "ignore-tag2")) + ignore_tag2 = 1; + else if (unformat (input, "cos1")) + cos1 = 1; + else if (unformat (input, "cos2")) + cos2 = 1; + else if (unformat (input, "dot1q")) + dot1q = 1; + else if (unformat (input, "dot1ad")) + dot1ad = 1; + else + break; + } + if ((src + dst + proto + tag1 + tag2 + dot1q + dot1ad + + ignore_tag1 + ignore_tag2 + cos1 + cos2) == 0) + return 0; + + if (tag1 || ignore_tag1 || cos1 || dot1q) + len = 18; + if (tag2 || ignore_tag2 || cos2 || dot1ad) + len = 22; + + vec_validate (mask, len-1); + + if (dst) + memset (mask, 0xff, 6); + + if (src) + memset (mask + 6, 0xff, 6); + + if (tag2 || dot1ad) + { + /* inner vlan tag */ + if (tag2) + { + mask[19] = 0xff; + mask[18] = 0x0f; + } + if (cos2) + mask[18] |= 0xe0; + if (proto) + mask[21] = mask [20] = 0xff; + if (tag1) + { + mask [15] = 0xff; + mask [14] = 0x0f; + } + if (cos1) + mask[14] |= 0xe0; + *maskp = mask; + return 1; + } + if (tag1 | dot1q) + { + if (tag1) + { + mask [15] = 0xff; + mask [14] = 0x0f; + } + if (cos1) + mask[14] |= 0xe0; + if (proto) + mask[16] = mask [17] = 0xff; + *maskp = mask; + return 1; + } + if (cos2) + mask[18] |= 0xe0; + if (cos1) + mask[14] |= 0xe0; + if (proto) + mask[12] = mask [13] = 0xff; + + *maskp = mask; + return 1; +} + +uword unformat_classify_mask (unformat_input_t * input, va_list * args) +{ + vnet_classify_main_t * CLIB_UNUSED(cm) + = va_arg (*args, vnet_classify_main_t *); + u8 ** maskp = va_arg (*args, u8 **); + u32 * skipp = va_arg (*args, u32 *); + u32 * matchp = va_arg (*args, u32 *); + u32 match; + u8 * mask = 0; + u8 * l2 = 0; + u8 * l3 = 0; + u8 * l4 = 0; + int i; + + while (unformat_check_input (input) != UNFORMAT_END_OF_INPUT) { + if (unformat (input, "hex %U", unformat_hex_string, &mask)) + ; + else if (unformat (input, "l2 %U", unformat_l2_mask, &l2)) + ; + else if (unformat (input, "l3 %U", unformat_l3_mask, &l3)) + ; + else if (unformat (input, "l4 %U", unformat_l4_mask, &l4)) + ; + else + break; + } + + if (l4 && !l3) { + vec_free (mask); + vec_free (l2); + vec_free (l4); + return 0; + } + + if (mask || l2 || l3 || l4) + { + if (l2 || l3 || l4) + { + /* "With a free Ethernet header in every package" */ + if (l2 == 0) + vec_validate (l2, 13); + mask = l2; + if (l3) + { + vec_append (mask, l3); + vec_free (l3); + } + if (l4) + { + vec_append (mask, l4); + vec_free (l4); + } + } + + /* Scan forward looking for the first significant mask octet */ + for (i = 0; i < vec_len (mask); i++) + if (mask[i]) + break; + + /* compute (skip, match) params */ + *skipp = i / sizeof(u32x4); + vec_delete (mask, *skipp * sizeof(u32x4), 0); + + /* Pad mask to an even multiple of the vector size */ + while (vec_len (mask) % sizeof (u32x4)) + vec_add1 (mask, 0); + + match = vec_len (mask) / sizeof (u32x4); + + for (i = match*sizeof(u32x4); i > 0; i-= sizeof(u32x4)) + { + u64 *tmp = (u64 *)(mask + (i-sizeof(u32x4))); + if (*tmp || *(tmp+1)) + break; + match--; + } + if (match == 0) + clib_warning ("BUG: match 0"); + + _vec_len (mask) = match * sizeof(u32x4); + + *matchp = match; + *maskp = mask; + + return 1; + } + + return 0; +} + +#define foreach_l2_input_next \ +_(drop, DROP) \ +_(ethernet, ETHERNET_INPUT) \ +_(ip4, IP4_INPUT) \ +_(ip6, IP6_INPUT) \ +_(li, LI) + +uword unformat_l2_input_next_index (unformat_input_t * input, va_list * args) +{ + vnet_classify_main_t * cm = &vnet_classify_main; + u32 * miss_next_indexp = va_arg (*args, u32 *); + u32 next_index = 0; + u32 tmp; + int i; + + /* First try registered unformat fns, allowing override... */ + for (i = 0; i < vec_len (cm->unformat_l2_next_index_fns); i++) + { + if (unformat (input, "%U", cm->unformat_l2_next_index_fns[i], &tmp)) + { + next_index = tmp; + goto out; + } + } + +#define _(n,N) \ + if (unformat (input, #n)) { next_index = L2_INPUT_CLASSIFY_NEXT_##N; goto out;} + foreach_l2_input_next; +#undef _ + + if (unformat (input, "%d", &tmp)) + { + next_index = tmp; + goto out; + } + + return 0; + + out: + *miss_next_indexp = next_index; + return 1; +} + +#define foreach_l2_output_next \ +_(drop, DROP) + +uword unformat_l2_output_next_index (unformat_input_t * input, va_list * args) +{ + vnet_classify_main_t * cm = &vnet_classify_main; + u32 * miss_next_indexp = va_arg (*args, u32 *); + u32 next_index = 0; + u32 tmp; + int i; + + /* First try registered unformat fns, allowing override... */ + for (i = 0; i < vec_len (cm->unformat_l2_next_index_fns); i++) + { + if (unformat (input, "%U", cm->unformat_l2_next_index_fns[i], &tmp)) + { + next_index = tmp; + goto out; + } + } + +#define _(n,N) \ + if (unformat (input, #n)) { next_index = L2_OUTPUT_CLASSIFY_NEXT_##N; goto out;} + foreach_l2_output_next; +#undef _ + + if (unformat (input, "%d", &tmp)) + { + next_index = tmp; + goto out; + } + + return 0; + + out: + *miss_next_indexp = next_index; + return 1; +} + +#define foreach_ip_next \ +_(drop, DROP) \ +_(rewrite, REWRITE) + +uword unformat_ip_next_index (unformat_input_t * input, va_list * args) +{ + u32 * miss_next_indexp = va_arg (*args, u32 *); + vnet_classify_main_t * cm = &vnet_classify_main; + u32 next_index = 0; + u32 tmp; + int i; + + /* First try registered unformat fns, allowing override... */ + for (i = 0; i < vec_len (cm->unformat_ip_next_index_fns); i++) + { + if (unformat (input, "%U", cm->unformat_ip_next_index_fns[i], &tmp)) + { + next_index = tmp; + goto out; + } + } + +#define _(n,N) \ + if (unformat (input, #n)) { next_index = IP_LOOKUP_NEXT_##N; goto out;} + foreach_ip_next; +#undef _ + + if (unformat (input, "%d", &tmp)) + { + next_index = tmp; + goto out; + } + + return 0; + + out: + *miss_next_indexp = next_index; + return 1; +} + +#define foreach_acl_next \ +_(deny, DENY) + +uword unformat_acl_next_index (unformat_input_t * input, va_list * args) +{ + u32 * next_indexp = va_arg (*args, u32 *); + vnet_classify_main_t * cm = &vnet_classify_main; + u32 next_index = 0; + u32 tmp; + int i; + + /* First try registered unformat fns, allowing override... */ + for (i = 0; i < vec_len (cm->unformat_acl_next_index_fns); i++) + { + if (unformat (input, "%U", cm->unformat_acl_next_index_fns[i], &tmp)) + { + next_index = tmp; + goto out; + } + } + +#define _(n,N) \ + if (unformat (input, #n)) { next_index = ACL_NEXT_INDEX_##N; goto out;} + foreach_acl_next; +#undef _ + + if (unformat (input, "permit")) + { + next_index = ~0; + goto out; + } + else if (unformat (input, "%d", &tmp)) + { + next_index = tmp; + goto out; + } + + return 0; + + out: + *next_indexp = next_index; + return 1; +} + +uword unformat_policer_next_index (unformat_input_t * input, va_list * args) +{ + u32 * next_indexp = va_arg (*args, u32 *); + vnet_classify_main_t * cm = &vnet_classify_main; + u32 next_index = 0; + u32 tmp; + int i; + + /* First try registered unformat fns, allowing override... */ + for (i = 0; i < vec_len (cm->unformat_policer_next_index_fns); i++) + { + if (unformat (input, "%U", cm->unformat_policer_next_index_fns[i], &tmp)) + { + next_index = tmp; + goto out; + } + } + + if (unformat (input, "%d", &tmp)) + { + next_index = tmp; + goto out; + } + + return 0; + + out: + *next_indexp = next_index; + return 1; +} + +static clib_error_t * +classify_table_command_fn (vlib_main_t * vm, + unformat_input_t * input, + vlib_cli_command_t * cmd) +{ + u32 nbuckets = 2; + u32 skip = ~0; + u32 match = ~0; + int is_add = 1; + int del_chain = 0; + u32 table_index = ~0; + u32 next_table_index = ~0; + u32 miss_next_index = ~0; + u32 memory_size = 2<<20; + u32 tmp; + u32 current_data_flag = 0; + int current_data_offset = 0; + + u8 * mask = 0; + vnet_classify_main_t * cm = &vnet_classify_main; + int rv; + + while (unformat_check_input (input) != UNFORMAT_END_OF_INPUT) { + if (unformat (input, "del")) + is_add = 0; + else if (unformat (input, "del-chain")) + { + is_add = 0; + del_chain = 1; + } + else if (unformat (input, "buckets %d", &nbuckets)) + ; + else if (unformat (input, "skip %d", &skip)) + ; + else if (unformat (input, "match %d", &match)) + ; + else if (unformat (input, "table %d", &table_index)) + ; + else if (unformat (input, "mask %U", unformat_classify_mask, + cm, &mask, &skip, &match)) + ; + else if (unformat (input, "memory-size %uM", &tmp)) + memory_size = tmp<<20; + else if (unformat (input, "memory-size %uG", &tmp)) + memory_size = tmp<<30; + else if (unformat (input, "next-table %d", &next_table_index)) + ; + else if (unformat (input, "miss-next %U", unformat_ip_next_index, + &miss_next_index)) + ; + else if (unformat (input, "l2-input-miss-next %U", unformat_l2_input_next_index, + &miss_next_index)) + ; + else if (unformat (input, "l2-output-miss-next %U", unformat_l2_output_next_index, + &miss_next_index)) + ; + else if (unformat (input, "acl-miss-next %U", unformat_acl_next_index, + &miss_next_index)) + ; + else if (unformat (input, "current-data-flag %d", ¤t_data_flag)) + ; + else if (unformat (input, "current-data-offset %d", ¤t_data_offset)) + ; + + else + break; + } + + if (is_add && mask == 0 && table_index == ~0) + return clib_error_return (0, "Mask required"); + + if (is_add && skip == ~0 && table_index == ~0) + return clib_error_return (0, "skip count required"); + + if (is_add && match == ~0 && table_index == ~0) + return clib_error_return (0, "match count required"); + + if (!is_add && table_index == ~0) + return clib_error_return (0, "table index required for delete"); + + rv = vnet_classify_add_del_table (cm, mask, nbuckets, memory_size, + skip, match, next_table_index, miss_next_index, &table_index, + current_data_flag, current_data_offset, is_add, del_chain); + switch (rv) + { + case 0: + break; + + default: + return clib_error_return (0, "vnet_classify_add_del_table returned %d", + rv); + } + return 0; +} + +VLIB_CLI_COMMAND (classify_table, static) = { + .path = "classify table", + .short_help = + "classify table [miss-next|l2-miss_next|acl-miss-next ]" + "\n mask buckets [skip ] [match ]" + "\n [current-data-flag ] [current-data-offset ] [table ]" + "\n [del] [del-chain]", + .function = classify_table_command_fn, +}; + +static u8 * format_vnet_classify_table (u8 * s, va_list * args) +{ + vnet_classify_main_t * cm = va_arg (*args, vnet_classify_main_t *); + int verbose = va_arg (*args, int); + u32 index = va_arg (*args, u32); + vnet_classify_table_t * t; + + if (index == ~0) + { + s = format (s, "%10s%10s%10s%10s", "TableIdx", "Sessions", "NextTbl", + "NextNode", verbose ? "Details" : ""); + return s; + } + + t = pool_elt_at_index (cm->tables, index); + s = format (s, "%10u%10d%10d%10d", index, t->active_elements, + t->next_table_index, t->miss_next_index); + + s = format (s, "\n Heap: %U", format_mheap, t->mheap, 0 /*verbose*/); + + s = format (s, "\n nbuckets %d, skip %d match %d flag %d offset %d", + t->nbuckets, t->skip_n_vectors, t->match_n_vectors, + t->current_data_flag, t->current_data_offset); + s = format (s, "\n mask %U", format_hex_bytes, t->mask, + t->match_n_vectors * sizeof (u32x4)); + + if (verbose == 0) + return s; + + s = format (s, "\n%U", format_classify_table, t, verbose); + + return s; +} + +static clib_error_t * +show_classify_tables_command_fn (vlib_main_t * vm, + unformat_input_t * input, + vlib_cli_command_t * cmd) +{ + vnet_classify_main_t * cm = &vnet_classify_main; + vnet_classify_table_t * t; + u32 match_index = ~0; + u32 * indices = 0; + int verbose = 0; + int i; + + while (unformat_check_input (input) != UNFORMAT_END_OF_INPUT) + { + if (unformat (input, "index %d", &match_index)) + ; + else if (unformat (input, "verbose %d", &verbose)) + ; + else if (unformat (input, "verbose")) + verbose = 1; + else + break; + } + + pool_foreach (t, cm->tables, + ({ + if (match_index == ~0 || (match_index == t - cm->tables)) + vec_add1 (indices, t - cm->tables); + })); + + if (vec_len(indices)) + { + vlib_cli_output (vm, "%U", format_vnet_classify_table, cm, verbose, + ~0 /* hdr */); + for (i = 0; i < vec_len (indices); i++) + vlib_cli_output (vm, "%U", format_vnet_classify_table, cm, + verbose, indices[i]); + } + else + vlib_cli_output (vm, "No classifier tables configured"); + + vec_free (indices); + + return 0; +} + +VLIB_CLI_COMMAND (show_classify_table_command, static) = { + .path = "show classify tables", + .short_help = "show classify tables [index ]", + .function = show_classify_tables_command_fn, +}; + +uword unformat_l4_match (unformat_input_t * input, va_list * args) +{ + u8 ** matchp = va_arg (*args, u8 **); + + u8 * proto_header = 0; + int src_port = 0; + int dst_port = 0; + + tcpudp_header_t h; + + while (unformat_check_input (input) != UNFORMAT_END_OF_INPUT) + { + if (unformat (input, "src_port %d", &src_port)) + ; + else if (unformat (input, "dst_port %d", &dst_port)) + ; + else + return 0; + } + + h.src_port = clib_host_to_net_u16(src_port); + h.dst_port = clib_host_to_net_u16(dst_port); + vec_validate(proto_header, sizeof(h)-1); + memcpy(proto_header, &h, sizeof(h)); + + *matchp = proto_header; + + return 1; +} + +uword unformat_ip4_match (unformat_input_t * input, va_list * args) +{ + u8 ** matchp = va_arg (*args, u8 **); + u8 * match = 0; + ip4_header_t * ip; + int version = 0; + u32 version_val; + int hdr_length = 0; + u32 hdr_length_val; + int src = 0, dst = 0; + ip4_address_t src_val, dst_val; + int proto = 0; + u32 proto_val; + int tos = 0; + u32 tos_val; + int length = 0; + u32 length_val; + int fragment_id = 0; + u32 fragment_id_val; + int ttl = 0; + int ttl_val; + int checksum = 0; + u32 checksum_val; + + while (unformat_check_input (input) != UNFORMAT_END_OF_INPUT) + { + if (unformat (input, "version %d", &version_val)) + version = 1; + else if (unformat (input, "hdr_length %d", &hdr_length_val)) + hdr_length = 1; + else if (unformat (input, "src %U", unformat_ip4_address, &src_val)) + src = 1; + else if (unformat (input, "dst %U", unformat_ip4_address, &dst_val)) + dst = 1; + else if (unformat (input, "proto %d", &proto_val)) + proto = 1; + else if (unformat (input, "tos %d", &tos_val)) + tos = 1; + else if (unformat (input, "length %d", &length_val)) + length = 1; + else if (unformat (input, "fragment_id %d", &fragment_id_val)) + fragment_id = 1; + else if (unformat (input, "ttl %d", &ttl_val)) + ttl = 1; + else if (unformat (input, "checksum %d", &checksum_val)) + checksum = 1; + else + break; + } + + if (version + hdr_length + src + dst + proto + tos + length + fragment_id + + ttl + checksum == 0) + return 0; + + /* + * Aligned because we use the real comparison functions + */ + vec_validate_aligned (match, sizeof (*ip) - 1, sizeof(u32x4)); + + ip = (ip4_header_t *) match; + + /* These are realistically matched in practice */ + if (src) + ip->src_address.as_u32 = src_val.as_u32; + + if (dst) + ip->dst_address.as_u32 = dst_val.as_u32; + + if (proto) + ip->protocol = proto_val; + + + /* These are not, but they're included for completeness */ + if (version) + ip->ip_version_and_header_length |= (version_val & 0xF)<<4; + + if (hdr_length) + ip->ip_version_and_header_length |= (hdr_length_val & 0xF); + + if (tos) + ip->tos = tos_val; + + if (length) + ip->length = clib_host_to_net_u16 (length_val); + + if (ttl) + ip->ttl = ttl_val; + + if (checksum) + ip->checksum = clib_host_to_net_u16 (checksum_val); + + *matchp = match; + return 1; +} + +uword unformat_ip6_match (unformat_input_t * input, va_list * args) +{ + u8 ** matchp = va_arg (*args, u8 **); + u8 * match = 0; + ip6_header_t * ip; + int version = 0; + u32 version_val; + u8 traffic_class = 0; + u32 traffic_class_val; + u8 flow_label = 0; + u8 flow_label_val; + int src = 0, dst = 0; + ip6_address_t src_val, dst_val; + int proto = 0; + u32 proto_val; + int payload_length = 0; + u32 payload_length_val; + int hop_limit = 0; + int hop_limit_val; + u32 ip_version_traffic_class_and_flow_label; + + while (unformat_check_input (input) != UNFORMAT_END_OF_INPUT) + { + if (unformat (input, "version %d", &version_val)) + version = 1; + else if (unformat (input, "traffic_class %d", &traffic_class_val)) + traffic_class = 1; + else if (unformat (input, "flow_label %d", &flow_label_val)) + flow_label = 1; + else if (unformat (input, "src %U", unformat_ip6_address, &src_val)) + src = 1; + else if (unformat (input, "dst %U", unformat_ip6_address, &dst_val)) + dst = 1; + else if (unformat (input, "proto %d", &proto_val)) + proto = 1; + else if (unformat (input, "payload_length %d", &payload_length_val)) + payload_length = 1; + else if (unformat (input, "hop_limit %d", &hop_limit_val)) + hop_limit = 1; + else + break; + } + + if (version + traffic_class + flow_label + src + dst + proto + + payload_length + hop_limit == 0) + return 0; + + /* + * Aligned because we use the real comparison functions + */ + vec_validate_aligned (match, sizeof (*ip) - 1, sizeof(u32x4)); + + ip = (ip6_header_t *) match; + + if (src) + clib_memcpy (&ip->src_address, &src_val, sizeof (ip->src_address)); + + if (dst) + clib_memcpy (&ip->dst_address, &dst_val, sizeof (ip->dst_address)); + + if (proto) + ip->protocol = proto_val; + + ip_version_traffic_class_and_flow_label = 0; + + if (version) + ip_version_traffic_class_and_flow_label |= (version_val & 0xF) << 28; + + if (traffic_class) + ip_version_traffic_class_and_flow_label |= (traffic_class_val & 0xFF) << 20; + + if (flow_label) + ip_version_traffic_class_and_flow_label |= (flow_label_val & 0xFFFFF); + + ip->ip_version_traffic_class_and_flow_label = + clib_host_to_net_u32 (ip_version_traffic_class_and_flow_label); + + if (payload_length) + ip->payload_length = clib_host_to_net_u16 (payload_length_val); + + if (hop_limit) + ip->hop_limit = hop_limit_val; + + *matchp = match; + return 1; +} + +uword unformat_l3_match (unformat_input_t * input, va_list * args) +{ + u8 ** matchp = va_arg (*args, u8 **); + + while (unformat_check_input (input) != UNFORMAT_END_OF_INPUT) { + if (unformat (input, "ip4 %U", unformat_ip4_match, matchp)) + return 1; + else if (unformat (input, "ip6 %U", unformat_ip6_match, matchp)) + return 1; + /* $$$$ add mpls */ + else + break; + } + return 0; +} + +uword unformat_vlan_tag (unformat_input_t * input, va_list * args) +{ + u8 * tagp = va_arg (*args, u8 *); + u32 tag; + + if (unformat(input, "%d", &tag)) + { + tagp[0] = (tag>>8) & 0x0F; + tagp[1] = tag & 0xFF; + return 1; + } + + return 0; +} + +uword unformat_l2_match (unformat_input_t * input, va_list * args) +{ + u8 ** matchp = va_arg (*args, u8 **); + u8 * match = 0; + u8 src = 0; + u8 src_val[6]; + u8 dst = 0; + u8 dst_val[6]; + u8 proto = 0; + u16 proto_val; + u8 tag1 = 0; + u8 tag1_val [2]; + u8 tag2 = 0; + u8 tag2_val [2]; + int len = 14; + u8 ignore_tag1 = 0; + u8 ignore_tag2 = 0; + u8 cos1 = 0; + u8 cos2 = 0; + u32 cos1_val = 0; + u32 cos2_val = 0; + + while (unformat_check_input (input) != UNFORMAT_END_OF_INPUT) { + if (unformat (input, "src %U", unformat_ethernet_address, &src_val)) + src = 1; + else if (unformat (input, "dst %U", unformat_ethernet_address, &dst_val)) + dst = 1; + else if (unformat (input, "proto %U", + unformat_ethernet_type_host_byte_order, &proto_val)) + proto = 1; + else if (unformat (input, "tag1 %U", unformat_vlan_tag, tag1_val)) + tag1 = 1; + else if (unformat (input, "tag2 %U", unformat_vlan_tag, tag2_val)) + tag2 = 1; + else if (unformat (input, "ignore-tag1")) + ignore_tag1 = 1; + else if (unformat (input, "ignore-tag2")) + ignore_tag2 = 1; + else if (unformat (input, "cos1 %d", &cos1_val)) + cos1 = 1; + else if (unformat (input, "cos2 %d", &cos2_val)) + cos2 = 1; + else + break; + } + if ((src + dst + proto + tag1 + tag2 + + ignore_tag1 + ignore_tag2 + cos1 + cos2) == 0) + return 0; + + if (tag1 || ignore_tag1 || cos1) + len = 18; + if (tag2 || ignore_tag2 || cos2) + len = 22; + + vec_validate_aligned (match, len-1, sizeof(u32x4)); + + if (dst) + clib_memcpy (match, dst_val, 6); + + if (src) + clib_memcpy (match + 6, src_val, 6); + + if (tag2) + { + /* inner vlan tag */ + match[19] = tag2_val[1]; + match[18] = tag2_val[0]; + if (cos2) + match [18] |= (cos2_val & 0x7) << 5; + if (proto) + { + match[21] = proto_val & 0xff; + match[20] = proto_val >> 8; + } + if (tag1) + { + match [15] = tag1_val[1]; + match [14] = tag1_val[0]; + } + if (cos1) + match [14] |= (cos1_val & 0x7) << 5; + *matchp = match; + return 1; + } + if (tag1) + { + match [15] = tag1_val[1]; + match [14] = tag1_val[0]; + if (proto) + { + match[17] = proto_val & 0xff; + match[16] = proto_val >> 8; + } + if (cos1) + match [14] |= (cos1_val & 0x7) << 5; + + *matchp = match; + return 1; + } + if (cos2) + match [18] |= (cos2_val & 0x7) << 5; + if (cos1) + match [14] |= (cos1_val & 0x7) << 5; + if (proto) + { + match[13] = proto_val & 0xff; + match[12] = proto_val >> 8; + } + + *matchp = match; + return 1; +} + + +uword unformat_classify_match (unformat_input_t * input, va_list * args) +{ + vnet_classify_main_t * cm = va_arg (*args, vnet_classify_main_t *); + u8 ** matchp = va_arg (*args, u8 **); + u32 table_index = va_arg (*args, u32); + vnet_classify_table_t * t; + + u8 * match = 0; + u8 * l2 = 0; + u8 * l3 = 0; + u8 * l4 = 0; + + if (pool_is_free_index (cm->tables, table_index)) + return 0; + + t = pool_elt_at_index (cm->tables, table_index); + + while (unformat_check_input (input) != UNFORMAT_END_OF_INPUT) { + if (unformat (input, "hex %U", unformat_hex_string, &match)) + ; + else if (unformat (input, "l2 %U", unformat_l2_match, &l2)) + ; + else if (unformat (input, "l3 %U", unformat_l3_match, &l3)) + ; + else if (unformat (input, "l4 %U", unformat_l4_match, &l4)) + ; + else + break; + } + + if (l4 && !l3) { + vec_free (match); + vec_free (l2); + vec_free (l4); + return 0; + } + + if (match || l2 || l3 || l4) + { + if (l2 || l3 || l4) + { + /* "Win a free Ethernet header in every packet" */ + if (l2 == 0) + vec_validate_aligned (l2, 13, sizeof(u32x4)); + match = l2; + if (l3) + { + vec_append_aligned (match, l3, sizeof(u32x4)); + vec_free (l3); + } + if (l4) + { + vec_append_aligned (match, l4, sizeof(u32x4)); + vec_free (l4); + } + } + + /* Make sure the vector is big enough even if key is all 0's */ + vec_validate_aligned + (match, ((t->match_n_vectors + t->skip_n_vectors) * sizeof(u32x4)) - 1, + sizeof(u32x4)); + + /* Set size, include skipped vectors*/ + _vec_len (match) = (t->match_n_vectors+t->skip_n_vectors) * sizeof(u32x4); + + *matchp = match; + + return 1; + } + + return 0; +} + +int vnet_classify_add_del_session (vnet_classify_main_t * cm, + u32 table_index, + u8 * match, + u32 hit_next_index, + u32 opaque_index, + i32 advance, + u8 action, + u32 metadata, + int is_add) +{ + vnet_classify_table_t * t; + vnet_classify_entry_5_t _max_e __attribute__((aligned (16))); + vnet_classify_entry_t * e; + int i, rv; + + if (pool_is_free_index (cm->tables, table_index)) + return VNET_API_ERROR_NO_SUCH_TABLE; + + t = pool_elt_at_index (cm->tables, table_index); + + e = (vnet_classify_entry_t *)&_max_e; + e->next_index = hit_next_index; + e->opaque_index = opaque_index; + e->advance = advance; + e->hits = 0; + e->last_heard = 0; + e->flags = 0; + e->action = action; + if (e->action == CLASSIFY_ACTION_SET_IP4_FIB_INDEX) + e->metadata = fib_table_find_or_create_and_lock (FIB_PROTOCOL_IP4, metadata); + else if (e->action == CLASSIFY_ACTION_SET_IP6_FIB_INDEX) + e->metadata = fib_table_find_or_create_and_lock (FIB_PROTOCOL_IP6, metadata); + + /* Copy key data, honoring skip_n_vectors */ + clib_memcpy (&e->key, match + t->skip_n_vectors * sizeof (u32x4), + t->match_n_vectors * sizeof (u32x4)); + + /* Clear don't-care bits; likely when dynamically creating sessions */ + for (i = 0; i < t->match_n_vectors; i++) + e->key[i] &= t->mask[i]; + + rv = vnet_classify_add_del (t, e, is_add); + if (rv) + return VNET_API_ERROR_NO_SUCH_ENTRY; + return 0; +} + +static clib_error_t * +classify_session_command_fn (vlib_main_t * vm, + unformat_input_t * input, + vlib_cli_command_t * cmd) +{ + vnet_classify_main_t * cm = &vnet_classify_main; + int is_add = 1; + u32 table_index = ~0; + u32 hit_next_index = ~0; + u64 opaque_index = ~0; + u8 * match = 0; + i32 advance = 0; + u32 action = 0; + u32 metadata = 0; + int i, rv; + + while (unformat_check_input (input) != UNFORMAT_END_OF_INPUT) + { + if (unformat (input, "del")) + is_add = 0; + else if (unformat (input, "hit-next %U", unformat_ip_next_index, + &hit_next_index)) + ; + else if (unformat (input, "l2-input-hit-next %U", unformat_l2_input_next_index, + &hit_next_index)) + ; + else if (unformat (input, "l2-output-hit-next %U", unformat_l2_output_next_index, + &hit_next_index)) + ; + else if (unformat (input, "acl-hit-next %U", unformat_acl_next_index, + &hit_next_index)) + ; + else if (unformat (input, "policer-hit-next %U", + unformat_policer_next_index, &hit_next_index)) + ; + else if (unformat (input, "opaque-index %lld", &opaque_index)) + ; + else if (unformat (input, "match %U", unformat_classify_match, + cm, &match, table_index)) + ; + else if (unformat (input, "advance %d", &advance)) + ; + else if (unformat (input, "table-index %d", &table_index)) + ; + else if (unformat (input, "action set-ip4-fib-id %d", &metadata)) + action = 1; + else if (unformat (input, "action set-ip6-fib-id %d", &metadata)) + action = 2; + else + { + /* Try registered opaque-index unformat fns */ + for (i = 0; i < vec_len (cm->unformat_opaque_index_fns); i++) + { + if (unformat (input, "%U", cm->unformat_opaque_index_fns[i], + &opaque_index)) + goto found_opaque; + } + break; + } + found_opaque: + ; + } + + if (table_index == ~0) + return clib_error_return (0, "Table index required"); + + if (is_add && match == 0) + return clib_error_return (0, "Match value required"); + + rv = vnet_classify_add_del_session (cm, table_index, match, + hit_next_index, + opaque_index, advance, + action, metadata, is_add); + + switch(rv) + { + case 0: + break; + + default: + return clib_error_return (0, "vnet_classify_add_del_session returned %d", + rv); + } + + return 0; +} + +VLIB_CLI_COMMAND (classify_session_command, static) = { + .path = "classify session", + .short_help = + "classify session [hit-next|l2-hit-next|" + "acl-hit-next |policer-hit-next ]" + "\n table-index match [hex] [l2] [l3 ip4] [opaque-index ]" + "\n [action set-ip4-fib-id ] [action set-ip6-fib-id ] [del]", + .function = classify_session_command_fn, +}; + +static uword +unformat_opaque_sw_if_index (unformat_input_t * input, va_list * args) +{ + u64 * opaquep = va_arg (*args, u64 *); + u32 sw_if_index; + + if (unformat (input, "opaque-sw_if_index %U", unformat_vnet_sw_interface, + vnet_get_main(), &sw_if_index)) + { + *opaquep = sw_if_index; + return 1; + } + return 0; +} + +static uword +unformat_ip_next_node (unformat_input_t * input, va_list * args) +{ + vnet_classify_main_t * cm = &vnet_classify_main; + u32 * next_indexp = va_arg (*args, u32 *); + u32 node_index; + u32 next_index = ~0; + + if (unformat (input, "ip6-node %U", unformat_vlib_node, + cm->vlib_main, &node_index)) + { + next_index = vlib_node_add_next (cm->vlib_main, + ip6_classify_node.index, node_index); + } + else if (unformat (input, "ip4-node %U", unformat_vlib_node, + cm->vlib_main, &node_index)) + { + next_index = vlib_node_add_next (cm->vlib_main, + ip4_classify_node.index, node_index); + } + else + return 0; + + *next_indexp = next_index; + return 1; +} + +static uword +unformat_acl_next_node (unformat_input_t * input, va_list * args) +{ + vnet_classify_main_t * cm = &vnet_classify_main; + u32 * next_indexp = va_arg (*args, u32 *); + u32 node_index; + u32 next_index; + + if (unformat (input, "ip6-node %U", unformat_vlib_node, + cm->vlib_main, &node_index)) + { + next_index = vlib_node_add_next (cm->vlib_main, + ip6_inacl_node.index, node_index); + } + else if (unformat (input, "ip4-node %U", unformat_vlib_node, + cm->vlib_main, &node_index)) + { + next_index = vlib_node_add_next (cm->vlib_main, + ip4_inacl_node.index, node_index); + } + else + return 0; + + *next_indexp = next_index; + return 1; +} + +static uword +unformat_l2_input_next_node (unformat_input_t * input, va_list * args) +{ + vnet_classify_main_t * cm = &vnet_classify_main; + u32 * next_indexp = va_arg (*args, u32 *); + u32 node_index; + u32 next_index; + + if (unformat (input, "input-node %U", unformat_vlib_node, + cm->vlib_main, &node_index)) + { + next_index = vlib_node_add_next + (cm->vlib_main, l2_input_classify_node.index, node_index); + + *next_indexp = next_index; + return 1; + } + return 0; +} + +static uword +unformat_l2_output_next_node (unformat_input_t * input, va_list * args) +{ + vnet_classify_main_t * cm = &vnet_classify_main; + u32 * next_indexp = va_arg (*args, u32 *); + u32 node_index; + u32 next_index; + + if (unformat (input, "output-node %U", unformat_vlib_node, + cm->vlib_main, &node_index)) + { + next_index = vlib_node_add_next + (cm->vlib_main, l2_output_classify_node.index, node_index); + + *next_indexp = next_index; + return 1; + } + return 0; +} + +static clib_error_t * +vnet_classify_init (vlib_main_t * vm) +{ + vnet_classify_main_t * cm = &vnet_classify_main; + + cm->vlib_main = vm; + cm->vnet_main = vnet_get_main(); + + vnet_classify_register_unformat_opaque_index_fn + (unformat_opaque_sw_if_index); + + vnet_classify_register_unformat_ip_next_index_fn + (unformat_ip_next_node); + + vnet_classify_register_unformat_l2_next_index_fn + (unformat_l2_input_next_node); + + vnet_classify_register_unformat_l2_next_index_fn + (unformat_l2_output_next_node); + + vnet_classify_register_unformat_acl_next_index_fn + (unformat_acl_next_node); + + return 0; +} + +VLIB_INIT_FUNCTION (vnet_classify_init); + +#define TEST_CODE 1 + +#if TEST_CODE > 0 +static clib_error_t * +test_classify_command_fn (vlib_main_t * vm, + unformat_input_t * input, + vlib_cli_command_t * cmd) +{ + u32 buckets = 2; + u32 sessions = 10; + int i, rv; + vnet_classify_table_t * t = 0; + classify_data_or_mask_t * mask; + classify_data_or_mask_t * data; + u8 *mp = 0, *dp = 0; + vnet_classify_main_t * cm = &vnet_classify_main; + vnet_classify_entry_t * e; + int is_add = 1; + u32 tmp; + u32 table_index = ~0; + ip4_address_t src; + u32 deleted = 0; + u32 memory_size = 64<<20; + + /* Default starting address 1.0.0.10 */ + src.as_u32 = clib_net_to_host_u32 (0x0100000A); + + while (unformat_check_input (input) != UNFORMAT_END_OF_INPUT) { + if (unformat (input, "sessions %d", &sessions)) + ; + else if (unformat (input, "src %U", unformat_ip4_address, &src)) + ; + else if (unformat (input, "buckets %d", &buckets)) + ; + else if (unformat (input, "memory-size %uM", &tmp)) + memory_size = tmp<<20; + else if (unformat (input, "memory-size %uG", &tmp)) + memory_size = tmp<<30; + else if (unformat (input, "del")) + is_add = 0; + else if (unformat (input, "table %d", &table_index)) + ; + else + break; + } + + vec_validate_aligned (mp, 3 * sizeof(u32x4), sizeof(u32x4)); + vec_validate_aligned (dp, 3 * sizeof(u32x4), sizeof(u32x4)); + + mask = (classify_data_or_mask_t *) mp; + data = (classify_data_or_mask_t *) dp; + + data->ip.src_address.as_u32 = src.as_u32; + + /* Mask on src address */ + memset (&mask->ip.src_address, 0xff, 4); + + buckets = 1<tables, table_index)) + { + vlib_cli_output (vm, "No such table %d", table_index); + goto out; + } + t = pool_elt_at_index (cm->tables, table_index); + } + + if (is_add) + { + if (t == 0) + { + t = vnet_classify_new_table (cm, (u8 *)mask, buckets, + memory_size, + 0 /* skip */, + 3 /* vectors to match */); + t->miss_next_index = IP_LOOKUP_NEXT_DROP; + vlib_cli_output (vm, "Create table %d", t - cm->tables); + } + + vlib_cli_output (vm, "Add %d sessions to %d buckets...", + sessions, buckets); + + for (i = 0; i < sessions; i++) + { + rv = vnet_classify_add_del_session (cm, t - cm->tables, (u8 *) data, + IP_LOOKUP_NEXT_DROP, + i+100 /* opaque_index */, + 0 /* advance */, 0, 0, + 1 /* is_add */); + + if (rv != 0) + clib_warning ("add: returned %d", rv); + + tmp = clib_net_to_host_u32 (data->ip.src_address.as_u32) + 1; + data->ip.src_address.as_u32 = clib_net_to_host_u32 (tmp); + } + goto out; + } + + if (t == 0) + { + vlib_cli_output (vm, "Must specify table index to delete sessions"); + goto out; + } + + vlib_cli_output (vm, "Try to delete %d sessions...", sessions); + + for (i = 0; i < sessions; i++) + { + u8 * key_minus_skip; + u64 hash; + + hash = vnet_classify_hash_packet (t, (u8 *) data); + + e = vnet_classify_find_entry (t, (u8 *) data, hash, 0 /* time_now */); + /* Previous delete, perhaps... */ + if (e == 0) + continue; + ASSERT (e->opaque_index == (i+100)); + + key_minus_skip = (u8 *)e->key; + key_minus_skip -= t->skip_n_vectors * sizeof (u32x4); + + rv = vnet_classify_add_del_session (cm, t - cm->tables, key_minus_skip, + IP_LOOKUP_NEXT_DROP, + i+100 /* opaque_index */, + 0 /* advance */, 0, 0, + 0 /* is_add */); + if (rv != 0) + clib_warning ("del: returned %d", rv); + + tmp = clib_net_to_host_u32 (data->ip.src_address.as_u32) + 1; + data->ip.src_address.as_u32 = clib_net_to_host_u32 (tmp); + deleted++; + } + + vlib_cli_output (vm, "Deleted %d sessions...", deleted); + + out: + vec_free (mp); + vec_free (dp); + + return 0; +} + +VLIB_CLI_COMMAND (test_classify_command, static) = { + .path = "test classify", + .short_help = + "test classify [src ] [sessions ] [buckets ] [table ] [del]", + .function = test_classify_command_fn, +}; +#endif /* TEST_CODE */ diff --git a/src/vnet/classify/vnet_classify.h b/src/vnet/classify/vnet_classify.h new file mode 100644 index 00000000..d0b896ed --- /dev/null +++ b/src/vnet/classify/vnet_classify.h @@ -0,0 +1,523 @@ +/* + * Copyright (c) 2015 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ +#ifndef __included_vnet_classify_h__ +#define __included_vnet_classify_h__ + +#include + +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include /* for API error numbers */ + +#include +#include +#include +#include + +extern vlib_node_registration_t ip4_classify_node; +extern vlib_node_registration_t ip6_classify_node; + +#define CLASSIFY_TRACE 0 + +#if !defined( __aarch64__) && !defined(__arm__) +#define CLASSIFY_USE_SSE //Allow usage of SSE operations +#endif + +#define U32X4_ALIGNED(p) PREDICT_TRUE((((intptr_t)p) & 0xf) == 0) + +/* + * Classify table option to process packets + * CLASSIFY_FLAG_USE_CURR_DATA: + * - classify packets starting from VPP node’s current data pointer + */ +#define CLASSIFY_FLAG_USE_CURR_DATA 1 + +/* + * Classify session action + * CLASSIFY_ACTION_SET_IP4_FIB_INDEX: + * - Classified IP packets will be looked up + * from the specified ipv4 fib table + * CLASSIFY_ACTION_SET_IP6_FIB_INDEX: + * - Classified IP packets will be looked up + * from the specified ipv6 fib table + */ +#define CLASSIFY_ACTION_SET_IP4_FIB_INDEX 1 +#define CLASSIFY_ACTION_SET_IP6_FIB_INDEX 2 + +struct _vnet_classify_main; +typedef struct _vnet_classify_main vnet_classify_main_t; + +#define foreach_size_in_u32x4 \ +_(1) \ +_(2) \ +_(3) \ +_(4) \ +_(5) + +typedef CLIB_PACKED(struct _vnet_classify_entry { + /* Graph node next index */ + u32 next_index; + + /* put into vnet_buffer(b)->l2_classfy.opaque_index */ + union { + struct { + u32 opaque_index; + /* advance on hit, note it's a signed quantity... */ + i32 advance; + }; + u64 opaque_count; + }; + + /* Really only need 1 bit */ + u8 flags; +#define VNET_CLASSIFY_ENTRY_FREE (1<<0) + + u8 action; + u16 metadata; + + /* Hit counter, last heard time */ + union { + u64 hits; + struct _vnet_classify_entry * next_free; + }; + + f64 last_heard; + + /* Must be aligned to a 16-octet boundary */ + u32x4 key[0]; +}) vnet_classify_entry_t; + +static inline int vnet_classify_entry_is_free (vnet_classify_entry_t * e) +{ + return e->flags & VNET_CLASSIFY_ENTRY_FREE; +} + +static inline int vnet_classify_entry_is_busy (vnet_classify_entry_t * e) +{ + return ((e->flags & VNET_CLASSIFY_ENTRY_FREE) == 0); +} + +/* Need these to con the vector allocator */ +#define _(size) \ +typedef CLIB_PACKED(struct { \ + u32 pad0[4]; \ + u64 pad1[2]; \ + u32x4 key[size]; \ +}) vnet_classify_entry_##size##_t; +foreach_size_in_u32x4; +#undef _ + +typedef struct { + union { + struct { + u32 offset; + u8 pad[3]; + u8 log2_pages; + }; + u64 as_u64; + }; +} vnet_classify_bucket_t; + +typedef struct { + /* Mask to apply after skipping N vectors */ + u32x4 *mask; + /* Buckets and entries */ + vnet_classify_bucket_t * buckets; + vnet_classify_entry_t * entries; + + /* Config parameters */ + u32 match_n_vectors; + u32 skip_n_vectors; + u32 nbuckets; + u32 log2_nbuckets; + int entries_per_page; + u32 active_elements; + u32 current_data_flag; + int current_data_offset; + u32 data_offset; + /* Index of next table to try */ + u32 next_table_index; + + /* Miss next index, return if next_table_index = 0 */ + u32 miss_next_index; + + /* Per-bucket working copies, one per thread */ + vnet_classify_entry_t ** working_copies; + vnet_classify_bucket_t saved_bucket; + + /* Free entry freelists */ + vnet_classify_entry_t **freelists; + + u8 * name; + + /* Private allocation arena, protected by the writer lock */ + void * mheap; + + /* Writer (only) lock for this table */ + volatile u32 * writer_lock; + +} vnet_classify_table_t; + +struct _vnet_classify_main { + /* Table pool */ + vnet_classify_table_t * tables; + + /* Registered next-index, opaque unformat fcns */ + unformat_function_t ** unformat_l2_next_index_fns; + unformat_function_t ** unformat_ip_next_index_fns; + unformat_function_t ** unformat_acl_next_index_fns; + unformat_function_t ** unformat_policer_next_index_fns; + unformat_function_t ** unformat_opaque_index_fns; + + /* convenience variables */ + vlib_main_t * vlib_main; + vnet_main_t * vnet_main; +}; + +extern vnet_classify_main_t vnet_classify_main; + +u8 * format_classify_table (u8 * s, va_list * args); + +u64 vnet_classify_hash_packet (vnet_classify_table_t * t, u8 * h); + +static inline u64 +vnet_classify_hash_packet_inline (vnet_classify_table_t * t, + u8 * h) +{ + u32x4 *mask; + + union { + u32x4 as_u32x4; + u64 as_u64[2]; + } xor_sum __attribute__((aligned(sizeof(u32x4)))); + + ASSERT(t); + mask = t->mask; +#ifdef CLASSIFY_USE_SSE + if (U32X4_ALIGNED(h)) { //SSE can't handle unaligned data + u32x4 *data = (u32x4 *)h; + xor_sum.as_u32x4 = data[0 + t->skip_n_vectors] & mask[0]; + switch (t->match_n_vectors) + { + case 5: + xor_sum.as_u32x4 ^= data[4 + t->skip_n_vectors] & mask[4]; + /* FALLTHROUGH */ + case 4: + xor_sum.as_u32x4 ^= data[3 + t->skip_n_vectors] & mask[3]; + /* FALLTHROUGH */ + case 3: + xor_sum.as_u32x4 ^= data[2 + t->skip_n_vectors] & mask[2]; + /* FALLTHROUGH */ + case 2: + xor_sum.as_u32x4 ^= data[1 + t->skip_n_vectors] & mask[1]; + /* FALLTHROUGH */ + case 1: + break; + default: + abort(); + } + } else +#endif /* CLASSIFY_USE_SSE */ + { + u32 skip_u64 = t->skip_n_vectors * 2; + u64 *data64 = (u64 *)h; + xor_sum.as_u64[0] = data64[0 + skip_u64] & ((u64 *)mask)[0]; + xor_sum.as_u64[1] = data64[1 + skip_u64] & ((u64 *)mask)[1]; + switch (t->match_n_vectors) + { + case 5: + xor_sum.as_u64[0] ^= data64[8 + skip_u64] & ((u64 *)mask)[8]; + xor_sum.as_u64[1] ^= data64[9 + skip_u64] & ((u64 *)mask)[9]; + /* FALLTHROUGH */ + case 4: + xor_sum.as_u64[0] ^= data64[6 + skip_u64] & ((u64 *)mask)[6]; + xor_sum.as_u64[1] ^= data64[7 + skip_u64] & ((u64 *)mask)[7]; + /* FALLTHROUGH */ + case 3: + xor_sum.as_u64[0] ^= data64[4 + skip_u64] & ((u64 *)mask)[4]; + xor_sum.as_u64[1] ^= data64[5 + skip_u64] & ((u64 *)mask)[5]; + /* FALLTHROUGH */ + case 2: + xor_sum.as_u64[0] ^= data64[2 + skip_u64] & ((u64 *)mask)[2]; + xor_sum.as_u64[1] ^= data64[3 + skip_u64] & ((u64 *)mask)[3]; + /* FALLTHROUGH */ + case 1: + break; + + default: + abort(); + } + } + + return clib_xxhash (xor_sum.as_u64[0] ^ xor_sum.as_u64[1]); +} + +static inline void +vnet_classify_prefetch_bucket (vnet_classify_table_t * t, u64 hash) +{ + u32 bucket_index; + + ASSERT (is_pow2(t->nbuckets)); + + bucket_index = hash & (t->nbuckets - 1); + + CLIB_PREFETCH(&t->buckets[bucket_index], CLIB_CACHE_LINE_BYTES, LOAD); +} + +static inline vnet_classify_entry_t * +vnet_classify_get_entry (vnet_classify_table_t * t, uword offset) +{ + u8 * hp = t->mheap; + u8 * vp = hp + offset; + + return (void *) vp; +} + +static inline uword vnet_classify_get_offset (vnet_classify_table_t * t, + vnet_classify_entry_t * v) +{ + u8 * hp, * vp; + + hp = (u8 *) t->mheap; + vp = (u8 *) v; + + ASSERT((vp - hp) < 0x100000000ULL); + return vp - hp; +} + +static inline vnet_classify_entry_t * +vnet_classify_entry_at_index (vnet_classify_table_t * t, + vnet_classify_entry_t * e, + u32 index) +{ + u8 * eu8; + + eu8 = (u8 *)e; + + eu8 += index * (sizeof (vnet_classify_entry_t) + + (t->match_n_vectors * sizeof (u32x4))); + + return (vnet_classify_entry_t *) eu8; +} + +static inline void +vnet_classify_prefetch_entry (vnet_classify_table_t * t, + u64 hash) +{ + u32 bucket_index; + u32 value_index; + vnet_classify_bucket_t * b; + vnet_classify_entry_t * e; + + bucket_index = hash & (t->nbuckets - 1); + + b = &t->buckets[bucket_index]; + + if (b->offset == 0) + return; + + hash >>= t->log2_nbuckets; + + e = vnet_classify_get_entry (t, b->offset); + value_index = hash & ((1<log2_pages)-1); + + e = vnet_classify_entry_at_index (t, e, value_index); + + CLIB_PREFETCH(e, CLIB_CACHE_LINE_BYTES, LOAD); +} + +vnet_classify_entry_t * +vnet_classify_find_entry (vnet_classify_table_t * t, + u8 * h, u64 hash, f64 now); + +static inline vnet_classify_entry_t * +vnet_classify_find_entry_inline (vnet_classify_table_t * t, + u8 * h, u64 hash, f64 now) + { + vnet_classify_entry_t * v; + u32x4 *mask, *key; + union { + u32x4 as_u32x4; + u64 as_u64[2]; + } result __attribute__((aligned(sizeof(u32x4)))); + vnet_classify_bucket_t * b; + u32 value_index; + u32 bucket_index; + int i; + + bucket_index = hash & (t->nbuckets-1); + b = &t->buckets[bucket_index]; + mask = t->mask; + + if (b->offset == 0) + return 0; + + hash >>= t->log2_nbuckets; + + v = vnet_classify_get_entry (t, b->offset); + value_index = hash & ((1<log2_pages)-1); + v = vnet_classify_entry_at_index (t, v, value_index); + +#ifdef CLASSIFY_USE_SSE + if (U32X4_ALIGNED(h)) { + u32x4 *data = (u32x4 *) h; + for (i = 0; i < t->entries_per_page; i++) { + key = v->key; + result.as_u32x4 = (data[0 + t->skip_n_vectors] & mask[0]) ^ key[0]; + switch (t->match_n_vectors) + { + case 5: + result.as_u32x4 |= (data[4 + t->skip_n_vectors] & mask[4]) ^ key[4]; + /* FALLTHROUGH */ + case 4: + result.as_u32x4 |= (data[3 + t->skip_n_vectors] & mask[3]) ^ key[3]; + /* FALLTHROUGH */ + case 3: + result.as_u32x4 |= (data[2 + t->skip_n_vectors] & mask[2]) ^ key[2]; + /* FALLTHROUGH */ + case 2: + result.as_u32x4 |= (data[1 + t->skip_n_vectors] & mask[1]) ^ key[1]; + /* FALLTHROUGH */ + case 1: + break; + default: + abort(); + } + + if (u32x4_zero_byte_mask (result.as_u32x4) == 0xffff) { + if (PREDICT_TRUE(now)) { + v->hits++; + v->last_heard = now; + } + return (v); + } + v = vnet_classify_entry_at_index (t, v, 1); + } + } else +#endif /* CLASSIFY_USE_SSE */ + { + u32 skip_u64 = t->skip_n_vectors * 2; + u64 *data64 = (u64 *)h; + for (i = 0; i < t->entries_per_page; i++) { + key = v->key; + + result.as_u64[0] = (data64[0 + skip_u64] & ((u64 *)mask)[0]) ^ ((u64 *)key)[0]; + result.as_u64[1] = (data64[1 + skip_u64] & ((u64 *)mask)[1]) ^ ((u64 *)key)[1]; + switch (t->match_n_vectors) + { + case 5: + result.as_u64[0] |= (data64[8 + skip_u64] & ((u64 *)mask)[8]) ^ ((u64 *)key)[8]; + result.as_u64[1] |= (data64[9 + skip_u64] & ((u64 *)mask)[9]) ^ ((u64 *)key)[9]; + /* FALLTHROUGH */ + case 4: + result.as_u64[0] |= (data64[6 + skip_u64] & ((u64 *)mask)[6]) ^ ((u64 *)key)[6]; + result.as_u64[1] |= (data64[7 + skip_u64] & ((u64 *)mask)[7]) ^ ((u64 *)key)[7]; + /* FALLTHROUGH */ + case 3: + result.as_u64[0] |= (data64[4 + skip_u64] & ((u64 *)mask)[4]) ^ ((u64 *)key)[4]; + result.as_u64[1] |= (data64[5 + skip_u64] & ((u64 *)mask)[5]) ^ ((u64 *)key)[5]; + /* FALLTHROUGH */ + case 2: + result.as_u64[0] |= (data64[2 + skip_u64] & ((u64 *)mask)[2]) ^ ((u64 *)key)[2]; + result.as_u64[1] |= (data64[3 + skip_u64] & ((u64 *)mask)[3]) ^ ((u64 *)key)[3]; + /* FALLTHROUGH */ + case 1: + break; + default: + abort(); + } + + if (result.as_u64[0] == 0 && result.as_u64[1] == 0) { + if (PREDICT_TRUE(now)) { + v->hits++; + v->last_heard = now; + } + return (v); + } + + v = vnet_classify_entry_at_index (t, v, 1); + } + } + return 0; + } + +vnet_classify_table_t * +vnet_classify_new_table (vnet_classify_main_t *cm, + u8 * mask, u32 nbuckets, u32 memory_size, + u32 skip_n_vectors, + u32 match_n_vectors); + +int vnet_classify_add_del_session (vnet_classify_main_t * cm, + u32 table_index, + u8 * match, + u32 hit_next_index, + u32 opaque_index, + i32 advance, + u8 action, + u32 metadata, + int is_add); + +int vnet_classify_add_del_table (vnet_classify_main_t * cm, + u8 * mask, + u32 nbuckets, + u32 memory_size, + u32 skip, + u32 match, + u32 next_table_index, + u32 miss_next_index, + u32 * table_index, + u8 current_data_flag, + i16 current_data_offset, + int is_add, + int del_chain); + +unformat_function_t unformat_ip4_mask; +unformat_function_t unformat_ip6_mask; +unformat_function_t unformat_l3_mask; +unformat_function_t unformat_l2_mask; +unformat_function_t unformat_classify_mask; +unformat_function_t unformat_l2_next_index; +unformat_function_t unformat_ip_next_index; +unformat_function_t unformat_ip4_match; +unformat_function_t unformat_ip6_match; +unformat_function_t unformat_l3_match; +unformat_function_t unformat_vlan_tag; +unformat_function_t unformat_l2_match; +unformat_function_t unformat_classify_match; + +void vnet_classify_register_unformat_ip_next_index_fn +(unformat_function_t * fn); + +void vnet_classify_register_unformat_l2_next_index_fn +(unformat_function_t * fn); + +void vnet_classify_register_unformat_acl_next_index_fn +(unformat_function_t * fn); + +void vnet_classify_register_unformat_policer_next_index_fn +(unformat_function_t * fn); + +void vnet_classify_register_unformat_opaque_index_fn (unformat_function_t * fn); + +#endif /* __included_vnet_classify_h__ */ -- cgit 1.2.3-korg From d2c97d988b2fbc28f0905d1826b428967d09348a Mon Sep 17 00:00:00 2001 From: Pavel Kotucek Date: Tue, 24 Jan 2017 10:58:12 +0100 Subject: API refactoring : classify Change-Id: Ib75197ef8e5057e7f0d9361a10705c3743d05333 Signed-off-by: Pavel Kotucek --- src/vnet.am | 8 +- src/vnet/classify/classify.api | 356 +++++++++++++++++++++++++ src/vnet/classify/classify_api.c | 555 +++++++++++++++++++++++++++++++++++++++ src/vnet/vnet_all_api_h.h | 1 + src/vpp/api/api.c | 456 -------------------------------- src/vpp/api/vpe.api | 340 +----------------------- 6 files changed, 921 insertions(+), 795 deletions(-) create mode 100644 src/vnet/classify/classify.api create mode 100644 src/vnet/classify/classify_api.c (limited to 'src/vnet/classify') diff --git a/src/vnet.am b/src/vnet.am index 93dd1e6c..660efcf5 100644 --- a/src/vnet.am +++ b/src/vnet.am @@ -270,13 +270,17 @@ libvnet_la_SOURCES += \ vnet/classify/policer_classify.c \ vnet/classify/flow_classify.c \ vnet/classify/flow_classify_node.c \ - vnet/classify/vnet_classify.h + vnet/classify/vnet_classify.h \ + vnet/classify/classify_api.c nobase_include_HEADERS += \ vnet/classify/vnet_classify.h \ vnet/classify/input_acl.h \ vnet/classify/policer_classify.h \ - vnet/classify/flow_classify.h + vnet/classify/flow_classify.h \ + vnet/classify/classify.api.h + +API_FILES += vnet/classify/classify.api ######################################## # Layer 3 protocols go here diff --git a/src/vnet/classify/classify.api b/src/vnet/classify/classify.api new file mode 100644 index 00000000..51ebd6c8 --- /dev/null +++ b/src/vnet/classify/classify.api @@ -0,0 +1,356 @@ +/* + * Copyright (c) 2015-2016 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +/** \brief Add/Delete classification table request + @param client_index - opaque cookie to identify the sender + @param context - sender context, to match reply w/ request + @param is_add- if non-zero add the table, else delete it + @param del_chain - if non-zero delete the whole chain of tables + @param table_index - if add, reuturns index of the created table, else specifies the table to delete + @param nbuckets - number of buckets when adding a table + @param memory_size - memory size when adding a table + @param match_n_vectors - number of match vectors + @param next_table_index - index of next table + @param miss_next_index - index of miss table + @param current_data_flag - option to use current node's packet payload + as the starting point from where packets are classified, + This option is only valid for L2/L3 input ACL for now. + 0: by default, classify data from the buffer's start location + 1: classify packets from VPP node’s current data pointer + @param current_data_offset - a signed value to shift the start location of + the packet to be classified + For example, if input IP ACL node is used, L2 header’s first byte + can be accessible by configuring current_data_offset to -14 + if there is no vlan tag. + This is valid only if current_data_flag is set to 1. + @param mask[] - match mask +*/ +define classify_add_del_table +{ + u32 client_index; + u32 context; + u8 is_add; + u8 del_chain; + u32 table_index; + u32 nbuckets; + u32 memory_size; + u32 skip_n_vectors; + u32 match_n_vectors; + u32 next_table_index; + u32 miss_next_index; + u32 current_data_flag; + i32 current_data_offset; + u8 mask[0]; +}; + +/** \brief Add/Delete classification table response + @param context - sender context, to match reply w/ request + @param retval - return code for the table add/del requst + @param new_table_index - for add, returned index of the new table + @param skip_n_vectors - for add, returned value of skip_n_vectors in table + @param match_n_vectors -for add, returned value of match_n_vectors in table +*/ +define classify_add_del_table_reply +{ + u32 context; + i32 retval; + u32 new_table_index; + u32 skip_n_vectors; + u32 match_n_vectors; +}; + +/** \brief Classify add / del session request + @param client_index - opaque cookie to identify the sender + @param context - sender context, to match reply w/ request + @param is_add - add session if non-zero, else delete + @param table_index - index of the table to add/del the session, required + @param hit_next_index - for add, hit_next_index of new session, required + @param opaque_index - for add, opaque_index of new session + @param advance -for add, advance value for session + @param action - + 0: no action (by default) + metadata is not used. + 1: Classified IP packets will be looked up from the + specified ipv4 fib table (configured by metadata as VRF id). + Only valid for L3 input ACL node + 2: Classified IP packets will be looked up from the + specified ipv6 fib table (configured by metadata as VRF id). + Only valid for L3 input ACL node + @param metadata - valid only if action != 0 + VRF id if action is 1 or 2. + @param match[] - for add, match value for session, required +*/ +define classify_add_del_session +{ + u32 client_index; + u32 context; + u8 is_add; + u32 table_index; + u32 hit_next_index; + u32 opaque_index; + i32 advance; + u8 action; + u32 metadata; + u8 match[0]; +}; + +/** \brief Classify add / del session response + @param context - sender context, to match reply w/ request + @param retval - return code for the add/del session request +*/ +define classify_add_del_session_reply +{ + u32 context; + i32 retval; +}; + +/** \brief Set/unset policer classify interface + @param client_index - opaque cookie to identify the sender + @param context - sender context, to match reply w/ request + @param sw_if_index - interface to set/unset policer classify + @param ip4_table_index - ip4 classify table index (~0 for skip) + @param ip6_table_index - ip6 classify table index (~0 for skip) + @param l2_table_index - l2 classify table index (~0 for skip) + @param is_add - Set if non-zero, else unset + Note: User is recommeneded to use just one valid table_index per call. + (ip4_table_index, ip6_table_index, or l2_table_index) +*/ +define policer_classify_set_interface +{ + u32 client_index; + u32 context; + u32 sw_if_index; + u32 ip4_table_index; + u32 ip6_table_index; + u32 l2_table_index; + u8 is_add; +}; + +/** \brief Set/unset policer classify interface response + @param context - sender context, to match reply w/ request + @param retval - return value for request +*/ +define policer_classify_set_interface_reply +{ + u32 context; + i32 retval; +}; + +/** \brief Get list of policer classify interfaces and tables + @param client_index - opaque cookie to identify the sender + @param context - sender context, to match reply w/ request + @param type - classify table type +*/ +define policer_classify_dump +{ + u32 client_index; + u32 context; + u8 type; +}; + +/** \brief Policer iclassify operational state response. + @param context - sender context, to match reply w/ request + @param sw_if_index - software interface index + @param table_index - classify table index +*/ +define policer_classify_details +{ + u32 context; + u32 sw_if_index; + u32 table_index; +}; + +/** \brief Classify get table IDs request + @param client_index - opaque cookie to identify the sender + @param context - sender context, to match reply w/ request +*/ +define classify_table_ids +{ + u32 client_index; + u32 context; +}; + +/** \brief Reply for classify get table IDs request + @param context - sender context which was passed in the request + @param count - number of ids returned in response + @param ids - array of classify table ids +*/ +define classify_table_ids_reply +{ + u32 context; + i32 retval; + u32 count; + u32 ids[count]; +}; + +/** \brief Classify table ids by interface index request + @param client_index - opaque cookie to identify the sender + @param context - sender context, to match reply w/ request + @param sw_if_index - index of the interface +*/ +define classify_table_by_interface +{ + u32 client_index; + u32 context; + u32 sw_if_index; +}; + +/** \brief Reply for classify table id by interface index request + @param context - sender context which was passed in the request + @param count - number of ids returned in response + @param sw_if_index - index of the interface + @param l2_table_id - l2 classify table index + @param ip4_table_id - ip4 classify table index + @param ip6_table_id - ip6 classify table index +*/ +define classify_table_by_interface_reply +{ + u32 context; + i32 retval; + u32 sw_if_index; + u32 l2_table_id; + u32 ip4_table_id; + u32 ip6_table_id; +}; + +/** \brief Classify table info + @param client_index - opaque cookie to identify the sender + @param context - sender context, to match reply w/ request + @param table_id - classify table index +*/ +define classify_table_info +{ + u32 client_index; + u32 context; + u32 table_id; +}; + +/** \brief Reply for classify table info request + @param context - sender context which was passed in the request + @param count - number of ids returned in response + @param table_id - classify table index + @param nbuckets - number of buckets when adding a table + @param match_n_vectors - number of match vectors + @param skip_n_vectors - number of skip_n_vectors + @param active_sessions - number of sessions (active entries) + @param next_table_index - index of next table + @param miss_next_index - index of miss table + @param mask[] - match mask +*/ +define classify_table_info_reply +{ + u32 context; + i32 retval; + u32 table_id; + u32 nbuckets; + u32 match_n_vectors; + u32 skip_n_vectors; + u32 active_sessions; + u32 next_table_index; + u32 miss_next_index; + u32 mask_length; + u8 mask[mask_length]; +}; + +/** \brief Classify sessions dump request + @param client_index - opaque cookie to identify the sender + @param context - sender context, to match reply w/ request + @param table_id - classify table index +*/ +define classify_session_dump +{ + u32 client_index; + u32 context; + u32 table_id; +}; + +/** \brief Reply for classify table session dump request + @param context - sender context which was passed in the request + @param count - number of ids returned in response + @param table_id - classify table index + @param hit_next_index - hit_next_index of session + @param opaque_index - for add, opaque_index of session + @param advance - advance value of session + @param match[] - match value for session +*/ +define classify_session_details +{ + u32 context; + i32 retval; + u32 table_id; + u32 hit_next_index; + i32 advance; + u32 opaque_index; + u32 match_length; + u8 match[match_length]; +}; + +/** \brief Set/unset flow classify interface + @param client_index - opaque cookie to identify the sender + @param context - sender context, to match reply w/ request + @param sw_if_index - interface to set/unset flow classify + @param ip4_table_index - ip4 classify table index (~0 for skip) + @param ip6_table_index - ip6 classify table index (~0 for skip) + @param l2_table_index - l2 classify table index (~0 for skip) + @param is_add - Set if non-zero, else unset + Note: User is recommeneded to use just one valid table_index per call. + (ip4_table_index, ip6_table_index, or l2_table_index) +*/ +define flow_classify_set_interface { + u32 client_index; + u32 context; + u32 sw_if_index; + u32 ip4_table_index; + u32 ip6_table_index; + u8 is_add; +}; + +/** \brief Set/unset flow classify interface response + @param context - sender context, to match reply w/ request + @param retval - return value for request +*/ +define flow_classify_set_interface_reply { + u32 context; + i32 retval; +}; + +/** \brief Get list of flow classify interfaces and tables + @param client_index - opaque cookie to identify the sender + @param context - sender context, to match reply w/ request + @param type - classify table type +*/ +define flow_classify_dump { + u32 client_index; + u32 context; + u8 type; +}; + +/** \brief Flow classify operational state response. + @param context - sender context, to match reply w/ request + @param sw_if_index - software interface index + @param table_index - classify table index +*/ +define flow_classify_details { + u32 context; + u32 sw_if_index; + u32 table_index; +}; + +/* + * Local Variables: + * eval: (c-set-style "gnu") + * End: + */ + \ No newline at end of file diff --git a/src/vnet/classify/classify_api.c b/src/vnet/classify/classify_api.c new file mode 100644 index 00000000..77a8b434 --- /dev/null +++ b/src/vnet/classify/classify_api.c @@ -0,0 +1,555 @@ +/* + *------------------------------------------------------------------ + * classify_api.c - classify api + * + * Copyright (c) 2016 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + *------------------------------------------------------------------ + */ + +#include +#include + +#include +#include + +#include +#include +#include +#include + +#include + +#define vl_typedefs /* define message structures */ +#include +#undef vl_typedefs + +#define vl_endianfun /* define message structures */ +#include +#undef vl_endianfun + +/* instantiate all the print functions we know about */ +#define vl_print(handle, ...) vlib_cli_output (handle, __VA_ARGS__) +#define vl_printfun +#include +#undef vl_printfun + +#include + +#define foreach_vpe_api_msg \ +_(CLASSIFY_ADD_DEL_TABLE, classify_add_del_table) \ +_(CLASSIFY_ADD_DEL_SESSION, classify_add_del_session) \ +_(CLASSIFY_TABLE_IDS,classify_table_ids) \ +_(CLASSIFY_TABLE_BY_INTERFACE, classify_table_by_interface) \ +_(CLASSIFY_TABLE_INFO,classify_table_info) \ +_(CLASSIFY_SESSION_DUMP,classify_session_dump) \ +_(CLASSIFY_SESSION_DETAILS,classify_session_details) \ +_(POLICER_CLASSIFY_SET_INTERFACE, policer_classify_set_interface) \ +_(POLICER_CLASSIFY_DUMP, policer_classify_dump) \ +_(FLOW_CLASSIFY_SET_INTERFACE, flow_classify_set_interface) \ +_(FLOW_CLASSIFY_DUMP, flow_classify_dump) + +#define foreach_classify_add_del_table_field \ +_(table_index) \ +_(nbuckets) \ +_(memory_size) \ +_(skip_n_vectors) \ +_(match_n_vectors) \ +_(next_table_index) \ +_(miss_next_index) \ +_(current_data_flag) \ +_(current_data_offset) + +static void vl_api_classify_add_del_table_t_handler + (vl_api_classify_add_del_table_t * mp) +{ + vl_api_classify_add_del_table_reply_t *rmp; + vnet_classify_main_t *cm = &vnet_classify_main; + vnet_classify_table_t *t; + int rv; + +#define _(a) u32 a; + foreach_classify_add_del_table_field; +#undef _ + +#define _(a) a = ntohl(mp->a); + foreach_classify_add_del_table_field; +#undef _ + + /* The underlying API fails silently, on purpose, so check here */ + if (mp->is_add == 0) /* delete */ + { + if (pool_is_free_index (cm->tables, table_index)) + { + rv = VNET_API_ERROR_NO_SUCH_TABLE; + goto out; + } + } + else /* add or update */ + { + if (table_index != ~0 && pool_is_free_index (cm->tables, table_index)) + table_index = ~0; + } + + rv = vnet_classify_add_del_table + (cm, mp->mask, nbuckets, memory_size, + skip_n_vectors, match_n_vectors, + next_table_index, miss_next_index, &table_index, + current_data_flag, current_data_offset, mp->is_add, mp->del_chain); + +out: + /* *INDENT-OFF* */ + REPLY_MACRO2(VL_API_CLASSIFY_ADD_DEL_TABLE_REPLY, + ({ + if (rv == 0 && mp->is_add) + { + t = pool_elt_at_index (cm->tables, table_index); + rmp->skip_n_vectors = ntohl(t->skip_n_vectors); + rmp->match_n_vectors = ntohl(t->match_n_vectors); + rmp->new_table_index = ntohl(table_index); + } + else + { + rmp->skip_n_vectors = ~0; + rmp->match_n_vectors = ~0; + rmp->new_table_index = ~0; + } + })); + /* *INDENT-ON* */ +} + +static void vl_api_classify_add_del_session_t_handler + (vl_api_classify_add_del_session_t * mp) +{ + vnet_classify_main_t *cm = &vnet_classify_main; + vl_api_classify_add_del_session_reply_t *rmp; + int rv; + u32 table_index, hit_next_index, opaque_index, metadata; + i32 advance; + u8 action; + + table_index = ntohl (mp->table_index); + hit_next_index = ntohl (mp->hit_next_index); + opaque_index = ntohl (mp->opaque_index); + advance = ntohl (mp->advance); + action = mp->action; + metadata = ntohl (mp->metadata); + + rv = vnet_classify_add_del_session + (cm, table_index, mp->match, hit_next_index, opaque_index, + advance, action, metadata, mp->is_add); + + REPLY_MACRO (VL_API_CLASSIFY_ADD_DEL_SESSION_REPLY); +} + +static void + vl_api_policer_classify_set_interface_t_handler + (vl_api_policer_classify_set_interface_t * mp) +{ + vlib_main_t *vm = vlib_get_main (); + vl_api_policer_classify_set_interface_reply_t *rmp; + int rv; + u32 sw_if_index, ip4_table_index, ip6_table_index, l2_table_index; + + ip4_table_index = ntohl (mp->ip4_table_index); + ip6_table_index = ntohl (mp->ip6_table_index); + l2_table_index = ntohl (mp->l2_table_index); + sw_if_index = ntohl (mp->sw_if_index); + + VALIDATE_SW_IF_INDEX (mp); + + rv = vnet_set_policer_classify_intfc (vm, sw_if_index, ip4_table_index, + ip6_table_index, l2_table_index, + mp->is_add); + + BAD_SW_IF_INDEX_LABEL; + + REPLY_MACRO (VL_API_POLICER_CLASSIFY_SET_INTERFACE_REPLY); +} + +static void +send_policer_classify_details (u32 sw_if_index, + u32 table_index, + unix_shared_memory_queue_t * q, u32 context) +{ + vl_api_policer_classify_details_t *mp; + + mp = vl_msg_api_alloc (sizeof (*mp)); + memset (mp, 0, sizeof (*mp)); + mp->_vl_msg_id = ntohs (VL_API_POLICER_CLASSIFY_DETAILS); + mp->context = context; + mp->sw_if_index = htonl (sw_if_index); + mp->table_index = htonl (table_index); + + vl_msg_api_send_shmem (q, (u8 *) & mp); +} + +static void +vl_api_policer_classify_dump_t_handler (vl_api_policer_classify_dump_t * mp) +{ + unix_shared_memory_queue_t *q; + policer_classify_main_t *pcm = &policer_classify_main; + u32 *vec_tbl; + int i; + + q = vl_api_client_index_to_input_queue (mp->client_index); + if (q == 0) + return; + + vec_tbl = pcm->classify_table_index_by_sw_if_index[mp->type]; + + if (vec_len (vec_tbl)) + { + for (i = 0; i < vec_len (vec_tbl); i++) + { + if (vec_elt (vec_tbl, i) == ~0) + continue; + + send_policer_classify_details (i, vec_elt (vec_tbl, i), q, + mp->context); + } + } +} + +static void +vl_api_classify_table_ids_t_handler (vl_api_classify_table_ids_t * mp) +{ + unix_shared_memory_queue_t *q; + + q = vl_api_client_index_to_input_queue (mp->client_index); + if (q == 0) + return; + + vnet_classify_main_t *cm = &vnet_classify_main; + vnet_classify_table_t *t; + u32 *table_ids = 0; + u32 count; + + /* *INDENT-OFF* */ + pool_foreach (t, cm->tables, + ({ + vec_add1 (table_ids, ntohl(t - cm->tables)); + })); + /* *INDENT-ON* */ + count = vec_len (table_ids); + + vl_api_classify_table_ids_reply_t *rmp; + rmp = vl_msg_api_alloc_as_if_client (sizeof (*rmp) + count * sizeof (u32)); + rmp->_vl_msg_id = ntohs (VL_API_CLASSIFY_TABLE_IDS_REPLY); + rmp->context = mp->context; + rmp->count = ntohl (count); + clib_memcpy (rmp->ids, table_ids, count * sizeof (u32)); + rmp->retval = 0; + + vl_msg_api_send_shmem (q, (u8 *) & rmp); + + vec_free (table_ids); +} + +static void + vl_api_classify_table_by_interface_t_handler + (vl_api_classify_table_by_interface_t * mp) +{ + vl_api_classify_table_by_interface_reply_t *rmp; + int rv = 0; + + u32 sw_if_index = ntohl (mp->sw_if_index); + u32 *acl = 0; + + vec_validate (acl, INPUT_ACL_N_TABLES - 1); + vec_set (acl, ~0); + + VALIDATE_SW_IF_INDEX (mp); + + input_acl_main_t *am = &input_acl_main; + + int if_idx; + u32 type; + + for (type = 0; type < INPUT_ACL_N_TABLES; type++) + { + u32 *vec_tbl = am->classify_table_index_by_sw_if_index[type]; + if (vec_len (vec_tbl)) + { + for (if_idx = 0; if_idx < vec_len (vec_tbl); if_idx++) + { + if (vec_elt (vec_tbl, if_idx) == ~0 || sw_if_index != if_idx) + { + continue; + } + acl[type] = vec_elt (vec_tbl, if_idx); + } + } + } + + BAD_SW_IF_INDEX_LABEL; + + /* *INDENT-OFF* */ + REPLY_MACRO2(VL_API_CLASSIFY_TABLE_BY_INTERFACE_REPLY, + ({ + rmp->sw_if_index = ntohl(sw_if_index); + rmp->l2_table_id = ntohl(acl[INPUT_ACL_TABLE_L2]); + rmp->ip4_table_id = ntohl(acl[INPUT_ACL_TABLE_IP4]); + rmp->ip6_table_id = ntohl(acl[INPUT_ACL_TABLE_IP6]); + })); + /* *INDENT-ON* */ + vec_free (acl); +} + +static void +vl_api_classify_table_info_t_handler (vl_api_classify_table_info_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_classify_table_info_reply_t *rmp = 0; + + vnet_classify_main_t *cm = &vnet_classify_main; + u32 table_id = ntohl (mp->table_id); + vnet_classify_table_t *t; + + /* *INDENT-OFF* */ + pool_foreach (t, cm->tables, + ({ + if (table_id == t - cm->tables) + { + rmp = vl_msg_api_alloc_as_if_client + (sizeof (*rmp) + t->match_n_vectors * sizeof (u32x4)); + rmp->_vl_msg_id = ntohs (VL_API_CLASSIFY_TABLE_INFO_REPLY); + rmp->context = mp->context; + rmp->table_id = ntohl(table_id); + rmp->nbuckets = ntohl(t->nbuckets); + rmp->match_n_vectors = ntohl(t->match_n_vectors); + rmp->skip_n_vectors = ntohl(t->skip_n_vectors); + rmp->active_sessions = ntohl(t->active_elements); + rmp->next_table_index = ntohl(t->next_table_index); + rmp->miss_next_index = ntohl(t->miss_next_index); + rmp->mask_length = ntohl(t->match_n_vectors * sizeof (u32x4)); + clib_memcpy(rmp->mask, t->mask, t->match_n_vectors * sizeof(u32x4)); + rmp->retval = 0; + break; + } + })); + /* *INDENT-ON* */ + + if (rmp == 0) + { + rmp = vl_msg_api_alloc (sizeof (*rmp)); + rmp->_vl_msg_id = ntohs ((VL_API_CLASSIFY_TABLE_INFO_REPLY)); + rmp->context = mp->context; + rmp->retval = ntohl (VNET_API_ERROR_CLASSIFY_TABLE_NOT_FOUND); + } + + vl_msg_api_send_shmem (q, (u8 *) & rmp); +} + +static void +vl_api_classify_session_details_t_handler (vl_api_classify_session_details_t * + mp) +{ + clib_warning ("BUG"); +} + +static void +send_classify_session_details (unix_shared_memory_queue_t * q, + u32 table_id, + u32 match_length, + vnet_classify_entry_t * e, u32 context) +{ + vl_api_classify_session_details_t *rmp; + + rmp = vl_msg_api_alloc (sizeof (*rmp)); + memset (rmp, 0, sizeof (*rmp)); + rmp->_vl_msg_id = ntohs (VL_API_CLASSIFY_SESSION_DETAILS); + rmp->context = context; + rmp->table_id = ntohl (table_id); + rmp->hit_next_index = ntohl (e->next_index); + rmp->advance = ntohl (e->advance); + rmp->opaque_index = ntohl (e->opaque_index); + rmp->match_length = ntohl (match_length); + clib_memcpy (rmp->match, e->key, match_length); + + vl_msg_api_send_shmem (q, (u8 *) & rmp); +} + +static void +vl_api_classify_session_dump_t_handler (vl_api_classify_session_dump_t * mp) +{ + vnet_classify_main_t *cm = &vnet_classify_main; + unix_shared_memory_queue_t *q; + + u32 table_id = ntohl (mp->table_id); + vnet_classify_table_t *t; + + q = vl_api_client_index_to_input_queue (mp->client_index); + if (!q) + return; + + /* *INDENT-OFF* */ + pool_foreach (t, cm->tables, + ({ + if (table_id == t - cm->tables) + { + vnet_classify_bucket_t * b; + vnet_classify_entry_t * v, * save_v; + int i, j, k; + + for (i = 0; i < t->nbuckets; i++) + { + b = &t->buckets [i]; + if (b->offset == 0) + continue; + + save_v = vnet_classify_get_entry (t, b->offset); + for (j = 0; j < (1<log2_pages); j++) + { + for (k = 0; k < t->entries_per_page; k++) + { + v = vnet_classify_entry_at_index + (t, save_v, j*t->entries_per_page + k); + if (vnet_classify_entry_is_free (v)) + continue; + + send_classify_session_details + (q, table_id, t->match_n_vectors * sizeof (u32x4), + v, mp->context); + } + } + } + break; + } + })); + /* *INDENT-ON* */ +} + +static void + vl_api_flow_classify_set_interface_t_handler + (vl_api_flow_classify_set_interface_t * mp) +{ + vlib_main_t *vm = vlib_get_main (); + vl_api_flow_classify_set_interface_reply_t *rmp; + int rv; + u32 sw_if_index, ip4_table_index, ip6_table_index; + + ip4_table_index = ntohl (mp->ip4_table_index); + ip6_table_index = ntohl (mp->ip6_table_index); + sw_if_index = ntohl (mp->sw_if_index); + + VALIDATE_SW_IF_INDEX (mp); + + rv = vnet_set_flow_classify_intfc (vm, sw_if_index, ip4_table_index, + ip6_table_index, mp->is_add); + + BAD_SW_IF_INDEX_LABEL; + + REPLY_MACRO (VL_API_FLOW_CLASSIFY_SET_INTERFACE_REPLY); +} + +static void +send_flow_classify_details (u32 sw_if_index, + u32 table_index, + unix_shared_memory_queue_t * q, u32 context) +{ + vl_api_flow_classify_details_t *mp; + + mp = vl_msg_api_alloc (sizeof (*mp)); + memset (mp, 0, sizeof (*mp)); + mp->_vl_msg_id = ntohs (VL_API_FLOW_CLASSIFY_DETAILS); + mp->context = context; + mp->sw_if_index = htonl (sw_if_index); + mp->table_index = htonl (table_index); + + vl_msg_api_send_shmem (q, (u8 *) & mp); +} + +static void +vl_api_flow_classify_dump_t_handler (vl_api_flow_classify_dump_t * mp) +{ + unix_shared_memory_queue_t *q; + flow_classify_main_t *pcm = &flow_classify_main; + u32 *vec_tbl; + int i; + + q = vl_api_client_index_to_input_queue (mp->client_index); + if (q == 0) + return; + + vec_tbl = pcm->classify_table_index_by_sw_if_index[mp->type]; + + if (vec_len (vec_tbl)) + { + for (i = 0; i < vec_len (vec_tbl); i++) + { + if (vec_elt (vec_tbl, i) == ~0) + continue; + + send_flow_classify_details (i, vec_elt (vec_tbl, i), q, + mp->context); + } + } +} + +/* + * classify_api_hookup + * Add vpe's API message handlers to the table. + * vlib has alread mapped shared memory and + * added the client registration handlers. + * See .../vlib-api/vlibmemory/memclnt_vlib.c:memclnt_process() + */ +#define vl_msg_name_crc_list +#include +#undef vl_msg_name_crc_list + +static void +setup_message_id_table (api_main_t * am) +{ +#define _(id,n,crc) vl_msg_api_add_msg_name_crc (am, #n "_" #crc, id); + foreach_vl_msg_name_crc_classify; +#undef _ +} + +static clib_error_t * +classify_api_hookup (vlib_main_t * vm) +{ + api_main_t *am = &api_main; + +#define _(N,n) \ + vl_msg_api_set_handlers(VL_API_##N, #n, \ + vl_api_##n##_t_handler, \ + vl_noop_handler, \ + vl_api_##n##_t_endian, \ + vl_api_##n##_t_print, \ + sizeof(vl_api_##n##_t), 1); + foreach_vpe_api_msg; +#undef _ + + /* + * Set up the (msg_name, crc, message-id) table + */ + setup_message_id_table (am); + + return 0; +} + +VLIB_API_INIT_FUNCTION (classify_api_hookup); + +/* + * fd.io coding-style-patch-verification: ON + * + * Local Variables: + * eval: (c-set-style "gnu") + * End: + */ diff --git a/src/vnet/vnet_all_api_h.h b/src/vnet/vnet_all_api_h.h index 1024f92c..c7789650 100644 --- a/src/vnet/vnet_all_api_h.h +++ b/src/vnet/vnet_all_api_h.h @@ -52,6 +52,7 @@ #include #include #include +#include /* * fd.io coding-style-patch-verification: ON diff --git a/src/vpp/api/api.c b/src/vpp/api/api.c index 3afc3383..6e7e9c10 100644 --- a/src/vpp/api/api.c +++ b/src/vpp/api/api.c @@ -62,10 +62,7 @@ #include #include #include -#include #include -#include -#include #include #include #include @@ -139,8 +136,6 @@ _(CLI_REQUEST, cli_request) \ _(CLI_INBAND, cli_inband) \ _(SET_ARP_NEIGHBOR_LIMIT, set_arp_neighbor_limit) \ _(L2_PATCH_ADD_DEL, l2_patch_add_del) \ -_(CLASSIFY_ADD_DEL_TABLE, classify_add_del_table) \ -_(CLASSIFY_ADD_DEL_SESSION, classify_add_del_session) \ _(CLASSIFY_SET_INTERFACE_IP_TABLE, classify_set_interface_ip_table) \ _(CLASSIFY_SET_INTERFACE_L2_TABLES, classify_set_interface_l2_tables) \ _(GET_NODE_INDEX, get_node_index) \ @@ -165,13 +160,6 @@ _(IOAM_ENABLE, ioam_enable) \ _(IOAM_DISABLE, ioam_disable) \ _(POLICER_ADD_DEL, policer_add_del) \ _(POLICER_DUMP, policer_dump) \ -_(POLICER_CLASSIFY_SET_INTERFACE, policer_classify_set_interface) \ -_(POLICER_CLASSIFY_DUMP, policer_classify_dump) \ -_(CLASSIFY_TABLE_IDS,classify_table_ids) \ -_(CLASSIFY_TABLE_BY_INTERFACE, classify_table_by_interface) \ -_(CLASSIFY_TABLE_INFO,classify_table_info) \ -_(CLASSIFY_SESSION_DUMP,classify_session_dump) \ -_(CLASSIFY_SESSION_DETAILS,classify_session_details) \ _(SET_IPFIX_EXPORTER, set_ipfix_exporter) \ _(IPFIX_EXPORTER_DUMP, ipfix_exporter_dump) \ _(SET_IPFIX_CLASSIFY_STREAM, set_ipfix_classify_stream) \ @@ -189,8 +177,6 @@ _(IP_SOURCE_AND_PORT_RANGE_CHECK_INTERFACE_ADD_DEL, \ _(DELETE_SUBIF, delete_subif) \ _(L2_INTERFACE_PBB_TAG_REWRITE, l2_interface_pbb_tag_rewrite) \ _(PUNT, punt) \ -_(FLOW_CLASSIFY_SET_INTERFACE, flow_classify_set_interface) \ -_(FLOW_CLASSIFY_DUMP, flow_classify_dump) \ _(FEATURE_ENABLE_DISABLE, feature_enable_disable) #define QUOTE_(x) #x @@ -1412,99 +1398,6 @@ 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); } -#define foreach_classify_add_del_table_field \ -_(table_index) \ -_(nbuckets) \ -_(memory_size) \ -_(skip_n_vectors) \ -_(match_n_vectors) \ -_(next_table_index) \ -_(miss_next_index) \ -_(current_data_flag) \ -_(current_data_offset) - -static void vl_api_classify_add_del_table_t_handler - (vl_api_classify_add_del_table_t * mp) -{ - vl_api_classify_add_del_table_reply_t *rmp; - vnet_classify_main_t *cm = &vnet_classify_main; - vnet_classify_table_t *t; - int rv; - -#define _(a) u32 a; - foreach_classify_add_del_table_field; -#undef _ - -#define _(a) a = ntohl(mp->a); - foreach_classify_add_del_table_field; -#undef _ - - /* The underlying API fails silently, on purpose, so check here */ - if (mp->is_add == 0) /* delete */ - { - if (pool_is_free_index (cm->tables, table_index)) - { - rv = VNET_API_ERROR_NO_SUCH_TABLE; - goto out; - } - } - else /* add or update */ - { - if (table_index != ~0 && pool_is_free_index (cm->tables, table_index)) - table_index = ~0; - } - - rv = vnet_classify_add_del_table - (cm, mp->mask, nbuckets, memory_size, - skip_n_vectors, match_n_vectors, - next_table_index, miss_next_index, &table_index, - current_data_flag, current_data_offset, mp->is_add, mp->del_chain); - -out: - /* *INDENT-OFF* */ - REPLY_MACRO2(VL_API_CLASSIFY_ADD_DEL_TABLE_REPLY, - ({ - if (rv == 0 && mp->is_add) - { - t = pool_elt_at_index (cm->tables, table_index); - rmp->skip_n_vectors = ntohl(t->skip_n_vectors); - rmp->match_n_vectors = ntohl(t->match_n_vectors); - rmp->new_table_index = ntohl(table_index); - } - else - { - rmp->skip_n_vectors = ~0; - rmp->match_n_vectors = ~0; - rmp->new_table_index = ~0; - } - })); - /* *INDENT-ON* */ -} - -static void vl_api_classify_add_del_session_t_handler - (vl_api_classify_add_del_session_t * mp) -{ - vnet_classify_main_t *cm = &vnet_classify_main; - vl_api_classify_add_del_session_reply_t *rmp; - int rv; - u32 table_index, hit_next_index, opaque_index, metadata; - i32 advance; - u8 action; - - table_index = ntohl (mp->table_index); - hit_next_index = ntohl (mp->hit_next_index); - opaque_index = ntohl (mp->opaque_index); - advance = ntohl (mp->advance); - action = mp->action; - metadata = ntohl (mp->metadata); - - rv = vnet_classify_add_del_session - (cm, table_index, mp->match, hit_next_index, opaque_index, - advance, action, metadata, mp->is_add); - - REPLY_MACRO (VL_API_CLASSIFY_ADD_DEL_SESSION_REPLY); -} - static void vl_api_classify_set_interface_ip_table_t_handler (vl_api_classify_set_interface_ip_table_t * mp) { @@ -2528,288 +2421,6 @@ vl_api_policer_dump_t_handler (vl_api_policer_dump_t * mp) } } -static void - vl_api_policer_classify_set_interface_t_handler - (vl_api_policer_classify_set_interface_t * mp) -{ - vlib_main_t *vm = vlib_get_main (); - vl_api_policer_classify_set_interface_reply_t *rmp; - int rv; - u32 sw_if_index, ip4_table_index, ip6_table_index, l2_table_index; - - ip4_table_index = ntohl (mp->ip4_table_index); - ip6_table_index = ntohl (mp->ip6_table_index); - l2_table_index = ntohl (mp->l2_table_index); - sw_if_index = ntohl (mp->sw_if_index); - - VALIDATE_SW_IF_INDEX (mp); - - rv = vnet_set_policer_classify_intfc (vm, sw_if_index, ip4_table_index, - ip6_table_index, l2_table_index, - mp->is_add); - - BAD_SW_IF_INDEX_LABEL; - - REPLY_MACRO (VL_API_POLICER_CLASSIFY_SET_INTERFACE_REPLY); -} - -static void -send_policer_classify_details (u32 sw_if_index, - u32 table_index, - unix_shared_memory_queue_t * q, u32 context) -{ - vl_api_policer_classify_details_t *mp; - - mp = vl_msg_api_alloc (sizeof (*mp)); - memset (mp, 0, sizeof (*mp)); - mp->_vl_msg_id = ntohs (VL_API_POLICER_CLASSIFY_DETAILS); - mp->context = context; - mp->sw_if_index = htonl (sw_if_index); - mp->table_index = htonl (table_index); - - vl_msg_api_send_shmem (q, (u8 *) & mp); -} - -static void -vl_api_policer_classify_dump_t_handler (vl_api_policer_classify_dump_t * mp) -{ - unix_shared_memory_queue_t *q; - policer_classify_main_t *pcm = &policer_classify_main; - u32 *vec_tbl; - int i; - - q = vl_api_client_index_to_input_queue (mp->client_index); - if (q == 0) - return; - - vec_tbl = pcm->classify_table_index_by_sw_if_index[mp->type]; - - if (vec_len (vec_tbl)) - { - for (i = 0; i < vec_len (vec_tbl); i++) - { - if (vec_elt (vec_tbl, i) == ~0) - continue; - - send_policer_classify_details (i, vec_elt (vec_tbl, i), q, - mp->context); - } - } -} - -static void -vl_api_classify_table_ids_t_handler (vl_api_classify_table_ids_t * mp) -{ - unix_shared_memory_queue_t *q; - - q = vl_api_client_index_to_input_queue (mp->client_index); - if (q == 0) - return; - - vnet_classify_main_t *cm = &vnet_classify_main; - vnet_classify_table_t *t; - u32 *table_ids = 0; - u32 count; - - /* *INDENT-OFF* */ - pool_foreach (t, cm->tables, - ({ - vec_add1 (table_ids, ntohl(t - cm->tables)); - })); - /* *INDENT-ON* */ - count = vec_len (table_ids); - - vl_api_classify_table_ids_reply_t *rmp; - rmp = vl_msg_api_alloc_as_if_client (sizeof (*rmp) + count * sizeof (u32)); - rmp->_vl_msg_id = ntohs (VL_API_CLASSIFY_TABLE_IDS_REPLY); - rmp->context = mp->context; - rmp->count = ntohl (count); - clib_memcpy (rmp->ids, table_ids, count * sizeof (u32)); - rmp->retval = 0; - - vl_msg_api_send_shmem (q, (u8 *) & rmp); - - vec_free (table_ids); -} - -static void - vl_api_classify_table_by_interface_t_handler - (vl_api_classify_table_by_interface_t * mp) -{ - vl_api_classify_table_by_interface_reply_t *rmp; - int rv = 0; - - u32 sw_if_index = ntohl (mp->sw_if_index); - u32 *acl = 0; - - vec_validate (acl, INPUT_ACL_N_TABLES - 1); - vec_set (acl, ~0); - - VALIDATE_SW_IF_INDEX (mp); - - input_acl_main_t *am = &input_acl_main; - - int if_idx; - u32 type; - - for (type = 0; type < INPUT_ACL_N_TABLES; type++) - { - u32 *vec_tbl = am->classify_table_index_by_sw_if_index[type]; - if (vec_len (vec_tbl)) - { - for (if_idx = 0; if_idx < vec_len (vec_tbl); if_idx++) - { - if (vec_elt (vec_tbl, if_idx) == ~0 || sw_if_index != if_idx) - { - continue; - } - acl[type] = vec_elt (vec_tbl, if_idx); - } - } - } - - BAD_SW_IF_INDEX_LABEL; - - /* *INDENT-OFF* */ - REPLY_MACRO2(VL_API_CLASSIFY_TABLE_BY_INTERFACE_REPLY, - ({ - rmp->sw_if_index = ntohl(sw_if_index); - rmp->l2_table_id = ntohl(acl[INPUT_ACL_TABLE_L2]); - rmp->ip4_table_id = ntohl(acl[INPUT_ACL_TABLE_IP4]); - rmp->ip6_table_id = ntohl(acl[INPUT_ACL_TABLE_IP6]); - })); - /* *INDENT-ON* */ - vec_free (acl); -} - -static void -vl_api_classify_table_info_t_handler (vl_api_classify_table_info_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_classify_table_info_reply_t *rmp = 0; - - vnet_classify_main_t *cm = &vnet_classify_main; - u32 table_id = ntohl (mp->table_id); - vnet_classify_table_t *t; - - /* *INDENT-OFF* */ - pool_foreach (t, cm->tables, - ({ - if (table_id == t - cm->tables) - { - rmp = vl_msg_api_alloc_as_if_client - (sizeof (*rmp) + t->match_n_vectors * sizeof (u32x4)); - rmp->_vl_msg_id = ntohs (VL_API_CLASSIFY_TABLE_INFO_REPLY); - rmp->context = mp->context; - rmp->table_id = ntohl(table_id); - rmp->nbuckets = ntohl(t->nbuckets); - rmp->match_n_vectors = ntohl(t->match_n_vectors); - rmp->skip_n_vectors = ntohl(t->skip_n_vectors); - rmp->active_sessions = ntohl(t->active_elements); - rmp->next_table_index = ntohl(t->next_table_index); - rmp->miss_next_index = ntohl(t->miss_next_index); - rmp->mask_length = ntohl(t->match_n_vectors * sizeof (u32x4)); - clib_memcpy(rmp->mask, t->mask, t->match_n_vectors * sizeof(u32x4)); - rmp->retval = 0; - break; - } - })); - /* *INDENT-ON* */ - - if (rmp == 0) - { - rmp = vl_msg_api_alloc (sizeof (*rmp)); - rmp->_vl_msg_id = ntohs ((VL_API_CLASSIFY_TABLE_INFO_REPLY)); - rmp->context = mp->context; - rmp->retval = ntohl (VNET_API_ERROR_CLASSIFY_TABLE_NOT_FOUND); - } - - vl_msg_api_send_shmem (q, (u8 *) & rmp); -} - -static void -vl_api_classify_session_details_t_handler (vl_api_classify_session_details_t * - mp) -{ - clib_warning ("BUG"); -} - -static void -send_classify_session_details (unix_shared_memory_queue_t * q, - u32 table_id, - u32 match_length, - vnet_classify_entry_t * e, u32 context) -{ - vl_api_classify_session_details_t *rmp; - - rmp = vl_msg_api_alloc (sizeof (*rmp)); - memset (rmp, 0, sizeof (*rmp)); - rmp->_vl_msg_id = ntohs (VL_API_CLASSIFY_SESSION_DETAILS); - rmp->context = context; - rmp->table_id = ntohl (table_id); - rmp->hit_next_index = ntohl (e->next_index); - rmp->advance = ntohl (e->advance); - rmp->opaque_index = ntohl (e->opaque_index); - rmp->match_length = ntohl (match_length); - clib_memcpy (rmp->match, e->key, match_length); - - vl_msg_api_send_shmem (q, (u8 *) & rmp); -} - -static void -vl_api_classify_session_dump_t_handler (vl_api_classify_session_dump_t * mp) -{ - vnet_classify_main_t *cm = &vnet_classify_main; - unix_shared_memory_queue_t *q; - - u32 table_id = ntohl (mp->table_id); - vnet_classify_table_t *t; - - q = vl_api_client_index_to_input_queue (mp->client_index); - if (!q) - return; - - /* *INDENT-OFF* */ - pool_foreach (t, cm->tables, - ({ - if (table_id == t - cm->tables) - { - vnet_classify_bucket_t * b; - vnet_classify_entry_t * v, * save_v; - int i, j, k; - - for (i = 0; i < t->nbuckets; i++) - { - b = &t->buckets [i]; - if (b->offset == 0) - continue; - - save_v = vnet_classify_get_entry (t, b->offset); - for (j = 0; j < (1<log2_pages); j++) - { - for (k = 0; k < t->entries_per_page; k++) - { - v = vnet_classify_entry_at_index - (t, save_v, j*t->entries_per_page + k); - if (vnet_classify_entry_is_free (v)) - continue; - - send_classify_session_details - (q, table_id, t->match_n_vectors * sizeof (u32x4), - v, mp->context); - } - } - } - break; - } - })); - /* *INDENT-ON* */ -} static void vl_api_set_ipfix_exporter_t_handler (vl_api_set_ipfix_exporter_t * mp) @@ -3399,73 +3010,6 @@ vl_api_punt_t_handler (vl_api_punt_t * mp) REPLY_MACRO (VL_API_PUNT_REPLY); } -static void - vl_api_flow_classify_set_interface_t_handler - (vl_api_flow_classify_set_interface_t * mp) -{ - vlib_main_t *vm = vlib_get_main (); - vl_api_flow_classify_set_interface_reply_t *rmp; - int rv; - u32 sw_if_index, ip4_table_index, ip6_table_index; - - ip4_table_index = ntohl (mp->ip4_table_index); - ip6_table_index = ntohl (mp->ip6_table_index); - sw_if_index = ntohl (mp->sw_if_index); - - VALIDATE_SW_IF_INDEX (mp); - - rv = vnet_set_flow_classify_intfc (vm, sw_if_index, ip4_table_index, - ip6_table_index, mp->is_add); - - BAD_SW_IF_INDEX_LABEL; - - REPLY_MACRO (VL_API_FLOW_CLASSIFY_SET_INTERFACE_REPLY); -} - -static void -send_flow_classify_details (u32 sw_if_index, - u32 table_index, - unix_shared_memory_queue_t * q, u32 context) -{ - vl_api_flow_classify_details_t *mp; - - mp = vl_msg_api_alloc (sizeof (*mp)); - memset (mp, 0, sizeof (*mp)); - mp->_vl_msg_id = ntohs (VL_API_FLOW_CLASSIFY_DETAILS); - mp->context = context; - mp->sw_if_index = htonl (sw_if_index); - mp->table_index = htonl (table_index); - - vl_msg_api_send_shmem (q, (u8 *) & mp); -} - -static void -vl_api_flow_classify_dump_t_handler (vl_api_flow_classify_dump_t * mp) -{ - unix_shared_memory_queue_t *q; - flow_classify_main_t *pcm = &flow_classify_main; - u32 *vec_tbl; - int i; - - q = vl_api_client_index_to_input_queue (mp->client_index); - if (q == 0) - return; - - vec_tbl = pcm->classify_table_index_by_sw_if_index[mp->type]; - - if (vec_len (vec_tbl)) - { - for (i = 0; i < vec_len (vec_tbl); i++) - { - if (vec_elt (vec_tbl, i) == ~0) - continue; - - send_flow_classify_details (i, vec_elt (vec_tbl, i), q, - mp->context); - } - } -} - static void vl_api_feature_enable_disable_t_handler (vl_api_feature_enable_disable_t * mp) { diff --git a/src/vpp/api/vpe.api b/src/vpp/api/vpe.api index e784fa01..a9b34d1f 100644 --- a/src/vpp/api/vpe.api +++ b/src/vpp/api/vpe.api @@ -26,7 +26,7 @@ * IP APIs: see .../src/vnet/ip/{ip.api, ip_api.c} * TAP APIs: see .../src/vnet/unix/{tap.api, tap_api.c} * VXLAN APIs: see .../src/vnet/vxlan/{vxlan.api, vxlan_api.c} - * AF-PACKET APIs: ... see /vnet/devices/af_packet/{af_packet.api, af_packet_api.c} + * AF-PACKET APIs: see ... /vnet/devices/af_packet/{af_packet.api, af_packet_api.c} * NETMAP APIs: see ... /src/vnet/devices/netmap/{netmap.api, netmap_api.c} * VHOST-USER APIs: see .../vnet/devices/virtio/{vhost_user.api, vhost_user_api.c} * VXLAN GPE APIs: see .../src/vnet/vxlan-gpe/{vxlan_gpe.api, vxlan_gpe_api.c} @@ -40,7 +40,8 @@ * LISP-GPE APIs: see .../src/vnet/lisp-gpe/{lisp_gpe.api, lisp_gpe_api.c} * MPLS APIs: see .../src/vnet/mpls/{mpls.api, mpls_api.c} * SR APIs: see .../src/vnet/sr/{sr.api, sr_api.c} - * DPDK APIs: ... see /src/vnet/devices/dpdk/{dpdk.api, dpdk_api.c} + * DPDK APIs: see ... /src/vnet/devices/dpdk/{dpdk.api, dpdk_api.c} + * CLASSIFY APIs: see ... /src/vnet/classify/{classify.api, classify_api.c} */ /** \brief Create a new subinterface with the given vlan id @@ -685,109 +686,6 @@ define bd_ip_mac_add_del_reply i32 retval; }; -/** \brief Add/Delete classification table request - @param client_index - opaque cookie to identify the sender - @param context - sender context, to match reply w/ request - @param is_add- if non-zero add the table, else delete it - @param del_chain - if non-zero delete the whole chain of tables - @param table_index - if add, reuturns index of the created table, else specifies the table to delete - @param nbuckets - number of buckets when adding a table - @param memory_size - memory size when adding a table - @param match_n_vectors - number of match vectors - @param next_table_index - index of next table - @param miss_next_index - index of miss table - @param current_data_flag - option to use current node's packet payload - as the starting point from where packets are classified, - This option is only valid for L2/L3 input ACL for now. - 0: by default, classify data from the buffer's start location - 1: classify packets from VPP node’s current data pointer - @param current_data_offset - a signed value to shift the start location of - the packet to be classified - For example, if input IP ACL node is used, L2 header’s first byte - can be accessible by configuring current_data_offset to -14 - if there is no vlan tag. - This is valid only if current_data_flag is set to 1. - @param mask[] - match mask -*/ -define classify_add_del_table -{ - u32 client_index; - u32 context; - u8 is_add; - u8 del_chain; - u32 table_index; - u32 nbuckets; - u32 memory_size; - u32 skip_n_vectors; - u32 match_n_vectors; - u32 next_table_index; - u32 miss_next_index; - u32 current_data_flag; - i32 current_data_offset; - u8 mask[0]; -}; - -/** \brief Add/Delete classification table response - @param context - sender context, to match reply w/ request - @param retval - return code for the table add/del requst - @param new_table_index - for add, returned index of the new table - @param skip_n_vectors - for add, returned value of skip_n_vectors in table - @param match_n_vectors -for add, returned value of match_n_vectors in table -*/ -define classify_add_del_table_reply -{ - u32 context; - i32 retval; - u32 new_table_index; - u32 skip_n_vectors; - u32 match_n_vectors; -}; - -/** \brief Classify add / del session request - @param client_index - opaque cookie to identify the sender - @param context - sender context, to match reply w/ request - @param is_add - add session if non-zero, else delete - @param table_index - index of the table to add/del the session, required - @param hit_next_index - for add, hit_next_index of new session, required - @param opaque_index - for add, opaque_index of new session - @param advance -for add, advance value for session - @param action - - 0: no action (by default) - metadata is not used. - 1: Classified IP packets will be looked up from the - specified ipv4 fib table (configured by metadata as VRF id). - Only valid for L3 input ACL node - 2: Classified IP packets will be looked up from the - specified ipv6 fib table (configured by metadata as VRF id). - Only valid for L3 input ACL node - @param metadata - valid only if action != 0 - VRF id if action is 1 or 2. - @param match[] - for add, match value for session, required -*/ -define classify_add_del_session -{ - u32 client_index; - u32 context; - u8 is_add; - u32 table_index; - u32 hit_next_index; - u32 opaque_index; - i32 advance; - u8 action; - u32 metadata; - u8 match[0]; -}; - -/** \brief Classify add / del session response - @param context - sender context, to match reply w/ request - @param retval - return code for the add/del session request -*/ -define classify_add_del_session_reply -{ - u32 context; - i32 retval; -}; - /** \brief Set/unset the classification table for an interface request @param client_index - opaque cookie to identify the sender @param context - sender context, to match reply w/ request @@ -1491,187 +1389,6 @@ define policer_details u64 last_update_time; }; -/** \brief Set/unset policer classify interface - @param client_index - opaque cookie to identify the sender - @param context - sender context, to match reply w/ request - @param sw_if_index - interface to set/unset policer classify - @param ip4_table_index - ip4 classify table index (~0 for skip) - @param ip6_table_index - ip6 classify table index (~0 for skip) - @param l2_table_index - l2 classify table index (~0 for skip) - @param is_add - Set if non-zero, else unset - Note: User is recommeneded to use just one valid table_index per call. - (ip4_table_index, ip6_table_index, or l2_table_index) -*/ -define policer_classify_set_interface -{ - u32 client_index; - u32 context; - u32 sw_if_index; - u32 ip4_table_index; - u32 ip6_table_index; - u32 l2_table_index; - u8 is_add; -}; - -/** \brief Set/unset policer classify interface response - @param context - sender context, to match reply w/ request - @param retval - return value for request -*/ -define policer_classify_set_interface_reply -{ - u32 context; - i32 retval; -}; - -/** \brief Get list of policer classify interfaces and tables - @param client_index - opaque cookie to identify the sender - @param context - sender context, to match reply w/ request - @param type - classify table type -*/ -define policer_classify_dump -{ - u32 client_index; - u32 context; - u8 type; -}; - -/** \brief Policer iclassify operational state response. - @param context - sender context, to match reply w/ request - @param sw_if_index - software interface index - @param table_index - classify table index -*/ -define policer_classify_details -{ - u32 context; - u32 sw_if_index; - u32 table_index; -}; - -/** \brief Classify get table IDs request - @param client_index - opaque cookie to identify the sender - @param context - sender context, to match reply w/ request -*/ -define classify_table_ids -{ - u32 client_index; - u32 context; -}; - -/** \brief Reply for classify get table IDs request - @param context - sender context which was passed in the request - @param count - number of ids returned in response - @param ids - array of classify table ids -*/ -define classify_table_ids_reply -{ - u32 context; - i32 retval; - u32 count; - u32 ids[count]; -}; - -/** \brief Classify table ids by interface index request - @param client_index - opaque cookie to identify the sender - @param context - sender context, to match reply w/ request - @param sw_if_index - index of the interface -*/ -define classify_table_by_interface -{ - u32 client_index; - u32 context; - u32 sw_if_index; -}; - -/** \brief Reply for classify table id by interface index request - @param context - sender context which was passed in the request - @param count - number of ids returned in response - @param sw_if_index - index of the interface - @param l2_table_id - l2 classify table index - @param ip4_table_id - ip4 classify table index - @param ip6_table_id - ip6 classify table index -*/ -define classify_table_by_interface_reply -{ - u32 context; - i32 retval; - u32 sw_if_index; - u32 l2_table_id; - u32 ip4_table_id; - u32 ip6_table_id; -}; - -/** \brief Classify table info - @param client_index - opaque cookie to identify the sender - @param context - sender context, to match reply w/ request - @param table_id - classify table index -*/ -define classify_table_info -{ - u32 client_index; - u32 context; - u32 table_id; -}; - -/** \brief Reply for classify table info request - @param context - sender context which was passed in the request - @param count - number of ids returned in response - @param table_id - classify table index - @param nbuckets - number of buckets when adding a table - @param match_n_vectors - number of match vectors - @param skip_n_vectors - number of skip_n_vectors - @param active_sessions - number of sessions (active entries) - @param next_table_index - index of next table - @param miss_next_index - index of miss table - @param mask[] - match mask -*/ -define classify_table_info_reply -{ - u32 context; - i32 retval; - u32 table_id; - u32 nbuckets; - u32 match_n_vectors; - u32 skip_n_vectors; - u32 active_sessions; - u32 next_table_index; - u32 miss_next_index; - u32 mask_length; - u8 mask[mask_length]; -}; - -/** \brief Classify sessions dump request - @param client_index - opaque cookie to identify the sender - @param context - sender context, to match reply w/ request - @param table_id - classify table index -*/ -define classify_session_dump -{ - u32 client_index; - u32 context; - u32 table_id; -}; - -/** \brief Reply for classify table session dump request - @param context - sender context which was passed in the request - @param count - number of ids returned in response - @param table_id - classify table index - @param hit_next_index - hit_next_index of session - @param opaque_index - for add, opaque_index of session - @param advance - advance value of session - @param match[] - match value for session -*/ -define classify_session_details -{ - u32 context; - i32 retval; - u32 table_id; - u32 hit_next_index; - i32 advance; - u32 opaque_index; - u32 match_length; - u8 match[match_length]; -}; - /** \brief Configure IPFIX exporter process request @param client_index - opaque cookie to identify the sender @param context - sender context, to match reply w/ request @@ -1825,57 +1542,6 @@ define ipfix_classify_table_details { u8 transport_protocol; }; -/** \brief Set/unset flow classify interface - @param client_index - opaque cookie to identify the sender - @param context - sender context, to match reply w/ request - @param sw_if_index - interface to set/unset flow classify - @param ip4_table_index - ip4 classify table index (~0 for skip) - @param ip6_table_index - ip6 classify table index (~0 for skip) - @param l2_table_index - l2 classify table index (~0 for skip) - @param is_add - Set if non-zero, else unset - Note: User is recommeneded to use just one valid table_index per call. - (ip4_table_index, ip6_table_index, or l2_table_index) -*/ -define flow_classify_set_interface { - u32 client_index; - u32 context; - u32 sw_if_index; - u32 ip4_table_index; - u32 ip6_table_index; - u8 is_add; -}; - -/** \brief Set/unset flow classify interface response - @param context - sender context, to match reply w/ request - @param retval - return value for request -*/ -define flow_classify_set_interface_reply { - u32 context; - i32 retval; -}; - -/** \brief Get list of flow classify interfaces and tables - @param client_index - opaque cookie to identify the sender - @param context - sender context, to match reply w/ request - @param type - classify table type -*/ -define flow_classify_dump { - u32 client_index; - u32 context; - u8 type; -}; - -/** \brief Flow classify operational state response. - @param context - sender context, to match reply w/ request - @param sw_if_index - software interface index - @param table_index - classify table index -*/ -define flow_classify_details { - u32 context; - u32 sw_if_index; - u32 table_index; -}; - /** \brief Query relative index via node names @param client_index - opaque cookie to identify the sender @param context - sender context, to match reply w/ request -- cgit 1.2.3-korg From 4a3f69c17f26e9790db26d0910befe36b5e9f85f Mon Sep 17 00:00:00 2001 From: Dave Barach Date: Wed, 22 Feb 2017 12:44:56 -0500 Subject: Fix vpp built-in version of api_unformat_sw_if_index(...) Change-Id: I103fe19a1ecbaf3746ec6b957fa1010458cc9fae Signed-off-by: Dave Barach --- src/vat/api_format.c | 50 ++++++++++++++++++++++++++------------- src/vnet/classify/vnet_classify.c | 4 +--- src/vnet/classify/vnet_classify.h | 1 + 3 files changed, 36 insertions(+), 19 deletions(-) (limited to 'src/vnet/classify') diff --git a/src/vat/api_format.c b/src/vat/api_format.c index 039fca06..226e129e 100644 --- a/src/vat/api_format.c +++ b/src/vat/api_format.c @@ -112,6 +112,7 @@ errmsg (char *fmt, ...) vec_free (s); } +#if VPP_API_TEST_BUILTIN == 0 static uword api_unformat_sw_if_index (unformat_input_t * input, va_list * args) { @@ -130,7 +131,6 @@ api_unformat_sw_if_index (unformat_input_t * input, va_list * args) return 1; } -#if VPP_API_TEST_BUILTIN == 0 /* Parse an IP4 address %d.%d.%d.%d. */ uword unformat_ip4_address (unformat_input_t * input, va_list * args) @@ -387,6 +387,21 @@ unformat_ikev2_id_type (unformat_input_t * input, va_list * args) return 0; return 1; } +#else /* VPP_API_TEST_BUILTIN == 1 */ +static uword +api_unformat_sw_if_index (unformat_input_t * input, va_list * args) +{ + vat_main_t *vam __attribute__ ((unused)) = va_arg (*args, vat_main_t *); + vnet_main_t *vnm = vnet_get_main (); + u32 *result = va_arg (*args, u32 *); + u32 sw_if_index; + + if (!unformat (input, "%U", unformat_vnet_sw_interface, vnm, &sw_if_index)) + return 0; + + *result = sw_if_index; + return 1; +} #endif /* VPP_API_TEST_BUILTIN */ static uword @@ -511,6 +526,7 @@ static const char *mfib_flag_long_names[] = MFIB_ENTRY_NAMES_LONG; static const char *mfib_itf_flag_long_names[] = MFIB_ITF_NAMES_LONG; static const char *mfib_itf_flag_names[] = MFIB_ITF_NAMES_SHORT; +#if (VPP_API_TEST_BUILTIN==0) uword unformat_mfib_itf_flags (unformat_input_t * input, va_list * args) { @@ -553,7 +569,6 @@ unformat_mfib_entry_flags (unformat_input_t * input, va_list * args) return (old == *eflags ? 0 : 1); } -#if (VPP_API_TEST_BUILTIN==0) u8 * format_ip4_address (u8 * s, va_list * args) { @@ -8724,6 +8739,12 @@ _(ttl) \ _(protocol) \ _(checksum) +typedef struct +{ + u16 src_port, dst_port; +} tcpudp_header_t; + +#if VPP_API_TEST_BUILTIN == 0 uword unformat_tcp_mask (unformat_input_t * input, va_list * args) { @@ -8806,11 +8827,6 @@ unformat_udp_mask (unformat_input_t * input, va_list * args) return 1; } -typedef struct -{ - u16 src_port, dst_port; -} tcpudp_header_t; - uword unformat_l4_mask (unformat_input_t * input, va_list * args) { @@ -9204,6 +9220,7 @@ unformat_classify_mask (unformat_input_t * input, va_list * args) return 0; } +#endif /* VPP_API_TEST_BUILTIN */ #define foreach_l2_next \ _(drop, DROP) \ @@ -9242,7 +9259,7 @@ _(local, LOCAL) \ _(rewrite, REWRITE) uword -unformat_ip_next_index (unformat_input_t * input, va_list * args) +api_unformat_ip_next_index (unformat_input_t * input, va_list * args) { u32 *miss_next_indexp = va_arg (*args, u32 *); u32 next_index = 0; @@ -9270,7 +9287,7 @@ out: _(deny, DENY) uword -unformat_acl_next_index (unformat_input_t * input, va_list * args) +api_unformat_acl_next_index (unformat_input_t * input, va_list * args) { u32 *miss_next_indexp = va_arg (*args, u32 *); u32 next_index = 0; @@ -9358,13 +9375,13 @@ api_classify_add_del_table (vat_main_t * vam) ; else if (unformat (i, "next-table %d", &next_table_index)) ; - else if (unformat (i, "miss-next %U", unformat_ip_next_index, + else if (unformat (i, "miss-next %U", api_unformat_ip_next_index, &miss_next_index)) ; else if (unformat (i, "l2-miss-next %U", unformat_l2_next_index, &miss_next_index)) ; - else if (unformat (i, "acl-miss-next %U", unformat_acl_next_index, + else if (unformat (i, "acl-miss-next %U", api_unformat_acl_next_index, &miss_next_index)) ; else if (unformat (i, "current-data-flag %d", ¤t_data_flag)) @@ -9421,6 +9438,7 @@ api_classify_add_del_table (vat_main_t * vam) return ret; } +#if VPP_API_TEST_BUILTIN == 0 uword unformat_l4_match (unformat_input_t * input, va_list * args) { @@ -9785,10 +9803,10 @@ unformat_l2_match (unformat_input_t * input, va_list * args) *matchp = match; return 1; } - +#endif uword -unformat_classify_match (unformat_input_t * input, va_list * args) +api_unformat_classify_match (unformat_input_t * input, va_list * args) { u8 **matchp = va_arg (*args, u8 **); u32 skip_n_vectors = va_arg (*args, u32); @@ -9884,13 +9902,13 @@ api_classify_add_del_session (vat_main_t * vam) { if (unformat (i, "del")) is_add = 0; - else if (unformat (i, "hit-next %U", unformat_ip_next_index, + else if (unformat (i, "hit-next %U", api_unformat_ip_next_index, &hit_next_index)) ; else if (unformat (i, "l2-hit-next %U", unformat_l2_next_index, &hit_next_index)) ; - else if (unformat (i, "acl-hit-next %U", unformat_acl_next_index, + else if (unformat (i, "acl-hit-next %U", api_unformat_acl_next_index, &hit_next_index)) ; else if (unformat (i, "policer-hit-next %d", &hit_next_index)) @@ -9903,7 +9921,7 @@ api_classify_add_del_session (vat_main_t * vam) ; else if (unformat (i, "match_n %d", &match_n_vectors)) ; - else if (unformat (i, "match %U", unformat_classify_match, + else if (unformat (i, "match %U", api_unformat_classify_match, &match, skip_n_vectors, match_n_vectors)) ; else if (unformat (i, "advance %d", &advance)) diff --git a/src/vnet/classify/vnet_classify.c b/src/vnet/classify/vnet_classify.c index ce38f9f1..6093e2ac 100644 --- a/src/vnet/classify/vnet_classify.c +++ b/src/vnet/classify/vnet_classify.c @@ -1093,8 +1093,6 @@ uword unformat_l2_mask (unformat_input_t * input, va_list * args) uword unformat_classify_mask (unformat_input_t * input, va_list * args) { - vnet_classify_main_t * CLIB_UNUSED(cm) - = va_arg (*args, vnet_classify_main_t *); u8 ** maskp = va_arg (*args, u8 **); u32 * skipp = va_arg (*args, u32 *); u32 * matchp = va_arg (*args, u32 *); @@ -1417,7 +1415,7 @@ classify_table_command_fn (vlib_main_t * vm, else if (unformat (input, "table %d", &table_index)) ; else if (unformat (input, "mask %U", unformat_classify_mask, - cm, &mask, &skip, &match)) + &mask, &skip, &match)) ; else if (unformat (input, "memory-size %uM", &tmp)) memory_size = tmp<<20; diff --git a/src/vnet/classify/vnet_classify.h b/src/vnet/classify/vnet_classify.h index d0b896ed..2c966632 100644 --- a/src/vnet/classify/vnet_classify.h +++ b/src/vnet/classify/vnet_classify.h @@ -502,6 +502,7 @@ unformat_function_t unformat_ip_next_index; unformat_function_t unformat_ip4_match; unformat_function_t unformat_ip6_match; unformat_function_t unformat_l3_match; +unformat_function_t unformat_l4_match; unformat_function_t unformat_vlan_tag; unformat_function_t unformat_l2_match; unformat_function_t unformat_classify_match; -- cgit 1.2.3-korg From 68b0fb0c620c7451ef1a6380c43c39de6614db51 Mon Sep 17 00:00:00 2001 From: Dave Barach Date: Tue, 28 Feb 2017 15:15:56 -0500 Subject: VPP-598: tcp stack initial commit Change-Id: I49e5ce0aae6e4ff634024387ceaf7dbc432a0351 Signed-off-by: Dave Barach Signed-off-by: Florin Coras --- src/Makefile.am | 1 + src/plugins/ioam/export-common/ioam_export.h | 2 +- src/plugins/ioam/ipfixcollector/ipfixcollector.c | 2 +- src/plugins/ioam/lib-vxlan-gpe/ioam_transit.c | 2 +- src/plugins/snat/in2out.c | 26 +- src/plugins/snat/out2in.c | 24 +- src/scripts/vnet/tcp | 18 +- src/scripts/vnet/udp | 19 + src/scripts/vnet/uri/tcp-setup.sh | 39 + src/scripts/vnet/uri/tcp_server | 4 + src/scripts/vnet/uri/udp | 19 + src/svm.am | 10 +- src/svm/ssvm.c | 16 + src/svm/ssvm.h | 18 +- src/svm/svm_fifo.c | 568 ++++++ src/svm/svm_fifo.h | 157 ++ src/svm/svm_fifo_segment.c | 193 ++ src/svm/svm_fifo_segment.h | 89 + src/svm/test_svm_fifo1.c | 361 ++++ src/uri.am | 22 + src/uri/uri_tcp_test.c | 916 +++++++++ src/uri/uri_udp_test.c | 553 ++++++ src/uri/uri_udp_test2.c | 954 +++++++++ src/uri/uritest.c | 484 +++++ src/vlib/buffer.c | 2 +- src/vlib/buffer.h | 68 + src/vlibmemory/unix_shared_memory_queue.c | 12 +- src/vlibmemory/unix_shared_memory_queue.h | 2 +- src/vnet.am | 66 +- src/vnet/api_errno.h | 21 +- src/vnet/bfd/bfd_udp.c | 4 +- src/vnet/buffer.h | 10 + src/vnet/classify/vnet_classify.c | 4 +- src/vnet/dhcp/dhcp_proxy.h | 2 +- src/vnet/flow/flow_report.h | 2 +- src/vnet/ip/ip.h | 4 +- src/vnet/ip/ip4.h | 42 +- src/vnet/ip/ip4_forward.c | 173 +- src/vnet/ip/ip4_packet.h | 26 +- src/vnet/ip/ip6.h | 44 +- src/vnet/ip/ip6_packet.h | 26 +- src/vnet/ip/punt.c | 2 +- src/vnet/ip/tcp_packet.h | 141 -- src/vnet/ip/udp.h | 315 --- src/vnet/ip/udp_error.def | 21 - src/vnet/ip/udp_format.c | 91 - src/vnet/ip/udp_init.c | 71 - src/vnet/ip/udp_local.c | 645 ------ src/vnet/ip/udp_packet.h | 65 - src/vnet/ip/udp_pg.c | 237 --- src/vnet/ipsec/ikev2.c | 2 +- src/vnet/ipsec/ikev2_cli.c | 2 +- src/vnet/ipsec/ikev2_crypto.c | 2 +- src/vnet/lisp-cp/packets.c | 65 +- src/vnet/lisp-cp/packets.h | 45 - src/vnet/lisp-gpe/interface.c | 2 +- src/vnet/lisp-gpe/lisp_gpe.h | 4 +- src/vnet/lisp-gpe/lisp_gpe_adjacency.c | 2 + src/vnet/session/application.c | 343 ++++ src/vnet/session/application.h | 120 ++ src/vnet/session/application_interface.c | 459 +++++ src/vnet/session/application_interface.h | 136 ++ src/vnet/session/hashes.c | 28 + src/vnet/session/node.c | 435 ++++ src/vnet/session/session.api | 429 ++++ src/vnet/session/session.c | 1286 ++++++++++++ src/vnet/session/session.h | 380 ++++ src/vnet/session/session_api.c | 821 ++++++++ src/vnet/session/session_cli.c | 189 ++ src/vnet/session/transport.c | 64 + src/vnet/session/transport.h | 250 +++ src/vnet/tcp/tcp.c | 708 +++++++ src/vnet/tcp/tcp.h | 624 ++++++ src/vnet/tcp/tcp_error.def | 35 + src/vnet/tcp/tcp_format.c | 136 ++ src/vnet/tcp/tcp_input.c | 2316 ++++++++++++++++++++++ src/vnet/tcp/tcp_newreno.c | 93 + src/vnet/tcp/tcp_output.c | 1412 +++++++++++++ src/vnet/tcp/tcp_packet.h | 184 ++ src/vnet/tcp/tcp_pg.c | 236 +++ src/vnet/tcp/tcp_syn_filter4.c | 542 +++++ src/vnet/tcp/tcp_timer.h | 29 + src/vnet/udp/builtin_server.c | 239 +++ src/vnet/udp/udp.c | 342 ++++ src/vnet/udp/udp.h | 362 ++++ src/vnet/udp/udp_error.def | 21 + src/vnet/udp/udp_format.c | 91 + src/vnet/udp/udp_input.c | 314 +++ src/vnet/udp/udp_local.c | 666 +++++++ src/vnet/udp/udp_packet.h | 65 + src/vnet/udp/udp_pg.c | 237 +++ src/vnet/vnet_all_api_h.h | 1 + src/vnet/vxlan-gpe/vxlan_gpe.h | 2 +- src/vnet/vxlan/vxlan.h | 2 +- src/vpp/api/vpe.api | 1 + src/vppinfra.am | 5 + src/vppinfra/bihash_16_8.h | 103 + src/vppinfra/bihash_48_8.h | 116 ++ src/vppinfra/tw_timer_16t_1w_2048sl.c | 26 + src/vppinfra/tw_timer_16t_1w_2048sl.h | 46 + 100 files changed, 18737 insertions(+), 1874 deletions(-) create mode 100644 src/scripts/vnet/udp create mode 100755 src/scripts/vnet/uri/tcp-setup.sh create mode 100644 src/scripts/vnet/uri/tcp_server create mode 100644 src/scripts/vnet/uri/udp create mode 100644 src/svm/svm_fifo.c create mode 100644 src/svm/svm_fifo.h create mode 100644 src/svm/svm_fifo_segment.c create mode 100644 src/svm/svm_fifo_segment.h create mode 100644 src/svm/test_svm_fifo1.c create mode 100644 src/uri.am create mode 100644 src/uri/uri_tcp_test.c create mode 100644 src/uri/uri_udp_test.c create mode 100644 src/uri/uri_udp_test2.c create mode 100644 src/uri/uritest.c delete mode 100644 src/vnet/ip/tcp_packet.h delete mode 100644 src/vnet/ip/udp.h delete mode 100644 src/vnet/ip/udp_error.def delete mode 100644 src/vnet/ip/udp_format.c delete mode 100644 src/vnet/ip/udp_init.c delete mode 100644 src/vnet/ip/udp_local.c delete mode 100644 src/vnet/ip/udp_packet.h delete mode 100644 src/vnet/ip/udp_pg.c create mode 100644 src/vnet/session/application.c create mode 100644 src/vnet/session/application.h create mode 100644 src/vnet/session/application_interface.c create mode 100644 src/vnet/session/application_interface.h create mode 100644 src/vnet/session/hashes.c create mode 100644 src/vnet/session/node.c create mode 100644 src/vnet/session/session.api create mode 100644 src/vnet/session/session.c create mode 100644 src/vnet/session/session.h create mode 100644 src/vnet/session/session_api.c create mode 100644 src/vnet/session/session_cli.c create mode 100644 src/vnet/session/transport.c create mode 100644 src/vnet/session/transport.h create mode 100644 src/vnet/tcp/tcp.c create mode 100644 src/vnet/tcp/tcp.h create mode 100644 src/vnet/tcp/tcp_error.def create mode 100644 src/vnet/tcp/tcp_format.c create mode 100644 src/vnet/tcp/tcp_input.c create mode 100644 src/vnet/tcp/tcp_newreno.c create mode 100644 src/vnet/tcp/tcp_output.c create mode 100644 src/vnet/tcp/tcp_packet.h create mode 100644 src/vnet/tcp/tcp_pg.c create mode 100644 src/vnet/tcp/tcp_syn_filter4.c create mode 100644 src/vnet/tcp/tcp_timer.h create mode 100644 src/vnet/udp/builtin_server.c create mode 100644 src/vnet/udp/udp.c create mode 100644 src/vnet/udp/udp.h create mode 100644 src/vnet/udp/udp_error.def create mode 100644 src/vnet/udp/udp_format.c create mode 100644 src/vnet/udp/udp_input.c create mode 100644 src/vnet/udp/udp_local.c create mode 100644 src/vnet/udp/udp_packet.h create mode 100644 src/vnet/udp/udp_pg.c create mode 100644 src/vppinfra/bihash_16_8.h create mode 100644 src/vppinfra/bihash_48_8.h create mode 100644 src/vppinfra/tw_timer_16t_1w_2048sl.c create mode 100644 src/vppinfra/tw_timer_16t_1w_2048sl.h (limited to 'src/vnet/classify') diff --git a/src/Makefile.am b/src/Makefile.am index 08feb29a..641707ed 100644 --- a/src/Makefile.am +++ b/src/Makefile.am @@ -88,6 +88,7 @@ include vlib-api.am include vnet.am include vpp.am include vpp-api-test.am +include uri.am SUBDIRS += plugins diff --git a/src/plugins/ioam/export-common/ioam_export.h b/src/plugins/ioam/export-common/ioam_export.h index e84dab0b..dd48a93b 100644 --- a/src/plugins/ioam/export-common/ioam_export.h +++ b/src/plugins/ioam/export-common/ioam_export.h @@ -21,7 +21,7 @@ #include #include #include -#include +#include #include #include diff --git a/src/plugins/ioam/ipfixcollector/ipfixcollector.c b/src/plugins/ioam/ipfixcollector/ipfixcollector.c index 4ae47edc..71b934ec 100644 --- a/src/plugins/ioam/ipfixcollector/ipfixcollector.c +++ b/src/plugins/ioam/ipfixcollector/ipfixcollector.c @@ -15,7 +15,7 @@ #include #include -#include +#include #include ipfix_collector_main_t ipfix_collector_main; diff --git a/src/plugins/ioam/lib-vxlan-gpe/ioam_transit.c b/src/plugins/ioam/lib-vxlan-gpe/ioam_transit.c index b42c357c..f334c983 100644 --- a/src/plugins/ioam/lib-vxlan-gpe/ioam_transit.c +++ b/src/plugins/ioam/lib-vxlan-gpe/ioam_transit.c @@ -16,7 +16,7 @@ #include #include #include -#include +#include #include #include #include diff --git a/src/plugins/snat/in2out.c b/src/plugins/snat/in2out.c index e30c913c..b4b7793d 100644 --- a/src/plugins/snat/in2out.c +++ b/src/plugins/snat/in2out.c @@ -689,12 +689,12 @@ snat_hairpinning (snat_main_t *sm, ip4_header_t, dst_address); ip0->checksum = ip_csum_fold (sum0); - old_dst_port0 = tcp0->ports.dst; + old_dst_port0 = tcp0->dst; if (PREDICT_TRUE(new_dst_port0 != old_dst_port0)) { if (PREDICT_TRUE(proto0 == SNAT_PROTOCOL_TCP)) { - tcp0->ports.dst = new_dst_port0; + tcp0->dst = new_dst_port0; sum0 = tcp0->checksum; sum0 = ip_csum_update (sum0, old_dst_addr0, new_dst_addr0, ip4_header_t, dst_address); @@ -872,9 +872,9 @@ snat_in2out_node_fn_inline (vlib_main_t * vm, if (PREDICT_TRUE(proto0 == SNAT_PROTOCOL_TCP)) { - old_port0 = tcp0->ports.src; - tcp0->ports.src = s0->out2in.port; - new_port0 = tcp0->ports.src; + old_port0 = tcp0->src_port; + tcp0->src_port = s0->out2in.port; + new_port0 = tcp0->src_port; sum0 = tcp0->checksum; sum0 = ip_csum_update (sum0, old_addr0, new_addr0, @@ -1012,9 +1012,9 @@ snat_in2out_node_fn_inline (vlib_main_t * vm, if (PREDICT_TRUE(proto1 == SNAT_PROTOCOL_TCP)) { - old_port1 = tcp1->ports.src; - tcp1->ports.src = s1->out2in.port; - new_port1 = tcp1->ports.src; + old_port1 = tcp1->src_port; + tcp1->src_port = s1->out2in.port; + new_port1 = tcp1->src_port; sum1 = tcp1->checksum; sum1 = ip_csum_update (sum1, old_addr1, new_addr1, @@ -1188,9 +1188,9 @@ snat_in2out_node_fn_inline (vlib_main_t * vm, if (PREDICT_TRUE(proto0 == SNAT_PROTOCOL_TCP)) { - old_port0 = tcp0->ports.src; - tcp0->ports.src = s0->out2in.port; - new_port0 = tcp0->ports.src; + old_port0 = tcp0->src_port; + tcp0->src_port = s0->out2in.port; + new_port0 = tcp0->src_port; sum0 = tcp0->checksum; sum0 = ip_csum_update (sum0, old_addr0, new_addr0, @@ -1667,8 +1667,8 @@ snat_in2out_fast_static_map_fn (vlib_main_t * vm, { if (PREDICT_TRUE(proto0 == SNAT_PROTOCOL_TCP)) { - old_port0 = tcp0->ports.src; - tcp0->ports.src = new_port0; + old_port0 = tcp0->src_port; + tcp0->src_port = new_port0; sum0 = tcp0->checksum; sum0 = ip_csum_update (sum0, old_addr0, new_addr0, diff --git a/src/plugins/snat/out2in.c b/src/plugins/snat/out2in.c index 328f5ba4..3bfc0aa3 100644 --- a/src/plugins/snat/out2in.c +++ b/src/plugins/snat/out2in.c @@ -19,7 +19,7 @@ #include #include -#include +#include #include #include #include @@ -602,9 +602,9 @@ snat_out2in_node_fn (vlib_main_t * vm, if (PREDICT_TRUE(proto0 == SNAT_PROTOCOL_TCP)) { - old_port0 = tcp0->ports.dst; - tcp0->ports.dst = s0->in2out.port; - new_port0 = tcp0->ports.dst; + old_port0 = tcp0->dst_port; + tcp0->dst_port = s0->in2out.port; + new_port0 = tcp0->dst_port; sum0 = tcp0->checksum; sum0 = ip_csum_update (sum0, old_addr0, new_addr0, @@ -737,9 +737,9 @@ snat_out2in_node_fn (vlib_main_t * vm, if (PREDICT_TRUE(proto1 == SNAT_PROTOCOL_TCP)) { - old_port1 = tcp1->ports.dst; - tcp1->ports.dst = s1->in2out.port; - new_port1 = tcp1->ports.dst; + old_port1 = tcp1->dst_port; + tcp1->dst_port = s1->in2out.port; + new_port1 = tcp1->dst_port; sum1 = tcp1->checksum; sum1 = ip_csum_update (sum1, old_addr1, new_addr1, @@ -907,9 +907,9 @@ snat_out2in_node_fn (vlib_main_t * vm, if (PREDICT_TRUE(proto0 == SNAT_PROTOCOL_TCP)) { - old_port0 = tcp0->ports.dst; - tcp0->ports.dst = s0->in2out.port; - new_port0 = tcp0->ports.dst; + old_port0 = tcp0->dst_port; + tcp0->dst_port = s0->in2out.port; + new_port0 = tcp0->dst_port; sum0 = tcp0->checksum; sum0 = ip_csum_update (sum0, old_addr0, new_addr0, @@ -1369,8 +1369,8 @@ snat_out2in_fast_node_fn (vlib_main_t * vm, { if (PREDICT_TRUE(proto0 == SNAT_PROTOCOL_TCP)) { - old_port0 = tcp0->ports.dst; - tcp0->ports.dst = new_port0; + old_port0 = tcp0->dst_port; + tcp0->dst_port = new_port0; sum0 = tcp0->checksum; sum0 = ip_csum_update (sum0, old_addr0, new_addr0, diff --git a/src/scripts/vnet/tcp b/src/scripts/vnet/tcp index a2ee8b2d..b9c23c3a 100644 --- a/src/scripts/vnet/tcp +++ b/src/scripts/vnet/tcp @@ -1,16 +1,18 @@ +loop create +set int ip address loop0 192.168.1.1/8 +set int state loop0 up + packet-generator new { name x - limit 1 + limit 2048 node ip4-input - size 64-64 + size 100-100 + interface loop0 no-recycle data { - TCP: 1.2.3.4 -> 5.6.7.8 - TCP: 1234 -> 5678 + TCP: 192.168.1.2 -> 192.168.1.1 + TCP: 32415 -> 80 + SYN incrementing 100 } } - -tr add pg-input 100 -ip route 5.6.7.8/32 via local -ip route 1.2.3.4/32 via local diff --git a/src/scripts/vnet/udp b/src/scripts/vnet/udp new file mode 100644 index 00000000..7dda1eec --- /dev/null +++ b/src/scripts/vnet/udp @@ -0,0 +1,19 @@ +loop create +set int ip address loop0 192.168.1.1/8 +set int state loop0 up + +packet-generator new { + name udp + limit 512 + rate 1e4 + node ip4-input + size 100-100 + interface loop0 + no-recycle + data { + UDP: 192.168.1.2 - 192.168.2.255 -> 192.168.1.1 + UDP: 4321 -> 1234 + length 72 + incrementing 100 + } +} diff --git a/src/scripts/vnet/uri/tcp-setup.sh b/src/scripts/vnet/uri/tcp-setup.sh new file mode 100755 index 00000000..e0b01588 --- /dev/null +++ b/src/scripts/vnet/uri/tcp-setup.sh @@ -0,0 +1,39 @@ +#!/usr/bin/env bash + +function topo_setup +{ + ip netns add vppns1 + ip link add veth_vpp1 type veth peer name vpp1 + ip link set dev vpp1 up + ip link set dev veth_vpp1 up netns vppns1 + + ip netns exec vppns1 \ + bash -c " + ip link set dev lo up + ip addr add 6.0.1.2/24 dev veth_vpp1 + " + + ethtool --offload vpp1 rx off tx off + ip netns exec vppns1 ethtool --offload veth_vpp1 rx off tx off + +} + +function topo_clean +{ + ip link del dev veth_vpp1 &> /dev/null + ip netns del vppns1 &> /dev/null +} + +if [ "$1" == "clean" ] ; then + topo_clean + exit 0 +else + topo_setup +fi + +# to test connectivity do: +# sudo ip netns exec vppns1 telnet 6.0.1.1 1234 +# to push traffic to the server +# dd if=/dev/zero bs=1024K count=512 | nc 6.0.1.1 +# to listen for incoming connection from vpp +# nc -l 1234 diff --git a/src/scripts/vnet/uri/tcp_server b/src/scripts/vnet/uri/tcp_server new file mode 100644 index 00000000..7f5a86de --- /dev/null +++ b/src/scripts/vnet/uri/tcp_server @@ -0,0 +1,4 @@ +create host-interface name vpp1 +set int state host-vpp1 up +set int ip address host-vpp1 6.0.1.1/24 +trace add af-packet-input 10 diff --git a/src/scripts/vnet/uri/udp b/src/scripts/vnet/uri/udp new file mode 100644 index 00000000..ca13b83c --- /dev/null +++ b/src/scripts/vnet/uri/udp @@ -0,0 +1,19 @@ +loop create +set int ip address loop0 10.0.0.1/32 +set int state loop0 up + +packet-generator new { + name udp + limit 512 + rate 1e4 + node ip4-input + size 100-100 + interface loop0 + no-recycle + data { + UDP: 192.168.1.2 - 192.168.2.255 -> 192.168.1.1 + UDP: 4321 -> 1234 + length 72 + incrementing 100 + } +} diff --git a/src/svm.am b/src/svm.am index 2cd385bd..442eba8e 100644 --- a/src/svm.am +++ b/src/svm.am @@ -13,13 +13,14 @@ bin_PROGRAMS += svmtool svmdbtool -nobase_include_HEADERS += svm/svm.h svm/ssvm.h svm/svmdb.h +nobase_include_HEADERS += svm/svm.h svm/ssvm.h svm/svmdb.h \ + svm/svm_fifo.h svm/svm_fifo_segment.h lib_LTLIBRARIES += libsvm.la libsvmdb.la +libsvm_la_SOURCES = svm/svm.c svm/ssvm.c svm/svm_fifo.c svm/svm_fifo_segment.c libsvm_la_LIBADD = libvppinfra.la -lrt -lpthread libsvm_la_DEPENDENCIES = libvppinfra.la -libsvm_la_SOURCES = svm/svm.c svm/ssvm.c svmtool_SOURCES = svm/svmtool.c svmtool_LDADD = libsvm.la libvppinfra.la -lpthread -lrt @@ -31,4 +32,9 @@ libsvmdb_la_SOURCES = svm/svmdb.c svmdbtool_SOURCES = svm/svmdbtool.c svmdbtool_LDADD = libsvmdb.la libsvm.la libvppinfra.la -lpthread -lrt +noinst_PROGRAMS += test_svm_fifo1 +test_svm_fifo1_SOURCES = svm/test_svm_fifo1.c +test_svm_fifo1_LDADD = libsvm.la libvppinfra.la -lpthread -lrt +test_svm_fifo1_LDFLAGS = -static + # vi:syntax=automake diff --git a/src/svm/ssvm.c b/src/svm/ssvm.c index 6f409eb6..6cda1f27 100644 --- a/src/svm/ssvm.c +++ b/src/svm/ssvm.c @@ -169,6 +169,22 @@ re_map_it: return 0; } +void +ssvm_delete (ssvm_private_t * ssvm) +{ + u8 *fn; + + fn = format (0, "/dev/shm/%s%c", ssvm->name, 0); + + /* Throw away the backing file */ + if (unlink ((char *) fn) < 0) + clib_unix_warning ("unlink segment '%s'", ssvm->name); + + munmap ((void *) ssvm->requested_va, ssvm->ssvm_size); + vec_free (fn); +} + + /* * fd.io coding-style-patch-verification: ON * diff --git a/src/svm/ssvm.h b/src/svm/ssvm.h index 9e61b9a0..bccfc164 100644 --- a/src/svm/ssvm.h +++ b/src/svm/ssvm.h @@ -38,7 +38,10 @@ #include #include -#define MMAP_PAGESIZE (4<<10) +#ifndef MMAP_PAGESIZE +#define MMAP_PAGESIZE (clib_mem_get_page_size()) +#endif + #define SSVM_N_OPAQUE 7 typedef struct @@ -125,12 +128,12 @@ ssvm_pop_heap (void *oldheap) } #define foreach_ssvm_api_error \ -_(NO_NAME, "No shared segment name", -10) \ -_(NO_SIZE, "Size not set (master)", -11) \ -_(CREATE_FAILURE, "Create failed", -12) \ -_(SET_SIZE, "Set size failed", -13) \ -_(MMAP, "mmap failed", -14) \ -_(SLAVE_TIMEOUT, "Slave map timeout", -15) +_(NO_NAME, "No shared segment name", -100) \ +_(NO_SIZE, "Size not set (master)", -101) \ +_(CREATE_FAILURE, "Create failed", -102) \ +_(SET_SIZE, "Set size failed", -103) \ +_(MMAP, "mmap failed", -104) \ +_(SLAVE_TIMEOUT, "Slave map timeout", -105) typedef enum { @@ -143,6 +146,7 @@ typedef enum int ssvm_master_init (ssvm_private_t * ssvm, u32 master_index); int ssvm_slave_init (ssvm_private_t * ssvm, int timeout_in_seconds); +void ssvm_delete (ssvm_private_t * ssvm); #endif /* __included_ssvm_h__ */ diff --git a/src/svm/svm_fifo.c b/src/svm/svm_fifo.c new file mode 100644 index 00000000..11f90193 --- /dev/null +++ b/src/svm/svm_fifo.c @@ -0,0 +1,568 @@ +/* + * Copyright (c) 2016 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +#include "svm_fifo.h" + +/** create an svm fifo, in the current heap. Fails vs blow up the process */ +svm_fifo_t * +svm_fifo_create (u32 data_size_in_bytes) +{ + svm_fifo_t *f; + pthread_mutexattr_t attr; + pthread_condattr_t cattr; + + f = clib_mem_alloc_aligned_or_null (sizeof (*f) + data_size_in_bytes, + CLIB_CACHE_LINE_BYTES); + if (f == 0) + return 0; + + memset (f, 0, sizeof (*f) + data_size_in_bytes); + f->nitems = data_size_in_bytes; + f->ooos_list_head = OOO_SEGMENT_INVALID_INDEX; + + memset (&attr, 0, sizeof (attr)); + memset (&cattr, 0, sizeof (cattr)); + + if (pthread_mutexattr_init (&attr)) + clib_unix_warning ("mutexattr_init"); + if (pthread_mutexattr_setpshared (&attr, PTHREAD_PROCESS_SHARED)) + clib_unix_warning ("pthread_mutexattr_setpshared"); + if (pthread_mutex_init (&f->mutex, &attr)) + clib_unix_warning ("mutex_init"); + if (pthread_mutexattr_destroy (&attr)) + clib_unix_warning ("mutexattr_destroy"); + if (pthread_condattr_init (&cattr)) + clib_unix_warning ("condattr_init"); + if (pthread_condattr_setpshared (&cattr, PTHREAD_PROCESS_SHARED)) + clib_unix_warning ("condattr_setpshared"); + if (pthread_cond_init (&f->condvar, &cattr)) + clib_unix_warning ("cond_init1"); + if (pthread_condattr_destroy (&cattr)) + clib_unix_warning ("cond_init2"); + + return (f); +} + +always_inline ooo_segment_t * +ooo_segment_new (svm_fifo_t * f, u32 start, u32 length) +{ + ooo_segment_t *s; + + pool_get (f->ooo_segments, s); + + s->fifo_position = start; + s->length = length; + + s->prev = s->next = OOO_SEGMENT_INVALID_INDEX; + + return s; +} + +always_inline void +ooo_segment_del (svm_fifo_t * f, u32 index) +{ + ooo_segment_t *cur, *prev = 0, *next = 0; + cur = pool_elt_at_index (f->ooo_segments, index); + + if (cur->next != OOO_SEGMENT_INVALID_INDEX) + { + next = pool_elt_at_index (f->ooo_segments, cur->next); + next->prev = cur->prev; + } + + if (cur->prev != OOO_SEGMENT_INVALID_INDEX) + { + prev = pool_elt_at_index (f->ooo_segments, cur->prev); + prev->next = cur->next; + } + else + { + f->ooos_list_head = cur->next; + } + + pool_put (f->ooo_segments, cur); +} + +/** + * Add segment to fifo's out-of-order segment list. Takes care of merging + * adjacent segments and removing overlapping ones. + */ +static void +ooo_segment_add (svm_fifo_t * f, u32 offset, u32 length) +{ + ooo_segment_t *s, *new_s, *prev, *next, *it; + u32 new_index, position, end_offset, s_sof, s_eof, s_index; + + position = (f->tail + offset) % f->nitems; + end_offset = offset + length; + + if (f->ooos_list_head == OOO_SEGMENT_INVALID_INDEX) + { + s = ooo_segment_new (f, position, length); + f->ooos_list_head = s - f->ooo_segments; + f->ooos_newest = f->ooos_list_head; + return; + } + + /* Find first segment that starts after new segment */ + s = pool_elt_at_index (f->ooo_segments, f->ooos_list_head); + while (s->next != OOO_SEGMENT_INVALID_INDEX + && ooo_segment_offset (f, s) <= offset) + s = pool_elt_at_index (f->ooo_segments, s->next); + + s_index = s - f->ooo_segments; + s_sof = ooo_segment_offset (f, s); + s_eof = ooo_segment_end_offset (f, s); + + /* No overlap, add before current segment */ + if (end_offset < s_sof) + { + new_s = ooo_segment_new (f, position, length); + new_index = new_s - f->ooo_segments; + + /* Pool might've moved, get segment again */ + s = pool_elt_at_index (f->ooo_segments, s_index); + + if (s->prev != OOO_SEGMENT_INVALID_INDEX) + { + new_s->prev = s->prev; + + prev = pool_elt_at_index (f->ooo_segments, new_s->prev); + prev->next = new_index; + } + else + { + /* New head */ + f->ooos_list_head = new_index; + } + + new_s->next = s - f->ooo_segments; + s->prev = new_index; + f->ooos_newest = new_index; + return; + } + /* No overlap, add after current segment */ + else if (s_eof < offset) + { + new_s = ooo_segment_new (f, position, length); + new_index = new_s - f->ooo_segments; + + /* Pool might've moved, get segment again */ + s = pool_elt_at_index (f->ooo_segments, s_index); + + if (s->next != OOO_SEGMENT_INVALID_INDEX) + { + new_s->next = s->next; + + next = pool_elt_at_index (f->ooo_segments, new_s->next); + next->prev = new_index; + } + + new_s->prev = s - f->ooo_segments; + s->next = new_index; + f->ooos_newest = new_index; + + return; + } + + /* + * Merge needed + */ + + /* Merge at head */ + if (offset <= s_sof) + { + /* If we have a previous, check if we overlap */ + if (s->prev != OOO_SEGMENT_INVALID_INDEX) + { + prev = pool_elt_at_index (f->ooo_segments, s->prev); + + /* New segment merges prev and current. Remove previous and + * update position of current. */ + if (ooo_segment_end_offset (f, prev) >= offset) + { + s->fifo_position = prev->fifo_position; + s->length = s_eof - ooo_segment_offset (f, prev); + ooo_segment_del (f, s->prev); + } + } + else + { + s->fifo_position = position; + s->length = s_eof - ooo_segment_offset (f, s); + } + + /* The new segment's tail may cover multiple smaller ones */ + if (s_eof < end_offset) + { + /* Remove segments completely covered */ + it = (s->next != OOO_SEGMENT_INVALID_INDEX) ? + pool_elt_at_index (f->ooo_segments, s->next) : 0; + while (it && ooo_segment_end_offset (f, it) < end_offset) + { + next = (it->next != OOO_SEGMENT_INVALID_INDEX) ? + pool_elt_at_index (f->ooo_segments, it->next) : 0; + ooo_segment_del (f, it - f->ooo_segments); + it = next; + } + + /* Update length. Segment's start might have changed. */ + s->length = end_offset - ooo_segment_offset (f, s); + + /* If partial overlap with last, merge */ + if (it && ooo_segment_offset (f, it) < end_offset) + { + s->length += + it->length - (ooo_segment_offset (f, it) - end_offset); + ooo_segment_del (f, it - f->ooo_segments); + } + } + } + /* Last but overlapping previous */ + else if (s_eof <= end_offset) + { + s->length = end_offset - ooo_segment_offset (f, s); + } + /* New segment completely covered by current one */ + else + { + /* Do Nothing */ + } + + /* Most recently updated segment */ + f->ooos_newest = s - f->ooo_segments; +} + +/** + * Removes segments that can now be enqueued because the fifo's tail has + * advanced. Returns the number of bytes added to tail. + */ +static int +ooo_segment_try_collect (svm_fifo_t * f, u32 n_bytes_enqueued) +{ + ooo_segment_t *s; + u32 index, bytes = 0, diff; + + s = pool_elt_at_index (f->ooo_segments, f->ooos_list_head); + + /* If last tail update overlaps one/multiple ooo segments, remove them */ + diff = (f->nitems + f->tail - s->fifo_position) % f->nitems; + while (0 < diff && diff < n_bytes_enqueued) + { + /* Segment end is beyond the tail. Advance tail and be done */ + if (diff < s->length) + { + f->tail += s->length - diff; + f->tail %= f->nitems; + break; + } + /* If we have next go on */ + else if (s->next != OOO_SEGMENT_INVALID_INDEX) + { + index = s - f->ooo_segments; + s = pool_elt_at_index (f->ooo_segments, s->next); + diff = (f->nitems + f->tail - s->fifo_position) % f->nitems; + ooo_segment_del (f, index); + } + /* End of search */ + else + { + break; + } + } + + /* If tail is adjacent to an ooo segment, 'consume' it */ + if (diff == 0) + { + bytes = ((f->nitems - f->cursize) >= s->length) ? s->length : + f->nitems - f->cursize; + + f->tail += bytes; + f->tail %= f->nitems; + + ooo_segment_del (f, s - f->ooo_segments); + } + + return bytes; +} + +static int +svm_fifo_enqueue_internal (svm_fifo_t * f, + int pid, u32 max_bytes, u8 * copy_from_here) +{ + u32 total_copy_bytes, first_copy_bytes, second_copy_bytes; + u32 cursize, nitems; + + if (PREDICT_FALSE (f->cursize == f->nitems)) + return -2; /* fifo stuffed */ + + /* read cursize, which can only decrease while we're working */ + cursize = f->cursize; + nitems = f->nitems; + + /* Number of bytes we're going to copy */ + total_copy_bytes = (nitems - cursize) < max_bytes ? + (nitems - cursize) : max_bytes; + + if (PREDICT_TRUE (copy_from_here != 0)) + { + /* Number of bytes in first copy segment */ + first_copy_bytes = ((nitems - f->tail) < total_copy_bytes) + ? (nitems - f->tail) : total_copy_bytes; + + clib_memcpy (&f->data[f->tail], copy_from_here, first_copy_bytes); + f->tail += first_copy_bytes; + f->tail = (f->tail == nitems) ? 0 : f->tail; + + /* Number of bytes in second copy segment, if any */ + second_copy_bytes = total_copy_bytes - first_copy_bytes; + if (second_copy_bytes) + { + clib_memcpy (&f->data[f->tail], copy_from_here + first_copy_bytes, + second_copy_bytes); + f->tail += second_copy_bytes; + f->tail = (f->tail == nitems) ? 0 : f->tail; + } + } + else + { + /* Account for a zero-copy enqueue done elsewhere */ + ASSERT (max_bytes <= (nitems - cursize)); + f->tail += max_bytes; + f->tail = f->tail % nitems; + total_copy_bytes = max_bytes; + } + + /* Any out-of-order segments to collect? */ + if (PREDICT_FALSE (f->ooos_list_head != OOO_SEGMENT_INVALID_INDEX)) + total_copy_bytes += ooo_segment_try_collect (f, total_copy_bytes); + + /* Atomically increase the queue length */ + __sync_fetch_and_add (&f->cursize, total_copy_bytes); + + return (total_copy_bytes); +} + +int +svm_fifo_enqueue_nowait (svm_fifo_t * f, + int pid, u32 max_bytes, u8 * copy_from_here) +{ + return svm_fifo_enqueue_internal (f, pid, max_bytes, copy_from_here); +} + +/** Enqueue a future segment. + * Two choices: either copies the entire segment, or copies nothing + * Returns 0 of the entire segment was copied + * Returns -1 if none of the segment was copied due to lack of space + */ + +static int +svm_fifo_enqueue_with_offset_internal2 (svm_fifo_t * f, + int pid, + u32 offset, + u32 required_bytes, + u8 * copy_from_here) +{ + u32 total_copy_bytes, first_copy_bytes, second_copy_bytes; + u32 cursize, nitems; + u32 tail_plus_offset; + + ASSERT (offset > 0); + + /* read cursize, which can only decrease while we're working */ + cursize = f->cursize; + nitems = f->nitems; + + /* Will this request fit? */ + if ((required_bytes + offset) > (nitems - cursize)) + return -1; + + ooo_segment_add (f, offset, required_bytes); + + /* Number of bytes we're going to copy */ + total_copy_bytes = required_bytes; + tail_plus_offset = (f->tail + offset) % nitems; + + /* Number of bytes in first copy segment */ + first_copy_bytes = ((nitems - tail_plus_offset) < total_copy_bytes) + ? (nitems - tail_plus_offset) : total_copy_bytes; + + clib_memcpy (&f->data[tail_plus_offset], copy_from_here, first_copy_bytes); + + /* Number of bytes in second copy segment, if any */ + second_copy_bytes = total_copy_bytes - first_copy_bytes; + if (second_copy_bytes) + { + tail_plus_offset += first_copy_bytes; + tail_plus_offset %= nitems; + + ASSERT (tail_plus_offset == 0); + + clib_memcpy (&f->data[tail_plus_offset], + copy_from_here + first_copy_bytes, second_copy_bytes); + } + + return (0); +} + + +int +svm_fifo_enqueue_with_offset (svm_fifo_t * f, + int pid, + u32 offset, + u32 required_bytes, u8 * copy_from_here) +{ + return svm_fifo_enqueue_with_offset_internal2 + (f, pid, offset, required_bytes, copy_from_here); +} + + +static int +svm_fifo_dequeue_internal2 (svm_fifo_t * f, + int pid, u32 max_bytes, u8 * copy_here) +{ + u32 total_copy_bytes, first_copy_bytes, second_copy_bytes; + u32 cursize, nitems; + + if (PREDICT_FALSE (f->cursize == 0)) + return -2; /* nothing in the fifo */ + + /* read cursize, which can only increase while we're working */ + cursize = f->cursize; + nitems = f->nitems; + + /* Number of bytes we're going to copy */ + total_copy_bytes = (cursize < max_bytes) ? cursize : max_bytes; + + if (PREDICT_TRUE (copy_here != 0)) + { + /* Number of bytes in first copy segment */ + first_copy_bytes = ((nitems - f->head) < total_copy_bytes) + ? (nitems - f->head) : total_copy_bytes; + clib_memcpy (copy_here, &f->data[f->head], first_copy_bytes); + f->head += first_copy_bytes; + f->head = (f->head == nitems) ? 0 : f->head; + + /* Number of bytes in second copy segment, if any */ + second_copy_bytes = total_copy_bytes - first_copy_bytes; + if (second_copy_bytes) + { + clib_memcpy (copy_here + first_copy_bytes, + &f->data[f->head], second_copy_bytes); + f->head += second_copy_bytes; + f->head = (f->head == nitems) ? 0 : f->head; + } + } + else + { + /* Account for a zero-copy dequeue done elsewhere */ + ASSERT (max_bytes <= cursize); + f->head += max_bytes; + f->head = f->head % nitems; + cursize -= max_bytes; + total_copy_bytes = max_bytes; + } + + __sync_fetch_and_sub (&f->cursize, total_copy_bytes); + + return (total_copy_bytes); +} + +int +svm_fifo_dequeue_nowait (svm_fifo_t * f, + int pid, u32 max_bytes, u8 * copy_here) +{ + return svm_fifo_dequeue_internal2 (f, pid, max_bytes, copy_here); +} + +int +svm_fifo_peek (svm_fifo_t * f, int pid, u32 offset, u32 max_bytes, + u8 * copy_here) +{ + u32 total_copy_bytes, first_copy_bytes, second_copy_bytes; + u32 cursize, nitems; + + if (PREDICT_FALSE (f->cursize == 0)) + return -2; /* nothing in the fifo */ + + /* read cursize, which can only increase while we're working */ + cursize = f->cursize; + nitems = f->nitems; + + /* Number of bytes we're going to copy */ + total_copy_bytes = (cursize < max_bytes) ? cursize : max_bytes; + + if (PREDICT_TRUE (copy_here != 0)) + { + /* Number of bytes in first copy segment */ + first_copy_bytes = + ((nitems - f->head) < total_copy_bytes) ? + (nitems - f->head) : total_copy_bytes; + clib_memcpy (copy_here, &f->data[f->head], first_copy_bytes); + + /* Number of bytes in second copy segment, if any */ + second_copy_bytes = total_copy_bytes - first_copy_bytes; + if (second_copy_bytes) + { + clib_memcpy (copy_here + first_copy_bytes, &f->data[0], + second_copy_bytes); + } + } + return total_copy_bytes; +} + +int +svm_fifo_dequeue_drop (svm_fifo_t * f, int pid, u32 max_bytes) +{ + u32 total_drop_bytes, first_drop_bytes, second_drop_bytes; + u32 cursize, nitems; + + if (PREDICT_FALSE (f->cursize == 0)) + return -2; /* nothing in the fifo */ + + /* read cursize, which can only increase while we're working */ + cursize = f->cursize; + nitems = f->nitems; + + /* Number of bytes we're going to drop */ + total_drop_bytes = (cursize < max_bytes) ? cursize : max_bytes; + + /* Number of bytes in first copy segment */ + first_drop_bytes = + ((nitems - f->head) < total_drop_bytes) ? + (nitems - f->head) : total_drop_bytes; + f->head += first_drop_bytes; + f->head = (f->head == nitems) ? 0 : f->head; + + /* Number of bytes in second drop segment, if any */ + second_drop_bytes = total_drop_bytes - first_drop_bytes; + if (second_drop_bytes) + { + f->head += second_drop_bytes; + f->head = (f->head == nitems) ? 0 : f->head; + } + + __sync_fetch_and_sub (&f->cursize, total_drop_bytes); + + return total_drop_bytes; +} + +/* + * fd.io coding-style-patch-verification: ON + * + * Local Variables: + * eval: (c-set-style "gnu") + * End: + */ diff --git a/src/svm/svm_fifo.h b/src/svm/svm_fifo.h new file mode 100644 index 00000000..70624b74 --- /dev/null +++ b/src/svm/svm_fifo.h @@ -0,0 +1,157 @@ +/* + * Copyright (c) 2016 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ +#ifndef __included_ssvm_fifo_h__ +#define __included_ssvm_fifo_h__ + +#include +#include +#include +#include +#include +#include +#include + +typedef enum +{ + SVM_FIFO_TAG_NOT_HELD = 0, + SVM_FIFO_TAG_DEQUEUE, + SVM_FIFO_TAG_ENQUEUE, +} svm_lock_tag_t; + +/** Out-of-order segment */ +typedef struct +{ + u32 next; /**< Next linked-list element pool index */ + u32 prev; /**< Previous linked-list element pool index */ + + u32 fifo_position; /**< Start of segment, normalized*/ + u32 length; /**< Length of segment */ +} ooo_segment_t; + +#define OOO_SEGMENT_INVALID_INDEX ((u32)~0) + +typedef struct +{ + pthread_mutex_t mutex; /* 8 bytes */ + pthread_cond_t condvar; /* 8 bytes */ + u32 owner_pid; + svm_lock_tag_t tag; + volatile u32 cursize; + u32 nitems; + + /* Backpointers */ + u32 server_session_index; + u32 client_session_index; + u8 server_thread_index; + u8 client_thread_index; + CLIB_CACHE_LINE_ALIGN_MARK (end_shared); + u32 head; + CLIB_CACHE_LINE_ALIGN_MARK (end_consumer); + + /* producer */ + u32 tail; + + ooo_segment_t *ooo_segments; /**< Pool of ooo segments */ + u32 ooos_list_head; /**< Head of out-of-order linked-list */ + u32 ooos_newest; /**< Last segment to have been updated */ + + CLIB_CACHE_LINE_ALIGN_MARK (data); +} svm_fifo_t; + +static inline int +svm_fifo_lock (svm_fifo_t * f, u32 pid, u32 tag, int nowait) +{ + if (PREDICT_TRUE (nowait == 0)) + pthread_mutex_lock (&f->mutex); + else + { + if (pthread_mutex_trylock (&f->mutex)) + return -1; + } + f->owner_pid = pid; + f->tag = tag; + return 0; +} + +static inline void +svm_fifo_unlock (svm_fifo_t * f) +{ + f->owner_pid = 0; + f->tag = 0; + CLIB_MEMORY_BARRIER (); + pthread_mutex_unlock (&f->mutex); +} + +static inline u32 +svm_fifo_max_dequeue (svm_fifo_t * f) +{ + return f->cursize; +} + +static inline u32 +svm_fifo_max_enqueue (svm_fifo_t * f) +{ + return f->nitems - f->cursize; +} + +static inline u8 +svm_fifo_has_ooo_data (svm_fifo_t * f) +{ + return f->ooos_list_head != OOO_SEGMENT_INVALID_INDEX; +} + +svm_fifo_t *svm_fifo_create (u32 data_size_in_bytes); + +int svm_fifo_enqueue_nowait (svm_fifo_t * f, int pid, u32 max_bytes, + u8 * copy_from_here); + +int svm_fifo_enqueue_with_offset (svm_fifo_t * f, int pid, + u32 offset, u32 required_bytes, + u8 * copy_from_here); + +int svm_fifo_dequeue_nowait (svm_fifo_t * f, int pid, u32 max_bytes, + u8 * copy_here); + +int svm_fifo_peek (svm_fifo_t * f, int pid, u32 offset, u32 max_bytes, + u8 * copy_here); +int svm_fifo_dequeue_drop (svm_fifo_t * f, int pid, u32 max_bytes); + +always_inline ooo_segment_t * +svm_fifo_newest_ooo_segment (svm_fifo_t * f) +{ + return f->ooo_segments + f->ooos_newest; +} + +always_inline u32 +ooo_segment_offset (svm_fifo_t * f, ooo_segment_t * s) +{ + return ((f->nitems + s->fifo_position - f->tail) % f->nitems); +} + +always_inline u32 +ooo_segment_end_offset (svm_fifo_t * f, ooo_segment_t * s) +{ + return ((f->nitems + s->fifo_position + s->length - f->tail) % f->nitems); +} + +#endif /* __included_ssvm_fifo_h__ */ + +/* + * fd.io coding-style-patch-verification: ON + * + * Local Variables: + * eval: (c-set-style "gnu") + * End: + */ diff --git a/src/svm/svm_fifo_segment.c b/src/svm/svm_fifo_segment.c new file mode 100644 index 00000000..acabb3bd --- /dev/null +++ b/src/svm/svm_fifo_segment.c @@ -0,0 +1,193 @@ +/* + * Copyright (c) 2016 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +#include + +svm_fifo_segment_main_t svm_fifo_segment_main; + +/** (master) create an svm fifo segment */ +int +svm_fifo_segment_create (svm_fifo_segment_create_args_t * a) +{ + int rv; + svm_fifo_segment_private_t *s; + svm_fifo_segment_main_t *sm = &svm_fifo_segment_main; + ssvm_shared_header_t *sh; + svm_fifo_segment_header_t *fsh; + void *oldheap; + + /* Allocate a fresh segment */ + pool_get (sm->segments, s); + memset (s, 0, sizeof (*s)); + + s->ssvm.ssvm_size = a->segment_size; + s->ssvm.i_am_master = 1; + s->ssvm.my_pid = getpid (); + s->ssvm.name = (u8 *) a->segment_name; + s->ssvm.requested_va = sm->next_baseva; + + rv = ssvm_master_init (&s->ssvm, s - sm->segments); + + if (rv) + { + _vec_len (s) = vec_len (s) - 1; + return (rv); + } + + /* Note; requested_va updated due to seg base addr randomization */ + sm->next_baseva = s->ssvm.requested_va + a->segment_size; + + sh = s->ssvm.sh; + oldheap = ssvm_push_heap (sh); + + /* Set up svm_fifo_segment shared header */ + fsh = clib_mem_alloc (sizeof (*fsh)); + memset (fsh, 0, sizeof (*fsh)); + sh->opaque[0] = fsh; + s->h = fsh; + fsh->segment_name = format (0, "%s%c", a->segment_name, 0); + + /* Avoid vec_add1(...) failure when adding a fifo, etc. */ + vec_validate (fsh->fifos, 64); + _vec_len (fsh->fifos) = 0; + + ssvm_pop_heap (oldheap); + + sh->ready = 1; + a->new_segment_index = s - sm->segments; + return (0); +} + +/** (slave) attach to an svm fifo segment */ +int +svm_fifo_segment_attach (svm_fifo_segment_create_args_t * a) +{ + int rv; + svm_fifo_segment_private_t *s; + svm_fifo_segment_main_t *sm = &svm_fifo_segment_main; + ssvm_shared_header_t *sh; + svm_fifo_segment_header_t *fsh; + + /* Allocate a fresh segment */ + pool_get (sm->segments, s); + + memset (s, 0, sizeof (*s)); + + s->ssvm.ssvm_size = a->segment_size; + s->ssvm.my_pid = getpid (); + s->ssvm.name = (u8 *) a->segment_name; + s->ssvm.requested_va = sm->next_baseva; + + rv = ssvm_slave_init (&s->ssvm, sm->timeout_in_seconds); + + if (rv) + { + _vec_len (s) = vec_len (s) - 1; + return (rv); + } + + /* Fish the segment header */ + sh = s->ssvm.sh; + fsh = (svm_fifo_segment_header_t *) sh->opaque[0]; + s->h = fsh; + + a->new_segment_index = s - sm->segments; + return (0); +} + +void +svm_fifo_segment_delete (svm_fifo_segment_private_t * s) +{ + svm_fifo_segment_main_t *sm = &svm_fifo_segment_main; + ssvm_delete (&s->ssvm); + pool_put (sm->segments, s); +} + +svm_fifo_t * +svm_fifo_segment_alloc_fifo (svm_fifo_segment_private_t * s, + u32 data_size_in_bytes) +{ + ssvm_shared_header_t *sh; + svm_fifo_segment_header_t *fsh; + svm_fifo_t *f; + void *oldheap; + + sh = s->ssvm.sh; + fsh = (svm_fifo_segment_header_t *) sh->opaque[0]; + oldheap = ssvm_push_heap (sh); + + /* Note: this can fail, in which case: create another segment */ + f = svm_fifo_create (data_size_in_bytes); + if (f == 0) + { + ssvm_pop_heap (oldheap); + return (0); + } + + vec_add1 (fsh->fifos, f); + + ssvm_pop_heap (oldheap); + return (f); +} + +void +svm_fifo_segment_free_fifo (svm_fifo_segment_private_t * s, svm_fifo_t * f) +{ + ssvm_shared_header_t *sh; + svm_fifo_segment_header_t *fsh; + void *oldheap; + int i; + + sh = s->ssvm.sh; + fsh = (svm_fifo_segment_header_t *) sh->opaque[0]; + oldheap = ssvm_push_heap (sh); + + for (i = 0; i < vec_len (fsh->fifos); i++) + { + if (fsh->fifos[i] == f) + { + vec_delete (fsh->fifos, 1, i); + goto found; + } + } + clib_warning ("fifo 0x%llx not found in fifo table...", f); + +found: + clib_mem_free (f); + ssvm_pop_heap (oldheap); +} + +void +svm_fifo_segment_init (u64 baseva, u32 timeout_in_seconds) +{ + svm_fifo_segment_main_t *sm = &svm_fifo_segment_main; + + sm->next_baseva = baseva; + sm->timeout_in_seconds = timeout_in_seconds; +} + +u32 +svm_fifo_segment_index (svm_fifo_segment_private_t * s) +{ + return s - svm_fifo_segment_main.segments; +} + +/* + * fd.io coding-style-patch-verification: ON + * + * Local Variables: + * eval: (c-set-style "gnu") + * End: + */ diff --git a/src/svm/svm_fifo_segment.h b/src/svm/svm_fifo_segment.h new file mode 100644 index 00000000..793fa7c8 --- /dev/null +++ b/src/svm/svm_fifo_segment.h @@ -0,0 +1,89 @@ +/* + * Copyright (c) 2016 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ +#ifndef __included_ssvm_fifo_segment_h__ +#define __included_ssvm_fifo_segment_h__ + +#include "svm_fifo.h" +#include "ssvm.h" + +typedef struct +{ + volatile svm_fifo_t **fifos; + u8 *segment_name; +} svm_fifo_segment_header_t; + +typedef struct +{ + ssvm_private_t ssvm; + svm_fifo_segment_header_t *h; +} svm_fifo_segment_private_t; + +typedef struct +{ + /** pool of segments */ + svm_fifo_segment_private_t *segments; + /* Where to put the next one */ + u64 next_baseva; + u32 timeout_in_seconds; +} svm_fifo_segment_main_t; + +extern svm_fifo_segment_main_t svm_fifo_segment_main; + +typedef struct +{ + char *segment_name; + u32 segment_size; + u32 new_segment_index; +} svm_fifo_segment_create_args_t; + +static inline svm_fifo_segment_private_t * +svm_fifo_get_segment (u32 segment_index) +{ + svm_fifo_segment_main_t *ssm = &svm_fifo_segment_main; + return vec_elt_at_index (ssm->segments, segment_index); +} + +#define foreach_ssvm_fifo_segment_api_error \ +_(OUT_OF_SPACE, "Out of space in segment", -200) + +typedef enum +{ +#define _(n,s,c) SSVM_FIFO_SEGMENT_API_ERROR_##n = c, + foreach_ssvm_fifo_segment_api_error +#undef _ +} ssvm_fifo_segment_api_error_enum_t; + +int svm_fifo_segment_create (svm_fifo_segment_create_args_t * a); +int svm_fifo_segment_attach (svm_fifo_segment_create_args_t * a); +void svm_fifo_segment_delete (svm_fifo_segment_private_t * s); + +svm_fifo_t *svm_fifo_segment_alloc_fifo (svm_fifo_segment_private_t * s, + u32 data_size_in_bytes); +void svm_fifo_segment_free_fifo (svm_fifo_segment_private_t * s, + svm_fifo_t * f); + +void svm_fifo_segment_init (u64 baseva, u32 timeout_in_seconds); + +u32 svm_fifo_segment_index (svm_fifo_segment_private_t * s); + +#endif /* __included_ssvm_fifo_segment_h__ */ + +/* + * fd.io coding-style-patch-verification: ON + * + * Local Variables: + * eval: (c-set-style "gnu") + * End: + */ diff --git a/src/svm/test_svm_fifo1.c b/src/svm/test_svm_fifo1.c new file mode 100644 index 00000000..355653df --- /dev/null +++ b/src/svm/test_svm_fifo1.c @@ -0,0 +1,361 @@ +/* + * Copyright (c) 2016 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +#include "svm_fifo_segment.h" + +clib_error_t * +hello_world (int verbose) +{ + svm_fifo_segment_create_args_t _a, *a = &_a; + svm_fifo_segment_private_t *sp; + svm_fifo_t *f; + int rv; + u8 *test_data; + u8 *retrieved_data = 0; + clib_error_t *error = 0; + int pid = getpid (); + + memset (a, 0, sizeof (*a)); + + a->segment_name = "fifo-test1"; + a->segment_size = 256 << 10; + + rv = svm_fifo_segment_create (a); + + if (rv) + return clib_error_return (0, "svm_fifo_segment_create returned %d", rv); + + sp = svm_fifo_get_segment (a->new_segment_index); + + f = svm_fifo_segment_alloc_fifo (sp, 4096); + + if (f == 0) + return clib_error_return (0, "svm_fifo_segment_alloc_fifo failed"); + + test_data = format (0, "Hello world%c", 0); + vec_validate (retrieved_data, vec_len (test_data) - 1); + + while (svm_fifo_max_enqueue (f) >= vec_len (test_data)) + svm_fifo_enqueue_nowait (f, pid, vec_len (test_data), test_data); + + while (svm_fifo_max_dequeue (f) >= vec_len (test_data)) + svm_fifo_dequeue_nowait (f, pid, vec_len (retrieved_data), + retrieved_data); + + while (svm_fifo_max_enqueue (f) >= vec_len (test_data)) + svm_fifo_enqueue_nowait (f, pid, vec_len (test_data), test_data); + + while (svm_fifo_max_dequeue (f) >= vec_len (test_data)) + svm_fifo_dequeue_nowait (f, pid, vec_len (retrieved_data), + retrieved_data); + + if (!memcmp (retrieved_data, test_data, vec_len (test_data))) + error = clib_error_return (0, "data test OK, got '%s'", retrieved_data); + else + error = clib_error_return (0, "data test FAIL!"); + + svm_fifo_segment_free_fifo (sp, f); + + return error; +} + +clib_error_t * +master (int verbose) +{ + svm_fifo_segment_create_args_t _a, *a = &_a; + svm_fifo_segment_private_t *sp; + svm_fifo_t *f; + int rv; + u8 *test_data; + u8 *retrieved_data = 0; + int i; + int pid = getpid (); + + memset (a, 0, sizeof (*a)); + + a->segment_name = "fifo-test1"; + a->segment_size = 256 << 10; + + rv = svm_fifo_segment_create (a); + + if (rv) + return clib_error_return (0, "svm_fifo_segment_create returned %d", rv); + + sp = svm_fifo_get_segment (a->new_segment_index); + + f = svm_fifo_segment_alloc_fifo (sp, 4096); + + if (f == 0) + return clib_error_return (0, "svm_fifo_segment_alloc_fifo failed"); + + test_data = format (0, "Hello world%c", 0); + vec_validate (retrieved_data, vec_len (test_data) - 1); + + for (i = 0; i < 1000; i++) + svm_fifo_enqueue_nowait (f, pid, vec_len (test_data), test_data); + + return clib_error_return (0, "master (enqueue) done"); +} + +clib_error_t * +mempig (int verbose) +{ + svm_fifo_segment_create_args_t _a, *a = &_a; + svm_fifo_segment_private_t *sp; + svm_fifo_t *f; + svm_fifo_t **flist = 0; + int rv; + int i; + + memset (a, 0, sizeof (*a)); + + a->segment_name = "fifo-test1"; + a->segment_size = 256 << 10; + + rv = svm_fifo_segment_create (a); + + if (rv) + return clib_error_return (0, "svm_fifo_segment_create returned %d", rv); + + sp = svm_fifo_get_segment (a->new_segment_index); + + for (i = 0; i < 1000; i++) + { + f = svm_fifo_segment_alloc_fifo (sp, 4096); + if (f == 0) + break; + vec_add1 (flist, f); + } + + fformat (stdout, "Try #1: created %d fifos...\n", vec_len (flist)); + for (i = 0; i < vec_len (flist); i++) + { + f = flist[i]; + svm_fifo_segment_free_fifo (sp, f); + } + + _vec_len (flist) = 0; + + for (i = 0; i < 1000; i++) + { + f = svm_fifo_segment_alloc_fifo (sp, 4096); + if (f == 0) + break; + vec_add1 (flist, f); + } + + fformat (stdout, "Try #2: created %d fifos...\n", vec_len (flist)); + for (i = 0; i < vec_len (flist); i++) + { + f = flist[i]; + svm_fifo_segment_free_fifo (sp, f); + } + + return 0; +} + +clib_error_t * +offset (int verbose) +{ + svm_fifo_segment_create_args_t _a, *a = &_a; + svm_fifo_segment_private_t *sp; + svm_fifo_t *f; + int rv; + u32 *test_data = 0; + u32 *recovered_data = 0; + int i; + int pid = getpid (); + + memset (a, 0, sizeof (*a)); + + a->segment_name = "fifo-test1"; + a->segment_size = 256 << 10; + + rv = svm_fifo_segment_create (a); + + if (rv) + return clib_error_return (0, "svm_fifo_segment_create returned %d", rv); + + sp = svm_fifo_get_segment (a->new_segment_index); + + f = svm_fifo_segment_alloc_fifo (sp, 200 << 10); + + if (f == 0) + return clib_error_return (0, "svm_fifo_segment_alloc_fifo failed"); + + for (i = 0; i < (3 * 1024); i++) + vec_add1 (test_data, i); + + /* Enqueue the first 1024 u32's */ + svm_fifo_enqueue_nowait (f, pid, 4096 /* bytes to enqueue */ , + (u8 *) test_data); + + /* Enqueue the third 1024 u32's 2048 ahead of the current tail */ + svm_fifo_enqueue_with_offset (f, pid, 4096, 4096, (u8 *) & test_data[2048]); + + /* Enqueue the second 1024 u32's at the current tail */ + svm_fifo_enqueue_nowait (f, pid, 4096 /* bytes to enqueue */ , + (u8 *) & test_data[1024]); + + vec_validate (recovered_data, (3 * 1024) - 1); + + svm_fifo_dequeue_nowait (f, pid, 3 * 4096, (u8 *) recovered_data); + + for (i = 0; i < (3 * 1024); i++) + { + if (recovered_data[i] != test_data[i]) + { + clib_warning ("[%d] expected %d recovered %d", i, + test_data[i], recovered_data[i]); + return clib_error_return (0, "offset test FAILED"); + } + } + + return clib_error_return (0, "offset test OK"); +} + +clib_error_t * +slave (int verbose) +{ + svm_fifo_segment_create_args_t _a, *a = &_a; + svm_fifo_segment_private_t *sp; + svm_fifo_segment_header_t *fsh; + svm_fifo_t *f; + ssvm_shared_header_t *sh; + int rv; + u8 *test_data; + u8 *retrieved_data = 0; + int pid = getpid (); + int i; + + memset (a, 0, sizeof (*a)); + + a->segment_name = "fifo-test1"; + + rv = svm_fifo_segment_attach (a); + + if (rv) + return clib_error_return (0, "svm_fifo_segment_attach returned %d", rv); + + sp = svm_fifo_get_segment (a->new_segment_index); + sh = sp->ssvm.sh; + fsh = (svm_fifo_segment_header_t *) sh->opaque[0]; + + /* might wanna wait.. */ + f = (svm_fifo_t *) fsh->fifos[0]; + + /* Lazy bastards united */ + test_data = format (0, "Hello world%c", 0); + vec_validate (retrieved_data, vec_len (test_data) - 1); + + for (i = 0; i < 1000; i++) + { + svm_fifo_dequeue_nowait (f, pid, vec_len (retrieved_data), + retrieved_data); + if (memcmp (retrieved_data, test_data, vec_len (retrieved_data))) + return clib_error_return (0, "retrieved data incorrect, '%s'", + retrieved_data); + } + + return clib_error_return (0, "slave (dequeue) done"); +} + + +int +test_ssvm_fifo1 (unformat_input_t * input) +{ + clib_error_t *error = 0; + int verbose = 0; + int test_id = 0; + + svm_fifo_segment_init (0x200000000ULL, 20); + + while (unformat_check_input (input) != UNFORMAT_END_OF_INPUT) + { + if (unformat (input, "verbose %d", &verbose)) + ; + else if (unformat (input, "verbose")) + verbose = 1; + else if (unformat (input, "master")) + test_id = 1; + else if (unformat (input, "slave")) + test_id = 2; + else if (unformat (input, "mempig")) + test_id = 3; + else if (unformat (input, "offset")) + test_id = 4; + else + { + error = clib_error_create ("unknown input `%U'\n", + format_unformat_error, input); + goto out; + } + } + + switch (test_id) + { + case 0: + error = hello_world (verbose); + break; + + case 1: + error = master (verbose); + break; + + case 2: + error = slave (verbose); + break; + + case 3: + error = mempig (verbose); + break; + + case 4: + error = offset (verbose); + break; + + default: + error = clib_error_return (0, "test id %d unknown", test_id); + break; + } + +out: + if (error) + clib_error_report (error); + + return 0; +} + + + +int +main (int argc, char *argv[]) +{ + unformat_input_t i; + int r; + + unformat_init_command_line (&i, argv); + r = test_ssvm_fifo1 (&i); + unformat_free (&i); + return r; +} + +/* + * fd.io coding-style-patch-verification: ON + * + * Local Variables: + * eval: (c-set-style "gnu") + * End: + */ diff --git a/src/uri.am b/src/uri.am new file mode 100644 index 00000000..8cdd77c6 --- /dev/null +++ b/src/uri.am @@ -0,0 +1,22 @@ +# Copyright (c) 2016 Cisco and/or its affiliates. +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at: +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +noinst_PROGRAMS += uri_udp_test2 uri_tcp_test + +uri_udp_test2_SOURCES = uri/uri_udp_test2.c +uri_udp_test2_LDADD = libvlibmemoryclient.la libvlibapi.la libsvm.la \ + libvppinfra.la -lpthread -lm -lrt + +uri_tcp_test_SOURCES = uri/uri_tcp_test.c +uri_tcp_test_LDADD = libvlibmemoryclient.la libvlibapi.la libsvm.la \ + libvppinfra.la -lpthread -lm -lrt diff --git a/src/uri/uri_tcp_test.c b/src/uri/uri_tcp_test.c new file mode 100644 index 00000000..ed5a37d8 --- /dev/null +++ b/src/uri/uri_tcp_test.c @@ -0,0 +1,916 @@ +/* + * Copyright (c) 2016 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +#include +#include +#include +#include +#include +#include +#include + +#include "../vnet/session/application_interface.h" + +#define vl_typedefs /* define message structures */ +#include +#undef vl_typedefs + +/* declare message handlers for each api */ + +#define vl_endianfun /* define message structures */ +#include +#undef vl_endianfun + +/* instantiate all the print functions we know about */ +#define vl_print(handle, ...) +#define vl_printfun +#include +#undef vl_printfun + +/* Satisfy external references when not linking with -lvlib */ +vlib_main_t vlib_global_main; +vlib_main_t **vlib_mains; + +typedef struct +{ + svm_fifo_t * server_rx_fifo; + svm_fifo_t * server_tx_fifo; + + u32 vpp_session_index; + u32 vpp_session_thread; +} session_t; + +typedef enum +{ + STATE_START, + STATE_READY, + STATE_DISCONNECTING, + STATE_FAILED +} connection_state_t; + +typedef struct +{ + /* vpe input queue */ + unix_shared_memory_queue_t *vl_input_queue; + + /* API client handle */ + u32 my_client_index; + + /* The URI we're playing with */ + u8 * uri; + + /* Session pool */ + session_t * sessions; + + /* Hash table for disconnect processing */ + uword * session_index_by_vpp_handles; + + /* intermediate rx buffer */ + u8 * rx_buf; + + /* URI for slave's connect */ + u8 * connect_uri; + + u32 connected_session_index; + + int i_am_master; + + /* drop all packets */ + int drop_packets; + + /* Our event queue */ + unix_shared_memory_queue_t * our_event_queue; + + /* $$$ single thread only for the moment */ + unix_shared_memory_queue_t * vpp_event_queue; + + pid_t my_pid; + + /* For deadman timers */ + clib_time_t clib_time; + + /* State of the connection, shared between msg RX thread and main thread */ + volatile connection_state_t state; + + /* Signal variables */ + volatile int time_to_stop; + volatile int time_to_print_stats; + + u32 configured_segment_size; + + /* VNET_API_ERROR_FOO -> "Foo" hash table */ + uword * error_string_by_error_number; + + /* convenience */ + svm_fifo_segment_main_t * segment_main; + + u8 *connect_test_data; +} uri_tcp_test_main_t; + +uri_tcp_test_main_t uri_tcp_test_main; + +#if CLIB_DEBUG > 0 +#define NITER 10000 +#else +#define NITER 4000000 +#endif + +int +wait_for_state_change (uri_tcp_test_main_t * utm, connection_state_t state) +{ +#if CLIB_DEBUG > 0 +#define TIMEOUT 600.0 +#else +#define TIMEOUT 600.0 +#endif + + f64 timeout = clib_time_now (&utm->clib_time) + TIMEOUT; + + while (clib_time_now (&utm->clib_time) < timeout) + { + if (utm->state == state) + return 0; + if (utm->state == STATE_FAILED) + return -1; + } + clib_warning ("timeout waiting for STATE_READY"); + return -1; +} + +static void +init_error_string_table (uri_tcp_test_main_t * utm) +{ + utm->error_string_by_error_number = hash_create (0, sizeof (uword)); + +#define _(n,v,s) hash_set (utm->error_string_by_error_number, -v, s); + foreach_vnet_api_error; +#undef _ + + hash_set (utm->error_string_by_error_number, 99, "Misc"); +} + +static void +stop_signal (int signum) +{ + uri_tcp_test_main_t *um = &uri_tcp_test_main; + + um->time_to_stop = 1; +} + +static void +stats_signal (int signum) +{ + uri_tcp_test_main_t *um = &uri_tcp_test_main; + + um->time_to_print_stats = 1; +} + +static clib_error_t * +setup_signal_handlers (void) +{ + signal (SIGINT, stats_signal); + signal (SIGQUIT, stop_signal); + signal (SIGTERM, stop_signal); + + return 0; +} + +void +vlib_cli_output (struct vlib_main_t *vm, char *fmt, ...) +{ + clib_warning ("BUG"); +} + +int +connect_to_vpp (char *name) +{ + uri_tcp_test_main_t *utm = &uri_tcp_test_main; + api_main_t *am = &api_main; + + if (vl_client_connect_to_vlib ("/vpe-api", name, 32) < 0) + return -1; + + utm->vl_input_queue = am->shmem_hdr->vl_input_queue; + utm->my_client_index = am->my_client_index; + + return 0; +} + +static void +vl_api_map_another_segment_t_handler (vl_api_map_another_segment_t *mp) +{ + svm_fifo_segment_create_args_t _a, *a = &_a; + int rv; + + a->segment_name = (char *) mp->segment_name; + a->segment_size = mp->segment_size; + /* Attach to the segment vpp created */ + rv = svm_fifo_segment_attach (a); + if (rv) + { + clib_warning ("svm_fifo_segment_attach ('%s') failed", + mp->segment_name); + return; + } + clib_warning ("Mapped new segment '%s' size %d", mp->segment_name, + mp->segment_size); +} + +static void +vl_api_disconnect_session_t_handler (vl_api_disconnect_session_t * mp) +{ + uri_tcp_test_main_t *utm = &uri_tcp_test_main; + session_t * session; + vl_api_disconnect_session_reply_t * rmp; + uword * p; + int rv = 0; + u64 key; + + key = (((u64)mp->session_thread_index) << 32) | (u64)mp->session_index; + + p = hash_get (utm->session_index_by_vpp_handles, key); + + if (p) + { + session = pool_elt_at_index (utm->sessions, p[0]); + hash_unset (utm->session_index_by_vpp_handles, key); + pool_put (utm->sessions, session); + } + else + { + clib_warning ("couldn't find session key %llx", key); + rv = -11; + } + + rmp = vl_msg_api_alloc (sizeof (*rmp)); + memset (rmp, 0, sizeof (*rmp)); + + rmp->_vl_msg_id = ntohs (VL_API_DISCONNECT_SESSION_REPLY); + rmp->retval = rv; + rmp->session_index = mp->session_index; + rmp->session_thread_index = mp->session_thread_index; + vl_msg_api_send_shmem (utm->vl_input_queue, (u8 *)&rmp); +} + +static void +vl_api_reset_session_t_handler (vl_api_reset_session_t * mp) +{ + uri_tcp_test_main_t *utm = &uri_tcp_test_main; + session_t * session; + vl_api_reset_session_reply_t * rmp; + uword * p; + int rv = 0; + u64 key; + + key = (((u64)mp->session_thread_index) << 32) | (u64)mp->session_index; + + p = hash_get(utm->session_index_by_vpp_handles, key); + + if (p) + { + session = pool_elt_at_index(utm->sessions, p[0]); + hash_unset(utm->session_index_by_vpp_handles, key); + pool_put(utm->sessions, session); + } + else + { + clib_warning("couldn't find session key %llx", key); + rv = -11; + } + + rmp = vl_msg_api_alloc (sizeof (*rmp)); + memset (rmp, 0, sizeof (*rmp)); + rmp->_vl_msg_id = ntohs (VL_API_DISCONNECT_SESSION_REPLY); + rmp->retval = rv; + rmp->session_index = mp->session_index; + rmp->session_thread_index = mp->session_thread_index; + vl_msg_api_send_shmem (utm->vl_input_queue, (u8 *)&rmp); +} + +void +handle_fifo_event_connect_rx (uri_tcp_test_main_t *utm, session_fifo_event_t * e) +{ + svm_fifo_t * rx_fifo; + int n_read, bytes; + + rx_fifo = e->fifo; + + bytes = e->enqueue_length; + do + { + n_read = svm_fifo_dequeue_nowait (rx_fifo, 0, vec_len(utm->rx_buf), + utm->rx_buf); + if (n_read > 0) + bytes -= n_read; + } + while (n_read < 0 || bytes > 0); + + // bytes_to_read = svm_fifo_max_dequeue (rx_fifo); + // + // bytes_to_read = vec_len(utm->rx_buf) > bytes_to_read ? + // bytes_to_read : vec_len(utm->rx_buf); + // + // buffer_offset = 0; + // while (bytes_to_read > 0) + // { + // rv = svm_fifo_dequeue_nowait2 (rx_fifo, mypid, + // bytes_to_read, + // utm->rx_buf + buffer_offset); + // if (rv > 0) + // { + // bytes_to_read -= rv; + // buffer_offset += rv; + // bytes_received += rv; + // } + // } + + + // while (bytes_received < bytes_sent) + // { + // rv = svm_fifo_dequeue_nowait2 (rx_fifo, mypid, + // vec_len (utm->rx_buf), + // utm->rx_buf); + // if (rv > 0) + // { + //#if CLIB_DEBUG > 0 + // int j; + // for (j = 0; j < rv; j++) + // { + // if (utm->rx_buf[j] != ((bytes_received + j) & 0xff)) + // { + // clib_warning ("error at byte %lld, 0x%x not 0x%x", + // bytes_received + j, + // utm->rx_buf[j], + // ((bytes_received + j )&0xff)); + // } + // } + //#endif + // bytes_received += (u64) rv; + // } + // } +} + +void +handle_connect_event_queue (uri_tcp_test_main_t * utm) +{ + session_fifo_event_t _e, *e = &_e;; + + unix_shared_memory_queue_sub (utm->our_event_queue, (u8 *) e, 0 /* nowait */); + switch (e->event_type) + { + case FIFO_EVENT_SERVER_RX: + handle_fifo_event_connect_rx (utm, e); + break; + + case FIFO_EVENT_SERVER_EXIT: + return; + + default: + clib_warning("unknown event type %d", e->event_type); + break; + } +} + +void +uri_tcp_connect_send (uri_tcp_test_main_t *utm) +{ + u8 *test_data = utm->connect_test_data; + u64 bytes_sent = 0; + int rv; + int mypid = getpid(); + session_t * session; + svm_fifo_t *tx_fifo; + int buffer_offset, bytes_to_send = 0; + session_fifo_event_t evt; + static int serial_number = 0; + int i; + u32 max_chunk = 64 << 10, write; + + session = pool_elt_at_index (utm->sessions, utm->connected_session_index); + tx_fifo = session->server_tx_fifo; + + vec_validate (utm->rx_buf, vec_len (test_data) - 1); + + for (i = 0; i < 10; i++) + { + bytes_to_send = vec_len (test_data); + buffer_offset = 0; + while (bytes_to_send > 0) + { + write = bytes_to_send > max_chunk ? max_chunk : bytes_to_send; + rv = svm_fifo_enqueue_nowait (tx_fifo, mypid, write, + test_data + buffer_offset); + + if (rv > 0) + { + bytes_to_send -= rv; + buffer_offset += rv; + bytes_sent += rv; + + /* Fabricate TX event, send to vpp */ + evt.fifo = tx_fifo; + evt.event_type = FIFO_EVENT_SERVER_TX; + /* $$$$ for event logging */ + evt.enqueue_length = rv; + evt.event_id = serial_number++; + + unix_shared_memory_queue_add (utm->vpp_event_queue, (u8 *) &evt, + 0 /* do wait for mutex */); + } + } + } +} + +static void +uri_tcp_client_test (uri_tcp_test_main_t * utm) +{ + vl_api_connect_uri_t * cmp; + vl_api_disconnect_session_t *dmp; + session_t *connected_session; + int i; + + cmp = vl_msg_api_alloc (sizeof (*cmp)); + memset (cmp, 0, sizeof (*cmp)); + + cmp->_vl_msg_id = ntohs (VL_API_CONNECT_URI); + cmp->client_index = utm->my_client_index; + cmp->context = ntohl(0xfeedface); + memcpy (cmp->uri, utm->connect_uri, vec_len (utm->connect_uri)); + vl_msg_api_send_shmem (utm->vl_input_queue, (u8 *)&cmp); + + if (wait_for_state_change (utm, STATE_READY)) + { + return; + } + + /* Init test data */ + vec_validate (utm->connect_test_data, 64 * 1024 - 1); + for (i = 0; i < vec_len (utm->connect_test_data); i++) + utm->connect_test_data[i] = i & 0xff; + + /* Start reader thread */ + /* handle_connect_event_queue (utm); */ + + /* Start send */ + uri_tcp_connect_send (utm); + + /* Disconnect */ + connected_session = pool_elt_at_index(utm->sessions, + utm->connected_session_index); + dmp = vl_msg_api_alloc (sizeof (*dmp)); + memset (dmp, 0, sizeof (*dmp)); + dmp->_vl_msg_id = ntohs (VL_API_DISCONNECT_SESSION); + dmp->client_index = utm->my_client_index; + dmp->session_index = connected_session->vpp_session_index; + dmp->session_thread_index = connected_session->vpp_session_thread; + vl_msg_api_send_shmem (utm->vl_input_queue, (u8 *)&dmp); +} + +void +handle_fifo_event_server_rx (uri_tcp_test_main_t *utm, session_fifo_event_t * e) +{ + svm_fifo_t * rx_fifo, * tx_fifo; + int n_read; + + session_fifo_event_t evt; + unix_shared_memory_queue_t *q; + int rv, bytes; + + rx_fifo = e->fifo; + tx_fifo = utm->sessions[rx_fifo->client_session_index].server_tx_fifo; + + bytes = e->enqueue_length; + do + { + n_read = svm_fifo_dequeue_nowait (rx_fifo, 0, vec_len(utm->rx_buf), + utm->rx_buf); + + /* Reflect if a non-drop session */ + if (!utm->drop_packets && n_read > 0) + { + do + { + rv = svm_fifo_enqueue_nowait (tx_fifo, 0, n_read, utm->rx_buf); + } + while (rv == -2); + + /* Fabricate TX event, send to vpp */ + evt.fifo = tx_fifo; + evt.event_type = FIFO_EVENT_SERVER_TX; + /* $$$$ for event logging */ + evt.enqueue_length = n_read; + evt.event_id = e->event_id; + q = utm->vpp_event_queue; + unix_shared_memory_queue_add (q, (u8 *) &evt, 0 /* do wait for mutex */); + } + + if (n_read > 0) + bytes -= n_read; + } + while (n_read < 0 || bytes > 0); +} + +void +handle_event_queue (uri_tcp_test_main_t * utm) +{ + session_fifo_event_t _e, *e = &_e;; + + while (1) + { + unix_shared_memory_queue_sub (utm->our_event_queue, (u8 *)e, + 0 /* nowait */); + switch (e->event_type) + { + case FIFO_EVENT_SERVER_RX: + handle_fifo_event_server_rx (utm, e); + break; + + case FIFO_EVENT_SERVER_EXIT: + return; + + default: + clib_warning ("unknown event type %d", e->event_type); + break; + } + if (PREDICT_FALSE(utm->time_to_stop == 1)) + break; + if (PREDICT_FALSE(utm->time_to_print_stats == 1)) + { + utm->time_to_print_stats = 0; + fformat(stdout, "%d connections\n", pool_elts (utm->sessions)); + } + } +} + +static void +vl_api_bind_uri_reply_t_handler (vl_api_bind_uri_reply_t * mp) +{ + uri_tcp_test_main_t *utm = &uri_tcp_test_main; + svm_fifo_segment_create_args_t _a, *a = &_a; + int rv; + + if (mp->retval) + { + clib_warning("bind failed: %d", mp->retval); + return; + } + + if (mp->segment_name_length == 0) + { + clib_warning("segment_name_length zero"); + return; + } + + a->segment_name = (char *) mp->segment_name; + a->segment_size = mp->segment_size; + + ASSERT(mp->server_event_queue_address); + + /* Attach to the segment vpp created */ + rv = svm_fifo_segment_attach (a); + if (rv) + { + clib_warning("svm_fifo_segment_attach ('%s') failed", mp->segment_name); + return; + } + + utm->our_event_queue = + (unix_shared_memory_queue_t *) mp->server_event_queue_address; + + utm->state = STATE_READY; +} + +static void +vl_api_connect_uri_reply_t_handler (vl_api_connect_uri_reply_t * mp) +{ + uri_tcp_test_main_t *utm = &uri_tcp_test_main; + svm_fifo_segment_create_args_t _a, *a = &_a; + session_t *session; + u32 session_index; + svm_fifo_t *rx_fifo, *tx_fifo; + int rv; + + if (mp->retval) + { + clib_warning ("connection failed with code: %d", mp->retval); + utm->state = STATE_FAILED; + return; + } + /* + * Attatch to segment + */ + + if (mp->segment_name_length == 0) + { + clib_warning ("segment_name_length zero"); + utm->state = STATE_FAILED; + return; + } + + a->segment_name = (char *) mp->segment_name; + a->segment_size = mp->segment_size; + + ASSERT(mp->client_event_queue_address); + + /* Attach to the segment vpp created */ + rv = svm_fifo_segment_attach (a); + if (rv) + { + clib_warning ("svm_fifo_segment_attach ('%s') failed", + mp->segment_name); + return; + } + + /* + * Save the queues + */ + + utm->our_event_queue = (unix_shared_memory_queue_t *) + mp->client_event_queue_address; + + utm->vpp_event_queue = (unix_shared_memory_queue_t *) + mp->vpp_event_queue_address; + + /* + * Setup session + */ + + pool_get (utm->sessions, session); + session_index = session - utm->sessions; + + rx_fifo = (svm_fifo_t *)mp->server_rx_fifo; + rx_fifo->client_session_index = session_index; + tx_fifo = (svm_fifo_t *)mp->server_tx_fifo; + tx_fifo->client_session_index = session_index; + + session->server_rx_fifo = rx_fifo; + session->server_tx_fifo = tx_fifo; + session->vpp_session_index = mp->session_index; + session->vpp_session_thread = mp->session_thread_index; + + /* Save handle */ + utm->connected_session_index = session_index; + + utm->state = STATE_READY; +} + +void +uri_tcp_bind (uri_tcp_test_main_t *utm) +{ + vl_api_bind_uri_t * bmp; + u32 fifo_size = 3 << 20; + bmp = vl_msg_api_alloc (sizeof (*bmp)); + memset (bmp, 0, sizeof (*bmp)); + + bmp->_vl_msg_id = ntohs (VL_API_BIND_URI); + bmp->client_index = utm->my_client_index; + bmp->context = ntohl(0xfeedface); + bmp->initial_segment_size = 256<<20; /* size of initial segment */ + bmp->options[SESSION_OPTIONS_FLAGS] = + SESSION_OPTIONS_FLAGS_USE_FIFO | SESSION_OPTIONS_FLAGS_ADD_SEGMENT; + bmp->options[SESSION_OPTIONS_RX_FIFO_SIZE] = fifo_size; + bmp->options[SESSION_OPTIONS_TX_FIFO_SIZE] = fifo_size; + bmp->options[SESSION_OPTIONS_ADD_SEGMENT_SIZE] = 128<<20; + memcpy (bmp->uri, utm->uri, vec_len (utm->uri)); + vl_msg_api_send_shmem (utm->vl_input_queue, (u8 *)&bmp); +} + +static void +vl_api_unbind_uri_reply_t_handler (vl_api_unbind_uri_reply_t *mp) +{ + uri_tcp_test_main_t *utm = &uri_tcp_test_main; + + if (mp->retval != 0) + clib_warning ("returned %d", ntohl(mp->retval)); + + utm->state = STATE_START; +} + +void +uri_tcp_unbind (uri_tcp_test_main_t *utm) +{ + vl_api_unbind_uri_t * ump; + + ump = vl_msg_api_alloc (sizeof (*ump)); + memset (ump, 0, sizeof (*ump)); + + ump->_vl_msg_id = ntohs (VL_API_UNBIND_URI); + ump->client_index = utm->my_client_index; + memcpy (ump->uri, utm->uri, vec_len (utm->uri)); + vl_msg_api_send_shmem (utm->vl_input_queue, (u8 *)&ump); +} + +static void +vl_api_accept_session_t_handler (vl_api_accept_session_t * mp) +{ + uri_tcp_test_main_t *utm = &uri_tcp_test_main; + vl_api_accept_session_reply_t *rmp; + svm_fifo_t * rx_fifo, * tx_fifo; + session_t * session; + static f64 start_time; + u64 key; + u32 session_index; + + if (start_time == 0.0) + start_time = clib_time_now (&utm->clib_time); + + utm->vpp_event_queue = (unix_shared_memory_queue_t *) + mp->vpp_event_queue_address; + + /* Allocate local session and set it up */ + pool_get (utm->sessions, session); + session_index = session - utm->sessions; + + rx_fifo = (svm_fifo_t *)mp->server_rx_fifo; + rx_fifo->client_session_index = session_index; + tx_fifo = (svm_fifo_t *)mp->server_tx_fifo; + tx_fifo->client_session_index = session_index; + + session->server_rx_fifo = rx_fifo; + session->server_tx_fifo = tx_fifo; + + /* Add it to lookup table */ + key = (((u64)mp->session_thread_index) << 32) | (u64)mp->session_index; + hash_set (utm->session_index_by_vpp_handles, key, session_index); + + utm->state = STATE_READY; + + /* Stats printing */ + if (pool_elts (utm->sessions) && (pool_elts(utm->sessions) % 20000) == 0) + { + f64 now = clib_time_now (&utm->clib_time); + fformat (stdout, "%d active sessions in %.2f seconds, %.2f/sec...\n", + pool_elts(utm->sessions), now - start_time, + (f64)pool_elts(utm->sessions) / (now - start_time)); + } + + /* Send accept reply to vpp */ + rmp = vl_msg_api_alloc (sizeof (*rmp)); + memset (rmp, 0, sizeof (*rmp)); + rmp->_vl_msg_id = ntohs (VL_API_ACCEPT_SESSION_REPLY); + rmp->session_type = mp->session_type; + rmp->session_index = mp->session_index; + rmp->session_thread_index = mp->session_thread_index; + vl_msg_api_send_shmem (utm->vl_input_queue, (u8 *)&rmp); +} + +void +uri_tcp_server_test (uri_tcp_test_main_t * utm) +{ + + /* Bind to uri */ + uri_tcp_bind (utm); + + if (wait_for_state_change (utm, STATE_READY)) + { + clib_warning ("timeout waiting for STATE_READY"); + return; + } + + /* Enter handle event loop */ + handle_event_queue (utm); + + /* Cleanup */ + uri_tcp_unbind (utm); + + if (wait_for_state_change (utm, STATE_START)) + { + clib_warning ("timeout waiting for STATE_START"); + return; + } + + fformat (stdout, "Test complete...\n"); +} + +#define foreach_uri_msg \ +_(BIND_URI_REPLY, bind_uri_reply) \ +_(UNBIND_URI_REPLY, unbind_uri_reply) \ +_(ACCEPT_SESSION, accept_session) \ +_(CONNECT_URI_REPLY, connect_uri_reply) \ +_(DISCONNECT_SESSION, disconnect_session) \ +_(RESET_SESSION, reset_session) \ +_(MAP_ANOTHER_SEGMENT, map_another_segment) + +void +uri_api_hookup (uri_tcp_test_main_t * utm) +{ +#define _(N,n) \ + vl_msg_api_set_handlers(VL_API_##N, #n, \ + vl_api_##n##_t_handler, \ + vl_noop_handler, \ + vl_api_##n##_t_endian, \ + vl_api_##n##_t_print, \ + sizeof(vl_api_##n##_t), 1); + foreach_uri_msg; +#undef _ +} + +int +main (int argc, char **argv) +{ + uri_tcp_test_main_t *utm = &uri_tcp_test_main; + unformat_input_t _argv, *a = &_argv; + u8 *chroot_prefix; + u8 *heap; + u8 * bind_name = (u8 *) "tcp://0.0.0.0/1234"; + u32 tmp; + mheap_t *h; + session_t * session; + int i; + int i_am_master = 1, drop_packets = 0; + + clib_mem_init (0, 256 << 20); + + heap = clib_mem_get_per_cpu_heap (); + h = mheap_header (heap); + + /* make the main heap thread-safe */ + h->flags |= MHEAP_FLAG_THREAD_SAFE; + + vec_validate (utm->rx_buf, 65536); + + utm->session_index_by_vpp_handles = + hash_create (0, sizeof(uword)); + + utm->my_pid = getpid(); + utm->configured_segment_size = 1<<20; + + clib_time_init (&utm->clib_time); + init_error_string_table (utm); + svm_fifo_segment_init(0x200000000ULL, 20); + unformat_init_command_line (a, argv); + + while (unformat_check_input (a) != UNFORMAT_END_OF_INPUT) + { + if (unformat (a, "chroot prefix %s", &chroot_prefix)) + { + vl_set_memory_root_path ((char *) chroot_prefix); + } + else if (unformat (a, "uri %s", &bind_name)) + ; + else if (unformat (a, "segment-size %dM", &tmp)) + utm->configured_segment_size = tmp<<20; + else if (unformat (a, "segment-size %dG", &tmp)) + utm->configured_segment_size = tmp<<30; + else if (unformat (a, "master")) + i_am_master = 1; + else if (unformat (a, "slave")) + i_am_master = 0; + else if (unformat (a, "drop")) + drop_packets = 1; + else + { + fformat (stderr, "%s: usage [master|slave]\n"); + exit (1); + } + } + + utm->uri = format (0, "%s%c", bind_name, 0); + utm->i_am_master = i_am_master; + utm->segment_main = &svm_fifo_segment_main; + utm->drop_packets = drop_packets; + + utm->connect_uri = format (0, "tcp://6.0.1.2/1234%c", 0); + + setup_signal_handlers(); + uri_api_hookup (utm); + + if (connect_to_vpp (i_am_master? "uri_tcp_server":"uri_tcp_client") < 0) + { + svm_region_exit (); + fformat (stderr, "Couldn't connect to vpe, exiting...\n"); + exit (1); + } + + if (i_am_master == 0) + { + uri_tcp_client_test (utm); + exit (0); + } + + /* $$$$ hack preallocation */ + for (i = 0; i < 200000; i++) + { + pool_get (utm->sessions, session); + memset (session, 0, sizeof (*session)); + } + for (i = 0; i < 200000; i++) + pool_put_index (utm->sessions, i); + + uri_tcp_server_test (utm); + + vl_client_disconnect_from_vlib (); + exit (0); +} diff --git a/src/uri/uri_udp_test.c b/src/uri/uri_udp_test.c new file mode 100644 index 00000000..6f5284c9 --- /dev/null +++ b/src/uri/uri_udp_test.c @@ -0,0 +1,553 @@ +/* + * Copyright (c) 2016 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include + +#include + +#define vl_typedefs /* define message structures */ +#include +#undef vl_typedefs + +/* declare message handlers for each api */ + +#define vl_endianfun /* define message structures */ +#include +#undef vl_endianfun + +/* instantiate all the print functions we know about */ +#define vl_print(handle, ...) +#define vl_printfun +#include +#undef vl_printfun + +/* Satisfy external references when not linking with -lvlib */ +vlib_main_t vlib_global_main; +vlib_main_t **vlib_mains; + +typedef enum +{ + STATE_START, + STATE_READY, + STATE_DISCONNECTING, +} connection_state_t; + +typedef struct +{ + svm_fifo_t *server_rx_fifo; + svm_fifo_t *server_tx_fifo; +} session_t; + +typedef struct +{ + /* vpe input queue */ + unix_shared_memory_queue_t *vl_input_queue; + + /* API client handle */ + u32 my_client_index; + + /* The URI we're playing with */ + u8 *uri; + + /* Session pool */ + session_t *sessions; + + /* Hash table for disconnect processing */ + uword *session_index_by_vpp_handles; + + /* fifo segment */ + svm_fifo_segment_private_t *seg; + + /* intermediate rx buffer */ + u8 *rx_buf; + + /* Our event queue */ + unix_shared_memory_queue_t *our_event_queue; + + /* $$$ single thread only for the moment */ + unix_shared_memory_queue_t *vpp_event_queue; + + /* For deadman timers */ + clib_time_t clib_time; + + /* State of the connection, shared between msg RX thread and main thread */ + volatile connection_state_t state; + + volatile int time_to_stop; + volatile int time_to_print_stats; + + /* VNET_API_ERROR_FOO -> "Foo" hash table */ + uword *error_string_by_error_number; +} uri_udp_test_main_t; + +#if CLIB_DEBUG > 0 +#define NITER 1000 +#else +#define NITER 1000000 +#endif + +uri_udp_test_main_t uri_udp_test_main; + +static void +stop_signal (int signum) +{ + uri_udp_test_main_t *um = &uri_udp_test_main; + + um->time_to_stop = 1; +} + +static void +stats_signal (int signum) +{ + uri_udp_test_main_t *um = &uri_udp_test_main; + + um->time_to_print_stats = 1; +} + +static clib_error_t * +setup_signal_handlers (void) +{ + signal (SIGINT, stats_signal); + signal (SIGQUIT, stop_signal); + signal (SIGTERM, stop_signal); + + return 0; +} + +u8 * +format_api_error (u8 * s, va_list * args) +{ + uri_udp_test_main_t *utm = va_arg (*args, uri_udp_test_main_t *); + i32 error = va_arg (*args, u32); + uword *p; + + p = hash_get (utm->error_string_by_error_number, -error); + + if (p) + s = format (s, "%s", p[0]); + else + s = format (s, "%d", error); + return s; +} + +int +wait_for_state_change (uri_udp_test_main_t * utm, connection_state_t state) +{ + f64 timeout = clib_time_now (&utm->clib_time) + 5.0; + + while (clib_time_now (&utm->clib_time) < timeout) + { + if (utm->state == state) + return 0; + } + return -1; +} + +static void +vl_api_bind_uri_reply_t_handler (vl_api_bind_uri_reply_t * mp) +{ + uri_udp_test_main_t *utm = &uri_udp_test_main; + svm_fifo_segment_create_args_t _a, *a = &_a; + int rv; + + if (mp->segment_name_length == 0) + { + clib_warning ("segment_name_length zero"); + return; + } + + a->segment_name = (char *) mp->segment_name; + + /* Attach to the segment vpp created */ + rv = svm_fifo_segment_attach (a); + if (rv) + { + clib_warning ("sm_fifo_segment_create ('%s') failed", mp->segment_name); + return; + } + + utm->our_event_queue = (unix_shared_memory_queue_t *) + mp->server_event_queue_address; + + utm->state = STATE_READY; +} + +static void +vl_api_unbind_uri_reply_t_handler (vl_api_unbind_uri_reply_t * mp) +{ + uri_udp_test_main_t *utm = &uri_udp_test_main; + + if (mp->retval != 0) + clib_warning ("returned %d", ntohl (mp->retval)); + + utm->state = STATE_START; +} + +static void +vl_api_accept_session_t_handler (vl_api_accept_session_t * mp) +{ + uri_udp_test_main_t *utm = &uri_udp_test_main; + vl_api_accept_session_reply_t *rmp; + svm_fifo_t *rx_fifo, *tx_fifo; + session_t *session; + static f64 start_time; + u64 key; + + if (start_time == 0.0) + start_time = clib_time_now (&utm->clib_time); + + utm->vpp_event_queue = (unix_shared_memory_queue_t *) + mp->vpp_event_queue_address; + + pool_get (utm->sessions, session); + + rx_fifo = (svm_fifo_t *) mp->server_rx_fifo; + rx_fifo->client_session_index = session - utm->sessions; + tx_fifo = (svm_fifo_t *) mp->server_tx_fifo; + tx_fifo->client_session_index = session - utm->sessions; + + session->server_rx_fifo = rx_fifo; + session->server_tx_fifo = tx_fifo; + + key = (((u64) mp->session_thread_index) << 32) | (u64) mp->session_index; + + hash_set (utm->session_index_by_vpp_handles, key, session - utm->sessions); + + utm->state = STATE_READY; + + if (pool_elts (utm->sessions) && (pool_elts (utm->sessions) % 20000) == 0) + { + f64 now = clib_time_now (&utm->clib_time); + fformat (stdout, "%d active sessions in %.2f seconds, %.2f/sec...\n", + pool_elts (utm->sessions), now - start_time, + (f64) pool_elts (utm->sessions) / (now - start_time)); + } + + rmp = vl_msg_api_alloc (sizeof (*rmp)); + memset (rmp, 0, sizeof (*rmp)); + rmp->_vl_msg_id = ntohs (VL_API_ACCEPT_SESSION_REPLY); + rmp->session_type = mp->session_type; + rmp->session_index = mp->session_index; + rmp->session_thread_index = mp->session_thread_index; + vl_msg_api_send_shmem (utm->vl_input_queue, (u8 *) & rmp); +} + +static void +vl_api_disconnect_session_t_handler (vl_api_disconnect_session_t * mp) +{ + uri_udp_test_main_t *utm = &uri_udp_test_main; + session_t *session; + vl_api_disconnect_session_reply_t *rmp; + uword *p; + int rv = 0; + u64 key; + + key = (((u64) mp->session_thread_index) << 32) | (u64) mp->session_index; + + p = hash_get (utm->session_index_by_vpp_handles, key); + + if (p) + { + session = pool_elt_at_index (utm->sessions, p[0]); + hash_unset (utm->session_index_by_vpp_handles, key); + pool_put (utm->sessions, session); + } + else + { + clib_warning ("couldn't find session key %llx", key); + rv = -11; + } + + rmp = vl_msg_api_alloc (sizeof (*rmp)); + memset (rmp, 0, sizeof (*rmp)); + rmp->_vl_msg_id = ntohs (VL_API_DISCONNECT_SESSION_REPLY); + rmp->retval = rv; + rmp->session_index = mp->session_index; + rmp->session_thread_index = mp->session_thread_index; + vl_msg_api_send_shmem (utm->vl_input_queue, (u8 *) & rmp); +} + +#define foreach_uri_msg \ +_(BIND_URI_REPLY, bind_uri_reply) \ +_(UNBIND_URI_REPLY, unbind_uri_reply) \ +_(ACCEPT_SESSION, accept_session) \ +_(DISCONNECT_SESSION, disconnect_session) + +void +uri_api_hookup (uri_udp_test_main_t * utm) +{ +#define _(N,n) \ + vl_msg_api_set_handlers(VL_API_##N, #n, \ + vl_api_##n##_t_handler, \ + vl_noop_handler, \ + vl_api_##n##_t_endian, \ + vl_api_##n##_t_print, \ + sizeof(vl_api_##n##_t), 1); + foreach_uri_msg; +#undef _ + +} + + +int +connect_to_vpp (char *name) +{ + uri_udp_test_main_t *utm = &uri_udp_test_main; + api_main_t *am = &api_main; + + if (vl_client_connect_to_vlib ("/vpe-api", name, 32) < 0) + return -1; + + utm->vl_input_queue = am->shmem_hdr->vl_input_queue; + utm->my_client_index = am->my_client_index; + + return 0; +} + +void +vlib_cli_output (struct vlib_main_t *vm, char *fmt, ...) +{ + clib_warning ("BUG"); +} + +static void +init_error_string_table (uri_udp_test_main_t * utm) +{ + utm->error_string_by_error_number = hash_create (0, sizeof (uword)); + +#define _(n,v,s) hash_set (utm->error_string_by_error_number, -v, s); + foreach_vnet_api_error; +#undef _ + + hash_set (utm->error_string_by_error_number, 99, "Misc"); +} + +void +handle_fifo_event_server_rx (uri_udp_test_main_t * utm, + session_fifo_event_t * e) +{ + svm_fifo_t *rx_fifo, *tx_fifo; + int nbytes; + + session_fifo_event_t evt; + unix_shared_memory_queue_t *q; + int rv; + + rx_fifo = e->fifo; + tx_fifo = utm->sessions[rx_fifo->client_session_index].server_tx_fifo; + + do + { + nbytes = svm_fifo_dequeue_nowait (rx_fifo, 0, + vec_len (utm->rx_buf), utm->rx_buf); + } + while (nbytes <= 0); + do + { + rv = svm_fifo_enqueue_nowait (tx_fifo, 0, nbytes, utm->rx_buf); + } + while (rv == -2); + + /* Fabricate TX event, send to vpp */ + evt.fifo = tx_fifo; + evt.event_type = FIFO_EVENT_SERVER_TX; + /* $$$$ for event logging */ + evt.enqueue_length = nbytes; + evt.event_id = e->event_id; + q = utm->vpp_event_queue; + unix_shared_memory_queue_add (q, (u8 *) & evt, 0 /* do wait for mutex */ ); +} + +void +handle_event_queue (uri_udp_test_main_t * utm) +{ + session_fifo_event_t _e, *e = &_e;; + + while (1) + { + unix_shared_memory_queue_sub (utm->our_event_queue, (u8 *) e, + 0 /* nowait */ ); + switch (e->event_type) + { + case FIFO_EVENT_SERVER_RX: + handle_fifo_event_server_rx (utm, e); + break; + + case FIFO_EVENT_SERVER_EXIT: + return; + + default: + clib_warning ("unknown event type %d", e->event_type); + break; + } + if (PREDICT_FALSE (utm->time_to_stop == 1)) + break; + if (PREDICT_FALSE (utm->time_to_print_stats == 1)) + { + utm->time_to_print_stats = 0; + fformat (stdout, "%d connections\n", pool_elts (utm->sessions)); + } + } +} + +void +uri_udp_test (uri_udp_test_main_t * utm) +{ + vl_api_bind_uri_t *bmp; + vl_api_unbind_uri_t *ump; + + bmp = vl_msg_api_alloc (sizeof (*bmp)); + memset (bmp, 0, sizeof (*bmp)); + + bmp->_vl_msg_id = ntohs (VL_API_BIND_URI); + bmp->client_index = utm->my_client_index; + bmp->context = ntohl (0xfeedface); + bmp->segment_size = 2 << 30; + memcpy (bmp->uri, utm->uri, vec_len (utm->uri)); + vl_msg_api_send_shmem (utm->vl_input_queue, (u8 *) & bmp); + + if (wait_for_state_change (utm, STATE_READY)) + { + clib_warning ("timeout waiting for STATE_READY"); + return; + } + + handle_event_queue (utm); + + ump = vl_msg_api_alloc (sizeof (*ump)); + memset (ump, 0, sizeof (*ump)); + + ump->_vl_msg_id = ntohs (VL_API_UNBIND_URI); + ump->client_index = utm->my_client_index; + memcpy (ump->uri, utm->uri, vec_len (utm->uri)); + vl_msg_api_send_shmem (utm->vl_input_queue, (u8 *) & ump); + + if (wait_for_state_change (utm, STATE_START)) + { + clib_warning ("timeout waiting for STATE_START"); + return; + } + + fformat (stdout, "Test complete...\n"); +} + +int +main (int argc, char **argv) +{ + uri_udp_test_main_t *utm = &uri_udp_test_main; + unformat_input_t _argv, *a = &_argv; + u8 *chroot_prefix; + u8 *heap; + u8 *bind_name = (u8 *) "udp4:1234"; + mheap_t *h; + session_t *session; + int i; + + clib_mem_init (0, 256 << 20); + + heap = clib_mem_get_per_cpu_heap (); + h = mheap_header (heap); + + /* make the main heap thread-safe */ + h->flags |= MHEAP_FLAG_THREAD_SAFE; + + vec_validate (utm->rx_buf, 8192); + + utm->session_index_by_vpp_handles = hash_create (0, sizeof (uword)); + + clib_time_init (&utm->clib_time); + init_error_string_table (utm); + svm_fifo_segment_init (0x200000000ULL, 20); + unformat_init_command_line (a, argv); + + while (unformat_check_input (a) != UNFORMAT_END_OF_INPUT) + { + if (unformat (a, "chroot prefix %s", &chroot_prefix)) + { + vl_set_memory_root_path ((char *) chroot_prefix); + } + else if (unformat (a, "uri %s", &bind_name)) + ; + else + { + fformat (stderr, "%s: usage [master|slave]\n"); + exit (1); + } + } + + utm->uri = format (0, "%s%c", bind_name, 0); + + setup_signal_handlers (); + + uri_api_hookup (utm); + + if (connect_to_vpp ("uri_udp_test") < 0) + { + svm_region_exit (); + fformat (stderr, "Couldn't connect to vpe, exiting...\n"); + exit (1); + } + + /* $$$$ hack preallocation */ + for (i = 0; i < 200000; i++) + { + pool_get (utm->sessions, session); + memset (session, 0, sizeof (*session)); + } + for (i = 0; i < 200000; i++) + pool_put_index (utm->sessions, i); + + uri_udp_test (utm); + + vl_client_disconnect_from_vlib (); + exit (0); +} + +#undef vl_api_version +#define vl_api_version(n,v) static u32 vpe_api_version = v; +#include +#undef vl_api_version + +void +vl_client_add_api_signatures (vl_api_memclnt_create_t * mp) +{ + /* + * Send the main API signature in slot 0. This bit of code must + * match the checks in ../vpe/api/api.c: vl_msg_api_version_check(). + */ + mp->api_versions[0] = clib_host_to_net_u32 (vpe_api_version); +} + +/* + * fd.io coding-style-patch-verification: ON + * + * Local Variables: + * eval: (c-set-style "gnu") + * End: + */ diff --git a/src/uri/uri_udp_test2.c b/src/uri/uri_udp_test2.c new file mode 100644 index 00000000..ddfffaa6 --- /dev/null +++ b/src/uri/uri_udp_test2.c @@ -0,0 +1,954 @@ +/* + * Copyright (c) 2016 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include + +#include "../vnet/session/application_interface.h" + +#define vl_typedefs /* define message structures */ +#include +#undef vl_typedefs + +/* declare message handlers for each api */ + +#define vl_endianfun /* define message structures */ +#include +#undef vl_endianfun + +/* instantiate all the print functions we know about */ +#define vl_print(handle, ...) +#define vl_printfun +#include +#undef vl_printfun + +/* Satisfy external references when not linking with -lvlib */ +vlib_main_t vlib_global_main; +vlib_main_t **vlib_mains; + +typedef enum +{ + STATE_START, + STATE_READY, + STATE_DISCONNECTING, +} connection_state_t; + +typedef struct +{ + svm_fifo_t *server_rx_fifo; + svm_fifo_t *server_tx_fifo; +} session_t; + +typedef struct +{ + /* vpe input queue */ + unix_shared_memory_queue_t *vl_input_queue; + + /* API client handle */ + u32 my_client_index; + + /* The URI we're playing with */ + u8 *uri; + + /* Session pool */ + session_t *sessions; + + /* Hash table for disconnect processing */ + uword *session_index_by_vpp_handles; + + /* fifo segment */ + svm_fifo_segment_private_t *seg; + + /* intermediate rx buffer */ + u8 *rx_buf; + + /* URI for connect */ + u8 *connect_uri; + + int i_am_master; + + /* Our event queue */ + unix_shared_memory_queue_t *our_event_queue; + + /* $$$ single thread only for the moment */ + unix_shared_memory_queue_t *vpp_event_queue; + + /* $$$$ hack: cut-through session index */ + volatile u32 cut_through_session_index; + + /* unique segment name counter */ + u32 unique_segment_index; + + pid_t my_pid; + + /* pthread handle */ + pthread_t cut_through_thread_handle; + + /* For deadman timers */ + clib_time_t clib_time; + + /* State of the connection, shared between msg RX thread and main thread */ + volatile connection_state_t state; + + volatile int time_to_stop; + volatile int time_to_print_stats; + + u32 configured_segment_size; + + /* VNET_API_ERROR_FOO -> "Foo" hash table */ + uword *error_string_by_error_number; + + /* convenience */ + svm_fifo_segment_main_t *segment_main; + +} uri_udp_test_main_t; + +#if CLIB_DEBUG > 0 +#define NITER 10000 +#else +#define NITER 4000000 +#endif + +uri_udp_test_main_t uri_udp_test_main; + +static void +stop_signal (int signum) +{ + uri_udp_test_main_t *um = &uri_udp_test_main; + + um->time_to_stop = 1; +} + +static void +stats_signal (int signum) +{ + uri_udp_test_main_t *um = &uri_udp_test_main; + + um->time_to_print_stats = 1; +} + +static clib_error_t * +setup_signal_handlers (void) +{ + signal (SIGINT, stats_signal); + signal (SIGQUIT, stop_signal); + signal (SIGTERM, stop_signal); + + return 0; +} + +u8 * +format_api_error (u8 * s, va_list * args) +{ + uri_udp_test_main_t *utm = va_arg (*args, uri_udp_test_main_t *); + i32 error = va_arg (*args, u32); + uword *p; + + p = hash_get (utm->error_string_by_error_number, -error); + + if (p) + s = format (s, "%s", p[0]); + else + s = format (s, "%d", error); + return s; +} + +int +wait_for_state_change (uri_udp_test_main_t * utm, connection_state_t state) +{ +#if CLIB_DEBUG > 0 +#define TIMEOUT 600.0 +#else +#define TIMEOUT 600.0 +#endif + + f64 timeout = clib_time_now (&utm->clib_time) + TIMEOUT; + + while (clib_time_now (&utm->clib_time) < timeout) + { + if (utm->state == state) + return 0; + } + return -1; +} + +u64 server_bytes_received, server_bytes_sent; + +static void * +cut_through_thread_fn (void *arg) +{ + session_t *s; + svm_fifo_t *rx_fifo; + svm_fifo_t *tx_fifo; + u8 *my_copy_buffer = 0; + uri_udp_test_main_t *utm = &uri_udp_test_main; + i32 actual_transfer; + int rv; + u32 buffer_offset; + + while (utm->cut_through_session_index == ~0) + ; + + s = pool_elt_at_index (utm->sessions, utm->cut_through_session_index); + + rx_fifo = s->server_rx_fifo; + tx_fifo = s->server_tx_fifo; + + vec_validate (my_copy_buffer, 64 * 1024 - 1); + + while (true) + { + /* We read from the tx fifo and write to the rx fifo */ + do + { + actual_transfer = svm_fifo_dequeue_nowait (tx_fifo, 0, + vec_len (my_copy_buffer), + my_copy_buffer); + } + while (actual_transfer <= 0); + + server_bytes_received += actual_transfer; + + buffer_offset = 0; + while (actual_transfer > 0) + { + rv = svm_fifo_enqueue_nowait (rx_fifo, 0, actual_transfer, + my_copy_buffer + buffer_offset); + if (rv > 0) + { + actual_transfer -= rv; + buffer_offset += rv; + server_bytes_sent += rv; + } + + } + if (PREDICT_FALSE (utm->time_to_stop)) + break; + } + + pthread_exit (0); +} + +static void +uri_udp_slave_test (uri_udp_test_main_t * utm) +{ + vl_api_connect_uri_t *cmp; + int i; + u8 *test_data = 0; + u64 bytes_received = 0, bytes_sent = 0; + i32 bytes_to_read; + int rv; + int mypid = getpid (); + f64 before, after, delta, bytes_per_second; + session_t *session; + svm_fifo_t *rx_fifo, *tx_fifo; + int buffer_offset, bytes_to_send = 0; + + vec_validate (test_data, 64 * 1024 - 1); + for (i = 0; i < vec_len (test_data); i++) + test_data[i] = i & 0xff; + + cmp = vl_msg_api_alloc (sizeof (*cmp)); + memset (cmp, 0, sizeof (*cmp)); + + cmp->_vl_msg_id = ntohs (VL_API_CONNECT_URI); + cmp->client_index = utm->my_client_index; + cmp->context = ntohl (0xfeedface); + memcpy (cmp->uri, utm->connect_uri, vec_len (utm->connect_uri)); + vl_msg_api_send_shmem (utm->vl_input_queue, (u8 *) & cmp); + + if (wait_for_state_change (utm, STATE_READY)) + { + clib_warning ("timeout waiting for STATE_READY"); + return; + } + + session = pool_elt_at_index (utm->sessions, utm->cut_through_session_index); + rx_fifo = session->server_rx_fifo; + tx_fifo = session->server_tx_fifo; + + before = clib_time_now (&utm->clib_time); + + vec_validate (utm->rx_buf, vec_len (test_data) - 1); + + for (i = 0; i < NITER; i++) + { + bytes_to_send = vec_len (test_data); + buffer_offset = 0; + while (bytes_to_send > 0) + { + rv = svm_fifo_enqueue_nowait (tx_fifo, mypid, + bytes_to_send, + test_data + buffer_offset); + + if (rv > 0) + { + bytes_to_send -= rv; + buffer_offset += rv; + bytes_sent += rv; + } + } + + bytes_to_read = svm_fifo_max_dequeue (rx_fifo); + + bytes_to_read = vec_len (utm->rx_buf) > bytes_to_read ? + bytes_to_read : vec_len (utm->rx_buf); + + buffer_offset = 0; + while (bytes_to_read > 0) + { + rv = svm_fifo_dequeue_nowait (rx_fifo, mypid, + bytes_to_read, + utm->rx_buf + buffer_offset); + if (rv > 0) + { + bytes_to_read -= rv; + buffer_offset += rv; + bytes_received += rv; + } + } + } + while (bytes_received < bytes_sent) + { + rv = svm_fifo_dequeue_nowait (rx_fifo, mypid, + vec_len (utm->rx_buf), utm->rx_buf); + if (rv > 0) + { +#if CLIB_DEBUG > 0 + int j; + for (j = 0; j < rv; j++) + { + if (utm->rx_buf[j] != ((bytes_received + j) & 0xff)) + { + clib_warning ("error at byte %lld, 0x%x not 0x%x", + bytes_received + j, + utm->rx_buf[j], + ((bytes_received + j) & 0xff)); + } + } +#endif + bytes_received += (u64) rv; + } + } + + after = clib_time_now (&utm->clib_time); + delta = after - before; + bytes_per_second = 0.0; + + if (delta > 0.0) + bytes_per_second = (f64) bytes_received / delta; + + fformat (stdout, + "Done: %lld recv bytes in %.2f seconds, %.2f bytes/sec...\n\n", + bytes_received, delta, bytes_per_second); + fformat (stdout, + "Done: %lld sent bytes in %.2f seconds, %.2f bytes/sec...\n\n", + bytes_sent, delta, bytes_per_second); + fformat (stdout, + "client -> server -> client round trip: %.2f Gbit/sec \n\n", + (bytes_per_second * 8.0) / 1e9); +} + +static void +vl_api_bind_uri_reply_t_handler (vl_api_bind_uri_reply_t * mp) +{ + uri_udp_test_main_t *utm = &uri_udp_test_main; + svm_fifo_segment_create_args_t _a, *a = &_a; + int rv; + + if (mp->segment_name_length == 0) + { + clib_warning ("segment_name_length zero"); + return; + } + + a->segment_name = (char *) mp->segment_name; + a->segment_size = mp->segment_size; + + ASSERT (mp->server_event_queue_address); + + /* Attach to the segment vpp created */ + rv = svm_fifo_segment_attach (a); + if (rv) + { + clib_warning ("svm_fifo_segment_attach ('%s') failed", + mp->segment_name); + return; + } + + utm->our_event_queue = (unix_shared_memory_queue_t *) + mp->server_event_queue_address; + + utm->state = STATE_READY; +} + +static void +vl_api_map_another_segment_t_handler (vl_api_map_another_segment_t * mp) +{ + svm_fifo_segment_create_args_t _a, *a = &_a; + int rv; + + a->segment_name = (char *) mp->segment_name; + a->segment_size = mp->segment_size; + /* Attach to the segment vpp created */ + rv = svm_fifo_segment_attach (a); + if (rv) + { + clib_warning ("svm_fifo_segment_attach ('%s') failed", + mp->segment_name); + return; + } + clib_warning ("Mapped new segment '%s' size %d", mp->segment_name, + mp->segment_size); +} + +static void +vl_api_connect_uri_t_handler (vl_api_connect_uri_t * mp) +{ + u32 segment_index; + uri_udp_test_main_t *utm = &uri_udp_test_main; + svm_fifo_segment_main_t *sm = &svm_fifo_segment_main; + svm_fifo_segment_create_args_t _a, *a = &_a; + svm_fifo_segment_private_t *seg; + unix_shared_memory_queue_t *client_q; + vl_api_connect_uri_reply_t *rmp; + session_t *session; + int rv = 0; + + /* Create the segment */ + a->segment_name = (char *) format (0, "%d:segment%d%c", utm->my_pid, + utm->unique_segment_index++, 0); + a->segment_size = utm->configured_segment_size; + + rv = svm_fifo_segment_create (a); + if (rv) + { + clib_warning ("sm_fifo_segment_create ('%s') failed", a->segment_name); + rv = VNET_API_ERROR_URI_FIFO_CREATE_FAILED; + goto send_reply; + } + + vec_add2 (utm->seg, seg, 1); + + segment_index = vec_len (sm->segments) - 1; + + memcpy (seg, sm->segments + segment_index, sizeof (utm->seg[0])); + + pool_get (utm->sessions, session); + + /* + * By construction the master's idea of the rx fifo ends up in + * fsh->fifos[0], and the master's idea of the tx fifo ends up in + * fsh->fifos[1]. + */ + session->server_rx_fifo = svm_fifo_segment_alloc_fifo (utm->seg, + 128 * 1024); + ASSERT (session->server_rx_fifo); + + session->server_tx_fifo = svm_fifo_segment_alloc_fifo (utm->seg, + 128 * 1024); + ASSERT (session->server_tx_fifo); + + session->server_rx_fifo->server_session_index = session - utm->sessions; + session->server_tx_fifo->server_session_index = session - utm->sessions; + utm->cut_through_session_index = session - utm->sessions; + + rv = pthread_create (&utm->cut_through_thread_handle, + NULL /*attr */ , cut_through_thread_fn, 0); + if (rv) + { + clib_warning ("pthread_create returned %d", rv); + rv = VNET_API_ERROR_SYSCALL_ERROR_1; + } + +send_reply: + rmp = vl_msg_api_alloc (sizeof (*rmp)); + memset (rmp, 0, sizeof (*rmp)); + + rmp->_vl_msg_id = ntohs (VL_API_CONNECT_URI_REPLY); + rmp->context = mp->context; + rmp->retval = ntohl (rv); + rmp->segment_name_length = vec_len (a->segment_name); + memcpy (rmp->segment_name, a->segment_name, vec_len (a->segment_name)); + + vec_free (a->segment_name); + + client_q = (unix_shared_memory_queue_t *) mp->client_queue_address; + vl_msg_api_send_shmem (client_q, (u8 *) & rmp); +} + +static void +vl_api_unbind_uri_reply_t_handler (vl_api_unbind_uri_reply_t * mp) +{ + uri_udp_test_main_t *utm = &uri_udp_test_main; + + if (mp->retval != 0) + clib_warning ("returned %d", ntohl (mp->retval)); + + utm->state = STATE_START; +} + +static void +vl_api_accept_session_t_handler (vl_api_accept_session_t * mp) +{ + uri_udp_test_main_t *utm = &uri_udp_test_main; + vl_api_accept_session_reply_t *rmp; + svm_fifo_t *rx_fifo, *tx_fifo; + session_t *session; + static f64 start_time; + u64 key; + + if (start_time == 0.0) + start_time = clib_time_now (&utm->clib_time); + + utm->vpp_event_queue = (unix_shared_memory_queue_t *) + mp->vpp_event_queue_address; + + pool_get (utm->sessions, session); + + rx_fifo = (svm_fifo_t *) mp->server_rx_fifo; + rx_fifo->client_session_index = session - utm->sessions; + tx_fifo = (svm_fifo_t *) mp->server_tx_fifo; + tx_fifo->client_session_index = session - utm->sessions; + + session->server_rx_fifo = rx_fifo; + session->server_tx_fifo = tx_fifo; + + key = (((u64) mp->session_thread_index) << 32) | (u64) mp->session_index; + + hash_set (utm->session_index_by_vpp_handles, key, session - utm->sessions); + + utm->state = STATE_READY; + + if (pool_elts (utm->sessions) && (pool_elts (utm->sessions) % 20000) == 0) + { + f64 now = clib_time_now (&utm->clib_time); + fformat (stdout, "%d active sessions in %.2f seconds, %.2f/sec...\n", + pool_elts (utm->sessions), now - start_time, + (f64) pool_elts (utm->sessions) / (now - start_time)); + } + + rmp = vl_msg_api_alloc (sizeof (*rmp)); + memset (rmp, 0, sizeof (*rmp)); + rmp->_vl_msg_id = ntohs (VL_API_ACCEPT_SESSION_REPLY); + rmp->session_type = mp->session_type; + rmp->session_index = mp->session_index; + rmp->session_thread_index = mp->session_thread_index; + vl_msg_api_send_shmem (utm->vl_input_queue, (u8 *) & rmp); +} + +static void +vl_api_disconnect_session_t_handler (vl_api_disconnect_session_t * mp) +{ + uri_udp_test_main_t *utm = &uri_udp_test_main; + session_t *session; + vl_api_disconnect_session_reply_t *rmp; + uword *p; + int rv = 0; + u64 key; + + key = (((u64) mp->session_thread_index) << 32) | (u64) mp->session_index; + + p = hash_get (utm->session_index_by_vpp_handles, key); + + if (p) + { + session = pool_elt_at_index (utm->sessions, p[0]); + hash_unset (utm->session_index_by_vpp_handles, key); + pool_put (utm->sessions, session); + } + else + { + clib_warning ("couldn't find session key %llx", key); + rv = -11; + } + + rmp = vl_msg_api_alloc (sizeof (*rmp)); + memset (rmp, 0, sizeof (*rmp)); + rmp->_vl_msg_id = ntohs (VL_API_DISCONNECT_SESSION_REPLY); + rmp->retval = rv; + rmp->session_index = mp->session_index; + rmp->session_thread_index = mp->session_thread_index; + vl_msg_api_send_shmem (utm->vl_input_queue, (u8 *) & rmp); +} + +static void +vl_api_connect_uri_reply_t_handler (vl_api_connect_uri_reply_t * mp) +{ + svm_fifo_segment_main_t *sm = &svm_fifo_segment_main; + uri_udp_test_main_t *utm = &uri_udp_test_main; + svm_fifo_segment_create_args_t _a, *a = &_a; + ssvm_shared_header_t *sh; + svm_fifo_segment_private_t *seg; + svm_fifo_segment_header_t *fsh; + session_t *session; + u32 segment_index; + int rv; + + ASSERT (utm->i_am_master == 0); + + if (mp->segment_name_length == 0) + { + clib_warning ("segment_name_length zero"); + return; + } + + memset (a, 0, sizeof (*a)); + + a->segment_name = (char *) mp->segment_name; + + sleep (1); + + rv = svm_fifo_segment_attach (a); + if (rv) + { + clib_warning ("sm_fifo_segment_create ('%v') failed", mp->segment_name); + return; + } + + segment_index = vec_len (sm->segments) - 1; + + vec_add2 (utm->seg, seg, 1); + + memcpy (seg, sm->segments + segment_index, sizeof (*seg)); + sh = seg->ssvm.sh; + fsh = (svm_fifo_segment_header_t *) sh->opaque[0]; + + while (vec_len (fsh->fifos) < 2) + sleep (1); + + pool_get (utm->sessions, session); + utm->cut_through_session_index = session - utm->sessions; + + session->server_rx_fifo = (svm_fifo_t *) fsh->fifos[0]; + ASSERT (session->server_rx_fifo); + session->server_tx_fifo = (svm_fifo_t *) fsh->fifos[1]; + ASSERT (session->server_tx_fifo); + + /* security: could unlink /dev/shm/segment_name> here, maybe */ + + utm->state = STATE_READY; +} + +#define foreach_uri_msg \ +_(BIND_URI_REPLY, bind_uri_reply) \ +_(CONNECT_URI, connect_uri) \ +_(CONNECT_URI_REPLY, connect_uri_reply) \ +_(UNBIND_URI_REPLY, unbind_uri_reply) \ +_(ACCEPT_SESSION, accept_session) \ +_(DISCONNECT_SESSION, disconnect_session) \ +_(MAP_ANOTHER_SEGMENT, map_another_segment) + +void +uri_api_hookup (uri_udp_test_main_t * utm) +{ +#define _(N,n) \ + vl_msg_api_set_handlers(VL_API_##N, #n, \ + vl_api_##n##_t_handler, \ + vl_noop_handler, \ + vl_api_##n##_t_endian, \ + vl_api_##n##_t_print, \ + sizeof(vl_api_##n##_t), 1); + foreach_uri_msg; +#undef _ + +} + + +int +connect_to_vpp (char *name) +{ + uri_udp_test_main_t *utm = &uri_udp_test_main; + api_main_t *am = &api_main; + + if (vl_client_connect_to_vlib ("/vpe-api", name, 32) < 0) + return -1; + + utm->vl_input_queue = am->shmem_hdr->vl_input_queue; + utm->my_client_index = am->my_client_index; + + return 0; +} + +void +vlib_cli_output (struct vlib_main_t *vm, char *fmt, ...) +{ + clib_warning ("BUG"); +} + +static void +init_error_string_table (uri_udp_test_main_t * utm) +{ + utm->error_string_by_error_number = hash_create (0, sizeof (uword)); + +#define _(n,v,s) hash_set (utm->error_string_by_error_number, -v, s); + foreach_vnet_api_error; +#undef _ + + hash_set (utm->error_string_by_error_number, 99, "Misc"); +} + +void +handle_fifo_event_server_rx (uri_udp_test_main_t * utm, + session_fifo_event_t * e) +{ + svm_fifo_t *rx_fifo, *tx_fifo; + int nbytes; + + session_fifo_event_t evt; + unix_shared_memory_queue_t *q; + int rv; + + rx_fifo = e->fifo; + tx_fifo = utm->sessions[rx_fifo->client_session_index].server_tx_fifo; + + do + { + nbytes = svm_fifo_dequeue_nowait (rx_fifo, 0, + vec_len (utm->rx_buf), utm->rx_buf); + } + while (nbytes <= 0); + do + { + rv = svm_fifo_enqueue_nowait (tx_fifo, 0, nbytes, utm->rx_buf); + } + while (rv == -2); + + /* Fabricate TX event, send to vpp */ + evt.fifo = tx_fifo; + evt.event_type = FIFO_EVENT_SERVER_TX; + /* $$$$ for event logging */ + evt.enqueue_length = nbytes; + evt.event_id = e->event_id; + q = utm->vpp_event_queue; + unix_shared_memory_queue_add (q, (u8 *) & evt, 0 /* do wait for mutex */ ); +} + +void +handle_event_queue (uri_udp_test_main_t * utm) +{ + session_fifo_event_t _e, *e = &_e;; + + while (1) + { + unix_shared_memory_queue_sub (utm->our_event_queue, (u8 *) e, + 0 /* nowait */ ); + switch (e->event_type) + { + case FIFO_EVENT_SERVER_RX: + handle_fifo_event_server_rx (utm, e); + break; + + case FIFO_EVENT_SERVER_EXIT: + return; + + default: + clib_warning ("unknown event type %d", e->event_type); + break; + } + if (PREDICT_FALSE (utm->time_to_stop == 1)) + break; + if (PREDICT_FALSE (utm->time_to_print_stats == 1)) + { + utm->time_to_print_stats = 0; + fformat (stdout, "%d connections\n", pool_elts (utm->sessions)); + } + } +} + +void +uri_udp_test (uri_udp_test_main_t * utm) +{ + vl_api_bind_uri_t *bmp; + vl_api_unbind_uri_t *ump; + + bmp = vl_msg_api_alloc (sizeof (*bmp)); + memset (bmp, 0, sizeof (*bmp)); + + bmp->_vl_msg_id = ntohs (VL_API_BIND_URI); + bmp->client_index = utm->my_client_index; + bmp->context = ntohl (0xfeedface); + bmp->initial_segment_size = 256 << 20; /* size of initial segment */ + bmp->options[SESSION_OPTIONS_FLAGS] = + SESSION_OPTIONS_FLAGS_USE_FIFO | SESSION_OPTIONS_FLAGS_ADD_SEGMENT; + bmp->options[SESSION_OPTIONS_RX_FIFO_SIZE] = 16 << 10; + bmp->options[SESSION_OPTIONS_TX_FIFO_SIZE] = 16 << 10; + bmp->options[SESSION_OPTIONS_ADD_SEGMENT_SIZE] = 128 << 20; + memcpy (bmp->uri, utm->uri, vec_len (utm->uri)); + vl_msg_api_send_shmem (utm->vl_input_queue, (u8 *) & bmp); + + if (wait_for_state_change (utm, STATE_READY)) + { + clib_warning ("timeout waiting for STATE_READY"); + return; + } + + handle_event_queue (utm); + + ump = vl_msg_api_alloc (sizeof (*ump)); + memset (ump, 0, sizeof (*ump)); + + ump->_vl_msg_id = ntohs (VL_API_UNBIND_URI); + ump->client_index = utm->my_client_index; + memcpy (ump->uri, utm->uri, vec_len (utm->uri)); + vl_msg_api_send_shmem (utm->vl_input_queue, (u8 *) & ump); + + if (wait_for_state_change (utm, STATE_START)) + { + clib_warning ("timeout waiting for STATE_START"); + return; + } + + fformat (stdout, "Test complete...\n"); +} + +int +main (int argc, char **argv) +{ + uri_udp_test_main_t *utm = &uri_udp_test_main; + unformat_input_t _argv, *a = &_argv; + u8 *chroot_prefix; + u8 *heap; + u8 *bind_name = (u8 *) "udp://0.0.0.0/1234"; + u32 tmp; + mheap_t *h; + session_t *session; + int i; + int i_am_master = 1; + + clib_mem_init (0, 256 << 20); + + heap = clib_mem_get_per_cpu_heap (); + h = mheap_header (heap); + + /* make the main heap thread-safe */ + h->flags |= MHEAP_FLAG_THREAD_SAFE; + + vec_validate (utm->rx_buf, 8192); + + utm->session_index_by_vpp_handles = hash_create (0, sizeof (uword)); + + utm->my_pid = getpid (); + utm->configured_segment_size = 1 << 20; + + clib_time_init (&utm->clib_time); + init_error_string_table (utm); + svm_fifo_segment_init (0x200000000ULL, 20); + unformat_init_command_line (a, argv); + + while (unformat_check_input (a) != UNFORMAT_END_OF_INPUT) + { + if (unformat (a, "chroot prefix %s", &chroot_prefix)) + { + vl_set_memory_root_path ((char *) chroot_prefix); + } + else if (unformat (a, "uri %s", &bind_name)) + ; + else if (unformat (a, "segment-size %dM", &tmp)) + utm->configured_segment_size = tmp << 20; + else if (unformat (a, "segment-size %dG", &tmp)) + utm->configured_segment_size = tmp << 30; + else if (unformat (a, "master")) + i_am_master = 1; + else if (unformat (a, "slave")) + i_am_master = 0; + else + { + fformat (stderr, "%s: usage [master|slave]\n"); + exit (1); + } + } + + utm->cut_through_session_index = ~0; + utm->uri = format (0, "%s%c", bind_name, 0); + utm->i_am_master = i_am_master; + utm->segment_main = &svm_fifo_segment_main; + + utm->connect_uri = format (0, "udp://10.0.0.1/1234%c", 0); + + setup_signal_handlers (); + + uri_api_hookup (utm); + + if (connect_to_vpp (i_am_master ? "uri_udp_master" : "uri_udp_slave") < 0) + { + svm_region_exit (); + fformat (stderr, "Couldn't connect to vpe, exiting...\n"); + exit (1); + } + + if (i_am_master == 0) + { + uri_udp_slave_test (utm); + exit (0); + } + + /* $$$$ hack preallocation */ + for (i = 0; i < 200000; i++) + { + pool_get (utm->sessions, session); + memset (session, 0, sizeof (*session)); + } + for (i = 0; i < 200000; i++) + pool_put_index (utm->sessions, i); + + uri_udp_test (utm); + + vl_client_disconnect_from_vlib (); + exit (0); +} + +#undef vl_api_version +#define vl_api_version(n,v) static u32 vpe_api_version = v; +#include +#undef vl_api_version + +void +vl_client_add_api_signatures (vl_api_memclnt_create_t * mp) +{ + /* + * Send the main API signature in slot 0. This bit of code must + * match the checks in ../vpe/api/api.c: vl_msg_api_version_check(). + */ + mp->api_versions[0] = clib_host_to_net_u32 (vpe_api_version); +} + +u32 +vl (void *p) +{ + return vec_len (p); +} + +/* + * fd.io coding-style-patch-verification: ON + * + * Local Variables: + * eval: (c-set-style "gnu") + * End: + */ diff --git a/src/uri/uritest.c b/src/uri/uritest.c new file mode 100644 index 00000000..edcdb3ad --- /dev/null +++ b/src/uri/uritest.c @@ -0,0 +1,484 @@ +/* + * Copyright (c) 2016 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include + +#define vl_typedefs /* define message structures */ +#include +#undef vl_typedefs + +/* declare message handlers for each api */ + +#define vl_endianfun /* define message structures */ +#include +#undef vl_endianfun + +/* instantiate all the print functions we know about */ +#define vl_print(handle, ...) +#define vl_printfun +#include +#undef vl_printfun + +typedef enum +{ + STATE_START, + STATE_READY, + STATE_DISCONNECTING, +} connection_state_t; + +typedef struct +{ + /* vpe input queue */ + unix_shared_memory_queue_t *vl_input_queue; + + /* API client handle */ + u32 my_client_index; + + /* role */ + int i_am_master; + + /* The URI we're playing with */ + u8 *uri; + + /* fifo segment */ + svm_fifo_segment_private_t *seg; + + svm_fifo_t *rx_fifo; + svm_fifo_t *tx_fifo; + + /* For deadman timers */ + clib_time_t clib_time; + + /* State of the connection, shared between msg RX thread and main thread */ + volatile connection_state_t state; + + /* VNET_API_ERROR_FOO -> "Foo" hash table */ + uword *error_string_by_error_number; +} uritest_main_t; + +#if CLIB_DEBUG > 0 +#define NITER 1000 +#else +#define NITER 1000000 +#endif + +uritest_main_t uritest_main; + +u8 * +format_api_error (u8 * s, va_list * args) +{ + uritest_main_t *utm = va_arg (*args, uritest_main_t *); + i32 error = va_arg (*args, u32); + uword *p; + + p = hash_get (utm->error_string_by_error_number, -error); + + if (p) + s = format (s, "%s", p[0]); + else + s = format (s, "%d", error); + return s; +} + +int +wait_for_state_change (uritest_main_t * utm, connection_state_t state) +{ + f64 timeout = clib_time_now (&utm->clib_time) + 1.0; + + while (clib_time_now (&utm->clib_time) < timeout) + { + if (utm->state == state) + return 0; + } + return -1; +} + +static void +vl_api_bind_uri_reply_t_handler (vl_api_bind_uri_reply_t * mp) +{ + uritest_main_t *utm = &uritest_main; + svm_fifo_segment_create_args_t _a, *a = &_a; + int rv; + + ASSERT (utm->i_am_master); + + if (mp->segment_name_length == 0) + { + clib_warning ("segment_name_length zero"); + return; + } + + a->segment_name = (char *) mp->segment_name; + a->segment_size = mp->segment_size; + + /* Create the segment */ + rv = svm_fifo_segment_create (a); + if (rv) + { + clib_warning ("sm_fifo_segment_create ('%s') failed", mp->segment_name); + return; + } + + vec_validate (utm->seg, 0); + + memcpy (utm->seg, a->rv, sizeof (*utm->seg)); + + /* + * By construction the master's idea of the rx fifo ends up in + * fsh->fifos[0], and the master's idea of the tx fifo ends up in + * fsh->fifos[1]. + */ + utm->rx_fifo = svm_fifo_segment_alloc_fifo (utm->seg, 10240); + ASSERT (utm->rx_fifo); + + utm->tx_fifo = svm_fifo_segment_alloc_fifo (utm->seg, 10240); + ASSERT (utm->tx_fifo); + + utm->state = STATE_READY; +} + +static void +vl_api_connect_uri_reply_t_handler (vl_api_connect_uri_reply_t * mp) +{ + uritest_main_t *utm = &uritest_main; + svm_fifo_segment_create_args_t _a, *a = &_a; + ssvm_shared_header_t *sh; + svm_fifo_segment_header_t *fsh; + int rv; + + ASSERT (utm->i_am_master == 0); + + if (mp->segment_name_length == 0) + { + clib_warning ("segment_name_length zero"); + return; + } + + memset (a, 0, sizeof (*a)); + + a->segment_name = (char *) mp->segment_name; + + rv = svm_fifo_segment_attach (a); + if (rv) + { + clib_warning ("sm_fifo_segment_create ('%s') failed", mp->segment_name); + return; + } + + vec_validate (utm->seg, 0); + + memcpy (utm->seg, a->rv, sizeof (*utm->seg)); + sh = utm->seg->ssvm.sh; + fsh = (svm_fifo_segment_header_t *) sh->opaque[0]; + + while (vec_len (fsh->fifos) < 2) + sleep (1); + + utm->rx_fifo = (svm_fifo_t *) fsh->fifos[1]; + ASSERT (utm->rx_fifo); + utm->tx_fifo = (svm_fifo_t *) fsh->fifos[0]; + ASSERT (utm->tx_fifo); + + /* security: could unlink /dev/shm/segment_name> here, maybe */ + + utm->state = STATE_READY; +} + +static void +vl_api_unbind_uri_reply_t_handler (vl_api_unbind_uri_reply_t * mp) +{ + uritest_main_t *utm = &uritest_main; + + if (mp->retval != 0) + clib_warning ("returned %d", ntohl (mp->retval)); + + utm->state = STATE_START; +} + +#define foreach_uri_msg \ +_(BIND_URI_REPLY, bind_uri_reply) \ +_(CONNECT_URI_REPLY, connect_uri_reply) \ +_(UNBIND_URI_REPLY, unbind_uri_reply) + +void +uri_api_hookup (uritest_main_t * utm) +{ +#define _(N,n) \ + vl_msg_api_set_handlers(VL_API_##N, #n, \ + vl_api_##n##_t_handler, \ + vl_noop_handler, \ + vl_api_##n##_t_endian, \ + vl_api_##n##_t_print, \ + sizeof(vl_api_##n##_t), 1); + foreach_uri_msg; +#undef _ + +} + + +int +connect_to_vpp (char *name) +{ + uritest_main_t *utm = &uritest_main; + api_main_t *am = &api_main; + + if (vl_client_connect_to_vlib ("/vpe-api", name, 32) < 0) + return -1; + + utm->vl_input_queue = am->shmem_hdr->vl_input_queue; + utm->my_client_index = am->my_client_index; + + return 0; +} + +void +vlib_cli_output (struct vlib_main_t *vm, char *fmt, ...) +{ + clib_warning ("BUG"); +} + +static void +init_error_string_table (uritest_main_t * utm) +{ + utm->error_string_by_error_number = hash_create (0, sizeof (uword)); + +#define _(n,v,s) hash_set (utm->error_string_by_error_number, -v, s); + foreach_vnet_api_error; +#undef _ + + hash_set (utm->error_string_by_error_number, 99, "Misc"); +} + +void +uritest_master (uritest_main_t * utm) +{ + vl_api_bind_uri_t *bmp; + vl_api_unbind_uri_t *ump; + int i; + u8 *test_data = 0; + u8 *reply = 0; + u32 reply_len; + int mypid = getpid (); + + for (i = 0; i < 2048; i++) + vec_add1 (test_data, 'a' + (i % 32)); + + bmp = vl_msg_api_alloc (sizeof (*bmp)); + memset (bmp, 0, sizeof (*bmp)); + + bmp->_vl_msg_id = ntohs (VL_API_BIND_URI); + bmp->client_index = utm->my_client_index; + bmp->context = ntohl (0xfeedface); + bmp->segment_size = 256 << 10; + memcpy (bmp->uri, utm->uri, vec_len (utm->uri)); + vl_msg_api_send_shmem (utm->vl_input_queue, (u8 *) & bmp); + + if (wait_for_state_change (utm, STATE_READY)) + { + clib_warning ("timeout waiting for STATE_READY"); + return; + } + + for (i = 0; i < NITER; i++) + svm_fifo_enqueue (utm->tx_fifo, mypid, vec_len (test_data), test_data); + + vec_validate (reply, 0); + + reply_len = svm_fifo_dequeue (utm->rx_fifo, mypid, vec_len (reply), reply); + + if (reply_len != 1) + clib_warning ("reply length %d", reply_len); + + if (reply[0] == 1) + fformat (stdout, "Test OK..."); + + ump = vl_msg_api_alloc (sizeof (*ump)); + memset (ump, 0, sizeof (*ump)); + + ump->_vl_msg_id = ntohs (VL_API_UNBIND_URI); + ump->client_index = utm->my_client_index; + memcpy (ump->uri, utm->uri, vec_len (utm->uri)); + vl_msg_api_send_shmem (utm->vl_input_queue, (u8 *) & ump); + + if (wait_for_state_change (utm, STATE_START)) + { + clib_warning ("timeout waiting for STATE_READY"); + return; + } + + fformat (stdout, "Master done...\n"); +} + +void +uritest_slave (uritest_main_t * utm) +{ + vl_api_connect_uri_t *cmp; + int i, j; + u8 *test_data = 0; + u8 *reply = 0; + u32 bytes_received = 0; + u32 actual_bytes; + int mypid = getpid (); + u8 ok; + f64 before, after, delta, bytes_per_second; + + vec_validate (test_data, 4095); + + cmp = vl_msg_api_alloc (sizeof (*cmp)); + memset (cmp, 0, sizeof (*cmp)); + + cmp->_vl_msg_id = ntohs (VL_API_CONNECT_URI); + cmp->client_index = utm->my_client_index; + cmp->context = ntohl (0xfeedface); + memcpy (cmp->uri, utm->uri, vec_len (utm->uri)); + vl_msg_api_send_shmem (utm->vl_input_queue, (u8 *) & cmp); + + if (wait_for_state_change (utm, STATE_READY)) + { + clib_warning ("timeout waiting for STATE_READY"); + return; + } + + ok = 1; + before = clib_time_now (&utm->clib_time); + for (i = 0; i < NITER; i++) + { + actual_bytes = svm_fifo_dequeue (utm->rx_fifo, mypid, + vec_len (test_data), test_data); + j = 0; + while (j < actual_bytes) + { + if (test_data[j] != ('a' + (bytes_received % 32))) + ok = 0; + bytes_received++; + j++; + } + if (bytes_received == NITER * 2048) + break; + } + + vec_add1 (reply, ok); + + svm_fifo_enqueue (utm->tx_fifo, mypid, vec_len (reply), reply); + after = clib_time_now (&utm->clib_time); + delta = after - before; + bytes_per_second = 0.0; + + if (delta > 0.0) + bytes_per_second = (f64) bytes_received / delta; + + fformat (stdout, + "Slave done, %d bytes in %.2f seconds, %.2f bytes/sec...\n", + bytes_received, delta, bytes_per_second); +} + +int +main (int argc, char **argv) +{ + uritest_main_t *utm = &uritest_main; + unformat_input_t _argv, *a = &_argv; + u8 *chroot_prefix; + u8 *heap; + char *bind_name = "fifo:uritest"; + mheap_t *h; + int i_am_master = 0; + + clib_mem_init (0, 128 << 20); + + heap = clib_mem_get_per_cpu_heap (); + h = mheap_header (heap); + + /* make the main heap thread-safe */ + h->flags |= MHEAP_FLAG_THREAD_SAFE; + + clib_time_init (&utm->clib_time); + init_error_string_table (utm); + svm_fifo_segment_init (0x200000000ULL, 20); + unformat_init_command_line (a, argv); + + utm->uri = format (0, "%s%c", bind_name, 0); + + while (unformat_check_input (a) != UNFORMAT_END_OF_INPUT) + { + if (unformat (a, "master")) + i_am_master = 1; + else if (unformat (a, "slave")) + i_am_master = 0; + else if (unformat (a, "chroot prefix %s", &chroot_prefix)) + { + vl_set_memory_root_path ((char *) chroot_prefix); + } + else + { + fformat (stderr, "%s: usage [master|slave]\n"); + exit (1); + } + } + + uri_api_hookup (utm); + + if (connect_to_vpp (i_am_master ? "uritest_master" : "uritest_slave") < 0) + { + svm_region_exit (); + fformat (stderr, "Couldn't connect to vpe, exiting...\n"); + exit (1); + } + + utm->i_am_master = i_am_master; + + if (i_am_master) + uritest_master (utm); + else + uritest_slave (utm); + + vl_client_disconnect_from_vlib (); + exit (0); +} + +#undef vl_api_version +#define vl_api_version(n,v) static u32 vpe_api_version = v; +#include +#undef vl_api_version + +void +vl_client_add_api_signatures (vl_api_memclnt_create_t * mp) +{ + /* + * Send the main API signature in slot 0. This bit of code must + * match the checks in ../vpe/api/api.c: vl_msg_api_version_check(). + */ + mp->api_versions[0] = clib_host_to_net_u32 (vpe_api_version); +} + +/* + * fd.io coding-style-patch-verification: ON + * + * Local Variables: + * eval: (c-set-style "gnu") + * End: + */ diff --git a/src/vlib/buffer.c b/src/vlib/buffer.c index 4f5eb09d..9f26bec7 100644 --- a/src/vlib/buffer.c +++ b/src/vlib/buffer.c @@ -360,7 +360,7 @@ vlib_buffer_create_free_list_helper (vlib_main_t * vm, memset (f, 0, sizeof (f[0])); f->index = f - bm->buffer_free_list_pool; f->n_data_bytes = vlib_buffer_round_size (n_data_bytes); - f->min_n_buffers_each_physmem_alloc = 16; + f->min_n_buffers_each_physmem_alloc = VLIB_FRAME_SIZE; f->name = clib_mem_is_heap_object (name) ? name : format (0, "%s", name); /* Setup free buffer template. */ diff --git a/src/vlib/buffer.h b/src/vlib/buffer.h index 1f723f3b..69c8c7cc 100644 --- a/src/vlib/buffer.h +++ b/src/vlib/buffer.h @@ -240,6 +240,74 @@ vlib_get_buffer_opaque2 (vlib_buffer_t * b) return (void *) b->opaque2; } +/** \brief Get pointer to the end of buffer's data + * @param b pointer to the buffer + * @return pointer to tail of packet's data + */ +always_inline u8 * +vlib_buffer_get_tail (vlib_buffer_t * b) +{ + return b->data + b->current_data + b->current_length; +} + +/** \brief Append uninitialized data to buffer + * @param b pointer to the buffer + * @param size number of uninitialized bytes + * @return pointer to beginning of uninitialized data + */ +always_inline void * +vlib_buffer_put_uninit (vlib_buffer_t * b, u8 size) +{ + void *p = vlib_buffer_get_tail (b); + /* XXX make sure there's enough space */ + b->current_length += size; + return p; +} + +/** \brief Prepend uninitialized data to buffer + * @param b pointer to the buffer + * @param size number of uninitialized bytes + * @return pointer to beginning of uninitialized data + */ +always_inline void * +vlib_buffer_push_uninit (vlib_buffer_t * b, u8 size) +{ + ASSERT (b->current_data + VLIB_BUFFER_PRE_DATA_SIZE >= size); + b->current_data -= size; + b->current_length += size; + + return vlib_buffer_get_current (b); +} + +/** \brief Make head room, typically for packet headers + * @param b pointer to the buffer + * @param size number of head room bytes + * @return pointer to start of buffer (current data) + */ +always_inline void * +vlib_buffer_make_headroom (vlib_buffer_t * b, u8 size) +{ + ASSERT (b->current_data + VLIB_BUFFER_PRE_DATA_SIZE >= size); + b->current_data += size; + return vlib_buffer_get_current (b); +} + +/** \brief Retrieve bytes from buffer head + * @param b pointer to the buffer + * @param size number of bytes to pull + * @return pointer to start of buffer (current data) + */ +always_inline void * +vlib_buffer_pull (vlib_buffer_t * b, u8 size) +{ + if (b->current_length + VLIB_BUFFER_PRE_DATA_SIZE < size) + return 0; + + void *data = vlib_buffer_get_current (b); + vlib_buffer_advance (b, size); + return data; +} + /* Forward declaration. */ struct vlib_main_t; diff --git a/src/vlibmemory/unix_shared_memory_queue.c b/src/vlibmemory/unix_shared_memory_queue.c index 25d28910..e86edec3 100644 --- a/src/vlibmemory/unix_shared_memory_queue.c +++ b/src/vlibmemory/unix_shared_memory_queue.c @@ -33,18 +33,13 @@ * nels = number of elements on the queue * elsize = element size, presumably 4 and cacheline-size will * be popular choices. - * coid = consumer coid, from ChannelCreate * pid = consumer pid - * pulse_code = pulse code consumer expects - * pulse_value = pulse value consumer expects - * consumer_prio = consumer's priority, so pulses won't change - * the consumer's priority. * * The idea is to call this function in the queue consumer, * and e-mail the queue pointer to the producer(s). * - * The spp process / main thread allocates one of these - * at startup; its main input queue. The spp main input queue + * The vpp process / main thread allocates one of these + * at startup; its main input queue. The vpp main input queue * has a pointer to it in the shared memory segment header. * * You probably want to be on an svm data heap before calling this @@ -70,7 +65,7 @@ unix_shared_memory_queue_init (int nels, q->signal_when_queue_non_empty = signal_when_queue_non_empty; memset (&attr, 0, sizeof (attr)); - memset (&cattr, 0, sizeof (attr)); + memset (&cattr, 0, sizeof (cattr)); if (pthread_mutexattr_init (&attr)) clib_unix_warning ("mutexattr_init"); @@ -277,6 +272,7 @@ unix_shared_memory_queue_sub (unix_shared_memory_queue_t * q, clib_memcpy (elem, headp, q->elsize); q->head++; + /* $$$$ JFC shouldn't this be == 0? */ if (q->cursize == q->maxsize) need_broadcast = 1; diff --git a/src/vlibmemory/unix_shared_memory_queue.h b/src/vlibmemory/unix_shared_memory_queue.h index f758f17c..13800065 100644 --- a/src/vlibmemory/unix_shared_memory_queue.h +++ b/src/vlibmemory/unix_shared_memory_queue.h @@ -29,7 +29,7 @@ typedef struct _unix_shared_memory_queue pthread_cond_t condvar; /* 8 bytes */ int head; int tail; - int cursize; + volatile int cursize; int maxsize; int elsize; int consumer_pid; diff --git a/src/vnet.am b/src/vnet.am index 64484e18..923f61d8 100644 --- a/src/vnet.am +++ b/src/vnet.am @@ -324,11 +324,7 @@ libvnet_la_SOURCES += \ vnet/ip/ip_input_acl.c \ vnet/ip/lookup.c \ vnet/ip/ping.c \ - vnet/ip/punt.c \ - vnet/ip/udp_format.c \ - vnet/ip/udp_init.c \ - vnet/ip/udp_local.c \ - vnet/ip/udp_pg.c + vnet/ip/punt.c nobase_include_HEADERS += \ vnet/ip/format.h \ @@ -354,11 +350,7 @@ nobase_include_HEADERS += \ vnet/ip/ports.def \ vnet/ip/protocols.def \ vnet/ip/punt_error.def \ - vnet/ip/punt.h \ - vnet/ip/tcp_packet.h \ - vnet/ip/udp_error.def \ - vnet/ip/udp.h \ - vnet/ip/udp_packet.h + vnet/ip/punt.h API_FILES += vnet/ip/ip.api @@ -473,6 +465,38 @@ test_map_LDADD = libvnet.la libvppinfra.la libvlib.la \ test_map_LDFLAGS = -static endif +######################################## +# Layer 4 protocol: tcp +######################################## +libvnet_la_SOURCES += \ + vnet/tcp/tcp_format.c \ + vnet/tcp/tcp_pg.c \ + vnet/tcp/tcp_syn_filter4.c \ + vnet/tcp/tcp_output.c \ + vnet/tcp/tcp_input.c \ + vnet/tcp/tcp_newreno.c \ + vnet/tcp/tcp.c + +nobase_include_HEADERS += \ + vnet/tcp/tcp_packet.h \ + vnet/tcp/tcp_timer.h \ + vnet/tcp/tcp.h + +######################################## +# Layer 4 protocol: udp +######################################## +libvnet_la_SOURCES += \ + vnet/udp/udp.c \ + vnet/udp/udp_input.c \ + vnet/udp/builtin_server.c \ + vnet/udp/udp_format.c \ + vnet/udp/udp_local.c \ + vnet/udp/udp_pg.c + +nobase_include_HEADERS += \ + vnet/udp/udp_error.def \ + vnet/udp/udp.h \ + vnet/udp/udp_packet.h ######################################## # Tunnel protocol: gre @@ -833,6 +857,28 @@ libvnet_la_SOURCES += \ nobase_include_HEADERS += \ vnet/devices/ssvm/ssvm_eth.h +######################################## +# session managmeent +######################################## + +libvnet_la_SOURCES += \ + vnet/session/session.c \ + vnet/session/node.c \ + vnet/session/transport.c \ + vnet/session/application.c \ + vnet/session/session_cli.c \ + vnet/session/hashes.c \ + vnet/session/application_interface.c \ + vnet/session/session_api.c + +nobase_include_HEADERS += \ + vnet/session/session.h \ + vnet/session/application.h \ + vnet/session/transport.h \ + vnet/session/application_interface.h + +API_FILES += vnet/session/session.api + ######################################## # Linux packet interface ######################################## diff --git a/src/vnet/api_errno.h b/src/vnet/api_errno.h index 8680ef7c..861a5767 100644 --- a/src/vnet/api_errno.h +++ b/src/vnet/api_errno.h @@ -91,14 +91,19 @@ _(INVALID_ADDRESS_FAMILY, -97, "Invalid address family") \ _(INVALID_SUB_SW_IF_INDEX, -98, "Invalid sub-interface sw_if_index") \ _(TABLE_TOO_BIG, -99, "Table too big") \ _(CANNOT_ENABLE_DISABLE_FEATURE, -100, "Cannot enable/disable feature") \ -_(BFD_EEXIST, -101, "Duplicate BFD object") \ -_(BFD_ENOENT, -102, "No such BFD object") \ -_(BFD_EINUSE, -103, "BFD object in use") \ -_(BFD_NOTSUPP, -104, "BFD feature not supported") \ -_(LISP_RLOC_LOCAL, -105, "RLOC address is local") \ -_(BFD_EAGAIN, -106, "BFD object cannot be manipulated at this time") \ -_(INVALID_GPE_MODE, -107, "Invalid GPE mode") \ -_(LISP_GPE_ENTRIES_PRESENT, -108, "LISP GPE entries are present") +_(BFD_EEXIST, -101, "Duplicate BFD object") \ +_(BFD_ENOENT, -102, "No such BFD object") \ +_(BFD_EINUSE, -103, "BFD object in use") \ +_(BFD_NOTSUPP, -104, "BFD feature not supported") \ +_(ADDRESS_IN_USE, -105, "Address in use") \ +_(ADDRESS_NOT_IN_USE, -106, "Address not in use") \ +_(QUEUE_FULL, -107, "Queue full") \ +_(UNKNOWN_URI_TYPE, -108, "Unknown URI type") \ +_(URI_FIFO_CREATE_FAILED, -109, "URI FIFO segment create failed") \ +_(LISP_RLOC_LOCAL, -110, "RLOC address is local") \ +_(BFD_EAGAIN, -111, "BFD object cannot be manipulated at this time") \ +_(INVALID_GPE_MODE, -112, "Invalid GPE mode") \ +_(LISP_GPE_ENTRIES_PRESENT, -113, "LISP GPE entries are present") typedef enum { diff --git a/src/vnet/bfd/bfd_udp.c b/src/vnet/bfd/bfd_udp.c index 146faad6..cf05089b 100644 --- a/src/vnet/bfd/bfd_udp.c +++ b/src/vnet/bfd/bfd_udp.c @@ -18,12 +18,12 @@ #include #include #include -#include +#include +#include #include #include #include #include -#include #include #include #include diff --git a/src/vnet/buffer.h b/src/vnet/buffer.h index f1cc6371..3de01f2a 100644 --- a/src/vnet/buffer.h +++ b/src/vnet/buffer.h @@ -277,6 +277,16 @@ typedef struct u16 buffer_advance; } device_input_feat; + /* TCP */ + struct + { + u32 connection_index; + u32 seq_number; + u32 seq_end; + u32 ack_number; + u8 flags; + } tcp; + u32 unused[6]; }; } vnet_buffer_opaque_t; diff --git a/src/vnet/classify/vnet_classify.c b/src/vnet/classify/vnet_classify.c index 6093e2ac..b651a1f1 100644 --- a/src/vnet/classify/vnet_classify.c +++ b/src/vnet/classify/vnet_classify.c @@ -695,8 +695,8 @@ int vnet_classify_add_del_table (vnet_classify_main_t * cm, } #define foreach_tcp_proto_field \ -_(src_port) \ -_(dst_port) +_(src) \ +_(dst) #define foreach_udp_proto_field \ _(src_port) \ diff --git a/src/vnet/dhcp/dhcp_proxy.h b/src/vnet/dhcp/dhcp_proxy.h index c0d79c41..4586d883 100644 --- a/src/vnet/dhcp/dhcp_proxy.h +++ b/src/vnet/dhcp/dhcp_proxy.h @@ -26,7 +26,7 @@ #include #include #include -#include +#include typedef enum { #define dhcp_proxy_error(n,s) DHCP_PROXY_ERROR_##n, diff --git a/src/vnet/flow/flow_report.h b/src/vnet/flow/flow_report.h index 4e764377..e8ed3818 100644 --- a/src/vnet/flow/flow_report.h +++ b/src/vnet/flow/flow_report.h @@ -23,7 +23,7 @@ #include #include #include -#include +#include #include #include #include diff --git a/src/vnet/ip/ip.h b/src/vnet/ip/ip.h index 02a1a963..70b4ccd8 100644 --- a/src/vnet/ip/ip.h +++ b/src/vnet/ip/ip.h @@ -50,8 +50,8 @@ #include #include -#include -#include +#include +#include #include #include diff --git a/src/vnet/ip/ip4.h b/src/vnet/ip/ip4.h index b184fbae..4e075d0f 100644 --- a/src/vnet/ip/ip4.h +++ b/src/vnet/ip/ip4.h @@ -309,8 +309,8 @@ ip4_compute_flow_hash (const ip4_header_t * ip, b = (flow_hash_config & IP_FLOW_HASH_REVERSE_SRC_DST) ? t1 : t2; b ^= (flow_hash_config & IP_FLOW_HASH_PROTO) ? ip->protocol : 0; - t1 = is_tcp_udp ? tcp->ports.src : 0; - t2 = is_tcp_udp ? tcp->ports.dst : 0; + t1 = is_tcp_udp ? tcp->src : 0; + t2 = is_tcp_udp ? tcp->dst : 0; t1 = (flow_hash_config & IP_FLOW_HASH_SRC_PORT) ? t1 : 0; t2 = (flow_hash_config & IP_FLOW_HASH_DST_PORT) ? t2 : 0; @@ -334,6 +334,44 @@ u8 *format_ip4_forward_next_trace (u8 * s, va_list * args); u32 ip4_tcp_udp_validate_checksum (vlib_main_t * vm, vlib_buffer_t * p0); +#define IP_DF 0x4000 /* don't fragment */ + +/** + * Push IPv4 header to buffer + * + * This does not support fragmentation. + * + * @param vm - vlib_main + * @param b - buffer to write the header to + * @param src - source IP + * @param dst - destination IP + * @param prot - payload proto + * + * @return - pointer to start of IP header + */ +always_inline void * +vlib_buffer_push_ip4 (vlib_main_t * vm, vlib_buffer_t * b, + ip4_address_t * src, ip4_address_t * dst, int proto) +{ + ip4_header_t *ih; + + /* make some room */ + ih = vlib_buffer_push_uninit (b, sizeof (ip4_header_t)); + + ih->ip_version_and_header_length = 0x45; + ih->tos = 0; + ih->length = clib_host_to_net_u16 (vlib_buffer_length_in_chain (vm, b)); + + /* No fragments */ + ih->flags_and_fragment_offset = clib_host_to_net_u16 (IP_DF); + ih->ttl = 255; + ih->protocol = proto; + ih->src_address.as_u32 = src->as_u32; + ih->dst_address.as_u32 = dst->as_u32; + + ih->checksum = ip4_header_checksum (ih); + return ih; +} #endif /* included_ip_ip4_h */ /* diff --git a/src/vnet/ip/ip4_forward.c b/src/vnet/ip/ip4_forward.c index 8081b34b..66d91ab6 100644 --- a/src/vnet/ip/ip4_forward.c +++ b/src/vnet/ip/ip4_forward.c @@ -1478,8 +1478,18 @@ ip4_tcp_udp_validate_checksum (vlib_main_t * vm, vlib_buffer_t * p0) return p0->flags; } -static uword -ip4_local (vlib_main_t * vm, vlib_node_runtime_t * node, vlib_frame_t * frame) +/* *INDENT-OFF* */ +VNET_FEATURE_ARC_INIT (ip4_local) = +{ + .arc_name = "ip4-local", + .start_nodes = VNET_FEATURES ("ip4-local"), +}; +/* *INDENT-ON* */ + +static inline uword +ip4_local_inline (vlib_main_t * vm, + vlib_node_runtime_t * node, + vlib_frame_t * frame, int head_of_feature_arc) { ip4_main_t *im = &ip4_main; ip_lookup_main_t *lm = &im->lookup_main; @@ -1487,6 +1497,7 @@ ip4_local (vlib_main_t * vm, vlib_node_runtime_t * node, vlib_frame_t * frame) u32 *from, *to_next, n_left_from, n_left_to_next; vlib_node_runtime_t *error_node = vlib_node_get_runtime (vm, ip4_input_node.index); + u8 arc_index = vnet_feat_arc_ip4_local.feature_arc_index; from = vlib_frame_vector_args (frame); n_left_from = frame->n_vectors; @@ -1513,7 +1524,7 @@ ip4_local (vlib_main_t * vm, vlib_node_runtime_t * node, vlib_frame_t * frame) i32 len_diff0, len_diff1; u8 error0, is_udp0, is_tcp_udp0, good_tcp_udp0, proto0; u8 error1, is_udp1, is_tcp_udp1, good_tcp_udp1, proto1; - u8 enqueue_code; + u32 sw_if_index0, sw_if_index1; pi0 = to_next[0] = from[0]; pi1 = to_next[1] = from[1]; @@ -1522,6 +1533,8 @@ ip4_local (vlib_main_t * vm, vlib_node_runtime_t * node, vlib_frame_t * frame) to_next += 2; n_left_to_next -= 2; + next0 = next1 = IP_LOCAL_NEXT_DROP; + p0 = vlib_get_buffer (vm, pi0); p1 = vlib_get_buffer (vm, pi1); @@ -1531,14 +1544,18 @@ ip4_local (vlib_main_t * vm, vlib_node_runtime_t * node, vlib_frame_t * frame) vnet_buffer (p0)->ip.start_of_ip_header = p0->current_data; vnet_buffer (p1)->ip.start_of_ip_header = p1->current_data; - fib_index0 = vec_elt (im->fib_index_by_sw_if_index, - vnet_buffer (p0)->sw_if_index[VLIB_RX]); + sw_if_index0 = vnet_buffer (p0)->sw_if_index[VLIB_RX]; + sw_if_index1 = vnet_buffer (p1)->sw_if_index[VLIB_RX]; + + fib_index0 = vec_elt (im->fib_index_by_sw_if_index, sw_if_index0); + fib_index1 = vec_elt (im->fib_index_by_sw_if_index, sw_if_index1); + + fib_index0 = vec_elt (im->fib_index_by_sw_if_index, sw_if_index0); fib_index0 = (vnet_buffer (p0)->sw_if_index[VLIB_TX] == (u32) ~ 0) ? fib_index0 : vnet_buffer (p0)->sw_if_index[VLIB_TX]; - fib_index1 = vec_elt (im->fib_index_by_sw_if_index, - vnet_buffer (p1)->sw_if_index[VLIB_RX]); + fib_index1 = vec_elt (im->fib_index_by_sw_if_index, sw_if_index1); fib_index1 = (vnet_buffer (p1)->sw_if_index[VLIB_TX] == (u32) ~ 0) ? fib_index1 : vnet_buffer (p1)->sw_if_index[VLIB_TX]; @@ -1557,6 +1574,13 @@ ip4_local (vlib_main_t * vm, vlib_node_runtime_t * node, vlib_frame_t * frame) until support of IP frag reassembly is implemented */ proto0 = ip4_is_fragment (ip0) ? 0xfe : ip0->protocol; proto1 = ip4_is_fragment (ip1) ? 0xfe : ip1->protocol; + + if (head_of_feature_arc == 0) + { + error0 = error1 = IP4_ERROR_UNKNOWN_PROTOCOL; + goto skip_checks; + } + is_udp0 = proto0 == IP_PROTOCOL_UDP; is_udp1 = proto1 == IP_PROTOCOL_UDP; is_tcp_udp0 = is_udp0 || proto0 == IP_PROTOCOL_TCP; @@ -1686,6 +1710,7 @@ ip4_local (vlib_main_t * vm, vlib_node_runtime_t * node, vlib_frame_t * frame) next0 = lm->local_next_by_ip_protocol[proto0]; next1 = lm->local_next_by_ip_protocol[proto1]; + skip_checks: next0 = error0 != IP4_ERROR_UNKNOWN_PROTOCOL ? IP_LOCAL_NEXT_DROP : next0; next1 = @@ -1694,44 +1719,17 @@ ip4_local (vlib_main_t * vm, vlib_node_runtime_t * node, vlib_frame_t * frame) p0->error = error0 ? error_node->errors[error0] : 0; p1->error = error1 ? error_node->errors[error1] : 0; - enqueue_code = (next0 != next_index) + 2 * (next1 != next_index); - - if (PREDICT_FALSE (enqueue_code != 0)) + if (head_of_feature_arc) { - switch (enqueue_code) - { - case 1: - /* A B A */ - to_next[-2] = pi1; - to_next -= 1; - n_left_to_next += 1; - vlib_set_next_frame_buffer (vm, node, next0, pi0); - break; - - case 2: - /* A A B */ - to_next -= 1; - n_left_to_next += 1; - vlib_set_next_frame_buffer (vm, node, next1, pi1); - break; - - case 3: - /* A B B or A B C */ - to_next -= 2; - n_left_to_next += 2; - vlib_set_next_frame_buffer (vm, node, next0, pi0); - vlib_set_next_frame_buffer (vm, node, next1, pi1); - if (next0 == next1) - { - vlib_put_next_frame (vm, node, next_index, - n_left_to_next); - next_index = next1; - vlib_get_next_frame (vm, node, next_index, to_next, - n_left_to_next); - } - break; - } + if (PREDICT_TRUE (error0 == (u8) IP4_ERROR_UNKNOWN_PROTOCOL)) + vnet_feature_arc_start (arc_index, sw_if_index0, &next0, p0); + if (PREDICT_TRUE (error1 == (u8) IP4_ERROR_UNKNOWN_PROTOCOL)) + vnet_feature_arc_start (arc_index, sw_if_index1, &next1, p1); } + + vlib_validate_buffer_enqueue_x2 (vm, node, next_index, to_next, + n_left_to_next, pi0, pi1, + next0, next1); } while (n_left_from > 0 && n_left_to_next > 0) @@ -1746,6 +1744,7 @@ ip4_local (vlib_main_t * vm, vlib_node_runtime_t * node, vlib_frame_t * frame) u8 error0, is_udp0, is_tcp_udp0, good_tcp_udp0, proto0; load_balance_t *lb0; const dpo_id_t *dpo0; + u32 sw_if_index0; pi0 = to_next[0] = from[0]; from += 1; @@ -1753,14 +1752,18 @@ ip4_local (vlib_main_t * vm, vlib_node_runtime_t * node, vlib_frame_t * frame) to_next += 1; n_left_to_next -= 1; + next0 = IP_LOCAL_NEXT_DROP; + p0 = vlib_get_buffer (vm, pi0); ip0 = vlib_buffer_get_current (p0); vnet_buffer (p0)->ip.start_of_ip_header = p0->current_data; - fib_index0 = vec_elt (im->fib_index_by_sw_if_index, - vnet_buffer (p0)->sw_if_index[VLIB_RX]); + sw_if_index0 = vnet_buffer (p0)->sw_if_index[VLIB_RX]; + + fib_index0 = vec_elt (im->fib_index_by_sw_if_index, sw_if_index0); + fib_index0 = (vnet_buffer (p0)->sw_if_index[VLIB_TX] == (u32) ~ 0) ? fib_index0 : vnet_buffer (p0)->sw_if_index[VLIB_TX]; @@ -1775,6 +1778,13 @@ ip4_local (vlib_main_t * vm, vlib_node_runtime_t * node, vlib_frame_t * frame) /* Treat IP frag packets as "experimental" protocol for now until support of IP frag reassembly is implemented */ proto0 = ip4_is_fragment (ip0) ? 0xfe : ip0->protocol; + + if (head_of_feature_arc == 0) + { + error0 = IP4_ERROR_UNKNOWN_PROTOCOL; + goto skip_check; + } + is_udp0 = proto0 == IP_PROTOCOL_UDP; is_tcp_udp0 = is_udp0 || proto0 == IP_PROTOCOL_TCP; @@ -1847,6 +1857,8 @@ ip4_local (vlib_main_t * vm, vlib_node_runtime_t * node, vlib_frame_t * frame) ip0->dst_address.as_u32 != 0xFFFFFFFF) ? IP4_ERROR_SRC_LOOKUP_MISS : error0); + skip_check: + next0 = lm->local_next_by_ip_protocol[proto0]; next0 = @@ -1854,18 +1866,15 @@ ip4_local (vlib_main_t * vm, vlib_node_runtime_t * node, vlib_frame_t * frame) p0->error = error0 ? error_node->errors[error0] : 0; - if (PREDICT_FALSE (next0 != next_index)) + if (head_of_feature_arc) { - n_left_to_next += 1; - vlib_put_next_frame (vm, node, next_index, n_left_to_next); - - next_index = next0; - vlib_get_next_frame (vm, node, next_index, to_next, - n_left_to_next); - to_next[0] = pi0; - to_next += 1; - n_left_to_next -= 1; + if (PREDICT_TRUE (error0 == (u8) IP4_ERROR_UNKNOWN_PROTOCOL)) + vnet_feature_arc_start (arc_index, sw_if_index0, &next0, p0); } + + vlib_validate_buffer_enqueue_x1 (vm, node, next_index, to_next, + n_left_to_next, pi0, next0); + } vlib_put_next_frame (vm, node, next_index, n_left_to_next); @@ -1874,21 +1883,57 @@ ip4_local (vlib_main_t * vm, vlib_node_runtime_t * node, vlib_frame_t * frame) return frame->n_vectors; } +static uword +ip4_local (vlib_main_t * vm, vlib_node_runtime_t * node, vlib_frame_t * frame) +{ + return ip4_local_inline (vm, node, frame, 1 /* head of feature arc */ ); +} + +/* *INDENT-OFF* */ VLIB_REGISTER_NODE (ip4_local_node) = { - .function = ip4_local,.name = "ip4-local",.vector_size = - sizeof (u32),.format_trace = - format_ip4_forward_next_trace,.n_next_nodes = - IP_LOCAL_N_NEXT,.next_nodes = + .function = ip4_local, + .name = "ip4-local", + .vector_size = sizeof (u32), + .format_trace = format_ip4_forward_next_trace, + .n_next_nodes = IP_LOCAL_N_NEXT, + .next_nodes = { - [IP_LOCAL_NEXT_DROP] = "error-drop", - [IP_LOCAL_NEXT_PUNT] = "error-punt", - [IP_LOCAL_NEXT_UDP_LOOKUP] = "ip4-udp-lookup", - [IP_LOCAL_NEXT_ICMP] = "ip4-icmp-input",} -,}; + [IP_LOCAL_NEXT_DROP] = "error-drop", + [IP_LOCAL_NEXT_PUNT] = "error-punt", + [IP_LOCAL_NEXT_UDP_LOOKUP] = "ip4-udp-lookup", + [IP_LOCAL_NEXT_ICMP] = "ip4-icmp-input",}, +}; +/* *INDENT-ON* */ VLIB_NODE_FUNCTION_MULTIARCH (ip4_local_node, ip4_local); +static uword +ip4_local_end_of_arc (vlib_main_t * vm, + vlib_node_runtime_t * node, vlib_frame_t * frame) +{ + return ip4_local_inline (vm, node, frame, 0 /* head of feature arc */ ); +} + +/* *INDENT-OFF* */ +VLIB_REGISTER_NODE (ip4_local_end_of_arc_node,static) = { + .function = ip4_local_end_of_arc, + .name = "ip4-local-end-of-arc", + .vector_size = sizeof (u32), + + .format_trace = format_ip4_forward_next_trace, + .sibling_of = "ip4-local", +}; + +VLIB_NODE_FUNCTION_MULTIARCH (ip4_local_end_of_arc_node, ip4_local_end_of_arc) + +VNET_FEATURE_INIT (ip4_local_end_of_arc, static) = { + .arc_name = "ip4-local", + .node_name = "ip4-local-end-of-arc", + .runs_before = 0, /* not before any other features */ +}; +/* *INDENT-ON* */ + void ip4_register_protocol (u32 protocol, u32 node_index) { diff --git a/src/vnet/ip/ip4_packet.h b/src/vnet/ip/ip4_packet.h index 8da788b4..b2c1fcd4 100644 --- a/src/vnet/ip/ip4_packet.h +++ b/src/vnet/ip/ip4_packet.h @@ -41,7 +41,7 @@ #define included_ip4_packet_h #include /* for ip_csum_t */ -#include /* for tcp_header_t */ +#include /* for tcp_header_t */ #include /* for clib_net_to_host_u16 */ /* IP4 address which can be accessed either as 4 bytes @@ -342,10 +342,10 @@ ip4_tcp_reply_x1 (ip4_header_t * ip0, tcp_header_t * tcp0) ip0->src_address.data_u32 = dst0; ip0->dst_address.data_u32 = src0; - src0 = tcp0->ports.src; - dst0 = tcp0->ports.dst; - tcp0->ports.src = dst0; - tcp0->ports.dst = src0; + src0 = tcp0->src; + dst0 = tcp0->dst; + tcp0->src = dst0; + tcp0->dst = src0; } always_inline void @@ -363,14 +363,14 @@ ip4_tcp_reply_x2 (ip4_header_t * ip0, ip4_header_t * ip1, ip0->dst_address.data_u32 = src0; ip1->dst_address.data_u32 = src1; - src0 = tcp0->ports.src; - src1 = tcp1->ports.src; - dst0 = tcp0->ports.dst; - dst1 = tcp1->ports.dst; - tcp0->ports.src = dst0; - tcp1->ports.src = dst1; - tcp0->ports.dst = src0; - tcp1->ports.dst = src1; + src0 = tcp0->src; + src1 = tcp1->src; + dst0 = tcp0->dst; + dst1 = tcp1->dst; + tcp0->src = dst0; + tcp1->src = dst1; + tcp0->dst = src0; + tcp1->dst = src1; } #endif /* included_ip4_packet_h */ diff --git a/src/vnet/ip/ip6.h b/src/vnet/ip/ip6.h index 5456f0f2..2615fbfa 100644 --- a/src/vnet/ip/ip6.h +++ b/src/vnet/ip/ip6.h @@ -461,8 +461,8 @@ ip6_compute_flow_hash (const ip6_header_t * ip, b = (flow_hash_config & IP_FLOW_HASH_REVERSE_SRC_DST) ? t1 : t2; b ^= (flow_hash_config & IP_FLOW_HASH_PROTO) ? ip->protocol : 0; - t1 = is_tcp_udp ? tcp->ports.src : 0; - t2 = is_tcp_udp ? tcp->ports.dst : 0; + t1 = is_tcp_udp ? tcp->src : 0; + t2 = is_tcp_udp ? tcp->dst : 0; t1 = (flow_hash_config & IP_FLOW_HASH_SRC_PORT) ? t1 : 0; t2 = (flow_hash_config & IP_FLOW_HASH_DST_PORT) ? t2 : 0; @@ -497,6 +497,46 @@ int ip6_hbh_register_option (u8 option, int ip6_hbh_unregister_option (u8 option); void ip6_hbh_set_next_override (uword next); +/** + * Push IPv6 header to buffer + * + * @param vm - vlib_main + * @param b - buffer to write the header to + * @param src - source IP + * @param dst - destination IP + * @param prot - payload proto + * + * @return - pointer to start of IP header + */ +always_inline void * +vlib_buffer_push_ip6 (vlib_main_t * vm, vlib_buffer_t * b, + ip6_address_t * src, ip6_address_t * dst, int proto) +{ + ip6_header_t *ip6h; + u16 payload_length; + + /* make some room */ + ip6h = vlib_buffer_push_uninit (b, sizeof (ip6_header_t)); + + ip6h->ip_version_traffic_class_and_flow_label = + clib_host_to_net_u32 (0x6 << 28); + + /* calculate ip6 payload length */ + payload_length = vlib_buffer_length_in_chain (vm, b); + payload_length -= sizeof (*ip6h); + + ip6h->payload_length = clib_host_to_net_u16 (payload_length); + + ip6h->hop_limit = 0xff; + ip6h->protocol = proto; + clib_memcpy (ip6h->src_address.as_u8, src->as_u8, + sizeof (ip6h->src_address)); + clib_memcpy (ip6h->dst_address.as_u8, dst->as_u8, + sizeof (ip6h->src_address)); + + return ip6h; +} + #endif /* included_ip_ip6_h */ /* diff --git a/src/vnet/ip/ip6_packet.h b/src/vnet/ip/ip6_packet.h index 1e551c8b..4fd14b96 100644 --- a/src/vnet/ip/ip6_packet.h +++ b/src/vnet/ip/ip6_packet.h @@ -40,7 +40,7 @@ #ifndef included_ip6_packet_h #define included_ip6_packet_h -#include +#include #include typedef union @@ -373,10 +373,10 @@ ip6_tcp_reply_x1 (ip6_header_t * ip0, tcp_header_t * tcp0) { u16 src0, dst0; - src0 = tcp0->ports.src; - dst0 = tcp0->ports.dst; - tcp0->ports.src = dst0; - tcp0->ports.dst = src0; + src0 = tcp0->src; + dst0 = tcp0->dst; + tcp0->src = dst0; + tcp0->dst = src0; } } @@ -400,14 +400,14 @@ ip6_tcp_reply_x2 (ip6_header_t * ip0, ip6_header_t * ip1, { u16 src0, dst0, src1, dst1; - src0 = tcp0->ports.src; - src1 = tcp1->ports.src; - dst0 = tcp0->ports.dst; - dst1 = tcp1->ports.dst; - tcp0->ports.src = dst0; - tcp1->ports.src = dst1; - tcp0->ports.dst = src0; - tcp1->ports.dst = src1; + src0 = tcp0->src; + src1 = tcp1->src; + dst0 = tcp0->dst; + dst1 = tcp1->dst; + tcp0->src = dst0; + tcp1->src = dst1; + tcp0->dst = src0; + tcp1->dst = src1; } } diff --git a/src/vnet/ip/punt.c b/src/vnet/ip/punt.c index 9c735128..48558401 100644 --- a/src/vnet/ip/punt.c +++ b/src/vnet/ip/punt.c @@ -23,7 +23,7 @@ */ #include #include -#include +#include #include #define foreach_punt_next \ diff --git a/src/vnet/ip/tcp_packet.h b/src/vnet/ip/tcp_packet.h deleted file mode 100644 index 93f73e01..00000000 --- a/src/vnet/ip/tcp_packet.h +++ /dev/null @@ -1,141 +0,0 @@ -/* - * Copyright (c) 2015 Cisco and/or its affiliates. - * Licensed under the Apache License, Version 2.0 (the "License"); - * you may not use this file except in compliance with the License. - * You may obtain a copy of the License at: - * - * http://www.apache.org/licenses/LICENSE-2.0 - * - * Unless required by applicable law or agreed to in writing, software - * distributed under the License is distributed on an "AS IS" BASIS, - * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. - * See the License for the specific language governing permissions and - * limitations under the License. - */ -/* - * ip4/tcp_packet.h: TCP packet format (see RFC 793) - * - * Copyright (c) 2008 Eliot Dresselhaus - * - * Permission is hereby granted, free of charge, to any person obtaining - * a copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sublicense, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice shall be - * included in all copies or substantial portions of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, - * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND - * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE - * LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION - * OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION - * WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - */ - -#ifndef included_tcp_packet_h -#define included_tcp_packet_h - -/* TCP flags bit 0 first. */ -#define foreach_tcp_flag \ - _ (FIN) \ - _ (SYN) \ - _ (RST) \ - _ (PSH) \ - _ (ACK) \ - _ (URG) \ - _ (ECE) \ - _ (CWR) - -enum -{ -#define _(f) TCP_FLAG_BIT_##f, - foreach_tcp_flag -#undef _ - TCP_N_FLAG_BITS, - -#define _(f) TCP_FLAG_##f = 1 << TCP_FLAG_BIT_##f, - foreach_tcp_flag -#undef _ -}; - -typedef struct -{ - /* Source and destination port. */ - union - { - union - { - struct - { - u16 src, dst; - }; - u32 src_and_dst; - } ports; - struct - { - u16 src_port, dst_port; - }; - }; - - /* Sequence and acknowledgment number. */ - u32 seq_number, ack_number; - - /* Size of TCP header in 32-bit units plus 4 reserved bits. */ - u8 tcp_header_u32s_and_reserved; - - /* see foreach_tcp_flag for enumation of tcp flags. */ - u8 flags; - - /* Current window advertised by sender. - This is the number of bytes sender is willing to receive - right now. */ - u16 window; - - /* Checksum of TCP pseudo header and data. */ - u16 checksum; - - u16 urgent_pointer; -} tcp_header_t; - -always_inline int -tcp_header_bytes (tcp_header_t * t) -{ - return (t->tcp_header_u32s_and_reserved >> 4) * sizeof (u32); -} - -/* TCP options. */ -typedef enum tcp_option_type -{ - TCP_OPTION_END = 0, - TCP_OPTION_NOP = 1, - TCP_OPTION_MSS = 2, - TCP_OPTION_WINDOW_SCALE = 3, - TCP_OPTION_SACK_PERMITTED = 4, - TCP_OPTION_SACK_BLOCK = 5, - TCP_OPTION_TIME_STAMP = 8, -} tcp_option_type_t; - -/* All except NOP and END have 1 byte length field. */ -typedef struct -{ - tcp_option_type_t type:8; - - /* Length of this option in bytes. */ - u8 length; -} tcp_option_with_length_t; - -#endif /* included_tcp_packet_h */ - - -/* - * fd.io coding-style-patch-verification: ON - * - * Local Variables: - * eval: (c-set-style "gnu") - * End: - */ diff --git a/src/vnet/ip/udp.h b/src/vnet/ip/udp.h deleted file mode 100644 index bad58b5d..00000000 --- a/src/vnet/ip/udp.h +++ /dev/null @@ -1,315 +0,0 @@ -/* - * ip/udp.h: udp protocol - * - * Copyright (c) 2013 Cisco and/or its affiliates. - * Licensed under the Apache License, Version 2.0 (the "License"); - * you may not use this file except in compliance with the License. - * You may obtain a copy of the License at: - * - * http://www.apache.org/licenses/LICENSE-2.0 - * - * Unless required by applicable law or agreed to in writing, software - * distributed under the License is distributed on an "AS IS" BASIS, - * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. - * See the License for the specific language governing permissions and - * limitations under the License. - */ - -#ifndef included_udp_h -#define included_udp_h - -#include -#include -#include -#include -#include -#include -#include - -typedef enum -{ -#define udp_error(n,s) UDP_ERROR_##n, -#include -#undef udp_error - UDP_N_ERROR, -} udp_error_t; - -#define foreach_udp4_dst_port \ -_ (67, dhcp_to_server) \ -_ (68, dhcp_to_client) \ -_ (500, ikev2) \ -_ (3784, bfd4) \ -_ (3785, bfd_echo4) \ -_ (4341, lisp_gpe) \ -_ (4342, lisp_cp) \ -_ (4739, ipfix) \ -_ (4789, vxlan) \ -_ (4789, vxlan6) \ -_ (4790, vxlan_gpe) \ -_ (6633, vpath_3) - - -#define foreach_udp6_dst_port \ -_ (547, dhcpv6_to_server) \ -_ (546, dhcpv6_to_client) \ -_ (3784, bfd6) \ -_ (3785, bfd_echo6) \ -_ (4341, lisp_gpe6) \ -_ (4342, lisp_cp6) \ -_ (4790, vxlan6_gpe) \ -_ (6633, vpath6_3) - -typedef enum -{ -#define _(n,f) UDP_DST_PORT_##f = n, - foreach_udp4_dst_port foreach_udp6_dst_port -#undef _ -} udp_dst_port_t; - -typedef enum -{ -#define _(n,f) UDP6_DST_PORT_##f = n, - foreach_udp6_dst_port -#undef _ -} udp6_dst_port_t; - -typedef struct -{ - /* Name (a c string). */ - char *name; - - /* GRE protocol type in host byte order. */ - udp_dst_port_t dst_port; - - /* Node which handles this type. */ - u32 node_index; - - /* Next index for this type. */ - u32 next_index; -} udp_dst_port_info_t; - -typedef enum -{ - UDP_IP6 = 0, - UDP_IP4, /* the code is full of is_ip4... */ - N_UDP_AF, -} udp_af_t; - -typedef struct -{ - udp_dst_port_info_t *dst_port_infos[N_UDP_AF]; - - /* Hash tables mapping name/protocol to protocol info index. */ - uword *dst_port_info_by_name[N_UDP_AF]; - uword *dst_port_info_by_dst_port[N_UDP_AF]; - - /* convenience */ - vlib_main_t *vlib_main; -} udp_main_t; - -always_inline udp_dst_port_info_t * -udp_get_dst_port_info (udp_main_t * um, udp_dst_port_t dst_port, u8 is_ip4) -{ - uword *p = hash_get (um->dst_port_info_by_dst_port[is_ip4], dst_port); - return p ? vec_elt_at_index (um->dst_port_infos[is_ip4], p[0]) : 0; -} - -format_function_t format_udp_header; -format_function_t format_udp_rx_trace; - -unformat_function_t unformat_udp_header; - -void udp_register_dst_port (vlib_main_t * vm, - udp_dst_port_t dst_port, - u32 node_index, u8 is_ip4); - -void udp_punt_unknown (vlib_main_t * vm, u8 is_ip4, u8 is_add); - -always_inline void -ip_udp_fixup_one (vlib_main_t * vm, vlib_buffer_t * b0, u8 is_ip4) -{ - u16 new_l0; - udp_header_t *udp0; - - if (is_ip4) - { - ip4_header_t *ip0; - ip_csum_t sum0; - u16 old_l0 = 0; - - ip0 = vlib_buffer_get_current (b0); - - /* fix the ing outer-IP checksum */ - sum0 = ip0->checksum; - /* old_l0 always 0, see the rewrite setup */ - new_l0 = clib_host_to_net_u16 (vlib_buffer_length_in_chain (vm, b0)); - - sum0 = ip_csum_update (sum0, old_l0, new_l0, ip4_header_t, - length /* changed member */ ); - ip0->checksum = ip_csum_fold (sum0); - ip0->length = new_l0; - - /* Fix UDP length */ - udp0 = (udp_header_t *) (ip0 + 1); - new_l0 = clib_host_to_net_u16 (vlib_buffer_length_in_chain (vm, b0) - - sizeof (*ip0)); - udp0->length = new_l0; - } - else - { - ip6_header_t *ip0; - int bogus0; - - ip0 = vlib_buffer_get_current (b0); - - new_l0 = clib_host_to_net_u16 (vlib_buffer_length_in_chain (vm, b0) - - sizeof (*ip0)); - ip0->payload_length = new_l0; - - /* Fix UDP length */ - udp0 = (udp_header_t *) (ip0 + 1); - udp0->length = new_l0; - - udp0->checksum = - ip6_tcp_udp_icmp_compute_checksum (vm, b0, ip0, &bogus0); - ASSERT (bogus0 == 0); - - if (udp0->checksum == 0) - udp0->checksum = 0xffff; - } -} - -always_inline void -ip_udp_encap_one (vlib_main_t * vm, vlib_buffer_t * b0, u8 * ec0, word ec_len, - u8 is_ip4) -{ - vlib_buffer_advance (b0, -ec_len); - - if (is_ip4) - { - ip4_header_t *ip0; - - ip0 = vlib_buffer_get_current (b0); - - /* Apply the encap string. */ - clib_memcpy (ip0, ec0, ec_len); - ip_udp_fixup_one (vm, b0, 1); - } - else - { - ip6_header_t *ip0; - - ip0 = vlib_buffer_get_current (b0); - - /* Apply the encap string. */ - clib_memcpy (ip0, ec0, ec_len); - ip_udp_fixup_one (vm, b0, 0); - } -} - -always_inline void -ip_udp_encap_two (vlib_main_t * vm, vlib_buffer_t * b0, vlib_buffer_t * b1, - u8 * ec0, u8 * ec1, word ec_len, u8 is_v4) -{ - u16 new_l0, new_l1; - udp_header_t *udp0, *udp1; - - ASSERT (_vec_len (ec0) == _vec_len (ec1)); - - vlib_buffer_advance (b0, -ec_len); - vlib_buffer_advance (b1, -ec_len); - - if (is_v4) - { - ip4_header_t *ip0, *ip1; - ip_csum_t sum0, sum1; - u16 old_l0 = 0, old_l1 = 0; - - ip0 = vlib_buffer_get_current (b0); - ip1 = vlib_buffer_get_current (b1); - - /* Apply the encap string */ - clib_memcpy (ip0, ec0, ec_len); - clib_memcpy (ip1, ec1, ec_len); - - /* fix the ing outer-IP checksum */ - sum0 = ip0->checksum; - sum1 = ip1->checksum; - - /* old_l0 always 0, see the rewrite setup */ - new_l0 = clib_host_to_net_u16 (vlib_buffer_length_in_chain (vm, b0)); - new_l1 = clib_host_to_net_u16 (vlib_buffer_length_in_chain (vm, b1)); - - sum0 = ip_csum_update (sum0, old_l0, new_l0, ip4_header_t, - length /* changed member */ ); - sum1 = ip_csum_update (sum1, old_l1, new_l1, ip4_header_t, - length /* changed member */ ); - - ip0->checksum = ip_csum_fold (sum0); - ip1->checksum = ip_csum_fold (sum1); - - ip0->length = new_l0; - ip1->length = new_l1; - - /* Fix UDP length */ - udp0 = (udp_header_t *) (ip0 + 1); - udp1 = (udp_header_t *) (ip1 + 1); - - new_l0 = - clib_host_to_net_u16 (vlib_buffer_length_in_chain (vm, b0) - - sizeof (*ip0)); - new_l1 = - clib_host_to_net_u16 (vlib_buffer_length_in_chain (vm, b1) - - sizeof (*ip1)); - udp0->length = new_l0; - udp1->length = new_l1; - } - else - { - ip6_header_t *ip0, *ip1; - int bogus0, bogus1; - - ip0 = vlib_buffer_get_current (b0); - ip1 = vlib_buffer_get_current (b1); - - /* Apply the encap string. */ - clib_memcpy (ip0, ec0, ec_len); - clib_memcpy (ip1, ec1, ec_len); - - new_l0 = clib_host_to_net_u16 (vlib_buffer_length_in_chain (vm, b0) - - sizeof (*ip0)); - new_l1 = clib_host_to_net_u16 (vlib_buffer_length_in_chain (vm, b1) - - sizeof (*ip1)); - ip0->payload_length = new_l0; - ip1->payload_length = new_l1; - - /* Fix UDP length */ - udp0 = (udp_header_t *) (ip0 + 1); - udp1 = (udp_header_t *) (ip1 + 1); - - udp0->length = new_l0; - udp1->length = new_l1; - - udp0->checksum = - ip6_tcp_udp_icmp_compute_checksum (vm, b0, ip0, &bogus0); - udp1->checksum = - ip6_tcp_udp_icmp_compute_checksum (vm, b1, ip1, &bogus1); - ASSERT (bogus0 == 0); - ASSERT (bogus1 == 0); - - if (udp0->checksum == 0) - udp0->checksum = 0xffff; - if (udp1->checksum == 0) - udp1->checksum = 0xffff; - } -} - -#endif /* included_udp_h */ - -/* - * fd.io coding-style-patch-verification: ON - * - * Local Variables: - * eval: (c-set-style "gnu") - * End: - */ diff --git a/src/vnet/ip/udp_error.def b/src/vnet/ip/udp_error.def deleted file mode 100644 index bfdae0ac..00000000 --- a/src/vnet/ip/udp_error.def +++ /dev/null @@ -1,21 +0,0 @@ -/* - * udp_error.def: udp errors - * - * Copyright (c) 2013-2016 Cisco and/or its affiliates. - * Licensed under the Apache License, Version 2.0 (the "License"); - * you may not use this file except in compliance with the License. - * You may obtain a copy of the License at: - * - * http://www.apache.org/licenses/LICENSE-2.0 - * - * Unless required by applicable law or agreed to in writing, software - * distributed under the License is distributed on an "AS IS" BASIS, - * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. - * See the License for the specific language governing permissions and - * limitations under the License. - */ - -udp_error (NONE, "no error") -udp_error (NO_LISTENER, "no listener for dst port") -udp_error (LENGTH_ERROR, "UDP packets with length errors") -udp_error (PUNT, "no listener punt") diff --git a/src/vnet/ip/udp_format.c b/src/vnet/ip/udp_format.c deleted file mode 100644 index abdf561e..00000000 --- a/src/vnet/ip/udp_format.c +++ /dev/null @@ -1,91 +0,0 @@ -/* - * Copyright (c) 2015 Cisco and/or its affiliates. - * Licensed under the Apache License, Version 2.0 (the "License"); - * you may not use this file except in compliance with the License. - * You may obtain a copy of the License at: - * - * http://www.apache.org/licenses/LICENSE-2.0 - * - * Unless required by applicable law or agreed to in writing, software - * distributed under the License is distributed on an "AS IS" BASIS, - * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. - * See the License for the specific language governing permissions and - * limitations under the License. - */ -/* - * ip/udp_format.c: udp formatting - * - * Copyright (c) 2008 Eliot Dresselhaus - * - * Permission is hereby granted, free of charge, to any person obtaining - * a copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sublicense, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice shall be - * included in all copies or substantial portions of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, - * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND - * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE - * LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION - * OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION - * WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - */ - -#include - -/* Format UDP header. */ -u8 * -format_udp_header (u8 * s, va_list * args) -{ - udp_header_t *udp = va_arg (*args, udp_header_t *); - u32 max_header_bytes = va_arg (*args, u32); - uword indent; - u32 header_bytes = sizeof (udp[0]); - - /* Nothing to do. */ - if (max_header_bytes < sizeof (udp[0])) - return format (s, "UDP header truncated"); - - indent = format_get_indent (s); - indent += 2; - - s = format (s, "UDP: %d -> %d", - clib_net_to_host_u16 (udp->src_port), - clib_net_to_host_u16 (udp->dst_port)); - - s = format (s, "\n%Ulength %d, checksum 0x%04x", - format_white_space, indent, - clib_net_to_host_u16 (udp->length), - clib_net_to_host_u16 (udp->checksum)); - - /* Recurse into next protocol layer. */ - if (max_header_bytes != 0 && header_bytes < max_header_bytes) - { - ip_main_t *im = &ip_main; - tcp_udp_port_info_t *pi; - - pi = ip_get_tcp_udp_port_info (im, udp->dst_port); - - if (pi && pi->format_header) - s = format (s, "\n%U%U", - format_white_space, indent - 2, pi->format_header, - /* next protocol header */ (udp + 1), - max_header_bytes - sizeof (udp[0])); - } - - return s; -} - -/* - * fd.io coding-style-patch-verification: ON - * - * Local Variables: - * eval: (c-set-style "gnu") - * End: - */ diff --git a/src/vnet/ip/udp_init.c b/src/vnet/ip/udp_init.c deleted file mode 100644 index 1241ca4a..00000000 --- a/src/vnet/ip/udp_init.c +++ /dev/null @@ -1,71 +0,0 @@ -/* - * Copyright (c) 2015 Cisco and/or its affiliates. - * Licensed under the Apache License, Version 2.0 (the "License"); - * you may not use this file except in compliance with the License. - * You may obtain a copy of the License at: - * - * http://www.apache.org/licenses/LICENSE-2.0 - * - * Unless required by applicable law or agreed to in writing, software - * distributed under the License is distributed on an "AS IS" BASIS, - * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. - * See the License for the specific language governing permissions and - * limitations under the License. - */ -/* - * ip/udp_init.c: udp initialization - * - * Copyright (c) 2008 Eliot Dresselhaus - * - * Permission is hereby granted, free of charge, to any person obtaining - * a copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sublicense, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice shall be - * included in all copies or substantial portions of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, - * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND - * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE - * LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION - * OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION - * WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - */ - -#include - -clib_error_t * -udp_init (vlib_main_t * vm) -{ - ip_main_t *im = &ip_main; - ip_protocol_info_t *pi; - clib_error_t *error; - - error = vlib_call_init_function (vm, ip_main_init); - - if (!error) - { - pi = ip_get_protocol_info (im, IP_PROTOCOL_UDP); - if (pi == 0) - return clib_error_return (0, "UDP protocol info AWOL"); - pi->format_header = format_udp_header; - pi->unformat_pg_edit = unformat_pg_udp_header; - } - - return 0; -} - -VLIB_INIT_FUNCTION (udp_init); - -/* - * fd.io coding-style-patch-verification: ON - * - * Local Variables: - * eval: (c-set-style "gnu") - * End: - */ diff --git a/src/vnet/ip/udp_local.c b/src/vnet/ip/udp_local.c deleted file mode 100644 index 13ab6e4f..00000000 --- a/src/vnet/ip/udp_local.c +++ /dev/null @@ -1,645 +0,0 @@ -/* - * node.c: udp packet processing - * - * Copyright (c) 2013 Cisco and/or its affiliates. - * Licensed under the Apache License, Version 2.0 (the "License"); - * you may not use this file except in compliance with the License. - * You may obtain a copy of the License at: - * - * http://www.apache.org/licenses/LICENSE-2.0 - * - * Unless required by applicable law or agreed to in writing, software - * distributed under the License is distributed on an "AS IS" BASIS, - * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. - * See the License for the specific language governing permissions and - * limitations under the License. - */ - -#include -#include -#include -#include -#include - -udp_main_t udp_main; - -#define foreach_udp_input_next \ - _ (PUNT, "error-punt") \ - _ (DROP, "error-drop") \ - _ (ICMP4_ERROR, "ip4-icmp-error") \ - _ (ICMP6_ERROR, "ip6-icmp-error") - -typedef enum -{ -#define _(s,n) UDP_INPUT_NEXT_##s, - foreach_udp_input_next -#undef _ - UDP_INPUT_N_NEXT, -} udp_input_next_t; - -typedef struct -{ - u16 src_port; - u16 dst_port; - u8 bound; -} udp_rx_trace_t; - -u8 * -format_udp_rx_trace (u8 * s, va_list * args) -{ - CLIB_UNUSED (vlib_main_t * vm) = va_arg (*args, vlib_main_t *); - CLIB_UNUSED (vlib_node_t * node) = va_arg (*args, vlib_node_t *); - udp_rx_trace_t *t = va_arg (*args, udp_rx_trace_t *); - - s = format (s, "UDP: src-port %d dst-port %d%s", - clib_net_to_host_u16 (t->src_port), - clib_net_to_host_u16 (t->dst_port), - t->bound ? "" : " (no listener)"); - return s; -} - -typedef struct -{ - /* Sparse vector mapping udp dst_port in network byte order - to next index. */ - u16 *next_by_dst_port; - u8 punt_unknown; -} udp_input_runtime_t; - -vlib_node_registration_t udp4_input_node; -vlib_node_registration_t udp6_input_node; - -always_inline uword -udp46_input_inline (vlib_main_t * vm, - vlib_node_runtime_t * node, - vlib_frame_t * from_frame, int is_ip4) -{ - udp_input_runtime_t *rt = is_ip4 ? - (void *) vlib_node_get_runtime_data (vm, udp4_input_node.index) - : (void *) vlib_node_get_runtime_data (vm, udp6_input_node.index); - __attribute__ ((unused)) u32 n_left_from, next_index, *from, *to_next; - word n_no_listener = 0; - u8 punt_unknown = rt->punt_unknown; - - from = vlib_frame_vector_args (from_frame); - n_left_from = from_frame->n_vectors; - - next_index = node->cached_next_index; - - while (n_left_from > 0) - { - u32 n_left_to_next; - - vlib_get_next_frame (vm, node, next_index, to_next, n_left_to_next); - - while (n_left_from >= 4 && n_left_to_next >= 2) - { - u32 bi0, bi1; - vlib_buffer_t *b0, *b1; - udp_header_t *h0 = 0, *h1 = 0; - u32 i0, i1, dst_port0, dst_port1; - u32 advance0, advance1; - u32 error0, next0, error1, next1; - - /* Prefetch next iteration. */ - { - vlib_buffer_t *p2, *p3; - - p2 = vlib_get_buffer (vm, from[2]); - p3 = vlib_get_buffer (vm, from[3]); - - vlib_prefetch_buffer_header (p2, LOAD); - vlib_prefetch_buffer_header (p3, LOAD); - - CLIB_PREFETCH (p2->data, sizeof (h0[0]), LOAD); - CLIB_PREFETCH (p3->data, sizeof (h1[0]), LOAD); - } - - bi0 = from[0]; - bi1 = from[1]; - to_next[0] = bi0; - to_next[1] = bi1; - from += 2; - to_next += 2; - n_left_to_next -= 2; - n_left_from -= 2; - - b0 = vlib_get_buffer (vm, bi0); - b1 = vlib_get_buffer (vm, bi1); - - /* ip4/6_local hands us the ip header, not the udp header */ - if (is_ip4) - { - advance0 = sizeof (ip4_header_t); - advance1 = sizeof (ip4_header_t); - } - else - { - advance0 = sizeof (ip6_header_t); - advance1 = sizeof (ip6_header_t); - } - - if (PREDICT_FALSE (b0->current_length < advance0 + sizeof (*h0))) - { - error0 = UDP_ERROR_LENGTH_ERROR; - next0 = UDP_INPUT_NEXT_DROP; - } - else - { - vlib_buffer_advance (b0, advance0); - h0 = vlib_buffer_get_current (b0); - error0 = next0 = 0; - if (PREDICT_FALSE (clib_net_to_host_u16 (h0->length) > - vlib_buffer_length_in_chain (vm, b0))) - { - error0 = UDP_ERROR_LENGTH_ERROR; - next0 = UDP_INPUT_NEXT_DROP; - } - } - - if (PREDICT_FALSE (b1->current_length < advance1 + sizeof (*h1))) - { - error1 = UDP_ERROR_LENGTH_ERROR; - next1 = UDP_INPUT_NEXT_DROP; - } - else - { - vlib_buffer_advance (b1, advance1); - h1 = vlib_buffer_get_current (b1); - error1 = next1 = 0; - if (PREDICT_FALSE (clib_net_to_host_u16 (h1->length) > - vlib_buffer_length_in_chain (vm, b1))) - { - error1 = UDP_ERROR_LENGTH_ERROR; - next1 = UDP_INPUT_NEXT_DROP; - } - } - - /* Index sparse array with network byte order. */ - dst_port0 = (error0 == 0) ? h0->dst_port : 0; - dst_port1 = (error1 == 0) ? h1->dst_port : 0; - sparse_vec_index2 (rt->next_by_dst_port, dst_port0, dst_port1, - &i0, &i1); - next0 = (error0 == 0) ? vec_elt (rt->next_by_dst_port, i0) : next0; - next1 = (error1 == 0) ? vec_elt (rt->next_by_dst_port, i1) : next1; - - if (PREDICT_FALSE (i0 == SPARSE_VEC_INVALID_INDEX)) - { - // move the pointer back so icmp-error can find the - // ip packet header - vlib_buffer_advance (b0, -(word) advance0); - - if (PREDICT_FALSE (punt_unknown)) - { - b0->error = node->errors[UDP_ERROR_PUNT]; - next0 = UDP_INPUT_NEXT_PUNT; - } - else if (is_ip4) - { - icmp4_error_set_vnet_buffer (b0, - ICMP4_destination_unreachable, - ICMP4_destination_unreachable_port_unreachable, - 0); - next0 = UDP_INPUT_NEXT_ICMP4_ERROR; - n_no_listener++; - } - else - { - icmp6_error_set_vnet_buffer (b0, - ICMP6_destination_unreachable, - ICMP6_destination_unreachable_port_unreachable, - 0); - next0 = UDP_INPUT_NEXT_ICMP6_ERROR; - n_no_listener++; - } - } - else - { - b0->error = node->errors[UDP_ERROR_NONE]; - // advance to the payload - vlib_buffer_advance (b0, sizeof (*h0)); - } - - if (PREDICT_FALSE (i1 == SPARSE_VEC_INVALID_INDEX)) - { - // move the pointer back so icmp-error can find the - // ip packet header - vlib_buffer_advance (b1, -(word) advance1); - - if (PREDICT_FALSE (punt_unknown)) - { - b1->error = node->errors[UDP_ERROR_PUNT]; - next1 = UDP_INPUT_NEXT_PUNT; - } - else if (is_ip4) - { - icmp4_error_set_vnet_buffer (b1, - ICMP4_destination_unreachable, - ICMP4_destination_unreachable_port_unreachable, - 0); - next1 = UDP_INPUT_NEXT_ICMP4_ERROR; - n_no_listener++; - } - else - { - icmp6_error_set_vnet_buffer (b1, - ICMP6_destination_unreachable, - ICMP6_destination_unreachable_port_unreachable, - 0); - next1 = UDP_INPUT_NEXT_ICMP6_ERROR; - n_no_listener++; - } - } - else - { - b1->error = node->errors[UDP_ERROR_NONE]; - // advance to the payload - vlib_buffer_advance (b1, sizeof (*h1)); - } - - if (PREDICT_FALSE (b0->flags & VLIB_BUFFER_IS_TRACED)) - { - udp_rx_trace_t *tr = vlib_add_trace (vm, node, - b0, sizeof (*tr)); - if (b0->error != node->errors[UDP_ERROR_LENGTH_ERROR]) - { - tr->src_port = h0 ? h0->src_port : 0; - tr->dst_port = h0 ? h0->dst_port : 0; - tr->bound = (next0 != UDP_INPUT_NEXT_ICMP4_ERROR && - next0 != UDP_INPUT_NEXT_ICMP6_ERROR); - } - } - if (PREDICT_FALSE (b1->flags & VLIB_BUFFER_IS_TRACED)) - { - udp_rx_trace_t *tr = vlib_add_trace (vm, node, - b1, sizeof (*tr)); - if (b1->error != node->errors[UDP_ERROR_LENGTH_ERROR]) - { - tr->src_port = h1 ? h1->src_port : 0; - tr->dst_port = h1 ? h1->dst_port : 0; - tr->bound = (next1 != UDP_INPUT_NEXT_ICMP4_ERROR && - next1 != UDP_INPUT_NEXT_ICMP6_ERROR); - } - } - - vlib_validate_buffer_enqueue_x2 (vm, node, next_index, - to_next, n_left_to_next, - bi0, bi1, next0, next1); - } - - while (n_left_from > 0 && n_left_to_next > 0) - { - u32 bi0; - vlib_buffer_t *b0; - udp_header_t *h0 = 0; - u32 i0, next0; - u32 advance0; - - bi0 = from[0]; - to_next[0] = bi0; - from += 1; - to_next += 1; - n_left_from -= 1; - n_left_to_next -= 1; - - b0 = vlib_get_buffer (vm, bi0); - - /* ip4/6_local hands us the ip header, not the udp header */ - if (is_ip4) - advance0 = sizeof (ip4_header_t); - else - advance0 = sizeof (ip6_header_t); - - if (PREDICT_FALSE (b0->current_length < advance0 + sizeof (*h0))) - { - b0->error = node->errors[UDP_ERROR_LENGTH_ERROR]; - next0 = UDP_INPUT_NEXT_DROP; - goto trace_x1; - } - - vlib_buffer_advance (b0, advance0); - - h0 = vlib_buffer_get_current (b0); - - if (PREDICT_TRUE (clib_net_to_host_u16 (h0->length) <= - vlib_buffer_length_in_chain (vm, b0))) - { - i0 = sparse_vec_index (rt->next_by_dst_port, h0->dst_port); - next0 = vec_elt (rt->next_by_dst_port, i0); - - if (PREDICT_FALSE (i0 == SPARSE_VEC_INVALID_INDEX)) - { - // move the pointer back so icmp-error can find the - // ip packet header - vlib_buffer_advance (b0, -(word) advance0); - - if (PREDICT_FALSE (punt_unknown)) - { - b0->error = node->errors[UDP_ERROR_PUNT]; - next0 = UDP_INPUT_NEXT_PUNT; - } - else if (is_ip4) - { - icmp4_error_set_vnet_buffer (b0, - ICMP4_destination_unreachable, - ICMP4_destination_unreachable_port_unreachable, - 0); - next0 = UDP_INPUT_NEXT_ICMP4_ERROR; - n_no_listener++; - } - else - { - icmp6_error_set_vnet_buffer (b0, - ICMP6_destination_unreachable, - ICMP6_destination_unreachable_port_unreachable, - 0); - next0 = UDP_INPUT_NEXT_ICMP6_ERROR; - n_no_listener++; - } - } - else - { - b0->error = node->errors[UDP_ERROR_NONE]; - // advance to the payload - vlib_buffer_advance (b0, sizeof (*h0)); - } - } - else - { - b0->error = node->errors[UDP_ERROR_LENGTH_ERROR]; - next0 = UDP_INPUT_NEXT_DROP; - } - - trace_x1: - if (PREDICT_FALSE (b0->flags & VLIB_BUFFER_IS_TRACED)) - { - udp_rx_trace_t *tr = vlib_add_trace (vm, node, - b0, sizeof (*tr)); - if (b0->error != node->errors[UDP_ERROR_LENGTH_ERROR]) - { - tr->src_port = h0->src_port; - tr->dst_port = h0->dst_port; - tr->bound = (next0 != UDP_INPUT_NEXT_ICMP4_ERROR && - next0 != UDP_INPUT_NEXT_ICMP6_ERROR); - } - } - - vlib_validate_buffer_enqueue_x1 (vm, node, next_index, - to_next, n_left_to_next, - bi0, next0); - } - - vlib_put_next_frame (vm, node, next_index, n_left_to_next); - } - vlib_error_count (vm, node->node_index, UDP_ERROR_NO_LISTENER, - n_no_listener); - return from_frame->n_vectors; -} - -static char *udp_error_strings[] = { -#define udp_error(n,s) s, -#include "udp_error.def" -#undef udp_error -}; - -static uword -udp4_input (vlib_main_t * vm, - vlib_node_runtime_t * node, vlib_frame_t * from_frame) -{ - return udp46_input_inline (vm, node, from_frame, 1 /* is_ip4 */ ); -} - -static uword -udp6_input (vlib_main_t * vm, - vlib_node_runtime_t * node, vlib_frame_t * from_frame) -{ - return udp46_input_inline (vm, node, from_frame, 0 /* is_ip4 */ ); -} - - -/* *INDENT-OFF* */ -VLIB_REGISTER_NODE (udp4_input_node) = { - .function = udp4_input, - .name = "ip4-udp-lookup", - /* Takes a vector of packets. */ - .vector_size = sizeof (u32), - - .runtime_data_bytes = sizeof (udp_input_runtime_t), - - .n_errors = UDP_N_ERROR, - .error_strings = udp_error_strings, - - .n_next_nodes = UDP_INPUT_N_NEXT, - .next_nodes = { -#define _(s,n) [UDP_INPUT_NEXT_##s] = n, - foreach_udp_input_next -#undef _ - }, - - .format_buffer = format_udp_header, - .format_trace = format_udp_rx_trace, - .unformat_buffer = unformat_udp_header, -}; -/* *INDENT-ON* */ - -VLIB_NODE_FUNCTION_MULTIARCH (udp4_input_node, udp4_input); - -/* *INDENT-OFF* */ -VLIB_REGISTER_NODE (udp6_input_node) = { - .function = udp6_input, - .name = "ip6-udp-lookup", - /* Takes a vector of packets. */ - .vector_size = sizeof (u32), - - .runtime_data_bytes = sizeof (udp_input_runtime_t), - - .n_errors = UDP_N_ERROR, - .error_strings = udp_error_strings, - - .n_next_nodes = UDP_INPUT_N_NEXT, - .next_nodes = { -#define _(s,n) [UDP_INPUT_NEXT_##s] = n, - foreach_udp_input_next -#undef _ - }, - - .format_buffer = format_udp_header, - .format_trace = format_udp_rx_trace, - .unformat_buffer = unformat_udp_header, -}; -/* *INDENT-ON* */ - -VLIB_NODE_FUNCTION_MULTIARCH (udp6_input_node, udp6_input); - -static void -add_dst_port (udp_main_t * um, - udp_dst_port_t dst_port, char *dst_port_name, u8 is_ip4) -{ - udp_dst_port_info_t *pi; - u32 i; - - vec_add2 (um->dst_port_infos[is_ip4], pi, 1); - i = pi - um->dst_port_infos[is_ip4]; - - pi->name = dst_port_name; - pi->dst_port = dst_port; - pi->next_index = pi->node_index = ~0; - - hash_set (um->dst_port_info_by_dst_port[is_ip4], dst_port, i); - - if (pi->name) - hash_set_mem (um->dst_port_info_by_name[is_ip4], pi->name, i); -} - -void -udp_register_dst_port (vlib_main_t * vm, - udp_dst_port_t dst_port, u32 node_index, u8 is_ip4) -{ - udp_main_t *um = &udp_main; - udp_dst_port_info_t *pi; - udp_input_runtime_t *rt; - u16 *n; - - { - clib_error_t *error = vlib_call_init_function (vm, udp_local_init); - if (error) - clib_error_report (error); - } - - pi = udp_get_dst_port_info (um, dst_port, is_ip4); - if (!pi) - { - add_dst_port (um, dst_port, 0, is_ip4); - pi = udp_get_dst_port_info (um, dst_port, is_ip4); - ASSERT (pi); - } - - pi->node_index = node_index; - pi->next_index = vlib_node_add_next (vm, - is_ip4 ? udp4_input_node.index - : udp6_input_node.index, node_index); - - /* Setup udp protocol -> next index sparse vector mapping. */ - rt = vlib_node_get_runtime_data - (vm, is_ip4 ? udp4_input_node.index : udp6_input_node.index); - n = sparse_vec_validate (rt->next_by_dst_port, - clib_host_to_net_u16 (dst_port)); - n[0] = pi->next_index; -} - -void -udp_punt_unknown (vlib_main_t * vm, u8 is_ip4, u8 is_add) -{ - udp_input_runtime_t *rt; - - { - clib_error_t *error = vlib_call_init_function (vm, udp_local_init); - if (error) - clib_error_report (error); - } - - rt = vlib_node_get_runtime_data - (vm, is_ip4 ? udp4_input_node.index : udp6_input_node.index); - - rt->punt_unknown = is_add; -} - -/* Parse a UDP header. */ -uword -unformat_udp_header (unformat_input_t * input, va_list * args) -{ - u8 **result = va_arg (*args, u8 **); - udp_header_t *udp; - __attribute__ ((unused)) int old_length; - u16 src_port, dst_port; - - /* Allocate space for IP header. */ - { - void *p; - - old_length = vec_len (*result); - vec_add2 (*result, p, sizeof (ip4_header_t)); - udp = p; - } - - memset (udp, 0, sizeof (udp[0])); - if (unformat (input, "src-port %d dst-port %d", &src_port, &dst_port)) - { - udp->src_port = clib_host_to_net_u16 (src_port); - udp->dst_port = clib_host_to_net_u16 (dst_port); - return 1; - } - return 0; -} - -static void -udp_setup_node (vlib_main_t * vm, u32 node_index) -{ - vlib_node_t *n = vlib_get_node (vm, node_index); - pg_node_t *pn = pg_get_node (node_index); - - n->format_buffer = format_udp_header; - n->unformat_buffer = unformat_udp_header; - pn->unformat_edit = unformat_pg_udp_header; -} - -clib_error_t * -udp_local_init (vlib_main_t * vm) -{ - udp_input_runtime_t *rt; - udp_main_t *um = &udp_main; - int i; - - { - clib_error_t *error; - error = vlib_call_init_function (vm, udp_init); - if (error) - clib_error_report (error); - } - - - for (i = 0; i < 2; i++) - { - um->dst_port_info_by_name[i] = hash_create_string (0, sizeof (uword)); - um->dst_port_info_by_dst_port[i] = hash_create (0, sizeof (uword)); - } - - udp_setup_node (vm, udp4_input_node.index); - udp_setup_node (vm, udp6_input_node.index); - - rt = vlib_node_get_runtime_data (vm, udp4_input_node.index); - - rt->next_by_dst_port = sparse_vec_new - ( /* elt bytes */ sizeof (rt->next_by_dst_port[0]), - /* bits in index */ BITS (((udp_header_t *) 0)->dst_port)); - - rt->punt_unknown = 0; - -#define _(n,s) add_dst_port (um, UDP_DST_PORT_##s, #s, 1 /* is_ip4 */); - foreach_udp4_dst_port -#undef _ - rt = vlib_node_get_runtime_data (vm, udp6_input_node.index); - - rt->next_by_dst_port = sparse_vec_new - ( /* elt bytes */ sizeof (rt->next_by_dst_port[0]), - /* bits in index */ BITS (((udp_header_t *) 0)->dst_port)); - - rt->punt_unknown = 0; - -#define _(n,s) add_dst_port (um, UDP_DST_PORT_##s, #s, 0 /* is_ip4 */); - foreach_udp6_dst_port -#undef _ - ip4_register_protocol (IP_PROTOCOL_UDP, udp4_input_node.index); - /* Note: ip6 differs from ip4, UDP is hotwired to ip6-udp-lookup */ - return 0; -} - -VLIB_INIT_FUNCTION (udp_local_init); - -/* - * fd.io coding-style-patch-verification: ON - * - * Local Variables: - * eval: (c-set-style "gnu") - * End: - */ diff --git a/src/vnet/ip/udp_packet.h b/src/vnet/ip/udp_packet.h deleted file mode 100644 index beea3059..00000000 --- a/src/vnet/ip/udp_packet.h +++ /dev/null @@ -1,65 +0,0 @@ -/* - * Copyright (c) 2015 Cisco and/or its affiliates. - * Licensed under the Apache License, Version 2.0 (the "License"); - * you may not use this file except in compliance with the License. - * You may obtain a copy of the License at: - * - * http://www.apache.org/licenses/LICENSE-2.0 - * - * Unless required by applicable law or agreed to in writing, software - * distributed under the License is distributed on an "AS IS" BASIS, - * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. - * See the License for the specific language governing permissions and - * limitations under the License. - */ -/* - * ip4/udp_packet.h: UDP packet format - * - * Copyright (c) 2008 Eliot Dresselhaus - * - * Permission is hereby granted, free of charge, to any person obtaining - * a copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sublicense, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice shall be - * included in all copies or substantial portions of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, - * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND - * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE - * LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION - * OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION - * WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - */ - -#ifndef included_udp_packet_h -#define included_udp_packet_h - -typedef struct -{ - /* Source and destination port. */ - u16 src_port, dst_port; - - /* Length of UDP header plus payload. */ - u16 length; - - /* Checksum of UDP pseudo-header and data or - zero if checksum is disabled. */ - u16 checksum; -} udp_header_t; - -#endif /* included_udp_packet_h */ - - -/* - * fd.io coding-style-patch-verification: ON - * - * Local Variables: - * eval: (c-set-style "gnu") - * End: - */ diff --git a/src/vnet/ip/udp_pg.c b/src/vnet/ip/udp_pg.c deleted file mode 100644 index c9d8d38c..00000000 --- a/src/vnet/ip/udp_pg.c +++ /dev/null @@ -1,237 +0,0 @@ -/* - * Copyright (c) 2015 Cisco and/or its affiliates. - * Licensed under the Apache License, Version 2.0 (the "License"); - * you may not use this file except in compliance with the License. - * You may obtain a copy of the License at: - * - * http://www.apache.org/licenses/LICENSE-2.0 - * - * Unless required by applicable law or agreed to in writing, software - * distributed under the License is distributed on an "AS IS" BASIS, - * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. - * See the License for the specific language governing permissions and - * limitations under the License. - */ -/* - * ip/udp_pg: UDP packet-generator interface - * - * Copyright (c) 2008 Eliot Dresselhaus - * - * Permission is hereby granted, free of charge, to any person obtaining - * a copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sublicense, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice shall be - * included in all copies or substantial portions of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, - * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND - * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE - * LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION - * OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION - * WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - */ - -#include -#include /* for unformat_udp_udp_port */ - -#define UDP_PG_EDIT_LENGTH (1 << 0) -#define UDP_PG_EDIT_CHECKSUM (1 << 1) - -always_inline void -udp_pg_edit_function_inline (pg_main_t * pg, - pg_stream_t * s, - pg_edit_group_t * g, - u32 * packets, u32 n_packets, u32 flags) -{ - vlib_main_t *vm = vlib_get_main (); - u32 ip_offset, udp_offset; - - udp_offset = g->start_byte_offset; - ip_offset = (g - 1)->start_byte_offset; - - while (n_packets >= 1) - { - vlib_buffer_t *p0; - ip4_header_t *ip0; - udp_header_t *udp0; - u32 udp_len0; - - p0 = vlib_get_buffer (vm, packets[0]); - n_packets -= 1; - packets += 1; - - ip0 = (void *) (p0->data + ip_offset); - udp0 = (void *) (p0->data + udp_offset); - udp_len0 = clib_net_to_host_u16 (ip0->length) - sizeof (ip0[0]); - - if (flags & UDP_PG_EDIT_LENGTH) - udp0->length = - clib_net_to_host_u16 (vlib_buffer_length_in_chain (vm, p0) - - ip_offset); - - /* Initialize checksum with header. */ - if (flags & UDP_PG_EDIT_CHECKSUM) - { - ip_csum_t sum0; - - sum0 = clib_mem_unaligned (&ip0->src_address, u64); - - sum0 = ip_csum_with_carry - (sum0, clib_host_to_net_u32 (udp_len0 + (ip0->protocol << 16))); - - /* Invalidate possibly old checksum. */ - udp0->checksum = 0; - - sum0 = - ip_incremental_checksum_buffer (vm, p0, udp_offset, udp_len0, - sum0); - - sum0 = ~ip_csum_fold (sum0); - - /* Zero checksum means checksumming disabled. */ - sum0 = sum0 != 0 ? sum0 : 0xffff; - - udp0->checksum = sum0; - } - } -} - -static void -udp_pg_edit_function (pg_main_t * pg, - pg_stream_t * s, - pg_edit_group_t * g, u32 * packets, u32 n_packets) -{ - switch (g->edit_function_opaque) - { - case UDP_PG_EDIT_LENGTH: - udp_pg_edit_function_inline (pg, s, g, packets, n_packets, - UDP_PG_EDIT_LENGTH); - break; - - case UDP_PG_EDIT_CHECKSUM: - udp_pg_edit_function_inline (pg, s, g, packets, n_packets, - UDP_PG_EDIT_CHECKSUM); - break; - - case UDP_PG_EDIT_CHECKSUM | UDP_PG_EDIT_LENGTH: - udp_pg_edit_function_inline (pg, s, g, packets, n_packets, - UDP_PG_EDIT_CHECKSUM | UDP_PG_EDIT_LENGTH); - break; - - default: - ASSERT (0); - break; - } -} - -typedef struct -{ - pg_edit_t src_port, dst_port; - pg_edit_t length; - pg_edit_t checksum; -} pg_udp_header_t; - -static inline void -pg_udp_header_init (pg_udp_header_t * p) -{ - /* Initialize fields that are not bit fields in the IP header. */ -#define _(f) pg_edit_init (&p->f, udp_header_t, f); - _(src_port); - _(dst_port); - _(length); - _(checksum); -#undef _ -} - -uword -unformat_pg_udp_header (unformat_input_t * input, va_list * args) -{ - pg_stream_t *s = va_arg (*args, pg_stream_t *); - pg_udp_header_t *p; - u32 group_index; - - p = pg_create_edit_group (s, sizeof (p[0]), sizeof (udp_header_t), - &group_index); - pg_udp_header_init (p); - - /* Defaults. */ - p->checksum.type = PG_EDIT_UNSPECIFIED; - p->length.type = PG_EDIT_UNSPECIFIED; - - if (!unformat (input, "UDP: %U -> %U", - unformat_pg_edit, - unformat_tcp_udp_port, &p->src_port, - unformat_pg_edit, unformat_tcp_udp_port, &p->dst_port)) - goto error; - - /* Parse options. */ - while (1) - { - if (unformat (input, "length %U", - unformat_pg_edit, unformat_pg_number, &p->length)) - ; - - else if (unformat (input, "checksum %U", - unformat_pg_edit, unformat_pg_number, &p->checksum)) - ; - - /* Can't parse input: try next protocol level. */ - else - break; - } - - { - ip_main_t *im = &ip_main; - u16 dst_port; - tcp_udp_port_info_t *pi; - - pi = 0; - if (p->dst_port.type == PG_EDIT_FIXED) - { - dst_port = pg_edit_get_value (&p->dst_port, PG_EDIT_LO); - pi = ip_get_tcp_udp_port_info (im, dst_port); - } - - if (pi && pi->unformat_pg_edit - && unformat_user (input, pi->unformat_pg_edit, s)) - ; - - else if (!unformat_user (input, unformat_pg_payload, s)) - goto error; - - p = pg_get_edit_group (s, group_index); - if (p->checksum.type == PG_EDIT_UNSPECIFIED - || p->length.type == PG_EDIT_UNSPECIFIED) - { - pg_edit_group_t *g = pg_stream_get_group (s, group_index); - g->edit_function = udp_pg_edit_function; - g->edit_function_opaque = 0; - if (p->checksum.type == PG_EDIT_UNSPECIFIED) - g->edit_function_opaque |= UDP_PG_EDIT_CHECKSUM; - if (p->length.type == PG_EDIT_UNSPECIFIED) - g->edit_function_opaque |= UDP_PG_EDIT_LENGTH; - } - - return 1; - } - -error: - /* Free up any edits we may have added. */ - pg_free_edit_group (s); - return 0; -} - - -/* - * fd.io coding-style-patch-verification: ON - * - * Local Variables: - * eval: (c-set-style "gnu") - * End: - */ diff --git a/src/vnet/ipsec/ikev2.c b/src/vnet/ipsec/ikev2.c index 09209334..2c1074d8 100644 --- a/src/vnet/ipsec/ikev2.c +++ b/src/vnet/ipsec/ikev2.c @@ -17,7 +17,7 @@ #include #include #include -#include +#include #include #include #include diff --git a/src/vnet/ipsec/ikev2_cli.c b/src/vnet/ipsec/ikev2_cli.c index 5c88d8d4..05ed4e60 100644 --- a/src/vnet/ipsec/ikev2_cli.c +++ b/src/vnet/ipsec/ikev2_cli.c @@ -16,7 +16,7 @@ #include #include #include -#include +#include #include #include diff --git a/src/vnet/ipsec/ikev2_crypto.c b/src/vnet/ipsec/ikev2_crypto.c index c201d3eb..ca56158f 100644 --- a/src/vnet/ipsec/ikev2_crypto.c +++ b/src/vnet/ipsec/ikev2_crypto.c @@ -17,7 +17,7 @@ #include #include #include -#include +#include #include #include #include diff --git a/src/vnet/lisp-cp/packets.c b/src/vnet/lisp-cp/packets.c index 3a4f421b..f24024f1 100644 --- a/src/vnet/lisp-cp/packets.c +++ b/src/vnet/lisp-cp/packets.c @@ -15,7 +15,7 @@ #include #include -#include +#include /* Returns IP ID for the packet */ /* static u16 ip_id = 0; @@ -141,61 +141,6 @@ pkt_push_udp (vlib_main_t * vm, vlib_buffer_t * b, u16 sp, u16 dp) return uh; } -void * -pkt_push_ipv4 (vlib_main_t * vm, vlib_buffer_t * b, ip4_address_t * src, - ip4_address_t * dst, int proto) -{ - ip4_header_t *ih; - - /* make some room */ - ih = vlib_buffer_push_uninit (b, sizeof (ip4_header_t)); - - ih->ip_version_and_header_length = 0x45; - ih->tos = 0; - ih->length = clib_host_to_net_u16 (vlib_buffer_length_in_chain (vm, b)); - - /* iph->fragment_id = clib_host_to_net_u16(get_IP_ID ()); */ - - /* TODO: decide if we allow fragments in case of control */ - ih->flags_and_fragment_offset = clib_host_to_net_u16 (IP_DF); - ih->ttl = 255; - ih->protocol = proto; - ih->src_address.as_u32 = src->as_u32; - ih->dst_address.as_u32 = dst->as_u32; - - ih->checksum = ip4_header_checksum (ih); - return ih; -} - -void * -pkt_push_ipv6 (vlib_main_t * vm, vlib_buffer_t * b, ip6_address_t * src, - ip6_address_t * dst, int proto) -{ - ip6_header_t *ip6h; - u16 payload_length; - - /* make some room */ - ip6h = vlib_buffer_push_uninit (b, sizeof (ip6_header_t)); - - ip6h->ip_version_traffic_class_and_flow_label = - clib_host_to_net_u32 (0x6 << 28); - - /* calculate ip6 payload length */ - payload_length = vlib_buffer_length_in_chain (vm, b); - payload_length -= sizeof (*ip6h); - - ip6h->payload_length = clib_host_to_net_u16 (payload_length); - - ip6h->hop_limit = 0xff; - ip6h->protocol = proto; - clib_memcpy (ip6h->src_address.as_u8, src->as_u8, - sizeof (ip6h->src_address)); - clib_memcpy (ip6h->dst_address.as_u8, dst->as_u8, - sizeof (ip6h->src_address)); - - return ip6h; -} - void * pkt_push_ip (vlib_main_t * vm, vlib_buffer_t * b, ip_address_t * src, ip_address_t * dst, u32 proto) @@ -210,12 +155,12 @@ pkt_push_ip (vlib_main_t * vm, vlib_buffer_t * b, ip_address_t * src, switch (ip_addr_version (src)) { case IP4: - return pkt_push_ipv4 (vm, b, &ip_addr_v4 (src), &ip_addr_v4 (dst), - proto); + return vlib_buffer_push_ip4 (vm, b, &ip_addr_v4 (src), + &ip_addr_v4 (dst), proto); break; case IP6: - return pkt_push_ipv6 (vm, b, &ip_addr_v6 (src), &ip_addr_v6 (dst), - proto); + return vlib_buffer_push_ip6 (vm, b, &ip_addr_v6 (src), + &ip_addr_v6 (dst), proto); break; } diff --git a/src/vnet/lisp-cp/packets.h b/src/vnet/lisp-cp/packets.h index 212a1d78..f6da3bf4 100644 --- a/src/vnet/lisp-cp/packets.h +++ b/src/vnet/lisp-cp/packets.h @@ -26,51 +26,6 @@ void *pkt_push_udp_and_ip (vlib_main_t * vm, vlib_buffer_t * b, u16 sp, void *pkt_push_ecm_hdr (vlib_buffer_t * b); -always_inline u8 * -vlib_buffer_get_tail (vlib_buffer_t * b) -{ - return b->data + b->current_data + b->current_length; -} - -always_inline void * -vlib_buffer_put_uninit (vlib_buffer_t * b, u8 size) -{ - /* XXX should make sure there's enough space! */ - void *p = vlib_buffer_get_tail (b); - b->current_length += size; - return p; -} - -always_inline void * -vlib_buffer_push_uninit (vlib_buffer_t * b, u8 size) -{ - /* XXX should make sure there's enough space! */ - ASSERT (b->current_data >= size); - b->current_data -= size; - b->current_length += size; - - return vlib_buffer_get_current (b); -} - -always_inline void * -vlib_buffer_make_headroom (vlib_buffer_t * b, u8 size) -{ - /* XXX should make sure there's enough space! */ - b->current_data += size; - return vlib_buffer_get_current (b); -} - -always_inline void * -vlib_buffer_pull (vlib_buffer_t * b, u8 size) -{ - if (b->current_length < size) - return 0; - - void *data = vlib_buffer_get_current (b); - vlib_buffer_advance (b, size); - return data; -} - /* *INDENT-ON* */ /* diff --git a/src/vnet/lisp-gpe/interface.c b/src/vnet/lisp-gpe/interface.c index 13359277..292c7e6a 100644 --- a/src/vnet/lisp-gpe/interface.c +++ b/src/vnet/lisp-gpe/interface.c @@ -23,7 +23,7 @@ #include #include #include -#include +#include #include #include #include diff --git a/src/vnet/lisp-gpe/lisp_gpe.h b/src/vnet/lisp-gpe/lisp_gpe.h index c898a7da..b5a50ec6 100644 --- a/src/vnet/lisp-gpe/lisp_gpe.h +++ b/src/vnet/lisp-gpe/lisp_gpe.h @@ -27,10 +27,12 @@ #include #include #include -#include +#include #include #include #include +#include +#include /** IP4-UDP-LISP encap header */ /* *INDENT-OFF* */ diff --git a/src/vnet/lisp-gpe/lisp_gpe_adjacency.c b/src/vnet/lisp-gpe/lisp_gpe_adjacency.c index 65006b81..dbcf7134 100644 --- a/src/vnet/lisp-gpe/lisp_gpe_adjacency.c +++ b/src/vnet/lisp-gpe/lisp_gpe_adjacency.c @@ -25,6 +25,8 @@ #include #include #include +#include +#include /** * Memory pool of all adjacencies diff --git a/src/vnet/session/application.c b/src/vnet/session/application.c new file mode 100644 index 00000000..a561e7d1 --- /dev/null +++ b/src/vnet/session/application.c @@ -0,0 +1,343 @@ +/* + * Copyright (c) 2017 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +#include +#include + +/* + * Pool from which we allocate all applications + */ +static application_t *app_pool; + +/* + * Hash table of apps by api client index + */ +static uword *app_by_api_client_index; + +int +application_api_queue_is_full (application_t * app) +{ + unix_shared_memory_queue_t *q; + + /* builtin servers are always OK */ + if (app->api_client_index == ~0) + return 0; + + q = vl_api_client_index_to_input_queue (app->api_client_index); + if (!q) + return 1; + + if (q->cursize == q->maxsize) + return 1; + return 0; +} + +static void +application_table_add (application_t * app) +{ + hash_set (app_by_api_client_index, app->api_client_index, app->index); +} + +static void +application_table_del (application_t * app) +{ + hash_unset (app_by_api_client_index, app->api_client_index); +} + +application_t * +application_lookup (u32 api_client_index) +{ + uword *p; + p = hash_get (app_by_api_client_index, api_client_index); + if (p) + return application_get (p[0]); + + return 0; +} + +void +application_del (application_t * app) +{ + session_manager_main_t *smm = vnet_get_session_manager_main (); + api_main_t *am = &api_main; + void *oldheap; + session_manager_t *sm; + + if (app->mode == APP_SERVER) + { + sm = session_manager_get (app->session_manager_index); + session_manager_del (smm, sm); + } + + /* Free the event fifo in the /vpe-api shared-memory segment */ + oldheap = svm_push_data_heap (am->vlib_rp); + if (app->event_queue) + unix_shared_memory_queue_free (app->event_queue); + svm_pop_heap (oldheap); + + application_table_del (app); + + pool_put (app_pool, app); +} + +application_t * +application_new (application_type_t type, session_type_t sst, + u32 api_client_index, u32 flags, session_cb_vft_t * cb_fns) +{ + session_manager_main_t *smm = vnet_get_session_manager_main (); + api_main_t *am = &api_main; + application_t *app; + void *oldheap; + session_manager_t *sm; + + pool_get (app_pool, app); + memset (app, 0, sizeof (*app)); + + /* Allocate event fifo in the /vpe-api shared-memory segment */ + oldheap = svm_push_data_heap (am->vlib_rp); + + /* Allocate server event queue */ + app->event_queue = + unix_shared_memory_queue_init (128 /* nels $$$$ config */ , + sizeof (session_fifo_event_t), + 0 /* consumer pid */ , + 0 + /* (do not) signal when queue non-empty */ + ); + + svm_pop_heap (oldheap); + + /* If a server, allocate session manager */ + if (type == APP_SERVER) + { + pool_get (smm->session_managers, sm); + memset (sm, 0, sizeof (*sm)); + + app->session_manager_index = sm - smm->session_managers; + } + else if (type == APP_CLIENT) + { + /* Allocate connect session manager if needed */ + if (smm->connect_manager_index[sst] == INVALID_INDEX) + connects_session_manager_init (smm, sst); + app->session_manager_index = smm->connect_manager_index[sst]; + } + + app->mode = type; + app->index = application_get_index (app); + app->session_type = sst; + app->api_client_index = api_client_index; + app->flags = flags; + app->cb_fns = *cb_fns; + + /* Add app to lookup by api_client_index table */ + application_table_add (app); + + return app; +} + +application_t * +application_get (u32 index) +{ + return pool_elt_at_index (app_pool, index); +} + +u32 +application_get_index (application_t * app) +{ + return app - app_pool; +} + +int +application_server_init (application_t * server, u32 segment_size, + u32 add_segment_size, u32 rx_fifo_size, + u32 tx_fifo_size, u8 ** segment_name) +{ + session_manager_main_t *smm = vnet_get_session_manager_main (); + session_manager_t *sm; + int rv; + + sm = session_manager_get (server->session_manager_index); + + /* Add first segment */ + if ((rv = session_manager_add_first_segment (smm, sm, segment_size, + segment_name))) + { + return rv; + } + + /* Setup session manager */ + sm->add_segment_size = add_segment_size; + sm->rx_fifo_size = rx_fifo_size; + sm->tx_fifo_size = tx_fifo_size; + sm->add_segment = sm->add_segment_size != 0; + return 0; +} + +u8 * +format_application_server (u8 * s, va_list * args) +{ + application_t *srv = va_arg (*args, application_t *); + int verbose = va_arg (*args, int); + vl_api_registration_t *regp; + stream_session_t *listener; + u8 *server_name, *str, *seg_name; + u32 segment_size; + + if (srv == 0) + { + if (verbose) + s = format (s, "%-40s%-20s%-15s%-15s%-10s", "Connection", "Server", + "Segment", "API Client", "Cookie"); + else + s = format (s, "%-40s%-20s", "Connection", "Server"); + + return s; + } + + regp = vl_api_client_index_to_registration (srv->api_client_index); + if (!regp) + server_name = format (0, "%s%c", regp->name, 0); + else + server_name = regp->name; + + listener = stream_session_listener_get (srv->session_type, + srv->session_index); + str = format (0, "%U", format_stream_session, listener, verbose); + + session_manager_get_segment_info (listener->server_segment_index, &seg_name, + &segment_size); + if (verbose) + { + s = format (s, "%-40s%-20s%-20s%-10d%-10d", str, server_name, + seg_name, srv->api_client_index, srv->accept_cookie); + } + else + s = format (s, "%-40s%-20s", str, server_name); + return s; +} + +u8 * +format_application_client (u8 * s, va_list * args) +{ + application_t *client = va_arg (*args, application_t *); + int verbose = va_arg (*args, int); + stream_session_t *session; + u8 *str, *seg_name; + u32 segment_size; + + if (client == 0) + { + if (verbose) + s = + format (s, "%-40s%-20s%-10s", "Connection", "Segment", + "API Client"); + else + s = format (s, "%-40s", "Connection"); + + return s; + } + + session = stream_session_get (client->session_index, client->thread_index); + str = format (0, "%U", format_stream_session, session, verbose); + + session_manager_get_segment_info (session->server_segment_index, &seg_name, + &segment_size); + if (verbose) + { + s = format (s, "%-40s%-20s%-10d%", str, seg_name, + client->api_client_index); + } + else + s = format (s, "%-40s", str); + return s; +} + +static clib_error_t * +show_app_command_fn (vlib_main_t * vm, unformat_input_t * input, + vlib_cli_command_t * cmd) +{ + application_t *app; + int do_server = 0; + int do_client = 0; + int verbose = 0; + + while (unformat_check_input (input) != UNFORMAT_END_OF_INPUT) + { + if (unformat (input, "server")) + do_server = 1; + else if (unformat (input, "client")) + do_client = 1; + else if (unformat (input, "verbose")) + verbose = 1; + else + break; + } + + if (do_server) + { + if (pool_elts (app_pool)) + { + vlib_cli_output (vm, "%U", format_application_server, + 0 /* header */ , + verbose); + /* *INDENT-OFF* */ + pool_foreach (app, app_pool, + ({ + if (app->mode == APP_SERVER) + vlib_cli_output (vm, "%U", format_application_server, app, + verbose); + })); + /* *INDENT-ON* */ + } + else + vlib_cli_output (vm, "No active server bindings"); + } + + if (do_client) + { + if (pool_elts (app_pool)) + { + vlib_cli_output (vm, "%U", format_application_client, + 0 /* header */ , + verbose); + /* *INDENT-OFF* */ + pool_foreach (app, app_pool, + ({ + if (app->mode == APP_CLIENT) + vlib_cli_output (vm, "%U", format_application_client, app, + verbose); + })); + /* *INDENT-ON* */ + } + else + vlib_cli_output (vm, "No active server bindings"); + } + + return 0; +} + +VLIB_CLI_COMMAND (show_app_command, static) = +{ +.path = "show app",.short_help = + "show app [server|client] [verbose]",.function = show_app_command_fn,}; + +/* + * fd.io coding-style-patch-verification: ON + * + * Local Variables: + * eval: (c-set-style "gnu") + * End: + */ diff --git a/src/vnet/session/application.h b/src/vnet/session/application.h new file mode 100644 index 00000000..027d6967 --- /dev/null +++ b/src/vnet/session/application.h @@ -0,0 +1,120 @@ +/* + * Copyright (c) 2017 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +#ifndef SRC_VNET_SESSION_APPLICATION_H_ +#define SRC_VNET_SESSION_APPLICATION_H_ + +#include +#include + +typedef enum +{ + APP_SERVER, + APP_CLIENT +} application_type_t; + +typedef struct _stream_session_cb_vft +{ + /** Notify server of new segment */ + int (*add_segment_callback) (u32 api_client_index, const u8 * seg_name, + u32 seg_size); + + /** Notify server of newly accepted session */ + int (*session_accept_callback) (stream_session_t * new_session); + + /* Connection request callback */ + int (*session_connected_callback) (u32 api_client_index, + stream_session_t * s, u8 code); + + /** Notify app that session is closing */ + void (*session_disconnect_callback) (stream_session_t * s); + + /** Notify app that session was reset */ + void (*session_reset_callback) (stream_session_t * s); + + /* Direct RX callback, for built-in servers */ + int (*builtin_server_rx_callback) (stream_session_t * session); + + /* Redirect connection to local server */ + int (*redirect_connect_callback) (u32 api_client_index, void *mp); +} session_cb_vft_t; + +typedef struct _application +{ + /** Index in server pool */ + u32 index; + + /** Flags */ + u32 flags; + + /** Binary API connection index, ~0 if internal */ + u32 api_client_index; + + /* */ + u32 api_context; + + /** Application listens for events on this svm queue */ + unix_shared_memory_queue_t *event_queue; + + /** Stream session type */ + u8 session_type; + + /* Stream server mode: accept or connect */ + u8 mode; + + u32 session_manager_index; + + /* + * Bind/Listen specific + */ + + /** Accept cookie, for multiple session flavors ($$$ maybe) */ + u32 accept_cookie; + + /** Index of the listen session or connect session */ + u32 session_index; + + /** Session thread index for client connect sessions */ + u32 thread_index; + + /* + * Callbacks: shoulder-taps for the server/client + */ + session_cb_vft_t cb_fns; +} application_t; + +application_t *application_new (application_type_t type, session_type_t sst, + u32 api_client_index, u32 flags, + session_cb_vft_t * cb_fns); +void application_del (application_t * app); +application_t *application_get (u32 index); +application_t *application_lookup (u32 api_client_index); +u32 application_get_index (application_t * app); + +int +application_server_init (application_t * server, u32 segment_size, + u32 add_segment_size, u32 rx_fifo_size, + u32 tx_fifo_size, u8 ** segment_name); +int application_api_queue_is_full (application_t * app); + +#endif /* SRC_VNET_SESSION_APPLICATION_H_ */ + +/* + * fd.io coding-style-patch-verification: ON + * + * Local Variables: + * eval: (c-set-style "gnu") + * End: + */ diff --git a/src/vnet/session/application_interface.c b/src/vnet/session/application_interface.c new file mode 100644 index 00000000..0ea77fd8 --- /dev/null +++ b/src/vnet/session/application_interface.c @@ -0,0 +1,459 @@ +/* + * Copyright (c) 2016 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ +#include + +#include +#include +#include +#include + +/** @file + VPP's application/session API bind/unbind/connect/disconnect calls +*/ + +static u8 +ip_is_zero (ip46_address_t * ip46_address, u8 is_ip4) +{ + if (is_ip4) + return (ip46_address->ip4.as_u32 == 0); + else + return (ip46_address->as_u64[0] == 0 && ip46_address->as_u64[1] == 0); +} + +static u8 +ip_is_local (ip46_address_t * ip46_address, u8 is_ip4) +{ + fib_node_index_t fei; + fib_entry_flag_t flags; + fib_prefix_t prefix; + + /* Check if requester is local */ + if (is_ip4) + { + prefix.fp_len = 32; + prefix.fp_proto = FIB_PROTOCOL_IP4; + } + else + { + prefix.fp_len = 128; + prefix.fp_proto = FIB_PROTOCOL_IP6; + } + + clib_memcpy (&prefix.fp_addr, ip46_address, sizeof (ip46_address)); + fei = fib_table_lookup (0, &prefix); + flags = fib_entry_get_flags (fei); + + return (flags & FIB_ENTRY_FLAG_LOCAL); +} + +int +api_parse_session_handle (u64 handle, u32 * session_index, u32 * thread_index) +{ + session_manager_main_t *smm = vnet_get_session_manager_main (); + stream_session_t *pool; + + *thread_index = handle & 0xFFFFFFFF; + *session_index = handle >> 32; + + if (*thread_index >= vec_len (smm->sessions)) + return VNET_API_ERROR_INVALID_VALUE; + + pool = smm->sessions[*thread_index]; + + if (pool_is_free_index (pool, *session_index)) + return VNET_API_ERROR_INVALID_VALUE_2; + + return 0; +} + +int +vnet_bind_i (u32 api_client_index, ip46_address_t * ip46, u16 port_host_order, + session_type_t sst, u64 * options, session_cb_vft_t * cb_fns, + application_t ** app, u32 * len_seg_name, char *seg_name) +{ + u8 *segment_name = 0; + application_t *server = 0; + stream_session_t *listener; + u8 is_ip4; + + listener = + stream_session_lookup_listener (ip46, + clib_host_to_net_u16 (port_host_order), + sst); + + if (listener) + return VNET_API_ERROR_ADDRESS_IN_USE; + + if (application_lookup (api_client_index)) + { + clib_warning ("Only one bind supported for now"); + return VNET_API_ERROR_ADDRESS_IN_USE; + } + + is_ip4 = SESSION_TYPE_IP4_UDP == sst || SESSION_TYPE_IP4_TCP == sst; + if (!ip_is_zero (ip46, is_ip4) && !ip_is_local (ip46, is_ip4)) + return VNET_API_ERROR_INVALID_VALUE; + + /* Allocate and initialize stream server */ + server = application_new (APP_SERVER, sst, api_client_index, + options[SESSION_OPTIONS_FLAGS], cb_fns); + + application_server_init (server, options[SESSION_OPTIONS_SEGMENT_SIZE], + options[SESSION_OPTIONS_ADD_SEGMENT_SIZE], + options[SESSION_OPTIONS_RX_FIFO_SIZE], + options[SESSION_OPTIONS_TX_FIFO_SIZE], + &segment_name); + + /* Setup listen path down to transport */ + stream_session_start_listen (server->index, ip46, port_host_order); + + /* + * Return values + */ + + ASSERT (vec_len (segment_name) <= 128); + *len_seg_name = vec_len (segment_name); + memcpy (seg_name, segment_name, *len_seg_name); + *app = server; + + return 0; +} + +int +vnet_unbind_i (u32 api_client_index) +{ + application_t *server; + + /* + * Find the stream_server_t corresponding to the api client + */ + server = application_lookup (api_client_index); + if (!server) + return VNET_API_ERROR_INVALID_VALUE_2; + + /* Clear the listener */ + stream_session_stop_listen (server->index); + application_del (server); + + return 0; +} + +int +vnet_connect_i (u32 api_client_index, u32 api_context, session_type_t sst, + ip46_address_t * ip46, u16 port, u64 * options, void *mp, + session_cb_vft_t * cb_fns) +{ + stream_session_t *listener; + application_t *server, *app; + + /* + * Figure out if connecting to a local server + */ + listener = stream_session_lookup_listener (ip46, + clib_host_to_net_u16 (port), + sst); + if (listener) + { + server = application_get (listener->app_index); + + /* + * Server is willing to have a direct fifo connection created + * instead of going through the state machine, etc. + */ + if (server->flags & SESSION_OPTIONS_FLAGS_USE_FIFO) + return server->cb_fns. + redirect_connect_callback (server->api_client_index, mp); + } + + /* Create client app */ + app = application_new (APP_CLIENT, sst, api_client_index, + options[SESSION_OPTIONS_FLAGS], cb_fns); + + app->api_context = api_context; + + /* + * Not connecting to a local server. Create regular session + */ + stream_session_open (sst, ip46, port, app->index); + + return 0; +} + +/** + * unformat a vnet URI + * + * fifo://name + * tcp://ip46-addr:port + * udp://ip46-addr:port + * + * u8 ip46_address[16]; + * u16 port_in_host_byte_order; + * stream_session_type_t sst; + * u8 *fifo_name; + * + * if (unformat (input, "%U", unformat_vnet_uri, &ip46_address, + * &sst, &port, &fifo_name)) + * etc... + * + */ +uword +unformat_vnet_uri (unformat_input_t * input, va_list * args) +{ + ip46_address_t *address = va_arg (*args, ip46_address_t *); + session_type_t *sst = va_arg (*args, session_type_t *); + u16 *port = va_arg (*args, u16 *); + + if (unformat (input, "tcp://%U/%d", unformat_ip4_address, &address->ip4, + port)) + { + *sst = SESSION_TYPE_IP4_TCP; + return 1; + } + if (unformat (input, "udp://%U/%d", unformat_ip4_address, &address->ip4, + port)) + { + *sst = SESSION_TYPE_IP4_UDP; + return 1; + } + if (unformat (input, "udp://%U/%d", unformat_ip6_address, &address->ip6, + port)) + { + *sst = SESSION_TYPE_IP6_UDP; + return 1; + } + if (unformat (input, "tcp://%U/%d", unformat_ip6_address, &address->ip6, + port)) + { + *sst = SESSION_TYPE_IP6_TCP; + return 1; + } + + return 0; +} + +int +parse_uri (char *uri, session_type_t * sst, ip46_address_t * addr, + u16 * port_number_host_byte_order) +{ + unformat_input_t _input, *input = &_input; + + /* Make sure */ + uri = (char *) format (0, "%s%c", uri, 0); + + /* Parse uri */ + unformat_init_string (input, uri, strlen (uri)); + if (!unformat (input, "%U", unformat_vnet_uri, addr, sst, + port_number_host_byte_order)) + { + unformat_free (input); + return VNET_API_ERROR_INVALID_VALUE; + } + unformat_free (input); + + return 0; +} + +int +vnet_bind_uri (vnet_bind_args_t * a) +{ + application_t *server = 0; + u16 port_host_order; + session_type_t sst = SESSION_N_TYPES; + ip46_address_t ip46; + int rv; + + memset (&ip46, 0, sizeof (ip46)); + rv = parse_uri (a->uri, &sst, &ip46, &port_host_order); + if (rv) + return rv; + + if ((rv = vnet_bind_i (a->api_client_index, &ip46, port_host_order, sst, + a->options, a->session_cb_vft, &server, + &a->segment_name_length, a->segment_name))) + return rv; + + a->server_event_queue_address = (u64) server->event_queue; + return 0; +} + +session_type_t +session_type_from_proto_and_ip (session_api_proto_t proto, u8 is_ip4) +{ + if (proto == SESSION_PROTO_TCP) + { + if (is_ip4) + return SESSION_TYPE_IP4_TCP; + else + return SESSION_TYPE_IP6_TCP; + } + else + { + if (is_ip4) + return SESSION_TYPE_IP4_UDP; + else + return SESSION_TYPE_IP6_UDP; + } + + return SESSION_N_TYPES; +} + +int +vnet_unbind_uri (char *uri, u32 api_client_index) +{ + u16 port_number_host_byte_order; + session_type_t sst = SESSION_N_TYPES; + ip46_address_t ip46_address; + stream_session_t *listener; + int rv; + + rv = parse_uri (uri, &sst, &ip46_address, &port_number_host_byte_order); + if (rv) + return rv; + + listener = + stream_session_lookup_listener (&ip46_address, + clib_host_to_net_u16 + (port_number_host_byte_order), sst); + + if (!listener) + return VNET_API_ERROR_ADDRESS_NOT_IN_USE; + + /* External client? */ + if (api_client_index != ~0) + { + ASSERT (vl_api_client_index_to_registration (api_client_index)); + } + + return vnet_unbind_i (api_client_index); +} + +int +vnet_connect_uri (vnet_connect_args_t * a) +{ + ip46_address_t ip46_address; + u16 port; + session_type_t sst; + application_t *app; + int rv; + + app = application_lookup (a->api_client_index); + if (app) + { + clib_warning ("Already have a connect from this app"); + return VNET_API_ERROR_INVALID_VALUE_2; + } + + /* Parse uri */ + rv = parse_uri (a->uri, &sst, &ip46_address, &port); + if (rv) + return rv; + + return vnet_connect_i (a->api_client_index, a->api_context, sst, + &ip46_address, port, a->options, a->mp, + a->session_cb_vft); +} + +int +vnet_disconnect_session (u32 client_index, u32 session_index, + u32 thread_index) +{ + stream_session_t *session; + + session = stream_session_get (session_index, thread_index); + stream_session_disconnect (session); + + return 0; +} + + +int +vnet_bind (vnet_bind_args_t * a) +{ + application_t *server = 0; + session_type_t sst = SESSION_N_TYPES; + int rv; + + sst = session_type_from_proto_and_ip (a->proto, a->tep.is_ip4); + if ((rv = vnet_bind_i (a->api_client_index, &a->tep.ip, a->tep.port, sst, + a->options, a->session_cb_vft, &server, + &a->segment_name_length, a->segment_name))) + return rv; + + a->server_event_queue_address = (u64) server->event_queue; + a->handle = (u64) a->tep.vrf << 32 | (u64) server->session_index; + return 0; +} + +int +vnet_unbind (vnet_unbind_args_t * a) +{ + application_t *server; + + if (a->api_client_index != ~0) + { + ASSERT (vl_api_client_index_to_registration (a->api_client_index)); + } + + /* Make sure this is the right one */ + server = application_lookup (a->api_client_index); + ASSERT (server->session_index == (0xFFFFFFFF & a->handle)); + + /* TODO use handle to disambiguate namespaces/vrfs */ + return vnet_unbind_i (a->api_client_index); +} + +int +vnet_connect (vnet_connect_args_t * a) +{ + session_type_t sst; + application_t *app; + + app = application_lookup (a->api_client_index); + if (app) + { + clib_warning ("Already have a connect from this app"); + return VNET_API_ERROR_INVALID_VALUE_2; + } + + sst = session_type_from_proto_and_ip (a->proto, a->tep.is_ip4); + return vnet_connect_i (a->api_client_index, a->api_context, sst, &a->tep.ip, + a->tep.port, a->options, a->mp, a->session_cb_vft); +} + +int +vnet_disconnect (vnet_disconnect_args_t * a) +{ + stream_session_t *session; + u32 session_index, thread_index; + + if (api_parse_session_handle (a->handle, &session_index, &thread_index)) + { + clib_warning ("Invalid handle"); + return -1; + } + + session = stream_session_get (session_index, thread_index); + stream_session_disconnect (session); + + return 0; +} + +/* + * fd.io coding-style-patch-verification: ON + * + * Local Variables: + * eval: (c-set-style "gnu") + * End: + */ diff --git a/src/vnet/session/application_interface.h b/src/vnet/session/application_interface.h new file mode 100644 index 00000000..8d87c067 --- /dev/null +++ b/src/vnet/session/application_interface.h @@ -0,0 +1,136 @@ +/* + * Copyright (c) 2016 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ +#ifndef __included_uri_h__ +#define __included_uri_h__ + +#include +#include +#include +#include +#include +#include + +typedef enum _session_api_proto +{ + SESSION_PROTO_TCP, + SESSION_PROTO_UDP +} session_api_proto_t; + +typedef struct _vnet_bind_args_t +{ + union + { + char *uri; + struct + { + transport_endpoint_t tep; + session_api_proto_t proto; + }; + }; + + u32 api_client_index; + u64 *options; + session_cb_vft_t *session_cb_vft; + + /* + * Results + */ + char *segment_name; + u32 segment_name_length; + u64 server_event_queue_address; + u64 handle; +} vnet_bind_args_t; + +typedef struct _vnet_unbind_args_t +{ + union + { + char *uri; + u64 handle; + }; + u32 api_client_index; +} vnet_unbind_args_t; + +typedef struct _vnet_connect_args +{ + union + { + char *uri; + struct + { + transport_endpoint_t tep; + session_api_proto_t proto; + }; + }; + u32 api_client_index; + u32 api_context; + u64 *options; + session_cb_vft_t *session_cb_vft; + + /* Used for redirects */ + void *mp; +} vnet_connect_args_t; + +typedef struct _vnet_disconnect_args_t +{ + u64 handle; + u32 api_client_index; +} vnet_disconnect_args_t; + +/* Bind / connect options */ +typedef enum +{ + SESSION_OPTIONS_FLAGS, + SESSION_OPTIONS_SEGMENT_SIZE, + SESSION_OPTIONS_ADD_SEGMENT_SIZE, + SESSION_OPTIONS_RX_FIFO_SIZE, + SESSION_OPTIONS_TX_FIFO_SIZE, + SESSION_OPTIONS_ACCEPT_COOKIE, + SESSION_OPTIONS_N_OPTIONS +} session_options_index_t; + +/** Server can handle delegated connect requests from local clients */ +#define SESSION_OPTIONS_FLAGS_USE_FIFO (1<<0) + +/** Server wants vpp to add segments when out of memory for fifos */ +#define SESSION_OPTIONS_FLAGS_ADD_SEGMENT (1<<1) + +#define VNET_CONNECT_REDIRECTED 123 + +int vnet_bind_uri (vnet_bind_args_t *); +int vnet_unbind_uri (char *uri, u32 api_client_index); +int vnet_connect_uri (vnet_connect_args_t * a); +int +vnet_disconnect_session (u32 client_index, u32 session_index, + u32 thread_index); + +int vnet_bind (vnet_bind_args_t * a); +int vnet_connect (vnet_connect_args_t * a); +int vnet_unbind (vnet_unbind_args_t * a); +int vnet_disconnect (vnet_disconnect_args_t * a); + +int +api_parse_session_handle (u64 handle, u32 * session_index, + u32 * thread_index); + +#endif /* __included_uri_h__ */ + +/* + * fd.io coding-style-patch-verification: ON + * + * Local Variables: + * eval: (c-set-style "gnu") + * End: + */ diff --git a/src/vnet/session/hashes.c b/src/vnet/session/hashes.c new file mode 100644 index 00000000..1808dd73 --- /dev/null +++ b/src/vnet/session/hashes.c @@ -0,0 +1,28 @@ +/* + * Copyright (c) 2016 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +/** Generate typed init functions for multiple hash table styles... */ + +#include +#include + +#include + +#undef __included_bihash_template_h__ + +#include +#include + +#include diff --git a/src/vnet/session/node.c b/src/vnet/session/node.c new file mode 100644 index 00000000..e467f4e9 --- /dev/null +++ b/src/vnet/session/node.c @@ -0,0 +1,435 @@ +/* + * Copyright (c) 2017 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +#include +#include +#include +#include + +#include + +#include +#include +#include +#include + +#include +#include +#include + +vlib_node_registration_t session_queue_node; + +typedef struct +{ + u32 session_index; + u32 server_thread_index; +} session_queue_trace_t; + +/* packet trace format function */ +static u8 * +format_session_queue_trace (u8 * s, va_list * args) +{ + CLIB_UNUSED (vlib_main_t * vm) = va_arg (*args, vlib_main_t *); + CLIB_UNUSED (vlib_node_t * node) = va_arg (*args, vlib_node_t *); + session_queue_trace_t *t = va_arg (*args, session_queue_trace_t *); + + s = format (s, "SESSION_QUEUE: session index %d, server thread index %d", + t->session_index, t->server_thread_index); + return s; +} + +vlib_node_registration_t session_queue_node; + +#define foreach_session_queue_error \ +_(TX, "Packets transmitted") \ +_(TIMER, "Timer events") + +typedef enum +{ +#define _(sym,str) SESSION_QUEUE_ERROR_##sym, + foreach_session_queue_error +#undef _ + SESSION_QUEUE_N_ERROR, +} session_queue_error_t; + +static char *session_queue_error_strings[] = { +#define _(sym,string) string, + foreach_session_queue_error +#undef _ +}; + +static u32 session_type_to_next[] = { + SESSION_QUEUE_NEXT_TCP_IP4_OUTPUT, + SESSION_QUEUE_NEXT_IP4_LOOKUP, + SESSION_QUEUE_NEXT_TCP_IP6_OUTPUT, + SESSION_QUEUE_NEXT_IP6_LOOKUP, +}; + +always_inline int +session_fifo_rx_i (vlib_main_t * vm, vlib_node_runtime_t * node, + session_manager_main_t * smm, session_fifo_event_t * e0, + stream_session_t * s0, u32 thread_index, int *n_tx_packets, + u8 peek_data) +{ + u32 n_trace = vlib_get_trace_count (vm, node); + u32 left_to_snd0, max_len_to_snd0, len_to_deq0, n_bufs, snd_space0; + u32 n_frame_bytes, n_frames_per_evt; + transport_connection_t *tc0; + transport_proto_vft_t *transport_vft; + u32 next_index, next0, *to_next, n_left_to_next, bi0; + vlib_buffer_t *b0; + u32 rx_offset; + u16 snd_mss0; + u8 *data0; + int i; + + next_index = next0 = session_type_to_next[s0->session_type]; + + transport_vft = session_get_transport_vft (s0->session_type); + tc0 = transport_vft->get_connection (s0->connection_index, thread_index); + + /* Make sure we have space to send and there's something to dequeue */ + snd_space0 = transport_vft->send_space (tc0); + snd_mss0 = transport_vft->send_mss (tc0); + + if (snd_space0 == 0 || svm_fifo_max_dequeue (s0->server_tx_fifo) == 0 + || snd_mss0 == 0) + return 0; + + ASSERT (e0->enqueue_length > 0); + + /* Ensure we're not writing more than transport window allows */ + max_len_to_snd0 = clib_min (e0->enqueue_length, snd_space0); + + if (peek_data) + { + /* Offset in rx fifo from where to peek data */ + rx_offset = transport_vft->rx_fifo_offset (tc0); + } + + /* TODO check if transport is willing to send len_to_snd0 + * bytes (Nagle) */ + + n_frame_bytes = snd_mss0 * VLIB_FRAME_SIZE; + n_frames_per_evt = ceil ((double) max_len_to_snd0 / n_frame_bytes); + + n_bufs = vec_len (smm->tx_buffers[thread_index]); + left_to_snd0 = max_len_to_snd0; + for (i = 0; i < n_frames_per_evt; i++) + { + /* Make sure we have at least one full frame of buffers ready */ + if (PREDICT_FALSE (n_bufs < VLIB_FRAME_SIZE)) + { + vec_validate (smm->tx_buffers[thread_index], + n_bufs + VLIB_FRAME_SIZE - 1); + n_bufs += + vlib_buffer_alloc (vm, &smm->tx_buffers[thread_index][n_bufs], + VLIB_FRAME_SIZE); + + /* buffer shortage + * XXX 0.9 because when debugging we might not get a full frame */ + if (PREDICT_FALSE (n_bufs < 0.9 * VLIB_FRAME_SIZE)) + { + /* Keep track of how much we've dequeued and exit */ + e0->enqueue_length -= max_len_to_snd0 - left_to_snd0; + return -1; + } + + _vec_len (smm->tx_buffers[thread_index]) = n_bufs; + } + + vlib_get_next_frame (vm, node, next_index, to_next, n_left_to_next); + while (left_to_snd0 && n_left_to_next) + { + /* Get free buffer */ + n_bufs--; + bi0 = smm->tx_buffers[thread_index][n_bufs]; + _vec_len (smm->tx_buffers[thread_index]) = n_bufs; + + b0 = vlib_get_buffer (vm, bi0); + b0->error = 0; + b0->flags = VLIB_BUFFER_TOTAL_LENGTH_VALID + | VNET_BUFFER_LOCALLY_ORIGINATED; + b0->current_data = 0; + + /* RX on the local interface. tx in default fib */ + vnet_buffer (b0)->sw_if_index[VLIB_RX] = 0; + vnet_buffer (b0)->sw_if_index[VLIB_TX] = (u32) ~ 0; + + /* usual speculation, or the enqueue_x1 macro will barf */ + to_next[0] = bi0; + to_next += 1; + n_left_to_next -= 1; + + VLIB_BUFFER_TRACE_TRAJECTORY_INIT (b0); + if (PREDICT_FALSE (n_trace > 0)) + { + session_queue_trace_t *t0; + vlib_trace_buffer (vm, node, next_index, b0, + 1 /* follow_chain */ ); + vlib_set_trace_count (vm, node, --n_trace); + t0 = vlib_add_trace (vm, node, b0, sizeof (*t0)); + t0->session_index = s0->session_index; + t0->server_thread_index = s0->thread_index; + } + + if (1) + { + ELOG_TYPE_DECLARE (e) = + { + .format = "evt-dequeue: id %d length %d",.format_args = + "i4i4",}; + struct + { + u32 data[2]; + } *ed; + ed = ELOG_DATA (&vm->elog_main, e); + ed->data[0] = e0->event_id; + ed->data[1] = e0->enqueue_length; + } + + len_to_deq0 = (left_to_snd0 < snd_mss0) ? left_to_snd0 : snd_mss0; + + /* Make room for headers */ + data0 = vlib_buffer_make_headroom (b0, MAX_HDRS_LEN); + + /* Dequeue the data + * TODO 1) peek instead of dequeue + * 2) buffer chains */ + if (peek_data) + { + int n_bytes_read; + n_bytes_read = svm_fifo_peek (s0->server_tx_fifo, s0->pid, + rx_offset, len_to_deq0, data0); + if (n_bytes_read < 0) + goto dequeue_fail; + + /* Keep track of progress locally, transport is also supposed to + * increment it independently when pushing header */ + rx_offset += n_bytes_read; + } + else + { + if (svm_fifo_dequeue_nowait (s0->server_tx_fifo, s0->pid, + len_to_deq0, data0) < 0) + goto dequeue_fail; + } + + b0->current_length = len_to_deq0; + + /* Ask transport to push header */ + transport_vft->push_header (tc0, b0); + + left_to_snd0 -= len_to_deq0; + *n_tx_packets = *n_tx_packets + 1; + + vlib_validate_buffer_enqueue_x1 (vm, node, next_index, + to_next, n_left_to_next, + bi0, next0); + } + vlib_put_next_frame (vm, node, next_index, n_left_to_next); + } + + /* If we couldn't dequeue all bytes store progress */ + if (max_len_to_snd0 < e0->enqueue_length) + { + e0->enqueue_length -= max_len_to_snd0; + vec_add1 (smm->evts_partially_read[thread_index], *e0); + } + return 0; + +dequeue_fail: + /* Can't read from fifo. Store event rx progress, save as partially read, + * return buff to free list and return */ + e0->enqueue_length -= max_len_to_snd0 - left_to_snd0; + vec_add1 (smm->evts_partially_read[thread_index], *e0); + + to_next -= 1; + n_left_to_next += 1; + _vec_len (smm->tx_buffers[thread_index]) += 1; + + clib_warning ("dequeue fail"); + return 0; +} + +int +session_fifo_rx_peek (vlib_main_t * vm, vlib_node_runtime_t * node, + session_manager_main_t * smm, session_fifo_event_t * e0, + stream_session_t * s0, u32 thread_index, int *n_tx_pkts) +{ + return session_fifo_rx_i (vm, node, smm, e0, s0, thread_index, n_tx_pkts, + 1); +} + +int +session_fifo_rx_dequeue (vlib_main_t * vm, vlib_node_runtime_t * node, + session_manager_main_t * smm, + session_fifo_event_t * e0, stream_session_t * s0, + u32 thread_index, int *n_tx_pkts) +{ + return session_fifo_rx_i (vm, node, smm, e0, s0, thread_index, n_tx_pkts, + 0); +} + +static uword +session_queue_node_fn (vlib_main_t * vm, vlib_node_runtime_t * node, + vlib_frame_t * frame) +{ + session_manager_main_t *smm = vnet_get_session_manager_main (); + session_fifo_event_t *my_fifo_events, *e; + u32 n_to_dequeue; + unix_shared_memory_queue_t *q; + int n_tx_packets = 0; + u32 my_thread_index = vm->cpu_index; + int i, rv; + + /* + * Update TCP time + */ + tcp_update_time (vlib_time_now (vm), my_thread_index); + + /* + * Get vpp queue events + */ + q = smm->vpp_event_queues[my_thread_index]; + if (PREDICT_FALSE (q == 0)) + return 0; + + /* min number of events we can dequeue without blocking */ + n_to_dequeue = q->cursize; + if (n_to_dequeue == 0) + return 0; + + my_fifo_events = smm->fifo_events[my_thread_index]; + + /* If we didn't manage to process previous events try going + * over them again without dequeuing new ones. + * XXX: Block senders to sessions that can't keep up */ + if (vec_len (my_fifo_events) >= 100) + goto skip_dequeue; + + /* See you in the next life, don't be late */ + if (pthread_mutex_trylock (&q->mutex)) + return 0; + + for (i = 0; i < n_to_dequeue; i++) + { + vec_add2 (my_fifo_events, e, 1); + unix_shared_memory_queue_sub_raw (q, (u8 *) e); + } + + /* The other side of the connection is not polling */ + if (q->cursize < (q->maxsize / 8)) + (void) pthread_cond_broadcast (&q->condvar); + pthread_mutex_unlock (&q->mutex); + + smm->fifo_events[my_thread_index] = my_fifo_events; + +skip_dequeue: + + for (i = 0; i < n_to_dequeue; i++) + { + svm_fifo_t *f0; /* $$$ prefetch 1 ahead maybe */ + stream_session_t *s0; + u32 server_session_index0, server_thread_index0; + session_fifo_event_t *e0; + + e0 = &my_fifo_events[i]; + f0 = e0->fifo; + server_session_index0 = f0->server_session_index; + server_thread_index0 = f0->server_thread_index; + + /* $$$ add multiple event queues, per vpp worker thread */ + ASSERT (server_thread_index0 == my_thread_index); + + s0 = pool_elt_at_index (smm->sessions[my_thread_index], + server_session_index0); + + ASSERT (s0->thread_index == my_thread_index); + + switch (e0->event_type) + { + case FIFO_EVENT_SERVER_TX: + /* Spray packets in per session type frames, since they go to + * different nodes */ + rv = (smm->session_rx_fns[s0->session_type]) (vm, node, smm, e0, s0, + my_thread_index, + &n_tx_packets); + if (rv < 0) + goto done; + + break; + + default: + clib_warning ("unhandled event type %d", e0->event_type); + } + } + +done: + + /* Couldn't process all events. Probably out of buffers */ + if (PREDICT_FALSE (i < n_to_dequeue)) + { + session_fifo_event_t *partially_read = + smm->evts_partially_read[my_thread_index]; + vec_add (partially_read, &my_fifo_events[i], n_to_dequeue - i); + vec_free (my_fifo_events); + smm->fifo_events[my_thread_index] = partially_read; + smm->evts_partially_read[my_thread_index] = 0; + } + else + { + vec_free (smm->fifo_events[my_thread_index]); + smm->fifo_events[my_thread_index] = + smm->evts_partially_read[my_thread_index]; + smm->evts_partially_read[my_thread_index] = 0; + } + + vlib_node_increment_counter (vm, session_queue_node.index, + SESSION_QUEUE_ERROR_TX, n_tx_packets); + + return n_tx_packets; +} + +/* *INDENT-OFF* */ +VLIB_REGISTER_NODE (session_queue_node) = +{ + .function = session_queue_node_fn, + .name = "session-queue", + .format_trace = format_session_queue_trace, + .type = VLIB_NODE_TYPE_INPUT, + .n_errors = ARRAY_LEN (session_queue_error_strings), + .error_strings = session_queue_error_strings, + .n_next_nodes = SESSION_QUEUE_N_NEXT, + /* .state = VLIB_NODE_STATE_DISABLED, enable on-demand? */ + /* edit / add dispositions here */ + .next_nodes = + { + [SESSION_QUEUE_NEXT_DROP] = "error-drop", + [SESSION_QUEUE_NEXT_IP4_LOOKUP] = "ip4-lookup", + [SESSION_QUEUE_NEXT_IP6_LOOKUP] = "ip6-lookup", + [SESSION_QUEUE_NEXT_TCP_IP4_OUTPUT] = "tcp4-output", + [SESSION_QUEUE_NEXT_TCP_IP6_OUTPUT] = "tcp6-output", + }, +}; +/* *INDENT-ON* */ + +/* + * fd.io coding-style-patch-verification: ON + * + * Local Variables: + * eval: (c-set-style "gnu") + * End: + */ diff --git a/src/vnet/session/session.api b/src/vnet/session/session.api new file mode 100644 index 00000000..a7b28c1d --- /dev/null +++ b/src/vnet/session/session.api @@ -0,0 +1,429 @@ +/* + * Copyright (c) 2015-2016 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + + /** \brief Bind to a given URI + @param client_index - opaque cookie to identify the sender + @param context - sender context, to match reply w/ request + @param accept_cookie - sender accept cookie, to identify this bind flavor + @param uri - a URI, e.g. "tcp://0.0.0.0/0/80" [ipv4] + "tcp://::/0/80" [ipv6] etc. + @param options - socket options, fifo sizes, etc. +*/ +define bind_uri { + u32 client_index; + u32 context; + u32 accept_cookie; + u32 initial_segment_size; + u8 uri[128]; + u64 options[16]; +}; + +/** \brief Unbind a given URI + @param client_index - opaque cookie to identify the sender + @param context - sender context, to match reply w/ request + @param uri - a URI, e.g. "tcp://0.0.0.0/0/80" [ipv4] + "tcp://::/0/80" [ipv6], etc. + @param options - socket options, fifo sizes, etc. +*/ +define unbind_uri { + u32 client_index; + u32 context; + u8 uri[128]; +}; + +/** \brief Connect to a given URI + @param client_index - opaque cookie to identify the sender + @param context - sender context, to match reply w/ request + @param accept_cookie - sender accept cookie, to identify this bind flavor + @param uri - a URI, e.g. "tcp4://0.0.0.0/0/80" + "tcp6://::/0/80" [ipv6], etc. + @param options - socket options, fifo sizes, etc. +*/ +define connect_uri { + u32 client_index; + u32 context; + u8 uri[128]; + u64 client_queue_address; + u64 options[16]; +}; + +/** \brief Bind reply + @param context - sender context, to match reply w/ request + @param retval - return code for the request + @param event_queue_address - vpp event queue address or 0 if this + connection shouldn't send events + @param segment_name_length - length of segment name + @param segment_name - name of segment client needs to attach to +*/ +define bind_uri_reply { + u32 context; + i32 retval; + u64 server_event_queue_address; + u8 segment_name_length; + u32 segment_size; + u8 segment_name[128]; +}; + +/** \brief unbind reply + @param context - sender context, to match reply w/ request + @param retval - return code for the request +*/ +define unbind_uri_reply { + u32 context; + i32 retval; +}; + +/** \brief vpp->client, connect reply + @param context - sender context, to match reply w/ request + @param retval - return code for the request + @param server_rx_fifo - rx (vpp -> vpp-client) fifo address + @param server_tx_fifo - tx (vpp-client -> vpp) fifo address + @param session_index - session index; + @param session_thread_index - session thread index + @param session_type - session thread type + @param vpp_event_queue_address - vpp's event queue address + @param client_event_queue_address - client's event queue address + @param segment_name_length - non-zero if the client needs to attach to + the fifo segment + @param segment_name - set if the client needs to attach to the segment +*/ +define connect_uri_reply { + u32 context; + i32 retval; + u64 server_rx_fifo; + u64 server_tx_fifo; + u32 session_index; + u32 session_thread_index; + u8 session_type; + u64 client_event_queue_address; + u64 vpp_event_queue_address; + u32 segment_size; + u8 segment_name_length; + u8 segment_name[128]; +}; + +/** \brief vpp->client, please map an additional shared memory segment + @param context - sender context, to match reply w/ request + @param segment_name - +*/ +define map_another_segment { + u32 client_index; + u32 context; + u32 segment_size; + u8 segment_name[128]; +}; + +/** \brief client->vpp + @param context - sender context, to match reply w/ request + @param retval - return code for the request +*/ +define map_another_segment_reply { + u32 context; + i32 retval; +}; + +/** \brief vpp->client, accept this session + @param context - sender context, to match reply w/ request + @param accept_cookie - tells client which bind flavor just occurred + @param rx_fifo_address - rx (vpp -> vpp-client) fifo address + @param tx_fifo_address - tx (vpp-client -> vpp) fifo address + @param session_index - index of new session + @param session_thread_index - thread index of new session + @param vpp_event_queue_address - vpp's event queue address + @param session_type - type of session + +*/ +define accept_session { + u32 client_index; + u32 context; + u32 accept_cookie; + u64 server_rx_fifo; + u64 server_tx_fifo; + u32 session_index; + u32 session_thread_index; + u64 vpp_event_queue_address; + u8 session_type; +}; + +/** \brief client->vpp, reply to an accept message + @param context - sender context, to match reply w/ request + @param retval - return code for the request + @param session_index - session index from accept_session / connect_reply + @param session_thread_index - thread index from accept_session / + connect_reply +*/ +define accept_session_reply { + u32 context; + i32 retval; + u8 session_type; + u8 session_thread_index; + u32 session_index; +}; + +/** \brief bidirectional disconnect API + @param client_index - opaque cookie to identify the sender + client to vpp direction only + @param context - sender context, to match reply w/ request + @param session_index - cookie #1 from accept_session / connect_reply + @param session_thread_index - cookie #2 +*/ +define disconnect_session { + u32 client_index; + u32 context; + u32 session_index; + u32 session_thread_index; +}; + +/** \brief bidirectional disconnect reply API + @param client_index - opaque cookie to identify the sender + client to vpp direction only + @param context - sender context, to match reply w/ request + @param retval - return code for the request + @param session_index - session index from accept_session / connect_reply + @param session_thread_index - thread index from accept_session / + connect_reply +*/ +define disconnect_session_reply { + u32 client_index; + u32 context; + i32 retval; + u32 session_index; + u32 session_thread_index; +}; + +/** \brief vpp->client reset session API + @param client_index - opaque cookie to identify the sender + client to vpp direction only + @param context - sender context, to match reply w/ request + @param session_index - session index from accept_session / connect_reply + @param session_thread_index - thread index from accept_session / + connect_reply +*/ +define reset_session { + u32 client_index; + u32 context; + u32 session_index; + u32 session_thread_index; +}; + +/** \brief client->vpp reset session reply + @param client_index - opaque cookie to identify the sender + client to vpp direction only + @param context - sender context, to match reply w/ request + @param retval - return code for the request + @param session_index - session index from accept_session / connect_reply + @param session_thread_index - thread index from accept_session / + connect_reply +*/ +define reset_session_reply { + u32 client_index; + u32 context; + i32 retval; + u32 session_index; + u32 session_thread_index; +}; + +/** \brief Bind to an ip:port pair for a given transport protocol + @param client_index - opaque cookie to identify the sender + @param context - sender context, to match reply w/ request + @param vrf - bind namespace + @param is_ip4 - flag that is 1 if ip address family is IPv4 + @param ip - ip address + @param port - port + @param proto - protocol 0 - TCP 1 - UDP + @param options - socket options, fifo sizes, etc. +*/ +define bind_sock { + u32 client_index; + u32 context; + u32 vrf; + u8 is_ip4; + u8 ip[16]; + u16 port; + u8 proto; + u64 options[16]; +}; + +/** \brief Unbind + @param client_index - opaque cookie to identify the sender + @param context - sender context, to match reply w/ request + @param handle - bind handle obtained from bind reply +*/ +define unbind_sock { + u32 client_index; + u32 context; + u64 handle; +}; + +/** \brief Connect to a remote peer + @param client_index - opaque cookie to identify the sender + @param context - sender context, to match reply w/ request + @param vrf - connection namespace + @param is_ip4 - flag that is 1 if ip address family is IPv4 + @param ip - ip address + @param port - port + @param proto - protocol 0 - TCP 1 - UDP + @param client_queue_address - client's API queue address. Non-zero when + used to perform redirects + @param options - socket options, fifo sizes, etc. +*/ +define connect_sock { + u32 client_index; + u32 context; + u32 vrf; + u8 is_ip4; + u8 ip[16]; + u16 port; + u8 proto; + u64 client_queue_address; + u64 options[16]; +}; + +/** \brief Bind reply + @param context - sender context, to match reply w/ request + @param handle - bind handle + @param retval - return code for the request + @param event_queue_address - vpp event queue address or 0 if this + connection shouldn't send events + @param segment_name_length - length of segment name + @param segment_name - name of segment client needs to attach to +*/ +define bind_sock_reply { + u32 context; + u64 handle; + i32 retval; + u64 server_event_queue_address; + u32 segment_size; + u8 segment_name_length; + u8 segment_name[128]; +}; + +/** \brief unbind reply + @param context - sender context, to match reply w/ request + @param retval - return code for the request +*/ +define unbind_sock_reply { + u32 context; + i32 retval; +}; + +/** \brief vpp/server->client, connect reply + @param context - sender context, to match reply w/ request + @param retval - return code for the request + @param handle - connection handle + @param server_rx_fifo - rx (vpp -> vpp-client) fifo address + @param server_tx_fifo - tx (vpp-client -> vpp) fifo address + @param vpp_event_queue_address - vpp's event queue address + @param client_event_queue_address - client's event queue address + @param segment_name_length - non-zero if the client needs to attach to + the fifo segment + @param segment_name - set if the client needs to attach to the segment +*/ +define connect_sock_reply { + u32 context; + i32 retval; + u64 handle; + u64 server_rx_fifo; + u64 server_tx_fifo; + u64 client_event_queue_address; + u64 vpp_event_queue_address; + u32 segment_size; + u8 segment_name_length; + u8 segment_name[128]; +}; + +/** \brief bidirectional disconnect API + @param client_index - opaque cookie to identify the sender + client to vpp direction only + @param context - sender context, to match reply w/ request + @param handle - session handle obtained through accept/connect +*/ +define disconnect_sock { + u32 client_index; + u32 context; + u64 handle; +}; + +/** \brief bidirectional disconnect reply API + @param client_index - opaque cookie to identify the sender + client to vpp direction only + @param client_context - sender context, to match reply w/ request + @param handle - session handle obtained through accept/connect +*/ +define disconnect_sock_reply { + u32 client_index; + u32 context; + i32 retval; + u64 handle; +}; + +/** \brief vpp->client, accept this session + @param context - sender context, to match reply w/ request + @param accept_cookie - tells client which bind flavor just occurred + @param handle - session handle obtained through accept/connect + @param rx_fifo_address - rx (vpp -> vpp-client) fifo address + @param tx_fifo_address - tx (vpp-client -> vpp) fifo address + @param vpp_event_queue_address - vpp's event queue address +*/ +define accept_sock { + u32 client_index; + u32 context; + u32 accept_cookie; + u64 handle; + u64 server_rx_fifo; + u64 server_tx_fifo; + u64 vpp_event_queue_address; +}; + +/** \brief client->vpp, reply to an accept message + @param context - sender context, to match reply w/ request + @param retval - return code for the request + @param handle - session handle obtained through accept/connect +*/ +define accept_sock_reply { + u32 context; + i32 retval; + u64 handle; +}; + +/** \brief vpp->client reset session API + @param client_index - opaque cookie to identify the sender + client to vpp direction only + @param context - sender context, to match reply w/ request + @param handle - session handle obtained through accept/connect +*/ +define reset_sock { + u32 client_index; + u32 context; + u64 handle; +}; + +/** \brief client->vpp reset session reply + @param client_index - opaque cookie to identify the sender + client to vpp direction only + @param context - sender context, to match reply w/ request + @param handle - session handle obtained through accept/connect +*/ +define reset_sock_reply { + u32 client_index; + u32 context; + i32 retval; + u64 handle; +}; +/* + * Local Variables: + * eval: (c-set-style "gnu") + * End: + */ \ No newline at end of file diff --git a/src/vnet/session/session.c b/src/vnet/session/session.c new file mode 100644 index 00000000..539da613 --- /dev/null +++ b/src/vnet/session/session.c @@ -0,0 +1,1286 @@ +/* + * Copyright (c) 2017 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ +/** + * @file + * @brief Session and session manager + */ + +#include +#include +#include +#include +#include + +/** + * Per-type vector of transport protocol virtual function tables + */ +static transport_proto_vft_t *tp_vfts; + +session_manager_main_t session_manager_main; + +/* + * Session lookup key; (src-ip, dst-ip, src-port, dst-port, session-type) + * Value: (owner thread index << 32 | session_index); + */ +static void +stream_session_table_add_for_tc (u8 sst, transport_connection_t * tc, + u64 value) +{ + session_manager_main_t *smm = &session_manager_main; + session_kv4_t kv4; + session_kv6_t kv6; + + switch (sst) + { + case SESSION_TYPE_IP4_UDP: + case SESSION_TYPE_IP4_TCP: + make_v4_ss_kv_from_tc (&kv4, tc); + kv4.value = value; + clib_bihash_add_del_16_8 (&smm->v4_session_hash, &kv4, 1 /* is_add */ ); + break; + case SESSION_TYPE_IP6_UDP: + case SESSION_TYPE_IP6_TCP: + make_v6_ss_kv_from_tc (&kv6, tc); + kv6.value = value; + clib_bihash_add_del_48_8 (&smm->v6_session_hash, &kv6, 1 /* is_add */ ); + break; + default: + clib_warning ("Session type not supported"); + ASSERT (0); + } +} + +void +stream_session_table_add (session_manager_main_t * smm, stream_session_t * s, + u64 value) +{ + transport_connection_t *tc; + + tc = tp_vfts[s->session_type].get_connection (s->connection_index, + s->thread_index); + stream_session_table_add_for_tc (s->session_type, tc, value); +} + +static void +stream_session_half_open_table_add (u8 sst, transport_connection_t * tc, + u64 value) +{ + session_manager_main_t *smm = &session_manager_main; + session_kv4_t kv4; + session_kv6_t kv6; + + switch (sst) + { + case SESSION_TYPE_IP4_UDP: + case SESSION_TYPE_IP4_TCP: + make_v4_ss_kv_from_tc (&kv4, tc); + kv4.value = value; + clib_bihash_add_del_16_8 (&smm->v4_half_open_hash, &kv4, + 1 /* is_add */ ); + break; + case SESSION_TYPE_IP6_UDP: + case SESSION_TYPE_IP6_TCP: + make_v6_ss_kv_from_tc (&kv6, tc); + kv6.value = value; + clib_bihash_add_del_48_8 (&smm->v6_half_open_hash, &kv6, + 1 /* is_add */ ); + break; + default: + clib_warning ("Session type not supported"); + ASSERT (0); + } +} + +static int +stream_session_table_del_for_tc (session_manager_main_t * smm, u8 sst, + transport_connection_t * tc) +{ + session_kv4_t kv4; + session_kv6_t kv6; + + switch (sst) + { + case SESSION_TYPE_IP4_UDP: + case SESSION_TYPE_IP4_TCP: + make_v4_ss_kv_from_tc (&kv4, tc); + return clib_bihash_add_del_16_8 (&smm->v4_session_hash, &kv4, + 0 /* is_add */ ); + break; + case SESSION_TYPE_IP6_UDP: + case SESSION_TYPE_IP6_TCP: + make_v6_ss_kv_from_tc (&kv6, tc); + return clib_bihash_add_del_48_8 (&smm->v6_session_hash, &kv6, + 0 /* is_add */ ); + break; + default: + clib_warning ("Session type not supported"); + ASSERT (0); + } + + return 0; +} + +static int +stream_session_table_del (session_manager_main_t * smm, stream_session_t * s) +{ + transport_connection_t *ts; + + ts = tp_vfts[s->session_type].get_connection (s->connection_index, + s->thread_index); + return stream_session_table_del_for_tc (smm, s->session_type, ts); +} + +static void +stream_session_half_open_table_del (session_manager_main_t * smm, u8 sst, + transport_connection_t * tc) +{ + session_kv4_t kv4; + session_kv6_t kv6; + + switch (sst) + { + case SESSION_TYPE_IP4_UDP: + case SESSION_TYPE_IP4_TCP: + make_v4_ss_kv_from_tc (&kv4, tc); + clib_bihash_add_del_16_8 (&smm->v4_half_open_hash, &kv4, + 0 /* is_add */ ); + break; + case SESSION_TYPE_IP6_UDP: + case SESSION_TYPE_IP6_TCP: + make_v6_ss_kv_from_tc (&kv6, tc); + clib_bihash_add_del_48_8 (&smm->v6_half_open_hash, &kv6, + 0 /* is_add */ ); + break; + default: + clib_warning ("Session type not supported"); + ASSERT (0); + } +} + +stream_session_t * +stream_session_lookup_listener4 (ip4_address_t * lcl, u16 lcl_port, u8 proto) +{ + session_manager_main_t *smm = &session_manager_main; + session_kv4_t kv4; + int rv; + + make_v4_listener_kv (&kv4, lcl, lcl_port, proto); + rv = clib_bihash_search_inline_16_8 (&smm->v4_session_hash, &kv4); + if (rv == 0) + return pool_elt_at_index (smm->listen_sessions[proto], (u32) kv4.value); + + /* Zero out the lcl ip */ + kv4.key[0] = 0; + rv = clib_bihash_search_inline_16_8 (&smm->v4_session_hash, &kv4); + if (rv == 0) + return pool_elt_at_index (smm->listen_sessions[proto], kv4.value); + + return 0; +} + +/** Looks up a session based on the 5-tuple passed as argument. + * + * First it tries to find an established session, if this fails, it tries + * finding a listener session if this fails, it tries a lookup with a + * wildcarded local source (listener bound to all interfaces) + */ +stream_session_t * +stream_session_lookup4 (ip4_address_t * lcl, ip4_address_t * rmt, + u16 lcl_port, u16 rmt_port, u8 proto, + u32 my_thread_index) +{ + session_manager_main_t *smm = &session_manager_main; + session_kv4_t kv4; + int rv; + + /* Lookup session amongst established ones */ + make_v4_ss_kv (&kv4, lcl, rmt, lcl_port, rmt_port, proto); + rv = clib_bihash_search_inline_16_8 (&smm->v4_session_hash, &kv4); + if (rv == 0) + return stream_session_get_tsi (kv4.value, my_thread_index); + + /* If nothing is found, check if any listener is available */ + return stream_session_lookup_listener4 (lcl, lcl_port, proto); +} + +stream_session_t * +stream_session_lookup_listener6 (ip6_address_t * lcl, u16 lcl_port, u8 proto) +{ + session_manager_main_t *smm = &session_manager_main; + session_kv6_t kv6; + int rv; + + make_v6_listener_kv (&kv6, lcl, lcl_port, proto); + rv = clib_bihash_search_inline_48_8 (&smm->v6_session_hash, &kv6); + if (rv == 0) + return pool_elt_at_index (smm->listen_sessions[proto], kv6.value); + + /* Zero out the lcl ip */ + kv6.key[0] = kv6.key[1] = 0; + rv = clib_bihash_search_inline_48_8 (&smm->v6_session_hash, &kv6); + if (rv == 0) + return pool_elt_at_index (smm->listen_sessions[proto], kv6.value); + + return 0; +} + +/* Looks up a session based on the 5-tuple passed as argument. + * First it tries to find an established session, if this fails, it tries + * finding a listener session if this fails, it tries a lookup with a + * wildcarded local source (listener bound to all interfaces) */ +stream_session_t * +stream_session_lookup6 (ip6_address_t * lcl, ip6_address_t * rmt, + u16 lcl_port, u16 rmt_port, u8 proto, + u32 my_thread_index) +{ + session_manager_main_t *smm = vnet_get_session_manager_main (); + session_kv6_t kv6; + int rv; + + make_v6_ss_kv (&kv6, lcl, rmt, lcl_port, rmt_port, proto); + rv = clib_bihash_search_inline_48_8 (&smm->v6_session_hash, &kv6); + if (rv == 0) + return stream_session_get_tsi (kv6.value, my_thread_index); + + /* If nothing is found, check if any listener is available */ + return stream_session_lookup_listener6 (lcl, lcl_port, proto); +} + +stream_session_t * +stream_session_lookup_listener (ip46_address_t * lcl, u16 lcl_port, u8 proto) +{ + switch (proto) + { + case SESSION_TYPE_IP4_UDP: + case SESSION_TYPE_IP4_TCP: + return stream_session_lookup_listener4 (&lcl->ip4, lcl_port, proto); + break; + case SESSION_TYPE_IP6_UDP: + case SESSION_TYPE_IP6_TCP: + return stream_session_lookup_listener6 (&lcl->ip6, lcl_port, proto); + break; + } + return 0; +} + +static u64 +stream_session_half_open_lookup (session_manager_main_t * smm, + ip46_address_t * lcl, ip46_address_t * rmt, + u16 lcl_port, u16 rmt_port, u8 proto) +{ + session_kv4_t kv4; + session_kv6_t kv6; + int rv; + + switch (proto) + { + case SESSION_TYPE_IP4_UDP: + case SESSION_TYPE_IP4_TCP: + make_v4_ss_kv (&kv4, &lcl->ip4, &rmt->ip4, lcl_port, rmt_port, proto); + rv = clib_bihash_search_inline_16_8 (&smm->v4_half_open_hash, &kv4); + + if (rv == 0) + return kv4.value; + + return (u64) ~ 0; + break; + case SESSION_TYPE_IP6_UDP: + case SESSION_TYPE_IP6_TCP: + make_v6_ss_kv (&kv6, &lcl->ip6, &rmt->ip6, lcl_port, rmt_port, proto); + rv = clib_bihash_search_inline_48_8 (&smm->v6_half_open_hash, &kv6); + + if (rv == 0) + return kv6.value; + + return (u64) ~ 0; + break; + } + return 0; +} + +transport_connection_t * +stream_session_lookup_transport4 (session_manager_main_t * smm, + ip4_address_t * lcl, ip4_address_t * rmt, + u16 lcl_port, u16 rmt_port, u8 proto, + u32 my_thread_index) +{ + session_kv4_t kv4; + stream_session_t *s; + int rv; + + /* Lookup session amongst established ones */ + make_v4_ss_kv (&kv4, lcl, rmt, lcl_port, rmt_port, proto); + rv = clib_bihash_search_inline_16_8 (&smm->v4_session_hash, &kv4); + if (rv == 0) + { + s = stream_session_get_tsi (kv4.value, my_thread_index); + + return tp_vfts[s->session_type].get_connection (s->connection_index, + my_thread_index); + } + + /* If nothing is found, check if any listener is available */ + s = stream_session_lookup_listener4 (lcl, lcl_port, proto); + if (s) + return tp_vfts[s->session_type].get_listener (s->connection_index); + + /* Finally, try half-open connections */ + rv = clib_bihash_search_inline_16_8 (&smm->v4_half_open_hash, &kv4); + if (rv == 0) + return tp_vfts[proto].get_half_open (kv4.value & 0xFFFFFFFF); + + return 0; +} + +transport_connection_t * +stream_session_lookup_transport6 (session_manager_main_t * smm, + ip6_address_t * lcl, ip6_address_t * rmt, + u16 lcl_port, u16 rmt_port, u8 proto, + u32 my_thread_index) +{ + stream_session_t *s; + session_kv6_t kv6; + int rv; + + make_v6_ss_kv (&kv6, lcl, rmt, lcl_port, rmt_port, proto); + rv = clib_bihash_search_inline_48_8 (&smm->v6_session_hash, &kv6); + if (rv == 0) + { + s = stream_session_get_tsi (kv6.value, my_thread_index); + + return tp_vfts[s->session_type].get_connection (s->connection_index, + my_thread_index); + } + + /* If nothing is found, check if any listener is available */ + s = stream_session_lookup_listener6 (lcl, lcl_port, proto); + if (s) + return tp_vfts[s->session_type].get_listener (s->connection_index); + + /* Finally, try half-open connections */ + rv = clib_bihash_search_inline_48_8 (&smm->v6_half_open_hash, &kv6); + if (rv == 0) + return tp_vfts[s->session_type].get_half_open (kv6.value & 0xFFFFFFFF); + + return 0; +} + +/** + * Allocate vpp event queue (once) per worker thread + */ +void +vpp_session_event_queue_allocate (session_manager_main_t * smm, + u32 thread_index) +{ + api_main_t *am = &api_main; + void *oldheap; + + if (smm->vpp_event_queues[thread_index] == 0) + { + /* Allocate event fifo in the /vpe-api shared-memory segment */ + oldheap = svm_push_data_heap (am->vlib_rp); + + smm->vpp_event_queues[thread_index] = + unix_shared_memory_queue_init (2048 /* nels $$$$ config */ , + sizeof (session_fifo_event_t), + 0 /* consumer pid */ , + 0 + /* (do not) send signal when queue non-empty */ + ); + + svm_pop_heap (oldheap); + } +} + +void +session_manager_get_segment_info (u32 index, u8 ** name, u32 * size) +{ + svm_fifo_segment_private_t *s; + s = svm_fifo_get_segment (index); + *name = s->h->segment_name; + *size = s->ssvm.ssvm_size; +} + +always_inline int +session_manager_add_segment_i (session_manager_main_t * smm, + session_manager_t * sm, + u32 segment_size, u8 * segment_name) +{ + svm_fifo_segment_create_args_t _ca, *ca = &_ca; + int rv; + + memset (ca, 0, sizeof (*ca)); + + ca->segment_name = (char *) segment_name; + ca->segment_size = segment_size; + + rv = svm_fifo_segment_create (ca); + if (rv) + { + clib_warning ("svm_fifo_segment_create ('%s', %d) failed", + ca->segment_name, ca->segment_size); + vec_free (segment_name); + return -1; + } + + vec_add1 (sm->segment_indices, ca->new_segment_index); + + return 0; +} + +static int +session_manager_add_segment (session_manager_main_t * smm, + session_manager_t * sm) +{ + u8 *segment_name; + svm_fifo_segment_create_args_t _ca, *ca = &_ca; + u32 add_segment_size; + u32 default_segment_size = 128 << 10; + + memset (ca, 0, sizeof (*ca)); + segment_name = format (0, "%d-%d%c", getpid (), + smm->unique_segment_name_counter++, 0); + add_segment_size = + sm->add_segment_size ? sm->add_segment_size : default_segment_size; + + return session_manager_add_segment_i (smm, sm, add_segment_size, + segment_name); +} + +int +session_manager_add_first_segment (session_manager_main_t * smm, + session_manager_t * sm, u32 segment_size, + u8 ** segment_name) +{ + svm_fifo_segment_create_args_t _ca, *ca = &_ca; + memset (ca, 0, sizeof (*ca)); + *segment_name = format (0, "%d-%d%c", getpid (), + smm->unique_segment_name_counter++, 0); + return session_manager_add_segment_i (smm, sm, segment_size, *segment_name); +} + +void +session_manager_del (session_manager_main_t * smm, session_manager_t * sm) +{ + u32 *deleted_sessions = 0; + u32 *deleted_thread_indices = 0; + int i, j; + + /* Across all fifo segments used by the server */ + for (j = 0; j < vec_len (sm->segment_indices); j++) + { + svm_fifo_segment_private_t *fifo_segment; + svm_fifo_t **fifos; + /* Vector of fifos allocated in the segment */ + fifo_segment = svm_fifo_get_segment (sm->segment_indices[j]); + fifos = (svm_fifo_t **) fifo_segment->h->fifos; + + /* + * Remove any residual sessions from the session lookup table + * Don't bother deleting the individual fifos, we're going to + * throw away the fifo segment in a minute. + */ + for (i = 0; i < vec_len (fifos); i++) + { + svm_fifo_t *fifo; + u32 session_index, thread_index; + stream_session_t *session; + + fifo = fifos[i]; + session_index = fifo->server_session_index; + thread_index = fifo->server_thread_index; + + session = pool_elt_at_index (smm->sessions[thread_index], + session_index); + + /* Add to the deleted_sessions vector (once!) */ + if (!session->is_deleted) + { + session->is_deleted = 1; + vec_add1 (deleted_sessions, + session - smm->sessions[thread_index]); + vec_add1 (deleted_thread_indices, thread_index); + } + } + + for (i = 0; i < vec_len (deleted_sessions); i++) + { + stream_session_t *session; + + session = + pool_elt_at_index (smm->sessions[deleted_thread_indices[i]], + deleted_sessions[i]); + + /* Instead of directly removing the session call disconnect */ + stream_session_disconnect (session); + + /* + stream_session_table_del (smm, session); + pool_put(smm->sessions[deleted_thread_indices[i]], session); + */ + } + + vec_reset_length (deleted_sessions); + vec_reset_length (deleted_thread_indices); + + /* Instead of removing the segment, test when removing the session if + * the segment can be removed + */ + /* svm_fifo_segment_delete (fifo_segment); */ + } + + vec_free (deleted_sessions); + vec_free (deleted_thread_indices); +} + +int +session_manager_allocate_session_fifos (session_manager_main_t * smm, + session_manager_t * sm, + svm_fifo_t ** server_rx_fifo, + svm_fifo_t ** server_tx_fifo, + u32 * fifo_segment_index, + u8 * added_a_segment) +{ + svm_fifo_segment_private_t *fifo_segment; + u32 fifo_size, default_fifo_size = 8192 /* TODO config */ ; + int i; + + *added_a_segment = 0; + + /* Allocate svm fifos */ + ASSERT (vec_len (sm->segment_indices)); + +again: + for (i = 0; i < vec_len (sm->segment_indices); i++) + { + *fifo_segment_index = sm->segment_indices[i]; + fifo_segment = svm_fifo_get_segment (*fifo_segment_index); + + fifo_size = sm->rx_fifo_size; + fifo_size = (fifo_size == 0) ? default_fifo_size : fifo_size; + *server_rx_fifo = svm_fifo_segment_alloc_fifo (fifo_segment, fifo_size); + + fifo_size = sm->tx_fifo_size; + fifo_size = (fifo_size == 0) ? default_fifo_size : fifo_size; + *server_tx_fifo = svm_fifo_segment_alloc_fifo (fifo_segment, fifo_size); + + if (*server_rx_fifo == 0) + { + /* This would be very odd, but handle it... */ + if (*server_tx_fifo != 0) + { + svm_fifo_segment_free_fifo (fifo_segment, *server_tx_fifo); + *server_tx_fifo = 0; + } + continue; + } + if (*server_tx_fifo == 0) + { + if (*server_rx_fifo != 0) + { + svm_fifo_segment_free_fifo (fifo_segment, *server_rx_fifo); + *server_rx_fifo = 0; + } + continue; + } + break; + } + + /* See if we're supposed to create another segment */ + if (*server_rx_fifo == 0) + { + if (sm->add_segment) + { + if (*added_a_segment) + { + clib_warning ("added a segment, still cant allocate a fifo"); + return SESSION_ERROR_NEW_SEG_NO_SPACE; + } + + if (session_manager_add_segment (smm, sm)) + return VNET_API_ERROR_URI_FIFO_CREATE_FAILED; + + *added_a_segment = 1; + goto again; + } + else + return SESSION_ERROR_NO_SPACE; + } + return 0; +} + +int +stream_session_create_i (session_manager_main_t * smm, application_t * app, + transport_connection_t * tc, + stream_session_t ** ret_s) +{ + int rv; + svm_fifo_t *server_rx_fifo = 0, *server_tx_fifo = 0; + u32 fifo_segment_index; + u32 pool_index, seg_size; + stream_session_t *s; + u64 value; + u32 thread_index = tc->thread_index; + session_manager_t *sm; + u8 segment_added; + u8 *seg_name; + + sm = session_manager_get (app->session_manager_index); + + /* Check the API queue */ + if (app->mode == APP_SERVER && application_api_queue_is_full (app)) + return SESSION_ERROR_API_QUEUE_FULL; + + if ((rv = session_manager_allocate_session_fifos (smm, sm, &server_rx_fifo, + &server_tx_fifo, + &fifo_segment_index, + &segment_added))) + return rv; + + if (segment_added && app->mode == APP_SERVER) + { + /* Send an API message to the external server, to map new segment */ + ASSERT (app->cb_fns.add_segment_callback); + + session_manager_get_segment_info (fifo_segment_index, &seg_name, + &seg_size); + if (app->cb_fns.add_segment_callback (app->api_client_index, seg_name, + seg_size)) + return VNET_API_ERROR_URI_FIFO_CREATE_FAILED; + } + + /* Create the session */ + pool_get (smm->sessions[thread_index], s); + memset (s, 0, sizeof (*s)); + + /* Initialize backpointers */ + pool_index = s - smm->sessions[thread_index]; + server_rx_fifo->server_session_index = pool_index; + server_rx_fifo->server_thread_index = thread_index; + + server_tx_fifo->server_session_index = pool_index; + server_tx_fifo->server_thread_index = thread_index; + + s->server_rx_fifo = server_rx_fifo; + s->server_tx_fifo = server_tx_fifo; + + /* Initialize state machine, such as it is... */ + s->session_type = app->session_type; + s->session_state = SESSION_STATE_CONNECTING; + s->app_index = application_get_index (app); + s->server_segment_index = fifo_segment_index; + s->thread_index = thread_index; + s->session_index = pool_index; + + /* Attach transport to session */ + s->connection_index = tc->c_index; + + /* Attach session to transport */ + tc->s_index = s->session_index; + + /* Add to the main lookup table */ + value = (((u64) thread_index) << 32) | (u64) s->session_index; + stream_session_table_add_for_tc (app->session_type, tc, value); + + *ret_s = s; + + return 0; +} + +/* + * Enqueue data for delivery to session peer. Does not notify peer of enqueue + * event but on request can queue notification events for later delivery by + * calling stream_server_flush_enqueue_events(). + * + * @param tc Transport connection which is to be enqueued data + * @param data Data to be enqueued + * @param len Length of data to be enqueued + * @param queue_event Flag to indicate if peer is to be notified or if event + * is to be queued. The former is useful when more data is + * enqueued and only one event is to be generated. + * @return Number of bytes enqueued or a negative value if enqueueing failed. + */ +int +stream_session_enqueue_data (transport_connection_t * tc, u8 * data, u16 len, + u8 queue_event) +{ + stream_session_t *s; + int enqueued; + + s = stream_session_get (tc->s_index, tc->thread_index); + + /* Make sure there's enough space left. We might've filled the pipes */ + if (PREDICT_FALSE (len > svm_fifo_max_enqueue (s->server_rx_fifo))) + return -1; + + enqueued = svm_fifo_enqueue_nowait (s->server_rx_fifo, s->pid, len, data); + + if (queue_event) + { + /* Queue RX event on this fifo. Eventually these will need to be flushed + * by calling stream_server_flush_enqueue_events () */ + session_manager_main_t *smm = vnet_get_session_manager_main (); + u32 thread_index = s->thread_index; + u32 my_enqueue_epoch = smm->current_enqueue_epoch[thread_index]; + + if (s->enqueue_epoch != my_enqueue_epoch) + { + s->enqueue_epoch = my_enqueue_epoch; + vec_add1 (smm->session_indices_to_enqueue_by_thread[thread_index], + s - smm->sessions[thread_index]); + } + } + + return enqueued; +} + +/** Check if we have space in rx fifo to push more bytes */ +u8 +stream_session_no_space (transport_connection_t * tc, u32 thread_index, + u16 data_len) +{ + stream_session_t *s = stream_session_get (tc->c_index, thread_index); + + if (PREDICT_FALSE (s->session_state != SESSION_STATE_READY)) + return 1; + + if (data_len > svm_fifo_max_enqueue (s->server_rx_fifo)) + return 1; + + return 0; +} + +u32 +stream_session_peek_bytes (transport_connection_t * tc, u8 * buffer, + u32 offset, u32 max_bytes) +{ + stream_session_t *s = stream_session_get (tc->s_index, tc->thread_index); + return svm_fifo_peek (s->server_tx_fifo, s->pid, offset, max_bytes, buffer); +} + +u32 +stream_session_dequeue_drop (transport_connection_t * tc, u32 max_bytes) +{ + stream_session_t *s = stream_session_get (tc->s_index, tc->thread_index); + return svm_fifo_dequeue_drop (s->server_tx_fifo, s->pid, max_bytes); +} + +/** + * Notify session peer that new data has been enqueued. + * + * @param s Stream session for which the event is to be generated. + * @param block Flag to indicate if call should block if event queue is full. + * + * @return 0 on succes or negative number if failed to send notification. + */ +static int +stream_session_enqueue_notify (stream_session_t * s, u8 block) +{ + application_t *app; + session_fifo_event_t evt; + unix_shared_memory_queue_t *q; + static u32 serial_number; + + if (PREDICT_FALSE (s->session_state == SESSION_STATE_CLOSED)) + return 0; + + /* Get session's server */ + app = application_get (s->app_index); + + /* Fabricate event */ + evt.fifo = s->server_rx_fifo; + evt.event_type = FIFO_EVENT_SERVER_RX; + evt.event_id = serial_number++; + evt.enqueue_length = svm_fifo_max_dequeue (s->server_rx_fifo); + + /* Add event to server's event queue */ + q = app->event_queue; + + /* Based on request block (or not) for lack of space */ + if (block || PREDICT_TRUE (q->cursize < q->maxsize)) + unix_shared_memory_queue_add (app->event_queue, (u8 *) & evt, + 0 /* do wait for mutex */ ); + else + return -1; + + if (1) + { + ELOG_TYPE_DECLARE (e) = + { + .format = "evt-enqueue: id %d length %d",.format_args = "i4i4",}; + struct + { + u32 data[2]; + } *ed; + ed = ELOG_DATA (&vlib_global_main.elog_main, e); + ed->data[0] = evt.event_id; + ed->data[1] = evt.enqueue_length; + } + + return 0; +} + +/** + * Flushes queue of sessions that are to be notified of new data + * enqueued events. + * + * @param thread_index Thread index for which the flush is to be performed. + * @return 0 on success or a positive number indicating the number of + * failures due to API queue being full. + */ +int +session_manager_flush_enqueue_events (u32 thread_index) +{ + session_manager_main_t *smm = &session_manager_main; + u32 *session_indices_to_enqueue; + int i, errors = 0; + + session_indices_to_enqueue = + smm->session_indices_to_enqueue_by_thread[thread_index]; + + for (i = 0; i < vec_len (session_indices_to_enqueue); i++) + { + stream_session_t *s0; + + /* Get session */ + s0 = stream_session_get (session_indices_to_enqueue[i], thread_index); + if (stream_session_enqueue_notify (s0, 0 /* don't block */ )) + { + errors++; + } + } + + vec_reset_length (session_indices_to_enqueue); + + smm->session_indices_to_enqueue_by_thread[thread_index] = + session_indices_to_enqueue; + + /* Increment enqueue epoch for next round */ + smm->current_enqueue_epoch[thread_index]++; + + return errors; +} + +/* + * Start listening on server's ip/port pair for requested transport. + * + * Creates a 'dummy' stream session with state LISTENING to be used in session + * lookups, prior to establishing connection. Requests transport to build + * it's own specific listening connection. + */ +int +stream_session_start_listen (u32 server_index, ip46_address_t * ip, u16 port) +{ + session_manager_main_t *smm = &session_manager_main; + stream_session_t *s; + transport_connection_t *tc; + application_t *srv; + u32 tci; + + srv = application_get (server_index); + + pool_get (smm->listen_sessions[srv->session_type], s); + memset (s, 0, sizeof (*s)); + + s->session_type = srv->session_type; + s->session_state = SESSION_STATE_LISTENING; + s->session_index = s - smm->listen_sessions[srv->session_type]; + s->app_index = srv->index; + + /* Transport bind/listen */ + tci = tp_vfts[srv->session_type].bind (smm->vlib_main, s->session_index, ip, + port); + + /* Attach transport to session */ + s->connection_index = tci; + tc = tp_vfts[srv->session_type].get_listener (tci); + + srv->session_index = s->session_index; + + /* Add to the main lookup table */ + stream_session_table_add_for_tc (s->session_type, tc, s->session_index); + + return 0; +} + +void +stream_session_stop_listen (u32 server_index) +{ + session_manager_main_t *smm = &session_manager_main; + stream_session_t *listener; + transport_connection_t *tc; + application_t *srv; + + srv = application_get (server_index); + listener = pool_elt_at_index (smm->listen_sessions[srv->session_type], + srv->session_index); + + tc = tp_vfts[srv->session_type].get_listener (listener->connection_index); + stream_session_table_del_for_tc (smm, listener->session_type, tc); + + tp_vfts[srv->session_type].unbind (smm->vlib_main, + listener->connection_index); + pool_put (smm->listen_sessions[srv->session_type], listener); +} + +int +connect_server_add_segment_cb (application_t * ss, char *segment_name, + u32 segment_size) +{ + /* Does exactly nothing, but die */ + ASSERT (0); + return 0; +} + +void +connects_session_manager_init (session_manager_main_t * smm, u8 session_type) +{ + session_manager_t *sm; + u32 connect_fifo_size = 8 << 10; /* Config? */ + u32 default_segment_size = 1 << 20; + + pool_get (smm->session_managers, sm); + memset (sm, 0, sizeof (*sm)); + + sm->add_segment_size = default_segment_size; + sm->rx_fifo_size = connect_fifo_size; + sm->tx_fifo_size = connect_fifo_size; + sm->add_segment = 1; + + session_manager_add_segment (smm, sm); + smm->connect_manager_index[session_type] = sm - smm->session_managers; +} + +void +stream_session_connect_notify (transport_connection_t * tc, u8 sst, + u8 is_fail) +{ + session_manager_main_t *smm = &session_manager_main; + application_t *app; + stream_session_t *new_s = 0; + u64 value; + + value = stream_session_half_open_lookup (smm, &tc->lcl_ip, &tc->rmt_ip, + tc->lcl_port, tc->rmt_port, + tc->proto); + if (value == HALF_OPEN_LOOKUP_INVALID_VALUE) + { + clib_warning ("This can't be good!"); + return; + } + + app = application_get (value >> 32); + + if (!is_fail) + { + /* Create new session (server segments are allocated if needed) */ + if (stream_session_create_i (smm, app, tc, &new_s)) + return; + + app->session_index = stream_session_get_index (new_s); + app->thread_index = new_s->thread_index; + + /* Allocate vpp event queue for this thread if needed */ + vpp_session_event_queue_allocate (smm, tc->thread_index); + } + + /* Notify client */ + app->cb_fns.session_connected_callback (app->api_client_index, new_s, + is_fail); + + /* Cleanup session lookup */ + stream_session_half_open_table_del (smm, sst, tc); +} + +void +stream_session_accept_notify (transport_connection_t * tc) +{ + application_t *server; + stream_session_t *s; + + s = stream_session_get (tc->s_index, tc->thread_index); + server = application_get (s->app_index); + server->cb_fns.session_accept_callback (s); +} + +/** + * Notification from transport that connection is being closed. + * + * A disconnect is sent to application but state is not removed. Once + * disconnect is acknowledged by application, session disconnect is called. + * Ultimately this leads to close being called on transport (passive close). + */ +void +stream_session_disconnect_notify (transport_connection_t * tc) +{ + application_t *server; + stream_session_t *s; + + s = stream_session_get (tc->s_index, tc->thread_index); + server = application_get (s->app_index); + server->cb_fns.session_disconnect_callback (s); +} + +/** + * Cleans up session and associated app if needed. + */ +void +stream_session_delete (stream_session_t * s) +{ + session_manager_main_t *smm = vnet_get_session_manager_main (); + svm_fifo_segment_private_t *fifo_segment; + application_t *app; + int rv; + + /* delete from the main lookup table */ + rv = stream_session_table_del (smm, s); + + if (rv) + clib_warning ("hash delete error, rv %d", rv); + + /* Cleanup fifo segments */ + fifo_segment = svm_fifo_get_segment (s->server_segment_index); + svm_fifo_segment_free_fifo (fifo_segment, s->server_rx_fifo); + svm_fifo_segment_free_fifo (fifo_segment, s->server_tx_fifo); + + /* Cleanup app if client */ + app = application_get (s->app_index); + if (app->mode == APP_CLIENT) + { + application_del (app); + } + else if (app->mode == APP_SERVER) + { + session_manager_t *sm; + svm_fifo_segment_private_t *fifo_segment; + svm_fifo_t **fifos; + u32 fifo_index; + + sm = session_manager_get (app->session_manager_index); + + /* Delete fifo */ + fifo_segment = svm_fifo_get_segment (s->server_segment_index); + fifos = (svm_fifo_t **) fifo_segment->h->fifos; + + fifo_index = svm_fifo_segment_index (fifo_segment); + + /* Remove segment only if it holds no fifos and not the first */ + if (sm->segment_indices[0] != fifo_index && vec_len (fifos) == 0) + svm_fifo_segment_delete (fifo_segment); + } + + pool_put (smm->sessions[s->thread_index], s); +} + +/** + * Notification from transport that connection is being deleted + * + * This should be called only on previously fully established sessions. For + * instance failed connects should call stream_session_connect_notify and + * indicate that the connect has failed. + */ +void +stream_session_delete_notify (transport_connection_t * tc) +{ + stream_session_t *s; + + s = stream_session_get_if_valid (tc->s_index, tc->thread_index); + if (!s) + { + clib_warning ("Surprised!"); + return; + } + stream_session_delete (s); +} + +/** + * Notify application that connection has been reset. + */ +void +stream_session_reset_notify (transport_connection_t * tc) +{ + stream_session_t *s; + application_t *app; + s = stream_session_get (tc->s_index, tc->thread_index); + + app = application_get (s->app_index); + app->cb_fns.session_reset_callback (s); +} + +/** + * Accept a stream session. Optionally ping the server by callback. + */ +int +stream_session_accept (transport_connection_t * tc, u32 listener_index, + u8 sst, u8 notify) +{ + session_manager_main_t *smm = &session_manager_main; + application_t *server; + stream_session_t *s, *listener; + + int rv; + + /* Find the server */ + listener = pool_elt_at_index (smm->listen_sessions[sst], listener_index); + server = application_get (listener->app_index); + + if ((rv = stream_session_create_i (smm, server, tc, &s))) + return rv; + + /* Allocate vpp event queue for this thread if needed */ + vpp_session_event_queue_allocate (smm, tc->thread_index); + + /* Shoulder-tap the server */ + if (notify) + { + server->cb_fns.session_accept_callback (s); + } + + return 0; +} + +void +stream_session_open (u8 sst, ip46_address_t * addr, u16 port_host_byte_order, + u32 app_index) +{ + transport_connection_t *tc; + u32 tci; + u64 value; + + /* Ask transport to open connection */ + tci = tp_vfts[sst].open (addr, port_host_byte_order); + + /* Get transport connection */ + tc = tp_vfts[sst].get_half_open (tci); + + /* Store api_client_index and transport connection index */ + value = (((u64) app_index) << 32) | (u64) tc->c_index; + + /* Add to the half-open lookup table */ + stream_session_half_open_table_add (sst, tc, value); +} + +/** + * Disconnect session and propagate to transport. This should eventually + * result in a delete notification that allows us to cleanup session state. + * Called for both active/passive disconnects. + */ +void +stream_session_disconnect (stream_session_t * s) +{ + tp_vfts[s->session_type].close (s->connection_index, s->thread_index); + s->session_state = SESSION_STATE_CLOSED; +} + +/** + * Cleanup transport and session state. + */ +void +stream_session_cleanup (stream_session_t * s) +{ + tp_vfts[s->session_type].cleanup (s->connection_index, s->thread_index); + stream_session_delete (s); +} + +void +session_register_transport (u8 type, const transport_proto_vft_t * vft) +{ + session_manager_main_t *smm = vnet_get_session_manager_main (); + + vec_validate (tp_vfts, type); + tp_vfts[type] = *vft; + + /* If an offset function is provided, then peek instead of dequeue */ + smm->session_rx_fns[type] = + (vft->rx_fifo_offset) ? session_fifo_rx_peek : session_fifo_rx_dequeue; +} + +transport_proto_vft_t * +session_get_transport_vft (u8 type) +{ + if (type >= vec_len (tp_vfts)) + return 0; + return &tp_vfts[type]; +} + +static clib_error_t * +session_manager_main_init (vlib_main_t * vm) +{ + u32 num_threads; + vlib_thread_main_t *vtm = vlib_get_thread_main (); + session_manager_main_t *smm = &session_manager_main; + int i; + + smm->vlib_main = vm; + smm->vnet_main = vnet_get_main (); + + num_threads = 1 /* main thread */ + vtm->n_threads; + + if (num_threads < 1) + return clib_error_return (0, "n_thread_stacks not set"); + + /* $$$ config parameters */ + svm_fifo_segment_init (0x200000000ULL /* first segment base VA */ , + 20 /* timeout in seconds */ ); + + /* configure per-thread ** vectors */ + vec_validate (smm->sessions, num_threads - 1); + vec_validate (smm->session_indices_to_enqueue_by_thread, num_threads - 1); + vec_validate (smm->tx_buffers, num_threads - 1); + vec_validate (smm->fifo_events, num_threads - 1); + vec_validate (smm->evts_partially_read, num_threads - 1); + vec_validate (smm->current_enqueue_epoch, num_threads - 1); + vec_validate (smm->vpp_event_queues, num_threads - 1); + + /* $$$$ preallocate hack config parameter */ + for (i = 0; i < 200000; i++) + { + stream_session_t *ss; + pool_get (smm->sessions[0], ss); + memset (ss, 0, sizeof (*ss)); + } + + for (i = 0; i < 200000; i++) + pool_put_index (smm->sessions[0], i); + + clib_bihash_init_16_8 (&smm->v4_session_hash, "v4 session table", + 200000 /* $$$$ config parameter nbuckets */ , + (64 << 20) /*$$$ config parameter table size */ ); + clib_bihash_init_48_8 (&smm->v6_session_hash, "v6 session table", + 200000 /* $$$$ config parameter nbuckets */ , + (64 << 20) /*$$$ config parameter table size */ ); + + clib_bihash_init_16_8 (&smm->v4_half_open_hash, "v4 half-open table", + 200000 /* $$$$ config parameter nbuckets */ , + (64 << 20) /*$$$ config parameter table size */ ); + clib_bihash_init_48_8 (&smm->v6_half_open_hash, "v6 half-open table", + 200000 /* $$$$ config parameter nbuckets */ , + (64 << 20) /*$$$ config parameter table size */ ); + + for (i = 0; i < SESSION_N_TYPES; i++) + smm->connect_manager_index[i] = INVALID_INDEX; + + return 0; +} + +VLIB_INIT_FUNCTION (session_manager_main_init); + +/* + * fd.io coding-style-patch-verification: ON + * + * Local Variables: + * eval: (c-set-style "gnu") + * End: + */ diff --git a/src/vnet/session/session.h b/src/vnet/session/session.h new file mode 100644 index 00000000..cf14cca9 --- /dev/null +++ b/src/vnet/session/session.h @@ -0,0 +1,380 @@ +/* + * Copyright (c) 2017 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ +#ifndef __included_session_h__ +#define __included_session_h__ + +#include +#include +#include +#include +#include + +#define HALF_OPEN_LOOKUP_INVALID_VALUE ((u64)~0) +#define INVALID_INDEX ((u32)~0) + +/* TODO decide how much since we have pre-data as well */ +#define MAX_HDRS_LEN 100 /* Max number of bytes for headers */ + +typedef enum +{ + FIFO_EVENT_SERVER_RX, + FIFO_EVENT_SERVER_TX, + FIFO_EVENT_TIMEOUT, + FIFO_EVENT_SERVER_EXIT, +} fifo_event_type_t; + +#define foreach_session_input_error \ +_(NO_SESSION, "No session drops") \ +_(NO_LISTENER, "No listener for dst port drops") \ +_(ENQUEUED, "Packets pushed into rx fifo") \ +_(NOT_READY, "Session not ready packets") \ +_(FIFO_FULL, "Packets dropped for lack of rx fifo space") \ +_(EVENT_FIFO_FULL, "Events not sent for lack of event fifo space") \ +_(API_QUEUE_FULL, "Sessions not created for lack of API queue space") \ +_(NEW_SEG_NO_SPACE, "Created segment, couldn't allocate a fifo pair") \ +_(NO_SPACE, "Couldn't allocate a fifo pair") + +typedef enum +{ +#define _(sym,str) SESSION_ERROR_##sym, + foreach_session_input_error +#undef _ + SESSION_N_ERROR, +} session_error_t; + +/* Event queue input node static next indices */ +typedef enum +{ + SESSION_QUEUE_NEXT_DROP, + SESSION_QUEUE_NEXT_TCP_IP4_OUTPUT, + SESSION_QUEUE_NEXT_IP4_LOOKUP, + SESSION_QUEUE_NEXT_TCP_IP6_OUTPUT, + SESSION_QUEUE_NEXT_IP6_LOOKUP, + SESSION_QUEUE_N_NEXT, +} session_queue_next_t; + +#define foreach_session_type \ + _(IP4_TCP, ip4_tcp) \ + _(IP4_UDP, ip4_udp) \ + _(IP6_TCP, ip6_tcp) \ + _(IP6_UDP, ip6_udp) + +typedef enum +{ +#define _(A, a) SESSION_TYPE_##A, + foreach_session_type +#undef _ + SESSION_N_TYPES, +} session_type_t; + +/* + * Application session state + */ +typedef enum +{ + SESSION_STATE_LISTENING, + SESSION_STATE_CONNECTING, + SESSION_STATE_READY, + SESSION_STATE_CLOSED, + SESSION_STATE_N_STATES, +} stream_session_state_t; + +typedef CLIB_PACKED (struct + { + svm_fifo_t * fifo; + u8 event_type; + /* $$$$ for event logging */ + u16 event_id; + u32 enqueue_length; + }) session_fifo_event_t; + +typedef struct _stream_session_t +{ + /** Type */ + u8 session_type; + + /** State */ + u8 session_state; + + /** Session index in per_thread pool */ + u32 session_index; + + /** Transport specific */ + u32 connection_index; + + u8 thread_index; + + /** Application specific */ + u32 pid; + + /** fifo pointers. Once allocated, these do not move */ + svm_fifo_t *server_rx_fifo; + svm_fifo_t *server_tx_fifo; + + /** To avoid n**2 "one event per frame" check */ + u8 enqueue_epoch; + + /** used during unbind processing */ + u8 is_deleted; + + /** stream server pool index */ + u32 app_index; + + /** svm segment index */ + u32 server_segment_index; +} stream_session_t; + +typedef struct _session_manager +{ + /** segments mapped by this server */ + u32 *segment_indices; + + /** Session fifo sizes. They are provided for binds and take default + * values for connects */ + u32 rx_fifo_size; + u32 tx_fifo_size; + + /** Configured additional segment size */ + u32 add_segment_size; + + /** Flag that indicates if additional segments should be created */ + u8 add_segment; +} session_manager_t; + +/* Forward definition */ +typedef struct _session_manager_main session_manager_main_t; + +typedef int + (session_fifo_rx_fn) (vlib_main_t * vm, vlib_node_runtime_t * node, + session_manager_main_t * smm, + session_fifo_event_t * e0, stream_session_t * s0, + u32 thread_index, int *n_tx_pkts); + +extern session_fifo_rx_fn session_fifo_rx_peek; +extern session_fifo_rx_fn session_fifo_rx_dequeue; + +struct _session_manager_main +{ + /** Lookup tables for established sessions and listeners */ + clib_bihash_16_8_t v4_session_hash; + clib_bihash_48_8_t v6_session_hash; + + /** Lookup tables for half-open sessions */ + clib_bihash_16_8_t v4_half_open_hash; + clib_bihash_48_8_t v6_half_open_hash; + + /** Per worker thread session pools */ + stream_session_t **sessions; + + /** Pool of listen sessions. Same type as stream sessions to ease lookups */ + stream_session_t *listen_sessions[SESSION_N_TYPES]; + + /** Sparse vector to map dst port to stream server */ + u16 *stream_server_by_dst_port[SESSION_N_TYPES]; + + /** per-worker enqueue epoch counters */ + u8 *current_enqueue_epoch; + + /** Per-worker thread vector of sessions to enqueue */ + u32 **session_indices_to_enqueue_by_thread; + + /** per-worker tx buffer free lists */ + u32 **tx_buffers; + + /** Per worker-thread vector of partially read events */ + session_fifo_event_t **evts_partially_read; + + /** per-worker active event vectors */ + session_fifo_event_t **fifo_events; + + /** vpp fifo event queue */ + unix_shared_memory_queue_t **vpp_event_queues; + + /** Unique segment name counter */ + u32 unique_segment_name_counter; + + /* Connection manager used by incoming connects */ + u32 connect_manager_index[SESSION_N_TYPES]; + + session_manager_t *session_managers; + + /** Per transport rx function that can either dequeue or peek */ + session_fifo_rx_fn *session_rx_fns[SESSION_N_TYPES]; + + /* Convenience */ + vlib_main_t *vlib_main; + vnet_main_t *vnet_main; +}; + +extern session_manager_main_t session_manager_main; + +/* + * Session manager function + */ +always_inline session_manager_main_t * +vnet_get_session_manager_main () +{ + return &session_manager_main; +} + +always_inline session_manager_t * +session_manager_get (u32 index) +{ + return pool_elt_at_index (session_manager_main.session_managers, index); +} + +always_inline unix_shared_memory_queue_t * +session_manager_get_vpp_event_queue (u32 thread_index) +{ + return session_manager_main.vpp_event_queues[thread_index]; +} + +always_inline session_manager_t * +connects_session_manager_get (session_manager_main_t * smm, + session_type_t session_type) +{ + return pool_elt_at_index (smm->session_managers, + smm->connect_manager_index[session_type]); +} + +void session_manager_get_segment_info (u32 index, u8 ** name, u32 * size); +int session_manager_flush_enqueue_events (u32 thread_index); +int +session_manager_add_first_segment (session_manager_main_t * smm, + session_manager_t * sm, u32 segment_size, + u8 ** segment_name); +void +session_manager_del (session_manager_main_t * smm, session_manager_t * sm); +void +connects_session_manager_init (session_manager_main_t * smm, u8 session_type); + +/* + * Stream session functions + */ + +stream_session_t *stream_session_lookup_listener4 (ip4_address_t * lcl, + u16 lcl_port, u8 proto); +stream_session_t *stream_session_lookup4 (ip4_address_t * lcl, + ip4_address_t * rmt, u16 lcl_port, + u16 rmt_port, u8 proto, + u32 thread_index); +stream_session_t *stream_session_lookup_listener6 (ip6_address_t * lcl, + u16 lcl_port, u8 proto); +stream_session_t *stream_session_lookup6 (ip6_address_t * lcl, + ip6_address_t * rmt, u16 lcl_port, + u16 rmt_port, u8, u32 thread_index); +transport_connection_t + * stream_session_lookup_transport4 (session_manager_main_t * smm, + ip4_address_t * lcl, + ip4_address_t * rmt, u16 lcl_port, + u16 rmt_port, u8 proto, + u32 thread_index); +transport_connection_t + * stream_session_lookup_transport6 (session_manager_main_t * smm, + ip6_address_t * lcl, + ip6_address_t * rmt, u16 lcl_port, + u16 rmt_port, u8 proto, + u32 thread_index); +stream_session_t *stream_session_lookup_listener (ip46_address_t * lcl, + u16 lcl_port, u8 proto); + +always_inline stream_session_t * +stream_session_get_tsi (u64 ti_and_si, u32 thread_index) +{ + ASSERT ((u32) (ti_and_si >> 32) == thread_index); + return pool_elt_at_index (session_manager_main.sessions[thread_index], + ti_and_si & 0xFFFFFFFFULL); +} + +always_inline stream_session_t * +stream_session_get (u64 si, u32 thread_index) +{ + return pool_elt_at_index (session_manager_main.sessions[thread_index], si); +} + +always_inline stream_session_t * +stream_session_get_if_valid (u64 si, u32 thread_index) +{ + if (thread_index >= vec_len (session_manager_main.sessions)) + return 0; + + if (pool_is_free_index (session_manager_main.sessions[thread_index], si)) + return 0; + + return pool_elt_at_index (session_manager_main.sessions[thread_index], si); +} + +always_inline stream_session_t * +stream_session_listener_get (u8 sst, u64 si) +{ + return pool_elt_at_index (session_manager_main.listen_sessions[sst], si); +} + +always_inline u32 +stream_session_get_index (stream_session_t * s) +{ + if (s->session_state == SESSION_STATE_LISTENING) + return s - session_manager_main.listen_sessions[s->session_type]; + + return s - session_manager_main.sessions[s->thread_index]; +} + +always_inline u32 +stream_session_max_enqueue (transport_connection_t * tc) +{ + stream_session_t *s = stream_session_get (tc->s_index, tc->thread_index); + return svm_fifo_max_enqueue (s->server_rx_fifo); +} + +int +stream_session_enqueue_data (transport_connection_t * tc, u8 * data, u16 len, + u8 queue_event); +u32 +stream_session_peek_bytes (transport_connection_t * tc, u8 * buffer, + u32 offset, u32 max_bytes); +u32 stream_session_dequeue_drop (transport_connection_t * tc, u32 max_bytes); + +void +stream_session_connect_notify (transport_connection_t * tc, u8 sst, + u8 is_fail); +void stream_session_accept_notify (transport_connection_t * tc); +void stream_session_disconnect_notify (transport_connection_t * tc); +void stream_session_delete_notify (transport_connection_t * tc); +void stream_session_reset_notify (transport_connection_t * tc); +int +stream_session_accept (transport_connection_t * tc, u32 listener_index, + u8 sst, u8 notify); +void stream_session_open (u8 sst, ip46_address_t * addr, + u16 port_host_byte_order, u32 api_client_index); +void stream_session_disconnect (stream_session_t * s); +void stream_session_cleanup (stream_session_t * s); +int +stream_session_start_listen (u32 server_index, ip46_address_t * ip, u16 port); +void stream_session_stop_listen (u32 server_index); + +u8 *format_stream_session (u8 * s, va_list * args); + +void session_register_transport (u8 type, const transport_proto_vft_t * vft); +transport_proto_vft_t *session_get_transport_vft (u8 type); + +#endif /* __included_session_h__ */ + +/* + * fd.io coding-style-patch-verification: ON + * + * Local Variables: + * eval: (c-set-style "gnu") + * End: + */ diff --git a/src/vnet/session/session_api.c b/src/vnet/session/session_api.c new file mode 100644 index 00000000..9d068684 --- /dev/null +++ b/src/vnet/session/session_api.c @@ -0,0 +1,821 @@ +/* + * Copyright (c) 2015-2016 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +#include +#include +#include + +#include +#include "application_interface.h" + +#define vl_typedefs /* define message structures */ +#include +#undef vl_typedefs + +#define vl_endianfun /* define message structures */ +#include +#undef vl_endianfun + +/* instantiate all the print functions we know about */ +#define vl_print(handle, ...) vlib_cli_output (handle, __VA_ARGS__) +#define vl_printfun +#include +#undef vl_printfun + +#include + +#define foreach_session_api_msg \ +_(MAP_ANOTHER_SEGMENT_REPLY, map_another_segment_reply) \ +_(BIND_URI, bind_uri) \ +_(UNBIND_URI, unbind_uri) \ +_(CONNECT_URI, connect_uri) \ +_(DISCONNECT_SESSION, disconnect_session) \ +_(DISCONNECT_SESSION_REPLY, disconnect_session_reply) \ +_(ACCEPT_SESSION_REPLY, accept_session_reply) \ +_(RESET_SESSION_REPLY, reset_session_reply) \ +_(BIND_SOCK, bind_sock) \ +_(UNBIND_SOCK, unbind_sock) \ +_(CONNECT_SOCK, connect_sock) \ +_(DISCONNECT_SOCK, disconnect_sock) \ +_(DISCONNECT_SOCK_REPLY, disconnect_sock_reply) \ +_(ACCEPT_SOCK_REPLY, accept_sock_reply) \ +_(RESET_SOCK_REPLY, reset_sock_reply) \ + +static int +send_add_segment_callback (u32 api_client_index, const u8 * segment_name, + u32 segment_size) +{ + vl_api_map_another_segment_t *mp; + unix_shared_memory_queue_t *q; + + q = vl_api_client_index_to_input_queue (api_client_index); + + if (!q) + return -1; + + mp = vl_msg_api_alloc (sizeof (*mp)); + memset (mp, 0, sizeof (*mp)); + mp->_vl_msg_id = clib_host_to_net_u16 (VL_API_MAP_ANOTHER_SEGMENT); + mp->segment_size = segment_size; + strncpy ((char *) mp->segment_name, (char *) segment_name, + sizeof (mp->segment_name) - 1); + + vl_msg_api_send_shmem (q, (u8 *) & mp); + + return 0; +} + +static int +send_session_accept_uri_callback (stream_session_t * s) +{ + vl_api_accept_session_t *mp; + unix_shared_memory_queue_t *q, *vpp_queue; + application_t *server = application_get (s->app_index); + + q = vl_api_client_index_to_input_queue (server->api_client_index); + vpp_queue = session_manager_get_vpp_event_queue (s->thread_index); + + if (!q) + return -1; + + mp = vl_msg_api_alloc (sizeof (*mp)); + mp->_vl_msg_id = clib_host_to_net_u16 (VL_API_ACCEPT_SESSION); + + /* Note: session_type is the first octet in all types of sessions */ + + mp->accept_cookie = server->accept_cookie; + mp->server_rx_fifo = (u64) s->server_rx_fifo; + mp->server_tx_fifo = (u64) s->server_tx_fifo; + mp->session_thread_index = s->thread_index; + mp->session_index = s->session_index; + mp->session_type = s->session_type; + mp->vpp_event_queue_address = (u64) vpp_queue; + vl_msg_api_send_shmem (q, (u8 *) & mp); + + return 0; +} + +static void +send_session_disconnect_uri_callback (stream_session_t * s) +{ + vl_api_disconnect_session_t *mp; + unix_shared_memory_queue_t *q; + application_t *app = application_get (s->app_index); + + q = vl_api_client_index_to_input_queue (app->api_client_index); + + if (!q) + return; + + mp = vl_msg_api_alloc (sizeof (*mp)); + memset (mp, 0, sizeof (*mp)); + mp->_vl_msg_id = clib_host_to_net_u16 (VL_API_DISCONNECT_SESSION); + + mp->session_thread_index = s->thread_index; + mp->session_index = s->session_index; + vl_msg_api_send_shmem (q, (u8 *) & mp); +} + +static int +send_session_connected_uri_callback (u32 api_client_index, + stream_session_t * s, u8 is_fail) +{ + vl_api_connect_uri_reply_t *mp; + unix_shared_memory_queue_t *q; + application_t *app = application_lookup (api_client_index); + u8 *seg_name; + unix_shared_memory_queue_t *vpp_queue; + + q = vl_api_client_index_to_input_queue (app->api_client_index); + + if (!q) + return -1; + + mp = vl_msg_api_alloc (sizeof (*mp)); + mp->_vl_msg_id = clib_host_to_net_u16 (VL_API_CONNECT_URI_REPLY); + mp->context = app->api_context; + mp->retval = is_fail; + if (!is_fail) + { + vpp_queue = session_manager_get_vpp_event_queue (s->thread_index); + mp->server_rx_fifo = (u64) s->server_rx_fifo; + mp->server_tx_fifo = (u64) s->server_tx_fifo; + mp->session_thread_index = s->thread_index; + mp->session_index = s->session_index; + mp->session_type = s->session_type; + mp->vpp_event_queue_address = (u64) vpp_queue; + mp->client_event_queue_address = (u64) app->event_queue; + + session_manager_get_segment_info (s->server_segment_index, &seg_name, + &mp->segment_size); + mp->segment_name_length = vec_len (seg_name); + if (mp->segment_name_length) + clib_memcpy (mp->segment_name, seg_name, mp->segment_name_length); + } + + vl_msg_api_send_shmem (q, (u8 *) & mp); + + /* Remove client if connect failed */ + if (is_fail) + application_del (app); + + return 0; +} + +/** + * Redirect a connect_uri message to the indicated server. + * Only sent if the server has bound the related port with + * URI_OPTIONS_FLAGS_USE_FIFO + */ +static int +redirect_connect_uri_callback (u32 server_api_client_index, void *mp_arg) +{ + vl_api_connect_uri_t *mp = mp_arg; + unix_shared_memory_queue_t *server_q, *client_q; + vlib_main_t *vm = vlib_get_main (); + f64 timeout = vlib_time_now (vm) + 0.5; + int rv = 0; + + server_q = vl_api_client_index_to_input_queue (server_api_client_index); + + if (!server_q) + { + rv = VNET_API_ERROR_INVALID_VALUE; + goto out; + } + + client_q = vl_api_client_index_to_input_queue (mp->client_index); + if (!client_q) + { + rv = VNET_API_ERROR_INVALID_VALUE_2; + goto out; + } + + /* Tell the server the client's API queue address, so it can reply */ + mp->client_queue_address = (u64) client_q; + + /* + * Bounce message handlers MUST NOT block the data-plane. + * Spin waiting for the queue lock, but + */ + + while (vlib_time_now (vm) < timeout) + { + rv = + unix_shared_memory_queue_add (server_q, (u8 *) & mp, 1 /*nowait */ ); + switch (rv) + { + /* correctly enqueued */ + case 0: + return VNET_CONNECT_REDIRECTED; + + /* continue spinning, wait for pthread_mutex_trylock to work */ + case -1: + continue; + + /* queue stuffed, drop the msg */ + case -2: + rv = VNET_API_ERROR_QUEUE_FULL; + goto out; + } + } +out: + /* Dispose of the message */ + vl_msg_api_free (mp); + return rv; +} + +static u64 +make_session_handle (stream_session_t * s) +{ + return (u64) s->session_index << 32 | (u64) s->thread_index; +} + +static int +send_session_accept_callback (stream_session_t * s) +{ + vl_api_accept_sock_t *mp; + unix_shared_memory_queue_t *q, *vpp_queue; + application_t *server = application_get (s->app_index); + + q = vl_api_client_index_to_input_queue (server->api_client_index); + vpp_queue = session_manager_get_vpp_event_queue (s->thread_index); + + if (!q) + return -1; + + mp = vl_msg_api_alloc (sizeof (*mp)); + mp->_vl_msg_id = clib_host_to_net_u16 (VL_API_ACCEPT_SOCK); + + /* Note: session_type is the first octet in all types of sessions */ + + mp->accept_cookie = server->accept_cookie; + mp->server_rx_fifo = (u64) s->server_rx_fifo; + mp->server_tx_fifo = (u64) s->server_tx_fifo; + mp->handle = make_session_handle (s); + mp->vpp_event_queue_address = (u64) vpp_queue; + vl_msg_api_send_shmem (q, (u8 *) & mp); + + return 0; +} + +static int +send_session_connected_callback (u32 api_client_index, stream_session_t * s, + u8 is_fail) +{ + vl_api_connect_sock_reply_t *mp; + unix_shared_memory_queue_t *q; + application_t *app = application_lookup (api_client_index); + u8 *seg_name; + unix_shared_memory_queue_t *vpp_queue; + + q = vl_api_client_index_to_input_queue (app->api_client_index); + + if (!q) + return -1; + + mp = vl_msg_api_alloc (sizeof (*mp)); + mp->_vl_msg_id = clib_host_to_net_u16 (VL_API_CONNECT_SOCK_REPLY); + mp->context = app->api_context; + mp->retval = is_fail; + if (!is_fail) + { + vpp_queue = session_manager_get_vpp_event_queue (s->thread_index); + mp->server_rx_fifo = (u64) s->server_rx_fifo; + mp->server_tx_fifo = (u64) s->server_tx_fifo; + mp->handle = make_session_handle (s); + mp->vpp_event_queue_address = (u64) vpp_queue; + mp->client_event_queue_address = (u64) app->event_queue; + + session_manager_get_segment_info (s->server_segment_index, &seg_name, + &mp->segment_size); + mp->segment_name_length = vec_len (seg_name); + if (mp->segment_name_length) + clib_memcpy (mp->segment_name, seg_name, mp->segment_name_length); + } + + vl_msg_api_send_shmem (q, (u8 *) & mp); + + /* Remove client if connect failed */ + if (is_fail) + application_del (app); + + return 0; +} + +static void +send_session_disconnect_callback (stream_session_t * s) +{ + vl_api_disconnect_sock_t *mp; + unix_shared_memory_queue_t *q; + application_t *app = application_get (s->app_index); + + q = vl_api_client_index_to_input_queue (app->api_client_index); + + if (!q) + return; + + mp = vl_msg_api_alloc (sizeof (*mp)); + memset (mp, 0, sizeof (*mp)); + mp->_vl_msg_id = clib_host_to_net_u16 (VL_API_DISCONNECT_SOCK); + + mp->handle = make_session_handle (s); + vl_msg_api_send_shmem (q, (u8 *) & mp); +} + +/** + * Redirect a connect_uri message to the indicated server. + * Only sent if the server has bound the related port with + * URI_OPTIONS_FLAGS_USE_FIFO + */ +static int +redirect_connect_callback (u32 server_api_client_index, void *mp_arg) +{ + vl_api_connect_sock_t *mp = mp_arg; + unix_shared_memory_queue_t *server_q, *client_q; + vlib_main_t *vm = vlib_get_main (); + f64 timeout = vlib_time_now (vm) + 0.5; + int rv = 0; + + server_q = vl_api_client_index_to_input_queue (server_api_client_index); + + if (!server_q) + { + rv = VNET_API_ERROR_INVALID_VALUE; + goto out; + } + + client_q = vl_api_client_index_to_input_queue (mp->client_index); + if (!client_q) + { + rv = VNET_API_ERROR_INVALID_VALUE_2; + goto out; + } + + /* Tell the server the client's API queue address, so it can reply */ + mp->client_queue_address = (u64) client_q; + + /* + * Bounce message handlers MUST NOT block the data-plane. + * Spin waiting for the queue lock, but + */ + + while (vlib_time_now (vm) < timeout) + { + rv = + unix_shared_memory_queue_add (server_q, (u8 *) & mp, 1 /*nowait */ ); + switch (rv) + { + /* correctly enqueued */ + case 0: + return VNET_CONNECT_REDIRECTED; + + /* continue spinning, wait for pthread_mutex_trylock to work */ + case -1: + continue; + + /* queue stuffed, drop the msg */ + case -2: + rv = VNET_API_ERROR_QUEUE_FULL; + goto out; + } + } +out: + /* Dispose of the message */ + vl_msg_api_free (mp); + return rv; +} + +static session_cb_vft_t uri_session_cb_vft = { + .session_accept_callback = send_session_accept_uri_callback, + .session_disconnect_callback = send_session_disconnect_uri_callback, + .session_connected_callback = send_session_connected_uri_callback, + .add_segment_callback = send_add_segment_callback, + .redirect_connect_callback = redirect_connect_uri_callback +}; + +static session_cb_vft_t session_cb_vft = { + .session_accept_callback = send_session_accept_callback, + .session_disconnect_callback = send_session_disconnect_callback, + .session_connected_callback = send_session_connected_callback, + .add_segment_callback = send_add_segment_callback, + .redirect_connect_callback = redirect_connect_callback +}; + +static int +api_session_not_valid (u32 session_index, u32 thread_index) +{ + session_manager_main_t *smm = vnet_get_session_manager_main (); + stream_session_t *pool; + + if (thread_index >= vec_len (smm->sessions)) + return VNET_API_ERROR_INVALID_VALUE; + + pool = smm->sessions[thread_index]; + + if (pool_is_free_index (pool, session_index)) + return VNET_API_ERROR_INVALID_VALUE_2; + + return 0; +} + +static void +vl_api_bind_uri_t_handler (vl_api_bind_uri_t * mp) +{ + vl_api_bind_uri_reply_t *rmp; + vnet_bind_args_t _a, *a = &_a; + char segment_name[128]; + u32 segment_name_length; + int rv; + + _Static_assert (sizeof (u64) * SESSION_OPTIONS_N_OPTIONS <= + sizeof (mp->options), + "Out of options, fix api message definition"); + + segment_name_length = ARRAY_LEN (segment_name); + + memset (a, 0, sizeof (*a)); + + a->uri = (char *) mp->uri; + a->api_client_index = mp->client_index; + a->options = mp->options; + a->segment_name = segment_name; + a->segment_name_length = segment_name_length; + a->session_cb_vft = &uri_session_cb_vft; + + a->options[SESSION_OPTIONS_SEGMENT_SIZE] = mp->initial_segment_size; + a->options[SESSION_OPTIONS_ACCEPT_COOKIE] = mp->accept_cookie; + rv = vnet_bind_uri (a); + + /* *INDENT-OFF* */ + REPLY_MACRO2 (VL_API_BIND_URI_REPLY, ({ + rmp->retval = rv; + if (!rv) + { + rmp->segment_name_length = 0; + /* $$$$ policy? */ + rmp->segment_size = mp->initial_segment_size; + if (segment_name_length) + { + memcpy (rmp->segment_name, segment_name, segment_name_length); + rmp->segment_name_length = segment_name_length; + } + rmp->server_event_queue_address = a->server_event_queue_address; + } + })); + /* *INDENT-ON* */ + +} + +static void +vl_api_unbind_uri_t_handler (vl_api_unbind_uri_t * mp) +{ + vl_api_unbind_uri_reply_t *rmp; + int rv; + + rv = vnet_unbind_uri ((char *) mp->uri, mp->client_index); + + REPLY_MACRO (VL_API_UNBIND_URI_REPLY); +} + +static void +vl_api_connect_uri_t_handler (vl_api_connect_uri_t * mp) +{ + vnet_connect_args_t _a, *a = &_a; + + a->uri = (char *) mp->uri; + a->api_client_index = mp->client_index; + a->api_context = mp->context; + a->options = mp->options; + a->session_cb_vft = &uri_session_cb_vft; + a->mp = mp; + vnet_connect_uri (a); +} + +static void +vl_api_disconnect_session_t_handler (vl_api_disconnect_session_t * mp) +{ + vl_api_disconnect_session_reply_t *rmp; + int rv; + + rv = api_session_not_valid (mp->session_index, mp->session_thread_index); + if (!rv) + rv = vnet_disconnect_session (mp->client_index, mp->session_index, + mp->session_thread_index); + + REPLY_MACRO (VL_API_DISCONNECT_SESSION_REPLY); +} + +static void +vl_api_disconnect_session_reply_t_handler (vl_api_disconnect_session_reply_t * + mp) +{ + if (api_session_not_valid (mp->session_index, mp->session_thread_index)) + { + clib_warning ("Invalid session!"); + return; + } + + /* Client objected to disconnecting the session, log and continue */ + if (mp->retval) + { + clib_warning ("client retval %d", mp->retval); + return; + } + + /* Disconnect has been confirmed. Confirm close to transport */ + vnet_disconnect_session (mp->client_index, mp->session_index, + mp->session_thread_index); +} + +static void +vl_api_reset_session_reply_t_handler (vl_api_reset_session_reply_t * mp) +{ + stream_session_t *s; + + if (api_session_not_valid (mp->session_index, mp->session_thread_index)) + { + clib_warning ("Invalid session!"); + return; + } + + /* Client objected to resetting the session, log and continue */ + if (mp->retval) + { + clib_warning ("client retval %d", mp->retval); + return; + } + + s = stream_session_get (mp->session_index, mp->session_thread_index); + + /* This comes as a response to a reset, transport only waiting for + * confirmation to remove connection state, no need to disconnect */ + stream_session_cleanup (s); +} + +static void +vl_api_accept_session_reply_t_handler (vl_api_accept_session_reply_t * mp) +{ + stream_session_t *s; + int rv; + + if (api_session_not_valid (mp->session_index, mp->session_thread_index)) + return; + + s = stream_session_get (mp->session_index, mp->session_thread_index); + rv = mp->retval; + + if (rv) + { + /* Server isn't interested, kill the session */ + stream_session_disconnect (s); + return; + } + + s->session_state = SESSION_STATE_READY; +} + +static void +vl_api_map_another_segment_reply_t_handler (vl_api_map_another_segment_reply_t + * mp) +{ + clib_warning ("not implemented"); +} + +static void +vl_api_bind_sock_t_handler (vl_api_bind_sock_t * mp) +{ + vl_api_bind_sock_reply_t *rmp; + vnet_bind_args_t _a, *a = &_a; + char segment_name[128]; + u32 segment_name_length; + int rv; + + STATIC_ASSERT (sizeof (u64) * SESSION_OPTIONS_N_OPTIONS <= + sizeof (mp->options), + "Out of options, fix api message definition"); + + segment_name_length = ARRAY_LEN (segment_name); + + memset (a, 0, sizeof (*a)); + + clib_memcpy (&a->tep.ip, mp->ip, + (mp->is_ip4 ? sizeof (ip4_address_t) : + sizeof (ip6_address_t))); + a->tep.is_ip4 = mp->is_ip4; + a->tep.port = mp->port; + a->tep.vrf = mp->vrf; + + a->api_client_index = mp->client_index; + a->options = mp->options; + a->segment_name = segment_name; + a->segment_name_length = segment_name_length; + a->session_cb_vft = &session_cb_vft; + + rv = vnet_bind_uri (a); + + /* *INDENT-OFF* */ + REPLY_MACRO2 (VL_API_BIND_SOCK_REPLY, ({ + rmp->retval = rv; + if (!rv) + { + rmp->segment_name_length = 0; + rmp->segment_size = mp->options[SESSION_OPTIONS_SEGMENT_SIZE]; + if (segment_name_length) + { + memcpy(rmp->segment_name, segment_name, segment_name_length); + rmp->segment_name_length = segment_name_length; + } + rmp->server_event_queue_address = a->server_event_queue_address; + } + })); + /* *INDENT-ON* */ +} + +static void +vl_api_unbind_sock_t_handler (vl_api_unbind_sock_t * mp) +{ + vl_api_unbind_sock_reply_t *rmp; + vnet_unbind_args_t _a, *a = &_a; + int rv; + + a->api_client_index = mp->client_index; + a->handle = mp->handle; + + rv = vnet_unbind (a); + + REPLY_MACRO (VL_API_UNBIND_SOCK_REPLY); +} + +static void +vl_api_connect_sock_t_handler (vl_api_connect_sock_t * mp) +{ + vnet_connect_args_t _a, *a = &_a; + + clib_memcpy (&a->tep.ip, mp->ip, + (mp->is_ip4 ? sizeof (ip4_address_t) : + sizeof (ip6_address_t))); + a->tep.is_ip4 = mp->is_ip4; + a->tep.port = mp->port; + a->tep.vrf = mp->vrf; + a->options = mp->options; + a->session_cb_vft = &session_cb_vft; + a->api_context = mp->context; + a->mp = mp; + + vnet_connect (a); +} + +static void +vl_api_disconnect_sock_t_handler (vl_api_disconnect_sock_t * mp) +{ + vnet_disconnect_args_t _a, *a = &_a; + vl_api_disconnect_sock_reply_t *rmp; + int rv; + + a->api_client_index = mp->client_index; + a->handle = mp->handle; + rv = vnet_disconnect (a); + + REPLY_MACRO (VL_API_DISCONNECT_SOCK_REPLY); +} + +static void +vl_api_disconnect_sock_reply_t_handler (vl_api_disconnect_sock_reply_t * mp) +{ + vnet_disconnect_args_t _a, *a = &_a; + + /* Client objected to disconnecting the session, log and continue */ + if (mp->retval) + { + clib_warning ("client retval %d", mp->retval); + return; + } + + a->api_client_index = mp->client_index; + a->handle = mp->handle; + + vnet_disconnect (a); +} + +static void +vl_api_reset_sock_reply_t_handler (vl_api_reset_sock_reply_t * mp) +{ + stream_session_t *s; + u32 session_index, thread_index; + + /* Client objected to resetting the session, log and continue */ + if (mp->retval) + { + clib_warning ("client retval %d", mp->retval); + return; + } + + if (api_parse_session_handle (mp->handle, &session_index, &thread_index)) + { + clib_warning ("Invalid handle"); + return; + } + + s = stream_session_get (session_index, thread_index); + + /* This comes as a response to a reset, transport only waiting for + * confirmation to remove connection state, no need to disconnect */ + stream_session_cleanup (s); +} + +static void +vl_api_accept_sock_reply_t_handler (vl_api_accept_sock_reply_t * mp) +{ + stream_session_t *s; + u32 session_index, thread_index; + + if (api_parse_session_handle (mp->handle, &session_index, &thread_index)) + { + clib_warning ("Invalid handle"); + return; + } + s = stream_session_get (session_index, thread_index); + + if (mp->retval) + { + /* Server isn't interested, kill the session */ + stream_session_disconnect (s); + return; + } + + s->session_state = SESSION_STATE_READY; +} + +#define vl_msg_name_crc_list +#include +#undef vl_msg_name_crc_list + +static void +setup_message_id_table (api_main_t * am) +{ +#define _(id,n,crc) vl_msg_api_add_msg_name_crc (am, #n "_" #crc, id); + foreach_vl_msg_name_crc_session; +#undef _ +} + +/* + * session_api_hookup + * Add uri's API message handlers to the table. + * vlib has alread mapped shared memory and + * added the client registration handlers. + * See .../open-repo/vlib/memclnt_vlib.c:memclnt_process() + */ +static clib_error_t * +session_api_hookup (vlib_main_t * vm) +{ + api_main_t *am = &api_main; + +#define _(N,n) \ + vl_msg_api_set_handlers(VL_API_##N, #n, \ + vl_api_##n##_t_handler, \ + vl_noop_handler, \ + vl_api_##n##_t_endian, \ + vl_api_##n##_t_print, \ + sizeof(vl_api_##n##_t), 1); + foreach_session_api_msg; +#undef _ + + /* + * Messages which bounce off the data-plane to + * an API client. Simply tells the message handling infra not + * to free the message. + * + * Bounced message handlers MUST NOT block the data plane + */ + am->message_bounce[VL_API_CONNECT_URI] = 1; + am->message_bounce[VL_API_CONNECT_SOCK] = 1; + + /* + * Set up the (msg_name, crc, message-id) table + */ + setup_message_id_table (am); + + return 0; +} + +VLIB_API_INIT_FUNCTION (session_api_hookup); +/* + * fd.io coding-style-patch-verification: ON + * + * Local Variables: + * eval: (c-set-style "gnu") + * End: + */ diff --git a/src/vnet/session/session_cli.c b/src/vnet/session/session_cli.c new file mode 100644 index 00000000..b2943a1c --- /dev/null +++ b/src/vnet/session/session_cli.c @@ -0,0 +1,189 @@ +/* + * Copyright (c) 2017 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ +#include +#include + +/** + * Format stream session as per the following format + * + * verbose: + * "Connection", "Rx fifo", "Tx fifo", "Session Index" + * non-verbose: + * "Connection" + */ +u8 * +format_stream_session (u8 * s, va_list * args) +{ + stream_session_t *ss = va_arg (*args, stream_session_t *); + int verbose = va_arg (*args, int); + transport_proto_vft_t *tp_vft; + u8 *str = 0; + + tp_vft = session_get_transport_vft (ss->session_type); + + if (verbose) + str = format (0, "%-20llp%-20llp%-15lld", ss->server_rx_fifo, + ss->server_tx_fifo, stream_session_get_index (ss)); + + if (ss->session_state == SESSION_STATE_READY) + { + s = format (s, "%-40U%v", tp_vft->format_connection, + ss->connection_index, ss->thread_index, str); + } + else if (ss->session_state == SESSION_STATE_LISTENING) + { + s = format (s, "%-40U%v", tp_vft->format_listener, ss->connection_index, + str); + } + else if (ss->session_state == SESSION_STATE_READY) + { + s = + format (s, "%-40U%v", tp_vft->format_half_open, ss->connection_index, + str); + } + else if (ss->session_state == SESSION_STATE_CLOSED) + { + s = format (s, "[CL] %-40U%v", tp_vft->format_connection, + ss->connection_index, ss->thread_index, str); + } + else + { + clib_warning ("Session in unknown state!"); + } + + vec_free (str); + + return s; +} + +static clib_error_t * +show_session_command_fn (vlib_main_t * vm, unformat_input_t * input, + vlib_cli_command_t * cmd) +{ + session_manager_main_t *smm = &session_manager_main; + int verbose = 0, i; + stream_session_t *pool; + stream_session_t *s; + u8 *str = 0; + + while (unformat_check_input (input) != UNFORMAT_END_OF_INPUT) + { + if (unformat (input, "verbose")) + verbose = 1; + else + break; + } + + for (i = 0; i < vec_len (smm->sessions); i++) + { + u32 once_per_pool; + pool = smm->sessions[i]; + + once_per_pool = 1; + + if (pool_elts (pool)) + { + + vlib_cli_output (vm, "Thread %d: %d active sessions", + i, pool_elts (pool)); + if (verbose) + { + if (once_per_pool) + { + str = format (str, "%-40s%-20s%-20s%-15s", + "Connection", "Rx fifo", "Tx fifo", + "Session Index"); + vlib_cli_output (vm, "%v", str); + vec_reset_length (str); + once_per_pool = 0; + } + + /* *INDENT-OFF* */ + pool_foreach (s, pool, + ({ + vlib_cli_output (vm, "%U", format_stream_session, s, verbose); + })); + /* *INDENT-ON* */ + } + } + else + vlib_cli_output (vm, "Thread %d: no active sessions", i); + } + vec_free (str); + + return 0; +} + +VLIB_CLI_COMMAND (show_uri_command, static) = +{ +.path = "show session",.short_help = "show session [verbose]",.function = + show_session_command_fn,}; + + +static clib_error_t * +clear_session_command_fn (vlib_main_t * vm, unformat_input_t * input, + vlib_cli_command_t * cmd) +{ + session_manager_main_t *smm = &session_manager_main; + u32 thread_index = 0; + u32 session_index = ~0; + stream_session_t *pool, *session; + application_t *server; + + while (unformat_check_input (input) != UNFORMAT_END_OF_INPUT) + { + if (unformat (input, "thread %d", &thread_index)) + ; + else if (unformat (input, "session %d", &session_index)) + ; + else + return clib_error_return (0, "unknown input `%U'", + format_unformat_error, input); + } + + if (session_index == ~0) + return clib_error_return (0, "session required, but not set."); + + if (thread_index > vec_len (smm->sessions)) + return clib_error_return (0, "thread %d out of range [0-%d]", + thread_index, vec_len (smm->sessions)); + + pool = smm->sessions[thread_index]; + + if (pool_is_free_index (pool, session_index)) + return clib_error_return (0, "session %d not active", session_index); + + session = pool_elt_at_index (pool, session_index); + server = application_get (session->app_index); + + /* Disconnect both app and transport */ + server->cb_fns.session_disconnect_callback (session); + + return 0; +} + +VLIB_CLI_COMMAND (clear_uri_session_command, static) = +{ +.path = "clear session",.short_help = + "clear session thread session ",.function = + clear_session_command_fn,}; + +/* + * fd.io coding-style-patch-verification: ON + * + * Local Variables: + * eval: (c-set-style "gnu") + * End: + */ diff --git a/src/vnet/session/transport.c b/src/vnet/session/transport.c new file mode 100644 index 00000000..abd94ba4 --- /dev/null +++ b/src/vnet/session/transport.c @@ -0,0 +1,64 @@ +/* + * Copyright (c) 2017 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +#include + +u32 +transport_endpoint_lookup (transport_endpoint_table_t *ht, ip46_address_t *ip, + u16 port) +{ + clib_bihash_kv_24_8_t kv; + int rv; + + kv.key[0] = ip->as_u64[0]; + kv.key[1] = ip->as_u64[1]; + kv.key[2] = port; + + rv = clib_bihash_search_inline_24_8 (ht, &kv); + if (rv == 0) + return kv.value; + + return TRANSPORT_ENDPOINT_INVALID_INDEX; +} + +void +transport_endpoint_table_add (transport_endpoint_table_t *ht, + transport_endpoint_t *te, u32 value) +{ + clib_bihash_kv_24_8_t kv; + + kv.key[0] = te->ip.as_u64[0]; + kv.key[1] = te->ip.as_u64[1]; + kv.key[2] = te->port; + kv.value = value; + + clib_bihash_add_del_24_8 (ht, &kv, 1); +} + +void +transport_endpoint_table_del (transport_endpoint_table_t *ht, + transport_endpoint_t *te) +{ + clib_bihash_kv_24_8_t kv; + + kv.key[0] = te->ip.as_u64[0]; + kv.key[1] = te->ip.as_u64[1]; + kv.key[2] = te->port; + + clib_bihash_add_del_24_8 (ht, &kv, 0); +} + + + diff --git a/src/vnet/session/transport.h b/src/vnet/session/transport.h new file mode 100644 index 00000000..2d4415ba --- /dev/null +++ b/src/vnet/session/transport.h @@ -0,0 +1,250 @@ +/* + * Copyright (c) 2016 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +#ifndef VNET_VNET_URI_TRANSPORT_H_ +#define VNET_VNET_URI_TRANSPORT_H_ + +#include +#include +#include +#include + +/* + * Protocol independent transport properties associated to a session + */ +typedef struct _transport_connection +{ + ip46_address_t rmt_ip; /**< Remote IP */ + ip46_address_t lcl_ip; /**< Local IP */ + u16 lcl_port; /**< Local port */ + u16 rmt_port; /**< Remote port */ + u8 proto; /**< Transport protocol id */ + + u32 s_index; /**< Parent session index */ + u32 c_index; /**< Connection index in transport pool */ + u8 is_ip4; /**< Flag if IP4 connection */ + u32 thread_index; /**< Worker-thread index */ + + /** Macros for 'derived classes' where base is named "connection" */ +#define c_lcl_ip connection.lcl_ip +#define c_rmt_ip connection.rmt_ip +#define c_lcl_ip4 connection.lcl_ip.ip4 +#define c_rmt_ip4 connection.rmt_ip.ip4 +#define c_lcl_ip6 connection.lcl_ip.ip6 +#define c_rmt_ip6 connection.rmt_ip.ip6 +#define c_lcl_port connection.lcl_port +#define c_rmt_port connection.rmt_port +#define c_proto connection.proto +#define c_state connection.state +#define c_s_index connection.s_index +#define c_c_index connection.c_index +#define c_is_ip4 connection.is_ip4 +#define c_thread_index connection.thread_index +} transport_connection_t; + +/* + * Transport protocol virtual function table + */ +typedef struct _transport_proto_vft +{ + /* + * Setup + */ + u32 (*bind) (vlib_main_t *, u32, ip46_address_t *, u16); + u32 (*unbind) (vlib_main_t *, u32); + int (*open) (ip46_address_t * addr, u16 port_host_byte_order); + void (*close) (u32 conn_index, u32 thread_index); + void (*cleanup) (u32 conn_index, u32 thread_index); + + /* + * Transmission + */ + u32 (*push_header) (transport_connection_t * tconn, vlib_buffer_t * b); + u16 (*send_mss) (transport_connection_t * tc); + u32 (*send_space) (transport_connection_t * tc); + u32 (*rx_fifo_offset) (transport_connection_t * tc); + + /* + * Connection retrieval + */ + transport_connection_t *(*get_connection) (u32 conn_idx, u32 thread_idx); + transport_connection_t *(*get_listener) (u32 conn_index); + transport_connection_t *(*get_half_open) (u32 conn_index); + + /* + * Format + */ + u8 *(*format_connection) (u8 * s, va_list * args); + u8 *(*format_listener) (u8 * s, va_list * args); + u8 *(*format_half_open) (u8 * s, va_list * args); + +} transport_proto_vft_t; + +/* 16 octets */ +typedef CLIB_PACKED (struct + { + union + { + struct + { + ip4_address_t src; ip4_address_t dst; + u16 src_port; + u16 dst_port; + /* align by making this 4 octets even though its a 1-bit field + * NOTE: avoid key overlap with other transports that use 5 tuples for + * session identification. + */ + u32 proto; + }; + u64 as_u64[2]; + }; + }) v4_connection_key_t; + +typedef CLIB_PACKED (struct + { + union + { + struct + { + /* 48 octets */ + ip6_address_t src; ip6_address_t dst; + u16 src_port; + u16 dst_port; u32 proto; u8 unused_for_now[8]; + }; u64 as_u64[6]; + }; + }) v6_connection_key_t; + +typedef clib_bihash_kv_16_8_t session_kv4_t; +typedef clib_bihash_kv_48_8_t session_kv6_t; + +always_inline void +make_v4_ss_kv (session_kv4_t * kv, ip4_address_t * lcl, ip4_address_t * rmt, + u16 lcl_port, u16 rmt_port, u8 proto) +{ + v4_connection_key_t key; + memset (&key, 0, sizeof (v4_connection_key_t)); + + key.src.as_u32 = lcl->as_u32; + key.dst.as_u32 = rmt->as_u32; + key.src_port = lcl_port; + key.dst_port = rmt_port; + key.proto = proto; + + kv->key[0] = key.as_u64[0]; + kv->key[1] = key.as_u64[1]; + kv->value = ~0ULL; +} + +always_inline void +make_v4_listener_kv (session_kv4_t * kv, ip4_address_t * lcl, u16 lcl_port, + u8 proto) +{ + v4_connection_key_t key; + memset (&key, 0, sizeof (v4_connection_key_t)); + + key.src.as_u32 = lcl->as_u32; + key.dst.as_u32 = 0; + key.src_port = lcl_port; + key.dst_port = 0; + key.proto = proto; + + kv->key[0] = key.as_u64[0]; + kv->key[1] = key.as_u64[1]; + kv->value = ~0ULL; +} + +always_inline void +make_v4_ss_kv_from_tc (session_kv4_t * kv, transport_connection_t * t) +{ + return make_v4_ss_kv (kv, &t->lcl_ip.ip4, &t->rmt_ip.ip4, t->lcl_port, + t->rmt_port, t->proto); +} + +always_inline void +make_v6_ss_kv (session_kv6_t * kv, ip6_address_t * lcl, ip6_address_t * rmt, + u16 lcl_port, u16 rmt_port, u8 proto) +{ + v6_connection_key_t key; + memset (&key, 0, sizeof (v6_connection_key_t)); + + key.src.as_u64[0] = lcl->as_u64[0]; + key.src.as_u64[1] = lcl->as_u64[1]; + key.dst.as_u64[0] = rmt->as_u64[0]; + key.dst.as_u64[1] = rmt->as_u64[1]; + key.src_port = lcl_port; + key.dst_port = rmt_port; + key.proto = proto; + + kv->key[0] = key.as_u64[0]; + kv->key[1] = key.as_u64[1]; + kv->value = ~0ULL; +} + +always_inline void +make_v6_listener_kv (session_kv6_t * kv, ip6_address_t * lcl, u16 lcl_port, + u8 proto) +{ + v6_connection_key_t key; + memset (&key, 0, sizeof (v6_connection_key_t)); + + key.src.as_u64[0] = lcl->as_u64[0]; + key.src.as_u64[1] = lcl->as_u64[1]; + key.dst.as_u64[0] = 0; + key.dst.as_u64[1] = 0; + key.src_port = lcl_port; + key.dst_port = 0; + key.proto = proto; + + kv->key[0] = key.as_u64[0]; + kv->key[1] = key.as_u64[1]; + kv->value = ~0ULL; +} + +always_inline void +make_v6_ss_kv_from_tc (session_kv6_t * kv, transport_connection_t * t) +{ + make_v6_ss_kv (kv, &t->lcl_ip.ip6, &t->rmt_ip.ip6, t->lcl_port, + t->rmt_port, t->proto); +} + +typedef struct _transport_endpoint +{ + ip46_address_t ip; + u16 port; + u8 is_ip4; + u32 vrf; +} transport_endpoint_t; + +typedef clib_bihash_24_8_t transport_endpoint_table_t; + +#define TRANSPORT_ENDPOINT_INVALID_INDEX ((u32)~0) + +u32 +transport_endpoint_lookup (transport_endpoint_table_t * ht, + ip46_address_t * ip, u16 port); +void transport_endpoint_table_add (transport_endpoint_table_t * ht, + transport_endpoint_t * te, u32 value); +void transport_endpoint_table_del (transport_endpoint_table_t * ht, + transport_endpoint_t * te); + +#endif /* VNET_VNET_URI_TRANSPORT_H_ */ + +/* + * fd.io coding-style-patch-verification: ON + * + * Local Variables: + * eval: (c-set-style "gnu") + * End: + */ diff --git a/src/vnet/tcp/tcp.c b/src/vnet/tcp/tcp.c new file mode 100644 index 00000000..0f9b7097 --- /dev/null +++ b/src/vnet/tcp/tcp.c @@ -0,0 +1,708 @@ +/* + * Copyright (c) 2016 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +#include +#include +#include +#include + +tcp_main_t tcp_main; + +static u32 +tcp_connection_bind (vlib_main_t * vm, u32 session_index, ip46_address_t * ip, + u16 port_host_byte_order, u8 is_ip4) +{ + tcp_main_t *tm = &tcp_main; + tcp_connection_t *listener; + + pool_get (tm->listener_pool, listener); + memset (listener, 0, sizeof (*listener)); + + listener->c_c_index = listener - tm->listener_pool; + listener->c_lcl_port = clib_host_to_net_u16 (port_host_byte_order); + + if (is_ip4) + listener->c_lcl_ip4.as_u32 = ip->ip4.as_u32; + else + clib_memcpy (&listener->c_lcl_ip6, &ip->ip6, sizeof (ip6_address_t)); + + listener->c_s_index = session_index; + listener->c_proto = SESSION_TYPE_IP4_TCP; + listener->state = TCP_STATE_LISTEN; + listener->c_is_ip4 = 1; + + return listener->c_c_index; +} + +u32 +tcp_session_bind_ip4 (vlib_main_t * vm, u32 session_index, + ip46_address_t * ip, u16 port_host_byte_order) +{ + return tcp_connection_bind (vm, session_index, ip, port_host_byte_order, 1); +} + +u32 +tcp_session_bind_ip6 (vlib_main_t * vm, u32 session_index, + ip46_address_t * ip, u16 port_host_byte_order) +{ + return tcp_connection_bind (vm, session_index, ip, port_host_byte_order, 0); + +} + +static void +tcp_session_unbind (u32 listener_index) +{ + tcp_main_t *tm = vnet_get_tcp_main (); + pool_put_index (tm->listener_pool, listener_index); +} + +u32 +tcp_session_unbind_ip4 (vlib_main_t * vm, u32 listener_index) +{ + tcp_session_unbind (listener_index); + return 0; +} + +u32 +tcp_session_unbind_ip6 (vlib_main_t * vm, u32 listener_index) +{ + tcp_session_unbind (listener_index); + return 0; +} + +transport_connection_t * +tcp_session_get_listener (u32 listener_index) +{ + tcp_main_t *tm = vnet_get_tcp_main (); + tcp_connection_t *tc; + tc = pool_elt_at_index (tm->listener_pool, listener_index); + return &tc->connection; +} + +/** + * Cleans up connection state. + * + * No notifications. + */ +void +tcp_connection_cleanup (tcp_connection_t * tc) +{ + tcp_main_t *tm = &tcp_main; + u32 tepi; + transport_endpoint_t *tep; + + /* Cleanup local endpoint if this was an active connect */ + tepi = transport_endpoint_lookup (&tm->local_endpoints_table, &tc->c_lcl_ip, + tc->c_lcl_port); + + /*XXX lock */ + if (tepi != TRANSPORT_ENDPOINT_INVALID_INDEX) + { + tep = pool_elt_at_index (tm->local_endpoints, tepi); + transport_endpoint_table_del (&tm->local_endpoints_table, tep); + pool_put (tm->local_endpoints, tep); + } + + /* Make sure all timers are cleared */ + tcp_connection_timers_reset (tc); + + /* Check if half-open */ + if (tc->state == TCP_STATE_SYN_SENT) + pool_put (tm->half_open_connections, tc); + else + pool_put (tm->connections[tc->c_thread_index], tc); +} + +/** + * Connection removal. + * + * This should be called only once connection enters CLOSED state. Note + * that it notifies the session of the removal event, so if the goal is to + * just remove the connection, call tcp_connection_cleanup instead. + */ +void +tcp_connection_del (tcp_connection_t * tc) +{ + stream_session_delete_notify (&tc->connection); + tcp_connection_cleanup (tc); +} + +/** + * Begin connection closing procedure. + * + * If at the end the connection is not in CLOSED state, it is not removed. + * Instead, we rely on on TCP to advance through state machine to either + * 1) LAST_ACK (passive close) whereby when the last ACK is received + * tcp_connection_del is called. This notifies session of the delete and + * calls cleanup. + * 2) TIME_WAIT (active close) whereby after 2MSL the 2MSL timer triggers + * and cleanup is called. + */ +void +tcp_connection_close (tcp_connection_t * tc) +{ + /* Send FIN if needed */ + if (tc->state == TCP_STATE_ESTABLISHED || tc->state == TCP_STATE_SYN_RCVD + || tc->state == TCP_STATE_CLOSE_WAIT) + tcp_send_fin (tc); + + /* Switch state */ + if (tc->state == TCP_STATE_ESTABLISHED || tc->state == TCP_STATE_SYN_RCVD) + tc->state = TCP_STATE_FIN_WAIT_1; + else if (tc->state == TCP_STATE_SYN_SENT) + tc->state = TCP_STATE_CLOSED; + else if (tc->state == TCP_STATE_CLOSE_WAIT) + tc->state = TCP_STATE_LAST_ACK; + + /* Half-close connections are not supported XXX */ + + if (tc->state == TCP_STATE_CLOSED) + tcp_connection_del (tc); +} + +void +tcp_session_close (u32 conn_index, u32 thread_index) +{ + tcp_connection_t *tc; + tc = tcp_connection_get (conn_index, thread_index); + tcp_connection_close (tc); +} + +void +tcp_session_cleanup (u32 conn_index, u32 thread_index) +{ + tcp_connection_t *tc; + tc = tcp_connection_get (conn_index, thread_index); + tcp_connection_cleanup (tc); +} + +void * +ip_interface_get_first_ip (u32 sw_if_index, u8 is_ip4) +{ + ip_lookup_main_t *lm4 = &ip4_main.lookup_main; + ip_lookup_main_t *lm6 = &ip6_main.lookup_main; + ip_interface_address_t *ia = 0; + + if (is_ip4) + { + /* *INDENT-OFF* */ + foreach_ip_interface_address (lm4, ia, sw_if_index, 1 /* unnumbered */ , + ({ + return ip_interface_address_get_address (lm4, ia); + })); + /* *INDENT-ON* */ + } + else + { + /* *INDENT-OFF* */ + foreach_ip_interface_address (lm6, ia, sw_if_index, 1 /* unnumbered */ , + ({ + return ip_interface_address_get_address (lm6, ia); + })); + /* *INDENT-ON* */ + } + + return 0; +} + +/** + * Allocate local port and add if successful add entry to local endpoint + * table to mark the pair as used. + */ +u16 +tcp_allocate_local_port (tcp_main_t * tm, ip46_address_t * ip) +{ + u8 unique = 0; + transport_endpoint_t *tep; + u32 time_now, tei; + u16 min = 1024, max = 65535, tries; /* XXX configurable ? */ + + tries = max - min; + time_now = tcp_time_now (); + + /* Start at random point or max */ + pool_get (tm->local_endpoints, tep); + clib_memcpy (&tep->ip, ip, sizeof (*ip)); + tep->port = random_u32 (&time_now) << 16; + tep->port = tep->port < min ? max : tep->port; + + /* Search for first free slot */ + while (tries) + { + tei = transport_endpoint_lookup (&tm->local_endpoints_table, &tep->ip, + tep->port); + if (tei == TRANSPORT_ENDPOINT_INVALID_INDEX) + { + unique = 1; + break; + } + + tep->port--; + + if (tep->port < min) + tep->port = max; + + tries--; + } + + if (unique) + { + transport_endpoint_table_add (&tm->local_endpoints_table, tep, + tep - tm->local_endpoints); + + return tep->port; + } + + /* Failed */ + pool_put (tm->local_endpoints, tep); + return -1; +} + +/** + * Initialize all connection timers as invalid + */ +void +tcp_connection_timers_init (tcp_connection_t * tc) +{ + int i; + + /* Set all to invalid */ + for (i = 0; i < TCP_N_TIMERS; i++) + { + tc->timers[i] = TCP_TIMER_HANDLE_INVALID; + } + + tc->rto = TCP_RTO_INIT; +} + +/** + * Stop all connection timers + */ +void +tcp_connection_timers_reset (tcp_connection_t * tc) +{ + int i; + for (i = 0; i < TCP_N_TIMERS; i++) + { + tcp_timer_reset (tc, i); + } +} + +/** Initialize tcp connection variables + * + * Should be called after having received a msg from the peer, i.e., a SYN or + * a SYNACK, such that connection options have already been exchanged. */ +void +tcp_connection_init_vars (tcp_connection_t * tc) +{ + tcp_connection_timers_init (tc); + tcp_set_snd_mss (tc); + tc->sack_sb.head = TCP_INVALID_SACK_HOLE_INDEX; + tcp_cc_init (tc); +} + +int +tcp_connection_open (ip46_address_t * rmt_addr, u16 rmt_port, u8 is_ip4) +{ + tcp_main_t *tm = vnet_get_tcp_main (); + tcp_connection_t *tc; + fib_prefix_t prefix; + u32 fei, sw_if_index; + ip46_address_t lcl_addr; + u16 lcl_port; + + /* + * Find the local address and allocate port + */ + memset (&lcl_addr, 0, sizeof (lcl_addr)); + + /* Find a FIB path to the destination */ + clib_memcpy (&prefix.fp_addr, rmt_addr, sizeof (*rmt_addr)); + prefix.fp_proto = is_ip4 ? FIB_PROTOCOL_IP4 : FIB_PROTOCOL_IP6; + prefix.fp_len = is_ip4 ? 32 : 128; + + fei = fib_table_lookup (0, &prefix); + + /* Couldn't find route to destination. Bail out. */ + if (fei == FIB_NODE_INDEX_INVALID) + return -1; + + sw_if_index = fib_entry_get_resolving_interface (fei); + + if (sw_if_index == (u32) ~ 0) + return -1; + + if (is_ip4) + { + ip4_address_t *ip4; + ip4 = ip_interface_get_first_ip (sw_if_index, 1); + lcl_addr.ip4.as_u32 = ip4->as_u32; + } + else + { + ip6_address_t *ip6; + ip6 = ip_interface_get_first_ip (sw_if_index, 0); + clib_memcpy (&lcl_addr.ip6, ip6, sizeof (*ip6)); + } + + /* Allocate source port */ + lcl_port = tcp_allocate_local_port (tm, &lcl_addr); + if (lcl_port < 1) + return -1; + + /* + * Create connection and send SYN + */ + + pool_get (tm->half_open_connections, tc); + memset (tc, 0, sizeof (*tc)); + + clib_memcpy (&tc->c_rmt_ip, rmt_addr, sizeof (ip46_address_t)); + clib_memcpy (&tc->c_lcl_ip, &lcl_addr, sizeof (ip46_address_t)); + tc->c_rmt_port = clib_host_to_net_u16 (rmt_port); + tc->c_lcl_port = clib_host_to_net_u16 (lcl_port); + tc->c_c_index = tc - tm->half_open_connections; + tc->c_is_ip4 = is_ip4; + + /* The other connection vars will be initialized after SYN ACK */ + tcp_connection_timers_init (tc); + + tcp_send_syn (tc); + + tc->state = TCP_STATE_SYN_SENT; + + return tc->c_c_index; +} + +int +tcp_session_open_ip4 (ip46_address_t * addr, u16 port) +{ + return tcp_connection_open (addr, port, 1); +} + +int +tcp_session_open_ip6 (ip46_address_t * addr, u16 port) +{ + return tcp_connection_open (addr, port, 0); +} + +u8 * +format_tcp_session_ip4 (u8 * s, va_list * args) +{ + u32 tci = va_arg (*args, u32); + u32 thread_index = va_arg (*args, u32); + tcp_connection_t *tc; + + tc = tcp_connection_get (tci, thread_index); + + s = format (s, "[%s] %U:%d->%U:%d", "tcp", format_ip4_address, + &tc->c_lcl_ip4, clib_net_to_host_u16 (tc->c_lcl_port), + format_ip4_address, &tc->c_rmt_ip4, + clib_net_to_host_u16 (tc->c_rmt_port)); + + return s; +} + +u8 * +format_tcp_session_ip6 (u8 * s, va_list * args) +{ + u32 tci = va_arg (*args, u32); + u32 thread_index = va_arg (*args, u32); + tcp_connection_t *tc = tcp_connection_get (tci, thread_index); + s = format (s, "[%s] %U:%d->%U:%d", "tcp", format_ip6_address, + &tc->c_lcl_ip6, clib_net_to_host_u16 (tc->c_lcl_port), + format_ip6_address, &tc->c_rmt_ip6, + clib_net_to_host_u16 (tc->c_rmt_port)); + return s; +} + +u8 * +format_tcp_listener_session_ip4 (u8 * s, va_list * args) +{ + u32 tci = va_arg (*args, u32); + tcp_connection_t *tc = tcp_listener_get (tci); + s = format (s, "[%s] %U:%d->%U:%d", "tcp", format_ip4_address, + &tc->c_lcl_ip4, clib_net_to_host_u16 (tc->c_lcl_port), + format_ip4_address, &tc->c_rmt_ip4, + clib_net_to_host_u16 (tc->c_rmt_port)); + return s; +} + +u8 * +format_tcp_listener_session_ip6 (u8 * s, va_list * args) +{ + u32 tci = va_arg (*args, u32); + tcp_connection_t *tc = tcp_listener_get (tci); + s = format (s, "[%s] %U:%d->%U:%d", "tcp", format_ip6_address, + &tc->c_lcl_ip6, clib_net_to_host_u16 (tc->c_lcl_port), + format_ip6_address, &tc->c_rmt_ip6, + clib_net_to_host_u16 (tc->c_rmt_port)); + return s; +} + +u8 * +format_tcp_half_open_session_ip4 (u8 * s, va_list * args) +{ + u32 tci = va_arg (*args, u32); + tcp_connection_t *tc = tcp_half_open_connection_get (tci); + s = format (s, "[%s] %U:%d->%U:%d", "tcp", format_ip4_address, + &tc->c_lcl_ip4, clib_net_to_host_u16 (tc->c_lcl_port), + format_ip4_address, &tc->c_rmt_ip4, + clib_net_to_host_u16 (tc->c_rmt_port)); + return s; +} + +u8 * +format_tcp_half_open_session_ip6 (u8 * s, va_list * args) +{ + u32 tci = va_arg (*args, u32); + tcp_connection_t *tc = tcp_half_open_connection_get (tci); + s = format (s, "[%s] %U:%d->%U:%d", "tcp", format_ip6_address, + &tc->c_lcl_ip6, clib_net_to_host_u16 (tc->c_lcl_port), + format_ip6_address, &tc->c_rmt_ip6, + clib_net_to_host_u16 (tc->c_rmt_port)); + return s; +} + +transport_connection_t * +tcp_session_get_transport (u32 conn_index, u32 thread_index) +{ + tcp_connection_t *tc = tcp_connection_get (conn_index, thread_index); + return &tc->connection; +} + +transport_connection_t * +tcp_half_open_session_get_transport (u32 conn_index) +{ + tcp_connection_t *tc = tcp_half_open_connection_get (conn_index); + return &tc->connection; +} + +u16 +tcp_session_send_mss (transport_connection_t * trans_conn) +{ + tcp_connection_t *tc = (tcp_connection_t *) trans_conn; + return tc->snd_mss; +} + +u32 +tcp_session_send_space (transport_connection_t * trans_conn) +{ + tcp_connection_t *tc = (tcp_connection_t *) trans_conn; + return tcp_available_snd_space (tc); +} + +u32 +tcp_session_rx_fifo_offset (transport_connection_t * trans_conn) +{ + tcp_connection_t *tc = (tcp_connection_t *) trans_conn; + return (tc->snd_una_max - tc->snd_una); +} + +/* *INDENT-OFF* */ +const static transport_proto_vft_t tcp4_proto = { + .bind = tcp_session_bind_ip4, + .unbind = tcp_session_unbind_ip4, + .push_header = tcp_push_header, + .get_connection = tcp_session_get_transport, + .get_listener = tcp_session_get_listener, + .get_half_open = tcp_half_open_session_get_transport, + .open = tcp_session_open_ip4, + .close = tcp_session_close, + .cleanup = tcp_session_cleanup, + .send_mss = tcp_session_send_mss, + .send_space = tcp_session_send_space, + .rx_fifo_offset = tcp_session_rx_fifo_offset, + .format_connection = format_tcp_session_ip4, + .format_listener = format_tcp_listener_session_ip4, + .format_half_open = format_tcp_half_open_session_ip4 +}; + +const static transport_proto_vft_t tcp6_proto = { + .bind = tcp_session_bind_ip6, + .unbind = tcp_session_unbind_ip6, + .push_header = tcp_push_header, + .get_connection = tcp_session_get_transport, + .get_listener = tcp_session_get_listener, + .get_half_open = tcp_half_open_session_get_transport, + .open = tcp_session_open_ip6, + .close = tcp_session_close, + .cleanup = tcp_session_cleanup, + .send_mss = tcp_session_send_mss, + .send_space = tcp_session_send_space, + .rx_fifo_offset = tcp_session_rx_fifo_offset, + .format_connection = format_tcp_session_ip6, + .format_listener = format_tcp_listener_session_ip6, + .format_half_open = format_tcp_half_open_session_ip6 +}; +/* *INDENT-ON* */ + +void +tcp_timer_keep_handler (u32 conn_index) +{ + u32 cpu_index = os_get_cpu_number (); + tcp_connection_t *tc; + + tc = tcp_connection_get (conn_index, cpu_index); + tc->timers[TCP_TIMER_KEEP] = TCP_TIMER_HANDLE_INVALID; + + tcp_connection_close (tc); +} + +void +tcp_timer_establish_handler (u32 conn_index) +{ + tcp_connection_t *tc; + u8 sst; + + tc = tcp_half_open_connection_get (conn_index); + tc->timers[TCP_TIMER_ESTABLISH] = TCP_TIMER_HANDLE_INVALID; + + ASSERT (tc->state == TCP_STATE_SYN_SENT); + + sst = tc->c_is_ip4 ? SESSION_TYPE_IP4_TCP : SESSION_TYPE_IP6_TCP; + stream_session_connect_notify (&tc->connection, sst, 1 /* fail */ ); + + tcp_connection_cleanup (tc); +} + +void +tcp_timer_2msl_handler (u32 conn_index) +{ + u32 cpu_index = os_get_cpu_number (); + tcp_connection_t *tc; + + tc = tcp_connection_get (conn_index, cpu_index); + tc->timers[TCP_TIMER_2MSL] = TCP_TIMER_HANDLE_INVALID; + + tcp_connection_del (tc); +} + +/* *INDENT-OFF* */ +static timer_expiration_handler *timer_expiration_handlers[TCP_N_TIMERS] = +{ + tcp_timer_retransmit_handler, + tcp_timer_delack_handler, + 0, + tcp_timer_keep_handler, + tcp_timer_2msl_handler, + tcp_timer_retransmit_syn_handler, + tcp_timer_establish_handler +}; +/* *INDENT-ON* */ + +static void +tcp_expired_timers_dispatch (u32 * expired_timers) +{ + int i; + u32 connection_index, timer_id; + + for (i = 0; i < vec_len (expired_timers); i++) + { + /* Get session index and timer id */ + connection_index = expired_timers[i] & 0x0FFFFFFF; + timer_id = expired_timers[i] >> 28; + + /* Handle expiration */ + (*timer_expiration_handlers[timer_id]) (connection_index); + } +} + +void +tcp_initialize_timer_wheels (tcp_main_t * tm) +{ + tw_timer_wheel_16t_2w_512sl_t *tw; + vec_foreach (tw, tm->timer_wheels) + { + tw_timer_wheel_init_16t_2w_512sl (tw, tcp_expired_timers_dispatch, + 100e-3 /* timer period 100ms */ , ~0); + tw->last_run_time = vlib_time_now (tm->vlib_main); + } +} + +clib_error_t * +tcp_init (vlib_main_t * vm) +{ + ip_main_t *im = &ip_main; + ip_protocol_info_t *pi; + tcp_main_t *tm = vnet_get_tcp_main (); + vlib_thread_main_t *vtm = vlib_get_thread_main (); + clib_error_t *error = 0; + u32 num_threads; + + tm->vlib_main = vm; + tm->vnet_main = vnet_get_main (); + + if ((error = vlib_call_init_function (vm, ip_main_init))) + return error; + if ((error = vlib_call_init_function (vm, ip4_lookup_init))) + return error; + if ((error = vlib_call_init_function (vm, ip6_lookup_init))) + return error; + + /* + * Registrations + */ + + /* Register with IP */ + pi = ip_get_protocol_info (im, IP_PROTOCOL_TCP); + if (pi == 0) + return clib_error_return (0, "TCP protocol info AWOL"); + pi->format_header = format_tcp_header; + pi->unformat_pg_edit = unformat_pg_tcp_header; + + ip4_register_protocol (IP_PROTOCOL_TCP, tcp4_input_node.index); + + /* Register as transport with URI */ + session_register_transport (SESSION_TYPE_IP4_TCP, &tcp4_proto); + session_register_transport (SESSION_TYPE_IP6_TCP, &tcp6_proto); + + /* + * Initialize data structures + */ + + num_threads = 1 /* main thread */ + vtm->n_threads; + vec_validate (tm->connections, num_threads - 1); + + /* Initialize per worker thread tx buffers (used for control messages) */ + vec_validate (tm->tx_buffers, num_threads - 1); + + /* Initialize timer wheels */ + vec_validate (tm->timer_wheels, num_threads - 1); + tcp_initialize_timer_wheels (tm); + + vec_validate (tm->delack_connections, num_threads - 1); + + /* Initialize clocks per tick for TCP timestamp. Used to compute + * monotonically increasing timestamps. */ + tm->tstamp_ticks_per_clock = vm->clib_time.seconds_per_clock + / TCP_TSTAMP_RESOLUTION; + + clib_bihash_init_24_8 (&tm->local_endpoints_table, "local endpoint table", + 200000 /* $$$$ config parameter nbuckets */ , + (64 << 20) /*$$$ config parameter table size */ ); + + return error; +} + +VLIB_INIT_FUNCTION (tcp_init); + +/* + * fd.io coding-style-patch-verification: ON + * + * Local Variables: + * eval: (c-set-style "gnu") + * End: + */ diff --git a/src/vnet/tcp/tcp.h b/src/vnet/tcp/tcp.h new file mode 100644 index 00000000..22f00a63 --- /dev/null +++ b/src/vnet/tcp/tcp.h @@ -0,0 +1,624 @@ +/* + * Copyright (c) 2016 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +#ifndef _vnet_tcp_h_ +#define _vnet_tcp_h_ + +#include +#include +#include +#include +#include +#include + +#define TCP_TICK 10e-3 /**< TCP tick period (s) */ +#define THZ 1/TCP_TICK /**< TCP tick frequency */ +#define TCP_TSTAMP_RESOLUTION TCP_TICK /**< Time stamp resolution */ +#define TCP_PAWS_IDLE 24 * 24 * 60 * 60 * THZ /**< 24 days */ +#define TCP_MAX_OPTION_SPACE 40 + +#define TCP_DUPACK_THRESHOLD 3 +#define TCP_DEFAULT_RX_FIFO_SIZE 64 << 10 + +/** TCP FSM state definitions as per RFC793. */ +#define foreach_tcp_fsm_state \ + _(CLOSED, "CLOSED") \ + _(LISTEN, "LISTEN") \ + _(SYN_SENT, "SYN_SENT") \ + _(SYN_RCVD, "SYN_RCVD") \ + _(ESTABLISHED, "ESTABLISHED") \ + _(CLOSE_WAIT, "CLOSE_WAIT") \ + _(FIN_WAIT_1, "FIN_WAIT_1") \ + _(LAST_ACK, "LAST_ACK") \ + _(CLOSING, "CLOSING") \ + _(FIN_WAIT_2, "FIN_WAIT_2") \ + _(TIME_WAIT, "TIME_WAIT") + +typedef enum _tcp_state +{ +#define _(sym, str) TCP_STATE_##sym, + foreach_tcp_fsm_state +#undef _ + TCP_N_STATES +} tcp_state_t; + +format_function_t format_tcp_state; + +/** TCP timers */ +#define foreach_tcp_timer \ + _(RETRANSMIT, "RETRANSMIT") \ + _(DELACK, "DELAYED ACK") \ + _(PERSIST, "PERSIST") \ + _(KEEP, "KEEP") \ + _(2MSL, "2MSL") \ + _(RETRANSMIT_SYN, "RETRANSMIT_SYN") \ + _(ESTABLISH, "ESTABLISH") + +typedef enum _tcp_timers +{ +#define _(sym, str) TCP_TIMER_##sym, + foreach_tcp_timer +#undef _ + TCP_N_TIMERS +} tcp_timers_e; + +typedef void (timer_expiration_handler) (u32 index); + +extern timer_expiration_handler tcp_timer_delack_handler; +extern timer_expiration_handler tcp_timer_retransmit_handler; +extern timer_expiration_handler tcp_timer_retransmit_syn_handler; + +#define TCP_TIMER_HANDLE_INVALID ((u32) ~0) + +/* Timer delays as multiples of 100ms */ +#define TCP_TO_TIMER_TICK TCP_TICK*10 /* Period for converting from TCP + * ticks to timer units */ +#define TCP_DELACK_TIME 1 /* 0.1s */ +#define TCP_ESTABLISH_TIME 750 /* 75s */ +#define TCP_2MSL_TIME 300 /* 30s */ + +#define TCP_RTO_MAX 60 * THZ /* Min max RTO (60s) as per RFC6298 */ +#define TCP_RTT_MAX 30 * THZ /* 30s (probably too much) */ +#define TCP_RTO_SYN_RETRIES 3 /* SYN retries without doubling RTO */ +#define TCP_RTO_INIT 1 * THZ /* Initial retransmit timer */ + +void tcp_update_time (f64 now, u32 thread_index); + +/** TCP connection flags */ +#define foreach_tcp_connection_flag \ + _(DELACK, "Delay ACK") \ + _(SNDACK, "Send ACK") \ + _(BURSTACK, "Burst ACK set") \ + _(SENT_RCV_WND0, "Sent 0 receive window") \ + _(RECOVERY, "Recovery on") \ + _(FAST_RECOVERY, "Fast Recovery on") + +typedef enum _tcp_connection_flag_bits +{ +#define _(sym, str) TCP_CONN_##sym##_BIT, + foreach_tcp_connection_flag +#undef _ + TCP_CONN_N_FLAG_BITS +} tcp_connection_flag_bits_e; + +typedef enum _tcp_connection_flag +{ +#define _(sym, str) TCP_CONN_##sym = 1 << TCP_CONN_##sym##_BIT, + foreach_tcp_connection_flag +#undef _ + TCP_CONN_N_FLAGS +} tcp_connection_flags_e; + +/** TCP buffer flags */ +#define foreach_tcp_buf_flag \ + _ (ACK) /**< Sending ACK. */ \ + _ (DUPACK) /**< Sending DUPACK. */ \ + +enum +{ +#define _(f) TCP_BUF_BIT_##f, + foreach_tcp_buf_flag +#undef _ + TCP_N_BUF_BITS, +}; + +enum +{ +#define _(f) TCP_BUF_FLAG_##f = 1 << TCP_BUF_BIT_##f, + foreach_tcp_buf_flag +#undef _ +}; + +#define TCP_MAX_SACK_BLOCKS 5 /**< Max number of SACK blocks stored */ +#define TCP_INVALID_SACK_HOLE_INDEX ((u32)~0) + +typedef struct _sack_scoreboard_hole +{ + u32 next; /**< Index for next entry in linked list */ + u32 prev; /**< Index for previous entry in linked list */ + u32 start; /**< Start sequence number */ + u32 end; /**< End sequence number */ +} sack_scoreboard_hole_t; + +typedef struct _sack_scoreboard +{ + sack_scoreboard_hole_t *holes; /**< Pool of holes */ + u32 head; /**< Index to first entry */ + u32 sacked_bytes; /**< Number of bytes sacked in sb */ +} sack_scoreboard_t; + +typedef enum _tcp_cc_algorithm_type +{ + TCP_CC_NEWRENO, +} tcp_cc_algorithm_type_e; + +typedef struct _tcp_cc_algorithm tcp_cc_algorithm_t; + +typedef enum _tcp_cc_ack_t +{ + TCP_CC_ACK, + TCP_CC_DUPACK, + TCP_CC_PARTIALACK +} tcp_cc_ack_t; + +typedef struct _tcp_connection +{ + transport_connection_t connection; /**< Common transport data. First! */ + + u8 state; /**< TCP state as per tcp_state_t */ + u16 flags; /**< Connection flags (see tcp_conn_flags_e) */ + u32 timers[TCP_N_TIMERS]; /**< Timer handles into timer wheel */ + + /* TODO RFC4898 */ + + /** Send sequence variables RFC793 */ + u32 snd_una; /**< oldest unacknowledged sequence number */ + u32 snd_una_max; /**< newest unacknowledged sequence number + 1*/ + u32 snd_wnd; /**< send window */ + u32 snd_wl1; /**< seq number used for last snd.wnd update */ + u32 snd_wl2; /**< ack number used for last snd.wnd update */ + u32 snd_nxt; /**< next seq number to be sent */ + + /** Receive sequence variables RFC793 */ + u32 rcv_nxt; /**< next sequence number expected */ + u32 rcv_wnd; /**< receive window we expect */ + + u32 rcv_las; /**< rcv_nxt at last ack sent/rcv_wnd update */ + u32 iss; /**< initial sent sequence */ + u32 irs; /**< initial remote sequence */ + + /* Options */ + tcp_options_t opt; /**< TCP connection options parsed */ + u8 rcv_wscale; /**< Window scale to advertise to peer */ + u8 snd_wscale; /**< Window scale to use when sending */ + u32 tsval_recent; /**< Last timestamp received */ + u32 tsval_recent_age; /**< When last updated tstamp_recent*/ + + sack_block_t *snd_sacks; /**< Vector of SACKs to send. XXX Fixed size? */ + sack_scoreboard_t sack_sb; /**< SACK "scoreboard" that tracks holes */ + + u8 rcv_dupacks; /**< Number of DUPACKs received */ + u8 snt_dupacks; /**< Number of DUPACKs sent in a burst */ + + /* Congestion control */ + u32 cwnd; /**< Congestion window */ + u32 ssthresh; /**< Slow-start threshold */ + u32 prev_ssthresh; /**< ssthresh before congestion */ + u32 bytes_acked; /**< Bytes acknowledged by current segment */ + u32 rtx_bytes; /**< Retransmitted bytes */ + u32 tsecr_last_ack; /**< Timestamp echoed to us in last health ACK */ + tcp_cc_algorithm_t *cc_algo; /**< Congestion control algorithm */ + + /* RTT and RTO */ + u32 rto; /**< Retransmission timeout */ + u32 rto_boff; /**< Index for RTO backoff */ + u32 srtt; /**< Smoothed RTT */ + u32 rttvar; /**< Smoothed mean RTT difference. Approximates variance */ + u32 rtt_ts; /**< Timestamp for tracked ACK */ + u32 rtt_seq; /**< Sequence number for tracked ACK */ + + u16 snd_mss; /**< Send MSS */ +} tcp_connection_t; + +struct _tcp_cc_algorithm +{ + void (*rcv_ack) (tcp_connection_t * tc); + void (*rcv_cong_ack) (tcp_connection_t * tc, tcp_cc_ack_t ack); + void (*congestion) (tcp_connection_t * tc); + void (*recovered) (tcp_connection_t * tc); + void (*init) (tcp_connection_t * tc); +}; + +#define tcp_fastrecovery_on(tc) (tc)->flags |= TCP_CONN_FAST_RECOVERY +#define tcp_fastrecovery_off(tc) (tc)->flags &= ~TCP_CONN_FAST_RECOVERY +#define tcp_in_fastrecovery(tc) ((tc)->flags & TCP_CONN_FAST_RECOVERY) +#define tcp_in_recovery(tc) ((tc)->flags & (TCP_CONN_FAST_RECOVERY | TCP_CONN_RECOVERY)) +#define tcp_recovery_off(tc) ((tc)->flags &= ~(TCP_CONN_FAST_RECOVERY | TCP_CONN_RECOVERY)) +#define tcp_in_slowstart(tc) (tc->cwnd < tc->ssthresh) + +typedef enum +{ + TCP_IP4, + TCP_IP6, + TCP_N_AF, +} tcp_af_t; + +typedef enum _tcp_error +{ +#define tcp_error(n,s) TCP_ERROR_##n, +#include +#undef tcp_error + TCP_N_ERROR, +} tcp_error_t; + +typedef struct _tcp_lookup_dispatch +{ + u8 next, error; +} tcp_lookup_dispatch_t; + +typedef struct _tcp_main +{ + /* Per-worker thread tcp connection pools */ + tcp_connection_t **connections; + + /* Pool of listeners. */ + tcp_connection_t *listener_pool; + + /** Dispatch table by state and flags */ + tcp_lookup_dispatch_t dispatch_table[TCP_N_STATES][64]; + + u8 log2_tstamp_clocks_per_tick; + f64 tstamp_ticks_per_clock; + + /** per-worker tx buffer free lists */ + u32 **tx_buffers; + + /* Per worker-thread timer wheel for connections timers */ + tw_timer_wheel_16t_2w_512sl_t *timer_wheels; + + /* Convenience per worker-thread vector of connections to DELACK */ + u32 **delack_connections; + + /* Pool of half-open connections on which we've sent a SYN */ + tcp_connection_t *half_open_connections; + + /* Pool of local TCP endpoints */ + transport_endpoint_t *local_endpoints; + + /* Local endpoints lookup table */ + transport_endpoint_table_t local_endpoints_table; + + /* Congestion control algorithms registered */ + tcp_cc_algorithm_t *cc_algos; + + /* convenience */ + vlib_main_t *vlib_main; + vnet_main_t *vnet_main; + ip4_main_t *ip4_main; + ip6_main_t *ip6_main; +} tcp_main_t; + +extern tcp_main_t tcp_main; +extern vlib_node_registration_t tcp4_input_node; +extern vlib_node_registration_t tcp6_input_node; +extern vlib_node_registration_t tcp4_output_node; +extern vlib_node_registration_t tcp6_output_node; + +always_inline tcp_main_t * +vnet_get_tcp_main () +{ + return &tcp_main; +} + +always_inline tcp_connection_t * +tcp_connection_get (u32 conn_index, u32 thread_index) +{ + return pool_elt_at_index (tcp_main.connections[thread_index], conn_index); +} + +always_inline tcp_connection_t * +tcp_connection_get_if_valid (u32 conn_index, u32 thread_index) +{ + if (tcp_main.connections[thread_index] == 0) + return 0; + if (pool_is_free_index (tcp_main.connections[thread_index], conn_index)) + return 0; + return pool_elt_at_index (tcp_main.connections[thread_index], conn_index); +} + +void tcp_connection_close (tcp_connection_t * tc); +void tcp_connection_cleanup (tcp_connection_t * tc); +void tcp_connection_del (tcp_connection_t * tc); + +always_inline tcp_connection_t * +tcp_listener_get (u32 tli) +{ + return pool_elt_at_index (tcp_main.listener_pool, tli); +} + +always_inline tcp_connection_t * +tcp_half_open_connection_get (u32 conn_index) +{ + return pool_elt_at_index (tcp_main.half_open_connections, conn_index); +} + +void tcp_make_ack (tcp_connection_t * ts, vlib_buffer_t * b); +void tcp_make_finack (tcp_connection_t * tc, vlib_buffer_t * b); +void tcp_make_synack (tcp_connection_t * ts, vlib_buffer_t * b); +void tcp_send_reset (vlib_buffer_t * pkt, u8 is_ip4); +void tcp_send_syn (tcp_connection_t * tc); +void tcp_send_fin (tcp_connection_t * tc); +void tcp_set_snd_mss (tcp_connection_t * tc); + +always_inline u32 +tcp_end_seq (tcp_header_t * th, u32 len) +{ + return th->seq_number + tcp_is_syn (th) + tcp_is_fin (th) + len; +} + +/* Modulo arithmetic for TCP sequence numbers */ +#define seq_lt(_s1, _s2) ((i32)((_s1)-(_s2)) < 0) +#define seq_leq(_s1, _s2) ((i32)((_s1)-(_s2)) <= 0) +#define seq_gt(_s1, _s2) ((i32)((_s1)-(_s2)) > 0) +#define seq_geq(_s1, _s2) ((i32)((_s1)-(_s2)) >= 0) + +/* Modulo arithmetic for timestamps */ +#define timestamp_lt(_t1, _t2) ((i32)((_t1)-(_t2)) < 0) +#define timestamp_leq(_t1, _t2) ((i32)((_t1)-(_t2)) <= 0) + +always_inline u32 +tcp_flight_size (const tcp_connection_t * tc) +{ + return tc->snd_una_max - tc->snd_una - tc->sack_sb.sacked_bytes + + tc->rtx_bytes; +} + +/** + * Initial cwnd as per RFC5681 + */ +always_inline u32 +tcp_initial_cwnd (const tcp_connection_t * tc) +{ + if (tc->snd_mss > 2190) + return 2 * tc->snd_mss; + else if (tc->snd_mss > 1095) + return 3 * tc->snd_mss; + else + return 4 * tc->snd_mss; +} + +always_inline u32 +tcp_loss_wnd (const tcp_connection_t * tc) +{ + return tc->snd_mss; +} + +always_inline u32 +tcp_available_wnd (const tcp_connection_t * tc) +{ + return clib_min (tc->cwnd, tc->snd_wnd); +} + +always_inline u32 +tcp_available_snd_space (const tcp_connection_t * tc) +{ + u32 available_wnd = tcp_available_wnd (tc); + u32 flight_size = tcp_flight_size (tc); + + if (available_wnd <= flight_size) + return 0; + + return available_wnd - flight_size; +} + +void tcp_retransmit_first_unacked (tcp_connection_t * tc); + +void tcp_fast_retransmit (tcp_connection_t * tc); + +always_inline u32 +tcp_time_now (void) +{ + return clib_cpu_time_now () * tcp_main.tstamp_ticks_per_clock; +} + +u32 tcp_push_header (transport_connection_t * tconn, vlib_buffer_t * b); + +u32 +tcp_prepare_retransmit_segment (tcp_connection_t * tc, vlib_buffer_t * b, + u32 max_bytes); + +void tcp_connection_timers_init (tcp_connection_t * tc); +void tcp_connection_timers_reset (tcp_connection_t * tc); + +void tcp_connection_init_vars (tcp_connection_t * tc); + +always_inline void +tcp_connection_force_ack (tcp_connection_t * tc, vlib_buffer_t * b) +{ + /* Reset flags, make sure ack is sent */ + tc->flags = TCP_CONN_SNDACK; + vnet_buffer (b)->tcp.flags &= ~TCP_BUF_FLAG_DUPACK; +} + +always_inline void +tcp_timer_set (tcp_connection_t * tc, u8 timer_id, u32 interval) +{ + tc->timers[timer_id] + = tw_timer_start_16t_2w_512sl (&tcp_main.timer_wheels[tc->c_thread_index], + tc->c_c_index, timer_id, interval); +} + +always_inline void +tcp_retransmit_timer_set (tcp_main_t * tm, tcp_connection_t * tc) +{ + /* XXX Switch to faster TW */ + tcp_timer_set (tc, TCP_TIMER_RETRANSMIT, + clib_max (tc->rto * TCP_TO_TIMER_TICK, 1)); +} + +always_inline void +tcp_timer_reset (tcp_connection_t * tc, u8 timer_id) +{ + if (tc->timers[timer_id] == TCP_TIMER_HANDLE_INVALID) + return; + + tw_timer_stop_16t_2w_512sl (&tcp_main.timer_wheels[tc->c_thread_index], + tc->timers[timer_id]); + tc->timers[timer_id] = TCP_TIMER_HANDLE_INVALID; +} + +always_inline void +tcp_timer_update (tcp_connection_t * tc, u8 timer_id, u32 interval) +{ + if (tc->timers[timer_id] != TCP_TIMER_HANDLE_INVALID) + tw_timer_stop_16t_2w_512sl (&tcp_main.timer_wheels[tc->c_thread_index], + tc->timers[timer_id]); + tc->timers[timer_id] = + tw_timer_start_16t_2w_512sl (&tcp_main.timer_wheels[tc->c_thread_index], + tc->c_c_index, timer_id, interval); +} + +always_inline u8 +tcp_timer_is_active (tcp_connection_t * tc, tcp_timers_e timer) +{ + return tc->timers[timer] != TCP_TIMER_HANDLE_INVALID; +} + +void +scoreboard_remove_hole (sack_scoreboard_t * sb, + sack_scoreboard_hole_t * hole); + +always_inline sack_scoreboard_hole_t * +scoreboard_next_hole (sack_scoreboard_t * sb, sack_scoreboard_hole_t * hole) +{ + if (hole->next != TCP_INVALID_SACK_HOLE_INDEX) + return pool_elt_at_index (sb->holes, hole->next); + return 0; +} + +always_inline sack_scoreboard_hole_t * +scoreboard_first_hole (sack_scoreboard_t * sb) +{ + if (sb->head != TCP_INVALID_SACK_HOLE_INDEX) + return pool_elt_at_index (sb->holes, sb->head); + return 0; +} + +always_inline void +scoreboard_clear (sack_scoreboard_t * sb) +{ + sack_scoreboard_hole_t *hole = scoreboard_first_hole (sb); + while ((hole = scoreboard_first_hole (sb))) + { + scoreboard_remove_hole (sb, hole); + } +} + +always_inline u32 +scoreboard_hole_bytes (sack_scoreboard_hole_t * hole) +{ + return hole->end - hole->start; +} + +always_inline void +tcp_cc_algo_register (tcp_cc_algorithm_type_e type, + const tcp_cc_algorithm_t * vft) +{ + tcp_main_t *tm = vnet_get_tcp_main (); + vec_validate (tm->cc_algos, type); + + tm->cc_algos[type] = *vft; +} + +always_inline tcp_cc_algorithm_t * +tcp_cc_algo_get (tcp_cc_algorithm_type_e type) +{ + tcp_main_t *tm = vnet_get_tcp_main (); + return &tm->cc_algos[type]; +} + +void tcp_cc_init (tcp_connection_t * tc); + +/** + * Push TCP header to buffer + * + * @param vm - vlib_main + * @param b - buffer to write the header to + * @param sp_net - source port net order + * @param dp_net - destination port net order + * @param seq - sequence number net order + * @param ack - ack number net order + * @param tcp_hdr_opts_len - header and options length in bytes + * @param flags - header flags + * @param wnd - window size + * + * @return - pointer to start of TCP header + */ +always_inline void * +vlib_buffer_push_tcp_net_order (vlib_buffer_t * b, u16 sp, u16 dp, u32 seq, + u32 ack, u8 tcp_hdr_opts_len, u8 flags, + u16 wnd) +{ + tcp_header_t *th; + + th = vlib_buffer_push_uninit (b, tcp_hdr_opts_len); + + th->src_port = sp; + th->dst_port = dp; + th->seq_number = seq; + th->ack_number = ack; + th->data_offset_and_reserved = (tcp_hdr_opts_len >> 2) << 4; + th->flags = flags; + th->window = wnd; + th->checksum = 0; + th->urgent_pointer = 0; + return th; +} + +/** + * Push TCP header to buffer + * + * @param vm - vlib_main + * @param b - buffer to write the header to + * @param sp_net - source port net order + * @param dp_net - destination port net order + * @param seq - sequence number host order + * @param ack - ack number host order + * @param tcp_hdr_opts_len - header and options length in bytes + * @param flags - header flags + * @param wnd - window size + * + * @return - pointer to start of TCP header + */ +always_inline void * +vlib_buffer_push_tcp (vlib_buffer_t * b, u16 sp_net, u16 dp_net, u32 seq, + u32 ack, u8 tcp_hdr_opts_len, u8 flags, u16 wnd) +{ + return vlib_buffer_push_tcp_net_order (b, sp_net, dp_net, + clib_host_to_net_u32 (seq), + clib_host_to_net_u32 (ack), + tcp_hdr_opts_len, flags, + clib_host_to_net_u16 (wnd)); +} + +#endif /* _vnet_tcp_h_ */ + +/* + * fd.io coding-style-patch-verification: ON + * + * Local Variables: + * eval: (c-set-style "gnu") + * End: + */ diff --git a/src/vnet/tcp/tcp_error.def b/src/vnet/tcp/tcp_error.def new file mode 100644 index 00000000..cff5ec13 --- /dev/null +++ b/src/vnet/tcp/tcp_error.def @@ -0,0 +1,35 @@ +/* + * Copyright (c) 2016 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +tcp_error (NONE, "no error") +tcp_error (NO_LISTENER, "no listener for dst port") +tcp_error (LOOKUP_DROPS, "lookup drops") +tcp_error (DISPATCH, "Dispatch error") +tcp_error (ENQUEUED, "Packets pushed into rx fifo") +tcp_error (PURE_ACK, "Pure acks") +tcp_error (SYNS_RCVD, "SYNs received") +tcp_error (SYN_ACKS_RCVD, "SYN-ACKs received") +tcp_error (NOT_READY, "Session not ready for packets") +tcp_error (FIFO_FULL, "Packets dropped for lack of rx fifo space") +tcp_error (EVENT_FIFO_FULL, "Events not sent for lack of event fifo space") +tcp_error (API_QUEUE_FULL, "Sessions not created for lack of API queue space") +tcp_error (CREATE_SESSION_FAIL, "Sessions couldn't be allocated") +tcp_error (SEGMENT_INVALID, "Invalid segment") +tcp_error (ACK_INVALID, "Invalid ACK") +tcp_error (ACK_DUP, "Duplicate ACK") +tcp_error (ACK_OLD, "Old ACK") +tcp_error (PKTS_SENT, "Packets sent") +tcp_error (FILTERED_DUPACKS, "Filtered duplicate ACKs") +tcp_error (RST_SENT, "Resets sent") \ No newline at end of file diff --git a/src/vnet/tcp/tcp_format.c b/src/vnet/tcp/tcp_format.c new file mode 100644 index 00000000..7136741d --- /dev/null +++ b/src/vnet/tcp/tcp_format.c @@ -0,0 +1,136 @@ +/* + * Copyright (c) 2015 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ +/* + * tcp/tcp_format.c: tcp formatting + * + * Copyright (c) 2008 Eliot Dresselhaus + * + * Permission is hereby granted, free of charge, to any person obtaining + * a copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sublicense, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice shall be + * included in all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND + * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE + * LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + * OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + * WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + */ + +#include +#include + +static u8 * +format_tcp_flags (u8 * s, va_list * args) +{ + int flags = va_arg (*args, int); + +#define _(f) if (flags & TCP_FLAG_##f) s = format (s, "%s, ", #f); + foreach_tcp_flag +#undef _ + return s; +} + +/* Format TCP header. */ +u8 * +format_tcp_header (u8 * s, va_list * args) +{ + tcp_header_t *tcp = va_arg (*args, tcp_header_t *); + u32 max_header_bytes = va_arg (*args, u32); + u32 header_bytes; + uword indent; + + /* Nothing to do. */ + if (max_header_bytes < sizeof (tcp[0])) + return format (s, "TCP header truncated"); + + indent = format_get_indent (s); + indent += 2; + header_bytes = tcp_header_bytes (tcp); + + s = format (s, "TCP: %d -> %d", clib_net_to_host_u16 (tcp->src), + clib_net_to_host_u16 (tcp->dst)); + + s = format (s, "\n%Useq. 0x%08x ack 0x%08x", format_white_space, indent, + clib_net_to_host_u32 (tcp->seq_number), + clib_net_to_host_u32 (tcp->ack_number)); + + s = format (s, "\n%Uflags %U, tcp header: %d bytes", format_white_space, + indent, format_tcp_flags, tcp->flags, header_bytes); + + s = format (s, "\n%Uwindow %d, checksum 0x%04x", format_white_space, indent, + clib_net_to_host_u16 (tcp->window), + clib_net_to_host_u16 (tcp->checksum)); + + +#if 0 + /* Format TCP options. */ + { + u8 *o; + u8 *option_start = (void *) (tcp + 1); + u8 *option_end = (void *) tcp + header_bytes; + + for (o = option_start; o < option_end;) + { + u32 length = o[1]; + switch (o[0]) + { + case TCP_OPTION_END: + length = 1; + o = option_end; + break; + + case TCP_OPTION_NOOP: + length = 1; + break; + + } + } + } +#endif + + /* Recurse into next protocol layer. */ + if (max_header_bytes != 0 && header_bytes < max_header_bytes) + { + ip_main_t *im = &ip_main; + tcp_udp_port_info_t *pi; + + pi = ip_get_tcp_udp_port_info (im, tcp->dst); + + if (pi && pi->format_header) + s = format (s, "\n%U%U", format_white_space, indent - 2, + pi->format_header, + /* next protocol header */ (void *) tcp + header_bytes, + max_header_bytes - header_bytes); + } + + return s; +} + +/* + * fd.io coding-style-patch-verification: ON + * + * Local Variables: + * eval: (c-set-style "gnu") + * End: + */ diff --git a/src/vnet/tcp/tcp_input.c b/src/vnet/tcp/tcp_input.c new file mode 100644 index 00000000..daa0683b --- /dev/null +++ b/src/vnet/tcp/tcp_input.c @@ -0,0 +1,2316 @@ +/* + * Copyright (c) 2016 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +#include +#include +#include +#include +#include + +static char *tcp_error_strings[] = { +#define tcp_error(n,s) s, +#include +#undef tcp_error +}; + +/* All TCP nodes have the same outgoing arcs */ +#define foreach_tcp_state_next \ + _ (DROP, "error-drop") \ + _ (TCP4_OUTPUT, "tcp4-output") \ + _ (TCP6_OUTPUT, "tcp6-output") + +typedef enum _tcp_established_next +{ +#define _(s,n) TCP_ESTABLISHED_NEXT_##s, + foreach_tcp_state_next +#undef _ + TCP_ESTABLISHED_N_NEXT, +} tcp_established_next_t; + +typedef enum _tcp_rcv_process_next +{ +#define _(s,n) TCP_RCV_PROCESS_NEXT_##s, + foreach_tcp_state_next +#undef _ + TCP_RCV_PROCESS_N_NEXT, +} tcp_rcv_process_next_t; + +typedef enum _tcp_syn_sent_next +{ +#define _(s,n) TCP_SYN_SENT_NEXT_##s, + foreach_tcp_state_next +#undef _ + TCP_SYN_SENT_N_NEXT, +} tcp_syn_sent_next_t; + +typedef enum _tcp_listen_next +{ +#define _(s,n) TCP_LISTEN_NEXT_##s, + foreach_tcp_state_next +#undef _ + TCP_LISTEN_N_NEXT, +} tcp_listen_next_t; + +/* Generic, state independent indices */ +typedef enum _tcp_state_next +{ +#define _(s,n) TCP_NEXT_##s, + foreach_tcp_state_next +#undef _ + TCP_STATE_N_NEXT, +} tcp_state_next_t; + +#define tcp_next_output(is_ip4) (is_ip4 ? TCP_NEXT_TCP4_OUTPUT \ + : TCP_NEXT_TCP6_OUTPUT) + +vlib_node_registration_t tcp4_established_node; +vlib_node_registration_t tcp6_established_node; + +/** + * Validate segment sequence number. As per RFC793: + * + * Segment Receive Test + * Length Window + * ------- ------- ------------------------------------------- + * 0 0 SEG.SEQ = RCV.NXT + * 0 >0 RCV.NXT =< SEG.SEQ < RCV.NXT+RCV.WND + * >0 0 not acceptable + * >0 >0 RCV.NXT =< SEG.SEQ < RCV.NXT+RCV.WND + * or RCV.NXT =< SEG.SEQ+SEG.LEN-1 < RCV.NXT+RCV.WND + * + * This ultimately consists in checking if segment falls within the window. + * The one important difference compared to RFC793 is that we use rcv_las, + * or the rcv_nxt at last ack sent instead of rcv_nxt since that's the + * peer's reference when computing our receive window. + * + * This accepts only segments within the window. + */ +always_inline u8 +tcp_segment_in_rcv_wnd (tcp_connection_t * tc, u32 seq, u32 end_seq) +{ + return seq_leq (end_seq, tc->rcv_las + tc->rcv_wnd) + && seq_geq (seq, tc->rcv_nxt); +} + +void +tcp_options_parse (tcp_header_t * th, tcp_options_t * to) +{ + const u8 *data; + u8 opt_len, opts_len, kind; + int j; + sack_block_t b; + + opts_len = (tcp_doff (th) << 2) - sizeof (tcp_header_t); + data = (const u8 *) (th + 1); + + /* Zero out all flags but those set in SYN */ + to->flags &= (TCP_OPTS_FLAG_SACK_PERMITTED | TCP_OPTS_FLAG_WSCALE); + + for (; opts_len > 0; opts_len -= opt_len, data += opt_len) + { + kind = data[0]; + + /* Get options length */ + if (kind == TCP_OPTION_EOL) + break; + else if (kind == TCP_OPTION_NOOP) + opt_len = 1; + else + { + /* broken options */ + if (opts_len < 2) + break; + opt_len = data[1]; + + /* weird option length */ + if (opt_len < 2 || opt_len > opts_len) + break; + } + + /* Parse options */ + switch (kind) + { + case TCP_OPTION_MSS: + if ((opt_len == TCP_OPTION_LEN_MSS) && tcp_syn (th)) + { + to->flags |= TCP_OPTS_FLAG_MSS; + to->mss = clib_net_to_host_u16 (*(u16 *) (data + 2)); + } + break; + case TCP_OPTION_WINDOW_SCALE: + if ((opt_len == TCP_OPTION_LEN_WINDOW_SCALE) && tcp_syn (th)) + { + to->flags |= TCP_OPTS_FLAG_WSCALE; + to->wscale = data[2]; + if (to->wscale > TCP_MAX_WND_SCALE) + { + clib_warning ("Illegal window scaling value: %d", + to->wscale); + to->wscale = TCP_MAX_WND_SCALE; + } + } + break; + case TCP_OPTION_TIMESTAMP: + if (opt_len == TCP_OPTION_LEN_TIMESTAMP) + { + to->flags |= TCP_OPTS_FLAG_TSTAMP; + to->tsval = clib_net_to_host_u32 (*(u32 *) (data + 2)); + to->tsecr = clib_net_to_host_u32 (*(u32 *) (data + 6)); + } + break; + case TCP_OPTION_SACK_PERMITTED: + if (opt_len == TCP_OPTION_LEN_SACK_PERMITTED && tcp_syn (th)) + to->flags |= TCP_OPTS_FLAG_SACK_PERMITTED; + break; + case TCP_OPTION_SACK_BLOCK: + /* If SACK permitted was not advertised or a SYN, break */ + if ((to->flags & TCP_OPTS_FLAG_SACK_PERMITTED) == 0 || tcp_syn (th)) + break; + + /* If too short or not correctly formatted, break */ + if (opt_len < 10 || ((opt_len - 2) % TCP_OPTION_LEN_SACK_BLOCK)) + break; + + to->flags |= TCP_OPTS_FLAG_SACK; + to->n_sack_blocks = (opt_len - 2) / TCP_OPTION_LEN_SACK_BLOCK; + vec_reset_length (to->sacks); + for (j = 0; j < to->n_sack_blocks; j++) + { + b.start = clib_net_to_host_u32 (*(u32 *) (data + 2 + 4 * j)); + b.end = clib_net_to_host_u32 (*(u32 *) (data + 6 + 4 * j)); + vec_add1 (to->sacks, b); + } + break; + default: + /* Nothing to see here */ + continue; + } + } +} + +always_inline int +tcp_segment_check_paws (tcp_connection_t * tc) +{ + /* XXX normally test for timestamp should be lt instead of leq, but for + * local testing this is not enough */ + return tcp_opts_tstamp (&tc->opt) && tc->tsval_recent + && timestamp_lt (tc->opt.tsval, tc->tsval_recent); +} + +/** + * Validate incoming segment as per RFC793 p. 69 and RFC1323 p. 19 + * + * It first verifies if segment has a wrapped sequence number (PAWS) and then + * does the processing associated to the first four steps (ignoring security + * and precedence): sequence number, rst bit and syn bit checks. + * + * @return 0 if segments passes validation. + */ +static int +tcp_segment_validate (vlib_main_t * vm, tcp_connection_t * tc0, + vlib_buffer_t * b0, tcp_header_t * th0, u32 * next0) +{ + u8 paws_failed; + + if (PREDICT_FALSE (!tcp_ack (th0) && !tcp_rst (th0) && !tcp_syn (th0))) + return -1; + + tcp_options_parse (th0, &tc0->opt); + + /* RFC1323: Check against wrapped sequence numbers (PAWS). If we have + * timestamp to echo and it's less than tsval_recent, drop segment + * but still send an ACK in order to retain TCP's mechanism for detecting + * and recovering from half-open connections */ + paws_failed = tcp_segment_check_paws (tc0); + if (paws_failed) + { + clib_warning ("paws failed"); + + /* If it just so happens that a segment updates tsval_recent for a + * segment over 24 days old, invalidate tsval_recent. */ + if (timestamp_lt (tc0->tsval_recent_age + TCP_PAWS_IDLE, + tcp_time_now ())) + { + /* Age isn't reset until we get a valid tsval (bsd inspired) */ + tc0->tsval_recent = 0; + } + else + { + /* Drop after ack if not rst */ + if (!tcp_rst (th0)) + { + tcp_make_ack (tc0, b0); + *next0 = tcp_next_output (tc0->c_is_ip4); + return -1; + } + } + } + + /* 1st: check sequence number */ + if (!tcp_segment_in_rcv_wnd (tc0, vnet_buffer (b0)->tcp.seq_number, + vnet_buffer (b0)->tcp.seq_end)) + { + if (!tcp_rst (th0)) + { + /* Send dup ack */ + tcp_make_ack (tc0, b0); + *next0 = tcp_next_output (tc0->c_is_ip4); + } + return -1; + } + + /* 2nd: check the RST bit */ + if (tcp_rst (th0)) + { + /* Notify session that connection has been reset. Switch + * state to closed and await for session to do the cleanup. */ + stream_session_reset_notify (&tc0->connection); + tc0->state = TCP_STATE_CLOSED; + return -1; + } + + /* 3rd: check security and precedence (skip) */ + + /* 4th: check the SYN bit */ + if (tcp_syn (th0)) + { + tcp_send_reset (b0, tc0->c_is_ip4); + return -1; + } + + /* If PAWS passed and segment in window, save timestamp */ + if (!paws_failed) + { + tc0->tsval_recent = tc0->opt.tsval; + tc0->tsval_recent_age = tcp_time_now (); + } + + return 0; +} + +always_inline int +tcp_rcv_ack_is_acceptable (tcp_connection_t * tc0, vlib_buffer_t * tb0) +{ + /* SND.UNA =< SEG.ACK =< SND.NXT */ + return (seq_leq (tc0->snd_una, vnet_buffer (tb0)->tcp.ack_number) + && seq_leq (vnet_buffer (tb0)->tcp.ack_number, tc0->snd_nxt)); +} + +/** + * Compute smoothed RTT as per VJ's '88 SIGCOMM and RFC6298 + * + * Note that although the original article, srtt and rttvar are scaled + * to minimize round-off errors, here we don't. Instead, we rely on + * better precision time measurements. + * + * TODO support us rtt resolution + */ +static void +tcp_estimate_rtt (tcp_connection_t * tc, u32 mrtt) +{ + int err; + + if (tc->srtt != 0) + { + err = mrtt - tc->srtt; + tc->srtt += err >> 3; + + /* XXX Drop in RTT results in RTTVAR increase and bigger RTO. + * The increase should be bound */ + tc->rttvar += (clib_abs (err) - tc->rttvar) >> 2; + } + else + { + /* First measurement. */ + tc->srtt = mrtt; + tc->rttvar = mrtt << 1; + } +} + +/** Update RTT estimate and RTO timer + * + * Measure RTT: We have two sources of RTT measurements: TSOPT and ACK + * timing. Middle boxes are known to fiddle with TCP options so we + * should give higher priority to ACK timing. + * + * return 1 if valid rtt 0 otherwise + */ +static int +tcp_update_rtt (tcp_connection_t * tc, u32 ack) +{ + u32 mrtt = 0; + + /* Karn's rule, part 1. Don't use retransmitted segments to estimate + * RTT because they're ambiguous. */ + if (tc->rtt_seq && seq_gt (ack, tc->rtt_seq) && !tc->rto_boff) + { + mrtt = tcp_time_now () - tc->rtt_ts; + tc->rtt_seq = 0; + } + + /* As per RFC7323 TSecr can be used for RTTM only if the segment advances + * snd_una, i.e., the left side of the send window: + * seq_lt (tc->snd_una, ack). Note: last condition could be dropped, we don't + * try to update rtt for dupacks */ + else if (tcp_opts_tstamp (&tc->opt) && tc->opt.tsecr && tc->bytes_acked) + { + mrtt = tcp_time_now () - tc->opt.tsecr; + } + + /* Ignore dubious measurements */ + if (mrtt == 0 || mrtt > TCP_RTT_MAX) + return 0; + + tcp_estimate_rtt (tc, mrtt); + + tc->rto = clib_min (tc->srtt + (tc->rttvar << 2), TCP_RTO_MAX); + + return 1; +} + +/** + * Dequeue bytes that have been acked and while at it update RTT estimates. + */ +static void +tcp_dequeue_acked (tcp_connection_t * tc, u32 ack) +{ + /* Dequeue the newly ACKed bytes */ + stream_session_dequeue_drop (&tc->connection, tc->bytes_acked); + + /* Update rtt and rto */ + if (tcp_update_rtt (tc, ack)) + { + /* Good ACK received and valid RTT, make sure retransmit backoff is 0 */ + tc->rto_boff = 0; + } +} + +/** Check if dupack as per RFC5681 Sec. 2 */ +always_inline u8 +tcp_ack_is_dupack (tcp_connection_t * tc, vlib_buffer_t * b, u32 new_snd_wnd) +{ + return ((vnet_buffer (b)->tcp.ack_number == tc->snd_una) + && seq_gt (tc->snd_una_max, tc->snd_una) + && (vnet_buffer (b)->tcp.seq_end == vnet_buffer (b)->tcp.seq_number) + && (new_snd_wnd == tc->snd_wnd)); +} + +void +scoreboard_remove_hole (sack_scoreboard_t * sb, sack_scoreboard_hole_t * hole) +{ + sack_scoreboard_hole_t *next, *prev; + + if (hole->next != TCP_INVALID_SACK_HOLE_INDEX) + { + next = pool_elt_at_index (sb->holes, hole->next); + next->prev = hole->prev; + } + + if (hole->prev != TCP_INVALID_SACK_HOLE_INDEX) + { + prev = pool_elt_at_index (sb->holes, hole->prev); + prev->next = hole->next; + } + else + { + sb->head = hole->next; + } + + pool_put (sb->holes, hole); +} + +sack_scoreboard_hole_t * +scoreboard_insert_hole (sack_scoreboard_t * sb, sack_scoreboard_hole_t * prev, + u32 start, u32 end) +{ + sack_scoreboard_hole_t *hole, *next; + u32 hole_index; + + pool_get (sb->holes, hole); + memset (hole, 0, sizeof (*hole)); + + hole->start = start; + hole->end = end; + hole_index = hole - sb->holes; + + if (prev) + { + hole->prev = prev - sb->holes; + hole->next = prev->next; + + if ((next = scoreboard_next_hole (sb, hole))) + next->prev = hole_index; + + prev->next = hole_index; + } + else + { + sb->head = hole_index; + hole->prev = TCP_INVALID_SACK_HOLE_INDEX; + hole->next = TCP_INVALID_SACK_HOLE_INDEX; + } + + return hole; +} + +static void +tcp_rcv_sacks (tcp_connection_t * tc, u32 ack) +{ + sack_scoreboard_t *sb = &tc->sack_sb; + sack_block_t *blk, tmp; + sack_scoreboard_hole_t *hole, *next_hole; + u32 blk_index = 0; + int i, j; + + if (!tcp_opts_sack (tc) && sb->head == TCP_INVALID_SACK_HOLE_INDEX) + return; + + /* Remove invalid blocks */ + vec_foreach (blk, tc->opt.sacks) + { + if (seq_lt (blk->start, blk->end) + && seq_gt (blk->start, tc->snd_una) + && seq_gt (blk->start, ack) && seq_lt (blk->end, tc->snd_nxt)) + continue; + + vec_del1 (tc->opt.sacks, blk - tc->opt.sacks); + } + + /* Add block for cumulative ack */ + if (seq_gt (ack, tc->snd_una)) + { + tmp.start = tc->snd_una; + tmp.end = ack; + vec_add1 (tc->opt.sacks, tmp); + } + + if (vec_len (tc->opt.sacks) == 0) + return; + + /* Make sure blocks are ordered */ + for (i = 0; i < vec_len (tc->opt.sacks); i++) + for (j = i; j < vec_len (tc->opt.sacks); j++) + if (seq_lt (tc->opt.sacks[j].start, tc->opt.sacks[i].start)) + { + tmp = tc->opt.sacks[i]; + tc->opt.sacks[i] = tc->opt.sacks[j]; + tc->opt.sacks[j] = tmp; + } + + /* If no holes, insert the first that covers all outstanding bytes */ + if (sb->head == TCP_INVALID_SACK_HOLE_INDEX) + { + scoreboard_insert_hole (sb, 0, tc->snd_una, tc->snd_una_max); + } + + /* Walk the holes with the SACK blocks */ + hole = pool_elt_at_index (sb->holes, sb->head); + while (hole && blk_index < vec_len (tc->opt.sacks)) + { + blk = &tc->opt.sacks[blk_index]; + + if (seq_leq (blk->start, hole->start)) + { + /* Block covers hole. Remove hole */ + if (seq_geq (blk->end, hole->end)) + { + next_hole = scoreboard_next_hole (sb, hole); + + /* Byte accounting */ + if (seq_lt (hole->end, ack)) + { + /* Bytes lost because snd wnd left edge advances */ + if (seq_lt (next_hole->start, ack)) + sb->sacked_bytes -= next_hole->start - hole->end; + else + sb->sacked_bytes -= ack - hole->end; + } + else + { + sb->sacked_bytes += scoreboard_hole_bytes (hole); + } + + scoreboard_remove_hole (sb, hole); + hole = next_hole; + } + /* Partial overlap */ + else + { + sb->sacked_bytes += blk->end - hole->start; + hole->start = blk->end; + blk_index++; + } + } + else + { + /* Hole must be split */ + if (seq_leq (blk->end, hole->end)) + { + sb->sacked_bytes += blk->end - blk->start; + scoreboard_insert_hole (sb, hole, blk->end, hole->end); + hole->end = blk->start - 1; + blk_index++; + } + else + { + sb->sacked_bytes += hole->end - blk->start + 1; + hole->end = blk->start - 1; + hole = scoreboard_next_hole (sb, hole); + } + } + } +} + +/** Update snd_wnd + * + * If (SND.WL1 < SEG.SEQ or (SND.WL1 = SEG.SEQ and SND.WL2 =< SEG.ACK)), set + * SND.WND <- SEG.WND, set SND.WL1 <- SEG.SEQ, and set SND.WL2 <- SEG.ACK */ +static void +tcp_update_snd_wnd (tcp_connection_t * tc, u32 seq, u32 ack, u32 snd_wnd) +{ + if (tc->snd_wl1 < seq || (tc->snd_wl1 == seq && tc->snd_wl2 <= ack)) + { + tc->snd_wnd = snd_wnd; + tc->snd_wl1 = seq; + tc->snd_wl2 = ack; + } +} + +static void +tcp_cc_congestion (tcp_connection_t * tc) +{ + tc->cc_algo->congestion (tc); +} + +static void +tcp_cc_recover (tcp_connection_t * tc) +{ + if (tcp_in_fastrecovery (tc)) + { + tc->cc_algo->recovered (tc); + tcp_recovery_off (tc); + } + else if (tcp_in_recovery (tc)) + { + tcp_recovery_off (tc); + tc->cwnd = tcp_loss_wnd (tc); + } +} + +static void +tcp_cc_rcv_ack (tcp_connection_t * tc) +{ + u8 partial_ack; + + if (tcp_in_recovery (tc)) + { + partial_ack = seq_lt (tc->snd_una, tc->snd_una_max); + if (!partial_ack) + { + /* Clear retransmitted bytes. */ + tc->rtx_bytes = 0; + tcp_cc_recover (tc); + } + else + { + /* Clear retransmitted bytes. XXX should we clear all? */ + tc->rtx_bytes = 0; + tc->cc_algo->rcv_cong_ack (tc, TCP_CC_PARTIALACK); + + /* Retransmit first unacked segment */ + tcp_retransmit_first_unacked (tc); + } + } + else + { + tc->cc_algo->rcv_ack (tc); + } + + tc->rcv_dupacks = 0; + tc->tsecr_last_ack = tc->opt.tsecr; +} + +static void +tcp_cc_rcv_dupack (tcp_connection_t * tc, u32 ack) +{ + ASSERT (tc->snd_una == ack); + + tc->rcv_dupacks++; + if (tc->rcv_dupacks == TCP_DUPACK_THRESHOLD) + { + /* RFC6582 NewReno heuristic to avoid multiple fast retransmits */ + if (tc->opt.tsecr != tc->tsecr_last_ack) + { + tc->rcv_dupacks = 0; + return; + } + + tcp_fastrecovery_on (tc); + + /* Handle congestion and dupack */ + tcp_cc_congestion (tc); + tc->cc_algo->rcv_cong_ack (tc, TCP_CC_DUPACK); + + tcp_fast_retransmit (tc); + + /* Post retransmit update cwnd to ssthresh and account for the + * three segments that have left the network and should've been + * buffered at the receiver */ + tc->cwnd = tc->ssthresh + TCP_DUPACK_THRESHOLD * tc->snd_mss; + } + else if (tc->rcv_dupacks > TCP_DUPACK_THRESHOLD) + { + ASSERT (tcp_in_fastrecovery (tc)); + + tc->cc_algo->rcv_cong_ack (tc, TCP_CC_DUPACK); + } +} + +void +tcp_cc_init (tcp_connection_t * tc) +{ + tc->cc_algo = tcp_cc_algo_get (TCP_CC_NEWRENO); + tc->cc_algo->init (tc); +} + +static int +tcp_rcv_ack (tcp_connection_t * tc, vlib_buffer_t * b, + tcp_header_t * th, u32 * next, u32 * error) +{ + u32 new_snd_wnd; + + /* If the ACK acks something not yet sent (SEG.ACK > SND.NXT) then send an + * ACK, drop the segment, and return */ + if (seq_gt (vnet_buffer (b)->tcp.ack_number, tc->snd_nxt)) + { + tcp_make_ack (tc, b); + *next = tcp_next_output (tc->c_is_ip4); + *error = TCP_ERROR_ACK_INVALID; + return -1; + } + + /* If old ACK, discard */ + if (seq_lt (vnet_buffer (b)->tcp.ack_number, tc->snd_una)) + { + *error = TCP_ERROR_ACK_OLD; + return -1; + } + + if (tcp_opts_sack_permitted (&tc->opt)) + tcp_rcv_sacks (tc, vnet_buffer (b)->tcp.ack_number); + + new_snd_wnd = clib_net_to_host_u32 (th->window) << tc->snd_wscale; + + if (tcp_ack_is_dupack (tc, b, new_snd_wnd)) + { + tcp_cc_rcv_dupack (tc, vnet_buffer (b)->tcp.ack_number); + *error = TCP_ERROR_ACK_DUP; + return -1; + } + + /* Valid ACK */ + tc->bytes_acked = vnet_buffer (b)->tcp.ack_number - tc->snd_una; + tc->snd_una = vnet_buffer (b)->tcp.ack_number; + + /* Dequeue ACKed packet and update RTT */ + tcp_dequeue_acked (tc, vnet_buffer (b)->tcp.ack_number); + + tcp_update_snd_wnd (tc, vnet_buffer (b)->tcp.seq_number, + vnet_buffer (b)->tcp.ack_number, new_snd_wnd); + + /* Updates congestion control (slow start/congestion avoidance) */ + tcp_cc_rcv_ack (tc); + + /* If everything has been acked, stop retransmit timer + * otherwise update */ + if (tc->snd_una == tc->snd_una_max) + tcp_timer_reset (tc, TCP_TIMER_RETRANSMIT); + else + tcp_timer_update (tc, TCP_TIMER_RETRANSMIT, tc->rto); + + return 0; +} + +/** + * Build SACK list as per RFC2018. + * + * Makes sure the first block contains the segment that generated the current + * ACK and the following ones are the ones most recently reported in SACK + * blocks. + * + * @param tc TCP connection for which the SACK list is updated + * @param start Start sequence number of the newest SACK block + * @param end End sequence of the newest SACK block + */ +static void +tcp_update_sack_list (tcp_connection_t * tc, u32 start, u32 end) +{ + sack_block_t *new_list = 0, block; + u32 n_elts; + int i; + u8 new_head = 0; + + /* If the first segment is ooo add it to the list. Last write might've moved + * rcv_nxt over the first segment. */ + if (seq_lt (tc->rcv_nxt, start)) + { + block.start = start; + block.end = end; + vec_add1 (new_list, block); + new_head = 1; + } + + /* Find the blocks still worth keeping. */ + for (i = 0; i < vec_len (tc->snd_sacks); i++) + { + /* Discard if: + * 1) rcv_nxt advanced beyond current block OR + * 2) Segment overlapped by the first segment, i.e., it has been merged + * into it.*/ + if (seq_leq (tc->snd_sacks[i].start, tc->rcv_nxt) + || seq_leq (tc->snd_sacks[i].start, end)) + continue; + + /* Save subsequent segments to new SACK list. */ + n_elts = clib_min (vec_len (tc->snd_sacks) - i, + TCP_MAX_SACK_BLOCKS - new_head); + vec_insert_elts (new_list, &tc->snd_sacks[i], n_elts, new_head); + break; + } + + /* Replace old vector with new one */ + vec_free (tc->snd_sacks); + tc->snd_sacks = new_list; +} + +/** Enqueue data for delivery to application */ +always_inline u32 +tcp_session_enqueue_data (tcp_connection_t * tc, vlib_buffer_t * b, + u16 data_len) +{ + int written; + + /* Pure ACK. Update rcv_nxt and be done. */ + if (PREDICT_FALSE (data_len == 0)) + { + tc->rcv_nxt = vnet_buffer (b)->tcp.seq_end; + return TCP_ERROR_PURE_ACK; + } + + written = stream_session_enqueue_data (&tc->connection, + vlib_buffer_get_current (b), + data_len, 1 /* queue event */ ); + + /* Update rcv_nxt */ + if (PREDICT_TRUE (written == data_len)) + { + tc->rcv_nxt = vnet_buffer (b)->tcp.seq_end; + } + /* If more data written than expected, account for out-of-order bytes. */ + else if (written > data_len) + { + tc->rcv_nxt = vnet_buffer (b)->tcp.seq_end + written - data_len; + + /* Send ACK confirming the update */ + tc->flags |= TCP_CONN_SNDACK; + + /* Update SACK list if need be */ + if (tcp_opts_sack_permitted (&tc->opt)) + { + /* Remove SACK blocks that have been delivered */ + tcp_update_sack_list (tc, tc->rcv_nxt, tc->rcv_nxt); + } + } + else + { + ASSERT (0); + return TCP_ERROR_FIFO_FULL; + } + + return TCP_ERROR_ENQUEUED; +} + +/** Enqueue out-of-order data */ +always_inline u32 +tcp_session_enqueue_ooo (tcp_connection_t * tc, vlib_buffer_t * b, + u16 data_len) +{ + stream_session_t *s0; + u32 offset, seq; + + s0 = stream_session_get (tc->c_s_index, tc->c_thread_index); + seq = vnet_buffer (b)->tcp.seq_number; + offset = seq - tc->rcv_nxt; + + if (svm_fifo_enqueue_with_offset (s0->server_rx_fifo, s0->pid, offset, + data_len, vlib_buffer_get_current (b))) + return TCP_ERROR_FIFO_FULL; + + /* Update SACK list if in use */ + if (tcp_opts_sack_permitted (&tc->opt)) + { + ooo_segment_t *newest; + u32 start, end; + + /* Get the newest segment from the fifo */ + newest = svm_fifo_newest_ooo_segment (s0->server_rx_fifo); + start = tc->rcv_nxt + ooo_segment_offset (s0->server_rx_fifo, newest); + end = tc->rcv_nxt + ooo_segment_end_offset (s0->server_rx_fifo, newest); + + tcp_update_sack_list (tc, start, end); + } + + return TCP_ERROR_ENQUEUED; +} + +/** + * Check if ACK could be delayed. DELACK timer is set only after frame is + * processed so this can return true for a full bursts of packets. + */ +always_inline int +tcp_can_delack (tcp_connection_t * tc) +{ + /* If there's no DELACK timer set and the last window sent wasn't 0 we + * can safely delay. */ + if (!tcp_timer_is_active (tc, TCP_TIMER_DELACK) + && (tc->flags & TCP_CONN_SENT_RCV_WND0) == 0 + && (tc->flags & TCP_CONN_SNDACK) == 0) + return 1; + + return 0; +} + +static int +tcp_segment_rcv (tcp_main_t * tm, tcp_connection_t * tc, vlib_buffer_t * b, + u16 n_data_bytes, u32 * next0) +{ + u32 error = 0; + + /* Handle out-of-order data */ + if (PREDICT_FALSE (vnet_buffer (b)->tcp.seq_number != tc->rcv_nxt)) + { + error = tcp_session_enqueue_ooo (tc, b, n_data_bytes); + + /* Don't send more than 3 dupacks per burst + * XXX decide if this is good */ + if (tc->snt_dupacks < 3) + { + /* RFC2581: Send DUPACK for fast retransmit */ + tcp_make_ack (tc, b); + *next0 = tcp_next_output (tc->c_is_ip4); + + /* Mark as DUPACK. We may filter these in output if + * the burst fills the holes. */ + vnet_buffer (b)->tcp.flags = TCP_BUF_FLAG_DUPACK; + + tc->snt_dupacks++; + } + + goto done; + } + + /* In order data, enqueue. Fifo figures out by itself if any out-of-order + * segments can be enqueued after fifo tail offset changes. */ + error = tcp_session_enqueue_data (tc, b, n_data_bytes); + + /* Check if ACK can be delayed */ + if (tcp_can_delack (tc)) + { + /* Nothing to do for pure ACKs */ + if (n_data_bytes == 0) + goto done; + + /* If connection has not been previously marked for delay ack + * add it to the list and flag it */ + if (!tc->flags & TCP_CONN_DELACK) + { + vec_add1 (tm->delack_connections[tc->c_thread_index], + tc->c_c_index); + tc->flags |= TCP_CONN_DELACK; + } + } + else + { + /* Check if a packet has already been enqueued to output for burst. + * If yes, then drop this one, otherwise, let it pass through to + * output */ + if ((tc->flags & TCP_CONN_BURSTACK) == 0) + { + *next0 = tcp_next_output (tc->c_is_ip4); + tcp_make_ack (tc, b); + error = TCP_ERROR_ENQUEUED; + + /* TODO: maybe add counter to ensure N acks will be sent/burst */ + tc->flags |= TCP_CONN_BURSTACK; + } + } + +done: + return error; +} + +void +delack_timers_init (tcp_main_t * tm, u32 thread_index) +{ + tcp_connection_t *tc; + u32 i, *conns; + tw_timer_wheel_16t_2w_512sl_t *tw; + + tw = &tm->timer_wheels[thread_index]; + conns = tm->delack_connections[thread_index]; + for (i = 0; i < vec_len (conns); i++) + { + tc = pool_elt_at_index (tm->connections[thread_index], conns[i]); + ASSERT (0 != tc); + + tc->timers[TCP_TIMER_DELACK] + = tw_timer_start_16t_2w_512sl (tw, conns[i], + TCP_TIMER_DELACK, TCP_DELACK_TIME); + } + vec_reset_length (tm->delack_connections[thread_index]); +} + +always_inline uword +tcp46_established_inline (vlib_main_t * vm, vlib_node_runtime_t * node, + vlib_frame_t * from_frame, int is_ip4) +{ + u32 n_left_from, next_index, *from, *to_next; + u32 my_thread_index = vm->cpu_index, errors = 0; + tcp_main_t *tm = vnet_get_tcp_main (); + + from = vlib_frame_vector_args (from_frame); + n_left_from = from_frame->n_vectors; + + next_index = node->cached_next_index; + + while (n_left_from > 0) + { + u32 n_left_to_next; + + vlib_get_next_frame (vm, node, next_index, to_next, n_left_to_next); + + while (n_left_from > 0 && n_left_to_next > 0) + { + u32 bi0; + vlib_buffer_t *b0; + tcp_header_t *th0 = 0; + tcp_connection_t *tc0; + ip4_header_t *ip40; + ip6_header_t *ip60; + u32 n_advance_bytes0, n_data_bytes0; + u32 next0 = TCP_ESTABLISHED_NEXT_DROP, error0 = TCP_ERROR_ENQUEUED; + + bi0 = from[0]; + to_next[0] = bi0; + from += 1; + to_next += 1; + n_left_from -= 1; + n_left_to_next -= 1; + + b0 = vlib_get_buffer (vm, bi0); + tc0 = tcp_connection_get (vnet_buffer (b0)->tcp.connection_index, + my_thread_index); + + /* Checksum computed by ipx_local no need to compute again */ + + if (is_ip4) + { + ip40 = vlib_buffer_get_current (b0); + th0 = ip4_next_header (ip40); + n_advance_bytes0 = (ip4_header_bytes (ip40) + + tcp_header_bytes (th0)); + n_data_bytes0 = clib_net_to_host_u16 (ip40->length) + - n_advance_bytes0; + } + else + { + ip60 = vlib_buffer_get_current (b0); + th0 = ip6_next_header (ip60); + n_advance_bytes0 = tcp_header_bytes (th0); + n_data_bytes0 = clib_net_to_host_u16 (ip60->payload_length) + - n_advance_bytes0; + n_advance_bytes0 += sizeof (ip60[0]); + } + + /* SYNs, FINs and data consume sequence numbers */ + vnet_buffer (b0)->tcp.seq_end = vnet_buffer (b0)->tcp.seq_number + + tcp_is_syn (th0) + tcp_is_fin (th0) + n_data_bytes0; + + /* TODO header prediction fast path */ + + /* 1-4: check SEQ, RST, SYN */ + if (PREDICT_FALSE (tcp_segment_validate (vm, tc0, b0, th0, &next0))) + { + error0 = TCP_ERROR_SEGMENT_INVALID; + goto drop; + } + + /* 5: check the ACK field */ + if (tcp_rcv_ack (tc0, b0, th0, &next0, &error0)) + { + goto drop; + } + + /* 6: check the URG bit TODO */ + + /* 7: process the segment text */ + vlib_buffer_advance (b0, n_advance_bytes0); + error0 = tcp_segment_rcv (tm, tc0, b0, n_data_bytes0, &next0); + + /* 8: check the FIN bit */ + if (tcp_fin (th0)) + { + /* Send ACK and enter CLOSE-WAIT */ + tcp_make_ack (tc0, b0); + tcp_connection_force_ack (tc0, b0); + next0 = tcp_next_output (tc0->c_is_ip4); + tc0->state = TCP_STATE_CLOSE_WAIT; + stream_session_disconnect_notify (&tc0->connection); + } + + drop: + b0->error = node->errors[error0]; + if (PREDICT_FALSE (b0->flags & VLIB_BUFFER_IS_TRACED)) + { + + } + + vlib_validate_buffer_enqueue_x1 (vm, node, next_index, to_next, + n_left_to_next, bi0, next0); + } + + vlib_put_next_frame (vm, node, next_index, n_left_to_next); + } + + errors = session_manager_flush_enqueue_events (my_thread_index); + if (errors) + { + if (is_ip4) + vlib_node_increment_counter (vm, tcp4_established_node.index, + TCP_ERROR_EVENT_FIFO_FULL, errors); + else + vlib_node_increment_counter (vm, tcp6_established_node.index, + TCP_ERROR_EVENT_FIFO_FULL, errors); + } + + delack_timers_init (tm, my_thread_index); + + return from_frame->n_vectors; +} + +static uword +tcp4_established (vlib_main_t * vm, vlib_node_runtime_t * node, + vlib_frame_t * from_frame) +{ + return tcp46_established_inline (vm, node, from_frame, 1 /* is_ip4 */ ); +} + +static uword +tcp6_established (vlib_main_t * vm, vlib_node_runtime_t * node, + vlib_frame_t * from_frame) +{ + return tcp46_established_inline (vm, node, from_frame, 0 /* is_ip4 */ ); +} + +/* *INDENT-OFF* */ +VLIB_REGISTER_NODE (tcp4_established_node) = +{ + .function = tcp4_established, + .name = "tcp4-established", + /* Takes a vector of packets. */ + .vector_size = sizeof (u32), + .n_errors = TCP_N_ERROR,.error_strings = tcp_error_strings, + .n_next_nodes = TCP_ESTABLISHED_N_NEXT, + .next_nodes = + { +#define _(s,n) [TCP_ESTABLISHED_NEXT_##s] = n, + foreach_tcp_state_next +#undef _ + }, +}; +/* *INDENT-ON* */ + +VLIB_NODE_FUNCTION_MULTIARCH (tcp4_established_node, tcp4_established); + +/* *INDENT-OFF* */ +VLIB_REGISTER_NODE (tcp6_established_node) = +{ + .function = tcp6_established, + .name = "tcp6-established", + /* Takes a vector of packets. */ + .vector_size = sizeof (u32), + .n_errors = TCP_N_ERROR, + .error_strings = tcp_error_strings, + .n_next_nodes = TCP_ESTABLISHED_N_NEXT, + .next_nodes = + { +#define _(s,n) [TCP_ESTABLISHED_NEXT_##s] = n, + foreach_tcp_state_next +#undef _ + }, +}; +/* *INDENT-ON* */ + + +VLIB_NODE_FUNCTION_MULTIARCH (tcp6_established_node, tcp6_established); + +vlib_node_registration_t tcp4_syn_sent_node; +vlib_node_registration_t tcp6_syn_sent_node; + +always_inline uword +tcp46_syn_sent_inline (vlib_main_t * vm, vlib_node_runtime_t * node, + vlib_frame_t * from_frame, int is_ip4) +{ + tcp_main_t *tm = vnet_get_tcp_main (); + u32 n_left_from, next_index, *from, *to_next; + u32 my_thread_index = vm->cpu_index, errors = 0; + u8 sst = is_ip4 ? SESSION_TYPE_IP4_TCP : SESSION_TYPE_IP6_TCP; + + from = vlib_frame_vector_args (from_frame); + n_left_from = from_frame->n_vectors; + + next_index = node->cached_next_index; + + while (n_left_from > 0) + { + u32 n_left_to_next; + + vlib_get_next_frame (vm, node, next_index, to_next, n_left_to_next); + + while (n_left_from > 0 && n_left_to_next > 0) + { + u32 bi0, ack0, seq0; + vlib_buffer_t *b0; + tcp_header_t *tcp0 = 0; + tcp_connection_t *tc0; + ip4_header_t *ip40; + ip6_header_t *ip60; + u32 n_advance_bytes0, n_data_bytes0; + tcp_connection_t *new_tc0; + u32 next0 = TCP_SYN_SENT_NEXT_DROP, error0 = TCP_ERROR_ENQUEUED; + + bi0 = from[0]; + to_next[0] = bi0; + from += 1; + to_next += 1; + n_left_from -= 1; + n_left_to_next -= 1; + + b0 = vlib_get_buffer (vm, bi0); + tc0 = + tcp_half_open_connection_get (vnet_buffer (b0)-> + tcp.connection_index); + + ack0 = vnet_buffer (b0)->tcp.ack_number; + seq0 = vnet_buffer (b0)->tcp.seq_number; + + /* Checksum computed by ipx_local no need to compute again */ + + if (is_ip4) + { + ip40 = vlib_buffer_get_current (b0); + tcp0 = ip4_next_header (ip40); + n_advance_bytes0 = (ip4_header_bytes (ip40) + + tcp_header_bytes (tcp0)); + n_data_bytes0 = clib_net_to_host_u16 (ip40->length) + - n_advance_bytes0; + } + else + { + ip60 = vlib_buffer_get_current (b0); + tcp0 = ip6_next_header (ip60); + n_advance_bytes0 = tcp_header_bytes (tcp0); + n_data_bytes0 = clib_net_to_host_u16 (ip60->payload_length) + - n_advance_bytes0; + n_advance_bytes0 += sizeof (ip60[0]); + } + + if (PREDICT_FALSE + (!tcp_ack (tcp0) && !tcp_rst (tcp0) && !tcp_syn (tcp0))) + goto drop; + + /* SYNs, FINs and data consume sequence numbers */ + vnet_buffer (b0)->tcp.seq_end = seq0 + tcp_is_syn (tcp0) + + tcp_is_fin (tcp0) + n_data_bytes0; + + /* + * 1. check the ACK bit + */ + + /* + * If the ACK bit is set + * If SEG.ACK =< ISS, or SEG.ACK > SND.NXT, send a reset (unless + * the RST bit is set, if so drop the segment and return) + * + * and discard the segment. Return. + * If SND.UNA =< SEG.ACK =< SND.NXT then the ACK is acceptable. + */ + if (tcp_ack (tcp0)) + { + if (ack0 <= tc0->iss || ack0 > tc0->snd_nxt) + { + if (!tcp_rst (tcp0)) + tcp_send_reset (b0, is_ip4); + + goto drop; + } + + /* Make sure ACK is valid */ + if (tc0->snd_una > ack0) + goto drop; + } + + /* + * 2. check the RST bit + */ + + if (tcp_rst (tcp0)) + { + /* If ACK is acceptable, signal client that peer is not + * willing to accept connection and drop connection*/ + if (tcp_ack (tcp0)) + { + stream_session_connect_notify (&tc0->connection, sst, + 1 /* fail */ ); + tcp_connection_cleanup (tc0); + } + goto drop; + } + + /* + * 3. check the security and precedence (skipped) + */ + + /* + * 4. check the SYN bit + */ + + /* No SYN flag. Drop. */ + if (!tcp_syn (tcp0)) + goto drop; + + /* Stop connection establishment and retransmit timers */ + tcp_timer_reset (tc0, TCP_TIMER_ESTABLISH); + tcp_timer_reset (tc0, TCP_TIMER_RETRANSMIT_SYN); + + /* Valid SYN or SYN-ACK. Move connection from half-open pool to + * current thread pool. */ + pool_get (tm->connections[my_thread_index], new_tc0); + clib_memcpy (new_tc0, tc0, sizeof (*new_tc0)); + + new_tc0->c_thread_index = my_thread_index; + + /* Cleanup half-open connection XXX lock */ + pool_put (tm->half_open_connections, tc0); + + new_tc0->rcv_nxt = vnet_buffer (b0)->tcp.seq_end; + new_tc0->irs = seq0; + + /* Parse options */ + tcp_options_parse (tcp0, &new_tc0->opt); + tcp_connection_init_vars (new_tc0); + + if (tcp_opts_tstamp (&new_tc0->opt)) + { + new_tc0->tsval_recent = new_tc0->opt.tsval; + new_tc0->tsval_recent_age = tcp_time_now (); + } + + if (tcp_opts_wscale (&new_tc0->opt)) + new_tc0->snd_wscale = new_tc0->opt.wscale; + + new_tc0->snd_wnd = clib_net_to_host_u32 (tcp0->window) + << new_tc0->snd_wscale; + new_tc0->snd_wl1 = seq0; + new_tc0->snd_wl2 = ack0; + + /* SYN-ACK: See if we can switch to ESTABLISHED state */ + if (tcp_ack (tcp0)) + { + /* Our SYN is ACKed: we have iss < ack = snd_una */ + + /* TODO Dequeue acknowledged segments if we support Fast Open */ + new_tc0->snd_una = ack0; + new_tc0->state = TCP_STATE_ESTABLISHED; + + /* Notify app that we have connection */ + stream_session_connect_notify (&new_tc0->connection, sst, 0); + + /* Make sure after data segment processing ACK is sent */ + new_tc0->flags |= TCP_CONN_SNDACK; + } + /* SYN: Simultaneous open. Change state to SYN-RCVD and send SYN-ACK */ + else + { + new_tc0->state = TCP_STATE_SYN_RCVD; + + /* Notify app that we have connection XXX */ + stream_session_connect_notify (&new_tc0->connection, sst, 0); + + tcp_make_synack (new_tc0, b0); + next0 = tcp_next_output (is_ip4); + + goto drop; + } + + /* Read data, if any */ + if (n_data_bytes0) + { + error0 = + tcp_segment_rcv (tm, new_tc0, b0, n_data_bytes0, &next0); + if (error0 == TCP_ERROR_PURE_ACK) + error0 = TCP_ERROR_SYN_ACKS_RCVD; + } + else + { + tcp_make_ack (new_tc0, b0); + next0 = tcp_next_output (new_tc0->c_is_ip4); + } + + drop: + + b0->error = error0 ? node->errors[error0] : 0; + if (PREDICT_FALSE (b0->flags & VLIB_BUFFER_IS_TRACED)) + { + + } + + vlib_validate_buffer_enqueue_x1 (vm, node, next_index, to_next, + n_left_to_next, bi0, next0); + } + + vlib_put_next_frame (vm, node, next_index, n_left_to_next); + } + + errors = session_manager_flush_enqueue_events (my_thread_index); + if (errors) + { + if (is_ip4) + vlib_node_increment_counter (vm, tcp4_established_node.index, + TCP_ERROR_EVENT_FIFO_FULL, errors); + else + vlib_node_increment_counter (vm, tcp6_established_node.index, + TCP_ERROR_EVENT_FIFO_FULL, errors); + } + + return from_frame->n_vectors; +} + +static uword +tcp4_syn_sent (vlib_main_t * vm, vlib_node_runtime_t * node, + vlib_frame_t * from_frame) +{ + return tcp46_syn_sent_inline (vm, node, from_frame, 1 /* is_ip4 */ ); +} + +static uword +tcp6_syn_sent_rcv (vlib_main_t * vm, vlib_node_runtime_t * node, + vlib_frame_t * from_frame) +{ + return tcp46_syn_sent_inline (vm, node, from_frame, 0 /* is_ip4 */ ); +} + +/* *INDENT-OFF* */ +VLIB_REGISTER_NODE (tcp4_syn_sent_node) = +{ + .function = tcp4_syn_sent, + .name = "tcp4-syn-sent", + /* Takes a vector of packets. */ + .vector_size = sizeof (u32), + .n_errors = TCP_N_ERROR, + .error_strings = tcp_error_strings, + .n_next_nodes = TCP_SYN_SENT_N_NEXT, + .next_nodes = + { +#define _(s,n) [TCP_SYN_SENT_NEXT_##s] = n, + foreach_tcp_state_next +#undef _ + }, +}; +/* *INDENT-ON* */ + +VLIB_NODE_FUNCTION_MULTIARCH (tcp4_syn_sent_node, tcp4_syn_sent); + +/* *INDENT-OFF* */ +VLIB_REGISTER_NODE (tcp6_syn_sent_node) = +{ + .function = tcp6_syn_sent_rcv, + .name = "tcp6-syn-sent", + /* Takes a vector of packets. */ + .vector_size = sizeof (u32), + .n_errors = TCP_N_ERROR, + .error_strings = tcp_error_strings, + .n_next_nodes = TCP_SYN_SENT_N_NEXT, + .next_nodes = + { +#define _(s,n) [TCP_SYN_SENT_NEXT_##s] = n, + foreach_tcp_state_next +#undef _ + } +,}; +/* *INDENT-ON* */ + +VLIB_NODE_FUNCTION_MULTIARCH (tcp6_syn_sent_node, tcp6_syn_sent_rcv); +/** + * Handles reception for all states except LISTEN, SYN-SEND and ESTABLISHED + * as per RFC793 p. 64 + */ +always_inline uword +tcp46_rcv_process_inline (vlib_main_t * vm, vlib_node_runtime_t * node, + vlib_frame_t * from_frame, int is_ip4) +{ + tcp_main_t *tm = vnet_get_tcp_main (); + u32 n_left_from, next_index, *from, *to_next; + u32 my_thread_index = vm->cpu_index, errors = 0; + + from = vlib_frame_vector_args (from_frame); + n_left_from = from_frame->n_vectors; + + next_index = node->cached_next_index; + + while (n_left_from > 0) + { + u32 n_left_to_next; + + vlib_get_next_frame (vm, node, next_index, to_next, n_left_to_next); + + while (n_left_from > 0 && n_left_to_next > 0) + { + u32 bi0; + vlib_buffer_t *b0; + tcp_header_t *tcp0 = 0; + tcp_connection_t *tc0; + ip4_header_t *ip40; + ip6_header_t *ip60; + u32 n_advance_bytes0, n_data_bytes0; + u32 next0 = TCP_RCV_PROCESS_NEXT_DROP, error0 = TCP_ERROR_ENQUEUED; + + bi0 = from[0]; + to_next[0] = bi0; + from += 1; + to_next += 1; + n_left_from -= 1; + n_left_to_next -= 1; + + b0 = vlib_get_buffer (vm, bi0); + tc0 = tcp_connection_get (vnet_buffer (b0)->tcp.connection_index, + my_thread_index); + + /* Checksum computed by ipx_local no need to compute again */ + + if (is_ip4) + { + ip40 = vlib_buffer_get_current (b0); + tcp0 = ip4_next_header (ip40); + n_advance_bytes0 = (ip4_header_bytes (ip40) + + tcp_header_bytes (tcp0)); + n_data_bytes0 = clib_net_to_host_u16 (ip40->length) + - n_advance_bytes0; + } + else + { + ip60 = vlib_buffer_get_current (b0); + tcp0 = ip6_next_header (ip60); + n_advance_bytes0 = tcp_header_bytes (tcp0); + n_data_bytes0 = clib_net_to_host_u16 (ip60->payload_length) + - n_advance_bytes0; + n_advance_bytes0 += sizeof (ip60[0]); + } + + /* SYNs, FINs and data consume sequence numbers */ + vnet_buffer (b0)->tcp.seq_end = vnet_buffer (b0)->tcp.seq_number + + tcp_is_syn (tcp0) + tcp_is_fin (tcp0) + n_data_bytes0; + + /* + * Special treatment for CLOSED + */ + switch (tc0->state) + { + case TCP_STATE_CLOSED: + goto drop; + break; + } + + /* + * For all other states (except LISTEN) + */ + + /* 1-4: check SEQ, RST, SYN */ + if (PREDICT_FALSE + (tcp_segment_validate (vm, tc0, b0, tcp0, &next0))) + { + error0 = TCP_ERROR_SEGMENT_INVALID; + goto drop; + } + + /* 5: check the ACK field */ + switch (tc0->state) + { + case TCP_STATE_SYN_RCVD: + /* + * If the segment acknowledgment is not acceptable, form a + * reset segment, + * + * and send it. + */ + if (!tcp_rcv_ack_is_acceptable (tc0, b0)) + { + tcp_send_reset (b0, is_ip4); + goto drop; + } + /* Switch state to ESTABLISHED */ + tc0->state = TCP_STATE_ESTABLISHED; + + /* Initialize session variables */ + tc0->snd_una = vnet_buffer (b0)->tcp.ack_number; + tc0->snd_wnd = clib_net_to_host_u32 (tcp0->window) + << tc0->opt.wscale; + tc0->snd_wl1 = vnet_buffer (b0)->tcp.seq_number; + tc0->snd_wl2 = vnet_buffer (b0)->tcp.ack_number; + + /* Shoulder tap the server */ + stream_session_accept_notify (&tc0->connection); + + tcp_timer_reset (tc0, TCP_TIMER_RETRANSMIT_SYN); + break; + case TCP_STATE_ESTABLISHED: + /* We can get packets in established state here because they + * were enqueued before state change */ + if (tcp_rcv_ack (tc0, b0, tcp0, &next0, &error0)) + goto drop; + + break; + case TCP_STATE_FIN_WAIT_1: + /* In addition to the processing for the ESTABLISHED state, if + * our FIN is now acknowledged then enter FIN-WAIT-2 and + * continue processing in that state. */ + if (tcp_rcv_ack (tc0, b0, tcp0, &next0, &error0)) + goto drop; + tc0->state = TCP_STATE_FIN_WAIT_2; + /* Stop all timers, 2MSL will be set lower */ + tcp_connection_timers_reset (tc0); + break; + case TCP_STATE_FIN_WAIT_2: + /* In addition to the processing for the ESTABLISHED state, if + * the retransmission queue is empty, the user's CLOSE can be + * acknowledged ("ok") but do not delete the TCB. */ + if (tcp_rcv_ack (tc0, b0, tcp0, &next0, &error0)) + goto drop; + /* check if rtx queue is empty and ack CLOSE TODO */ + break; + case TCP_STATE_CLOSE_WAIT: + /* Do the same processing as for the ESTABLISHED state. */ + if (tcp_rcv_ack (tc0, b0, tcp0, &next0, &error0)) + goto drop; + break; + case TCP_STATE_CLOSING: + /* In addition to the processing for the ESTABLISHED state, if + * the ACK acknowledges our FIN then enter the TIME-WAIT state, + * otherwise ignore the segment. */ + if (tcp_rcv_ack (tc0, b0, tcp0, &next0, &error0)) + goto drop; + + /* XXX test that send queue empty */ + tc0->state = TCP_STATE_TIME_WAIT; + goto drop; + + break; + case TCP_STATE_LAST_ACK: + /* The only thing that can arrive in this state is an + * acknowledgment of our FIN. If our FIN is now acknowledged, + * delete the TCB, enter the CLOSED state, and return. */ + + if (!tcp_rcv_ack_is_acceptable (tc0, b0)) + goto drop; + + tcp_connection_del (tc0); + goto drop; + + break; + case TCP_STATE_TIME_WAIT: + /* The only thing that can arrive in this state is a + * retransmission of the remote FIN. Acknowledge it, and restart + * the 2 MSL timeout. */ + + /* TODO */ + goto drop; + break; + default: + ASSERT (0); + } + + /* 6: check the URG bit TODO */ + + /* 7: process the segment text */ + switch (tc0->state) + { + case TCP_STATE_ESTABLISHED: + case TCP_STATE_FIN_WAIT_1: + case TCP_STATE_FIN_WAIT_2: + error0 = tcp_segment_rcv (tm, tc0, b0, n_data_bytes0, &next0); + break; + case TCP_STATE_CLOSE_WAIT: + case TCP_STATE_CLOSING: + case TCP_STATE_LAST_ACK: + case TCP_STATE_TIME_WAIT: + /* This should not occur, since a FIN has been received from the + * remote side. Ignore the segment text. */ + break; + } + + /* 8: check the FIN bit */ + if (!tcp_fin (tcp0)) + goto drop; + + switch (tc0->state) + { + case TCP_STATE_ESTABLISHED: + case TCP_STATE_SYN_RCVD: + /* Send FIN-ACK notify app and enter CLOSE-WAIT */ + tcp_connection_timers_reset (tc0); + tcp_make_finack (tc0, b0); + next0 = tcp_next_output (tc0->c_is_ip4); + stream_session_disconnect_notify (&tc0->connection); + tc0->state = TCP_STATE_CLOSE_WAIT; + break; + case TCP_STATE_CLOSE_WAIT: + case TCP_STATE_CLOSING: + case TCP_STATE_LAST_ACK: + /* move along .. */ + break; + case TCP_STATE_FIN_WAIT_1: + tc0->state = TCP_STATE_TIME_WAIT; + tcp_connection_timers_reset (tc0); + tcp_timer_set (tc0, TCP_TIMER_2MSL, TCP_2MSL_TIME); + break; + case TCP_STATE_FIN_WAIT_2: + /* Got FIN, send ACK! */ + tc0->state = TCP_STATE_TIME_WAIT; + tcp_timer_set (tc0, TCP_TIMER_2MSL, TCP_2MSL_TIME); + tcp_make_ack (tc0, b0); + next0 = tcp_next_output (is_ip4); + break; + case TCP_STATE_TIME_WAIT: + /* Remain in the TIME-WAIT state. Restart the 2 MSL time-wait + * timeout. + */ + tcp_timer_update (tc0, TCP_TIMER_2MSL, TCP_2MSL_TIME); + break; + } + + b0->error = error0 ? node->errors[error0] : 0; + + drop: + if (PREDICT_FALSE (b0->flags & VLIB_BUFFER_IS_TRACED)) + { + + } + + vlib_validate_buffer_enqueue_x1 (vm, node, next_index, to_next, + n_left_to_next, bi0, next0); + } + + vlib_put_next_frame (vm, node, next_index, n_left_to_next); + } + + errors = session_manager_flush_enqueue_events (my_thread_index); + if (errors) + { + if (is_ip4) + vlib_node_increment_counter (vm, tcp4_established_node.index, + TCP_ERROR_EVENT_FIFO_FULL, errors); + else + vlib_node_increment_counter (vm, tcp6_established_node.index, + TCP_ERROR_EVENT_FIFO_FULL, errors); + } + + return from_frame->n_vectors; +} + +static uword +tcp4_rcv_process (vlib_main_t * vm, vlib_node_runtime_t * node, + vlib_frame_t * from_frame) +{ + return tcp46_rcv_process_inline (vm, node, from_frame, 1 /* is_ip4 */ ); +} + +static uword +tcp6_rcv_process (vlib_main_t * vm, vlib_node_runtime_t * node, + vlib_frame_t * from_frame) +{ + return tcp46_rcv_process_inline (vm, node, from_frame, 0 /* is_ip4 */ ); +} + +/* *INDENT-OFF* */ +VLIB_REGISTER_NODE (tcp4_rcv_process_node) = +{ + .function = tcp4_rcv_process, + .name = "tcp4-rcv-process", + /* Takes a vector of packets. */ + .vector_size = sizeof (u32), + .n_errors = TCP_N_ERROR, + .error_strings = tcp_error_strings, + .n_next_nodes = TCP_RCV_PROCESS_N_NEXT, + .next_nodes = + { +#define _(s,n) [TCP_RCV_PROCESS_NEXT_##s] = n, + foreach_tcp_state_next +#undef _ + }, +}; +/* *INDENT-ON* */ + +VLIB_NODE_FUNCTION_MULTIARCH (tcp4_rcv_process_node, tcp4_rcv_process); + +/* *INDENT-OFF* */ +VLIB_REGISTER_NODE (tcp6_rcv_process_node) = +{ + .function = tcp6_rcv_process, + .name = "tcp6-rcv-process", + /* Takes a vector of packets. */ + .vector_size = sizeof (u32), + .n_errors = TCP_N_ERROR, + .error_strings = tcp_error_strings, + .n_next_nodes = TCP_RCV_PROCESS_N_NEXT, + .next_nodes = + { +#define _(s,n) [TCP_RCV_PROCESS_NEXT_##s] = n, + foreach_tcp_state_next +#undef _ + }, +}; +/* *INDENT-ON* */ + +VLIB_NODE_FUNCTION_MULTIARCH (tcp6_rcv_process_node, tcp6_rcv_process); + +vlib_node_registration_t tcp4_listen_node; +vlib_node_registration_t tcp6_listen_node; + +/** + * LISTEN state processing as per RFC 793 p. 65 + */ +always_inline uword +tcp46_listen_inline (vlib_main_t * vm, vlib_node_runtime_t * node, + vlib_frame_t * from_frame, int is_ip4) +{ + u32 n_left_from, next_index, *from, *to_next; + u32 my_thread_index = vm->cpu_index; + tcp_main_t *tm = vnet_get_tcp_main (); + u8 sst = is_ip4 ? SESSION_TYPE_IP4_TCP : SESSION_TYPE_IP6_TCP; + + from = vlib_frame_vector_args (from_frame); + n_left_from = from_frame->n_vectors; + + next_index = node->cached_next_index; + + while (n_left_from > 0) + { + u32 n_left_to_next; + + vlib_get_next_frame (vm, node, next_index, to_next, n_left_to_next); + + while (n_left_from > 0 && n_left_to_next > 0) + { + u32 bi0; + vlib_buffer_t *b0; + tcp_header_t *th0 = 0; + tcp_connection_t *lc0; + ip4_header_t *ip40; + ip6_header_t *ip60; + tcp_connection_t *child0; + u32 error0 = TCP_ERROR_SYNS_RCVD, next0 = TCP_LISTEN_NEXT_DROP; + + bi0 = from[0]; + to_next[0] = bi0; + from += 1; + to_next += 1; + n_left_from -= 1; + n_left_to_next -= 1; + + b0 = vlib_get_buffer (vm, bi0); + lc0 = tcp_listener_get (vnet_buffer (b0)->tcp.connection_index); + + if (is_ip4) + { + ip40 = vlib_buffer_get_current (b0); + th0 = ip4_next_header (ip40); + } + else + { + ip60 = vlib_buffer_get_current (b0); + th0 = ip6_next_header (ip60); + } + + /* Create child session. For syn-flood protection use filter */ + + /* 1. first check for an RST */ + if (tcp_rst (th0)) + goto drop; + + /* 2. second check for an ACK */ + if (tcp_ack (th0)) + { + tcp_send_reset (b0, is_ip4); + goto drop; + } + + /* 3. check for a SYN (did that already) */ + + /* Create child session and send SYN-ACK */ + pool_get (tm->connections[my_thread_index], child0); + memset (child0, 0, sizeof (*child0)); + + child0->c_c_index = child0 - tm->connections[my_thread_index]; + child0->c_lcl_port = lc0->c_lcl_port; + child0->c_rmt_port = th0->src_port; + child0->c_is_ip4 = is_ip4; + child0->c_thread_index = my_thread_index; + + if (is_ip4) + { + child0->c_lcl_ip4.as_u32 = ip40->dst_address.as_u32; + child0->c_rmt_ip4.as_u32 = ip40->src_address.as_u32; + } + else + { + clib_memcpy (&child0->c_lcl_ip6, &ip60->dst_address, + sizeof (ip6_address_t)); + clib_memcpy (&child0->c_rmt_ip6, &ip60->src_address, + sizeof (ip6_address_t)); + } + + if (stream_session_accept (&child0->connection, lc0->c_s_index, sst, + 0 /* notify */ )) + { + error0 = TCP_ERROR_CREATE_SESSION_FAIL; + goto drop; + } + + tcp_options_parse (th0, &child0->opt); + tcp_connection_init_vars (child0); + + child0->irs = vnet_buffer (b0)->tcp.seq_number; + child0->rcv_nxt = vnet_buffer (b0)->tcp.seq_number + 1; + child0->state = TCP_STATE_SYN_RCVD; + + /* RFC1323: TSval timestamps sent on {SYN} and {SYN,ACK} + * segments are used to initialize PAWS. */ + if (tcp_opts_tstamp (&child0->opt)) + { + child0->tsval_recent = child0->opt.tsval; + child0->tsval_recent_age = tcp_time_now (); + } + + /* Reuse buffer to make syn-ack and send */ + tcp_make_synack (child0, b0); + next0 = tcp_next_output (is_ip4); + + drop: + if (PREDICT_FALSE (b0->flags & VLIB_BUFFER_IS_TRACED)) + { + + } + + b0->error = error0 ? node->errors[error0] : 0; + + vlib_validate_buffer_enqueue_x1 (vm, node, next_index, to_next, + n_left_to_next, bi0, next0); + } + + vlib_put_next_frame (vm, node, next_index, n_left_to_next); + } + return from_frame->n_vectors; +} + +static uword +tcp4_listen (vlib_main_t * vm, vlib_node_runtime_t * node, + vlib_frame_t * from_frame) +{ + return tcp46_listen_inline (vm, node, from_frame, 1 /* is_ip4 */ ); +} + +static uword +tcp6_listen (vlib_main_t * vm, vlib_node_runtime_t * node, + vlib_frame_t * from_frame) +{ + return tcp46_listen_inline (vm, node, from_frame, 0 /* is_ip4 */ ); +} + +/* *INDENT-OFF* */ +VLIB_REGISTER_NODE (tcp4_listen_node) = +{ + .function = tcp4_listen, + .name = "tcp4-listen", + /* Takes a vector of packets. */ + .vector_size = sizeof (u32), + .n_errors = TCP_N_ERROR, + .error_strings = tcp_error_strings, + .n_next_nodes = TCP_LISTEN_N_NEXT, + .next_nodes = + { +#define _(s,n) [TCP_LISTEN_NEXT_##s] = n, + foreach_tcp_state_next +#undef _ + }, +}; +/* *INDENT-ON* */ + +VLIB_NODE_FUNCTION_MULTIARCH (tcp4_listen_node, tcp4_listen); + +/* *INDENT-OFF* */ +VLIB_REGISTER_NODE (tcp6_listen_node) = +{ + .function = tcp6_listen, + .name = "tcp6-listen", + /* Takes a vector of packets. */ + .vector_size = sizeof (u32), + .n_errors = TCP_N_ERROR, + .error_strings = tcp_error_strings, + .n_next_nodes = TCP_LISTEN_N_NEXT, + .next_nodes = + { +#define _(s,n) [TCP_LISTEN_NEXT_##s] = n, + foreach_tcp_state_next +#undef _ + }, +}; +/* *INDENT-ON* */ + +VLIB_NODE_FUNCTION_MULTIARCH (tcp6_listen_node, tcp6_listen); + +vlib_node_registration_t tcp4_input_node; +vlib_node_registration_t tcp6_input_node; + +typedef enum _tcp_input_next +{ + TCP_INPUT_NEXT_DROP, + TCP_INPUT_NEXT_LISTEN, + TCP_INPUT_NEXT_RCV_PROCESS, + TCP_INPUT_NEXT_SYN_SENT, + TCP_INPUT_NEXT_ESTABLISHED, + TCP_INPUT_NEXT_RESET, + TCP_INPUT_N_NEXT +} tcp_input_next_t; + +#define foreach_tcp4_input_next \ + _ (DROP, "error-drop") \ + _ (LISTEN, "tcp4-listen") \ + _ (RCV_PROCESS, "tcp4-rcv-process") \ + _ (SYN_SENT, "tcp4-syn-sent") \ + _ (ESTABLISHED, "tcp4-established") \ + _ (RESET, "tcp4-reset") + +#define foreach_tcp6_input_next \ + _ (DROP, "error-drop") \ + _ (LISTEN, "tcp6-listen") \ + _ (RCV_PROCESS, "tcp6-rcv-process") \ + _ (SYN_SENT, "tcp6-syn-sent") \ + _ (ESTABLISHED, "tcp6-established") \ + _ (RESET, "tcp6-reset") + +typedef struct +{ + u16 src_port; + u16 dst_port; + u8 state; +} tcp_rx_trace_t; + +const char *tcp_fsm_states[] = { +#define _(sym, str) str, + foreach_tcp_fsm_state +#undef _ +}; + +u8 * +format_tcp_state (u8 * s, va_list * args) +{ + tcp_state_t *state = va_arg (*args, tcp_state_t *); + + if (state[0] < TCP_N_STATES) + s = format (s, "%s", tcp_fsm_states[state[0]]); + else + s = format (s, "UNKNOWN"); + + return s; +} + +u8 * +format_tcp_rx_trace (u8 * s, va_list * args) +{ + CLIB_UNUSED (vlib_main_t * vm) = va_arg (*args, vlib_main_t *); + CLIB_UNUSED (vlib_node_t * node) = va_arg (*args, vlib_node_t *); + tcp_rx_trace_t *t = va_arg (*args, tcp_rx_trace_t *); + + s = format (s, "TCP: src-port %d dst-port %U%s\n", + clib_net_to_host_u16 (t->src_port), + clib_net_to_host_u16 (t->dst_port), format_tcp_state, t->state); + + return s; +} + +#define filter_flags (TCP_FLAG_SYN|TCP_FLAG_ACK|TCP_FLAG_RST|TCP_FLAG_FIN) + +always_inline uword +tcp46_input_inline (vlib_main_t * vm, vlib_node_runtime_t * node, + vlib_frame_t * from_frame, int is_ip4) +{ + u32 n_left_from, next_index, *from, *to_next; + u32 my_thread_index = vm->cpu_index; + tcp_main_t *tm = vnet_get_tcp_main (); + session_manager_main_t *ssm = vnet_get_session_manager_main (); + + from = vlib_frame_vector_args (from_frame); + n_left_from = from_frame->n_vectors; + + next_index = node->cached_next_index; + + while (n_left_from > 0) + { + u32 n_left_to_next; + + vlib_get_next_frame (vm, node, next_index, to_next, n_left_to_next); + + while (n_left_from > 0 && n_left_to_next > 0) + { + u32 bi0; + vlib_buffer_t *b0; + tcp_header_t *tcp0 = 0; + tcp_connection_t *tc0; + ip4_header_t *ip40; + ip6_header_t *ip60; + u32 error0 = TCP_ERROR_NO_LISTENER, next0 = TCP_INPUT_NEXT_DROP; + u8 flags0; + + bi0 = from[0]; + to_next[0] = bi0; + from += 1; + to_next += 1; + n_left_from -= 1; + n_left_to_next -= 1; + + b0 = vlib_get_buffer (vm, bi0); + + if (is_ip4) + { + ip40 = vlib_buffer_get_current (b0); + tcp0 = ip4_next_header (ip40); + + /* lookup session */ + tc0 = + (tcp_connection_t *) stream_session_lookup_transport4 (ssm, + &ip40->dst_address, + &ip40->src_address, + tcp0->dst_port, + tcp0->src_port, + SESSION_TYPE_IP4_TCP, + my_thread_index); + } + else + { + ip60 = vlib_buffer_get_current (b0); + tcp0 = ip6_next_header (ip60); + tc0 = + (tcp_connection_t *) stream_session_lookup_transport6 (ssm, + &ip60->src_address, + &ip60->dst_address, + tcp0->src_port, + tcp0->dst_port, + SESSION_TYPE_IP6_TCP, + my_thread_index); + } + + /* Session exists */ + if (PREDICT_TRUE (0 != tc0)) + { + /* Save connection index */ + vnet_buffer (b0)->tcp.connection_index = tc0->c_c_index; + vnet_buffer (b0)->tcp.seq_number = + clib_net_to_host_u32 (tcp0->seq_number); + vnet_buffer (b0)->tcp.ack_number = + clib_net_to_host_u32 (tcp0->ack_number); + + flags0 = tcp0->flags & filter_flags; + next0 = tm->dispatch_table[tc0->state][flags0].next; + error0 = tm->dispatch_table[tc0->state][flags0].error; + + if (PREDICT_FALSE (error0 == TCP_ERROR_DISPATCH)) + { + /* Overload tcp flags to store state */ + vnet_buffer (b0)->tcp.flags = tc0->state; + } + } + else + { + /* Send reset */ + next0 = TCP_INPUT_NEXT_RESET; + error0 = TCP_ERROR_NO_LISTENER; + vnet_buffer (b0)->tcp.flags = 0; + } + + b0->error = error0 ? node->errors[error0] : 0; + + if (PREDICT_FALSE (b0->flags & VLIB_BUFFER_IS_TRACED)) + { + + } + + vlib_validate_buffer_enqueue_x1 (vm, node, next_index, to_next, + n_left_to_next, bi0, next0); + } + + vlib_put_next_frame (vm, node, next_index, n_left_to_next); + } + + return from_frame->n_vectors; +} + +static uword +tcp4_input (vlib_main_t * vm, vlib_node_runtime_t * node, + vlib_frame_t * from_frame) +{ + return tcp46_input_inline (vm, node, from_frame, 1 /* is_ip4 */ ); +} + +static uword +tcp6_input (vlib_main_t * vm, vlib_node_runtime_t * node, + vlib_frame_t * from_frame) +{ + return tcp46_input_inline (vm, node, from_frame, 0 /* is_ip4 */ ); +} + +/* *INDENT-OFF* */ +VLIB_REGISTER_NODE (tcp4_input_node) = +{ + .function = tcp4_input, + .name = "tcp4-input", + /* Takes a vector of packets. */ + .vector_size = sizeof (u32), + .n_errors = TCP_N_ERROR, + .error_strings = tcp_error_strings, + .n_next_nodes = TCP_INPUT_N_NEXT, + .next_nodes = + { +#define _(s,n) [TCP_INPUT_NEXT_##s] = n, + foreach_tcp4_input_next +#undef _ + }, + .format_buffer = format_tcp_header, + .format_trace = format_tcp_rx_trace, +}; +/* *INDENT-ON* */ + +VLIB_NODE_FUNCTION_MULTIARCH (tcp4_input_node, tcp4_input); + +/* *INDENT-OFF* */ +VLIB_REGISTER_NODE (tcp6_input_node) = +{ + .function = tcp6_input, + .name = "tcp6-input", + /* Takes a vector of packets. */ + .vector_size = sizeof (u32), + .n_errors = TCP_N_ERROR, + .error_strings = tcp_error_strings, + .n_next_nodes = TCP_INPUT_N_NEXT, + .next_nodes = + { +#define _(s,n) [TCP_INPUT_NEXT_##s] = n, + foreach_tcp6_input_next +#undef _ + }, + .format_buffer = format_tcp_header, + .format_trace = format_tcp_rx_trace, +}; +/* *INDENT-ON* */ + +VLIB_NODE_FUNCTION_MULTIARCH (tcp6_input_node, tcp6_input); +void +tcp_update_time (f64 now, u32 thread_index) +{ + tcp_main_t *tm = vnet_get_tcp_main (); + tw_timer_expire_timers_16t_2w_512sl (&tm->timer_wheels[thread_index], now); +} + +static void +tcp_dispatch_table_init (tcp_main_t * tm) +{ + int i, j; + for (i = 0; i < ARRAY_LEN (tm->dispatch_table); i++) + for (j = 0; j < ARRAY_LEN (tm->dispatch_table[i]); j++) + { + tm->dispatch_table[i][j].next = TCP_INPUT_NEXT_DROP; + tm->dispatch_table[i][j].error = TCP_ERROR_DISPATCH; + } + +#define _(t,f,n,e) \ +do { \ + tm->dispatch_table[TCP_STATE_##t][f].next = (n); \ + tm->dispatch_table[TCP_STATE_##t][f].error = (e); \ +} while (0) + + /* SYNs for new connections -> tcp-listen. */ + _(LISTEN, TCP_FLAG_SYN, TCP_INPUT_NEXT_LISTEN, TCP_ERROR_NONE); + /* ACK for for a SYN-ACK -> tcp-rcv-process. */ + _(SYN_RCVD, TCP_FLAG_ACK, TCP_INPUT_NEXT_RCV_PROCESS, TCP_ERROR_NONE); + /* SYN-ACK for a SYN */ + _(SYN_SENT, TCP_FLAG_SYN | TCP_FLAG_ACK, TCP_INPUT_NEXT_SYN_SENT, + TCP_ERROR_NONE); + _(SYN_SENT, TCP_FLAG_ACK, TCP_INPUT_NEXT_SYN_SENT, TCP_ERROR_NONE); + _(SYN_SENT, TCP_FLAG_RST, TCP_INPUT_NEXT_SYN_SENT, TCP_ERROR_NONE); + _(SYN_SENT, TCP_FLAG_RST | TCP_FLAG_ACK, TCP_INPUT_NEXT_SYN_SENT, + TCP_ERROR_NONE); + /* ACK for for established connection -> tcp-established. */ + _(ESTABLISHED, TCP_FLAG_ACK, TCP_INPUT_NEXT_ESTABLISHED, TCP_ERROR_NONE); + /* FIN for for established connection -> tcp-established. */ + _(ESTABLISHED, TCP_FLAG_FIN, TCP_INPUT_NEXT_ESTABLISHED, TCP_ERROR_NONE); + _(ESTABLISHED, TCP_FLAG_FIN | TCP_FLAG_ACK, TCP_INPUT_NEXT_ESTABLISHED, + TCP_ERROR_NONE); + /* ACK or FIN-ACK to our FIN */ + _(FIN_WAIT_1, TCP_FLAG_ACK, TCP_INPUT_NEXT_RCV_PROCESS, TCP_ERROR_NONE); + _(FIN_WAIT_1, TCP_FLAG_ACK | TCP_FLAG_FIN, TCP_INPUT_NEXT_RCV_PROCESS, + TCP_ERROR_NONE); + /* FIN in reply to our FIN from the other side */ + _(FIN_WAIT_1, TCP_FLAG_FIN, TCP_INPUT_NEXT_RCV_PROCESS, TCP_ERROR_NONE); + /* FIN confirming that the peer (app) has closed */ + _(FIN_WAIT_2, TCP_FLAG_FIN, TCP_INPUT_NEXT_RCV_PROCESS, TCP_ERROR_NONE); + _(FIN_WAIT_2, TCP_FLAG_FIN | TCP_FLAG_ACK, TCP_INPUT_NEXT_RCV_PROCESS, + TCP_ERROR_NONE); + _(LAST_ACK, TCP_FLAG_ACK, TCP_INPUT_NEXT_RCV_PROCESS, TCP_ERROR_NONE); +#undef _ +} + +clib_error_t * +tcp_input_init (vlib_main_t * vm) +{ + clib_error_t *error = 0; + tcp_main_t *tm = vnet_get_tcp_main (); + + if ((error = vlib_call_init_function (vm, tcp_init))) + return error; + + /* Initialize dispatch table. */ + tcp_dispatch_table_init (tm); + + return error; +} + +VLIB_INIT_FUNCTION (tcp_input_init); + +/* + * fd.io coding-style-patch-verification: ON + * + * Local Variables: + * eval: (c-set-style "gnu") + * End: + */ diff --git a/src/vnet/tcp/tcp_newreno.c b/src/vnet/tcp/tcp_newreno.c new file mode 100644 index 00000000..856dffe4 --- /dev/null +++ b/src/vnet/tcp/tcp_newreno.c @@ -0,0 +1,93 @@ +/* + * Copyright (c) 2017 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +#include + +void +newreno_congestion (tcp_connection_t * tc) +{ + tc->prev_ssthresh = tc->ssthresh; + tc->ssthresh = clib_max (tcp_flight_size (tc) / 2, 2 * tc->snd_mss); +} + +void +newreno_recovered (tcp_connection_t * tc) +{ + tc->cwnd = tc->ssthresh; +} + +void +newreno_rcv_ack (tcp_connection_t * tc) +{ + if (tcp_in_slowstart (tc)) + { + tc->cwnd += clib_min (tc->snd_mss, tc->bytes_acked); + } + else + { + /* Round up to 1 if needed */ + tc->cwnd += clib_max (tc->snd_mss * tc->snd_mss / tc->cwnd, 1); + } +} + +void +newreno_rcv_cong_ack (tcp_connection_t * tc, tcp_cc_ack_t ack_type) +{ + if (ack_type == TCP_CC_DUPACK) + { + tc->cwnd += tc->snd_mss; + } + else if (ack_type == TCP_CC_PARTIALACK) + { + tc->cwnd -= tc->bytes_acked; + if (tc->bytes_acked > tc->snd_mss) + tc->bytes_acked += tc->snd_mss; + } +} + +void +newreno_conn_init (tcp_connection_t * tc) +{ + tc->ssthresh = tc->snd_wnd; + tc->cwnd = tcp_initial_cwnd (tc); +} + +const static tcp_cc_algorithm_t tcp_newreno = { + .congestion = newreno_congestion, + .recovered = newreno_recovered, + .rcv_ack = newreno_rcv_ack, + .rcv_cong_ack = newreno_rcv_cong_ack, + .init = newreno_conn_init +}; + +clib_error_t * +newreno_init (vlib_main_t * vm) +{ + clib_error_t *error = 0; + + tcp_cc_algo_register (TCP_CC_NEWRENO, &tcp_newreno); + + return error; +} + +VLIB_INIT_FUNCTION (newreno_init); + +/* + * fd.io coding-style-patch-verification: ON + * + * Local Variables: + * eval: (c-set-style "gnu") + * End: + */ diff --git a/src/vnet/tcp/tcp_output.c b/src/vnet/tcp/tcp_output.c new file mode 100644 index 00000000..dbcf1f74 --- /dev/null +++ b/src/vnet/tcp/tcp_output.c @@ -0,0 +1,1412 @@ +/* + * Copyright (c) 2016 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +#include +#include + +vlib_node_registration_t tcp4_output_node; +vlib_node_registration_t tcp6_output_node; + +typedef enum _tcp_output_nect +{ + TCP_OUTPUT_NEXT_DROP, + TCP_OUTPUT_NEXT_IP_LOOKUP, + TCP_OUTPUT_N_NEXT +} tcp_output_next_t; + +#define foreach_tcp4_output_next \ + _ (DROP, "error-drop") \ + _ (IP_LOOKUP, "ip4-lookup") + +#define foreach_tcp6_output_next \ + _ (DROP, "error-drop") \ + _ (IP_LOOKUP, "ip6-lookup") + +static char *tcp_error_strings[] = { +#define tcp_error(n,s) s, +#include +#undef tcp_error +}; + +typedef struct +{ + u16 src_port; + u16 dst_port; + u8 state; +} tcp_tx_trace_t; + +u16 dummy_mtu = 400; + +u8 * +format_tcp_tx_trace (u8 * s, va_list * args) +{ + CLIB_UNUSED (vlib_main_t * vm) = va_arg (*args, vlib_main_t *); + CLIB_UNUSED (vlib_node_t * node) = va_arg (*args, vlib_node_t *); + + s = format (s, "TBD\n"); + + return s; +} + +void +tcp_set_snd_mss (tcp_connection_t * tc) +{ + u16 snd_mss; + + /* TODO find our iface MTU */ + snd_mss = dummy_mtu; + + /* TODO cache mss and consider PMTU discovery */ + snd_mss = tc->opt.mss < snd_mss ? tc->opt.mss : snd_mss; + + tc->snd_mss = snd_mss; + + if (tc->snd_mss == 0) + { + clib_warning ("snd mss is 0"); + tc->snd_mss = dummy_mtu; + } +} + +static u8 +tcp_window_compute_scale (u32 available_space) +{ + u8 wnd_scale = 0; + while (wnd_scale < TCP_MAX_WND_SCALE + && (available_space >> wnd_scale) > TCP_WND_MAX) + wnd_scale++; + return wnd_scale; +} + +/** + * Compute initial window and scale factor. As per RFC1323, window field in + * SYN and SYN-ACK segments is never scaled. + */ +u32 +tcp_initial_window_to_advertise (tcp_connection_t * tc) +{ + u32 available_space; + + /* Initial wnd for SYN. Fifos are not allocated yet. + * Use some predefined value */ + if (tc->state != TCP_STATE_SYN_RCVD) + { + return TCP_DEFAULT_RX_FIFO_SIZE; + } + + available_space = stream_session_max_enqueue (&tc->connection); + tc->rcv_wscale = tcp_window_compute_scale (available_space); + tc->rcv_wnd = clib_min (available_space, TCP_WND_MAX << tc->rcv_wscale); + + return clib_min (tc->rcv_wnd, TCP_WND_MAX); +} + +/** + * Compute and return window to advertise, scaled as per RFC1323 + */ +u32 +tcp_window_to_advertise (tcp_connection_t * tc, tcp_state_t state) +{ + u32 available_space, wnd, scaled_space; + + if (state != TCP_STATE_ESTABLISHED) + return tcp_initial_window_to_advertise (tc); + + available_space = stream_session_max_enqueue (&tc->connection); + scaled_space = available_space >> tc->rcv_wscale; + + /* Need to update scale */ + if (PREDICT_FALSE ((scaled_space == 0 && available_space != 0)) + || (scaled_space >= TCP_WND_MAX)) + tc->rcv_wscale = tcp_window_compute_scale (available_space); + + wnd = clib_min (available_space, TCP_WND_MAX << tc->rcv_wscale); + tc->rcv_wnd = wnd; + + return wnd >> tc->rcv_wscale; +} + +/** + * Write TCP options to segment. + */ +u32 +tcp_options_write (u8 * data, tcp_options_t * opts) +{ + u32 opts_len = 0; + u32 buf, seq_len = 4; + + if (tcp_opts_mss (opts)) + { + *data++ = TCP_OPTION_MSS; + *data++ = TCP_OPTION_LEN_MSS; + buf = clib_host_to_net_u16 (opts->mss); + clib_memcpy (data, &buf, sizeof (opts->mss)); + data += sizeof (opts->mss); + opts_len += TCP_OPTION_LEN_MSS; + } + + if (tcp_opts_wscale (opts)) + { + *data++ = TCP_OPTION_WINDOW_SCALE; + *data++ = TCP_OPTION_LEN_WINDOW_SCALE; + *data++ = opts->wscale; + opts_len += TCP_OPTION_LEN_WINDOW_SCALE; + } + + if (tcp_opts_sack_permitted (opts)) + { + *data++ = TCP_OPTION_SACK_PERMITTED; + *data++ = TCP_OPTION_LEN_SACK_PERMITTED; + opts_len += TCP_OPTION_LEN_SACK_PERMITTED; + } + + if (tcp_opts_tstamp (opts)) + { + *data++ = TCP_OPTION_TIMESTAMP; + *data++ = TCP_OPTION_LEN_TIMESTAMP; + buf = clib_host_to_net_u32 (opts->tsval); + clib_memcpy (data, &buf, sizeof (opts->tsval)); + data += sizeof (opts->tsval); + buf = clib_host_to_net_u32 (opts->tsecr); + clib_memcpy (data, &buf, sizeof (opts->tsecr)); + data += sizeof (opts->tsecr); + opts_len += TCP_OPTION_LEN_TIMESTAMP; + } + + if (tcp_opts_sack (opts)) + { + int i; + u32 n_sack_blocks = clib_min (vec_len (opts->sacks), + TCP_OPTS_MAX_SACK_BLOCKS); + + if (n_sack_blocks != 0) + { + *data++ = TCP_OPTION_SACK_BLOCK; + *data++ = 2 + n_sack_blocks * TCP_OPTION_LEN_SACK_BLOCK; + for (i = 0; i < n_sack_blocks; i++) + { + buf = clib_host_to_net_u32 (opts->sacks[i].start); + clib_memcpy (data, &buf, seq_len); + data += seq_len; + buf = clib_host_to_net_u32 (opts->sacks[i].end); + clib_memcpy (data, &buf, seq_len); + data += seq_len; + } + opts_len += 2 + n_sack_blocks * TCP_OPTION_LEN_SACK_BLOCK; + } + } + + /* Terminate TCP options */ + if (opts_len % 4) + { + *data++ = TCP_OPTION_EOL; + opts_len += TCP_OPTION_LEN_EOL; + } + + /* Pad with zeroes to a u32 boundary */ + while (opts_len % 4) + { + *data++ = TCP_OPTION_NOOP; + opts_len += TCP_OPTION_LEN_NOOP; + } + return opts_len; +} + +always_inline int +tcp_make_syn_options (tcp_options_t * opts, u32 initial_wnd) +{ + u8 len = 0; + + opts->flags |= TCP_OPTS_FLAG_MSS; + opts->mss = dummy_mtu; /*XXX discover that */ + len += TCP_OPTION_LEN_MSS; + + opts->flags |= TCP_OPTS_FLAG_WSCALE; + opts->wscale = tcp_window_compute_scale (initial_wnd); + len += TCP_OPTION_LEN_WINDOW_SCALE; + + opts->flags |= TCP_OPTS_FLAG_TSTAMP; + opts->tsval = tcp_time_now (); + opts->tsecr = 0; + len += TCP_OPTION_LEN_TIMESTAMP; + + opts->flags |= TCP_OPTS_FLAG_SACK_PERMITTED; + len += TCP_OPTION_LEN_SACK_PERMITTED; + + /* Align to needed boundary */ + len += (TCP_OPTS_ALIGN - len % TCP_OPTS_ALIGN) % TCP_OPTS_ALIGN; + return len; +} + +always_inline int +tcp_make_synack_options (tcp_connection_t * tc, tcp_options_t * opts) +{ + u8 len = 0; + + opts->flags |= TCP_OPTS_FLAG_MSS; + opts->mss = dummy_mtu; /*XXX discover that */ + len += TCP_OPTION_LEN_MSS; + + if (tcp_opts_wscale (&tc->opt)) + { + opts->flags |= TCP_OPTS_FLAG_WSCALE; + opts->wscale = tc->rcv_wscale; + len += TCP_OPTION_LEN_WINDOW_SCALE; + } + + if (tcp_opts_tstamp (&tc->opt)) + { + opts->flags |= TCP_OPTS_FLAG_TSTAMP; + opts->tsval = tcp_time_now (); + opts->tsecr = tc->tsval_recent; + len += TCP_OPTION_LEN_TIMESTAMP; + } + + if (tcp_opts_sack_permitted (&tc->opt)) + { + opts->flags |= TCP_OPTS_FLAG_SACK_PERMITTED; + len += TCP_OPTION_LEN_SACK_PERMITTED; + } + + /* Align to needed boundary */ + len += (TCP_OPTS_ALIGN - len % TCP_OPTS_ALIGN) % TCP_OPTS_ALIGN; + return len; +} + +always_inline int +tcp_make_established_options (tcp_connection_t * tc, tcp_options_t * opts) +{ + u8 len = 0; + + opts->flags = 0; + + if (tcp_opts_tstamp (&tc->opt)) + { + opts->flags |= TCP_OPTS_FLAG_TSTAMP; + opts->tsval = tcp_time_now (); + opts->tsecr = tc->tsval_recent; + len += TCP_OPTION_LEN_TIMESTAMP; + } + if (tcp_opts_sack_permitted (&tc->opt)) + { + if (vec_len (tc->snd_sacks)) + { + opts->flags |= TCP_OPTS_FLAG_SACK; + opts->sacks = tc->snd_sacks; + opts->n_sack_blocks = vec_len (tc->snd_sacks); + len += 2 + TCP_OPTION_LEN_SACK_BLOCK * opts->n_sack_blocks; + } + } + + /* Align to needed boundary */ + len += (TCP_OPTS_ALIGN - len % TCP_OPTS_ALIGN) % TCP_OPTS_ALIGN; + return len; +} + +always_inline int +tcp_make_options (tcp_connection_t * tc, tcp_options_t * opts, + tcp_state_t state) +{ + switch (state) + { + case TCP_STATE_ESTABLISHED: + case TCP_STATE_FIN_WAIT_1: + return tcp_make_established_options (tc, opts); + case TCP_STATE_SYN_RCVD: + return tcp_make_synack_options (tc, opts); + case TCP_STATE_SYN_SENT: + return tcp_make_syn_options (opts, + tcp_initial_window_to_advertise (tc)); + default: + clib_warning ("Not handled!"); + return 0; + } +} + +#define tcp_get_free_buffer_index(tm, bidx) \ +do { \ + u32 *my_tx_buffers, n_free_buffers; \ + u32 cpu_index = tm->vlib_main->cpu_index; \ + my_tx_buffers = tm->tx_buffers[cpu_index]; \ + if (PREDICT_FALSE(vec_len (my_tx_buffers) == 0)) \ + { \ + n_free_buffers = 32; /* TODO config or macro */ \ + vec_validate (my_tx_buffers, n_free_buffers - 1); \ + _vec_len(my_tx_buffers) = vlib_buffer_alloc_from_free_list ( \ + tm->vlib_main, my_tx_buffers, n_free_buffers, \ + VLIB_BUFFER_DEFAULT_FREE_LIST_INDEX); \ + tm->tx_buffers[cpu_index] = my_tx_buffers; \ + } \ + /* buffer shortage */ \ + if (PREDICT_FALSE (vec_len (my_tx_buffers) == 0)) \ + return; \ + *bidx = my_tx_buffers[_vec_len (my_tx_buffers)-1]; \ + _vec_len (my_tx_buffers) -= 1; \ +} while (0) + +always_inline void +tcp_reuse_buffer (vlib_main_t * vm, vlib_buffer_t * b) +{ + vlib_buffer_t *it = b; + do + { + it->current_data = 0; + it->current_length = 0; + it->total_length_not_including_first_buffer = 0; + } + while ((it->flags & VLIB_BUFFER_NEXT_PRESENT) + && (it = vlib_get_buffer (vm, it->next_buffer))); + + /* Leave enough space for headers */ + vlib_buffer_make_headroom (b, MAX_HDRS_LEN); +} + +/** + * Prepare ACK + */ +void +tcp_make_ack_i (tcp_connection_t * tc, vlib_buffer_t * b, tcp_state_t state, + u8 flags) +{ + tcp_options_t _snd_opts, *snd_opts = &_snd_opts; + u8 tcp_opts_len, tcp_hdr_opts_len; + tcp_header_t *th; + u16 wnd; + + wnd = tcp_window_to_advertise (tc, state); + + /* Make and write options */ + tcp_opts_len = tcp_make_established_options (tc, snd_opts); + tcp_hdr_opts_len = tcp_opts_len + sizeof (tcp_header_t); + + th = vlib_buffer_push_tcp (b, tc->c_lcl_port, tc->c_rmt_port, tc->snd_nxt, + tc->rcv_nxt, tcp_hdr_opts_len, flags, wnd); + + tcp_options_write ((u8 *) (th + 1), snd_opts); + + /* Mark as ACK */ + vnet_buffer (b)->tcp.connection_index = tc->c_c_index; +} + +/** + * Convert buffer to ACK + */ +void +tcp_make_ack (tcp_connection_t * tc, vlib_buffer_t * b) +{ + tcp_main_t *tm = vnet_get_tcp_main (); + vlib_main_t *vm = tm->vlib_main; + + tcp_reuse_buffer (vm, b); + tcp_make_ack_i (tc, b, TCP_STATE_ESTABLISHED, TCP_FLAG_ACK); + vnet_buffer (b)->tcp.flags = TCP_BUF_FLAG_ACK; +} + +/** + * Convert buffer to FIN-ACK + */ +void +tcp_make_finack (tcp_connection_t * tc, vlib_buffer_t * b) +{ + tcp_main_t *tm = vnet_get_tcp_main (); + vlib_main_t *vm = tm->vlib_main; + + tcp_reuse_buffer (vm, b); + tcp_make_ack_i (tc, b, TCP_STATE_ESTABLISHED, TCP_FLAG_ACK | TCP_FLAG_FIN); + + /* Reset flags, make sure ack is sent */ + tc->flags = TCP_CONN_SNDACK; + vnet_buffer (b)->tcp.flags &= ~TCP_BUF_FLAG_DUPACK; + + tc->snd_nxt += 1; +} + +/** + * Convert buffer to SYN-ACK + */ +void +tcp_make_synack (tcp_connection_t * tc, vlib_buffer_t * b) +{ + tcp_main_t *tm = vnet_get_tcp_main (); + vlib_main_t *vm = tm->vlib_main; + tcp_options_t _snd_opts, *snd_opts = &_snd_opts; + u8 tcp_opts_len, tcp_hdr_opts_len; + tcp_header_t *th; + u16 initial_wnd; + u32 time_now; + + memset (snd_opts, 0, sizeof (*snd_opts)); + + tcp_reuse_buffer (vm, b); + + /* Set random initial sequence */ + time_now = tcp_time_now (); + + tc->iss = random_u32 (&time_now); + tc->snd_una = tc->iss; + tc->snd_nxt = tc->iss + 1; + tc->snd_una_max = tc->snd_nxt; + + initial_wnd = tcp_initial_window_to_advertise (tc); + + /* Make and write options */ + tcp_opts_len = tcp_make_synack_options (tc, snd_opts); + tcp_hdr_opts_len = tcp_opts_len + sizeof (tcp_header_t); + + th = vlib_buffer_push_tcp (b, tc->c_lcl_port, tc->c_rmt_port, tc->iss, + tc->rcv_nxt, tcp_hdr_opts_len, + TCP_FLAG_SYN | TCP_FLAG_ACK, initial_wnd); + + tcp_options_write ((u8 *) (th + 1), snd_opts); + + vnet_buffer (b)->tcp.connection_index = tc->c_c_index; + vnet_buffer (b)->tcp.flags = TCP_BUF_FLAG_ACK; + + /* Init retransmit timer */ + tcp_retransmit_timer_set (tm, tc); +} + +always_inline void +tcp_enqueue_to_ip_lookup (vlib_main_t * vm, vlib_buffer_t * b, u32 bi, + u8 is_ip4) +{ + u32 *to_next, next_index; + vlib_frame_t *f; + + b->flags |= VNET_BUFFER_LOCALLY_ORIGINATED; + b->error = 0; + + /* Default FIB for now */ + vnet_buffer (b)->sw_if_index[VLIB_TX] = 0; + + /* Send to IP lookup */ + next_index = is_ip4 ? ip4_lookup_node.index : ip6_lookup_node.index; + f = vlib_get_frame_to_node (vm, next_index); + + /* Enqueue the packet */ + to_next = vlib_frame_vector_args (f); + to_next[0] = bi; + f->n_vectors = 1; + vlib_put_frame_to_node (vm, next_index, f); +} + +int +tcp_make_reset_in_place (vlib_main_t * vm, vlib_buffer_t * b0, + tcp_state_t state, u32 my_thread_index, u8 is_ip4) +{ + u8 tcp_hdr_len = sizeof (tcp_header_t); + ip4_header_t *ih4; + ip6_header_t *ih6; + tcp_header_t *th0; + ip4_address_t src_ip40; + ip6_address_t src_ip60; + u16 src_port0; + u32 tmp; + + /* Find IP and TCP headers */ + if (is_ip4) + { + ih4 = vlib_buffer_get_current (b0); + th0 = ip4_next_header (ih4); + } + else + { + ih6 = vlib_buffer_get_current (b0); + th0 = ip6_next_header (ih6); + } + + /* Swap src and dst ip */ + if (is_ip4) + { + ASSERT ((ih4->ip_version_and_header_length & 0xF0) == 0x40); + src_ip40.as_u32 = ih4->src_address.as_u32; + ih4->src_address.as_u32 = ih4->dst_address.as_u32; + ih4->dst_address.as_u32 = src_ip40.as_u32; + + /* Chop the end of the pkt */ + b0->current_length += ip4_header_bytes (ih4) + tcp_hdr_len; + } + else + { + ASSERT ((ih6->ip_version_traffic_class_and_flow_label & 0xF0) == 0x60); + clib_memcpy (&src_ip60, &ih6->src_address, sizeof (ip6_address_t)); + clib_memcpy (&ih6->src_address, &ih6->dst_address, + sizeof (ip6_address_t)); + clib_memcpy (&ih6->dst_address, &src_ip60, sizeof (ip6_address_t)); + + /* Chop the end of the pkt */ + b0->current_length += sizeof (ip6_header_t) + tcp_hdr_len; + } + + /* Try to determine what/why we're actually resetting and swap + * src and dst ports */ + if (state == TCP_STATE_CLOSED) + { + if (!tcp_syn (th0)) + return -1; + + tmp = clib_net_to_host_u32 (th0->seq_number); + + /* Got a SYN for no listener. */ + th0->flags = TCP_FLAG_RST | TCP_FLAG_ACK; + th0->ack_number = clib_host_to_net_u32 (tmp + 1); + th0->seq_number = 0; + + } + else if (state >= TCP_STATE_SYN_SENT) + { + th0->flags = TCP_FLAG_RST | TCP_FLAG_ACK; + th0->seq_number = th0->ack_number; + th0->ack_number = 0; + } + + src_port0 = th0->src_port; + th0->src_port = th0->dst_port; + th0->dst_port = src_port0; + th0->window = 0; + th0->data_offset_and_reserved = (tcp_hdr_len >> 2) << 4; + th0->urgent_pointer = 0; + + /* Compute checksum */ + if (is_ip4) + { + th0->checksum = ip4_tcp_udp_compute_checksum (vm, b0, ih4); + } + else + { + int bogus = ~0; + th0->checksum = ip6_tcp_udp_icmp_compute_checksum (vm, b0, ih6, &bogus); + ASSERT (!bogus); + } + + return 0; +} + +/** + * Send reset without reusing existing buffer + */ +void +tcp_send_reset (vlib_buffer_t * pkt, u8 is_ip4) +{ + vlib_buffer_t *b; + u32 bi; + tcp_main_t *tm = vnet_get_tcp_main (); + vlib_main_t *vm = tm->vlib_main; + u8 tcp_hdr_len, flags = 0; + tcp_header_t *th, *pkt_th; + u32 seq, ack; + ip4_header_t *ih4, *pkt_ih4; + ip6_header_t *ih6, *pkt_ih6; + + tcp_get_free_buffer_index (tm, &bi); + b = vlib_get_buffer (vm, bi); + + /* Leave enough space for headers */ + vlib_buffer_make_headroom (b, MAX_HDRS_LEN); + + /* Make and write options */ + tcp_hdr_len = sizeof (tcp_header_t); + + if (is_ip4) + { + pkt_ih4 = vlib_buffer_get_current (pkt); + pkt_th = ip4_next_header (pkt_ih4); + } + else + { + pkt_ih6 = vlib_buffer_get_current (pkt); + pkt_th = ip6_next_header (pkt_ih6); + } + + if (tcp_ack (pkt_th)) + { + flags = TCP_FLAG_RST; + seq = pkt_th->ack_number; + ack = 0; + } + else + { + flags = TCP_FLAG_RST | TCP_FLAG_ACK; + seq = 0; + ack = clib_host_to_net_u32 (vnet_buffer (pkt)->tcp.seq_end); + } + + th = vlib_buffer_push_tcp_net_order (b, pkt_th->dst_port, pkt_th->src_port, + seq, ack, tcp_hdr_len, flags, 0); + + /* Swap src and dst ip */ + if (is_ip4) + { + ASSERT ((pkt_ih4->ip_version_and_header_length & 0xF0) == 0x40); + ih4 = vlib_buffer_push_ip4 (vm, b, &pkt_ih4->dst_address, + &pkt_ih4->src_address, IP_PROTOCOL_TCP); + th->checksum = ip4_tcp_udp_compute_checksum (vm, b, ih4); + } + else + { + int bogus = ~0; + pkt_ih6 = (ip6_header_t *) (pkt_th - 1); + ASSERT ((pkt_ih6->ip_version_traffic_class_and_flow_label & 0xF0) == + 0x60); + ih6 = + vlib_buffer_push_ip6 (vm, b, &pkt_ih6->dst_address, + &pkt_ih6->src_address, IP_PROTOCOL_TCP); + th->checksum = ip6_tcp_udp_icmp_compute_checksum (vm, b, ih6, &bogus); + ASSERT (!bogus); + } + + tcp_enqueue_to_ip_lookup (vm, b, bi, is_ip4); +} + +void +tcp_push_ip_hdr (tcp_main_t * tm, tcp_connection_t * tc, vlib_buffer_t * b) +{ + tcp_header_t *th = vlib_buffer_get_current (b); + + if (tc->c_is_ip4) + { + ip4_header_t *ih; + ih = vlib_buffer_push_ip4 (tm->vlib_main, b, &tc->c_lcl_ip4, + &tc->c_rmt_ip4, IP_PROTOCOL_TCP); + th->checksum = ip4_tcp_udp_compute_checksum (tm->vlib_main, b, ih); + } + else + { + ip6_header_t *ih; + int bogus = ~0; + + ih = vlib_buffer_push_ip6 (tm->vlib_main, b, &tc->c_lcl_ip6, + &tc->c_rmt_ip6, IP_PROTOCOL_TCP); + th->checksum = ip6_tcp_udp_icmp_compute_checksum (tm->vlib_main, b, ih, + &bogus); + ASSERT (!bogus); + } +} + +/** + * Send SYN + * + * Builds a SYN packet for a half-open connection and sends it to ipx_lookup. + * The packet is not forwarded through tcpx_output to avoid doing lookups + * in the half_open pool. + */ +void +tcp_send_syn (tcp_connection_t * tc) +{ + vlib_buffer_t *b; + u32 bi; + tcp_main_t *tm = vnet_get_tcp_main (); + vlib_main_t *vm = tm->vlib_main; + u8 tcp_hdr_opts_len, tcp_opts_len; + tcp_header_t *th; + u32 time_now; + u16 initial_wnd; + tcp_options_t snd_opts; + + tcp_get_free_buffer_index (tm, &bi); + b = vlib_get_buffer (vm, bi); + + /* Leave enough space for headers */ + vlib_buffer_make_headroom (b, MAX_HDRS_LEN); + + /* Set random initial sequence */ + time_now = tcp_time_now (); + + tc->iss = random_u32 (&time_now); + tc->snd_una = tc->iss; + tc->snd_una_max = tc->snd_nxt = tc->iss + 1; + + initial_wnd = tcp_initial_window_to_advertise (tc); + + /* Make and write options */ + memset (&snd_opts, 0, sizeof (snd_opts)); + tcp_opts_len = tcp_make_syn_options (&snd_opts, initial_wnd); + tcp_hdr_opts_len = tcp_opts_len + sizeof (tcp_header_t); + + th = vlib_buffer_push_tcp (b, tc->c_lcl_port, tc->c_rmt_port, tc->iss, + tc->rcv_nxt, tcp_hdr_opts_len, TCP_FLAG_SYN, + initial_wnd); + + tcp_options_write ((u8 *) (th + 1), &snd_opts); + + /* Measure RTT with this */ + tc->rtt_ts = tcp_time_now (); + tc->rtt_seq = tc->snd_nxt; + + /* Start retransmit trimer */ + tcp_timer_set (tc, TCP_TIMER_RETRANSMIT_SYN, tc->rto * TCP_TO_TIMER_TICK); + tc->rto_boff = 0; + + /* Set the connection establishment timer */ + tcp_timer_set (tc, TCP_TIMER_ESTABLISH, TCP_ESTABLISH_TIME); + + tcp_push_ip_hdr (tm, tc, b); + tcp_enqueue_to_ip_lookup (vm, b, bi, tc->c_is_ip4); +} + +always_inline void +tcp_enqueue_to_output (vlib_main_t * vm, vlib_buffer_t * b, u32 bi, u8 is_ip4) +{ + u32 *to_next, next_index; + vlib_frame_t *f; + + b->flags |= VNET_BUFFER_LOCALLY_ORIGINATED; + b->error = 0; + + /* Decide where to send the packet */ + next_index = is_ip4 ? tcp4_output_node.index : tcp6_output_node.index; + f = vlib_get_frame_to_node (vm, next_index); + + /* Enqueue the packet */ + to_next = vlib_frame_vector_args (f); + to_next[0] = bi; + f->n_vectors = 1; + vlib_put_frame_to_node (vm, next_index, f); +} + +/** + * Send FIN + */ +void +tcp_send_fin (tcp_connection_t * tc) +{ + vlib_buffer_t *b; + u32 bi; + tcp_main_t *tm = vnet_get_tcp_main (); + vlib_main_t *vm = tm->vlib_main; + + tcp_get_free_buffer_index (tm, &bi); + b = vlib_get_buffer (vm, bi); + + /* Leave enough space for headers */ + vlib_buffer_make_headroom (b, MAX_HDRS_LEN); + + tcp_make_finack (tc, b); + + tcp_enqueue_to_output (vm, b, bi, tc->c_is_ip4); +} + +always_inline u8 +tcp_make_state_flags (tcp_state_t next_state) +{ + switch (next_state) + { + case TCP_STATE_ESTABLISHED: + return TCP_FLAG_ACK; + case TCP_STATE_SYN_RCVD: + return TCP_FLAG_SYN | TCP_FLAG_ACK; + case TCP_STATE_SYN_SENT: + return TCP_FLAG_SYN; + case TCP_STATE_LAST_ACK: + case TCP_STATE_FIN_WAIT_1: + return TCP_FLAG_FIN; + default: + clib_warning ("Shouldn't be here!"); + } + return 0; +} + +/** + * Push TCP header and update connection variables + */ +static void +tcp_push_hdr_i (tcp_connection_t * tc, vlib_buffer_t * b, + tcp_state_t next_state) +{ + u32 advertise_wnd, data_len; + u8 tcp_opts_len, tcp_hdr_opts_len, opts_write_len, flags; + tcp_options_t _snd_opts, *snd_opts = &_snd_opts; + tcp_header_t *th; + + data_len = b->current_length; + vnet_buffer (b)->tcp.flags = 0; + + /* Make and write options */ + memset (snd_opts, 0, sizeof (*snd_opts)); + tcp_opts_len = tcp_make_options (tc, snd_opts, next_state); + tcp_hdr_opts_len = tcp_opts_len + sizeof (tcp_header_t); + + /* Get rcv window to advertise */ + advertise_wnd = tcp_window_to_advertise (tc, next_state); + flags = tcp_make_state_flags (next_state); + + /* Push header and options */ + th = vlib_buffer_push_tcp (b, tc->c_lcl_port, tc->c_rmt_port, tc->snd_nxt, + tc->rcv_nxt, tcp_hdr_opts_len, flags, + advertise_wnd); + + opts_write_len = tcp_options_write ((u8 *) (th + 1), snd_opts); + + ASSERT (opts_write_len == tcp_opts_len); + + /* Tag the buffer with the connection index */ + vnet_buffer (b)->tcp.connection_index = tc->c_c_index; + + tc->snd_nxt += data_len; +} + +/* Send delayed ACK when timer expires */ +void +tcp_timer_delack_handler (u32 index) +{ + tcp_main_t *tm = vnet_get_tcp_main (); + vlib_main_t *vm = tm->vlib_main; + u32 thread_index = os_get_cpu_number (); + tcp_connection_t *tc; + vlib_buffer_t *b; + u32 bi; + + tc = tcp_connection_get (index, thread_index); + + /* Get buffer */ + tcp_get_free_buffer_index (tm, &bi); + b = vlib_get_buffer (vm, bi); + + /* Fill in the ACK */ + tcp_make_ack (tc, b); + + tc->timers[TCP_TIMER_DELACK] = TCP_TIMER_HANDLE_INVALID; + tc->flags &= ~TCP_CONN_DELACK; + + tcp_enqueue_to_output (vm, b, bi, tc->c_is_ip4); +} + +/** Build a retransmit segment + * + * @return the number of bytes in the segment or 0 if there's nothing to + * retransmit + * */ +u32 +tcp_prepare_retransmit_segment (tcp_connection_t * tc, vlib_buffer_t * b, + u32 max_bytes) +{ + tcp_main_t *tm = vnet_get_tcp_main (); + vlib_main_t *vm = tm->vlib_main; + u32 n_bytes, offset = 0; + sack_scoreboard_hole_t *hole; + u32 hole_size; + + tcp_reuse_buffer (vm, b); + + ASSERT (tc->state == TCP_STATE_ESTABLISHED); + ASSERT (max_bytes != 0); + + if (tcp_opts_sack_permitted (&tc->opt)) + { + /* XXX get first hole not retransmitted yet */ + hole = scoreboard_first_hole (&tc->sack_sb); + if (!hole) + return 0; + + offset = hole->start - tc->snd_una; + hole_size = hole->end - hole->start; + + ASSERT (hole_size); + + if (hole_size < max_bytes) + max_bytes = hole_size; + } + else + { + if (seq_geq (tc->snd_nxt, tc->snd_una_max)) + return 0; + } + + n_bytes = stream_session_peek_bytes (&tc->connection, + vlib_buffer_get_current (b), offset, + max_bytes); + ASSERT (n_bytes != 0); + + tc->snd_nxt += n_bytes; + tcp_push_hdr_i (tc, b, tc->state); + + return n_bytes; +} + +static void +tcp_timer_retransmit_handler_i (u32 index, u8 is_syn) +{ + tcp_main_t *tm = vnet_get_tcp_main (); + vlib_main_t *vm = tm->vlib_main; + u32 thread_index = os_get_cpu_number (); + tcp_connection_t *tc; + vlib_buffer_t *b; + u32 bi, max_bytes, snd_space; + + if (is_syn) + { + tc = tcp_half_open_connection_get (index); + } + else + { + tc = tcp_connection_get (index, thread_index); + } + + /* Make sure timer handle is set to invalid */ + tc->timers[TCP_TIMER_RETRANSMIT] = TCP_TIMER_HANDLE_INVALID; + + /* Increment RTO backoff (also equal to number of retries) */ + tc->rto_boff += 1; + + /* Go back to first un-acked byte */ + tc->snd_nxt = tc->snd_una; + + /* Get buffer */ + tcp_get_free_buffer_index (tm, &bi); + b = vlib_get_buffer (vm, bi); + + if (tc->state == TCP_STATE_ESTABLISHED) + { + tcp_fastrecovery_off (tc); + + /* Exponential backoff */ + tc->rto = clib_min (tc->rto << 1, TCP_RTO_MAX); + + /* Figure out what and how many bytes we can send */ + snd_space = tcp_available_snd_space (tc); + max_bytes = clib_min (tc->snd_mss, snd_space); + tcp_prepare_retransmit_segment (tc, b, max_bytes); + + tc->rtx_bytes += max_bytes; + + /* No fancy recovery for now! */ + scoreboard_clear (&tc->sack_sb); + } + else + { + /* Retransmit for SYN/SYNACK */ + ASSERT (tc->state == TCP_STATE_SYN_RCVD + || tc->state == TCP_STATE_SYN_SENT); + + /* Try without increasing RTO a number of times. If this fails, + * start growing RTO exponentially */ + if (tc->rto_boff > TCP_RTO_SYN_RETRIES) + tc->rto = clib_min (tc->rto << 1, TCP_RTO_MAX); + + vlib_buffer_make_headroom (b, MAX_HDRS_LEN); + tcp_push_hdr_i (tc, b, tc->state); + } + + if (!is_syn) + { + tcp_enqueue_to_output (vm, b, bi, tc->c_is_ip4); + + /* Re-enable retransmit timer */ + tcp_retransmit_timer_set (tm, tc); + } + else + { + ASSERT (tc->state == TCP_STATE_SYN_SENT); + + /* This goes straight to ipx_lookup */ + tcp_push_ip_hdr (tm, tc, b); + tcp_enqueue_to_ip_lookup (vm, b, bi, tc->c_is_ip4); + + /* Re-enable retransmit timer */ + tcp_timer_set (tc, TCP_TIMER_RETRANSMIT_SYN, + tc->rto * TCP_TO_TIMER_TICK); + } +} + +void +tcp_timer_retransmit_handler (u32 index) +{ + tcp_timer_retransmit_handler_i (index, 0); +} + +void +tcp_timer_retransmit_syn_handler (u32 index) +{ + tcp_timer_retransmit_handler_i (index, 1); +} + +/** + * Retansmit first unacked segment */ +void +tcp_retransmit_first_unacked (tcp_connection_t * tc) +{ + tcp_main_t *tm = vnet_get_tcp_main (); + u32 snd_nxt = tc->snd_nxt; + vlib_buffer_t *b; + u32 bi; + + tc->snd_nxt = tc->snd_una; + + /* Get buffer */ + tcp_get_free_buffer_index (tm, &bi); + b = vlib_get_buffer (tm->vlib_main, bi); + + tcp_prepare_retransmit_segment (tc, b, tc->snd_mss); + tcp_enqueue_to_output (tm->vlib_main, b, bi, tc->c_is_ip4); + + tc->snd_nxt = snd_nxt; + tc->rtx_bytes += tc->snd_mss; +} + +void +tcp_fast_retransmit (tcp_connection_t * tc) +{ + tcp_main_t *tm = vnet_get_tcp_main (); + u32 snd_space, max_bytes, n_bytes, bi; + vlib_buffer_t *b; + + ASSERT (tcp_in_fastrecovery (tc)); + + clib_warning ("fast retransmit!"); + + /* Start resending from first un-acked segment */ + tc->snd_nxt = tc->snd_una; + + snd_space = tcp_available_snd_space (tc); + + while (snd_space) + { + tcp_get_free_buffer_index (tm, &bi); + b = vlib_get_buffer (tm->vlib_main, bi); + + max_bytes = clib_min (tc->snd_mss, snd_space); + n_bytes = tcp_prepare_retransmit_segment (tc, b, max_bytes); + + /* Nothing left to retransmit */ + if (n_bytes == 0) + return; + + tcp_enqueue_to_output (tm->vlib_main, b, bi, tc->c_is_ip4); + + snd_space -= n_bytes; + } + + /* If window allows, send new data */ + tc->snd_nxt = tc->snd_una_max; +} + +always_inline u32 +tcp_session_has_ooo_data (tcp_connection_t * tc) +{ + stream_session_t *s = + stream_session_get (tc->c_s_index, tc->c_thread_index); + return svm_fifo_has_ooo_data (s->server_rx_fifo); +} + +always_inline uword +tcp46_output_inline (vlib_main_t * vm, + vlib_node_runtime_t * node, + vlib_frame_t * from_frame, int is_ip4) +{ + tcp_main_t *tm = vnet_get_tcp_main (); + u32 n_left_from, next_index, *from, *to_next; + u32 my_thread_index = vm->cpu_index; + + from = vlib_frame_vector_args (from_frame); + n_left_from = from_frame->n_vectors; + + next_index = node->cached_next_index; + + while (n_left_from > 0) + { + u32 n_left_to_next; + + vlib_get_next_frame (vm, node, next_index, to_next, n_left_to_next); + + while (n_left_from > 0 && n_left_to_next > 0) + { + u32 bi0; + vlib_buffer_t *b0; + tcp_connection_t *tc0; + tcp_header_t *th0; + u32 error0 = TCP_ERROR_PKTS_SENT, next0 = TCP_OUTPUT_NEXT_IP_LOOKUP; + + bi0 = from[0]; + to_next[0] = bi0; + from += 1; + to_next += 1; + n_left_from -= 1; + n_left_to_next -= 1; + + b0 = vlib_get_buffer (vm, bi0); + tc0 = tcp_connection_get (vnet_buffer (b0)->tcp.connection_index, + my_thread_index); + th0 = vlib_buffer_get_current (b0); + + if (is_ip4) + { + ip4_header_t *ih0; + ih0 = vlib_buffer_push_ip4 (vm, b0, &tc0->c_lcl_ip4, + &tc0->c_rmt_ip4, IP_PROTOCOL_TCP); + th0->checksum = ip4_tcp_udp_compute_checksum (vm, b0, ih0); + } + else + { + ip6_header_t *ih0; + int bogus = ~0; + + ih0 = vlib_buffer_push_ip6 (vm, b0, &tc0->c_lcl_ip6, + &tc0->c_rmt_ip6, IP_PROTOCOL_TCP); + th0->checksum = ip6_tcp_udp_icmp_compute_checksum (vm, b0, ih0, + &bogus); + ASSERT (!bogus); + } + + /* Filter out DUPACKs if there are no OOO segments left */ + if (PREDICT_FALSE + (vnet_buffer (b0)->tcp.flags & TCP_BUF_FLAG_DUPACK)) + { + tc0->snt_dupacks--; + ASSERT (tc0->snt_dupacks >= 0); + if (!tcp_session_has_ooo_data (tc0)) + { + error0 = TCP_ERROR_FILTERED_DUPACKS; + next0 = TCP_OUTPUT_NEXT_DROP; + goto done; + } + } + + /* Retransmitted SYNs do reach this but it should be harmless */ + tc0->rcv_las = tc0->rcv_nxt; + + /* Stop DELACK timer and fix flags */ + tc0->flags &= + ~(TCP_CONN_SNDACK | TCP_CONN_DELACK | TCP_CONN_BURSTACK); + if (tcp_timer_is_active (tc0, TCP_TIMER_DELACK)) + { + tcp_timer_reset (tc0, TCP_TIMER_DELACK); + } + + /* If not retransmitting + * 1) update snd_una_max (SYN, SYNACK, new data, FIN) + * 2) If we're not tracking an ACK, start tracking */ + if (seq_lt (tc0->snd_una_max, tc0->snd_nxt)) + { + tc0->snd_una_max = tc0->snd_nxt; + if (tc0->rtt_ts == 0) + { + tc0->rtt_ts = tcp_time_now (); + tc0->rtt_seq = tc0->snd_nxt; + } + } + + /* Set the retransmit timer if not set already and not + * doing a pure ACK */ + if (!tcp_timer_is_active (tc0, TCP_TIMER_RETRANSMIT) + && tc0->snd_nxt != tc0->snd_una) + { + tcp_retransmit_timer_set (tm, tc0); + tc0->rto_boff = 0; + } + + /* set fib index to default and lookup node */ + /* XXX network virtualization (vrf/vni) */ + vnet_buffer (b0)->sw_if_index[VLIB_RX] = 0; + vnet_buffer (b0)->sw_if_index[VLIB_TX] = (u32) ~ 0; + + b0->flags |= VNET_BUFFER_LOCALLY_ORIGINATED; + + done: + b0->error = error0 != 0 ? node->errors[error0] : 0; + if (PREDICT_FALSE (b0->flags & VLIB_BUFFER_IS_TRACED)) + { + + } + + vlib_validate_buffer_enqueue_x1 (vm, node, next_index, to_next, + n_left_to_next, bi0, next0); + } + + vlib_put_next_frame (vm, node, next_index, n_left_to_next); + } + + return from_frame->n_vectors; +} + +static uword +tcp4_output (vlib_main_t * vm, vlib_node_runtime_t * node, + vlib_frame_t * from_frame) +{ + return tcp46_output_inline (vm, node, from_frame, 1 /* is_ip4 */ ); +} + +static uword +tcp6_output (vlib_main_t * vm, vlib_node_runtime_t * node, + vlib_frame_t * from_frame) +{ + return tcp46_output_inline (vm, node, from_frame, 0 /* is_ip4 */ ); +} + +VLIB_REGISTER_NODE (tcp4_output_node) = +{ + .function = tcp4_output,.name = "tcp4-output", + /* Takes a vector of packets. */ + .vector_size = sizeof (u32),.n_errors = TCP_N_ERROR,.error_strings = + tcp_error_strings,.n_next_nodes = TCP_OUTPUT_N_NEXT,.next_nodes = + { +#define _(s,n) [TCP_OUTPUT_NEXT_##s] = n, + foreach_tcp4_output_next +#undef _ + } +,.format_buffer = format_tcp_header,.format_trace = format_tcp_tx_trace,}; + +VLIB_NODE_FUNCTION_MULTIARCH (tcp4_output_node, tcp4_output) +VLIB_REGISTER_NODE (tcp6_output_node) = +{ + .function = tcp6_output,.name = "tcp6-output", + /* Takes a vector of packets. */ + .vector_size = sizeof (u32),.n_errors = TCP_N_ERROR,.error_strings = + tcp_error_strings,.n_next_nodes = TCP_OUTPUT_N_NEXT,.next_nodes = + { +#define _(s,n) [TCP_OUTPUT_NEXT_##s] = n, + foreach_tcp6_output_next +#undef _ + } +,.format_buffer = format_tcp_header,.format_trace = format_tcp_tx_trace,}; + +VLIB_NODE_FUNCTION_MULTIARCH (tcp6_output_node, tcp6_output) u32 +tcp_push_header (transport_connection_t * tconn, vlib_buffer_t * b) +{ + tcp_connection_t *tc; + + tc = (tcp_connection_t *) tconn; + tcp_push_hdr_i (tc, b, TCP_STATE_ESTABLISHED); + return 0; +} + +typedef enum _tcp_reset_next +{ + TCP_RESET_NEXT_DROP, + TCP_RESET_NEXT_IP_LOOKUP, + TCP_RESET_N_NEXT +} tcp_reset_next_t; + +#define foreach_tcp4_reset_next \ + _(DROP, "error-drop") \ + _(IP_LOOKUP, "ip4-lookup") + +#define foreach_tcp6_reset_next \ + _(DROP, "error-drop") \ + _(IP_LOOKUP, "ip6-lookup") + +static uword +tcp46_send_reset_inline (vlib_main_t * vm, vlib_node_runtime_t * node, + vlib_frame_t * from_frame, u8 is_ip4) +{ + u32 n_left_from, next_index, *from, *to_next; + u32 my_thread_index = vm->cpu_index; + + from = vlib_frame_vector_args (from_frame); + n_left_from = from_frame->n_vectors; + + next_index = node->cached_next_index; + + while (n_left_from > 0) + { + u32 n_left_to_next; + + vlib_get_next_frame (vm, node, next_index, to_next, n_left_to_next); + + while (n_left_from > 0 && n_left_to_next > 0) + { + u32 bi0; + vlib_buffer_t *b0; + u32 error0 = TCP_ERROR_RST_SENT, next0 = TCP_RESET_NEXT_IP_LOOKUP; + + bi0 = from[0]; + to_next[0] = bi0; + from += 1; + to_next += 1; + n_left_from -= 1; + n_left_to_next -= 1; + + b0 = vlib_get_buffer (vm, bi0); + + if (tcp_make_reset_in_place (vm, b0, vnet_buffer (b0)->tcp.flags, + my_thread_index, is_ip4)) + { + error0 = TCP_ERROR_LOOKUP_DROPS; + next0 = TCP_RESET_NEXT_DROP; + goto done; + } + + /* Prepare to send to IP lookup */ + vnet_buffer (b0)->sw_if_index[VLIB_TX] = 0; + next0 = TCP_RESET_NEXT_IP_LOOKUP; + + done: + b0->error = error0 != 0 ? node->errors[error0] : 0; + b0->flags |= VNET_BUFFER_LOCALLY_ORIGINATED; + if (PREDICT_FALSE (b0->flags & VLIB_BUFFER_IS_TRACED)) + { + + } + + vlib_validate_buffer_enqueue_x1 (vm, node, next_index, to_next, + n_left_to_next, bi0, next0); + } + vlib_put_next_frame (vm, node, next_index, n_left_to_next); + } + return from_frame->n_vectors; +} + +static uword +tcp4_send_reset (vlib_main_t * vm, vlib_node_runtime_t * node, + vlib_frame_t * from_frame) +{ + return tcp46_send_reset_inline (vm, node, from_frame, 1); +} + +static uword +tcp6_send_reset (vlib_main_t * vm, vlib_node_runtime_t * node, + vlib_frame_t * from_frame) +{ + return tcp46_send_reset_inline (vm, node, from_frame, 0); +} + +/* *INDENT-OFF* */ +VLIB_REGISTER_NODE (tcp4_reset_node) = { + .function = tcp4_send_reset, + .name = "tcp4-reset", + .vector_size = sizeof (u32), + .n_errors = TCP_N_ERROR, + .error_strings = tcp_error_strings, + .n_next_nodes = TCP_RESET_N_NEXT, + .next_nodes = { +#define _(s,n) [TCP_RESET_NEXT_##s] = n, + foreach_tcp4_reset_next +#undef _ + }, +}; +/* *INDENT-ON* */ + +/* *INDENT-OFF* */ +VLIB_REGISTER_NODE (tcp6_reset_node) = { + .function = tcp6_send_reset, + .name = "tcp6-reset", + .vector_size = sizeof (u32), + .n_errors = TCP_N_ERROR, + .error_strings = tcp_error_strings, + .n_next_nodes = TCP_RESET_N_NEXT, + .next_nodes = { +#define _(s,n) [TCP_RESET_NEXT_##s] = n, + foreach_tcp6_reset_next +#undef _ + }, +}; +/* *INDENT-ON* */ + +/* + * fd.io coding-style-patch-verification: ON + * + * Local Variables: + * eval: (c-set-style "gnu") + * End: + */ diff --git a/src/vnet/tcp/tcp_packet.h b/src/vnet/tcp/tcp_packet.h new file mode 100644 index 00000000..866c5fd6 --- /dev/null +++ b/src/vnet/tcp/tcp_packet.h @@ -0,0 +1,184 @@ +/* + * Copyright (c) 2016 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +#ifndef included_tcp_packet_h +#define included_tcp_packet_h + +#include + +/* TCP flags bit 0 first. */ +#define foreach_tcp_flag \ + _ (FIN) /**< No more data from sender. */ \ + _ (SYN) /**< Synchronize sequence numbers. */ \ + _ (RST) /**< Reset the connection. */ \ + _ (PSH) /**< Push function. */ \ + _ (ACK) /**< Ack field significant. */ \ + _ (URG) /**< Urgent pointer field significant. */ \ + _ (ECE) /**< ECN-echo. Receiver got CE packet */ \ + _ (CWR) /**< Sender reduced congestion window */ + +enum +{ +#define _(f) TCP_FLAG_BIT_##f, + foreach_tcp_flag +#undef _ + TCP_N_FLAG_BITS, +}; + +enum +{ +#define _(f) TCP_FLAG_##f = 1 << TCP_FLAG_BIT_##f, + foreach_tcp_flag +#undef _ +}; + +typedef struct _tcp_header +{ + union + { + struct + { + u16 src_port; /**< Source port. */ + u16 dst_port; /**< Destination port. */ + }; + struct + { + u16 src, dst; + }; + }; + + u32 seq_number; /**< Sequence number of the first data octet in this + * segment, except when SYN is present. If SYN + * is present the seq number is is the ISN and the + * first data octet is ISN+1 */ + u32 ack_number; /**< Acknowledgement number if ACK is set. It contains + * the value of the next sequence number the sender + * of the segment is expecting to receive. */ + u8 data_offset_and_reserved; + u8 flags; /**< Flags: see the macro above */ + u16 window; /**< Number of bytes sender is willing to receive. */ + + u16 checksum; /**< Checksum of TCP pseudo header and data. */ + u16 urgent_pointer; /**< Seq number of the byte after the urgent data. */ +} __attribute__ ((packed)) tcp_header_t; + +/* Flag tests that return 0 or !0 */ +#define tcp_doff(_th) ((_th)->data_offset_and_reserved >> 4) +#define tcp_fin(_th) ((_th)->flags & TCP_FLAG_FIN) +#define tcp_syn(_th) ((_th)->flags & TCP_FLAG_SYN) +#define tcp_rst(_th) ((_th)->flags & TCP_FLAG_RST) +#define tcp_psh(_th) ((_th)->flags & TCP_FLAG_PSH) +#define tcp_ack(_th) ((_th)->flags & TCP_FLAG_ACK) +#define tcp_urg(_th) ((_th)->flags & TCP_FLAG_URG) +#define tcp_ece(_th) ((_th)->flags & TCP_FLAG_ECE) +#define tcp_cwr(_th) ((_th)->flags & TCP_FLAG_CWR) + +/* Flag tests that return 0 or 1 */ +#define tcp_is_syn(_th) !!((_th)->flags & TCP_FLAG_SYN) +#define tcp_is_fin(_th) !!((_th)->flags & TCP_FLAG_FIN) + +always_inline int +tcp_header_bytes (tcp_header_t * t) +{ + return tcp_doff (t) * sizeof (u32); +} + +/* + * TCP options. + */ + +typedef enum tcp_option_type +{ + TCP_OPTION_EOL = 0, /**< End of options. */ + TCP_OPTION_NOOP = 1, /**< No operation. */ + TCP_OPTION_MSS = 2, /**< Limit MSS. */ + TCP_OPTION_WINDOW_SCALE = 3, /**< Window scale. */ + TCP_OPTION_SACK_PERMITTED = 4, /**< Selective Ack permitted. */ + TCP_OPTION_SACK_BLOCK = 5, /**< Selective Ack block. */ + TCP_OPTION_TIMESTAMP = 8, /**< Timestamps. */ + TCP_OPTION_UTO = 28, /**< User timeout. */ + TCP_OPTION_AO = 29, /**< Authentication Option. */ +} tcp_option_type_t; + +#define foreach_tcp_options_flag \ + _ (MSS) /**< MSS advertised in SYN */ \ + _ (TSTAMP) /**< Timestamp capability advertised in SYN */ \ + _ (WSCALE) /**< Wnd scale capability advertised in SYN */ \ + _ (SACK_PERMITTED) /**< SACK capability advertised in SYN */ \ + _ (SACK) /**< SACK present */ + +enum +{ +#define _(f) TCP_OPTS_FLAG_BIT_##f, + foreach_tcp_options_flag +#undef _ + TCP_OPTIONS_N_FLAG_BITS, +}; + +enum +{ +#define _(f) TCP_OPTS_FLAG_##f = 1 << TCP_OPTS_FLAG_BIT_##f, + foreach_tcp_options_flag +#undef _ +}; + +typedef struct _sack_block +{ + u32 start; /**< Start sequence number */ + u32 end; /**< End sequence number */ +} sack_block_t; + +typedef struct +{ + u8 flags; /** Option flags, see above */ + + /* Received options */ + u16 mss; /**< Maximum segment size advertised by peer */ + u8 wscale; /**< Window scale advertised by peer */ + u32 tsval; /**< Peer's timestamp value */ + u32 tsecr; /**< Echoed/reflected time stamp */ + sack_block_t *sacks; /**< SACK blocks received */ + u8 n_sack_blocks; /**< Number of SACKs blocks */ +} tcp_options_t; + +/* Flag tests that return 0 or !0 */ +#define tcp_opts_mss(_to) ((_to)->flags & TCP_OPTS_FLAG_MSS) +#define tcp_opts_tstamp(_to) ((_to)->flags & TCP_OPTS_FLAG_TSTAMP) +#define tcp_opts_wscale(_to) ((_to)->flags & TCP_OPTS_FLAG_WSCALE) +#define tcp_opts_sack(_to) ((_to)->flags & TCP_OPTS_FLAG_SACK) +#define tcp_opts_sack_permitted(_to) ((_to)->flags & TCP_OPTS_FLAG_SACK_PERMITTED) + +/* TCP option lengths */ +#define TCP_OPTION_LEN_EOL 1 +#define TCP_OPTION_LEN_NOOP 1 +#define TCP_OPTION_LEN_MSS 4 +#define TCP_OPTION_LEN_WINDOW_SCALE 3 +#define TCP_OPTION_LEN_SACK_PERMITTED 2 +#define TCP_OPTION_LEN_TIMESTAMP 10 +#define TCP_OPTION_LEN_SACK_BLOCK 8 + +#define TCP_WND_MAX 65535U +#define TCP_MAX_WND_SCALE 14 /* See RFC 1323 */ +#define TCP_OPTS_ALIGN 4 +#define TCP_OPTS_MAX_SACK_BLOCKS 3 +#endif /* included_tcp_packet_h */ + +/* + * fd.io coding-style-patch-verification: ON + * + * Local Variables: + * eval: (c-set-style "gnu") + * End: + */ diff --git a/src/vnet/tcp/tcp_pg.c b/src/vnet/tcp/tcp_pg.c new file mode 100644 index 00000000..dc324049 --- /dev/null +++ b/src/vnet/tcp/tcp_pg.c @@ -0,0 +1,236 @@ +/* + * Copyright (c) 2016 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ +/* + * ip/tcp_pg: TCP packet-generator interface + * + * Copyright (c) 2008 Eliot Dresselhaus + * + * Permission is hereby granted, free of charge, to any person obtaining + * a copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sublicense, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice shall be + * included in all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND + * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE + * LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + * OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + * WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + */ + +#include +#include + +/* TCP flags bit 0 first. */ +#define foreach_tcp_flag \ + _ (FIN) \ + _ (SYN) \ + _ (RST) \ + _ (PSH) \ + _ (ACK) \ + _ (URG) \ + _ (ECE) \ + _ (CWR) + +static void +tcp_pg_edit_function (pg_main_t * pg, + pg_stream_t * s, + pg_edit_group_t * g, + u32 * packets, + u32 n_packets) +{ + vlib_main_t * vm = vlib_get_main(); + u32 ip_offset, tcp_offset; + + tcp_offset = g->start_byte_offset; + ip_offset = (g-1)->start_byte_offset; + + while (n_packets >= 1) + { + vlib_buffer_t * p0; + ip4_header_t * ip0; + tcp_header_t * tcp0; + ip_csum_t sum0; + u32 tcp_len0; + + p0 = vlib_get_buffer (vm, packets[0]); + n_packets -= 1; + packets += 1; + + ASSERT (p0->current_data == 0); + ip0 = (void *) (p0->data + ip_offset); + tcp0 = (void *) (p0->data + tcp_offset); + tcp_len0 = clib_net_to_host_u16 (ip0->length) - sizeof (ip0[0]); + + /* Initialize checksum with header. */ + if (BITS (sum0) == 32) + { + sum0 = clib_mem_unaligned (&ip0->src_address, u32); + sum0 = ip_csum_with_carry (sum0, clib_mem_unaligned (&ip0->dst_address, u32)); + } + else + sum0 = clib_mem_unaligned (&ip0->src_address, u64); + + sum0 = ip_csum_with_carry + (sum0, clib_host_to_net_u32 (tcp_len0 + (ip0->protocol << 16))); + + /* Invalidate possibly old checksum. */ + tcp0->checksum = 0; + + sum0 = ip_incremental_checksum_buffer (vm, p0, tcp_offset, tcp_len0, sum0); + + tcp0->checksum = ~ ip_csum_fold (sum0); + } +} + +typedef struct { + pg_edit_t src, dst; + pg_edit_t seq_number, ack_number; + pg_edit_t data_offset_and_reserved; +#define _(f) pg_edit_t f##_flag; + foreach_tcp_flag +#undef _ + pg_edit_t window; + pg_edit_t checksum; + pg_edit_t urgent_pointer; +} pg_tcp_header_t; + +static inline void +pg_tcp_header_init (pg_tcp_header_t * p) +{ + /* Initialize fields that are not bit fields in the IP header. */ +#define _(f) pg_edit_init (&p->f, tcp_header_t, f); + _ (src); + _ (dst); + _ (seq_number); + _ (ack_number); + _ (window); + _ (checksum); + _ (urgent_pointer); +#undef _ + + /* Initialize bit fields. */ +#define _(f) \ + pg_edit_init_bitfield (&p->f##_flag, tcp_header_t, \ + flags, \ + TCP_FLAG_BIT_##f, 1); + + foreach_tcp_flag +#undef _ + + pg_edit_init_bitfield (&p->data_offset_and_reserved, tcp_header_t, + data_offset_and_reserved, + 4, 4); +} + +uword +unformat_pg_tcp_header (unformat_input_t * input, va_list * args) +{ + pg_stream_t * s = va_arg (*args, pg_stream_t *); + pg_tcp_header_t * p; + u32 group_index; + + p = pg_create_edit_group (s, sizeof (p[0]), sizeof (tcp_header_t), + &group_index); + pg_tcp_header_init (p); + + /* Defaults. */ + pg_edit_set_fixed (&p->seq_number, 0); + pg_edit_set_fixed (&p->ack_number, 0); + + pg_edit_set_fixed (&p->data_offset_and_reserved, + sizeof (tcp_header_t) / sizeof (u32)); + + pg_edit_set_fixed (&p->window, 4096); + pg_edit_set_fixed (&p->urgent_pointer, 0); + +#define _(f) pg_edit_set_fixed (&p->f##_flag, 0); + foreach_tcp_flag +#undef _ + + p->checksum.type = PG_EDIT_UNSPECIFIED; + + if (! unformat (input, "TCP: %U -> %U", + unformat_pg_edit, + unformat_tcp_udp_port, &p->src, + unformat_pg_edit, + unformat_tcp_udp_port, &p->dst)) + goto error; + + /* Parse options. */ + while (1) + { + if (unformat (input, "window %U", + unformat_pg_edit, + unformat_pg_number, &p->window)) + ; + + else if (unformat (input, "checksum %U", + unformat_pg_edit, + unformat_pg_number, &p->checksum)) + ; + + /* Flags. */ +#define _(f) else if (unformat (input, #f)) pg_edit_set_fixed (&p->f##_flag, 1); + foreach_tcp_flag +#undef _ + + /* Can't parse input: try next protocol level. */ + else + break; + } + + { + ip_main_t * im = &ip_main; + u16 dst_port; + tcp_udp_port_info_t * pi; + + pi = 0; + if (p->dst.type == PG_EDIT_FIXED) + { + dst_port = pg_edit_get_value (&p->dst, PG_EDIT_LO); + pi = ip_get_tcp_udp_port_info (im, dst_port); + } + + if (pi && pi->unformat_pg_edit + && unformat_user (input, pi->unformat_pg_edit, s)) + ; + + else if (! unformat_user (input, unformat_pg_payload, s)) + goto error; + + if (p->checksum.type == PG_EDIT_UNSPECIFIED) + { + pg_edit_group_t * g = pg_stream_get_group (s, group_index); + g->edit_function = tcp_pg_edit_function; + g->edit_function_opaque = 0; + } + + return 1; + } + + error: + /* Free up any edits we may have added. */ + pg_free_edit_group (s); + return 0; +} + diff --git a/src/vnet/tcp/tcp_syn_filter4.c b/src/vnet/tcp/tcp_syn_filter4.c new file mode 100644 index 00000000..c7605a30 --- /dev/null +++ b/src/vnet/tcp/tcp_syn_filter4.c @@ -0,0 +1,542 @@ +/* + * Copyright (c) 2016 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +#include +#include +#include +#include +#include +#include +#include + +typedef struct +{ + f64 next_reset; + f64 reset_interval; + u8 *syn_counts; +} syn_filter4_runtime_t; + +typedef struct +{ + u32 next_index; + int not_a_syn; + u8 filter_value; +} syn_filter4_trace_t; + +/* packet trace format function */ +static u8 * +format_syn_filter4_trace (u8 * s, va_list * args) +{ + CLIB_UNUSED (vlib_main_t * vm) = va_arg (*args, vlib_main_t *); + CLIB_UNUSED (vlib_node_t * node) = va_arg (*args, vlib_node_t *); + syn_filter4_trace_t *t = va_arg (*args, syn_filter4_trace_t *); + + s = format (s, "SYN_FILTER4: next index %d, %s", + t->next_index, t->not_a_syn ? "not a syn" : "syn"); + if (t->not_a_syn == 0) + s = format (s, ", filter value %d\n", t->filter_value); + else + s = format (s, "\n"); + return s; +} + +static vlib_node_registration_t syn_filter4_node; + +#define foreach_syn_filter_error \ +_(THROTTLED, "TCP SYN packet throttle drops") \ +_(OK, "TCP SYN packets passed") + +typedef enum +{ +#define _(sym,str) SYN_FILTER_ERROR_##sym, + foreach_syn_filter_error +#undef _ + SYN_FILTER_N_ERROR, +} syn_filter_error_t; + +static char *syn_filter4_error_strings[] = { +#define _(sym,string) string, + foreach_syn_filter_error +#undef _ +}; + +typedef enum +{ + SYN_FILTER_NEXT_DROP, + SYN_FILTER_N_NEXT, +} syn_filter_next_t; + +extern vnet_feature_arc_registration_t vnet_feat_arc_ip4_local; + +static uword +syn_filter4_node_fn (vlib_main_t * vm, + vlib_node_runtime_t * node, vlib_frame_t * frame) +{ + u32 n_left_from, *from, *to_next; + syn_filter_next_t next_index; + u32 ok_syn_packets = 0; + vnet_feature_main_t *fm = &feature_main; + u8 arc_index = vnet_feat_arc_ip4_local.feature_arc_index; + vnet_feature_config_main_t *cm = &fm->feature_config_mains[arc_index]; + syn_filter4_runtime_t *rt = (syn_filter4_runtime_t *) node->runtime_data; + f64 now = vlib_time_now (vm); + /* Shut up spurious gcc warnings. */ + u8 *c0 = 0, *c1 = 0, *c2 = 0, *c3 = 0; + + from = vlib_frame_vector_args (frame); + n_left_from = frame->n_vectors; + next_index = node->cached_next_index; + + if (now > rt->next_reset) + { + memset (rt->syn_counts, 0, vec_len (rt->syn_counts)); + rt->next_reset = now + rt->reset_interval; + } + + while (n_left_from > 0) + { + u32 n_left_to_next; + + vlib_get_next_frame (vm, node, next_index, to_next, n_left_to_next); + + while (n_left_from >= 8 && n_left_to_next >= 4) + { + u32 bi0, bi1, bi2, bi3; + vlib_buffer_t *b0, *b1, *b2, *b3; + u32 next0, next1, next2, next3; + ip4_header_t *ip0, *ip1, *ip2, *ip3; + tcp_header_t *tcp0, *tcp1, *tcp2, *tcp3; + u32 not_a_syn0 = 1, not_a_syn1 = 1, not_a_syn2 = 1, not_a_syn3 = 1; + u64 hash0, hash1, hash2, hash3; + + /* Prefetch next iteration. */ + { + vlib_buffer_t *p4, *p5, *p6, *p7; + + p4 = vlib_get_buffer (vm, from[4]); + p5 = vlib_get_buffer (vm, from[5]); + p6 = vlib_get_buffer (vm, from[6]); + p7 = vlib_get_buffer (vm, from[7]); + + vlib_prefetch_buffer_header (p4, LOAD); + vlib_prefetch_buffer_header (p5, LOAD); + vlib_prefetch_buffer_header (p6, LOAD); + vlib_prefetch_buffer_header (p7, LOAD); + + CLIB_PREFETCH (p4->data, CLIB_CACHE_LINE_BYTES, STORE); + CLIB_PREFETCH (p5->data, CLIB_CACHE_LINE_BYTES, STORE); + CLIB_PREFETCH (p6->data, CLIB_CACHE_LINE_BYTES, STORE); + CLIB_PREFETCH (p7->data, CLIB_CACHE_LINE_BYTES, STORE); + } + + /* speculatively enqueue b0 and b1 to the current next frame */ + to_next[0] = bi0 = from[0]; + to_next[1] = bi1 = from[1]; + to_next[2] = bi2 = from[2]; + to_next[3] = bi3 = from[3]; + from += 4; + to_next += 4; + n_left_from -= 4; + n_left_to_next -= 4; + + b0 = vlib_get_buffer (vm, bi0); + b1 = vlib_get_buffer (vm, bi1); + b2 = vlib_get_buffer (vm, bi2); + b3 = vlib_get_buffer (vm, bi3); + + vnet_get_config_data + (&cm->config_main, &b0->current_config_index, + &next0, 0 /* sizeof (c0[0]) */ ); + vnet_get_config_data + (&cm->config_main, &b1->current_config_index, + &next1, 0 /* sizeof (c0[0]) */ ); + vnet_get_config_data + (&cm->config_main, &b2->current_config_index, + &next2, 0 /* sizeof (c0[0]) */ ); + vnet_get_config_data + (&cm->config_main, &b3->current_config_index, + &next3, 0 /* sizeof (c0[0]) */ ); + + /* Not TCP? */ + ip0 = vlib_buffer_get_current (b0); + if (ip0->protocol != IP_PROTOCOL_TCP) + goto trace00; + + tcp0 = ip4_next_header (ip0); + /* + * Not a SYN? + * $$$$ hack: the TCP bitfield flags seem not to compile + * correct code. + */ + if (PREDICT_TRUE (!(tcp0->flags & 0x2))) + goto trace00; + + not_a_syn0 = 0; + hash0 = clib_xxhash ((u64) ip0->src_address.as_u32); + c0 = &rt->syn_counts[hash0 & (_vec_len (rt->syn_counts) - 1)]; + if (PREDICT_FALSE (*c0 >= 0x80)) + { + next0 = SYN_FILTER_NEXT_DROP; + b0->error = node->errors[SYN_FILTER_ERROR_THROTTLED]; + goto trace00; + } + *c0 += 1; + ok_syn_packets++; + + trace00: + if (PREDICT_FALSE ((node->flags & VLIB_NODE_FLAG_TRACE) + && (b0->flags & VLIB_BUFFER_IS_TRACED))) + { + syn_filter4_trace_t *t = + vlib_add_trace (vm, node, b0, sizeof (*t)); + t->not_a_syn = not_a_syn0; + t->next_index = next0; + t->filter_value = not_a_syn0 ? 0 : *c0; + } + + /* Not TCP? */ + ip1 = vlib_buffer_get_current (b1); + if (ip1->protocol != IP_PROTOCOL_TCP) + goto trace01; + + tcp1 = ip4_next_header (ip1); + /* + * Not a SYN? + * $$$$ hack: the TCP bitfield flags seem not to compile + * correct code. + */ + if (PREDICT_TRUE (!(tcp1->flags & 0x2))) + goto trace01; + + not_a_syn1 = 0; + hash1 = clib_xxhash ((u64) ip1->src_address.as_u32); + c1 = &rt->syn_counts[hash1 & (_vec_len (rt->syn_counts) - 1)]; + if (PREDICT_FALSE (*c1 >= 0x80)) + { + next1 = SYN_FILTER_NEXT_DROP; + b1->error = node->errors[SYN_FILTER_ERROR_THROTTLED]; + goto trace01; + } + *c1 += 1; + ok_syn_packets++; + + trace01: + if (PREDICT_FALSE ((node->flags & VLIB_NODE_FLAG_TRACE) + && (b1->flags & VLIB_BUFFER_IS_TRACED))) + { + syn_filter4_trace_t *t = + vlib_add_trace (vm, node, b1, sizeof (*t)); + t->not_a_syn = not_a_syn1; + t->next_index = next1; + t->filter_value = not_a_syn1 ? 0 : *c1; + } + + /* Not TCP? */ + ip2 = vlib_buffer_get_current (b2); + if (ip2->protocol != IP_PROTOCOL_TCP) + goto trace02; + + tcp2 = ip4_next_header (ip2); + /* + * Not a SYN? + * $$$$ hack: the TCP bitfield flags seem not to compile + * correct code. + */ + if (PREDICT_TRUE (!(tcp2->flags & 0x2))) + goto trace02; + + not_a_syn2 = 0; + hash2 = clib_xxhash ((u64) ip2->src_address.as_u32); + c2 = &rt->syn_counts[hash2 & (_vec_len (rt->syn_counts) - 1)]; + if (PREDICT_FALSE (*c2 >= 0x80)) + { + next2 = SYN_FILTER_NEXT_DROP; + b2->error = node->errors[SYN_FILTER_ERROR_THROTTLED]; + goto trace02; + } + *c2 += 1; + ok_syn_packets++; + + trace02: + if (PREDICT_FALSE ((node->flags & VLIB_NODE_FLAG_TRACE) + && (b2->flags & VLIB_BUFFER_IS_TRACED))) + { + syn_filter4_trace_t *t = + vlib_add_trace (vm, node, b2, sizeof (*t)); + t->not_a_syn = not_a_syn2; + t->next_index = next2; + t->filter_value = not_a_syn2 ? 0 : *c2; + } + + /* Not TCP? */ + ip3 = vlib_buffer_get_current (b3); + if (ip3->protocol != IP_PROTOCOL_TCP) + goto trace03; + + tcp3 = ip4_next_header (ip3); + /* + * Not a SYN? + * $$$$ hack: the TCP bitfield flags seem not to compile + * correct code. + */ + if (PREDICT_TRUE (!(tcp3->flags & 0x2))) + goto trace03; + + not_a_syn3 = 0; + hash3 = clib_xxhash ((u64) ip3->src_address.as_u32); + c3 = &rt->syn_counts[hash3 & (_vec_len (rt->syn_counts) - 1)]; + if (PREDICT_FALSE (*c3 >= 0x80)) + { + next3 = SYN_FILTER_NEXT_DROP; + b3->error = node->errors[SYN_FILTER_ERROR_THROTTLED]; + goto trace03; + } + *c3 += 1; + ok_syn_packets++; + + trace03: + if (PREDICT_FALSE ((node->flags & VLIB_NODE_FLAG_TRACE) + && (b3->flags & VLIB_BUFFER_IS_TRACED))) + { + syn_filter4_trace_t *t = + vlib_add_trace (vm, node, b3, sizeof (*t)); + t->not_a_syn = not_a_syn3; + t->next_index = next3; + t->filter_value = not_a_syn3 ? 0 : *c3; + } + vlib_validate_buffer_enqueue_x4 (vm, node, next_index, + to_next, n_left_to_next, + bi0, bi1, bi2, bi3, + next0, next1, next2, next3); + } + + while (n_left_from > 0 && n_left_to_next > 0) + { + u32 bi0; + vlib_buffer_t *b0; + u32 next0; + ip4_header_t *ip0; + tcp_header_t *tcp0; + u32 not_a_syn0 = 1; + u32 hash0; + u8 *c0; + + /* speculatively enqueue b0 to the current next frame */ + bi0 = from[0]; + to_next[0] = bi0; + from += 1; + to_next += 1; + n_left_from -= 1; + n_left_to_next -= 1; + + b0 = vlib_get_buffer (vm, bi0); + + vnet_get_config_data + (&cm->config_main, &b0->current_config_index, + &next0, 0 /* sizeof (c0[0]) */ ); + + /* Not TCP? */ + ip0 = vlib_buffer_get_current (b0); + if (ip0->protocol != IP_PROTOCOL_TCP) + goto trace0; + + tcp0 = ip4_next_header (ip0); + /* + * Not a SYN? + * $$$$ hack: the TCP bitfield flags seem not to compile + * correct code. + */ + if (PREDICT_TRUE (!(tcp0->flags & 0x2))) + goto trace0; + + not_a_syn0 = 0; + hash0 = clib_xxhash ((u64) ip0->src_address.as_u32); + c0 = &rt->syn_counts[hash0 & (_vec_len (rt->syn_counts) - 1)]; + if (PREDICT_FALSE (*c0 >= 0x80)) + { + next0 = SYN_FILTER_NEXT_DROP; + b0->error = node->errors[SYN_FILTER_ERROR_THROTTLED]; + goto trace0; + } + *c0 += 1; + ok_syn_packets++; + + trace0: + + if (PREDICT_FALSE ((node->flags & VLIB_NODE_FLAG_TRACE) + && (b0->flags & VLIB_BUFFER_IS_TRACED))) + { + syn_filter4_trace_t *t = + vlib_add_trace (vm, node, b0, sizeof (*t)); + t->not_a_syn = not_a_syn0; + t->next_index = next0; + t->filter_value = not_a_syn0 ? 0 : *c0; + } + + /* verify speculative enqueue, maybe switch current next frame */ + vlib_validate_buffer_enqueue_x1 (vm, node, next_index, + to_next, n_left_to_next, + bi0, next0); + } + + vlib_put_next_frame (vm, node, next_index, n_left_to_next); + } + + vlib_node_increment_counter (vm, syn_filter4_node.index, + SYN_FILTER_ERROR_OK, ok_syn_packets); + return frame->n_vectors; +} + +/* *INDENT-OFF* */ +VLIB_REGISTER_NODE (syn_filter4_node, static) = +{ + .function = syn_filter4_node_fn, + .name = "syn-filter-4", + .vector_size = sizeof (u32), + .format_trace = format_syn_filter4_trace, + .type = VLIB_NODE_TYPE_INTERNAL, + + .runtime_data_bytes = sizeof (syn_filter4_runtime_t), + .n_errors = ARRAY_LEN(syn_filter4_error_strings), + .error_strings = syn_filter4_error_strings, + + .n_next_nodes = SYN_FILTER_N_NEXT, + + /* edit / add dispositions here */ + .next_nodes = { + [SYN_FILTER_NEXT_DROP] = "error-drop", + }, +}; +/* *INDENT-ON* */ + +VLIB_NODE_FUNCTION_MULTIARCH (syn_filter4_node, syn_filter4_node_fn); + +/* *INDENT-OFF* */ +VNET_FEATURE_INIT (syn_filter_4, static) = +{ + .arc_name = "ip4-local", + .node_name = "syn-filter-4", + .runs_before = VNET_FEATURES("ip4-local-end-of-arc"), +}; +/* *INDENT-ON* */ + +int +syn_filter_enable_disable (u32 sw_if_index, int enable_disable) +{ + vnet_main_t *vnm = vnet_get_main (); + vnet_sw_interface_t *sw; + int rv = 0; + + /* Utterly wrong? */ + if (pool_is_free_index (vnm->interface_main.sw_interfaces, sw_if_index)) + return VNET_API_ERROR_INVALID_SW_IF_INDEX; + + /* Not a physical port? */ + sw = vnet_get_sw_interface (vnm, sw_if_index); + if (sw->type != VNET_SW_INTERFACE_TYPE_HARDWARE) + return VNET_API_ERROR_INVALID_SW_IF_INDEX; + + if (enable_disable) + { + vlib_main_t *vm = vlib_get_main (); + syn_filter4_runtime_t *rt; + + rt = vlib_node_get_runtime_data (vm, syn_filter4_node.index); + vec_validate (rt->syn_counts, 1023); + /* + * Given perfect disperson / optimal hashing results: + * Allow 128k (successful) syns/sec. 1024, buckets each of which + * absorb 128 syns before filtering. Reset table once a second. + * Reality bites, lets try resetting once every 100ms. + */ + rt->reset_interval = 0.1; /* reset interval in seconds */ + } + + rv = vnet_feature_enable_disable ("ip4-local", "syn-filter-4", + sw_if_index, enable_disable, 0, 0); + + return rv; +} + +static clib_error_t * +syn_filter_enable_disable_command_fn (vlib_main_t * vm, + unformat_input_t * input, + vlib_cli_command_t * cmd) +{ + vnet_main_t *vnm = vnet_get_main (); + u32 sw_if_index = ~0; + int enable_disable = 1; + int rv; + + while (unformat_check_input (input) != UNFORMAT_END_OF_INPUT) + { + if (unformat (input, "disable")) + enable_disable = 0; + else if (unformat (input, "%U", unformat_vnet_sw_interface, + vnm, &sw_if_index)) + ; + else + break; + } + + if (sw_if_index == ~0) + return clib_error_return (0, "Please specify an interface..."); + + rv = syn_filter_enable_disable (sw_if_index, enable_disable); + + switch (rv) + { + case 0: + break; + + case VNET_API_ERROR_INVALID_SW_IF_INDEX: + return clib_error_return + (0, "Invalid interface, only works on physical ports"); + break; + + case VNET_API_ERROR_UNIMPLEMENTED: + return clib_error_return (0, + "Device driver doesn't support redirection"); + break; + + case VNET_API_ERROR_INVALID_VALUE: + return clib_error_return (0, "feature arc not found"); + + case VNET_API_ERROR_INVALID_VALUE_2: + return clib_error_return (0, "feature node not found"); + + default: + return clib_error_return (0, "syn_filter_enable_disable returned %d", + rv); + } + return 0; +} + +/* *INDENT-OFF* */ +VLIB_CLI_COMMAND (sr_content_command, static) = +{ + .path = "ip syn filter", + .short_help = "ip syn filter [disable]", + .function = syn_filter_enable_disable_command_fn, +}; +/* *INDENT-ON* */ + +/* + * fd.io coding-style-patch-verification: ON + * + * Local Variables: + * eval: (c-set-style "gnu") + * End: + */ diff --git a/src/vnet/tcp/tcp_timer.h b/src/vnet/tcp/tcp_timer.h new file mode 100644 index 00000000..fa25268c --- /dev/null +++ b/src/vnet/tcp/tcp_timer.h @@ -0,0 +1,29 @@ +/* + * Copyright (c) 2016 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ +#ifndef __included_tcp_timer_h__ +#define __included_tcp_timer_h__ + +#include +#include + +#endif /* __included_tcp_timer_h__ */ + +/* + * fd.io coding-style-patch-verification: ON + * + * Local Variables: + * eval: (c-set-style "gnu") + * End: + */ diff --git a/src/vnet/udp/builtin_server.c b/src/vnet/udp/builtin_server.c new file mode 100644 index 00000000..afa66ba4 --- /dev/null +++ b/src/vnet/udp/builtin_server.c @@ -0,0 +1,239 @@ +/* + * Copyright (c) 2016 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +/** @file + udp builtin server +*/ + +#include +#include +#include + +/** per-worker built-in server copy buffers */ +u8 **copy_buffers; + +static int +builtin_session_create_callback (stream_session_t * s) +{ + /* Simple version: declare session ready-to-go... */ + s->session_state = SESSION_STATE_READY; + return 0; +} + +static void +builtin_session_disconnect_callback (stream_session_t * s) +{ + stream_session_disconnect (s); +} + +static int +builtin_server_rx_callback (stream_session_t * s) +{ + svm_fifo_t *rx_fifo, *tx_fifo; + u32 this_transfer; + int actual_transfer; + u8 *my_copy_buffer; + session_fifo_event_t evt; + unix_shared_memory_queue_t *q; + + my_copy_buffer = copy_buffers[s->thread_index]; + rx_fifo = s->server_rx_fifo; + tx_fifo = s->server_tx_fifo; + + this_transfer = svm_fifo_max_enqueue (tx_fifo) + < svm_fifo_max_dequeue (rx_fifo) ? + svm_fifo_max_enqueue (tx_fifo) : svm_fifo_max_dequeue (rx_fifo); + + vec_validate (my_copy_buffer, this_transfer - 1); + _vec_len (my_copy_buffer) = this_transfer; + + actual_transfer = svm_fifo_dequeue_nowait (rx_fifo, 0, this_transfer, + my_copy_buffer); + ASSERT (actual_transfer == this_transfer); + actual_transfer = svm_fifo_enqueue_nowait (tx_fifo, 0, this_transfer, + my_copy_buffer); + + copy_buffers[s->thread_index] = my_copy_buffer; + + /* Fabricate TX event, send to ourselves */ + evt.fifo = tx_fifo; + evt.event_type = FIFO_EVENT_SERVER_TX; + /* $$$$ for event logging */ + evt.enqueue_length = actual_transfer; + evt.event_id = 0; + q = session_manager_get_vpp_event_queue (s->thread_index); + unix_shared_memory_queue_add (q, (u8 *) & evt, 0 /* do wait for mutex */ ); + + return 0; +} + +/* *INDENT-OFF* */ +static session_cb_vft_t builtin_server = { + .session_accept_callback = builtin_session_create_callback, + .session_disconnect_callback = builtin_session_disconnect_callback, + .builtin_server_rx_callback = builtin_server_rx_callback +}; +/* *INDENT-ON* */ + +static int +bind_builtin_uri_server (u8 * uri) +{ + vnet_bind_args_t _a, *a = &_a; + char segment_name[128]; + u32 segment_name_length; + int rv; + u64 options[16]; + + segment_name_length = ARRAY_LEN (segment_name); + + memset (a, 0, sizeof (*a)); + memset (options, 0, sizeof (options)); + + a->uri = (char *) uri; + a->api_client_index = ~0; /* built-in server */ + a->segment_name = segment_name; + a->segment_name_length = segment_name_length; + a->session_cb_vft = &builtin_server; + + options[SESSION_OPTIONS_ACCEPT_COOKIE] = 0x12345678; + options[SESSION_OPTIONS_SEGMENT_SIZE] = (2 << 30); /*$$$$ config / arg */ + a->options = options; + + rv = vnet_bind_uri (a); + + return rv; +} + +static int +unbind_builtin_uri_server (u8 * uri) +{ + int rv; + + rv = vnet_unbind_uri ((char *) uri, ~0 /* client_index */ ); + + return rv; +} + +static clib_error_t * +builtin_server_init (vlib_main_t * vm) +{ + vlib_thread_main_t *vtm = vlib_get_thread_main (); + u32 num_threads; + + num_threads = 1 /* main thread */ + vtm->n_threads; + + vec_validate (copy_buffers, num_threads - 1); + return 0; +} + +VLIB_INIT_FUNCTION (builtin_server_init); + +static clib_error_t * +builtin_uri_bind_command_fn (vlib_main_t * vm, + unformat_input_t * input, + vlib_cli_command_t * cmd) +{ + u8 *uri = 0; + int rv; + + while (unformat_check_input (input) != UNFORMAT_END_OF_INPUT) + { + if (unformat (input, "uri %s", &uri)) + ; + else + break; + } + + if (uri == 0) + return clib_error_return (0, "uri to bind not specified..."); + + rv = bind_builtin_uri_server (uri); + + vec_free (uri); + + switch (rv) + { + case 0: + break; + + default: + return clib_error_return (0, "bind_uri_server returned %d", rv); + break; + } + + return 0; +} + +/* *INDENT-OFF* */ +VLIB_CLI_COMMAND (builtin_uri_bind_command, static) = +{ + .path = "builtin uri bind", + .short_help = "builtin uri bind", + .function = builtin_uri_bind_command_fn, +}; +/* *INDENT-ON* */ + +static clib_error_t * +builtin_uri_unbind_command_fn (vlib_main_t * vm, + unformat_input_t * input, + vlib_cli_command_t * cmd) +{ + u8 *uri = 0; + int rv; + + while (unformat_check_input (input) != UNFORMAT_END_OF_INPUT) + { + if (unformat (input, "uri %s", &uri)) + ; + else + break; + } + + if (uri == 0) + return clib_error_return (0, "uri to unbind not specified..."); + + rv = unbind_builtin_uri_server (uri); + + vec_free (uri); + + switch (rv) + { + case 0: + break; + + default: + return clib_error_return (0, "unbind_uri_server returned %d", rv); + break; + } + + return 0; +} + +/* *INDENT-OFF* */ +VLIB_CLI_COMMAND (builtin_uri_unbind_command, static) = +{ + .path = "builtin uri unbind", + .short_help = "builtin uri unbind", + .function = builtin_uri_unbind_command_fn, +}; +/* *INDENT-ON* */ + +/* + * fd.io coding-style-patch-verification: ON + * + * Local Variables: + * eval: (c-set-style "gnu") + * End: + */ diff --git a/src/vnet/udp/udp.c b/src/vnet/udp/udp.c new file mode 100644 index 00000000..9e740466 --- /dev/null +++ b/src/vnet/udp/udp.c @@ -0,0 +1,342 @@ +/* + * Copyright (c) 2016 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +/** @file + udp state machine, etc. +*/ + +#include +#include +#include +#include + +udp_uri_main_t udp_uri_main; + +u32 +udp_session_bind_ip4 (vlib_main_t * vm, u32 session_index, + ip46_address_t * ip, u16 port_number_host_byte_order) +{ + udp_uri_main_t *um = vnet_get_udp_main (); + udp_connection_t *listener; + + pool_get (um->udp_listeners, listener); + memset (listener, 0, sizeof (udp_connection_t)); + listener->c_lcl_port = clib_host_to_net_u16 (port_number_host_byte_order); + listener->c_lcl_ip4.as_u32 = ip->ip4.as_u32; + listener->c_proto = SESSION_TYPE_IP4_UDP; + udp_register_dst_port (um->vlib_main, port_number_host_byte_order, + udp4_uri_input_node.index, 1 /* is_ipv4 */ ); + return 0; +} + +u32 +udp_session_bind_ip6 (vlib_main_t * vm, u32 session_index, + ip46_address_t * ip, u16 port_number_host_byte_order) +{ + udp_uri_main_t *um = vnet_get_udp_main (); + udp_connection_t *listener; + + pool_get (um->udp_listeners, listener); + listener->c_lcl_port = clib_host_to_net_u16 (port_number_host_byte_order); + clib_memcpy (&listener->c_lcl_ip6, &ip->ip6, sizeof (ip6_address_t)); + listener->c_proto = SESSION_TYPE_IP6_UDP; + udp_register_dst_port (um->vlib_main, port_number_host_byte_order, + udp4_uri_input_node.index, 0 /* is_ipv4 */ ); + return 0; +} + +u32 +udp_session_unbind_ip4 (vlib_main_t * vm, u32 listener_index) +{ + udp_connection_t *listener; + listener = udp_listener_get (listener_index); + + /* deregister the udp_local mapping */ + udp_unregister_dst_port (vm, listener->c_lcl_port, 1 /* is_ipv4 */ ); + return 0; +} + +u32 +udp_session_unbind_ip6 (vlib_main_t * vm, u32 listener_index) +{ + udp_connection_t *listener; + + listener = udp_listener_get (listener_index); + + /* deregister the udp_local mapping */ + udp_unregister_dst_port (vm, listener->c_lcl_port, 0 /* is_ipv4 */ ); + return 0; +} + +transport_connection_t * +udp_session_get_listener (u32 listener_index) +{ + udp_connection_t *us; + + us = udp_listener_get (listener_index); + return &us->connection; +} + +u32 +udp_push_header (transport_connection_t * tconn, vlib_buffer_t * b) +{ + udp_connection_t *us; + u8 *data; + udp_header_t *udp; + + us = (udp_connection_t *) tconn; + + if (tconn->is_ip4) + { + ip4_header_t *ip; + + data = vlib_buffer_get_current (b); + udp = (udp_header_t *) (data - sizeof (udp_header_t)); + ip = (ip4_header_t *) ((u8 *) udp - sizeof (ip4_header_t)); + + /* Build packet header, swap rx key src + dst fields */ + ip->src_address.as_u32 = us->c_lcl_ip4.as_u32; + ip->dst_address.as_u32 = us->c_rmt_ip4.as_u32; + ip->ip_version_and_header_length = 0x45; + ip->ttl = 254; + ip->protocol = IP_PROTOCOL_UDP; + ip->length = clib_host_to_net_u16 (b->current_length + sizeof (*udp)); + ip->checksum = ip4_header_checksum (ip); + + udp->src_port = us->c_lcl_port; + udp->dst_port = us->c_rmt_port; + udp->length = clib_host_to_net_u16 (b->current_length); + udp->checksum = 0; + + b->current_length = sizeof (*ip) + sizeof (*udp); + return SESSION_QUEUE_NEXT_IP4_LOOKUP; + } + else + { + vlib_main_t *vm = vlib_get_main (); + ip6_header_t *ip; + u16 payload_length; + int bogus = ~0; + + data = vlib_buffer_get_current (b); + udp = (udp_header_t *) (data - sizeof (udp_header_t)); + ip = (ip6_header_t *) ((u8 *) udp - sizeof (ip6_header_t)); + + /* Build packet header, swap rx key src + dst fields */ + clib_memcpy (&ip->src_address, &us->c_lcl_ip6, sizeof (ip6_address_t)); + clib_memcpy (&ip->dst_address, &us->c_rmt_ip6, sizeof (ip6_address_t)); + + ip->ip_version_traffic_class_and_flow_label = + clib_host_to_net_u32 (0x6 << 28); + + ip->hop_limit = 0xff; + ip->protocol = IP_PROTOCOL_UDP; + + payload_length = vlib_buffer_length_in_chain (vm, b); + payload_length -= sizeof (*ip); + + ip->payload_length = clib_host_to_net_u16 (payload_length); + + udp->checksum = ip6_tcp_udp_icmp_compute_checksum (vm, b, ip, &bogus); + ASSERT (!bogus); + + udp->src_port = us->c_lcl_port; + udp->dst_port = us->c_rmt_port; + udp->length = clib_host_to_net_u16 (b->current_length); + udp->checksum = 0; + + b->current_length = sizeof (*ip) + sizeof (*udp); + + return SESSION_QUEUE_NEXT_IP6_LOOKUP; + } +} + +transport_connection_t * +udp_session_get (u32 connection_index, u32 my_thread_index) +{ + udp_uri_main_t *um = vnet_get_udp_main (); + + udp_connection_t *us; + us = + pool_elt_at_index (um->udp_sessions[my_thread_index], connection_index); + return &us->connection; +} + +void +udp_session_close (u32 connection_index, u32 my_thread_index) +{ + udp_uri_main_t *um = vnet_get_udp_main (); + pool_put_index (um->udp_sessions[my_thread_index], connection_index); +} + +u8 * +format_udp_session_ip4 (u8 * s, va_list * args) +{ + u32 uci = va_arg (*args, u32); + u32 thread_index = va_arg (*args, u32); + udp_connection_t *u4; + + u4 = udp_connection_get (uci, thread_index); + + s = format (s, "[%s] %U:%d->%U:%d", "udp", format_ip4_address, + &u4->c_lcl_ip4, clib_net_to_host_u16 (u4->c_lcl_port), + format_ip4_address, &u4->c_rmt_ip4, + clib_net_to_host_u16 (u4->c_rmt_port)); + return s; +} + +u8 * +format_udp_session_ip6 (u8 * s, va_list * args) +{ + u32 uci = va_arg (*args, u32); + u32 thread_index = va_arg (*args, u32); + udp_connection_t *tc = udp_connection_get (uci, thread_index); + s = format (s, "[%s] %U:%d->%U:%d", "udp", format_ip6_address, + &tc->c_lcl_ip6, clib_net_to_host_u16 (tc->c_lcl_port), + format_ip6_address, &tc->c_rmt_ip6, + clib_net_to_host_u16 (tc->c_rmt_port)); + return s; +} + +u8 * +format_udp_listener_session_ip4 (u8 * s, va_list * args) +{ + u32 tci = va_arg (*args, u32); + udp_connection_t *tc = udp_listener_get (tci); + s = format (s, "[%s] %U:%d->%U:%d", "udp", format_ip4_address, + &tc->c_lcl_ip4, clib_net_to_host_u16 (tc->c_lcl_port), + format_ip4_address, &tc->c_rmt_ip4, + clib_net_to_host_u16 (tc->c_rmt_port)); + return s; +} + +u8 * +format_udp_listener_session_ip6 (u8 * s, va_list * args) +{ + u32 tci = va_arg (*args, u32); + udp_connection_t *tc = udp_listener_get (tci); + s = format (s, "[%s] %U:%d->%U:%d", "udp", format_ip6_address, + &tc->c_lcl_ip6, clib_net_to_host_u16 (tc->c_lcl_port), + format_ip6_address, &tc->c_rmt_ip6, + clib_net_to_host_u16 (tc->c_rmt_port)); + return s; +} + +u16 +udp_send_mss_uri (transport_connection_t * t) +{ + /* TODO figure out MTU of output interface */ + return 400; +} + +u32 +udp_send_space_uri (transport_connection_t * t) +{ + /* No constraint on TX window */ + return ~0; +} + +int +udp_open_connection (ip46_address_t * addr, u16 port) +{ + clib_warning ("Not implemented"); + return 0; +} + +/* *INDENT-OFF* */ +const static transport_proto_vft_t udp4_proto = { + .bind = udp_session_bind_ip4, + .open = udp_open_connection, + .unbind = udp_session_unbind_ip4, + .push_header = udp_push_header, + .get_connection = udp_session_get, + .get_listener = udp_session_get_listener, + .close = udp_session_close, + .send_mss = udp_send_mss_uri, + .send_space = udp_send_space_uri, + .format_connection = format_udp_session_ip4, + .format_listener = format_udp_listener_session_ip4 +}; + +const static transport_proto_vft_t udp6_proto = { + .bind = udp_session_bind_ip6, + .open = udp_open_connection, + .unbind = udp_session_unbind_ip6, + .push_header = udp_push_header, + .get_connection = udp_session_get, + .get_listener = udp_session_get_listener, + .close = udp_session_close, + .send_mss = udp_send_mss_uri, + .send_space = udp_send_space_uri, + .format_connection = format_udp_session_ip6, + .format_listener = format_udp_listener_session_ip6 +}; +/* *INDENT-ON* */ + +static clib_error_t * +udp_init (vlib_main_t * vm) +{ + udp_uri_main_t *um = vnet_get_udp_main (); + ip_main_t *im = &ip_main; + vlib_thread_main_t *tm = vlib_get_thread_main (); + u32 num_threads; + clib_error_t *error = 0; + ip_protocol_info_t *pi; + + um->vlib_main = vm; + um->vnet_main = vnet_get_main (); + + if ((error = vlib_call_init_function (vm, ip_main_init))) + return error; + if ((error = vlib_call_init_function (vm, ip4_lookup_init))) + return error; + if ((error = vlib_call_init_function (vm, ip6_lookup_init))) + return error; + + /* + * Registrations + */ + + /* IP registration */ + pi = ip_get_protocol_info (im, IP_PROTOCOL_UDP); + if (pi == 0) + return clib_error_return (0, "UDP protocol info AWOL"); + pi->format_header = format_udp_header; + pi->unformat_pg_edit = unformat_pg_udp_header; + + + /* Register as transport with URI */ + session_register_transport (SESSION_TYPE_IP4_UDP, &udp4_proto); + session_register_transport (SESSION_TYPE_IP6_UDP, &udp6_proto); + + /* + * Initialize data structures + */ + + num_threads = 1 /* main thread */ + tm->n_threads; + vec_validate (um->udp_sessions, num_threads - 1); + + return error; +} + +VLIB_INIT_FUNCTION (udp_init); + +/* + * fd.io coding-style-patch-verification: ON + * + * Local Variables: + * eval: (c-set-style "gnu") + * End: + */ diff --git a/src/vnet/udp/udp.h b/src/vnet/udp/udp.h new file mode 100644 index 00000000..7ab26ce9 --- /dev/null +++ b/src/vnet/udp/udp.h @@ -0,0 +1,362 @@ +/* + * Copyright (c) 2017 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ +#ifndef __included_udp_h__ +#define __included_udp_h__ + +#include +#include +#include +#include +#include +#include +#include + +#include +#include + +typedef struct +{ + transport_connection_t connection; /** must be first */ + + /** ersatz MTU to limit fifo pushes to test data size */ + u32 mtu; +} udp_connection_t; + +typedef struct _udp_uri_main +{ + /* Per-worker thread udp connection pools */ + udp_connection_t **udp_sessions; + udp_connection_t *udp_listeners; + + /* convenience */ + vlib_main_t *vlib_main; + vnet_main_t *vnet_main; + ip4_main_t *ip4_main; + ip6_main_t *ip6_main; +} udp_uri_main_t; + +extern udp_uri_main_t udp_uri_main; +extern vlib_node_registration_t udp4_uri_input_node; + +always_inline udp_uri_main_t * +vnet_get_udp_main () +{ + return &udp_uri_main; +} + +always_inline udp_connection_t * +udp_connection_get (u32 conn_index, u32 thread_index) +{ + return pool_elt_at_index (udp_uri_main.udp_sessions[thread_index], + conn_index); +} + +always_inline udp_connection_t * +udp_listener_get (u32 conn_index) +{ + return pool_elt_at_index (udp_uri_main.udp_listeners, conn_index); +} + +typedef enum +{ +#define udp_error(n,s) UDP_ERROR_##n, +#include +#undef udp_error + UDP_N_ERROR, +} udp_error_t; + +#define foreach_udp4_dst_port \ +_ (67, dhcp_to_server) \ +_ (68, dhcp_to_client) \ +_ (500, ikev2) \ +_ (3784, bfd4) \ +_ (3785, bfd_echo4) \ +_ (4341, lisp_gpe) \ +_ (4342, lisp_cp) \ +_ (4739, ipfix) \ +_ (4789, vxlan) \ +_ (4789, vxlan6) \ +_ (4790, vxlan_gpe) \ +_ (6633, vpath_3) + + +#define foreach_udp6_dst_port \ +_ (547, dhcpv6_to_server) \ +_ (546, dhcpv6_to_client) \ +_ (3784, bfd6) \ +_ (3785, bfd_echo6) \ +_ (4341, lisp_gpe6) \ +_ (4342, lisp_cp6) \ +_ (4790, vxlan6_gpe) \ +_ (6633, vpath6_3) + +typedef enum +{ +#define _(n,f) UDP_DST_PORT_##f = n, + foreach_udp4_dst_port foreach_udp6_dst_port +#undef _ +} udp_dst_port_t; + +typedef enum +{ +#define _(n,f) UDP6_DST_PORT_##f = n, + foreach_udp6_dst_port +#undef _ +} udp6_dst_port_t; + +typedef struct +{ + /* Name (a c string). */ + char *name; + + /* GRE protocol type in host byte order. */ + udp_dst_port_t dst_port; + + /* Node which handles this type. */ + u32 node_index; + + /* Next index for this type. */ + u32 next_index; +} udp_dst_port_info_t; + +typedef enum +{ + UDP_IP6 = 0, + UDP_IP4, /* the code is full of is_ip4... */ + N_UDP_AF, +} udp_af_t; + +typedef struct +{ + udp_dst_port_info_t *dst_port_infos[N_UDP_AF]; + + /* Hash tables mapping name/protocol to protocol info index. */ + uword *dst_port_info_by_name[N_UDP_AF]; + uword *dst_port_info_by_dst_port[N_UDP_AF]; + + /* convenience */ + vlib_main_t *vlib_main; +} udp_main_t; + +always_inline udp_dst_port_info_t * +udp_get_dst_port_info (udp_main_t * um, udp_dst_port_t dst_port, u8 is_ip4) +{ + uword *p = hash_get (um->dst_port_info_by_dst_port[is_ip4], dst_port); + return p ? vec_elt_at_index (um->dst_port_infos[is_ip4], p[0]) : 0; +} + +format_function_t format_udp_header; +format_function_t format_udp_rx_trace; + +unformat_function_t unformat_udp_header; + +void udp_register_dst_port (vlib_main_t * vm, + udp_dst_port_t dst_port, + u32 node_index, u8 is_ip4); + +void +udp_unregister_dst_port (vlib_main_t * vm, + udp_dst_port_t dst_port, u8 is_ip4); + +void udp_punt_unknown (vlib_main_t * vm, u8 is_ip4, u8 is_add); + +always_inline void +ip_udp_fixup_one (vlib_main_t * vm, vlib_buffer_t * b0, u8 is_ip4) +{ + u16 new_l0; + udp_header_t *udp0; + + if (is_ip4) + { + ip4_header_t *ip0; + ip_csum_t sum0; + u16 old_l0 = 0; + + ip0 = vlib_buffer_get_current (b0); + + /* fix the ing outer-IP checksum */ + sum0 = ip0->checksum; + /* old_l0 always 0, see the rewrite setup */ + new_l0 = clib_host_to_net_u16 (vlib_buffer_length_in_chain (vm, b0)); + + sum0 = ip_csum_update (sum0, old_l0, new_l0, ip4_header_t, + length /* changed member */ ); + ip0->checksum = ip_csum_fold (sum0); + ip0->length = new_l0; + + /* Fix UDP length */ + udp0 = (udp_header_t *) (ip0 + 1); + new_l0 = clib_host_to_net_u16 (vlib_buffer_length_in_chain (vm, b0) + - sizeof (*ip0)); + udp0->length = new_l0; + } + else + { + ip6_header_t *ip0; + int bogus0; + + ip0 = vlib_buffer_get_current (b0); + + new_l0 = clib_host_to_net_u16 (vlib_buffer_length_in_chain (vm, b0) + - sizeof (*ip0)); + ip0->payload_length = new_l0; + + /* Fix UDP length */ + udp0 = (udp_header_t *) (ip0 + 1); + udp0->length = new_l0; + + udp0->checksum = + ip6_tcp_udp_icmp_compute_checksum (vm, b0, ip0, &bogus0); + ASSERT (bogus0 == 0); + + if (udp0->checksum == 0) + udp0->checksum = 0xffff; + } +} + +always_inline void +ip_udp_encap_one (vlib_main_t * vm, vlib_buffer_t * b0, u8 * ec0, word ec_len, + u8 is_ip4) +{ + vlib_buffer_advance (b0, -ec_len); + + if (is_ip4) + { + ip4_header_t *ip0; + + ip0 = vlib_buffer_get_current (b0); + + /* Apply the encap string. */ + clib_memcpy (ip0, ec0, ec_len); + ip_udp_fixup_one (vm, b0, 1); + } + else + { + ip6_header_t *ip0; + + ip0 = vlib_buffer_get_current (b0); + + /* Apply the encap string. */ + clib_memcpy (ip0, ec0, ec_len); + ip_udp_fixup_one (vm, b0, 0); + } +} + +always_inline void +ip_udp_encap_two (vlib_main_t * vm, vlib_buffer_t * b0, vlib_buffer_t * b1, + u8 * ec0, u8 * ec1, word ec_len, u8 is_v4) +{ + u16 new_l0, new_l1; + udp_header_t *udp0, *udp1; + + ASSERT (_vec_len (ec0) == _vec_len (ec1)); + + vlib_buffer_advance (b0, -ec_len); + vlib_buffer_advance (b1, -ec_len); + + if (is_v4) + { + ip4_header_t *ip0, *ip1; + ip_csum_t sum0, sum1; + u16 old_l0 = 0, old_l1 = 0; + + ip0 = vlib_buffer_get_current (b0); + ip1 = vlib_buffer_get_current (b1); + + /* Apply the encap string */ + clib_memcpy (ip0, ec0, ec_len); + clib_memcpy (ip1, ec1, ec_len); + + /* fix the ing outer-IP checksum */ + sum0 = ip0->checksum; + sum1 = ip1->checksum; + + /* old_l0 always 0, see the rewrite setup */ + new_l0 = clib_host_to_net_u16 (vlib_buffer_length_in_chain (vm, b0)); + new_l1 = clib_host_to_net_u16 (vlib_buffer_length_in_chain (vm, b1)); + + sum0 = ip_csum_update (sum0, old_l0, new_l0, ip4_header_t, + length /* changed member */ ); + sum1 = ip_csum_update (sum1, old_l1, new_l1, ip4_header_t, + length /* changed member */ ); + + ip0->checksum = ip_csum_fold (sum0); + ip1->checksum = ip_csum_fold (sum1); + + ip0->length = new_l0; + ip1->length = new_l1; + + /* Fix UDP length */ + udp0 = (udp_header_t *) (ip0 + 1); + udp1 = (udp_header_t *) (ip1 + 1); + + new_l0 = + clib_host_to_net_u16 (vlib_buffer_length_in_chain (vm, b0) - + sizeof (*ip0)); + new_l1 = + clib_host_to_net_u16 (vlib_buffer_length_in_chain (vm, b1) - + sizeof (*ip1)); + udp0->length = new_l0; + udp1->length = new_l1; + } + else + { + ip6_header_t *ip0, *ip1; + int bogus0, bogus1; + + ip0 = vlib_buffer_get_current (b0); + ip1 = vlib_buffer_get_current (b1); + + /* Apply the encap string. */ + clib_memcpy (ip0, ec0, ec_len); + clib_memcpy (ip1, ec1, ec_len); + + new_l0 = clib_host_to_net_u16 (vlib_buffer_length_in_chain (vm, b0) + - sizeof (*ip0)); + new_l1 = clib_host_to_net_u16 (vlib_buffer_length_in_chain (vm, b1) + - sizeof (*ip1)); + ip0->payload_length = new_l0; + ip1->payload_length = new_l1; + + /* Fix UDP length */ + udp0 = (udp_header_t *) (ip0 + 1); + udp1 = (udp_header_t *) (ip1 + 1); + + udp0->length = new_l0; + udp1->length = new_l1; + + udp0->checksum = + ip6_tcp_udp_icmp_compute_checksum (vm, b0, ip0, &bogus0); + udp1->checksum = + ip6_tcp_udp_icmp_compute_checksum (vm, b1, ip1, &bogus1); + ASSERT (bogus0 == 0); + ASSERT (bogus1 == 0); + + if (udp0->checksum == 0) + udp0->checksum = 0xffff; + if (udp1->checksum == 0) + udp1->checksum = 0xffff; + } +} + +/* + * fd.io coding-style-patch-verification: ON + * + * Local Variables: + * eval: (c-set-style "gnu") + * End: + */ + +#endif /* __included_udp_h__ */ diff --git a/src/vnet/udp/udp_error.def b/src/vnet/udp/udp_error.def new file mode 100644 index 00000000..bfdae0ac --- /dev/null +++ b/src/vnet/udp/udp_error.def @@ -0,0 +1,21 @@ +/* + * udp_error.def: udp errors + * + * Copyright (c) 2013-2016 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +udp_error (NONE, "no error") +udp_error (NO_LISTENER, "no listener for dst port") +udp_error (LENGTH_ERROR, "UDP packets with length errors") +udp_error (PUNT, "no listener punt") diff --git a/src/vnet/udp/udp_format.c b/src/vnet/udp/udp_format.c new file mode 100644 index 00000000..abdf561e --- /dev/null +++ b/src/vnet/udp/udp_format.c @@ -0,0 +1,91 @@ +/* + * Copyright (c) 2015 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ +/* + * ip/udp_format.c: udp formatting + * + * Copyright (c) 2008 Eliot Dresselhaus + * + * Permission is hereby granted, free of charge, to any person obtaining + * a copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sublicense, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice shall be + * included in all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND + * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE + * LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + * OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + * WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + */ + +#include + +/* Format UDP header. */ +u8 * +format_udp_header (u8 * s, va_list * args) +{ + udp_header_t *udp = va_arg (*args, udp_header_t *); + u32 max_header_bytes = va_arg (*args, u32); + uword indent; + u32 header_bytes = sizeof (udp[0]); + + /* Nothing to do. */ + if (max_header_bytes < sizeof (udp[0])) + return format (s, "UDP header truncated"); + + indent = format_get_indent (s); + indent += 2; + + s = format (s, "UDP: %d -> %d", + clib_net_to_host_u16 (udp->src_port), + clib_net_to_host_u16 (udp->dst_port)); + + s = format (s, "\n%Ulength %d, checksum 0x%04x", + format_white_space, indent, + clib_net_to_host_u16 (udp->length), + clib_net_to_host_u16 (udp->checksum)); + + /* Recurse into next protocol layer. */ + if (max_header_bytes != 0 && header_bytes < max_header_bytes) + { + ip_main_t *im = &ip_main; + tcp_udp_port_info_t *pi; + + pi = ip_get_tcp_udp_port_info (im, udp->dst_port); + + if (pi && pi->format_header) + s = format (s, "\n%U%U", + format_white_space, indent - 2, pi->format_header, + /* next protocol header */ (udp + 1), + max_header_bytes - sizeof (udp[0])); + } + + return s; +} + +/* + * fd.io coding-style-patch-verification: ON + * + * Local Variables: + * eval: (c-set-style "gnu") + * End: + */ diff --git a/src/vnet/udp/udp_input.c b/src/vnet/udp/udp_input.c new file mode 100644 index 00000000..4d509335 --- /dev/null +++ b/src/vnet/udp/udp_input.c @@ -0,0 +1,314 @@ +/* + * Copyright (c) 2016 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +#include +#include +#include +#include + +#include +#include +#include +#include + +#include + +#include +#include "../session/application_interface.h" + +vlib_node_registration_t udp4_uri_input_node; + +typedef struct +{ + u32 session; + u32 disposition; + u32 thread_index; +} udp4_uri_input_trace_t; + +/* packet trace format function */ +static u8 * +format_udp4_uri_input_trace (u8 * s, va_list * args) +{ + CLIB_UNUSED (vlib_main_t * vm) = va_arg (*args, vlib_main_t *); + CLIB_UNUSED (vlib_node_t * node) = va_arg (*args, vlib_node_t *); + udp4_uri_input_trace_t *t = va_arg (*args, udp4_uri_input_trace_t *); + + s = format (s, "UDP4_URI_INPUT: session %d, disposition %d, thread %d", + t->session, t->disposition, t->thread_index); + return s; +} + +typedef enum +{ + UDP4_URI_INPUT_NEXT_DROP, + UDP4_URI_INPUT_N_NEXT, +} udp4_uri_input_next_t; + +static char *udp4_uri_input_error_strings[] = { +#define _(sym,string) string, + foreach_session_input_error +#undef _ +}; + +static uword +udp4_uri_input_node_fn (vlib_main_t * vm, + vlib_node_runtime_t * node, vlib_frame_t * frame) +{ + u32 n_left_from, *from, *to_next; + udp4_uri_input_next_t next_index; + udp_uri_main_t *um = vnet_get_udp_main (); + session_manager_main_t *smm = vnet_get_session_manager_main (); + u32 my_thread_index = vm->cpu_index; + u8 my_enqueue_epoch; + u32 *session_indices_to_enqueue; + static u32 serial_number; + int i; + + my_enqueue_epoch = ++smm->current_enqueue_epoch[my_thread_index]; + + from = vlib_frame_vector_args (frame); + n_left_from = frame->n_vectors; + next_index = node->cached_next_index; + + while (n_left_from > 0) + { + u32 n_left_to_next; + + vlib_get_next_frame (vm, node, next_index, to_next, n_left_to_next); + + while (n_left_from > 0 && n_left_to_next > 0) + { + u32 bi0; + vlib_buffer_t *b0; + u32 next0 = UDP4_URI_INPUT_NEXT_DROP; + u32 error0 = SESSION_ERROR_ENQUEUED; + udp_header_t *udp0; + ip4_header_t *ip0; + stream_session_t *s0; + svm_fifo_t *f0; + u16 udp_len0; + u8 *data0; + + /* speculatively enqueue b0 to the current next frame */ + bi0 = from[0]; + to_next[0] = bi0; + from += 1; + to_next += 1; + n_left_from -= 1; + n_left_to_next -= 1; + + b0 = vlib_get_buffer (vm, bi0); + + /* udp_local hands us a pointer to the udp data */ + + data0 = vlib_buffer_get_current (b0); + udp0 = (udp_header_t *) (data0 - sizeof (*udp0)); + + /* $$$$ fixme: udp_local doesn't do ip options correctly anyhow */ + ip0 = (ip4_header_t *) (((u8 *) udp0) - sizeof (*ip0)); + s0 = 0; + + /* lookup session */ + s0 = stream_session_lookup4 (&ip0->dst_address, &ip0->src_address, + udp0->dst_port, udp0->src_port, + SESSION_TYPE_IP4_UDP, my_thread_index); + + /* no listener */ + if (PREDICT_FALSE (s0 == 0)) + { + error0 = SESSION_ERROR_NO_LISTENER; + goto trace0; + } + + f0 = s0->server_rx_fifo; + + /* established hit */ + if (PREDICT_TRUE (s0->session_state == SESSION_STATE_READY)) + { + udp_len0 = clib_net_to_host_u16 (udp0->length); + + if (PREDICT_FALSE (udp_len0 > svm_fifo_max_enqueue (f0))) + { + error0 = SESSION_ERROR_FIFO_FULL; + goto trace0; + } + + svm_fifo_enqueue_nowait (f0, 0 /* pid */ , + udp_len0 - sizeof (*udp0), + (u8 *) (udp0 + 1)); + + b0->error = node->errors[SESSION_ERROR_ENQUEUED]; + + /* We need to send an RX event on this fifo */ + if (s0->enqueue_epoch != my_enqueue_epoch) + { + s0->enqueue_epoch = my_enqueue_epoch; + + vec_add1 (smm->session_indices_to_enqueue_by_thread + [my_thread_index], + s0 - smm->sessions[my_thread_index]); + } + } + /* listener hit */ + else if (s0->session_state == SESSION_STATE_LISTENING) + { + udp_connection_t *us; + int rv; + + error0 = SESSION_ERROR_NOT_READY; + + /* + * create udp transport session + */ + pool_get (um->udp_sessions[my_thread_index], us); + + us->mtu = 1024; /* $$$$ policy */ + + us->c_lcl_ip4.as_u32 = ip0->dst_address.as_u32; + us->c_rmt_ip4.as_u32 = ip0->src_address.as_u32; + us->c_lcl_port = udp0->dst_port; + us->c_rmt_port = udp0->src_port; + us->c_proto = SESSION_TYPE_IP4_UDP; + us->c_c_index = us - um->udp_sessions[my_thread_index]; + + /* + * create stream session and attach the udp session to it + */ + rv = stream_session_accept (&us->connection, s0->session_index, + SESSION_TYPE_IP4_UDP, + 1 /*notify */ ); + if (rv) + error0 = rv; + + } + else + { + + error0 = SESSION_ERROR_NOT_READY; + goto trace0; + } + + trace0: + b0->error = node->errors[error0]; + + if (PREDICT_FALSE ((node->flags & VLIB_NODE_FLAG_TRACE) + && (b0->flags & VLIB_BUFFER_IS_TRACED))) + { + udp4_uri_input_trace_t *t = + vlib_add_trace (vm, node, b0, sizeof (*t)); + + t->session = ~0; + if (s0) + t->session = s0 - smm->sessions[my_thread_index]; + t->disposition = error0; + t->thread_index = my_thread_index; + } + + /* verify speculative enqueue, maybe switch current next frame */ + vlib_validate_buffer_enqueue_x1 (vm, node, next_index, + to_next, n_left_to_next, + bi0, next0); + } + + vlib_put_next_frame (vm, node, next_index, n_left_to_next); + } + + /* Send enqueue events */ + + session_indices_to_enqueue = + smm->session_indices_to_enqueue_by_thread[my_thread_index]; + + for (i = 0; i < vec_len (session_indices_to_enqueue); i++) + { + session_fifo_event_t evt; + unix_shared_memory_queue_t *q; + stream_session_t *s0; + application_t *server0; + + /* Get session */ + s0 = pool_elt_at_index (smm->sessions[my_thread_index], + session_indices_to_enqueue[i]); + + /* Get session's server */ + server0 = application_get (s0->app_index); + + /* Built-in server? Deliver the goods... */ + if (server0->cb_fns.builtin_server_rx_callback) + { + server0->cb_fns.builtin_server_rx_callback (s0); + continue; + } + + /* Fabricate event */ + evt.fifo = s0->server_rx_fifo; + evt.event_type = FIFO_EVENT_SERVER_RX; + evt.event_id = serial_number++; + evt.enqueue_length = svm_fifo_max_dequeue (s0->server_rx_fifo); + + /* Add event to server's event queue */ + q = server0->event_queue; + + /* Don't block for lack of space */ + if (PREDICT_TRUE (q->cursize < q->maxsize)) + unix_shared_memory_queue_add (server0->event_queue, (u8 *) & evt, + 0 /* do wait for mutex */ ); + else + { + vlib_node_increment_counter (vm, udp4_uri_input_node.index, + SESSION_ERROR_FIFO_FULL, 1); + } + if (1) + { + ELOG_TYPE_DECLARE (e) = + { + .format = "evt-enqueue: id %d length %d",.format_args = "i4i4",}; + struct + { + u32 data[2]; + } *ed; + ed = ELOG_DATA (&vlib_global_main.elog_main, e); + ed->data[0] = evt.event_id; + ed->data[1] = evt.enqueue_length; + } + } + + vec_reset_length (session_indices_to_enqueue); + + smm->session_indices_to_enqueue_by_thread[my_thread_index] = + session_indices_to_enqueue; + + return frame->n_vectors; +} + +VLIB_REGISTER_NODE (udp4_uri_input_node) = +{ + .function = udp4_uri_input_node_fn,.name = "udp4-uri-input",.vector_size = + sizeof (u32),.format_trace = format_udp4_uri_input_trace,.type = + VLIB_NODE_TYPE_INTERNAL,.n_errors = + ARRAY_LEN (udp4_uri_input_error_strings),.error_strings = + udp4_uri_input_error_strings,.n_next_nodes = UDP4_URI_INPUT_N_NEXT, + /* edit / add dispositions here */ + .next_nodes = + { + [UDP4_URI_INPUT_NEXT_DROP] = "error-drop",} +,}; + +/* + * fd.io coding-style-patch-verification: ON + * + * Local Variables: + * eval: (c-set-style "gnu") + * End: + */ diff --git a/src/vnet/udp/udp_local.c b/src/vnet/udp/udp_local.c new file mode 100644 index 00000000..6b239f73 --- /dev/null +++ b/src/vnet/udp/udp_local.c @@ -0,0 +1,666 @@ +/* + * node.c: udp packet processing + * + * Copyright (c) 2013 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +#include +#include +#include +#include +#include + +udp_main_t udp_main; + +#define foreach_udp_input_next \ + _ (PUNT, "error-punt") \ + _ (DROP, "error-drop") \ + _ (ICMP4_ERROR, "ip4-icmp-error") \ + _ (ICMP6_ERROR, "ip6-icmp-error") + +typedef enum +{ +#define _(s,n) UDP_INPUT_NEXT_##s, + foreach_udp_input_next +#undef _ + UDP_INPUT_N_NEXT, +} udp_input_next_t; + +typedef struct +{ + u16 src_port; + u16 dst_port; + u8 bound; +} udp_rx_trace_t; + +u8 * +format_udp_rx_trace (u8 * s, va_list * args) +{ + CLIB_UNUSED (vlib_main_t * vm) = va_arg (*args, vlib_main_t *); + CLIB_UNUSED (vlib_node_t * node) = va_arg (*args, vlib_node_t *); + udp_rx_trace_t *t = va_arg (*args, udp_rx_trace_t *); + + s = format (s, "UDP: src-port %d dst-port %d%s", + clib_net_to_host_u16 (t->src_port), + clib_net_to_host_u16 (t->dst_port), + t->bound ? "" : " (no listener)"); + return s; +} + +typedef struct +{ + /* Sparse vector mapping udp dst_port in network byte order + to next index. */ + u16 *next_by_dst_port; + u8 punt_unknown; +} udp_input_runtime_t; + +vlib_node_registration_t udp4_input_node; +vlib_node_registration_t udp6_input_node; + +always_inline uword +udp46_input_inline (vlib_main_t * vm, + vlib_node_runtime_t * node, + vlib_frame_t * from_frame, int is_ip4) +{ + udp_input_runtime_t *rt = is_ip4 ? + (void *) vlib_node_get_runtime_data (vm, udp4_input_node.index) + : (void *) vlib_node_get_runtime_data (vm, udp6_input_node.index); + __attribute__ ((unused)) u32 n_left_from, next_index, *from, *to_next; + word n_no_listener = 0; + u8 punt_unknown = rt->punt_unknown; + + from = vlib_frame_vector_args (from_frame); + n_left_from = from_frame->n_vectors; + + next_index = node->cached_next_index; + + while (n_left_from > 0) + { + u32 n_left_to_next; + + vlib_get_next_frame (vm, node, next_index, to_next, n_left_to_next); + + while (n_left_from >= 4 && n_left_to_next >= 2) + { + u32 bi0, bi1; + vlib_buffer_t *b0, *b1; + udp_header_t *h0 = 0, *h1 = 0; + u32 i0, i1, dst_port0, dst_port1; + u32 advance0, advance1; + u32 error0, next0, error1, next1; + + /* Prefetch next iteration. */ + { + vlib_buffer_t *p2, *p3; + + p2 = vlib_get_buffer (vm, from[2]); + p3 = vlib_get_buffer (vm, from[3]); + + vlib_prefetch_buffer_header (p2, LOAD); + vlib_prefetch_buffer_header (p3, LOAD); + + CLIB_PREFETCH (p2->data, sizeof (h0[0]), LOAD); + CLIB_PREFETCH (p3->data, sizeof (h1[0]), LOAD); + } + + bi0 = from[0]; + bi1 = from[1]; + to_next[0] = bi0; + to_next[1] = bi1; + from += 2; + to_next += 2; + n_left_to_next -= 2; + n_left_from -= 2; + + b0 = vlib_get_buffer (vm, bi0); + b1 = vlib_get_buffer (vm, bi1); + + /* ip4/6_local hands us the ip header, not the udp header */ + if (is_ip4) + { + advance0 = sizeof (ip4_header_t); + advance1 = sizeof (ip4_header_t); + } + else + { + advance0 = sizeof (ip6_header_t); + advance1 = sizeof (ip6_header_t); + } + + if (PREDICT_FALSE (b0->current_length < advance0 + sizeof (*h0))) + { + error0 = UDP_ERROR_LENGTH_ERROR; + next0 = UDP_INPUT_NEXT_DROP; + } + else + { + vlib_buffer_advance (b0, advance0); + h0 = vlib_buffer_get_current (b0); + error0 = next0 = 0; + if (PREDICT_FALSE (clib_net_to_host_u16 (h0->length) > + vlib_buffer_length_in_chain (vm, b0))) + { + error0 = UDP_ERROR_LENGTH_ERROR; + next0 = UDP_INPUT_NEXT_DROP; + } + } + + if (PREDICT_FALSE (b1->current_length < advance1 + sizeof (*h1))) + { + error1 = UDP_ERROR_LENGTH_ERROR; + next1 = UDP_INPUT_NEXT_DROP; + } + else + { + vlib_buffer_advance (b1, advance1); + h1 = vlib_buffer_get_current (b1); + error1 = next1 = 0; + if (PREDICT_FALSE (clib_net_to_host_u16 (h1->length) > + vlib_buffer_length_in_chain (vm, b1))) + { + error1 = UDP_ERROR_LENGTH_ERROR; + next1 = UDP_INPUT_NEXT_DROP; + } + } + + /* Index sparse array with network byte order. */ + dst_port0 = (error0 == 0) ? h0->dst_port : 0; + dst_port1 = (error1 == 0) ? h1->dst_port : 0; + sparse_vec_index2 (rt->next_by_dst_port, dst_port0, dst_port1, + &i0, &i1); + next0 = (error0 == 0) ? vec_elt (rt->next_by_dst_port, i0) : next0; + next1 = (error1 == 0) ? vec_elt (rt->next_by_dst_port, i1) : next1; + + if (PREDICT_FALSE (i0 == SPARSE_VEC_INVALID_INDEX)) + { + // move the pointer back so icmp-error can find the + // ip packet header + vlib_buffer_advance (b0, -(word) advance0); + + if (PREDICT_FALSE (punt_unknown)) + { + b0->error = node->errors[UDP_ERROR_PUNT]; + next0 = UDP_INPUT_NEXT_PUNT; + } + else if (is_ip4) + { + icmp4_error_set_vnet_buffer (b0, + ICMP4_destination_unreachable, + ICMP4_destination_unreachable_port_unreachable, + 0); + next0 = UDP_INPUT_NEXT_ICMP4_ERROR; + n_no_listener++; + } + else + { + icmp6_error_set_vnet_buffer (b0, + ICMP6_destination_unreachable, + ICMP6_destination_unreachable_port_unreachable, + 0); + next0 = UDP_INPUT_NEXT_ICMP6_ERROR; + n_no_listener++; + } + } + else + { + b0->error = node->errors[UDP_ERROR_NONE]; + // advance to the payload + vlib_buffer_advance (b0, sizeof (*h0)); + } + + if (PREDICT_FALSE (i1 == SPARSE_VEC_INVALID_INDEX)) + { + // move the pointer back so icmp-error can find the + // ip packet header + vlib_buffer_advance (b1, -(word) advance1); + + if (PREDICT_FALSE (punt_unknown)) + { + b1->error = node->errors[UDP_ERROR_PUNT]; + next1 = UDP_INPUT_NEXT_PUNT; + } + else if (is_ip4) + { + icmp4_error_set_vnet_buffer (b1, + ICMP4_destination_unreachable, + ICMP4_destination_unreachable_port_unreachable, + 0); + next1 = UDP_INPUT_NEXT_ICMP4_ERROR; + n_no_listener++; + } + else + { + icmp6_error_set_vnet_buffer (b1, + ICMP6_destination_unreachable, + ICMP6_destination_unreachable_port_unreachable, + 0); + next1 = UDP_INPUT_NEXT_ICMP6_ERROR; + n_no_listener++; + } + } + else + { + b1->error = node->errors[UDP_ERROR_NONE]; + // advance to the payload + vlib_buffer_advance (b1, sizeof (*h1)); + } + + if (PREDICT_FALSE (b0->flags & VLIB_BUFFER_IS_TRACED)) + { + udp_rx_trace_t *tr = vlib_add_trace (vm, node, + b0, sizeof (*tr)); + if (b0->error != node->errors[UDP_ERROR_LENGTH_ERROR]) + { + tr->src_port = h0 ? h0->src_port : 0; + tr->dst_port = h0 ? h0->dst_port : 0; + tr->bound = (next0 != UDP_INPUT_NEXT_ICMP4_ERROR && + next0 != UDP_INPUT_NEXT_ICMP6_ERROR); + } + } + if (PREDICT_FALSE (b1->flags & VLIB_BUFFER_IS_TRACED)) + { + udp_rx_trace_t *tr = vlib_add_trace (vm, node, + b1, sizeof (*tr)); + if (b1->error != node->errors[UDP_ERROR_LENGTH_ERROR]) + { + tr->src_port = h1 ? h1->src_port : 0; + tr->dst_port = h1 ? h1->dst_port : 0; + tr->bound = (next1 != UDP_INPUT_NEXT_ICMP4_ERROR && + next1 != UDP_INPUT_NEXT_ICMP6_ERROR); + } + } + + vlib_validate_buffer_enqueue_x2 (vm, node, next_index, + to_next, n_left_to_next, + bi0, bi1, next0, next1); + } + + while (n_left_from > 0 && n_left_to_next > 0) + { + u32 bi0; + vlib_buffer_t *b0; + udp_header_t *h0 = 0; + u32 i0, next0; + u32 advance0; + + bi0 = from[0]; + to_next[0] = bi0; + from += 1; + to_next += 1; + n_left_from -= 1; + n_left_to_next -= 1; + + b0 = vlib_get_buffer (vm, bi0); + + /* ip4/6_local hands us the ip header, not the udp header */ + if (is_ip4) + advance0 = sizeof (ip4_header_t); + else + advance0 = sizeof (ip6_header_t); + + if (PREDICT_FALSE (b0->current_length < advance0 + sizeof (*h0))) + { + b0->error = node->errors[UDP_ERROR_LENGTH_ERROR]; + next0 = UDP_INPUT_NEXT_DROP; + goto trace_x1; + } + + vlib_buffer_advance (b0, advance0); + + h0 = vlib_buffer_get_current (b0); + + if (PREDICT_TRUE (clib_net_to_host_u16 (h0->length) <= + vlib_buffer_length_in_chain (vm, b0))) + { + i0 = sparse_vec_index (rt->next_by_dst_port, h0->dst_port); + next0 = vec_elt (rt->next_by_dst_port, i0); + + if (PREDICT_FALSE (i0 == SPARSE_VEC_INVALID_INDEX)) + { + // move the pointer back so icmp-error can find the + // ip packet header + vlib_buffer_advance (b0, -(word) advance0); + + if (PREDICT_FALSE (punt_unknown)) + { + b0->error = node->errors[UDP_ERROR_PUNT]; + next0 = UDP_INPUT_NEXT_PUNT; + } + else if (is_ip4) + { + icmp4_error_set_vnet_buffer (b0, + ICMP4_destination_unreachable, + ICMP4_destination_unreachable_port_unreachable, + 0); + next0 = UDP_INPUT_NEXT_ICMP4_ERROR; + n_no_listener++; + } + else + { + icmp6_error_set_vnet_buffer (b0, + ICMP6_destination_unreachable, + ICMP6_destination_unreachable_port_unreachable, + 0); + next0 = UDP_INPUT_NEXT_ICMP6_ERROR; + n_no_listener++; + } + } + else + { + b0->error = node->errors[UDP_ERROR_NONE]; + // advance to the payload + vlib_buffer_advance (b0, sizeof (*h0)); + } + } + else + { + b0->error = node->errors[UDP_ERROR_LENGTH_ERROR]; + next0 = UDP_INPUT_NEXT_DROP; + } + + trace_x1: + if (PREDICT_FALSE (b0->flags & VLIB_BUFFER_IS_TRACED)) + { + udp_rx_trace_t *tr = vlib_add_trace (vm, node, + b0, sizeof (*tr)); + if (b0->error != node->errors[UDP_ERROR_LENGTH_ERROR]) + { + tr->src_port = h0->src_port; + tr->dst_port = h0->dst_port; + tr->bound = (next0 != UDP_INPUT_NEXT_ICMP4_ERROR && + next0 != UDP_INPUT_NEXT_ICMP6_ERROR); + } + } + + vlib_validate_buffer_enqueue_x1 (vm, node, next_index, + to_next, n_left_to_next, + bi0, next0); + } + + vlib_put_next_frame (vm, node, next_index, n_left_to_next); + } + vlib_error_count (vm, node->node_index, UDP_ERROR_NO_LISTENER, + n_no_listener); + return from_frame->n_vectors; +} + +static char *udp_error_strings[] = { +#define udp_error(n,s) s, +#include "udp_error.def" +#undef udp_error +}; + +static uword +udp4_input (vlib_main_t * vm, + vlib_node_runtime_t * node, vlib_frame_t * from_frame) +{ + return udp46_input_inline (vm, node, from_frame, 1 /* is_ip4 */ ); +} + +static uword +udp6_input (vlib_main_t * vm, + vlib_node_runtime_t * node, vlib_frame_t * from_frame) +{ + return udp46_input_inline (vm, node, from_frame, 0 /* is_ip4 */ ); +} + + +/* *INDENT-OFF* */ +VLIB_REGISTER_NODE (udp4_input_node) = { + .function = udp4_input, + .name = "ip4-udp-lookup", + /* Takes a vector of packets. */ + .vector_size = sizeof (u32), + + .runtime_data_bytes = sizeof (udp_input_runtime_t), + + .n_errors = UDP_N_ERROR, + .error_strings = udp_error_strings, + + .n_next_nodes = UDP_INPUT_N_NEXT, + .next_nodes = { +#define _(s,n) [UDP_INPUT_NEXT_##s] = n, + foreach_udp_input_next +#undef _ + }, + + .format_buffer = format_udp_header, + .format_trace = format_udp_rx_trace, + .unformat_buffer = unformat_udp_header, +}; +/* *INDENT-ON* */ + +VLIB_NODE_FUNCTION_MULTIARCH (udp4_input_node, udp4_input); + +/* *INDENT-OFF* */ +VLIB_REGISTER_NODE (udp6_input_node) = { + .function = udp6_input, + .name = "ip6-udp-lookup", + /* Takes a vector of packets. */ + .vector_size = sizeof (u32), + + .runtime_data_bytes = sizeof (udp_input_runtime_t), + + .n_errors = UDP_N_ERROR, + .error_strings = udp_error_strings, + + .n_next_nodes = UDP_INPUT_N_NEXT, + .next_nodes = { +#define _(s,n) [UDP_INPUT_NEXT_##s] = n, + foreach_udp_input_next +#undef _ + }, + + .format_buffer = format_udp_header, + .format_trace = format_udp_rx_trace, + .unformat_buffer = unformat_udp_header, +}; +/* *INDENT-ON* */ + +VLIB_NODE_FUNCTION_MULTIARCH (udp6_input_node, udp6_input); + +static void +add_dst_port (udp_main_t * um, + udp_dst_port_t dst_port, char *dst_port_name, u8 is_ip4) +{ + udp_dst_port_info_t *pi; + u32 i; + + vec_add2 (um->dst_port_infos[is_ip4], pi, 1); + i = pi - um->dst_port_infos[is_ip4]; + + pi->name = dst_port_name; + pi->dst_port = dst_port; + pi->next_index = pi->node_index = ~0; + + hash_set (um->dst_port_info_by_dst_port[is_ip4], dst_port, i); + + if (pi->name) + hash_set_mem (um->dst_port_info_by_name[is_ip4], pi->name, i); +} + +void +udp_register_dst_port (vlib_main_t * vm, + udp_dst_port_t dst_port, u32 node_index, u8 is_ip4) +{ + udp_main_t *um = &udp_main; + udp_dst_port_info_t *pi; + udp_input_runtime_t *rt; + u16 *n; + + { + clib_error_t *error = vlib_call_init_function (vm, udp_local_init); + if (error) + clib_error_report (error); + } + + pi = udp_get_dst_port_info (um, dst_port, is_ip4); + if (!pi) + { + add_dst_port (um, dst_port, 0, is_ip4); + pi = udp_get_dst_port_info (um, dst_port, is_ip4); + ASSERT (pi); + } + + pi->node_index = node_index; + pi->next_index = vlib_node_add_next (vm, + is_ip4 ? udp4_input_node.index + : udp6_input_node.index, node_index); + + /* Setup udp protocol -> next index sparse vector mapping. */ + rt = vlib_node_get_runtime_data + (vm, is_ip4 ? udp4_input_node.index : udp6_input_node.index); + n = sparse_vec_validate (rt->next_by_dst_port, + clib_host_to_net_u16 (dst_port)); + n[0] = pi->next_index; +} + +void +udp_unregister_dst_port (vlib_main_t * vm, udp_dst_port_t dst_port, u8 is_ip4) +{ + udp_main_t *um = &udp_main; + udp_dst_port_info_t *pi; + udp_input_runtime_t *rt; + u16 *n; + + pi = udp_get_dst_port_info (um, dst_port, is_ip4); + /* Not registered? Fagedaboudit */ + if (!pi) + return; + + /* Kill the mapping. Don't bother killing the pi, it may be back. */ + rt = vlib_node_get_runtime_data + (vm, is_ip4 ? udp4_input_node.index : udp6_input_node.index); + n = sparse_vec_validate (rt->next_by_dst_port, + clib_host_to_net_u16 (dst_port)); + n[0] = SPARSE_VEC_INVALID_INDEX; +} + +void +udp_punt_unknown (vlib_main_t * vm, u8 is_ip4, u8 is_add) +{ + udp_input_runtime_t *rt; + + { + clib_error_t *error = vlib_call_init_function (vm, udp_local_init); + if (error) + clib_error_report (error); + } + + rt = vlib_node_get_runtime_data + (vm, is_ip4 ? udp4_input_node.index : udp6_input_node.index); + + rt->punt_unknown = is_add; +} + +/* Parse a UDP header. */ +uword +unformat_udp_header (unformat_input_t * input, va_list * args) +{ + u8 **result = va_arg (*args, u8 **); + udp_header_t *udp; + __attribute__ ((unused)) int old_length; + u16 src_port, dst_port; + + /* Allocate space for IP header. */ + { + void *p; + + old_length = vec_len (*result); + vec_add2 (*result, p, sizeof (ip4_header_t)); + udp = p; + } + + memset (udp, 0, sizeof (udp[0])); + if (unformat (input, "src-port %d dst-port %d", &src_port, &dst_port)) + { + udp->src_port = clib_host_to_net_u16 (src_port); + udp->dst_port = clib_host_to_net_u16 (dst_port); + return 1; + } + return 0; +} + +static void +udp_setup_node (vlib_main_t * vm, u32 node_index) +{ + vlib_node_t *n = vlib_get_node (vm, node_index); + pg_node_t *pn = pg_get_node (node_index); + + n->format_buffer = format_udp_header; + n->unformat_buffer = unformat_udp_header; + pn->unformat_edit = unformat_pg_udp_header; +} + +clib_error_t * +udp_local_init (vlib_main_t * vm) +{ + udp_input_runtime_t *rt; + udp_main_t *um = &udp_main; + int i; + + { + clib_error_t *error; + error = vlib_call_init_function (vm, udp_init); + if (error) + clib_error_report (error); + } + + + for (i = 0; i < 2; i++) + { + um->dst_port_info_by_name[i] = hash_create_string (0, sizeof (uword)); + um->dst_port_info_by_dst_port[i] = hash_create (0, sizeof (uword)); + } + + udp_setup_node (vm, udp4_input_node.index); + udp_setup_node (vm, udp6_input_node.index); + + rt = vlib_node_get_runtime_data (vm, udp4_input_node.index); + + rt->next_by_dst_port = sparse_vec_new + ( /* elt bytes */ sizeof (rt->next_by_dst_port[0]), + /* bits in index */ BITS (((udp_header_t *) 0)->dst_port)); + + rt->punt_unknown = 0; + +#define _(n,s) add_dst_port (um, UDP_DST_PORT_##s, #s, 1 /* is_ip4 */); + foreach_udp4_dst_port +#undef _ + rt = vlib_node_get_runtime_data (vm, udp6_input_node.index); + + rt->next_by_dst_port = sparse_vec_new + ( /* elt bytes */ sizeof (rt->next_by_dst_port[0]), + /* bits in index */ BITS (((udp_header_t *) 0)->dst_port)); + + rt->punt_unknown = 0; + +#define _(n,s) add_dst_port (um, UDP_DST_PORT_##s, #s, 0 /* is_ip4 */); + foreach_udp6_dst_port +#undef _ + ip4_register_protocol (IP_PROTOCOL_UDP, udp4_input_node.index); + /* Note: ip6 differs from ip4, UDP is hotwired to ip6-udp-lookup */ + return 0; +} + +VLIB_INIT_FUNCTION (udp_local_init); + +/* + * fd.io coding-style-patch-verification: ON + * + * Local Variables: + * eval: (c-set-style "gnu") + * End: + */ diff --git a/src/vnet/udp/udp_packet.h b/src/vnet/udp/udp_packet.h new file mode 100644 index 00000000..beea3059 --- /dev/null +++ b/src/vnet/udp/udp_packet.h @@ -0,0 +1,65 @@ +/* + * Copyright (c) 2015 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ +/* + * ip4/udp_packet.h: UDP packet format + * + * Copyright (c) 2008 Eliot Dresselhaus + * + * Permission is hereby granted, free of charge, to any person obtaining + * a copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sublicense, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice shall be + * included in all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND + * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE + * LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + * OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + * WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + */ + +#ifndef included_udp_packet_h +#define included_udp_packet_h + +typedef struct +{ + /* Source and destination port. */ + u16 src_port, dst_port; + + /* Length of UDP header plus payload. */ + u16 length; + + /* Checksum of UDP pseudo-header and data or + zero if checksum is disabled. */ + u16 checksum; +} udp_header_t; + +#endif /* included_udp_packet_h */ + + +/* + * fd.io coding-style-patch-verification: ON + * + * Local Variables: + * eval: (c-set-style "gnu") + * End: + */ diff --git a/src/vnet/udp/udp_pg.c b/src/vnet/udp/udp_pg.c new file mode 100644 index 00000000..c9d8d38c --- /dev/null +++ b/src/vnet/udp/udp_pg.c @@ -0,0 +1,237 @@ +/* + * Copyright (c) 2015 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ +/* + * ip/udp_pg: UDP packet-generator interface + * + * Copyright (c) 2008 Eliot Dresselhaus + * + * Permission is hereby granted, free of charge, to any person obtaining + * a copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sublicense, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice shall be + * included in all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND + * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE + * LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + * OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + * WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + */ + +#include +#include /* for unformat_udp_udp_port */ + +#define UDP_PG_EDIT_LENGTH (1 << 0) +#define UDP_PG_EDIT_CHECKSUM (1 << 1) + +always_inline void +udp_pg_edit_function_inline (pg_main_t * pg, + pg_stream_t * s, + pg_edit_group_t * g, + u32 * packets, u32 n_packets, u32 flags) +{ + vlib_main_t *vm = vlib_get_main (); + u32 ip_offset, udp_offset; + + udp_offset = g->start_byte_offset; + ip_offset = (g - 1)->start_byte_offset; + + while (n_packets >= 1) + { + vlib_buffer_t *p0; + ip4_header_t *ip0; + udp_header_t *udp0; + u32 udp_len0; + + p0 = vlib_get_buffer (vm, packets[0]); + n_packets -= 1; + packets += 1; + + ip0 = (void *) (p0->data + ip_offset); + udp0 = (void *) (p0->data + udp_offset); + udp_len0 = clib_net_to_host_u16 (ip0->length) - sizeof (ip0[0]); + + if (flags & UDP_PG_EDIT_LENGTH) + udp0->length = + clib_net_to_host_u16 (vlib_buffer_length_in_chain (vm, p0) + - ip_offset); + + /* Initialize checksum with header. */ + if (flags & UDP_PG_EDIT_CHECKSUM) + { + ip_csum_t sum0; + + sum0 = clib_mem_unaligned (&ip0->src_address, u64); + + sum0 = ip_csum_with_carry + (sum0, clib_host_to_net_u32 (udp_len0 + (ip0->protocol << 16))); + + /* Invalidate possibly old checksum. */ + udp0->checksum = 0; + + sum0 = + ip_incremental_checksum_buffer (vm, p0, udp_offset, udp_len0, + sum0); + + sum0 = ~ip_csum_fold (sum0); + + /* Zero checksum means checksumming disabled. */ + sum0 = sum0 != 0 ? sum0 : 0xffff; + + udp0->checksum = sum0; + } + } +} + +static void +udp_pg_edit_function (pg_main_t * pg, + pg_stream_t * s, + pg_edit_group_t * g, u32 * packets, u32 n_packets) +{ + switch (g->edit_function_opaque) + { + case UDP_PG_EDIT_LENGTH: + udp_pg_edit_function_inline (pg, s, g, packets, n_packets, + UDP_PG_EDIT_LENGTH); + break; + + case UDP_PG_EDIT_CHECKSUM: + udp_pg_edit_function_inline (pg, s, g, packets, n_packets, + UDP_PG_EDIT_CHECKSUM); + break; + + case UDP_PG_EDIT_CHECKSUM | UDP_PG_EDIT_LENGTH: + udp_pg_edit_function_inline (pg, s, g, packets, n_packets, + UDP_PG_EDIT_CHECKSUM | UDP_PG_EDIT_LENGTH); + break; + + default: + ASSERT (0); + break; + } +} + +typedef struct +{ + pg_edit_t src_port, dst_port; + pg_edit_t length; + pg_edit_t checksum; +} pg_udp_header_t; + +static inline void +pg_udp_header_init (pg_udp_header_t * p) +{ + /* Initialize fields that are not bit fields in the IP header. */ +#define _(f) pg_edit_init (&p->f, udp_header_t, f); + _(src_port); + _(dst_port); + _(length); + _(checksum); +#undef _ +} + +uword +unformat_pg_udp_header (unformat_input_t * input, va_list * args) +{ + pg_stream_t *s = va_arg (*args, pg_stream_t *); + pg_udp_header_t *p; + u32 group_index; + + p = pg_create_edit_group (s, sizeof (p[0]), sizeof (udp_header_t), + &group_index); + pg_udp_header_init (p); + + /* Defaults. */ + p->checksum.type = PG_EDIT_UNSPECIFIED; + p->length.type = PG_EDIT_UNSPECIFIED; + + if (!unformat (input, "UDP: %U -> %U", + unformat_pg_edit, + unformat_tcp_udp_port, &p->src_port, + unformat_pg_edit, unformat_tcp_udp_port, &p->dst_port)) + goto error; + + /* Parse options. */ + while (1) + { + if (unformat (input, "length %U", + unformat_pg_edit, unformat_pg_number, &p->length)) + ; + + else if (unformat (input, "checksum %U", + unformat_pg_edit, unformat_pg_number, &p->checksum)) + ; + + /* Can't parse input: try next protocol level. */ + else + break; + } + + { + ip_main_t *im = &ip_main; + u16 dst_port; + tcp_udp_port_info_t *pi; + + pi = 0; + if (p->dst_port.type == PG_EDIT_FIXED) + { + dst_port = pg_edit_get_value (&p->dst_port, PG_EDIT_LO); + pi = ip_get_tcp_udp_port_info (im, dst_port); + } + + if (pi && pi->unformat_pg_edit + && unformat_user (input, pi->unformat_pg_edit, s)) + ; + + else if (!unformat_user (input, unformat_pg_payload, s)) + goto error; + + p = pg_get_edit_group (s, group_index); + if (p->checksum.type == PG_EDIT_UNSPECIFIED + || p->length.type == PG_EDIT_UNSPECIFIED) + { + pg_edit_group_t *g = pg_stream_get_group (s, group_index); + g->edit_function = udp_pg_edit_function; + g->edit_function_opaque = 0; + if (p->checksum.type == PG_EDIT_UNSPECIFIED) + g->edit_function_opaque |= UDP_PG_EDIT_CHECKSUM; + if (p->length.type == PG_EDIT_UNSPECIFIED) + g->edit_function_opaque |= UDP_PG_EDIT_LENGTH; + } + + return 1; + } + +error: + /* Free up any edits we may have added. */ + pg_free_edit_group (s); + return 0; +} + + +/* + * fd.io coding-style-patch-verification: ON + * + * Local Variables: + * eval: (c-set-style "gnu") + * End: + */ diff --git a/src/vnet/vnet_all_api_h.h b/src/vnet/vnet_all_api_h.h index 142acedc..c4075db6 100644 --- a/src/vnet/vnet_all_api_h.h +++ b/src/vnet/vnet_all_api_h.h @@ -51,6 +51,7 @@ #include #include #include +#include #include #include #include diff --git a/src/vnet/vxlan-gpe/vxlan_gpe.h b/src/vnet/vxlan-gpe/vxlan_gpe.h index 1b4bc44e..e768d230 100644 --- a/src/vnet/vxlan-gpe/vxlan_gpe.h +++ b/src/vnet/vxlan-gpe/vxlan_gpe.h @@ -29,7 +29,7 @@ #include #include #include -#include +#include /** * @brief VXLAN GPE header struct diff --git a/src/vnet/vxlan/vxlan.h b/src/vnet/vxlan/vxlan.h index adfa3a8e..dca1cd12 100644 --- a/src/vnet/vxlan/vxlan.h +++ b/src/vnet/vxlan/vxlan.h @@ -26,7 +26,7 @@ #include #include #include -#include +#include #include #include diff --git a/src/vpp/api/vpe.api b/src/vpp/api/vpe.api index 24f48293..2d6e4f37 100644 --- a/src/vpp/api/vpe.api +++ b/src/vpp/api/vpe.api @@ -38,6 +38,7 @@ * IPSEC-GRE APIs: see .../src/vnet/ipsec-gre/{ipsec_gre.api, ipsec_gre_api.c} * LISP APIs: see .../src/vnet/lisp/{lisp.api, lisp_api.c} * LISP-GPE APIs: see .../src/vnet/lisp-gpe/{lisp_gpe.api, lisp_gpe_api.c} + * SESSION APIs: .../vnet/session/{session.api session_api.c} * MPLS APIs: see .../src/vnet/mpls/{mpls.api, mpls_api.c} * SR APIs: see .../src/vnet/sr/{sr.api, sr_api.c} * DPDK APIs: see ... /src/vnet/devices/dpdk/{dpdk.api, dpdk_api.c} diff --git a/src/vppinfra.am b/src/vppinfra.am index 8d375958..4b9f0c29 100644 --- a/src/vppinfra.am +++ b/src/vppinfra.am @@ -157,7 +157,9 @@ nobase_include_HEADERS = \ vppinfra/asm_mips.h \ vppinfra/asm_x86.h \ vppinfra/bihash_8_8.h \ + vppinfra/bihash_16_8.h \ vppinfra/bihash_24_8.h \ + vppinfra/bihash_48_8.h \ vppinfra/bihash_template.h \ vppinfra/bihash_template.c \ vppinfra/bitmap.h \ @@ -206,6 +208,7 @@ nobase_include_HEADERS = \ vppinfra/timer.h \ vppinfra/tw_timer_2t_1w_2048sl.h \ vppinfra/tw_timer_16t_2w_512sl.h \ + vppinfra/tw_timer_16t_1w_2048sl.h \ vppinfra/tw_timer_template.h \ vppinfra/tw_timer_template.c \ vppinfra/types.h \ @@ -261,6 +264,8 @@ CLIB_CORE = \ vppinfra/tw_timer_2t_1w_2048sl.c \ vppinfra/tw_timer_16t_2w_512sl.h \ vppinfra/tw_timer_16t_2w_512sl.c \ + vppinfra/tw_timer_16t_1w_2048sl.h \ + vppinfra/tw_timer_16t_1w_2048sl.c \ vppinfra/unformat.c \ vppinfra/vec.c \ vppinfra/vector.c \ diff --git a/src/vppinfra/bihash_16_8.h b/src/vppinfra/bihash_16_8.h new file mode 100644 index 00000000..ce80f70e --- /dev/null +++ b/src/vppinfra/bihash_16_8.h @@ -0,0 +1,103 @@ +/* + * Copyright (c) 2016 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ +#undef BIHASH_TYPE + +#define BIHASH_TYPE _16_8 +#define BIHASH_KVP_PER_PAGE 4 + +#ifndef __included_bihash_16_8_h__ +#define __included_bihash_16_8_h__ + +#include +#include +#include +#include + +typedef struct +{ + u64 key[2]; + u64 value; +} clib_bihash_kv_16_8_t; + +static inline int +clib_bihash_is_free_16_8 (clib_bihash_kv_16_8_t * v) +{ + /* Free values are memset to 0xff, check a bit... */ + if (v->key[0] == ~0ULL && v->value == ~0ULL) + return 1; + return 0; +} + +#if __SSE4_2__ +#ifndef __defined_crc_u32__ +#define __defined_crc_u32__ +static inline u32 +crc_u32 (u32 data, u32 value) +{ + __asm__ volatile ("crc32l %[data], %[value];":[value] "+r" (value):[data] + "rm" (data)); + return value; +} +#endif /* __defined_crc_u32__ */ + +static inline u64 +clib_bihash_hash_16_8 (clib_bihash_kv_16_8_t * v) +{ + u32 *dp = (u32 *) & v->key[0]; + u32 value = 0; + + value = crc_u32 (dp[0], value); + value = crc_u32 (dp[1], value); + value = crc_u32 (dp[2], value); + value = crc_u32 (dp[3], value); + + return value; +} +#else +static inline u64 +clib_bihash_hash_16_8 (clib_bihash_kv_16_8_t * v) +{ + u64 tmp = v->key[0] ^ v->key[1]; + return clib_xxhash (tmp); +} +#endif + +static inline u8 * +format_bihash_kvp_16_8 (u8 * s, va_list * args) +{ + clib_bihash_kv_16_8_t *v = va_arg (*args, clib_bihash_kv_16_8_t *); + + s = format (s, "key %llu %llu value %llu", v->key[0], v->key[1], v->value); + return s; +} + +static inline int +clib_bihash_key_compare_16_8 (u64 * a, u64 * b) +{ + return ((a[0] ^ b[0]) | (a[1] ^ b[1])) == 0; +} + +#undef __included_bihash_template_h__ +#include + +#endif /* __included_bihash_16_8_h__ */ + +/* + * fd.io coding-style-patch-verification: ON + * + * Local Variables: + * eval: (c-set-style "gnu") + * End: + */ diff --git a/src/vppinfra/bihash_48_8.h b/src/vppinfra/bihash_48_8.h new file mode 100644 index 00000000..1a6e7691 --- /dev/null +++ b/src/vppinfra/bihash_48_8.h @@ -0,0 +1,116 @@ +/* + * Copyright (c) 2016 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +#undef BIHASH_TYPE + +#define BIHASH_TYPE _48_8 +#define BIHASH_KVP_PER_PAGE 4 + +#ifndef __included_bihash_48_8_h__ +#define __included_bihash_48_8_h__ + +#include +#include +#include +#include + +typedef struct +{ + u64 key[6]; + u64 value; +} clib_bihash_kv_48_8_t; + +static inline int +clib_bihash_is_free_48_8 (const clib_bihash_kv_48_8_t * v) +{ + /* Free values are memset to 0xff, check a bit... */ + if (v->key[0] == ~0ULL && v->value == ~0ULL) + return 1; + return 0; +} + +#if __SSE4_2__ +#ifndef __defined_crc_u32__ +#define __defined_crc_u32__ +static inline u32 +crc_u32 (u32 data, u32 value) +{ + __asm__ volatile ("crc32l %[data], %[value];":[value] "+r" (value):[data] + "rm" (data)); + return value; +} +#endif /* __defined_crc_u32__ */ + +static inline u64 +clib_bihash_hash_48_8 (const clib_bihash_kv_48_8_t * v) +{ + const u32 *dp = (const u32 *) &v->key[0]; + u32 value = 0; + + value = crc_u32 (dp[0], value); + value = crc_u32 (dp[1], value); + value = crc_u32 (dp[2], value); + value = crc_u32 (dp[3], value); + value = crc_u32 (dp[4], value); + value = crc_u32 (dp[5], value); + value = crc_u32 (dp[6], value); + value = crc_u32 (dp[7], value); + value = crc_u32 (dp[8], value); + value = crc_u32 (dp[9], value); + value = crc_u32 (dp[10], value); + value = crc_u32 (dp[11], value); + + return value; +} +#else +static inline u64 +clib_bihash_hash_48_8 (const clib_bihash_kv_48_8_t * v) +{ + u64 tmp = v->key[0] ^ v->key[1] ^ v->key[2] ^ v->key[3] ^ v->key[4] + ^ v->key[5]; + return clib_xxhash (tmp); +} +#endif + +static inline u8 * +format_bihash_kvp_48_8 (u8 * s, va_list * args) +{ + clib_bihash_kv_48_8_t *v = va_arg (*args, clib_bihash_kv_48_8_t *); + + s = format (s, "key %llu %llu %llu %llu %llu %llu value %llu", v->key[0], + v->key[1], v->key[2], v->key[3], v->key[4], v->key[5], + v->value); + return s; +} + +static inline int +clib_bihash_key_compare_48_8 (const u64 * a, const u64 * b) +{ + return ((a[0] ^ b[0]) | (a[1] ^ b[1]) | (a[2] ^ b[2]) | (a[3] ^ b[3]) + | (a[4] ^ b[4]) | (a[5] ^ b[5])) == 0; +} + +#undef __included_bihash_template_h__ +#include + +#endif /* __included_bihash_48_8_h__ */ + +/* + * fd.io coding-style-patch-verification: ON + * + * Local Variables: + * eval: (c-set-style "gnu") + * End: + */ diff --git a/src/vppinfra/tw_timer_16t_1w_2048sl.c b/src/vppinfra/tw_timer_16t_1w_2048sl.c new file mode 100644 index 00000000..3f342045 --- /dev/null +++ b/src/vppinfra/tw_timer_16t_1w_2048sl.c @@ -0,0 +1,26 @@ +/* + * Copyright (c) 2017 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +#include +#include "tw_timer_16t_1w_2048sl.h" +#include "tw_timer_template.c" + +/* + * fd.io coding-style-patch-verification: ON + * + * Local Variables: + * eval: (c-set-style "gnu") + * End: + */ diff --git a/src/vppinfra/tw_timer_16t_1w_2048sl.h b/src/vppinfra/tw_timer_16t_1w_2048sl.h new file mode 100644 index 00000000..685ac31e --- /dev/null +++ b/src/vppinfra/tw_timer_16t_1w_2048sl.h @@ -0,0 +1,46 @@ +/* + * Copyright (c) 2017 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +#ifndef __included_tw_timer_16t_2w_512sl_h__ +#define __included_tw_timer_16t_2w_512sl_h__ + +/* ... So that a client app can create multiple wheel geometries */ +#undef TW_TIMER_WHEELS +#undef TW_SLOTS_PER_RING +#undef TW_RING_SHIFT +#undef TW_RING_MASK +#undef TW_TIMERS_PER_OBJECT +#undef LOG2_TW_TIMERS_PER_OBJECT +#undef TW_SUFFIX + +#define TW_TIMER_WHEELS 1 +#define TW_SLOTS_PER_RING 2048 +#define TW_RING_SHIFT 11 +#define TW_RING_MASK (TW_SLOTS_PER_RING -1) +#define TW_TIMERS_PER_OBJECT 16 +#define LOG2_TW_TIMERS_PER_OBJECT 4 +#define TW_SUFFIX _16t_1w_2048sl + +#include + +#endif /* __included_tw_timer_16t_2w_512sl_h__ */ + +/* + * fd.io coding-style-patch-verification: ON + * + * Local Variables: + * eval: (c-set-style "gnu") + * End: + */ -- cgit 1.2.3-korg From a1a093d4e46e38503332a97ad216f80053a15f2b Mon Sep 17 00:00:00 2001 From: Dave Barach Date: Thu, 2 Mar 2017 13:13:23 -0500 Subject: Clean up binary api message handler registration issues Removed a fair number of "BUG" message handlers, due to conflicts with actual message handlers in api_format.c. Vpp itself had no business receiving certain messages, up to the point where we started building in relevant code from vpp_api_test. Eliminated all but one duplicate registration complaint. That one needs attention from the vxlan team since the duplicated handlers have diverged. Change-Id: Iafce5429d2f906270643b4ea5f0130e20beb4d1d Signed-off-by: Dave Barach --- src/vat/api_format.c | 43 ++++++- src/vlib/unix/input.c | 31 ++++- src/vlibapi/api.h | 15 +++ src/vlibapi/api_shared.c | 4 + src/vnet/classify/classify_api.c | 8 -- src/vnet/devices/virtio/vhost_user_api.c | 10 +- src/vnet/dhcp/dhcp_api.c | 8 -- src/vnet/interface_api.c | 7 - src/vnet/ip/ip_api.c | 83 ------------ src/vnet/l2/l2_api.c | 22 ---- src/vnet/mpls/mpls_api.c | 28 +--- src/vpp/api/api.c | 211 ------------------------------- src/vpp/api/api_main.c | 1 - src/vpp/stats/stats.c | 7 - 14 files changed, 81 insertions(+), 397 deletions(-) (limited to 'src/vnet/classify') diff --git a/src/vat/api_format.c b/src/vat/api_format.c index 1321bade..52436917 100644 --- a/src/vat/api_format.c +++ b/src/vat/api_format.c @@ -944,6 +944,7 @@ static void vl_api_sw_interface_details_t_handler_json } } +#if VPP_API_TEST_BUILTIN == 0 static void vl_api_sw_interface_set_flags_t_handler (vl_api_sw_interface_set_flags_t * mp) { @@ -954,6 +955,7 @@ static void vl_api_sw_interface_set_flags_t_handler mp->admin_up_down ? "admin-up" : "admin-down", mp->link_up_down ? "link-up" : "link-down"); } +#endif static void vl_api_sw_interface_set_flags_t_handler_json (vl_api_sw_interface_set_flags_t * mp) @@ -4009,7 +4011,6 @@ foreach_standard_reply_retval_handler; #define foreach_vpe_api_reply_msg \ _(CREATE_LOOPBACK_REPLY, create_loopback_reply) \ _(SW_INTERFACE_DETAILS, sw_interface_details) \ -_(SW_INTERFACE_SET_FLAGS, sw_interface_set_flags) \ _(SW_INTERFACE_SET_FLAGS_REPLY, sw_interface_set_flags_reply) \ _(CONTROL_PING_REPLY, control_ping_reply) \ _(CLI_REPLY, cli_reply) \ @@ -4126,11 +4127,6 @@ _(IKEV2_INITIATE_REKEY_CHILD_SA_REPLY, ikev2_initiate_rekey_child_sa_reply) \ _(DELETE_LOOPBACK_REPLY, delete_loopback_reply) \ _(BD_IP_MAC_ADD_DEL_REPLY, bd_ip_mac_add_del_reply) \ _(DHCP_COMPL_EVENT, dhcp_compl_event) \ -_(VNET_INTERFACE_COUNTERS, vnet_interface_counters) \ -_(VNET_IP4_FIB_COUNTERS, vnet_ip4_fib_counters) \ -_(VNET_IP6_FIB_COUNTERS, vnet_ip6_fib_counters) \ -_(VNET_IP4_NBR_COUNTERS, vnet_ip4_nbr_counters) \ -_(VNET_IP6_NBR_COUNTERS, vnet_ip6_nbr_counters) \ _(MAP_ADD_DOMAIN_REPLY, map_add_domain_reply) \ _(MAP_DEL_DOMAIN_REPLY, map_del_domain_reply) \ _(MAP_ADD_DEL_RULE_REPLY, map_add_del_rule_reply) \ @@ -4232,6 +4228,14 @@ _(SW_INTERFACE_SET_MTU_REPLY, sw_interface_set_mtu_reply) \ _(IP_NEIGHBOR_DETAILS, ip_neighbor_details) \ _(SW_INTERFACE_GET_TABLE_REPLY, sw_interface_get_table_reply) +#define foreach_standalone_reply_msg \ +_(SW_INTERFACE_SET_FLAGS, sw_interface_set_flags) \ +_(VNET_INTERFACE_COUNTERS, vnet_interface_counters) \ +_(VNET_IP4_FIB_COUNTERS, vnet_ip4_fib_counters) \ +_(VNET_IP6_FIB_COUNTERS, vnet_ip6_fib_counters) \ +_(VNET_IP4_NBR_COUNTERS, vnet_ip4_nbr_counters) \ +_(VNET_IP6_NBR_COUNTERS, vnet_ip6_nbr_counters) + typedef struct { u8 *name; @@ -15425,7 +15429,15 @@ api_af_packet_create (vat_main_t * vam) vec_free (host_if_name); S (mp); - W2 (ret, fprintf (vam->ofp, " new sw_if_index = %d ", vam->sw_if_index)); + + /* *INDENT-OFF* */ + W2 (ret, + ({ + if (ret == 0) + fprintf (vam->ofp ? vam->ofp : stderr, + " new sw_if_index = %d\n", vam->sw_if_index); + })); + /* *INDENT-ON* */ return ret; } @@ -18417,6 +18429,9 @@ _(unset, "usage: unset ") } \ } foreach_vpe_api_reply_msg; +#if VPP_API_TEST_BUILTIN == 0 +foreach_standalone_reply_msg; +#endif #undef _ void @@ -18430,6 +18445,9 @@ vat_api_hookup (vat_main_t * vam) vl_api_##n##_t_print, \ sizeof(vl_api_##n##_t), 1); foreach_vpe_api_reply_msg; +#if VPP_API_TEST_BUILTIN == 0 + foreach_standalone_reply_msg; +#endif #undef _ #if (VPP_API_TEST_BUILTIN==0) @@ -18463,6 +18481,17 @@ vat_api_hookup (vat_main_t * vam) #undef _ } +#if VPP_API_TEST_BUILTIN +static clib_error_t * +vat_api_hookup_shim (vlib_main_t * vm) +{ + vat_api_hookup (&vat_main); + return 0; +} + +VLIB_API_INIT_FUNCTION (vat_api_hookup_shim); +#endif + /* * fd.io coding-style-patch-verification: ON * diff --git a/src/vlib/unix/input.c b/src/vlib/unix/input.c index 07096ed2..7b4183a4 100644 --- a/src/vlib/unix/input.c +++ b/src/vlib/unix/input.c @@ -66,6 +66,7 @@ linux_epoll_file_update (unix_file_t * f, unix_file_update_type_t update_type) unix_main_t *um = &unix_main; linux_epoll_main_t *em = &linux_epoll_main; struct epoll_event e; + int op; memset (&e, 0, sizeof (e)); @@ -76,13 +77,29 @@ linux_epoll_file_update (unix_file_t * f, unix_file_update_type_t update_type) e.events |= EPOLLET; e.data.u32 = f - um->file_pool; - if (epoll_ctl (em->epoll_fd, - (update_type == UNIX_FILE_UPDATE_ADD - ? EPOLL_CTL_ADD - : (update_type == UNIX_FILE_UPDATE_MODIFY - ? EPOLL_CTL_MOD - : EPOLL_CTL_DEL)), f->file_descriptor, &e) < 0) - clib_warning ("epoll_ctl"); + op = -1; + + switch (update_type) + { + case UNIX_FILE_UPDATE_ADD: + op = EPOLL_CTL_ADD; + break; + + case UNIX_FILE_UPDATE_MODIFY: + op = EPOLL_CTL_MOD; + break; + + case UNIX_FILE_UPDATE_DELETE: + op = EPOLL_CTL_DEL; + break; + + default: + clib_warning ("unknown update_type %d", update_type); + return; + } + + if (epoll_ctl (em->epoll_fd, op, f->file_descriptor, &e) < 0) + clib_unix_warning ("epoll_ctl"); } static uword diff --git a/src/vlibapi/api.h b/src/vlibapi/api.h index fcb101d7..b40ece15 100644 --- a/src/vlibapi/api.h +++ b/src/vlibapi/api.h @@ -271,6 +271,21 @@ vlib_node_t **vlib_node_unserialize (u8 * vector); #define VLIB_API_INIT_FUNCTION(x) VLIB_DECLARE_INIT_FUNCTION(x,api_init) +/* Call given init function: used for init function dependencies. */ +#define vlib_call_api_init_function(vm, x) \ + ({ \ + extern vlib_init_function_t * _VLIB_INIT_FUNCTION_SYMBOL (x,api_init); \ + vlib_init_function_t * _f = _VLIB_INIT_FUNCTION_SYMBOL (x,api_init); \ + clib_error_t * _error = 0; \ + if (! hash_get (vm->init_functions_called, _f)) \ + { \ + hash_set1 (vm->init_functions_called, _f); \ + _error = _f (vm); \ + } \ + _error; \ + }) + + #endif /* included_api_h */ /* diff --git a/src/vlibapi/api_shared.c b/src/vlibapi/api_shared.c index 1a2740e2..79921afe 100644 --- a/src/vlibapi/api_shared.c +++ b/src/vlibapi/api_shared.c @@ -667,6 +667,10 @@ vl_msg_api_config (vl_msg_api_msg_config_t * c) foreach_msg_api_vector; #undef _ + if (am->msg_names[c->id]) + clib_warning ("BUG: multiple registrations of 'vl_api_%s_t_handler'", + c->name); + am->msg_names[c->id] = c->name; am->msg_handlers[c->id] = c->handler; am->msg_cleanup_handlers[c->id] = c->cleanup; diff --git a/src/vnet/classify/classify_api.c b/src/vnet/classify/classify_api.c index 77a8b434..24c7a2b9 100644 --- a/src/vnet/classify/classify_api.c +++ b/src/vnet/classify/classify_api.c @@ -53,7 +53,6 @@ _(CLASSIFY_TABLE_IDS,classify_table_ids) \ _(CLASSIFY_TABLE_BY_INTERFACE, classify_table_by_interface) \ _(CLASSIFY_TABLE_INFO,classify_table_info) \ _(CLASSIFY_SESSION_DUMP,classify_session_dump) \ -_(CLASSIFY_SESSION_DETAILS,classify_session_details) \ _(POLICER_CLASSIFY_SET_INTERFACE, policer_classify_set_interface) \ _(POLICER_CLASSIFY_DUMP, policer_classify_dump) \ _(FLOW_CLASSIFY_SET_INTERFACE, flow_classify_set_interface) \ @@ -356,13 +355,6 @@ vl_api_classify_table_info_t_handler (vl_api_classify_table_info_t * mp) vl_msg_api_send_shmem (q, (u8 *) & rmp); } -static void -vl_api_classify_session_details_t_handler (vl_api_classify_session_details_t * - mp) -{ - clib_warning ("BUG"); -} - static void send_classify_session_details (unix_shared_memory_queue_t * q, u32 table_id, diff --git a/src/vnet/devices/virtio/vhost_user_api.c b/src/vnet/devices/virtio/vhost_user_api.c index dd517c26..8dbd032b 100644 --- a/src/vnet/devices/virtio/vhost_user_api.c +++ b/src/vnet/devices/virtio/vhost_user_api.c @@ -46,8 +46,7 @@ _(CREATE_VHOST_USER_IF, create_vhost_user_if) \ _(MODIFY_VHOST_USER_IF, modify_vhost_user_if) \ _(DELETE_VHOST_USER_IF, delete_vhost_user_if) \ -_(SW_INTERFACE_VHOST_USER_DUMP, sw_interface_vhost_user_dump) \ -_(SW_INTERFACE_VHOST_USER_DETAILS, sw_interface_vhost_user_details) +_(SW_INTERFACE_VHOST_USER_DUMP, sw_interface_vhost_user_dump) /* * WARNING: replicated pending api refactor completion @@ -148,13 +147,6 @@ vl_api_delete_vhost_user_if_t_handler (vl_api_delete_vhost_user_if_t * mp) } } -static void - vl_api_sw_interface_vhost_user_details_t_handler - (vl_api_sw_interface_vhost_user_details_t * mp) -{ - clib_warning ("BUG"); -} - static void send_sw_interface_vhost_user_details (vpe_api_main_t * am, unix_shared_memory_queue_t * q, diff --git a/src/vnet/dhcp/dhcp_api.c b/src/vnet/dhcp/dhcp_api.c index bdf02cae..ce34f6a4 100644 --- a/src/vnet/dhcp/dhcp_api.c +++ b/src/vnet/dhcp/dhcp_api.c @@ -46,7 +46,6 @@ #define foreach_vpe_api_msg \ _(DHCP_PROXY_CONFIG,dhcp_proxy_config) \ _(DHCP_PROXY_DUMP,dhcp_proxy_dump) \ -_(DHCP_PROXY_DETAILS,dhcp_proxy_details) \ _(DHCP_PROXY_SET_VSS,dhcp_proxy_set_vss) \ _(DHCP_CLIENT_CONFIG, dhcp_client_config) @@ -158,13 +157,6 @@ dhcp_send_details (fib_protocol_t proto, vl_msg_api_send_shmem (q, (u8 *) & mp); } - -static void -vl_api_dhcp_proxy_details_t_handler (vl_api_dhcp_proxy_details_t * mp) -{ - clib_warning ("BUG"); -} - void dhcp_compl_event_callback (u32 client_index, u32 pid, u8 * hostname, u8 is_ipv6, u8 * host_address, u8 * router_address, diff --git a/src/vnet/interface_api.c b/src/vnet/interface_api.c index 63f7cad4..60cd6d40 100644 --- a/src/vnet/interface_api.c +++ b/src/vnet/interface_api.c @@ -50,7 +50,6 @@ _(SW_INTERFACE_SET_FLAGS, sw_interface_set_flags) \ _(SW_INTERFACE_SET_MTU, sw_interface_set_mtu) \ _(WANT_INTERFACE_EVENTS, want_interface_events) \ _(SW_INTERFACE_DUMP, sw_interface_dump) \ -_(SW_INTERFACE_DETAILS, sw_interface_details) \ _(SW_INTERFACE_ADD_DEL_ADDRESS, sw_interface_add_del_address) \ _(SW_INTERFACE_SET_TABLE, sw_interface_set_table) \ _(SW_INTERFACE_GET_TABLE, sw_interface_get_table) \ @@ -684,12 +683,6 @@ out: REPLY_MACRO (VL_API_SW_INTERFACE_TAG_ADD_DEL_REPLY); } -static void -vl_api_sw_interface_details_t_handler (vl_api_sw_interface_details_t * mp) -{ - clib_warning ("BUG"); -} - /* * vpe_api_hookup * Add vpe's API message handlers to the table. diff --git a/src/vnet/ip/ip_api.c b/src/vnet/ip/ip_api.c index 49d941c2..ab164a5f 100644 --- a/src/vnet/ip/ip_api.c +++ b/src/vnet/ip/ip_api.c @@ -59,17 +59,12 @@ #define foreach_ip_api_msg \ _(IP_FIB_DUMP, ip_fib_dump) \ -_(IP_FIB_DETAILS, ip_fib_details) \ _(IP6_FIB_DUMP, ip6_fib_dump) \ -_(IP6_FIB_DETAILS, ip6_fib_details) \ _(IP_MFIB_DUMP, ip_mfib_dump) \ -_(IP_MFIB_DETAILS, ip_mfib_details) \ _(IP6_MFIB_DUMP, ip6_mfib_dump) \ -_(IP6_MFIB_DETAILS, ip6_mfib_details) \ _(IP_NEIGHBOR_DUMP, ip_neighbor_dump) \ _(IP_MROUTE_ADD_DEL, ip_mroute_add_del) \ _(MFIB_SIGNAL_DUMP, mfib_signal_dump) \ -_(IP_NEIGHBOR_DETAILS, ip_neighbor_details) \ _(IP_ADDRESS_DUMP, ip_address_dump) \ _(IP_DUMP, ip_dump) \ _(IP_NEIGHBOR_ADD_DEL, ip_neighbor_add_del) \ @@ -105,12 +100,6 @@ send_ip_neighbor_details (u8 is_ipv6, vl_msg_api_send_shmem (q, (u8 *) & mp); } -static void -vl_api_ip_neighbor_details_t_handler (vl_api_ip_neighbor_details_t * mp) -{ - clib_warning ("BUG"); -} - static void vl_api_ip_neighbor_dump_t_handler (vl_api_ip_neighbor_dump_t * mp) { @@ -185,24 +174,6 @@ copy_fib_next_hop (fib_route_path_encode_t * api_rpath, void *fp_arg) sizeof (api_rpath->rpath.frp_addr.ip6)); } -static void -vl_api_ip_fib_details_t_handler (vl_api_ip_fib_details_t * mp) -{ - clib_warning ("BUG"); -} - -static void -vl_api_ip_fib_details_t_endian (vl_api_ip_fib_details_t * mp) -{ - clib_warning ("BUG"); -} - -static void -vl_api_ip_fib_details_t_print (vl_api_ip_fib_details_t * mp) -{ - clib_warning ("BUG"); -} - static void send_ip_fib_details (vpe_api_main_t * am, unix_shared_memory_queue_t * q, @@ -316,24 +287,6 @@ vl_api_ip_fib_dump_t_handler (vl_api_ip_fib_dump_t * mp) vec_free (lfeis); } -static void -vl_api_ip6_fib_details_t_handler (vl_api_ip6_fib_details_t * mp) -{ - clib_warning ("BUG"); -} - -static void -vl_api_ip6_fib_details_t_endian (vl_api_ip6_fib_details_t * mp) -{ - clib_warning ("BUG"); -} - -static void -vl_api_ip6_fib_details_t_print (vl_api_ip6_fib_details_t * mp) -{ - clib_warning ("BUG"); -} - static void send_ip6_fib_details (vpe_api_main_t * am, unix_shared_memory_queue_t * q, @@ -469,24 +422,6 @@ vl_api_ip6_fib_dump_t_handler (vl_api_ip6_fib_dump_t * mp) /* *INDENT-ON* */ } -static void -vl_api_ip_mfib_details_t_handler (vl_api_ip_mfib_details_t * mp) -{ - clib_warning ("BUG"); -} - -static void -vl_api_ip_mfib_details_t_endian (vl_api_ip_mfib_details_t * mp) -{ - clib_warning ("BUG"); -} - -static void -vl_api_ip_mfib_details_t_print (vl_api_ip_mfib_details_t * mp) -{ - clib_warning ("BUG"); -} - static void send_ip_mfib_details (vpe_api_main_t * am, unix_shared_memory_queue_t * q, @@ -591,24 +526,6 @@ vl_api_ip_mfib_dump_t_handler (vl_api_ip_mfib_dump_t * mp) vec_free (api_rpaths); } -static void -vl_api_ip6_mfib_details_t_handler (vl_api_ip6_mfib_details_t * mp) -{ - clib_warning ("BUG"); -} - -static void -vl_api_ip6_mfib_details_t_endian (vl_api_ip6_mfib_details_t * mp) -{ - clib_warning ("BUG"); -} - -static void -vl_api_ip6_mfib_details_t_print (vl_api_ip6_mfib_details_t * mp) -{ - clib_warning ("BUG"); -} - static void send_ip6_mfib_details (vpe_api_main_t * am, unix_shared_memory_queue_t * q, diff --git a/src/vnet/l2/l2_api.c b/src/vnet/l2/l2_api.c index a3cc49bf..a985852c 100644 --- a/src/vnet/l2/l2_api.c +++ b/src/vnet/l2/l2_api.c @@ -48,13 +48,10 @@ _(L2_XCONNECT_DUMP, l2_xconnect_dump) \ _(L2_FIB_CLEAR_TABLE, l2_fib_clear_table) \ _(L2_FIB_TABLE_DUMP, l2_fib_table_dump) \ -_(L2_FIB_TABLE_ENTRY, l2_fib_table_entry) \ _(L2FIB_ADD_DEL, l2fib_add_del) \ _(L2_FLAGS, l2_flags) \ _(BRIDGE_DOMAIN_ADD_DEL, bridge_domain_add_del) \ _(BRIDGE_DOMAIN_DUMP, bridge_domain_dump) \ -_(BRIDGE_DOMAIN_DETAILS, bridge_domain_details) \ -_(BRIDGE_DOMAIN_SW_IF_DETAILS, bridge_domain_sw_if_details) \ _(BRIDGE_FLAGS, bridge_flags) \ _(L2_INTERFACE_VLAN_TAG_REWRITE, l2_interface_vlan_tag_rewrite) \ _(L2_INTERFACE_PBB_TAG_REWRITE, l2_interface_pbb_tag_rewrite) @@ -140,12 +137,6 @@ send_l2fib_table_entry (vpe_api_main_t * am, vl_msg_api_send_shmem (q, (u8 *) & mp); } -static void -vl_api_l2_fib_table_entry_t_handler (vl_api_l2_fib_table_entry_t * mp) -{ - clib_warning ("BUG"); -} - static void vl_api_l2_fib_table_dump_t_handler (vl_api_l2_fib_table_dump_t * mp) { @@ -329,19 +320,6 @@ vl_api_bridge_domain_add_del_t_handler (vl_api_bridge_domain_add_del_t * mp) REPLY_MACRO (VL_API_BRIDGE_DOMAIN_ADD_DEL_REPLY); } -static void -vl_api_bridge_domain_details_t_handler (vl_api_bridge_domain_details_t * mp) -{ - clib_warning ("BUG"); -} - -static void - vl_api_bridge_domain_sw_if_details_t_handler - (vl_api_bridge_domain_sw_if_details_t * mp) -{ - clib_warning ("BUG"); -} - static void send_bridge_domain_details (unix_shared_memory_queue_t * q, l2_bridge_domain_t * bd_config, diff --git a/src/vnet/mpls/mpls_api.c b/src/vnet/mpls/mpls_api.c index ebbeba69..a36a5046 100644 --- a/src/vnet/mpls/mpls_api.c +++ b/src/vnet/mpls/mpls_api.c @@ -50,9 +50,7 @@ _(MPLS_IP_BIND_UNBIND, mpls_ip_bind_unbind) \ _(MPLS_ROUTE_ADD_DEL, mpls_route_add_del) \ _(MPLS_TUNNEL_ADD_DEL, mpls_tunnel_add_del) \ _(MPLS_TUNNEL_DUMP, mpls_tunnel_dump) \ -_(MPLS_TUNNEL_DETAILS, mpls_tunnel_details) \ -_(MPLS_FIB_DUMP, mpls_fib_dump) \ -_(MPLS_FIB_DETAILS, mpls_fib_details) +_(MPLS_FIB_DUMP, mpls_fib_dump) extern void stats_dslock_with_hint (int hint, int tag); extern void stats_dsunlock (void); @@ -280,12 +278,6 @@ vl_api_mpls_tunnel_add_del_t_handler (vl_api_mpls_tunnel_add_del_t * mp) /* *INDENT-ON* */ } -static void -vl_api_mpls_tunnel_details_t_handler (vl_api_mpls_tunnel_details_t * mp) -{ - clib_warning ("BUG"); -} - typedef struct mpls_tunnel_send_walk_ctx_t_ { unix_shared_memory_queue_t *q; @@ -340,24 +332,6 @@ vl_api_mpls_tunnel_dump_t_handler (vl_api_mpls_tunnel_dump_t * mp) mpls_tunnel_walk (send_mpls_tunnel_entry, &ctx); } -static void -vl_api_mpls_fib_details_t_handler (vl_api_mpls_fib_details_t * mp) -{ - clib_warning ("BUG"); -} - -static void -vl_api_mpls_fib_details_t_endian (vl_api_mpls_fib_details_t * mp) -{ - clib_warning ("BUG"); -} - -static void -vl_api_mpls_fib_details_t_print (vl_api_mpls_fib_details_t * mp) -{ - clib_warning ("BUG"); -} - static void send_mpls_fib_details (vpe_api_main_t * am, unix_shared_memory_queue_t * q, diff --git a/src/vpp/api/api.c b/src/vpp/api/api.c index 60fd0199..a8f471e8 100644 --- a/src/vpp/api/api.c +++ b/src/vpp/api/api.c @@ -128,12 +128,8 @@ _(CLASSIFY_SET_INTERFACE_IP_TABLE, classify_set_interface_ip_table) \ _(CLASSIFY_SET_INTERFACE_L2_TABLES, classify_set_interface_l2_tables) \ _(GET_NODE_INDEX, get_node_index) \ _(ADD_NODE_NEXT, add_node_next) \ -_(VXLAN_ADD_DEL_TUNNEL, vxlan_add_del_tunnel) \ -_(VXLAN_TUNNEL_DUMP, vxlan_tunnel_dump) \ _(L2_INTERFACE_EFP_FILTER, l2_interface_efp_filter) \ _(SHOW_VERSION, show_version) \ -_(VXLAN_GPE_ADD_DEL_TUNNEL, vxlan_gpe_add_del_tunnel) \ -_(VXLAN_GPE_TUNNEL_DUMP, vxlan_gpe_tunnel_dump) \ _(INTERFACE_NAME_RENUMBER, interface_name_renumber) \ _(WANT_IP4_ARP_EVENTS, want_ip4_arp_events) \ _(WANT_IP6_ND_EVENTS, want_ip6_nd_events) \ @@ -1436,62 +1432,6 @@ out: /* *INDENT-ON* */ } -static void vl_api_vxlan_add_del_tunnel_t_handler - (vl_api_vxlan_add_del_tunnel_t * mp) -{ - vl_api_vxlan_add_del_tunnel_reply_t *rmp; - int rv = 0; - vnet_vxlan_add_del_tunnel_args_t _a, *a = &_a; - u32 encap_fib_index; - uword *p; - ip4_main_t *im = &ip4_main; - vnet_main_t *vnm = vnet_get_main (); - u32 sw_if_index = ~0; - - p = hash_get (im->fib_index_by_table_id, ntohl (mp->encap_vrf_id)); - if (!p) - { - rv = VNET_API_ERROR_NO_SUCH_FIB; - goto out; - } - encap_fib_index = p[0]; - memset (a, 0, sizeof (*a)); - - a->is_add = mp->is_add; - a->is_ip6 = mp->is_ipv6; - - /* ip addresses sent in network byte order */ - ip46_from_addr_buf (mp->is_ipv6, mp->dst_address, &a->dst); - ip46_from_addr_buf (mp->is_ipv6, mp->src_address, &a->src); - - /* Check src & dst are different */ - if (ip46_address_cmp (&a->dst, &a->src) == 0) - { - rv = VNET_API_ERROR_SAME_SRC_DST; - goto out; - } - a->mcast_sw_if_index = ntohl (mp->mcast_sw_if_index); - if (ip46_address_is_multicast (&a->dst) && - pool_is_free_index (vnm->interface_main.sw_interfaces, - a->mcast_sw_if_index)) - { - rv = VNET_API_ERROR_INVALID_SW_IF_INDEX; - goto out; - } - a->encap_fib_index = encap_fib_index; - a->decap_next_index = ntohl (mp->decap_next_index); - a->vni = ntohl (mp->vni); - rv = vnet_vxlan_add_del_tunnel (a, &sw_if_index); - -out: - /* *INDENT-OFF* */ - REPLY_MACRO2(VL_API_VXLAN_ADD_DEL_TUNNEL_REPLY, - ({ - rmp->sw_if_index = ntohl (sw_if_index); - })); - /* *INDENT-ON* */ -} - static void send_vxlan_tunnel_details (vxlan_tunnel_t * t, unix_shared_memory_queue_t * q, u32 context) { @@ -1525,43 +1465,6 @@ static void send_vxlan_tunnel_details vl_msg_api_send_shmem (q, (u8 *) & rmp); } -static void vl_api_vxlan_tunnel_dump_t_handler - (vl_api_vxlan_tunnel_dump_t * mp) -{ - unix_shared_memory_queue_t *q; - vxlan_main_t *vxm = &vxlan_main; - vxlan_tunnel_t *t; - u32 sw_if_index; - - q = vl_api_client_index_to_input_queue (mp->client_index); - if (q == 0) - { - return; - } - - sw_if_index = ntohl (mp->sw_if_index); - - if (~0 == sw_if_index) - { - /* *INDENT-OFF* */ - pool_foreach (t, vxm->tunnels, - ({ - send_vxlan_tunnel_details(t, q, mp->context); - })); - /* *INDENT-ON* */ - } - else - { - if ((sw_if_index >= vec_len (vxm->tunnel_index_by_sw_if_index)) || - (~0 == vxm->tunnel_index_by_sw_if_index[sw_if_index])) - { - return; - } - t = &vxm->tunnels[vxm->tunnel_index_by_sw_if_index[sw_if_index]]; - send_vxlan_tunnel_details (t, q, mp->context); - } -} - static void vl_api_l2_patch_add_del_t_handler (vl_api_l2_patch_add_del_t * mp) { @@ -1585,83 +1488,6 @@ vl_api_l2_patch_add_del_t_handler (vl_api_l2_patch_add_del_t * mp) REPLY_MACRO (VL_API_L2_PATCH_ADD_DEL_REPLY); } -static void - vl_api_vxlan_gpe_add_del_tunnel_t_handler - (vl_api_vxlan_gpe_add_del_tunnel_t * mp) -{ - vl_api_vxlan_gpe_add_del_tunnel_reply_t *rmp; - int rv = 0; - vnet_vxlan_gpe_add_del_tunnel_args_t _a, *a = &_a; - u32 encap_fib_index, decap_fib_index; - u8 protocol; - uword *p; - ip4_main_t *im = &ip4_main; - u32 sw_if_index = ~0; - - - p = hash_get (im->fib_index_by_table_id, ntohl (mp->encap_vrf_id)); - if (!p) - { - rv = VNET_API_ERROR_NO_SUCH_FIB; - goto out; - } - encap_fib_index = p[0]; - - protocol = mp->protocol; - - /* Interpret decap_vrf_id as an opaque if sending to other-than-ip4-input */ - if (protocol == VXLAN_GPE_INPUT_NEXT_IP4_INPUT) - { - p = hash_get (im->fib_index_by_table_id, ntohl (mp->decap_vrf_id)); - if (!p) - { - rv = VNET_API_ERROR_NO_SUCH_INNER_FIB; - goto out; - } - decap_fib_index = p[0]; - } - else - { - decap_fib_index = ntohl (mp->decap_vrf_id); - } - - /* Check src & dst are different */ - if ((mp->is_ipv6 && memcmp (mp->local, mp->remote, 16) == 0) || - (!mp->is_ipv6 && memcmp (mp->local, mp->remote, 4) == 0)) - { - rv = VNET_API_ERROR_SAME_SRC_DST; - goto out; - } - memset (a, 0, sizeof (*a)); - - a->is_add = mp->is_add; - a->is_ip6 = mp->is_ipv6; - /* ip addresses sent in network byte order */ - if (a->is_ip6) - { - clib_memcpy (&(a->local.ip6), mp->local, 16); - clib_memcpy (&(a->remote.ip6), mp->remote, 16); - } - else - { - clib_memcpy (&(a->local.ip4), mp->local, 4); - clib_memcpy (&(a->remote.ip4), mp->remote, 4); - } - a->encap_fib_index = encap_fib_index; - a->decap_fib_index = decap_fib_index; - a->protocol = protocol; - a->vni = ntohl (mp->vni); - rv = vnet_vxlan_gpe_add_del_tunnel (a, &sw_if_index); - -out: - /* *INDENT-OFF* */ - REPLY_MACRO2(VL_API_VXLAN_GPE_ADD_DEL_TUNNEL_REPLY, - ({ - rmp->sw_if_index = ntohl (sw_if_index); - })); - /* *INDENT-ON* */ -} - static void send_vxlan_gpe_tunnel_details (vxlan_gpe_tunnel_t * t, unix_shared_memory_queue_t * q, u32 context) { @@ -1696,43 +1522,6 @@ static void send_vxlan_gpe_tunnel_details vl_msg_api_send_shmem (q, (u8 *) & rmp); } -static void vl_api_vxlan_gpe_tunnel_dump_t_handler - (vl_api_vxlan_gpe_tunnel_dump_t * mp) -{ - unix_shared_memory_queue_t *q; - vxlan_gpe_main_t *vgm = &vxlan_gpe_main; - vxlan_gpe_tunnel_t *t; - u32 sw_if_index; - - q = vl_api_client_index_to_input_queue (mp->client_index); - if (q == 0) - { - return; - } - - sw_if_index = ntohl (mp->sw_if_index); - - if (~0 == sw_if_index) - { - /* *INDENT-OFF* */ - pool_foreach (t, vgm->tunnels, - ({ - send_vxlan_gpe_tunnel_details(t, q, mp->context); - })); - /* *INDENT-ON* */ - } - else - { - if ((sw_if_index >= vec_len (vgm->tunnel_index_by_sw_if_index)) || - (~0 == vgm->tunnel_index_by_sw_if_index[sw_if_index])) - { - return; - } - t = &vgm->tunnels[vgm->tunnel_index_by_sw_if_index[sw_if_index]]; - send_vxlan_gpe_tunnel_details (t, q, mp->context); - } -} - static void vl_api_interface_name_renumber_t_handler (vl_api_interface_name_renumber_t * mp) diff --git a/src/vpp/api/api_main.c b/src/vpp/api/api_main.c index 97b501e0..6ae510b1 100644 --- a/src/vpp/api/api_main.c +++ b/src/vpp/api/api_main.c @@ -48,7 +48,6 @@ api_main_init (vlib_main_t * vm) vam->vlib_main = vm; vam->my_client_index = (u32) ~ 0; init_error_string_table (vam); - vat_api_hookup (vam); rv = vat_plugin_init (vam); if (rv) clib_warning ("vat_plugin_init returned %d", rv); diff --git a/src/vpp/stats/stats.c b/src/vpp/stats/stats.c index 5e9b0d69..c46d441a 100644 --- a/src/vpp/stats/stats.c +++ b/src/vpp/stats/stats.c @@ -46,7 +46,6 @@ stats_main_t stats_main; #define foreach_stats_msg \ _(WANT_STATS, want_stats) \ -_(WANT_STATS_REPLY, want_stats_reply) \ _(VNET_INTERFACE_COUNTERS, vnet_interface_counters) \ _(VNET_IP4_FIB_COUNTERS, vnet_ip4_fib_counters) \ _(VNET_IP6_FIB_COUNTERS, vnet_ip6_fib_counters) \ @@ -1226,12 +1225,6 @@ vl_api_vnet_ip6_nbr_counters_t_handler (vl_api_vnet_ip6_nbr_counters_t * mp) } } -static void -vl_api_want_stats_reply_t_handler (vl_api_want_stats_reply_t * mp) -{ - clib_warning ("BUG"); -} - static void vl_api_want_stats_t_handler (vl_api_want_stats_t * mp) { -- cgit 1.2.3-korg From 9a69a6095f67b8979a02f128f44e449889454273 Mon Sep 17 00:00:00 2001 From: Neale Ranns Date: Sun, 26 Mar 2017 10:56:33 -0700 Subject: Sub-net broadcast addresses for IPv4 Change-Id: Ib2189d01e8bc61de57404159690fb70f89c47277 Signed-off-by: Neale Ranns --- src/vnet/classify/vnet_classify.c | 2 +- src/vnet/fib/fib_entry_src_interface.c | 21 ++++++ src/vnet/ip/ip4_forward.c | 73 ++++++++++++++++++- test/framework.py | 2 +- test/test_ip4.py | 125 ++++++++++++++++++++++++++++++++- 5 files changed, 217 insertions(+), 6 deletions(-) (limited to 'src/vnet/classify') diff --git a/src/vnet/classify/vnet_classify.c b/src/vnet/classify/vnet_classify.c index b651a1f1..98842a48 100644 --- a/src/vnet/classify/vnet_classify.c +++ b/src/vnet/classify/vnet_classify.c @@ -1499,7 +1499,7 @@ static u8 * format_vnet_classify_table (u8 * s, va_list * args) s = format (s, "%10u%10d%10d%10d", index, t->active_elements, t->next_table_index, t->miss_next_index); - s = format (s, "\n Heap: %U", format_mheap, t->mheap, 0 /*verbose*/); + s = format (s, "\n Heap: %U", format_mheap, t->mheap, 0 /*verbose*/); s = format (s, "\n nbuckets %d, skip %d match %d flag %d offset %d", t->nbuckets, t->skip_n_vectors, t->match_n_vectors, diff --git a/src/vnet/fib/fib_entry_src_interface.c b/src/vnet/fib/fib_entry_src_interface.c index ca04716e..bb87818f 100644 --- a/src/vnet/fib/fib_entry_src_interface.c +++ b/src/vnet/fib/fib_entry_src_interface.c @@ -31,6 +31,25 @@ fib_entry_src_interface_init (fib_entry_src_t *src) src->interface.fesi_sibling = FIB_NODE_INDEX_INVALID; } +static void +fib_entry_src_interface_add (fib_entry_src_t *src, + const fib_entry_t *entry, + fib_entry_flag_t flags, + fib_protocol_t proto, + const dpo_id_t *dpo) +{ + src->fes_pl = fib_path_list_create_special( + proto, + fib_entry_src_flags_2_path_list_flags(flags), + dpo); +} + +static void +fib_entry_src_interface_remove (fib_entry_src_t *src) +{ + src->fes_pl = FIB_NODE_INDEX_INVALID; +} + static void fib_entry_src_interface_path_swap (fib_entry_src_t *src, const fib_entry_t *entry, @@ -176,6 +195,8 @@ fib_entry_src_interface_format (fib_entry_src_t *src, const static fib_entry_src_vft_t interface_src_vft = { .fesv_init = fib_entry_src_interface_init, + .fesv_add = fib_entry_src_interface_add, + .fesv_remove = fib_entry_src_interface_remove, .fesv_path_swap = fib_entry_src_interface_path_swap, .fesv_activate = fib_entry_src_interface_activate, .fesv_deactivate = fib_entry_src_interface_deactivate, diff --git a/src/vnet/ip/ip4_forward.c b/src/vnet/ip/ip4_forward.c index f2bc2a71..bbba4b70 100644 --- a/src/vnet/ip/ip4_forward.c +++ b/src/vnet/ip/ip4_forward.c @@ -745,8 +745,9 @@ ip4_add_interface_routes (u32 sw_if_index, a->neighbor_probe_adj_index = ~0; - if (pfx.fp_len < 32) + if (pfx.fp_len <= 30) { + /* a /30 or shorter - add a glean for the network address */ fib_node_index_t fei; fei = fib_table_entry_update_one_path (fib_index, &pfx, @@ -764,8 +765,50 @@ ip4_add_interface_routes (u32 sw_if_index, NULL, FIB_ROUTE_PATH_FLAG_NONE); a->neighbor_probe_adj_index = fib_entry_get_adj (fei); - } + /* Add the two broadcast addresses as drop */ + fib_prefix_t net_pfx = { + .fp_len = 32, + .fp_proto = FIB_PROTOCOL_IP4, + .fp_addr.ip4.as_u32 = address->as_u32 & im->fib_masks[pfx.fp_len], + }; + if (net_pfx.fp_addr.ip4.as_u32 != pfx.fp_addr.ip4.as_u32) + fib_table_entry_special_add(fib_index, + &net_pfx, + FIB_SOURCE_INTERFACE, + (FIB_ENTRY_FLAG_DROP | + FIB_ENTRY_FLAG_LOOSE_URPF_EXEMPT), + ADJ_INDEX_INVALID); + net_pfx.fp_addr.ip4.as_u32 |= ~im->fib_masks[pfx.fp_len]; + if (net_pfx.fp_addr.ip4.as_u32 != pfx.fp_addr.ip4.as_u32) + fib_table_entry_special_add(fib_index, + &net_pfx, + FIB_SOURCE_INTERFACE, + (FIB_ENTRY_FLAG_DROP | + FIB_ENTRY_FLAG_LOOSE_URPF_EXEMPT), + ADJ_INDEX_INVALID); + } + else if (pfx.fp_len == 31) + { + u32 mask = clib_host_to_net_u32(1); + fib_prefix_t net_pfx = pfx; + + net_pfx.fp_len = 32; + net_pfx.fp_addr.ip4.as_u32 ^= mask; + + /* a /31 - add the other end as an attached host */ + fib_table_entry_update_one_path (fib_index, &net_pfx, + FIB_SOURCE_INTERFACE, + (FIB_ENTRY_FLAG_ATTACHED), + FIB_PROTOCOL_IP4, + &net_pfx.fp_addr, + sw_if_index, + // invalid FIB index + ~0, + 1, + NULL, + FIB_ROUTE_PATH_FLAG_NONE); + } pfx.fp_len = 32; if (sw_if_index < vec_len (lm->classify_table_index_by_sw_if_index)) @@ -813,10 +856,34 @@ ip4_del_interface_routes (ip4_main_t * im, .fp_addr.ip4 = *address, }; - if (pfx.fp_len < 32) + if (pfx.fp_len <= 30) { + fib_prefix_t net_pfx = { + .fp_len = 32, + .fp_proto = FIB_PROTOCOL_IP4, + .fp_addr.ip4.as_u32 = address->as_u32 & im->fib_masks[pfx.fp_len], + }; + if (net_pfx.fp_addr.ip4.as_u32 != pfx.fp_addr.ip4.as_u32) + fib_table_entry_special_remove(fib_index, + &net_pfx, + FIB_SOURCE_INTERFACE); + net_pfx.fp_addr.ip4.as_u32 |= ~im->fib_masks[pfx.fp_len]; + if (net_pfx.fp_addr.ip4.as_u32 != pfx.fp_addr.ip4.as_u32) + fib_table_entry_special_remove(fib_index, + &net_pfx, + FIB_SOURCE_INTERFACE); fib_table_entry_delete (fib_index, &pfx, FIB_SOURCE_INTERFACE); } + else if (pfx.fp_len == 31) + { + u32 mask = clib_host_to_net_u32(1); + fib_prefix_t net_pfx = pfx; + + net_pfx.fp_len = 32; + net_pfx.fp_addr.ip4.as_u32 ^= mask; + + fib_table_entry_delete (fib_index, &net_pfx, FIB_SOURCE_INTERFACE); + } pfx.fp_len = 32; fib_table_entry_delete (fib_index, &pfx, FIB_SOURCE_INTERFACE); diff --git a/test/framework.py b/test/framework.py index eb3a9d7e..fbd21d23 100644 --- a/test/framework.py +++ b/test/framework.py @@ -359,7 +359,7 @@ class VppTestCase(unittest.TestCase): self._testMethodDoc)) if not self.vpp_dead: self.logger.debug(self.vapi.cli("show trace")) - self.logger.info(self.vapi.ppcli("show int")) + self.logger.info(self.vapi.ppcli("show interfaces")) self.logger.info(self.vapi.ppcli("show hardware")) self.logger.info(self.vapi.ppcli("show error")) self.logger.info(self.vapi.ppcli("show run")) diff --git a/test/test_ip4.py b/test/test_ip4.py index 79af5492..ed364b62 100644 --- a/test/test_ip4.py +++ b/test/test_ip4.py @@ -9,7 +9,7 @@ from vpp_ip_route import VppIpRoute, VppRoutePath, VppIpMRoute, \ VppMRoutePath, MRouteItfFlags, MRouteEntryFlags from scapy.packet import Raw -from scapy.layers.l2 import Ether, Dot1Q +from scapy.layers.l2 import Ether, Dot1Q, ARP from scapy.layers.inet import IP, UDP, ICMP, icmptypes, icmpcodes from util import ppp @@ -643,5 +643,128 @@ class TestIPDisabled(VppTestCase): self.send_and_assert_no_replies(self.pg1, pm, "IP disabled") +class TestIPSubNets(VppTestCase): + """ IPv4 Subnets """ + + def setUp(self): + super(TestIPSubNets, self).setUp() + + # create a 2 pg interfaces + self.create_pg_interfaces(range(2)) + + # pg0 we will use to experiemnt + self.pg0.admin_up() + + # pg1 is setup normally + self.pg1.admin_up() + self.pg1.config_ip4() + self.pg1.resolve_arp() + + def tearDown(self): + super(TestIPSubNets, self).tearDown() + for i in self.pg_interfaces: + i.admin_down() + + def send_and_assert_no_replies(self, intf, pkts, remark): + intf.add_stream(pkts) + self.pg_enable_capture(self.pg_interfaces) + self.pg_start() + for i in self.pg_interfaces: + i.get_capture(0) + i.assert_nothing_captured(remark=remark) + + def test_ip_sub_nets(self): + """ IP Sub Nets """ + + # + # Configure a covering route to forward so we know + # when we are dropping + # + cover_route = VppIpRoute(self, "10.0.0.0", 8, + [VppRoutePath(self.pg1.remote_ip4, + self.pg1.sw_if_index)]) + cover_route.add_vpp_config() + + p = (Ether(src=self.pg1.remote_mac, + dst=self.pg1.local_mac) / + IP(dst="10.10.10.10", src=self.pg0.local_ip4) / + UDP(sport=1234, dport=1234) / + Raw('\xa5' * 100)) + + self.pg1.add_stream(p) + self.pg_enable_capture(self.pg_interfaces) + self.pg_start() + rx = self.pg1.get_capture(1) + + # + # Configure some non-/24 subnets on an IP interface + # + ip_addr_n = socket.inet_pton(socket.AF_INET, "10.10.10.10") + + self.vapi.sw_interface_add_del_address(self.pg0.sw_if_index, + ip_addr_n, + 16) + + pn = (Ether(src=self.pg1.remote_mac, + dst=self.pg1.local_mac) / + IP(dst="10.10.0.0", src=self.pg0.local_ip4) / + UDP(sport=1234, dport=1234) / + Raw('\xa5' * 100)) + pb = (Ether(src=self.pg1.remote_mac, + dst=self.pg1.local_mac) / + IP(dst="10.10.255.255", src=self.pg0.local_ip4) / + UDP(sport=1234, dport=1234) / + Raw('\xa5' * 100)) + + self.send_and_assert_no_replies(self.pg1, pn, "IP Network address") + self.send_and_assert_no_replies(self.pg1, pb, "IP Broadcast address") + + # remove the sub-net and we are forwarding via the cover again + self.vapi.sw_interface_add_del_address(self.pg0.sw_if_index, + ip_addr_n, + 16, + is_add=0) + self.pg1.add_stream(pn) + self.pg_enable_capture(self.pg_interfaces) + self.pg_start() + rx = self.pg1.get_capture(1) + self.pg1.add_stream(pb) + self.pg_enable_capture(self.pg_interfaces) + self.pg_start() + rx = self.pg1.get_capture(1) + + # + # A /31 is a special case where the 'other-side' is an attached host + # packets to that peer generate ARP requests + # + ip_addr_n = socket.inet_pton(socket.AF_INET, "10.10.10.10") + + self.vapi.sw_interface_add_del_address(self.pg0.sw_if_index, + ip_addr_n, + 31) + + pn = (Ether(src=self.pg1.remote_mac, + dst=self.pg1.local_mac) / + IP(dst="10.10.10.11", src=self.pg0.local_ip4) / + UDP(sport=1234, dport=1234) / + Raw('\xa5' * 100)) + + self.pg1.add_stream(pn) + self.pg_enable_capture(self.pg_interfaces) + self.pg_start() + rx = self.pg0.get_capture(1) + rx[ARP] + + # remove the sub-net and we are forwarding via the cover again + self.vapi.sw_interface_add_del_address(self.pg0.sw_if_index, + ip_addr_n, + 31, + is_add=0) + self.pg1.add_stream(pn) + self.pg_enable_capture(self.pg_interfaces) + self.pg_start() + rx = self.pg1.get_capture(1) + + if __name__ == '__main__': unittest.main(testRunner=VppTestRunner) -- cgit 1.2.3-korg From 586afd762bfa149f5ca167bd5fd5a0cd59ce94fe Mon Sep 17 00:00:00 2001 From: Damjan Marion Date: Wed, 5 Apr 2017 19:18:20 +0200 Subject: Use thread local storage for thread index This patch deprecates stack-based thread identification, Also removes requirement that thread stacks are adjacent. Finally, possibly annoying for some folks, it renames all occurences of cpu_index and cpu_number with thread index. Using word "cpu" is misleading here as thread can be migrated ti different CPU, and also it is not related to linux cpu index. Change-Id: I68cdaf661e701d2336fc953dcb9978d10a70f7c1 Signed-off-by: Damjan Marion --- src/examples/srv6-sample-localsid/node.c | 4 +- src/plugins/dpdk/buffer.c | 2 +- src/plugins/dpdk/device/device.c | 8 +- src/plugins/dpdk/device/dpdk_priv.h | 8 +- src/plugins/dpdk/device/init.c | 2 +- src/plugins/dpdk/device/node.c | 32 +++--- src/plugins/dpdk/hqos/hqos.c | 16 +-- src/plugins/dpdk/ipsec/cli.c | 8 +- src/plugins/dpdk/ipsec/crypto_node.c | 4 +- src/plugins/dpdk/ipsec/esp.h | 4 +- src/plugins/dpdk/ipsec/esp_decrypt.c | 4 +- src/plugins/dpdk/ipsec/esp_encrypt.c | 5 +- src/plugins/dpdk/ipsec/ipsec.c | 2 +- src/plugins/dpdk/ipsec/ipsec.h | 4 +- src/plugins/dpdk/main.c | 2 +- src/plugins/flowperpkt/l2_node.c | 2 +- src/plugins/flowperpkt/node.c | 2 +- src/plugins/ioam/export-common/ioam_export.h | 6 +- .../ioam/ip6/ioam_cache_tunnel_select_node.c | 16 +-- src/plugins/ixge/ixge.c | 2 +- src/plugins/lb/lb.c | 8 +- src/plugins/lb/node.c | 22 ++-- src/plugins/lb/refcount.c | 8 +- src/plugins/lb/refcount.h | 4 +- src/plugins/memif/node.c | 35 +++--- src/plugins/snat/in2out.c | 110 +++++++++--------- src/plugins/snat/out2in.c | 102 ++++++++--------- src/plugins/snat/snat.h | 10 +- src/vlib/buffer.c | 6 +- src/vlib/buffer_funcs.h | 4 +- src/vlib/cli.c | 6 +- src/vlib/counter.h | 16 +-- src/vlib/error.c | 2 +- src/vlib/global_funcs.h | 2 +- src/vlib/main.c | 14 +-- src/vlib/main.h | 2 +- src/vlib/node.c | 2 +- src/vlib/node.h | 6 +- src/vlib/node_funcs.h | 8 +- src/vlib/threads.c | 69 ++++------- src/vlib/threads.h | 21 ++-- src/vlib/unix/cj.c | 7 +- src/vlib/unix/cj.h | 2 +- src/vlib/unix/main.c | 43 +++---- src/vnet/adj/adj_l2.c | 4 +- src/vnet/adj/adj_midchain.c | 8 +- src/vnet/adj/adj_nsh.c | 4 +- src/vnet/classify/vnet_classify.c | 16 +-- src/vnet/cop/ip4_whitelist.c | 8 +- src/vnet/cop/ip6_whitelist.c | 8 +- src/vnet/devices/af_packet/node.c | 20 ++-- src/vnet/devices/devices.c | 61 +++++----- src/vnet/devices/devices.h | 18 +-- src/vnet/devices/netmap/node.c | 24 ++-- src/vnet/devices/ssvm/node.c | 6 +- src/vnet/devices/virtio/vhost-user.c | 127 +++++++++++---------- src/vnet/dpo/lookup_dpo.c | 20 ++-- src/vnet/dpo/replicate_dpo.c | 12 +- src/vnet/ethernet/arp.c | 2 +- src/vnet/ethernet/interface.c | 7 +- src/vnet/ethernet/node.c | 14 +-- src/vnet/gre/node.c | 8 +- src/vnet/interface.h | 2 +- src/vnet/interface_output.c | 53 ++++----- src/vnet/ip/ip4_forward.c | 34 +++--- src/vnet/ip/ip4_input.c | 8 +- src/vnet/ip/ip6_forward.c | 24 ++-- src/vnet/ip/ip6_input.c | 8 +- src/vnet/ip/ip6_neighbor.c | 4 +- src/vnet/ipsec/esp.h | 8 +- src/vnet/ipsec/esp_decrypt.c | 13 ++- src/vnet/ipsec/esp_encrypt.c | 13 ++- src/vnet/ipsec/ikev2.c | 64 ++++++----- src/vnet/ipsec/ipsec.h | 12 +- src/vnet/ipsec/ipsec_if.c | 2 +- src/vnet/l2/l2_bvi.h | 2 +- src/vnet/l2/l2_input.c | 14 +-- src/vnet/l2/l2_output.c | 6 +- src/vnet/l2tp/decap.c | 2 +- src/vnet/l2tp/encap.c | 2 +- src/vnet/l2tp/l2tp.c | 6 +- src/vnet/lisp-gpe/decap.c | 16 +-- src/vnet/lldp/lldp_input.c | 2 +- src/vnet/map/ip4_map.c | 14 +-- src/vnet/map/ip4_map_t.c | 12 +- src/vnet/map/ip6_map.c | 19 +-- src/vnet/map/ip6_map_t.c | 12 +- src/vnet/mpls/mpls_input.c | 8 +- src/vnet/mpls/mpls_lookup.c | 20 ++-- src/vnet/mpls/mpls_output.c | 10 +- src/vnet/pg/input.c | 4 +- src/vnet/replication.c | 20 ++-- src/vnet/replication.h | 2 +- src/vnet/session/node.c | 2 +- src/vnet/sr/sr_localsid.c | 44 +++---- src/vnet/tcp/builtin_client.c | 2 +- src/vnet/tcp/tcp.c | 8 +- src/vnet/tcp/tcp_debug.h | 2 +- src/vnet/tcp/tcp_input.c | 10 +- src/vnet/tcp/tcp_output.c | 20 ++-- src/vnet/udp/udp_input.c | 2 +- src/vnet/unix/tapcli.c | 2 +- src/vnet/unix/tuntap.c | 4 +- src/vnet/vxlan-gpe/decap.c | 10 +- src/vnet/vxlan-gpe/encap.c | 12 +- src/vnet/vxlan/decap.c | 10 +- src/vnet/vxlan/encap.c | 12 +- src/vpp/stats/stats.c | 14 +-- src/vpp/stats/stats.h | 2 +- 109 files changed, 790 insertions(+), 791 deletions(-) (limited to 'src/vnet/classify') diff --git a/src/examples/srv6-sample-localsid/node.c b/src/examples/srv6-sample-localsid/node.c index 7bae9cd7..e83e2352 100644 --- a/src/examples/srv6-sample-localsid/node.c +++ b/src/examples/srv6-sample-localsid/node.c @@ -114,7 +114,7 @@ srv6_localsid_sample_fn (vlib_main_t * vm, vlib_node_runtime_t * node, vlib_fram from = vlib_frame_vector_args (frame); n_left_from = frame->n_vectors; next_index = node->cached_next_index; - u32 cpu_index = os_get_cpu_number (); + u32 thread_index = vlib_get_thread_index (); while (n_left_from > 0) { @@ -168,7 +168,7 @@ srv6_localsid_sample_fn (vlib_main_t * vm, vlib_node_runtime_t * node, vlib_fram /* This increments the SRv6 per LocalSID counters.*/ vlib_increment_combined_counter (((next0 == SRV6_SAMPLE_LOCALSID_NEXT_ERROR) ? &(sm->sr_ls_invalid_counters) : &(sm->sr_ls_valid_counters)), - cpu_index, + thread_index, ls0 - sm->localsids, 1, vlib_buffer_length_in_chain (vm, b0)); diff --git a/src/plugins/dpdk/buffer.c b/src/plugins/dpdk/buffer.c index 2765c292..c80b3fa8 100644 --- a/src/plugins/dpdk/buffer.c +++ b/src/plugins/dpdk/buffer.c @@ -132,7 +132,7 @@ dpdk_buffer_delete_free_list (vlib_main_t * vm, u32 free_list_index) u32 merge_index; int i; - ASSERT (os_get_cpu_number () == 0); + ASSERT (vlib_get_thread_index () == 0); f = vlib_buffer_get_free_list (vm, free_list_index); diff --git a/src/plugins/dpdk/device/device.c b/src/plugins/dpdk/device/device.c index 50b26689..91661246 100644 --- a/src/plugins/dpdk/device/device.c +++ b/src/plugins/dpdk/device/device.c @@ -243,7 +243,7 @@ static_always_inline ASSERT (ring->tx_tail == 0); n_retry = 16; - queue_id = vm->cpu_index; + queue_id = vm->thread_index; do { @@ -266,7 +266,7 @@ static_always_inline { /* no wrap, transmit in one burst */ dpdk_device_hqos_per_worker_thread_t *hqos = - &xd->hqos_wt[vm->cpu_index]; + &xd->hqos_wt[vm->thread_index]; ASSERT (hqos->swq != NULL); @@ -332,7 +332,7 @@ dpdk_buffer_recycle (vlib_main_t * vm, vlib_node_runtime_t * node, vlib_buffer_t * b, u32 bi, struct rte_mbuf **mbp) { dpdk_main_t *dm = &dpdk_main; - u32 my_cpu = vm->cpu_index; + u32 my_cpu = vm->thread_index; struct rte_mbuf *mb_new; if (PREDICT_FALSE (b->flags & VLIB_BUFFER_RECYCLE) == 0) @@ -376,7 +376,7 @@ dpdk_interface_tx (vlib_main_t * vm, tx_ring_hdr_t *ring; u32 n_on_ring; - my_cpu = vm->cpu_index; + my_cpu = vm->thread_index; queue_id = my_cpu; diff --git a/src/plugins/dpdk/device/dpdk_priv.h b/src/plugins/dpdk/device/dpdk_priv.h index dd40ff48..52b4ca4b 100644 --- a/src/plugins/dpdk/device/dpdk_priv.h +++ b/src/plugins/dpdk/device/dpdk_priv.h @@ -79,7 +79,7 @@ dpdk_update_counters (dpdk_device_t * xd, f64 now) { vlib_simple_counter_main_t *cm; vnet_main_t *vnm = vnet_get_main (); - u32 my_cpu = os_get_cpu_number (); + u32 thread_index = vlib_get_thread_index (); u64 rxerrors, last_rxerrors; /* only update counters for PMD interfaces */ @@ -96,7 +96,7 @@ dpdk_update_counters (dpdk_device_t * xd, f64 now) cm = vec_elt_at_index (vnm->interface_main.sw_if_counters, VNET_INTERFACE_COUNTER_RX_NO_BUF); - vlib_increment_simple_counter (cm, my_cpu, xd->vlib_sw_if_index, + vlib_increment_simple_counter (cm, thread_index, xd->vlib_sw_if_index, xd->stats.rx_nombuf - xd->last_stats.rx_nombuf); } @@ -107,7 +107,7 @@ dpdk_update_counters (dpdk_device_t * xd, f64 now) cm = vec_elt_at_index (vnm->interface_main.sw_if_counters, VNET_INTERFACE_COUNTER_RX_MISS); - vlib_increment_simple_counter (cm, my_cpu, xd->vlib_sw_if_index, + vlib_increment_simple_counter (cm, thread_index, xd->vlib_sw_if_index, xd->stats.imissed - xd->last_stats.imissed); } @@ -119,7 +119,7 @@ dpdk_update_counters (dpdk_device_t * xd, f64 now) cm = vec_elt_at_index (vnm->interface_main.sw_if_counters, VNET_INTERFACE_COUNTER_RX_ERROR); - vlib_increment_simple_counter (cm, my_cpu, xd->vlib_sw_if_index, + vlib_increment_simple_counter (cm, thread_index, xd->vlib_sw_if_index, rxerrors - last_rxerrors); } diff --git a/src/plugins/dpdk/device/init.c b/src/plugins/dpdk/device/init.c index 538db6cb..7eaf8da7 100755 --- a/src/plugins/dpdk/device/init.c +++ b/src/plugins/dpdk/device/init.c @@ -324,7 +324,7 @@ dpdk_port_setup (dpdk_main_t * dm, dpdk_device_t * xd) int rv; int j; - ASSERT (os_get_cpu_number () == 0); + ASSERT (vlib_get_thread_index () == 0); if (xd->flags & DPDK_DEVICE_FLAG_ADMIN_UP) { diff --git a/src/plugins/dpdk/device/node.c b/src/plugins/dpdk/device/node.c index e740fd18..b10e0fad 100644 --- a/src/plugins/dpdk/device/node.c +++ b/src/plugins/dpdk/device/node.c @@ -283,7 +283,7 @@ dpdk_buffer_init_from_template (void *d0, void *d1, void *d2, void *d3, */ static_always_inline u32 dpdk_device_input (dpdk_main_t * dm, dpdk_device_t * xd, - vlib_node_runtime_t * node, u32 cpu_index, u16 queue_id, + vlib_node_runtime_t * node, u32 thread_index, u16 queue_id, int maybe_multiseg) { u32 n_buffers; @@ -294,7 +294,7 @@ dpdk_device_input (dpdk_main_t * dm, dpdk_device_t * xd, uword n_rx_bytes = 0; u32 n_trace, trace_cnt __attribute__ ((unused)); vlib_buffer_free_list_t *fl; - vlib_buffer_t *bt = vec_elt_at_index (dm->buffer_templates, cpu_index); + vlib_buffer_t *bt = vec_elt_at_index (dm->buffer_templates, thread_index); if ((xd->flags & DPDK_DEVICE_FLAG_ADMIN_UP) == 0) return 0; @@ -306,7 +306,7 @@ dpdk_device_input (dpdk_main_t * dm, dpdk_device_t * xd, return 0; } - vec_reset_length (xd->d_trace_buffers[cpu_index]); + vec_reset_length (xd->d_trace_buffers[thread_index]); trace_cnt = n_trace = vlib_get_trace_count (vm, node); if (n_trace > 0) @@ -318,7 +318,7 @@ dpdk_device_input (dpdk_main_t * dm, dpdk_device_t * xd, { struct rte_mbuf *mb = xd->rx_vectors[queue_id][mb_index++]; vlib_buffer_t *b = vlib_buffer_from_rte_mbuf (mb); - vec_add1 (xd->d_trace_buffers[cpu_index], + vec_add1 (xd->d_trace_buffers[thread_index], vlib_get_buffer_index (vm, b)); } } @@ -546,20 +546,22 @@ dpdk_device_input (dpdk_main_t * dm, dpdk_device_t * xd, vlib_put_next_frame (vm, node, next_index, n_left_to_next); } - if (PREDICT_FALSE (vec_len (xd->d_trace_buffers[cpu_index]) > 0)) + if (PREDICT_FALSE (vec_len (xd->d_trace_buffers[thread_index]) > 0)) { - dpdk_rx_trace (dm, node, xd, queue_id, xd->d_trace_buffers[cpu_index], - vec_len (xd->d_trace_buffers[cpu_index])); - vlib_set_trace_count (vm, node, n_trace - - vec_len (xd->d_trace_buffers[cpu_index])); + dpdk_rx_trace (dm, node, xd, queue_id, + xd->d_trace_buffers[thread_index], + vec_len (xd->d_trace_buffers[thread_index])); + vlib_set_trace_count (vm, node, + n_trace - + vec_len (xd->d_trace_buffers[thread_index])); } vlib_increment_combined_counter (vnet_get_main ()->interface_main.combined_sw_if_counters + VNET_INTERFACE_COUNTER_RX, - cpu_index, xd->vlib_sw_if_index, mb_index, n_rx_bytes); + thread_index, xd->vlib_sw_if_index, mb_index, n_rx_bytes); - vnet_device_increment_rx_packets (cpu_index, mb_index); + vnet_device_increment_rx_packets (thread_index, mb_index); return mb_index; } @@ -630,19 +632,19 @@ dpdk_input (vlib_main_t * vm, vlib_node_runtime_t * node, vlib_frame_t * f) dpdk_device_t *xd; uword n_rx_packets = 0; dpdk_device_and_queue_t *dq; - u32 cpu_index = os_get_cpu_number (); + u32 thread_index = vlib_get_thread_index (); /* * Poll all devices on this cpu for input/interrupts. */ /* *INDENT-OFF* */ - vec_foreach (dq, dm->devices_by_cpu[cpu_index]) + vec_foreach (dq, dm->devices_by_cpu[thread_index]) { xd = vec_elt_at_index(dm->devices, dq->device); if (xd->flags & DPDK_DEVICE_FLAG_MAYBE_MULTISEG) - n_rx_packets += dpdk_device_input (dm, xd, node, cpu_index, dq->queue_id, /* maybe_multiseg */ 1); + n_rx_packets += dpdk_device_input (dm, xd, node, thread_index, dq->queue_id, /* maybe_multiseg */ 1); else - n_rx_packets += dpdk_device_input (dm, xd, node, cpu_index, dq->queue_id, /* maybe_multiseg */ 0); + n_rx_packets += dpdk_device_input (dm, xd, node, thread_index, dq->queue_id, /* maybe_multiseg */ 0); } /* *INDENT-ON* */ diff --git a/src/plugins/dpdk/hqos/hqos.c b/src/plugins/dpdk/hqos/hqos.c index a288fca7..8b251beb 100644 --- a/src/plugins/dpdk/hqos/hqos.c +++ b/src/plugins/dpdk/hqos/hqos.c @@ -397,7 +397,7 @@ static_always_inline void dpdk_hqos_thread_internal_hqos_dbg_bypass (vlib_main_t * vm) { dpdk_main_t *dm = &dpdk_main; - u32 cpu_index = vm->cpu_index; + u32 thread_index = vm->thread_index; u32 dev_pos; dev_pos = 0; @@ -405,12 +405,12 @@ dpdk_hqos_thread_internal_hqos_dbg_bypass (vlib_main_t * vm) { vlib_worker_thread_barrier_check (); - u32 n_devs = vec_len (dm->devices_by_hqos_cpu[cpu_index]); + u32 n_devs = vec_len (dm->devices_by_hqos_cpu[thread_index]); if (dev_pos >= n_devs) dev_pos = 0; dpdk_device_and_queue_t *dq = - vec_elt_at_index (dm->devices_by_hqos_cpu[cpu_index], dev_pos); + vec_elt_at_index (dm->devices_by_hqos_cpu[thread_index], dev_pos); dpdk_device_t *xd = vec_elt_at_index (dm->devices, dq->device); dpdk_device_hqos_per_hqos_thread_t *hqos = xd->hqos_ht; @@ -479,7 +479,7 @@ static_always_inline void dpdk_hqos_thread_internal (vlib_main_t * vm) { dpdk_main_t *dm = &dpdk_main; - u32 cpu_index = vm->cpu_index; + u32 thread_index = vm->thread_index; u32 dev_pos; dev_pos = 0; @@ -487,7 +487,7 @@ dpdk_hqos_thread_internal (vlib_main_t * vm) { vlib_worker_thread_barrier_check (); - u32 n_devs = vec_len (dm->devices_by_hqos_cpu[cpu_index]); + u32 n_devs = vec_len (dm->devices_by_hqos_cpu[thread_index]); if (PREDICT_FALSE (n_devs == 0)) { dev_pos = 0; @@ -497,7 +497,7 @@ dpdk_hqos_thread_internal (vlib_main_t * vm) dev_pos = 0; dpdk_device_and_queue_t *dq = - vec_elt_at_index (dm->devices_by_hqos_cpu[cpu_index], dev_pos); + vec_elt_at_index (dm->devices_by_hqos_cpu[thread_index], dev_pos); dpdk_device_t *xd = vec_elt_at_index (dm->devices, dq->device); dpdk_device_hqos_per_hqos_thread_t *hqos = xd->hqos_ht; @@ -586,7 +586,7 @@ dpdk_hqos_thread (vlib_worker_thread_t * w) vm = vlib_get_main (); - ASSERT (vm->cpu_index == os_get_cpu_number ()); + ASSERT (vm->thread_index == vlib_get_thread_index ()); clib_time_init (&vm->clib_time); clib_mem_set_heap (w->thread_mheap); @@ -595,7 +595,7 @@ dpdk_hqos_thread (vlib_worker_thread_t * w) while (tm->worker_thread_release == 0) vlib_worker_thread_barrier_check (); - if (vec_len (dm->devices_by_hqos_cpu[vm->cpu_index]) == 0) + if (vec_len (dm->devices_by_hqos_cpu[vm->thread_index]) == 0) return clib_error ("current I/O TX thread does not have any devices assigned to it"); diff --git a/src/plugins/dpdk/ipsec/cli.c b/src/plugins/dpdk/ipsec/cli.c index cd0a6037..3ae8c9b8 100644 --- a/src/plugins/dpdk/ipsec/cli.c +++ b/src/plugins/dpdk/ipsec/cli.c @@ -42,8 +42,8 @@ dpdk_ipsec_show_mapping (vlib_main_t * vm, u16 detail_display) for (i = 0; i < tm->n_vlib_mains; i++) { uword key, data; - u32 cpu_index = vlib_mains[i]->cpu_index; - crypto_worker_main_t *cwm = &dcm->workers_main[cpu_index]; + u32 thread_index = vlib_mains[i]->thread_index; + crypto_worker_main_t *cwm = &dcm->workers_main[thread_index]; u8 *s = 0; if (skip_master) @@ -57,7 +57,7 @@ dpdk_ipsec_show_mapping (vlib_main_t * vm, u16 detail_display) i32 last_cdev = -1; crypto_qp_data_t *qpd; - s = format (s, "%u\t", cpu_index); + s = format (s, "%u\t", thread_index); /* *INDENT-OFF* */ vec_foreach (qpd, cwm->qp_data) @@ -95,7 +95,7 @@ dpdk_ipsec_show_mapping (vlib_main_t * vm, u16 detail_display) cap.sym.auth.algo = p_key->auth_algo; check_algo_is_supported (&cap, auth_str); vlib_cli_output (vm, "%u\t%10s\t%15s\t%3s\t%u\t%u\n", - vlib_mains[i]->cpu_index, cipher_str, auth_str, + vlib_mains[i]->thread_index, cipher_str, auth_str, p_key->is_outbound ? "out" : "in", cwm->qp_data[data].dev_id, cwm->qp_data[data].qp_id); diff --git a/src/plugins/dpdk/ipsec/crypto_node.c b/src/plugins/dpdk/ipsec/crypto_node.c index dc3452b2..a3c45902 100644 --- a/src/plugins/dpdk/ipsec/crypto_node.c +++ b/src/plugins/dpdk/ipsec/crypto_node.c @@ -171,9 +171,9 @@ static uword dpdk_crypto_input_fn (vlib_main_t * vm, vlib_node_runtime_t * node, vlib_frame_t * frame) { - u32 cpu_index = os_get_cpu_number (); + u32 thread_index = vlib_get_thread_index (); dpdk_crypto_main_t *dcm = &dpdk_crypto_main; - crypto_worker_main_t *cwm = &dcm->workers_main[cpu_index]; + crypto_worker_main_t *cwm = &dcm->workers_main[thread_index]; crypto_qp_data_t *qpd; u32 n_deq = 0; diff --git a/src/plugins/dpdk/ipsec/esp.h b/src/plugins/dpdk/ipsec/esp.h index 320295b1..56f0c756 100644 --- a/src/plugins/dpdk/ipsec/esp.h +++ b/src/plugins/dpdk/ipsec/esp.h @@ -170,9 +170,9 @@ static_always_inline int create_sym_sess (ipsec_sa_t * sa, crypto_sa_session_t * sa_sess, u8 is_outbound) { - u32 cpu_index = os_get_cpu_number (); + u32 thread_index = vlib_get_thread_index (); dpdk_crypto_main_t *dcm = &dpdk_crypto_main; - crypto_worker_main_t *cwm = &dcm->workers_main[cpu_index]; + crypto_worker_main_t *cwm = &dcm->workers_main[thread_index]; struct rte_crypto_sym_xform cipher_xform = { 0 }; struct rte_crypto_sym_xform auth_xform = { 0 }; struct rte_crypto_sym_xform *xfs; diff --git a/src/plugins/dpdk/ipsec/esp_decrypt.c b/src/plugins/dpdk/ipsec/esp_decrypt.c index 286e03f8..bab76e3b 100644 --- a/src/plugins/dpdk/ipsec/esp_decrypt.c +++ b/src/plugins/dpdk/ipsec/esp_decrypt.c @@ -88,7 +88,7 @@ dpdk_esp_decrypt_node_fn (vlib_main_t * vm, { u32 n_left_from, *from, *to_next, next_index; ipsec_main_t *im = &ipsec_main; - u32 cpu_index = os_get_cpu_number(); + u32 thread_index = vlib_get_thread_index(); dpdk_crypto_main_t * dcm = &dpdk_crypto_main; dpdk_esp_main_t * em = &dpdk_esp_main; u32 i; @@ -104,7 +104,7 @@ dpdk_esp_decrypt_node_fn (vlib_main_t * vm, return n_left_from; } - crypto_worker_main_t *cwm = vec_elt_at_index(dcm->workers_main, cpu_index); + crypto_worker_main_t *cwm = vec_elt_at_index(dcm->workers_main, thread_index); u32 n_qps = vec_len(cwm->qp_data); struct rte_crypto_op ** cops_to_enq[n_qps]; u32 n_cop_qp[n_qps], * bi_to_enq[n_qps]; diff --git a/src/plugins/dpdk/ipsec/esp_encrypt.c b/src/plugins/dpdk/ipsec/esp_encrypt.c index 5b03de73..f996d7df 100644 --- a/src/plugins/dpdk/ipsec/esp_encrypt.c +++ b/src/plugins/dpdk/ipsec/esp_encrypt.c @@ -93,7 +93,7 @@ dpdk_esp_encrypt_node_fn (vlib_main_t * vm, { u32 n_left_from, *from, *to_next, next_index; ipsec_main_t *im = &ipsec_main; - u32 cpu_index = os_get_cpu_number (); + u32 thread_index = vlib_get_thread_index (); dpdk_crypto_main_t *dcm = &dpdk_crypto_main; dpdk_esp_main_t *em = &dpdk_esp_main; u32 i; @@ -111,7 +111,8 @@ dpdk_esp_encrypt_node_fn (vlib_main_t * vm, return n_left_from; } - crypto_worker_main_t *cwm = vec_elt_at_index (dcm->workers_main, cpu_index); + crypto_worker_main_t *cwm = + vec_elt_at_index (dcm->workers_main, thread_index); u32 n_qps = vec_len (cwm->qp_data); struct rte_crypto_op **cops_to_enq[n_qps]; u32 n_cop_qp[n_qps], *bi_to_enq[n_qps]; diff --git a/src/plugins/dpdk/ipsec/ipsec.c b/src/plugins/dpdk/ipsec/ipsec.c index b0aaaaec..5d8f4fba 100644 --- a/src/plugins/dpdk/ipsec/ipsec.c +++ b/src/plugins/dpdk/ipsec/ipsec.c @@ -289,7 +289,7 @@ dpdk_ipsec_process (vlib_main_t * vm, vlib_node_runtime_t * rt, if (!map) { clib_warning ("unable to create hash table for worker %u", - vlib_mains[i]->cpu_index); + vlib_mains[i]->thread_index); goto error; } cwm->algo_qp_map = map; diff --git a/src/plugins/dpdk/ipsec/ipsec.h b/src/plugins/dpdk/ipsec/ipsec.h index 28bffc80..f0f793c0 100644 --- a/src/plugins/dpdk/ipsec/ipsec.h +++ b/src/plugins/dpdk/ipsec/ipsec.h @@ -95,8 +95,8 @@ static_always_inline void crypto_alloc_cops () { dpdk_crypto_main_t *dcm = &dpdk_crypto_main; - u32 cpu_index = os_get_cpu_number (); - crypto_worker_main_t *cwm = &dcm->workers_main[cpu_index]; + u32 thread_index = vlib_get_thread_index (); + crypto_worker_main_t *cwm = &dcm->workers_main[thread_index]; unsigned socket_id = rte_socket_id (); crypto_qp_data_t *qpd; diff --git a/src/plugins/dpdk/main.c b/src/plugins/dpdk/main.c index 7ee2a785..942b8b2d 100644 --- a/src/plugins/dpdk/main.c +++ b/src/plugins/dpdk/main.c @@ -39,7 +39,7 @@ rte_delay_us_override (unsigned us) * thread then do not intercept. (Must not be called from an * independent pthread). */ - if (os_get_cpu_number () == 0) + if (vlib_get_thread_index () == 0) { /* * We're in the vlib main thread or a vlib process. Make sure diff --git a/src/plugins/flowperpkt/l2_node.c b/src/plugins/flowperpkt/l2_node.c index 1c2f681e..fdaf81d1 100644 --- a/src/plugins/flowperpkt/l2_node.c +++ b/src/plugins/flowperpkt/l2_node.c @@ -102,7 +102,7 @@ add_to_flow_record_l2 (vlib_main_t * vm, u8 * src_mac, u8 * dst_mac, u16 ethertype, u64 timestamp, u16 length, int do_flush) { - u32 my_cpu_number = vm->cpu_index; + u32 my_cpu_number = vm->thread_index; flow_report_main_t *frm = &flow_report_main; ip4_header_t *ip; udp_header_t *udp; diff --git a/src/plugins/flowperpkt/node.c b/src/plugins/flowperpkt/node.c index f77f087d..0277682d 100644 --- a/src/plugins/flowperpkt/node.c +++ b/src/plugins/flowperpkt/node.c @@ -101,7 +101,7 @@ add_to_flow_record_ipv4 (vlib_main_t * vm, u32 src_address, u32 dst_address, u8 tos, u64 timestamp, u16 length, int do_flush) { - u32 my_cpu_number = vm->cpu_index; + u32 my_cpu_number = vm->thread_index; flow_report_main_t *frm = &flow_report_main; ip4_header_t *ip; udp_header_t *udp; diff --git a/src/plugins/ioam/export-common/ioam_export.h b/src/plugins/ioam/export-common/ioam_export.h index 2bf3fd54..9de0d13b 100644 --- a/src/plugins/ioam/export-common/ioam_export.h +++ b/src/plugins/ioam/export-common/ioam_export.h @@ -477,8 +477,8 @@ do { \ from = vlib_frame_vector_args (F); \ n_left_from = (F)->n_vectors; \ next_index = (N)->cached_next_index; \ - while (__sync_lock_test_and_set ((EM)->lockp[(VM)->cpu_index], 1)); \ - my_buf = ioam_export_get_my_buffer (EM, (VM)->cpu_index); \ + while (__sync_lock_test_and_set ((EM)->lockp[(VM)->thread_index], 1)); \ + my_buf = ioam_export_get_my_buffer (EM, (VM)->thread_index); \ my_buf->touched_at = vlib_time_now (VM); \ while (n_left_from > 0) \ { \ @@ -620,7 +620,7 @@ do { \ } \ vlib_node_increment_counter (VM, export_node.index, \ EXPORT_ERROR_RECORDED, pkts_recorded); \ - *(EM)->lockp[(VM)->cpu_index] = 0; \ + *(EM)->lockp[(VM)->thread_index] = 0; \ } while(0) #endif /* __included_ioam_export_h__ */ diff --git a/src/plugins/ioam/ip6/ioam_cache_tunnel_select_node.c b/src/plugins/ioam/ip6/ioam_cache_tunnel_select_node.c index a56dc040..0cf742c9 100644 --- a/src/plugins/ioam/ip6/ioam_cache_tunnel_select_node.c +++ b/src/plugins/ioam/ip6/ioam_cache_tunnel_select_node.c @@ -396,7 +396,7 @@ ip6_reset_ts_hbh_node_fn (vlib_main_t * vm, clib_net_to_host_u32 (tcp0->seq_number) + 1, no_of_responses, now, - vm->cpu_index, &pool_index0)) + vm->thread_index, &pool_index0)) { cache_ts_added++; } @@ -419,7 +419,7 @@ ip6_reset_ts_hbh_node_fn (vlib_main_t * vm, e2e = (ioam_e2e_cache_option_t *) ((u8 *) hbh0 + cm->rewrite_pool_index_offset); - e2e->pool_id = (u8) vm->cpu_index; + e2e->pool_id = (u8) vm->thread_index; e2e->pool_index = pool_index0; ioam_e2e_id_rewrite_handler ((ioam_e2e_id_option_t *) ((u8 *) e2e + @@ -455,7 +455,7 @@ ip6_reset_ts_hbh_node_fn (vlib_main_t * vm, clib_net_to_host_u32 (tcp1->seq_number) + 1, no_of_responses, now, - vm->cpu_index, &pool_index1)) + vm->thread_index, &pool_index1)) { cache_ts_added++; } @@ -479,7 +479,7 @@ ip6_reset_ts_hbh_node_fn (vlib_main_t * vm, e2e = (ioam_e2e_cache_option_t *) ((u8 *) hbh1 + cm->rewrite_pool_index_offset); - e2e->pool_id = (u8) vm->cpu_index; + e2e->pool_id = (u8) vm->thread_index; e2e->pool_index = pool_index1; ioam_e2e_id_rewrite_handler ((ioam_e2e_id_option_t *) ((u8 *) e2e + @@ -562,7 +562,7 @@ ip6_reset_ts_hbh_node_fn (vlib_main_t * vm, clib_net_to_host_u32 (tcp0->seq_number) + 1, no_of_responses, now, - vm->cpu_index, &pool_index0)) + vm->thread_index, &pool_index0)) { cache_ts_added++; } @@ -585,7 +585,7 @@ ip6_reset_ts_hbh_node_fn (vlib_main_t * vm, e2e = (ioam_e2e_cache_option_t *) ((u8 *) hbh0 + cm->rewrite_pool_index_offset); - e2e->pool_id = (u8) vm->cpu_index; + e2e->pool_id = (u8) vm->thread_index; e2e->pool_index = pool_index0; ioam_e2e_id_rewrite_handler ((ioam_e2e_id_option_t *) ((u8 *) e2e + @@ -701,7 +701,7 @@ expired_cache_ts_timer_callback (u32 * expired_timers) ioam_cache_main_t *cm = &ioam_cache_main; int i; u32 pool_index; - u32 thread_index = os_get_cpu_number (); + u32 thread_index = vlib_get_thread_index (); u32 count = 0; for (i = 0; i < vec_len (expired_timers); i++) @@ -724,7 +724,7 @@ ioam_cache_ts_timer_tick_node_fn (vlib_main_t * vm, vlib_frame_t * f) { ioam_cache_main_t *cm = &ioam_cache_main; - u32 my_thread_index = os_get_cpu_number (); + u32 my_thread_index = vlib_get_thread_index (); struct timespec ts, tsrem; tw_timer_expire_timers_16t_2w_512sl (&cm->timer_wheels[my_thread_index], diff --git a/src/plugins/ixge/ixge.c b/src/plugins/ixge/ixge.c index f3c5cc09..08f5b692 100644 --- a/src/plugins/ixge/ixge.c +++ b/src/plugins/ixge/ixge.c @@ -1887,7 +1887,7 @@ done: vlib_increment_combined_counter (vnet_main. interface_main.combined_sw_if_counters + VNET_INTERFACE_COUNTER_RX, - 0 /* cpu_index */ , + 0 /* thread_index */ , xd->vlib_sw_if_index, n_packets, dq->rx.n_bytes); diff --git a/src/plugins/lb/lb.c b/src/plugins/lb/lb.c index add81236..addc2a42 100644 --- a/src/plugins/lb/lb.c +++ b/src/plugins/lb/lb.c @@ -63,11 +63,11 @@ u8 *format_lb_main (u8 * s, va_list * args) s = format(s, " #vips: %u\n", pool_elts(lbm->vips)); s = format(s, " #ass: %u\n", pool_elts(lbm->ass) - 1); - u32 cpu_index; - for(cpu_index = 0; cpu_index < tm->n_vlib_mains; cpu_index++ ) { - lb_hash_t *h = lbm->per_cpu[cpu_index].sticky_ht; + u32 thread_index; + for(thread_index = 0; thread_index < tm->n_vlib_mains; thread_index++ ) { + lb_hash_t *h = lbm->per_cpu[thread_index].sticky_ht; if (h) { - s = format(s, "core %d\n", cpu_index); + s = format(s, "core %d\n", thread_index); s = format(s, " timeout: %ds\n", h->timeout); s = format(s, " usage: %d / %d\n", lb_hash_elts(h, lb_hash_time_now(vlib_get_main())), lb_hash_size(h)); } diff --git a/src/plugins/lb/node.c b/src/plugins/lb/node.c index 8b763c53..3171148b 100644 --- a/src/plugins/lb/node.c +++ b/src/plugins/lb/node.c @@ -60,10 +60,10 @@ format_lb_trace (u8 * s, va_list * args) return s; } -lb_hash_t *lb_get_sticky_table(u32 cpu_index) +lb_hash_t *lb_get_sticky_table(u32 thread_index) { lb_main_t *lbm = &lb_main; - lb_hash_t *sticky_ht = lbm->per_cpu[cpu_index].sticky_ht; + lb_hash_t *sticky_ht = lbm->per_cpu[thread_index].sticky_ht; //Check if size changed if (PREDICT_FALSE(sticky_ht && (lbm->per_cpu_sticky_buckets != lb_hash_nbuckets(sticky_ht)))) { @@ -71,8 +71,8 @@ lb_hash_t *lb_get_sticky_table(u32 cpu_index) lb_hash_bucket_t *b; u32 i; lb_hash_foreach_entry(sticky_ht, b, i) { - vlib_refcount_add(&lbm->as_refcount, cpu_index, b->value[i], -1); - vlib_refcount_add(&lbm->as_refcount, cpu_index, 0, 1); + vlib_refcount_add(&lbm->as_refcount, thread_index, b->value[i], -1); + vlib_refcount_add(&lbm->as_refcount, thread_index, 0, 1); } lb_hash_free(sticky_ht); @@ -81,8 +81,8 @@ lb_hash_t *lb_get_sticky_table(u32 cpu_index) //Create if necessary if (PREDICT_FALSE(sticky_ht == NULL)) { - lbm->per_cpu[cpu_index].sticky_ht = lb_hash_alloc(lbm->per_cpu_sticky_buckets, lbm->flow_timeout); - sticky_ht = lbm->per_cpu[cpu_index].sticky_ht; + lbm->per_cpu[thread_index].sticky_ht = lb_hash_alloc(lbm->per_cpu_sticky_buckets, lbm->flow_timeout); + sticky_ht = lbm->per_cpu[thread_index].sticky_ht; clib_warning("Regenerated sticky table %p", sticky_ht); } @@ -153,10 +153,10 @@ lb_node_fn (vlib_main_t * vm, { lb_main_t *lbm = &lb_main; u32 n_left_from, *from, next_index, *to_next, n_left_to_next; - u32 cpu_index = os_get_cpu_number(); + u32 thread_index = vlib_get_thread_index(); u32 lb_time = lb_hash_time_now(vm); - lb_hash_t *sticky_ht = lb_get_sticky_table(cpu_index); + lb_hash_t *sticky_ht = lb_get_sticky_table(thread_index); from = vlib_frame_vector_args (frame); n_left_from = frame->n_vectors; next_index = node->cached_next_index; @@ -240,9 +240,9 @@ lb_node_fn (vlib_main_t * vm, //Configuration may be changed, vectors resized, etc... //Dereference previously used - vlib_refcount_add(&lbm->as_refcount, cpu_index, + vlib_refcount_add(&lbm->as_refcount, thread_index, lb_hash_available_value(sticky_ht, hash0, available_index0), -1); - vlib_refcount_add(&lbm->as_refcount, cpu_index, + vlib_refcount_add(&lbm->as_refcount, thread_index, asindex0, 1); //Add sticky entry @@ -260,7 +260,7 @@ lb_node_fn (vlib_main_t * vm, } vlib_increment_simple_counter(&lbm->vip_counters[counter], - cpu_index, + thread_index, vnet_buffer (p0)->ip.adj_index[VLIB_TX], 1); diff --git a/src/plugins/lb/refcount.c b/src/plugins/lb/refcount.c index 22415c88..6f01ab5a 100644 --- a/src/plugins/lb/refcount.c +++ b/src/plugins/lb/refcount.c @@ -31,10 +31,10 @@ u64 vlib_refcount_get(vlib_refcount_t *r, u32 index) { u64 count = 0; vlib_thread_main_t *tm = vlib_get_thread_main (); - u32 cpu_index; - for (cpu_index = 0; cpu_index < tm->n_vlib_mains; cpu_index++) { - if (r->per_cpu[cpu_index].length > index) - count += r->per_cpu[cpu_index].counters[index]; + u32 thread_index; + for (thread_index = 0; thread_index < tm->n_vlib_mains; thread_index++) { + if (r->per_cpu[thread_index].length > index) + count += r->per_cpu[thread_index].counters[index]; } return count; } diff --git a/src/plugins/lb/refcount.h b/src/plugins/lb/refcount.h index 8c26e7be..dcfcb3fe 100644 --- a/src/plugins/lb/refcount.h +++ b/src/plugins/lb/refcount.h @@ -45,9 +45,9 @@ typedef struct { void __vlib_refcount_resize(vlib_refcount_per_cpu_t *per_cpu, u32 size); static_always_inline -void vlib_refcount_add(vlib_refcount_t *r, u32 cpu_index, u32 counter_index, i32 v) +void vlib_refcount_add(vlib_refcount_t *r, u32 thread_index, u32 counter_index, i32 v) { - vlib_refcount_per_cpu_t *per_cpu = &r->per_cpu[cpu_index]; + vlib_refcount_per_cpu_t *per_cpu = &r->per_cpu[thread_index]; if (PREDICT_FALSE(counter_index >= per_cpu->length)) __vlib_refcount_resize(per_cpu, clib_max(counter_index + 16, per_cpu->length * 2)); diff --git a/src/plugins/memif/node.c b/src/plugins/memif/node.c index 659d5dfb..cee1f3d1 100644 --- a/src/plugins/memif/node.c +++ b/src/plugins/memif/node.c @@ -94,7 +94,7 @@ memif_device_input_inline (vlib_main_t * vm, vlib_node_runtime_t * node, u32 n_rx_bytes = 0; u32 *to_next = 0; u32 n_free_bufs; - u32 cpu_index = os_get_cpu_number (); + u32 thread_index = vlib_get_thread_index (); u32 bi0, bi1; vlib_buffer_t *b0, *b1; u16 ring_size = 1 << mif->log2_ring_size; @@ -105,14 +105,15 @@ memif_device_input_inline (vlib_main_t * vm, vlib_node_runtime_t * node, if (mif->per_interface_next_index != ~0) next_index = mif->per_interface_next_index; - n_free_bufs = vec_len (nm->rx_buffers[cpu_index]); + n_free_bufs = vec_len (nm->rx_buffers[thread_index]); if (PREDICT_FALSE (n_free_bufs < ring_size)) { - vec_validate (nm->rx_buffers[cpu_index], ring_size + n_free_bufs - 1); + vec_validate (nm->rx_buffers[thread_index], + ring_size + n_free_bufs - 1); n_free_bufs += - vlib_buffer_alloc (vm, &nm->rx_buffers[cpu_index][n_free_bufs], + vlib_buffer_alloc (vm, &nm->rx_buffers[thread_index][n_free_bufs], ring_size); - _vec_len (nm->rx_buffers[cpu_index]) = n_free_bufs; + _vec_len (nm->rx_buffers[thread_index]) = n_free_bufs; } head = ring->head; @@ -158,15 +159,15 @@ memif_device_input_inline (vlib_main_t * vm, vlib_node_runtime_t * node, CLIB_CACHE_LINE_BYTES, LOAD); } /* get empty buffer */ - u32 last_buf = vec_len (nm->rx_buffers[cpu_index]) - 1; - bi0 = nm->rx_buffers[cpu_index][last_buf]; - bi1 = nm->rx_buffers[cpu_index][last_buf - 1]; - _vec_len (nm->rx_buffers[cpu_index]) -= 2; + u32 last_buf = vec_len (nm->rx_buffers[thread_index]) - 1; + bi0 = nm->rx_buffers[thread_index][last_buf]; + bi1 = nm->rx_buffers[thread_index][last_buf - 1]; + _vec_len (nm->rx_buffers[thread_index]) -= 2; if (last_buf > 4) { - memif_prefetch (vm, nm->rx_buffers[cpu_index][last_buf - 2]); - memif_prefetch (vm, nm->rx_buffers[cpu_index][last_buf - 3]); + memif_prefetch (vm, nm->rx_buffers[thread_index][last_buf - 2]); + memif_prefetch (vm, nm->rx_buffers[thread_index][last_buf - 3]); } /* enqueue buffer */ @@ -256,9 +257,9 @@ memif_device_input_inline (vlib_main_t * vm, vlib_node_runtime_t * node, while (num_slots && n_left_to_next) { /* get empty buffer */ - u32 last_buf = vec_len (nm->rx_buffers[cpu_index]) - 1; - bi0 = nm->rx_buffers[cpu_index][last_buf]; - _vec_len (nm->rx_buffers[cpu_index]) = last_buf; + u32 last_buf = vec_len (nm->rx_buffers[thread_index]) - 1; + bi0 = nm->rx_buffers[thread_index][last_buf]; + _vec_len (nm->rx_buffers[thread_index]) = last_buf; /* enqueue buffer */ to_next[0] = bi0; @@ -315,7 +316,7 @@ memif_device_input_inline (vlib_main_t * vm, vlib_node_runtime_t * node, ring->tail = head; vlib_increment_combined_counter (vnm->interface_main.combined_sw_if_counters - + VNET_INTERFACE_COUNTER_RX, cpu_index, + + VNET_INTERFACE_COUNTER_RX, thread_index, mif->hw_if_index, n_rx_packets, n_rx_bytes); @@ -327,7 +328,7 @@ memif_input_fn (vlib_main_t * vm, vlib_node_runtime_t * node, vlib_frame_t * frame) { u32 n_rx_packets = 0; - u32 cpu_index = os_get_cpu_number (); + u32 thread_index = vlib_get_thread_index (); memif_main_t *nm = &memif_main; memif_if_t *mif; @@ -337,7 +338,7 @@ memif_input_fn (vlib_main_t * vm, vlib_node_runtime_t * node, if (mif->flags & MEMIF_IF_FLAG_ADMIN_UP && mif->flags & MEMIF_IF_FLAG_CONNECTED && (mif->if_index % nm->input_cpu_count) == - (cpu_index - nm->input_cpu_first_index)) + (thread_index - nm->input_cpu_first_index)) { if (mif->flags & MEMIF_IF_FLAG_IS_SLAVE) n_rx_packets += diff --git a/src/plugins/snat/in2out.c b/src/plugins/snat/in2out.c index b4961365..e5ee965f 100644 --- a/src/plugins/snat/in2out.c +++ b/src/plugins/snat/in2out.c @@ -212,7 +212,7 @@ static u32 slow_path (snat_main_t *sm, vlib_buffer_t *b0, snat_session_t ** sessionp, vlib_node_runtime_t * node, u32 next0, - u32 cpu_index) + u32 thread_index) { snat_user_t *u; snat_user_key_t user_key; @@ -246,27 +246,27 @@ static u32 slow_path (snat_main_t *sm, vlib_buffer_t *b0, if (clib_bihash_search_8_8 (&sm->user_hash, &kv0, &value0)) { /* no, make a new one */ - pool_get (sm->per_thread_data[cpu_index].users, u); + pool_get (sm->per_thread_data[thread_index].users, u); memset (u, 0, sizeof (*u)); u->addr = ip0->src_address; u->fib_index = rx_fib_index0; - pool_get (sm->per_thread_data[cpu_index].list_pool, per_user_list_head_elt); + pool_get (sm->per_thread_data[thread_index].list_pool, per_user_list_head_elt); u->sessions_per_user_list_head_index = per_user_list_head_elt - - sm->per_thread_data[cpu_index].list_pool; + sm->per_thread_data[thread_index].list_pool; - clib_dlist_init (sm->per_thread_data[cpu_index].list_pool, + clib_dlist_init (sm->per_thread_data[thread_index].list_pool, u->sessions_per_user_list_head_index); - kv0.value = u - sm->per_thread_data[cpu_index].users; + kv0.value = u - sm->per_thread_data[thread_index].users; /* add user */ clib_bihash_add_del_8_8 (&sm->user_hash, &kv0, 1 /* is_add */); } else { - u = pool_elt_at_index (sm->per_thread_data[cpu_index].users, + u = pool_elt_at_index (sm->per_thread_data[thread_index].users, value0.value); } @@ -276,25 +276,25 @@ static u32 slow_path (snat_main_t *sm, vlib_buffer_t *b0, /* Remove the oldest dynamic translation */ do { oldest_per_user_translation_list_index = - clib_dlist_remove_head (sm->per_thread_data[cpu_index].list_pool, + clib_dlist_remove_head (sm->per_thread_data[thread_index].list_pool, u->sessions_per_user_list_head_index); ASSERT (oldest_per_user_translation_list_index != ~0); /* add it back to the end of the LRU list */ - clib_dlist_addtail (sm->per_thread_data[cpu_index].list_pool, + clib_dlist_addtail (sm->per_thread_data[thread_index].list_pool, u->sessions_per_user_list_head_index, oldest_per_user_translation_list_index); /* Get the list element */ oldest_per_user_translation_list_elt = - pool_elt_at_index (sm->per_thread_data[cpu_index].list_pool, + pool_elt_at_index (sm->per_thread_data[thread_index].list_pool, oldest_per_user_translation_list_index); /* Get the session index from the list element */ session_index = oldest_per_user_translation_list_elt->value; /* Get the session */ - s = pool_elt_at_index (sm->per_thread_data[cpu_index].sessions, + s = pool_elt_at_index (sm->per_thread_data[thread_index].sessions, session_index); } while (snat_is_session_static (s)); @@ -346,7 +346,7 @@ static u32 slow_path (snat_main_t *sm, vlib_buffer_t *b0, } /* Create a new session */ - pool_get (sm->per_thread_data[cpu_index].sessions, s); + pool_get (sm->per_thread_data[thread_index].sessions, s); memset (s, 0, sizeof (*s)); s->outside_address_index = address_index; @@ -362,22 +362,22 @@ static u32 slow_path (snat_main_t *sm, vlib_buffer_t *b0, } /* Create list elts */ - pool_get (sm->per_thread_data[cpu_index].list_pool, + pool_get (sm->per_thread_data[thread_index].list_pool, per_user_translation_list_elt); - clib_dlist_init (sm->per_thread_data[cpu_index].list_pool, + clib_dlist_init (sm->per_thread_data[thread_index].list_pool, per_user_translation_list_elt - - sm->per_thread_data[cpu_index].list_pool); + sm->per_thread_data[thread_index].list_pool); per_user_translation_list_elt->value = - s - sm->per_thread_data[cpu_index].sessions; + s - sm->per_thread_data[thread_index].sessions; s->per_user_index = per_user_translation_list_elt - - sm->per_thread_data[cpu_index].list_pool; + sm->per_thread_data[thread_index].list_pool; s->per_user_list_head_index = u->sessions_per_user_list_head_index; - clib_dlist_addtail (sm->per_thread_data[cpu_index].list_pool, + clib_dlist_addtail (sm->per_thread_data[thread_index].list_pool, s->per_user_list_head_index, per_user_translation_list_elt - - sm->per_thread_data[cpu_index].list_pool); + sm->per_thread_data[thread_index].list_pool); } s->in2out = *key0; @@ -388,12 +388,12 @@ static u32 slow_path (snat_main_t *sm, vlib_buffer_t *b0, /* Add to translation hashes */ kv0.key = s->in2out.as_u64; - kv0.value = s - sm->per_thread_data[cpu_index].sessions; + kv0.value = s - sm->per_thread_data[thread_index].sessions; if (clib_bihash_add_del_8_8 (&sm->in2out, &kv0, 1 /* is_add */)) clib_warning ("in2out key add failed"); kv0.key = s->out2in.as_u64; - kv0.value = s - sm->per_thread_data[cpu_index].sessions; + kv0.value = s - sm->per_thread_data[thread_index].sessions; if (clib_bihash_add_del_8_8 (&sm->out2in, &kv0, 1 /* is_add */)) clib_warning ("out2in key add failed"); @@ -403,7 +403,7 @@ static u32 slow_path (snat_main_t *sm, vlib_buffer_t *b0, worker_by_out_key.port = s->out2in.port; worker_by_out_key.fib_index = s->out2in.fib_index; kv0.key = worker_by_out_key.as_u64; - kv0.value = cpu_index; + kv0.value = thread_index; clib_bihash_add_del_8_8 (&sm->worker_by_out, &kv0, 1); /* log NAT event */ @@ -465,7 +465,7 @@ snat_in2out_error_t icmp_get_key(icmp46_header_t *icmp0, * * @param[in,out] sm SNAT main * @param[in,out] node SNAT node runtime - * @param[in] cpu_index CPU index + * @param[in] thread_index thread index * @param[in,out] b0 buffer containing packet to be translated * @param[out] p_key address and port before NAT translation * @param[out] p_value address and port after NAT translation @@ -473,7 +473,7 @@ snat_in2out_error_t icmp_get_key(icmp46_header_t *icmp0, * @param d optional parameter */ u32 icmp_match_in2out_slow(snat_main_t *sm, vlib_node_runtime_t *node, - u32 cpu_index, vlib_buffer_t *b0, + u32 thread_index, vlib_buffer_t *b0, snat_session_key_t *p_key, snat_session_key_t *p_value, u8 *p_dont_translate, void *d) @@ -524,13 +524,13 @@ u32 icmp_match_in2out_slow(snat_main_t *sm, vlib_node_runtime_t *node, } next0 = slow_path (sm, b0, ip0, rx_fib_index0, &key0, - &s0, node, next0, cpu_index); + &s0, node, next0, thread_index); if (PREDICT_FALSE (next0 == SNAT_IN2OUT_NEXT_DROP)) goto out; } else - s0 = pool_elt_at_index (sm->per_thread_data[cpu_index].sessions, + s0 = pool_elt_at_index (sm->per_thread_data[thread_index].sessions, value0.value); out: @@ -548,7 +548,7 @@ out: * * @param[in] sm SNAT main * @param[in,out] node SNAT node runtime - * @param[in] cpu_index CPU index + * @param[in] thread_index thread index * @param[in,out] b0 buffer containing packet to be translated * @param[out] p_key address and port before NAT translation * @param[out] p_value address and port after NAT translation @@ -556,7 +556,7 @@ out: * @param d optional parameter */ u32 icmp_match_in2out_fast(snat_main_t *sm, vlib_node_runtime_t *node, - u32 cpu_index, vlib_buffer_t *b0, + u32 thread_index, vlib_buffer_t *b0, snat_session_key_t *p_key, snat_session_key_t *p_value, u8 *p_dont_translate, void *d) @@ -624,7 +624,7 @@ static inline u32 icmp_in2out (snat_main_t *sm, u32 rx_fib_index0, vlib_node_runtime_t * node, u32 next0, - u32 cpu_index, + u32 thread_index, void *d) { snat_session_key_t key0, sm0; @@ -641,7 +641,7 @@ static inline u32 icmp_in2out (snat_main_t *sm, echo0 = (icmp_echo_header_t *)(icmp0+1); - next0_tmp = sm->icmp_match_in2out_cb(sm, node, cpu_index, b0, + next0_tmp = sm->icmp_match_in2out_cb(sm, node, thread_index, b0, &key0, &sm0, &dont_translate, d); if (next0_tmp != ~0) next0 = next0_tmp; @@ -847,11 +847,11 @@ static inline u32 icmp_in2out_slow_path (snat_main_t *sm, vlib_node_runtime_t * node, u32 next0, f64 now, - u32 cpu_index, + u32 thread_index, snat_session_t ** p_s0) { next0 = icmp_in2out(sm, b0, ip0, icmp0, sw_if_index0, rx_fib_index0, node, - next0, cpu_index, p_s0); + next0, thread_index, p_s0); snat_session_t * s0 = *p_s0; if (PREDICT_TRUE(next0 != SNAT_IN2OUT_NEXT_DROP && s0)) { @@ -862,9 +862,9 @@ static inline u32 icmp_in2out_slow_path (snat_main_t *sm, /* Per-user LRU list maintenance for dynamic translations */ if (!snat_is_session_static (s0)) { - clib_dlist_remove (sm->per_thread_data[cpu_index].list_pool, + clib_dlist_remove (sm->per_thread_data[thread_index].list_pool, s0->per_user_index); - clib_dlist_addtail (sm->per_thread_data[cpu_index].list_pool, + clib_dlist_addtail (sm->per_thread_data[thread_index].list_pool, s0->per_user_list_head_index, s0->per_user_index); } @@ -884,7 +884,7 @@ snat_in2out_node_fn_inline (vlib_main_t * vm, snat_runtime_t * rt = (snat_runtime_t *)node->runtime_data; f64 now = vlib_time_now (vm); u32 stats_node_index; - u32 cpu_index = os_get_cpu_number (); + u32 thread_index = vlib_get_thread_index (); stats_node_index = is_slow_path ? snat_in2out_slowpath_node.index : snat_in2out_node.index; @@ -977,7 +977,7 @@ snat_in2out_node_fn_inline (vlib_main_t * vm, { next0 = icmp_in2out_slow_path (sm, b0, ip0, icmp0, sw_if_index0, rx_fib_index0, - node, next0, now, cpu_index, &s0); + node, next0, now, thread_index, &s0); goto trace00; } } @@ -1006,7 +1006,7 @@ snat_in2out_node_fn_inline (vlib_main_t * vm, goto trace00; next0 = slow_path (sm, b0, ip0, rx_fib_index0, &key0, - &s0, node, next0, cpu_index); + &s0, node, next0, thread_index); if (PREDICT_FALSE (next0 == SNAT_IN2OUT_NEXT_DROP)) goto trace00; } @@ -1017,7 +1017,7 @@ snat_in2out_node_fn_inline (vlib_main_t * vm, } } else - s0 = pool_elt_at_index (sm->per_thread_data[cpu_index].sessions, + s0 = pool_elt_at_index (sm->per_thread_data[thread_index].sessions, value0.value); old_addr0 = ip0->src_address.as_u32; @@ -1063,9 +1063,9 @@ snat_in2out_node_fn_inline (vlib_main_t * vm, /* Per-user LRU list maintenance for dynamic translation */ if (!snat_is_session_static (s0)) { - clib_dlist_remove (sm->per_thread_data[cpu_index].list_pool, + clib_dlist_remove (sm->per_thread_data[thread_index].list_pool, s0->per_user_index); - clib_dlist_addtail (sm->per_thread_data[cpu_index].list_pool, + clib_dlist_addtail (sm->per_thread_data[thread_index].list_pool, s0->per_user_list_head_index, s0->per_user_index); } @@ -1081,7 +1081,7 @@ snat_in2out_node_fn_inline (vlib_main_t * vm, t->next_index = next0; t->session_index = ~0; if (s0) - t->session_index = s0 - sm->per_thread_data[cpu_index].sessions; + t->session_index = s0 - sm->per_thread_data[thread_index].sessions; } pkts_processed += next0 != SNAT_IN2OUT_NEXT_DROP; @@ -1117,7 +1117,7 @@ snat_in2out_node_fn_inline (vlib_main_t * vm, { next1 = icmp_in2out_slow_path (sm, b1, ip1, icmp1, sw_if_index1, rx_fib_index1, node, - next1, now, cpu_index, &s1); + next1, now, thread_index, &s1); goto trace01; } } @@ -1146,7 +1146,7 @@ snat_in2out_node_fn_inline (vlib_main_t * vm, goto trace01; next1 = slow_path (sm, b1, ip1, rx_fib_index1, &key1, - &s1, node, next1, cpu_index); + &s1, node, next1, thread_index); if (PREDICT_FALSE (next1 == SNAT_IN2OUT_NEXT_DROP)) goto trace01; } @@ -1157,7 +1157,7 @@ snat_in2out_node_fn_inline (vlib_main_t * vm, } } else - s1 = pool_elt_at_index (sm->per_thread_data[cpu_index].sessions, + s1 = pool_elt_at_index (sm->per_thread_data[thread_index].sessions, value1.value); old_addr1 = ip1->src_address.as_u32; @@ -1203,9 +1203,9 @@ snat_in2out_node_fn_inline (vlib_main_t * vm, /* Per-user LRU list maintenance for dynamic translation */ if (!snat_is_session_static (s1)) { - clib_dlist_remove (sm->per_thread_data[cpu_index].list_pool, + clib_dlist_remove (sm->per_thread_data[thread_index].list_pool, s1->per_user_index); - clib_dlist_addtail (sm->per_thread_data[cpu_index].list_pool, + clib_dlist_addtail (sm->per_thread_data[thread_index].list_pool, s1->per_user_list_head_index, s1->per_user_index); } @@ -1220,7 +1220,7 @@ snat_in2out_node_fn_inline (vlib_main_t * vm, t->next_index = next1; t->session_index = ~0; if (s1) - t->session_index = s1 - sm->per_thread_data[cpu_index].sessions; + t->session_index = s1 - sm->per_thread_data[thread_index].sessions; } pkts_processed += next1 != SNAT_IN2OUT_NEXT_DROP; @@ -1292,7 +1292,7 @@ snat_in2out_node_fn_inline (vlib_main_t * vm, { next0 = icmp_in2out_slow_path (sm, b0, ip0, icmp0, sw_if_index0, rx_fib_index0, node, - next0, now, cpu_index, &s0); + next0, now, thread_index, &s0); goto trace0; } } @@ -1321,7 +1321,7 @@ snat_in2out_node_fn_inline (vlib_main_t * vm, goto trace0; next0 = slow_path (sm, b0, ip0, rx_fib_index0, &key0, - &s0, node, next0, cpu_index); + &s0, node, next0, thread_index); if (PREDICT_FALSE (next0 == SNAT_IN2OUT_NEXT_DROP)) goto trace0; @@ -1333,7 +1333,7 @@ snat_in2out_node_fn_inline (vlib_main_t * vm, } } else - s0 = pool_elt_at_index (sm->per_thread_data[cpu_index].sessions, + s0 = pool_elt_at_index (sm->per_thread_data[thread_index].sessions, value0.value); old_addr0 = ip0->src_address.as_u32; @@ -1379,9 +1379,9 @@ snat_in2out_node_fn_inline (vlib_main_t * vm, /* Per-user LRU list maintenance for dynamic translation */ if (!snat_is_session_static (s0)) { - clib_dlist_remove (sm->per_thread_data[cpu_index].list_pool, + clib_dlist_remove (sm->per_thread_data[thread_index].list_pool, s0->per_user_index); - clib_dlist_addtail (sm->per_thread_data[cpu_index].list_pool, + clib_dlist_addtail (sm->per_thread_data[thread_index].list_pool, s0->per_user_list_head_index, s0->per_user_index); } @@ -1397,7 +1397,7 @@ snat_in2out_node_fn_inline (vlib_main_t * vm, t->next_index = next0; t->session_index = ~0; if (s0) - t->session_index = s0 - sm->per_thread_data[cpu_index].sessions; + t->session_index = s0 - sm->per_thread_data[thread_index].sessions; } pkts_processed += next0 != SNAT_IN2OUT_NEXT_DROP; @@ -2010,7 +2010,7 @@ snat_in2out_worker_handoff_fn (vlib_main_t * vm, u32 n_left_to_next_worker = 0, *to_next_worker = 0; u32 next_worker_index = 0; u32 current_worker_index = ~0; - u32 cpu_index = os_get_cpu_number (); + u32 thread_index = vlib_get_thread_index (); ASSERT (vec_len (sm->workers)); @@ -2048,7 +2048,7 @@ snat_in2out_worker_handoff_fn (vlib_main_t * vm, next_worker_index = sm->worker_in2out_cb(ip0, rx_fib_index0); - if (PREDICT_FALSE (next_worker_index != cpu_index)) + if (PREDICT_FALSE (next_worker_index != thread_index)) { do_handoff = 1; diff --git a/src/plugins/snat/out2in.c b/src/plugins/snat/out2in.c index 656e42db..5d308d78 100644 --- a/src/plugins/snat/out2in.c +++ b/src/plugins/snat/out2in.c @@ -129,7 +129,7 @@ create_session_for_static_mapping (snat_main_t *sm, snat_session_key_t in2out, snat_session_key_t out2in, vlib_node_runtime_t * node, - u32 cpu_index) + u32 thread_index) { snat_user_t *u; snat_user_key_t user_key; @@ -146,36 +146,36 @@ create_session_for_static_mapping (snat_main_t *sm, if (clib_bihash_search_8_8 (&sm->user_hash, &kv0, &value0)) { /* no, make a new one */ - pool_get (sm->per_thread_data[cpu_index].users, u); + pool_get (sm->per_thread_data[thread_index].users, u); memset (u, 0, sizeof (*u)); u->addr = in2out.addr; u->fib_index = in2out.fib_index; - pool_get (sm->per_thread_data[cpu_index].list_pool, + pool_get (sm->per_thread_data[thread_index].list_pool, per_user_list_head_elt); u->sessions_per_user_list_head_index = per_user_list_head_elt - - sm->per_thread_data[cpu_index].list_pool; + sm->per_thread_data[thread_index].list_pool; - clib_dlist_init (sm->per_thread_data[cpu_index].list_pool, + clib_dlist_init (sm->per_thread_data[thread_index].list_pool, u->sessions_per_user_list_head_index); - kv0.value = u - sm->per_thread_data[cpu_index].users; + kv0.value = u - sm->per_thread_data[thread_index].users; /* add user */ clib_bihash_add_del_8_8 (&sm->user_hash, &kv0, 1 /* is_add */); /* add non-traslated packets worker lookup */ - kv0.value = cpu_index; + kv0.value = thread_index; clib_bihash_add_del_8_8 (&sm->worker_by_in, &kv0, 1); } else { - u = pool_elt_at_index (sm->per_thread_data[cpu_index].users, + u = pool_elt_at_index (sm->per_thread_data[thread_index].users, value0.value); } - pool_get (sm->per_thread_data[cpu_index].sessions, s); + pool_get (sm->per_thread_data[thread_index].sessions, s); memset (s, 0, sizeof (*s)); s->outside_address_index = ~0; @@ -183,22 +183,22 @@ create_session_for_static_mapping (snat_main_t *sm, u->nstaticsessions++; /* Create list elts */ - pool_get (sm->per_thread_data[cpu_index].list_pool, + pool_get (sm->per_thread_data[thread_index].list_pool, per_user_translation_list_elt); - clib_dlist_init (sm->per_thread_data[cpu_index].list_pool, + clib_dlist_init (sm->per_thread_data[thread_index].list_pool, per_user_translation_list_elt - - sm->per_thread_data[cpu_index].list_pool); + sm->per_thread_data[thread_index].list_pool); per_user_translation_list_elt->value = - s - sm->per_thread_data[cpu_index].sessions; + s - sm->per_thread_data[thread_index].sessions; s->per_user_index = - per_user_translation_list_elt - sm->per_thread_data[cpu_index].list_pool; + per_user_translation_list_elt - sm->per_thread_data[thread_index].list_pool; s->per_user_list_head_index = u->sessions_per_user_list_head_index; - clib_dlist_addtail (sm->per_thread_data[cpu_index].list_pool, + clib_dlist_addtail (sm->per_thread_data[thread_index].list_pool, s->per_user_list_head_index, per_user_translation_list_elt - - sm->per_thread_data[cpu_index].list_pool); + sm->per_thread_data[thread_index].list_pool); s->in2out = in2out; s->out2in = out2in; @@ -206,12 +206,12 @@ create_session_for_static_mapping (snat_main_t *sm, /* Add to translation hashes */ kv0.key = s->in2out.as_u64; - kv0.value = s - sm->per_thread_data[cpu_index].sessions; + kv0.value = s - sm->per_thread_data[thread_index].sessions; if (clib_bihash_add_del_8_8 (&sm->in2out, &kv0, 1 /* is_add */)) clib_warning ("in2out key add failed"); kv0.key = s->out2in.as_u64; - kv0.value = s - sm->per_thread_data[cpu_index].sessions; + kv0.value = s - sm->per_thread_data[thread_index].sessions; if (clib_bihash_add_del_8_8 (&sm->out2in, &kv0, 1 /* is_add */)) clib_warning ("out2in key add failed"); @@ -298,7 +298,7 @@ is_interface_addr(snat_main_t *sm, vlib_node_runtime_t *node, u32 sw_if_index0, * * @param[in,out] sm SNAT main * @param[in,out] node SNAT node runtime - * @param[in] cpu_index CPU index + * @param[in] thread_index thread index * @param[in,out] b0 buffer containing packet to be translated * @param[out] p_key address and port before NAT translation * @param[out] p_value address and port after NAT translation @@ -306,7 +306,7 @@ is_interface_addr(snat_main_t *sm, vlib_node_runtime_t *node, u32 sw_if_index0, * @param d optional parameter */ u32 icmp_match_out2in_slow(snat_main_t *sm, vlib_node_runtime_t *node, - u32 cpu_index, vlib_buffer_t *b0, + u32 thread_index, vlib_buffer_t *b0, snat_session_key_t *p_key, snat_session_key_t *p_value, u8 *p_dont_translate, void *d) @@ -366,7 +366,7 @@ u32 icmp_match_out2in_slow(snat_main_t *sm, vlib_node_runtime_t *node, /* Create session initiated by host from external network */ s0 = create_session_for_static_mapping(sm, b0, sm0, key0, - node, cpu_index); + node, thread_index); if (!s0) { @@ -375,7 +375,7 @@ u32 icmp_match_out2in_slow(snat_main_t *sm, vlib_node_runtime_t *node, } } else - s0 = pool_elt_at_index (sm->per_thread_data[cpu_index].sessions, + s0 = pool_elt_at_index (sm->per_thread_data[thread_index].sessions, value0.value); out: @@ -393,7 +393,7 @@ out: * * @param[in] sm SNAT main * @param[in,out] node SNAT node runtime - * @param[in] cpu_index CPU index + * @param[in] thread_index thread index * @param[in,out] b0 buffer containing packet to be translated * @param[out] p_key address and port before NAT translation * @param[out] p_value address and port after NAT translation @@ -401,7 +401,7 @@ out: * @param d optional parameter */ u32 icmp_match_out2in_fast(snat_main_t *sm, vlib_node_runtime_t *node, - u32 cpu_index, vlib_buffer_t *b0, + u32 thread_index, vlib_buffer_t *b0, snat_session_key_t *p_key, snat_session_key_t *p_value, u8 *p_dont_translate, void *d) @@ -460,7 +460,7 @@ static inline u32 icmp_out2in (snat_main_t *sm, u32 rx_fib_index0, vlib_node_runtime_t * node, u32 next0, - u32 cpu_index, + u32 thread_index, void *d) { snat_session_key_t key0, sm0; @@ -477,7 +477,7 @@ static inline u32 icmp_out2in (snat_main_t *sm, echo0 = (icmp_echo_header_t *)(icmp0+1); - next0_tmp = sm->icmp_match_out2in_cb(sm, node, cpu_index, b0, + next0_tmp = sm->icmp_match_out2in_cb(sm, node, thread_index, b0, &key0, &sm0, &dont_translate, d); if (next0_tmp != ~0) next0 = next0_tmp; @@ -589,11 +589,11 @@ static inline u32 icmp_out2in_slow_path (snat_main_t *sm, u32 rx_fib_index0, vlib_node_runtime_t * node, u32 next0, f64 now, - u32 cpu_index, + u32 thread_index, snat_session_t ** p_s0) { next0 = icmp_out2in(sm, b0, ip0, icmp0, sw_if_index0, rx_fib_index0, node, - next0, cpu_index, p_s0); + next0, thread_index, p_s0); snat_session_t * s0 = *p_s0; if (PREDICT_TRUE(next0 != SNAT_OUT2IN_NEXT_DROP && s0)) { @@ -604,9 +604,9 @@ static inline u32 icmp_out2in_slow_path (snat_main_t *sm, /* Per-user LRU list maintenance for dynamic translation */ if (!snat_is_session_static (s0)) { - clib_dlist_remove (sm->per_thread_data[cpu_index].list_pool, + clib_dlist_remove (sm->per_thread_data[thread_index].list_pool, s0->per_user_index); - clib_dlist_addtail (sm->per_thread_data[cpu_index].list_pool, + clib_dlist_addtail (sm->per_thread_data[thread_index].list_pool, s0->per_user_list_head_index, s0->per_user_index); } @@ -624,7 +624,7 @@ snat_out2in_node_fn (vlib_main_t * vm, u32 pkts_processed = 0; snat_main_t * sm = &snat_main; f64 now = vlib_time_now (vm); - u32 cpu_index = os_get_cpu_number (); + u32 thread_index = vlib_get_thread_index (); from = vlib_frame_vector_args (frame); n_left_from = frame->n_vectors; @@ -712,7 +712,7 @@ snat_out2in_node_fn (vlib_main_t * vm, { next0 = icmp_out2in_slow_path (sm, b0, ip0, icmp0, sw_if_index0, rx_fib_index0, node, - next0, now, cpu_index, &s0); + next0, now, thread_index, &s0); goto trace0; } @@ -743,7 +743,7 @@ snat_out2in_node_fn (vlib_main_t * vm, /* Create session initiated by host from external network */ s0 = create_session_for_static_mapping(sm, b0, sm0, key0, node, - cpu_index); + thread_index); if (!s0) { b0->error = node->errors[SNAT_OUT2IN_ERROR_NO_TRANSLATION]; @@ -752,7 +752,7 @@ snat_out2in_node_fn (vlib_main_t * vm, } } else - s0 = pool_elt_at_index (sm->per_thread_data[cpu_index].sessions, + s0 = pool_elt_at_index (sm->per_thread_data[thread_index].sessions, value0.value); old_addr0 = ip0->dst_address.as_u32; @@ -796,9 +796,9 @@ snat_out2in_node_fn (vlib_main_t * vm, /* Per-user LRU list maintenance for dynamic translation */ if (!snat_is_session_static (s0)) { - clib_dlist_remove (sm->per_thread_data[cpu_index].list_pool, + clib_dlist_remove (sm->per_thread_data[thread_index].list_pool, s0->per_user_index); - clib_dlist_addtail (sm->per_thread_data[cpu_index].list_pool, + clib_dlist_addtail (sm->per_thread_data[thread_index].list_pool, s0->per_user_list_head_index, s0->per_user_index); } @@ -813,7 +813,7 @@ snat_out2in_node_fn (vlib_main_t * vm, t->next_index = next0; t->session_index = ~0; if (s0) - t->session_index = s0 - sm->per_thread_data[cpu_index].sessions; + t->session_index = s0 - sm->per_thread_data[thread_index].sessions; } pkts_processed += next0 != SNAT_OUT2IN_NEXT_DROP; @@ -847,7 +847,7 @@ snat_out2in_node_fn (vlib_main_t * vm, { next1 = icmp_out2in_slow_path (sm, b1, ip1, icmp1, sw_if_index1, rx_fib_index1, node, - next1, now, cpu_index, &s1); + next1, now, thread_index, &s1); goto trace1; } @@ -878,7 +878,7 @@ snat_out2in_node_fn (vlib_main_t * vm, /* Create session initiated by host from external network */ s1 = create_session_for_static_mapping(sm, b1, sm1, key1, node, - cpu_index); + thread_index); if (!s1) { b1->error = node->errors[SNAT_OUT2IN_ERROR_NO_TRANSLATION]; @@ -887,7 +887,7 @@ snat_out2in_node_fn (vlib_main_t * vm, } } else - s1 = pool_elt_at_index (sm->per_thread_data[cpu_index].sessions, + s1 = pool_elt_at_index (sm->per_thread_data[thread_index].sessions, value1.value); old_addr1 = ip1->dst_address.as_u32; @@ -931,9 +931,9 @@ snat_out2in_node_fn (vlib_main_t * vm, /* Per-user LRU list maintenance for dynamic translation */ if (!snat_is_session_static (s1)) { - clib_dlist_remove (sm->per_thread_data[cpu_index].list_pool, + clib_dlist_remove (sm->per_thread_data[thread_index].list_pool, s1->per_user_index); - clib_dlist_addtail (sm->per_thread_data[cpu_index].list_pool, + clib_dlist_addtail (sm->per_thread_data[thread_index].list_pool, s1->per_user_list_head_index, s1->per_user_index); } @@ -948,7 +948,7 @@ snat_out2in_node_fn (vlib_main_t * vm, t->next_index = next1; t->session_index = ~0; if (s1) - t->session_index = s1 - sm->per_thread_data[cpu_index].sessions; + t->session_index = s1 - sm->per_thread_data[thread_index].sessions; } pkts_processed += next1 != SNAT_OUT2IN_NEXT_DROP; @@ -1016,7 +1016,7 @@ snat_out2in_node_fn (vlib_main_t * vm, { next0 = icmp_out2in_slow_path (sm, b0, ip0, icmp0, sw_if_index0, rx_fib_index0, node, - next0, now, cpu_index, &s0); + next0, now, thread_index, &s0); goto trace00; } @@ -1048,7 +1048,7 @@ snat_out2in_node_fn (vlib_main_t * vm, /* Create session initiated by host from external network */ s0 = create_session_for_static_mapping(sm, b0, sm0, key0, node, - cpu_index); + thread_index); if (!s0) { b0->error = node->errors[SNAT_OUT2IN_ERROR_NO_TRANSLATION]; @@ -1057,7 +1057,7 @@ snat_out2in_node_fn (vlib_main_t * vm, } } else - s0 = pool_elt_at_index (sm->per_thread_data[cpu_index].sessions, + s0 = pool_elt_at_index (sm->per_thread_data[thread_index].sessions, value0.value); old_addr0 = ip0->dst_address.as_u32; @@ -1101,9 +1101,9 @@ snat_out2in_node_fn (vlib_main_t * vm, /* Per-user LRU list maintenance for dynamic translation */ if (!snat_is_session_static (s0)) { - clib_dlist_remove (sm->per_thread_data[cpu_index].list_pool, + clib_dlist_remove (sm->per_thread_data[thread_index].list_pool, s0->per_user_index); - clib_dlist_addtail (sm->per_thread_data[cpu_index].list_pool, + clib_dlist_addtail (sm->per_thread_data[thread_index].list_pool, s0->per_user_list_head_index, s0->per_user_index); } @@ -1118,7 +1118,7 @@ snat_out2in_node_fn (vlib_main_t * vm, t->next_index = next0; t->session_index = ~0; if (s0) - t->session_index = s0 - sm->per_thread_data[cpu_index].sessions; + t->session_index = s0 - sm->per_thread_data[thread_index].sessions; } pkts_processed += next0 != SNAT_OUT2IN_NEXT_DROP; @@ -1599,7 +1599,7 @@ snat_out2in_worker_handoff_fn (vlib_main_t * vm, u32 n_left_to_next_worker = 0, *to_next_worker = 0; u32 next_worker_index = 0; u32 current_worker_index = ~0; - u32 cpu_index = os_get_cpu_number (); + u32 thread_index = vlib_get_thread_index (); ASSERT (vec_len (sm->workers)); @@ -1637,7 +1637,7 @@ snat_out2in_worker_handoff_fn (vlib_main_t * vm, next_worker_index = sm->worker_out2in_cb(ip0, rx_fib_index0); - if (PREDICT_FALSE (next_worker_index != cpu_index)) + if (PREDICT_FALSE (next_worker_index != thread_index)) { do_handoff = 1; diff --git a/src/plugins/snat/snat.h b/src/plugins/snat/snat.h index 017825c0..f4e1c5c0 100644 --- a/src/plugins/snat/snat.h +++ b/src/plugins/snat/snat.h @@ -221,7 +221,7 @@ struct snat_main_s; typedef u32 snat_icmp_match_function_t (struct snat_main_s *sm, vlib_node_runtime_t *node, - u32 cpu_index, + u32 thread_index, vlib_buffer_t *b0, snat_session_key_t *p_key, snat_session_key_t *p_value, @@ -402,22 +402,22 @@ typedef struct { } tcp_udp_header_t; u32 icmp_match_in2out_fast(snat_main_t *sm, vlib_node_runtime_t *node, - u32 cpu_index, vlib_buffer_t *b0, + u32 thread_index, vlib_buffer_t *b0, snat_session_key_t *p_key, snat_session_key_t *p_value, u8 *p_dont_translate, void *d); u32 icmp_match_in2out_slow(snat_main_t *sm, vlib_node_runtime_t *node, - u32 cpu_index, vlib_buffer_t *b0, + u32 thread_index, vlib_buffer_t *b0, snat_session_key_t *p_key, snat_session_key_t *p_value, u8 *p_dont_translate, void *d); u32 icmp_match_out2in_fast(snat_main_t *sm, vlib_node_runtime_t *node, - u32 cpu_index, vlib_buffer_t *b0, + u32 thread_index, vlib_buffer_t *b0, snat_session_key_t *p_key, snat_session_key_t *p_value, u8 *p_dont_translate, void *d); u32 icmp_match_out2in_slow(snat_main_t *sm, vlib_node_runtime_t *node, - u32 cpu_index, vlib_buffer_t *b0, + u32 thread_index, vlib_buffer_t *b0, snat_session_key_t *p_key, snat_session_key_t *p_value, u8 *p_dont_translate, void *d); diff --git a/src/vlib/buffer.c b/src/vlib/buffer.c index a517a597..be3b41ef 100644 --- a/src/vlib/buffer.c +++ b/src/vlib/buffer.c @@ -299,7 +299,7 @@ vlib_buffer_validate_alloc_free (vlib_main_t * vm, if (CLIB_DEBUG == 0) return; - ASSERT (os_get_cpu_number () == 0); + ASSERT (vlib_get_thread_index () == 0); /* smp disaster check */ if (vec_len (vlib_mains) > 1) @@ -355,7 +355,7 @@ vlib_buffer_create_free_list_helper (vlib_main_t * vm, vlib_buffer_free_list_t *f; int i; - ASSERT (os_get_cpu_number () == 0); + ASSERT (vlib_get_thread_index () == 0); if (!is_default && pool_elts (bm->buffer_free_list_pool) == 0) { @@ -474,7 +474,7 @@ vlib_buffer_delete_free_list_internal (vlib_main_t * vm, u32 free_list_index) u32 merge_index; int i; - ASSERT (os_get_cpu_number () == 0); + ASSERT (vlib_get_thread_index () == 0); f = vlib_buffer_get_free_list (vm, free_list_index); diff --git a/src/vlib/buffer_funcs.h b/src/vlib/buffer_funcs.h index 394c336a..328660a3 100644 --- a/src/vlib/buffer_funcs.h +++ b/src/vlib/buffer_funcs.h @@ -209,7 +209,7 @@ always_inline vlib_buffer_known_state_t vlib_buffer_is_known (vlib_main_t * vm, u32 buffer_index) { vlib_buffer_main_t *bm = vm->buffer_main; - ASSERT (os_get_cpu_number () == 0); + ASSERT (vlib_get_thread_index () == 0); uword *p = hash_get (bm->buffer_known_hash, buffer_index); return p ? p[0] : VLIB_BUFFER_UNKNOWN; @@ -221,7 +221,7 @@ vlib_buffer_set_known_state (vlib_main_t * vm, vlib_buffer_known_state_t state) { vlib_buffer_main_t *bm = vm->buffer_main; - ASSERT (os_get_cpu_number () == 0); + ASSERT (vlib_get_thread_index () == 0); hash_set (bm->buffer_known_hash, buffer_index, state); } diff --git a/src/vlib/cli.c b/src/vlib/cli.c index f853f655..3cc95076 100644 --- a/src/vlib/cli.c +++ b/src/vlib/cli.c @@ -709,7 +709,7 @@ test_heap_validate (vlib_main_t * vm, unformat_input_t * input, { /* *INDENT-OFF* */ foreach_vlib_main({ - heap = clib_per_cpu_mheaps[this_vlib_main->cpu_index]; + heap = clib_per_cpu_mheaps[this_vlib_main->thread_index]; mheap = mheap_header(heap); mheap->flags |= MHEAP_FLAG_VALIDATE; // Turn off small object cache because it delays detection of errors @@ -722,7 +722,7 @@ test_heap_validate (vlib_main_t * vm, unformat_input_t * input, { /* *INDENT-OFF* */ foreach_vlib_main({ - heap = clib_per_cpu_mheaps[this_vlib_main->cpu_index]; + heap = clib_per_cpu_mheaps[this_vlib_main->thread_index]; mheap = mheap_header(heap); mheap->flags &= ~MHEAP_FLAG_VALIDATE; mheap->flags |= MHEAP_FLAG_SMALL_OBJECT_CACHE; @@ -733,7 +733,7 @@ test_heap_validate (vlib_main_t * vm, unformat_input_t * input, { /* *INDENT-OFF* */ foreach_vlib_main({ - heap = clib_per_cpu_mheaps[this_vlib_main->cpu_index]; + heap = clib_per_cpu_mheaps[this_vlib_main->thread_index]; mheap = mheap_header(heap); mheap_validate(heap); }); diff --git a/src/vlib/counter.h b/src/vlib/counter.h index 17a85217..60e2055d 100644 --- a/src/vlib/counter.h +++ b/src/vlib/counter.h @@ -70,17 +70,17 @@ u32 vlib_simple_counter_n_counters (const vlib_simple_counter_main_t * cm); /** Increment a simple counter @param cm - (vlib_simple_counter_main_t *) simple counter main pointer - @param cpu_index - (u32) the current cpu index + @param thread_index - (u32) the current cpu index @param index - (u32) index of the counter to increment @param increment - (u64) quantitiy to add to the counter */ always_inline void vlib_increment_simple_counter (vlib_simple_counter_main_t * cm, - u32 cpu_index, u32 index, u64 increment) + u32 thread_index, u32 index, u64 increment) { counter_t *my_counters; - my_counters = cm->counters[cpu_index]; + my_counters = cm->counters[thread_index]; my_counters[index] += increment; } @@ -201,7 +201,7 @@ void vlib_clear_combined_counters (vlib_combined_counter_main_t * cm); /** Increment a combined counter @param cm - (vlib_combined_counter_main_t *) comined counter main pointer - @param cpu_index - (u32) the current cpu index + @param thread_index - (u32) the current cpu index @param index - (u32) index of the counter to increment @param packet_increment - (u64) number of packets to add to the counter @param byte_increment - (u64) number of bytes to add to the counter @@ -209,13 +209,13 @@ void vlib_clear_combined_counters (vlib_combined_counter_main_t * cm); always_inline void vlib_increment_combined_counter (vlib_combined_counter_main_t * cm, - u32 cpu_index, + u32 thread_index, u32 index, u64 n_packets, u64 n_bytes) { vlib_counter_t *my_counters; /* Use this CPU's counter array */ - my_counters = cm->counters[cpu_index]; + my_counters = cm->counters[thread_index]; my_counters[index].packets += n_packets; my_counters[index].bytes += n_bytes; @@ -224,14 +224,14 @@ vlib_increment_combined_counter (vlib_combined_counter_main_t * cm, /** Pre-fetch a per-thread combined counter for the given object index */ always_inline void vlib_prefetch_combined_counter (const vlib_combined_counter_main_t * cm, - u32 cpu_index, u32 index) + u32 thread_index, u32 index) { vlib_counter_t *cpu_counters; /* * This CPU's index is assumed to already be in cache */ - cpu_counters = cm->counters[cpu_index]; + cpu_counters = cm->counters[thread_index]; CLIB_PREFETCH (cpu_counters + index, CLIB_CACHE_LINE_BYTES, STORE); } diff --git a/src/vlib/error.c b/src/vlib/error.c index a2c23176..e4ed4ee3 100644 --- a/src/vlib/error.c +++ b/src/vlib/error.c @@ -149,7 +149,7 @@ vlib_register_errors (vlib_main_t * vm, vlib_node_t *n = vlib_get_node (vm, node_index); uword l; - ASSERT (os_get_cpu_number () == 0); + ASSERT (vlib_get_thread_index () == 0); /* Free up any previous error strings. */ if (n->n_errors > 0) diff --git a/src/vlib/global_funcs.h b/src/vlib/global_funcs.h index f51ec381..9dd01fbf 100644 --- a/src/vlib/global_funcs.h +++ b/src/vlib/global_funcs.h @@ -23,7 +23,7 @@ always_inline vlib_main_t * vlib_get_main (void) { vlib_main_t *vm; - vm = vlib_mains[os_get_cpu_number ()]; + vm = vlib_mains[vlib_get_thread_index ()]; ASSERT (vm); return vm; } diff --git a/src/vlib/main.c b/src/vlib/main.c index b22203f0..422d3e26 100644 --- a/src/vlib/main.c +++ b/src/vlib/main.c @@ -136,18 +136,18 @@ vlib_frame_alloc_to_node (vlib_main_t * vm, u32 to_node_index, else { f = clib_mem_alloc_aligned_no_fail (n, VLIB_FRAME_ALIGN); - f->cpu_index = vm->cpu_index; + f->thread_index = vm->thread_index; fi = vlib_frame_index_no_check (vm, f); } /* Poison frame when debugging. */ if (CLIB_DEBUG > 0) { - u32 save_cpu_index = f->cpu_index; + u32 save_thread_index = f->thread_index; memset (f, 0xfe, n); - f->cpu_index = save_cpu_index; + f->thread_index = save_thread_index; } /* Insert magic number. */ @@ -517,7 +517,7 @@ vlib_put_next_frame (vlib_main_t * vm, * a dangling frame reference. Each thread has its own copy of * the next_frames vector. */ - if (0 && r->cpu_index != next_runtime->cpu_index) + if (0 && r->thread_index != next_runtime->thread_index) { nf->frame_index = ~0; nf->flags &= ~(VLIB_FRAME_PENDING | VLIB_FRAME_IS_ALLOCATED); @@ -866,7 +866,7 @@ vlib_elog_main_loop_event (vlib_main_t * vm, : evm->node_call_elog_event_types, node_index), /* track */ - (vm->cpu_index ? &vlib_worker_threads[vm->cpu_index]. + (vm->thread_index ? &vlib_worker_threads[vm->thread_index]. elog_track : &em->default_track), /* data to log */ n_vectors); } @@ -963,7 +963,7 @@ dispatch_node (vlib_main_t * vm, vm->cpu_time_last_node_dispatch = last_time_stamp; - if (1 /* || vm->cpu_index == node->cpu_index */ ) + if (1 /* || vm->thread_index == node->thread_index */ ) { vlib_main_t *stat_vm; @@ -1029,7 +1029,7 @@ dispatch_node (vlib_main_t * vm, { u32 node_name, vector_length, is_polling; } *ed; - vlib_worker_thread_t *w = vlib_worker_threads + vm->cpu_index; + vlib_worker_thread_t *w = vlib_worker_threads + vm->thread_index; #endif if ((dispatch_state == VLIB_NODE_STATE_INTERRUPT diff --git a/src/vlib/main.h b/src/vlib/main.h index 0197b4f3..329bf073 100644 --- a/src/vlib/main.h +++ b/src/vlib/main.h @@ -156,7 +156,7 @@ typedef struct vlib_main_t uword *init_functions_called; /* to compare with node runtime */ - u32 cpu_index; + u32 thread_index; void **mbuf_alloc_list; diff --git a/src/vlib/node.c b/src/vlib/node.c index dc0a4de5..bbd3a42e 100644 --- a/src/vlib/node.c +++ b/src/vlib/node.c @@ -99,7 +99,7 @@ vlib_node_runtime_update (vlib_main_t * vm, u32 node_index, u32 next_index) vlib_pending_frame_t *pf; i32 i, j, n_insert; - ASSERT (os_get_cpu_number () == 0); + ASSERT (vlib_get_thread_index () == 0); vlib_worker_thread_barrier_sync (vm); diff --git a/src/vlib/node.h b/src/vlib/node.h index fc7e7da2..1e2f4c38 100644 --- a/src/vlib/node.h +++ b/src/vlib/node.h @@ -344,8 +344,8 @@ typedef struct vlib_frame_t /* Number of vector elements currently in frame. */ u16 n_vectors; - /* Owner cpuid / heap id */ - u16 cpu_index; + /* Owner thread / heap id */ + u16 thread_index; /* Scalar and vector arguments to next node. */ u8 arguments[0]; @@ -459,7 +459,7 @@ typedef struct vlib_node_runtime_t zero before first run of this node. */ - u16 cpu_index; /**< CPU this node runs on */ + u16 thread_index; /**< thread this node runs on */ u8 runtime_data[0]; /**< Function dependent node-runtime data. This data is diff --git a/src/vlib/node_funcs.h b/src/vlib/node_funcs.h index 1f7d94e1..54e36874 100644 --- a/src/vlib/node_funcs.h +++ b/src/vlib/node_funcs.h @@ -201,9 +201,9 @@ always_inline vlib_frame_t * vlib_get_frame_no_check (vlib_main_t * vm, uword frame_index) { vlib_frame_t *f; - u32 cpu_index = frame_index & VLIB_CPU_MASK; + u32 thread_index = frame_index & VLIB_CPU_MASK; u32 offset = frame_index & VLIB_OFFSET_MASK; - vm = vlib_mains[cpu_index]; + vm = vlib_mains[thread_index]; f = vm->heap_base + offset; return f; } @@ -215,10 +215,10 @@ vlib_frame_index_no_check (vlib_main_t * vm, vlib_frame_t * f) ASSERT (((uword) f & VLIB_CPU_MASK) == 0); - vm = vlib_mains[f->cpu_index]; + vm = vlib_mains[f->thread_index]; i = ((u8 *) f - (u8 *) vm->heap_base); - return i | f->cpu_index; + return i | f->thread_index; } always_inline vlib_frame_t * diff --git a/src/vlib/threads.c b/src/vlib/threads.c index ef3a24d3..4a111f8d 100644 --- a/src/vlib/threads.c +++ b/src/vlib/threads.c @@ -35,27 +35,12 @@ vl (void *p) vlib_worker_thread_t *vlib_worker_threads; vlib_thread_main_t vlib_thread_main; +__thread uword vlib_thread_index = 0; + uword os_get_cpu_number (void) { - void *sp; - uword n; - u32 len; - - len = vec_len (vlib_thread_stacks); - if (len == 0) - return 0; - - /* Get any old stack address. */ - sp = &sp; - - n = ((uword) sp - (uword) vlib_thread_stacks[0]) - >> VLIB_LOG2_THREAD_STACK_SIZE; - - /* "processes" have their own stacks, and they always run in thread 0 */ - n = n >= len ? 0 : n; - - return n; + return vlib_thread_index; } uword @@ -275,21 +260,6 @@ vlib_thread_init (vlib_main_t * vm) return 0; } -vlib_worker_thread_t * -vlib_alloc_thread (vlib_main_t * vm) -{ - vlib_worker_thread_t *w; - - if (vec_len (vlib_worker_threads) >= vec_len (vlib_thread_stacks)) - { - clib_warning ("out of worker threads... Quitting..."); - exit (1); - } - vec_add2 (vlib_worker_threads, w, 1); - w->thread_stack = vlib_thread_stacks[w - vlib_worker_threads]; - return w; -} - vlib_frame_queue_t * vlib_frame_queue_alloc (int nelts) { @@ -427,7 +397,7 @@ vlib_frame_queue_enqueue (vlib_main_t * vm, u32 node_runtime_index, f64 b4 = vlib_time_now_ticks (vm, before); vlib_worker_thread_barrier_check (vm, b4); /* Bad idea. Dequeue -> enqueue -> dequeue -> trouble */ - // vlib_frame_queue_dequeue (vm->cpu_index, vm, nm); + // vlib_frame_queue_dequeue (vm->thread_index, vm, nm); } elt = fq->elts + (new_tail & (fq->nelts - 1)); @@ -497,6 +467,8 @@ vlib_worker_thread_bootstrap_fn (void *arg) w->lwp = syscall (SYS_gettid); w->thread_id = pthread_self (); + vlib_thread_index = w - vlib_worker_threads; + rv = (void *) clib_calljmp ((uword (*)(uword)) w->thread_function, (uword) arg, w->thread_stack + VLIB_THREAD_STACK_SIZE); @@ -610,7 +582,9 @@ start_workers (vlib_main_t * vm) mheap_alloc (0 /* use VM */ , tr->mheap_size); else w->thread_mheap = main_heap; - w->thread_stack = vlib_thread_stacks[w - vlib_worker_threads]; + + w->thread_stack = + vlib_thread_stack_init (w - vlib_worker_threads); w->thread_function = tr->function; w->thread_function_arg = w; w->instance_id = k; @@ -630,7 +604,7 @@ start_workers (vlib_main_t * vm) vm_clone = clib_mem_alloc (sizeof (*vm_clone)); clib_memcpy (vm_clone, vlib_mains[0], sizeof (*vm_clone)); - vm_clone->cpu_index = worker_thread_index; + vm_clone->thread_index = worker_thread_index; vm_clone->heap_base = w->thread_mheap; vm_clone->mbuf_alloc_list = 0; vm_clone->init_functions_called = @@ -679,7 +653,7 @@ start_workers (vlib_main_t * vm) vec_foreach (rt, nm_clone->nodes_by_type[VLIB_NODE_TYPE_INPUT]) { vlib_node_t *n = vlib_get_node (vm, rt->node_index); - rt->cpu_index = vm_clone->cpu_index; + rt->thread_index = vm_clone->thread_index; /* copy initial runtime_data from node */ if (n->runtime_data && n->runtime_data_bytes > 0) clib_memcpy (rt->runtime_data, n->runtime_data, @@ -692,7 +666,7 @@ start_workers (vlib_main_t * vm) vec_foreach (rt, nm_clone->nodes_by_type[VLIB_NODE_TYPE_INPUT]) { vlib_node_t *n = vlib_get_node (vm, rt->node_index); - rt->cpu_index = vm_clone->cpu_index; + rt->thread_index = vm_clone->thread_index; /* copy initial runtime_data from node */ if (n->runtime_data && n->runtime_data_bytes > 0) clib_memcpy (rt->runtime_data, n->runtime_data, @@ -756,7 +730,8 @@ start_workers (vlib_main_t * vm) mheap_alloc (0 /* use VM */ , tr->mheap_size); else w->thread_mheap = main_heap; - w->thread_stack = vlib_thread_stacks[w - vlib_worker_threads]; + w->thread_stack = + vlib_thread_stack_init (w - vlib_worker_threads); w->thread_function = tr->function; w->thread_function_arg = w; w->instance_id = j; @@ -827,7 +802,7 @@ vlib_worker_thread_node_runtime_update (void) uword n_calls, uword n_vectors, uword n_clocks); - ASSERT (os_get_cpu_number () == 0); + ASSERT (vlib_get_thread_index () == 0); if (vec_len (vlib_mains) == 1) return; @@ -835,7 +810,7 @@ vlib_worker_thread_node_runtime_update (void) vm = vlib_mains[0]; nm = &vm->node_main; - ASSERT (os_get_cpu_number () == 0); + ASSERT (vlib_get_thread_index () == 0); ASSERT (*vlib_worker_threads->wait_at_barrier == 1); /* @@ -955,7 +930,7 @@ vlib_worker_thread_node_runtime_update (void) vec_foreach (rt, nm_clone->nodes_by_type[VLIB_NODE_TYPE_INTERNAL]) { vlib_node_t *n = vlib_get_node (vm, rt->node_index); - rt->cpu_index = vm_clone->cpu_index; + rt->thread_index = vm_clone->thread_index; /* copy runtime_data, will be overwritten later for existing rt */ if (n->runtime_data && n->runtime_data_bytes > 0) clib_memcpy (rt->runtime_data, n->runtime_data, @@ -981,7 +956,7 @@ vlib_worker_thread_node_runtime_update (void) vec_foreach (rt, nm_clone->nodes_by_type[VLIB_NODE_TYPE_INPUT]) { vlib_node_t *n = vlib_get_node (vm, rt->node_index); - rt->cpu_index = vm_clone->cpu_index; + rt->thread_index = vm_clone->thread_index; /* copy runtime_data, will be overwritten later for existing rt */ if (n->runtime_data && n->runtime_data_bytes > 0) clib_memcpy (rt->runtime_data, n->runtime_data, @@ -1180,7 +1155,7 @@ vlib_worker_thread_fork_fixup (vlib_fork_fixup_t which) if (vlib_mains == 0) return; - ASSERT (os_get_cpu_number () == 0); + ASSERT (vlib_get_thread_index () == 0); vlib_worker_thread_barrier_sync (vm); switch (which) @@ -1212,7 +1187,7 @@ vlib_worker_thread_barrier_sync (vlib_main_t * vm) vlib_worker_threads[0].barrier_sync_count++; - ASSERT (os_get_cpu_number () == 0); + ASSERT (vlib_get_thread_index () == 0); deadline = vlib_time_now (vm) + BARRIER_SYNC_TIMEOUT; @@ -1260,7 +1235,7 @@ vlib_worker_thread_barrier_release (vlib_main_t * vm) int vlib_frame_queue_dequeue (vlib_main_t * vm, vlib_frame_queue_main_t * fqm) { - u32 thread_id = vm->cpu_index; + u32 thread_id = vm->thread_index; vlib_frame_queue_t *fq = fqm->vlib_frame_queues[thread_id]; vlib_frame_queue_elt_t *elt; u32 *from, *to; @@ -1393,7 +1368,7 @@ vlib_worker_thread_fn (void *arg) vlib_main_t *vm = vlib_get_main (); clib_error_t *e; - ASSERT (vm->cpu_index == os_get_cpu_number ()); + ASSERT (vm->thread_index == vlib_get_thread_index ()); vlib_worker_thread_init (w); clib_time_init (&vm->clib_time); diff --git a/src/vlib/threads.h b/src/vlib/threads.h index eca4fc26..101d3d4a 100644 --- a/src/vlib/threads.h +++ b/src/vlib/threads.h @@ -153,8 +153,6 @@ typedef struct /* Called early, in thread 0's context */ clib_error_t *vlib_thread_init (vlib_main_t * vm); -vlib_worker_thread_t *vlib_alloc_thread (vlib_main_t * vm); - int vlib_frame_queue_enqueue (vlib_main_t * vm, u32 node_runtime_index, u32 frame_queue_index, vlib_frame_t * frame, vlib_frame_queue_msg_type_t type); @@ -183,12 +181,19 @@ u32 vlib_frame_queue_main_init (u32 node_index, u32 frame_queue_nelts); void vlib_worker_thread_barrier_sync (vlib_main_t * vm); void vlib_worker_thread_barrier_release (vlib_main_t * vm); +extern __thread uword vlib_thread_index; +static_always_inline uword +vlib_get_thread_index (void) +{ + return vlib_thread_index; +} + always_inline void vlib_smp_unsafe_warning (void) { if (CLIB_DEBUG > 0) { - if (os_get_cpu_number ()) + if (vlib_get_thread_index ()) fformat (stderr, "%s: SMP unsafe warning...\n", __FUNCTION__); } } @@ -331,21 +336,21 @@ vlib_num_workers () } always_inline u32 -vlib_get_worker_cpu_index (u32 worker_index) +vlib_get_worker_thread_index (u32 worker_index) { return worker_index + 1; } always_inline u32 -vlib_get_worker_index (u32 cpu_index) +vlib_get_worker_index (u32 thread_index) { - return cpu_index - 1; + return thread_index - 1; } always_inline u32 vlib_get_current_worker_index () { - return os_get_cpu_number () - 1; + return vlib_get_thread_index () - 1; } static inline void @@ -467,6 +472,8 @@ vlib_get_worker_handoff_queue_elt (u32 frame_queue_index, return elt; } +u8 *vlib_thread_stack_init (uword thread_index); + int vlib_thread_cb_register (struct vlib_main_t *vm, vlib_thread_callbacks_t * cb); diff --git a/src/vlib/unix/cj.c b/src/vlib/unix/cj.c index 33ba163a..7c1e9475 100644 --- a/src/vlib/unix/cj.c +++ b/src/vlib/unix/cj.c @@ -48,7 +48,7 @@ cj_log (u32 type, void *data0, void *data1) r = (cj_record_t *) & (cjm->records[new_tail & (cjm->num_records - 1)]); r->time = vlib_time_now (cjm->vlib_main); - r->cpu = os_get_cpu_number (); + r->thread_index = vlib_get_thread_index (); r->type = type; r->data[0] = pointer_to_uword (data0); r->data[1] = pointer_to_uword (data1); @@ -133,7 +133,8 @@ static inline void cj_dump_one_record (cj_record_t * r) { fprintf (stderr, "[%d]: %10.6f T%02d %llx %llx\n", - r->cpu, r->time, r->type, (long long unsigned int) r->data[0], + r->thread_index, r->time, r->type, + (long long unsigned int) r->data[0], (long long unsigned int) r->data[1]); } @@ -161,7 +162,7 @@ cj_dump_internal (u8 filter0_enable, u64 filter0, index = (cjm->tail + 1) & (cjm->num_records - 1); r = &(cjm->records[index]); - if (r->cpu != (u32) ~ 0) + if (r->thread_index != (u32) ~ 0) { /* Yes, dump from tail + 1 to the end */ for (i = index; i < cjm->num_records; i++) diff --git a/src/vlib/unix/cj.h b/src/vlib/unix/cj.h index 67626afe..d0a1d46e 100644 --- a/src/vlib/unix/cj.h +++ b/src/vlib/unix/cj.h @@ -23,7 +23,7 @@ typedef struct { f64 time; - u32 cpu; + u32 thread_index; u32 type; u64 data[2]; } cj_record_t; diff --git a/src/vlib/unix/main.c b/src/vlib/unix/main.c index 6b96cc0d..db5ddd64 100644 --- a/src/vlib/unix/main.c +++ b/src/vlib/unix/main.c @@ -510,13 +510,28 @@ thread0 (uword arg) return i; } +u8 * +vlib_thread_stack_init (uword thread_index) +{ + vec_validate (vlib_thread_stacks, thread_index); + vlib_thread_stacks[thread_index] = clib_mem_alloc_aligned + (VLIB_THREAD_STACK_SIZE, VLIB_THREAD_STACK_SIZE); + + /* + * Disallow writes to the bottom page of the stack, to + * catch stack overflows. + */ + if (mprotect (vlib_thread_stacks[thread_index], + clib_mem_get_page_size (), PROT_READ) < 0) + clib_unix_warning ("thread stack"); + return vlib_thread_stacks[thread_index]; +} + int vlib_unix_main (int argc, char *argv[]) { vlib_main_t *vm = &vlib_global_main; /* one and only time for this! */ - vlib_thread_main_t *tm = &vlib_thread_main; unformat_input_t input; - u8 *thread_stacks; clib_error_t *e; int i; @@ -548,29 +563,9 @@ vlib_unix_main (int argc, char *argv[]) } unformat_free (&input); - /* - * allocate n x VLIB_THREAD_STACK_SIZE stacks, aligned to a - * VLIB_THREAD_STACK_SIZE boundary - * See also: os_get_cpu_number() in vlib/vlib/threads.c - */ - thread_stacks = clib_mem_alloc_aligned - ((uword) tm->n_thread_stacks * VLIB_THREAD_STACK_SIZE, - VLIB_THREAD_STACK_SIZE); - - vec_validate (vlib_thread_stacks, tm->n_thread_stacks - 1); - for (i = 0; i < vec_len (vlib_thread_stacks); i++) - { - vlib_thread_stacks[i] = thread_stacks; - - /* - * Disallow writes to the bottom page of the stack, to - * catch stack overflows. - */ - if (mprotect (thread_stacks, clib_mem_get_page_size (), PROT_READ) < 0) - clib_unix_warning ("thread stack"); + vlib_thread_stack_init (0); - thread_stacks += VLIB_THREAD_STACK_SIZE; - } + vlib_thread_index = 0; i = clib_calljmp (thread0, (uword) vm, (void *) (vlib_thread_stacks[0] + diff --git a/src/vnet/adj/adj_l2.c b/src/vnet/adj/adj_l2.c index f68e54e0..20d70dd4 100644 --- a/src/vnet/adj/adj_l2.c +++ b/src/vnet/adj/adj_l2.c @@ -52,7 +52,7 @@ adj_l2_rewrite_inline (vlib_main_t * vm, { u32 * from = vlib_frame_vector_args (frame); u32 n_left_from, n_left_to_next, * to_next, next_index; - u32 cpu_index = os_get_cpu_number(); + u32 thread_index = vlib_get_thread_index(); ethernet_main_t * em = ðernet_main; n_left_from = frame->n_vectors; @@ -93,7 +93,7 @@ adj_l2_rewrite_inline (vlib_main_t * vm, vnet_buffer(p0)->sw_if_index[VLIB_TX] = adj0->rewrite_header.sw_if_index; vlib_increment_combined_counter(&adjacency_counters, - cpu_index, + thread_index, adj_index0, /* packet increment */ 0, /* byte increment */ rw_len0); diff --git a/src/vnet/adj/adj_midchain.c b/src/vnet/adj/adj_midchain.c index e8087f08..5756de43 100644 --- a/src/vnet/adj/adj_midchain.c +++ b/src/vnet/adj/adj_midchain.c @@ -49,7 +49,7 @@ adj_midchain_tx_inline (vlib_main_t * vm, u32 next_index; vnet_main_t *vnm = vnet_get_main (); vnet_interface_main_t *im = &vnm->interface_main; - u32 cpu_index = vm->cpu_index; + u32 thread_index = vm->thread_index; /* Vector of buffer / pkt indices we're supposed to process */ from = vlib_frame_vector_args (frame); @@ -124,13 +124,13 @@ adj_midchain_tx_inline (vlib_main_t * vm, { vlib_increment_combined_counter (im->combined_sw_if_counters + VNET_INTERFACE_COUNTER_TX, - cpu_index, + thread_index, adj0->rewrite_header.sw_if_index, 1, vlib_buffer_length_in_chain (vm, b0)); vlib_increment_combined_counter (im->combined_sw_if_counters + VNET_INTERFACE_COUNTER_TX, - cpu_index, + thread_index, adj1->rewrite_header.sw_if_index, 1, vlib_buffer_length_in_chain (vm, b1)); @@ -181,7 +181,7 @@ adj_midchain_tx_inline (vlib_main_t * vm, { vlib_increment_combined_counter (im->combined_sw_if_counters + VNET_INTERFACE_COUNTER_TX, - cpu_index, + thread_index, adj0->rewrite_header.sw_if_index, 1, vlib_buffer_length_in_chain (vm, b0)); diff --git a/src/vnet/adj/adj_nsh.c b/src/vnet/adj/adj_nsh.c index 9a0f9d8b..128570b0 100644 --- a/src/vnet/adj/adj_nsh.c +++ b/src/vnet/adj/adj_nsh.c @@ -53,7 +53,7 @@ adj_nsh_rewrite_inline (vlib_main_t * vm, { u32 * from = vlib_frame_vector_args (frame); u32 n_left_from, n_left_to_next, * to_next, next_index; - u32 cpu_index = os_get_cpu_number(); + u32 thread_index = vlib_get_thread_index(); n_left_from = frame->n_vectors; next_index = node->cached_next_index; @@ -94,7 +94,7 @@ adj_nsh_rewrite_inline (vlib_main_t * vm, vnet_buffer(p0)->ip.save_rewrite_length = rw_len0; vlib_increment_combined_counter(&adjacency_counters, - cpu_index, + thread_index, adj_index0, /* packet increment */ 0, /* byte increment */ rw_len0); diff --git a/src/vnet/classify/vnet_classify.c b/src/vnet/classify/vnet_classify.c index 98842a48..70a189b0 100644 --- a/src/vnet/classify/vnet_classify.c +++ b/src/vnet/classify/vnet_classify.c @@ -251,12 +251,12 @@ static inline void make_working_copy vnet_classify_entry_##size##_t * working_copy##size = 0; foreach_size_in_u32x4; #undef _ - u32 cpu_number = os_get_cpu_number(); + u32 thread_index = vlib_get_thread_index(); - if (cpu_number >= vec_len (t->working_copies)) + if (thread_index >= vec_len (t->working_copies)) { oldheap = clib_mem_set_heap (t->mheap); - vec_validate (t->working_copies, cpu_number); + vec_validate (t->working_copies, thread_index); clib_mem_set_heap (oldheap); } @@ -265,7 +265,7 @@ static inline void make_working_copy * updates from multiple threads will not result in sporadic, spurious * lookup failures. */ - working_copy = t->working_copies[cpu_number]; + working_copy = t->working_copies[thread_index]; t->saved_bucket.as_u64 = b->as_u64; oldheap = clib_mem_set_heap (t->mheap); @@ -290,7 +290,7 @@ static inline void make_working_copy default: abort(); } - t->working_copies[cpu_number] = working_copy; + t->working_copies[thread_index] = working_copy; } _vec_len(working_copy) = (1<log2_pages)*t->entries_per_page; @@ -318,7 +318,7 @@ static inline void make_working_copy working_bucket.offset = vnet_classify_get_offset (t, working_copy); CLIB_MEMORY_BARRIER(); b->as_u64 = working_bucket.as_u64; - t->working_copies[cpu_number] = working_copy; + t->working_copies[thread_index] = working_copy; } static vnet_classify_entry_t * @@ -387,7 +387,7 @@ int vnet_classify_add_del (vnet_classify_table_t * t, int i; u64 hash, new_hash; u32 new_log2_pages; - u32 cpu_number = os_get_cpu_number(); + u32 thread_index = vlib_get_thread_index(); u8 * key_minus_skip; ASSERT ((add_v->flags & VNET_CLASSIFY_ENTRY_FREE) == 0); @@ -498,7 +498,7 @@ int vnet_classify_add_del (vnet_classify_table_t * t, new_log2_pages = t->saved_bucket.log2_pages + 1; expand_again: - working_copy = t->working_copies[cpu_number]; + working_copy = t->working_copies[thread_index]; new_v = split_and_rehash (t, working_copy, new_log2_pages); if (new_v == 0) diff --git a/src/vnet/cop/ip4_whitelist.c b/src/vnet/cop/ip4_whitelist.c index 6ef3d7d7..1b5e336b 100644 --- a/src/vnet/cop/ip4_whitelist.c +++ b/src/vnet/cop/ip4_whitelist.c @@ -60,7 +60,7 @@ ip4_cop_whitelist_node_fn (vlib_main_t * vm, cop_feature_type_t next_index; cop_main_t *cm = &cop_main; vlib_combined_counter_main_t * vcm = &load_balance_main.lbm_via_counters; - u32 cpu_index = vm->cpu_index; + u32 thread_index = vm->thread_index; from = vlib_frame_vector_args (frame); n_left_from = frame->n_vectors; @@ -177,12 +177,12 @@ ip4_cop_whitelist_node_fn (vlib_main_t * vm, dpo1 = load_balance_get_bucket_i(lb1, 0); vlib_increment_combined_counter - (vcm, cpu_index, lb_index0, 1, + (vcm, thread_index, lb_index0, 1, vlib_buffer_length_in_chain (vm, b0) + sizeof(ethernet_header_t)); vlib_increment_combined_counter - (vcm, cpu_index, lb_index1, 1, + (vcm, thread_index, lb_index1, 1, vlib_buffer_length_in_chain (vm, b1) + sizeof(ethernet_header_t)); @@ -273,7 +273,7 @@ ip4_cop_whitelist_node_fn (vlib_main_t * vm, dpo0 = load_balance_get_bucket_i(lb0, 0); vlib_increment_combined_counter - (vcm, cpu_index, lb_index0, 1, + (vcm, thread_index, lb_index0, 1, vlib_buffer_length_in_chain (vm, b0) + sizeof(ethernet_header_t)); diff --git a/src/vnet/cop/ip6_whitelist.c b/src/vnet/cop/ip6_whitelist.c index c2e16ccf..f3fe62e3 100644 --- a/src/vnet/cop/ip6_whitelist.c +++ b/src/vnet/cop/ip6_whitelist.c @@ -61,7 +61,7 @@ ip6_cop_whitelist_node_fn (vlib_main_t * vm, cop_main_t *cm = &cop_main; ip6_main_t * im6 = &ip6_main; vlib_combined_counter_main_t * vcm = &load_balance_main.lbm_via_counters; - u32 cpu_index = vm->cpu_index; + u32 thread_index = vm->thread_index; from = vlib_frame_vector_args (frame); n_left_from = frame->n_vectors; @@ -153,12 +153,12 @@ ip6_cop_whitelist_node_fn (vlib_main_t * vm, dpo1 = load_balance_get_bucket_i(lb1, 0); vlib_increment_combined_counter - (vcm, cpu_index, lb_index0, 1, + (vcm, thread_index, lb_index0, 1, vlib_buffer_length_in_chain (vm, b0) + sizeof(ethernet_header_t)); vlib_increment_combined_counter - (vcm, cpu_index, lb_index1, 1, + (vcm, thread_index, lb_index1, 1, vlib_buffer_length_in_chain (vm, b1) + sizeof(ethernet_header_t)); @@ -233,7 +233,7 @@ ip6_cop_whitelist_node_fn (vlib_main_t * vm, dpo0 = load_balance_get_bucket_i(lb0, 0); vlib_increment_combined_counter - (vcm, cpu_index, lb_index0, 1, + (vcm, thread_index, lb_index0, 1, vlib_buffer_length_in_chain (vm, b0) + sizeof(ethernet_header_t)); diff --git a/src/vnet/devices/af_packet/node.c b/src/vnet/devices/af_packet/node.c index ba337f3f..76980102 100644 --- a/src/vnet/devices/af_packet/node.c +++ b/src/vnet/devices/af_packet/node.c @@ -124,7 +124,7 @@ af_packet_device_input_fn (vlib_main_t * vm, vlib_node_runtime_t * node, u32 frame_num = apif->rx_req->tp_frame_nr; u8 *block_start = apif->rx_ring + block * block_size; uword n_trace = vlib_get_trace_count (vm, node); - u32 cpu_index = os_get_cpu_number (); + u32 thread_index = vlib_get_thread_index (); u32 n_buffer_bytes = vlib_buffer_free_list_buffer_size (vm, VLIB_BUFFER_DEFAULT_FREE_LIST_INDEX); u32 min_bufs = apif->rx_req->tp_frame_size / n_buffer_bytes; @@ -132,15 +132,15 @@ af_packet_device_input_fn (vlib_main_t * vm, vlib_node_runtime_t * node, if (apif->per_interface_next_index != ~0) next_index = apif->per_interface_next_index; - n_free_bufs = vec_len (apm->rx_buffers[cpu_index]); + n_free_bufs = vec_len (apm->rx_buffers[thread_index]); if (PREDICT_FALSE (n_free_bufs < VLIB_FRAME_SIZE)) { - vec_validate (apm->rx_buffers[cpu_index], + vec_validate (apm->rx_buffers[thread_index], VLIB_FRAME_SIZE + n_free_bufs - 1); n_free_bufs += - vlib_buffer_alloc (vm, &apm->rx_buffers[cpu_index][n_free_bufs], + vlib_buffer_alloc (vm, &apm->rx_buffers[thread_index][n_free_bufs], VLIB_FRAME_SIZE); - _vec_len (apm->rx_buffers[cpu_index]) = n_free_bufs; + _vec_len (apm->rx_buffers[thread_index]) = n_free_bufs; } rx_frame = apif->next_rx_frame; @@ -163,11 +163,11 @@ af_packet_device_input_fn (vlib_main_t * vm, vlib_node_runtime_t * node, { /* grab free buffer */ u32 last_empty_buffer = - vec_len (apm->rx_buffers[cpu_index]) - 1; + vec_len (apm->rx_buffers[thread_index]) - 1; prev_bi0 = bi0; - bi0 = apm->rx_buffers[cpu_index][last_empty_buffer]; + bi0 = apm->rx_buffers[thread_index][last_empty_buffer]; b0 = vlib_get_buffer (vm, bi0); - _vec_len (apm->rx_buffers[cpu_index]) = last_empty_buffer; + _vec_len (apm->rx_buffers[thread_index]) = last_empty_buffer; n_free_bufs--; /* copy data */ @@ -236,9 +236,9 @@ af_packet_device_input_fn (vlib_main_t * vm, vlib_node_runtime_t * node, vlib_increment_combined_counter (vnet_get_main ()->interface_main.combined_sw_if_counters + VNET_INTERFACE_COUNTER_RX, - os_get_cpu_number (), apif->hw_if_index, n_rx_packets, n_rx_bytes); + vlib_get_thread_index (), apif->hw_if_index, n_rx_packets, n_rx_bytes); - vnet_device_increment_rx_packets (cpu_index, n_rx_packets); + vnet_device_increment_rx_packets (thread_index, n_rx_packets); return n_rx_packets; } diff --git a/src/vnet/devices/devices.c b/src/vnet/devices/devices.c index 41645220..5e5e812c 100644 --- a/src/vnet/devices/devices.c +++ b/src/vnet/devices/devices.c @@ -104,7 +104,7 @@ vnet_device_queue_sort (void *a1, void *a2) void vnet_device_input_assign_thread (u32 hw_if_index, - u16 queue_id, uword cpu_index) + u16 queue_id, uword thread_index) { vnet_main_t *vnm = vnet_get_main (); vnet_device_main_t *vdm = &vnet_device_main; @@ -115,19 +115,19 @@ vnet_device_input_assign_thread (u32 hw_if_index, ASSERT (hw->input_node_index > 0); - if (vdm->first_worker_cpu_index == 0) - cpu_index = 0; + if (vdm->first_worker_thread_index == 0) + thread_index = 0; - if (cpu_index != 0 && - (cpu_index < vdm->first_worker_cpu_index || - cpu_index > vdm->last_worker_cpu_index)) + if (thread_index != 0 && + (thread_index < vdm->first_worker_thread_index || + thread_index > vdm->last_worker_thread_index)) { - cpu_index = vdm->next_worker_cpu_index++; - if (vdm->next_worker_cpu_index > vdm->last_worker_cpu_index) - vdm->next_worker_cpu_index = vdm->first_worker_cpu_index; + thread_index = vdm->next_worker_thread_index++; + if (vdm->next_worker_thread_index > vdm->last_worker_thread_index) + vdm->next_worker_thread_index = vdm->first_worker_thread_index; } - vm = vlib_mains[cpu_index]; + vm = vlib_mains[thread_index]; rt = vlib_node_get_runtime_data (vm, hw->input_node_index); vec_add2 (rt->devices_and_queues, dq, 1); @@ -136,33 +136,33 @@ vnet_device_input_assign_thread (u32 hw_if_index, dq->queue_id = queue_id; vec_sort_with_function (rt->devices_and_queues, vnet_device_queue_sort); - vec_validate (hw->input_node_cpu_index_by_queue, queue_id); - hw->input_node_cpu_index_by_queue[queue_id] = cpu_index; + vec_validate (hw->input_node_thread_index_by_queue, queue_id); + hw->input_node_thread_index_by_queue[queue_id] = thread_index; } static int vnet_device_input_unassign_thread (u32 hw_if_index, u16 queue_id, - uword cpu_index) + uword thread_index) { vnet_main_t *vnm = vnet_get_main (); vnet_hw_interface_t *hw = vnet_get_hw_interface (vnm, hw_if_index); vnet_device_input_runtime_t *rt; vnet_device_and_queue_t *dq; - uword old_cpu_index; + uword old_thread_index; - if (hw->input_node_cpu_index_by_queue == 0) + if (hw->input_node_thread_index_by_queue == 0) return VNET_API_ERROR_INVALID_INTERFACE; - if (vec_len (hw->input_node_cpu_index_by_queue) < queue_id + 1) + if (vec_len (hw->input_node_thread_index_by_queue) < queue_id + 1) return VNET_API_ERROR_INVALID_INTERFACE; - old_cpu_index = hw->input_node_cpu_index_by_queue[queue_id]; + old_thread_index = hw->input_node_thread_index_by_queue[queue_id]; - if (old_cpu_index == cpu_index) + if (old_thread_index == thread_index) return 0; rt = - vlib_node_get_runtime_data (vlib_mains[old_cpu_index], + vlib_node_get_runtime_data (vlib_mains[old_thread_index], hw->input_node_index); vec_foreach (dq, rt->devices_and_queues) @@ -240,7 +240,7 @@ set_device_placement (vlib_main_t * vm, unformat_input_t * input, vnet_device_main_t *vdm = &vnet_device_main; u32 hw_if_index = (u32) ~ 0; u32 queue_id = (u32) 0; - u32 cpu_index = (u32) ~ 0; + u32 thread_index = (u32) ~ 0; int rv; if (!unformat_user (input, unformat_line_input, line_input)) @@ -253,10 +253,10 @@ set_device_placement (vlib_main_t * vm, unformat_input_t * input, ; else if (unformat (line_input, "queue %d", &queue_id)) ; - else if (unformat (line_input, "main", &cpu_index)) - cpu_index = 0; - else if (unformat (line_input, "worker %d", &cpu_index)) - cpu_index += vdm->first_worker_cpu_index; + else if (unformat (line_input, "main", &thread_index)) + thread_index = 0; + else if (unformat (line_input, "worker %d", &thread_index)) + thread_index += vdm->first_worker_thread_index; else { error = clib_error_return (0, "parse error: '%U'", @@ -271,16 +271,17 @@ set_device_placement (vlib_main_t * vm, unformat_input_t * input, if (hw_if_index == (u32) ~ 0) return clib_error_return (0, "please specify valid interface name"); - if (cpu_index > vdm->last_worker_cpu_index) + if (thread_index > vdm->last_worker_thread_index) return clib_error_return (0, "please specify valid worker thread or main"); - rv = vnet_device_input_unassign_thread (hw_if_index, queue_id, cpu_index); + rv = + vnet_device_input_unassign_thread (hw_if_index, queue_id, thread_index); if (rv) return clib_error_return (0, "not found"); - vnet_device_input_assign_thread (hw_if_index, queue_id, cpu_index); + vnet_device_input_assign_thread (hw_if_index, queue_id, thread_index); return 0; } @@ -326,9 +327,9 @@ vnet_device_init (vlib_main_t * vm) tr = p ? (vlib_thread_registration_t *) p[0] : 0; if (tr && tr->count > 0) { - vdm->first_worker_cpu_index = tr->first_index; - vdm->next_worker_cpu_index = tr->first_index; - vdm->last_worker_cpu_index = tr->first_index + tr->count - 1; + vdm->first_worker_thread_index = tr->first_index; + vdm->next_worker_thread_index = tr->first_index; + vdm->last_worker_thread_index = tr->first_index + tr->count - 1; } return 0; } diff --git a/src/vnet/devices/devices.h b/src/vnet/devices/devices.h index bbb29fe3..966f8302 100644 --- a/src/vnet/devices/devices.h +++ b/src/vnet/devices/devices.h @@ -50,9 +50,9 @@ typedef struct typedef struct { vnet_device_per_worker_data_t *workers; - uword first_worker_cpu_index; - uword last_worker_cpu_index; - uword next_worker_cpu_index; + uword first_worker_thread_index; + uword last_worker_thread_index; + uword next_worker_thread_index; } vnet_device_main_t; typedef struct @@ -80,7 +80,7 @@ vnet_set_device_input_node (u32 hw_if_index, u32 node_index) } void vnet_device_input_assign_thread (u32 hw_if_index, u16 queue_id, - uword cpu_index); + uword thread_index); static inline u64 vnet_get_aggregate_rx_packets (void) @@ -95,12 +95,12 @@ vnet_get_aggregate_rx_packets (void) } static inline void -vnet_device_increment_rx_packets (u32 cpu_index, u64 count) +vnet_device_increment_rx_packets (u32 thread_index, u64 count) { vnet_device_main_t *vdm = &vnet_device_main; vnet_device_per_worker_data_t *pwd; - pwd = vec_elt_at_index (vdm->workers, cpu_index); + pwd = vec_elt_at_index (vdm->workers, thread_index); pwd->aggregate_rx_packets += count; } @@ -117,9 +117,9 @@ vnet_device_input_set_interrupt_pending (vnet_main_t * vnm, u32 hw_if_index, { vnet_hw_interface_t *hw = vnet_get_hw_interface (vnm, hw_if_index); - ASSERT (queue_id < vec_len (hw->input_node_cpu_index_by_queue)); - u32 cpu_index = hw->input_node_cpu_index_by_queue[queue_id]; - vlib_node_set_interrupt_pending (vlib_mains[cpu_index], + ASSERT (queue_id < vec_len (hw->input_node_thread_index_by_queue)); + u32 thread_index = hw->input_node_thread_index_by_queue[queue_id]; + vlib_node_set_interrupt_pending (vlib_mains[thread_index], hw->input_node_index); } diff --git a/src/vnet/devices/netmap/node.c b/src/vnet/devices/netmap/node.c index 68ea7832..e120eeae 100644 --- a/src/vnet/devices/netmap/node.c +++ b/src/vnet/devices/netmap/node.c @@ -98,22 +98,22 @@ netmap_device_input_fn (vlib_main_t * vm, vlib_node_runtime_t * node, u32 n_free_bufs; struct netmap_ring *ring; int cur_ring; - u32 cpu_index = os_get_cpu_number (); + u32 thread_index = vlib_get_thread_index (); u32 n_buffer_bytes = vlib_buffer_free_list_buffer_size (vm, VLIB_BUFFER_DEFAULT_FREE_LIST_INDEX); if (nif->per_interface_next_index != ~0) next_index = nif->per_interface_next_index; - n_free_bufs = vec_len (nm->rx_buffers[cpu_index]); + n_free_bufs = vec_len (nm->rx_buffers[thread_index]); if (PREDICT_FALSE (n_free_bufs < VLIB_FRAME_SIZE)) { - vec_validate (nm->rx_buffers[cpu_index], + vec_validate (nm->rx_buffers[thread_index], VLIB_FRAME_SIZE + n_free_bufs - 1); n_free_bufs += - vlib_buffer_alloc (vm, &nm->rx_buffers[cpu_index][n_free_bufs], + vlib_buffer_alloc (vm, &nm->rx_buffers[thread_index][n_free_bufs], VLIB_FRAME_SIZE); - _vec_len (nm->rx_buffers[cpu_index]) = n_free_bufs; + _vec_len (nm->rx_buffers[thread_index]) = n_free_bufs; } cur_ring = nif->first_rx_ring; @@ -163,11 +163,11 @@ netmap_device_input_fn (vlib_main_t * vm, vlib_node_runtime_t * node, vlib_buffer_t *b0; /* grab free buffer */ u32 last_empty_buffer = - vec_len (nm->rx_buffers[cpu_index]) - 1; + vec_len (nm->rx_buffers[thread_index]) - 1; prev_bi0 = bi0; - bi0 = nm->rx_buffers[cpu_index][last_empty_buffer]; + bi0 = nm->rx_buffers[thread_index][last_empty_buffer]; b0 = vlib_get_buffer (vm, bi0); - _vec_len (nm->rx_buffers[cpu_index]) = last_empty_buffer; + _vec_len (nm->rx_buffers[thread_index]) = last_empty_buffer; n_free_bufs--; /* copy data */ @@ -247,9 +247,9 @@ netmap_device_input_fn (vlib_main_t * vm, vlib_node_runtime_t * node, vlib_increment_combined_counter (vnet_get_main ()->interface_main.combined_sw_if_counters + VNET_INTERFACE_COUNTER_RX, - os_get_cpu_number (), nif->hw_if_index, n_rx_packets, n_rx_bytes); + vlib_get_thread_index (), nif->hw_if_index, n_rx_packets, n_rx_bytes); - vnet_device_increment_rx_packets (cpu_index, n_rx_packets); + vnet_device_increment_rx_packets (thread_index, n_rx_packets); return n_rx_packets; } @@ -260,7 +260,7 @@ netmap_input_fn (vlib_main_t * vm, vlib_node_runtime_t * node, { int i; u32 n_rx_packets = 0; - u32 cpu_index = os_get_cpu_number (); + u32 thread_index = vlib_get_thread_index (); netmap_main_t *nm = &netmap_main; netmap_if_t *nmi; @@ -269,7 +269,7 @@ netmap_input_fn (vlib_main_t * vm, vlib_node_runtime_t * node, nmi = vec_elt_at_index (nm->interfaces, i); if (nmi->is_admin_up && (i % nm->input_cpu_count) == - (cpu_index - nm->input_cpu_first_index)) + (thread_index - nm->input_cpu_first_index)) n_rx_packets += netmap_device_input_fn (vm, node, frame, nmi); } diff --git a/src/vnet/devices/ssvm/node.c b/src/vnet/devices/ssvm/node.c index a6c9dfd7..539b4161 100644 --- a/src/vnet/devices/ssvm/node.c +++ b/src/vnet/devices/ssvm/node.c @@ -89,7 +89,7 @@ ssvm_eth_device_input (ssvm_eth_main_t * em, ethernet_header_t *eh0; u16 type0; u32 n_rx_bytes = 0, l3_offset0; - u32 cpu_index = os_get_cpu_number (); + u32 thread_index = vlib_get_thread_index (); u32 trace_cnt __attribute__ ((unused)) = vlib_get_trace_count (vm, node); volatile u32 *lock; u32 *elt_indices; @@ -284,10 +284,10 @@ out: vlib_increment_combined_counter (vnet_get_main ()->interface_main.combined_sw_if_counters - + VNET_INTERFACE_COUNTER_RX, cpu_index, + + VNET_INTERFACE_COUNTER_RX, thread_index, intfc->vlib_hw_if_index, rx_queue_index, n_rx_bytes); - vnet_device_increment_rx_packets (cpu_index, rx_queue_index); + vnet_device_increment_rx_packets (thread_index, rx_queue_index); return rx_queue_index; } diff --git a/src/vnet/devices/virtio/vhost-user.c b/src/vnet/devices/virtio/vhost-user.c index 00807dc0..5e720f65 100644 --- a/src/vnet/devices/virtio/vhost-user.c +++ b/src/vnet/devices/virtio/vhost-user.c @@ -331,7 +331,7 @@ vhost_user_tx_thread_placement (vhost_user_intf_t * vui) { //Let's try to assign one queue to each thread u32 qid = 0; - u32 cpu_index = 0; + u32 thread_index = 0; vui->use_tx_spinlock = 0; while (1) { @@ -341,20 +341,21 @@ vhost_user_tx_thread_placement (vhost_user_intf_t * vui) if (!rxvq->started || !rxvq->enabled) continue; - vui->per_cpu_tx_qid[cpu_index] = qid; - cpu_index++; - if (cpu_index == vlib_get_thread_main ()->n_vlib_mains) + vui->per_cpu_tx_qid[thread_index] = qid; + thread_index++; + if (thread_index == vlib_get_thread_main ()->n_vlib_mains) return; } //We need to loop, meaning the spinlock has to be used vui->use_tx_spinlock = 1; - if (cpu_index == 0) + if (thread_index == 0) { //Could not find a single valid one - for (cpu_index = 0; - cpu_index < vlib_get_thread_main ()->n_vlib_mains; cpu_index++) + for (thread_index = 0; + thread_index < vlib_get_thread_main ()->n_vlib_mains; + thread_index++) { - vui->per_cpu_tx_qid[cpu_index] = 0; + vui->per_cpu_tx_qid[thread_index] = 0; } return; } @@ -368,7 +369,7 @@ vhost_user_rx_thread_placement () vhost_user_intf_t *vui; vhost_cpu_t *vhc; u32 *workers = 0; - u32 cpu_index; + u32 thread_index; vlib_main_t *vm; //Let's list all workers cpu indexes @@ -400,9 +401,9 @@ vhost_user_rx_thread_placement () continue; i %= vec_len (vui_workers); - cpu_index = vui_workers[i]; + thread_index = vui_workers[i]; i++; - vhc = &vum->cpus[cpu_index]; + vhc = &vum->cpus[thread_index]; iaq.qid = qid; iaq.vhost_iface_index = vui - vum->vhost_user_interfaces; @@ -429,14 +430,14 @@ vhost_user_rx_thread_placement () vhc->operation_mode = mode; } - for (cpu_index = vum->input_cpu_first_index; - cpu_index < vum->input_cpu_first_index + vum->input_cpu_count; - cpu_index++) + for (thread_index = vum->input_cpu_first_index; + thread_index < vum->input_cpu_first_index + vum->input_cpu_count; + thread_index++) { vlib_node_state_t state = VLIB_NODE_STATE_POLLING; - vhc = &vum->cpus[cpu_index]; - vm = vlib_mains ? vlib_mains[cpu_index] : &vlib_global_main; + vhc = &vum->cpus[thread_index]; + vm = vlib_mains ? vlib_mains[thread_index] : &vlib_global_main; switch (vhc->operation_mode) { case VHOST_USER_INTERRUPT_MODE: @@ -532,7 +533,7 @@ vhost_user_set_interrupt_pending (vhost_user_intf_t * vui, u32 ifq) { vhost_user_main_t *vum = &vhost_user_main; vhost_cpu_t *vhc; - u32 cpu_index; + u32 thread_index; vhost_iface_and_queue_t *vhiq; vlib_main_t *vm; u32 ifq2; @@ -553,8 +554,8 @@ vhost_user_set_interrupt_pending (vhost_user_intf_t * vui, u32 ifq) if ((vhiq->vhost_iface_index == (ifq >> 8)) && (VHOST_VRING_IDX_TX (vhiq->qid) == (ifq & 0xff))) { - cpu_index = vhc - vum->cpus; - vm = vlib_mains ? vlib_mains[cpu_index] : &vlib_global_main; + thread_index = vhc - vum->cpus; + vm = vlib_mains ? vlib_mains[thread_index] : &vlib_global_main; /* * Convert RX virtqueue number in the lower byte to vring * queue index for the input node process. Top bytes contain @@ -1592,7 +1593,7 @@ vhost_user_if_input (vlib_main_t * vm, u32 n_trace = vlib_get_trace_count (vm, node); u16 qsz_mask; u32 map_hint = 0; - u16 cpu_index = os_get_cpu_number (); + u16 thread_index = vlib_get_thread_index (); u16 copy_len = 0; { @@ -1651,32 +1652,32 @@ vhost_user_if_input (vlib_main_t * vm, * in the loop and come back later. This is not an issue as for big packet, * processing cost really comes from the memory copy. */ - if (PREDICT_FALSE (vum->cpus[cpu_index].rx_buffers_len < n_left + 1)) + if (PREDICT_FALSE (vum->cpus[thread_index].rx_buffers_len < n_left + 1)) { - u32 curr_len = vum->cpus[cpu_index].rx_buffers_len; - vum->cpus[cpu_index].rx_buffers_len += + u32 curr_len = vum->cpus[thread_index].rx_buffers_len; + vum->cpus[thread_index].rx_buffers_len += vlib_buffer_alloc_from_free_list (vm, - vum->cpus[cpu_index].rx_buffers + + vum->cpus[thread_index].rx_buffers + curr_len, VHOST_USER_RX_BUFFERS_N - curr_len, VLIB_BUFFER_DEFAULT_FREE_LIST_INDEX); if (PREDICT_FALSE - (vum->cpus[cpu_index].rx_buffers_len < + (vum->cpus[thread_index].rx_buffers_len < VHOST_USER_RX_BUFFER_STARVATION)) { /* In case of buffer starvation, discard some packets from the queue * and log the event. * We keep doing best effort for the remaining packets. */ - u32 flush = (n_left + 1 > vum->cpus[cpu_index].rx_buffers_len) ? - n_left + 1 - vum->cpus[cpu_index].rx_buffers_len : 1; + u32 flush = (n_left + 1 > vum->cpus[thread_index].rx_buffers_len) ? + n_left + 1 - vum->cpus[thread_index].rx_buffers_len : 1; flush = vhost_user_rx_discard_packet (vm, vui, txvq, flush); n_left -= flush; vlib_increment_simple_counter (vnet_main. interface_main.sw_if_counters + VNET_INTERFACE_COUNTER_DROP, - os_get_cpu_number (), + vlib_get_thread_index (), vui->sw_if_index, flush); vlib_error_count (vm, vhost_user_input_node.index, @@ -1696,7 +1697,7 @@ vhost_user_if_input (vlib_main_t * vm, u32 desc_data_offset; vring_desc_t *desc_table = txvq->desc; - if (PREDICT_FALSE (vum->cpus[cpu_index].rx_buffers_len <= 1)) + if (PREDICT_FALSE (vum->cpus[thread_index].rx_buffers_len <= 1)) { /* Not enough rx_buffers * Note: We yeld on 1 so we don't need to do an additional @@ -1707,17 +1708,18 @@ vhost_user_if_input (vlib_main_t * vm, } desc_current = txvq->avail->ring[txvq->last_avail_idx & qsz_mask]; - vum->cpus[cpu_index].rx_buffers_len--; - bi_current = (vum->cpus[cpu_index].rx_buffers) - [vum->cpus[cpu_index].rx_buffers_len]; + vum->cpus[thread_index].rx_buffers_len--; + bi_current = (vum->cpus[thread_index].rx_buffers) + [vum->cpus[thread_index].rx_buffers_len]; b_head = b_current = vlib_get_buffer (vm, bi_current); to_next[0] = bi_current; //We do that now so we can forget about bi_current to_next++; n_left_to_next--; vlib_prefetch_buffer_with_index (vm, - (vum->cpus[cpu_index].rx_buffers) - [vum->cpus[cpu_index]. + (vum-> + cpus[thread_index].rx_buffers) + [vum->cpus[thread_index]. rx_buffers_len - 1], LOAD); /* Just preset the used descriptor id and length for later */ @@ -1791,7 +1793,7 @@ vhost_user_if_input (vlib_main_t * vm, (b_current->current_length == VLIB_BUFFER_DATA_SIZE)) { if (PREDICT_FALSE - (vum->cpus[cpu_index].rx_buffers_len == 0)) + (vum->cpus[thread_index].rx_buffers_len == 0)) { /* Cancel speculation */ to_next--; @@ -1805,17 +1807,18 @@ vhost_user_if_input (vlib_main_t * vm, * but valid. */ vhost_user_input_rewind_buffers (vm, - &vum->cpus[cpu_index], + &vum->cpus + [thread_index], b_head); n_left = 0; goto stop; } /* Get next output */ - vum->cpus[cpu_index].rx_buffers_len--; + vum->cpus[thread_index].rx_buffers_len--; u32 bi_next = - (vum->cpus[cpu_index].rx_buffers)[vum->cpus - [cpu_index].rx_buffers_len]; + (vum->cpus[thread_index].rx_buffers)[vum->cpus + [thread_index].rx_buffers_len]; b_current->next_buffer = bi_next; b_current->flags |= VLIB_BUFFER_NEXT_PRESENT; bi_current = bi_next; @@ -1823,7 +1826,7 @@ vhost_user_if_input (vlib_main_t * vm, } /* Prepare a copy order executed later for the data */ - vhost_copy_t *cpy = &vum->cpus[cpu_index].copy[copy_len]; + vhost_copy_t *cpy = &vum->cpus[thread_index].copy[copy_len]; copy_len++; u32 desc_data_l = desc_table[desc_current].len - desc_data_offset; @@ -1880,7 +1883,7 @@ vhost_user_if_input (vlib_main_t * vm, if (PREDICT_FALSE (copy_len >= VHOST_USER_RX_COPY_THRESHOLD)) { if (PREDICT_FALSE - (vhost_user_input_copy (vui, vum->cpus[cpu_index].copy, + (vhost_user_input_copy (vui, vum->cpus[thread_index].copy, copy_len, &map_hint))) { clib_warning @@ -1905,7 +1908,7 @@ vhost_user_if_input (vlib_main_t * vm, /* Do the memory copies */ if (PREDICT_FALSE - (vhost_user_input_copy (vui, vum->cpus[cpu_index].copy, + (vhost_user_input_copy (vui, vum->cpus[thread_index].copy, copy_len, &map_hint))) { clib_warning ("Memory mapping error on interface hw_if_index=%d " @@ -1933,9 +1936,9 @@ vhost_user_if_input (vlib_main_t * vm, vlib_increment_combined_counter (vnet_main.interface_main.combined_sw_if_counters + VNET_INTERFACE_COUNTER_RX, - os_get_cpu_number (), vui->sw_if_index, n_rx_packets, n_rx_bytes); + vlib_get_thread_index (), vui->sw_if_index, n_rx_packets, n_rx_bytes); - vnet_device_increment_rx_packets (cpu_index, n_rx_packets); + vnet_device_increment_rx_packets (thread_index, n_rx_packets); return n_rx_packets; } @@ -1946,15 +1949,15 @@ vhost_user_input (vlib_main_t * vm, { vhost_user_main_t *vum = &vhost_user_main; uword n_rx_packets = 0; - u32 cpu_index = os_get_cpu_number (); + u32 thread_index = vlib_get_thread_index (); vhost_iface_and_queue_t *vhiq; vhost_user_intf_t *vui; vhost_cpu_t *vhc; - vhc = &vum->cpus[cpu_index]; + vhc = &vum->cpus[thread_index]; if (PREDICT_TRUE (vhc->operation_mode == VHOST_USER_POLLING_MODE)) { - vec_foreach (vhiq, vum->cpus[cpu_index].rx_queues) + vec_foreach (vhiq, vum->cpus[thread_index].rx_queues) { vui = &vum->vhost_user_interfaces[vhiq->vhost_iface_index]; n_rx_packets += vhost_user_if_input (vm, vum, vui, vhiq->qid, node); @@ -2096,7 +2099,7 @@ vhost_user_tx (vlib_main_t * vm, vhost_user_vring_t *rxvq; u16 qsz_mask; u8 error; - u32 cpu_index = os_get_cpu_number (); + u32 thread_index = vlib_get_thread_index (); u32 map_hint = 0; u8 retry = 8; u16 copy_len; @@ -2116,7 +2119,7 @@ vhost_user_tx (vlib_main_t * vm, qid = VHOST_VRING_IDX_RX (*vec_elt_at_index - (vui->per_cpu_tx_qid, os_get_cpu_number ())); + (vui->per_cpu_tx_qid, vlib_get_thread_index ())); rxvq = &vui->vrings[qid]; if (PREDICT_FALSE (vui->use_tx_spinlock)) vhost_user_vring_lock (vui, qid); @@ -2143,10 +2146,10 @@ retry: if (PREDICT_FALSE (b0->flags & VLIB_BUFFER_IS_TRACED)) { - vum->cpus[cpu_index].current_trace = + vum->cpus[thread_index].current_trace = vlib_add_trace (vm, node, b0, - sizeof (*vum->cpus[cpu_index].current_trace)); - vhost_user_tx_trace (vum->cpus[cpu_index].current_trace, + sizeof (*vum->cpus[thread_index].current_trace)); + vhost_user_tx_trace (vum->cpus[thread_index].current_trace, vui, qid / 2, b0, rxvq); } @@ -2188,14 +2191,14 @@ retry: { // Get a header from the header array virtio_net_hdr_mrg_rxbuf_t *hdr = - &vum->cpus[cpu_index].tx_headers[tx_headers_len]; + &vum->cpus[thread_index].tx_headers[tx_headers_len]; tx_headers_len++; hdr->hdr.flags = 0; hdr->hdr.gso_type = 0; hdr->num_buffers = 1; //This is local, no need to check // Prepare a copy order executed later for the header - vhost_copy_t *cpy = &vum->cpus[cpu_index].copy[copy_len]; + vhost_copy_t *cpy = &vum->cpus[thread_index].copy[copy_len]; copy_len++; cpy->len = vui->virtio_net_hdr_sz; cpy->dst = buffer_map_addr; @@ -2220,7 +2223,7 @@ retry: else if (vui->virtio_net_hdr_sz == 12) //MRG is available { virtio_net_hdr_mrg_rxbuf_t *hdr = - &vum->cpus[cpu_index].tx_headers[tx_headers_len - 1]; + &vum->cpus[thread_index].tx_headers[tx_headers_len - 1]; //Move from available to used buffer rxvq->used->ring[rxvq->last_used_idx & qsz_mask].id = @@ -2282,7 +2285,7 @@ retry: } { - vhost_copy_t *cpy = &vum->cpus[cpu_index].copy[copy_len]; + vhost_copy_t *cpy = &vum->cpus[thread_index].copy[copy_len]; copy_len++; cpy->len = bytes_left; cpy->len = (cpy->len > buffer_len) ? buffer_len : cpy->len; @@ -2325,8 +2328,8 @@ retry: if (PREDICT_FALSE (b0->flags & VLIB_BUFFER_IS_TRACED)) { - vum->cpus[cpu_index].current_trace->hdr = - vum->cpus[cpu_index].tx_headers[tx_headers_len - 1]; + vum->cpus[thread_index].current_trace->hdr = + vum->cpus[thread_index].tx_headers[tx_headers_len - 1]; } n_left--; //At the end for error counting when 'goto done' is invoked @@ -2336,7 +2339,7 @@ retry: done: //Do the memory copies if (PREDICT_FALSE - (vhost_user_tx_copy (vui, vum->cpus[cpu_index].copy, + (vhost_user_tx_copy (vui, vum->cpus[thread_index].copy, copy_len, &map_hint))) { clib_warning ("Memory mapping error on interface hw_if_index=%d " @@ -2386,7 +2389,7 @@ done3: vlib_increment_simple_counter (vnet_main.interface_main.sw_if_counters + VNET_INTERFACE_COUNTER_DROP, - os_get_cpu_number (), vui->sw_if_index, n_left); + vlib_get_thread_index (), vui->sw_if_index, n_left); } vlib_buffer_free (vm, vlib_frame_args (frame), frame->n_vectors); @@ -2773,11 +2776,11 @@ vhost_user_send_interrupt_process (vlib_main_t * vm, case ~0: vec_foreach (vhc, vum->cpus) { - u32 cpu_index = vhc - vum->cpus; + u32 thread_index = vhc - vum->cpus; f64 next_timeout; next_timeout = timeout; - vec_foreach (vhiq, vum->cpus[cpu_index].rx_queues) + vec_foreach (vhiq, vum->cpus[thread_index].rx_queues) { vui = &vum->vhost_user_interfaces[vhiq->vhost_iface_index]; vhost_user_vring_t *rxvq = diff --git a/src/vnet/dpo/lookup_dpo.c b/src/vnet/dpo/lookup_dpo.c index e94e871c..97ad0a44 100644 --- a/src/vnet/dpo/lookup_dpo.c +++ b/src/vnet/dpo/lookup_dpo.c @@ -266,7 +266,7 @@ lookup_dpo_ip4_inline (vlib_main_t * vm, int table_from_interface) { u32 n_left_from, next_index, * from, * to_next; - u32 cpu_index = os_get_cpu_number(); + u32 thread_index = vlib_get_thread_index(); vlib_combined_counter_main_t * cm = &load_balance_main.lbm_to_counters; from = vlib_frame_vector_args (from_frame); @@ -407,10 +407,10 @@ lookup_dpo_ip4_inline (vlib_main_t * vm, vnet_buffer(b1)->ip.adj_index[VLIB_TX] = dpo1->dpoi_index; vlib_increment_combined_counter - (cm, cpu_index, lbi0, 1, + (cm, thread_index, lbi0, 1, vlib_buffer_length_in_chain (vm, b0)); vlib_increment_combined_counter - (cm, cpu_index, lbi1, 1, + (cm, thread_index, lbi1, 1, vlib_buffer_length_in_chain (vm, b1)); if (PREDICT_FALSE(b0->flags & VLIB_BUFFER_IS_TRACED)) @@ -511,7 +511,7 @@ lookup_dpo_ip4_inline (vlib_main_t * vm, vnet_buffer(b0)->ip.adj_index[VLIB_TX] = dpo0->dpoi_index; vlib_increment_combined_counter - (cm, cpu_index, lbi0, 1, + (cm, thread_index, lbi0, 1, vlib_buffer_length_in_chain (vm, b0)); if (PREDICT_FALSE(b0->flags & VLIB_BUFFER_IS_TRACED)) @@ -606,7 +606,7 @@ lookup_dpo_ip6_inline (vlib_main_t * vm, { vlib_combined_counter_main_t * cm = &load_balance_main.lbm_to_counters; u32 n_left_from, next_index, * from, * to_next; - u32 cpu_index = os_get_cpu_number(); + u32 thread_index = vlib_get_thread_index(); from = vlib_frame_vector_args (from_frame); n_left_from = from_frame->n_vectors; @@ -749,10 +749,10 @@ lookup_dpo_ip6_inline (vlib_main_t * vm, vnet_buffer(b1)->ip.adj_index[VLIB_TX] = dpo1->dpoi_index; vlib_increment_combined_counter - (cm, cpu_index, lbi0, 1, + (cm, thread_index, lbi0, 1, vlib_buffer_length_in_chain (vm, b0)); vlib_increment_combined_counter - (cm, cpu_index, lbi1, 1, + (cm, thread_index, lbi1, 1, vlib_buffer_length_in_chain (vm, b1)); if (PREDICT_FALSE(b0->flags & VLIB_BUFFER_IS_TRACED)) @@ -853,7 +853,7 @@ lookup_dpo_ip6_inline (vlib_main_t * vm, vnet_buffer(b0)->ip.adj_index[VLIB_TX] = dpo0->dpoi_index; vlib_increment_combined_counter - (cm, cpu_index, lbi0, 1, + (cm, thread_index, lbi0, 1, vlib_buffer_length_in_chain (vm, b0)); if (PREDICT_FALSE(b0->flags & VLIB_BUFFER_IS_TRACED)) @@ -930,7 +930,7 @@ lookup_dpo_mpls_inline (vlib_main_t * vm, int table_from_interface) { u32 n_left_from, next_index, * from, * to_next; - u32 cpu_index = os_get_cpu_number(); + u32 thread_index = vlib_get_thread_index(); vlib_combined_counter_main_t * cm = &load_balance_main.lbm_to_counters; from = vlib_frame_vector_args (from_frame); @@ -994,7 +994,7 @@ lookup_dpo_mpls_inline (vlib_main_t * vm, vnet_buffer(b0)->ip.adj_index[VLIB_TX] = dpo0->dpoi_index; vlib_increment_combined_counter - (cm, cpu_index, lbi0, 1, + (cm, thread_index, lbi0, 1, vlib_buffer_length_in_chain (vm, b0)); if (PREDICT_FALSE(b0->flags & VLIB_BUFFER_IS_TRACED)) diff --git a/src/vnet/dpo/replicate_dpo.c b/src/vnet/dpo/replicate_dpo.c index a9f334be..e25ceae9 100644 --- a/src/vnet/dpo/replicate_dpo.c +++ b/src/vnet/dpo/replicate_dpo.c @@ -627,7 +627,7 @@ replicate_inline (vlib_main_t * vm, vlib_combined_counter_main_t * cm = &replicate_main.repm_counters; replicate_main_t * rm = &replicate_main; u32 n_left_from, * from, * to_next, next_index; - u32 cpu_index = os_get_cpu_number(); + u32 thread_index = vlib_get_thread_index(); from = vlib_frame_vector_args (frame); n_left_from = frame->n_vectors; @@ -657,12 +657,12 @@ replicate_inline (vlib_main_t * vm, rep0 = replicate_get(repi0); vlib_increment_combined_counter( - cm, cpu_index, repi0, 1, + cm, thread_index, repi0, 1, vlib_buffer_length_in_chain(vm, b0)); - vec_validate (rm->clones[cpu_index], rep0->rep_n_buckets - 1); + vec_validate (rm->clones[thread_index], rep0->rep_n_buckets - 1); - num_cloned = vlib_buffer_clone (vm, bi0, rm->clones[cpu_index], rep0->rep_n_buckets, 128); + num_cloned = vlib_buffer_clone (vm, bi0, rm->clones[thread_index], rep0->rep_n_buckets, 128); if (num_cloned != rep0->rep_n_buckets) { @@ -673,7 +673,7 @@ replicate_inline (vlib_main_t * vm, for (bucket = 0; bucket < num_cloned; bucket++) { - ci0 = rm->clones[cpu_index][bucket]; + ci0 = rm->clones[thread_index][bucket]; c0 = vlib_get_buffer(vm, ci0); to_next[0] = ci0; @@ -700,7 +700,7 @@ replicate_inline (vlib_main_t * vm, vlib_get_next_frame (vm, node, next_index, to_next, n_left_to_next); } } - vec_reset_length (rm->clones[cpu_index]); + vec_reset_length (rm->clones[thread_index]); } vlib_put_next_frame (vm, node, next_index, n_left_to_next); diff --git a/src/vnet/ethernet/arp.c b/src/vnet/ethernet/arp.c index ee757505..c74a097e 100644 --- a/src/vnet/ethernet/arp.c +++ b/src/vnet/ethernet/arp.c @@ -1771,7 +1771,7 @@ set_ip4_over_ethernet_rpc_callback (vnet_arp_set_ip4_over_ethernet_rpc_args_t * a) { vnet_main_t *vm = vnet_get_main (); - ASSERT (os_get_cpu_number () == 0); + ASSERT (vlib_get_thread_index () == 0); if (a->flags & ETHERNET_ARP_ARGS_REMOVE) vnet_arp_unset_ip4_over_ethernet_internal (vm, a); diff --git a/src/vnet/ethernet/interface.c b/src/vnet/ethernet/interface.c index 9894e3c8..335e3f9f 100644 --- a/src/vnet/ethernet/interface.c +++ b/src/vnet/ethernet/interface.c @@ -362,7 +362,7 @@ simulated_ethernet_interface_tx (vlib_main_t * vm, u32 next_index = VNET_SIMULATED_ETHERNET_TX_NEXT_ETHERNET_INPUT; u32 i, next_node_index, bvi_flag, sw_if_index; u32 n_pkts = 0, n_bytes = 0; - u32 cpu_index = vm->cpu_index; + u32 thread_index = vm->thread_index; vnet_main_t *vnm = vnet_get_main (); vnet_interface_main_t *im = &vnm->interface_main; vlib_node_main_t *nm = &vm->node_main; @@ -420,8 +420,9 @@ simulated_ethernet_interface_tx (vlib_main_t * vm, /* increment TX interface stat */ vlib_increment_combined_counter (im->combined_sw_if_counters + - VNET_INTERFACE_COUNTER_TX, cpu_index, - sw_if_index, n_pkts, n_bytes); + VNET_INTERFACE_COUNTER_TX, + thread_index, sw_if_index, n_pkts, + n_bytes); } return n_left_from; diff --git a/src/vnet/ethernet/node.c b/src/vnet/ethernet/node.c index b699e381..f7787ed2 100755 --- a/src/vnet/ethernet/node.c +++ b/src/vnet/ethernet/node.c @@ -291,7 +291,7 @@ ethernet_input_inline (vlib_main_t * vm, vlib_node_runtime_t *error_node; u32 n_left_from, next_index, *from, *to_next; u32 stats_sw_if_index, stats_n_packets, stats_n_bytes; - u32 cpu_index = os_get_cpu_number (); + u32 thread_index = vlib_get_thread_index (); u32 cached_sw_if_index = ~0; u32 cached_is_l2 = 0; /* shut up gcc */ vnet_hw_interface_t *hi = NULL; /* used for main interface only */ @@ -510,7 +510,7 @@ ethernet_input_inline (vlib_main_t * vm, interface_main.combined_sw_if_counters + VNET_INTERFACE_COUNTER_RX, - cpu_index, + thread_index, new_sw_if_index0, 1, len0); if (new_sw_if_index1 != old_sw_if_index1 @@ -519,7 +519,7 @@ ethernet_input_inline (vlib_main_t * vm, interface_main.combined_sw_if_counters + VNET_INTERFACE_COUNTER_RX, - cpu_index, + thread_index, new_sw_if_index1, 1, len1); @@ -530,7 +530,7 @@ ethernet_input_inline (vlib_main_t * vm, vlib_increment_combined_counter (vnm->interface_main.combined_sw_if_counters + VNET_INTERFACE_COUNTER_RX, - cpu_index, + thread_index, stats_sw_if_index, stats_n_packets, stats_n_bytes); stats_n_packets = stats_n_bytes = 0; @@ -696,13 +696,13 @@ ethernet_input_inline (vlib_main_t * vm, vlib_increment_combined_counter (vnm->interface_main.combined_sw_if_counters + VNET_INTERFACE_COUNTER_RX, - cpu_index, new_sw_if_index0, 1, len0); + thread_index, new_sw_if_index0, 1, len0); if (stats_n_packets > 0) { vlib_increment_combined_counter (vnm->interface_main.combined_sw_if_counters + VNET_INTERFACE_COUNTER_RX, - cpu_index, + thread_index, stats_sw_if_index, stats_n_packets, stats_n_bytes); stats_n_packets = stats_n_bytes = 0; } @@ -734,7 +734,7 @@ ethernet_input_inline (vlib_main_t * vm, vlib_increment_combined_counter (vnm->interface_main.combined_sw_if_counters + VNET_INTERFACE_COUNTER_RX, - cpu_index, stats_sw_if_index, stats_n_packets, stats_n_bytes); + thread_index, stats_sw_if_index, stats_n_packets, stats_n_bytes); node->runtime_data[0] = stats_sw_if_index; } diff --git a/src/vnet/gre/node.c b/src/vnet/gre/node.c index 2683586e..acf15f24 100644 --- a/src/vnet/gre/node.c +++ b/src/vnet/gre/node.c @@ -75,7 +75,7 @@ gre_input (vlib_main_t * vm, u64 cached_tunnel_key6[4]; u32 cached_tunnel_sw_if_index = 0, tunnel_sw_if_index = 0; - u32 cpu_index = os_get_cpu_number(); + u32 thread_index = vlib_get_thread_index(); u32 len; vnet_interface_main_t *im = &gm->vnet_main->interface_main; @@ -257,7 +257,7 @@ gre_input (vlib_main_t * vm, len = vlib_buffer_length_in_chain (vm, b0); vlib_increment_combined_counter (im->combined_sw_if_counters + VNET_INTERFACE_COUNTER_RX, - cpu_index, + thread_index, tunnel_sw_if_index, 1 /* packets */, len /* bytes */); @@ -324,7 +324,7 @@ drop0: len = vlib_buffer_length_in_chain (vm, b1); vlib_increment_combined_counter (im->combined_sw_if_counters + VNET_INTERFACE_COUNTER_RX, - cpu_index, + thread_index, tunnel_sw_if_index, 1 /* packets */, len /* bytes */); @@ -502,7 +502,7 @@ drop1: len = vlib_buffer_length_in_chain (vm, b0); vlib_increment_combined_counter (im->combined_sw_if_counters + VNET_INTERFACE_COUNTER_RX, - cpu_index, + thread_index, tunnel_sw_if_index, 1 /* packets */, len /* bytes */); diff --git a/src/vnet/interface.h b/src/vnet/interface.h index a1ea2d61..08f08b10 100644 --- a/src/vnet/interface.h +++ b/src/vnet/interface.h @@ -468,7 +468,7 @@ typedef struct vnet_hw_interface_t u32 input_node_index; /* input node cpu index by queue */ - u32 *input_node_cpu_index_by_queue; + u32 *input_node_thread_index_by_queue; } vnet_hw_interface_t; diff --git a/src/vnet/interface_output.c b/src/vnet/interface_output.c index 03f2cdca..663dc309 100644 --- a/src/vnet/interface_output.c +++ b/src/vnet/interface_output.c @@ -196,7 +196,7 @@ slow_path (vlib_main_t * vm, */ static_always_inline void incr_output_stats (vnet_main_t * vnm, - u32 cpu_index, + u32 thread_index, u32 length, u32 sw_if_index, u32 * last_sw_if_index, u32 * n_packets, u32 * n_bytes) @@ -216,7 +216,7 @@ incr_output_stats (vnet_main_t * vnm, vlib_increment_combined_counter (im->combined_sw_if_counters + VNET_INTERFACE_COUNTER_TX, - cpu_index, + thread_index, *last_sw_if_index, *n_packets, *n_bytes); } @@ -240,7 +240,7 @@ vnet_interface_output_node_flatten (vlib_main_t * vm, u32 n_left_to_tx, *from, *from_end, *to_tx; u32 n_bytes, n_buffers, n_packets; u32 last_sw_if_index; - u32 cpu_index = vm->cpu_index; + u32 thread_index = vm->thread_index; n_buffers = frame->n_vectors; @@ -266,7 +266,7 @@ vnet_interface_output_node_flatten (vlib_main_t * vm, cm = vec_elt_at_index (vnm->interface_main.sw_if_counters, VNET_INTERFACE_COUNTER_TX_ERROR); - vlib_increment_simple_counter (cm, cpu_index, + vlib_increment_simple_counter (cm, thread_index, rt->sw_if_index, n_buffers); return vlib_error_drop_buffers (vm, node, from, /* buffer stride */ 1, @@ -341,18 +341,18 @@ vnet_interface_output_node_flatten (vlib_main_t * vm, from += 1; to_tx += n_buffers; n_left_to_tx -= n_buffers; - incr_output_stats (vnm, cpu_index, n_slow_bytes, + incr_output_stats (vnm, thread_index, n_slow_bytes, vnet_buffer (b)->sw_if_index[VLIB_TX], &last_sw_if_index, &n_packets, &n_bytes); } } else { - incr_output_stats (vnm, cpu_index, + incr_output_stats (vnm, thread_index, vlib_buffer_length_in_chain (vm, b0), vnet_buffer (b0)->sw_if_index[VLIB_TX], &last_sw_if_index, &n_packets, &n_bytes); - incr_output_stats (vnm, cpu_index, + incr_output_stats (vnm, thread_index, vlib_buffer_length_in_chain (vm, b0), vnet_buffer (b1)->sw_if_index[VLIB_TX], &last_sw_if_index, &n_packets, &n_bytes); @@ -396,7 +396,7 @@ vnet_interface_output_node_flatten (vlib_main_t * vm, to_tx += n_buffers; n_left_to_tx -= n_buffers; } - incr_output_stats (vnm, cpu_index, + incr_output_stats (vnm, thread_index, vlib_buffer_length_in_chain (vm, b0), vnet_buffer (b0)->sw_if_index[VLIB_TX], &last_sw_if_index, &n_packets, &n_bytes); @@ -408,7 +408,7 @@ vnet_interface_output_node_flatten (vlib_main_t * vm, } /* Final update of interface stats. */ - incr_output_stats (vnm, cpu_index, 0, ~0, /* ~0 will flush stats */ + incr_output_stats (vnm, thread_index, 0, ~0, /* ~0 will flush stats */ &last_sw_if_index, &n_packets, &n_bytes); return n_buffers; @@ -428,7 +428,7 @@ vnet_interface_output_node (vlib_main_t * vm, u32 n_left_to_tx, *from, *from_end, *to_tx; u32 n_bytes, n_buffers, n_packets; u32 n_bytes_b0, n_bytes_b1, n_bytes_b2, n_bytes_b3; - u32 cpu_index = vm->cpu_index; + u32 thread_index = vm->thread_index; vnet_interface_main_t *im = &vnm->interface_main; u32 next_index = VNET_INTERFACE_OUTPUT_NEXT_TX; u32 current_config_index = ~0; @@ -458,7 +458,7 @@ vnet_interface_output_node (vlib_main_t * vm, cm = vec_elt_at_index (vnm->interface_main.sw_if_counters, VNET_INTERFACE_COUNTER_TX_ERROR); - vlib_increment_simple_counter (cm, cpu_index, + vlib_increment_simple_counter (cm, thread_index, rt->sw_if_index, n_buffers); return vlib_error_drop_buffers (vm, node, from, @@ -558,7 +558,7 @@ vnet_interface_output_node (vlib_main_t * vm, { vlib_increment_combined_counter (im->combined_sw_if_counters + VNET_INTERFACE_COUNTER_TX, - cpu_index, tx_swif0, 1, + thread_index, tx_swif0, 1, n_bytes_b0); } @@ -567,7 +567,7 @@ vnet_interface_output_node (vlib_main_t * vm, vlib_increment_combined_counter (im->combined_sw_if_counters + VNET_INTERFACE_COUNTER_TX, - cpu_index, tx_swif1, 1, + thread_index, tx_swif1, 1, n_bytes_b1); } @@ -576,7 +576,7 @@ vnet_interface_output_node (vlib_main_t * vm, vlib_increment_combined_counter (im->combined_sw_if_counters + VNET_INTERFACE_COUNTER_TX, - cpu_index, tx_swif2, 1, + thread_index, tx_swif2, 1, n_bytes_b2); } if (PREDICT_FALSE (tx_swif3 != rt->sw_if_index)) @@ -584,7 +584,7 @@ vnet_interface_output_node (vlib_main_t * vm, vlib_increment_combined_counter (im->combined_sw_if_counters + VNET_INTERFACE_COUNTER_TX, - cpu_index, tx_swif3, 1, + thread_index, tx_swif3, 1, n_bytes_b3); } } @@ -623,7 +623,7 @@ vnet_interface_output_node (vlib_main_t * vm, vlib_increment_combined_counter (im->combined_sw_if_counters + VNET_INTERFACE_COUNTER_TX, - cpu_index, tx_swif0, 1, + thread_index, tx_swif0, 1, n_bytes_b0); } } @@ -634,7 +634,7 @@ vnet_interface_output_node (vlib_main_t * vm, /* Update main interface stats. */ vlib_increment_combined_counter (im->combined_sw_if_counters + VNET_INTERFACE_COUNTER_TX, - cpu_index, + thread_index, rt->sw_if_index, n_packets, n_bytes); return n_buffers; } @@ -893,7 +893,7 @@ process_drop_punt (vlib_main_t * vm, u32 current_sw_if_index, n_errors_current_sw_if_index; u64 current_counter; vlib_simple_counter_main_t *cm; - u32 cpu_index = vm->cpu_index; + u32 thread_index = vm->thread_index; static vlib_error_t memory[VNET_ERROR_N_DISPOSITION]; static char memory_init[VNET_ERROR_N_DISPOSITION]; @@ -965,19 +965,19 @@ process_drop_punt (vlib_main_t * vm, current_counter -= 2; n_errors_current_sw_if_index -= 2; - vlib_increment_simple_counter (cm, cpu_index, sw_if_index0, 1); - vlib_increment_simple_counter (cm, cpu_index, sw_if_index1, 1); + vlib_increment_simple_counter (cm, thread_index, sw_if_index0, 1); + vlib_increment_simple_counter (cm, thread_index, sw_if_index1, 1); /* Increment super-interface drop/punt counters for sub-interfaces. */ sw_if0 = vnet_get_sw_interface (vnm, sw_if_index0); vlib_increment_simple_counter - (cm, cpu_index, sw_if0->sup_sw_if_index, + (cm, thread_index, sw_if0->sup_sw_if_index, sw_if0->sup_sw_if_index != sw_if_index0); sw_if1 = vnet_get_sw_interface (vnm, sw_if_index1); vlib_increment_simple_counter - (cm, cpu_index, sw_if1->sup_sw_if_index, + (cm, thread_index, sw_if1->sup_sw_if_index, sw_if1->sup_sw_if_index != sw_if_index1); em->counters[current_counter_index] = current_counter; @@ -1013,11 +1013,12 @@ process_drop_punt (vlib_main_t * vm, sw_if_index0 = vnet_buffer (b0)->sw_if_index[VLIB_RX]; /* Increment drop/punt counters. */ - vlib_increment_simple_counter (cm, cpu_index, sw_if_index0, 1); + vlib_increment_simple_counter (cm, thread_index, sw_if_index0, 1); /* Increment super-interface drop/punt counters for sub-interfaces. */ sw_if0 = vnet_get_sw_interface (vnm, sw_if_index0); - vlib_increment_simple_counter (cm, cpu_index, sw_if0->sup_sw_if_index, + vlib_increment_simple_counter (cm, thread_index, + sw_if0->sup_sw_if_index, sw_if0->sup_sw_if_index != sw_if_index0); if (PREDICT_FALSE (e0 != current_error)) @@ -1041,12 +1042,12 @@ process_drop_punt (vlib_main_t * vm, { vnet_sw_interface_t *si; - vlib_increment_simple_counter (cm, cpu_index, current_sw_if_index, + vlib_increment_simple_counter (cm, thread_index, current_sw_if_index, n_errors_current_sw_if_index); si = vnet_get_sw_interface (vnm, current_sw_if_index); if (si->sup_sw_if_index != current_sw_if_index) - vlib_increment_simple_counter (cm, cpu_index, si->sup_sw_if_index, + vlib_increment_simple_counter (cm, thread_index, si->sup_sw_if_index, n_errors_current_sw_if_index); } diff --git a/src/vnet/ip/ip4_forward.c b/src/vnet/ip/ip4_forward.c index ee1703e7..fdfe7f63 100644 --- a/src/vnet/ip/ip4_forward.c +++ b/src/vnet/ip/ip4_forward.c @@ -75,7 +75,7 @@ ip4_lookup_inline (vlib_main_t * vm, vlib_combined_counter_main_t *cm = &load_balance_main.lbm_to_counters; u32 n_left_from, n_left_to_next, *from, *to_next; ip_lookup_next_t next; - u32 cpu_index = os_get_cpu_number (); + u32 thread_index = vlib_get_thread_index (); from = vlib_frame_vector_args (frame); n_left_from = frame->n_vectors; @@ -292,19 +292,19 @@ ip4_lookup_inline (vlib_main_t * vm, vnet_buffer (p3)->ip.adj_index[VLIB_TX] = dpo3->dpoi_index; vlib_increment_combined_counter - (cm, cpu_index, lb_index0, 1, + (cm, thread_index, lb_index0, 1, vlib_buffer_length_in_chain (vm, p0) + sizeof (ethernet_header_t)); vlib_increment_combined_counter - (cm, cpu_index, lb_index1, 1, + (cm, thread_index, lb_index1, 1, vlib_buffer_length_in_chain (vm, p1) + sizeof (ethernet_header_t)); vlib_increment_combined_counter - (cm, cpu_index, lb_index2, 1, + (cm, thread_index, lb_index2, 1, vlib_buffer_length_in_chain (vm, p2) + sizeof (ethernet_header_t)); vlib_increment_combined_counter - (cm, cpu_index, lb_index3, 1, + (cm, thread_index, lb_index3, 1, vlib_buffer_length_in_chain (vm, p3) + sizeof (ethernet_header_t)); @@ -392,7 +392,7 @@ ip4_lookup_inline (vlib_main_t * vm, vnet_buffer (p0)->ip.adj_index[VLIB_TX] = dpo0->dpoi_index; vlib_increment_combined_counter - (cm, cpu_index, lbi0, 1, vlib_buffer_length_in_chain (vm, p0)); + (cm, thread_index, lbi0, 1, vlib_buffer_length_in_chain (vm, p0)); from += 1; to_next += 1; @@ -479,7 +479,7 @@ ip4_load_balance (vlib_main_t * vm, vlib_combined_counter_main_t *cm = &load_balance_main.lbm_via_counters; u32 n_left_from, n_left_to_next, *from, *to_next; ip_lookup_next_t next; - u32 cpu_index = os_get_cpu_number (); + u32 thread_index = vlib_get_thread_index (); from = vlib_frame_vector_args (frame); n_left_from = frame->n_vectors; @@ -584,9 +584,9 @@ ip4_load_balance (vlib_main_t * vm, vnet_buffer (p1)->ip.adj_index[VLIB_TX] = dpo1->dpoi_index; vlib_increment_combined_counter - (cm, cpu_index, lbi0, 1, vlib_buffer_length_in_chain (vm, p0)); + (cm, thread_index, lbi0, 1, vlib_buffer_length_in_chain (vm, p0)); vlib_increment_combined_counter - (cm, cpu_index, lbi1, 1, vlib_buffer_length_in_chain (vm, p1)); + (cm, thread_index, lbi1, 1, vlib_buffer_length_in_chain (vm, p1)); vlib_validate_buffer_enqueue_x2 (vm, node, next, to_next, n_left_to_next, @@ -639,7 +639,7 @@ ip4_load_balance (vlib_main_t * vm, vnet_buffer (p0)->ip.adj_index[VLIB_TX] = dpo0->dpoi_index; vlib_increment_combined_counter - (cm, cpu_index, lbi0, 1, vlib_buffer_length_in_chain (vm, p0)); + (cm, thread_index, lbi0, 1, vlib_buffer_length_in_chain (vm, p0)); vlib_validate_buffer_enqueue_x1 (vm, node, next, to_next, n_left_to_next, @@ -2330,7 +2330,7 @@ ip4_rewrite_inline (vlib_main_t * vm, n_left_from = frame->n_vectors; next_index = node->cached_next_index; - u32 cpu_index = os_get_cpu_number (); + u32 thread_index = vlib_get_thread_index (); while (n_left_from > 0) { @@ -2379,9 +2379,9 @@ ip4_rewrite_inline (vlib_main_t * vm, if (do_counters) { vlib_prefetch_combined_counter (&adjacency_counters, - cpu_index, adj_index0); + thread_index, adj_index0); vlib_prefetch_combined_counter (&adjacency_counters, - cpu_index, adj_index1); + thread_index, adj_index1); } ip0 = vlib_buffer_get_current (p0); @@ -2527,13 +2527,13 @@ ip4_rewrite_inline (vlib_main_t * vm, { vlib_increment_combined_counter (&adjacency_counters, - cpu_index, + thread_index, adj_index0, 1, vlib_buffer_length_in_chain (vm, p0) + rw_len0); vlib_increment_combined_counter (&adjacency_counters, - cpu_index, + thread_index, adj_index1, 1, vlib_buffer_length_in_chain (vm, p1) + rw_len1); } @@ -2618,7 +2618,7 @@ ip4_rewrite_inline (vlib_main_t * vm, if (do_counters) vlib_prefetch_combined_counter (&adjacency_counters, - cpu_index, adj_index0); + thread_index, adj_index0); /* Guess we are only writing on simple Ethernet header. */ vnet_rewrite_one_header (adj0[0], ip0, sizeof (ethernet_header_t)); @@ -2637,7 +2637,7 @@ ip4_rewrite_inline (vlib_main_t * vm, if (do_counters) vlib_increment_combined_counter (&adjacency_counters, - cpu_index, adj_index0, 1, + thread_index, adj_index0, 1, vlib_buffer_length_in_chain (vm, p0) + rw_len0); /* Check MTU of outgoing interface. */ diff --git a/src/vnet/ip/ip4_input.c b/src/vnet/ip/ip4_input.c index ba200a9f..3b08f4b0 100644 --- a/src/vnet/ip/ip4_input.c +++ b/src/vnet/ip/ip4_input.c @@ -85,7 +85,7 @@ ip4_input_inline (vlib_main_t * vm, vlib_node_runtime_t *error_node = vlib_node_get_runtime (vm, ip4_input_node.index); vlib_simple_counter_main_t *cm; - u32 cpu_index = os_get_cpu_number (); + u32 thread_index = vlib_get_thread_index (); from = vlib_frame_vector_args (frame); n_left_from = frame->n_vectors; @@ -178,8 +178,8 @@ ip4_input_inline (vlib_main_t * vm, vnet_feature_arc_start (arc0, sw_if_index0, &next0, p0); vnet_feature_arc_start (arc1, sw_if_index1, &next1, p1); - vlib_increment_simple_counter (cm, cpu_index, sw_if_index0, 1); - vlib_increment_simple_counter (cm, cpu_index, sw_if_index1, 1); + vlib_increment_simple_counter (cm, thread_index, sw_if_index0, 1); + vlib_increment_simple_counter (cm, thread_index, sw_if_index1, 1); /* Punt packets with options or wrong version. */ if (PREDICT_FALSE (ip0->ip_version_and_header_length != 0x45)) @@ -299,7 +299,7 @@ ip4_input_inline (vlib_main_t * vm, vnet_buffer (p0)->ip.adj_index[VLIB_RX] = ~0; vnet_feature_arc_start (arc0, sw_if_index0, &next0, p0); - vlib_increment_simple_counter (cm, cpu_index, sw_if_index0, 1); + vlib_increment_simple_counter (cm, thread_index, sw_if_index0, 1); /* Punt packets with options or wrong version. */ if (PREDICT_FALSE (ip0->ip_version_and_header_length != 0x45)) diff --git a/src/vnet/ip/ip6_forward.c b/src/vnet/ip/ip6_forward.c index c120f12c..c2fc4f87 100644 --- a/src/vnet/ip/ip6_forward.c +++ b/src/vnet/ip/ip6_forward.c @@ -74,7 +74,7 @@ ip6_lookup_inline (vlib_main_t * vm, vlib_combined_counter_main_t *cm = &load_balance_main.lbm_to_counters; u32 n_left_from, n_left_to_next, *from, *to_next; ip_lookup_next_t next; - u32 cpu_index = os_get_cpu_number (); + u32 thread_index = vlib_get_thread_index (); from = vlib_frame_vector_args (frame); n_left_from = frame->n_vectors; @@ -185,9 +185,9 @@ ip6_lookup_inline (vlib_main_t * vm, vnet_buffer (p1)->ip.adj_index[VLIB_TX] = dpo1->dpoi_index; vlib_increment_combined_counter - (cm, cpu_index, lbi0, 1, vlib_buffer_length_in_chain (vm, p0)); + (cm, thread_index, lbi0, 1, vlib_buffer_length_in_chain (vm, p0)); vlib_increment_combined_counter - (cm, cpu_index, lbi1, 1, vlib_buffer_length_in_chain (vm, p1)); + (cm, thread_index, lbi1, 1, vlib_buffer_length_in_chain (vm, p1)); from += 2; to_next += 2; @@ -291,7 +291,7 @@ ip6_lookup_inline (vlib_main_t * vm, vnet_buffer (p0)->ip.adj_index[VLIB_TX] = dpo0->dpoi_index; vlib_increment_combined_counter - (cm, cpu_index, lbi0, 1, vlib_buffer_length_in_chain (vm, p0)); + (cm, thread_index, lbi0, 1, vlib_buffer_length_in_chain (vm, p0)); from += 1; to_next += 1; @@ -703,7 +703,7 @@ ip6_load_balance (vlib_main_t * vm, vlib_combined_counter_main_t *cm = &load_balance_main.lbm_via_counters; u32 n_left_from, n_left_to_next, *from, *to_next; ip_lookup_next_t next; - u32 cpu_index = os_get_cpu_number (); + u32 thread_index = vlib_get_thread_index (); ip6_main_t *im = &ip6_main; from = vlib_frame_vector_args (frame); @@ -824,9 +824,9 @@ ip6_load_balance (vlib_main_t * vm, vnet_buffer (p1)->ip.adj_index[VLIB_TX] = dpo1->dpoi_index; vlib_increment_combined_counter - (cm, cpu_index, lbi0, 1, vlib_buffer_length_in_chain (vm, p0)); + (cm, thread_index, lbi0, 1, vlib_buffer_length_in_chain (vm, p0)); vlib_increment_combined_counter - (cm, cpu_index, lbi1, 1, vlib_buffer_length_in_chain (vm, p1)); + (cm, thread_index, lbi1, 1, vlib_buffer_length_in_chain (vm, p1)); vlib_validate_buffer_enqueue_x2 (vm, node, next, to_next, n_left_to_next, @@ -886,7 +886,7 @@ ip6_load_balance (vlib_main_t * vm, } vlib_increment_combined_counter - (cm, cpu_index, lbi0, 1, vlib_buffer_length_in_chain (vm, p0)); + (cm, thread_index, lbi0, 1, vlib_buffer_length_in_chain (vm, p0)); vlib_validate_buffer_enqueue_x1 (vm, node, next, to_next, n_left_to_next, @@ -1897,7 +1897,7 @@ ip6_rewrite_inline (vlib_main_t * vm, n_left_from = frame->n_vectors; next_index = node->cached_next_index; - u32 cpu_index = os_get_cpu_number (); + u32 thread_index = vlib_get_thread_index (); while (n_left_from > 0) { @@ -2019,11 +2019,11 @@ ip6_rewrite_inline (vlib_main_t * vm, { vlib_increment_combined_counter (&adjacency_counters, - cpu_index, adj_index0, 1, + thread_index, adj_index0, 1, vlib_buffer_length_in_chain (vm, p0) + rw_len0); vlib_increment_combined_counter (&adjacency_counters, - cpu_index, adj_index1, 1, + thread_index, adj_index1, 1, vlib_buffer_length_in_chain (vm, p1) + rw_len1); } @@ -2156,7 +2156,7 @@ ip6_rewrite_inline (vlib_main_t * vm, { vlib_increment_combined_counter (&adjacency_counters, - cpu_index, adj_index0, 1, + thread_index, adj_index0, 1, vlib_buffer_length_in_chain (vm, p0) + rw_len0); } diff --git a/src/vnet/ip/ip6_input.c b/src/vnet/ip/ip6_input.c index 20306088..ffdc4727 100644 --- a/src/vnet/ip/ip6_input.c +++ b/src/vnet/ip/ip6_input.c @@ -82,7 +82,7 @@ ip6_input (vlib_main_t * vm, vlib_node_runtime_t * node, vlib_frame_t * frame) vlib_node_runtime_t *error_node = vlib_node_get_runtime (vm, ip6_input_node.index); vlib_simple_counter_main_t *cm; - u32 cpu_index = os_get_cpu_number (); + u32 thread_index = vlib_get_thread_index (); from = vlib_frame_vector_args (frame); n_left_from = frame->n_vectors; @@ -171,8 +171,8 @@ ip6_input (vlib_main_t * vm, vlib_node_runtime_t * node, vlib_frame_t * frame) vnet_feature_arc_start (arc0, sw_if_index0, &next0, p0); vnet_feature_arc_start (arc1, sw_if_index1, &next1, p1); - vlib_increment_simple_counter (cm, cpu_index, sw_if_index0, 1); - vlib_increment_simple_counter (cm, cpu_index, sw_if_index1, 1); + vlib_increment_simple_counter (cm, thread_index, sw_if_index0, 1); + vlib_increment_simple_counter (cm, thread_index, sw_if_index1, 1); error0 = error1 = IP6_ERROR_NONE; @@ -270,7 +270,7 @@ ip6_input (vlib_main_t * vm, vlib_node_runtime_t * node, vlib_frame_t * frame) vnet_buffer (p0)->ip.adj_index[VLIB_RX] = ~0; vnet_feature_arc_start (arc0, sw_if_index0, &next0, p0); - vlib_increment_simple_counter (cm, cpu_index, sw_if_index0, 1); + vlib_increment_simple_counter (cm, thread_index, sw_if_index0, 1); error0 = IP6_ERROR_NONE; /* Version != 6? Drop it. */ diff --git a/src/vnet/ip/ip6_neighbor.c b/src/vnet/ip/ip6_neighbor.c index 5d1fb6f8..2af546df 100644 --- a/src/vnet/ip/ip6_neighbor.c +++ b/src/vnet/ip/ip6_neighbor.c @@ -581,7 +581,7 @@ vnet_set_ip6_ethernet_neighbor (vlib_main_t * vm, u32 next_index; pending_resolution_t *pr, *mc; - if (os_get_cpu_number ()) + if (vlib_get_thread_index ()) { set_unset_ip6_neighbor_rpc (vm, sw_if_index, a, link_layer_address, 1 /* set new neighbor */ , is_static, @@ -722,7 +722,7 @@ vnet_unset_ip6_ethernet_neighbor (vlib_main_t * vm, uword *p; int rv = 0; - if (os_get_cpu_number ()) + if (vlib_get_thread_index ()) { set_unset_ip6_neighbor_rpc (vm, sw_if_index, a, link_layer_address, 0 /* unset */ , 0, 0); diff --git a/src/vnet/ipsec/esp.h b/src/vnet/ipsec/esp.h index 50cac806..799003b9 100644 --- a/src/vnet/ipsec/esp.h +++ b/src/vnet/ipsec/esp.h @@ -282,8 +282,8 @@ hmac_calc (ipsec_integ_alg_t alg, u8 * data, int data_len, u8 * signature, u8 use_esn, u32 seq_hi) { esp_main_t *em = &esp_main; - u32 cpu_index = os_get_cpu_number (); - HMAC_CTX *ctx = &(em->per_thread_data[cpu_index].hmac_ctx); + u32 thread_index = vlib_get_thread_index (); + HMAC_CTX *ctx = &(em->per_thread_data[thread_index].hmac_ctx); const EVP_MD *md = NULL; unsigned int len; @@ -292,10 +292,10 @@ hmac_calc (ipsec_integ_alg_t alg, if (PREDICT_FALSE (em->esp_integ_algs[alg].md == 0)) return 0; - if (PREDICT_FALSE (alg != em->per_thread_data[cpu_index].last_integ_alg)) + if (PREDICT_FALSE (alg != em->per_thread_data[thread_index].last_integ_alg)) { md = em->esp_integ_algs[alg].md; - em->per_thread_data[cpu_index].last_integ_alg = alg; + em->per_thread_data[thread_index].last_integ_alg = alg; } HMAC_Init (ctx, key, key_len, md); diff --git a/src/vnet/ipsec/esp_decrypt.c b/src/vnet/ipsec/esp_decrypt.c index 7289b260..925d2b45 100644 --- a/src/vnet/ipsec/esp_decrypt.c +++ b/src/vnet/ipsec/esp_decrypt.c @@ -85,8 +85,8 @@ esp_decrypt_aes_cbc (ipsec_crypto_alg_t alg, u8 * in, u8 * out, size_t in_len, u8 * key, u8 * iv) { esp_main_t *em = &esp_main; - u32 cpu_index = os_get_cpu_number (); - EVP_CIPHER_CTX *ctx = &(em->per_thread_data[cpu_index].decrypt_ctx); + u32 thread_index = vlib_get_thread_index (); + EVP_CIPHER_CTX *ctx = &(em->per_thread_data[thread_index].decrypt_ctx); const EVP_CIPHER *cipher = NULL; int out_len; @@ -95,10 +95,11 @@ esp_decrypt_aes_cbc (ipsec_crypto_alg_t alg, if (PREDICT_FALSE (em->esp_crypto_algs[alg].type == 0)) return; - if (PREDICT_FALSE (alg != em->per_thread_data[cpu_index].last_decrypt_alg)) + if (PREDICT_FALSE + (alg != em->per_thread_data[thread_index].last_decrypt_alg)) { cipher = em->esp_crypto_algs[alg].type; - em->per_thread_data[cpu_index].last_decrypt_alg = alg; + em->per_thread_data[thread_index].last_decrypt_alg = alg; } EVP_DecryptInit_ex (ctx, cipher, NULL, key, iv); @@ -117,11 +118,11 @@ esp_decrypt_node_fn (vlib_main_t * vm, u32 *recycle = 0; from = vlib_frame_vector_args (from_frame); n_left_from = from_frame->n_vectors; - u32 cpu_index = os_get_cpu_number (); + u32 thread_index = vlib_get_thread_index (); ipsec_alloc_empty_buffers (vm, im); - u32 *empty_buffers = im->empty_buffers[cpu_index]; + u32 *empty_buffers = im->empty_buffers[thread_index]; if (PREDICT_FALSE (vec_len (empty_buffers) < n_left_from)) { diff --git a/src/vnet/ipsec/esp_encrypt.c b/src/vnet/ipsec/esp_encrypt.c index 44ae2297..b2bc4e0b 100644 --- a/src/vnet/ipsec/esp_encrypt.c +++ b/src/vnet/ipsec/esp_encrypt.c @@ -88,8 +88,8 @@ esp_encrypt_aes_cbc (ipsec_crypto_alg_t alg, u8 * in, u8 * out, size_t in_len, u8 * key, u8 * iv) { esp_main_t *em = &esp_main; - u32 cpu_index = os_get_cpu_number (); - EVP_CIPHER_CTX *ctx = &(em->per_thread_data[cpu_index].encrypt_ctx); + u32 thread_index = vlib_get_thread_index (); + EVP_CIPHER_CTX *ctx = &(em->per_thread_data[thread_index].encrypt_ctx); const EVP_CIPHER *cipher = NULL; int out_len; @@ -98,10 +98,11 @@ esp_encrypt_aes_cbc (ipsec_crypto_alg_t alg, if (PREDICT_FALSE (em->esp_crypto_algs[alg].type == IPSEC_CRYPTO_ALG_NONE)) return; - if (PREDICT_FALSE (alg != em->per_thread_data[cpu_index].last_encrypt_alg)) + if (PREDICT_FALSE + (alg != em->per_thread_data[thread_index].last_encrypt_alg)) { cipher = em->esp_crypto_algs[alg].type; - em->per_thread_data[cpu_index].last_encrypt_alg = alg; + em->per_thread_data[thread_index].last_encrypt_alg = alg; } EVP_EncryptInit_ex (ctx, cipher, NULL, key, iv); @@ -119,11 +120,11 @@ esp_encrypt_node_fn (vlib_main_t * vm, n_left_from = from_frame->n_vectors; ipsec_main_t *im = &ipsec_main; u32 *recycle = 0; - u32 cpu_index = os_get_cpu_number (); + u32 thread_index = vlib_get_thread_index (); ipsec_alloc_empty_buffers (vm, im); - u32 *empty_buffers = im->empty_buffers[cpu_index]; + u32 *empty_buffers = im->empty_buffers[thread_index]; if (PREDICT_FALSE (vec_len (empty_buffers) < n_left_from)) { diff --git a/src/vnet/ipsec/ikev2.c b/src/vnet/ipsec/ikev2.c index 2c1074d8..3f9978a7 100644 --- a/src/vnet/ipsec/ikev2.c +++ b/src/vnet/ipsec/ikev2.c @@ -303,16 +303,16 @@ static void ikev2_delete_sa (ikev2_sa_t * sa) { ikev2_main_t *km = &ikev2_main; - u32 cpu_index = os_get_cpu_number (); + u32 thread_index = vlib_get_thread_index (); uword *p; ikev2_sa_free_all_vec (sa); - p = hash_get (km->per_thread_data[cpu_index].sa_by_rspi, sa->rspi); + p = hash_get (km->per_thread_data[thread_index].sa_by_rspi, sa->rspi); if (p) { - hash_unset (km->per_thread_data[cpu_index].sa_by_rspi, sa->rspi); - pool_put (km->per_thread_data[cpu_index].sas, sa); + hash_unset (km->per_thread_data[thread_index].sa_by_rspi, sa->rspi); + pool_put (km->per_thread_data[thread_index].sas, sa); } } @@ -776,29 +776,31 @@ ikev2_initial_contact_cleanup (ikev2_sa_t * sa) ikev2_sa_t *tmp; u32 i, *delete = 0; ikev2_child_sa_t *c; - u32 cpu_index = os_get_cpu_number (); + u32 thread_index = vlib_get_thread_index (); if (!sa->initial_contact) return; /* find old IKE SAs with the same authenticated identity */ /* *INDENT-OFF* */ - pool_foreach (tmp, km->per_thread_data[cpu_index].sas, ({ + pool_foreach (tmp, km->per_thread_data[thread_index].sas, ({ if (tmp->i_id.type != sa->i_id.type || vec_len(tmp->i_id.data) != vec_len(sa->i_id.data) || memcmp(sa->i_id.data, tmp->i_id.data, vec_len(sa->i_id.data))) continue; if (sa->rspi != tmp->rspi) - vec_add1(delete, tmp - km->per_thread_data[cpu_index].sas); + vec_add1(delete, tmp - km->per_thread_data[thread_index].sas); })); /* *INDENT-ON* */ for (i = 0; i < vec_len (delete); i++) { - tmp = pool_elt_at_index (km->per_thread_data[cpu_index].sas, delete[i]); - vec_foreach (c, tmp->childs) - ikev2_delete_tunnel_interface (km->vnet_main, tmp, c); + tmp = + pool_elt_at_index (km->per_thread_data[thread_index].sas, delete[i]); + vec_foreach (c, + tmp->childs) ikev2_delete_tunnel_interface (km->vnet_main, + tmp, c); ikev2_delete_sa (tmp); } @@ -1922,10 +1924,10 @@ ikev2_retransmit_sa_init (ike_header_t * ike, { ikev2_main_t *km = &ikev2_main; ikev2_sa_t *sa; - u32 cpu_index = os_get_cpu_number (); + u32 thread_index = vlib_get_thread_index (); /* *INDENT-OFF* */ - pool_foreach (sa, km->per_thread_data[cpu_index].sas, ({ + pool_foreach (sa, km->per_thread_data[thread_index].sas, ({ if (sa->ispi == clib_net_to_host_u64(ike->ispi) && sa->iaddr.as_u32 == iaddr.as_u32 && sa->raddr.as_u32 == raddr.as_u32) @@ -2036,7 +2038,7 @@ ikev2_node_fn (vlib_main_t * vm, u32 n_left_from, *from, *to_next; ikev2_next_t next_index; ikev2_main_t *km = &ikev2_main; - u32 cpu_index = os_get_cpu_number (); + u32 thread_index = vlib_get_thread_index (); from = vlib_frame_vector_args (frame); n_left_from = frame->n_vectors; @@ -2134,11 +2136,14 @@ ikev2_node_fn (vlib_main_t * vm, if (sa0->state == IKEV2_STATE_SA_INIT) { /* add SA to the pool */ - pool_get (km->per_thread_data[cpu_index].sas, sa0); + pool_get (km->per_thread_data[thread_index].sas, + sa0); clib_memcpy (sa0, &sa, sizeof (*sa0)); - hash_set (km->per_thread_data[cpu_index].sa_by_rspi, + hash_set (km-> + per_thread_data[thread_index].sa_by_rspi, sa0->rspi, - sa0 - km->per_thread_data[cpu_index].sas); + sa0 - + km->per_thread_data[thread_index].sas); } else { @@ -2169,11 +2174,11 @@ ikev2_node_fn (vlib_main_t * vm, if (sa0->state == IKEV2_STATE_SA_INIT) { /* add SA to the pool */ - pool_get (km->per_thread_data[cpu_index].sas, sa0); + pool_get (km->per_thread_data[thread_index].sas, sa0); clib_memcpy (sa0, &sa, sizeof (*sa0)); - hash_set (km->per_thread_data[cpu_index].sa_by_rspi, + hash_set (km->per_thread_data[thread_index].sa_by_rspi, sa0->rspi, - sa0 - km->per_thread_data[cpu_index].sas); + sa0 - km->per_thread_data[thread_index].sas); } else { @@ -2184,12 +2189,13 @@ ikev2_node_fn (vlib_main_t * vm, else if (ike0->exchange == IKEV2_EXCHANGE_IKE_AUTH) { uword *p; - p = hash_get (km->per_thread_data[cpu_index].sa_by_rspi, + p = hash_get (km->per_thread_data[thread_index].sa_by_rspi, clib_net_to_host_u64 (ike0->rspi)); if (p) { - sa0 = pool_elt_at_index (km->per_thread_data[cpu_index].sas, - p[0]); + sa0 = + pool_elt_at_index (km->per_thread_data[thread_index].sas, + p[0]); r = ikev2_retransmit_resp (sa0, ike0); if (r == 1) @@ -2240,12 +2246,13 @@ ikev2_node_fn (vlib_main_t * vm, else if (ike0->exchange == IKEV2_EXCHANGE_INFORMATIONAL) { uword *p; - p = hash_get (km->per_thread_data[cpu_index].sa_by_rspi, + p = hash_get (km->per_thread_data[thread_index].sa_by_rspi, clib_net_to_host_u64 (ike0->rspi)); if (p) { - sa0 = pool_elt_at_index (km->per_thread_data[cpu_index].sas, - p[0]); + sa0 = + pool_elt_at_index (km->per_thread_data[thread_index].sas, + p[0]); r = ikev2_retransmit_resp (sa0, ike0); if (r == 1) @@ -2305,12 +2312,13 @@ ikev2_node_fn (vlib_main_t * vm, else if (ike0->exchange == IKEV2_EXCHANGE_CREATE_CHILD_SA) { uword *p; - p = hash_get (km->per_thread_data[cpu_index].sa_by_rspi, + p = hash_get (km->per_thread_data[thread_index].sa_by_rspi, clib_net_to_host_u64 (ike0->rspi)); if (p) { - sa0 = pool_elt_at_index (km->per_thread_data[cpu_index].sas, - p[0]); + sa0 = + pool_elt_at_index (km->per_thread_data[thread_index].sas, + p[0]); r = ikev2_retransmit_resp (sa0, ike0); if (r == 1) diff --git a/src/vnet/ipsec/ipsec.h b/src/vnet/ipsec/ipsec.h index 58f0f145..c884e360 100644 --- a/src/vnet/ipsec/ipsec.h +++ b/src/vnet/ipsec/ipsec.h @@ -324,21 +324,21 @@ int ipsec_set_interface_key (vnet_main_t * vnm, u32 hw_if_index, always_inline void ipsec_alloc_empty_buffers (vlib_main_t * vm, ipsec_main_t * im) { - u32 cpu_index = os_get_cpu_number (); - uword l = vec_len (im->empty_buffers[cpu_index]); + u32 thread_index = vlib_get_thread_index (); + uword l = vec_len (im->empty_buffers[thread_index]); uword n_alloc = 0; if (PREDICT_FALSE (l < VLIB_FRAME_SIZE)) { - if (!im->empty_buffers[cpu_index]) + if (!im->empty_buffers[thread_index]) { - vec_alloc (im->empty_buffers[cpu_index], 2 * VLIB_FRAME_SIZE); + vec_alloc (im->empty_buffers[thread_index], 2 * VLIB_FRAME_SIZE); } - n_alloc = vlib_buffer_alloc (vm, im->empty_buffers[cpu_index] + l, + n_alloc = vlib_buffer_alloc (vm, im->empty_buffers[thread_index] + l, 2 * VLIB_FRAME_SIZE - l); - _vec_len (im->empty_buffers[cpu_index]) = l + n_alloc; + _vec_len (im->empty_buffers[thread_index]) = l + n_alloc; } } diff --git a/src/vnet/ipsec/ipsec_if.c b/src/vnet/ipsec/ipsec_if.c index dc882004..ed124894 100644 --- a/src/vnet/ipsec/ipsec_if.c +++ b/src/vnet/ipsec/ipsec_if.c @@ -99,7 +99,7 @@ static int ipsec_add_del_tunnel_if_rpc_callback (ipsec_add_del_tunnel_args_t * a) { vnet_main_t *vnm = vnet_get_main (); - ASSERT (os_get_cpu_number () == 0); + ASSERT (vlib_get_thread_index () == 0); return ipsec_add_del_tunnel_if_internal (vnm, a); } diff --git a/src/vnet/l2/l2_bvi.h b/src/vnet/l2/l2_bvi.h index dd1130a6..e21a1616 100644 --- a/src/vnet/l2/l2_bvi.h +++ b/src/vnet/l2/l2_bvi.h @@ -97,7 +97,7 @@ l2_to_bvi (vlib_main_t * vlib_main, vlib_increment_combined_counter (vnet_main->interface_main.combined_sw_if_counters + VNET_INTERFACE_COUNTER_RX, - vlib_main->cpu_index, + vlib_main->thread_index, vnet_buffer (b0)->sw_if_index[VLIB_RX], 1, vlib_buffer_length_in_chain (vlib_main, b0)); return TO_BVI_ERR_OK; diff --git a/src/vnet/l2/l2_input.c b/src/vnet/l2/l2_input.c index 041ff38d..e5d6878a 100644 --- a/src/vnet/l2/l2_input.c +++ b/src/vnet/l2/l2_input.c @@ -117,7 +117,7 @@ typedef enum static_always_inline void classify_and_dispatch (vlib_main_t * vm, vlib_node_runtime_t * node, - u32 cpu_index, + u32 thread_index, l2input_main_t * msm, vlib_buffer_t * b0, u32 * next0) { /* @@ -237,7 +237,7 @@ l2input_node_inline (vlib_main_t * vm, u32 n_left_from, *from, *to_next; l2input_next_t next_index; l2input_main_t *msm = &l2input_main; - u32 cpu_index = os_get_cpu_number (); + u32 thread_index = vlib_get_thread_index (); from = vlib_frame_vector_args (frame); n_left_from = frame->n_vectors; /* number of packets to process */ @@ -350,10 +350,10 @@ l2input_node_inline (vlib_main_t * vm, vlib_node_increment_counter (vm, l2input_node.index, L2INPUT_ERROR_L2INPUT, 4); - classify_and_dispatch (vm, node, cpu_index, msm, b0, &next0); - classify_and_dispatch (vm, node, cpu_index, msm, b1, &next1); - classify_and_dispatch (vm, node, cpu_index, msm, b2, &next2); - classify_and_dispatch (vm, node, cpu_index, msm, b3, &next3); + classify_and_dispatch (vm, node, thread_index, msm, b0, &next0); + classify_and_dispatch (vm, node, thread_index, msm, b1, &next1); + classify_and_dispatch (vm, node, thread_index, msm, b2, &next2); + classify_and_dispatch (vm, node, thread_index, msm, b3, &next3); /* verify speculative enqueues, maybe switch current next frame */ /* if next0==next1==next_index then nothing special needs to be done */ @@ -393,7 +393,7 @@ l2input_node_inline (vlib_main_t * vm, vlib_node_increment_counter (vm, l2input_node.index, L2INPUT_ERROR_L2INPUT, 1); - classify_and_dispatch (vm, node, cpu_index, msm, b0, &next0); + classify_and_dispatch (vm, node, thread_index, msm, b0, &next0); /* verify speculative enqueue, maybe switch current next frame */ vlib_validate_buffer_enqueue_x1 (vm, node, next_index, diff --git a/src/vnet/l2/l2_output.c b/src/vnet/l2/l2_output.c index 00f22571..e17b2a16 100644 --- a/src/vnet/l2/l2_output.c +++ b/src/vnet/l2/l2_output.c @@ -643,11 +643,11 @@ l2output_create_output_node_mapping (vlib_main_t * vlib_main, vnet_main_t * vnet hw0 = vnet_get_sup_hw_interface (vnet_main, sw_if_index); - uword cpu_number; + uword thread_index; - cpu_number = os_get_cpu_number (); + thread_index = vlib_get_thread_index (); - if (cpu_number) + if (thread_index) { u32 oldflags; diff --git a/src/vnet/l2tp/decap.c b/src/vnet/l2tp/decap.c index e8986935..46104129 100644 --- a/src/vnet/l2tp/decap.c +++ b/src/vnet/l2tp/decap.c @@ -149,7 +149,7 @@ last_stage (vlib_main_t * vm, vlib_node_runtime_t * node, u32 bi) /* per-mapping byte stats include the ethernet header */ vlib_increment_combined_counter (&lm->counter_main, - os_get_cpu_number (), + vlib_get_thread_index (), counter_index, 1 /* packet_increment */ , vlib_buffer_length_in_chain (vm, b) + sizeof (ethernet_header_t)); diff --git a/src/vnet/l2tp/encap.c b/src/vnet/l2tp/encap.c index ed7a9580..dcdfde4b 100644 --- a/src/vnet/l2tp/encap.c +++ b/src/vnet/l2tp/encap.c @@ -124,7 +124,7 @@ last_stage (vlib_main_t * vm, vlib_node_runtime_t * node, u32 bi) /* per-mapping byte stats include the ethernet header */ vlib_increment_combined_counter (&lm->counter_main, - os_get_cpu_number (), + vlib_get_thread_index (), counter_index, 1 /* packet_increment */ , vlib_buffer_length_in_chain (vm, b)); diff --git a/src/vnet/l2tp/l2tp.c b/src/vnet/l2tp/l2tp.c index cb94d7e7..3dedc447 100644 --- a/src/vnet/l2tp/l2tp.c +++ b/src/vnet/l2tp/l2tp.c @@ -157,7 +157,7 @@ test_counters_command_fn (vlib_main_t * vm, u32 session_index; u32 counter_index; u32 nincr = 0; - u32 cpu_index = os_get_cpu_number (); + u32 thread_index = vlib_get_thread_index (); /* *INDENT-OFF* */ pool_foreach (session, lm->sessions, @@ -167,11 +167,11 @@ test_counters_command_fn (vlib_main_t * vm, session_index_to_counter_index (session_index, SESSION_COUNTER_USER_TO_NETWORK); vlib_increment_combined_counter (&lm->counter_main, - cpu_index, + thread_index, counter_index, 1/*pkt*/, 1111 /*bytes*/); vlib_increment_combined_counter (&lm->counter_main, - cpu_index, + thread_index, counter_index+1, 1/*pkt*/, 2222 /*bytes*/); nincr++; diff --git a/src/vnet/lisp-gpe/decap.c b/src/vnet/lisp-gpe/decap.c index d887a95f..68769710 100644 --- a/src/vnet/lisp-gpe/decap.c +++ b/src/vnet/lisp-gpe/decap.c @@ -103,7 +103,7 @@ next_index_to_iface (lisp_gpe_main_t * lgm, u32 next_index) } static_always_inline void -incr_decap_stats (vnet_main_t * vnm, u32 cpu_index, u32 length, +incr_decap_stats (vnet_main_t * vnm, u32 thread_index, u32 length, u32 sw_if_index, u32 * last_sw_if_index, u32 * n_packets, u32 * n_bytes) { @@ -122,7 +122,7 @@ incr_decap_stats (vnet_main_t * vnm, u32 cpu_index, u32 length, vlib_increment_combined_counter (im->combined_sw_if_counters + VNET_INTERFACE_COUNTER_RX, - cpu_index, *last_sw_if_index, + thread_index, *last_sw_if_index, *n_packets, *n_bytes); } *last_sw_if_index = sw_if_index; @@ -150,11 +150,11 @@ static uword lisp_gpe_input_inline (vlib_main_t * vm, vlib_node_runtime_t * node, vlib_frame_t * from_frame, u8 is_v4) { - u32 n_left_from, next_index, *from, *to_next, cpu_index; + u32 n_left_from, next_index, *from, *to_next, thread_index; u32 n_bytes = 0, n_packets = 0, last_sw_if_index = ~0, drops = 0; lisp_gpe_main_t *lgm = vnet_lisp_gpe_get_main (); - cpu_index = os_get_cpu_number (); + thread_index = vlib_get_thread_index (); from = vlib_frame_vector_args (from_frame); n_left_from = from_frame->n_vectors; @@ -267,7 +267,7 @@ lisp_gpe_input_inline (vlib_main_t * vm, vlib_node_runtime_t * node, if (si0) { - incr_decap_stats (lgm->vnet_main, cpu_index, + incr_decap_stats (lgm->vnet_main, thread_index, vlib_buffer_length_in_chain (vm, b0), si0[0], &last_sw_if_index, &n_packets, &n_bytes); vnet_buffer (b0)->sw_if_index[VLIB_RX] = si0[0]; @@ -282,7 +282,7 @@ lisp_gpe_input_inline (vlib_main_t * vm, vlib_node_runtime_t * node, if (si1) { - incr_decap_stats (lgm->vnet_main, cpu_index, + incr_decap_stats (lgm->vnet_main, thread_index, vlib_buffer_length_in_chain (vm, b1), si1[0], &last_sw_if_index, &n_packets, &n_bytes); vnet_buffer (b1)->sw_if_index[VLIB_RX] = si1[0]; @@ -397,7 +397,7 @@ lisp_gpe_input_inline (vlib_main_t * vm, vlib_node_runtime_t * node, if (si0) { - incr_decap_stats (lgm->vnet_main, cpu_index, + incr_decap_stats (lgm->vnet_main, thread_index, vlib_buffer_length_in_chain (vm, b0), si0[0], &last_sw_if_index, &n_packets, &n_bytes); vnet_buffer (b0)->sw_if_index[VLIB_RX] = si0[0]; @@ -430,7 +430,7 @@ lisp_gpe_input_inline (vlib_main_t * vm, vlib_node_runtime_t * node, } /* flush iface stats */ - incr_decap_stats (lgm->vnet_main, cpu_index, 0, ~0, &last_sw_if_index, + incr_decap_stats (lgm->vnet_main, thread_index, 0, ~0, &last_sw_if_index, &n_packets, &n_bytes); vlib_node_increment_counter (vm, lisp_gpe_ip4_input_node.index, LISP_GPE_ERROR_NO_TUNNEL, drops); diff --git a/src/vnet/lldp/lldp_input.c b/src/vnet/lldp/lldp_input.c index 762743d0..e88f6fdb 100644 --- a/src/vnet/lldp/lldp_input.c +++ b/src/vnet/lldp/lldp_input.c @@ -35,7 +35,7 @@ typedef struct static void lldp_rpc_update_peer_cb (const lldp_intf_update_t * a) { - ASSERT (os_get_cpu_number () == 0); + ASSERT (vlib_get_thread_index () == 0); lldp_intf_t *n = lldp_get_intf (&lldp_main, a->hw_if_index); if (!n) diff --git a/src/vnet/map/ip4_map.c b/src/vnet/map/ip4_map.c index 1a20d704..e39b6f14 100644 --- a/src/vnet/map/ip4_map.c +++ b/src/vnet/map/ip4_map.c @@ -248,7 +248,7 @@ ip4_map (vlib_main_t * vm, vlib_node_runtime_t * node, vlib_frame_t * frame) next_index = node->cached_next_index; map_main_t *mm = &map_main; vlib_combined_counter_main_t *cm = mm->domain_counters; - u32 cpu_index = os_get_cpu_number (); + u32 thread_index = vlib_get_thread_index (); while (n_left_from > 0) { @@ -377,7 +377,7 @@ ip4_map (vlib_main_t * vm, vlib_node_runtime_t * node, vlib_frame_t * frame) ip40) ? IP4_MAP_NEXT_IP6_REWRITE : next0; vlib_increment_combined_counter (cm + MAP_DOMAIN_COUNTER_TX, - cpu_index, + thread_index, map_domain_index0, 1, clib_net_to_host_u16 (ip6h0->payload_length) + @@ -409,7 +409,7 @@ ip4_map (vlib_main_t * vm, vlib_node_runtime_t * node, vlib_frame_t * frame) ip41) ? IP4_MAP_NEXT_IP6_REWRITE : next1; vlib_increment_combined_counter (cm + MAP_DOMAIN_COUNTER_TX, - cpu_index, + thread_index, map_domain_index1, 1, clib_net_to_host_u16 (ip6h1->payload_length) + @@ -520,7 +520,7 @@ ip4_map (vlib_main_t * vm, vlib_node_runtime_t * node, vlib_frame_t * frame) ip40) ? IP4_MAP_NEXT_IP6_REWRITE : next0; vlib_increment_combined_counter (cm + MAP_DOMAIN_COUNTER_TX, - cpu_index, + thread_index, map_domain_index0, 1, clib_net_to_host_u16 (ip6h0->payload_length) + @@ -564,7 +564,7 @@ ip4_map_reass (vlib_main_t * vm, next_index = node->cached_next_index; map_main_t *mm = &map_main; vlib_combined_counter_main_t *cm = mm->domain_counters; - u32 cpu_index = os_get_cpu_number (); + u32 thread_index = vlib_get_thread_index (); u32 *fragments_to_drop = NULL; u32 *fragments_to_loopback = NULL; @@ -694,8 +694,8 @@ ip4_map_reass (vlib_main_t * vm, { if (error0 == MAP_ERROR_NONE) vlib_increment_combined_counter (cm + MAP_DOMAIN_COUNTER_TX, - cpu_index, map_domain_index0, - 1, + thread_index, + map_domain_index0, 1, clib_net_to_host_u16 (ip60->payload_length) + 40); next0 = diff --git a/src/vnet/map/ip4_map_t.c b/src/vnet/map/ip4_map_t.c index b63d76bf..5f2bcbf9 100644 --- a/src/vnet/map/ip4_map_t.c +++ b/src/vnet/map/ip4_map_t.c @@ -477,7 +477,7 @@ ip4_map_t_icmp (vlib_main_t * vm, n_left_from = frame->n_vectors; next_index = node->cached_next_index; vlib_combined_counter_main_t *cm = map_main.domain_counters; - u32 cpu_index = os_get_cpu_number (); + u32 thread_index = vlib_get_thread_index (); while (n_left_from > 0) { @@ -520,7 +520,7 @@ ip4_map_t_icmp (vlib_main_t * vm, if (PREDICT_TRUE (error0 == MAP_ERROR_NONE)) { vlib_increment_combined_counter (cm + MAP_DOMAIN_COUNTER_TX, - cpu_index, + thread_index, vnet_buffer (p0)->map_t. map_domain_index, 1, len0); } @@ -1051,7 +1051,7 @@ ip4_map_t (vlib_main_t * vm, vlib_node_runtime_t * node, vlib_frame_t * frame) n_left_from = frame->n_vectors; next_index = node->cached_next_index; vlib_combined_counter_main_t *cm = map_main.domain_counters; - u32 cpu_index = os_get_cpu_number (); + u32 thread_index = vlib_get_thread_index (); while (n_left_from > 0) { @@ -1158,7 +1158,7 @@ ip4_map_t (vlib_main_t * vm, vlib_node_runtime_t * node, vlib_frame_t * frame) (error0 == MAP_ERROR_NONE && next0 != IP4_MAPT_NEXT_MAPT_ICMP)) { vlib_increment_combined_counter (cm + MAP_DOMAIN_COUNTER_TX, - cpu_index, + thread_index, vnet_buffer (p0)->map_t. map_domain_index, 1, clib_net_to_host_u16 (ip40-> @@ -1169,7 +1169,7 @@ ip4_map_t (vlib_main_t * vm, vlib_node_runtime_t * node, vlib_frame_t * frame) (error1 == MAP_ERROR_NONE && next1 != IP4_MAPT_NEXT_MAPT_ICMP)) { vlib_increment_combined_counter (cm + MAP_DOMAIN_COUNTER_TX, - cpu_index, + thread_index, vnet_buffer (p1)->map_t. map_domain_index, 1, clib_net_to_host_u16 (ip41-> @@ -1252,7 +1252,7 @@ ip4_map_t (vlib_main_t * vm, vlib_node_runtime_t * node, vlib_frame_t * frame) (error0 == MAP_ERROR_NONE && next0 != IP4_MAPT_NEXT_MAPT_ICMP)) { vlib_increment_combined_counter (cm + MAP_DOMAIN_COUNTER_TX, - cpu_index, + thread_index, vnet_buffer (p0)->map_t. map_domain_index, 1, clib_net_to_host_u16 (ip40-> diff --git a/src/vnet/map/ip6_map.c b/src/vnet/map/ip6_map.c index f7eb768f..63ada962 100644 --- a/src/vnet/map/ip6_map.c +++ b/src/vnet/map/ip6_map.c @@ -172,7 +172,7 @@ ip6_map (vlib_main_t * vm, vlib_node_runtime_t * node, vlib_frame_t * frame) vlib_node_get_runtime (vm, ip6_map_node.index); map_main_t *mm = &map_main; vlib_combined_counter_main_t *cm = mm->domain_counters; - u32 cpu_index = os_get_cpu_number (); + u32 thread_index = vlib_get_thread_index (); from = vlib_frame_vector_args (frame); n_left_from = frame->n_vectors; @@ -319,7 +319,7 @@ ip6_map (vlib_main_t * vm, vlib_node_runtime_t * node, vlib_frame_t * frame) IP6_MAP_NEXT_IP4_REWRITE : next0; } vlib_increment_combined_counter (cm + MAP_DOMAIN_COUNTER_RX, - cpu_index, + thread_index, map_domain_index0, 1, clib_net_to_host_u16 (ip40->length)); @@ -352,7 +352,7 @@ ip6_map (vlib_main_t * vm, vlib_node_runtime_t * node, vlib_frame_t * frame) IP6_MAP_NEXT_IP4_REWRITE : next1; } vlib_increment_combined_counter (cm + MAP_DOMAIN_COUNTER_RX, - cpu_index, + thread_index, map_domain_index1, 1, clib_net_to_host_u16 (ip41->length)); @@ -505,7 +505,7 @@ ip6_map (vlib_main_t * vm, vlib_node_runtime_t * node, vlib_frame_t * frame) IP6_MAP_NEXT_IP4_REWRITE : next0; } vlib_increment_combined_counter (cm + MAP_DOMAIN_COUNTER_RX, - cpu_index, + thread_index, map_domain_index0, 1, clib_net_to_host_u16 (ip40->length)); @@ -820,7 +820,7 @@ ip6_map_ip4_reass (vlib_main_t * vm, vlib_node_get_runtime (vm, ip6_map_ip4_reass_node.index); map_main_t *mm = &map_main; vlib_combined_counter_main_t *cm = mm->domain_counters; - u32 cpu_index = os_get_cpu_number (); + u32 thread_index = vlib_get_thread_index (); u32 *fragments_to_drop = NULL; u32 *fragments_to_loopback = NULL; @@ -958,8 +958,8 @@ ip6_map_ip4_reass (vlib_main_t * vm, { if (error0 == MAP_ERROR_NONE) vlib_increment_combined_counter (cm + MAP_DOMAIN_COUNTER_RX, - cpu_index, map_domain_index0, - 1, + thread_index, + map_domain_index0, 1, clib_net_to_host_u16 (ip40->length)); next0 = @@ -1015,7 +1015,7 @@ ip6_map_icmp_relay (vlib_main_t * vm, vlib_node_runtime_t *error_node = vlib_node_get_runtime (vm, ip6_map_icmp_relay_node.index); map_main_t *mm = &map_main; - u32 cpu_index = os_get_cpu_number (); + u32 thread_index = vlib_get_thread_index (); u16 *fragment_ids, *fid; from = vlib_frame_vector_args (frame); @@ -1143,7 +1143,8 @@ ip6_map_icmp_relay (vlib_main_t * vm, ip_csum_t sum = ip_incremental_checksum (0, new_icmp40, nlen - 20); new_icmp40->checksum = ~ip_csum_fold (sum); - vlib_increment_simple_counter (&mm->icmp_relayed, cpu_index, 0, 1); + vlib_increment_simple_counter (&mm->icmp_relayed, thread_index, 0, + 1); error: if (PREDICT_FALSE (p0->flags & VLIB_BUFFER_IS_TRACED)) diff --git a/src/vnet/map/ip6_map_t.c b/src/vnet/map/ip6_map_t.c index eb3996c2..99151678 100644 --- a/src/vnet/map/ip6_map_t.c +++ b/src/vnet/map/ip6_map_t.c @@ -448,7 +448,7 @@ ip6_map_t_icmp (vlib_main_t * vm, n_left_from = frame->n_vectors; next_index = node->cached_next_index; vlib_combined_counter_main_t *cm = map_main.domain_counters; - u32 cpu_index = os_get_cpu_number (); + u32 thread_index = vlib_get_thread_index (); while (n_left_from > 0) { @@ -493,7 +493,7 @@ ip6_map_t_icmp (vlib_main_t * vm, if (PREDICT_TRUE (error0 == MAP_ERROR_NONE)) { vlib_increment_combined_counter (cm + MAP_DOMAIN_COUNTER_RX, - cpu_index, + thread_index, vnet_buffer (p0)-> map_t.map_domain_index, 1, len0); @@ -1051,7 +1051,7 @@ ip6_map_t (vlib_main_t * vm, vlib_node_runtime_t * node, vlib_frame_t * frame) vlib_node_runtime_t *error_node = vlib_node_get_runtime (vm, ip6_map_t_node.index); vlib_combined_counter_main_t *cm = map_main.domain_counters; - u32 cpu_index = os_get_cpu_number (); + u32 thread_index = vlib_get_thread_index (); from = vlib_frame_vector_args (frame); n_left_from = frame->n_vectors; @@ -1218,7 +1218,7 @@ ip6_map_t (vlib_main_t * vm, vlib_node_runtime_t * node, vlib_frame_t * frame) (error0 == MAP_ERROR_NONE && next0 != IP6_MAPT_NEXT_MAPT_ICMP)) { vlib_increment_combined_counter (cm + MAP_DOMAIN_COUNTER_RX, - cpu_index, + thread_index, vnet_buffer (p0)-> map_t.map_domain_index, 1, clib_net_to_host_u16 @@ -1229,7 +1229,7 @@ ip6_map_t (vlib_main_t * vm, vlib_node_runtime_t * node, vlib_frame_t * frame) (error1 == MAP_ERROR_NONE && next1 != IP6_MAPT_NEXT_MAPT_ICMP)) { vlib_increment_combined_counter (cm + MAP_DOMAIN_COUNTER_RX, - cpu_index, + thread_index, vnet_buffer (p1)-> map_t.map_domain_index, 1, clib_net_to_host_u16 @@ -1403,7 +1403,7 @@ ip6_map_t (vlib_main_t * vm, vlib_node_runtime_t * node, vlib_frame_t * frame) (error0 == MAP_ERROR_NONE && next0 != IP6_MAPT_NEXT_MAPT_ICMP)) { vlib_increment_combined_counter (cm + MAP_DOMAIN_COUNTER_RX, - cpu_index, + thread_index, vnet_buffer (p0)-> map_t.map_domain_index, 1, clib_net_to_host_u16 diff --git a/src/vnet/mpls/mpls_input.c b/src/vnet/mpls/mpls_input.c index 893c4511..1b9bdd05 100644 --- a/src/vnet/mpls/mpls_input.c +++ b/src/vnet/mpls/mpls_input.c @@ -76,7 +76,7 @@ mpls_input_inline (vlib_main_t * vm, u32 n_left_from, next_index, * from, * to_next; mpls_input_runtime_t * rt; mpls_main_t * mm; - u32 cpu_index = os_get_cpu_number(); + u32 thread_index = vlib_get_thread_index(); vlib_simple_counter_main_t * cm; vnet_main_t * vnm = vnet_get_main(); @@ -151,7 +151,7 @@ mpls_input_inline (vlib_main_t * vm, next0 = MPLS_INPUT_NEXT_LOOKUP; vnet_feature_arc_start(mm->input_feature_arc_index, sw_if_index0, &next0, b0); - vlib_increment_simple_counter (cm, cpu_index, sw_if_index0, 1); + vlib_increment_simple_counter (cm, thread_index, sw_if_index0, 1); } if (PREDICT_FALSE(h1[3] == 0)) @@ -164,7 +164,7 @@ mpls_input_inline (vlib_main_t * vm, next1 = MPLS_INPUT_NEXT_LOOKUP; vnet_feature_arc_start(mm->input_feature_arc_index, sw_if_index1, &next1, b1); - vlib_increment_simple_counter (cm, cpu_index, sw_if_index1, 1); + vlib_increment_simple_counter (cm, thread_index, sw_if_index1, 1); } if (PREDICT_FALSE(b0->flags & VLIB_BUFFER_IS_TRACED)) @@ -215,7 +215,7 @@ mpls_input_inline (vlib_main_t * vm, { next0 = MPLS_INPUT_NEXT_LOOKUP; vnet_feature_arc_start(mm->input_feature_arc_index, sw_if_index0, &next0, b0); - vlib_increment_simple_counter (cm, cpu_index, sw_if_index0, 1); + vlib_increment_simple_counter (cm, thread_index, sw_if_index0, 1); } if (PREDICT_FALSE(b0->flags & VLIB_BUFFER_IS_TRACED)) diff --git a/src/vnet/mpls/mpls_lookup.c b/src/vnet/mpls/mpls_lookup.c index 475bb204..ace6a70f 100644 --- a/src/vnet/mpls/mpls_lookup.c +++ b/src/vnet/mpls/mpls_lookup.c @@ -67,7 +67,7 @@ mpls_lookup (vlib_main_t * vm, vlib_combined_counter_main_t * cm = &load_balance_main.lbm_to_counters; u32 n_left_from, next_index, * from, * to_next; mpls_main_t * mm = &mpls_main; - u32 cpu_index = os_get_cpu_number(); + u32 thread_index = vlib_get_thread_index(); from = vlib_frame_vector_args (from_frame); n_left_from = from_frame->n_vectors; @@ -220,16 +220,16 @@ mpls_lookup (vlib_main_t * vm, vnet_buffer (b3)->ip.adj_index[VLIB_TX] = dpo3->dpoi_index; vlib_increment_combined_counter - (cm, cpu_index, lbi0, 1, + (cm, thread_index, lbi0, 1, vlib_buffer_length_in_chain (vm, b0)); vlib_increment_combined_counter - (cm, cpu_index, lbi1, 1, + (cm, thread_index, lbi1, 1, vlib_buffer_length_in_chain (vm, b1)); vlib_increment_combined_counter - (cm, cpu_index, lbi2, 1, + (cm, thread_index, lbi2, 1, vlib_buffer_length_in_chain (vm, b2)); vlib_increment_combined_counter - (cm, cpu_index, lbi3, 1, + (cm, thread_index, lbi3, 1, vlib_buffer_length_in_chain (vm, b3)); /* @@ -351,7 +351,7 @@ mpls_lookup (vlib_main_t * vm, vnet_buffer (b0)->ip.adj_index[VLIB_TX] = dpo0->dpoi_index; vlib_increment_combined_counter - (cm, cpu_index, lbi0, 1, + (cm, thread_index, lbi0, 1, vlib_buffer_length_in_chain (vm, b0)); /* @@ -440,7 +440,7 @@ mpls_load_balance (vlib_main_t * vm, { vlib_combined_counter_main_t * cm = &load_balance_main.lbm_via_counters; u32 n_left_from, n_left_to_next, * from, * to_next; - u32 cpu_index = os_get_cpu_number(); + u32 thread_index = vlib_get_thread_index(); u32 next; from = vlib_frame_vector_args (frame); @@ -536,10 +536,10 @@ mpls_load_balance (vlib_main_t * vm, vnet_buffer (p1)->ip.adj_index[VLIB_TX] = dpo1->dpoi_index; vlib_increment_combined_counter - (cm, cpu_index, lbi0, 1, + (cm, thread_index, lbi0, 1, vlib_buffer_length_in_chain (vm, p0)); vlib_increment_combined_counter - (cm, cpu_index, lbi1, 1, + (cm, thread_index, lbi1, 1, vlib_buffer_length_in_chain (vm, p1)); if (PREDICT_FALSE(p0->flags & VLIB_BUFFER_IS_TRACED)) @@ -597,7 +597,7 @@ mpls_load_balance (vlib_main_t * vm, vnet_buffer (p0)->ip.adj_index[VLIB_TX] = dpo0->dpoi_index; vlib_increment_combined_counter - (cm, cpu_index, lbi0, 1, + (cm, thread_index, lbi0, 1, vlib_buffer_length_in_chain (vm, p0)); vlib_validate_buffer_enqueue_x1 (vm, node, next, diff --git a/src/vnet/mpls/mpls_output.c b/src/vnet/mpls/mpls_output.c index 08018fd1..d90dec21 100644 --- a/src/vnet/mpls/mpls_output.c +++ b/src/vnet/mpls/mpls_output.c @@ -64,12 +64,12 @@ mpls_output_inline (vlib_main_t * vm, vlib_frame_t * from_frame, int is_midchain) { - u32 n_left_from, next_index, * from, * to_next, cpu_index; + u32 n_left_from, next_index, * from, * to_next, thread_index; vlib_node_runtime_t * error_node; u32 n_left_to_next; mpls_main_t *mm; - cpu_index = os_get_cpu_number(); + thread_index = vlib_get_thread_index(); error_node = vlib_node_get_runtime (vm, mpls_output_node.index); from = vlib_frame_vector_args (from_frame); n_left_from = from_frame->n_vectors; @@ -137,13 +137,13 @@ mpls_output_inline (vlib_main_t * vm, /* Bump the adj counters for packet and bytes */ vlib_increment_combined_counter (&adjacency_counters, - cpu_index, + thread_index, adj_index0, 1, vlib_buffer_length_in_chain (vm, p0) + rw_len0); vlib_increment_combined_counter (&adjacency_counters, - cpu_index, + thread_index, adj_index1, 1, vlib_buffer_length_in_chain (vm, p1) + rw_len1); @@ -245,7 +245,7 @@ mpls_output_inline (vlib_main_t * vm, vlib_increment_combined_counter (&adjacency_counters, - cpu_index, + thread_index, adj_index0, 1, vlib_buffer_length_in_chain (vm, p0) + rw_len0); diff --git a/src/vnet/pg/input.c b/src/vnet/pg/input.c index 2649798b..597ae060 100644 --- a/src/vnet/pg/input.c +++ b/src/vnet/pg/input.c @@ -893,7 +893,7 @@ pg_generate_set_lengths (pg_main_t * pg, vlib_increment_combined_counter (im->combined_sw_if_counters + VNET_INTERFACE_COUNTER_RX, - os_get_cpu_number (), + vlib_get_thread_index (), si->sw_if_index, n_buffers, length_sum); } @@ -1266,7 +1266,7 @@ pg_stream_fill_helper (pg_main_t * pg, l += vlib_buffer_index_length_in_chain (vm, buffers[i]); vlib_increment_combined_counter (im->combined_sw_if_counters + VNET_INTERFACE_COUNTER_RX, - os_get_cpu_number (), + vlib_get_thread_index (), si->sw_if_index, n_alloc, l); s->current_replay_packet_index += n_alloc; s->current_replay_packet_index %= diff --git a/src/vnet/replication.c b/src/vnet/replication.c index 86d922b5..233a8c2f 100644 --- a/src/vnet/replication.c +++ b/src/vnet/replication.c @@ -31,16 +31,16 @@ replication_prep (vlib_main_t * vm, { replication_main_t *rm = &replication_main; replication_context_t *ctx; - uword cpu_number = vm->cpu_index; + uword thread_index = vm->thread_index; ip4_header_t *ip; u32 ctx_id; /* Allocate a context, reserve context 0 */ - if (PREDICT_FALSE (rm->contexts[cpu_number] == 0)) - pool_get_aligned (rm->contexts[cpu_number], ctx, CLIB_CACHE_LINE_BYTES); + if (PREDICT_FALSE (rm->contexts[thread_index] == 0)) + pool_get_aligned (rm->contexts[thread_index], ctx, CLIB_CACHE_LINE_BYTES); - pool_get_aligned (rm->contexts[cpu_number], ctx, CLIB_CACHE_LINE_BYTES); - ctx_id = ctx - rm->contexts[cpu_number]; + pool_get_aligned (rm->contexts[thread_index], ctx, CLIB_CACHE_LINE_BYTES); + ctx_id = ctx - rm->contexts[thread_index]; /* Save state from vlib buffer */ ctx->saved_free_list_index = b0->free_list_index; @@ -94,11 +94,11 @@ replication_recycle (vlib_main_t * vm, vlib_buffer_t * b0, u32 is_last) { replication_main_t *rm = &replication_main; replication_context_t *ctx; - uword cpu_number = vm->cpu_index; + uword thread_index = vm->thread_index; ip4_header_t *ip; /* Get access to the replication context */ - ctx = pool_elt_at_index (rm->contexts[cpu_number], b0->recycle_count); + ctx = pool_elt_at_index (rm->contexts[thread_index], b0->recycle_count); /* Restore vnet buffer state */ clib_memcpy (vnet_buffer (b0), ctx->vnet_buffer, @@ -133,7 +133,7 @@ replication_recycle (vlib_main_t * vm, vlib_buffer_t * b0, u32 is_last) b0->flags &= ~VLIB_BUFFER_RECYCLE; /* Free context back to its pool */ - pool_put (rm->contexts[cpu_number], ctx); + pool_put (rm->contexts[thread_index], ctx); } return ctx; @@ -160,7 +160,7 @@ replication_recycle_callback (vlib_main_t * vm, vlib_buffer_free_list_t * fl) replication_main_t *rm = &replication_main; replication_context_t *ctx; u32 feature_node_index = 0; - uword cpu_number = vm->cpu_index; + uword thread_index = vm->thread_index; /* * All buffers in the list are destined to the same recycle node. @@ -172,7 +172,7 @@ replication_recycle_callback (vlib_main_t * vm, vlib_buffer_free_list_t * fl) { bi0 = fl->buffers[0]; b0 = vlib_get_buffer (vm, bi0); - ctx = pool_elt_at_index (rm->contexts[cpu_number], b0->recycle_count); + ctx = pool_elt_at_index (rm->contexts[thread_index], b0->recycle_count); feature_node_index = ctx->recycle_node_index; } diff --git a/src/vnet/replication.h b/src/vnet/replication.h index 5dc554c9..ce4b3ff1 100644 --- a/src/vnet/replication.h +++ b/src/vnet/replication.h @@ -100,7 +100,7 @@ replication_get_ctx (vlib_buffer_t * b0) replication_main_t *rm = &replication_main; return replication_is_recycled (b0) ? - pool_elt_at_index (rm->contexts[os_get_cpu_number ()], + pool_elt_at_index (rm->contexts[vlib_get_thread_index ()], b0->recycle_count) : 0; } diff --git a/src/vnet/session/node.c b/src/vnet/session/node.c index b86e87d9..dd211c51 100644 --- a/src/vnet/session/node.c +++ b/src/vnet/session/node.c @@ -311,7 +311,7 @@ session_queue_node_fn (vlib_main_t * vm, vlib_node_runtime_t * node, unix_shared_memory_queue_t *q; application_t *app; int n_tx_packets = 0; - u32 my_thread_index = vm->cpu_index; + u32 my_thread_index = vm->thread_index; int i, rv; f64 now = vlib_time_now (vm); diff --git a/src/vnet/sr/sr_localsid.c b/src/vnet/sr/sr_localsid.c index 2e3d56de..6d72a506 100755 --- a/src/vnet/sr/sr_localsid.c +++ b/src/vnet/sr/sr_localsid.c @@ -887,7 +887,7 @@ sr_localsid_d_fn (vlib_main_t * vm, vlib_node_runtime_t * node, from = vlib_frame_vector_args (from_frame); n_left_from = from_frame->n_vectors; next_index = node->cached_next_index; - u32 cpu_index = os_get_cpu_number (); + u32 thread_index = vlib_get_thread_index (); while (n_left_from > 0) { @@ -974,26 +974,26 @@ sr_localsid_d_fn (vlib_main_t * vm, vlib_node_runtime_t * node, vlib_increment_combined_counter (((next0 == SR_LOCALSID_NEXT_ERROR) ? &(sm->sr_ls_invalid_counters) : - &(sm->sr_ls_valid_counters)), cpu_index, ls0 - sm->localsids, 1, - vlib_buffer_length_in_chain (vm, b0)); + &(sm->sr_ls_valid_counters)), thread_index, ls0 - sm->localsids, + 1, vlib_buffer_length_in_chain (vm, b0)); vlib_increment_combined_counter (((next1 == SR_LOCALSID_NEXT_ERROR) ? &(sm->sr_ls_invalid_counters) : - &(sm->sr_ls_valid_counters)), cpu_index, ls1 - sm->localsids, 1, - vlib_buffer_length_in_chain (vm, b1)); + &(sm->sr_ls_valid_counters)), thread_index, ls1 - sm->localsids, + 1, vlib_buffer_length_in_chain (vm, b1)); vlib_increment_combined_counter (((next2 == SR_LOCALSID_NEXT_ERROR) ? &(sm->sr_ls_invalid_counters) : - &(sm->sr_ls_valid_counters)), cpu_index, ls2 - sm->localsids, 1, - vlib_buffer_length_in_chain (vm, b2)); + &(sm->sr_ls_valid_counters)), thread_index, ls2 - sm->localsids, + 1, vlib_buffer_length_in_chain (vm, b2)); vlib_increment_combined_counter (((next3 == SR_LOCALSID_NEXT_ERROR) ? &(sm->sr_ls_invalid_counters) : - &(sm->sr_ls_valid_counters)), cpu_index, ls3 - sm->localsids, 1, - vlib_buffer_length_in_chain (vm, b3)); + &(sm->sr_ls_valid_counters)), thread_index, ls3 - sm->localsids, + 1, vlib_buffer_length_in_chain (vm, b3)); vlib_validate_buffer_enqueue_x4 (vm, node, next_index, to_next, n_left_to_next, bi0, bi1, bi2, bi3, @@ -1062,8 +1062,8 @@ sr_localsid_d_fn (vlib_main_t * vm, vlib_node_runtime_t * node, vlib_increment_combined_counter (((next0 == SR_LOCALSID_NEXT_ERROR) ? &(sm->sr_ls_invalid_counters) : - &(sm->sr_ls_valid_counters)), cpu_index, ls0 - sm->localsids, 1, - vlib_buffer_length_in_chain (vm, b0)); + &(sm->sr_ls_valid_counters)), thread_index, ls0 - sm->localsids, + 1, vlib_buffer_length_in_chain (vm, b0)); vlib_validate_buffer_enqueue_x1 (vm, node, next_index, to_next, n_left_to_next, bi0, next0); @@ -1103,7 +1103,7 @@ sr_localsid_fn (vlib_main_t * vm, vlib_node_runtime_t * node, from = vlib_frame_vector_args (from_frame); n_left_from = from_frame->n_vectors; next_index = node->cached_next_index; - u32 cpu_index = os_get_cpu_number (); + u32 thread_index = vlib_get_thread_index (); while (n_left_from > 0) { @@ -1205,26 +1205,26 @@ sr_localsid_fn (vlib_main_t * vm, vlib_node_runtime_t * node, vlib_increment_combined_counter (((next0 == SR_LOCALSID_NEXT_ERROR) ? &(sm->sr_ls_invalid_counters) : - &(sm->sr_ls_valid_counters)), cpu_index, ls0 - sm->localsids, 1, - vlib_buffer_length_in_chain (vm, b0)); + &(sm->sr_ls_valid_counters)), thread_index, ls0 - sm->localsids, + 1, vlib_buffer_length_in_chain (vm, b0)); vlib_increment_combined_counter (((next1 == SR_LOCALSID_NEXT_ERROR) ? &(sm->sr_ls_invalid_counters) : - &(sm->sr_ls_valid_counters)), cpu_index, ls1 - sm->localsids, 1, - vlib_buffer_length_in_chain (vm, b1)); + &(sm->sr_ls_valid_counters)), thread_index, ls1 - sm->localsids, + 1, vlib_buffer_length_in_chain (vm, b1)); vlib_increment_combined_counter (((next2 == SR_LOCALSID_NEXT_ERROR) ? &(sm->sr_ls_invalid_counters) : - &(sm->sr_ls_valid_counters)), cpu_index, ls2 - sm->localsids, 1, - vlib_buffer_length_in_chain (vm, b2)); + &(sm->sr_ls_valid_counters)), thread_index, ls2 - sm->localsids, + 1, vlib_buffer_length_in_chain (vm, b2)); vlib_increment_combined_counter (((next3 == SR_LOCALSID_NEXT_ERROR) ? &(sm->sr_ls_invalid_counters) : - &(sm->sr_ls_valid_counters)), cpu_index, ls3 - sm->localsids, 1, - vlib_buffer_length_in_chain (vm, b3)); + &(sm->sr_ls_valid_counters)), thread_index, ls3 - sm->localsids, + 1, vlib_buffer_length_in_chain (vm, b3)); vlib_validate_buffer_enqueue_x4 (vm, node, next_index, to_next, n_left_to_next, bi0, bi1, bi2, bi3, @@ -1295,8 +1295,8 @@ sr_localsid_fn (vlib_main_t * vm, vlib_node_runtime_t * node, vlib_increment_combined_counter (((next0 == SR_LOCALSID_NEXT_ERROR) ? &(sm->sr_ls_invalid_counters) : - &(sm->sr_ls_valid_counters)), cpu_index, ls0 - sm->localsids, 1, - vlib_buffer_length_in_chain (vm, b0)); + &(sm->sr_ls_valid_counters)), thread_index, ls0 - sm->localsids, + 1, vlib_buffer_length_in_chain (vm, b0)); vlib_validate_buffer_enqueue_x1 (vm, node, next_index, to_next, n_left_to_next, bi0, next0); diff --git a/src/vnet/tcp/builtin_client.c b/src/vnet/tcp/builtin_client.c index e3705060..c1567aa0 100644 --- a/src/vnet/tcp/builtin_client.c +++ b/src/vnet/tcp/builtin_client.c @@ -174,7 +174,7 @@ tclient_thread_fn (void *arg) pthread_sigmask (SIG_SETMASK, &s, 0); } - clib_per_cpu_mheaps[os_get_cpu_number ()] = clib_per_cpu_mheaps[0]; + clib_per_cpu_mheaps[vlib_get_thread_index ()] = clib_per_cpu_mheaps[0]; while (1) { diff --git a/src/vnet/tcp/tcp.c b/src/vnet/tcp/tcp.c index b2a371e2..b6c34828 100644 --- a/src/vnet/tcp/tcp.c +++ b/src/vnet/tcp/tcp.c @@ -646,10 +646,10 @@ const static transport_proto_vft_t tcp6_proto = { void tcp_timer_keep_handler (u32 conn_index) { - u32 cpu_index = os_get_cpu_number (); + u32 thread_index = vlib_get_thread_index (); tcp_connection_t *tc; - tc = tcp_connection_get (conn_index, cpu_index); + tc = tcp_connection_get (conn_index, thread_index); tc->timers[TCP_TIMER_KEEP] = TCP_TIMER_HANDLE_INVALID; tcp_connection_close (tc); @@ -675,10 +675,10 @@ tcp_timer_establish_handler (u32 conn_index) void tcp_timer_waitclose_handler (u32 conn_index) { - u32 cpu_index = os_get_cpu_number (); + u32 thread_index = vlib_get_thread_index (); tcp_connection_t *tc; - tc = tcp_connection_get (conn_index, cpu_index); + tc = tcp_connection_get (conn_index, thread_index); tc->timers[TCP_TIMER_WAITCLOSE] = TCP_TIMER_HANDLE_INVALID; /* Session didn't come back with a close(). Send FIN either way diff --git a/src/vnet/tcp/tcp_debug.h b/src/vnet/tcp/tcp_debug.h index 0090e15e..eaca672c 100644 --- a/src/vnet/tcp/tcp_debug.h +++ b/src/vnet/tcp/tcp_debug.h @@ -343,7 +343,7 @@ typedef enum _tcp_dbg_evt } \ else \ { \ - u32 _thread_index = os_get_cpu_number (); \ + u32 _thread_index = vlib_get_thread_index (); \ _tc = tcp_connection_get (_tc_index, _thread_index); \ } \ ELOG_TYPE_DECLARE (_e) = \ diff --git a/src/vnet/tcp/tcp_input.c b/src/vnet/tcp/tcp_input.c index a8224dc2..7e9fa47b 100644 --- a/src/vnet/tcp/tcp_input.c +++ b/src/vnet/tcp/tcp_input.c @@ -1142,7 +1142,7 @@ tcp46_established_inline (vlib_main_t * vm, vlib_node_runtime_t * node, vlib_frame_t * from_frame, int is_ip4) { u32 n_left_from, next_index, *from, *to_next; - u32 my_thread_index = vm->cpu_index, errors = 0; + u32 my_thread_index = vm->thread_index, errors = 0; tcp_main_t *tm = vnet_get_tcp_main (); from = vlib_frame_vector_args (from_frame); @@ -1332,7 +1332,7 @@ tcp46_syn_sent_inline (vlib_main_t * vm, vlib_node_runtime_t * node, { tcp_main_t *tm = vnet_get_tcp_main (); u32 n_left_from, next_index, *from, *to_next; - u32 my_thread_index = vm->cpu_index, errors = 0; + u32 my_thread_index = vm->thread_index, errors = 0; u8 sst = is_ip4 ? SESSION_TYPE_IP4_TCP : SESSION_TYPE_IP6_TCP; from = vlib_frame_vector_args (from_frame); @@ -1634,7 +1634,7 @@ tcp46_rcv_process_inline (vlib_main_t * vm, vlib_node_runtime_t * node, { tcp_main_t *tm = vnet_get_tcp_main (); u32 n_left_from, next_index, *from, *to_next; - u32 my_thread_index = vm->cpu_index, errors = 0; + u32 my_thread_index = vm->thread_index, errors = 0; from = vlib_frame_vector_args (from_frame); n_left_from = from_frame->n_vectors; @@ -1989,7 +1989,7 @@ tcp46_listen_inline (vlib_main_t * vm, vlib_node_runtime_t * node, vlib_frame_t * from_frame, int is_ip4) { u32 n_left_from, next_index, *from, *to_next; - u32 my_thread_index = vm->cpu_index; + u32 my_thread_index = vm->thread_index; tcp_main_t *tm = vnet_get_tcp_main (); u8 sst = is_ip4 ? SESSION_TYPE_IP4_TCP : SESSION_TYPE_IP6_TCP; @@ -2243,7 +2243,7 @@ tcp46_input_inline (vlib_main_t * vm, vlib_node_runtime_t * node, vlib_frame_t * from_frame, int is_ip4) { u32 n_left_from, next_index, *from, *to_next; - u32 my_thread_index = vm->cpu_index; + u32 my_thread_index = vm->thread_index; tcp_main_t *tm = vnet_get_tcp_main (); from = vlib_frame_vector_args (from_frame); diff --git a/src/vnet/tcp/tcp_output.c b/src/vnet/tcp/tcp_output.c index ea157bd7..e18bfad7 100644 --- a/src/vnet/tcp/tcp_output.c +++ b/src/vnet/tcp/tcp_output.c @@ -387,8 +387,8 @@ tcp_make_options (tcp_connection_t * tc, tcp_options_t * opts, #define tcp_get_free_buffer_index(tm, bidx) \ do { \ u32 *my_tx_buffers, n_free_buffers; \ - u32 cpu_index = os_get_cpu_number(); \ - my_tx_buffers = tm->tx_buffers[cpu_index]; \ + u32 thread_index = vlib_get_thread_index(); \ + my_tx_buffers = tm->tx_buffers[thread_index]; \ if (PREDICT_FALSE(vec_len (my_tx_buffers) == 0)) \ { \ n_free_buffers = 32; /* TODO config or macro */ \ @@ -396,7 +396,7 @@ do { \ _vec_len(my_tx_buffers) = vlib_buffer_alloc_from_free_list ( \ tm->vlib_main, my_tx_buffers, n_free_buffers, \ VLIB_BUFFER_DEFAULT_FREE_LIST_INDEX); \ - tm->tx_buffers[cpu_index] = my_tx_buffers; \ + tm->tx_buffers[thread_index] = my_tx_buffers; \ } \ /* buffer shortage */ \ if (PREDICT_FALSE (vec_len (my_tx_buffers) == 0)) \ @@ -408,8 +408,8 @@ do { \ #define tcp_return_buffer(tm) \ do { \ u32 *my_tx_buffers; \ - u32 cpu_index = os_get_cpu_number(); \ - my_tx_buffers = tm->tx_buffers[cpu_index]; \ + u32 thread_index = vlib_get_thread_index(); \ + my_tx_buffers = tm->tx_buffers[thread_index]; \ _vec_len (my_tx_buffers) +=1; \ } while (0) @@ -942,7 +942,7 @@ tcp_send_ack (tcp_connection_t * tc) void tcp_timer_delack_handler (u32 index) { - u32 thread_index = os_get_cpu_number (); + u32 thread_index = vlib_get_thread_index (); tcp_connection_t *tc; tc = tcp_connection_get (index, thread_index); @@ -1022,7 +1022,7 @@ tcp_timer_retransmit_handler_i (u32 index, u8 is_syn) { tcp_main_t *tm = vnet_get_tcp_main (); vlib_main_t *vm = vlib_get_main (); - u32 thread_index = os_get_cpu_number (); + u32 thread_index = vlib_get_thread_index (); tcp_connection_t *tc; vlib_buffer_t *b; u32 bi, snd_space, n_bytes; @@ -1152,7 +1152,7 @@ tcp_timer_persist_handler (u32 index) { tcp_main_t *tm = vnet_get_tcp_main (); vlib_main_t *vm = vlib_get_main (); - u32 thread_index = os_get_cpu_number (); + u32 thread_index = vlib_get_thread_index (); tcp_connection_t *tc; vlib_buffer_t *b; u32 bi, n_bytes; @@ -1313,7 +1313,7 @@ tcp46_output_inline (vlib_main_t * vm, vlib_frame_t * from_frame, int is_ip4) { u32 n_left_from, next_index, *from, *to_next; - u32 my_thread_index = vm->cpu_index; + u32 my_thread_index = vm->thread_index; from = vlib_frame_vector_args (from_frame); n_left_from = from_frame->n_vectors; @@ -1524,7 +1524,7 @@ tcp46_send_reset_inline (vlib_main_t * vm, vlib_node_runtime_t * node, vlib_frame_t * from_frame, u8 is_ip4) { u32 n_left_from, next_index, *from, *to_next; - u32 my_thread_index = vm->cpu_index; + u32 my_thread_index = vm->thread_index; from = vlib_frame_vector_args (from_frame); n_left_from = from_frame->n_vectors; diff --git a/src/vnet/udp/udp_input.c b/src/vnet/udp/udp_input.c index 4b22109b..810278e6 100644 --- a/src/vnet/udp/udp_input.c +++ b/src/vnet/udp/udp_input.c @@ -70,7 +70,7 @@ udp4_uri_input_node_fn (vlib_main_t * vm, udp4_uri_input_next_t next_index; udp_uri_main_t *um = vnet_get_udp_main (); session_manager_main_t *smm = vnet_get_session_manager_main (); - u32 my_thread_index = vm->cpu_index; + u32 my_thread_index = vm->thread_index; u8 my_enqueue_epoch; u32 *session_indices_to_enqueue; static u32 serial_number; diff --git a/src/vnet/unix/tapcli.c b/src/vnet/unix/tapcli.c index fb1a8bac..0fc62f6c 100644 --- a/src/vnet/unix/tapcli.c +++ b/src/vnet/unix/tapcli.c @@ -366,7 +366,7 @@ static uword tapcli_rx_iface(vlib_main_t * vm, vlib_increment_combined_counter ( vnet_main.interface_main.combined_sw_if_counters + VNET_INTERFACE_COUNTER_RX, - os_get_cpu_number(), ti->sw_if_index, + vlib_get_thread_index(), ti->sw_if_index, 1, n_bytes_in_packet); if (PREDICT_FALSE(n_trace > 0)) { diff --git a/src/vnet/unix/tuntap.c b/src/vnet/unix/tuntap.c index 2cfcc92f..ac674653 100644 --- a/src/vnet/unix/tuntap.c +++ b/src/vnet/unix/tuntap.c @@ -189,7 +189,7 @@ tuntap_tx (vlib_main_t * vm, /* Update tuntap interface output stats. */ vlib_increment_combined_counter (im->combined_sw_if_counters + VNET_INTERFACE_COUNTER_TX, - vm->cpu_index, + vm->thread_index, tm->sw_if_index, n_packets, n_bytes); @@ -297,7 +297,7 @@ tuntap_rx (vlib_main_t * vm, vlib_increment_combined_counter (vnet_main.interface_main.combined_sw_if_counters + VNET_INTERFACE_COUNTER_RX, - os_get_cpu_number(), + vlib_get_thread_index(), tm->sw_if_index, 1, n_bytes_in_packet); diff --git a/src/vnet/vxlan-gpe/decap.c b/src/vnet/vxlan-gpe/decap.c index 22ab4b62..d4fe4231 100644 --- a/src/vnet/vxlan-gpe/decap.c +++ b/src/vnet/vxlan-gpe/decap.c @@ -115,7 +115,7 @@ vxlan_gpe_input (vlib_main_t * vm, vxlan4_gpe_tunnel_key_t last_key4; vxlan6_gpe_tunnel_key_t last_key6; u32 pkts_decapsulated = 0; - u32 cpu_index = os_get_cpu_number (); + u32 thread_index = vlib_get_thread_index (); u32 stats_sw_if_index, stats_n_packets, stats_n_bytes; if (is_ip4) @@ -342,7 +342,7 @@ vxlan_gpe_input (vlib_main_t * vm, if (stats_n_packets) vlib_increment_combined_counter ( im->combined_sw_if_counters + VNET_INTERFACE_COUNTER_RX, - cpu_index, stats_sw_if_index, stats_n_packets, stats_n_bytes); + thread_index, stats_sw_if_index, stats_n_packets, stats_n_bytes); stats_n_packets = 1; stats_n_bytes = len0; stats_sw_if_index = sw_if_index0; @@ -427,7 +427,7 @@ vxlan_gpe_input (vlib_main_t * vm, if (stats_n_packets) vlib_increment_combined_counter ( im->combined_sw_if_counters + VNET_INTERFACE_COUNTER_RX, - cpu_index, stats_sw_if_index, stats_n_packets, stats_n_bytes); + thread_index, stats_sw_if_index, stats_n_packets, stats_n_bytes); stats_n_packets = 1; stats_n_bytes = len1; stats_sw_if_index = sw_if_index1; @@ -588,7 +588,7 @@ vxlan_gpe_input (vlib_main_t * vm, if (stats_n_packets) vlib_increment_combined_counter ( im->combined_sw_if_counters + VNET_INTERFACE_COUNTER_RX, - cpu_index, stats_sw_if_index, stats_n_packets, stats_n_bytes); + thread_index, stats_sw_if_index, stats_n_packets, stats_n_bytes); stats_n_packets = 1; stats_n_bytes = len0; stats_sw_if_index = sw_if_index0; @@ -615,7 +615,7 @@ vxlan_gpe_input (vlib_main_t * vm, if (stats_n_packets) { vlib_increment_combined_counter ( - im->combined_sw_if_counters + VNET_INTERFACE_COUNTER_RX, cpu_index, + im->combined_sw_if_counters + VNET_INTERFACE_COUNTER_RX, thread_index, stats_sw_if_index, stats_n_packets, stats_n_bytes); node->runtime_data[0] = stats_sw_if_index; } diff --git a/src/vnet/vxlan-gpe/encap.c b/src/vnet/vxlan-gpe/encap.c index 3a486e56..67ed94b4 100644 --- a/src/vnet/vxlan-gpe/encap.c +++ b/src/vnet/vxlan-gpe/encap.c @@ -151,7 +151,7 @@ vxlan_gpe_encap (vlib_main_t * vm, vnet_main_t * vnm = ngm->vnet_main; vnet_interface_main_t * im = &vnm->interface_main; u32 pkts_encapsulated = 0; - u32 cpu_index = os_get_cpu_number (); + u32 thread_index = vlib_get_thread_index (); u32 stats_sw_if_index, stats_n_packets, stats_n_bytes; from = vlib_frame_vector_args (from_frame); @@ -253,7 +253,7 @@ vxlan_gpe_encap (vlib_main_t * vm, if (stats_n_packets) vlib_increment_combined_counter ( im->combined_sw_if_counters + VNET_INTERFACE_COUNTER_TX, - cpu_index, stats_sw_if_index, stats_n_packets, stats_n_bytes); + thread_index, stats_sw_if_index, stats_n_packets, stats_n_bytes); stats_sw_if_index = sw_if_index0; stats_n_packets = 2; stats_n_bytes = len0 + len1; @@ -262,10 +262,10 @@ vxlan_gpe_encap (vlib_main_t * vm, { vlib_increment_combined_counter ( im->combined_sw_if_counters + VNET_INTERFACE_COUNTER_TX, - cpu_index, sw_if_index0, 1, len0); + thread_index, sw_if_index0, 1, len0); vlib_increment_combined_counter ( im->combined_sw_if_counters + VNET_INTERFACE_COUNTER_TX, - cpu_index, sw_if_index1, 1, len1); + thread_index, sw_if_index1, 1, len1); } } @@ -335,7 +335,7 @@ vxlan_gpe_encap (vlib_main_t * vm, if (stats_n_packets) vlib_increment_combined_counter ( im->combined_sw_if_counters + VNET_INTERFACE_COUNTER_TX, - cpu_index, stats_sw_if_index, stats_n_packets, stats_n_bytes); + thread_index, stats_sw_if_index, stats_n_packets, stats_n_bytes); stats_n_packets = 1; stats_n_bytes = len0; stats_sw_if_index = sw_if_index0; @@ -359,7 +359,7 @@ vxlan_gpe_encap (vlib_main_t * vm, if (stats_n_packets) { vlib_increment_combined_counter ( - im->combined_sw_if_counters + VNET_INTERFACE_COUNTER_TX, cpu_index, + im->combined_sw_if_counters + VNET_INTERFACE_COUNTER_TX, thread_index, stats_sw_if_index, stats_n_packets, stats_n_bytes); node->runtime_data[0] = stats_sw_if_index; } diff --git a/src/vnet/vxlan/decap.c b/src/vnet/vxlan/decap.c index 514b2c99..2acb1f6f 100644 --- a/src/vnet/vxlan/decap.c +++ b/src/vnet/vxlan/decap.c @@ -81,7 +81,7 @@ vxlan_input (vlib_main_t * vm, vxlan4_tunnel_key_t last_key4; vxlan6_tunnel_key_t last_key6; u32 pkts_decapsulated = 0; - u32 cpu_index = os_get_cpu_number(); + u32 thread_index = vlib_get_thread_index(); u32 stats_sw_if_index, stats_n_packets, stats_n_bytes; if (is_ip4) @@ -314,7 +314,7 @@ vxlan_input (vlib_main_t * vm, if (stats_n_packets) vlib_increment_combined_counter (im->combined_sw_if_counters + VNET_INTERFACE_COUNTER_RX, - cpu_index, stats_sw_if_index, + thread_index, stats_sw_if_index, stats_n_packets, stats_n_bytes); stats_n_packets = 1; stats_n_bytes = len0; @@ -468,7 +468,7 @@ vxlan_input (vlib_main_t * vm, if (stats_n_packets) vlib_increment_combined_counter (im->combined_sw_if_counters + VNET_INTERFACE_COUNTER_RX, - cpu_index, stats_sw_if_index, + thread_index, stats_sw_if_index, stats_n_packets, stats_n_bytes); stats_n_packets = 1; stats_n_bytes = len1; @@ -674,7 +674,7 @@ vxlan_input (vlib_main_t * vm, if (stats_n_packets) vlib_increment_combined_counter (im->combined_sw_if_counters + VNET_INTERFACE_COUNTER_RX, - cpu_index, stats_sw_if_index, + thread_index, stats_sw_if_index, stats_n_packets, stats_n_bytes); stats_n_packets = 1; stats_n_bytes = len0; @@ -711,7 +711,7 @@ vxlan_input (vlib_main_t * vm, { vlib_increment_combined_counter (im->combined_sw_if_counters + VNET_INTERFACE_COUNTER_RX, - cpu_index, stats_sw_if_index, stats_n_packets, stats_n_bytes); + thread_index, stats_sw_if_index, stats_n_packets, stats_n_bytes); node->runtime_data[0] = stats_sw_if_index; } diff --git a/src/vnet/vxlan/encap.c b/src/vnet/vxlan/encap.c index 5b63064a..4cfbbc23 100644 --- a/src/vnet/vxlan/encap.c +++ b/src/vnet/vxlan/encap.c @@ -77,7 +77,7 @@ vxlan_encap_inline (vlib_main_t * vm, vnet_interface_main_t * im = &vnm->interface_main; u32 pkts_encapsulated = 0; u16 old_l0 = 0, old_l1 = 0; - u32 cpu_index = os_get_cpu_number(); + u32 thread_index = vlib_get_thread_index(); u32 stats_sw_if_index, stats_n_packets, stats_n_bytes; u32 sw_if_index0 = 0, sw_if_index1 = 0; u32 next0 = 0, next1 = 0; @@ -301,7 +301,7 @@ vxlan_encap_inline (vlib_main_t * vm, if (stats_n_packets) vlib_increment_combined_counter (im->combined_sw_if_counters + VNET_INTERFACE_COUNTER_TX, - cpu_index, stats_sw_if_index, + thread_index, stats_sw_if_index, stats_n_packets, stats_n_bytes); stats_sw_if_index = sw_if_index0; stats_n_packets = 2; @@ -311,10 +311,10 @@ vxlan_encap_inline (vlib_main_t * vm, { vlib_increment_combined_counter (im->combined_sw_if_counters + VNET_INTERFACE_COUNTER_TX, - cpu_index, sw_if_index0, 1, len0); + thread_index, sw_if_index0, 1, len0); vlib_increment_combined_counter (im->combined_sw_if_counters + VNET_INTERFACE_COUNTER_TX, - cpu_index, sw_if_index1, 1, len1); + thread_index, sw_if_index1, 1, len1); } } @@ -464,7 +464,7 @@ vxlan_encap_inline (vlib_main_t * vm, if (stats_n_packets) vlib_increment_combined_counter (im->combined_sw_if_counters + VNET_INTERFACE_COUNTER_TX, - cpu_index, stats_sw_if_index, + thread_index, stats_sw_if_index, stats_n_packets, stats_n_bytes); stats_n_packets = 1; stats_n_bytes = len0; @@ -496,7 +496,7 @@ vxlan_encap_inline (vlib_main_t * vm, { vlib_increment_combined_counter (im->combined_sw_if_counters + VNET_INTERFACE_COUNTER_TX, - cpu_index, stats_sw_if_index, stats_n_packets, stats_n_bytes); + thread_index, stats_sw_if_index, stats_n_packets, stats_n_bytes); node->runtime_data[0] = stats_sw_if_index; } diff --git a/src/vpp/stats/stats.c b/src/vpp/stats/stats.c index 042d02e2..4309cd51 100644 --- a/src/vpp/stats/stats.c +++ b/src/vpp/stats/stats.c @@ -66,14 +66,14 @@ _(VNET_IP6_NBR_COUNTERS, vnet_ip6_nbr_counters) void dslock (stats_main_t * sm, int release_hint, int tag) { - u32 thread_id; + u32 thread_index; data_structure_lock_t *l = sm->data_structure_lock; if (PREDICT_FALSE (l == 0)) return; - thread_id = os_get_cpu_number (); - if (l->lock && l->thread_id == thread_id) + thread_index = vlib_get_thread_index (); + if (l->lock && l->thread_index == thread_index) { l->count++; return; @@ -85,7 +85,7 @@ dslock (stats_main_t * sm, int release_hint, int tag) while (__sync_lock_test_and_set (&l->lock, 1)) /* zzzz */ ; l->tag = tag; - l->thread_id = thread_id; + l->thread_index = thread_index; l->count = 1; } @@ -99,14 +99,14 @@ stats_dslock_with_hint (int hint, int tag) void dsunlock (stats_main_t * sm) { - u32 thread_id; + u32 thread_index; data_structure_lock_t *l = sm->data_structure_lock; if (PREDICT_FALSE (l == 0)) return; - thread_id = os_get_cpu_number (); - ASSERT (l->lock && l->thread_id == thread_id); + thread_index = vlib_get_thread_index (); + ASSERT (l->lock && l->thread_index == thread_index); l->count--; if (l->count == 0) { diff --git a/src/vpp/stats/stats.h b/src/vpp/stats/stats.h index 118115be..024dc78e 100644 --- a/src/vpp/stats/stats.h +++ b/src/vpp/stats/stats.h @@ -30,7 +30,7 @@ typedef struct { volatile u32 lock; volatile u32 release_hint; - u32 thread_id; + u32 thread_index; u32 count; int tag; } data_structure_lock_t; -- cgit 1.2.3-korg From 11b8dbf78af49d270a0e72abe7dea73eec30d85f Mon Sep 17 00:00:00 2001 From: Dave Barach Date: Mon, 24 Apr 2017 10:46:54 -0400 Subject: "autoreply" flag: autogenerate standard xxx_reply_t messages Change-Id: I72298aaae7d172082ece3a8edea4217c11b28d79 Signed-off-by: Dave Barach --- src/examples/sample-plugin/sample/sample.api | 10 +- src/plugins/acl/acl.api | 60 +--- src/plugins/dpdk/api/dpdk.api | 35 +- src/plugins/flowperpkt/flowperpkt.api | 23 +- .../export-vxlan-gpe/vxlan_gpe_ioam_export.api | 10 +- src/plugins/ioam/export/ioam_export.api | 10 +- src/plugins/ioam/ip6/ioam_cache.api | 10 +- src/plugins/ioam/lib-pot/pot.api | 34 +- src/plugins/ioam/lib-trace/trace.api | 26 +- src/plugins/ioam/lib-vxlan-gpe/ioam_vxlan_gpe.api | 82 +---- src/plugins/lb/lb.api | 21 +- src/plugins/memif/memif.api | 12 +- src/plugins/snat/snat.api | 88 +---- src/tools/vppapigen/gram.y | 3 +- src/tools/vppapigen/lex.c | 57 +++- src/tools/vppapigen/lex.h | 1 + src/tools/vppapigen/node.c | 5 + src/tools/vppapigen/node.h | 2 + src/vlibmemory/memclnt.api | 7 +- src/vlibmemory/memory_vlib.c | 8 +- src/vnet/bfd/bfd.api | 132 +------- src/vnet/classify/classify.api | 37 +-- src/vnet/cop/cop.api | 28 +- src/vnet/devices/af_packet/af_packet.api | 12 +- src/vnet/devices/netmap/netmap.api | 24 +- src/vnet/devices/virtio/vhost_user.api | 24 +- src/vnet/dhcp/dhcp.api | 38 +-- src/vnet/flow/flow.api | 32 +- src/vnet/interface.api | 108 +----- src/vnet/ip/ip.api | 108 +----- src/vnet/ipsec/ipsec.api | 224 ++----------- src/vnet/l2/l2.api | 96 +----- src/vnet/l2tp/l2tp.api | 28 +- src/vnet/lisp-cp/lisp.api | 164 +-------- src/vnet/lisp-cp/one.api | 185 +---------- src/vnet/lisp-gpe/lisp_gpe.api | 48 +-- src/vnet/map/map.api | 22 +- src/vnet/mpls/mpls.api | 26 +- src/vnet/session/session.api | 68 +--- src/vnet/span/span.api | 10 +- src/vnet/sr/sr.api | 60 +--- src/vnet/unix/tap.api | 12 +- src/vnet/vxlan/vxlan.api | 12 +- src/vpp/api/vpe.api | 367 ++------------------- 44 files changed, 271 insertions(+), 2098 deletions(-) (limited to 'src/vnet/classify') diff --git a/src/examples/sample-plugin/sample/sample.api b/src/examples/sample-plugin/sample/sample.api index f99cdb38..d565c0b1 100644 --- a/src/examples/sample-plugin/sample/sample.api +++ b/src/examples/sample-plugin/sample/sample.api @@ -16,7 +16,7 @@ /* Define a simple binary API to control the feature */ -define sample_macswap_enable_disable { +autoreply define sample_macswap_enable_disable { /* Client identifier, set from api_main.my_client_index */ u32 client_index; @@ -29,11 +29,3 @@ define sample_macswap_enable_disable { /* Interface handle */ u32 sw_if_index; }; - -define sample_macswap_enable_disable_reply { - /* From the request */ - u32 context; - - /* Return value, zero means all OK */ - i32 retval; -}; diff --git a/src/plugins/acl/acl.api b/src/plugins/acl/acl.api index d981338d..3b334113 100644 --- a/src/plugins/acl/acl.api +++ b/src/plugins/acl/acl.api @@ -161,24 +161,13 @@ define acl_add_replace_reply @param acl_index - ACL index to delete */ -manual_print define acl_del +autoreply manual_print define acl_del { u32 client_index; u32 context; u32 acl_index; }; -/** \brief Reply to delete the ACL - @param context - returned sender context, to match reply w/ request - @param retval 0 - no error -*/ - -define acl_del_reply -{ - u32 context; - i32 retval; -}; - /* acl_interface_add_del(_reply) to be deprecated in lieu of acl_interface_set_acl_list */ /** \brief Use acl_interface_set_acl_list instead Append/remove an ACL index to/from the list of ACLs checked for an interface @@ -190,7 +179,7 @@ define acl_del_reply @param acl_index - index of ACL for the operation */ -manual_print define acl_interface_add_del +autoreply manual_print define acl_interface_add_del { u32 client_index; u32 context; @@ -204,17 +193,6 @@ manual_print define acl_interface_add_del u32 acl_index; }; -/** \brief Reply to alter the ACL list - @param context - returned sender context, to match reply w/ request - @param retval 0 - no error -*/ - -define acl_interface_add_del_reply -{ - u32 context; - i32 retval; -}; - /** \brief Set the vector of input/output ACLs checked for an interface @param client_index - opaque cookie to identify the sender @param context - sender context, to match reply w/ request @@ -224,7 +202,7 @@ define acl_interface_add_del_reply @param acls - vector of ACL indices */ -manual_print define acl_interface_set_acl_list +autoreply manual_print define acl_interface_set_acl_list { u32 client_index; u32 context; @@ -239,12 +217,6 @@ manual_print define acl_interface_set_acl_list @param retval 0 - no error */ -define acl_interface_set_acl_list_reply -{ - u32 context; - i32 retval; -}; - /** \brief Dump the specific ACL contents or all of the ACLs' contents @param client_index - opaque cookie to identify the sender @param context - sender context, to match reply w/ request @@ -341,24 +313,13 @@ define macip_acl_add_reply @param acl_index - MACIP ACL index to delete */ -manual_print define macip_acl_del +autoreply manual_print define macip_acl_del { u32 client_index; u32 context; u32 acl_index; }; -/** \brief Reply to delete the MACIP ACL - @param context - returned sender context, to match reply w/ request - @param retval 0 - no error -*/ - -define macip_acl_del_reply -{ - u32 context; - i32 retval; -}; - /** \brief Add or delete a MACIP ACL to/from interface @param client_index - opaque cookie to identify the sender @param context - sender context, to match reply w/ request @@ -367,7 +328,7 @@ define macip_acl_del_reply @param acl_index - MACIP ACL index */ -manual_print define macip_acl_interface_add_del +autoreply manual_print define macip_acl_interface_add_del { u32 client_index; u32 context; @@ -377,17 +338,6 @@ manual_print define macip_acl_interface_add_del u32 acl_index; }; -/** \brief Reply to apply/unapply the MACIP ACL - @param context - returned sender context, to match reply w/ request - @param retval 0 - no error -*/ - -define macip_acl_interface_add_del_reply -{ - u32 context; - i32 retval; -}; - /** \brief Dump one or all defined MACIP ACLs @param client_index - opaque cookie to identify the sender @param context - sender context, to match reply w/ request diff --git a/src/plugins/dpdk/api/dpdk.api b/src/plugins/dpdk/api/dpdk.api index 21215d45..d43f8a36 100644 --- a/src/plugins/dpdk/api/dpdk.api +++ b/src/plugins/dpdk/api/dpdk.api @@ -21,7 +21,7 @@ @param pipe - pipe ID within its subport @param profile - pipe profile ID */ -define sw_interface_set_dpdk_hqos_pipe { +autoreply define sw_interface_set_dpdk_hqos_pipe { u32 client_index; u32 context; u32 sw_if_index; @@ -30,15 +30,6 @@ define sw_interface_set_dpdk_hqos_pipe { u32 profile; }; -/** \brief DPDK interface HQoS pipe profile set reply - @param context - sender context, to match reply w/ request - @param retval - request return code -*/ -define sw_interface_set_dpdk_hqos_pipe_reply { - u32 context; - i32 retval; -}; - /** \brief DPDK interface HQoS subport parameters set request @param client_index - opaque cookie to identify the sender @param context - sender context, to match reply w/ request @@ -49,7 +40,7 @@ define sw_interface_set_dpdk_hqos_pipe_reply { @param tc_rate - subport traffic class 0 .. 3 rates (measured in bytes/second) @param tc_period - enforcement period for rates (measured in milliseconds) */ -define sw_interface_set_dpdk_hqos_subport { +autoreply define sw_interface_set_dpdk_hqos_subport { u32 client_index; u32 context; u32 sw_if_index; @@ -60,15 +51,6 @@ define sw_interface_set_dpdk_hqos_subport { u32 tc_period; }; -/** \brief DPDK interface HQoS subport parameters set reply - @param context - sender context, to match reply w/ request - @param retval - request return code -*/ -define sw_interface_set_dpdk_hqos_subport_reply { - u32 context; - i32 retval; -}; - /** \brief DPDK interface HQoS tctbl entry set request @param client_index - opaque cookie to identify the sender @param context - sender context, to match reply w/ request @@ -77,7 +59,7 @@ define sw_interface_set_dpdk_hqos_subport_reply { @param tc - traffic class (0 .. 3) @param queue - traffic class queue (0 .. 3) */ -define sw_interface_set_dpdk_hqos_tctbl { +autoreply define sw_interface_set_dpdk_hqos_tctbl { u32 client_index; u32 context; u32 sw_if_index; @@ -86,18 +68,9 @@ define sw_interface_set_dpdk_hqos_tctbl { u32 queue; }; -/** \brief DPDK interface HQoS tctbl entry set reply - @param context - sender context, to match reply w/ request - @param retval - request return code -*/ -define sw_interface_set_dpdk_hqos_tctbl_reply { - u32 context; - i32 retval; -}; - /* * Local Variables: * eval: (c-set-style "gnu") * End: */ - \ No newline at end of file + diff --git a/src/plugins/flowperpkt/flowperpkt.api b/src/plugins/flowperpkt/flowperpkt.api index 1cf62c54..3ff92dca 100644 --- a/src/plugins/flowperpkt/flowperpkt.api +++ b/src/plugins/flowperpkt/flowperpkt.api @@ -12,7 +12,7 @@ @param is_ipv6 - if non-zero the address is ipv6, else ipv4 @param sw_if_index - index of the interface */ -manual_print define flowperpkt_tx_interface_add_del +autoreply manual_print define flowperpkt_tx_interface_add_del { /* Client identifier, set from api_main.my_client_index */ u32 client_index; @@ -28,20 +28,7 @@ manual_print define flowperpkt_tx_interface_add_del u32 sw_if_index; }; -/** \brief Reply to enable/disable per-packet IPFIX recording messages - @param context - returned sender context, to match reply w/ request - @param retval - return code -*/ -define flowperpkt_tx_interface_add_del_reply -{ - /* From the request */ - u32 context; - - /* Return value, zero means all OK */ - i32 retval; -}; - -define flowperpkt_params +autoreply define flowperpkt_params { u32 client_index; u32 context; @@ -51,9 +38,3 @@ define flowperpkt_params u32 active_timer; /* ~0 is off, 0 is default */ u32 passive_timer; /* ~0 is off, 0 is default */ }; - -define flowperpkt_params_reply -{ - u32 context; - i32 retval; -}; diff --git a/src/plugins/ioam/export-vxlan-gpe/vxlan_gpe_ioam_export.api b/src/plugins/ioam/export-vxlan-gpe/vxlan_gpe_ioam_export.api index 7b17c3f7..caa97e6e 100644 --- a/src/plugins/ioam/export-vxlan-gpe/vxlan_gpe_ioam_export.api +++ b/src/plugins/ioam/export-vxlan-gpe/vxlan_gpe_ioam_export.api @@ -16,7 +16,7 @@ /* Define a simple binary API to control the feature */ -define vxlan_gpe_ioam_export_enable_disable { +autoreply define vxlan_gpe_ioam_export_enable_disable { /* Client identifier, set from api_main.my_client_index */ u32 client_index; @@ -32,11 +32,3 @@ define vxlan_gpe_ioam_export_enable_disable { /* Src ip address */ }; - -define vxlan_gpe_ioam_export_enable_disable_reply { - /* From the request */ - u32 context; - - /* Return value, zero means all OK */ - i32 retval; -}; \ No newline at end of file diff --git a/src/plugins/ioam/export/ioam_export.api b/src/plugins/ioam/export/ioam_export.api index f22d9fc8..bb830561 100644 --- a/src/plugins/ioam/export/ioam_export.api +++ b/src/plugins/ioam/export/ioam_export.api @@ -16,7 +16,7 @@ /* Define a simple binary API to control the feature */ -define ioam_export_ip6_enable_disable { +autoreply define ioam_export_ip6_enable_disable { /* Client identifier, set from api_main.my_client_index */ u32 client_index; @@ -32,11 +32,3 @@ define ioam_export_ip6_enable_disable { /* Src ip address */ }; - -define ioam_export_ip6_enable_disable_reply { - /* From the request */ - u32 context; - - /* Return value, zero means all OK */ - i32 retval; -}; diff --git a/src/plugins/ioam/ip6/ioam_cache.api b/src/plugins/ioam/ip6/ioam_cache.api index de50d57d..dd9c0186 100644 --- a/src/plugins/ioam/ip6/ioam_cache.api +++ b/src/plugins/ioam/ip6/ioam_cache.api @@ -16,7 +16,7 @@ /* API to control ioam caching */ -define ioam_cache_ip6_enable_disable { +autoreply define ioam_cache_ip6_enable_disable { /* Client identifier, set from api_main.my_client_index */ u32 client_index; @@ -27,11 +27,3 @@ define ioam_cache_ip6_enable_disable { u8 is_disable; }; - -define ioam_cache_ip6_enable_disable_reply { - /* From the request */ - u32 context; - - /* Return value, zero means all OK */ - i32 retval; -}; diff --git a/src/plugins/ioam/lib-pot/pot.api b/src/plugins/ioam/lib-pot/pot.api index fa2fc126..c377cde0 100644 --- a/src/plugins/ioam/lib-pot/pot.api +++ b/src/plugins/ioam/lib-pot/pot.api @@ -27,7 +27,7 @@ @param list_name_len - length of the name of this profile list @param list_name - name of this profile list */ -define pot_profile_add { +autoreply define pot_profile_add { u32 client_index; u32 context; u8 id; @@ -42,22 +42,12 @@ define pot_profile_add { u8 list_name[0]; }; -/** \brief Proof of Transit profile add / del response - @param context - sender context, to match reply w/ request - @param retval - return value for request -*/ -define pot_profile_add_reply { - u32 context; - i32 retval; -}; - - /** \brief Proof of Transit(POT): Activate POT profile in the list @param id - id of the profile @param list_name_len - length of the name of this profile list @param list_name - name of this profile list */ -define pot_profile_activate { +autoreply define pot_profile_activate { u32 client_index; u32 context; u8 id; @@ -65,37 +55,19 @@ define pot_profile_activate { u8 list_name[0]; }; -/** \brief Proof of Transit profile activate response - @param context - sender context, to match reply w/ request - @param retval - return value for request -*/ -define pot_profile_activate_reply { - u32 context; - i32 retval; -}; - /** \brief Delete POT Profile @param client_index - opaque cookie to identify the sender @param context - sender context, to match reply w/ request @param list_name_len - length of the name of the profile list @param list_name - name of profile list to delete */ -define pot_profile_del { +autoreply define pot_profile_del { u32 client_index; u32 context; u8 list_name_len; u8 list_name[0]; }; -/** \brief Proof of Transit profile add / del response - @param context - sender context, to match reply w/ request - @param retval - return value for request -*/ -define pot_profile_del_reply { - u32 context; - i32 retval; -}; - /** \brief Show POT Profiles @param client_index - opaque cookie to identify the sender @param context - sender context, to match reply w/ request diff --git a/src/plugins/ioam/lib-trace/trace.api b/src/plugins/ioam/lib-trace/trace.api index cb958325..2f45c6e2 100644 --- a/src/plugins/ioam/lib-trace/trace.api +++ b/src/plugins/ioam/lib-trace/trace.api @@ -22,7 +22,7 @@ @param trace_tsp- Timestamp resolution @param app_data - Application specific opaque */ -define trace_profile_add { +autoreply define trace_profile_add { u32 client_index; u32 context; u8 trace_type; @@ -32,37 +32,15 @@ define trace_profile_add { u32 app_data; }; -/** \brief Trace profile add / del response - @param context - sender context, to match reply w/ request - @param retval - return value for request -*/ -define trace_profile_add_reply { - u32 context; - i32 retval; -}; - - - /** \brief Delete trace Profile @param client_index - opaque cookie to identify the sender @param context - sender context, to match reply w/ request */ -define trace_profile_del { +autoreply define trace_profile_del { u32 client_index; u32 context; }; -/** \brief Trace profile add / del response - @param context - sender context, to match reply w/ request - @param retval - return value for request -*/ -define trace_profile_del_reply { - u32 context; - i32 retval; -}; - - - /** \brief Show trace Profile @param client_index - opaque cookie to identify the sender @param context - sender context, to match reply w/ request diff --git a/src/plugins/ioam/lib-vxlan-gpe/ioam_vxlan_gpe.api b/src/plugins/ioam/lib-vxlan-gpe/ioam_vxlan_gpe.api index 056529a4..a6761f07 100644 --- a/src/plugins/ioam/lib-vxlan-gpe/ioam_vxlan_gpe.api +++ b/src/plugins/ioam/lib-vxlan-gpe/ioam_vxlan_gpe.api @@ -24,7 +24,7 @@ @param trace_enable - iOAM Trace enabled or not flag */ -define vxlan_gpe_ioam_enable { +autoreply define vxlan_gpe_ioam_enable { u32 client_index; u32 context; u16 id; @@ -33,38 +33,18 @@ define vxlan_gpe_ioam_enable { u8 trace_enable; }; -/** \brief iOAM Over VxLAN-GPE - Set iOAM transport for VXLAN-GPE reply - @param context - sender context, to match reply w/ request - @param retval - return value for request -*/ -define vxlan_gpe_ioam_enable_reply { - u32 context; - i32 retval; -}; - - /** \brief iOAM for VxLAN-GPE disable @param client_index - opaque cookie to identify the sender @param context - sender context, to match reply w/ request @param id - profile id */ -define vxlan_gpe_ioam_disable +autoreply define vxlan_gpe_ioam_disable { u32 client_index; u32 context; u16 id; }; -/** \brief vxlan_gpe_ioam disable response - @param context - sender context, to match reply w/ request - @param retval - return value for request -*/ -define vxlan_gpe_ioam_disable_reply -{ - u32 context; - i32 retval; -}; - /** \brief Enable iOAM for a VNI (VXLAN-GPE) @param client_index - opaque cookie to identify the sender @param context - sender context, to match reply w/ request @@ -73,7 +53,7 @@ define vxlan_gpe_ioam_disable_reply @param remote - IPv4/6 Address of the remote VTEP */ -define vxlan_gpe_ioam_vni_enable { +autoreply define vxlan_gpe_ioam_vni_enable { u32 client_index; u32 context; u32 vni; @@ -82,18 +62,6 @@ define vxlan_gpe_ioam_vni_enable { u8 is_ipv6; }; -/** \brief Reply to enable iOAM for a VNI (VXLAN-GPE) - @param client_index - opaque cookie to identify the sender - @param context - sender context, to match reply w/ request - @param retval - return value for request - -*/ -define vxlan_gpe_ioam_vni_enable_reply { - u32 client_index; - u32 context; - i32 retval; -}; - /** \brief Disable iOAM for a VNI (VXLAN-GPE) @param client_index - opaque cookie to identify the sender @param context - sender context, to match reply w/ request @@ -102,7 +70,7 @@ define vxlan_gpe_ioam_vni_enable_reply { @param remote - IPv4/6 Address of the remote VTEP */ -define vxlan_gpe_ioam_vni_disable { +autoreply define vxlan_gpe_ioam_vni_disable { u32 client_index; u32 context; u32 vni; @@ -111,19 +79,6 @@ define vxlan_gpe_ioam_vni_disable { u8 is_ipv6; }; -/** \brief Reply to disable iOAM for a VNI (VXLAN-GPE) - @param client_index - opaque cookie to identify the sender - @param context - sender context, to match reply w/ request - @param retval - return value for request - -*/ -define vxlan_gpe_ioam_vni_disable_reply { - u32 client_index; - u32 context; - i32 retval; -}; - - /** \brief Enable iOAM for a VXLAN-GPE transit @param client_index - opaque cookie to identify the sender @param context - sender context, to match reply w/ request @@ -131,7 +86,7 @@ define vxlan_gpe_ioam_vni_disable_reply { @param outer_fib_index- FIB index */ -define vxlan_gpe_ioam_transit_enable { +autoreply define vxlan_gpe_ioam_transit_enable { u32 client_index; u32 context; u32 outer_fib_index; @@ -139,18 +94,6 @@ define vxlan_gpe_ioam_transit_enable { u8 is_ipv6; }; -/** \brief Reply to enable iOAM for VXLAN-GPE transit - @param client_index - opaque cookie to identify the sender - @param context - sender context, to match reply w/ request - @param retval - return value for request - -*/ -define vxlan_gpe_ioam_transit_enable_reply { - u32 client_index; - u32 context; - i32 retval; -}; - /** \brief Disable iOAM for VXLAN-GPE transit @param client_index - opaque cookie to identify the sender @param context - sender context, to match reply w/ request @@ -158,7 +101,7 @@ define vxlan_gpe_ioam_transit_enable_reply { @param outer_fib_index- FIB index */ -define vxlan_gpe_ioam_transit_disable { +autoreply define vxlan_gpe_ioam_transit_disable { u32 client_index; u32 context; u32 outer_fib_index; @@ -166,16 +109,3 @@ define vxlan_gpe_ioam_transit_disable { u8 is_ipv6; }; -/** \brief Reply to disable iOAM for VXLAN-GPE transit - @param client_index - opaque cookie to identify the sender - @param context - sender context, to match reply w/ request - @param retval - return value for request - -*/ -define vxlan_gpe_ioam_transit_disable_reply { - u32 client_index; - u32 context; - i32 retval; -}; - - diff --git a/src/plugins/lb/lb.api b/src/plugins/lb/lb.api index 39ee3c8f..32cc669b 100644 --- a/src/plugins/lb/lb.api +++ b/src/plugins/lb/lb.api @@ -8,7 +8,7 @@ @param flow_timeout - Time in seconds after which, if no packet is received for a given flow, the flow is removed from the established flow table. */ -define lb_conf +autoreply define lb_conf { u32 client_index; u32 context; @@ -18,11 +18,6 @@ define lb_conf u32 flow_timeout; }; -define lb_conf_reply { - u32 context; - i32 retval; -}; - /** \brief Add a virtual address (or prefix) @param client_index - opaque cookie to identify the sender @param context - sender context, to match reply w/ request @@ -33,7 +28,7 @@ define lb_conf_reply { for this VIP (must be power of 2). @param is_del - The VIP should be removed. */ -define lb_add_del_vip { +autoreply define lb_add_del_vip { u32 client_index; u32 context; u8 ip_prefix[16]; @@ -43,11 +38,6 @@ define lb_add_del_vip { u8 is_del; }; -define lb_add_del_vip_reply { - u32 context; - i32 retval; -}; - /** \brief Add an application server for a given VIP @param client_index - opaque cookie to identify the sender @param context - sender context, to match reply w/ request @@ -56,7 +46,7 @@ define lb_add_del_vip_reply { @param as_address - The application server address (IPv4 in lower order 32 bits). @param is_del - The AS should be removed. */ -define lb_add_del_as { +autoreply define lb_add_del_as { u32 client_index; u32 context; u8 vip_ip_prefix[16]; @@ -64,8 +54,3 @@ define lb_add_del_as { u8 as_address[16]; u8 is_del; }; - -define lb_add_del_as_reply { - u32 context; - i32 retval; -}; diff --git a/src/plugins/memif/memif.api b/src/plugins/memif/memif.api index 6f946421..95e016c3 100644 --- a/src/plugins/memif/memif.api +++ b/src/plugins/memif/memif.api @@ -57,7 +57,7 @@ define memif_create_reply @param context - sender context, to match reply w/ request @param sw_if_index - software index of the interface to delete */ -define memif_delete +autoreply define memif_delete { u32 client_index; u32 context; @@ -65,16 +65,6 @@ define memif_delete u32 sw_if_index; }; -/** \brief Delete host-interface response - @param context - sender context, to match reply w/ request - @param retval - return value for request -*/ -define memif_delete_reply -{ - u32 context; - i32 retval; -}; - /** \brief Memory interface details structure @param context - sender context, to match reply w/ request (memif_dump) @param sw_if_index - index of the interface diff --git a/src/plugins/snat/snat.api b/src/plugins/snat/snat.api index 9689f5f9..573b6753 100644 --- a/src/plugins/snat/snat.api +++ b/src/plugins/snat/snat.api @@ -29,7 +29,7 @@ @param vrf_id - VRF id of tenant, ~0 means independent of VRF @param is_add - 1 if add, 0 if delete */ -define snat_add_address_range { +autoreply define snat_add_address_range { u32 client_index; u32 context; u8 is_ip4; @@ -39,15 +39,6 @@ define snat_add_address_range { u8 is_add; }; -/** \brief Add S-NAT address range reply - @param context - sender context, to match reply w/ request - @param retval - return code -*/ -define snat_add_address_range_reply { - u32 context; - i32 retval; -}; - /** \brief Dump S-NAT addresses @param client_index - opaque cookie to identify the sender @param context - sender context, to match reply w/ request @@ -77,7 +68,7 @@ define snat_address_details { @param is_inside - 1 if inside, 0 if outside @param sw_if_index - software index of the interface */ -define snat_interface_add_del_feature { +autoreply define snat_interface_add_del_feature { u32 client_index; u32 context; u8 is_add; @@ -85,15 +76,6 @@ define snat_interface_add_del_feature { u32 sw_if_index; }; -/** \brief Enable/disable S-NAT feature on the interface reply - @param context - sender context, to match reply w/ request - @param retval - return code -*/ -define snat_interface_add_del_feature_reply { - u32 context; - i32 retval; -}; - /** \brief Dump interfaces with S-NAT feature @param client_index - opaque cookie to identify the sender @param context - sender context, to match reply w/ request @@ -130,7 +112,7 @@ define snat_interface_details { used) @param vfr_id - VRF ID */ -define snat_add_static_mapping { +autoreply define snat_add_static_mapping { u32 client_index; u32 context; u8 is_add; @@ -145,15 +127,6 @@ define snat_add_static_mapping { u32 vrf_id; }; -/** \brief Add/delete S-NAT static mapping reply - @param context - sender context, to match reply w/ request - @param retval - return code -*/ -define snat_add_static_mapping_reply { - u32 context; - i32 retval; -}; - /** \brief Dump S-NAT static mappings @param client_index - opaque cookie to identify the sender @param context - sender context, to match reply w/ request @@ -257,21 +230,12 @@ define snat_show_config_reply @param context - sender context, to match reply w/ request @param worker_mask - S-NAT workers mask */ -define snat_set_workers { +autoreply define snat_set_workers { u32 client_index; u32 context; u64 worker_mask; }; -/** \brief Set S-NAT workers reply - @param context - sender context, to match reply w/ request - @param retval - return code -*/ -define snat_set_workers_reply { - u32 context; - i32 retval; -}; - /** \brief Dump S-NAT workers @param client_index - opaque cookie to identify the sender @param context - sender context, to match reply w/ request @@ -300,7 +264,7 @@ define snat_worker_details { @param is_add - 1 if add, 0 if delete @param sw_if_index - software index of the interface */ -define snat_add_del_interface_addr { +autoreply define snat_add_del_interface_addr { u32 client_index; u32 context; u8 is_add; @@ -308,15 +272,6 @@ define snat_add_del_interface_addr { u32 sw_if_index; }; -/** \brief Add/delete S-NAT pool address from specific interfce reply - @param context - sender context, to match reply w/ request - @param retval - return code -*/ -define snat_add_del_interface_addr_reply { - u32 context; - i32 retval; -}; - /** \brief Dump S-NAT pool addresses interfaces @param client_index - opaque cookie to identify the sender @param context - sender context, to match reply w/ request @@ -342,7 +297,7 @@ define snat_interface_addr_details { @param src_port - source port number @param enable - 1 if enable, 0 if disable */ -define snat_ipfix_enable_disable { +autoreply define snat_ipfix_enable_disable { u32 client_index; u32 context; u32 domain_id; @@ -350,15 +305,6 @@ define snat_ipfix_enable_disable { u8 enable; }; -/** \brief Enable/disable S-NAT IPFIX logging reply - @param context - sender context, to match reply w/ request - @param retval - return code -*/ -define snat_ipfix_enable_disable_reply { - u32 context; - i32 retval; -}; - /** \brief Dump S-NAT users @param client_index - opaque cookie to identify the sender @param context - sender context, to match reply w/ request @@ -437,7 +383,7 @@ define snat_user_session_details { @param out_addr - outside IP address @param out_addr - outside IP address prefix length */ -define snat_add_det_map { +autoreply define snat_add_det_map { u32 client_index; u32 context; u8 is_add; @@ -449,15 +395,6 @@ define snat_add_det_map { u8 out_plen; }; -/** \brief Add/delete S-NAT deterministic mapping reply - @param context - sender context, to match reply w/ request - @param retval - return code -*/ -define snat_add_det_map_reply { - u32 context; - i32 retval; -}; - /** \brief Get outside address and port range from inside address @param client_index - opaque cookie to identify the sender @param context - sender context, to match reply w/ request @@ -556,7 +493,7 @@ define snat_det_map_details { @param tcp_transitory - TCP transitory timeout (default 240sec) @param icmp - ICMP timeout (default 60sec) */ -define snat_det_set_timeouts { +autoreply define snat_det_set_timeouts { u32 client_index; u32 context; u32 udp; @@ -565,15 +502,6 @@ define snat_det_set_timeouts { u32 icmp; }; -/** \brief Set values of timeouts for deterministic NAT reply - @param context - sender context, to match reply w/ request - @param retval - return code -*/ -define snat_det_set_timeouts_reply { - u32 context; - i32 retval; -}; - /** \brief Get values of timeouts for deterministic NAT (seconds) @param client_index - opaque cookie to identify the sender @param context - sender context, to match reply w/ request diff --git a/src/tools/vppapigen/gram.y b/src/tools/vppapigen/gram.y index de26af8d..9cea6023 100644 --- a/src/tools/vppapigen/gram.y +++ b/src/tools/vppapigen/gram.y @@ -38,7 +38,7 @@ void generate (YYSTYPE); %token NAME RPAR LPAR SEMI LBRACK RBRACK NUMBER PRIMTYPE BARF %token TPACKED DEFINE LCURLY RCURLY STRING UNION %token HELPER_STRING COMMA -%token NOVERSION MANUAL_PRINT MANUAL_ENDIAN TYPEONLY DONT_TRACE +%token NOVERSION MANUAL_PRINT MANUAL_ENDIAN TYPEONLY DONT_TRACE AUTOREPLY %% @@ -64,6 +64,7 @@ flag: | MANUAL_ENDIAN {$$ = $1;} | DONT_TRACE {$$ = $1;} | TYPEONLY {$$ = $1;} + | AUTOREPLY {$$ = $1;} ; defn: DEFINE NAME LCURLY defbody RCURLY SEMI diff --git a/src/tools/vppapigen/lex.c b/src/tools/vppapigen/lex.c index 733942ad..e6358143 100644 --- a/src/tools/vppapigen/lex.c +++ b/src/tools/vppapigen/lex.c @@ -27,6 +27,9 @@ #include "lex.h" #include "node.h" #include "tools/vppapigen/gram.h" +#include +#include +#include FILE *ifp, *ofp, *pythonfp, *jsonfp; char *vlib_app_name = "vpp"; @@ -38,6 +41,9 @@ int current_filename_allocated; unsigned long input_crc; unsigned long message_crc; int yydebug; +char *push_input_fifo; +char saved_ungetc_char; +char have_ungetc_char; /* * lexer variable definitions @@ -469,9 +475,50 @@ static char namebuf [MAXNAME]; static inline char getc_char (FILE *ifp) { + char rv; + + if (have_ungetc_char) { + have_ungetc_char = 0; + return saved_ungetc_char; + } + + if (clib_fifo_elts (push_input_fifo)) { + clib_fifo_sub1(push_input_fifo, rv); + return (rv & 0x7f); + } return ((char)(getc(ifp) & 0x7f)); } +u32 fe (char *fifo) +{ + return clib_fifo_elts (fifo); +} + +static inline void +ungetc_char (char c, FILE *ifp) +{ + saved_ungetc_char = c; + have_ungetc_char = 1; +} + +void autoreply (void *np_arg) +{ + static u8 *s; + node_t *np = (node_t *)np_arg; + int i; + + vec_reset_length (s); + + s = format (0, " define %s_reply\n", (char *)(np->data[0])); + s = format (s, "{\n"); + s = format (s, " u32 context;\n"); + s = format (s, " i32 retval;\n"); + s = format (s, "};\n"); + + for (i = 0; i < vec_len (s); i++) + clib_fifo_add1 (push_input_fifo, s[i]); +} + /* * yylex (well, yylex_1: The real yylex below does crc-hackery) */ @@ -595,7 +642,7 @@ static int yylex_1 (void) return (EOF); if (!isalnum (c) && c != '_') { - ungetc (c, ifp); + ungetc_char (c, ifp); namebuf [nameidx] = 0; the_lexer_state = START_STATE; return (name_check (namebuf, &yylval)); @@ -616,7 +663,7 @@ static int yylex_1 (void) return (EOF); if (!isdigit (c)) { - ungetc (c, ifp); + ungetc_char (c, ifp); namebuf [nameidx] = 0; the_lexer_state = START_STATE; yylval = (void *) atol(namebuf); @@ -889,6 +936,7 @@ int yylex (void) case MANUAL_ENDIAN: code = 276; break; case TYPEONLY: code = 278; break; case DONT_TRACE: code = 279; break; + case AUTOREPLY: code = 280; break; case EOF: code = ~0; break; /* hysterical compatibility */ @@ -929,6 +977,7 @@ static struct keytab { } keytab [] = /* Keep the table sorted, binary search used below! */ { + {"autoreply", NODE_AUTOREPLY}, {"define", NODE_DEFINE}, {"dont_trace", NODE_DONT_TRACE}, {"f64", NODE_F64}, @@ -1005,6 +1054,10 @@ static int name_check (const char *s, YYSTYPE *token_value) *token_value = (YYSTYPE) NODE_FLAG_DONT_TRACE; return(DONT_TRACE); + case NODE_AUTOREPLY: + *token_value = (YYSTYPE) NODE_FLAG_AUTOREPLY; + return(AUTOREPLY); + case NODE_NOVERSION: return(NOVERSION); diff --git a/src/tools/vppapigen/lex.h b/src/tools/vppapigen/lex.h index a0fdc735..275cf685 100644 --- a/src/tools/vppapigen/lex.h +++ b/src/tools/vppapigen/lex.h @@ -24,6 +24,7 @@ extern int yylex (void); extern void yyerror (char *); extern int yyparse (void); +extern void autoreply (void *); #ifndef YYSTYPE #define YYSTYPE void * diff --git a/src/tools/vppapigen/node.c b/src/tools/vppapigen/node.c index 359ac9c9..9f234037 100644 --- a/src/tools/vppapigen/node.c +++ b/src/tools/vppapigen/node.c @@ -1050,6 +1050,11 @@ YYSTYPE set_flags(YYSTYPE a1, YYSTYPE a2) flags = (int)(uword) a1; np->flags |= flags; + + /* Generate a foo_reply_t right here */ + if (flags & NODE_FLAG_AUTOREPLY) + autoreply(np); + return (a2); } /* diff --git a/src/tools/vppapigen/node.h b/src/tools/vppapigen/node.h index 297d6036..65bd5d10 100644 --- a/src/tools/vppapigen/node.h +++ b/src/tools/vppapigen/node.h @@ -53,6 +53,7 @@ enum node_subclass { /* WARNING: indices must match the vft... */ NODE_MANUAL_PRINT, NODE_MANUAL_ENDIAN, NODE_DONT_TRACE, + NODE_AUTOREPLY, }; enum passid { @@ -84,6 +85,7 @@ typedef struct node_ { #define NODE_FLAG_MANUAL_ENDIAN (1<<1) #define NODE_FLAG_TYPEONLY (1<<3) #define NODE_FLAG_DONT_TRACE (1<<4) +#define NODE_FLAG_AUTOREPLY (1<<5) typedef struct node_vft_ { void (*print)(struct node_ *); diff --git a/src/vlibmemory/memclnt.api b/src/vlibmemory/memclnt.api index c38b483c..32e51407 100644 --- a/src/vlibmemory/memclnt.api +++ b/src/vlibmemory/memclnt.api @@ -72,7 +72,7 @@ define memclnt_read_timeout { /* * RPC */ -define rpc_call { +autoreply define rpc_call { u32 client_index; u32 context; u64 function; @@ -82,11 +82,6 @@ define rpc_call { u8 data[0]; }; -define rpc_reply { - i32 retval; - u32 context; -}; - /* * Lookup message-ID base by name */ diff --git a/src/vlibmemory/memory_vlib.c b/src/vlibmemory/memory_vlib.c index 7a536ee8..43574dea 100644 --- a/src/vlibmemory/memory_vlib.c +++ b/src/vlibmemory/memory_vlib.c @@ -1275,7 +1275,7 @@ VLIB_CLI_COMMAND (cli_show_api_plugin_command, static) = { static void vl_api_rpc_call_t_handler (vl_api_rpc_call_t * mp) { - vl_api_rpc_reply_t *rmp; + vl_api_rpc_call_reply_t *rmp; int (*fp) (void *); i32 rv = 0; vlib_main_t *vm = vlib_get_main (); @@ -1305,7 +1305,7 @@ vl_api_rpc_call_t_handler (vl_api_rpc_call_t * mp) if (q) { rmp = vl_msg_api_alloc_as_if_client (sizeof (*rmp)); - rmp->_vl_msg_id = ntohs (VL_API_RPC_REPLY); + rmp->_vl_msg_id = ntohs (VL_API_RPC_CALL_REPLY); rmp->context = mp->context; rmp->retval = rv; vl_msg_api_send_shmem (q, (u8 *) & rmp); @@ -1318,7 +1318,7 @@ vl_api_rpc_call_t_handler (vl_api_rpc_call_t * mp) } static void -vl_api_rpc_reply_t_handler (vl_api_rpc_reply_t * mp) +vl_api_rpc_call_reply_t_handler (vl_api_rpc_call_reply_t * mp) { clib_warning ("unimplemented"); } @@ -1415,7 +1415,7 @@ vl_api_trace_plugin_msg_ids_t_handler (vl_api_trace_plugin_msg_ids_t * mp) #define foreach_rpc_api_msg \ _(RPC_CALL,rpc_call) \ -_(RPC_REPLY,rpc_reply) +_(RPC_CALL_REPLY,rpc_call_reply) #define foreach_plugin_trace_msg \ _(TRACE_PLUGIN_MSG_IDS,trace_plugin_msg_ids) diff --git a/src/vnet/bfd/bfd.api b/src/vnet/bfd/bfd.api index 2cdcfad3..7bcaa4c3 100644 --- a/src/vnet/bfd/bfd.api +++ b/src/vnet/bfd/bfd.api @@ -18,43 +18,23 @@ @param context - sender context, to match reply w/ request @param sw_if_index - interface to use as echo source */ -define bfd_udp_set_echo_source +autoreply define bfd_udp_set_echo_source { u32 client_index; u32 context; u32 sw_if_index; }; -/** \brief Set BFD feature response - @param context - sender context, to match reply w/ request - @param retval - return code for the request -*/ -define bfd_udp_set_echo_source_reply -{ - u32 context; - i32 retval; -}; - /** \brief Delete BFD echo source @param client_index - opaque cookie to identify the sender @param context - sender context, to match reply w/ request */ -define bfd_udp_del_echo_source +autoreply define bfd_udp_del_echo_source { u32 client_index; u32 context; }; -/** \brief Delete BFD echo source response - @param context - sender context, to match reply w/ request - @param retval - return code for the request -*/ -define bfd_udp_del_echo_source_reply -{ - u32 context; - i32 retval; -}; - /** \brief Add UDP BFD session on interface @param client_index - opaque cookie to identify the sender @param context - sender context, to match reply w/ request @@ -69,7 +49,7 @@ define bfd_udp_del_echo_source_reply @param bfd_key_id - key id sent out in BFD packets (if is_authenticated) @param conf_key_id - id of already configured key (if is_authenticated) */ -define bfd_udp_add +autoreply define bfd_udp_add { u32 client_index; u32 context; @@ -85,16 +65,6 @@ define bfd_udp_add u32 conf_key_id; }; -/** \brief Add UDP BFD session response - @param context - sender context, to match reply w/ request - @param retval - return code for the request -*/ -define bfd_udp_add_reply -{ - u32 context; - i32 retval; -}; - /** \brief Modify UDP BFD session on interface @param client_index - opaque cookie to identify the sender @param context - sender context, to match reply w/ request @@ -106,7 +76,7 @@ define bfd_udp_add_reply @param is_ipv6 - local_addr, peer_addr are IPv6 if non-zero, otherwise IPv4 @param detect_mult - detect multiplier (# of packets missed before connection goes down) */ -define bfd_udp_mod +autoreply define bfd_udp_mod { u32 client_index; u32 context; @@ -119,16 +89,6 @@ define bfd_udp_mod u8 detect_mult; }; -/** \brief Modify UDP BFD session response - @param context - sender context, to match reply w/ request - @param retval - return code for the request -*/ -define bfd_udp_mod_reply -{ - u32 context; - i32 retval; -}; - /** \brief Delete UDP BFD session on interface @param client_index - opaque cookie to identify the sender @param context - sender context, to match reply w/ request @@ -137,7 +97,7 @@ define bfd_udp_mod_reply @param peer_addr - peer address @param is_ipv6 - local_addr, peer_addr are IPv6 if non-zero, otherwise IPv4 */ -define bfd_udp_del +autoreply define bfd_udp_del { u32 client_index; u32 context; @@ -147,16 +107,6 @@ define bfd_udp_del u8 is_ipv6; }; -/** \brief Delete UDP BFD session response - @param context - sender context, to match reply w/ request - @param retval - return code for the request -*/ -define bfd_udp_del_reply -{ - u32 context; - i32 retval; -}; - /** \brief Get all BFD sessions @param client_index - opaque cookie to identify the sender @param context - sender context, to match reply w/ request @@ -206,7 +156,7 @@ define bfd_udp_session_details @param is_ipv6 - local_addr, peer_addr are IPv6 if non-zero, otherwise IPv4 @param admin_up_down - set the admin state, 1 = up, 0 = down */ -define bfd_udp_session_set_flags +autoreply define bfd_udp_session_set_flags { u32 client_index; u32 context; @@ -217,23 +167,13 @@ define bfd_udp_session_set_flags u8 admin_up_down; }; -/** \brief Reply to bfd_udp_session_set_flags - @param context - sender context which was passed in the request - @param retval - return code of the set flags request -*/ -define bfd_udp_session_set_flags_reply -{ - u32 context; - i32 retval; -}; - /** \brief Register for BFD events @param client_index - opaque cookie to identify the sender @param context - sender context, to match reply w/ request @param enable_disable - 1 => register for events, 0 => cancel registration @param pid - sender's pid */ -define want_bfd_events +autoreply define want_bfd_events { u32 client_index; u32 context; @@ -241,16 +181,6 @@ define want_bfd_events u32 pid; }; -/** \brief Reply for BFD events registration - @param context - returned sender context, to match reply w/ request - @param retval - return code -*/ -define want_bfd_events_reply -{ - u32 context; - i32 retval; -}; - /** \brief BFD UDP - add/replace key to configuration @param client_index - opaque cookie to identify the sender @param context - sender context, to match reply w/ request @@ -259,7 +189,7 @@ define want_bfd_events_reply @param auth_type - authentication type (RFC 5880/4.1/Auth Type) @param key - key data */ -define bfd_auth_set_key +autoreply define bfd_auth_set_key { u32 client_index; u32 context; @@ -269,16 +199,6 @@ define bfd_auth_set_key u8 key[20]; }; -/** \brief BFD UDP - add/replace key reply - @param context - returned sender context, to match reply w/ request - @param retval - return code -*/ -define bfd_auth_set_key_reply -{ - u32 context; - i32 retval; -}; - /** \brief BFD UDP - delete key from configuration @param client_index - opaque cookie to identify the sender @param context - sender context, to match reply w/ request @@ -286,23 +206,13 @@ define bfd_auth_set_key_reply @param key_len - length of key (must be non-zero) @param key - key data */ -define bfd_auth_del_key +autoreply define bfd_auth_del_key { u32 client_index; u32 context; u32 conf_key_id; }; -/** \brief BFD UDP - delete key reply - @param context - returned sender context, to match reply w/ request - @param retval - return code -*/ -define bfd_auth_del_key_reply -{ - u32 context; - i32 retval; -}; - /** \brief Get a list of configured authentication keys @param client_index - opaque cookie to identify the sender @param context - sender context, to match reply w/ request @@ -338,7 +248,7 @@ define bfd_auth_keys_details @param bfd_key_id - key id sent out in BFD packets @param conf_key_id - id of already configured key */ -define bfd_udp_auth_activate +autoreply define bfd_udp_auth_activate { u32 client_index; u32 context; @@ -351,16 +261,6 @@ define bfd_udp_auth_activate u32 conf_key_id; }; -/** \brief BFD UDP - activate/change authentication reply - @param context - returned sender context, to match reply w/ request - @param retval - return code -*/ -define bfd_udp_auth_activate_reply -{ - u32 context; - i32 retval; -}; - /** \brief BFD UDP - deactivate authentication @param client_index - opaque cookie to identify the sender @param context - sender context, to match reply w/ request @@ -370,7 +270,7 @@ define bfd_udp_auth_activate_reply @param is_ipv6 - local_addr, peer_addr are IPv6 if non-zero, otherwise IPv4 @param is_delayed - change is applied once peer applies the change (on first received non-authenticated packet) */ -define bfd_udp_auth_deactivate +autoreply define bfd_udp_auth_deactivate { u32 client_index; u32 context; @@ -381,16 +281,6 @@ define bfd_udp_auth_deactivate u8 is_delayed; }; -/** \brief BFD UDP - deactivate authentication reply - @param context - returned sender context, to match reply w/ request - @param retval - return code -*/ -define bfd_udp_auth_deactivate_reply -{ - u32 context; - i32 retval; -}; - /* * Local Variables: * eval: (c-set-style "gnu") diff --git a/src/vnet/classify/classify.api b/src/vnet/classify/classify.api index 51ebd6c8..cacb9bed 100644 --- a/src/vnet/classify/classify.api +++ b/src/vnet/classify/classify.api @@ -92,7 +92,7 @@ define classify_add_del_table_reply VRF id if action is 1 or 2. @param match[] - for add, match value for session, required */ -define classify_add_del_session +autoreply define classify_add_del_session { u32 client_index; u32 context; @@ -106,16 +106,6 @@ define classify_add_del_session u8 match[0]; }; -/** \brief Classify add / del session response - @param context - sender context, to match reply w/ request - @param retval - return code for the add/del session request -*/ -define classify_add_del_session_reply -{ - u32 context; - i32 retval; -}; - /** \brief Set/unset policer classify interface @param client_index - opaque cookie to identify the sender @param context - sender context, to match reply w/ request @@ -127,7 +117,7 @@ define classify_add_del_session_reply Note: User is recommeneded to use just one valid table_index per call. (ip4_table_index, ip6_table_index, or l2_table_index) */ -define policer_classify_set_interface +autoreply define policer_classify_set_interface { u32 client_index; u32 context; @@ -138,16 +128,6 @@ define policer_classify_set_interface u8 is_add; }; -/** \brief Set/unset policer classify interface response - @param context - sender context, to match reply w/ request - @param retval - return value for request -*/ -define policer_classify_set_interface_reply -{ - u32 context; - i32 retval; -}; - /** \brief Get list of policer classify interfaces and tables @param client_index - opaque cookie to identify the sender @param context - sender context, to match reply w/ request @@ -308,7 +288,7 @@ define classify_session_details Note: User is recommeneded to use just one valid table_index per call. (ip4_table_index, ip6_table_index, or l2_table_index) */ -define flow_classify_set_interface { +autoreply define flow_classify_set_interface { u32 client_index; u32 context; u32 sw_if_index; @@ -317,15 +297,6 @@ define flow_classify_set_interface { u8 is_add; }; -/** \brief Set/unset flow classify interface response - @param context - sender context, to match reply w/ request - @param retval - return value for request -*/ -define flow_classify_set_interface_reply { - u32 context; - i32 retval; -}; - /** \brief Get list of flow classify interfaces and tables @param client_index - opaque cookie to identify the sender @param context - sender context, to match reply w/ request @@ -353,4 +324,4 @@ define flow_classify_details { * eval: (c-set-style "gnu") * End: */ - \ No newline at end of file + diff --git a/src/vnet/cop/cop.api b/src/vnet/cop/cop.api index b34dae80..69316001 100644 --- a/src/vnet/cop/cop.api +++ b/src/vnet/cop/cop.api @@ -20,7 +20,7 @@ @param enable_disable - 1 => enable, 0 => disable */ -define cop_interface_enable_disable +autoreply define cop_interface_enable_disable { u32 client_index; u32 context; @@ -28,17 +28,6 @@ define cop_interface_enable_disable u8 enable_disable; }; -/** \brief cop: interface enable/disable junk filtration reply - @param context - returned sender context, to match reply w/ request - @param retval - return code -*/ - -define cop_interface_enable_disable_reply -{ - u32 context; - i32 retval; -}; - /** \brief cop: enable/disable whitelist filtration features on an interface Note: the supplied fib_id must match in order to remove the feature! @@ -51,7 +40,7 @@ define cop_interface_enable_disable_reply @param default_cop - 1 => enable non-ip4, non-ip6 filtration 0=> disable it */ -define cop_whitelist_enable_disable +autoreply define cop_whitelist_enable_disable { u32 client_index; u32 context; @@ -62,17 +51,6 @@ define cop_whitelist_enable_disable u8 default_cop; }; -/** \brief cop: interface enable/disable junk filtration reply - @param context - returned sender context, to match reply w/ request - @param retval - return code -*/ - -define cop_whitelist_enable_disable_reply -{ - u32 context; - i32 retval; -}; - /** \brief get_node_graph - get a copy of the vpp node graph including the current set of graph arcs. @@ -85,4 +63,4 @@ define cop_whitelist_enable_disable_reply * eval: (c-set-style "gnu") * End: */ - \ No newline at end of file + diff --git a/src/vnet/devices/af_packet/af_packet.api b/src/vnet/devices/af_packet/af_packet.api index 9fb2a207..8d40ad60 100644 --- a/src/vnet/devices/af_packet/af_packet.api +++ b/src/vnet/devices/af_packet/af_packet.api @@ -46,7 +46,7 @@ define af_packet_create_reply @param context - sender context, to match reply w/ request @param host_if_name - interface name */ -define af_packet_delete +autoreply define af_packet_delete { u32 client_index; u32 context; @@ -54,16 +54,6 @@ define af_packet_delete u8 host_if_name[64]; }; -/** \brief Delete host-interface response - @param context - sender context, to match reply w/ request - @param retval - return value for request -*/ -define af_packet_delete_reply -{ - u32 context; - i32 retval; -}; - /* * Local Variables: * eval: (c-set-style "gnu") diff --git a/src/vnet/devices/netmap/netmap.api b/src/vnet/devices/netmap/netmap.api index 377ccffd..8dc698b9 100644 --- a/src/vnet/devices/netmap/netmap.api +++ b/src/vnet/devices/netmap/netmap.api @@ -22,7 +22,7 @@ @param is_pipe - is pipe @param is_master - 0=slave, 1=master */ -define netmap_create +autoreply define netmap_create { u32 client_index; u32 context; @@ -34,22 +34,12 @@ define netmap_create u8 is_master; }; -/** \brief Create netmap response - @param context - sender context, to match reply w/ request - @param retval - return value for request -*/ -define netmap_create_reply -{ - u32 context; - i32 retval; -}; - /** \brief Delete netmap @param client_index - opaque cookie to identify the sender @param context - sender context, to match reply w/ request @param netmap_if_name - interface name */ -define netmap_delete +autoreply define netmap_delete { u32 client_index; u32 context; @@ -57,16 +47,6 @@ define netmap_delete u8 netmap_if_name[64]; }; -/** \brief Delete netmap response - @param context - sender context, to match reply w/ request - @param retval - return value for request -*/ -define netmap_delete_reply -{ - u32 context; - i32 retval; -}; - /* * Local Variables: * eval: (c-set-style "gnu") diff --git a/src/vnet/devices/virtio/vhost_user.api b/src/vnet/devices/virtio/vhost_user.api index 4f604e45..df7ce7ab 100644 --- a/src/vnet/devices/virtio/vhost_user.api +++ b/src/vnet/devices/virtio/vhost_user.api @@ -53,7 +53,7 @@ define create_vhost_user_if_reply @param sock_filename - unix socket filename, used to speak with frontend @param operation_mode - polling=0, interrupt=1, or adaptive=2 */ -define modify_vhost_user_if +autoreply define modify_vhost_user_if { u32 client_index; u32 context; @@ -65,36 +65,16 @@ define modify_vhost_user_if u8 operation_mode; }; -/** \brief vhost-user interface modify response - @param context - sender context, to match reply w/ request - @param retval - return code for the request -*/ -define modify_vhost_user_if_reply -{ - u32 context; - i32 retval; -}; - /** \brief vhost-user interface delete request @param client_index - opaque cookie to identify the sender */ -define delete_vhost_user_if +autoreply define delete_vhost_user_if { u32 client_index; u32 context; u32 sw_if_index; }; -/** \brief vhost-user interface delete response - @param context - sender context, to match reply w/ request - @param retval - return code for the request -*/ -define delete_vhost_user_if_reply -{ - u32 context; - i32 retval; -}; - /** \brief Vhost-user interface details structure (fix this) @param sw_if_index - index of the interface @param interface_name - name of interface diff --git a/src/vnet/dhcp/dhcp.api b/src/vnet/dhcp/dhcp.api index 2db85a79..eb0b070d 100644 --- a/src/vnet/dhcp/dhcp.api +++ b/src/vnet/dhcp/dhcp.api @@ -24,7 +24,7 @@ @param dhcp_server[] - server address @param dhcp_src_address[] - */ -define dhcp_proxy_config +autoreply define dhcp_proxy_config { u32 client_index; u32 context; @@ -36,16 +36,6 @@ define dhcp_proxy_config u8 dhcp_src_address[16]; }; -/** \brief DHCP Proxy config response - @param context - sender context, to match reply w/ request - @param retval - return code for the request -*/ -define dhcp_proxy_config_reply -{ - u32 context; - i32 retval; -}; - /** \brief DHCP Proxy set / unset vss request @param client_index - opaque cookie to identify the sender @param context - sender context, to match reply w/ request @@ -55,7 +45,7 @@ define dhcp_proxy_config_reply @param is_ipv6 - ip6 if non-zero, else ip4 @param is_add - set vss if non-zero, else delete */ -define dhcp_proxy_set_vss +autoreply define dhcp_proxy_set_vss { u32 client_index; u32 context; @@ -66,16 +56,6 @@ define dhcp_proxy_set_vss u8 is_add; }; -/** \brief DHCP proxy set / unset vss response - @param context - sender context, to match reply w/ request - @param retval - return code for the request -*/ -define dhcp_proxy_set_vss_reply -{ - u32 context; - i32 retval; -}; - /** \brief DHCP Client config add / del request @param client_index - opaque cookie to identify the sender @param context - sender context, to match reply w/ request @@ -86,7 +66,7 @@ define dhcp_proxy_set_vss_reply via dhcp_compl_event API message if non-zero @param pid - sender's pid */ -define dhcp_client_config +autoreply define dhcp_client_config { u32 client_index; u32 context; @@ -97,16 +77,6 @@ define dhcp_client_config u32 pid; }; -/** \brief DHCP Client config response - @param context - sender context, to match reply w/ request - @param retval - return code for the request -*/ -define dhcp_client_config_reply -{ - u32 context; - i32 retval; -}; - /** \brief Tell client about a DHCP completion event @param client_index - opaque cookie to identify the sender @param pid - client pid registered to receive notification @@ -162,4 +132,4 @@ manual_endian manual_print define dhcp_proxy_details * Local Variables: * eval: (c-set-style "gnu") * End: - */ \ No newline at end of file + */ diff --git a/src/vnet/flow/flow.api b/src/vnet/flow/flow.api index 0e0f99bf..1c5e8c5c 100644 --- a/src/vnet/flow/flow.api +++ b/src/vnet/flow/flow.api @@ -24,7 +24,7 @@ @param template_interval - number of seconds after which to resend template @param udp_checksum - UDP checksum calculation enable flag */ -define set_ipfix_exporter +autoreply define set_ipfix_exporter { u32 client_index; u32 context; @@ -37,15 +37,6 @@ define set_ipfix_exporter u8 udp_checksum; }; -/** \brief Reply to IPFIX exporter configure request - @param context - sender context which was passed in the request -*/ -define set_ipfix_exporter_reply -{ - u32 context; - i32 retval; -}; - /** \brief IPFIX exporter dump request @param client_index - opaque cookie to identify the sender @param context - sender context, to match reply w/ request @@ -84,22 +75,13 @@ define ipfix_exporter_details @param domain_id - domain ID reported in IPFIX messages for classify stream @param src_port - source port of UDP session for classify stream */ -define set_ipfix_classify_stream { +autoreply define set_ipfix_classify_stream { u32 client_index; u32 context; u32 domain_id; u16 src_port; }; -/** \brief IPFIX classify stream configure response - @param context - sender context, to match reply w/ request - @param retval - return value for request -*/ -define set_ipfix_classify_stream_reply { - u32 context; - i32 retval; -}; - /** \brief IPFIX classify stream dump request @param client_index - opaque cookie to identify the sender @param context - sender context, to match reply w/ request @@ -127,7 +109,7 @@ define ipfix_classify_stream_details { @param ip_version - version of IP used in the classifier table @param transport_protocol - transport protocol used in the classifier table or 255 for unspecified */ -define ipfix_classify_table_add_del { +autoreply define ipfix_classify_table_add_del { u32 client_index; u32 context; u32 table_id; @@ -136,14 +118,6 @@ define ipfix_classify_table_add_del { u8 is_add; }; -/** \brief IPFIX add classifier table response - @param context - sender context which was passed in the request -*/ -define ipfix_classify_table_add_del_reply { - u32 context; - i32 retval; -}; - /** \brief IPFIX classify tables dump request @param client_index - opaque cookie to identify the sender @param context - sender context, to match reply w/ request diff --git a/src/vnet/interface.api b/src/vnet/interface.api index 85fd73fb..9df63f18 100644 --- a/src/vnet/interface.api +++ b/src/vnet/interface.api @@ -6,7 +6,7 @@ @param link_up_down - Oper state sent on change event, not used in config. @param deleted - interface was deleted */ -define sw_interface_set_flags +autoreply define sw_interface_set_flags { u32 client_index; u32 context; @@ -17,23 +17,13 @@ define sw_interface_set_flags u8 deleted; }; -/** \brief Reply to sw_interface_set_flags - @param context - sender context which was passed in the request - @param retval - return code of the set flags request -*/ -define sw_interface_set_flags_reply -{ - u32 context; - i32 retval; -}; - /** \brief Set interface MTU @param client_index - opaque cookie to identify the sender @param context - sender context, to match reply w/ request @param sw_if_index - index of the interface to set MTU on @param mtu - MTU */ -define sw_interface_set_mtu +autoreply define sw_interface_set_mtu { u32 client_index; u32 context; @@ -41,23 +31,13 @@ define sw_interface_set_mtu u16 mtu; }; -/** \brief Reply to sw_interface_set_mtu - @param context - sender context which was passed in the request - @param retval - return code of the set flags request -*/ -define sw_interface_set_mtu_reply -{ - u32 context; - i32 retval; -}; - /** \brief Register for interface events @param client_index - opaque cookie to identify the sender @param context - sender context, to match reply w/ request @param enable_disable - 1 => register for events, 0 => cancel registration @param pid - sender's pid */ -define want_interface_events +autoreply define want_interface_events { u32 client_index; u32 context; @@ -65,16 +45,6 @@ define want_interface_events u32 pid; }; -/** \brief Reply for interface events registration - @param context - returned sender context, to match reply w/ request - @param retval - return code -*/ -define want_interface_events_reply -{ - u32 context; - i32 retval; -}; - /** \brief Interface details structure (fix this) @param sw_if_index - index of the interface @param sup_sw_if_index - index of parent interface if any, else same as sw_if_index @@ -184,7 +154,7 @@ define sw_interface_dump @param address_length - address length in bytes, 4 for ip4, 16 for ip6 @param address - array of address bytes */ -define sw_interface_add_del_address +autoreply define sw_interface_add_del_address { u32 client_index; u32 context; @@ -196,16 +166,6 @@ define sw_interface_add_del_address u8 address[16]; }; -/** \brief Reply to sw_interface_add_del_address - @param context - returned sender context, to match reply w/ request - @param retval - return code -*/ -define sw_interface_add_del_address_reply -{ - u32 context; - i32 retval; -}; - /** \brief Associate the specified interface with a fib table @param client_index - opaque cookie to identify the sender @param context - sender context, to match reply w/ request @@ -213,7 +173,7 @@ define sw_interface_add_del_address_reply @param is_ipv6 - if non-zero ipv6, else ipv4 @param vrf_id - fib table/vrd id to associate the interface with */ -define sw_interface_set_table +autoreply define sw_interface_set_table { u32 client_index; u32 context; @@ -222,16 +182,6 @@ define sw_interface_set_table u32 vrf_id; }; -/** \brief Reply to sw_interface_set_table - @param context - returned sender context, to match reply w/ request - @param retval - return code -*/ -define sw_interface_set_table_reply -{ - u32 context; - i32 retval; -}; - /** \brief Get VRF id assigned to interface @param client_index - opaque cookie to identify the sender @param context - sender context, to match reply w/ request @@ -280,7 +230,7 @@ define vnet_interface_counters @param unnumbered_sw_if_index - interface which will use the address @param is_add - if non-zero set the association, else unset it */ -define sw_interface_set_unnumbered +autoreply define sw_interface_set_unnumbered { u32 client_index; u32 context; @@ -289,38 +239,18 @@ define sw_interface_set_unnumbered u8 is_add; }; -/** \brief Set unnumbered interface add / del response - @param context - sender context, to match reply w/ request - @param retval - return code for the request -*/ -define sw_interface_set_unnumbered_reply -{ - u32 context; - i32 retval; -}; - /** \brief Clear interface statistics @param client_index - opaque cookie to identify the sender @param context - sender context, to match reply w/ request @param sw_if_index - index of the interface to clear statistics */ -define sw_interface_clear_stats +autoreply define sw_interface_clear_stats { u32 client_index; u32 context; u32 sw_if_index; }; -/** \brief Reply to sw_interface_clear_stats - @param context - sender context which was passed in the request - @param retval - return code of the set flags request -*/ -define sw_interface_clear_stats_reply -{ - u32 context; - i32 retval; -}; - /** \brief Set / clear software interface tag @param client_index - opaque cookie to identify the sender @param context - sender context, to match reply w/ request @@ -328,7 +258,7 @@ define sw_interface_clear_stats_reply @param add_del - 1 = add, 0 = delete @param tag - an ascii tag */ -define sw_interface_tag_add_del +autoreply define sw_interface_tag_add_del { u32 client_index; u32 context; @@ -337,23 +267,13 @@ define sw_interface_tag_add_del u8 tag[64]; }; -/** \brief Reply to set / clear software interface tag - @param context - sender context which was passed in the request - @param retval - return code for the request -*/ -define sw_interface_tag_add_del_reply -{ - u32 context; - i32 retval; -}; - /** \brief Set an interface's MAC address @param client_index - opaque cookie to identify the sender @param context - sender context, to match reply w/ request @param sw_if_index - the interface whose MAC will be set @param mac_addr - the new MAC address */ -define sw_interface_set_mac_address +autoreply define sw_interface_set_mac_address { u32 client_index; u32 context; @@ -361,16 +281,6 @@ define sw_interface_set_mac_address u8 mac_address[6]; }; -/** \brief Reply to setting an interface MAC address request - @param context - sender context which was passed in the request - @param retval - return code for the request -*/ -define sw_interface_set_mac_address_reply -{ - u32 context; - i32 retval; -}; - /* * Local Variables: * eval: (c-set-style "gnu") diff --git a/src/vnet/ip/ip.api b/src/vnet/ip/ip.api index 6af1714f..7097a130 100644 --- a/src/vnet/ip/ip.api +++ b/src/vnet/ip/ip.api @@ -136,7 +136,7 @@ define ip_neighbor_details { @param mac_address - l2 address of the neighbor @param dst_address - ip4 or ip6 address of the neighbor */ -define ip_neighbor_add_del +autoreply define ip_neighbor_add_del { u32 client_index; u32 context; @@ -150,16 +150,6 @@ define ip_neighbor_add_del u8 dst_address[16]; }; -/** \brief Reply for IP Neighbor add / delete request - @param context - returned sender context, to match reply w/ request - @param retval - return code -*/ -define ip_neighbor_add_del_reply -{ - u32 context; - i32 retval; -}; - /** \brief Set the ip flow hash config for a fib request @param client_index - opaque cookie to identify the sender @param context - sender context, to match reply w/ request @@ -172,7 +162,7 @@ define ip_neighbor_add_del_reply @param proto -if non-zero include proto in flow hash @param reverse - if non-zero include reverse in flow hash */ -define set_ip_flow_hash +autoreply define set_ip_flow_hash { u32 client_index; u32 context; @@ -186,16 +176,6 @@ define set_ip_flow_hash u8 reverse; }; -/** \brief Set the ip flow hash config for a fib response - @param context - sender context, to match reply w/ request - @param retval - return code for the request -*/ -define set_ip_flow_hash_reply -{ - u32 context; - i32 retval; -}; - /** \brief IPv6 router advertisement config request @param client_index - opaque cookie to identify the sender @param context - sender context, to match reply w/ request @@ -213,7 +193,7 @@ define set_ip_flow_hash_reply @param initial_count - @param initial_interval - */ -define sw_interface_ip6nd_ra_config +autoreply define sw_interface_ip6nd_ra_config { u32 client_index; u32 context; @@ -233,16 +213,6 @@ define sw_interface_ip6nd_ra_config u32 initial_interval; }; -/** \brief IPv6 router advertisement config response - @param context - sender context, to match reply w/ request - @param retval - return code for the request -*/ -define sw_interface_ip6nd_ra_config_reply -{ - u32 context; - i32 retval; -}; - /** \brief IPv6 router advertisement prefix config request @param client_index - opaque cookie to identify the sender @param context - sender context, to match reply w/ request @@ -272,7 +242,7 @@ define sw_interface_ip6nd_ra_config_reply preferred [ADDRCONF]. A value of all one bits (0xffffffff) represents infinity. */ -define sw_interface_ip6nd_ra_prefix +autoreply define sw_interface_ip6nd_ra_prefix { u32 client_index; u32 context; @@ -289,16 +259,6 @@ define sw_interface_ip6nd_ra_prefix u32 pref_lifetime; }; -/** \brief IPv6 router advertisement prefix config response - @param context - sender context, to match reply w/ request - @param retval - return code for the request -*/ -define sw_interface_ip6nd_ra_prefix_reply -{ - u32 context; - i32 retval; -}; - /** \brief IPv6 ND proxy config @param client_index - opaque cookie to identify the sender @param context - sender context, to match reply w/ request @@ -306,7 +266,7 @@ define sw_interface_ip6nd_ra_prefix_reply @param address - The address of the host for which to proxy for @param is_add - Adding or deleting */ -define ip6nd_proxy_add_del +autoreply define ip6nd_proxy_add_del { u32 client_index; u32 context; @@ -315,16 +275,6 @@ define ip6nd_proxy_add_del u8 address[16]; }; -/** \brief IPv6 ND proxy response - @param context - sender context, to match reply w/ request - @param retval - return code for the request -*/ -define ip6nd_proxy_add_del_reply -{ - u32 context; - i32 retval; -}; - /** \brief IPv6 ND proxy details returned after request @param context - sender context, to match reply w/ request @param retval - return code for the request @@ -355,7 +305,7 @@ define ip6nd_proxy_dump @param sw_if_index - interface used to reach neighbor @param enable - if non-zero enable ip6 on interface, else disable */ -define sw_interface_ip6_enable_disable +autoreply define sw_interface_ip6_enable_disable { u32 client_index; u32 context; @@ -363,23 +313,13 @@ define sw_interface_ip6_enable_disable u8 enable; /* set to true if enable */ }; -/** \brief IPv6 interface enable / disable response - @param context - sender context, to match reply w/ request - @param retval - return code for the request -*/ -define sw_interface_ip6_enable_disable_reply -{ - u32 context; - i32 retval; -}; - /** \brief IPv6 set link local address on interface request @param client_index - opaque cookie to identify the sender @param context - sender context, to match reply w/ request @param sw_if_index - interface to set link local on @param address[] - the new link local address */ -define sw_interface_ip6_set_link_local_address +autoreply define sw_interface_ip6_set_link_local_address { u32 client_index; u32 context; @@ -387,16 +327,6 @@ define sw_interface_ip6_set_link_local_address u8 address[16]; }; -/** \brief IPv6 set link local address on interface response - @param context - sender context, to match reply w/ request - @param retval - error code for the request -*/ -define sw_interface_ip6_set_link_local_address_reply -{ - u32 context; - i32 retval; -}; - /** \brief Add / del route request @param client_index - opaque cookie to identify the sender @param context - sender context, to match reply w/ request @@ -422,7 +352,7 @@ define sw_interface_ip6_set_link_local_address_reply @param next_hop_out_label_stack - the next-hop output label stack, outer most first @param next_hop_via_label - The next-hop is a resolved via a local label */ -define ip_add_del_route +autoreply define ip_add_del_route { u32 client_index; u32 context; @@ -452,16 +382,6 @@ define ip_add_del_route u32 next_hop_out_label_stack[next_hop_n_out_labels]; }; -/** \brief Reply for add / del route request - @param context - returned sender context, to match reply w/ request - @param retval - return code -*/ -define ip_add_del_route_reply -{ - u32 context; - i32 retval; -}; - /** \brief Add / del route request @param client_index - opaque cookie to identify the sender @param context - sender context, to match reply w/ request @@ -470,7 +390,7 @@ define ip_add_del_route_reply FIXME */ -define ip_mroute_add_del +autoreply define ip_mroute_add_del { u32 client_index; u32 context; @@ -488,16 +408,6 @@ define ip_mroute_add_del u8 src_address[16]; }; -/** \brief Reply for add / del mroute request - @param context - returned sender context, to match reply w/ request - @param retval - return code -*/ -define ip_mroute_add_del_reply -{ - u32 context; - i32 retval; -}; - /** \brief Dump IP multicast fib table @param client_index - opaque cookie to identify the sender */ diff --git a/src/vnet/ipsec/ipsec.api b/src/vnet/ipsec/ipsec.api index ef090f84..203c5272 100644 --- a/src/vnet/ipsec/ipsec.api +++ b/src/vnet/ipsec/ipsec.api @@ -20,7 +20,7 @@ @param spd_id - SPD instance id (control plane allocated) */ -define ipsec_spd_add_del +autoreply define ipsec_spd_add_del { u32 client_index; u32 context; @@ -28,17 +28,6 @@ define ipsec_spd_add_del u32 spd_id; }; -/** \brief Reply for IPsec: Add/delete Security Policy Database entry - @param context - returned sender context, to match reply w/ request - @param retval - return code -*/ - -define ipsec_spd_add_del_reply -{ - u32 context; - i32 retval; -}; - /** \brief IPsec: Add/delete SPD from interface @param client_index - opaque cookie to identify the sender @@ -49,7 +38,7 @@ define ipsec_spd_add_del_reply */ -define ipsec_interface_add_del_spd +autoreply define ipsec_interface_add_del_spd { u32 client_index; u32 context; @@ -59,17 +48,6 @@ define ipsec_interface_add_del_spd u32 spd_id; }; -/** \brief Reply for IPsec: Add/delete SPD from interface - @param context - returned sender context, to match reply w/ request - @param retval - return code -*/ - -define ipsec_interface_add_del_spd_reply -{ - u32 context; - i32 retval; -}; - /** \brief IPsec: Add/delete Security Policy Database entry See RFC 4301, 4.4.1.1 on how to match packet to selectors @@ -95,7 +73,7 @@ define ipsec_interface_add_del_spd_reply */ -define ipsec_spd_add_del_entry +autoreply define ipsec_spd_add_del_entry { u32 client_index; u32 context; @@ -125,17 +103,6 @@ define ipsec_spd_add_del_entry u32 sa_id; }; -/** \brief Reply for IPsec: Add/delete Security Policy Database entry - @param context - returned sender context, to match reply w/ request - @param retval - return code -*/ - -define ipsec_spd_add_del_entry_reply -{ - u32 context; - i32 retval; -}; - /** \brief IPsec: Add/delete Security Association Database entry @param client_index - opaque cookie to identify the sender @param context - sender context, to match reply w/ request @@ -167,7 +134,7 @@ define ipsec_spd_add_del_entry_reply IPsec tunnel address copy mode (to support GDOI) */ -define ipsec_sad_add_del_entry +autoreply define ipsec_sad_add_del_entry { u32 client_index; u32 context; @@ -195,17 +162,6 @@ define ipsec_sad_add_del_entry u8 tunnel_dst_address[16]; }; -/** \brief Reply for IPsec: Add/delete Security Association Database entry - @param context - returned sender context, to match reply w/ request - @param retval - return code -*/ - -define ipsec_sad_add_del_entry_reply -{ - u32 context; - i32 retval; -}; - /** \brief IPsec: Update Security Association keys @param client_index - opaque cookie to identify the sender @param context - sender context, to match reply w/ request @@ -219,7 +175,7 @@ define ipsec_sad_add_del_entry_reply @param integrity_key - integrity keying material */ -define ipsec_sa_set_key +autoreply define ipsec_sa_set_key { u32 client_index; u32 context; @@ -233,17 +189,6 @@ define ipsec_sa_set_key u8 integrity_key[128]; }; -/** \brief Reply for IPsec: Update Security Association keys - @param context - returned sender context, to match reply w/ request - @param retval - return code -*/ - -define ipsec_sa_set_key_reply -{ - u32 context; - i32 retval; -}; - /** \brief IKEv2: Add/delete profile @param client_index - opaque cookie to identify the sender @param context - sender context, to match reply w/ request @@ -251,7 +196,7 @@ define ipsec_sa_set_key_reply @param name - IKEv2 profile name @param is_add - Add IKEv2 profile if non-zero, else delete */ -define ikev2_profile_add_del +autoreply define ikev2_profile_add_del { u32 client_index; u32 context; @@ -260,16 +205,6 @@ define ikev2_profile_add_del u8 is_add; }; -/** \brief Reply for IKEv2: Add/delete profile - @param context - returned sender context, to match reply w/ request - @param retval - return code -*/ -define ikev2_profile_add_del_reply -{ - u32 context; - i32 retval; -}; - /** \brief IKEv2: Set IKEv2 profile authentication method @param client_index - opaque cookie to identify the sender @param context - sender context, to match reply w/ request @@ -280,7 +215,7 @@ define ikev2_profile_add_del_reply @param data_len - Authentication data length @param data - Authentication data (for rsa-sig cert file path) */ -define ikev2_profile_set_auth +autoreply define ikev2_profile_set_auth { u32 client_index; u32 context; @@ -292,16 +227,6 @@ define ikev2_profile_set_auth u8 data[0]; }; -/** \brief Reply for IKEv2: Set IKEv2 profile authentication method - @param context - returned sender context, to match reply w/ request - @param retval - return code -*/ -define ikev2_profile_set_auth_reply -{ - u32 context; - i32 retval; -}; - /** \brief IKEv2: Set IKEv2 profile local/remote identification @param client_index - opaque cookie to identify the sender @param context - sender context, to match reply w/ request @@ -312,7 +237,7 @@ define ikev2_profile_set_auth_reply @param data_len - Identification data length @param data - Identification data */ -define ikev2_profile_set_id +autoreply define ikev2_profile_set_id { u32 client_index; u32 context; @@ -324,16 +249,6 @@ define ikev2_profile_set_id u8 data[0]; }; -/** \brief Reply for IKEv2: - @param context - returned sender context, to match reply w/ request - @param retval - return code -*/ -define ikev2_profile_set_id_reply -{ - u32 context; - i32 retval; -}; - /** \brief IKEv2: Set IKEv2 profile traffic selector parameters @param client_index - opaque cookie to identify the sender @param context - sender context, to match reply w/ request @@ -346,7 +261,7 @@ define ikev2_profile_set_id_reply @param start_addr - The smallest address included in traffic selector @param end_addr - The largest address included in traffic selector */ -define ikev2_profile_set_ts +autoreply define ikev2_profile_set_ts { u32 client_index; u32 context; @@ -360,23 +275,13 @@ define ikev2_profile_set_ts u32 end_addr; }; -/** \brief Reply for IKEv2: Set IKEv2 profile traffic selector parameters - @param context - returned sender context, to match reply w/ request - @param retval - return code -*/ -define ikev2_profile_set_ts_reply -{ - u32 context; - i32 retval; -}; - /** \brief IKEv2: Set IKEv2 local RSA private key @param client_index - opaque cookie to identify the sender @param context - sender context, to match reply w/ request @param key_file - Key file absolute path */ -define ikev2_set_local_key +autoreply define ikev2_set_local_key { u32 client_index; u32 context; @@ -384,16 +289,6 @@ define ikev2_set_local_key u8 key_file[256]; }; -/** \brief Reply for IKEv2: Set IKEv2 local key - @param context - returned sender context, to match reply w/ request - @param retval - return code -*/ -define ikev2_set_local_key_reply -{ - u32 context; - i32 retval; -}; - /** \brief IKEv2: Set IKEv2 responder interface and IP address @param client_index - opaque cookie to identify the sender @param context - sender context, to match reply w/ request @@ -402,7 +297,7 @@ define ikev2_set_local_key_reply @param sw_if_index - interface index @param address - interface address */ -define ikev2_set_responder +autoreply define ikev2_set_responder { u32 client_index; u32 context; @@ -412,17 +307,6 @@ define ikev2_set_responder u8 address[4]; }; -/** \brief Reply for IKEv2: Set IKEv2 responder interface and IP address - @param context - returned sender context, to match reply w/ request - @param retval - return code -*/ -define ikev2_set_responder_reply -{ - u32 context; - i32 retval; -}; - - /** \brief IKEv2: Set IKEv2 IKE transforms in SA_INIT proposal (RFC 7296) @param client_index - opaque cookie to identify the sender @param context - sender context, to match reply w/ request @@ -434,7 +318,7 @@ define ikev2_set_responder_reply @param dh_group - Diffie-Hellman group */ -define ikev2_set_ike_transforms +autoreply define ikev2_set_ike_transforms { u32 client_index; u32 context; @@ -446,16 +330,6 @@ define ikev2_set_ike_transforms u32 dh_group; }; -/** \brief Reply for IKEv2: Set IKEv2 IKE transforms - @param context - returned sender context, to match reply w/ request - @param retval - return code -*/ -define ikev2_set_ike_transforms_reply -{ - u32 context; - i32 retval; -}; - /** \brief IKEv2: Set IKEv2 ESP transforms in SA_INIT proposal (RFC 7296) @param client_index - opaque cookie to identify the sender @param context - sender context, to match reply w/ request @@ -467,7 +341,7 @@ define ikev2_set_ike_transforms_reply @param dh_group - Diffie-Hellman group */ -define ikev2_set_esp_transforms +autoreply define ikev2_set_esp_transforms { u32 client_index; u32 context; @@ -479,16 +353,6 @@ define ikev2_set_esp_transforms u32 dh_group; }; -/** \brief Reply for IKEv2: Set IKEv2 ESP transforms - @param context - returned sender context, to match reply w/ request - @param retval - return code -*/ -define ikev2_set_esp_transforms_reply -{ - u32 context; - i32 retval; -}; - /** \brief IKEv2: Set Child SA lifetime, limited by time and/or data @param client_index - opaque cookie to identify the sender @param context - sender context, to match reply w/ request @@ -500,7 +364,7 @@ define ikev2_set_esp_transforms_reply @param lifetime_maxdata - SA maximum life time in bytes (0 to disable) */ -define ikev2_set_sa_lifetime +autoreply define ikev2_set_sa_lifetime { u32 client_index; u32 context; @@ -512,16 +376,6 @@ define ikev2_set_sa_lifetime u64 lifetime_maxdata; }; -/** \brief Reply for IKEv2: Set Child SA lifetime - @param context - returned sender context, to match reply w/ request - @param retval - return code -*/ -define ikev2_set_sa_lifetime_reply -{ - u32 context; - i32 retval; -}; - /** \brief IKEv2: Initiate the SA_INIT exchange @param client_index - opaque cookie to identify the sender @param context - sender context, to match reply w/ request @@ -529,7 +383,7 @@ define ikev2_set_sa_lifetime_reply @param name - IKEv2 profile name */ -define ikev2_initiate_sa_init +autoreply define ikev2_initiate_sa_init { u32 client_index; u32 context; @@ -537,16 +391,6 @@ define ikev2_initiate_sa_init u8 name[64]; }; -/** \brief Reply for IKEv2: Initiate the SA_INIT exchange - @param context - returned sender context, to match reply w/ request - @param retval - return code -*/ -define ikev2_initiate_sa_init_reply -{ - u32 context; - i32 retval; -}; - /** \brief IKEv2: Initiate the delete IKE SA exchange @param client_index - opaque cookie to identify the sender @param context - sender context, to match reply w/ request @@ -554,7 +398,7 @@ define ikev2_initiate_sa_init_reply @param ispi - IKE SA initiator SPI */ -define ikev2_initiate_del_ike_sa +autoreply define ikev2_initiate_del_ike_sa { u32 client_index; u32 context; @@ -562,16 +406,6 @@ define ikev2_initiate_del_ike_sa u64 ispi; }; -/** \brief Reply for IKEv2: Initiate the delete IKE SA exchange - @param context - returned sender context, to match reply w/ request - @param retval - return code -*/ -define ikev2_initiate_del_ike_sa_reply -{ - u32 context; - i32 retval; -}; - /** \brief IKEv2: Initiate the delete Child SA exchange @param client_index - opaque cookie to identify the sender @param context - sender context, to match reply w/ request @@ -579,7 +413,7 @@ define ikev2_initiate_del_ike_sa_reply @param ispi - Child SA initiator SPI */ -define ikev2_initiate_del_child_sa +autoreply define ikev2_initiate_del_child_sa { u32 client_index; u32 context; @@ -587,16 +421,6 @@ define ikev2_initiate_del_child_sa u32 ispi; }; -/** \brief Reply for IKEv2: Initiate the delete Child SA exchange - @param context - returned sender context, to match reply w/ request - @param retval - return code -*/ -define ikev2_initiate_del_child_sa_reply -{ - u32 context; - i32 retval; -}; - /** \brief IKEv2: Initiate the rekey Child SA exchange @param client_index - opaque cookie to identify the sender @param context - sender context, to match reply w/ request @@ -604,7 +428,7 @@ define ikev2_initiate_del_child_sa_reply @param ispi - Child SA initiator SPI */ -define ikev2_initiate_rekey_child_sa +autoreply define ikev2_initiate_rekey_child_sa { u32 client_index; u32 context; @@ -612,16 +436,6 @@ define ikev2_initiate_rekey_child_sa u32 ispi; }; -/** \brief Reply for IKEv2: Initiate the rekey Child SA exchange - @param context - returned sender context, to match reply w/ request - @param retval - return code -*/ -define ikev2_initiate_rekey_child_sa_reply -{ - u32 context; - i32 retval; -}; - /** \brief Dump ipsec policy database data @param client_index - opaque cookie to identify the sender @param context - sender context, to match reply w/ request @@ -682,4 +496,4 @@ define ipsec_spd_details { * eval: (c-set-style "gnu") * End: */ - \ No newline at end of file + diff --git a/src/vnet/l2/l2.api b/src/vnet/l2/l2.api index c23eebec..db42d635 100644 --- a/src/vnet/l2/l2.api +++ b/src/vnet/l2/l2.api @@ -70,66 +70,36 @@ define l2_fib_table_dump @param client_index - opaque cookie to identify the sender @param context - sender context, to match reply w/ request */ -define l2_fib_clear_table +autoreply define l2_fib_clear_table { u32 client_index; u32 context; }; -/** \brief L2 fib clear table response - @param context - sender context, to match reply w/ request - @param retval - return code for the request -*/ -define l2_fib_clear_table_reply -{ - u32 context; - i32 retval; -}; - /** \brief L2 FIB flush bridge domain entries @param client_index - opaque cookie to identify the sender @param context - sender context, to match reply w/ request @param bd_id - the entry's bridge domain id */ -define l2fib_flush_bd +autoreply define l2fib_flush_bd { u32 client_index; u32 context; u32 bd_id; }; -/** \brief L2 FIB flush bridge domain entries response - @param context - sender context, to match reply w/ request - @param retval - return code for the request -*/ -define l2fib_flush_bd_reply -{ - u32 context; - i32 retval; -}; - /** \brief L2 FIB flush interface entries @param client_index - opaque cookie to identify the sender @param context - sender context, to match reply w/ request @param bd_id - the entry's bridge domain id */ -define l2fib_flush_int +autoreply define l2fib_flush_int { u32 client_index; u32 context; u32 sw_if_index; }; -/** \brief L2 FIB flush interface entries response - @param context - sender context, to match reply w/ request - @param retval - return code for the request -*/ -define l2fib_flush_int_reply -{ - u32 context; - i32 retval; -}; - /** \brief L2 FIB add entry request @param client_index - opaque cookie to identify the sender @param context - sender context, to match reply w/ request @@ -140,7 +110,7 @@ define l2fib_flush_int_reply @param static_mac - @param filter_mac - */ -define l2fib_add_del +autoreply define l2fib_add_del { u32 client_index; u32 context; @@ -153,16 +123,6 @@ define l2fib_add_del u8 bvi_mac; }; -/** \brief L2 FIB add entry response - @param context - sender context, to match reply w/ request - @param retval - return code for the add l2fib entry request -*/ -define l2fib_add_del_reply -{ - u32 context; - i32 retval; -}; - /** \brief Set L2 flags request !!! TODO - need more info, feature bits in l2_input.h @param client_index - opaque cookie to identify the sender @param context - sender context, to match reply w/ request @@ -196,7 +156,7 @@ define l2_flags_reply @param bd_id - the bridge domain to create @param mac_age - mac aging time in min, 0 for disabled */ -define bridge_domain_set_mac_age +autoreply define bridge_domain_set_mac_age { u32 client_index; u32 context; @@ -204,16 +164,6 @@ define bridge_domain_set_mac_age u8 mac_age; }; -/** \brief Set bridge domain response - @param context - sender context, to match reply w/ request - @param retval - return code for the set l2 bits request -*/ -define bridge_domain_set_mac_age_reply -{ - u32 context; - i32 retval; -}; - /** \brief L2 bridge domain add or delete request @param client_index - opaque cookie to identify the sender @param context - sender context, to match reply w/ request @@ -226,7 +176,7 @@ define bridge_domain_set_mac_age_reply @param mac_age - mac aging time in min, 0 for disabled @param is_add - add or delete flag */ -define bridge_domain_add_del +autoreply define bridge_domain_add_del { u32 client_index; u32 context; @@ -240,16 +190,6 @@ define bridge_domain_add_del u8 is_add; }; -/** \brief L2 bridge domain add or delete response - @param context - sender context, to match reply w/ request - @param retval - return code for the set bridge flags request -*/ -define bridge_domain_add_del_reply -{ - u32 context; - i32 retval; -}; - /** \brief L2 bridge domain request operational state details @param client_index - opaque cookie to identify the sender @param context - sender context, to match reply w/ request @@ -337,7 +277,7 @@ define bridge_flags_reply @param tag1 - Needed for any push or translate vtr op @param tag2 - Needed for any push 2 or translate x-2 vtr ops */ -define l2_interface_vlan_tag_rewrite +autoreply define l2_interface_vlan_tag_rewrite { u32 client_index; u32 context; @@ -348,16 +288,6 @@ define l2_interface_vlan_tag_rewrite u32 tag2; // second pushed tag }; -/** \brief L2 interface vlan tag rewrite response - @param context - sender context, to match reply w/ request - @param retval - return code for the request -*/ -define l2_interface_vlan_tag_rewrite_reply -{ - u32 context; - i32 retval; -}; - /** \brief L2 interface pbb tag rewrite configure request @param client_index - opaque cookie to identify the sender @param context - sender context, to match reply w/ request @@ -370,7 +300,7 @@ define l2_interface_vlan_tag_rewrite_reply @param b_vlanid - B-tag vlanid, needed for any push or translate qinq vtr op @param i_sid - I-tag service id, needed for any push or translate qinq vtr op */ -define l2_interface_pbb_tag_rewrite +autoreply define l2_interface_pbb_tag_rewrite { u32 client_index; u32 context; @@ -383,16 +313,6 @@ define l2_interface_pbb_tag_rewrite u32 i_sid; }; -/** \brief L2 interface pbb tag rewrite response - @param context - sender context, to match reply w/ request - @param retval - return code for the request -*/ -define l2_interface_pbb_tag_rewrite_reply -{ - u32 context; - i32 retval; -}; - /* * Local Variables: * eval: (c-set-style "gnu") diff --git a/src/vnet/l2tp/l2tp.api b/src/vnet/l2tp/l2tp.api index 5a5a5a48..4587a807 100644 --- a/src/vnet/l2tp/l2tp.api +++ b/src/vnet/l2tp/l2tp.api @@ -52,7 +52,7 @@ define l2tpv3_create_tunnel_reply u32 sw_if_index; }; -define l2tpv3_set_tunnel_cookies +autoreply define l2tpv3_set_tunnel_cookies { u32 client_index; u32 context; @@ -61,16 +61,6 @@ define l2tpv3_set_tunnel_cookies u64 new_remote_cookie; }; -/** \brief L2TP tunnel set cookies response - @param context - sender context, to match reply w/ request - @param retval - return code for the request -*/ -define l2tpv3_set_tunnel_cookies_reply -{ - u32 context; - i32 retval; -}; - define sw_if_l2tpv3_tunnel_details { u32 context; @@ -91,7 +81,7 @@ define sw_if_l2tpv3_tunnel_dump u32 context; }; -define l2tpv3_interface_enable_disable +autoreply define l2tpv3_interface_enable_disable { u32 client_index; u32 context; @@ -99,13 +89,7 @@ define l2tpv3_interface_enable_disable u32 sw_if_index; }; -define l2tpv3_interface_enable_disable_reply -{ - u32 context; - i32 retval; -}; - -define l2tpv3_set_lookup_key +autoreply define l2tpv3_set_lookup_key { u32 client_index; u32 context; @@ -113,12 +97,6 @@ define l2tpv3_set_lookup_key u8 key; }; -define l2tpv3_set_lookup_key_reply -{ - u32 context; - i32 retval; -}; - /* * Local Variables: * eval: (c-set-style "gnu") diff --git a/src/vnet/lisp-cp/lisp.api b/src/vnet/lisp-cp/lisp.api index a50a5ccb..8bed71b3 100644 --- a/src/vnet/lisp-cp/lisp.api +++ b/src/vnet/lisp-cp/lisp.api @@ -59,7 +59,7 @@ define lisp_add_del_locator_set_reply @param priority - priority of the lisp locator @param weight - weight of the lisp locator */ -define lisp_add_del_locator +autoreply define lisp_add_del_locator { u32 client_index; u32 context; @@ -70,16 +70,6 @@ define lisp_add_del_locator u8 weight; }; -/** \brief Reply for locator add/del - @param context - returned sender context, to match reply w/ request - @param retval - return code -*/ -define lisp_add_del_locator_reply -{ - u32 context; - i32 retval; -}; - /** \brief add or delete lisp eid-table @param client_index - opaque cookie to identify the sender @param context - sender context, to match reply w/ request @@ -98,7 +88,7 @@ define lisp_add_del_locator_reply HMAC_SHA_256_128 2 @param key - secret key */ -define lisp_add_del_local_eid +autoreply define lisp_add_del_local_eid { u32 client_index; u32 context; @@ -112,16 +102,6 @@ define lisp_add_del_local_eid u8 key[64]; }; -/** \brief Reply for local_eid add/del - @param context - returned sender context, to match reply w/ request - @param retval - return code -*/ -define lisp_add_del_local_eid_reply -{ - u32 context; - i32 retval; -}; - /** \brief Add/delete map server @param client_index - opaque cookie to identify the sender @param context - sender context, to match reply w/ request @@ -129,7 +109,7 @@ define lisp_add_del_local_eid_reply @param is_ipv6 - if non-zero the address is ipv6, else ipv4 @param ip_address - map server IP address */ -define lisp_add_del_map_server +autoreply define lisp_add_del_map_server { u32 client_index; u32 context; @@ -138,16 +118,6 @@ define lisp_add_del_map_server u8 ip_address[16]; }; -/** \brief Reply for lisp_add_del_map_server - @param context - returned sender context, to match reply w/ request - @param retval - return code -*/ -define lisp_add_del_map_server_reply -{ - u32 context; - i32 retval; -}; - /** \brief add or delete map-resolver @param client_index - opaque cookie to identify the sender @param context - sender context, to match reply w/ request @@ -155,7 +125,7 @@ define lisp_add_del_map_server_reply @param is_ipv6 - if non-zero the address is ipv6, else ipv4 @param ip_address - array of address bytes */ -define lisp_add_del_map_resolver +autoreply define lisp_add_del_map_resolver { u32 client_index; u32 context; @@ -164,45 +134,25 @@ define lisp_add_del_map_resolver u8 ip_address[16]; }; -/** \brief Reply for map_resolver add/del - @param context - returned sender context, to match reply w/ request - @param retval - return code -*/ -define lisp_add_del_map_resolver_reply -{ - u32 context; - i32 retval; -}; - /** \brief enable or disable LISP feature @param client_index - opaque cookie to identify the sender @param context - sender context, to match reply w/ request @param is_en - enable protocol if non-zero, else disable */ -define lisp_enable_disable +autoreply define lisp_enable_disable { u32 client_index; u32 context; u8 is_en; }; -/** \brief Reply for gpe enable/disable - @param context - returned sender context, to match reply w/ request - @param retval - return code -*/ -define lisp_enable_disable_reply -{ - u32 context; - i32 retval; -}; - /** \brief configure or disable LISP PITR node @param client_index - opaque cookie to identify the sender @param context - sender context, to match reply w/ request @param ls_name - locator set name @param is_add - add locator set if non-zero, else disable pitr */ -define lisp_pitr_set_locator_set +autoreply define lisp_pitr_set_locator_set { u32 client_index; u32 context; @@ -210,16 +160,6 @@ define lisp_pitr_set_locator_set u8 ls_name[64]; }; -/** \brief Reply for lisp_pitr_set_locator_set - @param context - returned sender context, to match reply w/ request - @param retval - return code -*/ -define lisp_pitr_set_locator_set_reply -{ - u32 context; - i32 retval; -}; - /** \brief configure or disable use of PETR @param client_index - opaque cookie to identify the sender @param context - sender context, to match reply w/ request @@ -227,7 +167,7 @@ define lisp_pitr_set_locator_set_reply @param address - PETR IP address @param is_add - add locator set if non-zero, else disable pitr */ -define lisp_use_petr +autoreply define lisp_use_petr { u32 client_index; u32 context; @@ -236,16 +176,6 @@ define lisp_use_petr u8 is_add; }; -/** \brief Reply for lisp_pitr_set_locator_set - @param context - returned sender context, to match reply w/ request - @param retval - return code -*/ -define lisp_use_petr_reply -{ - u32 context; - i32 retval; -}; - /** \brief Request for LISP PETR status @param client_index - opaque cookie to identify the sender @param context - sender context, to match reply w/ request @@ -298,45 +228,25 @@ define show_lisp_rloc_probe_state_reply @param context - sender context, to match reply w/ request @param is_enable - enable if non-zero; disable otherwise */ -define lisp_rloc_probe_enable_disable +autoreply define lisp_rloc_probe_enable_disable { u32 client_index; u32 context; u8 is_enabled; }; -/** \brief Reply for lisp_rloc_probe_enable_disable - @param context - returned sender context, to match reply w/ request - @param retval - return code -*/ -define lisp_rloc_probe_enable_disable_reply -{ - u32 context; - i32 retval; -}; - /** \brief enable/disable LISP map-register @param client_index - opaque cookie to identify the sender @param context - sender context, to match reply w/ request @param is_enable - enable if non-zero; disable otherwise */ -define lisp_map_register_enable_disable +autoreply define lisp_map_register_enable_disable { u32 client_index; u32 context; u8 is_enabled; }; -/** \brief Reply for lisp_map_register_enable_disable - @param context - returned sender context, to match reply w/ request - @param retval - return code -*/ -define lisp_map_register_enable_disable_reply -{ - u32 context; - i32 retval; -}; - /** \brief Get state of LISP map-register @param client_index - opaque cookie to identify the sender @param context - sender context, to match reply w/ request @@ -366,23 +276,13 @@ define show_lisp_map_register_state_reply 0 - destination only 1 - source/destaination */ -define lisp_map_request_mode +autoreply define lisp_map_request_mode { u32 client_index; u32 context; u8 mode; }; -/** \brief Reply for lisp_map_request_mode - @param context - returned sender context, to match reply w/ request - @param retval - return code -*/ -define lisp_map_request_mode_reply -{ - u32 context; - i32 retval; -}; - /** \brief Request for LISP map-request mode @param client_index - opaque cookie to identify the sender @param context - sender context, to match reply w/ request @@ -430,7 +330,7 @@ typeonly manual_endian manual_print define remote_locator @param rloc_num - number of remote locators @param rlocs - remote locator records */ -manual_print manual_endian define lisp_add_del_remote_mapping +autoreply manual_print manual_endian define lisp_add_del_remote_mapping { u32 client_index; u32 context; @@ -448,16 +348,6 @@ manual_print manual_endian define lisp_add_del_remote_mapping vl_api_remote_locator_t rlocs[rloc_num]; }; -/** \brief Reply for lisp_add_del_remote_mapping - @param context - returned sender context, to match reply w/ request - @param retval - return code -*/ -define lisp_add_del_remote_mapping_reply -{ - u32 context; - i32 retval; -}; - /** \brief add or delete LISP adjacency adjacency @param client_index - opaque cookie to identify the sender @param context - sender context, to match reply w/ request @@ -470,7 +360,7 @@ define lisp_add_del_remote_mapping_reply @param reid - remote EID @param leid - local EID */ -define lisp_add_del_adjacency +autoreply define lisp_add_del_adjacency { u32 client_index; u32 context; @@ -483,23 +373,13 @@ define lisp_add_del_adjacency u8 leid_len; }; -/** \brief Reply for lisp_add_del_adjacency - @param context - returned sender context, to match reply w/ request - @param retval - return code -*/ -define lisp_add_del_adjacency_reply -{ - u32 context; - i32 retval; -}; - /** \brief add or delete map request itr rlocs @param client_index - opaque cookie to identify the sender @param context - sender context, to match reply w/ request @param is_add - add address if non-zero, else delete @param locator_set_name - locator set name */ -define lisp_add_del_map_request_itr_rlocs +autoreply define lisp_add_del_map_request_itr_rlocs { u32 client_index; u32 context; @@ -512,12 +392,6 @@ define lisp_add_del_map_request_itr_rlocs @param retval - return code */ -define lisp_add_del_map_request_itr_rlocs_reply -{ - u32 context; - i32 retval; -}; - /** \brief map/unmap vni/bd_index to vrf @param client_index - opaque cookie to identify the sender @param context - sender context, to match reply w/ request @@ -525,7 +399,7 @@ define lisp_add_del_map_request_itr_rlocs_reply @param dp_table - virtual network id/bridge domain index @param vrf - vrf */ -define lisp_eid_table_add_del_map +autoreply define lisp_eid_table_add_del_map { u32 client_index; u32 context; @@ -535,16 +409,6 @@ define lisp_eid_table_add_del_map u8 is_l2; }; -/** \brief Reply for lisp_eid_table_add_del_map - @param context - returned sender context, to match reply w/ request - @param retval - return code -*/ -define lisp_eid_table_add_del_map_reply -{ - u32 context; - i32 retval; -}; - /** \brief Request for map lisp locator status @param client_index - opaque cookie to identify the sender @param context - sender context, to match reply w/ request diff --git a/src/vnet/lisp-cp/one.api b/src/vnet/lisp-cp/one.api index ca82f694..2fa1edf6 100644 --- a/src/vnet/lisp-cp/one.api +++ b/src/vnet/lisp-cp/one.api @@ -59,7 +59,7 @@ define one_add_del_locator_set_reply @param priority - priority of the locator @param weight - weight of the locator */ -define one_add_del_locator +autoreply define one_add_del_locator { u32 client_index; u32 context; @@ -70,16 +70,6 @@ define one_add_del_locator u8 weight; }; -/** \brief Reply for locator add/del - @param context - returned sender context, to match reply w/ request - @param retval - return code -*/ -define one_add_del_locator_reply -{ - u32 context; - i32 retval; -}; - /** \brief add or delete ONE eid-table @param client_index - opaque cookie to identify the sender @param context - sender context, to match reply w/ request @@ -98,7 +88,7 @@ define one_add_del_locator_reply HMAC_SHA_256_128 2 @param key - secret key */ -define one_add_del_local_eid +autoreply define one_add_del_local_eid { u32 client_index; u32 context; @@ -112,16 +102,6 @@ define one_add_del_local_eid u8 key[64]; }; -/** \brief Reply for local_eid add/del - @param context - returned sender context, to match reply w/ request - @param retval - return code -*/ -define one_add_del_local_eid_reply -{ - u32 context; - i32 retval; -}; - /** \brief Add/delete map server @param client_index - opaque cookie to identify the sender @param context - sender context, to match reply w/ request @@ -129,7 +109,7 @@ define one_add_del_local_eid_reply @param is_ipv6 - if non-zero the address is ipv6, else ipv4 @param ip_address - map server IP address */ -define one_add_del_map_server +autoreply define one_add_del_map_server { u32 client_index; u32 context; @@ -138,16 +118,6 @@ define one_add_del_map_server u8 ip_address[16]; }; -/** \brief Reply for one_add_del_map_server - @param context - returned sender context, to match reply w/ request - @param retval - return code -*/ -define one_add_del_map_server_reply -{ - u32 context; - i32 retval; -}; - /** \brief add or delete map-resolver @param client_index - opaque cookie to identify the sender @param context - sender context, to match reply w/ request @@ -155,7 +125,7 @@ define one_add_del_map_server_reply @param is_ipv6 - if non-zero the address is ipv6, else ipv4 @param ip_address - array of address bytes */ -define one_add_del_map_resolver +autoreply define one_add_del_map_resolver { u32 client_index; u32 context; @@ -164,45 +134,25 @@ define one_add_del_map_resolver u8 ip_address[16]; }; -/** \brief Reply for map_resolver add/del - @param context - returned sender context, to match reply w/ request - @param retval - return code -*/ -define one_add_del_map_resolver_reply -{ - u32 context; - i32 retval; -}; - /** \brief enable or disable ONE feature @param client_index - opaque cookie to identify the sender @param context - sender context, to match reply w/ request @param is_en - enable protocol if non-zero, else disable */ -define one_enable_disable +autoreply define one_enable_disable { u32 client_index; u32 context; u8 is_en; }; -/** \brief Reply for gpe enable/disable - @param context - returned sender context, to match reply w/ request - @param retval - return code -*/ -define one_enable_disable_reply -{ - u32 context; - i32 retval; -}; - /** \brief configure or disable ONE PITR node @param client_index - opaque cookie to identify the sender @param context - sender context, to match reply w/ request @param ls_name - locator set name @param is_add - add locator set if non-zero, else disable pitr */ -define one_pitr_set_locator_set +autoreply define one_pitr_set_locator_set { u32 client_index; u32 context; @@ -210,16 +160,6 @@ define one_pitr_set_locator_set u8 ls_name[64]; }; -/** \brief Reply for one_pitr_set_locator_set - @param context - returned sender context, to match reply w/ request - @param retval - return code -*/ -define one_pitr_set_locator_set_reply -{ - u32 context; - i32 retval; -}; - /** \brief configure or disable use of PETR @param client_index - opaque cookie to identify the sender @param context - sender context, to match reply w/ request @@ -227,7 +167,7 @@ define one_pitr_set_locator_set_reply @param address - PETR IP address @param is_add - add locator set if non-zero, else disable PETR */ -define one_use_petr +autoreply define one_use_petr { u32 client_index; u32 context; @@ -236,16 +176,6 @@ define one_use_petr u8 is_add; }; -/** \brief Reply for one_use_petr - @param context - returned sender context, to match reply w/ request - @param retval - return code -*/ -define one_use_petr_reply -{ - u32 context; - i32 retval; -}; - /** \brief Request for ONE PETR status @param client_index - opaque cookie to identify the sender @param context - sender context, to match reply w/ request @@ -298,45 +228,25 @@ define show_one_rloc_probe_state_reply @param context - sender context, to match reply w/ request @param is_enable - enable if non-zero; disable otherwise */ -define one_rloc_probe_enable_disable +autoreply define one_rloc_probe_enable_disable { u32 client_index; u32 context; u8 is_enabled; }; -/** \brief Reply for one_rloc_probe_enable_disable - @param context - returned sender context, to match reply w/ request - @param retval - return code -*/ -define one_rloc_probe_enable_disable_reply -{ - u32 context; - i32 retval; -}; - /** \brief enable/disable ONE map-register @param client_index - opaque cookie to identify the sender @param context - sender context, to match reply w/ request @param is_enable - enable if non-zero; disable otherwise */ -define one_map_register_enable_disable +autoreply define one_map_register_enable_disable { u32 client_index; u32 context; u8 is_enabled; }; -/** \brief Reply for one_map_register_enable_disable - @param context - returned sender context, to match reply w/ request - @param retval - return code -*/ -define one_map_register_enable_disable_reply -{ - u32 context; - i32 retval; -}; - /** \brief Get state of ONE map-register @param client_index - opaque cookie to identify the sender @param context - sender context, to match reply w/ request @@ -366,23 +276,13 @@ define show_one_map_register_state_reply 0 - destination only 1 - source/destaination */ -define one_map_request_mode +autoreply define one_map_request_mode { u32 client_index; u32 context; u8 mode; }; -/** \brief Reply for one_map_request_mode - @param context - returned sender context, to match reply w/ request - @param retval - return code -*/ -define one_map_request_mode_reply -{ - u32 context; - i32 retval; -}; - /** \brief Request for ONE map-request mode @param client_index - opaque cookie to identify the sender @param context - sender context, to match reply w/ request @@ -430,7 +330,7 @@ typeonly manual_endian manual_print define one_remote_locator @param rloc_num - number of remote locators @param rlocs - remote locator records */ -manual_print manual_endian define one_add_del_remote_mapping +autoreply manual_print manual_endian define one_add_del_remote_mapping { u32 client_index; u32 context; @@ -448,16 +348,6 @@ manual_print manual_endian define one_add_del_remote_mapping vl_api_one_remote_locator_t rlocs[rloc_num]; }; -/** \brief Reply for one_add_del_remote_mapping - @param context - returned sender context, to match reply w/ request - @param retval - return code -*/ -define one_add_del_remote_mapping_reply -{ - u32 context; - i32 retval; -}; - /** \brief add or delete ONE adjacency adjacency @param client_index - opaque cookie to identify the sender @param context - sender context, to match reply w/ request @@ -470,7 +360,7 @@ define one_add_del_remote_mapping_reply @param reid - remote EID @param leid - local EID */ -define one_add_del_adjacency +autoreply define one_add_del_adjacency { u32 client_index; u32 context; @@ -483,23 +373,13 @@ define one_add_del_adjacency u8 leid_len; }; -/** \brief Reply for one_add_del_adjacency - @param context - returned sender context, to match reply w/ request - @param retval - return code -*/ -define one_add_del_adjacency_reply -{ - u32 context; - i32 retval; -}; - /** \brief add or delete map request itr rlocs @param client_index - opaque cookie to identify the sender @param context - sender context, to match reply w/ request @param is_add - add address if non-zero, else delete @param locator_set_name - locator set name */ -define one_add_del_map_request_itr_rlocs +autoreply define one_add_del_map_request_itr_rlocs { u32 client_index; u32 context; @@ -507,17 +387,6 @@ define one_add_del_map_request_itr_rlocs u8 locator_set_name[64]; }; -/** \brief Reply for one_add_del_map_request_itr_rlocs - @param context - returned sender context, to match reply w/ request - @param retval - return code -*/ - -define one_add_del_map_request_itr_rlocs_reply -{ - u32 context; - i32 retval; -}; - /** \brief map/unmap vni/bd_index to vrf @param client_index - opaque cookie to identify the sender @param context - sender context, to match reply w/ request @@ -525,7 +394,7 @@ define one_add_del_map_request_itr_rlocs_reply @param dp_table - virtual network id/bridge domain index @param vrf - vrf */ -define one_eid_table_add_del_map +autoreply define one_eid_table_add_del_map { u32 client_index; u32 context; @@ -535,16 +404,6 @@ define one_eid_table_add_del_map u8 is_l2; }; -/** \brief Reply for one_eid_table_add_del_map - @param context - returned sender context, to match reply w/ request - @param retval - return code -*/ -define one_eid_table_add_del_map_reply -{ - u32 context; - i32 retval; -}; - /** \brief Request for map one locator status @param client_index - opaque cookie to identify the sender @param context - sender context, to match reply w/ request @@ -901,31 +760,19 @@ define one_stats_details u32 bytes; }; -define one_stats_flush +autoreply define one_stats_flush { u32 client_index; u32 context; }; -define one_stats_flush_reply -{ - u32 context; - i32 retval; -}; - -define one_stats_enable_disable +autoreply define one_stats_enable_disable { u32 client_index; u32 context; u8 is_en; }; -define one_stats_enable_disable_reply -{ - u32 context; - i32 retval; -}; - define show_one_stats_enable_disable { u32 client_index; diff --git a/src/vnet/lisp-gpe/lisp_gpe.api b/src/vnet/lisp-gpe/lisp_gpe.api index 43a6a6cd..f79d18c1 100644 --- a/src/vnet/lisp-gpe/lisp_gpe.api +++ b/src/vnet/lisp-gpe/lisp_gpe.api @@ -43,7 +43,7 @@ typeonly manual_print manual_endian define gpe_locator @param loc_num - number of locators @param locs - array of remote locators */ -manual_print manual_endian define gpe_add_del_fwd_entry +autoreply manual_print manual_endian define gpe_add_del_fwd_entry { u32 client_index; u32 context; @@ -60,44 +60,24 @@ manual_print manual_endian define gpe_add_del_fwd_entry vl_api_gpe_locator_t locs[loc_num]; }; -/** \brief Reply for gpe_fwd_entry add/del - @param context - returned sender context, to match reply w/ request - @param retval - return code -*/ -define gpe_add_del_fwd_entry_reply -{ - u32 context; - i32 retval; -}; - /** \brief enable or disable gpe protocol @param client_index - opaque cookie to identify the sender @param context - sender context, to match reply w/ request @param is_en - enable protocol if non-zero, else disable */ -define gpe_enable_disable +autoreply define gpe_enable_disable { u32 client_index; u32 context; u8 is_en; }; -/** \brief Reply for gpe enable/disable - @param context - returned sender context, to match reply w/ request - @param retval - return code -*/ -define gpe_enable_disable_reply -{ - u32 context; - i32 retval; -}; - /** \brief add or delete gpe_iface @param client_index - opaque cookie to identify the sender @param context - sender context, to match reply w/ request @param is_add - add address if non-zero, else delete */ -define gpe_add_del_iface +autoreply define gpe_add_del_iface { u32 client_index; u32 context; @@ -107,16 +87,6 @@ define gpe_add_del_iface u32 vni; }; -/** \brief Reply for gpe_iface add/del - @param context - returned sender context, to match reply w/ request - @param retval - return code -*/ -define gpe_add_del_iface_reply -{ - u32 context; - i32 retval; -}; - define gpe_fwd_entries_get { u32 client_index; @@ -163,23 +133,13 @@ manual_endian manual_print define gpe_fwd_entry_path_details @param context - sender context, to match reply w/ request @param mode - LISP (value 0) or VXLAN (value 1) */ -define gpe_set_encap_mode +autoreply define gpe_set_encap_mode { u32 client_index; u32 context; u8 mode; }; -/** \brief Reply for set_encap_mode - @param context - returned sender context, to match reply w/ request - @param retval - return code -*/ -define gpe_set_encap_mode_reply -{ - u32 context; - i32 retval; -}; - /** \brief get GPE encapsulation mode @param client_index - opaque cookie to identify the sender @param context - sender context, to match reply w/ request diff --git a/src/vnet/map/map.api b/src/vnet/map/map.api index 4e4be85e..d68f13f0 100644 --- a/src/vnet/map/map.api +++ b/src/vnet/map/map.api @@ -62,22 +62,13 @@ define map_add_domain_reply @param context - sender context, to match reply w/ request @param index - MAP Domain index */ -define map_del_domain +autoreply define map_del_domain { u32 client_index; u32 context; u32 index; }; -/** \brief Reply for MAP domain del - @param context - returned sender context, to match reply w/ request - @param retval - return code -*/ -define map_del_domain_reply -{ - u32 context; - i32 retval; -}; /** \brief Add or Delete MAP rule from a domain (Only used for shared IPv4 per subscriber) @param client_index - opaque cookie to identify the sender @@ -87,7 +78,7 @@ define map_del_domain_reply @param ip6_dst - MAP CE IPv6 address @param psid - Rule PSID */ -define map_add_del_rule +autoreply define map_add_del_rule { u32 client_index; u32 context; @@ -97,15 +88,6 @@ define map_add_del_rule u16 psid; }; -/** \brief Reply for MAP rule add/del - @param context - returned sender context, to match reply w/ request - @param retval - return code -*/ -define map_add_del_rule_reply -{ - u32 context; - i32 retval; -}; /** \brief Get list of map domains @param client_index - opaque cookie to identify the sender diff --git a/src/vnet/mpls/mpls.api b/src/vnet/mpls/mpls.api index a1e1270a..c8a3ffb7 100644 --- a/src/vnet/mpls/mpls.api +++ b/src/vnet/mpls/mpls.api @@ -26,7 +26,7 @@ @param mb_address_length - Length of IP prefix @param mb_address[16] - IP prefix/ */ -define mpls_ip_bind_unbind +autoreply define mpls_ip_bind_unbind { u32 client_index; u32 context; @@ -40,16 +40,6 @@ define mpls_ip_bind_unbind u8 mb_address[16]; }; -/** \brief Reply for MPLS IP bind/unbind request - @param context - returned sender context, to match reply w/ request - @param retval - return code -*/ -define mpls_ip_bind_unbind_reply -{ - u32 context; - i32 retval; -}; - /** \brief MPLS tunnel Add / del route @param client_index - opaque cookie to identify the sender @param context - sender context, to match reply w/ request @@ -172,7 +162,7 @@ manual_endian manual_print define mpls_tunnel_details @param mr_next_hop_out_label_stack - the next-hop output label stack, outer most first @param next_hop_via_label - The next-hop is a resolved via a local label */ -define mpls_route_add_del +autoreply define mpls_route_add_del { u32 client_index; u32 context; @@ -199,16 +189,6 @@ define mpls_route_add_del u32 mr_next_hop_out_label_stack[mr_next_hop_n_out_labels]; }; -/** \brief Reply for MPLS route add / del request - @param context - returned sender context, to match reply w/ request - @param retval - return code -*/ -define mpls_route_add_del_reply -{ - u32 context; - i32 retval; -}; - /** \brief Dump MPLS fib table @param client_index - opaque cookie to identify the sender */ @@ -240,4 +220,4 @@ manual_endian manual_print define mpls_fib_details * eval: (c-set-style "gnu") * End: */ - \ No newline at end of file + diff --git a/src/vnet/session/session.api b/src/vnet/session/session.api index e207e46f..4aef09da 100644 --- a/src/vnet/session/session.api +++ b/src/vnet/session/session.api @@ -49,26 +49,17 @@ define application_attach_reply { @param client_index - opaque cookie to identify the sender @param context - sender context, to match reply w/ request */ - define application_detach { +autoreply define application_detach { u32 client_index; u32 context; }; - /** \brief detach reply - @param context - sender context, to match reply w/ request - @param retval - return code for the request -*/ -define application_detach_reply { - u32 context; - i32 retval; -}; - /** \brief vpp->client, please map an additional shared memory segment @param client_index - opaque cookie to identify the sender @param context - sender context, to match reply w/ request @param segment_name - */ -define map_another_segment { +autoreply define map_another_segment { u32 client_index; u32 context; u32 segment_size; @@ -83,7 +74,7 @@ define map_another_segment { "tcp://::/0/80" [ipv6] etc. @param options - socket options, fifo sizes, etc. */ -define bind_uri { +autoreply define bind_uri { u32 client_index; u32 context; u32 accept_cookie; @@ -97,7 +88,7 @@ define bind_uri { "tcp://::/0/80" [ipv6], etc. @param options - socket options, fifo sizes, etc. */ -define unbind_uri { +autoreply define unbind_uri { u32 client_index; u32 context; u8 uri[128]; @@ -122,24 +113,6 @@ define connect_uri { u64 options[16]; }; -/** \brief Bind reply - @param context - sender context, to match reply w/ request - @param retval - return code for the request -*/ -define bind_uri_reply { - u32 context; - i32 retval; -}; - -/** \brief unbind reply - @param context - sender context, to match reply w/ request - @param retval - return code for the request -*/ -define unbind_uri_reply { - u32 context; - i32 retval; -}; - /** \brief vpp->client, connect reply @param context - sender context, to match reply w/ request @param retval - return code for the request @@ -165,15 +138,6 @@ define connect_uri_reply { u8 segment_name[128]; }; -/** \brief client->vpp - @param context - sender context, to match reply w/ request - @param retval - return code for the request -*/ -define map_another_segment_reply { - u32 context; - i32 retval; -}; - /** \brief vpp->client, accept this session @param context - sender context, to match reply w/ request @param listener_handle - tells client which listener this pertains to @@ -290,7 +254,7 @@ define bind_sock { @param context - sender context, to match reply w/ request @param handle - bind handle obtained from bind reply */ -define unbind_sock { +autoreply define unbind_sock { u32 client_index; u32 context; u64 handle; @@ -339,15 +303,6 @@ define bind_sock_reply { u8 segment_name[128]; }; -/** \brief unbind reply - @param context - sender context, to match reply w/ request - @param retval - return code for the request -*/ -define unbind_sock_reply { - u32 context; - i32 retval; -}; - /** \brief vpp/server->client, connect reply @param context - sender context, to match reply w/ request @param retval - return code for the request @@ -378,23 +333,14 @@ define connect_sock_reply { @param context - sender context, to match reply w/ request @param is_enable - disable session layer if 0, enable otherwise */ -define session_enable_disable { +autoreply define session_enable_disable { u32 client_index; u32 context; u8 is_enable; }; -/** \brief Reply for session enable/disable - @param context - returned sender context, to match reply w/ request - @param retval - return code -*/ -define session_enable_disable_reply { - u32 context; - i32 retval; -}; - /* * Local Variables: * eval: (c-set-style "gnu") * End: - */ \ No newline at end of file + */ diff --git a/src/vnet/span/span.api b/src/vnet/span/span.api index 4babdd83..914fd8d0 100644 --- a/src/vnet/span/span.api +++ b/src/vnet/span/span.api @@ -21,7 +21,7 @@ @param sw_if_index_to - interface where the traffic is mirrored @param state - 0 = disabled, 1 = rx enabled, 2 = tx enabled, 3 tx & rx enabled */ -define sw_interface_span_enable_disable { +autoreply define sw_interface_span_enable_disable { u32 client_index; u32 context; u32 sw_if_index_from; @@ -29,14 +29,6 @@ define sw_interface_span_enable_disable { u8 state; }; -/** \brief Reply to SPAN enable/disable request - @param context - sender context which was passed in the request -*/ -define sw_interface_span_enable_disable_reply { - u32 context; - i32 retval; -}; - /** \brief SPAN dump request @param client_index - opaque cookie to identify the sender @param context - sender context, to match reply w/ request diff --git a/src/vnet/sr/sr.api b/src/vnet/sr/sr.api index 5feadcb0..9e900741 100644 --- a/src/vnet/sr/sr.api +++ b/src/vnet/sr/sr.api @@ -25,7 +25,7 @@ @param fib_table FIB table in which we should install the localsid entry @param nh_addr Next Hop IPv4/IPv6 address. Only for L2/L3 xconnect. */ -define sr_localsid_add_del +autoreply define sr_localsid_add_del { u32 client_index; u32 context; @@ -39,16 +39,6 @@ define sr_localsid_add_del u8 nh_addr[16]; }; -/** \brief IPv6 SR LocalSID add/del request response - @param context - sender context, to match reply w/ request - @param retval - return value for request -*/ -define sr_localsid_add_del_reply -{ - u32 context; - i32 retval; -}; - /** \brief IPv6 SR policy add @param client_index - opaque cookie to identify the sender @param context - sender context, to match reply w/ request @@ -59,7 +49,7 @@ define sr_localsid_add_del_reply @param fib_table is the VRF where to install the FIB entry for the BSID @param segments is a vector of IPv6 address composing the segment list */ -define sr_policy_add +autoreply define sr_policy_add { u32 client_index; u32 context; @@ -72,16 +62,6 @@ define sr_policy_add u8 segments[0]; }; -/** \brief IPv6 SR Policy add request response - @param context - sender context, to match reply w/ request - @param retval - return value for request -*/ -define sr_policy_add_reply -{ - u32 context; - i32 retval; -}; - /** \brief IPv6 SR policy modification @param client_index - opaque cookie to identify the sender @param context - sender context, to match reply w/ request @@ -94,7 +74,7 @@ define sr_policy_add_reply @param weight is the weight of the sid list. optional. @param is_encap Mode. Encapsulation or SRH insertion. */ -define sr_policy_mod +autoreply define sr_policy_mod { u32 client_index; u32 context; @@ -108,23 +88,13 @@ define sr_policy_mod u8 segments[0]; }; -/** \brief IPv6 SR Policy modification request response - @param context - sender context, to match reply w/ request - @param retval - return value for request -*/ -define sr_policy_mod_reply -{ - u32 context; - i32 retval; -}; - /** \brief IPv6 SR policy deletion @param client_index - opaque cookie to identify the sender @param context - sender context, to match reply w/ request @param bsid is the bindingSID of the SR Policy @param index is the index of the SR policy */ -define sr_policy_del +autoreply define sr_policy_del { u32 client_index; u32 context; @@ -132,16 +102,6 @@ define sr_policy_del u32 sr_policy_index; }; -/** \brief IPv6 SR Policy deletion request response - @param context - sender context, to match reply w/ request - @param retval - return value for request -*/ -define sr_policy_del_reply -{ - u32 context; - i32 retval; -}; - /** \brief IPv6 SR steering add/del @param client_index - opaque cookie to identify the sender @param context - sender context, to match reply w/ request @@ -154,7 +114,7 @@ define sr_policy_del_reply @param sw_if_index is the incoming interface for L2 traffic @param traffic_type describes the type of traffic */ -define sr_steering_add_del +autoreply define sr_steering_add_del { u32 client_index; u32 context; @@ -168,16 +128,6 @@ define sr_steering_add_del u8 traffic_type; }; -/** \brief IPv6 SR steering add/del request response - @param context - sender context, to match reply w/ request - @param retval - return value for request -*/ -define sr_steering_add_del_reply -{ - u32 context; - i32 retval; -}; - /** \brief Dump the list of SR LocalSIDs @param client_index - opaque cookie to identify the sender @param context - sender context, to match reply w/ request diff --git a/src/vnet/unix/tap.api b/src/vnet/unix/tap.api index 1fd0bb09..d9fba371 100644 --- a/src/vnet/unix/tap.api +++ b/src/vnet/unix/tap.api @@ -93,23 +93,13 @@ define tap_modify_reply @param context - sender context, to match reply w/ request @param sw_if_index - interface index of existing tap interface */ -define tap_delete +autoreply define tap_delete { u32 client_index; u32 context; u32 sw_if_index; }; -/** \brief Reply for tap delete request - @param context - returned sender context, to match reply w/ request - @param retval - return code -*/ -define tap_delete_reply -{ - u32 context; - i32 retval; -}; - /** \brief Dump tap interfaces request */ define sw_interface_tap_dump { diff --git a/src/vnet/vxlan/vxlan.api b/src/vnet/vxlan/vxlan.api index 048220fb..6c331a58 100644 --- a/src/vnet/vxlan/vxlan.api +++ b/src/vnet/vxlan/vxlan.api @@ -61,7 +61,7 @@ define vxlan_tunnel_details @param is_ipv6 - if non-zero, enable ipv6-vxlan-bypass, else ipv4-vxlan-bypass @param enable - if non-zero enable, else disable */ -define sw_interface_set_vxlan_bypass +autoreply define sw_interface_set_vxlan_bypass { u32 client_index; u32 context; @@ -69,13 +69,3 @@ define sw_interface_set_vxlan_bypass u8 is_ipv6; u8 enable; }; - -/** \brief Interface set vxlan-bypass response - @param context - sender context, to match reply w/ request - @param retval - return code for the request -*/ -define sw_interface_set_vxlan_bypass_reply -{ - u32 context; - i32 retval; -}; \ No newline at end of file diff --git a/src/vpp/api/vpe.api b/src/vpp/api/vpe.api index a4ba180d..7c07c822 100644 --- a/src/vpp/api/vpe.api +++ b/src/vpp/api/vpe.api @@ -80,7 +80,7 @@ define create_vlan_subif_reply @param sw_if_index - index of the interface @param enable - if non-zero enable, else disable */ -define sw_interface_set_mpls_enable +autoreply define sw_interface_set_mpls_enable { u32 client_index; u32 context; @@ -88,16 +88,6 @@ define sw_interface_set_mpls_enable u8 enable; }; -/** \brief Reply for MPLS state on an interface - @param context - returned sender context, to match reply w/ request - @param retval - return code -*/ -define sw_interface_set_mpls_enable_reply -{ - u32 context; - i32 retval; -}; - /** \brief Proxy ARP add / del request @param client_index - opaque cookie to identify the sender @param context - sender context, to match reply w/ request @@ -106,7 +96,7 @@ define sw_interface_set_mpls_enable_reply @param low_address[4] - Low address of the Proxy ARP range @param hi_address[4] - High address of the Proxy ARP range */ -define proxy_arp_add_del +autoreply define proxy_arp_add_del { u32 client_index; u32 context; @@ -116,23 +106,13 @@ define proxy_arp_add_del u8 hi_address[4]; }; -/** \brief Reply for proxy arp add / del request - @param context - returned sender context, to match reply w/ request - @param retval - return code -*/ -define proxy_arp_add_del_reply -{ - u32 context; - i32 retval; -}; - /** \brief Proxy ARP add / del request @param client_index - opaque cookie to identify the sender @param context - sender context, to match reply w/ request @param sw_if_index - Which interface to enable / disable Proxy Arp on @param enable_disable - 1 to enable Proxy ARP on interface, 0 to disable */ -define proxy_arp_intfc_enable_disable +autoreply define proxy_arp_intfc_enable_disable { u32 client_index; u32 context; @@ -141,23 +121,13 @@ define proxy_arp_intfc_enable_disable u8 enable_disable; }; -/** \brief Reply for Proxy ARP interface enable / disable request - @param context - returned sender context, to match reply w/ request - @param retval - return code -*/ -define proxy_arp_intfc_enable_disable_reply -{ - u32 context; - i32 retval; -}; - /** \brief Reset VRF (remove all routes etc) request @param client_index - opaque cookie to identify the sender @param context - sender context, to match reply w/ request @param is_ipv6 - 1 for IPv6 neighbor, 0 for IPv4 @param vrf_id - ID of th FIB table / VRF to reset */ -define reset_vrf +autoreply define reset_vrf { u32 client_index; u32 context; @@ -165,16 +135,6 @@ define reset_vrf u32 vrf_id; }; -/** \brief Reply for Reset VRF request - @param context - returned sender context, to match reply w/ request - @param retval - return code -*/ -define reset_vrf_reply -{ - u32 context; - i32 retval; -}; - /** \brief Is Address Reachable request - DISABLED @param client_index - opaque cookie to identify the sender @param context - sender context, to match reply w/ request @@ -200,7 +160,7 @@ define is_address_reachable @param enable_disable - 1 = enable stats, 0 = disable @param pid - pid of process requesting stats updates */ -define want_stats +autoreply define want_stats { u32 client_index; u32 context; @@ -208,16 +168,6 @@ define want_stats u32 pid; }; -/** \brief Reply for Want Stats request - @param context - returned sender context, to match reply w/ request - @param retval - return code -*/ -define want_stats_reply -{ - u32 context; - i32 retval; -}; - typeonly manual_print manual_endian define ip4_fib_counter { u32 address; @@ -331,7 +281,7 @@ define oam_event @param enable_disable- enable if non-zero, else disable @param pid - pid of the requesting process */ -define want_oam_events +autoreply define want_oam_events { u32 client_index; u32 context; @@ -339,16 +289,6 @@ define want_oam_events u32 pid; }; -/** \brief Want OAM events response - @param context - sender context, to match reply w/ request - @param retval - return code for the want oam stats request -*/ -define want_oam_events_reply -{ - u32 context; - i32 retval; -}; - /** \brief OAM add / del target request @param client_index - opaque cookie to identify the sender @param context - sender context, to match reply w/ request @@ -357,7 +297,7 @@ define want_oam_events_reply @param dst_address[] - destination address of the target @param is_add - add target if non-zero, else delete */ -define oam_add_del +autoreply define oam_add_del { u32 client_index; u32 context; @@ -367,23 +307,13 @@ define oam_add_del u8 is_add; }; -/** \brief OAM add / del target response - @param context - sender context, to match reply w/ request - @param retval - return code of the request -*/ -define oam_add_del_reply -{ - u32 context; - i32 retval; -}; - /** \brief Reset fib table request @param client_index - opaque cookie to identify the sender @param context - sender context, to match reply w/ request @param vrf_id - vrf/table id of the fib table to reset @param is_ipv6 - an ipv6 fib to reset if non-zero, else ipv4 */ -define reset_fib +autoreply define reset_fib { u32 client_index; u32 context; @@ -391,16 +321,6 @@ define reset_fib u8 is_ipv6; }; -/** \brief Reset fib response - @param context - sender context, to match reply w/ request - @param retval - return code for the reset bfib request -*/ -define reset_fib_reply -{ - u32 context; - i32 retval; -}; - /** \brief Create loopback interface request @param client_index - opaque cookie to identify the sender @param context - sender context, to match reply w/ request @@ -458,23 +378,13 @@ define create_loopback_instance_reply @param context - sender context, to match reply w/ request @param sw_if_index - sw index of the interface that was created */ -define delete_loopback +autoreply define delete_loopback { u32 client_index; u32 context; u32 sw_if_index; }; -/** \brief Delete loopback interface response - @param context - sender context, to match reply w/ request - @param retval - return code for the request -*/ -define delete_loopback_reply -{ - u32 context; - i32 retval; -}; - /** \brief Control ping from client to api server request @param client_index - opaque cookie to identify the sender @param context - sender context, to match reply w/ request @@ -543,7 +453,7 @@ define cli_inband_reply @param is_ipv6 - neighbor limit if non-zero, else ARP limit @param arp_neighbor_limit - the new limit, defaults are ~ 50k */ -define set_arp_neighbor_limit +autoreply define set_arp_neighbor_limit { u32 client_index; u32 context; @@ -551,16 +461,6 @@ define set_arp_neighbor_limit u32 arp_neighbor_limit; }; -/** \brief Set max allowed ARP or ip6 neighbor entries response - @param context - sender context, to match reply w/ request - @param retval - return code for request -*/ -define set_arp_neighbor_limit_reply -{ - u32 context; - i32 retval; -}; - /** \brief L2 interface patch add / del request @param client_index - opaque cookie to identify the sender @param context - sender context, to match reply w/ request @@ -568,7 +468,7 @@ define set_arp_neighbor_limit_reply @param tx_sw_if_index - transmit side interface @param is_add - if non-zero set up the interface patch, else remove it */ -define l2_patch_add_del +autoreply define l2_patch_add_del { u32 client_index; u32 context; @@ -577,23 +477,13 @@ define l2_patch_add_del u8 is_add; }; -/** \brief L2 interface patch add / del response - @param context - sender context, to match reply w/ request - @param retval - return code for the request -*/ -define l2_patch_add_del_reply -{ - u32 context; - i32 retval; -}; - /** \brief Interface set vpath request @param client_index - opaque cookie to identify the sender @param context - sender context, to match reply w/ request @param sw_if_index - interface used to reach neighbor @param enable - if non-zero enable, else disable */ -define sw_interface_set_vpath +autoreply define sw_interface_set_vpath { u32 client_index; u32 context; @@ -601,16 +491,6 @@ define sw_interface_set_vpath u8 enable; }; -/** \brief Interface set vpath response - @param context - sender context, to match reply w/ request - @param retval - return code for the request -*/ -define sw_interface_set_vpath_reply -{ - u32 context; - i32 retval; -}; - /** \brief Set L2 XConnect between two interfaces request @param client_index - opaque cookie to identify the sender @param context - sender context, to match reply w/ request @@ -618,7 +498,7 @@ define sw_interface_set_vpath_reply @param tx_sw_if_index - Transmit interface index @param enable - enable xconnect if not 0, else set to L3 mode */ -define sw_interface_set_l2_xconnect +autoreply define sw_interface_set_l2_xconnect { u32 client_index; u32 context; @@ -627,16 +507,6 @@ define sw_interface_set_l2_xconnect u8 enable; }; -/** \brief Set L2 XConnect response - @param context - sender context, to match reply w/ request - @param retval - L2 XConnect request return code -*/ -define sw_interface_set_l2_xconnect_reply -{ - u32 context; - i32 retval; -}; - /** \brief Interface bridge mode request @param client_index - opaque cookie to identify the sender @param context - sender context, to match reply w/ request @@ -646,7 +516,7 @@ define sw_interface_set_l2_xconnect_reply @param shg - Shared horizon group, for bridge mode only @param enable - Enable beige mode if not 0, else set to L3 mode */ -define sw_interface_set_l2_bridge +autoreply define sw_interface_set_l2_bridge { u32 client_index; u32 context; @@ -657,16 +527,6 @@ define sw_interface_set_l2_bridge u8 enable; }; -/** \brief Interface bridge mode response - @param context - sender context, to match reply w/ request - @param retval - Bridge mode request return code -*/ -define sw_interface_set_l2_bridge_reply -{ - u32 context; - i32 retval; -}; - /** \brief Set bridge domain ip to mac entry request @param client_index - opaque cookie to identify the sender @param context - sender context, to match reply w/ request @@ -676,7 +536,7 @@ define sw_interface_set_l2_bridge_reply @param mac_address - MAC address @param */ -define bd_ip_mac_add_del +autoreply define bd_ip_mac_add_del { u32 client_index; u32 context; @@ -687,16 +547,6 @@ define bd_ip_mac_add_del u8 mac_address[6]; }; -/** \brief Set bridge domain ip to mac entry response - @param context - sender context, to match reply w/ request - @param retval - return code for the set bridge flags request -*/ -define bd_ip_mac_add_del_reply -{ - u32 context; - i32 retval; -}; - /** \brief Set/unset the classification table for an interface request @param client_index - opaque cookie to identify the sender @param context - sender context, to match reply w/ request @@ -704,7 +554,7 @@ define bd_ip_mac_add_del_reply @param sw_if_index - interface to associate with the table @param table_index - index of the table, if ~0 unset the table */ -define classify_set_interface_ip_table +autoreply define classify_set_interface_ip_table { u32 client_index; u32 context; @@ -713,16 +563,6 @@ define classify_set_interface_ip_table u32 table_index; /* ~0 => off */ }; -/** \brief Set/unset interface classification table response - @param context - sender context, to match reply w/ request - @param retval - return code -*/ -define classify_set_interface_ip_table_reply -{ - u32 context; - i32 retval; -}; - /** \brief Set/unset l2 classification tables for an interface request @param client_index - opaque cookie to identify the sender @param context - sender context, to match reply w/ request @@ -731,7 +571,7 @@ define classify_set_interface_ip_table_reply @param ip6_table_index - ip6 index @param other_table_index - other index */ -define classify_set_interface_l2_tables +autoreply define classify_set_interface_l2_tables { u32 client_index; u32 context; @@ -743,16 +583,6 @@ define classify_set_interface_l2_tables u8 is_input; }; -/** \brief Set/unset l2 classification tables for an interface response - @param context - sender context, to match reply w/ request - @param retval - return code for the request -*/ -define classify_set_interface_l2_tables_reply -{ - u32 context; - i32 retval; -}; - /** \brief Get node index using name request @param client_index - opaque cookie to identify the sender @param context - sender context, to match reply w/ request @@ -809,7 +639,7 @@ define add_node_next_reply @param sw_if_index - interface to enable/disable filtering on @param enable_disable - if non-zero enable filtering, else disable */ -define l2_interface_efp_filter +autoreply define l2_interface_efp_filter { u32 client_index; u32 context; @@ -817,16 +647,6 @@ define l2_interface_efp_filter u32 enable_disable; }; -/** \brief L2 interface ethernet flow point filtering response - @param context - sender context, to match reply w/ request - @param retval - return code for the request -*/ -define l2_interface_efp_filter_reply -{ - u32 context; - i32 retval; -}; - define create_subif { u32 client_index; @@ -882,7 +702,7 @@ define show_version_reply }; /* Gross kludge, DGMS */ -define interface_name_renumber +autoreply define interface_name_renumber { u32 client_index; u32 context; @@ -890,12 +710,6 @@ define interface_name_renumber u32 new_show_dev_instance; }; -define interface_name_renumber_reply -{ - u32 context; - i32 retval; -}; - /** \brief Register for ip4 arp resolution events @param client_index - opaque cookie to identify the sender @param context - sender context, to match reply w/ request @@ -903,7 +717,7 @@ define interface_name_renumber_reply @param pid - sender's pid @param address - the exact ip4 address of interest */ -define want_ip4_arp_events +autoreply define want_ip4_arp_events { u32 client_index; u32 context; @@ -912,16 +726,6 @@ define want_ip4_arp_events u32 address; }; -/** \brief Reply for interface events registration - @param context - returned sender context, to match reply w/ request - @param retval - return code -*/ -define want_ip4_arp_events_reply -{ - u32 context; - i32 retval; -}; - /** \brief Tell client about an ip4 arp resolution event @param client_index - opaque cookie to identify the sender @param context - sender context, to match reply w/ request @@ -949,7 +753,7 @@ define ip4_arp_event @param pid - sender's pid @param address - the exact ip6 address of interest */ -define want_ip6_nd_events +autoreply define want_ip6_nd_events { u32 client_index; u32 context; @@ -958,16 +762,6 @@ define want_ip6_nd_events u8 address[16]; }; -/** \brief Reply for ip6 nd resolution events registration - @param context - returned sender context, to match reply w/ request - @param retval - return code -*/ -define want_ip6_nd_events_reply -{ - u32 context; - i32 retval; -}; - /** \brief Tell client about an ip6 nd resolution or mac/ip event @param client_index - opaque cookie to identify the sender @param context - sender context, to match reply w/ request @@ -999,7 +793,7 @@ define ip6_nd_event Note: User is recommeneded to use just one valid table_index per call. (ip4_table_index, ip6_table_index, or l2_table_index) */ -define input_acl_set_interface +autoreply define input_acl_set_interface { u32 client_index; u32 context; @@ -1010,16 +804,6 @@ define input_acl_set_interface u8 is_add; }; -/** \brief Set/unset input ACL interface response - @param context - sender context, to match reply w/ request - @param retval - return code for the request -*/ -define input_acl_set_interface_reply -{ - u32 context; - i32 retval; -}; - define get_node_graph { u32 client_index; @@ -1048,7 +832,7 @@ define get_node_graph_reply @param pow_enable - Proof of Work enabled or not flag @param trace_enable - iOAM Trace enabled or not flag */ -define ioam_enable +autoreply define ioam_enable { u32 client_index; u32 context; @@ -1060,38 +844,18 @@ define ioam_enable u32 node_id; }; -/** \brief iOAM Trace profile add / del response - @param context - sender context, to match reply w/ request - @param retval - return value for request -*/ -define ioam_enable_reply -{ - u32 context; - i32 retval; -}; - /** \brief iOAM disable @param client_index - opaque cookie to identify the sender @param context - sender context, to match reply w/ request @param index - MAP Domain index */ -define ioam_disable +autoreply define ioam_disable { u32 client_index; u32 context; u16 id; }; -/** \brief iOAM disable response - @param context - sender context, to match reply w/ request - @param retval - return value for request -*/ -define ioam_disable_reply -{ - u32 context; - i32 retval; -}; - /** \brief Query relative index via node names @param client_index - opaque cookie to identify the sender @param context - sender context, to match reply w/ request @@ -1149,7 +913,7 @@ define pg_create_interface_reply @param count - number of packets to be captured @param pcap_file - pacp file name to store captured packets */ -define pg_capture +autoreply define pg_capture { u32 client_index; u32 context; @@ -1160,23 +924,13 @@ define pg_capture u8 pcap_file_name[pcap_name_length]; }; -/** \brief PacketGenerator capture packets response - @param context - sender context, to match reply w/ request - @param retval - return value for request -*/ -define pg_capture_reply -{ - u32 context; - i32 retval; -}; - /** \brief Enable / disable packet generator request @param client_index - opaque cookie to identify the sender @param context - sender context, to match reply w/ request @param is_enabled - 1 if enabling streams, 0 if disabling @param stream - stream name to be enable/disabled, if not specified handle all streams */ -define pg_enable_disable +autoreply define pg_enable_disable { u32 client_index; u32 context; @@ -1185,16 +939,6 @@ define pg_enable_disable u8 stream_name[stream_name_length]; }; -/** \brief Reply for enable / disable packet generator - @param context - returned sender context, to match reply w/ request - @param retval - return code -*/ -define pg_enable_disable_reply -{ - u32 context; - i32 retval; -}; - /** \brief Configure IP source and L4 port-range check @param client_index - opaque cookie to identify the sender @param context - sender context, to match reply w/ request @@ -1208,7 +952,7 @@ define pg_enable_disable_reply @param vrf_id - fib table/vrf id to associate the source and port-range check with @note To specify a single port set low_port and high_port entry the same */ -define ip_source_and_port_range_check_add_del +autoreply define ip_source_and_port_range_check_add_del { u32 client_index; u32 context; @@ -1222,16 +966,6 @@ define ip_source_and_port_range_check_add_del u32 vrf_id; }; -/** \brief Configure IP source and L4 port-range check reply - @param context - returned sender context, to match reply w/ request - @param retval - return code -*/ -define ip_source_and_port_range_check_add_del_reply -{ - u32 context; - i32 retval; -}; - /** \brief Set interface source and L4 port-range request @param client_index - opaque cookie to identify the sender @param context - sender context, to match reply w/ request @@ -1239,7 +973,7 @@ define ip_source_and_port_range_check_add_del_reply @param tcp_vrf_id - VRF associated with source and TCP port-range check @param udp_vrf_id - VRF associated with source and TCP port-range check */ -define ip_source_and_port_range_check_interface_add_del +autoreply define ip_source_and_port_range_check_interface_add_del { u32 client_index; u32 context; @@ -1251,36 +985,17 @@ define ip_source_and_port_range_check_interface_add_del u32 udp_out_vrf_id; }; -/** \brief Set interface source and L4 port-range response - @param context - sender context, to match reply w/ request - @param retval - return value for request -*/ -define ip_source_and_port_range_check_interface_add_del_reply -{ - u32 context; - i32 retval; -}; - /** \brief Delete sub interface request @param client_index - opaque cookie to identify the sender @param context - sender context, to match reply w/ request @param sw_if_index - sw index of the interface that was created by create_subif */ -define delete_subif { +autoreply define delete_subif { u32 client_index; u32 context; u32 sw_if_index; }; -/** \brief Delete sub interface response - @param context - sender context, to match reply w/ request - @param retval - return code for the request -*/ -define delete_subif_reply { - u32 context; - i32 retval; -}; - /** \brief Punt traffic to the host @param client_index - opaque cookie to identify the sender @param context - sender context, to match reply w/ request @@ -1289,7 +1004,7 @@ define delete_subif_reply { @param l4_protocol - L4 protocol to be punted, only UDP (0x11) is supported @param l4_port - TCP/UDP port to be punted */ -define punt { +autoreply define punt { u32 client_index; u32 context; u8 is_add; @@ -1298,23 +1013,13 @@ define punt { u16 l4_port; }; -/** \brief Reply to the punt request - @param context - sender context which was passed in the request - @param retval - return code of punt request -*/ -define punt_reply -{ - u32 context; - i32 retval; -}; - /** \brief Feature path enable/disable request @param client_index - opaque cookie to identify the sender @param context - sender context, to match reply w/ request @param sw_if_index - the interface @param enable - 1 = on, 0 = off */ -define feature_enable_disable { +autoreply define feature_enable_disable { u32 client_index; u32 context; u32 sw_if_index; @@ -1323,16 +1028,6 @@ define feature_enable_disable { u8 feature_name[64]; }; -/** \brief Reply to the eature path enable/disable request - @param context - sender context which was passed in the request - @param retval - return code for the request -*/ -define feature_enable_disable_reply -{ - u32 context; - i32 retval; -}; - /* * Local Variables: * eval: (c-set-style "gnu") -- cgit 1.2.3-korg From cada2a0ae4d1c1add4ad1b37649555ae7e9afc14 Mon Sep 17 00:00:00 2001 From: Dave Barach Date: Thu, 18 May 2017 19:16:47 -0400 Subject: VPP-849: improve vnet classifier memory allocator performance Port the linear-scan bucket fix from bihash_template.c. Change-Id: Id8b2d1fe402401f098270ce6121c2f44f2f24c49 Signed-off-by: Dave Barach --- src/vnet/classify/vnet_classify.c | 621 ++++++++++++++++++++++++-------------- src/vnet/classify/vnet_classify.h | 103 ++++--- 2 files changed, 450 insertions(+), 274 deletions(-) (limited to 'src/vnet/classify') diff --git a/src/vnet/classify/vnet_classify.c b/src/vnet/classify/vnet_classify.c index 70a189b0..418e3da8 100644 --- a/src/vnet/classify/vnet_classify.c +++ b/src/vnet/classify/vnet_classify.c @@ -166,37 +166,21 @@ static vnet_classify_entry_t * vnet_classify_entry_alloc (vnet_classify_table_t * t, u32 log2_pages) { vnet_classify_entry_t * rv = 0; -#define _(size) \ - vnet_classify_entry_##size##_t * rv##size = 0; - foreach_size_in_u32x4; -#undef _ - + u32 required_length; void * oldheap; ASSERT (t->writer_lock[0]); + required_length = + (sizeof(vnet_classify_entry_t) + (t->match_n_vectors*sizeof(u32x4))) + * t->entries_per_page * (1<= vec_len (t->freelists) || t->freelists [log2_pages] == 0) { oldheap = clib_mem_set_heap (t->mheap); vec_validate (t->freelists, log2_pages); - switch(t->match_n_vectors) - { - /* Euchre the vector allocator into allocating the right sizes */ -#define _(size) \ - case size: \ - vec_validate_aligned \ - (rv##size, ((1<entries_per_page) - 1, \ - CLIB_CACHE_LINE_BYTES); \ - rv = (vnet_classify_entry_t *) rv##size; \ - break; - foreach_size_in_u32x4; -#undef _ - - default: - abort(); - } - + rv = clib_mem_alloc_aligned (required_length, CLIB_CACHE_LINE_BYTES); clib_mem_set_heap (oldheap); goto initialize; } @@ -205,39 +189,21 @@ vnet_classify_entry_alloc (vnet_classify_table_t * t, u32 log2_pages) initialize: ASSERT(rv); - ASSERT (vec_len(rv) == (1<entries_per_page); - switch (t->match_n_vectors) - { -#define _(size) \ - case size: \ - if(vec_len(rv)) \ - memset (rv, 0xff, sizeof (*rv##size) * vec_len(rv)); \ - break; - foreach_size_in_u32x4; -#undef _ - - default: - abort(); - } - + memset (rv, 0xff, required_length); return rv; } static void vnet_classify_entry_free (vnet_classify_table_t * t, - vnet_classify_entry_t * v) + vnet_classify_entry_t * v, u32 log2_pages) { - u32 free_list_index; - ASSERT (t->writer_lock[0]); - free_list_index = min_log2(vec_len(v)/t->entries_per_page); - - ASSERT(vec_len (t->freelists) > free_list_index); + ASSERT(vec_len (t->freelists) > log2_pages); - v->next_free = t->freelists[free_list_index]; - t->freelists[free_list_index] = v; + v->next_free = t->freelists[log2_pages]; + t->freelists[log2_pages] = v; } static inline void make_working_copy @@ -247,16 +213,15 @@ static inline void make_working_copy vnet_classify_bucket_t working_bucket __attribute__((aligned (8))); void * oldheap; vnet_classify_entry_t * working_copy; -#define _(size) \ - vnet_classify_entry_##size##_t * working_copy##size = 0; - foreach_size_in_u32x4; -#undef _ u32 thread_index = vlib_get_thread_index(); + int working_copy_length, required_length; if (thread_index >= vec_len (t->working_copies)) { oldheap = clib_mem_set_heap (t->mheap); vec_validate (t->working_copies, thread_index); + vec_validate (t->working_copy_lengths, thread_index); + t->working_copy_lengths[thread_index] = -1; clib_mem_set_heap (oldheap); } @@ -266,53 +231,28 @@ static inline void make_working_copy * lookup failures. */ working_copy = t->working_copies[thread_index]; + working_copy_length = t->working_copy_lengths[thread_index]; + required_length = + (sizeof(vnet_classify_entry_t) + (t->match_n_vectors*sizeof(u32x4))) + * t->entries_per_page * (1<log2_pages); t->saved_bucket.as_u64 = b->as_u64; oldheap = clib_mem_set_heap (t->mheap); - if ((1<log2_pages)*t->entries_per_page > vec_len (working_copy)) + if (required_length > working_copy_length) { - switch(t->match_n_vectors) - { - /* Euchre the vector allocator into allocating the right sizes */ -#define _(size) \ - case size: \ - working_copy##size = (void *) working_copy; \ - vec_validate_aligned \ - (working_copy##size, \ - ((1<log2_pages)*t->entries_per_page) - 1, \ - CLIB_CACHE_LINE_BYTES); \ - working_copy = (void *) working_copy##size; \ - break; - foreach_size_in_u32x4; -#undef _ - - default: - abort(); - } + if (working_copy) + clib_mem_free (working_copy); + working_copy = + clib_mem_alloc_aligned (required_length, CLIB_CACHE_LINE_BYTES); t->working_copies[thread_index] = working_copy; } - _vec_len(working_copy) = (1<log2_pages)*t->entries_per_page; clib_mem_set_heap (oldheap); v = vnet_classify_get_entry (t, b->offset); - switch(t->match_n_vectors) - { -#define _(size) \ - case size: \ - clib_memcpy (working_copy, v, \ - sizeof (vnet_classify_entry_##size##_t) \ - * (1<log2_pages) \ - * (t->entries_per_page)); \ - break; - foreach_size_in_u32x4 ; -#undef _ - - default: - abort(); - } + clib_memcpy (working_copy, v, required_length); working_bucket.as_u64 = b->as_u64; working_bucket.offset = vnet_classify_get_offset (t, working_copy); @@ -323,51 +263,48 @@ static inline void make_working_copy static vnet_classify_entry_t * split_and_rehash (vnet_classify_table_t * t, - vnet_classify_entry_t * old_values, + vnet_classify_entry_t * old_values, u32 old_log2_pages, u32 new_log2_pages) { vnet_classify_entry_t * new_values, * v, * new_v; - int i, j, k; + int i, j, length_in_entries; new_values = vnet_classify_entry_alloc (t, new_log2_pages); + length_in_entries = (1<entries_per_page; - for (i = 0; i < (vec_len (old_values)/t->entries_per_page); i++) + for (i = 0; i < length_in_entries; i++) { u64 new_hash; - for (j = 0; j < t->entries_per_page; j++) + v = vnet_classify_entry_at_index (t, old_values, i); + + if (vnet_classify_entry_is_busy (v)) { - v = vnet_classify_entry_at_index - (t, old_values, i * t->entries_per_page + j); + /* Hack so we can use the packet hash routine */ + u8 * key_minus_skip; + key_minus_skip = (u8 *) v->key; + key_minus_skip -= t->skip_n_vectors * sizeof (u32x4); + + new_hash = vnet_classify_hash_packet (t, key_minus_skip); + new_hash >>= t->log2_nbuckets; + new_hash &= (1<entries_per_page; j++) { - /* Hack so we can use the packet hash routine */ - u8 * key_minus_skip; - key_minus_skip = (u8 *) v->key; - key_minus_skip -= t->skip_n_vectors * sizeof (u32x4); - - new_hash = vnet_classify_hash_packet (t, key_minus_skip); - new_hash >>= t->log2_nbuckets; - new_hash &= (1<entries_per_page; k++) - { - new_v = vnet_classify_entry_at_index (t, new_values, - new_hash + k); + new_v = vnet_classify_entry_at_index (t, new_values, + new_hash + j); - if (vnet_classify_entry_is_free (new_v)) - { - clib_memcpy (new_v, v, sizeof (vnet_classify_entry_t) - + (t->match_n_vectors * sizeof (u32x4))); - new_v->flags &= ~(VNET_CLASSIFY_ENTRY_FREE); - goto doublebreak; - } + if (vnet_classify_entry_is_free (new_v)) + { + clib_memcpy (new_v, v, sizeof (vnet_classify_entry_t) + + (t->match_n_vectors * sizeof (u32x4))); + new_v->flags &= ~(VNET_CLASSIFY_ENTRY_FREE); + goto doublebreak; } - /* Crap. Tell caller to try again */ - vnet_classify_entry_free (t, new_values); - return 0; } + /* Crap. Tell caller to try again */ + vnet_classify_entry_free (t, new_values, new_log2_pages); + return 0; doublebreak: ; } @@ -375,6 +312,56 @@ split_and_rehash (vnet_classify_table_t * t, return new_values; } +static vnet_classify_entry_t * +split_and_rehash_linear (vnet_classify_table_t * t, + vnet_classify_entry_t * old_values, + u32 old_log2_pages, + u32 new_log2_pages) +{ + vnet_classify_entry_t * new_values, * v, * new_v; + int i, j, new_length_in_entries, old_length_in_entries; + + new_values = vnet_classify_entry_alloc (t, new_log2_pages); + new_length_in_entries = (1<entries_per_page; + old_length_in_entries = (1<entries_per_page; + + j = 0; + for (i = 0; i < old_length_in_entries; i++) + { + v = vnet_classify_entry_at_index (t, old_values, i); + + if (vnet_classify_entry_is_busy (v)) + { + for (; j < new_length_in_entries; j++) + { + new_v = vnet_classify_entry_at_index (t, new_values, j); + + if (vnet_classify_entry_is_busy (new_v)) + { + clib_warning ("BUG: linear rehash new entry not free!"); + continue; + } + clib_memcpy (new_v, v, sizeof (vnet_classify_entry_t) + + (t->match_n_vectors * sizeof (u32x4))); + new_v->flags &= ~(VNET_CLASSIFY_ENTRY_FREE); + j++; + goto doublebreak; + } + /* + * Crap. Tell caller to try again. + * This should never happen... + */ + clib_warning ("BUG: linear rehash failed!"); + vnet_classify_entry_free (t, new_values, new_log2_pages); + return 0; + } + doublebreak: + ; + } + + return new_values; +} + int vnet_classify_add_del (vnet_classify_table_t * t, vnet_classify_entry_t * add_v, int is_add) @@ -386,9 +373,12 @@ int vnet_classify_add_del (vnet_classify_table_t * t, int rv = 0; int i; u64 hash, new_hash; - u32 new_log2_pages; + u32 limit; + u32 old_log2_pages, new_log2_pages; u32 thread_index = vlib_get_thread_index(); u8 * key_minus_skip; + int resplit_once; + int mark_bucket_linear; ASSERT ((add_v->flags & VNET_CLASSIFY_ENTRY_FREE) == 0); @@ -432,6 +422,12 @@ int vnet_classify_add_del (vnet_classify_table_t * t, save_v = vnet_classify_get_entry (t, t->saved_bucket.offset); value_index = hash & ((1<saved_bucket.log2_pages)-1); + limit = t->entries_per_page; + if (PREDICT_FALSE (b->linear_search)) + { + value_index = 0; + limit *= (1<log2_pages); + } if (is_add) { @@ -440,7 +436,7 @@ int vnet_classify_add_del (vnet_classify_table_t * t, * replace an existing key, then look for an empty slot. */ - for (i = 0; i < t->entries_per_page; i++) + for (i = 0; i < limit; i++) { v = vnet_classify_entry_at_index (t, save_v, value_index + i); @@ -456,7 +452,7 @@ int vnet_classify_add_del (vnet_classify_table_t * t, goto unlock; } } - for (i = 0; i < t->entries_per_page; i++) + for (i = 0; i < limit; i++) { v = vnet_classify_entry_at_index (t, save_v, value_index + i); @@ -475,7 +471,7 @@ int vnet_classify_add_del (vnet_classify_table_t * t, } else { - for (i = 0; i < t->entries_per_page; i++) + for (i = 0; i < limit; i++) { v = vnet_classify_entry_at_index (t, save_v, value_index + i); @@ -495,16 +491,39 @@ int vnet_classify_add_del (vnet_classify_table_t * t, goto unlock; } - new_log2_pages = t->saved_bucket.log2_pages + 1; - - expand_again: + old_log2_pages = t->saved_bucket.log2_pages; + new_log2_pages = old_log2_pages + 1; working_copy = t->working_copies[thread_index]; - new_v = split_and_rehash (t, working_copy, new_log2_pages); + + if (t->saved_bucket.linear_search) + goto linear_resplit; + + mark_bucket_linear = 0; + + new_v = split_and_rehash (t, working_copy, old_log2_pages, new_log2_pages); if (new_v == 0) { + try_resplit: + resplit_once = 1; new_log2_pages++; - goto expand_again; + + new_v = split_and_rehash (t, working_copy, old_log2_pages, + new_log2_pages); + if (new_v == 0) + { + mark_linear: + new_log2_pages--; + + linear_resplit: + /* pinned collisions, use linear search */ + new_v = split_and_rehash_linear (t, working_copy, old_log2_pages, + new_log2_pages); + /* A new linear-search bucket? */ + if (!t->saved_bucket.linear_search) + t->linear_buckets ++; + mark_bucket_linear = 1; + } } /* Try to add the new entry */ @@ -515,9 +534,16 @@ int vnet_classify_add_del (vnet_classify_table_t * t, new_hash = vnet_classify_hash_packet_inline (t, key_minus_skip); new_hash >>= t->log2_nbuckets; - new_hash &= (1<entries_per_page))) - 1; + new_hash &= (1<entries_per_page; + if (mark_bucket_linear) + { + limit *= (1<entries_per_page; i++) + for (i = 0; i < limit; i++) { new_v = vnet_classify_entry_at_index (t, save_new_v, new_hash + i); @@ -530,23 +556,28 @@ int vnet_classify_add_del (vnet_classify_table_t * t, } } /* Crap. Try again */ + vnet_classify_entry_free (t, save_new_v, new_log2_pages); new_log2_pages++; - vnet_classify_entry_free (t, save_new_v); - goto expand_again; + + if (resplit_once) + goto mark_linear; + else + goto try_resplit; expand_ok: - tmp_b.log2_pages = min_log2 (vec_len (save_new_v)/t->entries_per_page); + tmp_b.log2_pages = new_log2_pages; tmp_b.offset = vnet_classify_get_offset (t, save_new_v); + tmp_b.linear_search = mark_bucket_linear; + CLIB_MEMORY_BARRIER(); b->as_u64 = tmp_b.as_u64; t->active_elements ++; v = vnet_classify_get_entry (t, t->saved_bucket.offset); - vnet_classify_entry_free (t, v); + vnet_classify_entry_free (t, v, old_log2_pages); unlock: CLIB_MEMORY_BARRIER(); t->writer_lock[0] = 0; - return rv; } @@ -610,8 +641,9 @@ u8 * format_classify_table (u8 * s, va_list * args) if (verbose) { - s = format (s, "[%d]: heap offset %d, len %d\n", i, - b->offset, (1<log2_pages)); + s = format (s, "[%d]: heap offset %d, elts %d, %s\n", i, + b->offset, (1<log2_pages)*t->entries_per_page, + b->linear_search ? "LINEAR" : "normal"); } save_v = vnet_classify_get_entry (t, b->offset); @@ -643,6 +675,7 @@ u8 * format_classify_table (u8 * s, va_list * args) s = format (s, " %lld active elements\n", active_elements); s = format (s, " %d free lists\n", vec_len (t->freelists)); + s = format (s, " %d linear-search buckets\n", t->linear_buckets); return s; } @@ -1506,6 +1539,7 @@ static u8 * format_vnet_classify_table (u8 * s, va_list * args) t->current_data_flag, t->current_data_offset); s = format (s, "\n mask %U", format_hex_bytes, t->mask, t->match_n_vectors * sizeof (u32x4)); + s = format (s, "\n linear-search buckets %d\n", t->linear_buckets); if (verbose == 0) return s; @@ -2029,6 +2063,8 @@ int vnet_classify_add_del_session (vnet_classify_main_t * cm, e->metadata = fib_table_find_or_create_and_lock (FIB_PROTOCOL_IP4, metadata); else if (e->action == CLASSIFY_ACTION_SET_IP6_FIB_INDEX) e->metadata = fib_table_find_or_create_and_lock (FIB_PROTOCOL_IP6, metadata); + else + e->metadata = 0; /* Copy key data, honoring skip_n_vectors */ clib_memcpy (&e->key, match + t->skip_n_vectors * sizeof (u32x4), @@ -2281,48 +2317,51 @@ VLIB_INIT_FUNCTION (vnet_classify_init); #define TEST_CODE 1 #if TEST_CODE > 0 -static clib_error_t * -test_classify_command_fn (vlib_main_t * vm, - unformat_input_t * input, - vlib_cli_command_t * cmd) + +typedef struct { - u32 buckets = 2; - u32 sessions = 10; - int i, rv; - vnet_classify_table_t * t = 0; + ip4_address_t addr; + int in_table; +} test_entry_t; + +typedef struct +{ + test_entry_t *entries; + + /* test parameters */ + u32 buckets; + u32 sessions; + u32 iterations; + u32 memory_size; + ip4_address_t src; + vnet_classify_table_t *table; + u32 table_index; + int verbose; + + /* Random seed */ + u32 seed; + + /* Test data */ classify_data_or_mask_t * mask; classify_data_or_mask_t * data; - u8 *mp = 0, *dp = 0; - vnet_classify_main_t * cm = &vnet_classify_main; - vnet_classify_entry_t * e; - int is_add = 1; - u32 tmp; - u32 table_index = ~0; - ip4_address_t src; - u32 deleted = 0; - u32 memory_size = 64<<20; - /* Default starting address 1.0.0.10 */ - src.as_u32 = clib_net_to_host_u32 (0x0100000A); + /* convenience */ + vnet_classify_main_t *classify_main; + vlib_main_t *vlib_main; - while (unformat_check_input (input) != UNFORMAT_END_OF_INPUT) { - if (unformat (input, "sessions %d", &sessions)) - ; - else if (unformat (input, "src %U", unformat_ip4_address, &src)) - ; - else if (unformat (input, "buckets %d", &buckets)) - ; - else if (unformat (input, "memory-size %uM", &tmp)) - memory_size = tmp<<20; - else if (unformat (input, "memory-size %uG", &tmp)) - memory_size = tmp<<30; - else if (unformat (input, "del")) - is_add = 0; - else if (unformat (input, "table %d", &table_index)) - ; - else - break; - } +} test_classify_main_t; + +static test_classify_main_t test_classify_main; + +static clib_error_t * +test_classify_churn (test_classify_main_t *tm) +{ + classify_data_or_mask_t *mask, *data; + vlib_main_t *vm = tm->vlib_main; + test_entry_t *ep; + u8 *mp = 0, *dp = 0; + u32 tmp; + int i, rv; vec_validate_aligned (mp, 3 * sizeof(u32x4), sizeof(u32x4)); vec_validate_aligned (dp, 3 * sizeof(u32x4), sizeof(u32x4)); @@ -2330,105 +2369,231 @@ test_classify_command_fn (vlib_main_t * vm, mask = (classify_data_or_mask_t *) mp; data = (classify_data_or_mask_t *) dp; - data->ip.src_address.as_u32 = src.as_u32; - /* Mask on src address */ memset (&mask->ip.src_address, 0xff, 4); - buckets = 1<src.as_u32); - if (table_index != ~0) + for (i = 0; i < tm->sessions; i++) { - if (pool_is_free_index (cm->tables, table_index)) - { - vlib_cli_output (vm, "No such table %d", table_index); - goto out; - } - t = pool_elt_at_index (cm->tables, table_index); + vec_add2 (tm->entries, ep, 1); + ep->addr.as_u32 = clib_host_to_net_u32 (tmp); + ep->in_table = 0; + tmp++; } - if (is_add) - { - if (t == 0) - { - t = vnet_classify_new_table (cm, (u8 *)mask, buckets, - memory_size, + tm->table = vnet_classify_new_table (tm->classify_main, + (u8 *)mask, + tm->buckets, + tm->memory_size, 0 /* skip */, 3 /* vectors to match */); - t->miss_next_index = IP_LOOKUP_NEXT_DROP; - vlib_cli_output (vm, "Create table %d", t - cm->tables); - } + tm->table->miss_next_index = IP_LOOKUP_NEXT_DROP; + tm->table_index = tm->table - tm->classify_main->tables; + vlib_cli_output (vm, "Created table %d, buckets %d", + tm->table_index, tm->buckets); + + vlib_cli_output (vm, "Initialize: add %d (approx. half of %d sessions)...", + tm->sessions/2, tm->sessions); - vlib_cli_output (vm, "Add %d sessions to %d buckets...", - sessions, buckets); + for (i = 0; i < tm->sessions/2; i++) + { + ep = vec_elt_at_index (tm->entries, i); + + data->ip.src_address.as_u32 = ep->addr.as_u32; + ep->in_table = 1; + + rv = vnet_classify_add_del_session (tm->classify_main, + tm->table_index, + (u8 *) data, + IP_LOOKUP_NEXT_DROP, + i /* opaque_index */, + 0 /* advance */, + 0 /* action*/, + 0 /* metadata */, + 1 /* is_add */); - for (i = 0; i < sessions; i++) - { - rv = vnet_classify_add_del_session (cm, t - cm->tables, (u8 *) data, - IP_LOOKUP_NEXT_DROP, - i+100 /* opaque_index */, - 0 /* advance */, 0, 0, - 1 /* is_add */); - - if (rv != 0) - clib_warning ("add: returned %d", rv); - - tmp = clib_net_to_host_u32 (data->ip.src_address.as_u32) + 1; - data->ip.src_address.as_u32 = clib_net_to_host_u32 (tmp); - } - goto out; + if (rv != 0) + clib_warning ("add: returned %d", rv); + + if (tm->verbose) + vlib_cli_output (vm, "add: %U", format_ip4_address, + &ep->addr.as_u32); } - if (t == 0) + vlib_cli_output (vm, "Execute %d random add/delete operations", + tm->iterations); + + for (i = 0; i < tm->iterations; i++) { - vlib_cli_output (vm, "Must specify table index to delete sessions"); - goto out; + int index, is_add; + + /* Pick a random entry */ + index = random_u32 (&tm->seed) % tm->sessions; + + ep = vec_elt_at_index (tm->entries, index); + + data->ip.src_address.as_u32 = ep->addr.as_u32; + + /* If it's in the table, remove it. Else, add it */ + is_add = !ep->in_table; + + if (tm->verbose) + vlib_cli_output (vm, "%s: %U", + is_add ? "add" : "del", + format_ip4_address, + &ep->addr.as_u32); + + rv = vnet_classify_add_del_session (tm->classify_main, + tm->table_index, + (u8 *) data, + IP_LOOKUP_NEXT_DROP, + i /* opaque_index */, + 0 /* advance */, + 0 /* action*/, + 0 /* metadata */, + is_add); + if (rv != 0) + vlib_cli_output (vm, + "%s[%d]: %U returned %d", is_add ? "add" : "del", + index, + format_ip4_address, + &ep->addr.as_u32, rv); + else + ep->in_table = is_add; } - vlib_cli_output (vm, "Try to delete %d sessions...", sessions); + vlib_cli_output (vm, "Remove remaining %d entries from the table", + tm->table->active_elements); - for (i = 0; i < sessions; i++) + for (i = 0; i < tm->sessions; i++) { u8 * key_minus_skip; u64 hash; + vnet_classify_entry_t * e; + + ep = tm->entries + i; + if (ep->in_table == 0) + continue; + + data->ip.src_address.as_u32 = ep->addr.as_u32; - hash = vnet_classify_hash_packet (t, (u8 *) data); + hash = vnet_classify_hash_packet (tm->table, (u8 *) data); - e = vnet_classify_find_entry (t, (u8 *) data, hash, 0 /* time_now */); - /* Previous delete, perhaps... */ + e = vnet_classify_find_entry (tm->table, + (u8 *) data, hash, 0 /* time_now */); if (e == 0) - continue; - ASSERT (e->opaque_index == (i+100)); + { + clib_warning ("Couldn't find %U index %d which should be present", + format_ip4_address, ep->addr, i); + continue; + } key_minus_skip = (u8 *)e->key; - key_minus_skip -= t->skip_n_vectors * sizeof (u32x4); + key_minus_skip -= tm->table->skip_n_vectors * sizeof (u32x4); + + rv = vnet_classify_add_del_session + (tm->classify_main, + tm->table_index, + key_minus_skip, + IP_LOOKUP_NEXT_DROP, + i /* opaque_index */, + 0 /* advance */, 0, 0, + 0 /* is_add */); - rv = vnet_classify_add_del_session (cm, t - cm->tables, key_minus_skip, - IP_LOOKUP_NEXT_DROP, - i+100 /* opaque_index */, - 0 /* advance */, 0, 0, - 0 /* is_add */); if (rv != 0) clib_warning ("del: returned %d", rv); - - tmp = clib_net_to_host_u32 (data->ip.src_address.as_u32) + 1; - data->ip.src_address.as_u32 = clib_net_to_host_u32 (tmp); - deleted++; + + if (tm->verbose) + vlib_cli_output (vm, "del: %U", format_ip4_address, + &ep->addr.as_u32); } - vlib_cli_output (vm, "Deleted %d sessions...", deleted); + vlib_cli_output (vm, "%d entries remain, MUST be zero", + tm->table->active_elements); + + vlib_cli_output (vm, "Table after cleanup: \n%U\n", + format_classify_table, tm->table, 0 /* verbose */); - out: vec_free (mp); vec_free (dp); + vnet_classify_delete_table_index (tm->classify_main, + tm->table_index, 1 /* del_chain */); + tm->table = 0; + tm->table_index = ~0; + vec_free(tm->entries); + return 0; } +static clib_error_t * +test_classify_command_fn (vlib_main_t * vm, + unformat_input_t * input, + vlib_cli_command_t * cmd) +{ + test_classify_main_t *tm = &test_classify_main; + vnet_classify_main_t * cm = &vnet_classify_main; + u32 tmp; + int which = 0; + clib_error_t * error = 0; + + tm->buckets = 1024; + tm->sessions = 8192; + tm->iterations = 8192; + tm->memory_size = 64<<20; + tm->src.as_u32 = clib_net_to_host_u32 (0x0100000A); + tm->table = 0; + tm->seed = 0xDEADDABE; + tm->classify_main = cm; + tm->vlib_main = vm; + tm->verbose = 0; + + /* Default starting address 1.0.0.10 */ + + while (unformat_check_input (input) != UNFORMAT_END_OF_INPUT) { + if (unformat (input, "sessions %d", &tmp)) + tm->sessions = tmp; + else if (unformat (input, "src %U", unformat_ip4_address, &tm->src.as_u32)) + ; + else if (unformat (input, "buckets %d", &tm->buckets)) + ; + else if (unformat (input, "memory-size %uM", &tmp)) + tm->memory_size = tmp<<20; + else if (unformat (input, "memory-size %uG", &tmp)) + tm->memory_size = tmp<<30; + else if (unformat (input, "seed %d", &tm->seed)) + ; + else if (unformat (input, "verbose")) + tm->verbose = 1; + + else if (unformat (input, "iterations %d", &tm->iterations)) + ; + else if (unformat (input, "churn-test")) + which = 0; + else + break; + } + + switch (which) + { + case 0: + error = test_classify_churn (tm); + break; + default: + error = clib_error_return (0, "No such test"); + break; + } + + return error; +} + VLIB_CLI_COMMAND (test_classify_command, static) = { .path = "test classify", .short_help = - "test classify [src ] [sessions ] [buckets ] [table ] [del]", + "test classify [src ] [sessions ] [buckets ] [seed ]\n" + " [memory-size [M|G]]\n" + " [churn-test]", .function = test_classify_command_fn, }; #endif /* TEST_CODE */ diff --git a/src/vnet/classify/vnet_classify.h b/src/vnet/classify/vnet_classify.h index 2c966632..ffe3dff6 100644 --- a/src/vnet/classify/vnet_classify.h +++ b/src/vnet/classify/vnet_classify.h @@ -132,7 +132,8 @@ typedef struct { union { struct { u32 offset; - u8 pad[3]; + u8 linear_search; + u8 pad[2]; u8 log2_pages; }; u64 as_u64; @@ -151,6 +152,7 @@ typedef struct { u32 skip_n_vectors; u32 nbuckets; u32 log2_nbuckets; + u32 linear_buckets; int entries_per_page; u32 active_elements; u32 current_data_flag; @@ -164,6 +166,7 @@ typedef struct { /* Per-bucket working copies, one per thread */ vnet_classify_entry_t ** working_copies; + int *working_copy_lengths; vnet_classify_bucket_t saved_bucket; /* Free entry freelists */ @@ -354,7 +357,7 @@ vnet_classify_find_entry (vnet_classify_table_t * t, static inline vnet_classify_entry_t * vnet_classify_find_entry_inline (vnet_classify_table_t * t, u8 * h, u64 hash, f64 now) - { +{ vnet_classify_entry_t * v; u32x4 *mask, *key; union { @@ -364,6 +367,7 @@ vnet_classify_find_entry_inline (vnet_classify_table_t * t, vnet_classify_bucket_t * b; u32 value_index; u32 bucket_index; + u32 limit; int i; bucket_index = hash & (t->nbuckets-1); @@ -377,16 +381,23 @@ vnet_classify_find_entry_inline (vnet_classify_table_t * t, v = vnet_classify_get_entry (t, b->offset); value_index = hash & ((1<log2_pages)-1); + limit = t->entries_per_page; + if (PREDICT_FALSE (b->linear_search)) + { + value_index = 0; + limit *= (1<log2_pages); + } + v = vnet_classify_entry_at_index (t, v, value_index); #ifdef CLASSIFY_USE_SSE if (U32X4_ALIGNED(h)) { u32x4 *data = (u32x4 *) h; - for (i = 0; i < t->entries_per_page; i++) { + for (i = 0; i < limit; i++) { key = v->key; result.as_u32x4 = (data[0 + t->skip_n_vectors] & mask[0]) ^ key[0]; switch (t->match_n_vectors) - { + { case 5: result.as_u32x4 |= (data[4 + t->skip_n_vectors] & mask[4]) ^ key[4]; /* FALLTHROUGH */ @@ -403,7 +414,7 @@ vnet_classify_find_entry_inline (vnet_classify_table_t * t, break; default: abort(); - } + } if (u32x4_zero_byte_mask (result.as_u32x4) == 0xffff) { if (PREDICT_TRUE(now)) { @@ -416,51 +427,51 @@ vnet_classify_find_entry_inline (vnet_classify_table_t * t, } } else #endif /* CLASSIFY_USE_SSE */ - { - u32 skip_u64 = t->skip_n_vectors * 2; - u64 *data64 = (u64 *)h; - for (i = 0; i < t->entries_per_page; i++) { - key = v->key; - - result.as_u64[0] = (data64[0 + skip_u64] & ((u64 *)mask)[0]) ^ ((u64 *)key)[0]; - result.as_u64[1] = (data64[1 + skip_u64] & ((u64 *)mask)[1]) ^ ((u64 *)key)[1]; - switch (t->match_n_vectors) - { - case 5: - result.as_u64[0] |= (data64[8 + skip_u64] & ((u64 *)mask)[8]) ^ ((u64 *)key)[8]; - result.as_u64[1] |= (data64[9 + skip_u64] & ((u64 *)mask)[9]) ^ ((u64 *)key)[9]; - /* FALLTHROUGH */ - case 4: - result.as_u64[0] |= (data64[6 + skip_u64] & ((u64 *)mask)[6]) ^ ((u64 *)key)[6]; - result.as_u64[1] |= (data64[7 + skip_u64] & ((u64 *)mask)[7]) ^ ((u64 *)key)[7]; - /* FALLTHROUGH */ - case 3: - result.as_u64[0] |= (data64[4 + skip_u64] & ((u64 *)mask)[4]) ^ ((u64 *)key)[4]; - result.as_u64[1] |= (data64[5 + skip_u64] & ((u64 *)mask)[5]) ^ ((u64 *)key)[5]; - /* FALLTHROUGH */ - case 2: - result.as_u64[0] |= (data64[2 + skip_u64] & ((u64 *)mask)[2]) ^ ((u64 *)key)[2]; - result.as_u64[1] |= (data64[3 + skip_u64] & ((u64 *)mask)[3]) ^ ((u64 *)key)[3]; - /* FALLTHROUGH */ - case 1: - break; - default: - abort(); - } - - if (result.as_u64[0] == 0 && result.as_u64[1] == 0) { - if (PREDICT_TRUE(now)) { - v->hits++; - v->last_heard = now; + { + u32 skip_u64 = t->skip_n_vectors * 2; + u64 *data64 = (u64 *)h; + for (i = 0; i < limit; i++) { + key = v->key; + + result.as_u64[0] = (data64[0 + skip_u64] & ((u64 *)mask)[0]) ^ ((u64 *)key)[0]; + result.as_u64[1] = (data64[1 + skip_u64] & ((u64 *)mask)[1]) ^ ((u64 *)key)[1]; + switch (t->match_n_vectors) + { + case 5: + result.as_u64[0] |= (data64[8 + skip_u64] & ((u64 *)mask)[8]) ^ ((u64 *)key)[8]; + result.as_u64[1] |= (data64[9 + skip_u64] & ((u64 *)mask)[9]) ^ ((u64 *)key)[9]; + /* FALLTHROUGH */ + case 4: + result.as_u64[0] |= (data64[6 + skip_u64] & ((u64 *)mask)[6]) ^ ((u64 *)key)[6]; + result.as_u64[1] |= (data64[7 + skip_u64] & ((u64 *)mask)[7]) ^ ((u64 *)key)[7]; + /* FALLTHROUGH */ + case 3: + result.as_u64[0] |= (data64[4 + skip_u64] & ((u64 *)mask)[4]) ^ ((u64 *)key)[4]; + result.as_u64[1] |= (data64[5 + skip_u64] & ((u64 *)mask)[5]) ^ ((u64 *)key)[5]; + /* FALLTHROUGH */ + case 2: + result.as_u64[0] |= (data64[2 + skip_u64] & ((u64 *)mask)[2]) ^ ((u64 *)key)[2]; + result.as_u64[1] |= (data64[3 + skip_u64] & ((u64 *)mask)[3]) ^ ((u64 *)key)[3]; + /* FALLTHROUGH */ + case 1: + break; + default: + abort(); + } + + if (result.as_u64[0] == 0 && result.as_u64[1] == 0) { + if (PREDICT_TRUE(now)) { + v->hits++; + v->last_heard = now; + } + return (v); } - return (v); - } - v = vnet_classify_entry_at_index (t, v, 1); + v = vnet_classify_entry_at_index (t, v, 1); + } } - } return 0; - } +} vnet_classify_table_t * vnet_classify_new_table (vnet_classify_main_t *cm, -- cgit 1.2.3-korg From 13eaf3e61decdb60bfff1741429cf05129e3cf38 Mon Sep 17 00:00:00 2001 From: Neale Ranns Date: Tue, 23 May 2017 06:10:33 -0700 Subject: Leak locks and tables in the Classifier Change-Id: Iae04c57bba87ab3665388eadd0805f75171636a5 Signed-off-by: Neale Ranns --- src/vnet/classify/vnet_classify.c | 39 +++++++++++++++++++++++++++++++++++++++ src/vnet/classify/vnet_classify.h | 9 ++++++--- src/vnet/fib/ip4_fib.c | 5 +++-- src/vnet/fib/ip6_fib.c | 5 +++-- test/test_classifier.py | 35 +++++++++++++++++++++++++++++++++-- 5 files changed, 84 insertions(+), 9 deletions(-) (limited to 'src/vnet/classify') diff --git a/src/vnet/classify/vnet_classify.c b/src/vnet/classify/vnet_classify.c index 418e3da8..99b10dc2 100644 --- a/src/vnet/classify/vnet_classify.c +++ b/src/vnet/classify/vnet_classify.c @@ -362,6 +362,34 @@ split_and_rehash_linear (vnet_classify_table_t * t, return new_values; } +static void +vnet_classify_entry_claim_resource (vnet_classify_entry_t *e) +{ + switch (e->action) + { + case CLASSIFY_ACTION_SET_IP4_FIB_INDEX: + fib_table_lock (e->metadata, FIB_PROTOCOL_IP4); + break; + case CLASSIFY_ACTION_SET_IP6_FIB_INDEX: + fib_table_lock (e->metadata, FIB_PROTOCOL_IP6); + break; + } +} + +static void +vnet_classify_entry_release_resource (vnet_classify_entry_t *e) +{ + switch (e->action) + { + case CLASSIFY_ACTION_SET_IP4_FIB_INDEX: + fib_table_unlock (e->metadata, FIB_PROTOCOL_IP4); + break; + case CLASSIFY_ACTION_SET_IP6_FIB_INDEX: + fib_table_unlock (e->metadata, FIB_PROTOCOL_IP6); + break; + } +} + int vnet_classify_add_del (vnet_classify_table_t * t, vnet_classify_entry_t * add_v, int is_add) @@ -408,6 +436,7 @@ int vnet_classify_add_del (vnet_classify_table_t * t, clib_memcpy (v, add_v, sizeof (vnet_classify_entry_t) + t->match_n_vectors * sizeof (u32x4)); v->flags &= ~(VNET_CLASSIFY_ENTRY_FREE); + vnet_classify_entry_claim_resource (v); tmp_b.as_u64 = 0; tmp_b.offset = vnet_classify_get_offset (t, v); @@ -445,6 +474,7 @@ int vnet_classify_add_del (vnet_classify_table_t * t, clib_memcpy (v, add_v, sizeof (vnet_classify_entry_t) + t->match_n_vectors * sizeof(u32x4)); v->flags &= ~(VNET_CLASSIFY_ENTRY_FREE); + vnet_classify_entry_claim_resource (v); CLIB_MEMORY_BARRIER(); /* Restore the previous (k,v) pairs */ @@ -461,6 +491,8 @@ int vnet_classify_add_del (vnet_classify_table_t * t, clib_memcpy (v, add_v, sizeof (vnet_classify_entry_t) + t->match_n_vectors * sizeof(u32x4)); v->flags &= ~(VNET_CLASSIFY_ENTRY_FREE); + vnet_classify_entry_claim_resource (v); + CLIB_MEMORY_BARRIER(); b->as_u64 = t->saved_bucket.as_u64; t->active_elements ++; @@ -477,9 +509,11 @@ int vnet_classify_add_del (vnet_classify_table_t * t, if (!memcmp (v->key, add_v->key, t->match_n_vectors * sizeof (u32x4))) { + vnet_classify_entry_release_resource (v); memset (v, 0xff, sizeof (vnet_classify_entry_t) + t->match_n_vectors * sizeof(u32x4)); v->flags |= VNET_CLASSIFY_ENTRY_FREE; + CLIB_MEMORY_BARRIER(); b->as_u64 = t->saved_bucket.as_u64; t->active_elements --; @@ -552,6 +586,8 @@ int vnet_classify_add_del (vnet_classify_table_t * t, clib_memcpy (new_v, add_v, sizeof (vnet_classify_entry_t) + t->match_n_vectors * sizeof(u32x4)); new_v->flags &= ~(VNET_CLASSIFY_ENTRY_FREE); + vnet_classify_entry_claim_resource (new_v); + goto expand_ok; } } @@ -2075,6 +2111,9 @@ int vnet_classify_add_del_session (vnet_classify_main_t * cm, e->key[i] &= t->mask[i]; rv = vnet_classify_add_del (t, e, is_add); + + vnet_classify_entry_release_resource(e); + if (rv) return VNET_API_ERROR_NO_SUCH_ENTRY; return 0; diff --git a/src/vnet/classify/vnet_classify.h b/src/vnet/classify/vnet_classify.h index ffe3dff6..1eb5b14d 100644 --- a/src/vnet/classify/vnet_classify.h +++ b/src/vnet/classify/vnet_classify.h @@ -62,8 +62,11 @@ extern vlib_node_registration_t ip6_classify_node; * - Classified IP packets will be looked up * from the specified ipv6 fib table */ -#define CLASSIFY_ACTION_SET_IP4_FIB_INDEX 1 -#define CLASSIFY_ACTION_SET_IP6_FIB_INDEX 2 +typedef enum vnet_classify_action_t_ +{ + CLASSIFY_ACTION_SET_IP4_FIB_INDEX = 1, + CLASSIFY_ACTION_SET_IP6_FIB_INDEX = 2, +} __attribute__ ((packed)) vnet_classify_action_t; struct _vnet_classify_main; typedef struct _vnet_classify_main vnet_classify_main_t; @@ -93,7 +96,7 @@ typedef CLIB_PACKED(struct _vnet_classify_entry { u8 flags; #define VNET_CLASSIFY_ENTRY_FREE (1<<0) - u8 action; + vnet_classify_action_t action; u16 metadata; /* Hit counter, last heard time */ diff --git a/src/vnet/fib/ip4_fib.c b/src/vnet/fib/ip4_fib.c index 878b4dbf..d563bafd 100644 --- a/src/vnet/fib/ip4_fib.c +++ b/src/vnet/fib/ip4_fib.c @@ -531,10 +531,11 @@ ip4_show_fib (vlib_main_t * vm, if (fib_index != ~0 && fib_index != (int)fib->index) continue; - vlib_cli_output (vm, "%U, fib_index %d, flow hash: %U", + vlib_cli_output (vm, "%U, fib_index:%d, flow hash:[%U] locks:%d", format_fib_table_name, fib->index, FIB_PROTOCOL_IP4, fib->index, - format_ip_flow_hash_config, fib_table->ft_flow_hash_config); + format_ip_flow_hash_config, fib_table->ft_flow_hash_config, + fib_table->ft_locks); /* Show summary? */ if (! verbose) diff --git a/src/vnet/fib/ip6_fib.c b/src/vnet/fib/ip6_fib.c index e046b349..4a24c212 100644 --- a/src/vnet/fib/ip6_fib.c +++ b/src/vnet/fib/ip6_fib.c @@ -633,9 +633,10 @@ ip6_show_fib (vlib_main_t * vm, if (fib_index != ~0 && fib_index != (int)fib->index) continue; - vlib_cli_output (vm, "%s, fib_index %d, flow hash: %U", + vlib_cli_output (vm, "%s, fib_index:%d, flow hash:[%U] locks:%d", fib_table->ft_desc, fib->index, - format_ip_flow_hash_config, fib_table->ft_flow_hash_config); + format_ip_flow_hash_config, fib_table->ft_flow_hash_config, + fib_table->ft_locks); /* Show summary? */ if (! verbose) diff --git a/test/test_classifier.py b/test/test_classifier.py index faa107dc..a43f7a3d 100644 --- a/test/test_classifier.py +++ b/test/test_classifier.py @@ -64,7 +64,7 @@ class TestClassifier(VppTestCase): self.logger.info(self.vapi.cli("show classify table verbose")) self.logger.info(self.vapi.cli("show ip fib")) - def config_pbr_fib_entry(self, intf): + def config_pbr_fib_entry(self, intf, is_add=1): """Configure fib entry to route traffic toward PBR VRF table :param VppInterface intf: destination interface to be routed for PBR. @@ -74,7 +74,8 @@ class TestClassifier(VppTestCase): self.vapi.ip_add_del_route(intf.local_ip4n, addr_len, intf.remote_ip4n, - table_id=self.pbr_vrfid) + table_id=self.pbr_vrfid, + is_add=is_add) def create_stream(self, src_if, dst_if, packet_sizes): """Create input packet stream for defined interfaces. @@ -139,6 +140,25 @@ class TestClassifier(VppTestCase): "Interface %s: Packet expected from interface %s " "didn't arrive" % (dst_if.name, i.name)) + def verify_vrf(self, vrf_id): + """ + Check if the FIB table / VRF ID is configured. + + :param int vrf_id: The FIB table / VRF ID to be verified. + :return: 1 if the FIB table / VRF ID is configured, otherwise return 0. + """ + ip_fib_dump = self.vapi.ip_fib_dump() + vrf_count = 0 + for ip_fib_details in ip_fib_dump: + if ip_fib_details[2] == vrf_id: + vrf_count += 1 + if vrf_count == 0: + self.logger.info("IPv4 VRF ID %d is not configured" % vrf_id) + return 0 + else: + self.logger.info("IPv4 VRF ID %d is configured" % vrf_id) + return 1 + @staticmethod def build_ip_mask(proto='', src_ip='', dst_ip='', src_port='', dst_port=''): @@ -332,10 +352,12 @@ class TestClassifier(VppTestCase): 'pbr', self.build_ip_mask( src_ip='ffffffff')) pbr_option = 1 + # this will create the VRF/table in which we will insert the route self.create_classify_session( self.pg0, self.acl_tbl_idx.get('pbr'), self.build_ip_match(src_ip=self.pg0.remote_ip4), pbr_option, self.pbr_vrfid) + self.assertTrue(self.verify_vrf(self.pbr_vrfid)) self.config_pbr_fib_entry(self.pg3) self.input_acl_set_interface(self.pg0, self.acl_tbl_idx.get('pbr')) @@ -349,6 +371,15 @@ class TestClassifier(VppTestCase): self.pg1.assert_nothing_captured(remark="packets forwarded") self.pg2.assert_nothing_captured(remark="packets forwarded") + # remove the classify session and the route + self.config_pbr_fib_entry(self.pg3, is_add=0) + self.create_classify_session( + self.pg0, self.acl_tbl_idx.get('pbr'), + self.build_ip_match(src_ip=self.pg0.remote_ip4), + pbr_option, self.pbr_vrfid, is_add=0) + + # and the table should be gone. + self.assertFalse(self.verify_vrf(self.pbr_vrfid)) if __name__ == '__main__': unittest.main(testRunner=VppTestRunner) -- cgit 1.2.3-korg From 48113e0c864fd18d9fa65eda950aa363b813bb15 Mon Sep 17 00:00:00 2001 From: Dave Barach Date: Wed, 7 Jun 2017 08:32:51 -0400 Subject: VPP-874: fix coverity warning in vnet_classify.c Change-Id: Icffd2862eadbe9ddfb3ee34f3cb19c9324b3d9b4 Signed-off-by: Dave Barach --- src/vnet/classify/vnet_classify.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src/vnet/classify') diff --git a/src/vnet/classify/vnet_classify.c b/src/vnet/classify/vnet_classify.c index 99b10dc2..879fba3c 100644 --- a/src/vnet/classify/vnet_classify.c +++ b/src/vnet/classify/vnet_classify.c @@ -405,7 +405,7 @@ int vnet_classify_add_del (vnet_classify_table_t * t, u32 old_log2_pages, new_log2_pages; u32 thread_index = vlib_get_thread_index(); u8 * key_minus_skip; - int resplit_once; + int resplit_once = 0; int mark_bucket_linear; ASSERT ((add_v->flags & VNET_CLASSIFY_ENTRY_FREE) == 0); -- cgit 1.2.3-korg From 1500254bee11355bbd69cc1dd9705be4f002f2bd Mon Sep 17 00:00:00 2001 From: Neale Ranns Date: Sun, 10 Sep 2017 04:39:11 -0700 Subject: FIB table add/delete API part 2; - this adds the code to create an IP and MPLS table via the API. - but the enforcement that the table must be created before it is used is still missing, this is so that CSIT can pass. Change-Id: Id124d884ade6cb7da947225200e3bb193454c555 Signed-off-by: Neale Ranns --- src/plugins/nat/nat.c | 17 +- src/plugins/nat/nat64.c | 13 +- src/vnet/classify/vnet_classify.c | 16 +- src/vnet/dhcp/dhcp4_proxy_node.c | 9 +- src/vnet/dhcp/dhcp6_proxy_node.c | 9 +- src/vnet/dhcp/dhcp_proxy.c | 19 ++- src/vnet/dpo/lookup_dpo.c | 20 ++- src/vnet/dpo/mpls_label_dpo.c | 12 +- src/vnet/ethernet/arp.c | 127 +++++++++++---- src/vnet/fib/fib_api.h | 1 - src/vnet/fib/fib_entry.c | 15 +- src/vnet/fib/fib_entry.h | 1 + src/vnet/fib/fib_entry_src_mpls.c | 7 +- src/vnet/fib/fib_table.c | 43 +++-- src/vnet/fib/fib_table.h | 32 +++- src/vnet/fib/fib_test.c | 27 ++-- src/vnet/fib/ip4_fib.c | 41 +++-- src/vnet/fib/ip4_fib.h | 5 +- src/vnet/fib/ip6_fib.c | 41 +++-- src/vnet/fib/ip6_fib.h | 5 +- src/vnet/fib/mpls_fib.c | 16 +- src/vnet/fib/mpls_fib.h | 5 +- src/vnet/interface_api.c | 177 ++++++++++++++++---- src/vnet/ip/ip.h | 7 + src/vnet/ip/ip4.h | 13 ++ src/vnet/ip/ip4_forward.c | 101 +----------- src/vnet/ip/ip4_source_and_port_range_check.c | 11 +- src/vnet/ip/ip6.h | 13 ++ src/vnet/ip/ip6_forward.c | 103 +----------- src/vnet/ip/ip6_neighbor.c | 108 +++++++++---- src/vnet/ip/ip_api.c | 122 +++++++++++--- src/vnet/ip/lookup.c | 225 ++++++++++++++++++++++++++ src/vnet/lisp-gpe/interface.c | 11 +- src/vnet/lisp-gpe/lisp_gpe_fwd_entry.c | 9 +- src/vnet/lisp-gpe/lisp_gpe_sub_interface.c | 11 +- src/vnet/mfib/ip4_mfib.c | 12 +- src/vnet/mfib/ip4_mfib.h | 5 +- src/vnet/mfib/ip6_mfib.c | 12 +- src/vnet/mfib/ip6_mfib.h | 5 +- src/vnet/mfib/mfib_entry.c | 11 ++ src/vnet/mfib/mfib_entry.h | 2 + src/vnet/mfib/mfib_table.c | 88 ++++++++-- src/vnet/mfib/mfib_table.h | 29 +++- src/vnet/mfib/mfib_test.c | 11 +- src/vnet/mfib/mfib_types.h | 8 +- src/vnet/mpls/interface.c | 26 ++- src/vnet/mpls/mpls.c | 76 ++++++++- src/vnet/mpls/mpls.h | 16 +- src/vnet/mpls/mpls_api.c | 66 ++++++-- src/vnet/srv6/sr_policy_rewrite.c | 6 +- src/vnet/srv6/sr_steering.c | 6 +- src/vpp/api/api.c | 5 +- src/vpp/api/custom_dump.c | 3 - test/test_dhcp.py | 24 ++- test/test_gre.py | 8 +- test/test_ip4.py | 11 +- test/test_ip4_vrf_multi_instance.py | 4 +- test/test_ip6.py | 7 +- test/test_ip6_vrf_multi_instance.py | 4 +- test/test_ip_mcast.py | 98 ++++++++++- test/test_mpls.py | 48 +++++- test/test_nat.py | 13 ++ test/test_neighbor.py | 66 +++++++- test/vpp_ip_route.py | 73 +++++++++ test/vpp_papi_provider.py | 46 ++++-- 65 files changed, 1643 insertions(+), 538 deletions(-) (limited to 'src/vnet/classify') diff --git a/src/plugins/nat/nat.c b/src/plugins/nat/nat.c index aa7ef10a..8aecac6d 100644 --- a/src/plugins/nat/nat.c +++ b/src/plugins/nat/nat.c @@ -167,7 +167,8 @@ void snat_add_address (snat_main_t *sm, ip4_address_t *addr, u32 vrf_id) ap->addr = *addr; if (vrf_id != ~0) ap->fib_index = - fib_table_find_or_create_and_lock (FIB_PROTOCOL_IP4, vrf_id); + fib_table_find_or_create_and_lock (FIB_PROTOCOL_IP4, vrf_id, + FIB_SOURCE_PLUGIN_HI); else ap->fib_index = ~0; #define _(N, i, n, s) \ @@ -625,7 +626,8 @@ int nat44_add_del_lb_static_mapping (ip4_address_t e_addr, u16 e_port, return VNET_API_ERROR_INVALID_VALUE; fib_index = fib_table_find_or_create_and_lock (FIB_PROTOCOL_IP4, - vrf_id); + vrf_id, + FIB_SOURCE_PLUGIN_HI); /* Find external address in allocated addresses and reserve port for address and port pair mapping when dynamic translations enabled */ @@ -754,7 +756,7 @@ int nat44_add_del_lb_static_mapping (ip4_address_t e_addr, u16 e_port, if (!m) return VNET_API_ERROR_NO_SUCH_ENTRY; - fib_table_unlock (m->fib_index, FIB_PROTOCOL_IP4); + fib_table_unlock (m->fib_index, FIB_PROTOCOL_IP4, FIB_SOURCE_PLUGIN_HI); /* Free external address port */ if (!sm->static_mapping_only) @@ -874,7 +876,8 @@ int snat_del_address (snat_main_t *sm, ip4_address_t addr, u8 delete_sm) } if (a->fib_index != ~0) - fib_table_unlock(a->fib_index, FIB_PROTOCOL_IP4); + fib_table_unlock(a->fib_index, FIB_PROTOCOL_IP4, + FIB_SOURCE_PLUGIN_HI); /* Delete sessions using address */ if (a->busy_tcp_ports || a->busy_udp_ports || a->busy_icmp_ports) @@ -2151,10 +2154,12 @@ snat_config (vlib_main_t * vm, unformat_input_t * input) sm->max_translations_per_user = max_translations_per_user; sm->outside_vrf_id = outside_vrf_id; sm->outside_fib_index = fib_table_find_or_create_and_lock (FIB_PROTOCOL_IP4, - outside_vrf_id); + outside_vrf_id, + FIB_SOURCE_PLUGIN_HI); sm->inside_vrf_id = inside_vrf_id; sm->inside_fib_index = fib_table_find_or_create_and_lock (FIB_PROTOCOL_IP4, - inside_vrf_id); + inside_vrf_id, + FIB_SOURCE_PLUGIN_HI); sm->static_mapping_only = static_mapping_only; sm->static_mapping_connection_tracking = static_mapping_connection_tracking; diff --git a/src/plugins/nat/nat64.c b/src/plugins/nat/nat64.c index b04901fa..bfcfa9b3 100644 --- a/src/plugins/nat/nat64.c +++ b/src/plugins/nat/nat64.c @@ -107,7 +107,8 @@ nat64_add_del_pool_addr (ip4_address_t * addr, u32 vrf_id, u8 is_add) a->fib_index = 0; if (vrf_id != ~0) a->fib_index = - fib_table_find_or_create_and_lock (FIB_PROTOCOL_IP6, vrf_id); + fib_table_find_or_create_and_lock (FIB_PROTOCOL_IP6, vrf_id, + FIB_SOURCE_PLUGIN_HI); #define _(N, i, n, s) \ clib_bitmap_alloc (a->busy_##n##_port_bitmap, 65535); foreach_snat_protocol @@ -119,7 +120,8 @@ nat64_add_del_pool_addr (ip4_address_t * addr, u32 vrf_id, u8 is_add) return VNET_API_ERROR_NO_SUCH_ENTRY; if (a->fib_index) - fib_table_unlock (a->fib_index, FIB_PROTOCOL_IP6); + fib_table_unlock (a->fib_index, FIB_PROTOCOL_IP6, + FIB_SOURCE_PLUGIN_HI); #define _(N, id, n, s) \ clib_bitmap_free (a->busy_##n##_port_bitmap); @@ -353,8 +355,8 @@ nat64_add_del_static_bib_entry (ip6_address_t * in_addr, { nat64_main_t *nm = &nat64_main; nat64_db_bib_entry_t *bibe; - u32 fib_index = - fib_table_find_or_create_and_lock (FIB_PROTOCOL_IP6, vrf_id); + u32 fib_index = fib_table_find_or_create_and_lock (FIB_PROTOCOL_IP6, vrf_id, + FIB_SOURCE_PLUGIN_HI); snat_protocol_t p = ip_proto_to_snat_proto (proto); ip46_address_t addr; int i; @@ -644,7 +646,8 @@ nat64_add_del_prefix (ip6_address_t * prefix, u8 plen, u32 vrf_id, u8 is_add) { vec_add2 (nm->pref64, p, 1); p->fib_index = - fib_table_find_or_create_and_lock (FIB_PROTOCOL_IP6, vrf_id); + fib_table_find_or_create_and_lock (FIB_PROTOCOL_IP6, vrf_id, + FIB_SOURCE_PLUGIN_HI); p->vrf_id = vrf_id; } diff --git a/src/vnet/classify/vnet_classify.c b/src/vnet/classify/vnet_classify.c index 879fba3c..57d86748 100644 --- a/src/vnet/classify/vnet_classify.c +++ b/src/vnet/classify/vnet_classify.c @@ -368,10 +368,10 @@ vnet_classify_entry_claim_resource (vnet_classify_entry_t *e) switch (e->action) { case CLASSIFY_ACTION_SET_IP4_FIB_INDEX: - fib_table_lock (e->metadata, FIB_PROTOCOL_IP4); + fib_table_lock (e->metadata, FIB_PROTOCOL_IP4, FIB_SOURCE_CLASSIFY); break; case CLASSIFY_ACTION_SET_IP6_FIB_INDEX: - fib_table_lock (e->metadata, FIB_PROTOCOL_IP6); + fib_table_lock (e->metadata, FIB_PROTOCOL_IP6, FIB_SOURCE_CLASSIFY); break; } } @@ -382,10 +382,10 @@ vnet_classify_entry_release_resource (vnet_classify_entry_t *e) switch (e->action) { case CLASSIFY_ACTION_SET_IP4_FIB_INDEX: - fib_table_unlock (e->metadata, FIB_PROTOCOL_IP4); + fib_table_unlock (e->metadata, FIB_PROTOCOL_IP4, FIB_SOURCE_CLASSIFY); break; case CLASSIFY_ACTION_SET_IP6_FIB_INDEX: - fib_table_unlock (e->metadata, FIB_PROTOCOL_IP6); + fib_table_unlock (e->metadata, FIB_PROTOCOL_IP6, FIB_SOURCE_CLASSIFY); break; } } @@ -2096,9 +2096,13 @@ int vnet_classify_add_del_session (vnet_classify_main_t * cm, e->flags = 0; e->action = action; if (e->action == CLASSIFY_ACTION_SET_IP4_FIB_INDEX) - e->metadata = fib_table_find_or_create_and_lock (FIB_PROTOCOL_IP4, metadata); + e->metadata = fib_table_find_or_create_and_lock (FIB_PROTOCOL_IP4, + metadata, + FIB_SOURCE_CLASSIFY); else if (e->action == CLASSIFY_ACTION_SET_IP6_FIB_INDEX) - e->metadata = fib_table_find_or_create_and_lock (FIB_PROTOCOL_IP6, metadata); + e->metadata = fib_table_find_or_create_and_lock (FIB_PROTOCOL_IP6, + metadata, + FIB_SOURCE_CLASSIFY); else e->metadata = 0; diff --git a/src/vnet/dhcp/dhcp4_proxy_node.c b/src/vnet/dhcp/dhcp4_proxy_node.c index 1b59cdea..339a7885 100644 --- a/src/vnet/dhcp/dhcp4_proxy_node.c +++ b/src/vnet/dhcp/dhcp4_proxy_node.c @@ -785,7 +785,8 @@ dhcp4_proxy_set_server (ip46_address_t *addr, return VNET_API_ERROR_INVALID_SRC_ADDRESS; rx_fib_index = fib_table_find_or_create_and_lock(FIB_PROTOCOL_IP4, - rx_table_id); + rx_table_id, + FIB_SOURCE_DHCP); if (is_del) { @@ -795,7 +796,7 @@ dhcp4_proxy_set_server (ip46_address_t *addr, fib_table_entry_special_remove(rx_fib_index, &all_1s, FIB_SOURCE_DHCP); - fib_table_unlock (rx_fib_index, FIB_PROTOCOL_IP4); + fib_table_unlock (rx_fib_index, FIB_PROTOCOL_IP4, FIB_SOURCE_DHCP); } } else @@ -808,10 +809,10 @@ dhcp4_proxy_set_server (ip46_address_t *addr, &all_1s, FIB_SOURCE_DHCP, FIB_ENTRY_FLAG_LOCAL); - fib_table_lock (rx_fib_index, FIB_PROTOCOL_IP4); + fib_table_lock (rx_fib_index, FIB_PROTOCOL_IP4, FIB_SOURCE_DHCP); } } - fib_table_unlock (rx_fib_index, FIB_PROTOCOL_IP4); + fib_table_unlock (rx_fib_index, FIB_PROTOCOL_IP4, FIB_SOURCE_DHCP); return (rc); } diff --git a/src/vnet/dhcp/dhcp6_proxy_node.c b/src/vnet/dhcp/dhcp6_proxy_node.c index 9c2f5220..ce7a8fca 100644 --- a/src/vnet/dhcp/dhcp6_proxy_node.c +++ b/src/vnet/dhcp/dhcp6_proxy_node.c @@ -841,7 +841,8 @@ dhcp6_proxy_set_server (ip46_address_t *addr, return VNET_API_ERROR_INVALID_SRC_ADDRESS; rx_fib_index = mfib_table_find_or_create_and_lock(FIB_PROTOCOL_IP6, - rx_table_id); + rx_table_id, + MFIB_SOURCE_DHCP); if (is_del) { @@ -851,7 +852,7 @@ dhcp6_proxy_set_server (ip46_address_t *addr, mfib_table_entry_delete(rx_fib_index, &all_dhcp_servers, MFIB_SOURCE_DHCP); - mfib_table_unlock(rx_fib_index, FIB_PROTOCOL_IP6); + mfib_table_unlock(rx_fib_index, FIB_PROTOCOL_IP6, MFIB_SOURCE_DHCP); } } else @@ -885,11 +886,11 @@ dhcp6_proxy_set_server (ip46_address_t *addr, MFIB_SOURCE_DHCP, MFIB_RPF_ID_NONE, MFIB_ENTRY_FLAG_ACCEPT_ALL_ITF); - mfib_table_lock(rx_fib_index, FIB_PROTOCOL_IP6); + mfib_table_lock(rx_fib_index, FIB_PROTOCOL_IP6, MFIB_SOURCE_DHCP); } } - mfib_table_unlock(rx_fib_index, FIB_PROTOCOL_IP6); + mfib_table_unlock(rx_fib_index, FIB_PROTOCOL_IP6, MFIB_SOURCE_DHCP); return (rc); } diff --git a/src/vnet/dhcp/dhcp_proxy.c b/src/vnet/dhcp/dhcp_proxy.c index ba7f354e..1784906b 100644 --- a/src/vnet/dhcp/dhcp_proxy.c +++ b/src/vnet/dhcp/dhcp_proxy.c @@ -29,9 +29,9 @@ dhcp_proxy_rx_table_lock (fib_protocol_t proto, u32 fib_index) { if (FIB_PROTOCOL_IP4 == proto) - fib_table_lock(fib_index, proto); + fib_table_lock(fib_index, proto, FIB_SOURCE_DHCP); else - mfib_table_lock(fib_index, proto); + mfib_table_lock(fib_index, proto, MFIB_SOURCE_DHCP); } static void @@ -39,9 +39,9 @@ dhcp_proxy_rx_table_unlock (fib_protocol_t proto, u32 fib_index) { if (FIB_PROTOCOL_IP4 == proto) - fib_table_unlock(fib_index, proto); + fib_table_unlock(fib_index, proto, FIB_SOURCE_DHCP); else - mfib_table_unlock(fib_index, proto); + mfib_table_unlock(fib_index, proto, MFIB_SOURCE_DHCP); } u32 @@ -169,7 +169,7 @@ dhcp_proxy_server_del (fib_protocol_t proto, if (~0 != index) { server = &proxy->dhcp_servers[index]; - fib_table_unlock (server->server_fib_index, proto); + fib_table_unlock (server->server_fib_index, proto, FIB_SOURCE_DHCP); vec_del1(proxy->dhcp_servers, index); @@ -228,7 +228,8 @@ dhcp_proxy_server_add (fib_protocol_t proto, dhcp_server_t server = { .dhcp_server = *addr, .server_fib_index = fib_table_find_or_create_and_lock(proto, - server_table_id), + server_table_id, + FIB_SOURCE_DHCP), }; vec_add1(proxy->dhcp_servers, server); @@ -297,9 +298,11 @@ int dhcp_proxy_set_vss (fib_protocol_t proto, int rc = 0; if (proto == FIB_PROTOCOL_IP4) - rx_fib_index = fib_table_find_or_create_and_lock(proto, tbl_id); + rx_fib_index = fib_table_find_or_create_and_lock(proto, tbl_id, + FIB_SOURCE_DHCP); else - rx_fib_index = mfib_table_find_or_create_and_lock(proto, tbl_id); + rx_fib_index = mfib_table_find_or_create_and_lock(proto, tbl_id, + MFIB_SOURCE_DHCP); v = dhcp_get_vss_info(dm, rx_fib_index, proto); if (NULL != v) diff --git a/src/vnet/dpo/lookup_dpo.c b/src/vnet/dpo/lookup_dpo.c index 26363a2f..af189eda 100644 --- a/src/vnet/dpo/lookup_dpo.c +++ b/src/vnet/dpo/lookup_dpo.c @@ -135,11 +135,15 @@ lookup_dpo_add_or_lock_w_fib_index (fib_node_index_t fib_index, { if (LOOKUP_UNICAST == cast) { - fib_table_lock(fib_index, dpo_proto_to_fib(proto)); + fib_table_lock(fib_index, + dpo_proto_to_fib(proto), + FIB_SOURCE_RR); } else { - mfib_table_lock(fib_index, dpo_proto_to_fib(proto)); + mfib_table_lock(fib_index, + dpo_proto_to_fib(proto), + MFIB_SOURCE_RR); } } lookup_dpo_add_or_lock_i(fib_index, proto, cast, input, table_config, dpo); @@ -161,13 +165,15 @@ lookup_dpo_add_or_lock_w_table_id (u32 table_id, { fib_index = fib_table_find_or_create_and_lock(dpo_proto_to_fib(proto), - table_id); + table_id, + FIB_SOURCE_RR); } else { fib_index = mfib_table_find_or_create_and_lock(dpo_proto_to_fib(proto), - table_id); + table_id, + MFIB_SOURCE_RR); } } @@ -238,12 +244,14 @@ lookup_dpo_unlock (dpo_id_t *dpo) if (LOOKUP_UNICAST == lkd->lkd_cast) { fib_table_unlock(lkd->lkd_fib_index, - dpo_proto_to_fib(lkd->lkd_proto)); + dpo_proto_to_fib(lkd->lkd_proto), + FIB_SOURCE_RR); } else { mfib_table_unlock(lkd->lkd_fib_index, - dpo_proto_to_fib(lkd->lkd_proto)); + dpo_proto_to_fib(lkd->lkd_proto), + MFIB_SOURCE_RR); } } pool_put(lookup_dpo_pool, lkd); diff --git a/src/vnet/dpo/mpls_label_dpo.c b/src/vnet/dpo/mpls_label_dpo.c index b178a902..2a6e7dd5 100644 --- a/src/vnet/dpo/mpls_label_dpo.c +++ b/src/vnet/dpo/mpls_label_dpo.c @@ -105,10 +105,18 @@ format_mpls_label_dpo (u8 *s, va_list *args) mpls_label_dpo_t *mld; u32 ii; - mld = mpls_label_dpo_get(index); - s = format(s, "mpls-label:[%d]:", index); + if (pool_is_free_index(mpls_label_dpo_pool, index)) + { + /* + * the packet trace can be printed after the DPO has been deleted + */ + return (s); + } + + mld = mpls_label_dpo_get(index); + for (ii = 0; ii < mld->mld_n_labels; ii++) { hdr.label_exp_s_ttl = diff --git a/src/vnet/ethernet/arp.c b/src/vnet/ethernet/arp.c index c84ff47b..08e91373 100644 --- a/src/vnet/ethernet/arp.c +++ b/src/vnet/ethernet/arp.c @@ -522,6 +522,24 @@ arp_update_adjacency (vnet_main_t * vnm, u32 sw_if_index, u32 ai) } } +static void +arp_adj_fib_add (ethernet_arp_ip4_entry_t * e, uint32_t fib_index) +{ + fib_prefix_t pfx = { + .fp_len = 32, + .fp_proto = FIB_PROTOCOL_IP4, + .fp_addr.ip4 = e->ip4_address, + }; + + e->fib_entry_index = + fib_table_entry_path_add (fib_index, &pfx, FIB_SOURCE_ADJ, + FIB_ENTRY_FLAG_ATTACHED, + DPO_PROTO_IP4, &pfx.fp_addr, + e->sw_if_index, ~0, 1, NULL, + FIB_ROUTE_PATH_FLAG_NONE); + fib_table_lock (fib_index, FIB_PROTOCOL_IP4, FIB_SOURCE_ADJ); +} + int vnet_arp_set_ip4_over_ethernet_internal (vnet_main_t * vnm, vnet_arp_set_ip4_over_ethernet_rpc_args_t @@ -576,21 +594,9 @@ vnet_arp_set_ip4_over_ethernet_internal (vnet_main_t * vnm, if (!is_no_fib_entry) { - fib_prefix_t pfx = { - .fp_len = 32, - .fp_proto = FIB_PROTOCOL_IP4, - .fp_addr.ip4 = a->ip4, - }; - u32 fib_index; - - fib_index = - ip4_fib_table_get_index_for_sw_if_index (e->sw_if_index); - e->fib_entry_index = - fib_table_entry_path_add (fib_index, &pfx, FIB_SOURCE_ADJ, - FIB_ENTRY_FLAG_ATTACHED, - DPO_PROTO_IP4, &pfx.fp_addr, - e->sw_if_index, ~0, 1, NULL, - FIB_ROUTE_PATH_FLAG_NONE); + arp_adj_fib_add (e, + ip4_fib_table_get_index_for_sw_if_index + (e->sw_if_index)); } else { @@ -1561,6 +1567,65 @@ arp_add_del_interface_address (ip4_main_t * im, } } +void +arp_adj_fib_remove (ethernet_arp_ip4_entry_t * e, uint32_t fib_index) +{ + if (FIB_NODE_INDEX_INVALID != e->fib_entry_index) + { + fib_prefix_t pfx = { + .fp_len = 32, + .fp_proto = FIB_PROTOCOL_IP4, + .fp_addr.ip4 = e->ip4_address, + }; + u32 fib_index; + + fib_index = ip4_fib_table_get_index_for_sw_if_index (e->sw_if_index); + + fib_table_entry_path_remove (fib_index, &pfx, + FIB_SOURCE_ADJ, + DPO_PROTO_IP4, + &pfx.fp_addr, + e->sw_if_index, ~0, 1, + FIB_ROUTE_PATH_FLAG_NONE); + fib_table_unlock (fib_index, FIB_PROTOCOL_IP4, FIB_SOURCE_ADJ); + } +} + +static void +arp_table_bind (ip4_main_t * im, + uword opaque, + u32 sw_if_index, u32 new_fib_index, u32 old_fib_index) +{ + ethernet_arp_main_t *am = ðernet_arp_main; + ethernet_arp_interface_t *eai; + ethernet_arp_ip4_entry_t *e; + hash_pair_t *pair; + + /* + * the IP table that the interface is bound to has changed. + * reinstall all the adj fibs. + */ + + if (vec_len (am->ethernet_arp_by_sw_if_index) <= sw_if_index) + return; + + eai = &am->ethernet_arp_by_sw_if_index[sw_if_index]; + + /* *INDENT-OFF* */ + hash_foreach_pair (pair, eai->arp_entries, + ({ + e = pool_elt_at_index(am->ip4_entry_pool, + pair->value[0]); + /* + * remove the adj-fib from the old table and add to the new + */ + arp_adj_fib_remove(e, old_fib_index); + arp_adj_fib_add(e, new_fib_index); + })); + /* *INDENT-ON* */ + +} + static clib_error_t * ethernet_arp_init (vlib_main_t * vm) { @@ -1606,6 +1671,11 @@ ethernet_arp_init (vlib_main_t * vm) cb.function_opaque = 0; vec_add1 (im->add_del_interface_address_callbacks, cb); + ip4_table_bind_callback_t cbt; + cbt.function = arp_table_bind; + cbt.function_opaque = 0; + vec_add1 (im->table_bind_callbacks, cbt); + return 0; } @@ -1616,24 +1686,9 @@ arp_entry_free (ethernet_arp_interface_t * eai, ethernet_arp_ip4_entry_t * e) { ethernet_arp_main_t *am = ðernet_arp_main; - if (FIB_NODE_INDEX_INVALID != e->fib_entry_index) - { - fib_prefix_t pfx = { - .fp_len = 32, - .fp_proto = FIB_PROTOCOL_IP4, - .fp_addr.ip4 = e->ip4_address, - }; - u32 fib_index; - - fib_index = ip4_fib_table_get_index_for_sw_if_index (e->sw_if_index); - - fib_table_entry_path_remove (fib_index, &pfx, - FIB_SOURCE_ADJ, - DPO_PROTO_IP4, - &pfx.fp_addr, - e->sw_if_index, ~0, 1, - FIB_ROUTE_PATH_FLAG_NONE); - } + arp_adj_fib_remove (e, + ip4_fib_table_get_index_for_sw_if_index + (e->sw_if_index)); hash_unset (eai->arp_entries, e->ip4_address.as_u32); pool_put (am->ip4_entry_pool, e); } @@ -1693,7 +1748,11 @@ vnet_arp_flush_ip4_over_ethernet_internal (vnet_main_t * vnm, * does in response to interface events. unset is only done * by the control plane. */ - if (e->flags & ETHERNET_ARP_IP4_ENTRY_FLAG_DYNAMIC) + if (e->flags & ETHERNET_ARP_IP4_ENTRY_FLAG_STATIC) + { + e->flags &= ETHERNET_ARP_IP4_ENTRY_FLAG_DYNAMIC; + } + else if (e->flags & ETHERNET_ARP_IP4_ENTRY_FLAG_DYNAMIC) { arp_entry_free (eai, e); } diff --git a/src/vnet/fib/fib_api.h b/src/vnet/fib/fib_api.h index d07d6cae..f5a107ca 100644 --- a/src/vnet/fib/fib_api.h +++ b/src/vnet/fib/fib_api.h @@ -23,7 +23,6 @@ add_del_route_check (fib_protocol_t table_proto, u32 next_hop_sw_if_index, dpo_proto_t next_hop_table_proto, u32 next_hop_table_id, - u8 create_missing_tables, u8 is_rpf_id, u32 * fib_index, u32 * next_hop_fib_index); diff --git a/src/vnet/fib/fib_entry.c b/src/vnet/fib/fib_entry.c index 2027f2be..4cb6cf60 100644 --- a/src/vnet/fib/fib_entry.c +++ b/src/vnet/fib/fib_entry.c @@ -89,6 +89,17 @@ fib_entry_get_default_chain_type (const fib_entry_t *fib_entry) return (FIB_FORW_CHAIN_TYPE_UNICAST_IP4); } +u8 * +format_fib_source (u8 * s, va_list * args) +{ + fib_source_t source = va_arg (*args, int); + + s = format (s, "\n src:%s ", + fib_source_names[source]); + + return (s); +} + u8 * format_fib_entry (u8 * s, va_list * args) { @@ -114,8 +125,8 @@ format_fib_entry (u8 * s, va_list * args) FOR_EACH_SRC_ADDED(fib_entry, src, source, ({ - s = format (s, "\n src:%s ", - fib_source_names[source]); + s = format (s, "\n src:%U ", + format_fib_source, source); s = fib_entry_src_format(fib_entry, source, s); s = format (s, " refs:%d ", src->fes_ref_count); if (FIB_ENTRY_FLAG_NONE != src->fes_entry_flags) { diff --git a/src/vnet/fib/fib_entry.h b/src/vnet/fib/fib_entry.h index 93b8016d..2f6e37fe 100644 --- a/src/vnet/fib/fib_entry.h +++ b/src/vnet/fib/fib_entry.h @@ -431,6 +431,7 @@ typedef struct fib_entry_t_ { #define FIB_ENTRY_FORMAT_DETAIL2 (0x2) extern u8 *format_fib_entry (u8 * s, va_list * args); +extern u8 *format_fib_source (u8 * s, va_list * args); extern fib_node_index_t fib_entry_create_special(u32 fib_index, const fib_prefix_t *prefix, diff --git a/src/vnet/fib/fib_entry_src_mpls.c b/src/vnet/fib/fib_entry_src_mpls.c index a616458f..6fdd5c0a 100644 --- a/src/vnet/fib/fib_entry_src_mpls.c +++ b/src/vnet/fib/fib_entry_src_mpls.c @@ -94,7 +94,9 @@ fib_entry_src_mpls_set_data (fib_entry_src_t *src, fib_table_entry_delete_index(src->mpls.fesm_lfes[eos], FIB_SOURCE_SPECIAL); } - fib_table_unlock(MPLS_FIB_DEFAULT_TABLE_ID, FIB_PROTOCOL_MPLS); + fib_table_unlock(MPLS_FIB_DEFAULT_TABLE_ID, + FIB_PROTOCOL_MPLS, + FIB_SOURCE_MPLS); src->mpls.fesm_label = label; } else @@ -113,7 +115,8 @@ fib_entry_src_mpls_set_data (fib_entry_src_t *src, { fib_index = fib_table_find_or_create_and_lock(FIB_PROTOCOL_MPLS, - MPLS_FIB_DEFAULT_TABLE_ID); + MPLS_FIB_DEFAULT_TABLE_ID, + FIB_SOURCE_MPLS); } else { diff --git a/src/vnet/fib/fib_table.c b/src/vnet/fib/fib_table.c index 6b6cc5cb..75d15628 100644 --- a/src/vnet/fib/fib_table.c +++ b/src/vnet/fib/fib_table.c @@ -1039,7 +1039,8 @@ fib_table_find (fib_protocol_t proto, u32 fib_table_find_or_create_and_lock (fib_protocol_t proto, - u32 table_id) + u32 table_id, + fib_source_t src) { fib_table_t *fib_table; fib_node_index_t fi; @@ -1047,13 +1048,13 @@ fib_table_find_or_create_and_lock (fib_protocol_t proto, switch (proto) { case FIB_PROTOCOL_IP4: - fi = ip4_fib_table_find_or_create_and_lock(table_id); + fi = ip4_fib_table_find_or_create_and_lock(table_id, src); break; case FIB_PROTOCOL_IP6: - fi = ip6_fib_table_find_or_create_and_lock(table_id); + fi = ip6_fib_table_find_or_create_and_lock(table_id, src); break; case FIB_PROTOCOL_MPLS: - fi = mpls_fib_table_find_or_create_and_lock(table_id); + fi = mpls_fib_table_find_or_create_and_lock(table_id, src); break; default: return (~0); @@ -1070,6 +1071,7 @@ fib_table_find_or_create_and_lock (fib_protocol_t proto, u32 fib_table_create_and_lock (fib_protocol_t proto, + fib_source_t src, const char *const fmt, ...) { @@ -1082,13 +1084,13 @@ fib_table_create_and_lock (fib_protocol_t proto, switch (proto) { case FIB_PROTOCOL_IP4: - fi = ip4_fib_table_create_and_lock(); + fi = ip4_fib_table_create_and_lock(src); break; case FIB_PROTOCOL_IP6: - fi = ip6_fib_table_create_and_lock(); + fi = ip6_fib_table_create_and_lock(src); break; case FIB_PROTOCOL_MPLS: - fi = mpls_fib_table_create_and_lock(); + fi = mpls_fib_table_create_and_lock(src); break; default: return (~0); @@ -1143,26 +1145,43 @@ fib_table_walk (u32 fib_index, void fib_table_unlock (u32 fib_index, - fib_protocol_t proto) + fib_protocol_t proto, + fib_source_t source) { fib_table_t *fib_table; fib_table = fib_table_get(fib_index, proto); - fib_table->ft_locks--; + fib_table->ft_locks[source]--; + fib_table->ft_locks[FIB_TABLE_TOTAL_LOCKS]--; - if (0 == fib_table->ft_locks) + if (0 == fib_table->ft_locks[source]) { + /* + * The source no longer needs the table. flush any routes + * from it just in case + */ + fib_table_flush(fib_index, proto, source); + } + + if (0 == fib_table->ft_locks[FIB_TABLE_TOTAL_LOCKS]) + { + /* + * no more locak from any source - kill it + */ fib_table_destroy(fib_table); } } + void fib_table_lock (u32 fib_index, - fib_protocol_t proto) + fib_protocol_t proto, + fib_source_t source) { fib_table_t *fib_table; fib_table = fib_table_get(fib_index, proto); - fib_table->ft_locks++; + fib_table->ft_locks[source]++; + fib_table->ft_locks[FIB_TABLE_TOTAL_LOCKS]++; } u32 diff --git a/src/vnet/fib/fib_table.h b/src/vnet/fib/fib_table.h index 579740e9..6b7011b3 100644 --- a/src/vnet/fib/fib_table.h +++ b/src/vnet/fib/fib_table.h @@ -22,6 +22,12 @@ #include #include +/** + * Keep a lock per-source and a total + */ +#define FIB_TABLE_N_LOCKS (FIB_SOURCE_MAX+1) +#define FIB_TABLE_TOTAL_LOCKS FIB_SOURCE_MAX + /** * @brief * A protocol Independent FIB table @@ -34,9 +40,9 @@ typedef struct fib_table_t_ fib_protocol_t ft_proto; /** - * number of locks on the table + * per-source number of locks on the table */ - u16 ft_locks; + u16 ft_locks[FIB_TABLE_N_LOCKS]; /** * Table ID (hash key) for this FIB. @@ -628,9 +634,13 @@ extern u32 fib_table_find(fib_protocol_t proto, u32 table_id); * * @return fib_index * The index of the FIB + * + * @param source + * The ID of the client/source. */ extern u32 fib_table_find_or_create_and_lock(fib_protocol_t proto, - u32 table_id); + u32 table_id, + fib_source_t source); /** * @brief @@ -643,10 +653,14 @@ extern u32 fib_table_find_or_create_and_lock(fib_protocol_t proto, * @param fmt * A string to describe the table * + * @param source + * The ID of the client/source. + * * @return fib_index * The index of the FIB */ extern u32 fib_table_create_and_lock(fib_protocol_t proto, + fib_source_t source, const char *const fmt, ...); @@ -704,9 +718,13 @@ extern void fib_table_set_flow_hash_config(u32 fib_index, * * @paran proto * The protocol of the FIB (and thus the entries therein) + * + * @param source + * The ID of the client/source. */ extern void fib_table_unlock(u32 fib_index, - fib_protocol_t proto); + fib_protocol_t proto, + fib_source_t source); /** * @brief @@ -718,9 +736,13 @@ extern void fib_table_unlock(u32 fib_index, * * @paran proto * The protocol of the FIB (and thus the entries therein) + * + * @param source + * The ID of the client/source. */ extern void fib_table_lock(u32 fib_index, - fib_protocol_t proto); + fib_protocol_t proto, + fib_source_t source); /** * @brief diff --git a/src/vnet/fib/fib_test.c b/src/vnet/fib/fib_test.c index 6867cca8..572d7f0d 100644 --- a/src/vnet/fib/fib_test.c +++ b/src/vnet/fib/fib_test.c @@ -739,7 +739,8 @@ fib_test_v4 (void) lb_count = pool_elts(load_balance_pool); /* Find or create FIB table 11 */ - fib_index = fib_table_find_or_create_and_lock(FIB_PROTOCOL_IP4, 11); + fib_index = fib_table_find_or_create_and_lock(FIB_PROTOCOL_IP4, 11, + FIB_SOURCE_API); for (ii = 0; ii < 4; ii++) { @@ -4150,7 +4151,7 @@ fib_test_v4 (void) FIB_SOURCE_INTERFACE)), "NO INterface Source'd prefixes"); - fib_table_unlock(fib_index, FIB_PROTOCOL_IP4); + fib_table_unlock(fib_index, FIB_PROTOCOL_IP4, FIB_SOURCE_API); FIB_TEST((0 == fib_path_list_db_size()), "path list DB population:%d", fib_path_list_db_size()); @@ -4201,7 +4202,8 @@ fib_test_v6 (void) dpo_drop = drop_dpo_get(DPO_PROTO_IP6); /* Find or create FIB table 11 */ - fib_index = fib_table_find_or_create_and_lock(FIB_PROTOCOL_IP6, 11); + fib_index = fib_table_find_or_create_and_lock(FIB_PROTOCOL_IP6, 11, + FIB_SOURCE_API); for (ii = 0; ii < 4; ii++) { @@ -5025,7 +5027,7 @@ fib_test_v6 (void) /* * now remove the VRF */ - fib_table_unlock(fib_index, FIB_PROTOCOL_IP6); + fib_table_unlock(fib_index, FIB_PROTOCOL_IP6, FIB_SOURCE_API); FIB_TEST((0 == fib_path_list_db_size()), "path list DB population:%d", fib_path_list_db_size()); @@ -5157,7 +5159,9 @@ fib_test_ae (void) */ u32 import_fib_index1; - import_fib_index1 = fib_table_find_or_create_and_lock(FIB_PROTOCOL_IP4, 11); + import_fib_index1 = fib_table_find_or_create_and_lock(FIB_PROTOCOL_IP4, + 11, + FIB_SOURCE_CLI); /* * Add an attached route in the import FIB @@ -5233,7 +5237,8 @@ fib_test_ae (void) */ u32 import_fib_index2; - import_fib_index2 = fib_table_find_or_create_and_lock(FIB_PROTOCOL_IP4, 12); + import_fib_index2 = fib_table_find_or_create_and_lock(FIB_PROTOCOL_IP4, 12, + FIB_SOURCE_CLI); /* * Add an attached route in the import FIB @@ -5595,8 +5600,8 @@ fib_test_ae (void) &local_pfx, FIB_SOURCE_API); - fib_table_unlock(import_fib_index1, FIB_PROTOCOL_IP4); - fib_table_unlock(import_fib_index2, FIB_PROTOCOL_IP4); + fib_table_unlock(import_fib_index1, FIB_PROTOCOL_IP4, FIB_SOURCE_CLI); + fib_table_unlock(import_fib_index2, FIB_PROTOCOL_IP4, FIB_SOURCE_CLI); FIB_TEST((0 == adj_nbr_db_size()), "ADJ DB size is %d", adj_nbr_db_size()); @@ -8168,9 +8173,10 @@ lfib_test (void) /* * MPLS enable an interface so we get the MPLS table created */ + mpls_table_create(MPLS_FIB_DEFAULT_TABLE_ID, FIB_SOURCE_API); mpls_sw_interface_enable_disable(&mpls_main, tm->hw[0]->sw_if_index, - 1); + 1, 1); ip46_address_t nh_10_10_10_1 = { .ip4.as_u32 = clib_host_to_net_u32(0x0a0a0a01), @@ -8662,7 +8668,8 @@ lfib_test (void) */ mpls_sw_interface_enable_disable(&mpls_main, tm->hw[0]->sw_if_index, - 0); + 0, 1); + mpls_table_delete(MPLS_FIB_DEFAULT_TABLE_ID, FIB_SOURCE_API); FIB_TEST(lb_count == pool_elts(load_balance_pool), "Load-balance resources freed %d of %d", diff --git a/src/vnet/fib/ip4_fib.c b/src/vnet/fib/ip4_fib.c index d563bafd..865e2dd5 100644 --- a/src/vnet/fib/ip4_fib.c +++ b/src/vnet/fib/ip4_fib.c @@ -101,7 +101,8 @@ static const ip4_fib_table_special_prefix_t ip4_specials[] = { static u32 -ip4_create_fib_with_table_id (u32 table_id) +ip4_create_fib_with_table_id (u32 table_id, + fib_source_t src) { fib_table_t *fib_table; ip4_fib_t *v4_fib; @@ -128,7 +129,7 @@ ip4_create_fib_with_table_id (u32 table_id) v4_fib->fwd_classify_table_index = ~0; v4_fib->rev_classify_table_index = ~0; - fib_table_lock(fib_table->ft_index, FIB_PROTOCOL_IP4); + fib_table_lock(fib_table->ft_index, FIB_PROTOCOL_IP4, src); ip4_mtrie_init(&v4_fib->mtrie); @@ -198,23 +199,24 @@ ip4_fib_table_destroy (u32 fib_index) u32 -ip4_fib_table_find_or_create_and_lock (u32 table_id) +ip4_fib_table_find_or_create_and_lock (u32 table_id, + fib_source_t src) { u32 index; index = ip4_fib_index_from_table_id(table_id); if (~0 == index) - return ip4_create_fib_with_table_id(table_id); + return ip4_create_fib_with_table_id(table_id, src); - fib_table_lock(index, FIB_PROTOCOL_IP4); + fib_table_lock(index, FIB_PROTOCOL_IP4, src); return (index); } u32 -ip4_fib_table_create_and_lock (void) +ip4_fib_table_create_and_lock (fib_source_t src) { - return (ip4_create_fib_with_table_id(~0)); + return (ip4_create_fib_with_table_id(~0, src)); } u32 @@ -525,17 +527,32 @@ ip4_show_fib (vlib_main_t * vm, pool_foreach (fib_table, im4->fibs, ({ ip4_fib_t *fib = pool_elt_at_index(im4->v4_fibs, fib_table->ft_index); + fib_source_t source; + u8 *s = NULL; if (table_id >= 0 && table_id != (int)fib->table_id) continue; if (fib_index != ~0 && fib_index != (int)fib->index) continue; - vlib_cli_output (vm, "%U, fib_index:%d, flow hash:[%U] locks:%d", - format_fib_table_name, fib->index, FIB_PROTOCOL_IP4, - fib->index, - format_ip_flow_hash_config, fib_table->ft_flow_hash_config, - fib_table->ft_locks); + s = format(s, "%U, fib_index:%d, flow hash:[%U] locks:[", + format_fib_table_name, fib->index, + FIB_PROTOCOL_IP4, + fib->index, + format_ip_flow_hash_config, + fib_table->ft_flow_hash_config); + FOR_EACH_FIB_SOURCE(source) + { + if (0 != fib_table->ft_locks[source]) + { + s = format(s, "%U:%d, ", + format_fib_source, source, + fib_table->ft_locks[source]); + } + } + s = format (s, "]"); + vlib_cli_output (vm, "%V", s); + vec_free(s); /* Show summary? */ if (! verbose) diff --git a/src/vnet/fib/ip4_fib.h b/src/vnet/fib/ip4_fib.h index 006163b4..495b45cc 100644 --- a/src/vnet/fib/ip4_fib.h +++ b/src/vnet/fib/ip4_fib.h @@ -127,8 +127,9 @@ ip4_fib_lookup (ip4_main_t * im, u32 sw_if_index, ip4_address_t * dst) * @returns A pointer to the retrieved or created fib. * */ -extern u32 ip4_fib_table_find_or_create_and_lock(u32 table_id); -extern u32 ip4_fib_table_create_and_lock(void); +extern u32 ip4_fib_table_find_or_create_and_lock(u32 table_id, + fib_source_t src); +extern u32 ip4_fib_table_create_and_lock(fib_source_t src); static inline diff --git a/src/vnet/fib/ip6_fib.c b/src/vnet/fib/ip6_fib.c index 8fde6f9f..3ddb8453 100644 --- a/src/vnet/fib/ip6_fib.c +++ b/src/vnet/fib/ip6_fib.c @@ -50,7 +50,8 @@ vnet_ip6_fib_init (u32 fib_index) } static u32 -create_fib_with_table_id (u32 table_id) +create_fib_with_table_id (u32 table_id, + fib_source_t src) { fib_table_t *fib_table; ip6_fib_t *v6_fib; @@ -77,29 +78,30 @@ create_fib_with_table_id (u32 table_id) fib_table->ft_flow_hash_config = IP_FLOW_HASH_DEFAULT; vnet_ip6_fib_init(fib_table->ft_index); - fib_table_lock(fib_table->ft_index, FIB_PROTOCOL_IP6); + fib_table_lock(fib_table->ft_index, FIB_PROTOCOL_IP6, src); return (fib_table->ft_index); } u32 -ip6_fib_table_find_or_create_and_lock (u32 table_id) +ip6_fib_table_find_or_create_and_lock (u32 table_id, + fib_source_t src) { uword * p; p = hash_get (ip6_main.fib_index_by_table_id, table_id); if (NULL == p) - return create_fib_with_table_id(table_id); + return create_fib_with_table_id(table_id, src); - fib_table_lock(p[0], FIB_PROTOCOL_IP6); + fib_table_lock(p[0], FIB_PROTOCOL_IP6, src); return (p[0]); } u32 -ip6_fib_table_create_and_lock (void) +ip6_fib_table_create_and_lock (fib_source_t src) { - return (create_fib_with_table_id(~0)); + return (create_fib_with_table_id(~0, src)); } void @@ -588,16 +590,33 @@ ip6_show_fib (vlib_main_t * vm, pool_foreach (fib_table, im6->fibs, ({ + fib_source_t source; + u8 *s = NULL; + fib = pool_elt_at_index(im6->v6_fibs, fib_table->ft_index); if (table_id >= 0 && table_id != (int)fib->table_id) continue; if (fib_index != ~0 && fib_index != (int)fib->index) continue; - vlib_cli_output (vm, "%s, fib_index:%d, flow hash:[%U] locks:%d", - fib_table->ft_desc, fib->index, - format_ip_flow_hash_config, fib_table->ft_flow_hash_config, - fib_table->ft_locks); + s = format(s, "%U, fib_index:%d, flow hash:[%U] locks:[", + format_fib_table_name, fib->index, + FIB_PROTOCOL_IP6, + fib->index, + format_ip_flow_hash_config, + fib_table->ft_flow_hash_config); + FOR_EACH_FIB_SOURCE(source) + { + if (0 != fib_table->ft_locks[source]) + { + s = format(s, "%U:%d, ", + format_fib_source, source, + fib_table->ft_locks[source]); + } + } + s = format (s, "]"); + vlib_cli_output (vm, "%V", s); + vec_free(s); /* Show summary? */ if (! verbose) diff --git a/src/vnet/fib/ip6_fib.h b/src/vnet/fib/ip6_fib.h index aad8305c..9728eecc 100644 --- a/src/vnet/fib/ip6_fib.h +++ b/src/vnet/fib/ip6_fib.h @@ -144,8 +144,9 @@ ip6_src_lookup_for_packet (ip6_main_t * im, * \returns A pointer to the retrieved or created fib. * */ -extern u32 ip6_fib_table_find_or_create_and_lock(u32 table_id); -extern u32 ip6_fib_table_create_and_lock(void); +extern u32 ip6_fib_table_find_or_create_and_lock(u32 table_id, + fib_source_t src); +extern u32 ip6_fib_table_create_and_lock(fib_source_t src); static inline ip6_fib_t * ip6_fib_get (fib_node_index_t index) diff --git a/src/vnet/fib/mpls_fib.c b/src/vnet/fib/mpls_fib.c index ca6271fe..4eeef7ab 100644 --- a/src/vnet/fib/mpls_fib.c +++ b/src/vnet/fib/mpls_fib.c @@ -83,7 +83,8 @@ mpls_fib_index_from_table_id (u32 table_id) } static u32 -mpls_fib_create_with_table_id (u32 table_id) +mpls_fib_create_with_table_id (u32 table_id, + fib_source_t src) { dpo_id_t dpo = DPO_INVALID; fib_table_t *fib_table; @@ -107,7 +108,7 @@ mpls_fib_create_with_table_id (u32 table_id) fib_table->ft_table_id = table_id; fib_table->ft_flow_hash_config = MPLS_FLOW_HASH_DEFAULT; - fib_table_lock(fib_table->ft_index, FIB_PROTOCOL_MPLS); + fib_table_lock(fib_table->ft_index, FIB_PROTOCOL_MPLS, src); if (INDEX_INVALID == mpls_fib_drop_dpo_index) { @@ -220,22 +221,23 @@ mpls_fib_create_with_table_id (u32 table_id) } u32 -mpls_fib_table_find_or_create_and_lock (u32 table_id) +mpls_fib_table_find_or_create_and_lock (u32 table_id, + fib_source_t src) { u32 index; index = mpls_fib_index_from_table_id(table_id); if (~0 == index) - return mpls_fib_create_with_table_id(table_id); + return mpls_fib_create_with_table_id(table_id, src); - fib_table_lock(index, FIB_PROTOCOL_MPLS); + fib_table_lock(index, FIB_PROTOCOL_MPLS, src); return (index); } u32 -mpls_fib_table_create_and_lock (void) +mpls_fib_table_create_and_lock (fib_source_t src) { - return (mpls_fib_create_with_table_id(~0)); + return (mpls_fib_create_with_table_id(~0, src)); } void diff --git a/src/vnet/fib/mpls_fib.h b/src/vnet/fib/mpls_fib.h index dfb8b7fc..29cd1d20 100644 --- a/src/vnet/fib/mpls_fib.h +++ b/src/vnet/fib/mpls_fib.h @@ -59,8 +59,9 @@ mpls_fib_get (fib_node_index_t index) return (pool_elt_at_index(mpls_main.mpls_fibs, index)); } -extern u32 mpls_fib_table_find_or_create_and_lock(u32 table_id); -extern u32 mpls_fib_table_create_and_lock(void); +extern u32 mpls_fib_table_find_or_create_and_lock(u32 table_id, + fib_source_t src); +extern u32 mpls_fib_table_create_and_lock(fib_source_t src); // extern mpls_fib_t * mpls_fib_find(u32 table_id); extern u32 mpls_fib_index_from_table_id(u32 table_id); diff --git a/src/vnet/interface_api.c b/src/vnet/interface_api.c index 113728cd..419fef94 100644 --- a/src/vnet/interface_api.c +++ b/src/vnet/interface_api.c @@ -320,68 +320,189 @@ stats_dsunlock (void) static void vl_api_sw_interface_set_table_t_handler (vl_api_sw_interface_set_table_t * mp) { - int rv = 0; - u32 table_id = ntohl (mp->vrf_id); - u32 sw_if_index = ntohl (mp->sw_if_index); vl_api_sw_interface_set_table_reply_t *rmp; - CLIB_UNUSED (ip_interface_address_t * ia); - u32 fib_index; + u32 sw_if_index = ntohl (mp->sw_if_index); + u32 table_id = ntohl (mp->vrf_id); + int rv = 0; VALIDATE_SW_IF_INDEX (mp); stats_dslock_with_hint (1 /* release hint */ , 4 /* tag */ ); if (mp->is_ipv6) + rv = ip_table_bind (FIB_PROTOCOL_IP6, sw_if_index, table_id, 1); + else + rv = ip_table_bind (FIB_PROTOCOL_IP4, sw_if_index, table_id, 1); + + stats_dsunlock (); + + BAD_SW_IF_INDEX_LABEL; + + REPLY_MACRO (VL_API_SW_INTERFACE_SET_TABLE_REPLY); +} + +int +ip_table_bind (fib_protocol_t fproto, + uint32_t sw_if_index, uint32_t table_id, u8 is_api) +{ + CLIB_UNUSED (ip_interface_address_t * ia); + u32 fib_index, mfib_index; + fib_source_t src; + mfib_source_t msrc; + + if (is_api) + { + src = FIB_SOURCE_API; + msrc = MFIB_SOURCE_API; + } + else + { + src = FIB_SOURCE_CLI; + msrc = MFIB_SOURCE_CLI; + } + + /* + * This is temporary whilst I do the song and dance with the CSIT version + */ + if (0 != table_id) { + fib_index = fib_table_find_or_create_and_lock (fproto, table_id, src); + mfib_index = + mfib_table_find_or_create_and_lock (fproto, table_id, msrc); + } + else + { + fib_index = 0; + mfib_index = 0; + } + + /* + * This if table does not exist = error is what we want in the end. + */ + /* fib_index = fib_table_find (fproto, table_id); */ + /* mfib_index = mfib_table_find (fproto, table_id); */ + + /* if (~0 == fib_index || ~0 == mfib_index) */ + /* { */ + /* return (VNET_API_ERROR_NO_SUCH_FIB); */ + /* } */ + + if (FIB_PROTOCOL_IP6 == fproto) + { + /* + * If the interface already has in IP address, then a change int + * VRF is not allowed. The IP address applied must first be removed. + * We do not do that automatically here, since VPP has no knowledge + * of whether thoses subnets are valid in the destination VRF. + */ /* *INDENT-OFF* */ foreach_ip_interface_address (&ip6_main.lookup_main, ia, sw_if_index, 1 /* honor unnumbered */ , ({ - rv = VNET_API_ERROR_ADDRESS_FOUND_FOR_INTERFACE; - goto done; + return (VNET_API_ERROR_ADDRESS_FOUND_FOR_INTERFACE); })); /* *INDENT-ON* */ - fib_index = fib_table_find_or_create_and_lock (FIB_PROTOCOL_IP6, - table_id); vec_validate (ip6_main.fib_index_by_sw_if_index, sw_if_index); - ip6_main.fib_index_by_sw_if_index[sw_if_index] = fib_index; - - fib_index = mfib_table_find_or_create_and_lock (FIB_PROTOCOL_IP6, - table_id); vec_validate (ip6_main.mfib_index_by_sw_if_index, sw_if_index); - ip6_main.mfib_index_by_sw_if_index[sw_if_index] = fib_index; + + /* + * tell those that are interested that the binding is changing. + */ + ip6_table_bind_callback_t *cb; + vec_foreach (cb, ip6_main.table_bind_callbacks) + cb->function (&ip6_main, cb->function_opaque, + sw_if_index, + fib_index, + ip6_main.fib_index_by_sw_if_index[sw_if_index]); + + if (0 == table_id) + { + /* reset back to default */ + if (0 != ip6_main.fib_index_by_sw_if_index[sw_if_index]) + fib_table_unlock (ip6_main.fib_index_by_sw_if_index[sw_if_index], + FIB_PROTOCOL_IP6, src); + if (0 != ip6_main.mfib_index_by_sw_if_index[sw_if_index]) + mfib_table_unlock (ip6_main.mfib_index_by_sw_if_index + [sw_if_index], FIB_PROTOCOL_IP6, msrc); + + } + else + { + /* we need to lock the table now it's inuse */ + fib_table_lock (fib_index, FIB_PROTOCOL_IP6, src); + mfib_table_lock (mfib_index, FIB_PROTOCOL_IP6, msrc); + } + + ip6_main.fib_index_by_sw_if_index[sw_if_index] = fib_index; + ip6_main.mfib_index_by_sw_if_index[sw_if_index] = mfib_index; } else { + /* + * If the interface already has in IP address, then a change int + * VRF is not allowed. The IP address applied must first be removed. + * We do not do that automatically here, since VPP has no knowledge + * of whether thoses subnets are valid in the destination VRF. + */ /* *INDENT-OFF* */ foreach_ip_interface_address (&ip4_main.lookup_main, ia, sw_if_index, 1 /* honor unnumbered */ , ({ - rv = VNET_API_ERROR_ADDRESS_FOUND_FOR_INTERFACE; - goto done; + return (VNET_API_ERROR_ADDRESS_FOUND_FOR_INTERFACE); })); /* *INDENT-ON* */ - fib_index = fib_table_find_or_create_and_lock (FIB_PROTOCOL_IP4, - table_id); vec_validate (ip4_main.fib_index_by_sw_if_index, sw_if_index); - ip4_main.fib_index_by_sw_if_index[sw_if_index] = fib_index; - - fib_index = mfib_table_find_or_create_and_lock (FIB_PROTOCOL_IP4, - table_id); vec_validate (ip4_main.mfib_index_by_sw_if_index, sw_if_index); - ip4_main.mfib_index_by_sw_if_index[sw_if_index] = fib_index; - } -done: - stats_dsunlock (); + /* + * tell those that are interested that the binding is changing. + */ + ip4_table_bind_callback_t *cb; + vec_foreach (cb, ip4_main.table_bind_callbacks) + cb->function (&ip4_main, cb->function_opaque, + sw_if_index, + fib_index, + ip4_main.fib_index_by_sw_if_index[sw_if_index]); + + if (0 == table_id) + { + /* reset back to default */ + if (0 != ip4_main.fib_index_by_sw_if_index[sw_if_index]) + fib_table_unlock (ip4_main.fib_index_by_sw_if_index[sw_if_index], + FIB_PROTOCOL_IP4, src); + if (0 != ip4_main.mfib_index_by_sw_if_index[sw_if_index]) + mfib_table_unlock (ip4_main.mfib_index_by_sw_if_index + [sw_if_index], FIB_PROTOCOL_IP4, msrc); - BAD_SW_IF_INDEX_LABEL; + } + else + { + /* we need to lock the table now it's inuse */ + fib_index = fib_table_find_or_create_and_lock (FIB_PROTOCOL_IP4, + table_id, src); - REPLY_MACRO (VL_API_SW_INTERFACE_SET_TABLE_REPLY); + mfib_index = mfib_table_find_or_create_and_lock (FIB_PROTOCOL_IP4, + table_id, msrc); + } + + ip4_main.fib_index_by_sw_if_index[sw_if_index] = fib_index; + ip4_main.mfib_index_by_sw_if_index[sw_if_index] = mfib_index; + } + + /* + * Temporary. undo the locks from the find and create at the staart + */ + if (0 != table_id) + { + fib_table_unlock (fib_index, fproto, src); + mfib_table_unlock (mfib_index, fproto, msrc); + } + + return (0); } static void diff --git a/src/vnet/ip/ip.h b/src/vnet/ip/ip.h index 70b4ccd8..7aae73ff 100644 --- a/src/vnet/ip/ip.h +++ b/src/vnet/ip/ip.h @@ -184,6 +184,13 @@ void ip_del_all_interface_addresses (vlib_main_t * vm, u32 sw_if_index); extern vlib_node_registration_t ip4_inacl_node; extern vlib_node_registration_t ip6_inacl_node; +void ip_table_create (fib_protocol_t fproto, uint32_t table_id, u8 is_api); + +void ip_table_delete (fib_protocol_t fproto, uint32_t table_id, u8 is_api); + +int ip_table_bind (fib_protocol_t fproto, + uint32_t sw_if_index, uint32_t table_id, u8 is_api); + #endif /* included_ip_main_h */ /* diff --git a/src/vnet/ip/ip4.h b/src/vnet/ip/ip4.h index 8f9a8e27..decb840b 100644 --- a/src/vnet/ip/ip4.h +++ b/src/vnet/ip/ip4.h @@ -72,6 +72,16 @@ typedef struct uword function_opaque; } ip4_add_del_interface_address_callback_t; +typedef void (ip4_table_bind_function_t) + (struct ip4_main_t * im, + uword opaque, u32 sw_if_index, u32 new_fib_index, u32 old_fib_index); + +typedef struct +{ + ip4_table_bind_function_t *function; + uword function_opaque; +} ip4_table_bind_callback_t; + /** * @brief IPv4 main type. * @@ -117,6 +127,9 @@ typedef struct ip4_main_t ip4_add_del_interface_address_callback_t * add_del_interface_address_callbacks; + /** Functions to call when interface to table biding changes. */ + ip4_table_bind_callback_t *table_bind_callbacks; + /** Template used to generate IP4 ARP packets. */ vlib_packet_template_t ip4_arp_request_packet_template; diff --git a/src/vnet/ip/ip4_forward.c b/src/vnet/ip/ip4_forward.c index 2d48e8a9..ec4287bb 100755 --- a/src/vnet/ip/ip4_forward.c +++ b/src/vnet/ip/ip4_forward.c @@ -1198,8 +1198,10 @@ ip4_lookup_init (vlib_main_t * vm) ip_lookup_init (&im->lookup_main, /* is_ip6 */ 0); /* Create FIB with index 0 and table id of 0. */ - fib_table_find_or_create_and_lock (FIB_PROTOCOL_IP4, 0); - mfib_table_find_or_create_and_lock (FIB_PROTOCOL_IP4, 0); + fib_table_find_or_create_and_lock (FIB_PROTOCOL_IP4, 0, + FIB_SOURCE_DEFAULT_ROUTE); + mfib_table_find_or_create_and_lock (FIB_PROTOCOL_IP4, 0, + MFIB_SOURCE_DEFAULT_ROUTE); { pg_node_t *pn; @@ -2794,101 +2796,6 @@ VLIB_REGISTER_NODE (ip4_midchain_node) = { VLIB_NODE_FUNCTION_MULTIARCH (ip4_midchain_node, ip4_midchain); /* *INDENT-ON */ -static clib_error_t * -add_del_interface_table (vlib_main_t * vm, - unformat_input_t * input, vlib_cli_command_t * cmd) -{ - vnet_main_t *vnm = vnet_get_main (); - ip_interface_address_t *ia; - clib_error_t *error = 0; - u32 sw_if_index, table_id; - - sw_if_index = ~0; - - if (!unformat_user (input, unformat_vnet_sw_interface, vnm, &sw_if_index)) - { - error = clib_error_return (0, "unknown interface `%U'", - format_unformat_error, input); - goto done; - } - - if (unformat (input, "%d", &table_id)) - ; - else - { - error = clib_error_return (0, "expected table id `%U'", - format_unformat_error, input); - goto done; - } - - /* - * If the interface already has in IP address, then a change int - * VRF is not allowed. The IP address applied must first be removed. - * We do not do that automatically here, since VPP has no knowledge - * of whether thoses subnets are valid in the destination VRF. - */ - /* *INDENT-OFF* */ - foreach_ip_interface_address (&ip4_main.lookup_main, - ia, sw_if_index, - 1 /* honor unnumbered */, - ({ - ip4_address_t * a; - - a = ip_interface_address_get_address (&ip4_main.lookup_main, ia); - error = clib_error_return (0, "interface %U has address %U", - format_vnet_sw_if_index_name, vnm, - sw_if_index, - format_ip4_address, a); - goto done; - })); - /* *INDENT-ON* */ - -{ - ip4_main_t *im = &ip4_main; - u32 fib_index; - - fib_index = fib_table_find_or_create_and_lock (FIB_PROTOCOL_IP4, table_id); - - vec_validate (im->fib_index_by_sw_if_index, sw_if_index); - im->fib_index_by_sw_if_index[sw_if_index] = fib_index; - - fib_index = mfib_table_find_or_create_and_lock (FIB_PROTOCOL_IP4, table_id); - vec_validate (im->mfib_index_by_sw_if_index, sw_if_index); - im->mfib_index_by_sw_if_index[sw_if_index] = fib_index; -} - -done: -return error; -} - -/*? - * Place the indicated interface into the supplied IPv4 FIB table (also known - * as a VRF). If the FIB table does not exist, this command creates it. To - * display the current IPv4 FIB table, use the command 'show ip fib'. - * FIB table will only be displayed if a route has been added to the table, or - * an IP Address is assigned to an interface in the table (which adds a route - * automatically). - * - * @note IP addresses added after setting the interface IP table are added to - * the indicated FIB table. If an IP address is added prior to changing the - * table then this is an error. The control plane must remove these addresses - * first and then change the table. VPP will not automatically move the - * addresses from the old to the new table as it does not know the validity - * of such a change. - * - * @cliexpar - * Example of how to add an interface to an IPv4 FIB table (where 2 is the table-id): - * @cliexcmd{set interface ip table GigabitEthernet2/0/0 2} - ?*/ -/* *INDENT-OFF* */ -VLIB_CLI_COMMAND (set_interface_ip_table_command, static) = -{ - .path = "set interface ip table", - .function = add_del_interface_table, - .short_help = "set interface ip table ", -}; -/* *INDENT-ON* */ - int ip4_lookup_validate (ip4_address_t * a, u32 fib_index0) { diff --git a/src/vnet/ip/ip4_source_and_port_range_check.c b/src/vnet/ip/ip4_source_and_port_range_check.c index ae836a11..9aa880ae 100644 --- a/src/vnet/ip/ip4_source_and_port_range_check.c +++ b/src/vnet/ip/ip4_source_and_port_range_check.c @@ -1126,6 +1126,14 @@ ip6_source_and_port_range_check_add_del (ip6_address_t * address, u16 * low_ports, u16 * high_ports, int is_add) { + uint32_t fib_index; + + fib_index = fib_table_find (FIB_PROTOCOL_IP4, vrf_id); + + ASSERT (~0 != fib_index); + + fib_table_unlock (fib_index, FIB_PROTOCOL_IP4, FIB_SOURCE_CLASSIFY); + return 0; } @@ -1138,7 +1146,8 @@ ip4_source_and_port_range_check_add_del (ip4_address_t * address, { u32 fib_index; - fib_index = fib_table_find_or_create_and_lock (FIB_PROTOCOL_IP4, vrf_id); + fib_index = fib_table_find_or_create_and_lock (FIB_PROTOCOL_IP4, vrf_id, + FIB_SOURCE_CLASSIFY); if (is_add == 0) { diff --git a/src/vnet/ip/ip6.h b/src/vnet/ip/ip6.h index fa922725..8aef53a9 100644 --- a/src/vnet/ip/ip6.h +++ b/src/vnet/ip/ip6.h @@ -103,6 +103,16 @@ typedef struct uword function_opaque; } ip6_add_del_interface_address_callback_t; +typedef void (ip6_table_bind_function_t) + (struct ip6_main_t * im, + uword opaque, u32 sw_if_index, u32 new_fib_index, u32 old_fib_index); + +typedef struct +{ + ip6_table_bind_function_t *function; + uword function_opaque; +} ip6_table_bind_callback_t; + /** * Enumeration of the FIB table instance types */ @@ -183,6 +193,9 @@ typedef struct ip6_main_t ip6_add_del_interface_address_callback_t * add_del_interface_address_callbacks; + /** Functions to call when interface to table biding changes. */ + ip6_table_bind_callback_t *table_bind_callbacks; + /* Template used to generate IP6 neighbor solicitation packets. */ vlib_packet_template_t discover_neighbor_packet_template; diff --git a/src/vnet/ip/ip6_forward.c b/src/vnet/ip/ip6_forward.c index 5832bd0b..1002f6b6 100644 --- a/src/vnet/ip/ip6_forward.c +++ b/src/vnet/ip/ip6_forward.c @@ -2999,8 +2999,10 @@ ip6_lookup_init (vlib_main_t * vm) im->lookup_table_nbuckets, im->lookup_table_size); /* Create FIB with index 0 and table id of 0. */ - fib_table_find_or_create_and_lock (FIB_PROTOCOL_IP6, 0); - mfib_table_find_or_create_and_lock (FIB_PROTOCOL_IP6, 0); + fib_table_find_or_create_and_lock (FIB_PROTOCOL_IP6, 0, + FIB_SOURCE_DEFAULT_ROUTE); + mfib_table_find_or_create_and_lock (FIB_PROTOCOL_IP6, 0, + MFIB_SOURCE_DEFAULT_ROUTE); { pg_node_t *pn; @@ -3045,103 +3047,6 @@ ip6_lookup_init (vlib_main_t * vm) VLIB_INIT_FUNCTION (ip6_lookup_init); -static clib_error_t * -add_del_ip6_interface_table (vlib_main_t * vm, - unformat_input_t * input, - vlib_cli_command_t * cmd) -{ - vnet_main_t *vnm = vnet_get_main (); - ip_interface_address_t *ia; - clib_error_t *error = 0; - u32 sw_if_index, table_id; - - sw_if_index = ~0; - - if (!unformat_user (input, unformat_vnet_sw_interface, vnm, &sw_if_index)) - { - error = clib_error_return (0, "unknown interface `%U'", - format_unformat_error, input); - goto done; - } - - if (unformat (input, "%d", &table_id)) - ; - else - { - error = clib_error_return (0, "expected table id `%U'", - format_unformat_error, input); - goto done; - } - - /* - * If the interface already has in IP address, then a change int - * VRF is not allowed. The IP address applied must first be removed. - * We do not do that automatically here, since VPP has no knowledge - * of whether thoses subnets are valid in the destination VRF. - */ - /* *INDENT-OFF* */ - foreach_ip_interface_address (&ip6_main.lookup_main, - ia, sw_if_index, - 1 /* honor unnumbered */, - ({ - ip4_address_t * a; - - a = ip_interface_address_get_address (&ip6_main.lookup_main, ia); - error = clib_error_return (0, "interface %U has address %U", - format_vnet_sw_if_index_name, vnm, - sw_if_index, - format_ip6_address, a); - goto done; - })); - /* *INDENT-ON* */ - - { - u32 fib_index = fib_table_find_or_create_and_lock (FIB_PROTOCOL_IP6, - table_id); - - vec_validate (ip6_main.fib_index_by_sw_if_index, sw_if_index); - ip6_main.fib_index_by_sw_if_index[sw_if_index] = fib_index; - - fib_index = mfib_table_find_or_create_and_lock (FIB_PROTOCOL_IP6, - table_id); - - vec_validate (ip6_main.mfib_index_by_sw_if_index, sw_if_index); - ip6_main.mfib_index_by_sw_if_index[sw_if_index] = fib_index; - } - - -done: - return error; -} - -/*? - * Place the indicated interface into the supplied IPv6 FIB table (also known - * as a VRF). If the FIB table does not exist, this command creates it. To - * display the current IPv6 FIB table, use the command 'show ip6 fib'. - * FIB table will only be displayed if a route has been added to the table, or - * an IP Address is assigned to an interface in the table (which adds a route - * automatically). - * - * @note IP addresses added after setting the interface IP table are added to - * the indicated FIB table. If an IP address is added prior to changing the - * table then this is an error. The control plane must remove these addresses - * first and then change the table. VPP will not automatically move the - * addresses from the old to the new table as it does not know the validity - * of such a change. - * - * @cliexpar - * Example of how to add an interface to an IPv6 FIB table (where 2 is the table-id): - * @cliexcmd{set interface ip6 table GigabitEthernet2/0/0 2} - ?*/ -/* *INDENT-OFF* */ -VLIB_CLI_COMMAND (set_interface_ip6_table_command, static) = -{ - .path = "set interface ip6 table", - .function = add_del_ip6_interface_table, - .short_help = "set interface ip6 table " -}; -/* *INDENT-ON* */ - void ip6_link_local_address_from_ethernet_mac_address (ip6_address_t * ip, u8 * mac) diff --git a/src/vnet/ip/ip6_neighbor.c b/src/vnet/ip/ip6_neighbor.c index 62cf23ac..56f33ac8 100644 --- a/src/vnet/ip/ip6_neighbor.c +++ b/src/vnet/ip/ip6_neighbor.c @@ -250,6 +250,26 @@ format_ip6_neighbor_ip6_entry (u8 * s, va_list * va) return s; } +static void +ip6_neighbor_adj_fib_remove (ip6_neighbor_t * n, uint32_t fib_index) +{ + if (FIB_NODE_INDEX_INVALID != n->fib_entry_index) + { + fib_prefix_t pfx = { + .fp_len = 128, + .fp_proto = FIB_PROTOCOL_IP6, + .fp_addr.ip6 = n->key.ip6_address, + }; + fib_table_entry_path_remove (fib_index, + &pfx, + FIB_SOURCE_ADJ, + DPO_PROTO_IP6, + &pfx.fp_addr, + n->key.sw_if_index, ~0, + 1, FIB_ROUTE_PATH_FLAG_NONE); + } +} + static clib_error_t * ip6_neighbor_sw_interface_up_down (vnet_main_t * vnm, u32 sw_if_index, u32 flags) @@ -273,22 +293,10 @@ ip6_neighbor_sw_interface_up_down (vnet_main_t * vnm, { n = pool_elt_at_index (nm->neighbor_pool, to_delete[i]); mhash_unset (&nm->neighbor_index_by_key, &n->key, 0); - if (FIB_NODE_INDEX_INVALID != n->fib_entry_index) - { - fib_prefix_t pfx = { - .fp_len = 128, - .fp_proto = FIB_PROTOCOL_IP6, - .fp_addr.ip6 = n->key.ip6_address, - }; - fib_table_entry_path_remove - (ip6_fib_table_get_index_for_sw_if_index (n->key.sw_if_index), - &pfx, - FIB_SOURCE_ADJ, - DPO_PROTO_IP6, - &pfx.fp_addr, - n->key.sw_if_index, ~0, 1, FIB_ROUTE_PATH_FLAG_NONE); - pool_put (nm->neighbor_pool, n); - } + ip6_neighbor_adj_fib_remove (n, + ip6_fib_table_get_index_for_sw_if_index + (n->key.sw_if_index)); + pool_put (nm->neighbor_pool, n); } vec_free (to_delete); } @@ -579,6 +587,24 @@ ip6_ethernet_update_adjacency (vnet_main_t * vnm, u32 sw_if_index, u32 ai) } } + +static void +ip6_neighbor_adj_fib_add (ip6_neighbor_t * n, uint32_t fib_index) +{ + fib_prefix_t pfx = { + .fp_len = 128, + .fp_proto = FIB_PROTOCOL_IP6, + .fp_addr.ip6 = n->key.ip6_address, + }; + + n->fib_entry_index = + fib_table_entry_path_add (fib_index, &pfx, FIB_SOURCE_ADJ, + FIB_ENTRY_FLAG_ATTACHED, + DPO_PROTO_IP6, &pfx.fp_addr, + n->key.sw_if_index, ~0, 1, NULL, + FIB_ROUTE_PATH_FLAG_NONE); +} + int vnet_set_ip6_ethernet_neighbor (vlib_main_t * vm, u32 sw_if_index, @@ -633,21 +659,9 @@ vnet_set_ip6_ethernet_neighbor (vlib_main_t * vm, */ if (!is_no_fib_entry) { - fib_prefix_t pfx = { - .fp_len = 128, - .fp_proto = FIB_PROTOCOL_IP6, - .fp_addr.ip6 = k.ip6_address, - }; - u32 fib_index; - - fib_index = - ip6_fib_table_get_index_for_sw_if_index (n->key.sw_if_index); - n->fib_entry_index = - fib_table_entry_path_add (fib_index, &pfx, FIB_SOURCE_ADJ, - FIB_ENTRY_FLAG_ATTACHED, - DPO_PROTO_IP6, &pfx.fp_addr, - n->key.sw_if_index, ~0, 1, NULL, - FIB_ROUTE_PATH_FLAG_NONE); + ip6_neighbor_adj_fib_add (n, + ip6_fib_table_get_index_for_sw_if_index + (n->key.sw_if_index)); } else { @@ -3843,6 +3857,33 @@ ip6_set_neighbor_limit (u32 neighbor_limit) return 0; } +static void +ip6_neighbor_table_bind (ip6_main_t * im, + uword opaque, + u32 sw_if_index, + u32 new_fib_index, u32 old_fib_index) +{ + ip6_neighbor_main_t *nm = &ip6_neighbor_main; + ip6_neighbor_t *n = NULL; + u32 i, *to_re_add = 0; + + /* *INDENT-OFF* */ + pool_foreach (n, nm->neighbor_pool, + ({ + if (n->key.sw_if_index == sw_if_index) + vec_add1 (to_re_add, n - nm->neighbor_pool); + })); + /* *INDENT-ON* */ + + for (i = 0; i < vec_len (to_re_add); i++) + { + n = pool_elt_at_index (nm->neighbor_pool, to_re_add[i]); + ip6_neighbor_adj_fib_remove (n, old_fib_index); + ip6_neighbor_adj_fib_add (n, new_fib_index); + } + vec_free (to_re_add); +} + static clib_error_t * ip6_neighbor_init (vlib_main_t * vm) { @@ -3874,6 +3915,11 @@ ip6_neighbor_init (vlib_main_t * vm) cb.function_opaque = 0; vec_add1 (im->add_del_interface_address_callbacks, cb); + ip6_table_bind_callback_t cbt; + cbt.function = ip6_neighbor_table_bind; + cbt.function_opaque = 0; + vec_add1 (im->table_bind_callbacks, cbt); + mhash_init (&nm->pending_resolutions_by_address, /* value size */ sizeof (uword), /* key size */ sizeof (ip6_address_t)); diff --git a/src/vnet/ip/ip_api.c b/src/vnet/ip/ip_api.c index bba65ab4..384ec3e0 100644 --- a/src/vnet/ip/ip_api.c +++ b/src/vnet/ip/ip_api.c @@ -699,12 +699,58 @@ vl_api_ip_neighbor_add_del_t_handler (vl_api_ip_neighbor_add_del_t * mp, REPLY_MACRO (VL_API_IP_NEIGHBOR_ADD_DEL_REPLY); } +void +ip_table_delete (fib_protocol_t fproto, u32 table_id, u8 is_api) +{ + u32 fib_index, mfib_index; + + /* + * ignore action on the default table - this is always present + * and cannot be added nor deleted from the API + */ + if (0 != table_id) + { + /* + * The API holds only one lock on the table. + * i.e. it can be added many times via the API but needs to be + * deleted only once. + * The FIB index for unicast and multicast is not necessarily the + * same, since internal VPP systesm (like LISP and SR) create + * their own unicast tables. + */ + fib_index = fib_table_find (fproto, table_id); + mfib_index = mfib_table_find (fproto, table_id); + + if (~0 != fib_index) + { + fib_table_unlock (fib_index, fproto, + (is_api ? FIB_SOURCE_API : FIB_SOURCE_CLI)); + } + if (~0 != mfib_index) + { + mfib_table_unlock (mfib_index, fproto, + (is_api ? MFIB_SOURCE_API : MFIB_SOURCE_CLI)); + } + } +} + void vl_api_ip_table_add_del_t_handler (vl_api_ip_table_add_del_t * mp) { vl_api_ip_table_add_del_reply_t *rmp; + fib_protocol_t fproto = (mp->is_ipv6 ? FIB_PROTOCOL_IP6 : FIB_PROTOCOL_IP4); + u32 table_id = ntohl (mp->table_id); int rv = 0; + if (mp->is_add) + { + ip_table_create (fproto, table_id, 1); + } + else + { + ip_table_delete (fproto, table_id, 1); + } + REPLY_MACRO (VL_API_IP_TABLE_ADD_DEL_REPLY); } @@ -866,18 +912,21 @@ add_del_route_check (fib_protocol_t table_proto, u32 next_hop_sw_if_index, dpo_proto_t next_hop_table_proto, u32 next_hop_table_id, - u8 create_missing_tables, u8 is_rpf_id, u32 * fib_index, u32 * next_hop_fib_index) { vnet_main_t *vnm = vnet_get_main (); + /* Temporaray whilst I do the CSIT dance */ + u8 create_missing_tables = 1; + *fib_index = fib_table_find (table_proto, ntohl (table_id)); if (~0 == *fib_index) { if (create_missing_tables) { *fib_index = fib_table_find_or_create_and_lock (table_proto, - ntohl (table_id)); + ntohl (table_id), + FIB_SOURCE_API); } else { @@ -918,12 +967,14 @@ add_del_route_check (fib_protocol_t table_proto, *next_hop_fib_index = mfib_table_find_or_create_and_lock (fib_nh_proto, ntohl - (next_hop_table_id)); + (next_hop_table_id), + MFIB_SOURCE_API); else *next_hop_fib_index = fib_table_find_or_create_and_lock (fib_nh_proto, ntohl - (next_hop_table_id)); + (next_hop_table_id), + FIB_SOURCE_API); } else { @@ -948,8 +999,7 @@ ip4_add_del_route_t_handler (vl_api_ip_add_del_route_t * mp) mp->next_hop_sw_if_index, DPO_PROTO_IP4, mp->next_hop_table_id, - mp->create_vrf_if_needed, 0, - &fib_index, &next_hop_fib_index); + 0, &fib_index, &next_hop_fib_index); if (0 != rv) return (rv); @@ -1008,8 +1058,7 @@ ip6_add_del_route_t_handler (vl_api_ip_add_del_route_t * mp) mp->next_hop_sw_if_index, DPO_PROTO_IP6, mp->next_hop_table_id, - mp->create_vrf_if_needed, 0, - &fib_index, &next_hop_fib_index); + 0, &fib_index, &next_hop_fib_index); if (0 != rv) return (rv); @@ -1074,27 +1123,57 @@ vl_api_ip_add_del_route_t_handler (vl_api_ip_add_del_route_t * mp) REPLY_MACRO (VL_API_IP_ADD_DEL_ROUTE_REPLY); } +void +ip_table_create (fib_protocol_t fproto, u32 table_id, u8 is_api) +{ + u32 fib_index, mfib_index; + + /* + * ignore action on the default table - this is always present + * and cannot be added nor deleted from the API + */ + if (0 != table_id) + { + /* + * The API holds only one lock on the table. + * i.e. it can be added many times via the API but needs to be + * deleted only once. + * The FIB index for unicast and multicast is not necessarily the + * same, since internal VPP systesm (like LISP and SR) create + * their own unicast tables. + */ + fib_index = fib_table_find (fproto, table_id); + mfib_index = mfib_table_find (fproto, table_id); + + if (~0 == fib_index) + { + fib_table_find_or_create_and_lock (fproto, table_id, + (is_api ? + FIB_SOURCE_API : + FIB_SOURCE_CLI)); + } + if (~0 == mfib_index) + { + mfib_table_find_or_create_and_lock (fproto, table_id, + (is_api ? + MFIB_SOURCE_API : + MFIB_SOURCE_CLI)); + } + } +} + static int add_del_mroute_check (fib_protocol_t table_proto, u32 table_id, - u32 next_hop_sw_if_index, - u8 is_local, u8 create_missing_tables, u32 * fib_index) + u32 next_hop_sw_if_index, u8 is_local, u32 * fib_index) { vnet_main_t *vnm = vnet_get_main (); *fib_index = mfib_table_find (table_proto, ntohl (table_id)); if (~0 == *fib_index) { - if (create_missing_tables) - { - *fib_index = mfib_table_find_or_create_and_lock (table_proto, - ntohl (table_id)); - } - else - { - /* No such VRF, and we weren't asked to create one */ - return VNET_API_ERROR_NO_SUCH_FIB; - } + /* No such table */ + return VNET_API_ERROR_NO_SUCH_FIB; } if (~0 != ntohl (next_hop_sw_if_index)) @@ -1163,8 +1242,7 @@ api_mroute_add_del_t_handler (vl_api_ip_mroute_add_del_t * mp) rv = add_del_mroute_check (fproto, mp->table_id, mp->next_hop_sw_if_index, - mp->is_local, - mp->create_vrf_if_needed, &fib_index); + mp->is_local, &fib_index); if (0 != rv) return (rv); diff --git a/src/vnet/ip/lookup.c b/src/vnet/ip/lookup.c index 5537bb04..667c6791 100755 --- a/src/vnet/ip/lookup.c +++ b/src/vnet/ip/lookup.c @@ -687,6 +687,78 @@ done: return error; } +clib_error_t * +vnet_ip_table_cmd (vlib_main_t * vm, + unformat_input_t * main_input, + vlib_cli_command_t * cmd, fib_protocol_t fproto) +{ + unformat_input_t _line_input, *line_input = &_line_input; + clib_error_t *error = NULL; + u32 table_id, is_add; + + is_add = 1; + table_id = ~0; + + /* Get a line of input. */ + if (!unformat_user (main_input, unformat_line_input, line_input)) + return 0; + + while (unformat_check_input (line_input) != UNFORMAT_END_OF_INPUT) + { + if (unformat (line_input, "%d", &table_id)) + ; + else if (unformat (line_input, "del")) + is_add = 0; + else if (unformat (line_input, "add")) + is_add = 1; + else + { + error = unformat_parse_error (line_input); + goto done; + } + } + + if (~0 == table_id) + { + error = clib_error_return (0, "No table id"); + goto done; + } + else if (0 == table_id) + { + error = clib_error_return (0, "Can't change the default table"); + goto done; + } + else + { + if (is_add) + { + ip_table_create (fproto, table_id, 0); + } + else + { + ip_table_delete (fproto, table_id, 0); + } + } + +done: + unformat_free (line_input); + return error; +} + +clib_error_t * +vnet_ip4_table_cmd (vlib_main_t * vm, + unformat_input_t * main_input, vlib_cli_command_t * cmd) +{ + return (vnet_ip_table_cmd (vm, main_input, cmd, FIB_PROTOCOL_IP4)); +} + +clib_error_t * +vnet_ip6_table_cmd (vlib_main_t * vm, + unformat_input_t * main_input, vlib_cli_command_t * cmd) +{ + return (vnet_ip_table_cmd (vm, main_input, cmd, FIB_PROTOCOL_IP6)); +} + /* *INDENT-OFF* */ VLIB_CLI_COMMAND (vlib_cli_ip_command, static) = { .path = "ip", @@ -749,6 +821,159 @@ VLIB_CLI_COMMAND (ip_route_command, static) = { .function = vnet_ip_route_cmd, .is_mp_safe = 1, }; + +/* *INDENT-ON* */ +/*? + * This command is used to add or delete IPv4 Tables. All + * Tables must be explicitly added before that can be used. Creating a + * table will add both unicast and multicast FIBs + * + ?*/ +/* *INDENT-OFF* */ +VLIB_CLI_COMMAND (ip4_table_command, static) = { + .path = "ip table", + .short_help = "ip table [add|del] ", + .function = vnet_ip4_table_cmd, + .is_mp_safe = 1, +}; +/* *INDENT-ON* */ + +/* *INDENT-ON* */ +/*? + * This command is used to add or delete IPv4 Tables. All + * Tables must be explicitly added before that can be used. Creating a + * table will add both unicast and multicast FIBs + * + ?*/ +/* *INDENT-OFF* */ +VLIB_CLI_COMMAND (ip6_table_command, static) = { + .path = "ip6 table", + .short_help = "ip6 table [add|del] ", + .function = vnet_ip6_table_cmd, + .is_mp_safe = 1, +}; + +static clib_error_t * +ip_table_bind_cmd (vlib_main_t * vm, + unformat_input_t * input, + vlib_cli_command_t * cmd, + fib_protocol_t fproto) +{ + vnet_main_t *vnm = vnet_get_main (); + clib_error_t *error = 0; + u32 sw_if_index, table_id; + int rv; + + sw_if_index = ~0; + + if (!unformat_user (input, unformat_vnet_sw_interface, vnm, &sw_if_index)) + { + error = clib_error_return (0, "unknown interface `%U'", + format_unformat_error, input); + goto done; + } + + if (unformat (input, "%d", &table_id)) + ; + else + { + error = clib_error_return (0, "expected table id `%U'", + format_unformat_error, input); + goto done; + } + + rv = ip_table_bind (fproto, sw_if_index, table_id, 0); + + if (VNET_API_ERROR_ADDRESS_FOUND_FOR_INTERFACE == rv) + { + error = clib_error_return (0, "IP addresses are still present on %U", + format_vnet_sw_if_index_name, + vnet_get_main(), + sw_if_index); + } + else if (VNET_API_ERROR_NO_SUCH_FIB == rv) + { + error = clib_error_return (0, "no such table %d", table_id); + } + else if (0 != rv) + { + error = clib_error_return (0, "unknown error"); + } + + done: + return error; +} + +static clib_error_t * +ip4_table_bind_cmd (vlib_main_t * vm, + unformat_input_t * input, + vlib_cli_command_t * cmd) +{ + return (ip_table_bind_cmd (vm , input, cmd, FIB_PROTOCOL_IP4)); +} + +static clib_error_t * +ip6_table_bind_cmd (vlib_main_t * vm, + unformat_input_t * input, + vlib_cli_command_t * cmd) +{ + return (ip_table_bind_cmd (vm , input, cmd, FIB_PROTOCOL_IP6)); +} + +/*? + * Place the indicated interface into the supplied IPv4 FIB table (also known + * as a VRF). If the FIB table does not exist, this command creates it. To + * display the current IPv4 FIB table, use the command 'show ip fib'. + * FIB table will only be displayed if a route has been added to the table, or + * an IP Address is assigned to an interface in the table (which adds a route + * automatically). + * + * @note IP addresses added after setting the interface IP table are added to + * the indicated FIB table. If an IP address is added prior to changing the + * table then this is an error. The control plane must remove these addresses + * first and then change the table. VPP will not automatically move the + * addresses from the old to the new table as it does not know the validity + * of such a change. + * + * @cliexpar + * Example of how to add an interface to an IPv4 FIB table (where 2 is the table-id): + * @cliexcmd{set interface ip table GigabitEthernet2/0/0 2} + ?*/ +/* *INDENT-OFF* */ +VLIB_CLI_COMMAND (set_interface_ip_table_command, static) = +{ + .path = "set interface ip table", + .function = ip4_table_bind_cmd, + .short_help = "set interface ip table ", +}; +/* *INDENT-ON* */ + +/*? + * Place the indicated interface into the supplied IPv6 FIB table (also known + * as a VRF). If the FIB table does not exist, this command creates it. To + * display the current IPv6 FIB table, use the command 'show ip6 fib'. + * FIB table will only be displayed if a route has been added to the table, or + * an IP Address is assigned to an interface in the table (which adds a route + * automatically). + * + * @note IP addresses added after setting the interface IP table are added to + * the indicated FIB table. If an IP address is added prior to changing the + * table then this is an error. The control plane must remove these addresses + * first and then change the table. VPP will not automatically move the + * addresses from the old to the new table as it does not know the validity + * of such a change. + * + * @cliexpar + * Example of how to add an interface to an IPv6 FIB table (where 2 is the table-id): + * @cliexcmd{set interface ip6 table GigabitEthernet2/0/0 2} + ?*/ +/* *INDENT-OFF* */ +VLIB_CLI_COMMAND (set_interface_ip6_table_command, static) = +{ + .path = "set interface ip6 table", + .function = ip6_table_bind_cmd, + .short_help = "set interface ip6 table " +}; /* *INDENT-ON* */ clib_error_t * diff --git a/src/vnet/lisp-gpe/interface.c b/src/vnet/lisp-gpe/interface.c index e832c23f..a0c05e85 100644 --- a/src/vnet/lisp-gpe/interface.c +++ b/src/vnet/lisp-gpe/interface.c @@ -505,12 +505,14 @@ lisp_gpe_iface_set_table (u32 sw_if_index, u32 table_id) { fib_node_index_t fib_index; - fib_index = fib_table_find_or_create_and_lock (FIB_PROTOCOL_IP4, table_id); + fib_index = fib_table_find_or_create_and_lock (FIB_PROTOCOL_IP4, table_id, + FIB_SOURCE_LISP); vec_validate (ip4_main.fib_index_by_sw_if_index, sw_if_index); ip4_main.fib_index_by_sw_if_index[sw_if_index] = fib_index; ip4_sw_interface_enable_disable (sw_if_index, 1); - fib_index = fib_table_find_or_create_and_lock (FIB_PROTOCOL_IP6, table_id); + fib_index = fib_table_find_or_create_and_lock (FIB_PROTOCOL_IP6, table_id, + FIB_SOURCE_LISP); vec_validate (ip6_main.fib_index_by_sw_if_index, sw_if_index); ip6_main.fib_index_by_sw_if_index[sw_if_index] = fib_index; ip6_sw_interface_enable_disable (sw_if_index, 1); @@ -530,7 +532,7 @@ lisp_gpe_tenant_del_default_routes (u32 table_id) fib_index = fib_table_find (prefix.fp_proto, table_id); fib_table_entry_special_remove (fib_index, &prefix, FIB_SOURCE_LISP); - fib_table_unlock (fib_index, prefix.fp_proto); + fib_table_unlock (fib_index, prefix.fp_proto, FIB_SOURCE_LISP); } } @@ -549,7 +551,8 @@ lisp_gpe_tenant_add_default_routes (u32 table_id) /* * Add a deafult route that results in a control plane punt DPO */ - fib_index = fib_table_find_or_create_and_lock (prefix.fp_proto, table_id); + fib_index = fib_table_find_or_create_and_lock (prefix.fp_proto, table_id, + FIB_SOURCE_LISP); fib_table_entry_special_dpo_add (fib_index, &prefix, FIB_SOURCE_LISP, FIB_ENTRY_FLAG_EXCLUSIVE, lisp_cp_dpo_get (fib_proto_to_dpo diff --git a/src/vnet/lisp-gpe/lisp_gpe_fwd_entry.c b/src/vnet/lisp-gpe/lisp_gpe_fwd_entry.c index d7d3cb86..0a8dc039 100644 --- a/src/vnet/lisp-gpe/lisp_gpe_fwd_entry.c +++ b/src/vnet/lisp-gpe/lisp_gpe_fwd_entry.c @@ -66,6 +66,7 @@ ip_dst_fib_add_route (u32 dst_fib_index, const ip_prefix_t * dst_prefix) /* create a new src FIB. */ src_fib_index = fib_table_create_and_lock (dst_fib_prefix.fp_proto, + FIB_SOURCE_LISP, "LISP-src for [%d,%U]", dst_fib_index, format_fib_prefix, &dst_fib_prefix); @@ -180,7 +181,8 @@ ip_src_dst_fib_del_route (u32 src_fib_index, */ fib_table_entry_special_remove (dst_fib_index, &dst_fib_prefix, FIB_SOURCE_LISP); - fib_table_unlock (src_fib_index, src_fib_prefix.fp_proto); + fib_table_unlock (src_fib_index, src_fib_prefix.fp_proto, + FIB_SOURCE_LISP); } } @@ -544,7 +546,8 @@ add_ip_fwd_entry (lisp_gpe_main_t * lgm, lfe->tenant = lisp_gpe_tenant_find_or_create (lfe->key->vni); lfe->eid_table_id = a->table_id; lfe->eid_fib_index = fib_table_find_or_create_and_lock (fproto, - lfe->eid_table_id); + lfe->eid_table_id, + FIB_SOURCE_LISP); lfe->is_src_dst = a->is_src_dst; if (LISP_GPE_FWD_ENTRY_TYPE_NEGATIVE != lfe->type) @@ -578,7 +581,7 @@ del_ip_fwd_entry_i (lisp_gpe_main_t * lgm, lisp_gpe_fwd_entry_t * lfe) fproto = (IP4 == ip_prefix_version (&fid_addr_ippref (&lfe->key->rmt)) ? FIB_PROTOCOL_IP4 : FIB_PROTOCOL_IP6); - fib_table_unlock (lfe->eid_fib_index, fproto); + fib_table_unlock (lfe->eid_fib_index, fproto, FIB_SOURCE_LISP); hash_unset_mem (lgm->lisp_gpe_fwd_entries, lfe->key); clib_mem_free (lfe->key); diff --git a/src/vnet/lisp-gpe/lisp_gpe_sub_interface.c b/src/vnet/lisp-gpe/lisp_gpe_sub_interface.c index b234d9dc..26664f53 100644 --- a/src/vnet/lisp-gpe/lisp_gpe_sub_interface.c +++ b/src/vnet/lisp-gpe/lisp_gpe_sub_interface.c @@ -89,13 +89,15 @@ lisp_gpe_sub_interface_set_table (u32 sw_if_index, u32 table_id) { fib_node_index_t fib_index; - fib_index = fib_table_find_or_create_and_lock (FIB_PROTOCOL_IP4, table_id); + fib_index = fib_table_find_or_create_and_lock (FIB_PROTOCOL_IP4, table_id, + FIB_SOURCE_LISP); ASSERT (FIB_NODE_INDEX_INVALID != fib_index); vec_validate (ip4_main.fib_index_by_sw_if_index, sw_if_index); ip4_main.fib_index_by_sw_if_index[sw_if_index] = fib_index; - fib_index = fib_table_find_or_create_and_lock (FIB_PROTOCOL_IP6, table_id); + fib_index = fib_table_find_or_create_and_lock (FIB_PROTOCOL_IP6, table_id, + FIB_SOURCE_LISP); ASSERT (FIB_NODE_INDEX_INVALID != fib_index); vec_validate (ip6_main.fib_index_by_sw_if_index, sw_if_index); @@ -105,9 +107,13 @@ lisp_gpe_sub_interface_set_table (u32 sw_if_index, u32 table_id) static void lisp_gpe_sub_interface_unset_table (u32 sw_if_index, u32 table_id) { + fib_table_unlock (ip4_main.fib_index_by_sw_if_index[sw_if_index], + FIB_PROTOCOL_IP4, FIB_SOURCE_LISP); ip4_main.fib_index_by_sw_if_index[sw_if_index] = 0; ip4_sw_interface_enable_disable (sw_if_index, 0); + fib_table_unlock (ip6_main.fib_index_by_sw_if_index[sw_if_index], + FIB_PROTOCOL_IP6, FIB_SOURCE_LISP); ip6_main.fib_index_by_sw_if_index[sw_if_index] = 0; ip6_sw_interface_enable_disable (sw_if_index, 0); } @@ -185,6 +191,7 @@ lisp_gpe_sub_interface_unlock (index_t l3si) l3s = lisp_gpe_sub_interface_get_i (l3si); + ASSERT (0 != l3s->locks); l3s->locks--; if (0 == l3s->locks) diff --git a/src/vnet/mfib/ip4_mfib.c b/src/vnet/mfib/ip4_mfib.c index 1849a3a4..b2482580 100644 --- a/src/vnet/mfib/ip4_mfib.c +++ b/src/vnet/mfib/ip4_mfib.c @@ -33,7 +33,8 @@ static const mfib_prefix_t ip4_specials[] = { }; static u32 -ip4_create_mfib_with_table_id (u32 table_id) +ip4_create_mfib_with_table_id (u32 table_id, + mfib_source_t src) { mfib_table_t *mfib_table; @@ -53,7 +54,7 @@ ip4_create_mfib_with_table_id (u32 table_id) mfib_table->v4.table_id = table_id; - mfib_table_lock(mfib_table->mft_index, FIB_PROTOCOL_IP4); + mfib_table_lock(mfib_table->mft_index, FIB_PROTOCOL_IP4, src); /* * add the special entries into the new FIB @@ -113,14 +114,15 @@ ip4_mfib_table_destroy (ip4_mfib_t *mfib) } u32 -ip4_mfib_table_find_or_create_and_lock (u32 table_id) +ip4_mfib_table_find_or_create_and_lock (u32 table_id, + mfib_source_t src) { u32 index; index = ip4_mfib_index_from_table_id(table_id); if (~0 == index) - return ip4_create_mfib_with_table_id(table_id); - mfib_table_lock(index, FIB_PROTOCOL_IP4); + return ip4_create_mfib_with_table_id(table_id, src); + mfib_table_lock(index, FIB_PROTOCOL_IP4, src); return (index); } diff --git a/src/vnet/mfib/ip4_mfib.h b/src/vnet/mfib/ip4_mfib.h index ea682651..e31fb744 100644 --- a/src/vnet/mfib/ip4_mfib.h +++ b/src/vnet/mfib/ip4_mfib.h @@ -72,8 +72,9 @@ ip4_mfib_get (u32 index) * @returns A pointer to the retrieved or created fib. * */ -extern u32 ip4_mfib_table_find_or_create_and_lock(u32 table_id); -extern u32 ip4_mfib_table_create_and_lock(void); +extern u32 ip4_mfib_table_find_or_create_and_lock(u32 table_id, + mfib_source_t src); +extern u32 ip4_mfib_table_create_and_lock(mfib_source_t src); static inline u32 ip4_mfib_index_from_table_id (u32 table_id) diff --git a/src/vnet/mfib/ip6_mfib.c b/src/vnet/mfib/ip6_mfib.c index 5e48e919..e4861330 100644 --- a/src/vnet/mfib/ip6_mfib.c +++ b/src/vnet/mfib/ip6_mfib.c @@ -151,7 +151,8 @@ static const ip6_mfib_special_t ip6_mfib_specials[] = static u32 -ip6_create_mfib_with_table_id (u32 table_id) +ip6_create_mfib_with_table_id (u32 table_id, + mfib_source_t src) { mfib_table_t *mfib_table; mfib_prefix_t pfx = { @@ -182,7 +183,7 @@ ip6_create_mfib_with_table_id (u32 table_id) mfib_table->v6.table_id = table_id; - mfib_table_lock(mfib_table->mft_index, FIB_PROTOCOL_IP6); + mfib_table_lock(mfib_table->mft_index, FIB_PROTOCOL_IP6, src); mfib_table->v6.rhead = clib_mem_alloc_aligned (sizeof(*mfib_table->v6.rhead), @@ -297,14 +298,15 @@ ip6_mfib_interface_enable_disable (u32 sw_if_index, int is_enable) } u32 -ip6_mfib_table_find_or_create_and_lock (u32 table_id) +ip6_mfib_table_find_or_create_and_lock (u32 table_id, + mfib_source_t src) { u32 index; index = ip6_mfib_index_from_table_id(table_id); if (~0 == index) - return ip6_create_mfib_with_table_id(table_id); - mfib_table_lock(index, FIB_PROTOCOL_IP6); + return ip6_create_mfib_with_table_id(table_id, src); + mfib_table_lock(index, FIB_PROTOCOL_IP6, src); return (index); } diff --git a/src/vnet/mfib/ip6_mfib.h b/src/vnet/mfib/ip6_mfib.h index adaa7ec2..ea81b553 100644 --- a/src/vnet/mfib/ip6_mfib.h +++ b/src/vnet/mfib/ip6_mfib.h @@ -79,8 +79,9 @@ ip6_mfib_get (u32 index) * @returns A pointer to the retrieved or created fib. * */ -extern u32 ip6_mfib_table_find_or_create_and_lock(u32 table_id); -extern u32 ip6_mfib_table_create_and_lock(void); +extern u32 ip6_mfib_table_find_or_create_and_lock(u32 table_id, + mfib_source_t src); +extern u32 ip6_mfib_table_create_and_lock(mfib_source_t src); static inline diff --git a/src/vnet/mfib/mfib_entry.c b/src/vnet/mfib/mfib_entry.c index 804e10ab..2302b9a1 100644 --- a/src/vnet/mfib/mfib_entry.c +++ b/src/vnet/mfib/mfib_entry.c @@ -334,6 +334,17 @@ mfib_entry_get_best_src (const mfib_entry_t *mfib_entry) return (bsrc); } +int +mfib_entry_is_sourced (fib_node_index_t mfib_entry_index, + mfib_source_t source) +{ + mfib_entry_t *mfib_entry; + + mfib_entry = mfib_entry_get(mfib_entry_index); + + return (NULL != mfib_entry_src_find(mfib_entry, source, NULL)); +} + static void mfib_entry_src_flush (mfib_entry_src_t *msrc) { diff --git a/src/vnet/mfib/mfib_entry.h b/src/vnet/mfib/mfib_entry.h index d4377878..96ee49f7 100644 --- a/src/vnet/mfib/mfib_entry.h +++ b/src/vnet/mfib/mfib_entry.h @@ -130,6 +130,8 @@ extern void mfib_entry_unlock(fib_node_index_t fib_entry_index); extern void mfib_entry_get_prefix(fib_node_index_t fib_entry_index, mfib_prefix_t *pfx); extern u32 mfib_entry_get_fib_index(fib_node_index_t fib_entry_index); +extern int mfib_entry_is_sourced(fib_node_index_t fib_entry_index, + mfib_source_t source); extern void mfib_entry_contribute_forwarding( fib_node_index_t mfib_entry_index, diff --git a/src/vnet/mfib/mfib_table.c b/src/vnet/mfib/mfib_table.c index 7ffe8941..e5550adc 100644 --- a/src/vnet/mfib/mfib_table.c +++ b/src/vnet/mfib/mfib_table.c @@ -424,7 +424,8 @@ mfib_table_find (fib_protocol_t proto, u32 mfib_table_find_or_create_and_lock (fib_protocol_t proto, - u32 table_id) + u32 table_id, + mfib_source_t src) { mfib_table_t *mfib_table; fib_node_index_t fi; @@ -432,10 +433,10 @@ mfib_table_find_or_create_and_lock (fib_protocol_t proto, switch (proto) { case FIB_PROTOCOL_IP4: - fi = ip4_mfib_table_find_or_create_and_lock(table_id); + fi = ip4_mfib_table_find_or_create_and_lock(table_id, src); break; case FIB_PROTOCOL_IP6: - fi = ip6_mfib_table_find_or_create_and_lock(table_id); + fi = ip6_mfib_table_find_or_create_and_lock(table_id, src); break; case FIB_PROTOCOL_MPLS: default: @@ -451,6 +452,59 @@ mfib_table_find_or_create_and_lock (fib_protocol_t proto, return (fi); } +/** + * @brief Table flush context. Store the indicies of matching FIB entries + * that need to be removed. + */ +typedef struct mfib_table_flush_ctx_t_ +{ + /** + * The list of entries to flush + */ + fib_node_index_t *mftf_entries; + + /** + * The source we are flushing + */ + mfib_source_t mftf_source; +} mfib_table_flush_ctx_t; + +static int +mfib_table_flush_cb (fib_node_index_t mfib_entry_index, + void *arg) +{ + mfib_table_flush_ctx_t *ctx = arg; + + if (mfib_entry_is_sourced(mfib_entry_index, ctx->mftf_source)) + { + vec_add1(ctx->mftf_entries, mfib_entry_index); + } + return (1); +} + +void +mfib_table_flush (u32 mfib_index, + fib_protocol_t proto, + mfib_source_t source) +{ + fib_node_index_t *mfib_entry_index; + mfib_table_flush_ctx_t ctx = { + .mftf_entries = NULL, + .mftf_source = source, + }; + + mfib_table_walk(mfib_index, proto, + mfib_table_flush_cb, + &ctx); + + vec_foreach(mfib_entry_index, ctx.mftf_entries) + { + mfib_table_entry_delete_index(*mfib_entry_index, source); + } + + vec_free(ctx.mftf_entries); +} + static void mfib_table_destroy (mfib_table_t *mfib_table) { @@ -472,27 +526,43 @@ mfib_table_destroy (mfib_table_t *mfib_table) void mfib_table_unlock (u32 fib_index, - fib_protocol_t proto) + fib_protocol_t proto, + mfib_source_t source) { mfib_table_t *mfib_table; mfib_table = mfib_table_get(fib_index, proto); - mfib_table->mft_locks--; + mfib_table->mft_locks[source]--; + mfib_table->mft_locks[MFIB_TABLE_TOTAL_LOCKS]--; + + if (0 == mfib_table->mft_locks[source]) + { + /* + * The source no longer needs the table. flush any routes + * from it just in case + */ + mfib_table_flush(fib_index, proto, source); + } - if (0 == mfib_table->mft_locks) + if (0 == mfib_table->mft_locks[MFIB_TABLE_TOTAL_LOCKS]) { - mfib_table_destroy(mfib_table); + /* + * no more locak from any source - kill it + */ + mfib_table_destroy(mfib_table); } } void mfib_table_lock (u32 fib_index, - fib_protocol_t proto) + fib_protocol_t proto, + mfib_source_t source) { mfib_table_t *mfib_table; mfib_table = mfib_table_get(fib_index, proto); - mfib_table->mft_locks++; + mfib_table->mft_locks[source]++; + mfib_table->mft_locks[MFIB_TABLE_TOTAL_LOCKS]++; } void diff --git a/src/vnet/mfib/mfib_table.h b/src/vnet/mfib/mfib_table.h index 83aa04ef..c6b0b097 100644 --- a/src/vnet/mfib/mfib_table.h +++ b/src/vnet/mfib/mfib_table.h @@ -22,6 +22,12 @@ #include +/** + * Keep a lock per-source and a total + */ +#define MFIB_TABLE_N_LOCKS (MFIB_N_SOURCES+1) +#define MFIB_TABLE_TOTAL_LOCKS MFIB_N_SOURCES + /** * @brief * A protocol Independent IP multicast FIB table @@ -47,7 +53,7 @@ typedef struct mfib_table_t_ /** * number of locks on the table */ - u16 mft_locks; + u16 mft_locks[MFIB_TABLE_N_LOCKS]; /** * Table ID (hash key) for this FIB. @@ -259,7 +265,8 @@ extern fib_node_index_t mfib_table_entry_special_add(u32 fib_index, * the source to flush */ extern void mfib_table_flush(u32 fib_index, - fib_protocol_t proto); + fib_protocol_t proto, + mfib_source_t source); /** * @brief @@ -307,9 +314,13 @@ extern u32 mfib_table_find(fib_protocol_t proto, u32 table_id); * * @return fib_index * The index of the FIB + * + * @param source + * The ID of the client/source. */ extern u32 mfib_table_find_or_create_and_lock(fib_protocol_t proto, - u32 table_id); + u32 table_id, + mfib_source_t source); /** @@ -321,9 +332,13 @@ extern u32 mfib_table_find_or_create_and_lock(fib_protocol_t proto, * * @paran proto * The protocol of the FIB (and thus the entries therein) + * + * @param source + * The ID of the client/source. */ extern void mfib_table_unlock(u32 fib_index, - fib_protocol_t proto); + fib_protocol_t proto, + mfib_source_t source); /** * @brief @@ -335,9 +350,13 @@ extern void mfib_table_unlock(u32 fib_index, * * @paran proto * The protocol of the FIB (and thus the entries therein) + * + * @param source + * The ID of the client/source. */ extern void mfib_table_lock(u32 fib_index, - fib_protocol_t proto); + fib_protocol_t proto, + mfib_source_t source); /** * @brief diff --git a/src/vnet/mfib/mfib_test.c b/src/vnet/mfib/mfib_test.c index 57787eca..3055844d 100644 --- a/src/vnet/mfib/mfib_test.c +++ b/src/vnet/mfib/mfib_test.c @@ -22,6 +22,7 @@ #include #include #include +#include #include #include @@ -366,7 +367,7 @@ mfib_test_i (fib_protocol_t PROTO, MFIB_TEST(3 == adj_mcast_db_size(), "3 MCAST adjs"); /* Find or create FIB table 11 */ - fib_index = mfib_table_find_or_create_and_lock(PROTO, 11); + fib_index = mfib_table_find_or_create_and_lock(PROTO, 11, MFIB_SOURCE_API); mfib_prefix_t pfx_dft = { .fp_len = 0, @@ -1113,9 +1114,10 @@ mfib_test_i (fib_protocol_t PROTO, /* * MPLS enable an interface so we get the MPLS table created */ + mpls_table_create(MPLS_FIB_DEFAULT_TABLE_ID, FIB_SOURCE_API); mpls_sw_interface_enable_disable(&mpls_main, tm->hw[0]->sw_if_index, - 1); + 1, 0); lfei = fib_table_entry_update_one_path(0, // default MPLS Table &pfx_3500, @@ -1192,7 +1194,7 @@ mfib_test_i (fib_protocol_t PROTO, /* * Unlock the table - it's the last lock so should be gone thereafter */ - mfib_table_unlock(fib_index, PROTO); + mfib_table_unlock(fib_index, PROTO, MFIB_SOURCE_API); MFIB_TEST((FIB_NODE_INDEX_INVALID == mfib_table_find(PROTO, fib_index)), @@ -1207,7 +1209,8 @@ mfib_test_i (fib_protocol_t PROTO, */ mpls_sw_interface_enable_disable(&mpls_main, tm->hw[0]->sw_if_index, - 0); + 0, 0); + mpls_table_delete(MPLS_FIB_DEFAULT_TABLE_ID, FIB_SOURCE_API); /* * test we've leaked no resources diff --git a/src/vnet/mfib/mfib_types.h b/src/vnet/mfib/mfib_types.h index 863fad16..50aede04 100644 --- a/src/vnet/mfib/mfib_types.h +++ b/src/vnet/mfib/mfib_types.h @@ -166,9 +166,10 @@ typedef enum mfib_source_t_ MFIB_SOURCE_VXLAN, MFIB_SOURCE_DHCP, MFIB_SOURCE_SRv6, - MFIB_SOURCE_DEFAULT_ROUTE, MFIB_SOURCE_GTPU, MFIB_SOURCE_VXLAN_GPE, + MFIB_SOURCE_RR, + MFIB_SOURCE_DEFAULT_ROUTE, } mfib_source_t; #define MFIB_SOURCE_NAMES { \ @@ -178,11 +179,14 @@ typedef enum mfib_source_t_ [MFIB_SOURCE_DHCP] = "DHCP", \ [MFIB_SOURCE_VXLAN] = "VXLAN", \ [MFIB_SOURCE_SRv6] = "SRv6", \ - [MFIB_SOURCE_DEFAULT_ROUTE] = "Default Route", \ [MFIB_SOURCE_GTPU] = "GTPU", \ [MFIB_SOURCE_VXLAN_GPE] = "VXLAN-GPE", \ + [MFIB_SOURCE_RR] = "Recursive-resolution", \ + [MFIB_SOURCE_DEFAULT_ROUTE] = "Default Route", \ } +#define MFIB_N_SOURCES (MFIB_SOURCE_DEFAULT_ROUTE) + /** * \brief Compare two prefixes for equality */ diff --git a/src/vnet/mpls/interface.c b/src/vnet/mpls/interface.c index a085aaa2..d7c8e7d3 100644 --- a/src/vnet/mpls/interface.c +++ b/src/vnet/mpls/interface.c @@ -35,25 +35,33 @@ mpls_sw_interface_is_enabled (u32 sw_if_index) return (mm->mpls_enabled_by_sw_if_index[sw_if_index]); } -void +int mpls_sw_interface_enable_disable (mpls_main_t * mm, u32 sw_if_index, - u8 is_enable) + u8 is_enable, + u8 is_api) { fib_node_index_t lfib_index; vec_validate_init_empty (mm->mpls_enabled_by_sw_if_index, sw_if_index, 0); + lfib_index = fib_table_find(FIB_PROTOCOL_MPLS, + MPLS_FIB_DEFAULT_TABLE_ID); + + if (~0 == lfib_index) + return VNET_API_ERROR_NO_SUCH_FIB; + /* * enable/disable only on the 1<->0 transition */ if (is_enable) { if (1 != ++mm->mpls_enabled_by_sw_if_index[sw_if_index]) - return; + return (0); + + fib_table_lock(lfib_index, FIB_PROTOCOL_MPLS, + (is_api? FIB_SOURCE_API: FIB_SOURCE_CLI)); - lfib_index = fib_table_find_or_create_and_lock(FIB_PROTOCOL_MPLS, - MPLS_FIB_DEFAULT_TABLE_ID); vec_validate(mm->fib_index_by_sw_if_index, 0); mm->fib_index_by_sw_if_index[sw_if_index] = lfib_index; } @@ -61,15 +69,17 @@ mpls_sw_interface_enable_disable (mpls_main_t * mm, { ASSERT(mm->mpls_enabled_by_sw_if_index[sw_if_index] > 0); if (0 != --mm->mpls_enabled_by_sw_if_index[sw_if_index]) - return; + return (0); fib_table_unlock(mm->fib_index_by_sw_if_index[sw_if_index], - FIB_PROTOCOL_MPLS); + FIB_PROTOCOL_MPLS, + (is_api? FIB_SOURCE_API: FIB_SOURCE_CLI)); } vnet_feature_enable_disable ("mpls-input", "mpls-not-enabled", sw_if_index, !is_enable, 0, 0); + return (0); } static clib_error_t * @@ -101,7 +111,7 @@ mpls_interface_enable_disable (vlib_main_t * vm, goto done; } - mpls_sw_interface_enable_disable(&mpls_main, sw_if_index, enable); + mpls_sw_interface_enable_disable(&mpls_main, sw_if_index, enable, 0); done: return error; diff --git a/src/vnet/mpls/mpls.c b/src/vnet/mpls/mpls.c index 5021ac23..7bdfd8c7 100644 --- a/src/vnet/mpls/mpls.c +++ b/src/vnet/mpls/mpls.c @@ -536,6 +536,78 @@ VLIB_CLI_COMMAND (mpls_local_label_command, static) = { .short_help = "Create/Delete MPL local labels", }; +clib_error_t * +vnet_mpls_table_cmd (vlib_main_t * vm, + unformat_input_t * main_input, + vlib_cli_command_t * cmdo) +{ + unformat_input_t _line_input, *line_input = &_line_input; + clib_error_t *error = NULL; + u32 table_id, is_add; + + is_add = 1; + table_id = ~0; + + /* Get a line of input. */ + if (!unformat_user (main_input, unformat_line_input, line_input)) + return 0; + + while (unformat_check_input (line_input) != UNFORMAT_END_OF_INPUT) + { + if (unformat (line_input, "%d", &table_id)) + ; + else if (unformat (line_input, "del")) + is_add = 0; + else if (unformat (line_input, "add")) + is_add = 1; + else + { + error = unformat_parse_error (line_input); + goto done; + } + } + + if (~0 == table_id) + { + error = clib_error_return (0, "No table id"); + goto done; + } + else if (0 == table_id) + { + error = clib_error_return (0, "Can't change the default table"); + goto done; + } + else + { + if (is_add) + { + mpls_table_create (table_id, 0); + } + else + { + mpls_table_delete (table_id, 0); + } + } + + done: + unformat_free (line_input); + return error; +} + +/* *INDENT-ON* */ +/*? + * This command is used to add or delete MPLS Tables. All + * Tables must be explicitly added before that can be used, + * Including the default table. + ?*/ +/* *INDENT-OFF* */ +VLIB_CLI_COMMAND (ip6_table_command, static) = { + .path = "mpla table", + .short_help = "mpls table [add|del] ", + .function = vnet_mpls_table_cmd, + .is_mp_safe = 1, +}; + int mpls_fib_reset_labels (u32 fib_id) { @@ -546,12 +618,8 @@ mpls_fib_reset_labels (u32 fib_id) static clib_error_t * mpls_init (vlib_main_t * vm) { - mpls_main_t * mm = &mpls_main; clib_error_t * error; - mm->vlib_main = vm; - mm->vnet_main = vnet_get_main(); - if ((error = vlib_call_init_function (vm, ip_main_init))) return error; diff --git a/src/vnet/mpls/mpls.h b/src/vnet/mpls/mpls.h index b0125e60..31cb1746 100644 --- a/src/vnet/mpls/mpls.h +++ b/src/vnet/mpls/mpls.h @@ -56,10 +56,6 @@ typedef struct { /* IP4 enabled count by software interface */ u8 * mpls_enabled_by_sw_if_index; - - /* convenience */ - vlib_main_t * vlib_main; - vnet_main_t * vnet_main; } mpls_main_t; extern mpls_main_t mpls_main; @@ -77,8 +73,6 @@ extern vlib_node_registration_t mpls_midchain_node; /* Parse mpls protocol as 0xXXXX or protocol name. In either host or network byte order. */ -unformat_function_t unformat_mpls_protocol_host_byte_order; -unformat_function_t unformat_mpls_protocol_net_byte_order; unformat_function_t unformat_mpls_label_net_byte_order; unformat_function_t unformat_mpls_unicast_label; @@ -86,9 +80,10 @@ unformat_function_t unformat_mpls_unicast_label; unformat_function_t unformat_mpls_header; unformat_function_t unformat_pg_mpls_header; -void mpls_sw_interface_enable_disable (mpls_main_t * mm, - u32 sw_if_index, - u8 is_enable); +int mpls_sw_interface_enable_disable (mpls_main_t * mm, + u32 sw_if_index, + u8 is_enable, + u8 is_api); u8 mpls_sw_interface_is_enabled (u32 sw_if_index); @@ -103,4 +98,7 @@ mpls_fib_index_cmp(void * a1, void * a2); int mpls_label_cmp(void * a1, void * a2); +void mpls_table_create(uint32_t table_id, u8 is_api); +void mpls_table_delete(uint32_t table_id, u8 is_api); + #endif /* included_vnet_mpls_h */ diff --git a/src/vnet/mpls/mpls_api.c b/src/vnet/mpls/mpls_api.c index a44b1a25..38f5b014 100644 --- a/src/vnet/mpls/mpls_api.c +++ b/src/vnet/mpls/mpls_api.c @@ -58,6 +58,29 @@ _(MPLS_FIB_DUMP, mpls_fib_dump) extern void stats_dslock_with_hint (int hint, int tag); extern void stats_dsunlock (void); +void +mpls_table_delete (u32 table_id, u8 is_api) +{ + u32 fib_index; + + /* + * The MPLS defult table must also be explicitly created via the API. + * So in contrast to IP, it gets no special treatment here. + * + * The API holds only one lock on the table. + * i.e. it can be added many times via the API but needs to be + * deleted only once. + */ + fib_index = fib_table_find (FIB_PROTOCOL_MPLS, table_id); + + if (~0 != fib_index) + { + fib_table_unlock (fib_index, + FIB_PROTOCOL_MPLS, + (is_api ? FIB_SOURCE_API : FIB_SOURCE_CLI)); + } +} + void vl_api_mpls_table_add_del_t_handler (vl_api_mpls_table_add_del_t * mp) { @@ -68,6 +91,13 @@ vl_api_mpls_table_add_del_t_handler (vl_api_mpls_table_add_del_t * mp) vnm = vnet_get_main (); vnm->api_errno = 0; + if (mp->mt_is_add) + mpls_table_create (ntohl (mp->mt_table_id), 1); + else + mpls_table_delete (ntohl (mp->mt_table_id), 1); + + rv = (rv == 0) ? vnm->api_errno : rv; + REPLY_MACRO (VL_API_MPLS_TABLE_ADD_DEL_REPLY); } @@ -82,14 +112,7 @@ mpls_ip_bind_unbind_handler (vnet_main_t * vnm, if (~0 == mpls_fib_index) { - if (mp->mb_create_table_if_needed) - { - mpls_fib_index = - fib_table_find_or_create_and_lock (FIB_PROTOCOL_MPLS, - ntohl (mp->mb_mpls_table_id)); - } - else - return VNET_API_ERROR_NO_SUCH_FIB; + return VNET_API_ERROR_NO_SUCH_FIB; } ip_fib_index = fib_table_find ((mp->mb_is_ip4 ? @@ -170,7 +193,6 @@ mpls_route_add_del_t_handler (vnet_main_t * vnm, mp->mr_next_hop_sw_if_index, pfx.fp_payload_proto, mp->mr_next_hop_table_id, - mp->mr_create_table_if_needed, mp->mr_is_rpf_id, &fib_index, &next_hop_fib_index); @@ -235,6 +257,32 @@ vl_api_mpls_route_add_del_t_handler (vl_api_mpls_route_add_del_t * mp) REPLY_MACRO (VL_API_MPLS_ROUTE_ADD_DEL_REPLY); } +void +mpls_table_create (u32 table_id, u8 is_api) +{ + u32 fib_index; + + /* + * The MPLS defult table must also be explicitly created via the API. + * So in contrast to IP, it gets no special treatment here. + */ + + /* + * The API holds only one lock on the table. + * i.e. it can be added many times via the API but needs to be + * deleted only once. + */ + fib_index = fib_table_find (FIB_PROTOCOL_MPLS, table_id); + + if (~0 == fib_index) + { + fib_table_find_or_create_and_lock (FIB_PROTOCOL_MPLS, + table_id, + (is_api ? + FIB_SOURCE_API : FIB_SOURCE_CLI)); + } +} + static void vl_api_mpls_tunnel_add_del_t_handler (vl_api_mpls_tunnel_add_del_t * mp) { diff --git a/src/vnet/srv6/sr_policy_rewrite.c b/src/vnet/srv6/sr_policy_rewrite.c index f427bbf3..2f90993a 100755 --- a/src/vnet/srv6/sr_policy_rewrite.c +++ b/src/vnet/srv6/sr_policy_rewrite.c @@ -595,8 +595,10 @@ sr_policy_add (ip6_address_t * bsid, ip6_address_t * segments, if (sm->fib_table_ip6 == (u32) ~ 0) { sm->fib_table_ip6 = fib_table_create_and_lock (FIB_PROTOCOL_IP6, + FIB_SOURCE_SR, "SRv6 steering of IP6 prefixes through BSIDs"); sm->fib_table_ip4 = fib_table_create_and_lock (FIB_PROTOCOL_IP6, + FIB_SOURCE_SR, "SRv6 steering of IP4 prefixes through BSIDs"); } @@ -684,8 +686,8 @@ sr_policy_del (ip6_address_t * bsid, u32 index) /* If FIB empty unlock it */ if (!pool_elts (sm->sr_policies) && !pool_elts (sm->steer_policies)) { - fib_table_unlock (sm->fib_table_ip6, FIB_PROTOCOL_IP6); - fib_table_unlock (sm->fib_table_ip4, FIB_PROTOCOL_IP6); + fib_table_unlock (sm->fib_table_ip6, FIB_PROTOCOL_IP6, FIB_SOURCE_SR); + fib_table_unlock (sm->fib_table_ip4, FIB_PROTOCOL_IP6, FIB_SOURCE_SR); sm->fib_table_ip6 = (u32) ~ 0; sm->fib_table_ip4 = (u32) ~ 0; } diff --git a/src/vnet/srv6/sr_steering.c b/src/vnet/srv6/sr_steering.c index 57fe21f6..cf4e81ab 100755 --- a/src/vnet/srv6/sr_steering.c +++ b/src/vnet/srv6/sr_steering.c @@ -159,8 +159,10 @@ sr_steering_policy (int is_del, ip6_address_t * bsid, u32 sr_policy_index, /* If no more SR policies or steering policies */ if (!pool_elts (sm->sr_policies) && !pool_elts (sm->steer_policies)) { - fib_table_unlock (sm->fib_table_ip6, FIB_PROTOCOL_IP6); - fib_table_unlock (sm->fib_table_ip4, FIB_PROTOCOL_IP6); + fib_table_unlock (sm->fib_table_ip6, + FIB_PROTOCOL_IP6, FIB_SOURCE_SR); + fib_table_unlock (sm->fib_table_ip4, + FIB_PROTOCOL_IP6, FIB_SOURCE_SR); sm->fib_table_ip6 = (u32) ~ 0; sm->fib_table_ip4 = (u32) ~ 0; } diff --git a/src/vpp/api/api.c b/src/vpp/api/api.c index f9c3129c..044ddb5b 100644 --- a/src/vpp/api/api.c +++ b/src/vpp/api/api.c @@ -699,8 +699,9 @@ static void VALIDATE_SW_IF_INDEX (mp); - mpls_sw_interface_enable_disable (&mpls_main, - ntohl (mp->sw_if_index), mp->enable); + rv = mpls_sw_interface_enable_disable (&mpls_main, + ntohl (mp->sw_if_index), + mp->enable, 1); BAD_SW_IF_INDEX_LABEL; REPLY_MACRO (VL_API_SW_INTERFACE_SET_MPLS_ENABLE_REPLY); diff --git a/src/vpp/api/custom_dump.c b/src/vpp/api/custom_dump.c index 1353fe28..be74b83a 100644 --- a/src/vpp/api/custom_dump.c +++ b/src/vpp/api/custom_dump.c @@ -560,9 +560,6 @@ static void *vl_api_ip_add_del_route_t_print if (mp->table_id != 0) s = format (s, "vrf %d ", ntohl (mp->table_id)); - if (mp->create_vrf_if_needed) - s = format (s, "create-vrf "); - if (mp->next_hop_weight != 1) s = format (s, "weight %d ", mp->next_hop_weight); diff --git a/test/test_dhcp.py b/test/test_dhcp.py index 6fc29182..fe97f6c9 100644 --- a/test/test_dhcp.py +++ b/test/test_dhcp.py @@ -6,7 +6,7 @@ import struct from framework import VppTestCase, VppTestRunner from vpp_neighbor import VppNeighbor -from vpp_ip_route import find_route +from vpp_ip_route import find_route, VppIpTable from util import mk_ll_addr from scapy.layers.l2 import Ether, getmacbyip, ARP @@ -34,9 +34,19 @@ class TestDHCP(VppTestCase): # create 3 pg interfaces self.create_pg_interfaces(range(4)) + self.tables = [] # pg0 and 1 are IP configured in VRF 0 and 1. # pg2 and 3 are non IP-configured in VRF 0 and 1 + table_id = 0 + for table_id in range(1, 4): + tbl4 = VppIpTable(self, table_id) + tbl4.add_vpp_config() + self.tables.append(tbl4) + tbl6 = VppIpTable(self, table_id, is_ip6=1) + tbl6.add_vpp_config() + self.tables.append(tbl6) + table_id = 0 for i in self.pg_interfaces[:2]: i.admin_up() @@ -56,11 +66,15 @@ class TestDHCP(VppTestCase): table_id += 1 def tearDown(self): - super(TestDHCP, self).tearDown() - for i in self.pg_interfaces: + for i in self.pg_interfaces[:2]: i.unconfig_ip4() i.unconfig_ip6() + + for i in self.pg_interfaces: + i.set_table_ip4(0) + i.set_table_ip6(0) i.admin_down() + super(TestDHCP, self).tearDown() def send_and_assert_no_replies(self, intf, pkts, remark): intf.add_stream(pkts) @@ -667,6 +681,8 @@ class TestDHCP(VppTestCase): "DHCP cleanup VRF 0") self.send_and_assert_no_replies(self.pg3, pkts_disc_vrf1, "DHCP cleanup VRF 1") + self.pg2.unconfig_ip4() + self.pg3.unconfig_ip4() def test_dhcp6_proxy(self): """ DHCPv6 Proxy""" @@ -1045,6 +1061,8 @@ class TestDHCP(VppTestCase): server_table_id=0, is_ipv6=1, is_add=0) + self.pg2.unconfig_ip6() + self.pg3.unconfig_ip6() def test_dhcp_client(self): """ DHCP Client""" diff --git a/test/test_gre.py b/test/test_gre.py index 1afc44fb..9046b05f 100644 --- a/test/test_gre.py +++ b/test/test_gre.py @@ -6,7 +6,7 @@ from logging import * from framework import VppTestCase, VppTestRunner from vpp_sub_interface import VppDot1QSubint from vpp_gre_interface import VppGreInterface, VppGre6Interface -from vpp_ip_route import VppIpRoute, VppRoutePath, DpoProto +from vpp_ip_route import VppIpRoute, VppRoutePath, DpoProto, VppIpTable from vpp_papi_provider import L2_VTR_OP from scapy.packet import Raw @@ -30,6 +30,9 @@ class TestGRE(VppTestCase): # create 3 pg interfaces - set one in a non-default table. self.create_pg_interfaces(range(3)) + + self.tbl = VppIpTable(self, 1) + self.tbl.add_vpp_config() self.pg1.set_table_ip4(1) for i in self.pg_interfaces: @@ -43,11 +46,12 @@ class TestGRE(VppTestCase): self.pg2.resolve_ndp() def tearDown(self): - super(TestGRE, self).tearDown() for i in self.pg_interfaces: i.unconfig_ip4() i.unconfig_ip6() i.admin_down() + self.pg1.set_table_ip4(0) + super(TestGRE, self).tearDown() def create_stream_ip4(self, src_if, src_ip, dst_ip): pkts = [] diff --git a/test/test_ip4.py b/test/test_ip4.py index 7a7098c3..55d16735 100644 --- a/test/test_ip4.py +++ b/test/test_ip4.py @@ -6,7 +6,8 @@ import unittest from framework import VppTestCase, VppTestRunner from vpp_sub_interface import VppSubInterface, VppDot1QSubint, VppDot1ADSubint from vpp_ip_route import VppIpRoute, VppRoutePath, VppIpMRoute, \ - VppMRoutePath, MRouteItfFlags, MRouteEntryFlags, VppMplsIpBind + VppMRoutePath, MRouteItfFlags, MRouteEntryFlags, VppMplsIpBind, \ + VppMplsTable from scapy.packet import Raw from scapy.layers.l2 import Ether, Dot1Q, ARP @@ -774,6 +775,8 @@ class TestIPLoadBalance(VppTestCase): super(TestIPLoadBalance, self).setUp() self.create_pg_interfaces(range(5)) + mpls_tbl = VppMplsTable(self, 0) + mpls_tbl.add_vpp_config() for i in self.pg_interfaces: i.admin_up() @@ -782,11 +785,11 @@ class TestIPLoadBalance(VppTestCase): i.enable_mpls() def tearDown(self): - super(TestIPLoadBalance, self).tearDown() for i in self.pg_interfaces: i.disable_mpls() i.unconfig_ip4() i.admin_down() + super(TestIPLoadBalance, self).tearDown() def send_and_expect_load_balancing(self, input, pkts, outputs): input.add_stream(pkts) @@ -966,6 +969,8 @@ class TestIPVlan0(VppTestCase): super(TestIPVlan0, self).setUp() self.create_pg_interfaces(range(2)) + mpls_tbl = VppMplsTable(self, 0) + mpls_tbl.add_vpp_config() for i in self.pg_interfaces: i.admin_up() @@ -974,11 +979,11 @@ class TestIPVlan0(VppTestCase): i.enable_mpls() def tearDown(self): - super(TestIPVlan0, self).tearDown() for i in self.pg_interfaces: i.disable_mpls() i.unconfig_ip4() i.admin_down() + super(TestIPVlan0, self).tearDown() def send_and_expect(self, input, pkts, output): input.add_stream(pkts) diff --git a/test/test_ip4_vrf_multi_instance.py b/test/test_ip4_vrf_multi_instance.py index b73ac948..5a8d6760 100644 --- a/test/test_ip4_vrf_multi_instance.py +++ b/test/test_ip4_vrf_multi_instance.py @@ -172,9 +172,10 @@ class TestIp4VrfMultiInst(VppTestCase): pg_if = self.pg_if_by_vrf_id[vrf_id][0] dest_addr = pg_if.remote_hosts[0].ip4n dest_addr_len = 24 + self.vapi.ip_table_add_del(vrf_id, is_add=1) self.vapi.ip_add_del_route( dest_addr, dest_addr_len, pg_if.local_ip4n, - table_id=vrf_id, create_vrf_if_needed=1, is_multipath=1) + table_id=vrf_id, is_multipath=1) self.logger.info("IPv4 VRF ID %d created" % vrf_id) if vrf_id not in self.vrf_list: self.vrf_list.append(vrf_id) @@ -216,6 +217,7 @@ class TestIp4VrfMultiInst(VppTestCase): self.logger.info("IPv4 VRF ID %d reset" % vrf_id) self.logger.debug(self.vapi.ppcli("show ip fib")) self.logger.debug(self.vapi.ppcli("show ip arp")) + self.vapi.ip_table_add_del(vrf_id, is_add=0) def create_stream(self, src_if, packet_sizes): """ diff --git a/test/test_ip6.py b/test/test_ip6.py index 285ce181..aad3713c 100644 --- a/test/test_ip6.py +++ b/test/test_ip6.py @@ -8,7 +8,7 @@ from vpp_sub_interface import VppSubInterface, VppDot1QSubint from vpp_pg_interface import is_ipv6_misc from vpp_ip_route import VppIpRoute, VppRoutePath, find_route, VppIpMRoute, \ VppMRoutePath, MRouteItfFlags, MRouteEntryFlags, VppMplsIpBind, \ - VppMplsRoute, DpoProto + VppMplsRoute, DpoProto, VppMplsTable from vpp_neighbor import find_nbr, VppNeighbor from scapy.packet import Raw @@ -1260,6 +1260,9 @@ class TestIP6LoadBalance(VppTestCase): self.create_pg_interfaces(range(5)) + mpls_tbl = VppMplsTable(self, 0) + mpls_tbl.add_vpp_config() + for i in self.pg_interfaces: i.admin_up() i.config_ip6() @@ -1267,11 +1270,11 @@ class TestIP6LoadBalance(VppTestCase): i.enable_mpls() def tearDown(self): - super(TestIP6LoadBalance, self).tearDown() for i in self.pg_interfaces: i.unconfig_ip6() i.admin_down() i.disable_mpls() + super(TestIP6LoadBalance, self).tearDown() def send_and_expect_load_balancing(self, input, pkts, outputs): input.add_stream(pkts) diff --git a/test/test_ip6_vrf_multi_instance.py b/test/test_ip6_vrf_multi_instance.py index af80b5ba..769cb2e5 100644 --- a/test/test_ip6_vrf_multi_instance.py +++ b/test/test_ip6_vrf_multi_instance.py @@ -187,9 +187,10 @@ class TestIP6VrfMultiInst(VppTestCase): pg_if = self.pg_if_by_vrf_id[vrf_id][0] dest_addr = pg_if.remote_hosts[0].ip6n dest_addr_len = 64 + self.vapi.ip_table_add_del(vrf_id, is_add=1, is_ipv6=1) self.vapi.ip_add_del_route( dest_addr, dest_addr_len, pg_if.local_ip6n, is_ipv6=1, - table_id=vrf_id, create_vrf_if_needed=1, is_multipath=1) + table_id=vrf_id, is_multipath=1) self.logger.info("IPv6 VRF ID %d created" % vrf_id) if vrf_id not in self.vrf_list: self.vrf_list.append(vrf_id) @@ -232,6 +233,7 @@ class TestIP6VrfMultiInst(VppTestCase): self.logger.info("IPv6 VRF ID %d reset" % vrf_id) self.logger.debug(self.vapi.ppcli("show ip6 fib")) self.logger.debug(self.vapi.ppcli("show ip6 neighbors")) + self.vapi.ip_table_add_del(vrf_id, is_add=0, is_ipv6=1) def create_stream(self, src_if, packet_sizes): """ diff --git a/test/test_ip_mcast.py b/test/test_ip_mcast.py index 276555d6..7cad683c 100644 --- a/test/test_ip_mcast.py +++ b/test/test_ip_mcast.py @@ -5,7 +5,7 @@ import unittest from framework import VppTestCase, VppTestRunner from vpp_sub_interface import VppSubInterface, VppDot1QSubint, VppDot1ADSubint from vpp_ip_route import VppIpMRoute, VppMRoutePath, VppMFibSignal, \ - MRouteItfFlags, MRouteEntryFlags + MRouteItfFlags, MRouteEntryFlags, VppIpTable from scapy.packet import Raw from scapy.layers.l2 import Ether @@ -44,16 +44,37 @@ class TestIPMcast(VppTestCase): super(TestIPMcast, self).setUp() # create 8 pg interfaces - self.create_pg_interfaces(range(8)) + self.create_pg_interfaces(range(9)) # setup interfaces - for i in self.pg_interfaces: + for i in self.pg_interfaces[:8]: i.admin_up() i.config_ip4() i.config_ip6() i.resolve_arp() i.resolve_ndp() + # one more in a vrf + tbl4 = VppIpTable(self, 10) + tbl4.add_vpp_config() + self.pg8.set_table_ip4(10) + self.pg8.config_ip4() + + tbl6 = VppIpTable(self, 10, is_ip6=1) + tbl6.add_vpp_config() + self.pg8.set_table_ip6(10) + self.pg8.config_ip6() + + def tearDown(self): + for i in self.pg_interfaces: + i.unconfig_ip4() + i.unconfig_ip6() + i.admin_down() + + self.pg8.set_table_ip4(0) + self.pg8.set_table_ip6(0) + super(TestIPMcast, self).tearDown() + def create_stream_ip4(self, src_if, src_ip, dst_ip, payload_size=0): pkts = [] # default to small packet sizes @@ -663,6 +684,77 @@ class TestIPMcast(VppTestCase): # route_232_1_1_1.remove_vpp_config() + def test_ip_mcast_vrf(self): + """ IP Multicast Replication in non-default table""" + + # + # An (S,G). + # one accepting interface, pg0, 2 forwarding interfaces + # + route_1_1_1_1_232_1_1_1 = VppIpMRoute( + self, + "1.1.1.1", + "232.1.1.1", 64, + MRouteEntryFlags.MFIB_ENTRY_FLAG_NONE, + [VppMRoutePath(self.pg8.sw_if_index, + MRouteItfFlags.MFIB_ITF_FLAG_ACCEPT), + VppMRoutePath(self.pg1.sw_if_index, + MRouteItfFlags.MFIB_ITF_FLAG_FORWARD), + VppMRoutePath(self.pg2.sw_if_index, + MRouteItfFlags.MFIB_ITF_FLAG_FORWARD)], + table_id=10) + route_1_1_1_1_232_1_1_1.add_vpp_config() + + # + # a stream that matches the route for (1.1.1.1,232.1.1.1) + # small packets + # + self.vapi.cli("clear trace") + tx = self.create_stream_ip4(self.pg8, "1.1.1.1", "232.1.1.1") + self.pg8.add_stream(tx) + + self.pg_enable_capture(self.pg_interfaces) + self.pg_start() + + # We expect replications on Pg1 & 2 + self.verify_capture_ip4(self.pg1, tx) + self.verify_capture_ip4(self.pg2, tx) + + def test_ip6_mcast_vrf(self): + """ IPv6 Multicast Replication in non-default table""" + + # + # An (S,G). + # one accepting interface, pg0, 2 forwarding interfaces + # + route_2001_ff01_1 = VppIpMRoute( + self, + "2001::1", + "ff01::1", 256, + MRouteEntryFlags.MFIB_ENTRY_FLAG_NONE, + [VppMRoutePath(self.pg8.sw_if_index, + MRouteItfFlags.MFIB_ITF_FLAG_ACCEPT), + VppMRoutePath(self.pg1.sw_if_index, + MRouteItfFlags.MFIB_ITF_FLAG_FORWARD), + VppMRoutePath(self.pg2.sw_if_index, + MRouteItfFlags.MFIB_ITF_FLAG_FORWARD)], + table_id=10, + is_ip6=1) + route_2001_ff01_1.add_vpp_config() + + # + # a stream that matches the route for (2001::1, ff00::1) + # + self.vapi.cli("clear trace") + tx = self.create_stream_ip6(self.pg8, "2001::1", "ff01::1") + self.pg8.add_stream(tx) + + self.pg_enable_capture(self.pg_interfaces) + self.pg_start() + + # We expect replications on Pg1, 2, + self.verify_capture_ip6(self.pg1, tx) + self.verify_capture_ip6(self.pg2, tx) if __name__ == '__main__': unittest.main(testRunner=VppTestRunner) diff --git a/test/test_mpls.py b/test/test_mpls.py index b2226a74..460a32d1 100644 --- a/test/test_mpls.py +++ b/test/test_mpls.py @@ -6,7 +6,7 @@ import socket from framework import VppTestCase, VppTestRunner from vpp_ip_route import VppIpRoute, VppRoutePath, VppMplsRoute, \ VppMplsIpBind, VppIpMRoute, VppMRoutePath, \ - MRouteItfFlags, MRouteEntryFlags, DpoProto + MRouteItfFlags, MRouteEntryFlags, DpoProto, VppIpTable, VppMplsTable from vpp_mpls_tunnel_interface import VppMPLSTunnelInterface from scapy.packet import Raw @@ -60,9 +60,23 @@ class TestMPLS(VppTestCase): # setup both interfaces # assign them different tables. table_id = 0 + self.tables = [] + + tbl = VppMplsTable(self, 0) + tbl.add_vpp_config() + self.tables.append(tbl) for i in self.pg_interfaces: i.admin_up() + + if table_id != 0: + tbl = VppIpTable(self, table_id) + tbl.add_vpp_config() + self.tables.append(tbl) + tbl = VppIpTable(self, table_id, is_ip6=1) + tbl.add_vpp_config() + self.tables.append(tbl) + i.set_table_ip4(table_id) i.set_table_ip6(table_id) i.config_ip4() @@ -73,12 +87,15 @@ class TestMPLS(VppTestCase): table_id += 1 def tearDown(self): - super(TestMPLS, self).tearDown() for i in self.pg_interfaces: i.unconfig_ip4() i.unconfig_ip6() i.ip6_disable() + i.set_table_ip4(0) + i.set_table_ip6(0) + i.disable_mpls() i.admin_down() + super(TestMPLS, self).tearDown() # the default of 64 matches the IP packet TTL default def create_stream_labelled_ip4( @@ -1092,6 +1109,9 @@ class TestMPLSDisabled(VppTestCase): # create 2 pg interfaces self.create_pg_interfaces(range(2)) + self.tbl = VppMplsTable(self, 0) + self.tbl.add_vpp_config() + # PG0 is MPLS enalbed self.pg0.admin_up() self.pg0.config_ip4() @@ -1102,11 +1122,13 @@ class TestMPLSDisabled(VppTestCase): self.pg1.admin_up() def tearDown(self): - super(TestMPLSDisabled, self).tearDown() for i in self.pg_interfaces: i.unconfig_ip4() i.admin_down() + self.pg0.disable_mpls() + super(TestMPLSDisabled, self).tearDown() + def send_and_assert_no_replies(self, intf, pkts, remark): intf.add_stream(pkts) self.pg_enable_capture(self.pg_interfaces) @@ -1174,6 +1196,13 @@ class TestMPLSPIC(VppTestCase): # create 2 pg interfaces self.create_pg_interfaces(range(4)) + mpls_tbl = VppMplsTable(self, 0) + mpls_tbl.add_vpp_config() + tbl4 = VppIpTable(self, 1) + tbl4.add_vpp_config() + tbl6 = VppIpTable(self, 1, is_ip6=1) + tbl6.add_vpp_config() + # core links self.pg0.admin_up() self.pg0.config_ip4() @@ -1201,14 +1230,15 @@ class TestMPLSPIC(VppTestCase): self.pg3.resolve_ndp() def tearDown(self): - super(TestMPLSPIC, self).tearDown() self.pg0.disable_mpls() + self.pg1.disable_mpls() for i in self.pg_interfaces: i.unconfig_ip4() i.unconfig_ip6() i.set_table_ip4(0) i.set_table_ip6(0) i.admin_down() + super(TestMPLSPIC, self).tearDown() def test_mpls_ibgp_pic(self): """ MPLS iBGP PIC edge convergence @@ -1534,24 +1564,30 @@ class TestMPLSL2(VppTestCase): # create 2 pg interfaces self.create_pg_interfaces(range(2)) + # create the default MPLS table + self.tables = [] + tbl = VppMplsTable(self, 0) + tbl.add_vpp_config() + self.tables.append(tbl) + # use pg0 as the core facing interface self.pg0.admin_up() self.pg0.config_ip4() self.pg0.resolve_arp() self.pg0.enable_mpls() - # use the other 2 for customer facg L2 links + # use the other 2 for customer facing L2 links for i in self.pg_interfaces[1:]: i.admin_up() def tearDown(self): - super(TestMPLSL2, self).tearDown() for i in self.pg_interfaces[1:]: i.admin_down() self.pg0.disable_mpls() self.pg0.unconfig_ip4() self.pg0.admin_down() + super(TestMPLSL2, self).tearDown() def verify_capture_tunneled_ethernet(self, capture, sent, mpls_labels, ttl=255, top=None): diff --git a/test/test_nat.py b/test/test_nat.py index 1f2d17ab..73e9e217 100644 --- a/test/test_nat.py +++ b/test/test_nat.py @@ -549,6 +549,8 @@ class TestNAT44(MethodHolder): cls.pg0.configure_ipv4_neighbors() cls.overlapping_interfaces = list(list(cls.pg_interfaces[4:7])) + cls.vapi.ip_table_add_del(10, is_add=1) + cls.vapi.ip_table_add_del(20, is_add=1) cls.pg4._local_ip4 = "172.16.255.1" cls.pg4._local_ip4n = socket.inet_pton(socket.AF_INET, i.local_ip4) @@ -1797,6 +1799,8 @@ class TestNAT44(MethodHolder): self.pg0.unconfig_ip4() self.pg1.unconfig_ip4() + self.vapi.ip_table_add_del(vrf_id1, is_add=1) + self.vapi.ip_table_add_del(vrf_id2, is_add=1) self.pg0.set_table_ip4(vrf_id1) self.pg1.set_table_ip4(vrf_id2) self.pg0.config_ip4() @@ -1825,6 +1829,13 @@ class TestNAT44(MethodHolder): capture = self.pg2.get_capture(len(pkts)) self.verify_capture_out(capture, nat_ip2) + self.pg0.unconfig_ip4() + self.pg1.unconfig_ip4() + self.pg0.set_table_ip4(0) + self.pg1.set_table_ip4(0) + self.vapi.ip_table_add_del(vrf_id1, is_add=0) + self.vapi.ip_table_add_del(vrf_id2, is_add=0) + def test_vrf_feature_independent(self): """ NAT44 tenant VRF independent address pool mode """ @@ -3042,6 +3053,8 @@ class TestNAT64(MethodHolder): cls.ip6_interfaces.append(cls.pg_interfaces[2]) cls.ip4_interfaces = list(cls.pg_interfaces[1:2]) + cls.vapi.ip_table_add_del(cls.vrf1_id, is_add=1, is_ipv6=1) + cls.pg_interfaces[2].set_table_ip6(cls.vrf1_id) cls.pg0.generate_remote_hosts(2) diff --git a/test/test_neighbor.py b/test/test_neighbor.py index 1c7cc267..68dde2fb 100644 --- a/test/test_neighbor.py +++ b/test/test_neighbor.py @@ -5,7 +5,8 @@ from socket import AF_INET, AF_INET6, inet_pton from framework import VppTestCase, VppTestRunner from vpp_neighbor import VppNeighbor, find_nbr -from vpp_ip_route import VppIpRoute, VppRoutePath, find_route +from vpp_ip_route import VppIpRoute, VppRoutePath, find_route, \ + VppIpTable from scapy.packet import Raw from scapy.layers.l2 import Ether, ARP, Dot1Q @@ -39,11 +40,13 @@ class ARPTestCase(VppTestCase): self.pg1.config_ip6() # pg3 in a different VRF + self.tbl = VppIpTable(self, 1) + self.tbl.add_vpp_config() + self.pg3.set_table_ip4(1) self.pg3.config_ip4() def tearDown(self): - super(ARPTestCase, self).tearDown() self.pg0.unconfig_ip4() self.pg0.unconfig_ip6() @@ -51,10 +54,13 @@ class ARPTestCase(VppTestCase): self.pg1.unconfig_ip6() self.pg3.unconfig_ip4() + self.pg3.set_table_ip4(0) for i in self.pg_interfaces: i.admin_down() + super(ARPTestCase, self).tearDown() + def verify_arp_req(self, rx, smac, sip, dip): ether = rx[Ether] self.assertEqual(ether.dst, "ff:ff:ff:ff:ff:ff") @@ -1080,6 +1086,62 @@ class ARPTestCase(VppTestCase): self.pg0.remote_ip4, self.pg1.remote_hosts[1].ip4) + def test_arp_static(self): + """ ARP Static""" + self.pg2.generate_remote_hosts(3) + + # + # Add a static ARP entry + # + static_arp = VppNeighbor(self, + self.pg2.sw_if_index, + self.pg2.remote_hosts[1].mac, + self.pg2.remote_hosts[1].ip4, + is_static=1) + static_arp.add_vpp_config() + + # + # Add the connected prefix to the interface + # + self.pg2.config_ip4() + + # + # We should now find the adj-fib + # + self.assertTrue(find_nbr(self, + self.pg2.sw_if_index, + self.pg2.remote_hosts[1].ip4, + is_static=1)) + self.assertTrue(find_route(self, + self.pg2.remote_hosts[1].ip4, + 32)) + + # + # remove the connected + # + self.pg2.unconfig_ip4() + + # + # put the interface into table 1 + # + self.pg2.set_table_ip4(1) + + # + # configure the same connected and expect to find the + # adj fib in the new table + # + self.pg2.config_ip4() + self.assertTrue(find_route(self, + self.pg2.remote_hosts[1].ip4, + 32, + table_id=1)) + + # + # clean-up + # + self.pg2.unconfig_ip4() + self.pg2.set_table_ip4(0) + if __name__ == '__main__': unittest.main(testRunner=VppTestRunner) diff --git a/test/vpp_ip_route.py b/test/vpp_ip_route.py index 2c489e3c..b7993793 100644 --- a/test/vpp_ip_route.py +++ b/test/vpp_ip_route.py @@ -54,6 +54,46 @@ def find_route(test, ip_addr, len, table_id=0, inet=AF_INET): return False +class VppIpTable(VppObject): + + def __init__(self, + test, + table_id, + is_ip6=0): + self._test = test + self.table_id = table_id + self.is_ip6 = is_ip6 + + def add_vpp_config(self): + self._test.vapi.ip_table_add_del( + self.table_id, + is_ipv6=self.is_ip6, + is_add=1) + self._test.registry.register(self, self._test.logger) + + def remove_vpp_config(self): + self._test.vapi.ip_table_add_del( + self.table_id, + is_ipv6=self.is_ip6, + is_add=0) + + def query_vpp_config(self): + # find the default route + return find_route(self._test, + "::" if self.is_ip6 else "0.0.0.0", + 0, + self.table_id, + inet=AF_INET6 if self.is_ip6 == 1 else AF_INET) + + def __str__(self): + return self.object_id() + + def object_id(self): + return ("table-%s-%d" % + ("v6" if self.is_ip6 == 1 else "v4", + self.table_id)) + + class VppRoutePath(object): def __init__( @@ -391,6 +431,39 @@ class VppMplsIpBind(VppObject): self.dest_addr_len)) +class VppMplsTable(VppObject): + + def __init__(self, + test, + table_id): + self._test = test + self.table_id = table_id + + def add_vpp_config(self): + self._test.vapi.mpls_table_add_del( + self.table_id, + is_add=1) + self._test.registry.register(self, self._test.logger) + + def remove_vpp_config(self): + self._test.vapi.mpls_table_add_del( + self.table_id, + is_add=0) + + def query_vpp_config(self): + # find the default route + dump = self._test.vapi.mpls_fib_dump() + if len(dump): + return True + return False + + def __str__(self): + return self.object_id() + + def object_id(self): + return ("table-mpls-%d" % (self.table_id)) + + class VppMplsRoute(VppObject): """ MPLS Route/LSP diff --git a/test/vpp_papi_provider.py b/test/vpp_papi_provider.py index b70da026..519aff80 100644 --- a/test/vpp_papi_provider.py +++ b/test/vpp_papi_provider.py @@ -652,6 +652,24 @@ class VppPapiProvider(object): return self.api(self.papi.delete_loopback, {'sw_if_index': sw_if_index, }) + def ip_table_add_del(self, + table_id, + is_add=1, + is_ipv6=0): + """ + + :param table_id + :param is_add: (Default value = 1) + :param is_ipv6: (Default value = 0) + + """ + + return self.api( + self.papi.ip_table_add_del, + {'table_id': table_id, + 'is_add': is_add, + 'is_ipv6': is_ipv6}) + def ip_add_del_route( self, dst_address, @@ -664,7 +682,6 @@ class VppPapiProvider(object): next_hop_n_out_labels=0, next_hop_out_label_stack=[], next_hop_via_label=MPLS_LABEL_INVALID, - create_vrf_if_needed=0, is_resolve_host=0, is_resolve_attached=0, classify_table_index=0xFFFFFFFF, @@ -687,7 +704,6 @@ class VppPapiProvider(object): :param vrf_id: (Default value = 0) :param lookup_in_vrf: (Default value = 0) :param classify_table_index: (Default value = 0xFFFFFFFF) - :param create_vrf_if_needed: (Default value = 0) :param is_add: (Default value = 1) :param is_drop: (Default value = 0) :param is_ipv6: (Default value = 0) @@ -707,7 +723,6 @@ class VppPapiProvider(object): 'table_id': table_id, 'classify_table_index': classify_table_index, 'next_hop_table_id': next_hop_table_id, - 'create_vrf_if_needed': create_vrf_if_needed, 'is_add': is_add, 'is_drop': is_drop, 'is_unreach': is_unreach, @@ -912,6 +927,22 @@ class VppPapiProvider(object): def mpls_fib_dump(self): return self.api(self.papi.mpls_fib_dump, {}) + def mpls_table_add_del( + self, + table_id, + is_add=1): + """ + + :param table_id + :param is_add: (Default value = 1) + + """ + + return self.api( + self.papi.mpls_table_add_del, + {'mt_table_id': table_id, + 'mt_is_add': is_add}) + def mpls_route_add_del( self, label, @@ -925,7 +956,6 @@ class VppPapiProvider(object): next_hop_n_out_labels=0, next_hop_out_label_stack=[], next_hop_via_label=MPLS_LABEL_INVALID, - create_vrf_if_needed=0, is_resolve_host=0, is_resolve_attached=0, is_interface_rx=0, @@ -947,7 +977,6 @@ class VppPapiProvider(object): :param vrf_id: (Default value = 0) :param lookup_in_vrf: (Default value = 0) :param classify_table_index: (Default value = 0xFFFFFFFF) - :param create_vrf_if_needed: (Default value = 0) :param is_add: (Default value = 1) :param is_drop: (Default value = 0) :param is_ipv6: (Default value = 0) @@ -968,7 +997,6 @@ class VppPapiProvider(object): 'mr_eos': eos, 'mr_table_id': table_id, 'mr_classify_table_index': classify_table_index, - 'mr_create_table_if_needed': create_vrf_if_needed, 'mr_is_add': is_add, 'mr_is_classify': is_classify, 'mr_is_multipath': is_multipath, @@ -994,7 +1022,6 @@ class VppPapiProvider(object): table_id=0, ip_table_id=0, is_ip4=1, - create_vrf_if_needed=0, is_bind=1): """ """ @@ -1003,7 +1030,6 @@ class VppPapiProvider(object): {'mb_mpls_table_id': table_id, 'mb_label': label, 'mb_ip_table_id': ip_table_id, - 'mb_create_table_if_needed': create_vrf_if_needed, 'mb_is_bind': is_bind, 'mb_is_ip4': is_ip4, 'mb_address_length': dst_address_length, @@ -1020,7 +1046,6 @@ class VppPapiProvider(object): next_hop_n_out_labels=0, next_hop_out_label_stack=[], next_hop_via_label=MPLS_LABEL_INVALID, - create_vrf_if_needed=0, is_add=1, l2_only=0, is_multicast=0): @@ -1034,7 +1059,6 @@ class VppPapiProvider(object): :param vrf_id: (Default value = 0) :param lookup_in_vrf: (Default value = 0) :param classify_table_index: (Default value = 0xFFFFFFFF) - :param create_vrf_if_needed: (Default value = 0) :param is_add: (Default value = 1) :param is_drop: (Default value = 0) :param is_ipv6: (Default value = 0) @@ -1844,7 +1868,6 @@ class VppPapiProvider(object): i_flags, rpf_id=0, table_id=0, - create_vrf_if_needed=0, is_add=1, is_ipv6=0, is_local=0): @@ -1857,7 +1880,6 @@ class VppPapiProvider(object): 'itf_flags': i_flags, 'table_id': table_id, 'rpf_id': rpf_id, - 'create_vrf_if_needed': create_vrf_if_needed, 'is_add': is_add, 'is_ipv6': is_ipv6, 'is_local': is_local, -- cgit 1.2.3-korg