]>
Commit | Line | Data |
---|---|---|
ad28e79a SW |
1 | /* Zebra Nexthop Group Code. |
2 | * Copyright (C) 2019 Cumulus Networks, Inc. | |
3 | * Donald Sharp | |
4 | * Stephen Worley | |
5 | * | |
6 | * This file is part of FRR. | |
7 | * | |
8 | * FRR is free software; you can redistribute it and/or modify it | |
9 | * under the terms of the GNU General Public License as published by the | |
10 | * Free Software Foundation; either version 2, or (at your option) any | |
11 | * later version. | |
12 | * | |
13 | * FRR is distributed in the hope that it will be useful, but | |
14 | * WITHOUT ANY WARRANTY; without even the implied warranty of | |
15 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU | |
16 | * General Public License for more details. | |
17 | * | |
18 | * You should have received a copy of the GNU General Public License | |
19 | * along with FRR; see the file COPYING. If not, write to the Free | |
20 | * Software Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA | |
21 | * 02111-1307, USA. | |
22 | */ | |
23 | #include <zebra.h> | |
24 | ||
25 | #include "lib/nexthop.h" | |
50d89650 | 26 | #include "lib/nexthop_group_private.h" |
ad28e79a | 27 | #include "lib/routemap.h" |
b43434ad | 28 | #include "lib/mpls.h" |
69171da2 | 29 | #include "lib/jhash.h" |
ad28e79a SW |
30 | |
31 | #include "zebra/connected.h" | |
32 | #include "zebra/debug.h" | |
33 | #include "zebra/zebra_router.h" | |
34 | #include "zebra/zebra_nhg.h" | |
35 | #include "zebra/zebra_rnh.h" | |
36 | #include "zebra/zebra_routemap.h" | |
37 | #include "zebra/rt.h" | |
d9f5b2f5 SW |
38 | #include "zebra_errors.h" |
39 | ||
40 | /** | |
41 | * zebra_nhg_lookup_id() - Lookup the nexthop group id in the id table | |
42 | * | |
43 | * @id: ID to look for | |
44 | * | |
45 | * Return: Nexthop hash entry if found/NULL if not found | |
46 | */ | |
47 | struct nhg_hash_entry *zebra_nhg_lookup_id(uint32_t id) | |
48 | { | |
49 | struct nhg_hash_entry lookup = {0}; | |
50 | ||
51 | lookup.id = id; | |
52 | return hash_lookup(zrouter.nhgs_id, &lookup); | |
53 | } | |
54 | ||
55 | /** | |
56 | * zebra_nhg_insert_id() - Insert a nhe into the id hashed table | |
57 | * | |
58 | * @nhe: The entry directly from the other table | |
59 | * | |
60 | * Return: Result status | |
61 | */ | |
62 | int zebra_nhg_insert_id(struct nhg_hash_entry *nhe) | |
63 | { | |
64 | if (hash_lookup(zrouter.nhgs_id, nhe)) { | |
65 | flog_err( | |
66 | EC_ZEBRA_NHG_TABLE_INSERT_FAILED, | |
67 | "Failed inserting NHG id=%u into the ID hash table, entry already exists", | |
68 | nhe->id); | |
69 | return -1; | |
70 | } | |
71 | ||
72 | hash_get(zrouter.nhgs_id, nhe, hash_alloc_intern); | |
73 | ||
74 | return 0; | |
75 | } | |
ad28e79a | 76 | |
4e49c8b8 DS |
77 | |
78 | static void *zebra_nhg_alloc(void *arg) | |
79 | { | |
d9f5b2f5 SW |
80 | /* lock for getiing and setting the id */ |
81 | static pthread_mutex_t lock = PTHREAD_MUTEX_INITIALIZER; | |
82 | /* id counter to keep in sync with kernel */ | |
83 | static uint32_t id_counter = 0; | |
4e49c8b8 DS |
84 | struct nhg_hash_entry *nhe; |
85 | struct nhg_hash_entry *copy = arg; | |
86 | ||
d9f5b2f5 SW |
87 | nhe = XCALLOC(MTYPE_TMP, sizeof(struct nhg_hash_entry)); |
88 | ||
89 | pthread_mutex_lock(&lock); /* Lock, set the id counter from kernel */ | |
90 | if (copy->id) { | |
91 | /* This is from the kernel if it has an id */ | |
92 | if (copy->id > id_counter) { | |
93 | /* Increase our counter so we don't try to create | |
94 | * an ID that already exists | |
95 | */ | |
96 | id_counter = copy->id; | |
97 | } | |
98 | nhe->id = copy->id; | |
99 | /* Mark as valid since from the kernel */ | |
100 | SET_FLAG(nhe->flags, NEXTHOP_GROUP_INSTALLED); | |
101 | SET_FLAG(nhe->flags, NEXTHOP_GROUP_VALID); | |
102 | } else { | |
103 | nhe->id = ++id_counter; | |
104 | } | |
105 | pthread_mutex_unlock(&lock); | |
4e49c8b8 DS |
106 | |
107 | nhe->vrf_id = copy->vrf_id; | |
4e49c8b8 | 108 | nhe->refcnt = 0; |
d9f5b2f5 | 109 | nhe->is_kernel_nh = false; |
4e49c8b8 DS |
110 | nhe->dplane_ref = zebra_router_get_next_sequence(); |
111 | nhe->nhg.nexthop = NULL; | |
112 | ||
113 | nexthop_group_copy(&nhe->nhg, ©->nhg); | |
114 | ||
d9f5b2f5 SW |
115 | /* Add to id table as well */ |
116 | zebra_nhg_insert_id(nhe); | |
117 | ||
4e49c8b8 DS |
118 | |
119 | return nhe; | |
120 | } | |
121 | ||
122 | static uint32_t zebra_nhg_hash_key_nexthop_group(struct nexthop_group *nhg) | |
123 | { | |
124 | struct nexthop *nh; | |
125 | uint32_t i; | |
126 | uint32_t key = 0; | |
127 | ||
128 | /* | |
129 | * We are not interested in hashing over any recursively | |
130 | * resolved nexthops | |
131 | */ | |
132 | for (nh = nhg->nexthop; nh; nh = nh->next) { | |
8b5bdc8b | 133 | key = jhash_1word(nh->type, key); |
4e49c8b8 DS |
134 | key = jhash_2words(nh->vrf_id, nh->nh_label_type, key); |
135 | /* gate and blackhole are together in a union */ | |
136 | key = jhash(&nh->gate, sizeof(nh->gate), key); | |
137 | key = jhash(&nh->src, sizeof(nh->src), key); | |
138 | key = jhash(&nh->rmap_src, sizeof(nh->rmap_src), key); | |
139 | if (nh->nh_label) { | |
140 | for (i = 0; i < nh->nh_label->num_labels; i++) | |
141 | key = jhash_1word(nh->nh_label->label[i], key); | |
142 | } | |
143 | switch (nh->type) { | |
144 | case NEXTHOP_TYPE_IPV4_IFINDEX: | |
145 | case NEXTHOP_TYPE_IPV6_IFINDEX: | |
146 | case NEXTHOP_TYPE_IFINDEX: | |
147 | key = jhash_1word(nh->ifindex, key); | |
148 | break; | |
149 | case NEXTHOP_TYPE_BLACKHOLE: | |
150 | case NEXTHOP_TYPE_IPV4: | |
151 | case NEXTHOP_TYPE_IPV6: | |
152 | break; | |
153 | } | |
154 | } | |
155 | return key; | |
156 | } | |
157 | ||
158 | uint32_t zebra_nhg_hash_key(const void *arg) | |
159 | { | |
160 | const struct nhg_hash_entry *nhe = arg; | |
d9f5b2f5 | 161 | |
4e49c8b8 DS |
162 | int key = 0x5a351234; |
163 | ||
8b5bdc8b | 164 | key = jhash_1word(nhe->vrf_id, key); |
4e49c8b8 | 165 | |
d9f5b2f5 SW |
166 | key = jhash_1word(zebra_nhg_hash_key_nexthop_group(&nhe->nhg), key); |
167 | ||
168 | ||
169 | return key; | |
4e49c8b8 DS |
170 | } |
171 | ||
a95b8020 SW |
172 | uint32_t zebra_nhg_id_key(const void *arg) |
173 | { | |
174 | const struct nhg_hash_entry *nhe = arg; | |
175 | ||
176 | return nhe->id; | |
177 | } | |
178 | ||
4e49c8b8 DS |
179 | bool zebra_nhg_hash_equal(const void *arg1, const void *arg2) |
180 | { | |
181 | const struct nhg_hash_entry *nhe1 = arg1; | |
182 | const struct nhg_hash_entry *nhe2 = arg2; | |
183 | struct nexthop *nh1, *nh2; | |
184 | uint32_t nh_count = 0; | |
185 | ||
186 | if (nhe1->vrf_id != nhe2->vrf_id) | |
187 | return false; | |
188 | ||
4e49c8b8 DS |
189 | /* |
190 | * Again we are not interested in looking at any recursively | |
191 | * resolved nexthops. Top level only | |
192 | */ | |
193 | for (nh1 = nhe1->nhg.nexthop; nh1; nh1 = nh1->next) { | |
194 | uint32_t inner_nh_count = 0; | |
195 | for (nh2 = nhe2->nhg.nexthop; nh2; nh2 = nh2->next) { | |
196 | if (inner_nh_count == nh_count) { | |
197 | break; | |
198 | } | |
199 | inner_nh_count++; | |
200 | } | |
201 | ||
202 | if (!nexthop_same(nh1, nh2)) | |
203 | return false; | |
204 | ||
205 | nh_count++; | |
206 | } | |
207 | ||
208 | return true; | |
209 | } | |
210 | ||
d9f5b2f5 | 211 | bool zebra_nhg_hash_id_equal(const void *arg1, const void *arg2) |
4e49c8b8 | 212 | { |
d9f5b2f5 SW |
213 | const struct nhg_hash_entry *nhe1 = arg1; |
214 | const struct nhg_hash_entry *nhe2 = arg2; | |
4e49c8b8 | 215 | |
d9f5b2f5 SW |
216 | return nhe1->id == nhe2->id; |
217 | } | |
4e49c8b8 | 218 | |
d9f5b2f5 SW |
219 | struct nhg_hash_entry *zebra_nhg_find_id(uint32_t id, struct nexthop_group *nhg) |
220 | { | |
221 | // TODO: How this will work is yet to be determined | |
222 | return NULL; | |
4e49c8b8 DS |
223 | } |
224 | ||
d9f5b2f5 SW |
225 | /** |
226 | * zebra_nhg_find() - Find the zebra nhg in our table, or create it | |
227 | * | |
228 | * @nhg: Nexthop group we lookup with | |
229 | * @vrf_id: VRF id | |
230 | * @id: ID we lookup with, 0 means its from us and we need to give it | |
231 | * an ID, otherwise its from the kernel as we use the ID it gave | |
232 | * us. | |
233 | * | |
234 | * Return: Hash entry found or created | |
235 | */ | |
236 | struct nhg_hash_entry *zebra_nhg_find(struct nexthop_group *nhg, | |
237 | vrf_id_t vrf_id, uint32_t id) | |
a95b8020 SW |
238 | { |
239 | struct nhg_hash_entry lookup = {0}; | |
d9f5b2f5 | 240 | struct nhg_hash_entry *nhe = NULL; |
a95b8020 | 241 | |
d9f5b2f5 SW |
242 | lookup.id = id; |
243 | lookup.vrf_id = vrf_id; | |
a95b8020 SW |
244 | lookup.nhg = *nhg; |
245 | ||
d9f5b2f5 SW |
246 | |
247 | nhe = hash_lookup(zrouter.nhgs, &lookup); | |
248 | ||
249 | if (!nhe) { | |
250 | nhe = hash_get(zrouter.nhgs, &lookup, zebra_nhg_alloc); | |
251 | } else { | |
252 | if (id) { | |
253 | /* Duplicate but with different ID from the kernel */ | |
254 | ||
255 | /* The kernel allows duplicate nexthops as long as they | |
256 | * have different IDs. We are ignoring those to prevent | |
257 | * syncing problems with the kernel changes. | |
258 | */ | |
259 | flog_warn( | |
260 | EC_ZEBRA_DUPLICATE_NHG_MESSAGE, | |
261 | "Nexthop Group from kernel with ID (%d) is a duplicate, ignoring", | |
262 | id); | |
263 | return NULL; | |
264 | } | |
265 | } | |
266 | ||
267 | return nhe; | |
a95b8020 SW |
268 | } |
269 | ||
d9f5b2f5 SW |
270 | /** |
271 | * zebra_nhg_free() - Free the nexthop group hash entry | |
272 | * | |
273 | * arg: Nexthop group entry to free | |
274 | */ | |
275 | void zebra_nhg_free(void *arg) | |
a95b8020 | 276 | { |
d9f5b2f5 | 277 | struct nhg_hash_entry *nhe = NULL; |
a95b8020 | 278 | |
d9f5b2f5 | 279 | nhe = (struct nhg_hash_entry *)arg; |
a95b8020 | 280 | |
d9f5b2f5 SW |
281 | nexthops_free(nhe->nhg.nexthop); |
282 | XFREE(MTYPE_TMP, nhe); | |
a95b8020 SW |
283 | } |
284 | ||
d9f5b2f5 SW |
285 | /** |
286 | * zebra_nhg_release() - Release a nhe from the tables | |
287 | * | |
288 | * @nhe: Nexthop group hash entry | |
289 | */ | |
290 | void zebra_nhg_release(struct nhg_hash_entry *nhe) | |
4e49c8b8 | 291 | { |
d9f5b2f5 SW |
292 | if (nhe->refcnt) { |
293 | flog_err( | |
294 | EC_ZEBRA_NHG_SYNC, | |
295 | "Kernel deleted a nexthop group with ID (%u) that we are still using for a route", | |
296 | nhe->id); | |
297 | // TODO: Re-send to kernel | |
298 | } | |
4e49c8b8 | 299 | |
d9f5b2f5 SW |
300 | hash_release(zrouter.nhgs, nhe); |
301 | hash_release(zrouter.nhgs_id, nhe); | |
302 | zebra_nhg_free(nhe); | |
303 | } | |
4e49c8b8 | 304 | |
d9f5b2f5 SW |
305 | /** |
306 | * zebra_nhg_decrement_ref() - Decrement the reference count, release if unused | |
307 | * | |
308 | * @nhe: Nexthop group hash entry | |
309 | * | |
310 | * If the counter hits 0 and is not a nexthop group that was created by the | |
311 | * kernel, we don't need to have it in our table anymore. | |
312 | */ | |
313 | void zebra_nhg_decrement_ref(struct nhg_hash_entry *nhe) | |
314 | { | |
4e49c8b8 DS |
315 | nhe->refcnt--; |
316 | ||
d9f5b2f5 SW |
317 | if (!nhe->is_kernel_nh && nhe->refcnt <= 0) { |
318 | zebra_nhg_release(nhe); | |
319 | } | |
320 | ||
4e49c8b8 DS |
321 | // re->ng = NULL; |
322 | } | |
323 | ||
ad28e79a SW |
324 | static void nexthop_set_resolved(afi_t afi, const struct nexthop *newhop, |
325 | struct nexthop *nexthop) | |
326 | { | |
327 | struct nexthop *resolved_hop; | |
b43434ad SW |
328 | uint8_t num_labels = 0; |
329 | mpls_label_t labels[MPLS_MAX_LABELS]; | |
330 | enum lsp_types_t label_type = ZEBRA_LSP_NONE; | |
331 | int i = 0; | |
ad28e79a SW |
332 | |
333 | resolved_hop = nexthop_new(); | |
334 | SET_FLAG(resolved_hop->flags, NEXTHOP_FLAG_ACTIVE); | |
335 | ||
336 | resolved_hop->vrf_id = nexthop->vrf_id; | |
337 | switch (newhop->type) { | |
338 | case NEXTHOP_TYPE_IPV4: | |
339 | case NEXTHOP_TYPE_IPV4_IFINDEX: | |
340 | /* If the resolving route specifies a gateway, use it */ | |
341 | resolved_hop->type = newhop->type; | |
342 | resolved_hop->gate.ipv4 = newhop->gate.ipv4; | |
343 | ||
344 | if (newhop->ifindex) { | |
345 | resolved_hop->type = NEXTHOP_TYPE_IPV4_IFINDEX; | |
346 | resolved_hop->ifindex = newhop->ifindex; | |
347 | } | |
348 | break; | |
349 | case NEXTHOP_TYPE_IPV6: | |
350 | case NEXTHOP_TYPE_IPV6_IFINDEX: | |
351 | resolved_hop->type = newhop->type; | |
352 | resolved_hop->gate.ipv6 = newhop->gate.ipv6; | |
353 | ||
354 | if (newhop->ifindex) { | |
355 | resolved_hop->type = NEXTHOP_TYPE_IPV6_IFINDEX; | |
356 | resolved_hop->ifindex = newhop->ifindex; | |
357 | } | |
358 | break; | |
359 | case NEXTHOP_TYPE_IFINDEX: | |
360 | /* If the resolving route is an interface route, | |
361 | * it means the gateway we are looking up is connected | |
362 | * to that interface. (The actual network is _not_ onlink). | |
363 | * Therefore, the resolved route should have the original | |
364 | * gateway as nexthop as it is directly connected. | |
365 | * | |
366 | * On Linux, we have to set the onlink netlink flag because | |
367 | * otherwise, the kernel won't accept the route. | |
368 | */ | |
369 | resolved_hop->flags |= NEXTHOP_FLAG_ONLINK; | |
370 | if (afi == AFI_IP) { | |
371 | resolved_hop->type = NEXTHOP_TYPE_IPV4_IFINDEX; | |
372 | resolved_hop->gate.ipv4 = nexthop->gate.ipv4; | |
373 | } else if (afi == AFI_IP6) { | |
374 | resolved_hop->type = NEXTHOP_TYPE_IPV6_IFINDEX; | |
375 | resolved_hop->gate.ipv6 = nexthop->gate.ipv6; | |
376 | } | |
377 | resolved_hop->ifindex = newhop->ifindex; | |
378 | break; | |
379 | case NEXTHOP_TYPE_BLACKHOLE: | |
380 | resolved_hop->type = NEXTHOP_TYPE_BLACKHOLE; | |
2dc359a6 | 381 | resolved_hop->bh_type = newhop->bh_type; |
ad28e79a SW |
382 | break; |
383 | } | |
384 | ||
385 | if (newhop->flags & NEXTHOP_FLAG_ONLINK) | |
386 | resolved_hop->flags |= NEXTHOP_FLAG_ONLINK; | |
387 | ||
b43434ad SW |
388 | /* Copy labels of the resolved route and the parent resolving to it */ |
389 | if (newhop->nh_label) { | |
390 | for (i = 0; i < newhop->nh_label->num_labels; i++) | |
391 | labels[num_labels++] = newhop->nh_label->label[i]; | |
392 | label_type = newhop->nh_label_type; | |
393 | } | |
394 | ||
395 | if (nexthop->nh_label) { | |
396 | for (i = 0; i < nexthop->nh_label->num_labels; i++) | |
397 | labels[num_labels++] = nexthop->nh_label->label[i]; | |
398 | ||
399 | /* If the parent has labels, use its type */ | |
400 | label_type = nexthop->nh_label_type; | |
401 | } | |
402 | ||
403 | if (num_labels) | |
404 | nexthop_add_labels(resolved_hop, label_type, num_labels, | |
405 | labels); | |
ad28e79a SW |
406 | |
407 | resolved_hop->rparent = nexthop; | |
50d89650 | 408 | _nexthop_add(&nexthop->resolved, resolved_hop); |
ad28e79a SW |
409 | } |
410 | ||
6913cb1b SW |
411 | /* Checks if nexthop we are trying to resolve to is valid */ |
412 | static bool nexthop_valid_resolve(const struct nexthop *nexthop, | |
413 | const struct nexthop *resolved) | |
414 | { | |
415 | /* Can't resolve to a recursive nexthop */ | |
416 | if (CHECK_FLAG(resolved->flags, NEXTHOP_FLAG_RECURSIVE)) | |
417 | return false; | |
418 | ||
419 | switch (nexthop->type) { | |
420 | case NEXTHOP_TYPE_IPV4_IFINDEX: | |
421 | case NEXTHOP_TYPE_IPV6_IFINDEX: | |
422 | /* If the nexthop we are resolving to does not match the | |
423 | * ifindex for the nexthop the route wanted, its not valid. | |
424 | */ | |
425 | if (nexthop->ifindex != resolved->ifindex) | |
426 | return false; | |
427 | break; | |
428 | case NEXTHOP_TYPE_IPV4: | |
429 | case NEXTHOP_TYPE_IPV6: | |
430 | case NEXTHOP_TYPE_IFINDEX: | |
431 | case NEXTHOP_TYPE_BLACKHOLE: | |
432 | break; | |
433 | } | |
434 | ||
435 | return true; | |
436 | } | |
437 | ||
ad28e79a SW |
438 | /* |
439 | * Given a nexthop we need to properly recursively resolve | |
440 | * the route. As such, do a table lookup to find and match | |
441 | * if at all possible. Set the nexthop->ifindex as appropriate | |
442 | */ | |
443 | static int nexthop_active(afi_t afi, struct route_entry *re, | |
444 | struct nexthop *nexthop, struct route_node *top) | |
445 | { | |
446 | struct prefix p; | |
447 | struct route_table *table; | |
448 | struct route_node *rn; | |
449 | struct route_entry *match = NULL; | |
450 | int resolved; | |
451 | struct nexthop *newhop; | |
452 | struct interface *ifp; | |
453 | rib_dest_t *dest; | |
5a0bdc78 | 454 | struct zebra_vrf *zvrf; |
ad28e79a SW |
455 | |
456 | if ((nexthop->type == NEXTHOP_TYPE_IPV4) | |
457 | || nexthop->type == NEXTHOP_TYPE_IPV6) | |
458 | nexthop->ifindex = 0; | |
459 | ||
460 | UNSET_FLAG(nexthop->flags, NEXTHOP_FLAG_RECURSIVE); | |
461 | nexthops_free(nexthop->resolved); | |
462 | nexthop->resolved = NULL; | |
463 | re->nexthop_mtu = 0; | |
464 | ||
465 | /* | |
a8c427ee | 466 | * If the kernel has sent us a NEW route, then |
ad28e79a | 467 | * by golly gee whiz it's a good route. |
a8c427ee SW |
468 | * |
469 | * If its an already INSTALLED route we have already handled, then the | |
470 | * kernel route's nexthop might have became unreachable | |
471 | * and we have to handle that. | |
ad28e79a | 472 | */ |
a8c427ee SW |
473 | if (!CHECK_FLAG(re->status, ROUTE_ENTRY_INSTALLED) |
474 | && (re->type == ZEBRA_ROUTE_KERNEL | |
475 | || re->type == ZEBRA_ROUTE_SYSTEM)) | |
ad28e79a SW |
476 | return 1; |
477 | ||
478 | /* | |
479 | * Check to see if we should trust the passed in information | |
480 | * for UNNUMBERED interfaces as that we won't find the GW | |
481 | * address in the routing table. | |
482 | * This check should suffice to handle IPv4 or IPv6 routes | |
483 | * sourced from EVPN routes which are installed with the | |
484 | * next hop as the remote VTEP IP. | |
485 | */ | |
486 | if (CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_ONLINK)) { | |
487 | ifp = if_lookup_by_index(nexthop->ifindex, nexthop->vrf_id); | |
488 | if (!ifp) { | |
489 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) | |
490 | zlog_debug( | |
491 | "\t%s: Onlink and interface: %u[%u] does not exist", | |
492 | __PRETTY_FUNCTION__, nexthop->ifindex, | |
493 | nexthop->vrf_id); | |
494 | return 0; | |
495 | } | |
496 | if (connected_is_unnumbered(ifp)) { | |
497 | if (if_is_operative(ifp)) | |
498 | return 1; | |
499 | else { | |
500 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) | |
501 | zlog_debug( | |
502 | "\t%s: Onlink and interface %s is not operative", | |
503 | __PRETTY_FUNCTION__, ifp->name); | |
504 | return 0; | |
505 | } | |
506 | } | |
507 | if (!if_is_operative(ifp)) { | |
508 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) | |
509 | zlog_debug( | |
510 | "\t%s: Interface %s is not unnumbered", | |
511 | __PRETTY_FUNCTION__, ifp->name); | |
512 | return 0; | |
513 | } | |
514 | } | |
515 | ||
516 | /* Make lookup prefix. */ | |
517 | memset(&p, 0, sizeof(struct prefix)); | |
518 | switch (afi) { | |
519 | case AFI_IP: | |
520 | p.family = AF_INET; | |
521 | p.prefixlen = IPV4_MAX_PREFIXLEN; | |
522 | p.u.prefix4 = nexthop->gate.ipv4; | |
523 | break; | |
524 | case AFI_IP6: | |
525 | p.family = AF_INET6; | |
526 | p.prefixlen = IPV6_MAX_PREFIXLEN; | |
527 | p.u.prefix6 = nexthop->gate.ipv6; | |
528 | break; | |
529 | default: | |
530 | assert(afi != AFI_IP && afi != AFI_IP6); | |
531 | break; | |
532 | } | |
533 | /* Lookup table. */ | |
534 | table = zebra_vrf_table(afi, SAFI_UNICAST, nexthop->vrf_id); | |
5a0bdc78 PG |
535 | /* get zvrf */ |
536 | zvrf = zebra_vrf_lookup_by_id(nexthop->vrf_id); | |
537 | if (!table || !zvrf) { | |
ad28e79a SW |
538 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) |
539 | zlog_debug("\t%s: Table not found", | |
540 | __PRETTY_FUNCTION__); | |
541 | return 0; | |
542 | } | |
543 | ||
544 | rn = route_node_match(table, (struct prefix *)&p); | |
545 | while (rn) { | |
546 | route_unlock_node(rn); | |
547 | ||
548 | /* Lookup should halt if we've matched against ourselves ('top', | |
549 | * if specified) - i.e., we cannot have a nexthop NH1 is | |
550 | * resolved by a route NH1. The exception is if the route is a | |
551 | * host route. | |
552 | */ | |
553 | if (top && rn == top) | |
554 | if (((afi == AFI_IP) && (rn->p.prefixlen != 32)) | |
555 | || ((afi == AFI_IP6) && (rn->p.prefixlen != 128))) { | |
556 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) | |
557 | zlog_debug( | |
558 | "\t%s: Matched against ourself and prefix length is not max bit length", | |
559 | __PRETTY_FUNCTION__); | |
560 | return 0; | |
561 | } | |
562 | ||
563 | /* Pick up selected route. */ | |
564 | /* However, do not resolve over default route unless explicitly | |
565 | * allowed. */ | |
566 | if (is_default_prefix(&rn->p) | |
5a0bdc78 | 567 | && !rnh_resolve_via_default(zvrf, p.family)) { |
ad28e79a SW |
568 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) |
569 | zlog_debug( | |
570 | "\t:%s: Resolved against default route", | |
571 | __PRETTY_FUNCTION__); | |
572 | return 0; | |
573 | } | |
574 | ||
575 | dest = rib_dest_from_rnode(rn); | |
576 | if (dest && dest->selected_fib | |
577 | && !CHECK_FLAG(dest->selected_fib->status, | |
578 | ROUTE_ENTRY_REMOVED) | |
579 | && dest->selected_fib->type != ZEBRA_ROUTE_TABLE) | |
580 | match = dest->selected_fib; | |
581 | ||
582 | /* If there is no selected route or matched route is EGP, go up | |
583 | tree. */ | |
584 | if (!match) { | |
585 | do { | |
586 | rn = rn->parent; | |
587 | } while (rn && rn->info == NULL); | |
588 | if (rn) | |
589 | route_lock_node(rn); | |
590 | ||
591 | continue; | |
592 | } | |
593 | ||
594 | if (match->type == ZEBRA_ROUTE_CONNECT) { | |
595 | /* Directly point connected route. */ | |
6b468511 | 596 | newhop = match->ng->nexthop; |
ad28e79a SW |
597 | if (newhop) { |
598 | if (nexthop->type == NEXTHOP_TYPE_IPV4 | |
599 | || nexthop->type == NEXTHOP_TYPE_IPV6) | |
600 | nexthop->ifindex = newhop->ifindex; | |
601 | } | |
602 | return 1; | |
603 | } else if (CHECK_FLAG(re->flags, ZEBRA_FLAG_ALLOW_RECURSION)) { | |
604 | resolved = 0; | |
6b468511 | 605 | for (ALL_NEXTHOPS_PTR(match->ng, newhop)) { |
ad28e79a SW |
606 | if (!CHECK_FLAG(match->status, |
607 | ROUTE_ENTRY_INSTALLED)) | |
608 | continue; | |
6913cb1b | 609 | if (!nexthop_valid_resolve(nexthop, newhop)) |
ad28e79a SW |
610 | continue; |
611 | ||
612 | SET_FLAG(nexthop->flags, | |
613 | NEXTHOP_FLAG_RECURSIVE); | |
ad28e79a SW |
614 | nexthop_set_resolved(afi, newhop, nexthop); |
615 | resolved = 1; | |
616 | } | |
617 | if (resolved) | |
618 | re->nexthop_mtu = match->mtu; | |
619 | if (!resolved && IS_ZEBRA_DEBUG_RIB_DETAILED) | |
620 | zlog_debug("\t%s: Recursion failed to find", | |
621 | __PRETTY_FUNCTION__); | |
622 | return resolved; | |
623 | } else if (re->type == ZEBRA_ROUTE_STATIC) { | |
624 | resolved = 0; | |
6b468511 | 625 | for (ALL_NEXTHOPS_PTR(match->ng, newhop)) { |
ad28e79a SW |
626 | if (!CHECK_FLAG(match->status, |
627 | ROUTE_ENTRY_INSTALLED)) | |
628 | continue; | |
6913cb1b | 629 | if (!nexthop_valid_resolve(nexthop, newhop)) |
ad28e79a SW |
630 | continue; |
631 | ||
632 | SET_FLAG(nexthop->flags, | |
633 | NEXTHOP_FLAG_RECURSIVE); | |
634 | nexthop_set_resolved(afi, newhop, nexthop); | |
635 | resolved = 1; | |
636 | } | |
637 | if (resolved) | |
638 | re->nexthop_mtu = match->mtu; | |
639 | ||
640 | if (!resolved && IS_ZEBRA_DEBUG_RIB_DETAILED) | |
641 | zlog_debug( | |
642 | "\t%s: Static route unable to resolve", | |
643 | __PRETTY_FUNCTION__); | |
644 | return resolved; | |
645 | } else { | |
646 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) { | |
647 | zlog_debug( | |
648 | "\t%s: Route Type %s has not turned on recursion", | |
649 | __PRETTY_FUNCTION__, | |
650 | zebra_route_string(re->type)); | |
651 | if (re->type == ZEBRA_ROUTE_BGP | |
652 | && !CHECK_FLAG(re->flags, ZEBRA_FLAG_IBGP)) | |
653 | zlog_debug( | |
654 | "\tEBGP: see \"disable-ebgp-connected-route-check\" or \"disable-connected-check\""); | |
655 | } | |
656 | return 0; | |
657 | } | |
658 | } | |
659 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) | |
660 | zlog_debug("\t%s: Nexthop did not lookup in table", | |
661 | __PRETTY_FUNCTION__); | |
662 | return 0; | |
663 | } | |
664 | ||
665 | /* This function verifies reachability of one given nexthop, which can be | |
666 | * numbered or unnumbered, IPv4 or IPv6. The result is unconditionally stored | |
667 | * in nexthop->flags field. The nexthop->ifindex will be updated | |
668 | * appropriately as well. An existing route map can turn | |
669 | * (otherwise active) nexthop into inactive, but not vice versa. | |
670 | * | |
671 | * The return value is the final value of 'ACTIVE' flag. | |
672 | */ | |
673 | static unsigned nexthop_active_check(struct route_node *rn, | |
674 | struct route_entry *re, | |
675 | struct nexthop *nexthop) | |
676 | { | |
677 | struct interface *ifp; | |
b68885f9 | 678 | route_map_result_t ret = RMAP_PERMITMATCH; |
ad28e79a SW |
679 | int family; |
680 | char buf[SRCDEST2STR_BUFFER]; | |
681 | const struct prefix *p, *src_p; | |
682 | struct zebra_vrf *zvrf; | |
683 | ||
684 | srcdest_rnode_prefixes(rn, &p, &src_p); | |
685 | ||
686 | if (rn->p.family == AF_INET) | |
687 | family = AFI_IP; | |
688 | else if (rn->p.family == AF_INET6) | |
689 | family = AFI_IP6; | |
690 | else | |
691 | family = 0; | |
692 | switch (nexthop->type) { | |
693 | case NEXTHOP_TYPE_IFINDEX: | |
694 | ifp = if_lookup_by_index(nexthop->ifindex, nexthop->vrf_id); | |
695 | if (ifp && if_is_operative(ifp)) | |
696 | SET_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE); | |
697 | else | |
698 | UNSET_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE); | |
699 | break; | |
700 | case NEXTHOP_TYPE_IPV4: | |
701 | case NEXTHOP_TYPE_IPV4_IFINDEX: | |
702 | family = AFI_IP; | |
703 | if (nexthop_active(AFI_IP, re, nexthop, rn)) | |
704 | SET_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE); | |
705 | else | |
706 | UNSET_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE); | |
707 | break; | |
708 | case NEXTHOP_TYPE_IPV6: | |
709 | family = AFI_IP6; | |
710 | if (nexthop_active(AFI_IP6, re, nexthop, rn)) | |
711 | SET_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE); | |
712 | else | |
713 | UNSET_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE); | |
714 | break; | |
715 | case NEXTHOP_TYPE_IPV6_IFINDEX: | |
716 | /* RFC 5549, v4 prefix with v6 NH */ | |
717 | if (rn->p.family != AF_INET) | |
718 | family = AFI_IP6; | |
719 | if (IN6_IS_ADDR_LINKLOCAL(&nexthop->gate.ipv6)) { | |
720 | ifp = if_lookup_by_index(nexthop->ifindex, | |
721 | nexthop->vrf_id); | |
722 | if (ifp && if_is_operative(ifp)) | |
723 | SET_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE); | |
724 | else | |
725 | UNSET_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE); | |
726 | } else { | |
727 | if (nexthop_active(AFI_IP6, re, nexthop, rn)) | |
728 | SET_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE); | |
729 | else | |
730 | UNSET_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE); | |
731 | } | |
732 | break; | |
733 | case NEXTHOP_TYPE_BLACKHOLE: | |
734 | SET_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE); | |
735 | break; | |
736 | default: | |
737 | break; | |
738 | } | |
739 | if (!CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE)) { | |
740 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) | |
741 | zlog_debug("\t%s: Unable to find a active nexthop", | |
742 | __PRETTY_FUNCTION__); | |
743 | return 0; | |
744 | } | |
745 | ||
746 | /* XXX: What exactly do those checks do? Do we support | |
747 | * e.g. IPv4 routes with IPv6 nexthops or vice versa? | |
748 | */ | |
749 | if (RIB_SYSTEM_ROUTE(re) || (family == AFI_IP && p->family != AF_INET) | |
750 | || (family == AFI_IP6 && p->family != AF_INET6)) | |
751 | return CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE); | |
752 | ||
753 | /* The original code didn't determine the family correctly | |
754 | * e.g. for NEXTHOP_TYPE_IFINDEX. Retrieve the correct afi | |
755 | * from the rib_table_info in those cases. | |
756 | * Possibly it may be better to use only the rib_table_info | |
757 | * in every case. | |
758 | */ | |
759 | if (!family) { | |
760 | rib_table_info_t *info; | |
761 | ||
762 | info = srcdest_rnode_table_info(rn); | |
763 | family = info->afi; | |
764 | } | |
765 | ||
766 | memset(&nexthop->rmap_src.ipv6, 0, sizeof(union g_addr)); | |
767 | ||
768 | zvrf = zebra_vrf_lookup_by_id(nexthop->vrf_id); | |
769 | if (!zvrf) { | |
770 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) | |
771 | zlog_debug("\t%s: zvrf is NULL", __PRETTY_FUNCTION__); | |
772 | return CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE); | |
773 | } | |
774 | ||
775 | /* It'll get set if required inside */ | |
776 | ret = zebra_route_map_check(family, re->type, re->instance, p, nexthop, | |
777 | zvrf, re->tag); | |
778 | if (ret == RMAP_DENYMATCH) { | |
779 | if (IS_ZEBRA_DEBUG_RIB) { | |
780 | srcdest_rnode2str(rn, buf, sizeof(buf)); | |
781 | zlog_debug( | |
782 | "%u:%s: Filtering out with NH out %s due to route map", | |
783 | re->vrf_id, buf, | |
784 | ifindex2ifname(nexthop->ifindex, | |
785 | nexthop->vrf_id)); | |
786 | } | |
787 | UNSET_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE); | |
788 | } | |
789 | return CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE); | |
790 | } | |
791 | ||
792 | /* | |
793 | * Iterate over all nexthops of the given RIB entry and refresh their | |
9a0d4dd3 DS |
794 | * ACTIVE flag. If any nexthop is found to toggle the ACTIVE flag, |
795 | * the whole re structure is flagged with ROUTE_ENTRY_CHANGED. | |
ad28e79a SW |
796 | * |
797 | * Return value is the new number of active nexthops. | |
798 | */ | |
799 | int nexthop_active_update(struct route_node *rn, struct route_entry *re) | |
800 | { | |
801 | struct nexthop *nexthop; | |
802 | union g_addr prev_src; | |
803 | unsigned int prev_active, new_active; | |
804 | ifindex_t prev_index; | |
9a0d4dd3 | 805 | uint8_t curr_active = 0; |
ad28e79a | 806 | |
ad28e79a SW |
807 | UNSET_FLAG(re->status, ROUTE_ENTRY_CHANGED); |
808 | ||
6b468511 | 809 | for (nexthop = re->ng->nexthop; nexthop; nexthop = nexthop->next) { |
ad28e79a SW |
810 | /* No protocol daemon provides src and so we're skipping |
811 | * tracking it */ | |
812 | prev_src = nexthop->rmap_src; | |
813 | prev_active = CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE); | |
814 | prev_index = nexthop->ifindex; | |
815 | /* | |
816 | * We need to respect the multipath_num here | |
817 | * as that what we should be able to install from | |
818 | * a multipath perpsective should not be a data plane | |
819 | * decision point. | |
820 | */ | |
821 | new_active = nexthop_active_check(rn, re, nexthop); | |
822 | if (new_active | |
9a0d4dd3 DS |
823 | && nexthop_group_active_nexthop_num(re->ng) |
824 | >= zrouter.multipath_num) { | |
ad28e79a SW |
825 | UNSET_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE); |
826 | new_active = 0; | |
827 | } | |
9a0d4dd3 | 828 | |
ad28e79a | 829 | if (new_active) |
9a0d4dd3 DS |
830 | curr_active++; |
831 | ||
ad28e79a SW |
832 | /* Don't allow src setting on IPv6 addr for now */ |
833 | if (prev_active != new_active || prev_index != nexthop->ifindex | |
834 | || ((nexthop->type >= NEXTHOP_TYPE_IFINDEX | |
835 | && nexthop->type < NEXTHOP_TYPE_IPV6) | |
836 | && prev_src.ipv4.s_addr | |
837 | != nexthop->rmap_src.ipv4.s_addr) | |
838 | || ((nexthop->type >= NEXTHOP_TYPE_IPV6 | |
839 | && nexthop->type < NEXTHOP_TYPE_BLACKHOLE) | |
840 | && !(IPV6_ADDR_SAME(&prev_src.ipv6, | |
841 | &nexthop->rmap_src.ipv6))) | |
42fc558e | 842 | || CHECK_FLAG(re->status, ROUTE_ENTRY_LABELS_CHANGED)) |
ad28e79a | 843 | SET_FLAG(re->status, ROUTE_ENTRY_CHANGED); |
ad28e79a SW |
844 | } |
845 | ||
9a0d4dd3 | 846 | return curr_active; |
ad28e79a | 847 | } |