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/vxlan-gpe/decap.c | 733 ++++++++++++++++++++++++++++ src/vnet/vxlan-gpe/dir.dox | 32 ++ src/vnet/vxlan-gpe/encap.c | 388 +++++++++++++++ src/vnet/vxlan-gpe/vxlan-gpe-rfc.txt | 868 +++++++++++++++++++++++++++++++++ src/vnet/vxlan-gpe/vxlan_gpe.api | 61 +++ src/vnet/vxlan-gpe/vxlan_gpe.c | 659 +++++++++++++++++++++++++ src/vnet/vxlan-gpe/vxlan_gpe.h | 221 +++++++++ src/vnet/vxlan-gpe/vxlan_gpe_api.c | 249 ++++++++++ src/vnet/vxlan-gpe/vxlan_gpe_error.def | 16 + src/vnet/vxlan-gpe/vxlan_gpe_packet.h | 110 +++++ 10 files changed, 3337 insertions(+) create mode 100644 src/vnet/vxlan-gpe/decap.c create mode 100644 src/vnet/vxlan-gpe/dir.dox create mode 100644 src/vnet/vxlan-gpe/encap.c create mode 100644 src/vnet/vxlan-gpe/vxlan-gpe-rfc.txt create mode 100644 src/vnet/vxlan-gpe/vxlan_gpe.api create mode 100644 src/vnet/vxlan-gpe/vxlan_gpe.c create mode 100644 src/vnet/vxlan-gpe/vxlan_gpe.h create mode 100644 src/vnet/vxlan-gpe/vxlan_gpe_api.c create mode 100644 src/vnet/vxlan-gpe/vxlan_gpe_error.def create mode 100644 src/vnet/vxlan-gpe/vxlan_gpe_packet.h (limited to 'src/vnet/vxlan-gpe') diff --git a/src/vnet/vxlan-gpe/decap.c b/src/vnet/vxlan-gpe/decap.c new file mode 100644 index 00000000..22ab4b62 --- /dev/null +++ b/src/vnet/vxlan-gpe/decap.c @@ -0,0 +1,733 @@ +/* + * decap.c - decapsulate VXLAN GPE + * + * Copyright (c) 2013 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ +/** + * @file + * @brief Functions for decapsulating VXLAN GPE tunnels + * +*/ + +#include +#include +#include + +vlib_node_registration_t vxlan_gpe_input_node; + +/** + * @brief Struct for VXLAN GPE decap packet tracing + * + */ +typedef struct { + u32 next_index; + u32 tunnel_index; + u32 error; +} vxlan_gpe_rx_trace_t; + +/** + * @brief Tracing function for VXLAN GPE packet decapsulation + * + * @param *s + * @param *args + * + * @return *s + * + */ +static u8 * format_vxlan_gpe_rx_trace (u8 * s, va_list * args) +{ + CLIB_UNUSED (vlib_main_t * vm) = va_arg (*args, vlib_main_t *); + CLIB_UNUSED (vlib_node_t * node) = va_arg (*args, vlib_node_t *); + vxlan_gpe_rx_trace_t * t = va_arg (*args, vxlan_gpe_rx_trace_t *); + + if (t->tunnel_index != ~0) + { + s = format (s, "VXLAN-GPE: tunnel %d next %d error %d", t->tunnel_index, + t->next_index, t->error); + } + else + { + s = format (s, "VXLAN-GPE: no tunnel next %d error %d\n", t->next_index, + t->error); + } + return s; +} + +/** + * @brief Tracing function for VXLAN GPE packet decapsulation including length + * + * @param *s + * @param *args + * + * @return *s + * + */ +static u8 * format_vxlan_gpe_with_length (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 *); + + + return s; +} + +/** + * @brief Common processing for IPv4 and IPv6 VXLAN GPE decap dispatch functions + * + * It is worth noting that other than trivial UDP forwarding (transit), VXLAN GPE + * tunnels are "terminate local". This means that there is no "TX" interface for this + * decap case, so that field in the buffer_metadata can be "used for something else". + * The something else in this case is, for the IPv4/IPv6 inner-packet type case, the + * FIB index used to look up the inner-packet's adjacency. + * + * vnet_buffer(b0)->sw_if_index[VLIB_TX] = t0->decap_fib_index; + * + * @param *vm + * @param *node + * @param *from_frame + * @param is_ip4 + * + * @return from_frame->n_vectors + * + */ +always_inline uword +vxlan_gpe_input (vlib_main_t * vm, + vlib_node_runtime_t * node, + vlib_frame_t * from_frame, + u8 is_ip4) +{ + u32 n_left_from, next_index, *from, *to_next; + vxlan_gpe_main_t * ngm = &vxlan_gpe_main; + vnet_main_t * vnm = ngm->vnet_main; + vnet_interface_main_t * im = &vnm->interface_main; + u32 last_tunnel_index = ~0; + vxlan4_gpe_tunnel_key_t last_key4; + vxlan6_gpe_tunnel_key_t last_key6; + u32 pkts_decapsulated = 0; + u32 cpu_index = os_get_cpu_number (); + u32 stats_sw_if_index, stats_n_packets, stats_n_bytes; + + if (is_ip4) + memset (&last_key4, 0xff, sizeof(last_key4)); + else + memset (&last_key6, 0xff, sizeof(last_key6)); + + from = vlib_frame_vector_args (from_frame); + n_left_from = from_frame->n_vectors; + + next_index = node->cached_next_index; + stats_sw_if_index = node->runtime_data[0]; + stats_n_packets = stats_n_bytes = 0; + + while (n_left_from > 0) + { + u32 n_left_to_next; + + vlib_get_next_frame(vm, node, next_index, to_next, n_left_to_next); + + while (n_left_from >= 4 && n_left_to_next >= 2) + { + u32 bi0, bi1; + vlib_buffer_t * b0, *b1; + u32 next0, next1; + ip4_vxlan_gpe_header_t * iuvn4_0, *iuvn4_1; + ip6_vxlan_gpe_header_t * iuvn6_0, *iuvn6_1; + uword * p0, *p1; + u32 tunnel_index0, tunnel_index1; + vxlan_gpe_tunnel_t * t0, *t1; + vxlan4_gpe_tunnel_key_t key4_0, key4_1; + vxlan6_gpe_tunnel_key_t key6_0, key6_1; + u32 error0, error1; + u32 sw_if_index0, sw_if_index1, len0, len1; + + /* Prefetch next iteration. */ + { + vlib_buffer_t * p2, *p3; + + p2 = vlib_get_buffer (vm, from[2]); + p3 = vlib_get_buffer (vm, from[3]); + + vlib_prefetch_buffer_header(p2, LOAD); + vlib_prefetch_buffer_header(p3, LOAD); + + CLIB_PREFETCH(p2->data, 2*CLIB_CACHE_LINE_BYTES, LOAD); + CLIB_PREFETCH(p3->data, 2*CLIB_CACHE_LINE_BYTES, LOAD); + } + + bi0 = from[0]; + bi1 = from[1]; + to_next[0] = bi0; + to_next[1] = bi1; + from += 2; + to_next += 2; + n_left_to_next -= 2; + n_left_from -= 2; + + b0 = vlib_get_buffer (vm, bi0); + b1 = vlib_get_buffer (vm, bi1); + + if (is_ip4) + { + /* udp leaves current_data pointing at the vxlan-gpe header */ + vlib_buffer_advance (b0, -(word) (sizeof(udp_header_t) + sizeof(ip4_header_t))); + vlib_buffer_advance (b1, -(word) (sizeof(udp_header_t) + sizeof(ip4_header_t))); + + iuvn4_0 = vlib_buffer_get_current (b0); + iuvn4_1 = vlib_buffer_get_current (b1); + + /* pop (ip, udp, vxlan) */ + vlib_buffer_advance (b0, sizeof(*iuvn4_0)); + vlib_buffer_advance (b1, sizeof(*iuvn4_1)); + } + else + { + /* udp leaves current_data pointing at the vxlan-gpe header */ + vlib_buffer_advance (b0, -(word) (sizeof(udp_header_t) + sizeof(ip6_header_t))); + vlib_buffer_advance (b1, -(word) (sizeof(udp_header_t) + sizeof(ip6_header_t))); + + iuvn6_0 = vlib_buffer_get_current (b0); + iuvn6_1 = vlib_buffer_get_current (b1); + + /* pop (ip, udp, vxlan) */ + vlib_buffer_advance (b0, sizeof(*iuvn6_0)); + vlib_buffer_advance (b1, sizeof(*iuvn6_1)); + } + + tunnel_index0 = ~0; + tunnel_index1 = ~0; + error0 = 0; + error1 = 0; + + if (is_ip4) + { + next0 = + (iuvn4_0->vxlan.protocol < VXLAN_GPE_PROTOCOL_MAX)? + ngm->decap_next_node_list[iuvn4_0->vxlan.protocol]: \ + VXLAN_GPE_INPUT_NEXT_DROP; + next1 = + (iuvn4_1->vxlan.protocol < VXLAN_GPE_PROTOCOL_MAX)? + ngm->decap_next_node_list[iuvn4_1->vxlan.protocol]: \ + VXLAN_GPE_INPUT_NEXT_DROP; + + key4_0.local = iuvn4_0->ip4.dst_address.as_u32; + key4_1.local = iuvn4_1->ip4.dst_address.as_u32; + + key4_0.remote = iuvn4_0->ip4.src_address.as_u32; + key4_1.remote = iuvn4_1->ip4.src_address.as_u32; + + key4_0.vni = iuvn4_0->vxlan.vni_res; + key4_1.vni = iuvn4_1->vxlan.vni_res; + + key4_0.pad = 0; + key4_1.pad = 0; + } + else /* is_ip6 */ + { + next0 = (iuvn6_0->vxlan.protocol < node->n_next_nodes) ? + iuvn6_0->vxlan.protocol : VXLAN_GPE_INPUT_NEXT_DROP; + next1 = (iuvn6_1->vxlan.protocol < node->n_next_nodes) ? + iuvn6_1->vxlan.protocol : VXLAN_GPE_INPUT_NEXT_DROP; + + key6_0.local.as_u64[0] = iuvn6_0->ip6.dst_address.as_u64[0]; + key6_0.local.as_u64[1] = iuvn6_0->ip6.dst_address.as_u64[1]; + key6_1.local.as_u64[0] = iuvn6_1->ip6.dst_address.as_u64[0]; + key6_1.local.as_u64[1] = iuvn6_1->ip6.dst_address.as_u64[1]; + + key6_0.remote.as_u64[0] = iuvn6_0->ip6.src_address.as_u64[0]; + key6_0.remote.as_u64[1] = iuvn6_0->ip6.src_address.as_u64[1]; + key6_1.remote.as_u64[0] = iuvn6_1->ip6.src_address.as_u64[0]; + key6_1.remote.as_u64[1] = iuvn6_1->ip6.src_address.as_u64[1]; + + key6_0.vni = iuvn6_0->vxlan.vni_res; + key6_1.vni = iuvn6_1->vxlan.vni_res; + } + + /* Processing packet 0*/ + if (is_ip4) + { + /* Processing for key4_0 */ + if (PREDICT_FALSE((key4_0.as_u64[0] != last_key4.as_u64[0]) + || (key4_0.as_u64[1] != last_key4.as_u64[1]))) + { + p0 = hash_get_mem(ngm->vxlan4_gpe_tunnel_by_key, &key4_0); + + if (p0 == 0) + { + error0 = VXLAN_GPE_ERROR_NO_SUCH_TUNNEL; + goto trace0; + } + + last_key4.as_u64[0] = key4_0.as_u64[0]; + last_key4.as_u64[1] = key4_0.as_u64[1]; + tunnel_index0 = last_tunnel_index = p0[0]; + } + else + tunnel_index0 = last_tunnel_index; + } + else /* is_ip6 */ + { + next0 = + (iuvn6_0->vxlan.protocol < VXLAN_GPE_PROTOCOL_MAX)? + ngm->decap_next_node_list[iuvn6_0->vxlan.protocol]: \ + VXLAN_GPE_INPUT_NEXT_DROP; + next1 = + (iuvn6_1->vxlan.protocol < VXLAN_GPE_PROTOCOL_MAX)? + ngm->decap_next_node_list[iuvn6_1->vxlan.protocol]: \ + VXLAN_GPE_INPUT_NEXT_DROP; + + key6_0.local.as_u64[0] = iuvn6_0->ip6.dst_address.as_u64[0]; + key6_0.local.as_u64[1] = iuvn6_0->ip6.dst_address.as_u64[1]; + key6_1.local.as_u64[0] = iuvn6_1->ip6.dst_address.as_u64[0]; + key6_1.local.as_u64[1] = iuvn6_1->ip6.dst_address.as_u64[1]; + + key6_0.remote.as_u64[0] = iuvn6_0->ip6.src_address.as_u64[0]; + key6_0.remote.as_u64[1] = iuvn6_0->ip6.src_address.as_u64[1]; + key6_1.remote.as_u64[0] = iuvn6_1->ip6.src_address.as_u64[0]; + key6_1.remote.as_u64[1] = iuvn6_1->ip6.src_address.as_u64[1]; + + key6_0.vni = iuvn6_0->vxlan.vni_res; + key6_1.vni = iuvn6_1->vxlan.vni_res; + + /* Processing for key6_0 */ + if (PREDICT_FALSE(memcmp (&key6_0, &last_key6, sizeof(last_key6)) != 0)) + { + p0 = hash_get_mem(ngm->vxlan6_gpe_tunnel_by_key, &key6_0); + + if (p0 == 0) + { + error0 = VXLAN_GPE_ERROR_NO_SUCH_TUNNEL; + goto trace0; + } + + memcpy (&last_key6, &key6_0, sizeof(key6_0)); + tunnel_index0 = last_tunnel_index = p0[0]; + } + else + tunnel_index0 = last_tunnel_index; + } + + t0 = pool_elt_at_index(ngm->tunnels, tunnel_index0); + + + sw_if_index0 = t0->sw_if_index; + len0 = vlib_buffer_length_in_chain (vm, b0); + + /* Required to make the l2 tag push / pop code work on l2 subifs */ + vnet_update_l2_len (b0); + + /** + * ip[46] lookup in the configured FIB + */ + vnet_buffer(b0)->sw_if_index[VLIB_TX] = t0->decap_fib_index; + + pkts_decapsulated++; + stats_n_packets += 1; + stats_n_bytes += len0; + + if (PREDICT_FALSE(sw_if_index0 != stats_sw_if_index)) + { + stats_n_packets -= 1; + stats_n_bytes -= len0; + if (stats_n_packets) + vlib_increment_combined_counter ( + im->combined_sw_if_counters + VNET_INTERFACE_COUNTER_RX, + cpu_index, stats_sw_if_index, stats_n_packets, stats_n_bytes); + stats_n_packets = 1; + stats_n_bytes = len0; + stats_sw_if_index = sw_if_index0; + } + + trace0: b0->error = error0 ? node->errors[error0] : 0; + + if (PREDICT_FALSE(b0->flags & VLIB_BUFFER_IS_TRACED)) + { + vxlan_gpe_rx_trace_t *tr = vlib_add_trace (vm, node, b0, sizeof(*tr)); + tr->next_index = next0; + tr->error = error0; + tr->tunnel_index = tunnel_index0; + } + + /* Process packet 1 */ + if (is_ip4) + { + /* Processing for key4_1 */ + if (PREDICT_FALSE( + (key4_1.as_u64[0] != last_key4.as_u64[0]) + || (key4_1.as_u64[1] != last_key4.as_u64[1]))) + { + p1 = hash_get_mem(ngm->vxlan4_gpe_tunnel_by_key, &key4_1); + + if (p1 == 0) + { + error1 = VXLAN_GPE_ERROR_NO_SUCH_TUNNEL; + goto trace1; + } + + last_key4.as_u64[0] = key4_1.as_u64[0]; + last_key4.as_u64[1] = key4_1.as_u64[1]; + tunnel_index1 = last_tunnel_index = p1[0]; + } + else + tunnel_index1 = last_tunnel_index; + } + else /* is_ip6 */ + { + /* Processing for key6_1 */ + if (PREDICT_FALSE(memcmp (&key6_1, &last_key6, sizeof(last_key6)) != 0)) + { + p1 = hash_get_mem(ngm->vxlan6_gpe_tunnel_by_key, &key6_1); + + if (p1 == 0) + { + error1 = VXLAN_GPE_ERROR_NO_SUCH_TUNNEL; + goto trace1; + } + + memcpy (&last_key6, &key6_1, sizeof(key6_1)); + tunnel_index1 = last_tunnel_index = p1[0]; + } + else + tunnel_index1 = last_tunnel_index; + } + + t1 = pool_elt_at_index(ngm->tunnels, tunnel_index1); + + sw_if_index1 = t1->sw_if_index; + len1 = vlib_buffer_length_in_chain (vm, b1); + + /* Required to make the l2 tag push / pop code work on l2 subifs */ + vnet_update_l2_len (b1); + + /* + * ip[46] lookup in the configured FIB + */ + vnet_buffer(b1)->sw_if_index[VLIB_TX] = t1->decap_fib_index; + + pkts_decapsulated++; + stats_n_packets += 1; + stats_n_bytes += len1; + + /* Batch stats increment on the same vxlan tunnel so counter + is not incremented per packet */ + if (PREDICT_FALSE(sw_if_index1 != stats_sw_if_index)) + { + stats_n_packets -= 1; + stats_n_bytes -= len1; + if (stats_n_packets) + vlib_increment_combined_counter ( + im->combined_sw_if_counters + VNET_INTERFACE_COUNTER_RX, + cpu_index, stats_sw_if_index, stats_n_packets, stats_n_bytes); + stats_n_packets = 1; + stats_n_bytes = len1; + stats_sw_if_index = sw_if_index1; + } + vnet_buffer(b1)->sw_if_index[VLIB_TX] = t1->decap_fib_index; + + trace1: b1->error = error1 ? node->errors[error1] : 0; + + if (PREDICT_FALSE(b1->flags & VLIB_BUFFER_IS_TRACED)) + { + vxlan_gpe_rx_trace_t *tr = vlib_add_trace (vm, node, b1, sizeof(*tr)); + tr->next_index = next1; + tr->error = error1; + tr->tunnel_index = tunnel_index1; + } + + vlib_validate_buffer_enqueue_x2(vm, node, next_index, to_next, + n_left_to_next, bi0, bi1, next0, next1); + } + + while (n_left_from > 0 && n_left_to_next > 0) + { + u32 bi0; + vlib_buffer_t * b0; + u32 next0; + ip4_vxlan_gpe_header_t * iuvn4_0; + ip6_vxlan_gpe_header_t * iuvn6_0; + uword * p0; + u32 tunnel_index0; + vxlan_gpe_tunnel_t * t0; + vxlan4_gpe_tunnel_key_t key4_0; + vxlan6_gpe_tunnel_key_t key6_0; + u32 error0; + u32 sw_if_index0, len0; + + bi0 = from[0]; + to_next[0] = bi0; + from += 1; + to_next += 1; + n_left_from -= 1; + n_left_to_next -= 1; + + b0 = vlib_get_buffer (vm, bi0); + + if (is_ip4) + { + /* udp leaves current_data pointing at the vxlan-gpe header */ + vlib_buffer_advance ( + b0, -(word) (sizeof(udp_header_t) + sizeof(ip4_header_t))); + + iuvn4_0 = vlib_buffer_get_current (b0); + + /* pop (ip, udp, vxlan) */ + vlib_buffer_advance (b0, sizeof(*iuvn4_0)); + } + else + { + /* udp leaves current_data pointing at the vxlan-gpe header */ + vlib_buffer_advance ( + b0, -(word) (sizeof(udp_header_t) + sizeof(ip6_header_t))); + + iuvn6_0 = vlib_buffer_get_current (b0); + + /* pop (ip, udp, vxlan) */ + vlib_buffer_advance (b0, sizeof(*iuvn6_0)); + } + + tunnel_index0 = ~0; + error0 = 0; + + if (is_ip4) + { + next0 = + (iuvn4_0->vxlan.protocol < VXLAN_GPE_PROTOCOL_MAX)? + ngm->decap_next_node_list[iuvn4_0->vxlan.protocol]: \ + VXLAN_GPE_INPUT_NEXT_DROP; + + key4_0.local = iuvn4_0->ip4.dst_address.as_u32; + key4_0.remote = iuvn4_0->ip4.src_address.as_u32; + key4_0.vni = iuvn4_0->vxlan.vni_res; + key4_0.pad = 0; + + /* Processing for key4_0 */ + if (PREDICT_FALSE( + (key4_0.as_u64[0] != last_key4.as_u64[0]) + || (key4_0.as_u64[1] != last_key4.as_u64[1]))) + { + p0 = hash_get_mem(ngm->vxlan4_gpe_tunnel_by_key, &key4_0); + + if (p0 == 0) + { + error0 = VXLAN_GPE_ERROR_NO_SUCH_TUNNEL; + goto trace00; + } + + last_key4.as_u64[0] = key4_0.as_u64[0]; + last_key4.as_u64[1] = key4_0.as_u64[1]; + tunnel_index0 = last_tunnel_index = p0[0]; + } + else + tunnel_index0 = last_tunnel_index; + } + else /* is_ip6 */ + { + next0 = + (iuvn6_0->vxlan.protocol < VXLAN_GPE_PROTOCOL_MAX)? + ngm->decap_next_node_list[iuvn6_0->vxlan.protocol]: \ + VXLAN_GPE_INPUT_NEXT_DROP; + + key6_0.local.as_u64[0] = iuvn6_0->ip6.dst_address.as_u64[0]; + key6_0.local.as_u64[1] = iuvn6_0->ip6.dst_address.as_u64[1]; + key6_0.remote.as_u64[0] = iuvn6_0->ip6.src_address.as_u64[0]; + key6_0.remote.as_u64[1] = iuvn6_0->ip6.src_address.as_u64[1]; + key6_0.vni = iuvn6_0->vxlan.vni_res; + + /* Processing for key6_0 */ + if (PREDICT_FALSE(memcmp (&key6_0, &last_key6, sizeof(last_key6)) != 0)) + { + p0 = hash_get_mem(ngm->vxlan6_gpe_tunnel_by_key, &key6_0); + + if (p0 == 0) + { + error0 = VXLAN_GPE_ERROR_NO_SUCH_TUNNEL; + goto trace00; + } + + memcpy (&last_key6, &key6_0, sizeof(key6_0)); + tunnel_index0 = last_tunnel_index = p0[0]; + } + else + tunnel_index0 = last_tunnel_index; + } + + t0 = pool_elt_at_index(ngm->tunnels, tunnel_index0); + + + sw_if_index0 = t0->sw_if_index; + len0 = vlib_buffer_length_in_chain (vm, b0); + + /* Required to make the l2 tag push / pop code work on l2 subifs */ + vnet_update_l2_len (b0); + + /* + * ip[46] lookup in the configured FIB + */ + vnet_buffer(b0)->sw_if_index[VLIB_TX] = t0->decap_fib_index; + + pkts_decapsulated++; + stats_n_packets += 1; + stats_n_bytes += len0; + + /* Batch stats increment on the same vxlan-gpe tunnel so counter + is not incremented per packet */ + if (PREDICT_FALSE(sw_if_index0 != stats_sw_if_index)) + { + stats_n_packets -= 1; + stats_n_bytes -= len0; + if (stats_n_packets) + vlib_increment_combined_counter ( + im->combined_sw_if_counters + VNET_INTERFACE_COUNTER_RX, + cpu_index, stats_sw_if_index, stats_n_packets, stats_n_bytes); + stats_n_packets = 1; + stats_n_bytes = len0; + stats_sw_if_index = sw_if_index0; + } + + trace00: b0->error = error0 ? node->errors[error0] : 0; + + if (PREDICT_FALSE(b0->flags & VLIB_BUFFER_IS_TRACED)) + { + vxlan_gpe_rx_trace_t *tr = vlib_add_trace (vm, node, b0, sizeof(*tr)); + tr->next_index = next0; + tr->error = error0; + tr->tunnel_index = tunnel_index0; + } + vlib_validate_buffer_enqueue_x1(vm, node, next_index, to_next, + n_left_to_next, bi0, next0); + } + + vlib_put_next_frame (vm, node, next_index, n_left_to_next); + } + vlib_node_increment_counter (vm, vxlan_gpe_input_node.index, + VXLAN_GPE_ERROR_DECAPSULATED, pkts_decapsulated); + /* Increment any remaining batch stats */ + if (stats_n_packets) + { + vlib_increment_combined_counter ( + im->combined_sw_if_counters + VNET_INTERFACE_COUNTER_RX, cpu_index, + stats_sw_if_index, stats_n_packets, stats_n_bytes); + node->runtime_data[0] = stats_sw_if_index; + } + return from_frame->n_vectors; +} + +/** + * @brief Graph processing dispatch function for IPv4 VXLAN GPE + * + * @node vxlan4-gpe-input + * @param *vm + * @param *node + * @param *from_frame + * + * @return from_frame->n_vectors + * + */ +static uword +vxlan4_gpe_input (vlib_main_t * vm, vlib_node_runtime_t * node, + vlib_frame_t * from_frame) +{ + return vxlan_gpe_input (vm, node, from_frame, /* is_ip4 */1); +} + + +void +vxlan_gpe_register_decap_protocol (u8 protocol_id, uword next_node_index) +{ + vxlan_gpe_main_t *hm = &vxlan_gpe_main; + hm->decap_next_node_list[protocol_id] = next_node_index; + return; +} + +void +vxlan_gpe_unregister_decap_protocol (u8 protocol_id, uword next_node_index) +{ + vxlan_gpe_main_t *hm = &vxlan_gpe_main; + hm->decap_next_node_list[protocol_id] = VXLAN_GPE_INPUT_NEXT_DROP; + return; +} + + +/** + * @brief Graph processing dispatch function for IPv6 VXLAN GPE + * + * @node vxlan6-gpe-input + * @param *vm + * @param *node + * @param *from_frame + * + * @return from_frame->n_vectors - uword + * + */ +static uword +vxlan6_gpe_input (vlib_main_t * vm, vlib_node_runtime_t * node, + vlib_frame_t * from_frame) +{ + return vxlan_gpe_input (vm, node, from_frame, /* is_ip4 */0); +} + +/** + * @brief VXLAN GPE error strings + */ +static char * vxlan_gpe_error_strings[] = { +#define vxlan_gpe_error(n,s) s, +#include +#undef vxlan_gpe_error +#undef _ +}; + +VLIB_REGISTER_NODE (vxlan4_gpe_input_node) = { + .function = vxlan4_gpe_input, + .name = "vxlan4-gpe-input", + /* Takes a vector of packets. */ + .vector_size = sizeof (u32), + .type = VLIB_NODE_TYPE_INTERNAL, + .n_errors = ARRAY_LEN(vxlan_gpe_error_strings), + .error_strings = vxlan_gpe_error_strings, + + .n_next_nodes = VXLAN_GPE_INPUT_N_NEXT, + .next_nodes = { +#define _(s,n) [VXLAN_GPE_INPUT_NEXT_##s] = n, + foreach_vxlan_gpe_input_next +#undef _ + }, + + .format_buffer = format_vxlan_gpe_with_length, + .format_trace = format_vxlan_gpe_rx_trace, + // $$$$ .unformat_buffer = unformat_vxlan_gpe_header, +}; + +VLIB_NODE_FUNCTION_MULTIARCH (vxlan4_gpe_input_node, vxlan4_gpe_input); + +VLIB_REGISTER_NODE (vxlan6_gpe_input_node) = { + .function = vxlan6_gpe_input, + .name = "vxlan6-gpe-input", + /* Takes a vector of packets. */ + .vector_size = sizeof (u32), + .type = VLIB_NODE_TYPE_INTERNAL, + .n_errors = ARRAY_LEN(vxlan_gpe_error_strings), + .error_strings = vxlan_gpe_error_strings, + + .n_next_nodes = VXLAN_GPE_INPUT_N_NEXT, + .next_nodes = { +#define _(s,n) [VXLAN_GPE_INPUT_NEXT_##s] = n, + foreach_vxlan_gpe_input_next +#undef _ + }, + + .format_buffer = format_vxlan_gpe_with_length, + .format_trace = format_vxlan_gpe_rx_trace, + // $$$$ .unformat_buffer = unformat_vxlan_gpe_header, +}; + +VLIB_NODE_FUNCTION_MULTIARCH (vxlan6_gpe_input_node, vxlan6_gpe_input); diff --git a/src/vnet/vxlan-gpe/dir.dox b/src/vnet/vxlan-gpe/dir.dox new file mode 100644 index 00000000..c154733b --- /dev/null +++ b/src/vnet/vxlan-gpe/dir.dox @@ -0,0 +1,32 @@ +/* + * + * Copyright (c) 2013 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ +/** + @dir + @brief VXLAN GPE + + Based on IETF: draft-quinn-vxlan-gpe-03.txt + +Abstract + + This draft describes extending Virtual eXtensible Local Area Network + (VXLAN), via changes to the VXLAN header, with three new + capabilities: support for multi-protocol encapsulation, operations, + administration and management (OAM) signaling and explicit + versioning. + + See file: vxlan-gpe-rfc.txt + +*/ \ No newline at end of file diff --git a/src/vnet/vxlan-gpe/encap.c b/src/vnet/vxlan-gpe/encap.c new file mode 100644 index 00000000..3a486e56 --- /dev/null +++ b/src/vnet/vxlan-gpe/encap.c @@ -0,0 +1,388 @@ +/* + * 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. + */ +/** + * @file + * @brief Functions for encapsulating VXLAN GPE tunnels + * +*/ +#include +#include +#include +#include +#include +#include + +/** Statistics (not really errors) */ +#define foreach_vxlan_gpe_encap_error \ +_(ENCAPSULATED, "good packets encapsulated") + +/** + * @brief VXLAN GPE encap error strings + */ +static char * vxlan_gpe_encap_error_strings[] = { +#define _(sym,string) string, + foreach_vxlan_gpe_encap_error +#undef _ +}; + +/** + * @brief Struct for VXLAN GPE errors/counters + */ +typedef enum { +#define _(sym,str) VXLAN_GPE_ENCAP_ERROR_##sym, + foreach_vxlan_gpe_encap_error +#undef _ + VXLAN_GPE_ENCAP_N_ERROR, +} vxlan_gpe_encap_error_t; + +/** + * @brief Struct for tracing VXLAN GPE encapsulated packets + */ +typedef struct { + u32 tunnel_index; +} vxlan_gpe_encap_trace_t; + +/** + * @brief Trace of packets encapsulated in VXLAN GPE + * + * @param *s + * @param *args + * + * @return *s + * + */ +u8 * format_vxlan_gpe_encap_trace (u8 * s, va_list * args) +{ + CLIB_UNUSED (vlib_main_t * vm) = va_arg (*args, vlib_main_t *); + CLIB_UNUSED (vlib_node_t * node) = va_arg (*args, vlib_node_t *); + vxlan_gpe_encap_trace_t * t + = va_arg (*args, vxlan_gpe_encap_trace_t *); + + s = format (s, "VXLAN-GPE-ENCAP: tunnel %d", t->tunnel_index); + return s; +} + +/** + * @brief Instantiates UDP + VXLAN-GPE header then set next node to IP4|6 lookup + * + * @param *ngm + * @param *b0 + * @param *t0 contains rewrite header + * @param *next0 relative index of next dispatch function (next node) + * @param is_v4 Is this IPv4? (or IPv6) + * + */ +always_inline void +vxlan_gpe_encap_one_inline (vxlan_gpe_main_t * ngm, vlib_buffer_t * b0, + vxlan_gpe_tunnel_t * t0, u32 * next0, + u8 is_v4) +{ + ASSERT(sizeof(ip4_vxlan_gpe_header_t) == 36); + ASSERT(sizeof(ip6_vxlan_gpe_header_t) == 56); + + ip_udp_encap_one (ngm->vlib_main, b0, t0->rewrite, t0->rewrite_size, is_v4); + next0[0] = t0->encap_next_node; +} + +/** + * @brief Instantiates UDP + VXLAN-GPE header then set next node to IP4|6 lookup for two packets + * + * @param *ngm + * @param *b0 Packet0 + * @param *b1 Packet1 + * @param *t0 contains rewrite header for Packet0 + * @param *t1 contains rewrite header for Packet1 + * @param *next0 relative index of next dispatch function (next node) for Packet0 + * @param *next1 relative index of next dispatch function (next node) for Packet1 + * @param is_v4 Is this IPv4? (or IPv6) + * + */ +always_inline void +vxlan_gpe_encap_two_inline (vxlan_gpe_main_t * ngm, vlib_buffer_t * b0, + vlib_buffer_t * b1, vxlan_gpe_tunnel_t * t0, + vxlan_gpe_tunnel_t * t1, u32 * next0, + u32 * next1, u8 is_v4) +{ + ASSERT(sizeof(ip4_vxlan_gpe_header_t) == 36); + ASSERT(sizeof(ip6_vxlan_gpe_header_t) == 56); + + ip_udp_encap_one (ngm->vlib_main, b0, t0->rewrite, t0->rewrite_size, is_v4); + ip_udp_encap_one (ngm->vlib_main, b1, t1->rewrite, t1->rewrite_size, is_v4); + next0[0] = next1[0] = t0->encap_next_node; +} + +/** + * @brief Common processing for IPv4 and IPv6 VXLAN GPE encap dispatch functions + * + * It is worth noting that other than trivial UDP forwarding (transit), VXLAN GPE + * tunnels are "establish local". This means that we don't have a TX interface as yet + * as we need to look up where the outer-header dest is. By setting the TX index in the + * buffer metadata to the encap FIB, we can do a lookup to get the adjacency and real TX. + * + * vnet_buffer(b0)->sw_if_index[VLIB_TX] = t0->encap_fib_index; + * + * @node vxlan-gpe-input + * @param *vm + * @param *node + * @param *from_frame + * + * @return from_frame->n_vectors + * + */ +static uword +vxlan_gpe_encap (vlib_main_t * vm, + vlib_node_runtime_t * node, + vlib_frame_t * from_frame) +{ + u32 n_left_from, next_index, *from, *to_next; + vxlan_gpe_main_t * ngm = &vxlan_gpe_main; + vnet_main_t * vnm = ngm->vnet_main; + vnet_interface_main_t * im = &vnm->interface_main; + u32 pkts_encapsulated = 0; + u32 cpu_index = os_get_cpu_number (); + u32 stats_sw_if_index, stats_n_packets, stats_n_bytes; + + from = vlib_frame_vector_args (from_frame); + n_left_from = from_frame->n_vectors; + + next_index = node->cached_next_index; + stats_sw_if_index = node->runtime_data[0]; + stats_n_packets = stats_n_bytes = 0; + + while (n_left_from > 0) + { + u32 n_left_to_next; + + vlib_get_next_frame(vm, node, next_index, to_next, n_left_to_next); + + while (n_left_from >= 4 && n_left_to_next >= 2) + { + u32 bi0, bi1; + vlib_buffer_t * b0, *b1; + u32 next0, next1; + u32 sw_if_index0, sw_if_index1, len0, len1; + vnet_hw_interface_t * hi0, *hi1; + vxlan_gpe_tunnel_t * t0, *t1; + u8 is_ip4_0, is_ip4_1; + + next0 = next1 = VXLAN_GPE_ENCAP_NEXT_IP4_LOOKUP; + + /* Prefetch next iteration. */ + { + vlib_buffer_t * p2, *p3; + + p2 = vlib_get_buffer (vm, from[2]); + p3 = vlib_get_buffer (vm, from[3]); + + vlib_prefetch_buffer_header(p2, LOAD); + vlib_prefetch_buffer_header(p3, LOAD); + + CLIB_PREFETCH(p2->data, 2*CLIB_CACHE_LINE_BYTES, LOAD); + CLIB_PREFETCH(p3->data, 2*CLIB_CACHE_LINE_BYTES, LOAD); + } + + bi0 = from[0]; + bi1 = from[1]; + to_next[0] = bi0; + to_next[1] = bi1; + from += 2; + to_next += 2; + n_left_to_next -= 2; + n_left_from -= 2; + + b0 = vlib_get_buffer (vm, bi0); + b1 = vlib_get_buffer (vm, bi1); + + /* 1-wide cache? */ + sw_if_index0 = vnet_buffer(b0)->sw_if_index[VLIB_TX]; + sw_if_index1 = vnet_buffer(b1)->sw_if_index[VLIB_TX]; + hi0 = vnet_get_sup_hw_interface (vnm, vnet_buffer(b0)->sw_if_index[VLIB_TX]); + hi1 = vnet_get_sup_hw_interface (vnm, vnet_buffer(b1)->sw_if_index[VLIB_TX]); + + t0 = pool_elt_at_index(ngm->tunnels, hi0->dev_instance); + t1 = pool_elt_at_index(ngm->tunnels, hi1->dev_instance); + + is_ip4_0 = (t0->flags & VXLAN_GPE_TUNNEL_IS_IPV4); + is_ip4_1 = (t1->flags & VXLAN_GPE_TUNNEL_IS_IPV4); + + if (PREDICT_TRUE(is_ip4_0 == is_ip4_1)) + { + vxlan_gpe_encap_two_inline (ngm, b0, b1, t0, t1, &next0, &next1,is_ip4_0); + } + else + { + vxlan_gpe_encap_one_inline (ngm, b0, t0, &next0, is_ip4_0); + vxlan_gpe_encap_one_inline (ngm, b1, t1, &next1, is_ip4_1); + } + + /* Reset to look up tunnel partner in the configured FIB */ + vnet_buffer(b0)->sw_if_index[VLIB_TX] = t0->encap_fib_index; + vnet_buffer(b1)->sw_if_index[VLIB_TX] = t1->encap_fib_index; + vnet_buffer(b0)->sw_if_index[VLIB_RX] = sw_if_index0; + vnet_buffer(b1)->sw_if_index[VLIB_RX] = sw_if_index1; + pkts_encapsulated += 2; + + len0 = vlib_buffer_length_in_chain (vm, b0); + len1 = vlib_buffer_length_in_chain (vm, b0); + stats_n_packets += 2; + stats_n_bytes += len0 + len1; + + /* Batch stats increment on the same vxlan tunnel so counter is not + incremented per packet. Note stats are still incremented for deleted + and admin-down tunnel where packets are dropped. It is not worthwhile + to check for this rare case and affect normal path performance. */ + if (PREDICT_FALSE((sw_if_index0 != stats_sw_if_index) + || (sw_if_index1 != stats_sw_if_index))) + { + stats_n_packets -= 2; + stats_n_bytes -= len0 + len1; + if (sw_if_index0 == sw_if_index1) + { + if (stats_n_packets) + vlib_increment_combined_counter ( + im->combined_sw_if_counters + VNET_INTERFACE_COUNTER_TX, + cpu_index, stats_sw_if_index, stats_n_packets, stats_n_bytes); + stats_sw_if_index = sw_if_index0; + stats_n_packets = 2; + stats_n_bytes = len0 + len1; + } + else + { + vlib_increment_combined_counter ( + im->combined_sw_if_counters + VNET_INTERFACE_COUNTER_TX, + cpu_index, sw_if_index0, 1, len0); + vlib_increment_combined_counter ( + im->combined_sw_if_counters + VNET_INTERFACE_COUNTER_TX, + cpu_index, sw_if_index1, 1, len1); + } + } + + if (PREDICT_FALSE(b0->flags & VLIB_BUFFER_IS_TRACED)) + { + vxlan_gpe_encap_trace_t *tr = vlib_add_trace (vm, node, b0, sizeof(*tr)); + tr->tunnel_index = t0 - ngm->tunnels; + } + + if (PREDICT_FALSE(b1->flags & VLIB_BUFFER_IS_TRACED)) + { + vxlan_gpe_encap_trace_t *tr = vlib_add_trace (vm, node, b1, + sizeof(*tr)); + tr->tunnel_index = t1 - ngm->tunnels; + } + + vlib_validate_buffer_enqueue_x2(vm, node, next_index, to_next, + n_left_to_next, bi0, bi1, next0, next1); + } + + while (n_left_from > 0 && n_left_to_next > 0) + { + u32 bi0; + vlib_buffer_t * b0; + u32 next0 = VXLAN_GPE_ENCAP_NEXT_IP4_LOOKUP; + u32 sw_if_index0, len0; + vnet_hw_interface_t * hi0; + vxlan_gpe_tunnel_t * t0; + u8 is_ip4_0; + + bi0 = from[0]; + to_next[0] = bi0; + from += 1; + to_next += 1; + n_left_from -= 1; + n_left_to_next -= 1; + + b0 = vlib_get_buffer (vm, bi0); + + /* 1-wide cache? */ + sw_if_index0 = vnet_buffer(b0)->sw_if_index[VLIB_TX]; + hi0 = vnet_get_sup_hw_interface (vnm, vnet_buffer(b0)->sw_if_index[VLIB_TX]); + + t0 = pool_elt_at_index(ngm->tunnels, hi0->dev_instance); + + is_ip4_0 = (t0->flags & VXLAN_GPE_TUNNEL_IS_IPV4); + + vxlan_gpe_encap_one_inline (ngm, b0, t0, &next0, is_ip4_0); + + /* Reset to look up tunnel partner in the configured FIB */ + vnet_buffer(b0)->sw_if_index[VLIB_TX] = t0->encap_fib_index; + vnet_buffer(b0)->sw_if_index[VLIB_RX] = sw_if_index0; + pkts_encapsulated++; + + len0 = vlib_buffer_length_in_chain (vm, b0); + stats_n_packets += 1; + stats_n_bytes += len0; + + /* Batch stats increment on the same vxlan tunnel so counter is not + * incremented per packet. Note stats are still incremented for deleted + * and admin-down tunnel where packets are dropped. It is not worthwhile + * to check for this rare case and affect normal path performance. */ + if (PREDICT_FALSE(sw_if_index0 != stats_sw_if_index)) + { + stats_n_packets -= 1; + stats_n_bytes -= len0; + if (stats_n_packets) + vlib_increment_combined_counter ( + im->combined_sw_if_counters + VNET_INTERFACE_COUNTER_TX, + cpu_index, stats_sw_if_index, stats_n_packets, stats_n_bytes); + stats_n_packets = 1; + stats_n_bytes = len0; + stats_sw_if_index = sw_if_index0; + } + if (PREDICT_FALSE(b0->flags & VLIB_BUFFER_IS_TRACED)) + { + vxlan_gpe_encap_trace_t *tr = vlib_add_trace (vm, node, b0, + sizeof(*tr)); + tr->tunnel_index = t0 - ngm->tunnels; + } + vlib_validate_buffer_enqueue_x1(vm, node, next_index, to_next, + n_left_to_next, bi0, next0); + } + + vlib_put_next_frame (vm, node, next_index, n_left_to_next); + } + vlib_node_increment_counter (vm, node->node_index, + VXLAN_GPE_ENCAP_ERROR_ENCAPSULATED, + pkts_encapsulated); + /* Increment any remaining batch stats */ + if (stats_n_packets) + { + vlib_increment_combined_counter ( + im->combined_sw_if_counters + VNET_INTERFACE_COUNTER_TX, cpu_index, + stats_sw_if_index, stats_n_packets, stats_n_bytes); + node->runtime_data[0] = stats_sw_if_index; + } + + return from_frame->n_vectors; +} + +VLIB_REGISTER_NODE (vxlan_gpe_encap_node) = { + .function = vxlan_gpe_encap, + .name = "vxlan-gpe-encap", + .vector_size = sizeof (u32), + .format_trace = format_vxlan_gpe_encap_trace, + .type = VLIB_NODE_TYPE_INTERNAL, + + .n_errors = ARRAY_LEN(vxlan_gpe_encap_error_strings), + .error_strings = vxlan_gpe_encap_error_strings, + + .n_next_nodes = VXLAN_GPE_ENCAP_N_NEXT, + + .next_nodes = { + [VXLAN_GPE_ENCAP_NEXT_IP4_LOOKUP] = "ip4-lookup", + [VXLAN_GPE_ENCAP_NEXT_IP6_LOOKUP] = "ip6-lookup", + [VXLAN_GPE_ENCAP_NEXT_DROP] = "error-drop", + }, +}; + diff --git a/src/vnet/vxlan-gpe/vxlan-gpe-rfc.txt b/src/vnet/vxlan-gpe/vxlan-gpe-rfc.txt new file mode 100644 index 00000000..35cee50f --- /dev/null +++ b/src/vnet/vxlan-gpe/vxlan-gpe-rfc.txt @@ -0,0 +1,868 @@ +Network Working Group P. Quinn +Internet-Draft Cisco Systems, Inc. +Intended status: Experimental P. Agarwal +Expires: January 4, 2015 Broadcom + R. Fernando + L. Kreeger + D. Lewis + F. Maino + M. Smith + N. Yadav + Cisco Systems, Inc. + L. Yong + Huawei USA + X. Xu + Huawei Technologies + U. Elzur + Intel + P. Garg + Microsoft + July 3, 2014 + + + Generic Protocol Extension for VXLAN + draft-quinn-vxlan-gpe-03.txt + +Abstract + + This draft describes extending Virtual eXtensible Local Area Network + (VXLAN), via changes to the VXLAN header, with three new + capabilities: support for multi-protocol encapsulation, operations, + administration and management (OAM) signaling and explicit + versioning. + +Status of this Memo + + This Internet-Draft is submitted in full conformance with the + provisions of BCP 78 and BCP 79. + + Internet-Drafts are working documents of the Internet Engineering + Task Force (IETF). Note that other groups may also distribute + working documents as Internet-Drafts. The list of current Internet- + Drafts is at http://datatracker.ietf.org/drafts/current/. + + Internet-Drafts are draft documents valid for a maximum of six months + and may be updated, replaced, or obsoleted by other documents at any + time. It is inappropriate to use Internet-Drafts as reference + material or to cite them other than as "work in progress." + + + + +Quinn, et al. Expires January 4, 2015 [Page 1] + +Internet-Draft Generic Protocol Extension for VXLAN July 2014 + + + This Internet-Draft will expire on January 4, 2015. + +Copyright Notice + + Copyright (c) 2014 IETF Trust and the persons identified as the + document authors. All rights reserved. + + This document is subject to BCP 78 and the IETF Trust's Legal + Provisions Relating to IETF Documents + (http://trustee.ietf.org/license-info) in effect on the date of + publication of this document. Please review these documents + carefully, as they describe your rights and restrictions with respect + to this document. Code Components extracted from this document must + include Simplified BSD License text as described in Section 4.e of + the Trust Legal Provisions and are provided without warranty as + described in the Simplified BSD License. + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + +Quinn, et al. Expires January 4, 2015 [Page 2] + +Internet-Draft Generic Protocol Extension for VXLAN July 2014 + + +Table of Contents + + 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . 4 + 2. VXLAN Without Protocol Extension . . . . . . . . . . . . . . . 5 + 3. Generic Protocol Extension VXLAN (VXLAN-gpe) . . . . . . . . . 6 + 3.1. Multi Protocol Support . . . . . . . . . . . . . . . . . . 6 + 3.2. OAM Support . . . . . . . . . . . . . . . . . . . . . . . 7 + 3.3. Version Bits . . . . . . . . . . . . . . . . . . . . . . . 7 + 4. Backward Compatibility . . . . . . . . . . . . . . . . . . . . 8 + 4.1. VXLAN VTEP to VXLAN-gpe VTEP . . . . . . . . . . . . . . . 8 + 4.2. VXLAN-gpe VTEP to VXLAN VTEP . . . . . . . . . . . . . . . 8 + 4.3. VXLAN-gpe UDP Ports . . . . . . . . . . . . . . . . . . . 8 + 4.4. VXLAN-gpe and Encapsulated IP Header Fields . . . . . . . 8 + 5. VXLAN-gpe Examples . . . . . . . . . . . . . . . . . . . . . . 9 + 6. Security Considerations . . . . . . . . . . . . . . . . . . . 11 + 7. Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . 12 + 8. IANA Considerations . . . . . . . . . . . . . . . . . . . . . 13 + 8.1. UDP Port . . . . . . . . . . . . . . . . . . . . . . . . . 13 + 8.2. VXLAN-gpe Next Protocol . . . . . . . . . . . . . . . . . 13 + 8.3. VXLAN-gpe Reserved Bits . . . . . . . . . . . . . . . . . 13 + 9. References . . . . . . . . . . . . . . . . . . . . . . . . . . 14 + 9.1. Normative References . . . . . . . . . . . . . . . . . . . 14 + 9.2. Informative References . . . . . . . . . . . . . . . . . . 14 + Authors' Addresses . . . . . . . . . . . . . . . . . . . . . . . . 15 + + + + + + + + + + + + + + + + + + + + + + + + + + + +Quinn, et al. Expires January 4, 2015 [Page 3] + +Internet-Draft Generic Protocol Extension for VXLAN July 2014 + + +1. Introduction + + Virtual eXtensible Local Area Network [VXLAN] defines an + encapsulation format that encapsulates Ethernet frames in an outer + UDP/IP transport. As data centers evolve, the need to carry other + protocols encapsulated in an IP packet is required, as well as the + need to provide increased visibility and diagnostic capabilities + within the overlay. The VXLAN header does not specify the protocol + being encapsulated and therefore is currently limited to + encapsulating only Ethernet frame payload, nor does it provide the + ability to define OAM protocols. Rather than defining yet another + encapsulation, VXLAN is extended to provide protocol typing and OAM + capabilities. + + This document describes extending VXLAN via the following changes: + + Next Protocol Bit (P bit): A reserved flag bit is allocated, and set + in the VXLAN-gpe header to indicate that a next protocol field is + present. + + OAM Flag Bit (O bit): A reserved flag bit is allocated, and set in + the VXLAN-gpe header, to indicate that the packet is an OAM + packet. + + Version: Two reserved bits are allocated, and set in the VXLAN-gpe + header, to indicate VXLAN-gpe protocol version. + + Next Protocol: A 8 bit next protocol field is present in the VXLAN- + gpe header. + + + + + + + + + + + + + + + + + + + + + + +Quinn, et al. Expires January 4, 2015 [Page 4] + +Internet-Draft Generic Protocol Extension for VXLAN July 2014 + + +2. VXLAN Without Protocol Extension + + As described in the introduction, the VXLAN header has no protocol + identifier that indicates the type of payload being carried by VXLAN. + Because of this, VXLAN is limited to an Ethernet payload. + Furthermore, the VXLAN header has no mechanism to signal OAM packets. + + The VXLAN header defines bits 0-7 as flags (some defined, some + reserved), the VXLAN network identifier (VNI) field and several + reserved bits. The flags provide flexibility to define how the + reserved bits can be used to change the definition of the VXLAN + header. + + + + 0 1 2 3 + 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 + +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ + |R|R|R|R|I|R|R|R| Reserved | + +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ + | VXLAN Network Identifier (VNI) | Reserved | + +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ + + + Figure 1: VXLAN Header + + + + + + + + + + + + + + + + + + + + + + + + + + +Quinn, et al. Expires January 4, 2015 [Page 5] + +Internet-Draft Generic Protocol Extension for VXLAN July 2014 + + +3. Generic Protocol Extension VXLAN (VXLAN-gpe) + +3.1. Multi Protocol Support + + This draft defines the following two changes to the VXLAN header in + order to support multi-protocol encapsulation: + + P Bit: Flag bit 5 is defined as the Next Protocol bit. The P bit + MUST be set to 1 to indicate the presence of the 8 bit next + protocol field. + + P = 0 indicates that the payload MUST conform to VXLAN as defined + in [VXLAN]. + + Flag bit 5 was chosen as the P bit because this flag bit is + currently reserved in VXLAN. + + Next Protocol Field: The lower 8 bits of the first word are used to + carry a next protocol. This next protocol field contains the + protocol of the encapsulated payload packet. A new protocol + registry will be requested from IANA. + + This draft defines the following Next Protocol values: + + 0x1 : IPv4 + 0x2 : IPv6 + 0x3 : Ethernet + 0x4 : Network Service Header [NSH] + + + + + 0 1 2 3 + 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 + +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ + |R|R|R|R|I|P|R|R| Reserved |Next Protocol | + +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ + | VXLAN Network Identifier (VNI) | Reserved | + +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ + + + + Figure 2: VXLAN-gpe Next Protocol + + + + + + + + +Quinn, et al. Expires January 4, 2015 [Page 6] + +Internet-Draft Generic Protocol Extension for VXLAN July 2014 + + +3.2. OAM Support + + Flag bit 7 is defined as the O bit. When the O bit is set to 1, the + packet is an OAM packet and OAM processing MUST occur. The OAM + protocol details are out of scope for this document. As with the + P-bit, bit 7 is currently a reserved flag in VXLAN. + + + + 0 1 2 3 + 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 + +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ + |R|R|R|R|I|P|R|O| Reserved |Next Protocol | + +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ + | VXLAN Network Identifier (VNI) | Reserved | + +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ + + + + Figure 3: VXLAN-gpe OAM Bit + +3.3. Version Bits + + VXLAN-gpe bits 8 and 9 are defined as version bits. These bits are + reserved in VXLAN. The version field is used to ensure backward + compatibility going forward with future VXLAN-gpe updates. + + The initial version for VXLAN-gpe is 0. + + + + 0 1 2 3 + 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 + +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ + |R|R|R|R|I|P|R|O|Ver| Reserved |Next Protocol | + +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ + | VXLAN Network Identifier (VNI) | Reserved | + +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ + + + + + + Figure 4: VXLAN-gpe Version Bits + + + + + + + +Quinn, et al. Expires January 4, 2015 [Page 7] + +Internet-Draft Generic Protocol Extension for VXLAN July 2014 + + +4. Backward Compatibility + +4.1. VXLAN VTEP to VXLAN-gpe VTEP + + As per VXLAN, reserved bits 5 and 7, VXLAN-gpe P and O-bits + respectively must be set to zero. The remaining reserved bits must + be zero, including the VXLAN-gpe version field, bits 8 and 9. The + encapsulated payload MUST be Ethernet. + +4.2. VXLAN-gpe VTEP to VXLAN VTEP + + A VXLAN-gpe VTEP MUST NOT encapsulate non-Ethernet frames to a VXLAN + VTEP. When encapsulating Ethernet frames to a VXLAN VTEP, the VXLAN- + gpe VTEP will set the P bit to 0, the Next Protocol to 0 and use UDP + destination port 4789. A VXLAN-gpe VTEP MUST also set O = 0 and Ver + = 0 when encapsulating Ethernet frames to VXLAN VTEP. The receiving + VXLAN VTEP will threat this packet as a VXLAN packet. + + A method for determining the capabilities of a VXLAN VTEP (gpe or + non-gpe) is out of the scope of this draft. + +4.3. VXLAN-gpe UDP Ports + + VXLAN-gpe uses a new UDP destination port (to be assigned by IANA) + when sending traffic to VXLAN-gpe VTEPs. + +4.4. VXLAN-gpe and Encapsulated IP Header Fields + + When encapsulating and decapsulating IPv4 and IPv6 packets, certain + fields, such as IPv4 Time to Live (TTL) from the inner IP header need + to be considered. VXLAN-gpe IP encapsulation and decapsulation + utilizes the techniques described in [RFC6830], section 5.3. + + + + + + + + + + + + + + + + + + + +Quinn, et al. Expires January 4, 2015 [Page 8] + +Internet-Draft Generic Protocol Extension for VXLAN July 2014 + + +5. VXLAN-gpe Examples + + This section provides three examples of protocols encapsulated using + the Generic Protocol Extension for VXLAN described in this document. + + + + 0 1 2 3 + 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 + +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ + |R|R|R|R|I|1|R|0|0|0| Reserved | NP = IPv4 | + +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ + | VXLAN Network Identifier (VNI) | Reserved | + +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ + | Original IPv4 Packet | + +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ + + + + Figure 5: IPv4 and VXLAN-gpe + + + + + 0 1 2 3 + 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 + +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ + |R|R|R|R|I|1|R|0|0|0| Reserved | NP = IPv6 | + +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ + | VXLAN Network Identifier (VNI) | Reserved | + +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ + | Original IPv6 Packet | + +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ + + + + Figure 6: IPv6 and VXLAN-gpe + + + + + + + + + + + + + + +Quinn, et al. Expires January 4, 2015 [Page 9] + +Internet-Draft Generic Protocol Extension for VXLAN July 2014 + + + 0 1 2 3 + 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 + +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ + |R|R|R|R|I|1|R|0|0|0| Reserved |NP = Ethernet | + +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ + | VXLAN Network Identifier (VNI) | Reserved | + +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ + | Original Ethernet Frame | + +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ + + + + Figure 7: Ethernet and VXLAN-gpe + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + +Quinn, et al. Expires January 4, 2015 [Page 10] + +Internet-Draft Generic Protocol Extension for VXLAN July 2014 + + +6. Security Considerations + + VXLAN's security is focused on issues around L2 encapsulation into + L3. With VXLAN-gpe, issues such as spoofing, flooding, and traffic + redirection are dependent on the particular protocol payload + encapsulated. + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + +Quinn, et al. Expires January 4, 2015 [Page 11] + +Internet-Draft Generic Protocol Extension for VXLAN July 2014 + + +7. Acknowledgments + + A special thank you goes to Dino Farinacci for his guidance and + detailed review. + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + +Quinn, et al. Expires January 4, 2015 [Page 12] + +Internet-Draft Generic Protocol Extension for VXLAN July 2014 + + +8. IANA Considerations + +8.1. UDP Port + + A new UDP port will be requested from IANA. + +8.2. VXLAN-gpe Next Protocol + + IANA is requested to set up a registry of "Next Protocol". These are + 8-bit values. Next Protocol values 0, 1, 2, 3 and 4 are defined in + this draft. New values are assigned via Standards Action [RFC5226]. + + +---------------+-------------+---------------+ + | Next Protocol | Description | Reference | + +---------------+-------------+---------------+ + | 0 | Reserved | This document | + | | | | + | 1 | IPv4 | This document | + | | | | + | 2 | IPv6 | This document | + | | | | + | 3 | Ethernet | This document | + | | | | + | 4 | NSH | This document | + | | | | + | 5..253 | Unassigned | | + +---------------+-------------+---------------+ + + Table 1 + +8.3. VXLAN-gpe Reserved Bits + + There are ten bits at the beginning of the VXLAN-gpe header. New + bits are assigned via Standards Action [RFC5226]. + + Bits 0-3 - Reserved + Bit 4 - Instance ID (I bit) + Bit 5 - Next Protocol (P bit) + Bit 6 - Reserved + Bit 7 - OAM (O bit) + Bits 8-9 - Version + + + + + + + + + + +Quinn, et al. Expires January 4, 2015 [Page 13] + +Internet-Draft Generic Protocol Extension for VXLAN July 2014 + + +9. References + +9.1. Normative References + + [RFC0768] Postel, J., "User Datagram Protocol", STD 6, RFC 768, + August 1980. + + [RFC0791] Postel, J., "Internet Protocol", STD 5, RFC 791, + September 1981. + + [RFC2119] Bradner, S., "Key words for use in RFCs to Indicate + Requirement Levels", BCP 14, RFC 2119, March 1997. + + [RFC5226] Narten, T. and H. Alvestrand, "Guidelines for Writing an + IANA Considerations Section in RFCs", BCP 26, RFC 5226, + May 2008. + +9.2. Informative References + + [NSH] Quinn, P. and et al. , "Network Service Header", 2014. + + [RFC1700] Reynolds, J. and J. Postel, "Assigned Numbers", RFC 1700, + October 1994. + + [RFC6830] Farinacci, D., Fuller, V., Meyer, D., and D. Lewis, "The + Locator/ID Separation Protocol (LISP)", RFC 6830, + January 2013. + + [VXLAN] Dutt, D., Mahalingam, M., Duda, K., Agarwal, P., Kreeger, + L., Sridhar, T., Bursell, M., and C. Wright, "VXLAN: A + Framework for Overlaying Virtualized Layer 2 Networks over + Layer 3 Networks", 2013. + + + + + + + + + + + + + + + + + + + +Quinn, et al. Expires January 4, 2015 [Page 14] + +Internet-Draft Generic Protocol Extension for VXLAN July 2014 + + +Authors' Addresses + + Paul Quinn + Cisco Systems, Inc. + + Email: paulq@cisco.com + + + Puneet Agarwal + Broadcom + + Email: pagarwal@broadcom.com + + + Rex Fernando + Cisco Systems, Inc. + + Email: rex@cisco.com + + + Larry Kreeger + Cisco Systems, Inc. + + Email: kreeger@cisco.com + + + Darrel Lewis + Cisco Systems, Inc. + + Email: darlewis@cisco.com + + + Fabio Maino + Cisco Systems, Inc. + + Email: kreeger@cisco.com + + + Michael Smith + Cisco Systems, Inc. + + Email: michsmit@cisco.com + + + + + + + + + +Quinn, et al. Expires January 4, 2015 [Page 15] + +Internet-Draft Generic Protocol Extension for VXLAN July 2014 + + + Navindra Yadav + Cisco Systems, Inc. + + Email: nyadav@cisco.com + + + Lucy Yong + Huawei USA + + Email: lucy.yong@huawei.com + + + Xiaohu Xu + Huawei Technologies + + Email: xuxiaohu@huawei.com + + + Uri Elzur + Intel + + Email: uri.elzur@intel.com + + + Pankaj Garg + Microsoft + + Email: Garg.Pankaj@microsoft.com diff --git a/src/vnet/vxlan-gpe/vxlan_gpe.api b/src/vnet/vxlan-gpe/vxlan_gpe.api new file mode 100644 index 00000000..6c6973f8 --- /dev/null +++ b/src/vnet/vxlan-gpe/vxlan_gpe.api @@ -0,0 +1,61 @@ +/* + * Copyright (c) 2015-2016 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +define vxlan_gpe_add_del_tunnel +{ + u32 client_index; + u32 context; + u8 is_ipv6; + u8 local[16]; + u8 remote[16]; + u32 encap_vrf_id; + u32 decap_vrf_id; + u8 protocol; + u32 vni; + u8 is_add; +}; + +define vxlan_gpe_add_del_tunnel_reply +{ + u32 context; + i32 retval; + u32 sw_if_index; +}; + +define vxlan_gpe_tunnel_dump +{ + u32 client_index; + u32 context; + u32 sw_if_index; +}; + +define vxlan_gpe_tunnel_details +{ + u32 context; + u32 sw_if_index; + u8 local[16]; + u8 remote[16]; + u32 vni; + u8 protocol; + u32 encap_vrf_id; + u32 decap_vrf_id; + u8 is_ipv6; +}; + +/* + * Local Variables: + * eval: (c-set-style "gnu") + * End: + */ \ No newline at end of file diff --git a/src/vnet/vxlan-gpe/vxlan_gpe.c b/src/vnet/vxlan-gpe/vxlan_gpe.c new file mode 100644 index 00000000..b97510c4 --- /dev/null +++ b/src/vnet/vxlan-gpe/vxlan_gpe.c @@ -0,0 +1,659 @@ +/* + * 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. + */ +/** + * @file + * @brief Common utility functions for IPv4 and IPv6 VXLAN GPE tunnels + * +*/ +#include +#include +#include + +vxlan_gpe_main_t vxlan_gpe_main; + +/** + * @brief Tracing function for VXLAN GPE tunnel packets + * + * @param *s formatting string + * @param *args + * + * @return *s formatted string + * + */ +u8 * format_vxlan_gpe_tunnel (u8 * s, va_list * args) +{ + vxlan_gpe_tunnel_t * t = va_arg (*args, vxlan_gpe_tunnel_t *); + vxlan_gpe_main_t * gm = &vxlan_gpe_main; + + s = format (s, "[%d] local: %U remote: %U ", + t - gm->tunnels, + format_ip46_address, &t->local, IP46_TYPE_ANY, + format_ip46_address, &t->remote, IP46_TYPE_ANY); + + s = format (s, " vxlan VNI %d ", t->vni); + + switch (t->protocol) + { + case VXLAN_GPE_PROTOCOL_IP4: + s = format (s, "next-protocol ip4"); + break; + case VXLAN_GPE_PROTOCOL_IP6: + s = format (s, "next-protocol ip6"); + break; + case VXLAN_GPE_PROTOCOL_ETHERNET: + s = format (s, "next-protocol ethernet"); + break; + case VXLAN_GPE_PROTOCOL_NSH: + s = format (s, "next-protocol nsh"); + break; + default: + s = format (s, "next-protocol unknown %d", t->protocol); + } + + s = format (s, " fibs: (encap %d, decap %d)", + t->encap_fib_index, + t->decap_fib_index); + + return s; +} + +/** + * @brief Naming for VXLAN GPE tunnel + * + * @param *s formatting string + * @param *args + * + * @return *s formatted string + * + */ +static u8 * format_vxlan_gpe_name (u8 * s, va_list * args) +{ + u32 dev_instance = va_arg (*args, u32); + return format (s, "vxlan_gpe_tunnel%d", dev_instance); +} + +static uword dummy_interface_tx (vlib_main_t * vm, + vlib_node_runtime_t * node, + vlib_frame_t * frame) +{ + clib_warning ("you shouldn't be here, leaking buffers..."); + return frame->n_vectors; +} + +/** + * @brief CLI function for VXLAN GPE admin up/down + * + * @param *vnm + * @param hw_if_index + * @param flag + * + * @return *rc + * + */ +static clib_error_t * +vxlan_gpe_interface_admin_up_down (vnet_main_t * vnm, u32 hw_if_index, u32 flags) +{ + if (flags & VNET_SW_INTERFACE_FLAG_ADMIN_UP) + vnet_hw_interface_set_flags (vnm, hw_if_index, VNET_HW_INTERFACE_FLAG_LINK_UP); + else + vnet_hw_interface_set_flags (vnm, hw_if_index, 0); + + return 0; +} + +VNET_DEVICE_CLASS (vxlan_gpe_device_class,static) = { + .name = "VXLAN_GPE", + .format_device_name = format_vxlan_gpe_name, + .format_tx_trace = format_vxlan_gpe_encap_trace, + .tx_function = dummy_interface_tx, + .admin_up_down_function = vxlan_gpe_interface_admin_up_down, +}; + + +/** + * @brief Formatting function for tracing VXLAN GPE with length + * + * @param *s + * @param *args + * + * @return *s + * + */ +static u8 * format_vxlan_gpe_header_with_length (u8 * s, va_list * args) +{ + u32 dev_instance = va_arg (*args, u32); + s = format (s, "unimplemented dev %u", dev_instance); + return s; +} + +VNET_HW_INTERFACE_CLASS (vxlan_gpe_hw_class) = { + .name = "VXLAN_GPE", + .format_header = format_vxlan_gpe_header_with_length, + .build_rewrite = default_build_rewrite, + .flags = VNET_HW_INTERFACE_CLASS_FLAG_P2P, +}; + + +#define foreach_gpe_copy_field \ +_(vni) \ +_(protocol) \ +_(encap_fib_index) \ +_(decap_fib_index) + +#define foreach_copy_ipv4 { \ + _(local.ip4.as_u32) \ + _(remote.ip4.as_u32) \ +} + +#define foreach_copy_ipv6 { \ + _(local.ip6.as_u64[0]) \ + _(local.ip6.as_u64[1]) \ + _(remote.ip6.as_u64[0]) \ + _(remote.ip6.as_u64[1]) \ +} + + +/** + * @brief Calculate IPv4 VXLAN GPE rewrite header + * + * @param *t + * + * @return rc + * + */ +int vxlan4_gpe_rewrite (vxlan_gpe_tunnel_t * t, u32 extension_size, + u8 protocol_override, uword encap_next_node) +{ + u8 *rw = 0; + ip4_header_t * ip0; + ip4_vxlan_gpe_header_t * h0; + int len; + + len = sizeof (*h0) + extension_size; + + vec_free(t->rewrite); + vec_validate_aligned (rw, len-1, CLIB_CACHE_LINE_BYTES); + + h0 = (ip4_vxlan_gpe_header_t *) rw; + + /* Fixed portion of the (outer) ip4 header */ + ip0 = &h0->ip4; + ip0->ip_version_and_header_length = 0x45; + ip0->ttl = 254; + ip0->protocol = IP_PROTOCOL_UDP; + + /* we fix up the ip4 header length and checksum after-the-fact */ + ip0->src_address.as_u32 = t->local.ip4.as_u32; + ip0->dst_address.as_u32 = t->remote.ip4.as_u32; + ip0->checksum = ip4_header_checksum (ip0); + + /* UDP header, randomize src port on something, maybe? */ + h0->udp.src_port = clib_host_to_net_u16 (4790); + h0->udp.dst_port = clib_host_to_net_u16 (UDP_DST_PORT_vxlan_gpe); + + /* VXLAN header. Are we having fun yet? */ + h0->vxlan.flags = VXLAN_GPE_FLAGS_I | VXLAN_GPE_FLAGS_P; + h0->vxlan.ver_res = VXLAN_GPE_VERSION; + if (protocol_override) + { + h0->vxlan.protocol = protocol_override; + } + else + { + h0->vxlan.protocol = t->protocol; + } + t->rewrite_size = sizeof(ip4_vxlan_gpe_header_t) + extension_size; + h0->vxlan.vni_res = clib_host_to_net_u32 (t->vni<<8); + + t->rewrite = rw; + t->encap_next_node = encap_next_node; + return (0); +} + +/** + * @brief Calculate IPv6 VXLAN GPE rewrite header + * + * @param *t + * + * @return rc + * + */ +int vxlan6_gpe_rewrite (vxlan_gpe_tunnel_t * t, u32 extension_size, + u8 protocol_override, uword encap_next_node) +{ + u8 *rw = 0; + ip6_header_t * ip0; + ip6_vxlan_gpe_header_t * h0; + int len; + + len = sizeof (*h0) + extension_size; + + vec_free(t->rewrite); + vec_validate_aligned (rw, len-1, CLIB_CACHE_LINE_BYTES); + + h0 = (ip6_vxlan_gpe_header_t *) rw; + + /* Fixed portion of the (outer) ip4 header */ + ip0 = &h0->ip6; + ip0->ip_version_traffic_class_and_flow_label = clib_host_to_net_u32(6 << 28); + ip0->hop_limit = 255; + ip0->protocol = IP_PROTOCOL_UDP; + + ip0->src_address.as_u64[0] = t->local.ip6.as_u64[0]; + ip0->src_address.as_u64[1] = t->local.ip6.as_u64[1]; + ip0->dst_address.as_u64[0] = t->remote.ip6.as_u64[0]; + ip0->dst_address.as_u64[1] = t->remote.ip6.as_u64[1]; + + /* UDP header, randomize src port on something, maybe? */ + h0->udp.src_port = clib_host_to_net_u16 (4790); + h0->udp.dst_port = clib_host_to_net_u16 (UDP_DST_PORT_vxlan_gpe); + + /* VXLAN header. Are we having fun yet? */ + h0->vxlan.flags = VXLAN_GPE_FLAGS_I | VXLAN_GPE_FLAGS_P; + h0->vxlan.ver_res = VXLAN_GPE_VERSION; + if (protocol_override) + { + h0->vxlan.protocol = t->protocol; + } + else + { + h0->vxlan.protocol = protocol_override; + } + t->rewrite_size = sizeof(ip4_vxlan_gpe_header_t) + extension_size; + h0->vxlan.vni_res = clib_host_to_net_u32 (t->vni<<8); + + t->rewrite = rw; + t->encap_next_node = encap_next_node; + return (0); +} + +/** + * @brief Add or Del a VXLAN GPE tunnel + * + * @param *a + * @param *sw_if_index + * + * @return rc + * + */ +int vnet_vxlan_gpe_add_del_tunnel +(vnet_vxlan_gpe_add_del_tunnel_args_t *a, u32 * sw_if_indexp) +{ + vxlan_gpe_main_t * gm = &vxlan_gpe_main; + vxlan_gpe_tunnel_t *t = 0; + vnet_main_t * vnm = gm->vnet_main; + vnet_hw_interface_t * hi; + uword * p; + u32 hw_if_index = ~0; + u32 sw_if_index = ~0; + int rv; + vxlan4_gpe_tunnel_key_t key4, *key4_copy; + vxlan6_gpe_tunnel_key_t key6, *key6_copy; + hash_pair_t *hp; + + if (!a->is_ip6) + { + key4.local = a->local.ip4.as_u32; + key4.remote = a->remote.ip4.as_u32; + key4.vni = clib_host_to_net_u32 (a->vni << 8); + key4.pad = 0; + + p = hash_get_mem(gm->vxlan4_gpe_tunnel_by_key, &key4); + } + else + { + key6.local.as_u64[0] = a->local.ip6.as_u64[0]; + key6.local.as_u64[1] = a->local.ip6.as_u64[1]; + key6.remote.as_u64[0] = a->remote.ip6.as_u64[0]; + key6.remote.as_u64[1] = a->remote.ip6.as_u64[1]; + key6.vni = clib_host_to_net_u32 (a->vni << 8); + + p = hash_get_mem(gm->vxlan6_gpe_tunnel_by_key, &key6); + } + + if (a->is_add) + { + /* adding a tunnel: tunnel must not already exist */ + if (p) + return VNET_API_ERROR_INVALID_VALUE; + + pool_get_aligned (gm->tunnels, t, CLIB_CACHE_LINE_BYTES); + memset (t, 0, sizeof (*t)); + + /* copy from arg structure */ +#define _(x) t->x = a->x; + foreach_gpe_copy_field; + if (!a->is_ip6) foreach_copy_ipv4 + else foreach_copy_ipv6 +#undef _ + + if (!a->is_ip6) t->flags |= VXLAN_GPE_TUNNEL_IS_IPV4; + + if (!a->is_ip6) { + rv = vxlan4_gpe_rewrite (t, 0, 0, VXLAN_GPE_ENCAP_NEXT_IP4_LOOKUP); + } else { + rv = vxlan6_gpe_rewrite (t, 0, 0, VXLAN_GPE_ENCAP_NEXT_IP6_LOOKUP); + } + + if (rv) + { + pool_put (gm->tunnels, t); + return rv; + } + + if (!a->is_ip6) + { + key4_copy = clib_mem_alloc (sizeof (*key4_copy)); + clib_memcpy (key4_copy, &key4, sizeof (*key4_copy)); + hash_set_mem (gm->vxlan4_gpe_tunnel_by_key, key4_copy, + t - gm->tunnels); + } + else + { + key6_copy = clib_mem_alloc (sizeof (*key6_copy)); + clib_memcpy (key6_copy, &key6, sizeof (*key6_copy)); + hash_set_mem (gm->vxlan6_gpe_tunnel_by_key, key6_copy, + t - gm->tunnels); + } + + if (vec_len (gm->free_vxlan_gpe_tunnel_hw_if_indices) > 0) + { + hw_if_index = gm->free_vxlan_gpe_tunnel_hw_if_indices + [vec_len (gm->free_vxlan_gpe_tunnel_hw_if_indices)-1]; + _vec_len (gm->free_vxlan_gpe_tunnel_hw_if_indices) -= 1; + + hi = vnet_get_hw_interface (vnm, hw_if_index); + hi->dev_instance = t - gm->tunnels; + hi->hw_instance = hi->dev_instance; + } + else + { + hw_if_index = vnet_register_interface + (vnm, vxlan_gpe_device_class.index, t - gm->tunnels, + vxlan_gpe_hw_class.index, t - gm->tunnels); + hi = vnet_get_hw_interface (vnm, hw_if_index); + hi->output_node_index = vxlan_gpe_encap_node.index; + } + + t->hw_if_index = hw_if_index; + t->sw_if_index = sw_if_index = hi->sw_if_index; + vec_validate_init_empty (gm->tunnel_index_by_sw_if_index, sw_if_index, ~0); + gm->tunnel_index_by_sw_if_index[sw_if_index] = t - gm->tunnels; + + vnet_sw_interface_set_flags (vnm, hi->sw_if_index, + VNET_SW_INTERFACE_FLAG_ADMIN_UP); + } + else + { + /* deleting a tunnel: tunnel must exist */ + if (!p) + return VNET_API_ERROR_NO_SUCH_ENTRY; + + t = pool_elt_at_index (gm->tunnels, p[0]); + + vnet_sw_interface_set_flags (vnm, t->sw_if_index, 0 /* down */); + vec_add1 (gm->free_vxlan_gpe_tunnel_hw_if_indices, t->hw_if_index); + + gm->tunnel_index_by_sw_if_index[t->sw_if_index] = ~0; + + if (!a->is_ip6) + { + hp = hash_get_pair (gm->vxlan4_gpe_tunnel_by_key, &key4); + key4_copy = (void *)(hp->key); + hash_unset_mem (gm->vxlan4_gpe_tunnel_by_key, &key4); + clib_mem_free (key4_copy); + } + else + { + hp = hash_get_pair (gm->vxlan6_gpe_tunnel_by_key, &key6); + key6_copy = (void *)(hp->key); + hash_unset_mem (gm->vxlan4_gpe_tunnel_by_key, &key6); + clib_mem_free (key6_copy); + } + + vec_free (t->rewrite); + pool_put (gm->tunnels, t); + } + + if (sw_if_indexp) + *sw_if_indexp = sw_if_index; + + return 0; +} + +static clib_error_t * +vxlan_gpe_add_del_tunnel_command_fn (vlib_main_t * vm, + unformat_input_t * input, + vlib_cli_command_t * cmd) +{ + unformat_input_t _line_input, * line_input = &_line_input; + u8 is_add = 1; + ip46_address_t local, remote; + u8 local_set = 0; + u8 remote_set = 0; + u8 ipv4_set = 0; + u8 ipv6_set = 0; + u32 encap_fib_index = 0; + u32 decap_fib_index = 0; + u8 protocol = VXLAN_GPE_PROTOCOL_IP4; + u32 vni; + u8 vni_set = 0; + int rv; + u32 tmp; + vnet_vxlan_gpe_add_del_tunnel_args_t _a, * a = &_a; + u32 sw_if_index; + + /* Get a line of input. */ + if (! unformat_user (input, unformat_line_input, line_input)) + return 0; + + while (unformat_check_input (line_input) != UNFORMAT_END_OF_INPUT) { + if (unformat (line_input, "del")) + is_add = 0; + else if (unformat (line_input, "local %U", + unformat_ip4_address, &local.ip4)) + { + local_set = 1; + ipv4_set = 1; + } + else if (unformat (line_input, "remote %U", + unformat_ip4_address, &remote.ip4)) + { + remote_set = 1; + ipv4_set = 1; + } + else if (unformat (line_input, "local %U", + unformat_ip6_address, &local.ip6)) + { + local_set = 1; + ipv6_set = 1; + } + else if (unformat (line_input, "remote %U", + unformat_ip6_address, &remote.ip6)) + { + remote_set = 1; + ipv6_set = 1; + } + else if (unformat (line_input, "encap-vrf-id %d", &tmp)) + { + if (ipv6_set) + encap_fib_index = ip6_fib_index_from_table_id (tmp); + else + encap_fib_index = ip4_fib_index_from_table_id (tmp); + + if (encap_fib_index == ~0) + return clib_error_return (0, "nonexistent encap fib id %d", tmp); + } + else if (unformat (line_input, "decap-vrf-id %d", &tmp)) + { + if (ipv6_set) + decap_fib_index = ip6_fib_index_from_table_id (tmp); + else + decap_fib_index = ip4_fib_index_from_table_id (tmp); + + if (decap_fib_index == ~0) + return clib_error_return (0, "nonexistent decap fib id %d", tmp); + } + else if (unformat (line_input, "vni %d", &vni)) + vni_set = 1; + else if (unformat(line_input, "next-ip4")) + protocol = VXLAN_GPE_PROTOCOL_IP4; + else if (unformat(line_input, "next-ip6")) + protocol = VXLAN_GPE_PROTOCOL_IP6; + else if (unformat(line_input, "next-ethernet")) + protocol = VXLAN_GPE_PROTOCOL_ETHERNET; + else if (unformat(line_input, "next-nsh")) + protocol = VXLAN_GPE_PROTOCOL_NSH; + else + return clib_error_return (0, "parse error: '%U'", + format_unformat_error, line_input); + } + + unformat_free (line_input); + + if (local_set == 0) + return clib_error_return (0, "tunnel local address not specified"); + + if (remote_set == 0) + return clib_error_return (0, "tunnel remote address not specified"); + + if (ipv4_set && ipv6_set) + return clib_error_return (0, "both IPv4 and IPv6 addresses specified"); + + if ((ipv4_set && memcmp(&local.ip4, &remote.ip4, sizeof(local.ip4)) == 0) || + (ipv6_set && memcmp(&local.ip6, &remote.ip6, sizeof(local.ip6)) == 0)) + return clib_error_return (0, "src and dst addresses are identical"); + + if (vni_set == 0) + return clib_error_return (0, "vni not specified"); + + memset (a, 0, sizeof (*a)); + + a->is_add = is_add; + a->is_ip6 = ipv6_set; + +#define _(x) a->x = x; + foreach_gpe_copy_field; + if (ipv4_set) foreach_copy_ipv4 + else foreach_copy_ipv6 +#undef _ + + rv = vnet_vxlan_gpe_add_del_tunnel (a, &sw_if_index); + + switch(rv) + { + case 0: + vlib_cli_output(vm, "%U\n", format_vnet_sw_if_index_name, vnet_get_main(), sw_if_index); + break; + case VNET_API_ERROR_INVALID_DECAP_NEXT: + return clib_error_return (0, "invalid decap-next..."); + + case VNET_API_ERROR_TUNNEL_EXIST: + return clib_error_return (0, "tunnel already exists..."); + + case VNET_API_ERROR_NO_SUCH_ENTRY: + return clib_error_return (0, "tunnel does not exist..."); + + default: + return clib_error_return + (0, "vnet_vxlan_gpe_add_del_tunnel returned %d", rv); + } + + return 0; +} + +VLIB_CLI_COMMAND (create_vxlan_gpe_tunnel_command, static) = { + .path = "create vxlan-gpe tunnel", + .short_help = + "create vxlan-gpe tunnel local remote " + " vni [next-ip4][next-ip6][next-ethernet][next-nsh]" + " [encap-vrf-id ] [decap-vrf-id ]" + " [del]\n", + .function = vxlan_gpe_add_del_tunnel_command_fn, +}; + +/** + * @brief CLI function for showing VXLAN GPE tunnels + * + * @param *vm + * @param *input + * @param *cmd + * + * @return error + * + */ +static clib_error_t * +show_vxlan_gpe_tunnel_command_fn (vlib_main_t * vm, + unformat_input_t * input, + vlib_cli_command_t * cmd) +{ + vxlan_gpe_main_t * gm = &vxlan_gpe_main; + vxlan_gpe_tunnel_t * t; + + if (pool_elts (gm->tunnels) == 0) + vlib_cli_output (vm, "No vxlan-gpe tunnels configured."); + + pool_foreach (t, gm->tunnels, + ({ + vlib_cli_output (vm, "%U", format_vxlan_gpe_tunnel, t); + })); + + return 0; +} + +VLIB_CLI_COMMAND (show_vxlan_gpe_tunnel_command, static) = { + .path = "show vxlan-gpe", + .function = show_vxlan_gpe_tunnel_command_fn, +}; + +/** + * @brief Feature init function for VXLAN GPE + * + * @param *vm + * + * @return error + * + */ +clib_error_t *vxlan_gpe_init (vlib_main_t *vm) +{ + vxlan_gpe_main_t *gm = &vxlan_gpe_main; + + gm->vnet_main = vnet_get_main(); + gm->vlib_main = vm; + + gm->vxlan4_gpe_tunnel_by_key + = hash_create_mem (0, sizeof(vxlan4_gpe_tunnel_key_t), sizeof (uword)); + + gm->vxlan6_gpe_tunnel_by_key + = hash_create_mem (0, sizeof(vxlan6_gpe_tunnel_key_t), sizeof (uword)); + + + udp_register_dst_port (vm, UDP_DST_PORT_vxlan_gpe, + vxlan4_gpe_input_node.index, 1 /* is_ip4 */); + udp_register_dst_port (vm, UDP_DST_PORT_vxlan6_gpe, + vxlan6_gpe_input_node.index, 0 /* is_ip4 */); + + /* Register the list of standard decap protocols supported */ + vxlan_gpe_register_decap_protocol (VXLAN_GPE_PROTOCOL_IP4, + VXLAN_GPE_INPUT_NEXT_IP4_INPUT); + vxlan_gpe_register_decap_protocol (VXLAN_GPE_PROTOCOL_IP6, + VXLAN_GPE_INPUT_NEXT_IP6_INPUT); + vxlan_gpe_register_decap_protocol (VXLAN_GPE_PROTOCOL_ETHERNET, + VXLAN_GPE_INPUT_NEXT_ETHERNET_INPUT); + return 0; +} + +VLIB_INIT_FUNCTION(vxlan_gpe_init); + diff --git a/src/vnet/vxlan-gpe/vxlan_gpe.h b/src/vnet/vxlan-gpe/vxlan_gpe.h new file mode 100644 index 00000000..1b4bc44e --- /dev/null +++ b/src/vnet/vxlan-gpe/vxlan_gpe.h @@ -0,0 +1,221 @@ +/* + * 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. + */ +/** + * @file + * @brief VXLAN GPE definitions + * +*/ +#ifndef included_vnet_vxlan_gpe_h +#define included_vnet_vxlan_gpe_h + +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include + +/** + * @brief VXLAN GPE header struct + * + */ +typedef CLIB_PACKED (struct { + /** 20 bytes */ + ip4_header_t ip4; + /** 8 bytes */ + udp_header_t udp; + /** 8 bytes */ + vxlan_gpe_header_t vxlan; +}) ip4_vxlan_gpe_header_t; + +typedef CLIB_PACKED (struct { + /** 40 bytes */ + ip6_header_t ip6; + /** 8 bytes */ + udp_header_t udp; + /** 8 bytes */ + vxlan_gpe_header_t vxlan; +}) ip6_vxlan_gpe_header_t; + +/** + * @brief Key struct for IPv4 VXLAN GPE tunnel. + * Key fields: local remote, vni + * all fields in NET byte order + * VNI shifted 8 bits + */ +typedef CLIB_PACKED(struct { + union { + struct { + u32 local; + u32 remote; + + u32 vni; + u32 pad; + }; + u64 as_u64[2]; + }; +}) vxlan4_gpe_tunnel_key_t; + +/** + * @brief Key struct for IPv6 VXLAN GPE tunnel. + * Key fields: local remote, vni + * all fields in NET byte order + * VNI shifted 8 bits + */ +typedef CLIB_PACKED(struct { + ip6_address_t local; + ip6_address_t remote; + u32 vni; +}) vxlan6_gpe_tunnel_key_t; + +/** + * @brief Struct for VXLAN GPE tunnel + */ +typedef struct { + /** Rewrite string. $$$$ embed vnet_rewrite header */ + u8 * rewrite; + + /** encapsulated protocol */ + u8 protocol; + + /** tunnel local address */ + ip46_address_t local; + /** tunnel remote address */ + ip46_address_t remote; + + /** FIB indices - tunnel partner lookup here */ + u32 encap_fib_index; + /** FIB indices - inner IP packet lookup here */ + u32 decap_fib_index; + + /** VXLAN GPE VNI in HOST byte order, shifted left 8 bits */ + u32 vni; + + /** vnet intfc hw_if_index */ + u32 hw_if_index; + /** vnet intfc sw_if_index */ + u32 sw_if_index; + + /** flags */ + u32 flags; + + /** rewrite size for dynamic plugins like iOAM */ + u8 rewrite_size; + + /** Next node after VxLAN-GPE encap */ + uword encap_next_node; +} vxlan_gpe_tunnel_t; + +/** Flags for vxlan_gpe_tunnel_t */ +#define VXLAN_GPE_TUNNEL_IS_IPV4 1 + +/** next nodes for VXLAN GPE input */ +#define foreach_vxlan_gpe_input_next \ +_(DROP, "error-drop") \ +_(IP4_INPUT, "ip4-input") \ +_(IP6_INPUT, "ip6-input") \ +_(ETHERNET_INPUT, "ethernet-input") + +/** struct for next nodes for VXLAN GPE input */ +typedef enum { +#define _(s,n) VXLAN_GPE_INPUT_NEXT_##s, + foreach_vxlan_gpe_input_next +#undef _ + VXLAN_GPE_INPUT_N_NEXT, +} vxlan_gpe_input_next_t; + +/** struct for VXLAN GPE errors */ +typedef enum { +#define vxlan_gpe_error(n,s) VXLAN_GPE_ERROR_##n, +#include +#undef vxlan_gpe_error + VXLAN_GPE_N_ERROR, +} vxlan_gpe_input_error_t; + +/** Struct for VXLAN GPE node state */ +typedef struct { + /** vector of encap tunnel instances */ + vxlan_gpe_tunnel_t *tunnels; + + /** lookup IPv4 VXLAN GPE tunnel by key */ + uword * vxlan4_gpe_tunnel_by_key; + /** lookup IPv6 VXLAN GPE tunnel by key */ + uword * vxlan6_gpe_tunnel_by_key; + + /** Free vlib hw_if_indices */ + u32 * free_vxlan_gpe_tunnel_hw_if_indices; + + /** Mapping from sw_if_index to tunnel index */ + u32 * tunnel_index_by_sw_if_index; + + /** State convenience vlib_main_t */ + vlib_main_t * vlib_main; + /** State convenience vnet_main_t */ + vnet_main_t * vnet_main; + + /** List of next nodes for the decap indexed on protocol */ + uword decap_next_node_list[VXLAN_GPE_PROTOCOL_MAX]; +} vxlan_gpe_main_t; + +vxlan_gpe_main_t vxlan_gpe_main; + +extern vlib_node_registration_t vxlan_gpe_encap_node; +extern vlib_node_registration_t vxlan4_gpe_input_node; +extern vlib_node_registration_t vxlan6_gpe_input_node; + +u8 * format_vxlan_gpe_encap_trace (u8 * s, va_list * args); + +/** Struct for VXLAN GPE add/del args */ +typedef struct { + u8 is_add; + u8 is_ip6; + ip46_address_t local, remote; + u8 protocol; + u32 encap_fib_index; + u32 decap_fib_index; + u32 vni; +} vnet_vxlan_gpe_add_del_tunnel_args_t; + + +int vnet_vxlan_gpe_add_del_tunnel +(vnet_vxlan_gpe_add_del_tunnel_args_t *a, u32 * sw_if_indexp); + + +int vxlan4_gpe_rewrite (vxlan_gpe_tunnel_t * t, u32 extension_size, + u8 protocol_override, uword encap_next_node); +int vxlan6_gpe_rewrite (vxlan_gpe_tunnel_t * t, u32 extension_size, + u8 protocol_override, uword encap_next_node); + +/** + * @brief Struct for defining VXLAN GPE next nodes + */ +typedef enum { + VXLAN_GPE_ENCAP_NEXT_IP4_LOOKUP, + VXLAN_GPE_ENCAP_NEXT_IP6_LOOKUP, + VXLAN_GPE_ENCAP_NEXT_DROP, + VXLAN_GPE_ENCAP_N_NEXT +} vxlan_gpe_encap_next_t; + + +void vxlan_gpe_unregister_decap_protocol (u8 protocol_id, uword next_node_index); + +void vxlan_gpe_register_decap_protocol (u8 protocol_id, uword next_node_index); + + +#endif /* included_vnet_vxlan_gpe_h */ diff --git a/src/vnet/vxlan-gpe/vxlan_gpe_api.c b/src/vnet/vxlan-gpe/vxlan_gpe_api.c new file mode 100644 index 00000000..012a41da --- /dev/null +++ b/src/vnet/vxlan-gpe/vxlan_gpe_api.c @@ -0,0 +1,249 @@ +/* + *------------------------------------------------------------------ + * vxlan_gpe_api.c - vxlan_gpe api + * + * Copyright (c) 2016 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + *------------------------------------------------------------------ + */ + +#include +#include + +#include +#include +#include +#include + +#include + +#define vl_typedefs /* define message structures */ +#include +#undef vl_typedefs + +#define vl_endianfun /* define message structures */ +#include +#undef vl_endianfun + +/* instantiate all the print functions we know about */ +#define vl_print(handle, ...) vlib_cli_output (handle, __VA_ARGS__) +#define vl_printfun +#include +#undef vl_printfun + +#include + +#define foreach_vpe_api_msg \ +_(VXLAN_GPE_ADD_DEL_TUNNEL, vxlan_gpe_add_del_tunnel) \ +_(VXLAN_GPE_TUNNEL_DUMP, vxlan_gpe_tunnel_dump) \ + +static void + vl_api_vxlan_gpe_add_del_tunnel_t_handler + (vl_api_vxlan_gpe_add_del_tunnel_t * mp) +{ + vl_api_vxlan_gpe_add_del_tunnel_reply_t *rmp; + int rv = 0; + vnet_vxlan_gpe_add_del_tunnel_args_t _a, *a = &_a; + u32 encap_fib_index, decap_fib_index; + u8 protocol; + uword *p; + ip4_main_t *im = &ip4_main; + u32 sw_if_index = ~0; + + + p = hash_get (im->fib_index_by_table_id, ntohl (mp->encap_vrf_id)); + if (!p) + { + rv = VNET_API_ERROR_NO_SUCH_FIB; + goto out; + } + encap_fib_index = p[0]; + + protocol = mp->protocol; + + /* Interpret decap_vrf_id as an opaque if sending to other-than-ip4-input */ + if (protocol == VXLAN_GPE_INPUT_NEXT_IP4_INPUT) + { + p = hash_get (im->fib_index_by_table_id, ntohl (mp->decap_vrf_id)); + if (!p) + { + rv = VNET_API_ERROR_NO_SUCH_INNER_FIB; + goto out; + } + decap_fib_index = p[0]; + } + else + { + decap_fib_index = ntohl (mp->decap_vrf_id); + } + + /* Check src & dst are different */ + if ((mp->is_ipv6 && memcmp (mp->local, mp->remote, 16) == 0) || + (!mp->is_ipv6 && memcmp (mp->local, mp->remote, 4) == 0)) + { + rv = VNET_API_ERROR_SAME_SRC_DST; + goto out; + } + memset (a, 0, sizeof (*a)); + + a->is_add = mp->is_add; + a->is_ip6 = mp->is_ipv6; + /* ip addresses sent in network byte order */ + if (a->is_ip6) + { + clib_memcpy (&(a->local.ip6), mp->local, 16); + clib_memcpy (&(a->remote.ip6), mp->remote, 16); + } + else + { + clib_memcpy (&(a->local.ip4), mp->local, 4); + clib_memcpy (&(a->remote.ip4), mp->remote, 4); + } + a->encap_fib_index = encap_fib_index; + a->decap_fib_index = decap_fib_index; + a->protocol = protocol; + a->vni = ntohl (mp->vni); + rv = vnet_vxlan_gpe_add_del_tunnel (a, &sw_if_index); + +out: + /* *INDENT-OFF* */ + REPLY_MACRO2(VL_API_VXLAN_GPE_ADD_DEL_TUNNEL_REPLY, + ({ + rmp->sw_if_index = ntohl (sw_if_index); + })); + /* *INDENT-ON* */ +} + +static void send_vxlan_gpe_tunnel_details + (vxlan_gpe_tunnel_t * t, unix_shared_memory_queue_t * q, u32 context) +{ + vl_api_vxlan_gpe_tunnel_details_t *rmp; + ip4_main_t *im4 = &ip4_main; + ip6_main_t *im6 = &ip6_main; + u8 is_ipv6 = !(t->flags & VXLAN_GPE_TUNNEL_IS_IPV4); + + rmp = vl_msg_api_alloc (sizeof (*rmp)); + memset (rmp, 0, sizeof (*rmp)); + rmp->_vl_msg_id = ntohs (VL_API_VXLAN_GPE_TUNNEL_DETAILS); + if (is_ipv6) + { + memcpy (rmp->local, &(t->local.ip6), 16); + memcpy (rmp->remote, &(t->remote.ip6), 16); + rmp->encap_vrf_id = htonl (im6->fibs[t->encap_fib_index].ft_table_id); + rmp->decap_vrf_id = htonl (im6->fibs[t->decap_fib_index].ft_table_id); + } + else + { + memcpy (rmp->local, &(t->local.ip4), 4); + memcpy (rmp->remote, &(t->remote.ip4), 4); + rmp->encap_vrf_id = htonl (im4->fibs[t->encap_fib_index].ft_table_id); + rmp->decap_vrf_id = htonl (im4->fibs[t->decap_fib_index].ft_table_id); + } + rmp->vni = htonl (t->vni); + rmp->protocol = t->protocol; + rmp->sw_if_index = htonl (t->sw_if_index); + rmp->is_ipv6 = is_ipv6; + rmp->context = context; + + vl_msg_api_send_shmem (q, (u8 *) & rmp); +} + +static void vl_api_vxlan_gpe_tunnel_dump_t_handler + (vl_api_vxlan_gpe_tunnel_dump_t * mp) +{ + unix_shared_memory_queue_t *q; + vxlan_gpe_main_t *vgm = &vxlan_gpe_main; + vxlan_gpe_tunnel_t *t; + u32 sw_if_index; + + q = vl_api_client_index_to_input_queue (mp->client_index); + if (q == 0) + { + return; + } + + sw_if_index = ntohl (mp->sw_if_index); + + if (~0 == sw_if_index) + { + /* *INDENT-OFF* */ + pool_foreach (t, vgm->tunnels, + ({ + send_vxlan_gpe_tunnel_details(t, q, mp->context); + })); + /* *INDENT-ON* */ + } + else + { + if ((sw_if_index >= vec_len (vgm->tunnel_index_by_sw_if_index)) || + (~0 == vgm->tunnel_index_by_sw_if_index[sw_if_index])) + { + return; + } + t = &vgm->tunnels[vgm->tunnel_index_by_sw_if_index[sw_if_index]]; + send_vxlan_gpe_tunnel_details (t, q, mp->context); + } +} + + +/* + * vpe_api_hookup + * Add vpe's API message handlers to the table. + * vlib has alread mapped shared memory and + * added the client registration handlers. + * See .../vlib-api/vlibmemory/memclnt_vlib.c:memclnt_process() + */ +#define vl_msg_name_crc_list +#include +#undef vl_msg_name_crc_list + +static void +setup_message_id_table (api_main_t * am) +{ +#define _(id,n,crc) vl_msg_api_add_msg_name_crc (am, #n "_" #crc, id); + foreach_vl_msg_name_crc_span; +#undef _ +} + +static clib_error_t * +vxlan_gpe_api_hookup (vlib_main_t * vm) +{ + api_main_t *am = &api_main; + +#define _(N,n) \ + vl_msg_api_set_handlers(VL_API_##N, #n, \ + vl_api_##n##_t_handler, \ + vl_noop_handler, \ + vl_api_##n##_t_endian, \ + vl_api_##n##_t_print, \ + sizeof(vl_api_##n##_t), 1); + foreach_vpe_api_msg; +#undef _ + + /* + * Set up the (msg_name, crc, message-id) table + */ + setup_message_id_table (am); + + return 0; +} + +VLIB_API_INIT_FUNCTION (vxlan_gpe_api_hookup); + +/* + * fd.io coding-style-patch-verification: ON + * + * Local Variables: + * eval: (c-set-style "gnu") + * End: + */ diff --git a/src/vnet/vxlan-gpe/vxlan_gpe_error.def b/src/vnet/vxlan-gpe/vxlan_gpe_error.def new file mode 100644 index 00000000..9cf1b1cb --- /dev/null +++ b/src/vnet/vxlan-gpe/vxlan_gpe_error.def @@ -0,0 +1,16 @@ +/* + * Copyright (c) 2015 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ +vxlan_gpe_error (DECAPSULATED, "good packets decapsulated") +vxlan_gpe_error (NO_SUCH_TUNNEL, "no such tunnel packets") diff --git a/src/vnet/vxlan-gpe/vxlan_gpe_packet.h b/src/vnet/vxlan-gpe/vxlan_gpe_packet.h new file mode 100644 index 00000000..ec3c2e58 --- /dev/null +++ b/src/vnet/vxlan-gpe/vxlan_gpe_packet.h @@ -0,0 +1,110 @@ +/* + * 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. + */ +/** + * @file + * @brief VXLAN GPE packet header structure + * +*/ +#ifndef included_vxlan_gpe_packet_h +#define included_vxlan_gpe_packet_h + +/** + * From draft-quinn-vxlan-gpe-03.txt + * + * 0 1 2 3 + * 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 + * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ + * |R|R|R|R|I|P|R|O|Ver| Reserved |Next Protocol | + * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ + * | VXLAN Network Identifier (VNI) | Reserved | + * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ + * + * I Bit: Flag bit 4 indicates that the VNI is valid. + * + * P Bit: Flag bit 5 is defined as the Next Protocol bit. The P bit + * MUST be set to 1 to indicate the presence of the 8 bit next + * protocol field. + * + * O Bit: Flag bit 7 is defined as the O bit. When the O bit is set to 1, + * + * the packet is an OAM packet and OAM processing MUST occur. The OAM + * protocol details are out of scope for this document. As with the + * P-bit, bit 7 is currently a reserved flag in VXLAN. + * + * VXLAN-gpe bits 8 and 9 are defined as version bits. These bits are + * reserved in VXLAN. The version field is used to ensure backward + * compatibility going forward with future VXLAN-gpe updates. + * + * The initial version for VXLAN-gpe is 0. + * + * This draft defines the following Next Protocol values: + * + * 0x1 : IPv4 + * 0x2 : IPv6 + * 0x3 : Ethernet + * 0x4 : Network Service Header [NSH] + */ + +/** + * @brief VXLAN GPE support inner protocol definition. + * 1 - IP4 + * 2 - IP6 + * 3 - ETHERNET + * 4 - NSH + */ +#define foreach_vxlan_gpe_protocol \ +_ (0x01, IP4) \ +_ (0x02, IP6) \ +_ (0x03, ETHERNET) \ +_ (0x04, NSH) \ +_ (0x05, IOAM) + + +/** + * @brief Struct for VXLAN GPE support inner protocol definition. + * 1 - IP4 + * 2 - IP6 + * 3 - ETHERNET + * 4 - NSH + * 5 - IOAM + */ +typedef enum { +#define _(n,f) VXLAN_GPE_PROTOCOL_##f = n, + foreach_vxlan_gpe_protocol +#undef _ + VXLAN_GPE_PROTOCOL_MAX, +} vxlan_gpe_protocol_t; + +/** + * @brief VXLAN GPE Header definition + */ +typedef struct { + u8 flags; + /** Version and Reserved */ + u8 ver_res; + /** Reserved */ + u8 res; + /** see vxlan_gpe_protocol_t */ + u8 protocol; + /** VNI and Reserved */ + u32 vni_res; +} vxlan_gpe_header_t; + +#define VXLAN_GPE_FLAGS_I 0x08 +#define VXLAN_GPE_FLAGS_P 0x04 +#define VXLAN_GPE_FLAGS_O 0x01 +#define VXLAN_GPE_VERSION 0x0 + +#endif /* included_vxlan_gpe_packet_h */ -- cgit 1.2.3-korg From da8b9a88d502ac575919bca25482d035330aea37 Mon Sep 17 00:00:00 2001 From: Gabriel Ganne Date: Tue, 3 Jan 2017 14:07:19 +0100 Subject: vlxlan-gpe - fix api registering (was registering span api twice instead) Change-Id: I196d520061d10872fd795c0b1115574591e8d978 Signed-off-by: Gabriel Ganne --- src/vnet/vxlan-gpe/vxlan_gpe_api.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src/vnet/vxlan-gpe') diff --git a/src/vnet/vxlan-gpe/vxlan_gpe_api.c b/src/vnet/vxlan-gpe/vxlan_gpe_api.c index 012a41da..0215054d 100644 --- a/src/vnet/vxlan-gpe/vxlan_gpe_api.c +++ b/src/vnet/vxlan-gpe/vxlan_gpe_api.c @@ -211,7 +211,7 @@ static void setup_message_id_table (api_main_t * am) { #define _(id,n,crc) vl_msg_api_add_msg_name_crc (am, #n "_" #crc, id); - foreach_vl_msg_name_crc_span; + foreach_vl_msg_name_crc_vxlan_gpe; #undef _ } -- cgit 1.2.3-korg From a9a20e7f69f4a91a4d5267ab5ce14125bdc7d6c6 Mon Sep 17 00:00:00 2001 From: Billy McFall Date: Wed, 15 Feb 2017 11:39:12 -0500 Subject: VPP-635: CLI Memory leak with invalid parameter In the CLI parsing, below is a common pattern: /* Get a line of input. */ if (!unformat_user (input, unformat_line_input, line_input)) return 0; while (unformat_check_input (line_input) != UNFORMAT_END_OF_INPUT) { if (unformat (line_input, "x")) x = 1; : else return clib_error_return (0, "unknown input `%U'", format_unformat_error, line_input); } unformat_free (line_input); The 'else' returns if an unknown string is encountered. There a memory leak because the 'unformat_free(line_input)' is not called. There is a large number of instances of this pattern. Replaced the previous pattern with: /* Get a line of input. */ if (!unformat_user (input, unformat_line_input, line_input)) return 0; while (unformat_check_input (line_input) != UNFORMAT_END_OF_INPUT) { if (unformat (line_input, "x")) x = 1; : else { error = clib_error_return (0, "unknown input `%U'", format_unformat_error, line_input); goto done: } } /* ...Remaining code... */ done: unformat_free (line_input); return error; } In multiple files, 'unformat_free (line_input);' was never called, so there was a memory leak whether an invalid string was entered or not. Also, there were multiple instance where: error = clib_error_return (0, "unknown input `%U'", format_unformat_error, line_input); used 'input' as the last parameter instead of 'line_input'. The result is that output did not contain the substring in error, instead just an empty string. Fixed all of those as well. There are a lot of file, and very mind numbing work, so tried to keep it to a pattern to avoid mistakes. Change-Id: I8902f0c32a47dd7fb3bb3471a89818571702f1d2 Signed-off-by: Billy McFall Signed-off-by: Dave Barach --- build-root/emacs-lisp/tunnel-c-skel.el | 19 ++- src/plugins/ila/ila.c | 25 ++- src/plugins/lb/cli.c | 99 ++++++----- src/plugins/sixrd/sixrd.c | 42 +++-- src/plugins/snat/snat.c | 139 +++++++++++----- src/vlib/threads_cli.c | 79 ++++++--- src/vlib/trace.c | 13 +- src/vlib/unix/cli.c | 22 ++- src/vnet/devices/af_packet/cli.c | 56 +++++-- src/vnet/devices/dpdk/cli.c | 290 +++++++++++++++++++++++---------- src/vnet/devices/dpdk/ipsec/cli.c | 15 +- src/vnet/devices/netmap/cli.c | 54 ++++-- src/vnet/devices/virtio/vhost-user.c | 62 +++++-- src/vnet/gre/interface.c | 35 ++-- src/vnet/ip/ip4_source_check.c | 6 +- src/vnet/ip/ip4_test.c | 15 +- src/vnet/ip/ip6_neighbor.c | 27 ++- src/vnet/ip/lookup.c | 34 ++-- src/vnet/ipsec-gre/interface.c | 34 ++-- src/vnet/ipsec/ipsec_cli.c | 177 +++++++++++++------- src/vnet/l2/l2_patch.c | 26 ++- src/vnet/l2/l2_xcrw.c | 34 +++- src/vnet/l2tp/l2tp.c | 39 +++-- src/vnet/lisp-cp/lisp_cli.c | 139 ++++++++++++---- src/vnet/lisp-gpe/interface.c | 58 +++++-- src/vnet/lisp-gpe/lisp_gpe.c | 13 +- src/vnet/map/map.c | 186 +++++++++++++++------ src/vnet/mpls/mpls.c | 2 + src/vnet/mpls/mpls_tunnel.c | 19 ++- src/vnet/pg/cli.c | 39 +++-- src/vnet/policer/node_funcs.c | 19 ++- src/vnet/policer/policer.c | 13 +- src/vnet/unix/tapcli.c | 57 +++++-- src/vnet/vxlan-gpe/vxlan_gpe.c | 62 +++++-- src/vnet/vxlan/vxlan.c | 81 ++++++--- src/vpp/app/l2t.c | 9 +- src/vpp/app/vpe_cli.c | 24 ++- 37 files changed, 1487 insertions(+), 576 deletions(-) (limited to 'src/vnet/vxlan-gpe') diff --git a/build-root/emacs-lisp/tunnel-c-skel.el b/build-root/emacs-lisp/tunnel-c-skel.el index aa260e53..a1b1757d 100644 --- a/build-root/emacs-lisp/tunnel-c-skel.el +++ b/build-root/emacs-lisp/tunnel-c-skel.el @@ -288,6 +288,7 @@ static clib_error_t * vlib_cli_command_t * cmd) { unformat_input_t _line_input, * line_input = &_line_input; + clib_error_t *error = 0; ip4_address_t src, dst; u8 is_add = 1; u8 src_set = 0; @@ -322,13 +323,19 @@ static clib_error_t * { encap_fib_index = fib_index_from_fib_id (tmp); if (encap_fib_index == ~0) - return clib_error_return (0, \"nonexistent encap fib id %d\", tmp); + { + unformat_free (line_input); + return clib_error_return (0, \"nonexistent encap fib id %d\", tmp); + } } else if (unformat (line_input, \"decap-vrf-id %d\", &tmp)) { decap_fib_index = fib_index_from_fib_id (tmp); if (decap_fib_index == ~0) - return clib_error_return (0, \"nonexistent decap fib id %d\", tmp); + { + unformat_free (line_input); + return clib_error_return (0, \"nonexistent decap fib id %d\", tmp); + } } else if (unformat (line_input, \"decap-next %U\", unformat_decap_next, &decap_next_index)) @@ -346,8 +353,12 @@ static clib_error_t * * in the " ENCAP_STACK " header */ else - return clib_error_return (0, \"parse error: '%U'\", - format_unformat_error, line_input); + { + error = clib_error_return (0, \"parse error: '%U'\", + format_unformat_error, line_input); + unformat_free (line_input); + return error; + } } unformat_free (line_input); diff --git a/src/plugins/ila/ila.c b/src/plugins/ila/ila.c index e0f3907f..52c7ea55 100644 --- a/src/plugins/ila/ila.c +++ b/src/plugins/ila/ila.c @@ -949,6 +949,7 @@ ila_entry_command_fn (vlib_main_t * vm, ila_add_del_entry_args_t args = { 0 }; u8 next_hop_set = 0; int ret; + clib_error_t *error = 0; args.type = ILA_TYPE_IID; args.csum_mode = ILA_CSUM_MODE_NO_ACTION; @@ -986,19 +987,29 @@ ila_entry_command_fn (vlib_main_t * vm, else if (unformat (line_input, "del")) args.is_del = 1; else - return clib_error_return (0, "parse error: '%U'", - format_unformat_error, line_input); + { + error = clib_error_return (0, "parse error: '%U'", + format_unformat_error, line_input); + goto done; + } } - unformat_free (line_input); - if (!next_hop_set) - return clib_error_return (0, "Specified a next hop"); + { + error = clib_error_return (0, "Specified a next hop"); + goto done; + } if ((ret = ila_add_del_entry (&args))) - return clib_error_return (0, "ila_add_del_entry returned error %d", ret); + { + error = clib_error_return (0, "ila_add_del_entry returned error %d", ret); + goto done; + } - return NULL; +done: + unformat_free (line_input); + + return error; } VLIB_CLI_COMMAND (ila_entry_command, static) = diff --git a/src/plugins/lb/cli.c b/src/plugins/lb/cli.c index b59c6426..6452a875 100644 --- a/src/plugins/lb/cli.c +++ b/src/plugins/lb/cli.c @@ -28,13 +28,16 @@ lb_vip_command_fn (vlib_main_t * vm, int ret; u32 gre4 = 0; lb_vip_type_t type; + clib_error_t *error = 0; if (!unformat_user (input, unformat_line_input, line_input)) return 0; - if (!unformat(line_input, "%U", unformat_ip46_prefix, &prefix, &plen, IP46_TYPE_ANY, &plen)) - return clib_error_return (0, "invalid vip prefix: '%U'", - format_unformat_error, line_input); + if (!unformat(line_input, "%U", unformat_ip46_prefix, &prefix, &plen, IP46_TYPE_ANY, &plen)) { + error = clib_error_return (0, "invalid vip prefix: '%U'", + format_unformat_error, line_input); + goto done; + } while (unformat_check_input (line_input) != UNFORMAT_END_OF_INPUT) { @@ -46,13 +49,13 @@ lb_vip_command_fn (vlib_main_t * vm, gre4 = 1; else if (unformat(line_input, "encap gre6")) gre4 = 0; - else - return clib_error_return (0, "parse error: '%U'", - format_unformat_error, line_input); + else { + error = clib_error_return (0, "parse error: '%U'", + format_unformat_error, line_input); + goto done; + } } - unformat_free (line_input); - if (ip46_prefix_is_ip4(&prefix, plen)) { type = (gre4)?LB_VIP_TYPE_IP4_GRE4:LB_VIP_TYPE_IP4_GRE6; @@ -65,17 +68,25 @@ lb_vip_command_fn (vlib_main_t * vm, u32 index; if (!del) { if ((ret = lb_vip_add(&prefix, plen, type, new_len, &index))) { - return clib_error_return (0, "lb_vip_add error %d", ret); + error = clib_error_return (0, "lb_vip_add error %d", ret); + goto done; } else { vlib_cli_output(vm, "lb_vip_add ok %d", index); } } else { - if ((ret = lb_vip_find_index(&prefix, plen, &index))) - return clib_error_return (0, "lb_vip_find_index error %d", ret); - else if ((ret = lb_vip_del(index))) - return clib_error_return (0, "lb_vip_del error %d", ret); + if ((ret = lb_vip_find_index(&prefix, plen, &index))) { + error = clib_error_return (0, "lb_vip_find_index error %d", ret); + goto done; + } else if ((ret = lb_vip_del(index))) { + error = clib_error_return (0, "lb_vip_del error %d", ret); + goto done; + } } - return NULL; + +done: + unformat_free (line_input); + + return error; } VLIB_CLI_COMMAND (lb_vip_command, static) = @@ -96,16 +107,21 @@ lb_as_command_fn (vlib_main_t * vm, u32 vip_index; u8 del = 0; int ret; + clib_error_t *error = 0; if (!unformat_user (input, unformat_line_input, line_input)) return 0; - if (!unformat(line_input, "%U", unformat_ip46_prefix, &vip_prefix, &vip_plen, IP46_TYPE_ANY)) - return clib_error_return (0, "invalid as address: '%U'", - format_unformat_error, line_input); + if (!unformat(line_input, "%U", unformat_ip46_prefix, &vip_prefix, &vip_plen, IP46_TYPE_ANY)) { + error = clib_error_return (0, "invalid as address: '%U'", + format_unformat_error, line_input); + goto done; + } - if ((ret = lb_vip_find_index(&vip_prefix, vip_plen, &vip_index))) - return clib_error_return (0, "lb_vip_find_index error %d", ret); + if ((ret = lb_vip_find_index(&vip_prefix, vip_plen, &vip_index))) { + error = clib_error_return (0, "lb_vip_find_index error %d", ret); + goto done; + } while (unformat_check_input (line_input) != UNFORMAT_END_OF_INPUT) { @@ -114,15 +130,15 @@ lb_as_command_fn (vlib_main_t * vm, } else if (unformat(line_input, "del")) { del = 1; } else { - vec_free(as_array); - return clib_error_return (0, "parse error: '%U'", - format_unformat_error, line_input); + error = clib_error_return (0, "parse error: '%U'", + format_unformat_error, line_input); + goto done; } } if (!vec_len(as_array)) { - vec_free(as_array); - return clib_error_return (0, "No AS address provided"); + error = clib_error_return (0, "No AS address provided"); + goto done; } lb_garbage_collection(); @@ -130,18 +146,21 @@ lb_as_command_fn (vlib_main_t * vm, if (del) { if ((ret = lb_vip_del_ass(vip_index, as_array, vec_len(as_array)))) { - vec_free(as_array); - return clib_error_return (0, "lb_vip_del_ass error %d", ret); + error = clib_error_return (0, "lb_vip_del_ass error %d", ret); + goto done; } } else { if ((ret = lb_vip_add_ass(vip_index, as_array, vec_len(as_array)))) { - vec_free(as_array); - return clib_error_return (0, "lb_vip_add_ass error %d", ret); + error = clib_error_return (0, "lb_vip_add_ass error %d", ret); + goto done; } } +done: + unformat_free (line_input); vec_free(as_array); - return 0; + + return error; } VLIB_CLI_COMMAND (lb_as_command, static) = @@ -163,6 +182,7 @@ lb_conf_command_fn (vlib_main_t * vm, u32 per_cpu_sticky_buckets_log2 = 0; u32 flow_timeout = lbm->flow_timeout; int ret; + clib_error_t *error = 0; if (!unformat_user (input, unformat_line_input, line_input)) return 0; @@ -181,19 +201,24 @@ lb_conf_command_fn (vlib_main_t * vm, per_cpu_sticky_buckets = 1 << per_cpu_sticky_buckets_log2; } else if (unformat(line_input, "timeout %d", &flow_timeout)) ; - else - return clib_error_return (0, "parse error: '%U'", - format_unformat_error, line_input); + else { + error = clib_error_return (0, "parse error: '%U'", + format_unformat_error, line_input); + goto done; + } } - unformat_free (line_input); - lb_garbage_collection(); - if ((ret = lb_conf(&ip4, &ip6, per_cpu_sticky_buckets, flow_timeout))) - return clib_error_return (0, "lb_conf error %d", ret); + if ((ret = lb_conf(&ip4, &ip6, per_cpu_sticky_buckets, flow_timeout))) { + error = clib_error_return (0, "lb_conf error %d", ret); + goto done; + } - return NULL; +done: + unformat_free (line_input); + + return error; } VLIB_CLI_COMMAND (lb_conf_command, static) = diff --git a/src/plugins/sixrd/sixrd.c b/src/plugins/sixrd/sixrd.c index 71fc181f..67a9a3ad 100644 --- a/src/plugins/sixrd/sixrd.c +++ b/src/plugins/sixrd/sixrd.c @@ -192,6 +192,7 @@ sixrd_add_domain_command_fn (vlib_main_t *vm, u32 num_m_args = 0; /* Optional arguments */ u32 mtu = 0; + clib_error_t *error = 0; /* Get a line of input. */ if (!unformat_user(input, unformat_line_input, line_input)) @@ -205,19 +206,25 @@ sixrd_add_domain_command_fn (vlib_main_t *vm, num_m_args++; else if (unformat(line_input, "mtu %d", &mtu)) num_m_args++; - else - return clib_error_return(0, "unknown input `%U'", - format_unformat_error, input); + else { + error = clib_error_return(0, "unknown input `%U'", + format_unformat_error, line_input); + goto done; + } } - unformat_free(line_input); - if (num_m_args < 3) - return clib_error_return(0, "mandatory argument(s) missing"); + if (num_m_args < 3) { + error = clib_error_return(0, "mandatory argument(s) missing"); + goto done; + } sixrd_create_domain(&ip6_prefix, ip6_prefix_len, &ip4_prefix, ip4_prefix_len, &ip4_src, &sixrd_domain_index, mtu); - return 0; +done: + unformat_free (line_input); + + return error; } static clib_error_t * @@ -228,6 +235,7 @@ sixrd_del_domain_command_fn (vlib_main_t *vm, unformat_input_t _line_input, *line_input = &_line_input; u32 num_m_args = 0; u32 sixrd_domain_index; + clib_error_t *error = 0; /* Get a line of input. */ if (! unformat_user(input, unformat_line_input, line_input)) @@ -236,18 +244,24 @@ sixrd_del_domain_command_fn (vlib_main_t *vm, while (unformat_check_input(line_input) != UNFORMAT_END_OF_INPUT) { if (unformat(line_input, "index %d", &sixrd_domain_index)) num_m_args++; - else - return clib_error_return(0, "unknown input `%U'", - format_unformat_error, input); + else { + error = clib_error_return(0, "unknown input `%U'", + format_unformat_error, line_input); + goto done; + } } - unformat_free(line_input); - if (num_m_args != 1) - return clib_error_return(0, "mandatory argument(s) missing"); + if (num_m_args != 1) { + error = clib_error_return(0, "mandatory argument(s) missing"); + goto done; + } sixrd_delete_domain(sixrd_domain_index); - return 0; +done: + unformat_free (line_input); + + return error; } static u8 * diff --git a/src/plugins/snat/snat.c b/src/plugins/snat/snat.c index 73854a7a..8c2bacdb 100644 --- a/src/plugins/snat/snat.c +++ b/src/plugins/snat/snat.c @@ -1705,6 +1705,7 @@ add_address_command_fn (vlib_main_t * vm, int i, count; int is_add = 1; int rv = 0; + clib_error_t *error = 0; /* Get a line of input. */ if (!unformat_user (input, unformat_line_input, line_input)) @@ -1721,19 +1722,27 @@ add_address_command_fn (vlib_main_t * vm, else if (unformat (line_input, "del")) is_add = 0; else - return clib_error_return (0, "unknown input '%U'", - format_unformat_error, input); + { + error = clib_error_return (0, "unknown input '%U'", + format_unformat_error, line_input); + goto done; + } } - unformat_free (line_input); if (sm->static_mapping_only) - return clib_error_return (0, "static mapping only mode"); + { + error = clib_error_return (0, "static mapping only mode"); + goto done; + } start_host_order = clib_host_to_net_u32 (start_addr.as_u32); end_host_order = clib_host_to_net_u32 (end_addr.as_u32); if (end_host_order < start_host_order) - return clib_error_return (0, "end address less than start address"); + { + error = clib_error_return (0, "end address less than start address"); + goto done; + } count = (end_host_order - start_host_order) + 1; @@ -1755,11 +1764,11 @@ add_address_command_fn (vlib_main_t * vm, switch (rv) { case VNET_API_ERROR_NO_SUCH_ENTRY: - return clib_error_return (0, "S-NAT address not exist."); - break; + error = clib_error_return (0, "S-NAT address not exist."); + goto done; case VNET_API_ERROR_UNSPECIFIED: - return clib_error_return (0, "S-NAT address used in static mapping."); - break; + error = clib_error_return (0, "S-NAT address used in static mapping."); + goto done; default: break; } @@ -1767,7 +1776,10 @@ add_address_command_fn (vlib_main_t * vm, increment_v4_address (&this_addr); } - return 0; +done: + unformat_free (line_input); + + return error; } VLIB_CLI_COMMAND (add_address_command, static) = { @@ -1807,10 +1819,12 @@ snat_feature_command_fn (vlib_main_t * vm, else if (unformat (line_input, "del")) is_del = 1; else - return clib_error_return (0, "unknown input '%U'", - format_unformat_error, input); + { + error = clib_error_return (0, "unknown input '%U'", + format_unformat_error, line_input); + goto done; + } } - unformat_free (line_input); if (vec_len (inside_sw_if_indices)) { @@ -1830,6 +1844,8 @@ snat_feature_command_fn (vlib_main_t * vm, } } +done: + unformat_free (line_input); vec_free (inside_sw_if_indices); vec_free (outside_sw_if_indices); @@ -1923,13 +1939,18 @@ add_static_mapping_command_fn (vlib_main_t * vm, else if (unformat (line_input, "del")) is_add = 0; else - return clib_error_return (0, "unknown input: '%U'", - format_unformat_error, line_input); + { + error = clib_error_return (0, "unknown input: '%U'", + format_unformat_error, line_input); + goto done; + } } - unformat_free (line_input); if (!addr_only && !proto_set) - return clib_error_return (0, "missing protocol"); + { + error = clib_error_return (0, "missing protocol"); + goto done; + } rv = snat_add_static_mapping(l_addr, e_addr, (u16) l_port, (u16) e_port, vrf_id, addr_only, sw_if_index, proto, is_add); @@ -1937,22 +1958,27 @@ add_static_mapping_command_fn (vlib_main_t * vm, switch (rv) { case VNET_API_ERROR_INVALID_VALUE: - return clib_error_return (0, "External port already in use."); - break; + error = clib_error_return (0, "External port already in use."); + goto done; case VNET_API_ERROR_NO_SUCH_ENTRY: if (is_add) - return clib_error_return (0, "External addres must be allocated."); + error = clib_error_return (0, "External addres must be allocated."); else - return clib_error_return (0, "Mapping not exist."); - break; + error = clib_error_return (0, "Mapping not exist."); + goto done; case VNET_API_ERROR_NO_SUCH_FIB: - return clib_error_return (0, "No such VRF id."); + error = clib_error_return (0, "No such VRF id."); + goto done; case VNET_API_ERROR_VALUE_EXIST: - return clib_error_return (0, "Mapping already exist."); + error = clib_error_return (0, "Mapping already exist."); + goto done; default: break; } +done: + unformat_free (line_input); + return error; } @@ -1985,6 +2011,7 @@ set_workers_command_fn (vlib_main_t * vm, unformat_input_t _line_input, *line_input = &_line_input; uword *bitmap = 0; int rv = 0; + clib_error_t *error = 0; /* Get a line of input. */ if (!unformat_user (input, unformat_line_input, line_input)) @@ -1995,13 +2022,18 @@ set_workers_command_fn (vlib_main_t * vm, if (unformat (line_input, "%U", unformat_bitmap_list, &bitmap)) ; else - return clib_error_return (0, "unknown input '%U'", - format_unformat_error, input); + { + error = clib_error_return (0, "unknown input '%U'", + format_unformat_error, line_input); + goto done; + } } - unformat_free (line_input); if (bitmap == 0) - return clib_error_return (0, "List of workers must be specified."); + { + error = clib_error_return (0, "List of workers must be specified."); + goto done; + } rv = snat_set_workers(bitmap); @@ -2010,17 +2042,20 @@ set_workers_command_fn (vlib_main_t * vm, switch (rv) { case VNET_API_ERROR_INVALID_WORKER: - return clib_error_return (0, "Invalid worker(s)."); - break; + error = clib_error_return (0, "Invalid worker(s)."); + goto done; case VNET_API_ERROR_FEATURE_DISABLED: - return clib_error_return (0, + error = clib_error_return (0, "Supported only if 2 or more workes available."); - break; + goto done; default: break; } - return 0; +done: + unformat_free (line_input); + + return error; } /*? @@ -2047,6 +2082,7 @@ snat_ipfix_logging_enable_disable_command_fn (vlib_main_t * vm, u32 src_port = 0; u8 enable = 1; int rv = 0; + clib_error_t *error = 0; /* Get a line of input. */ if (!unformat_user (input, unformat_line_input, line_input)) @@ -2061,17 +2097,25 @@ snat_ipfix_logging_enable_disable_command_fn (vlib_main_t * vm, else if (unformat (line_input, "disable")) enable = 0; else - return clib_error_return (0, "unknown input '%U'", - format_unformat_error, input); + { + error = clib_error_return (0, "unknown input '%U'", + format_unformat_error, line_input); + goto done; + } } - unformat_free (line_input); rv = snat_ipfix_logging_enable_disable (enable, domain_id, (u16) src_port); if (rv) - return clib_error_return (0, "ipfix logging enable failed"); + { + error = clib_error_return (0, "ipfix logging enable failed"); + goto done; + } - return 0; +done: + unformat_free (line_input); + + return error; } /*? @@ -2604,6 +2648,7 @@ snat_add_interface_address_command_fn (vlib_main_t * vm, u32 sw_if_index; int rv; int is_del = 0; + clib_error_t *error = 0; /* Get a line of input. */ if (!unformat_user (input, unformat_line_input, line_input)) @@ -2617,8 +2662,11 @@ snat_add_interface_address_command_fn (vlib_main_t * vm, else if (unformat (line_input, "del")) is_del = 1; else - return clib_error_return (0, "unknown input '%U'", - format_unformat_error, line_input); + { + error = clib_error_return (0, "unknown input '%U'", + format_unformat_error, line_input); + goto done; + } } rv = snat_add_interface_address (sm, sw_if_index, is_del); @@ -2629,10 +2677,15 @@ snat_add_interface_address_command_fn (vlib_main_t * vm, break; default: - return clib_error_return (0, "snat_add_interface_address returned %d", - rv); + error = clib_error_return (0, "snat_add_interface_address returned %d", + rv); + goto done; } - return 0; + +done: + unformat_free (line_input); + + return error; } VLIB_CLI_COMMAND (snat_add_interface_address_command, static) = { diff --git a/src/vlib/threads_cli.c b/src/vlib/threads_cli.c index 54cc1aed..36f8109e 100644 --- a/src/vlib/threads_cli.c +++ b/src/vlib/threads_cli.c @@ -163,21 +163,31 @@ trace_frame_queue (vlib_main_t * vm, unformat_input_t * input, else if (unformat (line_input, "index %u", &index)) ; else - return clib_error_return (0, "parse error: '%U'", - format_unformat_error, line_input); + { + error = clib_error_return (0, "parse error: '%U'", + format_unformat_error, line_input); + goto done; + } } - unformat_free (line_input); - if (enable > 1) - return clib_error_return (0, "expecting on or off"); + { + error = clib_error_return (0, "expecting on or off"); + goto done; + } if (vec_len (tm->frame_queue_mains) == 0) - return clib_error_return (0, "no worker handoffs exist"); + { + error = clib_error_return (0, "no worker handoffs exist"); + goto done; + } if (index > vec_len (tm->frame_queue_mains) - 1) - return clib_error_return (0, - "expecting valid worker handoff queue index"); + { + error = clib_error_return (0, + "expecting valid worker handoff queue index"); + goto done; + } fqm = vec_elt_at_index (tm->frame_queue_mains, index); @@ -185,7 +195,7 @@ trace_frame_queue (vlib_main_t * vm, unformat_input_t * input, if (num_fq == 0) { vlib_cli_output (vm, "No frame queues exist\n"); - return error; + goto done; } // Allocate storage for trace if necessary @@ -204,6 +214,10 @@ trace_frame_queue (vlib_main_t * vm, unformat_input_t * input, memset (fqh, 0, sizeof (*fqh)); fqm->vlib_frame_queues[fqix]->trace = enable; } + +done: + unformat_free (line_input); + return error; } @@ -432,28 +446,33 @@ test_frame_queue_nelts (vlib_main_t * vm, unformat_input_t * input, else if (unformat (line_input, "index %u", &index)) ; else - return clib_error_return (0, "parse error: '%U'", - format_unformat_error, line_input); + { + error = clib_error_return (0, "parse error: '%U'", + format_unformat_error, line_input); + goto done; + } } - unformat_free (line_input); - if (index > vec_len (tm->frame_queue_mains) - 1) - return clib_error_return (0, - "expecting valid worker handoff queue index"); + { + error = clib_error_return (0, + "expecting valid worker handoff queue index"); + goto done; + } fqm = vec_elt_at_index (tm->frame_queue_mains, index); if ((nelts != 4) && (nelts != 8) && (nelts != 16) && (nelts != 32)) { - return clib_error_return (0, "expecting 4,8,16,32"); + error = clib_error_return (0, "expecting 4,8,16,32"); + goto done; } num_fq = vec_len (fqm->vlib_frame_queues); if (num_fq == 0) { vlib_cli_output (vm, "No frame queues exist\n"); - return error; + goto done; } for (fqix = 0; fqix < num_fq; fqix++) @@ -461,6 +480,9 @@ test_frame_queue_nelts (vlib_main_t * vm, unformat_input_t * input, fqm->vlib_frame_queues[fqix]->nelts = nelts; } +done: + unformat_free (line_input); + return error; } @@ -499,15 +521,19 @@ test_frame_queue_threshold (vlib_main_t * vm, unformat_input_t * input, else if (unformat (line_input, "index %u", &index)) ; else - return clib_error_return (0, "parse error: '%U'", - format_unformat_error, line_input); + { + error = clib_error_return (0, "parse error: '%U'", + format_unformat_error, line_input); + goto done; + } } - unformat_free (line_input); - if (index > vec_len (tm->frame_queue_mains) - 1) - return clib_error_return (0, - "expecting valid worker handoff queue index"); + { + error = clib_error_return (0, + "expecting valid worker handoff queue index"); + goto done; + } fqm = vec_elt_at_index (tm->frame_queue_mains, index); @@ -515,7 +541,7 @@ test_frame_queue_threshold (vlib_main_t * vm, unformat_input_t * input, if (threshold == ~(u32) 0) { vlib_cli_output (vm, "expecting threshold value\n"); - return error; + goto done; } if (threshold == 0) @@ -525,7 +551,7 @@ test_frame_queue_threshold (vlib_main_t * vm, unformat_input_t * input, if (num_fq == 0) { vlib_cli_output (vm, "No frame queues exist\n"); - return error; + goto done; } for (fqix = 0; fqix < num_fq; fqix++) @@ -533,6 +559,9 @@ test_frame_queue_threshold (vlib_main_t * vm, unformat_input_t * input, fqm->vlib_frame_queues[fqix]->vector_threshold = threshold; } +done: + unformat_free (line_input); + return error; } diff --git a/src/vlib/trace.c b/src/vlib/trace.c index dcdb837f..6d487ae1 100644 --- a/src/vlib/trace.c +++ b/src/vlib/trace.c @@ -372,6 +372,7 @@ cli_add_trace_buffer (vlib_main_t * vm, vlib_trace_node_t *tn; u32 node_index, add; u8 verbose = 0; + clib_error_t *error = 0; if (!unformat_user (input, unformat_line_input, line_input)) return 0; @@ -384,8 +385,11 @@ cli_add_trace_buffer (vlib_main_t * vm, else if (unformat (line_input, "verbose")) verbose = 1; else - return clib_error_create ("expected NODE COUNT, got `%U'", - format_unformat_error, line_input); + { + error = clib_error_create ("expected NODE COUNT, got `%U'", + format_unformat_error, line_input); + goto done; + } } /* *INDENT-OFF* */ @@ -403,7 +407,10 @@ cli_add_trace_buffer (vlib_main_t * vm, })); /* *INDENT-ON* */ - return 0; +done: + unformat_free (line_input); + + return error; } /* *INDENT-OFF* */ diff --git a/src/vlib/unix/cli.c b/src/vlib/unix/cli.c index 69fca6ec..88e2453c 100644 --- a/src/vlib/unix/cli.c +++ b/src/vlib/unix/cli.c @@ -2835,6 +2835,7 @@ unix_cli_set_terminal_pager (vlib_main_t * vm, unix_cli_main_t *cm = &unix_cli_main; unix_cli_file_t *cf; unformat_input_t _line_input, *line_input = &_line_input; + clib_error_t *error = 0; if (!unformat_user (input, unformat_line_input, line_input)) return 0; @@ -2852,13 +2853,17 @@ unix_cli_set_terminal_pager (vlib_main_t * vm, "Pager limit set to %u lines; note, this is global.\n", um->cli_pager_buffer_limit); else - return clib_error_return (0, "unknown parameter: `%U`", - format_unformat_error, line_input); + { + error = clib_error_return (0, "unknown parameter: `%U`", + format_unformat_error, line_input); + goto done; + } } +done: unformat_free (line_input); - return 0; + return error; } /*? @@ -2886,6 +2891,7 @@ unix_cli_set_terminal_history (vlib_main_t * vm, unix_cli_file_t *cf; unformat_input_t _line_input, *line_input = &_line_input; u32 limit; + clib_error_t *error = 0; if (!unformat_user (input, unformat_line_input, line_input)) return 0; @@ -2901,8 +2907,11 @@ unix_cli_set_terminal_history (vlib_main_t * vm, else if (unformat (line_input, "limit %u", &cf->history_limit)) ; else - return clib_error_return (0, "unknown parameter: `%U`", - format_unformat_error, line_input); + { + error = clib_error_return (0, "unknown parameter: `%U`", + format_unformat_error, line_input); + goto done; + } /* If we reduced history size, or turned it off, purge the history */ limit = cf->has_history ? cf->history_limit : 0; @@ -2914,9 +2923,10 @@ unix_cli_set_terminal_history (vlib_main_t * vm, } } +done: unformat_free (line_input); - return 0; + return error; } /*? diff --git a/src/vnet/devices/af_packet/cli.c b/src/vnet/devices/af_packet/cli.c index 6baa26e1..d4aa7016 100644 --- a/src/vnet/devices/af_packet/cli.c +++ b/src/vnet/devices/af_packet/cli.c @@ -49,6 +49,7 @@ af_packet_create_command_fn (vlib_main_t * vm, unformat_input_t * input, u8 *hw_addr_ptr = 0; u32 sw_if_index; int r; + clib_error_t *error = NULL; /* Get a line of input. */ if (!unformat_user (input, unformat_line_input, line_input)) @@ -63,29 +64,47 @@ af_packet_create_command_fn (vlib_main_t * vm, unformat_input_t * input, (line_input, "hw-addr %U", unformat_ethernet_address, hwaddr)) hw_addr_ptr = hwaddr; else - return clib_error_return (0, "unknown input `%U'", - format_unformat_error, input); + { + error = clib_error_return (0, "unknown input `%U'", + format_unformat_error, line_input); + goto done; + } } - unformat_free (line_input); if (host_if_name == NULL) - return clib_error_return (0, "missing host interface name"); + { + error = clib_error_return (0, "missing host interface name"); + goto done; + } r = af_packet_create_if (vm, host_if_name, hw_addr_ptr, &sw_if_index); - vec_free (host_if_name); if (r == VNET_API_ERROR_SYSCALL_ERROR_1) - return clib_error_return (0, "%s (errno %d)", strerror (errno), errno); + { + error = clib_error_return (0, "%s (errno %d)", strerror (errno), errno); + goto done; + } if (r == VNET_API_ERROR_INVALID_INTERFACE) - return clib_error_return (0, "Invalid interface name"); + { + error = clib_error_return (0, "Invalid interface name"); + goto done; + } if (r == VNET_API_ERROR_SUBIF_ALREADY_EXISTS) - return clib_error_return (0, "Interface elready exists"); + { + error = clib_error_return (0, "Interface elready exists"); + goto done; + } vlib_cli_output (vm, "%U\n", format_vnet_sw_if_index_name, vnet_get_main (), sw_if_index); - return 0; + +done: + vec_free (host_if_name); + unformat_free (line_input); + + return error; } /*? @@ -124,6 +143,7 @@ af_packet_delete_command_fn (vlib_main_t * vm, unformat_input_t * input, { unformat_input_t _line_input, *line_input = &_line_input; u8 *host_if_name = NULL; + clib_error_t *error = NULL; /* Get a line of input. */ if (!unformat_user (input, unformat_line_input, line_input)) @@ -134,18 +154,26 @@ af_packet_delete_command_fn (vlib_main_t * vm, unformat_input_t * input, if (unformat (line_input, "name %s", &host_if_name)) ; else - return clib_error_return (0, "unknown input `%U'", - format_unformat_error, input); + { + error = clib_error_return (0, "unknown input `%U'", + format_unformat_error, line_input); + goto done; + } } - unformat_free (line_input); if (host_if_name == NULL) - return clib_error_return (0, "missing host interface name"); + { + error = clib_error_return (0, "missing host interface name"); + goto done; + } af_packet_delete_if (vm, host_if_name); + +done: vec_free (host_if_name); + unformat_free (line_input); - return 0; + return error; } /*? diff --git a/src/vnet/devices/dpdk/cli.c b/src/vnet/devices/dpdk/cli.c index d133cfd9..1fc665ac 100644 --- a/src/vnet/devices/dpdk/cli.c +++ b/src/vnet/devices/dpdk/cli.c @@ -398,7 +398,7 @@ set_dpdk_if_desc (vlib_main_t * vm, unformat_input_t * input, u32 hw_if_index = (u32) ~ 0; u32 nb_rx_desc = (u32) ~ 0; u32 nb_tx_desc = (u32) ~ 0; - clib_error_t *rv; + clib_error_t *error = NULL; if (!unformat_user (input, unformat_line_input, line_input)) return 0; @@ -414,25 +414,37 @@ set_dpdk_if_desc (vlib_main_t * vm, unformat_input_t * input, else if (unformat (line_input, "rx %d", &nb_rx_desc)) ; else - return clib_error_return (0, "parse error: '%U'", - format_unformat_error, line_input); + { + error = clib_error_return (0, "parse error: '%U'", + format_unformat_error, line_input); + goto done; + } } - unformat_free (line_input); - if (hw_if_index == (u32) ~ 0) - return clib_error_return (0, "please specify valid interface name"); + { + error = clib_error_return (0, "please specify valid interface name"); + goto done; + } hw = vnet_get_hw_interface (dm->vnet_main, hw_if_index); xd = vec_elt_at_index (dm->devices, hw->dev_instance); if ((xd->flags & DPDK_DEVICE_FLAG_PMD) == 0) - return clib_error_return (0, "number of descriptors can be set only for " - "physical devices"); + { + error = + clib_error_return (0, + "number of descriptors can be set only for " + "physical devices"); + goto done; + } if ((nb_rx_desc == (u32) ~ 0 || nb_rx_desc == xd->nb_rx_desc) && (nb_tx_desc == (u32) ~ 0 || nb_tx_desc == xd->nb_tx_desc)) - return clib_error_return (0, "nothing changed"); + { + error = clib_error_return (0, "nothing changed"); + goto done; + } if (nb_rx_desc != (u32) ~ 0) xd->nb_rx_desc = nb_rx_desc; @@ -440,9 +452,12 @@ set_dpdk_if_desc (vlib_main_t * vm, unformat_input_t * input, if (nb_tx_desc != (u32) ~ 0) xd->nb_tx_desc = nb_tx_desc; - rv = dpdk_port_setup (dm, xd); + error = dpdk_port_setup (dm, xd); + +done: + unformat_free (line_input); - return rv; + return error; } /* *INDENT-OFF* */ @@ -523,6 +538,7 @@ set_dpdk_if_placement (vlib_main_t * vm, unformat_input_t * input, u32 queue = (u32) 0; u32 cpu = (u32) ~ 0; int i; + clib_error_t *error = NULL; if (!unformat_user (input, unformat_line_input, line_input)) return 0; @@ -538,18 +554,25 @@ set_dpdk_if_placement (vlib_main_t * vm, unformat_input_t * input, else if (unformat (line_input, "thread %d", &cpu)) ; else - return clib_error_return (0, "parse error: '%U'", - format_unformat_error, line_input); + { + error = clib_error_return (0, "parse error: '%U'", + format_unformat_error, line_input); + goto done; + } } - unformat_free (line_input); - if (hw_if_index == (u32) ~ 0) - return clib_error_return (0, "please specify valid interface name"); + { + error = clib_error_return (0, "please specify valid interface name"); + goto done; + } if (cpu < dm->input_cpu_first_index || cpu >= (dm->input_cpu_first_index + dm->input_cpu_count)) - return clib_error_return (0, "please specify valid thread id"); + { + error = clib_error_return (0, "please specify valid thread id"); + goto done; + } hw = vnet_get_hw_interface (dm->vnet_main, hw_if_index); xd = vec_elt_at_index (dm->devices, hw->dev_instance); @@ -563,7 +586,7 @@ set_dpdk_if_placement (vlib_main_t * vm, unformat_input_t * input, queue == dq->queue_id) { if (cpu == i) /* nothing to do */ - return 0; + goto done; vec_del1(dm->devices_by_cpu[i], dq - dm->devices_by_cpu[i]); vec_add2(dm->devices_by_cpu[cpu], dq, 1); @@ -586,13 +609,18 @@ set_dpdk_if_placement (vlib_main_t * vm, unformat_input_t * input, vlib_node_set_state (vlib_mains[cpu], dpdk_input_node.index, VLIB_NODE_STATE_POLLING); - return 0; + goto done; } } /* *INDENT-ON* */ } - return clib_error_return (0, "not found"); + error = clib_error_return (0, "not found"); + +done: + unformat_free (line_input); + + return error; } /* *INDENT-OFF* */ @@ -653,6 +681,7 @@ set_dpdk_if_hqos_placement (vlib_main_t * vm, unformat_input_t * input, u32 hw_if_index = (u32) ~ 0; u32 cpu = (u32) ~ 0; int i; + clib_error_t *error = NULL; if (!unformat_user (input, unformat_line_input, line_input)) return 0; @@ -666,18 +695,22 @@ set_dpdk_if_hqos_placement (vlib_main_t * vm, unformat_input_t * input, else if (unformat (line_input, "thread %d", &cpu)) ; else - return clib_error_return (0, "parse error: '%U'", - format_unformat_error, line_input); + { + error = clib_error_return (0, "parse error: '%U'", + format_unformat_error, line_input); + goto done; + } } - unformat_free (line_input); - if (hw_if_index == (u32) ~ 0) return clib_error_return (0, "please specify valid interface name"); if (cpu < dm->hqos_cpu_first_index || cpu >= (dm->hqos_cpu_first_index + dm->hqos_cpu_count)) - return clib_error_return (0, "please specify valid thread id"); + { + error = clib_error_return (0, "please specify valid thread id"); + goto done; + } hw = vnet_get_hw_interface (dm->vnet_main, hw_if_index); xd = vec_elt_at_index (dm->devices, hw->dev_instance); @@ -689,7 +722,7 @@ set_dpdk_if_hqos_placement (vlib_main_t * vm, unformat_input_t * input, if (hw_if_index == dm->devices[dq->device].vlib_hw_if_index) { if (cpu == i) /* nothing to do */ - return 0; + goto done; vec_del1 (dm->devices_by_hqos_cpu[i], dq - dm->devices_by_hqos_cpu[i]); @@ -703,12 +736,17 @@ set_dpdk_if_hqos_placement (vlib_main_t * vm, unformat_input_t * input, vec_sort_with_function (dm->devices_by_hqos_cpu[cpu], dpdk_device_queue_sort); - return 0; + goto done; } } } - return clib_error_return (0, "not found"); + error = clib_error_return (0, "not found"); + +done: + unformat_free (line_input); + + return error; } /* *INDENT-OFF* */ @@ -732,6 +770,7 @@ set_dpdk_if_hqos_pipe (vlib_main_t * vm, unformat_input_t * input, u32 pipe_id = (u32) ~ 0; u32 profile_id = (u32) ~ 0; int rv; + clib_error_t *error = NULL; if (!unformat_user (input, unformat_line_input, line_input)) return 0; @@ -749,14 +788,18 @@ set_dpdk_if_hqos_pipe (vlib_main_t * vm, unformat_input_t * input, else if (unformat (line_input, "profile %d", &profile_id)) ; else - return clib_error_return (0, "parse error: '%U'", - format_unformat_error, line_input); + { + error = clib_error_return (0, "parse error: '%U'", + format_unformat_error, line_input); + goto done; + } } - unformat_free (line_input); - if (hw_if_index == (u32) ~ 0) - return clib_error_return (0, "please specify valid interface name"); + { + error = clib_error_return (0, "please specify valid interface name"); + goto done; + } hw = vnet_get_hw_interface (dm->vnet_main, hw_if_index); xd = vec_elt_at_index (dm->devices, hw->dev_instance); @@ -765,9 +808,15 @@ set_dpdk_if_hqos_pipe (vlib_main_t * vm, unformat_input_t * input, rte_sched_pipe_config (xd->hqos_ht->hqos, subport_id, pipe_id, profile_id); if (rv) - return clib_error_return (0, "pipe configuration failed"); + { + error = clib_error_return (0, "pipe configuration failed"); + goto done; + } - return 0; +done: + unformat_free (line_input); + + return error; } /* *INDENT-OFF* */ @@ -797,6 +846,7 @@ set_dpdk_if_hqos_subport (vlib_main_t * vm, unformat_input_t * input, .tc_period = 10, }; int rv; + clib_error_t *error = NULL; if (!unformat_user (input, unformat_line_input, line_input)) return 0; @@ -829,23 +879,33 @@ set_dpdk_if_hqos_subport (vlib_main_t * vm, unformat_input_t * input, else if (unformat (line_input, "period %d", &p.tc_period)) ; else - return clib_error_return (0, "parse error: '%U'", - format_unformat_error, line_input); + { + error = clib_error_return (0, "parse error: '%U'", + format_unformat_error, line_input); + goto done; + } } - unformat_free (line_input); - if (hw_if_index == (u32) ~ 0) - return clib_error_return (0, "please specify valid interface name"); + { + error = clib_error_return (0, "please specify valid interface name"); + goto done; + } hw = vnet_get_hw_interface (dm->vnet_main, hw_if_index); xd = vec_elt_at_index (dm->devices, hw->dev_instance); rv = rte_sched_subport_config (xd->hqos_ht->hqos, subport_id, &p); if (rv) - return clib_error_return (0, "subport configuration failed"); + { + error = clib_error_return (0, "subport configuration failed"); + goto done; + } - return 0; +done: + unformat_free (line_input); + + return error; } /* *INDENT-OFF* */ @@ -872,6 +932,7 @@ set_dpdk_if_hqos_tctbl (vlib_main_t * vm, unformat_input_t * input, u32 queue = (u32) ~ 0; u32 entry = (u32) ~ 0; u32 val, i; + clib_error_t *error = NULL; if (!unformat_user (input, unformat_line_input, line_input)) return 0; @@ -889,20 +950,33 @@ set_dpdk_if_hqos_tctbl (vlib_main_t * vm, unformat_input_t * input, else if (unformat (line_input, "queue %d", &queue)) ; else - return clib_error_return (0, "parse error: '%U'", - format_unformat_error, line_input); + { + error = clib_error_return (0, "parse error: '%U'", + format_unformat_error, line_input); + goto done; + } } - unformat_free (line_input); - if (hw_if_index == (u32) ~ 0) - return clib_error_return (0, "please specify valid interface name"); + { + error = clib_error_return (0, "please specify valid interface name"); + goto done; + } if (entry >= 64) - return clib_error_return (0, "invalid entry"); + { + error = clib_error_return (0, "invalid entry"); + goto done; + } if (tc >= RTE_SCHED_TRAFFIC_CLASSES_PER_PIPE) - return clib_error_return (0, "invalid traffic class"); + { + error = clib_error_return (0, "invalid traffic class"); + goto done; + } if (queue >= RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS) - return clib_error_return (0, "invalid traffic class"); + { + error = clib_error_return (0, "invalid traffic class"); + goto done; + } hw = vnet_get_hw_interface (dm->vnet_main, hw_if_index); xd = vec_elt_at_index (dm->devices, hw->dev_instance); @@ -911,7 +985,10 @@ set_dpdk_if_hqos_tctbl (vlib_main_t * vm, unformat_input_t * input, uword *p = hash_get_mem (tm->thread_registrations_by_name, "workers"); /* Should never happen, shut up Coverity warning */ if (p == 0) - return clib_error_return (0, "no worker registrations?"); + { + error = clib_error_return (0, "no worker registrations?"); + goto done; + } vlib_thread_registration_t *tr = (vlib_thread_registration_t *) p[0]; int worker_thread_first = tr->first_index; @@ -921,7 +998,10 @@ set_dpdk_if_hqos_tctbl (vlib_main_t * vm, unformat_input_t * input, for (i = 0; i < worker_thread_count; i++) xd->hqos_wt[worker_thread_first + i].hqos_tc_table[entry] = val; - return 0; +done: + unformat_free (line_input); + + return error; } /* *INDENT-OFF* */ @@ -939,6 +1019,7 @@ set_dpdk_if_hqos_pktfield (vlib_main_t * vm, unformat_input_t * input, unformat_input_t _line_input, *line_input = &_line_input; vlib_thread_main_t *tm = vlib_get_thread_main (); dpdk_main_t *dm = &dpdk_main; + clib_error_t *error = NULL; /* Device specific data */ struct rte_eth_dev_info dev_info; @@ -984,15 +1065,19 @@ set_dpdk_if_hqos_pktfield (vlib_main_t * vm, unformat_input_t * input, else if (unformat (line_input, "mask %llx", &mask)) ; else - return clib_error_return (0, "parse error: '%U'", - format_unformat_error, line_input); + { + error = clib_error_return (0, "parse error: '%U'", + format_unformat_error, line_input); + goto done; + } } - unformat_free (line_input); - /* Get interface */ if (hw_if_index == (u32) ~ 0) - return clib_error_return (0, "please specify valid interface name"); + { + error = clib_error_return (0, "please specify valid interface name"); + goto done; + } hw = vnet_get_hw_interface (dm->vnet_main, hw_if_index); xd = vec_elt_at_index (dm->devices, hw->dev_instance); @@ -1019,7 +1104,7 @@ set_dpdk_if_hqos_pktfield (vlib_main_t * vm, unformat_input_t * input, if (devconf->hqos_enabled == 0) { vlib_cli_output (vm, "HQoS disabled for this interface"); - return 0; + goto done; } n_subports_per_port = devconf->hqos.port.n_subports_per_port; @@ -1028,27 +1113,39 @@ set_dpdk_if_hqos_pktfield (vlib_main_t * vm, unformat_input_t * input, /* Validate packet field configuration: id, offset and mask */ if (id >= 3) - return clib_error_return (0, "invalid packet field id"); + { + error = clib_error_return (0, "invalid packet field id"); + goto done; + } switch (id) { case 0: if (dpdk_hqos_validate_mask (mask, n_subports_per_port) != 0) - return clib_error_return (0, "invalid subport ID mask " - "(n_subports_per_port = %u)", - n_subports_per_port); + { + error = clib_error_return (0, "invalid subport ID mask " + "(n_subports_per_port = %u)", + n_subports_per_port); + goto done; + } break; case 1: if (dpdk_hqos_validate_mask (mask, n_pipes_per_subport) != 0) - return clib_error_return (0, "invalid pipe ID mask " - "(n_pipes_per_subport = %u)", - n_pipes_per_subport); + { + error = clib_error_return (0, "invalid pipe ID mask " + "(n_pipes_per_subport = %u)", + n_pipes_per_subport); + goto done; + } break; case 2: default: if (dpdk_hqos_validate_mask (mask, tctbl_size) != 0) - return clib_error_return (0, "invalid TC table index mask " - "(TC table size = %u)", tctbl_size); + { + error = clib_error_return (0, "invalid TC table index mask " + "(TC table size = %u)", tctbl_size); + goto done; + } } /* Propagate packet field configuration to all workers */ @@ -1075,7 +1172,10 @@ set_dpdk_if_hqos_pktfield (vlib_main_t * vm, unformat_input_t * input, __builtin_ctzll (mask); } - return 0; +done: + unformat_free (line_input); + + return error; } /* *INDENT-OFF* */ @@ -1106,6 +1206,7 @@ show_dpdk_if_hqos (vlib_main_t * vm, unformat_input_t * input, dpdk_device_config_t *devconf = 0; vlib_thread_registration_t *tr; uword *p = 0; + clib_error_t *error = NULL; if (!unformat_user (input, unformat_line_input, line_input)) return 0; @@ -1117,14 +1218,18 @@ show_dpdk_if_hqos (vlib_main_t * vm, unformat_input_t * input, &hw_if_index)) ; else - return clib_error_return (0, "parse error: '%U'", - format_unformat_error, line_input); + { + error = clib_error_return (0, "parse error: '%U'", + format_unformat_error, line_input); + goto done; + } } - unformat_free (line_input); - if (hw_if_index == (u32) ~ 0) - return clib_error_return (0, "please specify interface name!!"); + { + error = clib_error_return (0, "please specify interface name!!"); + goto done; + } hw = vnet_get_hw_interface (dm->vnet_main, hw_if_index); xd = vec_elt_at_index (dm->devices, hw->dev_instance); @@ -1151,7 +1256,7 @@ show_dpdk_if_hqos (vlib_main_t * vm, unformat_input_t * input, if (devconf->hqos_enabled == 0) { vlib_cli_output (vm, "HQoS disabled for this interface"); - return 0; + goto done; } /* Detect the set of worker threads */ @@ -1159,7 +1264,10 @@ show_dpdk_if_hqos (vlib_main_t * vm, unformat_input_t * input, /* Should never happen, shut up Coverity warning */ if (p == 0) - return clib_error_return (0, "no worker registrations?"); + { + error = clib_error_return (0, "no worker registrations?"); + goto done; + } tr = (vlib_thread_registration_t *) p[0]; @@ -1284,7 +1392,10 @@ show_dpdk_if_hqos (vlib_main_t * vm, unformat_input_t * input, } #endif - return 0; +done: + unformat_free (line_input); + + return error; } /* *INDENT-OFF* */ @@ -1315,6 +1426,7 @@ show_dpdk_hqos_queue_stats (vlib_main_t * vm, unformat_input_t * input, u32 qindex; struct rte_sched_queue_stats stats; u16 qlen; + clib_error_t *error = NULL; if (!unformat_user (input, unformat_line_input, line_input)) return 0; @@ -1339,14 +1451,18 @@ show_dpdk_hqos_queue_stats (vlib_main_t * vm, unformat_input_t * input, ; else - return clib_error_return (0, "parse error: '%U'", - format_unformat_error, line_input); + { + error = clib_error_return (0, "parse error: '%U'", + format_unformat_error, line_input); + goto done; + } } - unformat_free (line_input); - if (hw_if_index == (u32) ~ 0) - return clib_error_return (0, "please specify interface name!!"); + { + error = clib_error_return (0, "please specify interface name!!"); + goto done; + } hw = vnet_get_hw_interface (dm->vnet_main, hw_if_index); xd = vec_elt_at_index (dm->devices, hw->dev_instance); @@ -1373,7 +1489,7 @@ show_dpdk_hqos_queue_stats (vlib_main_t * vm, unformat_input_t * input, if (devconf->hqos_enabled == 0) { vlib_cli_output (vm, "HQoS disabled for this interface"); - return 0; + goto done; } /* @@ -1386,7 +1502,10 @@ show_dpdk_hqos_queue_stats (vlib_main_t * vm, unformat_input_t * input, if (rte_sched_queue_read_stats (xd->hqos_ht->hqos, qindex, &stats, &qlen) != 0) - return clib_error_return (0, "failed to read stats"); + { + error = clib_error_return (0, "failed to read stats"); + goto done; + } vlib_cli_output (vm, "%=24s%=16s", "Stats Parameter", "Value"); vlib_cli_output (vm, "%=24s%=16d", "Packets", stats.n_pkts); @@ -1399,7 +1518,10 @@ show_dpdk_hqos_queue_stats (vlib_main_t * vm, unformat_input_t * input, vlib_cli_output (vm, "%=24s%=16d", "Bytes dropped", stats.n_bytes_dropped); - return 0; +done: + unformat_free (line_input); + + return error; } /* *INDENT-OFF* */ diff --git a/src/vnet/devices/dpdk/ipsec/cli.c b/src/vnet/devices/dpdk/ipsec/cli.c index 93df4a64..f9d3a5d0 100644 --- a/src/vnet/devices/dpdk/ipsec/cli.c +++ b/src/vnet/devices/dpdk/ipsec/cli.c @@ -111,6 +111,7 @@ lcore_cryptodev_map_fn (vlib_main_t * vm, unformat_input_t * input, { unformat_input_t _line_input, *line_input = &_line_input; u16 detail = 0; + clib_error_t *error = NULL; if (!unformat_user (input, unformat_line_input, line_input)) return 0; @@ -120,15 +121,19 @@ lcore_cryptodev_map_fn (vlib_main_t * vm, unformat_input_t * input, if (unformat (line_input, "verbose")) detail = 1; else - return clib_error_return (0, "parse error: '%U'", - format_unformat_error, line_input); + { + error = clib_error_return (0, "parse error: '%U'", + format_unformat_error, line_input); + goto done; + } } - unformat_free (line_input); - dpdk_ipsec_show_mapping (vm, detail); - return 0; +done: + unformat_free (line_input); + + return error; } /* *INDENT-OFF* */ diff --git a/src/vnet/devices/netmap/cli.c b/src/vnet/devices/netmap/cli.c index 6157f27c..71363294 100644 --- a/src/vnet/devices/netmap/cli.c +++ b/src/vnet/devices/netmap/cli.c @@ -37,6 +37,7 @@ netmap_create_command_fn (vlib_main_t * vm, unformat_input_t * input, u8 is_pipe = 0; u8 is_master = 0; u32 sw_if_index = ~0; + clib_error_t *error = NULL; /* Get a line of input. */ if (!unformat_user (input, unformat_line_input, line_input)) @@ -57,30 +58,48 @@ netmap_create_command_fn (vlib_main_t * vm, unformat_input_t * input, else if (unformat (line_input, "slave")) is_master = 0; else - return clib_error_return (0, "unknown input `%U'", - format_unformat_error, input); + { + error = clib_error_return (0, "unknown input `%U'", + format_unformat_error, line_input); + goto done; + } } - unformat_free (line_input); if (host_if_name == NULL) - return clib_error_return (0, "missing host interface name"); + { + error = clib_error_return (0, "missing host interface name"); + goto done; + } r = netmap_create_if (vm, host_if_name, hw_addr_ptr, is_pipe, is_master, &sw_if_index); if (r == VNET_API_ERROR_SYSCALL_ERROR_1) - return clib_error_return (0, "%s (errno %d)", strerror (errno), errno); + { + error = clib_error_return (0, "%s (errno %d)", strerror (errno), errno); + goto done; + } if (r == VNET_API_ERROR_INVALID_INTERFACE) - return clib_error_return (0, "Invalid interface name"); + { + error = clib_error_return (0, "Invalid interface name"); + goto done; + } if (r == VNET_API_ERROR_SUBIF_ALREADY_EXISTS) - return clib_error_return (0, "Interface already exists"); + { + error = clib_error_return (0, "Interface already exists"); + goto done; + } vlib_cli_output (vm, "%U\n", format_vnet_sw_if_index_name, vnet_get_main (), sw_if_index); - return 0; + +done: + unformat_free (line_input); + + return error; } /*? @@ -144,6 +163,7 @@ netmap_delete_command_fn (vlib_main_t * vm, unformat_input_t * input, { unformat_input_t _line_input, *line_input = &_line_input; u8 *host_if_name = NULL; + clib_error_t *error = NULL; /* Get a line of input. */ if (!unformat_user (input, unformat_line_input, line_input)) @@ -154,17 +174,25 @@ netmap_delete_command_fn (vlib_main_t * vm, unformat_input_t * input, if (unformat (line_input, "name %s", &host_if_name)) ; else - return clib_error_return (0, "unknown input `%U'", - format_unformat_error, input); + { + error = clib_error_return (0, "unknown input `%U'", + format_unformat_error, line_input); + goto done; + } } - unformat_free (line_input); if (host_if_name == NULL) - return clib_error_return (0, "missing host interface name"); + { + error = clib_error_return (0, "missing host interface name"); + goto done; + } netmap_delete_if (vm, host_if_name); - return 0; +done: + unformat_free (line_input); + + return error; } /*? diff --git a/src/vnet/devices/virtio/vhost-user.c b/src/vnet/devices/virtio/vhost-user.c index 315daa77..c43f6e67 100644 --- a/src/vnet/devices/virtio/vhost-user.c +++ b/src/vnet/devices/virtio/vhost-user.c @@ -2682,6 +2682,7 @@ vhost_user_connect_command_fn (vlib_main_t * vm, u32 custom_dev_instance = ~0; u8 hwaddr[6]; u8 *hw = NULL; + clib_error_t *error = NULL; /* Get a line of input. */ if (!unformat_user (input, unformat_line_input, line_input)) @@ -2704,10 +2705,12 @@ vhost_user_connect_command_fn (vlib_main_t * vm, renumber = 1; } else - return clib_error_return (0, "unknown input `%U'", - format_unformat_error, input); + { + error = clib_error_return (0, "unknown input `%U'", + format_unformat_error, line_input); + goto done; + } } - unformat_free (line_input); vnet_main_t *vnm = vnet_get_main (); @@ -2716,14 +2719,18 @@ vhost_user_connect_command_fn (vlib_main_t * vm, is_server, &sw_if_index, feature_mask, renumber, custom_dev_instance, hw))) { - vec_free (sock_filename); - return clib_error_return (0, "vhost_user_create_if returned %d", rv); + error = clib_error_return (0, "vhost_user_create_if returned %d", rv); + goto done; } - vec_free (sock_filename); vlib_cli_output (vm, "%U\n", format_vnet_sw_if_index_name, vnet_get_main (), sw_if_index); - return 0; + +done: + vec_free (sock_filename); + unformat_free (line_input); + + return error; } clib_error_t * @@ -2734,6 +2741,7 @@ vhost_user_delete_command_fn (vlib_main_t * vm, unformat_input_t _line_input, *line_input = &_line_input; u32 sw_if_index = ~0; vnet_main_t *vnm = vnet_get_main (); + clib_error_t *error = NULL; /* Get a line of input. */ if (!unformat_user (input, unformat_line_input, line_input)) @@ -2751,15 +2759,25 @@ vhost_user_delete_command_fn (vlib_main_t * vm, vnet_get_sup_hw_interface (vnm, sw_if_index); if (hwif == NULL || vhost_user_dev_class.index != hwif->dev_class_index) - return clib_error_return (0, "Not a vhost interface"); + { + error = clib_error_return (0, "Not a vhost interface"); + goto done; + } } else - return clib_error_return (0, "unknown input `%U'", - format_unformat_error, input); + { + error = clib_error_return (0, "unknown input `%U'", + format_unformat_error, line_input); + goto done; + } } - unformat_free (line_input); + vhost_user_delete_if (vnm, vm, sw_if_index); - return 0; + +done: + unformat_free (line_input); + + return error; } int @@ -3286,6 +3304,7 @@ vhost_thread_command_fn (vlib_main_t * vm, u32 sw_if_index; u8 del = 0; int rv; + clib_error_t *error = NULL; /* Get a line of input. */ if (!unformat_user (input, unformat_line_input, line_input)) @@ -3295,9 +3314,9 @@ vhost_thread_command_fn (vlib_main_t * vm, (line_input, "%U %d", unformat_vnet_sw_interface, vnet_get_main (), &sw_if_index, &worker_thread_index)) { - unformat_free (line_input); - return clib_error_return (0, "unknown input `%U'", - format_unformat_error, input); + error = clib_error_return (0, "unknown input `%U'", + format_unformat_error, line_input); + goto done; } if (unformat (line_input, "del")) @@ -3305,9 +3324,16 @@ vhost_thread_command_fn (vlib_main_t * vm, if ((rv = vhost_user_thread_placement (sw_if_index, worker_thread_index, del))) - return clib_error_return (0, "vhost_user_thread_placement returned %d", - rv); - return 0; + { + error = clib_error_return (0, "vhost_user_thread_placement returned %d", + rv); + goto done; + } + +done: + unformat_free (line_input); + + return error; } diff --git a/src/vnet/gre/interface.c b/src/vnet/gre/interface.c index d624587d..d4476ac4 100644 --- a/src/vnet/gre/interface.c +++ b/src/vnet/gre/interface.c @@ -491,6 +491,7 @@ create_gre_tunnel_command_fn (vlib_main_t * vm, u32 num_m_args = 0; u8 is_add = 1; u32 sw_if_index; + clib_error_t *error = NULL; /* Get a line of input. */ if (! unformat_user (input, unformat_line_input, line_input)) @@ -508,16 +509,24 @@ create_gre_tunnel_command_fn (vlib_main_t * vm, else if (unformat (line_input, "teb")) teb = 1; else - return clib_error_return (0, "unknown input `%U'", - format_unformat_error, input); + { + error = clib_error_return (0, "unknown input `%U'", + format_unformat_error, line_input); + goto done; + } } - unformat_free (line_input); if (num_m_args < 2) - return clib_error_return (0, "mandatory argument(s) missing"); + { + error = clib_error_return (0, "mandatory argument(s) missing"); + goto done; + } if (memcmp (&src, &dst, sizeof(src)) == 0) - return clib_error_return (0, "src and dst are identical"); + { + error = clib_error_return (0, "src and dst are identical"); + goto done; + } memset (a, 0, sizeof (*a)); a->outer_fib_id = outer_fib_id; @@ -536,15 +545,21 @@ create_gre_tunnel_command_fn (vlib_main_t * vm, vlib_cli_output(vm, "%U\n", format_vnet_sw_if_index_name, vnet_get_main(), sw_if_index); break; case VNET_API_ERROR_INVALID_VALUE: - return clib_error_return (0, "GRE tunnel already exists..."); + error = clib_error_return (0, "GRE tunnel already exists..."); + goto done; case VNET_API_ERROR_NO_SUCH_FIB: - return clib_error_return (0, "outer fib ID %d doesn't exist\n", - outer_fib_id); + error = clib_error_return (0, "outer fib ID %d doesn't exist\n", + outer_fib_id); + goto done; default: - return clib_error_return (0, "vnet_gre_add_del_tunnel returned %d", rv); + error = clib_error_return (0, "vnet_gre_add_del_tunnel returned %d", rv); + goto done; } - return 0; +done: + unformat_free (line_input); + + return error; } VLIB_CLI_COMMAND (create_gre_tunnel_command, static) = { diff --git a/src/vnet/ip/ip4_source_check.c b/src/vnet/ip/ip4_source_check.c index d461cc88..3af32f2e 100644 --- a/src/vnet/ip/ip4_source_check.c +++ b/src/vnet/ip/ip4_source_check.c @@ -399,6 +399,8 @@ set_ip_source_check (vlib_main_t * vm, vnet_feature_enable_disable ("ip4-unicast", feature_name, sw_if_index, is_del == 0, &config, sizeof (config)); done: + unformat_free (line_input); + return error; } @@ -531,7 +533,9 @@ ip_source_check_accept (vlib_main_t * vm, } done: - return (error); + unformat_free (line_input); + + return error; } /*? diff --git a/src/vnet/ip/ip4_test.c b/src/vnet/ip/ip4_test.c index 45d17113..73dabfdc 100644 --- a/src/vnet/ip/ip4_test.c +++ b/src/vnet/ip/ip4_test.c @@ -143,8 +143,11 @@ thrash (vlib_main_t * vm, else if (unformat (line_input, "verbose")) verbose = 1; else - return clib_error_return (0, "unknown input `%U'", - format_unformat_error, line_input); + { + error = clib_error_return (0, "unknown input `%U'", + format_unformat_error, line_input); + goto done; + } } } @@ -178,7 +181,7 @@ thrash (vlib_main_t * vm, if (p == 0) { vlib_cli_output (vm, "Couldn't map fib id %d to fib index\n", table_id); - return 0; + goto done; } table_index = p[0]; @@ -294,7 +297,11 @@ thrash (vlib_main_t * vm, pool_free (tm->route_pool); } - return 0; + +done: + unformat_free (line_input); + + return error; } /*? diff --git a/src/vnet/ip/ip6_neighbor.c b/src/vnet/ip/ip6_neighbor.c index 7229591e..6b53137f 100644 --- a/src/vnet/ip/ip6_neighbor.c +++ b/src/vnet/ip/ip6_neighbor.c @@ -2923,7 +2923,10 @@ ip6_neighbor_cmd (vlib_main_t * vm, unformat_input_t * main_input, else if (unformat (line_input, "ra-lifetime")) { if (!unformat (line_input, "%d", &ra_lifetime)) - return (error = unformat_parse_error (line_input)); + { + error = unformat_parse_error (line_input); + goto done; + } use_lifetime = 1; break; } @@ -2931,13 +2934,19 @@ ip6_neighbor_cmd (vlib_main_t * vm, unformat_input_t * main_input, { if (!unformat (line_input, "%d %d", &ra_initial_count, &ra_initial_interval)) - return (error = unformat_parse_error (line_input)); + { + error = unformat_parse_error (line_input); + goto done; + } break; } else if (unformat (line_input, "ra-interval")) { if (!unformat (line_input, "%d", &ra_max_interval)) - return (error = unformat_parse_error (line_input)); + { + error = unformat_parse_error (line_input); + goto done; + } if (!unformat (line_input, "%d", &ra_min_interval)) ra_min_interval = 0; @@ -2949,7 +2958,10 @@ ip6_neighbor_cmd (vlib_main_t * vm, unformat_input_t * main_input, break; } else - return (unformat_parse_error (line_input)); + { + error = unformat_parse_error (line_input); + goto done; + } } if (add_radv_info) @@ -3006,7 +3018,10 @@ ip6_neighbor_cmd (vlib_main_t * vm, unformat_input_t * main_input, else if (unformat (line_input, "no-onlink")) no_onlink = 1; else - return (unformat_parse_error (line_input)); + { + error = unformat_parse_error (line_input); + goto done; + } } ip6_neighbor_ra_prefix (vm, sw_if_index, @@ -3018,9 +3033,9 @@ ip6_neighbor_cmd (vlib_main_t * vm, unformat_input_t * main_input, off_link, no_autoconfig, no_onlink, is_no); } +done: unformat_free (line_input); -done: return error; } diff --git a/src/vnet/ip/lookup.c b/src/vnet/ip/lookup.c index 0ef0e7a6..807b87b6 100644 --- a/src/vnet/ip/lookup.c +++ b/src/vnet/ip/lookup.c @@ -568,8 +568,6 @@ vnet_ip_route_cmd (vlib_main_t * vm, } } - unformat_free (line_input); - if (vec_len (prefixs) == 0) { error = @@ -704,6 +702,7 @@ done: vec_free (dpos); vec_free (prefixs); vec_free (rpaths); + unformat_free (line_input); return error; } @@ -872,8 +871,6 @@ vnet_ip_mroute_cmd (vlib_main_t * vm, } } - unformat_free (line_input); - if (~0 == table_id) { /* @@ -970,6 +967,8 @@ vnet_ip_mroute_cmd (vlib_main_t * vm, (scount * gcount) / (timet[1] - timet[0])); done: + unformat_free (line_input); + return error; } @@ -1149,24 +1148,37 @@ probe_neighbor_address (vlib_main_t * vm, is_ip4 = 0; } else - return clib_error_return (0, "unknown input '%U'", - format_unformat_error, line_input); + { + error = clib_error_return (0, "unknown input '%U'", + format_unformat_error, line_input); + goto done; + } } - unformat_free (line_input); - if (sw_if_index == ~0) - return clib_error_return (0, "Interface required, not set."); + { + error = clib_error_return (0, "Interface required, not set."); + goto done; + } if (address_set == 0) - return clib_error_return (0, "ip address required, not set."); + { + error = clib_error_return (0, "ip address required, not set."); + goto done; + } if (address_set > 1) - return clib_error_return (0, "Multiple ip addresses not supported."); + { + error = clib_error_return (0, "Multiple ip addresses not supported."); + goto done; + } if (is_ip4) error = ip4_probe_neighbor_wait (vm, &a4, sw_if_index, retry_count); else error = ip6_probe_neighbor_wait (vm, &a6, sw_if_index, retry_count); +done: + unformat_free (line_input); + return error; } diff --git a/src/vnet/ipsec-gre/interface.c b/src/vnet/ipsec-gre/interface.c index 3b6e4ac2..0772ce73 100644 --- a/src/vnet/ipsec-gre/interface.c +++ b/src/vnet/ipsec-gre/interface.c @@ -232,6 +232,7 @@ create_ipsec_gre_tunnel_command_fn (vlib_main_t * vm, vnet_ipsec_gre_add_del_tunnel_args_t _a, *a = &_a; int rv; u32 sw_if_index; + clib_error_t *error = NULL; /* Get a line of input. */ if (!unformat_user (input, unformat_line_input, line_input)) @@ -250,16 +251,24 @@ create_ipsec_gre_tunnel_command_fn (vlib_main_t * vm, else if (unformat (line_input, "remote-sa %d", &rsa)) num_m_args++; else - return clib_error_return (0, "unknown input `%U'", - format_unformat_error, input); + { + error = clib_error_return (0, "unknown input `%U'", + format_unformat_error, line_input); + goto done; + } } - unformat_free (line_input); if (num_m_args < 4) - return clib_error_return (0, "mandatory argument(s) missing"); + { + error = clib_error_return (0, "mandatory argument(s) missing"); + goto done; + } if (memcmp (&src, &dst, sizeof (src)) == 0) - return clib_error_return (0, "src and dst are identical"); + { + error = clib_error_return (0, "src and dst are identical"); + goto done; + } memset (a, 0, sizeof (*a)); a->is_add = is_add; @@ -277,14 +286,19 @@ create_ipsec_gre_tunnel_command_fn (vlib_main_t * vm, vnet_get_main (), sw_if_index); break; case VNET_API_ERROR_INVALID_VALUE: - return clib_error_return (0, "GRE tunnel already exists..."); + error = clib_error_return (0, "GRE tunnel already exists..."); + goto done; default: - return clib_error_return (0, - "vnet_ipsec_gre_add_del_tunnel returned %d", - rv); + error = clib_error_return (0, + "vnet_ipsec_gre_add_del_tunnel returned %d", + rv); + goto done; } - return 0; +done: + unformat_free (line_input); + + return error; } /* *INDENT-OFF* */ diff --git a/src/vnet/ipsec/ipsec_cli.c b/src/vnet/ipsec/ipsec_cli.c index 3c1e26f2..0e034402 100644 --- a/src/vnet/ipsec/ipsec_cli.c +++ b/src/vnet/ipsec/ipsec_cli.c @@ -32,6 +32,7 @@ set_interface_spd_command_fn (vlib_main_t * vm, u32 sw_if_index = (u32) ~ 0; u32 spd_id; int is_add = 1; + clib_error_t *error = NULL; if (!unformat_user (input, unformat_line_input, line_input)) return 0; @@ -43,14 +44,18 @@ set_interface_spd_command_fn (vlib_main_t * vm, else if (unformat (line_input, "del")) is_add = 0; else - return clib_error_return (0, "parse error: '%U'", - format_unformat_error, line_input); - - unformat_free (line_input); + { + error = clib_error_return (0, "parse error: '%U'", + format_unformat_error, line_input); + goto done; + } ipsec_set_interface_spd (vm, sw_if_index, spd_id, is_add); - return 0; +done: + unformat_free (line_input); + + return error; } /* *INDENT-OFF* */ @@ -72,7 +77,7 @@ ipsec_sa_add_del_command_fn (vlib_main_t * vm, ipsec_sa_t sa; int is_add = ~0; u8 *ck = 0, *ik = 0; - clib_error_t *err = 0; + clib_error_t *error = NULL; memset (&sa, 0, sizeof (sa)); @@ -90,8 +95,11 @@ ipsec_sa_add_del_command_fn (vlib_main_t * vm, else if (unformat (line_input, "esp")) sa.protocol = IPSEC_PROTOCOL_ESP; else if (unformat (line_input, "ah")) - //sa.protocol = IPSEC_PROTOCOL_AH; - return clib_error_return (0, "unsupported security protocol 'AH'"); + { + //sa.protocol = IPSEC_PROTOCOL_AH; + error = clib_error_return (0, "unsupported security protocol 'AH'"); + goto done; + } else if (unformat (line_input, "crypto-key %U", unformat_hex_string, &ck)) sa.crypto_key_len = vec_len (ck); @@ -102,8 +110,12 @@ ipsec_sa_add_del_command_fn (vlib_main_t * vm, { if (sa.crypto_alg < IPSEC_CRYPTO_ALG_AES_CBC_128 || sa.crypto_alg >= IPSEC_CRYPTO_N_ALG) - return clib_error_return (0, "unsupported crypto-alg: '%U'", - format_ipsec_crypto_alg, sa.crypto_alg); + { + error = clib_error_return (0, "unsupported crypto-alg: '%U'", + format_ipsec_crypto_alg, + sa.crypto_alg); + goto done; + } } else if (unformat (line_input, "integ-key %U", unformat_hex_string, &ik)) @@ -113,8 +125,12 @@ ipsec_sa_add_del_command_fn (vlib_main_t * vm, { if (sa.integ_alg < IPSEC_INTEG_ALG_SHA1_96 || sa.integ_alg >= IPSEC_INTEG_N_ALG) - return clib_error_return (0, "unsupported integ-alg: '%U'", - format_ipsec_integ_alg, sa.integ_alg); + { + error = clib_error_return (0, "unsupported integ-alg: '%U'", + format_ipsec_integ_alg, + sa.integ_alg); + goto done; + } } else if (unformat (line_input, "tunnel-src %U", unformat_ip4_address, &sa.tunnel_src_addr.ip4)) @@ -135,12 +151,13 @@ ipsec_sa_add_del_command_fn (vlib_main_t * vm, sa.is_tunnel_ip6 = 1; } else - return clib_error_return (0, "parse error: '%U'", - format_unformat_error, line_input); + { + error = clib_error_return (0, "parse error: '%U'", + format_unformat_error, line_input); + goto done; + } } - unformat_free (line_input); - if (sa.crypto_key_len > sizeof (sa.crypto_key)) sa.crypto_key_len = sizeof (sa.crypto_key); @@ -156,14 +173,17 @@ ipsec_sa_add_del_command_fn (vlib_main_t * vm, if (is_add) { ASSERT (im->cb.check_support_cb); - err = im->cb.check_support_cb (&sa); - if (err) - return err; + error = im->cb.check_support_cb (&sa); + if (error) + goto done; } ipsec_add_del_sa (vm, &sa, is_add); - return 0; +done: + unformat_free (line_input); + + return error; } /* *INDENT-OFF* */ @@ -183,6 +203,7 @@ ipsec_spd_add_del_command_fn (vlib_main_t * vm, unformat_input_t _line_input, *line_input = &_line_input; u32 spd_id = ~0; int is_add = ~0; + clib_error_t *error = NULL; if (!unformat_user (input, unformat_line_input, line_input)) return 0; @@ -196,18 +217,25 @@ ipsec_spd_add_del_command_fn (vlib_main_t * vm, else if (unformat (line_input, "%u", &spd_id)) ; else - return clib_error_return (0, "parse error: '%U'", - format_unformat_error, line_input); + { + error = clib_error_return (0, "parse error: '%U'", + format_unformat_error, line_input); + goto done; + } } - unformat_free (line_input); - if (spd_id == ~0) - return clib_error_return (0, "please specify SPD ID"); + { + error = clib_error_return (0, "please specify SPD ID"); + goto done; + } ipsec_add_del_spd (vm, spd_id, is_add); - return 0; +done: + unformat_free (line_input); + + return error; } /* *INDENT-OFF* */ @@ -230,6 +258,7 @@ ipsec_policy_add_del_command_fn (vlib_main_t * vm, int is_add = 0; int is_ip_any = 1; u32 tmp, tmp2; + clib_error_t *error = NULL; memset (&p, 0, sizeof (p)); p.lport.stop = p.rport.stop = ~0; @@ -262,7 +291,10 @@ ipsec_policy_add_del_command_fn (vlib_main_t * vm, &p.policy)) { if (p.policy == IPSEC_POLICY_ACTION_RESOLVE) - return clib_error_return (0, "unsupported action: 'resolve'"); + { + error = clib_error_return (0, "unsupported action: 'resolve'"); + goto done; + } } else if (unformat (line_input, "sa %u", &p.sa_id)) ; @@ -300,19 +332,24 @@ ipsec_policy_add_del_command_fn (vlib_main_t * vm, p.rport.stop = tmp2; } else - return clib_error_return (0, "parse error: '%U'", - format_unformat_error, line_input); + { + error = clib_error_return (0, "parse error: '%U'", + format_unformat_error, line_input); + goto done; + } } - unformat_free (line_input); - ipsec_add_del_policy (vm, &p, is_add); if (is_ip_any) { p.is_ipv6 = 1; ipsec_add_del_policy (vm, &p, is_add); } - return 0; + +done: + unformat_free (line_input); + + return error; } /* *INDENT-OFF* */ @@ -332,6 +369,7 @@ set_ipsec_sa_key_command_fn (vlib_main_t * vm, unformat_input_t _line_input, *line_input = &_line_input; ipsec_sa_t sa; u8 *ck = 0, *ik = 0; + clib_error_t *error = NULL; memset (&sa, 0, sizeof (sa)); @@ -349,12 +387,13 @@ set_ipsec_sa_key_command_fn (vlib_main_t * vm, if (unformat (line_input, "integ-key %U", unformat_hex_string, &ik)) sa.integ_key_len = vec_len (ik); else - return clib_error_return (0, "parse error: '%U'", - format_unformat_error, line_input); + { + error = clib_error_return (0, "parse error: '%U'", + format_unformat_error, line_input); + goto done; + } } - unformat_free (line_input); - if (sa.crypto_key_len > sizeof (sa.crypto_key)) sa.crypto_key_len = sizeof (sa.crypto_key); @@ -369,7 +408,10 @@ set_ipsec_sa_key_command_fn (vlib_main_t * vm, ipsec_set_sa_key (vm, &sa); - return 0; +done: + unformat_free (line_input); + + return error; } /* *INDENT-OFF* */ @@ -649,6 +691,7 @@ create_ipsec_tunnel_command_fn (vlib_main_t * vm, ipsec_add_del_tunnel_args_t a; int rv; u32 num_m_args = 0; + clib_error_t *error = NULL; memset (&a, 0, sizeof (a)); a.is_add = 1; @@ -673,13 +716,18 @@ create_ipsec_tunnel_command_fn (vlib_main_t * vm, else if (unformat (line_input, "del")) a.is_add = 0; else - return clib_error_return (0, "unknown input `%U'", - format_unformat_error, input); + { + error = clib_error_return (0, "unknown input `%U'", + format_unformat_error, line_input); + goto done; + } } - unformat_free (line_input); if (num_m_args < 4) - return clib_error_return (0, "mandatory argument(s) missing"); + { + error = clib_error_return (0, "mandatory argument(s) missing"); + goto done; + } rv = ipsec_add_del_tunnel_if (&a); @@ -689,16 +737,21 @@ create_ipsec_tunnel_command_fn (vlib_main_t * vm, break; case VNET_API_ERROR_INVALID_VALUE: if (a.is_add) - return clib_error_return (0, - "IPSec tunnel interface already exists..."); + error = clib_error_return (0, + "IPSec tunnel interface already exists..."); else - return clib_error_return (0, "IPSec tunnel interface not exists..."); + error = clib_error_return (0, "IPSec tunnel interface not exists..."); + goto done; default: - return clib_error_return (0, "ipsec_register_interface returned %d", - rv); + error = clib_error_return (0, "ipsec_register_interface returned %d", + rv); + goto done; } - return 0; +done: + unformat_free (line_input); + + return error; } /* *INDENT-OFF* */ @@ -720,6 +773,7 @@ set_interface_key_command_fn (vlib_main_t * vm, u32 hw_if_index = (u32) ~ 0; u32 alg; u8 *key = 0; + clib_error_t *error = NULL; if (!unformat_user (input, unformat_line_input, line_input)) return 0; @@ -748,25 +802,38 @@ set_interface_key_command_fn (vlib_main_t * vm, else if (unformat (line_input, "%U", unformat_hex_string, &key)) ; else - return clib_error_return (0, "parse error: '%U'", - format_unformat_error, line_input); + { + error = clib_error_return (0, "parse error: '%U'", + format_unformat_error, line_input); + goto done; + } } - unformat_free (line_input); - if (type == IPSEC_IF_SET_KEY_TYPE_NONE) - return clib_error_return (0, "unknown key type"); + { + error = clib_error_return (0, "unknown key type"); + goto done; + } if (alg > 0 && vec_len (key) == 0) - return clib_error_return (0, "key is not specified"); + { + error = clib_error_return (0, "key is not specified"); + goto done; + } if (hw_if_index == (u32) ~ 0) - return clib_error_return (0, "interface not specified"); + { + error = clib_error_return (0, "interface not specified"); + goto done; + } ipsec_set_interface_key (im->vnet_main, hw_if_index, type, alg, key); + +done: vec_free (key); + unformat_free (line_input); - return 0; + return error; } /* *INDENT-OFF* */ diff --git a/src/vnet/l2/l2_patch.c b/src/vnet/l2/l2_patch.c index 5e4691f4..ff3d2f3a 100644 --- a/src/vnet/l2/l2_patch.c +++ b/src/vnet/l2/l2_patch.c @@ -315,6 +315,7 @@ test_patch_command_fn (vlib_main_t * vm, int rx_set = 0; int tx_set = 0; int is_add = 1; + clib_error_t *error = NULL; /* Get a line of input. */ if (!unformat_user (input, unformat_line_input, line_input)) @@ -335,10 +336,16 @@ test_patch_command_fn (vlib_main_t * vm, } if (rx_set == 0) - return clib_error_return (0, "rx interface not set"); + { + error = clib_error_return (0, "rx interface not set"); + goto done; + } if (tx_set == 0) - return clib_error_return (0, "tx interface not set"); + { + error = clib_error_return (0, "tx interface not set"); + goto done; + } rv = vnet_l2_patch_add_del (rx_sw_if_index, tx_sw_if_index, is_add); @@ -348,17 +355,24 @@ test_patch_command_fn (vlib_main_t * vm, break; case VNET_API_ERROR_INVALID_SW_IF_INDEX: - return clib_error_return (0, "rx interface not a physical port"); + error = clib_error_return (0, "rx interface not a physical port"); + goto done; case VNET_API_ERROR_INVALID_SW_IF_INDEX_2: - return clib_error_return (0, "tx interface not a physical port"); + error = clib_error_return (0, "tx interface not a physical port"); + goto done; default: - return clib_error_return + error = clib_error_return (0, "WARNING: vnet_l2_patch_add_del returned %d", rv); + goto done; } - return 0; + +done: + unformat_free (line_input); + + return error; } /*? diff --git a/src/vnet/l2/l2_xcrw.c b/src/vnet/l2/l2_xcrw.c index 70610a85..d08a5d8f 100644 --- a/src/vnet/l2/l2_xcrw.c +++ b/src/vnet/l2/l2_xcrw.c @@ -409,6 +409,7 @@ set_l2_xcrw_command_fn (vlib_main_t * vm, u8 *rw = 0; vnet_main_t *vnm = vnet_get_main (); int rv; + clib_error_t *error = NULL; if (!unformat_user (input, unformat_line_input, line_input)) @@ -416,8 +417,11 @@ set_l2_xcrw_command_fn (vlib_main_t * vm, if (!unformat (line_input, "%U", unformat_vnet_sw_interface, vnm, &l2_sw_if_index)) - return clib_error_return (0, "unknown input '%U'", - format_unformat_error, line_input); + { + error = clib_error_return (0, "unknown input '%U'", + format_unformat_error, line_input); + goto done; + } while (unformat_check_input (line_input) != UNFORMAT_END_OF_INPUT) { @@ -436,7 +440,10 @@ set_l2_xcrw_command_fn (vlib_main_t * vm, } if (next_node_index == ~0) - return clib_error_return (0, "next node not specified"); + { + error = clib_error_return (0, "next node not specified"); + goto done; + } if (tx_fib_id != ~0) { @@ -448,7 +455,11 @@ set_l2_xcrw_command_fn (vlib_main_t * vm, p = hash_get (ip4_main.fib_index_by_table_id, tx_fib_id); if (p == 0) - return clib_error_return (0, "nonexistent tx_fib_id %d", tx_fib_id); + { + error = + clib_error_return (0, "nonexistent tx_fib_id %d", tx_fib_id); + goto done; + } tx_fib_index = p[0]; } @@ -463,16 +474,21 @@ set_l2_xcrw_command_fn (vlib_main_t * vm, break; case VNET_API_ERROR_INVALID_SW_IF_INDEX: - return clib_error_return (0, "%U not cross-connected", - format_vnet_sw_if_index_name, - vnm, l2_sw_if_index); + error = clib_error_return (0, "%U not cross-connected", + format_vnet_sw_if_index_name, + vnm, l2_sw_if_index); + goto done; + default: - return clib_error_return (0, "vnet_configure_l2_xcrw returned %d", rv); + error = clib_error_return (0, "vnet_configure_l2_xcrw returned %d", rv); + goto done; } +done: vec_free (rw); + unformat_free (line_input); - return 0; + return error; } /*? diff --git a/src/vnet/l2tp/l2tp.c b/src/vnet/l2tp/l2tp.c index a4531dab..2d323397 100644 --- a/src/vnet/l2tp/l2tp.c +++ b/src/vnet/l2tp/l2tp.c @@ -427,6 +427,7 @@ create_l2tpv3_tunnel_command_fn (vlib_main_t * vm, u32 sw_if_index; u32 encap_fib_id = ~0; u32 encap_fib_index = ~0; + clib_error_t *error = NULL; /* Get a line of input. */ if (!unformat_user (input, unformat_line_input, line_input)) @@ -455,18 +456,22 @@ create_l2tpv3_tunnel_command_fn (vlib_main_t * vm, else if (unformat (line_input, "l2-sublayer-present")) l2_sublayer_present = 1; else - return clib_error_return (0, "parse error: '%U'", - format_unformat_error, line_input); + { + error = clib_error_return (0, "parse error: '%U'", + format_unformat_error, line_input); + goto done; + } } - unformat_free (line_input); - if (encap_fib_id != ~0) { uword *p; ip6_main_t *im = &ip6_main; if (!(p = hash_get (im->fib_index_by_table_id, encap_fib_id))) - return clib_error_return (0, "No fib with id %d", encap_fib_id); + { + error = clib_error_return (0, "No fib with id %d", encap_fib_id); + goto done; + } encap_fib_index = p[0]; } else @@ -475,9 +480,15 @@ create_l2tpv3_tunnel_command_fn (vlib_main_t * vm, } if (our_address_set == 0) - return clib_error_return (0, "our address not specified"); + { + error = clib_error_return (0, "our address not specified"); + goto done; + } if (client_address_set == 0) - return clib_error_return (0, "client address not specified"); + { + error = clib_error_return (0, "client address not specified"); + goto done; + } rv = create_l2tpv3_ipv6_tunnel (lm, &client_address, &our_address, local_session_id, remote_session_id, @@ -491,16 +502,22 @@ create_l2tpv3_tunnel_command_fn (vlib_main_t * vm, vnet_get_main (), sw_if_index); break; case VNET_API_ERROR_INVALID_VALUE: - return clib_error_return (0, "session already exists..."); + error = clib_error_return (0, "session already exists..."); + goto done; case VNET_API_ERROR_NO_SUCH_ENTRY: - return clib_error_return (0, "session does not exist..."); + error = clib_error_return (0, "session does not exist..."); + goto done; default: - return clib_error_return (0, "l2tp_session_add_del returned %d", rv); + error = clib_error_return (0, "l2tp_session_add_del returned %d", rv); + goto done; } - return 0; +done: + unformat_free (line_input); + + return error; } /* *INDENT-OFF* */ diff --git a/src/vnet/lisp-cp/lisp_cli.c b/src/vnet/lisp-cp/lisp_cli.c index 25d11c61..05df9fb6 100644 --- a/src/vnet/lisp-cp/lisp_cli.c +++ b/src/vnet/lisp-cp/lisp_cli.c @@ -25,6 +25,7 @@ lisp_show_adjacencies_command_fn (vlib_main_t * vm, vlib_cli_output (vm, "%s %40s\n", "leid", "reid"); unformat_input_t _line_input, *line_input = &_line_input; u32 vni = ~0; + clib_error_t *error = NULL; /* Get a line of input. */ if (!unformat_user (input, unformat_line_input, line_input)) @@ -38,14 +39,14 @@ lisp_show_adjacencies_command_fn (vlib_main_t * vm, { vlib_cli_output (vm, "parse error: '%U'", format_unformat_error, line_input); - return 0; + goto done; } } if (~0 == vni) { vlib_cli_output (vm, "error: no vni specified!"); - return 0; + goto done; } adjs = vnet_lisp_adjacencies_get_by_vni (vni); @@ -57,7 +58,10 @@ lisp_show_adjacencies_command_fn (vlib_main_t * vm, } vec_free (adjs); - return 0; +done: + unformat_free (line_input); + + return error; } /* *INDENT-OFF* */ @@ -77,6 +81,7 @@ lisp_add_del_map_server_command_fn (vlib_main_t * vm, u8 is_add = 1, ip_set = 0; ip_address_t ip; unformat_input_t _line_input, *line_input = &_line_input; + clib_error_t *error = NULL; /* Get a line of input. */ if (!unformat_user (input, unformat_line_input, line_input)) @@ -94,14 +99,14 @@ lisp_add_del_map_server_command_fn (vlib_main_t * vm, { vlib_cli_output (vm, "parse error: '%U'", format_unformat_error, line_input); - return 0; + goto done; } } if (!ip_set) { vlib_cli_output (vm, "map-server ip address not set!"); - return 0; + goto done; } rv = vnet_lisp_add_del_map_server (&ip, is_add); @@ -109,7 +114,10 @@ lisp_add_del_map_server_command_fn (vlib_main_t * vm, vlib_cli_output (vm, "failed to %s map-server!", is_add ? "add" : "delete"); - return 0; +done: + unformat_free (line_input); + + return error; } /* *INDENT-OFF* */ @@ -191,7 +199,7 @@ lisp_add_del_local_eid_command_fn (vlib_main_t * vm, unformat_input_t * input, if (key && (0 == key_id)) { vlib_cli_output (vm, "invalid key_id!"); - return 0; + goto done;; } gid_address_copy (&a->eid, &eid); @@ -213,6 +221,8 @@ done: vec_free (locator_set_name); gid_address_free (&a->eid); vec_free (a->key); + unformat_free (line_input); + return error; } @@ -233,6 +243,7 @@ lisp_eid_table_map_command_fn (vlib_main_t * vm, u8 is_add = 1, is_l2 = 0; u32 vni = 0, dp_id = 0; unformat_input_t _line_input, *line_input = &_line_input; + clib_error_t *error = NULL; /* Get a line of input. */ if (!unformat_user (input, unformat_line_input, line_input)) @@ -250,11 +261,16 @@ lisp_eid_table_map_command_fn (vlib_main_t * vm, is_l2 = 1; else { - return unformat_parse_error (line_input); + error = unformat_parse_error (line_input); + goto done; } } vnet_lisp_eid_table_map (vni, dp_id, is_l2, is_add); - return 0; + +done: + unformat_free (line_input); + + return error; } /* *INDENT-OFF* */ @@ -479,7 +495,7 @@ lisp_add_del_adjacency_command_fn (vlib_main_t * vm, unformat_input_t * input, != ip_prefix_version (leid_ippref))) { clib_warning ("remote and local EIDs are of different types!"); - return error; + goto done; } memset (a, 0, sizeof (a[0])); @@ -512,6 +528,7 @@ lisp_map_request_mode_command_fn (vlib_main_t * vm, { unformat_input_t _i, *i = &_i; map_request_mode_t mr_mode = _MR_MODE_MAX; + clib_error_t *error = NULL; /* Get a line of input. */ if (!unformat_user (input, unformat_line_input, i)) @@ -533,12 +550,15 @@ lisp_map_request_mode_command_fn (vlib_main_t * vm, if (_MR_MODE_MAX == mr_mode) { clib_warning ("No LISP map request mode entered!"); - return 0; + goto done; } vnet_lisp_set_map_request_mode (mr_mode); + done: - return 0; + unformat_free (i); + + return error; } /* *INDENT-OFF* */ @@ -630,7 +650,10 @@ lisp_pitr_set_locator_set_command_fn (vlib_main_t * vm, else if (unformat (line_input, "disable")) is_add = 0; else - return clib_error_return (0, "parse error"); + { + error = clib_error_return (0, "parse error"); + goto done; + } } if (!locator_name_set) @@ -648,6 +671,8 @@ lisp_pitr_set_locator_set_command_fn (vlib_main_t * vm, done: if (locator_set_name) vec_free (locator_set_name); + unformat_free (line_input); + return error; } @@ -771,6 +796,7 @@ lisp_show_eid_table_command_fn (vlib_main_t * vm, gid_address_t eid; u8 print_all = 1; u8 filter = 0; + clib_error_t *error = NULL; memset (&eid, 0, sizeof (eid)); @@ -787,8 +813,11 @@ lisp_show_eid_table_command_fn (vlib_main_t * vm, else if (unformat (line_input, "remote")) filter = 2; else - return clib_error_return (0, "parse error: '%U'", - format_unformat_error, line_input); + { + error = clib_error_return (0, "parse error: '%U'", + format_unformat_error, line_input); + goto done; + } } vlib_cli_output (vm, "%-35s%-20s%-30s%-20s%-s", @@ -818,7 +847,7 @@ lisp_show_eid_table_command_fn (vlib_main_t * vm, { mi = gid_dictionary_lookup (&lcm->mapping_index_by_gid, &eid); if ((u32) ~ 0 == mi) - return 0; + goto done; mapit = pool_elt_at_index (lcm->mapping_pool, mi); locator_set_t *ls = pool_elt_at_index (lcm->locator_set_pool, @@ -827,14 +856,17 @@ lisp_show_eid_table_command_fn (vlib_main_t * vm, if (filter && !((1 == filter && ls->local) || (2 == filter && !ls->local))) { - return 0; + goto done; } vlib_cli_output (vm, "%U,", format_eid_entry, lcm->vnet_main, lcm, mapit, ls); } - return 0; +done: + unformat_free (line_input); + + return error; } /* *INDENT-OFF* */ @@ -853,6 +885,7 @@ lisp_enable_disable_command_fn (vlib_main_t * vm, unformat_input_t * input, unformat_input_t _line_input, *line_input = &_line_input; u8 is_enabled = 0; u8 is_set = 0; + clib_error_t *error = NULL; /* Get a line of input. */ if (!unformat_user (input, unformat_line_input, line_input)) @@ -869,16 +902,24 @@ lisp_enable_disable_command_fn (vlib_main_t * vm, unformat_input_t * input, is_set = 1; else { - return clib_error_return (0, "parse error: '%U'", - format_unformat_error, line_input); + error = clib_error_return (0, "parse error: '%U'", + format_unformat_error, line_input); + goto done; } } if (!is_set) - return clib_error_return (0, "state not set"); + { + error = clib_error_return (0, "state not set"); + goto done; + } vnet_lisp_enable_disable (is_enabled); - return 0; + +done: + unformat_free (line_input); + + return error; } /* *INDENT-OFF* */ @@ -897,6 +938,7 @@ lisp_map_register_enable_disable_command_fn (vlib_main_t * vm, unformat_input_t _line_input, *line_input = &_line_input; u8 is_enabled = 0; u8 is_set = 0; + clib_error_t *error = NULL; /* Get a line of input. */ if (!unformat_user (input, unformat_line_input, line_input)) @@ -915,18 +957,22 @@ lisp_map_register_enable_disable_command_fn (vlib_main_t * vm, { vlib_cli_output (vm, "parse error: '%U'", format_unformat_error, line_input); - return 0; + goto done; } } if (!is_set) { vlib_cli_output (vm, "state not set!"); - return 0; + goto done; } vnet_lisp_map_register_enable_disable (is_enabled); - return 0; + +done: + unformat_free (line_input); + + return error; } /* *INDENT-OFF* */ @@ -945,6 +991,7 @@ lisp_rloc_probe_enable_disable_command_fn (vlib_main_t * vm, unformat_input_t _line_input, *line_input = &_line_input; u8 is_enabled = 0; u8 is_set = 0; + clib_error_t *error = NULL; /* Get a line of input. */ if (!unformat_user (input, unformat_line_input, line_input)) @@ -963,18 +1010,22 @@ lisp_rloc_probe_enable_disable_command_fn (vlib_main_t * vm, { vlib_cli_output (vm, "parse error: '%U'", format_unformat_error, line_input); - return 0; + goto done; } } if (!is_set) { vlib_cli_output (vm, "state not set!"); - return 0; + goto done; } vnet_lisp_rloc_probe_enable_disable (is_enabled); - return 0; + +done: + unformat_free (line_input); + + return error; } /* *INDENT-OFF* */ @@ -1022,6 +1073,7 @@ lisp_show_eid_table_map_command_fn (vlib_main_t * vm, lisp_cp_main_t *lcm = vnet_lisp_cp_get_main (); uword *vni_table = 0; u8 is_l2 = 0; + clib_error_t *error = NULL; /* Get a line of input. */ if (!unformat_user (input, unformat_line_input, line_input)) @@ -1040,14 +1092,17 @@ lisp_show_eid_table_map_command_fn (vlib_main_t * vm, is_l2 = 0; } else - return clib_error_return (0, "parse error: '%U'", - format_unformat_error, line_input); + { + error = clib_error_return (0, "parse error: '%U'", + format_unformat_error, line_input); + goto done; + } } if (!vni_table) { vlib_cli_output (vm, "Error: expected l2|l3 param!\n"); - return 0; + goto done; } vlib_cli_output (vm, "%=10s%=10s", "VNI", is_l2 ? "BD" : "VRF"); @@ -1059,7 +1114,10 @@ lisp_show_eid_table_map_command_fn (vlib_main_t * vm, })); /* *INDENT-ON* */ - return 0; +done: + unformat_free (line_input); + + return error; } /* *INDENT-OFF* */ @@ -1131,6 +1189,8 @@ done: vec_free (locators); if (locator_set_name) vec_free (locator_set_name); + unformat_free (line_input); + return error; } @@ -1205,6 +1265,8 @@ lisp_add_del_locator_in_set_command_fn (vlib_main_t * vm, done: vec_free (locators); vec_free (locator_set_name); + unformat_free (line_input); + return error; } @@ -1322,6 +1384,8 @@ lisp_add_del_map_resolver_command_fn (vlib_main_t * vm, } done: + unformat_free (line_input); + return error; } @@ -1372,11 +1436,11 @@ lisp_add_del_mreq_itr_rlocs_command_fn (vlib_main_t * vm, is_add ? "add" : "delete"); } +done: vec_free (locator_set_name); + unformat_free (line_input); -done: return error; - } /* *INDENT-OFF* */ @@ -1438,7 +1502,10 @@ lisp_use_petr_set_locator_set_command_fn (vlib_main_t * vm, else if (unformat (line_input, "disable")) is_add = 0; else - return clib_error_return (0, "parse error"); + { + error = clib_error_return (0, "parse error"); + goto done; + } } if (!ip_set) @@ -1454,6 +1521,8 @@ lisp_use_petr_set_locator_set_command_fn (vlib_main_t * vm, } done: + unformat_free (line_input); + return error; } diff --git a/src/vnet/lisp-gpe/interface.c b/src/vnet/lisp-gpe/interface.c index 2142e095..19ac22e7 100644 --- a/src/vnet/lisp-gpe/interface.c +++ b/src/vnet/lisp-gpe/interface.c @@ -794,6 +794,7 @@ lisp_gpe_add_del_iface_command_fn (vlib_main_t * vm, unformat_input_t * input, u32 table_id, vni, bd_id; u8 vni_is_set = 0, vrf_is_set = 0, bd_index_is_set = 0; u8 nsh_iface = 0; + clib_error_t *error = NULL; if (vnet_lisp_gpe_enable_disable_status () == 0) { @@ -828,8 +829,9 @@ lisp_gpe_add_del_iface_command_fn (vlib_main_t * vm, unformat_input_t * input, } else { - return clib_error_return (0, "parse error: '%U'", - format_unformat_error, line_input); + error = clib_error_return (0, "parse error: '%U'", + format_unformat_error, line_input); + goto done; } } @@ -839,7 +841,8 @@ lisp_gpe_add_del_iface_command_fn (vlib_main_t * vm, unformat_input_t * input, { if (~0 == lisp_gpe_add_nsh_iface (&lisp_gpe_main)) { - return clib_error_return (0, "NSH interface not created"); + error = clib_error_return (0, "NSH interface not created"); + goto done; } } else @@ -850,21 +853,34 @@ lisp_gpe_add_del_iface_command_fn (vlib_main_t * vm, unformat_input_t * input, } if (vrf_is_set && bd_index_is_set) - return clib_error_return (0, - "Cannot set both vrf and brdige domain index!"); + { + error = clib_error_return + (0, "Cannot set both vrf and brdige domain index!"); + goto done; + } if (!vni_is_set) - return clib_error_return (0, "vni must be set!"); + { + error = clib_error_return (0, "vni must be set!"); + goto done; + } if (!vrf_is_set && !bd_index_is_set) - return clib_error_return (0, "vrf or bridge domain index must be set!"); + { + error = + clib_error_return (0, "vrf or bridge domain index must be set!"); + goto done; + } if (bd_index_is_set) { if (is_add) { if (~0 == lisp_gpe_tenant_l2_iface_add_or_lock (vni, bd_id)) - return clib_error_return (0, "L2 interface not created"); + { + error = clib_error_return (0, "L2 interface not created"); + goto done; + } } else lisp_gpe_tenant_l2_iface_unlock (vni); @@ -874,13 +890,35 @@ lisp_gpe_add_del_iface_command_fn (vlib_main_t * vm, unformat_input_t * input, if (is_add) { if (~0 == lisp_gpe_tenant_l3_iface_add_or_lock (vni, table_id)) - return clib_error_return (0, "L3 interface not created"); + { + error = clib_error_return (0, "L3 interface not created"); + goto done; + } } else lisp_gpe_tenant_l3_iface_unlock (vni); } - return (NULL); + if (nsh_iface) + { + if (is_add) + { + if (~0 == lisp_gpe_add_nsh_iface (&lisp_gpe_main)) + { + error = clib_error_return (0, "NSH interface not created"); + goto done; + } + else + { + lisp_gpe_del_nsh_iface (&lisp_gpe_main); + } + } + } + +done: + unformat_free (line_input); + + return error; } /* *INDENT-OFF* */ diff --git a/src/vnet/lisp-gpe/lisp_gpe.c b/src/vnet/lisp-gpe/lisp_gpe.c index 1f8afdae..f2fbcbd5 100644 --- a/src/vnet/lisp-gpe/lisp_gpe.c +++ b/src/vnet/lisp-gpe/lisp_gpe.c @@ -218,6 +218,7 @@ lisp_gpe_enable_disable_command_fn (vlib_main_t * vm, unformat_input_t _line_input, *line_input = &_line_input; u8 is_en = 1; vnet_lisp_gpe_enable_disable_args_t _a, *a = &_a; + clib_error_t *error = NULL; /* Get a line of input. */ if (!unformat_user (input, unformat_line_input, line_input)) @@ -231,12 +232,18 @@ lisp_gpe_enable_disable_command_fn (vlib_main_t * vm, is_en = 0; else { - return clib_error_return (0, "parse error: '%U'", - format_unformat_error, line_input); + error = clib_error_return (0, "parse error: '%U'", + format_unformat_error, line_input); + goto done; } } a->is_en = is_en; - return vnet_lisp_gpe_enable_disable (a); + error = vnet_lisp_gpe_enable_disable (a); + +done: + unformat_free (line_input); + + return error; } /* *INDENT-OFF* */ diff --git a/src/vnet/map/map.c b/src/vnet/map/map.c index aeec6a94..a2d28118 100644 --- a/src/vnet/map/map.c +++ b/src/vnet/map/map.c @@ -465,6 +465,8 @@ map_security_check_command_fn (vlib_main_t * vm, { unformat_input_t _line_input, *line_input = &_line_input; map_main_t *mm = &map_main; + clib_error_t *error = NULL; + /* Get a line of input. */ if (!unformat_user (input, unformat_line_input, line_input)) return 0; @@ -476,11 +478,17 @@ map_security_check_command_fn (vlib_main_t * vm, else if (unformat (line_input, "on")) mm->sec_check = true; else - return clib_error_return (0, "unknown input `%U'", - format_unformat_error, input); + { + error = clib_error_return (0, "unknown input `%U'", + format_unformat_error, line_input); + goto done; + } } + +done: unformat_free (line_input); - return 0; + + return error; } static clib_error_t * @@ -490,6 +498,8 @@ map_security_check_frag_command_fn (vlib_main_t * vm, { unformat_input_t _line_input, *line_input = &_line_input; map_main_t *mm = &map_main; + clib_error_t *error = NULL; + /* Get a line of input. */ if (!unformat_user (input, unformat_line_input, line_input)) return 0; @@ -501,11 +511,17 @@ map_security_check_frag_command_fn (vlib_main_t * vm, else if (unformat (line_input, "on")) mm->sec_check_frag = true; else - return clib_error_return (0, "unknown input `%U'", - format_unformat_error, input); + { + error = clib_error_return (0, "unknown input `%U'", + format_unformat_error, line_input); + goto done; + } } + +done: unformat_free (line_input); - return 0; + + return error; } static clib_error_t * @@ -523,6 +539,7 @@ map_add_domain_command_fn (vlib_main_t * vm, u32 mtu = 0; u8 flags = 0; ip6_src_len = 128; + clib_error_t *error = NULL; /* Get a line of input. */ if (!unformat_user (input, unformat_line_input, line_input)) @@ -559,20 +576,28 @@ map_add_domain_command_fn (vlib_main_t * vm, else if (unformat (line_input, "map-t")) flags |= MAP_DOMAIN_TRANSLATION; else - return clib_error_return (0, "unknown input `%U'", - format_unformat_error, input); + { + error = clib_error_return (0, "unknown input `%U'", + format_unformat_error, line_input); + goto done; + } } - unformat_free (line_input); if (num_m_args < 3) - return clib_error_return (0, "mandatory argument(s) missing"); + { + error = clib_error_return (0, "mandatory argument(s) missing"); + goto done; + } map_create_domain (&ip4_prefix, ip4_prefix_len, &ip6_prefix, ip6_prefix_len, &ip6_src, ip6_src_len, ea_bits_len, psid_offset, psid_length, &map_domain_index, mtu, flags); - return 0; +done: + unformat_free (line_input); + + return error; } static clib_error_t * @@ -582,6 +607,7 @@ map_del_domain_command_fn (vlib_main_t * vm, unformat_input_t _line_input, *line_input = &_line_input; u32 num_m_args = 0; u32 map_domain_index; + clib_error_t *error = NULL; /* Get a line of input. */ if (!unformat_user (input, unformat_line_input, line_input)) @@ -592,17 +618,25 @@ map_del_domain_command_fn (vlib_main_t * vm, if (unformat (line_input, "index %d", &map_domain_index)) num_m_args++; else - return clib_error_return (0, "unknown input `%U'", - format_unformat_error, input); + { + error = clib_error_return (0, "unknown input `%U'", + format_unformat_error, line_input); + goto done; + } } - unformat_free (line_input); if (num_m_args != 1) - return clib_error_return (0, "mandatory argument(s) missing"); + { + error = clib_error_return (0, "mandatory argument(s) missing"); + goto done; + } map_delete_domain (map_domain_index); - return 0; +done: + unformat_free (line_input); + + return error; } static clib_error_t * @@ -613,6 +647,7 @@ map_add_rule_command_fn (vlib_main_t * vm, ip6_address_t tep; u32 num_m_args = 0; u32 psid = 0, map_domain_index; + clib_error_t *error = NULL; /* Get a line of input. */ if (!unformat_user (input, unformat_line_input, line_input)) @@ -628,19 +663,29 @@ map_add_rule_command_fn (vlib_main_t * vm, if (unformat (line_input, "ip6-dst %U", unformat_ip6_address, &tep)) num_m_args++; else - return clib_error_return (0, "unknown input `%U'", - format_unformat_error, input); + { + error = clib_error_return (0, "unknown input `%U'", + format_unformat_error, line_input); + goto done; + } } - unformat_free (line_input); if (num_m_args != 3) - return clib_error_return (0, "mandatory argument(s) missing"); + { + error = clib_error_return (0, "mandatory argument(s) missing"); + goto done; + } if (map_add_del_psid (map_domain_index, psid, &tep, 1) != 0) { - return clib_error_return (0, "Failing to add Mapping Rule"); + error = clib_error_return (0, "Failing to add Mapping Rule"); + goto done; } - return 0; + +done: + unformat_free (line_input); + + return error; } #if MAP_SKIP_IP6_LOOKUP @@ -653,6 +698,7 @@ map_pre_resolve_command_fn (vlib_main_t * vm, ip4_address_t ip4nh; ip6_address_t ip6nh; map_main_t *mm = &map_main; + clib_error_t *error = NULL; memset (&ip4nh, 0, sizeof (ip4nh)); memset (&ip6nh, 0, sizeof (ip6nh)); @@ -669,14 +715,19 @@ map_pre_resolve_command_fn (vlib_main_t * vm, if (unformat (line_input, "ip6-nh %U", unformat_ip6_address, &ip6nh)) mm->preresolve_ip6 = ip6nh; else - return clib_error_return (0, "unknown input `%U'", - format_unformat_error, input); + { + error = clib_error_return (0, "unknown input `%U'", + format_unformat_error, line_input); + goto done; + } } - unformat_free (line_input); map_pre_resolve (&ip4nh, &ip6nh); - return 0; +done: + unformat_free (line_input); + + return error; } #endif @@ -688,6 +739,7 @@ map_icmp_relay_source_address_command_fn (vlib_main_t * vm, unformat_input_t _line_input, *line_input = &_line_input; ip4_address_t icmp_src_address; map_main_t *mm = &map_main; + clib_error_t *error = NULL; mm->icmp4_src_address.as_u32 = 0; @@ -701,12 +753,17 @@ map_icmp_relay_source_address_command_fn (vlib_main_t * vm, (line_input, "%U", unformat_ip4_address, &icmp_src_address)) mm->icmp4_src_address = icmp_src_address; else - return clib_error_return (0, "unknown input `%U'", - format_unformat_error, input); + { + error = clib_error_return (0, "unknown input `%U'", + format_unformat_error, line_input); + goto done; + } } + +done: unformat_free (line_input); - return 0; + return error; } static clib_error_t * @@ -717,6 +774,7 @@ map_icmp_unreachables_command_fn (vlib_main_t * vm, unformat_input_t _line_input, *line_input = &_line_input; map_main_t *mm = &map_main; int num_m_args = 0; + clib_error_t *error = NULL; /* Get a line of input. */ if (!unformat_user (input, unformat_line_input, line_input)) @@ -730,16 +788,21 @@ map_icmp_unreachables_command_fn (vlib_main_t * vm, else if (unformat (line_input, "off")) mm->icmp6_enabled = false; else - return clib_error_return (0, "unknown input `%U'", - format_unformat_error, input); + { + error = clib_error_return (0, "unknown input `%U'", + format_unformat_error, line_input); + goto done; + } } - unformat_free (line_input); if (num_m_args != 1) - return clib_error_return (0, "mandatory argument(s) missing"); + error = clib_error_return (0, "mandatory argument(s) missing"); - return 0; +done: + unformat_free (line_input); + + return error; } static clib_error_t * @@ -748,6 +811,7 @@ map_fragment_command_fn (vlib_main_t * vm, { unformat_input_t _line_input, *line_input = &_line_input; map_main_t *mm = &map_main; + clib_error_t *error = NULL; /* Get a line of input. */ if (!unformat_user (input, unformat_line_input, line_input)) @@ -760,12 +824,17 @@ map_fragment_command_fn (vlib_main_t * vm, else if (unformat (line_input, "outer")) mm->frag_inner = false; else - return clib_error_return (0, "unknown input `%U'", - format_unformat_error, input); + { + error = clib_error_return (0, "unknown input `%U'", + format_unformat_error, line_input); + goto done; + } } + +done: unformat_free (line_input); - return 0; + return error; } static clib_error_t * @@ -775,6 +844,7 @@ map_fragment_df_command_fn (vlib_main_t * vm, { unformat_input_t _line_input, *line_input = &_line_input; map_main_t *mm = &map_main; + clib_error_t *error = NULL; /* Get a line of input. */ if (!unformat_user (input, unformat_line_input, line_input)) @@ -787,12 +857,17 @@ map_fragment_df_command_fn (vlib_main_t * vm, else if (unformat (line_input, "off")) mm->frag_ignore_df = false; else - return clib_error_return (0, "unknown input `%U'", - format_unformat_error, input); + { + error = clib_error_return (0, "unknown input `%U'", + format_unformat_error, line_input); + goto done; + } } + +done: unformat_free (line_input); - return 0; + return error; } static clib_error_t * @@ -803,6 +878,7 @@ map_traffic_class_command_fn (vlib_main_t * vm, unformat_input_t _line_input, *line_input = &_line_input; map_main_t *mm = &map_main; u32 tc = 0; + clib_error_t *error = NULL; mm->tc_copy = false; @@ -817,12 +893,17 @@ map_traffic_class_command_fn (vlib_main_t * vm, else if (unformat (line_input, "%x", &tc)) mm->tc = tc & 0xff; else - return clib_error_return (0, "unknown input `%U'", - format_unformat_error, input); + { + error = clib_error_return (0, "unknown input `%U'", + format_unformat_error, line_input); + goto done; + } } + +done: unformat_free (line_input); - return 0; + return error; } static u8 * @@ -922,6 +1003,7 @@ show_map_domain_command_fn (vlib_main_t * vm, unformat_input_t * input, map_domain_t *d; bool counters = false; u32 map_domain_index = ~0; + clib_error_t *error = NULL; /* Get a line of input. */ if (!unformat_user (input, unformat_line_input, line_input)) @@ -934,10 +1016,12 @@ show_map_domain_command_fn (vlib_main_t * vm, unformat_input_t * input, else if (unformat (line_input, "index %d", &map_domain_index)) ; else - return clib_error_return (0, "unknown input `%U'", - format_unformat_error, input); + { + error = clib_error_return (0, "unknown input `%U'", + format_unformat_error, line_input); + goto done; + } } - unformat_free (line_input); if (pool_elts (mm->domains) == 0) vlib_cli_output (vm, "No MAP domains are configured..."); @@ -952,15 +1036,19 @@ show_map_domain_command_fn (vlib_main_t * vm, unformat_input_t * input, { if (pool_is_free_index (mm->domains, map_domain_index)) { - return clib_error_return (0, "MAP domain does not exists %d", - map_domain_index); + error = clib_error_return (0, "MAP domain does not exists %d", + map_domain_index); + goto done; } d = pool_elt_at_index (mm->domains, map_domain_index); vlib_cli_output (vm, "%U", format_map_domain, d, counters); } - return 0; +done: + unformat_free (line_input); + + return error; } static clib_error_t * diff --git a/src/vnet/mpls/mpls.c b/src/vnet/mpls/mpls.c index 0e610e17..7ae4aa00 100644 --- a/src/vnet/mpls/mpls.c +++ b/src/vnet/mpls/mpls.c @@ -470,6 +470,8 @@ vnet_mpls_local_label (vlib_main_t * vm, } done: + unformat_free (line_input); + return error; } diff --git a/src/vnet/mpls/mpls_tunnel.c b/src/vnet/mpls/mpls_tunnel.c index 8d1e30a3..e488271d 100644 --- a/src/vnet/mpls/mpls_tunnel.c +++ b/src/vnet/mpls/mpls_tunnel.c @@ -535,6 +535,7 @@ vnet_create_mpls_tunnel_command_fn (vlib_main_t * vm, fib_route_path_t rpath, *rpaths = NULL; mpls_label_t out_label = MPLS_LABEL_INVALID, *labels = NULL; u32 sw_if_index; + clib_error_t *error = NULL; memset(&rpath, 0, sizeof(rpath)); @@ -595,8 +596,11 @@ vnet_create_mpls_tunnel_command_fn (vlib_main_t * vm, else if (unformat (line_input, "l2-only")) l2_only = 1; else - return clib_error_return (0, "unknown input '%U'", - format_unformat_error, line_input); + { + error = clib_error_return (0, "unknown input '%U'", + format_unformat_error, line_input); + goto done; + } } if (is_del) @@ -606,17 +610,22 @@ vnet_create_mpls_tunnel_command_fn (vlib_main_t * vm, else { if (0 == vec_len(labels)) - return clib_error_return (0, "No Output Labels '%U'", - format_unformat_error, line_input); + { + error = clib_error_return (0, "No Output Labels '%U'", + format_unformat_error, line_input); + goto done; + } vec_add1(rpaths, rpath); vnet_mpls_tunnel_add(rpaths, labels, l2_only, &sw_if_index); } +done: vec_free(labels); vec_free(rpaths); + unformat_free (line_input); - return (NULL); + return error; } /*? diff --git a/src/vnet/pg/cli.c b/src/vnet/pg/cli.c index f5896b43..3c249a7b 100644 --- a/src/vnet/pg/cli.c +++ b/src/vnet/pg/cli.c @@ -547,21 +547,30 @@ pg_capture_cmd_fn (vlib_main_t * vm, else { error = clib_error_create ("unknown input `%U'", - format_unformat_error, input); - return error; + format_unformat_error, line_input); + goto done; } } if (!hi) - return clib_error_return (0, "Please specify interface name"); + { + error = clib_error_return (0, "Please specify interface name"); + goto done; + } if (hi->dev_class_index != pg_dev_class.index) - return clib_error_return (0, "Please specify packet-generator interface"); + { + error = + clib_error_return (0, "Please specify packet-generator interface"); + goto done; + } if (!pcap_file_name && is_disable == 0) - return clib_error_return (0, "Please specify pcap file name"); + { + error = clib_error_return (0, "Please specify pcap file name"); + goto done; + } - unformat_free (line_input); pg_capture_args_t _a, *a = &_a; @@ -572,6 +581,10 @@ pg_capture_cmd_fn (vlib_main_t * vm, a->count = count; error = pg_capture (a); + +done: + unformat_free (line_input); + return error; } @@ -590,6 +603,7 @@ create_pg_if_cmd_fn (vlib_main_t * vm, pg_main_t *pg = &pg_main; unformat_input_t _line_input, *line_input = &_line_input; u32 if_id; + clib_error_t *error = NULL; if (!unformat_user (input, unformat_line_input, line_input)) return 0; @@ -600,14 +614,19 @@ create_pg_if_cmd_fn (vlib_main_t * vm, ; else - return clib_error_create ("unknown input `%U'", - format_unformat_error, input); + { + error = clib_error_create ("unknown input `%U'", + format_unformat_error, line_input); + goto done; + } } + pg_interface_add_or_get (pg, if_id); + +done: unformat_free (line_input); - pg_interface_add_or_get (pg, if_id); - return 0; + return error; } /* *INDENT-OFF* */ diff --git a/src/vnet/policer/node_funcs.c b/src/vnet/policer/node_funcs.c index 1f4997ff..457dd09f 100644 --- a/src/vnet/policer/node_funcs.c +++ b/src/vnet/policer/node_funcs.c @@ -447,6 +447,7 @@ test_policer_command_fn (vlib_main_t * vm, int rx_set = 0; int is_add = 1; int is_show = 0; + clib_error_t *error = NULL; /* Get a line of input. */ if (!unformat_user (input, unformat_line_input, line_input)) @@ -468,7 +469,10 @@ test_policer_command_fn (vlib_main_t * vm, } if (rx_set == 0) - return clib_error_return (0, "interface not set"); + { + error = clib_error_return (0, "interface not set"); + goto done; + } if (is_show) { @@ -477,12 +481,13 @@ test_policer_command_fn (vlib_main_t * vm, policer = pool_elt_at_index (pm->policers, pi); vlib_cli_output (vm, "%U", format_policer_instance, policer); - return 0; + goto done; } if (is_add && config_name == 0) { - return clib_error_return (0, "policer config name required"); + error = clib_error_return (0, "policer config name required"); + goto done; } rv = test_policer_add_del (rx_sw_if_index, config_name, is_add); @@ -493,11 +498,15 @@ test_policer_command_fn (vlib_main_t * vm, break; default: - return clib_error_return + error = clib_error_return (0, "WARNING: vnet_vnet_policer_add_del returned %d", rv); + goto done; } - return 0; +done: + unformat_free (line_input); + + return error; } /* *INDENT-OFF* */ diff --git a/src/vnet/policer/policer.c b/src/vnet/policer/policer.c index 290a6af5..cd754e29 100644 --- a/src/vnet/policer/policer.c +++ b/src/vnet/policer/policer.c @@ -413,6 +413,7 @@ configure_policer_command_fn (vlib_main_t * vm, u8 is_add = 1; u8 *name = 0; u32 pi; + clib_error_t *error = NULL; /* Get a line of input. */ if (!unformat_user (input, unformat_line_input, line_input)) @@ -433,13 +434,19 @@ configure_policer_command_fn (vlib_main_t * vm, foreach_config_param #undef _ else - return clib_error_return (0, "unknown input `%U'", - format_unformat_error, line_input); + { + error = clib_error_return (0, "unknown input `%U'", + format_unformat_error, line_input); + goto done; + } } + error = policer_add_del (vm, name, &c, &pi, is_add); + +done: unformat_free (line_input); - return policer_add_del (vm, name, &c, &pi, is_add); + return error; } /* *INDENT-OFF* */ diff --git a/src/vnet/unix/tapcli.c b/src/vnet/unix/tapcli.c index 48e81b50..25c930c6 100644 --- a/src/vnet/unix/tapcli.c +++ b/src/vnet/unix/tapcli.c @@ -1308,6 +1308,7 @@ tap_connect_command_fn (vlib_main_t * vm, int ip6_address_set = 0; u32 ip4_mask_width = 0; u32 ip6_mask_width = 0; + clib_error_t *error = NULL; if (tm->is_disabled) return clib_error_return (0, "device disabled..."); @@ -1336,12 +1337,18 @@ tap_connect_command_fn (vlib_main_t * vm, else if (unformat (line_input, "%s", &intfc_name)) ; else - return clib_error_return (0, "unknown input `%U'", - format_unformat_error, line_input); + { + error = clib_error_return (0, "unknown input `%U'", + format_unformat_error, line_input); + goto done; + } } if (intfc_name == 0) - return clib_error_return (0, "interface name must be specified"); + { + error = clib_error_return (0, "interface name must be specified"); + goto done; + } memset (ap, 0, sizeof (*ap)); @@ -1367,48 +1374,64 @@ tap_connect_command_fn (vlib_main_t * vm, switch (rv) { case VNET_API_ERROR_SYSCALL_ERROR_1: - return clib_error_return (0, "Couldn't open /dev/net/tun"); + error = clib_error_return (0, "Couldn't open /dev/net/tun"); + goto done; case VNET_API_ERROR_SYSCALL_ERROR_2: - return clib_error_return (0, "Error setting flags on '%s'", intfc_name); - + error = clib_error_return (0, "Error setting flags on '%s'", intfc_name); + goto done; + case VNET_API_ERROR_SYSCALL_ERROR_3: - return clib_error_return (0, "Couldn't open provisioning socket"); + error = clib_error_return (0, "Couldn't open provisioning socket"); + goto done; case VNET_API_ERROR_SYSCALL_ERROR_4: - return clib_error_return (0, "Couldn't get if_index"); + error = clib_error_return (0, "Couldn't get if_index"); + goto done; case VNET_API_ERROR_SYSCALL_ERROR_5: - return clib_error_return (0, "Couldn't bind provisioning socket"); + error = clib_error_return (0, "Couldn't bind provisioning socket"); + goto done; case VNET_API_ERROR_SYSCALL_ERROR_6: - return clib_error_return (0, "Couldn't set device non-blocking flag"); + error = clib_error_return (0, "Couldn't set device non-blocking flag"); + goto done; case VNET_API_ERROR_SYSCALL_ERROR_7: - return clib_error_return (0, "Couldn't set device MTU"); + error = clib_error_return (0, "Couldn't set device MTU"); + goto done; case VNET_API_ERROR_SYSCALL_ERROR_8: - return clib_error_return (0, "Couldn't get interface flags"); + error = clib_error_return (0, "Couldn't get interface flags"); + goto done; case VNET_API_ERROR_SYSCALL_ERROR_9: - return clib_error_return (0, "Couldn't set intfc admin state up"); + error = clib_error_return (0, "Couldn't set intfc admin state up"); + goto done; case VNET_API_ERROR_SYSCALL_ERROR_10: - return clib_error_return (0, "Couldn't set intfc address/mask"); + error = clib_error_return (0, "Couldn't set intfc address/mask"); + goto done; case VNET_API_ERROR_INVALID_REGISTRATION: - return clib_error_return (0, "Invalid registration"); + error = clib_error_return (0, "Invalid registration"); + goto done; case 0: break; default: - return clib_error_return (0, "Unknown error: %d", rv); + error = clib_error_return (0, "Unknown error: %d", rv); + goto done; } vlib_cli_output(vm, "%U\n", format_vnet_sw_if_index_name, vnet_get_main(), sw_if_index); - return 0; + +done: + unformat_free (line_input); + + return error; } VLIB_CLI_COMMAND (tap_connect_command, static) = { diff --git a/src/vnet/vxlan-gpe/vxlan_gpe.c b/src/vnet/vxlan-gpe/vxlan_gpe.c index b97510c4..2cba596f 100644 --- a/src/vnet/vxlan-gpe/vxlan_gpe.c +++ b/src/vnet/vxlan-gpe/vxlan_gpe.c @@ -454,6 +454,7 @@ vxlan_gpe_add_del_tunnel_command_fn (vlib_main_t * vm, u32 tmp; vnet_vxlan_gpe_add_del_tunnel_args_t _a, * a = &_a; u32 sw_if_index; + clib_error_t *error = NULL; /* Get a line of input. */ if (! unformat_user (input, unformat_line_input, line_input)) @@ -494,7 +495,10 @@ vxlan_gpe_add_del_tunnel_command_fn (vlib_main_t * vm, encap_fib_index = ip4_fib_index_from_table_id (tmp); if (encap_fib_index == ~0) - return clib_error_return (0, "nonexistent encap fib id %d", tmp); + { + error = clib_error_return (0, "nonexistent encap fib id %d", tmp); + goto done; + } } else if (unformat (line_input, "decap-vrf-id %d", &tmp)) { @@ -504,7 +508,10 @@ vxlan_gpe_add_del_tunnel_command_fn (vlib_main_t * vm, decap_fib_index = ip4_fib_index_from_table_id (tmp); if (decap_fib_index == ~0) - return clib_error_return (0, "nonexistent decap fib id %d", tmp); + { + error = clib_error_return (0, "nonexistent decap fib id %d", tmp); + goto done; + } } else if (unformat (line_input, "vni %d", &vni)) vni_set = 1; @@ -517,27 +524,43 @@ vxlan_gpe_add_del_tunnel_command_fn (vlib_main_t * vm, else if (unformat(line_input, "next-nsh")) protocol = VXLAN_GPE_PROTOCOL_NSH; else - return clib_error_return (0, "parse error: '%U'", - format_unformat_error, line_input); + { + error = clib_error_return (0, "parse error: '%U'", + format_unformat_error, line_input); + goto done; + } } - unformat_free (line_input); - if (local_set == 0) - return clib_error_return (0, "tunnel local address not specified"); + { + error = clib_error_return (0, "tunnel local address not specified"); + goto done; + } if (remote_set == 0) - return clib_error_return (0, "tunnel remote address not specified"); + { + error = clib_error_return (0, "tunnel remote address not specified"); + goto done; + } if (ipv4_set && ipv6_set) - return clib_error_return (0, "both IPv4 and IPv6 addresses specified"); + { + error = clib_error_return (0, "both IPv4 and IPv6 addresses specified"); + goto done; + } if ((ipv4_set && memcmp(&local.ip4, &remote.ip4, sizeof(local.ip4)) == 0) || (ipv6_set && memcmp(&local.ip6, &remote.ip6, sizeof(local.ip6)) == 0)) - return clib_error_return (0, "src and dst addresses are identical"); + { + error = clib_error_return (0, "src and dst addresses are identical"); + goto done; + } if (vni_set == 0) - return clib_error_return (0, "vni not specified"); + { + error = clib_error_return (0, "vni not specified"); + goto done; + } memset (a, 0, sizeof (*a)); @@ -558,20 +581,27 @@ vxlan_gpe_add_del_tunnel_command_fn (vlib_main_t * vm, vlib_cli_output(vm, "%U\n", format_vnet_sw_if_index_name, vnet_get_main(), sw_if_index); break; case VNET_API_ERROR_INVALID_DECAP_NEXT: - return clib_error_return (0, "invalid decap-next..."); + error = clib_error_return (0, "invalid decap-next..."); + goto done; case VNET_API_ERROR_TUNNEL_EXIST: - return clib_error_return (0, "tunnel already exists..."); + error = clib_error_return (0, "tunnel already exists..."); + goto done; case VNET_API_ERROR_NO_SUCH_ENTRY: - return clib_error_return (0, "tunnel does not exist..."); + error = clib_error_return (0, "tunnel does not exist..."); + goto done; default: - return clib_error_return + error = clib_error_return (0, "vnet_vxlan_gpe_add_del_tunnel returned %d", rv); + goto done; } - return 0; +done: + unformat_free (line_input); + + return error; } VLIB_CLI_COMMAND (create_vxlan_gpe_tunnel_command, static) = { diff --git a/src/vnet/vxlan/vxlan.c b/src/vnet/vxlan/vxlan.c index 849fc25d..eedc16f8 100644 --- a/src/vnet/vxlan/vxlan.c +++ b/src/vnet/vxlan/vxlan.c @@ -657,6 +657,7 @@ vxlan_add_del_tunnel_command_fn (vlib_main_t * vm, int rv; vnet_vxlan_add_del_tunnel_args_t _a, * a = &_a; u32 tunnel_sw_if_index; + clib_error_t *error = NULL; /* Cant "universally zero init" (={0}) due to GCC bug 53119 */ memset(&src, 0, sizeof src); @@ -715,7 +716,10 @@ vxlan_add_del_tunnel_command_fn (vlib_main_t * vm, { encap_fib_index = fib_table_find (fib_ip_proto (ipv6_set), tmp); if (encap_fib_index == ~0) - return clib_error_return (0, "nonexistent encap-vrf-id %d", tmp); + { + error = clib_error_return (0, "nonexistent encap-vrf-id %d", tmp); + goto done; + } } else if (unformat (line_input, "decap-next %U", unformat_decap_next, &decap_next_index, ipv4_set)) @@ -723,41 +727,72 @@ vxlan_add_del_tunnel_command_fn (vlib_main_t * vm, else if (unformat (line_input, "vni %d", &vni)) { if (vni >> 24) - return clib_error_return (0, "vni %d out of range", vni); + { + error = clib_error_return (0, "vni %d out of range", vni); + goto done; + } } else - return clib_error_return (0, "parse error: '%U'", - format_unformat_error, line_input); + { + error = clib_error_return (0, "parse error: '%U'", + format_unformat_error, line_input); + goto done; + } } - unformat_free (line_input); - if (src_set == 0) - return clib_error_return (0, "tunnel src address not specified"); + { + error = clib_error_return (0, "tunnel src address not specified"); + goto done; + } if (dst_set == 0) - return clib_error_return (0, "tunnel dst address not specified"); + { + error = clib_error_return (0, "tunnel dst address not specified"); + goto done; + } if (grp_set && !ip46_address_is_multicast(&dst)) - return clib_error_return (0, "tunnel group address not multicast"); + { + error = clib_error_return (0, "tunnel group address not multicast"); + goto done; + } if (grp_set == 0 && ip46_address_is_multicast(&dst)) - return clib_error_return (0, "dst address must be unicast"); + { + error = clib_error_return (0, "dst address must be unicast"); + goto done; + } if (grp_set && mcast_sw_if_index == ~0) - return clib_error_return (0, "tunnel nonexistent multicast device"); + { + error = clib_error_return (0, "tunnel nonexistent multicast device"); + goto done; + } if (ipv4_set && ipv6_set) - return clib_error_return (0, "both IPv4 and IPv6 addresses specified"); + { + error = clib_error_return (0, "both IPv4 and IPv6 addresses specified"); + goto done; + } if (ip46_address_cmp(&src, &dst) == 0) - return clib_error_return (0, "src and dst addresses are identical"); + { + error = clib_error_return (0, "src and dst addresses are identical"); + goto done; + } if (decap_next_index == ~0) - return clib_error_return (0, "next node not found"); + { + error = clib_error_return (0, "next node not found"); + goto done; + } if (vni == 0) - return clib_error_return (0, "vni not specified"); + { + error = clib_error_return (0, "vni not specified"); + goto done; + } memset (a, 0, sizeof (*a)); @@ -779,17 +814,23 @@ vxlan_add_del_tunnel_command_fn (vlib_main_t * vm, break; case VNET_API_ERROR_TUNNEL_EXIST: - return clib_error_return (0, "tunnel already exists..."); + error = clib_error_return (0, "tunnel already exists..."); + goto done; case VNET_API_ERROR_NO_SUCH_ENTRY: - return clib_error_return (0, "tunnel does not exist..."); + error = clib_error_return (0, "tunnel does not exist..."); + goto done; default: - return clib_error_return + error = clib_error_return (0, "vnet_vxlan_add_del_tunnel returned %d", rv); + goto done; } - return 0; +done: + unformat_free (line_input); + + return error; } /*? @@ -912,6 +953,8 @@ set_ip_vxlan_bypass (u32 is_ip6, vnet_int_vxlan_bypass_mode (sw_if_index, is_ip6, is_enable); done: + unformat_free (line_input); + return error; } diff --git a/src/vpp/app/l2t.c b/src/vpp/app/l2t.c index 45dd2807..e1eda155 100644 --- a/src/vpp/app/l2t.c +++ b/src/vpp/app/l2t.c @@ -254,6 +254,7 @@ l2tp_session_add_command_fn (vlib_main_t * vm, u32 local_session_id = 1, remote_session_id = 1; int our_address_set = 0, client_address_set = 0; int l2_sublayer_present = 0; + clib_error_t *error = NULL; /* Get a line of input. */ if (!unformat_user (input, unformat_line_input, line_input)) @@ -290,8 +291,12 @@ l2tp_session_add_command_fn (vlib_main_t * vm, else if (unformat (line_input, "l2-sublayer-present")) l2_sublayer_present = 1; else - return clib_error_return (0, "parse error: '%U'", - format_unformat_error, line_input); + { + error = clib_error_return (0, "parse error: '%U'", + format_unformat_error, line_input); + unformat_free (line_input); + return error; + } } unformat_free (line_input); diff --git a/src/vpp/app/vpe_cli.c b/src/vpp/app/vpe_cli.c index a26bf71f..94bdc84c 100644 --- a/src/vpp/app/vpe_cli.c +++ b/src/vpp/app/vpe_cli.c @@ -36,6 +36,7 @@ virtual_ip_cmd_fn_command_fn (vlib_main_t * vm, mac_addr_t *mac_addrs = 0; u32 sw_if_index; u32 i; + clib_error_t *error = NULL; next_hops = NULL; rpaths = NULL; @@ -49,7 +50,11 @@ virtual_ip_cmd_fn_command_fn (vlib_main_t * vm, if (!unformat (line_input, "%U %U", unformat_ip4_address, &prefix.fp_addr.ip4, unformat_vnet_sw_interface, vnm, &sw_if_index)) - goto barf; + { + error = clib_error_return (0, "unknown input `%U'", + format_unformat_error, line_input); + goto done; + } while (unformat_check_input (line_input) != UNFORMAT_END_OF_INPUT) { @@ -67,13 +72,18 @@ virtual_ip_cmd_fn_command_fn (vlib_main_t * vm, } else { - barf: - return clib_error_return (0, "unknown input `%U'", - format_unformat_error, input); + error = clib_error_return (0, "unknown input `%U'", + format_unformat_error, line_input); + goto done; } } + if (vec_len (mac_addrs) == 0 || vec_len (mac_addrs) != vec_len (next_hops)) - goto barf; + { + error = clib_error_return (0, "unknown input `%U'", + format_unformat_error, line_input); + goto done; + } /* Create / delete special interface route /32's */ @@ -100,10 +110,12 @@ virtual_ip_cmd_fn_command_fn (vlib_main_t * vm, &prefix, FIB_SOURCE_CLI, FIB_ENTRY_FLAG_NONE, rpaths); +done: vec_free (mac_addrs); vec_free (next_hops); + unformat_free (line_input); - return 0; + return error; } /* *INDENT-OFF* */ -- cgit 1.2.3-korg From 68b0fb0c620c7451ef1a6380c43c39de6614db51 Mon Sep 17 00:00:00 2001 From: Dave Barach Date: Tue, 28 Feb 2017 15:15:56 -0500 Subject: VPP-598: tcp stack initial commit Change-Id: I49e5ce0aae6e4ff634024387ceaf7dbc432a0351 Signed-off-by: Dave Barach Signed-off-by: Florin Coras --- src/Makefile.am | 1 + src/plugins/ioam/export-common/ioam_export.h | 2 +- src/plugins/ioam/ipfixcollector/ipfixcollector.c | 2 +- src/plugins/ioam/lib-vxlan-gpe/ioam_transit.c | 2 +- src/plugins/snat/in2out.c | 26 +- src/plugins/snat/out2in.c | 24 +- src/scripts/vnet/tcp | 18 +- src/scripts/vnet/udp | 19 + src/scripts/vnet/uri/tcp-setup.sh | 39 + src/scripts/vnet/uri/tcp_server | 4 + src/scripts/vnet/uri/udp | 19 + src/svm.am | 10 +- src/svm/ssvm.c | 16 + src/svm/ssvm.h | 18 +- src/svm/svm_fifo.c | 568 ++++++ src/svm/svm_fifo.h | 157 ++ src/svm/svm_fifo_segment.c | 193 ++ src/svm/svm_fifo_segment.h | 89 + src/svm/test_svm_fifo1.c | 361 ++++ src/uri.am | 22 + src/uri/uri_tcp_test.c | 916 +++++++++ src/uri/uri_udp_test.c | 553 ++++++ src/uri/uri_udp_test2.c | 954 +++++++++ src/uri/uritest.c | 484 +++++ src/vlib/buffer.c | 2 +- src/vlib/buffer.h | 68 + src/vlibmemory/unix_shared_memory_queue.c | 12 +- src/vlibmemory/unix_shared_memory_queue.h | 2 +- src/vnet.am | 66 +- src/vnet/api_errno.h | 21 +- src/vnet/bfd/bfd_udp.c | 4 +- src/vnet/buffer.h | 10 + src/vnet/classify/vnet_classify.c | 4 +- src/vnet/dhcp/dhcp_proxy.h | 2 +- src/vnet/flow/flow_report.h | 2 +- src/vnet/ip/ip.h | 4 +- src/vnet/ip/ip4.h | 42 +- src/vnet/ip/ip4_forward.c | 173 +- src/vnet/ip/ip4_packet.h | 26 +- src/vnet/ip/ip6.h | 44 +- src/vnet/ip/ip6_packet.h | 26 +- src/vnet/ip/punt.c | 2 +- src/vnet/ip/tcp_packet.h | 141 -- src/vnet/ip/udp.h | 315 --- src/vnet/ip/udp_error.def | 21 - src/vnet/ip/udp_format.c | 91 - src/vnet/ip/udp_init.c | 71 - src/vnet/ip/udp_local.c | 645 ------ src/vnet/ip/udp_packet.h | 65 - src/vnet/ip/udp_pg.c | 237 --- src/vnet/ipsec/ikev2.c | 2 +- src/vnet/ipsec/ikev2_cli.c | 2 +- src/vnet/ipsec/ikev2_crypto.c | 2 +- src/vnet/lisp-cp/packets.c | 65 +- src/vnet/lisp-cp/packets.h | 45 - src/vnet/lisp-gpe/interface.c | 2 +- src/vnet/lisp-gpe/lisp_gpe.h | 4 +- src/vnet/lisp-gpe/lisp_gpe_adjacency.c | 2 + src/vnet/session/application.c | 343 ++++ src/vnet/session/application.h | 120 ++ src/vnet/session/application_interface.c | 459 +++++ src/vnet/session/application_interface.h | 136 ++ src/vnet/session/hashes.c | 28 + src/vnet/session/node.c | 435 ++++ src/vnet/session/session.api | 429 ++++ src/vnet/session/session.c | 1286 ++++++++++++ src/vnet/session/session.h | 380 ++++ src/vnet/session/session_api.c | 821 ++++++++ src/vnet/session/session_cli.c | 189 ++ src/vnet/session/transport.c | 64 + src/vnet/session/transport.h | 250 +++ src/vnet/tcp/tcp.c | 708 +++++++ src/vnet/tcp/tcp.h | 624 ++++++ src/vnet/tcp/tcp_error.def | 35 + src/vnet/tcp/tcp_format.c | 136 ++ src/vnet/tcp/tcp_input.c | 2316 ++++++++++++++++++++++ src/vnet/tcp/tcp_newreno.c | 93 + src/vnet/tcp/tcp_output.c | 1412 +++++++++++++ src/vnet/tcp/tcp_packet.h | 184 ++ src/vnet/tcp/tcp_pg.c | 236 +++ src/vnet/tcp/tcp_syn_filter4.c | 542 +++++ src/vnet/tcp/tcp_timer.h | 29 + src/vnet/udp/builtin_server.c | 239 +++ src/vnet/udp/udp.c | 342 ++++ src/vnet/udp/udp.h | 362 ++++ src/vnet/udp/udp_error.def | 21 + src/vnet/udp/udp_format.c | 91 + src/vnet/udp/udp_input.c | 314 +++ src/vnet/udp/udp_local.c | 666 +++++++ src/vnet/udp/udp_packet.h | 65 + src/vnet/udp/udp_pg.c | 237 +++ src/vnet/vnet_all_api_h.h | 1 + src/vnet/vxlan-gpe/vxlan_gpe.h | 2 +- src/vnet/vxlan/vxlan.h | 2 +- src/vpp/api/vpe.api | 1 + src/vppinfra.am | 5 + src/vppinfra/bihash_16_8.h | 103 + src/vppinfra/bihash_48_8.h | 116 ++ src/vppinfra/tw_timer_16t_1w_2048sl.c | 26 + src/vppinfra/tw_timer_16t_1w_2048sl.h | 46 + 100 files changed, 18737 insertions(+), 1874 deletions(-) create mode 100644 src/scripts/vnet/udp create mode 100755 src/scripts/vnet/uri/tcp-setup.sh create mode 100644 src/scripts/vnet/uri/tcp_server create mode 100644 src/scripts/vnet/uri/udp create mode 100644 src/svm/svm_fifo.c create mode 100644 src/svm/svm_fifo.h create mode 100644 src/svm/svm_fifo_segment.c create mode 100644 src/svm/svm_fifo_segment.h create mode 100644 src/svm/test_svm_fifo1.c create mode 100644 src/uri.am create mode 100644 src/uri/uri_tcp_test.c create mode 100644 src/uri/uri_udp_test.c create mode 100644 src/uri/uri_udp_test2.c create mode 100644 src/uri/uritest.c delete mode 100644 src/vnet/ip/tcp_packet.h delete mode 100644 src/vnet/ip/udp.h delete mode 100644 src/vnet/ip/udp_error.def delete mode 100644 src/vnet/ip/udp_format.c delete mode 100644 src/vnet/ip/udp_init.c delete mode 100644 src/vnet/ip/udp_local.c delete mode 100644 src/vnet/ip/udp_packet.h delete mode 100644 src/vnet/ip/udp_pg.c create mode 100644 src/vnet/session/application.c create mode 100644 src/vnet/session/application.h create mode 100644 src/vnet/session/application_interface.c create mode 100644 src/vnet/session/application_interface.h create mode 100644 src/vnet/session/hashes.c create mode 100644 src/vnet/session/node.c create mode 100644 src/vnet/session/session.api create mode 100644 src/vnet/session/session.c create mode 100644 src/vnet/session/session.h create mode 100644 src/vnet/session/session_api.c create mode 100644 src/vnet/session/session_cli.c create mode 100644 src/vnet/session/transport.c create mode 100644 src/vnet/session/transport.h create mode 100644 src/vnet/tcp/tcp.c create mode 100644 src/vnet/tcp/tcp.h create mode 100644 src/vnet/tcp/tcp_error.def create mode 100644 src/vnet/tcp/tcp_format.c create mode 100644 src/vnet/tcp/tcp_input.c create mode 100644 src/vnet/tcp/tcp_newreno.c create mode 100644 src/vnet/tcp/tcp_output.c create mode 100644 src/vnet/tcp/tcp_packet.h create mode 100644 src/vnet/tcp/tcp_pg.c create mode 100644 src/vnet/tcp/tcp_syn_filter4.c create mode 100644 src/vnet/tcp/tcp_timer.h create mode 100644 src/vnet/udp/builtin_server.c create mode 100644 src/vnet/udp/udp.c create mode 100644 src/vnet/udp/udp.h create mode 100644 src/vnet/udp/udp_error.def create mode 100644 src/vnet/udp/udp_format.c create mode 100644 src/vnet/udp/udp_input.c create mode 100644 src/vnet/udp/udp_local.c create mode 100644 src/vnet/udp/udp_packet.h create mode 100644 src/vnet/udp/udp_pg.c create mode 100644 src/vppinfra/bihash_16_8.h create mode 100644 src/vppinfra/bihash_48_8.h create mode 100644 src/vppinfra/tw_timer_16t_1w_2048sl.c create mode 100644 src/vppinfra/tw_timer_16t_1w_2048sl.h (limited to 'src/vnet/vxlan-gpe') diff --git a/src/Makefile.am b/src/Makefile.am index 08feb29a..641707ed 100644 --- a/src/Makefile.am +++ b/src/Makefile.am @@ -88,6 +88,7 @@ include vlib-api.am include vnet.am include vpp.am include vpp-api-test.am +include uri.am SUBDIRS += plugins diff --git a/src/plugins/ioam/export-common/ioam_export.h b/src/plugins/ioam/export-common/ioam_export.h index e84dab0b..dd48a93b 100644 --- a/src/plugins/ioam/export-common/ioam_export.h +++ b/src/plugins/ioam/export-common/ioam_export.h @@ -21,7 +21,7 @@ #include #include #include -#include +#include #include #include diff --git a/src/plugins/ioam/ipfixcollector/ipfixcollector.c b/src/plugins/ioam/ipfixcollector/ipfixcollector.c index 4ae47edc..71b934ec 100644 --- a/src/plugins/ioam/ipfixcollector/ipfixcollector.c +++ b/src/plugins/ioam/ipfixcollector/ipfixcollector.c @@ -15,7 +15,7 @@ #include #include -#include +#include #include ipfix_collector_main_t ipfix_collector_main; diff --git a/src/plugins/ioam/lib-vxlan-gpe/ioam_transit.c b/src/plugins/ioam/lib-vxlan-gpe/ioam_transit.c index b42c357c..f334c983 100644 --- a/src/plugins/ioam/lib-vxlan-gpe/ioam_transit.c +++ b/src/plugins/ioam/lib-vxlan-gpe/ioam_transit.c @@ -16,7 +16,7 @@ #include #include #include -#include +#include #include #include #include diff --git a/src/plugins/snat/in2out.c b/src/plugins/snat/in2out.c index e30c913c..b4b7793d 100644 --- a/src/plugins/snat/in2out.c +++ b/src/plugins/snat/in2out.c @@ -689,12 +689,12 @@ snat_hairpinning (snat_main_t *sm, ip4_header_t, dst_address); ip0->checksum = ip_csum_fold (sum0); - old_dst_port0 = tcp0->ports.dst; + old_dst_port0 = tcp0->dst; if (PREDICT_TRUE(new_dst_port0 != old_dst_port0)) { if (PREDICT_TRUE(proto0 == SNAT_PROTOCOL_TCP)) { - tcp0->ports.dst = new_dst_port0; + tcp0->dst = new_dst_port0; sum0 = tcp0->checksum; sum0 = ip_csum_update (sum0, old_dst_addr0, new_dst_addr0, ip4_header_t, dst_address); @@ -872,9 +872,9 @@ snat_in2out_node_fn_inline (vlib_main_t * vm, if (PREDICT_TRUE(proto0 == SNAT_PROTOCOL_TCP)) { - old_port0 = tcp0->ports.src; - tcp0->ports.src = s0->out2in.port; - new_port0 = tcp0->ports.src; + old_port0 = tcp0->src_port; + tcp0->src_port = s0->out2in.port; + new_port0 = tcp0->src_port; sum0 = tcp0->checksum; sum0 = ip_csum_update (sum0, old_addr0, new_addr0, @@ -1012,9 +1012,9 @@ snat_in2out_node_fn_inline (vlib_main_t * vm, if (PREDICT_TRUE(proto1 == SNAT_PROTOCOL_TCP)) { - old_port1 = tcp1->ports.src; - tcp1->ports.src = s1->out2in.port; - new_port1 = tcp1->ports.src; + old_port1 = tcp1->src_port; + tcp1->src_port = s1->out2in.port; + new_port1 = tcp1->src_port; sum1 = tcp1->checksum; sum1 = ip_csum_update (sum1, old_addr1, new_addr1, @@ -1188,9 +1188,9 @@ snat_in2out_node_fn_inline (vlib_main_t * vm, if (PREDICT_TRUE(proto0 == SNAT_PROTOCOL_TCP)) { - old_port0 = tcp0->ports.src; - tcp0->ports.src = s0->out2in.port; - new_port0 = tcp0->ports.src; + old_port0 = tcp0->src_port; + tcp0->src_port = s0->out2in.port; + new_port0 = tcp0->src_port; sum0 = tcp0->checksum; sum0 = ip_csum_update (sum0, old_addr0, new_addr0, @@ -1667,8 +1667,8 @@ snat_in2out_fast_static_map_fn (vlib_main_t * vm, { if (PREDICT_TRUE(proto0 == SNAT_PROTOCOL_TCP)) { - old_port0 = tcp0->ports.src; - tcp0->ports.src = new_port0; + old_port0 = tcp0->src_port; + tcp0->src_port = new_port0; sum0 = tcp0->checksum; sum0 = ip_csum_update (sum0, old_addr0, new_addr0, diff --git a/src/plugins/snat/out2in.c b/src/plugins/snat/out2in.c index 328f5ba4..3bfc0aa3 100644 --- a/src/plugins/snat/out2in.c +++ b/src/plugins/snat/out2in.c @@ -19,7 +19,7 @@ #include #include -#include +#include #include #include #include @@ -602,9 +602,9 @@ snat_out2in_node_fn (vlib_main_t * vm, if (PREDICT_TRUE(proto0 == SNAT_PROTOCOL_TCP)) { - old_port0 = tcp0->ports.dst; - tcp0->ports.dst = s0->in2out.port; - new_port0 = tcp0->ports.dst; + old_port0 = tcp0->dst_port; + tcp0->dst_port = s0->in2out.port; + new_port0 = tcp0->dst_port; sum0 = tcp0->checksum; sum0 = ip_csum_update (sum0, old_addr0, new_addr0, @@ -737,9 +737,9 @@ snat_out2in_node_fn (vlib_main_t * vm, if (PREDICT_TRUE(proto1 == SNAT_PROTOCOL_TCP)) { - old_port1 = tcp1->ports.dst; - tcp1->ports.dst = s1->in2out.port; - new_port1 = tcp1->ports.dst; + old_port1 = tcp1->dst_port; + tcp1->dst_port = s1->in2out.port; + new_port1 = tcp1->dst_port; sum1 = tcp1->checksum; sum1 = ip_csum_update (sum1, old_addr1, new_addr1, @@ -907,9 +907,9 @@ snat_out2in_node_fn (vlib_main_t * vm, if (PREDICT_TRUE(proto0 == SNAT_PROTOCOL_TCP)) { - old_port0 = tcp0->ports.dst; - tcp0->ports.dst = s0->in2out.port; - new_port0 = tcp0->ports.dst; + old_port0 = tcp0->dst_port; + tcp0->dst_port = s0->in2out.port; + new_port0 = tcp0->dst_port; sum0 = tcp0->checksum; sum0 = ip_csum_update (sum0, old_addr0, new_addr0, @@ -1369,8 +1369,8 @@ snat_out2in_fast_node_fn (vlib_main_t * vm, { if (PREDICT_TRUE(proto0 == SNAT_PROTOCOL_TCP)) { - old_port0 = tcp0->ports.dst; - tcp0->ports.dst = new_port0; + old_port0 = tcp0->dst_port; + tcp0->dst_port = new_port0; sum0 = tcp0->checksum; sum0 = ip_csum_update (sum0, old_addr0, new_addr0, diff --git a/src/scripts/vnet/tcp b/src/scripts/vnet/tcp index a2ee8b2d..b9c23c3a 100644 --- a/src/scripts/vnet/tcp +++ b/src/scripts/vnet/tcp @@ -1,16 +1,18 @@ +loop create +set int ip address loop0 192.168.1.1/8 +set int state loop0 up + packet-generator new { name x - limit 1 + limit 2048 node ip4-input - size 64-64 + size 100-100 + interface loop0 no-recycle data { - TCP: 1.2.3.4 -> 5.6.7.8 - TCP: 1234 -> 5678 + TCP: 192.168.1.2 -> 192.168.1.1 + TCP: 32415 -> 80 + SYN incrementing 100 } } - -tr add pg-input 100 -ip route 5.6.7.8/32 via local -ip route 1.2.3.4/32 via local diff --git a/src/scripts/vnet/udp b/src/scripts/vnet/udp new file mode 100644 index 00000000..7dda1eec --- /dev/null +++ b/src/scripts/vnet/udp @@ -0,0 +1,19 @@ +loop create +set int ip address loop0 192.168.1.1/8 +set int state loop0 up + +packet-generator new { + name udp + limit 512 + rate 1e4 + node ip4-input + size 100-100 + interface loop0 + no-recycle + data { + UDP: 192.168.1.2 - 192.168.2.255 -> 192.168.1.1 + UDP: 4321 -> 1234 + length 72 + incrementing 100 + } +} diff --git a/src/scripts/vnet/uri/tcp-setup.sh b/src/scripts/vnet/uri/tcp-setup.sh new file mode 100755 index 00000000..e0b01588 --- /dev/null +++ b/src/scripts/vnet/uri/tcp-setup.sh @@ -0,0 +1,39 @@ +#!/usr/bin/env bash + +function topo_setup +{ + ip netns add vppns1 + ip link add veth_vpp1 type veth peer name vpp1 + ip link set dev vpp1 up + ip link set dev veth_vpp1 up netns vppns1 + + ip netns exec vppns1 \ + bash -c " + ip link set dev lo up + ip addr add 6.0.1.2/24 dev veth_vpp1 + " + + ethtool --offload vpp1 rx off tx off + ip netns exec vppns1 ethtool --offload veth_vpp1 rx off tx off + +} + +function topo_clean +{ + ip link del dev veth_vpp1 &> /dev/null + ip netns del vppns1 &> /dev/null +} + +if [ "$1" == "clean" ] ; then + topo_clean + exit 0 +else + topo_setup +fi + +# to test connectivity do: +# sudo ip netns exec vppns1 telnet 6.0.1.1 1234 +# to push traffic to the server +# dd if=/dev/zero bs=1024K count=512 | nc 6.0.1.1 +# to listen for incoming connection from vpp +# nc -l 1234 diff --git a/src/scripts/vnet/uri/tcp_server b/src/scripts/vnet/uri/tcp_server new file mode 100644 index 00000000..7f5a86de --- /dev/null +++ b/src/scripts/vnet/uri/tcp_server @@ -0,0 +1,4 @@ +create host-interface name vpp1 +set int state host-vpp1 up +set int ip address host-vpp1 6.0.1.1/24 +trace add af-packet-input 10 diff --git a/src/scripts/vnet/uri/udp b/src/scripts/vnet/uri/udp new file mode 100644 index 00000000..ca13b83c --- /dev/null +++ b/src/scripts/vnet/uri/udp @@ -0,0 +1,19 @@ +loop create +set int ip address loop0 10.0.0.1/32 +set int state loop0 up + +packet-generator new { + name udp + limit 512 + rate 1e4 + node ip4-input + size 100-100 + interface loop0 + no-recycle + data { + UDP: 192.168.1.2 - 192.168.2.255 -> 192.168.1.1 + UDP: 4321 -> 1234 + length 72 + incrementing 100 + } +} diff --git a/src/svm.am b/src/svm.am index 2cd385bd..442eba8e 100644 --- a/src/svm.am +++ b/src/svm.am @@ -13,13 +13,14 @@ bin_PROGRAMS += svmtool svmdbtool -nobase_include_HEADERS += svm/svm.h svm/ssvm.h svm/svmdb.h +nobase_include_HEADERS += svm/svm.h svm/ssvm.h svm/svmdb.h \ + svm/svm_fifo.h svm/svm_fifo_segment.h lib_LTLIBRARIES += libsvm.la libsvmdb.la +libsvm_la_SOURCES = svm/svm.c svm/ssvm.c svm/svm_fifo.c svm/svm_fifo_segment.c libsvm_la_LIBADD = libvppinfra.la -lrt -lpthread libsvm_la_DEPENDENCIES = libvppinfra.la -libsvm_la_SOURCES = svm/svm.c svm/ssvm.c svmtool_SOURCES = svm/svmtool.c svmtool_LDADD = libsvm.la libvppinfra.la -lpthread -lrt @@ -31,4 +32,9 @@ libsvmdb_la_SOURCES = svm/svmdb.c svmdbtool_SOURCES = svm/svmdbtool.c svmdbtool_LDADD = libsvmdb.la libsvm.la libvppinfra.la -lpthread -lrt +noinst_PROGRAMS += test_svm_fifo1 +test_svm_fifo1_SOURCES = svm/test_svm_fifo1.c +test_svm_fifo1_LDADD = libsvm.la libvppinfra.la -lpthread -lrt +test_svm_fifo1_LDFLAGS = -static + # vi:syntax=automake diff --git a/src/svm/ssvm.c b/src/svm/ssvm.c index 6f409eb6..6cda1f27 100644 --- a/src/svm/ssvm.c +++ b/src/svm/ssvm.c @@ -169,6 +169,22 @@ re_map_it: return 0; } +void +ssvm_delete (ssvm_private_t * ssvm) +{ + u8 *fn; + + fn = format (0, "/dev/shm/%s%c", ssvm->name, 0); + + /* Throw away the backing file */ + if (unlink ((char *) fn) < 0) + clib_unix_warning ("unlink segment '%s'", ssvm->name); + + munmap ((void *) ssvm->requested_va, ssvm->ssvm_size); + vec_free (fn); +} + + /* * fd.io coding-style-patch-verification: ON * diff --git a/src/svm/ssvm.h b/src/svm/ssvm.h index 9e61b9a0..bccfc164 100644 --- a/src/svm/ssvm.h +++ b/src/svm/ssvm.h @@ -38,7 +38,10 @@ #include #include -#define MMAP_PAGESIZE (4<<10) +#ifndef MMAP_PAGESIZE +#define MMAP_PAGESIZE (clib_mem_get_page_size()) +#endif + #define SSVM_N_OPAQUE 7 typedef struct @@ -125,12 +128,12 @@ ssvm_pop_heap (void *oldheap) } #define foreach_ssvm_api_error \ -_(NO_NAME, "No shared segment name", -10) \ -_(NO_SIZE, "Size not set (master)", -11) \ -_(CREATE_FAILURE, "Create failed", -12) \ -_(SET_SIZE, "Set size failed", -13) \ -_(MMAP, "mmap failed", -14) \ -_(SLAVE_TIMEOUT, "Slave map timeout", -15) +_(NO_NAME, "No shared segment name", -100) \ +_(NO_SIZE, "Size not set (master)", -101) \ +_(CREATE_FAILURE, "Create failed", -102) \ +_(SET_SIZE, "Set size failed", -103) \ +_(MMAP, "mmap failed", -104) \ +_(SLAVE_TIMEOUT, "Slave map timeout", -105) typedef enum { @@ -143,6 +146,7 @@ typedef enum int ssvm_master_init (ssvm_private_t * ssvm, u32 master_index); int ssvm_slave_init (ssvm_private_t * ssvm, int timeout_in_seconds); +void ssvm_delete (ssvm_private_t * ssvm); #endif /* __included_ssvm_h__ */ diff --git a/src/svm/svm_fifo.c b/src/svm/svm_fifo.c new file mode 100644 index 00000000..11f90193 --- /dev/null +++ b/src/svm/svm_fifo.c @@ -0,0 +1,568 @@ +/* + * Copyright (c) 2016 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +#include "svm_fifo.h" + +/** create an svm fifo, in the current heap. Fails vs blow up the process */ +svm_fifo_t * +svm_fifo_create (u32 data_size_in_bytes) +{ + svm_fifo_t *f; + pthread_mutexattr_t attr; + pthread_condattr_t cattr; + + f = clib_mem_alloc_aligned_or_null (sizeof (*f) + data_size_in_bytes, + CLIB_CACHE_LINE_BYTES); + if (f == 0) + return 0; + + memset (f, 0, sizeof (*f) + data_size_in_bytes); + f->nitems = data_size_in_bytes; + f->ooos_list_head = OOO_SEGMENT_INVALID_INDEX; + + memset (&attr, 0, sizeof (attr)); + memset (&cattr, 0, sizeof (cattr)); + + if (pthread_mutexattr_init (&attr)) + clib_unix_warning ("mutexattr_init"); + if (pthread_mutexattr_setpshared (&attr, PTHREAD_PROCESS_SHARED)) + clib_unix_warning ("pthread_mutexattr_setpshared"); + if (pthread_mutex_init (&f->mutex, &attr)) + clib_unix_warning ("mutex_init"); + if (pthread_mutexattr_destroy (&attr)) + clib_unix_warning ("mutexattr_destroy"); + if (pthread_condattr_init (&cattr)) + clib_unix_warning ("condattr_init"); + if (pthread_condattr_setpshared (&cattr, PTHREAD_PROCESS_SHARED)) + clib_unix_warning ("condattr_setpshared"); + if (pthread_cond_init (&f->condvar, &cattr)) + clib_unix_warning ("cond_init1"); + if (pthread_condattr_destroy (&cattr)) + clib_unix_warning ("cond_init2"); + + return (f); +} + +always_inline ooo_segment_t * +ooo_segment_new (svm_fifo_t * f, u32 start, u32 length) +{ + ooo_segment_t *s; + + pool_get (f->ooo_segments, s); + + s->fifo_position = start; + s->length = length; + + s->prev = s->next = OOO_SEGMENT_INVALID_INDEX; + + return s; +} + +always_inline void +ooo_segment_del (svm_fifo_t * f, u32 index) +{ + ooo_segment_t *cur, *prev = 0, *next = 0; + cur = pool_elt_at_index (f->ooo_segments, index); + + if (cur->next != OOO_SEGMENT_INVALID_INDEX) + { + next = pool_elt_at_index (f->ooo_segments, cur->next); + next->prev = cur->prev; + } + + if (cur->prev != OOO_SEGMENT_INVALID_INDEX) + { + prev = pool_elt_at_index (f->ooo_segments, cur->prev); + prev->next = cur->next; + } + else + { + f->ooos_list_head = cur->next; + } + + pool_put (f->ooo_segments, cur); +} + +/** + * Add segment to fifo's out-of-order segment list. Takes care of merging + * adjacent segments and removing overlapping ones. + */ +static void +ooo_segment_add (svm_fifo_t * f, u32 offset, u32 length) +{ + ooo_segment_t *s, *new_s, *prev, *next, *it; + u32 new_index, position, end_offset, s_sof, s_eof, s_index; + + position = (f->tail + offset) % f->nitems; + end_offset = offset + length; + + if (f->ooos_list_head == OOO_SEGMENT_INVALID_INDEX) + { + s = ooo_segment_new (f, position, length); + f->ooos_list_head = s - f->ooo_segments; + f->ooos_newest = f->ooos_list_head; + return; + } + + /* Find first segment that starts after new segment */ + s = pool_elt_at_index (f->ooo_segments, f->ooos_list_head); + while (s->next != OOO_SEGMENT_INVALID_INDEX + && ooo_segment_offset (f, s) <= offset) + s = pool_elt_at_index (f->ooo_segments, s->next); + + s_index = s - f->ooo_segments; + s_sof = ooo_segment_offset (f, s); + s_eof = ooo_segment_end_offset (f, s); + + /* No overlap, add before current segment */ + if (end_offset < s_sof) + { + new_s = ooo_segment_new (f, position, length); + new_index = new_s - f->ooo_segments; + + /* Pool might've moved, get segment again */ + s = pool_elt_at_index (f->ooo_segments, s_index); + + if (s->prev != OOO_SEGMENT_INVALID_INDEX) + { + new_s->prev = s->prev; + + prev = pool_elt_at_index (f->ooo_segments, new_s->prev); + prev->next = new_index; + } + else + { + /* New head */ + f->ooos_list_head = new_index; + } + + new_s->next = s - f->ooo_segments; + s->prev = new_index; + f->ooos_newest = new_index; + return; + } + /* No overlap, add after current segment */ + else if (s_eof < offset) + { + new_s = ooo_segment_new (f, position, length); + new_index = new_s - f->ooo_segments; + + /* Pool might've moved, get segment again */ + s = pool_elt_at_index (f->ooo_segments, s_index); + + if (s->next != OOO_SEGMENT_INVALID_INDEX) + { + new_s->next = s->next; + + next = pool_elt_at_index (f->ooo_segments, new_s->next); + next->prev = new_index; + } + + new_s->prev = s - f->ooo_segments; + s->next = new_index; + f->ooos_newest = new_index; + + return; + } + + /* + * Merge needed + */ + + /* Merge at head */ + if (offset <= s_sof) + { + /* If we have a previous, check if we overlap */ + if (s->prev != OOO_SEGMENT_INVALID_INDEX) + { + prev = pool_elt_at_index (f->ooo_segments, s->prev); + + /* New segment merges prev and current. Remove previous and + * update position of current. */ + if (ooo_segment_end_offset (f, prev) >= offset) + { + s->fifo_position = prev->fifo_position; + s->length = s_eof - ooo_segment_offset (f, prev); + ooo_segment_del (f, s->prev); + } + } + else + { + s->fifo_position = position; + s->length = s_eof - ooo_segment_offset (f, s); + } + + /* The new segment's tail may cover multiple smaller ones */ + if (s_eof < end_offset) + { + /* Remove segments completely covered */ + it = (s->next != OOO_SEGMENT_INVALID_INDEX) ? + pool_elt_at_index (f->ooo_segments, s->next) : 0; + while (it && ooo_segment_end_offset (f, it) < end_offset) + { + next = (it->next != OOO_SEGMENT_INVALID_INDEX) ? + pool_elt_at_index (f->ooo_segments, it->next) : 0; + ooo_segment_del (f, it - f->ooo_segments); + it = next; + } + + /* Update length. Segment's start might have changed. */ + s->length = end_offset - ooo_segment_offset (f, s); + + /* If partial overlap with last, merge */ + if (it && ooo_segment_offset (f, it) < end_offset) + { + s->length += + it->length - (ooo_segment_offset (f, it) - end_offset); + ooo_segment_del (f, it - f->ooo_segments); + } + } + } + /* Last but overlapping previous */ + else if (s_eof <= end_offset) + { + s->length = end_offset - ooo_segment_offset (f, s); + } + /* New segment completely covered by current one */ + else + { + /* Do Nothing */ + } + + /* Most recently updated segment */ + f->ooos_newest = s - f->ooo_segments; +} + +/** + * Removes segments that can now be enqueued because the fifo's tail has + * advanced. Returns the number of bytes added to tail. + */ +static int +ooo_segment_try_collect (svm_fifo_t * f, u32 n_bytes_enqueued) +{ + ooo_segment_t *s; + u32 index, bytes = 0, diff; + + s = pool_elt_at_index (f->ooo_segments, f->ooos_list_head); + + /* If last tail update overlaps one/multiple ooo segments, remove them */ + diff = (f->nitems + f->tail - s->fifo_position) % f->nitems; + while (0 < diff && diff < n_bytes_enqueued) + { + /* Segment end is beyond the tail. Advance tail and be done */ + if (diff < s->length) + { + f->tail += s->length - diff; + f->tail %= f->nitems; + break; + } + /* If we have next go on */ + else if (s->next != OOO_SEGMENT_INVALID_INDEX) + { + index = s - f->ooo_segments; + s = pool_elt_at_index (f->ooo_segments, s->next); + diff = (f->nitems + f->tail - s->fifo_position) % f->nitems; + ooo_segment_del (f, index); + } + /* End of search */ + else + { + break; + } + } + + /* If tail is adjacent to an ooo segment, 'consume' it */ + if (diff == 0) + { + bytes = ((f->nitems - f->cursize) >= s->length) ? s->length : + f->nitems - f->cursize; + + f->tail += bytes; + f->tail %= f->nitems; + + ooo_segment_del (f, s - f->ooo_segments); + } + + return bytes; +} + +static int +svm_fifo_enqueue_internal (svm_fifo_t * f, + int pid, u32 max_bytes, u8 * copy_from_here) +{ + u32 total_copy_bytes, first_copy_bytes, second_copy_bytes; + u32 cursize, nitems; + + if (PREDICT_FALSE (f->cursize == f->nitems)) + return -2; /* fifo stuffed */ + + /* read cursize, which can only decrease while we're working */ + cursize = f->cursize; + nitems = f->nitems; + + /* Number of bytes we're going to copy */ + total_copy_bytes = (nitems - cursize) < max_bytes ? + (nitems - cursize) : max_bytes; + + if (PREDICT_TRUE (copy_from_here != 0)) + { + /* Number of bytes in first copy segment */ + first_copy_bytes = ((nitems - f->tail) < total_copy_bytes) + ? (nitems - f->tail) : total_copy_bytes; + + clib_memcpy (&f->data[f->tail], copy_from_here, first_copy_bytes); + f->tail += first_copy_bytes; + f->tail = (f->tail == nitems) ? 0 : f->tail; + + /* Number of bytes in second copy segment, if any */ + second_copy_bytes = total_copy_bytes - first_copy_bytes; + if (second_copy_bytes) + { + clib_memcpy (&f->data[f->tail], copy_from_here + first_copy_bytes, + second_copy_bytes); + f->tail += second_copy_bytes; + f->tail = (f->tail == nitems) ? 0 : f->tail; + } + } + else + { + /* Account for a zero-copy enqueue done elsewhere */ + ASSERT (max_bytes <= (nitems - cursize)); + f->tail += max_bytes; + f->tail = f->tail % nitems; + total_copy_bytes = max_bytes; + } + + /* Any out-of-order segments to collect? */ + if (PREDICT_FALSE (f->ooos_list_head != OOO_SEGMENT_INVALID_INDEX)) + total_copy_bytes += ooo_segment_try_collect (f, total_copy_bytes); + + /* Atomically increase the queue length */ + __sync_fetch_and_add (&f->cursize, total_copy_bytes); + + return (total_copy_bytes); +} + +int +svm_fifo_enqueue_nowait (svm_fifo_t * f, + int pid, u32 max_bytes, u8 * copy_from_here) +{ + return svm_fifo_enqueue_internal (f, pid, max_bytes, copy_from_here); +} + +/** Enqueue a future segment. + * Two choices: either copies the entire segment, or copies nothing + * Returns 0 of the entire segment was copied + * Returns -1 if none of the segment was copied due to lack of space + */ + +static int +svm_fifo_enqueue_with_offset_internal2 (svm_fifo_t * f, + int pid, + u32 offset, + u32 required_bytes, + u8 * copy_from_here) +{ + u32 total_copy_bytes, first_copy_bytes, second_copy_bytes; + u32 cursize, nitems; + u32 tail_plus_offset; + + ASSERT (offset > 0); + + /* read cursize, which can only decrease while we're working */ + cursize = f->cursize; + nitems = f->nitems; + + /* Will this request fit? */ + if ((required_bytes + offset) > (nitems - cursize)) + return -1; + + ooo_segment_add (f, offset, required_bytes); + + /* Number of bytes we're going to copy */ + total_copy_bytes = required_bytes; + tail_plus_offset = (f->tail + offset) % nitems; + + /* Number of bytes in first copy segment */ + first_copy_bytes = ((nitems - tail_plus_offset) < total_copy_bytes) + ? (nitems - tail_plus_offset) : total_copy_bytes; + + clib_memcpy (&f->data[tail_plus_offset], copy_from_here, first_copy_bytes); + + /* Number of bytes in second copy segment, if any */ + second_copy_bytes = total_copy_bytes - first_copy_bytes; + if (second_copy_bytes) + { + tail_plus_offset += first_copy_bytes; + tail_plus_offset %= nitems; + + ASSERT (tail_plus_offset == 0); + + clib_memcpy (&f->data[tail_plus_offset], + copy_from_here + first_copy_bytes, second_copy_bytes); + } + + return (0); +} + + +int +svm_fifo_enqueue_with_offset (svm_fifo_t * f, + int pid, + u32 offset, + u32 required_bytes, u8 * copy_from_here) +{ + return svm_fifo_enqueue_with_offset_internal2 + (f, pid, offset, required_bytes, copy_from_here); +} + + +static int +svm_fifo_dequeue_internal2 (svm_fifo_t * f, + int pid, u32 max_bytes, u8 * copy_here) +{ + u32 total_copy_bytes, first_copy_bytes, second_copy_bytes; + u32 cursize, nitems; + + if (PREDICT_FALSE (f->cursize == 0)) + return -2; /* nothing in the fifo */ + + /* read cursize, which can only increase while we're working */ + cursize = f->cursize; + nitems = f->nitems; + + /* Number of bytes we're going to copy */ + total_copy_bytes = (cursize < max_bytes) ? cursize : max_bytes; + + if (PREDICT_TRUE (copy_here != 0)) + { + /* Number of bytes in first copy segment */ + first_copy_bytes = ((nitems - f->head) < total_copy_bytes) + ? (nitems - f->head) : total_copy_bytes; + clib_memcpy (copy_here, &f->data[f->head], first_copy_bytes); + f->head += first_copy_bytes; + f->head = (f->head == nitems) ? 0 : f->head; + + /* Number of bytes in second copy segment, if any */ + second_copy_bytes = total_copy_bytes - first_copy_bytes; + if (second_copy_bytes) + { + clib_memcpy (copy_here + first_copy_bytes, + &f->data[f->head], second_copy_bytes); + f->head += second_copy_bytes; + f->head = (f->head == nitems) ? 0 : f->head; + } + } + else + { + /* Account for a zero-copy dequeue done elsewhere */ + ASSERT (max_bytes <= cursize); + f->head += max_bytes; + f->head = f->head % nitems; + cursize -= max_bytes; + total_copy_bytes = max_bytes; + } + + __sync_fetch_and_sub (&f->cursize, total_copy_bytes); + + return (total_copy_bytes); +} + +int +svm_fifo_dequeue_nowait (svm_fifo_t * f, + int pid, u32 max_bytes, u8 * copy_here) +{ + return svm_fifo_dequeue_internal2 (f, pid, max_bytes, copy_here); +} + +int +svm_fifo_peek (svm_fifo_t * f, int pid, u32 offset, u32 max_bytes, + u8 * copy_here) +{ + u32 total_copy_bytes, first_copy_bytes, second_copy_bytes; + u32 cursize, nitems; + + if (PREDICT_FALSE (f->cursize == 0)) + return -2; /* nothing in the fifo */ + + /* read cursize, which can only increase while we're working */ + cursize = f->cursize; + nitems = f->nitems; + + /* Number of bytes we're going to copy */ + total_copy_bytes = (cursize < max_bytes) ? cursize : max_bytes; + + if (PREDICT_TRUE (copy_here != 0)) + { + /* Number of bytes in first copy segment */ + first_copy_bytes = + ((nitems - f->head) < total_copy_bytes) ? + (nitems - f->head) : total_copy_bytes; + clib_memcpy (copy_here, &f->data[f->head], first_copy_bytes); + + /* Number of bytes in second copy segment, if any */ + second_copy_bytes = total_copy_bytes - first_copy_bytes; + if (second_copy_bytes) + { + clib_memcpy (copy_here + first_copy_bytes, &f->data[0], + second_copy_bytes); + } + } + return total_copy_bytes; +} + +int +svm_fifo_dequeue_drop (svm_fifo_t * f, int pid, u32 max_bytes) +{ + u32 total_drop_bytes, first_drop_bytes, second_drop_bytes; + u32 cursize, nitems; + + if (PREDICT_FALSE (f->cursize == 0)) + return -2; /* nothing in the fifo */ + + /* read cursize, which can only increase while we're working */ + cursize = f->cursize; + nitems = f->nitems; + + /* Number of bytes we're going to drop */ + total_drop_bytes = (cursize < max_bytes) ? cursize : max_bytes; + + /* Number of bytes in first copy segment */ + first_drop_bytes = + ((nitems - f->head) < total_drop_bytes) ? + (nitems - f->head) : total_drop_bytes; + f->head += first_drop_bytes; + f->head = (f->head == nitems) ? 0 : f->head; + + /* Number of bytes in second drop segment, if any */ + second_drop_bytes = total_drop_bytes - first_drop_bytes; + if (second_drop_bytes) + { + f->head += second_drop_bytes; + f->head = (f->head == nitems) ? 0 : f->head; + } + + __sync_fetch_and_sub (&f->cursize, total_drop_bytes); + + return total_drop_bytes; +} + +/* + * fd.io coding-style-patch-verification: ON + * + * Local Variables: + * eval: (c-set-style "gnu") + * End: + */ diff --git a/src/svm/svm_fifo.h b/src/svm/svm_fifo.h new file mode 100644 index 00000000..70624b74 --- /dev/null +++ b/src/svm/svm_fifo.h @@ -0,0 +1,157 @@ +/* + * Copyright (c) 2016 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ +#ifndef __included_ssvm_fifo_h__ +#define __included_ssvm_fifo_h__ + +#include +#include +#include +#include +#include +#include +#include + +typedef enum +{ + SVM_FIFO_TAG_NOT_HELD = 0, + SVM_FIFO_TAG_DEQUEUE, + SVM_FIFO_TAG_ENQUEUE, +} svm_lock_tag_t; + +/** Out-of-order segment */ +typedef struct +{ + u32 next; /**< Next linked-list element pool index */ + u32 prev; /**< Previous linked-list element pool index */ + + u32 fifo_position; /**< Start of segment, normalized*/ + u32 length; /**< Length of segment */ +} ooo_segment_t; + +#define OOO_SEGMENT_INVALID_INDEX ((u32)~0) + +typedef struct +{ + pthread_mutex_t mutex; /* 8 bytes */ + pthread_cond_t condvar; /* 8 bytes */ + u32 owner_pid; + svm_lock_tag_t tag; + volatile u32 cursize; + u32 nitems; + + /* Backpointers */ + u32 server_session_index; + u32 client_session_index; + u8 server_thread_index; + u8 client_thread_index; + CLIB_CACHE_LINE_ALIGN_MARK (end_shared); + u32 head; + CLIB_CACHE_LINE_ALIGN_MARK (end_consumer); + + /* producer */ + u32 tail; + + ooo_segment_t *ooo_segments; /**< Pool of ooo segments */ + u32 ooos_list_head; /**< Head of out-of-order linked-list */ + u32 ooos_newest; /**< Last segment to have been updated */ + + CLIB_CACHE_LINE_ALIGN_MARK (data); +} svm_fifo_t; + +static inline int +svm_fifo_lock (svm_fifo_t * f, u32 pid, u32 tag, int nowait) +{ + if (PREDICT_TRUE (nowait == 0)) + pthread_mutex_lock (&f->mutex); + else + { + if (pthread_mutex_trylock (&f->mutex)) + return -1; + } + f->owner_pid = pid; + f->tag = tag; + return 0; +} + +static inline void +svm_fifo_unlock (svm_fifo_t * f) +{ + f->owner_pid = 0; + f->tag = 0; + CLIB_MEMORY_BARRIER (); + pthread_mutex_unlock (&f->mutex); +} + +static inline u32 +svm_fifo_max_dequeue (svm_fifo_t * f) +{ + return f->cursize; +} + +static inline u32 +svm_fifo_max_enqueue (svm_fifo_t * f) +{ + return f->nitems - f->cursize; +} + +static inline u8 +svm_fifo_has_ooo_data (svm_fifo_t * f) +{ + return f->ooos_list_head != OOO_SEGMENT_INVALID_INDEX; +} + +svm_fifo_t *svm_fifo_create (u32 data_size_in_bytes); + +int svm_fifo_enqueue_nowait (svm_fifo_t * f, int pid, u32 max_bytes, + u8 * copy_from_here); + +int svm_fifo_enqueue_with_offset (svm_fifo_t * f, int pid, + u32 offset, u32 required_bytes, + u8 * copy_from_here); + +int svm_fifo_dequeue_nowait (svm_fifo_t * f, int pid, u32 max_bytes, + u8 * copy_here); + +int svm_fifo_peek (svm_fifo_t * f, int pid, u32 offset, u32 max_bytes, + u8 * copy_here); +int svm_fifo_dequeue_drop (svm_fifo_t * f, int pid, u32 max_bytes); + +always_inline ooo_segment_t * +svm_fifo_newest_ooo_segment (svm_fifo_t * f) +{ + return f->ooo_segments + f->ooos_newest; +} + +always_inline u32 +ooo_segment_offset (svm_fifo_t * f, ooo_segment_t * s) +{ + return ((f->nitems + s->fifo_position - f->tail) % f->nitems); +} + +always_inline u32 +ooo_segment_end_offset (svm_fifo_t * f, ooo_segment_t * s) +{ + return ((f->nitems + s->fifo_position + s->length - f->tail) % f->nitems); +} + +#endif /* __included_ssvm_fifo_h__ */ + +/* + * fd.io coding-style-patch-verification: ON + * + * Local Variables: + * eval: (c-set-style "gnu") + * End: + */ diff --git a/src/svm/svm_fifo_segment.c b/src/svm/svm_fifo_segment.c new file mode 100644 index 00000000..acabb3bd --- /dev/null +++ b/src/svm/svm_fifo_segment.c @@ -0,0 +1,193 @@ +/* + * Copyright (c) 2016 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +#include + +svm_fifo_segment_main_t svm_fifo_segment_main; + +/** (master) create an svm fifo segment */ +int +svm_fifo_segment_create (svm_fifo_segment_create_args_t * a) +{ + int rv; + svm_fifo_segment_private_t *s; + svm_fifo_segment_main_t *sm = &svm_fifo_segment_main; + ssvm_shared_header_t *sh; + svm_fifo_segment_header_t *fsh; + void *oldheap; + + /* Allocate a fresh segment */ + pool_get (sm->segments, s); + memset (s, 0, sizeof (*s)); + + s->ssvm.ssvm_size = a->segment_size; + s->ssvm.i_am_master = 1; + s->ssvm.my_pid = getpid (); + s->ssvm.name = (u8 *) a->segment_name; + s->ssvm.requested_va = sm->next_baseva; + + rv = ssvm_master_init (&s->ssvm, s - sm->segments); + + if (rv) + { + _vec_len (s) = vec_len (s) - 1; + return (rv); + } + + /* Note; requested_va updated due to seg base addr randomization */ + sm->next_baseva = s->ssvm.requested_va + a->segment_size; + + sh = s->ssvm.sh; + oldheap = ssvm_push_heap (sh); + + /* Set up svm_fifo_segment shared header */ + fsh = clib_mem_alloc (sizeof (*fsh)); + memset (fsh, 0, sizeof (*fsh)); + sh->opaque[0] = fsh; + s->h = fsh; + fsh->segment_name = format (0, "%s%c", a->segment_name, 0); + + /* Avoid vec_add1(...) failure when adding a fifo, etc. */ + vec_validate (fsh->fifos, 64); + _vec_len (fsh->fifos) = 0; + + ssvm_pop_heap (oldheap); + + sh->ready = 1; + a->new_segment_index = s - sm->segments; + return (0); +} + +/** (slave) attach to an svm fifo segment */ +int +svm_fifo_segment_attach (svm_fifo_segment_create_args_t * a) +{ + int rv; + svm_fifo_segment_private_t *s; + svm_fifo_segment_main_t *sm = &svm_fifo_segment_main; + ssvm_shared_header_t *sh; + svm_fifo_segment_header_t *fsh; + + /* Allocate a fresh segment */ + pool_get (sm->segments, s); + + memset (s, 0, sizeof (*s)); + + s->ssvm.ssvm_size = a->segment_size; + s->ssvm.my_pid = getpid (); + s->ssvm.name = (u8 *) a->segment_name; + s->ssvm.requested_va = sm->next_baseva; + + rv = ssvm_slave_init (&s->ssvm, sm->timeout_in_seconds); + + if (rv) + { + _vec_len (s) = vec_len (s) - 1; + return (rv); + } + + /* Fish the segment header */ + sh = s->ssvm.sh; + fsh = (svm_fifo_segment_header_t *) sh->opaque[0]; + s->h = fsh; + + a->new_segment_index = s - sm->segments; + return (0); +} + +void +svm_fifo_segment_delete (svm_fifo_segment_private_t * s) +{ + svm_fifo_segment_main_t *sm = &svm_fifo_segment_main; + ssvm_delete (&s->ssvm); + pool_put (sm->segments, s); +} + +svm_fifo_t * +svm_fifo_segment_alloc_fifo (svm_fifo_segment_private_t * s, + u32 data_size_in_bytes) +{ + ssvm_shared_header_t *sh; + svm_fifo_segment_header_t *fsh; + svm_fifo_t *f; + void *oldheap; + + sh = s->ssvm.sh; + fsh = (svm_fifo_segment_header_t *) sh->opaque[0]; + oldheap = ssvm_push_heap (sh); + + /* Note: this can fail, in which case: create another segment */ + f = svm_fifo_create (data_size_in_bytes); + if (f == 0) + { + ssvm_pop_heap (oldheap); + return (0); + } + + vec_add1 (fsh->fifos, f); + + ssvm_pop_heap (oldheap); + return (f); +} + +void +svm_fifo_segment_free_fifo (svm_fifo_segment_private_t * s, svm_fifo_t * f) +{ + ssvm_shared_header_t *sh; + svm_fifo_segment_header_t *fsh; + void *oldheap; + int i; + + sh = s->ssvm.sh; + fsh = (svm_fifo_segment_header_t *) sh->opaque[0]; + oldheap = ssvm_push_heap (sh); + + for (i = 0; i < vec_len (fsh->fifos); i++) + { + if (fsh->fifos[i] == f) + { + vec_delete (fsh->fifos, 1, i); + goto found; + } + } + clib_warning ("fifo 0x%llx not found in fifo table...", f); + +found: + clib_mem_free (f); + ssvm_pop_heap (oldheap); +} + +void +svm_fifo_segment_init (u64 baseva, u32 timeout_in_seconds) +{ + svm_fifo_segment_main_t *sm = &svm_fifo_segment_main; + + sm->next_baseva = baseva; + sm->timeout_in_seconds = timeout_in_seconds; +} + +u32 +svm_fifo_segment_index (svm_fifo_segment_private_t * s) +{ + return s - svm_fifo_segment_main.segments; +} + +/* + * fd.io coding-style-patch-verification: ON + * + * Local Variables: + * eval: (c-set-style "gnu") + * End: + */ diff --git a/src/svm/svm_fifo_segment.h b/src/svm/svm_fifo_segment.h new file mode 100644 index 00000000..793fa7c8 --- /dev/null +++ b/src/svm/svm_fifo_segment.h @@ -0,0 +1,89 @@ +/* + * Copyright (c) 2016 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ +#ifndef __included_ssvm_fifo_segment_h__ +#define __included_ssvm_fifo_segment_h__ + +#include "svm_fifo.h" +#include "ssvm.h" + +typedef struct +{ + volatile svm_fifo_t **fifos; + u8 *segment_name; +} svm_fifo_segment_header_t; + +typedef struct +{ + ssvm_private_t ssvm; + svm_fifo_segment_header_t *h; +} svm_fifo_segment_private_t; + +typedef struct +{ + /** pool of segments */ + svm_fifo_segment_private_t *segments; + /* Where to put the next one */ + u64 next_baseva; + u32 timeout_in_seconds; +} svm_fifo_segment_main_t; + +extern svm_fifo_segment_main_t svm_fifo_segment_main; + +typedef struct +{ + char *segment_name; + u32 segment_size; + u32 new_segment_index; +} svm_fifo_segment_create_args_t; + +static inline svm_fifo_segment_private_t * +svm_fifo_get_segment (u32 segment_index) +{ + svm_fifo_segment_main_t *ssm = &svm_fifo_segment_main; + return vec_elt_at_index (ssm->segments, segment_index); +} + +#define foreach_ssvm_fifo_segment_api_error \ +_(OUT_OF_SPACE, "Out of space in segment", -200) + +typedef enum +{ +#define _(n,s,c) SSVM_FIFO_SEGMENT_API_ERROR_##n = c, + foreach_ssvm_fifo_segment_api_error +#undef _ +} ssvm_fifo_segment_api_error_enum_t; + +int svm_fifo_segment_create (svm_fifo_segment_create_args_t * a); +int svm_fifo_segment_attach (svm_fifo_segment_create_args_t * a); +void svm_fifo_segment_delete (svm_fifo_segment_private_t * s); + +svm_fifo_t *svm_fifo_segment_alloc_fifo (svm_fifo_segment_private_t * s, + u32 data_size_in_bytes); +void svm_fifo_segment_free_fifo (svm_fifo_segment_private_t * s, + svm_fifo_t * f); + +void svm_fifo_segment_init (u64 baseva, u32 timeout_in_seconds); + +u32 svm_fifo_segment_index (svm_fifo_segment_private_t * s); + +#endif /* __included_ssvm_fifo_segment_h__ */ + +/* + * fd.io coding-style-patch-verification: ON + * + * Local Variables: + * eval: (c-set-style "gnu") + * End: + */ diff --git a/src/svm/test_svm_fifo1.c b/src/svm/test_svm_fifo1.c new file mode 100644 index 00000000..355653df --- /dev/null +++ b/src/svm/test_svm_fifo1.c @@ -0,0 +1,361 @@ +/* + * Copyright (c) 2016 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +#include "svm_fifo_segment.h" + +clib_error_t * +hello_world (int verbose) +{ + svm_fifo_segment_create_args_t _a, *a = &_a; + svm_fifo_segment_private_t *sp; + svm_fifo_t *f; + int rv; + u8 *test_data; + u8 *retrieved_data = 0; + clib_error_t *error = 0; + int pid = getpid (); + + memset (a, 0, sizeof (*a)); + + a->segment_name = "fifo-test1"; + a->segment_size = 256 << 10; + + rv = svm_fifo_segment_create (a); + + if (rv) + return clib_error_return (0, "svm_fifo_segment_create returned %d", rv); + + sp = svm_fifo_get_segment (a->new_segment_index); + + f = svm_fifo_segment_alloc_fifo (sp, 4096); + + if (f == 0) + return clib_error_return (0, "svm_fifo_segment_alloc_fifo failed"); + + test_data = format (0, "Hello world%c", 0); + vec_validate (retrieved_data, vec_len (test_data) - 1); + + while (svm_fifo_max_enqueue (f) >= vec_len (test_data)) + svm_fifo_enqueue_nowait (f, pid, vec_len (test_data), test_data); + + while (svm_fifo_max_dequeue (f) >= vec_len (test_data)) + svm_fifo_dequeue_nowait (f, pid, vec_len (retrieved_data), + retrieved_data); + + while (svm_fifo_max_enqueue (f) >= vec_len (test_data)) + svm_fifo_enqueue_nowait (f, pid, vec_len (test_data), test_data); + + while (svm_fifo_max_dequeue (f) >= vec_len (test_data)) + svm_fifo_dequeue_nowait (f, pid, vec_len (retrieved_data), + retrieved_data); + + if (!memcmp (retrieved_data, test_data, vec_len (test_data))) + error = clib_error_return (0, "data test OK, got '%s'", retrieved_data); + else + error = clib_error_return (0, "data test FAIL!"); + + svm_fifo_segment_free_fifo (sp, f); + + return error; +} + +clib_error_t * +master (int verbose) +{ + svm_fifo_segment_create_args_t _a, *a = &_a; + svm_fifo_segment_private_t *sp; + svm_fifo_t *f; + int rv; + u8 *test_data; + u8 *retrieved_data = 0; + int i; + int pid = getpid (); + + memset (a, 0, sizeof (*a)); + + a->segment_name = "fifo-test1"; + a->segment_size = 256 << 10; + + rv = svm_fifo_segment_create (a); + + if (rv) + return clib_error_return (0, "svm_fifo_segment_create returned %d", rv); + + sp = svm_fifo_get_segment (a->new_segment_index); + + f = svm_fifo_segment_alloc_fifo (sp, 4096); + + if (f == 0) + return clib_error_return (0, "svm_fifo_segment_alloc_fifo failed"); + + test_data = format (0, "Hello world%c", 0); + vec_validate (retrieved_data, vec_len (test_data) - 1); + + for (i = 0; i < 1000; i++) + svm_fifo_enqueue_nowait (f, pid, vec_len (test_data), test_data); + + return clib_error_return (0, "master (enqueue) done"); +} + +clib_error_t * +mempig (int verbose) +{ + svm_fifo_segment_create_args_t _a, *a = &_a; + svm_fifo_segment_private_t *sp; + svm_fifo_t *f; + svm_fifo_t **flist = 0; + int rv; + int i; + + memset (a, 0, sizeof (*a)); + + a->segment_name = "fifo-test1"; + a->segment_size = 256 << 10; + + rv = svm_fifo_segment_create (a); + + if (rv) + return clib_error_return (0, "svm_fifo_segment_create returned %d", rv); + + sp = svm_fifo_get_segment (a->new_segment_index); + + for (i = 0; i < 1000; i++) + { + f = svm_fifo_segment_alloc_fifo (sp, 4096); + if (f == 0) + break; + vec_add1 (flist, f); + } + + fformat (stdout, "Try #1: created %d fifos...\n", vec_len (flist)); + for (i = 0; i < vec_len (flist); i++) + { + f = flist[i]; + svm_fifo_segment_free_fifo (sp, f); + } + + _vec_len (flist) = 0; + + for (i = 0; i < 1000; i++) + { + f = svm_fifo_segment_alloc_fifo (sp, 4096); + if (f == 0) + break; + vec_add1 (flist, f); + } + + fformat (stdout, "Try #2: created %d fifos...\n", vec_len (flist)); + for (i = 0; i < vec_len (flist); i++) + { + f = flist[i]; + svm_fifo_segment_free_fifo (sp, f); + } + + return 0; +} + +clib_error_t * +offset (int verbose) +{ + svm_fifo_segment_create_args_t _a, *a = &_a; + svm_fifo_segment_private_t *sp; + svm_fifo_t *f; + int rv; + u32 *test_data = 0; + u32 *recovered_data = 0; + int i; + int pid = getpid (); + + memset (a, 0, sizeof (*a)); + + a->segment_name = "fifo-test1"; + a->segment_size = 256 << 10; + + rv = svm_fifo_segment_create (a); + + if (rv) + return clib_error_return (0, "svm_fifo_segment_create returned %d", rv); + + sp = svm_fifo_get_segment (a->new_segment_index); + + f = svm_fifo_segment_alloc_fifo (sp, 200 << 10); + + if (f == 0) + return clib_error_return (0, "svm_fifo_segment_alloc_fifo failed"); + + for (i = 0; i < (3 * 1024); i++) + vec_add1 (test_data, i); + + /* Enqueue the first 1024 u32's */ + svm_fifo_enqueue_nowait (f, pid, 4096 /* bytes to enqueue */ , + (u8 *) test_data); + + /* Enqueue the third 1024 u32's 2048 ahead of the current tail */ + svm_fifo_enqueue_with_offset (f, pid, 4096, 4096, (u8 *) & test_data[2048]); + + /* Enqueue the second 1024 u32's at the current tail */ + svm_fifo_enqueue_nowait (f, pid, 4096 /* bytes to enqueue */ , + (u8 *) & test_data[1024]); + + vec_validate (recovered_data, (3 * 1024) - 1); + + svm_fifo_dequeue_nowait (f, pid, 3 * 4096, (u8 *) recovered_data); + + for (i = 0; i < (3 * 1024); i++) + { + if (recovered_data[i] != test_data[i]) + { + clib_warning ("[%d] expected %d recovered %d", i, + test_data[i], recovered_data[i]); + return clib_error_return (0, "offset test FAILED"); + } + } + + return clib_error_return (0, "offset test OK"); +} + +clib_error_t * +slave (int verbose) +{ + svm_fifo_segment_create_args_t _a, *a = &_a; + svm_fifo_segment_private_t *sp; + svm_fifo_segment_header_t *fsh; + svm_fifo_t *f; + ssvm_shared_header_t *sh; + int rv; + u8 *test_data; + u8 *retrieved_data = 0; + int pid = getpid (); + int i; + + memset (a, 0, sizeof (*a)); + + a->segment_name = "fifo-test1"; + + rv = svm_fifo_segment_attach (a); + + if (rv) + return clib_error_return (0, "svm_fifo_segment_attach returned %d", rv); + + sp = svm_fifo_get_segment (a->new_segment_index); + sh = sp->ssvm.sh; + fsh = (svm_fifo_segment_header_t *) sh->opaque[0]; + + /* might wanna wait.. */ + f = (svm_fifo_t *) fsh->fifos[0]; + + /* Lazy bastards united */ + test_data = format (0, "Hello world%c", 0); + vec_validate (retrieved_data, vec_len (test_data) - 1); + + for (i = 0; i < 1000; i++) + { + svm_fifo_dequeue_nowait (f, pid, vec_len (retrieved_data), + retrieved_data); + if (memcmp (retrieved_data, test_data, vec_len (retrieved_data))) + return clib_error_return (0, "retrieved data incorrect, '%s'", + retrieved_data); + } + + return clib_error_return (0, "slave (dequeue) done"); +} + + +int +test_ssvm_fifo1 (unformat_input_t * input) +{ + clib_error_t *error = 0; + int verbose = 0; + int test_id = 0; + + svm_fifo_segment_init (0x200000000ULL, 20); + + while (unformat_check_input (input) != UNFORMAT_END_OF_INPUT) + { + if (unformat (input, "verbose %d", &verbose)) + ; + else if (unformat (input, "verbose")) + verbose = 1; + else if (unformat (input, "master")) + test_id = 1; + else if (unformat (input, "slave")) + test_id = 2; + else if (unformat (input, "mempig")) + test_id = 3; + else if (unformat (input, "offset")) + test_id = 4; + else + { + error = clib_error_create ("unknown input `%U'\n", + format_unformat_error, input); + goto out; + } + } + + switch (test_id) + { + case 0: + error = hello_world (verbose); + break; + + case 1: + error = master (verbose); + break; + + case 2: + error = slave (verbose); + break; + + case 3: + error = mempig (verbose); + break; + + case 4: + error = offset (verbose); + break; + + default: + error = clib_error_return (0, "test id %d unknown", test_id); + break; + } + +out: + if (error) + clib_error_report (error); + + return 0; +} + + + +int +main (int argc, char *argv[]) +{ + unformat_input_t i; + int r; + + unformat_init_command_line (&i, argv); + r = test_ssvm_fifo1 (&i); + unformat_free (&i); + return r; +} + +/* + * fd.io coding-style-patch-verification: ON + * + * Local Variables: + * eval: (c-set-style "gnu") + * End: + */ diff --git a/src/uri.am b/src/uri.am new file mode 100644 index 00000000..8cdd77c6 --- /dev/null +++ b/src/uri.am @@ -0,0 +1,22 @@ +# Copyright (c) 2016 Cisco and/or its affiliates. +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at: +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +noinst_PROGRAMS += uri_udp_test2 uri_tcp_test + +uri_udp_test2_SOURCES = uri/uri_udp_test2.c +uri_udp_test2_LDADD = libvlibmemoryclient.la libvlibapi.la libsvm.la \ + libvppinfra.la -lpthread -lm -lrt + +uri_tcp_test_SOURCES = uri/uri_tcp_test.c +uri_tcp_test_LDADD = libvlibmemoryclient.la libvlibapi.la libsvm.la \ + libvppinfra.la -lpthread -lm -lrt diff --git a/src/uri/uri_tcp_test.c b/src/uri/uri_tcp_test.c new file mode 100644 index 00000000..ed5a37d8 --- /dev/null +++ b/src/uri/uri_tcp_test.c @@ -0,0 +1,916 @@ +/* + * Copyright (c) 2016 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +#include +#include +#include +#include +#include +#include +#include + +#include "../vnet/session/application_interface.h" + +#define vl_typedefs /* define message structures */ +#include +#undef vl_typedefs + +/* declare message handlers for each api */ + +#define vl_endianfun /* define message structures */ +#include +#undef vl_endianfun + +/* instantiate all the print functions we know about */ +#define vl_print(handle, ...) +#define vl_printfun +#include +#undef vl_printfun + +/* Satisfy external references when not linking with -lvlib */ +vlib_main_t vlib_global_main; +vlib_main_t **vlib_mains; + +typedef struct +{ + svm_fifo_t * server_rx_fifo; + svm_fifo_t * server_tx_fifo; + + u32 vpp_session_index; + u32 vpp_session_thread; +} session_t; + +typedef enum +{ + STATE_START, + STATE_READY, + STATE_DISCONNECTING, + STATE_FAILED +} connection_state_t; + +typedef struct +{ + /* vpe input queue */ + unix_shared_memory_queue_t *vl_input_queue; + + /* API client handle */ + u32 my_client_index; + + /* The URI we're playing with */ + u8 * uri; + + /* Session pool */ + session_t * sessions; + + /* Hash table for disconnect processing */ + uword * session_index_by_vpp_handles; + + /* intermediate rx buffer */ + u8 * rx_buf; + + /* URI for slave's connect */ + u8 * connect_uri; + + u32 connected_session_index; + + int i_am_master; + + /* drop all packets */ + int drop_packets; + + /* Our event queue */ + unix_shared_memory_queue_t * our_event_queue; + + /* $$$ single thread only for the moment */ + unix_shared_memory_queue_t * vpp_event_queue; + + pid_t my_pid; + + /* For deadman timers */ + clib_time_t clib_time; + + /* State of the connection, shared between msg RX thread and main thread */ + volatile connection_state_t state; + + /* Signal variables */ + volatile int time_to_stop; + volatile int time_to_print_stats; + + u32 configured_segment_size; + + /* VNET_API_ERROR_FOO -> "Foo" hash table */ + uword * error_string_by_error_number; + + /* convenience */ + svm_fifo_segment_main_t * segment_main; + + u8 *connect_test_data; +} uri_tcp_test_main_t; + +uri_tcp_test_main_t uri_tcp_test_main; + +#if CLIB_DEBUG > 0 +#define NITER 10000 +#else +#define NITER 4000000 +#endif + +int +wait_for_state_change (uri_tcp_test_main_t * utm, connection_state_t state) +{ +#if CLIB_DEBUG > 0 +#define TIMEOUT 600.0 +#else +#define TIMEOUT 600.0 +#endif + + f64 timeout = clib_time_now (&utm->clib_time) + TIMEOUT; + + while (clib_time_now (&utm->clib_time) < timeout) + { + if (utm->state == state) + return 0; + if (utm->state == STATE_FAILED) + return -1; + } + clib_warning ("timeout waiting for STATE_READY"); + return -1; +} + +static void +init_error_string_table (uri_tcp_test_main_t * utm) +{ + utm->error_string_by_error_number = hash_create (0, sizeof (uword)); + +#define _(n,v,s) hash_set (utm->error_string_by_error_number, -v, s); + foreach_vnet_api_error; +#undef _ + + hash_set (utm->error_string_by_error_number, 99, "Misc"); +} + +static void +stop_signal (int signum) +{ + uri_tcp_test_main_t *um = &uri_tcp_test_main; + + um->time_to_stop = 1; +} + +static void +stats_signal (int signum) +{ + uri_tcp_test_main_t *um = &uri_tcp_test_main; + + um->time_to_print_stats = 1; +} + +static clib_error_t * +setup_signal_handlers (void) +{ + signal (SIGINT, stats_signal); + signal (SIGQUIT, stop_signal); + signal (SIGTERM, stop_signal); + + return 0; +} + +void +vlib_cli_output (struct vlib_main_t *vm, char *fmt, ...) +{ + clib_warning ("BUG"); +} + +int +connect_to_vpp (char *name) +{ + uri_tcp_test_main_t *utm = &uri_tcp_test_main; + api_main_t *am = &api_main; + + if (vl_client_connect_to_vlib ("/vpe-api", name, 32) < 0) + return -1; + + utm->vl_input_queue = am->shmem_hdr->vl_input_queue; + utm->my_client_index = am->my_client_index; + + return 0; +} + +static void +vl_api_map_another_segment_t_handler (vl_api_map_another_segment_t *mp) +{ + svm_fifo_segment_create_args_t _a, *a = &_a; + int rv; + + a->segment_name = (char *) mp->segment_name; + a->segment_size = mp->segment_size; + /* Attach to the segment vpp created */ + rv = svm_fifo_segment_attach (a); + if (rv) + { + clib_warning ("svm_fifo_segment_attach ('%s') failed", + mp->segment_name); + return; + } + clib_warning ("Mapped new segment '%s' size %d", mp->segment_name, + mp->segment_size); +} + +static void +vl_api_disconnect_session_t_handler (vl_api_disconnect_session_t * mp) +{ + uri_tcp_test_main_t *utm = &uri_tcp_test_main; + session_t * session; + vl_api_disconnect_session_reply_t * rmp; + uword * p; + int rv = 0; + u64 key; + + key = (((u64)mp->session_thread_index) << 32) | (u64)mp->session_index; + + p = hash_get (utm->session_index_by_vpp_handles, key); + + if (p) + { + session = pool_elt_at_index (utm->sessions, p[0]); + hash_unset (utm->session_index_by_vpp_handles, key); + pool_put (utm->sessions, session); + } + else + { + clib_warning ("couldn't find session key %llx", key); + rv = -11; + } + + rmp = vl_msg_api_alloc (sizeof (*rmp)); + memset (rmp, 0, sizeof (*rmp)); + + rmp->_vl_msg_id = ntohs (VL_API_DISCONNECT_SESSION_REPLY); + rmp->retval = rv; + rmp->session_index = mp->session_index; + rmp->session_thread_index = mp->session_thread_index; + vl_msg_api_send_shmem (utm->vl_input_queue, (u8 *)&rmp); +} + +static void +vl_api_reset_session_t_handler (vl_api_reset_session_t * mp) +{ + uri_tcp_test_main_t *utm = &uri_tcp_test_main; + session_t * session; + vl_api_reset_session_reply_t * rmp; + uword * p; + int rv = 0; + u64 key; + + key = (((u64)mp->session_thread_index) << 32) | (u64)mp->session_index; + + p = hash_get(utm->session_index_by_vpp_handles, key); + + if (p) + { + session = pool_elt_at_index(utm->sessions, p[0]); + hash_unset(utm->session_index_by_vpp_handles, key); + pool_put(utm->sessions, session); + } + else + { + clib_warning("couldn't find session key %llx", key); + rv = -11; + } + + rmp = vl_msg_api_alloc (sizeof (*rmp)); + memset (rmp, 0, sizeof (*rmp)); + rmp->_vl_msg_id = ntohs (VL_API_DISCONNECT_SESSION_REPLY); + rmp->retval = rv; + rmp->session_index = mp->session_index; + rmp->session_thread_index = mp->session_thread_index; + vl_msg_api_send_shmem (utm->vl_input_queue, (u8 *)&rmp); +} + +void +handle_fifo_event_connect_rx (uri_tcp_test_main_t *utm, session_fifo_event_t * e) +{ + svm_fifo_t * rx_fifo; + int n_read, bytes; + + rx_fifo = e->fifo; + + bytes = e->enqueue_length; + do + { + n_read = svm_fifo_dequeue_nowait (rx_fifo, 0, vec_len(utm->rx_buf), + utm->rx_buf); + if (n_read > 0) + bytes -= n_read; + } + while (n_read < 0 || bytes > 0); + + // bytes_to_read = svm_fifo_max_dequeue (rx_fifo); + // + // bytes_to_read = vec_len(utm->rx_buf) > bytes_to_read ? + // bytes_to_read : vec_len(utm->rx_buf); + // + // buffer_offset = 0; + // while (bytes_to_read > 0) + // { + // rv = svm_fifo_dequeue_nowait2 (rx_fifo, mypid, + // bytes_to_read, + // utm->rx_buf + buffer_offset); + // if (rv > 0) + // { + // bytes_to_read -= rv; + // buffer_offset += rv; + // bytes_received += rv; + // } + // } + + + // while (bytes_received < bytes_sent) + // { + // rv = svm_fifo_dequeue_nowait2 (rx_fifo, mypid, + // vec_len (utm->rx_buf), + // utm->rx_buf); + // if (rv > 0) + // { + //#if CLIB_DEBUG > 0 + // int j; + // for (j = 0; j < rv; j++) + // { + // if (utm->rx_buf[j] != ((bytes_received + j) & 0xff)) + // { + // clib_warning ("error at byte %lld, 0x%x not 0x%x", + // bytes_received + j, + // utm->rx_buf[j], + // ((bytes_received + j )&0xff)); + // } + // } + //#endif + // bytes_received += (u64) rv; + // } + // } +} + +void +handle_connect_event_queue (uri_tcp_test_main_t * utm) +{ + session_fifo_event_t _e, *e = &_e;; + + unix_shared_memory_queue_sub (utm->our_event_queue, (u8 *) e, 0 /* nowait */); + switch (e->event_type) + { + case FIFO_EVENT_SERVER_RX: + handle_fifo_event_connect_rx (utm, e); + break; + + case FIFO_EVENT_SERVER_EXIT: + return; + + default: + clib_warning("unknown event type %d", e->event_type); + break; + } +} + +void +uri_tcp_connect_send (uri_tcp_test_main_t *utm) +{ + u8 *test_data = utm->connect_test_data; + u64 bytes_sent = 0; + int rv; + int mypid = getpid(); + session_t * session; + svm_fifo_t *tx_fifo; + int buffer_offset, bytes_to_send = 0; + session_fifo_event_t evt; + static int serial_number = 0; + int i; + u32 max_chunk = 64 << 10, write; + + session = pool_elt_at_index (utm->sessions, utm->connected_session_index); + tx_fifo = session->server_tx_fifo; + + vec_validate (utm->rx_buf, vec_len (test_data) - 1); + + for (i = 0; i < 10; i++) + { + bytes_to_send = vec_len (test_data); + buffer_offset = 0; + while (bytes_to_send > 0) + { + write = bytes_to_send > max_chunk ? max_chunk : bytes_to_send; + rv = svm_fifo_enqueue_nowait (tx_fifo, mypid, write, + test_data + buffer_offset); + + if (rv > 0) + { + bytes_to_send -= rv; + buffer_offset += rv; + bytes_sent += rv; + + /* Fabricate TX event, send to vpp */ + evt.fifo = tx_fifo; + evt.event_type = FIFO_EVENT_SERVER_TX; + /* $$$$ for event logging */ + evt.enqueue_length = rv; + evt.event_id = serial_number++; + + unix_shared_memory_queue_add (utm->vpp_event_queue, (u8 *) &evt, + 0 /* do wait for mutex */); + } + } + } +} + +static void +uri_tcp_client_test (uri_tcp_test_main_t * utm) +{ + vl_api_connect_uri_t * cmp; + vl_api_disconnect_session_t *dmp; + session_t *connected_session; + int i; + + cmp = vl_msg_api_alloc (sizeof (*cmp)); + memset (cmp, 0, sizeof (*cmp)); + + cmp->_vl_msg_id = ntohs (VL_API_CONNECT_URI); + cmp->client_index = utm->my_client_index; + cmp->context = ntohl(0xfeedface); + memcpy (cmp->uri, utm->connect_uri, vec_len (utm->connect_uri)); + vl_msg_api_send_shmem (utm->vl_input_queue, (u8 *)&cmp); + + if (wait_for_state_change (utm, STATE_READY)) + { + return; + } + + /* Init test data */ + vec_validate (utm->connect_test_data, 64 * 1024 - 1); + for (i = 0; i < vec_len (utm->connect_test_data); i++) + utm->connect_test_data[i] = i & 0xff; + + /* Start reader thread */ + /* handle_connect_event_queue (utm); */ + + /* Start send */ + uri_tcp_connect_send (utm); + + /* Disconnect */ + connected_session = pool_elt_at_index(utm->sessions, + utm->connected_session_index); + dmp = vl_msg_api_alloc (sizeof (*dmp)); + memset (dmp, 0, sizeof (*dmp)); + dmp->_vl_msg_id = ntohs (VL_API_DISCONNECT_SESSION); + dmp->client_index = utm->my_client_index; + dmp->session_index = connected_session->vpp_session_index; + dmp->session_thread_index = connected_session->vpp_session_thread; + vl_msg_api_send_shmem (utm->vl_input_queue, (u8 *)&dmp); +} + +void +handle_fifo_event_server_rx (uri_tcp_test_main_t *utm, session_fifo_event_t * e) +{ + svm_fifo_t * rx_fifo, * tx_fifo; + int n_read; + + session_fifo_event_t evt; + unix_shared_memory_queue_t *q; + int rv, bytes; + + rx_fifo = e->fifo; + tx_fifo = utm->sessions[rx_fifo->client_session_index].server_tx_fifo; + + bytes = e->enqueue_length; + do + { + n_read = svm_fifo_dequeue_nowait (rx_fifo, 0, vec_len(utm->rx_buf), + utm->rx_buf); + + /* Reflect if a non-drop session */ + if (!utm->drop_packets && n_read > 0) + { + do + { + rv = svm_fifo_enqueue_nowait (tx_fifo, 0, n_read, utm->rx_buf); + } + while (rv == -2); + + /* Fabricate TX event, send to vpp */ + evt.fifo = tx_fifo; + evt.event_type = FIFO_EVENT_SERVER_TX; + /* $$$$ for event logging */ + evt.enqueue_length = n_read; + evt.event_id = e->event_id; + q = utm->vpp_event_queue; + unix_shared_memory_queue_add (q, (u8 *) &evt, 0 /* do wait for mutex */); + } + + if (n_read > 0) + bytes -= n_read; + } + while (n_read < 0 || bytes > 0); +} + +void +handle_event_queue (uri_tcp_test_main_t * utm) +{ + session_fifo_event_t _e, *e = &_e;; + + while (1) + { + unix_shared_memory_queue_sub (utm->our_event_queue, (u8 *)e, + 0 /* nowait */); + switch (e->event_type) + { + case FIFO_EVENT_SERVER_RX: + handle_fifo_event_server_rx (utm, e); + break; + + case FIFO_EVENT_SERVER_EXIT: + return; + + default: + clib_warning ("unknown event type %d", e->event_type); + break; + } + if (PREDICT_FALSE(utm->time_to_stop == 1)) + break; + if (PREDICT_FALSE(utm->time_to_print_stats == 1)) + { + utm->time_to_print_stats = 0; + fformat(stdout, "%d connections\n", pool_elts (utm->sessions)); + } + } +} + +static void +vl_api_bind_uri_reply_t_handler (vl_api_bind_uri_reply_t * mp) +{ + uri_tcp_test_main_t *utm = &uri_tcp_test_main; + svm_fifo_segment_create_args_t _a, *a = &_a; + int rv; + + if (mp->retval) + { + clib_warning("bind failed: %d", mp->retval); + return; + } + + if (mp->segment_name_length == 0) + { + clib_warning("segment_name_length zero"); + return; + } + + a->segment_name = (char *) mp->segment_name; + a->segment_size = mp->segment_size; + + ASSERT(mp->server_event_queue_address); + + /* Attach to the segment vpp created */ + rv = svm_fifo_segment_attach (a); + if (rv) + { + clib_warning("svm_fifo_segment_attach ('%s') failed", mp->segment_name); + return; + } + + utm->our_event_queue = + (unix_shared_memory_queue_t *) mp->server_event_queue_address; + + utm->state = STATE_READY; +} + +static void +vl_api_connect_uri_reply_t_handler (vl_api_connect_uri_reply_t * mp) +{ + uri_tcp_test_main_t *utm = &uri_tcp_test_main; + svm_fifo_segment_create_args_t _a, *a = &_a; + session_t *session; + u32 session_index; + svm_fifo_t *rx_fifo, *tx_fifo; + int rv; + + if (mp->retval) + { + clib_warning ("connection failed with code: %d", mp->retval); + utm->state = STATE_FAILED; + return; + } + /* + * Attatch to segment + */ + + if (mp->segment_name_length == 0) + { + clib_warning ("segment_name_length zero"); + utm->state = STATE_FAILED; + return; + } + + a->segment_name = (char *) mp->segment_name; + a->segment_size = mp->segment_size; + + ASSERT(mp->client_event_queue_address); + + /* Attach to the segment vpp created */ + rv = svm_fifo_segment_attach (a); + if (rv) + { + clib_warning ("svm_fifo_segment_attach ('%s') failed", + mp->segment_name); + return; + } + + /* + * Save the queues + */ + + utm->our_event_queue = (unix_shared_memory_queue_t *) + mp->client_event_queue_address; + + utm->vpp_event_queue = (unix_shared_memory_queue_t *) + mp->vpp_event_queue_address; + + /* + * Setup session + */ + + pool_get (utm->sessions, session); + session_index = session - utm->sessions; + + rx_fifo = (svm_fifo_t *)mp->server_rx_fifo; + rx_fifo->client_session_index = session_index; + tx_fifo = (svm_fifo_t *)mp->server_tx_fifo; + tx_fifo->client_session_index = session_index; + + session->server_rx_fifo = rx_fifo; + session->server_tx_fifo = tx_fifo; + session->vpp_session_index = mp->session_index; + session->vpp_session_thread = mp->session_thread_index; + + /* Save handle */ + utm->connected_session_index = session_index; + + utm->state = STATE_READY; +} + +void +uri_tcp_bind (uri_tcp_test_main_t *utm) +{ + vl_api_bind_uri_t * bmp; + u32 fifo_size = 3 << 20; + bmp = vl_msg_api_alloc (sizeof (*bmp)); + memset (bmp, 0, sizeof (*bmp)); + + bmp->_vl_msg_id = ntohs (VL_API_BIND_URI); + bmp->client_index = utm->my_client_index; + bmp->context = ntohl(0xfeedface); + bmp->initial_segment_size = 256<<20; /* size of initial segment */ + bmp->options[SESSION_OPTIONS_FLAGS] = + SESSION_OPTIONS_FLAGS_USE_FIFO | SESSION_OPTIONS_FLAGS_ADD_SEGMENT; + bmp->options[SESSION_OPTIONS_RX_FIFO_SIZE] = fifo_size; + bmp->options[SESSION_OPTIONS_TX_FIFO_SIZE] = fifo_size; + bmp->options[SESSION_OPTIONS_ADD_SEGMENT_SIZE] = 128<<20; + memcpy (bmp->uri, utm->uri, vec_len (utm->uri)); + vl_msg_api_send_shmem (utm->vl_input_queue, (u8 *)&bmp); +} + +static void +vl_api_unbind_uri_reply_t_handler (vl_api_unbind_uri_reply_t *mp) +{ + uri_tcp_test_main_t *utm = &uri_tcp_test_main; + + if (mp->retval != 0) + clib_warning ("returned %d", ntohl(mp->retval)); + + utm->state = STATE_START; +} + +void +uri_tcp_unbind (uri_tcp_test_main_t *utm) +{ + vl_api_unbind_uri_t * ump; + + ump = vl_msg_api_alloc (sizeof (*ump)); + memset (ump, 0, sizeof (*ump)); + + ump->_vl_msg_id = ntohs (VL_API_UNBIND_URI); + ump->client_index = utm->my_client_index; + memcpy (ump->uri, utm->uri, vec_len (utm->uri)); + vl_msg_api_send_shmem (utm->vl_input_queue, (u8 *)&ump); +} + +static void +vl_api_accept_session_t_handler (vl_api_accept_session_t * mp) +{ + uri_tcp_test_main_t *utm = &uri_tcp_test_main; + vl_api_accept_session_reply_t *rmp; + svm_fifo_t * rx_fifo, * tx_fifo; + session_t * session; + static f64 start_time; + u64 key; + u32 session_index; + + if (start_time == 0.0) + start_time = clib_time_now (&utm->clib_time); + + utm->vpp_event_queue = (unix_shared_memory_queue_t *) + mp->vpp_event_queue_address; + + /* Allocate local session and set it up */ + pool_get (utm->sessions, session); + session_index = session - utm->sessions; + + rx_fifo = (svm_fifo_t *)mp->server_rx_fifo; + rx_fifo->client_session_index = session_index; + tx_fifo = (svm_fifo_t *)mp->server_tx_fifo; + tx_fifo->client_session_index = session_index; + + session->server_rx_fifo = rx_fifo; + session->server_tx_fifo = tx_fifo; + + /* Add it to lookup table */ + key = (((u64)mp->session_thread_index) << 32) | (u64)mp->session_index; + hash_set (utm->session_index_by_vpp_handles, key, session_index); + + utm->state = STATE_READY; + + /* Stats printing */ + if (pool_elts (utm->sessions) && (pool_elts(utm->sessions) % 20000) == 0) + { + f64 now = clib_time_now (&utm->clib_time); + fformat (stdout, "%d active sessions in %.2f seconds, %.2f/sec...\n", + pool_elts(utm->sessions), now - start_time, + (f64)pool_elts(utm->sessions) / (now - start_time)); + } + + /* Send accept reply to vpp */ + rmp = vl_msg_api_alloc (sizeof (*rmp)); + memset (rmp, 0, sizeof (*rmp)); + rmp->_vl_msg_id = ntohs (VL_API_ACCEPT_SESSION_REPLY); + rmp->session_type = mp->session_type; + rmp->session_index = mp->session_index; + rmp->session_thread_index = mp->session_thread_index; + vl_msg_api_send_shmem (utm->vl_input_queue, (u8 *)&rmp); +} + +void +uri_tcp_server_test (uri_tcp_test_main_t * utm) +{ + + /* Bind to uri */ + uri_tcp_bind (utm); + + if (wait_for_state_change (utm, STATE_READY)) + { + clib_warning ("timeout waiting for STATE_READY"); + return; + } + + /* Enter handle event loop */ + handle_event_queue (utm); + + /* Cleanup */ + uri_tcp_unbind (utm); + + if (wait_for_state_change (utm, STATE_START)) + { + clib_warning ("timeout waiting for STATE_START"); + return; + } + + fformat (stdout, "Test complete...\n"); +} + +#define foreach_uri_msg \ +_(BIND_URI_REPLY, bind_uri_reply) \ +_(UNBIND_URI_REPLY, unbind_uri_reply) \ +_(ACCEPT_SESSION, accept_session) \ +_(CONNECT_URI_REPLY, connect_uri_reply) \ +_(DISCONNECT_SESSION, disconnect_session) \ +_(RESET_SESSION, reset_session) \ +_(MAP_ANOTHER_SEGMENT, map_another_segment) + +void +uri_api_hookup (uri_tcp_test_main_t * utm) +{ +#define _(N,n) \ + vl_msg_api_set_handlers(VL_API_##N, #n, \ + vl_api_##n##_t_handler, \ + vl_noop_handler, \ + vl_api_##n##_t_endian, \ + vl_api_##n##_t_print, \ + sizeof(vl_api_##n##_t), 1); + foreach_uri_msg; +#undef _ +} + +int +main (int argc, char **argv) +{ + uri_tcp_test_main_t *utm = &uri_tcp_test_main; + unformat_input_t _argv, *a = &_argv; + u8 *chroot_prefix; + u8 *heap; + u8 * bind_name = (u8 *) "tcp://0.0.0.0/1234"; + u32 tmp; + mheap_t *h; + session_t * session; + int i; + int i_am_master = 1, drop_packets = 0; + + clib_mem_init (0, 256 << 20); + + heap = clib_mem_get_per_cpu_heap (); + h = mheap_header (heap); + + /* make the main heap thread-safe */ + h->flags |= MHEAP_FLAG_THREAD_SAFE; + + vec_validate (utm->rx_buf, 65536); + + utm->session_index_by_vpp_handles = + hash_create (0, sizeof(uword)); + + utm->my_pid = getpid(); + utm->configured_segment_size = 1<<20; + + clib_time_init (&utm->clib_time); + init_error_string_table (utm); + svm_fifo_segment_init(0x200000000ULL, 20); + unformat_init_command_line (a, argv); + + while (unformat_check_input (a) != UNFORMAT_END_OF_INPUT) + { + if (unformat (a, "chroot prefix %s", &chroot_prefix)) + { + vl_set_memory_root_path ((char *) chroot_prefix); + } + else if (unformat (a, "uri %s", &bind_name)) + ; + else if (unformat (a, "segment-size %dM", &tmp)) + utm->configured_segment_size = tmp<<20; + else if (unformat (a, "segment-size %dG", &tmp)) + utm->configured_segment_size = tmp<<30; + else if (unformat (a, "master")) + i_am_master = 1; + else if (unformat (a, "slave")) + i_am_master = 0; + else if (unformat (a, "drop")) + drop_packets = 1; + else + { + fformat (stderr, "%s: usage [master|slave]\n"); + exit (1); + } + } + + utm->uri = format (0, "%s%c", bind_name, 0); + utm->i_am_master = i_am_master; + utm->segment_main = &svm_fifo_segment_main; + utm->drop_packets = drop_packets; + + utm->connect_uri = format (0, "tcp://6.0.1.2/1234%c", 0); + + setup_signal_handlers(); + uri_api_hookup (utm); + + if (connect_to_vpp (i_am_master? "uri_tcp_server":"uri_tcp_client") < 0) + { + svm_region_exit (); + fformat (stderr, "Couldn't connect to vpe, exiting...\n"); + exit (1); + } + + if (i_am_master == 0) + { + uri_tcp_client_test (utm); + exit (0); + } + + /* $$$$ hack preallocation */ + for (i = 0; i < 200000; i++) + { + pool_get (utm->sessions, session); + memset (session, 0, sizeof (*session)); + } + for (i = 0; i < 200000; i++) + pool_put_index (utm->sessions, i); + + uri_tcp_server_test (utm); + + vl_client_disconnect_from_vlib (); + exit (0); +} diff --git a/src/uri/uri_udp_test.c b/src/uri/uri_udp_test.c new file mode 100644 index 00000000..6f5284c9 --- /dev/null +++ b/src/uri/uri_udp_test.c @@ -0,0 +1,553 @@ +/* + * Copyright (c) 2016 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include + +#include + +#define vl_typedefs /* define message structures */ +#include +#undef vl_typedefs + +/* declare message handlers for each api */ + +#define vl_endianfun /* define message structures */ +#include +#undef vl_endianfun + +/* instantiate all the print functions we know about */ +#define vl_print(handle, ...) +#define vl_printfun +#include +#undef vl_printfun + +/* Satisfy external references when not linking with -lvlib */ +vlib_main_t vlib_global_main; +vlib_main_t **vlib_mains; + +typedef enum +{ + STATE_START, + STATE_READY, + STATE_DISCONNECTING, +} connection_state_t; + +typedef struct +{ + svm_fifo_t *server_rx_fifo; + svm_fifo_t *server_tx_fifo; +} session_t; + +typedef struct +{ + /* vpe input queue */ + unix_shared_memory_queue_t *vl_input_queue; + + /* API client handle */ + u32 my_client_index; + + /* The URI we're playing with */ + u8 *uri; + + /* Session pool */ + session_t *sessions; + + /* Hash table for disconnect processing */ + uword *session_index_by_vpp_handles; + + /* fifo segment */ + svm_fifo_segment_private_t *seg; + + /* intermediate rx buffer */ + u8 *rx_buf; + + /* Our event queue */ + unix_shared_memory_queue_t *our_event_queue; + + /* $$$ single thread only for the moment */ + unix_shared_memory_queue_t *vpp_event_queue; + + /* For deadman timers */ + clib_time_t clib_time; + + /* State of the connection, shared between msg RX thread and main thread */ + volatile connection_state_t state; + + volatile int time_to_stop; + volatile int time_to_print_stats; + + /* VNET_API_ERROR_FOO -> "Foo" hash table */ + uword *error_string_by_error_number; +} uri_udp_test_main_t; + +#if CLIB_DEBUG > 0 +#define NITER 1000 +#else +#define NITER 1000000 +#endif + +uri_udp_test_main_t uri_udp_test_main; + +static void +stop_signal (int signum) +{ + uri_udp_test_main_t *um = &uri_udp_test_main; + + um->time_to_stop = 1; +} + +static void +stats_signal (int signum) +{ + uri_udp_test_main_t *um = &uri_udp_test_main; + + um->time_to_print_stats = 1; +} + +static clib_error_t * +setup_signal_handlers (void) +{ + signal (SIGINT, stats_signal); + signal (SIGQUIT, stop_signal); + signal (SIGTERM, stop_signal); + + return 0; +} + +u8 * +format_api_error (u8 * s, va_list * args) +{ + uri_udp_test_main_t *utm = va_arg (*args, uri_udp_test_main_t *); + i32 error = va_arg (*args, u32); + uword *p; + + p = hash_get (utm->error_string_by_error_number, -error); + + if (p) + s = format (s, "%s", p[0]); + else + s = format (s, "%d", error); + return s; +} + +int +wait_for_state_change (uri_udp_test_main_t * utm, connection_state_t state) +{ + f64 timeout = clib_time_now (&utm->clib_time) + 5.0; + + while (clib_time_now (&utm->clib_time) < timeout) + { + if (utm->state == state) + return 0; + } + return -1; +} + +static void +vl_api_bind_uri_reply_t_handler (vl_api_bind_uri_reply_t * mp) +{ + uri_udp_test_main_t *utm = &uri_udp_test_main; + svm_fifo_segment_create_args_t _a, *a = &_a; + int rv; + + if (mp->segment_name_length == 0) + { + clib_warning ("segment_name_length zero"); + return; + } + + a->segment_name = (char *) mp->segment_name; + + /* Attach to the segment vpp created */ + rv = svm_fifo_segment_attach (a); + if (rv) + { + clib_warning ("sm_fifo_segment_create ('%s') failed", mp->segment_name); + return; + } + + utm->our_event_queue = (unix_shared_memory_queue_t *) + mp->server_event_queue_address; + + utm->state = STATE_READY; +} + +static void +vl_api_unbind_uri_reply_t_handler (vl_api_unbind_uri_reply_t * mp) +{ + uri_udp_test_main_t *utm = &uri_udp_test_main; + + if (mp->retval != 0) + clib_warning ("returned %d", ntohl (mp->retval)); + + utm->state = STATE_START; +} + +static void +vl_api_accept_session_t_handler (vl_api_accept_session_t * mp) +{ + uri_udp_test_main_t *utm = &uri_udp_test_main; + vl_api_accept_session_reply_t *rmp; + svm_fifo_t *rx_fifo, *tx_fifo; + session_t *session; + static f64 start_time; + u64 key; + + if (start_time == 0.0) + start_time = clib_time_now (&utm->clib_time); + + utm->vpp_event_queue = (unix_shared_memory_queue_t *) + mp->vpp_event_queue_address; + + pool_get (utm->sessions, session); + + rx_fifo = (svm_fifo_t *) mp->server_rx_fifo; + rx_fifo->client_session_index = session - utm->sessions; + tx_fifo = (svm_fifo_t *) mp->server_tx_fifo; + tx_fifo->client_session_index = session - utm->sessions; + + session->server_rx_fifo = rx_fifo; + session->server_tx_fifo = tx_fifo; + + key = (((u64) mp->session_thread_index) << 32) | (u64) mp->session_index; + + hash_set (utm->session_index_by_vpp_handles, key, session - utm->sessions); + + utm->state = STATE_READY; + + if (pool_elts (utm->sessions) && (pool_elts (utm->sessions) % 20000) == 0) + { + f64 now = clib_time_now (&utm->clib_time); + fformat (stdout, "%d active sessions in %.2f seconds, %.2f/sec...\n", + pool_elts (utm->sessions), now - start_time, + (f64) pool_elts (utm->sessions) / (now - start_time)); + } + + rmp = vl_msg_api_alloc (sizeof (*rmp)); + memset (rmp, 0, sizeof (*rmp)); + rmp->_vl_msg_id = ntohs (VL_API_ACCEPT_SESSION_REPLY); + rmp->session_type = mp->session_type; + rmp->session_index = mp->session_index; + rmp->session_thread_index = mp->session_thread_index; + vl_msg_api_send_shmem (utm->vl_input_queue, (u8 *) & rmp); +} + +static void +vl_api_disconnect_session_t_handler (vl_api_disconnect_session_t * mp) +{ + uri_udp_test_main_t *utm = &uri_udp_test_main; + session_t *session; + vl_api_disconnect_session_reply_t *rmp; + uword *p; + int rv = 0; + u64 key; + + key = (((u64) mp->session_thread_index) << 32) | (u64) mp->session_index; + + p = hash_get (utm->session_index_by_vpp_handles, key); + + if (p) + { + session = pool_elt_at_index (utm->sessions, p[0]); + hash_unset (utm->session_index_by_vpp_handles, key); + pool_put (utm->sessions, session); + } + else + { + clib_warning ("couldn't find session key %llx", key); + rv = -11; + } + + rmp = vl_msg_api_alloc (sizeof (*rmp)); + memset (rmp, 0, sizeof (*rmp)); + rmp->_vl_msg_id = ntohs (VL_API_DISCONNECT_SESSION_REPLY); + rmp->retval = rv; + rmp->session_index = mp->session_index; + rmp->session_thread_index = mp->session_thread_index; + vl_msg_api_send_shmem (utm->vl_input_queue, (u8 *) & rmp); +} + +#define foreach_uri_msg \ +_(BIND_URI_REPLY, bind_uri_reply) \ +_(UNBIND_URI_REPLY, unbind_uri_reply) \ +_(ACCEPT_SESSION, accept_session) \ +_(DISCONNECT_SESSION, disconnect_session) + +void +uri_api_hookup (uri_udp_test_main_t * utm) +{ +#define _(N,n) \ + vl_msg_api_set_handlers(VL_API_##N, #n, \ + vl_api_##n##_t_handler, \ + vl_noop_handler, \ + vl_api_##n##_t_endian, \ + vl_api_##n##_t_print, \ + sizeof(vl_api_##n##_t), 1); + foreach_uri_msg; +#undef _ + +} + + +int +connect_to_vpp (char *name) +{ + uri_udp_test_main_t *utm = &uri_udp_test_main; + api_main_t *am = &api_main; + + if (vl_client_connect_to_vlib ("/vpe-api", name, 32) < 0) + return -1; + + utm->vl_input_queue = am->shmem_hdr->vl_input_queue; + utm->my_client_index = am->my_client_index; + + return 0; +} + +void +vlib_cli_output (struct vlib_main_t *vm, char *fmt, ...) +{ + clib_warning ("BUG"); +} + +static void +init_error_string_table (uri_udp_test_main_t * utm) +{ + utm->error_string_by_error_number = hash_create (0, sizeof (uword)); + +#define _(n,v,s) hash_set (utm->error_string_by_error_number, -v, s); + foreach_vnet_api_error; +#undef _ + + hash_set (utm->error_string_by_error_number, 99, "Misc"); +} + +void +handle_fifo_event_server_rx (uri_udp_test_main_t * utm, + session_fifo_event_t * e) +{ + svm_fifo_t *rx_fifo, *tx_fifo; + int nbytes; + + session_fifo_event_t evt; + unix_shared_memory_queue_t *q; + int rv; + + rx_fifo = e->fifo; + tx_fifo = utm->sessions[rx_fifo->client_session_index].server_tx_fifo; + + do + { + nbytes = svm_fifo_dequeue_nowait (rx_fifo, 0, + vec_len (utm->rx_buf), utm->rx_buf); + } + while (nbytes <= 0); + do + { + rv = svm_fifo_enqueue_nowait (tx_fifo, 0, nbytes, utm->rx_buf); + } + while (rv == -2); + + /* Fabricate TX event, send to vpp */ + evt.fifo = tx_fifo; + evt.event_type = FIFO_EVENT_SERVER_TX; + /* $$$$ for event logging */ + evt.enqueue_length = nbytes; + evt.event_id = e->event_id; + q = utm->vpp_event_queue; + unix_shared_memory_queue_add (q, (u8 *) & evt, 0 /* do wait for mutex */ ); +} + +void +handle_event_queue (uri_udp_test_main_t * utm) +{ + session_fifo_event_t _e, *e = &_e;; + + while (1) + { + unix_shared_memory_queue_sub (utm->our_event_queue, (u8 *) e, + 0 /* nowait */ ); + switch (e->event_type) + { + case FIFO_EVENT_SERVER_RX: + handle_fifo_event_server_rx (utm, e); + break; + + case FIFO_EVENT_SERVER_EXIT: + return; + + default: + clib_warning ("unknown event type %d", e->event_type); + break; + } + if (PREDICT_FALSE (utm->time_to_stop == 1)) + break; + if (PREDICT_FALSE (utm->time_to_print_stats == 1)) + { + utm->time_to_print_stats = 0; + fformat (stdout, "%d connections\n", pool_elts (utm->sessions)); + } + } +} + +void +uri_udp_test (uri_udp_test_main_t * utm) +{ + vl_api_bind_uri_t *bmp; + vl_api_unbind_uri_t *ump; + + bmp = vl_msg_api_alloc (sizeof (*bmp)); + memset (bmp, 0, sizeof (*bmp)); + + bmp->_vl_msg_id = ntohs (VL_API_BIND_URI); + bmp->client_index = utm->my_client_index; + bmp->context = ntohl (0xfeedface); + bmp->segment_size = 2 << 30; + memcpy (bmp->uri, utm->uri, vec_len (utm->uri)); + vl_msg_api_send_shmem (utm->vl_input_queue, (u8 *) & bmp); + + if (wait_for_state_change (utm, STATE_READY)) + { + clib_warning ("timeout waiting for STATE_READY"); + return; + } + + handle_event_queue (utm); + + ump = vl_msg_api_alloc (sizeof (*ump)); + memset (ump, 0, sizeof (*ump)); + + ump->_vl_msg_id = ntohs (VL_API_UNBIND_URI); + ump->client_index = utm->my_client_index; + memcpy (ump->uri, utm->uri, vec_len (utm->uri)); + vl_msg_api_send_shmem (utm->vl_input_queue, (u8 *) & ump); + + if (wait_for_state_change (utm, STATE_START)) + { + clib_warning ("timeout waiting for STATE_START"); + return; + } + + fformat (stdout, "Test complete...\n"); +} + +int +main (int argc, char **argv) +{ + uri_udp_test_main_t *utm = &uri_udp_test_main; + unformat_input_t _argv, *a = &_argv; + u8 *chroot_prefix; + u8 *heap; + u8 *bind_name = (u8 *) "udp4:1234"; + mheap_t *h; + session_t *session; + int i; + + clib_mem_init (0, 256 << 20); + + heap = clib_mem_get_per_cpu_heap (); + h = mheap_header (heap); + + /* make the main heap thread-safe */ + h->flags |= MHEAP_FLAG_THREAD_SAFE; + + vec_validate (utm->rx_buf, 8192); + + utm->session_index_by_vpp_handles = hash_create (0, sizeof (uword)); + + clib_time_init (&utm->clib_time); + init_error_string_table (utm); + svm_fifo_segment_init (0x200000000ULL, 20); + unformat_init_command_line (a, argv); + + while (unformat_check_input (a) != UNFORMAT_END_OF_INPUT) + { + if (unformat (a, "chroot prefix %s", &chroot_prefix)) + { + vl_set_memory_root_path ((char *) chroot_prefix); + } + else if (unformat (a, "uri %s", &bind_name)) + ; + else + { + fformat (stderr, "%s: usage [master|slave]\n"); + exit (1); + } + } + + utm->uri = format (0, "%s%c", bind_name, 0); + + setup_signal_handlers (); + + uri_api_hookup (utm); + + if (connect_to_vpp ("uri_udp_test") < 0) + { + svm_region_exit (); + fformat (stderr, "Couldn't connect to vpe, exiting...\n"); + exit (1); + } + + /* $$$$ hack preallocation */ + for (i = 0; i < 200000; i++) + { + pool_get (utm->sessions, session); + memset (session, 0, sizeof (*session)); + } + for (i = 0; i < 200000; i++) + pool_put_index (utm->sessions, i); + + uri_udp_test (utm); + + vl_client_disconnect_from_vlib (); + exit (0); +} + +#undef vl_api_version +#define vl_api_version(n,v) static u32 vpe_api_version = v; +#include +#undef vl_api_version + +void +vl_client_add_api_signatures (vl_api_memclnt_create_t * mp) +{ + /* + * Send the main API signature in slot 0. This bit of code must + * match the checks in ../vpe/api/api.c: vl_msg_api_version_check(). + */ + mp->api_versions[0] = clib_host_to_net_u32 (vpe_api_version); +} + +/* + * fd.io coding-style-patch-verification: ON + * + * Local Variables: + * eval: (c-set-style "gnu") + * End: + */ diff --git a/src/uri/uri_udp_test2.c b/src/uri/uri_udp_test2.c new file mode 100644 index 00000000..ddfffaa6 --- /dev/null +++ b/src/uri/uri_udp_test2.c @@ -0,0 +1,954 @@ +/* + * Copyright (c) 2016 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include + +#include "../vnet/session/application_interface.h" + +#define vl_typedefs /* define message structures */ +#include +#undef vl_typedefs + +/* declare message handlers for each api */ + +#define vl_endianfun /* define message structures */ +#include +#undef vl_endianfun + +/* instantiate all the print functions we know about */ +#define vl_print(handle, ...) +#define vl_printfun +#include +#undef vl_printfun + +/* Satisfy external references when not linking with -lvlib */ +vlib_main_t vlib_global_main; +vlib_main_t **vlib_mains; + +typedef enum +{ + STATE_START, + STATE_READY, + STATE_DISCONNECTING, +} connection_state_t; + +typedef struct +{ + svm_fifo_t *server_rx_fifo; + svm_fifo_t *server_tx_fifo; +} session_t; + +typedef struct +{ + /* vpe input queue */ + unix_shared_memory_queue_t *vl_input_queue; + + /* API client handle */ + u32 my_client_index; + + /* The URI we're playing with */ + u8 *uri; + + /* Session pool */ + session_t *sessions; + + /* Hash table for disconnect processing */ + uword *session_index_by_vpp_handles; + + /* fifo segment */ + svm_fifo_segment_private_t *seg; + + /* intermediate rx buffer */ + u8 *rx_buf; + + /* URI for connect */ + u8 *connect_uri; + + int i_am_master; + + /* Our event queue */ + unix_shared_memory_queue_t *our_event_queue; + + /* $$$ single thread only for the moment */ + unix_shared_memory_queue_t *vpp_event_queue; + + /* $$$$ hack: cut-through session index */ + volatile u32 cut_through_session_index; + + /* unique segment name counter */ + u32 unique_segment_index; + + pid_t my_pid; + + /* pthread handle */ + pthread_t cut_through_thread_handle; + + /* For deadman timers */ + clib_time_t clib_time; + + /* State of the connection, shared between msg RX thread and main thread */ + volatile connection_state_t state; + + volatile int time_to_stop; + volatile int time_to_print_stats; + + u32 configured_segment_size; + + /* VNET_API_ERROR_FOO -> "Foo" hash table */ + uword *error_string_by_error_number; + + /* convenience */ + svm_fifo_segment_main_t *segment_main; + +} uri_udp_test_main_t; + +#if CLIB_DEBUG > 0 +#define NITER 10000 +#else +#define NITER 4000000 +#endif + +uri_udp_test_main_t uri_udp_test_main; + +static void +stop_signal (int signum) +{ + uri_udp_test_main_t *um = &uri_udp_test_main; + + um->time_to_stop = 1; +} + +static void +stats_signal (int signum) +{ + uri_udp_test_main_t *um = &uri_udp_test_main; + + um->time_to_print_stats = 1; +} + +static clib_error_t * +setup_signal_handlers (void) +{ + signal (SIGINT, stats_signal); + signal (SIGQUIT, stop_signal); + signal (SIGTERM, stop_signal); + + return 0; +} + +u8 * +format_api_error (u8 * s, va_list * args) +{ + uri_udp_test_main_t *utm = va_arg (*args, uri_udp_test_main_t *); + i32 error = va_arg (*args, u32); + uword *p; + + p = hash_get (utm->error_string_by_error_number, -error); + + if (p) + s = format (s, "%s", p[0]); + else + s = format (s, "%d", error); + return s; +} + +int +wait_for_state_change (uri_udp_test_main_t * utm, connection_state_t state) +{ +#if CLIB_DEBUG > 0 +#define TIMEOUT 600.0 +#else +#define TIMEOUT 600.0 +#endif + + f64 timeout = clib_time_now (&utm->clib_time) + TIMEOUT; + + while (clib_time_now (&utm->clib_time) < timeout) + { + if (utm->state == state) + return 0; + } + return -1; +} + +u64 server_bytes_received, server_bytes_sent; + +static void * +cut_through_thread_fn (void *arg) +{ + session_t *s; + svm_fifo_t *rx_fifo; + svm_fifo_t *tx_fifo; + u8 *my_copy_buffer = 0; + uri_udp_test_main_t *utm = &uri_udp_test_main; + i32 actual_transfer; + int rv; + u32 buffer_offset; + + while (utm->cut_through_session_index == ~0) + ; + + s = pool_elt_at_index (utm->sessions, utm->cut_through_session_index); + + rx_fifo = s->server_rx_fifo; + tx_fifo = s->server_tx_fifo; + + vec_validate (my_copy_buffer, 64 * 1024 - 1); + + while (true) + { + /* We read from the tx fifo and write to the rx fifo */ + do + { + actual_transfer = svm_fifo_dequeue_nowait (tx_fifo, 0, + vec_len (my_copy_buffer), + my_copy_buffer); + } + while (actual_transfer <= 0); + + server_bytes_received += actual_transfer; + + buffer_offset = 0; + while (actual_transfer > 0) + { + rv = svm_fifo_enqueue_nowait (rx_fifo, 0, actual_transfer, + my_copy_buffer + buffer_offset); + if (rv > 0) + { + actual_transfer -= rv; + buffer_offset += rv; + server_bytes_sent += rv; + } + + } + if (PREDICT_FALSE (utm->time_to_stop)) + break; + } + + pthread_exit (0); +} + +static void +uri_udp_slave_test (uri_udp_test_main_t * utm) +{ + vl_api_connect_uri_t *cmp; + int i; + u8 *test_data = 0; + u64 bytes_received = 0, bytes_sent = 0; + i32 bytes_to_read; + int rv; + int mypid = getpid (); + f64 before, after, delta, bytes_per_second; + session_t *session; + svm_fifo_t *rx_fifo, *tx_fifo; + int buffer_offset, bytes_to_send = 0; + + vec_validate (test_data, 64 * 1024 - 1); + for (i = 0; i < vec_len (test_data); i++) + test_data[i] = i & 0xff; + + cmp = vl_msg_api_alloc (sizeof (*cmp)); + memset (cmp, 0, sizeof (*cmp)); + + cmp->_vl_msg_id = ntohs (VL_API_CONNECT_URI); + cmp->client_index = utm->my_client_index; + cmp->context = ntohl (0xfeedface); + memcpy (cmp->uri, utm->connect_uri, vec_len (utm->connect_uri)); + vl_msg_api_send_shmem (utm->vl_input_queue, (u8 *) & cmp); + + if (wait_for_state_change (utm, STATE_READY)) + { + clib_warning ("timeout waiting for STATE_READY"); + return; + } + + session = pool_elt_at_index (utm->sessions, utm->cut_through_session_index); + rx_fifo = session->server_rx_fifo; + tx_fifo = session->server_tx_fifo; + + before = clib_time_now (&utm->clib_time); + + vec_validate (utm->rx_buf, vec_len (test_data) - 1); + + for (i = 0; i < NITER; i++) + { + bytes_to_send = vec_len (test_data); + buffer_offset = 0; + while (bytes_to_send > 0) + { + rv = svm_fifo_enqueue_nowait (tx_fifo, mypid, + bytes_to_send, + test_data + buffer_offset); + + if (rv > 0) + { + bytes_to_send -= rv; + buffer_offset += rv; + bytes_sent += rv; + } + } + + bytes_to_read = svm_fifo_max_dequeue (rx_fifo); + + bytes_to_read = vec_len (utm->rx_buf) > bytes_to_read ? + bytes_to_read : vec_len (utm->rx_buf); + + buffer_offset = 0; + while (bytes_to_read > 0) + { + rv = svm_fifo_dequeue_nowait (rx_fifo, mypid, + bytes_to_read, + utm->rx_buf + buffer_offset); + if (rv > 0) + { + bytes_to_read -= rv; + buffer_offset += rv; + bytes_received += rv; + } + } + } + while (bytes_received < bytes_sent) + { + rv = svm_fifo_dequeue_nowait (rx_fifo, mypid, + vec_len (utm->rx_buf), utm->rx_buf); + if (rv > 0) + { +#if CLIB_DEBUG > 0 + int j; + for (j = 0; j < rv; j++) + { + if (utm->rx_buf[j] != ((bytes_received + j) & 0xff)) + { + clib_warning ("error at byte %lld, 0x%x not 0x%x", + bytes_received + j, + utm->rx_buf[j], + ((bytes_received + j) & 0xff)); + } + } +#endif + bytes_received += (u64) rv; + } + } + + after = clib_time_now (&utm->clib_time); + delta = after - before; + bytes_per_second = 0.0; + + if (delta > 0.0) + bytes_per_second = (f64) bytes_received / delta; + + fformat (stdout, + "Done: %lld recv bytes in %.2f seconds, %.2f bytes/sec...\n\n", + bytes_received, delta, bytes_per_second); + fformat (stdout, + "Done: %lld sent bytes in %.2f seconds, %.2f bytes/sec...\n\n", + bytes_sent, delta, bytes_per_second); + fformat (stdout, + "client -> server -> client round trip: %.2f Gbit/sec \n\n", + (bytes_per_second * 8.0) / 1e9); +} + +static void +vl_api_bind_uri_reply_t_handler (vl_api_bind_uri_reply_t * mp) +{ + uri_udp_test_main_t *utm = &uri_udp_test_main; + svm_fifo_segment_create_args_t _a, *a = &_a; + int rv; + + if (mp->segment_name_length == 0) + { + clib_warning ("segment_name_length zero"); + return; + } + + a->segment_name = (char *) mp->segment_name; + a->segment_size = mp->segment_size; + + ASSERT (mp->server_event_queue_address); + + /* Attach to the segment vpp created */ + rv = svm_fifo_segment_attach (a); + if (rv) + { + clib_warning ("svm_fifo_segment_attach ('%s') failed", + mp->segment_name); + return; + } + + utm->our_event_queue = (unix_shared_memory_queue_t *) + mp->server_event_queue_address; + + utm->state = STATE_READY; +} + +static void +vl_api_map_another_segment_t_handler (vl_api_map_another_segment_t * mp) +{ + svm_fifo_segment_create_args_t _a, *a = &_a; + int rv; + + a->segment_name = (char *) mp->segment_name; + a->segment_size = mp->segment_size; + /* Attach to the segment vpp created */ + rv = svm_fifo_segment_attach (a); + if (rv) + { + clib_warning ("svm_fifo_segment_attach ('%s') failed", + mp->segment_name); + return; + } + clib_warning ("Mapped new segment '%s' size %d", mp->segment_name, + mp->segment_size); +} + +static void +vl_api_connect_uri_t_handler (vl_api_connect_uri_t * mp) +{ + u32 segment_index; + uri_udp_test_main_t *utm = &uri_udp_test_main; + svm_fifo_segment_main_t *sm = &svm_fifo_segment_main; + svm_fifo_segment_create_args_t _a, *a = &_a; + svm_fifo_segment_private_t *seg; + unix_shared_memory_queue_t *client_q; + vl_api_connect_uri_reply_t *rmp; + session_t *session; + int rv = 0; + + /* Create the segment */ + a->segment_name = (char *) format (0, "%d:segment%d%c", utm->my_pid, + utm->unique_segment_index++, 0); + a->segment_size = utm->configured_segment_size; + + rv = svm_fifo_segment_create (a); + if (rv) + { + clib_warning ("sm_fifo_segment_create ('%s') failed", a->segment_name); + rv = VNET_API_ERROR_URI_FIFO_CREATE_FAILED; + goto send_reply; + } + + vec_add2 (utm->seg, seg, 1); + + segment_index = vec_len (sm->segments) - 1; + + memcpy (seg, sm->segments + segment_index, sizeof (utm->seg[0])); + + pool_get (utm->sessions, session); + + /* + * By construction the master's idea of the rx fifo ends up in + * fsh->fifos[0], and the master's idea of the tx fifo ends up in + * fsh->fifos[1]. + */ + session->server_rx_fifo = svm_fifo_segment_alloc_fifo (utm->seg, + 128 * 1024); + ASSERT (session->server_rx_fifo); + + session->server_tx_fifo = svm_fifo_segment_alloc_fifo (utm->seg, + 128 * 1024); + ASSERT (session->server_tx_fifo); + + session->server_rx_fifo->server_session_index = session - utm->sessions; + session->server_tx_fifo->server_session_index = session - utm->sessions; + utm->cut_through_session_index = session - utm->sessions; + + rv = pthread_create (&utm->cut_through_thread_handle, + NULL /*attr */ , cut_through_thread_fn, 0); + if (rv) + { + clib_warning ("pthread_create returned %d", rv); + rv = VNET_API_ERROR_SYSCALL_ERROR_1; + } + +send_reply: + rmp = vl_msg_api_alloc (sizeof (*rmp)); + memset (rmp, 0, sizeof (*rmp)); + + rmp->_vl_msg_id = ntohs (VL_API_CONNECT_URI_REPLY); + rmp->context = mp->context; + rmp->retval = ntohl (rv); + rmp->segment_name_length = vec_len (a->segment_name); + memcpy (rmp->segment_name, a->segment_name, vec_len (a->segment_name)); + + vec_free (a->segment_name); + + client_q = (unix_shared_memory_queue_t *) mp->client_queue_address; + vl_msg_api_send_shmem (client_q, (u8 *) & rmp); +} + +static void +vl_api_unbind_uri_reply_t_handler (vl_api_unbind_uri_reply_t * mp) +{ + uri_udp_test_main_t *utm = &uri_udp_test_main; + + if (mp->retval != 0) + clib_warning ("returned %d", ntohl (mp->retval)); + + utm->state = STATE_START; +} + +static void +vl_api_accept_session_t_handler (vl_api_accept_session_t * mp) +{ + uri_udp_test_main_t *utm = &uri_udp_test_main; + vl_api_accept_session_reply_t *rmp; + svm_fifo_t *rx_fifo, *tx_fifo; + session_t *session; + static f64 start_time; + u64 key; + + if (start_time == 0.0) + start_time = clib_time_now (&utm->clib_time); + + utm->vpp_event_queue = (unix_shared_memory_queue_t *) + mp->vpp_event_queue_address; + + pool_get (utm->sessions, session); + + rx_fifo = (svm_fifo_t *) mp->server_rx_fifo; + rx_fifo->client_session_index = session - utm->sessions; + tx_fifo = (svm_fifo_t *) mp->server_tx_fifo; + tx_fifo->client_session_index = session - utm->sessions; + + session->server_rx_fifo = rx_fifo; + session->server_tx_fifo = tx_fifo; + + key = (((u64) mp->session_thread_index) << 32) | (u64) mp->session_index; + + hash_set (utm->session_index_by_vpp_handles, key, session - utm->sessions); + + utm->state = STATE_READY; + + if (pool_elts (utm->sessions) && (pool_elts (utm->sessions) % 20000) == 0) + { + f64 now = clib_time_now (&utm->clib_time); + fformat (stdout, "%d active sessions in %.2f seconds, %.2f/sec...\n", + pool_elts (utm->sessions), now - start_time, + (f64) pool_elts (utm->sessions) / (now - start_time)); + } + + rmp = vl_msg_api_alloc (sizeof (*rmp)); + memset (rmp, 0, sizeof (*rmp)); + rmp->_vl_msg_id = ntohs (VL_API_ACCEPT_SESSION_REPLY); + rmp->session_type = mp->session_type; + rmp->session_index = mp->session_index; + rmp->session_thread_index = mp->session_thread_index; + vl_msg_api_send_shmem (utm->vl_input_queue, (u8 *) & rmp); +} + +static void +vl_api_disconnect_session_t_handler (vl_api_disconnect_session_t * mp) +{ + uri_udp_test_main_t *utm = &uri_udp_test_main; + session_t *session; + vl_api_disconnect_session_reply_t *rmp; + uword *p; + int rv = 0; + u64 key; + + key = (((u64) mp->session_thread_index) << 32) | (u64) mp->session_index; + + p = hash_get (utm->session_index_by_vpp_handles, key); + + if (p) + { + session = pool_elt_at_index (utm->sessions, p[0]); + hash_unset (utm->session_index_by_vpp_handles, key); + pool_put (utm->sessions, session); + } + else + { + clib_warning ("couldn't find session key %llx", key); + rv = -11; + } + + rmp = vl_msg_api_alloc (sizeof (*rmp)); + memset (rmp, 0, sizeof (*rmp)); + rmp->_vl_msg_id = ntohs (VL_API_DISCONNECT_SESSION_REPLY); + rmp->retval = rv; + rmp->session_index = mp->session_index; + rmp->session_thread_index = mp->session_thread_index; + vl_msg_api_send_shmem (utm->vl_input_queue, (u8 *) & rmp); +} + +static void +vl_api_connect_uri_reply_t_handler (vl_api_connect_uri_reply_t * mp) +{ + svm_fifo_segment_main_t *sm = &svm_fifo_segment_main; + uri_udp_test_main_t *utm = &uri_udp_test_main; + svm_fifo_segment_create_args_t _a, *a = &_a; + ssvm_shared_header_t *sh; + svm_fifo_segment_private_t *seg; + svm_fifo_segment_header_t *fsh; + session_t *session; + u32 segment_index; + int rv; + + ASSERT (utm->i_am_master == 0); + + if (mp->segment_name_length == 0) + { + clib_warning ("segment_name_length zero"); + return; + } + + memset (a, 0, sizeof (*a)); + + a->segment_name = (char *) mp->segment_name; + + sleep (1); + + rv = svm_fifo_segment_attach (a); + if (rv) + { + clib_warning ("sm_fifo_segment_create ('%v') failed", mp->segment_name); + return; + } + + segment_index = vec_len (sm->segments) - 1; + + vec_add2 (utm->seg, seg, 1); + + memcpy (seg, sm->segments + segment_index, sizeof (*seg)); + sh = seg->ssvm.sh; + fsh = (svm_fifo_segment_header_t *) sh->opaque[0]; + + while (vec_len (fsh->fifos) < 2) + sleep (1); + + pool_get (utm->sessions, session); + utm->cut_through_session_index = session - utm->sessions; + + session->server_rx_fifo = (svm_fifo_t *) fsh->fifos[0]; + ASSERT (session->server_rx_fifo); + session->server_tx_fifo = (svm_fifo_t *) fsh->fifos[1]; + ASSERT (session->server_tx_fifo); + + /* security: could unlink /dev/shm/segment_name> here, maybe */ + + utm->state = STATE_READY; +} + +#define foreach_uri_msg \ +_(BIND_URI_REPLY, bind_uri_reply) \ +_(CONNECT_URI, connect_uri) \ +_(CONNECT_URI_REPLY, connect_uri_reply) \ +_(UNBIND_URI_REPLY, unbind_uri_reply) \ +_(ACCEPT_SESSION, accept_session) \ +_(DISCONNECT_SESSION, disconnect_session) \ +_(MAP_ANOTHER_SEGMENT, map_another_segment) + +void +uri_api_hookup (uri_udp_test_main_t * utm) +{ +#define _(N,n) \ + vl_msg_api_set_handlers(VL_API_##N, #n, \ + vl_api_##n##_t_handler, \ + vl_noop_handler, \ + vl_api_##n##_t_endian, \ + vl_api_##n##_t_print, \ + sizeof(vl_api_##n##_t), 1); + foreach_uri_msg; +#undef _ + +} + + +int +connect_to_vpp (char *name) +{ + uri_udp_test_main_t *utm = &uri_udp_test_main; + api_main_t *am = &api_main; + + if (vl_client_connect_to_vlib ("/vpe-api", name, 32) < 0) + return -1; + + utm->vl_input_queue = am->shmem_hdr->vl_input_queue; + utm->my_client_index = am->my_client_index; + + return 0; +} + +void +vlib_cli_output (struct vlib_main_t *vm, char *fmt, ...) +{ + clib_warning ("BUG"); +} + +static void +init_error_string_table (uri_udp_test_main_t * utm) +{ + utm->error_string_by_error_number = hash_create (0, sizeof (uword)); + +#define _(n,v,s) hash_set (utm->error_string_by_error_number, -v, s); + foreach_vnet_api_error; +#undef _ + + hash_set (utm->error_string_by_error_number, 99, "Misc"); +} + +void +handle_fifo_event_server_rx (uri_udp_test_main_t * utm, + session_fifo_event_t * e) +{ + svm_fifo_t *rx_fifo, *tx_fifo; + int nbytes; + + session_fifo_event_t evt; + unix_shared_memory_queue_t *q; + int rv; + + rx_fifo = e->fifo; + tx_fifo = utm->sessions[rx_fifo->client_session_index].server_tx_fifo; + + do + { + nbytes = svm_fifo_dequeue_nowait (rx_fifo, 0, + vec_len (utm->rx_buf), utm->rx_buf); + } + while (nbytes <= 0); + do + { + rv = svm_fifo_enqueue_nowait (tx_fifo, 0, nbytes, utm->rx_buf); + } + while (rv == -2); + + /* Fabricate TX event, send to vpp */ + evt.fifo = tx_fifo; + evt.event_type = FIFO_EVENT_SERVER_TX; + /* $$$$ for event logging */ + evt.enqueue_length = nbytes; + evt.event_id = e->event_id; + q = utm->vpp_event_queue; + unix_shared_memory_queue_add (q, (u8 *) & evt, 0 /* do wait for mutex */ ); +} + +void +handle_event_queue (uri_udp_test_main_t * utm) +{ + session_fifo_event_t _e, *e = &_e;; + + while (1) + { + unix_shared_memory_queue_sub (utm->our_event_queue, (u8 *) e, + 0 /* nowait */ ); + switch (e->event_type) + { + case FIFO_EVENT_SERVER_RX: + handle_fifo_event_server_rx (utm, e); + break; + + case FIFO_EVENT_SERVER_EXIT: + return; + + default: + clib_warning ("unknown event type %d", e->event_type); + break; + } + if (PREDICT_FALSE (utm->time_to_stop == 1)) + break; + if (PREDICT_FALSE (utm->time_to_print_stats == 1)) + { + utm->time_to_print_stats = 0; + fformat (stdout, "%d connections\n", pool_elts (utm->sessions)); + } + } +} + +void +uri_udp_test (uri_udp_test_main_t * utm) +{ + vl_api_bind_uri_t *bmp; + vl_api_unbind_uri_t *ump; + + bmp = vl_msg_api_alloc (sizeof (*bmp)); + memset (bmp, 0, sizeof (*bmp)); + + bmp->_vl_msg_id = ntohs (VL_API_BIND_URI); + bmp->client_index = utm->my_client_index; + bmp->context = ntohl (0xfeedface); + bmp->initial_segment_size = 256 << 20; /* size of initial segment */ + bmp->options[SESSION_OPTIONS_FLAGS] = + SESSION_OPTIONS_FLAGS_USE_FIFO | SESSION_OPTIONS_FLAGS_ADD_SEGMENT; + bmp->options[SESSION_OPTIONS_RX_FIFO_SIZE] = 16 << 10; + bmp->options[SESSION_OPTIONS_TX_FIFO_SIZE] = 16 << 10; + bmp->options[SESSION_OPTIONS_ADD_SEGMENT_SIZE] = 128 << 20; + memcpy (bmp->uri, utm->uri, vec_len (utm->uri)); + vl_msg_api_send_shmem (utm->vl_input_queue, (u8 *) & bmp); + + if (wait_for_state_change (utm, STATE_READY)) + { + clib_warning ("timeout waiting for STATE_READY"); + return; + } + + handle_event_queue (utm); + + ump = vl_msg_api_alloc (sizeof (*ump)); + memset (ump, 0, sizeof (*ump)); + + ump->_vl_msg_id = ntohs (VL_API_UNBIND_URI); + ump->client_index = utm->my_client_index; + memcpy (ump->uri, utm->uri, vec_len (utm->uri)); + vl_msg_api_send_shmem (utm->vl_input_queue, (u8 *) & ump); + + if (wait_for_state_change (utm, STATE_START)) + { + clib_warning ("timeout waiting for STATE_START"); + return; + } + + fformat (stdout, "Test complete...\n"); +} + +int +main (int argc, char **argv) +{ + uri_udp_test_main_t *utm = &uri_udp_test_main; + unformat_input_t _argv, *a = &_argv; + u8 *chroot_prefix; + u8 *heap; + u8 *bind_name = (u8 *) "udp://0.0.0.0/1234"; + u32 tmp; + mheap_t *h; + session_t *session; + int i; + int i_am_master = 1; + + clib_mem_init (0, 256 << 20); + + heap = clib_mem_get_per_cpu_heap (); + h = mheap_header (heap); + + /* make the main heap thread-safe */ + h->flags |= MHEAP_FLAG_THREAD_SAFE; + + vec_validate (utm->rx_buf, 8192); + + utm->session_index_by_vpp_handles = hash_create (0, sizeof (uword)); + + utm->my_pid = getpid (); + utm->configured_segment_size = 1 << 20; + + clib_time_init (&utm->clib_time); + init_error_string_table (utm); + svm_fifo_segment_init (0x200000000ULL, 20); + unformat_init_command_line (a, argv); + + while (unformat_check_input (a) != UNFORMAT_END_OF_INPUT) + { + if (unformat (a, "chroot prefix %s", &chroot_prefix)) + { + vl_set_memory_root_path ((char *) chroot_prefix); + } + else if (unformat (a, "uri %s", &bind_name)) + ; + else if (unformat (a, "segment-size %dM", &tmp)) + utm->configured_segment_size = tmp << 20; + else if (unformat (a, "segment-size %dG", &tmp)) + utm->configured_segment_size = tmp << 30; + else if (unformat (a, "master")) + i_am_master = 1; + else if (unformat (a, "slave")) + i_am_master = 0; + else + { + fformat (stderr, "%s: usage [master|slave]\n"); + exit (1); + } + } + + utm->cut_through_session_index = ~0; + utm->uri = format (0, "%s%c", bind_name, 0); + utm->i_am_master = i_am_master; + utm->segment_main = &svm_fifo_segment_main; + + utm->connect_uri = format (0, "udp://10.0.0.1/1234%c", 0); + + setup_signal_handlers (); + + uri_api_hookup (utm); + + if (connect_to_vpp (i_am_master ? "uri_udp_master" : "uri_udp_slave") < 0) + { + svm_region_exit (); + fformat (stderr, "Couldn't connect to vpe, exiting...\n"); + exit (1); + } + + if (i_am_master == 0) + { + uri_udp_slave_test (utm); + exit (0); + } + + /* $$$$ hack preallocation */ + for (i = 0; i < 200000; i++) + { + pool_get (utm->sessions, session); + memset (session, 0, sizeof (*session)); + } + for (i = 0; i < 200000; i++) + pool_put_index (utm->sessions, i); + + uri_udp_test (utm); + + vl_client_disconnect_from_vlib (); + exit (0); +} + +#undef vl_api_version +#define vl_api_version(n,v) static u32 vpe_api_version = v; +#include +#undef vl_api_version + +void +vl_client_add_api_signatures (vl_api_memclnt_create_t * mp) +{ + /* + * Send the main API signature in slot 0. This bit of code must + * match the checks in ../vpe/api/api.c: vl_msg_api_version_check(). + */ + mp->api_versions[0] = clib_host_to_net_u32 (vpe_api_version); +} + +u32 +vl (void *p) +{ + return vec_len (p); +} + +/* + * fd.io coding-style-patch-verification: ON + * + * Local Variables: + * eval: (c-set-style "gnu") + * End: + */ diff --git a/src/uri/uritest.c b/src/uri/uritest.c new file mode 100644 index 00000000..edcdb3ad --- /dev/null +++ b/src/uri/uritest.c @@ -0,0 +1,484 @@ +/* + * Copyright (c) 2016 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include + +#define vl_typedefs /* define message structures */ +#include +#undef vl_typedefs + +/* declare message handlers for each api */ + +#define vl_endianfun /* define message structures */ +#include +#undef vl_endianfun + +/* instantiate all the print functions we know about */ +#define vl_print(handle, ...) +#define vl_printfun +#include +#undef vl_printfun + +typedef enum +{ + STATE_START, + STATE_READY, + STATE_DISCONNECTING, +} connection_state_t; + +typedef struct +{ + /* vpe input queue */ + unix_shared_memory_queue_t *vl_input_queue; + + /* API client handle */ + u32 my_client_index; + + /* role */ + int i_am_master; + + /* The URI we're playing with */ + u8 *uri; + + /* fifo segment */ + svm_fifo_segment_private_t *seg; + + svm_fifo_t *rx_fifo; + svm_fifo_t *tx_fifo; + + /* For deadman timers */ + clib_time_t clib_time; + + /* State of the connection, shared between msg RX thread and main thread */ + volatile connection_state_t state; + + /* VNET_API_ERROR_FOO -> "Foo" hash table */ + uword *error_string_by_error_number; +} uritest_main_t; + +#if CLIB_DEBUG > 0 +#define NITER 1000 +#else +#define NITER 1000000 +#endif + +uritest_main_t uritest_main; + +u8 * +format_api_error (u8 * s, va_list * args) +{ + uritest_main_t *utm = va_arg (*args, uritest_main_t *); + i32 error = va_arg (*args, u32); + uword *p; + + p = hash_get (utm->error_string_by_error_number, -error); + + if (p) + s = format (s, "%s", p[0]); + else + s = format (s, "%d", error); + return s; +} + +int +wait_for_state_change (uritest_main_t * utm, connection_state_t state) +{ + f64 timeout = clib_time_now (&utm->clib_time) + 1.0; + + while (clib_time_now (&utm->clib_time) < timeout) + { + if (utm->state == state) + return 0; + } + return -1; +} + +static void +vl_api_bind_uri_reply_t_handler (vl_api_bind_uri_reply_t * mp) +{ + uritest_main_t *utm = &uritest_main; + svm_fifo_segment_create_args_t _a, *a = &_a; + int rv; + + ASSERT (utm->i_am_master); + + if (mp->segment_name_length == 0) + { + clib_warning ("segment_name_length zero"); + return; + } + + a->segment_name = (char *) mp->segment_name; + a->segment_size = mp->segment_size; + + /* Create the segment */ + rv = svm_fifo_segment_create (a); + if (rv) + { + clib_warning ("sm_fifo_segment_create ('%s') failed", mp->segment_name); + return; + } + + vec_validate (utm->seg, 0); + + memcpy (utm->seg, a->rv, sizeof (*utm->seg)); + + /* + * By construction the master's idea of the rx fifo ends up in + * fsh->fifos[0], and the master's idea of the tx fifo ends up in + * fsh->fifos[1]. + */ + utm->rx_fifo = svm_fifo_segment_alloc_fifo (utm->seg, 10240); + ASSERT (utm->rx_fifo); + + utm->tx_fifo = svm_fifo_segment_alloc_fifo (utm->seg, 10240); + ASSERT (utm->tx_fifo); + + utm->state = STATE_READY; +} + +static void +vl_api_connect_uri_reply_t_handler (vl_api_connect_uri_reply_t * mp) +{ + uritest_main_t *utm = &uritest_main; + svm_fifo_segment_create_args_t _a, *a = &_a; + ssvm_shared_header_t *sh; + svm_fifo_segment_header_t *fsh; + int rv; + + ASSERT (utm->i_am_master == 0); + + if (mp->segment_name_length == 0) + { + clib_warning ("segment_name_length zero"); + return; + } + + memset (a, 0, sizeof (*a)); + + a->segment_name = (char *) mp->segment_name; + + rv = svm_fifo_segment_attach (a); + if (rv) + { + clib_warning ("sm_fifo_segment_create ('%s') failed", mp->segment_name); + return; + } + + vec_validate (utm->seg, 0); + + memcpy (utm->seg, a->rv, sizeof (*utm->seg)); + sh = utm->seg->ssvm.sh; + fsh = (svm_fifo_segment_header_t *) sh->opaque[0]; + + while (vec_len (fsh->fifos) < 2) + sleep (1); + + utm->rx_fifo = (svm_fifo_t *) fsh->fifos[1]; + ASSERT (utm->rx_fifo); + utm->tx_fifo = (svm_fifo_t *) fsh->fifos[0]; + ASSERT (utm->tx_fifo); + + /* security: could unlink /dev/shm/segment_name> here, maybe */ + + utm->state = STATE_READY; +} + +static void +vl_api_unbind_uri_reply_t_handler (vl_api_unbind_uri_reply_t * mp) +{ + uritest_main_t *utm = &uritest_main; + + if (mp->retval != 0) + clib_warning ("returned %d", ntohl (mp->retval)); + + utm->state = STATE_START; +} + +#define foreach_uri_msg \ +_(BIND_URI_REPLY, bind_uri_reply) \ +_(CONNECT_URI_REPLY, connect_uri_reply) \ +_(UNBIND_URI_REPLY, unbind_uri_reply) + +void +uri_api_hookup (uritest_main_t * utm) +{ +#define _(N,n) \ + vl_msg_api_set_handlers(VL_API_##N, #n, \ + vl_api_##n##_t_handler, \ + vl_noop_handler, \ + vl_api_##n##_t_endian, \ + vl_api_##n##_t_print, \ + sizeof(vl_api_##n##_t), 1); + foreach_uri_msg; +#undef _ + +} + + +int +connect_to_vpp (char *name) +{ + uritest_main_t *utm = &uritest_main; + api_main_t *am = &api_main; + + if (vl_client_connect_to_vlib ("/vpe-api", name, 32) < 0) + return -1; + + utm->vl_input_queue = am->shmem_hdr->vl_input_queue; + utm->my_client_index = am->my_client_index; + + return 0; +} + +void +vlib_cli_output (struct vlib_main_t *vm, char *fmt, ...) +{ + clib_warning ("BUG"); +} + +static void +init_error_string_table (uritest_main_t * utm) +{ + utm->error_string_by_error_number = hash_create (0, sizeof (uword)); + +#define _(n,v,s) hash_set (utm->error_string_by_error_number, -v, s); + foreach_vnet_api_error; +#undef _ + + hash_set (utm->error_string_by_error_number, 99, "Misc"); +} + +void +uritest_master (uritest_main_t * utm) +{ + vl_api_bind_uri_t *bmp; + vl_api_unbind_uri_t *ump; + int i; + u8 *test_data = 0; + u8 *reply = 0; + u32 reply_len; + int mypid = getpid (); + + for (i = 0; i < 2048; i++) + vec_add1 (test_data, 'a' + (i % 32)); + + bmp = vl_msg_api_alloc (sizeof (*bmp)); + memset (bmp, 0, sizeof (*bmp)); + + bmp->_vl_msg_id = ntohs (VL_API_BIND_URI); + bmp->client_index = utm->my_client_index; + bmp->context = ntohl (0xfeedface); + bmp->segment_size = 256 << 10; + memcpy (bmp->uri, utm->uri, vec_len (utm->uri)); + vl_msg_api_send_shmem (utm->vl_input_queue, (u8 *) & bmp); + + if (wait_for_state_change (utm, STATE_READY)) + { + clib_warning ("timeout waiting for STATE_READY"); + return; + } + + for (i = 0; i < NITER; i++) + svm_fifo_enqueue (utm->tx_fifo, mypid, vec_len (test_data), test_data); + + vec_validate (reply, 0); + + reply_len = svm_fifo_dequeue (utm->rx_fifo, mypid, vec_len (reply), reply); + + if (reply_len != 1) + clib_warning ("reply length %d", reply_len); + + if (reply[0] == 1) + fformat (stdout, "Test OK..."); + + ump = vl_msg_api_alloc (sizeof (*ump)); + memset (ump, 0, sizeof (*ump)); + + ump->_vl_msg_id = ntohs (VL_API_UNBIND_URI); + ump->client_index = utm->my_client_index; + memcpy (ump->uri, utm->uri, vec_len (utm->uri)); + vl_msg_api_send_shmem (utm->vl_input_queue, (u8 *) & ump); + + if (wait_for_state_change (utm, STATE_START)) + { + clib_warning ("timeout waiting for STATE_READY"); + return; + } + + fformat (stdout, "Master done...\n"); +} + +void +uritest_slave (uritest_main_t * utm) +{ + vl_api_connect_uri_t *cmp; + int i, j; + u8 *test_data = 0; + u8 *reply = 0; + u32 bytes_received = 0; + u32 actual_bytes; + int mypid = getpid (); + u8 ok; + f64 before, after, delta, bytes_per_second; + + vec_validate (test_data, 4095); + + cmp = vl_msg_api_alloc (sizeof (*cmp)); + memset (cmp, 0, sizeof (*cmp)); + + cmp->_vl_msg_id = ntohs (VL_API_CONNECT_URI); + cmp->client_index = utm->my_client_index; + cmp->context = ntohl (0xfeedface); + memcpy (cmp->uri, utm->uri, vec_len (utm->uri)); + vl_msg_api_send_shmem (utm->vl_input_queue, (u8 *) & cmp); + + if (wait_for_state_change (utm, STATE_READY)) + { + clib_warning ("timeout waiting for STATE_READY"); + return; + } + + ok = 1; + before = clib_time_now (&utm->clib_time); + for (i = 0; i < NITER; i++) + { + actual_bytes = svm_fifo_dequeue (utm->rx_fifo, mypid, + vec_len (test_data), test_data); + j = 0; + while (j < actual_bytes) + { + if (test_data[j] != ('a' + (bytes_received % 32))) + ok = 0; + bytes_received++; + j++; + } + if (bytes_received == NITER * 2048) + break; + } + + vec_add1 (reply, ok); + + svm_fifo_enqueue (utm->tx_fifo, mypid, vec_len (reply), reply); + after = clib_time_now (&utm->clib_time); + delta = after - before; + bytes_per_second = 0.0; + + if (delta > 0.0) + bytes_per_second = (f64) bytes_received / delta; + + fformat (stdout, + "Slave done, %d bytes in %.2f seconds, %.2f bytes/sec...\n", + bytes_received, delta, bytes_per_second); +} + +int +main (int argc, char **argv) +{ + uritest_main_t *utm = &uritest_main; + unformat_input_t _argv, *a = &_argv; + u8 *chroot_prefix; + u8 *heap; + char *bind_name = "fifo:uritest"; + mheap_t *h; + int i_am_master = 0; + + clib_mem_init (0, 128 << 20); + + heap = clib_mem_get_per_cpu_heap (); + h = mheap_header (heap); + + /* make the main heap thread-safe */ + h->flags |= MHEAP_FLAG_THREAD_SAFE; + + clib_time_init (&utm->clib_time); + init_error_string_table (utm); + svm_fifo_segment_init (0x200000000ULL, 20); + unformat_init_command_line (a, argv); + + utm->uri = format (0, "%s%c", bind_name, 0); + + while (unformat_check_input (a) != UNFORMAT_END_OF_INPUT) + { + if (unformat (a, "master")) + i_am_master = 1; + else if (unformat (a, "slave")) + i_am_master = 0; + else if (unformat (a, "chroot prefix %s", &chroot_prefix)) + { + vl_set_memory_root_path ((char *) chroot_prefix); + } + else + { + fformat (stderr, "%s: usage [master|slave]\n"); + exit (1); + } + } + + uri_api_hookup (utm); + + if (connect_to_vpp (i_am_master ? "uritest_master" : "uritest_slave") < 0) + { + svm_region_exit (); + fformat (stderr, "Couldn't connect to vpe, exiting...\n"); + exit (1); + } + + utm->i_am_master = i_am_master; + + if (i_am_master) + uritest_master (utm); + else + uritest_slave (utm); + + vl_client_disconnect_from_vlib (); + exit (0); +} + +#undef vl_api_version +#define vl_api_version(n,v) static u32 vpe_api_version = v; +#include +#undef vl_api_version + +void +vl_client_add_api_signatures (vl_api_memclnt_create_t * mp) +{ + /* + * Send the main API signature in slot 0. This bit of code must + * match the checks in ../vpe/api/api.c: vl_msg_api_version_check(). + */ + mp->api_versions[0] = clib_host_to_net_u32 (vpe_api_version); +} + +/* + * fd.io coding-style-patch-verification: ON + * + * Local Variables: + * eval: (c-set-style "gnu") + * End: + */ diff --git a/src/vlib/buffer.c b/src/vlib/buffer.c index 4f5eb09d..9f26bec7 100644 --- a/src/vlib/buffer.c +++ b/src/vlib/buffer.c @@ -360,7 +360,7 @@ vlib_buffer_create_free_list_helper (vlib_main_t * vm, memset (f, 0, sizeof (f[0])); f->index = f - bm->buffer_free_list_pool; f->n_data_bytes = vlib_buffer_round_size (n_data_bytes); - f->min_n_buffers_each_physmem_alloc = 16; + f->min_n_buffers_each_physmem_alloc = VLIB_FRAME_SIZE; f->name = clib_mem_is_heap_object (name) ? name : format (0, "%s", name); /* Setup free buffer template. */ diff --git a/src/vlib/buffer.h b/src/vlib/buffer.h index 1f723f3b..69c8c7cc 100644 --- a/src/vlib/buffer.h +++ b/src/vlib/buffer.h @@ -240,6 +240,74 @@ vlib_get_buffer_opaque2 (vlib_buffer_t * b) return (void *) b->opaque2; } +/** \brief Get pointer to the end of buffer's data + * @param b pointer to the buffer + * @return pointer to tail of packet's data + */ +always_inline u8 * +vlib_buffer_get_tail (vlib_buffer_t * b) +{ + return b->data + b->current_data + b->current_length; +} + +/** \brief Append uninitialized data to buffer + * @param b pointer to the buffer + * @param size number of uninitialized bytes + * @return pointer to beginning of uninitialized data + */ +always_inline void * +vlib_buffer_put_uninit (vlib_buffer_t * b, u8 size) +{ + void *p = vlib_buffer_get_tail (b); + /* XXX make sure there's enough space */ + b->current_length += size; + return p; +} + +/** \brief Prepend uninitialized data to buffer + * @param b pointer to the buffer + * @param size number of uninitialized bytes + * @return pointer to beginning of uninitialized data + */ +always_inline void * +vlib_buffer_push_uninit (vlib_buffer_t * b, u8 size) +{ + ASSERT (b->current_data + VLIB_BUFFER_PRE_DATA_SIZE >= size); + b->current_data -= size; + b->current_length += size; + + return vlib_buffer_get_current (b); +} + +/** \brief Make head room, typically for packet headers + * @param b pointer to the buffer + * @param size number of head room bytes + * @return pointer to start of buffer (current data) + */ +always_inline void * +vlib_buffer_make_headroom (vlib_buffer_t * b, u8 size) +{ + ASSERT (b->current_data + VLIB_BUFFER_PRE_DATA_SIZE >= size); + b->current_data += size; + return vlib_buffer_get_current (b); +} + +/** \brief Retrieve bytes from buffer head + * @param b pointer to the buffer + * @param size number of bytes to pull + * @return pointer to start of buffer (current data) + */ +always_inline void * +vlib_buffer_pull (vlib_buffer_t * b, u8 size) +{ + if (b->current_length + VLIB_BUFFER_PRE_DATA_SIZE < size) + return 0; + + void *data = vlib_buffer_get_current (b); + vlib_buffer_advance (b, size); + return data; +} + /* Forward declaration. */ struct vlib_main_t; diff --git a/src/vlibmemory/unix_shared_memory_queue.c b/src/vlibmemory/unix_shared_memory_queue.c index 25d28910..e86edec3 100644 --- a/src/vlibmemory/unix_shared_memory_queue.c +++ b/src/vlibmemory/unix_shared_memory_queue.c @@ -33,18 +33,13 @@ * nels = number of elements on the queue * elsize = element size, presumably 4 and cacheline-size will * be popular choices. - * coid = consumer coid, from ChannelCreate * pid = consumer pid - * pulse_code = pulse code consumer expects - * pulse_value = pulse value consumer expects - * consumer_prio = consumer's priority, so pulses won't change - * the consumer's priority. * * The idea is to call this function in the queue consumer, * and e-mail the queue pointer to the producer(s). * - * The spp process / main thread allocates one of these - * at startup; its main input queue. The spp main input queue + * The vpp process / main thread allocates one of these + * at startup; its main input queue. The vpp main input queue * has a pointer to it in the shared memory segment header. * * You probably want to be on an svm data heap before calling this @@ -70,7 +65,7 @@ unix_shared_memory_queue_init (int nels, q->signal_when_queue_non_empty = signal_when_queue_non_empty; memset (&attr, 0, sizeof (attr)); - memset (&cattr, 0, sizeof (attr)); + memset (&cattr, 0, sizeof (cattr)); if (pthread_mutexattr_init (&attr)) clib_unix_warning ("mutexattr_init"); @@ -277,6 +272,7 @@ unix_shared_memory_queue_sub (unix_shared_memory_queue_t * q, clib_memcpy (elem, headp, q->elsize); q->head++; + /* $$$$ JFC shouldn't this be == 0? */ if (q->cursize == q->maxsize) need_broadcast = 1; diff --git a/src/vlibmemory/unix_shared_memory_queue.h b/src/vlibmemory/unix_shared_memory_queue.h index f758f17c..13800065 100644 --- a/src/vlibmemory/unix_shared_memory_queue.h +++ b/src/vlibmemory/unix_shared_memory_queue.h @@ -29,7 +29,7 @@ typedef struct _unix_shared_memory_queue pthread_cond_t condvar; /* 8 bytes */ int head; int tail; - int cursize; + volatile int cursize; int maxsize; int elsize; int consumer_pid; diff --git a/src/vnet.am b/src/vnet.am index 64484e18..923f61d8 100644 --- a/src/vnet.am +++ b/src/vnet.am @@ -324,11 +324,7 @@ libvnet_la_SOURCES += \ vnet/ip/ip_input_acl.c \ vnet/ip/lookup.c \ vnet/ip/ping.c \ - vnet/ip/punt.c \ - vnet/ip/udp_format.c \ - vnet/ip/udp_init.c \ - vnet/ip/udp_local.c \ - vnet/ip/udp_pg.c + vnet/ip/punt.c nobase_include_HEADERS += \ vnet/ip/format.h \ @@ -354,11 +350,7 @@ nobase_include_HEADERS += \ vnet/ip/ports.def \ vnet/ip/protocols.def \ vnet/ip/punt_error.def \ - vnet/ip/punt.h \ - vnet/ip/tcp_packet.h \ - vnet/ip/udp_error.def \ - vnet/ip/udp.h \ - vnet/ip/udp_packet.h + vnet/ip/punt.h API_FILES += vnet/ip/ip.api @@ -473,6 +465,38 @@ test_map_LDADD = libvnet.la libvppinfra.la libvlib.la \ test_map_LDFLAGS = -static endif +######################################## +# Layer 4 protocol: tcp +######################################## +libvnet_la_SOURCES += \ + vnet/tcp/tcp_format.c \ + vnet/tcp/tcp_pg.c \ + vnet/tcp/tcp_syn_filter4.c \ + vnet/tcp/tcp_output.c \ + vnet/tcp/tcp_input.c \ + vnet/tcp/tcp_newreno.c \ + vnet/tcp/tcp.c + +nobase_include_HEADERS += \ + vnet/tcp/tcp_packet.h \ + vnet/tcp/tcp_timer.h \ + vnet/tcp/tcp.h + +######################################## +# Layer 4 protocol: udp +######################################## +libvnet_la_SOURCES += \ + vnet/udp/udp.c \ + vnet/udp/udp_input.c \ + vnet/udp/builtin_server.c \ + vnet/udp/udp_format.c \ + vnet/udp/udp_local.c \ + vnet/udp/udp_pg.c + +nobase_include_HEADERS += \ + vnet/udp/udp_error.def \ + vnet/udp/udp.h \ + vnet/udp/udp_packet.h ######################################## # Tunnel protocol: gre @@ -833,6 +857,28 @@ libvnet_la_SOURCES += \ nobase_include_HEADERS += \ vnet/devices/ssvm/ssvm_eth.h +######################################## +# session managmeent +######################################## + +libvnet_la_SOURCES += \ + vnet/session/session.c \ + vnet/session/node.c \ + vnet/session/transport.c \ + vnet/session/application.c \ + vnet/session/session_cli.c \ + vnet/session/hashes.c \ + vnet/session/application_interface.c \ + vnet/session/session_api.c + +nobase_include_HEADERS += \ + vnet/session/session.h \ + vnet/session/application.h \ + vnet/session/transport.h \ + vnet/session/application_interface.h + +API_FILES += vnet/session/session.api + ######################################## # Linux packet interface ######################################## diff --git a/src/vnet/api_errno.h b/src/vnet/api_errno.h index 8680ef7c..861a5767 100644 --- a/src/vnet/api_errno.h +++ b/src/vnet/api_errno.h @@ -91,14 +91,19 @@ _(INVALID_ADDRESS_FAMILY, -97, "Invalid address family") \ _(INVALID_SUB_SW_IF_INDEX, -98, "Invalid sub-interface sw_if_index") \ _(TABLE_TOO_BIG, -99, "Table too big") \ _(CANNOT_ENABLE_DISABLE_FEATURE, -100, "Cannot enable/disable feature") \ -_(BFD_EEXIST, -101, "Duplicate BFD object") \ -_(BFD_ENOENT, -102, "No such BFD object") \ -_(BFD_EINUSE, -103, "BFD object in use") \ -_(BFD_NOTSUPP, -104, "BFD feature not supported") \ -_(LISP_RLOC_LOCAL, -105, "RLOC address is local") \ -_(BFD_EAGAIN, -106, "BFD object cannot be manipulated at this time") \ -_(INVALID_GPE_MODE, -107, "Invalid GPE mode") \ -_(LISP_GPE_ENTRIES_PRESENT, -108, "LISP GPE entries are present") +_(BFD_EEXIST, -101, "Duplicate BFD object") \ +_(BFD_ENOENT, -102, "No such BFD object") \ +_(BFD_EINUSE, -103, "BFD object in use") \ +_(BFD_NOTSUPP, -104, "BFD feature not supported") \ +_(ADDRESS_IN_USE, -105, "Address in use") \ +_(ADDRESS_NOT_IN_USE, -106, "Address not in use") \ +_(QUEUE_FULL, -107, "Queue full") \ +_(UNKNOWN_URI_TYPE, -108, "Unknown URI type") \ +_(URI_FIFO_CREATE_FAILED, -109, "URI FIFO segment create failed") \ +_(LISP_RLOC_LOCAL, -110, "RLOC address is local") \ +_(BFD_EAGAIN, -111, "BFD object cannot be manipulated at this time") \ +_(INVALID_GPE_MODE, -112, "Invalid GPE mode") \ +_(LISP_GPE_ENTRIES_PRESENT, -113, "LISP GPE entries are present") typedef enum { diff --git a/src/vnet/bfd/bfd_udp.c b/src/vnet/bfd/bfd_udp.c index 146faad6..cf05089b 100644 --- a/src/vnet/bfd/bfd_udp.c +++ b/src/vnet/bfd/bfd_udp.c @@ -18,12 +18,12 @@ #include #include #include -#include +#include +#include #include #include #include #include -#include #include #include #include diff --git a/src/vnet/buffer.h b/src/vnet/buffer.h index f1cc6371..3de01f2a 100644 --- a/src/vnet/buffer.h +++ b/src/vnet/buffer.h @@ -277,6 +277,16 @@ typedef struct u16 buffer_advance; } device_input_feat; + /* TCP */ + struct + { + u32 connection_index; + u32 seq_number; + u32 seq_end; + u32 ack_number; + u8 flags; + } tcp; + u32 unused[6]; }; } vnet_buffer_opaque_t; diff --git a/src/vnet/classify/vnet_classify.c b/src/vnet/classify/vnet_classify.c index 6093e2ac..b651a1f1 100644 --- a/src/vnet/classify/vnet_classify.c +++ b/src/vnet/classify/vnet_classify.c @@ -695,8 +695,8 @@ int vnet_classify_add_del_table (vnet_classify_main_t * cm, } #define foreach_tcp_proto_field \ -_(src_port) \ -_(dst_port) +_(src) \ +_(dst) #define foreach_udp_proto_field \ _(src_port) \ diff --git a/src/vnet/dhcp/dhcp_proxy.h b/src/vnet/dhcp/dhcp_proxy.h index c0d79c41..4586d883 100644 --- a/src/vnet/dhcp/dhcp_proxy.h +++ b/src/vnet/dhcp/dhcp_proxy.h @@ -26,7 +26,7 @@ #include #include #include -#include +#include typedef enum { #define dhcp_proxy_error(n,s) DHCP_PROXY_ERROR_##n, diff --git a/src/vnet/flow/flow_report.h b/src/vnet/flow/flow_report.h index 4e764377..e8ed3818 100644 --- a/src/vnet/flow/flow_report.h +++ b/src/vnet/flow/flow_report.h @@ -23,7 +23,7 @@ #include #include #include -#include +#include #include #include #include diff --git a/src/vnet/ip/ip.h b/src/vnet/ip/ip.h index 02a1a963..70b4ccd8 100644 --- a/src/vnet/ip/ip.h +++ b/src/vnet/ip/ip.h @@ -50,8 +50,8 @@ #include #include -#include -#include +#include +#include #include #include diff --git a/src/vnet/ip/ip4.h b/src/vnet/ip/ip4.h index b184fbae..4e075d0f 100644 --- a/src/vnet/ip/ip4.h +++ b/src/vnet/ip/ip4.h @@ -309,8 +309,8 @@ ip4_compute_flow_hash (const ip4_header_t * ip, b = (flow_hash_config & IP_FLOW_HASH_REVERSE_SRC_DST) ? t1 : t2; b ^= (flow_hash_config & IP_FLOW_HASH_PROTO) ? ip->protocol : 0; - t1 = is_tcp_udp ? tcp->ports.src : 0; - t2 = is_tcp_udp ? tcp->ports.dst : 0; + t1 = is_tcp_udp ? tcp->src : 0; + t2 = is_tcp_udp ? tcp->dst : 0; t1 = (flow_hash_config & IP_FLOW_HASH_SRC_PORT) ? t1 : 0; t2 = (flow_hash_config & IP_FLOW_HASH_DST_PORT) ? t2 : 0; @@ -334,6 +334,44 @@ u8 *format_ip4_forward_next_trace (u8 * s, va_list * args); u32 ip4_tcp_udp_validate_checksum (vlib_main_t * vm, vlib_buffer_t * p0); +#define IP_DF 0x4000 /* don't fragment */ + +/** + * Push IPv4 header to buffer + * + * This does not support fragmentation. + * + * @param vm - vlib_main + * @param b - buffer to write the header to + * @param src - source IP + * @param dst - destination IP + * @param prot - payload proto + * + * @return - pointer to start of IP header + */ +always_inline void * +vlib_buffer_push_ip4 (vlib_main_t * vm, vlib_buffer_t * b, + ip4_address_t * src, ip4_address_t * dst, int proto) +{ + ip4_header_t *ih; + + /* make some room */ + ih = vlib_buffer_push_uninit (b, sizeof (ip4_header_t)); + + ih->ip_version_and_header_length = 0x45; + ih->tos = 0; + ih->length = clib_host_to_net_u16 (vlib_buffer_length_in_chain (vm, b)); + + /* No fragments */ + ih->flags_and_fragment_offset = clib_host_to_net_u16 (IP_DF); + ih->ttl = 255; + ih->protocol = proto; + ih->src_address.as_u32 = src->as_u32; + ih->dst_address.as_u32 = dst->as_u32; + + ih->checksum = ip4_header_checksum (ih); + return ih; +} #endif /* included_ip_ip4_h */ /* diff --git a/src/vnet/ip/ip4_forward.c b/src/vnet/ip/ip4_forward.c index 8081b34b..66d91ab6 100644 --- a/src/vnet/ip/ip4_forward.c +++ b/src/vnet/ip/ip4_forward.c @@ -1478,8 +1478,18 @@ ip4_tcp_udp_validate_checksum (vlib_main_t * vm, vlib_buffer_t * p0) return p0->flags; } -static uword -ip4_local (vlib_main_t * vm, vlib_node_runtime_t * node, vlib_frame_t * frame) +/* *INDENT-OFF* */ +VNET_FEATURE_ARC_INIT (ip4_local) = +{ + .arc_name = "ip4-local", + .start_nodes = VNET_FEATURES ("ip4-local"), +}; +/* *INDENT-ON* */ + +static inline uword +ip4_local_inline (vlib_main_t * vm, + vlib_node_runtime_t * node, + vlib_frame_t * frame, int head_of_feature_arc) { ip4_main_t *im = &ip4_main; ip_lookup_main_t *lm = &im->lookup_main; @@ -1487,6 +1497,7 @@ ip4_local (vlib_main_t * vm, vlib_node_runtime_t * node, vlib_frame_t * frame) u32 *from, *to_next, n_left_from, n_left_to_next; vlib_node_runtime_t *error_node = vlib_node_get_runtime (vm, ip4_input_node.index); + u8 arc_index = vnet_feat_arc_ip4_local.feature_arc_index; from = vlib_frame_vector_args (frame); n_left_from = frame->n_vectors; @@ -1513,7 +1524,7 @@ ip4_local (vlib_main_t * vm, vlib_node_runtime_t * node, vlib_frame_t * frame) i32 len_diff0, len_diff1; u8 error0, is_udp0, is_tcp_udp0, good_tcp_udp0, proto0; u8 error1, is_udp1, is_tcp_udp1, good_tcp_udp1, proto1; - u8 enqueue_code; + u32 sw_if_index0, sw_if_index1; pi0 = to_next[0] = from[0]; pi1 = to_next[1] = from[1]; @@ -1522,6 +1533,8 @@ ip4_local (vlib_main_t * vm, vlib_node_runtime_t * node, vlib_frame_t * frame) to_next += 2; n_left_to_next -= 2; + next0 = next1 = IP_LOCAL_NEXT_DROP; + p0 = vlib_get_buffer (vm, pi0); p1 = vlib_get_buffer (vm, pi1); @@ -1531,14 +1544,18 @@ ip4_local (vlib_main_t * vm, vlib_node_runtime_t * node, vlib_frame_t * frame) vnet_buffer (p0)->ip.start_of_ip_header = p0->current_data; vnet_buffer (p1)->ip.start_of_ip_header = p1->current_data; - fib_index0 = vec_elt (im->fib_index_by_sw_if_index, - vnet_buffer (p0)->sw_if_index[VLIB_RX]); + sw_if_index0 = vnet_buffer (p0)->sw_if_index[VLIB_RX]; + sw_if_index1 = vnet_buffer (p1)->sw_if_index[VLIB_RX]; + + fib_index0 = vec_elt (im->fib_index_by_sw_if_index, sw_if_index0); + fib_index1 = vec_elt (im->fib_index_by_sw_if_index, sw_if_index1); + + fib_index0 = vec_elt (im->fib_index_by_sw_if_index, sw_if_index0); fib_index0 = (vnet_buffer (p0)->sw_if_index[VLIB_TX] == (u32) ~ 0) ? fib_index0 : vnet_buffer (p0)->sw_if_index[VLIB_TX]; - fib_index1 = vec_elt (im->fib_index_by_sw_if_index, - vnet_buffer (p1)->sw_if_index[VLIB_RX]); + fib_index1 = vec_elt (im->fib_index_by_sw_if_index, sw_if_index1); fib_index1 = (vnet_buffer (p1)->sw_if_index[VLIB_TX] == (u32) ~ 0) ? fib_index1 : vnet_buffer (p1)->sw_if_index[VLIB_TX]; @@ -1557,6 +1574,13 @@ ip4_local (vlib_main_t * vm, vlib_node_runtime_t * node, vlib_frame_t * frame) until support of IP frag reassembly is implemented */ proto0 = ip4_is_fragment (ip0) ? 0xfe : ip0->protocol; proto1 = ip4_is_fragment (ip1) ? 0xfe : ip1->protocol; + + if (head_of_feature_arc == 0) + { + error0 = error1 = IP4_ERROR_UNKNOWN_PROTOCOL; + goto skip_checks; + } + is_udp0 = proto0 == IP_PROTOCOL_UDP; is_udp1 = proto1 == IP_PROTOCOL_UDP; is_tcp_udp0 = is_udp0 || proto0 == IP_PROTOCOL_TCP; @@ -1686,6 +1710,7 @@ ip4_local (vlib_main_t * vm, vlib_node_runtime_t * node, vlib_frame_t * frame) next0 = lm->local_next_by_ip_protocol[proto0]; next1 = lm->local_next_by_ip_protocol[proto1]; + skip_checks: next0 = error0 != IP4_ERROR_UNKNOWN_PROTOCOL ? IP_LOCAL_NEXT_DROP : next0; next1 = @@ -1694,44 +1719,17 @@ ip4_local (vlib_main_t * vm, vlib_node_runtime_t * node, vlib_frame_t * frame) p0->error = error0 ? error_node->errors[error0] : 0; p1->error = error1 ? error_node->errors[error1] : 0; - enqueue_code = (next0 != next_index) + 2 * (next1 != next_index); - - if (PREDICT_FALSE (enqueue_code != 0)) + if (head_of_feature_arc) { - switch (enqueue_code) - { - case 1: - /* A B A */ - to_next[-2] = pi1; - to_next -= 1; - n_left_to_next += 1; - vlib_set_next_frame_buffer (vm, node, next0, pi0); - break; - - case 2: - /* A A B */ - to_next -= 1; - n_left_to_next += 1; - vlib_set_next_frame_buffer (vm, node, next1, pi1); - break; - - case 3: - /* A B B or A B C */ - to_next -= 2; - n_left_to_next += 2; - vlib_set_next_frame_buffer (vm, node, next0, pi0); - vlib_set_next_frame_buffer (vm, node, next1, pi1); - if (next0 == next1) - { - vlib_put_next_frame (vm, node, next_index, - n_left_to_next); - next_index = next1; - vlib_get_next_frame (vm, node, next_index, to_next, - n_left_to_next); - } - break; - } + if (PREDICT_TRUE (error0 == (u8) IP4_ERROR_UNKNOWN_PROTOCOL)) + vnet_feature_arc_start (arc_index, sw_if_index0, &next0, p0); + if (PREDICT_TRUE (error1 == (u8) IP4_ERROR_UNKNOWN_PROTOCOL)) + vnet_feature_arc_start (arc_index, sw_if_index1, &next1, p1); } + + vlib_validate_buffer_enqueue_x2 (vm, node, next_index, to_next, + n_left_to_next, pi0, pi1, + next0, next1); } while (n_left_from > 0 && n_left_to_next > 0) @@ -1746,6 +1744,7 @@ ip4_local (vlib_main_t * vm, vlib_node_runtime_t * node, vlib_frame_t * frame) u8 error0, is_udp0, is_tcp_udp0, good_tcp_udp0, proto0; load_balance_t *lb0; const dpo_id_t *dpo0; + u32 sw_if_index0; pi0 = to_next[0] = from[0]; from += 1; @@ -1753,14 +1752,18 @@ ip4_local (vlib_main_t * vm, vlib_node_runtime_t * node, vlib_frame_t * frame) to_next += 1; n_left_to_next -= 1; + next0 = IP_LOCAL_NEXT_DROP; + p0 = vlib_get_buffer (vm, pi0); ip0 = vlib_buffer_get_current (p0); vnet_buffer (p0)->ip.start_of_ip_header = p0->current_data; - fib_index0 = vec_elt (im->fib_index_by_sw_if_index, - vnet_buffer (p0)->sw_if_index[VLIB_RX]); + sw_if_index0 = vnet_buffer (p0)->sw_if_index[VLIB_RX]; + + fib_index0 = vec_elt (im->fib_index_by_sw_if_index, sw_if_index0); + fib_index0 = (vnet_buffer (p0)->sw_if_index[VLIB_TX] == (u32) ~ 0) ? fib_index0 : vnet_buffer (p0)->sw_if_index[VLIB_TX]; @@ -1775,6 +1778,13 @@ ip4_local (vlib_main_t * vm, vlib_node_runtime_t * node, vlib_frame_t * frame) /* Treat IP frag packets as "experimental" protocol for now until support of IP frag reassembly is implemented */ proto0 = ip4_is_fragment (ip0) ? 0xfe : ip0->protocol; + + if (head_of_feature_arc == 0) + { + error0 = IP4_ERROR_UNKNOWN_PROTOCOL; + goto skip_check; + } + is_udp0 = proto0 == IP_PROTOCOL_UDP; is_tcp_udp0 = is_udp0 || proto0 == IP_PROTOCOL_TCP; @@ -1847,6 +1857,8 @@ ip4_local (vlib_main_t * vm, vlib_node_runtime_t * node, vlib_frame_t * frame) ip0->dst_address.as_u32 != 0xFFFFFFFF) ? IP4_ERROR_SRC_LOOKUP_MISS : error0); + skip_check: + next0 = lm->local_next_by_ip_protocol[proto0]; next0 = @@ -1854,18 +1866,15 @@ ip4_local (vlib_main_t * vm, vlib_node_runtime_t * node, vlib_frame_t * frame) p0->error = error0 ? error_node->errors[error0] : 0; - if (PREDICT_FALSE (next0 != next_index)) + if (head_of_feature_arc) { - n_left_to_next += 1; - vlib_put_next_frame (vm, node, next_index, n_left_to_next); - - next_index = next0; - vlib_get_next_frame (vm, node, next_index, to_next, - n_left_to_next); - to_next[0] = pi0; - to_next += 1; - n_left_to_next -= 1; + if (PREDICT_TRUE (error0 == (u8) IP4_ERROR_UNKNOWN_PROTOCOL)) + vnet_feature_arc_start (arc_index, sw_if_index0, &next0, p0); } + + vlib_validate_buffer_enqueue_x1 (vm, node, next_index, to_next, + n_left_to_next, pi0, next0); + } vlib_put_next_frame (vm, node, next_index, n_left_to_next); @@ -1874,21 +1883,57 @@ ip4_local (vlib_main_t * vm, vlib_node_runtime_t * node, vlib_frame_t * frame) return frame->n_vectors; } +static uword +ip4_local (vlib_main_t * vm, vlib_node_runtime_t * node, vlib_frame_t * frame) +{ + return ip4_local_inline (vm, node, frame, 1 /* head of feature arc */ ); +} + +/* *INDENT-OFF* */ VLIB_REGISTER_NODE (ip4_local_node) = { - .function = ip4_local,.name = "ip4-local",.vector_size = - sizeof (u32),.format_trace = - format_ip4_forward_next_trace,.n_next_nodes = - IP_LOCAL_N_NEXT,.next_nodes = + .function = ip4_local, + .name = "ip4-local", + .vector_size = sizeof (u32), + .format_trace = format_ip4_forward_next_trace, + .n_next_nodes = IP_LOCAL_N_NEXT, + .next_nodes = { - [IP_LOCAL_NEXT_DROP] = "error-drop", - [IP_LOCAL_NEXT_PUNT] = "error-punt", - [IP_LOCAL_NEXT_UDP_LOOKUP] = "ip4-udp-lookup", - [IP_LOCAL_NEXT_ICMP] = "ip4-icmp-input",} -,}; + [IP_LOCAL_NEXT_DROP] = "error-drop", + [IP_LOCAL_NEXT_PUNT] = "error-punt", + [IP_LOCAL_NEXT_UDP_LOOKUP] = "ip4-udp-lookup", + [IP_LOCAL_NEXT_ICMP] = "ip4-icmp-input",}, +}; +/* *INDENT-ON* */ VLIB_NODE_FUNCTION_MULTIARCH (ip4_local_node, ip4_local); +static uword +ip4_local_end_of_arc (vlib_main_t * vm, + vlib_node_runtime_t * node, vlib_frame_t * frame) +{ + return ip4_local_inline (vm, node, frame, 0 /* head of feature arc */ ); +} + +/* *INDENT-OFF* */ +VLIB_REGISTER_NODE (ip4_local_end_of_arc_node,static) = { + .function = ip4_local_end_of_arc, + .name = "ip4-local-end-of-arc", + .vector_size = sizeof (u32), + + .format_trace = format_ip4_forward_next_trace, + .sibling_of = "ip4-local", +}; + +VLIB_NODE_FUNCTION_MULTIARCH (ip4_local_end_of_arc_node, ip4_local_end_of_arc) + +VNET_FEATURE_INIT (ip4_local_end_of_arc, static) = { + .arc_name = "ip4-local", + .node_name = "ip4-local-end-of-arc", + .runs_before = 0, /* not before any other features */ +}; +/* *INDENT-ON* */ + void ip4_register_protocol (u32 protocol, u32 node_index) { diff --git a/src/vnet/ip/ip4_packet.h b/src/vnet/ip/ip4_packet.h index 8da788b4..b2c1fcd4 100644 --- a/src/vnet/ip/ip4_packet.h +++ b/src/vnet/ip/ip4_packet.h @@ -41,7 +41,7 @@ #define included_ip4_packet_h #include /* for ip_csum_t */ -#include /* for tcp_header_t */ +#include /* for tcp_header_t */ #include /* for clib_net_to_host_u16 */ /* IP4 address which can be accessed either as 4 bytes @@ -342,10 +342,10 @@ ip4_tcp_reply_x1 (ip4_header_t * ip0, tcp_header_t * tcp0) ip0->src_address.data_u32 = dst0; ip0->dst_address.data_u32 = src0; - src0 = tcp0->ports.src; - dst0 = tcp0->ports.dst; - tcp0->ports.src = dst0; - tcp0->ports.dst = src0; + src0 = tcp0->src; + dst0 = tcp0->dst; + tcp0->src = dst0; + tcp0->dst = src0; } always_inline void @@ -363,14 +363,14 @@ ip4_tcp_reply_x2 (ip4_header_t * ip0, ip4_header_t * ip1, ip0->dst_address.data_u32 = src0; ip1->dst_address.data_u32 = src1; - src0 = tcp0->ports.src; - src1 = tcp1->ports.src; - dst0 = tcp0->ports.dst; - dst1 = tcp1->ports.dst; - tcp0->ports.src = dst0; - tcp1->ports.src = dst1; - tcp0->ports.dst = src0; - tcp1->ports.dst = src1; + src0 = tcp0->src; + src1 = tcp1->src; + dst0 = tcp0->dst; + dst1 = tcp1->dst; + tcp0->src = dst0; + tcp1->src = dst1; + tcp0->dst = src0; + tcp1->dst = src1; } #endif /* included_ip4_packet_h */ diff --git a/src/vnet/ip/ip6.h b/src/vnet/ip/ip6.h index 5456f0f2..2615fbfa 100644 --- a/src/vnet/ip/ip6.h +++ b/src/vnet/ip/ip6.h @@ -461,8 +461,8 @@ ip6_compute_flow_hash (const ip6_header_t * ip, b = (flow_hash_config & IP_FLOW_HASH_REVERSE_SRC_DST) ? t1 : t2; b ^= (flow_hash_config & IP_FLOW_HASH_PROTO) ? ip->protocol : 0; - t1 = is_tcp_udp ? tcp->ports.src : 0; - t2 = is_tcp_udp ? tcp->ports.dst : 0; + t1 = is_tcp_udp ? tcp->src : 0; + t2 = is_tcp_udp ? tcp->dst : 0; t1 = (flow_hash_config & IP_FLOW_HASH_SRC_PORT) ? t1 : 0; t2 = (flow_hash_config & IP_FLOW_HASH_DST_PORT) ? t2 : 0; @@ -497,6 +497,46 @@ int ip6_hbh_register_option (u8 option, int ip6_hbh_unregister_option (u8 option); void ip6_hbh_set_next_override (uword next); +/** + * Push IPv6 header to buffer + * + * @param vm - vlib_main + * @param b - buffer to write the header to + * @param src - source IP + * @param dst - destination IP + * @param prot - payload proto + * + * @return - pointer to start of IP header + */ +always_inline void * +vlib_buffer_push_ip6 (vlib_main_t * vm, vlib_buffer_t * b, + ip6_address_t * src, ip6_address_t * dst, int proto) +{ + ip6_header_t *ip6h; + u16 payload_length; + + /* make some room */ + ip6h = vlib_buffer_push_uninit (b, sizeof (ip6_header_t)); + + ip6h->ip_version_traffic_class_and_flow_label = + clib_host_to_net_u32 (0x6 << 28); + + /* calculate ip6 payload length */ + payload_length = vlib_buffer_length_in_chain (vm, b); + payload_length -= sizeof (*ip6h); + + ip6h->payload_length = clib_host_to_net_u16 (payload_length); + + ip6h->hop_limit = 0xff; + ip6h->protocol = proto; + clib_memcpy (ip6h->src_address.as_u8, src->as_u8, + sizeof (ip6h->src_address)); + clib_memcpy (ip6h->dst_address.as_u8, dst->as_u8, + sizeof (ip6h->src_address)); + + return ip6h; +} + #endif /* included_ip_ip6_h */ /* diff --git a/src/vnet/ip/ip6_packet.h b/src/vnet/ip/ip6_packet.h index 1e551c8b..4fd14b96 100644 --- a/src/vnet/ip/ip6_packet.h +++ b/src/vnet/ip/ip6_packet.h @@ -40,7 +40,7 @@ #ifndef included_ip6_packet_h #define included_ip6_packet_h -#include +#include #include typedef union @@ -373,10 +373,10 @@ ip6_tcp_reply_x1 (ip6_header_t * ip0, tcp_header_t * tcp0) { u16 src0, dst0; - src0 = tcp0->ports.src; - dst0 = tcp0->ports.dst; - tcp0->ports.src = dst0; - tcp0->ports.dst = src0; + src0 = tcp0->src; + dst0 = tcp0->dst; + tcp0->src = dst0; + tcp0->dst = src0; } } @@ -400,14 +400,14 @@ ip6_tcp_reply_x2 (ip6_header_t * ip0, ip6_header_t * ip1, { u16 src0, dst0, src1, dst1; - src0 = tcp0->ports.src; - src1 = tcp1->ports.src; - dst0 = tcp0->ports.dst; - dst1 = tcp1->ports.dst; - tcp0->ports.src = dst0; - tcp1->ports.src = dst1; - tcp0->ports.dst = src0; - tcp1->ports.dst = src1; + src0 = tcp0->src; + src1 = tcp1->src; + dst0 = tcp0->dst; + dst1 = tcp1->dst; + tcp0->src = dst0; + tcp1->src = dst1; + tcp0->dst = src0; + tcp1->dst = src1; } } diff --git a/src/vnet/ip/punt.c b/src/vnet/ip/punt.c index 9c735128..48558401 100644 --- a/src/vnet/ip/punt.c +++ b/src/vnet/ip/punt.c @@ -23,7 +23,7 @@ */ #include #include -#include +#include #include #define foreach_punt_next \ diff --git a/src/vnet/ip/tcp_packet.h b/src/vnet/ip/tcp_packet.h deleted file mode 100644 index 93f73e01..00000000 --- a/src/vnet/ip/tcp_packet.h +++ /dev/null @@ -1,141 +0,0 @@ -/* - * Copyright (c) 2015 Cisco and/or its affiliates. - * Licensed under the Apache License, Version 2.0 (the "License"); - * you may not use this file except in compliance with the License. - * You may obtain a copy of the License at: - * - * http://www.apache.org/licenses/LICENSE-2.0 - * - * Unless required by applicable law or agreed to in writing, software - * distributed under the License is distributed on an "AS IS" BASIS, - * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. - * See the License for the specific language governing permissions and - * limitations under the License. - */ -/* - * ip4/tcp_packet.h: TCP packet format (see RFC 793) - * - * Copyright (c) 2008 Eliot Dresselhaus - * - * Permission is hereby granted, free of charge, to any person obtaining - * a copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sublicense, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice shall be - * included in all copies or substantial portions of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, - * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND - * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE - * LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION - * OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION - * WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - */ - -#ifndef included_tcp_packet_h -#define included_tcp_packet_h - -/* TCP flags bit 0 first. */ -#define foreach_tcp_flag \ - _ (FIN) \ - _ (SYN) \ - _ (RST) \ - _ (PSH) \ - _ (ACK) \ - _ (URG) \ - _ (ECE) \ - _ (CWR) - -enum -{ -#define _(f) TCP_FLAG_BIT_##f, - foreach_tcp_flag -#undef _ - TCP_N_FLAG_BITS, - -#define _(f) TCP_FLAG_##f = 1 << TCP_FLAG_BIT_##f, - foreach_tcp_flag -#undef _ -}; - -typedef struct -{ - /* Source and destination port. */ - union - { - union - { - struct - { - u16 src, dst; - }; - u32 src_and_dst; - } ports; - struct - { - u16 src_port, dst_port; - }; - }; - - /* Sequence and acknowledgment number. */ - u32 seq_number, ack_number; - - /* Size of TCP header in 32-bit units plus 4 reserved bits. */ - u8 tcp_header_u32s_and_reserved; - - /* see foreach_tcp_flag for enumation of tcp flags. */ - u8 flags; - - /* Current window advertised by sender. - This is the number of bytes sender is willing to receive - right now. */ - u16 window; - - /* Checksum of TCP pseudo header and data. */ - u16 checksum; - - u16 urgent_pointer; -} tcp_header_t; - -always_inline int -tcp_header_bytes (tcp_header_t * t) -{ - return (t->tcp_header_u32s_and_reserved >> 4) * sizeof (u32); -} - -/* TCP options. */ -typedef enum tcp_option_type -{ - TCP_OPTION_END = 0, - TCP_OPTION_NOP = 1, - TCP_OPTION_MSS = 2, - TCP_OPTION_WINDOW_SCALE = 3, - TCP_OPTION_SACK_PERMITTED = 4, - TCP_OPTION_SACK_BLOCK = 5, - TCP_OPTION_TIME_STAMP = 8, -} tcp_option_type_t; - -/* All except NOP and END have 1 byte length field. */ -typedef struct -{ - tcp_option_type_t type:8; - - /* Length of this option in bytes. */ - u8 length; -} tcp_option_with_length_t; - -#endif /* included_tcp_packet_h */ - - -/* - * fd.io coding-style-patch-verification: ON - * - * Local Variables: - * eval: (c-set-style "gnu") - * End: - */ diff --git a/src/vnet/ip/udp.h b/src/vnet/ip/udp.h deleted file mode 100644 index bad58b5d..00000000 --- a/src/vnet/ip/udp.h +++ /dev/null @@ -1,315 +0,0 @@ -/* - * ip/udp.h: udp protocol - * - * Copyright (c) 2013 Cisco and/or its affiliates. - * Licensed under the Apache License, Version 2.0 (the "License"); - * you may not use this file except in compliance with the License. - * You may obtain a copy of the License at: - * - * http://www.apache.org/licenses/LICENSE-2.0 - * - * Unless required by applicable law or agreed to in writing, software - * distributed under the License is distributed on an "AS IS" BASIS, - * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. - * See the License for the specific language governing permissions and - * limitations under the License. - */ - -#ifndef included_udp_h -#define included_udp_h - -#include -#include -#include -#include -#include -#include -#include - -typedef enum -{ -#define udp_error(n,s) UDP_ERROR_##n, -#include -#undef udp_error - UDP_N_ERROR, -} udp_error_t; - -#define foreach_udp4_dst_port \ -_ (67, dhcp_to_server) \ -_ (68, dhcp_to_client) \ -_ (500, ikev2) \ -_ (3784, bfd4) \ -_ (3785, bfd_echo4) \ -_ (4341, lisp_gpe) \ -_ (4342, lisp_cp) \ -_ (4739, ipfix) \ -_ (4789, vxlan) \ -_ (4789, vxlan6) \ -_ (4790, vxlan_gpe) \ -_ (6633, vpath_3) - - -#define foreach_udp6_dst_port \ -_ (547, dhcpv6_to_server) \ -_ (546, dhcpv6_to_client) \ -_ (3784, bfd6) \ -_ (3785, bfd_echo6) \ -_ (4341, lisp_gpe6) \ -_ (4342, lisp_cp6) \ -_ (4790, vxlan6_gpe) \ -_ (6633, vpath6_3) - -typedef enum -{ -#define _(n,f) UDP_DST_PORT_##f = n, - foreach_udp4_dst_port foreach_udp6_dst_port -#undef _ -} udp_dst_port_t; - -typedef enum -{ -#define _(n,f) UDP6_DST_PORT_##f = n, - foreach_udp6_dst_port -#undef _ -} udp6_dst_port_t; - -typedef struct -{ - /* Name (a c string). */ - char *name; - - /* GRE protocol type in host byte order. */ - udp_dst_port_t dst_port; - - /* Node which handles this type. */ - u32 node_index; - - /* Next index for this type. */ - u32 next_index; -} udp_dst_port_info_t; - -typedef enum -{ - UDP_IP6 = 0, - UDP_IP4, /* the code is full of is_ip4... */ - N_UDP_AF, -} udp_af_t; - -typedef struct -{ - udp_dst_port_info_t *dst_port_infos[N_UDP_AF]; - - /* Hash tables mapping name/protocol to protocol info index. */ - uword *dst_port_info_by_name[N_UDP_AF]; - uword *dst_port_info_by_dst_port[N_UDP_AF]; - - /* convenience */ - vlib_main_t *vlib_main; -} udp_main_t; - -always_inline udp_dst_port_info_t * -udp_get_dst_port_info (udp_main_t * um, udp_dst_port_t dst_port, u8 is_ip4) -{ - uword *p = hash_get (um->dst_port_info_by_dst_port[is_ip4], dst_port); - return p ? vec_elt_at_index (um->dst_port_infos[is_ip4], p[0]) : 0; -} - -format_function_t format_udp_header; -format_function_t format_udp_rx_trace; - -unformat_function_t unformat_udp_header; - -void udp_register_dst_port (vlib_main_t * vm, - udp_dst_port_t dst_port, - u32 node_index, u8 is_ip4); - -void udp_punt_unknown (vlib_main_t * vm, u8 is_ip4, u8 is_add); - -always_inline void -ip_udp_fixup_one (vlib_main_t * vm, vlib_buffer_t * b0, u8 is_ip4) -{ - u16 new_l0; - udp_header_t *udp0; - - if (is_ip4) - { - ip4_header_t *ip0; - ip_csum_t sum0; - u16 old_l0 = 0; - - ip0 = vlib_buffer_get_current (b0); - - /* fix the ing outer-IP checksum */ - sum0 = ip0->checksum; - /* old_l0 always 0, see the rewrite setup */ - new_l0 = clib_host_to_net_u16 (vlib_buffer_length_in_chain (vm, b0)); - - sum0 = ip_csum_update (sum0, old_l0, new_l0, ip4_header_t, - length /* changed member */ ); - ip0->checksum = ip_csum_fold (sum0); - ip0->length = new_l0; - - /* Fix UDP length */ - udp0 = (udp_header_t *) (ip0 + 1); - new_l0 = clib_host_to_net_u16 (vlib_buffer_length_in_chain (vm, b0) - - sizeof (*ip0)); - udp0->length = new_l0; - } - else - { - ip6_header_t *ip0; - int bogus0; - - ip0 = vlib_buffer_get_current (b0); - - new_l0 = clib_host_to_net_u16 (vlib_buffer_length_in_chain (vm, b0) - - sizeof (*ip0)); - ip0->payload_length = new_l0; - - /* Fix UDP length */ - udp0 = (udp_header_t *) (ip0 + 1); - udp0->length = new_l0; - - udp0->checksum = - ip6_tcp_udp_icmp_compute_checksum (vm, b0, ip0, &bogus0); - ASSERT (bogus0 == 0); - - if (udp0->checksum == 0) - udp0->checksum = 0xffff; - } -} - -always_inline void -ip_udp_encap_one (vlib_main_t * vm, vlib_buffer_t * b0, u8 * ec0, word ec_len, - u8 is_ip4) -{ - vlib_buffer_advance (b0, -ec_len); - - if (is_ip4) - { - ip4_header_t *ip0; - - ip0 = vlib_buffer_get_current (b0); - - /* Apply the encap string. */ - clib_memcpy (ip0, ec0, ec_len); - ip_udp_fixup_one (vm, b0, 1); - } - else - { - ip6_header_t *ip0; - - ip0 = vlib_buffer_get_current (b0); - - /* Apply the encap string. */ - clib_memcpy (ip0, ec0, ec_len); - ip_udp_fixup_one (vm, b0, 0); - } -} - -always_inline void -ip_udp_encap_two (vlib_main_t * vm, vlib_buffer_t * b0, vlib_buffer_t * b1, - u8 * ec0, u8 * ec1, word ec_len, u8 is_v4) -{ - u16 new_l0, new_l1; - udp_header_t *udp0, *udp1; - - ASSERT (_vec_len (ec0) == _vec_len (ec1)); - - vlib_buffer_advance (b0, -ec_len); - vlib_buffer_advance (b1, -ec_len); - - if (is_v4) - { - ip4_header_t *ip0, *ip1; - ip_csum_t sum0, sum1; - u16 old_l0 = 0, old_l1 = 0; - - ip0 = vlib_buffer_get_current (b0); - ip1 = vlib_buffer_get_current (b1); - - /* Apply the encap string */ - clib_memcpy (ip0, ec0, ec_len); - clib_memcpy (ip1, ec1, ec_len); - - /* fix the ing outer-IP checksum */ - sum0 = ip0->checksum; - sum1 = ip1->checksum; - - /* old_l0 always 0, see the rewrite setup */ - new_l0 = clib_host_to_net_u16 (vlib_buffer_length_in_chain (vm, b0)); - new_l1 = clib_host_to_net_u16 (vlib_buffer_length_in_chain (vm, b1)); - - sum0 = ip_csum_update (sum0, old_l0, new_l0, ip4_header_t, - length /* changed member */ ); - sum1 = ip_csum_update (sum1, old_l1, new_l1, ip4_header_t, - length /* changed member */ ); - - ip0->checksum = ip_csum_fold (sum0); - ip1->checksum = ip_csum_fold (sum1); - - ip0->length = new_l0; - ip1->length = new_l1; - - /* Fix UDP length */ - udp0 = (udp_header_t *) (ip0 + 1); - udp1 = (udp_header_t *) (ip1 + 1); - - new_l0 = - clib_host_to_net_u16 (vlib_buffer_length_in_chain (vm, b0) - - sizeof (*ip0)); - new_l1 = - clib_host_to_net_u16 (vlib_buffer_length_in_chain (vm, b1) - - sizeof (*ip1)); - udp0->length = new_l0; - udp1->length = new_l1; - } - else - { - ip6_header_t *ip0, *ip1; - int bogus0, bogus1; - - ip0 = vlib_buffer_get_current (b0); - ip1 = vlib_buffer_get_current (b1); - - /* Apply the encap string. */ - clib_memcpy (ip0, ec0, ec_len); - clib_memcpy (ip1, ec1, ec_len); - - new_l0 = clib_host_to_net_u16 (vlib_buffer_length_in_chain (vm, b0) - - sizeof (*ip0)); - new_l1 = clib_host_to_net_u16 (vlib_buffer_length_in_chain (vm, b1) - - sizeof (*ip1)); - ip0->payload_length = new_l0; - ip1->payload_length = new_l1; - - /* Fix UDP length */ - udp0 = (udp_header_t *) (ip0 + 1); - udp1 = (udp_header_t *) (ip1 + 1); - - udp0->length = new_l0; - udp1->length = new_l1; - - udp0->checksum = - ip6_tcp_udp_icmp_compute_checksum (vm, b0, ip0, &bogus0); - udp1->checksum = - ip6_tcp_udp_icmp_compute_checksum (vm, b1, ip1, &bogus1); - ASSERT (bogus0 == 0); - ASSERT (bogus1 == 0); - - if (udp0->checksum == 0) - udp0->checksum = 0xffff; - if (udp1->checksum == 0) - udp1->checksum = 0xffff; - } -} - -#endif /* included_udp_h */ - -/* - * fd.io coding-style-patch-verification: ON - * - * Local Variables: - * eval: (c-set-style "gnu") - * End: - */ diff --git a/src/vnet/ip/udp_error.def b/src/vnet/ip/udp_error.def deleted file mode 100644 index bfdae0ac..00000000 --- a/src/vnet/ip/udp_error.def +++ /dev/null @@ -1,21 +0,0 @@ -/* - * udp_error.def: udp errors - * - * Copyright (c) 2013-2016 Cisco and/or its affiliates. - * Licensed under the Apache License, Version 2.0 (the "License"); - * you may not use this file except in compliance with the License. - * You may obtain a copy of the License at: - * - * http://www.apache.org/licenses/LICENSE-2.0 - * - * Unless required by applicable law or agreed to in writing, software - * distributed under the License is distributed on an "AS IS" BASIS, - * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. - * See the License for the specific language governing permissions and - * limitations under the License. - */ - -udp_error (NONE, "no error") -udp_error (NO_LISTENER, "no listener for dst port") -udp_error (LENGTH_ERROR, "UDP packets with length errors") -udp_error (PUNT, "no listener punt") diff --git a/src/vnet/ip/udp_format.c b/src/vnet/ip/udp_format.c deleted file mode 100644 index abdf561e..00000000 --- a/src/vnet/ip/udp_format.c +++ /dev/null @@ -1,91 +0,0 @@ -/* - * Copyright (c) 2015 Cisco and/or its affiliates. - * Licensed under the Apache License, Version 2.0 (the "License"); - * you may not use this file except in compliance with the License. - * You may obtain a copy of the License at: - * - * http://www.apache.org/licenses/LICENSE-2.0 - * - * Unless required by applicable law or agreed to in writing, software - * distributed under the License is distributed on an "AS IS" BASIS, - * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. - * See the License for the specific language governing permissions and - * limitations under the License. - */ -/* - * ip/udp_format.c: udp formatting - * - * Copyright (c) 2008 Eliot Dresselhaus - * - * Permission is hereby granted, free of charge, to any person obtaining - * a copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sublicense, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice shall be - * included in all copies or substantial portions of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, - * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND - * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE - * LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION - * OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION - * WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - */ - -#include - -/* Format UDP header. */ -u8 * -format_udp_header (u8 * s, va_list * args) -{ - udp_header_t *udp = va_arg (*args, udp_header_t *); - u32 max_header_bytes = va_arg (*args, u32); - uword indent; - u32 header_bytes = sizeof (udp[0]); - - /* Nothing to do. */ - if (max_header_bytes < sizeof (udp[0])) - return format (s, "UDP header truncated"); - - indent = format_get_indent (s); - indent += 2; - - s = format (s, "UDP: %d -> %d", - clib_net_to_host_u16 (udp->src_port), - clib_net_to_host_u16 (udp->dst_port)); - - s = format (s, "\n%Ulength %d, checksum 0x%04x", - format_white_space, indent, - clib_net_to_host_u16 (udp->length), - clib_net_to_host_u16 (udp->checksum)); - - /* Recurse into next protocol layer. */ - if (max_header_bytes != 0 && header_bytes < max_header_bytes) - { - ip_main_t *im = &ip_main; - tcp_udp_port_info_t *pi; - - pi = ip_get_tcp_udp_port_info (im, udp->dst_port); - - if (pi && pi->format_header) - s = format (s, "\n%U%U", - format_white_space, indent - 2, pi->format_header, - /* next protocol header */ (udp + 1), - max_header_bytes - sizeof (udp[0])); - } - - return s; -} - -/* - * fd.io coding-style-patch-verification: ON - * - * Local Variables: - * eval: (c-set-style "gnu") - * End: - */ diff --git a/src/vnet/ip/udp_init.c b/src/vnet/ip/udp_init.c deleted file mode 100644 index 1241ca4a..00000000 --- a/src/vnet/ip/udp_init.c +++ /dev/null @@ -1,71 +0,0 @@ -/* - * Copyright (c) 2015 Cisco and/or its affiliates. - * Licensed under the Apache License, Version 2.0 (the "License"); - * you may not use this file except in compliance with the License. - * You may obtain a copy of the License at: - * - * http://www.apache.org/licenses/LICENSE-2.0 - * - * Unless required by applicable law or agreed to in writing, software - * distributed under the License is distributed on an "AS IS" BASIS, - * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. - * See the License for the specific language governing permissions and - * limitations under the License. - */ -/* - * ip/udp_init.c: udp initialization - * - * Copyright (c) 2008 Eliot Dresselhaus - * - * Permission is hereby granted, free of charge, to any person obtaining - * a copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sublicense, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice shall be - * included in all copies or substantial portions of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, - * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND - * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE - * LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION - * OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION - * WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - */ - -#include - -clib_error_t * -udp_init (vlib_main_t * vm) -{ - ip_main_t *im = &ip_main; - ip_protocol_info_t *pi; - clib_error_t *error; - - error = vlib_call_init_function (vm, ip_main_init); - - if (!error) - { - pi = ip_get_protocol_info (im, IP_PROTOCOL_UDP); - if (pi == 0) - return clib_error_return (0, "UDP protocol info AWOL"); - pi->format_header = format_udp_header; - pi->unformat_pg_edit = unformat_pg_udp_header; - } - - return 0; -} - -VLIB_INIT_FUNCTION (udp_init); - -/* - * fd.io coding-style-patch-verification: ON - * - * Local Variables: - * eval: (c-set-style "gnu") - * End: - */ diff --git a/src/vnet/ip/udp_local.c b/src/vnet/ip/udp_local.c deleted file mode 100644 index 13ab6e4f..00000000 --- a/src/vnet/ip/udp_local.c +++ /dev/null @@ -1,645 +0,0 @@ -/* - * node.c: udp packet processing - * - * Copyright (c) 2013 Cisco and/or its affiliates. - * Licensed under the Apache License, Version 2.0 (the "License"); - * you may not use this file except in compliance with the License. - * You may obtain a copy of the License at: - * - * http://www.apache.org/licenses/LICENSE-2.0 - * - * Unless required by applicable law or agreed to in writing, software - * distributed under the License is distributed on an "AS IS" BASIS, - * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. - * See the License for the specific language governing permissions and - * limitations under the License. - */ - -#include -#include -#include -#include -#include - -udp_main_t udp_main; - -#define foreach_udp_input_next \ - _ (PUNT, "error-punt") \ - _ (DROP, "error-drop") \ - _ (ICMP4_ERROR, "ip4-icmp-error") \ - _ (ICMP6_ERROR, "ip6-icmp-error") - -typedef enum -{ -#define _(s,n) UDP_INPUT_NEXT_##s, - foreach_udp_input_next -#undef _ - UDP_INPUT_N_NEXT, -} udp_input_next_t; - -typedef struct -{ - u16 src_port; - u16 dst_port; - u8 bound; -} udp_rx_trace_t; - -u8 * -format_udp_rx_trace (u8 * s, va_list * args) -{ - CLIB_UNUSED (vlib_main_t * vm) = va_arg (*args, vlib_main_t *); - CLIB_UNUSED (vlib_node_t * node) = va_arg (*args, vlib_node_t *); - udp_rx_trace_t *t = va_arg (*args, udp_rx_trace_t *); - - s = format (s, "UDP: src-port %d dst-port %d%s", - clib_net_to_host_u16 (t->src_port), - clib_net_to_host_u16 (t->dst_port), - t->bound ? "" : " (no listener)"); - return s; -} - -typedef struct -{ - /* Sparse vector mapping udp dst_port in network byte order - to next index. */ - u16 *next_by_dst_port; - u8 punt_unknown; -} udp_input_runtime_t; - -vlib_node_registration_t udp4_input_node; -vlib_node_registration_t udp6_input_node; - -always_inline uword -udp46_input_inline (vlib_main_t * vm, - vlib_node_runtime_t * node, - vlib_frame_t * from_frame, int is_ip4) -{ - udp_input_runtime_t *rt = is_ip4 ? - (void *) vlib_node_get_runtime_data (vm, udp4_input_node.index) - : (void *) vlib_node_get_runtime_data (vm, udp6_input_node.index); - __attribute__ ((unused)) u32 n_left_from, next_index, *from, *to_next; - word n_no_listener = 0; - u8 punt_unknown = rt->punt_unknown; - - from = vlib_frame_vector_args (from_frame); - n_left_from = from_frame->n_vectors; - - next_index = node->cached_next_index; - - while (n_left_from > 0) - { - u32 n_left_to_next; - - vlib_get_next_frame (vm, node, next_index, to_next, n_left_to_next); - - while (n_left_from >= 4 && n_left_to_next >= 2) - { - u32 bi0, bi1; - vlib_buffer_t *b0, *b1; - udp_header_t *h0 = 0, *h1 = 0; - u32 i0, i1, dst_port0, dst_port1; - u32 advance0, advance1; - u32 error0, next0, error1, next1; - - /* Prefetch next iteration. */ - { - vlib_buffer_t *p2, *p3; - - p2 = vlib_get_buffer (vm, from[2]); - p3 = vlib_get_buffer (vm, from[3]); - - vlib_prefetch_buffer_header (p2, LOAD); - vlib_prefetch_buffer_header (p3, LOAD); - - CLIB_PREFETCH (p2->data, sizeof (h0[0]), LOAD); - CLIB_PREFETCH (p3->data, sizeof (h1[0]), LOAD); - } - - bi0 = from[0]; - bi1 = from[1]; - to_next[0] = bi0; - to_next[1] = bi1; - from += 2; - to_next += 2; - n_left_to_next -= 2; - n_left_from -= 2; - - b0 = vlib_get_buffer (vm, bi0); - b1 = vlib_get_buffer (vm, bi1); - - /* ip4/6_local hands us the ip header, not the udp header */ - if (is_ip4) - { - advance0 = sizeof (ip4_header_t); - advance1 = sizeof (ip4_header_t); - } - else - { - advance0 = sizeof (ip6_header_t); - advance1 = sizeof (ip6_header_t); - } - - if (PREDICT_FALSE (b0->current_length < advance0 + sizeof (*h0))) - { - error0 = UDP_ERROR_LENGTH_ERROR; - next0 = UDP_INPUT_NEXT_DROP; - } - else - { - vlib_buffer_advance (b0, advance0); - h0 = vlib_buffer_get_current (b0); - error0 = next0 = 0; - if (PREDICT_FALSE (clib_net_to_host_u16 (h0->length) > - vlib_buffer_length_in_chain (vm, b0))) - { - error0 = UDP_ERROR_LENGTH_ERROR; - next0 = UDP_INPUT_NEXT_DROP; - } - } - - if (PREDICT_FALSE (b1->current_length < advance1 + sizeof (*h1))) - { - error1 = UDP_ERROR_LENGTH_ERROR; - next1 = UDP_INPUT_NEXT_DROP; - } - else - { - vlib_buffer_advance (b1, advance1); - h1 = vlib_buffer_get_current (b1); - error1 = next1 = 0; - if (PREDICT_FALSE (clib_net_to_host_u16 (h1->length) > - vlib_buffer_length_in_chain (vm, b1))) - { - error1 = UDP_ERROR_LENGTH_ERROR; - next1 = UDP_INPUT_NEXT_DROP; - } - } - - /* Index sparse array with network byte order. */ - dst_port0 = (error0 == 0) ? h0->dst_port : 0; - dst_port1 = (error1 == 0) ? h1->dst_port : 0; - sparse_vec_index2 (rt->next_by_dst_port, dst_port0, dst_port1, - &i0, &i1); - next0 = (error0 == 0) ? vec_elt (rt->next_by_dst_port, i0) : next0; - next1 = (error1 == 0) ? vec_elt (rt->next_by_dst_port, i1) : next1; - - if (PREDICT_FALSE (i0 == SPARSE_VEC_INVALID_INDEX)) - { - // move the pointer back so icmp-error can find the - // ip packet header - vlib_buffer_advance (b0, -(word) advance0); - - if (PREDICT_FALSE (punt_unknown)) - { - b0->error = node->errors[UDP_ERROR_PUNT]; - next0 = UDP_INPUT_NEXT_PUNT; - } - else if (is_ip4) - { - icmp4_error_set_vnet_buffer (b0, - ICMP4_destination_unreachable, - ICMP4_destination_unreachable_port_unreachable, - 0); - next0 = UDP_INPUT_NEXT_ICMP4_ERROR; - n_no_listener++; - } - else - { - icmp6_error_set_vnet_buffer (b0, - ICMP6_destination_unreachable, - ICMP6_destination_unreachable_port_unreachable, - 0); - next0 = UDP_INPUT_NEXT_ICMP6_ERROR; - n_no_listener++; - } - } - else - { - b0->error = node->errors[UDP_ERROR_NONE]; - // advance to the payload - vlib_buffer_advance (b0, sizeof (*h0)); - } - - if (PREDICT_FALSE (i1 == SPARSE_VEC_INVALID_INDEX)) - { - // move the pointer back so icmp-error can find the - // ip packet header - vlib_buffer_advance (b1, -(word) advance1); - - if (PREDICT_FALSE (punt_unknown)) - { - b1->error = node->errors[UDP_ERROR_PUNT]; - next1 = UDP_INPUT_NEXT_PUNT; - } - else if (is_ip4) - { - icmp4_error_set_vnet_buffer (b1, - ICMP4_destination_unreachable, - ICMP4_destination_unreachable_port_unreachable, - 0); - next1 = UDP_INPUT_NEXT_ICMP4_ERROR; - n_no_listener++; - } - else - { - icmp6_error_set_vnet_buffer (b1, - ICMP6_destination_unreachable, - ICMP6_destination_unreachable_port_unreachable, - 0); - next1 = UDP_INPUT_NEXT_ICMP6_ERROR; - n_no_listener++; - } - } - else - { - b1->error = node->errors[UDP_ERROR_NONE]; - // advance to the payload - vlib_buffer_advance (b1, sizeof (*h1)); - } - - if (PREDICT_FALSE (b0->flags & VLIB_BUFFER_IS_TRACED)) - { - udp_rx_trace_t *tr = vlib_add_trace (vm, node, - b0, sizeof (*tr)); - if (b0->error != node->errors[UDP_ERROR_LENGTH_ERROR]) - { - tr->src_port = h0 ? h0->src_port : 0; - tr->dst_port = h0 ? h0->dst_port : 0; - tr->bound = (next0 != UDP_INPUT_NEXT_ICMP4_ERROR && - next0 != UDP_INPUT_NEXT_ICMP6_ERROR); - } - } - if (PREDICT_FALSE (b1->flags & VLIB_BUFFER_IS_TRACED)) - { - udp_rx_trace_t *tr = vlib_add_trace (vm, node, - b1, sizeof (*tr)); - if (b1->error != node->errors[UDP_ERROR_LENGTH_ERROR]) - { - tr->src_port = h1 ? h1->src_port : 0; - tr->dst_port = h1 ? h1->dst_port : 0; - tr->bound = (next1 != UDP_INPUT_NEXT_ICMP4_ERROR && - next1 != UDP_INPUT_NEXT_ICMP6_ERROR); - } - } - - vlib_validate_buffer_enqueue_x2 (vm, node, next_index, - to_next, n_left_to_next, - bi0, bi1, next0, next1); - } - - while (n_left_from > 0 && n_left_to_next > 0) - { - u32 bi0; - vlib_buffer_t *b0; - udp_header_t *h0 = 0; - u32 i0, next0; - u32 advance0; - - bi0 = from[0]; - to_next[0] = bi0; - from += 1; - to_next += 1; - n_left_from -= 1; - n_left_to_next -= 1; - - b0 = vlib_get_buffer (vm, bi0); - - /* ip4/6_local hands us the ip header, not the udp header */ - if (is_ip4) - advance0 = sizeof (ip4_header_t); - else - advance0 = sizeof (ip6_header_t); - - if (PREDICT_FALSE (b0->current_length < advance0 + sizeof (*h0))) - { - b0->error = node->errors[UDP_ERROR_LENGTH_ERROR]; - next0 = UDP_INPUT_NEXT_DROP; - goto trace_x1; - } - - vlib_buffer_advance (b0, advance0); - - h0 = vlib_buffer_get_current (b0); - - if (PREDICT_TRUE (clib_net_to_host_u16 (h0->length) <= - vlib_buffer_length_in_chain (vm, b0))) - { - i0 = sparse_vec_index (rt->next_by_dst_port, h0->dst_port); - next0 = vec_elt (rt->next_by_dst_port, i0); - - if (PREDICT_FALSE (i0 == SPARSE_VEC_INVALID_INDEX)) - { - // move the pointer back so icmp-error can find the - // ip packet header - vlib_buffer_advance (b0, -(word) advance0); - - if (PREDICT_FALSE (punt_unknown)) - { - b0->error = node->errors[UDP_ERROR_PUNT]; - next0 = UDP_INPUT_NEXT_PUNT; - } - else if (is_ip4) - { - icmp4_error_set_vnet_buffer (b0, - ICMP4_destination_unreachable, - ICMP4_destination_unreachable_port_unreachable, - 0); - next0 = UDP_INPUT_NEXT_ICMP4_ERROR; - n_no_listener++; - } - else - { - icmp6_error_set_vnet_buffer (b0, - ICMP6_destination_unreachable, - ICMP6_destination_unreachable_port_unreachable, - 0); - next0 = UDP_INPUT_NEXT_ICMP6_ERROR; - n_no_listener++; - } - } - else - { - b0->error = node->errors[UDP_ERROR_NONE]; - // advance to the payload - vlib_buffer_advance (b0, sizeof (*h0)); - } - } - else - { - b0->error = node->errors[UDP_ERROR_LENGTH_ERROR]; - next0 = UDP_INPUT_NEXT_DROP; - } - - trace_x1: - if (PREDICT_FALSE (b0->flags & VLIB_BUFFER_IS_TRACED)) - { - udp_rx_trace_t *tr = vlib_add_trace (vm, node, - b0, sizeof (*tr)); - if (b0->error != node->errors[UDP_ERROR_LENGTH_ERROR]) - { - tr->src_port = h0->src_port; - tr->dst_port = h0->dst_port; - tr->bound = (next0 != UDP_INPUT_NEXT_ICMP4_ERROR && - next0 != UDP_INPUT_NEXT_ICMP6_ERROR); - } - } - - vlib_validate_buffer_enqueue_x1 (vm, node, next_index, - to_next, n_left_to_next, - bi0, next0); - } - - vlib_put_next_frame (vm, node, next_index, n_left_to_next); - } - vlib_error_count (vm, node->node_index, UDP_ERROR_NO_LISTENER, - n_no_listener); - return from_frame->n_vectors; -} - -static char *udp_error_strings[] = { -#define udp_error(n,s) s, -#include "udp_error.def" -#undef udp_error -}; - -static uword -udp4_input (vlib_main_t * vm, - vlib_node_runtime_t * node, vlib_frame_t * from_frame) -{ - return udp46_input_inline (vm, node, from_frame, 1 /* is_ip4 */ ); -} - -static uword -udp6_input (vlib_main_t * vm, - vlib_node_runtime_t * node, vlib_frame_t * from_frame) -{ - return udp46_input_inline (vm, node, from_frame, 0 /* is_ip4 */ ); -} - - -/* *INDENT-OFF* */ -VLIB_REGISTER_NODE (udp4_input_node) = { - .function = udp4_input, - .name = "ip4-udp-lookup", - /* Takes a vector of packets. */ - .vector_size = sizeof (u32), - - .runtime_data_bytes = sizeof (udp_input_runtime_t), - - .n_errors = UDP_N_ERROR, - .error_strings = udp_error_strings, - - .n_next_nodes = UDP_INPUT_N_NEXT, - .next_nodes = { -#define _(s,n) [UDP_INPUT_NEXT_##s] = n, - foreach_udp_input_next -#undef _ - }, - - .format_buffer = format_udp_header, - .format_trace = format_udp_rx_trace, - .unformat_buffer = unformat_udp_header, -}; -/* *INDENT-ON* */ - -VLIB_NODE_FUNCTION_MULTIARCH (udp4_input_node, udp4_input); - -/* *INDENT-OFF* */ -VLIB_REGISTER_NODE (udp6_input_node) = { - .function = udp6_input, - .name = "ip6-udp-lookup", - /* Takes a vector of packets. */ - .vector_size = sizeof (u32), - - .runtime_data_bytes = sizeof (udp_input_runtime_t), - - .n_errors = UDP_N_ERROR, - .error_strings = udp_error_strings, - - .n_next_nodes = UDP_INPUT_N_NEXT, - .next_nodes = { -#define _(s,n) [UDP_INPUT_NEXT_##s] = n, - foreach_udp_input_next -#undef _ - }, - - .format_buffer = format_udp_header, - .format_trace = format_udp_rx_trace, - .unformat_buffer = unformat_udp_header, -}; -/* *INDENT-ON* */ - -VLIB_NODE_FUNCTION_MULTIARCH (udp6_input_node, udp6_input); - -static void -add_dst_port (udp_main_t * um, - udp_dst_port_t dst_port, char *dst_port_name, u8 is_ip4) -{ - udp_dst_port_info_t *pi; - u32 i; - - vec_add2 (um->dst_port_infos[is_ip4], pi, 1); - i = pi - um->dst_port_infos[is_ip4]; - - pi->name = dst_port_name; - pi->dst_port = dst_port; - pi->next_index = pi->node_index = ~0; - - hash_set (um->dst_port_info_by_dst_port[is_ip4], dst_port, i); - - if (pi->name) - hash_set_mem (um->dst_port_info_by_name[is_ip4], pi->name, i); -} - -void -udp_register_dst_port (vlib_main_t * vm, - udp_dst_port_t dst_port, u32 node_index, u8 is_ip4) -{ - udp_main_t *um = &udp_main; - udp_dst_port_info_t *pi; - udp_input_runtime_t *rt; - u16 *n; - - { - clib_error_t *error = vlib_call_init_function (vm, udp_local_init); - if (error) - clib_error_report (error); - } - - pi = udp_get_dst_port_info (um, dst_port, is_ip4); - if (!pi) - { - add_dst_port (um, dst_port, 0, is_ip4); - pi = udp_get_dst_port_info (um, dst_port, is_ip4); - ASSERT (pi); - } - - pi->node_index = node_index; - pi->next_index = vlib_node_add_next (vm, - is_ip4 ? udp4_input_node.index - : udp6_input_node.index, node_index); - - /* Setup udp protocol -> next index sparse vector mapping. */ - rt = vlib_node_get_runtime_data - (vm, is_ip4 ? udp4_input_node.index : udp6_input_node.index); - n = sparse_vec_validate (rt->next_by_dst_port, - clib_host_to_net_u16 (dst_port)); - n[0] = pi->next_index; -} - -void -udp_punt_unknown (vlib_main_t * vm, u8 is_ip4, u8 is_add) -{ - udp_input_runtime_t *rt; - - { - clib_error_t *error = vlib_call_init_function (vm, udp_local_init); - if (error) - clib_error_report (error); - } - - rt = vlib_node_get_runtime_data - (vm, is_ip4 ? udp4_input_node.index : udp6_input_node.index); - - rt->punt_unknown = is_add; -} - -/* Parse a UDP header. */ -uword -unformat_udp_header (unformat_input_t * input, va_list * args) -{ - u8 **result = va_arg (*args, u8 **); - udp_header_t *udp; - __attribute__ ((unused)) int old_length; - u16 src_port, dst_port; - - /* Allocate space for IP header. */ - { - void *p; - - old_length = vec_len (*result); - vec_add2 (*result, p, sizeof (ip4_header_t)); - udp = p; - } - - memset (udp, 0, sizeof (udp[0])); - if (unformat (input, "src-port %d dst-port %d", &src_port, &dst_port)) - { - udp->src_port = clib_host_to_net_u16 (src_port); - udp->dst_port = clib_host_to_net_u16 (dst_port); - return 1; - } - return 0; -} - -static void -udp_setup_node (vlib_main_t * vm, u32 node_index) -{ - vlib_node_t *n = vlib_get_node (vm, node_index); - pg_node_t *pn = pg_get_node (node_index); - - n->format_buffer = format_udp_header; - n->unformat_buffer = unformat_udp_header; - pn->unformat_edit = unformat_pg_udp_header; -} - -clib_error_t * -udp_local_init (vlib_main_t * vm) -{ - udp_input_runtime_t *rt; - udp_main_t *um = &udp_main; - int i; - - { - clib_error_t *error; - error = vlib_call_init_function (vm, udp_init); - if (error) - clib_error_report (error); - } - - - for (i = 0; i < 2; i++) - { - um->dst_port_info_by_name[i] = hash_create_string (0, sizeof (uword)); - um->dst_port_info_by_dst_port[i] = hash_create (0, sizeof (uword)); - } - - udp_setup_node (vm, udp4_input_node.index); - udp_setup_node (vm, udp6_input_node.index); - - rt = vlib_node_get_runtime_data (vm, udp4_input_node.index); - - rt->next_by_dst_port = sparse_vec_new - ( /* elt bytes */ sizeof (rt->next_by_dst_port[0]), - /* bits in index */ BITS (((udp_header_t *) 0)->dst_port)); - - rt->punt_unknown = 0; - -#define _(n,s) add_dst_port (um, UDP_DST_PORT_##s, #s, 1 /* is_ip4 */); - foreach_udp4_dst_port -#undef _ - rt = vlib_node_get_runtime_data (vm, udp6_input_node.index); - - rt->next_by_dst_port = sparse_vec_new - ( /* elt bytes */ sizeof (rt->next_by_dst_port[0]), - /* bits in index */ BITS (((udp_header_t *) 0)->dst_port)); - - rt->punt_unknown = 0; - -#define _(n,s) add_dst_port (um, UDP_DST_PORT_##s, #s, 0 /* is_ip4 */); - foreach_udp6_dst_port -#undef _ - ip4_register_protocol (IP_PROTOCOL_UDP, udp4_input_node.index); - /* Note: ip6 differs from ip4, UDP is hotwired to ip6-udp-lookup */ - return 0; -} - -VLIB_INIT_FUNCTION (udp_local_init); - -/* - * fd.io coding-style-patch-verification: ON - * - * Local Variables: - * eval: (c-set-style "gnu") - * End: - */ diff --git a/src/vnet/ip/udp_packet.h b/src/vnet/ip/udp_packet.h deleted file mode 100644 index beea3059..00000000 --- a/src/vnet/ip/udp_packet.h +++ /dev/null @@ -1,65 +0,0 @@ -/* - * Copyright (c) 2015 Cisco and/or its affiliates. - * Licensed under the Apache License, Version 2.0 (the "License"); - * you may not use this file except in compliance with the License. - * You may obtain a copy of the License at: - * - * http://www.apache.org/licenses/LICENSE-2.0 - * - * Unless required by applicable law or agreed to in writing, software - * distributed under the License is distributed on an "AS IS" BASIS, - * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. - * See the License for the specific language governing permissions and - * limitations under the License. - */ -/* - * ip4/udp_packet.h: UDP packet format - * - * Copyright (c) 2008 Eliot Dresselhaus - * - * Permission is hereby granted, free of charge, to any person obtaining - * a copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sublicense, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice shall be - * included in all copies or substantial portions of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, - * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND - * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE - * LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION - * OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION - * WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - */ - -#ifndef included_udp_packet_h -#define included_udp_packet_h - -typedef struct -{ - /* Source and destination port. */ - u16 src_port, dst_port; - - /* Length of UDP header plus payload. */ - u16 length; - - /* Checksum of UDP pseudo-header and data or - zero if checksum is disabled. */ - u16 checksum; -} udp_header_t; - -#endif /* included_udp_packet_h */ - - -/* - * fd.io coding-style-patch-verification: ON - * - * Local Variables: - * eval: (c-set-style "gnu") - * End: - */ diff --git a/src/vnet/ip/udp_pg.c b/src/vnet/ip/udp_pg.c deleted file mode 100644 index c9d8d38c..00000000 --- a/src/vnet/ip/udp_pg.c +++ /dev/null @@ -1,237 +0,0 @@ -/* - * Copyright (c) 2015 Cisco and/or its affiliates. - * Licensed under the Apache License, Version 2.0 (the "License"); - * you may not use this file except in compliance with the License. - * You may obtain a copy of the License at: - * - * http://www.apache.org/licenses/LICENSE-2.0 - * - * Unless required by applicable law or agreed to in writing, software - * distributed under the License is distributed on an "AS IS" BASIS, - * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. - * See the License for the specific language governing permissions and - * limitations under the License. - */ -/* - * ip/udp_pg: UDP packet-generator interface - * - * Copyright (c) 2008 Eliot Dresselhaus - * - * Permission is hereby granted, free of charge, to any person obtaining - * a copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sublicense, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice shall be - * included in all copies or substantial portions of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, - * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND - * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE - * LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION - * OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION - * WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - */ - -#include -#include /* for unformat_udp_udp_port */ - -#define UDP_PG_EDIT_LENGTH (1 << 0) -#define UDP_PG_EDIT_CHECKSUM (1 << 1) - -always_inline void -udp_pg_edit_function_inline (pg_main_t * pg, - pg_stream_t * s, - pg_edit_group_t * g, - u32 * packets, u32 n_packets, u32 flags) -{ - vlib_main_t *vm = vlib_get_main (); - u32 ip_offset, udp_offset; - - udp_offset = g->start_byte_offset; - ip_offset = (g - 1)->start_byte_offset; - - while (n_packets >= 1) - { - vlib_buffer_t *p0; - ip4_header_t *ip0; - udp_header_t *udp0; - u32 udp_len0; - - p0 = vlib_get_buffer (vm, packets[0]); - n_packets -= 1; - packets += 1; - - ip0 = (void *) (p0->data + ip_offset); - udp0 = (void *) (p0->data + udp_offset); - udp_len0 = clib_net_to_host_u16 (ip0->length) - sizeof (ip0[0]); - - if (flags & UDP_PG_EDIT_LENGTH) - udp0->length = - clib_net_to_host_u16 (vlib_buffer_length_in_chain (vm, p0) - - ip_offset); - - /* Initialize checksum with header. */ - if (flags & UDP_PG_EDIT_CHECKSUM) - { - ip_csum_t sum0; - - sum0 = clib_mem_unaligned (&ip0->src_address, u64); - - sum0 = ip_csum_with_carry - (sum0, clib_host_to_net_u32 (udp_len0 + (ip0->protocol << 16))); - - /* Invalidate possibly old checksum. */ - udp0->checksum = 0; - - sum0 = - ip_incremental_checksum_buffer (vm, p0, udp_offset, udp_len0, - sum0); - - sum0 = ~ip_csum_fold (sum0); - - /* Zero checksum means checksumming disabled. */ - sum0 = sum0 != 0 ? sum0 : 0xffff; - - udp0->checksum = sum0; - } - } -} - -static void -udp_pg_edit_function (pg_main_t * pg, - pg_stream_t * s, - pg_edit_group_t * g, u32 * packets, u32 n_packets) -{ - switch (g->edit_function_opaque) - { - case UDP_PG_EDIT_LENGTH: - udp_pg_edit_function_inline (pg, s, g, packets, n_packets, - UDP_PG_EDIT_LENGTH); - break; - - case UDP_PG_EDIT_CHECKSUM: - udp_pg_edit_function_inline (pg, s, g, packets, n_packets, - UDP_PG_EDIT_CHECKSUM); - break; - - case UDP_PG_EDIT_CHECKSUM | UDP_PG_EDIT_LENGTH: - udp_pg_edit_function_inline (pg, s, g, packets, n_packets, - UDP_PG_EDIT_CHECKSUM | UDP_PG_EDIT_LENGTH); - break; - - default: - ASSERT (0); - break; - } -} - -typedef struct -{ - pg_edit_t src_port, dst_port; - pg_edit_t length; - pg_edit_t checksum; -} pg_udp_header_t; - -static inline void -pg_udp_header_init (pg_udp_header_t * p) -{ - /* Initialize fields that are not bit fields in the IP header. */ -#define _(f) pg_edit_init (&p->f, udp_header_t, f); - _(src_port); - _(dst_port); - _(length); - _(checksum); -#undef _ -} - -uword -unformat_pg_udp_header (unformat_input_t * input, va_list * args) -{ - pg_stream_t *s = va_arg (*args, pg_stream_t *); - pg_udp_header_t *p; - u32 group_index; - - p = pg_create_edit_group (s, sizeof (p[0]), sizeof (udp_header_t), - &group_index); - pg_udp_header_init (p); - - /* Defaults. */ - p->checksum.type = PG_EDIT_UNSPECIFIED; - p->length.type = PG_EDIT_UNSPECIFIED; - - if (!unformat (input, "UDP: %U -> %U", - unformat_pg_edit, - unformat_tcp_udp_port, &p->src_port, - unformat_pg_edit, unformat_tcp_udp_port, &p->dst_port)) - goto error; - - /* Parse options. */ - while (1) - { - if (unformat (input, "length %U", - unformat_pg_edit, unformat_pg_number, &p->length)) - ; - - else if (unformat (input, "checksum %U", - unformat_pg_edit, unformat_pg_number, &p->checksum)) - ; - - /* Can't parse input: try next protocol level. */ - else - break; - } - - { - ip_main_t *im = &ip_main; - u16 dst_port; - tcp_udp_port_info_t *pi; - - pi = 0; - if (p->dst_port.type == PG_EDIT_FIXED) - { - dst_port = pg_edit_get_value (&p->dst_port, PG_EDIT_LO); - pi = ip_get_tcp_udp_port_info (im, dst_port); - } - - if (pi && pi->unformat_pg_edit - && unformat_user (input, pi->unformat_pg_edit, s)) - ; - - else if (!unformat_user (input, unformat_pg_payload, s)) - goto error; - - p = pg_get_edit_group (s, group_index); - if (p->checksum.type == PG_EDIT_UNSPECIFIED - || p->length.type == PG_EDIT_UNSPECIFIED) - { - pg_edit_group_t *g = pg_stream_get_group (s, group_index); - g->edit_function = udp_pg_edit_function; - g->edit_function_opaque = 0; - if (p->checksum.type == PG_EDIT_UNSPECIFIED) - g->edit_function_opaque |= UDP_PG_EDIT_CHECKSUM; - if (p->length.type == PG_EDIT_UNSPECIFIED) - g->edit_function_opaque |= UDP_PG_EDIT_LENGTH; - } - - return 1; - } - -error: - /* Free up any edits we may have added. */ - pg_free_edit_group (s); - return 0; -} - - -/* - * fd.io coding-style-patch-verification: ON - * - * Local Variables: - * eval: (c-set-style "gnu") - * End: - */ diff --git a/src/vnet/ipsec/ikev2.c b/src/vnet/ipsec/ikev2.c index 09209334..2c1074d8 100644 --- a/src/vnet/ipsec/ikev2.c +++ b/src/vnet/ipsec/ikev2.c @@ -17,7 +17,7 @@ #include #include #include -#include +#include #include #include #include diff --git a/src/vnet/ipsec/ikev2_cli.c b/src/vnet/ipsec/ikev2_cli.c index 5c88d8d4..05ed4e60 100644 --- a/src/vnet/ipsec/ikev2_cli.c +++ b/src/vnet/ipsec/ikev2_cli.c @@ -16,7 +16,7 @@ #include #include #include -#include +#include #include #include diff --git a/src/vnet/ipsec/ikev2_crypto.c b/src/vnet/ipsec/ikev2_crypto.c index c201d3eb..ca56158f 100644 --- a/src/vnet/ipsec/ikev2_crypto.c +++ b/src/vnet/ipsec/ikev2_crypto.c @@ -17,7 +17,7 @@ #include #include #include -#include +#include #include #include #include diff --git a/src/vnet/lisp-cp/packets.c b/src/vnet/lisp-cp/packets.c index 3a4f421b..f24024f1 100644 --- a/src/vnet/lisp-cp/packets.c +++ b/src/vnet/lisp-cp/packets.c @@ -15,7 +15,7 @@ #include #include -#include +#include /* Returns IP ID for the packet */ /* static u16 ip_id = 0; @@ -141,61 +141,6 @@ pkt_push_udp (vlib_main_t * vm, vlib_buffer_t * b, u16 sp, u16 dp) return uh; } -void * -pkt_push_ipv4 (vlib_main_t * vm, vlib_buffer_t * b, ip4_address_t * src, - ip4_address_t * dst, int proto) -{ - ip4_header_t *ih; - - /* make some room */ - ih = vlib_buffer_push_uninit (b, sizeof (ip4_header_t)); - - ih->ip_version_and_header_length = 0x45; - ih->tos = 0; - ih->length = clib_host_to_net_u16 (vlib_buffer_length_in_chain (vm, b)); - - /* iph->fragment_id = clib_host_to_net_u16(get_IP_ID ()); */ - - /* TODO: decide if we allow fragments in case of control */ - ih->flags_and_fragment_offset = clib_host_to_net_u16 (IP_DF); - ih->ttl = 255; - ih->protocol = proto; - ih->src_address.as_u32 = src->as_u32; - ih->dst_address.as_u32 = dst->as_u32; - - ih->checksum = ip4_header_checksum (ih); - return ih; -} - -void * -pkt_push_ipv6 (vlib_main_t * vm, vlib_buffer_t * b, ip6_address_t * src, - ip6_address_t * dst, int proto) -{ - ip6_header_t *ip6h; - u16 payload_length; - - /* make some room */ - ip6h = vlib_buffer_push_uninit (b, sizeof (ip6_header_t)); - - ip6h->ip_version_traffic_class_and_flow_label = - clib_host_to_net_u32 (0x6 << 28); - - /* calculate ip6 payload length */ - payload_length = vlib_buffer_length_in_chain (vm, b); - payload_length -= sizeof (*ip6h); - - ip6h->payload_length = clib_host_to_net_u16 (payload_length); - - ip6h->hop_limit = 0xff; - ip6h->protocol = proto; - clib_memcpy (ip6h->src_address.as_u8, src->as_u8, - sizeof (ip6h->src_address)); - clib_memcpy (ip6h->dst_address.as_u8, dst->as_u8, - sizeof (ip6h->src_address)); - - return ip6h; -} - void * pkt_push_ip (vlib_main_t * vm, vlib_buffer_t * b, ip_address_t * src, ip_address_t * dst, u32 proto) @@ -210,12 +155,12 @@ pkt_push_ip (vlib_main_t * vm, vlib_buffer_t * b, ip_address_t * src, switch (ip_addr_version (src)) { case IP4: - return pkt_push_ipv4 (vm, b, &ip_addr_v4 (src), &ip_addr_v4 (dst), - proto); + return vlib_buffer_push_ip4 (vm, b, &ip_addr_v4 (src), + &ip_addr_v4 (dst), proto); break; case IP6: - return pkt_push_ipv6 (vm, b, &ip_addr_v6 (src), &ip_addr_v6 (dst), - proto); + return vlib_buffer_push_ip6 (vm, b, &ip_addr_v6 (src), + &ip_addr_v6 (dst), proto); break; } diff --git a/src/vnet/lisp-cp/packets.h b/src/vnet/lisp-cp/packets.h index 212a1d78..f6da3bf4 100644 --- a/src/vnet/lisp-cp/packets.h +++ b/src/vnet/lisp-cp/packets.h @@ -26,51 +26,6 @@ void *pkt_push_udp_and_ip (vlib_main_t * vm, vlib_buffer_t * b, u16 sp, void *pkt_push_ecm_hdr (vlib_buffer_t * b); -always_inline u8 * -vlib_buffer_get_tail (vlib_buffer_t * b) -{ - return b->data + b->current_data + b->current_length; -} - -always_inline void * -vlib_buffer_put_uninit (vlib_buffer_t * b, u8 size) -{ - /* XXX should make sure there's enough space! */ - void *p = vlib_buffer_get_tail (b); - b->current_length += size; - return p; -} - -always_inline void * -vlib_buffer_push_uninit (vlib_buffer_t * b, u8 size) -{ - /* XXX should make sure there's enough space! */ - ASSERT (b->current_data >= size); - b->current_data -= size; - b->current_length += size; - - return vlib_buffer_get_current (b); -} - -always_inline void * -vlib_buffer_make_headroom (vlib_buffer_t * b, u8 size) -{ - /* XXX should make sure there's enough space! */ - b->current_data += size; - return vlib_buffer_get_current (b); -} - -always_inline void * -vlib_buffer_pull (vlib_buffer_t * b, u8 size) -{ - if (b->current_length < size) - return 0; - - void *data = vlib_buffer_get_current (b); - vlib_buffer_advance (b, size); - return data; -} - /* *INDENT-ON* */ /* diff --git a/src/vnet/lisp-gpe/interface.c b/src/vnet/lisp-gpe/interface.c index 13359277..292c7e6a 100644 --- a/src/vnet/lisp-gpe/interface.c +++ b/src/vnet/lisp-gpe/interface.c @@ -23,7 +23,7 @@ #include #include #include -#include +#include #include #include #include diff --git a/src/vnet/lisp-gpe/lisp_gpe.h b/src/vnet/lisp-gpe/lisp_gpe.h index c898a7da..b5a50ec6 100644 --- a/src/vnet/lisp-gpe/lisp_gpe.h +++ b/src/vnet/lisp-gpe/lisp_gpe.h @@ -27,10 +27,12 @@ #include #include #include -#include +#include #include #include #include +#include +#include /** IP4-UDP-LISP encap header */ /* *INDENT-OFF* */ diff --git a/src/vnet/lisp-gpe/lisp_gpe_adjacency.c b/src/vnet/lisp-gpe/lisp_gpe_adjacency.c index 65006b81..dbcf7134 100644 --- a/src/vnet/lisp-gpe/lisp_gpe_adjacency.c +++ b/src/vnet/lisp-gpe/lisp_gpe_adjacency.c @@ -25,6 +25,8 @@ #include #include #include +#include +#include /** * Memory pool of all adjacencies diff --git a/src/vnet/session/application.c b/src/vnet/session/application.c new file mode 100644 index 00000000..a561e7d1 --- /dev/null +++ b/src/vnet/session/application.c @@ -0,0 +1,343 @@ +/* + * Copyright (c) 2017 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +#include +#include + +/* + * Pool from which we allocate all applications + */ +static application_t *app_pool; + +/* + * Hash table of apps by api client index + */ +static uword *app_by_api_client_index; + +int +application_api_queue_is_full (application_t * app) +{ + unix_shared_memory_queue_t *q; + + /* builtin servers are always OK */ + if (app->api_client_index == ~0) + return 0; + + q = vl_api_client_index_to_input_queue (app->api_client_index); + if (!q) + return 1; + + if (q->cursize == q->maxsize) + return 1; + return 0; +} + +static void +application_table_add (application_t * app) +{ + hash_set (app_by_api_client_index, app->api_client_index, app->index); +} + +static void +application_table_del (application_t * app) +{ + hash_unset (app_by_api_client_index, app->api_client_index); +} + +application_t * +application_lookup (u32 api_client_index) +{ + uword *p; + p = hash_get (app_by_api_client_index, api_client_index); + if (p) + return application_get (p[0]); + + return 0; +} + +void +application_del (application_t * app) +{ + session_manager_main_t *smm = vnet_get_session_manager_main (); + api_main_t *am = &api_main; + void *oldheap; + session_manager_t *sm; + + if (app->mode == APP_SERVER) + { + sm = session_manager_get (app->session_manager_index); + session_manager_del (smm, sm); + } + + /* Free the event fifo in the /vpe-api shared-memory segment */ + oldheap = svm_push_data_heap (am->vlib_rp); + if (app->event_queue) + unix_shared_memory_queue_free (app->event_queue); + svm_pop_heap (oldheap); + + application_table_del (app); + + pool_put (app_pool, app); +} + +application_t * +application_new (application_type_t type, session_type_t sst, + u32 api_client_index, u32 flags, session_cb_vft_t * cb_fns) +{ + session_manager_main_t *smm = vnet_get_session_manager_main (); + api_main_t *am = &api_main; + application_t *app; + void *oldheap; + session_manager_t *sm; + + pool_get (app_pool, app); + memset (app, 0, sizeof (*app)); + + /* Allocate event fifo in the /vpe-api shared-memory segment */ + oldheap = svm_push_data_heap (am->vlib_rp); + + /* Allocate server event queue */ + app->event_queue = + unix_shared_memory_queue_init (128 /* nels $$$$ config */ , + sizeof (session_fifo_event_t), + 0 /* consumer pid */ , + 0 + /* (do not) signal when queue non-empty */ + ); + + svm_pop_heap (oldheap); + + /* If a server, allocate session manager */ + if (type == APP_SERVER) + { + pool_get (smm->session_managers, sm); + memset (sm, 0, sizeof (*sm)); + + app->session_manager_index = sm - smm->session_managers; + } + else if (type == APP_CLIENT) + { + /* Allocate connect session manager if needed */ + if (smm->connect_manager_index[sst] == INVALID_INDEX) + connects_session_manager_init (smm, sst); + app->session_manager_index = smm->connect_manager_index[sst]; + } + + app->mode = type; + app->index = application_get_index (app); + app->session_type = sst; + app->api_client_index = api_client_index; + app->flags = flags; + app->cb_fns = *cb_fns; + + /* Add app to lookup by api_client_index table */ + application_table_add (app); + + return app; +} + +application_t * +application_get (u32 index) +{ + return pool_elt_at_index (app_pool, index); +} + +u32 +application_get_index (application_t * app) +{ + return app - app_pool; +} + +int +application_server_init (application_t * server, u32 segment_size, + u32 add_segment_size, u32 rx_fifo_size, + u32 tx_fifo_size, u8 ** segment_name) +{ + session_manager_main_t *smm = vnet_get_session_manager_main (); + session_manager_t *sm; + int rv; + + sm = session_manager_get (server->session_manager_index); + + /* Add first segment */ + if ((rv = session_manager_add_first_segment (smm, sm, segment_size, + segment_name))) + { + return rv; + } + + /* Setup session manager */ + sm->add_segment_size = add_segment_size; + sm->rx_fifo_size = rx_fifo_size; + sm->tx_fifo_size = tx_fifo_size; + sm->add_segment = sm->add_segment_size != 0; + return 0; +} + +u8 * +format_application_server (u8 * s, va_list * args) +{ + application_t *srv = va_arg (*args, application_t *); + int verbose = va_arg (*args, int); + vl_api_registration_t *regp; + stream_session_t *listener; + u8 *server_name, *str, *seg_name; + u32 segment_size; + + if (srv == 0) + { + if (verbose) + s = format (s, "%-40s%-20s%-15s%-15s%-10s", "Connection", "Server", + "Segment", "API Client", "Cookie"); + else + s = format (s, "%-40s%-20s", "Connection", "Server"); + + return s; + } + + regp = vl_api_client_index_to_registration (srv->api_client_index); + if (!regp) + server_name = format (0, "%s%c", regp->name, 0); + else + server_name = regp->name; + + listener = stream_session_listener_get (srv->session_type, + srv->session_index); + str = format (0, "%U", format_stream_session, listener, verbose); + + session_manager_get_segment_info (listener->server_segment_index, &seg_name, + &segment_size); + if (verbose) + { + s = format (s, "%-40s%-20s%-20s%-10d%-10d", str, server_name, + seg_name, srv->api_client_index, srv->accept_cookie); + } + else + s = format (s, "%-40s%-20s", str, server_name); + return s; +} + +u8 * +format_application_client (u8 * s, va_list * args) +{ + application_t *client = va_arg (*args, application_t *); + int verbose = va_arg (*args, int); + stream_session_t *session; + u8 *str, *seg_name; + u32 segment_size; + + if (client == 0) + { + if (verbose) + s = + format (s, "%-40s%-20s%-10s", "Connection", "Segment", + "API Client"); + else + s = format (s, "%-40s", "Connection"); + + return s; + } + + session = stream_session_get (client->session_index, client->thread_index); + str = format (0, "%U", format_stream_session, session, verbose); + + session_manager_get_segment_info (session->server_segment_index, &seg_name, + &segment_size); + if (verbose) + { + s = format (s, "%-40s%-20s%-10d%", str, seg_name, + client->api_client_index); + } + else + s = format (s, "%-40s", str); + return s; +} + +static clib_error_t * +show_app_command_fn (vlib_main_t * vm, unformat_input_t * input, + vlib_cli_command_t * cmd) +{ + application_t *app; + int do_server = 0; + int do_client = 0; + int verbose = 0; + + while (unformat_check_input (input) != UNFORMAT_END_OF_INPUT) + { + if (unformat (input, "server")) + do_server = 1; + else if (unformat (input, "client")) + do_client = 1; + else if (unformat (input, "verbose")) + verbose = 1; + else + break; + } + + if (do_server) + { + if (pool_elts (app_pool)) + { + vlib_cli_output (vm, "%U", format_application_server, + 0 /* header */ , + verbose); + /* *INDENT-OFF* */ + pool_foreach (app, app_pool, + ({ + if (app->mode == APP_SERVER) + vlib_cli_output (vm, "%U", format_application_server, app, + verbose); + })); + /* *INDENT-ON* */ + } + else + vlib_cli_output (vm, "No active server bindings"); + } + + if (do_client) + { + if (pool_elts (app_pool)) + { + vlib_cli_output (vm, "%U", format_application_client, + 0 /* header */ , + verbose); + /* *INDENT-OFF* */ + pool_foreach (app, app_pool, + ({ + if (app->mode == APP_CLIENT) + vlib_cli_output (vm, "%U", format_application_client, app, + verbose); + })); + /* *INDENT-ON* */ + } + else + vlib_cli_output (vm, "No active server bindings"); + } + + return 0; +} + +VLIB_CLI_COMMAND (show_app_command, static) = +{ +.path = "show app",.short_help = + "show app [server|client] [verbose]",.function = show_app_command_fn,}; + +/* + * fd.io coding-style-patch-verification: ON + * + * Local Variables: + * eval: (c-set-style "gnu") + * End: + */ diff --git a/src/vnet/session/application.h b/src/vnet/session/application.h new file mode 100644 index 00000000..027d6967 --- /dev/null +++ b/src/vnet/session/application.h @@ -0,0 +1,120 @@ +/* + * Copyright (c) 2017 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +#ifndef SRC_VNET_SESSION_APPLICATION_H_ +#define SRC_VNET_SESSION_APPLICATION_H_ + +#include +#include + +typedef enum +{ + APP_SERVER, + APP_CLIENT +} application_type_t; + +typedef struct _stream_session_cb_vft +{ + /** Notify server of new segment */ + int (*add_segment_callback) (u32 api_client_index, const u8 * seg_name, + u32 seg_size); + + /** Notify server of newly accepted session */ + int (*session_accept_callback) (stream_session_t * new_session); + + /* Connection request callback */ + int (*session_connected_callback) (u32 api_client_index, + stream_session_t * s, u8 code); + + /** Notify app that session is closing */ + void (*session_disconnect_callback) (stream_session_t * s); + + /** Notify app that session was reset */ + void (*session_reset_callback) (stream_session_t * s); + + /* Direct RX callback, for built-in servers */ + int (*builtin_server_rx_callback) (stream_session_t * session); + + /* Redirect connection to local server */ + int (*redirect_connect_callback) (u32 api_client_index, void *mp); +} session_cb_vft_t; + +typedef struct _application +{ + /** Index in server pool */ + u32 index; + + /** Flags */ + u32 flags; + + /** Binary API connection index, ~0 if internal */ + u32 api_client_index; + + /* */ + u32 api_context; + + /** Application listens for events on this svm queue */ + unix_shared_memory_queue_t *event_queue; + + /** Stream session type */ + u8 session_type; + + /* Stream server mode: accept or connect */ + u8 mode; + + u32 session_manager_index; + + /* + * Bind/Listen specific + */ + + /** Accept cookie, for multiple session flavors ($$$ maybe) */ + u32 accept_cookie; + + /** Index of the listen session or connect session */ + u32 session_index; + + /** Session thread index for client connect sessions */ + u32 thread_index; + + /* + * Callbacks: shoulder-taps for the server/client + */ + session_cb_vft_t cb_fns; +} application_t; + +application_t *application_new (application_type_t type, session_type_t sst, + u32 api_client_index, u32 flags, + session_cb_vft_t * cb_fns); +void application_del (application_t * app); +application_t *application_get (u32 index); +application_t *application_lookup (u32 api_client_index); +u32 application_get_index (application_t * app); + +int +application_server_init (application_t * server, u32 segment_size, + u32 add_segment_size, u32 rx_fifo_size, + u32 tx_fifo_size, u8 ** segment_name); +int application_api_queue_is_full (application_t * app); + +#endif /* SRC_VNET_SESSION_APPLICATION_H_ */ + +/* + * fd.io coding-style-patch-verification: ON + * + * Local Variables: + * eval: (c-set-style "gnu") + * End: + */ diff --git a/src/vnet/session/application_interface.c b/src/vnet/session/application_interface.c new file mode 100644 index 00000000..0ea77fd8 --- /dev/null +++ b/src/vnet/session/application_interface.c @@ -0,0 +1,459 @@ +/* + * Copyright (c) 2016 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ +#include + +#include +#include +#include +#include + +/** @file + VPP's application/session API bind/unbind/connect/disconnect calls +*/ + +static u8 +ip_is_zero (ip46_address_t * ip46_address, u8 is_ip4) +{ + if (is_ip4) + return (ip46_address->ip4.as_u32 == 0); + else + return (ip46_address->as_u64[0] == 0 && ip46_address->as_u64[1] == 0); +} + +static u8 +ip_is_local (ip46_address_t * ip46_address, u8 is_ip4) +{ + fib_node_index_t fei; + fib_entry_flag_t flags; + fib_prefix_t prefix; + + /* Check if requester is local */ + if (is_ip4) + { + prefix.fp_len = 32; + prefix.fp_proto = FIB_PROTOCOL_IP4; + } + else + { + prefix.fp_len = 128; + prefix.fp_proto = FIB_PROTOCOL_IP6; + } + + clib_memcpy (&prefix.fp_addr, ip46_address, sizeof (ip46_address)); + fei = fib_table_lookup (0, &prefix); + flags = fib_entry_get_flags (fei); + + return (flags & FIB_ENTRY_FLAG_LOCAL); +} + +int +api_parse_session_handle (u64 handle, u32 * session_index, u32 * thread_index) +{ + session_manager_main_t *smm = vnet_get_session_manager_main (); + stream_session_t *pool; + + *thread_index = handle & 0xFFFFFFFF; + *session_index = handle >> 32; + + if (*thread_index >= vec_len (smm->sessions)) + return VNET_API_ERROR_INVALID_VALUE; + + pool = smm->sessions[*thread_index]; + + if (pool_is_free_index (pool, *session_index)) + return VNET_API_ERROR_INVALID_VALUE_2; + + return 0; +} + +int +vnet_bind_i (u32 api_client_index, ip46_address_t * ip46, u16 port_host_order, + session_type_t sst, u64 * options, session_cb_vft_t * cb_fns, + application_t ** app, u32 * len_seg_name, char *seg_name) +{ + u8 *segment_name = 0; + application_t *server = 0; + stream_session_t *listener; + u8 is_ip4; + + listener = + stream_session_lookup_listener (ip46, + clib_host_to_net_u16 (port_host_order), + sst); + + if (listener) + return VNET_API_ERROR_ADDRESS_IN_USE; + + if (application_lookup (api_client_index)) + { + clib_warning ("Only one bind supported for now"); + return VNET_API_ERROR_ADDRESS_IN_USE; + } + + is_ip4 = SESSION_TYPE_IP4_UDP == sst || SESSION_TYPE_IP4_TCP == sst; + if (!ip_is_zero (ip46, is_ip4) && !ip_is_local (ip46, is_ip4)) + return VNET_API_ERROR_INVALID_VALUE; + + /* Allocate and initialize stream server */ + server = application_new (APP_SERVER, sst, api_client_index, + options[SESSION_OPTIONS_FLAGS], cb_fns); + + application_server_init (server, options[SESSION_OPTIONS_SEGMENT_SIZE], + options[SESSION_OPTIONS_ADD_SEGMENT_SIZE], + options[SESSION_OPTIONS_RX_FIFO_SIZE], + options[SESSION_OPTIONS_TX_FIFO_SIZE], + &segment_name); + + /* Setup listen path down to transport */ + stream_session_start_listen (server->index, ip46, port_host_order); + + /* + * Return values + */ + + ASSERT (vec_len (segment_name) <= 128); + *len_seg_name = vec_len (segment_name); + memcpy (seg_name, segment_name, *len_seg_name); + *app = server; + + return 0; +} + +int +vnet_unbind_i (u32 api_client_index) +{ + application_t *server; + + /* + * Find the stream_server_t corresponding to the api client + */ + server = application_lookup (api_client_index); + if (!server) + return VNET_API_ERROR_INVALID_VALUE_2; + + /* Clear the listener */ + stream_session_stop_listen (server->index); + application_del (server); + + return 0; +} + +int +vnet_connect_i (u32 api_client_index, u32 api_context, session_type_t sst, + ip46_address_t * ip46, u16 port, u64 * options, void *mp, + session_cb_vft_t * cb_fns) +{ + stream_session_t *listener; + application_t *server, *app; + + /* + * Figure out if connecting to a local server + */ + listener = stream_session_lookup_listener (ip46, + clib_host_to_net_u16 (port), + sst); + if (listener) + { + server = application_get (listener->app_index); + + /* + * Server is willing to have a direct fifo connection created + * instead of going through the state machine, etc. + */ + if (server->flags & SESSION_OPTIONS_FLAGS_USE_FIFO) + return server->cb_fns. + redirect_connect_callback (server->api_client_index, mp); + } + + /* Create client app */ + app = application_new (APP_CLIENT, sst, api_client_index, + options[SESSION_OPTIONS_FLAGS], cb_fns); + + app->api_context = api_context; + + /* + * Not connecting to a local server. Create regular session + */ + stream_session_open (sst, ip46, port, app->index); + + return 0; +} + +/** + * unformat a vnet URI + * + * fifo://name + * tcp://ip46-addr:port + * udp://ip46-addr:port + * + * u8 ip46_address[16]; + * u16 port_in_host_byte_order; + * stream_session_type_t sst; + * u8 *fifo_name; + * + * if (unformat (input, "%U", unformat_vnet_uri, &ip46_address, + * &sst, &port, &fifo_name)) + * etc... + * + */ +uword +unformat_vnet_uri (unformat_input_t * input, va_list * args) +{ + ip46_address_t *address = va_arg (*args, ip46_address_t *); + session_type_t *sst = va_arg (*args, session_type_t *); + u16 *port = va_arg (*args, u16 *); + + if (unformat (input, "tcp://%U/%d", unformat_ip4_address, &address->ip4, + port)) + { + *sst = SESSION_TYPE_IP4_TCP; + return 1; + } + if (unformat (input, "udp://%U/%d", unformat_ip4_address, &address->ip4, + port)) + { + *sst = SESSION_TYPE_IP4_UDP; + return 1; + } + if (unformat (input, "udp://%U/%d", unformat_ip6_address, &address->ip6, + port)) + { + *sst = SESSION_TYPE_IP6_UDP; + return 1; + } + if (unformat (input, "tcp://%U/%d", unformat_ip6_address, &address->ip6, + port)) + { + *sst = SESSION_TYPE_IP6_TCP; + return 1; + } + + return 0; +} + +int +parse_uri (char *uri, session_type_t * sst, ip46_address_t * addr, + u16 * port_number_host_byte_order) +{ + unformat_input_t _input, *input = &_input; + + /* Make sure */ + uri = (char *) format (0, "%s%c", uri, 0); + + /* Parse uri */ + unformat_init_string (input, uri, strlen (uri)); + if (!unformat (input, "%U", unformat_vnet_uri, addr, sst, + port_number_host_byte_order)) + { + unformat_free (input); + return VNET_API_ERROR_INVALID_VALUE; + } + unformat_free (input); + + return 0; +} + +int +vnet_bind_uri (vnet_bind_args_t * a) +{ + application_t *server = 0; + u16 port_host_order; + session_type_t sst = SESSION_N_TYPES; + ip46_address_t ip46; + int rv; + + memset (&ip46, 0, sizeof (ip46)); + rv = parse_uri (a->uri, &sst, &ip46, &port_host_order); + if (rv) + return rv; + + if ((rv = vnet_bind_i (a->api_client_index, &ip46, port_host_order, sst, + a->options, a->session_cb_vft, &server, + &a->segment_name_length, a->segment_name))) + return rv; + + a->server_event_queue_address = (u64) server->event_queue; + return 0; +} + +session_type_t +session_type_from_proto_and_ip (session_api_proto_t proto, u8 is_ip4) +{ + if (proto == SESSION_PROTO_TCP) + { + if (is_ip4) + return SESSION_TYPE_IP4_TCP; + else + return SESSION_TYPE_IP6_TCP; + } + else + { + if (is_ip4) + return SESSION_TYPE_IP4_UDP; + else + return SESSION_TYPE_IP6_UDP; + } + + return SESSION_N_TYPES; +} + +int +vnet_unbind_uri (char *uri, u32 api_client_index) +{ + u16 port_number_host_byte_order; + session_type_t sst = SESSION_N_TYPES; + ip46_address_t ip46_address; + stream_session_t *listener; + int rv; + + rv = parse_uri (uri, &sst, &ip46_address, &port_number_host_byte_order); + if (rv) + return rv; + + listener = + stream_session_lookup_listener (&ip46_address, + clib_host_to_net_u16 + (port_number_host_byte_order), sst); + + if (!listener) + return VNET_API_ERROR_ADDRESS_NOT_IN_USE; + + /* External client? */ + if (api_client_index != ~0) + { + ASSERT (vl_api_client_index_to_registration (api_client_index)); + } + + return vnet_unbind_i (api_client_index); +} + +int +vnet_connect_uri (vnet_connect_args_t * a) +{ + ip46_address_t ip46_address; + u16 port; + session_type_t sst; + application_t *app; + int rv; + + app = application_lookup (a->api_client_index); + if (app) + { + clib_warning ("Already have a connect from this app"); + return VNET_API_ERROR_INVALID_VALUE_2; + } + + /* Parse uri */ + rv = parse_uri (a->uri, &sst, &ip46_address, &port); + if (rv) + return rv; + + return vnet_connect_i (a->api_client_index, a->api_context, sst, + &ip46_address, port, a->options, a->mp, + a->session_cb_vft); +} + +int +vnet_disconnect_session (u32 client_index, u32 session_index, + u32 thread_index) +{ + stream_session_t *session; + + session = stream_session_get (session_index, thread_index); + stream_session_disconnect (session); + + return 0; +} + + +int +vnet_bind (vnet_bind_args_t * a) +{ + application_t *server = 0; + session_type_t sst = SESSION_N_TYPES; + int rv; + + sst = session_type_from_proto_and_ip (a->proto, a->tep.is_ip4); + if ((rv = vnet_bind_i (a->api_client_index, &a->tep.ip, a->tep.port, sst, + a->options, a->session_cb_vft, &server, + &a->segment_name_length, a->segment_name))) + return rv; + + a->server_event_queue_address = (u64) server->event_queue; + a->handle = (u64) a->tep.vrf << 32 | (u64) server->session_index; + return 0; +} + +int +vnet_unbind (vnet_unbind_args_t * a) +{ + application_t *server; + + if (a->api_client_index != ~0) + { + ASSERT (vl_api_client_index_to_registration (a->api_client_index)); + } + + /* Make sure this is the right one */ + server = application_lookup (a->api_client_index); + ASSERT (server->session_index == (0xFFFFFFFF & a->handle)); + + /* TODO use handle to disambiguate namespaces/vrfs */ + return vnet_unbind_i (a->api_client_index); +} + +int +vnet_connect (vnet_connect_args_t * a) +{ + session_type_t sst; + application_t *app; + + app = application_lookup (a->api_client_index); + if (app) + { + clib_warning ("Already have a connect from this app"); + return VNET_API_ERROR_INVALID_VALUE_2; + } + + sst = session_type_from_proto_and_ip (a->proto, a->tep.is_ip4); + return vnet_connect_i (a->api_client_index, a->api_context, sst, &a->tep.ip, + a->tep.port, a->options, a->mp, a->session_cb_vft); +} + +int +vnet_disconnect (vnet_disconnect_args_t * a) +{ + stream_session_t *session; + u32 session_index, thread_index; + + if (api_parse_session_handle (a->handle, &session_index, &thread_index)) + { + clib_warning ("Invalid handle"); + return -1; + } + + session = stream_session_get (session_index, thread_index); + stream_session_disconnect (session); + + return 0; +} + +/* + * fd.io coding-style-patch-verification: ON + * + * Local Variables: + * eval: (c-set-style "gnu") + * End: + */ diff --git a/src/vnet/session/application_interface.h b/src/vnet/session/application_interface.h new file mode 100644 index 00000000..8d87c067 --- /dev/null +++ b/src/vnet/session/application_interface.h @@ -0,0 +1,136 @@ +/* + * Copyright (c) 2016 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ +#ifndef __included_uri_h__ +#define __included_uri_h__ + +#include +#include +#include +#include +#include +#include + +typedef enum _session_api_proto +{ + SESSION_PROTO_TCP, + SESSION_PROTO_UDP +} session_api_proto_t; + +typedef struct _vnet_bind_args_t +{ + union + { + char *uri; + struct + { + transport_endpoint_t tep; + session_api_proto_t proto; + }; + }; + + u32 api_client_index; + u64 *options; + session_cb_vft_t *session_cb_vft; + + /* + * Results + */ + char *segment_name; + u32 segment_name_length; + u64 server_event_queue_address; + u64 handle; +} vnet_bind_args_t; + +typedef struct _vnet_unbind_args_t +{ + union + { + char *uri; + u64 handle; + }; + u32 api_client_index; +} vnet_unbind_args_t; + +typedef struct _vnet_connect_args +{ + union + { + char *uri; + struct + { + transport_endpoint_t tep; + session_api_proto_t proto; + }; + }; + u32 api_client_index; + u32 api_context; + u64 *options; + session_cb_vft_t *session_cb_vft; + + /* Used for redirects */ + void *mp; +} vnet_connect_args_t; + +typedef struct _vnet_disconnect_args_t +{ + u64 handle; + u32 api_client_index; +} vnet_disconnect_args_t; + +/* Bind / connect options */ +typedef enum +{ + SESSION_OPTIONS_FLAGS, + SESSION_OPTIONS_SEGMENT_SIZE, + SESSION_OPTIONS_ADD_SEGMENT_SIZE, + SESSION_OPTIONS_RX_FIFO_SIZE, + SESSION_OPTIONS_TX_FIFO_SIZE, + SESSION_OPTIONS_ACCEPT_COOKIE, + SESSION_OPTIONS_N_OPTIONS +} session_options_index_t; + +/** Server can handle delegated connect requests from local clients */ +#define SESSION_OPTIONS_FLAGS_USE_FIFO (1<<0) + +/** Server wants vpp to add segments when out of memory for fifos */ +#define SESSION_OPTIONS_FLAGS_ADD_SEGMENT (1<<1) + +#define VNET_CONNECT_REDIRECTED 123 + +int vnet_bind_uri (vnet_bind_args_t *); +int vnet_unbind_uri (char *uri, u32 api_client_index); +int vnet_connect_uri (vnet_connect_args_t * a); +int +vnet_disconnect_session (u32 client_index, u32 session_index, + u32 thread_index); + +int vnet_bind (vnet_bind_args_t * a); +int vnet_connect (vnet_connect_args_t * a); +int vnet_unbind (vnet_unbind_args_t * a); +int vnet_disconnect (vnet_disconnect_args_t * a); + +int +api_parse_session_handle (u64 handle, u32 * session_index, + u32 * thread_index); + +#endif /* __included_uri_h__ */ + +/* + * fd.io coding-style-patch-verification: ON + * + * Local Variables: + * eval: (c-set-style "gnu") + * End: + */ diff --git a/src/vnet/session/hashes.c b/src/vnet/session/hashes.c new file mode 100644 index 00000000..1808dd73 --- /dev/null +++ b/src/vnet/session/hashes.c @@ -0,0 +1,28 @@ +/* + * Copyright (c) 2016 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +/** Generate typed init functions for multiple hash table styles... */ + +#include +#include + +#include + +#undef __included_bihash_template_h__ + +#include +#include + +#include diff --git a/src/vnet/session/node.c b/src/vnet/session/node.c new file mode 100644 index 00000000..e467f4e9 --- /dev/null +++ b/src/vnet/session/node.c @@ -0,0 +1,435 @@ +/* + * Copyright (c) 2017 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +#include +#include +#include +#include + +#include + +#include +#include +#include +#include + +#include +#include +#include + +vlib_node_registration_t session_queue_node; + +typedef struct +{ + u32 session_index; + u32 server_thread_index; +} session_queue_trace_t; + +/* packet trace format function */ +static u8 * +format_session_queue_trace (u8 * s, va_list * args) +{ + CLIB_UNUSED (vlib_main_t * vm) = va_arg (*args, vlib_main_t *); + CLIB_UNUSED (vlib_node_t * node) = va_arg (*args, vlib_node_t *); + session_queue_trace_t *t = va_arg (*args, session_queue_trace_t *); + + s = format (s, "SESSION_QUEUE: session index %d, server thread index %d", + t->session_index, t->server_thread_index); + return s; +} + +vlib_node_registration_t session_queue_node; + +#define foreach_session_queue_error \ +_(TX, "Packets transmitted") \ +_(TIMER, "Timer events") + +typedef enum +{ +#define _(sym,str) SESSION_QUEUE_ERROR_##sym, + foreach_session_queue_error +#undef _ + SESSION_QUEUE_N_ERROR, +} session_queue_error_t; + +static char *session_queue_error_strings[] = { +#define _(sym,string) string, + foreach_session_queue_error +#undef _ +}; + +static u32 session_type_to_next[] = { + SESSION_QUEUE_NEXT_TCP_IP4_OUTPUT, + SESSION_QUEUE_NEXT_IP4_LOOKUP, + SESSION_QUEUE_NEXT_TCP_IP6_OUTPUT, + SESSION_QUEUE_NEXT_IP6_LOOKUP, +}; + +always_inline int +session_fifo_rx_i (vlib_main_t * vm, vlib_node_runtime_t * node, + session_manager_main_t * smm, session_fifo_event_t * e0, + stream_session_t * s0, u32 thread_index, int *n_tx_packets, + u8 peek_data) +{ + u32 n_trace = vlib_get_trace_count (vm, node); + u32 left_to_snd0, max_len_to_snd0, len_to_deq0, n_bufs, snd_space0; + u32 n_frame_bytes, n_frames_per_evt; + transport_connection_t *tc0; + transport_proto_vft_t *transport_vft; + u32 next_index, next0, *to_next, n_left_to_next, bi0; + vlib_buffer_t *b0; + u32 rx_offset; + u16 snd_mss0; + u8 *data0; + int i; + + next_index = next0 = session_type_to_next[s0->session_type]; + + transport_vft = session_get_transport_vft (s0->session_type); + tc0 = transport_vft->get_connection (s0->connection_index, thread_index); + + /* Make sure we have space to send and there's something to dequeue */ + snd_space0 = transport_vft->send_space (tc0); + snd_mss0 = transport_vft->send_mss (tc0); + + if (snd_space0 == 0 || svm_fifo_max_dequeue (s0->server_tx_fifo) == 0 + || snd_mss0 == 0) + return 0; + + ASSERT (e0->enqueue_length > 0); + + /* Ensure we're not writing more than transport window allows */ + max_len_to_snd0 = clib_min (e0->enqueue_length, snd_space0); + + if (peek_data) + { + /* Offset in rx fifo from where to peek data */ + rx_offset = transport_vft->rx_fifo_offset (tc0); + } + + /* TODO check if transport is willing to send len_to_snd0 + * bytes (Nagle) */ + + n_frame_bytes = snd_mss0 * VLIB_FRAME_SIZE; + n_frames_per_evt = ceil ((double) max_len_to_snd0 / n_frame_bytes); + + n_bufs = vec_len (smm->tx_buffers[thread_index]); + left_to_snd0 = max_len_to_snd0; + for (i = 0; i < n_frames_per_evt; i++) + { + /* Make sure we have at least one full frame of buffers ready */ + if (PREDICT_FALSE (n_bufs < VLIB_FRAME_SIZE)) + { + vec_validate (smm->tx_buffers[thread_index], + n_bufs + VLIB_FRAME_SIZE - 1); + n_bufs += + vlib_buffer_alloc (vm, &smm->tx_buffers[thread_index][n_bufs], + VLIB_FRAME_SIZE); + + /* buffer shortage + * XXX 0.9 because when debugging we might not get a full frame */ + if (PREDICT_FALSE (n_bufs < 0.9 * VLIB_FRAME_SIZE)) + { + /* Keep track of how much we've dequeued and exit */ + e0->enqueue_length -= max_len_to_snd0 - left_to_snd0; + return -1; + } + + _vec_len (smm->tx_buffers[thread_index]) = n_bufs; + } + + vlib_get_next_frame (vm, node, next_index, to_next, n_left_to_next); + while (left_to_snd0 && n_left_to_next) + { + /* Get free buffer */ + n_bufs--; + bi0 = smm->tx_buffers[thread_index][n_bufs]; + _vec_len (smm->tx_buffers[thread_index]) = n_bufs; + + b0 = vlib_get_buffer (vm, bi0); + b0->error = 0; + b0->flags = VLIB_BUFFER_TOTAL_LENGTH_VALID + | VNET_BUFFER_LOCALLY_ORIGINATED; + b0->current_data = 0; + + /* RX on the local interface. tx in default fib */ + vnet_buffer (b0)->sw_if_index[VLIB_RX] = 0; + vnet_buffer (b0)->sw_if_index[VLIB_TX] = (u32) ~ 0; + + /* usual speculation, or the enqueue_x1 macro will barf */ + to_next[0] = bi0; + to_next += 1; + n_left_to_next -= 1; + + VLIB_BUFFER_TRACE_TRAJECTORY_INIT (b0); + if (PREDICT_FALSE (n_trace > 0)) + { + session_queue_trace_t *t0; + vlib_trace_buffer (vm, node, next_index, b0, + 1 /* follow_chain */ ); + vlib_set_trace_count (vm, node, --n_trace); + t0 = vlib_add_trace (vm, node, b0, sizeof (*t0)); + t0->session_index = s0->session_index; + t0->server_thread_index = s0->thread_index; + } + + if (1) + { + ELOG_TYPE_DECLARE (e) = + { + .format = "evt-dequeue: id %d length %d",.format_args = + "i4i4",}; + struct + { + u32 data[2]; + } *ed; + ed = ELOG_DATA (&vm->elog_main, e); + ed->data[0] = e0->event_id; + ed->data[1] = e0->enqueue_length; + } + + len_to_deq0 = (left_to_snd0 < snd_mss0) ? left_to_snd0 : snd_mss0; + + /* Make room for headers */ + data0 = vlib_buffer_make_headroom (b0, MAX_HDRS_LEN); + + /* Dequeue the data + * TODO 1) peek instead of dequeue + * 2) buffer chains */ + if (peek_data) + { + int n_bytes_read; + n_bytes_read = svm_fifo_peek (s0->server_tx_fifo, s0->pid, + rx_offset, len_to_deq0, data0); + if (n_bytes_read < 0) + goto dequeue_fail; + + /* Keep track of progress locally, transport is also supposed to + * increment it independently when pushing header */ + rx_offset += n_bytes_read; + } + else + { + if (svm_fifo_dequeue_nowait (s0->server_tx_fifo, s0->pid, + len_to_deq0, data0) < 0) + goto dequeue_fail; + } + + b0->current_length = len_to_deq0; + + /* Ask transport to push header */ + transport_vft->push_header (tc0, b0); + + left_to_snd0 -= len_to_deq0; + *n_tx_packets = *n_tx_packets + 1; + + vlib_validate_buffer_enqueue_x1 (vm, node, next_index, + to_next, n_left_to_next, + bi0, next0); + } + vlib_put_next_frame (vm, node, next_index, n_left_to_next); + } + + /* If we couldn't dequeue all bytes store progress */ + if (max_len_to_snd0 < e0->enqueue_length) + { + e0->enqueue_length -= max_len_to_snd0; + vec_add1 (smm->evts_partially_read[thread_index], *e0); + } + return 0; + +dequeue_fail: + /* Can't read from fifo. Store event rx progress, save as partially read, + * return buff to free list and return */ + e0->enqueue_length -= max_len_to_snd0 - left_to_snd0; + vec_add1 (smm->evts_partially_read[thread_index], *e0); + + to_next -= 1; + n_left_to_next += 1; + _vec_len (smm->tx_buffers[thread_index]) += 1; + + clib_warning ("dequeue fail"); + return 0; +} + +int +session_fifo_rx_peek (vlib_main_t * vm, vlib_node_runtime_t * node, + session_manager_main_t * smm, session_fifo_event_t * e0, + stream_session_t * s0, u32 thread_index, int *n_tx_pkts) +{ + return session_fifo_rx_i (vm, node, smm, e0, s0, thread_index, n_tx_pkts, + 1); +} + +int +session_fifo_rx_dequeue (vlib_main_t * vm, vlib_node_runtime_t * node, + session_manager_main_t * smm, + session_fifo_event_t * e0, stream_session_t * s0, + u32 thread_index, int *n_tx_pkts) +{ + return session_fifo_rx_i (vm, node, smm, e0, s0, thread_index, n_tx_pkts, + 0); +} + +static uword +session_queue_node_fn (vlib_main_t * vm, vlib_node_runtime_t * node, + vlib_frame_t * frame) +{ + session_manager_main_t *smm = vnet_get_session_manager_main (); + session_fifo_event_t *my_fifo_events, *e; + u32 n_to_dequeue; + unix_shared_memory_queue_t *q; + int n_tx_packets = 0; + u32 my_thread_index = vm->cpu_index; + int i, rv; + + /* + * Update TCP time + */ + tcp_update_time (vlib_time_now (vm), my_thread_index); + + /* + * Get vpp queue events + */ + q = smm->vpp_event_queues[my_thread_index]; + if (PREDICT_FALSE (q == 0)) + return 0; + + /* min number of events we can dequeue without blocking */ + n_to_dequeue = q->cursize; + if (n_to_dequeue == 0) + return 0; + + my_fifo_events = smm->fifo_events[my_thread_index]; + + /* If we didn't manage to process previous events try going + * over them again without dequeuing new ones. + * XXX: Block senders to sessions that can't keep up */ + if (vec_len (my_fifo_events) >= 100) + goto skip_dequeue; + + /* See you in the next life, don't be late */ + if (pthread_mutex_trylock (&q->mutex)) + return 0; + + for (i = 0; i < n_to_dequeue; i++) + { + vec_add2 (my_fifo_events, e, 1); + unix_shared_memory_queue_sub_raw (q, (u8 *) e); + } + + /* The other side of the connection is not polling */ + if (q->cursize < (q->maxsize / 8)) + (void) pthread_cond_broadcast (&q->condvar); + pthread_mutex_unlock (&q->mutex); + + smm->fifo_events[my_thread_index] = my_fifo_events; + +skip_dequeue: + + for (i = 0; i < n_to_dequeue; i++) + { + svm_fifo_t *f0; /* $$$ prefetch 1 ahead maybe */ + stream_session_t *s0; + u32 server_session_index0, server_thread_index0; + session_fifo_event_t *e0; + + e0 = &my_fifo_events[i]; + f0 = e0->fifo; + server_session_index0 = f0->server_session_index; + server_thread_index0 = f0->server_thread_index; + + /* $$$ add multiple event queues, per vpp worker thread */ + ASSERT (server_thread_index0 == my_thread_index); + + s0 = pool_elt_at_index (smm->sessions[my_thread_index], + server_session_index0); + + ASSERT (s0->thread_index == my_thread_index); + + switch (e0->event_type) + { + case FIFO_EVENT_SERVER_TX: + /* Spray packets in per session type frames, since they go to + * different nodes */ + rv = (smm->session_rx_fns[s0->session_type]) (vm, node, smm, e0, s0, + my_thread_index, + &n_tx_packets); + if (rv < 0) + goto done; + + break; + + default: + clib_warning ("unhandled event type %d", e0->event_type); + } + } + +done: + + /* Couldn't process all events. Probably out of buffers */ + if (PREDICT_FALSE (i < n_to_dequeue)) + { + session_fifo_event_t *partially_read = + smm->evts_partially_read[my_thread_index]; + vec_add (partially_read, &my_fifo_events[i], n_to_dequeue - i); + vec_free (my_fifo_events); + smm->fifo_events[my_thread_index] = partially_read; + smm->evts_partially_read[my_thread_index] = 0; + } + else + { + vec_free (smm->fifo_events[my_thread_index]); + smm->fifo_events[my_thread_index] = + smm->evts_partially_read[my_thread_index]; + smm->evts_partially_read[my_thread_index] = 0; + } + + vlib_node_increment_counter (vm, session_queue_node.index, + SESSION_QUEUE_ERROR_TX, n_tx_packets); + + return n_tx_packets; +} + +/* *INDENT-OFF* */ +VLIB_REGISTER_NODE (session_queue_node) = +{ + .function = session_queue_node_fn, + .name = "session-queue", + .format_trace = format_session_queue_trace, + .type = VLIB_NODE_TYPE_INPUT, + .n_errors = ARRAY_LEN (session_queue_error_strings), + .error_strings = session_queue_error_strings, + .n_next_nodes = SESSION_QUEUE_N_NEXT, + /* .state = VLIB_NODE_STATE_DISABLED, enable on-demand? */ + /* edit / add dispositions here */ + .next_nodes = + { + [SESSION_QUEUE_NEXT_DROP] = "error-drop", + [SESSION_QUEUE_NEXT_IP4_LOOKUP] = "ip4-lookup", + [SESSION_QUEUE_NEXT_IP6_LOOKUP] = "ip6-lookup", + [SESSION_QUEUE_NEXT_TCP_IP4_OUTPUT] = "tcp4-output", + [SESSION_QUEUE_NEXT_TCP_IP6_OUTPUT] = "tcp6-output", + }, +}; +/* *INDENT-ON* */ + +/* + * fd.io coding-style-patch-verification: ON + * + * Local Variables: + * eval: (c-set-style "gnu") + * End: + */ diff --git a/src/vnet/session/session.api b/src/vnet/session/session.api new file mode 100644 index 00000000..a7b28c1d --- /dev/null +++ b/src/vnet/session/session.api @@ -0,0 +1,429 @@ +/* + * Copyright (c) 2015-2016 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + + /** \brief Bind to a given URI + @param client_index - opaque cookie to identify the sender + @param context - sender context, to match reply w/ request + @param accept_cookie - sender accept cookie, to identify this bind flavor + @param uri - a URI, e.g. "tcp://0.0.0.0/0/80" [ipv4] + "tcp://::/0/80" [ipv6] etc. + @param options - socket options, fifo sizes, etc. +*/ +define bind_uri { + u32 client_index; + u32 context; + u32 accept_cookie; + u32 initial_segment_size; + u8 uri[128]; + u64 options[16]; +}; + +/** \brief Unbind a given URI + @param client_index - opaque cookie to identify the sender + @param context - sender context, to match reply w/ request + @param uri - a URI, e.g. "tcp://0.0.0.0/0/80" [ipv4] + "tcp://::/0/80" [ipv6], etc. + @param options - socket options, fifo sizes, etc. +*/ +define unbind_uri { + u32 client_index; + u32 context; + u8 uri[128]; +}; + +/** \brief Connect to a given URI + @param client_index - opaque cookie to identify the sender + @param context - sender context, to match reply w/ request + @param accept_cookie - sender accept cookie, to identify this bind flavor + @param uri - a URI, e.g. "tcp4://0.0.0.0/0/80" + "tcp6://::/0/80" [ipv6], etc. + @param options - socket options, fifo sizes, etc. +*/ +define connect_uri { + u32 client_index; + u32 context; + u8 uri[128]; + u64 client_queue_address; + u64 options[16]; +}; + +/** \brief Bind reply + @param context - sender context, to match reply w/ request + @param retval - return code for the request + @param event_queue_address - vpp event queue address or 0 if this + connection shouldn't send events + @param segment_name_length - length of segment name + @param segment_name - name of segment client needs to attach to +*/ +define bind_uri_reply { + u32 context; + i32 retval; + u64 server_event_queue_address; + u8 segment_name_length; + u32 segment_size; + u8 segment_name[128]; +}; + +/** \brief unbind reply + @param context - sender context, to match reply w/ request + @param retval - return code for the request +*/ +define unbind_uri_reply { + u32 context; + i32 retval; +}; + +/** \brief vpp->client, connect reply + @param context - sender context, to match reply w/ request + @param retval - return code for the request + @param server_rx_fifo - rx (vpp -> vpp-client) fifo address + @param server_tx_fifo - tx (vpp-client -> vpp) fifo address + @param session_index - session index; + @param session_thread_index - session thread index + @param session_type - session thread type + @param vpp_event_queue_address - vpp's event queue address + @param client_event_queue_address - client's event queue address + @param segment_name_length - non-zero if the client needs to attach to + the fifo segment + @param segment_name - set if the client needs to attach to the segment +*/ +define connect_uri_reply { + u32 context; + i32 retval; + u64 server_rx_fifo; + u64 server_tx_fifo; + u32 session_index; + u32 session_thread_index; + u8 session_type; + u64 client_event_queue_address; + u64 vpp_event_queue_address; + u32 segment_size; + u8 segment_name_length; + u8 segment_name[128]; +}; + +/** \brief vpp->client, please map an additional shared memory segment + @param context - sender context, to match reply w/ request + @param segment_name - +*/ +define map_another_segment { + u32 client_index; + u32 context; + u32 segment_size; + u8 segment_name[128]; +}; + +/** \brief client->vpp + @param context - sender context, to match reply w/ request + @param retval - return code for the request +*/ +define map_another_segment_reply { + u32 context; + i32 retval; +}; + +/** \brief vpp->client, accept this session + @param context - sender context, to match reply w/ request + @param accept_cookie - tells client which bind flavor just occurred + @param rx_fifo_address - rx (vpp -> vpp-client) fifo address + @param tx_fifo_address - tx (vpp-client -> vpp) fifo address + @param session_index - index of new session + @param session_thread_index - thread index of new session + @param vpp_event_queue_address - vpp's event queue address + @param session_type - type of session + +*/ +define accept_session { + u32 client_index; + u32 context; + u32 accept_cookie; + u64 server_rx_fifo; + u64 server_tx_fifo; + u32 session_index; + u32 session_thread_index; + u64 vpp_event_queue_address; + u8 session_type; +}; + +/** \brief client->vpp, reply to an accept message + @param context - sender context, to match reply w/ request + @param retval - return code for the request + @param session_index - session index from accept_session / connect_reply + @param session_thread_index - thread index from accept_session / + connect_reply +*/ +define accept_session_reply { + u32 context; + i32 retval; + u8 session_type; + u8 session_thread_index; + u32 session_index; +}; + +/** \brief bidirectional disconnect API + @param client_index - opaque cookie to identify the sender + client to vpp direction only + @param context - sender context, to match reply w/ request + @param session_index - cookie #1 from accept_session / connect_reply + @param session_thread_index - cookie #2 +*/ +define disconnect_session { + u32 client_index; + u32 context; + u32 session_index; + u32 session_thread_index; +}; + +/** \brief bidirectional disconnect reply API + @param client_index - opaque cookie to identify the sender + client to vpp direction only + @param context - sender context, to match reply w/ request + @param retval - return code for the request + @param session_index - session index from accept_session / connect_reply + @param session_thread_index - thread index from accept_session / + connect_reply +*/ +define disconnect_session_reply { + u32 client_index; + u32 context; + i32 retval; + u32 session_index; + u32 session_thread_index; +}; + +/** \brief vpp->client reset session API + @param client_index - opaque cookie to identify the sender + client to vpp direction only + @param context - sender context, to match reply w/ request + @param session_index - session index from accept_session / connect_reply + @param session_thread_index - thread index from accept_session / + connect_reply +*/ +define reset_session { + u32 client_index; + u32 context; + u32 session_index; + u32 session_thread_index; +}; + +/** \brief client->vpp reset session reply + @param client_index - opaque cookie to identify the sender + client to vpp direction only + @param context - sender context, to match reply w/ request + @param retval - return code for the request + @param session_index - session index from accept_session / connect_reply + @param session_thread_index - thread index from accept_session / + connect_reply +*/ +define reset_session_reply { + u32 client_index; + u32 context; + i32 retval; + u32 session_index; + u32 session_thread_index; +}; + +/** \brief Bind to an ip:port pair for a given transport protocol + @param client_index - opaque cookie to identify the sender + @param context - sender context, to match reply w/ request + @param vrf - bind namespace + @param is_ip4 - flag that is 1 if ip address family is IPv4 + @param ip - ip address + @param port - port + @param proto - protocol 0 - TCP 1 - UDP + @param options - socket options, fifo sizes, etc. +*/ +define bind_sock { + u32 client_index; + u32 context; + u32 vrf; + u8 is_ip4; + u8 ip[16]; + u16 port; + u8 proto; + u64 options[16]; +}; + +/** \brief Unbind + @param client_index - opaque cookie to identify the sender + @param context - sender context, to match reply w/ request + @param handle - bind handle obtained from bind reply +*/ +define unbind_sock { + u32 client_index; + u32 context; + u64 handle; +}; + +/** \brief Connect to a remote peer + @param client_index - opaque cookie to identify the sender + @param context - sender context, to match reply w/ request + @param vrf - connection namespace + @param is_ip4 - flag that is 1 if ip address family is IPv4 + @param ip - ip address + @param port - port + @param proto - protocol 0 - TCP 1 - UDP + @param client_queue_address - client's API queue address. Non-zero when + used to perform redirects + @param options - socket options, fifo sizes, etc. +*/ +define connect_sock { + u32 client_index; + u32 context; + u32 vrf; + u8 is_ip4; + u8 ip[16]; + u16 port; + u8 proto; + u64 client_queue_address; + u64 options[16]; +}; + +/** \brief Bind reply + @param context - sender context, to match reply w/ request + @param handle - bind handle + @param retval - return code for the request + @param event_queue_address - vpp event queue address or 0 if this + connection shouldn't send events + @param segment_name_length - length of segment name + @param segment_name - name of segment client needs to attach to +*/ +define bind_sock_reply { + u32 context; + u64 handle; + i32 retval; + u64 server_event_queue_address; + u32 segment_size; + u8 segment_name_length; + u8 segment_name[128]; +}; + +/** \brief unbind reply + @param context - sender context, to match reply w/ request + @param retval - return code for the request +*/ +define unbind_sock_reply { + u32 context; + i32 retval; +}; + +/** \brief vpp/server->client, connect reply + @param context - sender context, to match reply w/ request + @param retval - return code for the request + @param handle - connection handle + @param server_rx_fifo - rx (vpp -> vpp-client) fifo address + @param server_tx_fifo - tx (vpp-client -> vpp) fifo address + @param vpp_event_queue_address - vpp's event queue address + @param client_event_queue_address - client's event queue address + @param segment_name_length - non-zero if the client needs to attach to + the fifo segment + @param segment_name - set if the client needs to attach to the segment +*/ +define connect_sock_reply { + u32 context; + i32 retval; + u64 handle; + u64 server_rx_fifo; + u64 server_tx_fifo; + u64 client_event_queue_address; + u64 vpp_event_queue_address; + u32 segment_size; + u8 segment_name_length; + u8 segment_name[128]; +}; + +/** \brief bidirectional disconnect API + @param client_index - opaque cookie to identify the sender + client to vpp direction only + @param context - sender context, to match reply w/ request + @param handle - session handle obtained through accept/connect +*/ +define disconnect_sock { + u32 client_index; + u32 context; + u64 handle; +}; + +/** \brief bidirectional disconnect reply API + @param client_index - opaque cookie to identify the sender + client to vpp direction only + @param client_context - sender context, to match reply w/ request + @param handle - session handle obtained through accept/connect +*/ +define disconnect_sock_reply { + u32 client_index; + u32 context; + i32 retval; + u64 handle; +}; + +/** \brief vpp->client, accept this session + @param context - sender context, to match reply w/ request + @param accept_cookie - tells client which bind flavor just occurred + @param handle - session handle obtained through accept/connect + @param rx_fifo_address - rx (vpp -> vpp-client) fifo address + @param tx_fifo_address - tx (vpp-client -> vpp) fifo address + @param vpp_event_queue_address - vpp's event queue address +*/ +define accept_sock { + u32 client_index; + u32 context; + u32 accept_cookie; + u64 handle; + u64 server_rx_fifo; + u64 server_tx_fifo; + u64 vpp_event_queue_address; +}; + +/** \brief client->vpp, reply to an accept message + @param context - sender context, to match reply w/ request + @param retval - return code for the request + @param handle - session handle obtained through accept/connect +*/ +define accept_sock_reply { + u32 context; + i32 retval; + u64 handle; +}; + +/** \brief vpp->client reset session API + @param client_index - opaque cookie to identify the sender + client to vpp direction only + @param context - sender context, to match reply w/ request + @param handle - session handle obtained through accept/connect +*/ +define reset_sock { + u32 client_index; + u32 context; + u64 handle; +}; + +/** \brief client->vpp reset session reply + @param client_index - opaque cookie to identify the sender + client to vpp direction only + @param context - sender context, to match reply w/ request + @param handle - session handle obtained through accept/connect +*/ +define reset_sock_reply { + u32 client_index; + u32 context; + i32 retval; + u64 handle; +}; +/* + * Local Variables: + * eval: (c-set-style "gnu") + * End: + */ \ No newline at end of file diff --git a/src/vnet/session/session.c b/src/vnet/session/session.c new file mode 100644 index 00000000..539da613 --- /dev/null +++ b/src/vnet/session/session.c @@ -0,0 +1,1286 @@ +/* + * Copyright (c) 2017 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ +/** + * @file + * @brief Session and session manager + */ + +#include +#include +#include +#include +#include + +/** + * Per-type vector of transport protocol virtual function tables + */ +static transport_proto_vft_t *tp_vfts; + +session_manager_main_t session_manager_main; + +/* + * Session lookup key; (src-ip, dst-ip, src-port, dst-port, session-type) + * Value: (owner thread index << 32 | session_index); + */ +static void +stream_session_table_add_for_tc (u8 sst, transport_connection_t * tc, + u64 value) +{ + session_manager_main_t *smm = &session_manager_main; + session_kv4_t kv4; + session_kv6_t kv6; + + switch (sst) + { + case SESSION_TYPE_IP4_UDP: + case SESSION_TYPE_IP4_TCP: + make_v4_ss_kv_from_tc (&kv4, tc); + kv4.value = value; + clib_bihash_add_del_16_8 (&smm->v4_session_hash, &kv4, 1 /* is_add */ ); + break; + case SESSION_TYPE_IP6_UDP: + case SESSION_TYPE_IP6_TCP: + make_v6_ss_kv_from_tc (&kv6, tc); + kv6.value = value; + clib_bihash_add_del_48_8 (&smm->v6_session_hash, &kv6, 1 /* is_add */ ); + break; + default: + clib_warning ("Session type not supported"); + ASSERT (0); + } +} + +void +stream_session_table_add (session_manager_main_t * smm, stream_session_t * s, + u64 value) +{ + transport_connection_t *tc; + + tc = tp_vfts[s->session_type].get_connection (s->connection_index, + s->thread_index); + stream_session_table_add_for_tc (s->session_type, tc, value); +} + +static void +stream_session_half_open_table_add (u8 sst, transport_connection_t * tc, + u64 value) +{ + session_manager_main_t *smm = &session_manager_main; + session_kv4_t kv4; + session_kv6_t kv6; + + switch (sst) + { + case SESSION_TYPE_IP4_UDP: + case SESSION_TYPE_IP4_TCP: + make_v4_ss_kv_from_tc (&kv4, tc); + kv4.value = value; + clib_bihash_add_del_16_8 (&smm->v4_half_open_hash, &kv4, + 1 /* is_add */ ); + break; + case SESSION_TYPE_IP6_UDP: + case SESSION_TYPE_IP6_TCP: + make_v6_ss_kv_from_tc (&kv6, tc); + kv6.value = value; + clib_bihash_add_del_48_8 (&smm->v6_half_open_hash, &kv6, + 1 /* is_add */ ); + break; + default: + clib_warning ("Session type not supported"); + ASSERT (0); + } +} + +static int +stream_session_table_del_for_tc (session_manager_main_t * smm, u8 sst, + transport_connection_t * tc) +{ + session_kv4_t kv4; + session_kv6_t kv6; + + switch (sst) + { + case SESSION_TYPE_IP4_UDP: + case SESSION_TYPE_IP4_TCP: + make_v4_ss_kv_from_tc (&kv4, tc); + return clib_bihash_add_del_16_8 (&smm->v4_session_hash, &kv4, + 0 /* is_add */ ); + break; + case SESSION_TYPE_IP6_UDP: + case SESSION_TYPE_IP6_TCP: + make_v6_ss_kv_from_tc (&kv6, tc); + return clib_bihash_add_del_48_8 (&smm->v6_session_hash, &kv6, + 0 /* is_add */ ); + break; + default: + clib_warning ("Session type not supported"); + ASSERT (0); + } + + return 0; +} + +static int +stream_session_table_del (session_manager_main_t * smm, stream_session_t * s) +{ + transport_connection_t *ts; + + ts = tp_vfts[s->session_type].get_connection (s->connection_index, + s->thread_index); + return stream_session_table_del_for_tc (smm, s->session_type, ts); +} + +static void +stream_session_half_open_table_del (session_manager_main_t * smm, u8 sst, + transport_connection_t * tc) +{ + session_kv4_t kv4; + session_kv6_t kv6; + + switch (sst) + { + case SESSION_TYPE_IP4_UDP: + case SESSION_TYPE_IP4_TCP: + make_v4_ss_kv_from_tc (&kv4, tc); + clib_bihash_add_del_16_8 (&smm->v4_half_open_hash, &kv4, + 0 /* is_add */ ); + break; + case SESSION_TYPE_IP6_UDP: + case SESSION_TYPE_IP6_TCP: + make_v6_ss_kv_from_tc (&kv6, tc); + clib_bihash_add_del_48_8 (&smm->v6_half_open_hash, &kv6, + 0 /* is_add */ ); + break; + default: + clib_warning ("Session type not supported"); + ASSERT (0); + } +} + +stream_session_t * +stream_session_lookup_listener4 (ip4_address_t * lcl, u16 lcl_port, u8 proto) +{ + session_manager_main_t *smm = &session_manager_main; + session_kv4_t kv4; + int rv; + + make_v4_listener_kv (&kv4, lcl, lcl_port, proto); + rv = clib_bihash_search_inline_16_8 (&smm->v4_session_hash, &kv4); + if (rv == 0) + return pool_elt_at_index (smm->listen_sessions[proto], (u32) kv4.value); + + /* Zero out the lcl ip */ + kv4.key[0] = 0; + rv = clib_bihash_search_inline_16_8 (&smm->v4_session_hash, &kv4); + if (rv == 0) + return pool_elt_at_index (smm->listen_sessions[proto], kv4.value); + + return 0; +} + +/** Looks up a session based on the 5-tuple passed as argument. + * + * First it tries to find an established session, if this fails, it tries + * finding a listener session if this fails, it tries a lookup with a + * wildcarded local source (listener bound to all interfaces) + */ +stream_session_t * +stream_session_lookup4 (ip4_address_t * lcl, ip4_address_t * rmt, + u16 lcl_port, u16 rmt_port, u8 proto, + u32 my_thread_index) +{ + session_manager_main_t *smm = &session_manager_main; + session_kv4_t kv4; + int rv; + + /* Lookup session amongst established ones */ + make_v4_ss_kv (&kv4, lcl, rmt, lcl_port, rmt_port, proto); + rv = clib_bihash_search_inline_16_8 (&smm->v4_session_hash, &kv4); + if (rv == 0) + return stream_session_get_tsi (kv4.value, my_thread_index); + + /* If nothing is found, check if any listener is available */ + return stream_session_lookup_listener4 (lcl, lcl_port, proto); +} + +stream_session_t * +stream_session_lookup_listener6 (ip6_address_t * lcl, u16 lcl_port, u8 proto) +{ + session_manager_main_t *smm = &session_manager_main; + session_kv6_t kv6; + int rv; + + make_v6_listener_kv (&kv6, lcl, lcl_port, proto); + rv = clib_bihash_search_inline_48_8 (&smm->v6_session_hash, &kv6); + if (rv == 0) + return pool_elt_at_index (smm->listen_sessions[proto], kv6.value); + + /* Zero out the lcl ip */ + kv6.key[0] = kv6.key[1] = 0; + rv = clib_bihash_search_inline_48_8 (&smm->v6_session_hash, &kv6); + if (rv == 0) + return pool_elt_at_index (smm->listen_sessions[proto], kv6.value); + + return 0; +} + +/* Looks up a session based on the 5-tuple passed as argument. + * First it tries to find an established session, if this fails, it tries + * finding a listener session if this fails, it tries a lookup with a + * wildcarded local source (listener bound to all interfaces) */ +stream_session_t * +stream_session_lookup6 (ip6_address_t * lcl, ip6_address_t * rmt, + u16 lcl_port, u16 rmt_port, u8 proto, + u32 my_thread_index) +{ + session_manager_main_t *smm = vnet_get_session_manager_main (); + session_kv6_t kv6; + int rv; + + make_v6_ss_kv (&kv6, lcl, rmt, lcl_port, rmt_port, proto); + rv = clib_bihash_search_inline_48_8 (&smm->v6_session_hash, &kv6); + if (rv == 0) + return stream_session_get_tsi (kv6.value, my_thread_index); + + /* If nothing is found, check if any listener is available */ + return stream_session_lookup_listener6 (lcl, lcl_port, proto); +} + +stream_session_t * +stream_session_lookup_listener (ip46_address_t * lcl, u16 lcl_port, u8 proto) +{ + switch (proto) + { + case SESSION_TYPE_IP4_UDP: + case SESSION_TYPE_IP4_TCP: + return stream_session_lookup_listener4 (&lcl->ip4, lcl_port, proto); + break; + case SESSION_TYPE_IP6_UDP: + case SESSION_TYPE_IP6_TCP: + return stream_session_lookup_listener6 (&lcl->ip6, lcl_port, proto); + break; + } + return 0; +} + +static u64 +stream_session_half_open_lookup (session_manager_main_t * smm, + ip46_address_t * lcl, ip46_address_t * rmt, + u16 lcl_port, u16 rmt_port, u8 proto) +{ + session_kv4_t kv4; + session_kv6_t kv6; + int rv; + + switch (proto) + { + case SESSION_TYPE_IP4_UDP: + case SESSION_TYPE_IP4_TCP: + make_v4_ss_kv (&kv4, &lcl->ip4, &rmt->ip4, lcl_port, rmt_port, proto); + rv = clib_bihash_search_inline_16_8 (&smm->v4_half_open_hash, &kv4); + + if (rv == 0) + return kv4.value; + + return (u64) ~ 0; + break; + case SESSION_TYPE_IP6_UDP: + case SESSION_TYPE_IP6_TCP: + make_v6_ss_kv (&kv6, &lcl->ip6, &rmt->ip6, lcl_port, rmt_port, proto); + rv = clib_bihash_search_inline_48_8 (&smm->v6_half_open_hash, &kv6); + + if (rv == 0) + return kv6.value; + + return (u64) ~ 0; + break; + } + return 0; +} + +transport_connection_t * +stream_session_lookup_transport4 (session_manager_main_t * smm, + ip4_address_t * lcl, ip4_address_t * rmt, + u16 lcl_port, u16 rmt_port, u8 proto, + u32 my_thread_index) +{ + session_kv4_t kv4; + stream_session_t *s; + int rv; + + /* Lookup session amongst established ones */ + make_v4_ss_kv (&kv4, lcl, rmt, lcl_port, rmt_port, proto); + rv = clib_bihash_search_inline_16_8 (&smm->v4_session_hash, &kv4); + if (rv == 0) + { + s = stream_session_get_tsi (kv4.value, my_thread_index); + + return tp_vfts[s->session_type].get_connection (s->connection_index, + my_thread_index); + } + + /* If nothing is found, check if any listener is available */ + s = stream_session_lookup_listener4 (lcl, lcl_port, proto); + if (s) + return tp_vfts[s->session_type].get_listener (s->connection_index); + + /* Finally, try half-open connections */ + rv = clib_bihash_search_inline_16_8 (&smm->v4_half_open_hash, &kv4); + if (rv == 0) + return tp_vfts[proto].get_half_open (kv4.value & 0xFFFFFFFF); + + return 0; +} + +transport_connection_t * +stream_session_lookup_transport6 (session_manager_main_t * smm, + ip6_address_t * lcl, ip6_address_t * rmt, + u16 lcl_port, u16 rmt_port, u8 proto, + u32 my_thread_index) +{ + stream_session_t *s; + session_kv6_t kv6; + int rv; + + make_v6_ss_kv (&kv6, lcl, rmt, lcl_port, rmt_port, proto); + rv = clib_bihash_search_inline_48_8 (&smm->v6_session_hash, &kv6); + if (rv == 0) + { + s = stream_session_get_tsi (kv6.value, my_thread_index); + + return tp_vfts[s->session_type].get_connection (s->connection_index, + my_thread_index); + } + + /* If nothing is found, check if any listener is available */ + s = stream_session_lookup_listener6 (lcl, lcl_port, proto); + if (s) + return tp_vfts[s->session_type].get_listener (s->connection_index); + + /* Finally, try half-open connections */ + rv = clib_bihash_search_inline_48_8 (&smm->v6_half_open_hash, &kv6); + if (rv == 0) + return tp_vfts[s->session_type].get_half_open (kv6.value & 0xFFFFFFFF); + + return 0; +} + +/** + * Allocate vpp event queue (once) per worker thread + */ +void +vpp_session_event_queue_allocate (session_manager_main_t * smm, + u32 thread_index) +{ + api_main_t *am = &api_main; + void *oldheap; + + if (smm->vpp_event_queues[thread_index] == 0) + { + /* Allocate event fifo in the /vpe-api shared-memory segment */ + oldheap = svm_push_data_heap (am->vlib_rp); + + smm->vpp_event_queues[thread_index] = + unix_shared_memory_queue_init (2048 /* nels $$$$ config */ , + sizeof (session_fifo_event_t), + 0 /* consumer pid */ , + 0 + /* (do not) send signal when queue non-empty */ + ); + + svm_pop_heap (oldheap); + } +} + +void +session_manager_get_segment_info (u32 index, u8 ** name, u32 * size) +{ + svm_fifo_segment_private_t *s; + s = svm_fifo_get_segment (index); + *name = s->h->segment_name; + *size = s->ssvm.ssvm_size; +} + +always_inline int +session_manager_add_segment_i (session_manager_main_t * smm, + session_manager_t * sm, + u32 segment_size, u8 * segment_name) +{ + svm_fifo_segment_create_args_t _ca, *ca = &_ca; + int rv; + + memset (ca, 0, sizeof (*ca)); + + ca->segment_name = (char *) segment_name; + ca->segment_size = segment_size; + + rv = svm_fifo_segment_create (ca); + if (rv) + { + clib_warning ("svm_fifo_segment_create ('%s', %d) failed", + ca->segment_name, ca->segment_size); + vec_free (segment_name); + return -1; + } + + vec_add1 (sm->segment_indices, ca->new_segment_index); + + return 0; +} + +static int +session_manager_add_segment (session_manager_main_t * smm, + session_manager_t * sm) +{ + u8 *segment_name; + svm_fifo_segment_create_args_t _ca, *ca = &_ca; + u32 add_segment_size; + u32 default_segment_size = 128 << 10; + + memset (ca, 0, sizeof (*ca)); + segment_name = format (0, "%d-%d%c", getpid (), + smm->unique_segment_name_counter++, 0); + add_segment_size = + sm->add_segment_size ? sm->add_segment_size : default_segment_size; + + return session_manager_add_segment_i (smm, sm, add_segment_size, + segment_name); +} + +int +session_manager_add_first_segment (session_manager_main_t * smm, + session_manager_t * sm, u32 segment_size, + u8 ** segment_name) +{ + svm_fifo_segment_create_args_t _ca, *ca = &_ca; + memset (ca, 0, sizeof (*ca)); + *segment_name = format (0, "%d-%d%c", getpid (), + smm->unique_segment_name_counter++, 0); + return session_manager_add_segment_i (smm, sm, segment_size, *segment_name); +} + +void +session_manager_del (session_manager_main_t * smm, session_manager_t * sm) +{ + u32 *deleted_sessions = 0; + u32 *deleted_thread_indices = 0; + int i, j; + + /* Across all fifo segments used by the server */ + for (j = 0; j < vec_len (sm->segment_indices); j++) + { + svm_fifo_segment_private_t *fifo_segment; + svm_fifo_t **fifos; + /* Vector of fifos allocated in the segment */ + fifo_segment = svm_fifo_get_segment (sm->segment_indices[j]); + fifos = (svm_fifo_t **) fifo_segment->h->fifos; + + /* + * Remove any residual sessions from the session lookup table + * Don't bother deleting the individual fifos, we're going to + * throw away the fifo segment in a minute. + */ + for (i = 0; i < vec_len (fifos); i++) + { + svm_fifo_t *fifo; + u32 session_index, thread_index; + stream_session_t *session; + + fifo = fifos[i]; + session_index = fifo->server_session_index; + thread_index = fifo->server_thread_index; + + session = pool_elt_at_index (smm->sessions[thread_index], + session_index); + + /* Add to the deleted_sessions vector (once!) */ + if (!session->is_deleted) + { + session->is_deleted = 1; + vec_add1 (deleted_sessions, + session - smm->sessions[thread_index]); + vec_add1 (deleted_thread_indices, thread_index); + } + } + + for (i = 0; i < vec_len (deleted_sessions); i++) + { + stream_session_t *session; + + session = + pool_elt_at_index (smm->sessions[deleted_thread_indices[i]], + deleted_sessions[i]); + + /* Instead of directly removing the session call disconnect */ + stream_session_disconnect (session); + + /* + stream_session_table_del (smm, session); + pool_put(smm->sessions[deleted_thread_indices[i]], session); + */ + } + + vec_reset_length (deleted_sessions); + vec_reset_length (deleted_thread_indices); + + /* Instead of removing the segment, test when removing the session if + * the segment can be removed + */ + /* svm_fifo_segment_delete (fifo_segment); */ + } + + vec_free (deleted_sessions); + vec_free (deleted_thread_indices); +} + +int +session_manager_allocate_session_fifos (session_manager_main_t * smm, + session_manager_t * sm, + svm_fifo_t ** server_rx_fifo, + svm_fifo_t ** server_tx_fifo, + u32 * fifo_segment_index, + u8 * added_a_segment) +{ + svm_fifo_segment_private_t *fifo_segment; + u32 fifo_size, default_fifo_size = 8192 /* TODO config */ ; + int i; + + *added_a_segment = 0; + + /* Allocate svm fifos */ + ASSERT (vec_len (sm->segment_indices)); + +again: + for (i = 0; i < vec_len (sm->segment_indices); i++) + { + *fifo_segment_index = sm->segment_indices[i]; + fifo_segment = svm_fifo_get_segment (*fifo_segment_index); + + fifo_size = sm->rx_fifo_size; + fifo_size = (fifo_size == 0) ? default_fifo_size : fifo_size; + *server_rx_fifo = svm_fifo_segment_alloc_fifo (fifo_segment, fifo_size); + + fifo_size = sm->tx_fifo_size; + fifo_size = (fifo_size == 0) ? default_fifo_size : fifo_size; + *server_tx_fifo = svm_fifo_segment_alloc_fifo (fifo_segment, fifo_size); + + if (*server_rx_fifo == 0) + { + /* This would be very odd, but handle it... */ + if (*server_tx_fifo != 0) + { + svm_fifo_segment_free_fifo (fifo_segment, *server_tx_fifo); + *server_tx_fifo = 0; + } + continue; + } + if (*server_tx_fifo == 0) + { + if (*server_rx_fifo != 0) + { + svm_fifo_segment_free_fifo (fifo_segment, *server_rx_fifo); + *server_rx_fifo = 0; + } + continue; + } + break; + } + + /* See if we're supposed to create another segment */ + if (*server_rx_fifo == 0) + { + if (sm->add_segment) + { + if (*added_a_segment) + { + clib_warning ("added a segment, still cant allocate a fifo"); + return SESSION_ERROR_NEW_SEG_NO_SPACE; + } + + if (session_manager_add_segment (smm, sm)) + return VNET_API_ERROR_URI_FIFO_CREATE_FAILED; + + *added_a_segment = 1; + goto again; + } + else + return SESSION_ERROR_NO_SPACE; + } + return 0; +} + +int +stream_session_create_i (session_manager_main_t * smm, application_t * app, + transport_connection_t * tc, + stream_session_t ** ret_s) +{ + int rv; + svm_fifo_t *server_rx_fifo = 0, *server_tx_fifo = 0; + u32 fifo_segment_index; + u32 pool_index, seg_size; + stream_session_t *s; + u64 value; + u32 thread_index = tc->thread_index; + session_manager_t *sm; + u8 segment_added; + u8 *seg_name; + + sm = session_manager_get (app->session_manager_index); + + /* Check the API queue */ + if (app->mode == APP_SERVER && application_api_queue_is_full (app)) + return SESSION_ERROR_API_QUEUE_FULL; + + if ((rv = session_manager_allocate_session_fifos (smm, sm, &server_rx_fifo, + &server_tx_fifo, + &fifo_segment_index, + &segment_added))) + return rv; + + if (segment_added && app->mode == APP_SERVER) + { + /* Send an API message to the external server, to map new segment */ + ASSERT (app->cb_fns.add_segment_callback); + + session_manager_get_segment_info (fifo_segment_index, &seg_name, + &seg_size); + if (app->cb_fns.add_segment_callback (app->api_client_index, seg_name, + seg_size)) + return VNET_API_ERROR_URI_FIFO_CREATE_FAILED; + } + + /* Create the session */ + pool_get (smm->sessions[thread_index], s); + memset (s, 0, sizeof (*s)); + + /* Initialize backpointers */ + pool_index = s - smm->sessions[thread_index]; + server_rx_fifo->server_session_index = pool_index; + server_rx_fifo->server_thread_index = thread_index; + + server_tx_fifo->server_session_index = pool_index; + server_tx_fifo->server_thread_index = thread_index; + + s->server_rx_fifo = server_rx_fifo; + s->server_tx_fifo = server_tx_fifo; + + /* Initialize state machine, such as it is... */ + s->session_type = app->session_type; + s->session_state = SESSION_STATE_CONNECTING; + s->app_index = application_get_index (app); + s->server_segment_index = fifo_segment_index; + s->thread_index = thread_index; + s->session_index = pool_index; + + /* Attach transport to session */ + s->connection_index = tc->c_index; + + /* Attach session to transport */ + tc->s_index = s->session_index; + + /* Add to the main lookup table */ + value = (((u64) thread_index) << 32) | (u64) s->session_index; + stream_session_table_add_for_tc (app->session_type, tc, value); + + *ret_s = s; + + return 0; +} + +/* + * Enqueue data for delivery to session peer. Does not notify peer of enqueue + * event but on request can queue notification events for later delivery by + * calling stream_server_flush_enqueue_events(). + * + * @param tc Transport connection which is to be enqueued data + * @param data Data to be enqueued + * @param len Length of data to be enqueued + * @param queue_event Flag to indicate if peer is to be notified or if event + * is to be queued. The former is useful when more data is + * enqueued and only one event is to be generated. + * @return Number of bytes enqueued or a negative value if enqueueing failed. + */ +int +stream_session_enqueue_data (transport_connection_t * tc, u8 * data, u16 len, + u8 queue_event) +{ + stream_session_t *s; + int enqueued; + + s = stream_session_get (tc->s_index, tc->thread_index); + + /* Make sure there's enough space left. We might've filled the pipes */ + if (PREDICT_FALSE (len > svm_fifo_max_enqueue (s->server_rx_fifo))) + return -1; + + enqueued = svm_fifo_enqueue_nowait (s->server_rx_fifo, s->pid, len, data); + + if (queue_event) + { + /* Queue RX event on this fifo. Eventually these will need to be flushed + * by calling stream_server_flush_enqueue_events () */ + session_manager_main_t *smm = vnet_get_session_manager_main (); + u32 thread_index = s->thread_index; + u32 my_enqueue_epoch = smm->current_enqueue_epoch[thread_index]; + + if (s->enqueue_epoch != my_enqueue_epoch) + { + s->enqueue_epoch = my_enqueue_epoch; + vec_add1 (smm->session_indices_to_enqueue_by_thread[thread_index], + s - smm->sessions[thread_index]); + } + } + + return enqueued; +} + +/** Check if we have space in rx fifo to push more bytes */ +u8 +stream_session_no_space (transport_connection_t * tc, u32 thread_index, + u16 data_len) +{ + stream_session_t *s = stream_session_get (tc->c_index, thread_index); + + if (PREDICT_FALSE (s->session_state != SESSION_STATE_READY)) + return 1; + + if (data_len > svm_fifo_max_enqueue (s->server_rx_fifo)) + return 1; + + return 0; +} + +u32 +stream_session_peek_bytes (transport_connection_t * tc, u8 * buffer, + u32 offset, u32 max_bytes) +{ + stream_session_t *s = stream_session_get (tc->s_index, tc->thread_index); + return svm_fifo_peek (s->server_tx_fifo, s->pid, offset, max_bytes, buffer); +} + +u32 +stream_session_dequeue_drop (transport_connection_t * tc, u32 max_bytes) +{ + stream_session_t *s = stream_session_get (tc->s_index, tc->thread_index); + return svm_fifo_dequeue_drop (s->server_tx_fifo, s->pid, max_bytes); +} + +/** + * Notify session peer that new data has been enqueued. + * + * @param s Stream session for which the event is to be generated. + * @param block Flag to indicate if call should block if event queue is full. + * + * @return 0 on succes or negative number if failed to send notification. + */ +static int +stream_session_enqueue_notify (stream_session_t * s, u8 block) +{ + application_t *app; + session_fifo_event_t evt; + unix_shared_memory_queue_t *q; + static u32 serial_number; + + if (PREDICT_FALSE (s->session_state == SESSION_STATE_CLOSED)) + return 0; + + /* Get session's server */ + app = application_get (s->app_index); + + /* Fabricate event */ + evt.fifo = s->server_rx_fifo; + evt.event_type = FIFO_EVENT_SERVER_RX; + evt.event_id = serial_number++; + evt.enqueue_length = svm_fifo_max_dequeue (s->server_rx_fifo); + + /* Add event to server's event queue */ + q = app->event_queue; + + /* Based on request block (or not) for lack of space */ + if (block || PREDICT_TRUE (q->cursize < q->maxsize)) + unix_shared_memory_queue_add (app->event_queue, (u8 *) & evt, + 0 /* do wait for mutex */ ); + else + return -1; + + if (1) + { + ELOG_TYPE_DECLARE (e) = + { + .format = "evt-enqueue: id %d length %d",.format_args = "i4i4",}; + struct + { + u32 data[2]; + } *ed; + ed = ELOG_DATA (&vlib_global_main.elog_main, e); + ed->data[0] = evt.event_id; + ed->data[1] = evt.enqueue_length; + } + + return 0; +} + +/** + * Flushes queue of sessions that are to be notified of new data + * enqueued events. + * + * @param thread_index Thread index for which the flush is to be performed. + * @return 0 on success or a positive number indicating the number of + * failures due to API queue being full. + */ +int +session_manager_flush_enqueue_events (u32 thread_index) +{ + session_manager_main_t *smm = &session_manager_main; + u32 *session_indices_to_enqueue; + int i, errors = 0; + + session_indices_to_enqueue = + smm->session_indices_to_enqueue_by_thread[thread_index]; + + for (i = 0; i < vec_len (session_indices_to_enqueue); i++) + { + stream_session_t *s0; + + /* Get session */ + s0 = stream_session_get (session_indices_to_enqueue[i], thread_index); + if (stream_session_enqueue_notify (s0, 0 /* don't block */ )) + { + errors++; + } + } + + vec_reset_length (session_indices_to_enqueue); + + smm->session_indices_to_enqueue_by_thread[thread_index] = + session_indices_to_enqueue; + + /* Increment enqueue epoch for next round */ + smm->current_enqueue_epoch[thread_index]++; + + return errors; +} + +/* + * Start listening on server's ip/port pair for requested transport. + * + * Creates a 'dummy' stream session with state LISTENING to be used in session + * lookups, prior to establishing connection. Requests transport to build + * it's own specific listening connection. + */ +int +stream_session_start_listen (u32 server_index, ip46_address_t * ip, u16 port) +{ + session_manager_main_t *smm = &session_manager_main; + stream_session_t *s; + transport_connection_t *tc; + application_t *srv; + u32 tci; + + srv = application_get (server_index); + + pool_get (smm->listen_sessions[srv->session_type], s); + memset (s, 0, sizeof (*s)); + + s->session_type = srv->session_type; + s->session_state = SESSION_STATE_LISTENING; + s->session_index = s - smm->listen_sessions[srv->session_type]; + s->app_index = srv->index; + + /* Transport bind/listen */ + tci = tp_vfts[srv->session_type].bind (smm->vlib_main, s->session_index, ip, + port); + + /* Attach transport to session */ + s->connection_index = tci; + tc = tp_vfts[srv->session_type].get_listener (tci); + + srv->session_index = s->session_index; + + /* Add to the main lookup table */ + stream_session_table_add_for_tc (s->session_type, tc, s->session_index); + + return 0; +} + +void +stream_session_stop_listen (u32 server_index) +{ + session_manager_main_t *smm = &session_manager_main; + stream_session_t *listener; + transport_connection_t *tc; + application_t *srv; + + srv = application_get (server_index); + listener = pool_elt_at_index (smm->listen_sessions[srv->session_type], + srv->session_index); + + tc = tp_vfts[srv->session_type].get_listener (listener->connection_index); + stream_session_table_del_for_tc (smm, listener->session_type, tc); + + tp_vfts[srv->session_type].unbind (smm->vlib_main, + listener->connection_index); + pool_put (smm->listen_sessions[srv->session_type], listener); +} + +int +connect_server_add_segment_cb (application_t * ss, char *segment_name, + u32 segment_size) +{ + /* Does exactly nothing, but die */ + ASSERT (0); + return 0; +} + +void +connects_session_manager_init (session_manager_main_t * smm, u8 session_type) +{ + session_manager_t *sm; + u32 connect_fifo_size = 8 << 10; /* Config? */ + u32 default_segment_size = 1 << 20; + + pool_get (smm->session_managers, sm); + memset (sm, 0, sizeof (*sm)); + + sm->add_segment_size = default_segment_size; + sm->rx_fifo_size = connect_fifo_size; + sm->tx_fifo_size = connect_fifo_size; + sm->add_segment = 1; + + session_manager_add_segment (smm, sm); + smm->connect_manager_index[session_type] = sm - smm->session_managers; +} + +void +stream_session_connect_notify (transport_connection_t * tc, u8 sst, + u8 is_fail) +{ + session_manager_main_t *smm = &session_manager_main; + application_t *app; + stream_session_t *new_s = 0; + u64 value; + + value = stream_session_half_open_lookup (smm, &tc->lcl_ip, &tc->rmt_ip, + tc->lcl_port, tc->rmt_port, + tc->proto); + if (value == HALF_OPEN_LOOKUP_INVALID_VALUE) + { + clib_warning ("This can't be good!"); + return; + } + + app = application_get (value >> 32); + + if (!is_fail) + { + /* Create new session (server segments are allocated if needed) */ + if (stream_session_create_i (smm, app, tc, &new_s)) + return; + + app->session_index = stream_session_get_index (new_s); + app->thread_index = new_s->thread_index; + + /* Allocate vpp event queue for this thread if needed */ + vpp_session_event_queue_allocate (smm, tc->thread_index); + } + + /* Notify client */ + app->cb_fns.session_connected_callback (app->api_client_index, new_s, + is_fail); + + /* Cleanup session lookup */ + stream_session_half_open_table_del (smm, sst, tc); +} + +void +stream_session_accept_notify (transport_connection_t * tc) +{ + application_t *server; + stream_session_t *s; + + s = stream_session_get (tc->s_index, tc->thread_index); + server = application_get (s->app_index); + server->cb_fns.session_accept_callback (s); +} + +/** + * Notification from transport that connection is being closed. + * + * A disconnect is sent to application but state is not removed. Once + * disconnect is acknowledged by application, session disconnect is called. + * Ultimately this leads to close being called on transport (passive close). + */ +void +stream_session_disconnect_notify (transport_connection_t * tc) +{ + application_t *server; + stream_session_t *s; + + s = stream_session_get (tc->s_index, tc->thread_index); + server = application_get (s->app_index); + server->cb_fns.session_disconnect_callback (s); +} + +/** + * Cleans up session and associated app if needed. + */ +void +stream_session_delete (stream_session_t * s) +{ + session_manager_main_t *smm = vnet_get_session_manager_main (); + svm_fifo_segment_private_t *fifo_segment; + application_t *app; + int rv; + + /* delete from the main lookup table */ + rv = stream_session_table_del (smm, s); + + if (rv) + clib_warning ("hash delete error, rv %d", rv); + + /* Cleanup fifo segments */ + fifo_segment = svm_fifo_get_segment (s->server_segment_index); + svm_fifo_segment_free_fifo (fifo_segment, s->server_rx_fifo); + svm_fifo_segment_free_fifo (fifo_segment, s->server_tx_fifo); + + /* Cleanup app if client */ + app = application_get (s->app_index); + if (app->mode == APP_CLIENT) + { + application_del (app); + } + else if (app->mode == APP_SERVER) + { + session_manager_t *sm; + svm_fifo_segment_private_t *fifo_segment; + svm_fifo_t **fifos; + u32 fifo_index; + + sm = session_manager_get (app->session_manager_index); + + /* Delete fifo */ + fifo_segment = svm_fifo_get_segment (s->server_segment_index); + fifos = (svm_fifo_t **) fifo_segment->h->fifos; + + fifo_index = svm_fifo_segment_index (fifo_segment); + + /* Remove segment only if it holds no fifos and not the first */ + if (sm->segment_indices[0] != fifo_index && vec_len (fifos) == 0) + svm_fifo_segment_delete (fifo_segment); + } + + pool_put (smm->sessions[s->thread_index], s); +} + +/** + * Notification from transport that connection is being deleted + * + * This should be called only on previously fully established sessions. For + * instance failed connects should call stream_session_connect_notify and + * indicate that the connect has failed. + */ +void +stream_session_delete_notify (transport_connection_t * tc) +{ + stream_session_t *s; + + s = stream_session_get_if_valid (tc->s_index, tc->thread_index); + if (!s) + { + clib_warning ("Surprised!"); + return; + } + stream_session_delete (s); +} + +/** + * Notify application that connection has been reset. + */ +void +stream_session_reset_notify (transport_connection_t * tc) +{ + stream_session_t *s; + application_t *app; + s = stream_session_get (tc->s_index, tc->thread_index); + + app = application_get (s->app_index); + app->cb_fns.session_reset_callback (s); +} + +/** + * Accept a stream session. Optionally ping the server by callback. + */ +int +stream_session_accept (transport_connection_t * tc, u32 listener_index, + u8 sst, u8 notify) +{ + session_manager_main_t *smm = &session_manager_main; + application_t *server; + stream_session_t *s, *listener; + + int rv; + + /* Find the server */ + listener = pool_elt_at_index (smm->listen_sessions[sst], listener_index); + server = application_get (listener->app_index); + + if ((rv = stream_session_create_i (smm, server, tc, &s))) + return rv; + + /* Allocate vpp event queue for this thread if needed */ + vpp_session_event_queue_allocate (smm, tc->thread_index); + + /* Shoulder-tap the server */ + if (notify) + { + server->cb_fns.session_accept_callback (s); + } + + return 0; +} + +void +stream_session_open (u8 sst, ip46_address_t * addr, u16 port_host_byte_order, + u32 app_index) +{ + transport_connection_t *tc; + u32 tci; + u64 value; + + /* Ask transport to open connection */ + tci = tp_vfts[sst].open (addr, port_host_byte_order); + + /* Get transport connection */ + tc = tp_vfts[sst].get_half_open (tci); + + /* Store api_client_index and transport connection index */ + value = (((u64) app_index) << 32) | (u64) tc->c_index; + + /* Add to the half-open lookup table */ + stream_session_half_open_table_add (sst, tc, value); +} + +/** + * Disconnect session and propagate to transport. This should eventually + * result in a delete notification that allows us to cleanup session state. + * Called for both active/passive disconnects. + */ +void +stream_session_disconnect (stream_session_t * s) +{ + tp_vfts[s->session_type].close (s->connection_index, s->thread_index); + s->session_state = SESSION_STATE_CLOSED; +} + +/** + * Cleanup transport and session state. + */ +void +stream_session_cleanup (stream_session_t * s) +{ + tp_vfts[s->session_type].cleanup (s->connection_index, s->thread_index); + stream_session_delete (s); +} + +void +session_register_transport (u8 type, const transport_proto_vft_t * vft) +{ + session_manager_main_t *smm = vnet_get_session_manager_main (); + + vec_validate (tp_vfts, type); + tp_vfts[type] = *vft; + + /* If an offset function is provided, then peek instead of dequeue */ + smm->session_rx_fns[type] = + (vft->rx_fifo_offset) ? session_fifo_rx_peek : session_fifo_rx_dequeue; +} + +transport_proto_vft_t * +session_get_transport_vft (u8 type) +{ + if (type >= vec_len (tp_vfts)) + return 0; + return &tp_vfts[type]; +} + +static clib_error_t * +session_manager_main_init (vlib_main_t * vm) +{ + u32 num_threads; + vlib_thread_main_t *vtm = vlib_get_thread_main (); + session_manager_main_t *smm = &session_manager_main; + int i; + + smm->vlib_main = vm; + smm->vnet_main = vnet_get_main (); + + num_threads = 1 /* main thread */ + vtm->n_threads; + + if (num_threads < 1) + return clib_error_return (0, "n_thread_stacks not set"); + + /* $$$ config parameters */ + svm_fifo_segment_init (0x200000000ULL /* first segment base VA */ , + 20 /* timeout in seconds */ ); + + /* configure per-thread ** vectors */ + vec_validate (smm->sessions, num_threads - 1); + vec_validate (smm->session_indices_to_enqueue_by_thread, num_threads - 1); + vec_validate (smm->tx_buffers, num_threads - 1); + vec_validate (smm->fifo_events, num_threads - 1); + vec_validate (smm->evts_partially_read, num_threads - 1); + vec_validate (smm->current_enqueue_epoch, num_threads - 1); + vec_validate (smm->vpp_event_queues, num_threads - 1); + + /* $$$$ preallocate hack config parameter */ + for (i = 0; i < 200000; i++) + { + stream_session_t *ss; + pool_get (smm->sessions[0], ss); + memset (ss, 0, sizeof (*ss)); + } + + for (i = 0; i < 200000; i++) + pool_put_index (smm->sessions[0], i); + + clib_bihash_init_16_8 (&smm->v4_session_hash, "v4 session table", + 200000 /* $$$$ config parameter nbuckets */ , + (64 << 20) /*$$$ config parameter table size */ ); + clib_bihash_init_48_8 (&smm->v6_session_hash, "v6 session table", + 200000 /* $$$$ config parameter nbuckets */ , + (64 << 20) /*$$$ config parameter table size */ ); + + clib_bihash_init_16_8 (&smm->v4_half_open_hash, "v4 half-open table", + 200000 /* $$$$ config parameter nbuckets */ , + (64 << 20) /*$$$ config parameter table size */ ); + clib_bihash_init_48_8 (&smm->v6_half_open_hash, "v6 half-open table", + 200000 /* $$$$ config parameter nbuckets */ , + (64 << 20) /*$$$ config parameter table size */ ); + + for (i = 0; i < SESSION_N_TYPES; i++) + smm->connect_manager_index[i] = INVALID_INDEX; + + return 0; +} + +VLIB_INIT_FUNCTION (session_manager_main_init); + +/* + * fd.io coding-style-patch-verification: ON + * + * Local Variables: + * eval: (c-set-style "gnu") + * End: + */ diff --git a/src/vnet/session/session.h b/src/vnet/session/session.h new file mode 100644 index 00000000..cf14cca9 --- /dev/null +++ b/src/vnet/session/session.h @@ -0,0 +1,380 @@ +/* + * Copyright (c) 2017 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ +#ifndef __included_session_h__ +#define __included_session_h__ + +#include +#include +#include +#include +#include + +#define HALF_OPEN_LOOKUP_INVALID_VALUE ((u64)~0) +#define INVALID_INDEX ((u32)~0) + +/* TODO decide how much since we have pre-data as well */ +#define MAX_HDRS_LEN 100 /* Max number of bytes for headers */ + +typedef enum +{ + FIFO_EVENT_SERVER_RX, + FIFO_EVENT_SERVER_TX, + FIFO_EVENT_TIMEOUT, + FIFO_EVENT_SERVER_EXIT, +} fifo_event_type_t; + +#define foreach_session_input_error \ +_(NO_SESSION, "No session drops") \ +_(NO_LISTENER, "No listener for dst port drops") \ +_(ENQUEUED, "Packets pushed into rx fifo") \ +_(NOT_READY, "Session not ready packets") \ +_(FIFO_FULL, "Packets dropped for lack of rx fifo space") \ +_(EVENT_FIFO_FULL, "Events not sent for lack of event fifo space") \ +_(API_QUEUE_FULL, "Sessions not created for lack of API queue space") \ +_(NEW_SEG_NO_SPACE, "Created segment, couldn't allocate a fifo pair") \ +_(NO_SPACE, "Couldn't allocate a fifo pair") + +typedef enum +{ +#define _(sym,str) SESSION_ERROR_##sym, + foreach_session_input_error +#undef _ + SESSION_N_ERROR, +} session_error_t; + +/* Event queue input node static next indices */ +typedef enum +{ + SESSION_QUEUE_NEXT_DROP, + SESSION_QUEUE_NEXT_TCP_IP4_OUTPUT, + SESSION_QUEUE_NEXT_IP4_LOOKUP, + SESSION_QUEUE_NEXT_TCP_IP6_OUTPUT, + SESSION_QUEUE_NEXT_IP6_LOOKUP, + SESSION_QUEUE_N_NEXT, +} session_queue_next_t; + +#define foreach_session_type \ + _(IP4_TCP, ip4_tcp) \ + _(IP4_UDP, ip4_udp) \ + _(IP6_TCP, ip6_tcp) \ + _(IP6_UDP, ip6_udp) + +typedef enum +{ +#define _(A, a) SESSION_TYPE_##A, + foreach_session_type +#undef _ + SESSION_N_TYPES, +} session_type_t; + +/* + * Application session state + */ +typedef enum +{ + SESSION_STATE_LISTENING, + SESSION_STATE_CONNECTING, + SESSION_STATE_READY, + SESSION_STATE_CLOSED, + SESSION_STATE_N_STATES, +} stream_session_state_t; + +typedef CLIB_PACKED (struct + { + svm_fifo_t * fifo; + u8 event_type; + /* $$$$ for event logging */ + u16 event_id; + u32 enqueue_length; + }) session_fifo_event_t; + +typedef struct _stream_session_t +{ + /** Type */ + u8 session_type; + + /** State */ + u8 session_state; + + /** Session index in per_thread pool */ + u32 session_index; + + /** Transport specific */ + u32 connection_index; + + u8 thread_index; + + /** Application specific */ + u32 pid; + + /** fifo pointers. Once allocated, these do not move */ + svm_fifo_t *server_rx_fifo; + svm_fifo_t *server_tx_fifo; + + /** To avoid n**2 "one event per frame" check */ + u8 enqueue_epoch; + + /** used during unbind processing */ + u8 is_deleted; + + /** stream server pool index */ + u32 app_index; + + /** svm segment index */ + u32 server_segment_index; +} stream_session_t; + +typedef struct _session_manager +{ + /** segments mapped by this server */ + u32 *segment_indices; + + /** Session fifo sizes. They are provided for binds and take default + * values for connects */ + u32 rx_fifo_size; + u32 tx_fifo_size; + + /** Configured additional segment size */ + u32 add_segment_size; + + /** Flag that indicates if additional segments should be created */ + u8 add_segment; +} session_manager_t; + +/* Forward definition */ +typedef struct _session_manager_main session_manager_main_t; + +typedef int + (session_fifo_rx_fn) (vlib_main_t * vm, vlib_node_runtime_t * node, + session_manager_main_t * smm, + session_fifo_event_t * e0, stream_session_t * s0, + u32 thread_index, int *n_tx_pkts); + +extern session_fifo_rx_fn session_fifo_rx_peek; +extern session_fifo_rx_fn session_fifo_rx_dequeue; + +struct _session_manager_main +{ + /** Lookup tables for established sessions and listeners */ + clib_bihash_16_8_t v4_session_hash; + clib_bihash_48_8_t v6_session_hash; + + /** Lookup tables for half-open sessions */ + clib_bihash_16_8_t v4_half_open_hash; + clib_bihash_48_8_t v6_half_open_hash; + + /** Per worker thread session pools */ + stream_session_t **sessions; + + /** Pool of listen sessions. Same type as stream sessions to ease lookups */ + stream_session_t *listen_sessions[SESSION_N_TYPES]; + + /** Sparse vector to map dst port to stream server */ + u16 *stream_server_by_dst_port[SESSION_N_TYPES]; + + /** per-worker enqueue epoch counters */ + u8 *current_enqueue_epoch; + + /** Per-worker thread vector of sessions to enqueue */ + u32 **session_indices_to_enqueue_by_thread; + + /** per-worker tx buffer free lists */ + u32 **tx_buffers; + + /** Per worker-thread vector of partially read events */ + session_fifo_event_t **evts_partially_read; + + /** per-worker active event vectors */ + session_fifo_event_t **fifo_events; + + /** vpp fifo event queue */ + unix_shared_memory_queue_t **vpp_event_queues; + + /** Unique segment name counter */ + u32 unique_segment_name_counter; + + /* Connection manager used by incoming connects */ + u32 connect_manager_index[SESSION_N_TYPES]; + + session_manager_t *session_managers; + + /** Per transport rx function that can either dequeue or peek */ + session_fifo_rx_fn *session_rx_fns[SESSION_N_TYPES]; + + /* Convenience */ + vlib_main_t *vlib_main; + vnet_main_t *vnet_main; +}; + +extern session_manager_main_t session_manager_main; + +/* + * Session manager function + */ +always_inline session_manager_main_t * +vnet_get_session_manager_main () +{ + return &session_manager_main; +} + +always_inline session_manager_t * +session_manager_get (u32 index) +{ + return pool_elt_at_index (session_manager_main.session_managers, index); +} + +always_inline unix_shared_memory_queue_t * +session_manager_get_vpp_event_queue (u32 thread_index) +{ + return session_manager_main.vpp_event_queues[thread_index]; +} + +always_inline session_manager_t * +connects_session_manager_get (session_manager_main_t * smm, + session_type_t session_type) +{ + return pool_elt_at_index (smm->session_managers, + smm->connect_manager_index[session_type]); +} + +void session_manager_get_segment_info (u32 index, u8 ** name, u32 * size); +int session_manager_flush_enqueue_events (u32 thread_index); +int +session_manager_add_first_segment (session_manager_main_t * smm, + session_manager_t * sm, u32 segment_size, + u8 ** segment_name); +void +session_manager_del (session_manager_main_t * smm, session_manager_t * sm); +void +connects_session_manager_init (session_manager_main_t * smm, u8 session_type); + +/* + * Stream session functions + */ + +stream_session_t *stream_session_lookup_listener4 (ip4_address_t * lcl, + u16 lcl_port, u8 proto); +stream_session_t *stream_session_lookup4 (ip4_address_t * lcl, + ip4_address_t * rmt, u16 lcl_port, + u16 rmt_port, u8 proto, + u32 thread_index); +stream_session_t *stream_session_lookup_listener6 (ip6_address_t * lcl, + u16 lcl_port, u8 proto); +stream_session_t *stream_session_lookup6 (ip6_address_t * lcl, + ip6_address_t * rmt, u16 lcl_port, + u16 rmt_port, u8, u32 thread_index); +transport_connection_t + * stream_session_lookup_transport4 (session_manager_main_t * smm, + ip4_address_t * lcl, + ip4_address_t * rmt, u16 lcl_port, + u16 rmt_port, u8 proto, + u32 thread_index); +transport_connection_t + * stream_session_lookup_transport6 (session_manager_main_t * smm, + ip6_address_t * lcl, + ip6_address_t * rmt, u16 lcl_port, + u16 rmt_port, u8 proto, + u32 thread_index); +stream_session_t *stream_session_lookup_listener (ip46_address_t * lcl, + u16 lcl_port, u8 proto); + +always_inline stream_session_t * +stream_session_get_tsi (u64 ti_and_si, u32 thread_index) +{ + ASSERT ((u32) (ti_and_si >> 32) == thread_index); + return pool_elt_at_index (session_manager_main.sessions[thread_index], + ti_and_si & 0xFFFFFFFFULL); +} + +always_inline stream_session_t * +stream_session_get (u64 si, u32 thread_index) +{ + return pool_elt_at_index (session_manager_main.sessions[thread_index], si); +} + +always_inline stream_session_t * +stream_session_get_if_valid (u64 si, u32 thread_index) +{ + if (thread_index >= vec_len (session_manager_main.sessions)) + return 0; + + if (pool_is_free_index (session_manager_main.sessions[thread_index], si)) + return 0; + + return pool_elt_at_index (session_manager_main.sessions[thread_index], si); +} + +always_inline stream_session_t * +stream_session_listener_get (u8 sst, u64 si) +{ + return pool_elt_at_index (session_manager_main.listen_sessions[sst], si); +} + +always_inline u32 +stream_session_get_index (stream_session_t * s) +{ + if (s->session_state == SESSION_STATE_LISTENING) + return s - session_manager_main.listen_sessions[s->session_type]; + + return s - session_manager_main.sessions[s->thread_index]; +} + +always_inline u32 +stream_session_max_enqueue (transport_connection_t * tc) +{ + stream_session_t *s = stream_session_get (tc->s_index, tc->thread_index); + return svm_fifo_max_enqueue (s->server_rx_fifo); +} + +int +stream_session_enqueue_data (transport_connection_t * tc, u8 * data, u16 len, + u8 queue_event); +u32 +stream_session_peek_bytes (transport_connection_t * tc, u8 * buffer, + u32 offset, u32 max_bytes); +u32 stream_session_dequeue_drop (transport_connection_t * tc, u32 max_bytes); + +void +stream_session_connect_notify (transport_connection_t * tc, u8 sst, + u8 is_fail); +void stream_session_accept_notify (transport_connection_t * tc); +void stream_session_disconnect_notify (transport_connection_t * tc); +void stream_session_delete_notify (transport_connection_t * tc); +void stream_session_reset_notify (transport_connection_t * tc); +int +stream_session_accept (transport_connection_t * tc, u32 listener_index, + u8 sst, u8 notify); +void stream_session_open (u8 sst, ip46_address_t * addr, + u16 port_host_byte_order, u32 api_client_index); +void stream_session_disconnect (stream_session_t * s); +void stream_session_cleanup (stream_session_t * s); +int +stream_session_start_listen (u32 server_index, ip46_address_t * ip, u16 port); +void stream_session_stop_listen (u32 server_index); + +u8 *format_stream_session (u8 * s, va_list * args); + +void session_register_transport (u8 type, const transport_proto_vft_t * vft); +transport_proto_vft_t *session_get_transport_vft (u8 type); + +#endif /* __included_session_h__ */ + +/* + * fd.io coding-style-patch-verification: ON + * + * Local Variables: + * eval: (c-set-style "gnu") + * End: + */ diff --git a/src/vnet/session/session_api.c b/src/vnet/session/session_api.c new file mode 100644 index 00000000..9d068684 --- /dev/null +++ b/src/vnet/session/session_api.c @@ -0,0 +1,821 @@ +/* + * Copyright (c) 2015-2016 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +#include +#include +#include + +#include +#include "application_interface.h" + +#define vl_typedefs /* define message structures */ +#include +#undef vl_typedefs + +#define vl_endianfun /* define message structures */ +#include +#undef vl_endianfun + +/* instantiate all the print functions we know about */ +#define vl_print(handle, ...) vlib_cli_output (handle, __VA_ARGS__) +#define vl_printfun +#include +#undef vl_printfun + +#include + +#define foreach_session_api_msg \ +_(MAP_ANOTHER_SEGMENT_REPLY, map_another_segment_reply) \ +_(BIND_URI, bind_uri) \ +_(UNBIND_URI, unbind_uri) \ +_(CONNECT_URI, connect_uri) \ +_(DISCONNECT_SESSION, disconnect_session) \ +_(DISCONNECT_SESSION_REPLY, disconnect_session_reply) \ +_(ACCEPT_SESSION_REPLY, accept_session_reply) \ +_(RESET_SESSION_REPLY, reset_session_reply) \ +_(BIND_SOCK, bind_sock) \ +_(UNBIND_SOCK, unbind_sock) \ +_(CONNECT_SOCK, connect_sock) \ +_(DISCONNECT_SOCK, disconnect_sock) \ +_(DISCONNECT_SOCK_REPLY, disconnect_sock_reply) \ +_(ACCEPT_SOCK_REPLY, accept_sock_reply) \ +_(RESET_SOCK_REPLY, reset_sock_reply) \ + +static int +send_add_segment_callback (u32 api_client_index, const u8 * segment_name, + u32 segment_size) +{ + vl_api_map_another_segment_t *mp; + unix_shared_memory_queue_t *q; + + q = vl_api_client_index_to_input_queue (api_client_index); + + if (!q) + return -1; + + mp = vl_msg_api_alloc (sizeof (*mp)); + memset (mp, 0, sizeof (*mp)); + mp->_vl_msg_id = clib_host_to_net_u16 (VL_API_MAP_ANOTHER_SEGMENT); + mp->segment_size = segment_size; + strncpy ((char *) mp->segment_name, (char *) segment_name, + sizeof (mp->segment_name) - 1); + + vl_msg_api_send_shmem (q, (u8 *) & mp); + + return 0; +} + +static int +send_session_accept_uri_callback (stream_session_t * s) +{ + vl_api_accept_session_t *mp; + unix_shared_memory_queue_t *q, *vpp_queue; + application_t *server = application_get (s->app_index); + + q = vl_api_client_index_to_input_queue (server->api_client_index); + vpp_queue = session_manager_get_vpp_event_queue (s->thread_index); + + if (!q) + return -1; + + mp = vl_msg_api_alloc (sizeof (*mp)); + mp->_vl_msg_id = clib_host_to_net_u16 (VL_API_ACCEPT_SESSION); + + /* Note: session_type is the first octet in all types of sessions */ + + mp->accept_cookie = server->accept_cookie; + mp->server_rx_fifo = (u64) s->server_rx_fifo; + mp->server_tx_fifo = (u64) s->server_tx_fifo; + mp->session_thread_index = s->thread_index; + mp->session_index = s->session_index; + mp->session_type = s->session_type; + mp->vpp_event_queue_address = (u64) vpp_queue; + vl_msg_api_send_shmem (q, (u8 *) & mp); + + return 0; +} + +static void +send_session_disconnect_uri_callback (stream_session_t * s) +{ + vl_api_disconnect_session_t *mp; + unix_shared_memory_queue_t *q; + application_t *app = application_get (s->app_index); + + q = vl_api_client_index_to_input_queue (app->api_client_index); + + if (!q) + return; + + mp = vl_msg_api_alloc (sizeof (*mp)); + memset (mp, 0, sizeof (*mp)); + mp->_vl_msg_id = clib_host_to_net_u16 (VL_API_DISCONNECT_SESSION); + + mp->session_thread_index = s->thread_index; + mp->session_index = s->session_index; + vl_msg_api_send_shmem (q, (u8 *) & mp); +} + +static int +send_session_connected_uri_callback (u32 api_client_index, + stream_session_t * s, u8 is_fail) +{ + vl_api_connect_uri_reply_t *mp; + unix_shared_memory_queue_t *q; + application_t *app = application_lookup (api_client_index); + u8 *seg_name; + unix_shared_memory_queue_t *vpp_queue; + + q = vl_api_client_index_to_input_queue (app->api_client_index); + + if (!q) + return -1; + + mp = vl_msg_api_alloc (sizeof (*mp)); + mp->_vl_msg_id = clib_host_to_net_u16 (VL_API_CONNECT_URI_REPLY); + mp->context = app->api_context; + mp->retval = is_fail; + if (!is_fail) + { + vpp_queue = session_manager_get_vpp_event_queue (s->thread_index); + mp->server_rx_fifo = (u64) s->server_rx_fifo; + mp->server_tx_fifo = (u64) s->server_tx_fifo; + mp->session_thread_index = s->thread_index; + mp->session_index = s->session_index; + mp->session_type = s->session_type; + mp->vpp_event_queue_address = (u64) vpp_queue; + mp->client_event_queue_address = (u64) app->event_queue; + + session_manager_get_segment_info (s->server_segment_index, &seg_name, + &mp->segment_size); + mp->segment_name_length = vec_len (seg_name); + if (mp->segment_name_length) + clib_memcpy (mp->segment_name, seg_name, mp->segment_name_length); + } + + vl_msg_api_send_shmem (q, (u8 *) & mp); + + /* Remove client if connect failed */ + if (is_fail) + application_del (app); + + return 0; +} + +/** + * Redirect a connect_uri message to the indicated server. + * Only sent if the server has bound the related port with + * URI_OPTIONS_FLAGS_USE_FIFO + */ +static int +redirect_connect_uri_callback (u32 server_api_client_index, void *mp_arg) +{ + vl_api_connect_uri_t *mp = mp_arg; + unix_shared_memory_queue_t *server_q, *client_q; + vlib_main_t *vm = vlib_get_main (); + f64 timeout = vlib_time_now (vm) + 0.5; + int rv = 0; + + server_q = vl_api_client_index_to_input_queue (server_api_client_index); + + if (!server_q) + { + rv = VNET_API_ERROR_INVALID_VALUE; + goto out; + } + + client_q = vl_api_client_index_to_input_queue (mp->client_index); + if (!client_q) + { + rv = VNET_API_ERROR_INVALID_VALUE_2; + goto out; + } + + /* Tell the server the client's API queue address, so it can reply */ + mp->client_queue_address = (u64) client_q; + + /* + * Bounce message handlers MUST NOT block the data-plane. + * Spin waiting for the queue lock, but + */ + + while (vlib_time_now (vm) < timeout) + { + rv = + unix_shared_memory_queue_add (server_q, (u8 *) & mp, 1 /*nowait */ ); + switch (rv) + { + /* correctly enqueued */ + case 0: + return VNET_CONNECT_REDIRECTED; + + /* continue spinning, wait for pthread_mutex_trylock to work */ + case -1: + continue; + + /* queue stuffed, drop the msg */ + case -2: + rv = VNET_API_ERROR_QUEUE_FULL; + goto out; + } + } +out: + /* Dispose of the message */ + vl_msg_api_free (mp); + return rv; +} + +static u64 +make_session_handle (stream_session_t * s) +{ + return (u64) s->session_index << 32 | (u64) s->thread_index; +} + +static int +send_session_accept_callback (stream_session_t * s) +{ + vl_api_accept_sock_t *mp; + unix_shared_memory_queue_t *q, *vpp_queue; + application_t *server = application_get (s->app_index); + + q = vl_api_client_index_to_input_queue (server->api_client_index); + vpp_queue = session_manager_get_vpp_event_queue (s->thread_index); + + if (!q) + return -1; + + mp = vl_msg_api_alloc (sizeof (*mp)); + mp->_vl_msg_id = clib_host_to_net_u16 (VL_API_ACCEPT_SOCK); + + /* Note: session_type is the first octet in all types of sessions */ + + mp->accept_cookie = server->accept_cookie; + mp->server_rx_fifo = (u64) s->server_rx_fifo; + mp->server_tx_fifo = (u64) s->server_tx_fifo; + mp->handle = make_session_handle (s); + mp->vpp_event_queue_address = (u64) vpp_queue; + vl_msg_api_send_shmem (q, (u8 *) & mp); + + return 0; +} + +static int +send_session_connected_callback (u32 api_client_index, stream_session_t * s, + u8 is_fail) +{ + vl_api_connect_sock_reply_t *mp; + unix_shared_memory_queue_t *q; + application_t *app = application_lookup (api_client_index); + u8 *seg_name; + unix_shared_memory_queue_t *vpp_queue; + + q = vl_api_client_index_to_input_queue (app->api_client_index); + + if (!q) + return -1; + + mp = vl_msg_api_alloc (sizeof (*mp)); + mp->_vl_msg_id = clib_host_to_net_u16 (VL_API_CONNECT_SOCK_REPLY); + mp->context = app->api_context; + mp->retval = is_fail; + if (!is_fail) + { + vpp_queue = session_manager_get_vpp_event_queue (s->thread_index); + mp->server_rx_fifo = (u64) s->server_rx_fifo; + mp->server_tx_fifo = (u64) s->server_tx_fifo; + mp->handle = make_session_handle (s); + mp->vpp_event_queue_address = (u64) vpp_queue; + mp->client_event_queue_address = (u64) app->event_queue; + + session_manager_get_segment_info (s->server_segment_index, &seg_name, + &mp->segment_size); + mp->segment_name_length = vec_len (seg_name); + if (mp->segment_name_length) + clib_memcpy (mp->segment_name, seg_name, mp->segment_name_length); + } + + vl_msg_api_send_shmem (q, (u8 *) & mp); + + /* Remove client if connect failed */ + if (is_fail) + application_del (app); + + return 0; +} + +static void +send_session_disconnect_callback (stream_session_t * s) +{ + vl_api_disconnect_sock_t *mp; + unix_shared_memory_queue_t *q; + application_t *app = application_get (s->app_index); + + q = vl_api_client_index_to_input_queue (app->api_client_index); + + if (!q) + return; + + mp = vl_msg_api_alloc (sizeof (*mp)); + memset (mp, 0, sizeof (*mp)); + mp->_vl_msg_id = clib_host_to_net_u16 (VL_API_DISCONNECT_SOCK); + + mp->handle = make_session_handle (s); + vl_msg_api_send_shmem (q, (u8 *) & mp); +} + +/** + * Redirect a connect_uri message to the indicated server. + * Only sent if the server has bound the related port with + * URI_OPTIONS_FLAGS_USE_FIFO + */ +static int +redirect_connect_callback (u32 server_api_client_index, void *mp_arg) +{ + vl_api_connect_sock_t *mp = mp_arg; + unix_shared_memory_queue_t *server_q, *client_q; + vlib_main_t *vm = vlib_get_main (); + f64 timeout = vlib_time_now (vm) + 0.5; + int rv = 0; + + server_q = vl_api_client_index_to_input_queue (server_api_client_index); + + if (!server_q) + { + rv = VNET_API_ERROR_INVALID_VALUE; + goto out; + } + + client_q = vl_api_client_index_to_input_queue (mp->client_index); + if (!client_q) + { + rv = VNET_API_ERROR_INVALID_VALUE_2; + goto out; + } + + /* Tell the server the client's API queue address, so it can reply */ + mp->client_queue_address = (u64) client_q; + + /* + * Bounce message handlers MUST NOT block the data-plane. + * Spin waiting for the queue lock, but + */ + + while (vlib_time_now (vm) < timeout) + { + rv = + unix_shared_memory_queue_add (server_q, (u8 *) & mp, 1 /*nowait */ ); + switch (rv) + { + /* correctly enqueued */ + case 0: + return VNET_CONNECT_REDIRECTED; + + /* continue spinning, wait for pthread_mutex_trylock to work */ + case -1: + continue; + + /* queue stuffed, drop the msg */ + case -2: + rv = VNET_API_ERROR_QUEUE_FULL; + goto out; + } + } +out: + /* Dispose of the message */ + vl_msg_api_free (mp); + return rv; +} + +static session_cb_vft_t uri_session_cb_vft = { + .session_accept_callback = send_session_accept_uri_callback, + .session_disconnect_callback = send_session_disconnect_uri_callback, + .session_connected_callback = send_session_connected_uri_callback, + .add_segment_callback = send_add_segment_callback, + .redirect_connect_callback = redirect_connect_uri_callback +}; + +static session_cb_vft_t session_cb_vft = { + .session_accept_callback = send_session_accept_callback, + .session_disconnect_callback = send_session_disconnect_callback, + .session_connected_callback = send_session_connected_callback, + .add_segment_callback = send_add_segment_callback, + .redirect_connect_callback = redirect_connect_callback +}; + +static int +api_session_not_valid (u32 session_index, u32 thread_index) +{ + session_manager_main_t *smm = vnet_get_session_manager_main (); + stream_session_t *pool; + + if (thread_index >= vec_len (smm->sessions)) + return VNET_API_ERROR_INVALID_VALUE; + + pool = smm->sessions[thread_index]; + + if (pool_is_free_index (pool, session_index)) + return VNET_API_ERROR_INVALID_VALUE_2; + + return 0; +} + +static void +vl_api_bind_uri_t_handler (vl_api_bind_uri_t * mp) +{ + vl_api_bind_uri_reply_t *rmp; + vnet_bind_args_t _a, *a = &_a; + char segment_name[128]; + u32 segment_name_length; + int rv; + + _Static_assert (sizeof (u64) * SESSION_OPTIONS_N_OPTIONS <= + sizeof (mp->options), + "Out of options, fix api message definition"); + + segment_name_length = ARRAY_LEN (segment_name); + + memset (a, 0, sizeof (*a)); + + a->uri = (char *) mp->uri; + a->api_client_index = mp->client_index; + a->options = mp->options; + a->segment_name = segment_name; + a->segment_name_length = segment_name_length; + a->session_cb_vft = &uri_session_cb_vft; + + a->options[SESSION_OPTIONS_SEGMENT_SIZE] = mp->initial_segment_size; + a->options[SESSION_OPTIONS_ACCEPT_COOKIE] = mp->accept_cookie; + rv = vnet_bind_uri (a); + + /* *INDENT-OFF* */ + REPLY_MACRO2 (VL_API_BIND_URI_REPLY, ({ + rmp->retval = rv; + if (!rv) + { + rmp->segment_name_length = 0; + /* $$$$ policy? */ + rmp->segment_size = mp->initial_segment_size; + if (segment_name_length) + { + memcpy (rmp->segment_name, segment_name, segment_name_length); + rmp->segment_name_length = segment_name_length; + } + rmp->server_event_queue_address = a->server_event_queue_address; + } + })); + /* *INDENT-ON* */ + +} + +static void +vl_api_unbind_uri_t_handler (vl_api_unbind_uri_t * mp) +{ + vl_api_unbind_uri_reply_t *rmp; + int rv; + + rv = vnet_unbind_uri ((char *) mp->uri, mp->client_index); + + REPLY_MACRO (VL_API_UNBIND_URI_REPLY); +} + +static void +vl_api_connect_uri_t_handler (vl_api_connect_uri_t * mp) +{ + vnet_connect_args_t _a, *a = &_a; + + a->uri = (char *) mp->uri; + a->api_client_index = mp->client_index; + a->api_context = mp->context; + a->options = mp->options; + a->session_cb_vft = &uri_session_cb_vft; + a->mp = mp; + vnet_connect_uri (a); +} + +static void +vl_api_disconnect_session_t_handler (vl_api_disconnect_session_t * mp) +{ + vl_api_disconnect_session_reply_t *rmp; + int rv; + + rv = api_session_not_valid (mp->session_index, mp->session_thread_index); + if (!rv) + rv = vnet_disconnect_session (mp->client_index, mp->session_index, + mp->session_thread_index); + + REPLY_MACRO (VL_API_DISCONNECT_SESSION_REPLY); +} + +static void +vl_api_disconnect_session_reply_t_handler (vl_api_disconnect_session_reply_t * + mp) +{ + if (api_session_not_valid (mp->session_index, mp->session_thread_index)) + { + clib_warning ("Invalid session!"); + return; + } + + /* Client objected to disconnecting the session, log and continue */ + if (mp->retval) + { + clib_warning ("client retval %d", mp->retval); + return; + } + + /* Disconnect has been confirmed. Confirm close to transport */ + vnet_disconnect_session (mp->client_index, mp->session_index, + mp->session_thread_index); +} + +static void +vl_api_reset_session_reply_t_handler (vl_api_reset_session_reply_t * mp) +{ + stream_session_t *s; + + if (api_session_not_valid (mp->session_index, mp->session_thread_index)) + { + clib_warning ("Invalid session!"); + return; + } + + /* Client objected to resetting the session, log and continue */ + if (mp->retval) + { + clib_warning ("client retval %d", mp->retval); + return; + } + + s = stream_session_get (mp->session_index, mp->session_thread_index); + + /* This comes as a response to a reset, transport only waiting for + * confirmation to remove connection state, no need to disconnect */ + stream_session_cleanup (s); +} + +static void +vl_api_accept_session_reply_t_handler (vl_api_accept_session_reply_t * mp) +{ + stream_session_t *s; + int rv; + + if (api_session_not_valid (mp->session_index, mp->session_thread_index)) + return; + + s = stream_session_get (mp->session_index, mp->session_thread_index); + rv = mp->retval; + + if (rv) + { + /* Server isn't interested, kill the session */ + stream_session_disconnect (s); + return; + } + + s->session_state = SESSION_STATE_READY; +} + +static void +vl_api_map_another_segment_reply_t_handler (vl_api_map_another_segment_reply_t + * mp) +{ + clib_warning ("not implemented"); +} + +static void +vl_api_bind_sock_t_handler (vl_api_bind_sock_t * mp) +{ + vl_api_bind_sock_reply_t *rmp; + vnet_bind_args_t _a, *a = &_a; + char segment_name[128]; + u32 segment_name_length; + int rv; + + STATIC_ASSERT (sizeof (u64) * SESSION_OPTIONS_N_OPTIONS <= + sizeof (mp->options), + "Out of options, fix api message definition"); + + segment_name_length = ARRAY_LEN (segment_name); + + memset (a, 0, sizeof (*a)); + + clib_memcpy (&a->tep.ip, mp->ip, + (mp->is_ip4 ? sizeof (ip4_address_t) : + sizeof (ip6_address_t))); + a->tep.is_ip4 = mp->is_ip4; + a->tep.port = mp->port; + a->tep.vrf = mp->vrf; + + a->api_client_index = mp->client_index; + a->options = mp->options; + a->segment_name = segment_name; + a->segment_name_length = segment_name_length; + a->session_cb_vft = &session_cb_vft; + + rv = vnet_bind_uri (a); + + /* *INDENT-OFF* */ + REPLY_MACRO2 (VL_API_BIND_SOCK_REPLY, ({ + rmp->retval = rv; + if (!rv) + { + rmp->segment_name_length = 0; + rmp->segment_size = mp->options[SESSION_OPTIONS_SEGMENT_SIZE]; + if (segment_name_length) + { + memcpy(rmp->segment_name, segment_name, segment_name_length); + rmp->segment_name_length = segment_name_length; + } + rmp->server_event_queue_address = a->server_event_queue_address; + } + })); + /* *INDENT-ON* */ +} + +static void +vl_api_unbind_sock_t_handler (vl_api_unbind_sock_t * mp) +{ + vl_api_unbind_sock_reply_t *rmp; + vnet_unbind_args_t _a, *a = &_a; + int rv; + + a->api_client_index = mp->client_index; + a->handle = mp->handle; + + rv = vnet_unbind (a); + + REPLY_MACRO (VL_API_UNBIND_SOCK_REPLY); +} + +static void +vl_api_connect_sock_t_handler (vl_api_connect_sock_t * mp) +{ + vnet_connect_args_t _a, *a = &_a; + + clib_memcpy (&a->tep.ip, mp->ip, + (mp->is_ip4 ? sizeof (ip4_address_t) : + sizeof (ip6_address_t))); + a->tep.is_ip4 = mp->is_ip4; + a->tep.port = mp->port; + a->tep.vrf = mp->vrf; + a->options = mp->options; + a->session_cb_vft = &session_cb_vft; + a->api_context = mp->context; + a->mp = mp; + + vnet_connect (a); +} + +static void +vl_api_disconnect_sock_t_handler (vl_api_disconnect_sock_t * mp) +{ + vnet_disconnect_args_t _a, *a = &_a; + vl_api_disconnect_sock_reply_t *rmp; + int rv; + + a->api_client_index = mp->client_index; + a->handle = mp->handle; + rv = vnet_disconnect (a); + + REPLY_MACRO (VL_API_DISCONNECT_SOCK_REPLY); +} + +static void +vl_api_disconnect_sock_reply_t_handler (vl_api_disconnect_sock_reply_t * mp) +{ + vnet_disconnect_args_t _a, *a = &_a; + + /* Client objected to disconnecting the session, log and continue */ + if (mp->retval) + { + clib_warning ("client retval %d", mp->retval); + return; + } + + a->api_client_index = mp->client_index; + a->handle = mp->handle; + + vnet_disconnect (a); +} + +static void +vl_api_reset_sock_reply_t_handler (vl_api_reset_sock_reply_t * mp) +{ + stream_session_t *s; + u32 session_index, thread_index; + + /* Client objected to resetting the session, log and continue */ + if (mp->retval) + { + clib_warning ("client retval %d", mp->retval); + return; + } + + if (api_parse_session_handle (mp->handle, &session_index, &thread_index)) + { + clib_warning ("Invalid handle"); + return; + } + + s = stream_session_get (session_index, thread_index); + + /* This comes as a response to a reset, transport only waiting for + * confirmation to remove connection state, no need to disconnect */ + stream_session_cleanup (s); +} + +static void +vl_api_accept_sock_reply_t_handler (vl_api_accept_sock_reply_t * mp) +{ + stream_session_t *s; + u32 session_index, thread_index; + + if (api_parse_session_handle (mp->handle, &session_index, &thread_index)) + { + clib_warning ("Invalid handle"); + return; + } + s = stream_session_get (session_index, thread_index); + + if (mp->retval) + { + /* Server isn't interested, kill the session */ + stream_session_disconnect (s); + return; + } + + s->session_state = SESSION_STATE_READY; +} + +#define vl_msg_name_crc_list +#include +#undef vl_msg_name_crc_list + +static void +setup_message_id_table (api_main_t * am) +{ +#define _(id,n,crc) vl_msg_api_add_msg_name_crc (am, #n "_" #crc, id); + foreach_vl_msg_name_crc_session; +#undef _ +} + +/* + * session_api_hookup + * Add uri's API message handlers to the table. + * vlib has alread mapped shared memory and + * added the client registration handlers. + * See .../open-repo/vlib/memclnt_vlib.c:memclnt_process() + */ +static clib_error_t * +session_api_hookup (vlib_main_t * vm) +{ + api_main_t *am = &api_main; + +#define _(N,n) \ + vl_msg_api_set_handlers(VL_API_##N, #n, \ + vl_api_##n##_t_handler, \ + vl_noop_handler, \ + vl_api_##n##_t_endian, \ + vl_api_##n##_t_print, \ + sizeof(vl_api_##n##_t), 1); + foreach_session_api_msg; +#undef _ + + /* + * Messages which bounce off the data-plane to + * an API client. Simply tells the message handling infra not + * to free the message. + * + * Bounced message handlers MUST NOT block the data plane + */ + am->message_bounce[VL_API_CONNECT_URI] = 1; + am->message_bounce[VL_API_CONNECT_SOCK] = 1; + + /* + * Set up the (msg_name, crc, message-id) table + */ + setup_message_id_table (am); + + return 0; +} + +VLIB_API_INIT_FUNCTION (session_api_hookup); +/* + * fd.io coding-style-patch-verification: ON + * + * Local Variables: + * eval: (c-set-style "gnu") + * End: + */ diff --git a/src/vnet/session/session_cli.c b/src/vnet/session/session_cli.c new file mode 100644 index 00000000..b2943a1c --- /dev/null +++ b/src/vnet/session/session_cli.c @@ -0,0 +1,189 @@ +/* + * Copyright (c) 2017 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ +#include +#include + +/** + * Format stream session as per the following format + * + * verbose: + * "Connection", "Rx fifo", "Tx fifo", "Session Index" + * non-verbose: + * "Connection" + */ +u8 * +format_stream_session (u8 * s, va_list * args) +{ + stream_session_t *ss = va_arg (*args, stream_session_t *); + int verbose = va_arg (*args, int); + transport_proto_vft_t *tp_vft; + u8 *str = 0; + + tp_vft = session_get_transport_vft (ss->session_type); + + if (verbose) + str = format (0, "%-20llp%-20llp%-15lld", ss->server_rx_fifo, + ss->server_tx_fifo, stream_session_get_index (ss)); + + if (ss->session_state == SESSION_STATE_READY) + { + s = format (s, "%-40U%v", tp_vft->format_connection, + ss->connection_index, ss->thread_index, str); + } + else if (ss->session_state == SESSION_STATE_LISTENING) + { + s = format (s, "%-40U%v", tp_vft->format_listener, ss->connection_index, + str); + } + else if (ss->session_state == SESSION_STATE_READY) + { + s = + format (s, "%-40U%v", tp_vft->format_half_open, ss->connection_index, + str); + } + else if (ss->session_state == SESSION_STATE_CLOSED) + { + s = format (s, "[CL] %-40U%v", tp_vft->format_connection, + ss->connection_index, ss->thread_index, str); + } + else + { + clib_warning ("Session in unknown state!"); + } + + vec_free (str); + + return s; +} + +static clib_error_t * +show_session_command_fn (vlib_main_t * vm, unformat_input_t * input, + vlib_cli_command_t * cmd) +{ + session_manager_main_t *smm = &session_manager_main; + int verbose = 0, i; + stream_session_t *pool; + stream_session_t *s; + u8 *str = 0; + + while (unformat_check_input (input) != UNFORMAT_END_OF_INPUT) + { + if (unformat (input, "verbose")) + verbose = 1; + else + break; + } + + for (i = 0; i < vec_len (smm->sessions); i++) + { + u32 once_per_pool; + pool = smm->sessions[i]; + + once_per_pool = 1; + + if (pool_elts (pool)) + { + + vlib_cli_output (vm, "Thread %d: %d active sessions", + i, pool_elts (pool)); + if (verbose) + { + if (once_per_pool) + { + str = format (str, "%-40s%-20s%-20s%-15s", + "Connection", "Rx fifo", "Tx fifo", + "Session Index"); + vlib_cli_output (vm, "%v", str); + vec_reset_length (str); + once_per_pool = 0; + } + + /* *INDENT-OFF* */ + pool_foreach (s, pool, + ({ + vlib_cli_output (vm, "%U", format_stream_session, s, verbose); + })); + /* *INDENT-ON* */ + } + } + else + vlib_cli_output (vm, "Thread %d: no active sessions", i); + } + vec_free (str); + + return 0; +} + +VLIB_CLI_COMMAND (show_uri_command, static) = +{ +.path = "show session",.short_help = "show session [verbose]",.function = + show_session_command_fn,}; + + +static clib_error_t * +clear_session_command_fn (vlib_main_t * vm, unformat_input_t * input, + vlib_cli_command_t * cmd) +{ + session_manager_main_t *smm = &session_manager_main; + u32 thread_index = 0; + u32 session_index = ~0; + stream_session_t *pool, *session; + application_t *server; + + while (unformat_check_input (input) != UNFORMAT_END_OF_INPUT) + { + if (unformat (input, "thread %d", &thread_index)) + ; + else if (unformat (input, "session %d", &session_index)) + ; + else + return clib_error_return (0, "unknown input `%U'", + format_unformat_error, input); + } + + if (session_index == ~0) + return clib_error_return (0, "session required, but not set."); + + if (thread_index > vec_len (smm->sessions)) + return clib_error_return (0, "thread %d out of range [0-%d]", + thread_index, vec_len (smm->sessions)); + + pool = smm->sessions[thread_index]; + + if (pool_is_free_index (pool, session_index)) + return clib_error_return (0, "session %d not active", session_index); + + session = pool_elt_at_index (pool, session_index); + server = application_get (session->app_index); + + /* Disconnect both app and transport */ + server->cb_fns.session_disconnect_callback (session); + + return 0; +} + +VLIB_CLI_COMMAND (clear_uri_session_command, static) = +{ +.path = "clear session",.short_help = + "clear session thread session ",.function = + clear_session_command_fn,}; + +/* + * fd.io coding-style-patch-verification: ON + * + * Local Variables: + * eval: (c-set-style "gnu") + * End: + */ diff --git a/src/vnet/session/transport.c b/src/vnet/session/transport.c new file mode 100644 index 00000000..abd94ba4 --- /dev/null +++ b/src/vnet/session/transport.c @@ -0,0 +1,64 @@ +/* + * Copyright (c) 2017 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +#include + +u32 +transport_endpoint_lookup (transport_endpoint_table_t *ht, ip46_address_t *ip, + u16 port) +{ + clib_bihash_kv_24_8_t kv; + int rv; + + kv.key[0] = ip->as_u64[0]; + kv.key[1] = ip->as_u64[1]; + kv.key[2] = port; + + rv = clib_bihash_search_inline_24_8 (ht, &kv); + if (rv == 0) + return kv.value; + + return TRANSPORT_ENDPOINT_INVALID_INDEX; +} + +void +transport_endpoint_table_add (transport_endpoint_table_t *ht, + transport_endpoint_t *te, u32 value) +{ + clib_bihash_kv_24_8_t kv; + + kv.key[0] = te->ip.as_u64[0]; + kv.key[1] = te->ip.as_u64[1]; + kv.key[2] = te->port; + kv.value = value; + + clib_bihash_add_del_24_8 (ht, &kv, 1); +} + +void +transport_endpoint_table_del (transport_endpoint_table_t *ht, + transport_endpoint_t *te) +{ + clib_bihash_kv_24_8_t kv; + + kv.key[0] = te->ip.as_u64[0]; + kv.key[1] = te->ip.as_u64[1]; + kv.key[2] = te->port; + + clib_bihash_add_del_24_8 (ht, &kv, 0); +} + + + diff --git a/src/vnet/session/transport.h b/src/vnet/session/transport.h new file mode 100644 index 00000000..2d4415ba --- /dev/null +++ b/src/vnet/session/transport.h @@ -0,0 +1,250 @@ +/* + * Copyright (c) 2016 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +#ifndef VNET_VNET_URI_TRANSPORT_H_ +#define VNET_VNET_URI_TRANSPORT_H_ + +#include +#include +#include +#include + +/* + * Protocol independent transport properties associated to a session + */ +typedef struct _transport_connection +{ + ip46_address_t rmt_ip; /**< Remote IP */ + ip46_address_t lcl_ip; /**< Local IP */ + u16 lcl_port; /**< Local port */ + u16 rmt_port; /**< Remote port */ + u8 proto; /**< Transport protocol id */ + + u32 s_index; /**< Parent session index */ + u32 c_index; /**< Connection index in transport pool */ + u8 is_ip4; /**< Flag if IP4 connection */ + u32 thread_index; /**< Worker-thread index */ + + /** Macros for 'derived classes' where base is named "connection" */ +#define c_lcl_ip connection.lcl_ip +#define c_rmt_ip connection.rmt_ip +#define c_lcl_ip4 connection.lcl_ip.ip4 +#define c_rmt_ip4 connection.rmt_ip.ip4 +#define c_lcl_ip6 connection.lcl_ip.ip6 +#define c_rmt_ip6 connection.rmt_ip.ip6 +#define c_lcl_port connection.lcl_port +#define c_rmt_port connection.rmt_port +#define c_proto connection.proto +#define c_state connection.state +#define c_s_index connection.s_index +#define c_c_index connection.c_index +#define c_is_ip4 connection.is_ip4 +#define c_thread_index connection.thread_index +} transport_connection_t; + +/* + * Transport protocol virtual function table + */ +typedef struct _transport_proto_vft +{ + /* + * Setup + */ + u32 (*bind) (vlib_main_t *, u32, ip46_address_t *, u16); + u32 (*unbind) (vlib_main_t *, u32); + int (*open) (ip46_address_t * addr, u16 port_host_byte_order); + void (*close) (u32 conn_index, u32 thread_index); + void (*cleanup) (u32 conn_index, u32 thread_index); + + /* + * Transmission + */ + u32 (*push_header) (transport_connection_t * tconn, vlib_buffer_t * b); + u16 (*send_mss) (transport_connection_t * tc); + u32 (*send_space) (transport_connection_t * tc); + u32 (*rx_fifo_offset) (transport_connection_t * tc); + + /* + * Connection retrieval + */ + transport_connection_t *(*get_connection) (u32 conn_idx, u32 thread_idx); + transport_connection_t *(*get_listener) (u32 conn_index); + transport_connection_t *(*get_half_open) (u32 conn_index); + + /* + * Format + */ + u8 *(*format_connection) (u8 * s, va_list * args); + u8 *(*format_listener) (u8 * s, va_list * args); + u8 *(*format_half_open) (u8 * s, va_list * args); + +} transport_proto_vft_t; + +/* 16 octets */ +typedef CLIB_PACKED (struct + { + union + { + struct + { + ip4_address_t src; ip4_address_t dst; + u16 src_port; + u16 dst_port; + /* align by making this 4 octets even though its a 1-bit field + * NOTE: avoid key overlap with other transports that use 5 tuples for + * session identification. + */ + u32 proto; + }; + u64 as_u64[2]; + }; + }) v4_connection_key_t; + +typedef CLIB_PACKED (struct + { + union + { + struct + { + /* 48 octets */ + ip6_address_t src; ip6_address_t dst; + u16 src_port; + u16 dst_port; u32 proto; u8 unused_for_now[8]; + }; u64 as_u64[6]; + }; + }) v6_connection_key_t; + +typedef clib_bihash_kv_16_8_t session_kv4_t; +typedef clib_bihash_kv_48_8_t session_kv6_t; + +always_inline void +make_v4_ss_kv (session_kv4_t * kv, ip4_address_t * lcl, ip4_address_t * rmt, + u16 lcl_port, u16 rmt_port, u8 proto) +{ + v4_connection_key_t key; + memset (&key, 0, sizeof (v4_connection_key_t)); + + key.src.as_u32 = lcl->as_u32; + key.dst.as_u32 = rmt->as_u32; + key.src_port = lcl_port; + key.dst_port = rmt_port; + key.proto = proto; + + kv->key[0] = key.as_u64[0]; + kv->key[1] = key.as_u64[1]; + kv->value = ~0ULL; +} + +always_inline void +make_v4_listener_kv (session_kv4_t * kv, ip4_address_t * lcl, u16 lcl_port, + u8 proto) +{ + v4_connection_key_t key; + memset (&key, 0, sizeof (v4_connection_key_t)); + + key.src.as_u32 = lcl->as_u32; + key.dst.as_u32 = 0; + key.src_port = lcl_port; + key.dst_port = 0; + key.proto = proto; + + kv->key[0] = key.as_u64[0]; + kv->key[1] = key.as_u64[1]; + kv->value = ~0ULL; +} + +always_inline void +make_v4_ss_kv_from_tc (session_kv4_t * kv, transport_connection_t * t) +{ + return make_v4_ss_kv (kv, &t->lcl_ip.ip4, &t->rmt_ip.ip4, t->lcl_port, + t->rmt_port, t->proto); +} + +always_inline void +make_v6_ss_kv (session_kv6_t * kv, ip6_address_t * lcl, ip6_address_t * rmt, + u16 lcl_port, u16 rmt_port, u8 proto) +{ + v6_connection_key_t key; + memset (&key, 0, sizeof (v6_connection_key_t)); + + key.src.as_u64[0] = lcl->as_u64[0]; + key.src.as_u64[1] = lcl->as_u64[1]; + key.dst.as_u64[0] = rmt->as_u64[0]; + key.dst.as_u64[1] = rmt->as_u64[1]; + key.src_port = lcl_port; + key.dst_port = rmt_port; + key.proto = proto; + + kv->key[0] = key.as_u64[0]; + kv->key[1] = key.as_u64[1]; + kv->value = ~0ULL; +} + +always_inline void +make_v6_listener_kv (session_kv6_t * kv, ip6_address_t * lcl, u16 lcl_port, + u8 proto) +{ + v6_connection_key_t key; + memset (&key, 0, sizeof (v6_connection_key_t)); + + key.src.as_u64[0] = lcl->as_u64[0]; + key.src.as_u64[1] = lcl->as_u64[1]; + key.dst.as_u64[0] = 0; + key.dst.as_u64[1] = 0; + key.src_port = lcl_port; + key.dst_port = 0; + key.proto = proto; + + kv->key[0] = key.as_u64[0]; + kv->key[1] = key.as_u64[1]; + kv->value = ~0ULL; +} + +always_inline void +make_v6_ss_kv_from_tc (session_kv6_t * kv, transport_connection_t * t) +{ + make_v6_ss_kv (kv, &t->lcl_ip.ip6, &t->rmt_ip.ip6, t->lcl_port, + t->rmt_port, t->proto); +} + +typedef struct _transport_endpoint +{ + ip46_address_t ip; + u16 port; + u8 is_ip4; + u32 vrf; +} transport_endpoint_t; + +typedef clib_bihash_24_8_t transport_endpoint_table_t; + +#define TRANSPORT_ENDPOINT_INVALID_INDEX ((u32)~0) + +u32 +transport_endpoint_lookup (transport_endpoint_table_t * ht, + ip46_address_t * ip, u16 port); +void transport_endpoint_table_add (transport_endpoint_table_t * ht, + transport_endpoint_t * te, u32 value); +void transport_endpoint_table_del (transport_endpoint_table_t * ht, + transport_endpoint_t * te); + +#endif /* VNET_VNET_URI_TRANSPORT_H_ */ + +/* + * fd.io coding-style-patch-verification: ON + * + * Local Variables: + * eval: (c-set-style "gnu") + * End: + */ diff --git a/src/vnet/tcp/tcp.c b/src/vnet/tcp/tcp.c new file mode 100644 index 00000000..0f9b7097 --- /dev/null +++ b/src/vnet/tcp/tcp.c @@ -0,0 +1,708 @@ +/* + * Copyright (c) 2016 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +#include +#include +#include +#include + +tcp_main_t tcp_main; + +static u32 +tcp_connection_bind (vlib_main_t * vm, u32 session_index, ip46_address_t * ip, + u16 port_host_byte_order, u8 is_ip4) +{ + tcp_main_t *tm = &tcp_main; + tcp_connection_t *listener; + + pool_get (tm->listener_pool, listener); + memset (listener, 0, sizeof (*listener)); + + listener->c_c_index = listener - tm->listener_pool; + listener->c_lcl_port = clib_host_to_net_u16 (port_host_byte_order); + + if (is_ip4) + listener->c_lcl_ip4.as_u32 = ip->ip4.as_u32; + else + clib_memcpy (&listener->c_lcl_ip6, &ip->ip6, sizeof (ip6_address_t)); + + listener->c_s_index = session_index; + listener->c_proto = SESSION_TYPE_IP4_TCP; + listener->state = TCP_STATE_LISTEN; + listener->c_is_ip4 = 1; + + return listener->c_c_index; +} + +u32 +tcp_session_bind_ip4 (vlib_main_t * vm, u32 session_index, + ip46_address_t * ip, u16 port_host_byte_order) +{ + return tcp_connection_bind (vm, session_index, ip, port_host_byte_order, 1); +} + +u32 +tcp_session_bind_ip6 (vlib_main_t * vm, u32 session_index, + ip46_address_t * ip, u16 port_host_byte_order) +{ + return tcp_connection_bind (vm, session_index, ip, port_host_byte_order, 0); + +} + +static void +tcp_session_unbind (u32 listener_index) +{ + tcp_main_t *tm = vnet_get_tcp_main (); + pool_put_index (tm->listener_pool, listener_index); +} + +u32 +tcp_session_unbind_ip4 (vlib_main_t * vm, u32 listener_index) +{ + tcp_session_unbind (listener_index); + return 0; +} + +u32 +tcp_session_unbind_ip6 (vlib_main_t * vm, u32 listener_index) +{ + tcp_session_unbind (listener_index); + return 0; +} + +transport_connection_t * +tcp_session_get_listener (u32 listener_index) +{ + tcp_main_t *tm = vnet_get_tcp_main (); + tcp_connection_t *tc; + tc = pool_elt_at_index (tm->listener_pool, listener_index); + return &tc->connection; +} + +/** + * Cleans up connection state. + * + * No notifications. + */ +void +tcp_connection_cleanup (tcp_connection_t * tc) +{ + tcp_main_t *tm = &tcp_main; + u32 tepi; + transport_endpoint_t *tep; + + /* Cleanup local endpoint if this was an active connect */ + tepi = transport_endpoint_lookup (&tm->local_endpoints_table, &tc->c_lcl_ip, + tc->c_lcl_port); + + /*XXX lock */ + if (tepi != TRANSPORT_ENDPOINT_INVALID_INDEX) + { + tep = pool_elt_at_index (tm->local_endpoints, tepi); + transport_endpoint_table_del (&tm->local_endpoints_table, tep); + pool_put (tm->local_endpoints, tep); + } + + /* Make sure all timers are cleared */ + tcp_connection_timers_reset (tc); + + /* Check if half-open */ + if (tc->state == TCP_STATE_SYN_SENT) + pool_put (tm->half_open_connections, tc); + else + pool_put (tm->connections[tc->c_thread_index], tc); +} + +/** + * Connection removal. + * + * This should be called only once connection enters CLOSED state. Note + * that it notifies the session of the removal event, so if the goal is to + * just remove the connection, call tcp_connection_cleanup instead. + */ +void +tcp_connection_del (tcp_connection_t * tc) +{ + stream_session_delete_notify (&tc->connection); + tcp_connection_cleanup (tc); +} + +/** + * Begin connection closing procedure. + * + * If at the end the connection is not in CLOSED state, it is not removed. + * Instead, we rely on on TCP to advance through state machine to either + * 1) LAST_ACK (passive close) whereby when the last ACK is received + * tcp_connection_del is called. This notifies session of the delete and + * calls cleanup. + * 2) TIME_WAIT (active close) whereby after 2MSL the 2MSL timer triggers + * and cleanup is called. + */ +void +tcp_connection_close (tcp_connection_t * tc) +{ + /* Send FIN if needed */ + if (tc->state == TCP_STATE_ESTABLISHED || tc->state == TCP_STATE_SYN_RCVD + || tc->state == TCP_STATE_CLOSE_WAIT) + tcp_send_fin (tc); + + /* Switch state */ + if (tc->state == TCP_STATE_ESTABLISHED || tc->state == TCP_STATE_SYN_RCVD) + tc->state = TCP_STATE_FIN_WAIT_1; + else if (tc->state == TCP_STATE_SYN_SENT) + tc->state = TCP_STATE_CLOSED; + else if (tc->state == TCP_STATE_CLOSE_WAIT) + tc->state = TCP_STATE_LAST_ACK; + + /* Half-close connections are not supported XXX */ + + if (tc->state == TCP_STATE_CLOSED) + tcp_connection_del (tc); +} + +void +tcp_session_close (u32 conn_index, u32 thread_index) +{ + tcp_connection_t *tc; + tc = tcp_connection_get (conn_index, thread_index); + tcp_connection_close (tc); +} + +void +tcp_session_cleanup (u32 conn_index, u32 thread_index) +{ + tcp_connection_t *tc; + tc = tcp_connection_get (conn_index, thread_index); + tcp_connection_cleanup (tc); +} + +void * +ip_interface_get_first_ip (u32 sw_if_index, u8 is_ip4) +{ + ip_lookup_main_t *lm4 = &ip4_main.lookup_main; + ip_lookup_main_t *lm6 = &ip6_main.lookup_main; + ip_interface_address_t *ia = 0; + + if (is_ip4) + { + /* *INDENT-OFF* */ + foreach_ip_interface_address (lm4, ia, sw_if_index, 1 /* unnumbered */ , + ({ + return ip_interface_address_get_address (lm4, ia); + })); + /* *INDENT-ON* */ + } + else + { + /* *INDENT-OFF* */ + foreach_ip_interface_address (lm6, ia, sw_if_index, 1 /* unnumbered */ , + ({ + return ip_interface_address_get_address (lm6, ia); + })); + /* *INDENT-ON* */ + } + + return 0; +} + +/** + * Allocate local port and add if successful add entry to local endpoint + * table to mark the pair as used. + */ +u16 +tcp_allocate_local_port (tcp_main_t * tm, ip46_address_t * ip) +{ + u8 unique = 0; + transport_endpoint_t *tep; + u32 time_now, tei; + u16 min = 1024, max = 65535, tries; /* XXX configurable ? */ + + tries = max - min; + time_now = tcp_time_now (); + + /* Start at random point or max */ + pool_get (tm->local_endpoints, tep); + clib_memcpy (&tep->ip, ip, sizeof (*ip)); + tep->port = random_u32 (&time_now) << 16; + tep->port = tep->port < min ? max : tep->port; + + /* Search for first free slot */ + while (tries) + { + tei = transport_endpoint_lookup (&tm->local_endpoints_table, &tep->ip, + tep->port); + if (tei == TRANSPORT_ENDPOINT_INVALID_INDEX) + { + unique = 1; + break; + } + + tep->port--; + + if (tep->port < min) + tep->port = max; + + tries--; + } + + if (unique) + { + transport_endpoint_table_add (&tm->local_endpoints_table, tep, + tep - tm->local_endpoints); + + return tep->port; + } + + /* Failed */ + pool_put (tm->local_endpoints, tep); + return -1; +} + +/** + * Initialize all connection timers as invalid + */ +void +tcp_connection_timers_init (tcp_connection_t * tc) +{ + int i; + + /* Set all to invalid */ + for (i = 0; i < TCP_N_TIMERS; i++) + { + tc->timers[i] = TCP_TIMER_HANDLE_INVALID; + } + + tc->rto = TCP_RTO_INIT; +} + +/** + * Stop all connection timers + */ +void +tcp_connection_timers_reset (tcp_connection_t * tc) +{ + int i; + for (i = 0; i < TCP_N_TIMERS; i++) + { + tcp_timer_reset (tc, i); + } +} + +/** Initialize tcp connection variables + * + * Should be called after having received a msg from the peer, i.e., a SYN or + * a SYNACK, such that connection options have already been exchanged. */ +void +tcp_connection_init_vars (tcp_connection_t * tc) +{ + tcp_connection_timers_init (tc); + tcp_set_snd_mss (tc); + tc->sack_sb.head = TCP_INVALID_SACK_HOLE_INDEX; + tcp_cc_init (tc); +} + +int +tcp_connection_open (ip46_address_t * rmt_addr, u16 rmt_port, u8 is_ip4) +{ + tcp_main_t *tm = vnet_get_tcp_main (); + tcp_connection_t *tc; + fib_prefix_t prefix; + u32 fei, sw_if_index; + ip46_address_t lcl_addr; + u16 lcl_port; + + /* + * Find the local address and allocate port + */ + memset (&lcl_addr, 0, sizeof (lcl_addr)); + + /* Find a FIB path to the destination */ + clib_memcpy (&prefix.fp_addr, rmt_addr, sizeof (*rmt_addr)); + prefix.fp_proto = is_ip4 ? FIB_PROTOCOL_IP4 : FIB_PROTOCOL_IP6; + prefix.fp_len = is_ip4 ? 32 : 128; + + fei = fib_table_lookup (0, &prefix); + + /* Couldn't find route to destination. Bail out. */ + if (fei == FIB_NODE_INDEX_INVALID) + return -1; + + sw_if_index = fib_entry_get_resolving_interface (fei); + + if (sw_if_index == (u32) ~ 0) + return -1; + + if (is_ip4) + { + ip4_address_t *ip4; + ip4 = ip_interface_get_first_ip (sw_if_index, 1); + lcl_addr.ip4.as_u32 = ip4->as_u32; + } + else + { + ip6_address_t *ip6; + ip6 = ip_interface_get_first_ip (sw_if_index, 0); + clib_memcpy (&lcl_addr.ip6, ip6, sizeof (*ip6)); + } + + /* Allocate source port */ + lcl_port = tcp_allocate_local_port (tm, &lcl_addr); + if (lcl_port < 1) + return -1; + + /* + * Create connection and send SYN + */ + + pool_get (tm->half_open_connections, tc); + memset (tc, 0, sizeof (*tc)); + + clib_memcpy (&tc->c_rmt_ip, rmt_addr, sizeof (ip46_address_t)); + clib_memcpy (&tc->c_lcl_ip, &lcl_addr, sizeof (ip46_address_t)); + tc->c_rmt_port = clib_host_to_net_u16 (rmt_port); + tc->c_lcl_port = clib_host_to_net_u16 (lcl_port); + tc->c_c_index = tc - tm->half_open_connections; + tc->c_is_ip4 = is_ip4; + + /* The other connection vars will be initialized after SYN ACK */ + tcp_connection_timers_init (tc); + + tcp_send_syn (tc); + + tc->state = TCP_STATE_SYN_SENT; + + return tc->c_c_index; +} + +int +tcp_session_open_ip4 (ip46_address_t * addr, u16 port) +{ + return tcp_connection_open (addr, port, 1); +} + +int +tcp_session_open_ip6 (ip46_address_t * addr, u16 port) +{ + return tcp_connection_open (addr, port, 0); +} + +u8 * +format_tcp_session_ip4 (u8 * s, va_list * args) +{ + u32 tci = va_arg (*args, u32); + u32 thread_index = va_arg (*args, u32); + tcp_connection_t *tc; + + tc = tcp_connection_get (tci, thread_index); + + s = format (s, "[%s] %U:%d->%U:%d", "tcp", format_ip4_address, + &tc->c_lcl_ip4, clib_net_to_host_u16 (tc->c_lcl_port), + format_ip4_address, &tc->c_rmt_ip4, + clib_net_to_host_u16 (tc->c_rmt_port)); + + return s; +} + +u8 * +format_tcp_session_ip6 (u8 * s, va_list * args) +{ + u32 tci = va_arg (*args, u32); + u32 thread_index = va_arg (*args, u32); + tcp_connection_t *tc = tcp_connection_get (tci, thread_index); + s = format (s, "[%s] %U:%d->%U:%d", "tcp", format_ip6_address, + &tc->c_lcl_ip6, clib_net_to_host_u16 (tc->c_lcl_port), + format_ip6_address, &tc->c_rmt_ip6, + clib_net_to_host_u16 (tc->c_rmt_port)); + return s; +} + +u8 * +format_tcp_listener_session_ip4 (u8 * s, va_list * args) +{ + u32 tci = va_arg (*args, u32); + tcp_connection_t *tc = tcp_listener_get (tci); + s = format (s, "[%s] %U:%d->%U:%d", "tcp", format_ip4_address, + &tc->c_lcl_ip4, clib_net_to_host_u16 (tc->c_lcl_port), + format_ip4_address, &tc->c_rmt_ip4, + clib_net_to_host_u16 (tc->c_rmt_port)); + return s; +} + +u8 * +format_tcp_listener_session_ip6 (u8 * s, va_list * args) +{ + u32 tci = va_arg (*args, u32); + tcp_connection_t *tc = tcp_listener_get (tci); + s = format (s, "[%s] %U:%d->%U:%d", "tcp", format_ip6_address, + &tc->c_lcl_ip6, clib_net_to_host_u16 (tc->c_lcl_port), + format_ip6_address, &tc->c_rmt_ip6, + clib_net_to_host_u16 (tc->c_rmt_port)); + return s; +} + +u8 * +format_tcp_half_open_session_ip4 (u8 * s, va_list * args) +{ + u32 tci = va_arg (*args, u32); + tcp_connection_t *tc = tcp_half_open_connection_get (tci); + s = format (s, "[%s] %U:%d->%U:%d", "tcp", format_ip4_address, + &tc->c_lcl_ip4, clib_net_to_host_u16 (tc->c_lcl_port), + format_ip4_address, &tc->c_rmt_ip4, + clib_net_to_host_u16 (tc->c_rmt_port)); + return s; +} + +u8 * +format_tcp_half_open_session_ip6 (u8 * s, va_list * args) +{ + u32 tci = va_arg (*args, u32); + tcp_connection_t *tc = tcp_half_open_connection_get (tci); + s = format (s, "[%s] %U:%d->%U:%d", "tcp", format_ip6_address, + &tc->c_lcl_ip6, clib_net_to_host_u16 (tc->c_lcl_port), + format_ip6_address, &tc->c_rmt_ip6, + clib_net_to_host_u16 (tc->c_rmt_port)); + return s; +} + +transport_connection_t * +tcp_session_get_transport (u32 conn_index, u32 thread_index) +{ + tcp_connection_t *tc = tcp_connection_get (conn_index, thread_index); + return &tc->connection; +} + +transport_connection_t * +tcp_half_open_session_get_transport (u32 conn_index) +{ + tcp_connection_t *tc = tcp_half_open_connection_get (conn_index); + return &tc->connection; +} + +u16 +tcp_session_send_mss (transport_connection_t * trans_conn) +{ + tcp_connection_t *tc = (tcp_connection_t *) trans_conn; + return tc->snd_mss; +} + +u32 +tcp_session_send_space (transport_connection_t * trans_conn) +{ + tcp_connection_t *tc = (tcp_connection_t *) trans_conn; + return tcp_available_snd_space (tc); +} + +u32 +tcp_session_rx_fifo_offset (transport_connection_t * trans_conn) +{ + tcp_connection_t *tc = (tcp_connection_t *) trans_conn; + return (tc->snd_una_max - tc->snd_una); +} + +/* *INDENT-OFF* */ +const static transport_proto_vft_t tcp4_proto = { + .bind = tcp_session_bind_ip4, + .unbind = tcp_session_unbind_ip4, + .push_header = tcp_push_header, + .get_connection = tcp_session_get_transport, + .get_listener = tcp_session_get_listener, + .get_half_open = tcp_half_open_session_get_transport, + .open = tcp_session_open_ip4, + .close = tcp_session_close, + .cleanup = tcp_session_cleanup, + .send_mss = tcp_session_send_mss, + .send_space = tcp_session_send_space, + .rx_fifo_offset = tcp_session_rx_fifo_offset, + .format_connection = format_tcp_session_ip4, + .format_listener = format_tcp_listener_session_ip4, + .format_half_open = format_tcp_half_open_session_ip4 +}; + +const static transport_proto_vft_t tcp6_proto = { + .bind = tcp_session_bind_ip6, + .unbind = tcp_session_unbind_ip6, + .push_header = tcp_push_header, + .get_connection = tcp_session_get_transport, + .get_listener = tcp_session_get_listener, + .get_half_open = tcp_half_open_session_get_transport, + .open = tcp_session_open_ip6, + .close = tcp_session_close, + .cleanup = tcp_session_cleanup, + .send_mss = tcp_session_send_mss, + .send_space = tcp_session_send_space, + .rx_fifo_offset = tcp_session_rx_fifo_offset, + .format_connection = format_tcp_session_ip6, + .format_listener = format_tcp_listener_session_ip6, + .format_half_open = format_tcp_half_open_session_ip6 +}; +/* *INDENT-ON* */ + +void +tcp_timer_keep_handler (u32 conn_index) +{ + u32 cpu_index = os_get_cpu_number (); + tcp_connection_t *tc; + + tc = tcp_connection_get (conn_index, cpu_index); + tc->timers[TCP_TIMER_KEEP] = TCP_TIMER_HANDLE_INVALID; + + tcp_connection_close (tc); +} + +void +tcp_timer_establish_handler (u32 conn_index) +{ + tcp_connection_t *tc; + u8 sst; + + tc = tcp_half_open_connection_get (conn_index); + tc->timers[TCP_TIMER_ESTABLISH] = TCP_TIMER_HANDLE_INVALID; + + ASSERT (tc->state == TCP_STATE_SYN_SENT); + + sst = tc->c_is_ip4 ? SESSION_TYPE_IP4_TCP : SESSION_TYPE_IP6_TCP; + stream_session_connect_notify (&tc->connection, sst, 1 /* fail */ ); + + tcp_connection_cleanup (tc); +} + +void +tcp_timer_2msl_handler (u32 conn_index) +{ + u32 cpu_index = os_get_cpu_number (); + tcp_connection_t *tc; + + tc = tcp_connection_get (conn_index, cpu_index); + tc->timers[TCP_TIMER_2MSL] = TCP_TIMER_HANDLE_INVALID; + + tcp_connection_del (tc); +} + +/* *INDENT-OFF* */ +static timer_expiration_handler *timer_expiration_handlers[TCP_N_TIMERS] = +{ + tcp_timer_retransmit_handler, + tcp_timer_delack_handler, + 0, + tcp_timer_keep_handler, + tcp_timer_2msl_handler, + tcp_timer_retransmit_syn_handler, + tcp_timer_establish_handler +}; +/* *INDENT-ON* */ + +static void +tcp_expired_timers_dispatch (u32 * expired_timers) +{ + int i; + u32 connection_index, timer_id; + + for (i = 0; i < vec_len (expired_timers); i++) + { + /* Get session index and timer id */ + connection_index = expired_timers[i] & 0x0FFFFFFF; + timer_id = expired_timers[i] >> 28; + + /* Handle expiration */ + (*timer_expiration_handlers[timer_id]) (connection_index); + } +} + +void +tcp_initialize_timer_wheels (tcp_main_t * tm) +{ + tw_timer_wheel_16t_2w_512sl_t *tw; + vec_foreach (tw, tm->timer_wheels) + { + tw_timer_wheel_init_16t_2w_512sl (tw, tcp_expired_timers_dispatch, + 100e-3 /* timer period 100ms */ , ~0); + tw->last_run_time = vlib_time_now (tm->vlib_main); + } +} + +clib_error_t * +tcp_init (vlib_main_t * vm) +{ + ip_main_t *im = &ip_main; + ip_protocol_info_t *pi; + tcp_main_t *tm = vnet_get_tcp_main (); + vlib_thread_main_t *vtm = vlib_get_thread_main (); + clib_error_t *error = 0; + u32 num_threads; + + tm->vlib_main = vm; + tm->vnet_main = vnet_get_main (); + + if ((error = vlib_call_init_function (vm, ip_main_init))) + return error; + if ((error = vlib_call_init_function (vm, ip4_lookup_init))) + return error; + if ((error = vlib_call_init_function (vm, ip6_lookup_init))) + return error; + + /* + * Registrations + */ + + /* Register with IP */ + pi = ip_get_protocol_info (im, IP_PROTOCOL_TCP); + if (pi == 0) + return clib_error_return (0, "TCP protocol info AWOL"); + pi->format_header = format_tcp_header; + pi->unformat_pg_edit = unformat_pg_tcp_header; + + ip4_register_protocol (IP_PROTOCOL_TCP, tcp4_input_node.index); + + /* Register as transport with URI */ + session_register_transport (SESSION_TYPE_IP4_TCP, &tcp4_proto); + session_register_transport (SESSION_TYPE_IP6_TCP, &tcp6_proto); + + /* + * Initialize data structures + */ + + num_threads = 1 /* main thread */ + vtm->n_threads; + vec_validate (tm->connections, num_threads - 1); + + /* Initialize per worker thread tx buffers (used for control messages) */ + vec_validate (tm->tx_buffers, num_threads - 1); + + /* Initialize timer wheels */ + vec_validate (tm->timer_wheels, num_threads - 1); + tcp_initialize_timer_wheels (tm); + + vec_validate (tm->delack_connections, num_threads - 1); + + /* Initialize clocks per tick for TCP timestamp. Used to compute + * monotonically increasing timestamps. */ + tm->tstamp_ticks_per_clock = vm->clib_time.seconds_per_clock + / TCP_TSTAMP_RESOLUTION; + + clib_bihash_init_24_8 (&tm->local_endpoints_table, "local endpoint table", + 200000 /* $$$$ config parameter nbuckets */ , + (64 << 20) /*$$$ config parameter table size */ ); + + return error; +} + +VLIB_INIT_FUNCTION (tcp_init); + +/* + * fd.io coding-style-patch-verification: ON + * + * Local Variables: + * eval: (c-set-style "gnu") + * End: + */ diff --git a/src/vnet/tcp/tcp.h b/src/vnet/tcp/tcp.h new file mode 100644 index 00000000..22f00a63 --- /dev/null +++ b/src/vnet/tcp/tcp.h @@ -0,0 +1,624 @@ +/* + * Copyright (c) 2016 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +#ifndef _vnet_tcp_h_ +#define _vnet_tcp_h_ + +#include +#include +#include +#include +#include +#include + +#define TCP_TICK 10e-3 /**< TCP tick period (s) */ +#define THZ 1/TCP_TICK /**< TCP tick frequency */ +#define TCP_TSTAMP_RESOLUTION TCP_TICK /**< Time stamp resolution */ +#define TCP_PAWS_IDLE 24 * 24 * 60 * 60 * THZ /**< 24 days */ +#define TCP_MAX_OPTION_SPACE 40 + +#define TCP_DUPACK_THRESHOLD 3 +#define TCP_DEFAULT_RX_FIFO_SIZE 64 << 10 + +/** TCP FSM state definitions as per RFC793. */ +#define foreach_tcp_fsm_state \ + _(CLOSED, "CLOSED") \ + _(LISTEN, "LISTEN") \ + _(SYN_SENT, "SYN_SENT") \ + _(SYN_RCVD, "SYN_RCVD") \ + _(ESTABLISHED, "ESTABLISHED") \ + _(CLOSE_WAIT, "CLOSE_WAIT") \ + _(FIN_WAIT_1, "FIN_WAIT_1") \ + _(LAST_ACK, "LAST_ACK") \ + _(CLOSING, "CLOSING") \ + _(FIN_WAIT_2, "FIN_WAIT_2") \ + _(TIME_WAIT, "TIME_WAIT") + +typedef enum _tcp_state +{ +#define _(sym, str) TCP_STATE_##sym, + foreach_tcp_fsm_state +#undef _ + TCP_N_STATES +} tcp_state_t; + +format_function_t format_tcp_state; + +/** TCP timers */ +#define foreach_tcp_timer \ + _(RETRANSMIT, "RETRANSMIT") \ + _(DELACK, "DELAYED ACK") \ + _(PERSIST, "PERSIST") \ + _(KEEP, "KEEP") \ + _(2MSL, "2MSL") \ + _(RETRANSMIT_SYN, "RETRANSMIT_SYN") \ + _(ESTABLISH, "ESTABLISH") + +typedef enum _tcp_timers +{ +#define _(sym, str) TCP_TIMER_##sym, + foreach_tcp_timer +#undef _ + TCP_N_TIMERS +} tcp_timers_e; + +typedef void (timer_expiration_handler) (u32 index); + +extern timer_expiration_handler tcp_timer_delack_handler; +extern timer_expiration_handler tcp_timer_retransmit_handler; +extern timer_expiration_handler tcp_timer_retransmit_syn_handler; + +#define TCP_TIMER_HANDLE_INVALID ((u32) ~0) + +/* Timer delays as multiples of 100ms */ +#define TCP_TO_TIMER_TICK TCP_TICK*10 /* Period for converting from TCP + * ticks to timer units */ +#define TCP_DELACK_TIME 1 /* 0.1s */ +#define TCP_ESTABLISH_TIME 750 /* 75s */ +#define TCP_2MSL_TIME 300 /* 30s */ + +#define TCP_RTO_MAX 60 * THZ /* Min max RTO (60s) as per RFC6298 */ +#define TCP_RTT_MAX 30 * THZ /* 30s (probably too much) */ +#define TCP_RTO_SYN_RETRIES 3 /* SYN retries without doubling RTO */ +#define TCP_RTO_INIT 1 * THZ /* Initial retransmit timer */ + +void tcp_update_time (f64 now, u32 thread_index); + +/** TCP connection flags */ +#define foreach_tcp_connection_flag \ + _(DELACK, "Delay ACK") \ + _(SNDACK, "Send ACK") \ + _(BURSTACK, "Burst ACK set") \ + _(SENT_RCV_WND0, "Sent 0 receive window") \ + _(RECOVERY, "Recovery on") \ + _(FAST_RECOVERY, "Fast Recovery on") + +typedef enum _tcp_connection_flag_bits +{ +#define _(sym, str) TCP_CONN_##sym##_BIT, + foreach_tcp_connection_flag +#undef _ + TCP_CONN_N_FLAG_BITS +} tcp_connection_flag_bits_e; + +typedef enum _tcp_connection_flag +{ +#define _(sym, str) TCP_CONN_##sym = 1 << TCP_CONN_##sym##_BIT, + foreach_tcp_connection_flag +#undef _ + TCP_CONN_N_FLAGS +} tcp_connection_flags_e; + +/** TCP buffer flags */ +#define foreach_tcp_buf_flag \ + _ (ACK) /**< Sending ACK. */ \ + _ (DUPACK) /**< Sending DUPACK. */ \ + +enum +{ +#define _(f) TCP_BUF_BIT_##f, + foreach_tcp_buf_flag +#undef _ + TCP_N_BUF_BITS, +}; + +enum +{ +#define _(f) TCP_BUF_FLAG_##f = 1 << TCP_BUF_BIT_##f, + foreach_tcp_buf_flag +#undef _ +}; + +#define TCP_MAX_SACK_BLOCKS 5 /**< Max number of SACK blocks stored */ +#define TCP_INVALID_SACK_HOLE_INDEX ((u32)~0) + +typedef struct _sack_scoreboard_hole +{ + u32 next; /**< Index for next entry in linked list */ + u32 prev; /**< Index for previous entry in linked list */ + u32 start; /**< Start sequence number */ + u32 end; /**< End sequence number */ +} sack_scoreboard_hole_t; + +typedef struct _sack_scoreboard +{ + sack_scoreboard_hole_t *holes; /**< Pool of holes */ + u32 head; /**< Index to first entry */ + u32 sacked_bytes; /**< Number of bytes sacked in sb */ +} sack_scoreboard_t; + +typedef enum _tcp_cc_algorithm_type +{ + TCP_CC_NEWRENO, +} tcp_cc_algorithm_type_e; + +typedef struct _tcp_cc_algorithm tcp_cc_algorithm_t; + +typedef enum _tcp_cc_ack_t +{ + TCP_CC_ACK, + TCP_CC_DUPACK, + TCP_CC_PARTIALACK +} tcp_cc_ack_t; + +typedef struct _tcp_connection +{ + transport_connection_t connection; /**< Common transport data. First! */ + + u8 state; /**< TCP state as per tcp_state_t */ + u16 flags; /**< Connection flags (see tcp_conn_flags_e) */ + u32 timers[TCP_N_TIMERS]; /**< Timer handles into timer wheel */ + + /* TODO RFC4898 */ + + /** Send sequence variables RFC793 */ + u32 snd_una; /**< oldest unacknowledged sequence number */ + u32 snd_una_max; /**< newest unacknowledged sequence number + 1*/ + u32 snd_wnd; /**< send window */ + u32 snd_wl1; /**< seq number used for last snd.wnd update */ + u32 snd_wl2; /**< ack number used for last snd.wnd update */ + u32 snd_nxt; /**< next seq number to be sent */ + + /** Receive sequence variables RFC793 */ + u32 rcv_nxt; /**< next sequence number expected */ + u32 rcv_wnd; /**< receive window we expect */ + + u32 rcv_las; /**< rcv_nxt at last ack sent/rcv_wnd update */ + u32 iss; /**< initial sent sequence */ + u32 irs; /**< initial remote sequence */ + + /* Options */ + tcp_options_t opt; /**< TCP connection options parsed */ + u8 rcv_wscale; /**< Window scale to advertise to peer */ + u8 snd_wscale; /**< Window scale to use when sending */ + u32 tsval_recent; /**< Last timestamp received */ + u32 tsval_recent_age; /**< When last updated tstamp_recent*/ + + sack_block_t *snd_sacks; /**< Vector of SACKs to send. XXX Fixed size? */ + sack_scoreboard_t sack_sb; /**< SACK "scoreboard" that tracks holes */ + + u8 rcv_dupacks; /**< Number of DUPACKs received */ + u8 snt_dupacks; /**< Number of DUPACKs sent in a burst */ + + /* Congestion control */ + u32 cwnd; /**< Congestion window */ + u32 ssthresh; /**< Slow-start threshold */ + u32 prev_ssthresh; /**< ssthresh before congestion */ + u32 bytes_acked; /**< Bytes acknowledged by current segment */ + u32 rtx_bytes; /**< Retransmitted bytes */ + u32 tsecr_last_ack; /**< Timestamp echoed to us in last health ACK */ + tcp_cc_algorithm_t *cc_algo; /**< Congestion control algorithm */ + + /* RTT and RTO */ + u32 rto; /**< Retransmission timeout */ + u32 rto_boff; /**< Index for RTO backoff */ + u32 srtt; /**< Smoothed RTT */ + u32 rttvar; /**< Smoothed mean RTT difference. Approximates variance */ + u32 rtt_ts; /**< Timestamp for tracked ACK */ + u32 rtt_seq; /**< Sequence number for tracked ACK */ + + u16 snd_mss; /**< Send MSS */ +} tcp_connection_t; + +struct _tcp_cc_algorithm +{ + void (*rcv_ack) (tcp_connection_t * tc); + void (*rcv_cong_ack) (tcp_connection_t * tc, tcp_cc_ack_t ack); + void (*congestion) (tcp_connection_t * tc); + void (*recovered) (tcp_connection_t * tc); + void (*init) (tcp_connection_t * tc); +}; + +#define tcp_fastrecovery_on(tc) (tc)->flags |= TCP_CONN_FAST_RECOVERY +#define tcp_fastrecovery_off(tc) (tc)->flags &= ~TCP_CONN_FAST_RECOVERY +#define tcp_in_fastrecovery(tc) ((tc)->flags & TCP_CONN_FAST_RECOVERY) +#define tcp_in_recovery(tc) ((tc)->flags & (TCP_CONN_FAST_RECOVERY | TCP_CONN_RECOVERY)) +#define tcp_recovery_off(tc) ((tc)->flags &= ~(TCP_CONN_FAST_RECOVERY | TCP_CONN_RECOVERY)) +#define tcp_in_slowstart(tc) (tc->cwnd < tc->ssthresh) + +typedef enum +{ + TCP_IP4, + TCP_IP6, + TCP_N_AF, +} tcp_af_t; + +typedef enum _tcp_error +{ +#define tcp_error(n,s) TCP_ERROR_##n, +#include +#undef tcp_error + TCP_N_ERROR, +} tcp_error_t; + +typedef struct _tcp_lookup_dispatch +{ + u8 next, error; +} tcp_lookup_dispatch_t; + +typedef struct _tcp_main +{ + /* Per-worker thread tcp connection pools */ + tcp_connection_t **connections; + + /* Pool of listeners. */ + tcp_connection_t *listener_pool; + + /** Dispatch table by state and flags */ + tcp_lookup_dispatch_t dispatch_table[TCP_N_STATES][64]; + + u8 log2_tstamp_clocks_per_tick; + f64 tstamp_ticks_per_clock; + + /** per-worker tx buffer free lists */ + u32 **tx_buffers; + + /* Per worker-thread timer wheel for connections timers */ + tw_timer_wheel_16t_2w_512sl_t *timer_wheels; + + /* Convenience per worker-thread vector of connections to DELACK */ + u32 **delack_connections; + + /* Pool of half-open connections on which we've sent a SYN */ + tcp_connection_t *half_open_connections; + + /* Pool of local TCP endpoints */ + transport_endpoint_t *local_endpoints; + + /* Local endpoints lookup table */ + transport_endpoint_table_t local_endpoints_table; + + /* Congestion control algorithms registered */ + tcp_cc_algorithm_t *cc_algos; + + /* convenience */ + vlib_main_t *vlib_main; + vnet_main_t *vnet_main; + ip4_main_t *ip4_main; + ip6_main_t *ip6_main; +} tcp_main_t; + +extern tcp_main_t tcp_main; +extern vlib_node_registration_t tcp4_input_node; +extern vlib_node_registration_t tcp6_input_node; +extern vlib_node_registration_t tcp4_output_node; +extern vlib_node_registration_t tcp6_output_node; + +always_inline tcp_main_t * +vnet_get_tcp_main () +{ + return &tcp_main; +} + +always_inline tcp_connection_t * +tcp_connection_get (u32 conn_index, u32 thread_index) +{ + return pool_elt_at_index (tcp_main.connections[thread_index], conn_index); +} + +always_inline tcp_connection_t * +tcp_connection_get_if_valid (u32 conn_index, u32 thread_index) +{ + if (tcp_main.connections[thread_index] == 0) + return 0; + if (pool_is_free_index (tcp_main.connections[thread_index], conn_index)) + return 0; + return pool_elt_at_index (tcp_main.connections[thread_index], conn_index); +} + +void tcp_connection_close (tcp_connection_t * tc); +void tcp_connection_cleanup (tcp_connection_t * tc); +void tcp_connection_del (tcp_connection_t * tc); + +always_inline tcp_connection_t * +tcp_listener_get (u32 tli) +{ + return pool_elt_at_index (tcp_main.listener_pool, tli); +} + +always_inline tcp_connection_t * +tcp_half_open_connection_get (u32 conn_index) +{ + return pool_elt_at_index (tcp_main.half_open_connections, conn_index); +} + +void tcp_make_ack (tcp_connection_t * ts, vlib_buffer_t * b); +void tcp_make_finack (tcp_connection_t * tc, vlib_buffer_t * b); +void tcp_make_synack (tcp_connection_t * ts, vlib_buffer_t * b); +void tcp_send_reset (vlib_buffer_t * pkt, u8 is_ip4); +void tcp_send_syn (tcp_connection_t * tc); +void tcp_send_fin (tcp_connection_t * tc); +void tcp_set_snd_mss (tcp_connection_t * tc); + +always_inline u32 +tcp_end_seq (tcp_header_t * th, u32 len) +{ + return th->seq_number + tcp_is_syn (th) + tcp_is_fin (th) + len; +} + +/* Modulo arithmetic for TCP sequence numbers */ +#define seq_lt(_s1, _s2) ((i32)((_s1)-(_s2)) < 0) +#define seq_leq(_s1, _s2) ((i32)((_s1)-(_s2)) <= 0) +#define seq_gt(_s1, _s2) ((i32)((_s1)-(_s2)) > 0) +#define seq_geq(_s1, _s2) ((i32)((_s1)-(_s2)) >= 0) + +/* Modulo arithmetic for timestamps */ +#define timestamp_lt(_t1, _t2) ((i32)((_t1)-(_t2)) < 0) +#define timestamp_leq(_t1, _t2) ((i32)((_t1)-(_t2)) <= 0) + +always_inline u32 +tcp_flight_size (const tcp_connection_t * tc) +{ + return tc->snd_una_max - tc->snd_una - tc->sack_sb.sacked_bytes + + tc->rtx_bytes; +} + +/** + * Initial cwnd as per RFC5681 + */ +always_inline u32 +tcp_initial_cwnd (const tcp_connection_t * tc) +{ + if (tc->snd_mss > 2190) + return 2 * tc->snd_mss; + else if (tc->snd_mss > 1095) + return 3 * tc->snd_mss; + else + return 4 * tc->snd_mss; +} + +always_inline u32 +tcp_loss_wnd (const tcp_connection_t * tc) +{ + return tc->snd_mss; +} + +always_inline u32 +tcp_available_wnd (const tcp_connection_t * tc) +{ + return clib_min (tc->cwnd, tc->snd_wnd); +} + +always_inline u32 +tcp_available_snd_space (const tcp_connection_t * tc) +{ + u32 available_wnd = tcp_available_wnd (tc); + u32 flight_size = tcp_flight_size (tc); + + if (available_wnd <= flight_size) + return 0; + + return available_wnd - flight_size; +} + +void tcp_retransmit_first_unacked (tcp_connection_t * tc); + +void tcp_fast_retransmit (tcp_connection_t * tc); + +always_inline u32 +tcp_time_now (void) +{ + return clib_cpu_time_now () * tcp_main.tstamp_ticks_per_clock; +} + +u32 tcp_push_header (transport_connection_t * tconn, vlib_buffer_t * b); + +u32 +tcp_prepare_retransmit_segment (tcp_connection_t * tc, vlib_buffer_t * b, + u32 max_bytes); + +void tcp_connection_timers_init (tcp_connection_t * tc); +void tcp_connection_timers_reset (tcp_connection_t * tc); + +void tcp_connection_init_vars (tcp_connection_t * tc); + +always_inline void +tcp_connection_force_ack (tcp_connection_t * tc, vlib_buffer_t * b) +{ + /* Reset flags, make sure ack is sent */ + tc->flags = TCP_CONN_SNDACK; + vnet_buffer (b)->tcp.flags &= ~TCP_BUF_FLAG_DUPACK; +} + +always_inline void +tcp_timer_set (tcp_connection_t * tc, u8 timer_id, u32 interval) +{ + tc->timers[timer_id] + = tw_timer_start_16t_2w_512sl (&tcp_main.timer_wheels[tc->c_thread_index], + tc->c_c_index, timer_id, interval); +} + +always_inline void +tcp_retransmit_timer_set (tcp_main_t * tm, tcp_connection_t * tc) +{ + /* XXX Switch to faster TW */ + tcp_timer_set (tc, TCP_TIMER_RETRANSMIT, + clib_max (tc->rto * TCP_TO_TIMER_TICK, 1)); +} + +always_inline void +tcp_timer_reset (tcp_connection_t * tc, u8 timer_id) +{ + if (tc->timers[timer_id] == TCP_TIMER_HANDLE_INVALID) + return; + + tw_timer_stop_16t_2w_512sl (&tcp_main.timer_wheels[tc->c_thread_index], + tc->timers[timer_id]); + tc->timers[timer_id] = TCP_TIMER_HANDLE_INVALID; +} + +always_inline void +tcp_timer_update (tcp_connection_t * tc, u8 timer_id, u32 interval) +{ + if (tc->timers[timer_id] != TCP_TIMER_HANDLE_INVALID) + tw_timer_stop_16t_2w_512sl (&tcp_main.timer_wheels[tc->c_thread_index], + tc->timers[timer_id]); + tc->timers[timer_id] = + tw_timer_start_16t_2w_512sl (&tcp_main.timer_wheels[tc->c_thread_index], + tc->c_c_index, timer_id, interval); +} + +always_inline u8 +tcp_timer_is_active (tcp_connection_t * tc, tcp_timers_e timer) +{ + return tc->timers[timer] != TCP_TIMER_HANDLE_INVALID; +} + +void +scoreboard_remove_hole (sack_scoreboard_t * sb, + sack_scoreboard_hole_t * hole); + +always_inline sack_scoreboard_hole_t * +scoreboard_next_hole (sack_scoreboard_t * sb, sack_scoreboard_hole_t * hole) +{ + if (hole->next != TCP_INVALID_SACK_HOLE_INDEX) + return pool_elt_at_index (sb->holes, hole->next); + return 0; +} + +always_inline sack_scoreboard_hole_t * +scoreboard_first_hole (sack_scoreboard_t * sb) +{ + if (sb->head != TCP_INVALID_SACK_HOLE_INDEX) + return pool_elt_at_index (sb->holes, sb->head); + return 0; +} + +always_inline void +scoreboard_clear (sack_scoreboard_t * sb) +{ + sack_scoreboard_hole_t *hole = scoreboard_first_hole (sb); + while ((hole = scoreboard_first_hole (sb))) + { + scoreboard_remove_hole (sb, hole); + } +} + +always_inline u32 +scoreboard_hole_bytes (sack_scoreboard_hole_t * hole) +{ + return hole->end - hole->start; +} + +always_inline void +tcp_cc_algo_register (tcp_cc_algorithm_type_e type, + const tcp_cc_algorithm_t * vft) +{ + tcp_main_t *tm = vnet_get_tcp_main (); + vec_validate (tm->cc_algos, type); + + tm->cc_algos[type] = *vft; +} + +always_inline tcp_cc_algorithm_t * +tcp_cc_algo_get (tcp_cc_algorithm_type_e type) +{ + tcp_main_t *tm = vnet_get_tcp_main (); + return &tm->cc_algos[type]; +} + +void tcp_cc_init (tcp_connection_t * tc); + +/** + * Push TCP header to buffer + * + * @param vm - vlib_main + * @param b - buffer to write the header to + * @param sp_net - source port net order + * @param dp_net - destination port net order + * @param seq - sequence number net order + * @param ack - ack number net order + * @param tcp_hdr_opts_len - header and options length in bytes + * @param flags - header flags + * @param wnd - window size + * + * @return - pointer to start of TCP header + */ +always_inline void * +vlib_buffer_push_tcp_net_order (vlib_buffer_t * b, u16 sp, u16 dp, u32 seq, + u32 ack, u8 tcp_hdr_opts_len, u8 flags, + u16 wnd) +{ + tcp_header_t *th; + + th = vlib_buffer_push_uninit (b, tcp_hdr_opts_len); + + th->src_port = sp; + th->dst_port = dp; + th->seq_number = seq; + th->ack_number = ack; + th->data_offset_and_reserved = (tcp_hdr_opts_len >> 2) << 4; + th->flags = flags; + th->window = wnd; + th->checksum = 0; + th->urgent_pointer = 0; + return th; +} + +/** + * Push TCP header to buffer + * + * @param vm - vlib_main + * @param b - buffer to write the header to + * @param sp_net - source port net order + * @param dp_net - destination port net order + * @param seq - sequence number host order + * @param ack - ack number host order + * @param tcp_hdr_opts_len - header and options length in bytes + * @param flags - header flags + * @param wnd - window size + * + * @return - pointer to start of TCP header + */ +always_inline void * +vlib_buffer_push_tcp (vlib_buffer_t * b, u16 sp_net, u16 dp_net, u32 seq, + u32 ack, u8 tcp_hdr_opts_len, u8 flags, u16 wnd) +{ + return vlib_buffer_push_tcp_net_order (b, sp_net, dp_net, + clib_host_to_net_u32 (seq), + clib_host_to_net_u32 (ack), + tcp_hdr_opts_len, flags, + clib_host_to_net_u16 (wnd)); +} + +#endif /* _vnet_tcp_h_ */ + +/* + * fd.io coding-style-patch-verification: ON + * + * Local Variables: + * eval: (c-set-style "gnu") + * End: + */ diff --git a/src/vnet/tcp/tcp_error.def b/src/vnet/tcp/tcp_error.def new file mode 100644 index 00000000..cff5ec13 --- /dev/null +++ b/src/vnet/tcp/tcp_error.def @@ -0,0 +1,35 @@ +/* + * Copyright (c) 2016 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +tcp_error (NONE, "no error") +tcp_error (NO_LISTENER, "no listener for dst port") +tcp_error (LOOKUP_DROPS, "lookup drops") +tcp_error (DISPATCH, "Dispatch error") +tcp_error (ENQUEUED, "Packets pushed into rx fifo") +tcp_error (PURE_ACK, "Pure acks") +tcp_error (SYNS_RCVD, "SYNs received") +tcp_error (SYN_ACKS_RCVD, "SYN-ACKs received") +tcp_error (NOT_READY, "Session not ready for packets") +tcp_error (FIFO_FULL, "Packets dropped for lack of rx fifo space") +tcp_error (EVENT_FIFO_FULL, "Events not sent for lack of event fifo space") +tcp_error (API_QUEUE_FULL, "Sessions not created for lack of API queue space") +tcp_error (CREATE_SESSION_FAIL, "Sessions couldn't be allocated") +tcp_error (SEGMENT_INVALID, "Invalid segment") +tcp_error (ACK_INVALID, "Invalid ACK") +tcp_error (ACK_DUP, "Duplicate ACK") +tcp_error (ACK_OLD, "Old ACK") +tcp_error (PKTS_SENT, "Packets sent") +tcp_error (FILTERED_DUPACKS, "Filtered duplicate ACKs") +tcp_error (RST_SENT, "Resets sent") \ No newline at end of file diff --git a/src/vnet/tcp/tcp_format.c b/src/vnet/tcp/tcp_format.c new file mode 100644 index 00000000..7136741d --- /dev/null +++ b/src/vnet/tcp/tcp_format.c @@ -0,0 +1,136 @@ +/* + * Copyright (c) 2015 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ +/* + * tcp/tcp_format.c: tcp formatting + * + * Copyright (c) 2008 Eliot Dresselhaus + * + * Permission is hereby granted, free of charge, to any person obtaining + * a copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sublicense, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice shall be + * included in all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND + * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE + * LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + * OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + * WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + */ + +#include +#include + +static u8 * +format_tcp_flags (u8 * s, va_list * args) +{ + int flags = va_arg (*args, int); + +#define _(f) if (flags & TCP_FLAG_##f) s = format (s, "%s, ", #f); + foreach_tcp_flag +#undef _ + return s; +} + +/* Format TCP header. */ +u8 * +format_tcp_header (u8 * s, va_list * args) +{ + tcp_header_t *tcp = va_arg (*args, tcp_header_t *); + u32 max_header_bytes = va_arg (*args, u32); + u32 header_bytes; + uword indent; + + /* Nothing to do. */ + if (max_header_bytes < sizeof (tcp[0])) + return format (s, "TCP header truncated"); + + indent = format_get_indent (s); + indent += 2; + header_bytes = tcp_header_bytes (tcp); + + s = format (s, "TCP: %d -> %d", clib_net_to_host_u16 (tcp->src), + clib_net_to_host_u16 (tcp->dst)); + + s = format (s, "\n%Useq. 0x%08x ack 0x%08x", format_white_space, indent, + clib_net_to_host_u32 (tcp->seq_number), + clib_net_to_host_u32 (tcp->ack_number)); + + s = format (s, "\n%Uflags %U, tcp header: %d bytes", format_white_space, + indent, format_tcp_flags, tcp->flags, header_bytes); + + s = format (s, "\n%Uwindow %d, checksum 0x%04x", format_white_space, indent, + clib_net_to_host_u16 (tcp->window), + clib_net_to_host_u16 (tcp->checksum)); + + +#if 0 + /* Format TCP options. */ + { + u8 *o; + u8 *option_start = (void *) (tcp + 1); + u8 *option_end = (void *) tcp + header_bytes; + + for (o = option_start; o < option_end;) + { + u32 length = o[1]; + switch (o[0]) + { + case TCP_OPTION_END: + length = 1; + o = option_end; + break; + + case TCP_OPTION_NOOP: + length = 1; + break; + + } + } + } +#endif + + /* Recurse into next protocol layer. */ + if (max_header_bytes != 0 && header_bytes < max_header_bytes) + { + ip_main_t *im = &ip_main; + tcp_udp_port_info_t *pi; + + pi = ip_get_tcp_udp_port_info (im, tcp->dst); + + if (pi && pi->format_header) + s = format (s, "\n%U%U", format_white_space, indent - 2, + pi->format_header, + /* next protocol header */ (void *) tcp + header_bytes, + max_header_bytes - header_bytes); + } + + return s; +} + +/* + * fd.io coding-style-patch-verification: ON + * + * Local Variables: + * eval: (c-set-style "gnu") + * End: + */ diff --git a/src/vnet/tcp/tcp_input.c b/src/vnet/tcp/tcp_input.c new file mode 100644 index 00000000..daa0683b --- /dev/null +++ b/src/vnet/tcp/tcp_input.c @@ -0,0 +1,2316 @@ +/* + * Copyright (c) 2016 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +#include +#include +#include +#include +#include + +static char *tcp_error_strings[] = { +#define tcp_error(n,s) s, +#include +#undef tcp_error +}; + +/* All TCP nodes have the same outgoing arcs */ +#define foreach_tcp_state_next \ + _ (DROP, "error-drop") \ + _ (TCP4_OUTPUT, "tcp4-output") \ + _ (TCP6_OUTPUT, "tcp6-output") + +typedef enum _tcp_established_next +{ +#define _(s,n) TCP_ESTABLISHED_NEXT_##s, + foreach_tcp_state_next +#undef _ + TCP_ESTABLISHED_N_NEXT, +} tcp_established_next_t; + +typedef enum _tcp_rcv_process_next +{ +#define _(s,n) TCP_RCV_PROCESS_NEXT_##s, + foreach_tcp_state_next +#undef _ + TCP_RCV_PROCESS_N_NEXT, +} tcp_rcv_process_next_t; + +typedef enum _tcp_syn_sent_next +{ +#define _(s,n) TCP_SYN_SENT_NEXT_##s, + foreach_tcp_state_next +#undef _ + TCP_SYN_SENT_N_NEXT, +} tcp_syn_sent_next_t; + +typedef enum _tcp_listen_next +{ +#define _(s,n) TCP_LISTEN_NEXT_##s, + foreach_tcp_state_next +#undef _ + TCP_LISTEN_N_NEXT, +} tcp_listen_next_t; + +/* Generic, state independent indices */ +typedef enum _tcp_state_next +{ +#define _(s,n) TCP_NEXT_##s, + foreach_tcp_state_next +#undef _ + TCP_STATE_N_NEXT, +} tcp_state_next_t; + +#define tcp_next_output(is_ip4) (is_ip4 ? TCP_NEXT_TCP4_OUTPUT \ + : TCP_NEXT_TCP6_OUTPUT) + +vlib_node_registration_t tcp4_established_node; +vlib_node_registration_t tcp6_established_node; + +/** + * Validate segment sequence number. As per RFC793: + * + * Segment Receive Test + * Length Window + * ------- ------- ------------------------------------------- + * 0 0 SEG.SEQ = RCV.NXT + * 0 >0 RCV.NXT =< SEG.SEQ < RCV.NXT+RCV.WND + * >0 0 not acceptable + * >0 >0 RCV.NXT =< SEG.SEQ < RCV.NXT+RCV.WND + * or RCV.NXT =< SEG.SEQ+SEG.LEN-1 < RCV.NXT+RCV.WND + * + * This ultimately consists in checking if segment falls within the window. + * The one important difference compared to RFC793 is that we use rcv_las, + * or the rcv_nxt at last ack sent instead of rcv_nxt since that's the + * peer's reference when computing our receive window. + * + * This accepts only segments within the window. + */ +always_inline u8 +tcp_segment_in_rcv_wnd (tcp_connection_t * tc, u32 seq, u32 end_seq) +{ + return seq_leq (end_seq, tc->rcv_las + tc->rcv_wnd) + && seq_geq (seq, tc->rcv_nxt); +} + +void +tcp_options_parse (tcp_header_t * th, tcp_options_t * to) +{ + const u8 *data; + u8 opt_len, opts_len, kind; + int j; + sack_block_t b; + + opts_len = (tcp_doff (th) << 2) - sizeof (tcp_header_t); + data = (const u8 *) (th + 1); + + /* Zero out all flags but those set in SYN */ + to->flags &= (TCP_OPTS_FLAG_SACK_PERMITTED | TCP_OPTS_FLAG_WSCALE); + + for (; opts_len > 0; opts_len -= opt_len, data += opt_len) + { + kind = data[0]; + + /* Get options length */ + if (kind == TCP_OPTION_EOL) + break; + else if (kind == TCP_OPTION_NOOP) + opt_len = 1; + else + { + /* broken options */ + if (opts_len < 2) + break; + opt_len = data[1]; + + /* weird option length */ + if (opt_len < 2 || opt_len > opts_len) + break; + } + + /* Parse options */ + switch (kind) + { + case TCP_OPTION_MSS: + if ((opt_len == TCP_OPTION_LEN_MSS) && tcp_syn (th)) + { + to->flags |= TCP_OPTS_FLAG_MSS; + to->mss = clib_net_to_host_u16 (*(u16 *) (data + 2)); + } + break; + case TCP_OPTION_WINDOW_SCALE: + if ((opt_len == TCP_OPTION_LEN_WINDOW_SCALE) && tcp_syn (th)) + { + to->flags |= TCP_OPTS_FLAG_WSCALE; + to->wscale = data[2]; + if (to->wscale > TCP_MAX_WND_SCALE) + { + clib_warning ("Illegal window scaling value: %d", + to->wscale); + to->wscale = TCP_MAX_WND_SCALE; + } + } + break; + case TCP_OPTION_TIMESTAMP: + if (opt_len == TCP_OPTION_LEN_TIMESTAMP) + { + to->flags |= TCP_OPTS_FLAG_TSTAMP; + to->tsval = clib_net_to_host_u32 (*(u32 *) (data + 2)); + to->tsecr = clib_net_to_host_u32 (*(u32 *) (data + 6)); + } + break; + case TCP_OPTION_SACK_PERMITTED: + if (opt_len == TCP_OPTION_LEN_SACK_PERMITTED && tcp_syn (th)) + to->flags |= TCP_OPTS_FLAG_SACK_PERMITTED; + break; + case TCP_OPTION_SACK_BLOCK: + /* If SACK permitted was not advertised or a SYN, break */ + if ((to->flags & TCP_OPTS_FLAG_SACK_PERMITTED) == 0 || tcp_syn (th)) + break; + + /* If too short or not correctly formatted, break */ + if (opt_len < 10 || ((opt_len - 2) % TCP_OPTION_LEN_SACK_BLOCK)) + break; + + to->flags |= TCP_OPTS_FLAG_SACK; + to->n_sack_blocks = (opt_len - 2) / TCP_OPTION_LEN_SACK_BLOCK; + vec_reset_length (to->sacks); + for (j = 0; j < to->n_sack_blocks; j++) + { + b.start = clib_net_to_host_u32 (*(u32 *) (data + 2 + 4 * j)); + b.end = clib_net_to_host_u32 (*(u32 *) (data + 6 + 4 * j)); + vec_add1 (to->sacks, b); + } + break; + default: + /* Nothing to see here */ + continue; + } + } +} + +always_inline int +tcp_segment_check_paws (tcp_connection_t * tc) +{ + /* XXX normally test for timestamp should be lt instead of leq, but for + * local testing this is not enough */ + return tcp_opts_tstamp (&tc->opt) && tc->tsval_recent + && timestamp_lt (tc->opt.tsval, tc->tsval_recent); +} + +/** + * Validate incoming segment as per RFC793 p. 69 and RFC1323 p. 19 + * + * It first verifies if segment has a wrapped sequence number (PAWS) and then + * does the processing associated to the first four steps (ignoring security + * and precedence): sequence number, rst bit and syn bit checks. + * + * @return 0 if segments passes validation. + */ +static int +tcp_segment_validate (vlib_main_t * vm, tcp_connection_t * tc0, + vlib_buffer_t * b0, tcp_header_t * th0, u32 * next0) +{ + u8 paws_failed; + + if (PREDICT_FALSE (!tcp_ack (th0) && !tcp_rst (th0) && !tcp_syn (th0))) + return -1; + + tcp_options_parse (th0, &tc0->opt); + + /* RFC1323: Check against wrapped sequence numbers (PAWS). If we have + * timestamp to echo and it's less than tsval_recent, drop segment + * but still send an ACK in order to retain TCP's mechanism for detecting + * and recovering from half-open connections */ + paws_failed = tcp_segment_check_paws (tc0); + if (paws_failed) + { + clib_warning ("paws failed"); + + /* If it just so happens that a segment updates tsval_recent for a + * segment over 24 days old, invalidate tsval_recent. */ + if (timestamp_lt (tc0->tsval_recent_age + TCP_PAWS_IDLE, + tcp_time_now ())) + { + /* Age isn't reset until we get a valid tsval (bsd inspired) */ + tc0->tsval_recent = 0; + } + else + { + /* Drop after ack if not rst */ + if (!tcp_rst (th0)) + { + tcp_make_ack (tc0, b0); + *next0 = tcp_next_output (tc0->c_is_ip4); + return -1; + } + } + } + + /* 1st: check sequence number */ + if (!tcp_segment_in_rcv_wnd (tc0, vnet_buffer (b0)->tcp.seq_number, + vnet_buffer (b0)->tcp.seq_end)) + { + if (!tcp_rst (th0)) + { + /* Send dup ack */ + tcp_make_ack (tc0, b0); + *next0 = tcp_next_output (tc0->c_is_ip4); + } + return -1; + } + + /* 2nd: check the RST bit */ + if (tcp_rst (th0)) + { + /* Notify session that connection has been reset. Switch + * state to closed and await for session to do the cleanup. */ + stream_session_reset_notify (&tc0->connection); + tc0->state = TCP_STATE_CLOSED; + return -1; + } + + /* 3rd: check security and precedence (skip) */ + + /* 4th: check the SYN bit */ + if (tcp_syn (th0)) + { + tcp_send_reset (b0, tc0->c_is_ip4); + return -1; + } + + /* If PAWS passed and segment in window, save timestamp */ + if (!paws_failed) + { + tc0->tsval_recent = tc0->opt.tsval; + tc0->tsval_recent_age = tcp_time_now (); + } + + return 0; +} + +always_inline int +tcp_rcv_ack_is_acceptable (tcp_connection_t * tc0, vlib_buffer_t * tb0) +{ + /* SND.UNA =< SEG.ACK =< SND.NXT */ + return (seq_leq (tc0->snd_una, vnet_buffer (tb0)->tcp.ack_number) + && seq_leq (vnet_buffer (tb0)->tcp.ack_number, tc0->snd_nxt)); +} + +/** + * Compute smoothed RTT as per VJ's '88 SIGCOMM and RFC6298 + * + * Note that although the original article, srtt and rttvar are scaled + * to minimize round-off errors, here we don't. Instead, we rely on + * better precision time measurements. + * + * TODO support us rtt resolution + */ +static void +tcp_estimate_rtt (tcp_connection_t * tc, u32 mrtt) +{ + int err; + + if (tc->srtt != 0) + { + err = mrtt - tc->srtt; + tc->srtt += err >> 3; + + /* XXX Drop in RTT results in RTTVAR increase and bigger RTO. + * The increase should be bound */ + tc->rttvar += (clib_abs (err) - tc->rttvar) >> 2; + } + else + { + /* First measurement. */ + tc->srtt = mrtt; + tc->rttvar = mrtt << 1; + } +} + +/** Update RTT estimate and RTO timer + * + * Measure RTT: We have two sources of RTT measurements: TSOPT and ACK + * timing. Middle boxes are known to fiddle with TCP options so we + * should give higher priority to ACK timing. + * + * return 1 if valid rtt 0 otherwise + */ +static int +tcp_update_rtt (tcp_connection_t * tc, u32 ack) +{ + u32 mrtt = 0; + + /* Karn's rule, part 1. Don't use retransmitted segments to estimate + * RTT because they're ambiguous. */ + if (tc->rtt_seq && seq_gt (ack, tc->rtt_seq) && !tc->rto_boff) + { + mrtt = tcp_time_now () - tc->rtt_ts; + tc->rtt_seq = 0; + } + + /* As per RFC7323 TSecr can be used for RTTM only if the segment advances + * snd_una, i.e., the left side of the send window: + * seq_lt (tc->snd_una, ack). Note: last condition could be dropped, we don't + * try to update rtt for dupacks */ + else if (tcp_opts_tstamp (&tc->opt) && tc->opt.tsecr && tc->bytes_acked) + { + mrtt = tcp_time_now () - tc->opt.tsecr; + } + + /* Ignore dubious measurements */ + if (mrtt == 0 || mrtt > TCP_RTT_MAX) + return 0; + + tcp_estimate_rtt (tc, mrtt); + + tc->rto = clib_min (tc->srtt + (tc->rttvar << 2), TCP_RTO_MAX); + + return 1; +} + +/** + * Dequeue bytes that have been acked and while at it update RTT estimates. + */ +static void +tcp_dequeue_acked (tcp_connection_t * tc, u32 ack) +{ + /* Dequeue the newly ACKed bytes */ + stream_session_dequeue_drop (&tc->connection, tc->bytes_acked); + + /* Update rtt and rto */ + if (tcp_update_rtt (tc, ack)) + { + /* Good ACK received and valid RTT, make sure retransmit backoff is 0 */ + tc->rto_boff = 0; + } +} + +/** Check if dupack as per RFC5681 Sec. 2 */ +always_inline u8 +tcp_ack_is_dupack (tcp_connection_t * tc, vlib_buffer_t * b, u32 new_snd_wnd) +{ + return ((vnet_buffer (b)->tcp.ack_number == tc->snd_una) + && seq_gt (tc->snd_una_max, tc->snd_una) + && (vnet_buffer (b)->tcp.seq_end == vnet_buffer (b)->tcp.seq_number) + && (new_snd_wnd == tc->snd_wnd)); +} + +void +scoreboard_remove_hole (sack_scoreboard_t * sb, sack_scoreboard_hole_t * hole) +{ + sack_scoreboard_hole_t *next, *prev; + + if (hole->next != TCP_INVALID_SACK_HOLE_INDEX) + { + next = pool_elt_at_index (sb->holes, hole->next); + next->prev = hole->prev; + } + + if (hole->prev != TCP_INVALID_SACK_HOLE_INDEX) + { + prev = pool_elt_at_index (sb->holes, hole->prev); + prev->next = hole->next; + } + else + { + sb->head = hole->next; + } + + pool_put (sb->holes, hole); +} + +sack_scoreboard_hole_t * +scoreboard_insert_hole (sack_scoreboard_t * sb, sack_scoreboard_hole_t * prev, + u32 start, u32 end) +{ + sack_scoreboard_hole_t *hole, *next; + u32 hole_index; + + pool_get (sb->holes, hole); + memset (hole, 0, sizeof (*hole)); + + hole->start = start; + hole->end = end; + hole_index = hole - sb->holes; + + if (prev) + { + hole->prev = prev - sb->holes; + hole->next = prev->next; + + if ((next = scoreboard_next_hole (sb, hole))) + next->prev = hole_index; + + prev->next = hole_index; + } + else + { + sb->head = hole_index; + hole->prev = TCP_INVALID_SACK_HOLE_INDEX; + hole->next = TCP_INVALID_SACK_HOLE_INDEX; + } + + return hole; +} + +static void +tcp_rcv_sacks (tcp_connection_t * tc, u32 ack) +{ + sack_scoreboard_t *sb = &tc->sack_sb; + sack_block_t *blk, tmp; + sack_scoreboard_hole_t *hole, *next_hole; + u32 blk_index = 0; + int i, j; + + if (!tcp_opts_sack (tc) && sb->head == TCP_INVALID_SACK_HOLE_INDEX) + return; + + /* Remove invalid blocks */ + vec_foreach (blk, tc->opt.sacks) + { + if (seq_lt (blk->start, blk->end) + && seq_gt (blk->start, tc->snd_una) + && seq_gt (blk->start, ack) && seq_lt (blk->end, tc->snd_nxt)) + continue; + + vec_del1 (tc->opt.sacks, blk - tc->opt.sacks); + } + + /* Add block for cumulative ack */ + if (seq_gt (ack, tc->snd_una)) + { + tmp.start = tc->snd_una; + tmp.end = ack; + vec_add1 (tc->opt.sacks, tmp); + } + + if (vec_len (tc->opt.sacks) == 0) + return; + + /* Make sure blocks are ordered */ + for (i = 0; i < vec_len (tc->opt.sacks); i++) + for (j = i; j < vec_len (tc->opt.sacks); j++) + if (seq_lt (tc->opt.sacks[j].start, tc->opt.sacks[i].start)) + { + tmp = tc->opt.sacks[i]; + tc->opt.sacks[i] = tc->opt.sacks[j]; + tc->opt.sacks[j] = tmp; + } + + /* If no holes, insert the first that covers all outstanding bytes */ + if (sb->head == TCP_INVALID_SACK_HOLE_INDEX) + { + scoreboard_insert_hole (sb, 0, tc->snd_una, tc->snd_una_max); + } + + /* Walk the holes with the SACK blocks */ + hole = pool_elt_at_index (sb->holes, sb->head); + while (hole && blk_index < vec_len (tc->opt.sacks)) + { + blk = &tc->opt.sacks[blk_index]; + + if (seq_leq (blk->start, hole->start)) + { + /* Block covers hole. Remove hole */ + if (seq_geq (blk->end, hole->end)) + { + next_hole = scoreboard_next_hole (sb, hole); + + /* Byte accounting */ + if (seq_lt (hole->end, ack)) + { + /* Bytes lost because snd wnd left edge advances */ + if (seq_lt (next_hole->start, ack)) + sb->sacked_bytes -= next_hole->start - hole->end; + else + sb->sacked_bytes -= ack - hole->end; + } + else + { + sb->sacked_bytes += scoreboard_hole_bytes (hole); + } + + scoreboard_remove_hole (sb, hole); + hole = next_hole; + } + /* Partial overlap */ + else + { + sb->sacked_bytes += blk->end - hole->start; + hole->start = blk->end; + blk_index++; + } + } + else + { + /* Hole must be split */ + if (seq_leq (blk->end, hole->end)) + { + sb->sacked_bytes += blk->end - blk->start; + scoreboard_insert_hole (sb, hole, blk->end, hole->end); + hole->end = blk->start - 1; + blk_index++; + } + else + { + sb->sacked_bytes += hole->end - blk->start + 1; + hole->end = blk->start - 1; + hole = scoreboard_next_hole (sb, hole); + } + } + } +} + +/** Update snd_wnd + * + * If (SND.WL1 < SEG.SEQ or (SND.WL1 = SEG.SEQ and SND.WL2 =< SEG.ACK)), set + * SND.WND <- SEG.WND, set SND.WL1 <- SEG.SEQ, and set SND.WL2 <- SEG.ACK */ +static void +tcp_update_snd_wnd (tcp_connection_t * tc, u32 seq, u32 ack, u32 snd_wnd) +{ + if (tc->snd_wl1 < seq || (tc->snd_wl1 == seq && tc->snd_wl2 <= ack)) + { + tc->snd_wnd = snd_wnd; + tc->snd_wl1 = seq; + tc->snd_wl2 = ack; + } +} + +static void +tcp_cc_congestion (tcp_connection_t * tc) +{ + tc->cc_algo->congestion (tc); +} + +static void +tcp_cc_recover (tcp_connection_t * tc) +{ + if (tcp_in_fastrecovery (tc)) + { + tc->cc_algo->recovered (tc); + tcp_recovery_off (tc); + } + else if (tcp_in_recovery (tc)) + { + tcp_recovery_off (tc); + tc->cwnd = tcp_loss_wnd (tc); + } +} + +static void +tcp_cc_rcv_ack (tcp_connection_t * tc) +{ + u8 partial_ack; + + if (tcp_in_recovery (tc)) + { + partial_ack = seq_lt (tc->snd_una, tc->snd_una_max); + if (!partial_ack) + { + /* Clear retransmitted bytes. */ + tc->rtx_bytes = 0; + tcp_cc_recover (tc); + } + else + { + /* Clear retransmitted bytes. XXX should we clear all? */ + tc->rtx_bytes = 0; + tc->cc_algo->rcv_cong_ack (tc, TCP_CC_PARTIALACK); + + /* Retransmit first unacked segment */ + tcp_retransmit_first_unacked (tc); + } + } + else + { + tc->cc_algo->rcv_ack (tc); + } + + tc->rcv_dupacks = 0; + tc->tsecr_last_ack = tc->opt.tsecr; +} + +static void +tcp_cc_rcv_dupack (tcp_connection_t * tc, u32 ack) +{ + ASSERT (tc->snd_una == ack); + + tc->rcv_dupacks++; + if (tc->rcv_dupacks == TCP_DUPACK_THRESHOLD) + { + /* RFC6582 NewReno heuristic to avoid multiple fast retransmits */ + if (tc->opt.tsecr != tc->tsecr_last_ack) + { + tc->rcv_dupacks = 0; + return; + } + + tcp_fastrecovery_on (tc); + + /* Handle congestion and dupack */ + tcp_cc_congestion (tc); + tc->cc_algo->rcv_cong_ack (tc, TCP_CC_DUPACK); + + tcp_fast_retransmit (tc); + + /* Post retransmit update cwnd to ssthresh and account for the + * three segments that have left the network and should've been + * buffered at the receiver */ + tc->cwnd = tc->ssthresh + TCP_DUPACK_THRESHOLD * tc->snd_mss; + } + else if (tc->rcv_dupacks > TCP_DUPACK_THRESHOLD) + { + ASSERT (tcp_in_fastrecovery (tc)); + + tc->cc_algo->rcv_cong_ack (tc, TCP_CC_DUPACK); + } +} + +void +tcp_cc_init (tcp_connection_t * tc) +{ + tc->cc_algo = tcp_cc_algo_get (TCP_CC_NEWRENO); + tc->cc_algo->init (tc); +} + +static int +tcp_rcv_ack (tcp_connection_t * tc, vlib_buffer_t * b, + tcp_header_t * th, u32 * next, u32 * error) +{ + u32 new_snd_wnd; + + /* If the ACK acks something not yet sent (SEG.ACK > SND.NXT) then send an + * ACK, drop the segment, and return */ + if (seq_gt (vnet_buffer (b)->tcp.ack_number, tc->snd_nxt)) + { + tcp_make_ack (tc, b); + *next = tcp_next_output (tc->c_is_ip4); + *error = TCP_ERROR_ACK_INVALID; + return -1; + } + + /* If old ACK, discard */ + if (seq_lt (vnet_buffer (b)->tcp.ack_number, tc->snd_una)) + { + *error = TCP_ERROR_ACK_OLD; + return -1; + } + + if (tcp_opts_sack_permitted (&tc->opt)) + tcp_rcv_sacks (tc, vnet_buffer (b)->tcp.ack_number); + + new_snd_wnd = clib_net_to_host_u32 (th->window) << tc->snd_wscale; + + if (tcp_ack_is_dupack (tc, b, new_snd_wnd)) + { + tcp_cc_rcv_dupack (tc, vnet_buffer (b)->tcp.ack_number); + *error = TCP_ERROR_ACK_DUP; + return -1; + } + + /* Valid ACK */ + tc->bytes_acked = vnet_buffer (b)->tcp.ack_number - tc->snd_una; + tc->snd_una = vnet_buffer (b)->tcp.ack_number; + + /* Dequeue ACKed packet and update RTT */ + tcp_dequeue_acked (tc, vnet_buffer (b)->tcp.ack_number); + + tcp_update_snd_wnd (tc, vnet_buffer (b)->tcp.seq_number, + vnet_buffer (b)->tcp.ack_number, new_snd_wnd); + + /* Updates congestion control (slow start/congestion avoidance) */ + tcp_cc_rcv_ack (tc); + + /* If everything has been acked, stop retransmit timer + * otherwise update */ + if (tc->snd_una == tc->snd_una_max) + tcp_timer_reset (tc, TCP_TIMER_RETRANSMIT); + else + tcp_timer_update (tc, TCP_TIMER_RETRANSMIT, tc->rto); + + return 0; +} + +/** + * Build SACK list as per RFC2018. + * + * Makes sure the first block contains the segment that generated the current + * ACK and the following ones are the ones most recently reported in SACK + * blocks. + * + * @param tc TCP connection for which the SACK list is updated + * @param start Start sequence number of the newest SACK block + * @param end End sequence of the newest SACK block + */ +static void +tcp_update_sack_list (tcp_connection_t * tc, u32 start, u32 end) +{ + sack_block_t *new_list = 0, block; + u32 n_elts; + int i; + u8 new_head = 0; + + /* If the first segment is ooo add it to the list. Last write might've moved + * rcv_nxt over the first segment. */ + if (seq_lt (tc->rcv_nxt, start)) + { + block.start = start; + block.end = end; + vec_add1 (new_list, block); + new_head = 1; + } + + /* Find the blocks still worth keeping. */ + for (i = 0; i < vec_len (tc->snd_sacks); i++) + { + /* Discard if: + * 1) rcv_nxt advanced beyond current block OR + * 2) Segment overlapped by the first segment, i.e., it has been merged + * into it.*/ + if (seq_leq (tc->snd_sacks[i].start, tc->rcv_nxt) + || seq_leq (tc->snd_sacks[i].start, end)) + continue; + + /* Save subsequent segments to new SACK list. */ + n_elts = clib_min (vec_len (tc->snd_sacks) - i, + TCP_MAX_SACK_BLOCKS - new_head); + vec_insert_elts (new_list, &tc->snd_sacks[i], n_elts, new_head); + break; + } + + /* Replace old vector with new one */ + vec_free (tc->snd_sacks); + tc->snd_sacks = new_list; +} + +/** Enqueue data for delivery to application */ +always_inline u32 +tcp_session_enqueue_data (tcp_connection_t * tc, vlib_buffer_t * b, + u16 data_len) +{ + int written; + + /* Pure ACK. Update rcv_nxt and be done. */ + if (PREDICT_FALSE (data_len == 0)) + { + tc->rcv_nxt = vnet_buffer (b)->tcp.seq_end; + return TCP_ERROR_PURE_ACK; + } + + written = stream_session_enqueue_data (&tc->connection, + vlib_buffer_get_current (b), + data_len, 1 /* queue event */ ); + + /* Update rcv_nxt */ + if (PREDICT_TRUE (written == data_len)) + { + tc->rcv_nxt = vnet_buffer (b)->tcp.seq_end; + } + /* If more data written than expected, account for out-of-order bytes. */ + else if (written > data_len) + { + tc->rcv_nxt = vnet_buffer (b)->tcp.seq_end + written - data_len; + + /* Send ACK confirming the update */ + tc->flags |= TCP_CONN_SNDACK; + + /* Update SACK list if need be */ + if (tcp_opts_sack_permitted (&tc->opt)) + { + /* Remove SACK blocks that have been delivered */ + tcp_update_sack_list (tc, tc->rcv_nxt, tc->rcv_nxt); + } + } + else + { + ASSERT (0); + return TCP_ERROR_FIFO_FULL; + } + + return TCP_ERROR_ENQUEUED; +} + +/** Enqueue out-of-order data */ +always_inline u32 +tcp_session_enqueue_ooo (tcp_connection_t * tc, vlib_buffer_t * b, + u16 data_len) +{ + stream_session_t *s0; + u32 offset, seq; + + s0 = stream_session_get (tc->c_s_index, tc->c_thread_index); + seq = vnet_buffer (b)->tcp.seq_number; + offset = seq - tc->rcv_nxt; + + if (svm_fifo_enqueue_with_offset (s0->server_rx_fifo, s0->pid, offset, + data_len, vlib_buffer_get_current (b))) + return TCP_ERROR_FIFO_FULL; + + /* Update SACK list if in use */ + if (tcp_opts_sack_permitted (&tc->opt)) + { + ooo_segment_t *newest; + u32 start, end; + + /* Get the newest segment from the fifo */ + newest = svm_fifo_newest_ooo_segment (s0->server_rx_fifo); + start = tc->rcv_nxt + ooo_segment_offset (s0->server_rx_fifo, newest); + end = tc->rcv_nxt + ooo_segment_end_offset (s0->server_rx_fifo, newest); + + tcp_update_sack_list (tc, start, end); + } + + return TCP_ERROR_ENQUEUED; +} + +/** + * Check if ACK could be delayed. DELACK timer is set only after frame is + * processed so this can return true for a full bursts of packets. + */ +always_inline int +tcp_can_delack (tcp_connection_t * tc) +{ + /* If there's no DELACK timer set and the last window sent wasn't 0 we + * can safely delay. */ + if (!tcp_timer_is_active (tc, TCP_TIMER_DELACK) + && (tc->flags & TCP_CONN_SENT_RCV_WND0) == 0 + && (tc->flags & TCP_CONN_SNDACK) == 0) + return 1; + + return 0; +} + +static int +tcp_segment_rcv (tcp_main_t * tm, tcp_connection_t * tc, vlib_buffer_t * b, + u16 n_data_bytes, u32 * next0) +{ + u32 error = 0; + + /* Handle out-of-order data */ + if (PREDICT_FALSE (vnet_buffer (b)->tcp.seq_number != tc->rcv_nxt)) + { + error = tcp_session_enqueue_ooo (tc, b, n_data_bytes); + + /* Don't send more than 3 dupacks per burst + * XXX decide if this is good */ + if (tc->snt_dupacks < 3) + { + /* RFC2581: Send DUPACK for fast retransmit */ + tcp_make_ack (tc, b); + *next0 = tcp_next_output (tc->c_is_ip4); + + /* Mark as DUPACK. We may filter these in output if + * the burst fills the holes. */ + vnet_buffer (b)->tcp.flags = TCP_BUF_FLAG_DUPACK; + + tc->snt_dupacks++; + } + + goto done; + } + + /* In order data, enqueue. Fifo figures out by itself if any out-of-order + * segments can be enqueued after fifo tail offset changes. */ + error = tcp_session_enqueue_data (tc, b, n_data_bytes); + + /* Check if ACK can be delayed */ + if (tcp_can_delack (tc)) + { + /* Nothing to do for pure ACKs */ + if (n_data_bytes == 0) + goto done; + + /* If connection has not been previously marked for delay ack + * add it to the list and flag it */ + if (!tc->flags & TCP_CONN_DELACK) + { + vec_add1 (tm->delack_connections[tc->c_thread_index], + tc->c_c_index); + tc->flags |= TCP_CONN_DELACK; + } + } + else + { + /* Check if a packet has already been enqueued to output for burst. + * If yes, then drop this one, otherwise, let it pass through to + * output */ + if ((tc->flags & TCP_CONN_BURSTACK) == 0) + { + *next0 = tcp_next_output (tc->c_is_ip4); + tcp_make_ack (tc, b); + error = TCP_ERROR_ENQUEUED; + + /* TODO: maybe add counter to ensure N acks will be sent/burst */ + tc->flags |= TCP_CONN_BURSTACK; + } + } + +done: + return error; +} + +void +delack_timers_init (tcp_main_t * tm, u32 thread_index) +{ + tcp_connection_t *tc; + u32 i, *conns; + tw_timer_wheel_16t_2w_512sl_t *tw; + + tw = &tm->timer_wheels[thread_index]; + conns = tm->delack_connections[thread_index]; + for (i = 0; i < vec_len (conns); i++) + { + tc = pool_elt_at_index (tm->connections[thread_index], conns[i]); + ASSERT (0 != tc); + + tc->timers[TCP_TIMER_DELACK] + = tw_timer_start_16t_2w_512sl (tw, conns[i], + TCP_TIMER_DELACK, TCP_DELACK_TIME); + } + vec_reset_length (tm->delack_connections[thread_index]); +} + +always_inline uword +tcp46_established_inline (vlib_main_t * vm, vlib_node_runtime_t * node, + vlib_frame_t * from_frame, int is_ip4) +{ + u32 n_left_from, next_index, *from, *to_next; + u32 my_thread_index = vm->cpu_index, errors = 0; + tcp_main_t *tm = vnet_get_tcp_main (); + + from = vlib_frame_vector_args (from_frame); + n_left_from = from_frame->n_vectors; + + next_index = node->cached_next_index; + + while (n_left_from > 0) + { + u32 n_left_to_next; + + vlib_get_next_frame (vm, node, next_index, to_next, n_left_to_next); + + while (n_left_from > 0 && n_left_to_next > 0) + { + u32 bi0; + vlib_buffer_t *b0; + tcp_header_t *th0 = 0; + tcp_connection_t *tc0; + ip4_header_t *ip40; + ip6_header_t *ip60; + u32 n_advance_bytes0, n_data_bytes0; + u32 next0 = TCP_ESTABLISHED_NEXT_DROP, error0 = TCP_ERROR_ENQUEUED; + + bi0 = from[0]; + to_next[0] = bi0; + from += 1; + to_next += 1; + n_left_from -= 1; + n_left_to_next -= 1; + + b0 = vlib_get_buffer (vm, bi0); + tc0 = tcp_connection_get (vnet_buffer (b0)->tcp.connection_index, + my_thread_index); + + /* Checksum computed by ipx_local no need to compute again */ + + if (is_ip4) + { + ip40 = vlib_buffer_get_current (b0); + th0 = ip4_next_header (ip40); + n_advance_bytes0 = (ip4_header_bytes (ip40) + + tcp_header_bytes (th0)); + n_data_bytes0 = clib_net_to_host_u16 (ip40->length) + - n_advance_bytes0; + } + else + { + ip60 = vlib_buffer_get_current (b0); + th0 = ip6_next_header (ip60); + n_advance_bytes0 = tcp_header_bytes (th0); + n_data_bytes0 = clib_net_to_host_u16 (ip60->payload_length) + - n_advance_bytes0; + n_advance_bytes0 += sizeof (ip60[0]); + } + + /* SYNs, FINs and data consume sequence numbers */ + vnet_buffer (b0)->tcp.seq_end = vnet_buffer (b0)->tcp.seq_number + + tcp_is_syn (th0) + tcp_is_fin (th0) + n_data_bytes0; + + /* TODO header prediction fast path */ + + /* 1-4: check SEQ, RST, SYN */ + if (PREDICT_FALSE (tcp_segment_validate (vm, tc0, b0, th0, &next0))) + { + error0 = TCP_ERROR_SEGMENT_INVALID; + goto drop; + } + + /* 5: check the ACK field */ + if (tcp_rcv_ack (tc0, b0, th0, &next0, &error0)) + { + goto drop; + } + + /* 6: check the URG bit TODO */ + + /* 7: process the segment text */ + vlib_buffer_advance (b0, n_advance_bytes0); + error0 = tcp_segment_rcv (tm, tc0, b0, n_data_bytes0, &next0); + + /* 8: check the FIN bit */ + if (tcp_fin (th0)) + { + /* Send ACK and enter CLOSE-WAIT */ + tcp_make_ack (tc0, b0); + tcp_connection_force_ack (tc0, b0); + next0 = tcp_next_output (tc0->c_is_ip4); + tc0->state = TCP_STATE_CLOSE_WAIT; + stream_session_disconnect_notify (&tc0->connection); + } + + drop: + b0->error = node->errors[error0]; + if (PREDICT_FALSE (b0->flags & VLIB_BUFFER_IS_TRACED)) + { + + } + + vlib_validate_buffer_enqueue_x1 (vm, node, next_index, to_next, + n_left_to_next, bi0, next0); + } + + vlib_put_next_frame (vm, node, next_index, n_left_to_next); + } + + errors = session_manager_flush_enqueue_events (my_thread_index); + if (errors) + { + if (is_ip4) + vlib_node_increment_counter (vm, tcp4_established_node.index, + TCP_ERROR_EVENT_FIFO_FULL, errors); + else + vlib_node_increment_counter (vm, tcp6_established_node.index, + TCP_ERROR_EVENT_FIFO_FULL, errors); + } + + delack_timers_init (tm, my_thread_index); + + return from_frame->n_vectors; +} + +static uword +tcp4_established (vlib_main_t * vm, vlib_node_runtime_t * node, + vlib_frame_t * from_frame) +{ + return tcp46_established_inline (vm, node, from_frame, 1 /* is_ip4 */ ); +} + +static uword +tcp6_established (vlib_main_t * vm, vlib_node_runtime_t * node, + vlib_frame_t * from_frame) +{ + return tcp46_established_inline (vm, node, from_frame, 0 /* is_ip4 */ ); +} + +/* *INDENT-OFF* */ +VLIB_REGISTER_NODE (tcp4_established_node) = +{ + .function = tcp4_established, + .name = "tcp4-established", + /* Takes a vector of packets. */ + .vector_size = sizeof (u32), + .n_errors = TCP_N_ERROR,.error_strings = tcp_error_strings, + .n_next_nodes = TCP_ESTABLISHED_N_NEXT, + .next_nodes = + { +#define _(s,n) [TCP_ESTABLISHED_NEXT_##s] = n, + foreach_tcp_state_next +#undef _ + }, +}; +/* *INDENT-ON* */ + +VLIB_NODE_FUNCTION_MULTIARCH (tcp4_established_node, tcp4_established); + +/* *INDENT-OFF* */ +VLIB_REGISTER_NODE (tcp6_established_node) = +{ + .function = tcp6_established, + .name = "tcp6-established", + /* Takes a vector of packets. */ + .vector_size = sizeof (u32), + .n_errors = TCP_N_ERROR, + .error_strings = tcp_error_strings, + .n_next_nodes = TCP_ESTABLISHED_N_NEXT, + .next_nodes = + { +#define _(s,n) [TCP_ESTABLISHED_NEXT_##s] = n, + foreach_tcp_state_next +#undef _ + }, +}; +/* *INDENT-ON* */ + + +VLIB_NODE_FUNCTION_MULTIARCH (tcp6_established_node, tcp6_established); + +vlib_node_registration_t tcp4_syn_sent_node; +vlib_node_registration_t tcp6_syn_sent_node; + +always_inline uword +tcp46_syn_sent_inline (vlib_main_t * vm, vlib_node_runtime_t * node, + vlib_frame_t * from_frame, int is_ip4) +{ + tcp_main_t *tm = vnet_get_tcp_main (); + u32 n_left_from, next_index, *from, *to_next; + u32 my_thread_index = vm->cpu_index, errors = 0; + u8 sst = is_ip4 ? SESSION_TYPE_IP4_TCP : SESSION_TYPE_IP6_TCP; + + from = vlib_frame_vector_args (from_frame); + n_left_from = from_frame->n_vectors; + + next_index = node->cached_next_index; + + while (n_left_from > 0) + { + u32 n_left_to_next; + + vlib_get_next_frame (vm, node, next_index, to_next, n_left_to_next); + + while (n_left_from > 0 && n_left_to_next > 0) + { + u32 bi0, ack0, seq0; + vlib_buffer_t *b0; + tcp_header_t *tcp0 = 0; + tcp_connection_t *tc0; + ip4_header_t *ip40; + ip6_header_t *ip60; + u32 n_advance_bytes0, n_data_bytes0; + tcp_connection_t *new_tc0; + u32 next0 = TCP_SYN_SENT_NEXT_DROP, error0 = TCP_ERROR_ENQUEUED; + + bi0 = from[0]; + to_next[0] = bi0; + from += 1; + to_next += 1; + n_left_from -= 1; + n_left_to_next -= 1; + + b0 = vlib_get_buffer (vm, bi0); + tc0 = + tcp_half_open_connection_get (vnet_buffer (b0)-> + tcp.connection_index); + + ack0 = vnet_buffer (b0)->tcp.ack_number; + seq0 = vnet_buffer (b0)->tcp.seq_number; + + /* Checksum computed by ipx_local no need to compute again */ + + if (is_ip4) + { + ip40 = vlib_buffer_get_current (b0); + tcp0 = ip4_next_header (ip40); + n_advance_bytes0 = (ip4_header_bytes (ip40) + + tcp_header_bytes (tcp0)); + n_data_bytes0 = clib_net_to_host_u16 (ip40->length) + - n_advance_bytes0; + } + else + { + ip60 = vlib_buffer_get_current (b0); + tcp0 = ip6_next_header (ip60); + n_advance_bytes0 = tcp_header_bytes (tcp0); + n_data_bytes0 = clib_net_to_host_u16 (ip60->payload_length) + - n_advance_bytes0; + n_advance_bytes0 += sizeof (ip60[0]); + } + + if (PREDICT_FALSE + (!tcp_ack (tcp0) && !tcp_rst (tcp0) && !tcp_syn (tcp0))) + goto drop; + + /* SYNs, FINs and data consume sequence numbers */ + vnet_buffer (b0)->tcp.seq_end = seq0 + tcp_is_syn (tcp0) + + tcp_is_fin (tcp0) + n_data_bytes0; + + /* + * 1. check the ACK bit + */ + + /* + * If the ACK bit is set + * If SEG.ACK =< ISS, or SEG.ACK > SND.NXT, send a reset (unless + * the RST bit is set, if so drop the segment and return) + * + * and discard the segment. Return. + * If SND.UNA =< SEG.ACK =< SND.NXT then the ACK is acceptable. + */ + if (tcp_ack (tcp0)) + { + if (ack0 <= tc0->iss || ack0 > tc0->snd_nxt) + { + if (!tcp_rst (tcp0)) + tcp_send_reset (b0, is_ip4); + + goto drop; + } + + /* Make sure ACK is valid */ + if (tc0->snd_una > ack0) + goto drop; + } + + /* + * 2. check the RST bit + */ + + if (tcp_rst (tcp0)) + { + /* If ACK is acceptable, signal client that peer is not + * willing to accept connection and drop connection*/ + if (tcp_ack (tcp0)) + { + stream_session_connect_notify (&tc0->connection, sst, + 1 /* fail */ ); + tcp_connection_cleanup (tc0); + } + goto drop; + } + + /* + * 3. check the security and precedence (skipped) + */ + + /* + * 4. check the SYN bit + */ + + /* No SYN flag. Drop. */ + if (!tcp_syn (tcp0)) + goto drop; + + /* Stop connection establishment and retransmit timers */ + tcp_timer_reset (tc0, TCP_TIMER_ESTABLISH); + tcp_timer_reset (tc0, TCP_TIMER_RETRANSMIT_SYN); + + /* Valid SYN or SYN-ACK. Move connection from half-open pool to + * current thread pool. */ + pool_get (tm->connections[my_thread_index], new_tc0); + clib_memcpy (new_tc0, tc0, sizeof (*new_tc0)); + + new_tc0->c_thread_index = my_thread_index; + + /* Cleanup half-open connection XXX lock */ + pool_put (tm->half_open_connections, tc0); + + new_tc0->rcv_nxt = vnet_buffer (b0)->tcp.seq_end; + new_tc0->irs = seq0; + + /* Parse options */ + tcp_options_parse (tcp0, &new_tc0->opt); + tcp_connection_init_vars (new_tc0); + + if (tcp_opts_tstamp (&new_tc0->opt)) + { + new_tc0->tsval_recent = new_tc0->opt.tsval; + new_tc0->tsval_recent_age = tcp_time_now (); + } + + if (tcp_opts_wscale (&new_tc0->opt)) + new_tc0->snd_wscale = new_tc0->opt.wscale; + + new_tc0->snd_wnd = clib_net_to_host_u32 (tcp0->window) + << new_tc0->snd_wscale; + new_tc0->snd_wl1 = seq0; + new_tc0->snd_wl2 = ack0; + + /* SYN-ACK: See if we can switch to ESTABLISHED state */ + if (tcp_ack (tcp0)) + { + /* Our SYN is ACKed: we have iss < ack = snd_una */ + + /* TODO Dequeue acknowledged segments if we support Fast Open */ + new_tc0->snd_una = ack0; + new_tc0->state = TCP_STATE_ESTABLISHED; + + /* Notify app that we have connection */ + stream_session_connect_notify (&new_tc0->connection, sst, 0); + + /* Make sure after data segment processing ACK is sent */ + new_tc0->flags |= TCP_CONN_SNDACK; + } + /* SYN: Simultaneous open. Change state to SYN-RCVD and send SYN-ACK */ + else + { + new_tc0->state = TCP_STATE_SYN_RCVD; + + /* Notify app that we have connection XXX */ + stream_session_connect_notify (&new_tc0->connection, sst, 0); + + tcp_make_synack (new_tc0, b0); + next0 = tcp_next_output (is_ip4); + + goto drop; + } + + /* Read data, if any */ + if (n_data_bytes0) + { + error0 = + tcp_segment_rcv (tm, new_tc0, b0, n_data_bytes0, &next0); + if (error0 == TCP_ERROR_PURE_ACK) + error0 = TCP_ERROR_SYN_ACKS_RCVD; + } + else + { + tcp_make_ack (new_tc0, b0); + next0 = tcp_next_output (new_tc0->c_is_ip4); + } + + drop: + + b0->error = error0 ? node->errors[error0] : 0; + if (PREDICT_FALSE (b0->flags & VLIB_BUFFER_IS_TRACED)) + { + + } + + vlib_validate_buffer_enqueue_x1 (vm, node, next_index, to_next, + n_left_to_next, bi0, next0); + } + + vlib_put_next_frame (vm, node, next_index, n_left_to_next); + } + + errors = session_manager_flush_enqueue_events (my_thread_index); + if (errors) + { + if (is_ip4) + vlib_node_increment_counter (vm, tcp4_established_node.index, + TCP_ERROR_EVENT_FIFO_FULL, errors); + else + vlib_node_increment_counter (vm, tcp6_established_node.index, + TCP_ERROR_EVENT_FIFO_FULL, errors); + } + + return from_frame->n_vectors; +} + +static uword +tcp4_syn_sent (vlib_main_t * vm, vlib_node_runtime_t * node, + vlib_frame_t * from_frame) +{ + return tcp46_syn_sent_inline (vm, node, from_frame, 1 /* is_ip4 */ ); +} + +static uword +tcp6_syn_sent_rcv (vlib_main_t * vm, vlib_node_runtime_t * node, + vlib_frame_t * from_frame) +{ + return tcp46_syn_sent_inline (vm, node, from_frame, 0 /* is_ip4 */ ); +} + +/* *INDENT-OFF* */ +VLIB_REGISTER_NODE (tcp4_syn_sent_node) = +{ + .function = tcp4_syn_sent, + .name = "tcp4-syn-sent", + /* Takes a vector of packets. */ + .vector_size = sizeof (u32), + .n_errors = TCP_N_ERROR, + .error_strings = tcp_error_strings, + .n_next_nodes = TCP_SYN_SENT_N_NEXT, + .next_nodes = + { +#define _(s,n) [TCP_SYN_SENT_NEXT_##s] = n, + foreach_tcp_state_next +#undef _ + }, +}; +/* *INDENT-ON* */ + +VLIB_NODE_FUNCTION_MULTIARCH (tcp4_syn_sent_node, tcp4_syn_sent); + +/* *INDENT-OFF* */ +VLIB_REGISTER_NODE (tcp6_syn_sent_node) = +{ + .function = tcp6_syn_sent_rcv, + .name = "tcp6-syn-sent", + /* Takes a vector of packets. */ + .vector_size = sizeof (u32), + .n_errors = TCP_N_ERROR, + .error_strings = tcp_error_strings, + .n_next_nodes = TCP_SYN_SENT_N_NEXT, + .next_nodes = + { +#define _(s,n) [TCP_SYN_SENT_NEXT_##s] = n, + foreach_tcp_state_next +#undef _ + } +,}; +/* *INDENT-ON* */ + +VLIB_NODE_FUNCTION_MULTIARCH (tcp6_syn_sent_node, tcp6_syn_sent_rcv); +/** + * Handles reception for all states except LISTEN, SYN-SEND and ESTABLISHED + * as per RFC793 p. 64 + */ +always_inline uword +tcp46_rcv_process_inline (vlib_main_t * vm, vlib_node_runtime_t * node, + vlib_frame_t * from_frame, int is_ip4) +{ + tcp_main_t *tm = vnet_get_tcp_main (); + u32 n_left_from, next_index, *from, *to_next; + u32 my_thread_index = vm->cpu_index, errors = 0; + + from = vlib_frame_vector_args (from_frame); + n_left_from = from_frame->n_vectors; + + next_index = node->cached_next_index; + + while (n_left_from > 0) + { + u32 n_left_to_next; + + vlib_get_next_frame (vm, node, next_index, to_next, n_left_to_next); + + while (n_left_from > 0 && n_left_to_next > 0) + { + u32 bi0; + vlib_buffer_t *b0; + tcp_header_t *tcp0 = 0; + tcp_connection_t *tc0; + ip4_header_t *ip40; + ip6_header_t *ip60; + u32 n_advance_bytes0, n_data_bytes0; + u32 next0 = TCP_RCV_PROCESS_NEXT_DROP, error0 = TCP_ERROR_ENQUEUED; + + bi0 = from[0]; + to_next[0] = bi0; + from += 1; + to_next += 1; + n_left_from -= 1; + n_left_to_next -= 1; + + b0 = vlib_get_buffer (vm, bi0); + tc0 = tcp_connection_get (vnet_buffer (b0)->tcp.connection_index, + my_thread_index); + + /* Checksum computed by ipx_local no need to compute again */ + + if (is_ip4) + { + ip40 = vlib_buffer_get_current (b0); + tcp0 = ip4_next_header (ip40); + n_advance_bytes0 = (ip4_header_bytes (ip40) + + tcp_header_bytes (tcp0)); + n_data_bytes0 = clib_net_to_host_u16 (ip40->length) + - n_advance_bytes0; + } + else + { + ip60 = vlib_buffer_get_current (b0); + tcp0 = ip6_next_header (ip60); + n_advance_bytes0 = tcp_header_bytes (tcp0); + n_data_bytes0 = clib_net_to_host_u16 (ip60->payload_length) + - n_advance_bytes0; + n_advance_bytes0 += sizeof (ip60[0]); + } + + /* SYNs, FINs and data consume sequence numbers */ + vnet_buffer (b0)->tcp.seq_end = vnet_buffer (b0)->tcp.seq_number + + tcp_is_syn (tcp0) + tcp_is_fin (tcp0) + n_data_bytes0; + + /* + * Special treatment for CLOSED + */ + switch (tc0->state) + { + case TCP_STATE_CLOSED: + goto drop; + break; + } + + /* + * For all other states (except LISTEN) + */ + + /* 1-4: check SEQ, RST, SYN */ + if (PREDICT_FALSE + (tcp_segment_validate (vm, tc0, b0, tcp0, &next0))) + { + error0 = TCP_ERROR_SEGMENT_INVALID; + goto drop; + } + + /* 5: check the ACK field */ + switch (tc0->state) + { + case TCP_STATE_SYN_RCVD: + /* + * If the segment acknowledgment is not acceptable, form a + * reset segment, + * + * and send it. + */ + if (!tcp_rcv_ack_is_acceptable (tc0, b0)) + { + tcp_send_reset (b0, is_ip4); + goto drop; + } + /* Switch state to ESTABLISHED */ + tc0->state = TCP_STATE_ESTABLISHED; + + /* Initialize session variables */ + tc0->snd_una = vnet_buffer (b0)->tcp.ack_number; + tc0->snd_wnd = clib_net_to_host_u32 (tcp0->window) + << tc0->opt.wscale; + tc0->snd_wl1 = vnet_buffer (b0)->tcp.seq_number; + tc0->snd_wl2 = vnet_buffer (b0)->tcp.ack_number; + + /* Shoulder tap the server */ + stream_session_accept_notify (&tc0->connection); + + tcp_timer_reset (tc0, TCP_TIMER_RETRANSMIT_SYN); + break; + case TCP_STATE_ESTABLISHED: + /* We can get packets in established state here because they + * were enqueued before state change */ + if (tcp_rcv_ack (tc0, b0, tcp0, &next0, &error0)) + goto drop; + + break; + case TCP_STATE_FIN_WAIT_1: + /* In addition to the processing for the ESTABLISHED state, if + * our FIN is now acknowledged then enter FIN-WAIT-2 and + * continue processing in that state. */ + if (tcp_rcv_ack (tc0, b0, tcp0, &next0, &error0)) + goto drop; + tc0->state = TCP_STATE_FIN_WAIT_2; + /* Stop all timers, 2MSL will be set lower */ + tcp_connection_timers_reset (tc0); + break; + case TCP_STATE_FIN_WAIT_2: + /* In addition to the processing for the ESTABLISHED state, if + * the retransmission queue is empty, the user's CLOSE can be + * acknowledged ("ok") but do not delete the TCB. */ + if (tcp_rcv_ack (tc0, b0, tcp0, &next0, &error0)) + goto drop; + /* check if rtx queue is empty and ack CLOSE TODO */ + break; + case TCP_STATE_CLOSE_WAIT: + /* Do the same processing as for the ESTABLISHED state. */ + if (tcp_rcv_ack (tc0, b0, tcp0, &next0, &error0)) + goto drop; + break; + case TCP_STATE_CLOSING: + /* In addition to the processing for the ESTABLISHED state, if + * the ACK acknowledges our FIN then enter the TIME-WAIT state, + * otherwise ignore the segment. */ + if (tcp_rcv_ack (tc0, b0, tcp0, &next0, &error0)) + goto drop; + + /* XXX test that send queue empty */ + tc0->state = TCP_STATE_TIME_WAIT; + goto drop; + + break; + case TCP_STATE_LAST_ACK: + /* The only thing that can arrive in this state is an + * acknowledgment of our FIN. If our FIN is now acknowledged, + * delete the TCB, enter the CLOSED state, and return. */ + + if (!tcp_rcv_ack_is_acceptable (tc0, b0)) + goto drop; + + tcp_connection_del (tc0); + goto drop; + + break; + case TCP_STATE_TIME_WAIT: + /* The only thing that can arrive in this state is a + * retransmission of the remote FIN. Acknowledge it, and restart + * the 2 MSL timeout. */ + + /* TODO */ + goto drop; + break; + default: + ASSERT (0); + } + + /* 6: check the URG bit TODO */ + + /* 7: process the segment text */ + switch (tc0->state) + { + case TCP_STATE_ESTABLISHED: + case TCP_STATE_FIN_WAIT_1: + case TCP_STATE_FIN_WAIT_2: + error0 = tcp_segment_rcv (tm, tc0, b0, n_data_bytes0, &next0); + break; + case TCP_STATE_CLOSE_WAIT: + case TCP_STATE_CLOSING: + case TCP_STATE_LAST_ACK: + case TCP_STATE_TIME_WAIT: + /* This should not occur, since a FIN has been received from the + * remote side. Ignore the segment text. */ + break; + } + + /* 8: check the FIN bit */ + if (!tcp_fin (tcp0)) + goto drop; + + switch (tc0->state) + { + case TCP_STATE_ESTABLISHED: + case TCP_STATE_SYN_RCVD: + /* Send FIN-ACK notify app and enter CLOSE-WAIT */ + tcp_connection_timers_reset (tc0); + tcp_make_finack (tc0, b0); + next0 = tcp_next_output (tc0->c_is_ip4); + stream_session_disconnect_notify (&tc0->connection); + tc0->state = TCP_STATE_CLOSE_WAIT; + break; + case TCP_STATE_CLOSE_WAIT: + case TCP_STATE_CLOSING: + case TCP_STATE_LAST_ACK: + /* move along .. */ + break; + case TCP_STATE_FIN_WAIT_1: + tc0->state = TCP_STATE_TIME_WAIT; + tcp_connection_timers_reset (tc0); + tcp_timer_set (tc0, TCP_TIMER_2MSL, TCP_2MSL_TIME); + break; + case TCP_STATE_FIN_WAIT_2: + /* Got FIN, send ACK! */ + tc0->state = TCP_STATE_TIME_WAIT; + tcp_timer_set (tc0, TCP_TIMER_2MSL, TCP_2MSL_TIME); + tcp_make_ack (tc0, b0); + next0 = tcp_next_output (is_ip4); + break; + case TCP_STATE_TIME_WAIT: + /* Remain in the TIME-WAIT state. Restart the 2 MSL time-wait + * timeout. + */ + tcp_timer_update (tc0, TCP_TIMER_2MSL, TCP_2MSL_TIME); + break; + } + + b0->error = error0 ? node->errors[error0] : 0; + + drop: + if (PREDICT_FALSE (b0->flags & VLIB_BUFFER_IS_TRACED)) + { + + } + + vlib_validate_buffer_enqueue_x1 (vm, node, next_index, to_next, + n_left_to_next, bi0, next0); + } + + vlib_put_next_frame (vm, node, next_index, n_left_to_next); + } + + errors = session_manager_flush_enqueue_events (my_thread_index); + if (errors) + { + if (is_ip4) + vlib_node_increment_counter (vm, tcp4_established_node.index, + TCP_ERROR_EVENT_FIFO_FULL, errors); + else + vlib_node_increment_counter (vm, tcp6_established_node.index, + TCP_ERROR_EVENT_FIFO_FULL, errors); + } + + return from_frame->n_vectors; +} + +static uword +tcp4_rcv_process (vlib_main_t * vm, vlib_node_runtime_t * node, + vlib_frame_t * from_frame) +{ + return tcp46_rcv_process_inline (vm, node, from_frame, 1 /* is_ip4 */ ); +} + +static uword +tcp6_rcv_process (vlib_main_t * vm, vlib_node_runtime_t * node, + vlib_frame_t * from_frame) +{ + return tcp46_rcv_process_inline (vm, node, from_frame, 0 /* is_ip4 */ ); +} + +/* *INDENT-OFF* */ +VLIB_REGISTER_NODE (tcp4_rcv_process_node) = +{ + .function = tcp4_rcv_process, + .name = "tcp4-rcv-process", + /* Takes a vector of packets. */ + .vector_size = sizeof (u32), + .n_errors = TCP_N_ERROR, + .error_strings = tcp_error_strings, + .n_next_nodes = TCP_RCV_PROCESS_N_NEXT, + .next_nodes = + { +#define _(s,n) [TCP_RCV_PROCESS_NEXT_##s] = n, + foreach_tcp_state_next +#undef _ + }, +}; +/* *INDENT-ON* */ + +VLIB_NODE_FUNCTION_MULTIARCH (tcp4_rcv_process_node, tcp4_rcv_process); + +/* *INDENT-OFF* */ +VLIB_REGISTER_NODE (tcp6_rcv_process_node) = +{ + .function = tcp6_rcv_process, + .name = "tcp6-rcv-process", + /* Takes a vector of packets. */ + .vector_size = sizeof (u32), + .n_errors = TCP_N_ERROR, + .error_strings = tcp_error_strings, + .n_next_nodes = TCP_RCV_PROCESS_N_NEXT, + .next_nodes = + { +#define _(s,n) [TCP_RCV_PROCESS_NEXT_##s] = n, + foreach_tcp_state_next +#undef _ + }, +}; +/* *INDENT-ON* */ + +VLIB_NODE_FUNCTION_MULTIARCH (tcp6_rcv_process_node, tcp6_rcv_process); + +vlib_node_registration_t tcp4_listen_node; +vlib_node_registration_t tcp6_listen_node; + +/** + * LISTEN state processing as per RFC 793 p. 65 + */ +always_inline uword +tcp46_listen_inline (vlib_main_t * vm, vlib_node_runtime_t * node, + vlib_frame_t * from_frame, int is_ip4) +{ + u32 n_left_from, next_index, *from, *to_next; + u32 my_thread_index = vm->cpu_index; + tcp_main_t *tm = vnet_get_tcp_main (); + u8 sst = is_ip4 ? SESSION_TYPE_IP4_TCP : SESSION_TYPE_IP6_TCP; + + from = vlib_frame_vector_args (from_frame); + n_left_from = from_frame->n_vectors; + + next_index = node->cached_next_index; + + while (n_left_from > 0) + { + u32 n_left_to_next; + + vlib_get_next_frame (vm, node, next_index, to_next, n_left_to_next); + + while (n_left_from > 0 && n_left_to_next > 0) + { + u32 bi0; + vlib_buffer_t *b0; + tcp_header_t *th0 = 0; + tcp_connection_t *lc0; + ip4_header_t *ip40; + ip6_header_t *ip60; + tcp_connection_t *child0; + u32 error0 = TCP_ERROR_SYNS_RCVD, next0 = TCP_LISTEN_NEXT_DROP; + + bi0 = from[0]; + to_next[0] = bi0; + from += 1; + to_next += 1; + n_left_from -= 1; + n_left_to_next -= 1; + + b0 = vlib_get_buffer (vm, bi0); + lc0 = tcp_listener_get (vnet_buffer (b0)->tcp.connection_index); + + if (is_ip4) + { + ip40 = vlib_buffer_get_current (b0); + th0 = ip4_next_header (ip40); + } + else + { + ip60 = vlib_buffer_get_current (b0); + th0 = ip6_next_header (ip60); + } + + /* Create child session. For syn-flood protection use filter */ + + /* 1. first check for an RST */ + if (tcp_rst (th0)) + goto drop; + + /* 2. second check for an ACK */ + if (tcp_ack (th0)) + { + tcp_send_reset (b0, is_ip4); + goto drop; + } + + /* 3. check for a SYN (did that already) */ + + /* Create child session and send SYN-ACK */ + pool_get (tm->connections[my_thread_index], child0); + memset (child0, 0, sizeof (*child0)); + + child0->c_c_index = child0 - tm->connections[my_thread_index]; + child0->c_lcl_port = lc0->c_lcl_port; + child0->c_rmt_port = th0->src_port; + child0->c_is_ip4 = is_ip4; + child0->c_thread_index = my_thread_index; + + if (is_ip4) + { + child0->c_lcl_ip4.as_u32 = ip40->dst_address.as_u32; + child0->c_rmt_ip4.as_u32 = ip40->src_address.as_u32; + } + else + { + clib_memcpy (&child0->c_lcl_ip6, &ip60->dst_address, + sizeof (ip6_address_t)); + clib_memcpy (&child0->c_rmt_ip6, &ip60->src_address, + sizeof (ip6_address_t)); + } + + if (stream_session_accept (&child0->connection, lc0->c_s_index, sst, + 0 /* notify */ )) + { + error0 = TCP_ERROR_CREATE_SESSION_FAIL; + goto drop; + } + + tcp_options_parse (th0, &child0->opt); + tcp_connection_init_vars (child0); + + child0->irs = vnet_buffer (b0)->tcp.seq_number; + child0->rcv_nxt = vnet_buffer (b0)->tcp.seq_number + 1; + child0->state = TCP_STATE_SYN_RCVD; + + /* RFC1323: TSval timestamps sent on {SYN} and {SYN,ACK} + * segments are used to initialize PAWS. */ + if (tcp_opts_tstamp (&child0->opt)) + { + child0->tsval_recent = child0->opt.tsval; + child0->tsval_recent_age = tcp_time_now (); + } + + /* Reuse buffer to make syn-ack and send */ + tcp_make_synack (child0, b0); + next0 = tcp_next_output (is_ip4); + + drop: + if (PREDICT_FALSE (b0->flags & VLIB_BUFFER_IS_TRACED)) + { + + } + + b0->error = error0 ? node->errors[error0] : 0; + + vlib_validate_buffer_enqueue_x1 (vm, node, next_index, to_next, + n_left_to_next, bi0, next0); + } + + vlib_put_next_frame (vm, node, next_index, n_left_to_next); + } + return from_frame->n_vectors; +} + +static uword +tcp4_listen (vlib_main_t * vm, vlib_node_runtime_t * node, + vlib_frame_t * from_frame) +{ + return tcp46_listen_inline (vm, node, from_frame, 1 /* is_ip4 */ ); +} + +static uword +tcp6_listen (vlib_main_t * vm, vlib_node_runtime_t * node, + vlib_frame_t * from_frame) +{ + return tcp46_listen_inline (vm, node, from_frame, 0 /* is_ip4 */ ); +} + +/* *INDENT-OFF* */ +VLIB_REGISTER_NODE (tcp4_listen_node) = +{ + .function = tcp4_listen, + .name = "tcp4-listen", + /* Takes a vector of packets. */ + .vector_size = sizeof (u32), + .n_errors = TCP_N_ERROR, + .error_strings = tcp_error_strings, + .n_next_nodes = TCP_LISTEN_N_NEXT, + .next_nodes = + { +#define _(s,n) [TCP_LISTEN_NEXT_##s] = n, + foreach_tcp_state_next +#undef _ + }, +}; +/* *INDENT-ON* */ + +VLIB_NODE_FUNCTION_MULTIARCH (tcp4_listen_node, tcp4_listen); + +/* *INDENT-OFF* */ +VLIB_REGISTER_NODE (tcp6_listen_node) = +{ + .function = tcp6_listen, + .name = "tcp6-listen", + /* Takes a vector of packets. */ + .vector_size = sizeof (u32), + .n_errors = TCP_N_ERROR, + .error_strings = tcp_error_strings, + .n_next_nodes = TCP_LISTEN_N_NEXT, + .next_nodes = + { +#define _(s,n) [TCP_LISTEN_NEXT_##s] = n, + foreach_tcp_state_next +#undef _ + }, +}; +/* *INDENT-ON* */ + +VLIB_NODE_FUNCTION_MULTIARCH (tcp6_listen_node, tcp6_listen); + +vlib_node_registration_t tcp4_input_node; +vlib_node_registration_t tcp6_input_node; + +typedef enum _tcp_input_next +{ + TCP_INPUT_NEXT_DROP, + TCP_INPUT_NEXT_LISTEN, + TCP_INPUT_NEXT_RCV_PROCESS, + TCP_INPUT_NEXT_SYN_SENT, + TCP_INPUT_NEXT_ESTABLISHED, + TCP_INPUT_NEXT_RESET, + TCP_INPUT_N_NEXT +} tcp_input_next_t; + +#define foreach_tcp4_input_next \ + _ (DROP, "error-drop") \ + _ (LISTEN, "tcp4-listen") \ + _ (RCV_PROCESS, "tcp4-rcv-process") \ + _ (SYN_SENT, "tcp4-syn-sent") \ + _ (ESTABLISHED, "tcp4-established") \ + _ (RESET, "tcp4-reset") + +#define foreach_tcp6_input_next \ + _ (DROP, "error-drop") \ + _ (LISTEN, "tcp6-listen") \ + _ (RCV_PROCESS, "tcp6-rcv-process") \ + _ (SYN_SENT, "tcp6-syn-sent") \ + _ (ESTABLISHED, "tcp6-established") \ + _ (RESET, "tcp6-reset") + +typedef struct +{ + u16 src_port; + u16 dst_port; + u8 state; +} tcp_rx_trace_t; + +const char *tcp_fsm_states[] = { +#define _(sym, str) str, + foreach_tcp_fsm_state +#undef _ +}; + +u8 * +format_tcp_state (u8 * s, va_list * args) +{ + tcp_state_t *state = va_arg (*args, tcp_state_t *); + + if (state[0] < TCP_N_STATES) + s = format (s, "%s", tcp_fsm_states[state[0]]); + else + s = format (s, "UNKNOWN"); + + return s; +} + +u8 * +format_tcp_rx_trace (u8 * s, va_list * args) +{ + CLIB_UNUSED (vlib_main_t * vm) = va_arg (*args, vlib_main_t *); + CLIB_UNUSED (vlib_node_t * node) = va_arg (*args, vlib_node_t *); + tcp_rx_trace_t *t = va_arg (*args, tcp_rx_trace_t *); + + s = format (s, "TCP: src-port %d dst-port %U%s\n", + clib_net_to_host_u16 (t->src_port), + clib_net_to_host_u16 (t->dst_port), format_tcp_state, t->state); + + return s; +} + +#define filter_flags (TCP_FLAG_SYN|TCP_FLAG_ACK|TCP_FLAG_RST|TCP_FLAG_FIN) + +always_inline uword +tcp46_input_inline (vlib_main_t * vm, vlib_node_runtime_t * node, + vlib_frame_t * from_frame, int is_ip4) +{ + u32 n_left_from, next_index, *from, *to_next; + u32 my_thread_index = vm->cpu_index; + tcp_main_t *tm = vnet_get_tcp_main (); + session_manager_main_t *ssm = vnet_get_session_manager_main (); + + from = vlib_frame_vector_args (from_frame); + n_left_from = from_frame->n_vectors; + + next_index = node->cached_next_index; + + while (n_left_from > 0) + { + u32 n_left_to_next; + + vlib_get_next_frame (vm, node, next_index, to_next, n_left_to_next); + + while (n_left_from > 0 && n_left_to_next > 0) + { + u32 bi0; + vlib_buffer_t *b0; + tcp_header_t *tcp0 = 0; + tcp_connection_t *tc0; + ip4_header_t *ip40; + ip6_header_t *ip60; + u32 error0 = TCP_ERROR_NO_LISTENER, next0 = TCP_INPUT_NEXT_DROP; + u8 flags0; + + bi0 = from[0]; + to_next[0] = bi0; + from += 1; + to_next += 1; + n_left_from -= 1; + n_left_to_next -= 1; + + b0 = vlib_get_buffer (vm, bi0); + + if (is_ip4) + { + ip40 = vlib_buffer_get_current (b0); + tcp0 = ip4_next_header (ip40); + + /* lookup session */ + tc0 = + (tcp_connection_t *) stream_session_lookup_transport4 (ssm, + &ip40->dst_address, + &ip40->src_address, + tcp0->dst_port, + tcp0->src_port, + SESSION_TYPE_IP4_TCP, + my_thread_index); + } + else + { + ip60 = vlib_buffer_get_current (b0); + tcp0 = ip6_next_header (ip60); + tc0 = + (tcp_connection_t *) stream_session_lookup_transport6 (ssm, + &ip60->src_address, + &ip60->dst_address, + tcp0->src_port, + tcp0->dst_port, + SESSION_TYPE_IP6_TCP, + my_thread_index); + } + + /* Session exists */ + if (PREDICT_TRUE (0 != tc0)) + { + /* Save connection index */ + vnet_buffer (b0)->tcp.connection_index = tc0->c_c_index; + vnet_buffer (b0)->tcp.seq_number = + clib_net_to_host_u32 (tcp0->seq_number); + vnet_buffer (b0)->tcp.ack_number = + clib_net_to_host_u32 (tcp0->ack_number); + + flags0 = tcp0->flags & filter_flags; + next0 = tm->dispatch_table[tc0->state][flags0].next; + error0 = tm->dispatch_table[tc0->state][flags0].error; + + if (PREDICT_FALSE (error0 == TCP_ERROR_DISPATCH)) + { + /* Overload tcp flags to store state */ + vnet_buffer (b0)->tcp.flags = tc0->state; + } + } + else + { + /* Send reset */ + next0 = TCP_INPUT_NEXT_RESET; + error0 = TCP_ERROR_NO_LISTENER; + vnet_buffer (b0)->tcp.flags = 0; + } + + b0->error = error0 ? node->errors[error0] : 0; + + if (PREDICT_FALSE (b0->flags & VLIB_BUFFER_IS_TRACED)) + { + + } + + vlib_validate_buffer_enqueue_x1 (vm, node, next_index, to_next, + n_left_to_next, bi0, next0); + } + + vlib_put_next_frame (vm, node, next_index, n_left_to_next); + } + + return from_frame->n_vectors; +} + +static uword +tcp4_input (vlib_main_t * vm, vlib_node_runtime_t * node, + vlib_frame_t * from_frame) +{ + return tcp46_input_inline (vm, node, from_frame, 1 /* is_ip4 */ ); +} + +static uword +tcp6_input (vlib_main_t * vm, vlib_node_runtime_t * node, + vlib_frame_t * from_frame) +{ + return tcp46_input_inline (vm, node, from_frame, 0 /* is_ip4 */ ); +} + +/* *INDENT-OFF* */ +VLIB_REGISTER_NODE (tcp4_input_node) = +{ + .function = tcp4_input, + .name = "tcp4-input", + /* Takes a vector of packets. */ + .vector_size = sizeof (u32), + .n_errors = TCP_N_ERROR, + .error_strings = tcp_error_strings, + .n_next_nodes = TCP_INPUT_N_NEXT, + .next_nodes = + { +#define _(s,n) [TCP_INPUT_NEXT_##s] = n, + foreach_tcp4_input_next +#undef _ + }, + .format_buffer = format_tcp_header, + .format_trace = format_tcp_rx_trace, +}; +/* *INDENT-ON* */ + +VLIB_NODE_FUNCTION_MULTIARCH (tcp4_input_node, tcp4_input); + +/* *INDENT-OFF* */ +VLIB_REGISTER_NODE (tcp6_input_node) = +{ + .function = tcp6_input, + .name = "tcp6-input", + /* Takes a vector of packets. */ + .vector_size = sizeof (u32), + .n_errors = TCP_N_ERROR, + .error_strings = tcp_error_strings, + .n_next_nodes = TCP_INPUT_N_NEXT, + .next_nodes = + { +#define _(s,n) [TCP_INPUT_NEXT_##s] = n, + foreach_tcp6_input_next +#undef _ + }, + .format_buffer = format_tcp_header, + .format_trace = format_tcp_rx_trace, +}; +/* *INDENT-ON* */ + +VLIB_NODE_FUNCTION_MULTIARCH (tcp6_input_node, tcp6_input); +void +tcp_update_time (f64 now, u32 thread_index) +{ + tcp_main_t *tm = vnet_get_tcp_main (); + tw_timer_expire_timers_16t_2w_512sl (&tm->timer_wheels[thread_index], now); +} + +static void +tcp_dispatch_table_init (tcp_main_t * tm) +{ + int i, j; + for (i = 0; i < ARRAY_LEN (tm->dispatch_table); i++) + for (j = 0; j < ARRAY_LEN (tm->dispatch_table[i]); j++) + { + tm->dispatch_table[i][j].next = TCP_INPUT_NEXT_DROP; + tm->dispatch_table[i][j].error = TCP_ERROR_DISPATCH; + } + +#define _(t,f,n,e) \ +do { \ + tm->dispatch_table[TCP_STATE_##t][f].next = (n); \ + tm->dispatch_table[TCP_STATE_##t][f].error = (e); \ +} while (0) + + /* SYNs for new connections -> tcp-listen. */ + _(LISTEN, TCP_FLAG_SYN, TCP_INPUT_NEXT_LISTEN, TCP_ERROR_NONE); + /* ACK for for a SYN-ACK -> tcp-rcv-process. */ + _(SYN_RCVD, TCP_FLAG_ACK, TCP_INPUT_NEXT_RCV_PROCESS, TCP_ERROR_NONE); + /* SYN-ACK for a SYN */ + _(SYN_SENT, TCP_FLAG_SYN | TCP_FLAG_ACK, TCP_INPUT_NEXT_SYN_SENT, + TCP_ERROR_NONE); + _(SYN_SENT, TCP_FLAG_ACK, TCP_INPUT_NEXT_SYN_SENT, TCP_ERROR_NONE); + _(SYN_SENT, TCP_FLAG_RST, TCP_INPUT_NEXT_SYN_SENT, TCP_ERROR_NONE); + _(SYN_SENT, TCP_FLAG_RST | TCP_FLAG_ACK, TCP_INPUT_NEXT_SYN_SENT, + TCP_ERROR_NONE); + /* ACK for for established connection -> tcp-established. */ + _(ESTABLISHED, TCP_FLAG_ACK, TCP_INPUT_NEXT_ESTABLISHED, TCP_ERROR_NONE); + /* FIN for for established connection -> tcp-established. */ + _(ESTABLISHED, TCP_FLAG_FIN, TCP_INPUT_NEXT_ESTABLISHED, TCP_ERROR_NONE); + _(ESTABLISHED, TCP_FLAG_FIN | TCP_FLAG_ACK, TCP_INPUT_NEXT_ESTABLISHED, + TCP_ERROR_NONE); + /* ACK or FIN-ACK to our FIN */ + _(FIN_WAIT_1, TCP_FLAG_ACK, TCP_INPUT_NEXT_RCV_PROCESS, TCP_ERROR_NONE); + _(FIN_WAIT_1, TCP_FLAG_ACK | TCP_FLAG_FIN, TCP_INPUT_NEXT_RCV_PROCESS, + TCP_ERROR_NONE); + /* FIN in reply to our FIN from the other side */ + _(FIN_WAIT_1, TCP_FLAG_FIN, TCP_INPUT_NEXT_RCV_PROCESS, TCP_ERROR_NONE); + /* FIN confirming that the peer (app) has closed */ + _(FIN_WAIT_2, TCP_FLAG_FIN, TCP_INPUT_NEXT_RCV_PROCESS, TCP_ERROR_NONE); + _(FIN_WAIT_2, TCP_FLAG_FIN | TCP_FLAG_ACK, TCP_INPUT_NEXT_RCV_PROCESS, + TCP_ERROR_NONE); + _(LAST_ACK, TCP_FLAG_ACK, TCP_INPUT_NEXT_RCV_PROCESS, TCP_ERROR_NONE); +#undef _ +} + +clib_error_t * +tcp_input_init (vlib_main_t * vm) +{ + clib_error_t *error = 0; + tcp_main_t *tm = vnet_get_tcp_main (); + + if ((error = vlib_call_init_function (vm, tcp_init))) + return error; + + /* Initialize dispatch table. */ + tcp_dispatch_table_init (tm); + + return error; +} + +VLIB_INIT_FUNCTION (tcp_input_init); + +/* + * fd.io coding-style-patch-verification: ON + * + * Local Variables: + * eval: (c-set-style "gnu") + * End: + */ diff --git a/src/vnet/tcp/tcp_newreno.c b/src/vnet/tcp/tcp_newreno.c new file mode 100644 index 00000000..856dffe4 --- /dev/null +++ b/src/vnet/tcp/tcp_newreno.c @@ -0,0 +1,93 @@ +/* + * Copyright (c) 2017 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +#include + +void +newreno_congestion (tcp_connection_t * tc) +{ + tc->prev_ssthresh = tc->ssthresh; + tc->ssthresh = clib_max (tcp_flight_size (tc) / 2, 2 * tc->snd_mss); +} + +void +newreno_recovered (tcp_connection_t * tc) +{ + tc->cwnd = tc->ssthresh; +} + +void +newreno_rcv_ack (tcp_connection_t * tc) +{ + if (tcp_in_slowstart (tc)) + { + tc->cwnd += clib_min (tc->snd_mss, tc->bytes_acked); + } + else + { + /* Round up to 1 if needed */ + tc->cwnd += clib_max (tc->snd_mss * tc->snd_mss / tc->cwnd, 1); + } +} + +void +newreno_rcv_cong_ack (tcp_connection_t * tc, tcp_cc_ack_t ack_type) +{ + if (ack_type == TCP_CC_DUPACK) + { + tc->cwnd += tc->snd_mss; + } + else if (ack_type == TCP_CC_PARTIALACK) + { + tc->cwnd -= tc->bytes_acked; + if (tc->bytes_acked > tc->snd_mss) + tc->bytes_acked += tc->snd_mss; + } +} + +void +newreno_conn_init (tcp_connection_t * tc) +{ + tc->ssthresh = tc->snd_wnd; + tc->cwnd = tcp_initial_cwnd (tc); +} + +const static tcp_cc_algorithm_t tcp_newreno = { + .congestion = newreno_congestion, + .recovered = newreno_recovered, + .rcv_ack = newreno_rcv_ack, + .rcv_cong_ack = newreno_rcv_cong_ack, + .init = newreno_conn_init +}; + +clib_error_t * +newreno_init (vlib_main_t * vm) +{ + clib_error_t *error = 0; + + tcp_cc_algo_register (TCP_CC_NEWRENO, &tcp_newreno); + + return error; +} + +VLIB_INIT_FUNCTION (newreno_init); + +/* + * fd.io coding-style-patch-verification: ON + * + * Local Variables: + * eval: (c-set-style "gnu") + * End: + */ diff --git a/src/vnet/tcp/tcp_output.c b/src/vnet/tcp/tcp_output.c new file mode 100644 index 00000000..dbcf1f74 --- /dev/null +++ b/src/vnet/tcp/tcp_output.c @@ -0,0 +1,1412 @@ +/* + * Copyright (c) 2016 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +#include +#include + +vlib_node_registration_t tcp4_output_node; +vlib_node_registration_t tcp6_output_node; + +typedef enum _tcp_output_nect +{ + TCP_OUTPUT_NEXT_DROP, + TCP_OUTPUT_NEXT_IP_LOOKUP, + TCP_OUTPUT_N_NEXT +} tcp_output_next_t; + +#define foreach_tcp4_output_next \ + _ (DROP, "error-drop") \ + _ (IP_LOOKUP, "ip4-lookup") + +#define foreach_tcp6_output_next \ + _ (DROP, "error-drop") \ + _ (IP_LOOKUP, "ip6-lookup") + +static char *tcp_error_strings[] = { +#define tcp_error(n,s) s, +#include +#undef tcp_error +}; + +typedef struct +{ + u16 src_port; + u16 dst_port; + u8 state; +} tcp_tx_trace_t; + +u16 dummy_mtu = 400; + +u8 * +format_tcp_tx_trace (u8 * s, va_list * args) +{ + CLIB_UNUSED (vlib_main_t * vm) = va_arg (*args, vlib_main_t *); + CLIB_UNUSED (vlib_node_t * node) = va_arg (*args, vlib_node_t *); + + s = format (s, "TBD\n"); + + return s; +} + +void +tcp_set_snd_mss (tcp_connection_t * tc) +{ + u16 snd_mss; + + /* TODO find our iface MTU */ + snd_mss = dummy_mtu; + + /* TODO cache mss and consider PMTU discovery */ + snd_mss = tc->opt.mss < snd_mss ? tc->opt.mss : snd_mss; + + tc->snd_mss = snd_mss; + + if (tc->snd_mss == 0) + { + clib_warning ("snd mss is 0"); + tc->snd_mss = dummy_mtu; + } +} + +static u8 +tcp_window_compute_scale (u32 available_space) +{ + u8 wnd_scale = 0; + while (wnd_scale < TCP_MAX_WND_SCALE + && (available_space >> wnd_scale) > TCP_WND_MAX) + wnd_scale++; + return wnd_scale; +} + +/** + * Compute initial window and scale factor. As per RFC1323, window field in + * SYN and SYN-ACK segments is never scaled. + */ +u32 +tcp_initial_window_to_advertise (tcp_connection_t * tc) +{ + u32 available_space; + + /* Initial wnd for SYN. Fifos are not allocated yet. + * Use some predefined value */ + if (tc->state != TCP_STATE_SYN_RCVD) + { + return TCP_DEFAULT_RX_FIFO_SIZE; + } + + available_space = stream_session_max_enqueue (&tc->connection); + tc->rcv_wscale = tcp_window_compute_scale (available_space); + tc->rcv_wnd = clib_min (available_space, TCP_WND_MAX << tc->rcv_wscale); + + return clib_min (tc->rcv_wnd, TCP_WND_MAX); +} + +/** + * Compute and return window to advertise, scaled as per RFC1323 + */ +u32 +tcp_window_to_advertise (tcp_connection_t * tc, tcp_state_t state) +{ + u32 available_space, wnd, scaled_space; + + if (state != TCP_STATE_ESTABLISHED) + return tcp_initial_window_to_advertise (tc); + + available_space = stream_session_max_enqueue (&tc->connection); + scaled_space = available_space >> tc->rcv_wscale; + + /* Need to update scale */ + if (PREDICT_FALSE ((scaled_space == 0 && available_space != 0)) + || (scaled_space >= TCP_WND_MAX)) + tc->rcv_wscale = tcp_window_compute_scale (available_space); + + wnd = clib_min (available_space, TCP_WND_MAX << tc->rcv_wscale); + tc->rcv_wnd = wnd; + + return wnd >> tc->rcv_wscale; +} + +/** + * Write TCP options to segment. + */ +u32 +tcp_options_write (u8 * data, tcp_options_t * opts) +{ + u32 opts_len = 0; + u32 buf, seq_len = 4; + + if (tcp_opts_mss (opts)) + { + *data++ = TCP_OPTION_MSS; + *data++ = TCP_OPTION_LEN_MSS; + buf = clib_host_to_net_u16 (opts->mss); + clib_memcpy (data, &buf, sizeof (opts->mss)); + data += sizeof (opts->mss); + opts_len += TCP_OPTION_LEN_MSS; + } + + if (tcp_opts_wscale (opts)) + { + *data++ = TCP_OPTION_WINDOW_SCALE; + *data++ = TCP_OPTION_LEN_WINDOW_SCALE; + *data++ = opts->wscale; + opts_len += TCP_OPTION_LEN_WINDOW_SCALE; + } + + if (tcp_opts_sack_permitted (opts)) + { + *data++ = TCP_OPTION_SACK_PERMITTED; + *data++ = TCP_OPTION_LEN_SACK_PERMITTED; + opts_len += TCP_OPTION_LEN_SACK_PERMITTED; + } + + if (tcp_opts_tstamp (opts)) + { + *data++ = TCP_OPTION_TIMESTAMP; + *data++ = TCP_OPTION_LEN_TIMESTAMP; + buf = clib_host_to_net_u32 (opts->tsval); + clib_memcpy (data, &buf, sizeof (opts->tsval)); + data += sizeof (opts->tsval); + buf = clib_host_to_net_u32 (opts->tsecr); + clib_memcpy (data, &buf, sizeof (opts->tsecr)); + data += sizeof (opts->tsecr); + opts_len += TCP_OPTION_LEN_TIMESTAMP; + } + + if (tcp_opts_sack (opts)) + { + int i; + u32 n_sack_blocks = clib_min (vec_len (opts->sacks), + TCP_OPTS_MAX_SACK_BLOCKS); + + if (n_sack_blocks != 0) + { + *data++ = TCP_OPTION_SACK_BLOCK; + *data++ = 2 + n_sack_blocks * TCP_OPTION_LEN_SACK_BLOCK; + for (i = 0; i < n_sack_blocks; i++) + { + buf = clib_host_to_net_u32 (opts->sacks[i].start); + clib_memcpy (data, &buf, seq_len); + data += seq_len; + buf = clib_host_to_net_u32 (opts->sacks[i].end); + clib_memcpy (data, &buf, seq_len); + data += seq_len; + } + opts_len += 2 + n_sack_blocks * TCP_OPTION_LEN_SACK_BLOCK; + } + } + + /* Terminate TCP options */ + if (opts_len % 4) + { + *data++ = TCP_OPTION_EOL; + opts_len += TCP_OPTION_LEN_EOL; + } + + /* Pad with zeroes to a u32 boundary */ + while (opts_len % 4) + { + *data++ = TCP_OPTION_NOOP; + opts_len += TCP_OPTION_LEN_NOOP; + } + return opts_len; +} + +always_inline int +tcp_make_syn_options (tcp_options_t * opts, u32 initial_wnd) +{ + u8 len = 0; + + opts->flags |= TCP_OPTS_FLAG_MSS; + opts->mss = dummy_mtu; /*XXX discover that */ + len += TCP_OPTION_LEN_MSS; + + opts->flags |= TCP_OPTS_FLAG_WSCALE; + opts->wscale = tcp_window_compute_scale (initial_wnd); + len += TCP_OPTION_LEN_WINDOW_SCALE; + + opts->flags |= TCP_OPTS_FLAG_TSTAMP; + opts->tsval = tcp_time_now (); + opts->tsecr = 0; + len += TCP_OPTION_LEN_TIMESTAMP; + + opts->flags |= TCP_OPTS_FLAG_SACK_PERMITTED; + len += TCP_OPTION_LEN_SACK_PERMITTED; + + /* Align to needed boundary */ + len += (TCP_OPTS_ALIGN - len % TCP_OPTS_ALIGN) % TCP_OPTS_ALIGN; + return len; +} + +always_inline int +tcp_make_synack_options (tcp_connection_t * tc, tcp_options_t * opts) +{ + u8 len = 0; + + opts->flags |= TCP_OPTS_FLAG_MSS; + opts->mss = dummy_mtu; /*XXX discover that */ + len += TCP_OPTION_LEN_MSS; + + if (tcp_opts_wscale (&tc->opt)) + { + opts->flags |= TCP_OPTS_FLAG_WSCALE; + opts->wscale = tc->rcv_wscale; + len += TCP_OPTION_LEN_WINDOW_SCALE; + } + + if (tcp_opts_tstamp (&tc->opt)) + { + opts->flags |= TCP_OPTS_FLAG_TSTAMP; + opts->tsval = tcp_time_now (); + opts->tsecr = tc->tsval_recent; + len += TCP_OPTION_LEN_TIMESTAMP; + } + + if (tcp_opts_sack_permitted (&tc->opt)) + { + opts->flags |= TCP_OPTS_FLAG_SACK_PERMITTED; + len += TCP_OPTION_LEN_SACK_PERMITTED; + } + + /* Align to needed boundary */ + len += (TCP_OPTS_ALIGN - len % TCP_OPTS_ALIGN) % TCP_OPTS_ALIGN; + return len; +} + +always_inline int +tcp_make_established_options (tcp_connection_t * tc, tcp_options_t * opts) +{ + u8 len = 0; + + opts->flags = 0; + + if (tcp_opts_tstamp (&tc->opt)) + { + opts->flags |= TCP_OPTS_FLAG_TSTAMP; + opts->tsval = tcp_time_now (); + opts->tsecr = tc->tsval_recent; + len += TCP_OPTION_LEN_TIMESTAMP; + } + if (tcp_opts_sack_permitted (&tc->opt)) + { + if (vec_len (tc->snd_sacks)) + { + opts->flags |= TCP_OPTS_FLAG_SACK; + opts->sacks = tc->snd_sacks; + opts->n_sack_blocks = vec_len (tc->snd_sacks); + len += 2 + TCP_OPTION_LEN_SACK_BLOCK * opts->n_sack_blocks; + } + } + + /* Align to needed boundary */ + len += (TCP_OPTS_ALIGN - len % TCP_OPTS_ALIGN) % TCP_OPTS_ALIGN; + return len; +} + +always_inline int +tcp_make_options (tcp_connection_t * tc, tcp_options_t * opts, + tcp_state_t state) +{ + switch (state) + { + case TCP_STATE_ESTABLISHED: + case TCP_STATE_FIN_WAIT_1: + return tcp_make_established_options (tc, opts); + case TCP_STATE_SYN_RCVD: + return tcp_make_synack_options (tc, opts); + case TCP_STATE_SYN_SENT: + return tcp_make_syn_options (opts, + tcp_initial_window_to_advertise (tc)); + default: + clib_warning ("Not handled!"); + return 0; + } +} + +#define tcp_get_free_buffer_index(tm, bidx) \ +do { \ + u32 *my_tx_buffers, n_free_buffers; \ + u32 cpu_index = tm->vlib_main->cpu_index; \ + my_tx_buffers = tm->tx_buffers[cpu_index]; \ + if (PREDICT_FALSE(vec_len (my_tx_buffers) == 0)) \ + { \ + n_free_buffers = 32; /* TODO config or macro */ \ + vec_validate (my_tx_buffers, n_free_buffers - 1); \ + _vec_len(my_tx_buffers) = vlib_buffer_alloc_from_free_list ( \ + tm->vlib_main, my_tx_buffers, n_free_buffers, \ + VLIB_BUFFER_DEFAULT_FREE_LIST_INDEX); \ + tm->tx_buffers[cpu_index] = my_tx_buffers; \ + } \ + /* buffer shortage */ \ + if (PREDICT_FALSE (vec_len (my_tx_buffers) == 0)) \ + return; \ + *bidx = my_tx_buffers[_vec_len (my_tx_buffers)-1]; \ + _vec_len (my_tx_buffers) -= 1; \ +} while (0) + +always_inline void +tcp_reuse_buffer (vlib_main_t * vm, vlib_buffer_t * b) +{ + vlib_buffer_t *it = b; + do + { + it->current_data = 0; + it->current_length = 0; + it->total_length_not_including_first_buffer = 0; + } + while ((it->flags & VLIB_BUFFER_NEXT_PRESENT) + && (it = vlib_get_buffer (vm, it->next_buffer))); + + /* Leave enough space for headers */ + vlib_buffer_make_headroom (b, MAX_HDRS_LEN); +} + +/** + * Prepare ACK + */ +void +tcp_make_ack_i (tcp_connection_t * tc, vlib_buffer_t * b, tcp_state_t state, + u8 flags) +{ + tcp_options_t _snd_opts, *snd_opts = &_snd_opts; + u8 tcp_opts_len, tcp_hdr_opts_len; + tcp_header_t *th; + u16 wnd; + + wnd = tcp_window_to_advertise (tc, state); + + /* Make and write options */ + tcp_opts_len = tcp_make_established_options (tc, snd_opts); + tcp_hdr_opts_len = tcp_opts_len + sizeof (tcp_header_t); + + th = vlib_buffer_push_tcp (b, tc->c_lcl_port, tc->c_rmt_port, tc->snd_nxt, + tc->rcv_nxt, tcp_hdr_opts_len, flags, wnd); + + tcp_options_write ((u8 *) (th + 1), snd_opts); + + /* Mark as ACK */ + vnet_buffer (b)->tcp.connection_index = tc->c_c_index; +} + +/** + * Convert buffer to ACK + */ +void +tcp_make_ack (tcp_connection_t * tc, vlib_buffer_t * b) +{ + tcp_main_t *tm = vnet_get_tcp_main (); + vlib_main_t *vm = tm->vlib_main; + + tcp_reuse_buffer (vm, b); + tcp_make_ack_i (tc, b, TCP_STATE_ESTABLISHED, TCP_FLAG_ACK); + vnet_buffer (b)->tcp.flags = TCP_BUF_FLAG_ACK; +} + +/** + * Convert buffer to FIN-ACK + */ +void +tcp_make_finack (tcp_connection_t * tc, vlib_buffer_t * b) +{ + tcp_main_t *tm = vnet_get_tcp_main (); + vlib_main_t *vm = tm->vlib_main; + + tcp_reuse_buffer (vm, b); + tcp_make_ack_i (tc, b, TCP_STATE_ESTABLISHED, TCP_FLAG_ACK | TCP_FLAG_FIN); + + /* Reset flags, make sure ack is sent */ + tc->flags = TCP_CONN_SNDACK; + vnet_buffer (b)->tcp.flags &= ~TCP_BUF_FLAG_DUPACK; + + tc->snd_nxt += 1; +} + +/** + * Convert buffer to SYN-ACK + */ +void +tcp_make_synack (tcp_connection_t * tc, vlib_buffer_t * b) +{ + tcp_main_t *tm = vnet_get_tcp_main (); + vlib_main_t *vm = tm->vlib_main; + tcp_options_t _snd_opts, *snd_opts = &_snd_opts; + u8 tcp_opts_len, tcp_hdr_opts_len; + tcp_header_t *th; + u16 initial_wnd; + u32 time_now; + + memset (snd_opts, 0, sizeof (*snd_opts)); + + tcp_reuse_buffer (vm, b); + + /* Set random initial sequence */ + time_now = tcp_time_now (); + + tc->iss = random_u32 (&time_now); + tc->snd_una = tc->iss; + tc->snd_nxt = tc->iss + 1; + tc->snd_una_max = tc->snd_nxt; + + initial_wnd = tcp_initial_window_to_advertise (tc); + + /* Make and write options */ + tcp_opts_len = tcp_make_synack_options (tc, snd_opts); + tcp_hdr_opts_len = tcp_opts_len + sizeof (tcp_header_t); + + th = vlib_buffer_push_tcp (b, tc->c_lcl_port, tc->c_rmt_port, tc->iss, + tc->rcv_nxt, tcp_hdr_opts_len, + TCP_FLAG_SYN | TCP_FLAG_ACK, initial_wnd); + + tcp_options_write ((u8 *) (th + 1), snd_opts); + + vnet_buffer (b)->tcp.connection_index = tc->c_c_index; + vnet_buffer (b)->tcp.flags = TCP_BUF_FLAG_ACK; + + /* Init retransmit timer */ + tcp_retransmit_timer_set (tm, tc); +} + +always_inline void +tcp_enqueue_to_ip_lookup (vlib_main_t * vm, vlib_buffer_t * b, u32 bi, + u8 is_ip4) +{ + u32 *to_next, next_index; + vlib_frame_t *f; + + b->flags |= VNET_BUFFER_LOCALLY_ORIGINATED; + b->error = 0; + + /* Default FIB for now */ + vnet_buffer (b)->sw_if_index[VLIB_TX] = 0; + + /* Send to IP lookup */ + next_index = is_ip4 ? ip4_lookup_node.index : ip6_lookup_node.index; + f = vlib_get_frame_to_node (vm, next_index); + + /* Enqueue the packet */ + to_next = vlib_frame_vector_args (f); + to_next[0] = bi; + f->n_vectors = 1; + vlib_put_frame_to_node (vm, next_index, f); +} + +int +tcp_make_reset_in_place (vlib_main_t * vm, vlib_buffer_t * b0, + tcp_state_t state, u32 my_thread_index, u8 is_ip4) +{ + u8 tcp_hdr_len = sizeof (tcp_header_t); + ip4_header_t *ih4; + ip6_header_t *ih6; + tcp_header_t *th0; + ip4_address_t src_ip40; + ip6_address_t src_ip60; + u16 src_port0; + u32 tmp; + + /* Find IP and TCP headers */ + if (is_ip4) + { + ih4 = vlib_buffer_get_current (b0); + th0 = ip4_next_header (ih4); + } + else + { + ih6 = vlib_buffer_get_current (b0); + th0 = ip6_next_header (ih6); + } + + /* Swap src and dst ip */ + if (is_ip4) + { + ASSERT ((ih4->ip_version_and_header_length & 0xF0) == 0x40); + src_ip40.as_u32 = ih4->src_address.as_u32; + ih4->src_address.as_u32 = ih4->dst_address.as_u32; + ih4->dst_address.as_u32 = src_ip40.as_u32; + + /* Chop the end of the pkt */ + b0->current_length += ip4_header_bytes (ih4) + tcp_hdr_len; + } + else + { + ASSERT ((ih6->ip_version_traffic_class_and_flow_label & 0xF0) == 0x60); + clib_memcpy (&src_ip60, &ih6->src_address, sizeof (ip6_address_t)); + clib_memcpy (&ih6->src_address, &ih6->dst_address, + sizeof (ip6_address_t)); + clib_memcpy (&ih6->dst_address, &src_ip60, sizeof (ip6_address_t)); + + /* Chop the end of the pkt */ + b0->current_length += sizeof (ip6_header_t) + tcp_hdr_len; + } + + /* Try to determine what/why we're actually resetting and swap + * src and dst ports */ + if (state == TCP_STATE_CLOSED) + { + if (!tcp_syn (th0)) + return -1; + + tmp = clib_net_to_host_u32 (th0->seq_number); + + /* Got a SYN for no listener. */ + th0->flags = TCP_FLAG_RST | TCP_FLAG_ACK; + th0->ack_number = clib_host_to_net_u32 (tmp + 1); + th0->seq_number = 0; + + } + else if (state >= TCP_STATE_SYN_SENT) + { + th0->flags = TCP_FLAG_RST | TCP_FLAG_ACK; + th0->seq_number = th0->ack_number; + th0->ack_number = 0; + } + + src_port0 = th0->src_port; + th0->src_port = th0->dst_port; + th0->dst_port = src_port0; + th0->window = 0; + th0->data_offset_and_reserved = (tcp_hdr_len >> 2) << 4; + th0->urgent_pointer = 0; + + /* Compute checksum */ + if (is_ip4) + { + th0->checksum = ip4_tcp_udp_compute_checksum (vm, b0, ih4); + } + else + { + int bogus = ~0; + th0->checksum = ip6_tcp_udp_icmp_compute_checksum (vm, b0, ih6, &bogus); + ASSERT (!bogus); + } + + return 0; +} + +/** + * Send reset without reusing existing buffer + */ +void +tcp_send_reset (vlib_buffer_t * pkt, u8 is_ip4) +{ + vlib_buffer_t *b; + u32 bi; + tcp_main_t *tm = vnet_get_tcp_main (); + vlib_main_t *vm = tm->vlib_main; + u8 tcp_hdr_len, flags = 0; + tcp_header_t *th, *pkt_th; + u32 seq, ack; + ip4_header_t *ih4, *pkt_ih4; + ip6_header_t *ih6, *pkt_ih6; + + tcp_get_free_buffer_index (tm, &bi); + b = vlib_get_buffer (vm, bi); + + /* Leave enough space for headers */ + vlib_buffer_make_headroom (b, MAX_HDRS_LEN); + + /* Make and write options */ + tcp_hdr_len = sizeof (tcp_header_t); + + if (is_ip4) + { + pkt_ih4 = vlib_buffer_get_current (pkt); + pkt_th = ip4_next_header (pkt_ih4); + } + else + { + pkt_ih6 = vlib_buffer_get_current (pkt); + pkt_th = ip6_next_header (pkt_ih6); + } + + if (tcp_ack (pkt_th)) + { + flags = TCP_FLAG_RST; + seq = pkt_th->ack_number; + ack = 0; + } + else + { + flags = TCP_FLAG_RST | TCP_FLAG_ACK; + seq = 0; + ack = clib_host_to_net_u32 (vnet_buffer (pkt)->tcp.seq_end); + } + + th = vlib_buffer_push_tcp_net_order (b, pkt_th->dst_port, pkt_th->src_port, + seq, ack, tcp_hdr_len, flags, 0); + + /* Swap src and dst ip */ + if (is_ip4) + { + ASSERT ((pkt_ih4->ip_version_and_header_length & 0xF0) == 0x40); + ih4 = vlib_buffer_push_ip4 (vm, b, &pkt_ih4->dst_address, + &pkt_ih4->src_address, IP_PROTOCOL_TCP); + th->checksum = ip4_tcp_udp_compute_checksum (vm, b, ih4); + } + else + { + int bogus = ~0; + pkt_ih6 = (ip6_header_t *) (pkt_th - 1); + ASSERT ((pkt_ih6->ip_version_traffic_class_and_flow_label & 0xF0) == + 0x60); + ih6 = + vlib_buffer_push_ip6 (vm, b, &pkt_ih6->dst_address, + &pkt_ih6->src_address, IP_PROTOCOL_TCP); + th->checksum = ip6_tcp_udp_icmp_compute_checksum (vm, b, ih6, &bogus); + ASSERT (!bogus); + } + + tcp_enqueue_to_ip_lookup (vm, b, bi, is_ip4); +} + +void +tcp_push_ip_hdr (tcp_main_t * tm, tcp_connection_t * tc, vlib_buffer_t * b) +{ + tcp_header_t *th = vlib_buffer_get_current (b); + + if (tc->c_is_ip4) + { + ip4_header_t *ih; + ih = vlib_buffer_push_ip4 (tm->vlib_main, b, &tc->c_lcl_ip4, + &tc->c_rmt_ip4, IP_PROTOCOL_TCP); + th->checksum = ip4_tcp_udp_compute_checksum (tm->vlib_main, b, ih); + } + else + { + ip6_header_t *ih; + int bogus = ~0; + + ih = vlib_buffer_push_ip6 (tm->vlib_main, b, &tc->c_lcl_ip6, + &tc->c_rmt_ip6, IP_PROTOCOL_TCP); + th->checksum = ip6_tcp_udp_icmp_compute_checksum (tm->vlib_main, b, ih, + &bogus); + ASSERT (!bogus); + } +} + +/** + * Send SYN + * + * Builds a SYN packet for a half-open connection and sends it to ipx_lookup. + * The packet is not forwarded through tcpx_output to avoid doing lookups + * in the half_open pool. + */ +void +tcp_send_syn (tcp_connection_t * tc) +{ + vlib_buffer_t *b; + u32 bi; + tcp_main_t *tm = vnet_get_tcp_main (); + vlib_main_t *vm = tm->vlib_main; + u8 tcp_hdr_opts_len, tcp_opts_len; + tcp_header_t *th; + u32 time_now; + u16 initial_wnd; + tcp_options_t snd_opts; + + tcp_get_free_buffer_index (tm, &bi); + b = vlib_get_buffer (vm, bi); + + /* Leave enough space for headers */ + vlib_buffer_make_headroom (b, MAX_HDRS_LEN); + + /* Set random initial sequence */ + time_now = tcp_time_now (); + + tc->iss = random_u32 (&time_now); + tc->snd_una = tc->iss; + tc->snd_una_max = tc->snd_nxt = tc->iss + 1; + + initial_wnd = tcp_initial_window_to_advertise (tc); + + /* Make and write options */ + memset (&snd_opts, 0, sizeof (snd_opts)); + tcp_opts_len = tcp_make_syn_options (&snd_opts, initial_wnd); + tcp_hdr_opts_len = tcp_opts_len + sizeof (tcp_header_t); + + th = vlib_buffer_push_tcp (b, tc->c_lcl_port, tc->c_rmt_port, tc->iss, + tc->rcv_nxt, tcp_hdr_opts_len, TCP_FLAG_SYN, + initial_wnd); + + tcp_options_write ((u8 *) (th + 1), &snd_opts); + + /* Measure RTT with this */ + tc->rtt_ts = tcp_time_now (); + tc->rtt_seq = tc->snd_nxt; + + /* Start retransmit trimer */ + tcp_timer_set (tc, TCP_TIMER_RETRANSMIT_SYN, tc->rto * TCP_TO_TIMER_TICK); + tc->rto_boff = 0; + + /* Set the connection establishment timer */ + tcp_timer_set (tc, TCP_TIMER_ESTABLISH, TCP_ESTABLISH_TIME); + + tcp_push_ip_hdr (tm, tc, b); + tcp_enqueue_to_ip_lookup (vm, b, bi, tc->c_is_ip4); +} + +always_inline void +tcp_enqueue_to_output (vlib_main_t * vm, vlib_buffer_t * b, u32 bi, u8 is_ip4) +{ + u32 *to_next, next_index; + vlib_frame_t *f; + + b->flags |= VNET_BUFFER_LOCALLY_ORIGINATED; + b->error = 0; + + /* Decide where to send the packet */ + next_index = is_ip4 ? tcp4_output_node.index : tcp6_output_node.index; + f = vlib_get_frame_to_node (vm, next_index); + + /* Enqueue the packet */ + to_next = vlib_frame_vector_args (f); + to_next[0] = bi; + f->n_vectors = 1; + vlib_put_frame_to_node (vm, next_index, f); +} + +/** + * Send FIN + */ +void +tcp_send_fin (tcp_connection_t * tc) +{ + vlib_buffer_t *b; + u32 bi; + tcp_main_t *tm = vnet_get_tcp_main (); + vlib_main_t *vm = tm->vlib_main; + + tcp_get_free_buffer_index (tm, &bi); + b = vlib_get_buffer (vm, bi); + + /* Leave enough space for headers */ + vlib_buffer_make_headroom (b, MAX_HDRS_LEN); + + tcp_make_finack (tc, b); + + tcp_enqueue_to_output (vm, b, bi, tc->c_is_ip4); +} + +always_inline u8 +tcp_make_state_flags (tcp_state_t next_state) +{ + switch (next_state) + { + case TCP_STATE_ESTABLISHED: + return TCP_FLAG_ACK; + case TCP_STATE_SYN_RCVD: + return TCP_FLAG_SYN | TCP_FLAG_ACK; + case TCP_STATE_SYN_SENT: + return TCP_FLAG_SYN; + case TCP_STATE_LAST_ACK: + case TCP_STATE_FIN_WAIT_1: + return TCP_FLAG_FIN; + default: + clib_warning ("Shouldn't be here!"); + } + return 0; +} + +/** + * Push TCP header and update connection variables + */ +static void +tcp_push_hdr_i (tcp_connection_t * tc, vlib_buffer_t * b, + tcp_state_t next_state) +{ + u32 advertise_wnd, data_len; + u8 tcp_opts_len, tcp_hdr_opts_len, opts_write_len, flags; + tcp_options_t _snd_opts, *snd_opts = &_snd_opts; + tcp_header_t *th; + + data_len = b->current_length; + vnet_buffer (b)->tcp.flags = 0; + + /* Make and write options */ + memset (snd_opts, 0, sizeof (*snd_opts)); + tcp_opts_len = tcp_make_options (tc, snd_opts, next_state); + tcp_hdr_opts_len = tcp_opts_len + sizeof (tcp_header_t); + + /* Get rcv window to advertise */ + advertise_wnd = tcp_window_to_advertise (tc, next_state); + flags = tcp_make_state_flags (next_state); + + /* Push header and options */ + th = vlib_buffer_push_tcp (b, tc->c_lcl_port, tc->c_rmt_port, tc->snd_nxt, + tc->rcv_nxt, tcp_hdr_opts_len, flags, + advertise_wnd); + + opts_write_len = tcp_options_write ((u8 *) (th + 1), snd_opts); + + ASSERT (opts_write_len == tcp_opts_len); + + /* Tag the buffer with the connection index */ + vnet_buffer (b)->tcp.connection_index = tc->c_c_index; + + tc->snd_nxt += data_len; +} + +/* Send delayed ACK when timer expires */ +void +tcp_timer_delack_handler (u32 index) +{ + tcp_main_t *tm = vnet_get_tcp_main (); + vlib_main_t *vm = tm->vlib_main; + u32 thread_index = os_get_cpu_number (); + tcp_connection_t *tc; + vlib_buffer_t *b; + u32 bi; + + tc = tcp_connection_get (index, thread_index); + + /* Get buffer */ + tcp_get_free_buffer_index (tm, &bi); + b = vlib_get_buffer (vm, bi); + + /* Fill in the ACK */ + tcp_make_ack (tc, b); + + tc->timers[TCP_TIMER_DELACK] = TCP_TIMER_HANDLE_INVALID; + tc->flags &= ~TCP_CONN_DELACK; + + tcp_enqueue_to_output (vm, b, bi, tc->c_is_ip4); +} + +/** Build a retransmit segment + * + * @return the number of bytes in the segment or 0 if there's nothing to + * retransmit + * */ +u32 +tcp_prepare_retransmit_segment (tcp_connection_t * tc, vlib_buffer_t * b, + u32 max_bytes) +{ + tcp_main_t *tm = vnet_get_tcp_main (); + vlib_main_t *vm = tm->vlib_main; + u32 n_bytes, offset = 0; + sack_scoreboard_hole_t *hole; + u32 hole_size; + + tcp_reuse_buffer (vm, b); + + ASSERT (tc->state == TCP_STATE_ESTABLISHED); + ASSERT (max_bytes != 0); + + if (tcp_opts_sack_permitted (&tc->opt)) + { + /* XXX get first hole not retransmitted yet */ + hole = scoreboard_first_hole (&tc->sack_sb); + if (!hole) + return 0; + + offset = hole->start - tc->snd_una; + hole_size = hole->end - hole->start; + + ASSERT (hole_size); + + if (hole_size < max_bytes) + max_bytes = hole_size; + } + else + { + if (seq_geq (tc->snd_nxt, tc->snd_una_max)) + return 0; + } + + n_bytes = stream_session_peek_bytes (&tc->connection, + vlib_buffer_get_current (b), offset, + max_bytes); + ASSERT (n_bytes != 0); + + tc->snd_nxt += n_bytes; + tcp_push_hdr_i (tc, b, tc->state); + + return n_bytes; +} + +static void +tcp_timer_retransmit_handler_i (u32 index, u8 is_syn) +{ + tcp_main_t *tm = vnet_get_tcp_main (); + vlib_main_t *vm = tm->vlib_main; + u32 thread_index = os_get_cpu_number (); + tcp_connection_t *tc; + vlib_buffer_t *b; + u32 bi, max_bytes, snd_space; + + if (is_syn) + { + tc = tcp_half_open_connection_get (index); + } + else + { + tc = tcp_connection_get (index, thread_index); + } + + /* Make sure timer handle is set to invalid */ + tc->timers[TCP_TIMER_RETRANSMIT] = TCP_TIMER_HANDLE_INVALID; + + /* Increment RTO backoff (also equal to number of retries) */ + tc->rto_boff += 1; + + /* Go back to first un-acked byte */ + tc->snd_nxt = tc->snd_una; + + /* Get buffer */ + tcp_get_free_buffer_index (tm, &bi); + b = vlib_get_buffer (vm, bi); + + if (tc->state == TCP_STATE_ESTABLISHED) + { + tcp_fastrecovery_off (tc); + + /* Exponential backoff */ + tc->rto = clib_min (tc->rto << 1, TCP_RTO_MAX); + + /* Figure out what and how many bytes we can send */ + snd_space = tcp_available_snd_space (tc); + max_bytes = clib_min (tc->snd_mss, snd_space); + tcp_prepare_retransmit_segment (tc, b, max_bytes); + + tc->rtx_bytes += max_bytes; + + /* No fancy recovery for now! */ + scoreboard_clear (&tc->sack_sb); + } + else + { + /* Retransmit for SYN/SYNACK */ + ASSERT (tc->state == TCP_STATE_SYN_RCVD + || tc->state == TCP_STATE_SYN_SENT); + + /* Try without increasing RTO a number of times. If this fails, + * start growing RTO exponentially */ + if (tc->rto_boff > TCP_RTO_SYN_RETRIES) + tc->rto = clib_min (tc->rto << 1, TCP_RTO_MAX); + + vlib_buffer_make_headroom (b, MAX_HDRS_LEN); + tcp_push_hdr_i (tc, b, tc->state); + } + + if (!is_syn) + { + tcp_enqueue_to_output (vm, b, bi, tc->c_is_ip4); + + /* Re-enable retransmit timer */ + tcp_retransmit_timer_set (tm, tc); + } + else + { + ASSERT (tc->state == TCP_STATE_SYN_SENT); + + /* This goes straight to ipx_lookup */ + tcp_push_ip_hdr (tm, tc, b); + tcp_enqueue_to_ip_lookup (vm, b, bi, tc->c_is_ip4); + + /* Re-enable retransmit timer */ + tcp_timer_set (tc, TCP_TIMER_RETRANSMIT_SYN, + tc->rto * TCP_TO_TIMER_TICK); + } +} + +void +tcp_timer_retransmit_handler (u32 index) +{ + tcp_timer_retransmit_handler_i (index, 0); +} + +void +tcp_timer_retransmit_syn_handler (u32 index) +{ + tcp_timer_retransmit_handler_i (index, 1); +} + +/** + * Retansmit first unacked segment */ +void +tcp_retransmit_first_unacked (tcp_connection_t * tc) +{ + tcp_main_t *tm = vnet_get_tcp_main (); + u32 snd_nxt = tc->snd_nxt; + vlib_buffer_t *b; + u32 bi; + + tc->snd_nxt = tc->snd_una; + + /* Get buffer */ + tcp_get_free_buffer_index (tm, &bi); + b = vlib_get_buffer (tm->vlib_main, bi); + + tcp_prepare_retransmit_segment (tc, b, tc->snd_mss); + tcp_enqueue_to_output (tm->vlib_main, b, bi, tc->c_is_ip4); + + tc->snd_nxt = snd_nxt; + tc->rtx_bytes += tc->snd_mss; +} + +void +tcp_fast_retransmit (tcp_connection_t * tc) +{ + tcp_main_t *tm = vnet_get_tcp_main (); + u32 snd_space, max_bytes, n_bytes, bi; + vlib_buffer_t *b; + + ASSERT (tcp_in_fastrecovery (tc)); + + clib_warning ("fast retransmit!"); + + /* Start resending from first un-acked segment */ + tc->snd_nxt = tc->snd_una; + + snd_space = tcp_available_snd_space (tc); + + while (snd_space) + { + tcp_get_free_buffer_index (tm, &bi); + b = vlib_get_buffer (tm->vlib_main, bi); + + max_bytes = clib_min (tc->snd_mss, snd_space); + n_bytes = tcp_prepare_retransmit_segment (tc, b, max_bytes); + + /* Nothing left to retransmit */ + if (n_bytes == 0) + return; + + tcp_enqueue_to_output (tm->vlib_main, b, bi, tc->c_is_ip4); + + snd_space -= n_bytes; + } + + /* If window allows, send new data */ + tc->snd_nxt = tc->snd_una_max; +} + +always_inline u32 +tcp_session_has_ooo_data (tcp_connection_t * tc) +{ + stream_session_t *s = + stream_session_get (tc->c_s_index, tc->c_thread_index); + return svm_fifo_has_ooo_data (s->server_rx_fifo); +} + +always_inline uword +tcp46_output_inline (vlib_main_t * vm, + vlib_node_runtime_t * node, + vlib_frame_t * from_frame, int is_ip4) +{ + tcp_main_t *tm = vnet_get_tcp_main (); + u32 n_left_from, next_index, *from, *to_next; + u32 my_thread_index = vm->cpu_index; + + from = vlib_frame_vector_args (from_frame); + n_left_from = from_frame->n_vectors; + + next_index = node->cached_next_index; + + while (n_left_from > 0) + { + u32 n_left_to_next; + + vlib_get_next_frame (vm, node, next_index, to_next, n_left_to_next); + + while (n_left_from > 0 && n_left_to_next > 0) + { + u32 bi0; + vlib_buffer_t *b0; + tcp_connection_t *tc0; + tcp_header_t *th0; + u32 error0 = TCP_ERROR_PKTS_SENT, next0 = TCP_OUTPUT_NEXT_IP_LOOKUP; + + bi0 = from[0]; + to_next[0] = bi0; + from += 1; + to_next += 1; + n_left_from -= 1; + n_left_to_next -= 1; + + b0 = vlib_get_buffer (vm, bi0); + tc0 = tcp_connection_get (vnet_buffer (b0)->tcp.connection_index, + my_thread_index); + th0 = vlib_buffer_get_current (b0); + + if (is_ip4) + { + ip4_header_t *ih0; + ih0 = vlib_buffer_push_ip4 (vm, b0, &tc0->c_lcl_ip4, + &tc0->c_rmt_ip4, IP_PROTOCOL_TCP); + th0->checksum = ip4_tcp_udp_compute_checksum (vm, b0, ih0); + } + else + { + ip6_header_t *ih0; + int bogus = ~0; + + ih0 = vlib_buffer_push_ip6 (vm, b0, &tc0->c_lcl_ip6, + &tc0->c_rmt_ip6, IP_PROTOCOL_TCP); + th0->checksum = ip6_tcp_udp_icmp_compute_checksum (vm, b0, ih0, + &bogus); + ASSERT (!bogus); + } + + /* Filter out DUPACKs if there are no OOO segments left */ + if (PREDICT_FALSE + (vnet_buffer (b0)->tcp.flags & TCP_BUF_FLAG_DUPACK)) + { + tc0->snt_dupacks--; + ASSERT (tc0->snt_dupacks >= 0); + if (!tcp_session_has_ooo_data (tc0)) + { + error0 = TCP_ERROR_FILTERED_DUPACKS; + next0 = TCP_OUTPUT_NEXT_DROP; + goto done; + } + } + + /* Retransmitted SYNs do reach this but it should be harmless */ + tc0->rcv_las = tc0->rcv_nxt; + + /* Stop DELACK timer and fix flags */ + tc0->flags &= + ~(TCP_CONN_SNDACK | TCP_CONN_DELACK | TCP_CONN_BURSTACK); + if (tcp_timer_is_active (tc0, TCP_TIMER_DELACK)) + { + tcp_timer_reset (tc0, TCP_TIMER_DELACK); + } + + /* If not retransmitting + * 1) update snd_una_max (SYN, SYNACK, new data, FIN) + * 2) If we're not tracking an ACK, start tracking */ + if (seq_lt (tc0->snd_una_max, tc0->snd_nxt)) + { + tc0->snd_una_max = tc0->snd_nxt; + if (tc0->rtt_ts == 0) + { + tc0->rtt_ts = tcp_time_now (); + tc0->rtt_seq = tc0->snd_nxt; + } + } + + /* Set the retransmit timer if not set already and not + * doing a pure ACK */ + if (!tcp_timer_is_active (tc0, TCP_TIMER_RETRANSMIT) + && tc0->snd_nxt != tc0->snd_una) + { + tcp_retransmit_timer_set (tm, tc0); + tc0->rto_boff = 0; + } + + /* set fib index to default and lookup node */ + /* XXX network virtualization (vrf/vni) */ + vnet_buffer (b0)->sw_if_index[VLIB_RX] = 0; + vnet_buffer (b0)->sw_if_index[VLIB_TX] = (u32) ~ 0; + + b0->flags |= VNET_BUFFER_LOCALLY_ORIGINATED; + + done: + b0->error = error0 != 0 ? node->errors[error0] : 0; + if (PREDICT_FALSE (b0->flags & VLIB_BUFFER_IS_TRACED)) + { + + } + + vlib_validate_buffer_enqueue_x1 (vm, node, next_index, to_next, + n_left_to_next, bi0, next0); + } + + vlib_put_next_frame (vm, node, next_index, n_left_to_next); + } + + return from_frame->n_vectors; +} + +static uword +tcp4_output (vlib_main_t * vm, vlib_node_runtime_t * node, + vlib_frame_t * from_frame) +{ + return tcp46_output_inline (vm, node, from_frame, 1 /* is_ip4 */ ); +} + +static uword +tcp6_output (vlib_main_t * vm, vlib_node_runtime_t * node, + vlib_frame_t * from_frame) +{ + return tcp46_output_inline (vm, node, from_frame, 0 /* is_ip4 */ ); +} + +VLIB_REGISTER_NODE (tcp4_output_node) = +{ + .function = tcp4_output,.name = "tcp4-output", + /* Takes a vector of packets. */ + .vector_size = sizeof (u32),.n_errors = TCP_N_ERROR,.error_strings = + tcp_error_strings,.n_next_nodes = TCP_OUTPUT_N_NEXT,.next_nodes = + { +#define _(s,n) [TCP_OUTPUT_NEXT_##s] = n, + foreach_tcp4_output_next +#undef _ + } +,.format_buffer = format_tcp_header,.format_trace = format_tcp_tx_trace,}; + +VLIB_NODE_FUNCTION_MULTIARCH (tcp4_output_node, tcp4_output) +VLIB_REGISTER_NODE (tcp6_output_node) = +{ + .function = tcp6_output,.name = "tcp6-output", + /* Takes a vector of packets. */ + .vector_size = sizeof (u32),.n_errors = TCP_N_ERROR,.error_strings = + tcp_error_strings,.n_next_nodes = TCP_OUTPUT_N_NEXT,.next_nodes = + { +#define _(s,n) [TCP_OUTPUT_NEXT_##s] = n, + foreach_tcp6_output_next +#undef _ + } +,.format_buffer = format_tcp_header,.format_trace = format_tcp_tx_trace,}; + +VLIB_NODE_FUNCTION_MULTIARCH (tcp6_output_node, tcp6_output) u32 +tcp_push_header (transport_connection_t * tconn, vlib_buffer_t * b) +{ + tcp_connection_t *tc; + + tc = (tcp_connection_t *) tconn; + tcp_push_hdr_i (tc, b, TCP_STATE_ESTABLISHED); + return 0; +} + +typedef enum _tcp_reset_next +{ + TCP_RESET_NEXT_DROP, + TCP_RESET_NEXT_IP_LOOKUP, + TCP_RESET_N_NEXT +} tcp_reset_next_t; + +#define foreach_tcp4_reset_next \ + _(DROP, "error-drop") \ + _(IP_LOOKUP, "ip4-lookup") + +#define foreach_tcp6_reset_next \ + _(DROP, "error-drop") \ + _(IP_LOOKUP, "ip6-lookup") + +static uword +tcp46_send_reset_inline (vlib_main_t * vm, vlib_node_runtime_t * node, + vlib_frame_t * from_frame, u8 is_ip4) +{ + u32 n_left_from, next_index, *from, *to_next; + u32 my_thread_index = vm->cpu_index; + + from = vlib_frame_vector_args (from_frame); + n_left_from = from_frame->n_vectors; + + next_index = node->cached_next_index; + + while (n_left_from > 0) + { + u32 n_left_to_next; + + vlib_get_next_frame (vm, node, next_index, to_next, n_left_to_next); + + while (n_left_from > 0 && n_left_to_next > 0) + { + u32 bi0; + vlib_buffer_t *b0; + u32 error0 = TCP_ERROR_RST_SENT, next0 = TCP_RESET_NEXT_IP_LOOKUP; + + bi0 = from[0]; + to_next[0] = bi0; + from += 1; + to_next += 1; + n_left_from -= 1; + n_left_to_next -= 1; + + b0 = vlib_get_buffer (vm, bi0); + + if (tcp_make_reset_in_place (vm, b0, vnet_buffer (b0)->tcp.flags, + my_thread_index, is_ip4)) + { + error0 = TCP_ERROR_LOOKUP_DROPS; + next0 = TCP_RESET_NEXT_DROP; + goto done; + } + + /* Prepare to send to IP lookup */ + vnet_buffer (b0)->sw_if_index[VLIB_TX] = 0; + next0 = TCP_RESET_NEXT_IP_LOOKUP; + + done: + b0->error = error0 != 0 ? node->errors[error0] : 0; + b0->flags |= VNET_BUFFER_LOCALLY_ORIGINATED; + if (PREDICT_FALSE (b0->flags & VLIB_BUFFER_IS_TRACED)) + { + + } + + vlib_validate_buffer_enqueue_x1 (vm, node, next_index, to_next, + n_left_to_next, bi0, next0); + } + vlib_put_next_frame (vm, node, next_index, n_left_to_next); + } + return from_frame->n_vectors; +} + +static uword +tcp4_send_reset (vlib_main_t * vm, vlib_node_runtime_t * node, + vlib_frame_t * from_frame) +{ + return tcp46_send_reset_inline (vm, node, from_frame, 1); +} + +static uword +tcp6_send_reset (vlib_main_t * vm, vlib_node_runtime_t * node, + vlib_frame_t * from_frame) +{ + return tcp46_send_reset_inline (vm, node, from_frame, 0); +} + +/* *INDENT-OFF* */ +VLIB_REGISTER_NODE (tcp4_reset_node) = { + .function = tcp4_send_reset, + .name = "tcp4-reset", + .vector_size = sizeof (u32), + .n_errors = TCP_N_ERROR, + .error_strings = tcp_error_strings, + .n_next_nodes = TCP_RESET_N_NEXT, + .next_nodes = { +#define _(s,n) [TCP_RESET_NEXT_##s] = n, + foreach_tcp4_reset_next +#undef _ + }, +}; +/* *INDENT-ON* */ + +/* *INDENT-OFF* */ +VLIB_REGISTER_NODE (tcp6_reset_node) = { + .function = tcp6_send_reset, + .name = "tcp6-reset", + .vector_size = sizeof (u32), + .n_errors = TCP_N_ERROR, + .error_strings = tcp_error_strings, + .n_next_nodes = TCP_RESET_N_NEXT, + .next_nodes = { +#define _(s,n) [TCP_RESET_NEXT_##s] = n, + foreach_tcp6_reset_next +#undef _ + }, +}; +/* *INDENT-ON* */ + +/* + * fd.io coding-style-patch-verification: ON + * + * Local Variables: + * eval: (c-set-style "gnu") + * End: + */ diff --git a/src/vnet/tcp/tcp_packet.h b/src/vnet/tcp/tcp_packet.h new file mode 100644 index 00000000..866c5fd6 --- /dev/null +++ b/src/vnet/tcp/tcp_packet.h @@ -0,0 +1,184 @@ +/* + * Copyright (c) 2016 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +#ifndef included_tcp_packet_h +#define included_tcp_packet_h + +#include + +/* TCP flags bit 0 first. */ +#define foreach_tcp_flag \ + _ (FIN) /**< No more data from sender. */ \ + _ (SYN) /**< Synchronize sequence numbers. */ \ + _ (RST) /**< Reset the connection. */ \ + _ (PSH) /**< Push function. */ \ + _ (ACK) /**< Ack field significant. */ \ + _ (URG) /**< Urgent pointer field significant. */ \ + _ (ECE) /**< ECN-echo. Receiver got CE packet */ \ + _ (CWR) /**< Sender reduced congestion window */ + +enum +{ +#define _(f) TCP_FLAG_BIT_##f, + foreach_tcp_flag +#undef _ + TCP_N_FLAG_BITS, +}; + +enum +{ +#define _(f) TCP_FLAG_##f = 1 << TCP_FLAG_BIT_##f, + foreach_tcp_flag +#undef _ +}; + +typedef struct _tcp_header +{ + union + { + struct + { + u16 src_port; /**< Source port. */ + u16 dst_port; /**< Destination port. */ + }; + struct + { + u16 src, dst; + }; + }; + + u32 seq_number; /**< Sequence number of the first data octet in this + * segment, except when SYN is present. If SYN + * is present the seq number is is the ISN and the + * first data octet is ISN+1 */ + u32 ack_number; /**< Acknowledgement number if ACK is set. It contains + * the value of the next sequence number the sender + * of the segment is expecting to receive. */ + u8 data_offset_and_reserved; + u8 flags; /**< Flags: see the macro above */ + u16 window; /**< Number of bytes sender is willing to receive. */ + + u16 checksum; /**< Checksum of TCP pseudo header and data. */ + u16 urgent_pointer; /**< Seq number of the byte after the urgent data. */ +} __attribute__ ((packed)) tcp_header_t; + +/* Flag tests that return 0 or !0 */ +#define tcp_doff(_th) ((_th)->data_offset_and_reserved >> 4) +#define tcp_fin(_th) ((_th)->flags & TCP_FLAG_FIN) +#define tcp_syn(_th) ((_th)->flags & TCP_FLAG_SYN) +#define tcp_rst(_th) ((_th)->flags & TCP_FLAG_RST) +#define tcp_psh(_th) ((_th)->flags & TCP_FLAG_PSH) +#define tcp_ack(_th) ((_th)->flags & TCP_FLAG_ACK) +#define tcp_urg(_th) ((_th)->flags & TCP_FLAG_URG) +#define tcp_ece(_th) ((_th)->flags & TCP_FLAG_ECE) +#define tcp_cwr(_th) ((_th)->flags & TCP_FLAG_CWR) + +/* Flag tests that return 0 or 1 */ +#define tcp_is_syn(_th) !!((_th)->flags & TCP_FLAG_SYN) +#define tcp_is_fin(_th) !!((_th)->flags & TCP_FLAG_FIN) + +always_inline int +tcp_header_bytes (tcp_header_t * t) +{ + return tcp_doff (t) * sizeof (u32); +} + +/* + * TCP options. + */ + +typedef enum tcp_option_type +{ + TCP_OPTION_EOL = 0, /**< End of options. */ + TCP_OPTION_NOOP = 1, /**< No operation. */ + TCP_OPTION_MSS = 2, /**< Limit MSS. */ + TCP_OPTION_WINDOW_SCALE = 3, /**< Window scale. */ + TCP_OPTION_SACK_PERMITTED = 4, /**< Selective Ack permitted. */ + TCP_OPTION_SACK_BLOCK = 5, /**< Selective Ack block. */ + TCP_OPTION_TIMESTAMP = 8, /**< Timestamps. */ + TCP_OPTION_UTO = 28, /**< User timeout. */ + TCP_OPTION_AO = 29, /**< Authentication Option. */ +} tcp_option_type_t; + +#define foreach_tcp_options_flag \ + _ (MSS) /**< MSS advertised in SYN */ \ + _ (TSTAMP) /**< Timestamp capability advertised in SYN */ \ + _ (WSCALE) /**< Wnd scale capability advertised in SYN */ \ + _ (SACK_PERMITTED) /**< SACK capability advertised in SYN */ \ + _ (SACK) /**< SACK present */ + +enum +{ +#define _(f) TCP_OPTS_FLAG_BIT_##f, + foreach_tcp_options_flag +#undef _ + TCP_OPTIONS_N_FLAG_BITS, +}; + +enum +{ +#define _(f) TCP_OPTS_FLAG_##f = 1 << TCP_OPTS_FLAG_BIT_##f, + foreach_tcp_options_flag +#undef _ +}; + +typedef struct _sack_block +{ + u32 start; /**< Start sequence number */ + u32 end; /**< End sequence number */ +} sack_block_t; + +typedef struct +{ + u8 flags; /** Option flags, see above */ + + /* Received options */ + u16 mss; /**< Maximum segment size advertised by peer */ + u8 wscale; /**< Window scale advertised by peer */ + u32 tsval; /**< Peer's timestamp value */ + u32 tsecr; /**< Echoed/reflected time stamp */ + sack_block_t *sacks; /**< SACK blocks received */ + u8 n_sack_blocks; /**< Number of SACKs blocks */ +} tcp_options_t; + +/* Flag tests that return 0 or !0 */ +#define tcp_opts_mss(_to) ((_to)->flags & TCP_OPTS_FLAG_MSS) +#define tcp_opts_tstamp(_to) ((_to)->flags & TCP_OPTS_FLAG_TSTAMP) +#define tcp_opts_wscale(_to) ((_to)->flags & TCP_OPTS_FLAG_WSCALE) +#define tcp_opts_sack(_to) ((_to)->flags & TCP_OPTS_FLAG_SACK) +#define tcp_opts_sack_permitted(_to) ((_to)->flags & TCP_OPTS_FLAG_SACK_PERMITTED) + +/* TCP option lengths */ +#define TCP_OPTION_LEN_EOL 1 +#define TCP_OPTION_LEN_NOOP 1 +#define TCP_OPTION_LEN_MSS 4 +#define TCP_OPTION_LEN_WINDOW_SCALE 3 +#define TCP_OPTION_LEN_SACK_PERMITTED 2 +#define TCP_OPTION_LEN_TIMESTAMP 10 +#define TCP_OPTION_LEN_SACK_BLOCK 8 + +#define TCP_WND_MAX 65535U +#define TCP_MAX_WND_SCALE 14 /* See RFC 1323 */ +#define TCP_OPTS_ALIGN 4 +#define TCP_OPTS_MAX_SACK_BLOCKS 3 +#endif /* included_tcp_packet_h */ + +/* + * fd.io coding-style-patch-verification: ON + * + * Local Variables: + * eval: (c-set-style "gnu") + * End: + */ diff --git a/src/vnet/tcp/tcp_pg.c b/src/vnet/tcp/tcp_pg.c new file mode 100644 index 00000000..dc324049 --- /dev/null +++ b/src/vnet/tcp/tcp_pg.c @@ -0,0 +1,236 @@ +/* + * Copyright (c) 2016 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ +/* + * ip/tcp_pg: TCP packet-generator interface + * + * Copyright (c) 2008 Eliot Dresselhaus + * + * Permission is hereby granted, free of charge, to any person obtaining + * a copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sublicense, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice shall be + * included in all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND + * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE + * LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + * OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + * WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + */ + +#include +#include + +/* TCP flags bit 0 first. */ +#define foreach_tcp_flag \ + _ (FIN) \ + _ (SYN) \ + _ (RST) \ + _ (PSH) \ + _ (ACK) \ + _ (URG) \ + _ (ECE) \ + _ (CWR) + +static void +tcp_pg_edit_function (pg_main_t * pg, + pg_stream_t * s, + pg_edit_group_t * g, + u32 * packets, + u32 n_packets) +{ + vlib_main_t * vm = vlib_get_main(); + u32 ip_offset, tcp_offset; + + tcp_offset = g->start_byte_offset; + ip_offset = (g-1)->start_byte_offset; + + while (n_packets >= 1) + { + vlib_buffer_t * p0; + ip4_header_t * ip0; + tcp_header_t * tcp0; + ip_csum_t sum0; + u32 tcp_len0; + + p0 = vlib_get_buffer (vm, packets[0]); + n_packets -= 1; + packets += 1; + + ASSERT (p0->current_data == 0); + ip0 = (void *) (p0->data + ip_offset); + tcp0 = (void *) (p0->data + tcp_offset); + tcp_len0 = clib_net_to_host_u16 (ip0->length) - sizeof (ip0[0]); + + /* Initialize checksum with header. */ + if (BITS (sum0) == 32) + { + sum0 = clib_mem_unaligned (&ip0->src_address, u32); + sum0 = ip_csum_with_carry (sum0, clib_mem_unaligned (&ip0->dst_address, u32)); + } + else + sum0 = clib_mem_unaligned (&ip0->src_address, u64); + + sum0 = ip_csum_with_carry + (sum0, clib_host_to_net_u32 (tcp_len0 + (ip0->protocol << 16))); + + /* Invalidate possibly old checksum. */ + tcp0->checksum = 0; + + sum0 = ip_incremental_checksum_buffer (vm, p0, tcp_offset, tcp_len0, sum0); + + tcp0->checksum = ~ ip_csum_fold (sum0); + } +} + +typedef struct { + pg_edit_t src, dst; + pg_edit_t seq_number, ack_number; + pg_edit_t data_offset_and_reserved; +#define _(f) pg_edit_t f##_flag; + foreach_tcp_flag +#undef _ + pg_edit_t window; + pg_edit_t checksum; + pg_edit_t urgent_pointer; +} pg_tcp_header_t; + +static inline void +pg_tcp_header_init (pg_tcp_header_t * p) +{ + /* Initialize fields that are not bit fields in the IP header. */ +#define _(f) pg_edit_init (&p->f, tcp_header_t, f); + _ (src); + _ (dst); + _ (seq_number); + _ (ack_number); + _ (window); + _ (checksum); + _ (urgent_pointer); +#undef _ + + /* Initialize bit fields. */ +#define _(f) \ + pg_edit_init_bitfield (&p->f##_flag, tcp_header_t, \ + flags, \ + TCP_FLAG_BIT_##f, 1); + + foreach_tcp_flag +#undef _ + + pg_edit_init_bitfield (&p->data_offset_and_reserved, tcp_header_t, + data_offset_and_reserved, + 4, 4); +} + +uword +unformat_pg_tcp_header (unformat_input_t * input, va_list * args) +{ + pg_stream_t * s = va_arg (*args, pg_stream_t *); + pg_tcp_header_t * p; + u32 group_index; + + p = pg_create_edit_group (s, sizeof (p[0]), sizeof (tcp_header_t), + &group_index); + pg_tcp_header_init (p); + + /* Defaults. */ + pg_edit_set_fixed (&p->seq_number, 0); + pg_edit_set_fixed (&p->ack_number, 0); + + pg_edit_set_fixed (&p->data_offset_and_reserved, + sizeof (tcp_header_t) / sizeof (u32)); + + pg_edit_set_fixed (&p->window, 4096); + pg_edit_set_fixed (&p->urgent_pointer, 0); + +#define _(f) pg_edit_set_fixed (&p->f##_flag, 0); + foreach_tcp_flag +#undef _ + + p->checksum.type = PG_EDIT_UNSPECIFIED; + + if (! unformat (input, "TCP: %U -> %U", + unformat_pg_edit, + unformat_tcp_udp_port, &p->src, + unformat_pg_edit, + unformat_tcp_udp_port, &p->dst)) + goto error; + + /* Parse options. */ + while (1) + { + if (unformat (input, "window %U", + unformat_pg_edit, + unformat_pg_number, &p->window)) + ; + + else if (unformat (input, "checksum %U", + unformat_pg_edit, + unformat_pg_number, &p->checksum)) + ; + + /* Flags. */ +#define _(f) else if (unformat (input, #f)) pg_edit_set_fixed (&p->f##_flag, 1); + foreach_tcp_flag +#undef _ + + /* Can't parse input: try next protocol level. */ + else + break; + } + + { + ip_main_t * im = &ip_main; + u16 dst_port; + tcp_udp_port_info_t * pi; + + pi = 0; + if (p->dst.type == PG_EDIT_FIXED) + { + dst_port = pg_edit_get_value (&p->dst, PG_EDIT_LO); + pi = ip_get_tcp_udp_port_info (im, dst_port); + } + + if (pi && pi->unformat_pg_edit + && unformat_user (input, pi->unformat_pg_edit, s)) + ; + + else if (! unformat_user (input, unformat_pg_payload, s)) + goto error; + + if (p->checksum.type == PG_EDIT_UNSPECIFIED) + { + pg_edit_group_t * g = pg_stream_get_group (s, group_index); + g->edit_function = tcp_pg_edit_function; + g->edit_function_opaque = 0; + } + + return 1; + } + + error: + /* Free up any edits we may have added. */ + pg_free_edit_group (s); + return 0; +} + diff --git a/src/vnet/tcp/tcp_syn_filter4.c b/src/vnet/tcp/tcp_syn_filter4.c new file mode 100644 index 00000000..c7605a30 --- /dev/null +++ b/src/vnet/tcp/tcp_syn_filter4.c @@ -0,0 +1,542 @@ +/* + * Copyright (c) 2016 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +#include +#include +#include +#include +#include +#include +#include + +typedef struct +{ + f64 next_reset; + f64 reset_interval; + u8 *syn_counts; +} syn_filter4_runtime_t; + +typedef struct +{ + u32 next_index; + int not_a_syn; + u8 filter_value; +} syn_filter4_trace_t; + +/* packet trace format function */ +static u8 * +format_syn_filter4_trace (u8 * s, va_list * args) +{ + CLIB_UNUSED (vlib_main_t * vm) = va_arg (*args, vlib_main_t *); + CLIB_UNUSED (vlib_node_t * node) = va_arg (*args, vlib_node_t *); + syn_filter4_trace_t *t = va_arg (*args, syn_filter4_trace_t *); + + s = format (s, "SYN_FILTER4: next index %d, %s", + t->next_index, t->not_a_syn ? "not a syn" : "syn"); + if (t->not_a_syn == 0) + s = format (s, ", filter value %d\n", t->filter_value); + else + s = format (s, "\n"); + return s; +} + +static vlib_node_registration_t syn_filter4_node; + +#define foreach_syn_filter_error \ +_(THROTTLED, "TCP SYN packet throttle drops") \ +_(OK, "TCP SYN packets passed") + +typedef enum +{ +#define _(sym,str) SYN_FILTER_ERROR_##sym, + foreach_syn_filter_error +#undef _ + SYN_FILTER_N_ERROR, +} syn_filter_error_t; + +static char *syn_filter4_error_strings[] = { +#define _(sym,string) string, + foreach_syn_filter_error +#undef _ +}; + +typedef enum +{ + SYN_FILTER_NEXT_DROP, + SYN_FILTER_N_NEXT, +} syn_filter_next_t; + +extern vnet_feature_arc_registration_t vnet_feat_arc_ip4_local; + +static uword +syn_filter4_node_fn (vlib_main_t * vm, + vlib_node_runtime_t * node, vlib_frame_t * frame) +{ + u32 n_left_from, *from, *to_next; + syn_filter_next_t next_index; + u32 ok_syn_packets = 0; + vnet_feature_main_t *fm = &feature_main; + u8 arc_index = vnet_feat_arc_ip4_local.feature_arc_index; + vnet_feature_config_main_t *cm = &fm->feature_config_mains[arc_index]; + syn_filter4_runtime_t *rt = (syn_filter4_runtime_t *) node->runtime_data; + f64 now = vlib_time_now (vm); + /* Shut up spurious gcc warnings. */ + u8 *c0 = 0, *c1 = 0, *c2 = 0, *c3 = 0; + + from = vlib_frame_vector_args (frame); + n_left_from = frame->n_vectors; + next_index = node->cached_next_index; + + if (now > rt->next_reset) + { + memset (rt->syn_counts, 0, vec_len (rt->syn_counts)); + rt->next_reset = now + rt->reset_interval; + } + + while (n_left_from > 0) + { + u32 n_left_to_next; + + vlib_get_next_frame (vm, node, next_index, to_next, n_left_to_next); + + while (n_left_from >= 8 && n_left_to_next >= 4) + { + u32 bi0, bi1, bi2, bi3; + vlib_buffer_t *b0, *b1, *b2, *b3; + u32 next0, next1, next2, next3; + ip4_header_t *ip0, *ip1, *ip2, *ip3; + tcp_header_t *tcp0, *tcp1, *tcp2, *tcp3; + u32 not_a_syn0 = 1, not_a_syn1 = 1, not_a_syn2 = 1, not_a_syn3 = 1; + u64 hash0, hash1, hash2, hash3; + + /* Prefetch next iteration. */ + { + vlib_buffer_t *p4, *p5, *p6, *p7; + + p4 = vlib_get_buffer (vm, from[4]); + p5 = vlib_get_buffer (vm, from[5]); + p6 = vlib_get_buffer (vm, from[6]); + p7 = vlib_get_buffer (vm, from[7]); + + vlib_prefetch_buffer_header (p4, LOAD); + vlib_prefetch_buffer_header (p5, LOAD); + vlib_prefetch_buffer_header (p6, LOAD); + vlib_prefetch_buffer_header (p7, LOAD); + + CLIB_PREFETCH (p4->data, CLIB_CACHE_LINE_BYTES, STORE); + CLIB_PREFETCH (p5->data, CLIB_CACHE_LINE_BYTES, STORE); + CLIB_PREFETCH (p6->data, CLIB_CACHE_LINE_BYTES, STORE); + CLIB_PREFETCH (p7->data, CLIB_CACHE_LINE_BYTES, STORE); + } + + /* speculatively enqueue b0 and b1 to the current next frame */ + to_next[0] = bi0 = from[0]; + to_next[1] = bi1 = from[1]; + to_next[2] = bi2 = from[2]; + to_next[3] = bi3 = from[3]; + from += 4; + to_next += 4; + n_left_from -= 4; + n_left_to_next -= 4; + + b0 = vlib_get_buffer (vm, bi0); + b1 = vlib_get_buffer (vm, bi1); + b2 = vlib_get_buffer (vm, bi2); + b3 = vlib_get_buffer (vm, bi3); + + vnet_get_config_data + (&cm->config_main, &b0->current_config_index, + &next0, 0 /* sizeof (c0[0]) */ ); + vnet_get_config_data + (&cm->config_main, &b1->current_config_index, + &next1, 0 /* sizeof (c0[0]) */ ); + vnet_get_config_data + (&cm->config_main, &b2->current_config_index, + &next2, 0 /* sizeof (c0[0]) */ ); + vnet_get_config_data + (&cm->config_main, &b3->current_config_index, + &next3, 0 /* sizeof (c0[0]) */ ); + + /* Not TCP? */ + ip0 = vlib_buffer_get_current (b0); + if (ip0->protocol != IP_PROTOCOL_TCP) + goto trace00; + + tcp0 = ip4_next_header (ip0); + /* + * Not a SYN? + * $$$$ hack: the TCP bitfield flags seem not to compile + * correct code. + */ + if (PREDICT_TRUE (!(tcp0->flags & 0x2))) + goto trace00; + + not_a_syn0 = 0; + hash0 = clib_xxhash ((u64) ip0->src_address.as_u32); + c0 = &rt->syn_counts[hash0 & (_vec_len (rt->syn_counts) - 1)]; + if (PREDICT_FALSE (*c0 >= 0x80)) + { + next0 = SYN_FILTER_NEXT_DROP; + b0->error = node->errors[SYN_FILTER_ERROR_THROTTLED]; + goto trace00; + } + *c0 += 1; + ok_syn_packets++; + + trace00: + if (PREDICT_FALSE ((node->flags & VLIB_NODE_FLAG_TRACE) + && (b0->flags & VLIB_BUFFER_IS_TRACED))) + { + syn_filter4_trace_t *t = + vlib_add_trace (vm, node, b0, sizeof (*t)); + t->not_a_syn = not_a_syn0; + t->next_index = next0; + t->filter_value = not_a_syn0 ? 0 : *c0; + } + + /* Not TCP? */ + ip1 = vlib_buffer_get_current (b1); + if (ip1->protocol != IP_PROTOCOL_TCP) + goto trace01; + + tcp1 = ip4_next_header (ip1); + /* + * Not a SYN? + * $$$$ hack: the TCP bitfield flags seem not to compile + * correct code. + */ + if (PREDICT_TRUE (!(tcp1->flags & 0x2))) + goto trace01; + + not_a_syn1 = 0; + hash1 = clib_xxhash ((u64) ip1->src_address.as_u32); + c1 = &rt->syn_counts[hash1 & (_vec_len (rt->syn_counts) - 1)]; + if (PREDICT_FALSE (*c1 >= 0x80)) + { + next1 = SYN_FILTER_NEXT_DROP; + b1->error = node->errors[SYN_FILTER_ERROR_THROTTLED]; + goto trace01; + } + *c1 += 1; + ok_syn_packets++; + + trace01: + if (PREDICT_FALSE ((node->flags & VLIB_NODE_FLAG_TRACE) + && (b1->flags & VLIB_BUFFER_IS_TRACED))) + { + syn_filter4_trace_t *t = + vlib_add_trace (vm, node, b1, sizeof (*t)); + t->not_a_syn = not_a_syn1; + t->next_index = next1; + t->filter_value = not_a_syn1 ? 0 : *c1; + } + + /* Not TCP? */ + ip2 = vlib_buffer_get_current (b2); + if (ip2->protocol != IP_PROTOCOL_TCP) + goto trace02; + + tcp2 = ip4_next_header (ip2); + /* + * Not a SYN? + * $$$$ hack: the TCP bitfield flags seem not to compile + * correct code. + */ + if (PREDICT_TRUE (!(tcp2->flags & 0x2))) + goto trace02; + + not_a_syn2 = 0; + hash2 = clib_xxhash ((u64) ip2->src_address.as_u32); + c2 = &rt->syn_counts[hash2 & (_vec_len (rt->syn_counts) - 1)]; + if (PREDICT_FALSE (*c2 >= 0x80)) + { + next2 = SYN_FILTER_NEXT_DROP; + b2->error = node->errors[SYN_FILTER_ERROR_THROTTLED]; + goto trace02; + } + *c2 += 1; + ok_syn_packets++; + + trace02: + if (PREDICT_FALSE ((node->flags & VLIB_NODE_FLAG_TRACE) + && (b2->flags & VLIB_BUFFER_IS_TRACED))) + { + syn_filter4_trace_t *t = + vlib_add_trace (vm, node, b2, sizeof (*t)); + t->not_a_syn = not_a_syn2; + t->next_index = next2; + t->filter_value = not_a_syn2 ? 0 : *c2; + } + + /* Not TCP? */ + ip3 = vlib_buffer_get_current (b3); + if (ip3->protocol != IP_PROTOCOL_TCP) + goto trace03; + + tcp3 = ip4_next_header (ip3); + /* + * Not a SYN? + * $$$$ hack: the TCP bitfield flags seem not to compile + * correct code. + */ + if (PREDICT_TRUE (!(tcp3->flags & 0x2))) + goto trace03; + + not_a_syn3 = 0; + hash3 = clib_xxhash ((u64) ip3->src_address.as_u32); + c3 = &rt->syn_counts[hash3 & (_vec_len (rt->syn_counts) - 1)]; + if (PREDICT_FALSE (*c3 >= 0x80)) + { + next3 = SYN_FILTER_NEXT_DROP; + b3->error = node->errors[SYN_FILTER_ERROR_THROTTLED]; + goto trace03; + } + *c3 += 1; + ok_syn_packets++; + + trace03: + if (PREDICT_FALSE ((node->flags & VLIB_NODE_FLAG_TRACE) + && (b3->flags & VLIB_BUFFER_IS_TRACED))) + { + syn_filter4_trace_t *t = + vlib_add_trace (vm, node, b3, sizeof (*t)); + t->not_a_syn = not_a_syn3; + t->next_index = next3; + t->filter_value = not_a_syn3 ? 0 : *c3; + } + vlib_validate_buffer_enqueue_x4 (vm, node, next_index, + to_next, n_left_to_next, + bi0, bi1, bi2, bi3, + next0, next1, next2, next3); + } + + while (n_left_from > 0 && n_left_to_next > 0) + { + u32 bi0; + vlib_buffer_t *b0; + u32 next0; + ip4_header_t *ip0; + tcp_header_t *tcp0; + u32 not_a_syn0 = 1; + u32 hash0; + u8 *c0; + + /* speculatively enqueue b0 to the current next frame */ + bi0 = from[0]; + to_next[0] = bi0; + from += 1; + to_next += 1; + n_left_from -= 1; + n_left_to_next -= 1; + + b0 = vlib_get_buffer (vm, bi0); + + vnet_get_config_data + (&cm->config_main, &b0->current_config_index, + &next0, 0 /* sizeof (c0[0]) */ ); + + /* Not TCP? */ + ip0 = vlib_buffer_get_current (b0); + if (ip0->protocol != IP_PROTOCOL_TCP) + goto trace0; + + tcp0 = ip4_next_header (ip0); + /* + * Not a SYN? + * $$$$ hack: the TCP bitfield flags seem not to compile + * correct code. + */ + if (PREDICT_TRUE (!(tcp0->flags & 0x2))) + goto trace0; + + not_a_syn0 = 0; + hash0 = clib_xxhash ((u64) ip0->src_address.as_u32); + c0 = &rt->syn_counts[hash0 & (_vec_len (rt->syn_counts) - 1)]; + if (PREDICT_FALSE (*c0 >= 0x80)) + { + next0 = SYN_FILTER_NEXT_DROP; + b0->error = node->errors[SYN_FILTER_ERROR_THROTTLED]; + goto trace0; + } + *c0 += 1; + ok_syn_packets++; + + trace0: + + if (PREDICT_FALSE ((node->flags & VLIB_NODE_FLAG_TRACE) + && (b0->flags & VLIB_BUFFER_IS_TRACED))) + { + syn_filter4_trace_t *t = + vlib_add_trace (vm, node, b0, sizeof (*t)); + t->not_a_syn = not_a_syn0; + t->next_index = next0; + t->filter_value = not_a_syn0 ? 0 : *c0; + } + + /* verify speculative enqueue, maybe switch current next frame */ + vlib_validate_buffer_enqueue_x1 (vm, node, next_index, + to_next, n_left_to_next, + bi0, next0); + } + + vlib_put_next_frame (vm, node, next_index, n_left_to_next); + } + + vlib_node_increment_counter (vm, syn_filter4_node.index, + SYN_FILTER_ERROR_OK, ok_syn_packets); + return frame->n_vectors; +} + +/* *INDENT-OFF* */ +VLIB_REGISTER_NODE (syn_filter4_node, static) = +{ + .function = syn_filter4_node_fn, + .name = "syn-filter-4", + .vector_size = sizeof (u32), + .format_trace = format_syn_filter4_trace, + .type = VLIB_NODE_TYPE_INTERNAL, + + .runtime_data_bytes = sizeof (syn_filter4_runtime_t), + .n_errors = ARRAY_LEN(syn_filter4_error_strings), + .error_strings = syn_filter4_error_strings, + + .n_next_nodes = SYN_FILTER_N_NEXT, + + /* edit / add dispositions here */ + .next_nodes = { + [SYN_FILTER_NEXT_DROP] = "error-drop", + }, +}; +/* *INDENT-ON* */ + +VLIB_NODE_FUNCTION_MULTIARCH (syn_filter4_node, syn_filter4_node_fn); + +/* *INDENT-OFF* */ +VNET_FEATURE_INIT (syn_filter_4, static) = +{ + .arc_name = "ip4-local", + .node_name = "syn-filter-4", + .runs_before = VNET_FEATURES("ip4-local-end-of-arc"), +}; +/* *INDENT-ON* */ + +int +syn_filter_enable_disable (u32 sw_if_index, int enable_disable) +{ + vnet_main_t *vnm = vnet_get_main (); + vnet_sw_interface_t *sw; + int rv = 0; + + /* Utterly wrong? */ + if (pool_is_free_index (vnm->interface_main.sw_interfaces, sw_if_index)) + return VNET_API_ERROR_INVALID_SW_IF_INDEX; + + /* Not a physical port? */ + sw = vnet_get_sw_interface (vnm, sw_if_index); + if (sw->type != VNET_SW_INTERFACE_TYPE_HARDWARE) + return VNET_API_ERROR_INVALID_SW_IF_INDEX; + + if (enable_disable) + { + vlib_main_t *vm = vlib_get_main (); + syn_filter4_runtime_t *rt; + + rt = vlib_node_get_runtime_data (vm, syn_filter4_node.index); + vec_validate (rt->syn_counts, 1023); + /* + * Given perfect disperson / optimal hashing results: + * Allow 128k (successful) syns/sec. 1024, buckets each of which + * absorb 128 syns before filtering. Reset table once a second. + * Reality bites, lets try resetting once every 100ms. + */ + rt->reset_interval = 0.1; /* reset interval in seconds */ + } + + rv = vnet_feature_enable_disable ("ip4-local", "syn-filter-4", + sw_if_index, enable_disable, 0, 0); + + return rv; +} + +static clib_error_t * +syn_filter_enable_disable_command_fn (vlib_main_t * vm, + unformat_input_t * input, + vlib_cli_command_t * cmd) +{ + vnet_main_t *vnm = vnet_get_main (); + u32 sw_if_index = ~0; + int enable_disable = 1; + int rv; + + while (unformat_check_input (input) != UNFORMAT_END_OF_INPUT) + { + if (unformat (input, "disable")) + enable_disable = 0; + else if (unformat (input, "%U", unformat_vnet_sw_interface, + vnm, &sw_if_index)) + ; + else + break; + } + + if (sw_if_index == ~0) + return clib_error_return (0, "Please specify an interface..."); + + rv = syn_filter_enable_disable (sw_if_index, enable_disable); + + switch (rv) + { + case 0: + break; + + case VNET_API_ERROR_INVALID_SW_IF_INDEX: + return clib_error_return + (0, "Invalid interface, only works on physical ports"); + break; + + case VNET_API_ERROR_UNIMPLEMENTED: + return clib_error_return (0, + "Device driver doesn't support redirection"); + break; + + case VNET_API_ERROR_INVALID_VALUE: + return clib_error_return (0, "feature arc not found"); + + case VNET_API_ERROR_INVALID_VALUE_2: + return clib_error_return (0, "feature node not found"); + + default: + return clib_error_return (0, "syn_filter_enable_disable returned %d", + rv); + } + return 0; +} + +/* *INDENT-OFF* */ +VLIB_CLI_COMMAND (sr_content_command, static) = +{ + .path = "ip syn filter", + .short_help = "ip syn filter [disable]", + .function = syn_filter_enable_disable_command_fn, +}; +/* *INDENT-ON* */ + +/* + * fd.io coding-style-patch-verification: ON + * + * Local Variables: + * eval: (c-set-style "gnu") + * End: + */ diff --git a/src/vnet/tcp/tcp_timer.h b/src/vnet/tcp/tcp_timer.h new file mode 100644 index 00000000..fa25268c --- /dev/null +++ b/src/vnet/tcp/tcp_timer.h @@ -0,0 +1,29 @@ +/* + * Copyright (c) 2016 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ +#ifndef __included_tcp_timer_h__ +#define __included_tcp_timer_h__ + +#include +#include + +#endif /* __included_tcp_timer_h__ */ + +/* + * fd.io coding-style-patch-verification: ON + * + * Local Variables: + * eval: (c-set-style "gnu") + * End: + */ diff --git a/src/vnet/udp/builtin_server.c b/src/vnet/udp/builtin_server.c new file mode 100644 index 00000000..afa66ba4 --- /dev/null +++ b/src/vnet/udp/builtin_server.c @@ -0,0 +1,239 @@ +/* + * Copyright (c) 2016 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +/** @file + udp builtin server +*/ + +#include +#include +#include + +/** per-worker built-in server copy buffers */ +u8 **copy_buffers; + +static int +builtin_session_create_callback (stream_session_t * s) +{ + /* Simple version: declare session ready-to-go... */ + s->session_state = SESSION_STATE_READY; + return 0; +} + +static void +builtin_session_disconnect_callback (stream_session_t * s) +{ + stream_session_disconnect (s); +} + +static int +builtin_server_rx_callback (stream_session_t * s) +{ + svm_fifo_t *rx_fifo, *tx_fifo; + u32 this_transfer; + int actual_transfer; + u8 *my_copy_buffer; + session_fifo_event_t evt; + unix_shared_memory_queue_t *q; + + my_copy_buffer = copy_buffers[s->thread_index]; + rx_fifo = s->server_rx_fifo; + tx_fifo = s->server_tx_fifo; + + this_transfer = svm_fifo_max_enqueue (tx_fifo) + < svm_fifo_max_dequeue (rx_fifo) ? + svm_fifo_max_enqueue (tx_fifo) : svm_fifo_max_dequeue (rx_fifo); + + vec_validate (my_copy_buffer, this_transfer - 1); + _vec_len (my_copy_buffer) = this_transfer; + + actual_transfer = svm_fifo_dequeue_nowait (rx_fifo, 0, this_transfer, + my_copy_buffer); + ASSERT (actual_transfer == this_transfer); + actual_transfer = svm_fifo_enqueue_nowait (tx_fifo, 0, this_transfer, + my_copy_buffer); + + copy_buffers[s->thread_index] = my_copy_buffer; + + /* Fabricate TX event, send to ourselves */ + evt.fifo = tx_fifo; + evt.event_type = FIFO_EVENT_SERVER_TX; + /* $$$$ for event logging */ + evt.enqueue_length = actual_transfer; + evt.event_id = 0; + q = session_manager_get_vpp_event_queue (s->thread_index); + unix_shared_memory_queue_add (q, (u8 *) & evt, 0 /* do wait for mutex */ ); + + return 0; +} + +/* *INDENT-OFF* */ +static session_cb_vft_t builtin_server = { + .session_accept_callback = builtin_session_create_callback, + .session_disconnect_callback = builtin_session_disconnect_callback, + .builtin_server_rx_callback = builtin_server_rx_callback +}; +/* *INDENT-ON* */ + +static int +bind_builtin_uri_server (u8 * uri) +{ + vnet_bind_args_t _a, *a = &_a; + char segment_name[128]; + u32 segment_name_length; + int rv; + u64 options[16]; + + segment_name_length = ARRAY_LEN (segment_name); + + memset (a, 0, sizeof (*a)); + memset (options, 0, sizeof (options)); + + a->uri = (char *) uri; + a->api_client_index = ~0; /* built-in server */ + a->segment_name = segment_name; + a->segment_name_length = segment_name_length; + a->session_cb_vft = &builtin_server; + + options[SESSION_OPTIONS_ACCEPT_COOKIE] = 0x12345678; + options[SESSION_OPTIONS_SEGMENT_SIZE] = (2 << 30); /*$$$$ config / arg */ + a->options = options; + + rv = vnet_bind_uri (a); + + return rv; +} + +static int +unbind_builtin_uri_server (u8 * uri) +{ + int rv; + + rv = vnet_unbind_uri ((char *) uri, ~0 /* client_index */ ); + + return rv; +} + +static clib_error_t * +builtin_server_init (vlib_main_t * vm) +{ + vlib_thread_main_t *vtm = vlib_get_thread_main (); + u32 num_threads; + + num_threads = 1 /* main thread */ + vtm->n_threads; + + vec_validate (copy_buffers, num_threads - 1); + return 0; +} + +VLIB_INIT_FUNCTION (builtin_server_init); + +static clib_error_t * +builtin_uri_bind_command_fn (vlib_main_t * vm, + unformat_input_t * input, + vlib_cli_command_t * cmd) +{ + u8 *uri = 0; + int rv; + + while (unformat_check_input (input) != UNFORMAT_END_OF_INPUT) + { + if (unformat (input, "uri %s", &uri)) + ; + else + break; + } + + if (uri == 0) + return clib_error_return (0, "uri to bind not specified..."); + + rv = bind_builtin_uri_server (uri); + + vec_free (uri); + + switch (rv) + { + case 0: + break; + + default: + return clib_error_return (0, "bind_uri_server returned %d", rv); + break; + } + + return 0; +} + +/* *INDENT-OFF* */ +VLIB_CLI_COMMAND (builtin_uri_bind_command, static) = +{ + .path = "builtin uri bind", + .short_help = "builtin uri bind", + .function = builtin_uri_bind_command_fn, +}; +/* *INDENT-ON* */ + +static clib_error_t * +builtin_uri_unbind_command_fn (vlib_main_t * vm, + unformat_input_t * input, + vlib_cli_command_t * cmd) +{ + u8 *uri = 0; + int rv; + + while (unformat_check_input (input) != UNFORMAT_END_OF_INPUT) + { + if (unformat (input, "uri %s", &uri)) + ; + else + break; + } + + if (uri == 0) + return clib_error_return (0, "uri to unbind not specified..."); + + rv = unbind_builtin_uri_server (uri); + + vec_free (uri); + + switch (rv) + { + case 0: + break; + + default: + return clib_error_return (0, "unbind_uri_server returned %d", rv); + break; + } + + return 0; +} + +/* *INDENT-OFF* */ +VLIB_CLI_COMMAND (builtin_uri_unbind_command, static) = +{ + .path = "builtin uri unbind", + .short_help = "builtin uri unbind", + .function = builtin_uri_unbind_command_fn, +}; +/* *INDENT-ON* */ + +/* + * fd.io coding-style-patch-verification: ON + * + * Local Variables: + * eval: (c-set-style "gnu") + * End: + */ diff --git a/src/vnet/udp/udp.c b/src/vnet/udp/udp.c new file mode 100644 index 00000000..9e740466 --- /dev/null +++ b/src/vnet/udp/udp.c @@ -0,0 +1,342 @@ +/* + * Copyright (c) 2016 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +/** @file + udp state machine, etc. +*/ + +#include +#include +#include +#include + +udp_uri_main_t udp_uri_main; + +u32 +udp_session_bind_ip4 (vlib_main_t * vm, u32 session_index, + ip46_address_t * ip, u16 port_number_host_byte_order) +{ + udp_uri_main_t *um = vnet_get_udp_main (); + udp_connection_t *listener; + + pool_get (um->udp_listeners, listener); + memset (listener, 0, sizeof (udp_connection_t)); + listener->c_lcl_port = clib_host_to_net_u16 (port_number_host_byte_order); + listener->c_lcl_ip4.as_u32 = ip->ip4.as_u32; + listener->c_proto = SESSION_TYPE_IP4_UDP; + udp_register_dst_port (um->vlib_main, port_number_host_byte_order, + udp4_uri_input_node.index, 1 /* is_ipv4 */ ); + return 0; +} + +u32 +udp_session_bind_ip6 (vlib_main_t * vm, u32 session_index, + ip46_address_t * ip, u16 port_number_host_byte_order) +{ + udp_uri_main_t *um = vnet_get_udp_main (); + udp_connection_t *listener; + + pool_get (um->udp_listeners, listener); + listener->c_lcl_port = clib_host_to_net_u16 (port_number_host_byte_order); + clib_memcpy (&listener->c_lcl_ip6, &ip->ip6, sizeof (ip6_address_t)); + listener->c_proto = SESSION_TYPE_IP6_UDP; + udp_register_dst_port (um->vlib_main, port_number_host_byte_order, + udp4_uri_input_node.index, 0 /* is_ipv4 */ ); + return 0; +} + +u32 +udp_session_unbind_ip4 (vlib_main_t * vm, u32 listener_index) +{ + udp_connection_t *listener; + listener = udp_listener_get (listener_index); + + /* deregister the udp_local mapping */ + udp_unregister_dst_port (vm, listener->c_lcl_port, 1 /* is_ipv4 */ ); + return 0; +} + +u32 +udp_session_unbind_ip6 (vlib_main_t * vm, u32 listener_index) +{ + udp_connection_t *listener; + + listener = udp_listener_get (listener_index); + + /* deregister the udp_local mapping */ + udp_unregister_dst_port (vm, listener->c_lcl_port, 0 /* is_ipv4 */ ); + return 0; +} + +transport_connection_t * +udp_session_get_listener (u32 listener_index) +{ + udp_connection_t *us; + + us = udp_listener_get (listener_index); + return &us->connection; +} + +u32 +udp_push_header (transport_connection_t * tconn, vlib_buffer_t * b) +{ + udp_connection_t *us; + u8 *data; + udp_header_t *udp; + + us = (udp_connection_t *) tconn; + + if (tconn->is_ip4) + { + ip4_header_t *ip; + + data = vlib_buffer_get_current (b); + udp = (udp_header_t *) (data - sizeof (udp_header_t)); + ip = (ip4_header_t *) ((u8 *) udp - sizeof (ip4_header_t)); + + /* Build packet header, swap rx key src + dst fields */ + ip->src_address.as_u32 = us->c_lcl_ip4.as_u32; + ip->dst_address.as_u32 = us->c_rmt_ip4.as_u32; + ip->ip_version_and_header_length = 0x45; + ip->ttl = 254; + ip->protocol = IP_PROTOCOL_UDP; + ip->length = clib_host_to_net_u16 (b->current_length + sizeof (*udp)); + ip->checksum = ip4_header_checksum (ip); + + udp->src_port = us->c_lcl_port; + udp->dst_port = us->c_rmt_port; + udp->length = clib_host_to_net_u16 (b->current_length); + udp->checksum = 0; + + b->current_length = sizeof (*ip) + sizeof (*udp); + return SESSION_QUEUE_NEXT_IP4_LOOKUP; + } + else + { + vlib_main_t *vm = vlib_get_main (); + ip6_header_t *ip; + u16 payload_length; + int bogus = ~0; + + data = vlib_buffer_get_current (b); + udp = (udp_header_t *) (data - sizeof (udp_header_t)); + ip = (ip6_header_t *) ((u8 *) udp - sizeof (ip6_header_t)); + + /* Build packet header, swap rx key src + dst fields */ + clib_memcpy (&ip->src_address, &us->c_lcl_ip6, sizeof (ip6_address_t)); + clib_memcpy (&ip->dst_address, &us->c_rmt_ip6, sizeof (ip6_address_t)); + + ip->ip_version_traffic_class_and_flow_label = + clib_host_to_net_u32 (0x6 << 28); + + ip->hop_limit = 0xff; + ip->protocol = IP_PROTOCOL_UDP; + + payload_length = vlib_buffer_length_in_chain (vm, b); + payload_length -= sizeof (*ip); + + ip->payload_length = clib_host_to_net_u16 (payload_length); + + udp->checksum = ip6_tcp_udp_icmp_compute_checksum (vm, b, ip, &bogus); + ASSERT (!bogus); + + udp->src_port = us->c_lcl_port; + udp->dst_port = us->c_rmt_port; + udp->length = clib_host_to_net_u16 (b->current_length); + udp->checksum = 0; + + b->current_length = sizeof (*ip) + sizeof (*udp); + + return SESSION_QUEUE_NEXT_IP6_LOOKUP; + } +} + +transport_connection_t * +udp_session_get (u32 connection_index, u32 my_thread_index) +{ + udp_uri_main_t *um = vnet_get_udp_main (); + + udp_connection_t *us; + us = + pool_elt_at_index (um->udp_sessions[my_thread_index], connection_index); + return &us->connection; +} + +void +udp_session_close (u32 connection_index, u32 my_thread_index) +{ + udp_uri_main_t *um = vnet_get_udp_main (); + pool_put_index (um->udp_sessions[my_thread_index], connection_index); +} + +u8 * +format_udp_session_ip4 (u8 * s, va_list * args) +{ + u32 uci = va_arg (*args, u32); + u32 thread_index = va_arg (*args, u32); + udp_connection_t *u4; + + u4 = udp_connection_get (uci, thread_index); + + s = format (s, "[%s] %U:%d->%U:%d", "udp", format_ip4_address, + &u4->c_lcl_ip4, clib_net_to_host_u16 (u4->c_lcl_port), + format_ip4_address, &u4->c_rmt_ip4, + clib_net_to_host_u16 (u4->c_rmt_port)); + return s; +} + +u8 * +format_udp_session_ip6 (u8 * s, va_list * args) +{ + u32 uci = va_arg (*args, u32); + u32 thread_index = va_arg (*args, u32); + udp_connection_t *tc = udp_connection_get (uci, thread_index); + s = format (s, "[%s] %U:%d->%U:%d", "udp", format_ip6_address, + &tc->c_lcl_ip6, clib_net_to_host_u16 (tc->c_lcl_port), + format_ip6_address, &tc->c_rmt_ip6, + clib_net_to_host_u16 (tc->c_rmt_port)); + return s; +} + +u8 * +format_udp_listener_session_ip4 (u8 * s, va_list * args) +{ + u32 tci = va_arg (*args, u32); + udp_connection_t *tc = udp_listener_get (tci); + s = format (s, "[%s] %U:%d->%U:%d", "udp", format_ip4_address, + &tc->c_lcl_ip4, clib_net_to_host_u16 (tc->c_lcl_port), + format_ip4_address, &tc->c_rmt_ip4, + clib_net_to_host_u16 (tc->c_rmt_port)); + return s; +} + +u8 * +format_udp_listener_session_ip6 (u8 * s, va_list * args) +{ + u32 tci = va_arg (*args, u32); + udp_connection_t *tc = udp_listener_get (tci); + s = format (s, "[%s] %U:%d->%U:%d", "udp", format_ip6_address, + &tc->c_lcl_ip6, clib_net_to_host_u16 (tc->c_lcl_port), + format_ip6_address, &tc->c_rmt_ip6, + clib_net_to_host_u16 (tc->c_rmt_port)); + return s; +} + +u16 +udp_send_mss_uri (transport_connection_t * t) +{ + /* TODO figure out MTU of output interface */ + return 400; +} + +u32 +udp_send_space_uri (transport_connection_t * t) +{ + /* No constraint on TX window */ + return ~0; +} + +int +udp_open_connection (ip46_address_t * addr, u16 port) +{ + clib_warning ("Not implemented"); + return 0; +} + +/* *INDENT-OFF* */ +const static transport_proto_vft_t udp4_proto = { + .bind = udp_session_bind_ip4, + .open = udp_open_connection, + .unbind = udp_session_unbind_ip4, + .push_header = udp_push_header, + .get_connection = udp_session_get, + .get_listener = udp_session_get_listener, + .close = udp_session_close, + .send_mss = udp_send_mss_uri, + .send_space = udp_send_space_uri, + .format_connection = format_udp_session_ip4, + .format_listener = format_udp_listener_session_ip4 +}; + +const static transport_proto_vft_t udp6_proto = { + .bind = udp_session_bind_ip6, + .open = udp_open_connection, + .unbind = udp_session_unbind_ip6, + .push_header = udp_push_header, + .get_connection = udp_session_get, + .get_listener = udp_session_get_listener, + .close = udp_session_close, + .send_mss = udp_send_mss_uri, + .send_space = udp_send_space_uri, + .format_connection = format_udp_session_ip6, + .format_listener = format_udp_listener_session_ip6 +}; +/* *INDENT-ON* */ + +static clib_error_t * +udp_init (vlib_main_t * vm) +{ + udp_uri_main_t *um = vnet_get_udp_main (); + ip_main_t *im = &ip_main; + vlib_thread_main_t *tm = vlib_get_thread_main (); + u32 num_threads; + clib_error_t *error = 0; + ip_protocol_info_t *pi; + + um->vlib_main = vm; + um->vnet_main = vnet_get_main (); + + if ((error = vlib_call_init_function (vm, ip_main_init))) + return error; + if ((error = vlib_call_init_function (vm, ip4_lookup_init))) + return error; + if ((error = vlib_call_init_function (vm, ip6_lookup_init))) + return error; + + /* + * Registrations + */ + + /* IP registration */ + pi = ip_get_protocol_info (im, IP_PROTOCOL_UDP); + if (pi == 0) + return clib_error_return (0, "UDP protocol info AWOL"); + pi->format_header = format_udp_header; + pi->unformat_pg_edit = unformat_pg_udp_header; + + + /* Register as transport with URI */ + session_register_transport (SESSION_TYPE_IP4_UDP, &udp4_proto); + session_register_transport (SESSION_TYPE_IP6_UDP, &udp6_proto); + + /* + * Initialize data structures + */ + + num_threads = 1 /* main thread */ + tm->n_threads; + vec_validate (um->udp_sessions, num_threads - 1); + + return error; +} + +VLIB_INIT_FUNCTION (udp_init); + +/* + * fd.io coding-style-patch-verification: ON + * + * Local Variables: + * eval: (c-set-style "gnu") + * End: + */ diff --git a/src/vnet/udp/udp.h b/src/vnet/udp/udp.h new file mode 100644 index 00000000..7ab26ce9 --- /dev/null +++ b/src/vnet/udp/udp.h @@ -0,0 +1,362 @@ +/* + * Copyright (c) 2017 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ +#ifndef __included_udp_h__ +#define __included_udp_h__ + +#include +#include +#include +#include +#include +#include +#include + +#include +#include + +typedef struct +{ + transport_connection_t connection; /** must be first */ + + /** ersatz MTU to limit fifo pushes to test data size */ + u32 mtu; +} udp_connection_t; + +typedef struct _udp_uri_main +{ + /* Per-worker thread udp connection pools */ + udp_connection_t **udp_sessions; + udp_connection_t *udp_listeners; + + /* convenience */ + vlib_main_t *vlib_main; + vnet_main_t *vnet_main; + ip4_main_t *ip4_main; + ip6_main_t *ip6_main; +} udp_uri_main_t; + +extern udp_uri_main_t udp_uri_main; +extern vlib_node_registration_t udp4_uri_input_node; + +always_inline udp_uri_main_t * +vnet_get_udp_main () +{ + return &udp_uri_main; +} + +always_inline udp_connection_t * +udp_connection_get (u32 conn_index, u32 thread_index) +{ + return pool_elt_at_index (udp_uri_main.udp_sessions[thread_index], + conn_index); +} + +always_inline udp_connection_t * +udp_listener_get (u32 conn_index) +{ + return pool_elt_at_index (udp_uri_main.udp_listeners, conn_index); +} + +typedef enum +{ +#define udp_error(n,s) UDP_ERROR_##n, +#include +#undef udp_error + UDP_N_ERROR, +} udp_error_t; + +#define foreach_udp4_dst_port \ +_ (67, dhcp_to_server) \ +_ (68, dhcp_to_client) \ +_ (500, ikev2) \ +_ (3784, bfd4) \ +_ (3785, bfd_echo4) \ +_ (4341, lisp_gpe) \ +_ (4342, lisp_cp) \ +_ (4739, ipfix) \ +_ (4789, vxlan) \ +_ (4789, vxlan6) \ +_ (4790, vxlan_gpe) \ +_ (6633, vpath_3) + + +#define foreach_udp6_dst_port \ +_ (547, dhcpv6_to_server) \ +_ (546, dhcpv6_to_client) \ +_ (3784, bfd6) \ +_ (3785, bfd_echo6) \ +_ (4341, lisp_gpe6) \ +_ (4342, lisp_cp6) \ +_ (4790, vxlan6_gpe) \ +_ (6633, vpath6_3) + +typedef enum +{ +#define _(n,f) UDP_DST_PORT_##f = n, + foreach_udp4_dst_port foreach_udp6_dst_port +#undef _ +} udp_dst_port_t; + +typedef enum +{ +#define _(n,f) UDP6_DST_PORT_##f = n, + foreach_udp6_dst_port +#undef _ +} udp6_dst_port_t; + +typedef struct +{ + /* Name (a c string). */ + char *name; + + /* GRE protocol type in host byte order. */ + udp_dst_port_t dst_port; + + /* Node which handles this type. */ + u32 node_index; + + /* Next index for this type. */ + u32 next_index; +} udp_dst_port_info_t; + +typedef enum +{ + UDP_IP6 = 0, + UDP_IP4, /* the code is full of is_ip4... */ + N_UDP_AF, +} udp_af_t; + +typedef struct +{ + udp_dst_port_info_t *dst_port_infos[N_UDP_AF]; + + /* Hash tables mapping name/protocol to protocol info index. */ + uword *dst_port_info_by_name[N_UDP_AF]; + uword *dst_port_info_by_dst_port[N_UDP_AF]; + + /* convenience */ + vlib_main_t *vlib_main; +} udp_main_t; + +always_inline udp_dst_port_info_t * +udp_get_dst_port_info (udp_main_t * um, udp_dst_port_t dst_port, u8 is_ip4) +{ + uword *p = hash_get (um->dst_port_info_by_dst_port[is_ip4], dst_port); + return p ? vec_elt_at_index (um->dst_port_infos[is_ip4], p[0]) : 0; +} + +format_function_t format_udp_header; +format_function_t format_udp_rx_trace; + +unformat_function_t unformat_udp_header; + +void udp_register_dst_port (vlib_main_t * vm, + udp_dst_port_t dst_port, + u32 node_index, u8 is_ip4); + +void +udp_unregister_dst_port (vlib_main_t * vm, + udp_dst_port_t dst_port, u8 is_ip4); + +void udp_punt_unknown (vlib_main_t * vm, u8 is_ip4, u8 is_add); + +always_inline void +ip_udp_fixup_one (vlib_main_t * vm, vlib_buffer_t * b0, u8 is_ip4) +{ + u16 new_l0; + udp_header_t *udp0; + + if (is_ip4) + { + ip4_header_t *ip0; + ip_csum_t sum0; + u16 old_l0 = 0; + + ip0 = vlib_buffer_get_current (b0); + + /* fix the ing outer-IP checksum */ + sum0 = ip0->checksum; + /* old_l0 always 0, see the rewrite setup */ + new_l0 = clib_host_to_net_u16 (vlib_buffer_length_in_chain (vm, b0)); + + sum0 = ip_csum_update (sum0, old_l0, new_l0, ip4_header_t, + length /* changed member */ ); + ip0->checksum = ip_csum_fold (sum0); + ip0->length = new_l0; + + /* Fix UDP length */ + udp0 = (udp_header_t *) (ip0 + 1); + new_l0 = clib_host_to_net_u16 (vlib_buffer_length_in_chain (vm, b0) + - sizeof (*ip0)); + udp0->length = new_l0; + } + else + { + ip6_header_t *ip0; + int bogus0; + + ip0 = vlib_buffer_get_current (b0); + + new_l0 = clib_host_to_net_u16 (vlib_buffer_length_in_chain (vm, b0) + - sizeof (*ip0)); + ip0->payload_length = new_l0; + + /* Fix UDP length */ + udp0 = (udp_header_t *) (ip0 + 1); + udp0->length = new_l0; + + udp0->checksum = + ip6_tcp_udp_icmp_compute_checksum (vm, b0, ip0, &bogus0); + ASSERT (bogus0 == 0); + + if (udp0->checksum == 0) + udp0->checksum = 0xffff; + } +} + +always_inline void +ip_udp_encap_one (vlib_main_t * vm, vlib_buffer_t * b0, u8 * ec0, word ec_len, + u8 is_ip4) +{ + vlib_buffer_advance (b0, -ec_len); + + if (is_ip4) + { + ip4_header_t *ip0; + + ip0 = vlib_buffer_get_current (b0); + + /* Apply the encap string. */ + clib_memcpy (ip0, ec0, ec_len); + ip_udp_fixup_one (vm, b0, 1); + } + else + { + ip6_header_t *ip0; + + ip0 = vlib_buffer_get_current (b0); + + /* Apply the encap string. */ + clib_memcpy (ip0, ec0, ec_len); + ip_udp_fixup_one (vm, b0, 0); + } +} + +always_inline void +ip_udp_encap_two (vlib_main_t * vm, vlib_buffer_t * b0, vlib_buffer_t * b1, + u8 * ec0, u8 * ec1, word ec_len, u8 is_v4) +{ + u16 new_l0, new_l1; + udp_header_t *udp0, *udp1; + + ASSERT (_vec_len (ec0) == _vec_len (ec1)); + + vlib_buffer_advance (b0, -ec_len); + vlib_buffer_advance (b1, -ec_len); + + if (is_v4) + { + ip4_header_t *ip0, *ip1; + ip_csum_t sum0, sum1; + u16 old_l0 = 0, old_l1 = 0; + + ip0 = vlib_buffer_get_current (b0); + ip1 = vlib_buffer_get_current (b1); + + /* Apply the encap string */ + clib_memcpy (ip0, ec0, ec_len); + clib_memcpy (ip1, ec1, ec_len); + + /* fix the ing outer-IP checksum */ + sum0 = ip0->checksum; + sum1 = ip1->checksum; + + /* old_l0 always 0, see the rewrite setup */ + new_l0 = clib_host_to_net_u16 (vlib_buffer_length_in_chain (vm, b0)); + new_l1 = clib_host_to_net_u16 (vlib_buffer_length_in_chain (vm, b1)); + + sum0 = ip_csum_update (sum0, old_l0, new_l0, ip4_header_t, + length /* changed member */ ); + sum1 = ip_csum_update (sum1, old_l1, new_l1, ip4_header_t, + length /* changed member */ ); + + ip0->checksum = ip_csum_fold (sum0); + ip1->checksum = ip_csum_fold (sum1); + + ip0->length = new_l0; + ip1->length = new_l1; + + /* Fix UDP length */ + udp0 = (udp_header_t *) (ip0 + 1); + udp1 = (udp_header_t *) (ip1 + 1); + + new_l0 = + clib_host_to_net_u16 (vlib_buffer_length_in_chain (vm, b0) - + sizeof (*ip0)); + new_l1 = + clib_host_to_net_u16 (vlib_buffer_length_in_chain (vm, b1) - + sizeof (*ip1)); + udp0->length = new_l0; + udp1->length = new_l1; + } + else + { + ip6_header_t *ip0, *ip1; + int bogus0, bogus1; + + ip0 = vlib_buffer_get_current (b0); + ip1 = vlib_buffer_get_current (b1); + + /* Apply the encap string. */ + clib_memcpy (ip0, ec0, ec_len); + clib_memcpy (ip1, ec1, ec_len); + + new_l0 = clib_host_to_net_u16 (vlib_buffer_length_in_chain (vm, b0) + - sizeof (*ip0)); + new_l1 = clib_host_to_net_u16 (vlib_buffer_length_in_chain (vm, b1) + - sizeof (*ip1)); + ip0->payload_length = new_l0; + ip1->payload_length = new_l1; + + /* Fix UDP length */ + udp0 = (udp_header_t *) (ip0 + 1); + udp1 = (udp_header_t *) (ip1 + 1); + + udp0->length = new_l0; + udp1->length = new_l1; + + udp0->checksum = + ip6_tcp_udp_icmp_compute_checksum (vm, b0, ip0, &bogus0); + udp1->checksum = + ip6_tcp_udp_icmp_compute_checksum (vm, b1, ip1, &bogus1); + ASSERT (bogus0 == 0); + ASSERT (bogus1 == 0); + + if (udp0->checksum == 0) + udp0->checksum = 0xffff; + if (udp1->checksum == 0) + udp1->checksum = 0xffff; + } +} + +/* + * fd.io coding-style-patch-verification: ON + * + * Local Variables: + * eval: (c-set-style "gnu") + * End: + */ + +#endif /* __included_udp_h__ */ diff --git a/src/vnet/udp/udp_error.def b/src/vnet/udp/udp_error.def new file mode 100644 index 00000000..bfdae0ac --- /dev/null +++ b/src/vnet/udp/udp_error.def @@ -0,0 +1,21 @@ +/* + * udp_error.def: udp errors + * + * Copyright (c) 2013-2016 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +udp_error (NONE, "no error") +udp_error (NO_LISTENER, "no listener for dst port") +udp_error (LENGTH_ERROR, "UDP packets with length errors") +udp_error (PUNT, "no listener punt") diff --git a/src/vnet/udp/udp_format.c b/src/vnet/udp/udp_format.c new file mode 100644 index 00000000..abdf561e --- /dev/null +++ b/src/vnet/udp/udp_format.c @@ -0,0 +1,91 @@ +/* + * Copyright (c) 2015 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ +/* + * ip/udp_format.c: udp formatting + * + * Copyright (c) 2008 Eliot Dresselhaus + * + * Permission is hereby granted, free of charge, to any person obtaining + * a copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sublicense, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice shall be + * included in all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND + * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE + * LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + * OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + * WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + */ + +#include + +/* Format UDP header. */ +u8 * +format_udp_header (u8 * s, va_list * args) +{ + udp_header_t *udp = va_arg (*args, udp_header_t *); + u32 max_header_bytes = va_arg (*args, u32); + uword indent; + u32 header_bytes = sizeof (udp[0]); + + /* Nothing to do. */ + if (max_header_bytes < sizeof (udp[0])) + return format (s, "UDP header truncated"); + + indent = format_get_indent (s); + indent += 2; + + s = format (s, "UDP: %d -> %d", + clib_net_to_host_u16 (udp->src_port), + clib_net_to_host_u16 (udp->dst_port)); + + s = format (s, "\n%Ulength %d, checksum 0x%04x", + format_white_space, indent, + clib_net_to_host_u16 (udp->length), + clib_net_to_host_u16 (udp->checksum)); + + /* Recurse into next protocol layer. */ + if (max_header_bytes != 0 && header_bytes < max_header_bytes) + { + ip_main_t *im = &ip_main; + tcp_udp_port_info_t *pi; + + pi = ip_get_tcp_udp_port_info (im, udp->dst_port); + + if (pi && pi->format_header) + s = format (s, "\n%U%U", + format_white_space, indent - 2, pi->format_header, + /* next protocol header */ (udp + 1), + max_header_bytes - sizeof (udp[0])); + } + + return s; +} + +/* + * fd.io coding-style-patch-verification: ON + * + * Local Variables: + * eval: (c-set-style "gnu") + * End: + */ diff --git a/src/vnet/udp/udp_input.c b/src/vnet/udp/udp_input.c new file mode 100644 index 00000000..4d509335 --- /dev/null +++ b/src/vnet/udp/udp_input.c @@ -0,0 +1,314 @@ +/* + * Copyright (c) 2016 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +#include +#include +#include +#include + +#include +#include +#include +#include + +#include + +#include +#include "../session/application_interface.h" + +vlib_node_registration_t udp4_uri_input_node; + +typedef struct +{ + u32 session; + u32 disposition; + u32 thread_index; +} udp4_uri_input_trace_t; + +/* packet trace format function */ +static u8 * +format_udp4_uri_input_trace (u8 * s, va_list * args) +{ + CLIB_UNUSED (vlib_main_t * vm) = va_arg (*args, vlib_main_t *); + CLIB_UNUSED (vlib_node_t * node) = va_arg (*args, vlib_node_t *); + udp4_uri_input_trace_t *t = va_arg (*args, udp4_uri_input_trace_t *); + + s = format (s, "UDP4_URI_INPUT: session %d, disposition %d, thread %d", + t->session, t->disposition, t->thread_index); + return s; +} + +typedef enum +{ + UDP4_URI_INPUT_NEXT_DROP, + UDP4_URI_INPUT_N_NEXT, +} udp4_uri_input_next_t; + +static char *udp4_uri_input_error_strings[] = { +#define _(sym,string) string, + foreach_session_input_error +#undef _ +}; + +static uword +udp4_uri_input_node_fn (vlib_main_t * vm, + vlib_node_runtime_t * node, vlib_frame_t * frame) +{ + u32 n_left_from, *from, *to_next; + udp4_uri_input_next_t next_index; + udp_uri_main_t *um = vnet_get_udp_main (); + session_manager_main_t *smm = vnet_get_session_manager_main (); + u32 my_thread_index = vm->cpu_index; + u8 my_enqueue_epoch; + u32 *session_indices_to_enqueue; + static u32 serial_number; + int i; + + my_enqueue_epoch = ++smm->current_enqueue_epoch[my_thread_index]; + + from = vlib_frame_vector_args (frame); + n_left_from = frame->n_vectors; + next_index = node->cached_next_index; + + while (n_left_from > 0) + { + u32 n_left_to_next; + + vlib_get_next_frame (vm, node, next_index, to_next, n_left_to_next); + + while (n_left_from > 0 && n_left_to_next > 0) + { + u32 bi0; + vlib_buffer_t *b0; + u32 next0 = UDP4_URI_INPUT_NEXT_DROP; + u32 error0 = SESSION_ERROR_ENQUEUED; + udp_header_t *udp0; + ip4_header_t *ip0; + stream_session_t *s0; + svm_fifo_t *f0; + u16 udp_len0; + u8 *data0; + + /* speculatively enqueue b0 to the current next frame */ + bi0 = from[0]; + to_next[0] = bi0; + from += 1; + to_next += 1; + n_left_from -= 1; + n_left_to_next -= 1; + + b0 = vlib_get_buffer (vm, bi0); + + /* udp_local hands us a pointer to the udp data */ + + data0 = vlib_buffer_get_current (b0); + udp0 = (udp_header_t *) (data0 - sizeof (*udp0)); + + /* $$$$ fixme: udp_local doesn't do ip options correctly anyhow */ + ip0 = (ip4_header_t *) (((u8 *) udp0) - sizeof (*ip0)); + s0 = 0; + + /* lookup session */ + s0 = stream_session_lookup4 (&ip0->dst_address, &ip0->src_address, + udp0->dst_port, udp0->src_port, + SESSION_TYPE_IP4_UDP, my_thread_index); + + /* no listener */ + if (PREDICT_FALSE (s0 == 0)) + { + error0 = SESSION_ERROR_NO_LISTENER; + goto trace0; + } + + f0 = s0->server_rx_fifo; + + /* established hit */ + if (PREDICT_TRUE (s0->session_state == SESSION_STATE_READY)) + { + udp_len0 = clib_net_to_host_u16 (udp0->length); + + if (PREDICT_FALSE (udp_len0 > svm_fifo_max_enqueue (f0))) + { + error0 = SESSION_ERROR_FIFO_FULL; + goto trace0; + } + + svm_fifo_enqueue_nowait (f0, 0 /* pid */ , + udp_len0 - sizeof (*udp0), + (u8 *) (udp0 + 1)); + + b0->error = node->errors[SESSION_ERROR_ENQUEUED]; + + /* We need to send an RX event on this fifo */ + if (s0->enqueue_epoch != my_enqueue_epoch) + { + s0->enqueue_epoch = my_enqueue_epoch; + + vec_add1 (smm->session_indices_to_enqueue_by_thread + [my_thread_index], + s0 - smm->sessions[my_thread_index]); + } + } + /* listener hit */ + else if (s0->session_state == SESSION_STATE_LISTENING) + { + udp_connection_t *us; + int rv; + + error0 = SESSION_ERROR_NOT_READY; + + /* + * create udp transport session + */ + pool_get (um->udp_sessions[my_thread_index], us); + + us->mtu = 1024; /* $$$$ policy */ + + us->c_lcl_ip4.as_u32 = ip0->dst_address.as_u32; + us->c_rmt_ip4.as_u32 = ip0->src_address.as_u32; + us->c_lcl_port = udp0->dst_port; + us->c_rmt_port = udp0->src_port; + us->c_proto = SESSION_TYPE_IP4_UDP; + us->c_c_index = us - um->udp_sessions[my_thread_index]; + + /* + * create stream session and attach the udp session to it + */ + rv = stream_session_accept (&us->connection, s0->session_index, + SESSION_TYPE_IP4_UDP, + 1 /*notify */ ); + if (rv) + error0 = rv; + + } + else + { + + error0 = SESSION_ERROR_NOT_READY; + goto trace0; + } + + trace0: + b0->error = node->errors[error0]; + + if (PREDICT_FALSE ((node->flags & VLIB_NODE_FLAG_TRACE) + && (b0->flags & VLIB_BUFFER_IS_TRACED))) + { + udp4_uri_input_trace_t *t = + vlib_add_trace (vm, node, b0, sizeof (*t)); + + t->session = ~0; + if (s0) + t->session = s0 - smm->sessions[my_thread_index]; + t->disposition = error0; + t->thread_index = my_thread_index; + } + + /* verify speculative enqueue, maybe switch current next frame */ + vlib_validate_buffer_enqueue_x1 (vm, node, next_index, + to_next, n_left_to_next, + bi0, next0); + } + + vlib_put_next_frame (vm, node, next_index, n_left_to_next); + } + + /* Send enqueue events */ + + session_indices_to_enqueue = + smm->session_indices_to_enqueue_by_thread[my_thread_index]; + + for (i = 0; i < vec_len (session_indices_to_enqueue); i++) + { + session_fifo_event_t evt; + unix_shared_memory_queue_t *q; + stream_session_t *s0; + application_t *server0; + + /* Get session */ + s0 = pool_elt_at_index (smm->sessions[my_thread_index], + session_indices_to_enqueue[i]); + + /* Get session's server */ + server0 = application_get (s0->app_index); + + /* Built-in server? Deliver the goods... */ + if (server0->cb_fns.builtin_server_rx_callback) + { + server0->cb_fns.builtin_server_rx_callback (s0); + continue; + } + + /* Fabricate event */ + evt.fifo = s0->server_rx_fifo; + evt.event_type = FIFO_EVENT_SERVER_RX; + evt.event_id = serial_number++; + evt.enqueue_length = svm_fifo_max_dequeue (s0->server_rx_fifo); + + /* Add event to server's event queue */ + q = server0->event_queue; + + /* Don't block for lack of space */ + if (PREDICT_TRUE (q->cursize < q->maxsize)) + unix_shared_memory_queue_add (server0->event_queue, (u8 *) & evt, + 0 /* do wait for mutex */ ); + else + { + vlib_node_increment_counter (vm, udp4_uri_input_node.index, + SESSION_ERROR_FIFO_FULL, 1); + } + if (1) + { + ELOG_TYPE_DECLARE (e) = + { + .format = "evt-enqueue: id %d length %d",.format_args = "i4i4",}; + struct + { + u32 data[2]; + } *ed; + ed = ELOG_DATA (&vlib_global_main.elog_main, e); + ed->data[0] = evt.event_id; + ed->data[1] = evt.enqueue_length; + } + } + + vec_reset_length (session_indices_to_enqueue); + + smm->session_indices_to_enqueue_by_thread[my_thread_index] = + session_indices_to_enqueue; + + return frame->n_vectors; +} + +VLIB_REGISTER_NODE (udp4_uri_input_node) = +{ + .function = udp4_uri_input_node_fn,.name = "udp4-uri-input",.vector_size = + sizeof (u32),.format_trace = format_udp4_uri_input_trace,.type = + VLIB_NODE_TYPE_INTERNAL,.n_errors = + ARRAY_LEN (udp4_uri_input_error_strings),.error_strings = + udp4_uri_input_error_strings,.n_next_nodes = UDP4_URI_INPUT_N_NEXT, + /* edit / add dispositions here */ + .next_nodes = + { + [UDP4_URI_INPUT_NEXT_DROP] = "error-drop",} +,}; + +/* + * fd.io coding-style-patch-verification: ON + * + * Local Variables: + * eval: (c-set-style "gnu") + * End: + */ diff --git a/src/vnet/udp/udp_local.c b/src/vnet/udp/udp_local.c new file mode 100644 index 00000000..6b239f73 --- /dev/null +++ b/src/vnet/udp/udp_local.c @@ -0,0 +1,666 @@ +/* + * node.c: udp packet processing + * + * Copyright (c) 2013 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +#include +#include +#include +#include +#include + +udp_main_t udp_main; + +#define foreach_udp_input_next \ + _ (PUNT, "error-punt") \ + _ (DROP, "error-drop") \ + _ (ICMP4_ERROR, "ip4-icmp-error") \ + _ (ICMP6_ERROR, "ip6-icmp-error") + +typedef enum +{ +#define _(s,n) UDP_INPUT_NEXT_##s, + foreach_udp_input_next +#undef _ + UDP_INPUT_N_NEXT, +} udp_input_next_t; + +typedef struct +{ + u16 src_port; + u16 dst_port; + u8 bound; +} udp_rx_trace_t; + +u8 * +format_udp_rx_trace (u8 * s, va_list * args) +{ + CLIB_UNUSED (vlib_main_t * vm) = va_arg (*args, vlib_main_t *); + CLIB_UNUSED (vlib_node_t * node) = va_arg (*args, vlib_node_t *); + udp_rx_trace_t *t = va_arg (*args, udp_rx_trace_t *); + + s = format (s, "UDP: src-port %d dst-port %d%s", + clib_net_to_host_u16 (t->src_port), + clib_net_to_host_u16 (t->dst_port), + t->bound ? "" : " (no listener)"); + return s; +} + +typedef struct +{ + /* Sparse vector mapping udp dst_port in network byte order + to next index. */ + u16 *next_by_dst_port; + u8 punt_unknown; +} udp_input_runtime_t; + +vlib_node_registration_t udp4_input_node; +vlib_node_registration_t udp6_input_node; + +always_inline uword +udp46_input_inline (vlib_main_t * vm, + vlib_node_runtime_t * node, + vlib_frame_t * from_frame, int is_ip4) +{ + udp_input_runtime_t *rt = is_ip4 ? + (void *) vlib_node_get_runtime_data (vm, udp4_input_node.index) + : (void *) vlib_node_get_runtime_data (vm, udp6_input_node.index); + __attribute__ ((unused)) u32 n_left_from, next_index, *from, *to_next; + word n_no_listener = 0; + u8 punt_unknown = rt->punt_unknown; + + from = vlib_frame_vector_args (from_frame); + n_left_from = from_frame->n_vectors; + + next_index = node->cached_next_index; + + while (n_left_from > 0) + { + u32 n_left_to_next; + + vlib_get_next_frame (vm, node, next_index, to_next, n_left_to_next); + + while (n_left_from >= 4 && n_left_to_next >= 2) + { + u32 bi0, bi1; + vlib_buffer_t *b0, *b1; + udp_header_t *h0 = 0, *h1 = 0; + u32 i0, i1, dst_port0, dst_port1; + u32 advance0, advance1; + u32 error0, next0, error1, next1; + + /* Prefetch next iteration. */ + { + vlib_buffer_t *p2, *p3; + + p2 = vlib_get_buffer (vm, from[2]); + p3 = vlib_get_buffer (vm, from[3]); + + vlib_prefetch_buffer_header (p2, LOAD); + vlib_prefetch_buffer_header (p3, LOAD); + + CLIB_PREFETCH (p2->data, sizeof (h0[0]), LOAD); + CLIB_PREFETCH (p3->data, sizeof (h1[0]), LOAD); + } + + bi0 = from[0]; + bi1 = from[1]; + to_next[0] = bi0; + to_next[1] = bi1; + from += 2; + to_next += 2; + n_left_to_next -= 2; + n_left_from -= 2; + + b0 = vlib_get_buffer (vm, bi0); + b1 = vlib_get_buffer (vm, bi1); + + /* ip4/6_local hands us the ip header, not the udp header */ + if (is_ip4) + { + advance0 = sizeof (ip4_header_t); + advance1 = sizeof (ip4_header_t); + } + else + { + advance0 = sizeof (ip6_header_t); + advance1 = sizeof (ip6_header_t); + } + + if (PREDICT_FALSE (b0->current_length < advance0 + sizeof (*h0))) + { + error0 = UDP_ERROR_LENGTH_ERROR; + next0 = UDP_INPUT_NEXT_DROP; + } + else + { + vlib_buffer_advance (b0, advance0); + h0 = vlib_buffer_get_current (b0); + error0 = next0 = 0; + if (PREDICT_FALSE (clib_net_to_host_u16 (h0->length) > + vlib_buffer_length_in_chain (vm, b0))) + { + error0 = UDP_ERROR_LENGTH_ERROR; + next0 = UDP_INPUT_NEXT_DROP; + } + } + + if (PREDICT_FALSE (b1->current_length < advance1 + sizeof (*h1))) + { + error1 = UDP_ERROR_LENGTH_ERROR; + next1 = UDP_INPUT_NEXT_DROP; + } + else + { + vlib_buffer_advance (b1, advance1); + h1 = vlib_buffer_get_current (b1); + error1 = next1 = 0; + if (PREDICT_FALSE (clib_net_to_host_u16 (h1->length) > + vlib_buffer_length_in_chain (vm, b1))) + { + error1 = UDP_ERROR_LENGTH_ERROR; + next1 = UDP_INPUT_NEXT_DROP; + } + } + + /* Index sparse array with network byte order. */ + dst_port0 = (error0 == 0) ? h0->dst_port : 0; + dst_port1 = (error1 == 0) ? h1->dst_port : 0; + sparse_vec_index2 (rt->next_by_dst_port, dst_port0, dst_port1, + &i0, &i1); + next0 = (error0 == 0) ? vec_elt (rt->next_by_dst_port, i0) : next0; + next1 = (error1 == 0) ? vec_elt (rt->next_by_dst_port, i1) : next1; + + if (PREDICT_FALSE (i0 == SPARSE_VEC_INVALID_INDEX)) + { + // move the pointer back so icmp-error can find the + // ip packet header + vlib_buffer_advance (b0, -(word) advance0); + + if (PREDICT_FALSE (punt_unknown)) + { + b0->error = node->errors[UDP_ERROR_PUNT]; + next0 = UDP_INPUT_NEXT_PUNT; + } + else if (is_ip4) + { + icmp4_error_set_vnet_buffer (b0, + ICMP4_destination_unreachable, + ICMP4_destination_unreachable_port_unreachable, + 0); + next0 = UDP_INPUT_NEXT_ICMP4_ERROR; + n_no_listener++; + } + else + { + icmp6_error_set_vnet_buffer (b0, + ICMP6_destination_unreachable, + ICMP6_destination_unreachable_port_unreachable, + 0); + next0 = UDP_INPUT_NEXT_ICMP6_ERROR; + n_no_listener++; + } + } + else + { + b0->error = node->errors[UDP_ERROR_NONE]; + // advance to the payload + vlib_buffer_advance (b0, sizeof (*h0)); + } + + if (PREDICT_FALSE (i1 == SPARSE_VEC_INVALID_INDEX)) + { + // move the pointer back so icmp-error can find the + // ip packet header + vlib_buffer_advance (b1, -(word) advance1); + + if (PREDICT_FALSE (punt_unknown)) + { + b1->error = node->errors[UDP_ERROR_PUNT]; + next1 = UDP_INPUT_NEXT_PUNT; + } + else if (is_ip4) + { + icmp4_error_set_vnet_buffer (b1, + ICMP4_destination_unreachable, + ICMP4_destination_unreachable_port_unreachable, + 0); + next1 = UDP_INPUT_NEXT_ICMP4_ERROR; + n_no_listener++; + } + else + { + icmp6_error_set_vnet_buffer (b1, + ICMP6_destination_unreachable, + ICMP6_destination_unreachable_port_unreachable, + 0); + next1 = UDP_INPUT_NEXT_ICMP6_ERROR; + n_no_listener++; + } + } + else + { + b1->error = node->errors[UDP_ERROR_NONE]; + // advance to the payload + vlib_buffer_advance (b1, sizeof (*h1)); + } + + if (PREDICT_FALSE (b0->flags & VLIB_BUFFER_IS_TRACED)) + { + udp_rx_trace_t *tr = vlib_add_trace (vm, node, + b0, sizeof (*tr)); + if (b0->error != node->errors[UDP_ERROR_LENGTH_ERROR]) + { + tr->src_port = h0 ? h0->src_port : 0; + tr->dst_port = h0 ? h0->dst_port : 0; + tr->bound = (next0 != UDP_INPUT_NEXT_ICMP4_ERROR && + next0 != UDP_INPUT_NEXT_ICMP6_ERROR); + } + } + if (PREDICT_FALSE (b1->flags & VLIB_BUFFER_IS_TRACED)) + { + udp_rx_trace_t *tr = vlib_add_trace (vm, node, + b1, sizeof (*tr)); + if (b1->error != node->errors[UDP_ERROR_LENGTH_ERROR]) + { + tr->src_port = h1 ? h1->src_port : 0; + tr->dst_port = h1 ? h1->dst_port : 0; + tr->bound = (next1 != UDP_INPUT_NEXT_ICMP4_ERROR && + next1 != UDP_INPUT_NEXT_ICMP6_ERROR); + } + } + + vlib_validate_buffer_enqueue_x2 (vm, node, next_index, + to_next, n_left_to_next, + bi0, bi1, next0, next1); + } + + while (n_left_from > 0 && n_left_to_next > 0) + { + u32 bi0; + vlib_buffer_t *b0; + udp_header_t *h0 = 0; + u32 i0, next0; + u32 advance0; + + bi0 = from[0]; + to_next[0] = bi0; + from += 1; + to_next += 1; + n_left_from -= 1; + n_left_to_next -= 1; + + b0 = vlib_get_buffer (vm, bi0); + + /* ip4/6_local hands us the ip header, not the udp header */ + if (is_ip4) + advance0 = sizeof (ip4_header_t); + else + advance0 = sizeof (ip6_header_t); + + if (PREDICT_FALSE (b0->current_length < advance0 + sizeof (*h0))) + { + b0->error = node->errors[UDP_ERROR_LENGTH_ERROR]; + next0 = UDP_INPUT_NEXT_DROP; + goto trace_x1; + } + + vlib_buffer_advance (b0, advance0); + + h0 = vlib_buffer_get_current (b0); + + if (PREDICT_TRUE (clib_net_to_host_u16 (h0->length) <= + vlib_buffer_length_in_chain (vm, b0))) + { + i0 = sparse_vec_index (rt->next_by_dst_port, h0->dst_port); + next0 = vec_elt (rt->next_by_dst_port, i0); + + if (PREDICT_FALSE (i0 == SPARSE_VEC_INVALID_INDEX)) + { + // move the pointer back so icmp-error can find the + // ip packet header + vlib_buffer_advance (b0, -(word) advance0); + + if (PREDICT_FALSE (punt_unknown)) + { + b0->error = node->errors[UDP_ERROR_PUNT]; + next0 = UDP_INPUT_NEXT_PUNT; + } + else if (is_ip4) + { + icmp4_error_set_vnet_buffer (b0, + ICMP4_destination_unreachable, + ICMP4_destination_unreachable_port_unreachable, + 0); + next0 = UDP_INPUT_NEXT_ICMP4_ERROR; + n_no_listener++; + } + else + { + icmp6_error_set_vnet_buffer (b0, + ICMP6_destination_unreachable, + ICMP6_destination_unreachable_port_unreachable, + 0); + next0 = UDP_INPUT_NEXT_ICMP6_ERROR; + n_no_listener++; + } + } + else + { + b0->error = node->errors[UDP_ERROR_NONE]; + // advance to the payload + vlib_buffer_advance (b0, sizeof (*h0)); + } + } + else + { + b0->error = node->errors[UDP_ERROR_LENGTH_ERROR]; + next0 = UDP_INPUT_NEXT_DROP; + } + + trace_x1: + if (PREDICT_FALSE (b0->flags & VLIB_BUFFER_IS_TRACED)) + { + udp_rx_trace_t *tr = vlib_add_trace (vm, node, + b0, sizeof (*tr)); + if (b0->error != node->errors[UDP_ERROR_LENGTH_ERROR]) + { + tr->src_port = h0->src_port; + tr->dst_port = h0->dst_port; + tr->bound = (next0 != UDP_INPUT_NEXT_ICMP4_ERROR && + next0 != UDP_INPUT_NEXT_ICMP6_ERROR); + } + } + + vlib_validate_buffer_enqueue_x1 (vm, node, next_index, + to_next, n_left_to_next, + bi0, next0); + } + + vlib_put_next_frame (vm, node, next_index, n_left_to_next); + } + vlib_error_count (vm, node->node_index, UDP_ERROR_NO_LISTENER, + n_no_listener); + return from_frame->n_vectors; +} + +static char *udp_error_strings[] = { +#define udp_error(n,s) s, +#include "udp_error.def" +#undef udp_error +}; + +static uword +udp4_input (vlib_main_t * vm, + vlib_node_runtime_t * node, vlib_frame_t * from_frame) +{ + return udp46_input_inline (vm, node, from_frame, 1 /* is_ip4 */ ); +} + +static uword +udp6_input (vlib_main_t * vm, + vlib_node_runtime_t * node, vlib_frame_t * from_frame) +{ + return udp46_input_inline (vm, node, from_frame, 0 /* is_ip4 */ ); +} + + +/* *INDENT-OFF* */ +VLIB_REGISTER_NODE (udp4_input_node) = { + .function = udp4_input, + .name = "ip4-udp-lookup", + /* Takes a vector of packets. */ + .vector_size = sizeof (u32), + + .runtime_data_bytes = sizeof (udp_input_runtime_t), + + .n_errors = UDP_N_ERROR, + .error_strings = udp_error_strings, + + .n_next_nodes = UDP_INPUT_N_NEXT, + .next_nodes = { +#define _(s,n) [UDP_INPUT_NEXT_##s] = n, + foreach_udp_input_next +#undef _ + }, + + .format_buffer = format_udp_header, + .format_trace = format_udp_rx_trace, + .unformat_buffer = unformat_udp_header, +}; +/* *INDENT-ON* */ + +VLIB_NODE_FUNCTION_MULTIARCH (udp4_input_node, udp4_input); + +/* *INDENT-OFF* */ +VLIB_REGISTER_NODE (udp6_input_node) = { + .function = udp6_input, + .name = "ip6-udp-lookup", + /* Takes a vector of packets. */ + .vector_size = sizeof (u32), + + .runtime_data_bytes = sizeof (udp_input_runtime_t), + + .n_errors = UDP_N_ERROR, + .error_strings = udp_error_strings, + + .n_next_nodes = UDP_INPUT_N_NEXT, + .next_nodes = { +#define _(s,n) [UDP_INPUT_NEXT_##s] = n, + foreach_udp_input_next +#undef _ + }, + + .format_buffer = format_udp_header, + .format_trace = format_udp_rx_trace, + .unformat_buffer = unformat_udp_header, +}; +/* *INDENT-ON* */ + +VLIB_NODE_FUNCTION_MULTIARCH (udp6_input_node, udp6_input); + +static void +add_dst_port (udp_main_t * um, + udp_dst_port_t dst_port, char *dst_port_name, u8 is_ip4) +{ + udp_dst_port_info_t *pi; + u32 i; + + vec_add2 (um->dst_port_infos[is_ip4], pi, 1); + i = pi - um->dst_port_infos[is_ip4]; + + pi->name = dst_port_name; + pi->dst_port = dst_port; + pi->next_index = pi->node_index = ~0; + + hash_set (um->dst_port_info_by_dst_port[is_ip4], dst_port, i); + + if (pi->name) + hash_set_mem (um->dst_port_info_by_name[is_ip4], pi->name, i); +} + +void +udp_register_dst_port (vlib_main_t * vm, + udp_dst_port_t dst_port, u32 node_index, u8 is_ip4) +{ + udp_main_t *um = &udp_main; + udp_dst_port_info_t *pi; + udp_input_runtime_t *rt; + u16 *n; + + { + clib_error_t *error = vlib_call_init_function (vm, udp_local_init); + if (error) + clib_error_report (error); + } + + pi = udp_get_dst_port_info (um, dst_port, is_ip4); + if (!pi) + { + add_dst_port (um, dst_port, 0, is_ip4); + pi = udp_get_dst_port_info (um, dst_port, is_ip4); + ASSERT (pi); + } + + pi->node_index = node_index; + pi->next_index = vlib_node_add_next (vm, + is_ip4 ? udp4_input_node.index + : udp6_input_node.index, node_index); + + /* Setup udp protocol -> next index sparse vector mapping. */ + rt = vlib_node_get_runtime_data + (vm, is_ip4 ? udp4_input_node.index : udp6_input_node.index); + n = sparse_vec_validate (rt->next_by_dst_port, + clib_host_to_net_u16 (dst_port)); + n[0] = pi->next_index; +} + +void +udp_unregister_dst_port (vlib_main_t * vm, udp_dst_port_t dst_port, u8 is_ip4) +{ + udp_main_t *um = &udp_main; + udp_dst_port_info_t *pi; + udp_input_runtime_t *rt; + u16 *n; + + pi = udp_get_dst_port_info (um, dst_port, is_ip4); + /* Not registered? Fagedaboudit */ + if (!pi) + return; + + /* Kill the mapping. Don't bother killing the pi, it may be back. */ + rt = vlib_node_get_runtime_data + (vm, is_ip4 ? udp4_input_node.index : udp6_input_node.index); + n = sparse_vec_validate (rt->next_by_dst_port, + clib_host_to_net_u16 (dst_port)); + n[0] = SPARSE_VEC_INVALID_INDEX; +} + +void +udp_punt_unknown (vlib_main_t * vm, u8 is_ip4, u8 is_add) +{ + udp_input_runtime_t *rt; + + { + clib_error_t *error = vlib_call_init_function (vm, udp_local_init); + if (error) + clib_error_report (error); + } + + rt = vlib_node_get_runtime_data + (vm, is_ip4 ? udp4_input_node.index : udp6_input_node.index); + + rt->punt_unknown = is_add; +} + +/* Parse a UDP header. */ +uword +unformat_udp_header (unformat_input_t * input, va_list * args) +{ + u8 **result = va_arg (*args, u8 **); + udp_header_t *udp; + __attribute__ ((unused)) int old_length; + u16 src_port, dst_port; + + /* Allocate space for IP header. */ + { + void *p; + + old_length = vec_len (*result); + vec_add2 (*result, p, sizeof (ip4_header_t)); + udp = p; + } + + memset (udp, 0, sizeof (udp[0])); + if (unformat (input, "src-port %d dst-port %d", &src_port, &dst_port)) + { + udp->src_port = clib_host_to_net_u16 (src_port); + udp->dst_port = clib_host_to_net_u16 (dst_port); + return 1; + } + return 0; +} + +static void +udp_setup_node (vlib_main_t * vm, u32 node_index) +{ + vlib_node_t *n = vlib_get_node (vm, node_index); + pg_node_t *pn = pg_get_node (node_index); + + n->format_buffer = format_udp_header; + n->unformat_buffer = unformat_udp_header; + pn->unformat_edit = unformat_pg_udp_header; +} + +clib_error_t * +udp_local_init (vlib_main_t * vm) +{ + udp_input_runtime_t *rt; + udp_main_t *um = &udp_main; + int i; + + { + clib_error_t *error; + error = vlib_call_init_function (vm, udp_init); + if (error) + clib_error_report (error); + } + + + for (i = 0; i < 2; i++) + { + um->dst_port_info_by_name[i] = hash_create_string (0, sizeof (uword)); + um->dst_port_info_by_dst_port[i] = hash_create (0, sizeof (uword)); + } + + udp_setup_node (vm, udp4_input_node.index); + udp_setup_node (vm, udp6_input_node.index); + + rt = vlib_node_get_runtime_data (vm, udp4_input_node.index); + + rt->next_by_dst_port = sparse_vec_new + ( /* elt bytes */ sizeof (rt->next_by_dst_port[0]), + /* bits in index */ BITS (((udp_header_t *) 0)->dst_port)); + + rt->punt_unknown = 0; + +#define _(n,s) add_dst_port (um, UDP_DST_PORT_##s, #s, 1 /* is_ip4 */); + foreach_udp4_dst_port +#undef _ + rt = vlib_node_get_runtime_data (vm, udp6_input_node.index); + + rt->next_by_dst_port = sparse_vec_new + ( /* elt bytes */ sizeof (rt->next_by_dst_port[0]), + /* bits in index */ BITS (((udp_header_t *) 0)->dst_port)); + + rt->punt_unknown = 0; + +#define _(n,s) add_dst_port (um, UDP_DST_PORT_##s, #s, 0 /* is_ip4 */); + foreach_udp6_dst_port +#undef _ + ip4_register_protocol (IP_PROTOCOL_UDP, udp4_input_node.index); + /* Note: ip6 differs from ip4, UDP is hotwired to ip6-udp-lookup */ + return 0; +} + +VLIB_INIT_FUNCTION (udp_local_init); + +/* + * fd.io coding-style-patch-verification: ON + * + * Local Variables: + * eval: (c-set-style "gnu") + * End: + */ diff --git a/src/vnet/udp/udp_packet.h b/src/vnet/udp/udp_packet.h new file mode 100644 index 00000000..beea3059 --- /dev/null +++ b/src/vnet/udp/udp_packet.h @@ -0,0 +1,65 @@ +/* + * Copyright (c) 2015 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ +/* + * ip4/udp_packet.h: UDP packet format + * + * Copyright (c) 2008 Eliot Dresselhaus + * + * Permission is hereby granted, free of charge, to any person obtaining + * a copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sublicense, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice shall be + * included in all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND + * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE + * LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + * OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + * WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + */ + +#ifndef included_udp_packet_h +#define included_udp_packet_h + +typedef struct +{ + /* Source and destination port. */ + u16 src_port, dst_port; + + /* Length of UDP header plus payload. */ + u16 length; + + /* Checksum of UDP pseudo-header and data or + zero if checksum is disabled. */ + u16 checksum; +} udp_header_t; + +#endif /* included_udp_packet_h */ + + +/* + * fd.io coding-style-patch-verification: ON + * + * Local Variables: + * eval: (c-set-style "gnu") + * End: + */ diff --git a/src/vnet/udp/udp_pg.c b/src/vnet/udp/udp_pg.c new file mode 100644 index 00000000..c9d8d38c --- /dev/null +++ b/src/vnet/udp/udp_pg.c @@ -0,0 +1,237 @@ +/* + * Copyright (c) 2015 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ +/* + * ip/udp_pg: UDP packet-generator interface + * + * Copyright (c) 2008 Eliot Dresselhaus + * + * Permission is hereby granted, free of charge, to any person obtaining + * a copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sublicense, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice shall be + * included in all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND + * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE + * LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + * OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + * WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + */ + +#include +#include /* for unformat_udp_udp_port */ + +#define UDP_PG_EDIT_LENGTH (1 << 0) +#define UDP_PG_EDIT_CHECKSUM (1 << 1) + +always_inline void +udp_pg_edit_function_inline (pg_main_t * pg, + pg_stream_t * s, + pg_edit_group_t * g, + u32 * packets, u32 n_packets, u32 flags) +{ + vlib_main_t *vm = vlib_get_main (); + u32 ip_offset, udp_offset; + + udp_offset = g->start_byte_offset; + ip_offset = (g - 1)->start_byte_offset; + + while (n_packets >= 1) + { + vlib_buffer_t *p0; + ip4_header_t *ip0; + udp_header_t *udp0; + u32 udp_len0; + + p0 = vlib_get_buffer (vm, packets[0]); + n_packets -= 1; + packets += 1; + + ip0 = (void *) (p0->data + ip_offset); + udp0 = (void *) (p0->data + udp_offset); + udp_len0 = clib_net_to_host_u16 (ip0->length) - sizeof (ip0[0]); + + if (flags & UDP_PG_EDIT_LENGTH) + udp0->length = + clib_net_to_host_u16 (vlib_buffer_length_in_chain (vm, p0) + - ip_offset); + + /* Initialize checksum with header. */ + if (flags & UDP_PG_EDIT_CHECKSUM) + { + ip_csum_t sum0; + + sum0 = clib_mem_unaligned (&ip0->src_address, u64); + + sum0 = ip_csum_with_carry + (sum0, clib_host_to_net_u32 (udp_len0 + (ip0->protocol << 16))); + + /* Invalidate possibly old checksum. */ + udp0->checksum = 0; + + sum0 = + ip_incremental_checksum_buffer (vm, p0, udp_offset, udp_len0, + sum0); + + sum0 = ~ip_csum_fold (sum0); + + /* Zero checksum means checksumming disabled. */ + sum0 = sum0 != 0 ? sum0 : 0xffff; + + udp0->checksum = sum0; + } + } +} + +static void +udp_pg_edit_function (pg_main_t * pg, + pg_stream_t * s, + pg_edit_group_t * g, u32 * packets, u32 n_packets) +{ + switch (g->edit_function_opaque) + { + case UDP_PG_EDIT_LENGTH: + udp_pg_edit_function_inline (pg, s, g, packets, n_packets, + UDP_PG_EDIT_LENGTH); + break; + + case UDP_PG_EDIT_CHECKSUM: + udp_pg_edit_function_inline (pg, s, g, packets, n_packets, + UDP_PG_EDIT_CHECKSUM); + break; + + case UDP_PG_EDIT_CHECKSUM | UDP_PG_EDIT_LENGTH: + udp_pg_edit_function_inline (pg, s, g, packets, n_packets, + UDP_PG_EDIT_CHECKSUM | UDP_PG_EDIT_LENGTH); + break; + + default: + ASSERT (0); + break; + } +} + +typedef struct +{ + pg_edit_t src_port, dst_port; + pg_edit_t length; + pg_edit_t checksum; +} pg_udp_header_t; + +static inline void +pg_udp_header_init (pg_udp_header_t * p) +{ + /* Initialize fields that are not bit fields in the IP header. */ +#define _(f) pg_edit_init (&p->f, udp_header_t, f); + _(src_port); + _(dst_port); + _(length); + _(checksum); +#undef _ +} + +uword +unformat_pg_udp_header (unformat_input_t * input, va_list * args) +{ + pg_stream_t *s = va_arg (*args, pg_stream_t *); + pg_udp_header_t *p; + u32 group_index; + + p = pg_create_edit_group (s, sizeof (p[0]), sizeof (udp_header_t), + &group_index); + pg_udp_header_init (p); + + /* Defaults. */ + p->checksum.type = PG_EDIT_UNSPECIFIED; + p->length.type = PG_EDIT_UNSPECIFIED; + + if (!unformat (input, "UDP: %U -> %U", + unformat_pg_edit, + unformat_tcp_udp_port, &p->src_port, + unformat_pg_edit, unformat_tcp_udp_port, &p->dst_port)) + goto error; + + /* Parse options. */ + while (1) + { + if (unformat (input, "length %U", + unformat_pg_edit, unformat_pg_number, &p->length)) + ; + + else if (unformat (input, "checksum %U", + unformat_pg_edit, unformat_pg_number, &p->checksum)) + ; + + /* Can't parse input: try next protocol level. */ + else + break; + } + + { + ip_main_t *im = &ip_main; + u16 dst_port; + tcp_udp_port_info_t *pi; + + pi = 0; + if (p->dst_port.type == PG_EDIT_FIXED) + { + dst_port = pg_edit_get_value (&p->dst_port, PG_EDIT_LO); + pi = ip_get_tcp_udp_port_info (im, dst_port); + } + + if (pi && pi->unformat_pg_edit + && unformat_user (input, pi->unformat_pg_edit, s)) + ; + + else if (!unformat_user (input, unformat_pg_payload, s)) + goto error; + + p = pg_get_edit_group (s, group_index); + if (p->checksum.type == PG_EDIT_UNSPECIFIED + || p->length.type == PG_EDIT_UNSPECIFIED) + { + pg_edit_group_t *g = pg_stream_get_group (s, group_index); + g->edit_function = udp_pg_edit_function; + g->edit_function_opaque = 0; + if (p->checksum.type == PG_EDIT_UNSPECIFIED) + g->edit_function_opaque |= UDP_PG_EDIT_CHECKSUM; + if (p->length.type == PG_EDIT_UNSPECIFIED) + g->edit_function_opaque |= UDP_PG_EDIT_LENGTH; + } + + return 1; + } + +error: + /* Free up any edits we may have added. */ + pg_free_edit_group (s); + return 0; +} + + +/* + * fd.io coding-style-patch-verification: ON + * + * Local Variables: + * eval: (c-set-style "gnu") + * End: + */ diff --git a/src/vnet/vnet_all_api_h.h b/src/vnet/vnet_all_api_h.h index 142acedc..c4075db6 100644 --- a/src/vnet/vnet_all_api_h.h +++ b/src/vnet/vnet_all_api_h.h @@ -51,6 +51,7 @@ #include #include #include +#include #include #include #include diff --git a/src/vnet/vxlan-gpe/vxlan_gpe.h b/src/vnet/vxlan-gpe/vxlan_gpe.h index 1b4bc44e..e768d230 100644 --- a/src/vnet/vxlan-gpe/vxlan_gpe.h +++ b/src/vnet/vxlan-gpe/vxlan_gpe.h @@ -29,7 +29,7 @@ #include #include #include -#include +#include /** * @brief VXLAN GPE header struct diff --git a/src/vnet/vxlan/vxlan.h b/src/vnet/vxlan/vxlan.h index adfa3a8e..dca1cd12 100644 --- a/src/vnet/vxlan/vxlan.h +++ b/src/vnet/vxlan/vxlan.h @@ -26,7 +26,7 @@ #include #include #include -#include +#include #include #include diff --git a/src/vpp/api/vpe.api b/src/vpp/api/vpe.api index 24f48293..2d6e4f37 100644 --- a/src/vpp/api/vpe.api +++ b/src/vpp/api/vpe.api @@ -38,6 +38,7 @@ * IPSEC-GRE APIs: see .../src/vnet/ipsec-gre/{ipsec_gre.api, ipsec_gre_api.c} * LISP APIs: see .../src/vnet/lisp/{lisp.api, lisp_api.c} * LISP-GPE APIs: see .../src/vnet/lisp-gpe/{lisp_gpe.api, lisp_gpe_api.c} + * SESSION APIs: .../vnet/session/{session.api session_api.c} * MPLS APIs: see .../src/vnet/mpls/{mpls.api, mpls_api.c} * SR APIs: see .../src/vnet/sr/{sr.api, sr_api.c} * DPDK APIs: see ... /src/vnet/devices/dpdk/{dpdk.api, dpdk_api.c} diff --git a/src/vppinfra.am b/src/vppinfra.am index 8d375958..4b9f0c29 100644 --- a/src/vppinfra.am +++ b/src/vppinfra.am @@ -157,7 +157,9 @@ nobase_include_HEADERS = \ vppinfra/asm_mips.h \ vppinfra/asm_x86.h \ vppinfra/bihash_8_8.h \ + vppinfra/bihash_16_8.h \ vppinfra/bihash_24_8.h \ + vppinfra/bihash_48_8.h \ vppinfra/bihash_template.h \ vppinfra/bihash_template.c \ vppinfra/bitmap.h \ @@ -206,6 +208,7 @@ nobase_include_HEADERS = \ vppinfra/timer.h \ vppinfra/tw_timer_2t_1w_2048sl.h \ vppinfra/tw_timer_16t_2w_512sl.h \ + vppinfra/tw_timer_16t_1w_2048sl.h \ vppinfra/tw_timer_template.h \ vppinfra/tw_timer_template.c \ vppinfra/types.h \ @@ -261,6 +264,8 @@ CLIB_CORE = \ vppinfra/tw_timer_2t_1w_2048sl.c \ vppinfra/tw_timer_16t_2w_512sl.h \ vppinfra/tw_timer_16t_2w_512sl.c \ + vppinfra/tw_timer_16t_1w_2048sl.h \ + vppinfra/tw_timer_16t_1w_2048sl.c \ vppinfra/unformat.c \ vppinfra/vec.c \ vppinfra/vector.c \ diff --git a/src/vppinfra/bihash_16_8.h b/src/vppinfra/bihash_16_8.h new file mode 100644 index 00000000..ce80f70e --- /dev/null +++ b/src/vppinfra/bihash_16_8.h @@ -0,0 +1,103 @@ +/* + * Copyright (c) 2016 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ +#undef BIHASH_TYPE + +#define BIHASH_TYPE _16_8 +#define BIHASH_KVP_PER_PAGE 4 + +#ifndef __included_bihash_16_8_h__ +#define __included_bihash_16_8_h__ + +#include +#include +#include +#include + +typedef struct +{ + u64 key[2]; + u64 value; +} clib_bihash_kv_16_8_t; + +static inline int +clib_bihash_is_free_16_8 (clib_bihash_kv_16_8_t * v) +{ + /* Free values are memset to 0xff, check a bit... */ + if (v->key[0] == ~0ULL && v->value == ~0ULL) + return 1; + return 0; +} + +#if __SSE4_2__ +#ifndef __defined_crc_u32__ +#define __defined_crc_u32__ +static inline u32 +crc_u32 (u32 data, u32 value) +{ + __asm__ volatile ("crc32l %[data], %[value];":[value] "+r" (value):[data] + "rm" (data)); + return value; +} +#endif /* __defined_crc_u32__ */ + +static inline u64 +clib_bihash_hash_16_8 (clib_bihash_kv_16_8_t * v) +{ + u32 *dp = (u32 *) & v->key[0]; + u32 value = 0; + + value = crc_u32 (dp[0], value); + value = crc_u32 (dp[1], value); + value = crc_u32 (dp[2], value); + value = crc_u32 (dp[3], value); + + return value; +} +#else +static inline u64 +clib_bihash_hash_16_8 (clib_bihash_kv_16_8_t * v) +{ + u64 tmp = v->key[0] ^ v->key[1]; + return clib_xxhash (tmp); +} +#endif + +static inline u8 * +format_bihash_kvp_16_8 (u8 * s, va_list * args) +{ + clib_bihash_kv_16_8_t *v = va_arg (*args, clib_bihash_kv_16_8_t *); + + s = format (s, "key %llu %llu value %llu", v->key[0], v->key[1], v->value); + return s; +} + +static inline int +clib_bihash_key_compare_16_8 (u64 * a, u64 * b) +{ + return ((a[0] ^ b[0]) | (a[1] ^ b[1])) == 0; +} + +#undef __included_bihash_template_h__ +#include + +#endif /* __included_bihash_16_8_h__ */ + +/* + * fd.io coding-style-patch-verification: ON + * + * Local Variables: + * eval: (c-set-style "gnu") + * End: + */ diff --git a/src/vppinfra/bihash_48_8.h b/src/vppinfra/bihash_48_8.h new file mode 100644 index 00000000..1a6e7691 --- /dev/null +++ b/src/vppinfra/bihash_48_8.h @@ -0,0 +1,116 @@ +/* + * Copyright (c) 2016 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +#undef BIHASH_TYPE + +#define BIHASH_TYPE _48_8 +#define BIHASH_KVP_PER_PAGE 4 + +#ifndef __included_bihash_48_8_h__ +#define __included_bihash_48_8_h__ + +#include +#include +#include +#include + +typedef struct +{ + u64 key[6]; + u64 value; +} clib_bihash_kv_48_8_t; + +static inline int +clib_bihash_is_free_48_8 (const clib_bihash_kv_48_8_t * v) +{ + /* Free values are memset to 0xff, check a bit... */ + if (v->key[0] == ~0ULL && v->value == ~0ULL) + return 1; + return 0; +} + +#if __SSE4_2__ +#ifndef __defined_crc_u32__ +#define __defined_crc_u32__ +static inline u32 +crc_u32 (u32 data, u32 value) +{ + __asm__ volatile ("crc32l %[data], %[value];":[value] "+r" (value):[data] + "rm" (data)); + return value; +} +#endif /* __defined_crc_u32__ */ + +static inline u64 +clib_bihash_hash_48_8 (const clib_bihash_kv_48_8_t * v) +{ + const u32 *dp = (const u32 *) &v->key[0]; + u32 value = 0; + + value = crc_u32 (dp[0], value); + value = crc_u32 (dp[1], value); + value = crc_u32 (dp[2], value); + value = crc_u32 (dp[3], value); + value = crc_u32 (dp[4], value); + value = crc_u32 (dp[5], value); + value = crc_u32 (dp[6], value); + value = crc_u32 (dp[7], value); + value = crc_u32 (dp[8], value); + value = crc_u32 (dp[9], value); + value = crc_u32 (dp[10], value); + value = crc_u32 (dp[11], value); + + return value; +} +#else +static inline u64 +clib_bihash_hash_48_8 (const clib_bihash_kv_48_8_t * v) +{ + u64 tmp = v->key[0] ^ v->key[1] ^ v->key[2] ^ v->key[3] ^ v->key[4] + ^ v->key[5]; + return clib_xxhash (tmp); +} +#endif + +static inline u8 * +format_bihash_kvp_48_8 (u8 * s, va_list * args) +{ + clib_bihash_kv_48_8_t *v = va_arg (*args, clib_bihash_kv_48_8_t *); + + s = format (s, "key %llu %llu %llu %llu %llu %llu value %llu", v->key[0], + v->key[1], v->key[2], v->key[3], v->key[4], v->key[5], + v->value); + return s; +} + +static inline int +clib_bihash_key_compare_48_8 (const u64 * a, const u64 * b) +{ + return ((a[0] ^ b[0]) | (a[1] ^ b[1]) | (a[2] ^ b[2]) | (a[3] ^ b[3]) + | (a[4] ^ b[4]) | (a[5] ^ b[5])) == 0; +} + +#undef __included_bihash_template_h__ +#include + +#endif /* __included_bihash_48_8_h__ */ + +/* + * fd.io coding-style-patch-verification: ON + * + * Local Variables: + * eval: (c-set-style "gnu") + * End: + */ diff --git a/src/vppinfra/tw_timer_16t_1w_2048sl.c b/src/vppinfra/tw_timer_16t_1w_2048sl.c new file mode 100644 index 00000000..3f342045 --- /dev/null +++ b/src/vppinfra/tw_timer_16t_1w_2048sl.c @@ -0,0 +1,26 @@ +/* + * Copyright (c) 2017 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +#include +#include "tw_timer_16t_1w_2048sl.h" +#include "tw_timer_template.c" + +/* + * fd.io coding-style-patch-verification: ON + * + * Local Variables: + * eval: (c-set-style "gnu") + * End: + */ diff --git a/src/vppinfra/tw_timer_16t_1w_2048sl.h b/src/vppinfra/tw_timer_16t_1w_2048sl.h new file mode 100644 index 00000000..685ac31e --- /dev/null +++ b/src/vppinfra/tw_timer_16t_1w_2048sl.h @@ -0,0 +1,46 @@ +/* + * Copyright (c) 2017 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +#ifndef __included_tw_timer_16t_2w_512sl_h__ +#define __included_tw_timer_16t_2w_512sl_h__ + +/* ... So that a client app can create multiple wheel geometries */ +#undef TW_TIMER_WHEELS +#undef TW_SLOTS_PER_RING +#undef TW_RING_SHIFT +#undef TW_RING_MASK +#undef TW_TIMERS_PER_OBJECT +#undef LOG2_TW_TIMERS_PER_OBJECT +#undef TW_SUFFIX + +#define TW_TIMER_WHEELS 1 +#define TW_SLOTS_PER_RING 2048 +#define TW_RING_SHIFT 11 +#define TW_RING_MASK (TW_SLOTS_PER_RING -1) +#define TW_TIMERS_PER_OBJECT 16 +#define LOG2_TW_TIMERS_PER_OBJECT 4 +#define TW_SUFFIX _16t_1w_2048sl + +#include + +#endif /* __included_tw_timer_16t_2w_512sl_h__ */ + +/* + * fd.io coding-style-patch-verification: ON + * + * Local Variables: + * eval: (c-set-style "gnu") + * End: + */ -- cgit 1.2.3-korg From 586afd762bfa149f5ca167bd5fd5a0cd59ce94fe Mon Sep 17 00:00:00 2001 From: Damjan Marion Date: Wed, 5 Apr 2017 19:18:20 +0200 Subject: Use thread local storage for thread index This patch deprecates stack-based thread identification, Also removes requirement that thread stacks are adjacent. Finally, possibly annoying for some folks, it renames all occurences of cpu_index and cpu_number with thread index. Using word "cpu" is misleading here as thread can be migrated ti different CPU, and also it is not related to linux cpu index. Change-Id: I68cdaf661e701d2336fc953dcb9978d10a70f7c1 Signed-off-by: Damjan Marion --- src/examples/srv6-sample-localsid/node.c | 4 +- src/plugins/dpdk/buffer.c | 2 +- src/plugins/dpdk/device/device.c | 8 +- src/plugins/dpdk/device/dpdk_priv.h | 8 +- src/plugins/dpdk/device/init.c | 2 +- src/plugins/dpdk/device/node.c | 32 +++--- src/plugins/dpdk/hqos/hqos.c | 16 +-- src/plugins/dpdk/ipsec/cli.c | 8 +- src/plugins/dpdk/ipsec/crypto_node.c | 4 +- src/plugins/dpdk/ipsec/esp.h | 4 +- src/plugins/dpdk/ipsec/esp_decrypt.c | 4 +- src/plugins/dpdk/ipsec/esp_encrypt.c | 5 +- src/plugins/dpdk/ipsec/ipsec.c | 2 +- src/plugins/dpdk/ipsec/ipsec.h | 4 +- src/plugins/dpdk/main.c | 2 +- src/plugins/flowperpkt/l2_node.c | 2 +- src/plugins/flowperpkt/node.c | 2 +- src/plugins/ioam/export-common/ioam_export.h | 6 +- .../ioam/ip6/ioam_cache_tunnel_select_node.c | 16 +-- src/plugins/ixge/ixge.c | 2 +- src/plugins/lb/lb.c | 8 +- src/plugins/lb/node.c | 22 ++-- src/plugins/lb/refcount.c | 8 +- src/plugins/lb/refcount.h | 4 +- src/plugins/memif/node.c | 35 +++--- src/plugins/snat/in2out.c | 110 +++++++++--------- src/plugins/snat/out2in.c | 102 ++++++++--------- src/plugins/snat/snat.h | 10 +- src/vlib/buffer.c | 6 +- src/vlib/buffer_funcs.h | 4 +- src/vlib/cli.c | 6 +- src/vlib/counter.h | 16 +-- src/vlib/error.c | 2 +- src/vlib/global_funcs.h | 2 +- src/vlib/main.c | 14 +-- src/vlib/main.h | 2 +- src/vlib/node.c | 2 +- src/vlib/node.h | 6 +- src/vlib/node_funcs.h | 8 +- src/vlib/threads.c | 69 ++++------- src/vlib/threads.h | 21 ++-- src/vlib/unix/cj.c | 7 +- src/vlib/unix/cj.h | 2 +- src/vlib/unix/main.c | 43 +++---- src/vnet/adj/adj_l2.c | 4 +- src/vnet/adj/adj_midchain.c | 8 +- src/vnet/adj/adj_nsh.c | 4 +- src/vnet/classify/vnet_classify.c | 16 +-- src/vnet/cop/ip4_whitelist.c | 8 +- src/vnet/cop/ip6_whitelist.c | 8 +- src/vnet/devices/af_packet/node.c | 20 ++-- src/vnet/devices/devices.c | 61 +++++----- src/vnet/devices/devices.h | 18 +-- src/vnet/devices/netmap/node.c | 24 ++-- src/vnet/devices/ssvm/node.c | 6 +- src/vnet/devices/virtio/vhost-user.c | 127 +++++++++++---------- src/vnet/dpo/lookup_dpo.c | 20 ++-- src/vnet/dpo/replicate_dpo.c | 12 +- src/vnet/ethernet/arp.c | 2 +- src/vnet/ethernet/interface.c | 7 +- src/vnet/ethernet/node.c | 14 +-- src/vnet/gre/node.c | 8 +- src/vnet/interface.h | 2 +- src/vnet/interface_output.c | 53 ++++----- src/vnet/ip/ip4_forward.c | 34 +++--- src/vnet/ip/ip4_input.c | 8 +- src/vnet/ip/ip6_forward.c | 24 ++-- src/vnet/ip/ip6_input.c | 8 +- src/vnet/ip/ip6_neighbor.c | 4 +- src/vnet/ipsec/esp.h | 8 +- src/vnet/ipsec/esp_decrypt.c | 13 ++- src/vnet/ipsec/esp_encrypt.c | 13 ++- src/vnet/ipsec/ikev2.c | 64 ++++++----- src/vnet/ipsec/ipsec.h | 12 +- src/vnet/ipsec/ipsec_if.c | 2 +- src/vnet/l2/l2_bvi.h | 2 +- src/vnet/l2/l2_input.c | 14 +-- src/vnet/l2/l2_output.c | 6 +- src/vnet/l2tp/decap.c | 2 +- src/vnet/l2tp/encap.c | 2 +- src/vnet/l2tp/l2tp.c | 6 +- src/vnet/lisp-gpe/decap.c | 16 +-- src/vnet/lldp/lldp_input.c | 2 +- src/vnet/map/ip4_map.c | 14 +-- src/vnet/map/ip4_map_t.c | 12 +- src/vnet/map/ip6_map.c | 19 +-- src/vnet/map/ip6_map_t.c | 12 +- src/vnet/mpls/mpls_input.c | 8 +- src/vnet/mpls/mpls_lookup.c | 20 ++-- src/vnet/mpls/mpls_output.c | 10 +- src/vnet/pg/input.c | 4 +- src/vnet/replication.c | 20 ++-- src/vnet/replication.h | 2 +- src/vnet/session/node.c | 2 +- src/vnet/sr/sr_localsid.c | 44 +++---- src/vnet/tcp/builtin_client.c | 2 +- src/vnet/tcp/tcp.c | 8 +- src/vnet/tcp/tcp_debug.h | 2 +- src/vnet/tcp/tcp_input.c | 10 +- src/vnet/tcp/tcp_output.c | 20 ++-- src/vnet/udp/udp_input.c | 2 +- src/vnet/unix/tapcli.c | 2 +- src/vnet/unix/tuntap.c | 4 +- src/vnet/vxlan-gpe/decap.c | 10 +- src/vnet/vxlan-gpe/encap.c | 12 +- src/vnet/vxlan/decap.c | 10 +- src/vnet/vxlan/encap.c | 12 +- src/vpp/stats/stats.c | 14 +-- src/vpp/stats/stats.h | 2 +- 109 files changed, 790 insertions(+), 791 deletions(-) (limited to 'src/vnet/vxlan-gpe') diff --git a/src/examples/srv6-sample-localsid/node.c b/src/examples/srv6-sample-localsid/node.c index 7bae9cd7..e83e2352 100644 --- a/src/examples/srv6-sample-localsid/node.c +++ b/src/examples/srv6-sample-localsid/node.c @@ -114,7 +114,7 @@ srv6_localsid_sample_fn (vlib_main_t * vm, vlib_node_runtime_t * node, vlib_fram from = vlib_frame_vector_args (frame); n_left_from = frame->n_vectors; next_index = node->cached_next_index; - u32 cpu_index = os_get_cpu_number (); + u32 thread_index = vlib_get_thread_index (); while (n_left_from > 0) { @@ -168,7 +168,7 @@ srv6_localsid_sample_fn (vlib_main_t * vm, vlib_node_runtime_t * node, vlib_fram /* This increments the SRv6 per LocalSID counters.*/ vlib_increment_combined_counter (((next0 == SRV6_SAMPLE_LOCALSID_NEXT_ERROR) ? &(sm->sr_ls_invalid_counters) : &(sm->sr_ls_valid_counters)), - cpu_index, + thread_index, ls0 - sm->localsids, 1, vlib_buffer_length_in_chain (vm, b0)); diff --git a/src/plugins/dpdk/buffer.c b/src/plugins/dpdk/buffer.c index 2765c292..c80b3fa8 100644 --- a/src/plugins/dpdk/buffer.c +++ b/src/plugins/dpdk/buffer.c @@ -132,7 +132,7 @@ dpdk_buffer_delete_free_list (vlib_main_t * vm, u32 free_list_index) u32 merge_index; int i; - ASSERT (os_get_cpu_number () == 0); + ASSERT (vlib_get_thread_index () == 0); f = vlib_buffer_get_free_list (vm, free_list_index); diff --git a/src/plugins/dpdk/device/device.c b/src/plugins/dpdk/device/device.c index 50b26689..91661246 100644 --- a/src/plugins/dpdk/device/device.c +++ b/src/plugins/dpdk/device/device.c @@ -243,7 +243,7 @@ static_always_inline ASSERT (ring->tx_tail == 0); n_retry = 16; - queue_id = vm->cpu_index; + queue_id = vm->thread_index; do { @@ -266,7 +266,7 @@ static_always_inline { /* no wrap, transmit in one burst */ dpdk_device_hqos_per_worker_thread_t *hqos = - &xd->hqos_wt[vm->cpu_index]; + &xd->hqos_wt[vm->thread_index]; ASSERT (hqos->swq != NULL); @@ -332,7 +332,7 @@ dpdk_buffer_recycle (vlib_main_t * vm, vlib_node_runtime_t * node, vlib_buffer_t * b, u32 bi, struct rte_mbuf **mbp) { dpdk_main_t *dm = &dpdk_main; - u32 my_cpu = vm->cpu_index; + u32 my_cpu = vm->thread_index; struct rte_mbuf *mb_new; if (PREDICT_FALSE (b->flags & VLIB_BUFFER_RECYCLE) == 0) @@ -376,7 +376,7 @@ dpdk_interface_tx (vlib_main_t * vm, tx_ring_hdr_t *ring; u32 n_on_ring; - my_cpu = vm->cpu_index; + my_cpu = vm->thread_index; queue_id = my_cpu; diff --git a/src/plugins/dpdk/device/dpdk_priv.h b/src/plugins/dpdk/device/dpdk_priv.h index dd40ff48..52b4ca4b 100644 --- a/src/plugins/dpdk/device/dpdk_priv.h +++ b/src/plugins/dpdk/device/dpdk_priv.h @@ -79,7 +79,7 @@ dpdk_update_counters (dpdk_device_t * xd, f64 now) { vlib_simple_counter_main_t *cm; vnet_main_t *vnm = vnet_get_main (); - u32 my_cpu = os_get_cpu_number (); + u32 thread_index = vlib_get_thread_index (); u64 rxerrors, last_rxerrors; /* only update counters for PMD interfaces */ @@ -96,7 +96,7 @@ dpdk_update_counters (dpdk_device_t * xd, f64 now) cm = vec_elt_at_index (vnm->interface_main.sw_if_counters, VNET_INTERFACE_COUNTER_RX_NO_BUF); - vlib_increment_simple_counter (cm, my_cpu, xd->vlib_sw_if_index, + vlib_increment_simple_counter (cm, thread_index, xd->vlib_sw_if_index, xd->stats.rx_nombuf - xd->last_stats.rx_nombuf); } @@ -107,7 +107,7 @@ dpdk_update_counters (dpdk_device_t * xd, f64 now) cm = vec_elt_at_index (vnm->interface_main.sw_if_counters, VNET_INTERFACE_COUNTER_RX_MISS); - vlib_increment_simple_counter (cm, my_cpu, xd->vlib_sw_if_index, + vlib_increment_simple_counter (cm, thread_index, xd->vlib_sw_if_index, xd->stats.imissed - xd->last_stats.imissed); } @@ -119,7 +119,7 @@ dpdk_update_counters (dpdk_device_t * xd, f64 now) cm = vec_elt_at_index (vnm->interface_main.sw_if_counters, VNET_INTERFACE_COUNTER_RX_ERROR); - vlib_increment_simple_counter (cm, my_cpu, xd->vlib_sw_if_index, + vlib_increment_simple_counter (cm, thread_index, xd->vlib_sw_if_index, rxerrors - last_rxerrors); } diff --git a/src/plugins/dpdk/device/init.c b/src/plugins/dpdk/device/init.c index 538db6cb..7eaf8da7 100755 --- a/src/plugins/dpdk/device/init.c +++ b/src/plugins/dpdk/device/init.c @@ -324,7 +324,7 @@ dpdk_port_setup (dpdk_main_t * dm, dpdk_device_t * xd) int rv; int j; - ASSERT (os_get_cpu_number () == 0); + ASSERT (vlib_get_thread_index () == 0); if (xd->flags & DPDK_DEVICE_FLAG_ADMIN_UP) { diff --git a/src/plugins/dpdk/device/node.c b/src/plugins/dpdk/device/node.c index e740fd18..b10e0fad 100644 --- a/src/plugins/dpdk/device/node.c +++ b/src/plugins/dpdk/device/node.c @@ -283,7 +283,7 @@ dpdk_buffer_init_from_template (void *d0, void *d1, void *d2, void *d3, */ static_always_inline u32 dpdk_device_input (dpdk_main_t * dm, dpdk_device_t * xd, - vlib_node_runtime_t * node, u32 cpu_index, u16 queue_id, + vlib_node_runtime_t * node, u32 thread_index, u16 queue_id, int maybe_multiseg) { u32 n_buffers; @@ -294,7 +294,7 @@ dpdk_device_input (dpdk_main_t * dm, dpdk_device_t * xd, uword n_rx_bytes = 0; u32 n_trace, trace_cnt __attribute__ ((unused)); vlib_buffer_free_list_t *fl; - vlib_buffer_t *bt = vec_elt_at_index (dm->buffer_templates, cpu_index); + vlib_buffer_t *bt = vec_elt_at_index (dm->buffer_templates, thread_index); if ((xd->flags & DPDK_DEVICE_FLAG_ADMIN_UP) == 0) return 0; @@ -306,7 +306,7 @@ dpdk_device_input (dpdk_main_t * dm, dpdk_device_t * xd, return 0; } - vec_reset_length (xd->d_trace_buffers[cpu_index]); + vec_reset_length (xd->d_trace_buffers[thread_index]); trace_cnt = n_trace = vlib_get_trace_count (vm, node); if (n_trace > 0) @@ -318,7 +318,7 @@ dpdk_device_input (dpdk_main_t * dm, dpdk_device_t * xd, { struct rte_mbuf *mb = xd->rx_vectors[queue_id][mb_index++]; vlib_buffer_t *b = vlib_buffer_from_rte_mbuf (mb); - vec_add1 (xd->d_trace_buffers[cpu_index], + vec_add1 (xd->d_trace_buffers[thread_index], vlib_get_buffer_index (vm, b)); } } @@ -546,20 +546,22 @@ dpdk_device_input (dpdk_main_t * dm, dpdk_device_t * xd, vlib_put_next_frame (vm, node, next_index, n_left_to_next); } - if (PREDICT_FALSE (vec_len (xd->d_trace_buffers[cpu_index]) > 0)) + if (PREDICT_FALSE (vec_len (xd->d_trace_buffers[thread_index]) > 0)) { - dpdk_rx_trace (dm, node, xd, queue_id, xd->d_trace_buffers[cpu_index], - vec_len (xd->d_trace_buffers[cpu_index])); - vlib_set_trace_count (vm, node, n_trace - - vec_len (xd->d_trace_buffers[cpu_index])); + dpdk_rx_trace (dm, node, xd, queue_id, + xd->d_trace_buffers[thread_index], + vec_len (xd->d_trace_buffers[thread_index])); + vlib_set_trace_count (vm, node, + n_trace - + vec_len (xd->d_trace_buffers[thread_index])); } vlib_increment_combined_counter (vnet_get_main ()->interface_main.combined_sw_if_counters + VNET_INTERFACE_COUNTER_RX, - cpu_index, xd->vlib_sw_if_index, mb_index, n_rx_bytes); + thread_index, xd->vlib_sw_if_index, mb_index, n_rx_bytes); - vnet_device_increment_rx_packets (cpu_index, mb_index); + vnet_device_increment_rx_packets (thread_index, mb_index); return mb_index; } @@ -630,19 +632,19 @@ dpdk_input (vlib_main_t * vm, vlib_node_runtime_t * node, vlib_frame_t * f) dpdk_device_t *xd; uword n_rx_packets = 0; dpdk_device_and_queue_t *dq; - u32 cpu_index = os_get_cpu_number (); + u32 thread_index = vlib_get_thread_index (); /* * Poll all devices on this cpu for input/interrupts. */ /* *INDENT-OFF* */ - vec_foreach (dq, dm->devices_by_cpu[cpu_index]) + vec_foreach (dq, dm->devices_by_cpu[thread_index]) { xd = vec_elt_at_index(dm->devices, dq->device); if (xd->flags & DPDK_DEVICE_FLAG_MAYBE_MULTISEG) - n_rx_packets += dpdk_device_input (dm, xd, node, cpu_index, dq->queue_id, /* maybe_multiseg */ 1); + n_rx_packets += dpdk_device_input (dm, xd, node, thread_index, dq->queue_id, /* maybe_multiseg */ 1); else - n_rx_packets += dpdk_device_input (dm, xd, node, cpu_index, dq->queue_id, /* maybe_multiseg */ 0); + n_rx_packets += dpdk_device_input (dm, xd, node, thread_index, dq->queue_id, /* maybe_multiseg */ 0); } /* *INDENT-ON* */ diff --git a/src/plugins/dpdk/hqos/hqos.c b/src/plugins/dpdk/hqos/hqos.c index a288fca7..8b251beb 100644 --- a/src/plugins/dpdk/hqos/hqos.c +++ b/src/plugins/dpdk/hqos/hqos.c @@ -397,7 +397,7 @@ static_always_inline void dpdk_hqos_thread_internal_hqos_dbg_bypass (vlib_main_t * vm) { dpdk_main_t *dm = &dpdk_main; - u32 cpu_index = vm->cpu_index; + u32 thread_index = vm->thread_index; u32 dev_pos; dev_pos = 0; @@ -405,12 +405,12 @@ dpdk_hqos_thread_internal_hqos_dbg_bypass (vlib_main_t * vm) { vlib_worker_thread_barrier_check (); - u32 n_devs = vec_len (dm->devices_by_hqos_cpu[cpu_index]); + u32 n_devs = vec_len (dm->devices_by_hqos_cpu[thread_index]); if (dev_pos >= n_devs) dev_pos = 0; dpdk_device_and_queue_t *dq = - vec_elt_at_index (dm->devices_by_hqos_cpu[cpu_index], dev_pos); + vec_elt_at_index (dm->devices_by_hqos_cpu[thread_index], dev_pos); dpdk_device_t *xd = vec_elt_at_index (dm->devices, dq->device); dpdk_device_hqos_per_hqos_thread_t *hqos = xd->hqos_ht; @@ -479,7 +479,7 @@ static_always_inline void dpdk_hqos_thread_internal (vlib_main_t * vm) { dpdk_main_t *dm = &dpdk_main; - u32 cpu_index = vm->cpu_index; + u32 thread_index = vm->thread_index; u32 dev_pos; dev_pos = 0; @@ -487,7 +487,7 @@ dpdk_hqos_thread_internal (vlib_main_t * vm) { vlib_worker_thread_barrier_check (); - u32 n_devs = vec_len (dm->devices_by_hqos_cpu[cpu_index]); + u32 n_devs = vec_len (dm->devices_by_hqos_cpu[thread_index]); if (PREDICT_FALSE (n_devs == 0)) { dev_pos = 0; @@ -497,7 +497,7 @@ dpdk_hqos_thread_internal (vlib_main_t * vm) dev_pos = 0; dpdk_device_and_queue_t *dq = - vec_elt_at_index (dm->devices_by_hqos_cpu[cpu_index], dev_pos); + vec_elt_at_index (dm->devices_by_hqos_cpu[thread_index], dev_pos); dpdk_device_t *xd = vec_elt_at_index (dm->devices, dq->device); dpdk_device_hqos_per_hqos_thread_t *hqos = xd->hqos_ht; @@ -586,7 +586,7 @@ dpdk_hqos_thread (vlib_worker_thread_t * w) vm = vlib_get_main (); - ASSERT (vm->cpu_index == os_get_cpu_number ()); + ASSERT (vm->thread_index == vlib_get_thread_index ()); clib_time_init (&vm->clib_time); clib_mem_set_heap (w->thread_mheap); @@ -595,7 +595,7 @@ dpdk_hqos_thread (vlib_worker_thread_t * w) while (tm->worker_thread_release == 0) vlib_worker_thread_barrier_check (); - if (vec_len (dm->devices_by_hqos_cpu[vm->cpu_index]) == 0) + if (vec_len (dm->devices_by_hqos_cpu[vm->thread_index]) == 0) return clib_error ("current I/O TX thread does not have any devices assigned to it"); diff --git a/src/plugins/dpdk/ipsec/cli.c b/src/plugins/dpdk/ipsec/cli.c index cd0a6037..3ae8c9b8 100644 --- a/src/plugins/dpdk/ipsec/cli.c +++ b/src/plugins/dpdk/ipsec/cli.c @@ -42,8 +42,8 @@ dpdk_ipsec_show_mapping (vlib_main_t * vm, u16 detail_display) for (i = 0; i < tm->n_vlib_mains; i++) { uword key, data; - u32 cpu_index = vlib_mains[i]->cpu_index; - crypto_worker_main_t *cwm = &dcm->workers_main[cpu_index]; + u32 thread_index = vlib_mains[i]->thread_index; + crypto_worker_main_t *cwm = &dcm->workers_main[thread_index]; u8 *s = 0; if (skip_master) @@ -57,7 +57,7 @@ dpdk_ipsec_show_mapping (vlib_main_t * vm, u16 detail_display) i32 last_cdev = -1; crypto_qp_data_t *qpd; - s = format (s, "%u\t", cpu_index); + s = format (s, "%u\t", thread_index); /* *INDENT-OFF* */ vec_foreach (qpd, cwm->qp_data) @@ -95,7 +95,7 @@ dpdk_ipsec_show_mapping (vlib_main_t * vm, u16 detail_display) cap.sym.auth.algo = p_key->auth_algo; check_algo_is_supported (&cap, auth_str); vlib_cli_output (vm, "%u\t%10s\t%15s\t%3s\t%u\t%u\n", - vlib_mains[i]->cpu_index, cipher_str, auth_str, + vlib_mains[i]->thread_index, cipher_str, auth_str, p_key->is_outbound ? "out" : "in", cwm->qp_data[data].dev_id, cwm->qp_data[data].qp_id); diff --git a/src/plugins/dpdk/ipsec/crypto_node.c b/src/plugins/dpdk/ipsec/crypto_node.c index dc3452b2..a3c45902 100644 --- a/src/plugins/dpdk/ipsec/crypto_node.c +++ b/src/plugins/dpdk/ipsec/crypto_node.c @@ -171,9 +171,9 @@ static uword dpdk_crypto_input_fn (vlib_main_t * vm, vlib_node_runtime_t * node, vlib_frame_t * frame) { - u32 cpu_index = os_get_cpu_number (); + u32 thread_index = vlib_get_thread_index (); dpdk_crypto_main_t *dcm = &dpdk_crypto_main; - crypto_worker_main_t *cwm = &dcm->workers_main[cpu_index]; + crypto_worker_main_t *cwm = &dcm->workers_main[thread_index]; crypto_qp_data_t *qpd; u32 n_deq = 0; diff --git a/src/plugins/dpdk/ipsec/esp.h b/src/plugins/dpdk/ipsec/esp.h index 320295b1..56f0c756 100644 --- a/src/plugins/dpdk/ipsec/esp.h +++ b/src/plugins/dpdk/ipsec/esp.h @@ -170,9 +170,9 @@ static_always_inline int create_sym_sess (ipsec_sa_t * sa, crypto_sa_session_t * sa_sess, u8 is_outbound) { - u32 cpu_index = os_get_cpu_number (); + u32 thread_index = vlib_get_thread_index (); dpdk_crypto_main_t *dcm = &dpdk_crypto_main; - crypto_worker_main_t *cwm = &dcm->workers_main[cpu_index]; + crypto_worker_main_t *cwm = &dcm->workers_main[thread_index]; struct rte_crypto_sym_xform cipher_xform = { 0 }; struct rte_crypto_sym_xform auth_xform = { 0 }; struct rte_crypto_sym_xform *xfs; diff --git a/src/plugins/dpdk/ipsec/esp_decrypt.c b/src/plugins/dpdk/ipsec/esp_decrypt.c index 286e03f8..bab76e3b 100644 --- a/src/plugins/dpdk/ipsec/esp_decrypt.c +++ b/src/plugins/dpdk/ipsec/esp_decrypt.c @@ -88,7 +88,7 @@ dpdk_esp_decrypt_node_fn (vlib_main_t * vm, { u32 n_left_from, *from, *to_next, next_index; ipsec_main_t *im = &ipsec_main; - u32 cpu_index = os_get_cpu_number(); + u32 thread_index = vlib_get_thread_index(); dpdk_crypto_main_t * dcm = &dpdk_crypto_main; dpdk_esp_main_t * em = &dpdk_esp_main; u32 i; @@ -104,7 +104,7 @@ dpdk_esp_decrypt_node_fn (vlib_main_t * vm, return n_left_from; } - crypto_worker_main_t *cwm = vec_elt_at_index(dcm->workers_main, cpu_index); + crypto_worker_main_t *cwm = vec_elt_at_index(dcm->workers_main, thread_index); u32 n_qps = vec_len(cwm->qp_data); struct rte_crypto_op ** cops_to_enq[n_qps]; u32 n_cop_qp[n_qps], * bi_to_enq[n_qps]; diff --git a/src/plugins/dpdk/ipsec/esp_encrypt.c b/src/plugins/dpdk/ipsec/esp_encrypt.c index 5b03de73..f996d7df 100644 --- a/src/plugins/dpdk/ipsec/esp_encrypt.c +++ b/src/plugins/dpdk/ipsec/esp_encrypt.c @@ -93,7 +93,7 @@ dpdk_esp_encrypt_node_fn (vlib_main_t * vm, { u32 n_left_from, *from, *to_next, next_index; ipsec_main_t *im = &ipsec_main; - u32 cpu_index = os_get_cpu_number (); + u32 thread_index = vlib_get_thread_index (); dpdk_crypto_main_t *dcm = &dpdk_crypto_main; dpdk_esp_main_t *em = &dpdk_esp_main; u32 i; @@ -111,7 +111,8 @@ dpdk_esp_encrypt_node_fn (vlib_main_t * vm, return n_left_from; } - crypto_worker_main_t *cwm = vec_elt_at_index (dcm->workers_main, cpu_index); + crypto_worker_main_t *cwm = + vec_elt_at_index (dcm->workers_main, thread_index); u32 n_qps = vec_len (cwm->qp_data); struct rte_crypto_op **cops_to_enq[n_qps]; u32 n_cop_qp[n_qps], *bi_to_enq[n_qps]; diff --git a/src/plugins/dpdk/ipsec/ipsec.c b/src/plugins/dpdk/ipsec/ipsec.c index b0aaaaec..5d8f4fba 100644 --- a/src/plugins/dpdk/ipsec/ipsec.c +++ b/src/plugins/dpdk/ipsec/ipsec.c @@ -289,7 +289,7 @@ dpdk_ipsec_process (vlib_main_t * vm, vlib_node_runtime_t * rt, if (!map) { clib_warning ("unable to create hash table for worker %u", - vlib_mains[i]->cpu_index); + vlib_mains[i]->thread_index); goto error; } cwm->algo_qp_map = map; diff --git a/src/plugins/dpdk/ipsec/ipsec.h b/src/plugins/dpdk/ipsec/ipsec.h index 28bffc80..f0f793c0 100644 --- a/src/plugins/dpdk/ipsec/ipsec.h +++ b/src/plugins/dpdk/ipsec/ipsec.h @@ -95,8 +95,8 @@ static_always_inline void crypto_alloc_cops () { dpdk_crypto_main_t *dcm = &dpdk_crypto_main; - u32 cpu_index = os_get_cpu_number (); - crypto_worker_main_t *cwm = &dcm->workers_main[cpu_index]; + u32 thread_index = vlib_get_thread_index (); + crypto_worker_main_t *cwm = &dcm->workers_main[thread_index]; unsigned socket_id = rte_socket_id (); crypto_qp_data_t *qpd; diff --git a/src/plugins/dpdk/main.c b/src/plugins/dpdk/main.c index 7ee2a785..942b8b2d 100644 --- a/src/plugins/dpdk/main.c +++ b/src/plugins/dpdk/main.c @@ -39,7 +39,7 @@ rte_delay_us_override (unsigned us) * thread then do not intercept. (Must not be called from an * independent pthread). */ - if (os_get_cpu_number () == 0) + if (vlib_get_thread_index () == 0) { /* * We're in the vlib main thread or a vlib process. Make sure diff --git a/src/plugins/flowperpkt/l2_node.c b/src/plugins/flowperpkt/l2_node.c index 1c2f681e..fdaf81d1 100644 --- a/src/plugins/flowperpkt/l2_node.c +++ b/src/plugins/flowperpkt/l2_node.c @@ -102,7 +102,7 @@ add_to_flow_record_l2 (vlib_main_t * vm, u8 * src_mac, u8 * dst_mac, u16 ethertype, u64 timestamp, u16 length, int do_flush) { - u32 my_cpu_number = vm->cpu_index; + u32 my_cpu_number = vm->thread_index; flow_report_main_t *frm = &flow_report_main; ip4_header_t *ip; udp_header_t *udp; diff --git a/src/plugins/flowperpkt/node.c b/src/plugins/flowperpkt/node.c index f77f087d..0277682d 100644 --- a/src/plugins/flowperpkt/node.c +++ b/src/plugins/flowperpkt/node.c @@ -101,7 +101,7 @@ add_to_flow_record_ipv4 (vlib_main_t * vm, u32 src_address, u32 dst_address, u8 tos, u64 timestamp, u16 length, int do_flush) { - u32 my_cpu_number = vm->cpu_index; + u32 my_cpu_number = vm->thread_index; flow_report_main_t *frm = &flow_report_main; ip4_header_t *ip; udp_header_t *udp; diff --git a/src/plugins/ioam/export-common/ioam_export.h b/src/plugins/ioam/export-common/ioam_export.h index 2bf3fd54..9de0d13b 100644 --- a/src/plugins/ioam/export-common/ioam_export.h +++ b/src/plugins/ioam/export-common/ioam_export.h @@ -477,8 +477,8 @@ do { \ from = vlib_frame_vector_args (F); \ n_left_from = (F)->n_vectors; \ next_index = (N)->cached_next_index; \ - while (__sync_lock_test_and_set ((EM)->lockp[(VM)->cpu_index], 1)); \ - my_buf = ioam_export_get_my_buffer (EM, (VM)->cpu_index); \ + while (__sync_lock_test_and_set ((EM)->lockp[(VM)->thread_index], 1)); \ + my_buf = ioam_export_get_my_buffer (EM, (VM)->thread_index); \ my_buf->touched_at = vlib_time_now (VM); \ while (n_left_from > 0) \ { \ @@ -620,7 +620,7 @@ do { \ } \ vlib_node_increment_counter (VM, export_node.index, \ EXPORT_ERROR_RECORDED, pkts_recorded); \ - *(EM)->lockp[(VM)->cpu_index] = 0; \ + *(EM)->lockp[(VM)->thread_index] = 0; \ } while(0) #endif /* __included_ioam_export_h__ */ diff --git a/src/plugins/ioam/ip6/ioam_cache_tunnel_select_node.c b/src/plugins/ioam/ip6/ioam_cache_tunnel_select_node.c index a56dc040..0cf742c9 100644 --- a/src/plugins/ioam/ip6/ioam_cache_tunnel_select_node.c +++ b/src/plugins/ioam/ip6/ioam_cache_tunnel_select_node.c @@ -396,7 +396,7 @@ ip6_reset_ts_hbh_node_fn (vlib_main_t * vm, clib_net_to_host_u32 (tcp0->seq_number) + 1, no_of_responses, now, - vm->cpu_index, &pool_index0)) + vm->thread_index, &pool_index0)) { cache_ts_added++; } @@ -419,7 +419,7 @@ ip6_reset_ts_hbh_node_fn (vlib_main_t * vm, e2e = (ioam_e2e_cache_option_t *) ((u8 *) hbh0 + cm->rewrite_pool_index_offset); - e2e->pool_id = (u8) vm->cpu_index; + e2e->pool_id = (u8) vm->thread_index; e2e->pool_index = pool_index0; ioam_e2e_id_rewrite_handler ((ioam_e2e_id_option_t *) ((u8 *) e2e + @@ -455,7 +455,7 @@ ip6_reset_ts_hbh_node_fn (vlib_main_t * vm, clib_net_to_host_u32 (tcp1->seq_number) + 1, no_of_responses, now, - vm->cpu_index, &pool_index1)) + vm->thread_index, &pool_index1)) { cache_ts_added++; } @@ -479,7 +479,7 @@ ip6_reset_ts_hbh_node_fn (vlib_main_t * vm, e2e = (ioam_e2e_cache_option_t *) ((u8 *) hbh1 + cm->rewrite_pool_index_offset); - e2e->pool_id = (u8) vm->cpu_index; + e2e->pool_id = (u8) vm->thread_index; e2e->pool_index = pool_index1; ioam_e2e_id_rewrite_handler ((ioam_e2e_id_option_t *) ((u8 *) e2e + @@ -562,7 +562,7 @@ ip6_reset_ts_hbh_node_fn (vlib_main_t * vm, clib_net_to_host_u32 (tcp0->seq_number) + 1, no_of_responses, now, - vm->cpu_index, &pool_index0)) + vm->thread_index, &pool_index0)) { cache_ts_added++; } @@ -585,7 +585,7 @@ ip6_reset_ts_hbh_node_fn (vlib_main_t * vm, e2e = (ioam_e2e_cache_option_t *) ((u8 *) hbh0 + cm->rewrite_pool_index_offset); - e2e->pool_id = (u8) vm->cpu_index; + e2e->pool_id = (u8) vm->thread_index; e2e->pool_index = pool_index0; ioam_e2e_id_rewrite_handler ((ioam_e2e_id_option_t *) ((u8 *) e2e + @@ -701,7 +701,7 @@ expired_cache_ts_timer_callback (u32 * expired_timers) ioam_cache_main_t *cm = &ioam_cache_main; int i; u32 pool_index; - u32 thread_index = os_get_cpu_number (); + u32 thread_index = vlib_get_thread_index (); u32 count = 0; for (i = 0; i < vec_len (expired_timers); i++) @@ -724,7 +724,7 @@ ioam_cache_ts_timer_tick_node_fn (vlib_main_t * vm, vlib_frame_t * f) { ioam_cache_main_t *cm = &ioam_cache_main; - u32 my_thread_index = os_get_cpu_number (); + u32 my_thread_index = vlib_get_thread_index (); struct timespec ts, tsrem; tw_timer_expire_timers_16t_2w_512sl (&cm->timer_wheels[my_thread_index], diff --git a/src/plugins/ixge/ixge.c b/src/plugins/ixge/ixge.c index f3c5cc09..08f5b692 100644 --- a/src/plugins/ixge/ixge.c +++ b/src/plugins/ixge/ixge.c @@ -1887,7 +1887,7 @@ done: vlib_increment_combined_counter (vnet_main. interface_main.combined_sw_if_counters + VNET_INTERFACE_COUNTER_RX, - 0 /* cpu_index */ , + 0 /* thread_index */ , xd->vlib_sw_if_index, n_packets, dq->rx.n_bytes); diff --git a/src/plugins/lb/lb.c b/src/plugins/lb/lb.c index add81236..addc2a42 100644 --- a/src/plugins/lb/lb.c +++ b/src/plugins/lb/lb.c @@ -63,11 +63,11 @@ u8 *format_lb_main (u8 * s, va_list * args) s = format(s, " #vips: %u\n", pool_elts(lbm->vips)); s = format(s, " #ass: %u\n", pool_elts(lbm->ass) - 1); - u32 cpu_index; - for(cpu_index = 0; cpu_index < tm->n_vlib_mains; cpu_index++ ) { - lb_hash_t *h = lbm->per_cpu[cpu_index].sticky_ht; + u32 thread_index; + for(thread_index = 0; thread_index < tm->n_vlib_mains; thread_index++ ) { + lb_hash_t *h = lbm->per_cpu[thread_index].sticky_ht; if (h) { - s = format(s, "core %d\n", cpu_index); + s = format(s, "core %d\n", thread_index); s = format(s, " timeout: %ds\n", h->timeout); s = format(s, " usage: %d / %d\n", lb_hash_elts(h, lb_hash_time_now(vlib_get_main())), lb_hash_size(h)); } diff --git a/src/plugins/lb/node.c b/src/plugins/lb/node.c index 8b763c53..3171148b 100644 --- a/src/plugins/lb/node.c +++ b/src/plugins/lb/node.c @@ -60,10 +60,10 @@ format_lb_trace (u8 * s, va_list * args) return s; } -lb_hash_t *lb_get_sticky_table(u32 cpu_index) +lb_hash_t *lb_get_sticky_table(u32 thread_index) { lb_main_t *lbm = &lb_main; - lb_hash_t *sticky_ht = lbm->per_cpu[cpu_index].sticky_ht; + lb_hash_t *sticky_ht = lbm->per_cpu[thread_index].sticky_ht; //Check if size changed if (PREDICT_FALSE(sticky_ht && (lbm->per_cpu_sticky_buckets != lb_hash_nbuckets(sticky_ht)))) { @@ -71,8 +71,8 @@ lb_hash_t *lb_get_sticky_table(u32 cpu_index) lb_hash_bucket_t *b; u32 i; lb_hash_foreach_entry(sticky_ht, b, i) { - vlib_refcount_add(&lbm->as_refcount, cpu_index, b->value[i], -1); - vlib_refcount_add(&lbm->as_refcount, cpu_index, 0, 1); + vlib_refcount_add(&lbm->as_refcount, thread_index, b->value[i], -1); + vlib_refcount_add(&lbm->as_refcount, thread_index, 0, 1); } lb_hash_free(sticky_ht); @@ -81,8 +81,8 @@ lb_hash_t *lb_get_sticky_table(u32 cpu_index) //Create if necessary if (PREDICT_FALSE(sticky_ht == NULL)) { - lbm->per_cpu[cpu_index].sticky_ht = lb_hash_alloc(lbm->per_cpu_sticky_buckets, lbm->flow_timeout); - sticky_ht = lbm->per_cpu[cpu_index].sticky_ht; + lbm->per_cpu[thread_index].sticky_ht = lb_hash_alloc(lbm->per_cpu_sticky_buckets, lbm->flow_timeout); + sticky_ht = lbm->per_cpu[thread_index].sticky_ht; clib_warning("Regenerated sticky table %p", sticky_ht); } @@ -153,10 +153,10 @@ lb_node_fn (vlib_main_t * vm, { lb_main_t *lbm = &lb_main; u32 n_left_from, *from, next_index, *to_next, n_left_to_next; - u32 cpu_index = os_get_cpu_number(); + u32 thread_index = vlib_get_thread_index(); u32 lb_time = lb_hash_time_now(vm); - lb_hash_t *sticky_ht = lb_get_sticky_table(cpu_index); + lb_hash_t *sticky_ht = lb_get_sticky_table(thread_index); from = vlib_frame_vector_args (frame); n_left_from = frame->n_vectors; next_index = node->cached_next_index; @@ -240,9 +240,9 @@ lb_node_fn (vlib_main_t * vm, //Configuration may be changed, vectors resized, etc... //Dereference previously used - vlib_refcount_add(&lbm->as_refcount, cpu_index, + vlib_refcount_add(&lbm->as_refcount, thread_index, lb_hash_available_value(sticky_ht, hash0, available_index0), -1); - vlib_refcount_add(&lbm->as_refcount, cpu_index, + vlib_refcount_add(&lbm->as_refcount, thread_index, asindex0, 1); //Add sticky entry @@ -260,7 +260,7 @@ lb_node_fn (vlib_main_t * vm, } vlib_increment_simple_counter(&lbm->vip_counters[counter], - cpu_index, + thread_index, vnet_buffer (p0)->ip.adj_index[VLIB_TX], 1); diff --git a/src/plugins/lb/refcount.c b/src/plugins/lb/refcount.c index 22415c88..6f01ab5a 100644 --- a/src/plugins/lb/refcount.c +++ b/src/plugins/lb/refcount.c @@ -31,10 +31,10 @@ u64 vlib_refcount_get(vlib_refcount_t *r, u32 index) { u64 count = 0; vlib_thread_main_t *tm = vlib_get_thread_main (); - u32 cpu_index; - for (cpu_index = 0; cpu_index < tm->n_vlib_mains; cpu_index++) { - if (r->per_cpu[cpu_index].length > index) - count += r->per_cpu[cpu_index].counters[index]; + u32 thread_index; + for (thread_index = 0; thread_index < tm->n_vlib_mains; thread_index++) { + if (r->per_cpu[thread_index].length > index) + count += r->per_cpu[thread_index].counters[index]; } return count; } diff --git a/src/plugins/lb/refcount.h b/src/plugins/lb/refcount.h index 8c26e7be..dcfcb3fe 100644 --- a/src/plugins/lb/refcount.h +++ b/src/plugins/lb/refcount.h @@ -45,9 +45,9 @@ typedef struct { void __vlib_refcount_resize(vlib_refcount_per_cpu_t *per_cpu, u32 size); static_always_inline -void vlib_refcount_add(vlib_refcount_t *r, u32 cpu_index, u32 counter_index, i32 v) +void vlib_refcount_add(vlib_refcount_t *r, u32 thread_index, u32 counter_index, i32 v) { - vlib_refcount_per_cpu_t *per_cpu = &r->per_cpu[cpu_index]; + vlib_refcount_per_cpu_t *per_cpu = &r->per_cpu[thread_index]; if (PREDICT_FALSE(counter_index >= per_cpu->length)) __vlib_refcount_resize(per_cpu, clib_max(counter_index + 16, per_cpu->length * 2)); diff --git a/src/plugins/memif/node.c b/src/plugins/memif/node.c index 659d5dfb..cee1f3d1 100644 --- a/src/plugins/memif/node.c +++ b/src/plugins/memif/node.c @@ -94,7 +94,7 @@ memif_device_input_inline (vlib_main_t * vm, vlib_node_runtime_t * node, u32 n_rx_bytes = 0; u32 *to_next = 0; u32 n_free_bufs; - u32 cpu_index = os_get_cpu_number (); + u32 thread_index = vlib_get_thread_index (); u32 bi0, bi1; vlib_buffer_t *b0, *b1; u16 ring_size = 1 << mif->log2_ring_size; @@ -105,14 +105,15 @@ memif_device_input_inline (vlib_main_t * vm, vlib_node_runtime_t * node, if (mif->per_interface_next_index != ~0) next_index = mif->per_interface_next_index; - n_free_bufs = vec_len (nm->rx_buffers[cpu_index]); + n_free_bufs = vec_len (nm->rx_buffers[thread_index]); if (PREDICT_FALSE (n_free_bufs < ring_size)) { - vec_validate (nm->rx_buffers[cpu_index], ring_size + n_free_bufs - 1); + vec_validate (nm->rx_buffers[thread_index], + ring_size + n_free_bufs - 1); n_free_bufs += - vlib_buffer_alloc (vm, &nm->rx_buffers[cpu_index][n_free_bufs], + vlib_buffer_alloc (vm, &nm->rx_buffers[thread_index][n_free_bufs], ring_size); - _vec_len (nm->rx_buffers[cpu_index]) = n_free_bufs; + _vec_len (nm->rx_buffers[thread_index]) = n_free_bufs; } head = ring->head; @@ -158,15 +159,15 @@ memif_device_input_inline (vlib_main_t * vm, vlib_node_runtime_t * node, CLIB_CACHE_LINE_BYTES, LOAD); } /* get empty buffer */ - u32 last_buf = vec_len (nm->rx_buffers[cpu_index]) - 1; - bi0 = nm->rx_buffers[cpu_index][last_buf]; - bi1 = nm->rx_buffers[cpu_index][last_buf - 1]; - _vec_len (nm->rx_buffers[cpu_index]) -= 2; + u32 last_buf = vec_len (nm->rx_buffers[thread_index]) - 1; + bi0 = nm->rx_buffers[thread_index][last_buf]; + bi1 = nm->rx_buffers[thread_index][last_buf - 1]; + _vec_len (nm->rx_buffers[thread_index]) -= 2; if (last_buf > 4) { - memif_prefetch (vm, nm->rx_buffers[cpu_index][last_buf - 2]); - memif_prefetch (vm, nm->rx_buffers[cpu_index][last_buf - 3]); + memif_prefetch (vm, nm->rx_buffers[thread_index][last_buf - 2]); + memif_prefetch (vm, nm->rx_buffers[thread_index][last_buf - 3]); } /* enqueue buffer */ @@ -256,9 +257,9 @@ memif_device_input_inline (vlib_main_t * vm, vlib_node_runtime_t * node, while (num_slots && n_left_to_next) { /* get empty buffer */ - u32 last_buf = vec_len (nm->rx_buffers[cpu_index]) - 1; - bi0 = nm->rx_buffers[cpu_index][last_buf]; - _vec_len (nm->rx_buffers[cpu_index]) = last_buf; + u32 last_buf = vec_len (nm->rx_buffers[thread_index]) - 1; + bi0 = nm->rx_buffers[thread_index][last_buf]; + _vec_len (nm->rx_buffers[thread_index]) = last_buf; /* enqueue buffer */ to_next[0] = bi0; @@ -315,7 +316,7 @@ memif_device_input_inline (vlib_main_t * vm, vlib_node_runtime_t * node, ring->tail = head; vlib_increment_combined_counter (vnm->interface_main.combined_sw_if_counters - + VNET_INTERFACE_COUNTER_RX, cpu_index, + + VNET_INTERFACE_COUNTER_RX, thread_index, mif->hw_if_index, n_rx_packets, n_rx_bytes); @@ -327,7 +328,7 @@ memif_input_fn (vlib_main_t * vm, vlib_node_runtime_t * node, vlib_frame_t * frame) { u32 n_rx_packets = 0; - u32 cpu_index = os_get_cpu_number (); + u32 thread_index = vlib_get_thread_index (); memif_main_t *nm = &memif_main; memif_if_t *mif; @@ -337,7 +338,7 @@ memif_input_fn (vlib_main_t * vm, vlib_node_runtime_t * node, if (mif->flags & MEMIF_IF_FLAG_ADMIN_UP && mif->flags & MEMIF_IF_FLAG_CONNECTED && (mif->if_index % nm->input_cpu_count) == - (cpu_index - nm->input_cpu_first_index)) + (thread_index - nm->input_cpu_first_index)) { if (mif->flags & MEMIF_IF_FLAG_IS_SLAVE) n_rx_packets += diff --git a/src/plugins/snat/in2out.c b/src/plugins/snat/in2out.c index b4961365..e5ee965f 100644 --- a/src/plugins/snat/in2out.c +++ b/src/plugins/snat/in2out.c @@ -212,7 +212,7 @@ static u32 slow_path (snat_main_t *sm, vlib_buffer_t *b0, snat_session_t ** sessionp, vlib_node_runtime_t * node, u32 next0, - u32 cpu_index) + u32 thread_index) { snat_user_t *u; snat_user_key_t user_key; @@ -246,27 +246,27 @@ static u32 slow_path (snat_main_t *sm, vlib_buffer_t *b0, if (clib_bihash_search_8_8 (&sm->user_hash, &kv0, &value0)) { /* no, make a new one */ - pool_get (sm->per_thread_data[cpu_index].users, u); + pool_get (sm->per_thread_data[thread_index].users, u); memset (u, 0, sizeof (*u)); u->addr = ip0->src_address; u->fib_index = rx_fib_index0; - pool_get (sm->per_thread_data[cpu_index].list_pool, per_user_list_head_elt); + pool_get (sm->per_thread_data[thread_index].list_pool, per_user_list_head_elt); u->sessions_per_user_list_head_index = per_user_list_head_elt - - sm->per_thread_data[cpu_index].list_pool; + sm->per_thread_data[thread_index].list_pool; - clib_dlist_init (sm->per_thread_data[cpu_index].list_pool, + clib_dlist_init (sm->per_thread_data[thread_index].list_pool, u->sessions_per_user_list_head_index); - kv0.value = u - sm->per_thread_data[cpu_index].users; + kv0.value = u - sm->per_thread_data[thread_index].users; /* add user */ clib_bihash_add_del_8_8 (&sm->user_hash, &kv0, 1 /* is_add */); } else { - u = pool_elt_at_index (sm->per_thread_data[cpu_index].users, + u = pool_elt_at_index (sm->per_thread_data[thread_index].users, value0.value); } @@ -276,25 +276,25 @@ static u32 slow_path (snat_main_t *sm, vlib_buffer_t *b0, /* Remove the oldest dynamic translation */ do { oldest_per_user_translation_list_index = - clib_dlist_remove_head (sm->per_thread_data[cpu_index].list_pool, + clib_dlist_remove_head (sm->per_thread_data[thread_index].list_pool, u->sessions_per_user_list_head_index); ASSERT (oldest_per_user_translation_list_index != ~0); /* add it back to the end of the LRU list */ - clib_dlist_addtail (sm->per_thread_data[cpu_index].list_pool, + clib_dlist_addtail (sm->per_thread_data[thread_index].list_pool, u->sessions_per_user_list_head_index, oldest_per_user_translation_list_index); /* Get the list element */ oldest_per_user_translation_list_elt = - pool_elt_at_index (sm->per_thread_data[cpu_index].list_pool, + pool_elt_at_index (sm->per_thread_data[thread_index].list_pool, oldest_per_user_translation_list_index); /* Get the session index from the list element */ session_index = oldest_per_user_translation_list_elt->value; /* Get the session */ - s = pool_elt_at_index (sm->per_thread_data[cpu_index].sessions, + s = pool_elt_at_index (sm->per_thread_data[thread_index].sessions, session_index); } while (snat_is_session_static (s)); @@ -346,7 +346,7 @@ static u32 slow_path (snat_main_t *sm, vlib_buffer_t *b0, } /* Create a new session */ - pool_get (sm->per_thread_data[cpu_index].sessions, s); + pool_get (sm->per_thread_data[thread_index].sessions, s); memset (s, 0, sizeof (*s)); s->outside_address_index = address_index; @@ -362,22 +362,22 @@ static u32 slow_path (snat_main_t *sm, vlib_buffer_t *b0, } /* Create list elts */ - pool_get (sm->per_thread_data[cpu_index].list_pool, + pool_get (sm->per_thread_data[thread_index].list_pool, per_user_translation_list_elt); - clib_dlist_init (sm->per_thread_data[cpu_index].list_pool, + clib_dlist_init (sm->per_thread_data[thread_index].list_pool, per_user_translation_list_elt - - sm->per_thread_data[cpu_index].list_pool); + sm->per_thread_data[thread_index].list_pool); per_user_translation_list_elt->value = - s - sm->per_thread_data[cpu_index].sessions; + s - sm->per_thread_data[thread_index].sessions; s->per_user_index = per_user_translation_list_elt - - sm->per_thread_data[cpu_index].list_pool; + sm->per_thread_data[thread_index].list_pool; s->per_user_list_head_index = u->sessions_per_user_list_head_index; - clib_dlist_addtail (sm->per_thread_data[cpu_index].list_pool, + clib_dlist_addtail (sm->per_thread_data[thread_index].list_pool, s->per_user_list_head_index, per_user_translation_list_elt - - sm->per_thread_data[cpu_index].list_pool); + sm->per_thread_data[thread_index].list_pool); } s->in2out = *key0; @@ -388,12 +388,12 @@ static u32 slow_path (snat_main_t *sm, vlib_buffer_t *b0, /* Add to translation hashes */ kv0.key = s->in2out.as_u64; - kv0.value = s - sm->per_thread_data[cpu_index].sessions; + kv0.value = s - sm->per_thread_data[thread_index].sessions; if (clib_bihash_add_del_8_8 (&sm->in2out, &kv0, 1 /* is_add */)) clib_warning ("in2out key add failed"); kv0.key = s->out2in.as_u64; - kv0.value = s - sm->per_thread_data[cpu_index].sessions; + kv0.value = s - sm->per_thread_data[thread_index].sessions; if (clib_bihash_add_del_8_8 (&sm->out2in, &kv0, 1 /* is_add */)) clib_warning ("out2in key add failed"); @@ -403,7 +403,7 @@ static u32 slow_path (snat_main_t *sm, vlib_buffer_t *b0, worker_by_out_key.port = s->out2in.port; worker_by_out_key.fib_index = s->out2in.fib_index; kv0.key = worker_by_out_key.as_u64; - kv0.value = cpu_index; + kv0.value = thread_index; clib_bihash_add_del_8_8 (&sm->worker_by_out, &kv0, 1); /* log NAT event */ @@ -465,7 +465,7 @@ snat_in2out_error_t icmp_get_key(icmp46_header_t *icmp0, * * @param[in,out] sm SNAT main * @param[in,out] node SNAT node runtime - * @param[in] cpu_index CPU index + * @param[in] thread_index thread index * @param[in,out] b0 buffer containing packet to be translated * @param[out] p_key address and port before NAT translation * @param[out] p_value address and port after NAT translation @@ -473,7 +473,7 @@ snat_in2out_error_t icmp_get_key(icmp46_header_t *icmp0, * @param d optional parameter */ u32 icmp_match_in2out_slow(snat_main_t *sm, vlib_node_runtime_t *node, - u32 cpu_index, vlib_buffer_t *b0, + u32 thread_index, vlib_buffer_t *b0, snat_session_key_t *p_key, snat_session_key_t *p_value, u8 *p_dont_translate, void *d) @@ -524,13 +524,13 @@ u32 icmp_match_in2out_slow(snat_main_t *sm, vlib_node_runtime_t *node, } next0 = slow_path (sm, b0, ip0, rx_fib_index0, &key0, - &s0, node, next0, cpu_index); + &s0, node, next0, thread_index); if (PREDICT_FALSE (next0 == SNAT_IN2OUT_NEXT_DROP)) goto out; } else - s0 = pool_elt_at_index (sm->per_thread_data[cpu_index].sessions, + s0 = pool_elt_at_index (sm->per_thread_data[thread_index].sessions, value0.value); out: @@ -548,7 +548,7 @@ out: * * @param[in] sm SNAT main * @param[in,out] node SNAT node runtime - * @param[in] cpu_index CPU index + * @param[in] thread_index thread index * @param[in,out] b0 buffer containing packet to be translated * @param[out] p_key address and port before NAT translation * @param[out] p_value address and port after NAT translation @@ -556,7 +556,7 @@ out: * @param d optional parameter */ u32 icmp_match_in2out_fast(snat_main_t *sm, vlib_node_runtime_t *node, - u32 cpu_index, vlib_buffer_t *b0, + u32 thread_index, vlib_buffer_t *b0, snat_session_key_t *p_key, snat_session_key_t *p_value, u8 *p_dont_translate, void *d) @@ -624,7 +624,7 @@ static inline u32 icmp_in2out (snat_main_t *sm, u32 rx_fib_index0, vlib_node_runtime_t * node, u32 next0, - u32 cpu_index, + u32 thread_index, void *d) { snat_session_key_t key0, sm0; @@ -641,7 +641,7 @@ static inline u32 icmp_in2out (snat_main_t *sm, echo0 = (icmp_echo_header_t *)(icmp0+1); - next0_tmp = sm->icmp_match_in2out_cb(sm, node, cpu_index, b0, + next0_tmp = sm->icmp_match_in2out_cb(sm, node, thread_index, b0, &key0, &sm0, &dont_translate, d); if (next0_tmp != ~0) next0 = next0_tmp; @@ -847,11 +847,11 @@ static inline u32 icmp_in2out_slow_path (snat_main_t *sm, vlib_node_runtime_t * node, u32 next0, f64 now, - u32 cpu_index, + u32 thread_index, snat_session_t ** p_s0) { next0 = icmp_in2out(sm, b0, ip0, icmp0, sw_if_index0, rx_fib_index0, node, - next0, cpu_index, p_s0); + next0, thread_index, p_s0); snat_session_t * s0 = *p_s0; if (PREDICT_TRUE(next0 != SNAT_IN2OUT_NEXT_DROP && s0)) { @@ -862,9 +862,9 @@ static inline u32 icmp_in2out_slow_path (snat_main_t *sm, /* Per-user LRU list maintenance for dynamic translations */ if (!snat_is_session_static (s0)) { - clib_dlist_remove (sm->per_thread_data[cpu_index].list_pool, + clib_dlist_remove (sm->per_thread_data[thread_index].list_pool, s0->per_user_index); - clib_dlist_addtail (sm->per_thread_data[cpu_index].list_pool, + clib_dlist_addtail (sm->per_thread_data[thread_index].list_pool, s0->per_user_list_head_index, s0->per_user_index); } @@ -884,7 +884,7 @@ snat_in2out_node_fn_inline (vlib_main_t * vm, snat_runtime_t * rt = (snat_runtime_t *)node->runtime_data; f64 now = vlib_time_now (vm); u32 stats_node_index; - u32 cpu_index = os_get_cpu_number (); + u32 thread_index = vlib_get_thread_index (); stats_node_index = is_slow_path ? snat_in2out_slowpath_node.index : snat_in2out_node.index; @@ -977,7 +977,7 @@ snat_in2out_node_fn_inline (vlib_main_t * vm, { next0 = icmp_in2out_slow_path (sm, b0, ip0, icmp0, sw_if_index0, rx_fib_index0, - node, next0, now, cpu_index, &s0); + node, next0, now, thread_index, &s0); goto trace00; } } @@ -1006,7 +1006,7 @@ snat_in2out_node_fn_inline (vlib_main_t * vm, goto trace00; next0 = slow_path (sm, b0, ip0, rx_fib_index0, &key0, - &s0, node, next0, cpu_index); + &s0, node, next0, thread_index); if (PREDICT_FALSE (next0 == SNAT_IN2OUT_NEXT_DROP)) goto trace00; } @@ -1017,7 +1017,7 @@ snat_in2out_node_fn_inline (vlib_main_t * vm, } } else - s0 = pool_elt_at_index (sm->per_thread_data[cpu_index].sessions, + s0 = pool_elt_at_index (sm->per_thread_data[thread_index].sessions, value0.value); old_addr0 = ip0->src_address.as_u32; @@ -1063,9 +1063,9 @@ snat_in2out_node_fn_inline (vlib_main_t * vm, /* Per-user LRU list maintenance for dynamic translation */ if (!snat_is_session_static (s0)) { - clib_dlist_remove (sm->per_thread_data[cpu_index].list_pool, + clib_dlist_remove (sm->per_thread_data[thread_index].list_pool, s0->per_user_index); - clib_dlist_addtail (sm->per_thread_data[cpu_index].list_pool, + clib_dlist_addtail (sm->per_thread_data[thread_index].list_pool, s0->per_user_list_head_index, s0->per_user_index); } @@ -1081,7 +1081,7 @@ snat_in2out_node_fn_inline (vlib_main_t * vm, t->next_index = next0; t->session_index = ~0; if (s0) - t->session_index = s0 - sm->per_thread_data[cpu_index].sessions; + t->session_index = s0 - sm->per_thread_data[thread_index].sessions; } pkts_processed += next0 != SNAT_IN2OUT_NEXT_DROP; @@ -1117,7 +1117,7 @@ snat_in2out_node_fn_inline (vlib_main_t * vm, { next1 = icmp_in2out_slow_path (sm, b1, ip1, icmp1, sw_if_index1, rx_fib_index1, node, - next1, now, cpu_index, &s1); + next1, now, thread_index, &s1); goto trace01; } } @@ -1146,7 +1146,7 @@ snat_in2out_node_fn_inline (vlib_main_t * vm, goto trace01; next1 = slow_path (sm, b1, ip1, rx_fib_index1, &key1, - &s1, node, next1, cpu_index); + &s1, node, next1, thread_index); if (PREDICT_FALSE (next1 == SNAT_IN2OUT_NEXT_DROP)) goto trace01; } @@ -1157,7 +1157,7 @@ snat_in2out_node_fn_inline (vlib_main_t * vm, } } else - s1 = pool_elt_at_index (sm->per_thread_data[cpu_index].sessions, + s1 = pool_elt_at_index (sm->per_thread_data[thread_index].sessions, value1.value); old_addr1 = ip1->src_address.as_u32; @@ -1203,9 +1203,9 @@ snat_in2out_node_fn_inline (vlib_main_t * vm, /* Per-user LRU list maintenance for dynamic translation */ if (!snat_is_session_static (s1)) { - clib_dlist_remove (sm->per_thread_data[cpu_index].list_pool, + clib_dlist_remove (sm->per_thread_data[thread_index].list_pool, s1->per_user_index); - clib_dlist_addtail (sm->per_thread_data[cpu_index].list_pool, + clib_dlist_addtail (sm->per_thread_data[thread_index].list_pool, s1->per_user_list_head_index, s1->per_user_index); } @@ -1220,7 +1220,7 @@ snat_in2out_node_fn_inline (vlib_main_t * vm, t->next_index = next1; t->session_index = ~0; if (s1) - t->session_index = s1 - sm->per_thread_data[cpu_index].sessions; + t->session_index = s1 - sm->per_thread_data[thread_index].sessions; } pkts_processed += next1 != SNAT_IN2OUT_NEXT_DROP; @@ -1292,7 +1292,7 @@ snat_in2out_node_fn_inline (vlib_main_t * vm, { next0 = icmp_in2out_slow_path (sm, b0, ip0, icmp0, sw_if_index0, rx_fib_index0, node, - next0, now, cpu_index, &s0); + next0, now, thread_index, &s0); goto trace0; } } @@ -1321,7 +1321,7 @@ snat_in2out_node_fn_inline (vlib_main_t * vm, goto trace0; next0 = slow_path (sm, b0, ip0, rx_fib_index0, &key0, - &s0, node, next0, cpu_index); + &s0, node, next0, thread_index); if (PREDICT_FALSE (next0 == SNAT_IN2OUT_NEXT_DROP)) goto trace0; @@ -1333,7 +1333,7 @@ snat_in2out_node_fn_inline (vlib_main_t * vm, } } else - s0 = pool_elt_at_index (sm->per_thread_data[cpu_index].sessions, + s0 = pool_elt_at_index (sm->per_thread_data[thread_index].sessions, value0.value); old_addr0 = ip0->src_address.as_u32; @@ -1379,9 +1379,9 @@ snat_in2out_node_fn_inline (vlib_main_t * vm, /* Per-user LRU list maintenance for dynamic translation */ if (!snat_is_session_static (s0)) { - clib_dlist_remove (sm->per_thread_data[cpu_index].list_pool, + clib_dlist_remove (sm->per_thread_data[thread_index].list_pool, s0->per_user_index); - clib_dlist_addtail (sm->per_thread_data[cpu_index].list_pool, + clib_dlist_addtail (sm->per_thread_data[thread_index].list_pool, s0->per_user_list_head_index, s0->per_user_index); } @@ -1397,7 +1397,7 @@ snat_in2out_node_fn_inline (vlib_main_t * vm, t->next_index = next0; t->session_index = ~0; if (s0) - t->session_index = s0 - sm->per_thread_data[cpu_index].sessions; + t->session_index = s0 - sm->per_thread_data[thread_index].sessions; } pkts_processed += next0 != SNAT_IN2OUT_NEXT_DROP; @@ -2010,7 +2010,7 @@ snat_in2out_worker_handoff_fn (vlib_main_t * vm, u32 n_left_to_next_worker = 0, *to_next_worker = 0; u32 next_worker_index = 0; u32 current_worker_index = ~0; - u32 cpu_index = os_get_cpu_number (); + u32 thread_index = vlib_get_thread_index (); ASSERT (vec_len (sm->workers)); @@ -2048,7 +2048,7 @@ snat_in2out_worker_handoff_fn (vlib_main_t * vm, next_worker_index = sm->worker_in2out_cb(ip0, rx_fib_index0); - if (PREDICT_FALSE (next_worker_index != cpu_index)) + if (PREDICT_FALSE (next_worker_index != thread_index)) { do_handoff = 1; diff --git a/src/plugins/snat/out2in.c b/src/plugins/snat/out2in.c index 656e42db..5d308d78 100644 --- a/src/plugins/snat/out2in.c +++ b/src/plugins/snat/out2in.c @@ -129,7 +129,7 @@ create_session_for_static_mapping (snat_main_t *sm, snat_session_key_t in2out, snat_session_key_t out2in, vlib_node_runtime_t * node, - u32 cpu_index) + u32 thread_index) { snat_user_t *u; snat_user_key_t user_key; @@ -146,36 +146,36 @@ create_session_for_static_mapping (snat_main_t *sm, if (clib_bihash_search_8_8 (&sm->user_hash, &kv0, &value0)) { /* no, make a new one */ - pool_get (sm->per_thread_data[cpu_index].users, u); + pool_get (sm->per_thread_data[thread_index].users, u); memset (u, 0, sizeof (*u)); u->addr = in2out.addr; u->fib_index = in2out.fib_index; - pool_get (sm->per_thread_data[cpu_index].list_pool, + pool_get (sm->per_thread_data[thread_index].list_pool, per_user_list_head_elt); u->sessions_per_user_list_head_index = per_user_list_head_elt - - sm->per_thread_data[cpu_index].list_pool; + sm->per_thread_data[thread_index].list_pool; - clib_dlist_init (sm->per_thread_data[cpu_index].list_pool, + clib_dlist_init (sm->per_thread_data[thread_index].list_pool, u->sessions_per_user_list_head_index); - kv0.value = u - sm->per_thread_data[cpu_index].users; + kv0.value = u - sm->per_thread_data[thread_index].users; /* add user */ clib_bihash_add_del_8_8 (&sm->user_hash, &kv0, 1 /* is_add */); /* add non-traslated packets worker lookup */ - kv0.value = cpu_index; + kv0.value = thread_index; clib_bihash_add_del_8_8 (&sm->worker_by_in, &kv0, 1); } else { - u = pool_elt_at_index (sm->per_thread_data[cpu_index].users, + u = pool_elt_at_index (sm->per_thread_data[thread_index].users, value0.value); } - pool_get (sm->per_thread_data[cpu_index].sessions, s); + pool_get (sm->per_thread_data[thread_index].sessions, s); memset (s, 0, sizeof (*s)); s->outside_address_index = ~0; @@ -183,22 +183,22 @@ create_session_for_static_mapping (snat_main_t *sm, u->nstaticsessions++; /* Create list elts */ - pool_get (sm->per_thread_data[cpu_index].list_pool, + pool_get (sm->per_thread_data[thread_index].list_pool, per_user_translation_list_elt); - clib_dlist_init (sm->per_thread_data[cpu_index].list_pool, + clib_dlist_init (sm->per_thread_data[thread_index].list_pool, per_user_translation_list_elt - - sm->per_thread_data[cpu_index].list_pool); + sm->per_thread_data[thread_index].list_pool); per_user_translation_list_elt->value = - s - sm->per_thread_data[cpu_index].sessions; + s - sm->per_thread_data[thread_index].sessions; s->per_user_index = - per_user_translation_list_elt - sm->per_thread_data[cpu_index].list_pool; + per_user_translation_list_elt - sm->per_thread_data[thread_index].list_pool; s->per_user_list_head_index = u->sessions_per_user_list_head_index; - clib_dlist_addtail (sm->per_thread_data[cpu_index].list_pool, + clib_dlist_addtail (sm->per_thread_data[thread_index].list_pool, s->per_user_list_head_index, per_user_translation_list_elt - - sm->per_thread_data[cpu_index].list_pool); + sm->per_thread_data[thread_index].list_pool); s->in2out = in2out; s->out2in = out2in; @@ -206,12 +206,12 @@ create_session_for_static_mapping (snat_main_t *sm, /* Add to translation hashes */ kv0.key = s->in2out.as_u64; - kv0.value = s - sm->per_thread_data[cpu_index].sessions; + kv0.value = s - sm->per_thread_data[thread_index].sessions; if (clib_bihash_add_del_8_8 (&sm->in2out, &kv0, 1 /* is_add */)) clib_warning ("in2out key add failed"); kv0.key = s->out2in.as_u64; - kv0.value = s - sm->per_thread_data[cpu_index].sessions; + kv0.value = s - sm->per_thread_data[thread_index].sessions; if (clib_bihash_add_del_8_8 (&sm->out2in, &kv0, 1 /* is_add */)) clib_warning ("out2in key add failed"); @@ -298,7 +298,7 @@ is_interface_addr(snat_main_t *sm, vlib_node_runtime_t *node, u32 sw_if_index0, * * @param[in,out] sm SNAT main * @param[in,out] node SNAT node runtime - * @param[in] cpu_index CPU index + * @param[in] thread_index thread index * @param[in,out] b0 buffer containing packet to be translated * @param[out] p_key address and port before NAT translation * @param[out] p_value address and port after NAT translation @@ -306,7 +306,7 @@ is_interface_addr(snat_main_t *sm, vlib_node_runtime_t *node, u32 sw_if_index0, * @param d optional parameter */ u32 icmp_match_out2in_slow(snat_main_t *sm, vlib_node_runtime_t *node, - u32 cpu_index, vlib_buffer_t *b0, + u32 thread_index, vlib_buffer_t *b0, snat_session_key_t *p_key, snat_session_key_t *p_value, u8 *p_dont_translate, void *d) @@ -366,7 +366,7 @@ u32 icmp_match_out2in_slow(snat_main_t *sm, vlib_node_runtime_t *node, /* Create session initiated by host from external network */ s0 = create_session_for_static_mapping(sm, b0, sm0, key0, - node, cpu_index); + node, thread_index); if (!s0) { @@ -375,7 +375,7 @@ u32 icmp_match_out2in_slow(snat_main_t *sm, vlib_node_runtime_t *node, } } else - s0 = pool_elt_at_index (sm->per_thread_data[cpu_index].sessions, + s0 = pool_elt_at_index (sm->per_thread_data[thread_index].sessions, value0.value); out: @@ -393,7 +393,7 @@ out: * * @param[in] sm SNAT main * @param[in,out] node SNAT node runtime - * @param[in] cpu_index CPU index + * @param[in] thread_index thread index * @param[in,out] b0 buffer containing packet to be translated * @param[out] p_key address and port before NAT translation * @param[out] p_value address and port after NAT translation @@ -401,7 +401,7 @@ out: * @param d optional parameter */ u32 icmp_match_out2in_fast(snat_main_t *sm, vlib_node_runtime_t *node, - u32 cpu_index, vlib_buffer_t *b0, + u32 thread_index, vlib_buffer_t *b0, snat_session_key_t *p_key, snat_session_key_t *p_value, u8 *p_dont_translate, void *d) @@ -460,7 +460,7 @@ static inline u32 icmp_out2in (snat_main_t *sm, u32 rx_fib_index0, vlib_node_runtime_t * node, u32 next0, - u32 cpu_index, + u32 thread_index, void *d) { snat_session_key_t key0, sm0; @@ -477,7 +477,7 @@ static inline u32 icmp_out2in (snat_main_t *sm, echo0 = (icmp_echo_header_t *)(icmp0+1); - next0_tmp = sm->icmp_match_out2in_cb(sm, node, cpu_index, b0, + next0_tmp = sm->icmp_match_out2in_cb(sm, node, thread_index, b0, &key0, &sm0, &dont_translate, d); if (next0_tmp != ~0) next0 = next0_tmp; @@ -589,11 +589,11 @@ static inline u32 icmp_out2in_slow_path (snat_main_t *sm, u32 rx_fib_index0, vlib_node_runtime_t * node, u32 next0, f64 now, - u32 cpu_index, + u32 thread_index, snat_session_t ** p_s0) { next0 = icmp_out2in(sm, b0, ip0, icmp0, sw_if_index0, rx_fib_index0, node, - next0, cpu_index, p_s0); + next0, thread_index, p_s0); snat_session_t * s0 = *p_s0; if (PREDICT_TRUE(next0 != SNAT_OUT2IN_NEXT_DROP && s0)) { @@ -604,9 +604,9 @@ static inline u32 icmp_out2in_slow_path (snat_main_t *sm, /* Per-user LRU list maintenance for dynamic translation */ if (!snat_is_session_static (s0)) { - clib_dlist_remove (sm->per_thread_data[cpu_index].list_pool, + clib_dlist_remove (sm->per_thread_data[thread_index].list_pool, s0->per_user_index); - clib_dlist_addtail (sm->per_thread_data[cpu_index].list_pool, + clib_dlist_addtail (sm->per_thread_data[thread_index].list_pool, s0->per_user_list_head_index, s0->per_user_index); } @@ -624,7 +624,7 @@ snat_out2in_node_fn (vlib_main_t * vm, u32 pkts_processed = 0; snat_main_t * sm = &snat_main; f64 now = vlib_time_now (vm); - u32 cpu_index = os_get_cpu_number (); + u32 thread_index = vlib_get_thread_index (); from = vlib_frame_vector_args (frame); n_left_from = frame->n_vectors; @@ -712,7 +712,7 @@ snat_out2in_node_fn (vlib_main_t * vm, { next0 = icmp_out2in_slow_path (sm, b0, ip0, icmp0, sw_if_index0, rx_fib_index0, node, - next0, now, cpu_index, &s0); + next0, now, thread_index, &s0); goto trace0; } @@ -743,7 +743,7 @@ snat_out2in_node_fn (vlib_main_t * vm, /* Create session initiated by host from external network */ s0 = create_session_for_static_mapping(sm, b0, sm0, key0, node, - cpu_index); + thread_index); if (!s0) { b0->error = node->errors[SNAT_OUT2IN_ERROR_NO_TRANSLATION]; @@ -752,7 +752,7 @@ snat_out2in_node_fn (vlib_main_t * vm, } } else - s0 = pool_elt_at_index (sm->per_thread_data[cpu_index].sessions, + s0 = pool_elt_at_index (sm->per_thread_data[thread_index].sessions, value0.value); old_addr0 = ip0->dst_address.as_u32; @@ -796,9 +796,9 @@ snat_out2in_node_fn (vlib_main_t * vm, /* Per-user LRU list maintenance for dynamic translation */ if (!snat_is_session_static (s0)) { - clib_dlist_remove (sm->per_thread_data[cpu_index].list_pool, + clib_dlist_remove (sm->per_thread_data[thread_index].list_pool, s0->per_user_index); - clib_dlist_addtail (sm->per_thread_data[cpu_index].list_pool, + clib_dlist_addtail (sm->per_thread_data[thread_index].list_pool, s0->per_user_list_head_index, s0->per_user_index); } @@ -813,7 +813,7 @@ snat_out2in_node_fn (vlib_main_t * vm, t->next_index = next0; t->session_index = ~0; if (s0) - t->session_index = s0 - sm->per_thread_data[cpu_index].sessions; + t->session_index = s0 - sm->per_thread_data[thread_index].sessions; } pkts_processed += next0 != SNAT_OUT2IN_NEXT_DROP; @@ -847,7 +847,7 @@ snat_out2in_node_fn (vlib_main_t * vm, { next1 = icmp_out2in_slow_path (sm, b1, ip1, icmp1, sw_if_index1, rx_fib_index1, node, - next1, now, cpu_index, &s1); + next1, now, thread_index, &s1); goto trace1; } @@ -878,7 +878,7 @@ snat_out2in_node_fn (vlib_main_t * vm, /* Create session initiated by host from external network */ s1 = create_session_for_static_mapping(sm, b1, sm1, key1, node, - cpu_index); + thread_index); if (!s1) { b1->error = node->errors[SNAT_OUT2IN_ERROR_NO_TRANSLATION]; @@ -887,7 +887,7 @@ snat_out2in_node_fn (vlib_main_t * vm, } } else - s1 = pool_elt_at_index (sm->per_thread_data[cpu_index].sessions, + s1 = pool_elt_at_index (sm->per_thread_data[thread_index].sessions, value1.value); old_addr1 = ip1->dst_address.as_u32; @@ -931,9 +931,9 @@ snat_out2in_node_fn (vlib_main_t * vm, /* Per-user LRU list maintenance for dynamic translation */ if (!snat_is_session_static (s1)) { - clib_dlist_remove (sm->per_thread_data[cpu_index].list_pool, + clib_dlist_remove (sm->per_thread_data[thread_index].list_pool, s1->per_user_index); - clib_dlist_addtail (sm->per_thread_data[cpu_index].list_pool, + clib_dlist_addtail (sm->per_thread_data[thread_index].list_pool, s1->per_user_list_head_index, s1->per_user_index); } @@ -948,7 +948,7 @@ snat_out2in_node_fn (vlib_main_t * vm, t->next_index = next1; t->session_index = ~0; if (s1) - t->session_index = s1 - sm->per_thread_data[cpu_index].sessions; + t->session_index = s1 - sm->per_thread_data[thread_index].sessions; } pkts_processed += next1 != SNAT_OUT2IN_NEXT_DROP; @@ -1016,7 +1016,7 @@ snat_out2in_node_fn (vlib_main_t * vm, { next0 = icmp_out2in_slow_path (sm, b0, ip0, icmp0, sw_if_index0, rx_fib_index0, node, - next0, now, cpu_index, &s0); + next0, now, thread_index, &s0); goto trace00; } @@ -1048,7 +1048,7 @@ snat_out2in_node_fn (vlib_main_t * vm, /* Create session initiated by host from external network */ s0 = create_session_for_static_mapping(sm, b0, sm0, key0, node, - cpu_index); + thread_index); if (!s0) { b0->error = node->errors[SNAT_OUT2IN_ERROR_NO_TRANSLATION]; @@ -1057,7 +1057,7 @@ snat_out2in_node_fn (vlib_main_t * vm, } } else - s0 = pool_elt_at_index (sm->per_thread_data[cpu_index].sessions, + s0 = pool_elt_at_index (sm->per_thread_data[thread_index].sessions, value0.value); old_addr0 = ip0->dst_address.as_u32; @@ -1101,9 +1101,9 @@ snat_out2in_node_fn (vlib_main_t * vm, /* Per-user LRU list maintenance for dynamic translation */ if (!snat_is_session_static (s0)) { - clib_dlist_remove (sm->per_thread_data[cpu_index].list_pool, + clib_dlist_remove (sm->per_thread_data[thread_index].list_pool, s0->per_user_index); - clib_dlist_addtail (sm->per_thread_data[cpu_index].list_pool, + clib_dlist_addtail (sm->per_thread_data[thread_index].list_pool, s0->per_user_list_head_index, s0->per_user_index); } @@ -1118,7 +1118,7 @@ snat_out2in_node_fn (vlib_main_t * vm, t->next_index = next0; t->session_index = ~0; if (s0) - t->session_index = s0 - sm->per_thread_data[cpu_index].sessions; + t->session_index = s0 - sm->per_thread_data[thread_index].sessions; } pkts_processed += next0 != SNAT_OUT2IN_NEXT_DROP; @@ -1599,7 +1599,7 @@ snat_out2in_worker_handoff_fn (vlib_main_t * vm, u32 n_left_to_next_worker = 0, *to_next_worker = 0; u32 next_worker_index = 0; u32 current_worker_index = ~0; - u32 cpu_index = os_get_cpu_number (); + u32 thread_index = vlib_get_thread_index (); ASSERT (vec_len (sm->workers)); @@ -1637,7 +1637,7 @@ snat_out2in_worker_handoff_fn (vlib_main_t * vm, next_worker_index = sm->worker_out2in_cb(ip0, rx_fib_index0); - if (PREDICT_FALSE (next_worker_index != cpu_index)) + if (PREDICT_FALSE (next_worker_index != thread_index)) { do_handoff = 1; diff --git a/src/plugins/snat/snat.h b/src/plugins/snat/snat.h index 017825c0..f4e1c5c0 100644 --- a/src/plugins/snat/snat.h +++ b/src/plugins/snat/snat.h @@ -221,7 +221,7 @@ struct snat_main_s; typedef u32 snat_icmp_match_function_t (struct snat_main_s *sm, vlib_node_runtime_t *node, - u32 cpu_index, + u32 thread_index, vlib_buffer_t *b0, snat_session_key_t *p_key, snat_session_key_t *p_value, @@ -402,22 +402,22 @@ typedef struct { } tcp_udp_header_t; u32 icmp_match_in2out_fast(snat_main_t *sm, vlib_node_runtime_t *node, - u32 cpu_index, vlib_buffer_t *b0, + u32 thread_index, vlib_buffer_t *b0, snat_session_key_t *p_key, snat_session_key_t *p_value, u8 *p_dont_translate, void *d); u32 icmp_match_in2out_slow(snat_main_t *sm, vlib_node_runtime_t *node, - u32 cpu_index, vlib_buffer_t *b0, + u32 thread_index, vlib_buffer_t *b0, snat_session_key_t *p_key, snat_session_key_t *p_value, u8 *p_dont_translate, void *d); u32 icmp_match_out2in_fast(snat_main_t *sm, vlib_node_runtime_t *node, - u32 cpu_index, vlib_buffer_t *b0, + u32 thread_index, vlib_buffer_t *b0, snat_session_key_t *p_key, snat_session_key_t *p_value, u8 *p_dont_translate, void *d); u32 icmp_match_out2in_slow(snat_main_t *sm, vlib_node_runtime_t *node, - u32 cpu_index, vlib_buffer_t *b0, + u32 thread_index, vlib_buffer_t *b0, snat_session_key_t *p_key, snat_session_key_t *p_value, u8 *p_dont_translate, void *d); diff --git a/src/vlib/buffer.c b/src/vlib/buffer.c index a517a597..be3b41ef 100644 --- a/src/vlib/buffer.c +++ b/src/vlib/buffer.c @@ -299,7 +299,7 @@ vlib_buffer_validate_alloc_free (vlib_main_t * vm, if (CLIB_DEBUG == 0) return; - ASSERT (os_get_cpu_number () == 0); + ASSERT (vlib_get_thread_index () == 0); /* smp disaster check */ if (vec_len (vlib_mains) > 1) @@ -355,7 +355,7 @@ vlib_buffer_create_free_list_helper (vlib_main_t * vm, vlib_buffer_free_list_t *f; int i; - ASSERT (os_get_cpu_number () == 0); + ASSERT (vlib_get_thread_index () == 0); if (!is_default && pool_elts (bm->buffer_free_list_pool) == 0) { @@ -474,7 +474,7 @@ vlib_buffer_delete_free_list_internal (vlib_main_t * vm, u32 free_list_index) u32 merge_index; int i; - ASSERT (os_get_cpu_number () == 0); + ASSERT (vlib_get_thread_index () == 0); f = vlib_buffer_get_free_list (vm, free_list_index); diff --git a/src/vlib/buffer_funcs.h b/src/vlib/buffer_funcs.h index 394c336a..328660a3 100644 --- a/src/vlib/buffer_funcs.h +++ b/src/vlib/buffer_funcs.h @@ -209,7 +209,7 @@ always_inline vlib_buffer_known_state_t vlib_buffer_is_known (vlib_main_t * vm, u32 buffer_index) { vlib_buffer_main_t *bm = vm->buffer_main; - ASSERT (os_get_cpu_number () == 0); + ASSERT (vlib_get_thread_index () == 0); uword *p = hash_get (bm->buffer_known_hash, buffer_index); return p ? p[0] : VLIB_BUFFER_UNKNOWN; @@ -221,7 +221,7 @@ vlib_buffer_set_known_state (vlib_main_t * vm, vlib_buffer_known_state_t state) { vlib_buffer_main_t *bm = vm->buffer_main; - ASSERT (os_get_cpu_number () == 0); + ASSERT (vlib_get_thread_index () == 0); hash_set (bm->buffer_known_hash, buffer_index, state); } diff --git a/src/vlib/cli.c b/src/vlib/cli.c index f853f655..3cc95076 100644 --- a/src/vlib/cli.c +++ b/src/vlib/cli.c @@ -709,7 +709,7 @@ test_heap_validate (vlib_main_t * vm, unformat_input_t * input, { /* *INDENT-OFF* */ foreach_vlib_main({ - heap = clib_per_cpu_mheaps[this_vlib_main->cpu_index]; + heap = clib_per_cpu_mheaps[this_vlib_main->thread_index]; mheap = mheap_header(heap); mheap->flags |= MHEAP_FLAG_VALIDATE; // Turn off small object cache because it delays detection of errors @@ -722,7 +722,7 @@ test_heap_validate (vlib_main_t * vm, unformat_input_t * input, { /* *INDENT-OFF* */ foreach_vlib_main({ - heap = clib_per_cpu_mheaps[this_vlib_main->cpu_index]; + heap = clib_per_cpu_mheaps[this_vlib_main->thread_index]; mheap = mheap_header(heap); mheap->flags &= ~MHEAP_FLAG_VALIDATE; mheap->flags |= MHEAP_FLAG_SMALL_OBJECT_CACHE; @@ -733,7 +733,7 @@ test_heap_validate (vlib_main_t * vm, unformat_input_t * input, { /* *INDENT-OFF* */ foreach_vlib_main({ - heap = clib_per_cpu_mheaps[this_vlib_main->cpu_index]; + heap = clib_per_cpu_mheaps[this_vlib_main->thread_index]; mheap = mheap_header(heap); mheap_validate(heap); }); diff --git a/src/vlib/counter.h b/src/vlib/counter.h index 17a85217..60e2055d 100644 --- a/src/vlib/counter.h +++ b/src/vlib/counter.h @@ -70,17 +70,17 @@ u32 vlib_simple_counter_n_counters (const vlib_simple_counter_main_t * cm); /** Increment a simple counter @param cm - (vlib_simple_counter_main_t *) simple counter main pointer - @param cpu_index - (u32) the current cpu index + @param thread_index - (u32) the current cpu index @param index - (u32) index of the counter to increment @param increment - (u64) quantitiy to add to the counter */ always_inline void vlib_increment_simple_counter (vlib_simple_counter_main_t * cm, - u32 cpu_index, u32 index, u64 increment) + u32 thread_index, u32 index, u64 increment) { counter_t *my_counters; - my_counters = cm->counters[cpu_index]; + my_counters = cm->counters[thread_index]; my_counters[index] += increment; } @@ -201,7 +201,7 @@ void vlib_clear_combined_counters (vlib_combined_counter_main_t * cm); /** Increment a combined counter @param cm - (vlib_combined_counter_main_t *) comined counter main pointer - @param cpu_index - (u32) the current cpu index + @param thread_index - (u32) the current cpu index @param index - (u32) index of the counter to increment @param packet_increment - (u64) number of packets to add to the counter @param byte_increment - (u64) number of bytes to add to the counter @@ -209,13 +209,13 @@ void vlib_clear_combined_counters (vlib_combined_counter_main_t * cm); always_inline void vlib_increment_combined_counter (vlib_combined_counter_main_t * cm, - u32 cpu_index, + u32 thread_index, u32 index, u64 n_packets, u64 n_bytes) { vlib_counter_t *my_counters; /* Use this CPU's counter array */ - my_counters = cm->counters[cpu_index]; + my_counters = cm->counters[thread_index]; my_counters[index].packets += n_packets; my_counters[index].bytes += n_bytes; @@ -224,14 +224,14 @@ vlib_increment_combined_counter (vlib_combined_counter_main_t * cm, /** Pre-fetch a per-thread combined counter for the given object index */ always_inline void vlib_prefetch_combined_counter (const vlib_combined_counter_main_t * cm, - u32 cpu_index, u32 index) + u32 thread_index, u32 index) { vlib_counter_t *cpu_counters; /* * This CPU's index is assumed to already be in cache */ - cpu_counters = cm->counters[cpu_index]; + cpu_counters = cm->counters[thread_index]; CLIB_PREFETCH (cpu_counters + index, CLIB_CACHE_LINE_BYTES, STORE); } diff --git a/src/vlib/error.c b/src/vlib/error.c index a2c23176..e4ed4ee3 100644 --- a/src/vlib/error.c +++ b/src/vlib/error.c @@ -149,7 +149,7 @@ vlib_register_errors (vlib_main_t * vm, vlib_node_t *n = vlib_get_node (vm, node_index); uword l; - ASSERT (os_get_cpu_number () == 0); + ASSERT (vlib_get_thread_index () == 0); /* Free up any previous error strings. */ if (n->n_errors > 0) diff --git a/src/vlib/global_funcs.h b/src/vlib/global_funcs.h index f51ec381..9dd01fbf 100644 --- a/src/vlib/global_funcs.h +++ b/src/vlib/global_funcs.h @@ -23,7 +23,7 @@ always_inline vlib_main_t * vlib_get_main (void) { vlib_main_t *vm; - vm = vlib_mains[os_get_cpu_number ()]; + vm = vlib_mains[vlib_get_thread_index ()]; ASSERT (vm); return vm; } diff --git a/src/vlib/main.c b/src/vlib/main.c index b22203f0..422d3e26 100644 --- a/src/vlib/main.c +++ b/src/vlib/main.c @@ -136,18 +136,18 @@ vlib_frame_alloc_to_node (vlib_main_t * vm, u32 to_node_index, else { f = clib_mem_alloc_aligned_no_fail (n, VLIB_FRAME_ALIGN); - f->cpu_index = vm->cpu_index; + f->thread_index = vm->thread_index; fi = vlib_frame_index_no_check (vm, f); } /* Poison frame when debugging. */ if (CLIB_DEBUG > 0) { - u32 save_cpu_index = f->cpu_index; + u32 save_thread_index = f->thread_index; memset (f, 0xfe, n); - f->cpu_index = save_cpu_index; + f->thread_index = save_thread_index; } /* Insert magic number. */ @@ -517,7 +517,7 @@ vlib_put_next_frame (vlib_main_t * vm, * a dangling frame reference. Each thread has its own copy of * the next_frames vector. */ - if (0 && r->cpu_index != next_runtime->cpu_index) + if (0 && r->thread_index != next_runtime->thread_index) { nf->frame_index = ~0; nf->flags &= ~(VLIB_FRAME_PENDING | VLIB_FRAME_IS_ALLOCATED); @@ -866,7 +866,7 @@ vlib_elog_main_loop_event (vlib_main_t * vm, : evm->node_call_elog_event_types, node_index), /* track */ - (vm->cpu_index ? &vlib_worker_threads[vm->cpu_index]. + (vm->thread_index ? &vlib_worker_threads[vm->thread_index]. elog_track : &em->default_track), /* data to log */ n_vectors); } @@ -963,7 +963,7 @@ dispatch_node (vlib_main_t * vm, vm->cpu_time_last_node_dispatch = last_time_stamp; - if (1 /* || vm->cpu_index == node->cpu_index */ ) + if (1 /* || vm->thread_index == node->thread_index */ ) { vlib_main_t *stat_vm; @@ -1029,7 +1029,7 @@ dispatch_node (vlib_main_t * vm, { u32 node_name, vector_length, is_polling; } *ed; - vlib_worker_thread_t *w = vlib_worker_threads + vm->cpu_index; + vlib_worker_thread_t *w = vlib_worker_threads + vm->thread_index; #endif if ((dispatch_state == VLIB_NODE_STATE_INTERRUPT diff --git a/src/vlib/main.h b/src/vlib/main.h index 0197b4f3..329bf073 100644 --- a/src/vlib/main.h +++ b/src/vlib/main.h @@ -156,7 +156,7 @@ typedef struct vlib_main_t uword *init_functions_called; /* to compare with node runtime */ - u32 cpu_index; + u32 thread_index; void **mbuf_alloc_list; diff --git a/src/vlib/node.c b/src/vlib/node.c index dc0a4de5..bbd3a42e 100644 --- a/src/vlib/node.c +++ b/src/vlib/node.c @@ -99,7 +99,7 @@ vlib_node_runtime_update (vlib_main_t * vm, u32 node_index, u32 next_index) vlib_pending_frame_t *pf; i32 i, j, n_insert; - ASSERT (os_get_cpu_number () == 0); + ASSERT (vlib_get_thread_index () == 0); vlib_worker_thread_barrier_sync (vm); diff --git a/src/vlib/node.h b/src/vlib/node.h index fc7e7da2..1e2f4c38 100644 --- a/src/vlib/node.h +++ b/src/vlib/node.h @@ -344,8 +344,8 @@ typedef struct vlib_frame_t /* Number of vector elements currently in frame. */ u16 n_vectors; - /* Owner cpuid / heap id */ - u16 cpu_index; + /* Owner thread / heap id */ + u16 thread_index; /* Scalar and vector arguments to next node. */ u8 arguments[0]; @@ -459,7 +459,7 @@ typedef struct vlib_node_runtime_t zero before first run of this node. */ - u16 cpu_index; /**< CPU this node runs on */ + u16 thread_index; /**< thread this node runs on */ u8 runtime_data[0]; /**< Function dependent node-runtime data. This data is diff --git a/src/vlib/node_funcs.h b/src/vlib/node_funcs.h index 1f7d94e1..54e36874 100644 --- a/src/vlib/node_funcs.h +++ b/src/vlib/node_funcs.h @@ -201,9 +201,9 @@ always_inline vlib_frame_t * vlib_get_frame_no_check (vlib_main_t * vm, uword frame_index) { vlib_frame_t *f; - u32 cpu_index = frame_index & VLIB_CPU_MASK; + u32 thread_index = frame_index & VLIB_CPU_MASK; u32 offset = frame_index & VLIB_OFFSET_MASK; - vm = vlib_mains[cpu_index]; + vm = vlib_mains[thread_index]; f = vm->heap_base + offset; return f; } @@ -215,10 +215,10 @@ vlib_frame_index_no_check (vlib_main_t * vm, vlib_frame_t * f) ASSERT (((uword) f & VLIB_CPU_MASK) == 0); - vm = vlib_mains[f->cpu_index]; + vm = vlib_mains[f->thread_index]; i = ((u8 *) f - (u8 *) vm->heap_base); - return i | f->cpu_index; + return i | f->thread_index; } always_inline vlib_frame_t * diff --git a/src/vlib/threads.c b/src/vlib/threads.c index ef3a24d3..4a111f8d 100644 --- a/src/vlib/threads.c +++ b/src/vlib/threads.c @@ -35,27 +35,12 @@ vl (void *p) vlib_worker_thread_t *vlib_worker_threads; vlib_thread_main_t vlib_thread_main; +__thread uword vlib_thread_index = 0; + uword os_get_cpu_number (void) { - void *sp; - uword n; - u32 len; - - len = vec_len (vlib_thread_stacks); - if (len == 0) - return 0; - - /* Get any old stack address. */ - sp = &sp; - - n = ((uword) sp - (uword) vlib_thread_stacks[0]) - >> VLIB_LOG2_THREAD_STACK_SIZE; - - /* "processes" have their own stacks, and they always run in thread 0 */ - n = n >= len ? 0 : n; - - return n; + return vlib_thread_index; } uword @@ -275,21 +260,6 @@ vlib_thread_init (vlib_main_t * vm) return 0; } -vlib_worker_thread_t * -vlib_alloc_thread (vlib_main_t * vm) -{ - vlib_worker_thread_t *w; - - if (vec_len (vlib_worker_threads) >= vec_len (vlib_thread_stacks)) - { - clib_warning ("out of worker threads... Quitting..."); - exit (1); - } - vec_add2 (vlib_worker_threads, w, 1); - w->thread_stack = vlib_thread_stacks[w - vlib_worker_threads]; - return w; -} - vlib_frame_queue_t * vlib_frame_queue_alloc (int nelts) { @@ -427,7 +397,7 @@ vlib_frame_queue_enqueue (vlib_main_t * vm, u32 node_runtime_index, f64 b4 = vlib_time_now_ticks (vm, before); vlib_worker_thread_barrier_check (vm, b4); /* Bad idea. Dequeue -> enqueue -> dequeue -> trouble */ - // vlib_frame_queue_dequeue (vm->cpu_index, vm, nm); + // vlib_frame_queue_dequeue (vm->thread_index, vm, nm); } elt = fq->elts + (new_tail & (fq->nelts - 1)); @@ -497,6 +467,8 @@ vlib_worker_thread_bootstrap_fn (void *arg) w->lwp = syscall (SYS_gettid); w->thread_id = pthread_self (); + vlib_thread_index = w - vlib_worker_threads; + rv = (void *) clib_calljmp ((uword (*)(uword)) w->thread_function, (uword) arg, w->thread_stack + VLIB_THREAD_STACK_SIZE); @@ -610,7 +582,9 @@ start_workers (vlib_main_t * vm) mheap_alloc (0 /* use VM */ , tr->mheap_size); else w->thread_mheap = main_heap; - w->thread_stack = vlib_thread_stacks[w - vlib_worker_threads]; + + w->thread_stack = + vlib_thread_stack_init (w - vlib_worker_threads); w->thread_function = tr->function; w->thread_function_arg = w; w->instance_id = k; @@ -630,7 +604,7 @@ start_workers (vlib_main_t * vm) vm_clone = clib_mem_alloc (sizeof (*vm_clone)); clib_memcpy (vm_clone, vlib_mains[0], sizeof (*vm_clone)); - vm_clone->cpu_index = worker_thread_index; + vm_clone->thread_index = worker_thread_index; vm_clone->heap_base = w->thread_mheap; vm_clone->mbuf_alloc_list = 0; vm_clone->init_functions_called = @@ -679,7 +653,7 @@ start_workers (vlib_main_t * vm) vec_foreach (rt, nm_clone->nodes_by_type[VLIB_NODE_TYPE_INPUT]) { vlib_node_t *n = vlib_get_node (vm, rt->node_index); - rt->cpu_index = vm_clone->cpu_index; + rt->thread_index = vm_clone->thread_index; /* copy initial runtime_data from node */ if (n->runtime_data && n->runtime_data_bytes > 0) clib_memcpy (rt->runtime_data, n->runtime_data, @@ -692,7 +666,7 @@ start_workers (vlib_main_t * vm) vec_foreach (rt, nm_clone->nodes_by_type[VLIB_NODE_TYPE_INPUT]) { vlib_node_t *n = vlib_get_node (vm, rt->node_index); - rt->cpu_index = vm_clone->cpu_index; + rt->thread_index = vm_clone->thread_index; /* copy initial runtime_data from node */ if (n->runtime_data && n->runtime_data_bytes > 0) clib_memcpy (rt->runtime_data, n->runtime_data, @@ -756,7 +730,8 @@ start_workers (vlib_main_t * vm) mheap_alloc (0 /* use VM */ , tr->mheap_size); else w->thread_mheap = main_heap; - w->thread_stack = vlib_thread_stacks[w - vlib_worker_threads]; + w->thread_stack = + vlib_thread_stack_init (w - vlib_worker_threads); w->thread_function = tr->function; w->thread_function_arg = w; w->instance_id = j; @@ -827,7 +802,7 @@ vlib_worker_thread_node_runtime_update (void) uword n_calls, uword n_vectors, uword n_clocks); - ASSERT (os_get_cpu_number () == 0); + ASSERT (vlib_get_thread_index () == 0); if (vec_len (vlib_mains) == 1) return; @@ -835,7 +810,7 @@ vlib_worker_thread_node_runtime_update (void) vm = vlib_mains[0]; nm = &vm->node_main; - ASSERT (os_get_cpu_number () == 0); + ASSERT (vlib_get_thread_index () == 0); ASSERT (*vlib_worker_threads->wait_at_barrier == 1); /* @@ -955,7 +930,7 @@ vlib_worker_thread_node_runtime_update (void) vec_foreach (rt, nm_clone->nodes_by_type[VLIB_NODE_TYPE_INTERNAL]) { vlib_node_t *n = vlib_get_node (vm, rt->node_index); - rt->cpu_index = vm_clone->cpu_index; + rt->thread_index = vm_clone->thread_index; /* copy runtime_data, will be overwritten later for existing rt */ if (n->runtime_data && n->runtime_data_bytes > 0) clib_memcpy (rt->runtime_data, n->runtime_data, @@ -981,7 +956,7 @@ vlib_worker_thread_node_runtime_update (void) vec_foreach (rt, nm_clone->nodes_by_type[VLIB_NODE_TYPE_INPUT]) { vlib_node_t *n = vlib_get_node (vm, rt->node_index); - rt->cpu_index = vm_clone->cpu_index; + rt->thread_index = vm_clone->thread_index; /* copy runtime_data, will be overwritten later for existing rt */ if (n->runtime_data && n->runtime_data_bytes > 0) clib_memcpy (rt->runtime_data, n->runtime_data, @@ -1180,7 +1155,7 @@ vlib_worker_thread_fork_fixup (vlib_fork_fixup_t which) if (vlib_mains == 0) return; - ASSERT (os_get_cpu_number () == 0); + ASSERT (vlib_get_thread_index () == 0); vlib_worker_thread_barrier_sync (vm); switch (which) @@ -1212,7 +1187,7 @@ vlib_worker_thread_barrier_sync (vlib_main_t * vm) vlib_worker_threads[0].barrier_sync_count++; - ASSERT (os_get_cpu_number () == 0); + ASSERT (vlib_get_thread_index () == 0); deadline = vlib_time_now (vm) + BARRIER_SYNC_TIMEOUT; @@ -1260,7 +1235,7 @@ vlib_worker_thread_barrier_release (vlib_main_t * vm) int vlib_frame_queue_dequeue (vlib_main_t * vm, vlib_frame_queue_main_t * fqm) { - u32 thread_id = vm->cpu_index; + u32 thread_id = vm->thread_index; vlib_frame_queue_t *fq = fqm->vlib_frame_queues[thread_id]; vlib_frame_queue_elt_t *elt; u32 *from, *to; @@ -1393,7 +1368,7 @@ vlib_worker_thread_fn (void *arg) vlib_main_t *vm = vlib_get_main (); clib_error_t *e; - ASSERT (vm->cpu_index == os_get_cpu_number ()); + ASSERT (vm->thread_index == vlib_get_thread_index ()); vlib_worker_thread_init (w); clib_time_init (&vm->clib_time); diff --git a/src/vlib/threads.h b/src/vlib/threads.h index eca4fc26..101d3d4a 100644 --- a/src/vlib/threads.h +++ b/src/vlib/threads.h @@ -153,8 +153,6 @@ typedef struct /* Called early, in thread 0's context */ clib_error_t *vlib_thread_init (vlib_main_t * vm); -vlib_worker_thread_t *vlib_alloc_thread (vlib_main_t * vm); - int vlib_frame_queue_enqueue (vlib_main_t * vm, u32 node_runtime_index, u32 frame_queue_index, vlib_frame_t * frame, vlib_frame_queue_msg_type_t type); @@ -183,12 +181,19 @@ u32 vlib_frame_queue_main_init (u32 node_index, u32 frame_queue_nelts); void vlib_worker_thread_barrier_sync (vlib_main_t * vm); void vlib_worker_thread_barrier_release (vlib_main_t * vm); +extern __thread uword vlib_thread_index; +static_always_inline uword +vlib_get_thread_index (void) +{ + return vlib_thread_index; +} + always_inline void vlib_smp_unsafe_warning (void) { if (CLIB_DEBUG > 0) { - if (os_get_cpu_number ()) + if (vlib_get_thread_index ()) fformat (stderr, "%s: SMP unsafe warning...\n", __FUNCTION__); } } @@ -331,21 +336,21 @@ vlib_num_workers () } always_inline u32 -vlib_get_worker_cpu_index (u32 worker_index) +vlib_get_worker_thread_index (u32 worker_index) { return worker_index + 1; } always_inline u32 -vlib_get_worker_index (u32 cpu_index) +vlib_get_worker_index (u32 thread_index) { - return cpu_index - 1; + return thread_index - 1; } always_inline u32 vlib_get_current_worker_index () { - return os_get_cpu_number () - 1; + return vlib_get_thread_index () - 1; } static inline void @@ -467,6 +472,8 @@ vlib_get_worker_handoff_queue_elt (u32 frame_queue_index, return elt; } +u8 *vlib_thread_stack_init (uword thread_index); + int vlib_thread_cb_register (struct vlib_main_t *vm, vlib_thread_callbacks_t * cb); diff --git a/src/vlib/unix/cj.c b/src/vlib/unix/cj.c index 33ba163a..7c1e9475 100644 --- a/src/vlib/unix/cj.c +++ b/src/vlib/unix/cj.c @@ -48,7 +48,7 @@ cj_log (u32 type, void *data0, void *data1) r = (cj_record_t *) & (cjm->records[new_tail & (cjm->num_records - 1)]); r->time = vlib_time_now (cjm->vlib_main); - r->cpu = os_get_cpu_number (); + r->thread_index = vlib_get_thread_index (); r->type = type; r->data[0] = pointer_to_uword (data0); r->data[1] = pointer_to_uword (data1); @@ -133,7 +133,8 @@ static inline void cj_dump_one_record (cj_record_t * r) { fprintf (stderr, "[%d]: %10.6f T%02d %llx %llx\n", - r->cpu, r->time, r->type, (long long unsigned int) r->data[0], + r->thread_index, r->time, r->type, + (long long unsigned int) r->data[0], (long long unsigned int) r->data[1]); } @@ -161,7 +162,7 @@ cj_dump_internal (u8 filter0_enable, u64 filter0, index = (cjm->tail + 1) & (cjm->num_records - 1); r = &(cjm->records[index]); - if (r->cpu != (u32) ~ 0) + if (r->thread_index != (u32) ~ 0) { /* Yes, dump from tail + 1 to the end */ for (i = index; i < cjm->num_records; i++) diff --git a/src/vlib/unix/cj.h b/src/vlib/unix/cj.h index 67626afe..d0a1d46e 100644 --- a/src/vlib/unix/cj.h +++ b/src/vlib/unix/cj.h @@ -23,7 +23,7 @@ typedef struct { f64 time; - u32 cpu; + u32 thread_index; u32 type; u64 data[2]; } cj_record_t; diff --git a/src/vlib/unix/main.c b/src/vlib/unix/main.c index 6b96cc0d..db5ddd64 100644 --- a/src/vlib/unix/main.c +++ b/src/vlib/unix/main.c @@ -510,13 +510,28 @@ thread0 (uword arg) return i; } +u8 * +vlib_thread_stack_init (uword thread_index) +{ + vec_validate (vlib_thread_stacks, thread_index); + vlib_thread_stacks[thread_index] = clib_mem_alloc_aligned + (VLIB_THREAD_STACK_SIZE, VLIB_THREAD_STACK_SIZE); + + /* + * Disallow writes to the bottom page of the stack, to + * catch stack overflows. + */ + if (mprotect (vlib_thread_stacks[thread_index], + clib_mem_get_page_size (), PROT_READ) < 0) + clib_unix_warning ("thread stack"); + return vlib_thread_stacks[thread_index]; +} + int vlib_unix_main (int argc, char *argv[]) { vlib_main_t *vm = &vlib_global_main; /* one and only time for this! */ - vlib_thread_main_t *tm = &vlib_thread_main; unformat_input_t input; - u8 *thread_stacks; clib_error_t *e; int i; @@ -548,29 +563,9 @@ vlib_unix_main (int argc, char *argv[]) } unformat_free (&input); - /* - * allocate n x VLIB_THREAD_STACK_SIZE stacks, aligned to a - * VLIB_THREAD_STACK_SIZE boundary - * See also: os_get_cpu_number() in vlib/vlib/threads.c - */ - thread_stacks = clib_mem_alloc_aligned - ((uword) tm->n_thread_stacks * VLIB_THREAD_STACK_SIZE, - VLIB_THREAD_STACK_SIZE); - - vec_validate (vlib_thread_stacks, tm->n_thread_stacks - 1); - for (i = 0; i < vec_len (vlib_thread_stacks); i++) - { - vlib_thread_stacks[i] = thread_stacks; - - /* - * Disallow writes to the bottom page of the stack, to - * catch stack overflows. - */ - if (mprotect (thread_stacks, clib_mem_get_page_size (), PROT_READ) < 0) - clib_unix_warning ("thread stack"); + vlib_thread_stack_init (0); - thread_stacks += VLIB_THREAD_STACK_SIZE; - } + vlib_thread_index = 0; i = clib_calljmp (thread0, (uword) vm, (void *) (vlib_thread_stacks[0] + diff --git a/src/vnet/adj/adj_l2.c b/src/vnet/adj/adj_l2.c index f68e54e0..20d70dd4 100644 --- a/src/vnet/adj/adj_l2.c +++ b/src/vnet/adj/adj_l2.c @@ -52,7 +52,7 @@ adj_l2_rewrite_inline (vlib_main_t * vm, { u32 * from = vlib_frame_vector_args (frame); u32 n_left_from, n_left_to_next, * to_next, next_index; - u32 cpu_index = os_get_cpu_number(); + u32 thread_index = vlib_get_thread_index(); ethernet_main_t * em = ðernet_main; n_left_from = frame->n_vectors; @@ -93,7 +93,7 @@ adj_l2_rewrite_inline (vlib_main_t * vm, vnet_buffer(p0)->sw_if_index[VLIB_TX] = adj0->rewrite_header.sw_if_index; vlib_increment_combined_counter(&adjacency_counters, - cpu_index, + thread_index, adj_index0, /* packet increment */ 0, /* byte increment */ rw_len0); diff --git a/src/vnet/adj/adj_midchain.c b/src/vnet/adj/adj_midchain.c index e8087f08..5756de43 100644 --- a/src/vnet/adj/adj_midchain.c +++ b/src/vnet/adj/adj_midchain.c @@ -49,7 +49,7 @@ adj_midchain_tx_inline (vlib_main_t * vm, u32 next_index; vnet_main_t *vnm = vnet_get_main (); vnet_interface_main_t *im = &vnm->interface_main; - u32 cpu_index = vm->cpu_index; + u32 thread_index = vm->thread_index; /* Vector of buffer / pkt indices we're supposed to process */ from = vlib_frame_vector_args (frame); @@ -124,13 +124,13 @@ adj_midchain_tx_inline (vlib_main_t * vm, { vlib_increment_combined_counter (im->combined_sw_if_counters + VNET_INTERFACE_COUNTER_TX, - cpu_index, + thread_index, adj0->rewrite_header.sw_if_index, 1, vlib_buffer_length_in_chain (vm, b0)); vlib_increment_combined_counter (im->combined_sw_if_counters + VNET_INTERFACE_COUNTER_TX, - cpu_index, + thread_index, adj1->rewrite_header.sw_if_index, 1, vlib_buffer_length_in_chain (vm, b1)); @@ -181,7 +181,7 @@ adj_midchain_tx_inline (vlib_main_t * vm, { vlib_increment_combined_counter (im->combined_sw_if_counters + VNET_INTERFACE_COUNTER_TX, - cpu_index, + thread_index, adj0->rewrite_header.sw_if_index, 1, vlib_buffer_length_in_chain (vm, b0)); diff --git a/src/vnet/adj/adj_nsh.c b/src/vnet/adj/adj_nsh.c index 9a0f9d8b..128570b0 100644 --- a/src/vnet/adj/adj_nsh.c +++ b/src/vnet/adj/adj_nsh.c @@ -53,7 +53,7 @@ adj_nsh_rewrite_inline (vlib_main_t * vm, { u32 * from = vlib_frame_vector_args (frame); u32 n_left_from, n_left_to_next, * to_next, next_index; - u32 cpu_index = os_get_cpu_number(); + u32 thread_index = vlib_get_thread_index(); n_left_from = frame->n_vectors; next_index = node->cached_next_index; @@ -94,7 +94,7 @@ adj_nsh_rewrite_inline (vlib_main_t * vm, vnet_buffer(p0)->ip.save_rewrite_length = rw_len0; vlib_increment_combined_counter(&adjacency_counters, - cpu_index, + thread_index, adj_index0, /* packet increment */ 0, /* byte increment */ rw_len0); diff --git a/src/vnet/classify/vnet_classify.c b/src/vnet/classify/vnet_classify.c index 98842a48..70a189b0 100644 --- a/src/vnet/classify/vnet_classify.c +++ b/src/vnet/classify/vnet_classify.c @@ -251,12 +251,12 @@ static inline void make_working_copy vnet_classify_entry_##size##_t * working_copy##size = 0; foreach_size_in_u32x4; #undef _ - u32 cpu_number = os_get_cpu_number(); + u32 thread_index = vlib_get_thread_index(); - if (cpu_number >= vec_len (t->working_copies)) + if (thread_index >= vec_len (t->working_copies)) { oldheap = clib_mem_set_heap (t->mheap); - vec_validate (t->working_copies, cpu_number); + vec_validate (t->working_copies, thread_index); clib_mem_set_heap (oldheap); } @@ -265,7 +265,7 @@ static inline void make_working_copy * updates from multiple threads will not result in sporadic, spurious * lookup failures. */ - working_copy = t->working_copies[cpu_number]; + working_copy = t->working_copies[thread_index]; t->saved_bucket.as_u64 = b->as_u64; oldheap = clib_mem_set_heap (t->mheap); @@ -290,7 +290,7 @@ static inline void make_working_copy default: abort(); } - t->working_copies[cpu_number] = working_copy; + t->working_copies[thread_index] = working_copy; } _vec_len(working_copy) = (1<log2_pages)*t->entries_per_page; @@ -318,7 +318,7 @@ static inline void make_working_copy working_bucket.offset = vnet_classify_get_offset (t, working_copy); CLIB_MEMORY_BARRIER(); b->as_u64 = working_bucket.as_u64; - t->working_copies[cpu_number] = working_copy; + t->working_copies[thread_index] = working_copy; } static vnet_classify_entry_t * @@ -387,7 +387,7 @@ int vnet_classify_add_del (vnet_classify_table_t * t, int i; u64 hash, new_hash; u32 new_log2_pages; - u32 cpu_number = os_get_cpu_number(); + u32 thread_index = vlib_get_thread_index(); u8 * key_minus_skip; ASSERT ((add_v->flags & VNET_CLASSIFY_ENTRY_FREE) == 0); @@ -498,7 +498,7 @@ int vnet_classify_add_del (vnet_classify_table_t * t, new_log2_pages = t->saved_bucket.log2_pages + 1; expand_again: - working_copy = t->working_copies[cpu_number]; + working_copy = t->working_copies[thread_index]; new_v = split_and_rehash (t, working_copy, new_log2_pages); if (new_v == 0) diff --git a/src/vnet/cop/ip4_whitelist.c b/src/vnet/cop/ip4_whitelist.c index 6ef3d7d7..1b5e336b 100644 --- a/src/vnet/cop/ip4_whitelist.c +++ b/src/vnet/cop/ip4_whitelist.c @@ -60,7 +60,7 @@ ip4_cop_whitelist_node_fn (vlib_main_t * vm, cop_feature_type_t next_index; cop_main_t *cm = &cop_main; vlib_combined_counter_main_t * vcm = &load_balance_main.lbm_via_counters; - u32 cpu_index = vm->cpu_index; + u32 thread_index = vm->thread_index; from = vlib_frame_vector_args (frame); n_left_from = frame->n_vectors; @@ -177,12 +177,12 @@ ip4_cop_whitelist_node_fn (vlib_main_t * vm, dpo1 = load_balance_get_bucket_i(lb1, 0); vlib_increment_combined_counter - (vcm, cpu_index, lb_index0, 1, + (vcm, thread_index, lb_index0, 1, vlib_buffer_length_in_chain (vm, b0) + sizeof(ethernet_header_t)); vlib_increment_combined_counter - (vcm, cpu_index, lb_index1, 1, + (vcm, thread_index, lb_index1, 1, vlib_buffer_length_in_chain (vm, b1) + sizeof(ethernet_header_t)); @@ -273,7 +273,7 @@ ip4_cop_whitelist_node_fn (vlib_main_t * vm, dpo0 = load_balance_get_bucket_i(lb0, 0); vlib_increment_combined_counter - (vcm, cpu_index, lb_index0, 1, + (vcm, thread_index, lb_index0, 1, vlib_buffer_length_in_chain (vm, b0) + sizeof(ethernet_header_t)); diff --git a/src/vnet/cop/ip6_whitelist.c b/src/vnet/cop/ip6_whitelist.c index c2e16ccf..f3fe62e3 100644 --- a/src/vnet/cop/ip6_whitelist.c +++ b/src/vnet/cop/ip6_whitelist.c @@ -61,7 +61,7 @@ ip6_cop_whitelist_node_fn (vlib_main_t * vm, cop_main_t *cm = &cop_main; ip6_main_t * im6 = &ip6_main; vlib_combined_counter_main_t * vcm = &load_balance_main.lbm_via_counters; - u32 cpu_index = vm->cpu_index; + u32 thread_index = vm->thread_index; from = vlib_frame_vector_args (frame); n_left_from = frame->n_vectors; @@ -153,12 +153,12 @@ ip6_cop_whitelist_node_fn (vlib_main_t * vm, dpo1 = load_balance_get_bucket_i(lb1, 0); vlib_increment_combined_counter - (vcm, cpu_index, lb_index0, 1, + (vcm, thread_index, lb_index0, 1, vlib_buffer_length_in_chain (vm, b0) + sizeof(ethernet_header_t)); vlib_increment_combined_counter - (vcm, cpu_index, lb_index1, 1, + (vcm, thread_index, lb_index1, 1, vlib_buffer_length_in_chain (vm, b1) + sizeof(ethernet_header_t)); @@ -233,7 +233,7 @@ ip6_cop_whitelist_node_fn (vlib_main_t * vm, dpo0 = load_balance_get_bucket_i(lb0, 0); vlib_increment_combined_counter - (vcm, cpu_index, lb_index0, 1, + (vcm, thread_index, lb_index0, 1, vlib_buffer_length_in_chain (vm, b0) + sizeof(ethernet_header_t)); diff --git a/src/vnet/devices/af_packet/node.c b/src/vnet/devices/af_packet/node.c index ba337f3f..76980102 100644 --- a/src/vnet/devices/af_packet/node.c +++ b/src/vnet/devices/af_packet/node.c @@ -124,7 +124,7 @@ af_packet_device_input_fn (vlib_main_t * vm, vlib_node_runtime_t * node, u32 frame_num = apif->rx_req->tp_frame_nr; u8 *block_start = apif->rx_ring + block * block_size; uword n_trace = vlib_get_trace_count (vm, node); - u32 cpu_index = os_get_cpu_number (); + u32 thread_index = vlib_get_thread_index (); u32 n_buffer_bytes = vlib_buffer_free_list_buffer_size (vm, VLIB_BUFFER_DEFAULT_FREE_LIST_INDEX); u32 min_bufs = apif->rx_req->tp_frame_size / n_buffer_bytes; @@ -132,15 +132,15 @@ af_packet_device_input_fn (vlib_main_t * vm, vlib_node_runtime_t * node, if (apif->per_interface_next_index != ~0) next_index = apif->per_interface_next_index; - n_free_bufs = vec_len (apm->rx_buffers[cpu_index]); + n_free_bufs = vec_len (apm->rx_buffers[thread_index]); if (PREDICT_FALSE (n_free_bufs < VLIB_FRAME_SIZE)) { - vec_validate (apm->rx_buffers[cpu_index], + vec_validate (apm->rx_buffers[thread_index], VLIB_FRAME_SIZE + n_free_bufs - 1); n_free_bufs += - vlib_buffer_alloc (vm, &apm->rx_buffers[cpu_index][n_free_bufs], + vlib_buffer_alloc (vm, &apm->rx_buffers[thread_index][n_free_bufs], VLIB_FRAME_SIZE); - _vec_len (apm->rx_buffers[cpu_index]) = n_free_bufs; + _vec_len (apm->rx_buffers[thread_index]) = n_free_bufs; } rx_frame = apif->next_rx_frame; @@ -163,11 +163,11 @@ af_packet_device_input_fn (vlib_main_t * vm, vlib_node_runtime_t * node, { /* grab free buffer */ u32 last_empty_buffer = - vec_len (apm->rx_buffers[cpu_index]) - 1; + vec_len (apm->rx_buffers[thread_index]) - 1; prev_bi0 = bi0; - bi0 = apm->rx_buffers[cpu_index][last_empty_buffer]; + bi0 = apm->rx_buffers[thread_index][last_empty_buffer]; b0 = vlib_get_buffer (vm, bi0); - _vec_len (apm->rx_buffers[cpu_index]) = last_empty_buffer; + _vec_len (apm->rx_buffers[thread_index]) = last_empty_buffer; n_free_bufs--; /* copy data */ @@ -236,9 +236,9 @@ af_packet_device_input_fn (vlib_main_t * vm, vlib_node_runtime_t * node, vlib_increment_combined_counter (vnet_get_main ()->interface_main.combined_sw_if_counters + VNET_INTERFACE_COUNTER_RX, - os_get_cpu_number (), apif->hw_if_index, n_rx_packets, n_rx_bytes); + vlib_get_thread_index (), apif->hw_if_index, n_rx_packets, n_rx_bytes); - vnet_device_increment_rx_packets (cpu_index, n_rx_packets); + vnet_device_increment_rx_packets (thread_index, n_rx_packets); return n_rx_packets; } diff --git a/src/vnet/devices/devices.c b/src/vnet/devices/devices.c index 41645220..5e5e812c 100644 --- a/src/vnet/devices/devices.c +++ b/src/vnet/devices/devices.c @@ -104,7 +104,7 @@ vnet_device_queue_sort (void *a1, void *a2) void vnet_device_input_assign_thread (u32 hw_if_index, - u16 queue_id, uword cpu_index) + u16 queue_id, uword thread_index) { vnet_main_t *vnm = vnet_get_main (); vnet_device_main_t *vdm = &vnet_device_main; @@ -115,19 +115,19 @@ vnet_device_input_assign_thread (u32 hw_if_index, ASSERT (hw->input_node_index > 0); - if (vdm->first_worker_cpu_index == 0) - cpu_index = 0; + if (vdm->first_worker_thread_index == 0) + thread_index = 0; - if (cpu_index != 0 && - (cpu_index < vdm->first_worker_cpu_index || - cpu_index > vdm->last_worker_cpu_index)) + if (thread_index != 0 && + (thread_index < vdm->first_worker_thread_index || + thread_index > vdm->last_worker_thread_index)) { - cpu_index = vdm->next_worker_cpu_index++; - if (vdm->next_worker_cpu_index > vdm->last_worker_cpu_index) - vdm->next_worker_cpu_index = vdm->first_worker_cpu_index; + thread_index = vdm->next_worker_thread_index++; + if (vdm->next_worker_thread_index > vdm->last_worker_thread_index) + vdm->next_worker_thread_index = vdm->first_worker_thread_index; } - vm = vlib_mains[cpu_index]; + vm = vlib_mains[thread_index]; rt = vlib_node_get_runtime_data (vm, hw->input_node_index); vec_add2 (rt->devices_and_queues, dq, 1); @@ -136,33 +136,33 @@ vnet_device_input_assign_thread (u32 hw_if_index, dq->queue_id = queue_id; vec_sort_with_function (rt->devices_and_queues, vnet_device_queue_sort); - vec_validate (hw->input_node_cpu_index_by_queue, queue_id); - hw->input_node_cpu_index_by_queue[queue_id] = cpu_index; + vec_validate (hw->input_node_thread_index_by_queue, queue_id); + hw->input_node_thread_index_by_queue[queue_id] = thread_index; } static int vnet_device_input_unassign_thread (u32 hw_if_index, u16 queue_id, - uword cpu_index) + uword thread_index) { vnet_main_t *vnm = vnet_get_main (); vnet_hw_interface_t *hw = vnet_get_hw_interface (vnm, hw_if_index); vnet_device_input_runtime_t *rt; vnet_device_and_queue_t *dq; - uword old_cpu_index; + uword old_thread_index; - if (hw->input_node_cpu_index_by_queue == 0) + if (hw->input_node_thread_index_by_queue == 0) return VNET_API_ERROR_INVALID_INTERFACE; - if (vec_len (hw->input_node_cpu_index_by_queue) < queue_id + 1) + if (vec_len (hw->input_node_thread_index_by_queue) < queue_id + 1) return VNET_API_ERROR_INVALID_INTERFACE; - old_cpu_index = hw->input_node_cpu_index_by_queue[queue_id]; + old_thread_index = hw->input_node_thread_index_by_queue[queue_id]; - if (old_cpu_index == cpu_index) + if (old_thread_index == thread_index) return 0; rt = - vlib_node_get_runtime_data (vlib_mains[old_cpu_index], + vlib_node_get_runtime_data (vlib_mains[old_thread_index], hw->input_node_index); vec_foreach (dq, rt->devices_and_queues) @@ -240,7 +240,7 @@ set_device_placement (vlib_main_t * vm, unformat_input_t * input, vnet_device_main_t *vdm = &vnet_device_main; u32 hw_if_index = (u32) ~ 0; u32 queue_id = (u32) 0; - u32 cpu_index = (u32) ~ 0; + u32 thread_index = (u32) ~ 0; int rv; if (!unformat_user (input, unformat_line_input, line_input)) @@ -253,10 +253,10 @@ set_device_placement (vlib_main_t * vm, unformat_input_t * input, ; else if (unformat (line_input, "queue %d", &queue_id)) ; - else if (unformat (line_input, "main", &cpu_index)) - cpu_index = 0; - else if (unformat (line_input, "worker %d", &cpu_index)) - cpu_index += vdm->first_worker_cpu_index; + else if (unformat (line_input, "main", &thread_index)) + thread_index = 0; + else if (unformat (line_input, "worker %d", &thread_index)) + thread_index += vdm->first_worker_thread_index; else { error = clib_error_return (0, "parse error: '%U'", @@ -271,16 +271,17 @@ set_device_placement (vlib_main_t * vm, unformat_input_t * input, if (hw_if_index == (u32) ~ 0) return clib_error_return (0, "please specify valid interface name"); - if (cpu_index > vdm->last_worker_cpu_index) + if (thread_index > vdm->last_worker_thread_index) return clib_error_return (0, "please specify valid worker thread or main"); - rv = vnet_device_input_unassign_thread (hw_if_index, queue_id, cpu_index); + rv = + vnet_device_input_unassign_thread (hw_if_index, queue_id, thread_index); if (rv) return clib_error_return (0, "not found"); - vnet_device_input_assign_thread (hw_if_index, queue_id, cpu_index); + vnet_device_input_assign_thread (hw_if_index, queue_id, thread_index); return 0; } @@ -326,9 +327,9 @@ vnet_device_init (vlib_main_t * vm) tr = p ? (vlib_thread_registration_t *) p[0] : 0; if (tr && tr->count > 0) { - vdm->first_worker_cpu_index = tr->first_index; - vdm->next_worker_cpu_index = tr->first_index; - vdm->last_worker_cpu_index = tr->first_index + tr->count - 1; + vdm->first_worker_thread_index = tr->first_index; + vdm->next_worker_thread_index = tr->first_index; + vdm->last_worker_thread_index = tr->first_index + tr->count - 1; } return 0; } diff --git a/src/vnet/devices/devices.h b/src/vnet/devices/devices.h index bbb29fe3..966f8302 100644 --- a/src/vnet/devices/devices.h +++ b/src/vnet/devices/devices.h @@ -50,9 +50,9 @@ typedef struct typedef struct { vnet_device_per_worker_data_t *workers; - uword first_worker_cpu_index; - uword last_worker_cpu_index; - uword next_worker_cpu_index; + uword first_worker_thread_index; + uword last_worker_thread_index; + uword next_worker_thread_index; } vnet_device_main_t; typedef struct @@ -80,7 +80,7 @@ vnet_set_device_input_node (u32 hw_if_index, u32 node_index) } void vnet_device_input_assign_thread (u32 hw_if_index, u16 queue_id, - uword cpu_index); + uword thread_index); static inline u64 vnet_get_aggregate_rx_packets (void) @@ -95,12 +95,12 @@ vnet_get_aggregate_rx_packets (void) } static inline void -vnet_device_increment_rx_packets (u32 cpu_index, u64 count) +vnet_device_increment_rx_packets (u32 thread_index, u64 count) { vnet_device_main_t *vdm = &vnet_device_main; vnet_device_per_worker_data_t *pwd; - pwd = vec_elt_at_index (vdm->workers, cpu_index); + pwd = vec_elt_at_index (vdm->workers, thread_index); pwd->aggregate_rx_packets += count; } @@ -117,9 +117,9 @@ vnet_device_input_set_interrupt_pending (vnet_main_t * vnm, u32 hw_if_index, { vnet_hw_interface_t *hw = vnet_get_hw_interface (vnm, hw_if_index); - ASSERT (queue_id < vec_len (hw->input_node_cpu_index_by_queue)); - u32 cpu_index = hw->input_node_cpu_index_by_queue[queue_id]; - vlib_node_set_interrupt_pending (vlib_mains[cpu_index], + ASSERT (queue_id < vec_len (hw->input_node_thread_index_by_queue)); + u32 thread_index = hw->input_node_thread_index_by_queue[queue_id]; + vlib_node_set_interrupt_pending (vlib_mains[thread_index], hw->input_node_index); } diff --git a/src/vnet/devices/netmap/node.c b/src/vnet/devices/netmap/node.c index 68ea7832..e120eeae 100644 --- a/src/vnet/devices/netmap/node.c +++ b/src/vnet/devices/netmap/node.c @@ -98,22 +98,22 @@ netmap_device_input_fn (vlib_main_t * vm, vlib_node_runtime_t * node, u32 n_free_bufs; struct netmap_ring *ring; int cur_ring; - u32 cpu_index = os_get_cpu_number (); + u32 thread_index = vlib_get_thread_index (); u32 n_buffer_bytes = vlib_buffer_free_list_buffer_size (vm, VLIB_BUFFER_DEFAULT_FREE_LIST_INDEX); if (nif->per_interface_next_index != ~0) next_index = nif->per_interface_next_index; - n_free_bufs = vec_len (nm->rx_buffers[cpu_index]); + n_free_bufs = vec_len (nm->rx_buffers[thread_index]); if (PREDICT_FALSE (n_free_bufs < VLIB_FRAME_SIZE)) { - vec_validate (nm->rx_buffers[cpu_index], + vec_validate (nm->rx_buffers[thread_index], VLIB_FRAME_SIZE + n_free_bufs - 1); n_free_bufs += - vlib_buffer_alloc (vm, &nm->rx_buffers[cpu_index][n_free_bufs], + vlib_buffer_alloc (vm, &nm->rx_buffers[thread_index][n_free_bufs], VLIB_FRAME_SIZE); - _vec_len (nm->rx_buffers[cpu_index]) = n_free_bufs; + _vec_len (nm->rx_buffers[thread_index]) = n_free_bufs; } cur_ring = nif->first_rx_ring; @@ -163,11 +163,11 @@ netmap_device_input_fn (vlib_main_t * vm, vlib_node_runtime_t * node, vlib_buffer_t *b0; /* grab free buffer */ u32 last_empty_buffer = - vec_len (nm->rx_buffers[cpu_index]) - 1; + vec_len (nm->rx_buffers[thread_index]) - 1; prev_bi0 = bi0; - bi0 = nm->rx_buffers[cpu_index][last_empty_buffer]; + bi0 = nm->rx_buffers[thread_index][last_empty_buffer]; b0 = vlib_get_buffer (vm, bi0); - _vec_len (nm->rx_buffers[cpu_index]) = last_empty_buffer; + _vec_len (nm->rx_buffers[thread_index]) = last_empty_buffer; n_free_bufs--; /* copy data */ @@ -247,9 +247,9 @@ netmap_device_input_fn (vlib_main_t * vm, vlib_node_runtime_t * node, vlib_increment_combined_counter (vnet_get_main ()->interface_main.combined_sw_if_counters + VNET_INTERFACE_COUNTER_RX, - os_get_cpu_number (), nif->hw_if_index, n_rx_packets, n_rx_bytes); + vlib_get_thread_index (), nif->hw_if_index, n_rx_packets, n_rx_bytes); - vnet_device_increment_rx_packets (cpu_index, n_rx_packets); + vnet_device_increment_rx_packets (thread_index, n_rx_packets); return n_rx_packets; } @@ -260,7 +260,7 @@ netmap_input_fn (vlib_main_t * vm, vlib_node_runtime_t * node, { int i; u32 n_rx_packets = 0; - u32 cpu_index = os_get_cpu_number (); + u32 thread_index = vlib_get_thread_index (); netmap_main_t *nm = &netmap_main; netmap_if_t *nmi; @@ -269,7 +269,7 @@ netmap_input_fn (vlib_main_t * vm, vlib_node_runtime_t * node, nmi = vec_elt_at_index (nm->interfaces, i); if (nmi->is_admin_up && (i % nm->input_cpu_count) == - (cpu_index - nm->input_cpu_first_index)) + (thread_index - nm->input_cpu_first_index)) n_rx_packets += netmap_device_input_fn (vm, node, frame, nmi); } diff --git a/src/vnet/devices/ssvm/node.c b/src/vnet/devices/ssvm/node.c index a6c9dfd7..539b4161 100644 --- a/src/vnet/devices/ssvm/node.c +++ b/src/vnet/devices/ssvm/node.c @@ -89,7 +89,7 @@ ssvm_eth_device_input (ssvm_eth_main_t * em, ethernet_header_t *eh0; u16 type0; u32 n_rx_bytes = 0, l3_offset0; - u32 cpu_index = os_get_cpu_number (); + u32 thread_index = vlib_get_thread_index (); u32 trace_cnt __attribute__ ((unused)) = vlib_get_trace_count (vm, node); volatile u32 *lock; u32 *elt_indices; @@ -284,10 +284,10 @@ out: vlib_increment_combined_counter (vnet_get_main ()->interface_main.combined_sw_if_counters - + VNET_INTERFACE_COUNTER_RX, cpu_index, + + VNET_INTERFACE_COUNTER_RX, thread_index, intfc->vlib_hw_if_index, rx_queue_index, n_rx_bytes); - vnet_device_increment_rx_packets (cpu_index, rx_queue_index); + vnet_device_increment_rx_packets (thread_index, rx_queue_index); return rx_queue_index; } diff --git a/src/vnet/devices/virtio/vhost-user.c b/src/vnet/devices/virtio/vhost-user.c index 00807dc0..5e720f65 100644 --- a/src/vnet/devices/virtio/vhost-user.c +++ b/src/vnet/devices/virtio/vhost-user.c @@ -331,7 +331,7 @@ vhost_user_tx_thread_placement (vhost_user_intf_t * vui) { //Let's try to assign one queue to each thread u32 qid = 0; - u32 cpu_index = 0; + u32 thread_index = 0; vui->use_tx_spinlock = 0; while (1) { @@ -341,20 +341,21 @@ vhost_user_tx_thread_placement (vhost_user_intf_t * vui) if (!rxvq->started || !rxvq->enabled) continue; - vui->per_cpu_tx_qid[cpu_index] = qid; - cpu_index++; - if (cpu_index == vlib_get_thread_main ()->n_vlib_mains) + vui->per_cpu_tx_qid[thread_index] = qid; + thread_index++; + if (thread_index == vlib_get_thread_main ()->n_vlib_mains) return; } //We need to loop, meaning the spinlock has to be used vui->use_tx_spinlock = 1; - if (cpu_index == 0) + if (thread_index == 0) { //Could not find a single valid one - for (cpu_index = 0; - cpu_index < vlib_get_thread_main ()->n_vlib_mains; cpu_index++) + for (thread_index = 0; + thread_index < vlib_get_thread_main ()->n_vlib_mains; + thread_index++) { - vui->per_cpu_tx_qid[cpu_index] = 0; + vui->per_cpu_tx_qid[thread_index] = 0; } return; } @@ -368,7 +369,7 @@ vhost_user_rx_thread_placement () vhost_user_intf_t *vui; vhost_cpu_t *vhc; u32 *workers = 0; - u32 cpu_index; + u32 thread_index; vlib_main_t *vm; //Let's list all workers cpu indexes @@ -400,9 +401,9 @@ vhost_user_rx_thread_placement () continue; i %= vec_len (vui_workers); - cpu_index = vui_workers[i]; + thread_index = vui_workers[i]; i++; - vhc = &vum->cpus[cpu_index]; + vhc = &vum->cpus[thread_index]; iaq.qid = qid; iaq.vhost_iface_index = vui - vum->vhost_user_interfaces; @@ -429,14 +430,14 @@ vhost_user_rx_thread_placement () vhc->operation_mode = mode; } - for (cpu_index = vum->input_cpu_first_index; - cpu_index < vum->input_cpu_first_index + vum->input_cpu_count; - cpu_index++) + for (thread_index = vum->input_cpu_first_index; + thread_index < vum->input_cpu_first_index + vum->input_cpu_count; + thread_index++) { vlib_node_state_t state = VLIB_NODE_STATE_POLLING; - vhc = &vum->cpus[cpu_index]; - vm = vlib_mains ? vlib_mains[cpu_index] : &vlib_global_main; + vhc = &vum->cpus[thread_index]; + vm = vlib_mains ? vlib_mains[thread_index] : &vlib_global_main; switch (vhc->operation_mode) { case VHOST_USER_INTERRUPT_MODE: @@ -532,7 +533,7 @@ vhost_user_set_interrupt_pending (vhost_user_intf_t * vui, u32 ifq) { vhost_user_main_t *vum = &vhost_user_main; vhost_cpu_t *vhc; - u32 cpu_index; + u32 thread_index; vhost_iface_and_queue_t *vhiq; vlib_main_t *vm; u32 ifq2; @@ -553,8 +554,8 @@ vhost_user_set_interrupt_pending (vhost_user_intf_t * vui, u32 ifq) if ((vhiq->vhost_iface_index == (ifq >> 8)) && (VHOST_VRING_IDX_TX (vhiq->qid) == (ifq & 0xff))) { - cpu_index = vhc - vum->cpus; - vm = vlib_mains ? vlib_mains[cpu_index] : &vlib_global_main; + thread_index = vhc - vum->cpus; + vm = vlib_mains ? vlib_mains[thread_index] : &vlib_global_main; /* * Convert RX virtqueue number in the lower byte to vring * queue index for the input node process. Top bytes contain @@ -1592,7 +1593,7 @@ vhost_user_if_input (vlib_main_t * vm, u32 n_trace = vlib_get_trace_count (vm, node); u16 qsz_mask; u32 map_hint = 0; - u16 cpu_index = os_get_cpu_number (); + u16 thread_index = vlib_get_thread_index (); u16 copy_len = 0; { @@ -1651,32 +1652,32 @@ vhost_user_if_input (vlib_main_t * vm, * in the loop and come back later. This is not an issue as for big packet, * processing cost really comes from the memory copy. */ - if (PREDICT_FALSE (vum->cpus[cpu_index].rx_buffers_len < n_left + 1)) + if (PREDICT_FALSE (vum->cpus[thread_index].rx_buffers_len < n_left + 1)) { - u32 curr_len = vum->cpus[cpu_index].rx_buffers_len; - vum->cpus[cpu_index].rx_buffers_len += + u32 curr_len = vum->cpus[thread_index].rx_buffers_len; + vum->cpus[thread_index].rx_buffers_len += vlib_buffer_alloc_from_free_list (vm, - vum->cpus[cpu_index].rx_buffers + + vum->cpus[thread_index].rx_buffers + curr_len, VHOST_USER_RX_BUFFERS_N - curr_len, VLIB_BUFFER_DEFAULT_FREE_LIST_INDEX); if (PREDICT_FALSE - (vum->cpus[cpu_index].rx_buffers_len < + (vum->cpus[thread_index].rx_buffers_len < VHOST_USER_RX_BUFFER_STARVATION)) { /* In case of buffer starvation, discard some packets from the queue * and log the event. * We keep doing best effort for the remaining packets. */ - u32 flush = (n_left + 1 > vum->cpus[cpu_index].rx_buffers_len) ? - n_left + 1 - vum->cpus[cpu_index].rx_buffers_len : 1; + u32 flush = (n_left + 1 > vum->cpus[thread_index].rx_buffers_len) ? + n_left + 1 - vum->cpus[thread_index].rx_buffers_len : 1; flush = vhost_user_rx_discard_packet (vm, vui, txvq, flush); n_left -= flush; vlib_increment_simple_counter (vnet_main. interface_main.sw_if_counters + VNET_INTERFACE_COUNTER_DROP, - os_get_cpu_number (), + vlib_get_thread_index (), vui->sw_if_index, flush); vlib_error_count (vm, vhost_user_input_node.index, @@ -1696,7 +1697,7 @@ vhost_user_if_input (vlib_main_t * vm, u32 desc_data_offset; vring_desc_t *desc_table = txvq->desc; - if (PREDICT_FALSE (vum->cpus[cpu_index].rx_buffers_len <= 1)) + if (PREDICT_FALSE (vum->cpus[thread_index].rx_buffers_len <= 1)) { /* Not enough rx_buffers * Note: We yeld on 1 so we don't need to do an additional @@ -1707,17 +1708,18 @@ vhost_user_if_input (vlib_main_t * vm, } desc_current = txvq->avail->ring[txvq->last_avail_idx & qsz_mask]; - vum->cpus[cpu_index].rx_buffers_len--; - bi_current = (vum->cpus[cpu_index].rx_buffers) - [vum->cpus[cpu_index].rx_buffers_len]; + vum->cpus[thread_index].rx_buffers_len--; + bi_current = (vum->cpus[thread_index].rx_buffers) + [vum->cpus[thread_index].rx_buffers_len]; b_head = b_current = vlib_get_buffer (vm, bi_current); to_next[0] = bi_current; //We do that now so we can forget about bi_current to_next++; n_left_to_next--; vlib_prefetch_buffer_with_index (vm, - (vum->cpus[cpu_index].rx_buffers) - [vum->cpus[cpu_index]. + (vum-> + cpus[thread_index].rx_buffers) + [vum->cpus[thread_index]. rx_buffers_len - 1], LOAD); /* Just preset the used descriptor id and length for later */ @@ -1791,7 +1793,7 @@ vhost_user_if_input (vlib_main_t * vm, (b_current->current_length == VLIB_BUFFER_DATA_SIZE)) { if (PREDICT_FALSE - (vum->cpus[cpu_index].rx_buffers_len == 0)) + (vum->cpus[thread_index].rx_buffers_len == 0)) { /* Cancel speculation */ to_next--; @@ -1805,17 +1807,18 @@ vhost_user_if_input (vlib_main_t * vm, * but valid. */ vhost_user_input_rewind_buffers (vm, - &vum->cpus[cpu_index], + &vum->cpus + [thread_index], b_head); n_left = 0; goto stop; } /* Get next output */ - vum->cpus[cpu_index].rx_buffers_len--; + vum->cpus[thread_index].rx_buffers_len--; u32 bi_next = - (vum->cpus[cpu_index].rx_buffers)[vum->cpus - [cpu_index].rx_buffers_len]; + (vum->cpus[thread_index].rx_buffers)[vum->cpus + [thread_index].rx_buffers_len]; b_current->next_buffer = bi_next; b_current->flags |= VLIB_BUFFER_NEXT_PRESENT; bi_current = bi_next; @@ -1823,7 +1826,7 @@ vhost_user_if_input (vlib_main_t * vm, } /* Prepare a copy order executed later for the data */ - vhost_copy_t *cpy = &vum->cpus[cpu_index].copy[copy_len]; + vhost_copy_t *cpy = &vum->cpus[thread_index].copy[copy_len]; copy_len++; u32 desc_data_l = desc_table[desc_current].len - desc_data_offset; @@ -1880,7 +1883,7 @@ vhost_user_if_input (vlib_main_t * vm, if (PREDICT_FALSE (copy_len >= VHOST_USER_RX_COPY_THRESHOLD)) { if (PREDICT_FALSE - (vhost_user_input_copy (vui, vum->cpus[cpu_index].copy, + (vhost_user_input_copy (vui, vum->cpus[thread_index].copy, copy_len, &map_hint))) { clib_warning @@ -1905,7 +1908,7 @@ vhost_user_if_input (vlib_main_t * vm, /* Do the memory copies */ if (PREDICT_FALSE - (vhost_user_input_copy (vui, vum->cpus[cpu_index].copy, + (vhost_user_input_copy (vui, vum->cpus[thread_index].copy, copy_len, &map_hint))) { clib_warning ("Memory mapping error on interface hw_if_index=%d " @@ -1933,9 +1936,9 @@ vhost_user_if_input (vlib_main_t * vm, vlib_increment_combined_counter (vnet_main.interface_main.combined_sw_if_counters + VNET_INTERFACE_COUNTER_RX, - os_get_cpu_number (), vui->sw_if_index, n_rx_packets, n_rx_bytes); + vlib_get_thread_index (), vui->sw_if_index, n_rx_packets, n_rx_bytes); - vnet_device_increment_rx_packets (cpu_index, n_rx_packets); + vnet_device_increment_rx_packets (thread_index, n_rx_packets); return n_rx_packets; } @@ -1946,15 +1949,15 @@ vhost_user_input (vlib_main_t * vm, { vhost_user_main_t *vum = &vhost_user_main; uword n_rx_packets = 0; - u32 cpu_index = os_get_cpu_number (); + u32 thread_index = vlib_get_thread_index (); vhost_iface_and_queue_t *vhiq; vhost_user_intf_t *vui; vhost_cpu_t *vhc; - vhc = &vum->cpus[cpu_index]; + vhc = &vum->cpus[thread_index]; if (PREDICT_TRUE (vhc->operation_mode == VHOST_USER_POLLING_MODE)) { - vec_foreach (vhiq, vum->cpus[cpu_index].rx_queues) + vec_foreach (vhiq, vum->cpus[thread_index].rx_queues) { vui = &vum->vhost_user_interfaces[vhiq->vhost_iface_index]; n_rx_packets += vhost_user_if_input (vm, vum, vui, vhiq->qid, node); @@ -2096,7 +2099,7 @@ vhost_user_tx (vlib_main_t * vm, vhost_user_vring_t *rxvq; u16 qsz_mask; u8 error; - u32 cpu_index = os_get_cpu_number (); + u32 thread_index = vlib_get_thread_index (); u32 map_hint = 0; u8 retry = 8; u16 copy_len; @@ -2116,7 +2119,7 @@ vhost_user_tx (vlib_main_t * vm, qid = VHOST_VRING_IDX_RX (*vec_elt_at_index - (vui->per_cpu_tx_qid, os_get_cpu_number ())); + (vui->per_cpu_tx_qid, vlib_get_thread_index ())); rxvq = &vui->vrings[qid]; if (PREDICT_FALSE (vui->use_tx_spinlock)) vhost_user_vring_lock (vui, qid); @@ -2143,10 +2146,10 @@ retry: if (PREDICT_FALSE (b0->flags & VLIB_BUFFER_IS_TRACED)) { - vum->cpus[cpu_index].current_trace = + vum->cpus[thread_index].current_trace = vlib_add_trace (vm, node, b0, - sizeof (*vum->cpus[cpu_index].current_trace)); - vhost_user_tx_trace (vum->cpus[cpu_index].current_trace, + sizeof (*vum->cpus[thread_index].current_trace)); + vhost_user_tx_trace (vum->cpus[thread_index].current_trace, vui, qid / 2, b0, rxvq); } @@ -2188,14 +2191,14 @@ retry: { // Get a header from the header array virtio_net_hdr_mrg_rxbuf_t *hdr = - &vum->cpus[cpu_index].tx_headers[tx_headers_len]; + &vum->cpus[thread_index].tx_headers[tx_headers_len]; tx_headers_len++; hdr->hdr.flags = 0; hdr->hdr.gso_type = 0; hdr->num_buffers = 1; //This is local, no need to check // Prepare a copy order executed later for the header - vhost_copy_t *cpy = &vum->cpus[cpu_index].copy[copy_len]; + vhost_copy_t *cpy = &vum->cpus[thread_index].copy[copy_len]; copy_len++; cpy->len = vui->virtio_net_hdr_sz; cpy->dst = buffer_map_addr; @@ -2220,7 +2223,7 @@ retry: else if (vui->virtio_net_hdr_sz == 12) //MRG is available { virtio_net_hdr_mrg_rxbuf_t *hdr = - &vum->cpus[cpu_index].tx_headers[tx_headers_len - 1]; + &vum->cpus[thread_index].tx_headers[tx_headers_len - 1]; //Move from available to used buffer rxvq->used->ring[rxvq->last_used_idx & qsz_mask].id = @@ -2282,7 +2285,7 @@ retry: } { - vhost_copy_t *cpy = &vum->cpus[cpu_index].copy[copy_len]; + vhost_copy_t *cpy = &vum->cpus[thread_index].copy[copy_len]; copy_len++; cpy->len = bytes_left; cpy->len = (cpy->len > buffer_len) ? buffer_len : cpy->len; @@ -2325,8 +2328,8 @@ retry: if (PREDICT_FALSE (b0->flags & VLIB_BUFFER_IS_TRACED)) { - vum->cpus[cpu_index].current_trace->hdr = - vum->cpus[cpu_index].tx_headers[tx_headers_len - 1]; + vum->cpus[thread_index].current_trace->hdr = + vum->cpus[thread_index].tx_headers[tx_headers_len - 1]; } n_left--; //At the end for error counting when 'goto done' is invoked @@ -2336,7 +2339,7 @@ retry: done: //Do the memory copies if (PREDICT_FALSE - (vhost_user_tx_copy (vui, vum->cpus[cpu_index].copy, + (vhost_user_tx_copy (vui, vum->cpus[thread_index].copy, copy_len, &map_hint))) { clib_warning ("Memory mapping error on interface hw_if_index=%d " @@ -2386,7 +2389,7 @@ done3: vlib_increment_simple_counter (vnet_main.interface_main.sw_if_counters + VNET_INTERFACE_COUNTER_DROP, - os_get_cpu_number (), vui->sw_if_index, n_left); + vlib_get_thread_index (), vui->sw_if_index, n_left); } vlib_buffer_free (vm, vlib_frame_args (frame), frame->n_vectors); @@ -2773,11 +2776,11 @@ vhost_user_send_interrupt_process (vlib_main_t * vm, case ~0: vec_foreach (vhc, vum->cpus) { - u32 cpu_index = vhc - vum->cpus; + u32 thread_index = vhc - vum->cpus; f64 next_timeout; next_timeout = timeout; - vec_foreach (vhiq, vum->cpus[cpu_index].rx_queues) + vec_foreach (vhiq, vum->cpus[thread_index].rx_queues) { vui = &vum->vhost_user_interfaces[vhiq->vhost_iface_index]; vhost_user_vring_t *rxvq = diff --git a/src/vnet/dpo/lookup_dpo.c b/src/vnet/dpo/lookup_dpo.c index e94e871c..97ad0a44 100644 --- a/src/vnet/dpo/lookup_dpo.c +++ b/src/vnet/dpo/lookup_dpo.c @@ -266,7 +266,7 @@ lookup_dpo_ip4_inline (vlib_main_t * vm, int table_from_interface) { u32 n_left_from, next_index, * from, * to_next; - u32 cpu_index = os_get_cpu_number(); + u32 thread_index = vlib_get_thread_index(); vlib_combined_counter_main_t * cm = &load_balance_main.lbm_to_counters; from = vlib_frame_vector_args (from_frame); @@ -407,10 +407,10 @@ lookup_dpo_ip4_inline (vlib_main_t * vm, vnet_buffer(b1)->ip.adj_index[VLIB_TX] = dpo1->dpoi_index; vlib_increment_combined_counter - (cm, cpu_index, lbi0, 1, + (cm, thread_index, lbi0, 1, vlib_buffer_length_in_chain (vm, b0)); vlib_increment_combined_counter - (cm, cpu_index, lbi1, 1, + (cm, thread_index, lbi1, 1, vlib_buffer_length_in_chain (vm, b1)); if (PREDICT_FALSE(b0->flags & VLIB_BUFFER_IS_TRACED)) @@ -511,7 +511,7 @@ lookup_dpo_ip4_inline (vlib_main_t * vm, vnet_buffer(b0)->ip.adj_index[VLIB_TX] = dpo0->dpoi_index; vlib_increment_combined_counter - (cm, cpu_index, lbi0, 1, + (cm, thread_index, lbi0, 1, vlib_buffer_length_in_chain (vm, b0)); if (PREDICT_FALSE(b0->flags & VLIB_BUFFER_IS_TRACED)) @@ -606,7 +606,7 @@ lookup_dpo_ip6_inline (vlib_main_t * vm, { vlib_combined_counter_main_t * cm = &load_balance_main.lbm_to_counters; u32 n_left_from, next_index, * from, * to_next; - u32 cpu_index = os_get_cpu_number(); + u32 thread_index = vlib_get_thread_index(); from = vlib_frame_vector_args (from_frame); n_left_from = from_frame->n_vectors; @@ -749,10 +749,10 @@ lookup_dpo_ip6_inline (vlib_main_t * vm, vnet_buffer(b1)->ip.adj_index[VLIB_TX] = dpo1->dpoi_index; vlib_increment_combined_counter - (cm, cpu_index, lbi0, 1, + (cm, thread_index, lbi0, 1, vlib_buffer_length_in_chain (vm, b0)); vlib_increment_combined_counter - (cm, cpu_index, lbi1, 1, + (cm, thread_index, lbi1, 1, vlib_buffer_length_in_chain (vm, b1)); if (PREDICT_FALSE(b0->flags & VLIB_BUFFER_IS_TRACED)) @@ -853,7 +853,7 @@ lookup_dpo_ip6_inline (vlib_main_t * vm, vnet_buffer(b0)->ip.adj_index[VLIB_TX] = dpo0->dpoi_index; vlib_increment_combined_counter - (cm, cpu_index, lbi0, 1, + (cm, thread_index, lbi0, 1, vlib_buffer_length_in_chain (vm, b0)); if (PREDICT_FALSE(b0->flags & VLIB_BUFFER_IS_TRACED)) @@ -930,7 +930,7 @@ lookup_dpo_mpls_inline (vlib_main_t * vm, int table_from_interface) { u32 n_left_from, next_index, * from, * to_next; - u32 cpu_index = os_get_cpu_number(); + u32 thread_index = vlib_get_thread_index(); vlib_combined_counter_main_t * cm = &load_balance_main.lbm_to_counters; from = vlib_frame_vector_args (from_frame); @@ -994,7 +994,7 @@ lookup_dpo_mpls_inline (vlib_main_t * vm, vnet_buffer(b0)->ip.adj_index[VLIB_TX] = dpo0->dpoi_index; vlib_increment_combined_counter - (cm, cpu_index, lbi0, 1, + (cm, thread_index, lbi0, 1, vlib_buffer_length_in_chain (vm, b0)); if (PREDICT_FALSE(b0->flags & VLIB_BUFFER_IS_TRACED)) diff --git a/src/vnet/dpo/replicate_dpo.c b/src/vnet/dpo/replicate_dpo.c index a9f334be..e25ceae9 100644 --- a/src/vnet/dpo/replicate_dpo.c +++ b/src/vnet/dpo/replicate_dpo.c @@ -627,7 +627,7 @@ replicate_inline (vlib_main_t * vm, vlib_combined_counter_main_t * cm = &replicate_main.repm_counters; replicate_main_t * rm = &replicate_main; u32 n_left_from, * from, * to_next, next_index; - u32 cpu_index = os_get_cpu_number(); + u32 thread_index = vlib_get_thread_index(); from = vlib_frame_vector_args (frame); n_left_from = frame->n_vectors; @@ -657,12 +657,12 @@ replicate_inline (vlib_main_t * vm, rep0 = replicate_get(repi0); vlib_increment_combined_counter( - cm, cpu_index, repi0, 1, + cm, thread_index, repi0, 1, vlib_buffer_length_in_chain(vm, b0)); - vec_validate (rm->clones[cpu_index], rep0->rep_n_buckets - 1); + vec_validate (rm->clones[thread_index], rep0->rep_n_buckets - 1); - num_cloned = vlib_buffer_clone (vm, bi0, rm->clones[cpu_index], rep0->rep_n_buckets, 128); + num_cloned = vlib_buffer_clone (vm, bi0, rm->clones[thread_index], rep0->rep_n_buckets, 128); if (num_cloned != rep0->rep_n_buckets) { @@ -673,7 +673,7 @@ replicate_inline (vlib_main_t * vm, for (bucket = 0; bucket < num_cloned; bucket++) { - ci0 = rm->clones[cpu_index][bucket]; + ci0 = rm->clones[thread_index][bucket]; c0 = vlib_get_buffer(vm, ci0); to_next[0] = ci0; @@ -700,7 +700,7 @@ replicate_inline (vlib_main_t * vm, vlib_get_next_frame (vm, node, next_index, to_next, n_left_to_next); } } - vec_reset_length (rm->clones[cpu_index]); + vec_reset_length (rm->clones[thread_index]); } vlib_put_next_frame (vm, node, next_index, n_left_to_next); diff --git a/src/vnet/ethernet/arp.c b/src/vnet/ethernet/arp.c index ee757505..c74a097e 100644 --- a/src/vnet/ethernet/arp.c +++ b/src/vnet/ethernet/arp.c @@ -1771,7 +1771,7 @@ set_ip4_over_ethernet_rpc_callback (vnet_arp_set_ip4_over_ethernet_rpc_args_t * a) { vnet_main_t *vm = vnet_get_main (); - ASSERT (os_get_cpu_number () == 0); + ASSERT (vlib_get_thread_index () == 0); if (a->flags & ETHERNET_ARP_ARGS_REMOVE) vnet_arp_unset_ip4_over_ethernet_internal (vm, a); diff --git a/src/vnet/ethernet/interface.c b/src/vnet/ethernet/interface.c index 9894e3c8..335e3f9f 100644 --- a/src/vnet/ethernet/interface.c +++ b/src/vnet/ethernet/interface.c @@ -362,7 +362,7 @@ simulated_ethernet_interface_tx (vlib_main_t * vm, u32 next_index = VNET_SIMULATED_ETHERNET_TX_NEXT_ETHERNET_INPUT; u32 i, next_node_index, bvi_flag, sw_if_index; u32 n_pkts = 0, n_bytes = 0; - u32 cpu_index = vm->cpu_index; + u32 thread_index = vm->thread_index; vnet_main_t *vnm = vnet_get_main (); vnet_interface_main_t *im = &vnm->interface_main; vlib_node_main_t *nm = &vm->node_main; @@ -420,8 +420,9 @@ simulated_ethernet_interface_tx (vlib_main_t * vm, /* increment TX interface stat */ vlib_increment_combined_counter (im->combined_sw_if_counters + - VNET_INTERFACE_COUNTER_TX, cpu_index, - sw_if_index, n_pkts, n_bytes); + VNET_INTERFACE_COUNTER_TX, + thread_index, sw_if_index, n_pkts, + n_bytes); } return n_left_from; diff --git a/src/vnet/ethernet/node.c b/src/vnet/ethernet/node.c index b699e381..f7787ed2 100755 --- a/src/vnet/ethernet/node.c +++ b/src/vnet/ethernet/node.c @@ -291,7 +291,7 @@ ethernet_input_inline (vlib_main_t * vm, vlib_node_runtime_t *error_node; u32 n_left_from, next_index, *from, *to_next; u32 stats_sw_if_index, stats_n_packets, stats_n_bytes; - u32 cpu_index = os_get_cpu_number (); + u32 thread_index = vlib_get_thread_index (); u32 cached_sw_if_index = ~0; u32 cached_is_l2 = 0; /* shut up gcc */ vnet_hw_interface_t *hi = NULL; /* used for main interface only */ @@ -510,7 +510,7 @@ ethernet_input_inline (vlib_main_t * vm, interface_main.combined_sw_if_counters + VNET_INTERFACE_COUNTER_RX, - cpu_index, + thread_index, new_sw_if_index0, 1, len0); if (new_sw_if_index1 != old_sw_if_index1 @@ -519,7 +519,7 @@ ethernet_input_inline (vlib_main_t * vm, interface_main.combined_sw_if_counters + VNET_INTERFACE_COUNTER_RX, - cpu_index, + thread_index, new_sw_if_index1, 1, len1); @@ -530,7 +530,7 @@ ethernet_input_inline (vlib_main_t * vm, vlib_increment_combined_counter (vnm->interface_main.combined_sw_if_counters + VNET_INTERFACE_COUNTER_RX, - cpu_index, + thread_index, stats_sw_if_index, stats_n_packets, stats_n_bytes); stats_n_packets = stats_n_bytes = 0; @@ -696,13 +696,13 @@ ethernet_input_inline (vlib_main_t * vm, vlib_increment_combined_counter (vnm->interface_main.combined_sw_if_counters + VNET_INTERFACE_COUNTER_RX, - cpu_index, new_sw_if_index0, 1, len0); + thread_index, new_sw_if_index0, 1, len0); if (stats_n_packets > 0) { vlib_increment_combined_counter (vnm->interface_main.combined_sw_if_counters + VNET_INTERFACE_COUNTER_RX, - cpu_index, + thread_index, stats_sw_if_index, stats_n_packets, stats_n_bytes); stats_n_packets = stats_n_bytes = 0; } @@ -734,7 +734,7 @@ ethernet_input_inline (vlib_main_t * vm, vlib_increment_combined_counter (vnm->interface_main.combined_sw_if_counters + VNET_INTERFACE_COUNTER_RX, - cpu_index, stats_sw_if_index, stats_n_packets, stats_n_bytes); + thread_index, stats_sw_if_index, stats_n_packets, stats_n_bytes); node->runtime_data[0] = stats_sw_if_index; } diff --git a/src/vnet/gre/node.c b/src/vnet/gre/node.c index 2683586e..acf15f24 100644 --- a/src/vnet/gre/node.c +++ b/src/vnet/gre/node.c @@ -75,7 +75,7 @@ gre_input (vlib_main_t * vm, u64 cached_tunnel_key6[4]; u32 cached_tunnel_sw_if_index = 0, tunnel_sw_if_index = 0; - u32 cpu_index = os_get_cpu_number(); + u32 thread_index = vlib_get_thread_index(); u32 len; vnet_interface_main_t *im = &gm->vnet_main->interface_main; @@ -257,7 +257,7 @@ gre_input (vlib_main_t * vm, len = vlib_buffer_length_in_chain (vm, b0); vlib_increment_combined_counter (im->combined_sw_if_counters + VNET_INTERFACE_COUNTER_RX, - cpu_index, + thread_index, tunnel_sw_if_index, 1 /* packets */, len /* bytes */); @@ -324,7 +324,7 @@ drop0: len = vlib_buffer_length_in_chain (vm, b1); vlib_increment_combined_counter (im->combined_sw_if_counters + VNET_INTERFACE_COUNTER_RX, - cpu_index, + thread_index, tunnel_sw_if_index, 1 /* packets */, len /* bytes */); @@ -502,7 +502,7 @@ drop1: len = vlib_buffer_length_in_chain (vm, b0); vlib_increment_combined_counter (im->combined_sw_if_counters + VNET_INTERFACE_COUNTER_RX, - cpu_index, + thread_index, tunnel_sw_if_index, 1 /* packets */, len /* bytes */); diff --git a/src/vnet/interface.h b/src/vnet/interface.h index a1ea2d61..08f08b10 100644 --- a/src/vnet/interface.h +++ b/src/vnet/interface.h @@ -468,7 +468,7 @@ typedef struct vnet_hw_interface_t u32 input_node_index; /* input node cpu index by queue */ - u32 *input_node_cpu_index_by_queue; + u32 *input_node_thread_index_by_queue; } vnet_hw_interface_t; diff --git a/src/vnet/interface_output.c b/src/vnet/interface_output.c index 03f2cdca..663dc309 100644 --- a/src/vnet/interface_output.c +++ b/src/vnet/interface_output.c @@ -196,7 +196,7 @@ slow_path (vlib_main_t * vm, */ static_always_inline void incr_output_stats (vnet_main_t * vnm, - u32 cpu_index, + u32 thread_index, u32 length, u32 sw_if_index, u32 * last_sw_if_index, u32 * n_packets, u32 * n_bytes) @@ -216,7 +216,7 @@ incr_output_stats (vnet_main_t * vnm, vlib_increment_combined_counter (im->combined_sw_if_counters + VNET_INTERFACE_COUNTER_TX, - cpu_index, + thread_index, *last_sw_if_index, *n_packets, *n_bytes); } @@ -240,7 +240,7 @@ vnet_interface_output_node_flatten (vlib_main_t * vm, u32 n_left_to_tx, *from, *from_end, *to_tx; u32 n_bytes, n_buffers, n_packets; u32 last_sw_if_index; - u32 cpu_index = vm->cpu_index; + u32 thread_index = vm->thread_index; n_buffers = frame->n_vectors; @@ -266,7 +266,7 @@ vnet_interface_output_node_flatten (vlib_main_t * vm, cm = vec_elt_at_index (vnm->interface_main.sw_if_counters, VNET_INTERFACE_COUNTER_TX_ERROR); - vlib_increment_simple_counter (cm, cpu_index, + vlib_increment_simple_counter (cm, thread_index, rt->sw_if_index, n_buffers); return vlib_error_drop_buffers (vm, node, from, /* buffer stride */ 1, @@ -341,18 +341,18 @@ vnet_interface_output_node_flatten (vlib_main_t * vm, from += 1; to_tx += n_buffers; n_left_to_tx -= n_buffers; - incr_output_stats (vnm, cpu_index, n_slow_bytes, + incr_output_stats (vnm, thread_index, n_slow_bytes, vnet_buffer (b)->sw_if_index[VLIB_TX], &last_sw_if_index, &n_packets, &n_bytes); } } else { - incr_output_stats (vnm, cpu_index, + incr_output_stats (vnm, thread_index, vlib_buffer_length_in_chain (vm, b0), vnet_buffer (b0)->sw_if_index[VLIB_TX], &last_sw_if_index, &n_packets, &n_bytes); - incr_output_stats (vnm, cpu_index, + incr_output_stats (vnm, thread_index, vlib_buffer_length_in_chain (vm, b0), vnet_buffer (b1)->sw_if_index[VLIB_TX], &last_sw_if_index, &n_packets, &n_bytes); @@ -396,7 +396,7 @@ vnet_interface_output_node_flatten (vlib_main_t * vm, to_tx += n_buffers; n_left_to_tx -= n_buffers; } - incr_output_stats (vnm, cpu_index, + incr_output_stats (vnm, thread_index, vlib_buffer_length_in_chain (vm, b0), vnet_buffer (b0)->sw_if_index[VLIB_TX], &last_sw_if_index, &n_packets, &n_bytes); @@ -408,7 +408,7 @@ vnet_interface_output_node_flatten (vlib_main_t * vm, } /* Final update of interface stats. */ - incr_output_stats (vnm, cpu_index, 0, ~0, /* ~0 will flush stats */ + incr_output_stats (vnm, thread_index, 0, ~0, /* ~0 will flush stats */ &last_sw_if_index, &n_packets, &n_bytes); return n_buffers; @@ -428,7 +428,7 @@ vnet_interface_output_node (vlib_main_t * vm, u32 n_left_to_tx, *from, *from_end, *to_tx; u32 n_bytes, n_buffers, n_packets; u32 n_bytes_b0, n_bytes_b1, n_bytes_b2, n_bytes_b3; - u32 cpu_index = vm->cpu_index; + u32 thread_index = vm->thread_index; vnet_interface_main_t *im = &vnm->interface_main; u32 next_index = VNET_INTERFACE_OUTPUT_NEXT_TX; u32 current_config_index = ~0; @@ -458,7 +458,7 @@ vnet_interface_output_node (vlib_main_t * vm, cm = vec_elt_at_index (vnm->interface_main.sw_if_counters, VNET_INTERFACE_COUNTER_TX_ERROR); - vlib_increment_simple_counter (cm, cpu_index, + vlib_increment_simple_counter (cm, thread_index, rt->sw_if_index, n_buffers); return vlib_error_drop_buffers (vm, node, from, @@ -558,7 +558,7 @@ vnet_interface_output_node (vlib_main_t * vm, { vlib_increment_combined_counter (im->combined_sw_if_counters + VNET_INTERFACE_COUNTER_TX, - cpu_index, tx_swif0, 1, + thread_index, tx_swif0, 1, n_bytes_b0); } @@ -567,7 +567,7 @@ vnet_interface_output_node (vlib_main_t * vm, vlib_increment_combined_counter (im->combined_sw_if_counters + VNET_INTERFACE_COUNTER_TX, - cpu_index, tx_swif1, 1, + thread_index, tx_swif1, 1, n_bytes_b1); } @@ -576,7 +576,7 @@ vnet_interface_output_node (vlib_main_t * vm, vlib_increment_combined_counter (im->combined_sw_if_counters + VNET_INTERFACE_COUNTER_TX, - cpu_index, tx_swif2, 1, + thread_index, tx_swif2, 1, n_bytes_b2); } if (PREDICT_FALSE (tx_swif3 != rt->sw_if_index)) @@ -584,7 +584,7 @@ vnet_interface_output_node (vlib_main_t * vm, vlib_increment_combined_counter (im->combined_sw_if_counters + VNET_INTERFACE_COUNTER_TX, - cpu_index, tx_swif3, 1, + thread_index, tx_swif3, 1, n_bytes_b3); } } @@ -623,7 +623,7 @@ vnet_interface_output_node (vlib_main_t * vm, vlib_increment_combined_counter (im->combined_sw_if_counters + VNET_INTERFACE_COUNTER_TX, - cpu_index, tx_swif0, 1, + thread_index, tx_swif0, 1, n_bytes_b0); } } @@ -634,7 +634,7 @@ vnet_interface_output_node (vlib_main_t * vm, /* Update main interface stats. */ vlib_increment_combined_counter (im->combined_sw_if_counters + VNET_INTERFACE_COUNTER_TX, - cpu_index, + thread_index, rt->sw_if_index, n_packets, n_bytes); return n_buffers; } @@ -893,7 +893,7 @@ process_drop_punt (vlib_main_t * vm, u32 current_sw_if_index, n_errors_current_sw_if_index; u64 current_counter; vlib_simple_counter_main_t *cm; - u32 cpu_index = vm->cpu_index; + u32 thread_index = vm->thread_index; static vlib_error_t memory[VNET_ERROR_N_DISPOSITION]; static char memory_init[VNET_ERROR_N_DISPOSITION]; @@ -965,19 +965,19 @@ process_drop_punt (vlib_main_t * vm, current_counter -= 2; n_errors_current_sw_if_index -= 2; - vlib_increment_simple_counter (cm, cpu_index, sw_if_index0, 1); - vlib_increment_simple_counter (cm, cpu_index, sw_if_index1, 1); + vlib_increment_simple_counter (cm, thread_index, sw_if_index0, 1); + vlib_increment_simple_counter (cm, thread_index, sw_if_index1, 1); /* Increment super-interface drop/punt counters for sub-interfaces. */ sw_if0 = vnet_get_sw_interface (vnm, sw_if_index0); vlib_increment_simple_counter - (cm, cpu_index, sw_if0->sup_sw_if_index, + (cm, thread_index, sw_if0->sup_sw_if_index, sw_if0->sup_sw_if_index != sw_if_index0); sw_if1 = vnet_get_sw_interface (vnm, sw_if_index1); vlib_increment_simple_counter - (cm, cpu_index, sw_if1->sup_sw_if_index, + (cm, thread_index, sw_if1->sup_sw_if_index, sw_if1->sup_sw_if_index != sw_if_index1); em->counters[current_counter_index] = current_counter; @@ -1013,11 +1013,12 @@ process_drop_punt (vlib_main_t * vm, sw_if_index0 = vnet_buffer (b0)->sw_if_index[VLIB_RX]; /* Increment drop/punt counters. */ - vlib_increment_simple_counter (cm, cpu_index, sw_if_index0, 1); + vlib_increment_simple_counter (cm, thread_index, sw_if_index0, 1); /* Increment super-interface drop/punt counters for sub-interfaces. */ sw_if0 = vnet_get_sw_interface (vnm, sw_if_index0); - vlib_increment_simple_counter (cm, cpu_index, sw_if0->sup_sw_if_index, + vlib_increment_simple_counter (cm, thread_index, + sw_if0->sup_sw_if_index, sw_if0->sup_sw_if_index != sw_if_index0); if (PREDICT_FALSE (e0 != current_error)) @@ -1041,12 +1042,12 @@ process_drop_punt (vlib_main_t * vm, { vnet_sw_interface_t *si; - vlib_increment_simple_counter (cm, cpu_index, current_sw_if_index, + vlib_increment_simple_counter (cm, thread_index, current_sw_if_index, n_errors_current_sw_if_index); si = vnet_get_sw_interface (vnm, current_sw_if_index); if (si->sup_sw_if_index != current_sw_if_index) - vlib_increment_simple_counter (cm, cpu_index, si->sup_sw_if_index, + vlib_increment_simple_counter (cm, thread_index, si->sup_sw_if_index, n_errors_current_sw_if_index); } diff --git a/src/vnet/ip/ip4_forward.c b/src/vnet/ip/ip4_forward.c index ee1703e7..fdfe7f63 100644 --- a/src/vnet/ip/ip4_forward.c +++ b/src/vnet/ip/ip4_forward.c @@ -75,7 +75,7 @@ ip4_lookup_inline (vlib_main_t * vm, vlib_combined_counter_main_t *cm = &load_balance_main.lbm_to_counters; u32 n_left_from, n_left_to_next, *from, *to_next; ip_lookup_next_t next; - u32 cpu_index = os_get_cpu_number (); + u32 thread_index = vlib_get_thread_index (); from = vlib_frame_vector_args (frame); n_left_from = frame->n_vectors; @@ -292,19 +292,19 @@ ip4_lookup_inline (vlib_main_t * vm, vnet_buffer (p3)->ip.adj_index[VLIB_TX] = dpo3->dpoi_index; vlib_increment_combined_counter - (cm, cpu_index, lb_index0, 1, + (cm, thread_index, lb_index0, 1, vlib_buffer_length_in_chain (vm, p0) + sizeof (ethernet_header_t)); vlib_increment_combined_counter - (cm, cpu_index, lb_index1, 1, + (cm, thread_index, lb_index1, 1, vlib_buffer_length_in_chain (vm, p1) + sizeof (ethernet_header_t)); vlib_increment_combined_counter - (cm, cpu_index, lb_index2, 1, + (cm, thread_index, lb_index2, 1, vlib_buffer_length_in_chain (vm, p2) + sizeof (ethernet_header_t)); vlib_increment_combined_counter - (cm, cpu_index, lb_index3, 1, + (cm, thread_index, lb_index3, 1, vlib_buffer_length_in_chain (vm, p3) + sizeof (ethernet_header_t)); @@ -392,7 +392,7 @@ ip4_lookup_inline (vlib_main_t * vm, vnet_buffer (p0)->ip.adj_index[VLIB_TX] = dpo0->dpoi_index; vlib_increment_combined_counter - (cm, cpu_index, lbi0, 1, vlib_buffer_length_in_chain (vm, p0)); + (cm, thread_index, lbi0, 1, vlib_buffer_length_in_chain (vm, p0)); from += 1; to_next += 1; @@ -479,7 +479,7 @@ ip4_load_balance (vlib_main_t * vm, vlib_combined_counter_main_t *cm = &load_balance_main.lbm_via_counters; u32 n_left_from, n_left_to_next, *from, *to_next; ip_lookup_next_t next; - u32 cpu_index = os_get_cpu_number (); + u32 thread_index = vlib_get_thread_index (); from = vlib_frame_vector_args (frame); n_left_from = frame->n_vectors; @@ -584,9 +584,9 @@ ip4_load_balance (vlib_main_t * vm, vnet_buffer (p1)->ip.adj_index[VLIB_TX] = dpo1->dpoi_index; vlib_increment_combined_counter - (cm, cpu_index, lbi0, 1, vlib_buffer_length_in_chain (vm, p0)); + (cm, thread_index, lbi0, 1, vlib_buffer_length_in_chain (vm, p0)); vlib_increment_combined_counter - (cm, cpu_index, lbi1, 1, vlib_buffer_length_in_chain (vm, p1)); + (cm, thread_index, lbi1, 1, vlib_buffer_length_in_chain (vm, p1)); vlib_validate_buffer_enqueue_x2 (vm, node, next, to_next, n_left_to_next, @@ -639,7 +639,7 @@ ip4_load_balance (vlib_main_t * vm, vnet_buffer (p0)->ip.adj_index[VLIB_TX] = dpo0->dpoi_index; vlib_increment_combined_counter - (cm, cpu_index, lbi0, 1, vlib_buffer_length_in_chain (vm, p0)); + (cm, thread_index, lbi0, 1, vlib_buffer_length_in_chain (vm, p0)); vlib_validate_buffer_enqueue_x1 (vm, node, next, to_next, n_left_to_next, @@ -2330,7 +2330,7 @@ ip4_rewrite_inline (vlib_main_t * vm, n_left_from = frame->n_vectors; next_index = node->cached_next_index; - u32 cpu_index = os_get_cpu_number (); + u32 thread_index = vlib_get_thread_index (); while (n_left_from > 0) { @@ -2379,9 +2379,9 @@ ip4_rewrite_inline (vlib_main_t * vm, if (do_counters) { vlib_prefetch_combined_counter (&adjacency_counters, - cpu_index, adj_index0); + thread_index, adj_index0); vlib_prefetch_combined_counter (&adjacency_counters, - cpu_index, adj_index1); + thread_index, adj_index1); } ip0 = vlib_buffer_get_current (p0); @@ -2527,13 +2527,13 @@ ip4_rewrite_inline (vlib_main_t * vm, { vlib_increment_combined_counter (&adjacency_counters, - cpu_index, + thread_index, adj_index0, 1, vlib_buffer_length_in_chain (vm, p0) + rw_len0); vlib_increment_combined_counter (&adjacency_counters, - cpu_index, + thread_index, adj_index1, 1, vlib_buffer_length_in_chain (vm, p1) + rw_len1); } @@ -2618,7 +2618,7 @@ ip4_rewrite_inline (vlib_main_t * vm, if (do_counters) vlib_prefetch_combined_counter (&adjacency_counters, - cpu_index, adj_index0); + thread_index, adj_index0); /* Guess we are only writing on simple Ethernet header. */ vnet_rewrite_one_header (adj0[0], ip0, sizeof (ethernet_header_t)); @@ -2637,7 +2637,7 @@ ip4_rewrite_inline (vlib_main_t * vm, if (do_counters) vlib_increment_combined_counter (&adjacency_counters, - cpu_index, adj_index0, 1, + thread_index, adj_index0, 1, vlib_buffer_length_in_chain (vm, p0) + rw_len0); /* Check MTU of outgoing interface. */ diff --git a/src/vnet/ip/ip4_input.c b/src/vnet/ip/ip4_input.c index ba200a9f..3b08f4b0 100644 --- a/src/vnet/ip/ip4_input.c +++ b/src/vnet/ip/ip4_input.c @@ -85,7 +85,7 @@ ip4_input_inline (vlib_main_t * vm, vlib_node_runtime_t *error_node = vlib_node_get_runtime (vm, ip4_input_node.index); vlib_simple_counter_main_t *cm; - u32 cpu_index = os_get_cpu_number (); + u32 thread_index = vlib_get_thread_index (); from = vlib_frame_vector_args (frame); n_left_from = frame->n_vectors; @@ -178,8 +178,8 @@ ip4_input_inline (vlib_main_t * vm, vnet_feature_arc_start (arc0, sw_if_index0, &next0, p0); vnet_feature_arc_start (arc1, sw_if_index1, &next1, p1); - vlib_increment_simple_counter (cm, cpu_index, sw_if_index0, 1); - vlib_increment_simple_counter (cm, cpu_index, sw_if_index1, 1); + vlib_increment_simple_counter (cm, thread_index, sw_if_index0, 1); + vlib_increment_simple_counter (cm, thread_index, sw_if_index1, 1); /* Punt packets with options or wrong version. */ if (PREDICT_FALSE (ip0->ip_version_and_header_length != 0x45)) @@ -299,7 +299,7 @@ ip4_input_inline (vlib_main_t * vm, vnet_buffer (p0)->ip.adj_index[VLIB_RX] = ~0; vnet_feature_arc_start (arc0, sw_if_index0, &next0, p0); - vlib_increment_simple_counter (cm, cpu_index, sw_if_index0, 1); + vlib_increment_simple_counter (cm, thread_index, sw_if_index0, 1); /* Punt packets with options or wrong version. */ if (PREDICT_FALSE (ip0->ip_version_and_header_length != 0x45)) diff --git a/src/vnet/ip/ip6_forward.c b/src/vnet/ip/ip6_forward.c index c120f12c..c2fc4f87 100644 --- a/src/vnet/ip/ip6_forward.c +++ b/src/vnet/ip/ip6_forward.c @@ -74,7 +74,7 @@ ip6_lookup_inline (vlib_main_t * vm, vlib_combined_counter_main_t *cm = &load_balance_main.lbm_to_counters; u32 n_left_from, n_left_to_next, *from, *to_next; ip_lookup_next_t next; - u32 cpu_index = os_get_cpu_number (); + u32 thread_index = vlib_get_thread_index (); from = vlib_frame_vector_args (frame); n_left_from = frame->n_vectors; @@ -185,9 +185,9 @@ ip6_lookup_inline (vlib_main_t * vm, vnet_buffer (p1)->ip.adj_index[VLIB_TX] = dpo1->dpoi_index; vlib_increment_combined_counter - (cm, cpu_index, lbi0, 1, vlib_buffer_length_in_chain (vm, p0)); + (cm, thread_index, lbi0, 1, vlib_buffer_length_in_chain (vm, p0)); vlib_increment_combined_counter - (cm, cpu_index, lbi1, 1, vlib_buffer_length_in_chain (vm, p1)); + (cm, thread_index, lbi1, 1, vlib_buffer_length_in_chain (vm, p1)); from += 2; to_next += 2; @@ -291,7 +291,7 @@ ip6_lookup_inline (vlib_main_t * vm, vnet_buffer (p0)->ip.adj_index[VLIB_TX] = dpo0->dpoi_index; vlib_increment_combined_counter - (cm, cpu_index, lbi0, 1, vlib_buffer_length_in_chain (vm, p0)); + (cm, thread_index, lbi0, 1, vlib_buffer_length_in_chain (vm, p0)); from += 1; to_next += 1; @@ -703,7 +703,7 @@ ip6_load_balance (vlib_main_t * vm, vlib_combined_counter_main_t *cm = &load_balance_main.lbm_via_counters; u32 n_left_from, n_left_to_next, *from, *to_next; ip_lookup_next_t next; - u32 cpu_index = os_get_cpu_number (); + u32 thread_index = vlib_get_thread_index (); ip6_main_t *im = &ip6_main; from = vlib_frame_vector_args (frame); @@ -824,9 +824,9 @@ ip6_load_balance (vlib_main_t * vm, vnet_buffer (p1)->ip.adj_index[VLIB_TX] = dpo1->dpoi_index; vlib_increment_combined_counter - (cm, cpu_index, lbi0, 1, vlib_buffer_length_in_chain (vm, p0)); + (cm, thread_index, lbi0, 1, vlib_buffer_length_in_chain (vm, p0)); vlib_increment_combined_counter - (cm, cpu_index, lbi1, 1, vlib_buffer_length_in_chain (vm, p1)); + (cm, thread_index, lbi1, 1, vlib_buffer_length_in_chain (vm, p1)); vlib_validate_buffer_enqueue_x2 (vm, node, next, to_next, n_left_to_next, @@ -886,7 +886,7 @@ ip6_load_balance (vlib_main_t * vm, } vlib_increment_combined_counter - (cm, cpu_index, lbi0, 1, vlib_buffer_length_in_chain (vm, p0)); + (cm, thread_index, lbi0, 1, vlib_buffer_length_in_chain (vm, p0)); vlib_validate_buffer_enqueue_x1 (vm, node, next, to_next, n_left_to_next, @@ -1897,7 +1897,7 @@ ip6_rewrite_inline (vlib_main_t * vm, n_left_from = frame->n_vectors; next_index = node->cached_next_index; - u32 cpu_index = os_get_cpu_number (); + u32 thread_index = vlib_get_thread_index (); while (n_left_from > 0) { @@ -2019,11 +2019,11 @@ ip6_rewrite_inline (vlib_main_t * vm, { vlib_increment_combined_counter (&adjacency_counters, - cpu_index, adj_index0, 1, + thread_index, adj_index0, 1, vlib_buffer_length_in_chain (vm, p0) + rw_len0); vlib_increment_combined_counter (&adjacency_counters, - cpu_index, adj_index1, 1, + thread_index, adj_index1, 1, vlib_buffer_length_in_chain (vm, p1) + rw_len1); } @@ -2156,7 +2156,7 @@ ip6_rewrite_inline (vlib_main_t * vm, { vlib_increment_combined_counter (&adjacency_counters, - cpu_index, adj_index0, 1, + thread_index, adj_index0, 1, vlib_buffer_length_in_chain (vm, p0) + rw_len0); } diff --git a/src/vnet/ip/ip6_input.c b/src/vnet/ip/ip6_input.c index 20306088..ffdc4727 100644 --- a/src/vnet/ip/ip6_input.c +++ b/src/vnet/ip/ip6_input.c @@ -82,7 +82,7 @@ ip6_input (vlib_main_t * vm, vlib_node_runtime_t * node, vlib_frame_t * frame) vlib_node_runtime_t *error_node = vlib_node_get_runtime (vm, ip6_input_node.index); vlib_simple_counter_main_t *cm; - u32 cpu_index = os_get_cpu_number (); + u32 thread_index = vlib_get_thread_index (); from = vlib_frame_vector_args (frame); n_left_from = frame->n_vectors; @@ -171,8 +171,8 @@ ip6_input (vlib_main_t * vm, vlib_node_runtime_t * node, vlib_frame_t * frame) vnet_feature_arc_start (arc0, sw_if_index0, &next0, p0); vnet_feature_arc_start (arc1, sw_if_index1, &next1, p1); - vlib_increment_simple_counter (cm, cpu_index, sw_if_index0, 1); - vlib_increment_simple_counter (cm, cpu_index, sw_if_index1, 1); + vlib_increment_simple_counter (cm, thread_index, sw_if_index0, 1); + vlib_increment_simple_counter (cm, thread_index, sw_if_index1, 1); error0 = error1 = IP6_ERROR_NONE; @@ -270,7 +270,7 @@ ip6_input (vlib_main_t * vm, vlib_node_runtime_t * node, vlib_frame_t * frame) vnet_buffer (p0)->ip.adj_index[VLIB_RX] = ~0; vnet_feature_arc_start (arc0, sw_if_index0, &next0, p0); - vlib_increment_simple_counter (cm, cpu_index, sw_if_index0, 1); + vlib_increment_simple_counter (cm, thread_index, sw_if_index0, 1); error0 = IP6_ERROR_NONE; /* Version != 6? Drop it. */ diff --git a/src/vnet/ip/ip6_neighbor.c b/src/vnet/ip/ip6_neighbor.c index 5d1fb6f8..2af546df 100644 --- a/src/vnet/ip/ip6_neighbor.c +++ b/src/vnet/ip/ip6_neighbor.c @@ -581,7 +581,7 @@ vnet_set_ip6_ethernet_neighbor (vlib_main_t * vm, u32 next_index; pending_resolution_t *pr, *mc; - if (os_get_cpu_number ()) + if (vlib_get_thread_index ()) { set_unset_ip6_neighbor_rpc (vm, sw_if_index, a, link_layer_address, 1 /* set new neighbor */ , is_static, @@ -722,7 +722,7 @@ vnet_unset_ip6_ethernet_neighbor (vlib_main_t * vm, uword *p; int rv = 0; - if (os_get_cpu_number ()) + if (vlib_get_thread_index ()) { set_unset_ip6_neighbor_rpc (vm, sw_if_index, a, link_layer_address, 0 /* unset */ , 0, 0); diff --git a/src/vnet/ipsec/esp.h b/src/vnet/ipsec/esp.h index 50cac806..799003b9 100644 --- a/src/vnet/ipsec/esp.h +++ b/src/vnet/ipsec/esp.h @@ -282,8 +282,8 @@ hmac_calc (ipsec_integ_alg_t alg, u8 * data, int data_len, u8 * signature, u8 use_esn, u32 seq_hi) { esp_main_t *em = &esp_main; - u32 cpu_index = os_get_cpu_number (); - HMAC_CTX *ctx = &(em->per_thread_data[cpu_index].hmac_ctx); + u32 thread_index = vlib_get_thread_index (); + HMAC_CTX *ctx = &(em->per_thread_data[thread_index].hmac_ctx); const EVP_MD *md = NULL; unsigned int len; @@ -292,10 +292,10 @@ hmac_calc (ipsec_integ_alg_t alg, if (PREDICT_FALSE (em->esp_integ_algs[alg].md == 0)) return 0; - if (PREDICT_FALSE (alg != em->per_thread_data[cpu_index].last_integ_alg)) + if (PREDICT_FALSE (alg != em->per_thread_data[thread_index].last_integ_alg)) { md = em->esp_integ_algs[alg].md; - em->per_thread_data[cpu_index].last_integ_alg = alg; + em->per_thread_data[thread_index].last_integ_alg = alg; } HMAC_Init (ctx, key, key_len, md); diff --git a/src/vnet/ipsec/esp_decrypt.c b/src/vnet/ipsec/esp_decrypt.c index 7289b260..925d2b45 100644 --- a/src/vnet/ipsec/esp_decrypt.c +++ b/src/vnet/ipsec/esp_decrypt.c @@ -85,8 +85,8 @@ esp_decrypt_aes_cbc (ipsec_crypto_alg_t alg, u8 * in, u8 * out, size_t in_len, u8 * key, u8 * iv) { esp_main_t *em = &esp_main; - u32 cpu_index = os_get_cpu_number (); - EVP_CIPHER_CTX *ctx = &(em->per_thread_data[cpu_index].decrypt_ctx); + u32 thread_index = vlib_get_thread_index (); + EVP_CIPHER_CTX *ctx = &(em->per_thread_data[thread_index].decrypt_ctx); const EVP_CIPHER *cipher = NULL; int out_len; @@ -95,10 +95,11 @@ esp_decrypt_aes_cbc (ipsec_crypto_alg_t alg, if (PREDICT_FALSE (em->esp_crypto_algs[alg].type == 0)) return; - if (PREDICT_FALSE (alg != em->per_thread_data[cpu_index].last_decrypt_alg)) + if (PREDICT_FALSE + (alg != em->per_thread_data[thread_index].last_decrypt_alg)) { cipher = em->esp_crypto_algs[alg].type; - em->per_thread_data[cpu_index].last_decrypt_alg = alg; + em->per_thread_data[thread_index].last_decrypt_alg = alg; } EVP_DecryptInit_ex (ctx, cipher, NULL, key, iv); @@ -117,11 +118,11 @@ esp_decrypt_node_fn (vlib_main_t * vm, u32 *recycle = 0; from = vlib_frame_vector_args (from_frame); n_left_from = from_frame->n_vectors; - u32 cpu_index = os_get_cpu_number (); + u32 thread_index = vlib_get_thread_index (); ipsec_alloc_empty_buffers (vm, im); - u32 *empty_buffers = im->empty_buffers[cpu_index]; + u32 *empty_buffers = im->empty_buffers[thread_index]; if (PREDICT_FALSE (vec_len (empty_buffers) < n_left_from)) { diff --git a/src/vnet/ipsec/esp_encrypt.c b/src/vnet/ipsec/esp_encrypt.c index 44ae2297..b2bc4e0b 100644 --- a/src/vnet/ipsec/esp_encrypt.c +++ b/src/vnet/ipsec/esp_encrypt.c @@ -88,8 +88,8 @@ esp_encrypt_aes_cbc (ipsec_crypto_alg_t alg, u8 * in, u8 * out, size_t in_len, u8 * key, u8 * iv) { esp_main_t *em = &esp_main; - u32 cpu_index = os_get_cpu_number (); - EVP_CIPHER_CTX *ctx = &(em->per_thread_data[cpu_index].encrypt_ctx); + u32 thread_index = vlib_get_thread_index (); + EVP_CIPHER_CTX *ctx = &(em->per_thread_data[thread_index].encrypt_ctx); const EVP_CIPHER *cipher = NULL; int out_len; @@ -98,10 +98,11 @@ esp_encrypt_aes_cbc (ipsec_crypto_alg_t alg, if (PREDICT_FALSE (em->esp_crypto_algs[alg].type == IPSEC_CRYPTO_ALG_NONE)) return; - if (PREDICT_FALSE (alg != em->per_thread_data[cpu_index].last_encrypt_alg)) + if (PREDICT_FALSE + (alg != em->per_thread_data[thread_index].last_encrypt_alg)) { cipher = em->esp_crypto_algs[alg].type; - em->per_thread_data[cpu_index].last_encrypt_alg = alg; + em->per_thread_data[thread_index].last_encrypt_alg = alg; } EVP_EncryptInit_ex (ctx, cipher, NULL, key, iv); @@ -119,11 +120,11 @@ esp_encrypt_node_fn (vlib_main_t * vm, n_left_from = from_frame->n_vectors; ipsec_main_t *im = &ipsec_main; u32 *recycle = 0; - u32 cpu_index = os_get_cpu_number (); + u32 thread_index = vlib_get_thread_index (); ipsec_alloc_empty_buffers (vm, im); - u32 *empty_buffers = im->empty_buffers[cpu_index]; + u32 *empty_buffers = im->empty_buffers[thread_index]; if (PREDICT_FALSE (vec_len (empty_buffers) < n_left_from)) { diff --git a/src/vnet/ipsec/ikev2.c b/src/vnet/ipsec/ikev2.c index 2c1074d8..3f9978a7 100644 --- a/src/vnet/ipsec/ikev2.c +++ b/src/vnet/ipsec/ikev2.c @@ -303,16 +303,16 @@ static void ikev2_delete_sa (ikev2_sa_t * sa) { ikev2_main_t *km = &ikev2_main; - u32 cpu_index = os_get_cpu_number (); + u32 thread_index = vlib_get_thread_index (); uword *p; ikev2_sa_free_all_vec (sa); - p = hash_get (km->per_thread_data[cpu_index].sa_by_rspi, sa->rspi); + p = hash_get (km->per_thread_data[thread_index].sa_by_rspi, sa->rspi); if (p) { - hash_unset (km->per_thread_data[cpu_index].sa_by_rspi, sa->rspi); - pool_put (km->per_thread_data[cpu_index].sas, sa); + hash_unset (km->per_thread_data[thread_index].sa_by_rspi, sa->rspi); + pool_put (km->per_thread_data[thread_index].sas, sa); } } @@ -776,29 +776,31 @@ ikev2_initial_contact_cleanup (ikev2_sa_t * sa) ikev2_sa_t *tmp; u32 i, *delete = 0; ikev2_child_sa_t *c; - u32 cpu_index = os_get_cpu_number (); + u32 thread_index = vlib_get_thread_index (); if (!sa->initial_contact) return; /* find old IKE SAs with the same authenticated identity */ /* *INDENT-OFF* */ - pool_foreach (tmp, km->per_thread_data[cpu_index].sas, ({ + pool_foreach (tmp, km->per_thread_data[thread_index].sas, ({ if (tmp->i_id.type != sa->i_id.type || vec_len(tmp->i_id.data) != vec_len(sa->i_id.data) || memcmp(sa->i_id.data, tmp->i_id.data, vec_len(sa->i_id.data))) continue; if (sa->rspi != tmp->rspi) - vec_add1(delete, tmp - km->per_thread_data[cpu_index].sas); + vec_add1(delete, tmp - km->per_thread_data[thread_index].sas); })); /* *INDENT-ON* */ for (i = 0; i < vec_len (delete); i++) { - tmp = pool_elt_at_index (km->per_thread_data[cpu_index].sas, delete[i]); - vec_foreach (c, tmp->childs) - ikev2_delete_tunnel_interface (km->vnet_main, tmp, c); + tmp = + pool_elt_at_index (km->per_thread_data[thread_index].sas, delete[i]); + vec_foreach (c, + tmp->childs) ikev2_delete_tunnel_interface (km->vnet_main, + tmp, c); ikev2_delete_sa (tmp); } @@ -1922,10 +1924,10 @@ ikev2_retransmit_sa_init (ike_header_t * ike, { ikev2_main_t *km = &ikev2_main; ikev2_sa_t *sa; - u32 cpu_index = os_get_cpu_number (); + u32 thread_index = vlib_get_thread_index (); /* *INDENT-OFF* */ - pool_foreach (sa, km->per_thread_data[cpu_index].sas, ({ + pool_foreach (sa, km->per_thread_data[thread_index].sas, ({ if (sa->ispi == clib_net_to_host_u64(ike->ispi) && sa->iaddr.as_u32 == iaddr.as_u32 && sa->raddr.as_u32 == raddr.as_u32) @@ -2036,7 +2038,7 @@ ikev2_node_fn (vlib_main_t * vm, u32 n_left_from, *from, *to_next; ikev2_next_t next_index; ikev2_main_t *km = &ikev2_main; - u32 cpu_index = os_get_cpu_number (); + u32 thread_index = vlib_get_thread_index (); from = vlib_frame_vector_args (frame); n_left_from = frame->n_vectors; @@ -2134,11 +2136,14 @@ ikev2_node_fn (vlib_main_t * vm, if (sa0->state == IKEV2_STATE_SA_INIT) { /* add SA to the pool */ - pool_get (km->per_thread_data[cpu_index].sas, sa0); + pool_get (km->per_thread_data[thread_index].sas, + sa0); clib_memcpy (sa0, &sa, sizeof (*sa0)); - hash_set (km->per_thread_data[cpu_index].sa_by_rspi, + hash_set (km-> + per_thread_data[thread_index].sa_by_rspi, sa0->rspi, - sa0 - km->per_thread_data[cpu_index].sas); + sa0 - + km->per_thread_data[thread_index].sas); } else { @@ -2169,11 +2174,11 @@ ikev2_node_fn (vlib_main_t * vm, if (sa0->state == IKEV2_STATE_SA_INIT) { /* add SA to the pool */ - pool_get (km->per_thread_data[cpu_index].sas, sa0); + pool_get (km->per_thread_data[thread_index].sas, sa0); clib_memcpy (sa0, &sa, sizeof (*sa0)); - hash_set (km->per_thread_data[cpu_index].sa_by_rspi, + hash_set (km->per_thread_data[thread_index].sa_by_rspi, sa0->rspi, - sa0 - km->per_thread_data[cpu_index].sas); + sa0 - km->per_thread_data[thread_index].sas); } else { @@ -2184,12 +2189,13 @@ ikev2_node_fn (vlib_main_t * vm, else if (ike0->exchange == IKEV2_EXCHANGE_IKE_AUTH) { uword *p; - p = hash_get (km->per_thread_data[cpu_index].sa_by_rspi, + p = hash_get (km->per_thread_data[thread_index].sa_by_rspi, clib_net_to_host_u64 (ike0->rspi)); if (p) { - sa0 = pool_elt_at_index (km->per_thread_data[cpu_index].sas, - p[0]); + sa0 = + pool_elt_at_index (km->per_thread_data[thread_index].sas, + p[0]); r = ikev2_retransmit_resp (sa0, ike0); if (r == 1) @@ -2240,12 +2246,13 @@ ikev2_node_fn (vlib_main_t * vm, else if (ike0->exchange == IKEV2_EXCHANGE_INFORMATIONAL) { uword *p; - p = hash_get (km->per_thread_data[cpu_index].sa_by_rspi, + p = hash_get (km->per_thread_data[thread_index].sa_by_rspi, clib_net_to_host_u64 (ike0->rspi)); if (p) { - sa0 = pool_elt_at_index (km->per_thread_data[cpu_index].sas, - p[0]); + sa0 = + pool_elt_at_index (km->per_thread_data[thread_index].sas, + p[0]); r = ikev2_retransmit_resp (sa0, ike0); if (r == 1) @@ -2305,12 +2312,13 @@ ikev2_node_fn (vlib_main_t * vm, else if (ike0->exchange == IKEV2_EXCHANGE_CREATE_CHILD_SA) { uword *p; - p = hash_get (km->per_thread_data[cpu_index].sa_by_rspi, + p = hash_get (km->per_thread_data[thread_index].sa_by_rspi, clib_net_to_host_u64 (ike0->rspi)); if (p) { - sa0 = pool_elt_at_index (km->per_thread_data[cpu_index].sas, - p[0]); + sa0 = + pool_elt_at_index (km->per_thread_data[thread_index].sas, + p[0]); r = ikev2_retransmit_resp (sa0, ike0); if (r == 1) diff --git a/src/vnet/ipsec/ipsec.h b/src/vnet/ipsec/ipsec.h index 58f0f145..c884e360 100644 --- a/src/vnet/ipsec/ipsec.h +++ b/src/vnet/ipsec/ipsec.h @@ -324,21 +324,21 @@ int ipsec_set_interface_key (vnet_main_t * vnm, u32 hw_if_index, always_inline void ipsec_alloc_empty_buffers (vlib_main_t * vm, ipsec_main_t * im) { - u32 cpu_index = os_get_cpu_number (); - uword l = vec_len (im->empty_buffers[cpu_index]); + u32 thread_index = vlib_get_thread_index (); + uword l = vec_len (im->empty_buffers[thread_index]); uword n_alloc = 0; if (PREDICT_FALSE (l < VLIB_FRAME_SIZE)) { - if (!im->empty_buffers[cpu_index]) + if (!im->empty_buffers[thread_index]) { - vec_alloc (im->empty_buffers[cpu_index], 2 * VLIB_FRAME_SIZE); + vec_alloc (im->empty_buffers[thread_index], 2 * VLIB_FRAME_SIZE); } - n_alloc = vlib_buffer_alloc (vm, im->empty_buffers[cpu_index] + l, + n_alloc = vlib_buffer_alloc (vm, im->empty_buffers[thread_index] + l, 2 * VLIB_FRAME_SIZE - l); - _vec_len (im->empty_buffers[cpu_index]) = l + n_alloc; + _vec_len (im->empty_buffers[thread_index]) = l + n_alloc; } } diff --git a/src/vnet/ipsec/ipsec_if.c b/src/vnet/ipsec/ipsec_if.c index dc882004..ed124894 100644 --- a/src/vnet/ipsec/ipsec_if.c +++ b/src/vnet/ipsec/ipsec_if.c @@ -99,7 +99,7 @@ static int ipsec_add_del_tunnel_if_rpc_callback (ipsec_add_del_tunnel_args_t * a) { vnet_main_t *vnm = vnet_get_main (); - ASSERT (os_get_cpu_number () == 0); + ASSERT (vlib_get_thread_index () == 0); return ipsec_add_del_tunnel_if_internal (vnm, a); } diff --git a/src/vnet/l2/l2_bvi.h b/src/vnet/l2/l2_bvi.h index dd1130a6..e21a1616 100644 --- a/src/vnet/l2/l2_bvi.h +++ b/src/vnet/l2/l2_bvi.h @@ -97,7 +97,7 @@ l2_to_bvi (vlib_main_t * vlib_main, vlib_increment_combined_counter (vnet_main->interface_main.combined_sw_if_counters + VNET_INTERFACE_COUNTER_RX, - vlib_main->cpu_index, + vlib_main->thread_index, vnet_buffer (b0)->sw_if_index[VLIB_RX], 1, vlib_buffer_length_in_chain (vlib_main, b0)); return TO_BVI_ERR_OK; diff --git a/src/vnet/l2/l2_input.c b/src/vnet/l2/l2_input.c index 041ff38d..e5d6878a 100644 --- a/src/vnet/l2/l2_input.c +++ b/src/vnet/l2/l2_input.c @@ -117,7 +117,7 @@ typedef enum static_always_inline void classify_and_dispatch (vlib_main_t * vm, vlib_node_runtime_t * node, - u32 cpu_index, + u32 thread_index, l2input_main_t * msm, vlib_buffer_t * b0, u32 * next0) { /* @@ -237,7 +237,7 @@ l2input_node_inline (vlib_main_t * vm, u32 n_left_from, *from, *to_next; l2input_next_t next_index; l2input_main_t *msm = &l2input_main; - u32 cpu_index = os_get_cpu_number (); + u32 thread_index = vlib_get_thread_index (); from = vlib_frame_vector_args (frame); n_left_from = frame->n_vectors; /* number of packets to process */ @@ -350,10 +350,10 @@ l2input_node_inline (vlib_main_t * vm, vlib_node_increment_counter (vm, l2input_node.index, L2INPUT_ERROR_L2INPUT, 4); - classify_and_dispatch (vm, node, cpu_index, msm, b0, &next0); - classify_and_dispatch (vm, node, cpu_index, msm, b1, &next1); - classify_and_dispatch (vm, node, cpu_index, msm, b2, &next2); - classify_and_dispatch (vm, node, cpu_index, msm, b3, &next3); + classify_and_dispatch (vm, node, thread_index, msm, b0, &next0); + classify_and_dispatch (vm, node, thread_index, msm, b1, &next1); + classify_and_dispatch (vm, node, thread_index, msm, b2, &next2); + classify_and_dispatch (vm, node, thread_index, msm, b3, &next3); /* verify speculative enqueues, maybe switch current next frame */ /* if next0==next1==next_index then nothing special needs to be done */ @@ -393,7 +393,7 @@ l2input_node_inline (vlib_main_t * vm, vlib_node_increment_counter (vm, l2input_node.index, L2INPUT_ERROR_L2INPUT, 1); - classify_and_dispatch (vm, node, cpu_index, msm, b0, &next0); + classify_and_dispatch (vm, node, thread_index, msm, b0, &next0); /* verify speculative enqueue, maybe switch current next frame */ vlib_validate_buffer_enqueue_x1 (vm, node, next_index, diff --git a/src/vnet/l2/l2_output.c b/src/vnet/l2/l2_output.c index 00f22571..e17b2a16 100644 --- a/src/vnet/l2/l2_output.c +++ b/src/vnet/l2/l2_output.c @@ -643,11 +643,11 @@ l2output_create_output_node_mapping (vlib_main_t * vlib_main, vnet_main_t * vnet hw0 = vnet_get_sup_hw_interface (vnet_main, sw_if_index); - uword cpu_number; + uword thread_index; - cpu_number = os_get_cpu_number (); + thread_index = vlib_get_thread_index (); - if (cpu_number) + if (thread_index) { u32 oldflags; diff --git a/src/vnet/l2tp/decap.c b/src/vnet/l2tp/decap.c index e8986935..46104129 100644 --- a/src/vnet/l2tp/decap.c +++ b/src/vnet/l2tp/decap.c @@ -149,7 +149,7 @@ last_stage (vlib_main_t * vm, vlib_node_runtime_t * node, u32 bi) /* per-mapping byte stats include the ethernet header */ vlib_increment_combined_counter (&lm->counter_main, - os_get_cpu_number (), + vlib_get_thread_index (), counter_index, 1 /* packet_increment */ , vlib_buffer_length_in_chain (vm, b) + sizeof (ethernet_header_t)); diff --git a/src/vnet/l2tp/encap.c b/src/vnet/l2tp/encap.c index ed7a9580..dcdfde4b 100644 --- a/src/vnet/l2tp/encap.c +++ b/src/vnet/l2tp/encap.c @@ -124,7 +124,7 @@ last_stage (vlib_main_t * vm, vlib_node_runtime_t * node, u32 bi) /* per-mapping byte stats include the ethernet header */ vlib_increment_combined_counter (&lm->counter_main, - os_get_cpu_number (), + vlib_get_thread_index (), counter_index, 1 /* packet_increment */ , vlib_buffer_length_in_chain (vm, b)); diff --git a/src/vnet/l2tp/l2tp.c b/src/vnet/l2tp/l2tp.c index cb94d7e7..3dedc447 100644 --- a/src/vnet/l2tp/l2tp.c +++ b/src/vnet/l2tp/l2tp.c @@ -157,7 +157,7 @@ test_counters_command_fn (vlib_main_t * vm, u32 session_index; u32 counter_index; u32 nincr = 0; - u32 cpu_index = os_get_cpu_number (); + u32 thread_index = vlib_get_thread_index (); /* *INDENT-OFF* */ pool_foreach (session, lm->sessions, @@ -167,11 +167,11 @@ test_counters_command_fn (vlib_main_t * vm, session_index_to_counter_index (session_index, SESSION_COUNTER_USER_TO_NETWORK); vlib_increment_combined_counter (&lm->counter_main, - cpu_index, + thread_index, counter_index, 1/*pkt*/, 1111 /*bytes*/); vlib_increment_combined_counter (&lm->counter_main, - cpu_index, + thread_index, counter_index+1, 1/*pkt*/, 2222 /*bytes*/); nincr++; diff --git a/src/vnet/lisp-gpe/decap.c b/src/vnet/lisp-gpe/decap.c index d887a95f..68769710 100644 --- a/src/vnet/lisp-gpe/decap.c +++ b/src/vnet/lisp-gpe/decap.c @@ -103,7 +103,7 @@ next_index_to_iface (lisp_gpe_main_t * lgm, u32 next_index) } static_always_inline void -incr_decap_stats (vnet_main_t * vnm, u32 cpu_index, u32 length, +incr_decap_stats (vnet_main_t * vnm, u32 thread_index, u32 length, u32 sw_if_index, u32 * last_sw_if_index, u32 * n_packets, u32 * n_bytes) { @@ -122,7 +122,7 @@ incr_decap_stats (vnet_main_t * vnm, u32 cpu_index, u32 length, vlib_increment_combined_counter (im->combined_sw_if_counters + VNET_INTERFACE_COUNTER_RX, - cpu_index, *last_sw_if_index, + thread_index, *last_sw_if_index, *n_packets, *n_bytes); } *last_sw_if_index = sw_if_index; @@ -150,11 +150,11 @@ static uword lisp_gpe_input_inline (vlib_main_t * vm, vlib_node_runtime_t * node, vlib_frame_t * from_frame, u8 is_v4) { - u32 n_left_from, next_index, *from, *to_next, cpu_index; + u32 n_left_from, next_index, *from, *to_next, thread_index; u32 n_bytes = 0, n_packets = 0, last_sw_if_index = ~0, drops = 0; lisp_gpe_main_t *lgm = vnet_lisp_gpe_get_main (); - cpu_index = os_get_cpu_number (); + thread_index = vlib_get_thread_index (); from = vlib_frame_vector_args (from_frame); n_left_from = from_frame->n_vectors; @@ -267,7 +267,7 @@ lisp_gpe_input_inline (vlib_main_t * vm, vlib_node_runtime_t * node, if (si0) { - incr_decap_stats (lgm->vnet_main, cpu_index, + incr_decap_stats (lgm->vnet_main, thread_index, vlib_buffer_length_in_chain (vm, b0), si0[0], &last_sw_if_index, &n_packets, &n_bytes); vnet_buffer (b0)->sw_if_index[VLIB_RX] = si0[0]; @@ -282,7 +282,7 @@ lisp_gpe_input_inline (vlib_main_t * vm, vlib_node_runtime_t * node, if (si1) { - incr_decap_stats (lgm->vnet_main, cpu_index, + incr_decap_stats (lgm->vnet_main, thread_index, vlib_buffer_length_in_chain (vm, b1), si1[0], &last_sw_if_index, &n_packets, &n_bytes); vnet_buffer (b1)->sw_if_index[VLIB_RX] = si1[0]; @@ -397,7 +397,7 @@ lisp_gpe_input_inline (vlib_main_t * vm, vlib_node_runtime_t * node, if (si0) { - incr_decap_stats (lgm->vnet_main, cpu_index, + incr_decap_stats (lgm->vnet_main, thread_index, vlib_buffer_length_in_chain (vm, b0), si0[0], &last_sw_if_index, &n_packets, &n_bytes); vnet_buffer (b0)->sw_if_index[VLIB_RX] = si0[0]; @@ -430,7 +430,7 @@ lisp_gpe_input_inline (vlib_main_t * vm, vlib_node_runtime_t * node, } /* flush iface stats */ - incr_decap_stats (lgm->vnet_main, cpu_index, 0, ~0, &last_sw_if_index, + incr_decap_stats (lgm->vnet_main, thread_index, 0, ~0, &last_sw_if_index, &n_packets, &n_bytes); vlib_node_increment_counter (vm, lisp_gpe_ip4_input_node.index, LISP_GPE_ERROR_NO_TUNNEL, drops); diff --git a/src/vnet/lldp/lldp_input.c b/src/vnet/lldp/lldp_input.c index 762743d0..e88f6fdb 100644 --- a/src/vnet/lldp/lldp_input.c +++ b/src/vnet/lldp/lldp_input.c @@ -35,7 +35,7 @@ typedef struct static void lldp_rpc_update_peer_cb (const lldp_intf_update_t * a) { - ASSERT (os_get_cpu_number () == 0); + ASSERT (vlib_get_thread_index () == 0); lldp_intf_t *n = lldp_get_intf (&lldp_main, a->hw_if_index); if (!n) diff --git a/src/vnet/map/ip4_map.c b/src/vnet/map/ip4_map.c index 1a20d704..e39b6f14 100644 --- a/src/vnet/map/ip4_map.c +++ b/src/vnet/map/ip4_map.c @@ -248,7 +248,7 @@ ip4_map (vlib_main_t * vm, vlib_node_runtime_t * node, vlib_frame_t * frame) next_index = node->cached_next_index; map_main_t *mm = &map_main; vlib_combined_counter_main_t *cm = mm->domain_counters; - u32 cpu_index = os_get_cpu_number (); + u32 thread_index = vlib_get_thread_index (); while (n_left_from > 0) { @@ -377,7 +377,7 @@ ip4_map (vlib_main_t * vm, vlib_node_runtime_t * node, vlib_frame_t * frame) ip40) ? IP4_MAP_NEXT_IP6_REWRITE : next0; vlib_increment_combined_counter (cm + MAP_DOMAIN_COUNTER_TX, - cpu_index, + thread_index, map_domain_index0, 1, clib_net_to_host_u16 (ip6h0->payload_length) + @@ -409,7 +409,7 @@ ip4_map (vlib_main_t * vm, vlib_node_runtime_t * node, vlib_frame_t * frame) ip41) ? IP4_MAP_NEXT_IP6_REWRITE : next1; vlib_increment_combined_counter (cm + MAP_DOMAIN_COUNTER_TX, - cpu_index, + thread_index, map_domain_index1, 1, clib_net_to_host_u16 (ip6h1->payload_length) + @@ -520,7 +520,7 @@ ip4_map (vlib_main_t * vm, vlib_node_runtime_t * node, vlib_frame_t * frame) ip40) ? IP4_MAP_NEXT_IP6_REWRITE : next0; vlib_increment_combined_counter (cm + MAP_DOMAIN_COUNTER_TX, - cpu_index, + thread_index, map_domain_index0, 1, clib_net_to_host_u16 (ip6h0->payload_length) + @@ -564,7 +564,7 @@ ip4_map_reass (vlib_main_t * vm, next_index = node->cached_next_index; map_main_t *mm = &map_main; vlib_combined_counter_main_t *cm = mm->domain_counters; - u32 cpu_index = os_get_cpu_number (); + u32 thread_index = vlib_get_thread_index (); u32 *fragments_to_drop = NULL; u32 *fragments_to_loopback = NULL; @@ -694,8 +694,8 @@ ip4_map_reass (vlib_main_t * vm, { if (error0 == MAP_ERROR_NONE) vlib_increment_combined_counter (cm + MAP_DOMAIN_COUNTER_TX, - cpu_index, map_domain_index0, - 1, + thread_index, + map_domain_index0, 1, clib_net_to_host_u16 (ip60->payload_length) + 40); next0 = diff --git a/src/vnet/map/ip4_map_t.c b/src/vnet/map/ip4_map_t.c index b63d76bf..5f2bcbf9 100644 --- a/src/vnet/map/ip4_map_t.c +++ b/src/vnet/map/ip4_map_t.c @@ -477,7 +477,7 @@ ip4_map_t_icmp (vlib_main_t * vm, n_left_from = frame->n_vectors; next_index = node->cached_next_index; vlib_combined_counter_main_t *cm = map_main.domain_counters; - u32 cpu_index = os_get_cpu_number (); + u32 thread_index = vlib_get_thread_index (); while (n_left_from > 0) { @@ -520,7 +520,7 @@ ip4_map_t_icmp (vlib_main_t * vm, if (PREDICT_TRUE (error0 == MAP_ERROR_NONE)) { vlib_increment_combined_counter (cm + MAP_DOMAIN_COUNTER_TX, - cpu_index, + thread_index, vnet_buffer (p0)->map_t. map_domain_index, 1, len0); } @@ -1051,7 +1051,7 @@ ip4_map_t (vlib_main_t * vm, vlib_node_runtime_t * node, vlib_frame_t * frame) n_left_from = frame->n_vectors; next_index = node->cached_next_index; vlib_combined_counter_main_t *cm = map_main.domain_counters; - u32 cpu_index = os_get_cpu_number (); + u32 thread_index = vlib_get_thread_index (); while (n_left_from > 0) { @@ -1158,7 +1158,7 @@ ip4_map_t (vlib_main_t * vm, vlib_node_runtime_t * node, vlib_frame_t * frame) (error0 == MAP_ERROR_NONE && next0 != IP4_MAPT_NEXT_MAPT_ICMP)) { vlib_increment_combined_counter (cm + MAP_DOMAIN_COUNTER_TX, - cpu_index, + thread_index, vnet_buffer (p0)->map_t. map_domain_index, 1, clib_net_to_host_u16 (ip40-> @@ -1169,7 +1169,7 @@ ip4_map_t (vlib_main_t * vm, vlib_node_runtime_t * node, vlib_frame_t * frame) (error1 == MAP_ERROR_NONE && next1 != IP4_MAPT_NEXT_MAPT_ICMP)) { vlib_increment_combined_counter (cm + MAP_DOMAIN_COUNTER_TX, - cpu_index, + thread_index, vnet_buffer (p1)->map_t. map_domain_index, 1, clib_net_to_host_u16 (ip41-> @@ -1252,7 +1252,7 @@ ip4_map_t (vlib_main_t * vm, vlib_node_runtime_t * node, vlib_frame_t * frame) (error0 == MAP_ERROR_NONE && next0 != IP4_MAPT_NEXT_MAPT_ICMP)) { vlib_increment_combined_counter (cm + MAP_DOMAIN_COUNTER_TX, - cpu_index, + thread_index, vnet_buffer (p0)->map_t. map_domain_index, 1, clib_net_to_host_u16 (ip40-> diff --git a/src/vnet/map/ip6_map.c b/src/vnet/map/ip6_map.c index f7eb768f..63ada962 100644 --- a/src/vnet/map/ip6_map.c +++ b/src/vnet/map/ip6_map.c @@ -172,7 +172,7 @@ ip6_map (vlib_main_t * vm, vlib_node_runtime_t * node, vlib_frame_t * frame) vlib_node_get_runtime (vm, ip6_map_node.index); map_main_t *mm = &map_main; vlib_combined_counter_main_t *cm = mm->domain_counters; - u32 cpu_index = os_get_cpu_number (); + u32 thread_index = vlib_get_thread_index (); from = vlib_frame_vector_args (frame); n_left_from = frame->n_vectors; @@ -319,7 +319,7 @@ ip6_map (vlib_main_t * vm, vlib_node_runtime_t * node, vlib_frame_t * frame) IP6_MAP_NEXT_IP4_REWRITE : next0; } vlib_increment_combined_counter (cm + MAP_DOMAIN_COUNTER_RX, - cpu_index, + thread_index, map_domain_index0, 1, clib_net_to_host_u16 (ip40->length)); @@ -352,7 +352,7 @@ ip6_map (vlib_main_t * vm, vlib_node_runtime_t * node, vlib_frame_t * frame) IP6_MAP_NEXT_IP4_REWRITE : next1; } vlib_increment_combined_counter (cm + MAP_DOMAIN_COUNTER_RX, - cpu_index, + thread_index, map_domain_index1, 1, clib_net_to_host_u16 (ip41->length)); @@ -505,7 +505,7 @@ ip6_map (vlib_main_t * vm, vlib_node_runtime_t * node, vlib_frame_t * frame) IP6_MAP_NEXT_IP4_REWRITE : next0; } vlib_increment_combined_counter (cm + MAP_DOMAIN_COUNTER_RX, - cpu_index, + thread_index, map_domain_index0, 1, clib_net_to_host_u16 (ip40->length)); @@ -820,7 +820,7 @@ ip6_map_ip4_reass (vlib_main_t * vm, vlib_node_get_runtime (vm, ip6_map_ip4_reass_node.index); map_main_t *mm = &map_main; vlib_combined_counter_main_t *cm = mm->domain_counters; - u32 cpu_index = os_get_cpu_number (); + u32 thread_index = vlib_get_thread_index (); u32 *fragments_to_drop = NULL; u32 *fragments_to_loopback = NULL; @@ -958,8 +958,8 @@ ip6_map_ip4_reass (vlib_main_t * vm, { if (error0 == MAP_ERROR_NONE) vlib_increment_combined_counter (cm + MAP_DOMAIN_COUNTER_RX, - cpu_index, map_domain_index0, - 1, + thread_index, + map_domain_index0, 1, clib_net_to_host_u16 (ip40->length)); next0 = @@ -1015,7 +1015,7 @@ ip6_map_icmp_relay (vlib_main_t * vm, vlib_node_runtime_t *error_node = vlib_node_get_runtime (vm, ip6_map_icmp_relay_node.index); map_main_t *mm = &map_main; - u32 cpu_index = os_get_cpu_number (); + u32 thread_index = vlib_get_thread_index (); u16 *fragment_ids, *fid; from = vlib_frame_vector_args (frame); @@ -1143,7 +1143,8 @@ ip6_map_icmp_relay (vlib_main_t * vm, ip_csum_t sum = ip_incremental_checksum (0, new_icmp40, nlen - 20); new_icmp40->checksum = ~ip_csum_fold (sum); - vlib_increment_simple_counter (&mm->icmp_relayed, cpu_index, 0, 1); + vlib_increment_simple_counter (&mm->icmp_relayed, thread_index, 0, + 1); error: if (PREDICT_FALSE (p0->flags & VLIB_BUFFER_IS_TRACED)) diff --git a/src/vnet/map/ip6_map_t.c b/src/vnet/map/ip6_map_t.c index eb3996c2..99151678 100644 --- a/src/vnet/map/ip6_map_t.c +++ b/src/vnet/map/ip6_map_t.c @@ -448,7 +448,7 @@ ip6_map_t_icmp (vlib_main_t * vm, n_left_from = frame->n_vectors; next_index = node->cached_next_index; vlib_combined_counter_main_t *cm = map_main.domain_counters; - u32 cpu_index = os_get_cpu_number (); + u32 thread_index = vlib_get_thread_index (); while (n_left_from > 0) { @@ -493,7 +493,7 @@ ip6_map_t_icmp (vlib_main_t * vm, if (PREDICT_TRUE (error0 == MAP_ERROR_NONE)) { vlib_increment_combined_counter (cm + MAP_DOMAIN_COUNTER_RX, - cpu_index, + thread_index, vnet_buffer (p0)-> map_t.map_domain_index, 1, len0); @@ -1051,7 +1051,7 @@ ip6_map_t (vlib_main_t * vm, vlib_node_runtime_t * node, vlib_frame_t * frame) vlib_node_runtime_t *error_node = vlib_node_get_runtime (vm, ip6_map_t_node.index); vlib_combined_counter_main_t *cm = map_main.domain_counters; - u32 cpu_index = os_get_cpu_number (); + u32 thread_index = vlib_get_thread_index (); from = vlib_frame_vector_args (frame); n_left_from = frame->n_vectors; @@ -1218,7 +1218,7 @@ ip6_map_t (vlib_main_t * vm, vlib_node_runtime_t * node, vlib_frame_t * frame) (error0 == MAP_ERROR_NONE && next0 != IP6_MAPT_NEXT_MAPT_ICMP)) { vlib_increment_combined_counter (cm + MAP_DOMAIN_COUNTER_RX, - cpu_index, + thread_index, vnet_buffer (p0)-> map_t.map_domain_index, 1, clib_net_to_host_u16 @@ -1229,7 +1229,7 @@ ip6_map_t (vlib_main_t * vm, vlib_node_runtime_t * node, vlib_frame_t * frame) (error1 == MAP_ERROR_NONE && next1 != IP6_MAPT_NEXT_MAPT_ICMP)) { vlib_increment_combined_counter (cm + MAP_DOMAIN_COUNTER_RX, - cpu_index, + thread_index, vnet_buffer (p1)-> map_t.map_domain_index, 1, clib_net_to_host_u16 @@ -1403,7 +1403,7 @@ ip6_map_t (vlib_main_t * vm, vlib_node_runtime_t * node, vlib_frame_t * frame) (error0 == MAP_ERROR_NONE && next0 != IP6_MAPT_NEXT_MAPT_ICMP)) { vlib_increment_combined_counter (cm + MAP_DOMAIN_COUNTER_RX, - cpu_index, + thread_index, vnet_buffer (p0)-> map_t.map_domain_index, 1, clib_net_to_host_u16 diff --git a/src/vnet/mpls/mpls_input.c b/src/vnet/mpls/mpls_input.c index 893c4511..1b9bdd05 100644 --- a/src/vnet/mpls/mpls_input.c +++ b/src/vnet/mpls/mpls_input.c @@ -76,7 +76,7 @@ mpls_input_inline (vlib_main_t * vm, u32 n_left_from, next_index, * from, * to_next; mpls_input_runtime_t * rt; mpls_main_t * mm; - u32 cpu_index = os_get_cpu_number(); + u32 thread_index = vlib_get_thread_index(); vlib_simple_counter_main_t * cm; vnet_main_t * vnm = vnet_get_main(); @@ -151,7 +151,7 @@ mpls_input_inline (vlib_main_t * vm, next0 = MPLS_INPUT_NEXT_LOOKUP; vnet_feature_arc_start(mm->input_feature_arc_index, sw_if_index0, &next0, b0); - vlib_increment_simple_counter (cm, cpu_index, sw_if_index0, 1); + vlib_increment_simple_counter (cm, thread_index, sw_if_index0, 1); } if (PREDICT_FALSE(h1[3] == 0)) @@ -164,7 +164,7 @@ mpls_input_inline (vlib_main_t * vm, next1 = MPLS_INPUT_NEXT_LOOKUP; vnet_feature_arc_start(mm->input_feature_arc_index, sw_if_index1, &next1, b1); - vlib_increment_simple_counter (cm, cpu_index, sw_if_index1, 1); + vlib_increment_simple_counter (cm, thread_index, sw_if_index1, 1); } if (PREDICT_FALSE(b0->flags & VLIB_BUFFER_IS_TRACED)) @@ -215,7 +215,7 @@ mpls_input_inline (vlib_main_t * vm, { next0 = MPLS_INPUT_NEXT_LOOKUP; vnet_feature_arc_start(mm->input_feature_arc_index, sw_if_index0, &next0, b0); - vlib_increment_simple_counter (cm, cpu_index, sw_if_index0, 1); + vlib_increment_simple_counter (cm, thread_index, sw_if_index0, 1); } if (PREDICT_FALSE(b0->flags & VLIB_BUFFER_IS_TRACED)) diff --git a/src/vnet/mpls/mpls_lookup.c b/src/vnet/mpls/mpls_lookup.c index 475bb204..ace6a70f 100644 --- a/src/vnet/mpls/mpls_lookup.c +++ b/src/vnet/mpls/mpls_lookup.c @@ -67,7 +67,7 @@ mpls_lookup (vlib_main_t * vm, vlib_combined_counter_main_t * cm = &load_balance_main.lbm_to_counters; u32 n_left_from, next_index, * from, * to_next; mpls_main_t * mm = &mpls_main; - u32 cpu_index = os_get_cpu_number(); + u32 thread_index = vlib_get_thread_index(); from = vlib_frame_vector_args (from_frame); n_left_from = from_frame->n_vectors; @@ -220,16 +220,16 @@ mpls_lookup (vlib_main_t * vm, vnet_buffer (b3)->ip.adj_index[VLIB_TX] = dpo3->dpoi_index; vlib_increment_combined_counter - (cm, cpu_index, lbi0, 1, + (cm, thread_index, lbi0, 1, vlib_buffer_length_in_chain (vm, b0)); vlib_increment_combined_counter - (cm, cpu_index, lbi1, 1, + (cm, thread_index, lbi1, 1, vlib_buffer_length_in_chain (vm, b1)); vlib_increment_combined_counter - (cm, cpu_index, lbi2, 1, + (cm, thread_index, lbi2, 1, vlib_buffer_length_in_chain (vm, b2)); vlib_increment_combined_counter - (cm, cpu_index, lbi3, 1, + (cm, thread_index, lbi3, 1, vlib_buffer_length_in_chain (vm, b3)); /* @@ -351,7 +351,7 @@ mpls_lookup (vlib_main_t * vm, vnet_buffer (b0)->ip.adj_index[VLIB_TX] = dpo0->dpoi_index; vlib_increment_combined_counter - (cm, cpu_index, lbi0, 1, + (cm, thread_index, lbi0, 1, vlib_buffer_length_in_chain (vm, b0)); /* @@ -440,7 +440,7 @@ mpls_load_balance (vlib_main_t * vm, { vlib_combined_counter_main_t * cm = &load_balance_main.lbm_via_counters; u32 n_left_from, n_left_to_next, * from, * to_next; - u32 cpu_index = os_get_cpu_number(); + u32 thread_index = vlib_get_thread_index(); u32 next; from = vlib_frame_vector_args (frame); @@ -536,10 +536,10 @@ mpls_load_balance (vlib_main_t * vm, vnet_buffer (p1)->ip.adj_index[VLIB_TX] = dpo1->dpoi_index; vlib_increment_combined_counter - (cm, cpu_index, lbi0, 1, + (cm, thread_index, lbi0, 1, vlib_buffer_length_in_chain (vm, p0)); vlib_increment_combined_counter - (cm, cpu_index, lbi1, 1, + (cm, thread_index, lbi1, 1, vlib_buffer_length_in_chain (vm, p1)); if (PREDICT_FALSE(p0->flags & VLIB_BUFFER_IS_TRACED)) @@ -597,7 +597,7 @@ mpls_load_balance (vlib_main_t * vm, vnet_buffer (p0)->ip.adj_index[VLIB_TX] = dpo0->dpoi_index; vlib_increment_combined_counter - (cm, cpu_index, lbi0, 1, + (cm, thread_index, lbi0, 1, vlib_buffer_length_in_chain (vm, p0)); vlib_validate_buffer_enqueue_x1 (vm, node, next, diff --git a/src/vnet/mpls/mpls_output.c b/src/vnet/mpls/mpls_output.c index 08018fd1..d90dec21 100644 --- a/src/vnet/mpls/mpls_output.c +++ b/src/vnet/mpls/mpls_output.c @@ -64,12 +64,12 @@ mpls_output_inline (vlib_main_t * vm, vlib_frame_t * from_frame, int is_midchain) { - u32 n_left_from, next_index, * from, * to_next, cpu_index; + u32 n_left_from, next_index, * from, * to_next, thread_index; vlib_node_runtime_t * error_node; u32 n_left_to_next; mpls_main_t *mm; - cpu_index = os_get_cpu_number(); + thread_index = vlib_get_thread_index(); error_node = vlib_node_get_runtime (vm, mpls_output_node.index); from = vlib_frame_vector_args (from_frame); n_left_from = from_frame->n_vectors; @@ -137,13 +137,13 @@ mpls_output_inline (vlib_main_t * vm, /* Bump the adj counters for packet and bytes */ vlib_increment_combined_counter (&adjacency_counters, - cpu_index, + thread_index, adj_index0, 1, vlib_buffer_length_in_chain (vm, p0) + rw_len0); vlib_increment_combined_counter (&adjacency_counters, - cpu_index, + thread_index, adj_index1, 1, vlib_buffer_length_in_chain (vm, p1) + rw_len1); @@ -245,7 +245,7 @@ mpls_output_inline (vlib_main_t * vm, vlib_increment_combined_counter (&adjacency_counters, - cpu_index, + thread_index, adj_index0, 1, vlib_buffer_length_in_chain (vm, p0) + rw_len0); diff --git a/src/vnet/pg/input.c b/src/vnet/pg/input.c index 2649798b..597ae060 100644 --- a/src/vnet/pg/input.c +++ b/src/vnet/pg/input.c @@ -893,7 +893,7 @@ pg_generate_set_lengths (pg_main_t * pg, vlib_increment_combined_counter (im->combined_sw_if_counters + VNET_INTERFACE_COUNTER_RX, - os_get_cpu_number (), + vlib_get_thread_index (), si->sw_if_index, n_buffers, length_sum); } @@ -1266,7 +1266,7 @@ pg_stream_fill_helper (pg_main_t * pg, l += vlib_buffer_index_length_in_chain (vm, buffers[i]); vlib_increment_combined_counter (im->combined_sw_if_counters + VNET_INTERFACE_COUNTER_RX, - os_get_cpu_number (), + vlib_get_thread_index (), si->sw_if_index, n_alloc, l); s->current_replay_packet_index += n_alloc; s->current_replay_packet_index %= diff --git a/src/vnet/replication.c b/src/vnet/replication.c index 86d922b5..233a8c2f 100644 --- a/src/vnet/replication.c +++ b/src/vnet/replication.c @@ -31,16 +31,16 @@ replication_prep (vlib_main_t * vm, { replication_main_t *rm = &replication_main; replication_context_t *ctx; - uword cpu_number = vm->cpu_index; + uword thread_index = vm->thread_index; ip4_header_t *ip; u32 ctx_id; /* Allocate a context, reserve context 0 */ - if (PREDICT_FALSE (rm->contexts[cpu_number] == 0)) - pool_get_aligned (rm->contexts[cpu_number], ctx, CLIB_CACHE_LINE_BYTES); + if (PREDICT_FALSE (rm->contexts[thread_index] == 0)) + pool_get_aligned (rm->contexts[thread_index], ctx, CLIB_CACHE_LINE_BYTES); - pool_get_aligned (rm->contexts[cpu_number], ctx, CLIB_CACHE_LINE_BYTES); - ctx_id = ctx - rm->contexts[cpu_number]; + pool_get_aligned (rm->contexts[thread_index], ctx, CLIB_CACHE_LINE_BYTES); + ctx_id = ctx - rm->contexts[thread_index]; /* Save state from vlib buffer */ ctx->saved_free_list_index = b0->free_list_index; @@ -94,11 +94,11 @@ replication_recycle (vlib_main_t * vm, vlib_buffer_t * b0, u32 is_last) { replication_main_t *rm = &replication_main; replication_context_t *ctx; - uword cpu_number = vm->cpu_index; + uword thread_index = vm->thread_index; ip4_header_t *ip; /* Get access to the replication context */ - ctx = pool_elt_at_index (rm->contexts[cpu_number], b0->recycle_count); + ctx = pool_elt_at_index (rm->contexts[thread_index], b0->recycle_count); /* Restore vnet buffer state */ clib_memcpy (vnet_buffer (b0), ctx->vnet_buffer, @@ -133,7 +133,7 @@ replication_recycle (vlib_main_t * vm, vlib_buffer_t * b0, u32 is_last) b0->flags &= ~VLIB_BUFFER_RECYCLE; /* Free context back to its pool */ - pool_put (rm->contexts[cpu_number], ctx); + pool_put (rm->contexts[thread_index], ctx); } return ctx; @@ -160,7 +160,7 @@ replication_recycle_callback (vlib_main_t * vm, vlib_buffer_free_list_t * fl) replication_main_t *rm = &replication_main; replication_context_t *ctx; u32 feature_node_index = 0; - uword cpu_number = vm->cpu_index; + uword thread_index = vm->thread_index; /* * All buffers in the list are destined to the same recycle node. @@ -172,7 +172,7 @@ replication_recycle_callback (vlib_main_t * vm, vlib_buffer_free_list_t * fl) { bi0 = fl->buffers[0]; b0 = vlib_get_buffer (vm, bi0); - ctx = pool_elt_at_index (rm->contexts[cpu_number], b0->recycle_count); + ctx = pool_elt_at_index (rm->contexts[thread_index], b0->recycle_count); feature_node_index = ctx->recycle_node_index; } diff --git a/src/vnet/replication.h b/src/vnet/replication.h index 5dc554c9..ce4b3ff1 100644 --- a/src/vnet/replication.h +++ b/src/vnet/replication.h @@ -100,7 +100,7 @@ replication_get_ctx (vlib_buffer_t * b0) replication_main_t *rm = &replication_main; return replication_is_recycled (b0) ? - pool_elt_at_index (rm->contexts[os_get_cpu_number ()], + pool_elt_at_index (rm->contexts[vlib_get_thread_index ()], b0->recycle_count) : 0; } diff --git a/src/vnet/session/node.c b/src/vnet/session/node.c index b86e87d9..dd211c51 100644 --- a/src/vnet/session/node.c +++ b/src/vnet/session/node.c @@ -311,7 +311,7 @@ session_queue_node_fn (vlib_main_t * vm, vlib_node_runtime_t * node, unix_shared_memory_queue_t *q; application_t *app; int n_tx_packets = 0; - u32 my_thread_index = vm->cpu_index; + u32 my_thread_index = vm->thread_index; int i, rv; f64 now = vlib_time_now (vm); diff --git a/src/vnet/sr/sr_localsid.c b/src/vnet/sr/sr_localsid.c index 2e3d56de..6d72a506 100755 --- a/src/vnet/sr/sr_localsid.c +++ b/src/vnet/sr/sr_localsid.c @@ -887,7 +887,7 @@ sr_localsid_d_fn (vlib_main_t * vm, vlib_node_runtime_t * node, from = vlib_frame_vector_args (from_frame); n_left_from = from_frame->n_vectors; next_index = node->cached_next_index; - u32 cpu_index = os_get_cpu_number (); + u32 thread_index = vlib_get_thread_index (); while (n_left_from > 0) { @@ -974,26 +974,26 @@ sr_localsid_d_fn (vlib_main_t * vm, vlib_node_runtime_t * node, vlib_increment_combined_counter (((next0 == SR_LOCALSID_NEXT_ERROR) ? &(sm->sr_ls_invalid_counters) : - &(sm->sr_ls_valid_counters)), cpu_index, ls0 - sm->localsids, 1, - vlib_buffer_length_in_chain (vm, b0)); + &(sm->sr_ls_valid_counters)), thread_index, ls0 - sm->localsids, + 1, vlib_buffer_length_in_chain (vm, b0)); vlib_increment_combined_counter (((next1 == SR_LOCALSID_NEXT_ERROR) ? &(sm->sr_ls_invalid_counters) : - &(sm->sr_ls_valid_counters)), cpu_index, ls1 - sm->localsids, 1, - vlib_buffer_length_in_chain (vm, b1)); + &(sm->sr_ls_valid_counters)), thread_index, ls1 - sm->localsids, + 1, vlib_buffer_length_in_chain (vm, b1)); vlib_increment_combined_counter (((next2 == SR_LOCALSID_NEXT_ERROR) ? &(sm->sr_ls_invalid_counters) : - &(sm->sr_ls_valid_counters)), cpu_index, ls2 - sm->localsids, 1, - vlib_buffer_length_in_chain (vm, b2)); + &(sm->sr_ls_valid_counters)), thread_index, ls2 - sm->localsids, + 1, vlib_buffer_length_in_chain (vm, b2)); vlib_increment_combined_counter (((next3 == SR_LOCALSID_NEXT_ERROR) ? &(sm->sr_ls_invalid_counters) : - &(sm->sr_ls_valid_counters)), cpu_index, ls3 - sm->localsids, 1, - vlib_buffer_length_in_chain (vm, b3)); + &(sm->sr_ls_valid_counters)), thread_index, ls3 - sm->localsids, + 1, vlib_buffer_length_in_chain (vm, b3)); vlib_validate_buffer_enqueue_x4 (vm, node, next_index, to_next, n_left_to_next, bi0, bi1, bi2, bi3, @@ -1062,8 +1062,8 @@ sr_localsid_d_fn (vlib_main_t * vm, vlib_node_runtime_t * node, vlib_increment_combined_counter (((next0 == SR_LOCALSID_NEXT_ERROR) ? &(sm->sr_ls_invalid_counters) : - &(sm->sr_ls_valid_counters)), cpu_index, ls0 - sm->localsids, 1, - vlib_buffer_length_in_chain (vm, b0)); + &(sm->sr_ls_valid_counters)), thread_index, ls0 - sm->localsids, + 1, vlib_buffer_length_in_chain (vm, b0)); vlib_validate_buffer_enqueue_x1 (vm, node, next_index, to_next, n_left_to_next, bi0, next0); @@ -1103,7 +1103,7 @@ sr_localsid_fn (vlib_main_t * vm, vlib_node_runtime_t * node, from = vlib_frame_vector_args (from_frame); n_left_from = from_frame->n_vectors; next_index = node->cached_next_index; - u32 cpu_index = os_get_cpu_number (); + u32 thread_index = vlib_get_thread_index (); while (n_left_from > 0) { @@ -1205,26 +1205,26 @@ sr_localsid_fn (vlib_main_t * vm, vlib_node_runtime_t * node, vlib_increment_combined_counter (((next0 == SR_LOCALSID_NEXT_ERROR) ? &(sm->sr_ls_invalid_counters) : - &(sm->sr_ls_valid_counters)), cpu_index, ls0 - sm->localsids, 1, - vlib_buffer_length_in_chain (vm, b0)); + &(sm->sr_ls_valid_counters)), thread_index, ls0 - sm->localsids, + 1, vlib_buffer_length_in_chain (vm, b0)); vlib_increment_combined_counter (((next1 == SR_LOCALSID_NEXT_ERROR) ? &(sm->sr_ls_invalid_counters) : - &(sm->sr_ls_valid_counters)), cpu_index, ls1 - sm->localsids, 1, - vlib_buffer_length_in_chain (vm, b1)); + &(sm->sr_ls_valid_counters)), thread_index, ls1 - sm->localsids, + 1, vlib_buffer_length_in_chain (vm, b1)); vlib_increment_combined_counter (((next2 == SR_LOCALSID_NEXT_ERROR) ? &(sm->sr_ls_invalid_counters) : - &(sm->sr_ls_valid_counters)), cpu_index, ls2 - sm->localsids, 1, - vlib_buffer_length_in_chain (vm, b2)); + &(sm->sr_ls_valid_counters)), thread_index, ls2 - sm->localsids, + 1, vlib_buffer_length_in_chain (vm, b2)); vlib_increment_combined_counter (((next3 == SR_LOCALSID_NEXT_ERROR) ? &(sm->sr_ls_invalid_counters) : - &(sm->sr_ls_valid_counters)), cpu_index, ls3 - sm->localsids, 1, - vlib_buffer_length_in_chain (vm, b3)); + &(sm->sr_ls_valid_counters)), thread_index, ls3 - sm->localsids, + 1, vlib_buffer_length_in_chain (vm, b3)); vlib_validate_buffer_enqueue_x4 (vm, node, next_index, to_next, n_left_to_next, bi0, bi1, bi2, bi3, @@ -1295,8 +1295,8 @@ sr_localsid_fn (vlib_main_t * vm, vlib_node_runtime_t * node, vlib_increment_combined_counter (((next0 == SR_LOCALSID_NEXT_ERROR) ? &(sm->sr_ls_invalid_counters) : - &(sm->sr_ls_valid_counters)), cpu_index, ls0 - sm->localsids, 1, - vlib_buffer_length_in_chain (vm, b0)); + &(sm->sr_ls_valid_counters)), thread_index, ls0 - sm->localsids, + 1, vlib_buffer_length_in_chain (vm, b0)); vlib_validate_buffer_enqueue_x1 (vm, node, next_index, to_next, n_left_to_next, bi0, next0); diff --git a/src/vnet/tcp/builtin_client.c b/src/vnet/tcp/builtin_client.c index e3705060..c1567aa0 100644 --- a/src/vnet/tcp/builtin_client.c +++ b/src/vnet/tcp/builtin_client.c @@ -174,7 +174,7 @@ tclient_thread_fn (void *arg) pthread_sigmask (SIG_SETMASK, &s, 0); } - clib_per_cpu_mheaps[os_get_cpu_number ()] = clib_per_cpu_mheaps[0]; + clib_per_cpu_mheaps[vlib_get_thread_index ()] = clib_per_cpu_mheaps[0]; while (1) { diff --git a/src/vnet/tcp/tcp.c b/src/vnet/tcp/tcp.c index b2a371e2..b6c34828 100644 --- a/src/vnet/tcp/tcp.c +++ b/src/vnet/tcp/tcp.c @@ -646,10 +646,10 @@ const static transport_proto_vft_t tcp6_proto = { void tcp_timer_keep_handler (u32 conn_index) { - u32 cpu_index = os_get_cpu_number (); + u32 thread_index = vlib_get_thread_index (); tcp_connection_t *tc; - tc = tcp_connection_get (conn_index, cpu_index); + tc = tcp_connection_get (conn_index, thread_index); tc->timers[TCP_TIMER_KEEP] = TCP_TIMER_HANDLE_INVALID; tcp_connection_close (tc); @@ -675,10 +675,10 @@ tcp_timer_establish_handler (u32 conn_index) void tcp_timer_waitclose_handler (u32 conn_index) { - u32 cpu_index = os_get_cpu_number (); + u32 thread_index = vlib_get_thread_index (); tcp_connection_t *tc; - tc = tcp_connection_get (conn_index, cpu_index); + tc = tcp_connection_get (conn_index, thread_index); tc->timers[TCP_TIMER_WAITCLOSE] = TCP_TIMER_HANDLE_INVALID; /* Session didn't come back with a close(). Send FIN either way diff --git a/src/vnet/tcp/tcp_debug.h b/src/vnet/tcp/tcp_debug.h index 0090e15e..eaca672c 100644 --- a/src/vnet/tcp/tcp_debug.h +++ b/src/vnet/tcp/tcp_debug.h @@ -343,7 +343,7 @@ typedef enum _tcp_dbg_evt } \ else \ { \ - u32 _thread_index = os_get_cpu_number (); \ + u32 _thread_index = vlib_get_thread_index (); \ _tc = tcp_connection_get (_tc_index, _thread_index); \ } \ ELOG_TYPE_DECLARE (_e) = \ diff --git a/src/vnet/tcp/tcp_input.c b/src/vnet/tcp/tcp_input.c index a8224dc2..7e9fa47b 100644 --- a/src/vnet/tcp/tcp_input.c +++ b/src/vnet/tcp/tcp_input.c @@ -1142,7 +1142,7 @@ tcp46_established_inline (vlib_main_t * vm, vlib_node_runtime_t * node, vlib_frame_t * from_frame, int is_ip4) { u32 n_left_from, next_index, *from, *to_next; - u32 my_thread_index = vm->cpu_index, errors = 0; + u32 my_thread_index = vm->thread_index, errors = 0; tcp_main_t *tm = vnet_get_tcp_main (); from = vlib_frame_vector_args (from_frame); @@ -1332,7 +1332,7 @@ tcp46_syn_sent_inline (vlib_main_t * vm, vlib_node_runtime_t * node, { tcp_main_t *tm = vnet_get_tcp_main (); u32 n_left_from, next_index, *from, *to_next; - u32 my_thread_index = vm->cpu_index, errors = 0; + u32 my_thread_index = vm->thread_index, errors = 0; u8 sst = is_ip4 ? SESSION_TYPE_IP4_TCP : SESSION_TYPE_IP6_TCP; from = vlib_frame_vector_args (from_frame); @@ -1634,7 +1634,7 @@ tcp46_rcv_process_inline (vlib_main_t * vm, vlib_node_runtime_t * node, { tcp_main_t *tm = vnet_get_tcp_main (); u32 n_left_from, next_index, *from, *to_next; - u32 my_thread_index = vm->cpu_index, errors = 0; + u32 my_thread_index = vm->thread_index, errors = 0; from = vlib_frame_vector_args (from_frame); n_left_from = from_frame->n_vectors; @@ -1989,7 +1989,7 @@ tcp46_listen_inline (vlib_main_t * vm, vlib_node_runtime_t * node, vlib_frame_t * from_frame, int is_ip4) { u32 n_left_from, next_index, *from, *to_next; - u32 my_thread_index = vm->cpu_index; + u32 my_thread_index = vm->thread_index; tcp_main_t *tm = vnet_get_tcp_main (); u8 sst = is_ip4 ? SESSION_TYPE_IP4_TCP : SESSION_TYPE_IP6_TCP; @@ -2243,7 +2243,7 @@ tcp46_input_inline (vlib_main_t * vm, vlib_node_runtime_t * node, vlib_frame_t * from_frame, int is_ip4) { u32 n_left_from, next_index, *from, *to_next; - u32 my_thread_index = vm->cpu_index; + u32 my_thread_index = vm->thread_index; tcp_main_t *tm = vnet_get_tcp_main (); from = vlib_frame_vector_args (from_frame); diff --git a/src/vnet/tcp/tcp_output.c b/src/vnet/tcp/tcp_output.c index ea157bd7..e18bfad7 100644 --- a/src/vnet/tcp/tcp_output.c +++ b/src/vnet/tcp/tcp_output.c @@ -387,8 +387,8 @@ tcp_make_options (tcp_connection_t * tc, tcp_options_t * opts, #define tcp_get_free_buffer_index(tm, bidx) \ do { \ u32 *my_tx_buffers, n_free_buffers; \ - u32 cpu_index = os_get_cpu_number(); \ - my_tx_buffers = tm->tx_buffers[cpu_index]; \ + u32 thread_index = vlib_get_thread_index(); \ + my_tx_buffers = tm->tx_buffers[thread_index]; \ if (PREDICT_FALSE(vec_len (my_tx_buffers) == 0)) \ { \ n_free_buffers = 32; /* TODO config or macro */ \ @@ -396,7 +396,7 @@ do { \ _vec_len(my_tx_buffers) = vlib_buffer_alloc_from_free_list ( \ tm->vlib_main, my_tx_buffers, n_free_buffers, \ VLIB_BUFFER_DEFAULT_FREE_LIST_INDEX); \ - tm->tx_buffers[cpu_index] = my_tx_buffers; \ + tm->tx_buffers[thread_index] = my_tx_buffers; \ } \ /* buffer shortage */ \ if (PREDICT_FALSE (vec_len (my_tx_buffers) == 0)) \ @@ -408,8 +408,8 @@ do { \ #define tcp_return_buffer(tm) \ do { \ u32 *my_tx_buffers; \ - u32 cpu_index = os_get_cpu_number(); \ - my_tx_buffers = tm->tx_buffers[cpu_index]; \ + u32 thread_index = vlib_get_thread_index(); \ + my_tx_buffers = tm->tx_buffers[thread_index]; \ _vec_len (my_tx_buffers) +=1; \ } while (0) @@ -942,7 +942,7 @@ tcp_send_ack (tcp_connection_t * tc) void tcp_timer_delack_handler (u32 index) { - u32 thread_index = os_get_cpu_number (); + u32 thread_index = vlib_get_thread_index (); tcp_connection_t *tc; tc = tcp_connection_get (index, thread_index); @@ -1022,7 +1022,7 @@ tcp_timer_retransmit_handler_i (u32 index, u8 is_syn) { tcp_main_t *tm = vnet_get_tcp_main (); vlib_main_t *vm = vlib_get_main (); - u32 thread_index = os_get_cpu_number (); + u32 thread_index = vlib_get_thread_index (); tcp_connection_t *tc; vlib_buffer_t *b; u32 bi, snd_space, n_bytes; @@ -1152,7 +1152,7 @@ tcp_timer_persist_handler (u32 index) { tcp_main_t *tm = vnet_get_tcp_main (); vlib_main_t *vm = vlib_get_main (); - u32 thread_index = os_get_cpu_number (); + u32 thread_index = vlib_get_thread_index (); tcp_connection_t *tc; vlib_buffer_t *b; u32 bi, n_bytes; @@ -1313,7 +1313,7 @@ tcp46_output_inline (vlib_main_t * vm, vlib_frame_t * from_frame, int is_ip4) { u32 n_left_from, next_index, *from, *to_next; - u32 my_thread_index = vm->cpu_index; + u32 my_thread_index = vm->thread_index; from = vlib_frame_vector_args (from_frame); n_left_from = from_frame->n_vectors; @@ -1524,7 +1524,7 @@ tcp46_send_reset_inline (vlib_main_t * vm, vlib_node_runtime_t * node, vlib_frame_t * from_frame, u8 is_ip4) { u32 n_left_from, next_index, *from, *to_next; - u32 my_thread_index = vm->cpu_index; + u32 my_thread_index = vm->thread_index; from = vlib_frame_vector_args (from_frame); n_left_from = from_frame->n_vectors; diff --git a/src/vnet/udp/udp_input.c b/src/vnet/udp/udp_input.c index 4b22109b..810278e6 100644 --- a/src/vnet/udp/udp_input.c +++ b/src/vnet/udp/udp_input.c @@ -70,7 +70,7 @@ udp4_uri_input_node_fn (vlib_main_t * vm, udp4_uri_input_next_t next_index; udp_uri_main_t *um = vnet_get_udp_main (); session_manager_main_t *smm = vnet_get_session_manager_main (); - u32 my_thread_index = vm->cpu_index; + u32 my_thread_index = vm->thread_index; u8 my_enqueue_epoch; u32 *session_indices_to_enqueue; static u32 serial_number; diff --git a/src/vnet/unix/tapcli.c b/src/vnet/unix/tapcli.c index fb1a8bac..0fc62f6c 100644 --- a/src/vnet/unix/tapcli.c +++ b/src/vnet/unix/tapcli.c @@ -366,7 +366,7 @@ static uword tapcli_rx_iface(vlib_main_t * vm, vlib_increment_combined_counter ( vnet_main.interface_main.combined_sw_if_counters + VNET_INTERFACE_COUNTER_RX, - os_get_cpu_number(), ti->sw_if_index, + vlib_get_thread_index(), ti->sw_if_index, 1, n_bytes_in_packet); if (PREDICT_FALSE(n_trace > 0)) { diff --git a/src/vnet/unix/tuntap.c b/src/vnet/unix/tuntap.c index 2cfcc92f..ac674653 100644 --- a/src/vnet/unix/tuntap.c +++ b/src/vnet/unix/tuntap.c @@ -189,7 +189,7 @@ tuntap_tx (vlib_main_t * vm, /* Update tuntap interface output stats. */ vlib_increment_combined_counter (im->combined_sw_if_counters + VNET_INTERFACE_COUNTER_TX, - vm->cpu_index, + vm->thread_index, tm->sw_if_index, n_packets, n_bytes); @@ -297,7 +297,7 @@ tuntap_rx (vlib_main_t * vm, vlib_increment_combined_counter (vnet_main.interface_main.combined_sw_if_counters + VNET_INTERFACE_COUNTER_RX, - os_get_cpu_number(), + vlib_get_thread_index(), tm->sw_if_index, 1, n_bytes_in_packet); diff --git a/src/vnet/vxlan-gpe/decap.c b/src/vnet/vxlan-gpe/decap.c index 22ab4b62..d4fe4231 100644 --- a/src/vnet/vxlan-gpe/decap.c +++ b/src/vnet/vxlan-gpe/decap.c @@ -115,7 +115,7 @@ vxlan_gpe_input (vlib_main_t * vm, vxlan4_gpe_tunnel_key_t last_key4; vxlan6_gpe_tunnel_key_t last_key6; u32 pkts_decapsulated = 0; - u32 cpu_index = os_get_cpu_number (); + u32 thread_index = vlib_get_thread_index (); u32 stats_sw_if_index, stats_n_packets, stats_n_bytes; if (is_ip4) @@ -342,7 +342,7 @@ vxlan_gpe_input (vlib_main_t * vm, if (stats_n_packets) vlib_increment_combined_counter ( im->combined_sw_if_counters + VNET_INTERFACE_COUNTER_RX, - cpu_index, stats_sw_if_index, stats_n_packets, stats_n_bytes); + thread_index, stats_sw_if_index, stats_n_packets, stats_n_bytes); stats_n_packets = 1; stats_n_bytes = len0; stats_sw_if_index = sw_if_index0; @@ -427,7 +427,7 @@ vxlan_gpe_input (vlib_main_t * vm, if (stats_n_packets) vlib_increment_combined_counter ( im->combined_sw_if_counters + VNET_INTERFACE_COUNTER_RX, - cpu_index, stats_sw_if_index, stats_n_packets, stats_n_bytes); + thread_index, stats_sw_if_index, stats_n_packets, stats_n_bytes); stats_n_packets = 1; stats_n_bytes = len1; stats_sw_if_index = sw_if_index1; @@ -588,7 +588,7 @@ vxlan_gpe_input (vlib_main_t * vm, if (stats_n_packets) vlib_increment_combined_counter ( im->combined_sw_if_counters + VNET_INTERFACE_COUNTER_RX, - cpu_index, stats_sw_if_index, stats_n_packets, stats_n_bytes); + thread_index, stats_sw_if_index, stats_n_packets, stats_n_bytes); stats_n_packets = 1; stats_n_bytes = len0; stats_sw_if_index = sw_if_index0; @@ -615,7 +615,7 @@ vxlan_gpe_input (vlib_main_t * vm, if (stats_n_packets) { vlib_increment_combined_counter ( - im->combined_sw_if_counters + VNET_INTERFACE_COUNTER_RX, cpu_index, + im->combined_sw_if_counters + VNET_INTERFACE_COUNTER_RX, thread_index, stats_sw_if_index, stats_n_packets, stats_n_bytes); node->runtime_data[0] = stats_sw_if_index; } diff --git a/src/vnet/vxlan-gpe/encap.c b/src/vnet/vxlan-gpe/encap.c index 3a486e56..67ed94b4 100644 --- a/src/vnet/vxlan-gpe/encap.c +++ b/src/vnet/vxlan-gpe/encap.c @@ -151,7 +151,7 @@ vxlan_gpe_encap (vlib_main_t * vm, vnet_main_t * vnm = ngm->vnet_main; vnet_interface_main_t * im = &vnm->interface_main; u32 pkts_encapsulated = 0; - u32 cpu_index = os_get_cpu_number (); + u32 thread_index = vlib_get_thread_index (); u32 stats_sw_if_index, stats_n_packets, stats_n_bytes; from = vlib_frame_vector_args (from_frame); @@ -253,7 +253,7 @@ vxlan_gpe_encap (vlib_main_t * vm, if (stats_n_packets) vlib_increment_combined_counter ( im->combined_sw_if_counters + VNET_INTERFACE_COUNTER_TX, - cpu_index, stats_sw_if_index, stats_n_packets, stats_n_bytes); + thread_index, stats_sw_if_index, stats_n_packets, stats_n_bytes); stats_sw_if_index = sw_if_index0; stats_n_packets = 2; stats_n_bytes = len0 + len1; @@ -262,10 +262,10 @@ vxlan_gpe_encap (vlib_main_t * vm, { vlib_increment_combined_counter ( im->combined_sw_if_counters + VNET_INTERFACE_COUNTER_TX, - cpu_index, sw_if_index0, 1, len0); + thread_index, sw_if_index0, 1, len0); vlib_increment_combined_counter ( im->combined_sw_if_counters + VNET_INTERFACE_COUNTER_TX, - cpu_index, sw_if_index1, 1, len1); + thread_index, sw_if_index1, 1, len1); } } @@ -335,7 +335,7 @@ vxlan_gpe_encap (vlib_main_t * vm, if (stats_n_packets) vlib_increment_combined_counter ( im->combined_sw_if_counters + VNET_INTERFACE_COUNTER_TX, - cpu_index, stats_sw_if_index, stats_n_packets, stats_n_bytes); + thread_index, stats_sw_if_index, stats_n_packets, stats_n_bytes); stats_n_packets = 1; stats_n_bytes = len0; stats_sw_if_index = sw_if_index0; @@ -359,7 +359,7 @@ vxlan_gpe_encap (vlib_main_t * vm, if (stats_n_packets) { vlib_increment_combined_counter ( - im->combined_sw_if_counters + VNET_INTERFACE_COUNTER_TX, cpu_index, + im->combined_sw_if_counters + VNET_INTERFACE_COUNTER_TX, thread_index, stats_sw_if_index, stats_n_packets, stats_n_bytes); node->runtime_data[0] = stats_sw_if_index; } diff --git a/src/vnet/vxlan/decap.c b/src/vnet/vxlan/decap.c index 514b2c99..2acb1f6f 100644 --- a/src/vnet/vxlan/decap.c +++ b/src/vnet/vxlan/decap.c @@ -81,7 +81,7 @@ vxlan_input (vlib_main_t * vm, vxlan4_tunnel_key_t last_key4; vxlan6_tunnel_key_t last_key6; u32 pkts_decapsulated = 0; - u32 cpu_index = os_get_cpu_number(); + u32 thread_index = vlib_get_thread_index(); u32 stats_sw_if_index, stats_n_packets, stats_n_bytes; if (is_ip4) @@ -314,7 +314,7 @@ vxlan_input (vlib_main_t * vm, if (stats_n_packets) vlib_increment_combined_counter (im->combined_sw_if_counters + VNET_INTERFACE_COUNTER_RX, - cpu_index, stats_sw_if_index, + thread_index, stats_sw_if_index, stats_n_packets, stats_n_bytes); stats_n_packets = 1; stats_n_bytes = len0; @@ -468,7 +468,7 @@ vxlan_input (vlib_main_t * vm, if (stats_n_packets) vlib_increment_combined_counter (im->combined_sw_if_counters + VNET_INTERFACE_COUNTER_RX, - cpu_index, stats_sw_if_index, + thread_index, stats_sw_if_index, stats_n_packets, stats_n_bytes); stats_n_packets = 1; stats_n_bytes = len1; @@ -674,7 +674,7 @@ vxlan_input (vlib_main_t * vm, if (stats_n_packets) vlib_increment_combined_counter (im->combined_sw_if_counters + VNET_INTERFACE_COUNTER_RX, - cpu_index, stats_sw_if_index, + thread_index, stats_sw_if_index, stats_n_packets, stats_n_bytes); stats_n_packets = 1; stats_n_bytes = len0; @@ -711,7 +711,7 @@ vxlan_input (vlib_main_t * vm, { vlib_increment_combined_counter (im->combined_sw_if_counters + VNET_INTERFACE_COUNTER_RX, - cpu_index, stats_sw_if_index, stats_n_packets, stats_n_bytes); + thread_index, stats_sw_if_index, stats_n_packets, stats_n_bytes); node->runtime_data[0] = stats_sw_if_index; } diff --git a/src/vnet/vxlan/encap.c b/src/vnet/vxlan/encap.c index 5b63064a..4cfbbc23 100644 --- a/src/vnet/vxlan/encap.c +++ b/src/vnet/vxlan/encap.c @@ -77,7 +77,7 @@ vxlan_encap_inline (vlib_main_t * vm, vnet_interface_main_t * im = &vnm->interface_main; u32 pkts_encapsulated = 0; u16 old_l0 = 0, old_l1 = 0; - u32 cpu_index = os_get_cpu_number(); + u32 thread_index = vlib_get_thread_index(); u32 stats_sw_if_index, stats_n_packets, stats_n_bytes; u32 sw_if_index0 = 0, sw_if_index1 = 0; u32 next0 = 0, next1 = 0; @@ -301,7 +301,7 @@ vxlan_encap_inline (vlib_main_t * vm, if (stats_n_packets) vlib_increment_combined_counter (im->combined_sw_if_counters + VNET_INTERFACE_COUNTER_TX, - cpu_index, stats_sw_if_index, + thread_index, stats_sw_if_index, stats_n_packets, stats_n_bytes); stats_sw_if_index = sw_if_index0; stats_n_packets = 2; @@ -311,10 +311,10 @@ vxlan_encap_inline (vlib_main_t * vm, { vlib_increment_combined_counter (im->combined_sw_if_counters + VNET_INTERFACE_COUNTER_TX, - cpu_index, sw_if_index0, 1, len0); + thread_index, sw_if_index0, 1, len0); vlib_increment_combined_counter (im->combined_sw_if_counters + VNET_INTERFACE_COUNTER_TX, - cpu_index, sw_if_index1, 1, len1); + thread_index, sw_if_index1, 1, len1); } } @@ -464,7 +464,7 @@ vxlan_encap_inline (vlib_main_t * vm, if (stats_n_packets) vlib_increment_combined_counter (im->combined_sw_if_counters + VNET_INTERFACE_COUNTER_TX, - cpu_index, stats_sw_if_index, + thread_index, stats_sw_if_index, stats_n_packets, stats_n_bytes); stats_n_packets = 1; stats_n_bytes = len0; @@ -496,7 +496,7 @@ vxlan_encap_inline (vlib_main_t * vm, { vlib_increment_combined_counter (im->combined_sw_if_counters + VNET_INTERFACE_COUNTER_TX, - cpu_index, stats_sw_if_index, stats_n_packets, stats_n_bytes); + thread_index, stats_sw_if_index, stats_n_packets, stats_n_bytes); node->runtime_data[0] = stats_sw_if_index; } diff --git a/src/vpp/stats/stats.c b/src/vpp/stats/stats.c index 042d02e2..4309cd51 100644 --- a/src/vpp/stats/stats.c +++ b/src/vpp/stats/stats.c @@ -66,14 +66,14 @@ _(VNET_IP6_NBR_COUNTERS, vnet_ip6_nbr_counters) void dslock (stats_main_t * sm, int release_hint, int tag) { - u32 thread_id; + u32 thread_index; data_structure_lock_t *l = sm->data_structure_lock; if (PREDICT_FALSE (l == 0)) return; - thread_id = os_get_cpu_number (); - if (l->lock && l->thread_id == thread_id) + thread_index = vlib_get_thread_index (); + if (l->lock && l->thread_index == thread_index) { l->count++; return; @@ -85,7 +85,7 @@ dslock (stats_main_t * sm, int release_hint, int tag) while (__sync_lock_test_and_set (&l->lock, 1)) /* zzzz */ ; l->tag = tag; - l->thread_id = thread_id; + l->thread_index = thread_index; l->count = 1; } @@ -99,14 +99,14 @@ stats_dslock_with_hint (int hint, int tag) void dsunlock (stats_main_t * sm) { - u32 thread_id; + u32 thread_index; data_structure_lock_t *l = sm->data_structure_lock; if (PREDICT_FALSE (l == 0)) return; - thread_id = os_get_cpu_number (); - ASSERT (l->lock && l->thread_id == thread_id); + thread_index = vlib_get_thread_index (); + ASSERT (l->lock && l->thread_index == thread_index); l->count--; if (l->count == 0) { diff --git a/src/vpp/stats/stats.h b/src/vpp/stats/stats.h index 118115be..024dc78e 100644 --- a/src/vpp/stats/stats.h +++ b/src/vpp/stats/stats.h @@ -30,7 +30,7 @@ typedef struct { volatile u32 lock; volatile u32 release_hint; - u32 thread_id; + u32 thread_index; u32 count; int tag; } data_structure_lock_t; -- cgit 1.2.3-korg From 630198f04916deb35c5b7774823ae1a5dd168a6c Mon Sep 17 00:00:00 2001 From: Neale Ranns Date: Mon, 22 May 2017 09:20:20 -0400 Subject: IPv6 Performance bugs - inline the FIB lookup function; this requires access to the bihash, so for files that use more than one type this casues problems. those files that include ip6_fib.h unnecessarily have been updated - better use of the feature arcs. ip6-lookup and interface-output are now sentinels (end-node-index in the cm speak) rather than enabled features. Change-Id: I9d1375fee63f7dbb2d327da6124d8e60b63367ec Signed-off-by: Neale Ranns --- src/plugins/ioam/lib-vxlan-gpe/ioam_transit.c | 1 - src/plugins/ioam/lib-vxlan-gpe/vxlan_gpe_ioam.c | 2 +- src/vnet/dhcp/dhcp6_proxy_node.c | 9 +++--- src/vnet/fib/fib.h | 2 -- src/vnet/fib/ip6_fib.c | 39 ----------------------- src/vnet/fib/ip6_fib.h | 42 +++++++++++++++++++++++-- src/vnet/ip/ip6_forward.c | 32 +++++++++---------- src/vnet/vxlan-gpe/vxlan_gpe.c | 8 ++--- src/vpp/api/api.c | 2 -- 9 files changed, 63 insertions(+), 74 deletions(-) (limited to 'src/vnet/vxlan-gpe') diff --git a/src/plugins/ioam/lib-vxlan-gpe/ioam_transit.c b/src/plugins/ioam/lib-vxlan-gpe/ioam_transit.c index f334c983..d90cd5e4 100644 --- a/src/plugins/ioam/lib-vxlan-gpe/ioam_transit.c +++ b/src/plugins/ioam/lib-vxlan-gpe/ioam_transit.c @@ -22,7 +22,6 @@ #include #include #include -#include #include #include diff --git a/src/plugins/ioam/lib-vxlan-gpe/vxlan_gpe_ioam.c b/src/plugins/ioam/lib-vxlan-gpe/vxlan_gpe_ioam.c index cfc550cd..8558c505 100644 --- a/src/plugins/ioam/lib-vxlan-gpe/vxlan_gpe_ioam.c +++ b/src/plugins/ioam/lib-vxlan-gpe/vxlan_gpe_ioam.c @@ -16,7 +16,7 @@ #include #include #include -#include +#include #include #include diff --git a/src/vnet/dhcp/dhcp6_proxy_node.c b/src/vnet/dhcp/dhcp6_proxy_node.c index de73154d..885313a5 100644 --- a/src/vnet/dhcp/dhcp6_proxy_node.c +++ b/src/vnet/dhcp/dhcp6_proxy_node.c @@ -19,9 +19,9 @@ #include #include #include -#include #include #include +#include static char * dhcpv6_proxy_error_strings[] = { #define dhcpv6_proxy_error(n,s) s, @@ -966,7 +966,7 @@ static u8 * format_dhcp6_proxy_server (u8 * s, va_list * args) { dhcp_proxy_t * proxy = va_arg (*args, dhcp_proxy_t *); - ip6_fib_t *server_fib; + fib_table_t *server_fib; dhcp_server_t *server; ip6_mfib_t *rx_fib; @@ -985,9 +985,10 @@ format_dhcp6_proxy_server (u8 * s, va_list * args) vec_foreach(server, proxy->dhcp_servers) { - server_fib = ip6_fib_get(server->server_fib_index); + server_fib = fib_table_get(server->server_fib_index, + FIB_PROTOCOL_IP6); s = format (s, "%u,%U ", - server_fib->table_id, + server_fib->ft_table_id, format_ip46_address, &server->dhcp_server, IP46_TYPE_ANY); } diff --git a/src/vnet/fib/fib.h b/src/vnet/fib/fib.h index 7cf1d136..ec97c565 100644 --- a/src/vnet/fib/fib.h +++ b/src/vnet/fib/fib.h @@ -646,7 +646,5 @@ #include #include -#include -#include #endif diff --git a/src/vnet/fib/ip6_fib.c b/src/vnet/fib/ip6_fib.c index 4a24c212..527f9114 100644 --- a/src/vnet/fib/ip6_fib.c +++ b/src/vnet/fib/ip6_fib.c @@ -341,45 +341,6 @@ ip6_fib_table_entry_insert (u32 fib_index, compute_prefix_lengths_in_search_order (table); } -u32 -ip6_fib_table_fwding_lookup (ip6_main_t * im, - u32 fib_index, - const ip6_address_t * dst) -{ - const ip6_fib_table_instance_t *table; - int i, len; - int rv; - BVT(clib_bihash_kv) kv, value; - u64 fib; - - table = &ip6_main.ip6_table[IP6_FIB_TABLE_FWDING]; - len = vec_len (table->prefix_lengths_in_search_order); - - kv.key[0] = dst->as_u64[0]; - kv.key[1] = dst->as_u64[1]; - fib = ((u64)((fib_index))<<32); - - for (i = 0; i < len; i++) - { - int dst_address_length = table->prefix_lengths_in_search_order[i]; - ip6_address_t * mask = &ip6_main.fib_masks[dst_address_length]; - - ASSERT(dst_address_length >= 0 && dst_address_length <= 128); - //As lengths are decreasing, masks are increasingly specific. - kv.key[0] &= mask->as_u64[0]; - kv.key[1] &= mask->as_u64[1]; - kv.key[2] = fib | dst_address_length; - - rv = BV(clib_bihash_search_inline_2)(&table->ip6_hash, &kv, &value); - if (rv == 0) - return value.value; - } - - /* default route is always present */ - ASSERT(0); - return 0; -} - u32 ip6_fib_table_fwding_lookup_with_if_index (ip6_main_t * im, u32 sw_if_index, const ip6_address_t * dst) diff --git a/src/vnet/fib/ip6_fib.h b/src/vnet/fib/ip6_fib.h index 2bf8ef78..9789da4f 100644 --- a/src/vnet/fib/ip6_fib.h +++ b/src/vnet/fib/ip6_fib.h @@ -53,9 +53,6 @@ extern void ip6_fib_table_fwding_dpo_remove(u32 fib_index, u32 ip6_fib_table_fwding_lookup_with_if_index(ip6_main_t * im, u32 sw_if_index, const ip6_address_t * dst); -u32 ip6_fib_table_fwding_lookup(ip6_main_t * im, - u32 fib_index, - const ip6_address_t * dst); /** * @brief Walk all entries in a FIB table @@ -66,6 +63,45 @@ extern void ip6_fib_table_walk(u32 fib_index, fib_table_walk_fn_t fn, void *ctx); +always_inline u32 +ip6_fib_table_fwding_lookup (ip6_main_t * im, + u32 fib_index, + const ip6_address_t * dst) +{ + const ip6_fib_table_instance_t *table; + int i, len; + int rv; + BVT(clib_bihash_kv) kv, value; + u64 fib; + + table = &ip6_main.ip6_table[IP6_FIB_TABLE_FWDING]; + len = vec_len (table->prefix_lengths_in_search_order); + + kv.key[0] = dst->as_u64[0]; + kv.key[1] = dst->as_u64[1]; + fib = ((u64)((fib_index))<<32); + + for (i = 0; i < len; i++) + { + int dst_address_length = table->prefix_lengths_in_search_order[i]; + ip6_address_t * mask = &ip6_main.fib_masks[dst_address_length]; + + ASSERT(dst_address_length >= 0 && dst_address_length <= 128); + //As lengths are decreasing, masks are increasingly specific. + kv.key[0] &= mask->as_u64[0]; + kv.key[1] &= mask->as_u64[1]; + kv.key[2] = fib | dst_address_length; + + rv = BV(clib_bihash_search_inline_2)(&table->ip6_hash, &kv, &value); + if (rv == 0) + return value.value; + } + + /* default route is always present */ + ASSERT(0); + return 0; +} + /** * @brief return the DPO that the LB stacks on. */ diff --git a/src/vnet/ip/ip6_forward.c b/src/vnet/ip/ip6_forward.c index 25714e48..28c84d1c 100644 --- a/src/vnet/ip/ip6_forward.c +++ b/src/vnet/ip/ip6_forward.c @@ -444,12 +444,11 @@ ip6_sw_interface_enable_disable (u32 sw_if_index, u32 is_enable) return; } - vnet_feature_enable_disable ("ip6-unicast", "ip6-lookup", sw_if_index, - is_enable, 0, 0); - - vnet_feature_enable_disable ("ip6-multicast", "ip6-mfib-forward-lookup", - sw_if_index, is_enable, 0, 0); + vnet_feature_enable_disable ("ip6-unicast", "ip6-drop", sw_if_index, + !is_enable, 0, 0); + vnet_feature_enable_disable ("ip6-multicast", "ip6-drop", sw_if_index, + !is_enable, 0, 0); } /* get first interface address */ @@ -624,17 +623,17 @@ VNET_FEATURE_INIT (ip6_vxlan_bypass, static) = .runs_before = VNET_FEATURES ("ip6-lookup"), }; -VNET_FEATURE_INIT (ip6_lookup, static) = +VNET_FEATURE_INIT (ip6_drop, static) = { .arc_name = "ip6-unicast", - .node_name = "ip6-lookup", - .runs_before = VNET_FEATURES ("ip6-drop"), + .node_name = "ip6-drop", + .runs_before = VNET_FEATURES ("ip6-lookup"), }; -VNET_FEATURE_INIT (ip6_drop, static) = +VNET_FEATURE_INIT (ip6_lookup, static) = { .arc_name = "ip6-unicast", - .node_name = "ip6-drop", + .node_name = "ip6-lookup", .runs_before = 0, /*last feature*/ }; @@ -652,15 +651,15 @@ VNET_FEATURE_INIT (ip6_vpath_mc, static) = { .runs_before = VNET_FEATURES ("ip6-mfib-forward-lookup"), }; -VNET_FEATURE_INIT (ip6_mc_lookup, static) = { +VNET_FEATURE_INIT (ip6_drop_mc, static) = { .arc_name = "ip6-multicast", - .node_name = "ip6-mfib-forward-lookup", - .runs_before = VNET_FEATURES ("ip6-drop"), + .node_name = "ip6-drop", + .runs_before = VNET_FEATURES ("ip6-mfib-forward-lookup"), }; -VNET_FEATURE_INIT (ip6_drop_mc, static) = { +VNET_FEATURE_INIT (ip6_mc_lookup, static) = { .arc_name = "ip6-multicast", - .node_name = "ip6-drop", + .node_name = "ip6-mfib-forward-lookup", .runs_before = 0, /* last feature */ }; @@ -699,9 +698,6 @@ ip6_sw_interface_add_del (vnet_main_t * vnm, u32 sw_if_index, u32 is_add) vnet_feature_enable_disable ("ip6-multicast", "ip6-drop", sw_if_index, is_add, 0, 0); - vnet_feature_enable_disable ("ip6-output", "interface-output", sw_if_index, - is_add, 0, 0); - return /* no error */ 0; } diff --git a/src/vnet/vxlan-gpe/vxlan_gpe.c b/src/vnet/vxlan-gpe/vxlan_gpe.c index 2cba596f..1e674085 100644 --- a/src/vnet/vxlan-gpe/vxlan_gpe.c +++ b/src/vnet/vxlan-gpe/vxlan_gpe.c @@ -490,9 +490,9 @@ vxlan_gpe_add_del_tunnel_command_fn (vlib_main_t * vm, else if (unformat (line_input, "encap-vrf-id %d", &tmp)) { if (ipv6_set) - encap_fib_index = ip6_fib_index_from_table_id (tmp); + encap_fib_index = fib_table_find (FIB_PROTOCOL_IP6, tmp); else - encap_fib_index = ip4_fib_index_from_table_id (tmp); + encap_fib_index = fib_table_find (FIB_PROTOCOL_IP4, tmp); if (encap_fib_index == ~0) { @@ -503,9 +503,9 @@ vxlan_gpe_add_del_tunnel_command_fn (vlib_main_t * vm, else if (unformat (line_input, "decap-vrf-id %d", &tmp)) { if (ipv6_set) - decap_fib_index = ip6_fib_index_from_table_id (tmp); + decap_fib_index = fib_table_find (FIB_PROTOCOL_IP6, tmp); else - decap_fib_index = ip4_fib_index_from_table_id (tmp); + decap_fib_index = fib_table_find (FIB_PROTOCOL_IP4, tmp); if (decap_fib_index == ~0) { diff --git a/src/vpp/api/api.c b/src/vpp/api/api.c index 60eb5331..c1dcfb03 100644 --- a/src/vpp/api/api.c +++ b/src/vpp/api/api.c @@ -81,8 +81,6 @@ #include #include #include -#include -#include #include #include #include -- cgit 1.2.3-korg From 8a0a0ae60b8dd9da7cf53c895e85dc6daf67143d Mon Sep 17 00:00:00 2001 From: Hongjun Ni Date: Sat, 27 May 2017 20:23:09 +0800 Subject: Rework vxlan-gpe to support FIB 2.0 and bypass mode Change-Id: I0324f945bdb4dd3b19151be6f3ce24a47a000104 Signed-off-by: Hongjun Ni --- src/plugins/ioam/lib-vxlan-gpe/ioam_transit.c | 2 +- src/vnet/fib/fib_node.h | 4 +- src/vnet/mfib/mfib_types.h | 2 + src/vnet/udp/udp.h | 4 +- src/vnet/vxlan-gpe/decap.c | 479 +++++++++++++++++- src/vnet/vxlan-gpe/vxlan_gpe.api | 19 + src/vnet/vxlan-gpe/vxlan_gpe.c | 688 +++++++++++++++++++++++--- src/vnet/vxlan-gpe/vxlan_gpe.h | 38 ++ src/vnet/vxlan-gpe/vxlan_gpe_api.c | 20 + test/test_vxlan_gpe.py | 313 ++++++++++++ test/vpp_papi_provider.py | 35 ++ 11 files changed, 1517 insertions(+), 87 deletions(-) create mode 100644 test/test_vxlan_gpe.py (limited to 'src/vnet/vxlan-gpe') diff --git a/src/plugins/ioam/lib-vxlan-gpe/ioam_transit.c b/src/plugins/ioam/lib-vxlan-gpe/ioam_transit.c index d90cd5e4..60eabc22 100644 --- a/src/plugins/ioam/lib-vxlan-gpe/ioam_transit.c +++ b/src/plugins/ioam/lib-vxlan-gpe/ioam_transit.c @@ -107,7 +107,7 @@ vxlan_gpe_transit_ioam (vlib_main_t * vm, if (PREDICT_FALSE ((ip0->protocol == IP_PROTOCOL_UDP) && (clib_net_to_host_u16 (udp_hdr0->dst_port) == - UDP_DST_PORT_vxlan_gpe))) + UDP_DST_PORT_VXLAN_GPE))) { /* Check the iOAM header */ diff --git a/src/vnet/fib/fib_node.h b/src/vnet/fib/fib_node.h index 496929ad..ec517e15 100644 --- a/src/vnet/fib/fib_node.h +++ b/src/vnet/fib/fib_node.h @@ -42,6 +42,7 @@ typedef enum fib_node_type_t_ { FIB_NODE_TYPE_GRE_TUNNEL, FIB_NODE_TYPE_VXLAN_TUNNEL, FIB_NODE_TYPE_MAP_E, + FIB_NODE_TYPE_VXLAN_GPE_TUNNEL, /** * Marker. New types before this one. leave the test last. */ @@ -65,6 +66,7 @@ typedef enum fib_node_type_t_ { [FIB_NODE_TYPE_GRE_TUNNEL] = "gre-tunnel", \ [FIB_NODE_TYPE_VXLAN_TUNNEL] = "vxlan-tunnel", \ [FIB_NODE_TYPE_MAP_E] = "map-e", \ + [FIB_NODE_TYPE_VXLAN_GPE_TUNNEL] = "vxlan-gpe-tunnel", \ } /** @@ -272,7 +274,7 @@ typedef struct fib_node_vft_t_ { /** * An node in the FIB graph * - * Objects in the FIB form a graph. + * Objects in the FIB form a graph. */ typedef struct fib_node_t_ { #if CLIB_DEBUG > 0 diff --git a/src/vnet/mfib/mfib_types.h b/src/vnet/mfib/mfib_types.h index e278af7f..863fad16 100644 --- a/src/vnet/mfib/mfib_types.h +++ b/src/vnet/mfib/mfib_types.h @@ -168,6 +168,7 @@ typedef enum mfib_source_t_ MFIB_SOURCE_SRv6, MFIB_SOURCE_DEFAULT_ROUTE, MFIB_SOURCE_GTPU, + MFIB_SOURCE_VXLAN_GPE, } mfib_source_t; #define MFIB_SOURCE_NAMES { \ @@ -179,6 +180,7 @@ typedef enum mfib_source_t_ [MFIB_SOURCE_SRv6] = "SRv6", \ [MFIB_SOURCE_DEFAULT_ROUTE] = "Default Route", \ [MFIB_SOURCE_GTPU] = "GTPU", \ + [MFIB_SOURCE_VXLAN_GPE] = "VXLAN-GPE", \ } /** diff --git a/src/vnet/udp/udp.h b/src/vnet/udp/udp.h index 5b63f44d..a9a62c27 100644 --- a/src/vnet/udp/udp.h +++ b/src/vnet/udp/udp.h @@ -89,7 +89,7 @@ _ (4342, lisp_cp) \ _ (4739, ipfix) \ _ (4789, vxlan) \ _ (4789, vxlan6) \ -_ (4790, vxlan_gpe) \ +_ (4790, VXLAN_GPE) \ _ (6633, vpath_3) @@ -101,7 +101,7 @@ _ (3784, bfd6) \ _ (3785, bfd_echo6) \ _ (4341, lisp_gpe6) \ _ (4342, lisp_cp6) \ -_ (4790, vxlan6_gpe) \ +_ (4790, VXLAN6_GPE) \ _ (6633, vpath6_3) typedef enum diff --git a/src/vnet/vxlan-gpe/decap.c b/src/vnet/vxlan-gpe/decap.c index d4fe4231..075b0f51 100644 --- a/src/vnet/vxlan-gpe/decap.c +++ b/src/vnet/vxlan-gpe/decap.c @@ -108,8 +108,8 @@ vxlan_gpe_input (vlib_main_t * vm, u8 is_ip4) { u32 n_left_from, next_index, *from, *to_next; - vxlan_gpe_main_t * ngm = &vxlan_gpe_main; - vnet_main_t * vnm = ngm->vnet_main; + vxlan_gpe_main_t * nngm = &vxlan_gpe_main; + vnet_main_t * vnm = nngm->vnet_main; vnet_interface_main_t * im = &vnm->interface_main; u32 last_tunnel_index = ~0; vxlan4_gpe_tunnel_key_t last_key4; @@ -213,11 +213,11 @@ vxlan_gpe_input (vlib_main_t * vm, { next0 = (iuvn4_0->vxlan.protocol < VXLAN_GPE_PROTOCOL_MAX)? - ngm->decap_next_node_list[iuvn4_0->vxlan.protocol]: \ + nngm->decap_next_node_list[iuvn4_0->vxlan.protocol]: \ VXLAN_GPE_INPUT_NEXT_DROP; next1 = (iuvn4_1->vxlan.protocol < VXLAN_GPE_PROTOCOL_MAX)? - ngm->decap_next_node_list[iuvn4_1->vxlan.protocol]: \ + nngm->decap_next_node_list[iuvn4_1->vxlan.protocol]: \ VXLAN_GPE_INPUT_NEXT_DROP; key4_0.local = iuvn4_0->ip4.dst_address.as_u32; @@ -260,7 +260,7 @@ vxlan_gpe_input (vlib_main_t * vm, if (PREDICT_FALSE((key4_0.as_u64[0] != last_key4.as_u64[0]) || (key4_0.as_u64[1] != last_key4.as_u64[1]))) { - p0 = hash_get_mem(ngm->vxlan4_gpe_tunnel_by_key, &key4_0); + p0 = hash_get_mem(nngm->vxlan4_gpe_tunnel_by_key, &key4_0); if (p0 == 0) { @@ -279,11 +279,11 @@ vxlan_gpe_input (vlib_main_t * vm, { next0 = (iuvn6_0->vxlan.protocol < VXLAN_GPE_PROTOCOL_MAX)? - ngm->decap_next_node_list[iuvn6_0->vxlan.protocol]: \ + nngm->decap_next_node_list[iuvn6_0->vxlan.protocol]: \ VXLAN_GPE_INPUT_NEXT_DROP; next1 = (iuvn6_1->vxlan.protocol < VXLAN_GPE_PROTOCOL_MAX)? - ngm->decap_next_node_list[iuvn6_1->vxlan.protocol]: \ + nngm->decap_next_node_list[iuvn6_1->vxlan.protocol]: \ VXLAN_GPE_INPUT_NEXT_DROP; key6_0.local.as_u64[0] = iuvn6_0->ip6.dst_address.as_u64[0]; @@ -302,7 +302,7 @@ vxlan_gpe_input (vlib_main_t * vm, /* Processing for key6_0 */ if (PREDICT_FALSE(memcmp (&key6_0, &last_key6, sizeof(last_key6)) != 0)) { - p0 = hash_get_mem(ngm->vxlan6_gpe_tunnel_by_key, &key6_0); + p0 = hash_get_mem(nngm->vxlan6_gpe_tunnel_by_key, &key6_0); if (p0 == 0) { @@ -317,7 +317,7 @@ vxlan_gpe_input (vlib_main_t * vm, tunnel_index0 = last_tunnel_index; } - t0 = pool_elt_at_index(ngm->tunnels, tunnel_index0); + t0 = pool_elt_at_index(nngm->tunnels, tunnel_index0); sw_if_index0 = t0->sw_if_index; @@ -366,7 +366,7 @@ vxlan_gpe_input (vlib_main_t * vm, (key4_1.as_u64[0] != last_key4.as_u64[0]) || (key4_1.as_u64[1] != last_key4.as_u64[1]))) { - p1 = hash_get_mem(ngm->vxlan4_gpe_tunnel_by_key, &key4_1); + p1 = hash_get_mem(nngm->vxlan4_gpe_tunnel_by_key, &key4_1); if (p1 == 0) { @@ -386,7 +386,7 @@ vxlan_gpe_input (vlib_main_t * vm, /* Processing for key6_1 */ if (PREDICT_FALSE(memcmp (&key6_1, &last_key6, sizeof(last_key6)) != 0)) { - p1 = hash_get_mem(ngm->vxlan6_gpe_tunnel_by_key, &key6_1); + p1 = hash_get_mem(nngm->vxlan6_gpe_tunnel_by_key, &key6_1); if (p1 == 0) { @@ -401,7 +401,7 @@ vxlan_gpe_input (vlib_main_t * vm, tunnel_index1 = last_tunnel_index; } - t1 = pool_elt_at_index(ngm->tunnels, tunnel_index1); + t1 = pool_elt_at_index(nngm->tunnels, tunnel_index1); sw_if_index1 = t1->sw_if_index; len1 = vlib_buffer_length_in_chain (vm, b1); @@ -502,7 +502,7 @@ vxlan_gpe_input (vlib_main_t * vm, { next0 = (iuvn4_0->vxlan.protocol < VXLAN_GPE_PROTOCOL_MAX)? - ngm->decap_next_node_list[iuvn4_0->vxlan.protocol]: \ + nngm->decap_next_node_list[iuvn4_0->vxlan.protocol]: \ VXLAN_GPE_INPUT_NEXT_DROP; key4_0.local = iuvn4_0->ip4.dst_address.as_u32; @@ -515,7 +515,7 @@ vxlan_gpe_input (vlib_main_t * vm, (key4_0.as_u64[0] != last_key4.as_u64[0]) || (key4_0.as_u64[1] != last_key4.as_u64[1]))) { - p0 = hash_get_mem(ngm->vxlan4_gpe_tunnel_by_key, &key4_0); + p0 = hash_get_mem(nngm->vxlan4_gpe_tunnel_by_key, &key4_0); if (p0 == 0) { @@ -534,7 +534,7 @@ vxlan_gpe_input (vlib_main_t * vm, { next0 = (iuvn6_0->vxlan.protocol < VXLAN_GPE_PROTOCOL_MAX)? - ngm->decap_next_node_list[iuvn6_0->vxlan.protocol]: \ + nngm->decap_next_node_list[iuvn6_0->vxlan.protocol]: \ VXLAN_GPE_INPUT_NEXT_DROP; key6_0.local.as_u64[0] = iuvn6_0->ip6.dst_address.as_u64[0]; @@ -546,7 +546,7 @@ vxlan_gpe_input (vlib_main_t * vm, /* Processing for key6_0 */ if (PREDICT_FALSE(memcmp (&key6_0, &last_key6, sizeof(last_key6)) != 0)) { - p0 = hash_get_mem(ngm->vxlan6_gpe_tunnel_by_key, &key6_0); + p0 = hash_get_mem(nngm->vxlan6_gpe_tunnel_by_key, &key6_0); if (p0 == 0) { @@ -561,7 +561,7 @@ vxlan_gpe_input (vlib_main_t * vm, tunnel_index0 = last_tunnel_index; } - t0 = pool_elt_at_index(ngm->tunnels, tunnel_index0); + t0 = pool_elt_at_index(nngm->tunnels, tunnel_index0); sw_if_index0 = t0->sw_if_index; @@ -731,3 +731,448 @@ VLIB_REGISTER_NODE (vxlan6_gpe_input_node) = { }; VLIB_NODE_FUNCTION_MULTIARCH (vxlan6_gpe_input_node, vxlan6_gpe_input); +typedef enum { + IP_VXLAN_BYPASS_NEXT_DROP, + IP_VXLAN_BYPASS_NEXT_VXLAN, + IP_VXLAN_BYPASS_N_NEXT, +} ip_vxan_bypass_next_t; + +always_inline uword +ip_vxlan_gpe_bypass_inline (vlib_main_t * vm, + vlib_node_runtime_t * node, + vlib_frame_t * frame, + u32 is_ip4) +{ + vxlan_gpe_main_t * ngm = &vxlan_gpe_main; + u32 * from, * to_next, n_left_from, n_left_to_next, next_index; + vlib_node_runtime_t * error_node = vlib_node_get_runtime (vm, ip4_input_node.index); + ip4_address_t addr4; /* last IPv4 address matching a local VTEP address */ + ip6_address_t addr6; /* last IPv6 address matching a local VTEP address */ + + 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); + + if (is_ip4) addr4.data_u32 = ~0; + else ip6_address_set_zero (&addr6); + + 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 * b0, * b1; + ip4_header_t * ip40, * ip41; + ip6_header_t * ip60, * ip61; + udp_header_t * udp0, * udp1; + u32 bi0, ip_len0, udp_len0, flags0, next0; + u32 bi1, ip_len1, udp_len1, flags1, next1; + i32 len_diff0, len_diff1; + u8 error0, good_udp0, proto0; + u8 error1, good_udp1, proto1; + + /* Prefetch next iteration. */ + { + vlib_buffer_t * p2, * p3; + + p2 = vlib_get_buffer (vm, from[2]); + p3 = vlib_get_buffer (vm, from[3]); + + vlib_prefetch_buffer_header (p2, LOAD); + vlib_prefetch_buffer_header (p3, LOAD); + + CLIB_PREFETCH (p2->data, 2*CLIB_CACHE_LINE_BYTES, LOAD); + CLIB_PREFETCH (p3->data, 2*CLIB_CACHE_LINE_BYTES, LOAD); + } + + bi0 = to_next[0] = from[0]; + bi1 = to_next[1] = from[1]; + from += 2; + n_left_from -= 2; + to_next += 2; + n_left_to_next -= 2; + + b0 = vlib_get_buffer (vm, bi0); + b1 = vlib_get_buffer (vm, bi1); + if (is_ip4) + { + ip40 = vlib_buffer_get_current (b0); + ip41 = vlib_buffer_get_current (b1); + } + else + { + ip60 = vlib_buffer_get_current (b0); + ip61 = vlib_buffer_get_current (b1); + } + + /* Setup packet for next IP feature */ + vnet_feature_next(vnet_buffer(b0)->sw_if_index[VLIB_RX], &next0, b0); + vnet_feature_next(vnet_buffer(b1)->sw_if_index[VLIB_RX], &next1, b1); + + if (is_ip4) + { + proto0 = ip40->protocol; + proto1 = ip41->protocol; + } + else + { + proto0 = ip60->protocol; + proto1 = ip61->protocol; + } + + /* Process packet 0 */ + if (proto0 != IP_PROTOCOL_UDP) + goto exit0; /* not UDP packet */ + + if (is_ip4) + udp0 = ip4_next_header (ip40); + else + udp0 = ip6_next_header (ip60); + + if (udp0->dst_port != clib_host_to_net_u16 (UDP_DST_PORT_VXLAN_GPE)) + goto exit0; /* not VXLAN packet */ + + /* Validate DIP against VTEPs*/ + if (is_ip4) + { + if (addr4.as_u32 != ip40->dst_address.as_u32) + { + if (!hash_get (ngm->vtep4, ip40->dst_address.as_u32)) + goto exit0; /* no local VTEP for VXLAN packet */ + addr4 = ip40->dst_address; + } + } + else + { + if (!ip6_address_is_equal (&addr6, &ip60->dst_address)) + { + if (!hash_get_mem (ngm->vtep6, &ip60->dst_address)) + goto exit0; /* no local VTEP for VXLAN packet */ + addr6 = ip60->dst_address; + } + } + + flags0 = b0->flags; + good_udp0 = (flags0 & IP_BUFFER_L4_CHECKSUM_CORRECT) != 0; + + /* Don't verify UDP checksum for packets with explicit zero checksum. */ + good_udp0 |= udp0->checksum == 0; + + /* Verify UDP length */ + if (is_ip4) + ip_len0 = clib_net_to_host_u16 (ip40->length); + else + ip_len0 = clib_net_to_host_u16 (ip60->payload_length); + udp_len0 = clib_net_to_host_u16 (udp0->length); + len_diff0 = ip_len0 - udp_len0; + + /* Verify UDP checksum */ + if (PREDICT_FALSE (!good_udp0)) + { + if ((flags0 & IP_BUFFER_L4_CHECKSUM_COMPUTED) == 0) + { + if (is_ip4) + flags0 = ip4_tcp_udp_validate_checksum (vm, b0); + else + flags0 = ip6_tcp_udp_icmp_validate_checksum (vm, b0); + good_udp0 = + (flags0 & IP_BUFFER_L4_CHECKSUM_CORRECT) != 0; + } + } + + if (is_ip4) + { + error0 = good_udp0 ? 0 : IP4_ERROR_UDP_CHECKSUM; + error0 = (len_diff0 >= 0) ? error0 : IP4_ERROR_UDP_LENGTH; + } + else + { + error0 = good_udp0 ? 0 : IP6_ERROR_UDP_CHECKSUM; + error0 = (len_diff0 >= 0) ? error0 : IP6_ERROR_UDP_LENGTH; + } + + next0 = error0 ? + IP_VXLAN_BYPASS_NEXT_DROP : IP_VXLAN_BYPASS_NEXT_VXLAN; + b0->error = error0 ? error_node->errors[error0] : 0; + + /* vxlan_gpe-input node expect current at VXLAN header */ + if (is_ip4) + vlib_buffer_advance (b0, sizeof(ip4_header_t)+sizeof(udp_header_t)); + else + vlib_buffer_advance (b0, sizeof(ip6_header_t)+sizeof(udp_header_t)); + + exit0: + /* Process packet 1 */ + if (proto1 != IP_PROTOCOL_UDP) + goto exit1; /* not UDP packet */ + + if (is_ip4) + udp1 = ip4_next_header (ip41); + else + udp1 = ip6_next_header (ip61); + + if (udp1->dst_port != clib_host_to_net_u16 (UDP_DST_PORT_VXLAN_GPE)) + goto exit1; /* not VXLAN packet */ + + /* Validate DIP against VTEPs*/ + if (is_ip4) + { + if (addr4.as_u32 != ip41->dst_address.as_u32) + { + if (!hash_get (ngm->vtep4, ip41->dst_address.as_u32)) + goto exit1; /* no local VTEP for VXLAN packet */ + addr4 = ip41->dst_address; + } + } + else + { + if (!ip6_address_is_equal (&addr6, &ip61->dst_address)) + { + if (!hash_get_mem (ngm->vtep6, &ip61->dst_address)) + goto exit1; /* no local VTEP for VXLAN packet */ + addr6 = ip61->dst_address; + } + } + + flags1 = b1->flags; + good_udp1 = (flags1 & IP_BUFFER_L4_CHECKSUM_CORRECT) != 0; + + /* Don't verify UDP checksum for packets with explicit zero checksum. */ + good_udp1 |= udp1->checksum == 0; + + /* Verify UDP length */ + if (is_ip4) + ip_len1 = clib_net_to_host_u16 (ip41->length); + else + ip_len1 = clib_net_to_host_u16 (ip61->payload_length); + udp_len1 = clib_net_to_host_u16 (udp1->length); + len_diff1 = ip_len1 - udp_len1; + + /* Verify UDP checksum */ + if (PREDICT_FALSE (!good_udp1)) + { + if ((flags1 & IP_BUFFER_L4_CHECKSUM_COMPUTED) == 0) + { + if (is_ip4) + flags1 = ip4_tcp_udp_validate_checksum (vm, b1); + else + flags1 = ip6_tcp_udp_icmp_validate_checksum (vm, b1); + good_udp1 = + (flags1 & IP_BUFFER_L4_CHECKSUM_CORRECT) != 0; + } + } + + if (is_ip4) + { + error1 = good_udp1 ? 0 : IP4_ERROR_UDP_CHECKSUM; + error1 = (len_diff1 >= 0) ? error1 : IP4_ERROR_UDP_LENGTH; + } + else + { + error1 = good_udp1 ? 0 : IP6_ERROR_UDP_CHECKSUM; + error1 = (len_diff1 >= 0) ? error1 : IP6_ERROR_UDP_LENGTH; + } + + next1 = error1 ? + IP_VXLAN_BYPASS_NEXT_DROP : IP_VXLAN_BYPASS_NEXT_VXLAN; + b1->error = error1 ? error_node->errors[error1] : 0; + + /* vxlan_gpe-input node expect current at VXLAN header */ + if (is_ip4) + vlib_buffer_advance (b1, sizeof(ip4_header_t)+sizeof(udp_header_t)); + else + vlib_buffer_advance (b1, sizeof(ip6_header_t)+sizeof(udp_header_t)); + + exit1: + vlib_validate_buffer_enqueue_x2 (vm, node, next_index, + to_next, n_left_to_next, + bi0, bi1, next0, next1); + } + + while (n_left_from > 0 && n_left_to_next > 0) + { + vlib_buffer_t * b0; + ip4_header_t * ip40; + ip6_header_t * ip60; + udp_header_t * udp0; + u32 bi0, ip_len0, udp_len0, flags0, next0; + i32 len_diff0; + u8 error0, good_udp0, proto0; + + bi0 = to_next[0] = from[0]; + from += 1; + n_left_from -= 1; + to_next += 1; + n_left_to_next -= 1; + + b0 = vlib_get_buffer (vm, bi0); + if (is_ip4) + ip40 = vlib_buffer_get_current (b0); + else + ip60 = vlib_buffer_get_current (b0); + + /* Setup packet for next IP feature */ + vnet_feature_next(vnet_buffer(b0)->sw_if_index[VLIB_RX], &next0, b0); + + if (is_ip4) + proto0 = ip40->protocol; + else + proto0 = ip60->protocol; + + if (proto0 != IP_PROTOCOL_UDP) + goto exit; /* not UDP packet */ + + if (is_ip4) + udp0 = ip4_next_header (ip40); + else + udp0 = ip6_next_header (ip60); + + if (udp0->dst_port != clib_host_to_net_u16 (UDP_DST_PORT_VXLAN_GPE)) + goto exit; /* not VXLAN packet */ + + /* Validate DIP against VTEPs*/ + if (is_ip4) + { + if (addr4.as_u32 != ip40->dst_address.as_u32) + { + if (!hash_get (ngm->vtep4, ip40->dst_address.as_u32)) + goto exit; /* no local VTEP for VXLAN packet */ + addr4 = ip40->dst_address; + } + } + else + { + if (!ip6_address_is_equal (&addr6, &ip60->dst_address)) + { + if (!hash_get_mem (ngm->vtep6, &ip60->dst_address)) + goto exit; /* no local VTEP for VXLAN packet */ + addr6 = ip60->dst_address; + } + } + + flags0 = b0->flags; + good_udp0 = (flags0 & IP_BUFFER_L4_CHECKSUM_CORRECT) != 0; + + /* Don't verify UDP checksum for packets with explicit zero checksum. */ + good_udp0 |= udp0->checksum == 0; + + /* Verify UDP length */ + if (is_ip4) + ip_len0 = clib_net_to_host_u16 (ip40->length); + else + ip_len0 = clib_net_to_host_u16 (ip60->payload_length); + udp_len0 = clib_net_to_host_u16 (udp0->length); + len_diff0 = ip_len0 - udp_len0; + + /* Verify UDP checksum */ + if (PREDICT_FALSE (!good_udp0)) + { + if ((flags0 & IP_BUFFER_L4_CHECKSUM_COMPUTED) == 0) + { + if (is_ip4) + flags0 = ip4_tcp_udp_validate_checksum (vm, b0); + else + flags0 = ip6_tcp_udp_icmp_validate_checksum (vm, b0); + good_udp0 = + (flags0 & IP_BUFFER_L4_CHECKSUM_CORRECT) != 0; + } + } + + if (is_ip4) + { + error0 = good_udp0 ? 0 : IP4_ERROR_UDP_CHECKSUM; + error0 = (len_diff0 >= 0) ? error0 : IP4_ERROR_UDP_LENGTH; + } + else + { + error0 = good_udp0 ? 0 : IP6_ERROR_UDP_CHECKSUM; + error0 = (len_diff0 >= 0) ? error0 : IP6_ERROR_UDP_LENGTH; + } + + next0 = error0 ? + IP_VXLAN_BYPASS_NEXT_DROP : IP_VXLAN_BYPASS_NEXT_VXLAN; + b0->error = error0 ? error_node->errors[error0] : 0; + + /* vxlan_gpe-input node expect current at VXLAN header */ + if (is_ip4) + vlib_buffer_advance (b0, sizeof(ip4_header_t)+sizeof(udp_header_t)); + else + vlib_buffer_advance (b0, sizeof(ip6_header_t)+sizeof(udp_header_t)); + + exit: + vlib_validate_buffer_enqueue_x1 (vm, node, next_index, + to_next, n_left_to_next, + bi0, next0); + } + + vlib_put_next_frame (vm, node, next_index, n_left_to_next); + } + + return frame->n_vectors; +} + +static uword +ip4_vxlan_gpe_bypass (vlib_main_t * vm, + vlib_node_runtime_t * node, + vlib_frame_t * frame) +{ + return ip_vxlan_gpe_bypass_inline (vm, node, frame, /* is_ip4 */ 1); +} + +VLIB_REGISTER_NODE (ip4_vxlan_gpe_bypass_node) = { + .function = ip4_vxlan_gpe_bypass, + .name = "ip4-vxlan-gpe-bypass", + .vector_size = sizeof (u32), + + .n_next_nodes = IP_VXLAN_BYPASS_N_NEXT, + .next_nodes = { + [IP_VXLAN_BYPASS_NEXT_DROP] = "error-drop", + [IP_VXLAN_BYPASS_NEXT_VXLAN] = "vxlan4-gpe-input", + }, + + .format_buffer = format_ip4_header, + .format_trace = format_ip4_forward_next_trace, +}; + +VLIB_NODE_FUNCTION_MULTIARCH (ip4_vxlan_gpe_bypass_node,ip4_vxlan_gpe_bypass) + +/* Dummy init function to get us linked in. */ +clib_error_t * ip4_vxlan_gpe_bypass_init (vlib_main_t * vm) +{ return 0; } + +VLIB_INIT_FUNCTION (ip4_vxlan_gpe_bypass_init); + +static uword +ip6_vxlan_gpe_bypass (vlib_main_t * vm, + vlib_node_runtime_t * node, + vlib_frame_t * frame) +{ + return ip_vxlan_gpe_bypass_inline (vm, node, frame, /* is_ip4 */ 0); +} + +VLIB_REGISTER_NODE (ip6_vxlan_gpe_bypass_node) = { + .function = ip6_vxlan_gpe_bypass, + .name = "ip6-vxlan-gpe-bypass", + .vector_size = sizeof (u32), + + .n_next_nodes = IP_VXLAN_BYPASS_N_NEXT, + .next_nodes = { + [IP_VXLAN_BYPASS_NEXT_DROP] = "error-drop", + [IP_VXLAN_BYPASS_NEXT_VXLAN] = "vxlan6-gpe-input", + }, + + .format_buffer = format_ip6_header, + .format_trace = format_ip6_forward_next_trace, +}; + +VLIB_NODE_FUNCTION_MULTIARCH (ip6_vxlan_gpe_bypass_node,ip6_vxlan_gpe_bypass) + +/* Dummy init function to get us linked in. */ +clib_error_t * ip6_vxlan_gpe_bypass_init (vlib_main_t * vm) +{ return 0; } + +VLIB_INIT_FUNCTION (ip6_vxlan_gpe_bypass_init); diff --git a/src/vnet/vxlan-gpe/vxlan_gpe.api b/src/vnet/vxlan-gpe/vxlan_gpe.api index 6c6973f8..41b10316 100644 --- a/src/vnet/vxlan-gpe/vxlan_gpe.api +++ b/src/vnet/vxlan-gpe/vxlan_gpe.api @@ -20,6 +20,7 @@ define vxlan_gpe_add_del_tunnel u8 is_ipv6; u8 local[16]; u8 remote[16]; + u32 mcast_sw_if_index; u32 encap_vrf_id; u32 decap_vrf_id; u8 protocol; @@ -49,11 +50,29 @@ define vxlan_gpe_tunnel_details u8 remote[16]; u32 vni; u8 protocol; + u32 mcast_sw_if_index; u32 encap_vrf_id; u32 decap_vrf_id; + u8 is_ipv6; }; +/** \brief Interface set vxlan-gpe-bypass request + @param client_index - opaque cookie to identify the sender + @param context - sender context, to match reply w/ request + @param sw_if_index - interface used to reach neighbor + @param is_ipv6 - if non-zero, enable ipv6-vxlan-bypass, else ipv4-vxlan-bypass + @param enable - if non-zero enable, else disable +*/ +autoreply define sw_interface_set_vxlan_gpe_bypass +{ + u32 client_index; + u32 context; + u32 sw_if_index; + u8 is_ipv6; + u8 enable; +}; + /* * Local Variables: * eval: (c-set-style "gnu") diff --git a/src/vnet/vxlan-gpe/vxlan_gpe.c b/src/vnet/vxlan-gpe/vxlan_gpe.c index 1e674085..3a92c88f 100644 --- a/src/vnet/vxlan-gpe/vxlan_gpe.c +++ b/src/vnet/vxlan-gpe/vxlan_gpe.c @@ -20,6 +20,27 @@ #include #include #include +#include +#include +#include +#include +#include +#include + +/** + * @file + * @brief VXLAN-GPE. + * + * VXLAN-GPE provides the features needed to allow L2 bridge domains (BDs) + * to span multiple servers. This is done by building an L2 overlay on + * top of an L3 network underlay using VXLAN-GPE tunnels. + * + * This makes it possible for servers to be co-located in the same data + * center or be separated geographically as long as they are reachable + * through the underlay L3 network. + * + * You can refer to this kind of L2 overlay bridge domain as a VXLAN-GPE segment. + */ vxlan_gpe_main_t vxlan_gpe_main; @@ -35,10 +56,10 @@ vxlan_gpe_main_t vxlan_gpe_main; u8 * format_vxlan_gpe_tunnel (u8 * s, va_list * args) { vxlan_gpe_tunnel_t * t = va_arg (*args, vxlan_gpe_tunnel_t *); - vxlan_gpe_main_t * gm = &vxlan_gpe_main; + vxlan_gpe_main_t * ngm = &vxlan_gpe_main; s = format (s, "[%d] local: %U remote: %U ", - t - gm->tunnels, + t - ngm->tunnels, format_ip46_address, &t->local, IP46_TYPE_ANY, format_ip46_address, &t->remote, IP46_TYPE_ANY); @@ -105,10 +126,9 @@ static uword dummy_interface_tx (vlib_main_t * vm, static clib_error_t * vxlan_gpe_interface_admin_up_down (vnet_main_t * vnm, u32 hw_if_index, u32 flags) { - if (flags & VNET_SW_INTERFACE_FLAG_ADMIN_UP) - vnet_hw_interface_set_flags (vnm, hw_if_index, VNET_HW_INTERFACE_FLAG_LINK_UP); - else - vnet_hw_interface_set_flags (vnm, hw_if_index, 0); + u32 hw_flags = (flags & VNET_SW_INTERFACE_FLAG_ADMIN_UP) ? + VNET_HW_INTERFACE_FLAG_LINK_UP : 0; + vnet_hw_interface_set_flags (vnm, hw_if_index, hw_flags); return 0; } @@ -142,13 +162,84 @@ VNET_HW_INTERFACE_CLASS (vxlan_gpe_hw_class) = { .name = "VXLAN_GPE", .format_header = format_vxlan_gpe_header_with_length, .build_rewrite = default_build_rewrite, - .flags = VNET_HW_INTERFACE_CLASS_FLAG_P2P, }; +static void +vxlan_gpe_tunnel_restack_dpo(vxlan_gpe_tunnel_t * t) +{ + dpo_id_t dpo = DPO_INVALID; + u32 encap_index = vxlan_gpe_encap_node.index; + fib_forward_chain_type_t forw_type = ip46_address_is_ip4(&t->remote) ? + FIB_FORW_CHAIN_TYPE_UNICAST_IP4 : FIB_FORW_CHAIN_TYPE_UNICAST_IP6; + + fib_entry_contribute_forwarding (t->fib_entry_index, forw_type, &dpo); + dpo_stack_from_node (encap_index, &t->next_dpo, &dpo); + dpo_reset(&dpo); +} + +static vxlan_gpe_tunnel_t * +vxlan_gpe_tunnel_from_fib_node (fib_node_t *node) +{ +#if (CLIB_DEBUG > 0) + ASSERT(FIB_NODE_TYPE_VXLAN_GPE_TUNNEL == node->fn_type); +#endif + return ((vxlan_gpe_tunnel_t*) (((char*)node) - + STRUCT_OFFSET_OF(vxlan_gpe_tunnel_t, node))); +} + +/** + * Function definition to backwalk a FIB node - + * Here we will restack the new dpo of VXLAN_GPE DIP to encap node. + */ +static fib_node_back_walk_rc_t +vxlan_gpe_tunnel_back_walk (fib_node_t *node, + fib_node_back_walk_ctx_t *ctx) +{ + vxlan_gpe_tunnel_restack_dpo(vxlan_gpe_tunnel_from_fib_node(node)); + return (FIB_NODE_BACK_WALK_CONTINUE); +} + +/** + * Function definition to get a FIB node from its index + */ +static fib_node_t* +vxlan_gpe_tunnel_fib_node_get (fib_node_index_t index) +{ + vxlan_gpe_tunnel_t * t; + vxlan_gpe_main_t * ngm = &vxlan_gpe_main; + + t = pool_elt_at_index(ngm->tunnels, index); + + return (&t->node); +} + +/** + * Function definition to inform the FIB node that its last lock has gone. + */ +static void +vxlan_gpe_tunnel_last_lock_gone (fib_node_t *node) +{ + /* + * The VXLAN_GPE tunnel is a root of the graph. As such + * it never has children and thus is never locked. + */ + ASSERT(0); +} + +/* + * Virtual function table registered by VXLAN_GPE tunnels + * for participation in the FIB object graph. + */ +const static fib_node_vft_t vxlan_gpe_vft = { + .fnv_get = vxlan_gpe_tunnel_fib_node_get, + .fnv_last_lock = vxlan_gpe_tunnel_last_lock_gone, + .fnv_back_walk = vxlan_gpe_tunnel_back_walk, +}; #define foreach_gpe_copy_field \ _(vni) \ -_(protocol) \ +_(protocol) \ +_(mcast_sw_if_index) \ _(encap_fib_index) \ _(decap_fib_index) @@ -173,7 +264,7 @@ _(decap_fib_index) * @return rc * */ -int vxlan4_gpe_rewrite (vxlan_gpe_tunnel_t * t, u32 extension_size, +int vxlan4_gpe_rewrite (vxlan_gpe_tunnel_t * t, u32 extension_size, u8 protocol_override, uword encap_next_node) { u8 *rw = 0; @@ -201,7 +292,7 @@ int vxlan4_gpe_rewrite (vxlan_gpe_tunnel_t * t, u32 extension_size, /* UDP header, randomize src port on something, maybe? */ h0->udp.src_port = clib_host_to_net_u16 (4790); - h0->udp.dst_port = clib_host_to_net_u16 (UDP_DST_PORT_vxlan_gpe); + h0->udp.dst_port = clib_host_to_net_u16 (UDP_DST_PORT_VXLAN_GPE); /* VXLAN header. Are we having fun yet? */ h0->vxlan.flags = VXLAN_GPE_FLAGS_I | VXLAN_GPE_FLAGS_P; @@ -230,7 +321,7 @@ int vxlan4_gpe_rewrite (vxlan_gpe_tunnel_t * t, u32 extension_size, * @return rc * */ -int vxlan6_gpe_rewrite (vxlan_gpe_tunnel_t * t, u32 extension_size, +int vxlan6_gpe_rewrite (vxlan_gpe_tunnel_t * t, u32 extension_size, u8 protocol_override, uword encap_next_node) { u8 *rw = 0; @@ -258,7 +349,7 @@ int vxlan6_gpe_rewrite (vxlan_gpe_tunnel_t * t, u32 extension_size, /* UDP header, randomize src port on something, maybe? */ h0->udp.src_port = clib_host_to_net_u16 (4790); - h0->udp.dst_port = clib_host_to_net_u16 (UDP_DST_PORT_vxlan_gpe); + h0->udp.dst_port = clib_host_to_net_u16 (UDP_DST_PORT_VXLAN_GPE); /* VXLAN header. Are we having fun yet? */ h0->vxlan.flags = VXLAN_GPE_FLAGS_I | VXLAN_GPE_FLAGS_P; @@ -279,6 +370,100 @@ int vxlan6_gpe_rewrite (vxlan_gpe_tunnel_t * t, u32 extension_size, return (0); } +static void +hash_set_key_copy (uword ** h, void * key, uword v) { + size_t ksz = hash_header(*h)->user; + void * copy = clib_mem_alloc (ksz); + clib_memcpy (copy, key, ksz); + hash_set_mem (*h, copy, v); +} + +static void +hash_unset_key_free (uword ** h, void * key) { + hash_pair_t * hp = hash_get_pair_mem (*h, key); + ASSERT (hp); + key = uword_to_pointer (hp->key, void *); + hash_unset_mem (*h, key); + clib_mem_free (key); +} + +static uword +vtep_addr_ref(ip46_address_t *ip) +{ + uword *vtep = ip46_address_is_ip4(ip) ? + hash_get (vxlan_gpe_main.vtep4, ip->ip4.as_u32) : + hash_get_mem (vxlan_gpe_main.vtep6, &ip->ip6); + if (vtep) + return ++(*vtep); + ip46_address_is_ip4(ip) ? + hash_set (vxlan_gpe_main.vtep4, ip->ip4.as_u32, 1) : + hash_set_key_copy (&vxlan_gpe_main.vtep6, &ip->ip6, 1); + return 1; +} + +static uword +vtep_addr_unref(ip46_address_t *ip) +{ + uword *vtep = ip46_address_is_ip4(ip) ? + hash_get (vxlan_gpe_main.vtep4, ip->ip4.as_u32) : + hash_get_mem (vxlan_gpe_main.vtep6, &ip->ip6); + ASSERT(vtep); + if (--(*vtep) != 0) + return *vtep; + ip46_address_is_ip4(ip) ? + hash_unset (vxlan_gpe_main.vtep4, ip->ip4.as_u32) : + hash_unset_key_free (&vxlan_gpe_main.vtep6, &ip->ip6); + return 0; +} + +typedef CLIB_PACKED(union { + struct { + fib_node_index_t mfib_entry_index; + adj_index_t mcast_adj_index; + }; + u64 as_u64; +}) mcast_shared_t; + +static inline mcast_shared_t +mcast_shared_get(ip46_address_t * ip) +{ + ASSERT(ip46_address_is_multicast(ip)); + uword * p = hash_get_mem (vxlan_gpe_main.mcast_shared, ip); + ASSERT(p); + return (mcast_shared_t) { .as_u64 = *p }; +} + +static inline void +mcast_shared_add(ip46_address_t *remote, + fib_node_index_t mfei, + adj_index_t ai) +{ + mcast_shared_t new_ep = { + .mcast_adj_index = ai, + .mfib_entry_index = mfei, + }; + + hash_set_key_copy (&vxlan_gpe_main.mcast_shared, remote, new_ep.as_u64); +} + +static inline void +mcast_shared_remove(ip46_address_t *remote) +{ + mcast_shared_t ep = mcast_shared_get(remote); + + adj_unlock(ep.mcast_adj_index); + mfib_table_entry_delete_index(ep.mfib_entry_index, + MFIB_SOURCE_VXLAN_GPE); + + hash_unset_key_free (&vxlan_gpe_main.mcast_shared, remote); +} + +static inline fib_protocol_t +fib_ip_proto(bool is_ip6) +{ + return (is_ip6) ? FIB_PROTOCOL_IP6 : FIB_PROTOCOL_IP4; +} + /** * @brief Add or Del a VXLAN GPE tunnel * @@ -291,9 +476,9 @@ int vxlan6_gpe_rewrite (vxlan_gpe_tunnel_t * t, u32 extension_size, int vnet_vxlan_gpe_add_del_tunnel (vnet_vxlan_gpe_add_del_tunnel_args_t *a, u32 * sw_if_indexp) { - vxlan_gpe_main_t * gm = &vxlan_gpe_main; + vxlan_gpe_main_t * ngm = &vxlan_gpe_main; vxlan_gpe_tunnel_t *t = 0; - vnet_main_t * vnm = gm->vnet_main; + vnet_main_t * vnm = ngm->vnet_main; vnet_hw_interface_t * hi; uword * p; u32 hw_if_index = ~0; @@ -301,16 +486,16 @@ int vnet_vxlan_gpe_add_del_tunnel int rv; vxlan4_gpe_tunnel_key_t key4, *key4_copy; vxlan6_gpe_tunnel_key_t key6, *key6_copy; - hash_pair_t *hp; + u32 is_ip6 = a->is_ip6; - if (!a->is_ip6) + if (!is_ip6) { key4.local = a->local.ip4.as_u32; key4.remote = a->remote.ip4.as_u32; key4.vni = clib_host_to_net_u32 (a->vni << 8); key4.pad = 0; - p = hash_get_mem(gm->vxlan4_gpe_tunnel_by_key, &key4); + p = hash_get_mem(ngm->vxlan4_gpe_tunnel_by_key, &key4); } else { @@ -320,16 +505,18 @@ int vnet_vxlan_gpe_add_del_tunnel key6.remote.as_u64[1] = a->remote.ip6.as_u64[1]; key6.vni = clib_host_to_net_u32 (a->vni << 8); - p = hash_get_mem(gm->vxlan6_gpe_tunnel_by_key, &key6); + p = hash_get_mem(ngm->vxlan6_gpe_tunnel_by_key, &key6); } if (a->is_add) { + l2input_main_t * l2im = &l2input_main; + /* adding a tunnel: tunnel must not already exist */ if (p) - return VNET_API_ERROR_INVALID_VALUE; + return VNET_API_ERROR_TUNNEL_EXIST; - pool_get_aligned (gm->tunnels, t, CLIB_CACHE_LINE_BYTES); + pool_get_aligned (ngm->tunnels, t, CLIB_CACHE_LINE_BYTES); memset (t, 0, sizeof (*t)); /* copy from arg structure */ @@ -349,51 +536,167 @@ int vnet_vxlan_gpe_add_del_tunnel if (rv) { - pool_put (gm->tunnels, t); + pool_put (ngm->tunnels, t); return rv; } - if (!a->is_ip6) + if (!is_ip6) { key4_copy = clib_mem_alloc (sizeof (*key4_copy)); clib_memcpy (key4_copy, &key4, sizeof (*key4_copy)); - hash_set_mem (gm->vxlan4_gpe_tunnel_by_key, key4_copy, - t - gm->tunnels); + hash_set_mem (ngm->vxlan4_gpe_tunnel_by_key, key4_copy, + t - ngm->tunnels); } else { key6_copy = clib_mem_alloc (sizeof (*key6_copy)); clib_memcpy (key6_copy, &key6, sizeof (*key6_copy)); - hash_set_mem (gm->vxlan6_gpe_tunnel_by_key, key6_copy, - t - gm->tunnels); + hash_set_mem (ngm->vxlan6_gpe_tunnel_by_key, key6_copy, + t - ngm->tunnels); } - if (vec_len (gm->free_vxlan_gpe_tunnel_hw_if_indices) > 0) + if (vec_len (ngm->free_vxlan_gpe_tunnel_hw_if_indices) > 0) { - hw_if_index = gm->free_vxlan_gpe_tunnel_hw_if_indices - [vec_len (gm->free_vxlan_gpe_tunnel_hw_if_indices)-1]; - _vec_len (gm->free_vxlan_gpe_tunnel_hw_if_indices) -= 1; + vnet_interface_main_t * im = &vnm->interface_main; + hw_if_index = ngm->free_vxlan_gpe_tunnel_hw_if_indices + [vec_len (ngm->free_vxlan_gpe_tunnel_hw_if_indices)-1]; + _vec_len (ngm->free_vxlan_gpe_tunnel_hw_if_indices) -= 1; hi = vnet_get_hw_interface (vnm, hw_if_index); - hi->dev_instance = t - gm->tunnels; + hi->dev_instance = t - ngm->tunnels; hi->hw_instance = hi->dev_instance; + /* clear old stats of freed tunnel before reuse */ + sw_if_index = hi->sw_if_index; + vnet_interface_counter_lock(im); + vlib_zero_combined_counter + (&im->combined_sw_if_counters[VNET_INTERFACE_COUNTER_TX], sw_if_index); + vlib_zero_combined_counter + (&im->combined_sw_if_counters[VNET_INTERFACE_COUNTER_RX], sw_if_index); + vlib_zero_simple_counter + (&im->sw_if_counters[VNET_INTERFACE_COUNTER_DROP], sw_if_index); + vnet_interface_counter_unlock(im); } else { hw_if_index = vnet_register_interface - (vnm, vxlan_gpe_device_class.index, t - gm->tunnels, - vxlan_gpe_hw_class.index, t - gm->tunnels); + (vnm, vxlan_gpe_device_class.index, t - ngm->tunnels, + vxlan_gpe_hw_class.index, t - ngm->tunnels); hi = vnet_get_hw_interface (vnm, hw_if_index); hi->output_node_index = vxlan_gpe_encap_node.index; } t->hw_if_index = hw_if_index; t->sw_if_index = sw_if_index = hi->sw_if_index; - vec_validate_init_empty (gm->tunnel_index_by_sw_if_index, sw_if_index, ~0); - gm->tunnel_index_by_sw_if_index[sw_if_index] = t - gm->tunnels; + vec_validate_init_empty (ngm->tunnel_index_by_sw_if_index, sw_if_index, ~0); + ngm->tunnel_index_by_sw_if_index[sw_if_index] = t - ngm->tunnels; + + /* setup l2 input config with l2 feature and bd 0 to drop packet */ + vec_validate (l2im->configs, sw_if_index); + l2im->configs[sw_if_index].feature_bitmap = L2INPUT_FEAT_DROP; + l2im->configs[sw_if_index].bd_index = 0; + vnet_sw_interface_t * si = vnet_get_sw_interface (vnm, sw_if_index); + si->flags &= ~VNET_SW_INTERFACE_FLAG_HIDDEN; vnet_sw_interface_set_flags (vnm, hi->sw_if_index, VNET_SW_INTERFACE_FLAG_ADMIN_UP); + fib_node_init(&t->node, FIB_NODE_TYPE_VXLAN_GPE_TUNNEL); + fib_prefix_t tun_remote_pfx; + u32 encap_index = vxlan_gpe_encap_node.index; + vnet_flood_class_t flood_class = VNET_FLOOD_CLASS_TUNNEL_NORMAL; + + fib_prefix_from_ip46_addr(&t->remote, &tun_remote_pfx); + if (!ip46_address_is_multicast(&t->remote)) + { + /* Unicast tunnel - + * source the FIB entry for the tunnel's destination + * and become a child thereof. The tunnel will then get poked + * when the forwarding for the entry updates, and the tunnel can + * re-stack accordingly + */ + vtep_addr_ref(&t->local); + t->fib_entry_index = fib_table_entry_special_add + (t->encap_fib_index, &tun_remote_pfx, FIB_SOURCE_RR, + FIB_ENTRY_FLAG_NONE); + t->sibling_index = fib_entry_child_add + (t->fib_entry_index, FIB_NODE_TYPE_VXLAN_GPE_TUNNEL, t - ngm->tunnels); + vxlan_gpe_tunnel_restack_dpo(t); + } + else + { + /* Multicast tunnel - + * as the same mcast group can be used for mutiple mcast tunnels + * with different VNIs, create the output fib adjecency only if + * it does not already exist + */ + fib_protocol_t fp = fib_ip_proto(is_ip6); + + if (vtep_addr_ref(&t->remote) == 1) + { + fib_node_index_t mfei; + adj_index_t ai; + fib_route_path_t path = { + .frp_proto = fp, + .frp_addr = zero_addr, + .frp_sw_if_index = 0xffffffff, + .frp_fib_index = ~0, + .frp_weight = 0, + .frp_flags = FIB_ROUTE_PATH_LOCAL, + }; + const mfib_prefix_t mpfx = { + .fp_proto = fp, + .fp_len = (is_ip6 ? 128 : 32), + .fp_grp_addr = tun_remote_pfx.fp_addr, + }; + + /* + * Setup the (*,G) to receive traffic on the mcast group + * - the forwarding interface is for-us + * - the accepting interface is that from the API + */ + mfib_table_entry_path_update(t->encap_fib_index, + &mpfx, + MFIB_SOURCE_VXLAN_GPE, + &path, + MFIB_ITF_FLAG_FORWARD); + + path.frp_sw_if_index = a->mcast_sw_if_index; + path.frp_flags = FIB_ROUTE_PATH_FLAG_NONE; + mfei = mfib_table_entry_path_update(t->encap_fib_index, + &mpfx, + MFIB_SOURCE_VXLAN_GPE, + &path, + MFIB_ITF_FLAG_ACCEPT); + + /* + * Create the mcast adjacency to send traffic to the group + */ + ai = adj_mcast_add_or_lock(fp, + fib_proto_to_link(fp), + a->mcast_sw_if_index); + + /* + * create a new end-point + */ + mcast_shared_add(&t->remote, mfei, ai); + } + + dpo_id_t dpo = DPO_INVALID; + mcast_shared_t ep = mcast_shared_get(&t->remote); + + /* Stack shared mcast remote mac addr rewrite on encap */ + dpo_set (&dpo, DPO_ADJACENCY_MCAST, + fib_proto_to_dpo(fp), + ep.mcast_adj_index); + + dpo_stack_from_node (encap_index, &t->next_dpo, &dpo); + dpo_reset (&dpo); + flood_class = VNET_FLOOD_CLASS_TUNNEL_MASTER; + } + + /* Set vxlan tunnel output node */ + hi->output_node_index = encap_index; + + vnet_get_sw_interface (vnet_get_main(), sw_if_index)->flood_class = flood_class; } else { @@ -401,30 +704,36 @@ int vnet_vxlan_gpe_add_del_tunnel if (!p) return VNET_API_ERROR_NO_SUCH_ENTRY; - t = pool_elt_at_index (gm->tunnels, p[0]); + t = pool_elt_at_index (ngm->tunnels, p[0]); + sw_if_index = t->sw_if_index; vnet_sw_interface_set_flags (vnm, t->sw_if_index, 0 /* down */); - vec_add1 (gm->free_vxlan_gpe_tunnel_hw_if_indices, t->hw_if_index); + vnet_sw_interface_t * si = vnet_get_sw_interface (vnm, t->sw_if_index); + si->flags |= VNET_SW_INTERFACE_FLAG_HIDDEN; + set_int_l2_mode(ngm->vlib_main, vnm, MODE_L3, t->sw_if_index, 0, 0, 0, 0); + vec_add1 (ngm->free_vxlan_gpe_tunnel_hw_if_indices, t->hw_if_index); - gm->tunnel_index_by_sw_if_index[t->sw_if_index] = ~0; + ngm->tunnel_index_by_sw_if_index[t->sw_if_index] = ~0; - if (!a->is_ip6) - { - hp = hash_get_pair (gm->vxlan4_gpe_tunnel_by_key, &key4); - key4_copy = (void *)(hp->key); - hash_unset_mem (gm->vxlan4_gpe_tunnel_by_key, &key4); - clib_mem_free (key4_copy); - } + if (!is_ip6) + hash_unset (ngm->vxlan4_gpe_tunnel_by_key, key4.as_u64); else - { - hp = hash_get_pair (gm->vxlan6_gpe_tunnel_by_key, &key6); - key6_copy = (void *)(hp->key); - hash_unset_mem (gm->vxlan4_gpe_tunnel_by_key, &key6); - clib_mem_free (key6_copy); - } + hash_unset_key_free (&ngm->vxlan6_gpe_tunnel_by_key, &key6); + + if (!ip46_address_is_multicast(&t->remote)) + { + vtep_addr_unref(&t->local); + fib_entry_child_remove(t->fib_entry_index, t->sibling_index); + fib_table_entry_delete_index(t->fib_entry_index, FIB_SOURCE_RR); + } + else if (vtep_addr_unref(&t->remote) == 0) + { + mcast_shared_remove(&t->remote); + } + fib_node_deinit(&t->node); vec_free (t->rewrite); - pool_put (gm->tunnels, t); + pool_put (ngm->tunnels, t); } if (sw_if_indexp) @@ -443,8 +752,10 @@ vxlan_gpe_add_del_tunnel_command_fn (vlib_main_t * vm, ip46_address_t local, remote; u8 local_set = 0; u8 remote_set = 0; + u8 grp_set = 0; u8 ipv4_set = 0; u8 ipv6_set = 0; + u32 mcast_sw_if_index = ~0; u32 encap_fib_index = 0; u32 decap_fib_index = 0; u8 protocol = VXLAN_GPE_PROTOCOL_IP4; @@ -487,6 +798,22 @@ vxlan_gpe_add_del_tunnel_command_fn (vlib_main_t * vm, remote_set = 1; ipv6_set = 1; } + else if (unformat (line_input, "group %U %U", + unformat_ip4_address, &remote.ip4, + unformat_vnet_sw_interface, + vnet_get_main(), &mcast_sw_if_index)) + { + grp_set = remote_set = 1; + ipv4_set = 1; + } + else if (unformat (line_input, "group %U %U", + unformat_ip6_address, &remote.ip6, + unformat_vnet_sw_interface, + vnet_get_main(), &mcast_sw_if_index)) + { + grp_set = remote_set = 1; + ipv6_set = 1; + } else if (unformat (line_input, "encap-vrf-id %d", &tmp)) { if (ipv6_set) @@ -543,6 +870,23 @@ vxlan_gpe_add_del_tunnel_command_fn (vlib_main_t * vm, goto done; } + if (grp_set && !ip46_address_is_multicast(&remote)) + { + error = clib_error_return (0, "tunnel group address not multicast"); + goto done; + } + + if (grp_set == 0 && ip46_address_is_multicast(&remote)) + { + error = clib_error_return (0, "remote address must be unicast"); + goto done; + } + + if (grp_set && mcast_sw_if_index == ~0) + { + error = clib_error_return (0, "tunnel nonexistent multicast device"); + goto done; + } if (ipv4_set && ipv6_set) { error = clib_error_return (0, "both IPv4 and IPv6 addresses specified"); @@ -552,7 +896,7 @@ vxlan_gpe_add_del_tunnel_command_fn (vlib_main_t * vm, if ((ipv4_set && memcmp(&local.ip4, &remote.ip4, sizeof(local.ip4)) == 0) || (ipv6_set && memcmp(&local.ip6, &remote.ip6, sizeof(local.ip6)) == 0)) { - error = clib_error_return (0, "src and dst addresses are identical"); + error = clib_error_return (0, "src and remote addresses are identical"); goto done; } @@ -604,15 +948,36 @@ done: return error; } +/*? + * Add or delete a VXLAN-GPE Tunnel. + * + * VXLAN-GPE provides the features needed to allow L2 bridge domains (BDs) + * to span multiple servers. This is done by building an L2 overlay on + * top of an L3 network underlay using VXLAN-GPE tunnels. + * + * This makes it possible for servers to be co-located in the same data + * center or be separated geographically as long as they are reachable + * through the underlay L3 network. + * + * You can refer to this kind of L2 overlay bridge domain as a VXLAN-GPE sengment. + * + * @cliexpar + * Example of how to create a VXLAN-GPE Tunnel: + * @cliexcmd{create vxlan-gpe tunnel local 10.0.3.1 local 10.0.3.3 vni 13 encap-vrf-id 7} + * Example of how to delete a VXLAN Tunnel: + * @cliexcmd{create vxlan tunnel src 10.0.3.1 remote 10.0.3.3 vni 13 del} + ?*/ +/* *INDENT-OFF* */ VLIB_CLI_COMMAND (create_vxlan_gpe_tunnel_command, static) = { .path = "create vxlan-gpe tunnel", .short_help = - "create vxlan-gpe tunnel local remote " + "create vxlan-gpe tunnel local " + " {remote |group }" " vni [next-ip4][next-ip6][next-ethernet][next-nsh]" - " [encap-vrf-id ] [decap-vrf-id ]" - " [del]\n", + " [encap-vrf-id ] [decap-vrf-id ] [del]\n", .function = vxlan_gpe_add_del_tunnel_command_fn, }; +/* *INDENT-ON* */ /** * @brief CLI function for showing VXLAN GPE tunnels @@ -629,13 +994,13 @@ show_vxlan_gpe_tunnel_command_fn (vlib_main_t * vm, unformat_input_t * input, vlib_cli_command_t * cmd) { - vxlan_gpe_main_t * gm = &vxlan_gpe_main; + vxlan_gpe_main_t * ngm = &vxlan_gpe_main; vxlan_gpe_tunnel_t * t; - if (pool_elts (gm->tunnels) == 0) + if (pool_elts (ngm->tunnels) == 0) vlib_cli_output (vm, "No vxlan-gpe tunnels configured."); - pool_foreach (t, gm->tunnels, + pool_foreach (t, ngm->tunnels, ({ vlib_cli_output (vm, "%U", format_vxlan_gpe_tunnel, t); })); @@ -643,10 +1008,194 @@ show_vxlan_gpe_tunnel_command_fn (vlib_main_t * vm, return 0; } +/*? + * Display all the VXLAN-GPE Tunnel entries. + * + * @cliexpar + * Example of how to display the VXLAN-GPE Tunnel entries: + * @cliexstart{show vxlan-gpe tunnel} + * [0] local 10.0.3.1 remote 10.0.3.3 vni 13 encap_fib_index 0 sw_if_index 5 decap_next l2 + * @cliexend + ?*/ +/* *INDENT-OFF* */ VLIB_CLI_COMMAND (show_vxlan_gpe_tunnel_command, static) = { .path = "show vxlan-gpe", .function = show_vxlan_gpe_tunnel_command_fn, }; +/* *INDENT-ON* */ + +void vnet_int_vxlan_gpe_bypass_mode (u32 sw_if_index, + u8 is_ip6, + u8 is_enable) +{ + if (is_ip6) + vnet_feature_enable_disable ("ip6-unicast", "ip6-vxlan-gpe-bypass", + sw_if_index, is_enable, 0, 0); + else + vnet_feature_enable_disable ("ip4-unicast", "ip4-vxlan-gpe-bypass", + sw_if_index, is_enable, 0, 0); +} + + +static clib_error_t * +set_ip_vxlan_gpe_bypass (u32 is_ip6, + unformat_input_t * input, + vlib_cli_command_t * cmd) +{ + unformat_input_t _line_input, * line_input = &_line_input; + vnet_main_t * vnm = vnet_get_main(); + clib_error_t * error = 0; + u32 sw_if_index, is_enable; + + sw_if_index = ~0; + is_enable = 1; + + if (! unformat_user (input, unformat_line_input, line_input)) + return 0; + + while (unformat_check_input (line_input) != UNFORMAT_END_OF_INPUT) + { + if (unformat_user (line_input, unformat_vnet_sw_interface, vnm, &sw_if_index)) + ; + else if (unformat (line_input, "del")) + is_enable = 0; + else + { + error = unformat_parse_error (line_input); + goto done; + } + } + + if (~0 == sw_if_index) + { + error = clib_error_return (0, "unknown interface `%U'", + format_unformat_error, line_input); + goto done; + } + + vnet_int_vxlan_gpe_bypass_mode (sw_if_index, is_ip6, is_enable); + + done: + unformat_free (line_input); + + return error; +} + +static clib_error_t * +set_ip4_vxlan_gpe_bypass (vlib_main_t * vm, + unformat_input_t * input, + vlib_cli_command_t * cmd) +{ + return set_ip_vxlan_gpe_bypass (0, input, cmd); +} + +/*? + * This command adds the 'ip4-vxlan-gpe-bypass' graph node for a given interface. + * By adding the IPv4 vxlan-gpe-bypass graph node to an interface, the node checks + * for and validate input vxlan_gpe packet and bypass ip4-lookup, ip4-local, + * ip4-udp-lookup nodes to speedup vxlan_gpe packet forwarding. This node will + * cause extra overhead to for non-vxlan_gpe packets which is kept at a minimum. + * + * @cliexpar + * @parblock + * Example of graph node before ip4-vxlan-gpe-bypass is enabled: + * @cliexstart{show vlib graph ip4-vxlan-gpe-bypass} + * Name Next Previous + * ip4-vxlan-gpe-bypass error-drop [0] + * vxlan4-gpe-input [1] + * ip4-lookup [2] + * @cliexend + * + * Example of how to enable ip4-vxlan-gpe-bypass on an interface: + * @cliexcmd{set interface ip vxlan-gpe-bypass GigabitEthernet2/0/0} + * + * Example of graph node after ip4-vxlan-gpe-bypass is enabled: + * @cliexstart{show vlib graph ip4-vxlan-gpe-bypass} + * Name Next Previous + * ip4-vxlan-gpe-bypass error-drop [0] ip4-input + * vxlan4-gpe-input [1] ip4-input-no-checksum + * ip4-lookup [2] + * @cliexend + * + * Example of how to display the feature enabed on an interface: + * @cliexstart{show ip interface features GigabitEthernet2/0/0} + * IP feature paths configured on GigabitEthernet2/0/0... + * ... + * ipv4 unicast: + * ip4-vxlan-gpe-bypass + * ip4-lookup + * ... + * @cliexend + * + * Example of how to disable ip4-vxlan-gpe-bypass on an interface: + * @cliexcmd{set interface ip vxlan-gpe-bypass GigabitEthernet2/0/0 del} + * @endparblock +?*/ +/* *INDENT-OFF* */ +VLIB_CLI_COMMAND (set_interface_ip_vxlan_gpe_bypass_command, static) = { + .path = "set interface ip vxlan-gpe-bypass", + .function = set_ip4_vxlan_gpe_bypass, + .short_help = "set interface ip vxlan-gpe-bypass [del]", +}; +/* *INDENT-ON* */ + +static clib_error_t * +set_ip6_vxlan_gpe_bypass (vlib_main_t * vm, + unformat_input_t * input, + vlib_cli_command_t * cmd) +{ + return set_ip_vxlan_gpe_bypass (1, input, cmd); +} + +/*? + * This command adds the 'ip6-vxlan-gpe-bypass' graph node for a given interface. + * By adding the IPv6 vxlan-gpe-bypass graph node to an interface, the node checks + * for and validate input vxlan_gpe packet and bypass ip6-lookup, ip6-local, + * ip6-udp-lookup nodes to speedup vxlan_gpe packet forwarding. This node will + * cause extra overhead to for non-vxlan_gpe packets which is kept at a minimum. + * + * @cliexpar + * @parblock + * Example of graph node before ip6-vxlan-gpe-bypass is enabled: + * @cliexstart{show vlib graph ip6-vxlan-gpe-bypass} + * Name Next Previous + * ip6-vxlan-gpe-bypass error-drop [0] + * vxlan6-gpe-input [1] + * ip6-lookup [2] + * @cliexend + * + * Example of how to enable ip6-vxlan-gpe-bypass on an interface: + * @cliexcmd{set interface ip6 vxlan-gpe-bypass GigabitEthernet2/0/0} + * + * Example of graph node after ip6-vxlan-gpe-bypass is enabled: + * @cliexstart{show vlib graph ip6-vxlan-gpe-bypass} + * Name Next Previous + * ip6-vxlan-gpe-bypass error-drop [0] ip6-input + * vxlan6-gpe-input [1] ip4-input-no-checksum + * ip6-lookup [2] + * @cliexend + * + * Example of how to display the feature enabed on an interface: + * @cliexstart{show ip interface features GigabitEthernet2/0/0} + * IP feature paths configured on GigabitEthernet2/0/0... + * ... + * ipv6 unicast: + * ip6-vxlan-gpe-bypass + * ip6-lookup + * ... + * @cliexend + * + * Example of how to disable ip6-vxlan-gpe-bypass on an interface: + * @cliexcmd{set interface ip6 vxlan-gpe-bypass GigabitEthernet2/0/0 del} + * @endparblock +?*/ +/* *INDENT-OFF* */ +VLIB_CLI_COMMAND (set_interface_ip6_vxlan_gpe_bypass_command, static) = { + .path = "set interface ip6 vxlan-gpe-bypass", + .function = set_ip6_vxlan_gpe_bypass, + .short_help = "set interface ip vxlan-gpe-bypass [del]", +}; +/* *INDENT-ON* */ /** * @brief Feature init function for VXLAN GPE @@ -658,21 +1207,25 @@ VLIB_CLI_COMMAND (show_vxlan_gpe_tunnel_command, static) = { */ clib_error_t *vxlan_gpe_init (vlib_main_t *vm) { - vxlan_gpe_main_t *gm = &vxlan_gpe_main; + vxlan_gpe_main_t *ngm = &vxlan_gpe_main; - gm->vnet_main = vnet_get_main(); - gm->vlib_main = vm; + ngm->vnet_main = vnet_get_main(); + ngm->vlib_main = vm; - gm->vxlan4_gpe_tunnel_by_key + ngm->vxlan4_gpe_tunnel_by_key = hash_create_mem (0, sizeof(vxlan4_gpe_tunnel_key_t), sizeof (uword)); - gm->vxlan6_gpe_tunnel_by_key + ngm->vxlan6_gpe_tunnel_by_key = hash_create_mem (0, sizeof(vxlan6_gpe_tunnel_key_t), sizeof (uword)); - udp_register_dst_port (vm, UDP_DST_PORT_vxlan_gpe, + ngm->mcast_shared = hash_create_mem(0, + sizeof(ip46_address_t), + sizeof(mcast_shared_t)); + + udp_register_dst_port (vm, UDP_DST_PORT_VXLAN_GPE, vxlan4_gpe_input_node.index, 1 /* is_ip4 */); - udp_register_dst_port (vm, UDP_DST_PORT_vxlan6_gpe, + udp_register_dst_port (vm, UDP_DST_PORT_VXLAN6_GPE, vxlan6_gpe_input_node.index, 0 /* is_ip4 */); /* Register the list of standard decap protocols supported */ @@ -682,6 +1235,9 @@ clib_error_t *vxlan_gpe_init (vlib_main_t *vm) VXLAN_GPE_INPUT_NEXT_IP6_INPUT); vxlan_gpe_register_decap_protocol (VXLAN_GPE_PROTOCOL_ETHERNET, VXLAN_GPE_INPUT_NEXT_ETHERNET_INPUT); + + fib_node_register_type(FIB_NODE_TYPE_VXLAN_GPE_TUNNEL, &vxlan_gpe_vft); + return 0; } diff --git a/src/vnet/vxlan-gpe/vxlan_gpe.h b/src/vnet/vxlan-gpe/vxlan_gpe.h index e768d230..c348b5d5 100644 --- a/src/vnet/vxlan-gpe/vxlan_gpe.h +++ b/src/vnet/vxlan-gpe/vxlan_gpe.h @@ -25,11 +25,15 @@ #include #include #include +#include +#include #include #include #include #include #include +#include +#include /** * @brief VXLAN GPE header struct @@ -94,11 +98,16 @@ typedef struct { /** encapsulated protocol */ u8 protocol; + /* FIB DPO for IP forwarding of VXLAN-GPE encap packet */ + dpo_id_t next_dpo; /** tunnel local address */ ip46_address_t local; /** tunnel remote address */ ip46_address_t remote; + /* mcast packet output intfc index (used only if dst is mcast) */ + u32 mcast_sw_if_index; + /** FIB indices - tunnel partner lookup here */ u32 encap_fib_index; /** FIB indices - inner IP packet lookup here */ @@ -120,6 +129,27 @@ typedef struct { /** Next node after VxLAN-GPE encap */ uword encap_next_node; + + /** + * Linkage into the FIB object graph + */ + fib_node_t node; + + /* + * The FIB entry for (depending on VXLAN-GPE tunnel is unicast or mcast) + * sending unicast VXLAN-GPE encap packets or receiving mcast VXLAN-GPE packets + */ + fib_node_index_t fib_entry_index; + adj_index_t mcast_adj_index; + + /** + * The tunnel is a child of the FIB entry for its desintion. This is + * so it receives updates when the forwarding information for that entry + * changes. + * The tunnels sibling index on the FIB entry's dependency list. + */ + u32 sibling_index; + } vxlan_gpe_tunnel_t; /** Flags for vxlan_gpe_tunnel_t */ @@ -158,6 +188,12 @@ typedef struct { /** lookup IPv6 VXLAN GPE tunnel by key */ uword * vxlan6_gpe_tunnel_by_key; + /* local VTEP IPs ref count used by vxlan-bypass node to check if + received VXLAN packet DIP matches any local VTEP address */ + uword * vtep4; /* local ip4 VTEPs keyed on their ip4 addr */ + uword * vtep6; /* local ip6 VTEPs keyed on their ip6 addr */ + /* mcast shared info */ + uword * mcast_shared; /* keyed on mcast ip46 addr */ /** Free vlib hw_if_indices */ u32 * free_vxlan_gpe_tunnel_hw_if_indices; @@ -187,6 +223,7 @@ typedef struct { u8 is_ip6; ip46_address_t local, remote; u8 protocol; + u32 mcast_sw_if_index; u32 encap_fib_index; u32 decap_fib_index; u32 vni; @@ -217,5 +254,6 @@ void vxlan_gpe_unregister_decap_protocol (u8 protocol_id, uword next_node_index) void vxlan_gpe_register_decap_protocol (u8 protocol_id, uword next_node_index); +void vnet_int_vxlan_gpe_bypass_mode (u32 sw_if_index, u8 is_ip6, u8 is_enable); #endif /* included_vnet_vxlan_gpe_h */ diff --git a/src/vnet/vxlan-gpe/vxlan_gpe_api.c b/src/vnet/vxlan-gpe/vxlan_gpe_api.c index 0215054d..3675fc55 100644 --- a/src/vnet/vxlan-gpe/vxlan_gpe_api.c +++ b/src/vnet/vxlan-gpe/vxlan_gpe_api.c @@ -22,6 +22,7 @@ #include #include +#include #include #include @@ -44,9 +45,26 @@ #include #define foreach_vpe_api_msg \ +_(SW_INTERFACE_SET_VXLAN_GPE_BYPASS, sw_interface_set_vxlan_gpe_bypass) \ _(VXLAN_GPE_ADD_DEL_TUNNEL, vxlan_gpe_add_del_tunnel) \ _(VXLAN_GPE_TUNNEL_DUMP, vxlan_gpe_tunnel_dump) \ +static void + vl_api_sw_interface_set_vxlan_gpe_bypass_t_handler + (vl_api_sw_interface_set_vxlan_gpe_bypass_t * mp) +{ + vl_api_sw_interface_set_vxlan_gpe_bypass_reply_t *rmp; + int rv = 0; + u32 sw_if_index = ntohl (mp->sw_if_index); + + VALIDATE_SW_IF_INDEX (mp); + + vnet_int_vxlan_gpe_bypass_mode (sw_if_index, mp->is_ipv6, mp->enable); + BAD_SW_IF_INDEX_LABEL; + + REPLY_MACRO (VL_API_SW_INTERFACE_SET_VXLAN_GPE_BYPASS_REPLY); +} + static void vl_api_vxlan_gpe_add_del_tunnel_t_handler (vl_api_vxlan_gpe_add_del_tunnel_t * mp) @@ -109,6 +127,7 @@ static void clib_memcpy (&(a->local.ip4), mp->local, 4); clib_memcpy (&(a->remote.ip4), mp->remote, 4); } + a->mcast_sw_if_index = ntohl (mp->mcast_sw_if_index); a->encap_fib_index = encap_fib_index; a->decap_fib_index = decap_fib_index; a->protocol = protocol; @@ -149,6 +168,7 @@ static void send_vxlan_gpe_tunnel_details rmp->encap_vrf_id = htonl (im4->fibs[t->encap_fib_index].ft_table_id); rmp->decap_vrf_id = htonl (im4->fibs[t->decap_fib_index].ft_table_id); } + rmp->mcast_sw_if_index = htonl (t->mcast_sw_if_index); rmp->vni = htonl (t->vni); rmp->protocol = t->protocol; rmp->sw_if_index = htonl (t->sw_if_index); diff --git a/test/test_vxlan_gpe.py b/test/test_vxlan_gpe.py new file mode 100644 index 00000000..40975bf1 --- /dev/null +++ b/test/test_vxlan_gpe.py @@ -0,0 +1,313 @@ +#!/usr/bin/env python + +import socket +from util import ip4n_range +import unittest +from framework import VppTestCase, VppTestRunner, running_extended_tests +from template_bd import BridgeDomain + +from scapy.layers.l2 import Ether, Raw +from scapy.layers.inet import IP, UDP +from scapy.layers.vxlan import VXLAN +from scapy.utils import atol + + +@unittest.skipUnless(running_extended_tests(), "part of extended tests") +class TestVxlanGpe(BridgeDomain, VppTestCase): + """ VXLAN-GPE Test Case """ + + def __init__(self, *args): + BridgeDomain.__init__(self) + VppTestCase.__init__(self, *args) + + def encapsulate(self, pkt, vni): + """ + Encapsulate the original payload frame by adding VXLAN-GPE header + with its UDP, IP and Ethernet fields + """ + return (Ether(src=self.pg0.remote_mac, dst=self.pg0.local_mac) / + IP(src=self.pg0.remote_ip4, dst=self.pg0.local_ip4) / + UDP(sport=self.dport, dport=self.dport, chksum=0) / + VXLAN(vni=vni, flags=self.flags) / + pkt) + + def encap_mcast(self, pkt, src_ip, src_mac, vni): + """ + Encapsulate the original payload frame by adding VXLAN-GPE header + with its UDP, IP and Ethernet fields + """ + return (Ether(src=src_mac, dst=self.mcast_mac) / + IP(src=src_ip, dst=self.mcast_ip4) / + UDP(sport=self.dport, dport=self.dport, chksum=0) / + VXLAN(vni=vni, flags=self.flags) / + pkt) + + def decapsulate(self, pkt): + """ + Decapsulate the original payload frame by removing VXLAN-GPE header + """ + # check if is set I and P flag + self.assertEqual(pkt[VXLAN].flags, int('0xc', 16)) + return pkt[VXLAN].payload + + # Method for checking VXLAN-GPE encapsulation. + # + def check_encapsulation(self, pkt, vni, local_only=False, mcast_pkt=False): + # Verify source MAC is VPP_MAC and destination MAC is MY_MAC resolved + # by VPP using ARP. + self.assertEqual(pkt[Ether].src, self.pg0.local_mac) + if not local_only: + if not mcast_pkt: + self.assertEqual(pkt[Ether].dst, self.pg0.remote_mac) + else: + self.assertEqual(pkt[Ether].dst, type(self).mcast_mac) + # Verify VXLAN-GPE tunnel src IP is VPP_IP and dst IP is MY_IP. + self.assertEqual(pkt[IP].src, self.pg0.local_ip4) + if not local_only: + if not mcast_pkt: + self.assertEqual(pkt[IP].dst, self.pg0.remote_ip4) + else: + self.assertEqual(pkt[IP].dst, type(self).mcast_ip4) + # Verify UDP destination port is VXLAN-GPE 4790, source UDP port + # could be arbitrary. + self.assertEqual(pkt[UDP].dport, type(self).dport) + # Verify VNI + self.assertEqual(pkt[VXLAN].VNI, vni) + + def test_decap(self): + """ Decapsulation test + Send encapsulated frames from pg0 + Verify receipt of decapsulated frames on pg1 + """ + + encapsulated_pkt = self.encapsulate(self.frame_request, + self.single_tunnel_bd) + + self.pg0.add_stream([encapsulated_pkt, ]) + + self.pg1.enable_capture() + + self.pg_start() + + # Pick first received frame and check if it's the non-encapsulated + # frame + out = self.pg1.get_capture(1) + pkt = out[0] + self.assert_eq_pkts(pkt, self.frame_request) + + def test_encap(self): + """ Encapsulation test + Send frames from pg1 + Verify receipt of encapsulated frames on pg0 + """ + self.pg1.add_stream([self.frame_reply]) + + self.pg0.enable_capture() + + self.pg_start() + + # Pick first received frame and check if it's corectly encapsulated. + out = self.pg0.get_capture(1) + pkt = out[0] + self.check_encapsulation(pkt, self.single_tunnel_bd) + + # payload = self.decapsulate(pkt) + # self.assert_eq_pkts(payload, self.frame_reply) + + def test_ucast_flood(self): + """ Unicast flood test + Send frames from pg3 + Verify receipt of encapsulated frames on pg0 + """ + self.pg3.add_stream([self.frame_reply]) + + self.pg0.enable_capture() + + self.pg_start() + + # Get packet from each tunnel and assert it's corectly encapsulated. + out = self.pg0.get_capture(self.n_ucast_tunnels) + for pkt in out: + self.check_encapsulation(pkt, self.ucast_flood_bd, True) + # payload = self.decapsulate(pkt) + # self.assert_eq_pkts(payload, self.frame_reply) + + def test_mcast_flood(self): + """ Multicast flood test + Send frames from pg2 + Verify receipt of encapsulated frames on pg0 + """ + self.pg2.add_stream([self.frame_reply]) + + self.pg0.enable_capture() + + self.pg_start() + + # Pick first received frame and check if it's corectly encapsulated. + out = self.pg0.get_capture(1) + pkt = out[0] + self.check_encapsulation(pkt, self.mcast_flood_bd, + local_only=False, mcast_pkt=True) + + # payload = self.decapsulate(pkt) + # self.assert_eq_pkts(payload, self.frame_reply) + + @classmethod + def create_vxlan_gpe_flood_test_bd(cls, vni, n_ucast_tunnels): + # Create 10 ucast vxlan_gpe tunnels under bd + ip_range_start = 10 + ip_range_end = ip_range_start + n_ucast_tunnels + next_hop_address = cls.pg0.remote_ip4n + for dest_ip4n in ip4n_range(next_hop_address, ip_range_start, + ip_range_end): + # add host route so dest_ip4n will not be resolved + cls.vapi.ip_add_del_route(dest_ip4n, 32, next_hop_address) + r = cls.vapi.vxlan_gpe_add_del_tunnel( + src_addr=cls.pg0.local_ip4n, + dst_addr=dest_ip4n, + vni=vni) + cls.vapi.sw_interface_set_l2_bridge(r.sw_if_index, bd_id=vni) + + @classmethod + def add_del_shared_mcast_dst_load(cls, is_add): + """ + add or del tunnels sharing the same mcast dst + to test vxlan_gpe ref_count mechanism + """ + n_shared_dst_tunnels = 20 + vni_start = 1000 + vni_end = vni_start + n_shared_dst_tunnels + for vni in range(vni_start, vni_end): + r = cls.vapi.vxlan_gpe_add_del_tunnel( + src_addr=cls.pg0.local_ip4n, + dst_addr=cls.mcast_ip4n, + mcast_sw_if_index=1, + vni=vni, + is_add=is_add) + if r.sw_if_index == 0xffffffff: + raise "bad sw_if_index" + + @classmethod + def add_shared_mcast_dst_load(cls): + cls.add_del_shared_mcast_dst_load(is_add=1) + + @classmethod + def del_shared_mcast_dst_load(cls): + cls.add_del_shared_mcast_dst_load(is_add=0) + + @classmethod + def add_del_mcast_tunnels_load(cls, is_add): + """ + add or del tunnels to test vxlan_gpe stability + """ + n_distinct_dst_tunnels = 20 + ip_range_start = 10 + ip_range_end = ip_range_start + n_distinct_dst_tunnels + for dest_ip4n in ip4n_range(cls.mcast_ip4n, ip_range_start, + ip_range_end): + vni = bytearray(dest_ip4n)[3] + cls.vapi.vxlan_gpe_add_del_tunnel( + src_addr=cls.pg0.local_ip4n, + dst_addr=dest_ip4n, + mcast_sw_if_index=1, + vni=vni, + is_add=is_add) + + @classmethod + def add_mcast_tunnels_load(cls): + cls.add_del_mcast_tunnels_load(is_add=1) + + @classmethod + def del_mcast_tunnels_load(cls): + cls.add_del_mcast_tunnels_load(is_add=0) + + # Class method to start the VXLAN-GPE test case. + # Overrides setUpClass method in VppTestCase class. + # Python try..except statement is used to ensure that the tear down of + # the class will be executed even if exception is raised. + # @param cls The class pointer. + @classmethod + def setUpClass(cls): + super(TestVxlanGpe, cls).setUpClass() + + try: + cls.dport = 4790 + cls.flags = 0xc + + # Create 2 pg interfaces. + cls.create_pg_interfaces(range(4)) + for pg in cls.pg_interfaces: + pg.admin_up() + + # Configure IPv4 addresses on VPP pg0. + cls.pg0.config_ip4() + + # Resolve MAC address for VPP's IP address on pg0. + cls.pg0.resolve_arp() + + # Our Multicast address + cls.mcast_ip4 = '239.1.1.1' + cls.mcast_ip4n = socket.inet_pton(socket.AF_INET, cls.mcast_ip4) + iplong = atol(cls.mcast_ip4) + cls.mcast_mac = "01:00:5e:%02x:%02x:%02x" % ( + (iplong >> 16) & 0x7F, (iplong >> 8) & 0xFF, iplong & 0xFF) + + # Create VXLAN-GPE VTEP on VPP pg0, and put vxlan_gpe_tunnel0 + # and pg1 into BD. + cls.single_tunnel_bd = 11 + r = cls.vapi.vxlan_gpe_add_del_tunnel( + src_addr=cls.pg0.local_ip4n, + dst_addr=cls.pg0.remote_ip4n, + vni=cls.single_tunnel_bd) + cls.vapi.sw_interface_set_l2_bridge(r.sw_if_index, + bd_id=cls.single_tunnel_bd) + cls.vapi.sw_interface_set_l2_bridge(cls.pg1.sw_if_index, + bd_id=cls.single_tunnel_bd) + + # Setup vni 2 to test multicast flooding + cls.n_ucast_tunnels = 10 + cls.mcast_flood_bd = 12 + cls.create_vxlan_gpe_flood_test_bd(cls.mcast_flood_bd, + cls.n_ucast_tunnels) + r = cls.vapi.vxlan_gpe_add_del_tunnel( + src_addr=cls.pg0.local_ip4n, + dst_addr=cls.mcast_ip4n, + mcast_sw_if_index=1, + vni=cls.mcast_flood_bd) + cls.vapi.sw_interface_set_l2_bridge(r.sw_if_index, + bd_id=cls.mcast_flood_bd) + cls.vapi.sw_interface_set_l2_bridge(cls.pg2.sw_if_index, + bd_id=cls.mcast_flood_bd) + + # Add and delete mcast tunnels to check stability + cls.add_shared_mcast_dst_load() + cls.add_mcast_tunnels_load() + cls.del_shared_mcast_dst_load() + cls.del_mcast_tunnels_load() + + # Setup vni 3 to test unicast flooding + cls.ucast_flood_bd = 13 + cls.create_vxlan_gpe_flood_test_bd(cls.ucast_flood_bd, + cls.n_ucast_tunnels) + cls.vapi.sw_interface_set_l2_bridge(cls.pg3.sw_if_index, + bd_id=cls.ucast_flood_bd) + except Exception: + super(TestVxlanGpe, cls).tearDownClass() + raise + + # Method to define VPP actions before tear down of the test case. + # Overrides tearDown method in VppTestCase class. + # @param self The object pointer. + def tearDown(self): + super(TestVxlanGpe, self).tearDown() + if not self.vpp_dead: + self.logger.info(self.vapi.cli("show bridge-domain 11 detail")) + self.logger.info(self.vapi.cli("show bridge-domain 12 detail")) + self.logger.info(self.vapi.cli("show bridge-domain 13 detail")) + self.logger.info(self.vapi.cli("show int")) + self.logger.info(self.vapi.cli("show vxlan-gpe tunnel")) + self.logger.info(self.vapi.cli("show trace")) + + +if __name__ == '__main__': + unittest.main(testRunner=VppTestRunner) diff --git a/test/vpp_papi_provider.py b/test/vpp_papi_provider.py index b8a7a470..b310d094 100644 --- a/test/vpp_papi_provider.py +++ b/test/vpp_papi_provider.py @@ -1828,3 +1828,38 @@ class VppPapiProvider(object): 'encap_vrf_id': encap_vrf_id, 'decap_next_index': decap_next_index, 'teid': teid}) + + def vxlan_gpe_add_del_tunnel( + self, + src_addr, + dst_addr, + mcast_sw_if_index=0xFFFFFFFF, + is_add=1, + is_ipv6=0, + encap_vrf_id=0, + decap_vrf_id=0, + protocol=3, + vni=0): + """ + + :param local: + :param remote: + :param is_add: (Default value = 1) + :param is_ipv6: (Default value = 0) + :param encap_vrf_id: (Default value = 0) + :param decap_vrf_id: (Default value = 0) + :param mcast_sw_if_index: (Default value = 0xFFFFFFFF) + :param protocol: (Default value = 3) + :param vni: (Default value = 0) + + """ + return self.api(self.papi.vxlan_gpe_add_del_tunnel, + {'is_add': is_add, + 'is_ipv6': is_ipv6, + 'local': src_addr, + 'remote': dst_addr, + 'mcast_sw_if_index': mcast_sw_if_index, + 'encap_vrf_id': encap_vrf_id, + 'decap_vrf_id': decap_vrf_id, + 'protocol': protocol, + 'vni': vni}) -- cgit 1.2.3-korg From 04ffd0ad83b2d87edb669a9d76eee85f5c589564 Mon Sep 17 00:00:00 2001 From: Hongjun Ni Date: Fri, 23 Jun 2017 00:18:40 +0800 Subject: VPP crash on creating vxlan gpe interface. VPP-875 Change-Id: I6b19634ecb03860a7624d9408e09b52e95f47aef Signed-off-by: Hongjun Ni --- src/vat/api_format.c | 108 +++++++++++++++++++++++++++++++++---- src/vnet/vxlan-gpe/vxlan_gpe.api | 1 - src/vnet/vxlan-gpe/vxlan_gpe.c | 3 ++ src/vnet/vxlan-gpe/vxlan_gpe_api.c | 13 +++-- src/vpp/api/custom_dump.c | 12 ++++- 5 files changed, 118 insertions(+), 19 deletions(-) (limited to 'src/vnet/vxlan-gpe') diff --git a/src/vat/api_format.c b/src/vat/api_format.c index 0a38353c..937d7c5d 100644 --- a/src/vat/api_format.c +++ b/src/vat/api_format.c @@ -1795,6 +1795,40 @@ static void vl_api_vxlan_add_del_tunnel_reply_t_handler_json vam->result_ready = 1; } +static void vl_api_vxlan_gpe_add_del_tunnel_reply_t_handler + (vl_api_vxlan_gpe_add_del_tunnel_reply_t * mp) +{ + vat_main_t *vam = &vat_main; + i32 retval = ntohl (mp->retval); + if (vam->async_mode) + { + vam->async_errors += (retval < 0); + } + else + { + vam->retval = retval; + vam->sw_if_index = ntohl (mp->sw_if_index); + vam->result_ready = 1; + } +} + +static void vl_api_vxlan_gpe_add_del_tunnel_reply_t_handler_json + (vl_api_vxlan_gpe_add_del_tunnel_reply_t * mp) +{ + vat_main_t *vam = &vat_main; + vat_json_node_t node; + + vat_json_init_object (&node); + vat_json_object_add_int (&node, "retval", ntohl (mp->retval)); + vat_json_object_add_uint (&node, "sw_if_index", ntohl (mp->sw_if_index)); + + vat_json_print (vam->ofp, &node); + vat_json_free (&node); + + vam->retval = ntohl (mp->retval); + vam->result_ready = 1; +} + static void vl_api_gre_add_del_tunnel_reply_t_handler (vl_api_gre_add_del_tunnel_reply_t * mp) { @@ -4480,6 +4514,7 @@ _(sw_interface_set_table_reply) \ _(sw_interface_set_mpls_enable_reply) \ _(sw_interface_set_vpath_reply) \ _(sw_interface_set_vxlan_bypass_reply) \ +_(sw_interface_set_vxlan_gpe_bypass_reply) \ _(sw_interface_set_l2_bridge_reply) \ _(bridge_domain_add_del_reply) \ _(sw_interface_set_l2_xconnect_reply) \ @@ -4578,7 +4613,6 @@ _(gpe_enable_disable_reply) \ _(gpe_set_encap_mode_reply) \ _(gpe_add_del_iface_reply) \ _(gpe_add_del_native_fwd_rpath_reply) \ -_(vxlan_gpe_add_del_tunnel_reply) \ _(af_packet_delete_reply) \ _(policer_classify_set_interface_reply) \ _(netmap_create_reply) \ @@ -4651,6 +4685,7 @@ _(SW_INTERFACE_SET_TABLE_REPLY, sw_interface_set_table_reply) \ _(SW_INTERFACE_SET_MPLS_ENABLE_REPLY, sw_interface_set_mpls_enable_reply) \ _(SW_INTERFACE_SET_VPATH_REPLY, sw_interface_set_vpath_reply) \ _(SW_INTERFACE_SET_VXLAN_BYPASS_REPLY, sw_interface_set_vxlan_bypass_reply) \ +_(SW_INTERFACE_SET_VXLAN_GPE_BYPASS_REPLY, sw_interface_set_vxlan_gpe_bypass_reply) \ _(SW_INTERFACE_SET_L2_XCONNECT_REPLY, \ sw_interface_set_l2_xconnect_reply) \ _(SW_INTERFACE_SET_L2_BRIDGE_REPLY, \ @@ -6039,6 +6074,7 @@ api_sw_interface_set_vxlan_bypass (vat_main_t * vam) return ret; } + static int api_sw_interface_set_l2_xconnect (vat_main_t * vam) { @@ -12002,6 +12038,8 @@ api_vxlan_gpe_add_del_tunnel (vat_main_t * vam) u8 ipv4_set = 0, ipv6_set = 0; u8 local_set = 0; u8 remote_set = 0; + u8 grp_set = 0; + u32 mcast_sw_if_index = ~0; u32 encap_vrf_id = 0; u32 decap_vrf_id = 0; u8 protocol = ~0; @@ -12009,6 +12047,12 @@ api_vxlan_gpe_add_del_tunnel (vat_main_t * vam) u8 vni_set = 0; int ret; + /* Can't "universally zero init" (={0}) due to GCC bug 53119 */ + memset (&local4, 0, sizeof local4); + memset (&remote4, 0, sizeof remote4); + memset (&local6, 0, sizeof local6); + memset (&remote6, 0, sizeof remote6); + while (unformat_check_input (line_input) != UNFORMAT_END_OF_INPUT) { if (unformat (line_input, "del")) @@ -12037,6 +12081,35 @@ api_vxlan_gpe_add_del_tunnel (vat_main_t * vam) remote_set = 1; ipv6_set = 1; } + else if (unformat (line_input, "group %U %U", + unformat_ip4_address, &remote4, + api_unformat_sw_if_index, vam, &mcast_sw_if_index)) + { + grp_set = remote_set = 1; + ipv4_set = 1; + } + else if (unformat (line_input, "group %U", + unformat_ip4_address, &remote4)) + { + grp_set = remote_set = 1; + ipv4_set = 1; + } + else if (unformat (line_input, "group %U %U", + unformat_ip6_address, &remote6, + api_unformat_sw_if_index, vam, &mcast_sw_if_index)) + { + grp_set = remote_set = 1; + ipv6_set = 1; + } + else if (unformat (line_input, "group %U", + unformat_ip6_address, &remote6)) + { + grp_set = remote_set = 1; + ipv6_set = 1; + } + else + if (unformat (line_input, "mcast_sw_if_index %u", &mcast_sw_if_index)) + ; else if (unformat (line_input, "encap-vrf-id %d", &encap_vrf_id)) ; else if (unformat (line_input, "decap-vrf-id %d", &decap_vrf_id)) @@ -12068,6 +12141,11 @@ api_vxlan_gpe_add_del_tunnel (vat_main_t * vam) errmsg ("tunnel remote address not specified"); return -99; } + if (grp_set && mcast_sw_if_index == ~0) + { + errmsg ("tunnel nonexistent multicast device"); + return -99; + } if (ipv4_set && ipv6_set) { errmsg ("both IPv4 and IPv6 addresses specified"); @@ -12094,6 +12172,7 @@ api_vxlan_gpe_add_del_tunnel (vat_main_t * vam) clib_memcpy (&mp->remote, &remote4, sizeof (remote4)); } + mp->mcast_sw_if_index = ntohl (mcast_sw_if_index); mp->encap_vrf_id = ntohl (encap_vrf_id); mp->decap_vrf_id = ntohl (decap_vrf_id); mp->protocol = protocol; @@ -12110,16 +12189,19 @@ static void vl_api_vxlan_gpe_tunnel_details_t_handler (vl_api_vxlan_gpe_tunnel_details_t * mp) { vat_main_t *vam = &vat_main; + ip46_address_t local = to_ip46 (mp->is_ipv6, mp->local); + ip46_address_t remote = to_ip46 (mp->is_ipv6, mp->remote); - print (vam->ofp, "%11d%24U%24U%13d%12d%14d%14d", + print (vam->ofp, "%11d%24U%24U%13d%12d%19d%14d%14d", ntohl (mp->sw_if_index), - format_ip46_address, &(mp->local[0]), - format_ip46_address, &(mp->remote[0]), - ntohl (mp->vni), - ntohl (mp->protocol), + format_ip46_address, &local, IP46_TYPE_ANY, + format_ip46_address, &remote, IP46_TYPE_ANY, + ntohl (mp->vni), mp->protocol, + ntohl (mp->mcast_sw_if_index), ntohl (mp->encap_vrf_id), ntohl (mp->decap_vrf_id)); } + static void vl_api_vxlan_gpe_tunnel_details_t_handler_json (vl_api_vxlan_gpe_tunnel_details_t * mp) { @@ -12153,6 +12235,8 @@ static void vl_api_vxlan_gpe_tunnel_details_t_handler_json } vat_json_object_add_uint (node, "vni", ntohl (mp->vni)); vat_json_object_add_uint (node, "protocol", ntohl (mp->protocol)); + vat_json_object_add_uint (node, "mcast_sw_if_index", + ntohl (mp->mcast_sw_if_index)); vat_json_object_add_uint (node, "encap_vrf_id", ntohl (mp->encap_vrf_id)); vat_json_object_add_uint (node, "decap_vrf_id", ntohl (mp->decap_vrf_id)); vat_json_object_add_uint (node, "is_ipv6", mp->is_ipv6 ? 1 : 0); @@ -12184,9 +12268,9 @@ api_vxlan_gpe_tunnel_dump (vat_main_t * vam) if (!vam->json_output) { - print (vam->ofp, "%11s%24s%24s%13s%15s%14s%14s", + print (vam->ofp, "%11s%24s%24s%13s%15s%19s%14s%14s", "sw_if_index", "local", "remote", "vni", - "protocol", "encap_vrf_id", "decap_vrf_id"); + "protocol", "mcast_sw_if_index", "encap_vrf_id", "decap_vrf_id"); } /* Get list of vxlan-tunnel interfaces */ @@ -12204,6 +12288,7 @@ api_vxlan_gpe_tunnel_dump (vat_main_t * vam) return ret; } + u8 * format_l2_fib_mac_address (u8 * s, va_list * args) { @@ -19673,9 +19758,10 @@ _(delete_vhost_user_if, " | sw_if_index ") \ _(sw_interface_vhost_user_dump, "") \ _(show_version, "") \ _(vxlan_gpe_add_del_tunnel, \ - "local remote vni \n" \ - "[encap-vrf-id ] [decap-vrf-id ] [next-ip4][next-ip6]" \ - "[next-ethernet] [next-nsh]\n") \ + "local remote | group \n" \ + "{ | mcast_sw_if_index } }\n" \ + "vni [encap-vrf-id ] [decap-vrf-id ]\n" \ + "[next-ip4][next-ip6][next-ethernet] [next-nsh] [del]\n") \ _(vxlan_gpe_tunnel_dump, "[ | sw_if_index ]") \ _(l2_fib_table_dump, "bd_id ") \ _(interface_name_renumber, \ diff --git a/src/vnet/vxlan-gpe/vxlan_gpe.api b/src/vnet/vxlan-gpe/vxlan_gpe.api index 41b10316..04082d69 100644 --- a/src/vnet/vxlan-gpe/vxlan_gpe.api +++ b/src/vnet/vxlan-gpe/vxlan_gpe.api @@ -53,7 +53,6 @@ define vxlan_gpe_tunnel_details u32 mcast_sw_if_index; u32 encap_vrf_id; u32 decap_vrf_id; - u8 is_ipv6; }; diff --git a/src/vnet/vxlan-gpe/vxlan_gpe.c b/src/vnet/vxlan-gpe/vxlan_gpe.c index 3a92c88f..fab887ad 100644 --- a/src/vnet/vxlan-gpe/vxlan_gpe.c +++ b/src/vnet/vxlan-gpe/vxlan_gpe.c @@ -83,6 +83,9 @@ u8 * format_vxlan_gpe_tunnel (u8 * s, va_list * args) s = format (s, "next-protocol unknown %d", t->protocol); } + if (ip46_address_is_multicast (&t->remote)) + s = format (s, "mcast_sw_if_index %d ", t->mcast_sw_if_index); + s = format (s, " fibs: (encap %d, decap %d)", t->encap_fib_index, t->decap_fib_index); diff --git a/src/vnet/vxlan-gpe/vxlan_gpe_api.c b/src/vnet/vxlan-gpe/vxlan_gpe_api.c index 3675fc55..8e268418 100644 --- a/src/vnet/vxlan-gpe/vxlan_gpe_api.c +++ b/src/vnet/vxlan-gpe/vxlan_gpe_api.c @@ -47,7 +47,7 @@ #define foreach_vpe_api_msg \ _(SW_INTERFACE_SET_VXLAN_GPE_BYPASS, sw_interface_set_vxlan_gpe_bypass) \ _(VXLAN_GPE_ADD_DEL_TUNNEL, vxlan_gpe_add_del_tunnel) \ -_(VXLAN_GPE_TUNNEL_DUMP, vxlan_gpe_tunnel_dump) \ +_(VXLAN_GPE_TUNNEL_DUMP, vxlan_gpe_tunnel_dump) static void vl_api_sw_interface_set_vxlan_gpe_bypass_t_handler @@ -156,15 +156,15 @@ static void send_vxlan_gpe_tunnel_details rmp->_vl_msg_id = ntohs (VL_API_VXLAN_GPE_TUNNEL_DETAILS); if (is_ipv6) { - memcpy (rmp->local, &(t->local.ip6), 16); - memcpy (rmp->remote, &(t->remote.ip6), 16); + memcpy (rmp->local, &(t->local.ip6.as_u8), 16); + memcpy (rmp->remote, &(t->remote.ip6.as_u8), 16); rmp->encap_vrf_id = htonl (im6->fibs[t->encap_fib_index].ft_table_id); rmp->decap_vrf_id = htonl (im6->fibs[t->decap_fib_index].ft_table_id); } else { - memcpy (rmp->local, &(t->local.ip4), 4); - memcpy (rmp->remote, &(t->remote.ip4), 4); + memcpy (rmp->local, &(t->local.ip4.as_u8), 4); + memcpy (rmp->remote, &(t->remote.ip4.as_u8), 4); rmp->encap_vrf_id = htonl (im4->fibs[t->encap_fib_index].ft_table_id); rmp->decap_vrf_id = htonl (im4->fibs[t->decap_fib_index].ft_table_id); } @@ -250,6 +250,9 @@ vxlan_gpe_api_hookup (vlib_main_t * vm) foreach_vpe_api_msg; #undef _ + am->api_trace_cfg[VL_API_VXLAN_GPE_ADD_DEL_TUNNEL].size += + 17 * sizeof (u32); + /* * Set up the (msg_name, crc, message-id) table */ diff --git a/src/vpp/api/custom_dump.c b/src/vpp/api/custom_dump.c index dc9fee9a..3ac8874e 100644 --- a/src/vpp/api/custom_dump.c +++ b/src/vpp/api/custom_dump.c @@ -1703,10 +1703,18 @@ static void *vl_api_vxlan_gpe_add_del_tunnel_t_print s = format (0, "SCRIPT: vxlan_gpe_add_del_tunnel "); - s = format (s, "local %U ", format_ip46_address, &mp->local, mp->is_ipv6); + ip46_address_t local = to_ip46 (mp->is_ipv6, mp->local); + ip46_address_t remote = to_ip46 (mp->is_ipv6, mp->remote); - s = format (s, "remote %U ", format_ip46_address, &mp->remote, mp->is_ipv6); + u8 is_grp = ip46_address_is_multicast (&remote); + char *remote_name = is_grp ? "group" : "remote"; + s = format (s, "local %U ", format_ip46_address, &local, IP46_TYPE_ANY); + s = format (s, "%s %U ", remote_name, format_ip46_address, + &remote, IP46_TYPE_ANY); + + if (is_grp) + s = format (s, "mcast_sw_if_index %d ", ntohl (mp->mcast_sw_if_index)); s = format (s, "protocol %d ", ntohl (mp->protocol)); s = format (s, "vni %d ", ntohl (mp->vni)); -- cgit 1.2.3-korg From 213b5aae860c2a9d5de8d4d070d0d2091af699f5 Mon Sep 17 00:00:00 2001 From: Damjan Marion Date: Thu, 13 Jul 2017 21:19:27 +0200 Subject: vnet_buffer_t flags cleanup Change-Id: I123eccea98abafeb31f25d2a162501e2eded60d4 Signed-off-by: Damjan Marion --- src/plugins/dpdk/device/init.c | 7 +++--- src/plugins/gtpu/gtpu_decap.c | 18 ++++++++-------- src/plugins/ixge/ixge.c | 12 +++++------ src/vnet/bfd/bfd_udp.c | 4 ++-- src/vnet/buffer.h | 49 +++++++++++++++++++++--------------------- src/vnet/ethernet/ethernet.h | 14 ++++++------ src/vnet/handoff.c | 2 +- src/vnet/ip/icmp4.c | 6 +++--- src/vnet/ip/ip4_forward.c | 38 ++++++++++++++++---------------- src/vnet/ip/ip6_forward.c | 44 +++++++++++++++++++++---------------- src/vnet/ip/ip6_neighbor.c | 4 ++-- src/vnet/session/node.c | 2 +- src/vnet/span/node.c | 4 ++-- src/vnet/tcp/tcp_output.c | 8 +++---- src/vnet/vxlan-gpe/decap.c | 18 ++++++++-------- src/vnet/vxlan/decap.c | 18 ++++++++-------- 16 files changed, 127 insertions(+), 121 deletions(-) (limited to 'src/vnet/vxlan-gpe') diff --git a/src/plugins/dpdk/device/init.c b/src/plugins/dpdk/device/init.c index 2ec1664b..7ca3d358 100755 --- a/src/plugins/dpdk/device/init.c +++ b/src/plugins/dpdk/device/init.c @@ -196,8 +196,8 @@ dpdk_lib_init (dpdk_main_t * dm) "dpdk rx"); if (dm->conf->enable_tcp_udp_checksum) - dm->buffer_flags_template &= ~(IP_BUFFER_L4_CHECKSUM_CORRECT - | IP_BUFFER_L4_CHECKSUM_COMPUTED); + dm->buffer_flags_template &= ~(VNET_BUFFER_F_L4_CHECKSUM_CORRECT + | VNET_BUFFER_F_L4_CHECKSUM_COMPUTED); /* vlib_buffer_t template */ vec_validate_aligned (dm->buffer_templates, tm->n_vlib_mains - 1, @@ -1544,7 +1544,8 @@ dpdk_init (vlib_main_t * vm) /* Default vlib_buffer_t flags, DISABLES tcp/udp checksumming... */ dm->buffer_flags_template = (VLIB_BUFFER_TOTAL_LENGTH_VALID | VLIB_BUFFER_EXT_HDR_VALID - | IP_BUFFER_L4_CHECKSUM_COMPUTED | IP_BUFFER_L4_CHECKSUM_CORRECT); + | VNET_BUFFER_F_L4_CHECKSUM_COMPUTED | + VNET_BUFFER_F_L4_CHECKSUM_CORRECT); dm->stat_poll_interval = DPDK_STATS_POLL_INTERVAL; dm->link_state_poll_interval = DPDK_LINK_POLL_INTERVAL; diff --git a/src/plugins/gtpu/gtpu_decap.c b/src/plugins/gtpu/gtpu_decap.c index fc74e7cb..de235889 100644 --- a/src/plugins/gtpu/gtpu_decap.c +++ b/src/plugins/gtpu/gtpu_decap.c @@ -982,7 +982,7 @@ ip_gtpu_bypass_inline (vlib_main_t * vm, } flags0 = b0->flags; - good_udp0 = (flags0 & IP_BUFFER_L4_CHECKSUM_CORRECT) != 0; + good_udp0 = (flags0 & VNET_BUFFER_F_L4_CHECKSUM_CORRECT) != 0; /* Don't verify UDP checksum for packets with explicit zero checksum. */ good_udp0 |= udp0->checksum == 0; @@ -998,14 +998,14 @@ ip_gtpu_bypass_inline (vlib_main_t * vm, /* Verify UDP checksum */ if (PREDICT_FALSE (!good_udp0)) { - if ((flags0 & IP_BUFFER_L4_CHECKSUM_COMPUTED) == 0) + if ((flags0 & VNET_BUFFER_F_L4_CHECKSUM_COMPUTED) == 0) { if (is_ip4) flags0 = ip4_tcp_udp_validate_checksum (vm, b0); else flags0 = ip6_tcp_udp_icmp_validate_checksum (vm, b0); good_udp0 = - (flags0 & IP_BUFFER_L4_CHECKSUM_CORRECT) != 0; + (flags0 & VNET_BUFFER_F_L4_CHECKSUM_CORRECT) != 0; } } @@ -1064,7 +1064,7 @@ ip_gtpu_bypass_inline (vlib_main_t * vm, } flags1 = b1->flags; - good_udp1 = (flags1 & IP_BUFFER_L4_CHECKSUM_CORRECT) != 0; + good_udp1 = (flags1 & VNET_BUFFER_F_L4_CHECKSUM_CORRECT) != 0; /* Don't verify UDP checksum for packets with explicit zero checksum. */ good_udp1 |= udp1->checksum == 0; @@ -1080,14 +1080,14 @@ ip_gtpu_bypass_inline (vlib_main_t * vm, /* Verify UDP checksum */ if (PREDICT_FALSE (!good_udp1)) { - if ((flags1 & IP_BUFFER_L4_CHECKSUM_COMPUTED) == 0) + if ((flags1 & VNET_BUFFER_F_L4_CHECKSUM_COMPUTED) == 0) { if (is_ip4) flags1 = ip4_tcp_udp_validate_checksum (vm, b1); else flags1 = ip6_tcp_udp_icmp_validate_checksum (vm, b1); good_udp1 = - (flags1 & IP_BUFFER_L4_CHECKSUM_CORRECT) != 0; + (flags1 & VNET_BUFFER_F_L4_CHECKSUM_CORRECT) != 0; } } @@ -1182,7 +1182,7 @@ ip_gtpu_bypass_inline (vlib_main_t * vm, } flags0 = b0->flags; - good_udp0 = (flags0 & IP_BUFFER_L4_CHECKSUM_CORRECT) != 0; + good_udp0 = (flags0 & VNET_BUFFER_F_L4_CHECKSUM_CORRECT) != 0; /* Don't verify UDP checksum for packets with explicit zero checksum. */ good_udp0 |= udp0->checksum == 0; @@ -1198,14 +1198,14 @@ ip_gtpu_bypass_inline (vlib_main_t * vm, /* Verify UDP checksum */ if (PREDICT_FALSE (!good_udp0)) { - if ((flags0 & IP_BUFFER_L4_CHECKSUM_COMPUTED) == 0) + if ((flags0 & VNET_BUFFER_F_L4_CHECKSUM_COMPUTED) == 0) { if (is_ip4) flags0 = ip4_tcp_udp_validate_checksum (vm, b0); else flags0 = ip6_tcp_udp_icmp_validate_checksum (vm, b0); good_udp0 = - (flags0 & IP_BUFFER_L4_CHECKSUM_CORRECT) != 0; + (flags0 & VNET_BUFFER_F_L4_CHECKSUM_CORRECT) != 0; } } diff --git a/src/plugins/ixge/ixge.c b/src/plugins/ixge/ixge.c index 628d6d71..e0150f41 100644 --- a/src/plugins/ixge/ixge.c +++ b/src/plugins/ixge/ixge.c @@ -656,11 +656,11 @@ ixge_rx_next_and_error_from_status_x1 (ixge_device_t * xd, f0 = ((s02 & (IXGE_RX_DESCRIPTOR_STATUS2_IS_TCP_CHECKSUMMED | IXGE_RX_DESCRIPTOR_STATUS2_IS_UDP_CHECKSUMMED)) - ? IP_BUFFER_L4_CHECKSUM_COMPUTED : 0); + ? VNET_BUFFER_F_L4_CHECKSUM_COMPUTED : 0); f0 |= ((s02 & (IXGE_RX_DESCRIPTOR_STATUS2_TCP_CHECKSUM_ERROR | IXGE_RX_DESCRIPTOR_STATUS2_UDP_CHECKSUM_ERROR)) - ? 0 : IP_BUFFER_L4_CHECKSUM_CORRECT); + ? 0 : VNET_BUFFER_F_L4_CHECKSUM_CORRECT); *error0 = e0; *next0 = n0; @@ -715,17 +715,17 @@ ixge_rx_next_and_error_from_status_x2 (ixge_device_t * xd, f0 = ((s02 & (IXGE_RX_DESCRIPTOR_STATUS2_IS_TCP_CHECKSUMMED | IXGE_RX_DESCRIPTOR_STATUS2_IS_UDP_CHECKSUMMED)) - ? IP_BUFFER_L4_CHECKSUM_COMPUTED : 0); + ? VNET_BUFFER_F_L4_CHECKSUM_COMPUTED : 0); f1 = ((s12 & (IXGE_RX_DESCRIPTOR_STATUS2_IS_TCP_CHECKSUMMED | IXGE_RX_DESCRIPTOR_STATUS2_IS_UDP_CHECKSUMMED)) - ? IP_BUFFER_L4_CHECKSUM_COMPUTED : 0); + ? VNET_BUFFER_F_L4_CHECKSUM_COMPUTED : 0); f0 |= ((s02 & (IXGE_RX_DESCRIPTOR_STATUS2_TCP_CHECKSUM_ERROR | IXGE_RX_DESCRIPTOR_STATUS2_UDP_CHECKSUM_ERROR)) - ? 0 : IP_BUFFER_L4_CHECKSUM_CORRECT); + ? 0 : VNET_BUFFER_F_L4_CHECKSUM_CORRECT); f1 |= ((s12 & (IXGE_RX_DESCRIPTOR_STATUS2_TCP_CHECKSUM_ERROR | IXGE_RX_DESCRIPTOR_STATUS2_UDP_CHECKSUM_ERROR)) - ? 0 : IP_BUFFER_L4_CHECKSUM_CORRECT); + ? 0 : VNET_BUFFER_F_L4_CHECKSUM_CORRECT); *flags0 = f0; *flags1 = f1; diff --git a/src/vnet/bfd/bfd_udp.c b/src/vnet/bfd/bfd_udp.c index 06b843c6..533d98d6 100644 --- a/src/vnet/bfd/bfd_udp.c +++ b/src/vnet/bfd/bfd_udp.c @@ -246,7 +246,7 @@ bfd_add_udp4_transport (vlib_main_t * vm, u32 bi, const bfd_session_t * bs, const bfd_udp_key_t *key = &bus->key; vlib_buffer_t *b = vlib_get_buffer (vm, bi); - b->flags |= VNET_BUFFER_LOCALLY_ORIGINATED; + b->flags |= VNET_BUFFER_F_LOCALLY_ORIGINATED; vnet_buffer (b)->ip.adj_index[VLIB_RX] = bus->adj_index; vnet_buffer (b)->ip.adj_index[VLIB_TX] = bus->adj_index; vnet_buffer (b)->sw_if_index[VLIB_RX] = 0; @@ -301,7 +301,7 @@ bfd_add_udp6_transport (vlib_main_t * vm, u32 bi, const bfd_session_t * bs, const bfd_udp_key_t *key = &bus->key; vlib_buffer_t *b = vlib_get_buffer (vm, bi); - b->flags |= VNET_BUFFER_LOCALLY_ORIGINATED; + b->flags |= VNET_BUFFER_F_LOCALLY_ORIGINATED; vnet_buffer (b)->ip.adj_index[VLIB_RX] = bus->adj_index; vnet_buffer (b)->ip.adj_index[VLIB_TX] = bus->adj_index; vnet_buffer (b)->sw_if_index[VLIB_RX] = 0; diff --git a/src/vnet/buffer.h b/src/vnet/buffer.h index 8647db00..52dada30 100644 --- a/src/vnet/buffer.h +++ b/src/vnet/buffer.h @@ -42,33 +42,32 @@ #include -/* VLIB buffer flags for ip4/ip6 packets. Set by input interfaces for ip4/ip6 - tcp/udp packets with hardware computed checksums. */ -#define LOG2_IP_BUFFER_L4_CHECKSUM_COMPUTED LOG2_VLIB_BUFFER_FLAG_USER(1) -#define LOG2_IP_BUFFER_L4_CHECKSUM_CORRECT LOG2_VLIB_BUFFER_FLAG_USER(2) -#define IP_BUFFER_L4_CHECKSUM_COMPUTED (1 << LOG2_IP_BUFFER_L4_CHECKSUM_COMPUTED) -#define IP_BUFFER_L4_CHECKSUM_CORRECT (1 << LOG2_IP_BUFFER_L4_CHECKSUM_CORRECT) - -/* VLAN header flags. - * These bits are zeroed in vlib_buffer_init_for_free_list() - * meaning wherever the buffer comes from they have a reasonable - * value (eg, if ip4/ip6 generates the packet.) - */ -#define LOG2_ETH_BUFFER_VLAN_2_DEEP LOG2_VLIB_BUFFER_FLAG_USER(3) -#define LOG2_ETH_BUFFER_VLAN_1_DEEP LOG2_VLIB_BUFFER_FLAG_USER(4) -#define ETH_BUFFER_VLAN_2_DEEP (1 << LOG2_ETH_BUFFER_VLAN_2_DEEP) -#define ETH_BUFFER_VLAN_1_DEEP (1 << LOG2_ETH_BUFFER_VLAN_1_DEEP) -#define ETH_BUFFER_VLAN_BITS (ETH_BUFFER_VLAN_1_DEEP | \ - ETH_BUFFER_VLAN_2_DEEP) - -#define LOG2_BUFFER_HANDOFF_NEXT_VALID LOG2_VLIB_BUFFER_FLAG_USER(6) -#define BUFFER_HANDOFF_NEXT_VALID (1 << LOG2_BUFFER_HANDOFF_NEXT_VALID) +#define foreach_vnet_buffer_field \ + _( 1, L4_CHECKSUM_COMPUTED) \ + _( 2, L4_CHECKSUM_CORRECT) \ + _( 3, VLAN_2_DEEP) \ + _( 4, VLAN_1_DEEP) \ + _( 6, HANDOFF_NEXT_VALID) \ + _( 7, LOCALLY_ORIGINATED) \ + _( 8, SPAN_CLONE) + +#define VNET_BUFFER_FLAGS_VLAN_BITS \ + (VNET_BUFFER_F_VLAN_1_DEEP | VNET_BUFFER_F_VLAN_2_DEEP) + +enum +{ +#define _(bit, name) VNET_BUFFER_F_##name = (1 << LOG2_VLIB_BUFFER_FLAG_USER(bit)), + foreach_vnet_buffer_field +#undef _ +}; -#define LOG2_VNET_BUFFER_LOCALLY_ORIGINATED LOG2_VLIB_BUFFER_FLAG_USER(7) -#define VNET_BUFFER_LOCALLY_ORIGINATED (1 << LOG2_VNET_BUFFER_LOCALLY_ORIGINATED) +enum +{ +#define _(bit, name) VNET_BUFFER_F_LOG2_##name = LOG2_VLIB_BUFFER_FLAG_USER(bit), + foreach_vnet_buffer_field +#undef _ +}; -#define LOG2_VNET_BUFFER_SPAN_CLONE LOG2_VLIB_BUFFER_FLAG_USER(8) -#define VNET_BUFFER_SPAN_CLONE (1 << LOG2_VNET_BUFFER_SPAN_CLONE) #define foreach_buffer_opaque_union_subtype \ _(ip) \ diff --git a/src/vnet/ethernet/ethernet.h b/src/vnet/ethernet/ethernet.h index 2fc5b804..d9ab8c10 100644 --- a/src/vnet/ethernet/ethernet.h +++ b/src/vnet/ethernet/ethernet.h @@ -352,7 +352,7 @@ ethernet_buffer_get_header (vlib_buffer_t * b) * the number of headers is not known. */ #define ethernet_buffer_get_vlan_count(b) ( \ - ((b)->flags & ETH_BUFFER_VLAN_BITS) >> LOG2_ETH_BUFFER_VLAN_1_DEEP \ + ((b)->flags & VNET_BUFFER_FLAGS_VLAN_BITS) >> VNET_BUFFER_F_LOG2_VLAN_1_DEEP \ ) /** Sets the number of VLAN headers in the current Ethernet frame in the @@ -360,8 +360,8 @@ ethernet_buffer_get_header (vlib_buffer_t * b) * the number of headers is not known. */ #define ethernet_buffer_set_vlan_count(b, v) ( \ - (b)->flags = ((b)->flags & ~ETH_BUFFER_VLAN_BITS) | \ - (((v) << LOG2_ETH_BUFFER_VLAN_1_DEEP) & ETH_BUFFER_VLAN_BITS) \ + (b)->flags = ((b)->flags & ~VNET_BUFFER_FLAGS_VLAN_BITS) | \ + (((v) << VNET_BUFFER_F_LOG2_VLAN_1_DEEP) & VNET_BUFFER_FLAGS_VLAN_BITS) \ ) /** Adjusts the vlan count by the delta in 'v' */ @@ -372,10 +372,10 @@ ethernet_buffer_get_header (vlib_buffer_t * b) /** Adjusts the vlan count by the header size byte delta in 'v' */ #define ethernet_buffer_adjust_vlan_count_by_bytes(b, v) ( \ - (b)->flags = ((b)->flags & ~ETH_BUFFER_VLAN_BITS) | (( \ - ((b)->flags & ETH_BUFFER_VLAN_BITS) + \ - ((v) << (LOG2_ETH_BUFFER_VLAN_1_DEEP - 2)) \ - ) & ETH_BUFFER_VLAN_BITS) \ + (b)->flags = ((b)->flags & ~VNET_BUFFER_FLAGS_VLAN_BITS) | (( \ + ((b)->flags & VNET_BUFFER_FLAGS_VLAN_BITS) + \ + ((v) << (VNET_BUFFER_F_LOG2_VLAN_1_DEEP - 2)) \ + ) & VNET_BUFFER_FLAGS_VLAN_BITS) \ ) /** diff --git a/src/vnet/handoff.c b/src/vnet/handoff.c index 9f3c93b4..81cb9f55 100644 --- a/src/vnet/handoff.c +++ b/src/vnet/handoff.c @@ -130,7 +130,7 @@ worker_handoff_node_fn (vlib_main_t * vm, /* if input node did not specify next index, then packet should go to eternet-input */ - if (PREDICT_FALSE ((b0->flags & BUFFER_HANDOFF_NEXT_VALID) == 0)) + if (PREDICT_FALSE ((b0->flags & VNET_BUFFER_F_HANDOFF_NEXT_VALID) == 0)) vnet_buffer (b0)->handoff.next_index = HANDOFF_DISPATCH_NEXT_ETHERNET_INPUT; else if (vnet_buffer (b0)->handoff.next_index == diff --git a/src/vnet/ip/icmp4.c b/src/vnet/ip/icmp4.c index c3afff72..bbeab32b 100644 --- a/src/vnet/ip/icmp4.c +++ b/src/vnet/ip/icmp4.c @@ -336,8 +336,8 @@ ip4_icmp_echo_request (vlib_main_t * vm, ASSERT (ip0->checksum == ip4_header_checksum (ip0)); ASSERT (ip1->checksum == ip4_header_checksum (ip1)); - p0->flags |= VNET_BUFFER_LOCALLY_ORIGINATED; - p1->flags |= VNET_BUFFER_LOCALLY_ORIGINATED; + p0->flags |= VNET_BUFFER_F_LOCALLY_ORIGINATED; + p1->flags |= VNET_BUFFER_F_LOCALLY_ORIGINATED; } while (n_left_from > 0 && n_left_to_next > 0) @@ -392,7 +392,7 @@ ip4_icmp_echo_request (vlib_main_t * vm, ASSERT (ip0->checksum == ip4_header_checksum (ip0)); - p0->flags |= VNET_BUFFER_LOCALLY_ORIGINATED; + p0->flags |= VNET_BUFFER_F_LOCALLY_ORIGINATED; } vlib_put_next_frame (vm, node, next, n_left_to_next); diff --git a/src/vnet/ip/ip4_forward.c b/src/vnet/ip/ip4_forward.c index b8dfa847..8dd927d4 100755 --- a/src/vnet/ip/ip4_forward.c +++ b/src/vnet/ip/ip4_forward.c @@ -1509,15 +1509,15 @@ ip4_tcp_udp_validate_checksum (vlib_main_t * vm, vlib_buffer_t * p0) 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); + p0->flags |= (VNET_BUFFER_F_L4_CHECKSUM_COMPUTED + | VNET_BUFFER_F_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)); + p0->flags |= (VNET_BUFFER_F_L4_CHECKSUM_COMPUTED + | ((sum16 == 0) << VNET_BUFFER_F_LOG2_L4_CHECKSUM_CORRECT)); return p0->flags; } @@ -1629,8 +1629,8 @@ ip4_local_inline (vlib_main_t * vm, 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; + good_tcp_udp0 = (flags0 & VNET_BUFFER_F_L4_CHECKSUM_CORRECT) != 0; + good_tcp_udp1 = (flags1 & VNET_BUFFER_F_L4_CHECKSUM_CORRECT) != 0; udp0 = ip4_next_header (ip0); udp1 = ip4_next_header (ip1); @@ -1657,19 +1657,19 @@ ip4_local_inline (vlib_main_t * vm, if (is_tcp_udp0) { if (is_tcp_udp0 - && !(flags0 & IP_BUFFER_L4_CHECKSUM_COMPUTED)) + && !(flags0 & VNET_BUFFER_F_L4_CHECKSUM_COMPUTED)) flags0 = ip4_tcp_udp_validate_checksum (vm, p0); good_tcp_udp0 = - (flags0 & IP_BUFFER_L4_CHECKSUM_CORRECT) != 0; + (flags0 & VNET_BUFFER_F_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 & VNET_BUFFER_F_L4_CHECKSUM_COMPUTED)) flags1 = ip4_tcp_udp_validate_checksum (vm, p1); good_tcp_udp1 = - (flags1 & IP_BUFFER_L4_CHECKSUM_CORRECT) != 0; + (flags1 & VNET_BUFFER_F_L4_CHECKSUM_CORRECT) != 0; good_tcp_udp1 |= is_udp1 && udp1->checksum == 0; } } @@ -1817,7 +1817,7 @@ ip4_local_inline (vlib_main_t * vm, flags0 = p0->flags; - good_tcp_udp0 = (flags0 & IP_BUFFER_L4_CHECKSUM_CORRECT) != 0; + good_tcp_udp0 = (flags0 & VNET_BUFFER_F_L4_CHECKSUM_CORRECT) != 0; udp0 = ip4_next_header (ip0); @@ -1837,10 +1837,10 @@ ip4_local_inline (vlib_main_t * vm, if (is_tcp_udp0) { if (is_tcp_udp0 - && !(flags0 & IP_BUFFER_L4_CHECKSUM_COMPUTED)) + && !(flags0 & VNET_BUFFER_F_L4_CHECKSUM_COMPUTED)) flags0 = ip4_tcp_udp_validate_checksum (vm, p0); good_tcp_udp0 = - (flags0 & IP_BUFFER_L4_CHECKSUM_CORRECT) != 0; + (flags0 & VNET_BUFFER_F_L4_CHECKSUM_CORRECT) != 0; good_tcp_udp0 |= is_udp0 && udp0->checksum == 0; } } @@ -2428,7 +2428,7 @@ ip4_rewrite_inline (vlib_main_t * vm, /* Decrement TTL & update checksum. Works either endian, so no need for byte swap. */ - if (PREDICT_TRUE (!(p0->flags & VNET_BUFFER_LOCALLY_ORIGINATED))) + if (PREDICT_TRUE (!(p0->flags & VNET_BUFFER_F_LOCALLY_ORIGINATED))) { i32 ttl0 = ip0->ttl; @@ -2461,9 +2461,9 @@ ip4_rewrite_inline (vlib_main_t * vm, } else { - p0->flags &= ~VNET_BUFFER_LOCALLY_ORIGINATED; + p0->flags &= ~VNET_BUFFER_F_LOCALLY_ORIGINATED; } - if (PREDICT_TRUE (!(p1->flags & VNET_BUFFER_LOCALLY_ORIGINATED))) + if (PREDICT_TRUE (!(p1->flags & VNET_BUFFER_F_LOCALLY_ORIGINATED))) { i32 ttl1 = ip1->ttl; @@ -2496,7 +2496,7 @@ ip4_rewrite_inline (vlib_main_t * vm, } else { - p1->flags &= ~VNET_BUFFER_LOCALLY_ORIGINATED; + p1->flags &= ~VNET_BUFFER_F_LOCALLY_ORIGINATED; } /* Rewrite packet header and updates lengths. */ @@ -2614,7 +2614,7 @@ ip4_rewrite_inline (vlib_main_t * vm, next0 = IP4_REWRITE_NEXT_DROP; /* drop on error */ /* Decrement TTL & update checksum. */ - if (PREDICT_TRUE (!(p0->flags & VNET_BUFFER_LOCALLY_ORIGINATED))) + if (PREDICT_TRUE (!(p0->flags & VNET_BUFFER_F_LOCALLY_ORIGINATED))) { i32 ttl0 = ip0->ttl; @@ -2648,7 +2648,7 @@ ip4_rewrite_inline (vlib_main_t * vm, } else { - p0->flags &= ~VNET_BUFFER_LOCALLY_ORIGINATED; + p0->flags &= ~VNET_BUFFER_F_LOCALLY_ORIGINATED; } if (do_counters) diff --git a/src/vnet/ip/ip6_forward.c b/src/vnet/ip/ip6_forward.c index 2b8c2bd2..604e1492 100644 --- a/src/vnet/ip/ip6_forward.c +++ b/src/vnet/ip/ip6_forward.c @@ -1284,15 +1284,15 @@ ip6_tcp_udp_icmp_validate_checksum (vlib_main_t * vm, vlib_buffer_t * p0) 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); + p0->flags |= (VNET_BUFFER_F_L4_CHECKSUM_COMPUTED + | VNET_BUFFER_F_L4_CHECKSUM_CORRECT); return p0->flags; } sum16 = ip6_tcp_udp_icmp_compute_checksum (vm, p0, ip0, &bogus_length); - p0->flags |= (IP_BUFFER_L4_CHECKSUM_COMPUTED - | ((sum16 == 0) << LOG2_IP_BUFFER_L4_CHECKSUM_CORRECT)); + p0->flags |= (VNET_BUFFER_F_L4_CHECKSUM_COMPUTED + | ((sum16 == 0) << VNET_BUFFER_F_LOG2_L4_CHECKSUM_CORRECT)); return p0->flags; } @@ -1374,8 +1374,10 @@ ip6_local (vlib_main_t * vm, vlib_node_runtime_t * node, vlib_frame_t * frame) flags0 = p0->flags; flags1 = p1->flags; - good_l4_checksum0 = (flags0 & IP_BUFFER_L4_CHECKSUM_CORRECT) != 0; - good_l4_checksum1 = (flags1 & IP_BUFFER_L4_CHECKSUM_CORRECT) != 0; + good_l4_checksum0 = + (flags0 & VNET_BUFFER_F_L4_CHECKSUM_CORRECT) != 0; + good_l4_checksum1 = + (flags1 & VNET_BUFFER_F_L4_CHECKSUM_CORRECT) != 0; len_diff0 = 0; len_diff1 = 0; @@ -1414,19 +1416,21 @@ ip6_local (vlib_main_t * vm, vlib_node_runtime_t * node, vlib_frame_t * frame) if (PREDICT_FALSE (type0 != IP_BUILTIN_PROTOCOL_UNKNOWN && !good_l4_checksum0 - && !(flags0 & IP_BUFFER_L4_CHECKSUM_COMPUTED))) + && !(flags0 & + VNET_BUFFER_F_L4_CHECKSUM_COMPUTED))) { flags0 = ip6_tcp_udp_icmp_validate_checksum (vm, p0); good_l4_checksum0 = - (flags0 & IP_BUFFER_L4_CHECKSUM_CORRECT) != 0; + (flags0 & VNET_BUFFER_F_L4_CHECKSUM_CORRECT) != 0; } if (PREDICT_FALSE (type1 != IP_BUILTIN_PROTOCOL_UNKNOWN && !good_l4_checksum1 - && !(flags1 & IP_BUFFER_L4_CHECKSUM_COMPUTED))) + && !(flags1 & + VNET_BUFFER_F_L4_CHECKSUM_COMPUTED))) { flags1 = ip6_tcp_udp_icmp_validate_checksum (vm, p1); good_l4_checksum1 = - (flags1 & IP_BUFFER_L4_CHECKSUM_CORRECT) != 0; + (flags1 & VNET_BUFFER_F_L4_CHECKSUM_CORRECT) != 0; } error0 = error1 = IP6_ERROR_UNKNOWN_PROTOCOL; @@ -1500,7 +1504,8 @@ ip6_local (vlib_main_t * vm, vlib_node_runtime_t * node, vlib_frame_t * frame) flags0 = p0->flags; - good_l4_checksum0 = (flags0 & IP_BUFFER_L4_CHECKSUM_CORRECT) != 0; + good_l4_checksum0 = + (flags0 & VNET_BUFFER_F_L4_CHECKSUM_CORRECT) != 0; len_diff0 = 0; if (PREDICT_TRUE (IP_PROTOCOL_UDP == ip6_locate_header (p0, ip0, @@ -1522,11 +1527,12 @@ ip6_local (vlib_main_t * vm, vlib_node_runtime_t * node, vlib_frame_t * frame) if (PREDICT_FALSE (type0 != IP_BUILTIN_PROTOCOL_UNKNOWN && !good_l4_checksum0 - && !(flags0 & IP_BUFFER_L4_CHECKSUM_COMPUTED))) + && !(flags0 & + VNET_BUFFER_F_L4_CHECKSUM_COMPUTED))) { flags0 = ip6_tcp_udp_icmp_validate_checksum (vm, p0); good_l4_checksum0 = - (flags0 & IP_BUFFER_L4_CHECKSUM_CORRECT) != 0; + (flags0 & VNET_BUFFER_F_L4_CHECKSUM_CORRECT) != 0; } error0 = IP6_ERROR_UNKNOWN_PROTOCOL; @@ -2019,7 +2025,7 @@ ip6_rewrite_inline (vlib_main_t * vm, error0 = error1 = IP6_ERROR_NONE; next0 = next1 = IP6_REWRITE_NEXT_DROP; - if (PREDICT_TRUE (!(p0->flags & VNET_BUFFER_LOCALLY_ORIGINATED))) + if (PREDICT_TRUE (!(p0->flags & VNET_BUFFER_F_LOCALLY_ORIGINATED))) { i32 hop_limit0 = ip0->hop_limit; @@ -2046,9 +2052,9 @@ ip6_rewrite_inline (vlib_main_t * vm, } else { - p0->flags &= ~VNET_BUFFER_LOCALLY_ORIGINATED; + p0->flags &= ~VNET_BUFFER_F_LOCALLY_ORIGINATED; } - if (PREDICT_TRUE (!(p1->flags & VNET_BUFFER_LOCALLY_ORIGINATED))) + if (PREDICT_TRUE (!(p1->flags & VNET_BUFFER_F_LOCALLY_ORIGINATED))) { i32 hop_limit1 = ip1->hop_limit; @@ -2075,7 +2081,7 @@ ip6_rewrite_inline (vlib_main_t * vm, } else { - p1->flags &= ~VNET_BUFFER_LOCALLY_ORIGINATED; + p1->flags &= ~VNET_BUFFER_F_LOCALLY_ORIGINATED; } adj0 = adj_get (adj_index0); adj1 = adj_get (adj_index1); @@ -2186,7 +2192,7 @@ ip6_rewrite_inline (vlib_main_t * vm, next0 = IP6_REWRITE_NEXT_DROP; /* Check hop limit */ - if (PREDICT_TRUE (!(p0->flags & VNET_BUFFER_LOCALLY_ORIGINATED))) + if (PREDICT_TRUE (!(p0->flags & VNET_BUFFER_F_LOCALLY_ORIGINATED))) { i32 hop_limit0 = ip0->hop_limit; @@ -2212,7 +2218,7 @@ ip6_rewrite_inline (vlib_main_t * vm, } else { - p0->flags &= ~VNET_BUFFER_LOCALLY_ORIGINATED; + p0->flags &= ~VNET_BUFFER_F_LOCALLY_ORIGINATED; } /* Guess we are only writing on simple Ethernet header. */ diff --git a/src/vnet/ip/ip6_neighbor.c b/src/vnet/ip/ip6_neighbor.c index 68a8cbbc..e8eebd4e 100644 --- a/src/vnet/ip/ip6_neighbor.c +++ b/src/vnet/ip/ip6_neighbor.c @@ -1649,7 +1649,7 @@ icmp6_router_solicitation (vlib_main_t * vm, adj_index0; } } - p0->flags |= VNET_BUFFER_LOCALLY_ORIGINATED; + p0->flags |= VNET_BUFFER_F_LOCALLY_ORIGINATED; radv_info->n_solicitations_dropped += is_dropped; radv_info->n_solicitations_rcvd += is_solicitation; @@ -2348,7 +2348,7 @@ ip6_neighbor_send_mldpv2_report (u32 sw_if_index) vnet_main.local_interface_sw_if_index; vnet_buffer (b0)->ip.adj_index[VLIB_TX] = radv_info->mcast_adj_index; - b0->flags |= VNET_BUFFER_LOCALLY_ORIGINATED; + b0->flags |= VNET_BUFFER_F_LOCALLY_ORIGINATED; vlib_node_t *node = vlib_get_node_by_name (vm, (u8 *) "ip6-rewrite-mcast"); diff --git a/src/vnet/session/node.c b/src/vnet/session/node.c index 56e62637..983b78b8 100644 --- a/src/vnet/session/node.c +++ b/src/vnet/session/node.c @@ -256,7 +256,7 @@ session_tx_fifo_read_and_snd_i (vlib_main_t * vm, vlib_node_runtime_t * node, b0 = vlib_get_buffer (vm, bi0); b0->error = 0; b0->flags = VLIB_BUFFER_TOTAL_LENGTH_VALID - | VNET_BUFFER_LOCALLY_ORIGINATED; + | VNET_BUFFER_F_LOCALLY_ORIGINATED; b0->current_data = 0; b0->total_length_not_including_first_buffer = 0; diff --git a/src/vnet/span/node.c b/src/vnet/span/node.c index 2a43b6e3..3a461b0a 100644 --- a/src/vnet/span/node.c +++ b/src/vnet/span/node.c @@ -77,7 +77,7 @@ span_mirror (vlib_main_t * vm, vlib_node_runtime_t * node, u32 sw_if_index0, return; /* Don't do it again */ - if (PREDICT_FALSE (b0->flags & VNET_BUFFER_SPAN_CLONE)) + if (PREDICT_FALSE (b0->flags & VNET_BUFFER_F_SPAN_CLONE)) return; /* *INDENT-OFF* */ @@ -92,7 +92,7 @@ span_mirror (vlib_main_t * vm, vlib_node_runtime_t * node, u32 sw_if_index0, if (PREDICT_TRUE(c0 != 0)) { vnet_buffer (c0)->sw_if_index[VLIB_TX] = i; - c0->flags |= VNET_BUFFER_SPAN_CLONE; + c0->flags |= VNET_BUFFER_F_SPAN_CLONE; to_mirror_next[0] = vlib_get_buffer_index (vm, c0); mirror_frames[i]->n_vectors++; if (PREDICT_FALSE (b0->flags & VLIB_BUFFER_IS_TRACED)) diff --git a/src/vnet/tcp/tcp_output.c b/src/vnet/tcp/tcp_output.c index b418e8ba..f34eb797 100644 --- a/src/vnet/tcp/tcp_output.c +++ b/src/vnet/tcp/tcp_output.c @@ -585,7 +585,7 @@ tcp_enqueue_to_ip_lookup (vlib_main_t * vm, vlib_buffer_t * b, u32 bi, u32 *to_next, next_index; vlib_frame_t *f; - b->flags |= VNET_BUFFER_LOCALLY_ORIGINATED; + b->flags |= VNET_BUFFER_F_LOCALLY_ORIGINATED; b->error = 0; /* Default FIB for now */ @@ -847,7 +847,7 @@ tcp_enqueue_to_output (vlib_main_t * vm, vlib_buffer_t * b, u32 bi, u8 is_ip4) u32 *to_next, next_index; vlib_frame_t *f; - b->flags |= VNET_BUFFER_LOCALLY_ORIGINATED; + b->flags |= VNET_BUFFER_F_LOCALLY_ORIGINATED; b->error = 0; /* Decide where to send the packet */ @@ -1563,7 +1563,7 @@ tcp46_output_inline (vlib_main_t * vm, vnet_buffer (b0)->sw_if_index[VLIB_RX] = 0; vnet_buffer (b0)->sw_if_index[VLIB_TX] = ~0; - b0->flags |= VNET_BUFFER_LOCALLY_ORIGINATED; + b0->flags |= VNET_BUFFER_F_LOCALLY_ORIGINATED; done: b0->error = node->errors[error0]; if (PREDICT_FALSE (b0->flags & VLIB_BUFFER_IS_TRACED)) @@ -1728,7 +1728,7 @@ tcp46_send_reset_inline (vlib_main_t * vm, vlib_node_runtime_t * node, done: b0->error = node->errors[error0]; - b0->flags |= VNET_BUFFER_LOCALLY_ORIGINATED; + b0->flags |= VNET_BUFFER_F_LOCALLY_ORIGINATED; if (PREDICT_FALSE (b0->flags & VLIB_BUFFER_IS_TRACED)) { th0 = vlib_buffer_get_current (b0); diff --git a/src/vnet/vxlan-gpe/decap.c b/src/vnet/vxlan-gpe/decap.c index 075b0f51..1b3a8b00 100644 --- a/src/vnet/vxlan-gpe/decap.c +++ b/src/vnet/vxlan-gpe/decap.c @@ -857,7 +857,7 @@ ip_vxlan_gpe_bypass_inline (vlib_main_t * vm, } flags0 = b0->flags; - good_udp0 = (flags0 & IP_BUFFER_L4_CHECKSUM_CORRECT) != 0; + good_udp0 = (flags0 & VNET_BUFFER_F_L4_CHECKSUM_CORRECT) != 0; /* Don't verify UDP checksum for packets with explicit zero checksum. */ good_udp0 |= udp0->checksum == 0; @@ -873,14 +873,14 @@ ip_vxlan_gpe_bypass_inline (vlib_main_t * vm, /* Verify UDP checksum */ if (PREDICT_FALSE (!good_udp0)) { - if ((flags0 & IP_BUFFER_L4_CHECKSUM_COMPUTED) == 0) + if ((flags0 & VNET_BUFFER_F_L4_CHECKSUM_COMPUTED) == 0) { if (is_ip4) flags0 = ip4_tcp_udp_validate_checksum (vm, b0); else flags0 = ip6_tcp_udp_icmp_validate_checksum (vm, b0); good_udp0 = - (flags0 & IP_BUFFER_L4_CHECKSUM_CORRECT) != 0; + (flags0 & VNET_BUFFER_F_L4_CHECKSUM_CORRECT) != 0; } } @@ -939,7 +939,7 @@ ip_vxlan_gpe_bypass_inline (vlib_main_t * vm, } flags1 = b1->flags; - good_udp1 = (flags1 & IP_BUFFER_L4_CHECKSUM_CORRECT) != 0; + good_udp1 = (flags1 & VNET_BUFFER_F_L4_CHECKSUM_CORRECT) != 0; /* Don't verify UDP checksum for packets with explicit zero checksum. */ good_udp1 |= udp1->checksum == 0; @@ -955,14 +955,14 @@ ip_vxlan_gpe_bypass_inline (vlib_main_t * vm, /* Verify UDP checksum */ if (PREDICT_FALSE (!good_udp1)) { - if ((flags1 & IP_BUFFER_L4_CHECKSUM_COMPUTED) == 0) + if ((flags1 & VNET_BUFFER_F_L4_CHECKSUM_COMPUTED) == 0) { if (is_ip4) flags1 = ip4_tcp_udp_validate_checksum (vm, b1); else flags1 = ip6_tcp_udp_icmp_validate_checksum (vm, b1); good_udp1 = - (flags1 & IP_BUFFER_L4_CHECKSUM_CORRECT) != 0; + (flags1 & VNET_BUFFER_F_L4_CHECKSUM_CORRECT) != 0; } } @@ -1055,7 +1055,7 @@ ip_vxlan_gpe_bypass_inline (vlib_main_t * vm, } flags0 = b0->flags; - good_udp0 = (flags0 & IP_BUFFER_L4_CHECKSUM_CORRECT) != 0; + good_udp0 = (flags0 & VNET_BUFFER_F_L4_CHECKSUM_CORRECT) != 0; /* Don't verify UDP checksum for packets with explicit zero checksum. */ good_udp0 |= udp0->checksum == 0; @@ -1071,14 +1071,14 @@ ip_vxlan_gpe_bypass_inline (vlib_main_t * vm, /* Verify UDP checksum */ if (PREDICT_FALSE (!good_udp0)) { - if ((flags0 & IP_BUFFER_L4_CHECKSUM_COMPUTED) == 0) + if ((flags0 & VNET_BUFFER_F_L4_CHECKSUM_COMPUTED) == 0) { if (is_ip4) flags0 = ip4_tcp_udp_validate_checksum (vm, b0); else flags0 = ip6_tcp_udp_icmp_validate_checksum (vm, b0); good_udp0 = - (flags0 & IP_BUFFER_L4_CHECKSUM_CORRECT) != 0; + (flags0 & VNET_BUFFER_F_L4_CHECKSUM_CORRECT) != 0; } } diff --git a/src/vnet/vxlan/decap.c b/src/vnet/vxlan/decap.c index 2acb1f6f..0dc89d3f 100644 --- a/src/vnet/vxlan/decap.c +++ b/src/vnet/vxlan/decap.c @@ -916,7 +916,7 @@ ip_vxlan_bypass_inline (vlib_main_t * vm, } flags0 = b0->flags; - good_udp0 = (flags0 & IP_BUFFER_L4_CHECKSUM_CORRECT) != 0; + good_udp0 = (flags0 & VNET_BUFFER_F_L4_CHECKSUM_CORRECT) != 0; /* Don't verify UDP checksum for packets with explicit zero checksum. */ good_udp0 |= udp0->checksum == 0; @@ -932,14 +932,14 @@ ip_vxlan_bypass_inline (vlib_main_t * vm, /* Verify UDP checksum */ if (PREDICT_FALSE (!good_udp0)) { - if ((flags0 & IP_BUFFER_L4_CHECKSUM_COMPUTED) == 0) + if ((flags0 & VNET_BUFFER_F_L4_CHECKSUM_COMPUTED) == 0) { if (is_ip4) flags0 = ip4_tcp_udp_validate_checksum (vm, b0); else flags0 = ip6_tcp_udp_icmp_validate_checksum (vm, b0); good_udp0 = - (flags0 & IP_BUFFER_L4_CHECKSUM_CORRECT) != 0; + (flags0 & VNET_BUFFER_F_L4_CHECKSUM_CORRECT) != 0; } } @@ -998,7 +998,7 @@ ip_vxlan_bypass_inline (vlib_main_t * vm, } flags1 = b1->flags; - good_udp1 = (flags1 & IP_BUFFER_L4_CHECKSUM_CORRECT) != 0; + good_udp1 = (flags1 & VNET_BUFFER_F_L4_CHECKSUM_CORRECT) != 0; /* Don't verify UDP checksum for packets with explicit zero checksum. */ good_udp1 |= udp1->checksum == 0; @@ -1014,14 +1014,14 @@ ip_vxlan_bypass_inline (vlib_main_t * vm, /* Verify UDP checksum */ if (PREDICT_FALSE (!good_udp1)) { - if ((flags1 & IP_BUFFER_L4_CHECKSUM_COMPUTED) == 0) + if ((flags1 & VNET_BUFFER_F_L4_CHECKSUM_COMPUTED) == 0) { if (is_ip4) flags1 = ip4_tcp_udp_validate_checksum (vm, b1); else flags1 = ip6_tcp_udp_icmp_validate_checksum (vm, b1); good_udp1 = - (flags1 & IP_BUFFER_L4_CHECKSUM_CORRECT) != 0; + (flags1 & VNET_BUFFER_F_L4_CHECKSUM_CORRECT) != 0; } } @@ -1116,7 +1116,7 @@ ip_vxlan_bypass_inline (vlib_main_t * vm, } flags0 = b0->flags; - good_udp0 = (flags0 & IP_BUFFER_L4_CHECKSUM_CORRECT) != 0; + good_udp0 = (flags0 & VNET_BUFFER_F_L4_CHECKSUM_CORRECT) != 0; /* Don't verify UDP checksum for packets with explicit zero checksum. */ good_udp0 |= udp0->checksum == 0; @@ -1132,14 +1132,14 @@ ip_vxlan_bypass_inline (vlib_main_t * vm, /* Verify UDP checksum */ if (PREDICT_FALSE (!good_udp0)) { - if ((flags0 & IP_BUFFER_L4_CHECKSUM_COMPUTED) == 0) + if ((flags0 & VNET_BUFFER_F_L4_CHECKSUM_COMPUTED) == 0) { if (is_ip4) flags0 = ip4_tcp_udp_validate_checksum (vm, b0); else flags0 = ip6_tcp_udp_icmp_validate_checksum (vm, b0); good_udp0 = - (flags0 & IP_BUFFER_L4_CHECKSUM_CORRECT) != 0; + (flags0 & VNET_BUFFER_F_L4_CHECKSUM_CORRECT) != 0; } } -- cgit 1.2.3-korg From e1bf572999362b7a0f485613d70736c95e47a2ae Mon Sep 17 00:00:00 2001 From: Hongjun Ni Date: Thu, 3 Aug 2017 20:34:45 +0800 Subject: Initialize vxlan-gpe bypass mode Change-Id: I2a75357d49a971818c8d96d56d184c5d01fbc775 Signed-off-by: Hongjun Ni --- src/vnet/vxlan-gpe/vxlan_gpe.c | 16 ++++++++++++++++ 1 file changed, 16 insertions(+) (limited to 'src/vnet/vxlan-gpe') diff --git a/src/vnet/vxlan-gpe/vxlan_gpe.c b/src/vnet/vxlan-gpe/vxlan_gpe.c index fab887ad..97bb1b15 100644 --- a/src/vnet/vxlan-gpe/vxlan_gpe.c +++ b/src/vnet/vxlan-gpe/vxlan_gpe.c @@ -1200,6 +1200,22 @@ VLIB_CLI_COMMAND (set_interface_ip6_vxlan_gpe_bypass_command, static) = { }; /* *INDENT-ON* */ +/* *INDENT-OFF* */ +VNET_FEATURE_INIT (ip4_vxlan_gpe_bypass, static) = +{ + .arc_name = "ip4-unicast", + .node_name = "ip4-vxlan-gpe-bypass", + .runs_before = VNET_FEATURES ("ip4-lookup"), +}; + +VNET_FEATURE_INIT (ip6_vxlan_gpe_bypass, static) = +{ + .arc_name = "ip6-unicast", + .node_name = "ip6-vxlan-gpe-bypass", + .runs_before = VNET_FEATURES ("ip6-lookup"), +}; +/* *INDENT-ON* */ + /** * @brief Feature init function for VXLAN GPE * -- cgit 1.2.3-korg From da78f957e46c686434149d332a477d7ea055d76a Mon Sep 17 00:00:00 2001 From: Neale Ranns Date: Wed, 24 May 2017 09:15:43 -0700 Subject: L2 over MPLS [support for VPWS/VPLS] - switch to using dpo_proto_t rather than fib_protocol_t in fib_paths so that we can describe L2 paths - VLIB nodes to handle pop/push of MPLS labels to L2 Change-Id: Id050d06a11fd2c9c1c81ce5a0654e6c5ae6afa6e Signed-off-by: Neale Ranns --- src/plugins/gtpu/gtpu.c | 2 +- src/plugins/snat/snat.c | 2 +- src/vat/api_format.c | 17 +- src/vnet/dhcp/client.c | 6 +- src/vnet/dhcp/dhcp6_proxy_node.c | 2 +- src/vnet/dpo/dpo.c | 19 ++ src/vnet/dpo/dpo.h | 11 +- src/vnet/dpo/interface_dpo.c | 30 +++ src/vnet/dpo/mpls_label_dpo.c | 45 +++- src/vnet/ethernet/arp.c | 4 +- src/vnet/fib/fib_api.h | 4 +- src/vnet/fib/fib_entry.c | 8 +- src/vnet/fib/fib_entry_src.c | 16 +- src/vnet/fib/fib_entry_src.h | 4 +- src/vnet/fib/fib_entry_src_api.c | 2 +- src/vnet/fib/fib_entry_src_default_route.c | 2 +- src/vnet/fib/fib_entry_src_interface.c | 2 +- src/vnet/fib/fib_entry_src_lisp.c | 8 +- src/vnet/fib/fib_entry_src_mpls.c | 4 +- src/vnet/fib/fib_entry_src_rr.c | 15 +- src/vnet/fib/fib_entry_src_special.c | 2 +- src/vnet/fib/fib_path.c | 79 +++---- src/vnet/fib/fib_path.h | 11 +- src/vnet/fib/fib_path_ext.c | 3 + src/vnet/fib/fib_path_list.c | 4 +- src/vnet/fib/fib_path_list.h | 4 +- src/vnet/fib/fib_table.c | 6 +- src/vnet/fib/fib_table.h | 6 +- src/vnet/fib/fib_test.c | 338 ++++++++++++++--------------- src/vnet/fib/fib_types.h | 8 +- src/vnet/interface_format.c | 12 +- src/vnet/ip/ip4_forward.c | 6 +- src/vnet/ip/ip6_forward.c | 4 +- src/vnet/ip/ip6_neighbor.c | 10 +- src/vnet/ip/ip_api.c | 38 ++-- src/vnet/ip/lookup.c | 18 +- src/vnet/lisp-gpe/lisp_gpe.c | 13 +- src/vnet/lisp-gpe/lisp_gpe_api.c | 10 +- src/vnet/lisp-gpe/lisp_gpe_fwd_entry.c | 5 +- src/vnet/mfib/ip6_mfib.c | 6 +- src/vnet/mfib/mfib_entry.c | 10 +- src/vnet/mfib/mfib_test.c | 14 +- src/vnet/mpls/mpls.api | 4 +- src/vnet/mpls/mpls.c | 36 +-- src/vnet/mpls/mpls_api.c | 21 +- src/vnet/mpls/mpls_tunnel.c | 75 ++++--- src/vnet/mpls/mpls_tunnel.h | 38 ++-- src/vnet/srmpls/sr_mpls_policy.c | 6 +- src/vnet/srmpls/sr_mpls_steering.c | 2 +- src/vnet/srv6/sr_steering.c | 4 +- src/vnet/vxlan-gpe/vxlan_gpe.c | 2 +- src/vnet/vxlan/vxlan.c | 2 +- src/vpp/app/vpe_cli.c | 2 +- test/test_bfd.py | 6 +- test/test_gre.py | 24 +- test/test_ip6.py | 26 +-- test/test_map.py | 19 +- test/test_mpls.py | 318 ++++++++++++++++++++++----- test/test_p2p_ethernet.py | 16 +- test/vpp_ip_route.py | 24 +- test/vpp_mpls_tunnel_interface.py | 6 +- test/vpp_papi_provider.py | 4 +- 62 files changed, 889 insertions(+), 556 deletions(-) (limited to 'src/vnet/vxlan-gpe') diff --git a/src/plugins/gtpu/gtpu.c b/src/plugins/gtpu/gtpu.c index 84745bd8..3dfb4210 100755 --- a/src/plugins/gtpu/gtpu.c +++ b/src/plugins/gtpu/gtpu.c @@ -534,7 +534,7 @@ int vnet_gtpu_add_del_tunnel fib_node_index_t mfei; adj_index_t ai; fib_route_path_t path = { - .frp_proto = fp, + .frp_proto = fib_proto_to_dpo (fp), .frp_addr = zero_addr, .frp_sw_if_index = 0xffffffff, .frp_fib_index = ~0, diff --git a/src/plugins/snat/snat.c b/src/plugins/snat/snat.c index 9fbc1e54..f196b5c2 100644 --- a/src/plugins/snat/snat.c +++ b/src/plugins/snat/snat.c @@ -135,7 +135,7 @@ snat_add_del_addr_to_fib (ip4_address_t * addr, u8 p_len, u32 sw_if_index, (FIB_ENTRY_FLAG_CONNECTED | FIB_ENTRY_FLAG_LOCAL | FIB_ENTRY_FLAG_EXCLUSIVE), - FIB_PROTOCOL_IP4, + DPO_PROTO_IP4, NULL, sw_if_index, ~0, diff --git a/src/vat/api_format.c b/src/vat/api_format.c index f97cdeef..009cf173 100644 --- a/src/vat/api_format.c +++ b/src/vat/api_format.c @@ -7498,7 +7498,7 @@ api_mpls_route_add_del (vat_main_t * vam) mpls_label_t *next_hop_out_label_stack = NULL; mpls_label_t local_label = MPLS_LABEL_INVALID; u8 is_eos = 0; - u8 next_hop_proto_is_ip4 = 1; + dpo_proto_t next_hop_proto = DPO_PROTO_IP4; /* Parse args required to build the message */ while (unformat_check_input (i) != UNFORMAT_END_OF_INPUT) @@ -7517,13 +7517,13 @@ api_mpls_route_add_del (vat_main_t * vam) &v4_next_hop_address)) { next_hop_set = 1; - next_hop_proto_is_ip4 = 1; + next_hop_proto = DPO_PROTO_IP4; } else if (unformat (i, "via %U", unformat_ip6_address, &v6_next_hop_address)) { next_hop_set = 1; - next_hop_proto_is_ip4 = 0; + next_hop_proto = DPO_PROTO_IP6; } else if (unformat (i, "weight %d", &next_hop_weight)) ; @@ -7548,12 +7548,12 @@ api_mpls_route_add_del (vat_main_t * vam) else if (unformat (i, "lookup-in-ip4-table %d", &next_hop_table_id)) { next_hop_set = 1; - next_hop_proto_is_ip4 = 1; + next_hop_proto = DPO_PROTO_IP4; } else if (unformat (i, "lookup-in-ip6-table %d", &next_hop_table_id)) { next_hop_set = 1; - next_hop_proto_is_ip4 = 0; + next_hop_proto = DPO_PROTO_IP6; } else if (unformat (i, "next-hop-table %d", &next_hop_table_id)) ; @@ -7599,7 +7599,7 @@ api_mpls_route_add_del (vat_main_t * vam) mp->mr_create_table_if_needed = create_table_if_needed; mp->mr_is_add = is_add; - mp->mr_next_hop_proto_is_ip4 = next_hop_proto_is_ip4; + mp->mr_next_hop_proto = next_hop_proto; mp->mr_is_classify = is_classify; mp->mr_is_multipath = is_multipath; mp->mr_is_resolve_host = resolve_host; @@ -7622,13 +7622,14 @@ api_mpls_route_add_del (vat_main_t * vam) if (next_hop_set) { - if (next_hop_proto_is_ip4) + if (DPO_PROTO_IP4 == next_hop_proto) { clib_memcpy (mp->mr_next_hop, &v4_next_hop_address, sizeof (v4_next_hop_address)); } - else + else if (DPO_PROTO_IP6 == next_hop_proto) + { clib_memcpy (mp->mr_next_hop, &v6_next_hop_address, diff --git a/src/vnet/dhcp/client.c b/src/vnet/dhcp/client.c index cfe62a6f..dd5e99f2 100644 --- a/src/vnet/dhcp/client.c +++ b/src/vnet/dhcp/client.c @@ -296,7 +296,7 @@ int dhcp_client_for_us (u32 bi, vlib_buffer_t * b, &all_0s, FIB_SOURCE_DHCP, FIB_ENTRY_FLAG_NONE, - FIB_PROTOCOL_IP4, + DPO_PROTO_IP4, &nh, c->sw_if_index, ~0, @@ -605,7 +605,7 @@ dhcp_bound_state (dhcp_client_main_t * dcm, dhcp_client_t * c, f64 now) c->sw_if_index), &all_0s, FIB_SOURCE_DHCP, - FIB_PROTOCOL_IP4, + DPO_PROTO_IP4, &nh, c->sw_if_index, ~0, @@ -900,7 +900,7 @@ int dhcp_client_add_del (dhcp_client_add_del_args_t * a) c->sw_if_index), &all_0s, FIB_SOURCE_DHCP, - FIB_PROTOCOL_IP4, + DPO_PROTO_IP4, &nh, c->sw_if_index, ~0, diff --git a/src/vnet/dhcp/dhcp6_proxy_node.c b/src/vnet/dhcp/dhcp6_proxy_node.c index e109cc4c..9c2f5220 100644 --- a/src/vnet/dhcp/dhcp6_proxy_node.c +++ b/src/vnet/dhcp/dhcp6_proxy_node.c @@ -857,7 +857,7 @@ dhcp6_proxy_set_server (ip46_address_t *addr, else { const fib_route_path_t path_for_us = { - .frp_proto = FIB_PROTOCOL_IP6, + .frp_proto = DPO_PROTO_IP6, .frp_addr = zero_addr, .frp_sw_if_index = 0xffffffff, .frp_fib_index = ~0, diff --git a/src/vnet/dpo/dpo.c b/src/vnet/dpo/dpo.c index 389f995b..aa770838 100644 --- a/src/vnet/dpo/dpo.c +++ b/src/vnet/dpo/dpo.c @@ -109,6 +109,25 @@ vnet_link_to_dpo_proto (vnet_link_t linkt) return (0); } +vnet_link_t +dpo_proto_to_link (dpo_proto_t dp) +{ + switch (dp) + { + case DPO_PROTO_IP6: + return (VNET_LINK_IP6); + case DPO_PROTO_IP4: + return (VNET_LINK_IP4); + case DPO_PROTO_MPLS: + return (VNET_LINK_MPLS); + case DPO_PROTO_ETHERNET: + return (VNET_LINK_ETHERNET); + case DPO_PROTO_NSH: + return (VNET_LINK_NSH); + } + return (~0); +} + u8 * format_dpo_type (u8 * s, va_list * args) { diff --git a/src/vnet/dpo/dpo.h b/src/vnet/dpo/dpo.h index 5aa4e2d2..42fc51d4 100644 --- a/src/vnet/dpo/dpo.h +++ b/src/vnet/dpo/dpo.h @@ -59,14 +59,10 @@ typedef u32 index_t; */ typedef enum dpo_proto_t_ { -#if CLIB_DEBUG > 0 - DPO_PROTO_IP4 = 1, -#else DPO_PROTO_IP4 = 0, -#endif DPO_PROTO_IP6, - DPO_PROTO_ETHERNET, DPO_PROTO_MPLS, + DPO_PROTO_ETHERNET, DPO_PROTO_NSH, } __attribute__((packed)) dpo_proto_t; @@ -272,6 +268,11 @@ extern u8 *format_dpo_type(u8 * s, va_list * args); */ extern u8 *format_dpo_proto(u8 * s, va_list * args); +/** + * @brief format a DPO protocol + */ +extern vnet_link_t dpo_proto_to_link(dpo_proto_t dp); + /** * @brief * Set and stack a DPO. diff --git a/src/vnet/dpo/interface_dpo.c b/src/vnet/dpo/interface_dpo.c index 8d700c23..780bfa2a 100644 --- a/src/vnet/dpo/interface_dpo.c +++ b/src/vnet/dpo/interface_dpo.c @@ -195,11 +195,17 @@ const static char* const interface_dpo_ip6_nodes[] = "interface-dpo-ip4", NULL, }; +const static char* const interface_dpo_l2_nodes[] = +{ + "interface-dpo-l2", + NULL, +}; const static char* const * const interface_dpo_nodes[DPO_PROTO_NUM] = { [DPO_PROTO_IP4] = interface_dpo_ip4_nodes, [DPO_PROTO_IP6] = interface_dpo_ip6_nodes, + [DPO_PROTO_ETHERNET] = interface_dpo_l2_nodes, [DPO_PROTO_MPLS] = NULL, }; @@ -382,6 +388,14 @@ interface_dpo_ip6 (vlib_main_t * vm, return (interface_dpo_inline(vm, node, from_frame)); } +static uword +interface_dpo_l2 (vlib_main_t * vm, + vlib_node_runtime_t * node, + vlib_frame_t * from_frame) +{ + return (interface_dpo_inline(vm, node, from_frame)); +} + VLIB_REGISTER_NODE (interface_dpo_ip4_node) = { .function = interface_dpo_ip4, .name = "interface-dpo-ip4", @@ -414,3 +428,19 @@ VLIB_REGISTER_NODE (interface_dpo_ip6_node) = { VLIB_NODE_FUNCTION_MULTIARCH (interface_dpo_ip6_node, interface_dpo_ip6) +VLIB_REGISTER_NODE (interface_dpo_l2_node) = { + .function = interface_dpo_l2, + .name = "interface-dpo-l2", + .vector_size = sizeof (u32), + .format_trace = format_interface_dpo_trace, + + .n_next_nodes = 2, + .next_nodes = { + [INTERFACE_DPO_DROP] = "error-drop", + [INTERFACE_DPO_INPUT] = "l2-input", + }, +}; + +VLIB_NODE_FUNCTION_MULTIARCH (interface_dpo_l2_node, + interface_dpo_l2) + diff --git a/src/vnet/dpo/mpls_label_dpo.c b/src/vnet/dpo/mpls_label_dpo.c index 1c451a51..b178a902 100644 --- a/src/vnet/dpo/mpls_label_dpo.c +++ b/src/vnet/dpo/mpls_label_dpo.c @@ -192,7 +192,8 @@ mpls_label_imposition_inline (vlib_main_t * vm, vlib_node_runtime_t * node, vlib_frame_t * from_frame, u8 payload_is_ip4, - u8 payload_is_ip6) + u8 payload_is_ip6, + u8 payload_is_ethernet) { u32 n_left_from, next_index, * from, * to_next; @@ -320,6 +321,13 @@ mpls_label_imposition_inline (vlib_main_t * vm, ttl2 = ip2->hop_limit; ttl3 = ip3->hop_limit; } + else if (payload_is_ethernet) + { + /* + * nothing to chang ein the ethernet header + */ + ttl0 = ttl1 = ttl2 = ttl3 = 255; + } else { /* @@ -551,7 +559,7 @@ mpls_label_imposition (vlib_main_t * vm, vlib_node_runtime_t * node, vlib_frame_t * frame) { - return (mpls_label_imposition_inline(vm, node, frame, 0, 0)); + return (mpls_label_imposition_inline(vm, node, frame, 0, 0, 0)); } VLIB_REGISTER_NODE (mpls_label_imposition_node) = { @@ -573,7 +581,7 @@ ip4_mpls_label_imposition (vlib_main_t * vm, vlib_node_runtime_t * node, vlib_frame_t * frame) { - return (mpls_label_imposition_inline(vm, node, frame, 1, 0)); + return (mpls_label_imposition_inline(vm, node, frame, 1, 0, 0)); } VLIB_REGISTER_NODE (ip4_mpls_label_imposition_node) = { @@ -595,7 +603,7 @@ ip6_mpls_label_imposition (vlib_main_t * vm, vlib_node_runtime_t * node, vlib_frame_t * frame) { - return (mpls_label_imposition_inline(vm, node, frame, 0, 1)); + return (mpls_label_imposition_inline(vm, node, frame, 0, 1, 0)); } VLIB_REGISTER_NODE (ip6_mpls_label_imposition_node) = { @@ -612,6 +620,28 @@ VLIB_REGISTER_NODE (ip6_mpls_label_imposition_node) = { VLIB_NODE_FUNCTION_MULTIARCH (ip6_mpls_label_imposition_node, ip6_mpls_label_imposition) +static uword +ethernet_mpls_label_imposition (vlib_main_t * vm, + vlib_node_runtime_t * node, + vlib_frame_t * frame) +{ + return (mpls_label_imposition_inline(vm, node, frame, 0, 0, 1)); +} + +VLIB_REGISTER_NODE (ethernet_mpls_label_imposition_node) = { + .function = ethernet_mpls_label_imposition, + .name = "ethernet-mpls-label-imposition", + .vector_size = sizeof (u32), + + .format_trace = format_mpls_label_imposition_trace, + .n_next_nodes = 1, + .next_nodes = { + [0] = "error-drop", + } +}; +VLIB_NODE_FUNCTION_MULTIARCH (ethernet_mpls_label_imposition_node, + ethernet_mpls_label_imposition) + static void mpls_label_dpo_mem_show (void) { @@ -643,11 +673,18 @@ const static char* const mpls_label_imp_mpls_nodes[] = "mpls-label-imposition", NULL, }; +const static char* const mpls_label_imp_ethernet_nodes[] = +{ + "ethernet-mpls-label-imposition", + NULL, +}; + const static char* const * const mpls_label_imp_nodes[DPO_PROTO_NUM] = { [DPO_PROTO_IP4] = mpls_label_imp_ip4_nodes, [DPO_PROTO_IP6] = mpls_label_imp_ip6_nodes, [DPO_PROTO_MPLS] = mpls_label_imp_mpls_nodes, + [DPO_PROTO_ETHERNET] = mpls_label_imp_ethernet_nodes, }; diff --git a/src/vnet/ethernet/arp.c b/src/vnet/ethernet/arp.c index 4d9edaf5..8a394006 100644 --- a/src/vnet/ethernet/arp.c +++ b/src/vnet/ethernet/arp.c @@ -588,7 +588,7 @@ vnet_arp_set_ip4_over_ethernet_internal (vnet_main_t * vnm, e->fib_entry_index = fib_table_entry_path_add (fib_index, &pfx, FIB_SOURCE_ADJ, FIB_ENTRY_FLAG_ATTACHED, - FIB_PROTOCOL_IP4, &pfx.fp_addr, + DPO_PROTO_IP4, &pfx.fp_addr, e->sw_if_index, ~0, 1, NULL, FIB_ROUTE_PATH_FLAG_NONE); } @@ -1621,7 +1621,7 @@ arp_entry_free (ethernet_arp_interface_t * eai, ethernet_arp_ip4_entry_t * e) fib_table_entry_path_remove (fib_index, &pfx, FIB_SOURCE_ADJ, - FIB_PROTOCOL_IP4, + DPO_PROTO_IP4, &pfx.fp_addr, e->sw_if_index, ~0, 1, FIB_ROUTE_PATH_FLAG_NONE); diff --git a/src/vnet/fib/fib_api.h b/src/vnet/fib/fib_api.h index 73d76a42..d07d6cae 100644 --- a/src/vnet/fib/fib_api.h +++ b/src/vnet/fib/fib_api.h @@ -21,7 +21,7 @@ int add_del_route_check (fib_protocol_t table_proto, u32 table_id, u32 next_hop_sw_if_index, - fib_protocol_t next_hop_table_proto, + dpo_proto_t next_hop_table_proto, u32 next_hop_table_id, u8 create_missing_tables, u8 is_rpf_id, @@ -43,7 +43,7 @@ add_del_route_t_handler (u8 is_multipath, u8 is_rpf_id, u32 fib_index, const fib_prefix_t * prefix, - u8 next_hop_proto_is_ip4, + dpo_proto_t next_hop_proto, const ip46_address_t * next_hop, u32 next_hop_sw_if_index, u8 next_hop_fib_index, diff --git a/src/vnet/fib/fib_entry.c b/src/vnet/fib/fib_entry.c index d7ff1c8c..2027f2be 100644 --- a/src/vnet/fib/fib_entry.c +++ b/src/vnet/fib/fib_entry.c @@ -58,12 +58,18 @@ fib_entry_get_index (const fib_entry_t * fib_entry) return (fib_entry - fib_entry_pool); } -static fib_protocol_t +fib_protocol_t fib_entry_get_proto (const fib_entry_t * fib_entry) { return (fib_entry->fe_prefix.fp_proto); } +dpo_proto_t +fib_entry_get_dpo_proto (const fib_entry_t * fib_entry) +{ + return (fib_proto_to_dpo(fib_entry->fe_prefix.fp_proto)); +} + fib_forward_chain_type_t fib_entry_get_default_chain_type (const fib_entry_t *fib_entry) { diff --git a/src/vnet/fib/fib_entry_src.c b/src/vnet/fib/fib_entry_src.c index ff73cbf9..173df74f 100644 --- a/src/vnet/fib/fib_entry_src.c +++ b/src/vnet/fib/fib_entry_src.c @@ -29,12 +29,6 @@ */ static fib_entry_src_vft_t fib_entry_src_vft[FIB_SOURCE_MAX]; -static fib_protocol_t -fib_entry_get_proto (const fib_entry_t * fib_entry) -{ - return (fib_entry->fe_prefix.fp_proto); -} - void fib_entry_src_register (fib_source_t source, const fib_entry_src_vft_t *vft) @@ -861,7 +855,7 @@ fib_entry_src_action_add (fib_entry_t *fib_entry, fib_entry_src_vft[source].fesv_add(esrc, fib_entry, flags, - fib_entry_get_proto(fib_entry), + fib_entry_get_dpo_proto(fib_entry), dpo); } @@ -914,7 +908,7 @@ fib_entry_src_action_update (fib_entry_t *fib_entry, fib_entry_src_vft[source].fesv_add(esrc, fib_entry, flags, - fib_entry_get_proto(fib_entry), + fib_entry_get_dpo_proto(fib_entry), dpo); } @@ -1106,8 +1100,7 @@ fib_entry_src_action_path_add (fib_entry_t *fib_entry, source, flags, drop_dpo_get( - fib_proto_to_dpo( - fib_entry_get_proto(fib_entry)))); + fib_entry_get_dpo_proto(fib_entry))); esrc = fib_entry_src_find(fib_entry, source, NULL); } @@ -1166,8 +1159,7 @@ fib_entry_src_action_path_swap (fib_entry_t *fib_entry, source, flags, drop_dpo_get( - fib_proto_to_dpo( - fib_entry_get_proto(fib_entry)))); + fib_entry_get_dpo_proto(fib_entry))); esrc = fib_entry_src_find(fib_entry, source, NULL); } diff --git a/src/vnet/fib/fib_entry_src.h b/src/vnet/fib/fib_entry_src.h index 640c174d..35c43936 100644 --- a/src/vnet/fib/fib_entry_src.h +++ b/src/vnet/fib/fib_entry_src.h @@ -73,7 +73,7 @@ typedef void (*fib_entry_src_deactivate_t)(fib_entry_src_t *src, typedef void (*fib_entry_src_add_t)(fib_entry_src_t *src, const fib_entry_t *entry, fib_entry_flag_t flags, - fib_protocol_t proto, + dpo_proto_t proto, const dpo_id_t *dpo); /** @@ -277,6 +277,8 @@ extern void fib_entry_src_mk_lb (fib_entry_t *fib_entry, fib_forward_chain_type_t fct, dpo_id_t *dpo_lb); +extern fib_protocol_t fib_entry_get_proto(const fib_entry_t * fib_entry); +extern dpo_proto_t fib_entry_get_dpo_proto(const fib_entry_t * fib_entry); /* * Per-source registration. declared here so we save a separate .h file for each diff --git a/src/vnet/fib/fib_entry_src_api.c b/src/vnet/fib/fib_entry_src_api.c index f895886b..1cdcfbde 100644 --- a/src/vnet/fib/fib_entry_src_api.c +++ b/src/vnet/fib/fib_entry_src_api.c @@ -131,7 +131,7 @@ static void fib_entry_src_api_add (fib_entry_src_t *src, const fib_entry_t *entry, fib_entry_flag_t flags, - fib_protocol_t proto, + dpo_proto_t proto, const dpo_id_t *dpo) { if (FIB_ENTRY_FLAG_NONE != flags) diff --git a/src/vnet/fib/fib_entry_src_default_route.c b/src/vnet/fib/fib_entry_src_default_route.c index 9f4e7c36..431abb66 100644 --- a/src/vnet/fib/fib_entry_src_default_route.c +++ b/src/vnet/fib/fib_entry_src_default_route.c @@ -35,7 +35,7 @@ static void fib_entry_src_default_route_add (fib_entry_src_t *src, const fib_entry_t *entry, fib_entry_flag_t flags, - fib_protocol_t proto, + dpo_proto_t proto, const dpo_id_t *dpo) { src->fes_pl = fib_path_list_create_special(proto, diff --git a/src/vnet/fib/fib_entry_src_interface.c b/src/vnet/fib/fib_entry_src_interface.c index bb87818f..6c087f34 100644 --- a/src/vnet/fib/fib_entry_src_interface.c +++ b/src/vnet/fib/fib_entry_src_interface.c @@ -35,7 +35,7 @@ static void fib_entry_src_interface_add (fib_entry_src_t *src, const fib_entry_t *entry, fib_entry_flag_t flags, - fib_protocol_t proto, + dpo_proto_t proto, const dpo_id_t *dpo) { src->fes_pl = fib_path_list_create_special( diff --git a/src/vnet/fib/fib_entry_src_lisp.c b/src/vnet/fib/fib_entry_src_lisp.c index 7f8b91bb..e72dce63 100644 --- a/src/vnet/fib/fib_entry_src_lisp.c +++ b/src/vnet/fib/fib_entry_src_lisp.c @@ -79,10 +79,10 @@ fib_entry_src_lisp_path_remove (fib_entry_src_t *src, static void fib_entry_src_lisp_add (fib_entry_src_t *src, - const fib_entry_t *entry, - fib_entry_flag_t flags, - fib_protocol_t proto, - const dpo_id_t *dpo) + const fib_entry_t *entry, + fib_entry_flag_t flags, + dpo_proto_t proto, + const dpo_id_t *dpo) { if (FIB_ENTRY_FLAG_NONE != flags) { diff --git a/src/vnet/fib/fib_entry_src_mpls.c b/src/vnet/fib/fib_entry_src_mpls.c index 14c7310f..a616458f 100644 --- a/src/vnet/fib/fib_entry_src_mpls.c +++ b/src/vnet/fib/fib_entry_src_mpls.c @@ -57,13 +57,13 @@ static void fib_entry_src_mpls_add (fib_entry_src_t *src, const fib_entry_t *entry, fib_entry_flag_t flags, - fib_protocol_t proto, + dpo_proto_t proto, const dpo_id_t *dpo) { src->fes_pl = fib_path_list_create_special(proto, FIB_PATH_LIST_FLAG_DROP, - drop_dpo_get(fib_proto_to_dpo(proto))); + drop_dpo_get(proto)); } static void diff --git a/src/vnet/fib/fib_entry_src_rr.c b/src/vnet/fib/fib_entry_src_rr.c index d66ef7b1..1153f3f1 100644 --- a/src/vnet/fib/fib_entry_src_rr.c +++ b/src/vnet/fib/fib_entry_src_rr.c @@ -35,7 +35,7 @@ fib_entry_src_rr_resolve_via_connected (fib_entry_src_t *src, const fib_entry_t *cover) { const fib_route_path_t path = { - .frp_proto = fib_entry->fe_prefix.fp_proto, + .frp_proto = fib_proto_to_dpo(fib_entry->fe_prefix.fp_proto), .frp_addr = fib_entry->fe_prefix.fp_addr, .frp_sw_if_index = fib_entry_get_resolving_interface( fib_entry_get_index(cover)), @@ -90,18 +90,17 @@ fib_entry_src_rr_use_covers_pl (fib_entry_src_t *src, const fib_entry_t *cover) { fib_node_index_t *entries = NULL; - fib_protocol_t proto; + dpo_proto_t proto; - proto = fib_entry->fe_prefix.fp_proto; + proto = fib_proto_to_dpo(fib_entry->fe_prefix.fp_proto); vec_add1(entries, fib_entry_get_index(fib_entry)); if (fib_path_list_recursive_loop_detect(cover->fe_parent, &entries)) { - src->fes_pl = fib_path_list_create_special( - proto, - FIB_PATH_LIST_FLAG_DROP, - drop_dpo_get(fib_proto_to_dpo(proto))); + src->fes_pl = fib_path_list_create_special(proto, + FIB_PATH_LIST_FLAG_DROP, + drop_dpo_get(proto)); } else { @@ -126,7 +125,7 @@ fib_entry_src_rr_activate (fib_entry_src_t *src, */ if (FIB_PROTOCOL_MPLS == fib_entry->fe_prefix.fp_proto) { - src->fes_pl = fib_path_list_create_special(FIB_PROTOCOL_MPLS, + src->fes_pl = fib_path_list_create_special(DPO_PROTO_MPLS, FIB_PATH_LIST_FLAG_DROP, NULL); fib_path_list_lock(src->fes_pl); diff --git a/src/vnet/fib/fib_entry_src_special.c b/src/vnet/fib/fib_entry_src_special.c index 75605d7f..e979e18f 100644 --- a/src/vnet/fib/fib_entry_src_special.c +++ b/src/vnet/fib/fib_entry_src_special.c @@ -43,7 +43,7 @@ static void fib_entry_src_special_add (fib_entry_src_t *src, const fib_entry_t *entry, fib_entry_flag_t flags, - fib_protocol_t proto, + dpo_proto_t proto, const dpo_id_t *dpo) { src->fes_pl = diff --git a/src/vnet/fib/fib_path.c b/src/vnet/fib/fib_path.c index 3a67a544..58050ccb 100644 --- a/src/vnet/fib/fib_path.c +++ b/src/vnet/fib/fib_path.c @@ -193,7 +193,7 @@ typedef struct fib_path_t_ { * next-hop's address. We can't derive this from the address itself * since the address can be all zeros */ - fib_protocol_t fp_nh_proto; + dpo_proto_t fp_nh_proto; /** * UCMP [unnormalised] weigth @@ -381,7 +381,7 @@ format_fib_path (u8 * s, va_list * args) s = format (s, " index:%d ", fib_path_get_index(path)); s = format (s, "pl-index:%d ", path->fp_pl_index); - s = format (s, "%U ", format_fib_protocol, path->fp_nh_proto); + s = format (s, "%U ", format_dpo_proto, path->fp_nh_proto); s = format (s, "weight=%d ", path->fp_weight); s = format (s, "pref=%d ", path->fp_preference); s = format (s, "%s: ", fib_path_type_names[path->fp_type]); @@ -454,7 +454,7 @@ format_fib_path (u8 * s, va_list * args) } break; case FIB_PATH_TYPE_RECURSIVE: - if (FIB_PROTOCOL_MPLS == path->fp_nh_proto) + if (DPO_PROTO_MPLS == path->fp_nh_proto) { s = format (s, "via %U %U", format_mpls_unicast_label, @@ -552,14 +552,14 @@ fib_path_attached_next_hop_get_adj (fib_path_t *path, * the subnet address (the attached route) links to the * auto-adj (see below), we want that adj here too. */ - return (adj_nbr_add_or_lock(path->fp_nh_proto, + return (adj_nbr_add_or_lock(dpo_proto_to_fib(path->fp_nh_proto), link, &zero_addr, path->attached_next_hop.fp_interface)); } else { - return (adj_nbr_add_or_lock(path->fp_nh_proto, + return (adj_nbr_add_or_lock(dpo_proto_to_fib(path->fp_nh_proto), link, &path->attached_next_hop.fp_nh, path->attached_next_hop.fp_interface)); @@ -575,10 +575,10 @@ fib_path_attached_next_hop_set (fib_path_t *path) */ dpo_set(&path->fp_dpo, DPO_ADJACENCY, - fib_proto_to_dpo(path->fp_nh_proto), + path->fp_nh_proto, fib_path_attached_next_hop_get_adj( path, - fib_proto_to_link(path->fp_nh_proto))); + dpo_proto_to_link(path->fp_nh_proto))); /* * become a child of the adjacency so we receive updates @@ -607,14 +607,14 @@ fib_path_attached_get_adj (fib_path_t *path, * point-2-point interfaces do not require a glean, since * there is nothing to ARP. Install a rewrite/nbr adj instead */ - return (adj_nbr_add_or_lock(path->fp_nh_proto, + return (adj_nbr_add_or_lock(dpo_proto_to_fib(path->fp_nh_proto), link, &zero_addr, path->attached.fp_interface)); } else { - return (adj_glean_add_or_lock(path->fp_nh_proto, + return (adj_glean_add_or_lock(dpo_proto_to_fib(path->fp_nh_proto), path->attached.fp_interface, NULL)); } @@ -650,7 +650,7 @@ fib_path_recursive_adj_update (fib_path_t *path, if (path->fp_oper_flags & FIB_PATH_OPER_FLAG_RECURSIVE_LOOP) { path->fp_oper_flags &= ~FIB_PATH_OPER_FLAG_RESOLVED; - dpo_copy(&via_dpo, drop_dpo_get(fib_proto_to_dpo(path->fp_nh_proto))); + dpo_copy(&via_dpo, drop_dpo_get(path->fp_nh_proto)); } else if (path->fp_cfg_flags & FIB_PATH_CFG_FLAG_RESOLVE_HOST) { @@ -668,7 +668,7 @@ fib_path_recursive_adj_update (fib_path_t *path, if (fib_entry_get_best_source(path->fp_via_fib) >= FIB_SOURCE_RR) { path->fp_oper_flags &= ~FIB_PATH_OPER_FLAG_RESOLVED; - dpo_copy(&via_dpo, drop_dpo_get(fib_proto_to_dpo(path->fp_nh_proto))); + dpo_copy(&via_dpo, drop_dpo_get(path->fp_nh_proto)); /* * PIC edge trigger. let the load-balance maps know @@ -685,7 +685,7 @@ fib_path_recursive_adj_update (fib_path_t *path, if (!(FIB_ENTRY_FLAG_ATTACHED & fib_entry_get_flags(path->fp_via_fib))) { path->fp_oper_flags &= ~FIB_PATH_OPER_FLAG_RESOLVED; - dpo_copy(&via_dpo, drop_dpo_get(fib_proto_to_dpo(path->fp_nh_proto))); + dpo_copy(&via_dpo, drop_dpo_get(path->fp_nh_proto)); /* * PIC edge trigger. let the load-balance maps know @@ -699,7 +699,7 @@ fib_path_recursive_adj_update (fib_path_t *path, if (!fib_entry_is_resolved(path->fp_via_fib)) { path->fp_oper_flags &= ~FIB_PATH_OPER_FLAG_RESOLVED; - dpo_copy(&via_dpo, drop_dpo_get(fib_proto_to_dpo(path->fp_nh_proto))); + dpo_copy(&via_dpo, drop_dpo_get(path->fp_nh_proto)); /* * PIC edge trigger. let the load-balance maps know @@ -720,9 +720,7 @@ fib_path_recursive_adj_update (fib_path_t *path, */ dpo_copy(dpo, &via_dpo); - FIB_PATH_DBG(path, "recursive update: %U", - fib_get_lookup_main(path->fp_nh_proto), - &path->fp_dpo, 2); + FIB_PATH_DBG(path, "recursive update:"); dpo_reset(&via_dpo); } @@ -804,13 +802,8 @@ fib_path_unresolve (fib_path_t *path) static fib_forward_chain_type_t fib_path_to_chain_type (const fib_path_t *path) { - switch (path->fp_nh_proto) + if (DPO_PROTO_MPLS == path->fp_nh_proto) { - case FIB_PROTOCOL_IP4: - return (FIB_FORW_CHAIN_TYPE_UNICAST_IP4); - case FIB_PROTOCOL_IP6: - return (FIB_FORW_CHAIN_TYPE_UNICAST_IP6); - case FIB_PROTOCOL_MPLS: if (FIB_PATH_TYPE_RECURSIVE == path->fp_type && MPLS_EOS == path->recursive.fp_nh.fp_eos) { @@ -821,7 +814,10 @@ fib_path_to_chain_type (const fib_path_t *path) return (FIB_FORW_CHAIN_TYPE_MPLS_NON_EOS); } } - return (FIB_FORW_CHAIN_TYPE_UNICAST_IP4); + else + { + return (fib_forw_chain_type_from_dpo_proto(path->fp_nh_proto)); + } } /* @@ -927,7 +923,7 @@ FIXME comment ai = fib_path_attached_next_hop_get_adj( path, - fib_proto_to_link(path->fp_nh_proto)); + dpo_proto_to_link(path->fp_nh_proto)); path->fp_oper_flags &= ~FIB_PATH_OPER_FLAG_RESOLVED; if (if_is_up && adj_is_up(ai)) @@ -935,9 +931,7 @@ FIXME comment path->fp_oper_flags |= FIB_PATH_OPER_FLAG_RESOLVED; } - dpo_set(&path->fp_dpo, DPO_ADJACENCY, - fib_proto_to_dpo(path->fp_nh_proto), - ai); + dpo_set(&path->fp_dpo, DPO_ADJACENCY, path->fp_nh_proto, ai); adj_unlock(ai); if (!if_is_up) @@ -1141,7 +1135,7 @@ fib_path_create (fib_node_index_t pl_index, else { path->fp_type = FIB_PATH_TYPE_RECURSIVE; - if (FIB_PROTOCOL_MPLS == path->fp_nh_proto) + if (DPO_PROTO_MPLS == path->fp_nh_proto) { path->recursive.fp_nh.fp_local_label = rpath->frp_local_label; path->recursive.fp_nh.fp_eos = rpath->frp_eos; @@ -1167,7 +1161,7 @@ fib_path_create (fib_node_index_t pl_index, */ fib_node_index_t fib_path_create_special (fib_node_index_t pl_index, - fib_protocol_t nh_proto, + dpo_proto_t nh_proto, fib_path_cfg_flags_t flags, const dpo_id_t *dpo) { @@ -1433,7 +1427,7 @@ fib_path_cmp_w_route_path (fib_node_index_t path_index, res = (path->attached.fp_interface - rpath->frp_sw_if_index); break; case FIB_PATH_TYPE_RECURSIVE: - if (FIB_PROTOCOL_MPLS == path->fp_nh_proto) + if (DPO_PROTO_MPLS == path->fp_nh_proto) { res = path->recursive.fp_nh.fp_local_label - rpath->frp_local_label; @@ -1535,8 +1529,7 @@ fib_path_recursive_loop_detect (fib_node_index_t path_index, FIB_PATH_DBG(path, "recursive loop formed"); path->fp_oper_flags |= FIB_PATH_OPER_FLAG_RECURSIVE_LOOP; - dpo_copy(&path->fp_dpo, - drop_dpo_get(fib_proto_to_dpo(path->fp_nh_proto))); + dpo_copy(&path->fp_dpo, drop_dpo_get(path->fp_nh_proto)); } else { @@ -1590,8 +1583,7 @@ fib_path_resolve (fib_node_index_t path_index) */ if (fib_path_is_permanent_drop(path)) { - dpo_copy(&path->fp_dpo, - drop_dpo_get(fib_proto_to_dpo(path->fp_nh_proto))); + dpo_copy(&path->fp_dpo, drop_dpo_get(path->fp_nh_proto)); path->fp_oper_flags &= ~FIB_PATH_OPER_FLAG_RESOLVED; return (fib_path_is_resolved(path_index)); } @@ -1612,9 +1604,9 @@ fib_path_resolve (fib_node_index_t path_index) } dpo_set(&path->fp_dpo, DPO_ADJACENCY, - fib_proto_to_dpo(path->fp_nh_proto), + path->fp_nh_proto, fib_path_attached_get_adj(path, - fib_proto_to_link(path->fp_nh_proto))); + dpo_proto_to_link(path->fp_nh_proto))); /* * become a child of the adjacency so we receive updates @@ -1639,7 +1631,7 @@ fib_path_resolve (fib_node_index_t path_index) ASSERT(FIB_NODE_INDEX_INVALID == path->fp_via_fib); - if (FIB_PROTOCOL_MPLS == path->fp_nh_proto) + if (DPO_PROTO_MPLS == path->fp_nh_proto) { fib_prefix_from_mpls_label(path->recursive.fp_nh.fp_local_label, path->recursive.fp_nh.fp_eos, @@ -1680,8 +1672,7 @@ fib_path_resolve (fib_node_index_t path_index) /* * Resolve via the drop */ - dpo_copy(&path->fp_dpo, - drop_dpo_get(fib_proto_to_dpo(path->fp_nh_proto))); + dpo_copy(&path->fp_dpo, drop_dpo_get(path->fp_nh_proto)); break; case FIB_PATH_TYPE_DEAG: { @@ -1696,7 +1687,7 @@ fib_path_resolve (fib_node_index_t path_index) LOOKUP_UNICAST); lookup_dpo_add_or_lock_w_fib_index(path->deag.fp_tbl_id, - fib_proto_to_dpo(path->fp_nh_proto), + path->fp_nh_proto, cast, LOOKUP_INPUT_DST_ADDR, LOOKUP_TABLE_FROM_CONFIG, @@ -1707,7 +1698,7 @@ fib_path_resolve (fib_node_index_t path_index) /* * Resolve via a receive DPO. */ - receive_dpo_add_or_lock(fib_proto_to_dpo(path->fp_nh_proto), + receive_dpo_add_or_lock(path->fp_nh_proto, path->receive.fp_interface, &path->receive.fp_addr, &path->fp_dpo); @@ -1716,7 +1707,7 @@ fib_path_resolve (fib_node_index_t path_index) /* * Resolve via a receive DPO. */ - interface_dpo_add_or_lock(fib_proto_to_dpo(path->fp_nh_proto), + interface_dpo_add_or_lock(path->fp_nh_proto, path->intf_rx.fp_interface, &path->fp_dpo); break; @@ -2035,7 +2026,7 @@ fib_path_contribute_forwarding (fib_node_index_t path_index, /* * Create the adj needed for sending IP multicast traffic */ - ai = adj_mcast_add_or_lock(path->fp_nh_proto, + ai = adj_mcast_add_or_lock(dpo_proto_to_fib(path->fp_nh_proto), fib_forw_chain_type_to_link_type(fct), path->attached.fp_interface); dpo_set(dpo, DPO_ADJACENCY, @@ -2187,7 +2178,7 @@ fib_path_encode (fib_node_index_t path_list_index, return (FIB_PATH_LIST_WALK_CONTINUE); } -fib_protocol_t +dpo_proto_t fib_path_get_proto (fib_node_index_t path_index) { fib_path_t *path; diff --git a/src/vnet/fib/fib_path.h b/src/vnet/fib/fib_path.h index a34cb43f..f986e437 100644 --- a/src/vnet/fib/fib_path.h +++ b/src/vnet/fib/fib_path.h @@ -78,6 +78,11 @@ typedef enum fib_path_cfg_attribute_t_ { * The path is an interface recieve */ FIB_PATH_CFG_ATTRIBUTE_LOCAL, + /** + * The path is L2. i.e. the parameters therein are to be interpreted as + * pertaining to L2 config. + */ + FIB_PATH_CFG_ATTRIBUTE_L2, /** * Marker. Add new types before this one, then update it. */ @@ -98,6 +103,7 @@ typedef enum fib_path_cfg_attribute_t_ { [FIB_PATH_CFG_ATTRIBUTE_ATTACHED] = "attached", \ [FIB_PATH_CFG_ATTRIBUTE_INTF_RX] = "interface-rx", \ [FIB_PATH_CFG_ATTRIBUTE_RPF_ID] = "rpf-id", \ + [FIB_PATH_CFG_ATTRIBUTE_L2] = "l2", \ } #define FOR_EACH_FIB_PATH_CFG_ATTRIBUTE(_item) \ @@ -118,6 +124,7 @@ typedef enum fib_path_cfg_flags_t_ { FIB_PATH_CFG_FLAG_ATTACHED = (1 << FIB_PATH_CFG_ATTRIBUTE_ATTACHED), FIB_PATH_CFG_FLAG_INTF_RX = (1 << FIB_PATH_CFG_ATTRIBUTE_INTF_RX), FIB_PATH_CFG_FLAG_RPF_ID = (1 << FIB_PATH_CFG_ATTRIBUTE_RPF_ID), + FIB_PATH_CFG_FLAG_L2 = (1 << FIB_PATH_CFG_ATTRIBUTE_L2), } __attribute__ ((packed)) fib_path_cfg_flags_t; @@ -131,7 +138,7 @@ extern u8 * format_fib_path(u8 * s, va_list * args); extern fib_node_index_t fib_path_create(fib_node_index_t pl_index, const fib_route_path_t *path); extern fib_node_index_t fib_path_create_special(fib_node_index_t pl_index, - fib_protocol_t nh_proto, + dpo_proto_t nh_proto, fib_path_cfg_flags_t flags, const dpo_id_t *dpo); @@ -148,7 +155,7 @@ extern int fib_path_is_recursive_constrained(fib_node_index_t path_index); extern int fib_path_is_exclusive(fib_node_index_t path_index); extern int fib_path_is_deag(fib_node_index_t path_index); extern int fib_path_is_looped(fib_node_index_t path_index); -extern fib_protocol_t fib_path_get_proto(fib_node_index_t path_index); +extern dpo_proto_t fib_path_get_proto(fib_node_index_t path_index); extern void fib_path_destroy(fib_node_index_t path_index); extern uword fib_path_hash(fib_node_index_t path_index); extern load_balance_path_t * fib_path_append_nh_for_multipath_hash( diff --git a/src/vnet/fib/fib_path_ext.c b/src/vnet/fib/fib_path_ext.c index 26f2b9b6..4438671b 100644 --- a/src/vnet/fib/fib_path_ext.c +++ b/src/vnet/fib/fib_path_ext.c @@ -191,6 +191,9 @@ fib_path_ext_stack (fib_path_ext_t *path_ext, case FIB_FORW_CHAIN_TYPE_MPLS_NON_EOS: parent_fct = child_fct; break; + case FIB_FORW_CHAIN_TYPE_ETHERNET: + parent_fct = FIB_FORW_CHAIN_TYPE_MPLS_NON_EOS; + break; default: return (nhs); break; diff --git a/src/vnet/fib/fib_path_list.c b/src/vnet/fib/fib_path_list.c index 7a9c328c..f30fd7ea 100644 --- a/src/vnet/fib/fib_path_list.c +++ b/src/vnet/fib/fib_path_list.c @@ -611,7 +611,7 @@ fib_path_list_get_resolving_interface (fib_node_index_t path_list_index) return (sw_if_index); } -fib_protocol_t +dpo_proto_t fib_path_list_get_proto (fib_node_index_t path_list_index) { fib_path_list_t *path_list; @@ -753,7 +753,7 @@ fib_path_list_flags_2_path_flags (fib_path_list_flags_t plf) } fib_node_index_t -fib_path_list_create_special (fib_protocol_t nh_proto, +fib_path_list_create_special (dpo_proto_t nh_proto, fib_path_list_flags_t flags, const dpo_id_t *dpo) { diff --git a/src/vnet/fib/fib_path_list.h b/src/vnet/fib/fib_path_list.h index b4b6985b..a54b79e2 100644 --- a/src/vnet/fib/fib_path_list.h +++ b/src/vnet/fib/fib_path_list.h @@ -106,7 +106,7 @@ typedef enum fib_path_list_flags_t_ { extern fib_node_index_t fib_path_list_create(fib_path_list_flags_t flags, const fib_route_path_t *paths); -extern fib_node_index_t fib_path_list_create_special(fib_protocol_t nh_proto, +extern fib_node_index_t fib_path_list_create_special(dpo_proto_t nh_proto, fib_path_list_flags_t flags, const dpo_id_t *dpo); @@ -150,7 +150,7 @@ extern int fib_path_list_recursive_loop_detect(fib_node_index_t path_list_index, extern u32 fib_path_list_get_resolving_interface(fib_node_index_t path_list_index); extern int fib_path_list_is_looped(fib_node_index_t path_list_index); extern int fib_path_list_is_popular(fib_node_index_t path_list_index); -extern fib_protocol_t fib_path_list_get_proto(fib_node_index_t path_list_index); +extern dpo_proto_t fib_path_list_get_proto(fib_node_index_t path_list_index); extern u8 * fib_path_list_format(fib_node_index_t pl_index, u8 * s); extern index_t fib_path_list_lb_map_add_or_lock(fib_node_index_t pl_index, diff --git a/src/vnet/fib/fib_table.c b/src/vnet/fib/fib_table.c index 5aa02dd0..6b6cc5cb 100644 --- a/src/vnet/fib/fib_table.c +++ b/src/vnet/fib/fib_table.c @@ -505,7 +505,7 @@ fib_table_entry_path_add (u32 fib_index, const fib_prefix_t *prefix, fib_source_t source, fib_entry_flag_t flags, - fib_protocol_t next_hop_proto, + dpo_proto_t next_hop_proto, const ip46_address_t *next_hop, u32 next_hop_sw_if_index, u32 next_hop_fib_index, @@ -664,7 +664,7 @@ void fib_table_entry_path_remove (u32 fib_index, const fib_prefix_t *prefix, fib_source_t source, - fib_protocol_t next_hop_proto, + dpo_proto_t next_hop_proto, const ip46_address_t *next_hop, u32 next_hop_sw_if_index, u32 next_hop_fib_index, @@ -755,7 +755,7 @@ fib_table_entry_update_one_path (u32 fib_index, const fib_prefix_t *prefix, fib_source_t source, fib_entry_flag_t flags, - fib_protocol_t next_hop_proto, + dpo_proto_t next_hop_proto, const ip46_address_t *next_hop, u32 next_hop_sw_if_index, u32 next_hop_fib_index, diff --git a/src/vnet/fib/fib_table.h b/src/vnet/fib/fib_table.h index a65fea74..579740e9 100644 --- a/src/vnet/fib/fib_table.h +++ b/src/vnet/fib/fib_table.h @@ -288,7 +288,7 @@ extern fib_node_index_t fib_table_entry_path_add(u32 fib_index, const fib_prefix_t *prefix, fib_source_t source, fib_entry_flag_t flags, - fib_protocol_t next_hop_proto, + dpo_proto_t next_hop_proto, const ip46_address_t *next_hop, u32 next_hop_sw_if_index, u32 next_hop_fib_index, @@ -364,7 +364,7 @@ extern fib_node_index_t fib_table_entry_path_add2(u32 fib_index, extern void fib_table_entry_path_remove(u32 fib_index, const fib_prefix_t *prefix, fib_source_t source, - fib_protocol_t next_hop_proto, + dpo_proto_t next_hop_proto, const ip46_address_t *next_hop, u32 next_hop_sw_if_index, u32 next_hop_fib_index, @@ -471,7 +471,7 @@ extern fib_node_index_t fib_table_entry_update_one_path(u32 fib_index, const fib_prefix_t *prefix, fib_source_t source, fib_entry_flag_t flags, - fib_protocol_t next_hop_proto, + dpo_proto_t next_hop_proto, const ip46_address_t *next_hop, u32 next_hop_sw_if_index, u32 next_hop_fib_index, diff --git a/src/vnet/fib/fib_test.c b/src/vnet/fib/fib_test.c index 4c891667..59d5da2a 100644 --- a/src/vnet/fib/fib_test.c +++ b/src/vnet/fib/fib_test.c @@ -833,7 +833,7 @@ fib_test_v4 (void) FIB_SOURCE_INTERFACE, (FIB_ENTRY_FLAG_CONNECTED | FIB_ENTRY_FLAG_ATTACHED), - FIB_PROTOCOL_IP4, + DPO_PROTO_IP4, NULL, tm->hw[0]->sw_if_index, ~0, // invalid fib index @@ -861,7 +861,7 @@ fib_test_v4 (void) FIB_SOURCE_INTERFACE, (FIB_ENTRY_FLAG_CONNECTED | FIB_ENTRY_FLAG_LOCAL), - FIB_PROTOCOL_IP4, + DPO_PROTO_IP4, NULL, tm->hw[0]->sw_if_index, ~0, // invalid fib index @@ -911,7 +911,7 @@ fib_test_v4 (void) fib_table_entry_path_add(fib_index, &pfx, FIB_SOURCE_API, FIB_ENTRY_FLAG_NONE, - FIB_PROTOCOL_IP4, + DPO_PROTO_IP4, &nh_10_10_10_1, tm->hw[0]->sw_if_index, ~0, // invalid fib index @@ -963,7 +963,7 @@ fib_test_v4 (void) pfx.fp_len = 0; fib_table_entry_path_remove(fib_index, &pfx, FIB_SOURCE_API, - FIB_PROTOCOL_IP4, + DPO_PROTO_IP4, &nh_10_10_10_1, tm->hw[0]->sw_if_index, ~0, // non-recursive path, so no FIB index @@ -1029,7 +1029,7 @@ fib_test_v4 (void) &pfx_11_11_11_11_s_32, FIB_SOURCE_API, FIB_ENTRY_FLAG_ATTACHED, - FIB_PROTOCOL_IP4, + DPO_PROTO_IP4, &pfx_10_10_10_1_s_32.fp_addr, tm->hw[0]->sw_if_index, ~0, // invalid fib index @@ -1095,7 +1095,7 @@ fib_test_v4 (void) &pfx_10_10_10_1_s_32, FIB_SOURCE_ADJ, FIB_ENTRY_FLAG_ATTACHED, - FIB_PROTOCOL_IP4, + DPO_PROTO_IP4, &pfx_10_10_10_1_s_32.fp_addr, tm->hw[0]->sw_if_index, ~0, // invalid fib index @@ -1110,7 +1110,7 @@ fib_test_v4 (void) fib_table_entry_path_remove(fib_index, &pfx_11_11_11_11_s_32, FIB_SOURCE_API, - FIB_PROTOCOL_IP4, + DPO_PROTO_IP4, &pfx_10_10_10_1_s_32.fp_addr, tm->hw[0]->sw_if_index, ~0, // invalid fib index @@ -1144,7 +1144,7 @@ fib_test_v4 (void) &pfx_10_10_10_2_s_32, FIB_SOURCE_ADJ, FIB_ENTRY_FLAG_ATTACHED, - FIB_PROTOCOL_IP4, + DPO_PROTO_IP4, &pfx_10_10_10_2_s_32.fp_addr, tm->hw[0]->sw_if_index, ~0, // invalid fib index @@ -1181,7 +1181,7 @@ fib_test_v4 (void) &pfx_1_1_1_1_s_32, FIB_SOURCE_API, FIB_ENTRY_FLAG_NONE, - FIB_PROTOCOL_IP4, + DPO_PROTO_IP4, &nh_10_10_10_1, tm->hw[0]->sw_if_index, ~0, // invalid fib index @@ -1214,7 +1214,7 @@ fib_test_v4 (void) &pfx_1_1_2_0_s_24, FIB_SOURCE_API, FIB_ENTRY_FLAG_NONE, - FIB_PROTOCOL_IP4, + DPO_PROTO_IP4, &nh_10_10_10_1, tm->hw[0]->sw_if_index, ~0, // invalid fib index @@ -1241,7 +1241,7 @@ fib_test_v4 (void) &pfx_1_1_2_0_s_24, FIB_SOURCE_API, FIB_ENTRY_FLAG_NONE, - FIB_PROTOCOL_IP4, + DPO_PROTO_IP4, &nh_10_10_10_2, tm->hw[0]->sw_if_index, ~0, // invalid fib index @@ -1280,7 +1280,7 @@ fib_test_v4 (void) fib_table_entry_path_remove(fib_index, &pfx_1_1_2_0_s_24, FIB_SOURCE_API, - FIB_PROTOCOL_IP4, + DPO_PROTO_IP4, &nh_10_10_10_2, tm->hw[0]->sw_if_index, ~0, @@ -1327,7 +1327,7 @@ fib_test_v4 (void) &bgp_100_pfx, FIB_SOURCE_API, FIB_ENTRY_FLAG_NONE, - FIB_PROTOCOL_IP4, + DPO_PROTO_IP4, &nh_1_1_1_1, ~0, // no index provided. fib_index, // nexthop in same fib as route @@ -1363,7 +1363,7 @@ fib_test_v4 (void) &bgp_101_pfx, FIB_SOURCE_API, FIB_ENTRY_FLAG_NONE, - FIB_PROTOCOL_IP4, + DPO_PROTO_IP4, &nh_1_1_1_1, ~0, // no index provided. fib_index, // nexthop in same fib as route @@ -1487,7 +1487,7 @@ fib_test_v4 (void) &bgp_200_pfx, FIB_SOURCE_API, FIB_ENTRY_FLAG_NONE, - FIB_PROTOCOL_IP4, + DPO_PROTO_IP4, &pfx_1_1_1_2_s_32.fp_addr, ~0, // no index provided. fib_index, // nexthop in same fib as route @@ -1534,7 +1534,7 @@ fib_test_v4 (void) &pfx_1_2_3_4_s_32, FIB_SOURCE_API, FIB_ENTRY_FLAG_NONE, - FIB_PROTOCOL_IP4, + DPO_PROTO_IP4, &nh_10_10_10_1, tm->hw[0]->sw_if_index, ~0, @@ -1545,7 +1545,7 @@ fib_test_v4 (void) &pfx_1_2_3_4_s_32, FIB_SOURCE_API, FIB_ENTRY_FLAG_NONE, - FIB_PROTOCOL_IP4, + DPO_PROTO_IP4, &nh_12_12_12_12, tm->hw[1]->sw_if_index, ~0, @@ -1586,7 +1586,7 @@ fib_test_v4 (void) &pfx_1_2_3_5_s_32, FIB_SOURCE_API, FIB_ENTRY_FLAG_NONE, - FIB_PROTOCOL_IP4, + DPO_PROTO_IP4, &nh_12_12_12_12, tm->hw[1]->sw_if_index, ~0, @@ -1597,7 +1597,7 @@ fib_test_v4 (void) &pfx_1_2_3_5_s_32, FIB_SOURCE_API, FIB_ENTRY_FLAG_NONE, - FIB_PROTOCOL_IP4, + DPO_PROTO_IP4, &nh_10_10_10_1, tm->hw[0]->sw_if_index, ~0, @@ -1669,7 +1669,7 @@ fib_test_v4 (void) &pfx_6_6_6_6_s_32, FIB_SOURCE_API, FIB_ENTRY_FLAG_NONE, - FIB_PROTOCOL_IP4, + DPO_PROTO_IP4, &nh_10_10_10_1, tm->hw[0]->sw_if_index, ~0, // invalid fib index @@ -1688,7 +1688,7 @@ fib_test_v4 (void) &pfx_6_6_6_6_s_32, FIB_SOURCE_API, FIB_ENTRY_FLAG_NONE, - FIB_PROTOCOL_IP4, + DPO_PROTO_IP4, &nh_10_10_10_2, tm->hw[0]->sw_if_index, ~0, // invalid fib index @@ -1770,7 +1770,7 @@ fib_test_v4 (void) &pfx_6_6_6_6_s_32, FIB_SOURCE_API, FIB_ENTRY_FLAG_NONE, - FIB_PROTOCOL_IP4, + DPO_PROTO_IP4, &nh_12_12_12_12, tm->hw[1]->sw_if_index, ~0, // invalid fib index @@ -1915,7 +1915,7 @@ fib_test_v4 (void) fib_table_entry_path_remove(fib_index, &pfx_6_6_6_6_s_32, FIB_SOURCE_API, - FIB_PROTOCOL_IP4, + DPO_PROTO_IP4, &nh_12_12_12_12, tm->hw[1]->sw_if_index, ~0, // invalid fib index @@ -1995,7 +1995,7 @@ fib_test_v4 (void) fib_table_entry_path_remove(fib_index, &pfx_6_6_6_6_s_32, FIB_SOURCE_API, - FIB_PROTOCOL_IP4, + DPO_PROTO_IP4, &nh_10_10_10_2, tm->hw[0]->sw_if_index, ~0, // invalid fib index @@ -2026,7 +2026,7 @@ fib_test_v4 (void) &bgp_44_s_32, FIB_SOURCE_API, FIB_ENTRY_FLAG_NONE, - FIB_PROTOCOL_IP4, + DPO_PROTO_IP4, &pfx_1_2_3_4_s_32.fp_addr, ~0, fib_index, @@ -2037,7 +2037,7 @@ fib_test_v4 (void) &bgp_44_s_32, FIB_SOURCE_API, FIB_ENTRY_FLAG_NONE, - FIB_PROTOCOL_IP4, + DPO_PROTO_IP4, &pfx_1_2_3_5_s_32.fp_addr, ~0, fib_index, @@ -2107,7 +2107,7 @@ fib_test_v4 (void) &bgp_201_pfx, FIB_SOURCE_API, FIB_ENTRY_FLAG_NONE, - FIB_PROTOCOL_IP4, + DPO_PROTO_IP4, &pfx_1_1_1_200_s_32.fp_addr, ~0, // no index provided. fib_index, // nexthop in same fib as route @@ -2151,7 +2151,7 @@ fib_test_v4 (void) &pfx_1_1_1_0_s_24, FIB_SOURCE_API, FIB_ENTRY_FLAG_NONE, - FIB_PROTOCOL_IP4, + DPO_PROTO_IP4, &nh_10_10_10_1, tm->hw[0]->sw_if_index, ~0, // invalid fib index @@ -2209,7 +2209,7 @@ fib_test_v4 (void) &pfx_1_1_1_0_s_28, FIB_SOURCE_API, FIB_ENTRY_FLAG_NONE, - FIB_PROTOCOL_IP4, + DPO_PROTO_IP4, &nh_10_10_10_2, tm->hw[0]->sw_if_index, ~0, // invalid fib index @@ -2244,7 +2244,7 @@ fib_test_v4 (void) fib_table_entry_path_remove(fib_index, &pfx_1_1_1_0_s_28, FIB_SOURCE_API, - FIB_PROTOCOL_IP4, + DPO_PROTO_IP4, &nh_10_10_10_2, tm->hw[0]->sw_if_index, ~0, @@ -2275,7 +2275,7 @@ fib_test_v4 (void) fib_table_entry_path_remove(fib_index, &pfx_1_1_1_0_s_24, FIB_SOURCE_API, - FIB_PROTOCOL_IP4, + DPO_PROTO_IP4, &nh_10_10_10_1, tm->hw[0]->sw_if_index, ~0, @@ -2316,7 +2316,7 @@ fib_test_v4 (void) &pfx_1_1_1_2_s_32, FIB_SOURCE_API, FIB_ENTRY_FLAG_NONE, - FIB_PROTOCOL_IP4, + DPO_PROTO_IP4, &nh_10_10_10_1, tm->hw[0]->sw_if_index, ~0, // invalid fib index @@ -2351,7 +2351,7 @@ fib_test_v4 (void) &bgp_201_pfx, FIB_SOURCE_API, FIB_ENTRY_FLAG_NONE, - FIB_PROTOCOL_IP4, + DPO_PROTO_IP4, &pfx_1_1_1_2_s_32.fp_addr, ~0, fib_index, @@ -2362,7 +2362,7 @@ fib_test_v4 (void) fib_table_entry_path_remove(fib_index, &bgp_201_pfx, FIB_SOURCE_API, - FIB_PROTOCOL_IP4, + DPO_PROTO_IP4, &pfx_1_1_1_2_s_32.fp_addr, ~0, fib_index, @@ -2375,7 +2375,7 @@ fib_test_v4 (void) fib_table_entry_path_remove(fib_index, &bgp_201_pfx, FIB_SOURCE_API, - FIB_PROTOCOL_IP4, + DPO_PROTO_IP4, &pfx_1_1_1_200_s_32.fp_addr, ~0, // no index provided. fib_index, @@ -2405,7 +2405,7 @@ fib_test_v4 (void) fib_table_entry_path_remove(fib_index, &bgp_200_pfx, FIB_SOURCE_API, - FIB_PROTOCOL_IP4, + DPO_PROTO_IP4, &pfx_1_1_1_2_s_32.fp_addr, ~0, // no index provided. fib_index, @@ -2446,7 +2446,7 @@ fib_test_v4 (void) &bgp_102, FIB_SOURCE_API, FIB_ENTRY_FLAG_NONE, - FIB_PROTOCOL_IP4, + DPO_PROTO_IP4, &pfx_1_1_1_1_s_32.fp_addr, ~0, // no index provided. fib_index, // same as route @@ -2457,7 +2457,7 @@ fib_test_v4 (void) &bgp_102, FIB_SOURCE_API, FIB_ENTRY_FLAG_NONE, - FIB_PROTOCOL_IP4, + DPO_PROTO_IP4, &pfx_1_1_1_2_s_32.fp_addr, ~0, // no index provided. fib_index, // same as route's FIB @@ -2483,7 +2483,7 @@ fib_test_v4 (void) fib_table_entry_path_remove(fib_index, &bgp_102, FIB_SOURCE_API, - FIB_PROTOCOL_IP4, + DPO_PROTO_IP4, &pfx_1_1_1_1_s_32.fp_addr, ~0, // no index provided. fib_index, // same as route's FIB @@ -2492,7 +2492,7 @@ fib_test_v4 (void) fib_table_entry_path_remove(fib_index, &bgp_102, FIB_SOURCE_API, - FIB_PROTOCOL_IP4, + DPO_PROTO_IP4, &pfx_1_1_1_2_s_32.fp_addr, ~0, // no index provided. fib_index, // same as route's FIB @@ -2507,7 +2507,7 @@ fib_test_v4 (void) fib_table_entry_path_remove(fib_index, &bgp_100_pfx, FIB_SOURCE_API, - FIB_PROTOCOL_IP4, + DPO_PROTO_IP4, &pfx_1_1_1_1_s_32.fp_addr, ~0, // no index provided. fib_index, // same as route's FIB @@ -2516,7 +2516,7 @@ fib_test_v4 (void) fib_table_entry_path_remove(fib_index, &bgp_101_pfx, FIB_SOURCE_API, - FIB_PROTOCOL_IP4, + DPO_PROTO_IP4, &pfx_1_1_1_1_s_32.fp_addr, ~0, // no index provided. fib_index, // same as route's FIB @@ -2546,7 +2546,7 @@ fib_test_v4 (void) &bgp_200_pfx, FIB_SOURCE_API, FIB_ENTRY_FLAG_NONE, - FIB_PROTOCOL_IP4, + DPO_PROTO_IP4, &nh_10_10_10_1, ~0, // no index provided. fib_index, // Same as route's FIB @@ -2593,7 +2593,7 @@ fib_test_v4 (void) &bgp_201_pfx, FIB_SOURCE_API, FIB_ENTRY_FLAG_NONE, - FIB_PROTOCOL_IP4, + DPO_PROTO_IP4, &nh_10_10_10_3, ~0, // no index provided. fib_index, @@ -2639,7 +2639,7 @@ fib_test_v4 (void) fib_table_entry_path_remove(fib_index, &bgp_200_pfx, FIB_SOURCE_API, - FIB_PROTOCOL_IP4, + DPO_PROTO_IP4, &nh_10_10_10_1, ~0, // no index provided. fib_index, // same as route's FIB @@ -2648,7 +2648,7 @@ fib_test_v4 (void) fib_table_entry_path_remove(fib_index, &bgp_201_pfx, FIB_SOURCE_API, - FIB_PROTOCOL_IP4, + DPO_PROTO_IP4, &nh_10_10_10_3, ~0, // no index provided. fib_index, // same as route's FIB @@ -2707,7 +2707,7 @@ fib_test_v4 (void) &pfx_5_5_5_5_s_32, FIB_SOURCE_API, FIB_ENTRY_FLAG_NONE, - FIB_PROTOCOL_IP4, + DPO_PROTO_IP4, &pfx_5_5_5_6_s_32.fp_addr, ~0, // no index provided. fib_index, @@ -2718,7 +2718,7 @@ fib_test_v4 (void) &pfx_5_5_5_6_s_32, FIB_SOURCE_API, FIB_ENTRY_FLAG_NONE, - FIB_PROTOCOL_IP4, + DPO_PROTO_IP4, &pfx_5_5_5_7_s_32.fp_addr, ~0, // no index provided. fib_index, @@ -2729,7 +2729,7 @@ fib_test_v4 (void) &pfx_5_5_5_7_s_32, FIB_SOURCE_API, FIB_ENTRY_FLAG_NONE, - FIB_PROTOCOL_IP4, + DPO_PROTO_IP4, &pfx_5_5_5_5_s_32.fp_addr, ~0, // no index provided. fib_index, @@ -2768,7 +2768,7 @@ fib_test_v4 (void) &pfx_5_5_5_6_s_32, FIB_SOURCE_API, FIB_ENTRY_FLAG_NONE, - FIB_PROTOCOL_IP4, + DPO_PROTO_IP4, &nh_10_10_10_1, tm->hw[0]->sw_if_index, ~0, @@ -2801,7 +2801,7 @@ fib_test_v4 (void) fib_table_entry_path_remove(fib_index, &pfx_5_5_5_6_s_32, FIB_SOURCE_API, - FIB_PROTOCOL_IP4, + DPO_PROTO_IP4, &nh_10_10_10_1, tm->hw[0]->sw_if_index, ~0, @@ -2826,7 +2826,7 @@ fib_test_v4 (void) &pfx_5_5_5_5_s_32, FIB_SOURCE_API, FIB_ENTRY_FLAG_NONE, - FIB_PROTOCOL_IP4, + DPO_PROTO_IP4, &nh_10_10_10_1, tm->hw[0]->sw_if_index, ~0, // invalid fib index @@ -2868,7 +2868,7 @@ fib_test_v4 (void) &pfx_5_5_5_5_s_32, FIB_SOURCE_API, FIB_ENTRY_FLAG_NONE, - FIB_PROTOCOL_IP4, + DPO_PROTO_IP4, &pfx_5_5_5_6_s_32.fp_addr, ~0, // no index provided. fib_index, @@ -2892,7 +2892,7 @@ fib_test_v4 (void) fib_table_entry_path_remove(fib_index, &pfx_5_5_5_5_s_32, FIB_SOURCE_API, - FIB_PROTOCOL_IP4, + DPO_PROTO_IP4, &pfx_5_5_5_6_s_32.fp_addr, ~0, // no index provided. fib_index, // same as route's FIB @@ -2901,7 +2901,7 @@ fib_test_v4 (void) fib_table_entry_path_remove(fib_index, &pfx_5_5_5_6_s_32, FIB_SOURCE_API, - FIB_PROTOCOL_IP4, + DPO_PROTO_IP4, &pfx_5_5_5_7_s_32.fp_addr, ~0, // no index provided. fib_index, // same as route's FIB @@ -2910,7 +2910,7 @@ fib_test_v4 (void) fib_table_entry_path_remove(fib_index, &pfx_5_5_5_7_s_32, FIB_SOURCE_API, - FIB_PROTOCOL_IP4, + DPO_PROTO_IP4, &pfx_5_5_5_5_s_32.fp_addr, ~0, // no index provided. fib_index, // same as route's FIB @@ -2919,7 +2919,7 @@ fib_test_v4 (void) fib_table_entry_path_remove(fib_index, &pfx_5_5_5_6_s_32, FIB_SOURCE_API, - FIB_PROTOCOL_IP4, + DPO_PROTO_IP4, &nh_10_10_10_2, ~0, // no index provided. fib_index, // same as route's FIB @@ -2943,7 +2943,7 @@ fib_test_v4 (void) &pfx_5_5_5_6_s_32, FIB_SOURCE_API, FIB_ENTRY_FLAG_NONE, - FIB_PROTOCOL_IP4, + DPO_PROTO_IP4, &pfx_5_5_5_6_s_32.fp_addr, ~0, // no index provided. fib_index, @@ -2957,7 +2957,7 @@ fib_test_v4 (void) fib_table_entry_path_remove(fib_index, &pfx_5_5_5_6_s_32, FIB_SOURCE_API, - FIB_PROTOCOL_IP4, + DPO_PROTO_IP4, &pfx_5_5_5_6_s_32.fp_addr, ~0, // no index provided. fib_index, // same as route's FIB @@ -2991,7 +2991,7 @@ fib_test_v4 (void) &pfx_23_23_23_0_s_24, FIB_SOURCE_API, FIB_ENTRY_FLAG_NONE, - FIB_PROTOCOL_IP4, + DPO_PROTO_IP4, &pfx_23_23_23_23_s_32.fp_addr, ~0, // recursive fib_index, @@ -3021,7 +3021,7 @@ fib_test_v4 (void) &pfx_0_0_0_0_s_0, FIB_SOURCE_API, FIB_ENTRY_FLAG_NONE, - FIB_PROTOCOL_IP4, + DPO_PROTO_IP4, &pfx_23_23_23_23_s_32.fp_addr, ~0, // recursive fib_index, @@ -3051,7 +3051,7 @@ fib_test_v4 (void) &bgp_200_pfx, FIB_SOURCE_API, FIB_ENTRY_FLAG_NONE, - FIB_PROTOCOL_IP4, + DPO_PROTO_IP4, &nh_1_1_1_1, ~0, fib_index, @@ -3081,7 +3081,7 @@ fib_test_v4 (void) &pfx_1_1_1_0_s_28, FIB_SOURCE_API, FIB_ENTRY_FLAG_NONE, - FIB_PROTOCOL_IP4, + DPO_PROTO_IP4, &nh_10_10_10_1, tm->hw[0]->sw_if_index, ~0, // invalid fib index @@ -3099,7 +3099,7 @@ fib_test_v4 (void) fib_table_entry_path_remove(fib_index, &pfx_1_1_1_1_s_32, FIB_SOURCE_API, - FIB_PROTOCOL_IP4, + DPO_PROTO_IP4, &nh_10_10_10_1, tm->hw[0]->sw_if_index, ~0, // invalid fib index @@ -3116,7 +3116,7 @@ fib_test_v4 (void) &pfx_1_1_1_1_s_32, FIB_SOURCE_API, FIB_ENTRY_FLAG_NONE, - FIB_PROTOCOL_IP4, + DPO_PROTO_IP4, &nh_10_10_10_1, tm->hw[0]->sw_if_index, ~0, // invalid fib index @@ -3140,7 +3140,7 @@ fib_test_v4 (void) &pfx_1_1_1_3_s_32, FIB_SOURCE_API, FIB_ENTRY_FLAG_NONE, - FIB_PROTOCOL_IP4, + DPO_PROTO_IP4, &nh_10_10_10_2, tm->hw[0]->sw_if_index, ~0, // invalid fib index @@ -3152,7 +3152,7 @@ fib_test_v4 (void) &bgp_200_pfx, FIB_SOURCE_API, FIB_ENTRY_FLAG_NONE, - FIB_PROTOCOL_IP4, + DPO_PROTO_IP4, &pfx_1_1_1_3_s_32.fp_addr, ~0, fib_index, @@ -3177,7 +3177,7 @@ fib_test_v4 (void) &bgp_78s[ii], FIB_SOURCE_API, FIB_ENTRY_FLAG_NONE, - FIB_PROTOCOL_IP4, + DPO_PROTO_IP4, &pfx_1_1_1_3_s_32.fp_addr, ~0, fib_index, @@ -3188,7 +3188,7 @@ fib_test_v4 (void) &bgp_78s[ii], FIB_SOURCE_API, FIB_ENTRY_FLAG_NONE, - FIB_PROTOCOL_IP4, + DPO_PROTO_IP4, &nh_1_1_1_1, ~0, fib_index, @@ -3238,7 +3238,7 @@ fib_test_v4 (void) fib_table_entry_path_remove(fib_index, &pfx_1_1_1_1_s_32, FIB_SOURCE_API, - FIB_PROTOCOL_IP4, + DPO_PROTO_IP4, &nh_10_10_10_1, tm->hw[0]->sw_if_index, ~0, // invalid fib index @@ -3277,7 +3277,7 @@ fib_test_v4 (void) &pfx_1_1_1_1_s_32, FIB_SOURCE_API, FIB_ENTRY_FLAG_NONE, - FIB_PROTOCOL_IP4, + DPO_PROTO_IP4, &nh_10_10_10_1, tm->hw[0]->sw_if_index, ~0, // invalid fib index @@ -3307,7 +3307,7 @@ fib_test_v4 (void) &bgp_200_pfx, FIB_SOURCE_API, FIB_ENTRY_FLAG_NONE, - FIB_PROTOCOL_IP4, + DPO_PROTO_IP4, &pfx_1_1_1_2_s_32.fp_addr, ~0, fib_index, @@ -3320,7 +3320,7 @@ fib_test_v4 (void) &bgp_78s[ii], FIB_SOURCE_API, FIB_ENTRY_FLAG_NONE, - FIB_PROTOCOL_IP4, + DPO_PROTO_IP4, &pfx_1_1_1_2_s_32.fp_addr, ~0, fib_index, @@ -3354,7 +3354,7 @@ fib_test_v4 (void) fib_table_entry_path_remove(fib_index, &pfx_1_1_1_1_s_32, FIB_SOURCE_API, - FIB_PROTOCOL_IP4, + DPO_PROTO_IP4, &nh_10_10_10_1, tm->hw[0]->sw_if_index, ~0, @@ -3391,7 +3391,7 @@ fib_test_v4 (void) &pfx_1_1_1_1_s_32, FIB_SOURCE_API, FIB_ENTRY_FLAG_NONE, - FIB_PROTOCOL_IP4, + DPO_PROTO_IP4, &nh_10_10_10_1, tm->hw[0]->sw_if_index, ~0, @@ -3412,7 +3412,7 @@ fib_test_v4 (void) fib_table_entry_path_remove(fib_index, &bgp_200_pfx, FIB_SOURCE_API, - FIB_PROTOCOL_IP4, + DPO_PROTO_IP4, &pfx_1_1_1_2_s_32.fp_addr, ~0, fib_index, @@ -3421,7 +3421,7 @@ fib_test_v4 (void) fib_table_entry_path_remove(fib_index, &bgp_200_pfx, FIB_SOURCE_API, - FIB_PROTOCOL_IP4, + DPO_PROTO_IP4, &nh_1_1_1_1, ~0, fib_index, @@ -3430,7 +3430,7 @@ fib_test_v4 (void) fib_table_entry_path_remove(fib_index, &bgp_200_pfx, FIB_SOURCE_API, - FIB_PROTOCOL_IP4, + DPO_PROTO_IP4, &pfx_1_1_1_3_s_32.fp_addr, ~0, fib_index, @@ -3481,7 +3481,7 @@ fib_test_v4 (void) &pfx_4_4_4_4_s_32, FIB_SOURCE_API, FIB_ENTRY_FLAG_NONE, - FIB_PROTOCOL_IP4, + DPO_PROTO_IP4, &nh_10_10_10_1, tm->hw[0]->sw_if_index, ~0, @@ -3492,7 +3492,7 @@ fib_test_v4 (void) &pfx_4_4_4_4_s_32, FIB_SOURCE_API, FIB_ENTRY_FLAG_NONE, - FIB_PROTOCOL_IP4, + DPO_PROTO_IP4, &nh_10_10_10_2, tm->hw[0]->sw_if_index, ~0, @@ -3503,7 +3503,7 @@ fib_test_v4 (void) &pfx_4_4_4_4_s_32, FIB_SOURCE_API, FIB_ENTRY_FLAG_NONE, - FIB_PROTOCOL_IP4, + DPO_PROTO_IP4, &nh_10_10_10_3, tm->hw[0]->sw_if_index, ~0, @@ -3539,7 +3539,7 @@ fib_test_v4 (void) for (ii = 0; ii < 4; ii++) { fib_route_path_t r_path = { - .frp_proto = FIB_PROTOCOL_IP4, + .frp_proto = DPO_PROTO_IP4, .frp_addr = { .ip4.as_u32 = clib_host_to_net_u32(0x0a0a0a02 + ii), }, @@ -3588,7 +3588,7 @@ fib_test_v4 (void) &pfx_4_4_4_4_s_32, FIB_SOURCE_API, FIB_ENTRY_FLAG_NONE, - FIB_PROTOCOL_IP4, + DPO_PROTO_IP4, &zero_addr, ~0, fib_index, @@ -3648,7 +3648,7 @@ fib_test_v4 (void) &pfx_34_34_1_1_s_32, FIB_SOURCE_API, FIB_ENTRY_FLAG_NONE, - FIB_PROTOCOL_IP4, + DPO_PROTO_IP4, &nh_10_10_10_1, tm->hw[0]->sw_if_index, 0, @@ -3659,7 +3659,7 @@ fib_test_v4 (void) &pfx_34_1_1_1_s_32, FIB_SOURCE_API, FIB_ENTRY_FLAG_NONE, - FIB_PROTOCOL_IP4, + DPO_PROTO_IP4, &pfx_34_34_1_1_s_32.fp_addr, ~0, fib_index, @@ -3670,7 +3670,7 @@ fib_test_v4 (void) &pfx_34_1_1_1_s_32, FIB_SOURCE_API, FIB_ENTRY_FLAG_NONE, - FIB_PROTOCOL_IP4, + DPO_PROTO_IP4, &pfx_34_34_1_1_s_32.fp_addr, ~0, fib_index, @@ -3691,7 +3691,7 @@ fib_test_v4 (void) fib_table_entry_path_remove(fib_index, &pfx_1_1_1_2_s_32, FIB_SOURCE_API, - FIB_PROTOCOL_IP4, + DPO_PROTO_IP4, &nh_10_10_10_1, tm->hw[0]->sw_if_index, ~0, @@ -3700,7 +3700,7 @@ fib_test_v4 (void) fib_table_entry_path_remove(fib_index, &pfx_1_1_1_1_s_32, FIB_SOURCE_API, - FIB_PROTOCOL_IP4, + DPO_PROTO_IP4, &nh_10_10_10_1, tm->hw[0]->sw_if_index, ~0, @@ -3709,7 +3709,7 @@ fib_test_v4 (void) fib_table_entry_path_remove(fib_index, &pfx_1_1_2_0_s_24, FIB_SOURCE_API, - FIB_PROTOCOL_IP4, + DPO_PROTO_IP4, &nh_10_10_10_1, tm->hw[0]->sw_if_index, ~0, @@ -3751,7 +3751,7 @@ fib_test_v4 (void) &pfx_4_1_1_1_s_32, FIB_SOURCE_API, FIB_ENTRY_FLAG_NONE, - FIB_PROTOCOL_IP4, + DPO_PROTO_IP4, &zero_addr, tm->hw[0]->sw_if_index, fib_index, @@ -3805,7 +3805,7 @@ fib_test_v4 (void) &pfx_2001_s_64, FIB_SOURCE_API, FIB_ENTRY_FLAG_NONE, - FIB_PROTOCOL_IP4, + DPO_PROTO_IP4, &nh_10_10_10_1, tm->hw[0]->sw_if_index, fib_index, @@ -3863,7 +3863,7 @@ fib_test_v4 (void) &pfx_12_10_10_2_s_32, FIB_SOURCE_ADJ, FIB_ENTRY_FLAG_ATTACHED, - FIB_PROTOCOL_IP4, + DPO_PROTO_IP4, &pfx_12_10_10_2_s_32.fp_addr, tm->hw[0]->sw_if_index, ~0, // invalid fib index @@ -3897,7 +3897,7 @@ fib_test_v4 (void) &pfx_10_10_10_127_s_32, FIB_SOURCE_ADJ, FIB_ENTRY_FLAG_ATTACHED, - FIB_PROTOCOL_IP4, + DPO_PROTO_IP4, &pfx_10_10_10_127_s_32.fp_addr, tm->hw[1]->sw_if_index, ~0, // invalid fib index @@ -3945,7 +3945,7 @@ fib_test_v4 (void) &pfx_10_10_10_3_s_32, FIB_SOURCE_ADJ, FIB_ENTRY_FLAG_NONE, - FIB_PROTOCOL_IP4, + DPO_PROTO_IP4, &nh_10_10_10_3, tm->hw[0]->sw_if_index, fib_index, @@ -3956,7 +3956,7 @@ fib_test_v4 (void) &pfx_10_10_10_3_s_32, FIB_SOURCE_ADJ, FIB_ENTRY_FLAG_NONE, - FIB_PROTOCOL_IP4, + DPO_PROTO_IP4, &nh_12_12_12_12, tm->hw[1]->sw_if_index, fib_index, @@ -3975,7 +3975,7 @@ fib_test_v4 (void) fib_table_entry_path_remove(fib_index, &pfx_10_10_10_3_s_32, FIB_SOURCE_ADJ, - FIB_PROTOCOL_IP4, + DPO_PROTO_IP4, &nh_10_10_10_3, tm->hw[0]->sw_if_index, fib_index, @@ -3992,7 +3992,7 @@ fib_test_v4 (void) &pfx_10_10_10_3_s_32, FIB_SOURCE_ADJ, FIB_ENTRY_FLAG_NONE, - FIB_PROTOCOL_IP4, + DPO_PROTO_IP4, &nh_10_10_10_3, tm->hw[0]->sw_if_index, fib_index, @@ -4011,7 +4011,7 @@ fib_test_v4 (void) fib_table_entry_path_remove(fib_index, &pfx_10_10_10_3_s_32, FIB_SOURCE_ADJ, - FIB_PROTOCOL_IP4, + DPO_PROTO_IP4, &nh_12_12_12_12, tm->hw[1]->sw_if_index, fib_index, @@ -4030,7 +4030,7 @@ fib_test_v4 (void) fib_table_entry_path_remove(fib_index, &pfx_10_10_10_3_s_32, FIB_SOURCE_ADJ, - FIB_PROTOCOL_IP4, + DPO_PROTO_IP4, &nh_10_10_10_3, tm->hw[0]->sw_if_index, fib_index, @@ -4269,7 +4269,7 @@ fib_test_v6 (void) FIB_SOURCE_INTERFACE, (FIB_ENTRY_FLAG_CONNECTED | FIB_ENTRY_FLAG_ATTACHED), - FIB_PROTOCOL_IP6, + DPO_PROTO_IP6, NULL, tm->hw[0]->sw_if_index, ~0, @@ -4300,7 +4300,7 @@ fib_test_v6 (void) FIB_SOURCE_INTERFACE, (FIB_ENTRY_FLAG_CONNECTED | FIB_ENTRY_FLAG_LOCAL), - FIB_PROTOCOL_IP6, + DPO_PROTO_IP6, NULL, tm->hw[0]->sw_if_index, ~0, // invalid fib index @@ -4345,7 +4345,7 @@ fib_test_v6 (void) fib_table_entry_path_add(fib_index, &pfx_0_0, FIB_SOURCE_API, FIB_ENTRY_FLAG_NONE, - FIB_PROTOCOL_IP6, + DPO_PROTO_IP6, &nh_2001_2, tm->hw[0]->sw_if_index, ~0, @@ -4389,7 +4389,7 @@ fib_test_v6 (void) */ fib_table_entry_path_remove(fib_index, &pfx_0_0, FIB_SOURCE_API, - FIB_PROTOCOL_IP6, + DPO_PROTO_IP6, &nh_2001_2, tm->hw[0]->sw_if_index, ~0, @@ -4466,7 +4466,7 @@ fib_test_v6 (void) &pfx_2001_1_2_s_128, FIB_SOURCE_ADJ, FIB_ENTRY_FLAG_ATTACHED, - FIB_PROTOCOL_IP6, + DPO_PROTO_IP6, &pfx_2001_1_2_s_128.fp_addr, tm->hw[0]->sw_if_index, ~0, @@ -4505,7 +4505,7 @@ fib_test_v6 (void) &pfx_2001_1_3_s_128, FIB_SOURCE_ADJ, FIB_ENTRY_FLAG_ATTACHED, - FIB_PROTOCOL_IP6, + DPO_PROTO_IP6, &pfx_2001_1_3_s_128.fp_addr, tm->hw[0]->sw_if_index, ~0, @@ -4559,7 +4559,7 @@ fib_test_v6 (void) &pfx_2001_a_s_64, FIB_SOURCE_API, FIB_ENTRY_FLAG_NONE, - FIB_PROTOCOL_IP6, + DPO_PROTO_IP6, &nh_2001_2, tm->hw[0]->sw_if_index, ~0, @@ -4573,7 +4573,7 @@ fib_test_v6 (void) &pfx_2001_b_s_64, FIB_SOURCE_API, FIB_ENTRY_FLAG_NONE, - FIB_PROTOCOL_IP6, + DPO_PROTO_IP6, &nh_2001_2, tm->hw[0]->sw_if_index, ~0, @@ -4608,7 +4608,7 @@ fib_test_v6 (void) &pfx_1_1_1_1_s_32, FIB_SOURCE_API, FIB_ENTRY_FLAG_NONE, - FIB_PROTOCOL_IP6, + DPO_PROTO_IP6, &nh_2001_2, tm->hw[0]->sw_if_index, ~0, @@ -4646,7 +4646,7 @@ fib_test_v6 (void) &pfx_2001_c_s_64, FIB_SOURCE_CLI, FIB_ENTRY_FLAG_ATTACHED, - FIB_PROTOCOL_IP6, + DPO_PROTO_IP6, NULL, tm->hw[0]->sw_if_index, ~0, @@ -4663,7 +4663,7 @@ fib_test_v6 (void) fib_table_entry_path_remove(fib_index, &pfx_2001_c_s_64, FIB_SOURCE_CLI, - FIB_PROTOCOL_IP6, + DPO_PROTO_IP6, NULL, tm->hw[0]->sw_if_index, ~0, @@ -4748,7 +4748,7 @@ fib_test_v6 (void) FIB_SOURCE_INTERFACE, (FIB_ENTRY_FLAG_CONNECTED | FIB_ENTRY_FLAG_ATTACHED), - FIB_PROTOCOL_IP6, + DPO_PROTO_IP6, NULL, tm->hw[1]->sw_if_index, ~0, @@ -4767,7 +4767,7 @@ fib_test_v6 (void) FIB_SOURCE_INTERFACE, (FIB_ENTRY_FLAG_CONNECTED | FIB_ENTRY_FLAG_LOCAL), - FIB_PROTOCOL_IP6, + DPO_PROTO_IP6, NULL, tm->hw[0]->sw_if_index, ~0, // invalid fib index @@ -5095,7 +5095,7 @@ fib_test_ae (void) FIB_SOURCE_INTERFACE, (FIB_ENTRY_FLAG_CONNECTED | FIB_ENTRY_FLAG_ATTACHED), - FIB_PROTOCOL_IP4, + DPO_PROTO_IP4, NULL, tm->hw[0]->sw_if_index, ~0, @@ -5111,7 +5111,7 @@ fib_test_ae (void) FIB_SOURCE_INTERFACE, (FIB_ENTRY_FLAG_CONNECTED | FIB_ENTRY_FLAG_LOCAL), - FIB_PROTOCOL_IP4, + DPO_PROTO_IP4, NULL, tm->hw[0]->sw_if_index, ~0, // invalid fib index @@ -5140,7 +5140,7 @@ fib_test_ae (void) &pfx_10_10_10_1_s_32, FIB_SOURCE_ADJ, FIB_ENTRY_FLAG_ATTACHED, - FIB_PROTOCOL_IP4, + DPO_PROTO_IP4, &pfx_10_10_10_1_s_32.fp_addr, tm->hw[0]->sw_if_index, ~0, // invalid fib index @@ -5167,7 +5167,7 @@ fib_test_ae (void) &local_pfx, FIB_SOURCE_API, FIB_ENTRY_FLAG_NONE, - FIB_PROTOCOL_IP4, + DPO_PROTO_IP4, NULL, tm->hw[0]->sw_if_index, ~0, // invalid fib index @@ -5209,7 +5209,7 @@ fib_test_ae (void) &pfx_10_10_10_2_s_32, FIB_SOURCE_ADJ, FIB_ENTRY_FLAG_ATTACHED, - FIB_PROTOCOL_IP4, + DPO_PROTO_IP4, &pfx_10_10_10_2_s_32.fp_addr, tm->hw[0]->sw_if_index, ~0, // invalid fib index @@ -5243,7 +5243,7 @@ fib_test_ae (void) &local_pfx, FIB_SOURCE_API, FIB_ENTRY_FLAG_NONE, - FIB_PROTOCOL_IP4, + DPO_PROTO_IP4, NULL, tm->hw[0]->sw_if_index, ~0, // invalid fib index @@ -5280,7 +5280,7 @@ fib_test_ae (void) &pfx_10_10_10_3_s_32, FIB_SOURCE_ADJ, FIB_ENTRY_FLAG_ATTACHED, - FIB_PROTOCOL_IP4, + DPO_PROTO_IP4, &pfx_10_10_10_3_s_32.fp_addr, tm->hw[0]->sw_if_index, ~0, // invalid fib index @@ -5352,7 +5352,7 @@ fib_test_ae (void) &local_pfx, FIB_SOURCE_API, FIB_ENTRY_FLAG_NONE, - FIB_PROTOCOL_IP4, + DPO_PROTO_IP4, &pfx_10_10_10_2_s_32.fp_addr, tm->hw[0]->sw_if_index, ~0, // invalid fib index @@ -5375,7 +5375,7 @@ fib_test_ae (void) &local_pfx, FIB_SOURCE_API, FIB_ENTRY_FLAG_NONE, - FIB_PROTOCOL_IP4, + DPO_PROTO_IP4, NULL, tm->hw[0]->sw_if_index, ~0, // invalid fib index @@ -5407,7 +5407,7 @@ fib_test_ae (void) &pfx_10_0_0_0_s_8, FIB_SOURCE_API, FIB_ENTRY_FLAG_NONE, - FIB_PROTOCOL_IP4, + DPO_PROTO_IP4, &pfx_10_10_10_3_s_32.fp_addr, tm->hw[0]->sw_if_index, ~0, // invalid fib index @@ -5463,7 +5463,7 @@ fib_test_ae (void) &local_pfx, FIB_SOURCE_API, FIB_ENTRY_FLAG_NONE, - FIB_PROTOCOL_IP4, + DPO_PROTO_IP4, &pfx_10_10_10_1_s_32.fp_addr, tm->hw[0]->sw_if_index, ~0, // invalid fib index @@ -5499,7 +5499,7 @@ fib_test_ae (void) &local_pfx, FIB_SOURCE_API, FIB_ENTRY_FLAG_NONE, - FIB_PROTOCOL_IP4, + DPO_PROTO_IP4, NULL, tm->hw[0]->sw_if_index, ~0, // invalid fib index @@ -5538,7 +5538,7 @@ fib_test_ae (void) FIB_SOURCE_INTERFACE, (FIB_ENTRY_FLAG_CONNECTED | FIB_ENTRY_FLAG_ATTACHED), - FIB_PROTOCOL_IP4, + DPO_PROTO_IP4, NULL, tm->hw[0]->sw_if_index, ~0, @@ -5626,7 +5626,7 @@ fib_test_pref (void) * 2 high, 2 medium and 2 low preference non-recursive paths */ fib_route_path_t nr_path_hi_1 = { - .frp_proto = FIB_PROTOCOL_IP4, + .frp_proto = DPO_PROTO_IP4, .frp_sw_if_index = tm->hw[0]->sw_if_index, .frp_fib_index = ~0, .frp_weight = 1, @@ -5637,7 +5637,7 @@ fib_test_pref (void) }, }; fib_route_path_t nr_path_hi_2 = { - .frp_proto = FIB_PROTOCOL_IP4, + .frp_proto = DPO_PROTO_IP4, .frp_sw_if_index = tm->hw[0]->sw_if_index, .frp_fib_index = ~0, .frp_weight = 1, @@ -5648,7 +5648,7 @@ fib_test_pref (void) }, }; fib_route_path_t nr_path_med_1 = { - .frp_proto = FIB_PROTOCOL_IP4, + .frp_proto = DPO_PROTO_IP4, .frp_sw_if_index = tm->hw[1]->sw_if_index, .frp_fib_index = ~0, .frp_weight = 1, @@ -5659,7 +5659,7 @@ fib_test_pref (void) }, }; fib_route_path_t nr_path_med_2 = { - .frp_proto = FIB_PROTOCOL_IP4, + .frp_proto = DPO_PROTO_IP4, .frp_sw_if_index = tm->hw[1]->sw_if_index, .frp_fib_index = ~0, .frp_weight = 1, @@ -5670,7 +5670,7 @@ fib_test_pref (void) }, }; fib_route_path_t nr_path_low_1 = { - .frp_proto = FIB_PROTOCOL_IP4, + .frp_proto = DPO_PROTO_IP4, .frp_sw_if_index = tm->hw[2]->sw_if_index, .frp_fib_index = ~0, .frp_weight = 1, @@ -5681,7 +5681,7 @@ fib_test_pref (void) }, }; fib_route_path_t nr_path_low_2 = { - .frp_proto = FIB_PROTOCOL_IP4, + .frp_proto = DPO_PROTO_IP4, .frp_sw_if_index = tm->hw[2]->sw_if_index, .frp_fib_index = ~0, .frp_weight = 1, @@ -5897,7 +5897,7 @@ fib_test_pref (void) }, }; fib_route_path_t r_path_hi = { - .frp_proto = FIB_PROTOCOL_IP4, + .frp_proto = DPO_PROTO_IP4, .frp_sw_if_index = ~0, .frp_fib_index = 0, .frp_weight = 1, @@ -5906,7 +5906,7 @@ fib_test_pref (void) .frp_addr = pfx_1_1_1_1_s_32.fp_addr, }; fib_route_path_t r_path_med = { - .frp_proto = FIB_PROTOCOL_IP4, + .frp_proto = DPO_PROTO_IP4, .frp_sw_if_index = ~0, .frp_fib_index = 0, .frp_weight = 1, @@ -5915,7 +5915,7 @@ fib_test_pref (void) .frp_addr = pfx_1_1_1_2_s_32.fp_addr, }; fib_route_path_t r_path_low = { - .frp_proto = FIB_PROTOCOL_IP4, + .frp_proto = DPO_PROTO_IP4, .frp_sw_if_index = ~0, .frp_fib_index = 0, .frp_weight = 1, @@ -6099,7 +6099,7 @@ fib_test_label (void) FIB_SOURCE_INTERFACE, (FIB_ENTRY_FLAG_CONNECTED | FIB_ENTRY_FLAG_ATTACHED), - FIB_PROTOCOL_IP4, + DPO_PROTO_IP4, NULL, tm->hw[0]->sw_if_index, ~0, @@ -6115,7 +6115,7 @@ fib_test_label (void) FIB_SOURCE_INTERFACE, (FIB_ENTRY_FLAG_CONNECTED | FIB_ENTRY_FLAG_LOCAL), - FIB_PROTOCOL_IP4, + DPO_PROTO_IP4, NULL, tm->hw[0]->sw_if_index, ~0, // invalid fib index @@ -6145,7 +6145,7 @@ fib_test_label (void) FIB_SOURCE_INTERFACE, (FIB_ENTRY_FLAG_CONNECTED | FIB_ENTRY_FLAG_ATTACHED), - FIB_PROTOCOL_IP4, + DPO_PROTO_IP4, NULL, tm->hw[1]->sw_if_index, ~0, @@ -6161,7 +6161,7 @@ fib_test_label (void) FIB_SOURCE_INTERFACE, (FIB_ENTRY_FLAG_CONNECTED | FIB_ENTRY_FLAG_LOCAL), - FIB_PROTOCOL_IP4, + DPO_PROTO_IP4, NULL, tm->hw[1]->sw_if_index, ~0, // invalid fib index @@ -6243,7 +6243,7 @@ fib_test_label (void) &pfx_1_1_1_1_s_32, FIB_SOURCE_API, FIB_ENTRY_FLAG_NONE, - FIB_PROTOCOL_IP4, + DPO_PROTO_IP4, &nh_10_10_10_1, tm->hw[0]->sw_if_index, ~0, // invalid fib index @@ -6282,7 +6282,7 @@ fib_test_label (void) &pfx_1_1_1_1_s_32, FIB_SOURCE_API, FIB_ENTRY_FLAG_NONE, - FIB_PROTOCOL_IP4, + DPO_PROTO_IP4, &nh_10_10_11_1, tm->hw[1]->sw_if_index, ~0, // invalid fib index @@ -6360,7 +6360,7 @@ fib_test_label (void) &pfx_1_1_1_1_s_32, FIB_SOURCE_API, FIB_ENTRY_FLAG_NONE, - FIB_PROTOCOL_IP4, + DPO_PROTO_IP4, &nh_10_10_11_2, tm->hw[1]->sw_if_index, ~0, // invalid fib index @@ -6440,7 +6440,7 @@ fib_test_label (void) &pfx_2_2_2_2_s_32, FIB_SOURCE_API, FIB_ENTRY_FLAG_NONE, - FIB_PROTOCOL_IP4, + DPO_PROTO_IP4, &pfx_1_1_1_1_s_32.fp_addr, ~0, fib_index, @@ -6612,7 +6612,7 @@ fib_test_label (void) fib_table_entry_path_remove(fib_index, &pfx_1_1_1_1_s_32, FIB_SOURCE_API, - FIB_PROTOCOL_IP4, + DPO_PROTO_IP4, &nh_10_10_10_1, tm->hw[0]->sw_if_index, ~0, // invalid fib index @@ -6669,7 +6669,7 @@ fib_test_label (void) fib_table_entry_path_remove(fib_index, &pfx_1_1_1_1_s_32, FIB_SOURCE_API, - FIB_PROTOCOL_IP4, + DPO_PROTO_IP4, &nh_10_10_11_1, tm->hw[1]->sw_if_index, ~0, // invalid fib index @@ -6711,7 +6711,7 @@ fib_test_label (void) &pfx_1_1_1_1_s_32, FIB_SOURCE_API, FIB_ENTRY_FLAG_NONE, - FIB_PROTOCOL_IP4, + DPO_PROTO_IP4, &nh_10_10_10_1, tm->hw[0]->sw_if_index, ~0, // invalid fib index @@ -6842,7 +6842,7 @@ fib_test_label (void) &pfx_1_1_1_2_s_32, FIB_SOURCE_API, FIB_ENTRY_FLAG_NONE, - FIB_PROTOCOL_IP4, + DPO_PROTO_IP4, &nh_10_10_10_1, tm->hw[0]->sw_if_index, ~0, // invalid fib index @@ -6884,7 +6884,7 @@ fib_test_label (void) &pfx_2_2_2_2_s_32, FIB_SOURCE_API, FIB_ENTRY_FLAG_NONE, - FIB_PROTOCOL_IP4, + DPO_PROTO_IP4, &pfx_1_1_1_2_s_32.fp_addr, ~0, fib_index, @@ -6912,7 +6912,7 @@ fib_test_label (void) &pfx_1_1_1_2_s_32, FIB_SOURCE_API, FIB_ENTRY_FLAG_NONE, - FIB_PROTOCOL_IP4, + DPO_PROTO_IP4, &nh_10_10_11_1, tm->hw[1]->sw_if_index, ~0, // invalid fib index @@ -6945,7 +6945,7 @@ fib_test_label (void) &pfx_1_1_1_2_s_32, FIB_SOURCE_API, FIB_ENTRY_FLAG_NONE, - FIB_PROTOCOL_IP4, + DPO_PROTO_IP4, &nh_10_10_11_1, tm->hw[1]->sw_if_index, ~0, // invalid fib index @@ -6987,7 +6987,7 @@ fib_test_label (void) &pfx_2_2_2_3_s_32, FIB_SOURCE_API, FIB_ENTRY_FLAG_NONE, - FIB_PROTOCOL_IP4, + DPO_PROTO_IP4, &pfx_1_1_1_1_s_32.fp_addr, ~0, fib_index, @@ -7031,7 +7031,7 @@ fib_test_label (void) &pfx_2_2_2_4_s_32, FIB_SOURCE_API, FIB_ENTRY_FLAG_NONE, - FIB_PROTOCOL_IP4, + DPO_PROTO_IP4, &pfx_1_1_1_1_s_32.fp_addr, ~0, fib_index, @@ -7081,7 +7081,7 @@ fib_test_label (void) &pfx_2_2_5_5_s_32, FIB_SOURCE_API, FIB_ENTRY_FLAG_NONE, - FIB_PROTOCOL_IP4, + DPO_PROTO_IP4, &nh_10_10_11_1, tm->hw[1]->sw_if_index, ~0, // invalid fib index @@ -7689,7 +7689,7 @@ fib_test_bfd (void) FIB_SOURCE_INTERFACE, (FIB_ENTRY_FLAG_CONNECTED | FIB_ENTRY_FLAG_ATTACHED), - FIB_PROTOCOL_IP4, + DPO_PROTO_IP4, NULL, tm->hw[0]->sw_if_index, ~0, // invalid fib index @@ -7706,7 +7706,7 @@ fib_test_bfd (void) FIB_SOURCE_INTERFACE, (FIB_ENTRY_FLAG_CONNECTED | FIB_ENTRY_FLAG_LOCAL), - FIB_PROTOCOL_IP4, + DPO_PROTO_IP4, NULL, tm->hw[0]->sw_if_index, ~0, // invalid fib index @@ -7780,7 +7780,7 @@ fib_test_bfd (void) &pfx_10_10_10_1_s_32, FIB_SOURCE_ADJ, FIB_ENTRY_FLAG_ATTACHED, - FIB_PROTOCOL_IP4, + DPO_PROTO_IP4, &nh_10_10_10_1, tm->hw[0]->sw_if_index, ~0, // invalid fib index @@ -7819,7 +7819,7 @@ fib_test_bfd (void) &pfx_10_10_10_2_s_32, FIB_SOURCE_ADJ, FIB_ENTRY_FLAG_ATTACHED, - FIB_PROTOCOL_IP4, + DPO_PROTO_IP4, &nh_10_10_10_2, tm->hw[0]->sw_if_index, ~0, // invalid fib index @@ -7851,7 +7851,7 @@ fib_test_bfd (void) &pfx_10_10_10_2_s_32, FIB_SOURCE_ADJ, FIB_ENTRY_FLAG_ATTACHED, - FIB_PROTOCOL_IP4, + DPO_PROTO_IP4, &nh_10_10_10_2, tm->hw[0]->sw_if_index, ~0, // invalid fib index @@ -7907,7 +7907,7 @@ fib_test_bfd (void) &pfx_200_0_0_0_s_24, FIB_SOURCE_API, FIB_ENTRY_FLAG_NONE, - FIB_PROTOCOL_IP4, + DPO_PROTO_IP4, &nh_10_10_10_2, ~0, // recursive 0, // default fib index @@ -7926,7 +7926,7 @@ fib_test_bfd (void) &pfx_200_0_0_0_s_24, FIB_SOURCE_API, FIB_ENTRY_FLAG_NONE, - FIB_PROTOCOL_IP4, + DPO_PROTO_IP4, &nh_10_10_10_1, ~0, // recursive 0, // default fib index @@ -8065,7 +8065,7 @@ fib_test_bfd (void) &pfx_5_5_5_5_s_32, FIB_SOURCE_CLI, FIB_ENTRY_FLAG_NONE, - FIB_PROTOCOL_IP4, + DPO_PROTO_IP4, &nh_10_10_10_1, tm->hw[0]->sw_if_index, ~0, // invalid fib index @@ -8096,7 +8096,7 @@ fib_test_bfd (void) &pfx_5_5_5_5_s_32, FIB_SOURCE_CLI, FIB_ENTRY_FLAG_NONE, - FIB_PROTOCOL_IP4, + DPO_PROTO_IP4, &nh_10_10_10_2, tm->hw[0]->sw_if_index, ~0, // invalid fib index @@ -8234,7 +8234,7 @@ lfib_test (void) &pfx, FIB_SOURCE_CLI, FIB_ENTRY_FLAG_NONE, - FIB_PROTOCOL_IP4, + DPO_PROTO_IP4, &zero_addr, ~0, fib_index, @@ -8285,7 +8285,7 @@ lfib_test (void) &pfx, FIB_SOURCE_CLI, FIB_ENTRY_FLAG_NONE, - FIB_PROTOCOL_IP4, + DPO_PROTO_IP4, &zero_addr, ~0, lfib_index, @@ -8363,7 +8363,7 @@ lfib_test (void) &pfx_1200, FIB_SOURCE_API, FIB_ENTRY_FLAG_NONE, - FIB_PROTOCOL_IP4, + DPO_PROTO_IP4, &nh_10_10_10_1, tm->hw[0]->sw_if_index, ~0, // invalid fib index @@ -8389,7 +8389,7 @@ lfib_test (void) }, }; fib_route_path_t *rpaths = NULL, rpath = { - .frp_proto = FIB_PROTOCOL_MPLS, + .frp_proto = DPO_PROTO_MPLS, .frp_local_label = 1200, .frp_eos = MPLS_NON_EOS, .frp_sw_if_index = ~0, // recurive @@ -8545,7 +8545,7 @@ lfib_test (void) &pfx_2500, FIB_SOURCE_API, FIB_ENTRY_FLAG_NONE, - FIB_PROTOCOL_IP4, + DPO_PROTO_IP4, NULL, tm->hw[0]->sw_if_index, ~0, // invalid fib index @@ -8590,7 +8590,7 @@ lfib_test (void) &pfx_3500, FIB_SOURCE_API, FIB_ENTRY_FLAG_MULTICAST, - FIB_PROTOCOL_IP4, + DPO_PROTO_IP4, &nh_10_10_10_1, tm->hw[0]->sw_if_index, ~0, // invalid fib index @@ -8610,7 +8610,7 @@ lfib_test (void) &pfx_3500, FIB_SOURCE_API, FIB_ENTRY_FLAG_MULTICAST, - FIB_PROTOCOL_IP4, + DPO_PROTO_IP4, NULL, tm->hw[0]->sw_if_index, ~0, // invalid fib index @@ -8637,7 +8637,7 @@ lfib_test (void) &pfx_3500, FIB_SOURCE_API, FIB_ENTRY_FLAG_MULTICAST, - FIB_PROTOCOL_IP4, + DPO_PROTO_IP4, NULL, 5, // rpf-id 0, // default table diff --git a/src/vnet/fib/fib_types.h b/src/vnet/fib/fib_types.h index a209ff3c..f11a55da 100644 --- a/src/vnet/fib/fib_types.h +++ b/src/vnet/fib/fib_types.h @@ -32,9 +32,9 @@ typedef u32 fib_node_index_t; * Protocol Type. packed so it consumes a u8 only */ typedef enum fib_protocol_t_ { - FIB_PROTOCOL_IP4 = 0, - FIB_PROTOCOL_IP6, - FIB_PROTOCOL_MPLS, + FIB_PROTOCOL_IP4 = DPO_PROTO_IP4, + FIB_PROTOCOL_IP6 = DPO_PROTO_IP6, + FIB_PROTOCOL_MPLS = DPO_PROTO_MPLS, } __attribute__ ((packed)) fib_protocol_t; #define FIB_PROTOCOLS { \ @@ -338,7 +338,7 @@ typedef struct fib_route_path_t_ { * The protocol of the address below. We need this since the all * zeros address is ambiguous. */ - fib_protocol_t frp_proto; + dpo_proto_t frp_proto; union { /** diff --git a/src/vnet/interface_format.c b/src/vnet/interface_format.c index df7e9388..5694bb2f 100644 --- a/src/vnet/interface_format.c +++ b/src/vnet/interface_format.c @@ -165,9 +165,15 @@ format_vnet_sw_if_index_name (u8 * s, va_list * args) { vnet_main_t *vnm = va_arg (*args, vnet_main_t *); u32 sw_if_index = va_arg (*args, u32); - return format (s, "%U", - format_vnet_sw_interface_name, vnm, - vnet_get_sw_interface (vnm, sw_if_index)); + vnet_sw_interface_t *si; + + si = vnet_get_sw_interface_safe (vnm, sw_if_index); + + if (NULL == si) + { + return format (s, "DELETED"); + } + return format (s, "%U", format_vnet_sw_interface_name, vnm, si); } u8 * diff --git a/src/vnet/ip/ip4_forward.c b/src/vnet/ip/ip4_forward.c index ee17ea88..7a8d7a0c 100755 --- a/src/vnet/ip/ip4_forward.c +++ b/src/vnet/ip/ip4_forward.c @@ -724,7 +724,7 @@ ip4_add_interface_routes (u32 sw_if_index, FIB_SOURCE_INTERFACE, (FIB_ENTRY_FLAG_CONNECTED | FIB_ENTRY_FLAG_ATTACHED), - FIB_PROTOCOL_IP4, + DPO_PROTO_IP4, /* No next-hop address */ NULL, sw_if_index, @@ -767,7 +767,7 @@ ip4_add_interface_routes (u32 sw_if_index, fib_table_entry_update_one_path (fib_index, &net_pfx, FIB_SOURCE_INTERFACE, (FIB_ENTRY_FLAG_ATTACHED), - FIB_PROTOCOL_IP4, + DPO_PROTO_IP4, &net_pfx.fp_addr, sw_if_index, // invalid FIB index @@ -803,7 +803,7 @@ ip4_add_interface_routes (u32 sw_if_index, FIB_SOURCE_INTERFACE, (FIB_ENTRY_FLAG_CONNECTED | FIB_ENTRY_FLAG_LOCAL), - FIB_PROTOCOL_IP4, + DPO_PROTO_IP4, &pfx.fp_addr, sw_if_index, // invalid FIB index diff --git a/src/vnet/ip/ip6_forward.c b/src/vnet/ip/ip6_forward.c index bc66416e..8ae08a01 100644 --- a/src/vnet/ip/ip6_forward.c +++ b/src/vnet/ip/ip6_forward.c @@ -355,7 +355,7 @@ ip6_add_interface_routes (vnet_main_t * vnm, u32 sw_if_index, FIB_SOURCE_INTERFACE, (FIB_ENTRY_FLAG_CONNECTED | FIB_ENTRY_FLAG_ATTACHED), - FIB_PROTOCOL_IP6, + DPO_PROTO_IP6, /* No next-hop address */ NULL, sw_if_index, /* invalid FIB index */ @@ -390,7 +390,7 @@ ip6_add_interface_routes (vnet_main_t * vnm, u32 sw_if_index, FIB_SOURCE_INTERFACE, (FIB_ENTRY_FLAG_CONNECTED | FIB_ENTRY_FLAG_LOCAL), - FIB_PROTOCOL_IP6, + DPO_PROTO_IP6, &pfx.fp_addr, sw_if_index, ~0, 1, NULL, FIB_ROUTE_PATH_FLAG_NONE); diff --git a/src/vnet/ip/ip6_neighbor.c b/src/vnet/ip/ip6_neighbor.c index e8eebd4e..6a9139ab 100644 --- a/src/vnet/ip/ip6_neighbor.c +++ b/src/vnet/ip/ip6_neighbor.c @@ -284,7 +284,7 @@ ip6_neighbor_sw_interface_up_down (vnet_main_t * vnm, (ip6_fib_table_get_index_for_sw_if_index (n->key.sw_if_index), &pfx, FIB_SOURCE_ADJ, - FIB_PROTOCOL_IP6, + DPO_PROTO_IP6, &pfx.fp_addr, n->key.sw_if_index, ~0, 1, FIB_ROUTE_PATH_FLAG_NONE); pool_put (nm->neighbor_pool, n); @@ -645,7 +645,7 @@ vnet_set_ip6_ethernet_neighbor (vlib_main_t * vm, n->fib_entry_index = fib_table_entry_path_add (fib_index, &pfx, FIB_SOURCE_ADJ, FIB_ENTRY_FLAG_ATTACHED, - FIB_PROTOCOL_IP6, &pfx.fp_addr, + DPO_PROTO_IP6, &pfx.fp_addr, n->key.sw_if_index, ~0, 1, NULL, FIB_ROUTE_PATH_FLAG_NONE); } @@ -776,7 +776,7 @@ vnet_unset_ip6_ethernet_neighbor (vlib_main_t * vm, (ip6_fib_table_get_index_for_sw_if_index (n->key.sw_if_index), &pfx, FIB_SOURCE_ADJ, - FIB_PROTOCOL_IP6, + DPO_PROTO_IP6, &pfx.fp_addr, n->key.sw_if_index, ~0, 1, FIB_ROUTE_PATH_FLAG_NONE); } pool_put (nm->neighbor_pool, n); @@ -4110,7 +4110,7 @@ ip6_neighbor_proxy_add_del (u32 sw_if_index, ip6_address_t * addr, u8 is_del) fib_table_entry_path_remove (fib_index, &pfx, FIB_SOURCE_IP6_ND_PROXY, - FIB_PROTOCOL_IP6, + DPO_PROTO_IP6, &nh, sw_if_index, ~0, 1, FIB_ROUTE_PATH_FLAG_NONE); @@ -4124,7 +4124,7 @@ ip6_neighbor_proxy_add_del (u32 sw_if_index, ip6_address_t * addr, u8 is_del) &pfx, FIB_SOURCE_IP6_ND_PROXY, FIB_ENTRY_FLAG_NONE, - FIB_PROTOCOL_IP6, + DPO_PROTO_IP6, &nh, sw_if_index, ~0, 1, NULL, FIB_ROUTE_PATH_FLAG_NONE); diff --git a/src/vnet/ip/ip_api.c b/src/vnet/ip/ip_api.c index 4cbf75a3..0676a387 100644 --- a/src/vnet/ip/ip_api.c +++ b/src/vnet/ip/ip_api.c @@ -156,9 +156,9 @@ copy_fib_next_hop (fib_route_path_encode_t * api_rpath, void *fp_arg) int is_ip4; vl_api_fib_path_t *fp = (vl_api_fib_path_t *) fp_arg; - if (api_rpath->rpath.frp_proto == FIB_PROTOCOL_IP4) + if (api_rpath->rpath.frp_proto == DPO_PROTO_IP4) fp->afi = IP46_TYPE_IP4; - else if (api_rpath->rpath.frp_proto == FIB_PROTOCOL_IP6) + else if (api_rpath->rpath.frp_proto == DPO_PROTO_IP6) fp->afi = IP46_TYPE_IP6; else { @@ -714,7 +714,7 @@ add_del_route_t_handler (u8 is_multipath, u8 is_rpf_id, u32 fib_index, const fib_prefix_t * prefix, - u8 next_hop_proto_is_ip4, + dpo_proto_t next_hop_proto, const ip46_address_t * next_hop, u32 next_hop_sw_if_index, u8 next_hop_fib_index, @@ -726,8 +726,7 @@ add_del_route_t_handler (u8 is_multipath, vnet_classify_main_t *cm = &vnet_classify_main; fib_route_path_flags_t path_flags = FIB_ROUTE_PATH_FLAG_NONE; fib_route_path_t path = { - .frp_proto = (next_hop_proto_is_ip4 ? - FIB_PROTOCOL_IP4 : FIB_PROTOCOL_IP6), + .frp_proto = next_hop_proto, .frp_addr = (NULL == next_hop ? zero_addr : *next_hop), .frp_sw_if_index = next_hop_sw_if_index, .frp_fib_index = next_hop_fib_index, @@ -740,7 +739,7 @@ add_del_route_t_handler (u8 is_multipath, if (MPLS_LABEL_INVALID != next_hop_via_label) { - path.frp_proto = FIB_PROTOCOL_MPLS; + path.frp_proto = DPO_PROTO_MPLS; path.frp_local_label = next_hop_via_label; path.frp_eos = MPLS_NON_EOS; } @@ -855,7 +854,7 @@ int add_del_route_check (fib_protocol_t table_proto, u32 table_id, u32 next_hop_sw_if_index, - fib_protocol_t next_hop_table_proto, + dpo_proto_t next_hop_table_proto, u32 next_hop_table_id, u8 create_missing_tables, u8 is_rpf_id, u32 * fib_index, u32 * next_hop_fib_index) @@ -887,11 +886,18 @@ add_del_route_check (fib_protocol_t table_proto, } else { + fib_protocol_t fib_nh_proto; + + if (next_hop_table_proto > DPO_PROTO_MPLS) + return (0); + + fib_nh_proto = dpo_proto_to_fib (next_hop_table_proto); + if (is_rpf_id) - *next_hop_fib_index = mfib_table_find (next_hop_table_proto, + *next_hop_fib_index = mfib_table_find (fib_nh_proto, ntohl (next_hop_table_id)); else - *next_hop_fib_index = fib_table_find (next_hop_table_proto, + *next_hop_fib_index = fib_table_find (fib_nh_proto, ntohl (next_hop_table_id)); if (~0 == *next_hop_fib_index) @@ -900,12 +906,12 @@ add_del_route_check (fib_protocol_t table_proto, { if (is_rpf_id) *next_hop_fib_index = - mfib_table_find_or_create_and_lock (next_hop_table_proto, + mfib_table_find_or_create_and_lock (fib_nh_proto, ntohl (next_hop_table_id)); else *next_hop_fib_index = - fib_table_find_or_create_and_lock (next_hop_table_proto, + fib_table_find_or_create_and_lock (fib_nh_proto, ntohl (next_hop_table_id)); } @@ -930,7 +936,7 @@ ip4_add_del_route_t_handler (vl_api_ip_add_del_route_t * mp) rv = add_del_route_check (FIB_PROTOCOL_IP4, mp->table_id, mp->next_hop_sw_if_index, - FIB_PROTOCOL_IP4, + DPO_PROTO_IP4, mp->next_hop_table_id, mp->create_vrf_if_needed, 0, &fib_index, &next_hop_fib_index); @@ -970,7 +976,7 @@ ip4_add_del_route_t_handler (vl_api_ip_add_del_route_t * mp) mp->classify_table_index, mp->is_resolve_host, mp->is_resolve_attached, 0, 0, - fib_index, &pfx, 1, + fib_index, &pfx, DPO_PROTO_IP4, &nh, ntohl (mp->next_hop_sw_if_index), next_hop_fib_index, @@ -990,7 +996,7 @@ ip6_add_del_route_t_handler (vl_api_ip_add_del_route_t * mp) rv = add_del_route_check (FIB_PROTOCOL_IP6, mp->table_id, mp->next_hop_sw_if_index, - FIB_PROTOCOL_IP6, + DPO_PROTO_IP6, mp->next_hop_table_id, mp->create_vrf_if_needed, 0, &fib_index, &next_hop_fib_index); @@ -1030,7 +1036,7 @@ ip6_add_del_route_t_handler (vl_api_ip_add_del_route_t * mp) mp->classify_table_index, mp->is_resolve_host, mp->is_resolve_attached, 0, 0, - fib_index, &pfx, 0, + fib_index, &pfx, DPO_PROTO_IP6, &nh, ntohl (mp->next_hop_sw_if_index), next_hop_fib_index, mp->next_hop_weight, @@ -1106,7 +1112,7 @@ mroute_add_del_handler (u8 is_add, fib_route_path_t path = { .frp_sw_if_index = next_hop_sw_if_index, - .frp_proto = prefix->fp_proto, + .frp_proto = fib_proto_to_dpo (prefix->fp_proto), }; if (is_local) diff --git a/src/vnet/ip/lookup.c b/src/vnet/ip/lookup.c index 533d010a..41e46070 100755 --- a/src/vnet/ip/lookup.c +++ b/src/vnet/ip/lookup.c @@ -423,7 +423,7 @@ vnet_ip_route_cmd (vlib_main_t * vm, { rpath.frp_weight = 1; rpath.frp_eos = MPLS_NON_EOS; - rpath.frp_proto = FIB_PROTOCOL_MPLS; + rpath.frp_proto = DPO_PROTO_MPLS; rpath.frp_sw_if_index = ~0; vec_add1 (rpaths, rpath); } @@ -449,7 +449,7 @@ vnet_ip_route_cmd (vlib_main_t * vm, &rpath.frp_sw_if_index)) { rpath.frp_weight = 1; - rpath.frp_proto = FIB_PROTOCOL_IP4; + rpath.frp_proto = DPO_PROTO_IP4; vec_add1 (rpaths, rpath); } @@ -460,7 +460,7 @@ vnet_ip_route_cmd (vlib_main_t * vm, &rpath.frp_sw_if_index)) { rpath.frp_weight = 1; - rpath.frp_proto = FIB_PROTOCOL_IP6; + rpath.frp_proto = DPO_PROTO_IP6; vec_add1 (rpaths, rpath); } else if (unformat (line_input, "weight %u", &weight)) @@ -479,7 +479,7 @@ vnet_ip_route_cmd (vlib_main_t * vm, { rpath.frp_weight = 1; rpath.frp_sw_if_index = ~0; - rpath.frp_proto = FIB_PROTOCOL_IP4; + rpath.frp_proto = DPO_PROTO_IP4; vec_add1 (rpaths, rpath); } else if (unformat (line_input, "via %U next-hop-table %d", @@ -488,7 +488,7 @@ vnet_ip_route_cmd (vlib_main_t * vm, { rpath.frp_weight = 1; rpath.frp_sw_if_index = ~0; - rpath.frp_proto = FIB_PROTOCOL_IP6; + rpath.frp_proto = DPO_PROTO_IP6; vec_add1 (rpaths, rpath); } else if (unformat (line_input, "via %U", @@ -501,7 +501,7 @@ vnet_ip_route_cmd (vlib_main_t * vm, rpath.frp_fib_index = table_id; rpath.frp_weight = 1; rpath.frp_sw_if_index = ~0; - rpath.frp_proto = FIB_PROTOCOL_IP4; + rpath.frp_proto = DPO_PROTO_IP4; vec_add1 (rpaths, rpath); } else if (unformat (line_input, "via %U", @@ -510,13 +510,13 @@ vnet_ip_route_cmd (vlib_main_t * vm, rpath.frp_fib_index = table_id; rpath.frp_weight = 1; rpath.frp_sw_if_index = ~0; - rpath.frp_proto = FIB_PROTOCOL_IP6; + rpath.frp_proto = DPO_PROTO_IP6; vec_add1 (rpaths, rpath); } else if (unformat (line_input, "lookup in table %d", &rpath.frp_fib_index)) { - rpath.frp_proto = pfx.fp_proto; + rpath.frp_proto = fib_proto_to_dpo (pfx.fp_proto); rpath.frp_sw_if_index = ~0; vec_add1 (rpaths, rpath); } @@ -526,7 +526,7 @@ vnet_ip_route_cmd (vlib_main_t * vm, &rpath.frp_sw_if_index)) { rpath.frp_weight = 1; - rpath.frp_proto = prefixs[0].fp_proto; + rpath.frp_proto = fib_proto_to_dpo (prefixs[0].fp_proto); vec_add1 (rpaths, rpath); } else if (vec_len (prefixs) > 0 && diff --git a/src/vnet/lisp-gpe/lisp_gpe.c b/src/vnet/lisp-gpe/lisp_gpe.c index 0acc7349..018895ad 100644 --- a/src/vnet/lisp-gpe/lisp_gpe.c +++ b/src/vnet/lisp-gpe/lisp_gpe.c @@ -454,7 +454,7 @@ vnet_gpe_add_del_native_fwd_rpath (vnet_gpe_native_fwd_rpath_args_t * a) fib_route_path_t *rpath; u8 ip_version; - ip_version = a->rpath.frp_proto == FIB_PROTOCOL_IP4 ? IP4 : IP6; + ip_version = a->rpath.frp_proto == DPO_PROTO_IP4 ? IP4 : IP6; if (a->is_add) { @@ -511,7 +511,7 @@ gpe_native_forward_command_fn (vlib_main_t * vm, unformat_input_t * input, &rpath.frp_sw_if_index)) { rpath.frp_weight = 1; - rpath.frp_proto = FIB_PROTOCOL_IP4; + rpath.frp_proto = DPO_PROTO_IP4; } else if (unformat (line_input, "via %U %U", unformat_ip6_address, @@ -520,21 +520,21 @@ gpe_native_forward_command_fn (vlib_main_t * vm, unformat_input_t * input, &rpath.frp_sw_if_index)) { rpath.frp_weight = 1; - rpath.frp_proto = FIB_PROTOCOL_IP6; + rpath.frp_proto = DPO_PROTO_IP6; } else if (unformat (line_input, "via %U", unformat_ip4_address, &rpath.frp_addr.ip4)) { rpath.frp_weight = 1; rpath.frp_sw_if_index = ~0; - rpath.frp_proto = FIB_PROTOCOL_IP4; + rpath.frp_proto = DPO_PROTO_IP4; } else if (unformat (line_input, "via %U", unformat_ip6_address, &rpath.frp_addr.ip6)) { rpath.frp_weight = 1; rpath.frp_sw_if_index = ~0; - rpath.frp_proto = FIB_PROTOCOL_IP6; + rpath.frp_proto = DPO_PROTO_IP6; } else { @@ -549,7 +549,8 @@ gpe_native_forward_command_fn (vlib_main_t * vm, unformat_input_t * input, } else { - rpath.frp_fib_index = fib_table_find (rpath.frp_proto, table_id); + rpath.frp_fib_index = + fib_table_find (dpo_proto_to_fib (rpath.frp_proto), table_id); if ((u32) ~ 0 == rpath.frp_fib_index) { error = clib_error_return (0, "Nonexistent table id %d", table_id); diff --git a/src/vnet/lisp-gpe/lisp_gpe_api.c b/src/vnet/lisp-gpe/lisp_gpe_api.c index f1663699..4367a719 100644 --- a/src/vnet/lisp-gpe/lisp_gpe_api.c +++ b/src/vnet/lisp-gpe/lisp_gpe_api.c @@ -455,10 +455,10 @@ static void clib_memcpy (&a->rpath.frp_addr.ip6, mp->nh_addr, sizeof (ip6_address_t)); a->is_add = mp->is_add; - a->rpath.frp_proto = mp->is_ip4 ? FIB_PROTOCOL_IP4 : FIB_PROTOCOL_IP6; - a->rpath.frp_fib_index = fib_table_find (a->rpath.frp_proto, - clib_net_to_host_u32 - (mp->table_id)); + a->rpath.frp_proto = mp->is_ip4 ? DPO_PROTO_IP4 : DPO_PROTO_IP6; + a->rpath.frp_fib_index = + fib_table_find (dpo_proto_to_fib (a->rpath.frp_proto), + clib_net_to_host_u32 (mp->table_id)); if (~0 == a->rpath.frp_fib_index) { rv = VNET_API_ERROR_INVALID_VALUE; @@ -484,7 +484,7 @@ gpe_native_fwd_rpaths_copy (vl_api_gpe_native_fwd_rpath_t * dst, vec_foreach (e, src) { memset (&dst[i], 0, sizeof (*dst)); - table = fib_table_get (e->frp_fib_index, e->frp_proto); + table = fib_table_get (e->frp_fib_index, dpo_proto_to_fib (e->frp_proto)); dst[i].fib_index = table->ft_table_id; dst[i].nh_sw_if_index = e->frp_sw_if_index; dst[i].is_ip4 = is_ip4; diff --git a/src/vnet/lisp-gpe/lisp_gpe_fwd_entry.c b/src/vnet/lisp-gpe/lisp_gpe_fwd_entry.c index 395b493a..ac048149 100644 --- a/src/vnet/lisp-gpe/lisp_gpe_fwd_entry.c +++ b/src/vnet/lisp-gpe/lisp_gpe_fwd_entry.c @@ -225,6 +225,7 @@ lisp_gpe_mk_fib_paths (const lisp_fwd_path_t * paths) { const lisp_gpe_adjacency_t *ladj; fib_route_path_t *rpaths = NULL; + fib_protocol_t fp; u8 best_priority; u32 ii; @@ -239,9 +240,9 @@ lisp_gpe_mk_fib_paths (const lisp_fwd_path_t * paths) ladj = lisp_gpe_adjacency_get (paths[ii].lisp_adj); - ip_address_to_46 (&ladj->remote_rloc, - &rpaths[ii].frp_addr, &rpaths[ii].frp_proto); + ip_address_to_46 (&ladj->remote_rloc, &rpaths[ii].frp_addr, &fp); + rpaths[ii].frp_proto = fib_proto_to_dpo (fp); rpaths[ii].frp_sw_if_index = ladj->sw_if_index; rpaths[ii].frp_weight = (paths[ii].weight ? paths[ii].weight : 1); } diff --git a/src/vnet/mfib/ip6_mfib.c b/src/vnet/mfib/ip6_mfib.c index 5c6f8126..5e48e919 100644 --- a/src/vnet/mfib/ip6_mfib.c +++ b/src/vnet/mfib/ip6_mfib.c @@ -158,7 +158,7 @@ ip6_create_mfib_with_table_id (u32 table_id) .fp_proto = FIB_PROTOCOL_IP6, }; const fib_route_path_t path_for_us = { - .frp_proto = FIB_PROTOCOL_IP6, + .frp_proto = DPO_PROTO_IP6, .frp_addr = zero_addr, .frp_sw_if_index = 0xffffffff, .frp_fib_index = ~0, @@ -222,7 +222,7 @@ ip6_mfib_table_destroy (ip6_mfib_t *mfib) .fp_proto = FIB_PROTOCOL_IP6, }; const fib_route_path_t path_for_us = { - .frp_proto = FIB_PROTOCOL_IP6, + .frp_proto = DPO_PROTO_IP6, .frp_addr = zero_addr, .frp_sw_if_index = 0xffffffff, .frp_fib_index = ~0, @@ -259,7 +259,7 @@ void ip6_mfib_interface_enable_disable (u32 sw_if_index, int is_enable) { const fib_route_path_t path = { - .frp_proto = FIB_PROTOCOL_IP6, + .frp_proto = DPO_PROTO_IP6, .frp_addr = zero_addr, .frp_sw_if_index = sw_if_index, .frp_fib_index = ~0, diff --git a/src/vnet/mfib/mfib_entry.c b/src/vnet/mfib/mfib_entry.c index cf25b67a..b37f8825 100644 --- a/src/vnet/mfib/mfib_entry.c +++ b/src/vnet/mfib/mfib_entry.c @@ -764,18 +764,16 @@ mfib_entry_update (fib_node_index_t mfib_entry_index, * entry */ fib_node_index_t old_pl_index; - fib_protocol_t fp; + dpo_proto_t dp; dpo_id_t dpo = DPO_INVALID; - fp = mfib_entry_get_proto(mfib_entry); + dp = fib_proto_to_dpo(mfib_entry_get_proto(mfib_entry)); old_pl_index = msrc->mfes_pl; - dpo_set(&dpo, DPO_REPLICATE, - fib_proto_to_dpo(fp), - repi); + dpo_set(&dpo, DPO_REPLICATE, dp, repi); msrc->mfes_pl = - fib_path_list_create_special(fp, + fib_path_list_create_special(dp, FIB_PATH_LIST_FLAG_EXCLUSIVE, &dpo); diff --git a/src/vnet/mfib/mfib_test.c b/src/vnet/mfib/mfib_test.c index 7c92ae99..57787eca 100644 --- a/src/vnet/mfib/mfib_test.c +++ b/src/vnet/mfib/mfib_test.c @@ -387,7 +387,7 @@ mfib_test_i (fib_protocol_t PROTO, fib_route_path_t path_via_if0 = { - .frp_proto = PROTO, + .frp_proto = fib_proto_to_dpo(PROTO), .frp_addr = zero_addr, .frp_sw_if_index = tm->hw[0]->sw_if_index, .frp_fib_index = ~0, @@ -411,7 +411,7 @@ mfib_test_i (fib_protocol_t PROTO, MFIB_ITF_FLAG_ACCEPT)); fib_route_path_t path_via_if1 = { - .frp_proto = PROTO, + .frp_proto = fib_proto_to_dpo(PROTO), .frp_addr = zero_addr, .frp_sw_if_index = tm->hw[1]->sw_if_index, .frp_fib_index = ~0, @@ -419,7 +419,7 @@ mfib_test_i (fib_protocol_t PROTO, .frp_flags = 0, }; fib_route_path_t path_via_if2 = { - .frp_proto = PROTO, + .frp_proto = fib_proto_to_dpo(PROTO), .frp_addr = zero_addr, .frp_sw_if_index = tm->hw[2]->sw_if_index, .frp_fib_index = ~0, @@ -427,7 +427,7 @@ mfib_test_i (fib_protocol_t PROTO, .frp_flags = 0, }; fib_route_path_t path_via_if3 = { - .frp_proto = PROTO, + .frp_proto = fib_proto_to_dpo(PROTO), .frp_addr = zero_addr, .frp_sw_if_index = tm->hw[3]->sw_if_index, .frp_fib_index = ~0, @@ -435,7 +435,7 @@ mfib_test_i (fib_protocol_t PROTO, .frp_flags = 0, }; fib_route_path_t path_for_us = { - .frp_proto = PROTO, + .frp_proto = fib_proto_to_dpo(PROTO), .frp_addr = zero_addr, .frp_sw_if_index = 0xffffffff, .frp_fib_index = ~0, @@ -1121,7 +1121,7 @@ mfib_test_i (fib_protocol_t PROTO, &pfx_3500, FIB_SOURCE_API, FIB_ENTRY_FLAG_MULTICAST, - FIB_PROTOCOL_IP4, + DPO_PROTO_IP4, &nh_10_10_10_1, tm->hw[0]->sw_if_index, ~0, // invalid fib index @@ -1138,7 +1138,7 @@ mfib_test_i (fib_protocol_t PROTO, * An (S,G) that resolves via the mLDP head-end */ fib_route_path_t path_via_mldp = { - .frp_proto = FIB_PROTOCOL_MPLS, + .frp_proto = DPO_PROTO_MPLS, .frp_local_label = pfx_3500.fp_label, .frp_eos = MPLS_EOS, .frp_sw_if_index = 0xffffffff, diff --git a/src/vnet/mpls/mpls.api b/src/vnet/mpls/mpls.api index 67f1045d..5973a0a6 100644 --- a/src/vnet/mpls/mpls.api +++ b/src/vnet/mpls/mpls.api @@ -156,7 +156,7 @@ manual_endian manual_print define mpls_tunnel_details @param mr_is_interface_rx - Interface Receive path @param mr_is_interface_rx - RPF-ID Receive path. The next-hop interface is used as the RPF-ID - @param mr_next_hop_proto_is_ip4 - The next-hop is IPV4 + @param mr_next_hop_proto - The next-hop protocol, of type dpo_proto_t @param mr_next_hop_weight - The weight, for UCMP @param mr_next_hop[16] - the nextop address @param mr_next_hop_sw_if_index - the next-hop SW interface @@ -182,7 +182,7 @@ autoreply define mpls_route_add_del u8 mr_is_resolve_attached; u8 mr_is_interface_rx; u8 mr_is_rpf_id; - u8 mr_next_hop_proto_is_ip4; + u8 mr_next_hop_proto; u8 mr_next_hop_weight; u8 mr_next_hop_preference; u8 mr_next_hop[16]; diff --git a/src/vnet/mpls/mpls.c b/src/vnet/mpls/mpls.c index 068d31f4..266ba42c 100644 --- a/src/vnet/mpls/mpls.c +++ b/src/vnet/mpls/mpls.c @@ -261,7 +261,7 @@ vnet_mpls_local_label (vlib_main_t * vm, &rpath.frp_sw_if_index, &rpath.frp_weight)) { - rpath.frp_proto = FIB_PROTOCOL_IP4; + rpath.frp_proto = DPO_PROTO_IP4; vec_add1(rpaths, rpath); } @@ -272,7 +272,7 @@ vnet_mpls_local_label (vlib_main_t * vm, &rpath.frp_sw_if_index, &rpath.frp_weight)) { - rpath.frp_proto = FIB_PROTOCOL_IP6; + rpath.frp_proto = DPO_PROTO_IP6; vec_add1(rpaths, rpath); } @@ -283,7 +283,7 @@ vnet_mpls_local_label (vlib_main_t * vm, &rpath.frp_sw_if_index)) { rpath.frp_weight = 1; - rpath.frp_proto = FIB_PROTOCOL_IP4; + rpath.frp_proto = DPO_PROTO_IP4; vec_add1(rpaths, rpath); } else if (unformat (line_input, "rx-ip4 %U", @@ -291,7 +291,7 @@ vnet_mpls_local_label (vlib_main_t * vm, &rpath.frp_sw_if_index)) { rpath.frp_weight = 1; - rpath.frp_proto = FIB_PROTOCOL_IP4; + rpath.frp_proto = DPO_PROTO_IP4; rpath.frp_flags = FIB_ROUTE_PATH_INTF_RX; vec_add1(rpaths, rpath); } @@ -302,7 +302,7 @@ vnet_mpls_local_label (vlib_main_t * vm, &rpath.frp_sw_if_index)) { rpath.frp_weight = 1; - rpath.frp_proto = FIB_PROTOCOL_IP6; + rpath.frp_proto = DPO_PROTO_IP6; vec_add1(rpaths, rpath); } else if (unformat (line_input, "via %U next-hop-table %d", @@ -312,7 +312,7 @@ vnet_mpls_local_label (vlib_main_t * vm, { rpath.frp_weight = 1; rpath.frp_sw_if_index = ~0; - rpath.frp_proto = FIB_PROTOCOL_IP4; + rpath.frp_proto = DPO_PROTO_IP4; vec_add1(rpaths, rpath); } else if (unformat (line_input, "via %U next-hop-table %d", @@ -322,7 +322,7 @@ vnet_mpls_local_label (vlib_main_t * vm, { rpath.frp_weight = 1; rpath.frp_sw_if_index = ~0; - rpath.frp_proto = FIB_PROTOCOL_IP6; + rpath.frp_proto = DPO_PROTO_IP6; vec_add1(rpaths, rpath); } else if (unformat (line_input, "via %U", @@ -336,7 +336,7 @@ vnet_mpls_local_label (vlib_main_t * vm, rpath.frp_fib_index = table_id; rpath.frp_weight = 1; rpath.frp_sw_if_index = ~0; - rpath.frp_proto = FIB_PROTOCOL_IP4; + rpath.frp_proto = DPO_PROTO_IP4; vec_add1(rpaths, rpath); } else if (unformat (line_input, "via %U", @@ -346,7 +346,7 @@ vnet_mpls_local_label (vlib_main_t * vm, rpath.frp_fib_index = table_id; rpath.frp_weight = 1; rpath.frp_sw_if_index = ~0; - rpath.frp_proto = FIB_PROTOCOL_IP6; + rpath.frp_proto = DPO_PROTO_IP6; vec_add1(rpaths, rpath); } else if (unformat (line_input, "%d", &local_label)) @@ -355,7 +355,7 @@ vnet_mpls_local_label (vlib_main_t * vm, "ip4-lookup-in-table %d", &rpath.frp_fib_index)) { - rpath.frp_proto = FIB_PROTOCOL_IP4; + rpath.frp_proto = DPO_PROTO_IP4; rpath.frp_sw_if_index = FIB_NODE_INDEX_INVALID; pfx.fp_payload_proto = DPO_PROTO_IP4; vec_add1(rpaths, rpath); @@ -364,7 +364,7 @@ vnet_mpls_local_label (vlib_main_t * vm, "ip6-lookup-in-table %d", &rpath.frp_fib_index)) { - rpath.frp_proto = FIB_PROTOCOL_IP6; + rpath.frp_proto = DPO_PROTO_IP6; rpath.frp_sw_if_index = FIB_NODE_INDEX_INVALID; vec_add1(rpaths, rpath); pfx.fp_payload_proto = DPO_PROTO_IP6; @@ -373,11 +373,21 @@ vnet_mpls_local_label (vlib_main_t * vm, "mpls-lookup-in-table %d", &rpath.frp_fib_index)) { - rpath.frp_proto = FIB_PROTOCOL_MPLS; + rpath.frp_proto = DPO_PROTO_MPLS; rpath.frp_sw_if_index = FIB_NODE_INDEX_INVALID; pfx.fp_payload_proto = DPO_PROTO_MPLS; vec_add1(rpaths, rpath); } + else if (unformat (line_input, + "l2-input-on %U", + unformat_vnet_sw_interface, vnm, + &rpath.frp_sw_if_index)) + { + rpath.frp_proto = DPO_PROTO_ETHERNET; + pfx.fp_payload_proto = DPO_PROTO_ETHERNET; + rpath.frp_flags = FIB_ROUTE_PATH_INTF_RX; + vec_add1(rpaths, rpath); + } else if (unformat (line_input, "out-label %U", unformat_mpls_unicast_label, &out_label)) @@ -440,7 +450,7 @@ vnet_mpls_local_label (vlib_main_t * vm, pfx.fp_proto = FIB_PROTOCOL_MPLS; pfx.fp_len = 21; pfx.fp_label = local_label; - pfx.fp_payload_proto = fib_proto_to_dpo(rpaths[0].frp_proto); + pfx.fp_payload_proto = rpaths[0].frp_proto; /* * the CLI parsing stored table Ids, swap to FIB indicies diff --git a/src/vnet/mpls/mpls_api.c b/src/vnet/mpls/mpls_api.c index 92fb24a6..737299e6 100644 --- a/src/vnet/mpls/mpls_api.c +++ b/src/vnet/mpls/mpls_api.c @@ -144,14 +144,7 @@ mpls_route_add_del_t_handler (vnet_main_t * vnm, }; if (pfx.fp_eos) { - if (mp->mr_next_hop_proto_is_ip4) - { - pfx.fp_payload_proto = DPO_PROTO_IP4; - } - else - { - pfx.fp_payload_proto = DPO_PROTO_IP6; - } + pfx.fp_payload_proto = mp->mr_next_hop_proto; } else { @@ -161,7 +154,7 @@ mpls_route_add_del_t_handler (vnet_main_t * vnm, rv = add_del_route_check (FIB_PROTOCOL_MPLS, mp->mr_table_id, mp->mr_next_hop_sw_if_index, - dpo_proto_to_fib (pfx.fp_payload_proto), + pfx.fp_payload_proto, mp->mr_next_hop_table_id, mp->mr_create_table_if_needed, mp->mr_is_rpf_id, @@ -173,9 +166,9 @@ mpls_route_add_del_t_handler (vnet_main_t * vnm, ip46_address_t nh; memset (&nh, 0, sizeof (nh)); - if (mp->mr_next_hop_proto_is_ip4) + if (DPO_PROTO_IP4 == mp->mr_next_hop_proto) memcpy (&nh.ip4, mp->mr_next_hop, sizeof (nh.ip4)); - else + else if (DPO_PROTO_IP6 == mp->mr_next_hop_proto) memcpy (&nh.ip6, mp->mr_next_hop, sizeof (nh.ip6)); n_labels = mp->mr_next_hop_n_out_labels; @@ -202,7 +195,7 @@ mpls_route_add_del_t_handler (vnet_main_t * vnm, mp->mr_is_interface_rx, mp->mr_is_rpf_id, fib_index, &pfx, - mp->mr_next_hop_proto_is_ip4, + mp->mr_next_hop_proto, &nh, ntohl (mp->mr_next_hop_sw_if_index), next_hop_fib_index, mp->mr_next_hop_weight, @@ -243,13 +236,13 @@ vl_api_mpls_tunnel_add_del_t_handler (vl_api_mpls_tunnel_add_del_t * mp) if (mp->mt_next_hop_proto_is_ip4) { - rpath.frp_proto = FIB_PROTOCOL_IP4; + rpath.frp_proto = DPO_PROTO_IP4; clib_memcpy (&rpath.frp_addr.ip4, mp->mt_next_hop, sizeof (rpath.frp_addr.ip4)); } else { - rpath.frp_proto = FIB_PROTOCOL_IP6; + rpath.frp_proto = DPO_PROTO_IP6; clib_memcpy (&rpath.frp_addr.ip6, mp->mt_next_hop, sizeof (rpath.frp_addr.ip6)); } diff --git a/src/vnet/mpls/mpls_tunnel.c b/src/vnet/mpls/mpls_tunnel.c index c025cc58..6452a60b 100644 --- a/src/vnet/mpls/mpls_tunnel.c +++ b/src/vnet/mpls/mpls_tunnel.c @@ -171,7 +171,7 @@ mpls_tunnel_mk_lb (mpls_tunnel_t *mt, vec_validate(ctx.next_hops, fib_path_list_get_n_paths(mt->mt_path_list)); vec_reset_length(ctx.next_hops); - lb_proto = vnet_link_to_dpo_proto(linkt); + lb_proto = fib_forw_chain_type_to_dpo_proto(fct); fib_path_list_walk(mt->mt_path_list, mpls_tunnel_collect_forwarding, @@ -313,12 +313,34 @@ mpls_tunnel_restack (mpls_tunnel_t *mt) /* * walk all the adjacencies on the MPLS interface and restack them */ - FOR_EACH_FIB_PROTOCOL(proto) + if (mt->mt_flags & MPLS_TUNNEL_FLAG_L2) { - adj_nbr_walk(mt->mt_sw_if_index, - proto, - mpls_adj_walk_cb, - NULL); + /* + * Stack a load-balance that drops, whilst we have no paths + */ + vnet_hw_interface_t * hi; + dpo_id_t dpo = DPO_INVALID; + + mpls_tunnel_mk_lb(mt, + VNET_LINK_MPLS, + FIB_FORW_CHAIN_TYPE_ETHERNET, + &dpo); + + hi = vnet_get_hw_interface(vnet_get_main(), mt->mt_hw_if_index); + dpo_stack_from_node(hi->tx_node_index, + &mt->mt_l2_lb, + &dpo); + dpo_reset(&dpo); + } + else + { + FOR_EACH_FIB_PROTOCOL(proto) + { + adj_nbr_walk(mt->mt_sw_if_index, + proto, + mpls_adj_walk_cb, + NULL); + } } } @@ -495,7 +517,7 @@ mpls_tunnel_tx (vlib_main_t * vm, b0 = vlib_get_buffer(vm, bi0); - vnet_buffer(b0)->ip.adj_index[VLIB_TX] = mt->mt_l2_adj; + vnet_buffer(b0)->ip.adj_index[VLIB_TX] = mt->mt_l2_lb.dpoi_index; if (PREDICT_FALSE(b0->flags & VLIB_BUFFER_IS_TRACED)) { @@ -506,7 +528,7 @@ mpls_tunnel_tx (vlib_main_t * vm, vlib_validate_buffer_enqueue_x1 (vm, node, next_index, to_next, n_left_to_next, - bi0, mt->mt_l2_tx_arc); + bi0, mt->mt_l2_lb.dpoi_next_node); } vlib_put_next_frame (vm, node, next_index, n_left_to_next); @@ -565,8 +587,7 @@ vnet_mpls_tunnel_del (u32 sw_if_index) if (FIB_NODE_INDEX_INVALID != mt->mt_path_list) fib_path_list_child_remove(mt->mt_path_list, mt->mt_sibling_index); - if (ADJ_INDEX_INVALID != mt->mt_l2_adj) - adj_unlock(mt->mt_l2_adj); + dpo_reset(&mt->mt_l2_lb); vec_add1 (mpls_tunnel_free_hw_if_indices, mt->mt_hw_if_index); pool_put(mpls_tunnel_pool, mt); @@ -587,12 +608,13 @@ vnet_mpls_tunnel_create (u8 l2_only, memset (mt, 0, sizeof (*mt)); mti = mt - mpls_tunnel_pool; fib_node_init(&mt->mt_node, FIB_NODE_TYPE_MPLS_TUNNEL); - mt->mt_l2_adj = ADJ_INDEX_INVALID; mt->mt_path_list = FIB_NODE_INDEX_INVALID; mt->mt_sibling_index = FIB_NODE_INDEX_INVALID; if (is_multicast) mt->mt_flags |= MPLS_TUNNEL_FLAG_MCAST; + if (l2_only) + mt->mt_flags |= MPLS_TUNNEL_FLAG_L2; /* * Create a new, or re=use and old, tunnel HW interface @@ -614,7 +636,7 @@ vnet_mpls_tunnel_create (u8 l2_only, mti, mpls_tunnel_hw_interface_class.index, mti); - hi = vnet_get_hw_interface(vnm, mt->mt_hw_if_index); + hi = vnet_get_hw_interface (vnm, mt->mt_hw_if_index); } /* @@ -624,19 +646,6 @@ vnet_mpls_tunnel_create (u8 l2_only, vec_validate_init_empty(mpls_tunnel_db, mt->mt_sw_if_index, ~0); mpls_tunnel_db[mt->mt_sw_if_index] = mti; - if (l2_only) - { - mt->mt_l2_adj = - adj_nbr_add_or_lock(fib_path_list_get_proto(mt->mt_path_list), - VNET_LINK_ETHERNET, - &zero_addr, - mt->mt_sw_if_index); - - mt->mt_l2_tx_arc = vlib_node_add_named_next(vlib_get_main(), - hi->tx_node_index, - "adj-l2-midchain"); - } - return (mt->mt_sw_if_index); } @@ -803,7 +812,7 @@ vnet_create_mpls_tunnel_command_fn (vlib_main_t * vm, &rpath.frp_sw_if_index)) { rpath.frp_weight = 1; - rpath.frp_proto = FIB_PROTOCOL_IP4; + rpath.frp_proto = DPO_PROTO_IP4; } else if (unformat (line_input, "via %U %U", @@ -813,7 +822,7 @@ vnet_create_mpls_tunnel_command_fn (vlib_main_t * vm, &rpath.frp_sw_if_index)) { rpath.frp_weight = 1; - rpath.frp_proto = FIB_PROTOCOL_IP6; + rpath.frp_proto = DPO_PROTO_IP6; } else if (unformat (line_input, "via %U", unformat_ip6_address, @@ -822,7 +831,7 @@ vnet_create_mpls_tunnel_command_fn (vlib_main_t * vm, rpath.frp_fib_index = 0; rpath.frp_weight = 1; rpath.frp_sw_if_index = ~0; - rpath.frp_proto = FIB_PROTOCOL_IP6; + rpath.frp_proto = DPO_PROTO_IP6; } else if (unformat (line_input, "via %U", unformat_ip4_address, @@ -831,7 +840,7 @@ vnet_create_mpls_tunnel_command_fn (vlib_main_t * vm, rpath.frp_fib_index = 0; rpath.frp_weight = 1; rpath.frp_sw_if_index = ~0; - rpath.frp_proto = FIB_PROTOCOL_IP4; + rpath.frp_proto = DPO_PROTO_IP4; } else if (unformat (line_input, "l2-only")) l2_only = 1; @@ -915,6 +924,14 @@ format_mpls_tunnel (u8 * s, va_list * args) s = format(s, "%U", format_fib_path_ext_list, &mt->mt_path_exts); s = format(s, "\n"); + if (mt->mt_flags & MPLS_TUNNEL_FLAG_L2) + { + s = format(s, " forwarding: %U\n", + format_fib_forw_chain_type, + FIB_FORW_CHAIN_TYPE_ETHERNET); + s = format(s, " %U\n", format_dpo_id, &mt->mt_l2_lb, 2); + } + return (s); } diff --git a/src/vnet/mpls/mpls_tunnel.h b/src/vnet/mpls/mpls_tunnel.h index 4cb0a860..285817c3 100644 --- a/src/vnet/mpls/mpls_tunnel.h +++ b/src/vnet/mpls/mpls_tunnel.h @@ -22,15 +22,20 @@ typedef enum mpls_tunnel_attribute_t_ { MPLS_TUNNEL_ATTRIBUTE_FIRST = 0, + /** + * @brief The tunnel is L2 only + */ + MPLS_TUNNEL_ATTRIBUTE_L2 = MPLS_TUNNEL_ATTRIBUTE_FIRST, /** * @brief The tunnel has an underlying multicast LSP */ - MPLS_TUNNEL_ATTRIBUTE_MCAST = MPLS_TUNNEL_ATTRIBUTE_FIRST, + MPLS_TUNNEL_ATTRIBUTE_MCAST, MPLS_TUNNEL_ATTRIBUTE_LAST = MPLS_TUNNEL_ATTRIBUTE_MCAST, } mpls_tunnel_attribute_t; #define MPLS_TUNNEL_ATTRIBUTES { \ [MPLS_TUNNEL_ATTRIBUTE_MCAST] = "multicast", \ + [MPLS_TUNNEL_ATTRIBUTE_L2] = "L2", \ } #define FOR_EACH_MPLS_TUNNEL_ATTRIBUTE(_item) \ for (_item = MPLS_TUNNEL_ATTRIBUTE_FIRST; \ @@ -39,6 +44,7 @@ typedef enum mpls_tunnel_attribute_t_ typedef enum mpls_tunnel_flag_t_ { MPLS_TUNNEL_FLAG_NONE = 0, + MPLS_TUNNEL_FLAG_L2 = (1 << MPLS_TUNNEL_ATTRIBUTE_L2), MPLS_TUNNEL_FLAG_MCAST = (1 << MPLS_TUNNEL_ATTRIBUTE_MCAST), } __attribute__ ((packed)) mpls_tunnel_flags_t; @@ -60,14 +66,19 @@ typedef struct mpls_tunnel_t_ /** * @brief If the tunnel is an L2 tunnel, this is the link type ETHERNET - * adjacency + * load-balance + */ + dpo_id_t mt_l2_lb; + + /** + * @brief The HW interface index of the tunnel interfaces */ - adj_index_t mt_l2_adj; + u32 mt_hw_if_index; /** - * @brief on a L2 tunnel this is the VLIB arc from the L2-tx to the l2-midchain + * @brief The SW interface index of the tunnel interfaces */ - u32 mt_l2_tx_arc; + u32 mt_sw_if_index; /** * @brief The path-list over which the tunnel's destination is reachable @@ -83,23 +94,6 @@ typedef struct mpls_tunnel_t_ * A vector of path extensions o hold the label stack for each path */ fib_path_ext_list_t mt_path_exts; - - /** - * @brief Flag to indicate the tunnel is only for L2 traffic, that is - * this tunnel belongs in a bridge domain. - */ - u8 mt_l2_only; - - /** - * @brief The HW interface index of the tunnel interfaces - */ - u32 mt_hw_if_index; - - /** - * @brief The SW interface index of the tunnel interfaces - */ - u32 mt_sw_if_index; - } mpls_tunnel_t; /** diff --git a/src/vnet/srmpls/sr_mpls_policy.c b/src/vnet/srmpls/sr_mpls_policy.c index 5ebbc60d..db4ad2a7 100755 --- a/src/vnet/srmpls/sr_mpls_policy.c +++ b/src/vnet/srmpls/sr_mpls_policy.c @@ -75,7 +75,7 @@ create_sl (mpls_sr_policy_t * sr_policy, mpls_label_t * sl, u32 weight) segment_list->segments = vec_dup (sl); fib_route_path_t path = { - .frp_proto = FIB_PROTOCOL_MPLS, + .frp_proto = DPO_PROTO_MPLS, .frp_sw_if_index = ~0, .frp_fib_index = 0, .frp_weight = segment_list->weight, @@ -203,7 +203,7 @@ sr_mpls_policy_del (mpls_label_t bsid, u32 index) segment_list = pool_elt_at_index (sm->sid_lists, *sl_index); fib_route_path_t path = { - .frp_proto = FIB_PROTOCOL_MPLS, + .frp_proto = DPO_PROTO_MPLS, .frp_sw_if_index = ~0, .frp_fib_index = 0, .frp_weight = segment_list->weight, @@ -308,7 +308,7 @@ sr_mpls_policy_mod (mpls_label_t bsid, u32 index, u8 operation, mpls_eos_bit_t eos; fib_route_path_t path = { - .frp_proto = FIB_PROTOCOL_MPLS, + .frp_proto = DPO_PROTO_MPLS, .frp_sw_if_index = ~0, .frp_fib_index = 0, .frp_weight = segment_list->weight, diff --git a/src/vnet/srmpls/sr_mpls_steering.c b/src/vnet/srmpls/sr_mpls_steering.c index 37707049..3a9aea2d 100755 --- a/src/vnet/srmpls/sr_mpls_steering.c +++ b/src/vnet/srmpls/sr_mpls_steering.c @@ -218,7 +218,7 @@ sr_mpls_steering_policy (int is_del, mpls_label_t bsid, u32 sr_policy_index, update_fib:; fib_route_path_t path = { - .frp_proto = FIB_PROTOCOL_MPLS, + .frp_proto = DPO_PROTO_MPLS, .frp_local_label = sr_policy->bsid, .frp_eos = MPLS_EOS, .frp_sw_if_index = ~0, diff --git a/src/vnet/srv6/sr_steering.c b/src/vnet/srv6/sr_steering.c index a7903751..704adaa7 100755 --- a/src/vnet/srv6/sr_steering.c +++ b/src/vnet/srv6/sr_steering.c @@ -310,7 +310,7 @@ update_fib: table_id : 0)), &pfx, FIB_SOURCE_SR, FIB_ENTRY_FLAG_LOOSE_URPF_EXEMPT, - FIB_PROTOCOL_IP6, + DPO_PROTO_IP6, (ip46_address_t *) & sr_policy->bsid, ~0, sm->fib_table_ip6, 1, NULL, FIB_ROUTE_PATH_FLAG_NONE); @@ -327,7 +327,7 @@ update_fib: table_id : 0)), &pfx, FIB_SOURCE_SR, FIB_ENTRY_FLAG_LOOSE_URPF_EXEMPT, - FIB_PROTOCOL_IP6, + DPO_PROTO_IP6, (ip46_address_t *) & sr_policy->bsid, ~0, sm->fib_table_ip4, 1, NULL, FIB_ROUTE_PATH_FLAG_NONE); diff --git a/src/vnet/vxlan-gpe/vxlan_gpe.c b/src/vnet/vxlan-gpe/vxlan_gpe.c index 97bb1b15..462c79a0 100644 --- a/src/vnet/vxlan-gpe/vxlan_gpe.c +++ b/src/vnet/vxlan-gpe/vxlan_gpe.c @@ -638,7 +638,7 @@ int vnet_vxlan_gpe_add_del_tunnel fib_node_index_t mfei; adj_index_t ai; fib_route_path_t path = { - .frp_proto = fp, + .frp_proto = fib_proto_to_dpo(fp), .frp_addr = zero_addr, .frp_sw_if_index = 0xffffffff, .frp_fib_index = ~0, diff --git a/src/vnet/vxlan/vxlan.c b/src/vnet/vxlan/vxlan.c index 1b3df2a8..dc973372 100644 --- a/src/vnet/vxlan/vxlan.c +++ b/src/vnet/vxlan/vxlan.c @@ -505,7 +505,7 @@ int vnet_vxlan_add_del_tunnel fib_node_index_t mfei; adj_index_t ai; fib_route_path_t path = { - .frp_proto = fp, + .frp_proto = fib_proto_to_dpo(fp), .frp_addr = zero_addr, .frp_sw_if_index = 0xffffffff, .frp_fib_index = ~0, diff --git a/src/vpp/app/vpe_cli.c b/src/vpp/app/vpe_cli.c index 94bdc84c..fcc496ad 100644 --- a/src/vpp/app/vpe_cli.c +++ b/src/vpp/app/vpe_cli.c @@ -98,7 +98,7 @@ virtual_ip_cmd_fn_command_fn (vlib_main_t * vm, vec_add2 (rpaths, rpath, 1); - rpath->frp_proto = FIB_PROTOCOL_IP4; + rpath->frp_proto = DPO_PROTO_IP4; rpath->frp_addr = next_hops[i]; rpath->frp_sw_if_index = sw_if_index; rpath->frp_fib_index = ~0; diff --git a/test/test_bfd.py b/test/test_bfd.py index be42cdad..4cb6d379 100644 --- a/test/test_bfd.py +++ b/test/test_bfd.py @@ -20,7 +20,7 @@ from vpp_pg_interface import CaptureTimeoutError, is_ipv6_misc from vpp_lo_interface import VppLoInterface from util import ppp from vpp_papi_provider import UnexpectedApiReturnValueError -from vpp_ip_route import VppIpRoute, VppRoutePath +from vpp_ip_route import VppIpRoute, VppRoutePath, DpoProto USEC_IN_SEC = 1000000 @@ -1678,12 +1678,12 @@ class BFDFIBTestCase(VppTestCase): ip_2001_s_64 = VppIpRoute(self, "2001::", 64, [VppRoutePath(self.pg0.remote_ip6, self.pg0.sw_if_index, - is_ip6=1)], + proto=DPO_PROTO_IP6)], is_ip6=1) ip_2002_s_64 = VppIpRoute(self, "2002::", 64, [VppRoutePath(self.pg0.remote_ip6, 0xffffffff, - is_ip6=1)], + proto=DPO_PROTO_IP6)], is_ip6=1) ip_2001_s_64.add_vpp_config() ip_2002_s_64.add_vpp_config() diff --git a/test/test_gre.py b/test/test_gre.py index 18b67dbd..1afc44fb 100644 --- a/test/test_gre.py +++ b/test/test_gre.py @@ -6,7 +6,7 @@ from logging import * from framework import VppTestCase, VppTestRunner from vpp_sub_interface import VppDot1QSubint from vpp_gre_interface import VppGreInterface, VppGre6Interface -from vpp_ip_route import VppIpRoute, VppRoutePath +from vpp_ip_route import VppIpRoute, VppRoutePath, DpoProto from vpp_papi_provider import L2_VTR_OP from scapy.packet import Raw @@ -516,11 +516,12 @@ class TestGRE(VppTestCase): gre_if.admin_up() gre_if.config_ip6() - route_via_tun = VppIpRoute(self, "4004::1", 128, - [VppRoutePath("0::0", - gre_if.sw_if_index, - is_ip6=1)], - is_ip6=1) + route_via_tun = VppIpRoute( + self, "4004::1", 128, + [VppRoutePath("0::0", + gre_if.sw_if_index, + proto=DpoProto.DPO_PROTO_IP6)], + is_ip6=1) route_via_tun.add_vpp_config() @@ -542,11 +543,12 @@ class TestGRE(VppTestCase): # # Add a route that resolves the tunnel's destination # - route_tun_dst = VppIpRoute(self, "1002::1", 128, - [VppRoutePath(self.pg2.remote_ip6, - self.pg2.sw_if_index, - is_ip6=1)], - is_ip6=1) + route_tun_dst = VppIpRoute( + self, "1002::1", 128, + [VppRoutePath(self.pg2.remote_ip6, + self.pg2.sw_if_index, + proto=DpoProto.DPO_PROTO_IP6)], + is_ip6=1) route_tun_dst.add_vpp_config() # diff --git a/test/test_ip6.py b/test/test_ip6.py index 593f6868..285ce181 100644 --- a/test/test_ip6.py +++ b/test/test_ip6.py @@ -8,7 +8,7 @@ from vpp_sub_interface import VppSubInterface, VppDot1QSubint from vpp_pg_interface import is_ipv6_misc from vpp_ip_route import VppIpRoute, VppRoutePath, find_route, VppIpMRoute, \ VppMRoutePath, MRouteItfFlags, MRouteEntryFlags, VppMplsIpBind, \ - VppMplsRoute + VppMplsRoute, DpoProto from vpp_neighbor import find_nbr, VppNeighbor from scapy.packet import Raw @@ -490,7 +490,7 @@ class TestIPv6(TestIPv6ND): inet=AF_INET6)) def test_ns_duplicates(self): - """ ARP Duplicates""" + """ ND Duplicates""" # # Generate some hosts on the LAN @@ -537,7 +537,7 @@ class TestIPv6(TestIPv6ND): # # remove the duplicate on pg1 - # packet stream shoud generate ARPs out of pg1 + # packet stream shoud generate NSs out of pg1 # ns_pg1.remove_vpp_config() @@ -1347,10 +1347,10 @@ class TestIP6LoadBalance(VppTestCase): route_3000_1 = VppIpRoute(self, "3000::1", 128, [VppRoutePath(self.pg1.remote_ip6, self.pg1.sw_if_index, - is_ip6=1), + proto=DpoProto.DPO_PROTO_IP6), VppRoutePath(self.pg2.remote_ip6, self.pg2.sw_if_index, - is_ip6=1)], + proto=DpoProto.DPO_PROTO_IP6)], is_ip6=1) route_3000_1.add_vpp_config() @@ -1367,11 +1367,11 @@ class TestIP6LoadBalance(VppTestCase): [VppRoutePath(self.pg1.remote_ip6, self.pg1.sw_if_index, labels=[67], - is_ip6=1), + proto=DpoProto.DPO_PROTO_IP6), VppRoutePath(self.pg2.remote_ip6, self.pg2.sw_if_index, labels=[67], - is_ip6=1)]) + proto=DpoProto.DPO_PROTO_IP6)]) route_67.add_vpp_config() # @@ -1441,20 +1441,20 @@ class TestIP6LoadBalance(VppTestCase): route_3000_2 = VppIpRoute(self, "3000::2", 128, [VppRoutePath(self.pg3.remote_ip6, self.pg3.sw_if_index, - is_ip6=1), + proto=DpoProto.DPO_PROTO_IP6), VppRoutePath(self.pg4.remote_ip6, self.pg4.sw_if_index, - is_ip6=1)], + proto=DpoProto.DPO_PROTO_IP6)], is_ip6=1) route_3000_2.add_vpp_config() route_4000_1 = VppIpRoute(self, "4000::1", 128, [VppRoutePath("3000::1", 0xffffffff, - is_ip6=1), + proto=DpoProto.DPO_PROTO_IP6), VppRoutePath("3000::2", 0xffffffff, - is_ip6=1)], + proto=DpoProto.DPO_PROTO_IP6)], is_ip6=1) route_4000_1.add_vpp_config() @@ -1485,14 +1485,14 @@ class TestIP6LoadBalance(VppTestCase): route_5000_2 = VppIpRoute(self, "5000::2", 128, [VppRoutePath(self.pg3.remote_ip6, self.pg3.sw_if_index, - is_ip6=1)], + proto=DpoProto.DPO_PROTO_IP6)], is_ip6=1) route_5000_2.add_vpp_config() route_6000_1 = VppIpRoute(self, "6000::1", 128, [VppRoutePath("5000::2", 0xffffffff, - is_ip6=1)], + proto=DpoProto.DPO_PROTO_IP6)], is_ip6=1) route_6000_1.add_vpp_config() diff --git a/test/test_map.py b/test/test_map.py index 9ac3948a..bbf4aec2 100644 --- a/test/test_map.py +++ b/test/test_map.py @@ -4,7 +4,7 @@ import unittest import socket from framework import VppTestCase, VppTestRunner -from vpp_ip_route import VppIpRoute, VppRoutePath +from vpp_ip_route import VppIpRoute, VppRoutePath, DpoProto from scapy.layers.l2 import Ether, Raw from scapy.layers.inet import IP, UDP, ICMP @@ -75,7 +75,7 @@ class TestMAP(VppTestCase): map_br_pfx_len, [VppRoutePath(self.pg1.remote_ip6, self.pg1.sw_if_index, - is_ip6=1)], + proto=DpoProto.DPO_PROTO_IP6)], is_ip6=1) map_route.add_vpp_config() @@ -138,13 +138,12 @@ class TestMAP(VppTestCase): # Add a route to 4001::1. Expect the encapped traffic to be # sent via that routes next-hop # - pre_res_route = VppIpRoute(self, - "4001::1", - 128, - [VppRoutePath(self.pg1.remote_hosts[2].ip6, - self.pg1.sw_if_index, - is_ip6=1)], - is_ip6=1) + pre_res_route = VppIpRoute( + self, "4001::1", 128, + [VppRoutePath(self.pg1.remote_hosts[2].ip6, + self.pg1.sw_if_index, + proto=DpoProto.DPO_PROTO_IP6)], + is_ip6=1) pre_res_route.add_vpp_config() self.send_and_assert_encapped(v4, map_src, @@ -156,7 +155,7 @@ class TestMAP(VppTestCase): # pre_res_route.modify([VppRoutePath(self.pg1.remote_hosts[3].ip6, self.pg1.sw_if_index, - is_ip6=1)]) + proto=DpoProto.DPO_PROTO_IP6)]) pre_res_route.add_vpp_config() self.send_and_assert_encapped(v4, map_src, diff --git a/test/test_mpls.py b/test/test_mpls.py index e3d013af..b2226a74 100644 --- a/test/test_mpls.py +++ b/test/test_mpls.py @@ -6,7 +6,7 @@ import socket from framework import VppTestCase, VppTestRunner from vpp_ip_route import VppIpRoute, VppRoutePath, VppMplsRoute, \ VppMplsIpBind, VppIpMRoute, VppMRoutePath, \ - MRouteItfFlags, MRouteEntryFlags + MRouteItfFlags, MRouteEntryFlags, DpoProto from vpp_mpls_tunnel_interface import VppMPLSTunnelInterface from scapy.packet import Raw @@ -16,6 +16,38 @@ from scapy.layers.inet6 import IPv6 from scapy.contrib.mpls import MPLS +def verify_filter(capture, sent): + if not len(capture) == len(sent): + # filter out any IPv6 RAs from the capture + for p in capture: + if p.haslayer(IPv6): + capture.remove(p) + return capture + + +def verify_mpls_stack(tst, rx, mpls_labels, ttl=255, num=0): + # the rx'd packet has the MPLS label popped + eth = rx[Ether] + tst.assertEqual(eth.type, 0x8847) + + rx_mpls = rx[MPLS] + + for ii in range(len(mpls_labels)): + tst.assertEqual(rx_mpls.label, mpls_labels[ii]) + tst.assertEqual(rx_mpls.cos, 0) + if ii == num: + tst.assertEqual(rx_mpls.ttl, ttl) + else: + tst.assertEqual(rx_mpls.ttl, 255) + if ii == len(mpls_labels) - 1: + tst.assertEqual(rx_mpls.s, 1) + else: + # not end of stack + tst.assertEqual(rx_mpls.s, 0) + # pop the label to expose the next + rx_mpls = rx_mpls[MPLS].payload + + class TestMPLS(VppTestCase): """ MPLS Test Case """ @@ -120,18 +152,9 @@ class TestMPLS(VppTestCase): pkts.append(p) return pkts - @staticmethod - def verify_filter(capture, sent): - if not len(capture) == len(sent): - # filter out any IPv6 RAs from the capture - for p in capture: - if p.haslayer(IPv6): - capture.remove(p) - return capture - def verify_capture_ip4(self, src_if, capture, sent, ping_resp=0): try: - capture = self.verify_filter(capture, sent) + capture = verify_filter(capture, sent) self.assertEqual(len(capture), len(sent)) @@ -158,33 +181,10 @@ class TestMPLS(VppTestCase): except: raise - def verify_mpls_stack(self, rx, mpls_labels, ttl=255, num=0): - # the rx'd packet has the MPLS label popped - eth = rx[Ether] - self.assertEqual(eth.type, 0x8847) - - rx_mpls = rx[MPLS] - - for ii in range(len(mpls_labels)): - self.assertEqual(rx_mpls.label, mpls_labels[ii]) - self.assertEqual(rx_mpls.cos, 0) - if ii == num: - self.assertEqual(rx_mpls.ttl, ttl) - else: - self.assertEqual(rx_mpls.ttl, 255) - - if ii == len(mpls_labels) - 1: - self.assertEqual(rx_mpls.s, 1) - else: - # not end of stack - self.assertEqual(rx_mpls.s, 0) - # pop the label to expose the next - rx_mpls = rx_mpls[MPLS].payload - def verify_capture_labelled_ip4(self, src_if, capture, sent, mpls_labels): try: - capture = self.verify_filter(capture, sent) + capture = verify_filter(capture, sent) self.assertEqual(len(capture), len(sent)) @@ -195,8 +195,8 @@ class TestMPLS(VppTestCase): rx_ip = rx[IP] # the MPLS TTL is copied from the IP - self.verify_mpls_stack( - rx, mpls_labels, rx_ip.ttl, len(mpls_labels) - 1) + verify_mpls_stack(self, rx, mpls_labels, rx_ip.ttl, + len(mpls_labels) - 1) self.assertEqual(rx_ip.src, tx_ip.src) self.assertEqual(rx_ip.dst, tx_ip.dst) @@ -211,7 +211,7 @@ class TestMPLS(VppTestCase): if top is None: top = len(mpls_labels) - 1 try: - capture = self.verify_filter(capture, sent) + capture = verify_filter(capture, sent) self.assertEqual(len(capture), len(sent)) @@ -222,8 +222,7 @@ class TestMPLS(VppTestCase): rx_ip = rx[IP] # the MPLS TTL is 255 since it enters a new tunnel - self.verify_mpls_stack( - rx, mpls_labels, ttl, top) + verify_mpls_stack(self, rx, mpls_labels, ttl, top) self.assertEqual(rx_ip.src, tx_ip.src) self.assertEqual(rx_ip.dst, tx_ip.dst) @@ -236,13 +235,13 @@ class TestMPLS(VppTestCase): def verify_capture_labelled(self, src_if, capture, sent, mpls_labels, ttl=254, num=0): try: - capture = self.verify_filter(capture, sent) + capture = verify_filter(capture, sent) self.assertEqual(len(capture), len(sent)) for i in range(len(capture)): rx = capture[i] - self.verify_mpls_stack(rx, mpls_labels, ttl, num) + verify_mpls_stack(self, rx, mpls_labels, ttl, num) except: raise @@ -1049,7 +1048,7 @@ class TestMPLS(VppTestCase): self.pg1.sw_if_index, nh_table_id=1, rpf_id=55, - is_ip6=1)], + proto=DpoProto.DPO_PROTO_IP6)], is_multicast=1) route_34_eos.add_vpp_config() @@ -1440,19 +1439,20 @@ class TestMPLSPIC(VppTestCase): for ii in range(64): dst = "3000::%d" % ii local_label = 1600 + ii - vpn_routes.append(VppIpRoute(self, dst, 128, - [VppRoutePath(self.pg2.remote_ip6, - 0xffffffff, - nh_table_id=1, - is_resolve_attached=1, - is_ip6=1), - VppRoutePath(self.pg3.remote_ip6, - 0xffffffff, - nh_table_id=1, - is_ip6=1, - is_resolve_attached=1)], - table_id=1, - is_ip6=1)) + vpn_routes.append(VppIpRoute( + self, dst, 128, + [VppRoutePath(self.pg2.remote_ip6, + 0xffffffff, + nh_table_id=1, + is_resolve_attached=1, + proto=DpoProto.DPO_PROTO_IP6), + VppRoutePath(self.pg3.remote_ip6, + 0xffffffff, + nh_table_id=1, + proto=DpoProto.DPO_PROTO_IP6, + is_resolve_attached=1)], + table_id=1, + is_ip6=1)) vpn_routes[ii].add_vpp_config() vpn_bindings.append(VppMplsIpBind(self, local_label, dst, 128, @@ -1525,5 +1525,211 @@ class TestMPLSPIC(VppTestCase): self.assertNotEqual(0, len(rx1)) +class TestMPLSL2(VppTestCase): + """ MPLS-L2 """ + + def setUp(self): + super(TestMPLSL2, self).setUp() + + # create 2 pg interfaces + self.create_pg_interfaces(range(2)) + + # use pg0 as the core facing interface + self.pg0.admin_up() + self.pg0.config_ip4() + self.pg0.resolve_arp() + self.pg0.enable_mpls() + + # use the other 2 for customer facg L2 links + for i in self.pg_interfaces[1:]: + i.admin_up() + + def tearDown(self): + super(TestMPLSL2, self).tearDown() + for i in self.pg_interfaces[1:]: + i.admin_down() + + self.pg0.disable_mpls() + self.pg0.unconfig_ip4() + self.pg0.admin_down() + + def verify_capture_tunneled_ethernet(self, capture, sent, mpls_labels, + ttl=255, top=None): + if top is None: + top = len(mpls_labels) - 1 + + capture = verify_filter(capture, sent) + + self.assertEqual(len(capture), len(sent)) + + for i in range(len(capture)): + tx = sent[i] + rx = capture[i] + + # the MPLS TTL is 255 since it enters a new tunnel + verify_mpls_stack(self, rx, mpls_labels, ttl, top) + + tx_eth = tx[Ether] + rx_eth = Ether(str(rx[MPLS].payload)) + + self.assertEqual(rx_eth.src, tx_eth.src) + self.assertEqual(rx_eth.dst, tx_eth.dst) + + def test_vpws(self): + """ Virtual Private Wire Service """ + + # + # Create an MPLS tunnel that pushes 1 label + # + mpls_tun_1 = VppMPLSTunnelInterface(self, + [VppRoutePath(self.pg0.remote_ip4, + self.pg0.sw_if_index, + labels=[42])], + is_l2=1) + mpls_tun_1.add_vpp_config() + mpls_tun_1.admin_up() + + # + # Create a label entry to for 55 that does L2 input to the tunnel + # + route_55_eos = VppMplsRoute( + self, 55, 1, + [VppRoutePath("0.0.0.0", + mpls_tun_1.sw_if_index, + is_interface_rx=1, + proto=DpoProto.DPO_PROTO_ETHERNET)]) + route_55_eos.add_vpp_config() + + # + # Cross-connect the tunnel with one of the customers L2 interfaces + # + self.vapi.sw_interface_set_l2_xconnect(self.pg1.sw_if_index, + mpls_tun_1.sw_if_index, + enable=1) + self.vapi.sw_interface_set_l2_xconnect(mpls_tun_1.sw_if_index, + self.pg1.sw_if_index, + enable=1) + + # + # inject a packet from the core + # + pcore = (Ether(dst=self.pg0.local_mac, + src=self.pg0.remote_mac) / + MPLS(label=55, ttl=64) / + Ether(dst="00:00:de:ad:ba:be", + src="00:00:de:ad:be:ef") / + IP(src="10.10.10.10", dst="11.11.11.11") / + UDP(sport=1234, dport=1234) / + Raw('\xa5' * 100)) + + self.pg0.add_stream(pcore * 65) + self.pg_enable_capture(self.pg_interfaces) + self.pg_start() + + rx0 = self.pg1.get_capture(65) + tx = pcore[MPLS].payload + + self.assertEqual(rx0[0][Ether].dst, tx[Ether].dst) + self.assertEqual(rx0[0][Ether].src, tx[Ether].src) + + # + # Inject a packet from the custoer/L2 side + # + self.pg1.add_stream(tx * 65) + self.pg_enable_capture(self.pg_interfaces) + self.pg_start() + + rx0 = self.pg0.get_capture(65) + + self.verify_capture_tunneled_ethernet(rx0, tx*65, [42]) + + def test_vpls(self): + """ Virtual Private LAN Service """ + # + # Create an L2 MPLS tunnel + # + mpls_tun = VppMPLSTunnelInterface(self, + [VppRoutePath(self.pg0.remote_ip4, + self.pg0.sw_if_index, + labels=[42])], + is_l2=1) + mpls_tun.add_vpp_config() + mpls_tun.admin_up() + + # + # Create a label entry to for 55 that does L2 input to the tunnel + # + route_55_eos = VppMplsRoute( + self, 55, 1, + [VppRoutePath("0.0.0.0", + mpls_tun.sw_if_index, + is_interface_rx=1, + proto=DpoProto.DPO_PROTO_ETHERNET)]) + route_55_eos.add_vpp_config() + + # + # add to tunnel to the customers bridge-domain + # + self.vapi.sw_interface_set_l2_bridge(mpls_tun.sw_if_index, + bd_id=1) + self.vapi.sw_interface_set_l2_bridge(self.pg1.sw_if_index, + bd_id=1) + + # + # Packet from the customer interface and from the core + # + p_cust = (Ether(dst="00:00:de:ad:ba:be", + src="00:00:de:ad:be:ef") / + IP(src="10.10.10.10", dst="11.11.11.11") / + UDP(sport=1234, dport=1234) / + Raw('\xa5' * 100)) + p_core = (Ether(src="00:00:de:ad:ba:be", + dst="00:00:de:ad:be:ef") / + IP(dst="10.10.10.10", src="11.11.11.11") / + UDP(sport=1234, dport=1234) / + Raw('\xa5' * 100)) + + # + # The BD is learning, so send in one of each packet to learn + # + p_core_encap = (Ether(dst=self.pg0.local_mac, + src=self.pg0.remote_mac) / + MPLS(label=55, ttl=64) / + p_core) + + self.pg1.add_stream(p_cust) + self.pg_enable_capture(self.pg_interfaces) + self.pg_start() + self.pg0.add_stream(p_core_encap) + self.pg_enable_capture(self.pg_interfaces) + self.pg_start() + + # we've learnt this so expect it be be forwarded + rx0 = self.pg1.get_capture(1) + + self.assertEqual(rx0[0][Ether].dst, p_core[Ether].dst) + self.assertEqual(rx0[0][Ether].src, p_core[Ether].src) + + # + # now a stream in each direction + # + self.pg1.add_stream(p_cust * 65) + self.pg_enable_capture(self.pg_interfaces) + self.pg_start() + + rx0 = self.pg0.get_capture(65) + + self.verify_capture_tunneled_ethernet(rx0, p_cust*65, [42]) + + # + # remove interfaces from customers bridge-domain + # + self.vapi.sw_interface_set_l2_bridge(mpls_tun.sw_if_index, + bd_id=1, + enable=0) + self.vapi.sw_interface_set_l2_bridge(self.pg1.sw_if_index, + bd_id=1, + enable=0) + if __name__ == '__main__': unittest.main(testRunner=VppTestRunner) diff --git a/test/test_p2p_ethernet.py b/test/test_p2p_ethernet.py index 37a1d18b..8688f7e6 100644 --- a/test/test_p2p_ethernet.py +++ b/test/test_p2p_ethernet.py @@ -11,7 +11,7 @@ from scapy.layers.inet6 import IPv6 from framework import VppTestCase, VppTestRunner, running_extended_tests from vpp_sub_interface import VppP2PSubint -from vpp_ip_route import VppIpRoute, VppRoutePath +from vpp_ip_route import VppIpRoute, VppRoutePath, DpoProto from util import mactobinary @@ -219,7 +219,7 @@ class P2PEthernetIPV6(VppTestCase): route_8000 = VppIpRoute(self, "8000::", 64, [VppRoutePath(self.pg0.remote_ip6, self.pg0.sw_if_index, - is_ip6=1)], + proto=DpoProto.DPO_PROTO_IP6)], is_ip6=1) route_8000.add_vpp_config() @@ -239,7 +239,7 @@ class P2PEthernetIPV6(VppTestCase): route_9001 = VppIpRoute(self, "9001::", 64, [VppRoutePath(self.pg1.remote_ip6, self.pg1.sw_if_index, - is_ip6=1)], + proto=DpoProto.DPO_PROTO_IP6)], is_ip6=1) route_9001.add_vpp_config() @@ -264,7 +264,7 @@ class P2PEthernetIPV6(VppTestCase): route_3 = VppIpRoute(self, "9000::", 64, [VppRoutePath(self.pg1._remote_hosts[0].ip6, self.pg1.sw_if_index, - is_ip6=1)], + proto=DpoProto.DPO_PROTO_IP6)], is_ip6=1) route_3.add_vpp_config() @@ -289,7 +289,7 @@ class P2PEthernetIPV6(VppTestCase): route_9001 = VppIpRoute(self, "9000::", 64, [VppRoutePath(self.pg1._remote_hosts[0].ip6, self.pg1.sw_if_index, - is_ip6=1)], + proto=DpoProto.DPO_PROTO_IP6)], is_ip6=1) route_9001.add_vpp_config() @@ -310,19 +310,19 @@ class P2PEthernetIPV6(VppTestCase): route_8000 = VppIpRoute(self, "8000::", 64, [VppRoutePath(self.pg0.remote_ip6, self.pg0.sw_if_index, - is_ip6=1)], + proto=DpoProto.DPO_PROTO_IP6)], is_ip6=1) route_8000.add_vpp_config() route_8001 = VppIpRoute(self, "8001::", 64, [VppRoutePath(self.p2p_sub_ifs[0].remote_ip6, self.p2p_sub_ifs[0].sw_if_index, - is_ip6=1)], + proto=DpoProto.DPO_PROTO_IP6)], is_ip6=1) route_8001.add_vpp_config() route_8002 = VppIpRoute(self, "8002::", 64, [VppRoutePath(self.p2p_sub_ifs[1].remote_ip6, self.p2p_sub_ifs[1].sw_if_index, - is_ip6=1)], + proto=DpoProto.DPO_PROTO_IP6)], is_ip6=1) route_8002.add_vpp_config() diff --git a/test/vpp_ip_route.py b/test/vpp_ip_route.py index badb3102..2c489e3c 100644 --- a/test/vpp_ip_route.py +++ b/test/vpp_ip_route.py @@ -29,6 +29,14 @@ class MRouteEntryFlags: MFIB_ENTRY_FLAG_INHERIT_ACCEPT = 8 +class DpoProto: + DPO_PROTO_IP4 = 0 + DPO_PROTO_IP6 = 1 + DPO_PROTO_MPLS = 2 + DPO_PROTO_ETHERNET = 3 + DPO_PROTO_NSH = 4 + + def find_route(test, ip_addr, len, table_id=0, inet=AF_INET): if inet == AF_INET: s = 4 @@ -55,22 +63,24 @@ class VppRoutePath(object): nh_table_id=0, labels=[], nh_via_label=MPLS_LABEL_INVALID, - is_ip6=0, rpf_id=0, is_interface_rx=0, is_resolve_host=0, - is_resolve_attached=0): + is_resolve_attached=0, + proto=DpoProto.DPO_PROTO_IP4): self.nh_itf = nh_sw_if_index self.nh_table_id = nh_table_id self.nh_via_label = nh_via_label self.nh_labels = labels self.weight = 1 self.rpf_id = rpf_id - self.is_ip4 = 1 if is_ip6 == 0 else 0 - if self.is_ip4: + self.proto = proto + if self.proto is DpoProto.DPO_PROTO_IP6: + self.nh_addr = inet_pton(AF_INET6, nh_addr) + elif self.proto is DpoProto.DPO_PROTO_IP4: self.nh_addr = inet_pton(AF_INET, nh_addr) else: - self.nh_addr = inet_pton(AF_INET6, nh_addr) + self.nh_addr = inet_pton(AF_INET6, "::") self.is_resolve_host = is_resolve_host self.is_resolve_attached = is_resolve_attached self.is_interface_rx = is_interface_rx @@ -401,7 +411,7 @@ class VppMplsRoute(VppObject): self._test.vapi.mpls_route_add_del( self.local_label, self.eos_bit, - path.is_ip4, + path.proto, path.nh_addr, path.nh_itf, is_multicast=self.is_multicast, @@ -420,7 +430,7 @@ class VppMplsRoute(VppObject): for path in self.paths: self._test.vapi.mpls_route_add_del(self.local_label, self.eos_bit, - 1, + path.proto, path.nh_addr, path.nh_itf, is_rpf_id=path.is_rpf_id, diff --git a/test/vpp_mpls_tunnel_interface.py b/test/vpp_mpls_tunnel_interface.py index f2001574..0542b05c 100644 --- a/test/vpp_mpls_tunnel_interface.py +++ b/test/vpp_mpls_tunnel_interface.py @@ -9,13 +9,14 @@ class VppMPLSTunnelInterface(VppInterface): VPP MPLS Tunnel interface """ - def __init__(self, test, paths, is_multicast=0): + def __init__(self, test, paths, is_multicast=0, is_l2=0): """ Create MPLS Tunnel interface """ self._sw_if_index = 0 super(VppMPLSTunnelInterface, self).__init__(test) self._test = test self.t_paths = paths self.is_multicast = is_multicast + self.is_l2 = is_l2 def add_vpp_config(self): self._sw_if_index = 0xffffffff @@ -29,7 +30,8 @@ class VppMPLSTunnelInterface(VppInterface): path.weight, next_hop_out_label_stack=path.nh_labels, next_hop_n_out_labels=len(path.nh_labels), - is_multicast=self.is_multicast) + is_multicast=self.is_multicast, + l2_only=self.is_l2) self._sw_if_index = reply.sw_if_index def remove_vpp_config(self): diff --git a/test/vpp_papi_provider.py b/test/vpp_papi_provider.py index 801a6c2d..3ba2ad4a 100644 --- a/test/vpp_papi_provider.py +++ b/test/vpp_papi_provider.py @@ -921,7 +921,7 @@ class VppPapiProvider(object): self, label, eos, - next_hop_proto_is_ip4, + next_hop_proto, next_hop_address, next_hop_sw_if_index=0xFFFFFFFF, table_id=0, @@ -982,7 +982,7 @@ class VppPapiProvider(object): 'mr_is_resolve_attached': is_resolve_attached, 'mr_is_interface_rx': is_interface_rx, 'mr_is_rpf_id': is_rpf_id, - 'mr_next_hop_proto_is_ip4': next_hop_proto_is_ip4, + 'mr_next_hop_proto': next_hop_proto, 'mr_next_hop_weight': next_hop_weight, 'mr_next_hop': next_hop_address, 'mr_next_hop_n_out_labels': next_hop_n_out_labels, -- cgit 1.2.3-korg