diff options
Diffstat (limited to 'vnet/vnet/vxlan')
| -rw-r--r-- | vnet/vnet/vxlan/decap.c | 429 | ||||
| -rw-r--r-- | vnet/vnet/vxlan/encap.c | 449 | ||||
| -rw-r--r-- | vnet/vnet/vxlan/vxlan.c | 436 | ||||
| -rw-r--r-- | vnet/vnet/vxlan/vxlan.h | 128 | ||||
| -rw-r--r-- | vnet/vnet/vxlan/vxlan_error.def | 16 | ||||
| -rw-r--r-- | vnet/vnet/vxlan/vxlan_packet.h | 66 |
6 files changed, 1524 insertions, 0 deletions
diff --git a/vnet/vnet/vxlan/decap.c b/vnet/vnet/vxlan/decap.c new file mode 100644 index 00000000000..7789bed9310 --- /dev/null +++ b/vnet/vnet/vxlan/decap.c @@ -0,0 +1,429 @@ +/* + * 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 vxlan_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: tunnel %d vni %d next %d error %d", + t->tunnel_index, t->vni, t->next_index, t->error); + } + else + { + s = format (s, "VXLAN: no tunnel for vni %d next %d error %d", + t->vni, t->next_index, t->error); + } + return s; +} + +static uword +vxlan_input (vlib_main_t * vm, + vlib_node_runtime_t * node, + vlib_frame_t * from_frame) +{ + 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; + vxlan_tunnel_key_t last_key; + u32 pkts_decapsulated = 0; + u32 cpu_index = os_get_cpu_number(); + u32 stats_sw_if_index, stats_n_packets, stats_n_bytes; + + last_key.as_u64 = ~0; + + 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 * ip0, * ip1; + vxlan_header_t * vxlan0, * vxlan1; + uword * p0, * p1; + u32 tunnel_index0, tunnel_index1; + vxlan_tunnel_t * t0, * t1; + vxlan_tunnel_key_t key0, key1; + 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); + + 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))); + ip0 = vlib_buffer_get_current (b0); + ip1 = vlib_buffer_get_current (b1); + + /* pop (ip, udp, vxlan) */ + vlib_buffer_advance + (b0, sizeof(*ip0)+sizeof(udp_header_t)+sizeof(*vxlan0)); + vlib_buffer_advance + (b1, sizeof(*ip1)+sizeof(udp_header_t)+sizeof(*vxlan1)); + + tunnel_index0 = ~0; + error0 = 0; + + tunnel_index1 = ~0; + error1 = 0; + + key0.src = ip0->src_address.as_u32; + key0.vni = vxlan0->vni_reserved; + + if (PREDICT_FALSE (key0.as_u64 != last_key.as_u64)) + { + p0 = hash_get (vxm->vxlan_tunnel_by_key, key0.as_u64); + + if (p0 == 0) + { + error0 = VXLAN_ERROR_NO_SUCH_TUNNEL; + next0 = VXLAN_INPUT_NEXT_DROP; + goto trace0; + } + + last_key.as_u64 = key0.as_u64; + tunnel_index0 = last_tunnel_index = p0[0]; + } + else + tunnel_index0 = last_tunnel_index; + + t0 = pool_elt_at_index (vxm->tunnels, tunnel_index0); + + 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 */ + vnet_update_l2_len (b0); + + /* Set input sw_if_index to VXLAN tunnel for learning */ + vnet_buffer(b0)->sw_if_index[VLIB_RX] = 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); + } + + key1.src = ip1->src_address.as_u32; + key1.vni = vxlan1->vni_reserved; + + if (PREDICT_FALSE (key1.as_u64 != last_key.as_u64)) + { + p1 = hash_get (vxm->vxlan_tunnel_by_key, key1.as_u64); + + if (p1 == 0) + { + error1 = VXLAN_ERROR_NO_SUCH_TUNNEL; + next1 = VXLAN_INPUT_NEXT_DROP; + goto trace1; + } + + last_key.as_u64 = key1.as_u64; + tunnel_index1 = last_tunnel_index = p1[0]; + } + else + tunnel_index1 = last_tunnel_index; + + t1 = pool_elt_at_index (vxm->tunnels, tunnel_index1); + + 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 */ + vnet_update_l2_len (b1); + + /* Set input sw_if_index to VXLAN tunnel for learning */ + vnet_buffer(b1)->sw_if_index[VLIB_RX] = 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 * ip0; + vxlan_header_t * vxlan0; + uword * p0; + u32 tunnel_index0; + vxlan_tunnel_t * t0; + vxlan_tunnel_key_t key0; + 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); + + vlib_buffer_advance + (b0, -(word)(sizeof(udp_header_t)+sizeof(ip4_header_t))); + ip0 = vlib_buffer_get_current (b0); + + /* pop (ip, udp, vxlan) */ + vlib_buffer_advance + (b0, sizeof(*ip0)+sizeof(udp_header_t)+sizeof(*vxlan0)); + + tunnel_index0 = ~0; + error0 = 0; + + key0.src = ip0->src_address.as_u32; + key0.vni = vxlan0->vni_reserved; + + if (PREDICT_FALSE (key0.as_u64 != last_key.as_u64)) + { + p0 = hash_get (vxm->vxlan_tunnel_by_key, key0.as_u64); + + if (p0 == 0) + { + error0 = VXLAN_ERROR_NO_SUCH_TUNNEL; + next0 = VXLAN_INPUT_NEXT_DROP; + goto trace00; + } + + last_key.as_u64 = key0.as_u64; + tunnel_index0 = last_tunnel_index = p0[0]; + } + else + tunnel_index0 = last_tunnel_index; + + t0 = pool_elt_at_index (vxm->tunnels, tunnel_index0); + + 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 */ + vnet_update_l2_len (b0); + + /* Set input sw_if_index to VXLAN tunnel for learning */ + vnet_buffer(b0)->sw_if_index[VLIB_RX] = 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, vxlan_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 char * vxlan_error_strings[] = { +#define vxlan_error(n,s) s, +#include <vnet/vxlan/vxlan_error.def> +#undef vxlan_error +#undef _ +}; + +VLIB_REGISTER_NODE (vxlan_input_node) = { + .function = vxlan_input, + .name = "vxlan-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, +}; diff --git a/vnet/vnet/vxlan/encap.c b/vnet/vnet/vxlan/encap.c new file mode 100644 index 00000000000..4b475f83886 --- /dev/null +++ b/vnet/vnet/vxlan/encap.c @@ -0,0 +1,449 @@ +/* + * 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") \ +_(DEL_TUNNEL, "deleted tunnel packets") + +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_IP4_LOOKUP, + 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: tunnel %d vni %d", t->tunnel_index, t->vni); + return s; +} + +#define foreach_fixed_header_offset \ + _(0) _(1) _(2) _(3) + +static uword +vxlan_encap (vlib_main_t * vm, + vlib_node_runtime_t * node, + vlib_frame_t * from_frame) +{ + u32 n_left_from, next_index, * from, * to_next; + vxlan_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; + + 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 next0 = VXLAN_ENCAP_NEXT_IP4_LOOKUP; + u32 next1 = VXLAN_ENCAP_NEXT_IP4_LOOKUP; + u32 sw_if_index0, sw_if_index1, len0, len1; + vnet_hw_interface_t * hi0, * hi1; + ip4_header_t * ip0, * ip1; + 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; + vxlan_tunnel_t * t0, * t1; + 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); + + /* 1-wide cache? */ + sw_if_index0 = vnet_buffer(b0)->sw_if_index[VLIB_TX]; + sw_if_index1 = vnet_buffer(b1)->sw_if_index[VLIB_TX]; + hi0 = vnet_get_sup_hw_interface (vnm, sw_if_index0); + hi1 = vnet_get_sup_hw_interface (vnm, sw_if_index1); + + t0 = &vxm->tunnels[hi0->dev_instance]; + t1 = &vxm->tunnels[hi1->dev_instance]; + + /* Check rewrite string and drop packet if tunnel is deleted */ + if (PREDICT_FALSE(t0->rewrite == vxlan_dummy_rewrite)) + { + next0 = VXLAN_ENCAP_NEXT_DROP; + b0->error = node->errors[VXLAN_ENCAP_ERROR_DEL_TUNNEL]; + pkts_encapsulated --; + } /* Still go through normal encap with dummy rewrite */ + if (PREDICT_FALSE(t1->rewrite == vxlan_dummy_rewrite)) + { + next1 = VXLAN_ENCAP_NEXT_DROP; + b1->error = node->errors[VXLAN_ENCAP_ERROR_DEL_TUNNEL]; + pkts_encapsulated --; + } /* Still go through normal encap with dummy rewrite */ + + /* IP4 VXLAN header sizeof(ip4_vxlan_header_t) should be 36 octects */ + ASSERT(vec_len(t0->rewrite) == 36); + ASSERT(vec_len(t1->rewrite) == 36); + + /* Apply the rewrite string. $$$$ vnet_rewrite? */ + vlib_buffer_advance (b0, -(word)_vec_len(t0->rewrite)); + vlib_buffer_advance (b1, -(word)_vec_len(t1->rewrite)); + + ip0 = vlib_buffer_get_current(b0); + ip1 = vlib_buffer_get_current(b1); + /* Copy the fixed header */ + copy_dst0 = (u64 *) ip0; + copy_src0 = (u64 *) t0->rewrite; + copy_dst1 = (u64 *) ip1; + copy_src1 = (u64 *) t1->rewrite; + + /* Copy first 32 octets 8-bytes at a time */ +#define _(offs) copy_dst0[offs] = copy_src0[offs]; + foreach_fixed_header_offset; +#undef _ +#define _(offs) copy_dst1[offs] = copy_src1[offs]; + foreach_fixed_header_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_last1 = (u32 *)(©_dst1[4]); + copy_src_last1 = (u32 *)(©_src1[4]); + + copy_dst_last0[0] = copy_src_last0[0]; + copy_dst_last1[0] = copy_src_last1[0]; + + /* fix the <bleep>ing outer-IP checksum */ + sum0 = ip0->checksum; + /* old_l0 always 0, see the rewrite setup */ + new_l0 = + clib_host_to_net_u16 (vlib_buffer_length_in_chain (vm, b0)); + + sum0 = ip_csum_update (sum0, old_l0, new_l0, ip4_header_t, + length /* changed member */); + ip0->checksum = ip_csum_fold (sum0); + ip0->length = new_l0; + + sum1 = ip1->checksum; + /* old_l1 always 0, see the rewrite setup */ + new_l1 = + 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 */); + ip1->checksum = ip_csum_fold (sum1); + ip1->length = new_l1; + + /* Fix UDP length */ + udp0 = (udp_header_t *)(ip0+1); + new_l0 = clib_host_to_net_u16 (vlib_buffer_length_in_chain (vm, b0) + - sizeof (*ip0)); + udp1 = (udp_header_t *)(ip1+1); + new_l1 = clib_host_to_net_u16 (vlib_buffer_length_in_chain (vm, b1) + - sizeof (*ip1)); + + udp0->length = new_l0; + udp0->src_port = flow_hash0; + + udp1->length = new_l1; + udp1->src_port = flow_hash1; + + /* Reset to look up tunnel partner in the configured FIB */ + vnet_buffer(b0)->sw_if_index[VLIB_TX] = t0->encap_fib_index; + vnet_buffer(b1)->sw_if_index[VLIB_TX] = t1->encap_fib_index; + pkts_encapsulated += 2; + + len0 = vlib_buffer_length_in_chain (vm, b0); + len1 = vlib_buffer_length_in_chain (vm, b0); + stats_n_packets += 2; + stats_n_bytes += len0 + len1; + + /* Batch stats increment on the same vxlan tunnel so counter is not + incremented per packet. Note stats are still incremented for deleted + and admin-down tunnel where packets are dropped. It is not worthwhile + to check for this rare case and affect normal path performance. */ + if (PREDICT_FALSE ((sw_if_index0 != stats_sw_if_index) || + (sw_if_index0 != 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 next0 = VXLAN_ENCAP_NEXT_IP4_LOOKUP; + u32 sw_if_index0, len0; + vnet_hw_interface_t * hi0; + ip4_header_t * ip0; + udp_header_t * udp0; + u64 * copy_src0, * copy_dst0; + u32 * copy_src_last0, * copy_dst_last0; + vxlan_tunnel_t * t0; + 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); + + /* 1-wide cache? */ + 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]; + + /* Check rewrite string and drop packet if tunnel is deleted */ + if (PREDICT_FALSE(t0->rewrite == vxlan_dummy_rewrite)) + { + next0 = VXLAN_ENCAP_NEXT_DROP; + b0->error = node->errors[VXLAN_ENCAP_ERROR_DEL_TUNNEL]; + pkts_encapsulated --; + } /* Still go through normal encap with dummy rewrite */ + + /* IP4 VXLAN header sizeof(ip4_vxlan_header_t) should be 36 octects */ + ASSERT(vec_len(t0->rewrite) == 36); + + /* Apply the rewrite string. $$$$ vnet_rewrite? */ + vlib_buffer_advance (b0, -(word)_vec_len(t0->rewrite)); + + ip0 = vlib_buffer_get_current(b0); + /* Copy the fixed header */ + copy_dst0 = (u64 *) ip0; + copy_src0 = (u64 *) t0->rewrite; + + /* Copy first 32 octets 8-bytes at a time */ +#define _(offs) copy_dst0[offs] = copy_src0[offs]; + foreach_fixed_header_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 <bleep>ing outer-IP checksum */ + sum0 = ip0->checksum; + /* old_l0 always 0, see the rewrite setup */ + new_l0 = + clib_host_to_net_u16 (vlib_buffer_length_in_chain (vm, b0)); + + sum0 = ip_csum_update (sum0, old_l0, new_l0, ip4_header_t, + length /* changed member */); + ip0->checksum = ip_csum_fold (sum0); + ip0->length = new_l0; + + /* Fix UDP length */ + udp0 = (udp_header_t *)(ip0+1); + new_l0 = clib_host_to_net_u16 (vlib_buffer_length_in_chain (vm, b0) + - sizeof (*ip0)); + + udp0->length = new_l0; + udp0->src_port = flow_hash0; + + /* Reset to look up tunnel partner in the configured FIB */ + vnet_buffer(b0)->sw_if_index[VLIB_TX] = t0->encap_fib_index; + 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; +} + +VLIB_REGISTER_NODE (vxlan_encap_node) = { + .function = vxlan_encap, + .name = "vxlan-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_IP4_LOOKUP] = "ip4-lookup", + [VXLAN_ENCAP_NEXT_DROP] = "error-drop", + }, +}; diff --git a/vnet/vnet/vxlan/vxlan.c b/vnet/vnet/vxlan/vxlan.c new file mode 100644 index 00000000000..75cbc6c6499 --- /dev/null +++ b/vnet/vnet/vxlan/vxlan.c @@ -0,0 +1,436 @@ +/* + * 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> + +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"); + case VXLAN_INPUT_NEXT_IP4_INPUT: + return format (s, "ip4"); + case VXLAN_INPUT_NEXT_IP6_INPUT: + return format (s, "ip6"); + default: + return format (s, "unknown %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] %U (src) %U (dst) vni %d encap_fib_index %d", + t - ngm->tunnels, + format_ip4_address, &t->src, + format_ip4_address, &t->dst, + t->vni, + t->encap_fib_index); + s = format (s, " decap_next %U\n", 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; +} + +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, +}; + +static uword dummy_set_rewrite (vnet_main_t * vnm, + u32 sw_if_index, + u32 l3_type, + void * dst_address, + void * rewrite, + uword max_rewrite_bytes) +{ + return 0; +} + +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, + .set_rewrite = dummy_set_rewrite, +}; + +#define foreach_copy_field \ +_(src.as_u32) \ +_(dst.as_u32) \ +_(vni) \ +_(encap_fib_index) \ +_(decap_next_index) + +static int vxlan_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.as_u32; + ip0->dst_address.as_u32 = t->dst.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); +} + +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; + vnet_hw_interface_t * hi; + uword * p; + u32 hw_if_index = ~0; + u32 sw_if_index = ~0; + int rv; + vxlan_tunnel_key_t key; + + key.src = a->dst.as_u32; /* decap src in key is encap dst in config */ + key.vni = clib_host_to_net_u32 (a->vni << 8); + + p = hash_get (vxm->vxlan_tunnel_by_key, key.as_u64); + + if (a->is_add) + { + /* adding a tunnel: tunnel must not already exist */ + if (p) + return VNET_API_ERROR_TUNNEL_EXIST; + + if (a->decap_next_index == ~0) + a->decap_next_index = VXLAN_INPUT_NEXT_L2_INPUT; + + if (a->decap_next_index >= VXLAN_INPUT_N_NEXT) + 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 _ + + rv = vxlan_rewrite (t); + + if (rv) + { + pool_put (vxm->tunnels, t); + return rv; + } + + hash_set (vxm->vxlan_tunnel_by_key, key.as_u64, t - vxm->tunnels); + + 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); + hi->output_node_index = vxlan_encap_node.index; + } + + t->hw_if_index = hw_if_index; + t->sw_if_index = sw_if_index = hi->sw_if_index; + + if (a->decap_next_index == VXLAN_INPUT_NEXT_L2_INPUT) + { + l2input_main_t * l2im = &l2input_main; + /* 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); + } + 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); + + hash_unset (vxm->vxlan_tunnel_by_key, key.as_u64); + + vec_free (t->rewrite); + t->rewrite = vxlan_dummy_rewrite; + pool_put (vxm->tunnels, t); + } + + if (sw_if_indexp) + *sw_if_indexp = sw_if_index; + + return 0; +} + +static u32 fib_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 uword unformat_decap_next (unformat_input_t * input, va_list * args) +{ + u32 * result = va_arg (*args, u32 *); + u32 tmp; + + if (unformat (input, "l2")) + *result = VXLAN_INPUT_NEXT_L2_INPUT; + else if (unformat (input, "drop")) + *result = VXLAN_INPUT_NEXT_DROP; + else if (unformat (input, "ip4")) + *result = VXLAN_INPUT_NEXT_IP4_INPUT; + else if (unformat (input, "ip6")) + *result = VXLAN_INPUT_NEXT_IP6_INPUT; + 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; + ip4_address_t src, dst; + u8 is_add = 1; + u8 src_set = 0; + u8 dst_set = 0; + u32 encap_fib_index = 0; + u32 decap_next_index = ~0; + u32 vni = 0; + u32 tmp; + int rv; + vnet_vxlan_add_del_tunnel_args_t _a, * a = &_a; + + /* 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)) + src_set = 1; + else if (unformat (line_input, "dst %U", + unformat_ip4_address, &dst)) + dst_set = 1; + else if (unformat (line_input, "encap-vrf-id %d", &tmp)) + { + encap_fib_index = fib_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)) + ; + 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 (vni == 0) + return clib_error_return (0, "vni not specified"); + + memset (a, 0, sizeof (*a)); + + a->is_add = is_add; + +#define _(x) a->x = x; + foreach_copy_field; +#undef _ + + rv = vnet_vxlan_add_del_tunnel (a, 0 /* hw_if_indexp */); + + switch(rv) + { + case 0: + break; + case VNET_API_ERROR_INVALID_DECAP_NEXT: + return clib_error_return (0, "invalid decap-next..."); + + case VNET_API_ERROR_TUNNEL_EXIST: + return clib_error_return (0, "tunnel already exists..."); + + case VNET_API_ERROR_NO_SUCH_ENTRY: + return clib_error_return (0, "tunnel does not exist..."); + + default: + return clib_error_return + (0, "vnet_vxlan_add_del_tunnel returned %d", rv); + } + + return 0; +} + +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> vni <nn>" + " [encap-vrf-id <nn>] [decap-next [l2|ip4|ip6] [del]\n", + .function = vxlan_add_del_tunnel_command_fn, +}; + +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; +} + +VLIB_CLI_COMMAND (show_vxlan_tunnel_command, static) = { + .path = "show vxlan tunnel", + .function = show_vxlan_tunnel_command_fn, +}; + +clib_error_t *vxlan_init (vlib_main_t *vm) +{ + vxlan_main_t * vxm = &vxlan_main; + ip4_vxlan_header_t * hdr; + ip4_header_t * ip; + + vxm->vnet_main = vnet_get_main(); + vxm->vlib_main = vm; + + /* init dummy rewrite string for deleted vxlan tunnels */ + _vec_len(vxlan_dummy_rewrite) = sizeof(ip4_vxlan_header_t); + hdr = (ip4_vxlan_header_t *) vxlan_dummy_rewrite; + ip = &hdr->ip4; + /* minimal rewrite setup, see vxlan_rewite() above as reference */ + ip->ip_version_and_header_length = 0x45; + ip->checksum = ip4_header_checksum (ip); + + udp_register_dst_port (vm, UDP_DST_PORT_vxlan, + vxlan_input_node.index, 1 /* is_ip4 */); + return 0; +} + +VLIB_INIT_FUNCTION(vxlan_init); + diff --git a/vnet/vnet/vxlan/vxlan.h b/vnet/vnet/vxlan/vxlan.h new file mode 100644 index 00000000000..5c82a3dfd9d --- /dev/null +++ b/vnet/vnet/vxlan/vxlan.h @@ -0,0 +1,128 @@ +/* + * 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_bd.h> +#include <vnet/ethernet/ethernet.h> +#include <vnet/vxlan/vxlan_packet.h> +#include <vnet/ip/ip4_packet.h> +#include <vnet/ip/udp.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 { + /* + * 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; + }; +}) vxlan_tunnel_key_t; + +typedef struct { + /* Rewrite string. $$$$ embed vnet_rewrite header */ + u8 * rewrite; + + /* decap next index */ + u32 decap_next_index; + + /* tunnel src and dst addresses */ + ip4_address_t src; + ip4_address_t dst; + + /* vxlan VNI in HOST byte order */ + u32 vni; + + /* L3 FIB index and L2 BD ID */ + u16 encap_fib_index; /* tunnel partner IP lookup here */ + + /* vnet intfc hw/sw_if_index */ + u16 hw_if_index; + u32 sw_if_index; +} vxlan_tunnel_t; + +#define foreach_vxlan_input_next \ +_(DROP, "error-drop") \ +_(L2_INPUT, "l2-input") \ +_(IP4_INPUT, "ip4-input") \ +_(IP6_INPUT, "ip6-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 { + /* vector of encap tunnel instances */ + vxlan_tunnel_t *tunnels; + + /* lookup tunnel by key */ + uword * vxlan_tunnel_by_key; + + /* Free vlib hw_if_indices */ + u32 * free_vxlan_tunnel_hw_if_indices; + + /* Dummy rewrite for deleted vxlan_tunnels with hw_if_indices as above */ + u64 dummy_str [sizeof(ip4_vxlan_header_t)/sizeof(u64) + 2]; +#define vxlan_dummy_rewrite ((u8 *) &vxlan_main.dummy_str[1]) + + /* convenience */ + vlib_main_t * vlib_main; + vnet_main_t * vnet_main; +} vxlan_main_t; + +vxlan_main_t vxlan_main; + +vlib_node_registration_t vxlan_input_node; +vlib_node_registration_t vxlan_encap_node; + +u8 * format_vxlan_encap_trace (u8 * s, va_list * args); + +typedef struct { + u8 is_add; + ip4_address_t src, dst; + 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/vnet/vnet/vxlan/vxlan_error.def b/vnet/vnet/vxlan/vxlan_error.def new file mode 100644 index 00000000000..3ead986cca8 --- /dev/null +++ b/vnet/vnet/vxlan/vxlan_error.def @@ -0,0 +1,16 @@ +/* + * Copyright (c) 2015 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ +vxlan_error (DECAPSULATED, "good packets decapsulated") +vxlan_error (NO_SUCH_TUNNEL, "no such tunnel packets") diff --git a/vnet/vnet/vxlan/vxlan_packet.h b/vnet/vnet/vxlan/vxlan_packet.h new file mode 100644 index 00000000000..8a9a3b80532 --- /dev/null +++ b/vnet/vnet/vxlan/vxlan_packet.h @@ -0,0 +1,66 @@ +/* + * 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 { + u32 flags; + u32 vni_reserved; +} vxlan_header_t; + +#define VXLAN_FLAGS_VALID_HOST_BYTE_ORDER (1<<27) +#define VXLAN_FLAGS_VALID_NET_BYTE_ORDER (clib_host_to_net_u32(1<<27)) + +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); + h->flags = VXLAN_FLAGS_VALID_NET_BYTE_ORDER; +} + +#endif /* __included_vxlan_packet_h__ */ |