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/ip/ip4_forward.c | 3345 +++++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 3345 insertions(+) create mode 100644 src/vnet/ip/ip4_forward.c (limited to 'src/vnet/ip/ip4_forward.c') diff --git a/src/vnet/ip/ip4_forward.c b/src/vnet/ip/ip4_forward.c new file mode 100644 index 00000000..6e91b9e9 --- /dev/null +++ b/src/vnet/ip/ip4_forward.c @@ -0,0 +1,3345 @@ +/* + * 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/ip4_forward.c: IP v4 forwarding + * + * 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 +#include /* for ethernet_header_t */ +#include /* for ethernet_arp_header_t */ +#include +#include /* for srp_hw_interface_class */ +#include /* for API error numbers */ +#include /* for FIB table and entry creation */ +#include /* for FIB table and entry creation */ +#include /* for FIB uRPF check */ +#include +#include +#include + +/** + * @file + * @brief IPv4 Forwarding. + * + * This file contains the source code for IPv4 forwarding. + */ + +void +ip4_forward_next_trace (vlib_main_t * vm, + vlib_node_runtime_t * node, + vlib_frame_t * frame, + vlib_rx_or_tx_t which_adj_index); + +always_inline uword +ip4_lookup_inline (vlib_main_t * vm, + vlib_node_runtime_t * node, + vlib_frame_t * frame, + int lookup_for_responses_to_locally_received_packets) +{ + ip4_main_t *im = &ip4_main; + 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 (); + + from = vlib_frame_vector_args (frame); + n_left_from = frame->n_vectors; + next = node->cached_next_index; + + while (n_left_from > 0) + { + vlib_get_next_frame (vm, node, next, to_next, n_left_to_next); + + while (n_left_from >= 8 && n_left_to_next >= 4) + { + vlib_buffer_t *p0, *p1, *p2, *p3; + ip4_header_t *ip0, *ip1, *ip2, *ip3; + __attribute__ ((unused)) tcp_header_t *tcp0, *tcp1, *tcp2, *tcp3; + ip_lookup_next_t next0, next1, next2, next3; + const load_balance_t *lb0, *lb1, *lb2, *lb3; + ip4_fib_mtrie_t *mtrie0, *mtrie1, *mtrie2, *mtrie3; + ip4_fib_mtrie_leaf_t leaf0, leaf1, leaf2, leaf3; + ip4_address_t *dst_addr0, *dst_addr1, *dst_addr2, *dst_addr3; + __attribute__ ((unused)) u32 pi0, fib_index0, lb_index0, + is_tcp_udp0; + __attribute__ ((unused)) u32 pi1, fib_index1, lb_index1, + is_tcp_udp1; + __attribute__ ((unused)) u32 pi2, fib_index2, lb_index2, + is_tcp_udp2; + __attribute__ ((unused)) u32 pi3, fib_index3, lb_index3, + is_tcp_udp3; + flow_hash_config_t flow_hash_config0, flow_hash_config1; + flow_hash_config_t flow_hash_config2, flow_hash_config3; + u32 hash_c0, hash_c1, hash_c2, hash_c3; + const dpo_id_t *dpo0, *dpo1, *dpo2, *dpo3; + + /* 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, sizeof (ip0[0]), LOAD); + CLIB_PREFETCH (p5->data, sizeof (ip0[0]), LOAD); + CLIB_PREFETCH (p6->data, sizeof (ip0[0]), LOAD); + CLIB_PREFETCH (p7->data, sizeof (ip0[0]), LOAD); + } + + pi0 = to_next[0] = from[0]; + pi1 = to_next[1] = from[1]; + pi2 = to_next[2] = from[2]; + pi3 = to_next[3] = from[3]; + + from += 4; + to_next += 4; + n_left_to_next -= 4; + n_left_from -= 4; + + p0 = vlib_get_buffer (vm, pi0); + p1 = vlib_get_buffer (vm, pi1); + p2 = vlib_get_buffer (vm, pi2); + p3 = vlib_get_buffer (vm, pi3); + + ip0 = vlib_buffer_get_current (p0); + ip1 = vlib_buffer_get_current (p1); + ip2 = vlib_buffer_get_current (p2); + ip3 = vlib_buffer_get_current (p3); + + dst_addr0 = &ip0->dst_address; + dst_addr1 = &ip1->dst_address; + dst_addr2 = &ip2->dst_address; + dst_addr3 = &ip3->dst_address; + + fib_index0 = + vec_elt (im->fib_index_by_sw_if_index, + vnet_buffer (p0)->sw_if_index[VLIB_RX]); + fib_index1 = + vec_elt (im->fib_index_by_sw_if_index, + vnet_buffer (p1)->sw_if_index[VLIB_RX]); + fib_index2 = + vec_elt (im->fib_index_by_sw_if_index, + vnet_buffer (p2)->sw_if_index[VLIB_RX]); + fib_index3 = + vec_elt (im->fib_index_by_sw_if_index, + vnet_buffer (p3)->sw_if_index[VLIB_RX]); + fib_index0 = + (vnet_buffer (p0)->sw_if_index[VLIB_TX] == + (u32) ~ 0) ? fib_index0 : vnet_buffer (p0)->sw_if_index[VLIB_TX]; + fib_index1 = + (vnet_buffer (p1)->sw_if_index[VLIB_TX] == + (u32) ~ 0) ? fib_index1 : vnet_buffer (p1)->sw_if_index[VLIB_TX]; + fib_index2 = + (vnet_buffer (p2)->sw_if_index[VLIB_TX] == + (u32) ~ 0) ? fib_index2 : vnet_buffer (p2)->sw_if_index[VLIB_TX]; + fib_index3 = + (vnet_buffer (p3)->sw_if_index[VLIB_TX] == + (u32) ~ 0) ? fib_index3 : vnet_buffer (p3)->sw_if_index[VLIB_TX]; + + + if (!lookup_for_responses_to_locally_received_packets) + { + mtrie0 = &ip4_fib_get (fib_index0)->mtrie; + mtrie1 = &ip4_fib_get (fib_index1)->mtrie; + mtrie2 = &ip4_fib_get (fib_index2)->mtrie; + mtrie3 = &ip4_fib_get (fib_index3)->mtrie; + + leaf0 = leaf1 = leaf2 = leaf3 = IP4_FIB_MTRIE_LEAF_ROOT; + + leaf0 = ip4_fib_mtrie_lookup_step (mtrie0, leaf0, dst_addr0, 0); + leaf1 = ip4_fib_mtrie_lookup_step (mtrie1, leaf1, dst_addr1, 0); + leaf2 = ip4_fib_mtrie_lookup_step (mtrie2, leaf2, dst_addr2, 0); + leaf3 = ip4_fib_mtrie_lookup_step (mtrie3, leaf3, dst_addr3, 0); + } + + tcp0 = (void *) (ip0 + 1); + tcp1 = (void *) (ip1 + 1); + tcp2 = (void *) (ip2 + 1); + tcp3 = (void *) (ip3 + 1); + + is_tcp_udp0 = (ip0->protocol == IP_PROTOCOL_TCP + || ip0->protocol == IP_PROTOCOL_UDP); + is_tcp_udp1 = (ip1->protocol == IP_PROTOCOL_TCP + || ip1->protocol == IP_PROTOCOL_UDP); + is_tcp_udp2 = (ip2->protocol == IP_PROTOCOL_TCP + || ip2->protocol == IP_PROTOCOL_UDP); + is_tcp_udp3 = (ip1->protocol == IP_PROTOCOL_TCP + || ip1->protocol == IP_PROTOCOL_UDP); + + if (!lookup_for_responses_to_locally_received_packets) + { + leaf0 = ip4_fib_mtrie_lookup_step (mtrie0, leaf0, dst_addr0, 1); + leaf1 = ip4_fib_mtrie_lookup_step (mtrie1, leaf1, dst_addr1, 1); + leaf2 = ip4_fib_mtrie_lookup_step (mtrie2, leaf2, dst_addr2, 1); + leaf3 = ip4_fib_mtrie_lookup_step (mtrie3, leaf3, dst_addr3, 1); + } + + if (!lookup_for_responses_to_locally_received_packets) + { + leaf0 = ip4_fib_mtrie_lookup_step (mtrie0, leaf0, dst_addr0, 2); + leaf1 = ip4_fib_mtrie_lookup_step (mtrie1, leaf1, dst_addr1, 2); + leaf2 = ip4_fib_mtrie_lookup_step (mtrie2, leaf2, dst_addr2, 2); + leaf3 = ip4_fib_mtrie_lookup_step (mtrie3, leaf3, dst_addr3, 2); + } + + if (!lookup_for_responses_to_locally_received_packets) + { + leaf0 = ip4_fib_mtrie_lookup_step (mtrie0, leaf0, dst_addr0, 3); + leaf1 = ip4_fib_mtrie_lookup_step (mtrie1, leaf1, dst_addr1, 3); + leaf2 = ip4_fib_mtrie_lookup_step (mtrie2, leaf2, dst_addr2, 3); + leaf3 = ip4_fib_mtrie_lookup_step (mtrie3, leaf3, dst_addr3, 3); + } + + if (lookup_for_responses_to_locally_received_packets) + { + lb_index0 = vnet_buffer (p0)->ip.adj_index[VLIB_RX]; + lb_index1 = vnet_buffer (p1)->ip.adj_index[VLIB_RX]; + lb_index2 = vnet_buffer (p2)->ip.adj_index[VLIB_RX]; + lb_index3 = vnet_buffer (p3)->ip.adj_index[VLIB_RX]; + } + else + { + /* Handle default route. */ + leaf0 = + (leaf0 == + IP4_FIB_MTRIE_LEAF_EMPTY ? mtrie0->default_leaf : leaf0); + leaf1 = + (leaf1 == + IP4_FIB_MTRIE_LEAF_EMPTY ? mtrie1->default_leaf : leaf1); + leaf2 = + (leaf2 == + IP4_FIB_MTRIE_LEAF_EMPTY ? mtrie2->default_leaf : leaf2); + leaf3 = + (leaf3 == + IP4_FIB_MTRIE_LEAF_EMPTY ? mtrie3->default_leaf : leaf3); + lb_index0 = ip4_fib_mtrie_leaf_get_adj_index (leaf0); + lb_index1 = ip4_fib_mtrie_leaf_get_adj_index (leaf1); + lb_index2 = ip4_fib_mtrie_leaf_get_adj_index (leaf2); + lb_index3 = ip4_fib_mtrie_leaf_get_adj_index (leaf3); + } + + lb0 = load_balance_get (lb_index0); + lb1 = load_balance_get (lb_index1); + lb2 = load_balance_get (lb_index2); + lb3 = load_balance_get (lb_index3); + + /* Use flow hash to compute multipath adjacency. */ + hash_c0 = vnet_buffer (p0)->ip.flow_hash = 0; + hash_c1 = vnet_buffer (p1)->ip.flow_hash = 0; + hash_c2 = vnet_buffer (p2)->ip.flow_hash = 0; + hash_c3 = vnet_buffer (p3)->ip.flow_hash = 0; + if (PREDICT_FALSE (lb0->lb_n_buckets > 1)) + { + flow_hash_config0 = lb0->lb_hash_config; + hash_c0 = vnet_buffer (p0)->ip.flow_hash = + ip4_compute_flow_hash (ip0, flow_hash_config0); + } + if (PREDICT_FALSE (lb1->lb_n_buckets > 1)) + { + flow_hash_config1 = lb1->lb_hash_config; + hash_c1 = vnet_buffer (p1)->ip.flow_hash = + ip4_compute_flow_hash (ip1, flow_hash_config1); + } + if (PREDICT_FALSE (lb2->lb_n_buckets > 1)) + { + flow_hash_config2 = lb2->lb_hash_config; + hash_c2 = vnet_buffer (p2)->ip.flow_hash = + ip4_compute_flow_hash (ip2, flow_hash_config2); + } + if (PREDICT_FALSE (lb3->lb_n_buckets > 1)) + { + flow_hash_config3 = lb3->lb_hash_config; + hash_c3 = vnet_buffer (p3)->ip.flow_hash = + ip4_compute_flow_hash (ip3, flow_hash_config3); + } + + ASSERT (lb0->lb_n_buckets > 0); + ASSERT (is_pow2 (lb0->lb_n_buckets)); + ASSERT (lb1->lb_n_buckets > 0); + ASSERT (is_pow2 (lb1->lb_n_buckets)); + ASSERT (lb2->lb_n_buckets > 0); + ASSERT (is_pow2 (lb2->lb_n_buckets)); + ASSERT (lb3->lb_n_buckets > 0); + ASSERT (is_pow2 (lb3->lb_n_buckets)); + + dpo0 = load_balance_get_bucket_i (lb0, + (hash_c0 & + (lb0->lb_n_buckets_minus_1))); + dpo1 = load_balance_get_bucket_i (lb1, + (hash_c1 & + (lb1->lb_n_buckets_minus_1))); + dpo2 = load_balance_get_bucket_i (lb2, + (hash_c2 & + (lb2->lb_n_buckets_minus_1))); + dpo3 = load_balance_get_bucket_i (lb3, + (hash_c3 & + (lb3->lb_n_buckets_minus_1))); + + next0 = dpo0->dpoi_next_node; + vnet_buffer (p0)->ip.adj_index[VLIB_TX] = dpo0->dpoi_index; + next1 = dpo1->dpoi_next_node; + vnet_buffer (p1)->ip.adj_index[VLIB_TX] = dpo1->dpoi_index; + next2 = dpo2->dpoi_next_node; + vnet_buffer (p2)->ip.adj_index[VLIB_TX] = dpo2->dpoi_index; + next3 = dpo3->dpoi_next_node; + vnet_buffer (p3)->ip.adj_index[VLIB_TX] = dpo3->dpoi_index; + + vlib_increment_combined_counter + (cm, cpu_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, + vlib_buffer_length_in_chain (vm, p1) + + sizeof (ethernet_header_t)); + vlib_increment_combined_counter + (cm, cpu_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, + vlib_buffer_length_in_chain (vm, p3) + + sizeof (ethernet_header_t)); + + vlib_validate_buffer_enqueue_x4 (vm, node, next, + to_next, n_left_to_next, + pi0, pi1, pi2, pi3, + next0, next1, next2, next3); + } + + while (n_left_from > 0 && n_left_to_next > 0) + { + vlib_buffer_t *p0; + ip4_header_t *ip0; + __attribute__ ((unused)) tcp_header_t *tcp0; + ip_lookup_next_t next0; + const load_balance_t *lb0; + ip4_fib_mtrie_t *mtrie0; + ip4_fib_mtrie_leaf_t leaf0; + ip4_address_t *dst_addr0; + __attribute__ ((unused)) u32 pi0, fib_index0, is_tcp_udp0, lbi0; + flow_hash_config_t flow_hash_config0; + const dpo_id_t *dpo0; + u32 hash_c0; + + pi0 = from[0]; + to_next[0] = pi0; + + p0 = vlib_get_buffer (vm, pi0); + + ip0 = vlib_buffer_get_current (p0); + + dst_addr0 = &ip0->dst_address; + + fib_index0 = + vec_elt (im->fib_index_by_sw_if_index, + vnet_buffer (p0)->sw_if_index[VLIB_RX]); + fib_index0 = + (vnet_buffer (p0)->sw_if_index[VLIB_TX] == + (u32) ~ 0) ? fib_index0 : vnet_buffer (p0)->sw_if_index[VLIB_TX]; + + if (!lookup_for_responses_to_locally_received_packets) + { + mtrie0 = &ip4_fib_get (fib_index0)->mtrie; + + leaf0 = IP4_FIB_MTRIE_LEAF_ROOT; + + leaf0 = ip4_fib_mtrie_lookup_step (mtrie0, leaf0, dst_addr0, 0); + } + + tcp0 = (void *) (ip0 + 1); + + is_tcp_udp0 = (ip0->protocol == IP_PROTOCOL_TCP + || ip0->protocol == IP_PROTOCOL_UDP); + + if (!lookup_for_responses_to_locally_received_packets) + leaf0 = ip4_fib_mtrie_lookup_step (mtrie0, leaf0, dst_addr0, 1); + + if (!lookup_for_responses_to_locally_received_packets) + leaf0 = ip4_fib_mtrie_lookup_step (mtrie0, leaf0, dst_addr0, 2); + + if (!lookup_for_responses_to_locally_received_packets) + leaf0 = ip4_fib_mtrie_lookup_step (mtrie0, leaf0, dst_addr0, 3); + + if (lookup_for_responses_to_locally_received_packets) + lbi0 = vnet_buffer (p0)->ip.adj_index[VLIB_RX]; + else + { + /* Handle default route. */ + leaf0 = + (leaf0 == + IP4_FIB_MTRIE_LEAF_EMPTY ? mtrie0->default_leaf : leaf0); + lbi0 = ip4_fib_mtrie_leaf_get_adj_index (leaf0); + } + + lb0 = load_balance_get (lbi0); + + /* Use flow hash to compute multipath adjacency. */ + hash_c0 = vnet_buffer (p0)->ip.flow_hash = 0; + if (PREDICT_FALSE (lb0->lb_n_buckets > 1)) + { + flow_hash_config0 = lb0->lb_hash_config; + + hash_c0 = vnet_buffer (p0)->ip.flow_hash = + ip4_compute_flow_hash (ip0, flow_hash_config0); + } + + ASSERT (lb0->lb_n_buckets > 0); + ASSERT (is_pow2 (lb0->lb_n_buckets)); + + dpo0 = load_balance_get_bucket_i (lb0, + (hash_c0 & + (lb0->lb_n_buckets_minus_1))); + + next0 = dpo0->dpoi_next_node; + 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)); + + from += 1; + to_next += 1; + n_left_to_next -= 1; + n_left_from -= 1; + + if (PREDICT_FALSE (next0 != next)) + { + n_left_to_next += 1; + vlib_put_next_frame (vm, node, next, n_left_to_next); + next = next0; + vlib_get_next_frame (vm, node, next, to_next, n_left_to_next); + to_next[0] = pi0; + to_next += 1; + n_left_to_next -= 1; + } + } + + vlib_put_next_frame (vm, node, next, n_left_to_next); + } + + if (node->flags & VLIB_NODE_FLAG_TRACE) + ip4_forward_next_trace (vm, node, frame, VLIB_TX); + + return frame->n_vectors; +} + +/** @brief IPv4 lookup node. + @node ip4-lookup + + This is the main IPv4 lookup dispatch node. + + @param vm vlib_main_t corresponding to the current thread + @param node vlib_node_runtime_t + @param frame vlib_frame_t whose contents should be dispatched + + @par Graph mechanics: buffer metadata, next index usage + + @em Uses: + - vnet_buffer(b)->sw_if_index[VLIB_RX] + - Indicates the @c sw_if_index value of the interface that the + packet was received on. + - vnet_buffer(b)->sw_if_index[VLIB_TX] + - When the value is @c ~0 then the node performs a longest prefix + match (LPM) for the packet destination address in the FIB attached + to the receive interface. + - Otherwise perform LPM for the packet destination address in the + indicated FIB. In this case [VLIB_TX] is a FIB index + value (0, 1, ...) and not a VRF id. + + @em Sets: + - vnet_buffer(b)->ip.adj_index[VLIB_TX] + - The lookup result adjacency index. + + Next Index: + - Dispatches the packet to the node index found in + ip_adjacency_t @c adj->lookup_next_index + (where @c adj is the lookup result adjacency). +*/ +static uword +ip4_lookup (vlib_main_t * vm, + vlib_node_runtime_t * node, vlib_frame_t * frame) +{ + return ip4_lookup_inline (vm, node, frame, + /* lookup_for_responses_to_locally_received_packets */ + 0); + +} + +static u8 *format_ip4_lookup_trace (u8 * s, va_list * args); + +VLIB_REGISTER_NODE (ip4_lookup_node) = +{ +.function = ip4_lookup,.name = "ip4-lookup",.vector_size = + sizeof (u32),.format_trace = format_ip4_lookup_trace,.n_next_nodes = + IP_LOOKUP_N_NEXT,.next_nodes = IP4_LOOKUP_NEXT_NODES,}; + +VLIB_NODE_FUNCTION_MULTIARCH (ip4_lookup_node, ip4_lookup); + +always_inline uword +ip4_load_balance (vlib_main_t * vm, + vlib_node_runtime_t * node, vlib_frame_t * frame) +{ + 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 (); + + from = vlib_frame_vector_args (frame); + n_left_from = frame->n_vectors; + next = node->cached_next_index; + + if (node->flags & VLIB_NODE_FLAG_TRACE) + ip4_forward_next_trace (vm, node, frame, VLIB_TX); + + while (n_left_from > 0) + { + vlib_get_next_frame (vm, node, next, to_next, n_left_to_next); + + + while (n_left_from >= 4 && n_left_to_next >= 2) + { + ip_lookup_next_t next0, next1; + const load_balance_t *lb0, *lb1; + vlib_buffer_t *p0, *p1; + u32 pi0, lbi0, hc0, pi1, lbi1, hc1; + const ip4_header_t *ip0, *ip1; + const dpo_id_t *dpo0, *dpo1; + + /* 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, STORE); + vlib_prefetch_buffer_header (p3, STORE); + + CLIB_PREFETCH (p2->data, sizeof (ip0[0]), STORE); + CLIB_PREFETCH (p3->data, sizeof (ip0[0]), STORE); + } + + pi0 = to_next[0] = from[0]; + pi1 = to_next[1] = from[1]; + + from += 2; + n_left_from -= 2; + to_next += 2; + n_left_to_next -= 2; + + p0 = vlib_get_buffer (vm, pi0); + p1 = vlib_get_buffer (vm, pi1); + + ip0 = vlib_buffer_get_current (p0); + ip1 = vlib_buffer_get_current (p1); + lbi0 = vnet_buffer (p0)->ip.adj_index[VLIB_TX]; + lbi1 = vnet_buffer (p1)->ip.adj_index[VLIB_TX]; + + lb0 = load_balance_get (lbi0); + lb1 = load_balance_get (lbi1); + + /* + * this node is for via FIBs we can re-use the hash value from the + * to node if present. + * We don't want to use the same hash value at each level in the recursion + * graph as that would lead to polarisation + */ + hc0 = vnet_buffer (p0)->ip.flow_hash = 0; + hc1 = vnet_buffer (p1)->ip.flow_hash = 0; + + if (PREDICT_FALSE (lb0->lb_n_buckets > 1)) + { + if (PREDICT_TRUE (vnet_buffer (p0)->ip.flow_hash)) + { + hc0 = vnet_buffer (p0)->ip.flow_hash = + vnet_buffer (p0)->ip.flow_hash >> 1; + } + else + { + hc0 = vnet_buffer (p0)->ip.flow_hash = + ip4_compute_flow_hash (ip0, hc0); + } + } + if (PREDICT_FALSE (lb1->lb_n_buckets > 1)) + { + if (PREDICT_TRUE (vnet_buffer (p1)->ip.flow_hash)) + { + hc1 = vnet_buffer (p1)->ip.flow_hash = + vnet_buffer (p1)->ip.flow_hash >> 1; + } + else + { + hc1 = vnet_buffer (p1)->ip.flow_hash = + ip4_compute_flow_hash (ip1, hc1); + } + } + + dpo0 = + load_balance_get_bucket_i (lb0, + hc0 & (lb0->lb_n_buckets_minus_1)); + dpo1 = + load_balance_get_bucket_i (lb1, + hc1 & (lb1->lb_n_buckets_minus_1)); + + next0 = dpo0->dpoi_next_node; + next1 = dpo1->dpoi_next_node; + + vnet_buffer (p0)->ip.adj_index[VLIB_TX] = dpo0->dpoi_index; + 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)); + vlib_increment_combined_counter + (cm, cpu_index, lbi1, 1, vlib_buffer_length_in_chain (vm, p1)); + + vlib_validate_buffer_enqueue_x2 (vm, node, next, + to_next, n_left_to_next, + pi0, pi1, next0, next1); + } + + while (n_left_from > 0 && n_left_to_next > 0) + { + ip_lookup_next_t next0; + const load_balance_t *lb0; + vlib_buffer_t *p0; + u32 pi0, lbi0, hc0; + const ip4_header_t *ip0; + const dpo_id_t *dpo0; + + pi0 = from[0]; + to_next[0] = pi0; + from += 1; + to_next += 1; + n_left_to_next -= 1; + n_left_from -= 1; + + p0 = vlib_get_buffer (vm, pi0); + + ip0 = vlib_buffer_get_current (p0); + lbi0 = vnet_buffer (p0)->ip.adj_index[VLIB_TX]; + + lb0 = load_balance_get (lbi0); + + hc0 = vnet_buffer (p0)->ip.flow_hash = 0; + if (PREDICT_FALSE (lb0->lb_n_buckets > 1)) + { + if (PREDICT_TRUE (vnet_buffer (p0)->ip.flow_hash)) + { + hc0 = vnet_buffer (p0)->ip.flow_hash = + vnet_buffer (p0)->ip.flow_hash >> 1; + } + else + { + hc0 = vnet_buffer (p0)->ip.flow_hash = + ip4_compute_flow_hash (ip0, hc0); + } + } + + dpo0 = + load_balance_get_bucket_i (lb0, + hc0 & (lb0->lb_n_buckets_minus_1)); + + next0 = dpo0->dpoi_next_node; + 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)); + + vlib_validate_buffer_enqueue_x1 (vm, node, next, + to_next, n_left_to_next, + pi0, next0); + } + + vlib_put_next_frame (vm, node, next, n_left_to_next); + } + + return frame->n_vectors; +} + +VLIB_REGISTER_NODE (ip4_load_balance_node) = +{ +.function = ip4_load_balance,.name = "ip4-load-balance",.vector_size = + sizeof (u32),.sibling_of = "ip4-lookup",.format_trace = + format_ip4_lookup_trace,}; + +VLIB_NODE_FUNCTION_MULTIARCH (ip4_load_balance_node, ip4_load_balance); + +/* get first interface address */ +ip4_address_t * +ip4_interface_first_address (ip4_main_t * im, u32 sw_if_index, + ip_interface_address_t ** result_ia) +{ + ip_lookup_main_t *lm = &im->lookup_main; + ip_interface_address_t *ia = 0; + ip4_address_t *result = 0; + + foreach_ip_interface_address (lm, ia, sw_if_index, + 1 /* honor unnumbered */ , + ( + { + ip4_address_t * a = + ip_interface_address_get_address (lm, ia); + result = a; + break; + } + )); + if (result_ia) + *result_ia = result ? ia : 0; + return result; +} + +static void +ip4_add_interface_routes (u32 sw_if_index, + ip4_main_t * im, u32 fib_index, + ip_interface_address_t * a) +{ + ip_lookup_main_t *lm = &im->lookup_main; + ip4_address_t *address = ip_interface_address_get_address (lm, a); + fib_prefix_t pfx = { + .fp_len = a->address_length, + .fp_proto = FIB_PROTOCOL_IP4, + .fp_addr.ip4 = *address, + }; + + a->neighbor_probe_adj_index = ~0; + + if (pfx.fp_len < 32) + { + fib_node_index_t fei; + + fei = fib_table_entry_update_one_path (fib_index, &pfx, FIB_SOURCE_INTERFACE, (FIB_ENTRY_FLAG_CONNECTED | FIB_ENTRY_FLAG_ATTACHED), FIB_PROTOCOL_IP4, NULL, /* No next-hop address */ + sw_if_index, ~0, // invalid FIB index + 1, NULL, // no out-label stack + FIB_ROUTE_PATH_FLAG_NONE); + a->neighbor_probe_adj_index = fib_entry_get_adj (fei); + } + + pfx.fp_len = 32; + + if (sw_if_index < vec_len (lm->classify_table_index_by_sw_if_index)) + { + u32 classify_table_index = + lm->classify_table_index_by_sw_if_index[sw_if_index]; + if (classify_table_index != (u32) ~ 0) + { + dpo_id_t dpo = DPO_INVALID; + + dpo_set (&dpo, + DPO_CLASSIFY, + DPO_PROTO_IP4, + classify_dpo_create (DPO_PROTO_IP4, classify_table_index)); + + fib_table_entry_special_dpo_add (fib_index, + &pfx, + FIB_SOURCE_CLASSIFY, + FIB_ENTRY_FLAG_NONE, &dpo); + dpo_reset (&dpo); + } + } + + fib_table_entry_update_one_path (fib_index, &pfx, FIB_SOURCE_INTERFACE, (FIB_ENTRY_FLAG_CONNECTED | FIB_ENTRY_FLAG_LOCAL), FIB_PROTOCOL_IP4, &pfx.fp_addr, sw_if_index, ~0, // invalid FIB index + 1, NULL, // no out-label stack + FIB_ROUTE_PATH_FLAG_NONE); +} + +static void +ip4_del_interface_routes (ip4_main_t * im, + u32 fib_index, + ip4_address_t * address, u32 address_length) +{ + fib_prefix_t pfx = { + .fp_len = address_length, + .fp_proto = FIB_PROTOCOL_IP4, + .fp_addr.ip4 = *address, + }; + + if (pfx.fp_len < 32) + { + fib_table_entry_delete (fib_index, &pfx, FIB_SOURCE_INTERFACE); + } + + pfx.fp_len = 32; + fib_table_entry_delete (fib_index, &pfx, FIB_SOURCE_INTERFACE); +} + +void +ip4_sw_interface_enable_disable (u32 sw_if_index, u32 is_enable) +{ + ip4_main_t *im = &ip4_main; + + vec_validate_init_empty (im->ip_enabled_by_sw_if_index, sw_if_index, 0); + + /* + * enable/disable only on the 1<->0 transition + */ + if (is_enable) + { + if (1 != ++im->ip_enabled_by_sw_if_index[sw_if_index]) + return; + } + else + { + ASSERT (im->ip_enabled_by_sw_if_index[sw_if_index] > 0); + if (0 != --im->ip_enabled_by_sw_if_index[sw_if_index]) + return; + } + vnet_feature_enable_disable ("ip4-unicast", "ip4-drop", sw_if_index, + !is_enable, 0, 0); + + vnet_feature_enable_disable ("ip4-multicast", "ip4-drop", sw_if_index, + !is_enable, 0, 0); + +} + +static clib_error_t * +ip4_add_del_interface_address_internal (vlib_main_t * vm, + u32 sw_if_index, + ip4_address_t * address, + u32 address_length, u32 is_del) +{ + vnet_main_t *vnm = vnet_get_main (); + ip4_main_t *im = &ip4_main; + ip_lookup_main_t *lm = &im->lookup_main; + clib_error_t *error = 0; + u32 if_address_index, elts_before; + ip4_address_fib_t ip4_af, *addr_fib = 0; + + vec_validate (im->fib_index_by_sw_if_index, sw_if_index); + ip4_addr_fib_init (&ip4_af, address, + vec_elt (im->fib_index_by_sw_if_index, sw_if_index)); + vec_add1 (addr_fib, ip4_af); + + /* FIXME-LATER + * there is no support for adj-fib handling in the presence of overlapping + * subnets on interfaces. Easy fix - disallow overlapping subnets, like + * most routers do. + */ + if (!is_del) + { + /* When adding an address check that it does not conflict + with an existing address. */ + ip_interface_address_t *ia; + foreach_ip_interface_address (&im->lookup_main, ia, sw_if_index, + 0 /* honor unnumbered */ , + ( + { + ip4_address_t * x = + ip_interface_address_get_address + (&im->lookup_main, ia); + if (ip4_destination_matches_route + (im, address, x, ia->address_length) + || + ip4_destination_matches_route (im, + x, + address, + address_length)) + return + clib_error_create + ("failed to add %U which conflicts with %U for interface %U", + format_ip4_address_and_length, address, + address_length, + format_ip4_address_and_length, x, + ia->address_length, + format_vnet_sw_if_index_name, vnm, + sw_if_index);} + )); + } + + elts_before = pool_elts (lm->if_address_pool); + + error = ip_interface_address_add_del + (lm, sw_if_index, addr_fib, address_length, is_del, &if_address_index); + if (error) + goto done; + + ip4_sw_interface_enable_disable (sw_if_index, !is_del); + + if (is_del) + ip4_del_interface_routes (im, ip4_af.fib_index, address, address_length); + else + ip4_add_interface_routes (sw_if_index, + im, ip4_af.fib_index, + pool_elt_at_index + (lm->if_address_pool, if_address_index)); + + /* If pool did not grow/shrink: add duplicate address. */ + if (elts_before != pool_elts (lm->if_address_pool)) + { + ip4_add_del_interface_address_callback_t *cb; + vec_foreach (cb, im->add_del_interface_address_callbacks) + cb->function (im, cb->function_opaque, sw_if_index, + address, address_length, if_address_index, is_del); + } + +done: + vec_free (addr_fib); + return error; +} + +clib_error_t * +ip4_add_del_interface_address (vlib_main_t * vm, u32 sw_if_index, + ip4_address_t * address, u32 address_length, + u32 is_del) +{ + return ip4_add_del_interface_address_internal + (vm, sw_if_index, address, address_length, is_del); +} + +/* Built-in ip4 unicast rx feature path definition */ +/* *INDENT-OFF* */ +VNET_FEATURE_ARC_INIT (ip4_unicast, static) = +{ + .arc_name = "ip4-unicast", + .start_nodes = VNET_FEATURES ("ip4-input", "ip4-input-no-checksum"), + .end_node = "ip4-lookup", + .arc_index_ptr = &ip4_main.lookup_main.ucast_feature_arc_index, +}; + +VNET_FEATURE_INIT (ip4_flow_classify, static) = +{ + .arc_name = "ip4-unicast", + .node_name = "ip4-flow-classify", + .runs_before = VNET_FEATURES ("ip4-inacl"), +}; + +VNET_FEATURE_INIT (ip4_inacl, static) = +{ + .arc_name = "ip4-unicast", + .node_name = "ip4-inacl", + .runs_before = VNET_FEATURES ("ip4-source-check-via-rx"), +}; + +VNET_FEATURE_INIT (ip4_source_check_1, static) = +{ + .arc_name = "ip4-unicast", + .node_name = "ip4-source-check-via-rx", + .runs_before = VNET_FEATURES ("ip4-source-check-via-any"), +}; + +VNET_FEATURE_INIT (ip4_source_check_2, static) = +{ + .arc_name = "ip4-unicast", + .node_name = "ip4-source-check-via-any", + .runs_before = VNET_FEATURES ("ip4-policer-classify"), +}; + +VNET_FEATURE_INIT (ip4_source_and_port_range_check_rx, static) = +{ + .arc_name = "ip4-unicast", + .node_name = "ip4-source-and-port-range-check-rx", + .runs_before = VNET_FEATURES ("ip4-policer-classify"), +}; + +VNET_FEATURE_INIT (ip4_policer_classify, static) = +{ + .arc_name = "ip4-unicast", + .node_name = "ip4-policer-classify", + .runs_before = VNET_FEATURES ("ipsec-input-ip4"), +}; + +VNET_FEATURE_INIT (ip4_ipsec, static) = +{ + .arc_name = "ip4-unicast", + .node_name = "ipsec-input-ip4", + .runs_before = VNET_FEATURES ("vpath-input-ip4"), +}; + +VNET_FEATURE_INIT (ip4_vpath, static) = +{ + .arc_name = "ip4-unicast", + .node_name = "vpath-input-ip4", + .runs_before = VNET_FEATURES ("ip4-vxlan-bypass"), +}; + +VNET_FEATURE_INIT (ip4_vxlan_bypass, static) = +{ + .arc_name = "ip4-unicast", + .node_name = "ip4-vxlan-bypass", + .runs_before = VNET_FEATURES ("ip4-lookup"), +}; + +VNET_FEATURE_INIT (ip4_lookup, static) = +{ + .arc_name = "ip4-unicast", + .node_name = "ip4-lookup", + .runs_before = VNET_FEATURES ("ip4-drop"), +}; + +VNET_FEATURE_INIT (ip4_drop, static) = +{ + .arc_name = "ip4-unicast", + .node_name = "ip4-drop", + .runs_before = 0, /* not before any other features */ +}; + + +/* Built-in ip4 multicast rx feature path definition */ +VNET_FEATURE_ARC_INIT (ip4_multicast, static) = +{ + .arc_name = "ip4-multicast", + .start_nodes = VNET_FEATURES ("ip4-input", "ip4-input-no-checksum"), + .end_node = "ip4-lookup-multicast", + .arc_index_ptr = &ip4_main.lookup_main.mcast_feature_arc_index, +}; + +VNET_FEATURE_INIT (ip4_vpath_mc, static) = +{ + .arc_name = "ip4-multicast", + .node_name = "vpath-input-ip4", + .runs_before = VNET_FEATURES ("ip4-lookup-multicast"), +}; + +VNET_FEATURE_INIT (ip4_lookup_mc, static) = +{ + .arc_name = "ip4-multicast", + .node_name = "ip4-lookup-multicast", + .runs_before = VNET_FEATURES ("ip4-drop"), +}; + +VNET_FEATURE_INIT (ip4_mc_drop, static) = +{ + .arc_name = "ip4-multicast", + .node_name = "ip4-drop", + .runs_before = 0, /* last feature */ +}; + +/* Source and port-range check ip4 tx feature path definition */ +VNET_FEATURE_ARC_INIT (ip4_output, static) = +{ + .arc_name = "ip4-output", + .start_nodes = VNET_FEATURES ("ip4-rewrite", "ip4-midchain"), + .end_node = "interface-output", + .arc_index_ptr = &ip4_main.lookup_main.output_feature_arc_index, +}; + +VNET_FEATURE_INIT (ip4_source_and_port_range_check_tx, static) = +{ + .arc_name = "ip4-output", + .node_name = "ip4-source-and-port-range-check-tx", + .runs_before = VNET_FEATURES ("ipsec-output-ip4"), +}; + +VNET_FEATURE_INIT (ip4_ipsec_output, static) = +{ + .arc_name = "ip4-output", + .node_name = "ipsec-output-ip4", + .runs_before = VNET_FEATURES ("interface-output"), +}; + +/* Built-in ip4 tx feature path definition */ +VNET_FEATURE_INIT (ip4_interface_output, static) = +{ + .arc_name = "ip4-output", + .node_name = "interface-output", + .runs_before = 0, /* not before any other features */ +}; +/* *INDENT-ON* */ + +static clib_error_t * +ip4_sw_interface_add_del (vnet_main_t * vnm, u32 sw_if_index, u32 is_add) +{ + ip4_main_t *im = &ip4_main; + + /* Fill in lookup tables with default table (0). */ + vec_validate (im->fib_index_by_sw_if_index, sw_if_index); + + vnet_feature_enable_disable ("ip4-unicast", "ip4-drop", sw_if_index, + is_add, 0, 0); + + vnet_feature_enable_disable ("ip4-multicast", "ip4-drop", sw_if_index, + is_add, 0, 0); + + return /* no error */ 0; +} + +VNET_SW_INTERFACE_ADD_DEL_FUNCTION (ip4_sw_interface_add_del); + +/* Global IP4 main. */ +ip4_main_t ip4_main; + +clib_error_t * +ip4_lookup_init (vlib_main_t * vm) +{ + ip4_main_t *im = &ip4_main; + clib_error_t *error; + uword i; + + if ((error = vlib_call_init_function (vm, vnet_feature_init))) + return error; + + for (i = 0; i < ARRAY_LEN (im->fib_masks); i++) + { + u32 m; + + if (i < 32) + m = pow2_mask (i) << (32 - i); + else + m = ~0; + im->fib_masks[i] = clib_host_to_net_u32 (m); + } + + 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); + + { + pg_node_t *pn; + pn = pg_get_node (ip4_lookup_node.index); + pn->unformat_edit = unformat_pg_ip4_header; + } + + { + ethernet_arp_header_t h; + + memset (&h, 0, sizeof (h)); + + /* Set target ethernet address to all zeros. */ + memset (h.ip4_over_ethernet[1].ethernet, 0, + sizeof (h.ip4_over_ethernet[1].ethernet)); + +#define _16(f,v) h.f = clib_host_to_net_u16 (v); +#define _8(f,v) h.f = v; + _16 (l2_type, ETHERNET_ARP_HARDWARE_TYPE_ethernet); + _16 (l3_type, ETHERNET_TYPE_IP4); + _8 (n_l2_address_bytes, 6); + _8 (n_l3_address_bytes, 4); + _16 (opcode, ETHERNET_ARP_OPCODE_request); +#undef _16 +#undef _8 + + vlib_packet_template_init (vm, &im->ip4_arp_request_packet_template, + /* data */ &h, + sizeof (h), + /* alloc chunk size */ 8, + "ip4 arp"); + } + + return error; +} + +VLIB_INIT_FUNCTION (ip4_lookup_init); + +typedef struct +{ + /* Adjacency taken. */ + u32 dpo_index; + u32 flow_hash; + u32 fib_index; + + /* Packet data, possibly *after* rewrite. */ + u8 packet_data[64 - 1 * sizeof (u32)]; +} +ip4_forward_next_trace_t; + +u8 * +format_ip4_forward_next_trace (u8 * s, va_list * args) +{ + CLIB_UNUSED (vlib_main_t * vm) = va_arg (*args, vlib_main_t *); + CLIB_UNUSED (vlib_node_t * node) = va_arg (*args, vlib_node_t *); + ip4_forward_next_trace_t *t = va_arg (*args, ip4_forward_next_trace_t *); + uword indent = format_get_indent (s); + s = format (s, "%U%U", + format_white_space, indent, + format_ip4_header, t->packet_data, sizeof (t->packet_data)); + return s; +} + +static u8 * +format_ip4_lookup_trace (u8 * s, va_list * args) +{ + CLIB_UNUSED (vlib_main_t * vm) = va_arg (*args, vlib_main_t *); + CLIB_UNUSED (vlib_node_t * node) = va_arg (*args, vlib_node_t *); + ip4_forward_next_trace_t *t = va_arg (*args, ip4_forward_next_trace_t *); + uword indent = format_get_indent (s); + + s = format (s, "fib %d dpo-idx %d flow hash: 0x%08x", + t->fib_index, t->dpo_index, t->flow_hash); + s = format (s, "\n%U%U", + format_white_space, indent, + format_ip4_header, t->packet_data, sizeof (t->packet_data)); + return s; +} + +static u8 * +format_ip4_rewrite_trace (u8 * s, va_list * args) +{ + CLIB_UNUSED (vlib_main_t * vm) = va_arg (*args, vlib_main_t *); + CLIB_UNUSED (vlib_node_t * node) = va_arg (*args, vlib_node_t *); + ip4_forward_next_trace_t *t = va_arg (*args, ip4_forward_next_trace_t *); + vnet_main_t *vnm = vnet_get_main (); + uword indent = format_get_indent (s); + + s = format (s, "tx_sw_if_index %d dpo-idx %d : %U flow hash: 0x%08x", + t->fib_index, t->dpo_index, format_ip_adjacency, + t->dpo_index, FORMAT_IP_ADJACENCY_NONE, t->flow_hash); + s = format (s, "\n%U%U", + format_white_space, indent, + format_ip_adjacency_packet_data, + vnm, t->dpo_index, t->packet_data, sizeof (t->packet_data)); + return s; +} + +/* Common trace function for all ip4-forward next nodes. */ +void +ip4_forward_next_trace (vlib_main_t * vm, + vlib_node_runtime_t * node, + vlib_frame_t * frame, vlib_rx_or_tx_t which_adj_index) +{ + u32 *from, n_left; + ip4_main_t *im = &ip4_main; + + n_left = frame->n_vectors; + from = vlib_frame_vector_args (frame); + + while (n_left >= 4) + { + u32 bi0, bi1; + vlib_buffer_t *b0, *b1; + ip4_forward_next_trace_t *t0, *t1; + + /* Prefetch next iteration. */ + vlib_prefetch_buffer_with_index (vm, from[2], LOAD); + vlib_prefetch_buffer_with_index (vm, from[3], LOAD); + + bi0 = from[0]; + bi1 = from[1]; + + b0 = vlib_get_buffer (vm, bi0); + b1 = vlib_get_buffer (vm, bi1); + + if (b0->flags & VLIB_BUFFER_IS_TRACED) + { + t0 = vlib_add_trace (vm, node, b0, sizeof (t0[0])); + t0->dpo_index = vnet_buffer (b0)->ip.adj_index[which_adj_index]; + t0->flow_hash = vnet_buffer (b0)->ip.flow_hash; + t0->fib_index = + (vnet_buffer (b0)->sw_if_index[VLIB_TX] != + (u32) ~ 0) ? vnet_buffer (b0)->sw_if_index[VLIB_TX] : + vec_elt (im->fib_index_by_sw_if_index, + vnet_buffer (b0)->sw_if_index[VLIB_RX]); + + clib_memcpy (t0->packet_data, + vlib_buffer_get_current (b0), + sizeof (t0->packet_data)); + } + if (b1->flags & VLIB_BUFFER_IS_TRACED) + { + t1 = vlib_add_trace (vm, node, b1, sizeof (t1[0])); + t1->dpo_index = vnet_buffer (b1)->ip.adj_index[which_adj_index]; + t1->flow_hash = vnet_buffer (b1)->ip.flow_hash; + t1->fib_index = + (vnet_buffer (b1)->sw_if_index[VLIB_TX] != + (u32) ~ 0) ? vnet_buffer (b1)->sw_if_index[VLIB_TX] : + vec_elt (im->fib_index_by_sw_if_index, + vnet_buffer (b1)->sw_if_index[VLIB_RX]); + clib_memcpy (t1->packet_data, vlib_buffer_get_current (b1), + sizeof (t1->packet_data)); + } + from += 2; + n_left -= 2; + } + + while (n_left >= 1) + { + u32 bi0; + vlib_buffer_t *b0; + ip4_forward_next_trace_t *t0; + + bi0 = from[0]; + + b0 = vlib_get_buffer (vm, bi0); + + if (b0->flags & VLIB_BUFFER_IS_TRACED) + { + t0 = vlib_add_trace (vm, node, b0, sizeof (t0[0])); + t0->dpo_index = vnet_buffer (b0)->ip.adj_index[which_adj_index]; + t0->flow_hash = vnet_buffer (b0)->ip.flow_hash; + t0->fib_index = + (vnet_buffer (b0)->sw_if_index[VLIB_TX] != + (u32) ~ 0) ? vnet_buffer (b0)->sw_if_index[VLIB_TX] : + vec_elt (im->fib_index_by_sw_if_index, + vnet_buffer (b0)->sw_if_index[VLIB_RX]); + clib_memcpy (t0->packet_data, vlib_buffer_get_current (b0), + sizeof (t0->packet_data)); + } + from += 1; + n_left -= 1; + } +} + +static uword +ip4_drop_or_punt (vlib_main_t * vm, + vlib_node_runtime_t * node, + vlib_frame_t * frame, ip4_error_t error_code) +{ + u32 *buffers = vlib_frame_vector_args (frame); + uword n_packets = frame->n_vectors; + + vlib_error_drop_buffers (vm, node, buffers, + /* stride */ 1, + n_packets, + /* next */ 0, + ip4_input_node.index, error_code); + + if (node->flags & VLIB_NODE_FLAG_TRACE) + ip4_forward_next_trace (vm, node, frame, VLIB_TX); + + return n_packets; +} + +static uword +ip4_drop (vlib_main_t * vm, vlib_node_runtime_t * node, vlib_frame_t * frame) +{ + return ip4_drop_or_punt (vm, node, frame, IP4_ERROR_ADJACENCY_DROP); +} + +static uword +ip4_punt (vlib_main_t * vm, vlib_node_runtime_t * node, vlib_frame_t * frame) +{ + return ip4_drop_or_punt (vm, node, frame, IP4_ERROR_ADJACENCY_PUNT); +} + +VLIB_REGISTER_NODE (ip4_drop_node, static) = +{ + .function = ip4_drop,.name = "ip4-drop",.vector_size = + sizeof (u32),.format_trace = format_ip4_forward_next_trace,.n_next_nodes = + 1,.next_nodes = + { + [0] = "error-drop",} +,}; + +VLIB_NODE_FUNCTION_MULTIARCH (ip4_drop_node, ip4_drop); + +VLIB_REGISTER_NODE (ip4_punt_node, static) = +{ + .function = ip4_punt,.name = "ip4-punt",.vector_size = + sizeof (u32),.format_trace = format_ip4_forward_next_trace,.n_next_nodes = + 1,.next_nodes = + { + [0] = "error-punt",} +,}; + +VLIB_NODE_FUNCTION_MULTIARCH (ip4_punt_node, ip4_punt); + +/* Compute TCP/UDP/ICMP4 checksum in software. */ +u16 +ip4_tcp_udp_compute_checksum (vlib_main_t * vm, vlib_buffer_t * p0, + ip4_header_t * ip0) +{ + ip_csum_t sum0; + u32 ip_header_length, payload_length_host_byte_order; + u32 n_this_buffer, n_bytes_left; + u16 sum16; + void *data_this_buffer; + + /* Initialize checksum with ip header. */ + ip_header_length = ip4_header_bytes (ip0); + payload_length_host_byte_order = + clib_net_to_host_u16 (ip0->length) - ip_header_length; + sum0 = + clib_host_to_net_u32 (payload_length_host_byte_order + + (ip0->protocol << 16)); + + if (BITS (uword) == 32) + { + sum0 = + ip_csum_with_carry (sum0, + clib_mem_unaligned (&ip0->src_address, u32)); + sum0 = + ip_csum_with_carry (sum0, + clib_mem_unaligned (&ip0->dst_address, u32)); + } + else + sum0 = + ip_csum_with_carry (sum0, clib_mem_unaligned (&ip0->src_address, u64)); + + n_bytes_left = n_this_buffer = payload_length_host_byte_order; + data_this_buffer = (void *) ip0 + ip_header_length; + if (n_this_buffer + ip_header_length > p0->current_length) + n_this_buffer = + p0->current_length > + ip_header_length ? p0->current_length - ip_header_length : 0; + while (1) + { + sum0 = ip_incremental_checksum (sum0, data_this_buffer, n_this_buffer); + n_bytes_left -= n_this_buffer; + if (n_bytes_left == 0) + break; + + ASSERT (p0->flags & VLIB_BUFFER_NEXT_PRESENT); + p0 = vlib_get_buffer (vm, p0->next_buffer); + data_this_buffer = vlib_buffer_get_current (p0); + n_this_buffer = p0->current_length; + } + + sum16 = ~ip_csum_fold (sum0); + + return sum16; +} + +u32 +ip4_tcp_udp_validate_checksum (vlib_main_t * vm, vlib_buffer_t * p0) +{ + ip4_header_t *ip0 = vlib_buffer_get_current (p0); + udp_header_t *udp0; + u16 sum16; + + ASSERT (ip0->protocol == IP_PROTOCOL_TCP + || ip0->protocol == IP_PROTOCOL_UDP); + + udp0 = (void *) (ip0 + 1); + if (ip0->protocol == IP_PROTOCOL_UDP && udp0->checksum == 0) + { + p0->flags |= (IP_BUFFER_L4_CHECKSUM_COMPUTED + | IP_BUFFER_L4_CHECKSUM_CORRECT); + return p0->flags; + } + + sum16 = ip4_tcp_udp_compute_checksum (vm, p0, ip0); + + p0->flags |= (IP_BUFFER_L4_CHECKSUM_COMPUTED + | ((sum16 == 0) << LOG2_IP_BUFFER_L4_CHECKSUM_CORRECT)); + + return p0->flags; +} + +static uword +ip4_local (vlib_main_t * vm, vlib_node_runtime_t * node, vlib_frame_t * frame) +{ + ip4_main_t *im = &ip4_main; + ip_lookup_main_t *lm = &im->lookup_main; + ip_local_next_t next_index; + 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); + + from = vlib_frame_vector_args (frame); + n_left_from = frame->n_vectors; + next_index = node->cached_next_index; + + if (node->flags & VLIB_NODE_FLAG_TRACE) + ip4_forward_next_trace (vm, node, frame, VLIB_TX); + + while (n_left_from > 0) + { + vlib_get_next_frame (vm, node, next_index, to_next, n_left_to_next); + + while (n_left_from >= 4 && n_left_to_next >= 2) + { + vlib_buffer_t *p0, *p1; + ip4_header_t *ip0, *ip1; + udp_header_t *udp0, *udp1; + ip4_fib_mtrie_t *mtrie0, *mtrie1; + ip4_fib_mtrie_leaf_t leaf0, leaf1; + const dpo_id_t *dpo0, *dpo1; + const load_balance_t *lb0, *lb1; + u32 pi0, ip_len0, udp_len0, flags0, next0, fib_index0, lbi0; + u32 pi1, ip_len1, udp_len1, flags1, next1, fib_index1, lbi1; + 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; + + pi0 = to_next[0] = from[0]; + pi1 = to_next[1] = from[1]; + from += 2; + n_left_from -= 2; + to_next += 2; + n_left_to_next -= 2; + + p0 = vlib_get_buffer (vm, pi0); + p1 = vlib_get_buffer (vm, pi1); + + ip0 = vlib_buffer_get_current (p0); + ip1 = vlib_buffer_get_current (p1); + + 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]); + 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 = (vnet_buffer (p1)->sw_if_index[VLIB_TX] == (u32) ~ 0) ? + fib_index1 : vnet_buffer (p1)->sw_if_index[VLIB_TX]; + + mtrie0 = &ip4_fib_get (fib_index0)->mtrie; + mtrie1 = &ip4_fib_get (fib_index1)->mtrie; + + leaf0 = leaf1 = IP4_FIB_MTRIE_LEAF_ROOT; + + leaf0 = + ip4_fib_mtrie_lookup_step (mtrie0, leaf0, &ip0->src_address, 0); + leaf1 = + ip4_fib_mtrie_lookup_step (mtrie1, leaf1, &ip1->src_address, 0); + + /* Treat IP frag packets as "experimental" protocol for now + until support of IP frag reassembly is implemented */ + proto0 = ip4_is_fragment (ip0) ? 0xfe : ip0->protocol; + proto1 = ip4_is_fragment (ip1) ? 0xfe : ip1->protocol; + is_udp0 = proto0 == IP_PROTOCOL_UDP; + is_udp1 = proto1 == IP_PROTOCOL_UDP; + is_tcp_udp0 = is_udp0 || proto0 == IP_PROTOCOL_TCP; + is_tcp_udp1 = is_udp1 || proto1 == IP_PROTOCOL_TCP; + + flags0 = p0->flags; + flags1 = p1->flags; + + good_tcp_udp0 = (flags0 & IP_BUFFER_L4_CHECKSUM_CORRECT) != 0; + good_tcp_udp1 = (flags1 & IP_BUFFER_L4_CHECKSUM_CORRECT) != 0; + + udp0 = ip4_next_header (ip0); + udp1 = ip4_next_header (ip1); + + /* Don't verify UDP checksum for packets with explicit zero checksum. */ + good_tcp_udp0 |= is_udp0 && udp0->checksum == 0; + good_tcp_udp1 |= is_udp1 && udp1->checksum == 0; + + leaf0 = + ip4_fib_mtrie_lookup_step (mtrie0, leaf0, &ip0->src_address, 1); + leaf1 = + ip4_fib_mtrie_lookup_step (mtrie1, leaf1, &ip1->src_address, 1); + + /* Verify UDP length. */ + ip_len0 = clib_net_to_host_u16 (ip0->length); + ip_len1 = clib_net_to_host_u16 (ip1->length); + udp_len0 = clib_net_to_host_u16 (udp0->length); + udp_len1 = clib_net_to_host_u16 (udp1->length); + + len_diff0 = ip_len0 - udp_len0; + len_diff1 = ip_len1 - udp_len1; + + len_diff0 = is_udp0 ? len_diff0 : 0; + len_diff1 = is_udp1 ? len_diff1 : 0; + + if (PREDICT_FALSE (!(is_tcp_udp0 & is_tcp_udp1 + & good_tcp_udp0 & good_tcp_udp1))) + { + if (is_tcp_udp0) + { + if (is_tcp_udp0 + && !(flags0 & IP_BUFFER_L4_CHECKSUM_COMPUTED)) + flags0 = ip4_tcp_udp_validate_checksum (vm, p0); + good_tcp_udp0 = + (flags0 & IP_BUFFER_L4_CHECKSUM_CORRECT) != 0; + good_tcp_udp0 |= is_udp0 && udp0->checksum == 0; + } + if (is_tcp_udp1) + { + if (is_tcp_udp1 + && !(flags1 & IP_BUFFER_L4_CHECKSUM_COMPUTED)) + flags1 = ip4_tcp_udp_validate_checksum (vm, p1); + good_tcp_udp1 = + (flags1 & IP_BUFFER_L4_CHECKSUM_CORRECT) != 0; + good_tcp_udp1 |= is_udp1 && udp1->checksum == 0; + } + } + + good_tcp_udp0 &= len_diff0 >= 0; + good_tcp_udp1 &= len_diff1 >= 0; + + leaf0 = + ip4_fib_mtrie_lookup_step (mtrie0, leaf0, &ip0->src_address, 2); + leaf1 = + ip4_fib_mtrie_lookup_step (mtrie1, leaf1, &ip1->src_address, 2); + + error0 = error1 = IP4_ERROR_UNKNOWN_PROTOCOL; + + error0 = len_diff0 < 0 ? IP4_ERROR_UDP_LENGTH : error0; + error1 = len_diff1 < 0 ? IP4_ERROR_UDP_LENGTH : error1; + + ASSERT (IP4_ERROR_TCP_CHECKSUM + 1 == IP4_ERROR_UDP_CHECKSUM); + error0 = (is_tcp_udp0 && !good_tcp_udp0 + ? IP4_ERROR_TCP_CHECKSUM + is_udp0 : error0); + error1 = (is_tcp_udp1 && !good_tcp_udp1 + ? IP4_ERROR_TCP_CHECKSUM + is_udp1 : error1); + + leaf0 = + ip4_fib_mtrie_lookup_step (mtrie0, leaf0, &ip0->src_address, 3); + leaf1 = + ip4_fib_mtrie_lookup_step (mtrie1, leaf1, &ip1->src_address, 3); + leaf0 = + (leaf0 == + IP4_FIB_MTRIE_LEAF_EMPTY ? mtrie0->default_leaf : leaf0); + leaf1 = + (leaf1 == + IP4_FIB_MTRIE_LEAF_EMPTY ? mtrie1->default_leaf : leaf1); + + vnet_buffer (p0)->ip.adj_index[VLIB_RX] = lbi0 = + ip4_fib_mtrie_leaf_get_adj_index (leaf0); + vnet_buffer (p0)->ip.adj_index[VLIB_TX] = lbi0; + + vnet_buffer (p1)->ip.adj_index[VLIB_RX] = lbi1 = + ip4_fib_mtrie_leaf_get_adj_index (leaf1); + vnet_buffer (p1)->ip.adj_index[VLIB_TX] = lbi1; + + lb0 = load_balance_get (lbi0); + lb1 = load_balance_get (lbi1); + dpo0 = load_balance_get_bucket_i (lb0, 0); + dpo1 = load_balance_get_bucket_i (lb1, 0); + + /* + * Must have a route to source otherwise we drop the packet. + * ip4 broadcasts are accepted, e.g. to make dhcp client work + * + * The checks are: + * - the source is a recieve => it's from us => bogus, do this + * first since it sets a different error code. + * - uRPF check for any route to source - accept if passes. + * - allow packets destined to the broadcast address from unknown sources + */ + error0 = ((error0 == IP4_ERROR_UNKNOWN_PROTOCOL && + dpo0->dpoi_type == DPO_RECEIVE) ? + IP4_ERROR_SPOOFED_LOCAL_PACKETS : error0); + error0 = ((error0 == IP4_ERROR_UNKNOWN_PROTOCOL && + !fib_urpf_check_size (lb0->lb_urpf) && + ip0->dst_address.as_u32 != 0xFFFFFFFF) + ? IP4_ERROR_SRC_LOOKUP_MISS : error0); + error1 = ((error1 == IP4_ERROR_UNKNOWN_PROTOCOL && + dpo1->dpoi_type == DPO_RECEIVE) ? + IP4_ERROR_SPOOFED_LOCAL_PACKETS : error1); + error1 = ((error1 == IP4_ERROR_UNKNOWN_PROTOCOL && + !fib_urpf_check_size (lb1->lb_urpf) && + ip1->dst_address.as_u32 != 0xFFFFFFFF) + ? IP4_ERROR_SRC_LOOKUP_MISS : error1); + + next0 = lm->local_next_by_ip_protocol[proto0]; + next1 = lm->local_next_by_ip_protocol[proto1]; + + next0 = + error0 != IP4_ERROR_UNKNOWN_PROTOCOL ? IP_LOCAL_NEXT_DROP : next0; + next1 = + error1 != IP4_ERROR_UNKNOWN_PROTOCOL ? IP_LOCAL_NEXT_DROP : next1; + + 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)) + { + 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; + } + } + } + + while (n_left_from > 0 && n_left_to_next > 0) + { + vlib_buffer_t *p0; + ip4_header_t *ip0; + udp_header_t *udp0; + ip4_fib_mtrie_t *mtrie0; + ip4_fib_mtrie_leaf_t leaf0; + u32 pi0, next0, ip_len0, udp_len0, flags0, fib_index0, lbi0; + i32 len_diff0; + u8 error0, is_udp0, is_tcp_udp0, good_tcp_udp0, proto0; + load_balance_t *lb0; + const dpo_id_t *dpo0; + + pi0 = to_next[0] = from[0]; + from += 1; + n_left_from -= 1; + to_next += 1; + n_left_to_next -= 1; + + p0 = vlib_get_buffer (vm, pi0); + + 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]); + fib_index0 = (vnet_buffer (p0)->sw_if_index[VLIB_TX] == (u32) ~ 0) ? + fib_index0 : vnet_buffer (p0)->sw_if_index[VLIB_TX]; + + mtrie0 = &ip4_fib_get (fib_index0)->mtrie; + + leaf0 = IP4_FIB_MTRIE_LEAF_ROOT; + + leaf0 = + ip4_fib_mtrie_lookup_step (mtrie0, leaf0, &ip0->src_address, 0); + + /* 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; + is_udp0 = proto0 == IP_PROTOCOL_UDP; + is_tcp_udp0 = is_udp0 || proto0 == IP_PROTOCOL_TCP; + + flags0 = p0->flags; + + good_tcp_udp0 = (flags0 & IP_BUFFER_L4_CHECKSUM_CORRECT) != 0; + + udp0 = ip4_next_header (ip0); + + /* Don't verify UDP checksum for packets with explicit zero checksum. */ + good_tcp_udp0 |= is_udp0 && udp0->checksum == 0; + + leaf0 = + ip4_fib_mtrie_lookup_step (mtrie0, leaf0, &ip0->src_address, 1); + + /* Verify UDP length. */ + ip_len0 = clib_net_to_host_u16 (ip0->length); + udp_len0 = clib_net_to_host_u16 (udp0->length); + + len_diff0 = ip_len0 - udp_len0; + + len_diff0 = is_udp0 ? len_diff0 : 0; + + if (PREDICT_FALSE (!(is_tcp_udp0 & good_tcp_udp0))) + { + if (is_tcp_udp0) + { + if (is_tcp_udp0 + && !(flags0 & IP_BUFFER_L4_CHECKSUM_COMPUTED)) + flags0 = ip4_tcp_udp_validate_checksum (vm, p0); + good_tcp_udp0 = + (flags0 & IP_BUFFER_L4_CHECKSUM_CORRECT) != 0; + good_tcp_udp0 |= is_udp0 && udp0->checksum == 0; + } + } + + good_tcp_udp0 &= len_diff0 >= 0; + + leaf0 = + ip4_fib_mtrie_lookup_step (mtrie0, leaf0, &ip0->src_address, 2); + + error0 = IP4_ERROR_UNKNOWN_PROTOCOL; + + error0 = len_diff0 < 0 ? IP4_ERROR_UDP_LENGTH : error0; + + ASSERT (IP4_ERROR_TCP_CHECKSUM + 1 == IP4_ERROR_UDP_CHECKSUM); + error0 = (is_tcp_udp0 && !good_tcp_udp0 + ? IP4_ERROR_TCP_CHECKSUM + is_udp0 : error0); + + leaf0 = + ip4_fib_mtrie_lookup_step (mtrie0, leaf0, &ip0->src_address, 3); + leaf0 = + (leaf0 == + IP4_FIB_MTRIE_LEAF_EMPTY ? mtrie0->default_leaf : leaf0); + + lbi0 = ip4_fib_mtrie_leaf_get_adj_index (leaf0); + vnet_buffer (p0)->ip.adj_index[VLIB_TX] = lbi0; + + lb0 = load_balance_get (lbi0); + dpo0 = load_balance_get_bucket_i (lb0, 0); + + vnet_buffer (p0)->ip.adj_index[VLIB_TX] = + vnet_buffer (p0)->ip.adj_index[VLIB_RX] = lbi0; + + error0 = ((error0 == IP4_ERROR_UNKNOWN_PROTOCOL && + dpo0->dpoi_type == DPO_RECEIVE) ? + IP4_ERROR_SPOOFED_LOCAL_PACKETS : error0); + error0 = ((error0 == IP4_ERROR_UNKNOWN_PROTOCOL && + !fib_urpf_check_size (lb0->lb_urpf) && + ip0->dst_address.as_u32 != 0xFFFFFFFF) + ? IP4_ERROR_SRC_LOOKUP_MISS : error0); + + next0 = lm->local_next_by_ip_protocol[proto0]; + + next0 = + error0 != IP4_ERROR_UNKNOWN_PROTOCOL ? IP_LOCAL_NEXT_DROP : next0; + + p0->error = error0 ? error_node->errors[error0] : 0; + + if (PREDICT_FALSE (next0 != next_index)) + { + 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; + } + } + + vlib_put_next_frame (vm, node, next_index, n_left_to_next); + } + + return frame->n_vectors; +} + +VLIB_REGISTER_NODE (ip4_local_node, static) = +{ + .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",} +,}; + +VLIB_NODE_FUNCTION_MULTIARCH (ip4_local_node, ip4_local); + +void +ip4_register_protocol (u32 protocol, u32 node_index) +{ + vlib_main_t *vm = vlib_get_main (); + ip4_main_t *im = &ip4_main; + ip_lookup_main_t *lm = &im->lookup_main; + + ASSERT (protocol < ARRAY_LEN (lm->local_next_by_ip_protocol)); + lm->local_next_by_ip_protocol[protocol] = + vlib_node_add_next (vm, ip4_local_node.index, node_index); +} + +static clib_error_t * +show_ip_local_command_fn (vlib_main_t * vm, + unformat_input_t * input, vlib_cli_command_t * cmd) +{ + ip4_main_t *im = &ip4_main; + ip_lookup_main_t *lm = &im->lookup_main; + int i; + + vlib_cli_output (vm, "Protocols handled by ip4_local"); + for (i = 0; i < ARRAY_LEN (lm->local_next_by_ip_protocol); i++) + { + if (lm->local_next_by_ip_protocol[i] != IP_LOCAL_NEXT_PUNT) + vlib_cli_output (vm, "%d", i); + } + return 0; +} + + + +/*? + * Display the set of protocols handled by the local IPv4 stack. + * + * @cliexpar + * Example of how to display local protocol table: + * @cliexstart{show ip local} + * Protocols handled by ip4_local + * 1 + * 17 + * 47 + * @cliexend +?*/ +/* *INDENT-OFF* */ +VLIB_CLI_COMMAND (show_ip_local, static) = +{ + .path = "show ip local", + .function = show_ip_local_command_fn, + .short_help = "show ip local", +}; +/* *INDENT-ON* */ + +always_inline uword +ip4_arp_inline (vlib_main_t * vm, + vlib_node_runtime_t * node, + vlib_frame_t * frame, int is_glean) +{ + vnet_main_t *vnm = vnet_get_main (); + ip4_main_t *im = &ip4_main; + ip_lookup_main_t *lm = &im->lookup_main; + u32 *from, *to_next_drop; + uword n_left_from, n_left_to_next_drop, next_index; + static f64 time_last_seed_change = -1e100; + static u32 hash_seeds[3]; + static uword hash_bitmap[256 / BITS (uword)]; + f64 time_now; + + if (node->flags & VLIB_NODE_FLAG_TRACE) + ip4_forward_next_trace (vm, node, frame, VLIB_TX); + + time_now = vlib_time_now (vm); + if (time_now - time_last_seed_change > 1e-3) + { + uword i; + u32 *r = clib_random_buffer_get_data (&vm->random_buffer, + sizeof (hash_seeds)); + for (i = 0; i < ARRAY_LEN (hash_seeds); i++) + hash_seeds[i] = r[i]; + + /* Mark all hash keys as been no-seen before. */ + for (i = 0; i < ARRAY_LEN (hash_bitmap); i++) + hash_bitmap[i] = 0; + + time_last_seed_change = time_now; + } + + from = vlib_frame_vector_args (frame); + n_left_from = frame->n_vectors; + next_index = node->cached_next_index; + if (next_index == IP4_ARP_NEXT_DROP) + next_index = IP4_ARP_N_NEXT; /* point to first interface */ + + while (n_left_from > 0) + { + vlib_get_next_frame (vm, node, IP4_ARP_NEXT_DROP, + to_next_drop, n_left_to_next_drop); + + while (n_left_from > 0 && n_left_to_next_drop > 0) + { + u32 pi0, adj_index0, a0, b0, c0, m0, sw_if_index0, drop0; + ip_adjacency_t *adj0; + vlib_buffer_t *p0; + ip4_header_t *ip0; + uword bm0; + + pi0 = from[0]; + + p0 = vlib_get_buffer (vm, pi0); + + adj_index0 = vnet_buffer (p0)->ip.adj_index[VLIB_TX]; + adj0 = ip_get_adjacency (lm, adj_index0); + ip0 = vlib_buffer_get_current (p0); + + a0 = hash_seeds[0]; + b0 = hash_seeds[1]; + c0 = hash_seeds[2]; + + sw_if_index0 = adj0->rewrite_header.sw_if_index; + vnet_buffer (p0)->sw_if_index[VLIB_TX] = sw_if_index0; + + if (is_glean) + { + /* + * this is the Glean case, so we are ARPing for the + * packet's destination + */ + a0 ^= ip0->dst_address.data_u32; + } + else + { + a0 ^= adj0->sub_type.nbr.next_hop.ip4.data_u32; + } + b0 ^= sw_if_index0; + + hash_v3_finalize32 (a0, b0, c0); + + c0 &= BITS (hash_bitmap) - 1; + c0 = c0 / BITS (uword); + m0 = (uword) 1 << (c0 % BITS (uword)); + + bm0 = hash_bitmap[c0]; + drop0 = (bm0 & m0) != 0; + + /* Mark it as seen. */ + hash_bitmap[c0] = bm0 | m0; + + from += 1; + n_left_from -= 1; + to_next_drop[0] = pi0; + to_next_drop += 1; + n_left_to_next_drop -= 1; + + p0->error = + node->errors[drop0 ? IP4_ARP_ERROR_DROP : + IP4_ARP_ERROR_REQUEST_SENT]; + + /* + * the adj has been updated to a rewrite but the node the DPO that got + * us here hasn't - yet. no big deal. we'll drop while we wait. + */ + if (IP_LOOKUP_NEXT_REWRITE == adj0->lookup_next_index) + continue; + + if (drop0) + continue; + + /* + * Can happen if the control-plane is programming tables + * with traffic flowing; at least that's today's lame excuse. + */ + if ((is_glean && adj0->lookup_next_index != IP_LOOKUP_NEXT_GLEAN) || + (!is_glean && adj0->lookup_next_index != IP_LOOKUP_NEXT_ARP)) + { + p0->error = node->errors[IP4_ARP_ERROR_NON_ARP_ADJ]; + } + else + /* Send ARP request. */ + { + u32 bi0 = 0; + vlib_buffer_t *b0; + ethernet_arp_header_t *h0; + vnet_hw_interface_t *hw_if0; + + h0 = + vlib_packet_template_get_packet (vm, + &im->ip4_arp_request_packet_template, + &bi0); + + /* Add rewrite/encap string for ARP packet. */ + vnet_rewrite_one_header (adj0[0], h0, + sizeof (ethernet_header_t)); + + hw_if0 = vnet_get_sup_hw_interface (vnm, sw_if_index0); + + /* Src ethernet address in ARP header. */ + clib_memcpy (h0->ip4_over_ethernet[0].ethernet, + hw_if0->hw_address, + sizeof (h0->ip4_over_ethernet[0].ethernet)); + + if (is_glean) + { + /* The interface's source address is stashed in the Glean Adj */ + h0->ip4_over_ethernet[0].ip4 = + adj0->sub_type.glean.receive_addr.ip4; + + /* Copy in destination address we are requesting. This is the + * glean case, so it's the packet's destination.*/ + h0->ip4_over_ethernet[1].ip4.data_u32 = + ip0->dst_address.data_u32; + } + else + { + /* Src IP address in ARP header. */ + if (ip4_src_address_for_packet (lm, sw_if_index0, + &h0-> + ip4_over_ethernet[0].ip4)) + { + /* No source address available */ + p0->error = + node->errors[IP4_ARP_ERROR_NO_SOURCE_ADDRESS]; + vlib_buffer_free (vm, &bi0, 1); + continue; + } + + /* Copy in destination address we are requesting from the + incomplete adj */ + h0->ip4_over_ethernet[1].ip4.data_u32 = + adj0->sub_type.nbr.next_hop.ip4.as_u32; + } + + vlib_buffer_copy_trace_flag (vm, p0, bi0); + b0 = vlib_get_buffer (vm, bi0); + vnet_buffer (b0)->sw_if_index[VLIB_TX] = sw_if_index0; + + vlib_buffer_advance (b0, -adj0->rewrite_header.data_bytes); + + vlib_set_next_frame_buffer (vm, node, + adj0->rewrite_header.next_index, + bi0); + } + } + + vlib_put_next_frame (vm, node, IP4_ARP_NEXT_DROP, n_left_to_next_drop); + } + + return frame->n_vectors; +} + +static uword +ip4_arp (vlib_main_t * vm, vlib_node_runtime_t * node, vlib_frame_t * frame) +{ + return (ip4_arp_inline (vm, node, frame, 0)); +} + +static uword +ip4_glean (vlib_main_t * vm, vlib_node_runtime_t * node, vlib_frame_t * frame) +{ + return (ip4_arp_inline (vm, node, frame, 1)); +} + +static char *ip4_arp_error_strings[] = { + [IP4_ARP_ERROR_DROP] = "address overflow drops", + [IP4_ARP_ERROR_REQUEST_SENT] = "ARP requests sent", + [IP4_ARP_ERROR_NON_ARP_ADJ] = "ARPs to non-ARP adjacencies", + [IP4_ARP_ERROR_REPLICATE_DROP] = "ARP replication completed", + [IP4_ARP_ERROR_REPLICATE_FAIL] = "ARP replication failed", + [IP4_ARP_ERROR_NO_SOURCE_ADDRESS] = "no source address for ARP request", +}; + +VLIB_REGISTER_NODE (ip4_arp_node) = +{ + .function = ip4_arp,.name = "ip4-arp",.vector_size = + sizeof (u32),.format_trace = format_ip4_forward_next_trace,.n_errors = + ARRAY_LEN (ip4_arp_error_strings),.error_strings = + ip4_arp_error_strings,.n_next_nodes = IP4_ARP_N_NEXT,.next_nodes = + { + [IP4_ARP_NEXT_DROP] = "error-drop",} +,}; + +VLIB_REGISTER_NODE (ip4_glean_node) = +{ + .function = ip4_glean,.name = "ip4-glean",.vector_size = + sizeof (u32),.format_trace = format_ip4_forward_next_trace,.n_errors = + ARRAY_LEN (ip4_arp_error_strings),.error_strings = + ip4_arp_error_strings,.n_next_nodes = IP4_ARP_N_NEXT,.next_nodes = + { + [IP4_ARP_NEXT_DROP] = "error-drop",} +,}; + +#define foreach_notrace_ip4_arp_error \ +_(DROP) \ +_(REQUEST_SENT) \ +_(REPLICATE_DROP) \ +_(REPLICATE_FAIL) + +clib_error_t * +arp_notrace_init (vlib_main_t * vm) +{ + vlib_node_runtime_t *rt = vlib_node_get_runtime (vm, ip4_arp_node.index); + + /* don't trace ARP request packets */ +#define _(a) \ + vnet_pcap_drop_trace_filter_add_del \ + (rt->errors[IP4_ARP_ERROR_##a], \ + 1 /* is_add */); + foreach_notrace_ip4_arp_error; +#undef _ + return 0; +} + +VLIB_INIT_FUNCTION (arp_notrace_init); + + +/* Send an ARP request to see if given destination is reachable on given interface. */ +clib_error_t * +ip4_probe_neighbor (vlib_main_t * vm, ip4_address_t * dst, u32 sw_if_index) +{ + vnet_main_t *vnm = vnet_get_main (); + ip4_main_t *im = &ip4_main; + ethernet_arp_header_t *h; + ip4_address_t *src; + ip_interface_address_t *ia; + ip_adjacency_t *adj; + vnet_hw_interface_t *hi; + vnet_sw_interface_t *si; + vlib_buffer_t *b; + u32 bi = 0; + + si = vnet_get_sw_interface (vnm, sw_if_index); + + if (!(si->flags & VNET_SW_INTERFACE_FLAG_ADMIN_UP)) + { + return clib_error_return (0, "%U: interface %U down", + format_ip4_address, dst, + format_vnet_sw_if_index_name, vnm, + sw_if_index); + } + + src = + ip4_interface_address_matching_destination (im, dst, sw_if_index, &ia); + if (!src) + { + vnm->api_errno = VNET_API_ERROR_NO_MATCHING_INTERFACE; + return clib_error_return + (0, "no matching interface address for destination %U (interface %U)", + format_ip4_address, dst, + format_vnet_sw_if_index_name, vnm, sw_if_index); + } + + adj = ip_get_adjacency (&im->lookup_main, ia->neighbor_probe_adj_index); + + h = + vlib_packet_template_get_packet (vm, &im->ip4_arp_request_packet_template, + &bi); + + hi = vnet_get_sup_hw_interface (vnm, sw_if_index); + + clib_memcpy (h->ip4_over_ethernet[0].ethernet, hi->hw_address, + sizeof (h->ip4_over_ethernet[0].ethernet)); + + h->ip4_over_ethernet[0].ip4 = src[0]; + h->ip4_over_ethernet[1].ip4 = dst[0]; + + b = vlib_get_buffer (vm, bi); + vnet_buffer (b)->sw_if_index[VLIB_RX] = + vnet_buffer (b)->sw_if_index[VLIB_TX] = sw_if_index; + + /* Add encapsulation string for software interface (e.g. ethernet header). */ + vnet_rewrite_one_header (adj[0], h, sizeof (ethernet_header_t)); + vlib_buffer_advance (b, -adj->rewrite_header.data_bytes); + + { + vlib_frame_t *f = vlib_get_frame_to_node (vm, hi->output_node_index); + u32 *to_next = vlib_frame_vector_args (f); + to_next[0] = bi; + f->n_vectors = 1; + vlib_put_frame_to_node (vm, hi->output_node_index, f); + } + + return /* no error */ 0; +} + +typedef enum +{ + IP4_REWRITE_NEXT_DROP, + IP4_REWRITE_NEXT_ICMP_ERROR, +} ip4_rewrite_next_t; + +always_inline uword +ip4_rewrite_inline (vlib_main_t * vm, + vlib_node_runtime_t * node, + vlib_frame_t * frame, int is_midchain) +{ + ip_lookup_main_t *lm = &ip4_main.lookup_main; + u32 *from = vlib_frame_vector_args (frame); + u32 n_left_from, n_left_to_next, *to_next, next_index; + vlib_node_runtime_t *error_node = + vlib_node_get_runtime (vm, ip4_input_node.index); + + n_left_from = frame->n_vectors; + next_index = node->cached_next_index; + u32 cpu_index = os_get_cpu_number (); + + while (n_left_from > 0) + { + vlib_get_next_frame (vm, node, next_index, to_next, n_left_to_next); + + while (n_left_from >= 4 && n_left_to_next >= 2) + { + ip_adjacency_t *adj0, *adj1; + vlib_buffer_t *p0, *p1; + ip4_header_t *ip0, *ip1; + u32 pi0, rw_len0, next0, error0, checksum0, adj_index0; + u32 pi1, rw_len1, next1, error1, checksum1, adj_index1; + u32 tx_sw_if_index0, tx_sw_if_index1; + + /* 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, STORE); + vlib_prefetch_buffer_header (p3, STORE); + + CLIB_PREFETCH (p2->data, sizeof (ip0[0]), STORE); + CLIB_PREFETCH (p3->data, sizeof (ip0[0]), STORE); + } + + pi0 = to_next[0] = from[0]; + pi1 = to_next[1] = from[1]; + + from += 2; + n_left_from -= 2; + to_next += 2; + n_left_to_next -= 2; + + p0 = vlib_get_buffer (vm, pi0); + p1 = vlib_get_buffer (vm, pi1); + + adj_index0 = vnet_buffer (p0)->ip.adj_index[VLIB_TX]; + adj_index1 = vnet_buffer (p1)->ip.adj_index[VLIB_TX]; + + /* We should never rewrite a pkt using the MISS adjacency */ + ASSERT (adj_index0 && adj_index1); + + ip0 = vlib_buffer_get_current (p0); + ip1 = vlib_buffer_get_current (p1); + + error0 = error1 = IP4_ERROR_NONE; + next0 = next1 = IP4_REWRITE_NEXT_DROP; + + /* Decrement TTL & update checksum. + Works either endian, so no need for byte swap. */ + if (PREDICT_TRUE (!(p0->flags & VNET_BUFFER_LOCALLY_ORIGINATED))) + { + i32 ttl0 = ip0->ttl; + + /* Input node should have reject packets with ttl 0. */ + ASSERT (ip0->ttl > 0); + + checksum0 = ip0->checksum + clib_host_to_net_u16 (0x0100); + checksum0 += checksum0 >= 0xffff; + + ip0->checksum = checksum0; + ttl0 -= 1; + ip0->ttl = ttl0; + + /* + * If the ttl drops below 1 when forwarding, generate + * an ICMP response. + */ + if (PREDICT_FALSE (ttl0 <= 0)) + { + error0 = IP4_ERROR_TIME_EXPIRED; + vnet_buffer (p0)->sw_if_index[VLIB_TX] = (u32) ~ 0; + icmp4_error_set_vnet_buffer (p0, ICMP4_time_exceeded, + ICMP4_time_exceeded_ttl_exceeded_in_transit, + 0); + next0 = IP4_REWRITE_NEXT_ICMP_ERROR; + } + + /* Verify checksum. */ + ASSERT (ip0->checksum == ip4_header_checksum (ip0)); + } + else + { + p0->flags &= ~VNET_BUFFER_LOCALLY_ORIGINATED; + } + if (PREDICT_TRUE (!(p1->flags & VNET_BUFFER_LOCALLY_ORIGINATED))) + { + i32 ttl1 = ip1->ttl; + + /* Input node should have reject packets with ttl 0. */ + ASSERT (ip1->ttl > 0); + + checksum1 = ip1->checksum + clib_host_to_net_u16 (0x0100); + checksum1 += checksum1 >= 0xffff; + + ip1->checksum = checksum1; + ttl1 -= 1; + ip1->ttl = ttl1; + + /* + * If the ttl drops below 1 when forwarding, generate + * an ICMP response. + */ + if (PREDICT_FALSE (ttl1 <= 0)) + { + error1 = IP4_ERROR_TIME_EXPIRED; + vnet_buffer (p1)->sw_if_index[VLIB_TX] = (u32) ~ 0; + icmp4_error_set_vnet_buffer (p1, ICMP4_time_exceeded, + ICMP4_time_exceeded_ttl_exceeded_in_transit, + 0); + next1 = IP4_REWRITE_NEXT_ICMP_ERROR; + } + + /* Verify checksum. */ + ASSERT (ip0->checksum == ip4_header_checksum (ip0)); + ASSERT (ip1->checksum == ip4_header_checksum (ip1)); + } + else + { + p1->flags &= ~VNET_BUFFER_LOCALLY_ORIGINATED; + } + + /* Rewrite packet header and updates lengths. */ + adj0 = ip_get_adjacency (lm, adj_index0); + adj1 = ip_get_adjacency (lm, adj_index1); + + /* Worth pipelining. No guarantee that adj0,1 are hot... */ + rw_len0 = adj0[0].rewrite_header.data_bytes; + rw_len1 = adj1[0].rewrite_header.data_bytes; + vnet_buffer (p0)->ip.save_rewrite_length = rw_len0; + vnet_buffer (p1)->ip.save_rewrite_length = rw_len1; + + /* Check MTU of outgoing interface. */ + error0 = + (vlib_buffer_length_in_chain (vm, p0) > + adj0[0]. + rewrite_header.max_l3_packet_bytes ? IP4_ERROR_MTU_EXCEEDED : + error0); + error1 = + (vlib_buffer_length_in_chain (vm, p1) > + adj1[0]. + rewrite_header.max_l3_packet_bytes ? IP4_ERROR_MTU_EXCEEDED : + error1); + + /* + * We've already accounted for an ethernet_header_t elsewhere + */ + if (PREDICT_FALSE (rw_len0 > sizeof (ethernet_header_t))) + vlib_increment_combined_counter + (&adjacency_counters, cpu_index, adj_index0, + /* packet increment */ 0, + /* byte increment */ rw_len0 - sizeof (ethernet_header_t)); + + if (PREDICT_FALSE (rw_len1 > sizeof (ethernet_header_t))) + vlib_increment_combined_counter + (&adjacency_counters, cpu_index, adj_index1, + /* packet increment */ 0, + /* byte increment */ rw_len1 - sizeof (ethernet_header_t)); + + /* Don't adjust the buffer for ttl issue; icmp-error node wants + * to see the IP headerr */ + if (PREDICT_TRUE (error0 == IP4_ERROR_NONE)) + { + next0 = adj0[0].rewrite_header.next_index; + p0->current_data -= rw_len0; + p0->current_length += rw_len0; + tx_sw_if_index0 = adj0[0].rewrite_header.sw_if_index; + vnet_buffer (p0)->sw_if_index[VLIB_TX] = tx_sw_if_index0; + + vnet_feature_arc_start (lm->output_feature_arc_index, + tx_sw_if_index0, &next0, p0); + } + if (PREDICT_TRUE (error1 == IP4_ERROR_NONE)) + { + next1 = adj1[0].rewrite_header.next_index; + p1->current_data -= rw_len1; + p1->current_length += rw_len1; + + tx_sw_if_index1 = adj1[0].rewrite_header.sw_if_index; + vnet_buffer (p1)->sw_if_index[VLIB_TX] = tx_sw_if_index1; + + vnet_feature_arc_start (lm->output_feature_arc_index, + tx_sw_if_index1, &next1, p1); + } + + /* Guess we are only writing on simple Ethernet header. */ + vnet_rewrite_two_headers (adj0[0], adj1[0], + ip0, ip1, sizeof (ethernet_header_t)); + + if (is_midchain) + { + adj0->sub_type.midchain.fixup_func (vm, adj0, p0); + adj1->sub_type.midchain.fixup_func (vm, adj1, 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) + { + ip_adjacency_t *adj0; + vlib_buffer_t *p0; + ip4_header_t *ip0; + u32 pi0, rw_len0, adj_index0, next0, error0, checksum0; + u32 tx_sw_if_index0; + + pi0 = to_next[0] = from[0]; + + p0 = vlib_get_buffer (vm, pi0); + + adj_index0 = vnet_buffer (p0)->ip.adj_index[VLIB_TX]; + + /* We should never rewrite a pkt using the MISS adjacency */ + ASSERT (adj_index0); + + adj0 = ip_get_adjacency (lm, adj_index0); + + ip0 = vlib_buffer_get_current (p0); + + error0 = IP4_ERROR_NONE; + next0 = IP4_REWRITE_NEXT_DROP; /* drop on error */ + + /* Decrement TTL & update checksum. */ + if (PREDICT_TRUE (!(p0->flags & VNET_BUFFER_LOCALLY_ORIGINATED))) + { + i32 ttl0 = ip0->ttl; + + checksum0 = ip0->checksum + clib_host_to_net_u16 (0x0100); + + checksum0 += checksum0 >= 0xffff; + + ip0->checksum = checksum0; + + ASSERT (ip0->ttl > 0); + + ttl0 -= 1; + + ip0->ttl = ttl0; + + ASSERT (ip0->checksum == ip4_header_checksum (ip0)); + + if (PREDICT_FALSE (ttl0 <= 0)) + { + /* + * If the ttl drops below 1 when forwarding, generate + * an ICMP response. + */ + error0 = IP4_ERROR_TIME_EXPIRED; + next0 = IP4_REWRITE_NEXT_ICMP_ERROR; + vnet_buffer (p0)->sw_if_index[VLIB_TX] = (u32) ~ 0; + icmp4_error_set_vnet_buffer (p0, ICMP4_time_exceeded, + ICMP4_time_exceeded_ttl_exceeded_in_transit, + 0); + } + } + else + { + p0->flags &= ~VNET_BUFFER_LOCALLY_ORIGINATED; + } + + /* Guess we are only writing on simple Ethernet header. */ + vnet_rewrite_one_header (adj0[0], ip0, sizeof (ethernet_header_t)); + + /* Update packet buffer attributes/set output interface. */ + rw_len0 = adj0[0].rewrite_header.data_bytes; + vnet_buffer (p0)->ip.save_rewrite_length = rw_len0; + + if (PREDICT_FALSE (rw_len0 > sizeof (ethernet_header_t))) + vlib_increment_combined_counter + (&adjacency_counters, cpu_index, adj_index0, + /* packet increment */ 0, + /* byte increment */ rw_len0 - sizeof (ethernet_header_t)); + + /* Check MTU of outgoing interface. */ + error0 = (vlib_buffer_length_in_chain (vm, p0) + > adj0[0].rewrite_header.max_l3_packet_bytes + ? IP4_ERROR_MTU_EXCEEDED : error0); + + p0->error = error_node->errors[error0]; + + /* Don't adjust the buffer for ttl issue; icmp-error node wants + * to see the IP headerr */ + if (PREDICT_TRUE (error0 == IP4_ERROR_NONE)) + { + p0->current_data -= rw_len0; + p0->current_length += rw_len0; + tx_sw_if_index0 = adj0[0].rewrite_header.sw_if_index; + + vnet_buffer (p0)->sw_if_index[VLIB_TX] = tx_sw_if_index0; + next0 = adj0[0].rewrite_header.next_index; + + if (is_midchain) + { + adj0->sub_type.midchain.fixup_func (vm, adj0, p0); + } + + vnet_feature_arc_start (lm->output_feature_arc_index, + tx_sw_if_index0, &next0, p0); + + } + + from += 1; + n_left_from -= 1; + to_next += 1; + n_left_to_next -= 1; + + 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); + } + + /* Need to do trace after rewrites to pick up new packet data. */ + if (node->flags & VLIB_NODE_FLAG_TRACE) + ip4_forward_next_trace (vm, node, frame, VLIB_TX); + + return frame->n_vectors; +} + + +/** @brief IPv4 rewrite node. + @node ip4-rewrite + + This is the IPv4 transit-rewrite node: decrement TTL, fix the ipv4 + header checksum, fetch the ip adjacency, check the outbound mtu, + apply the adjacency rewrite, and send pkts to the adjacency + rewrite header's rewrite_next_index. + + @param vm vlib_main_t corresponding to the current thread + @param node vlib_node_runtime_t + @param frame vlib_frame_t whose contents should be dispatched + + @par Graph mechanics: buffer metadata, next index usage + + @em Uses: + - vnet_buffer(b)->ip.adj_index[VLIB_TX] + - the rewrite adjacency index + - adj->lookup_next_index + - Must be IP_LOOKUP_NEXT_REWRITE or IP_LOOKUP_NEXT_ARP, otherwise + the packet will be dropped. + - adj->rewrite_header + - Rewrite string length, rewrite string, next_index + + @em Sets: + - b->current_data, b->current_length + - Updated net of applying the rewrite string + + Next Indices: + - adj->rewrite_header.next_index + or @c error-drop +*/ +static uword +ip4_rewrite (vlib_main_t * vm, + vlib_node_runtime_t * node, vlib_frame_t * frame) +{ + return ip4_rewrite_inline (vm, node, frame, 0); +} + +static uword +ip4_midchain (vlib_main_t * vm, + vlib_node_runtime_t * node, vlib_frame_t * frame) +{ + return ip4_rewrite_inline (vm, node, frame, 1); +} + + +VLIB_REGISTER_NODE (ip4_rewrite_node) = +{ + .function = ip4_rewrite,.name = "ip4-rewrite",.vector_size = + sizeof (u32),.format_trace = format_ip4_rewrite_trace,.n_next_nodes = + 2,.next_nodes = + { + [IP4_REWRITE_NEXT_DROP] = "error-drop", + [IP4_REWRITE_NEXT_ICMP_ERROR] = "ip4-icmp-error",} +,}; + +VLIB_NODE_FUNCTION_MULTIARCH (ip4_rewrite_node, ip4_rewrite); + +VLIB_REGISTER_NODE (ip4_midchain_node) = +{ +.function = ip4_midchain,.name = "ip4-midchain",.vector_size = + sizeof (u32),.format_trace = format_ip4_forward_next_trace,.sibling_of = + "ip4-rewrite",}; + +VLIB_NODE_FUNCTION_MULTIARCH (ip4_midchain_node, ip4_midchain); + +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 (); + 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; + } + + { + ip4_main_t *im = &ip4_main; + u32 fib_index; + + fib_index = fib_table_find_or_create_and_lock (FIB_PROTOCOL_IP4, + table_id); + + // + // FIXME-LATER + // changing an interface's table has consequences for any connecteds + // and adj-fibs already installed. + // + vec_validate (im->fib_index_by_sw_if_index, sw_if_index); + im->fib_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 end up in + * the indicated FIB table. If the IP address is added prior to adding the + * interface to the FIB table, it will NOT be part of the FIB table. Predictable + * but potentially counter-intuitive results occur if you provision interface + * addresses in multiple FIBs. Upon RX, packets will be processed in the last + * IP table ID provisioned. It might be marginally useful to evade source RPF + * drops to put an interface address into multiple FIBs. + * + * @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* */ + + +static uword +ip4_lookup_multicast (vlib_main_t * vm, + vlib_node_runtime_t * node, vlib_frame_t * frame) +{ + ip4_main_t *im = &ip4_main; + 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 (); + + from = vlib_frame_vector_args (frame); + n_left_from = frame->n_vectors; + next = node->cached_next_index; + + while (n_left_from > 0) + { + vlib_get_next_frame (vm, node, next, to_next, n_left_to_next); + + while (n_left_from >= 4 && n_left_to_next >= 2) + { + vlib_buffer_t *p0, *p1; + u32 pi0, pi1, lb_index0, lb_index1, wrong_next; + ip_lookup_next_t next0, next1; + ip4_header_t *ip0, *ip1; + u32 fib_index0, fib_index1; + const dpo_id_t *dpo0, *dpo1; + const load_balance_t *lb0, *lb1; + + /* 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 (ip0[0]), LOAD); + CLIB_PREFETCH (p3->data, sizeof (ip0[0]), LOAD); + } + + pi0 = to_next[0] = from[0]; + pi1 = to_next[1] = from[1]; + + p0 = vlib_get_buffer (vm, pi0); + p1 = vlib_get_buffer (vm, pi1); + + ip0 = vlib_buffer_get_current (p0); + ip1 = vlib_buffer_get_current (p1); + + fib_index0 = + vec_elt (im->fib_index_by_sw_if_index, + vnet_buffer (p0)->sw_if_index[VLIB_RX]); + fib_index1 = + vec_elt (im->fib_index_by_sw_if_index, + vnet_buffer (p1)->sw_if_index[VLIB_RX]); + fib_index0 = + (vnet_buffer (p0)->sw_if_index[VLIB_TX] == + (u32) ~ 0) ? fib_index0 : vnet_buffer (p0)->sw_if_index[VLIB_TX]; + fib_index1 = + (vnet_buffer (p1)->sw_if_index[VLIB_TX] == + (u32) ~ 0) ? fib_index1 : vnet_buffer (p1)->sw_if_index[VLIB_TX]; + + lb_index0 = ip4_fib_table_lookup_lb (ip4_fib_get (fib_index0), + &ip0->dst_address); + lb_index1 = ip4_fib_table_lookup_lb (ip4_fib_get (fib_index1), + &ip1->dst_address); + + lb0 = load_balance_get (lb_index0); + lb1 = load_balance_get (lb_index1); + + ASSERT (lb0->lb_n_buckets > 0); + ASSERT (is_pow2 (lb0->lb_n_buckets)); + ASSERT (lb1->lb_n_buckets > 0); + ASSERT (is_pow2 (lb1->lb_n_buckets)); + + vnet_buffer (p0)->ip.flow_hash = ip4_compute_flow_hash + (ip0, lb0->lb_hash_config); + + vnet_buffer (p1)->ip.flow_hash = ip4_compute_flow_hash + (ip1, lb1->lb_hash_config); + + dpo0 = load_balance_get_bucket_i (lb0, + (vnet_buffer (p0)->ip.flow_hash & + (lb0->lb_n_buckets_minus_1))); + dpo1 = load_balance_get_bucket_i (lb1, + (vnet_buffer (p1)->ip.flow_hash & + (lb1->lb_n_buckets_minus_1))); + + next0 = dpo0->dpoi_next_node; + vnet_buffer (p0)->ip.adj_index[VLIB_TX] = dpo0->dpoi_index; + next1 = dpo1->dpoi_next_node; + vnet_buffer (p1)->ip.adj_index[VLIB_TX] = dpo1->dpoi_index; + + if (1) /* $$$$$$ HACK FIXME */ + vlib_increment_combined_counter + (cm, cpu_index, lb_index0, 1, + vlib_buffer_length_in_chain (vm, p0)); + if (1) /* $$$$$$ HACK FIXME */ + vlib_increment_combined_counter + (cm, cpu_index, lb_index1, 1, + vlib_buffer_length_in_chain (vm, p1)); + + from += 2; + to_next += 2; + n_left_to_next -= 2; + n_left_from -= 2; + + wrong_next = (next0 != next) + 2 * (next1 != next); + if (PREDICT_FALSE (wrong_next != 0)) + { + switch (wrong_next) + { + 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 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) + { + /* A B B */ + vlib_put_next_frame (vm, node, next, n_left_to_next); + next = next1; + vlib_get_next_frame (vm, node, next, to_next, + n_left_to_next); + } + } + } + } + + while (n_left_from > 0 && n_left_to_next > 0) + { + vlib_buffer_t *p0; + ip4_header_t *ip0; + u32 pi0, lb_index0; + ip_lookup_next_t next0; + u32 fib_index0; + const dpo_id_t *dpo0; + const load_balance_t *lb0; + + pi0 = from[0]; + to_next[0] = pi0; + + p0 = vlib_get_buffer (vm, pi0); + + ip0 = vlib_buffer_get_current (p0); + + fib_index0 = vec_elt (im->fib_index_by_sw_if_index, + vnet_buffer (p0)->sw_if_index[VLIB_RX]); + fib_index0 = (vnet_buffer (p0)->sw_if_index[VLIB_TX] == (u32) ~ 0) ? + fib_index0 : vnet_buffer (p0)->sw_if_index[VLIB_TX]; + + lb_index0 = ip4_fib_table_lookup_lb (ip4_fib_get (fib_index0), + &ip0->dst_address); + + lb0 = load_balance_get (lb_index0); + + ASSERT (lb0->lb_n_buckets > 0); + ASSERT (is_pow2 (lb0->lb_n_buckets)); + + vnet_buffer (p0)->ip.flow_hash = ip4_compute_flow_hash + (ip0, lb0->lb_hash_config); + + dpo0 = load_balance_get_bucket_i (lb0, + (vnet_buffer (p0)->ip.flow_hash & + (lb0->lb_n_buckets_minus_1))); + + next0 = dpo0->dpoi_next_node; + vnet_buffer (p0)->ip.adj_index[VLIB_TX] = dpo0->dpoi_index; + + if (1) /* $$$$$$ HACK FIXME */ + vlib_increment_combined_counter + (cm, cpu_index, lb_index0, 1, + vlib_buffer_length_in_chain (vm, p0)); + + from += 1; + to_next += 1; + n_left_to_next -= 1; + n_left_from -= 1; + + if (PREDICT_FALSE (next0 != next)) + { + n_left_to_next += 1; + vlib_put_next_frame (vm, node, next, n_left_to_next); + next = next0; + vlib_get_next_frame (vm, node, next, to_next, n_left_to_next); + to_next[0] = pi0; + to_next += 1; + n_left_to_next -= 1; + } + } + + vlib_put_next_frame (vm, node, next, n_left_to_next); + } + + if (node->flags & VLIB_NODE_FLAG_TRACE) + ip4_forward_next_trace (vm, node, frame, VLIB_TX); + + return frame->n_vectors; +} + +VLIB_REGISTER_NODE (ip4_lookup_multicast_node, static) = +{ +.function = ip4_lookup_multicast,.name = + "ip4-lookup-multicast",.vector_size = sizeof (u32),.sibling_of = + "ip4-lookup",.format_trace = format_ip4_lookup_trace,.n_next_nodes = 0,}; + +VLIB_NODE_FUNCTION_MULTIARCH (ip4_lookup_multicast_node, + ip4_lookup_multicast); + +VLIB_REGISTER_NODE (ip4_multicast_node, static) = +{ + .function = ip4_drop,.name = "ip4-multicast",.vector_size = + sizeof (u32),.format_trace = format_ip4_forward_next_trace,.n_next_nodes = + 1,.next_nodes = + { + [0] = "error-drop",} +,}; + +int +ip4_lookup_validate (ip4_address_t * a, u32 fib_index0) +{ + ip4_fib_mtrie_t *mtrie0; + ip4_fib_mtrie_leaf_t leaf0; + u32 lbi0; + + mtrie0 = &ip4_fib_get (fib_index0)->mtrie; + + leaf0 = IP4_FIB_MTRIE_LEAF_ROOT; + leaf0 = ip4_fib_mtrie_lookup_step (mtrie0, leaf0, a, 0); + leaf0 = ip4_fib_mtrie_lookup_step (mtrie0, leaf0, a, 1); + leaf0 = ip4_fib_mtrie_lookup_step (mtrie0, leaf0, a, 2); + leaf0 = ip4_fib_mtrie_lookup_step (mtrie0, leaf0, a, 3); + + /* Handle default route. */ + leaf0 = (leaf0 == IP4_FIB_MTRIE_LEAF_EMPTY ? mtrie0->default_leaf : leaf0); + + lbi0 = ip4_fib_mtrie_leaf_get_adj_index (leaf0); + + return lbi0 == ip4_fib_table_lookup_lb (ip4_fib_get (fib_index0), a); +} + +static clib_error_t * +test_lookup_command_fn (vlib_main_t * vm, + unformat_input_t * input, vlib_cli_command_t * cmd) +{ + ip4_fib_t *fib; + u32 table_id = 0; + f64 count = 1; + u32 n; + int i; + ip4_address_t ip4_base_address; + u64 errors = 0; + + while (unformat_check_input (input) != UNFORMAT_END_OF_INPUT) + { + if (unformat (input, "table %d", &table_id)) + { + /* Make sure the entry exists. */ + fib = ip4_fib_get (table_id); + if ((fib) && (fib->index != table_id)) + return clib_error_return (0, " %d does not exist", + table_id); + } + else if (unformat (input, "count %f", &count)) + ; + + else if (unformat (input, "%U", + unformat_ip4_address, &ip4_base_address)) + ; + else + return clib_error_return (0, "unknown input `%U'", + format_unformat_error, input); + } + + n = count; + + for (i = 0; i < n; i++) + { + if (!ip4_lookup_validate (&ip4_base_address, table_id)) + errors++; + + ip4_base_address.as_u32 = + clib_host_to_net_u32 (1 + + clib_net_to_host_u32 (ip4_base_address.as_u32)); + } + + if (errors) + vlib_cli_output (vm, "%llu errors out of %d lookups\n", errors, n); + else + vlib_cli_output (vm, "No errors in %d lookups\n", n); + + return 0; +} + +/*? + * Perform a lookup of an IPv4 Address (or range of addresses) in the + * given FIB table to determine if there is a conflict with the + * adjacency table. The fib-id can be determined by using the + * 'show ip fib' command. If fib-id is not entered, default value + * of 0 is used. + * + * @todo This command uses fib-id, other commands use table-id (not + * just a name, they are different indexes). Would like to change this + * to table-id for consistency. + * + * @cliexpar + * Example of how to run the test lookup command: + * @cliexstart{test lookup 172.16.1.1 table 1 count 2} + * No errors in 2 lookups + * @cliexend +?*/ +/* *INDENT-OFF* */ +VLIB_CLI_COMMAND (lookup_test_command, static) = +{ + .path = "test lookup", + .short_help = "test lookup [table ] [count ]", + .function = test_lookup_command_fn, +}; +/* *INDENT-ON* */ + +int +vnet_set_ip4_flow_hash (u32 table_id, u32 flow_hash_config) +{ + ip4_main_t *im4 = &ip4_main; + ip4_fib_t *fib; + uword *p = hash_get (im4->fib_index_by_table_id, table_id); + + if (p == 0) + return VNET_API_ERROR_NO_SUCH_FIB; + + fib = ip4_fib_get (p[0]); + + fib->flow_hash_config = flow_hash_config; + return 0; +} + +static clib_error_t * +set_ip_flow_hash_command_fn (vlib_main_t * vm, + unformat_input_t * input, + vlib_cli_command_t * cmd) +{ + int matched = 0; + u32 table_id = 0; + u32 flow_hash_config = 0; + int rv; + + while (unformat_check_input (input) != UNFORMAT_END_OF_INPUT) + { + if (unformat (input, "table %d", &table_id)) + matched = 1; +#define _(a,v) \ + else if (unformat (input, #a)) { flow_hash_config |= v; matched=1;} + foreach_flow_hash_bit +#undef _ + else + break; + } + + if (matched == 0) + return clib_error_return (0, "unknown input `%U'", + format_unformat_error, input); + + rv = vnet_set_ip4_flow_hash (table_id, flow_hash_config); + switch (rv) + { + case 0: + break; + + case VNET_API_ERROR_NO_SUCH_FIB: + return clib_error_return (0, "no such FIB table %d", table_id); + + default: + clib_warning ("BUG: illegal flow hash config 0x%x", flow_hash_config); + break; + } + + return 0; +} + +/*? + * Configure the set of IPv4 fields used by the flow hash. + * + * @cliexpar + * Example of how to set the flow hash on a given table: + * @cliexcmd{set ip flow-hash table 7 dst sport dport proto} + * Example of display the configured flow hash: + * @cliexstart{show ip fib} + * ipv4-VRF:0, fib_index 0, flow hash: src dst sport dport proto + * 0.0.0.0/0 + * unicast-ip4-chain + * [@0]: dpo-load-balance: [index:0 buckets:1 uRPF:0 to:[0:0]] + * [0] [@0]: dpo-drop ip6 + * 0.0.0.0/32 + * unicast-ip4-chain + * [@0]: dpo-load-balance: [index:1 buckets:1 uRPF:1 to:[0:0]] + * [0] [@0]: dpo-drop ip6 + * 224.0.0.0/8 + * unicast-ip4-chain + * [@0]: dpo-load-balance: [index:3 buckets:1 uRPF:3 to:[0:0]] + * [0] [@0]: dpo-drop ip6 + * 6.0.1.2/32 + * unicast-ip4-chain + * [@0]: dpo-load-balance: [index:30 buckets:1 uRPF:29 to:[0:0]] + * [0] [@3]: arp-ipv4: via 6.0.0.1 af_packet0 + * 7.0.0.1/32 + * unicast-ip4-chain + * [@0]: dpo-load-balance: [index:31 buckets:4 uRPF:30 to:[0:0]] + * [0] [@3]: arp-ipv4: via 6.0.0.2 af_packet0 + * [1] [@3]: arp-ipv4: via 6.0.0.2 af_packet0 + * [2] [@3]: arp-ipv4: via 6.0.0.2 af_packet0 + * [3] [@3]: arp-ipv4: via 6.0.0.1 af_packet0 + * 240.0.0.0/8 + * unicast-ip4-chain + * [@0]: dpo-load-balance: [index:2 buckets:1 uRPF:2 to:[0:0]] + * [0] [@0]: dpo-drop ip6 + * 255.255.255.255/32 + * unicast-ip4-chain + * [@0]: dpo-load-balance: [index:4 buckets:1 uRPF:4 to:[0:0]] + * [0] [@0]: dpo-drop ip6 + * ipv4-VRF:7, fib_index 1, flow hash: dst sport dport proto + * 0.0.0.0/0 + * unicast-ip4-chain + * [@0]: dpo-load-balance: [index:12 buckets:1 uRPF:11 to:[0:0]] + * [0] [@0]: dpo-drop ip6 + * 0.0.0.0/32 + * unicast-ip4-chain + * [@0]: dpo-load-balance: [index:13 buckets:1 uRPF:12 to:[0:0]] + * [0] [@0]: dpo-drop ip6 + * 172.16.1.0/24 + * unicast-ip4-chain + * [@0]: dpo-load-balance: [index:17 buckets:1 uRPF:16 to:[0:0]] + * [0] [@4]: ipv4-glean: af_packet0 + * 172.16.1.1/32 + * unicast-ip4-chain + * [@0]: dpo-load-balance: [index:18 buckets:1 uRPF:17 to:[1:84]] + * [0] [@2]: dpo-receive: 172.16.1.1 on af_packet0 + * 172.16.1.2/32 + * unicast-ip4-chain + * [@0]: dpo-load-balance: [index:21 buckets:1 uRPF:20 to:[0:0]] + * [0] [@5]: ipv4 via 172.16.1.2 af_packet0: IP4: 02:fe:9e:70:7a:2b -> 26:a5:f6:9c:3a:36 + * 172.16.2.0/24 + * unicast-ip4-chain + * [@0]: dpo-load-balance: [index:19 buckets:1 uRPF:18 to:[0:0]] + * [0] [@4]: ipv4-glean: af_packet1 + * 172.16.2.1/32 + * unicast-ip4-chain + * [@0]: dpo-load-balance: [index:20 buckets:1 uRPF:19 to:[0:0]] + * [0] [@2]: dpo-receive: 172.16.2.1 on af_packet1 + * 224.0.0.0/8 + * unicast-ip4-chain + * [@0]: dpo-load-balance: [index:15 buckets:1 uRPF:14 to:[0:0]] + * [0] [@0]: dpo-drop ip6 + * 240.0.0.0/8 + * unicast-ip4-chain + * [@0]: dpo-load-balance: [index:14 buckets:1 uRPF:13 to:[0:0]] + * [0] [@0]: dpo-drop ip6 + * 255.255.255.255/32 + * unicast-ip4-chain + * [@0]: dpo-load-balance: [index:16 buckets:1 uRPF:15 to:[0:0]] + * [0] [@0]: dpo-drop ip6 + * @cliexend +?*/ +/* *INDENT-OFF* */ +VLIB_CLI_COMMAND (set_ip_flow_hash_command, static) = +{ + .path = "set ip flow-hash", + .short_help = + "set ip flow-hash table [src] [dst] [sport] [dport] [proto] [reverse]", + .function = set_ip_flow_hash_command_fn, +}; +/* *INDENT-ON* */ + +int +vnet_set_ip4_classify_intfc (vlib_main_t * vm, u32 sw_if_index, + u32 table_index) +{ + vnet_main_t *vnm = vnet_get_main (); + vnet_interface_main_t *im = &vnm->interface_main; + ip4_main_t *ipm = &ip4_main; + ip_lookup_main_t *lm = &ipm->lookup_main; + vnet_classify_main_t *cm = &vnet_classify_main; + ip4_address_t *if_addr; + + if (pool_is_free_index (im->sw_interfaces, sw_if_index)) + return VNET_API_ERROR_NO_MATCHING_INTERFACE; + + if (table_index != ~0 && pool_is_free_index (cm->tables, table_index)) + return VNET_API_ERROR_NO_SUCH_ENTRY; + + vec_validate (lm->classify_table_index_by_sw_if_index, sw_if_index); + lm->classify_table_index_by_sw_if_index[sw_if_index] = table_index; + + if_addr = ip4_interface_first_address (ipm, sw_if_index, NULL); + + if (NULL != if_addr) + { + fib_prefix_t pfx = { + .fp_len = 32, + .fp_proto = FIB_PROTOCOL_IP4, + .fp_addr.ip4 = *if_addr, + }; + u32 fib_index; + + fib_index = fib_table_get_index_for_sw_if_index (FIB_PROTOCOL_IP4, + sw_if_index); + + + if (table_index != (u32) ~ 0) + { + dpo_id_t dpo = DPO_INVALID; + + dpo_set (&dpo, + DPO_CLASSIFY, + DPO_PROTO_IP4, + classify_dpo_create (DPO_PROTO_IP4, table_index)); + + fib_table_entry_special_dpo_add (fib_index, + &pfx, + FIB_SOURCE_CLASSIFY, + FIB_ENTRY_FLAG_NONE, &dpo); + dpo_reset (&dpo); + } + else + { + fib_table_entry_special_remove (fib_index, + &pfx, FIB_SOURCE_CLASSIFY); + } + } + + return 0; +} + +static clib_error_t * +set_ip_classify_command_fn (vlib_main_t * vm, + unformat_input_t * input, + vlib_cli_command_t * cmd) +{ + u32 table_index = ~0; + int table_index_set = 0; + u32 sw_if_index = ~0; + int rv; + + while (unformat_check_input (input) != UNFORMAT_END_OF_INPUT) + { + if (unformat (input, "table-index %d", &table_index)) + table_index_set = 1; + else if (unformat (input, "intfc %U", unformat_vnet_sw_interface, + vnet_get_main (), &sw_if_index)) + ; + else + break; + } + + if (table_index_set == 0) + return clib_error_return (0, "classify table-index must be specified"); + + if (sw_if_index == ~0) + return clib_error_return (0, "interface / subif must be specified"); + + rv = vnet_set_ip4_classify_intfc (vm, sw_if_index, table_index); + + 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; +} + +/*? + * Assign a classification table to an interface. The classification + * table is created using the 'classify table' and 'classify session' + * commands. Once the table is create, use this command to filter packets + * on an interface. + * + * @cliexpar + * Example of how to assign a classification table to an interface: + * @cliexcmd{set ip classify intfc GigabitEthernet2/0/0 table-index 1} +?*/ +/* *INDENT-OFF* */ +VLIB_CLI_COMMAND (set_ip_classify_command, static) = +{ + .path = "set ip classify", + .short_help = + "set ip classify intfc table-index ", + .function = set_ip_classify_command_fn, +}; +/* *INDENT-ON* */ + +/* + * fd.io coding-style-patch-verification: ON + * + * Local Variables: + * eval: (c-set-style "gnu") + * End: + */ -- cgit 1.2.3-korg From 044183faacc5eb8a055658f9deebefb56b254adc Mon Sep 17 00:00:00 2001 From: Neale Ranns Date: Tue, 24 Jan 2017 01:34:25 -0800 Subject: [re]Enable per-Adjacency/neighbour counters Change-Id: I953b3888bbc6d8a5f53f684a5edc8742b382f323 Signed-off-by: Neale Ranns --- src/scripts/vnet/ip6 | 18 +- src/vat/api_format.c | 124 ++++++++++++++ src/vat/vat.h | 18 ++ src/vlib/counter.h | 14 ++ src/vnet/adj/adj.c | 4 + src/vnet/adj/adj_l2.c | 10 +- src/vnet/ip/ip4_forward.c | 42 +++-- src/vnet/ip/ip6_forward.c | 24 +-- src/vnet/mpls/mpls_output.c | 37 ++-- src/vpp/api/vpe.api | 41 +++++ src/vpp/stats/stats.c | 405 +++++++++++++++++++++++++++++++++++++++++++- 11 files changed, 675 insertions(+), 62 deletions(-) (limited to 'src/vnet/ip/ip4_forward.c') diff --git a/src/scripts/vnet/ip6 b/src/scripts/vnet/ip6 index 4f9f3ee5..adb27225 100644 --- a/src/scripts/vnet/ip6 +++ b/src/scripts/vnet/ip6 @@ -6,10 +6,24 @@ packet-generator new { no-recycle data { IP6: 1.2.3 -> 4.5.6 - ICMP: ::1 -> ::2 + ICMP: 3002::2 -> 3001::2 ICMP echo_request incrementing 100 } } -tr add pg-input 100 + +loop create +loop create +set int state loop0 up +set int state loop1 up + +set int ip address loop0 2001:1::1/64 +set int ip address loop1 2001:2::1/64 + +set ip6 neighbor loop0 2001:1::2 00:00:DD:EE:AA:DD +set ip6 neighbor loop1 2001:2::2 00:00:DD:EE:AA:EE + +ip route add 3001::/64 via 2001:2::2 loop1 + +trace add pg-input 100 diff --git a/src/vat/api_format.c b/src/vat/api_format.c index 653cf79f..839bcdaa 100644 --- a/src/vat/api_format.c +++ b/src/vat/api_format.c @@ -2042,6 +2042,42 @@ static void vl_api_vnet_ip4_fib_counters_t_handler_json } } +static void vl_api_vnet_ip4_nbr_counters_t_handler + (vl_api_vnet_ip4_nbr_counters_t * mp) +{ + /* not supported */ +} + +static void vl_api_vnet_ip4_nbr_counters_t_handler_json + (vl_api_vnet_ip4_nbr_counters_t * mp) +{ + vat_main_t *vam = &vat_main; + vl_api_ip4_nbr_counter_t *v; + ip4_nbr_counter_t *counter; + u32 sw_if_index; + u32 count; + int i; + + sw_if_index = ntohl (mp->sw_if_index); + count = ntohl (mp->count); + vec_validate (vam->ip4_nbr_counters, sw_if_index); + + if (mp->begin) + vec_free (vam->ip4_nbr_counters[sw_if_index]); + + v = (vl_api_ip4_nbr_counter_t *) & mp->c; + for (i = 0; i < count; i++) + { + vec_validate (vam->ip4_nbr_counters[sw_if_index], i); + counter = &vam->ip4_nbr_counters[sw_if_index][i]; + counter->address.s_addr = v->address; + counter->packets = clib_net_to_host_u64 (v->packets); + counter->bytes = clib_net_to_host_u64 (v->bytes); + counter->linkt = v->link_type; + v++; + } +} + static void vl_api_vnet_ip6_fib_counters_t_handler (vl_api_vnet_ip6_fib_counters_t * mp) { @@ -2087,6 +2123,43 @@ static void vl_api_vnet_ip6_fib_counters_t_handler_json } } +static void vl_api_vnet_ip6_nbr_counters_t_handler + (vl_api_vnet_ip6_nbr_counters_t * mp) +{ + /* not supported */ +} + +static void vl_api_vnet_ip6_nbr_counters_t_handler_json + (vl_api_vnet_ip6_nbr_counters_t * mp) +{ + vat_main_t *vam = &vat_main; + vl_api_ip6_nbr_counter_t *v; + ip6_nbr_counter_t *counter; + struct in6_addr ip6; + u32 sw_if_index; + u32 count; + int i; + + sw_if_index = ntohl (mp->sw_if_index); + count = ntohl (mp->count); + vec_validate (vam->ip6_nbr_counters, sw_if_index); + + if (mp->begin) + vec_free (vam->ip6_nbr_counters[sw_if_index]); + + v = (vl_api_ip6_nbr_counter_t *) & mp->c; + for (i = 0; i < count; i++) + { + vec_validate (vam->ip6_nbr_counters[sw_if_index], i); + counter = &vam->ip6_nbr_counters[sw_if_index][i]; + clib_memcpy (&ip6, &v->address, sizeof (ip6)); + counter->address = ip6; + counter->packets = clib_net_to_host_u64 (v->packets); + counter->bytes = clib_net_to_host_u64 (v->bytes); + v++; + } +} + static void vl_api_get_first_msg_id_reply_t_handler (vl_api_get_first_msg_id_reply_t * mp) { @@ -3490,6 +3563,10 @@ static void vl_api_flow_classify_details_t_handler_json #define vl_api_vnet_ip4_fib_counters_t_print vl_noop_handler #define vl_api_vnet_ip6_fib_counters_t_endian vl_noop_handler #define vl_api_vnet_ip6_fib_counters_t_print vl_noop_handler +#define vl_api_vnet_ip4_nbr_counters_t_endian vl_noop_handler +#define vl_api_vnet_ip4_nbr_counters_t_print vl_noop_handler +#define vl_api_vnet_ip6_nbr_counters_t_endian vl_noop_handler +#define vl_api_vnet_ip6_nbr_counters_t_print vl_noop_handler #define vl_api_lisp_adjacencies_get_reply_t_endian vl_noop_handler #define vl_api_lisp_adjacencies_get_reply_t_print vl_noop_handler @@ -3799,6 +3876,8 @@ _(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) \ @@ -4131,6 +4210,8 @@ dump_stats_table (vat_main_t * vam) u64 packets; ip4_fib_counter_t *c4; ip6_fib_counter_t *c6; + ip4_nbr_counter_t *n4; + ip6_nbr_counter_t *n6; int i, j; if (!vam->json_output) @@ -4226,6 +4307,49 @@ dump_stats_table (vat_main_t * vam) } } + /* ip4 nbr counters */ + msg_array = vat_json_object_add (&node, "ip4_nbr_counters"); + vat_json_init_array (msg_array); + for (i = 0; i < vec_len (vam->ip4_nbr_counters); i++) + { + msg = vat_json_array_add (msg_array); + vat_json_init_object (msg); + vat_json_object_add_uint (msg, "sw_if_index", i); + counter_array = vat_json_object_add (msg, "c"); + vat_json_init_array (counter_array); + for (j = 0; j < vec_len (vam->ip4_nbr_counters[i]); j++) + { + counter = vat_json_array_add (counter_array); + vat_json_init_object (counter); + n4 = &vam->ip4_nbr_counters[i][j]; + vat_json_object_add_ip4 (counter, "address", n4->address); + vat_json_object_add_uint (counter, "link-type", n4->linkt); + vat_json_object_add_uint (counter, "packets", n4->packets); + vat_json_object_add_uint (counter, "bytes", n4->bytes); + } + } + + /* ip6 nbr counters */ + msg_array = vat_json_object_add (&node, "ip6_nbr_counters"); + vat_json_init_array (msg_array); + for (i = 0; i < vec_len (vam->ip6_nbr_counters); i++) + { + msg = vat_json_array_add (msg_array); + vat_json_init_object (msg); + vat_json_object_add_uint (msg, "sw_if_index", i); + counter_array = vat_json_object_add (msg, "c"); + vat_json_init_array (counter_array); + for (j = 0; j < vec_len (vam->ip6_nbr_counters[i]); j++) + { + counter = vat_json_array_add (counter_array); + vat_json_init_object (counter); + n6 = &vam->ip6_nbr_counters[i][j]; + vat_json_object_add_ip6 (counter, "address", n6->address); + vat_json_object_add_uint (counter, "packets", n6->packets); + vat_json_object_add_uint (counter, "bytes", n6->bytes); + } + } + vat_json_print (vam->ofp, &node); vat_json_free (&node); diff --git a/src/vat/vat.h b/src/vat/vat.h index 3d7d96ae..831bdf50 100644 --- a/src/vat/vat.h +++ b/src/vat/vat.h @@ -95,6 +95,22 @@ typedef struct u64 bytes; } ip6_fib_counter_t; +typedef struct +{ + struct in_addr address; + vnet_link_t linkt; + u64 packets; + u64 bytes; +} ip4_nbr_counter_t; + +typedef struct +{ + struct in6_addr address; + vnet_link_t linkt; + u64 packets; + u64 bytes; +} ip6_nbr_counter_t; + typedef struct { /* vpe input queue */ @@ -185,6 +201,8 @@ typedef struct u32 *ip4_fib_counters_vrf_id_by_index; ip6_fib_counter_t **ip6_fib_counters; u32 *ip6_fib_counters_vrf_id_by_index; + ip4_nbr_counter_t **ip4_nbr_counters; + ip6_nbr_counter_t **ip6_nbr_counters; /* Convenience */ vlib_main_t *vlib_main; diff --git a/src/vlib/counter.h b/src/vlib/counter.h index a7903206..abfa89ee 100644 --- a/src/vlib/counter.h +++ b/src/vlib/counter.h @@ -273,6 +273,20 @@ vlib_increment_combined_counter (vlib_combined_counter_main_t * cm, } } +#define vlib_prefetch_combined_counter(_cm, _cpu_index, _index) \ +{ \ + vlib_mini_counter_t *_cpu_minis; \ + \ + /* \ + * This CPU's mini index is assumed to already be in cache \ + */ \ + _cpu_minis = (_cm)->minis[(_cpu_index)]; \ + CLIB_PREFETCH(_cpu_minis + (_index), \ + sizeof(*_cpu_minis), \ + STORE); \ +} + + /** Get the value of a combined counter, never called in the speed path Scrapes the entire set of mini counters. Innacurate unless worker threads which might increment the counter are diff --git a/src/vnet/adj/adj.c b/src/vnet/adj/adj.c index e740c4cb..d0be0f0e 100644 --- a/src/vnet/adj/adj.c +++ b/src/vnet/adj/adj.c @@ -122,6 +122,10 @@ format_ip_adjacency (u8 * s, va_list * args) if (fiaf & FORMAT_IP_ADJACENCY_DETAIL) { + vlib_counter_t counts; + + vlib_get_combined_counter(&adjacency_counters, adj_index, &counts); + s = format (s, "\n counts:[%Ld:%Ld]", counts.packets, counts.bytes); s = format (s, "\n locks:%d", adj->ia_node.fn_locks); s = format (s, " node:[%d]:%U", adj->rewrite_header.node_index, diff --git a/src/vnet/adj/adj_l2.c b/src/vnet/adj/adj_l2.c index 4d2dd708..5a083643 100644 --- a/src/vnet/adj/adj_l2.c +++ b/src/vnet/adj/adj_l2.c @@ -94,11 +94,11 @@ adj_l2_rewrite_inline (vlib_main_t * vm, rw_len0 = adj0[0].rewrite_header.data_bytes; vnet_buffer(p0)->ip.save_rewrite_length = rw_len0; - vlib_increment_combined_counter - (&adjacency_counters, - cpu_index, adj_index0, - /* packet increment */ 0, - /* byte increment */ rw_len0-sizeof(ethernet_header_t)); + vlib_increment_combined_counter(&adjacency_counters, + cpu_index, + adj_index0, + /* packet increment */ 0, + /* byte increment */ rw_len0); /* Check MTU of outgoing interface. */ if (PREDICT_TRUE((vlib_buffer_length_in_chain (vm, p0) <= diff --git a/src/vnet/ip/ip4_forward.c b/src/vnet/ip/ip4_forward.c index 6e91b9e9..87b345bd 100644 --- a/src/vnet/ip/ip4_forward.c +++ b/src/vnet/ip/ip4_forward.c @@ -2402,19 +2402,12 @@ ip4_rewrite_inline (vlib_main_t * vm, error1); /* - * We've already accounted for an ethernet_header_t elsewhere + * pre-fetch the per-adjacency counters */ - if (PREDICT_FALSE (rw_len0 > sizeof (ethernet_header_t))) - vlib_increment_combined_counter - (&adjacency_counters, cpu_index, adj_index0, - /* packet increment */ 0, - /* byte increment */ rw_len0 - sizeof (ethernet_header_t)); - - if (PREDICT_FALSE (rw_len1 > sizeof (ethernet_header_t))) - vlib_increment_combined_counter - (&adjacency_counters, cpu_index, adj_index1, - /* packet increment */ 0, - /* byte increment */ rw_len1 - sizeof (ethernet_header_t)); + vlib_prefetch_combined_counter (&adjacency_counters, + cpu_index, adj_index0); + vlib_prefetch_combined_counter (&adjacency_counters, + cpu_index, adj_index1); /* Don't adjust the buffer for ttl issue; icmp-error node wants * to see the IP headerr */ @@ -2446,6 +2439,19 @@ ip4_rewrite_inline (vlib_main_t * vm, vnet_rewrite_two_headers (adj0[0], adj1[0], ip0, ip1, sizeof (ethernet_header_t)); + /* + * Bump the per-adjacency counters + */ + vlib_increment_combined_counter + (&adjacency_counters, + cpu_index, + adj_index0, 1, vlib_buffer_length_in_chain (vm, p0) + rw_len0); + + vlib_increment_combined_counter + (&adjacency_counters, + cpu_index, + adj_index1, 1, vlib_buffer_length_in_chain (vm, p1) + rw_len1); + if (is_midchain) { adj0->sub_type.midchain.fixup_func (vm, adj0, p0); @@ -2519,6 +2525,9 @@ ip4_rewrite_inline (vlib_main_t * vm, p0->flags &= ~VNET_BUFFER_LOCALLY_ORIGINATED; } + vlib_prefetch_combined_counter (&adjacency_counters, + cpu_index, adj_index0); + /* Guess we are only writing on simple Ethernet header. */ vnet_rewrite_one_header (adj0[0], ip0, sizeof (ethernet_header_t)); @@ -2526,11 +2535,10 @@ ip4_rewrite_inline (vlib_main_t * vm, rw_len0 = adj0[0].rewrite_header.data_bytes; vnet_buffer (p0)->ip.save_rewrite_length = rw_len0; - if (PREDICT_FALSE (rw_len0 > sizeof (ethernet_header_t))) - vlib_increment_combined_counter - (&adjacency_counters, cpu_index, adj_index0, - /* packet increment */ 0, - /* byte increment */ rw_len0 - sizeof (ethernet_header_t)); + vlib_increment_combined_counter + (&adjacency_counters, + cpu_index, + adj_index0, 1, vlib_buffer_length_in_chain (vm, p0) + rw_len0); /* Check MTU of outgoing interface. */ error0 = (vlib_buffer_length_in_chain (vm, p0) diff --git a/src/vnet/ip/ip6_forward.c b/src/vnet/ip/ip6_forward.c index 197a9b79..232f7283 100644 --- a/src/vnet/ip/ip6_forward.c +++ b/src/vnet/ip/ip6_forward.c @@ -2108,14 +2108,14 @@ ip6_rewrite_inline (vlib_main_t * vm, vnet_buffer (p0)->ip.save_rewrite_length = rw_len0; vnet_buffer (p1)->ip.save_rewrite_length = rw_len1; - vlib_increment_combined_counter (&adjacency_counters, - cpu_index, adj_index0, - /* packet increment */ 0, - /* byte increment */ rw_len0); - vlib_increment_combined_counter (&adjacency_counters, - cpu_index, adj_index1, - /* packet increment */ 0, - /* byte increment */ rw_len1); + vlib_increment_combined_counter + (&adjacency_counters, + cpu_index, + adj_index0, 1, vlib_buffer_length_in_chain (vm, p0) + rw_len0); + vlib_increment_combined_counter + (&adjacency_counters, + cpu_index, adj_index1, + 1, vlib_buffer_length_in_chain (vm, p1) + rw_len1); /* Check MTU of outgoing interface. */ error0 = @@ -2233,10 +2233,10 @@ ip6_rewrite_inline (vlib_main_t * vm, rw_len0 = adj0[0].rewrite_header.data_bytes; vnet_buffer (p0)->ip.save_rewrite_length = rw_len0; - vlib_increment_combined_counter (&adjacency_counters, - cpu_index, adj_index0, - /* packet increment */ 0, - /* byte increment */ rw_len0); + vlib_increment_combined_counter + (&adjacency_counters, + cpu_index, + adj_index0, 1, vlib_buffer_length_in_chain (vm, p0) + rw_len0); /* Check MTU of outgoing interface. */ error0 = diff --git a/src/vnet/mpls/mpls_output.c b/src/vnet/mpls/mpls_output.c index 8292a0cb..c06cf917 100644 --- a/src/vnet/mpls/mpls_output.c +++ b/src/vnet/mpls/mpls_output.c @@ -128,18 +128,19 @@ mpls_output_inline (vlib_main_t * vm, rw_len0 = adj0[0].rewrite_header.data_bytes; rw_len1 = adj1[0].rewrite_header.data_bytes; - if (PREDICT_FALSE (rw_len0 > sizeof(ethernet_header_t))) - vlib_increment_combined_counter - (&adjacency_counters, - cpu_index, adj_index0, - /* packet increment */ 0, - /* byte increment */ rw_len0-sizeof(ethernet_header_t)); - if (PREDICT_FALSE (rw_len1 > sizeof(ethernet_header_t))) - vlib_increment_combined_counter - (&adjacency_counters, - cpu_index, adj_index1, - /* packet increment */ 0, - /* byte increment */ rw_len1-sizeof(ethernet_header_t)); + /* Bump the adj counters for packet and bytes */ + vlib_increment_combined_counter + (&adjacency_counters, + cpu_index, + adj_index0, + 1, + vlib_buffer_length_in_chain (vm, p0) + rw_len0); + vlib_increment_combined_counter + (&adjacency_counters, + cpu_index, + adj_index1, + 1, + vlib_buffer_length_in_chain (vm, p1) + rw_len1); /* Check MTU of outgoing interface. */ if (PREDICT_TRUE(vlib_buffer_length_in_chain (vm, p0) <= @@ -234,12 +235,12 @@ mpls_output_inline (vlib_main_t * vm, /* Update packet buffer attributes/set output interface. */ rw_len0 = adj0[0].rewrite_header.data_bytes; - if (PREDICT_FALSE (rw_len0 > sizeof(ethernet_header_t))) - vlib_increment_combined_counter - (&adjacency_counters, - cpu_index, adj_index0, - /* packet increment */ 0, - /* byte increment */ rw_len0-sizeof(ethernet_header_t)); + vlib_increment_combined_counter + (&adjacency_counters, + cpu_index, + adj_index0, + 1, + vlib_buffer_length_in_chain (vm, p0) + rw_len0); /* Check MTU of outgoing interface. */ if (PREDICT_TRUE(vlib_buffer_length_in_chain (vm, p0) <= diff --git a/src/vpp/api/vpe.api b/src/vpp/api/vpe.api index f32ba670..a00033c5 100644 --- a/src/vpp/api/vpe.api +++ b/src/vpp/api/vpe.api @@ -230,6 +230,31 @@ manual_print manual_endian define vnet_ip4_fib_counters vl_api_ip4_fib_counter_t c[count]; }; +typeonly manual_print manual_endian define ip4_nbr_counter +{ + u32 address; + u8 link_type; + u64 packets; + u64 bytes; +}; + +/** + * @brief Per-neighbour (i.e. per-adjacency) coutners + * @param count The size of the array of counters + * @param sw_if_index The interface the adjacency is on + * @param begin Flag to indicate this is the first set of stats for this + * interface. If this flag is not set the it is a continuation of + * stats for this interface + * @param c counters + */ +manual_print manual_endian define vnet_ip4_nbr_counters +{ + u32 count; + u32 sw_if_index; + u8 begin; + vl_api_ip4_nbr_counter_t c[count]; +}; + typeonly manual_print manual_endian define ip6_fib_counter { u64 address[2]; @@ -245,6 +270,22 @@ manual_print manual_endian define vnet_ip6_fib_counters vl_api_ip6_fib_counter_t c[count]; }; +typeonly manual_print manual_endian define ip6_nbr_counter +{ + u64 address[2]; + u8 link_type; + u64 packets; + u64 bytes; +}; + +manual_print manual_endian define vnet_ip6_nbr_counters +{ + u32 count; + u32 sw_if_index; + u8 begin; + vl_api_ip6_nbr_counter_t c[count]; +}; + /** \brief Request for a single block of summary stats @param client_index - opaque cookie to identify the sender @param context - sender context, to match reply w/ request diff --git a/src/vpp/stats/stats.c b/src/vpp/stats/stats.c index 391e02f6..5e9b0d69 100644 --- a/src/vpp/stats/stats.c +++ b/src/vpp/stats/stats.c @@ -49,7 +49,9 @@ _(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) +_(VNET_IP6_FIB_COUNTERS, vnet_ip6_fib_counters) \ +_(VNET_IP4_NBR_COUNTERS, vnet_ip4_nbr_counters) \ +_(VNET_IP6_NBR_COUNTERS, vnet_ip6_nbr_counters) /* These constants ensure msg sizes <= 1024, aka ring allocation */ #define SIMPLE_COUNTER_BATCH_SIZE 126 @@ -258,6 +260,313 @@ ip46_fib_stats_delay (stats_main_t * sm, u32 sec, u32 nsec) } } +/** + * @brief The context passed when collecting adjacency counters + */ +typedef struct ip4_nbr_stats_ctx_t_ +{ + /** + * The SW IF index all these adjs belong to + */ + u32 sw_if_index; + + /** + * A vector of ip4 nbr counters + */ + vl_api_ip4_nbr_counter_t *counters; +} ip4_nbr_stats_ctx_t; + +static adj_walk_rc_t +ip4_nbr_stats_cb (adj_index_t ai, void *arg) +{ + vl_api_ip4_nbr_counter_t *vl_counter; + vlib_counter_t adj_counter; + ip4_nbr_stats_ctx_t *ctx; + ip_adjacency_t *adj; + + ctx = arg; + vlib_get_combined_counter (&adjacency_counters, ai, &adj_counter); + + if (0 != adj_counter.packets) + { + vec_add2 (ctx->counters, vl_counter, 1); + adj = adj_get (ai); + + vl_counter->packets = clib_host_to_net_u64 (adj_counter.packets); + vl_counter->bytes = clib_host_to_net_u64 (adj_counter.bytes); + vl_counter->address = adj->sub_type.nbr.next_hop.ip4.as_u32; + vl_counter->link_type = adj->ia_link; + } + return (ADJ_WALK_RC_CONTINUE); +} + +#define MIN(x,y) (((x)<(y))?(x):(y)) + +static void +ip4_nbr_ship (stats_main_t * sm, ip4_nbr_stats_ctx_t * ctx) +{ + api_main_t *am = sm->api_main; + vl_shmem_hdr_t *shmem_hdr = am->shmem_hdr; + unix_shared_memory_queue_t *q = shmem_hdr->vl_input_queue; + vl_api_vnet_ip4_nbr_counters_t *mp = 0; + int first = 0; + + /* + * If the walk context has counters, which may be left over from the last + * suspend, then we continue from there. + */ + while (0 != vec_len (ctx->counters)) + { + u32 n_items = MIN (vec_len (ctx->counters), + IP4_FIB_COUNTER_BATCH_SIZE); + u8 pause = 0; + + dslock (sm, 0 /* release hint */ , 1 /* tag */ ); + + mp = vl_msg_api_alloc_as_if_client (sizeof (*mp) + + (n_items * + sizeof + (vl_api_ip4_nbr_counter_t))); + mp->_vl_msg_id = ntohs (VL_API_VNET_IP4_NBR_COUNTERS); + mp->count = ntohl (n_items); + mp->sw_if_index = ntohl (ctx->sw_if_index); + mp->begin = first; + first = 0; + + /* + * copy the counters from the back of the context, then we can easily + * 'erase' them by resetting the vector length. + * The order we push the stats to the caller is not important. + */ + clib_memcpy (mp->c, + &ctx->counters[vec_len (ctx->counters) - n_items], + n_items * sizeof (*ctx->counters)); + + _vec_len (ctx->counters) = vec_len (ctx->counters) - n_items; + + /* + * send to the shm q + */ + unix_shared_memory_queue_lock (q); + pause = unix_shared_memory_queue_is_full (q); + + vl_msg_api_send_shmem_nolock (q, (u8 *) & mp); + unix_shared_memory_queue_unlock (q); + dsunlock (sm); + + if (pause) + ip46_fib_stats_delay (sm, 0 /* sec */ , + STATS_RELEASE_DELAY_NS); + } +} + +static void +do_ip4_nbrs (stats_main_t * sm) +{ + vnet_main_t *vnm = vnet_get_main (); + vnet_interface_main_t *im = &vnm->interface_main; + vnet_sw_interface_t *si; + + ip4_nbr_stats_ctx_t ctx = { + .sw_if_index = 0, + .counters = NULL, + }; + + /* *INDENT-OFF* */ + pool_foreach (si, im->sw_interfaces, + ({ + /* + * update the interface we are now concerned with + */ + ctx.sw_if_index = si->sw_if_index; + + /* + * we are about to walk another interface, so we shouldn't have any pending + * stats to export. + */ + ASSERT(ctx.counters == NULL); + + /* + * visit each neighbour adjacency on the interface and collect + * its current stats. + * Because we hold the lock the walk is synchronous, so safe to routing + * updates. It's limited in work by the number of adjacenies on an + * interface, which is typically not huge. + */ + dslock (sm, 0 /* release hint */ , 1 /* tag */ ); + adj_nbr_walk (si->sw_if_index, + FIB_PROTOCOL_IP4, + ip4_nbr_stats_cb, + &ctx); + dsunlock (sm); + + /* + * if this interface has some adjacencies with counters then ship them, + * else continue to the next interface. + */ + if (NULL != ctx.counters) + { + ip4_nbr_ship(sm, &ctx); + } + })); + /* *INDENT-OFF* */ +} + +/** + * @brief The context passed when collecting adjacency counters + */ +typedef struct ip6_nbr_stats_ctx_t_ +{ + /** + * The SW IF index all these adjs belong to + */ + u32 sw_if_index; + + /** + * A vector of ip6 nbr counters + */ + vl_api_ip6_nbr_counter_t *counters; +} ip6_nbr_stats_ctx_t; + +static adj_walk_rc_t +ip6_nbr_stats_cb (adj_index_t ai, + void *arg) +{ + vl_api_ip6_nbr_counter_t *vl_counter; + vlib_counter_t adj_counter; + ip6_nbr_stats_ctx_t *ctx; + ip_adjacency_t *adj; + + ctx = arg; + vlib_get_combined_counter(&adjacency_counters, ai, &adj_counter); + + if (0 != adj_counter.packets) + { + vec_add2(ctx->counters, vl_counter, 1); + adj = adj_get(ai); + + vl_counter->packets = clib_host_to_net_u64(adj_counter.packets); + vl_counter->bytes = clib_host_to_net_u64(adj_counter.bytes); + vl_counter->address[0] = adj->sub_type.nbr.next_hop.ip6.as_u64[0]; + vl_counter->address[1] = adj->sub_type.nbr.next_hop.ip6.as_u64[1]; + vl_counter->link_type = adj->ia_link; + } + return (ADJ_WALK_RC_CONTINUE); +} + +#define MIN(x,y) (((x)<(y))?(x):(y)) + +static void +ip6_nbr_ship (stats_main_t * sm, + ip6_nbr_stats_ctx_t *ctx) +{ + api_main_t *am = sm->api_main; + vl_shmem_hdr_t *shmem_hdr = am->shmem_hdr; + unix_shared_memory_queue_t *q = shmem_hdr->vl_input_queue; + vl_api_vnet_ip6_nbr_counters_t *mp = 0; + int first = 0; + + /* + * If the walk context has counters, which may be left over from the last + * suspend, then we continue from there. + */ + while (0 != vec_len(ctx->counters)) + { + u32 n_items = MIN (vec_len (ctx->counters), + IP6_FIB_COUNTER_BATCH_SIZE); + u8 pause = 0; + + dslock (sm, 0 /* release hint */ , 1 /* tag */ ); + + mp = vl_msg_api_alloc_as_if_client (sizeof (*mp) + + (n_items * + sizeof + (vl_api_ip6_nbr_counter_t))); + mp->_vl_msg_id = ntohs (VL_API_VNET_IP6_NBR_COUNTERS); + mp->count = ntohl (n_items); + mp->sw_if_index = ntohl (ctx->sw_if_index); + mp->begin = first; + first = 0; + + /* + * copy the counters from the back of the context, then we can easily + * 'erase' them by resetting the vector length. + * The order we push the stats to the caller is not important. + */ + clib_memcpy (mp->c, + &ctx->counters[vec_len (ctx->counters) - n_items], + n_items * sizeof (*ctx->counters)); + + _vec_len (ctx->counters) = vec_len (ctx->counters) - n_items; + + /* + * send to the shm q + */ + unix_shared_memory_queue_lock (q); + pause = unix_shared_memory_queue_is_full (q); + + vl_msg_api_send_shmem_nolock (q, (u8 *) & mp); + unix_shared_memory_queue_unlock (q); + dsunlock (sm); + + if (pause) + ip46_fib_stats_delay (sm, 0 /* sec */ , + STATS_RELEASE_DELAY_NS); + } +} + +static void +do_ip6_nbrs (stats_main_t * sm) +{ + vnet_main_t *vnm = vnet_get_main (); + vnet_interface_main_t *im = &vnm->interface_main; + vnet_sw_interface_t *si; + + ip6_nbr_stats_ctx_t ctx = { + .sw_if_index = 0, + .counters = NULL, + }; + + /* *INDENT-OFF* */ + pool_foreach (si, im->sw_interfaces, + ({ + /* + * update the interface we are now concerned with + */ + ctx.sw_if_index = si->sw_if_index; + + /* + * we are about to walk another interface, so we shouldn't have any pending + * stats to export. + */ + ASSERT(ctx.counters == NULL); + + /* + * visit each neighbour adjacency on the interface and collect + * its current stats. + * Because we hold the lock the walk is synchronous, so safe to routing + * updates. It's limited in work by the number of adjacenies on an + * interface, which is typically not huge. + */ + dslock (sm, 0 /* release hint */ , 1 /* tag */ ); + adj_nbr_walk (si->sw_if_index, + FIB_PROTOCOL_IP6, + ip6_nbr_stats_cb, + &ctx); + dsunlock (sm); + + /* + * if this interface has some adjacencies with counters then ship them, + * else continue to the next interface. + */ + if (NULL != ctx.counters) + { + ip6_nbr_ship(sm, &ctx); + } + })); + /* *INDENT-OFF* */ +} + static void do_ip4_fibs (stats_main_t * sm) { @@ -318,13 +627,7 @@ again: hash_foreach_pair (p, hash, ({ x.address.data_u32 = p->key; - if (lm->fib_result_n_words > 1) - { - x.index = vec_len (results); - vec_add (results, p->value, lm->fib_result_n_words); - } - else - x.index = p->value[0]; + x.index = p->value[0]; vec_add1 (routes, x); if (sm->data_structure_lock->release_hint) @@ -631,6 +934,8 @@ stats_thread_fn (void *arg) do_combined_interface_counters (sm); do_ip4_fibs (sm); do_ip6_fibs (sm); + do_ip4_nbrs (sm); + do_ip6_nbrs (sm); } } @@ -804,6 +1109,45 @@ vl_api_vnet_ip4_fib_counters_t_handler (vl_api_vnet_ip4_fib_counters_t * mp) } } +static void +vl_api_vnet_ip4_nbr_counters_t_handler (vl_api_vnet_ip4_nbr_counters_t * mp) +{ + vpe_client_registration_t *reg; + stats_main_t *sm = &stats_main; + unix_shared_memory_queue_t *q, *q_prev = NULL; + vl_api_vnet_ip4_nbr_counters_t *mp_copy = NULL; + u32 mp_size; + + mp_size = sizeof (*mp_copy) + + ntohl (mp->count) * sizeof (vl_api_ip4_nbr_counter_t); + + /* *INDENT-OFF* */ + pool_foreach(reg, sm->stats_registrations, + ({ + q = vl_api_client_index_to_input_queue (reg->client_index); + if (q) + { + if (q_prev && (q_prev->cursize < q_prev->maxsize)) + { + mp_copy = vl_msg_api_alloc_as_if_client(mp_size); + clib_memcpy(mp_copy, mp, mp_size); + vl_msg_api_send_shmem (q_prev, (u8 *)&mp); + mp = mp_copy; + } + q_prev = q; + } + })); + /* *INDENT-ON* */ + if (q_prev && (q_prev->cursize < q_prev->maxsize)) + { + vl_msg_api_send_shmem (q_prev, (u8 *) & mp); + } + else + { + vl_msg_api_free (mp); + } +} + static void vl_api_vnet_ip6_fib_counters_t_handler (vl_api_vnet_ip6_fib_counters_t * mp) { @@ -843,6 +1187,45 @@ vl_api_vnet_ip6_fib_counters_t_handler (vl_api_vnet_ip6_fib_counters_t * mp) } } +static void +vl_api_vnet_ip6_nbr_counters_t_handler (vl_api_vnet_ip6_nbr_counters_t * mp) +{ + vpe_client_registration_t *reg; + stats_main_t *sm = &stats_main; + unix_shared_memory_queue_t *q, *q_prev = NULL; + vl_api_vnet_ip6_nbr_counters_t *mp_copy = NULL; + u32 mp_size; + + mp_size = sizeof (*mp_copy) + + ntohl (mp->count) * sizeof (vl_api_ip6_nbr_counter_t); + + /* *INDENT-OFF* */ + pool_foreach(reg, sm->stats_registrations, + ({ + q = vl_api_client_index_to_input_queue (reg->client_index); + if (q) + { + if (q_prev && (q_prev->cursize < q_prev->maxsize)) + { + mp_copy = vl_msg_api_alloc_as_if_client(mp_size); + clib_memcpy(mp_copy, mp, mp_size); + vl_msg_api_send_shmem (q_prev, (u8 *)&mp); + mp = mp_copy; + } + q_prev = q; + } + })); + /* *INDENT-ON* */ + if (q_prev && (q_prev->cursize < q_prev->maxsize)) + { + vl_msg_api_send_shmem (q_prev, (u8 *) & mp); + } + else + { + vl_msg_api_free (mp); + } +} + static void vl_api_want_stats_reply_t_handler (vl_api_want_stats_reply_t * mp) { @@ -929,6 +1312,10 @@ stats_memclnt_delete_callback (u32 client_index) #define vl_api_vnet_ip4_fib_counters_t_print vl_noop_handler #define vl_api_vnet_ip6_fib_counters_t_endian vl_noop_handler #define vl_api_vnet_ip6_fib_counters_t_print vl_noop_handler +#define vl_api_vnet_ip4_nbr_counters_t_endian vl_noop_handler +#define vl_api_vnet_ip4_nbr_counters_t_print vl_noop_handler +#define vl_api_vnet_ip6_nbr_counters_t_endian vl_noop_handler +#define vl_api_vnet_ip6_nbr_counters_t_print vl_noop_handler static clib_error_t * stats_init (vlib_main_t * vm) @@ -961,6 +1348,8 @@ stats_init (vlib_main_t * vm) am->message_bounce[VL_API_VNET_INTERFACE_COUNTERS] = 1; am->message_bounce[VL_API_VNET_IP4_FIB_COUNTERS] = 1; am->message_bounce[VL_API_VNET_IP6_FIB_COUNTERS] = 1; + am->message_bounce[VL_API_VNET_IP4_NBR_COUNTERS] = 1; + am->message_bounce[VL_API_VNET_IP6_NBR_COUNTERS] = 1; return 0; } -- cgit 1.2.3-korg From 32e1c010b0c34fd0984f7fc45fae648a182025c5 Mon Sep 17 00:00:00 2001 From: Neale Ranns Date: Tue, 22 Nov 2016 17:07:28 +0000 Subject: IP Multicast FIB (mfib) - IPv[46] mfib tables with support for (*,G/m), (*,G) and (S,G) exact and longest prefix match - Replication represented via a new replicate DPO. - RPF configuration and data-plane checking - data-plane signals sent to listening control planes. The functions of multicast forwarding entries differ from their unicast conterparts, so we introduce a new mfib_table_t and mfib_entry_t objects. However, we re-use the fib_path_list to resolve and build the entry's output list. the fib_path_list provides the service to construct a replicate DPO for multicast. 'make tests' is added to with two new suites; TEST=mfib, this is invocation of the CLI command 'test mfib' which deals with many path add/remove, flag set/unset scenarios, TEST=ip-mcast, data-plane forwarding tests. Updated applications to use the new MIFB functions; - IPv6 NS/RA. - DHCPv6 unit tests for these are undated accordingly. Change-Id: I49ec37b01f1b170335a5697541c8fd30e6d3a961 Signed-off-by: Neale Ranns --- src/scripts/vnet/mcast/ip4 | 22 + src/vat/api_format.c | 173 +++++ src/vnet.am | 35 +- src/vnet/adj/adj.c | 11 +- src/vnet/adj/adj.h | 1 - src/vnet/adj/adj_internal.h | 2 + src/vnet/adj/adj_mcast.c | 346 ++++++++++ src/vnet/adj/adj_mcast.h | 78 +++ src/vnet/adj/adj_nbr.c | 2 +- src/vnet/adj/adj_rewrite.c | 53 -- src/vnet/adj/adj_rewrite.h | 49 -- src/vnet/dhcpv6/proxy_node.c | 46 +- src/vnet/dpo/dpo.c | 2 + src/vnet/dpo/dpo.h | 8 +- src/vnet/dpo/load_balance.c | 13 +- src/vnet/dpo/load_balance.h | 8 + src/vnet/dpo/replicate_dpo.c | 759 ++++++++++++++++++++++ src/vnet/dpo/replicate_dpo.h | 143 +++++ src/vnet/ethernet/arp.c | 84 ++- src/vnet/ethernet/ethernet.h | 2 + src/vnet/ethernet/interface.c | 20 + src/vnet/fib/fib_attached_export.c | 4 +- src/vnet/fib/fib_entry.h | 2 +- src/vnet/fib/fib_entry_delegate.c | 3 + src/vnet/fib/fib_entry_src.c | 4 + src/vnet/fib/fib_node.h | 2 + src/vnet/fib/fib_path.c | 91 ++- src/vnet/fib/fib_path_list.c | 9 +- src/vnet/fib/fib_path_list.h | 6 + src/vnet/fib/fib_table.c | 69 +- src/vnet/fib/fib_table.h | 16 + src/vnet/fib/fib_test.c | 207 +++--- src/vnet/fib/fib_types.c | 4 + src/vnet/fib/fib_types.h | 14 + src/vnet/fib/fib_urpf_list.c | 20 +- src/vnet/fib/ip4_fib.c | 50 +- src/vnet/fib/ip4_fib.h | 9 + src/vnet/fib/ip6_fib.c | 117 ++-- src/vnet/fib/ip6_fib.h | 11 +- src/vnet/fib/mpls_fib.c | 17 +- src/vnet/fib/mpls_fib.h | 9 + src/vnet/ip/ip.api | 53 ++ src/vnet/ip/ip4.h | 24 + src/vnet/ip/ip4_forward.c | 498 +++++---------- src/vnet/ip/ip4_input.c | 4 +- src/vnet/ip/ip6.h | 27 + src/vnet/ip/ip6_forward.c | 83 ++- src/vnet/ip/ip6_input.c | 43 +- src/vnet/ip/ip6_neighbor.c | 134 ++-- src/vnet/ip/ip_api.c | 210 +++++++ src/vnet/ip/lookup.c | 171 +++++ src/vnet/ip/lookup.h | 82 +-- src/vnet/mcast/mcast.c | 565 ----------------- src/vnet/mcast/mcast.h | 50 -- src/vnet/mcast/mcast_test.c | 149 ----- src/vnet/mfib/ip4_mfib.c | 465 ++++++++++++++ src/vnet/mfib/ip4_mfib.h | 95 +++ src/vnet/mfib/ip6_mfib.c | 663 +++++++++++++++++++ src/vnet/mfib/ip6_mfib.h | 109 ++++ src/vnet/mfib/mfib_entry.c | 1096 ++++++++++++++++++++++++++++++++ src/vnet/mfib/mfib_entry.h | 172 +++++ src/vnet/mfib/mfib_forward.c | 512 +++++++++++++++ src/vnet/mfib/mfib_itf.c | 119 ++++ src/vnet/mfib/mfib_itf.h | 63 ++ src/vnet/mfib/mfib_signal.c | 201 ++++++ src/vnet/mfib/mfib_signal.h | 59 ++ src/vnet/mfib/mfib_table.c | 489 ++++++++++++++ src/vnet/mfib/mfib_table.h | 331 ++++++++++ src/vnet/mfib/mfib_test.c | 1225 ++++++++++++++++++++++++++++++++++++ src/vnet/mfib/mfib_types.c | 213 +++++++ src/vnet/mfib/mfib_types.h | 185 ++++++ src/vnet/misc.c | 3 + src/vnet/rewrite.h | 31 + src/vnet/sr/sr.c | 4 +- src/vnet/util/radix.c | 1104 ++++++++++++++++++++++++++++++++ src/vnet/util/radix.h | 147 +++++ src/vnet/vxlan/vxlan.c | 112 +++- src/vpp/api/api.c | 14 +- src/vppinfra.am | 2 +- src/vppinfra/dlist.h | 2 +- src/vppinfra/format.c | 8 +- src/vppinfra/format.h | 4 +- src/vppinfra/unformat.c | 16 +- test/test_dhcp.py | 16 - test/test_ip6.py | 131 ++-- test/test_ip_mcast.py | 612 ++++++++++++++++++ test/test_mfib.py | 23 + test/vpp_interface.py | 3 +- test/vpp_ip_route.py | 101 ++- test/vpp_papi_provider.py | 34 +- 90 files changed, 11211 insertions(+), 1767 deletions(-) create mode 100644 src/scripts/vnet/mcast/ip4 create mode 100644 src/vnet/adj/adj_mcast.c create mode 100644 src/vnet/adj/adj_mcast.h delete mode 100644 src/vnet/adj/adj_rewrite.c delete mode 100644 src/vnet/adj/adj_rewrite.h create mode 100644 src/vnet/dpo/replicate_dpo.c create mode 100644 src/vnet/dpo/replicate_dpo.h delete mode 100644 src/vnet/mcast/mcast.c delete mode 100644 src/vnet/mcast/mcast.h delete mode 100644 src/vnet/mcast/mcast_test.c create mode 100644 src/vnet/mfib/ip4_mfib.c create mode 100644 src/vnet/mfib/ip4_mfib.h create mode 100644 src/vnet/mfib/ip6_mfib.c create mode 100644 src/vnet/mfib/ip6_mfib.h create mode 100644 src/vnet/mfib/mfib_entry.c create mode 100644 src/vnet/mfib/mfib_entry.h create mode 100644 src/vnet/mfib/mfib_forward.c create mode 100644 src/vnet/mfib/mfib_itf.c create mode 100644 src/vnet/mfib/mfib_itf.h create mode 100644 src/vnet/mfib/mfib_signal.c create mode 100644 src/vnet/mfib/mfib_signal.h create mode 100644 src/vnet/mfib/mfib_table.c create mode 100644 src/vnet/mfib/mfib_table.h create mode 100644 src/vnet/mfib/mfib_test.c create mode 100644 src/vnet/mfib/mfib_types.c create mode 100644 src/vnet/mfib/mfib_types.h create mode 100644 src/vnet/util/radix.c create mode 100644 src/vnet/util/radix.h create mode 100644 test/test_ip_mcast.py create mode 100644 test/test_mfib.py (limited to 'src/vnet/ip/ip4_forward.c') diff --git a/src/scripts/vnet/mcast/ip4 b/src/scripts/vnet/mcast/ip4 new file mode 100644 index 00000000..69f1ee00 --- /dev/null +++ b/src/scripts/vnet/mcast/ip4 @@ -0,0 +1,22 @@ +packet-generator new { + name x + limit 1 + node ip4-input + size 64-64 + no-recycle + data { + ICMP: 1.0.0.2 -> 232.1.1.1 + ICMP echo_request + incrementing 100 + } +} + +trace add pg-input 100 +loop create +loop create +set int state loop0 up +set int state loop1 up + +ip mroute add 232.1.1.1 via pg0 Accept +ip mroute add 232.1.1.1 via loop0 Forward +ip mroute add 232.1.1.1 via loop1 Forward diff --git a/src/vat/api_format.c b/src/vat/api_format.c index b83313de..4cfe4a58 100644 --- a/src/vat/api_format.c +++ b/src/vat/api_format.c @@ -48,6 +48,7 @@ #include #include #include +#include #include "vat/json_format.h" @@ -505,6 +506,53 @@ unformat_flow_classify_table_type (unformat_input_t * input, va_list * va) return 1; } +static const char *mfib_flag_names[] = MFIB_ENTRY_NAMES_SHORT; +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; + +uword +unformat_mfib_itf_flags (unformat_input_t * input, va_list * args) +{ + mfib_itf_flags_t old, *iflags = va_arg (*args, mfib_itf_flags_t *); + mfib_itf_attribute_t attr; + + old = *iflags; + FOR_EACH_MFIB_ITF_ATTRIBUTE (attr) + { + if (unformat (input, mfib_itf_flag_long_names[attr])) + *iflags |= (1 << attr); + } + FOR_EACH_MFIB_ITF_ATTRIBUTE (attr) + { + if (unformat (input, mfib_itf_flag_names[attr])) + *iflags |= (1 << attr); + } + + return (old == *iflags ? 0 : 1); +} + +uword +unformat_mfib_entry_flags (unformat_input_t * input, va_list * args) +{ + mfib_entry_flags_t old, *eflags = va_arg (*args, mfib_entry_flags_t *); + mfib_entry_attribute_t attr; + + old = *eflags; + FOR_EACH_MFIB_ATTRIBUTE (attr) + { + if (unformat (input, mfib_flag_long_names[attr])) + *eflags |= (1 << attr); + } + FOR_EACH_MFIB_ATTRIBUTE (attr) + { + if (unformat (input, mfib_flag_names[attr])) + *eflags |= (1 << attr); + } + + return (old == *eflags ? 0 : 1); +} + #if (VPP_API_TEST_BUILTIN==0) u8 * format_ip4_address (u8 * s, va_list * args) @@ -3592,6 +3640,7 @@ _(bridge_domain_add_del_reply) \ _(sw_interface_set_l2_xconnect_reply) \ _(l2fib_add_del_reply) \ _(ip_add_del_route_reply) \ +_(ip_mroute_add_del_reply) \ _(mpls_route_add_del_reply) \ _(mpls_ip_bind_unbind_reply) \ _(proxy_arp_add_del_reply) \ @@ -3792,6 +3841,7 @@ _(TAP_MODIFY_REPLY, tap_modify_reply) \ _(TAP_DELETE_REPLY, tap_delete_reply) \ _(SW_INTERFACE_TAP_DETAILS, sw_interface_tap_details) \ _(IP_ADD_DEL_ROUTE_REPLY, ip_add_del_route_reply) \ +_(IP_MROUTE_ADD_DEL_REPLY, ip_mroute_add_del_reply) \ _(MPLS_ROUTE_ADD_DEL_REPLY, mpls_route_add_del_reply) \ _(MPLS_IP_BIND_UNBIND_REPLY, mpls_ip_bind_unbind_reply) \ _(PROXY_ARP_ADD_DEL_REPLY, proxy_arp_add_del_reply) \ @@ -6383,6 +6433,126 @@ api_ip_add_del_route (vat_main_t * vam) return (vam->retval); } +static int +api_ip_mroute_add_del (vat_main_t * vam) +{ + unformat_input_t *i = vam->input; + vl_api_ip_mroute_add_del_t *mp; + f64 timeout; + u32 sw_if_index = ~0, vrf_id = 0; + u8 is_ipv6 = 0; + u8 is_local = 0; + u8 create_vrf_if_needed = 0; + u8 is_add = 1; + u8 address_set = 0; + u32 grp_address_length = 0; + ip4_address_t v4_grp_address, v4_src_address; + ip6_address_t v6_grp_address, v6_src_address; + mfib_itf_flags_t iflags = 0; + mfib_entry_flags_t eflags = 0; + + /* Parse args required to build the message */ + while (unformat_check_input (i) != UNFORMAT_END_OF_INPUT) + { + if (unformat (i, "sw_if_index %d", &sw_if_index)) + ; + else if (unformat (i, "%U %U", + unformat_ip4_address, &v4_src_address, + unformat_ip4_address, &v4_grp_address)) + { + grp_address_length = 64; + address_set = 1; + is_ipv6 = 0; + } + else if (unformat (i, "%U %U", + unformat_ip6_address, &v6_src_address, + unformat_ip6_address, &v6_grp_address)) + { + grp_address_length = 256; + address_set = 1; + is_ipv6 = 1; + } + else if (unformat (i, "%U", unformat_ip4_address, &v4_grp_address)) + { + memset (&v4_src_address, 0, sizeof (v4_src_address)); + grp_address_length = 32; + address_set = 1; + is_ipv6 = 0; + } + else if (unformat (i, "%U", unformat_ip6_address, &v6_grp_address)) + { + memset (&v6_src_address, 0, sizeof (v6_src_address)); + grp_address_length = 128; + address_set = 1; + is_ipv6 = 1; + } + else if (unformat (i, "/%d", &grp_address_length)) + ; + else if (unformat (i, "local")) + { + is_local = 1; + } + else if (unformat (i, "del")) + is_add = 0; + else if (unformat (i, "add")) + is_add = 1; + else if (unformat (i, "vrf %d", &vrf_id)) + ; + else if (unformat (i, "create-vrf")) + create_vrf_if_needed = 1; + else if (unformat (i, "%U", unformat_mfib_itf_flags, &iflags)) + ; + else if (unformat (i, "%U", unformat_mfib_entry_flags, &eflags)) + ; + else + { + clib_warning ("parse error '%U'", format_unformat_error, i); + return -99; + } + } + + if (address_set == 0) + { + errmsg ("missing addresses\n"); + return -99; + } + + /* Construct the API message */ + M (IP_MROUTE_ADD_DEL, ip_mroute_add_del); + + mp->next_hop_sw_if_index = ntohl (sw_if_index); + mp->table_id = ntohl (vrf_id); + mp->create_vrf_if_needed = create_vrf_if_needed; + + mp->is_add = is_add; + mp->is_ipv6 = is_ipv6; + mp->is_local = is_local; + mp->itf_flags = ntohl (iflags); + mp->entry_flags = ntohl (eflags); + mp->grp_address_length = grp_address_length; + mp->grp_address_length = ntohs (mp->grp_address_length); + + if (is_ipv6) + { + clib_memcpy (mp->grp_address, &v6_grp_address, sizeof (v6_grp_address)); + clib_memcpy (mp->src_address, &v6_src_address, sizeof (v6_src_address)); + } + else + { + clib_memcpy (mp->grp_address, &v4_grp_address, sizeof (v4_grp_address)); + clib_memcpy (mp->src_address, &v4_src_address, sizeof (v4_src_address)); + + } + + /* send it... */ + S; + /* Wait for a reply... */ + W; + + /* Return the good/bad news */ + return (vam->retval); +} + static int api_mpls_route_add_del (vat_main_t * vam) { @@ -17512,6 +17682,9 @@ _(ip_add_del_route, \ "[ | sw_if_index ] [resolve-attempts ]\n" \ "[weight ] [drop] [local] [classify ] [del]\n" \ "[multipath] [count ]") \ +_(ip_mroute_add_del, \ + " / [table-id ]\n" \ + "[ | sw_if_index ] [local] [del]") \ _(mpls_route_add_del, \ "