blob: dcf480578a7e4ff0c83a1e4436842ad2f814fbc9 [file] [log] [blame]
Ed Warnickecb9cada2015-12-08 15:45:58 -07001/*
2 * Copyright (c) 2015 Cisco and/or its affiliates.
3 * Licensed under the Apache License, Version 2.0 (the "License");
4 * you may not use this file except in compliance with the License.
5 * You may obtain a copy of the License at:
6 *
7 * http://www.apache.org/licenses/LICENSE-2.0
8 *
9 * Unless required by applicable law or agreed to in writing, software
10 * distributed under the License is distributed on an "AS IS" BASIS,
11 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
12 * See the License for the specific language governing permissions and
13 * limitations under the License.
14 */
15#include <vnet/vxlan/vxlan.h>
Chris Luke99cb3352016-04-26 10:49:53 -040016#include <vnet/ip/format.h>
John Loc42912d2016-11-07 18:30:47 -050017#include <vnet/fib/fib_entry.h>
18#include <vnet/fib/fib_table.h>
Neale Ranns1f50bf82019-07-16 15:28:52 +000019#include <vnet/fib/fib_entry_track.h>
Neale Ranns32e1c012016-11-22 17:07:28 +000020#include <vnet/mfib/mfib_table.h>
21#include <vnet/adj/adj_mcast.h>
eyal bari82e21d72018-04-26 13:14:55 +030022#include <vnet/adj/rewrite.h>
Neale Ranns1b5ca982020-12-16 13:06:58 +000023#include <vnet/dpo/drop_dpo.h>
Eyal Bari3212c572017-03-06 11:47:50 +020024#include <vnet/interface.h>
eyal bariaf86a482018-04-17 11:20:27 +030025#include <vnet/flow/flow.h>
Florin Corasb040f982020-10-20 14:59:43 -070026#include <vnet/udp/udp_local.h>
Hongjun Nibeb4bf72016-11-25 00:03:46 +080027#include <vlib/vlib.h>
Ed Warnickecb9cada2015-12-08 15:45:58 -070028
Billy McFallc5d9cda2016-09-14 10:39:58 -040029/**
30 * @file
31 * @brief VXLAN.
32 *
33 * VXLAN provides the features needed to allow L2 bridge domains (BDs)
34 * to span multiple servers. This is done by building an L2 overlay on
35 * top of an L3 network underlay using VXLAN tunnels.
36 *
37 * This makes it possible for servers to be co-located in the same data
38 * center or be separated geographically as long as they are reachable
39 * through the underlay L3 network.
40 *
41 * You can refer to this kind of L2 overlay bridge domain as a VXLAN
42 * (Virtual eXtensible VLAN) segment.
43 */
44
45
Ed Warnickecb9cada2015-12-08 15:45:58 -070046vxlan_main_t vxlan_main;
47
Ed Warnickea4b05412021-01-18 11:56:22 -060048static u32
49vxlan_eth_flag_change (vnet_main_t *vnm, vnet_hw_interface_t *hi, u32 flags)
50{
51 /* nothing for now */
52 return 0;
53}
54
Eyal Bari8f576482018-05-07 10:13:44 +030055static u8 *
56format_decap_next (u8 * s, va_list * args)
Hongjun Nibeb4bf72016-11-25 00:03:46 +080057{
58 u32 next_index = va_arg (*args, u32);
59
John Lo4478d8e2018-01-12 17:15:25 -050060 if (next_index == VXLAN_INPUT_NEXT_DROP)
61 return format (s, "drop");
62 else
63 return format (s, "index %d", next_index);
Hongjun Nibeb4bf72016-11-25 00:03:46 +080064 return s;
65}
66
Eyal Bari8f576482018-05-07 10:13:44 +030067u8 *
68format_vxlan_tunnel (u8 * s, va_list * args)
Ed Warnickecb9cada2015-12-08 15:45:58 -070069{
Eyal Bari8f576482018-05-07 10:13:44 +030070 vxlan_tunnel_t *t = va_arg (*args, vxlan_tunnel_t *);
Jon Loeliger3d460bd2018-02-01 16:36:12 -060071
72 s = format (s,
Artem Glazychev839dcc02020-12-01 02:39:21 +070073 "[%d] instance %d src %U dst %U src_port %d dst_port %d vni %d "
74 "fib-idx %d sw-if-idx %d ",
75 t->dev_instance, t->user_instance, format_ip46_address, &t->src,
76 IP46_TYPE_ANY, format_ip46_address, &t->dst, IP46_TYPE_ANY,
77 t->src_port, t->dst_port, t->vni, t->encap_fib_index,
78 t->sw_if_index);
John Lo56912c82016-12-08 16:10:02 -050079
John Lo4478d8e2018-01-12 17:15:25 -050080 s = format (s, "encap-dpo-idx %d ", t->next_dpo.dpoi_index);
John Lo56912c82016-12-08 16:10:02 -050081
John Lo4478d8e2018-01-12 17:15:25 -050082 if (PREDICT_FALSE (t->decap_next_index != VXLAN_INPUT_NEXT_L2_INPUT))
83 s = format (s, "decap-next-%U ", format_decap_next, t->decap_next_index);
84
85 if (PREDICT_FALSE (ip46_address_is_multicast (&t->dst)))
86 s = format (s, "mcast-sw-if-idx %d ", t->mcast_sw_if_index);
87
eyal bariaf86a482018-04-17 11:20:27 +030088 if (t->flow_index != ~0)
89 s = format (s, "flow-index %d [%U]", t->flow_index,
90 format_flow_enabled_hw, t->flow_index);
91
Ed Warnickecb9cada2015-12-08 15:45:58 -070092 return s;
93}
94
Eyal Bari8f576482018-05-07 10:13:44 +030095static u8 *
96format_vxlan_name (u8 * s, va_list * args)
Ed Warnickecb9cada2015-12-08 15:45:58 -070097{
98 u32 dev_instance = va_arg (*args, u32);
Eyal Bari8f576482018-05-07 10:13:44 +030099 vxlan_main_t *vxm = &vxlan_main;
Jon Loeliger3d460bd2018-02-01 16:36:12 -0600100 vxlan_tunnel_t *t;
101
102 if (dev_instance == ~0)
Eyal Bari8f576482018-05-07 10:13:44 +0300103 return format (s, "<cached-unused>");
Jon Loeliger3d460bd2018-02-01 16:36:12 -0600104
Eyal Bari8f576482018-05-07 10:13:44 +0300105 if (dev_instance >= vec_len (vxm->tunnels))
106 return format (s, "<improperly-referenced>");
Jon Loeliger3d460bd2018-02-01 16:36:12 -0600107
Eyal Bari8f576482018-05-07 10:13:44 +0300108 t = pool_elt_at_index (vxm->tunnels, dev_instance);
Jon Loeliger3d460bd2018-02-01 16:36:12 -0600109
110 return format (s, "vxlan_tunnel%d", t->user_instance);
Ed Warnickecb9cada2015-12-08 15:45:58 -0700111}
112
Pavel Kotucek988a7c42016-02-29 15:03:08 +0100113static clib_error_t *
114vxlan_interface_admin_up_down (vnet_main_t * vnm, u32 hw_if_index, u32 flags)
115{
Eyal Baria93ea422017-02-01 13:36:15 +0200116 u32 hw_flags = (flags & VNET_SW_INTERFACE_FLAG_ADMIN_UP) ?
117 VNET_HW_INTERFACE_FLAG_LINK_UP : 0;
118 vnet_hw_interface_set_flags (vnm, hw_if_index, hw_flags);
Pavel Kotucek988a7c42016-02-29 15:03:08 +0100119
120 return /* no error */ 0;
121}
122
Eyal Bari8f576482018-05-07 10:13:44 +0300123/* *INDENT-OFF* */
124VNET_DEVICE_CLASS (vxlan_device_class, static) = {
Ed Warnickecb9cada2015-12-08 15:45:58 -0700125 .name = "VXLAN",
126 .format_device_name = format_vxlan_name,
127 .format_tx_trace = format_vxlan_encap_trace,
Pavel Kotucek988a7c42016-02-29 15:03:08 +0100128 .admin_up_down_function = vxlan_interface_admin_up_down,
Ed Warnickecb9cada2015-12-08 15:45:58 -0700129};
Eyal Bari8f576482018-05-07 10:13:44 +0300130/* *INDENT-ON* */
Ed Warnickecb9cada2015-12-08 15:45:58 -0700131
Eyal Bari8f576482018-05-07 10:13:44 +0300132static u8 *
133format_vxlan_header_with_length (u8 * s, va_list * args)
Ed Warnickecb9cada2015-12-08 15:45:58 -0700134{
135 u32 dev_instance = va_arg (*args, u32);
136 s = format (s, "unimplemented dev %u", dev_instance);
137 return s;
138}
139
Eyal Bari8f576482018-05-07 10:13:44 +0300140/* *INDENT-OFF* */
Ed Warnickecb9cada2015-12-08 15:45:58 -0700141VNET_HW_INTERFACE_CLASS (vxlan_hw_class) = {
142 .name = "VXLAN",
143 .format_header = format_vxlan_header_with_length,
Neale Rannsb80c5362016-10-08 13:03:40 +0100144 .build_rewrite = default_build_rewrite,
Ed Warnickecb9cada2015-12-08 15:45:58 -0700145};
Eyal Bari8f576482018-05-07 10:13:44 +0300146/* *INDENT-ON* */
Ed Warnickecb9cada2015-12-08 15:45:58 -0700147
Eyal Baric5b13602016-11-24 19:42:43 +0200148static void
Eyal Bari8f576482018-05-07 10:13:44 +0300149vxlan_tunnel_restack_dpo (vxlan_tunnel_t * t)
Eyal Baric5b13602016-11-24 19:42:43 +0200150{
Eyal Bari8f576482018-05-07 10:13:44 +0300151 u8 is_ip4 = ip46_address_is_ip4 (&t->dst);
Eyal Bari99ed4862018-04-25 13:50:57 +0300152 dpo_id_t dpo = DPO_INVALID;
153 fib_forward_chain_type_t forw_type = is_ip4 ?
Eyal Bari8f576482018-05-07 10:13:44 +0300154 FIB_FORW_CHAIN_TYPE_UNICAST_IP4 : FIB_FORW_CHAIN_TYPE_UNICAST_IP6;
Eyal Baric5b13602016-11-24 19:42:43 +0200155
Eyal Bari99ed4862018-04-25 13:50:57 +0300156 fib_entry_contribute_forwarding (t->fib_entry_index, forw_type, &dpo);
157
158 /* vxlan uses the payload hash as the udp source port
159 * hence the packet's hash is unknown
160 * skip single bucket load balance dpo's */
161 while (DPO_LOAD_BALANCE == dpo.dpoi_type)
162 {
Neale Ranns1b5ca982020-12-16 13:06:58 +0000163 const load_balance_t *lb;
164 const dpo_id_t *choice;
165
166 lb = load_balance_get (dpo.dpoi_index);
Eyal Bari99ed4862018-04-25 13:50:57 +0300167 if (lb->lb_n_buckets > 1)
Eyal Bari8f576482018-05-07 10:13:44 +0300168 break;
Eyal Bari99ed4862018-04-25 13:50:57 +0300169
Neale Ranns1b5ca982020-12-16 13:06:58 +0000170 choice = load_balance_get_bucket_i (lb, 0);
171
172 if (DPO_RECEIVE == choice->dpoi_type)
173 dpo_copy (&dpo, drop_dpo_get (choice->dpoi_proto));
174 else
175 dpo_copy (&dpo, choice);
Eyal Bari99ed4862018-04-25 13:50:57 +0300176 }
177
178 u32 encap_index = is_ip4 ?
Eyal Bari8f576482018-05-07 10:13:44 +0300179 vxlan4_encap_node.index : vxlan6_encap_node.index;
Eyal Bari99ed4862018-04-25 13:50:57 +0300180 dpo_stack_from_node (encap_index, &t->next_dpo, &dpo);
Eyal Bari8f576482018-05-07 10:13:44 +0300181 dpo_reset (&dpo);
Eyal Baric5b13602016-11-24 19:42:43 +0200182}
John Loc42912d2016-11-07 18:30:47 -0500183
184static vxlan_tunnel_t *
Eyal Bari8f576482018-05-07 10:13:44 +0300185vxlan_tunnel_from_fib_node (fib_node_t * node)
John Loc42912d2016-11-07 18:30:47 -0500186{
Eyal Bari8f576482018-05-07 10:13:44 +0300187 ASSERT (FIB_NODE_TYPE_VXLAN_TUNNEL == node->fn_type);
188 return ((vxlan_tunnel_t *) (((char *) node) -
189 STRUCT_OFFSET_OF (vxlan_tunnel_t, node)));
John Loc42912d2016-11-07 18:30:47 -0500190}
191
192/**
193 * Function definition to backwalk a FIB node -
194 * Here we will restack the new dpo of VXLAN DIP to encap node.
195 */
196static fib_node_back_walk_rc_t
Eyal Bari8f576482018-05-07 10:13:44 +0300197vxlan_tunnel_back_walk (fib_node_t * node, fib_node_back_walk_ctx_t * ctx)
John Loc42912d2016-11-07 18:30:47 -0500198{
Eyal Bari8f576482018-05-07 10:13:44 +0300199 vxlan_tunnel_restack_dpo (vxlan_tunnel_from_fib_node (node));
200 return (FIB_NODE_BACK_WALK_CONTINUE);
John Loc42912d2016-11-07 18:30:47 -0500201}
202
203/**
204 * Function definition to get a FIB node from its index
205 */
Eyal Bari8f576482018-05-07 10:13:44 +0300206static fib_node_t *
John Loc42912d2016-11-07 18:30:47 -0500207vxlan_tunnel_fib_node_get (fib_node_index_t index)
208{
Eyal Bari8f576482018-05-07 10:13:44 +0300209 vxlan_tunnel_t *t;
210 vxlan_main_t *vxm = &vxlan_main;
John Loc42912d2016-11-07 18:30:47 -0500211
Eyal Bari8f576482018-05-07 10:13:44 +0300212 t = pool_elt_at_index (vxm->tunnels, index);
John Loc42912d2016-11-07 18:30:47 -0500213
Eyal Bari8f576482018-05-07 10:13:44 +0300214 return (&t->node);
John Loc42912d2016-11-07 18:30:47 -0500215}
216
217/**
218 * Function definition to inform the FIB node that its last lock has gone.
219 */
220static void
Eyal Bari8f576482018-05-07 10:13:44 +0300221vxlan_tunnel_last_lock_gone (fib_node_t * node)
John Loc42912d2016-11-07 18:30:47 -0500222{
Eyal Bari8f576482018-05-07 10:13:44 +0300223 /*
224 * The VXLAN tunnel is a root of the graph. As such
225 * it never has children and thus is never locked.
226 */
227 ASSERT (0);
John Loc42912d2016-11-07 18:30:47 -0500228}
229
230/*
231 * Virtual function table registered by VXLAN tunnels
232 * for participation in the FIB object graph.
233 */
234const static fib_node_vft_t vxlan_vft = {
Eyal Bari8f576482018-05-07 10:13:44 +0300235 .fnv_get = vxlan_tunnel_fib_node_get,
236 .fnv_last_lock = vxlan_tunnel_last_lock_gone,
237 .fnv_back_walk = vxlan_tunnel_back_walk,
John Loc42912d2016-11-07 18:30:47 -0500238};
239
Artem Glazychev839dcc02020-12-01 02:39:21 +0700240#define foreach_copy_field \
241 _ (vni) \
242 _ (mcast_sw_if_index) \
243 _ (encap_fib_index) \
244 _ (decap_next_index) \
245 _ (src) \
246 _ (dst) \
247 _ (src_port) \
248 _ (dst_port)
Chris Luke99cb3352016-04-26 10:49:53 -0400249
Eyal Bari554075a2018-02-13 15:17:17 +0200250static void
Eyal Baria93ea422017-02-01 13:36:15 +0200251vxlan_rewrite (vxlan_tunnel_t * t, bool is_ip6)
Ed Warnickecb9cada2015-12-08 15:45:58 -0700252{
Eyal Bari8f576482018-05-07 10:13:44 +0300253 union
254 {
eyal bari82e21d72018-04-26 13:14:55 +0300255 ip4_vxlan_header_t h4;
256 ip6_vxlan_header_t h6;
Matthew Smithec8aea12018-05-02 09:55:01 -0500257 } h;
eyal bari82e21d72018-04-26 13:14:55 +0300258 int len = is_ip6 ? sizeof h.h6 : sizeof h.h4;
Ed Warnickecb9cada2015-12-08 15:45:58 -0700259
Eyal Bari8f576482018-05-07 10:13:44 +0300260 udp_header_t *udp;
261 vxlan_header_t *vxlan;
Eyal Baria93ea422017-02-01 13:36:15 +0200262 /* Fixed portion of the (outer) ip header */
Matthew Smithec8aea12018-05-02 09:55:01 -0500263
Dave Barachb7b92992018-10-17 10:38:51 -0400264 clib_memset (&h, 0, sizeof (h));
Eyal Bari8f576482018-05-07 10:13:44 +0300265 if (!is_ip6)
Hongjun Nibeb4bf72016-11-25 00:03:46 +0800266 {
Eyal Bari8f576482018-05-07 10:13:44 +0300267 ip4_header_t *ip = &h.h4.ip4;
eyal bari82e21d72018-04-26 13:14:55 +0300268 udp = &h.h4.udp, vxlan = &h.h4.vxlan;
Eyal Baria93ea422017-02-01 13:36:15 +0200269 ip->ip_version_and_header_length = 0x45;
270 ip->ttl = 254;
271 ip->protocol = IP_PROTOCOL_UDP;
Eyal Bari8f576482018-05-07 10:13:44 +0300272
Eyal Baria93ea422017-02-01 13:36:15 +0200273 ip->src_address = t->src.ip4;
274 ip->dst_address = t->dst.ip4;
275
276 /* we fix up the ip4 header length and checksum after-the-fact */
277 ip->checksum = ip4_header_checksum (ip);
Hongjun Nibeb4bf72016-11-25 00:03:46 +0800278 }
279 else
280 {
Eyal Bari8f576482018-05-07 10:13:44 +0300281 ip6_header_t *ip = &h.h6.ip6;
eyal bari82e21d72018-04-26 13:14:55 +0300282 udp = &h.h6.udp, vxlan = &h.h6.vxlan;
Eyal Bari8f576482018-05-07 10:13:44 +0300283 ip->ip_version_traffic_class_and_flow_label =
284 clib_host_to_net_u32 (6 << 28);
Eyal Baria93ea422017-02-01 13:36:15 +0200285 ip->hop_limit = 255;
286 ip->protocol = IP_PROTOCOL_UDP;
Eyal Bari8f576482018-05-07 10:13:44 +0300287
Eyal Baria93ea422017-02-01 13:36:15 +0200288 ip->src_address = t->src.ip6;
289 ip->dst_address = t->dst.ip6;
Hongjun Nibeb4bf72016-11-25 00:03:46 +0800290 }
291
Eyal Baria93ea422017-02-01 13:36:15 +0200292 /* UDP header, randomize src port on something, maybe? */
Artem Glazychev839dcc02020-12-01 02:39:21 +0700293 udp->src_port = clib_host_to_net_u16 (t->src_port);
294 udp->dst_port = clib_host_to_net_u16 (t->dst_port);
Eyal Baria93ea422017-02-01 13:36:15 +0200295
296 /* VXLAN header */
Eyal Bari8f576482018-05-07 10:13:44 +0300297 vnet_set_vni_and_flags (vxlan, t->vni);
eyal bari82e21d72018-04-26 13:14:55 +0300298 vnet_rewrite_set_data (*t, &h, len);
Eyal Baria93ea422017-02-01 13:36:15 +0200299}
300
301static bool
Eyal Bari8f576482018-05-07 10:13:44 +0300302vxlan_decap_next_is_valid (vxlan_main_t * vxm, u32 is_ip6,
303 u32 decap_next_index)
Eyal Baria93ea422017-02-01 13:36:15 +0200304{
Eyal Bari8f576482018-05-07 10:13:44 +0300305 vlib_main_t *vm = vxm->vlib_main;
306 u32 input_idx = (!is_ip6) ?
307 vxlan4_input_node.index : vxlan6_input_node.index;
Eyal Baria93ea422017-02-01 13:36:15 +0200308 vlib_node_runtime_t *r = vlib_node_get_runtime (vm, input_idx);
309
310 return decap_next_index < r->n_next_nodes;
Hongjun Nibeb4bf72016-11-25 00:03:46 +0800311}
312
Eyal Bari8f576482018-05-07 10:13:44 +0300313/* *INDENT-OFF* */
314typedef CLIB_PACKED(union
315{
316 struct
317 {
Neale Ranns32e1c012016-11-22 17:07:28 +0000318 fib_node_index_t mfib_entry_index;
Eyal Bari0ded8512017-01-19 17:01:09 +0200319 adj_index_t mcast_adj_index;
320 };
321 u64 as_u64;
322}) mcast_shared_t;
Eyal Bari8f576482018-05-07 10:13:44 +0300323/* *INDENT-ON* */
Eyal Bari0ded8512017-01-19 17:01:09 +0200324
325static inline mcast_shared_t
Eyal Bari8f576482018-05-07 10:13:44 +0300326mcast_shared_get (ip46_address_t * ip)
Eyal Barifff67c82016-12-21 12:45:47 +0200327{
Eyal Bari8f576482018-05-07 10:13:44 +0300328 ASSERT (ip46_address_is_multicast (ip));
329 uword *p = hash_get_mem (vxlan_main.mcast_shared, ip);
Dave Barach47d41ad2020-02-17 09:13:26 -0500330 ALWAYS_ASSERT (p);
Eyal Bari8f576482018-05-07 10:13:44 +0300331 mcast_shared_t ret = {.as_u64 = *p };
332 return ret;
Eyal Barifff67c82016-12-21 12:45:47 +0200333}
334
Eyal Bari0ded8512017-01-19 17:01:09 +0200335static inline void
Eyal Bari8f576482018-05-07 10:13:44 +0300336mcast_shared_add (ip46_address_t * dst, fib_node_index_t mfei, adj_index_t ai)
Neale Ranns32e1c012016-11-22 17:07:28 +0000337{
Eyal Bari8f576482018-05-07 10:13:44 +0300338 mcast_shared_t new_ep = {
339 .mcast_adj_index = ai,
340 .mfib_entry_index = mfei,
341 };
Neale Ranns32e1c012016-11-22 17:07:28 +0000342
Eyal Bari8f576482018-05-07 10:13:44 +0300343 hash_set_mem_alloc (&vxlan_main.mcast_shared, dst, new_ep.as_u64);
Neale Ranns32e1c012016-11-22 17:07:28 +0000344}
345
346static inline void
Eyal Bari8f576482018-05-07 10:13:44 +0300347mcast_shared_remove (ip46_address_t * dst)
Neale Ranns32e1c012016-11-22 17:07:28 +0000348{
Eyal Bari8f576482018-05-07 10:13:44 +0300349 mcast_shared_t ep = mcast_shared_get (dst);
Neale Ranns32e1c012016-11-22 17:07:28 +0000350
Eyal Bari8f576482018-05-07 10:13:44 +0300351 adj_unlock (ep.mcast_adj_index);
352 mfib_table_entry_delete_index (ep.mfib_entry_index, MFIB_SOURCE_VXLAN);
Neale Ranns32e1c012016-11-22 17:07:28 +0000353
Eyal Bari8f576482018-05-07 10:13:44 +0300354 hash_unset_mem_free (&vxlan_main.mcast_shared, dst);
Eyal Barifff67c82016-12-21 12:45:47 +0200355}
356
Eyal Bari8f576482018-05-07 10:13:44 +0300357int vnet_vxlan_add_del_tunnel
358 (vnet_vxlan_add_del_tunnel_args_t * a, u32 * sw_if_indexp)
Ed Warnickecb9cada2015-12-08 15:45:58 -0700359{
Eyal Bari8f576482018-05-07 10:13:44 +0300360 vxlan_main_t *vxm = &vxlan_main;
Eyal Bari8f576482018-05-07 10:13:44 +0300361 vnet_main_t *vnm = vxm->vnet_main;
Eyal Bariefd9cf32018-10-02 12:23:06 +0300362 vxlan_decap_info_t *p;
Ed Warnickecb9cada2015-12-08 15:45:58 -0700363 u32 sw_if_index = ~0;
Chris Luke99cb3352016-04-26 10:49:53 -0400364 vxlan4_tunnel_key_t key4;
365 vxlan6_tunnel_key_t key6;
John Loc42912d2016-11-07 18:30:47 -0500366 u32 is_ip6 = a->is_ip6;
Ed Warnickea4b05412021-01-18 11:56:22 -0600367 vlib_main_t *vm = vlib_get_main ();
368 u8 hw_addr[6];
Ed Warnickecb9cada2015-12-08 15:45:58 -0700369
Artem Glazychev839dcc02020-12-01 02:39:21 +0700370 /* Set udp-ports */
371 if (a->src_port == 0)
372 a->src_port = is_ip6 ? UDP_DST_PORT_vxlan6 : UDP_DST_PORT_vxlan;
373
374 if (a->dst_port == 0)
375 a->dst_port = is_ip6 ? UDP_DST_PORT_vxlan6 : UDP_DST_PORT_vxlan;
376
Eyal Baridd47eca2018-07-08 08:15:56 +0300377 int not_found;
John Lo56912c82016-12-08 16:10:02 -0500378 if (!is_ip6)
379 {
Eyal Bariefd9cf32018-10-02 12:23:06 +0300380 /* ip4 mcast is indexed by mcast addr only */
381 key4.key[0] = ip46_address_is_multicast (&a->dst) ?
Artem Glazychev839dcc02020-12-01 02:39:21 +0700382 a->dst.ip4.as_u32 :
383 a->dst.ip4.as_u32 | (((u64) a->src.ip4.as_u32) << 32);
384 key4.key[1] = ((u64) clib_host_to_net_u16 (a->src_port) << 48) |
385 (((u64) a->encap_fib_index) << 32) |
386 clib_host_to_net_u32 (a->vni << 8);
Eyal Baridd47eca2018-07-08 08:15:56 +0300387 not_found =
388 clib_bihash_search_inline_16_8 (&vxm->vxlan4_tunnel_by_key, &key4);
Eyal Bariefd9cf32018-10-02 12:23:06 +0300389 p = (void *) &key4.value;
Eyal Bari8f576482018-05-07 10:13:44 +0300390 }
391 else
John Lo56912c82016-12-08 16:10:02 -0500392 {
Eyal Bari0fa56782018-06-04 12:25:05 +0300393 key6.key[0] = a->dst.ip6.as_u64[0];
394 key6.key[1] = a->dst.ip6.as_u64[1];
Artem Glazychev839dcc02020-12-01 02:39:21 +0700395 key6.key[2] = (((u64) clib_host_to_net_u16 (a->src_port) << 48) |
396 ((u64) a->encap_fib_index) << 32) |
397 clib_host_to_net_u32 (a->vni << 8);
Eyal Baridd47eca2018-07-08 08:15:56 +0300398 not_found =
399 clib_bihash_search_inline_24_8 (&vxm->vxlan6_tunnel_by_key, &key6);
Eyal Bariefd9cf32018-10-02 12:23:06 +0300400 p = (void *) &key6.value;
John Lo56912c82016-12-08 16:10:02 -0500401 }
Jon Loeliger3d460bd2018-02-01 16:36:12 -0600402
Eyal Baridd47eca2018-07-08 08:15:56 +0300403 if (not_found)
404 p = 0;
405
Ed Warnickecb9cada2015-12-08 15:45:58 -0700406 if (a->is_add)
407 {
Eyal Bari8f576482018-05-07 10:13:44 +0300408 l2input_main_t *l2im = &l2input_main;
Jon Loeliger3d460bd2018-02-01 16:36:12 -0600409 u32 dev_instance; /* real dev instance tunnel index */
410 u32 user_instance; /* request and actual instance number */
John Loc42912d2016-11-07 18:30:47 -0500411
Ed Warnickecb9cada2015-12-08 15:45:58 -0700412 /* adding a tunnel: tunnel must not already exist */
413 if (p)
Eyal Bari8f576482018-05-07 10:13:44 +0300414 return VNET_API_ERROR_TUNNEL_EXIST;
Ed Warnickecb9cada2015-12-08 15:45:58 -0700415
Hongjun Nibeb4bf72016-11-25 00:03:46 +0800416 /*if not set explicitly, default to l2 */
Eyal Bari8f576482018-05-07 10:13:44 +0300417 if (a->decap_next_index == ~0)
Hongjun Nibeb4bf72016-11-25 00:03:46 +0800418 a->decap_next_index = VXLAN_INPUT_NEXT_L2_INPUT;
Eyal Bari8f576482018-05-07 10:13:44 +0300419 if (!vxlan_decap_next_is_valid (vxm, is_ip6, a->decap_next_index))
420 return VNET_API_ERROR_INVALID_DECAP_NEXT;
Hongjun Nibeb4bf72016-11-25 00:03:46 +0800421
Eyal Bariefd9cf32018-10-02 12:23:06 +0300422 vxlan_tunnel_t *t;
Ed Warnickecb9cada2015-12-08 15:45:58 -0700423 pool_get_aligned (vxm->tunnels, t, CLIB_CACHE_LINE_BYTES);
Dave Barachb7b92992018-10-17 10:38:51 -0400424 clib_memset (t, 0, sizeof (*t));
Jon Loeliger3d460bd2018-02-01 16:36:12 -0600425 dev_instance = t - vxm->tunnels;
426
Ed Warnickecb9cada2015-12-08 15:45:58 -0700427 /* copy from arg structure */
428#define _(x) t->x = a->x;
429 foreach_copy_field;
430#undef _
Chris Luke99cb3352016-04-26 10:49:53 -0400431
Eyal Bari554075a2018-02-13 15:17:17 +0200432 vxlan_rewrite (t, is_ip6);
Jon Loeliger3d460bd2018-02-01 16:36:12 -0600433 /*
434 * Reconcile the real dev_instance and a possible requested instance.
435 */
436 user_instance = a->instance;
Eyal Bari8f576482018-05-07 10:13:44 +0300437 if (user_instance == ~0)
Jon Loeliger3d460bd2018-02-01 16:36:12 -0600438 user_instance = dev_instance;
439 if (hash_get (vxm->instance_used, user_instance))
440 {
441 pool_put (vxm->tunnels, t);
442 return VNET_API_ERROR_INSTANCE_IN_USE;
443 }
Ed Warnickea4b05412021-01-18 11:56:22 -0600444
Ray Kinsella15cb3da2021-01-28 17:09:45 +0000445 hash_set (vxm->instance_used, user_instance, 1);
446
447 t->dev_instance = dev_instance; /* actual */
448 t->user_instance = user_instance; /* name */
449 t->flow_index = ~0;
450
Nathan Skrzypczak3e384222021-04-21 19:56:34 +0200451 if (a->is_l3)
452 t->hw_if_index =
453 vnet_register_interface (vnm, vxlan_device_class.index, dev_instance,
454 vxlan_hw_class.index, dev_instance);
455 else
Ed Warnickea4b05412021-01-18 11:56:22 -0600456 {
Nathan Skrzypczak3e384222021-04-21 19:56:34 +0200457 f64 now = vlib_time_now (vm);
458 u32 rnd;
459 rnd = (u32) (now * 1e6);
460 rnd = random_u32 (&rnd);
461 memcpy (hw_addr + 2, &rnd, sizeof (rnd));
462 hw_addr[0] = 2;
463 hw_addr[1] = 0xfe;
464 if (ethernet_register_interface (
465 vnm, vxlan_device_class.index, dev_instance, hw_addr,
466 &t->hw_if_index, vxlan_eth_flag_change))
467 {
468 hash_unset (vxm->instance_used, t->user_instance);
Ray Kinsella15cb3da2021-01-28 17:09:45 +0000469
Nathan Skrzypczak3e384222021-04-21 19:56:34 +0200470 pool_put (vxm->tunnels, t);
471 return VNET_API_ERROR_SYSCALL_ERROR_2;
472 }
Ed Warnickea4b05412021-01-18 11:56:22 -0600473 }
474
Eyal Bari8f576482018-05-07 10:13:44 +0300475 vnet_hw_interface_t *hi = vnet_get_hw_interface (vnm, t->hw_if_index);
John Loe5453d02018-01-23 19:21:34 -0500476
477 /* Set vxlan tunnel output node */
478 u32 encap_index = !is_ip6 ?
Eyal Bari8f576482018-05-07 10:13:44 +0300479 vxlan4_encap_node.index : vxlan6_encap_node.index;
Eyal Bari554075a2018-02-13 15:17:17 +0200480 vnet_set_interface_output_node (vnm, t->hw_if_index, encap_index);
John Loe5453d02018-01-23 19:21:34 -0500481
Ed Warnickecb9cada2015-12-08 15:45:58 -0700482 t->sw_if_index = sw_if_index = hi->sw_if_index;
John Loe5453d02018-01-23 19:21:34 -0500483
Eyal Bariefd9cf32018-10-02 12:23:06 +0300484 /* copy the key */
485 int add_failed;
486 if (is_ip6)
487 {
488 key6.value = (u64) dev_instance;
489 add_failed = clib_bihash_add_del_24_8 (&vxm->vxlan6_tunnel_by_key,
490 &key6, 1 /*add */ );
491 }
492 else
493 {
494 vxlan_decap_info_t di = {.sw_if_index = t->sw_if_index, };
495 if (ip46_address_is_multicast (&t->dst))
496 di.local_ip = t->src.ip4;
497 else
498 di.next_index = t->decap_next_index;
499 key4.value = di.as_u64;
500 add_failed = clib_bihash_add_del_16_8 (&vxm->vxlan4_tunnel_by_key,
501 &key4, 1 /*add */ );
502 }
503
504 if (add_failed)
505 {
Nathan Skrzypczak3e384222021-04-21 19:56:34 +0200506 if (a->is_l3)
507 vnet_delete_hw_interface (vnm, t->hw_if_index);
508 else
509 ethernet_delete_interface (vnm, t->hw_if_index);
Eyal Bariefd9cf32018-10-02 12:23:06 +0300510 hash_unset (vxm->instance_used, t->user_instance);
511 pool_put (vxm->tunnels, t);
512 return VNET_API_ERROR_INVALID_REGISTRATION;
513 }
514
Eyal Bari8f576482018-05-07 10:13:44 +0300515 vec_validate_init_empty (vxm->tunnel_index_by_sw_if_index, sw_if_index,
516 ~0);
Jon Loeliger3d460bd2018-02-01 16:36:12 -0600517 vxm->tunnel_index_by_sw_if_index[sw_if_index] = dev_instance;
Dave Wallace60231f32015-12-17 21:04:30 -0500518
John Loc42912d2016-11-07 18:30:47 -0500519 /* setup l2 input config with l2 feature and bd 0 to drop packet */
520 vec_validate (l2im->configs, sw_if_index);
521 l2im->configs[sw_if_index].feature_bitmap = L2INPUT_FEAT_DROP;
522 l2im->configs[sw_if_index].bd_index = 0;
John Loe5453d02018-01-23 19:21:34 -0500523
Eyal Bari8f576482018-05-07 10:13:44 +0300524 vnet_sw_interface_t *si = vnet_get_sw_interface (vnm, sw_if_index);
Eyal Bari3212c572017-03-06 11:47:50 +0200525 si->flags &= ~VNET_SW_INTERFACE_FLAG_HIDDEN;
John Loe5453d02018-01-23 19:21:34 -0500526 vnet_sw_interface_set_flags (vnm, sw_if_index,
Eyal Bari8f576482018-05-07 10:13:44 +0300527 VNET_SW_INTERFACE_FLAG_ADMIN_UP);
Eyal Bari3212c572017-03-06 11:47:50 +0200528
Eyal Bari8f576482018-05-07 10:13:44 +0300529 fib_node_init (&t->node, FIB_NODE_TYPE_VXLAN_TUNNEL);
Eyal Baric5b13602016-11-24 19:42:43 +0200530 fib_prefix_t tun_dst_pfx;
Eyal Baric5b13602016-11-24 19:42:43 +0200531 vnet_flood_class_t flood_class = VNET_FLOOD_CLASS_TUNNEL_NORMAL;
John Loc42912d2016-11-07 18:30:47 -0500532
Eyal Bari8f576482018-05-07 10:13:44 +0300533 fib_prefix_from_ip46_addr (&t->dst, &tun_dst_pfx);
534 if (!ip46_address_is_multicast (&t->dst))
535 {
536 /* Unicast tunnel -
537 * source the FIB entry for the tunnel's destination
538 * and become a child thereof. The tunnel will then get poked
539 * when the forwarding for the entry updates, and the tunnel can
540 * re-stack accordingly
541 */
Nick Zavaritsky27518c22020-02-27 15:54:58 +0000542 vtep_addr_ref (&vxm->vtep_table, t->encap_fib_index, &t->src);
Neale Ranns1f50bf82019-07-16 15:28:52 +0000543 t->fib_entry_index = fib_entry_track (t->encap_fib_index,
544 &tun_dst_pfx,
545 FIB_NODE_TYPE_VXLAN_TUNNEL,
546 dev_instance,
547 &t->sibling_index);
Eyal Bari8f576482018-05-07 10:13:44 +0300548 vxlan_tunnel_restack_dpo (t);
549 }
Eyal Barifff67c82016-12-21 12:45:47 +0200550 else
Eyal Bari8f576482018-05-07 10:13:44 +0300551 {
552 /* Multicast tunnel -
Paul Vinciguerrabdc0e6b2018-09-22 05:32:50 -0700553 * as the same mcast group can be used for multiple mcast tunnels
554 * with different VNIs, create the output fib adjacency only if
John Lo56912c82016-12-08 16:10:02 -0500555 * it does not already exist
556 */
Eyal Bari8f576482018-05-07 10:13:44 +0300557 fib_protocol_t fp = fib_ip_proto (is_ip6);
Eyal Baric5b13602016-11-24 19:42:43 +0200558
Nick Zavaritsky27518c22020-02-27 15:54:58 +0000559 if (vtep_addr_ref (&vxm->vtep_table,
560 t->encap_fib_index, &t->dst) == 1)
Eyal Bari8f576482018-05-07 10:13:44 +0300561 {
562 fib_node_index_t mfei;
563 adj_index_t ai;
564 fib_route_path_t path = {
565 .frp_proto = fib_proto_to_dpo (fp),
566 .frp_addr = zero_addr,
567 .frp_sw_if_index = 0xffffffff,
568 .frp_fib_index = ~0,
Neale Ranns097fa662018-05-01 05:17:55 -0700569 .frp_weight = 1,
Eyal Bari8f576482018-05-07 10:13:44 +0300570 .frp_flags = FIB_ROUTE_PATH_LOCAL,
Neale Ranns097fa662018-05-01 05:17:55 -0700571 .frp_mitf_flags = MFIB_ITF_FLAG_FORWARD,
Eyal Bari8f576482018-05-07 10:13:44 +0300572 };
573 const mfib_prefix_t mpfx = {
574 .fp_proto = fp,
575 .fp_len = (is_ip6 ? 128 : 32),
576 .fp_grp_addr = tun_dst_pfx.fp_addr,
577 };
Eyal Barifff67c82016-12-21 12:45:47 +0200578
Eyal Bari8f576482018-05-07 10:13:44 +0300579 /*
580 * Setup the (*,G) to receive traffic on the mcast group
581 * - the forwarding interface is for-us
582 * - the accepting interface is that from the API
583 */
584 mfib_table_entry_path_update (t->encap_fib_index,
Neale Ranns097fa662018-05-01 05:17:55 -0700585 &mpfx, MFIB_SOURCE_VXLAN, &path);
Neale Ranns32e1c012016-11-22 17:07:28 +0000586
Eyal Bari8f576482018-05-07 10:13:44 +0300587 path.frp_sw_if_index = a->mcast_sw_if_index;
588 path.frp_flags = FIB_ROUTE_PATH_FLAG_NONE;
Neale Ranns097fa662018-05-01 05:17:55 -0700589 path.frp_mitf_flags = MFIB_ITF_FLAG_ACCEPT;
Eyal Bari8f576482018-05-07 10:13:44 +0300590 mfei = mfib_table_entry_path_update (t->encap_fib_index,
591 &mpfx,
Neale Ranns097fa662018-05-01 05:17:55 -0700592 MFIB_SOURCE_VXLAN, &path);
Neale Ranns32e1c012016-11-22 17:07:28 +0000593
Eyal Bari8f576482018-05-07 10:13:44 +0300594 /*
595 * Create the mcast adjacency to send traffic to the group
596 */
597 ai = adj_mcast_add_or_lock (fp,
598 fib_proto_to_link (fp),
599 a->mcast_sw_if_index);
Neale Ranns32e1c012016-11-22 17:07:28 +0000600
Eyal Bari8f576482018-05-07 10:13:44 +0300601 /*
602 * create a new end-point
603 */
604 mcast_shared_add (&t->dst, mfei, ai);
605 }
Neale Ranns32e1c012016-11-22 17:07:28 +0000606
Eyal Bari8f576482018-05-07 10:13:44 +0300607 dpo_id_t dpo = DPO_INVALID;
608 mcast_shared_t ep = mcast_shared_get (&t->dst);
Neale Ranns32e1c012016-11-22 17:07:28 +0000609
Eyal Bari8f576482018-05-07 10:13:44 +0300610 /* Stack shared mcast dst mac addr rewrite on encap */
611 dpo_set (&dpo, DPO_ADJACENCY_MCAST,
612 fib_proto_to_dpo (fp), ep.mcast_adj_index);
Neale Ranns32e1c012016-11-22 17:07:28 +0000613
Eyal Bari8f576482018-05-07 10:13:44 +0300614 dpo_stack_from_node (encap_index, &t->next_dpo, &dpo);
615 dpo_reset (&dpo);
Eyal Barifff67c82016-12-21 12:45:47 +0200616 flood_class = VNET_FLOOD_CLASS_TUNNEL_MASTER;
John Lo56912c82016-12-08 16:10:02 -0500617 }
618
Eyal Bari8f576482018-05-07 10:13:44 +0300619 vnet_get_sw_interface (vnet_get_main (), sw_if_index)->flood_class =
620 flood_class;
Ed Warnickecb9cada2015-12-08 15:45:58 -0700621 }
622 else
623 {
624 /* deleting a tunnel: tunnel must exist */
John Lo37682e12016-11-30 12:51:39 -0500625 if (!p)
Eyal Bari8f576482018-05-07 10:13:44 +0300626 return VNET_API_ERROR_NO_SUCH_ENTRY;
Ed Warnickecb9cada2015-12-08 15:45:58 -0700627
Eyal Barif8d5e212018-10-14 10:54:32 +0300628 u32 instance = is_ip6 ? key6.value :
629 vxm->tunnel_index_by_sw_if_index[p->sw_if_index];
Eyal Bariefd9cf32018-10-02 12:23:06 +0300630 vxlan_tunnel_t *t = pool_elt_at_index (vxm->tunnels, instance);
Ed Warnickecb9cada2015-12-08 15:45:58 -0700631
Eyal Baribc799c92017-04-06 03:26:59 +0300632 sw_if_index = t->sw_if_index;
Eyal Bari8f576482018-05-07 10:13:44 +0300633 vnet_sw_interface_set_flags (vnm, sw_if_index, 0 /* down */ );
Ed Warnickecb9cada2015-12-08 15:45:58 -0700634
Jon Loeliger3d460bd2018-02-01 16:36:12 -0600635 vxm->tunnel_index_by_sw_if_index[sw_if_index] = ~0;
Dave Wallace60231f32015-12-17 21:04:30 -0500636
John Loc42912d2016-11-07 18:30:47 -0500637 if (!is_ip6)
Eyal Baridd47eca2018-07-08 08:15:56 +0300638 clib_bihash_add_del_16_8 (&vxm->vxlan4_tunnel_by_key, &key4,
639 0 /*del */ );
Chris Luke99cb3352016-04-26 10:49:53 -0400640 else
Eyal Baridd47eca2018-07-08 08:15:56 +0300641 clib_bihash_add_del_24_8 (&vxm->vxlan6_tunnel_by_key, &key6,
Eyal Bari0fa56782018-06-04 12:25:05 +0300642 0 /*del */ );
Eyal Barifff67c82016-12-21 12:45:47 +0200643
Eyal Bari8f576482018-05-07 10:13:44 +0300644 if (!ip46_address_is_multicast (&t->dst))
645 {
eyal bariaf86a482018-04-17 11:20:27 +0300646 if (t->flow_index != ~0)
647 vnet_flow_del (vnm, t->flow_index);
648
Nick Zavaritsky27518c22020-02-27 15:54:58 +0000649 vtep_addr_unref (&vxm->vtep_table, t->encap_fib_index, &t->src);
Neale Ranns1f50bf82019-07-16 15:28:52 +0000650 fib_entry_untrack (t->fib_entry_index, t->sibling_index);
Eyal Bari8f576482018-05-07 10:13:44 +0300651 }
Nick Zavaritsky27518c22020-02-27 15:54:58 +0000652 else if (vtep_addr_unref (&vxm->vtep_table,
653 t->encap_fib_index, &t->dst) == 0)
Eyal Bari8f576482018-05-07 10:13:44 +0300654 {
655 mcast_shared_remove (&t->dst);
656 }
Eyal Barifff67c82016-12-21 12:45:47 +0200657
Nathan Skrzypczak3e384222021-04-21 19:56:34 +0200658 vnet_hw_interface_t *hw = vnet_get_hw_interface (vnm, t->hw_if_index);
659 if (hw->dev_class_index == vxlan_device_class.index)
660 vnet_delete_hw_interface (vnm, t->hw_if_index);
661 else
662 ethernet_delete_interface (vnm, t->hw_if_index);
Jon Loeliger25ef2b52018-02-23 17:02:41 -0600663 hash_unset (vxm->instance_used, t->user_instance);
Jon Loeliger3d460bd2018-02-01 16:36:12 -0600664
Eyal Bari8f576482018-05-07 10:13:44 +0300665 fib_node_deinit (&t->node);
Ed Warnickecb9cada2015-12-08 15:45:58 -0700666 pool_put (vxm->tunnels, t);
667 }
668
669 if (sw_if_indexp)
Eyal Bari8f576482018-05-07 10:13:44 +0300670 *sw_if_indexp = sw_if_index;
Ed Warnickecb9cada2015-12-08 15:45:58 -0700671
Jakub Grajciardf3ca232019-06-04 13:16:42 +0200672 if (a->is_add)
673 {
674 /* register udp ports */
Artem Glazychev839dcc02020-12-01 02:39:21 +0700675 if (!is_ip6 && !udp_is_valid_dst_port (a->src_port, 1))
676 udp_register_dst_port (vxm->vlib_main, a->src_port,
Jakub Grajciardf3ca232019-06-04 13:16:42 +0200677 vxlan4_input_node.index, 1);
Artem Glazychev839dcc02020-12-01 02:39:21 +0700678 if (is_ip6 && !udp_is_valid_dst_port (a->src_port, 0))
679 udp_register_dst_port (vxm->vlib_main, a->src_port,
Jakub Grajciardf3ca232019-06-04 13:16:42 +0200680 vxlan6_input_node.index, 0);
681 }
682
Ed Warnickecb9cada2015-12-08 15:45:58 -0700683 return 0;
684}
685
Eyal Bari8f576482018-05-07 10:13:44 +0300686static uword
687get_decap_next_for_node (u32 node_index, u32 ipv4_set)
Hongjun Nibeb4bf72016-11-25 00:03:46 +0800688{
Eyal Bari8f576482018-05-07 10:13:44 +0300689 vxlan_main_t *vxm = &vxlan_main;
690 vlib_main_t *vm = vxm->vlib_main;
Eyal Bari0765c6e2017-01-29 12:40:50 +0200691 uword input_node = (ipv4_set) ? vxlan4_input_node.index :
692 vxlan6_input_node.index;
Hongjun Nibeb4bf72016-11-25 00:03:46 +0800693
Eyal Bari0765c6e2017-01-29 12:40:50 +0200694 return vlib_node_add_next (vm, input_node, node_index);
Hongjun Nibeb4bf72016-11-25 00:03:46 +0800695}
696
Eyal Bari8f576482018-05-07 10:13:44 +0300697static uword
698unformat_decap_next (unformat_input_t * input, va_list * args)
Ed Warnickecb9cada2015-12-08 15:45:58 -0700699{
Eyal Bari8f576482018-05-07 10:13:44 +0300700 u32 *result = va_arg (*args, u32 *);
Hongjun Nibeb4bf72016-11-25 00:03:46 +0800701 u32 ipv4_set = va_arg (*args, int);
Eyal Bari8f576482018-05-07 10:13:44 +0300702 vxlan_main_t *vxm = &vxlan_main;
703 vlib_main_t *vm = vxm->vlib_main;
Hongjun Nibeb4bf72016-11-25 00:03:46 +0800704 u32 node_index;
Ed Warnickecb9cada2015-12-08 15:45:58 -0700705 u32 tmp;
Hongjun Nibeb4bf72016-11-25 00:03:46 +0800706
Ed Warnickecb9cada2015-12-08 15:45:58 -0700707 if (unformat (input, "l2"))
708 *result = VXLAN_INPUT_NEXT_L2_INPUT;
Hongjun Nibeb4bf72016-11-25 00:03:46 +0800709 else if (unformat (input, "node %U", unformat_vlib_node, vm, &node_index))
Eyal Bari8f576482018-05-07 10:13:44 +0300710 *result = get_decap_next_for_node (node_index, ipv4_set);
Ed Warnickecb9cada2015-12-08 15:45:58 -0700711 else if (unformat (input, "%d", &tmp))
712 *result = tmp;
713 else
714 return 0;
715 return 1;
716}
717
718static clib_error_t *
719vxlan_add_del_tunnel_command_fn (vlib_main_t * vm,
Eyal Bari8f576482018-05-07 10:13:44 +0300720 unformat_input_t * input,
721 vlib_cli_command_t * cmd)
Ed Warnickecb9cada2015-12-08 15:45:58 -0700722{
Eyal Bari8f576482018-05-07 10:13:44 +0300723 unformat_input_t _line_input, *line_input = &_line_input;
724 ip46_address_t src = ip46_address_initializer, dst =
725 ip46_address_initializer;
Ed Warnickecb9cada2015-12-08 15:45:58 -0700726 u8 is_add = 1;
727 u8 src_set = 0;
728 u8 dst_set = 0;
Eyal Baric5b13602016-11-24 19:42:43 +0200729 u8 grp_set = 0;
Chris Luke99cb3352016-04-26 10:49:53 -0400730 u8 ipv4_set = 0;
731 u8 ipv6_set = 0;
Nathan Skrzypczak3e384222021-04-21 19:56:34 +0200732 u8 is_l3 = 0;
Jon Loeliger3d460bd2018-02-01 16:36:12 -0600733 u32 instance = ~0;
Ed Warnickecb9cada2015-12-08 15:45:58 -0700734 u32 encap_fib_index = 0;
Eyal Baric5b13602016-11-24 19:42:43 +0200735 u32 mcast_sw_if_index = ~0;
Hongjun Nibeb4bf72016-11-25 00:03:46 +0800736 u32 decap_next_index = VXLAN_INPUT_NEXT_L2_INPUT;
Ed Warnickecb9cada2015-12-08 15:45:58 -0700737 u32 vni = 0;
Artem Glazychev839dcc02020-12-01 02:39:21 +0700738 u32 src_port = 0;
739 u32 dst_port = 0;
Eyal Bariaa0180b2018-03-27 11:43:57 +0300740 u32 table_id;
Eyal Bari0fa56782018-06-04 12:25:05 +0300741 clib_error_t *parse_error = NULL;
Eyal Baric5b13602016-11-24 19:42:43 +0200742
Ed Warnickecb9cada2015-12-08 15:45:58 -0700743 /* Get a line of input. */
Eyal Bari8f576482018-05-07 10:13:44 +0300744 if (!unformat_user (input, unformat_line_input, line_input))
Ed Warnickecb9cada2015-12-08 15:45:58 -0700745 return 0;
746
Eyal Bari8f576482018-05-07 10:13:44 +0300747 while (unformat_check_input (line_input) != UNFORMAT_END_OF_INPUT)
748 {
749 if (unformat (line_input, "del"))
750 {
751 is_add = 0;
752 }
753 else if (unformat (line_input, "instance %d", &instance))
754 ;
755 else if (unformat (line_input, "src %U",
756 unformat_ip46_address, &src, IP46_TYPE_ANY))
757 {
758 src_set = 1;
759 ip46_address_is_ip4 (&src) ? (ipv4_set = 1) : (ipv6_set = 1);
760 }
761 else if (unformat (line_input, "dst %U",
762 unformat_ip46_address, &dst, IP46_TYPE_ANY))
763 {
764 dst_set = 1;
765 ip46_address_is_ip4 (&dst) ? (ipv4_set = 1) : (ipv6_set = 1);
766 }
767 else if (unformat (line_input, "group %U %U",
768 unformat_ip46_address, &dst, IP46_TYPE_ANY,
769 unformat_vnet_sw_interface,
770 vnet_get_main (), &mcast_sw_if_index))
771 {
772 grp_set = dst_set = 1;
773 ip46_address_is_ip4 (&dst) ? (ipv4_set = 1) : (ipv6_set = 1);
774 }
775 else if (unformat (line_input, "encap-vrf-id %d", &table_id))
776 {
777 encap_fib_index =
778 fib_table_find (fib_ip_proto (ipv6_set), table_id);
Eyal Bari8f576482018-05-07 10:13:44 +0300779 }
Nathan Skrzypczak3e384222021-04-21 19:56:34 +0200780 else if (unformat (line_input, "l3"))
781 is_l3 = 1;
Eyal Bari8f576482018-05-07 10:13:44 +0300782 else if (unformat (line_input, "decap-next %U", unformat_decap_next,
783 &decap_next_index, ipv4_set))
784 ;
785 else if (unformat (line_input, "vni %d", &vni))
786 ;
Artem Glazychev839dcc02020-12-01 02:39:21 +0700787 else if (unformat (line_input, "src_port %d", &src_port))
788 ;
789 else if (unformat (line_input, "dst_port %d", &dst_port))
790 ;
Eyal Bari8f576482018-05-07 10:13:44 +0300791 else
792 {
Eyal Bari0fa56782018-06-04 12:25:05 +0300793 parse_error = clib_error_return (0, "parse error: '%U'",
794 format_unformat_error, line_input);
Eyal Bari8f576482018-05-07 10:13:44 +0300795 break;
796 }
797 }
Ed Warnickecb9cada2015-12-08 15:45:58 -0700798
Eyal Bariaa0180b2018-03-27 11:43:57 +0300799 unformat_free (line_input);
800
Eyal Bari0fa56782018-06-04 12:25:05 +0300801 if (parse_error)
802 return parse_error;
803
Nathan Skrzypczak3e384222021-04-21 19:56:34 +0200804 if (is_l3 && decap_next_index == VXLAN_INPUT_NEXT_L2_INPUT)
805 {
806 vlib_node_t *node = vlib_get_node_by_name (
807 vm, (u8 *) (ipv4_set ? "ip4-input" : "ip6-input"));
808 decap_next_index = get_decap_next_for_node (node->index, ipv4_set);
809 }
810
Eyal Bari0fa56782018-06-04 12:25:05 +0300811 if (encap_fib_index == ~0)
812 return clib_error_return (0, "nonexistent encap-vrf-id %d", table_id);
Eyal Bariaa0180b2018-03-27 11:43:57 +0300813
Ed Warnickecb9cada2015-12-08 15:45:58 -0700814 if (src_set == 0)
Eyal Bariaa0180b2018-03-27 11:43:57 +0300815 return clib_error_return (0, "tunnel src address not specified");
Ed Warnickecb9cada2015-12-08 15:45:58 -0700816
817 if (dst_set == 0)
Eyal Bariaa0180b2018-03-27 11:43:57 +0300818 return clib_error_return (0, "tunnel dst address not specified");
Ed Warnickecb9cada2015-12-08 15:45:58 -0700819
Eyal Bari8f576482018-05-07 10:13:44 +0300820 if (grp_set && !ip46_address_is_multicast (&dst))
Eyal Bariaa0180b2018-03-27 11:43:57 +0300821 return clib_error_return (0, "tunnel group address not multicast");
Eyal Baric5b13602016-11-24 19:42:43 +0200822
Eyal Bari8f576482018-05-07 10:13:44 +0300823 if (grp_set == 0 && ip46_address_is_multicast (&dst))
Eyal Bariaa0180b2018-03-27 11:43:57 +0300824 return clib_error_return (0, "dst address must be unicast");
John Lo56912c82016-12-08 16:10:02 -0500825
Eyal Baric5b13602016-11-24 19:42:43 +0200826 if (grp_set && mcast_sw_if_index == ~0)
Eyal Bariaa0180b2018-03-27 11:43:57 +0300827 return clib_error_return (0, "tunnel nonexistent multicast device");
Eyal Baric5b13602016-11-24 19:42:43 +0200828
Chris Luke99cb3352016-04-26 10:49:53 -0400829 if (ipv4_set && ipv6_set)
Eyal Bariaa0180b2018-03-27 11:43:57 +0300830 return clib_error_return (0, "both IPv4 and IPv6 addresses specified");
Chris Luke99cb3352016-04-26 10:49:53 -0400831
Eyal Bari8f576482018-05-07 10:13:44 +0300832 if (ip46_address_cmp (&src, &dst) == 0)
Eyal Bariaa0180b2018-03-27 11:43:57 +0300833 return clib_error_return (0, "src and dst addresses are identical");
Chris Luke99cb3352016-04-26 10:49:53 -0400834
Hongjun Nibeb4bf72016-11-25 00:03:46 +0800835 if (decap_next_index == ~0)
Eyal Bariaa0180b2018-03-27 11:43:57 +0300836 return clib_error_return (0, "next node not found");
Hongjun Nibeb4bf72016-11-25 00:03:46 +0800837
Ed Warnickecb9cada2015-12-08 15:45:58 -0700838 if (vni == 0)
Eyal Bariaa0180b2018-03-27 11:43:57 +0300839 return clib_error_return (0, "vni not specified");
Ed Warnickecb9cada2015-12-08 15:45:58 -0700840
Eyal Bari8f576482018-05-07 10:13:44 +0300841 if (vni >> 24)
Eyal Bariaa0180b2018-03-27 11:43:57 +0300842 return clib_error_return (0, "vni %d out of range", vni);
Ed Warnickecb9cada2015-12-08 15:45:58 -0700843
Nathan Skrzypczak3e384222021-04-21 19:56:34 +0200844 vnet_vxlan_add_del_tunnel_args_t a = { .is_add = is_add,
845 .is_ip6 = ipv6_set,
846 .is_l3 = is_l3,
847 .instance = instance,
Eyal Bariaa0180b2018-03-27 11:43:57 +0300848#define _(x) .x = x,
Nathan Skrzypczak3e384222021-04-21 19:56:34 +0200849 foreach_copy_field
Ed Warnickecb9cada2015-12-08 15:45:58 -0700850#undef _
Eyal Bariaa0180b2018-03-27 11:43:57 +0300851 };
Jon Loeliger3d460bd2018-02-01 16:36:12 -0600852
Eyal Bariaa0180b2018-03-27 11:43:57 +0300853 u32 tunnel_sw_if_index;
854 int rv = vnet_vxlan_add_del_tunnel (&a, &tunnel_sw_if_index);
Ed Warnickecb9cada2015-12-08 15:45:58 -0700855
Eyal Bari8f576482018-05-07 10:13:44 +0300856 switch (rv)
Ed Warnickecb9cada2015-12-08 15:45:58 -0700857 {
858 case 0:
Pierre Pfister78ea9c22016-05-23 12:51:54 +0100859 if (is_add)
Eyal Bari8f576482018-05-07 10:13:44 +0300860 vlib_cli_output (vm, "%U\n", format_vnet_sw_if_index_name,
861 vnet_get_main (), tunnel_sw_if_index);
Ed Warnickecb9cada2015-12-08 15:45:58 -0700862 break;
Ed Warnickecb9cada2015-12-08 15:45:58 -0700863
864 case VNET_API_ERROR_TUNNEL_EXIST:
Eyal Bariaa0180b2018-03-27 11:43:57 +0300865 return clib_error_return (0, "tunnel already exists...");
Ed Warnickecb9cada2015-12-08 15:45:58 -0700866
867 case VNET_API_ERROR_NO_SUCH_ENTRY:
Eyal Bariaa0180b2018-03-27 11:43:57 +0300868 return clib_error_return (0, "tunnel does not exist...");
Ed Warnickecb9cada2015-12-08 15:45:58 -0700869
Jon Loeliger3d460bd2018-02-01 16:36:12 -0600870 case VNET_API_ERROR_INSTANCE_IN_USE:
Eyal Bariaa0180b2018-03-27 11:43:57 +0300871 return clib_error_return (0, "Instance is in use");
Jon Loeliger3d460bd2018-02-01 16:36:12 -0600872
Ed Warnickecb9cada2015-12-08 15:45:58 -0700873 default:
Eyal Bariaa0180b2018-03-27 11:43:57 +0300874 return clib_error_return
Eyal Bari8f576482018-05-07 10:13:44 +0300875 (0, "vnet_vxlan_add_del_tunnel returned %d", rv);
Ed Warnickecb9cada2015-12-08 15:45:58 -0700876 }
877
Eyal Bari0fa56782018-06-04 12:25:05 +0300878 return 0;
Ed Warnickecb9cada2015-12-08 15:45:58 -0700879}
880
Billy McFallc5d9cda2016-09-14 10:39:58 -0400881/*?
882 * Add or delete a VXLAN Tunnel.
883 *
884 * VXLAN provides the features needed to allow L2 bridge domains (BDs)
885 * to span multiple servers. This is done by building an L2 overlay on
886 * top of an L3 network underlay using VXLAN tunnels.
887 *
888 * This makes it possible for servers to be co-located in the same data
889 * center or be separated geographically as long as they are reachable
890 * through the underlay L3 network.
891 *
892 * You can refer to this kind of L2 overlay bridge domain as a VXLAN
893 * (Virtual eXtensible VLAN) segment.
894 *
895 * @cliexpar
896 * Example of how to create a VXLAN Tunnel:
Artem Glazychev839dcc02020-12-01 02:39:21 +0700897 * @cliexcmd{create vxlan tunnel src 10.0.3.1 dst 10.0.3.3 vni 13 encap-vrf-id
898 7}
Jon Loeliger3d460bd2018-02-01 16:36:12 -0600899 * Example of how to create a VXLAN Tunnel with a known name, vxlan_tunnel42:
900 * @cliexcmd{create vxlan tunnel src 10.0.3.1 dst 10.0.3.3 instance 42}
Artem Glazychev839dcc02020-12-01 02:39:21 +0700901 * Example of how to create a multicast VXLAN Tunnel with a known name,
902 vxlan_tunnel23:
903 * @cliexcmd{create vxlan tunnel src 10.0.3.1 group 239.1.1.1
904 GigabitEthernet0/8/0 instance 23}
905 * Example of how to create a VXLAN Tunnel with custom udp-ports:
906 * @cliexcmd{create vxlan tunnel src 10.0.3.1 dst 10.0.3.3 vni 13 src_port
907 59000 dst_port 59001}
Billy McFallc5d9cda2016-09-14 10:39:58 -0400908 * Example of how to delete a VXLAN Tunnel:
909 * @cliexcmd{create vxlan tunnel src 10.0.3.1 dst 10.0.3.3 vni 13 del}
910 ?*/
911/* *INDENT-OFF* */
Ed Warnickecb9cada2015-12-08 15:45:58 -0700912VLIB_CLI_COMMAND (create_vxlan_tunnel_command, static) = {
913 .path = "create vxlan tunnel",
Eyal Bari8f576482018-05-07 10:13:44 +0300914 .short_help =
Artem Glazychev839dcc02020-12-01 02:39:21 +0700915 "create vxlan tunnel src <local-vtep-addr>"
916 " {dst <remote-vtep-addr>|group <mcast-vtep-addr> <intf-name>} vni <nn>"
917 " [instance <id>]"
Nathan Skrzypczak3e384222021-04-21 19:56:34 +0200918 " [encap-vrf-id <nn>] [decap-next [l2|node <name>]] [del] [l3]"
Artem Glazychev839dcc02020-12-01 02:39:21 +0700919 " [src_port <local-vtep-udp-port>] [dst_port <remote-vtep-udp-port>]",
Ed Warnickecb9cada2015-12-08 15:45:58 -0700920 .function = vxlan_add_del_tunnel_command_fn,
921};
Billy McFallc5d9cda2016-09-14 10:39:58 -0400922/* *INDENT-ON* */
Ed Warnickecb9cada2015-12-08 15:45:58 -0700923
924static clib_error_t *
925show_vxlan_tunnel_command_fn (vlib_main_t * vm,
Eyal Bari8f576482018-05-07 10:13:44 +0300926 unformat_input_t * input,
927 vlib_cli_command_t * cmd)
Ed Warnickecb9cada2015-12-08 15:45:58 -0700928{
Eyal Bari8f576482018-05-07 10:13:44 +0300929 vxlan_main_t *vxm = &vxlan_main;
930 vxlan_tunnel_t *t;
Eyal Bari0fa56782018-06-04 12:25:05 +0300931 int raw = 0;
Eyal Bari0fa56782018-06-04 12:25:05 +0300932
Neale Ranns16be62e2018-07-30 11:59:22 -0700933 while (unformat_check_input (input) != UNFORMAT_END_OF_INPUT)
Eyal Bari0fa56782018-06-04 12:25:05 +0300934 {
Neale Ranns16be62e2018-07-30 11:59:22 -0700935 if (unformat (input, "raw"))
Eyal Bari0fa56782018-06-04 12:25:05 +0300936 raw = 1;
937 else
Neale Ranns16be62e2018-07-30 11:59:22 -0700938 return clib_error_return (0, "parse error: '%U'",
939 format_unformat_error, input);
Eyal Bari0fa56782018-06-04 12:25:05 +0300940 }
941
Ed Warnickecb9cada2015-12-08 15:45:58 -0700942 if (pool_elts (vxm->tunnels) == 0)
943 vlib_cli_output (vm, "No vxlan tunnels configured...");
944
Eyal Bari8f576482018-05-07 10:13:44 +0300945/* *INDENT-OFF* */
Damjan Marionb2c31b62020-12-13 21:47:40 +0100946 pool_foreach (t, vxm->tunnels)
947 {
Ed Warnickecb9cada2015-12-08 15:45:58 -0700948 vlib_cli_output (vm, "%U", format_vxlan_tunnel, t);
Damjan Marionb2c31b62020-12-13 21:47:40 +0100949 }
Eyal Bari8f576482018-05-07 10:13:44 +0300950/* *INDENT-ON* */
951
Eyal Bari0fa56782018-06-04 12:25:05 +0300952 if (raw)
Eyal Baridd47eca2018-07-08 08:15:56 +0300953 {
954 vlib_cli_output (vm, "Raw IPv4 Hash Table:\n%U\n",
955 format_bihash_16_8, &vxm->vxlan4_tunnel_by_key,
956 1 /* verbose */ );
957 vlib_cli_output (vm, "Raw IPv6 Hash Table:\n%U\n",
958 format_bihash_24_8, &vxm->vxlan6_tunnel_by_key,
959 1 /* verbose */ );
960 }
Eyal Bari0fa56782018-06-04 12:25:05 +0300961
Ed Warnickecb9cada2015-12-08 15:45:58 -0700962 return 0;
963}
964
Billy McFallc5d9cda2016-09-14 10:39:58 -0400965/*?
966 * Display all the VXLAN Tunnel entries.
967 *
968 * @cliexpar
969 * Example of how to display the VXLAN Tunnel entries:
970 * @cliexstart{show vxlan tunnel}
Artem Glazychev839dcc02020-12-01 02:39:21 +0700971 * [0] src 10.0.3.1 dst 10.0.3.3 src_port 4789 dst_port 4789 vni 13
972 encap_fib_index 0 sw_if_index 5 decap_next l2
Billy McFallc5d9cda2016-09-14 10:39:58 -0400973 * @cliexend
974 ?*/
975/* *INDENT-OFF* */
Ed Warnickecb9cada2015-12-08 15:45:58 -0700976VLIB_CLI_COMMAND (show_vxlan_tunnel_command, static) = {
977 .path = "show vxlan tunnel",
Eyal Bari0fa56782018-06-04 12:25:05 +0300978 .short_help = "show vxlan tunnel [raw]",
Ed Warnickecb9cada2015-12-08 15:45:58 -0700979 .function = show_vxlan_tunnel_command_fn,
980};
Billy McFallc5d9cda2016-09-14 10:39:58 -0400981/* *INDENT-ON* */
Ed Warnickecb9cada2015-12-08 15:45:58 -0700982
Chris Luke99cb3352016-04-26 10:49:53 -0400983
Eyal Bari8f576482018-05-07 10:13:44 +0300984void
985vnet_int_vxlan_bypass_mode (u32 sw_if_index, u8 is_ip6, u8 is_enable)
John Lo2b81eb82017-01-30 13:12:10 -0500986{
Jon Loeligere3034cd2019-01-03 12:56:02 -0600987 vxlan_main_t *vxm = &vxlan_main;
988
989 if (pool_is_free_index (vxm->vnet_main->interface_main.sw_interfaces,
990 sw_if_index))
991 return;
992
993 is_enable = ! !is_enable;
994
John Lo2b81eb82017-01-30 13:12:10 -0500995 if (is_ip6)
Jon Loeligere3034cd2019-01-03 12:56:02 -0600996 {
997 if (clib_bitmap_get (vxm->bm_ip6_bypass_enabled_by_sw_if, sw_if_index)
998 != is_enable)
999 {
1000 vnet_feature_enable_disable ("ip6-unicast", "ip6-vxlan-bypass",
1001 sw_if_index, is_enable, 0, 0);
1002 vxm->bm_ip6_bypass_enabled_by_sw_if =
1003 clib_bitmap_set (vxm->bm_ip6_bypass_enabled_by_sw_if,
1004 sw_if_index, is_enable);
1005 }
1006 }
John Lo2b81eb82017-01-30 13:12:10 -05001007 else
Jon Loeligere3034cd2019-01-03 12:56:02 -06001008 {
1009 if (clib_bitmap_get (vxm->bm_ip4_bypass_enabled_by_sw_if, sw_if_index)
1010 != is_enable)
1011 {
1012 vnet_feature_enable_disable ("ip4-unicast", "ip4-vxlan-bypass",
1013 sw_if_index, is_enable, 0, 0);
1014 vxm->bm_ip4_bypass_enabled_by_sw_if =
1015 clib_bitmap_set (vxm->bm_ip4_bypass_enabled_by_sw_if,
1016 sw_if_index, is_enable);
1017 }
1018 }
John Lo2b81eb82017-01-30 13:12:10 -05001019}
1020
1021
1022static clib_error_t *
1023set_ip_vxlan_bypass (u32 is_ip6,
Eyal Bari8f576482018-05-07 10:13:44 +03001024 unformat_input_t * input, vlib_cli_command_t * cmd)
John Lo2b81eb82017-01-30 13:12:10 -05001025{
Eyal Bari8f576482018-05-07 10:13:44 +03001026 unformat_input_t _line_input, *line_input = &_line_input;
1027 vnet_main_t *vnm = vnet_get_main ();
1028 clib_error_t *error = 0;
John Lo2b81eb82017-01-30 13:12:10 -05001029 u32 sw_if_index, is_enable;
1030
1031 sw_if_index = ~0;
1032 is_enable = 1;
1033
Eyal Bari8f576482018-05-07 10:13:44 +03001034 if (!unformat_user (input, unformat_line_input, line_input))
John Lo2b81eb82017-01-30 13:12:10 -05001035 return 0;
1036
1037 while (unformat_check_input (line_input) != UNFORMAT_END_OF_INPUT)
1038 {
Eyal Bari8f576482018-05-07 10:13:44 +03001039 if (unformat_user
1040 (line_input, unformat_vnet_sw_interface, vnm, &sw_if_index))
1041 ;
John Lo2b81eb82017-01-30 13:12:10 -05001042 else if (unformat (line_input, "del"))
Eyal Bari8f576482018-05-07 10:13:44 +03001043 is_enable = 0;
John Lo2b81eb82017-01-30 13:12:10 -05001044 else
Eyal Bari8f576482018-05-07 10:13:44 +03001045 {
John Lo2b81eb82017-01-30 13:12:10 -05001046 error = unformat_parse_error (line_input);
1047 goto done;
1048 }
1049 }
1050
1051 if (~0 == sw_if_index)
1052 {
1053 error = clib_error_return (0, "unknown interface `%U'",
1054 format_unformat_error, line_input);
1055 goto done;
1056 }
1057
1058 vnet_int_vxlan_bypass_mode (sw_if_index, is_ip6, is_enable);
1059
Eyal Bari8f576482018-05-07 10:13:44 +03001060done:
Billy McFalla9a20e72017-02-15 11:39:12 -05001061 unformat_free (line_input);
1062
John Lo2b81eb82017-01-30 13:12:10 -05001063 return error;
1064}
1065
1066static clib_error_t *
1067set_ip4_vxlan_bypass (vlib_main_t * vm,
Eyal Bari8f576482018-05-07 10:13:44 +03001068 unformat_input_t * input, vlib_cli_command_t * cmd)
John Lo2b81eb82017-01-30 13:12:10 -05001069{
1070 return set_ip_vxlan_bypass (0, input, cmd);
1071}
1072
1073/*?
Eyal Bari8f576482018-05-07 10:13:44 +03001074 * This command adds the 'ip4-vxlan-bypass' graph node for a given interface.
John Lo2b81eb82017-01-30 13:12:10 -05001075 * By adding the IPv4 vxlan-bypass graph node to an interface, the node checks
Eyal Bari8f576482018-05-07 10:13:44 +03001076 * for and validate input vxlan packet and bypass ip4-lookup, ip4-local,
1077 * ip4-udp-lookup nodes to speedup vxlan packet forwarding. This node will
John Lo2b81eb82017-01-30 13:12:10 -05001078 * cause extra overhead to for non-vxlan packets which is kept at a minimum.
1079 *
1080 * @cliexpar
1081 * @parblock
1082 * Example of graph node before ip4-vxlan-bypass is enabled:
1083 * @cliexstart{show vlib graph ip4-vxlan-bypass}
1084 * Name Next Previous
1085 * ip4-vxlan-bypass error-drop [0]
1086 * vxlan4-input [1]
Eyal Bari8f576482018-05-07 10:13:44 +03001087 * ip4-lookup [2]
John Lo2b81eb82017-01-30 13:12:10 -05001088 * @cliexend
1089 *
1090 * Example of how to enable ip4-vxlan-bypass on an interface:
1091 * @cliexcmd{set interface ip vxlan-bypass GigabitEthernet2/0/0}
1092 *
1093 * Example of graph node after ip4-vxlan-bypass is enabled:
1094 * @cliexstart{show vlib graph ip4-vxlan-bypass}
Eyal Bari8f576482018-05-07 10:13:44 +03001095 * Name Next Previous
1096 * ip4-vxlan-bypass error-drop [0] ip4-input
1097 * vxlan4-input [1] ip4-input-no-checksum
1098 * ip4-lookup [2]
John Lo2b81eb82017-01-30 13:12:10 -05001099 * @cliexend
1100 *
Paul Vinciguerrabdc0e6b2018-09-22 05:32:50 -07001101 * Example of how to display the feature enabled on an interface:
John Lo2b81eb82017-01-30 13:12:10 -05001102 * @cliexstart{show ip interface features GigabitEthernet2/0/0}
1103 * IP feature paths configured on GigabitEthernet2/0/0...
1104 * ...
1105 * ipv4 unicast:
1106 * ip4-vxlan-bypass
1107 * ip4-lookup
1108 * ...
1109 * @cliexend
1110 *
1111 * Example of how to disable ip4-vxlan-bypass on an interface:
1112 * @cliexcmd{set interface ip vxlan-bypass GigabitEthernet2/0/0 del}
1113 * @endparblock
1114?*/
1115/* *INDENT-OFF* */
1116VLIB_CLI_COMMAND (set_interface_ip_vxlan_bypass_command, static) = {
1117 .path = "set interface ip vxlan-bypass",
1118 .function = set_ip4_vxlan_bypass,
1119 .short_help = "set interface ip vxlan-bypass <interface> [del]",
1120};
1121/* *INDENT-ON* */
1122
1123static clib_error_t *
1124set_ip6_vxlan_bypass (vlib_main_t * vm,
Eyal Bari8f576482018-05-07 10:13:44 +03001125 unformat_input_t * input, vlib_cli_command_t * cmd)
John Lo2b81eb82017-01-30 13:12:10 -05001126{
1127 return set_ip_vxlan_bypass (1, input, cmd);
1128}
1129
1130/*?
Eyal Bari8f576482018-05-07 10:13:44 +03001131 * This command adds the 'ip6-vxlan-bypass' graph node for a given interface.
John Lo2b81eb82017-01-30 13:12:10 -05001132 * By adding the IPv6 vxlan-bypass graph node to an interface, the node checks
Eyal Bari8f576482018-05-07 10:13:44 +03001133 * for and validate input vxlan packet and bypass ip6-lookup, ip6-local,
1134 * ip6-udp-lookup nodes to speedup vxlan packet forwarding. This node will
John Lo2b81eb82017-01-30 13:12:10 -05001135 * cause extra overhead to for non-vxlan packets which is kept at a minimum.
1136 *
1137 * @cliexpar
1138 * @parblock
1139 * Example of graph node before ip6-vxlan-bypass is enabled:
1140 * @cliexstart{show vlib graph ip6-vxlan-bypass}
1141 * Name Next Previous
1142 * ip6-vxlan-bypass error-drop [0]
1143 * vxlan6-input [1]
Eyal Bari8f576482018-05-07 10:13:44 +03001144 * ip6-lookup [2]
John Lo2b81eb82017-01-30 13:12:10 -05001145 * @cliexend
1146 *
1147 * Example of how to enable ip6-vxlan-bypass on an interface:
1148 * @cliexcmd{set interface ip6 vxlan-bypass GigabitEthernet2/0/0}
1149 *
1150 * Example of graph node after ip6-vxlan-bypass is enabled:
1151 * @cliexstart{show vlib graph ip6-vxlan-bypass}
Eyal Bari8f576482018-05-07 10:13:44 +03001152 * Name Next Previous
1153 * ip6-vxlan-bypass error-drop [0] ip6-input
1154 * vxlan6-input [1] ip4-input-no-checksum
1155 * ip6-lookup [2]
John Lo2b81eb82017-01-30 13:12:10 -05001156 * @cliexend
1157 *
Paul Vinciguerrabdc0e6b2018-09-22 05:32:50 -07001158 * Example of how to display the feature enabled on an interface:
John Lo2b81eb82017-01-30 13:12:10 -05001159 * @cliexstart{show ip interface features GigabitEthernet2/0/0}
1160 * IP feature paths configured on GigabitEthernet2/0/0...
1161 * ...
1162 * ipv6 unicast:
1163 * ip6-vxlan-bypass
1164 * ip6-lookup
1165 * ...
1166 * @cliexend
1167 *
1168 * Example of how to disable ip6-vxlan-bypass on an interface:
1169 * @cliexcmd{set interface ip6 vxlan-bypass GigabitEthernet2/0/0 del}
1170 * @endparblock
1171?*/
1172/* *INDENT-OFF* */
1173VLIB_CLI_COMMAND (set_interface_ip6_vxlan_bypass_command, static) = {
1174 .path = "set interface ip6 vxlan-bypass",
1175 .function = set_ip6_vxlan_bypass,
Paul Vinciguerra5fb22782019-11-07 17:20:48 -05001176 .short_help = "set interface ip6 vxlan-bypass <interface> [del]",
John Lo2b81eb82017-01-30 13:12:10 -05001177};
1178/* *INDENT-ON* */
1179
eyal bariaf86a482018-04-17 11:20:27 +03001180int
1181vnet_vxlan_add_del_rx_flow (u32 hw_if_index, u32 t_index, int is_add)
1182{
1183 vxlan_main_t *vxm = &vxlan_main;
1184 vxlan_tunnel_t *t = pool_elt_at_index (vxm->tunnels, t_index);
1185 vnet_main_t *vnm = vnet_get_main ();
1186 if (is_add)
1187 {
1188 if (t->flow_index == ~0)
1189 {
1190 vxlan_main_t *vxm = &vxlan_main;
1191 vnet_flow_t flow = {
1192 .actions =
Chenmin Sun1ec9fdb2019-12-05 01:41:35 +08001193 VNET_FLOW_ACTION_REDIRECT_TO_NODE | VNET_FLOW_ACTION_MARK |
1194 VNET_FLOW_ACTION_BUFFER_ADVANCE,
eyal bariaf86a482018-04-17 11:20:27 +03001195 .mark_flow_id = t->dev_instance + vxm->flow_id_start,
1196 .redirect_node_index = vxlan4_flow_input_node.index,
Chenmin Sun1ec9fdb2019-12-05 01:41:35 +08001197 .buffer_advance = sizeof (ethernet_header_t),
eyal bariaf86a482018-04-17 11:20:27 +03001198 .type = VNET_FLOW_TYPE_IP4_VXLAN,
1199 .ip4_vxlan = {
Chenmin Sun34bfa502020-07-27 17:40:17 +08001200 .protocol.prot = IP_PROTOCOL_UDP,
1201 .src_addr.addr = t->dst.ip4,
1202 .dst_addr.addr = t->src.ip4,
1203 .src_addr.mask.as_u32 = ~0,
1204 .dst_addr.mask.as_u32 = ~0,
Artem Glazychev839dcc02020-12-01 02:39:21 +07001205 .dst_port.port = t->src_port,
Chenmin Sun34bfa502020-07-27 17:40:17 +08001206 .dst_port.mask = 0xFF,
eyal bariaf86a482018-04-17 11:20:27 +03001207 .vni = t->vni,
1208 }
1209 ,
1210 };
1211 vnet_flow_add (vnm, &flow, &t->flow_index);
1212 }
1213 return vnet_flow_enable (vnm, t->flow_index, hw_if_index);
1214 }
1215 /* flow index is removed when the tunnel is deleted */
1216 return vnet_flow_disable (vnm, t->flow_index, hw_if_index);
1217}
1218
1219u32
1220vnet_vxlan_get_tunnel_index (u32 sw_if_index)
1221{
1222 vxlan_main_t *vxm = &vxlan_main;
1223
Eyal Baricd307742018-07-22 12:45:15 +03001224 if (sw_if_index >= vec_len (vxm->tunnel_index_by_sw_if_index))
eyal bariaf86a482018-04-17 11:20:27 +03001225 return ~0;
1226 return vxm->tunnel_index_by_sw_if_index[sw_if_index];
1227}
1228
1229static clib_error_t *
1230vxlan_offload_command_fn (vlib_main_t * vm,
1231 unformat_input_t * input, vlib_cli_command_t * cmd)
1232{
1233 unformat_input_t _line_input, *line_input = &_line_input;
1234
1235 /* Get a line of input. */
1236 if (!unformat_user (input, unformat_line_input, line_input))
1237 return 0;
1238
1239 vnet_main_t *vnm = vnet_get_main ();
1240 u32 rx_sw_if_index = ~0;
1241 u32 hw_if_index = ~0;
1242 int is_add = 1;
1243
1244 while (unformat_check_input (line_input) != UNFORMAT_END_OF_INPUT)
1245 {
1246 if (unformat (line_input, "hw %U", unformat_vnet_hw_interface, vnm,
1247 &hw_if_index))
1248 continue;
1249 if (unformat (line_input, "rx %U", unformat_vnet_sw_interface, vnm,
1250 &rx_sw_if_index))
1251 continue;
1252 if (unformat (line_input, "del"))
1253 {
1254 is_add = 0;
1255 continue;
1256 }
1257 return clib_error_return (0, "unknown input `%U'",
1258 format_unformat_error, line_input);
1259 }
1260
1261 if (rx_sw_if_index == ~0)
1262 return clib_error_return (0, "missing rx interface");
1263 if (hw_if_index == ~0)
1264 return clib_error_return (0, "missing hw interface");
1265
1266 u32 t_index = vnet_vxlan_get_tunnel_index (rx_sw_if_index);;
1267 if (t_index == ~0)
1268 return clib_error_return (0, "%U is not a vxlan tunnel",
1269 format_vnet_sw_if_index_name, vnm,
1270 rx_sw_if_index);
1271
1272 vxlan_main_t *vxm = &vxlan_main;
1273 vxlan_tunnel_t *t = pool_elt_at_index (vxm->tunnels, t_index);
1274
1275 if (!ip46_address_is_ip4 (&t->dst))
1276 return clib_error_return (0, "currently only IPV4 tunnels are supported");
1277
1278 vnet_hw_interface_t *hw_if = vnet_get_hw_interface (vnm, hw_if_index);
1279 ip4_main_t *im = &ip4_main;
1280 u32 rx_fib_index =
1281 vec_elt (im->fib_index_by_sw_if_index, hw_if->sw_if_index);
1282
1283 if (t->encap_fib_index != rx_fib_index)
1284 return clib_error_return (0, "interface/tunnel fib mismatch");
1285
1286 if (vnet_vxlan_add_del_rx_flow (hw_if_index, t_index, is_add))
1287 return clib_error_return (0, "error %s flow",
1288 is_add ? "enabling" : "disabling");
1289
1290 return 0;
1291}
1292
1293/* *INDENT-OFF* */
1294VLIB_CLI_COMMAND (vxlan_offload_command, static) = {
1295 .path = "set flow-offload vxlan",
1296 .short_help =
Paul Vinciguerrabdc0e6b2018-09-22 05:32:50 -07001297 "set flow-offload vxlan hw <interface-name> rx <tunnel-name> [del]",
eyal bariaf86a482018-04-17 11:20:27 +03001298 .function = vxlan_offload_command_fn,
1299};
1300/* *INDENT-ON* */
1301
Eyal Bari0fa56782018-06-04 12:25:05 +03001302#define VXLAN_HASH_NUM_BUCKETS (2 * 1024)
1303#define VXLAN_HASH_MEMORY_SIZE (1 << 20)
1304
Eyal Bari8f576482018-05-07 10:13:44 +03001305clib_error_t *
1306vxlan_init (vlib_main_t * vm)
Ed Warnickecb9cada2015-12-08 15:45:58 -07001307{
Eyal Bari8f576482018-05-07 10:13:44 +03001308 vxlan_main_t *vxm = &vxlan_main;
1309
1310 vxm->vnet_main = vnet_get_main ();
Ed Warnickecb9cada2015-12-08 15:45:58 -07001311 vxm->vlib_main = vm;
1312
eyal bariaf86a482018-04-17 11:20:27 +03001313 vnet_flow_get_range (vxm->vnet_main, "vxlan", 1024 * 1024,
1314 &vxm->flow_id_start);
1315
Jon Loeligere3034cd2019-01-03 12:56:02 -06001316 vxm->bm_ip4_bypass_enabled_by_sw_if = 0;
1317 vxm->bm_ip6_bypass_enabled_by_sw_if = 0;
1318
Chris Luke99cb3352016-04-26 10:49:53 -04001319 /* initialize the ip6 hash */
Eyal Baridd47eca2018-07-08 08:15:56 +03001320 clib_bihash_init_16_8 (&vxm->vxlan4_tunnel_by_key, "vxlan4",
1321 VXLAN_HASH_NUM_BUCKETS, VXLAN_HASH_MEMORY_SIZE);
1322 clib_bihash_init_24_8 (&vxm->vxlan6_tunnel_by_key, "vxlan6",
Eyal Bari0fa56782018-06-04 12:25:05 +03001323 VXLAN_HASH_NUM_BUCKETS, VXLAN_HASH_MEMORY_SIZE);
Nick Zavaritsky27518c22020-02-27 15:54:58 +00001324 vxm->vtep_table = vtep_table_create ();
Eyal Bari8f576482018-05-07 10:13:44 +03001325 vxm->mcast_shared = hash_create_mem (0,
1326 sizeof (ip46_address_t),
1327 sizeof (mcast_shared_t));
Chris Luke99cb3352016-04-26 10:49:53 -04001328
Eyal Bari8f576482018-05-07 10:13:44 +03001329 fib_node_register_type (FIB_NODE_TYPE_VXLAN_TUNNEL, &vxlan_vft);
John Loc42912d2016-11-07 18:30:47 -05001330
Ed Warnickecb9cada2015-12-08 15:45:58 -07001331 return 0;
1332}
1333
Eyal Bari8f576482018-05-07 10:13:44 +03001334VLIB_INIT_FUNCTION (vxlan_init);
Jon Loeliger3d460bd2018-02-01 16:36:12 -06001335
1336/*
Eyal Bari8f576482018-05-07 10:13:44 +03001337 * fd.io coding-style-patch-verification: ON
1338 *
Jon Loeliger3d460bd2018-02-01 16:36:12 -06001339 * Local Variables:
1340 * eval: (c-set-style "gnu")
1341 * End:
1342 */