blob: 8568d5b0a983e372252bc3691e99e1e0b617dfb9 [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 */
Paul Atkins8e2b1b12021-10-12 14:32:11 +0100584 mfib_table_entry_path_update (t->encap_fib_index, &mpfx,
585 MFIB_SOURCE_VXLAN,
586 MFIB_ENTRY_FLAG_NONE, &path);
Neale Ranns32e1c012016-11-22 17:07:28 +0000587
Eyal Bari8f576482018-05-07 10:13:44 +0300588 path.frp_sw_if_index = a->mcast_sw_if_index;
589 path.frp_flags = FIB_ROUTE_PATH_FLAG_NONE;
Neale Ranns097fa662018-05-01 05:17:55 -0700590 path.frp_mitf_flags = MFIB_ITF_FLAG_ACCEPT;
Paul Atkins8e2b1b12021-10-12 14:32:11 +0100591 mfei = mfib_table_entry_path_update (
592 t->encap_fib_index, &mpfx, MFIB_SOURCE_VXLAN,
593 MFIB_ENTRY_FLAG_NONE, &path);
Neale Ranns32e1c012016-11-22 17:07:28 +0000594
Eyal Bari8f576482018-05-07 10:13:44 +0300595 /*
596 * Create the mcast adjacency to send traffic to the group
597 */
598 ai = adj_mcast_add_or_lock (fp,
599 fib_proto_to_link (fp),
600 a->mcast_sw_if_index);
Neale Ranns32e1c012016-11-22 17:07:28 +0000601
Eyal Bari8f576482018-05-07 10:13:44 +0300602 /*
603 * create a new end-point
604 */
605 mcast_shared_add (&t->dst, mfei, ai);
606 }
Neale Ranns32e1c012016-11-22 17:07:28 +0000607
Eyal Bari8f576482018-05-07 10:13:44 +0300608 dpo_id_t dpo = DPO_INVALID;
609 mcast_shared_t ep = mcast_shared_get (&t->dst);
Neale Ranns32e1c012016-11-22 17:07:28 +0000610
Eyal Bari8f576482018-05-07 10:13:44 +0300611 /* Stack shared mcast dst mac addr rewrite on encap */
612 dpo_set (&dpo, DPO_ADJACENCY_MCAST,
613 fib_proto_to_dpo (fp), ep.mcast_adj_index);
Neale Ranns32e1c012016-11-22 17:07:28 +0000614
Eyal Bari8f576482018-05-07 10:13:44 +0300615 dpo_stack_from_node (encap_index, &t->next_dpo, &dpo);
616 dpo_reset (&dpo);
Eyal Barifff67c82016-12-21 12:45:47 +0200617 flood_class = VNET_FLOOD_CLASS_TUNNEL_MASTER;
John Lo56912c82016-12-08 16:10:02 -0500618 }
619
Eyal Bari8f576482018-05-07 10:13:44 +0300620 vnet_get_sw_interface (vnet_get_main (), sw_if_index)->flood_class =
621 flood_class;
Ed Warnickecb9cada2015-12-08 15:45:58 -0700622 }
623 else
624 {
625 /* deleting a tunnel: tunnel must exist */
John Lo37682e12016-11-30 12:51:39 -0500626 if (!p)
Eyal Bari8f576482018-05-07 10:13:44 +0300627 return VNET_API_ERROR_NO_SUCH_ENTRY;
Ed Warnickecb9cada2015-12-08 15:45:58 -0700628
Eyal Barif8d5e212018-10-14 10:54:32 +0300629 u32 instance = is_ip6 ? key6.value :
630 vxm->tunnel_index_by_sw_if_index[p->sw_if_index];
Eyal Bariefd9cf32018-10-02 12:23:06 +0300631 vxlan_tunnel_t *t = pool_elt_at_index (vxm->tunnels, instance);
Ed Warnickecb9cada2015-12-08 15:45:58 -0700632
Eyal Baribc799c92017-04-06 03:26:59 +0300633 sw_if_index = t->sw_if_index;
Eyal Bari8f576482018-05-07 10:13:44 +0300634 vnet_sw_interface_set_flags (vnm, sw_if_index, 0 /* down */ );
Ed Warnickecb9cada2015-12-08 15:45:58 -0700635
Jon Loeliger3d460bd2018-02-01 16:36:12 -0600636 vxm->tunnel_index_by_sw_if_index[sw_if_index] = ~0;
Dave Wallace60231f32015-12-17 21:04:30 -0500637
John Loc42912d2016-11-07 18:30:47 -0500638 if (!is_ip6)
Eyal Baridd47eca2018-07-08 08:15:56 +0300639 clib_bihash_add_del_16_8 (&vxm->vxlan4_tunnel_by_key, &key4,
640 0 /*del */ );
Chris Luke99cb3352016-04-26 10:49:53 -0400641 else
Eyal Baridd47eca2018-07-08 08:15:56 +0300642 clib_bihash_add_del_24_8 (&vxm->vxlan6_tunnel_by_key, &key6,
Eyal Bari0fa56782018-06-04 12:25:05 +0300643 0 /*del */ );
Eyal Barifff67c82016-12-21 12:45:47 +0200644
Eyal Bari8f576482018-05-07 10:13:44 +0300645 if (!ip46_address_is_multicast (&t->dst))
646 {
eyal bariaf86a482018-04-17 11:20:27 +0300647 if (t->flow_index != ~0)
648 vnet_flow_del (vnm, t->flow_index);
649
Nick Zavaritsky27518c22020-02-27 15:54:58 +0000650 vtep_addr_unref (&vxm->vtep_table, t->encap_fib_index, &t->src);
Neale Ranns1f50bf82019-07-16 15:28:52 +0000651 fib_entry_untrack (t->fib_entry_index, t->sibling_index);
Eyal Bari8f576482018-05-07 10:13:44 +0300652 }
Nick Zavaritsky27518c22020-02-27 15:54:58 +0000653 else if (vtep_addr_unref (&vxm->vtep_table,
654 t->encap_fib_index, &t->dst) == 0)
Eyal Bari8f576482018-05-07 10:13:44 +0300655 {
656 mcast_shared_remove (&t->dst);
657 }
Eyal Barifff67c82016-12-21 12:45:47 +0200658
Nathan Skrzypczak3e384222021-04-21 19:56:34 +0200659 vnet_hw_interface_t *hw = vnet_get_hw_interface (vnm, t->hw_if_index);
660 if (hw->dev_class_index == vxlan_device_class.index)
661 vnet_delete_hw_interface (vnm, t->hw_if_index);
662 else
663 ethernet_delete_interface (vnm, t->hw_if_index);
Jon Loeliger25ef2b52018-02-23 17:02:41 -0600664 hash_unset (vxm->instance_used, t->user_instance);
Jon Loeliger3d460bd2018-02-01 16:36:12 -0600665
Eyal Bari8f576482018-05-07 10:13:44 +0300666 fib_node_deinit (&t->node);
Ed Warnickecb9cada2015-12-08 15:45:58 -0700667 pool_put (vxm->tunnels, t);
668 }
669
670 if (sw_if_indexp)
Eyal Bari8f576482018-05-07 10:13:44 +0300671 *sw_if_indexp = sw_if_index;
Ed Warnickecb9cada2015-12-08 15:45:58 -0700672
Jakub Grajciardf3ca232019-06-04 13:16:42 +0200673 if (a->is_add)
674 {
675 /* register udp ports */
Artem Glazychev839dcc02020-12-01 02:39:21 +0700676 if (!is_ip6 && !udp_is_valid_dst_port (a->src_port, 1))
677 udp_register_dst_port (vxm->vlib_main, a->src_port,
Jakub Grajciardf3ca232019-06-04 13:16:42 +0200678 vxlan4_input_node.index, 1);
Artem Glazychev839dcc02020-12-01 02:39:21 +0700679 if (is_ip6 && !udp_is_valid_dst_port (a->src_port, 0))
680 udp_register_dst_port (vxm->vlib_main, a->src_port,
Jakub Grajciardf3ca232019-06-04 13:16:42 +0200681 vxlan6_input_node.index, 0);
682 }
683
Ed Warnickecb9cada2015-12-08 15:45:58 -0700684 return 0;
685}
686
Eyal Bari8f576482018-05-07 10:13:44 +0300687static uword
688get_decap_next_for_node (u32 node_index, u32 ipv4_set)
Hongjun Nibeb4bf72016-11-25 00:03:46 +0800689{
Eyal Bari8f576482018-05-07 10:13:44 +0300690 vxlan_main_t *vxm = &vxlan_main;
691 vlib_main_t *vm = vxm->vlib_main;
Eyal Bari0765c6e2017-01-29 12:40:50 +0200692 uword input_node = (ipv4_set) ? vxlan4_input_node.index :
693 vxlan6_input_node.index;
Hongjun Nibeb4bf72016-11-25 00:03:46 +0800694
Eyal Bari0765c6e2017-01-29 12:40:50 +0200695 return vlib_node_add_next (vm, input_node, node_index);
Hongjun Nibeb4bf72016-11-25 00:03:46 +0800696}
697
Eyal Bari8f576482018-05-07 10:13:44 +0300698static uword
699unformat_decap_next (unformat_input_t * input, va_list * args)
Ed Warnickecb9cada2015-12-08 15:45:58 -0700700{
Eyal Bari8f576482018-05-07 10:13:44 +0300701 u32 *result = va_arg (*args, u32 *);
Hongjun Nibeb4bf72016-11-25 00:03:46 +0800702 u32 ipv4_set = va_arg (*args, int);
Eyal Bari8f576482018-05-07 10:13:44 +0300703 vxlan_main_t *vxm = &vxlan_main;
704 vlib_main_t *vm = vxm->vlib_main;
Hongjun Nibeb4bf72016-11-25 00:03:46 +0800705 u32 node_index;
Ed Warnickecb9cada2015-12-08 15:45:58 -0700706 u32 tmp;
Hongjun Nibeb4bf72016-11-25 00:03:46 +0800707
Ed Warnickecb9cada2015-12-08 15:45:58 -0700708 if (unformat (input, "l2"))
709 *result = VXLAN_INPUT_NEXT_L2_INPUT;
Hongjun Nibeb4bf72016-11-25 00:03:46 +0800710 else if (unformat (input, "node %U", unformat_vlib_node, vm, &node_index))
Eyal Bari8f576482018-05-07 10:13:44 +0300711 *result = get_decap_next_for_node (node_index, ipv4_set);
Ed Warnickecb9cada2015-12-08 15:45:58 -0700712 else if (unformat (input, "%d", &tmp))
713 *result = tmp;
714 else
715 return 0;
716 return 1;
717}
718
719static clib_error_t *
720vxlan_add_del_tunnel_command_fn (vlib_main_t * vm,
Eyal Bari8f576482018-05-07 10:13:44 +0300721 unformat_input_t * input,
722 vlib_cli_command_t * cmd)
Ed Warnickecb9cada2015-12-08 15:45:58 -0700723{
Eyal Bari8f576482018-05-07 10:13:44 +0300724 unformat_input_t _line_input, *line_input = &_line_input;
725 ip46_address_t src = ip46_address_initializer, dst =
726 ip46_address_initializer;
Ed Warnickecb9cada2015-12-08 15:45:58 -0700727 u8 is_add = 1;
728 u8 src_set = 0;
729 u8 dst_set = 0;
Eyal Baric5b13602016-11-24 19:42:43 +0200730 u8 grp_set = 0;
Chris Luke99cb3352016-04-26 10:49:53 -0400731 u8 ipv4_set = 0;
732 u8 ipv6_set = 0;
Nathan Skrzypczak3e384222021-04-21 19:56:34 +0200733 u8 is_l3 = 0;
Jon Loeliger3d460bd2018-02-01 16:36:12 -0600734 u32 instance = ~0;
Ed Warnickecb9cada2015-12-08 15:45:58 -0700735 u32 encap_fib_index = 0;
Eyal Baric5b13602016-11-24 19:42:43 +0200736 u32 mcast_sw_if_index = ~0;
Hongjun Nibeb4bf72016-11-25 00:03:46 +0800737 u32 decap_next_index = VXLAN_INPUT_NEXT_L2_INPUT;
Ed Warnickecb9cada2015-12-08 15:45:58 -0700738 u32 vni = 0;
Artem Glazychev839dcc02020-12-01 02:39:21 +0700739 u32 src_port = 0;
740 u32 dst_port = 0;
Eyal Bariaa0180b2018-03-27 11:43:57 +0300741 u32 table_id;
Eyal Bari0fa56782018-06-04 12:25:05 +0300742 clib_error_t *parse_error = NULL;
Eyal Baric5b13602016-11-24 19:42:43 +0200743
Ed Warnickecb9cada2015-12-08 15:45:58 -0700744 /* Get a line of input. */
Eyal Bari8f576482018-05-07 10:13:44 +0300745 if (!unformat_user (input, unformat_line_input, line_input))
Ed Warnickecb9cada2015-12-08 15:45:58 -0700746 return 0;
747
Eyal Bari8f576482018-05-07 10:13:44 +0300748 while (unformat_check_input (line_input) != UNFORMAT_END_OF_INPUT)
749 {
750 if (unformat (line_input, "del"))
751 {
752 is_add = 0;
753 }
754 else if (unformat (line_input, "instance %d", &instance))
755 ;
756 else if (unformat (line_input, "src %U",
757 unformat_ip46_address, &src, IP46_TYPE_ANY))
758 {
759 src_set = 1;
760 ip46_address_is_ip4 (&src) ? (ipv4_set = 1) : (ipv6_set = 1);
761 }
762 else if (unformat (line_input, "dst %U",
763 unformat_ip46_address, &dst, IP46_TYPE_ANY))
764 {
765 dst_set = 1;
766 ip46_address_is_ip4 (&dst) ? (ipv4_set = 1) : (ipv6_set = 1);
767 }
768 else if (unformat (line_input, "group %U %U",
769 unformat_ip46_address, &dst, IP46_TYPE_ANY,
770 unformat_vnet_sw_interface,
771 vnet_get_main (), &mcast_sw_if_index))
772 {
773 grp_set = dst_set = 1;
774 ip46_address_is_ip4 (&dst) ? (ipv4_set = 1) : (ipv6_set = 1);
775 }
776 else if (unformat (line_input, "encap-vrf-id %d", &table_id))
777 {
778 encap_fib_index =
779 fib_table_find (fib_ip_proto (ipv6_set), table_id);
Eyal Bari8f576482018-05-07 10:13:44 +0300780 }
Nathan Skrzypczak3e384222021-04-21 19:56:34 +0200781 else if (unformat (line_input, "l3"))
782 is_l3 = 1;
Eyal Bari8f576482018-05-07 10:13:44 +0300783 else if (unformat (line_input, "decap-next %U", unformat_decap_next,
784 &decap_next_index, ipv4_set))
785 ;
786 else if (unformat (line_input, "vni %d", &vni))
787 ;
Artem Glazychev839dcc02020-12-01 02:39:21 +0700788 else if (unformat (line_input, "src_port %d", &src_port))
789 ;
790 else if (unformat (line_input, "dst_port %d", &dst_port))
791 ;
Eyal Bari8f576482018-05-07 10:13:44 +0300792 else
793 {
Eyal Bari0fa56782018-06-04 12:25:05 +0300794 parse_error = clib_error_return (0, "parse error: '%U'",
795 format_unformat_error, line_input);
Eyal Bari8f576482018-05-07 10:13:44 +0300796 break;
797 }
798 }
Ed Warnickecb9cada2015-12-08 15:45:58 -0700799
Eyal Bariaa0180b2018-03-27 11:43:57 +0300800 unformat_free (line_input);
801
Eyal Bari0fa56782018-06-04 12:25:05 +0300802 if (parse_error)
803 return parse_error;
804
Nathan Skrzypczak3e384222021-04-21 19:56:34 +0200805 if (is_l3 && decap_next_index == VXLAN_INPUT_NEXT_L2_INPUT)
806 {
807 vlib_node_t *node = vlib_get_node_by_name (
808 vm, (u8 *) (ipv4_set ? "ip4-input" : "ip6-input"));
809 decap_next_index = get_decap_next_for_node (node->index, ipv4_set);
810 }
811
Eyal Bari0fa56782018-06-04 12:25:05 +0300812 if (encap_fib_index == ~0)
813 return clib_error_return (0, "nonexistent encap-vrf-id %d", table_id);
Eyal Bariaa0180b2018-03-27 11:43:57 +0300814
Ed Warnickecb9cada2015-12-08 15:45:58 -0700815 if (src_set == 0)
Eyal Bariaa0180b2018-03-27 11:43:57 +0300816 return clib_error_return (0, "tunnel src address not specified");
Ed Warnickecb9cada2015-12-08 15:45:58 -0700817
818 if (dst_set == 0)
Eyal Bariaa0180b2018-03-27 11:43:57 +0300819 return clib_error_return (0, "tunnel dst address not specified");
Ed Warnickecb9cada2015-12-08 15:45:58 -0700820
Eyal Bari8f576482018-05-07 10:13:44 +0300821 if (grp_set && !ip46_address_is_multicast (&dst))
Eyal Bariaa0180b2018-03-27 11:43:57 +0300822 return clib_error_return (0, "tunnel group address not multicast");
Eyal Baric5b13602016-11-24 19:42:43 +0200823
Eyal Bari8f576482018-05-07 10:13:44 +0300824 if (grp_set == 0 && ip46_address_is_multicast (&dst))
Eyal Bariaa0180b2018-03-27 11:43:57 +0300825 return clib_error_return (0, "dst address must be unicast");
John Lo56912c82016-12-08 16:10:02 -0500826
Eyal Baric5b13602016-11-24 19:42:43 +0200827 if (grp_set && mcast_sw_if_index == ~0)
Eyal Bariaa0180b2018-03-27 11:43:57 +0300828 return clib_error_return (0, "tunnel nonexistent multicast device");
Eyal Baric5b13602016-11-24 19:42:43 +0200829
Chris Luke99cb3352016-04-26 10:49:53 -0400830 if (ipv4_set && ipv6_set)
Eyal Bariaa0180b2018-03-27 11:43:57 +0300831 return clib_error_return (0, "both IPv4 and IPv6 addresses specified");
Chris Luke99cb3352016-04-26 10:49:53 -0400832
Eyal Bari8f576482018-05-07 10:13:44 +0300833 if (ip46_address_cmp (&src, &dst) == 0)
Eyal Bariaa0180b2018-03-27 11:43:57 +0300834 return clib_error_return (0, "src and dst addresses are identical");
Chris Luke99cb3352016-04-26 10:49:53 -0400835
Hongjun Nibeb4bf72016-11-25 00:03:46 +0800836 if (decap_next_index == ~0)
Eyal Bariaa0180b2018-03-27 11:43:57 +0300837 return clib_error_return (0, "next node not found");
Hongjun Nibeb4bf72016-11-25 00:03:46 +0800838
Ed Warnickecb9cada2015-12-08 15:45:58 -0700839 if (vni == 0)
Eyal Bariaa0180b2018-03-27 11:43:57 +0300840 return clib_error_return (0, "vni not specified");
Ed Warnickecb9cada2015-12-08 15:45:58 -0700841
Eyal Bari8f576482018-05-07 10:13:44 +0300842 if (vni >> 24)
Eyal Bariaa0180b2018-03-27 11:43:57 +0300843 return clib_error_return (0, "vni %d out of range", vni);
Ed Warnickecb9cada2015-12-08 15:45:58 -0700844
Nathan Skrzypczak3e384222021-04-21 19:56:34 +0200845 vnet_vxlan_add_del_tunnel_args_t a = { .is_add = is_add,
846 .is_ip6 = ipv6_set,
847 .is_l3 = is_l3,
848 .instance = instance,
Eyal Bariaa0180b2018-03-27 11:43:57 +0300849#define _(x) .x = x,
Nathan Skrzypczak3e384222021-04-21 19:56:34 +0200850 foreach_copy_field
Ed Warnickecb9cada2015-12-08 15:45:58 -0700851#undef _
Eyal Bariaa0180b2018-03-27 11:43:57 +0300852 };
Jon Loeliger3d460bd2018-02-01 16:36:12 -0600853
Eyal Bariaa0180b2018-03-27 11:43:57 +0300854 u32 tunnel_sw_if_index;
855 int rv = vnet_vxlan_add_del_tunnel (&a, &tunnel_sw_if_index);
Ed Warnickecb9cada2015-12-08 15:45:58 -0700856
Eyal Bari8f576482018-05-07 10:13:44 +0300857 switch (rv)
Ed Warnickecb9cada2015-12-08 15:45:58 -0700858 {
859 case 0:
Pierre Pfister78ea9c22016-05-23 12:51:54 +0100860 if (is_add)
Eyal Bari8f576482018-05-07 10:13:44 +0300861 vlib_cli_output (vm, "%U\n", format_vnet_sw_if_index_name,
862 vnet_get_main (), tunnel_sw_if_index);
Ed Warnickecb9cada2015-12-08 15:45:58 -0700863 break;
Ed Warnickecb9cada2015-12-08 15:45:58 -0700864
865 case VNET_API_ERROR_TUNNEL_EXIST:
Eyal Bariaa0180b2018-03-27 11:43:57 +0300866 return clib_error_return (0, "tunnel already exists...");
Ed Warnickecb9cada2015-12-08 15:45:58 -0700867
868 case VNET_API_ERROR_NO_SUCH_ENTRY:
Eyal Bariaa0180b2018-03-27 11:43:57 +0300869 return clib_error_return (0, "tunnel does not exist...");
Ed Warnickecb9cada2015-12-08 15:45:58 -0700870
Jon Loeliger3d460bd2018-02-01 16:36:12 -0600871 case VNET_API_ERROR_INSTANCE_IN_USE:
Eyal Bariaa0180b2018-03-27 11:43:57 +0300872 return clib_error_return (0, "Instance is in use");
Jon Loeliger3d460bd2018-02-01 16:36:12 -0600873
Ed Warnickecb9cada2015-12-08 15:45:58 -0700874 default:
Eyal Bariaa0180b2018-03-27 11:43:57 +0300875 return clib_error_return
Eyal Bari8f576482018-05-07 10:13:44 +0300876 (0, "vnet_vxlan_add_del_tunnel returned %d", rv);
Ed Warnickecb9cada2015-12-08 15:45:58 -0700877 }
878
Eyal Bari0fa56782018-06-04 12:25:05 +0300879 return 0;
Ed Warnickecb9cada2015-12-08 15:45:58 -0700880}
881
Billy McFallc5d9cda2016-09-14 10:39:58 -0400882/*?
883 * Add or delete a VXLAN Tunnel.
884 *
885 * VXLAN provides the features needed to allow L2 bridge domains (BDs)
886 * to span multiple servers. This is done by building an L2 overlay on
887 * top of an L3 network underlay using VXLAN tunnels.
888 *
889 * This makes it possible for servers to be co-located in the same data
890 * center or be separated geographically as long as they are reachable
891 * through the underlay L3 network.
892 *
893 * You can refer to this kind of L2 overlay bridge domain as a VXLAN
894 * (Virtual eXtensible VLAN) segment.
895 *
896 * @cliexpar
897 * Example of how to create a VXLAN Tunnel:
Artem Glazychev839dcc02020-12-01 02:39:21 +0700898 * @cliexcmd{create vxlan tunnel src 10.0.3.1 dst 10.0.3.3 vni 13 encap-vrf-id
899 7}
Jon Loeliger3d460bd2018-02-01 16:36:12 -0600900 * Example of how to create a VXLAN Tunnel with a known name, vxlan_tunnel42:
901 * @cliexcmd{create vxlan tunnel src 10.0.3.1 dst 10.0.3.3 instance 42}
Artem Glazychev839dcc02020-12-01 02:39:21 +0700902 * Example of how to create a multicast VXLAN Tunnel with a known name,
903 vxlan_tunnel23:
904 * @cliexcmd{create vxlan tunnel src 10.0.3.1 group 239.1.1.1
905 GigabitEthernet0/8/0 instance 23}
906 * Example of how to create a VXLAN Tunnel with custom udp-ports:
907 * @cliexcmd{create vxlan tunnel src 10.0.3.1 dst 10.0.3.3 vni 13 src_port
908 59000 dst_port 59001}
Billy McFallc5d9cda2016-09-14 10:39:58 -0400909 * Example of how to delete a VXLAN Tunnel:
910 * @cliexcmd{create vxlan tunnel src 10.0.3.1 dst 10.0.3.3 vni 13 del}
911 ?*/
912/* *INDENT-OFF* */
Ed Warnickecb9cada2015-12-08 15:45:58 -0700913VLIB_CLI_COMMAND (create_vxlan_tunnel_command, static) = {
914 .path = "create vxlan tunnel",
Eyal Bari8f576482018-05-07 10:13:44 +0300915 .short_help =
Artem Glazychev839dcc02020-12-01 02:39:21 +0700916 "create vxlan tunnel src <local-vtep-addr>"
917 " {dst <remote-vtep-addr>|group <mcast-vtep-addr> <intf-name>} vni <nn>"
918 " [instance <id>]"
Nathan Skrzypczak3e384222021-04-21 19:56:34 +0200919 " [encap-vrf-id <nn>] [decap-next [l2|node <name>]] [del] [l3]"
Artem Glazychev839dcc02020-12-01 02:39:21 +0700920 " [src_port <local-vtep-udp-port>] [dst_port <remote-vtep-udp-port>]",
Ed Warnickecb9cada2015-12-08 15:45:58 -0700921 .function = vxlan_add_del_tunnel_command_fn,
922};
Billy McFallc5d9cda2016-09-14 10:39:58 -0400923/* *INDENT-ON* */
Ed Warnickecb9cada2015-12-08 15:45:58 -0700924
925static clib_error_t *
926show_vxlan_tunnel_command_fn (vlib_main_t * vm,
Eyal Bari8f576482018-05-07 10:13:44 +0300927 unformat_input_t * input,
928 vlib_cli_command_t * cmd)
Ed Warnickecb9cada2015-12-08 15:45:58 -0700929{
Eyal Bari8f576482018-05-07 10:13:44 +0300930 vxlan_main_t *vxm = &vxlan_main;
931 vxlan_tunnel_t *t;
Eyal Bari0fa56782018-06-04 12:25:05 +0300932 int raw = 0;
Eyal Bari0fa56782018-06-04 12:25:05 +0300933
Neale Ranns16be62e2018-07-30 11:59:22 -0700934 while (unformat_check_input (input) != UNFORMAT_END_OF_INPUT)
Eyal Bari0fa56782018-06-04 12:25:05 +0300935 {
Neale Ranns16be62e2018-07-30 11:59:22 -0700936 if (unformat (input, "raw"))
Eyal Bari0fa56782018-06-04 12:25:05 +0300937 raw = 1;
938 else
Neale Ranns16be62e2018-07-30 11:59:22 -0700939 return clib_error_return (0, "parse error: '%U'",
940 format_unformat_error, input);
Eyal Bari0fa56782018-06-04 12:25:05 +0300941 }
942
Ed Warnickecb9cada2015-12-08 15:45:58 -0700943 if (pool_elts (vxm->tunnels) == 0)
944 vlib_cli_output (vm, "No vxlan tunnels configured...");
945
Eyal Bari8f576482018-05-07 10:13:44 +0300946/* *INDENT-OFF* */
Damjan Marionb2c31b62020-12-13 21:47:40 +0100947 pool_foreach (t, vxm->tunnels)
948 {
Ed Warnickecb9cada2015-12-08 15:45:58 -0700949 vlib_cli_output (vm, "%U", format_vxlan_tunnel, t);
Damjan Marionb2c31b62020-12-13 21:47:40 +0100950 }
Eyal Bari8f576482018-05-07 10:13:44 +0300951/* *INDENT-ON* */
952
Eyal Bari0fa56782018-06-04 12:25:05 +0300953 if (raw)
Eyal Baridd47eca2018-07-08 08:15:56 +0300954 {
955 vlib_cli_output (vm, "Raw IPv4 Hash Table:\n%U\n",
956 format_bihash_16_8, &vxm->vxlan4_tunnel_by_key,
957 1 /* verbose */ );
958 vlib_cli_output (vm, "Raw IPv6 Hash Table:\n%U\n",
959 format_bihash_24_8, &vxm->vxlan6_tunnel_by_key,
960 1 /* verbose */ );
961 }
Eyal Bari0fa56782018-06-04 12:25:05 +0300962
Ed Warnickecb9cada2015-12-08 15:45:58 -0700963 return 0;
964}
965
Billy McFallc5d9cda2016-09-14 10:39:58 -0400966/*?
967 * Display all the VXLAN Tunnel entries.
968 *
969 * @cliexpar
970 * Example of how to display the VXLAN Tunnel entries:
971 * @cliexstart{show vxlan tunnel}
Artem Glazychev839dcc02020-12-01 02:39:21 +0700972 * [0] src 10.0.3.1 dst 10.0.3.3 src_port 4789 dst_port 4789 vni 13
973 encap_fib_index 0 sw_if_index 5 decap_next l2
Billy McFallc5d9cda2016-09-14 10:39:58 -0400974 * @cliexend
975 ?*/
976/* *INDENT-OFF* */
Ed Warnickecb9cada2015-12-08 15:45:58 -0700977VLIB_CLI_COMMAND (show_vxlan_tunnel_command, static) = {
978 .path = "show vxlan tunnel",
Eyal Bari0fa56782018-06-04 12:25:05 +0300979 .short_help = "show vxlan tunnel [raw]",
Ed Warnickecb9cada2015-12-08 15:45:58 -0700980 .function = show_vxlan_tunnel_command_fn,
981};
Billy McFallc5d9cda2016-09-14 10:39:58 -0400982/* *INDENT-ON* */
Ed Warnickecb9cada2015-12-08 15:45:58 -0700983
Chris Luke99cb3352016-04-26 10:49:53 -0400984
Eyal Bari8f576482018-05-07 10:13:44 +0300985void
986vnet_int_vxlan_bypass_mode (u32 sw_if_index, u8 is_ip6, u8 is_enable)
John Lo2b81eb82017-01-30 13:12:10 -0500987{
Jon Loeligere3034cd2019-01-03 12:56:02 -0600988 vxlan_main_t *vxm = &vxlan_main;
989
990 if (pool_is_free_index (vxm->vnet_main->interface_main.sw_interfaces,
991 sw_if_index))
992 return;
993
994 is_enable = ! !is_enable;
995
John Lo2b81eb82017-01-30 13:12:10 -0500996 if (is_ip6)
Jon Loeligere3034cd2019-01-03 12:56:02 -0600997 {
998 if (clib_bitmap_get (vxm->bm_ip6_bypass_enabled_by_sw_if, sw_if_index)
999 != is_enable)
1000 {
1001 vnet_feature_enable_disable ("ip6-unicast", "ip6-vxlan-bypass",
1002 sw_if_index, is_enable, 0, 0);
1003 vxm->bm_ip6_bypass_enabled_by_sw_if =
1004 clib_bitmap_set (vxm->bm_ip6_bypass_enabled_by_sw_if,
1005 sw_if_index, is_enable);
1006 }
1007 }
John Lo2b81eb82017-01-30 13:12:10 -05001008 else
Jon Loeligere3034cd2019-01-03 12:56:02 -06001009 {
1010 if (clib_bitmap_get (vxm->bm_ip4_bypass_enabled_by_sw_if, sw_if_index)
1011 != is_enable)
1012 {
1013 vnet_feature_enable_disable ("ip4-unicast", "ip4-vxlan-bypass",
1014 sw_if_index, is_enable, 0, 0);
1015 vxm->bm_ip4_bypass_enabled_by_sw_if =
1016 clib_bitmap_set (vxm->bm_ip4_bypass_enabled_by_sw_if,
1017 sw_if_index, is_enable);
1018 }
1019 }
John Lo2b81eb82017-01-30 13:12:10 -05001020}
1021
1022
1023static clib_error_t *
1024set_ip_vxlan_bypass (u32 is_ip6,
Eyal Bari8f576482018-05-07 10:13:44 +03001025 unformat_input_t * input, vlib_cli_command_t * cmd)
John Lo2b81eb82017-01-30 13:12:10 -05001026{
Eyal Bari8f576482018-05-07 10:13:44 +03001027 unformat_input_t _line_input, *line_input = &_line_input;
1028 vnet_main_t *vnm = vnet_get_main ();
1029 clib_error_t *error = 0;
John Lo2b81eb82017-01-30 13:12:10 -05001030 u32 sw_if_index, is_enable;
1031
1032 sw_if_index = ~0;
1033 is_enable = 1;
1034
Eyal Bari8f576482018-05-07 10:13:44 +03001035 if (!unformat_user (input, unformat_line_input, line_input))
John Lo2b81eb82017-01-30 13:12:10 -05001036 return 0;
1037
1038 while (unformat_check_input (line_input) != UNFORMAT_END_OF_INPUT)
1039 {
Eyal Bari8f576482018-05-07 10:13:44 +03001040 if (unformat_user
1041 (line_input, unformat_vnet_sw_interface, vnm, &sw_if_index))
1042 ;
John Lo2b81eb82017-01-30 13:12:10 -05001043 else if (unformat (line_input, "del"))
Eyal Bari8f576482018-05-07 10:13:44 +03001044 is_enable = 0;
John Lo2b81eb82017-01-30 13:12:10 -05001045 else
Eyal Bari8f576482018-05-07 10:13:44 +03001046 {
John Lo2b81eb82017-01-30 13:12:10 -05001047 error = unformat_parse_error (line_input);
1048 goto done;
1049 }
1050 }
1051
1052 if (~0 == sw_if_index)
1053 {
1054 error = clib_error_return (0, "unknown interface `%U'",
1055 format_unformat_error, line_input);
1056 goto done;
1057 }
1058
1059 vnet_int_vxlan_bypass_mode (sw_if_index, is_ip6, is_enable);
1060
Eyal Bari8f576482018-05-07 10:13:44 +03001061done:
Billy McFalla9a20e72017-02-15 11:39:12 -05001062 unformat_free (line_input);
1063
John Lo2b81eb82017-01-30 13:12:10 -05001064 return error;
1065}
1066
1067static clib_error_t *
1068set_ip4_vxlan_bypass (vlib_main_t * vm,
Eyal Bari8f576482018-05-07 10:13:44 +03001069 unformat_input_t * input, vlib_cli_command_t * cmd)
John Lo2b81eb82017-01-30 13:12:10 -05001070{
1071 return set_ip_vxlan_bypass (0, input, cmd);
1072}
1073
1074/*?
Eyal Bari8f576482018-05-07 10:13:44 +03001075 * This command adds the 'ip4-vxlan-bypass' graph node for a given interface.
John Lo2b81eb82017-01-30 13:12:10 -05001076 * By adding the IPv4 vxlan-bypass graph node to an interface, the node checks
Nathan Skrzypczak5e6a1652021-10-06 15:03:35 +02001077 * for and validate input vxlan packet and bypass ip4-lookup, ip4-local,
Eyal Bari8f576482018-05-07 10:13:44 +03001078 * ip4-udp-lookup nodes to speedup vxlan packet forwarding. This node will
John Lo2b81eb82017-01-30 13:12:10 -05001079 * cause extra overhead to for non-vxlan packets which is kept at a minimum.
1080 *
1081 * @cliexpar
1082 * @parblock
1083 * Example of graph node before ip4-vxlan-bypass is enabled:
1084 * @cliexstart{show vlib graph ip4-vxlan-bypass}
1085 * Name Next Previous
1086 * ip4-vxlan-bypass error-drop [0]
1087 * vxlan4-input [1]
Eyal Bari8f576482018-05-07 10:13:44 +03001088 * ip4-lookup [2]
John Lo2b81eb82017-01-30 13:12:10 -05001089 * @cliexend
1090 *
1091 * Example of how to enable ip4-vxlan-bypass on an interface:
1092 * @cliexcmd{set interface ip vxlan-bypass GigabitEthernet2/0/0}
1093 *
1094 * Example of graph node after ip4-vxlan-bypass is enabled:
1095 * @cliexstart{show vlib graph ip4-vxlan-bypass}
Eyal Bari8f576482018-05-07 10:13:44 +03001096 * Name Next Previous
1097 * ip4-vxlan-bypass error-drop [0] ip4-input
1098 * vxlan4-input [1] ip4-input-no-checksum
1099 * ip4-lookup [2]
John Lo2b81eb82017-01-30 13:12:10 -05001100 * @cliexend
1101 *
Paul Vinciguerrabdc0e6b2018-09-22 05:32:50 -07001102 * Example of how to display the feature enabled on an interface:
John Lo2b81eb82017-01-30 13:12:10 -05001103 * @cliexstart{show ip interface features GigabitEthernet2/0/0}
1104 * IP feature paths configured on GigabitEthernet2/0/0...
1105 * ...
1106 * ipv4 unicast:
1107 * ip4-vxlan-bypass
1108 * ip4-lookup
1109 * ...
1110 * @cliexend
1111 *
1112 * Example of how to disable ip4-vxlan-bypass on an interface:
1113 * @cliexcmd{set interface ip vxlan-bypass GigabitEthernet2/0/0 del}
1114 * @endparblock
1115?*/
1116/* *INDENT-OFF* */
1117VLIB_CLI_COMMAND (set_interface_ip_vxlan_bypass_command, static) = {
1118 .path = "set interface ip vxlan-bypass",
1119 .function = set_ip4_vxlan_bypass,
1120 .short_help = "set interface ip vxlan-bypass <interface> [del]",
1121};
1122/* *INDENT-ON* */
1123
1124static clib_error_t *
1125set_ip6_vxlan_bypass (vlib_main_t * vm,
Eyal Bari8f576482018-05-07 10:13:44 +03001126 unformat_input_t * input, vlib_cli_command_t * cmd)
John Lo2b81eb82017-01-30 13:12:10 -05001127{
1128 return set_ip_vxlan_bypass (1, input, cmd);
1129}
1130
1131/*?
Eyal Bari8f576482018-05-07 10:13:44 +03001132 * This command adds the 'ip6-vxlan-bypass' graph node for a given interface.
John Lo2b81eb82017-01-30 13:12:10 -05001133 * By adding the IPv6 vxlan-bypass graph node to an interface, the node checks
Nathan Skrzypczak5e6a1652021-10-06 15:03:35 +02001134 * for and validate input vxlan packet and bypass ip6-lookup, ip6-local,
Eyal Bari8f576482018-05-07 10:13:44 +03001135 * ip6-udp-lookup nodes to speedup vxlan packet forwarding. This node will
John Lo2b81eb82017-01-30 13:12:10 -05001136 * cause extra overhead to for non-vxlan packets which is kept at a minimum.
1137 *
1138 * @cliexpar
1139 * @parblock
1140 * Example of graph node before ip6-vxlan-bypass is enabled:
1141 * @cliexstart{show vlib graph ip6-vxlan-bypass}
1142 * Name Next Previous
1143 * ip6-vxlan-bypass error-drop [0]
1144 * vxlan6-input [1]
Eyal Bari8f576482018-05-07 10:13:44 +03001145 * ip6-lookup [2]
John Lo2b81eb82017-01-30 13:12:10 -05001146 * @cliexend
1147 *
1148 * Example of how to enable ip6-vxlan-bypass on an interface:
1149 * @cliexcmd{set interface ip6 vxlan-bypass GigabitEthernet2/0/0}
1150 *
1151 * Example of graph node after ip6-vxlan-bypass is enabled:
1152 * @cliexstart{show vlib graph ip6-vxlan-bypass}
Eyal Bari8f576482018-05-07 10:13:44 +03001153 * Name Next Previous
1154 * ip6-vxlan-bypass error-drop [0] ip6-input
1155 * vxlan6-input [1] ip4-input-no-checksum
1156 * ip6-lookup [2]
John Lo2b81eb82017-01-30 13:12:10 -05001157 * @cliexend
1158 *
Paul Vinciguerrabdc0e6b2018-09-22 05:32:50 -07001159 * Example of how to display the feature enabled on an interface:
John Lo2b81eb82017-01-30 13:12:10 -05001160 * @cliexstart{show ip interface features GigabitEthernet2/0/0}
1161 * IP feature paths configured on GigabitEthernet2/0/0...
1162 * ...
1163 * ipv6 unicast:
1164 * ip6-vxlan-bypass
1165 * ip6-lookup
1166 * ...
1167 * @cliexend
1168 *
1169 * Example of how to disable ip6-vxlan-bypass on an interface:
1170 * @cliexcmd{set interface ip6 vxlan-bypass GigabitEthernet2/0/0 del}
1171 * @endparblock
1172?*/
1173/* *INDENT-OFF* */
1174VLIB_CLI_COMMAND (set_interface_ip6_vxlan_bypass_command, static) = {
1175 .path = "set interface ip6 vxlan-bypass",
1176 .function = set_ip6_vxlan_bypass,
Paul Vinciguerra5fb22782019-11-07 17:20:48 -05001177 .short_help = "set interface ip6 vxlan-bypass <interface> [del]",
John Lo2b81eb82017-01-30 13:12:10 -05001178};
1179/* *INDENT-ON* */
1180
eyal bariaf86a482018-04-17 11:20:27 +03001181int
1182vnet_vxlan_add_del_rx_flow (u32 hw_if_index, u32 t_index, int is_add)
1183{
1184 vxlan_main_t *vxm = &vxlan_main;
1185 vxlan_tunnel_t *t = pool_elt_at_index (vxm->tunnels, t_index);
1186 vnet_main_t *vnm = vnet_get_main ();
1187 if (is_add)
1188 {
1189 if (t->flow_index == ~0)
1190 {
1191 vxlan_main_t *vxm = &vxlan_main;
1192 vnet_flow_t flow = {
1193 .actions =
Chenmin Sun1ec9fdb2019-12-05 01:41:35 +08001194 VNET_FLOW_ACTION_REDIRECT_TO_NODE | VNET_FLOW_ACTION_MARK |
1195 VNET_FLOW_ACTION_BUFFER_ADVANCE,
eyal bariaf86a482018-04-17 11:20:27 +03001196 .mark_flow_id = t->dev_instance + vxm->flow_id_start,
1197 .redirect_node_index = vxlan4_flow_input_node.index,
Chenmin Sun1ec9fdb2019-12-05 01:41:35 +08001198 .buffer_advance = sizeof (ethernet_header_t),
eyal bariaf86a482018-04-17 11:20:27 +03001199 .type = VNET_FLOW_TYPE_IP4_VXLAN,
1200 .ip4_vxlan = {
Chenmin Sun34bfa502020-07-27 17:40:17 +08001201 .protocol.prot = IP_PROTOCOL_UDP,
1202 .src_addr.addr = t->dst.ip4,
1203 .dst_addr.addr = t->src.ip4,
1204 .src_addr.mask.as_u32 = ~0,
1205 .dst_addr.mask.as_u32 = ~0,
Artem Glazychev839dcc02020-12-01 02:39:21 +07001206 .dst_port.port = t->src_port,
Chenmin Sun34bfa502020-07-27 17:40:17 +08001207 .dst_port.mask = 0xFF,
eyal bariaf86a482018-04-17 11:20:27 +03001208 .vni = t->vni,
1209 }
1210 ,
1211 };
1212 vnet_flow_add (vnm, &flow, &t->flow_index);
1213 }
1214 return vnet_flow_enable (vnm, t->flow_index, hw_if_index);
1215 }
1216 /* flow index is removed when the tunnel is deleted */
1217 return vnet_flow_disable (vnm, t->flow_index, hw_if_index);
1218}
1219
1220u32
1221vnet_vxlan_get_tunnel_index (u32 sw_if_index)
1222{
1223 vxlan_main_t *vxm = &vxlan_main;
1224
Eyal Baricd307742018-07-22 12:45:15 +03001225 if (sw_if_index >= vec_len (vxm->tunnel_index_by_sw_if_index))
eyal bariaf86a482018-04-17 11:20:27 +03001226 return ~0;
1227 return vxm->tunnel_index_by_sw_if_index[sw_if_index];
1228}
1229
1230static clib_error_t *
1231vxlan_offload_command_fn (vlib_main_t * vm,
1232 unformat_input_t * input, vlib_cli_command_t * cmd)
1233{
1234 unformat_input_t _line_input, *line_input = &_line_input;
1235
1236 /* Get a line of input. */
1237 if (!unformat_user (input, unformat_line_input, line_input))
1238 return 0;
1239
1240 vnet_main_t *vnm = vnet_get_main ();
1241 u32 rx_sw_if_index = ~0;
1242 u32 hw_if_index = ~0;
1243 int is_add = 1;
1244
1245 while (unformat_check_input (line_input) != UNFORMAT_END_OF_INPUT)
1246 {
1247 if (unformat (line_input, "hw %U", unformat_vnet_hw_interface, vnm,
1248 &hw_if_index))
1249 continue;
1250 if (unformat (line_input, "rx %U", unformat_vnet_sw_interface, vnm,
1251 &rx_sw_if_index))
1252 continue;
1253 if (unformat (line_input, "del"))
1254 {
1255 is_add = 0;
1256 continue;
1257 }
1258 return clib_error_return (0, "unknown input `%U'",
1259 format_unformat_error, line_input);
1260 }
1261
1262 if (rx_sw_if_index == ~0)
1263 return clib_error_return (0, "missing rx interface");
1264 if (hw_if_index == ~0)
1265 return clib_error_return (0, "missing hw interface");
1266
1267 u32 t_index = vnet_vxlan_get_tunnel_index (rx_sw_if_index);;
1268 if (t_index == ~0)
1269 return clib_error_return (0, "%U is not a vxlan tunnel",
1270 format_vnet_sw_if_index_name, vnm,
1271 rx_sw_if_index);
1272
1273 vxlan_main_t *vxm = &vxlan_main;
1274 vxlan_tunnel_t *t = pool_elt_at_index (vxm->tunnels, t_index);
1275
1276 if (!ip46_address_is_ip4 (&t->dst))
1277 return clib_error_return (0, "currently only IPV4 tunnels are supported");
1278
1279 vnet_hw_interface_t *hw_if = vnet_get_hw_interface (vnm, hw_if_index);
1280 ip4_main_t *im = &ip4_main;
1281 u32 rx_fib_index =
1282 vec_elt (im->fib_index_by_sw_if_index, hw_if->sw_if_index);
1283
1284 if (t->encap_fib_index != rx_fib_index)
1285 return clib_error_return (0, "interface/tunnel fib mismatch");
1286
1287 if (vnet_vxlan_add_del_rx_flow (hw_if_index, t_index, is_add))
1288 return clib_error_return (0, "error %s flow",
1289 is_add ? "enabling" : "disabling");
1290
1291 return 0;
1292}
1293
1294/* *INDENT-OFF* */
1295VLIB_CLI_COMMAND (vxlan_offload_command, static) = {
1296 .path = "set flow-offload vxlan",
1297 .short_help =
Paul Vinciguerrabdc0e6b2018-09-22 05:32:50 -07001298 "set flow-offload vxlan hw <interface-name> rx <tunnel-name> [del]",
eyal bariaf86a482018-04-17 11:20:27 +03001299 .function = vxlan_offload_command_fn,
1300};
1301/* *INDENT-ON* */
1302
Eyal Bari0fa56782018-06-04 12:25:05 +03001303#define VXLAN_HASH_NUM_BUCKETS (2 * 1024)
1304#define VXLAN_HASH_MEMORY_SIZE (1 << 20)
1305
Eyal Bari8f576482018-05-07 10:13:44 +03001306clib_error_t *
1307vxlan_init (vlib_main_t * vm)
Ed Warnickecb9cada2015-12-08 15:45:58 -07001308{
Eyal Bari8f576482018-05-07 10:13:44 +03001309 vxlan_main_t *vxm = &vxlan_main;
1310
1311 vxm->vnet_main = vnet_get_main ();
Ed Warnickecb9cada2015-12-08 15:45:58 -07001312 vxm->vlib_main = vm;
1313
eyal bariaf86a482018-04-17 11:20:27 +03001314 vnet_flow_get_range (vxm->vnet_main, "vxlan", 1024 * 1024,
1315 &vxm->flow_id_start);
1316
Jon Loeligere3034cd2019-01-03 12:56:02 -06001317 vxm->bm_ip4_bypass_enabled_by_sw_if = 0;
1318 vxm->bm_ip6_bypass_enabled_by_sw_if = 0;
1319
Chris Luke99cb3352016-04-26 10:49:53 -04001320 /* initialize the ip6 hash */
Eyal Baridd47eca2018-07-08 08:15:56 +03001321 clib_bihash_init_16_8 (&vxm->vxlan4_tunnel_by_key, "vxlan4",
1322 VXLAN_HASH_NUM_BUCKETS, VXLAN_HASH_MEMORY_SIZE);
1323 clib_bihash_init_24_8 (&vxm->vxlan6_tunnel_by_key, "vxlan6",
Eyal Bari0fa56782018-06-04 12:25:05 +03001324 VXLAN_HASH_NUM_BUCKETS, VXLAN_HASH_MEMORY_SIZE);
Nick Zavaritsky27518c22020-02-27 15:54:58 +00001325 vxm->vtep_table = vtep_table_create ();
Eyal Bari8f576482018-05-07 10:13:44 +03001326 vxm->mcast_shared = hash_create_mem (0,
1327 sizeof (ip46_address_t),
1328 sizeof (mcast_shared_t));
Chris Luke99cb3352016-04-26 10:49:53 -04001329
Eyal Bari8f576482018-05-07 10:13:44 +03001330 fib_node_register_type (FIB_NODE_TYPE_VXLAN_TUNNEL, &vxlan_vft);
John Loc42912d2016-11-07 18:30:47 -05001331
Ed Warnickecb9cada2015-12-08 15:45:58 -07001332 return 0;
1333}
1334
Eyal Bari8f576482018-05-07 10:13:44 +03001335VLIB_INIT_FUNCTION (vxlan_init);
Jon Loeliger3d460bd2018-02-01 16:36:12 -06001336
1337/*
Eyal Bari8f576482018-05-07 10:13:44 +03001338 * fd.io coding-style-patch-verification: ON
1339 *
Jon Loeliger3d460bd2018-02-01 16:36:12 -06001340 * Local Variables:
1341 * eval: (c-set-style "gnu")
1342 * End:
1343 */