diff options
Diffstat (limited to 'src/vnet/vxlan')
| -rw-r--r-- | src/vnet/vxlan/decap.c | 1130 | ||||
| -rw-r--r-- | src/vnet/vxlan/dir.dox | 24 | ||||
| -rw-r--r-- | src/vnet/vxlan/encap.c | 553 | ||||
| -rw-r--r-- | src/vnet/vxlan/vxlan.api | 81 | ||||
| -rw-r--r-- | src/vnet/vxlan/vxlan.c | 899 | ||||
| -rw-r--r-- | src/vnet/vxlan/vxlan.h | 199 | ||||
| -rw-r--r-- | src/vnet/vxlan/vxlan_api.c | 253 | ||||
| -rw-r--r-- | src/vnet/vxlan/vxlan_error.def | 17 | ||||
| -rw-r--r-- | src/vnet/vxlan/vxlan_packet.h | 69 |
9 files changed, 3225 insertions, 0 deletions
diff --git a/src/vnet/vxlan/decap.c b/src/vnet/vxlan/decap.c new file mode 100644 index 00000000000..73e50ffe31f --- /dev/null +++ b/src/vnet/vxlan/decap.c @@ -0,0 +1,1130 @@ +/* + * decap.c: vxlan tunnel decap 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 <vlib/vlib.h> +#include <vnet/pg/pg.h> +#include <vnet/vxlan/vxlan.h> + +vlib_node_registration_t vxlan4_input_node; +vlib_node_registration_t vxlan6_input_node; + +typedef struct { + u32 next_index; + u32 tunnel_index; + u32 error; + u32 vni; +} vxlan_rx_trace_t; + +static u8 * format_vxlan_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_rx_trace_t * t = va_arg (*args, vxlan_rx_trace_t *); + + if (t->tunnel_index != ~0) + { + s = format (s, "VXLAN decap from vxlan_tunnel%d vni %d next %d error %d", + t->tunnel_index, t->vni, t->next_index, t->error); + } + else + { + s = format (s, "VXLAN decap error - tunnel for vni %d does not exist", + t->vni); + } + return s; +} + +always_inline uword +vxlan_input (vlib_main_t * vm, + vlib_node_runtime_t * node, + vlib_frame_t * from_frame, + u32 is_ip4) +{ + u32 n_left_from, next_index, * from, * to_next; + vxlan_main_t * vxm = &vxlan_main; + vnet_main_t * vnm = vxm->vnet_main; + vnet_interface_main_t * im = &vnm->interface_main; + u32 last_tunnel_index = ~0; + vxlan4_tunnel_key_t last_key4; + vxlan6_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) + last_key4.as_u64 = ~0; + 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_header_t * ip4_0, * ip4_1; + ip6_header_t * ip6_0, * ip6_1; + vxlan_header_t * vxlan0, * vxlan1; + uword * p0, * p1; + u32 tunnel_index0, tunnel_index1; + vxlan_tunnel_t * t0, * t1, * mt0 = NULL, * mt1 = NULL; + vxlan4_tunnel_key_t key4_0, key4_1; + vxlan6_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); + + /* udp leaves current_data pointing at the vxlan header */ + vxlan0 = vlib_buffer_get_current (b0); + vxlan1 = vlib_buffer_get_current (b1); + if (is_ip4) { + 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))); + ip4_0 = vlib_buffer_get_current (b0); + ip4_1 = vlib_buffer_get_current (b1); + } else { + 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))); + ip6_0 = vlib_buffer_get_current (b0); + ip6_1 = vlib_buffer_get_current (b1); + } + + /* pop (ip, udp, vxlan) */ + if (is_ip4) { + vlib_buffer_advance + (b0, sizeof(*ip4_0)+sizeof(udp_header_t)+sizeof(*vxlan0)); + vlib_buffer_advance + (b1, sizeof(*ip4_1)+sizeof(udp_header_t)+sizeof(*vxlan1)); + } else { + vlib_buffer_advance + (b0, sizeof(*ip6_0)+sizeof(udp_header_t)+sizeof(*vxlan0)); + vlib_buffer_advance + (b1, sizeof(*ip6_1)+sizeof(udp_header_t)+sizeof(*vxlan1)); + } + + tunnel_index0 = ~0; + error0 = 0; + + tunnel_index1 = ~0; + error1 = 0; + + if (PREDICT_FALSE (vxlan0->flags != VXLAN_FLAGS_I)) + { + error0 = VXLAN_ERROR_BAD_FLAGS; + next0 = VXLAN_INPUT_NEXT_DROP; + goto trace0; + } + + if (is_ip4) { + key4_0.src = ip4_0->src_address.as_u32; + key4_0.vni = vxlan0->vni_reserved; + + /* Make sure VXLAN tunnel exist according to packet SIP and VNI */ + if (PREDICT_FALSE (key4_0.as_u64 != last_key4.as_u64)) + { + p0 = hash_get (vxm->vxlan4_tunnel_by_key, key4_0.as_u64); + if (PREDICT_FALSE (p0 == NULL)) + { + error0 = VXLAN_ERROR_NO_SUCH_TUNNEL; + next0 = VXLAN_INPUT_NEXT_DROP; + goto trace0; + } + last_key4.as_u64 = key4_0.as_u64; + tunnel_index0 = last_tunnel_index = p0[0]; + } + else + tunnel_index0 = last_tunnel_index; + t0 = pool_elt_at_index (vxm->tunnels, tunnel_index0); + + /* Validate VXLAN tunnel SIP against packet DIP */ + if (PREDICT_TRUE (ip4_0->dst_address.as_u32 == t0->src.ip4.as_u32)) + goto next0; /* valid packet */ + if (PREDICT_FALSE (ip4_address_is_multicast (&ip4_0->dst_address))) + { + key4_0.src = ip4_0->dst_address.as_u32; + key4_0.vni = vxlan0->vni_reserved; + /* Make sure mcast VXLAN tunnel exist by packet DIP and VNI */ + p0 = hash_get (vxm->vxlan4_tunnel_by_key, key4_0.as_u64); + if (PREDICT_TRUE (p0 != NULL)) + { + mt0 = pool_elt_at_index (vxm->tunnels, p0[0]); + goto next0; /* valid packet */ + } + } + error0 = VXLAN_ERROR_NO_SUCH_TUNNEL; + next0 = VXLAN_INPUT_NEXT_DROP; + goto trace0; + + } else /* !is_ip4 */ { + key6_0.src.as_u64[0] = ip6_0->src_address.as_u64[0]; + key6_0.src.as_u64[1] = ip6_0->src_address.as_u64[1]; + key6_0.vni = vxlan0->vni_reserved; + + /* Make sure VXLAN tunnel exist according to packet SIP and VNI */ + if (PREDICT_FALSE (memcmp(&key6_0, &last_key6, sizeof(last_key6)) != 0)) + { + p0 = hash_get_mem (vxm->vxlan6_tunnel_by_key, &key6_0); + if (PREDICT_FALSE (p0 == NULL)) + { + error0 = VXLAN_ERROR_NO_SUCH_TUNNEL; + next0 = VXLAN_INPUT_NEXT_DROP; + goto trace0; + } + clib_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 (vxm->tunnels, tunnel_index0); + + /* Validate VXLAN tunnel SIP against packet DIP */ + if (PREDICT_TRUE (ip6_address_is_equal (&ip6_0->dst_address, + &t0->src.ip6))) + goto next0; /* valid packet */ + if (PREDICT_FALSE (ip6_address_is_multicast (&ip6_0->dst_address))) + { + key6_0.src.as_u64[0] = ip6_0->dst_address.as_u64[0]; + key6_0.src.as_u64[1] = ip6_0->dst_address.as_u64[1]; + key6_0.vni = vxlan0->vni_reserved; + p0 = hash_get (vxm->vxlan6_tunnel_by_key, &key6_0); + if (PREDICT_TRUE (p0 != NULL)) + { + mt0 = pool_elt_at_index (vxm->tunnels, p0[0]); + goto next0; /* valid packet */ + } + } + error0 = VXLAN_ERROR_NO_SUCH_TUNNEL; + next0 = VXLAN_INPUT_NEXT_DROP; + goto trace0; + } + + next0: + next0 = t0->decap_next_index; + 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 */ + if (PREDICT_TRUE(next0 == VXLAN_INPUT_NEXT_L2_INPUT)) + vnet_update_l2_len (b0); + + /* Set packet input sw_if_index to unicast VXLAN tunnel for learning */ + vnet_buffer(b0)->sw_if_index[VLIB_RX] = sw_if_index0; + sw_if_index0 = (mt0) ? mt0->sw_if_index : sw_if_index0; + + pkts_decapsulated ++; + stats_n_packets += 1; + stats_n_bytes += len0; + + /* Batch stats increment on the same vxlan 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; + } + + trace0: + b0->error = error0 ? node->errors[error0] : 0; + + if (PREDICT_FALSE(b0->flags & VLIB_BUFFER_IS_TRACED)) + { + vxlan_rx_trace_t *tr + = vlib_add_trace (vm, node, b0, sizeof (*tr)); + tr->next_index = next0; + tr->error = error0; + tr->tunnel_index = tunnel_index0; + tr->vni = vnet_get_vni (vxlan0); + } + + if (PREDICT_FALSE (vxlan1->flags != VXLAN_FLAGS_I)) + { + error1 = VXLAN_ERROR_BAD_FLAGS; + next1 = VXLAN_INPUT_NEXT_DROP; + goto trace1; + } + + if (is_ip4) { + key4_1.src = ip4_1->src_address.as_u32; + key4_1.vni = vxlan1->vni_reserved; + + /* Make sure unicast VXLAN tunnel exist by packet SIP and VNI */ + if (PREDICT_FALSE (key4_1.as_u64 != last_key4.as_u64)) + { + p1 = hash_get (vxm->vxlan4_tunnel_by_key, key4_1.as_u64); + if (PREDICT_FALSE (p1 == NULL)) + { + error1 = VXLAN_ERROR_NO_SUCH_TUNNEL; + next1 = VXLAN_INPUT_NEXT_DROP; + goto trace1; + } + last_key4.as_u64 = key4_1.as_u64; + tunnel_index1 = last_tunnel_index = p1[0]; + } + else + tunnel_index1 = last_tunnel_index; + t1 = pool_elt_at_index (vxm->tunnels, tunnel_index1); + + /* Validate VXLAN tunnel SIP against packet DIP */ + if (PREDICT_TRUE (ip4_1->dst_address.as_u32 == t1->src.ip4.as_u32)) + goto next1; /* valid packet */ + if (PREDICT_FALSE (ip4_address_is_multicast (&ip4_1->dst_address))) + { + key4_1.src = ip4_1->dst_address.as_u32; + key4_1.vni = vxlan1->vni_reserved; + /* Make sure mcast VXLAN tunnel exist by packet DIP and VNI */ + p1 = hash_get (vxm->vxlan4_tunnel_by_key, key4_1.as_u64); + if (PREDICT_TRUE (p1 != NULL)) + { + mt1 = pool_elt_at_index (vxm->tunnels, p1[0]); + goto next1; /* valid packet */ + } + } + error1 = VXLAN_ERROR_NO_SUCH_TUNNEL; + next1 = VXLAN_INPUT_NEXT_DROP; + goto trace1; + + } else /* !is_ip4 */ { + key6_1.src.as_u64[0] = ip6_1->src_address.as_u64[0]; + key6_1.src.as_u64[1] = ip6_1->src_address.as_u64[1]; + key6_1.vni = vxlan1->vni_reserved; + + /* Make sure VXLAN tunnel exist according to packet SIP and VNI */ + if (PREDICT_FALSE (memcmp(&key6_1, &last_key6, sizeof(last_key6)) != 0)) + { + p1 = hash_get_mem (vxm->vxlan6_tunnel_by_key, &key6_1); + + if (PREDICT_FALSE (p1 == NULL)) + { + error1 = VXLAN_ERROR_NO_SUCH_TUNNEL; + next1 = VXLAN_INPUT_NEXT_DROP; + goto trace1; + } + + clib_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 (vxm->tunnels, tunnel_index1); + + /* Validate VXLAN tunnel SIP against packet DIP */ + if (PREDICT_TRUE (ip6_address_is_equal (&ip6_1->dst_address, + &t1->src.ip6))) + goto next1; /* valid packet */ + if (PREDICT_FALSE (ip6_address_is_multicast (&ip6_1->dst_address))) + { + key6_1.src.as_u64[0] = ip6_1->dst_address.as_u64[0]; + key6_1.src.as_u64[1] = ip6_1->dst_address.as_u64[1]; + key6_1.vni = vxlan1->vni_reserved; + p1 = hash_get (vxm->vxlan6_tunnel_by_key, &key6_1); + if (PREDICT_TRUE (p1 != NULL)) + { + mt1 = pool_elt_at_index (vxm->tunnels, p1[0]); + goto next1; /* valid packet */ + } + } + error1 = VXLAN_ERROR_NO_SUCH_TUNNEL; + next1 = VXLAN_INPUT_NEXT_DROP; + goto trace1; + } + + next1: + next1 = t1->decap_next_index; + 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 */ + if (PREDICT_TRUE(next1 == VXLAN_INPUT_NEXT_L2_INPUT)) + vnet_update_l2_len (b1); + + /* Set packet input sw_if_index to unicast VXLAN tunnel for learning */ + vnet_buffer(b1)->sw_if_index[VLIB_RX] = sw_if_index1; + sw_if_index1 = (mt1) ? mt1->sw_if_index : sw_if_index1; + + 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; + } + + trace1: + b1->error = error1 ? node->errors[error1] : 0; + + if (PREDICT_FALSE(b1->flags & VLIB_BUFFER_IS_TRACED)) + { + vxlan_rx_trace_t *tr + = vlib_add_trace (vm, node, b1, sizeof (*tr)); + tr->next_index = next1; + tr->error = error1; + tr->tunnel_index = tunnel_index1; + tr->vni = vnet_get_vni (vxlan1); + } + + 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_header_t * ip4_0; + ip6_header_t * ip6_0; + vxlan_header_t * vxlan0; + uword * p0; + u32 tunnel_index0; + vxlan_tunnel_t * t0, * mt0 = NULL; + vxlan4_tunnel_key_t key4_0; + vxlan6_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); + + /* udp leaves current_data pointing at the vxlan header */ + vxlan0 = vlib_buffer_get_current (b0); + if (is_ip4) { + vlib_buffer_advance + (b0, -(word)(sizeof(udp_header_t)+sizeof(ip4_header_t))); + ip4_0 = vlib_buffer_get_current (b0); + } else { + vlib_buffer_advance + (b0, -(word)(sizeof(udp_header_t)+sizeof(ip6_header_t))); + ip6_0 = vlib_buffer_get_current (b0); + } + + /* pop (ip, udp, vxlan) */ + if (is_ip4) { + vlib_buffer_advance + (b0, sizeof(*ip4_0)+sizeof(udp_header_t)+sizeof(*vxlan0)); + } else { + vlib_buffer_advance + (b0, sizeof(*ip6_0)+sizeof(udp_header_t)+sizeof(*vxlan0)); + } + + tunnel_index0 = ~0; + error0 = 0; + + if (PREDICT_FALSE (vxlan0->flags != VXLAN_FLAGS_I)) + { + error0 = VXLAN_ERROR_BAD_FLAGS; + next0 = VXLAN_INPUT_NEXT_DROP; + goto trace00; + } + + if (is_ip4) { + key4_0.src = ip4_0->src_address.as_u32; + key4_0.vni = vxlan0->vni_reserved; + + /* Make sure unicast VXLAN tunnel exist by packet SIP and VNI */ + if (PREDICT_FALSE (key4_0.as_u64 != last_key4.as_u64)) + { + p0 = hash_get (vxm->vxlan4_tunnel_by_key, key4_0.as_u64); + if (PREDICT_FALSE (p0 == NULL)) + { + error0 = VXLAN_ERROR_NO_SUCH_TUNNEL; + next0 = VXLAN_INPUT_NEXT_DROP; + goto trace00; + } + last_key4.as_u64 = key4_0.as_u64; + tunnel_index0 = last_tunnel_index = p0[0]; + } + else + tunnel_index0 = last_tunnel_index; + t0 = pool_elt_at_index (vxm->tunnels, tunnel_index0); + + /* Validate VXLAN tunnel SIP against packet DIP */ + if (PREDICT_TRUE (ip4_0->dst_address.as_u32 == t0->src.ip4.as_u32)) + goto next00; /* valid packet */ + if (PREDICT_FALSE (ip4_address_is_multicast (&ip4_0->dst_address))) + { + key4_0.src = ip4_0->dst_address.as_u32; + key4_0.vni = vxlan0->vni_reserved; + /* Make sure mcast VXLAN tunnel exist by packet DIP and VNI */ + p0 = hash_get (vxm->vxlan4_tunnel_by_key, key4_0.as_u64); + if (PREDICT_TRUE (p0 != NULL)) + { + mt0 = pool_elt_at_index (vxm->tunnels, p0[0]); + goto next00; /* valid packet */ + } + } + error0 = VXLAN_ERROR_NO_SUCH_TUNNEL; + next0 = VXLAN_INPUT_NEXT_DROP; + goto trace00; + + } else /* !is_ip4 */ { + key6_0.src.as_u64[0] = ip6_0->src_address.as_u64[0]; + key6_0.src.as_u64[1] = ip6_0->src_address.as_u64[1]; + key6_0.vni = vxlan0->vni_reserved; + + /* Make sure VXLAN tunnel exist according to packet SIP and VNI */ + if (PREDICT_FALSE (memcmp(&key6_0, &last_key6, sizeof(last_key6)) != 0)) + { + p0 = hash_get_mem (vxm->vxlan6_tunnel_by_key, &key6_0); + if (PREDICT_FALSE (p0 == NULL)) + { + error0 = VXLAN_ERROR_NO_SUCH_TUNNEL; + next0 = VXLAN_INPUT_NEXT_DROP; + goto trace00; + } + clib_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 (vxm->tunnels, tunnel_index0); + + /* Validate VXLAN tunnel SIP against packet DIP */ + if (PREDICT_TRUE (ip6_address_is_equal (&ip6_0->dst_address, + &t0->src.ip6))) + goto next00; /* valid packet */ + if (PREDICT_FALSE (ip6_address_is_multicast (&ip6_0->dst_address))) + { + key6_0.src.as_u64[0] = ip6_0->dst_address.as_u64[0]; + key6_0.src.as_u64[1] = ip6_0->dst_address.as_u64[1]; + key6_0.vni = vxlan0->vni_reserved; + p0 = hash_get (vxm->vxlan6_tunnel_by_key, &key6_0); + if (PREDICT_TRUE (p0 != NULL)) + { + mt0 = pool_elt_at_index (vxm->tunnels, p0[0]); + goto next00; /* valid packet */ + } + } + error0 = VXLAN_ERROR_NO_SUCH_TUNNEL; + next0 = VXLAN_INPUT_NEXT_DROP; + goto trace00; + } + + next00: + next0 = t0->decap_next_index; + 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 */ + if (PREDICT_TRUE(next0 == VXLAN_INPUT_NEXT_L2_INPUT)) + vnet_update_l2_len (b0); + + /* Set packet input sw_if_index to unicast VXLAN tunnel for learning */ + vnet_buffer(b0)->sw_if_index[VLIB_RX] = sw_if_index0; + sw_if_index0 = (mt0) ? mt0->sw_if_index : sw_if_index0; + + pkts_decapsulated ++; + stats_n_packets += 1; + stats_n_bytes += len0; + + /* Batch stats increment on the same vxlan 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_rx_trace_t *tr + = vlib_add_trace (vm, node, b0, sizeof (*tr)); + tr->next_index = next0; + tr->error = error0; + tr->tunnel_index = tunnel_index0; + tr->vni = vnet_get_vni (vxlan0); + } + 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); + } + /* Do we still need this now that tunnel tx stats is kept? */ + vlib_node_increment_counter (vm, is_ip4? + vxlan4_input_node.index:vxlan6_input_node.index, + VXLAN_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; +} + +static uword +vxlan4_input (vlib_main_t * vm, + vlib_node_runtime_t * node, + vlib_frame_t * from_frame) +{ + return vxlan_input(vm, node, from_frame, /* is_ip4 */ 1); +} + +static uword +vxlan6_input (vlib_main_t * vm, + vlib_node_runtime_t * node, + vlib_frame_t * from_frame) +{ + return vxlan_input(vm, node, from_frame, /* is_ip4 */ 0); +} + +static char * vxlan_error_strings[] = { +#define vxlan_error(n,s) s, +#include <vnet/vxlan/vxlan_error.def> +#undef vxlan_error +#undef _ +}; + +VLIB_REGISTER_NODE (vxlan4_input_node) = { + .function = vxlan4_input, + .name = "vxlan4-input", + /* Takes a vector of packets. */ + .vector_size = sizeof (u32), + + .n_errors = VXLAN_N_ERROR, + .error_strings = vxlan_error_strings, + + .n_next_nodes = VXLAN_INPUT_N_NEXT, + .next_nodes = { +#define _(s,n) [VXLAN_INPUT_NEXT_##s] = n, + foreach_vxlan_input_next +#undef _ + }, + +//temp .format_buffer = format_vxlan_header, + .format_trace = format_vxlan_rx_trace, + // $$$$ .unformat_buffer = unformat_vxlan_header, +}; + +VLIB_NODE_FUNCTION_MULTIARCH (vxlan4_input_node, vxlan4_input) + +VLIB_REGISTER_NODE (vxlan6_input_node) = { + .function = vxlan6_input, + .name = "vxlan6-input", + /* Takes a vector of packets. */ + .vector_size = sizeof (u32), + + .n_errors = VXLAN_N_ERROR, + .error_strings = vxlan_error_strings, + + .n_next_nodes = VXLAN_INPUT_N_NEXT, + .next_nodes = { +#define _(s,n) [VXLAN_INPUT_NEXT_##s] = n, + foreach_vxlan_input_next +#undef _ + }, + +//temp .format_buffer = format_vxlan_header, + .format_trace = format_vxlan_rx_trace, + // $$$$ .unformat_buffer = unformat_vxlan_header, +}; + +VLIB_NODE_FUNCTION_MULTIARCH (vxlan6_input_node, vxlan6_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_bypass_inline (vlib_main_t * vm, + vlib_node_runtime_t * node, + vlib_frame_t * frame, + u32 is_ip4) +{ + vxlan_main_t * vxm = &vxlan_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 * ip0, * ip1; + 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); + ip0 = vlib_buffer_get_current (b0); + ip1 = 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); + + /* Treat IP frag packets as "experimental" protocol for now + until support of IP frag reassembly is implemented */ + proto0 = ip4_is_fragment(ip0) ? 0xfe : ip0->protocol; + proto1 = ip4_is_fragment(ip1) ? 0xfe : ip1->protocol; + + /* Process packet 0 */ + if (proto0 != IP_PROTOCOL_UDP) + goto exit0; /* not UDP packet */ + + udp0 = ip4_next_header (ip0); + if (udp0->dst_port != clib_host_to_net_u16 (UDP_DST_PORT_vxlan)) + goto exit0; /* not VXLAN packet */ + + if (is_ip4) + { + if (addr4.as_u32 != ip0->dst_address.as_u32) + { + if (!hash_get (vxm->vtep4, ip0->dst_address.as_u32)) + goto exit0; /* no local VTEP for VXLAN packet */ + addr4 = ip0->dst_address; + } + } + else goto exit0; /* IP6 VXLAN bypass not yet supported */ + + /* vxlan-input node expect current at VXLAN header */ + vlib_buffer_advance (b0, sizeof(ip4_header_t)+sizeof(udp_header_t)); + + 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 */ + ip_len0 = clib_net_to_host_u16 (ip0->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)) + flags0 = ip4_tcp_udp_validate_checksum (vm, b0); + good_udp0 = + (flags0 & IP_BUFFER_L4_CHECKSUM_CORRECT) != 0; + } + + error0 = good_udp0 ? 0 : IP4_ERROR_UDP_CHECKSUM; + error0 = (len_diff0 >= 0) ? error0 : IP4_ERROR_UDP_LENGTH; + + next0 = error0 ? + IP_VXLAN_BYPASS_NEXT_DROP : IP_VXLAN_BYPASS_NEXT_VXLAN; + b0->error = error0 ? error_node->errors[error0] : 0; + + exit0: + /* Process packet 1 */ + if (proto1 != IP_PROTOCOL_UDP) + goto exit1; /* not UDP packet */ + + udp1 = ip4_next_header (ip1); + if (udp1->dst_port != clib_host_to_net_u16 (UDP_DST_PORT_vxlan)) + goto exit1; /* not VXLAN packet */ + + if (is_ip4) + { + if (addr4.as_u32 != ip1->dst_address.as_u32) + { + if (!hash_get (vxm->vtep4, ip1->dst_address.as_u32)) + goto exit1; /* no local VTEP for VXLAN packet */ + addr4 = ip1->dst_address; + } + } + else goto exit1; /* IP6 VXLAN bypass not yet supported */ + + /* vxlan-input node expect current at VXLAN header */ + vlib_buffer_advance (b1, sizeof(ip4_header_t)+sizeof(udp_header_t)); + + 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 */ + ip_len1 = clib_net_to_host_u16 (ip1->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)) + flags1 = ip4_tcp_udp_validate_checksum (vm, b1); + good_udp1 = + (flags1 & IP_BUFFER_L4_CHECKSUM_CORRECT) != 0; + } + + error1 = good_udp1 ? 0 : IP4_ERROR_UDP_CHECKSUM; + error1 = (len_diff1 >= 0) ? error1 : IP4_ERROR_UDP_LENGTH; + + next1 = error1 ? + IP_VXLAN_BYPASS_NEXT_DROP : IP_VXLAN_BYPASS_NEXT_VXLAN; + b1->error = error1 ? error_node->errors[error1] : 0; + + 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 * ip0; + 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); + ip0 = vlib_buffer_get_current (b0); + + /* Setup packet for next IP feature */ + vnet_feature_next(vnet_buffer(b0)->sw_if_index[VLIB_RX], &next0, b0); + + /* 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 (proto0 != IP_PROTOCOL_UDP) + goto exit; /* not UDP packet */ + + udp0 = ip4_next_header (ip0); + if (udp0->dst_port != clib_host_to_net_u16 (UDP_DST_PORT_vxlan)) + goto exit; /* not VXLAN packet */ + + if (is_ip4) + { + if (addr4.as_u32 != ip0->dst_address.as_u32) + { + if (!hash_get (vxm->vtep4, ip0->dst_address.as_u32)) + goto exit; /* no local VTEP for VXLAN packet */ + addr4 = ip0->dst_address; + } + } + else goto exit; /* IP6 VXLAN bypass not yet supported */ + + /* vxlan-input node expect current at VXLAN header */ + vlib_buffer_advance (b0, sizeof(ip4_header_t)+sizeof(udp_header_t)); + + 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 */ + ip_len0 = clib_net_to_host_u16 (ip0->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)) + flags0 = ip4_tcp_udp_validate_checksum (vm, b0); + good_udp0 = + (flags0 & IP_BUFFER_L4_CHECKSUM_CORRECT) != 0; + } + + error0 = good_udp0 ? 0 : IP4_ERROR_UDP_CHECKSUM; + error0 = (len_diff0 >= 0) ? error0 : IP4_ERROR_UDP_LENGTH; + + next0 = error0 ? + IP_VXLAN_BYPASS_NEXT_DROP : IP_VXLAN_BYPASS_NEXT_VXLAN; + b0->error = error0 ? error_node->errors[error0] : 0; + + 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_bypass (vlib_main_t * vm, + vlib_node_runtime_t * node, + vlib_frame_t * frame) +{ + return ip_vxlan_bypass_inline (vm, node, frame, /* is_ip4 */ 1); +} + +VLIB_REGISTER_NODE (ip4_vxlan_bypass_node) = { + .function = ip4_vxlan_bypass, + .name = "ip4-vxlan-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-input", + }, + + .format_buffer = format_ip4_header, + .format_trace = format_ip4_forward_next_trace, +}; + +VLIB_NODE_FUNCTION_MULTIARCH (ip4_vxlan_bypass_node,ip4_vxlan_bypass) + + +static clib_error_t * +set_ip_vxlan_bypass (vlib_main_t * vm, + unformat_input_t * input, + vlib_cli_command_t * cmd) +{ + unformat_input_t _line_input, * line_input = &_line_input; + vnet_main_t * vnm = vnet_get_main(); + clib_error_t * error = 0; + u32 sw_if_index, is_del; + + sw_if_index = ~0; + is_del = 0; + + 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_del = 1; + 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_feature_enable_disable ("ip4-unicast", "ip4-vxlan-bypass", sw_if_index, + is_del == 0, 0, 0); + done: + return error; +} + +/*? + * This command adds the 'ip4-vxlan-bypass' graph node for a given interface. + * By adding the IPv4 vxlan-bypass graph node to an interface, the node checks + * for and validate input vxlan packet and bypass ip4-lookup, ip4-local, + * ip4-udp-lookup nodes to speedup vxlan packet forwarding. This node will + * cause extra overhead to for non-vxlan packets which is kept at a minimum. + * + * @cliexpar + * @parblock + * Example of graph node before ip4-vxlan-bypass is enabled: + * @cliexstart{show vlib graph ip4-vxlan-bypass} + * Name Next Previous + * ip4-vxlan-bypass error-drop [0] + * vxlan4-input [1] + * ip4-lookup [2] + * @cliexend + * + * Example of how to enable ip4-vxlan-bypass on an interface: + * @cliexcmd{set interface ip vxlan-bypass GigabitEthernet2/0/0} + * + * Example of graph node after ip4-vxlan-bypass is enabled: + * @cliexstart{show vlib graph ip4-vxlan-bypass} + * Name Next Previous + * ip4-vxlan-bypass error-drop [0] ip4-input + * vxlan4-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-bypass + * ip4-lookup + * + * ipv4 multicast: + * ip4-lookup-multicast + * + * ipv4 multicast: + * interface-output + * + * ipv6 unicast: + * ip6-lookup + * + * ipv6 multicast: + * ip6-lookup + * + * ipv6 multicast: + * interface-output + * @cliexend + * + * Example of how to disable unicast source checking on an interface: + * @cliexcmd{set interface ip vxlan-bypass GigabitEthernet2/0/0 del} + * @endparblock +?*/ +/* *INDENT-OFF* */ +VLIB_CLI_COMMAND (set_interface_ip_vxlan_bypass_command, static) = { + .path = "set interface ip vxlan-bypass", + .function = set_ip_vxlan_bypass, + .short_help = "set interface ip vxlan-bypass <interface> [del]", +}; + +/* Dummy init function to get us linked in. */ +clib_error_t * ip4_vxlan_bypass_init (vlib_main_t * vm) +{ return 0; } + +VLIB_INIT_FUNCTION (ip4_vxlan_bypass_init); diff --git a/src/vnet/vxlan/dir.dox b/src/vnet/vxlan/dir.dox new file mode 100644 index 00000000000..31a9e2b6112 --- /dev/null +++ b/src/vnet/vxlan/dir.dox @@ -0,0 +1,24 @@ +/* + * Copyright (c) 2016 Cisco and/or its affiliates. + * + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +/** +@dir +@brief VXLAN Code. + +This directory contains source code to support VXLAN. + +*/ +/*? %%clicmd:group_label VXLAN CLI %% ?*/ diff --git a/src/vnet/vxlan/encap.c b/src/vnet/vxlan/encap.c new file mode 100644 index 00000000000..5b63064a848 --- /dev/null +++ b/src/vnet/vxlan/encap.c @@ -0,0 +1,553 @@ +/* + * Copyright (c) 2015 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ +#include <vppinfra/error.h> +#include <vppinfra/hash.h> +#include <vnet/vnet.h> +#include <vnet/ip/ip.h> +#include <vnet/ethernet/ethernet.h> +#include <vnet/vxlan/vxlan.h> + +/* Statistics (not all errors) */ +#define foreach_vxlan_encap_error \ +_(ENCAPSULATED, "good packets encapsulated") + +static char * vxlan_encap_error_strings[] = { +#define _(sym,string) string, + foreach_vxlan_encap_error +#undef _ +}; + +typedef enum { +#define _(sym,str) VXLAN_ENCAP_ERROR_##sym, + foreach_vxlan_encap_error +#undef _ + VXLAN_ENCAP_N_ERROR, +} vxlan_encap_error_t; + +typedef enum { + VXLAN_ENCAP_NEXT_DROP, + VXLAN_ENCAP_N_NEXT, +} vxlan_encap_next_t; + +typedef struct { + u32 tunnel_index; + u32 vni; +} vxlan_encap_trace_t; + +u8 * format_vxlan_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_encap_trace_t * t + = va_arg (*args, vxlan_encap_trace_t *); + + s = format (s, "VXLAN encap to vxlan_tunnel%d vni %d", + t->tunnel_index, t->vni); + return s; +} + + +#define foreach_fixed_header4_offset \ + _(0) _(1) _(2) _(3) + +#define foreach_fixed_header6_offset \ + _(0) _(1) _(2) _(3) _(4) _(5) _(6) + +always_inline uword +vxlan_encap_inline (vlib_main_t * vm, + vlib_node_runtime_t * node, + vlib_frame_t * from_frame, + u32 is_ip4) +{ + u32 n_left_from, next_index, * from, * to_next; + vxlan_main_t * vxm = &vxlan_main; + vnet_main_t * vnm = vxm->vnet_main; + 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 stats_sw_if_index, stats_n_packets, stats_n_bytes; + u32 sw_if_index0 = 0, sw_if_index1 = 0; + u32 next0 = 0, next1 = 0; + vnet_hw_interface_t * hi0, * hi1; + vxlan_tunnel_t * t0 = NULL, * t1 = NULL; + + 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 flow_hash0, flow_hash1; + u32 len0, len1; + ip4_header_t * ip4_0, * ip4_1; + ip6_header_t * ip6_0, * ip6_1; + udp_header_t * udp0, * udp1; + u64 * copy_src0, * copy_dst0; + u64 * copy_src1, * copy_dst1; + u32 * copy_src_last0, * copy_dst_last0; + u32 * copy_src_last1, * copy_dst_last1; + u16 new_l0, new_l1; + ip_csum_t sum0, sum1; + + /* 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); + + flow_hash0 = vnet_l2_compute_flow_hash (b0); + flow_hash1 = vnet_l2_compute_flow_hash (b1); + + /* Get next node index and adj index from tunnel next_dpo */ + if (sw_if_index0 != vnet_buffer(b0)->sw_if_index[VLIB_TX]) + { + sw_if_index0 = vnet_buffer(b0)->sw_if_index[VLIB_TX]; + hi0 = vnet_get_sup_hw_interface (vnm, sw_if_index0); + t0 = &vxm->tunnels[hi0->dev_instance]; + /* Note: change to always set next0 if it may be set to drop */ + next0 = t0->next_dpo.dpoi_next_node; + } + vnet_buffer(b0)->ip.adj_index[VLIB_TX] = t0->next_dpo.dpoi_index; + + /* Get next node index and adj index from tunnel next_dpo */ + if (sw_if_index1 != vnet_buffer(b1)->sw_if_index[VLIB_TX]) + { + sw_if_index1 = vnet_buffer(b1)->sw_if_index[VLIB_TX]; + hi1 = vnet_get_sup_hw_interface (vnm, sw_if_index1); + t1 = &vxm->tunnels[hi1->dev_instance]; + /* Note: change to always set next1 if it may be set to drop */ + next1 = t1->next_dpo.dpoi_next_node; + } + vnet_buffer(b1)->ip.adj_index[VLIB_TX] = t1->next_dpo.dpoi_index; + + /* Apply the rewrite string. $$$$ vnet_rewrite? */ + vlib_buffer_advance (b0, -(word)_vec_len(t0->rewrite)); + vlib_buffer_advance (b1, -(word)_vec_len(t1->rewrite)); + + if (is_ip4) + { + /* IP4 VXLAN header should be 36 octects */ + ASSERT(sizeof(ip4_vxlan_header_t) == 36); + ASSERT(vec_len(t0->rewrite) == sizeof(ip4_vxlan_header_t)); + ASSERT(vec_len(t1->rewrite) == sizeof(ip4_vxlan_header_t)); + + ip4_0 = vlib_buffer_get_current(b0); + ip4_1 = vlib_buffer_get_current(b1); + + /* Copy the fixed header */ + copy_dst0 = (u64 *) ip4_0; + copy_src0 = (u64 *) t0->rewrite; + copy_dst1 = (u64 *) ip4_1; + copy_src1 = (u64 *) t1->rewrite; + /* Copy first 32 octets 8-bytes at a time */ +#define _(offs) copy_dst0[offs] = copy_src0[offs]; + foreach_fixed_header4_offset; +#undef _ +#define _(offs) copy_dst1[offs] = copy_src1[offs]; + foreach_fixed_header4_offset; +#undef _ + /* Last 4 octets. Hopefully gcc will be our friend */ + copy_dst_last0 = (u32 *)(©_dst0[4]); + copy_src_last0 = (u32 *)(©_src0[4]); + copy_dst_last0[0] = copy_src_last0[0]; + copy_dst_last1 = (u32 *)(©_dst1[4]); + copy_src_last1 = (u32 *)(©_src1[4]); + copy_dst_last1[0] = copy_src_last1[0]; + + /* Fix the IP4 checksum and length */ + sum0 = ip4_0->checksum; + new_l0 = /* old_l0 always 0, see the rewrite setup */ + 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 */); + ip4_0->checksum = ip_csum_fold (sum0); + ip4_0->length = new_l0; + sum1 = ip4_1->checksum; + new_l1 = /* old_l1 always 0, see the rewrite setup */ + clib_host_to_net_u16 (vlib_buffer_length_in_chain (vm, b1)); + sum1 = ip_csum_update (sum1, old_l1, new_l1, ip4_header_t, + length /* changed member */); + ip4_1->checksum = ip_csum_fold (sum1); + ip4_1->length = new_l1; + + /* Fix UDP length and set source port */ + udp0 = (udp_header_t *)(ip4_0+1); + new_l0 = clib_host_to_net_u16 (vlib_buffer_length_in_chain(vm, b0) + - sizeof (*ip4_0)); + udp0->length = new_l0; + udp0->src_port = flow_hash0; + udp1 = (udp_header_t *)(ip4_1+1); + new_l1 = clib_host_to_net_u16 (vlib_buffer_length_in_chain(vm, b1) + - sizeof (*ip4_1)); + udp1->length = new_l1; + udp1->src_port = flow_hash1; + } + else /* ipv6 */ + { + int bogus = 0; + + /* IP6 VXLAN header should be 56 octects */ + ASSERT(sizeof(ip6_vxlan_header_t) == 56); + ASSERT(vec_len(t0->rewrite) == sizeof(ip6_vxlan_header_t)); + ASSERT(vec_len(t1->rewrite) == sizeof(ip6_vxlan_header_t)); + ip6_0 = vlib_buffer_get_current(b0); + ip6_1 = vlib_buffer_get_current(b1); + + /* Copy the fixed header */ + copy_dst0 = (u64 *) ip6_0; + copy_src0 = (u64 *) t0->rewrite; + copy_dst1 = (u64 *) ip6_1; + copy_src1 = (u64 *) t1->rewrite; + /* Copy first 56 (ip6) octets 8-bytes at a time */ +#define _(offs) copy_dst0[offs] = copy_src0[offs]; + foreach_fixed_header6_offset; +#undef _ +#define _(offs) copy_dst1[offs] = copy_src1[offs]; + foreach_fixed_header6_offset; +#undef _ + /* Fix IP6 payload length */ + new_l0 = + clib_host_to_net_u16 (vlib_buffer_length_in_chain (vm, b0) + - sizeof(*ip6_0)); + ip6_0->payload_length = new_l0; + new_l1 = + clib_host_to_net_u16 (vlib_buffer_length_in_chain (vm, b1) + - sizeof(*ip6_1)); + ip6_1->payload_length = new_l1; + + /* Fix UDP length and set source port */ + udp0 = (udp_header_t *)(ip6_0+1); + udp0->length = new_l0; + udp0->src_port = flow_hash0; + udp1 = (udp_header_t *)(ip6_1+1); + udp1->length = new_l1; + udp1->src_port = flow_hash1; + + /* IPv6 UDP checksum is mandatory */ + udp0->checksum = ip6_tcp_udp_icmp_compute_checksum(vm, b0, + ip6_0, &bogus); + ASSERT(bogus == 0); + if (udp0->checksum == 0) + udp0->checksum = 0xffff; + udp1->checksum = ip6_tcp_udp_icmp_compute_checksum(vm, b1, + ip6_1, &bogus); + ASSERT(bogus == 0); + if (udp1->checksum == 0) + udp1->checksum = 0xffff; + } + + pkts_encapsulated += 2; + len0 = vlib_buffer_length_in_chain (vm, b0); + len1 = vlib_buffer_length_in_chain (vm, b1); + 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_encap_trace_t *tr = + vlib_add_trace (vm, node, b0, sizeof (*tr)); + tr->tunnel_index = t0 - vxm->tunnels; + tr->vni = t0->vni; + } + + if (PREDICT_FALSE(b1->flags & VLIB_BUFFER_IS_TRACED)) + { + vxlan_encap_trace_t *tr = + vlib_add_trace (vm, node, b1, sizeof (*tr)); + tr->tunnel_index = t1 - vxm->tunnels; + tr->vni = t1->vni; + } + + 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 flow_hash0; + u32 len0; + ip4_header_t * ip4_0; + ip6_header_t * ip6_0; + udp_header_t * udp0; + u64 * copy_src0, * copy_dst0; + u32 * copy_src_last0, * copy_dst_last0; + u16 new_l0; + ip_csum_t sum0; + + 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); + + flow_hash0 = vnet_l2_compute_flow_hash(b0); + + /* Get next node index and adj index from tunnel next_dpo */ + if (sw_if_index0 != vnet_buffer(b0)->sw_if_index[VLIB_TX]) + { + sw_if_index0 = vnet_buffer(b0)->sw_if_index[VLIB_TX]; + hi0 = vnet_get_sup_hw_interface (vnm, sw_if_index0); + t0 = &vxm->tunnels[hi0->dev_instance]; + /* Note: change to always set next0 if it may be set to drop */ + next0 = t0->next_dpo.dpoi_next_node; + } + vnet_buffer(b0)->ip.adj_index[VLIB_TX] = t0->next_dpo.dpoi_index; + + /* Apply the rewrite string. $$$$ vnet_rewrite? */ + vlib_buffer_advance (b0, -(word)_vec_len(t0->rewrite)); + + if (is_ip4) + { + /* IP4 VXLAN header should be 36 octects */ + ASSERT(sizeof(ip4_vxlan_header_t) == 36); + ASSERT(vec_len(t0->rewrite) == sizeof(ip4_vxlan_header_t)); + ip4_0 = vlib_buffer_get_current(b0); + + /* Copy the fixed header */ + copy_dst0 = (u64 *) ip4_0; + copy_src0 = (u64 *) t0->rewrite; + /* Copy first 32 octets 8-bytes at a time */ +#define _(offs) copy_dst0[offs] = copy_src0[offs]; + foreach_fixed_header4_offset; +#undef _ + /* Last 4 octets. Hopefully gcc will be our friend */ + copy_dst_last0 = (u32 *)(©_dst0[4]); + copy_src_last0 = (u32 *)(©_src0[4]); + copy_dst_last0[0] = copy_src_last0[0]; + + /* Fix the IP4 checksum and length */ + sum0 = ip4_0->checksum; + new_l0 = /* old_l0 always 0, see the rewrite setup */ + 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 */); + ip4_0->checksum = ip_csum_fold (sum0); + ip4_0->length = new_l0; + + /* Fix UDP length and set source port */ + udp0 = (udp_header_t *)(ip4_0+1); + new_l0 = clib_host_to_net_u16 (vlib_buffer_length_in_chain(vm, b0) + - sizeof (*ip4_0)); + udp0->length = new_l0; + udp0->src_port = flow_hash0; + } + + else /* ip6 path */ + { + int bogus = 0; + + /* IP6 VXLAN header should be 56 octects */ + ASSERT(sizeof(ip6_vxlan_header_t) == 56); + ASSERT(vec_len(t0->rewrite) == sizeof(ip6_vxlan_header_t)); + ip6_0 = vlib_buffer_get_current(b0); + /* Copy the fixed header */ + copy_dst0 = (u64 *) ip6_0; + copy_src0 = (u64 *) t0->rewrite; + /* Copy first 56 (ip6) octets 8-bytes at a time */ +#define _(offs) copy_dst0[offs] = copy_src0[offs]; + foreach_fixed_header6_offset; +#undef _ + /* Fix IP6 payload length */ + new_l0 = + clib_host_to_net_u16 (vlib_buffer_length_in_chain (vm, b0) + - sizeof(*ip6_0)); + ip6_0->payload_length = new_l0; + + /* Fix UDP length and set source port */ + udp0 = (udp_header_t *)(ip6_0+1); + udp0->length = new_l0; + udp0->src_port = flow_hash0; + + /* IPv6 UDP checksum is mandatory */ + udp0->checksum = ip6_tcp_udp_icmp_compute_checksum(vm, b0, + ip6_0, &bogus); + ASSERT(bogus == 0); + if (udp0->checksum == 0) + udp0->checksum = 0xffff; + } + + 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_encap_trace_t *tr = + vlib_add_trace (vm, node, b0, sizeof (*tr)); + tr->tunnel_index = t0 - vxm->tunnels; + tr->vni = t0->vni; + } + 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); + } + + /* Do we still need this now that tunnel tx stats is kept? */ + vlib_node_increment_counter (vm, node->node_index, + VXLAN_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; +} + +static uword +vxlan4_encap (vlib_main_t * vm, + vlib_node_runtime_t * node, + vlib_frame_t * from_frame) +{ + return vxlan_encap_inline (vm, node, from_frame, /* is_ip4 */ 1); +} + +static uword +vxlan6_encap (vlib_main_t * vm, + vlib_node_runtime_t * node, + vlib_frame_t * from_frame) +{ + return vxlan_encap_inline (vm, node, from_frame, /* is_ip4 */ 0); +} + +VLIB_REGISTER_NODE (vxlan4_encap_node) = { + .function = vxlan4_encap, + .name = "vxlan4-encap", + .vector_size = sizeof (u32), + .format_trace = format_vxlan_encap_trace, + .type = VLIB_NODE_TYPE_INTERNAL, + .n_errors = ARRAY_LEN(vxlan_encap_error_strings), + .error_strings = vxlan_encap_error_strings, + .n_next_nodes = VXLAN_ENCAP_N_NEXT, + .next_nodes = { + [VXLAN_ENCAP_NEXT_DROP] = "error-drop", + }, +}; + +VLIB_NODE_FUNCTION_MULTIARCH (vxlan4_encap_node, vxlan4_encap) + +VLIB_REGISTER_NODE (vxlan6_encap_node) = { + .function = vxlan6_encap, + .name = "vxlan6-encap", + .vector_size = sizeof (u32), + .format_trace = format_vxlan_encap_trace, + .type = VLIB_NODE_TYPE_INTERNAL, + .n_errors = ARRAY_LEN(vxlan_encap_error_strings), + .error_strings = vxlan_encap_error_strings, + .n_next_nodes = VXLAN_ENCAP_N_NEXT, + .next_nodes = { + [VXLAN_ENCAP_NEXT_DROP] = "error-drop", + }, +}; + +VLIB_NODE_FUNCTION_MULTIARCH (vxlan6_encap_node, vxlan6_encap) + diff --git a/src/vnet/vxlan/vxlan.api b/src/vnet/vxlan/vxlan.api new file mode 100644 index 00000000000..048220fba52 --- /dev/null +++ b/src/vnet/vxlan/vxlan.api @@ -0,0 +1,81 @@ +/* + * 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_add_del_tunnel +{ + u32 client_index; + u32 context; + u8 is_add; + u8 is_ipv6; + u8 src_address[16]; + u8 dst_address[16]; + u32 mcast_sw_if_index; + u32 encap_vrf_id; + u32 decap_next_index; + u32 vni; +}; + +define vxlan_add_del_tunnel_reply +{ + u32 context; + i32 retval; + u32 sw_if_index; +}; + +define vxlan_tunnel_dump +{ + u32 client_index; + u32 context; + u32 sw_if_index; +}; + +define vxlan_tunnel_details +{ + u32 context; + u32 sw_if_index; + u8 src_address[16]; + u8 dst_address[16]; + u32 mcast_sw_if_index; + u32 encap_vrf_id; + u32 decap_next_index; + u32 vni; + u8 is_ipv6; +}; + +/** \brief Interface set vxlan-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 +*/ +define sw_interface_set_vxlan_bypass +{ + u32 client_index; + u32 context; + u32 sw_if_index; + u8 is_ipv6; + u8 enable; +}; + +/** \brief Interface set vxlan-bypass response + @param context - sender context, to match reply w/ request + @param retval - return code for the request +*/ +define sw_interface_set_vxlan_bypass_reply +{ + u32 context; + i32 retval; +};
\ No newline at end of file diff --git a/src/vnet/vxlan/vxlan.c b/src/vnet/vxlan/vxlan.c new file mode 100644 index 00000000000..abf0182f91f --- /dev/null +++ b/src/vnet/vxlan/vxlan.c @@ -0,0 +1,899 @@ +/* + * Copyright (c) 2015 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ +#include <vnet/vxlan/vxlan.h> +#include <vnet/ip/format.h> +#include <vnet/fib/fib_entry.h> +#include <vnet/fib/fib_table.h> +#include <vnet/dpo/receive_dpo.h> +#include <vlib/vlib.h> + +/** + * @file + * @brief VXLAN. + * + * VXLAN 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 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 + * (Virtual eXtensible VLAN) segment. + */ + + +vxlan_main_t vxlan_main; + +static u8 * format_decap_next (u8 * s, va_list * args) +{ + u32 next_index = va_arg (*args, u32); + + switch (next_index) + { + case VXLAN_INPUT_NEXT_DROP: + return format (s, "drop"); + case VXLAN_INPUT_NEXT_L2_INPUT: + return format (s, "l2"); + default: + return format (s, "index %d", next_index); + } + return s; +} + +u8 * format_vxlan_tunnel (u8 * s, va_list * args) +{ + vxlan_tunnel_t * t = va_arg (*args, vxlan_tunnel_t *); + vxlan_main_t * ngm = &vxlan_main; + + s = format (s, "[%d] src %U dst %U vni %d sw_if_index %d ", + t - ngm->tunnels, + format_ip46_address, &t->src, IP46_TYPE_ANY, + format_ip46_address, &t->dst, IP46_TYPE_ANY, + t->vni, t->sw_if_index); + + if (ip46_address_is_multicast (&t->dst)) + s = format (s, "mcast_sw_if_index %d ", t->mcast_sw_if_index); + + s = format (s, "encap_fib_index %d fib_entry_index %d decap_next %U\n", + t->encap_fib_index, t->fib_entry_index, + format_decap_next, t->decap_next_index); + return s; +} + +static u8 * format_vxlan_name (u8 * s, va_list * args) +{ + u32 dev_instance = va_arg (*args, u32); + return format (s, "vxlan_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; +} + +static clib_error_t * +vxlan_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 /* no error */ 0; +} + +VNET_DEVICE_CLASS (vxlan_device_class,static) = { + .name = "VXLAN", + .format_device_name = format_vxlan_name, + .format_tx_trace = format_vxlan_encap_trace, + .tx_function = dummy_interface_tx, + .admin_up_down_function = vxlan_interface_admin_up_down, +}; + +static u8 * format_vxlan_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_hw_class) = { + .name = "VXLAN", + .format_header = format_vxlan_header_with_length, + .build_rewrite = default_build_rewrite, +}; + +static void +vxlan_tunnel_restack_dpo(vxlan_tunnel_t * t) +{ + dpo_id_t dpo = DPO_INVALID; + u32 encap_index = ip46_address_is_ip4(&t->dst) ? + vxlan4_encap_node.index : vxlan6_encap_node.index; + fib_forward_chain_type_t forw_type = ip46_address_is_ip4(&t->dst) ? + 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_tunnel_t * +vxlan_tunnel_from_fib_node (fib_node_t *node) +{ +#if (CLIB_DEBUG > 0) + ASSERT(FIB_NODE_TYPE_VXLAN_TUNNEL == node->fn_type); +#endif + return ((vxlan_tunnel_t*) (((char*)node) - + STRUCT_OFFSET_OF(vxlan_tunnel_t, node))); +} + +/** + * Function definition to backwalk a FIB node - + * Here we will restack the new dpo of VXLAN DIP to encap node. + */ +static fib_node_back_walk_rc_t +vxlan_tunnel_back_walk (fib_node_t *node, + fib_node_back_walk_ctx_t *ctx) +{ + vxlan_tunnel_restack_dpo(vxlan_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_tunnel_fib_node_get (fib_node_index_t index) +{ + vxlan_tunnel_t * t; + vxlan_main_t * vxm = &vxlan_main; + + t = pool_elt_at_index(vxm->tunnels, index); + + return (&t->node); +} + +/** + * Function definition to inform the FIB node that its last lock has gone. + */ +static void +vxlan_tunnel_last_lock_gone (fib_node_t *node) +{ + /* + * The VXLAN 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 tunnels + * for participation in the FIB object graph. + */ +const static fib_node_vft_t vxlan_vft = { + .fnv_get = vxlan_tunnel_fib_node_get, + .fnv_last_lock = vxlan_tunnel_last_lock_gone, + .fnv_back_walk = vxlan_tunnel_back_walk, +}; + + +#define foreach_copy_field \ +_(vni) \ +_(mcast_sw_if_index) \ +_(encap_fib_index) \ +_(decap_next_index) \ +_(src) \ +_(dst) + +static int vxlan4_rewrite (vxlan_tunnel_t * t) +{ + u8 *rw = 0; + ip4_header_t * ip0; + ip4_vxlan_header_t * h0; + int len = sizeof (*h0); + + vec_validate_aligned (rw, len-1, CLIB_CACHE_LINE_BYTES); + + h0 = (ip4_vxlan_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->src.ip4.as_u32; + ip0->dst_address.as_u32 = t->dst.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 (4789); + h0->udp.dst_port = clib_host_to_net_u16 (UDP_DST_PORT_vxlan); + + /* VXLAN header */ + vnet_set_vni_and_flags(&h0->vxlan, t->vni); + + t->rewrite = rw; + return (0); +} + +static int vxlan6_rewrite (vxlan_tunnel_t * t) +{ + u8 *rw = 0; + ip6_header_t * ip0; + ip6_vxlan_header_t * h0; + int len = sizeof (*h0); + + vec_validate_aligned (rw, len-1, CLIB_CACHE_LINE_BYTES); + + h0 = (ip6_vxlan_header_t *) rw; + + /* Fixed portion of the (outer) ip6 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->src.ip6.as_u64[0]; + ip0->src_address.as_u64[1] = t->src.ip6.as_u64[1]; + ip0->dst_address.as_u64[0] = t->dst.ip6.as_u64[0]; + ip0->dst_address.as_u64[1] = t->dst.ip6.as_u64[1]; + + /* UDP header, randomize src port on something, maybe? */ + h0->udp.src_port = clib_host_to_net_u16 (4789); + h0->udp.dst_port = clib_host_to_net_u16 (UDP_DST_PORT_vxlan); + + /* VXLAN header */ + vnet_set_vni_and_flags(&h0->vxlan, t->vni); + + t->rewrite = rw; + return (0); +} + +static int vxlan_check_decap_next(vxlan_main_t * vxm, u32 is_ip6, u32 decap_next_index) +{ + vlib_main_t * vm = vxm->vlib_main; + vlib_node_runtime_t *r; + + if(!is_ip6) + { + r = vlib_node_get_runtime (vm, vxlan4_input_node.index); + if(decap_next_index >= r->n_next_nodes) + return 1; + } + else + { + r = vlib_node_get_runtime (vm, vxlan6_input_node.index); + if(decap_next_index >= r->n_next_nodes) + return 1; + } + + return 0; +} + +static uword +vtep_addr_ref(ip46_address_t *ip) +{ + if (!ip46_address_is_ip4(ip)) + return 1; /* always create */ + uword *pvtep = hash_get (vxlan_main.vtep4, ip->ip4.as_u32); + if (pvtep) + return ++pvtep[0]; + hash_set (vxlan_main.vtep4, ip->ip4.as_u32, 1); + return 1; +} + +static uword +vtep_addr_unref(ip46_address_t *ip) +{ + if (!ip46_address_is_ip4(ip)) + return 0; /* alwways destroy */ + uword *pvtep = hash_get (vxlan_main.vtep4, ip->ip4.as_u32); + ASSERT(pvtep); + if (!(--pvtep[0])) + hash_unset (vxlan_main.vtep4, ip->ip4.as_u32); + return pvtep[0]; +} + +static +mcast_remote_t * +mcast_ep_get(ip46_address_t * ip) +{ + ASSERT(ip46_address_is_multicast(ip)); + uword * ep_idx = hash_get_mem (vxlan_main.mcast_ep_by_ip, ip); + ASSERT(ep_idx); + return pool_elt_at_index(vxlan_main.mcast_eps, *ep_idx); +} + +static void +mcast_ep_add(mcast_remote_t * new_ep) +{ + mcast_remote_t * ep; + + pool_get_aligned (vxlan_main.mcast_eps, ep, CLIB_CACHE_LINE_BYTES); + *ep = *new_ep; + hash_set_mem (vxlan_main.mcast_ep_by_ip, &ep->ip, ep - vxlan_main.mcast_eps); +} + +static void +mcast_ep_remove(mcast_remote_t * ep) +{ + hash_unset_mem (vxlan_main.mcast_ep_by_ip, &ep->ip); + pool_put (vxlan_main.mcast_eps, ep); +} + +static void +ip46_multicast_ethernet_address(u8 * ethernet_address, ip46_address_t * ip) { + if (ip46_address_is_ip4(ip)) + ip4_multicast_ethernet_address(ethernet_address, &ip->ip4); + else + ip6_multicast_ethernet_address(ethernet_address, ip->ip6.as_u32[0]); +} + +int vnet_vxlan_add_del_tunnel +(vnet_vxlan_add_del_tunnel_args_t *a, u32 * sw_if_indexp) +{ + vxlan_main_t * vxm = &vxlan_main; + vxlan_tunnel_t *t = 0; + vnet_main_t * vnm = vxm->vnet_main; + uword * p; + u32 hw_if_index = ~0; + u32 sw_if_index = ~0; + int rv; + vxlan4_tunnel_key_t key4; + vxlan6_tunnel_key_t key6; + u32 is_ip6 = a->is_ip6; + + if (!is_ip6) + { + key4.src = a->dst.ip4.as_u32; /* decap src in key is encap dst in config */ + key4.vni = clib_host_to_net_u32 (a->vni << 8); + p = hash_get (vxm->vxlan4_tunnel_by_key, key4.as_u64); + } + else + { + key6.src.as_u64[0] = a->dst.ip6.as_u64[0]; + key6.src.as_u64[1] = a->dst.ip6.as_u64[1]; + key6.vni = clib_host_to_net_u32 (a->vni << 8); + p = hash_get_mem (vxm->vxlan6_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_TUNNEL_EXIST; + + /*if not set explicitly, default to l2 */ + if(a->decap_next_index == ~0) + a->decap_next_index = VXLAN_INPUT_NEXT_L2_INPUT; + if (vxlan_check_decap_next(vxm, is_ip6, a->decap_next_index)) + return VNET_API_ERROR_INVALID_DECAP_NEXT; + + pool_get_aligned (vxm->tunnels, t, CLIB_CACHE_LINE_BYTES); + memset (t, 0, sizeof (*t)); + + /* copy from arg structure */ +#define _(x) t->x = a->x; + foreach_copy_field; +#undef _ + + /* copy the key */ + if (is_ip6) + { + t->key6 = clib_mem_alloc (sizeof(vxlan6_tunnel_key_t)); + clib_memcpy (t->key6, &key6, sizeof(key6)); + } + else + { + t->key4 = 0; /* not yet used */ + } + + if (!is_ip6) + rv = vxlan4_rewrite (t); + else + rv = vxlan6_rewrite (t); + + if (rv) + { + pool_put (vxm->tunnels, t); + return rv; + } + + if (!is_ip6) + hash_set (vxm->vxlan4_tunnel_by_key, key4.as_u64, t - vxm->tunnels); + else + hash_set_mem (vxm->vxlan6_tunnel_by_key, t->key6, t - vxm->tunnels); + + vnet_hw_interface_t * hi; + if (vec_len (vxm->free_vxlan_tunnel_hw_if_indices) > 0) + { + vnet_interface_main_t * im = &vnm->interface_main; + hw_if_index = vxm->free_vxlan_tunnel_hw_if_indices + [vec_len (vxm->free_vxlan_tunnel_hw_if_indices)-1]; + _vec_len (vxm->free_vxlan_tunnel_hw_if_indices) -= 1; + + hi = vnet_get_hw_interface (vnm, hw_if_index); + hi->dev_instance = t - vxm->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_device_class.index, t - vxm->tunnels, + vxlan_hw_class.index, t - vxm->tunnels); + hi = vnet_get_hw_interface (vnm, hw_if_index); + } + + t->hw_if_index = hw_if_index; + t->sw_if_index = sw_if_index = hi->sw_if_index; + + vec_validate_init_empty (vxm->tunnel_index_by_sw_if_index, sw_if_index, ~0); + vxm->tunnel_index_by_sw_if_index[sw_if_index] = t - vxm->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_set_flags (vnm, sw_if_index, + VNET_SW_INTERFACE_FLAG_ADMIN_UP); + fib_node_init(&t->node, FIB_NODE_TYPE_VXLAN_TUNNEL); + fib_prefix_t tun_dst_pfx; + u32 encap_index = !is_ip6 ? + vxlan4_encap_node.index : vxlan6_encap_node.index; + vnet_flood_class_t flood_class = VNET_FLOOD_CLASS_TUNNEL_NORMAL; + + fib_prefix_from_ip46_addr(&t->dst, &tun_dst_pfx); + if (!ip46_address_is_multicast(&t->dst)) + { + /* 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->src); + t->fib_entry_index = fib_table_entry_special_add + (t->encap_fib_index, &tun_dst_pfx, FIB_SOURCE_RR, + FIB_ENTRY_FLAG_NONE, ADJ_INDEX_INVALID); + t->sibling_index = fib_entry_child_add + (t->fib_entry_index, FIB_NODE_TYPE_VXLAN_TUNNEL, t - vxm->tunnels); + vxlan_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 = (is_ip6) ? FIB_PROTOCOL_IP6 : FIB_PROTOCOL_IP4; + dpo_id_t dpo = DPO_INVALID; + dpo_proto_t dproto = fib_proto_to_dpo(fp); + + if (vtep_addr_ref(&t->dst) == 1) + { + u8 mcast_mac[6]; + + ip46_multicast_ethernet_address(mcast_mac, &t->dst); + receive_dpo_add_or_lock(dproto, ~0, NULL, &dpo); + mcast_remote_t new_ep = { + .ip = t->dst, + .mcast_adj_index = adj_rewrite_add_and_lock + (fp, fib_proto_to_link(fp), a->mcast_sw_if_index, mcast_mac), + /* Add VRF local mcast adj. */ + .fib_entry_index = fib_table_entry_special_dpo_add + (t->encap_fib_index, &tun_dst_pfx, + FIB_SOURCE_SPECIAL, FIB_ENTRY_FLAG_NONE, &dpo) + }; + mcast_ep_add(&new_ep); + dpo_reset(&dpo); + } + /* Stack shared mcast dst mac addr rewrite on encap */ + dpo_set (&dpo, DPO_ADJACENCY, dproto, + mcast_ep_get(&t->dst)->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 + { + /* deleting a tunnel: tunnel must exist */ + if (!p) + return VNET_API_ERROR_NO_SUCH_ENTRY; + + t = pool_elt_at_index (vxm->tunnels, p[0]); + + vnet_sw_interface_set_flags (vnm, t->sw_if_index, 0 /* down */); + /* make sure tunnel is removed from l2 bd or xconnect */ + set_int_l2_mode(vxm->vlib_main, vnm, MODE_L3, t->sw_if_index, 0, 0, 0, 0); + vec_add1 (vxm->free_vxlan_tunnel_hw_if_indices, t->hw_if_index); + + vxm->tunnel_index_by_sw_if_index[t->sw_if_index] = ~0; + + if (!is_ip6) + hash_unset (vxm->vxlan4_tunnel_by_key, key4.as_u64); + else + { + hash_unset_mem (vxm->vxlan6_tunnel_by_key, t->key6); + clib_mem_free (t->key6); + } + + if (!ip46_address_is_multicast(&t->dst)) + { + vtep_addr_unref(&a->src); + 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->dst) == 0) + { + mcast_remote_t* ep = mcast_ep_get(&t->dst); + adj_unlock(ep->mcast_adj_index); + fib_table_entry_delete_index(ep->fib_entry_index, FIB_SOURCE_SPECIAL); + mcast_ep_remove(ep); + } + + fib_node_deinit(&t->node); + vec_free (t->rewrite); + pool_put (vxm->tunnels, t); + } + + if (sw_if_indexp) + *sw_if_indexp = sw_if_index; + + return 0; +} + +static u32 fib4_index_from_fib_id (u32 fib_id) +{ + ip4_main_t * im = &ip4_main; + uword * p; + + p = hash_get (im->fib_index_by_table_id, fib_id); + if (!p) + return ~0; + + return p[0]; +} + +static u32 fib6_index_from_fib_id (u32 fib_id) +{ + ip6_main_t * im = &ip6_main; + uword * p; + + p = hash_get (im->fib_index_by_table_id, fib_id); + if (!p) + return ~0; + + return p[0]; +} + +static uword get_decap_next_for_node(u32 node_index, u32 ipv4_set) +{ + vxlan_main_t * vxm = &vxlan_main; + vlib_main_t * vm = vxm->vlib_main; + uword next_index = ~0; + + if (ipv4_set) + { + next_index = vlib_node_add_next (vm, vxlan4_input_node.index, node_index); + } + else + { + next_index = vlib_node_add_next (vm, vxlan6_input_node.index, node_index); + } + + return next_index; +} + +static uword unformat_decap_next (unformat_input_t * input, va_list * args) +{ + u32 * result = va_arg (*args, u32 *); + u32 ipv4_set = va_arg (*args, int); + vxlan_main_t * vxm = &vxlan_main; + vlib_main_t * vm = vxm->vlib_main; + u32 node_index; + u32 tmp; + + if (unformat (input, "l2")) + *result = VXLAN_INPUT_NEXT_L2_INPUT; + else if (unformat (input, "node %U", unformat_vlib_node, vm, &node_index)) + *result = get_decap_next_for_node(node_index, ipv4_set); + else if (unformat (input, "%d", &tmp)) + *result = tmp; + else + return 0; + return 1; +} + +static clib_error_t * +vxlan_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; + ip46_address_t src , dst; + u8 is_add = 1; + u8 src_set = 0; + u8 dst_set = 0; + u8 grp_set = 0; + u8 ipv4_set = 0; + u8 ipv6_set = 0; + u32 encap_fib_index = 0; + u32 mcast_sw_if_index = ~0; + u32 decap_next_index = VXLAN_INPUT_NEXT_L2_INPUT; + u32 vni = 0; + u32 tmp; + int rv; + vnet_vxlan_add_del_tunnel_args_t _a, * a = &_a; + u32 tunnel_sw_if_index; + + /* Cant "universally zero init" (={0}) due to GCC bug 53119 */ + memset(&src, 0, sizeof src); + memset(&dst, 0, sizeof dst); + + /* 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, "src %U", + unformat_ip4_address, &src.ip4)) + { + src_set = 1; + ipv4_set = 1; + } + else if (unformat (line_input, "dst %U", + unformat_ip4_address, &dst.ip4)) + { + dst_set = 1; + ipv4_set = 1; + } + else if (unformat (line_input, "src %U", + unformat_ip6_address, &src.ip6)) + { + src_set = 1; + ipv6_set = 1; + } + else if (unformat (line_input, "dst %U", + unformat_ip6_address, &dst.ip6)) + { + dst_set = 1; + ipv6_set = 1; + } + else if (unformat (line_input, "group %U %U", + unformat_ip4_address, &dst.ip4, + unformat_vnet_sw_interface, + vnet_get_main(), &mcast_sw_if_index)) + { + grp_set = dst_set = 1; + ipv4_set = 1; + } + else if (unformat (line_input, "group %U %U", + unformat_ip6_address, &dst.ip6, + unformat_vnet_sw_interface, + vnet_get_main(), &mcast_sw_if_index)) + { + grp_set = dst_set = 1; + ipv6_set = 1; + } + else if (unformat (line_input, "encap-vrf-id %d", &tmp)) + { + if (ipv6_set) + encap_fib_index = fib6_index_from_fib_id (tmp); + else + encap_fib_index = fib4_index_from_fib_id (tmp); + if (encap_fib_index == ~0) + return clib_error_return (0, "nonexistent encap-vrf-id %d", tmp); + } + else if (unformat (line_input, "decap-next %U", unformat_decap_next, + &decap_next_index, ipv4_set)) + ; + else if (unformat (line_input, "vni %d", &vni)) + { + if (vni >> 24) + return clib_error_return (0, "vni %d out of range", vni); + } + else + return clib_error_return (0, "parse error: '%U'", + format_unformat_error, line_input); + } + + unformat_free (line_input); + + if (src_set == 0) + return clib_error_return (0, "tunnel src address not specified"); + + if (dst_set == 0) + return clib_error_return (0, "tunnel dst address not specified"); + + if (grp_set && !ip46_address_is_multicast(&dst)) + return clib_error_return (0, "tunnel group address not multicast"); + + if (grp_set == 0 && ip46_address_is_multicast(&dst)) + return clib_error_return (0, "dst address must be unicast"); + + if (grp_set && mcast_sw_if_index == ~0) + return clib_error_return (0, "tunnel nonexistent multicast device"); + + if (ipv4_set && ipv6_set) + return clib_error_return (0, "both IPv4 and IPv6 addresses specified"); + + if (ip46_address_cmp(&src, &dst) == 0) + return clib_error_return (0, "src and dst addresses are identical"); + + if (decap_next_index == ~0) + return clib_error_return (0, "next node not found"); + + if (vni == 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_copy_field; +#undef _ + + rv = vnet_vxlan_add_del_tunnel (a, &tunnel_sw_if_index); + + switch(rv) + { + case 0: + if (is_add) + vlib_cli_output(vm, "%U\n", format_vnet_sw_if_index_name, + vnet_get_main(), tunnel_sw_if_index); + break; + + 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_add_del_tunnel returned %d", rv); + } + + return 0; +} + +/*? + * Add or delete a VXLAN Tunnel. + * + * VXLAN 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 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 + * (Virtual eXtensible VLAN) segment. + * + * @cliexpar + * Example of how to create a VXLAN Tunnel: + * @cliexcmd{create vxlan tunnel src 10.0.3.1 dst 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 dst 10.0.3.3 vni 13 del} + ?*/ +/* *INDENT-OFF* */ +VLIB_CLI_COMMAND (create_vxlan_tunnel_command, static) = { + .path = "create vxlan tunnel", + .short_help = + "create vxlan tunnel src <local-vtep-addr>" + " {dst <remote-vtep-addr>|group <mcast-vtep-addr> <intf-name>} vni <nn>" + " [encap-vrf-id <nn>] [decap-next [l2|node <name>]] [del]", + .function = vxlan_add_del_tunnel_command_fn, +}; +/* *INDENT-ON* */ + +static clib_error_t * +show_vxlan_tunnel_command_fn (vlib_main_t * vm, + unformat_input_t * input, + vlib_cli_command_t * cmd) +{ + vxlan_main_t * vxm = &vxlan_main; + vxlan_tunnel_t * t; + + if (pool_elts (vxm->tunnels) == 0) + vlib_cli_output (vm, "No vxlan tunnels configured..."); + + pool_foreach (t, vxm->tunnels, + ({ + vlib_cli_output (vm, "%U", format_vxlan_tunnel, t); + })); + + return 0; +} + +/*? + * Display all the VXLAN Tunnel entries. + * + * @cliexpar + * Example of how to display the VXLAN Tunnel entries: + * @cliexstart{show vxlan tunnel} + * [0] src 10.0.3.1 dst 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_tunnel_command, static) = { + .path = "show vxlan tunnel", + .short_help = "show vxlan tunnel", + .function = show_vxlan_tunnel_command_fn, +}; +/* *INDENT-ON* */ + + +clib_error_t *vxlan_init (vlib_main_t *vm) +{ + vxlan_main_t * vxm = &vxlan_main; + + vxm->vnet_main = vnet_get_main(); + vxm->vlib_main = vm; + + /* initialize the ip6 hash */ + vxm->vxlan6_tunnel_by_key = hash_create_mem(0, + sizeof(vxlan6_tunnel_key_t), + sizeof(uword)); + vxm->mcast_ep_by_ip = hash_create_mem(0, + sizeof(ip46_address_t), + sizeof(uword)); + + udp_register_dst_port (vm, UDP_DST_PORT_vxlan, + vxlan4_input_node.index, /* is_ip4 */ 1); + udp_register_dst_port (vm, UDP_DST_PORT_vxlan6, + vxlan6_input_node.index, /* is_ip4 */ 0); + + fib_node_register_type(FIB_NODE_TYPE_VXLAN_TUNNEL, &vxlan_vft); + + return 0; +} + +VLIB_INIT_FUNCTION(vxlan_init); diff --git a/src/vnet/vxlan/vxlan.h b/src/vnet/vxlan/vxlan.h new file mode 100644 index 00000000000..7605d94aa98 --- /dev/null +++ b/src/vnet/vxlan/vxlan.h @@ -0,0 +1,199 @@ +/* + * Copyright (c) 2015 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ +#ifndef included_vnet_vxlan_h +#define included_vnet_vxlan_h + +#include <vppinfra/error.h> +#include <vppinfra/hash.h> +#include <vnet/vnet.h> +#include <vnet/ip/ip.h> +#include <vnet/l2/l2_input.h> +#include <vnet/l2/l2_output.h> +#include <vnet/l2/l2_bd.h> +#include <vnet/ethernet/ethernet.h> +#include <vnet/vxlan/vxlan_packet.h> +#include <vnet/ip/ip4_packet.h> +#include <vnet/ip/ip6_packet.h> +#include <vnet/ip/udp.h> +#include <vnet/dpo/dpo.h> +#include <vnet/adj/adj_types.h> + +typedef CLIB_PACKED (struct { + ip4_header_t ip4; /* 20 bytes */ + udp_header_t udp; /* 8 bytes */ + vxlan_header_t vxlan; /* 8 bytes */ +}) ip4_vxlan_header_t; + +typedef CLIB_PACKED (struct { + ip6_header_t ip6; /* 40 bytes */ + udp_header_t udp; /* 8 bytes */ + vxlan_header_t vxlan; /* 8 bytes */ +}) ip6_vxlan_header_t; + +typedef CLIB_PACKED(struct { + /* + * Key fields: ip src and vxlan vni on incoming VXLAN packet + * all fields in NET byte order + */ + union { + struct { + u32 src; + u32 vni; /* shifted left 8 bits */ + }; + u64 as_u64; + }; +}) vxlan4_tunnel_key_t; + +typedef CLIB_PACKED(struct { + /* + * Key fields: ip src and vxlan vni on incoming VXLAN packet + * all fields in NET byte order + */ + ip6_address_t src; + u32 vni; /* shifted left 8 bits */ +}) vxlan6_tunnel_key_t; + +typedef struct { + /* Rewrite string. $$$$ embed vnet_rewrite header */ + u8 * rewrite; + + /* FIB DPO for IP forwarding of VXLAN encap packet */ + dpo_id_t next_dpo; + + /* storage for the hash key */ + union { + vxlan4_tunnel_key_t *key4; /* unused for now */ + vxlan6_tunnel_key_t *key6; + }; + + /* vxlan VNI in HOST byte order */ + u32 vni; + + /* tunnel src and dst addresses */ + ip46_address_t src; + ip46_address_t dst; + + /* mcast packet output intfc index (used only if dst is mcast) */ + u32 mcast_sw_if_index; + + /* decap next index */ + u32 decap_next_index; + + /* The FIB index for src/dst addresses */ + u32 encap_fib_index; + + /* vnet intfc index */ + u32 sw_if_index; + u32 hw_if_index; + + /** + * Linkage into the FIB object graph + */ + fib_node_t node; + + /* + * The FIB entry for (depending on VXLAN tunnel is unicast or mcast) + * sending unicast VXLAN encap packets or receiving mcast VXLAN 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_tunnel_t; + +#define foreach_vxlan_input_next \ +_(DROP, "error-drop") \ +_(L2_INPUT, "l2-input") + +typedef enum { +#define _(s,n) VXLAN_INPUT_NEXT_##s, + foreach_vxlan_input_next +#undef _ + VXLAN_INPUT_N_NEXT, +} vxlan_input_next_t; + +typedef enum { +#define vxlan_error(n,s) VXLAN_ERROR_##n, +#include <vnet/vxlan/vxlan_error.def> +#undef vxlan_error + VXLAN_N_ERROR, +} vxlan_input_error_t; + +typedef struct { + ip46_address_t ip; + fib_node_index_t fib_entry_index; + adj_index_t mcast_adj_index; +} mcast_remote_t; + +typedef struct { + /* vector of encap tunnel instances */ + vxlan_tunnel_t * tunnels; + + /* lookup tunnel by key */ + uword * vxlan4_tunnel_by_key; /* keyed on ipv4.dst + vni */ + uword * vxlan6_tunnel_by_key; /* keyed on ipv6.dst + vni */ + + /* 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 */ + + /* set of active remote mcast VTEP */ + mcast_remote_t * mcast_eps; + uword * mcast_ep_by_ip; /* mcast VTEPs keyed on their ip46 addr */ + + /* Free vlib hw_if_indices */ + u32 * free_vxlan_tunnel_hw_if_indices; + + /* Mapping from sw_if_index to tunnel index */ + u32 * tunnel_index_by_sw_if_index; + + /* convenience */ + vlib_main_t * vlib_main; + vnet_main_t * vnet_main; +} vxlan_main_t; + +vxlan_main_t vxlan_main; + +extern vlib_node_registration_t vxlan4_input_node; +extern vlib_node_registration_t vxlan6_input_node; +extern vlib_node_registration_t vxlan4_encap_node; +extern vlib_node_registration_t vxlan6_encap_node; + +u8 * format_vxlan_encap_trace (u8 * s, va_list * args); + +typedef struct { + u8 is_add; + + /* we normally use is_ip4, but since this adds to the + * structure, this seems less of abreaking change */ + u8 is_ip6; + ip46_address_t src, dst; + u32 mcast_sw_if_index; + u32 encap_fib_index; + u32 decap_next_index; + u32 vni; +} vnet_vxlan_add_del_tunnel_args_t; + +int vnet_vxlan_add_del_tunnel +(vnet_vxlan_add_del_tunnel_args_t *a, u32 * sw_if_indexp); + +#endif /* included_vnet_vxlan_h */ diff --git a/src/vnet/vxlan/vxlan_api.c b/src/vnet/vxlan/vxlan_api.c new file mode 100644 index 00000000000..6c9cbd79764 --- /dev/null +++ b/src/vnet/vxlan/vxlan_api.c @@ -0,0 +1,253 @@ +/* + *------------------------------------------------------------------ + * vxlan_api.c - vxlan 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 <vnet/vnet.h> +#include <vlibmemory/api.h> + +#include <vnet/interface.h> +#include <vnet/api_errno.h> +#include <vnet/feature/feature.h> +#include <vnet/vxlan/vxlan.h> +#include <vnet/fib/fib_table.h> + +#include <vnet/vnet_msg_enum.h> + +#define vl_typedefs /* define message structures */ +#include <vnet/vnet_all_api_h.h> +#undef vl_typedefs + +#define vl_endianfun /* define message structures */ +#include <vnet/vnet_all_api_h.h> +#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 <vnet/vnet_all_api_h.h> +#undef vl_printfun + +#include <vlibapi/api_helper_macros.h> + +#define foreach_vpe_api_msg \ +_(SW_INTERFACE_SET_VXLAN_BYPASS, sw_interface_set_vxlan_bypass) \ +_(VXLAN_ADD_DEL_TUNNEL, vxlan_add_del_tunnel) \ +_(VXLAN_TUNNEL_DUMP, vxlan_tunnel_dump) + +static void + vl_api_sw_interface_set_vxlan_bypass_t_handler + (vl_api_sw_interface_set_vxlan_bypass_t * mp) +{ + vl_api_sw_interface_set_vxlan_bypass_reply_t *rmp; + int rv = 0; + u32 sw_if_index = ntohl (mp->sw_if_index); + + VALIDATE_SW_IF_INDEX (mp); + + if (mp->is_ipv6) + { + /* not yet implemented */ + } + else + vnet_feature_enable_disable ("ip4-unicast", "ip4-vxlan-bypass", + sw_if_index, mp->enable, 0, 0); + + BAD_SW_IF_INDEX_LABEL; + + REPLY_MACRO (VL_API_SW_INTERFACE_SET_VXLAN_BYPASS_REPLY); +} + +static void vl_api_vxlan_add_del_tunnel_t_handler + (vl_api_vxlan_add_del_tunnel_t * mp) +{ + vl_api_vxlan_add_del_tunnel_reply_t *rmp; + int rv = 0; + vnet_vxlan_add_del_tunnel_args_t _a, *a = &_a; + u32 encap_fib_index; + uword *p; + ip4_main_t *im = &ip4_main; + vnet_main_t *vnm = vnet_get_main (); + u32 sw_if_index = ~0; + + p = hash_get (im->fib_index_by_table_id, ntohl (mp->encap_vrf_id)); + if (!p) + { + rv = VNET_API_ERROR_NO_SUCH_FIB; + goto out; + } + encap_fib_index = p[0]; + memset (a, 0, sizeof (*a)); + + a->is_add = mp->is_add; + a->is_ip6 = mp->is_ipv6; + + /* ip addresses sent in network byte order */ + ip46_from_addr_buf (mp->is_ipv6, mp->dst_address, &a->dst); + ip46_from_addr_buf (mp->is_ipv6, mp->src_address, &a->src); + + /* Check src & dst are different */ + if (ip46_address_cmp (&a->dst, &a->src) == 0) + { + rv = VNET_API_ERROR_SAME_SRC_DST; + goto out; + } + a->mcast_sw_if_index = ntohl (mp->mcast_sw_if_index); + if (ip46_address_is_multicast (&a->dst) && + pool_is_free_index (vnm->interface_main.sw_interfaces, + a->mcast_sw_if_index)) + { + rv = VNET_API_ERROR_INVALID_SW_IF_INDEX; + goto out; + } + a->encap_fib_index = encap_fib_index; + a->decap_next_index = ntohl (mp->decap_next_index); + a->vni = ntohl (mp->vni); + rv = vnet_vxlan_add_del_tunnel (a, &sw_if_index); + +out: + /* *INDENT-OFF* */ + REPLY_MACRO2(VL_API_VXLAN_ADD_DEL_TUNNEL_REPLY, + ({ + rmp->sw_if_index = ntohl (sw_if_index); + })); + /* *INDENT-ON* */ +} + +static void send_vxlan_tunnel_details + (vxlan_tunnel_t * t, unix_shared_memory_queue_t * q, u32 context) +{ + vl_api_vxlan_tunnel_details_t *rmp; + ip4_main_t *im4 = &ip4_main; + ip6_main_t *im6 = &ip6_main; + u8 is_ipv6 = !ip46_address_is_ip4 (&t->dst); + + rmp = vl_msg_api_alloc (sizeof (*rmp)); + memset (rmp, 0, sizeof (*rmp)); + rmp->_vl_msg_id = ntohs (VL_API_VXLAN_TUNNEL_DETAILS); + if (is_ipv6) + { + memcpy (rmp->src_address, t->src.ip6.as_u8, 16); + memcpy (rmp->dst_address, t->dst.ip6.as_u8, 16); + rmp->encap_vrf_id = htonl (im6->fibs[t->encap_fib_index].ft_table_id); + } + else + { + memcpy (rmp->src_address, t->src.ip4.as_u8, 4); + memcpy (rmp->dst_address, t->dst.ip4.as_u8, 4); + rmp->encap_vrf_id = htonl (im4->fibs[t->encap_fib_index].ft_table_id); + } + rmp->mcast_sw_if_index = htonl (t->mcast_sw_if_index); + rmp->vni = htonl (t->vni); + rmp->decap_next_index = htonl (t->decap_next_index); + 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_tunnel_dump_t_handler + (vl_api_vxlan_tunnel_dump_t * mp) +{ + unix_shared_memory_queue_t *q; + vxlan_main_t *vxm = &vxlan_main; + vxlan_tunnel_t *t; + u32 sw_if_index; + + q = vl_api_client_index_to_input_queue (mp->client_index); + if (q == 0) + { + return; + } + + sw_if_index = ntohl (mp->sw_if_index); + + if (~0 == sw_if_index) + { + /* *INDENT-OFF* */ + pool_foreach (t, vxm->tunnels, + ({ + send_vxlan_tunnel_details(t, q, mp->context); + })); + /* *INDENT-ON* */ + } + else + { + if ((sw_if_index >= vec_len (vxm->tunnel_index_by_sw_if_index)) || + (~0 == vxm->tunnel_index_by_sw_if_index[sw_if_index])) + { + return; + } + t = &vxm->tunnels[vxm->tunnel_index_by_sw_if_index[sw_if_index]]; + send_vxlan_tunnel_details (t, q, mp->context); + } +} + +/* + * 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 <vnet/vnet_all_api_h.h> +#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_vxlan; +#undef _ +} + +static clib_error_t * +vxlan_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 _ + + am->api_trace_cfg[VL_API_VXLAN_ADD_DEL_TUNNEL].size += 16 * sizeof (u32); + + /* + * Set up the (msg_name, crc, message-id) table + */ + setup_message_id_table (am); + + return 0; +} + +VLIB_API_INIT_FUNCTION (vxlan_api_hookup); + +/* + * fd.io coding-style-patch-verification: ON + * + * Local Variables: + * eval: (c-set-style "gnu") + * End: + */ diff --git a/src/vnet/vxlan/vxlan_error.def b/src/vnet/vxlan/vxlan_error.def new file mode 100644 index 00000000000..17f905950f5 --- /dev/null +++ b/src/vnet/vxlan/vxlan_error.def @@ -0,0 +1,17 @@ +/* + * 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_error (DECAPSULATED, "good packets decapsulated") +vxlan_error (NO_SUCH_TUNNEL, "no such tunnel packets") +vxlan_error (BAD_FLAGS, "packets with bad flags field in vxlan header") diff --git a/src/vnet/vxlan/vxlan_packet.h b/src/vnet/vxlan/vxlan_packet.h new file mode 100644 index 00000000000..5f93a36fd8f --- /dev/null +++ b/src/vnet/vxlan/vxlan_packet.h @@ -0,0 +1,69 @@ +/* + * Copyright (c) 2015 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ +#ifndef __included_vxlan_packet_h__ +#define __included_vxlan_packet_h__ 1 + +/* + * From RFC-7384 + * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ + * |R|R|R|R|I|R|R|R| Reserved | + * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ + * | VXLAN Network Identifier (VNI) | Reserved | + * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ + * + * VXLAN Header: This is an 8-byte field that has: + * + * - Flags (8 bits): where the I flag MUST be set to 1 for a valid + * VXLAN Network ID (VNI). The other 7 bits (designated "R") are + * reserved fields and MUST be set to zero on transmission and + * ignored on receipt. + * + * - VXLAN Segment ID/VXLAN Network Identifier (VNI): this is a + * 24-bit value used to designate the individual VXLAN overlay + * network on which the communicating VMs are situated. VMs in + * different VXLAN overlay networks cannot communicate with each + * other. + * + * - Reserved fields (24 bits and 8 bits): MUST be set to zero on + * transmission and ignored on receipt. + * + */ + +typedef struct { + u8 flags; + u8 res1; + u8 res2; + u8 res3; + u32 vni_reserved; +} vxlan_header_t; + +#define VXLAN_FLAGS_I 0x08 + +static inline u32 vnet_get_vni (vxlan_header_t * h) +{ + u32 vni_reserved_host_byte_order; + + vni_reserved_host_byte_order = clib_net_to_host_u32 (h->vni_reserved); + return vni_reserved_host_byte_order >> 8; +} + +static inline void vnet_set_vni_and_flags (vxlan_header_t * h, u32 vni) +{ + h->vni_reserved = clib_host_to_net_u32 (vni<<8); + * (u32 *) h = 0; + h->flags = VXLAN_FLAGS_I; +} + +#endif /* __included_vxlan_packet_h__ */ |