Line data Source code
1 : /*
2 : * Copyright (c) 2016 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 :
16 : #include <vnet/adj/adj_mcast.h>
17 : #include <vnet/adj/adj_internal.h>
18 : #include <vnet/fib/fib_walk.h>
19 : #include <vnet/ip/ip.h>
20 :
21 : /*
22 : * The 'DB' of all mcast adjs.
23 : * There is only one mcast per-interface per-protocol, so this is a per-interface
24 : * vector
25 : */
26 : static adj_index_t *adj_mcasts[FIB_PROTOCOL_MAX];
27 :
28 : static u32
29 3886 : adj_get_mcast_node (fib_protocol_t proto)
30 : {
31 3886 : switch (proto) {
32 88 : case FIB_PROTOCOL_IP4:
33 88 : return (ip4_rewrite_mcast_node.index);
34 3798 : case FIB_PROTOCOL_IP6:
35 3798 : return (ip6_rewrite_mcast_node.index);
36 0 : case FIB_PROTOCOL_MPLS:
37 0 : break;
38 : }
39 0 : ASSERT(0);
40 0 : return (0);
41 : }
42 :
43 : /*
44 : * adj_mcast_add_or_lock
45 : *
46 : * The next_hop address here is used for source address selection in the DP.
47 : * The mcast adj is added to an interface's connected prefix, the next-hop
48 : * passed here is the local prefix on the same interface.
49 : */
50 : adj_index_t
51 7283 : adj_mcast_add_or_lock (fib_protocol_t proto,
52 : vnet_link_t link_type,
53 : u32 sw_if_index)
54 : {
55 : ip_adjacency_t * adj;
56 :
57 8772 : vec_validate_init_empty(adj_mcasts[proto], sw_if_index, ADJ_INDEX_INVALID);
58 :
59 7283 : if (ADJ_INDEX_INVALID == adj_mcasts[proto][sw_if_index])
60 : {
61 : vnet_main_t *vnm;
62 :
63 1943 : vnm = vnet_get_main();
64 1943 : adj = adj_alloc(proto);
65 :
66 1943 : adj->lookup_next_index = IP_LOOKUP_NEXT_MCAST;
67 1943 : adj->ia_nh_proto = proto;
68 1943 : adj->ia_link = link_type;
69 1943 : adj->ia_node_index = adj_get_mcast_node(proto);
70 1943 : adj_mcasts[proto][sw_if_index] = adj_get_index(adj);
71 1943 : adj_lock(adj_get_index(adj));
72 :
73 1943 : vnet_rewrite_init(vnm, sw_if_index, link_type,
74 : adj->ia_node_index,
75 : vnet_tx_node_index_for_sw_interface(vnm, sw_if_index),
76 : &adj->rewrite_header);
77 :
78 : /*
79 : * we need a rewrite where the destination IP address is converted
80 : * to the appropriate link-layer address. This is interface specific.
81 : * So ask the interface to do it.
82 : */
83 1943 : vnet_update_adjacency_for_sw_interface(vnm, sw_if_index,
84 : adj_get_index(adj));
85 :
86 1943 : adj_delegate_adj_created(adj);
87 : }
88 : else
89 : {
90 5340 : adj = adj_get(adj_mcasts[proto][sw_if_index]);
91 5340 : adj_lock(adj_get_index(adj));
92 : }
93 :
94 7283 : return (adj_get_index(adj));
95 : }
96 :
97 : /**
98 : * adj_mcast_update_rewrite
99 : *
100 : * Update the adjacency's rewrite string. A NULL string implies the
101 : * rewrite is reset (i.e. when ARP/ND entry is gone).
102 : * NB: the adj being updated may be handling traffic in the DP.
103 : */
104 : void
105 1943 : adj_mcast_update_rewrite (adj_index_t adj_index,
106 : u8 *rewrite,
107 : u8 offset)
108 : {
109 : ip_adjacency_t *adj;
110 :
111 1943 : ASSERT(ADJ_INDEX_INVALID != adj_index);
112 :
113 1943 : adj = adj_get(adj_index);
114 :
115 : /*
116 : * update the adj's rewrite string and build the arc
117 : * from the rewrite node to the interface's TX node
118 : */
119 3886 : adj_nbr_update_rewrite_internal(adj, IP_LOOKUP_NEXT_MCAST,
120 1943 : adj_get_mcast_node(adj->ia_nh_proto),
121 : vnet_tx_node_index_for_sw_interface(
122 1943 : vnet_get_main(),
123 : adj->rewrite_header.sw_if_index),
124 : rewrite);
125 : /*
126 : * set the offset corresponding to the mcast IP address rewrite
127 : */
128 1943 : adj->rewrite_header.dst_mcast_offset = offset;
129 1943 : }
130 :
131 : /**
132 : * adj_mcast_midchain_update_rewrite
133 : *
134 : * Update the adjacency's rewrite string. A NULL string implies the
135 : * rewrite is reset (i.e. when ARP/ND entry is gone).
136 : * NB: the adj being updated may be handling traffic in the DP.
137 : */
138 : void
139 0 : adj_mcast_midchain_update_rewrite (adj_index_t adj_index,
140 : adj_midchain_fixup_t fixup,
141 : const void *fixup_data,
142 : adj_flags_t flags,
143 : u8 *rewrite,
144 : u8 offset,
145 : u32 mask)
146 : {
147 : ip_adjacency_t *adj;
148 :
149 0 : ASSERT(ADJ_INDEX_INVALID != adj_index);
150 :
151 0 : adj = adj_get(adj_index);
152 :
153 : /*
154 : * one time only update. since we don't support changing the tunnel
155 : * src,dst, this is all we need.
156 : */
157 0 : ASSERT(adj->lookup_next_index == IP_LOOKUP_NEXT_MCAST);
158 : /*
159 : * tunnels can always provide a rewrite.
160 : */
161 0 : ASSERT(NULL != rewrite);
162 :
163 0 : adj_midchain_setup(adj_index, fixup, fixup_data, flags);
164 :
165 : /*
166 : * update the adj's rewrite string and build the arc
167 : * from the rewrite node to the interface's TX node
168 : */
169 0 : adj_nbr_update_rewrite_internal(adj, IP_LOOKUP_NEXT_MCAST_MIDCHAIN,
170 0 : adj_get_mcast_node(adj->ia_nh_proto),
171 : vnet_tx_node_index_for_sw_interface(
172 0 : vnet_get_main(),
173 : adj->rewrite_header.sw_if_index),
174 : rewrite);
175 :
176 0 : adj->rewrite_header.dst_mcast_offset = offset;
177 0 : }
178 :
179 : void
180 1761 : adj_mcast_remove (fib_protocol_t proto,
181 : u32 sw_if_index)
182 : {
183 1761 : ASSERT(sw_if_index < vec_len(adj_mcasts[proto]));
184 :
185 1761 : adj_mcasts[proto][sw_if_index] = ADJ_INDEX_INVALID;
186 1761 : }
187 :
188 : static clib_error_t *
189 26450 : adj_mcast_interface_state_change (vnet_main_t * vnm,
190 : u32 sw_if_index,
191 : u32 flags)
192 : {
193 : /*
194 : * for each mcast on the interface trigger a walk back to the children
195 : */
196 : fib_protocol_t proto;
197 : ip_adjacency_t *adj;
198 :
199 :
200 79350 : for (proto = FIB_PROTOCOL_IP4; proto <= FIB_PROTOCOL_IP6; proto++)
201 : {
202 52900 : if (sw_if_index >= vec_len(adj_mcasts[proto]) ||
203 7938 : ADJ_INDEX_INVALID == adj_mcasts[proto][sw_if_index])
204 50402 : continue;
205 :
206 2498 : adj = adj_get(adj_mcasts[proto][sw_if_index]);
207 :
208 4996 : fib_node_back_walk_ctx_t bw_ctx = {
209 2498 : .fnbw_reason = (flags & VNET_SW_INTERFACE_FLAG_ADMIN_UP ?
210 2498 : FIB_NODE_BW_REASON_FLAG_INTERFACE_UP :
211 : FIB_NODE_BW_REASON_FLAG_INTERFACE_DOWN),
212 : };
213 :
214 2498 : fib_walk_sync(FIB_NODE_TYPE_ADJ, adj_get_index(adj), &bw_ctx);
215 : }
216 :
217 26450 : return (NULL);
218 : }
219 :
220 2801 : VNET_SW_INTERFACE_ADMIN_UP_DOWN_FUNCTION(adj_mcast_interface_state_change);
221 :
222 : /**
223 : * @brief Invoked on each SW interface of a HW interface when the
224 : * HW interface state changes
225 : */
226 : static walk_rc_t
227 13182 : adj_mcast_hw_sw_interface_state_change (vnet_main_t * vnm,
228 : u32 sw_if_index,
229 : void *arg)
230 : {
231 13182 : adj_mcast_interface_state_change(vnm, sw_if_index, (uword) arg);
232 :
233 13182 : return (WALK_CONTINUE);
234 : }
235 :
236 : /**
237 : * @brief Registered callback for HW interface state changes
238 : */
239 : static clib_error_t *
240 13092 : adj_mcast_hw_interface_state_change (vnet_main_t * vnm,
241 : u32 hw_if_index,
242 : u32 flags)
243 : {
244 : /*
245 : * walk SW interfaces on the HW
246 : */
247 : uword sw_flags;
248 :
249 13092 : sw_flags = ((flags & VNET_HW_INTERFACE_FLAG_LINK_UP) ?
250 13092 : VNET_SW_INTERFACE_FLAG_ADMIN_UP :
251 : 0);
252 :
253 13092 : vnet_hw_interface_walk_sw(vnm, hw_if_index,
254 : adj_mcast_hw_sw_interface_state_change,
255 : (void*) sw_flags);
256 :
257 13092 : return (NULL);
258 : }
259 :
260 2801 : VNET_HW_INTERFACE_LINK_UP_DOWN_FUNCTION(
261 : adj_mcast_hw_interface_state_change);
262 :
263 : static clib_error_t *
264 11597 : adj_mcast_interface_delete (vnet_main_t * vnm,
265 : u32 sw_if_index,
266 : u32 is_add)
267 : {
268 : /*
269 : * for each mcast on the interface trigger a walk back to the children
270 : */
271 : fib_protocol_t proto;
272 : ip_adjacency_t *adj;
273 :
274 11597 : if (is_add)
275 : {
276 : /*
277 : * not interested in interface additions. we will not back walk
278 : * to resolve paths through newly added interfaces. Why? The control
279 : * plane should have the brains to add interfaces first, then routes.
280 : * So the case where there are paths with a interface that matches
281 : * one just created is the case where the path resolved through an
282 : * interface that was deleted, and still has not been removed. The
283 : * new interface added, is NO GUARANTEE that the interface being
284 : * added now, even though it may have the same sw_if_index, is the
285 : * same interface that the path needs. So tough!
286 : * If the control plane wants these routes to resolve it needs to
287 : * remove and add them again.
288 : */
289 7418 : return (NULL);
290 : }
291 :
292 12537 : for (proto = FIB_PROTOCOL_IP4; proto <= FIB_PROTOCOL_IP6; proto++)
293 : {
294 8358 : if (sw_if_index >= vec_len(adj_mcasts[proto]) ||
295 580 : ADJ_INDEX_INVALID == adj_mcasts[proto][sw_if_index])
296 8358 : continue;
297 :
298 0 : adj = adj_get(adj_mcasts[proto][sw_if_index]);
299 :
300 0 : fib_node_back_walk_ctx_t bw_ctx = {
301 : .fnbw_reason = FIB_NODE_BW_REASON_FLAG_INTERFACE_DELETE,
302 : };
303 :
304 0 : fib_walk_sync(FIB_NODE_TYPE_ADJ, adj_get_index(adj), &bw_ctx);
305 : }
306 :
307 4179 : return (NULL);
308 : }
309 :
310 3363 : VNET_SW_INTERFACE_ADD_DEL_FUNCTION(adj_mcast_interface_delete);
311 :
312 : /**
313 : * @brief Walk the multicast Adjacencies on a given interface
314 : */
315 : void
316 274800 : adj_mcast_walk (u32 sw_if_index,
317 : fib_protocol_t proto,
318 : adj_walk_cb_t cb,
319 : void *ctx)
320 : {
321 274800 : if (vec_len(adj_mcasts[proto]) > sw_if_index)
322 : {
323 42414 : if (ADJ_INDEX_INVALID != adj_mcasts[proto][sw_if_index])
324 : {
325 23002 : cb(adj_mcasts[proto][sw_if_index], ctx);
326 : }
327 : }
328 274800 : }
329 :
330 : u8*
331 2191 : format_adj_mcast (u8* s, va_list *ap)
332 : {
333 2191 : index_t index = va_arg(*ap, index_t);
334 2191 : CLIB_UNUSED(u32 indent) = va_arg(*ap, u32);
335 : ip_adjacency_t * adj;
336 :
337 2191 : if (!adj_is_valid(index))
338 1616 : return format(s, "<invalid adjacency>");
339 :
340 575 : adj = adj_get(index);
341 :
342 575 : s = format(s, "%U-mcast: ",
343 575 : format_fib_protocol, adj->ia_nh_proto);
344 575 : if (adj->rewrite_header.flags & VNET_REWRITE_HAS_FEATURES)
345 0 : s = format(s, "[features] ");
346 575 : s = format (s, "%U",
347 : format_vnet_rewrite,
348 : &adj->rewrite_header, sizeof (adj->rewrite_data), 0);
349 :
350 575 : return (s);
351 : }
352 :
353 : u8*
354 0 : format_adj_mcast_midchain (u8* s, va_list *ap)
355 : {
356 0 : index_t index = va_arg(*ap, index_t);
357 0 : CLIB_UNUSED(u32 indent) = va_arg(*ap, u32);
358 0 : ip_adjacency_t * adj = adj_get(index);
359 :
360 0 : s = format(s, "%U-mcast-midchain: ",
361 0 : format_fib_protocol, adj->ia_nh_proto);
362 0 : s = format (s, "%U",
363 : format_vnet_rewrite,
364 : &adj->rewrite_header,
365 : sizeof (adj->rewrite_data), 0);
366 0 : s = format (s, "\n%Ustacked-on:\n%U%U",
367 : format_white_space, indent,
368 : format_white_space, indent+2,
369 : format_dpo_id, &adj->sub_type.midchain.next_dpo, indent+2);
370 :
371 0 : return (s);
372 : }
373 :
374 :
375 : static void
376 16766 : adj_dpo_lock (dpo_id_t *dpo)
377 : {
378 16766 : adj_lock(dpo->dpoi_index);
379 16766 : }
380 : static void
381 14262 : adj_dpo_unlock (dpo_id_t *dpo)
382 : {
383 14262 : adj_unlock(dpo->dpoi_index);
384 14262 : }
385 :
386 : const static dpo_vft_t adj_mcast_dpo_vft = {
387 : .dv_lock = adj_dpo_lock,
388 : .dv_unlock = adj_dpo_unlock,
389 : .dv_format = format_adj_mcast,
390 : .dv_get_urpf = adj_dpo_get_urpf,
391 : .dv_get_mtu = adj_dpo_get_mtu,
392 : };
393 : const static dpo_vft_t adj_mcast_midchain_dpo_vft = {
394 : .dv_lock = adj_dpo_lock,
395 : .dv_unlock = adj_dpo_unlock,
396 : .dv_format = format_adj_mcast_midchain,
397 : .dv_get_urpf = adj_dpo_get_urpf,
398 : .dv_get_mtu = adj_dpo_get_mtu,
399 : };
400 :
401 : /**
402 : * @brief The per-protocol VLIB graph nodes that are assigned to a mcast
403 : * object.
404 : *
405 : * this means that these graph nodes are ones from which a mcast is the
406 : * parent object in the DPO-graph.
407 : */
408 : const static char* const adj_mcast_ip4_nodes[] =
409 : {
410 : "ip4-rewrite-mcast",
411 : NULL,
412 : };
413 : const static char* const adj_mcast_ip6_nodes[] =
414 : {
415 : "ip6-rewrite-mcast",
416 : NULL,
417 : };
418 :
419 : const static char* const * const adj_mcast_nodes[DPO_PROTO_NUM] =
420 : {
421 : [DPO_PROTO_IP4] = adj_mcast_ip4_nodes,
422 : [DPO_PROTO_IP6] = adj_mcast_ip6_nodes,
423 : [DPO_PROTO_MPLS] = NULL,
424 : };
425 :
426 : /**
427 : * @brief The per-protocol VLIB graph nodes that are assigned to a mcast
428 : * object.
429 : *
430 : * this means that these graph nodes are ones from which a mcast is the
431 : * parent object in the DPO-graph.
432 : */
433 : const static char* const adj_mcast_midchain_ip4_nodes[] =
434 : {
435 : "ip4-mcast-midchain",
436 : NULL,
437 : };
438 : const static char* const adj_mcast_midchain_ip6_nodes[] =
439 : {
440 : "ip6-mcast-midchain",
441 : NULL,
442 : };
443 :
444 : const static char* const * const adj_mcast_midchain_nodes[DPO_PROTO_NUM] =
445 : {
446 : [DPO_PROTO_IP4] = adj_mcast_midchain_ip4_nodes,
447 : [DPO_PROTO_IP6] = adj_mcast_midchain_ip6_nodes,
448 : [DPO_PROTO_MPLS] = NULL,
449 : };
450 :
451 : /**
452 : * @brief Return the size of the adj DB.
453 : * This is only for testing purposes so an efficient implementation is not needed
454 : */
455 : u32
456 6 : adj_mcast_db_size (void)
457 : {
458 : u32 n_adjs, sw_if_index;
459 : fib_protocol_t proto;
460 :
461 6 : n_adjs = 0;
462 18 : for (proto = FIB_PROTOCOL_IP4; proto <= FIB_PROTOCOL_IP6; proto++)
463 : {
464 74 : for (sw_if_index = 0;
465 62 : sw_if_index < vec_len(adj_mcasts[proto]);
466 50 : sw_if_index++)
467 : {
468 50 : if (ADJ_INDEX_INVALID != adj_mcasts[proto][sw_if_index])
469 : {
470 6 : n_adjs++;
471 : }
472 : }
473 : }
474 :
475 6 : return (n_adjs);
476 : }
477 :
478 : void
479 559 : adj_mcast_module_init (void)
480 : {
481 559 : dpo_register(DPO_ADJACENCY_MCAST,
482 : &adj_mcast_dpo_vft,
483 : adj_mcast_nodes);
484 559 : dpo_register(DPO_ADJACENCY_MCAST_MIDCHAIN,
485 : &adj_mcast_midchain_dpo_vft,
486 : adj_mcast_midchain_nodes);
487 559 : }
|
