Line data Source code
1 : /*
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 <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/fib/fib_entry_track.h>
20 : #include <vnet/mfib/mfib_table.h>
21 : #include <vnet/adj/adj_mcast.h>
22 : #include <vnet/interface.h>
23 : #include <vlib/vlib.h>
24 :
25 : /**
26 : * @file
27 : * @brief GENEVE.
28 : *
29 : * GENEVE provides the features needed to allow L2 bridge domains (BDs)
30 : * to span multiple servers. This is done by building an L2 overlay on
31 : * top of an L3 network underlay using GENEVE tunnels.
32 : *
33 : * This makes it possible for servers to be co-located in the same data
34 : * center or be separated geographically as long as they are reachable
35 : * through the underlay L3 network.
36 : */
37 :
38 :
39 : geneve_main_t geneve_main;
40 :
41 : u8 *
42 14 : format_geneve_encap_trace (u8 * s, va_list * args)
43 : {
44 14 : CLIB_UNUSED (vlib_main_t * vm) = va_arg (*args, vlib_main_t *);
45 14 : CLIB_UNUSED (vlib_node_t * node) = va_arg (*args, vlib_node_t *);
46 14 : geneve_encap_trace_t *t = va_arg (*args, geneve_encap_trace_t *);
47 :
48 14 : s = format (s, "GENEVE encap to geneve_tunnel%d vni %d",
49 : t->tunnel_index, t->vni);
50 14 : return s;
51 : }
52 :
53 : static u8 *
54 110 : format_decap_next (u8 * s, va_list * args)
55 : {
56 110 : u32 next_index = va_arg (*args, u32);
57 :
58 110 : switch (next_index)
59 : {
60 0 : case GENEVE_INPUT_NEXT_DROP:
61 0 : return format (s, "drop");
62 110 : case GENEVE_INPUT_NEXT_L2_INPUT:
63 110 : return format (s, "l2");
64 0 : default:
65 0 : return format (s, "index %d", next_index);
66 : }
67 : return s;
68 : }
69 :
70 : u8 *
71 110 : format_geneve_tunnel (u8 * s, va_list * args)
72 : {
73 110 : geneve_tunnel_t *t = va_arg (*args, geneve_tunnel_t *);
74 110 : geneve_main_t *ngm = &geneve_main;
75 :
76 110 : s = format (s, "[%d] lcl %U rmt %U vni %d fib-idx %d sw-if-idx %d ",
77 110 : t - ngm->tunnels,
78 : format_ip46_address, &t->local, IP46_TYPE_ANY,
79 : format_ip46_address, &t->remote, IP46_TYPE_ANY,
80 : t->vni, t->encap_fib_index, t->sw_if_index);
81 :
82 110 : s = format (s, "encap-dpo-idx %d ", t->next_dpo.dpoi_index);
83 110 : s = format (s, "decap-next-%U ", format_decap_next, t->decap_next_index);
84 110 : s = format (s, "l3-mode %u ", t->l3_mode);
85 :
86 110 : if (PREDICT_FALSE (ip46_address_is_multicast (&t->remote)))
87 5 : s = format (s, "mcast-sw-if-idx %d ", t->mcast_sw_if_index);
88 :
89 110 : return s;
90 : }
91 :
92 : static u8 *
93 153 : format_geneve_name (u8 * s, va_list * args)
94 : {
95 153 : u32 dev_instance = va_arg (*args, u32);
96 153 : return format (s, "geneve_tunnel%d", dev_instance);
97 : }
98 :
99 : static clib_error_t *
100 64 : geneve_interface_admin_up_down (vnet_main_t * vnm, u32 hw_if_index, u32 flags)
101 : {
102 64 : u32 hw_flags = (flags & VNET_SW_INTERFACE_FLAG_ADMIN_UP) ?
103 : VNET_HW_INTERFACE_FLAG_LINK_UP : 0;
104 64 : vnet_hw_interface_set_flags (vnm, hw_if_index, hw_flags);
105 :
106 64 : return /* no error */ 0;
107 : }
108 :
109 : static clib_error_t *
110 0 : geneve_mac_change (vnet_hw_interface_t * hi,
111 : const u8 * old_address, const u8 * mac_address)
112 : {
113 0 : l2input_interface_mac_change (hi->sw_if_index, old_address, mac_address);
114 0 : return (NULL);
115 : }
116 :
117 : /* *INDENT-OFF* */
118 9791 : VNET_DEVICE_CLASS (geneve_device_class, static) = {
119 : .name = "GENEVE",
120 : .format_device_name = format_geneve_name,
121 : .format_tx_trace = format_geneve_encap_trace,
122 : .admin_up_down_function = geneve_interface_admin_up_down,
123 : .mac_addr_change_function = geneve_mac_change,
124 : };
125 : /* *INDENT-ON* */
126 :
127 : static u8 *
128 0 : format_geneve_header_with_length (u8 * s, va_list * args)
129 : {
130 0 : u32 dev_instance = va_arg (*args, u32);
131 0 : s = format (s, "unimplemented dev %u", dev_instance);
132 0 : return s;
133 : }
134 :
135 : /* *INDENT-OFF* */
136 6911 : VNET_HW_INTERFACE_CLASS (geneve_hw_class) = {
137 : .name = "GENEVE",
138 : .format_header = format_geneve_header_with_length,
139 : .build_rewrite = default_build_rewrite,
140 : };
141 : /* *INDENT-ON* */
142 :
143 : static void
144 32 : geneve_tunnel_restack_dpo (geneve_tunnel_t * t)
145 : {
146 32 : dpo_id_t dpo = DPO_INVALID;
147 32 : u32 encap_index = ip46_address_is_ip4 (&t->remote) ?
148 32 : geneve4_encap_node.index : geneve6_encap_node.index;
149 32 : fib_forward_chain_type_t forw_type = ip46_address_is_ip4 (&t->remote) ?
150 32 : FIB_FORW_CHAIN_TYPE_UNICAST_IP4 : FIB_FORW_CHAIN_TYPE_UNICAST_IP6;
151 :
152 32 : fib_entry_contribute_forwarding (t->fib_entry_index, forw_type, &dpo);
153 :
154 : /* geneve uses the flow hash as the udp source port
155 : * hence the packet's hash is unknown at this time.
156 : * However, we can still skip single bucket load balance dpo's */
157 94 : while (DPO_LOAD_BALANCE == dpo.dpoi_type)
158 : {
159 62 : load_balance_t *lb = load_balance_get (dpo.dpoi_index);
160 62 : if (lb->lb_n_buckets > 1)
161 0 : break;
162 :
163 62 : dpo_copy (&dpo, load_balance_get_bucket_i (lb, 0));
164 : }
165 32 : dpo_stack_from_node (encap_index, &t->next_dpo, &dpo);
166 32 : dpo_reset (&dpo);
167 32 : }
168 :
169 : static geneve_tunnel_t *
170 10 : geneve_tunnel_from_fib_node (fib_node_t * node)
171 : {
172 10 : ASSERT (FIB_NODE_TYPE_GENEVE_TUNNEL == node->fn_type);
173 10 : return ((geneve_tunnel_t *) (((char *) node) -
174 : STRUCT_OFFSET_OF (geneve_tunnel_t, node)));
175 : }
176 :
177 : /**
178 : * Function definition to backwalk a FIB node -
179 : * Here we will restack the new dpo of GENEVE DIP to encap node.
180 : */
181 : static fib_node_back_walk_rc_t
182 10 : geneve_tunnel_back_walk (fib_node_t * node, fib_node_back_walk_ctx_t * ctx)
183 : {
184 10 : geneve_tunnel_restack_dpo (geneve_tunnel_from_fib_node (node));
185 10 : return (FIB_NODE_BACK_WALK_CONTINUE);
186 : }
187 :
188 : /**
189 : * Function definition to get a FIB node from its index
190 : */
191 : static fib_node_t *
192 10 : geneve_tunnel_fib_node_get (fib_node_index_t index)
193 : {
194 : geneve_tunnel_t *t;
195 10 : geneve_main_t *vxm = &geneve_main;
196 :
197 10 : t = pool_elt_at_index (vxm->tunnels, index);
198 :
199 10 : return (&t->node);
200 : }
201 :
202 : /**
203 : * Function definition to inform the FIB node that its last lock has gone.
204 : */
205 : static void
206 0 : geneve_tunnel_last_lock_gone (fib_node_t * node)
207 : {
208 : /*
209 : * The GENEVE tunnel is a root of the graph. As such
210 : * it never has children and thus is never locked.
211 : */
212 0 : ASSERT (0);
213 0 : }
214 :
215 : /*
216 : * Virtual function table registered by GENEVE tunnels
217 : * for participation in the FIB object graph.
218 : */
219 : const static fib_node_vft_t geneve_vft = {
220 : .fnv_get = geneve_tunnel_fib_node_get,
221 : .fnv_last_lock = geneve_tunnel_last_lock_gone,
222 : .fnv_back_walk = geneve_tunnel_back_walk,
223 : };
224 :
225 :
226 : #define foreach_copy_field \
227 : _(vni) \
228 : _(mcast_sw_if_index) \
229 : _(encap_fib_index) \
230 : _(decap_next_index) \
231 : _(local) \
232 : _(remote) \
233 : _(l3_mode)
234 :
235 : static int
236 43 : geneve_rewrite (geneve_tunnel_t * t, bool is_ip6)
237 : {
238 : union
239 : {
240 : ip4_geneve_header_t *h4;
241 : ip6_geneve_header_t *h6;
242 : u8 *rw;
243 43 : } r =
244 : {
245 : .rw = 0};
246 43 : int len = is_ip6 ? sizeof *r.h6 : sizeof *r.h4;
247 : #if SUPPORT_OPTIONS_HEADER==1
248 : len += t->options_len;
249 : #endif
250 :
251 43 : vec_validate_aligned (r.rw, len - 1, CLIB_CACHE_LINE_BYTES);
252 :
253 : udp_header_t *udp;
254 : geneve_header_t *geneve;
255 : /* Fixed portion of the (outer) ip header */
256 43 : if (!is_ip6)
257 : {
258 43 : ip4_header_t *ip = &r.h4->ip4;
259 43 : udp = &r.h4->udp, geneve = &r.h4->geneve;
260 43 : ip->ip_version_and_header_length = 0x45;
261 43 : ip->ttl = 254;
262 43 : ip->protocol = IP_PROTOCOL_UDP;
263 :
264 43 : ip->src_address = t->local.ip4;
265 43 : ip->dst_address = t->remote.ip4;
266 :
267 : /* we fix up the ip4 header length and checksum after-the-fact */
268 43 : ip->checksum = ip4_header_checksum (ip);
269 : }
270 : else
271 : {
272 0 : ip6_header_t *ip = &r.h6->ip6;
273 0 : udp = &r.h6->udp, geneve = &r.h6->geneve;
274 0 : ip->ip_version_traffic_class_and_flow_label =
275 0 : clib_host_to_net_u32 (6 << 28);
276 0 : ip->hop_limit = 255;
277 0 : ip->protocol = IP_PROTOCOL_UDP;
278 :
279 0 : ip->src_address = t->local.ip6;
280 0 : ip->dst_address = t->remote.ip6;
281 : }
282 :
283 : /* UDP header, randomize local port on something, maybe? */
284 43 : udp->src_port = clib_host_to_net_u16 (5251);
285 43 : udp->dst_port = clib_host_to_net_u16 (UDP_DST_PORT_geneve);
286 :
287 : /* GENEVE header */
288 43 : vnet_set_geneve_version (geneve, GENEVE_VERSION);
289 : #if SUPPORT_OPTIONS_HEADER==1
290 : vnet_set_geneve_options_len (geneve, t->options_len);
291 : #else
292 43 : vnet_set_geneve_options_len (geneve, 0);
293 : #endif
294 43 : vnet_set_geneve_oamframe_bit (geneve, 0);
295 43 : vnet_set_geneve_critical_bit (geneve, 0);
296 43 : vnet_set_geneve_protocol (geneve, GENEVE_ETH_PROTOCOL);
297 :
298 43 : vnet_geneve_hdr_1word_hton (geneve);
299 :
300 43 : vnet_set_geneve_vni (geneve, t->vni);
301 :
302 43 : t->rewrite = r.rw;
303 43 : return (0);
304 : }
305 :
306 : static bool
307 43 : geneve_decap_next_is_valid (geneve_main_t * vxm, u32 is_ip6,
308 : u32 decap_next_index)
309 : {
310 43 : vlib_main_t *vm = vxm->vlib_main;
311 43 : u32 input_idx =
312 43 : (!is_ip6) ? geneve4_input_node.index : geneve6_input_node.index;
313 43 : vlib_node_runtime_t *r = vlib_node_get_runtime (vm, input_idx);
314 :
315 43 : return decap_next_index < r->n_next_nodes;
316 : }
317 :
318 : typedef union
319 : {
320 : struct
321 : {
322 : fib_node_index_t mfib_entry_index;
323 : adj_index_t mcast_adj_index;
324 : };
325 : u64 as_u64;
326 : } __clib_packed mcast_shared_t;
327 :
328 : static inline mcast_shared_t
329 31 : mcast_shared_get (ip46_address_t * ip)
330 : {
331 31 : ASSERT (ip46_address_is_multicast (ip));
332 31 : uword *p = hash_get_mem (geneve_main.mcast_shared, ip);
333 31 : ALWAYS_ASSERT (p);
334 62 : return (mcast_shared_t)
335 : {
336 31 : .as_u64 = *p};
337 : }
338 :
339 : static inline void
340 11 : mcast_shared_add (ip46_address_t * remote,
341 : fib_node_index_t mfei, adj_index_t ai)
342 : {
343 11 : mcast_shared_t new_ep = {
344 : .mcast_adj_index = ai,
345 : .mfib_entry_index = mfei,
346 : };
347 :
348 11 : hash_set_mem_alloc (&geneve_main.mcast_shared, remote, new_ep.as_u64);
349 11 : }
350 :
351 : static inline void
352 10 : mcast_shared_remove (ip46_address_t * remote)
353 : {
354 10 : mcast_shared_t ep = mcast_shared_get (remote);
355 :
356 10 : adj_unlock (ep.mcast_adj_index);
357 10 : mfib_table_entry_delete_index (ep.mfib_entry_index, MFIB_SOURCE_GENEVE);
358 :
359 10 : hash_unset_mem_free (&geneve_main.mcast_shared, remote);
360 10 : }
361 :
362 64 : int vnet_geneve_add_del_tunnel
363 : (vnet_geneve_add_del_tunnel_args_t * a, u32 * sw_if_indexp)
364 : {
365 64 : geneve_main_t *vxm = &geneve_main;
366 64 : geneve_tunnel_t *t = 0;
367 64 : vnet_main_t *vnm = vxm->vnet_main;
368 : uword *p;
369 64 : u32 hw_if_index = ~0;
370 64 : u32 sw_if_index = ~0;
371 : int rv;
372 : geneve4_tunnel_key_t key4;
373 : geneve6_tunnel_key_t key6;
374 64 : u32 is_ip6 = a->is_ip6;
375 :
376 64 : if (!is_ip6)
377 : {
378 64 : key4.remote = a->remote.ip4.as_u32;
379 64 : key4.vni = clib_host_to_net_u32 (a->vni << GENEVE_VNI_SHIFT);
380 64 : p = hash_get (vxm->geneve4_tunnel_by_key, key4.as_u64);
381 : }
382 : else
383 : {
384 0 : key6.remote = a->remote.ip6;
385 0 : key6.vni = clib_host_to_net_u32 (a->vni << GENEVE_VNI_SHIFT);
386 0 : p = hash_get_mem (vxm->geneve6_tunnel_by_key, &key6);
387 : }
388 :
389 64 : if (a->is_add)
390 : {
391 43 : l2input_main_t *l2im = &l2input_main;
392 :
393 : /* adding a tunnel: tunnel must not already exist */
394 43 : if (p)
395 0 : return VNET_API_ERROR_TUNNEL_EXIST;
396 :
397 : /*if not set explicitly, default to l2 */
398 43 : if (a->decap_next_index == ~0)
399 42 : a->decap_next_index = GENEVE_INPUT_NEXT_L2_INPUT;
400 43 : if (!geneve_decap_next_is_valid (vxm, is_ip6, a->decap_next_index))
401 0 : return VNET_API_ERROR_INVALID_DECAP_NEXT;
402 :
403 43 : pool_get_aligned (vxm->tunnels, t, CLIB_CACHE_LINE_BYTES);
404 43 : clib_memset (t, 0, sizeof (*t));
405 :
406 : /* copy from arg structure */
407 : #define _(x) t->x = a->x;
408 43 : foreach_copy_field;
409 : #undef _
410 :
411 43 : rv = geneve_rewrite (t, is_ip6);
412 43 : if (rv)
413 : {
414 0 : pool_put (vxm->tunnels, t);
415 0 : return rv;
416 : }
417 :
418 : /* copy the key */
419 43 : if (is_ip6)
420 0 : hash_set_mem_alloc (&vxm->geneve6_tunnel_by_key, &key6,
421 0 : t - vxm->tunnels);
422 : else
423 43 : hash_set (vxm->geneve4_tunnel_by_key, key4.as_u64, t - vxm->tunnels);
424 :
425 : vnet_hw_interface_t *hi;
426 43 : if (a->l3_mode)
427 : {
428 1 : vnet_eth_interface_registration_t eir = {};
429 1 : u32 t_idx = t - vxm->tunnels;
430 1 : u8 address[6] =
431 1 : { 0xd0, 0x0b, 0xee, 0xd0, (u8) (t_idx >> 8), (u8) t_idx };
432 :
433 1 : eir.dev_class_index = geneve_device_class.index;
434 1 : eir.dev_instance = t_idx;
435 1 : eir.address = address;
436 1 : hw_if_index = vnet_eth_register_interface (vnm, &eir);
437 : }
438 : else
439 : {
440 42 : hw_if_index = vnet_register_interface
441 42 : (vnm, geneve_device_class.index, t - vxm->tunnels,
442 42 : geneve_hw_class.index, t - vxm->tunnels);
443 : }
444 :
445 43 : hi = vnet_get_hw_interface (vnm, hw_if_index);
446 :
447 : /* Set geneve tunnel output node */
448 43 : u32 encap_index = !is_ip6 ?
449 43 : geneve4_encap_node.index : geneve6_encap_node.index;
450 43 : vnet_set_interface_output_node (vnm, hw_if_index, encap_index);
451 :
452 43 : t->hw_if_index = hw_if_index;
453 43 : t->sw_if_index = sw_if_index = hi->sw_if_index;
454 :
455 84 : vec_validate_init_empty (vxm->tunnel_index_by_sw_if_index, sw_if_index,
456 : ~0);
457 43 : vxm->tunnel_index_by_sw_if_index[sw_if_index] = t - vxm->tunnels;
458 :
459 : /* setup l2 input config with l2 feature and bd 0 to drop packet */
460 43 : vec_validate (l2im->configs, sw_if_index);
461 43 : l2im->configs[sw_if_index].feature_bitmap = L2INPUT_FEAT_DROP;
462 43 : l2im->configs[sw_if_index].bd_index = 0;
463 :
464 43 : vnet_sw_interface_t *si = vnet_get_sw_interface (vnm, sw_if_index);
465 43 : si->flags &= ~VNET_SW_INTERFACE_FLAG_HIDDEN;
466 43 : vnet_sw_interface_set_flags (vnm, sw_if_index,
467 : VNET_SW_INTERFACE_FLAG_ADMIN_UP);
468 :
469 43 : fib_node_init (&t->node, FIB_NODE_TYPE_GENEVE_TUNNEL);
470 : fib_prefix_t tun_remote_pfx;
471 43 : vnet_flood_class_t flood_class = VNET_FLOOD_CLASS_TUNNEL_NORMAL;
472 :
473 43 : fib_protocol_t fp = fib_ip_proto (is_ip6);
474 43 : fib_prefix_from_ip46_addr (fp, &t->remote, &tun_remote_pfx);
475 43 : if (!ip46_address_is_multicast (&t->remote))
476 : {
477 : /* Unicast tunnel -
478 : * source the FIB entry for the tunnel's destination
479 : * and become a child thereof. The tunnel will then get poked
480 : * when the forwarding for the entry updates, and the tunnel can
481 : * re-stack accordingly
482 : */
483 22 : vtep_addr_ref (&vxm->vtep_table, t->encap_fib_index, &t->local);
484 44 : t->fib_entry_index = fib_entry_track (t->encap_fib_index,
485 : &tun_remote_pfx,
486 : FIB_NODE_TYPE_GENEVE_TUNNEL,
487 22 : t - vxm->tunnels,
488 22 : &t->sibling_index);
489 22 : geneve_tunnel_restack_dpo (t);
490 : }
491 : else
492 : {
493 : /* Multicast tunnel -
494 : * as the same mcast group can be used for mutiple mcast tunnels
495 : * with different VNIs, create the output fib adjecency only if
496 : * it does not already exist
497 : */
498 21 : if (vtep_addr_ref (&vxm->vtep_table,
499 21 : t->encap_fib_index, &t->remote) == 1)
500 : {
501 : fib_node_index_t mfei;
502 : adj_index_t ai;
503 22 : fib_route_path_t path = {
504 11 : .frp_proto = fib_proto_to_dpo (fp),
505 : .frp_addr = zero_addr,
506 : .frp_sw_if_index = 0xffffffff,
507 : .frp_fib_index = ~0,
508 : .frp_weight = 1,
509 : .frp_flags = FIB_ROUTE_PATH_LOCAL,
510 : .frp_mitf_flags = MFIB_ITF_FLAG_FORWARD,
511 : };
512 11 : const mfib_prefix_t mpfx = {
513 : .fp_proto = fp,
514 : .fp_len = (is_ip6 ? 128 : 32),
515 : .fp_grp_addr = tun_remote_pfx.fp_addr,
516 : };
517 :
518 : /*
519 : * Setup the (*,G) to receive traffic on the mcast group
520 : * - the forwarding interface is for-us
521 : * - the accepting interface is that from the API
522 : */
523 11 : mfib_table_entry_path_update (t->encap_fib_index, &mpfx,
524 : MFIB_SOURCE_GENEVE,
525 : MFIB_ENTRY_FLAG_NONE, &path);
526 :
527 11 : path.frp_sw_if_index = a->mcast_sw_if_index;
528 11 : path.frp_flags = FIB_ROUTE_PATH_FLAG_NONE;
529 11 : path.frp_mitf_flags = MFIB_ITF_FLAG_ACCEPT;
530 11 : mfei = mfib_table_entry_path_update (
531 11 : t->encap_fib_index, &mpfx, MFIB_SOURCE_GENEVE,
532 : MFIB_ENTRY_FLAG_NONE, &path);
533 :
534 : /*
535 : * Create the mcast adjacency to send traffic to the group
536 : */
537 11 : ai = adj_mcast_add_or_lock (fp,
538 11 : fib_proto_to_link (fp),
539 : a->mcast_sw_if_index);
540 :
541 : /*
542 : * create a new end-point
543 : */
544 11 : mcast_shared_add (&t->remote, mfei, ai);
545 : }
546 :
547 21 : dpo_id_t dpo = DPO_INVALID;
548 21 : mcast_shared_t ep = mcast_shared_get (&t->remote);
549 :
550 : /* Stack shared mcast remote mac addr rewrite on encap */
551 21 : dpo_set (&dpo, DPO_ADJACENCY_MCAST,
552 21 : fib_proto_to_dpo (fp), ep.mcast_adj_index);
553 :
554 21 : dpo_stack_from_node (encap_index, &t->next_dpo, &dpo);
555 21 : dpo_reset (&dpo);
556 21 : flood_class = VNET_FLOOD_CLASS_TUNNEL_MASTER;
557 : }
558 :
559 43 : vnet_get_sw_interface (vnet_get_main (), sw_if_index)->flood_class =
560 : flood_class;
561 : }
562 : else
563 : {
564 : /* deleting a tunnel: tunnel must exist */
565 21 : if (!p)
566 0 : return VNET_API_ERROR_NO_SUCH_ENTRY;
567 :
568 21 : t = pool_elt_at_index (vxm->tunnels, p[0]);
569 :
570 21 : sw_if_index = t->sw_if_index;
571 21 : vnet_sw_interface_set_flags (vnm, t->sw_if_index, 0 /* down */ );
572 21 : vnet_sw_interface_t *si = vnet_get_sw_interface (vnm, t->sw_if_index);
573 21 : si->flags |= VNET_SW_INTERFACE_FLAG_HIDDEN;
574 :
575 : /* make sure tunnel is removed from l2 bd or xconnect */
576 21 : set_int_l2_mode (vxm->vlib_main, vnm, MODE_L3, t->sw_if_index, 0,
577 : L2_BD_PORT_TYPE_NORMAL, 0, 0);
578 :
579 21 : if (t->l3_mode)
580 1 : ethernet_delete_interface (vnm, t->hw_if_index);
581 : else
582 20 : vnet_delete_hw_interface (vnm, t->hw_if_index);
583 :
584 21 : vxm->tunnel_index_by_sw_if_index[t->sw_if_index] = ~0;
585 :
586 21 : if (!is_ip6)
587 21 : hash_unset (vxm->geneve4_tunnel_by_key, key4.as_u64);
588 : else
589 0 : hash_unset_mem_free (&vxm->geneve6_tunnel_by_key, &key6);
590 :
591 21 : if (!ip46_address_is_multicast (&t->remote))
592 : {
593 1 : vtep_addr_unref (&vxm->vtep_table, t->encap_fib_index, &t->local);
594 1 : fib_entry_untrack (t->fib_entry_index, t->sibling_index);
595 : }
596 20 : else if (vtep_addr_unref (&vxm->vtep_table,
597 20 : t->encap_fib_index, &t->remote) == 0)
598 : {
599 10 : mcast_shared_remove (&t->remote);
600 : }
601 :
602 21 : fib_node_deinit (&t->node);
603 21 : vec_free (t->rewrite);
604 21 : pool_put (vxm->tunnels, t);
605 : }
606 :
607 64 : if (sw_if_indexp)
608 64 : *sw_if_indexp = sw_if_index;
609 :
610 64 : if (a->is_add)
611 : {
612 : /* register udp ports */
613 43 : if (!is_ip6 && !udp_is_valid_dst_port (UDP_DST_PORT_geneve, 1))
614 2 : udp_register_dst_port (vxm->vlib_main, UDP_DST_PORT_geneve,
615 : geneve4_input_node.index, 1);
616 43 : if (is_ip6 && !udp_is_valid_dst_port (UDP_DST_PORT_geneve6, 0))
617 0 : udp_register_dst_port (vxm->vlib_main, UDP_DST_PORT_geneve6,
618 : geneve6_input_node.index, 0);
619 : }
620 :
621 64 : return 0;
622 : }
623 :
624 : static uword
625 0 : get_decap_next_for_node (u32 node_index, u32 ipv4_set)
626 : {
627 0 : geneve_main_t *vxm = &geneve_main;
628 0 : vlib_main_t *vm = vxm->vlib_main;
629 0 : uword input_node = (ipv4_set) ? geneve4_input_node.index :
630 0 : geneve6_input_node.index;
631 :
632 0 : return vlib_node_add_next (vm, input_node, node_index);
633 : }
634 :
635 : static uword
636 0 : unformat_decap_next (unformat_input_t * input, va_list * args)
637 : {
638 0 : u32 *result = va_arg (*args, u32 *);
639 0 : u32 ipv4_set = va_arg (*args, int);
640 0 : geneve_main_t *vxm = &geneve_main;
641 0 : vlib_main_t *vm = vxm->vlib_main;
642 : u32 node_index;
643 : u32 tmp;
644 :
645 0 : if (unformat (input, "l2"))
646 0 : *result = GENEVE_INPUT_NEXT_L2_INPUT;
647 0 : else if (unformat (input, "node %U", unformat_vlib_node, vm, &node_index))
648 0 : *result = get_decap_next_for_node (node_index, ipv4_set);
649 0 : else if (unformat (input, "%d", &tmp))
650 0 : *result = tmp;
651 : else
652 0 : return 0;
653 0 : return 1;
654 : }
655 :
656 : static clib_error_t *
657 0 : geneve_add_del_tunnel_command_fn (vlib_main_t * vm,
658 : unformat_input_t * input,
659 : vlib_cli_command_t * cmd)
660 : {
661 0 : unformat_input_t _line_input, *line_input = &_line_input;
662 : ip46_address_t local, remote;
663 0 : u8 is_add = 1;
664 0 : u8 local_set = 0;
665 0 : u8 remote_set = 0;
666 0 : u8 grp_set = 0;
667 0 : u8 ipv4_set = 0;
668 0 : u8 ipv6_set = 0;
669 0 : u8 l3_mode = 0;
670 0 : u32 encap_fib_index = 0;
671 0 : u32 mcast_sw_if_index = ~0;
672 0 : u32 decap_next_index = GENEVE_INPUT_NEXT_L2_INPUT;
673 0 : u32 vni = 0;
674 : u32 tmp;
675 : int rv;
676 0 : vnet_geneve_add_del_tunnel_args_t _a, *a = &_a;
677 : u32 tunnel_sw_if_index;
678 0 : clib_error_t *error = NULL;
679 :
680 : /* Cant "universally zero init" (={0}) due to GCC bug 53119 */
681 0 : clib_memset (&local, 0, sizeof local);
682 0 : clib_memset (&remote, 0, sizeof remote);
683 :
684 : /* Get a line of input. */
685 0 : if (!unformat_user (input, unformat_line_input, line_input))
686 0 : return 0;
687 :
688 0 : while (unformat_check_input (line_input) != UNFORMAT_END_OF_INPUT)
689 : {
690 0 : if (unformat (line_input, "del"))
691 : {
692 0 : is_add = 0;
693 : }
694 0 : else if (unformat (line_input, "local %U",
695 : unformat_ip4_address, &local.ip4))
696 : {
697 0 : local_set = 1;
698 0 : ipv4_set = 1;
699 : }
700 0 : else if (unformat (line_input, "remote %U",
701 : unformat_ip4_address, &remote.ip4))
702 : {
703 0 : remote_set = 1;
704 0 : ipv4_set = 1;
705 : }
706 0 : else if (unformat (line_input, "local %U",
707 : unformat_ip6_address, &local.ip6))
708 : {
709 0 : local_set = 1;
710 0 : ipv6_set = 1;
711 : }
712 0 : else if (unformat (line_input, "remote %U",
713 : unformat_ip6_address, &remote.ip6))
714 : {
715 0 : remote_set = 1;
716 0 : ipv6_set = 1;
717 : }
718 0 : else if (unformat (line_input, "group %U %U",
719 : unformat_ip4_address, &remote.ip4,
720 : unformat_vnet_sw_interface,
721 : vnet_get_main (), &mcast_sw_if_index))
722 : {
723 0 : grp_set = remote_set = 1;
724 0 : ipv4_set = 1;
725 : }
726 0 : else if (unformat (line_input, "group %U %U",
727 : unformat_ip6_address, &remote.ip6,
728 : unformat_vnet_sw_interface,
729 : vnet_get_main (), &mcast_sw_if_index))
730 : {
731 0 : grp_set = remote_set = 1;
732 0 : ipv6_set = 1;
733 : }
734 0 : else if (unformat (line_input, "encap-vrf-id %d", &tmp))
735 : {
736 0 : encap_fib_index = fib_table_find (fib_ip_proto (ipv6_set), tmp);
737 0 : if (encap_fib_index == ~0)
738 : {
739 : error =
740 0 : clib_error_return (0, "nonexistent encap-vrf-id %d", tmp);
741 0 : goto done;
742 : }
743 : }
744 0 : else if (unformat (line_input, "decap-next %U", unformat_decap_next,
745 : &decap_next_index, ipv4_set))
746 : ;
747 0 : else if (unformat (line_input, "vni %d", &vni))
748 : {
749 0 : if (vni >> 24)
750 : {
751 0 : error = clib_error_return (0, "vni %d out of range", vni);
752 0 : goto done;
753 : }
754 : }
755 0 : else if (unformat (line_input, "l3-mode"))
756 : {
757 0 : l3_mode = 1;
758 : }
759 : else
760 : {
761 0 : error = clib_error_return (0, "parse error: '%U'",
762 : format_unformat_error, line_input);
763 0 : goto done;
764 : }
765 : }
766 :
767 0 : if (local_set == 0)
768 : {
769 0 : error = clib_error_return (0, "tunnel local address not specified");
770 0 : goto done;
771 : }
772 :
773 0 : if (remote_set == 0)
774 : {
775 0 : error = clib_error_return (0, "tunnel remote address not specified");
776 0 : goto done;
777 : }
778 :
779 0 : if (grp_set && !ip46_address_is_multicast (&remote))
780 : {
781 0 : error = clib_error_return (0, "tunnel group address not multicast");
782 0 : goto done;
783 : }
784 :
785 0 : if (grp_set == 0 && ip46_address_is_multicast (&remote))
786 : {
787 0 : error = clib_error_return (0, "remote address must be unicast");
788 0 : goto done;
789 : }
790 :
791 0 : if (grp_set && mcast_sw_if_index == ~0)
792 : {
793 0 : error = clib_error_return (0, "tunnel nonexistent multicast device");
794 0 : goto done;
795 : }
796 :
797 0 : if (ipv4_set && ipv6_set)
798 : {
799 0 : error = clib_error_return (0, "both IPv4 and IPv6 addresses specified");
800 0 : goto done;
801 : }
802 :
803 0 : if (ip46_address_cmp (&local, &remote) == 0)
804 : {
805 : error =
806 0 : clib_error_return (0, "local and remote addresses are identical");
807 0 : goto done;
808 : }
809 :
810 0 : if (decap_next_index == ~0)
811 : {
812 0 : error = clib_error_return (0, "next node not found");
813 0 : goto done;
814 : }
815 :
816 0 : if (vni == 0)
817 : {
818 0 : error = clib_error_return (0, "vni not specified");
819 0 : goto done;
820 : }
821 :
822 0 : clib_memset (a, 0, sizeof (*a));
823 :
824 0 : a->is_add = is_add;
825 0 : a->is_ip6 = ipv6_set;
826 :
827 : #define _(x) a->x = x;
828 0 : foreach_copy_field;
829 : #undef _
830 :
831 0 : rv = vnet_geneve_add_del_tunnel (a, &tunnel_sw_if_index);
832 :
833 0 : switch (rv)
834 : {
835 0 : case 0:
836 0 : if (is_add)
837 0 : vlib_cli_output (vm, "%U\n", format_vnet_sw_if_index_name,
838 : vnet_get_main (), tunnel_sw_if_index);
839 0 : break;
840 :
841 0 : case VNET_API_ERROR_TUNNEL_EXIST:
842 0 : error = clib_error_return (0, "tunnel already exists...");
843 0 : goto done;
844 :
845 0 : case VNET_API_ERROR_NO_SUCH_ENTRY:
846 0 : error = clib_error_return (0, "tunnel does not exist...");
847 0 : goto done;
848 :
849 0 : default:
850 0 : error = clib_error_return
851 : (0, "vnet_geneve_add_del_tunnel returned %d", rv);
852 0 : goto done;
853 : }
854 :
855 0 : done:
856 0 : unformat_free (line_input);
857 :
858 0 : return error;
859 : }
860 :
861 : /*?
862 : * Add or delete a GENEVE Tunnel.
863 : *
864 : * GENEVE provides the features needed to allow L2 bridge domains (BDs)
865 : * to span multiple servers. This is done by building an L2 overlay on
866 : * top of an L3 network underlay using GENEVE tunnels.
867 : *
868 : * This makes it possible for servers to be co-located in the same data
869 : * center or be separated geographically as long as they are reachable
870 : * through the underlay L3 network.
871 : *
872 : * You can refer to this kind of L2 overlay bridge domain as a GENEVE
873 : * segment.
874 : *
875 : * @cliexpar
876 : * Example of how to create a GENEVE Tunnel:
877 : * @cliexcmd{create geneve tunnel local 10.0.3.1 remote 10.0.3.3 vni 13 encap-vrf-id 7}
878 : * Example of how to delete a GENEVE Tunnel:
879 : * @cliexcmd{create geneve tunnel local 10.0.3.1 remote 10.0.3.3 vni 13 del}
880 : ?*/
881 : /* *INDENT-OFF* */
882 221929 : VLIB_CLI_COMMAND (create_geneve_tunnel_command, static) = {
883 : .path = "create geneve tunnel",
884 : .short_help =
885 : "create geneve tunnel local <local-vtep-addr>"
886 : " {remote <remote-vtep-addr>|group <mcast-vtep-addr> <intf-name>} vni <nn>"
887 : " [encap-vrf-id <nn>] [decap-next [l2|node <name>]] [l3-mode] [del]",
888 : .function = geneve_add_del_tunnel_command_fn,
889 : };
890 : /* *INDENT-ON* */
891 :
892 : static clib_error_t *
893 6 : show_geneve_tunnel_command_fn (vlib_main_t * vm,
894 : unformat_input_t * input,
895 : vlib_cli_command_t * cmd)
896 : {
897 6 : geneve_main_t *vxm = &geneve_main;
898 : geneve_tunnel_t *t;
899 :
900 6 : if (pool_elts (vxm->tunnels) == 0)
901 1 : vlib_cli_output (vm, "No geneve tunnels configured...");
902 :
903 116 : pool_foreach (t, vxm->tunnels)
904 : {
905 110 : vlib_cli_output (vm, "%U", format_geneve_tunnel, t);
906 : }
907 :
908 6 : return 0;
909 : }
910 :
911 : /*?
912 : * Display all the GENEVE Tunnel entries.
913 : *
914 : * @cliexpar
915 : * Example of how to display the GENEVE Tunnel entries:
916 : * @cliexstart{show geneve tunnel}
917 : * [0] local 10.0.3.1 remote 10.0.3.3 vni 13 encap_fib_index 0 sw_if_index 5 decap_next l2
918 : * @cliexend
919 : ?*/
920 : /* *INDENT-OFF* */
921 221929 : VLIB_CLI_COMMAND (show_geneve_tunnel_command, static) = {
922 : .path = "show geneve tunnel",
923 : .short_help = "show geneve tunnel",
924 : .function = show_geneve_tunnel_command_fn,
925 : };
926 : /* *INDENT-ON* */
927 :
928 :
929 : void
930 0 : vnet_int_geneve_bypass_mode (u32 sw_if_index, u8 is_ip6, u8 is_enable)
931 : {
932 0 : if (is_ip6)
933 0 : vnet_feature_enable_disable ("ip6-unicast", "ip6-geneve-bypass",
934 : sw_if_index, is_enable, 0, 0);
935 : else
936 0 : vnet_feature_enable_disable ("ip4-unicast", "ip4-geneve-bypass",
937 : sw_if_index, is_enable, 0, 0);
938 0 : }
939 :
940 :
941 : static clib_error_t *
942 0 : set_ip_geneve_bypass (u32 is_ip6,
943 : unformat_input_t * input, vlib_cli_command_t * cmd)
944 : {
945 0 : unformat_input_t _line_input, *line_input = &_line_input;
946 0 : vnet_main_t *vnm = vnet_get_main ();
947 0 : clib_error_t *error = 0;
948 : u32 sw_if_index, is_enable;
949 :
950 0 : sw_if_index = ~0;
951 0 : is_enable = 1;
952 :
953 0 : if (!unformat_user (input, unformat_line_input, line_input))
954 0 : return 0;
955 :
956 0 : while (unformat_check_input (line_input) != UNFORMAT_END_OF_INPUT)
957 : {
958 0 : if (unformat_user
959 : (line_input, unformat_vnet_sw_interface, vnm, &sw_if_index))
960 : ;
961 0 : else if (unformat (line_input, "del"))
962 0 : is_enable = 0;
963 : else
964 : {
965 0 : error = unformat_parse_error (line_input);
966 0 : goto done;
967 : }
968 : }
969 :
970 0 : if (~0 == sw_if_index)
971 : {
972 0 : error = clib_error_return (0, "unknown interface `%U'",
973 : format_unformat_error, line_input);
974 0 : goto done;
975 : }
976 :
977 0 : vnet_int_geneve_bypass_mode (sw_if_index, is_ip6, is_enable);
978 :
979 0 : done:
980 0 : unformat_free (line_input);
981 :
982 0 : return error;
983 : }
984 :
985 : static clib_error_t *
986 0 : set_ip4_geneve_bypass (vlib_main_t * vm,
987 : unformat_input_t * input, vlib_cli_command_t * cmd)
988 : {
989 0 : return set_ip_geneve_bypass (0, input, cmd);
990 : }
991 :
992 : /*?
993 : * This command adds the 'ip4-geneve-bypass' graph node for a given interface.
994 : * By adding the IPv4 geneve-bypass graph node to an interface, the node checks
995 : * for and validate input geneve packet and bypass ip4-lookup, ip4-local,
996 : * ip4-udp-lookup nodes to speedup geneve packet forwarding. This node will
997 : * cause extra overhead to for non-geneve packets which is kept at a minimum.
998 : *
999 : * @cliexpar
1000 : * @parblock
1001 : * Example of graph node before ip4-geneve-bypass is enabled:
1002 : * @cliexstart{show vlib graph ip4-geneve-bypass}
1003 : * Name Next Previous
1004 : * ip4-geneve-bypass error-drop [0]
1005 : * geneve4-input [1]
1006 : * ip4-lookup [2]
1007 : * @cliexend
1008 : *
1009 : * Example of how to enable ip4-geneve-bypass on an interface:
1010 : * @cliexcmd{set interface ip geneve-bypass GigabitEthernet2/0/0}
1011 : *
1012 : * Example of graph node after ip4-geneve-bypass is enabled:
1013 : * @cliexstart{show vlib graph ip4-geneve-bypass}
1014 : * Name Next Previous
1015 : * ip4-geneve-bypass error-drop [0] ip4-input
1016 : * geneve4-input [1] ip4-input-no-checksum
1017 : * ip4-lookup [2]
1018 : * @cliexend
1019 : *
1020 : * Example of how to display the feature enabled on an interface:
1021 : * @cliexstart{show ip interface features GigabitEthernet2/0/0}
1022 : * IP feature paths configured on GigabitEthernet2/0/0...
1023 : * ...
1024 : * ipv4 unicast:
1025 : * ip4-geneve-bypass
1026 : * ip4-lookup
1027 : * ...
1028 : * @cliexend
1029 : *
1030 : * Example of how to disable ip4-geneve-bypass on an interface:
1031 : * @cliexcmd{set interface ip geneve-bypass GigabitEthernet2/0/0 del}
1032 : * @endparblock
1033 : ?*/
1034 : /* *INDENT-OFF* */
1035 221929 : VLIB_CLI_COMMAND (set_interface_ip_geneve_bypass_command, static) = {
1036 : .path = "set interface ip geneve-bypass",
1037 : .function = set_ip4_geneve_bypass,
1038 : .short_help = "set interface ip geneve-bypass <interface> [del]",
1039 : };
1040 : /* *INDENT-ON* */
1041 :
1042 : static clib_error_t *
1043 0 : set_ip6_geneve_bypass (vlib_main_t * vm,
1044 : unformat_input_t * input, vlib_cli_command_t * cmd)
1045 : {
1046 0 : return set_ip_geneve_bypass (1, input, cmd);
1047 : }
1048 :
1049 : /*?
1050 : * This command adds the 'ip6-geneve-bypass' graph node for a given interface.
1051 : * By adding the IPv6 geneve-bypass graph node to an interface, the node checks
1052 : * for and validate input geneve packet and bypass ip6-lookup, ip6-local,
1053 : * ip6-udp-lookup nodes to speedup geneve packet forwarding. This node will
1054 : * cause extra overhead to for non-geneve packets which is kept at a minimum.
1055 : *
1056 : * @cliexpar
1057 : * @parblock
1058 : * Example of graph node before ip6-geneve-bypass is enabled:
1059 : * @cliexstart{show vlib graph ip6-geneve-bypass}
1060 : * Name Next Previous
1061 : * ip6-geneve-bypass error-drop [0]
1062 : * geneve6-input [1]
1063 : * ip6-lookup [2]
1064 : * @cliexend
1065 : *
1066 : * Example of how to enable ip6-geneve-bypass on an interface:
1067 : * @cliexcmd{set interface ip6 geneve-bypass GigabitEthernet2/0/0}
1068 : *
1069 : * Example of graph node after ip6-geneve-bypass is enabled:
1070 : * @cliexstart{show vlib graph ip6-geneve-bypass}
1071 : * Name Next Previous
1072 : * ip6-geneve-bypass error-drop [0] ip6-input
1073 : * geneve6-input [1] ip4-input-no-checksum
1074 : * ip6-lookup [2]
1075 : * @cliexend
1076 : *
1077 : * Example of how to display the feature enabled on an interface:
1078 : * @cliexstart{show ip interface features GigabitEthernet2/0/0}
1079 : * IP feature paths configured on GigabitEthernet2/0/0...
1080 : * ...
1081 : * ipv6 unicast:
1082 : * ip6-geneve-bypass
1083 : * ip6-lookup
1084 : * ...
1085 : * @cliexend
1086 : *
1087 : * Example of how to disable ip6-geneve-bypass on an interface:
1088 : * @cliexcmd{set interface ip6 geneve-bypass GigabitEthernet2/0/0 del}
1089 : * @endparblock
1090 : ?*/
1091 : /* *INDENT-OFF* */
1092 221929 : VLIB_CLI_COMMAND (set_interface_ip6_geneve_bypass_command, static) = {
1093 : .path = "set interface ip6 geneve-bypass",
1094 : .function = set_ip6_geneve_bypass,
1095 : .short_help = "set interface ip6 geneve-bypass <interface> [del]",
1096 : };
1097 : /* *INDENT-ON* */
1098 :
1099 : clib_error_t *
1100 575 : geneve_init (vlib_main_t * vm)
1101 : {
1102 575 : geneve_main_t *vxm = &geneve_main;
1103 :
1104 575 : vxm->vnet_main = vnet_get_main ();
1105 575 : vxm->vlib_main = vm;
1106 :
1107 : /* initialize the ip6 hash */
1108 575 : vxm->geneve6_tunnel_by_key = hash_create_mem (0,
1109 : sizeof (geneve6_tunnel_key_t),
1110 : sizeof (uword));
1111 575 : vxm->vtep_table = vtep_table_create ();
1112 575 : vxm->mcast_shared = hash_create_mem (0,
1113 : sizeof (ip46_address_t),
1114 : sizeof (mcast_shared_t));
1115 :
1116 575 : fib_node_register_type (FIB_NODE_TYPE_GENEVE_TUNNEL, &geneve_vft);
1117 :
1118 575 : return 0;
1119 : }
1120 :
1121 1151 : VLIB_INIT_FUNCTION (geneve_init);
1122 :
1123 : /*
1124 : * fd.io coding-style-patch-verification: ON
1125 : *
1126 : * Local Variables:
1127 : * eval: (c-set-style "gnu")
1128 : * End:
1129 : */
|
