blob: 6bc4b67d10dc4e6084ccc52d9e76f36dd25b6c19 [file] [log] [blame]
Marco Varleseb598f1d2017-09-19 14:25:28 +02001/*
2 * Copyright (c) 2017 SUSE LLC.
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/geneve/geneve.h>
16#include <vnet/ip/format.h>
17#include <vnet/fib/fib_entry.h>
18#include <vnet/fib/fib_table.h>
19#include <vnet/mfib/mfib_table.h>
20#include <vnet/adj/adj_mcast.h>
21#include <vnet/interface.h>
22#include <vlib/vlib.h>
23
24/**
25 * @file
26 * @brief GENEVE.
27 *
28 * GENEVE provides the features needed to allow L2 bridge domains (BDs)
29 * to span multiple servers. This is done by building an L2 overlay on
30 * top of an L3 network underlay using GENEVE tunnels.
31 *
32 * This makes it possible for servers to be co-located in the same data
33 * center or be separated geographically as long as they are reachable
34 * through the underlay L3 network.
Marco Varleseb598f1d2017-09-19 14:25:28 +020035 */
36
37
38geneve_main_t geneve_main;
39
40static u8 *
41format_decap_next (u8 * s, va_list * args)
42{
43 u32 next_index = va_arg (*args, u32);
44
45 switch (next_index)
46 {
47 case GENEVE_INPUT_NEXT_DROP:
48 return format (s, "drop");
49 case GENEVE_INPUT_NEXT_L2_INPUT:
50 return format (s, "l2");
51 default:
52 return format (s, "index %d", next_index);
53 }
54 return s;
55}
56
57u8 *
58format_geneve_tunnel (u8 * s, va_list * args)
59{
60 geneve_tunnel_t *t = va_arg (*args, geneve_tunnel_t *);
61 geneve_main_t *ngm = &geneve_main;
62
John Lo4478d8e2018-01-12 17:15:25 -050063 s = format (s, "[%d] lcl %U rmt %U vni %d fib-idx %d sw-if-idx %d ",
Marco Varleseb598f1d2017-09-19 14:25:28 +020064 t - ngm->tunnels,
65 format_ip46_address, &t->local, IP46_TYPE_ANY,
66 format_ip46_address, &t->remote, IP46_TYPE_ANY,
John Lo4478d8e2018-01-12 17:15:25 -050067 t->vni, t->encap_fib_index, t->sw_if_index);
Marco Varleseb598f1d2017-09-19 14:25:28 +020068
John Lo4478d8e2018-01-12 17:15:25 -050069 s = format (s, "encap-dpo-idx %d ", t->next_dpo.dpoi_index);
70 s = format (s, "decap-next-%U ", format_decap_next, t->decap_next_index);
Marco Varleseb598f1d2017-09-19 14:25:28 +020071
John Lo4478d8e2018-01-12 17:15:25 -050072 if (PREDICT_FALSE (ip46_address_is_multicast (&t->remote)))
73 s = format (s, "mcast-sw-if-idx %d ", t->mcast_sw_if_index);
74
Marco Varleseb598f1d2017-09-19 14:25:28 +020075 return s;
76}
77
78static u8 *
79format_geneve_name (u8 * s, va_list * args)
80{
81 u32 dev_instance = va_arg (*args, u32);
82 return format (s, "geneve_tunnel%d", dev_instance);
83}
84
85static uword
86dummy_interface_tx (vlib_main_t * vm,
87 vlib_node_runtime_t * node, vlib_frame_t * frame)
88{
89 clib_warning ("you shouldn't be here, leaking buffers...");
90 return frame->n_vectors;
91}
92
93static clib_error_t *
94geneve_interface_admin_up_down (vnet_main_t * vnm, u32 hw_if_index, u32 flags)
95{
96 u32 hw_flags = (flags & VNET_SW_INTERFACE_FLAG_ADMIN_UP) ?
97 VNET_HW_INTERFACE_FLAG_LINK_UP : 0;
98 vnet_hw_interface_set_flags (vnm, hw_if_index, hw_flags);
99
100 return /* no error */ 0;
101}
102
103/* *INDENT-OFF* */
104VNET_DEVICE_CLASS (geneve_device_class, static) = {
105 .name = "GENEVE",
106 .format_device_name = format_geneve_name,
107 .format_tx_trace = format_geneve_encap_trace,
108 .tx_function = dummy_interface_tx,
109 .admin_up_down_function = geneve_interface_admin_up_down,
110};
111/* *INDENT-ON* */
112
113static u8 *
114format_geneve_header_with_length (u8 * s, va_list * args)
115{
116 u32 dev_instance = va_arg (*args, u32);
117 s = format (s, "unimplemented dev %u", dev_instance);
118 return s;
119}
120
121/* *INDENT-OFF* */
122VNET_HW_INTERFACE_CLASS (geneve_hw_class) = {
123 .name = "GENEVE",
124 .format_header = format_geneve_header_with_length,
125 .build_rewrite = default_build_rewrite,
126};
127/* *INDENT-ON* */
128
129static void
130geneve_tunnel_restack_dpo (geneve_tunnel_t * t)
131{
132 dpo_id_t dpo = DPO_INVALID;
133 u32 encap_index = ip46_address_is_ip4 (&t->remote) ?
134 geneve4_encap_node.index : geneve6_encap_node.index;
135 fib_forward_chain_type_t forw_type = ip46_address_is_ip4 (&t->remote) ?
136 FIB_FORW_CHAIN_TYPE_UNICAST_IP4 : FIB_FORW_CHAIN_TYPE_UNICAST_IP6;
137
138 fib_entry_contribute_forwarding (t->fib_entry_index, forw_type, &dpo);
139 dpo_stack_from_node (encap_index, &t->next_dpo, &dpo);
140 dpo_reset (&dpo);
141}
142
143static geneve_tunnel_t *
144geneve_tunnel_from_fib_node (fib_node_t * node)
145{
Marco Varleseb598f1d2017-09-19 14:25:28 +0200146 ASSERT (FIB_NODE_TYPE_GENEVE_TUNNEL == node->fn_type);
Marco Varleseb598f1d2017-09-19 14:25:28 +0200147 return ((geneve_tunnel_t *) (((char *) node) -
148 STRUCT_OFFSET_OF (geneve_tunnel_t, node)));
149}
150
151/**
152 * Function definition to backwalk a FIB node -
153 * Here we will restack the new dpo of GENEVE DIP to encap node.
154 */
155static fib_node_back_walk_rc_t
156geneve_tunnel_back_walk (fib_node_t * node, fib_node_back_walk_ctx_t * ctx)
157{
158 geneve_tunnel_restack_dpo (geneve_tunnel_from_fib_node (node));
159 return (FIB_NODE_BACK_WALK_CONTINUE);
160}
161
162/**
163 * Function definition to get a FIB node from its index
164 */
165static fib_node_t *
166geneve_tunnel_fib_node_get (fib_node_index_t index)
167{
168 geneve_tunnel_t *t;
169 geneve_main_t *vxm = &geneve_main;
170
171 t = pool_elt_at_index (vxm->tunnels, index);
172
173 return (&t->node);
174}
175
176/**
177 * Function definition to inform the FIB node that its last lock has gone.
178 */
179static void
180geneve_tunnel_last_lock_gone (fib_node_t * node)
181{
182 /*
183 * The GENEVE tunnel is a root of the graph. As such
184 * it never has children and thus is never locked.
185 */
186 ASSERT (0);
187}
188
189/*
190 * Virtual function table registered by GENEVE tunnels
191 * for participation in the FIB object graph.
192 */
193const static fib_node_vft_t geneve_vft = {
194 .fnv_get = geneve_tunnel_fib_node_get,
195 .fnv_last_lock = geneve_tunnel_last_lock_gone,
196 .fnv_back_walk = geneve_tunnel_back_walk,
197};
198
199
200#define foreach_copy_field \
201_(vni) \
202_(mcast_sw_if_index) \
203_(encap_fib_index) \
204_(decap_next_index) \
205_(local) \
206_(remote)
207
208static int
209geneve_rewrite (geneve_tunnel_t * t, bool is_ip6)
210{
211 union
212 {
213 ip4_geneve_header_t *h4;
214 ip6_geneve_header_t *h6;
215 u8 *rw;
216 } r =
217 {
218 .rw = 0};
219 int len = is_ip6 ? sizeof *r.h6 : sizeof *r.h4;
220#if SUPPORT_OPTIONS_HEADER==1
221 len += t->options_len;
222#endif
223
224 vec_validate_aligned (r.rw, len - 1, CLIB_CACHE_LINE_BYTES);
225
226 udp_header_t *udp;
227 geneve_header_t *geneve;
228 /* Fixed portion of the (outer) ip header */
229 if (!is_ip6)
230 {
231 ip4_header_t *ip = &r.h4->ip4;
232 udp = &r.h4->udp, geneve = &r.h4->geneve;
233 ip->ip_version_and_header_length = 0x45;
234 ip->ttl = 254;
235 ip->protocol = IP_PROTOCOL_UDP;
236
237 ip->src_address = t->local.ip4;
238 ip->dst_address = t->remote.ip4;
239
240 /* we fix up the ip4 header length and checksum after-the-fact */
241 ip->checksum = ip4_header_checksum (ip);
242 }
243 else
244 {
245 ip6_header_t *ip = &r.h6->ip6;
246 udp = &r.h6->udp, geneve = &r.h6->geneve;
247 ip->ip_version_traffic_class_and_flow_label =
248 clib_host_to_net_u32 (6 << 28);
249 ip->hop_limit = 255;
250 ip->protocol = IP_PROTOCOL_UDP;
251
252 ip->src_address = t->local.ip6;
253 ip->dst_address = t->remote.ip6;
254 }
255
256 /* UDP header, randomize local port on something, maybe? */
257 udp->src_port = clib_host_to_net_u16 (5251);
258 udp->dst_port = clib_host_to_net_u16 (UDP_DST_PORT_geneve);
259
260 /* GENEVE header */
261 vnet_set_geneve_version (geneve, GENEVE_VERSION);
262#if SUPPORT_OPTIONS_HEADER==1
263 vnet_set_geneve_options_len (geneve, t->options_len);
264#else
265 vnet_set_geneve_options_len (geneve, 0);
266#endif
267 vnet_set_geneve_oamframe_bit (geneve, 0);
268 vnet_set_geneve_critical_bit (geneve, 0);
269 vnet_set_geneve_protocol (geneve, GENEVE_ETH_PROTOCOL);
Marco Varlese60d48bb2017-11-20 09:20:38 +0100270
271 vnet_geneve_hdr_1word_hton (geneve);
272
Marco Varleseb598f1d2017-09-19 14:25:28 +0200273 vnet_set_geneve_vni (geneve, t->vni);
274
275 t->rewrite = r.rw;
276 return (0);
277}
278
279static bool
280geneve_decap_next_is_valid (geneve_main_t * vxm, u32 is_ip6,
281 u32 decap_next_index)
282{
283 vlib_main_t *vm = vxm->vlib_main;
284 u32 input_idx =
285 (!is_ip6) ? geneve4_input_node.index : geneve6_input_node.index;
286 vlib_node_runtime_t *r = vlib_node_get_runtime (vm, input_idx);
287
288 return decap_next_index < r->n_next_nodes;
289}
290
Marco Varleseb598f1d2017-09-19 14:25:28 +0200291static uword
292vtep_addr_ref (ip46_address_t * ip)
293{
294 uword *vtep = ip46_address_is_ip4 (ip) ?
295 hash_get (geneve_main.vtep4, ip->ip4.as_u32) :
296 hash_get_mem (geneve_main.vtep6, &ip->ip6);
297 if (vtep)
298 return ++(*vtep);
299 ip46_address_is_ip4 (ip) ?
300 hash_set (geneve_main.vtep4, ip->ip4.as_u32, 1) :
John Loe6bfeab2018-01-04 16:39:42 -0500301 hash_set_mem_alloc (&geneve_main.vtep6, &ip->ip6, 1);
Marco Varleseb598f1d2017-09-19 14:25:28 +0200302 return 1;
303}
304
305static uword
306vtep_addr_unref (ip46_address_t * ip)
307{
308 uword *vtep = ip46_address_is_ip4 (ip) ?
309 hash_get (geneve_main.vtep4, ip->ip4.as_u32) :
310 hash_get_mem (geneve_main.vtep6, &ip->ip6);
311 ASSERT (vtep);
312 if (--(*vtep) != 0)
313 return *vtep;
314 ip46_address_is_ip4 (ip) ?
315 hash_unset (geneve_main.vtep4, ip->ip4.as_u32) :
John Loe6bfeab2018-01-04 16:39:42 -0500316 hash_unset_mem_free (&geneve_main.vtep6, &ip->ip6);
Marco Varleseb598f1d2017-09-19 14:25:28 +0200317 return 0;
318}
319
320typedef CLIB_PACKED (union
321 {
322 struct
323 {
324 fib_node_index_t mfib_entry_index;
325 adj_index_t mcast_adj_index;
326 }; u64 as_u64;
327 }) mcast_shared_t;
328
329static inline mcast_shared_t
330mcast_shared_get (ip46_address_t * ip)
331{
332 ASSERT (ip46_address_is_multicast (ip));
333 uword *p = hash_get_mem (geneve_main.mcast_shared, ip);
334 ASSERT (p);
335 return (mcast_shared_t)
336 {
337 .as_u64 = *p};
338}
339
340static inline void
341mcast_shared_add (ip46_address_t * remote,
342 fib_node_index_t mfei, adj_index_t ai)
343{
344 mcast_shared_t new_ep = {
345 .mcast_adj_index = ai,
346 .mfib_entry_index = mfei,
347 };
348
John Loe6bfeab2018-01-04 16:39:42 -0500349 hash_set_mem_alloc (&geneve_main.mcast_shared, remote, new_ep.as_u64);
Marco Varleseb598f1d2017-09-19 14:25:28 +0200350}
351
352static inline void
353mcast_shared_remove (ip46_address_t * remote)
354{
355 mcast_shared_t ep = mcast_shared_get (remote);
356
357 adj_unlock (ep.mcast_adj_index);
358 mfib_table_entry_delete_index (ep.mfib_entry_index, MFIB_SOURCE_GENEVE);
359
John Loe6bfeab2018-01-04 16:39:42 -0500360 hash_unset_mem_free (&geneve_main.mcast_shared, remote);
Marco Varleseb598f1d2017-09-19 14:25:28 +0200361}
362
363static inline fib_protocol_t
364fib_ip_proto (bool is_ip6)
365{
366 return (is_ip6) ? FIB_PROTOCOL_IP6 : FIB_PROTOCOL_IP4;
367}
368
369int vnet_geneve_add_del_tunnel
370 (vnet_geneve_add_del_tunnel_args_t * a, u32 * sw_if_indexp)
371{
372 geneve_main_t *vxm = &geneve_main;
373 geneve_tunnel_t *t = 0;
374 vnet_main_t *vnm = vxm->vnet_main;
375 uword *p;
376 u32 hw_if_index = ~0;
377 u32 sw_if_index = ~0;
378 int rv;
379 geneve4_tunnel_key_t key4;
380 geneve6_tunnel_key_t key6;
381 u32 is_ip6 = a->is_ip6;
382
383 if (!is_ip6)
384 {
385 key4.remote = a->remote.ip4.as_u32;
Marco Varlese60d48bb2017-11-20 09:20:38 +0100386 key4.vni =
387 clib_host_to_net_u32 ((a->vni << GENEVE_VNI_SHIFT) & GENEVE_VNI_MASK);
Marco Varleseb598f1d2017-09-19 14:25:28 +0200388 p = hash_get (vxm->geneve4_tunnel_by_key, key4.as_u64);
389 }
390 else
391 {
392 key6.remote = a->remote.ip6;
Marco Varlese60d48bb2017-11-20 09:20:38 +0100393 key6.vni =
394 clib_host_to_net_u32 ((a->vni << GENEVE_VNI_SHIFT) & GENEVE_VNI_MASK);
Marco Varleseb598f1d2017-09-19 14:25:28 +0200395 p = hash_get_mem (vxm->geneve6_tunnel_by_key, &key6);
396 }
397
398 if (a->is_add)
399 {
400 l2input_main_t *l2im = &l2input_main;
401
402 /* adding a tunnel: tunnel must not already exist */
403 if (p)
404 return VNET_API_ERROR_TUNNEL_EXIST;
405
406 /*if not set explicitly, default to l2 */
407 if (a->decap_next_index == ~0)
408 a->decap_next_index = GENEVE_INPUT_NEXT_L2_INPUT;
409 if (!geneve_decap_next_is_valid (vxm, is_ip6, a->decap_next_index))
410 return VNET_API_ERROR_INVALID_DECAP_NEXT;
411
412 pool_get_aligned (vxm->tunnels, t, CLIB_CACHE_LINE_BYTES);
413 memset (t, 0, sizeof (*t));
414
415 /* copy from arg structure */
416#define _(x) t->x = a->x;
417 foreach_copy_field;
418#undef _
419
420 rv = geneve_rewrite (t, is_ip6);
421 if (rv)
422 {
423 pool_put (vxm->tunnels, t);
424 return rv;
425 }
426
427 /* copy the key */
428 if (is_ip6)
John Loe6bfeab2018-01-04 16:39:42 -0500429 hash_set_mem_alloc (&vxm->geneve6_tunnel_by_key, &key6,
430 t - vxm->tunnels);
Marco Varleseb598f1d2017-09-19 14:25:28 +0200431 else
432 hash_set (vxm->geneve4_tunnel_by_key, key4.as_u64, t - vxm->tunnels);
433
434 vnet_hw_interface_t *hi;
435 if (vec_len (vxm->free_geneve_tunnel_hw_if_indices) > 0)
436 {
437 vnet_interface_main_t *im = &vnm->interface_main;
438 hw_if_index = vxm->free_geneve_tunnel_hw_if_indices
439 [vec_len (vxm->free_geneve_tunnel_hw_if_indices) - 1];
440 _vec_len (vxm->free_geneve_tunnel_hw_if_indices) -= 1;
441
442 hi = vnet_get_hw_interface (vnm, hw_if_index);
443 hi->dev_instance = t - vxm->tunnels;
444 hi->hw_instance = hi->dev_instance;
445
446 /* clear old stats of freed tunnel before reuse */
447 sw_if_index = hi->sw_if_index;
448 vnet_interface_counter_lock (im);
449 vlib_zero_combined_counter
450 (&im->combined_sw_if_counters[VNET_INTERFACE_COUNTER_TX],
451 sw_if_index);
452 vlib_zero_combined_counter (&im->combined_sw_if_counters
453 [VNET_INTERFACE_COUNTER_RX],
454 sw_if_index);
455 vlib_zero_simple_counter (&im->sw_if_counters
456 [VNET_INTERFACE_COUNTER_DROP],
457 sw_if_index);
458 vnet_interface_counter_unlock (im);
459 }
460 else
461 {
462 hw_if_index = vnet_register_interface
463 (vnm, geneve_device_class.index, t - vxm->tunnels,
464 geneve_hw_class.index, t - vxm->tunnels);
465 hi = vnet_get_hw_interface (vnm, hw_if_index);
466 }
467
468 t->hw_if_index = hw_if_index;
469 t->sw_if_index = sw_if_index = hi->sw_if_index;
470
471 vec_validate_init_empty (vxm->tunnel_index_by_sw_if_index, sw_if_index,
472 ~0);
473 vxm->tunnel_index_by_sw_if_index[sw_if_index] = t - vxm->tunnels;
474
475 /* setup l2 input config with l2 feature and bd 0 to drop packet */
476 vec_validate (l2im->configs, sw_if_index);
477 l2im->configs[sw_if_index].feature_bitmap = L2INPUT_FEAT_DROP;
478 l2im->configs[sw_if_index].bd_index = 0;
479
480 vnet_sw_interface_t *si = vnet_get_sw_interface (vnm, sw_if_index);
481 si->flags &= ~VNET_SW_INTERFACE_FLAG_HIDDEN;
482 vnet_sw_interface_set_flags (vnm, sw_if_index,
483 VNET_SW_INTERFACE_FLAG_ADMIN_UP);
484
485 fib_node_init (&t->node, FIB_NODE_TYPE_GENEVE_TUNNEL);
486 fib_prefix_t tun_remote_pfx;
487 u32 encap_index = !is_ip6 ?
488 geneve4_encap_node.index : geneve6_encap_node.index;
489 vnet_flood_class_t flood_class = VNET_FLOOD_CLASS_TUNNEL_NORMAL;
490
491 fib_prefix_from_ip46_addr (&t->remote, &tun_remote_pfx);
492 if (!ip46_address_is_multicast (&t->remote))
493 {
494 /* Unicast tunnel -
495 * source the FIB entry for the tunnel's destination
496 * and become a child thereof. The tunnel will then get poked
497 * when the forwarding for the entry updates, and the tunnel can
498 * re-stack accordingly
499 */
500 vtep_addr_ref (&t->local);
501 t->fib_entry_index = fib_table_entry_special_add
502 (t->encap_fib_index, &tun_remote_pfx, FIB_SOURCE_RR,
503 FIB_ENTRY_FLAG_NONE);
504 t->sibling_index = fib_entry_child_add
505 (t->fib_entry_index, FIB_NODE_TYPE_GENEVE_TUNNEL,
506 t - vxm->tunnels);
507 geneve_tunnel_restack_dpo (t);
508 }
509 else
510 {
511 /* Multicast tunnel -
512 * as the same mcast group can be used for mutiple mcast tunnels
513 * with different VNIs, create the output fib adjecency only if
514 * it does not already exist
515 */
516 fib_protocol_t fp = fib_ip_proto (is_ip6);
517
518 if (vtep_addr_ref (&t->remote) == 1)
519 {
520 fib_node_index_t mfei;
521 adj_index_t ai;
522 fib_route_path_t path = {
523 .frp_proto = fib_proto_to_dpo (fp),
524 .frp_addr = zero_addr,
525 .frp_sw_if_index = 0xffffffff,
526 .frp_fib_index = ~0,
527 .frp_weight = 0,
528 .frp_flags = FIB_ROUTE_PATH_LOCAL,
529 };
530 const mfib_prefix_t mpfx = {
531 .fp_proto = fp,
532 .fp_len = (is_ip6 ? 128 : 32),
533 .fp_grp_addr = tun_remote_pfx.fp_addr,
534 };
535
536 /*
537 * Setup the (*,G) to receive traffic on the mcast group
538 * - the forwarding interface is for-us
539 * - the accepting interface is that from the API
540 */
541 mfib_table_entry_path_update (t->encap_fib_index,
542 &mpfx,
543 MFIB_SOURCE_GENEVE,
544 &path, MFIB_ITF_FLAG_FORWARD);
545
546 path.frp_sw_if_index = a->mcast_sw_if_index;
547 path.frp_flags = FIB_ROUTE_PATH_FLAG_NONE;
548 mfei = mfib_table_entry_path_update (t->encap_fib_index,
549 &mpfx,
550 MFIB_SOURCE_GENEVE,
551 &path,
552 MFIB_ITF_FLAG_ACCEPT);
553
554 /*
555 * Create the mcast adjacency to send traffic to the group
556 */
557 ai = adj_mcast_add_or_lock (fp,
558 fib_proto_to_link (fp),
559 a->mcast_sw_if_index);
560
561 /*
562 * create a new end-point
563 */
564 mcast_shared_add (&t->remote, mfei, ai);
565 }
566
567 dpo_id_t dpo = DPO_INVALID;
568 mcast_shared_t ep = mcast_shared_get (&t->remote);
569
570 /* Stack shared mcast remote mac addr rewrite on encap */
571 dpo_set (&dpo, DPO_ADJACENCY_MCAST,
572 fib_proto_to_dpo (fp), ep.mcast_adj_index);
573
574 dpo_stack_from_node (encap_index, &t->next_dpo, &dpo);
575 dpo_reset (&dpo);
576 flood_class = VNET_FLOOD_CLASS_TUNNEL_MASTER;
577 }
578
579 /* Set geneve tunnel output node */
580 hi->output_node_index = encap_index;
581
582 vnet_get_sw_interface (vnet_get_main (), sw_if_index)->flood_class =
583 flood_class;
584 }
585 else
586 {
587 /* deleting a tunnel: tunnel must exist */
588 if (!p)
589 return VNET_API_ERROR_NO_SUCH_ENTRY;
590
591 t = pool_elt_at_index (vxm->tunnels, p[0]);
592
593 sw_if_index = t->sw_if_index;
594 vnet_sw_interface_set_flags (vnm, t->sw_if_index, 0 /* down */ );
595 vnet_sw_interface_t *si = vnet_get_sw_interface (vnm, t->sw_if_index);
596 si->flags |= VNET_SW_INTERFACE_FLAG_HIDDEN;
597
598 /* make sure tunnel is removed from l2 bd or xconnect */
599 set_int_l2_mode (vxm->vlib_main, vnm, MODE_L3, t->sw_if_index, 0, 0, 0,
600 0);
601 vec_add1 (vxm->free_geneve_tunnel_hw_if_indices, t->hw_if_index);
602
603 vxm->tunnel_index_by_sw_if_index[t->sw_if_index] = ~0;
604
605 if (!is_ip6)
606 hash_unset (vxm->geneve4_tunnel_by_key, key4.as_u64);
607 else
John Loe6bfeab2018-01-04 16:39:42 -0500608 hash_unset_mem_free (&vxm->geneve6_tunnel_by_key, &key6);
Marco Varleseb598f1d2017-09-19 14:25:28 +0200609
610 if (!ip46_address_is_multicast (&t->remote))
611 {
612 vtep_addr_unref (&t->local);
613 fib_entry_child_remove (t->fib_entry_index, t->sibling_index);
614 fib_table_entry_delete_index (t->fib_entry_index, FIB_SOURCE_RR);
615 }
616 else if (vtep_addr_unref (&t->remote) == 0)
617 {
618 mcast_shared_remove (&t->remote);
619 }
620
621 fib_node_deinit (&t->node);
622 vec_free (t->rewrite);
623 pool_put (vxm->tunnels, t);
624 }
625
626 if (sw_if_indexp)
627 *sw_if_indexp = sw_if_index;
628
629 return 0;
630}
631
632static uword
633get_decap_next_for_node (u32 node_index, u32 ipv4_set)
634{
635 geneve_main_t *vxm = &geneve_main;
636 vlib_main_t *vm = vxm->vlib_main;
637 uword input_node = (ipv4_set) ? geneve4_input_node.index :
638 geneve6_input_node.index;
639
640 return vlib_node_add_next (vm, input_node, node_index);
641}
642
643static uword
644unformat_decap_next (unformat_input_t * input, va_list * args)
645{
646 u32 *result = va_arg (*args, u32 *);
647 u32 ipv4_set = va_arg (*args, int);
648 geneve_main_t *vxm = &geneve_main;
649 vlib_main_t *vm = vxm->vlib_main;
650 u32 node_index;
651 u32 tmp;
652
653 if (unformat (input, "l2"))
654 *result = GENEVE_INPUT_NEXT_L2_INPUT;
655 else if (unformat (input, "node %U", unformat_vlib_node, vm, &node_index))
656 *result = get_decap_next_for_node (node_index, ipv4_set);
657 else if (unformat (input, "%d", &tmp))
658 *result = tmp;
659 else
660 return 0;
661 return 1;
662}
663
664static clib_error_t *
665geneve_add_del_tunnel_command_fn (vlib_main_t * vm,
666 unformat_input_t * input,
667 vlib_cli_command_t * cmd)
668{
669 unformat_input_t _line_input, *line_input = &_line_input;
670 ip46_address_t local, remote;
671 u8 is_add = 1;
672 u8 local_set = 0;
673 u8 remote_set = 0;
674 u8 grp_set = 0;
675 u8 ipv4_set = 0;
676 u8 ipv6_set = 0;
677 u32 encap_fib_index = 0;
678 u32 mcast_sw_if_index = ~0;
679 u32 decap_next_index = GENEVE_INPUT_NEXT_L2_INPUT;
680 u32 vni = 0;
681 u32 tmp;
682 int rv;
683 vnet_geneve_add_del_tunnel_args_t _a, *a = &_a;
684 u32 tunnel_sw_if_index;
685 clib_error_t *error = NULL;
686
687 /* Cant "universally zero init" (={0}) due to GCC bug 53119 */
688 memset (&local, 0, sizeof local);
689 memset (&remote, 0, sizeof remote);
690
691 /* Get a line of input. */
692 if (!unformat_user (input, unformat_line_input, line_input))
693 return 0;
694
695 while (unformat_check_input (line_input) != UNFORMAT_END_OF_INPUT)
696 {
697 if (unformat (line_input, "del"))
698 {
699 is_add = 0;
700 }
701 else if (unformat (line_input, "local %U",
702 unformat_ip4_address, &local.ip4))
703 {
704 local_set = 1;
705 ipv4_set = 1;
706 }
707 else if (unformat (line_input, "remote %U",
708 unformat_ip4_address, &remote.ip4))
709 {
710 remote_set = 1;
711 ipv4_set = 1;
712 }
713 else if (unformat (line_input, "local %U",
714 unformat_ip6_address, &local.ip6))
715 {
716 local_set = 1;
717 ipv6_set = 1;
718 }
719 else if (unformat (line_input, "remote %U",
720 unformat_ip6_address, &remote.ip6))
721 {
722 remote_set = 1;
723 ipv6_set = 1;
724 }
725 else if (unformat (line_input, "group %U %U",
726 unformat_ip4_address, &remote.ip4,
727 unformat_vnet_sw_interface,
728 vnet_get_main (), &mcast_sw_if_index))
729 {
730 grp_set = remote_set = 1;
731 ipv4_set = 1;
732 }
733 else if (unformat (line_input, "group %U %U",
734 unformat_ip6_address, &remote.ip6,
735 unformat_vnet_sw_interface,
736 vnet_get_main (), &mcast_sw_if_index))
737 {
738 grp_set = remote_set = 1;
739 ipv6_set = 1;
740 }
741 else if (unformat (line_input, "encap-vrf-id %d", &tmp))
742 {
743 encap_fib_index = fib_table_find (fib_ip_proto (ipv6_set), tmp);
744 if (encap_fib_index == ~0)
745 {
746 error =
747 clib_error_return (0, "nonexistent encap-vrf-id %d", tmp);
748 goto done;
749 }
750 }
751 else if (unformat (line_input, "decap-next %U", unformat_decap_next,
752 &decap_next_index, ipv4_set))
753 ;
754 else if (unformat (line_input, "vni %d", &vni))
755 {
756 if (vni >> 24)
757 {
758 error = clib_error_return (0, "vni %d out of range", vni);
759 goto done;
760 }
761 }
762 else
763 {
764 error = clib_error_return (0, "parse error: '%U'",
765 format_unformat_error, line_input);
766 goto done;
767 }
768 }
769
770 if (local_set == 0)
771 {
772 error = clib_error_return (0, "tunnel local address not specified");
773 goto done;
774 }
775
776 if (remote_set == 0)
777 {
778 error = clib_error_return (0, "tunnel remote address not specified");
779 goto done;
780 }
781
782 if (grp_set && !ip46_address_is_multicast (&remote))
783 {
784 error = clib_error_return (0, "tunnel group address not multicast");
785 goto done;
786 }
787
788 if (grp_set == 0 && ip46_address_is_multicast (&remote))
789 {
790 error = clib_error_return (0, "remote address must be unicast");
791 goto done;
792 }
793
794 if (grp_set && mcast_sw_if_index == ~0)
795 {
796 error = clib_error_return (0, "tunnel nonexistent multicast device");
797 goto done;
798 }
799
800 if (ipv4_set && ipv6_set)
801 {
802 error = clib_error_return (0, "both IPv4 and IPv6 addresses specified");
803 goto done;
804 }
805
806 if (ip46_address_cmp (&local, &remote) == 0)
807 {
808 error =
809 clib_error_return (0, "local and remote addresses are identical");
810 goto done;
811 }
812
813 if (decap_next_index == ~0)
814 {
815 error = clib_error_return (0, "next node not found");
816 goto done;
817 }
818
819 if (vni == 0)
820 {
821 error = clib_error_return (0, "vni not specified");
822 goto done;
823 }
824
825 memset (a, 0, sizeof (*a));
826
827 a->is_add = is_add;
828 a->is_ip6 = ipv6_set;
829
830#define _(x) a->x = x;
831 foreach_copy_field;
832#undef _
833
834 rv = vnet_geneve_add_del_tunnel (a, &tunnel_sw_if_index);
835
836 switch (rv)
837 {
838 case 0:
839 if (is_add)
840 vlib_cli_output (vm, "%U\n", format_vnet_sw_if_index_name,
841 vnet_get_main (), tunnel_sw_if_index);
842 break;
843
844 case VNET_API_ERROR_TUNNEL_EXIST:
845 error = clib_error_return (0, "tunnel already exists...");
846 goto done;
847
848 case VNET_API_ERROR_NO_SUCH_ENTRY:
849 error = clib_error_return (0, "tunnel does not exist...");
850 goto done;
851
852 default:
853 error = clib_error_return
854 (0, "vnet_geneve_add_del_tunnel returned %d", rv);
855 goto done;
856 }
857
858done:
859 unformat_free (line_input);
860
861 return error;
862}
863
864/*?
865 * Add or delete a GENEVE Tunnel.
866 *
867 * GENEVE provides the features needed to allow L2 bridge domains (BDs)
868 * to span multiple servers. This is done by building an L2 overlay on
869 * top of an L3 network underlay using GENEVE tunnels.
870 *
871 * This makes it possible for servers to be co-located in the same data
872 * center or be separated geographically as long as they are reachable
873 * through the underlay L3 network.
874 *
875 * You can refer to this kind of L2 overlay bridge domain as a GENEVE
876 * segment.
877 *
878 * @cliexpar
879 * Example of how to create a GENEVE Tunnel:
880 * @cliexcmd{create geneve tunnel local 10.0.3.1 remote 10.0.3.3 vni 13 encap-vrf-id 7}
881 * Example of how to delete a GENEVE Tunnel:
882 * @cliexcmd{create geneve tunnel local 10.0.3.1 remote 10.0.3.3 vni 13 del}
883 ?*/
884/* *INDENT-OFF* */
885VLIB_CLI_COMMAND (create_geneve_tunnel_command, static) = {
886 .path = "create geneve tunnel",
887 .short_help =
888 "create geneve tunnel local <local-vtep-addr>"
889 " {remote <remote-vtep-addr>|group <mcast-vtep-addr> <intf-name>} vni <nn>"
890 " [encap-vrf-id <nn>] [decap-next [l2|node <name>]] [del]",
891 .function = geneve_add_del_tunnel_command_fn,
892};
893/* *INDENT-ON* */
894
895static clib_error_t *
896show_geneve_tunnel_command_fn (vlib_main_t * vm,
897 unformat_input_t * input,
898 vlib_cli_command_t * cmd)
899{
900 geneve_main_t *vxm = &geneve_main;
901 geneve_tunnel_t *t;
902
903 if (pool_elts (vxm->tunnels) == 0)
904 vlib_cli_output (vm, "No geneve tunnels configured...");
905
906 pool_foreach (t, vxm->tunnels, (
907 {
908 vlib_cli_output (vm, "%U",
909 format_geneve_tunnel, t);
910 }
911 ));
912
913 return 0;
914}
915
916/*?
917 * Display all the GENEVE Tunnel entries.
918 *
919 * @cliexpar
920 * Example of how to display the GENEVE Tunnel entries:
921 * @cliexstart{show geneve tunnel}
922 * [0] local 10.0.3.1 remote 10.0.3.3 vni 13 encap_fib_index 0 sw_if_index 5 decap_next l2
923 * @cliexend
924 ?*/
925/* *INDENT-OFF* */
926VLIB_CLI_COMMAND (show_geneve_tunnel_command, static) = {
927 .path = "show geneve tunnel",
928 .short_help = "show geneve tunnel",
929 .function = show_geneve_tunnel_command_fn,
930};
931/* *INDENT-ON* */
932
933
934void
935vnet_int_geneve_bypass_mode (u32 sw_if_index, u8 is_ip6, u8 is_enable)
936{
937 if (is_ip6)
938 vnet_feature_enable_disable ("ip6-unicast", "ip6-geneve-bypass",
939 sw_if_index, is_enable, 0, 0);
940 else
941 vnet_feature_enable_disable ("ip4-unicast", "ip4-geneve-bypass",
942 sw_if_index, is_enable, 0, 0);
943}
944
945
946static clib_error_t *
947set_ip_geneve_bypass (u32 is_ip6,
948 unformat_input_t * input, vlib_cli_command_t * cmd)
949{
950 unformat_input_t _line_input, *line_input = &_line_input;
951 vnet_main_t *vnm = vnet_get_main ();
952 clib_error_t *error = 0;
953 u32 sw_if_index, is_enable;
954
955 sw_if_index = ~0;
956 is_enable = 1;
957
958 if (!unformat_user (input, unformat_line_input, line_input))
959 return 0;
960
961 while (unformat_check_input (line_input) != UNFORMAT_END_OF_INPUT)
962 {
963 if (unformat_user
964 (line_input, unformat_vnet_sw_interface, vnm, &sw_if_index))
965 ;
966 else if (unformat (line_input, "del"))
967 is_enable = 0;
968 else
969 {
970 error = unformat_parse_error (line_input);
971 goto done;
972 }
973 }
974
975 if (~0 == sw_if_index)
976 {
977 error = clib_error_return (0, "unknown interface `%U'",
978 format_unformat_error, line_input);
979 goto done;
980 }
981
982 vnet_int_geneve_bypass_mode (sw_if_index, is_ip6, is_enable);
983
984done:
985 unformat_free (line_input);
986
987 return error;
988}
989
990static clib_error_t *
991set_ip4_geneve_bypass (vlib_main_t * vm,
992 unformat_input_t * input, vlib_cli_command_t * cmd)
993{
994 return set_ip_geneve_bypass (0, input, cmd);
995}
996
997/*?
998 * This command adds the 'ip4-geneve-bypass' graph node for a given interface.
999 * By adding the IPv4 geneve-bypass graph node to an interface, the node checks
1000 * for and validate input geneve packet and bypass ip4-lookup, ip4-local,
1001 * ip4-udp-lookup nodes to speedup geneve packet forwarding. This node will
1002 * cause extra overhead to for non-geneve packets which is kept at a minimum.
1003 *
1004 * @cliexpar
1005 * @parblock
1006 * Example of graph node before ip4-geneve-bypass is enabled:
1007 * @cliexstart{show vlib graph ip4-geneve-bypass}
1008 * Name Next Previous
1009 * ip4-geneve-bypass error-drop [0]
1010 * geneve4-input [1]
1011 * ip4-lookup [2]
1012 * @cliexend
1013 *
1014 * Example of how to enable ip4-geneve-bypass on an interface:
1015 * @cliexcmd{set interface ip geneve-bypass GigabitEthernet2/0/0}
1016 *
1017 * Example of graph node after ip4-geneve-bypass is enabled:
1018 * @cliexstart{show vlib graph ip4-geneve-bypass}
1019 * Name Next Previous
1020 * ip4-geneve-bypass error-drop [0] ip4-input
1021 * geneve4-input [1] ip4-input-no-checksum
1022 * ip4-lookup [2]
1023 * @cliexend
1024 *
1025 * Example of how to display the feature enabed on an interface:
1026 * @cliexstart{show ip interface features GigabitEthernet2/0/0}
1027 * IP feature paths configured on GigabitEthernet2/0/0...
1028 * ...
1029 * ipv4 unicast:
1030 * ip4-geneve-bypass
1031 * ip4-lookup
1032 * ...
1033 * @cliexend
1034 *
1035 * Example of how to disable ip4-geneve-bypass on an interface:
1036 * @cliexcmd{set interface ip geneve-bypass GigabitEthernet2/0/0 del}
1037 * @endparblock
1038?*/
1039/* *INDENT-OFF* */
1040VLIB_CLI_COMMAND (set_interface_ip_geneve_bypass_command, static) = {
1041 .path = "set interface ip geneve-bypass",
1042 .function = set_ip4_geneve_bypass,
1043 .short_help = "set interface ip geneve-bypass <interface> [del]",
1044};
1045/* *INDENT-ON* */
1046
1047static clib_error_t *
1048set_ip6_geneve_bypass (vlib_main_t * vm,
1049 unformat_input_t * input, vlib_cli_command_t * cmd)
1050{
1051 return set_ip_geneve_bypass (1, input, cmd);
1052}
1053
1054/*?
1055 * This command adds the 'ip6-geneve-bypass' graph node for a given interface.
1056 * By adding the IPv6 geneve-bypass graph node to an interface, the node checks
1057 * for and validate input geneve packet and bypass ip6-lookup, ip6-local,
1058 * ip6-udp-lookup nodes to speedup geneve packet forwarding. This node will
1059 * cause extra overhead to for non-geneve packets which is kept at a minimum.
1060 *
1061 * @cliexpar
1062 * @parblock
1063 * Example of graph node before ip6-geneve-bypass is enabled:
1064 * @cliexstart{show vlib graph ip6-geneve-bypass}
1065 * Name Next Previous
1066 * ip6-geneve-bypass error-drop [0]
1067 * geneve6-input [1]
1068 * ip6-lookup [2]
1069 * @cliexend
1070 *
1071 * Example of how to enable ip6-geneve-bypass on an interface:
1072 * @cliexcmd{set interface ip6 geneve-bypass GigabitEthernet2/0/0}
1073 *
1074 * Example of graph node after ip6-geneve-bypass is enabled:
1075 * @cliexstart{show vlib graph ip6-geneve-bypass}
1076 * Name Next Previous
1077 * ip6-geneve-bypass error-drop [0] ip6-input
1078 * geneve6-input [1] ip4-input-no-checksum
1079 * ip6-lookup [2]
1080 * @cliexend
1081 *
1082 * Example of how to display the feature enabed on an interface:
1083 * @cliexstart{show ip interface features GigabitEthernet2/0/0}
1084 * IP feature paths configured on GigabitEthernet2/0/0...
1085 * ...
1086 * ipv6 unicast:
1087 * ip6-geneve-bypass
1088 * ip6-lookup
1089 * ...
1090 * @cliexend
1091 *
1092 * Example of how to disable ip6-geneve-bypass on an interface:
1093 * @cliexcmd{set interface ip6 geneve-bypass GigabitEthernet2/0/0 del}
1094 * @endparblock
1095?*/
1096/* *INDENT-OFF* */
1097VLIB_CLI_COMMAND (set_interface_ip6_geneve_bypass_command, static) = {
1098 .path = "set interface ip6 geneve-bypass",
1099 .function = set_ip6_geneve_bypass,
1100 .short_help = "set interface ip geneve-bypass <interface> [del]",
1101};
1102/* *INDENT-ON* */
1103
1104clib_error_t *
1105geneve_init (vlib_main_t * vm)
1106{
1107 geneve_main_t *vxm = &geneve_main;
1108
1109 vxm->vnet_main = vnet_get_main ();
1110 vxm->vlib_main = vm;
1111
1112 /* initialize the ip6 hash */
1113 vxm->geneve6_tunnel_by_key = hash_create_mem (0,
1114 sizeof (geneve6_tunnel_key_t),
1115 sizeof (uword));
1116 vxm->vtep6 = hash_create_mem (0, sizeof (ip6_address_t), sizeof (uword));
1117 vxm->mcast_shared = hash_create_mem (0,
1118 sizeof (ip46_address_t),
1119 sizeof (mcast_shared_t));
1120
1121 udp_register_dst_port (vm, UDP_DST_PORT_geneve,
1122 geneve4_input_node.index, /* is_ip4 */ 1);
1123 udp_register_dst_port (vm, UDP_DST_PORT_geneve6,
1124 geneve6_input_node.index, /* is_ip4 */ 0);
1125
1126 fib_node_register_type (FIB_NODE_TYPE_GENEVE_TUNNEL, &geneve_vft);
1127
1128 return 0;
1129}
1130
1131VLIB_INIT_FUNCTION (geneve_init);
1132
1133/*
1134 * fd.io coding-style-patch-verification: ON
1135 *
1136 * Local Variables:
1137 * eval: (c-set-style "gnu")
1138 * End:
1139 */