]>
Commit | Line | Data |
---|---|---|
8ef0791c | 1 | /* |
165b5fff JB |
2 | * BGP Multipath |
3 | * Copyright (C) 2010 Google Inc. | |
4 | * | |
5 | * This file is part of Quagga | |
6 | * | |
7 | * Quagga is free software; you can redistribute it and/or modify it | |
8 | * under the terms of the GNU General Public License as published by the | |
9 | * Free Software Foundation; either version 2, or (at your option) any | |
10 | * later version. | |
11 | * | |
12 | * Quagga is distributed in the hope that it will be useful, but | |
13 | * WITHOUT ANY WARRANTY; without even the implied warranty of | |
14 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU | |
15 | * General Public License for more details. | |
16 | * | |
17 | * You should have received a copy of the GNU General Public License | |
18 | * along with Quagga; see the file COPYING. If not, write to the Free | |
19 | * Software Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA | |
20 | * 02111-1307, USA. | |
21 | */ | |
22 | ||
23 | #include <zebra.h> | |
24 | ||
25 | #include "command.h" | |
96450faf JB |
26 | #include "prefix.h" |
27 | #include "linklist.h" | |
28 | #include "sockunion.h" | |
de8d5dff | 29 | #include "memory.h" |
3f9c7369 | 30 | #include "queue.h" |
039f3a34 | 31 | #include "filter.h" |
165b5fff JB |
32 | |
33 | #include "bgpd/bgpd.h" | |
96450faf JB |
34 | #include "bgpd/bgp_table.h" |
35 | #include "bgpd/bgp_route.h" | |
36 | #include "bgpd/bgp_attr.h" | |
de8d5dff | 37 | #include "bgpd/bgp_debug.h" |
0b597ef0 JB |
38 | #include "bgpd/bgp_aspath.h" |
39 | #include "bgpd/bgp_community.h" | |
40 | #include "bgpd/bgp_ecommunity.h" | |
57d187bc | 41 | #include "bgpd/bgp_lcommunity.h" |
165b5fff JB |
42 | #include "bgpd/bgp_mpath.h" |
43 | ||
44 | /* | |
45 | * bgp_maximum_paths_set | |
46 | * | |
47 | * Record maximum-paths configuration for BGP instance | |
48 | */ | |
49 | int | |
50 | bgp_maximum_paths_set (struct bgp *bgp, afi_t afi, safi_t safi, | |
5e242b0d | 51 | int peertype, u_int16_t maxpaths, u_int16_t options) |
165b5fff JB |
52 | { |
53 | if (!bgp || (afi >= AFI_MAX) || (safi >= SAFI_MAX)) | |
54 | return -1; | |
55 | ||
56 | switch (peertype) | |
57 | { | |
58 | case BGP_PEER_IBGP: | |
59 | bgp->maxpaths[afi][safi].maxpaths_ibgp = maxpaths; | |
5e242b0d | 60 | bgp->maxpaths[afi][safi].ibgp_flags |= options; |
165b5fff JB |
61 | break; |
62 | case BGP_PEER_EBGP: | |
63 | bgp->maxpaths[afi][safi].maxpaths_ebgp = maxpaths; | |
64 | break; | |
65 | default: | |
66 | return -1; | |
67 | } | |
68 | ||
69 | return 0; | |
70 | } | |
71 | ||
72 | /* | |
73 | * bgp_maximum_paths_unset | |
74 | * | |
75 | * Remove maximum-paths configuration from BGP instance | |
76 | */ | |
77 | int | |
78 | bgp_maximum_paths_unset (struct bgp *bgp, afi_t afi, safi_t safi, | |
79 | int peertype) | |
80 | { | |
81 | if (!bgp || (afi >= AFI_MAX) || (safi >= SAFI_MAX)) | |
82 | return -1; | |
83 | ||
84 | switch (peertype) | |
85 | { | |
86 | case BGP_PEER_IBGP: | |
d5b77cc2 | 87 | bgp->maxpaths[afi][safi].maxpaths_ibgp = MULTIPATH_NUM; |
5e242b0d | 88 | bgp->maxpaths[afi][safi].ibgp_flags = 0; |
165b5fff JB |
89 | break; |
90 | case BGP_PEER_EBGP: | |
d5b77cc2 | 91 | bgp->maxpaths[afi][safi].maxpaths_ebgp = MULTIPATH_NUM; |
165b5fff JB |
92 | break; |
93 | default: | |
94 | return -1; | |
95 | } | |
96 | ||
97 | return 0; | |
98 | } | |
96450faf JB |
99 | |
100 | /* | |
101 | * bgp_info_nexthop_cmp | |
102 | * | |
103 | * Compare the nexthops of two paths. Return value is less than, equal to, | |
104 | * or greater than zero if bi1 is respectively less than, equal to, | |
105 | * or greater than bi2. | |
106 | */ | |
7dc9d4e4 | 107 | int |
96450faf JB |
108 | bgp_info_nexthop_cmp (struct bgp_info *bi1, struct bgp_info *bi2) |
109 | { | |
110 | struct attr_extra *ae1, *ae2; | |
111 | int compare; | |
112 | ||
0b597ef0 JB |
113 | ae1 = bi1->attr->extra; |
114 | ae2 = bi2->attr->extra; | |
96450faf JB |
115 | |
116 | compare = IPV4_ADDR_CMP (&bi1->attr->nexthop, &bi2->attr->nexthop); | |
e328382f | 117 | if (!compare && ae1 && ae2) |
96450faf | 118 | { |
e328382f | 119 | if (ae1->mp_nexthop_len == ae2->mp_nexthop_len) |
96450faf | 120 | { |
e328382f DS |
121 | switch (ae1->mp_nexthop_len) |
122 | { | |
801a9bcc DS |
123 | case BGP_ATTR_NHLEN_IPV4: |
124 | case BGP_ATTR_NHLEN_VPNV4: | |
e328382f DS |
125 | compare = IPV4_ADDR_CMP (&ae1->mp_nexthop_global_in, |
126 | &ae2->mp_nexthop_global_in); | |
127 | break; | |
801a9bcc | 128 | case BGP_ATTR_NHLEN_IPV6_GLOBAL: |
af37568a | 129 | case BGP_ATTR_NHLEN_VPNV6_GLOBAL: |
e328382f DS |
130 | compare = IPV6_ADDR_CMP (&ae1->mp_nexthop_global, |
131 | &ae2->mp_nexthop_global); | |
132 | break; | |
801a9bcc | 133 | case BGP_ATTR_NHLEN_IPV6_GLOBAL_AND_LL: |
e328382f DS |
134 | compare = IPV6_ADDR_CMP (&ae1->mp_nexthop_global, |
135 | &ae2->mp_nexthop_global); | |
136 | if (!compare) | |
137 | compare = IPV6_ADDR_CMP (&ae1->mp_nexthop_local, | |
138 | &ae2->mp_nexthop_local); | |
139 | break; | |
e328382f DS |
140 | } |
141 | } | |
142 | ||
e328382f DS |
143 | /* This can happen if one IPv6 peer sends you global and link-local |
144 | * nexthops but another IPv6 peer only sends you global | |
145 | */ | |
801a9bcc DS |
146 | else if (ae1->mp_nexthop_len == BGP_ATTR_NHLEN_IPV6_GLOBAL || |
147 | ae1->mp_nexthop_len == BGP_ATTR_NHLEN_IPV6_GLOBAL_AND_LL) | |
e328382f | 148 | { |
96450faf JB |
149 | compare = IPV6_ADDR_CMP (&ae1->mp_nexthop_global, |
150 | &ae2->mp_nexthop_global); | |
151 | if (!compare) | |
e328382f DS |
152 | { |
153 | if (ae1->mp_nexthop_len < ae2->mp_nexthop_len) | |
154 | compare = -1; | |
155 | else | |
156 | compare = 1; | |
157 | } | |
96450faf JB |
158 | } |
159 | } | |
160 | ||
161 | return compare; | |
162 | } | |
163 | ||
164 | /* | |
165 | * bgp_info_mpath_cmp | |
166 | * | |
167 | * This function determines our multipath list ordering. By ordering | |
168 | * the list we can deterministically select which paths are included | |
169 | * in the multipath set. The ordering also helps in detecting changes | |
170 | * in the multipath selection so we can detect whether to send an | |
171 | * update to zebra. | |
172 | * | |
173 | * The order of paths is determined first by received nexthop, and then | |
174 | * by peer address if the nexthops are the same. | |
175 | */ | |
176 | static int | |
177 | bgp_info_mpath_cmp (void *val1, void *val2) | |
178 | { | |
179 | struct bgp_info *bi1, *bi2; | |
180 | int compare; | |
181 | ||
182 | bi1 = val1; | |
183 | bi2 = val2; | |
184 | ||
185 | compare = bgp_info_nexthop_cmp (bi1, bi2); | |
186 | ||
187 | if (!compare) | |
a3cb01d8 | 188 | { |
189 | if (!bi1->peer->su_remote && !bi2->peer->su_remote) | |
190 | compare = 0; | |
191 | else if (!bi1->peer->su_remote) | |
192 | compare = 1; | |
193 | else if (!bi2->peer->su_remote) | |
194 | compare = -1; | |
195 | else | |
196 | compare = sockunion_cmp (bi1->peer->su_remote, bi2->peer->su_remote); | |
197 | } | |
96450faf JB |
198 | |
199 | return compare; | |
200 | } | |
201 | ||
202 | /* | |
203 | * bgp_mp_list_init | |
204 | * | |
205 | * Initialize the mp_list, which holds the list of multipaths | |
206 | * selected by bgp_best_selection | |
207 | */ | |
208 | void | |
209 | bgp_mp_list_init (struct list *mp_list) | |
210 | { | |
211 | assert (mp_list); | |
212 | memset (mp_list, 0, sizeof (struct list)); | |
213 | mp_list->cmp = bgp_info_mpath_cmp; | |
214 | } | |
215 | ||
216 | /* | |
217 | * bgp_mp_list_clear | |
218 | * | |
219 | * Clears all entries out of the mp_list | |
220 | */ | |
221 | void | |
222 | bgp_mp_list_clear (struct list *mp_list) | |
223 | { | |
224 | assert (mp_list); | |
225 | list_delete_all_node (mp_list); | |
226 | } | |
227 | ||
228 | /* | |
229 | * bgp_mp_list_add | |
230 | * | |
231 | * Adds a multipath entry to the mp_list | |
232 | */ | |
233 | void | |
234 | bgp_mp_list_add (struct list *mp_list, struct bgp_info *mpinfo) | |
235 | { | |
236 | assert (mp_list && mpinfo); | |
237 | listnode_add_sort (mp_list, mpinfo); | |
238 | } | |
de8d5dff JB |
239 | |
240 | /* | |
241 | * bgp_info_mpath_new | |
242 | * | |
243 | * Allocate and zero memory for a new bgp_info_mpath element | |
244 | */ | |
245 | static struct bgp_info_mpath * | |
246 | bgp_info_mpath_new (void) | |
247 | { | |
248 | struct bgp_info_mpath *new_mpath; | |
249 | new_mpath = XCALLOC (MTYPE_BGP_MPATH_INFO, sizeof (struct bgp_info_mpath)); | |
250 | return new_mpath; | |
251 | } | |
252 | ||
253 | /* | |
254 | * bgp_info_mpath_free | |
255 | * | |
256 | * Release resources for a bgp_info_mpath element and zero out pointer | |
257 | */ | |
258 | void | |
259 | bgp_info_mpath_free (struct bgp_info_mpath **mpath) | |
260 | { | |
261 | if (mpath && *mpath) | |
262 | { | |
0b597ef0 | 263 | if ((*mpath)->mp_attr) |
fac3c245 | 264 | bgp_attr_unintern (&(*mpath)->mp_attr); |
de8d5dff JB |
265 | XFREE (MTYPE_BGP_MPATH_INFO, *mpath); |
266 | *mpath = NULL; | |
267 | } | |
268 | } | |
269 | ||
270 | /* | |
271 | * bgp_info_mpath_get | |
272 | * | |
273 | * Fetch the mpath element for the given bgp_info. Used for | |
274 | * doing lazy allocation. | |
275 | */ | |
276 | static struct bgp_info_mpath * | |
277 | bgp_info_mpath_get (struct bgp_info *binfo) | |
278 | { | |
279 | struct bgp_info_mpath *mpath; | |
280 | if (!binfo->mpath) | |
281 | { | |
282 | mpath = bgp_info_mpath_new(); | |
283 | if (!mpath) | |
284 | return NULL; | |
285 | binfo->mpath = mpath; | |
286 | mpath->mp_info = binfo; | |
287 | } | |
288 | return binfo->mpath; | |
289 | } | |
290 | ||
291 | /* | |
292 | * bgp_info_mpath_enqueue | |
293 | * | |
294 | * Enqueue a path onto the multipath list given the previous multipath | |
295 | * list entry | |
296 | */ | |
297 | static void | |
298 | bgp_info_mpath_enqueue (struct bgp_info *prev_info, struct bgp_info *binfo) | |
299 | { | |
300 | struct bgp_info_mpath *prev, *mpath; | |
301 | ||
302 | prev = bgp_info_mpath_get (prev_info); | |
303 | mpath = bgp_info_mpath_get (binfo); | |
304 | if (!prev || !mpath) | |
305 | return; | |
306 | ||
307 | mpath->mp_next = prev->mp_next; | |
308 | mpath->mp_prev = prev; | |
309 | if (prev->mp_next) | |
310 | prev->mp_next->mp_prev = mpath; | |
311 | prev->mp_next = mpath; | |
312 | ||
313 | SET_FLAG (binfo->flags, BGP_INFO_MULTIPATH); | |
314 | } | |
315 | ||
316 | /* | |
317 | * bgp_info_mpath_dequeue | |
318 | * | |
319 | * Remove a path from the multipath list | |
320 | */ | |
321 | void | |
322 | bgp_info_mpath_dequeue (struct bgp_info *binfo) | |
323 | { | |
324 | struct bgp_info_mpath *mpath = binfo->mpath; | |
325 | if (!mpath) | |
326 | return; | |
327 | if (mpath->mp_prev) | |
328 | mpath->mp_prev->mp_next = mpath->mp_next; | |
329 | if (mpath->mp_next) | |
330 | mpath->mp_next->mp_prev = mpath->mp_prev; | |
331 | mpath->mp_next = mpath->mp_prev = NULL; | |
332 | UNSET_FLAG (binfo->flags, BGP_INFO_MULTIPATH); | |
333 | } | |
334 | ||
335 | /* | |
336 | * bgp_info_mpath_next | |
337 | * | |
338 | * Given a bgp_info, return the next multipath entry | |
339 | */ | |
340 | struct bgp_info * | |
341 | bgp_info_mpath_next (struct bgp_info *binfo) | |
342 | { | |
343 | if (!binfo->mpath || !binfo->mpath->mp_next) | |
344 | return NULL; | |
345 | return binfo->mpath->mp_next->mp_info; | |
346 | } | |
347 | ||
348 | /* | |
349 | * bgp_info_mpath_first | |
350 | * | |
351 | * Given bestpath bgp_info, return the first multipath entry. | |
352 | */ | |
353 | struct bgp_info * | |
354 | bgp_info_mpath_first (struct bgp_info *binfo) | |
355 | { | |
356 | return bgp_info_mpath_next (binfo); | |
357 | } | |
358 | ||
359 | /* | |
360 | * bgp_info_mpath_count | |
361 | * | |
362 | * Given the bestpath bgp_info, return the number of multipath entries | |
363 | */ | |
364 | u_int32_t | |
365 | bgp_info_mpath_count (struct bgp_info *binfo) | |
366 | { | |
367 | if (!binfo->mpath) | |
368 | return 0; | |
369 | return binfo->mpath->mp_count; | |
370 | } | |
371 | ||
372 | /* | |
373 | * bgp_info_mpath_count_set | |
374 | * | |
375 | * Sets the count of multipaths into bestpath's mpath element | |
376 | */ | |
377 | static void | |
378 | bgp_info_mpath_count_set (struct bgp_info *binfo, u_int32_t count) | |
379 | { | |
380 | struct bgp_info_mpath *mpath; | |
381 | if (!count && !binfo->mpath) | |
382 | return; | |
383 | mpath = bgp_info_mpath_get (binfo); | |
384 | if (!mpath) | |
385 | return; | |
386 | mpath->mp_count = count; | |
387 | } | |
388 | ||
0b597ef0 JB |
389 | /* |
390 | * bgp_info_mpath_attr | |
391 | * | |
392 | * Given bestpath bgp_info, return aggregated attribute set used | |
393 | * for advertising the multipath route | |
394 | */ | |
395 | struct attr * | |
396 | bgp_info_mpath_attr (struct bgp_info *binfo) | |
397 | { | |
398 | if (!binfo->mpath) | |
399 | return NULL; | |
400 | return binfo->mpath->mp_attr; | |
401 | } | |
402 | ||
403 | /* | |
404 | * bgp_info_mpath_attr_set | |
405 | * | |
406 | * Sets the aggregated attribute into bestpath's mpath element | |
407 | */ | |
408 | static void | |
409 | bgp_info_mpath_attr_set (struct bgp_info *binfo, struct attr *attr) | |
410 | { | |
411 | struct bgp_info_mpath *mpath; | |
412 | if (!attr && !binfo->mpath) | |
413 | return; | |
414 | mpath = bgp_info_mpath_get (binfo); | |
415 | if (!mpath) | |
416 | return; | |
417 | mpath->mp_attr = attr; | |
418 | } | |
419 | ||
de8d5dff JB |
420 | /* |
421 | * bgp_info_mpath_update | |
422 | * | |
423 | * Compare and sync up the multipath list with the mp_list generated by | |
424 | * bgp_best_selection | |
425 | */ | |
426 | void | |
427 | bgp_info_mpath_update (struct bgp_node *rn, struct bgp_info *new_best, | |
428 | struct bgp_info *old_best, struct list *mp_list, | |
429 | struct bgp_maxpaths_cfg *mpath_cfg) | |
430 | { | |
431 | u_int16_t maxpaths, mpath_count, old_mpath_count; | |
432 | struct listnode *mp_node, *mp_next_node; | |
433 | struct bgp_info *cur_mpath, *new_mpath, *next_mpath, *prev_mpath; | |
434 | int mpath_changed, debug; | |
4690c7d7 | 435 | char pfx_buf[PREFIX2STR_BUFFER], nh_buf[2][INET6_ADDRSTRLEN]; |
2ec1e66f | 436 | char path_buf[PATH_ADDPATH_STR_BUFFER]; |
de8d5dff JB |
437 | |
438 | mpath_changed = 0; | |
d5b77cc2 | 439 | maxpaths = MULTIPATH_NUM; |
de8d5dff JB |
440 | mpath_count = 0; |
441 | cur_mpath = NULL; | |
442 | old_mpath_count = 0; | |
443 | prev_mpath = new_best; | |
444 | mp_node = listhead (mp_list); | |
9fbdd100 | 445 | debug = bgp_debug_bestpath(&rn->p); |
de8d5dff JB |
446 | |
447 | if (debug) | |
448 | prefix2str (&rn->p, pfx_buf, sizeof (pfx_buf)); | |
449 | ||
450 | if (new_best) | |
451 | { | |
452 | mpath_count++; | |
453 | if (new_best != old_best) | |
454 | bgp_info_mpath_dequeue (new_best); | |
6d85b15b | 455 | maxpaths = (new_best->peer->sort == BGP_PEER_IBGP) ? |
de8d5dff JB |
456 | mpath_cfg->maxpaths_ibgp : mpath_cfg->maxpaths_ebgp; |
457 | } | |
458 | ||
459 | if (old_best) | |
460 | { | |
461 | cur_mpath = bgp_info_mpath_first (old_best); | |
462 | old_mpath_count = bgp_info_mpath_count (old_best); | |
463 | bgp_info_mpath_count_set (old_best, 0); | |
464 | bgp_info_mpath_dequeue (old_best); | |
465 | } | |
466 | ||
a6086ad4 | 467 | if (debug) |
468 | zlog_debug("%s: starting mpath update, newbest %s num candidates %d old-mpath-count %d", | |
469 | pfx_buf, new_best ? new_best->peer->host : "NONE", | |
470 | listcount (mp_list), old_mpath_count); | |
471 | ||
de8d5dff JB |
472 | /* |
473 | * We perform an ordered walk through both lists in parallel. | |
474 | * The reason for the ordered walk is that if there are paths | |
475 | * that were previously multipaths and are still multipaths, the walk | |
476 | * should encounter them in both lists at the same time. Otherwise | |
477 | * there will be paths that are in one list or another, and we | |
478 | * will deal with these separately. | |
479 | * | |
480 | * Note that new_best might be somewhere in the mp_list, so we need | |
481 | * to skip over it | |
482 | */ | |
483 | while (mp_node || cur_mpath) | |
484 | { | |
a6086ad4 | 485 | struct bgp_info *tmp_info; |
486 | ||
de8d5dff JB |
487 | /* |
488 | * We can bail out of this loop if all existing paths on the | |
489 | * multipath list have been visited (for cleanup purposes) and | |
490 | * the maxpath requirement is fulfulled | |
491 | */ | |
492 | if (!cur_mpath && (mpath_count >= maxpaths)) | |
493 | break; | |
494 | ||
495 | mp_next_node = mp_node ? listnextnode (mp_node) : NULL; | |
496 | next_mpath = cur_mpath ? bgp_info_mpath_next (cur_mpath) : NULL; | |
a6086ad4 | 497 | tmp_info = mp_node ? listgetdata (mp_node) : NULL; |
498 | ||
499 | if (debug) | |
500 | zlog_debug("%s: comparing candidate %s with existing mpath %s", | |
501 | pfx_buf, tmp_info ? tmp_info->peer->host : "NONE", | |
502 | cur_mpath ? cur_mpath->peer->host : "NONE"); | |
de8d5dff JB |
503 | |
504 | /* | |
505 | * If equal, the path was a multipath and is still a multipath. | |
506 | * Insert onto new multipath list if maxpaths allows. | |
507 | */ | |
508 | if (mp_node && (listgetdata (mp_node) == cur_mpath)) | |
509 | { | |
510 | list_delete_node (mp_list, mp_node); | |
511 | bgp_info_mpath_dequeue (cur_mpath); | |
512 | if ((mpath_count < maxpaths) && | |
513 | bgp_info_nexthop_cmp (prev_mpath, cur_mpath)) | |
514 | { | |
515 | bgp_info_mpath_enqueue (prev_mpath, cur_mpath); | |
516 | prev_mpath = cur_mpath; | |
517 | mpath_count++; | |
a6086ad4 | 518 | if (debug) |
519 | { | |
520 | bgp_info_path_with_addpath_rx_str(cur_mpath, path_buf); | |
521 | zlog_debug("%s: %s is still multipath, cur count %d", | |
522 | pfx_buf, path_buf, mpath_count); | |
523 | } | |
de8d5dff JB |
524 | } |
525 | else | |
526 | { | |
527 | mpath_changed = 1; | |
528 | if (debug) | |
2ec1e66f DW |
529 | { |
530 | bgp_info_path_with_addpath_rx_str(cur_mpath, path_buf); | |
a6086ad4 | 531 | zlog_debug ("%s: remove mpath %s nexthop %s, cur count %d", |
532 | pfx_buf, path_buf, | |
2ec1e66f DW |
533 | inet_ntop (AF_INET, &cur_mpath->attr->nexthop, |
534 | nh_buf[0], sizeof (nh_buf[0])), | |
a6086ad4 | 535 | mpath_count); |
2ec1e66f | 536 | } |
de8d5dff JB |
537 | } |
538 | mp_node = mp_next_node; | |
539 | cur_mpath = next_mpath; | |
540 | continue; | |
541 | } | |
542 | ||
543 | if (cur_mpath && (!mp_node || | |
544 | (bgp_info_mpath_cmp (cur_mpath, | |
545 | listgetdata (mp_node)) < 0))) | |
546 | { | |
547 | /* | |
548 | * If here, we have an old multipath and either the mp_list | |
549 | * is finished or the next mp_node points to a later | |
550 | * multipath, so we need to purge this path from the | |
551 | * multipath list | |
552 | */ | |
553 | bgp_info_mpath_dequeue (cur_mpath); | |
554 | mpath_changed = 1; | |
555 | if (debug) | |
2ec1e66f DW |
556 | { |
557 | bgp_info_path_with_addpath_rx_str(cur_mpath, path_buf); | |
a6086ad4 | 558 | zlog_debug ("%s: remove mpath %s nexthop %s, cur count %d", |
559 | pfx_buf, path_buf, | |
2ec1e66f DW |
560 | inet_ntop (AF_INET, &cur_mpath->attr->nexthop, |
561 | nh_buf[0], sizeof (nh_buf[0])), | |
a6086ad4 | 562 | mpath_count); |
2ec1e66f | 563 | } |
de8d5dff JB |
564 | cur_mpath = next_mpath; |
565 | } | |
566 | else | |
567 | { | |
568 | /* | |
569 | * If here, we have a path on the mp_list that was not previously | |
570 | * a multipath (due to non-equivalance or maxpaths exceeded), | |
571 | * or the matching multipath is sorted later in the multipath | |
572 | * list. Before we enqueue the path on the new multipath list, | |
573 | * make sure its not on the old_best multipath list or referenced | |
574 | * via next_mpath: | |
575 | * - If next_mpath points to this new path, update next_mpath to | |
576 | * point to the multipath after this one | |
577 | * - Dequeue the path from the multipath list just to make sure | |
578 | */ | |
579 | new_mpath = listgetdata (mp_node); | |
580 | list_delete_node (mp_list, mp_node); | |
de8d5dff JB |
581 | if ((mpath_count < maxpaths) && (new_mpath != new_best) && |
582 | bgp_info_nexthop_cmp (prev_mpath, new_mpath)) | |
583 | { | |
78d92e17 | 584 | if (new_mpath == next_mpath) |
18f1dc06 | 585 | bgp_info_mpath_next (new_mpath); |
78d92e17 JB |
586 | bgp_info_mpath_dequeue (new_mpath); |
587 | ||
de8d5dff JB |
588 | bgp_info_mpath_enqueue (prev_mpath, new_mpath); |
589 | prev_mpath = new_mpath; | |
590 | mpath_changed = 1; | |
591 | mpath_count++; | |
592 | if (debug) | |
2ec1e66f DW |
593 | { |
594 | bgp_info_path_with_addpath_rx_str(new_mpath, path_buf); | |
a6086ad4 | 595 | zlog_debug ("%s: add mpath %s nexthop %s, cur count %d", |
596 | pfx_buf, path_buf, | |
2ec1e66f DW |
597 | inet_ntop (AF_INET, &new_mpath->attr->nexthop, |
598 | nh_buf[0], sizeof (nh_buf[0])), | |
a6086ad4 | 599 | mpath_count); |
2ec1e66f | 600 | } |
de8d5dff JB |
601 | } |
602 | mp_node = mp_next_node; | |
603 | } | |
604 | } | |
605 | ||
606 | if (new_best) | |
607 | { | |
a6086ad4 | 608 | if (debug) |
609 | zlog_debug("%s: New mpath count (incl newbest) %d mpath-change %s", | |
610 | pfx_buf, mpath_count, mpath_changed ? "YES" : "NO"); | |
611 | ||
de8d5dff JB |
612 | bgp_info_mpath_count_set (new_best, mpath_count-1); |
613 | if (mpath_changed || (bgp_info_mpath_count (new_best) != old_mpath_count)) | |
614 | SET_FLAG (new_best->flags, BGP_INFO_MULTIPATH_CHG); | |
615 | } | |
616 | } | |
6918e74b JB |
617 | |
618 | /* | |
619 | * bgp_mp_dmed_deselect | |
620 | * | |
621 | * Clean up multipath information for BGP_INFO_DMED_SELECTED path that | |
622 | * is not selected as best path | |
623 | */ | |
624 | void | |
625 | bgp_mp_dmed_deselect (struct bgp_info *dmed_best) | |
626 | { | |
627 | struct bgp_info *mpinfo, *mpnext; | |
628 | ||
629 | if (!dmed_best) | |
630 | return; | |
631 | ||
632 | for (mpinfo = bgp_info_mpath_first (dmed_best); mpinfo; mpinfo = mpnext) | |
633 | { | |
634 | mpnext = bgp_info_mpath_next (mpinfo); | |
635 | bgp_info_mpath_dequeue (mpinfo); | |
636 | } | |
637 | ||
638 | bgp_info_mpath_count_set (dmed_best, 0); | |
639 | UNSET_FLAG (dmed_best->flags, BGP_INFO_MULTIPATH_CHG); | |
640 | assert (bgp_info_mpath_first (dmed_best) == 0); | |
641 | } | |
0b597ef0 JB |
642 | |
643 | /* | |
644 | * bgp_info_mpath_aggregate_update | |
645 | * | |
646 | * Set the multipath aggregate attribute. We need to see if the | |
647 | * aggregate has changed and then set the ATTR_CHANGED flag on the | |
648 | * bestpath info so that a peer update will be generated. The | |
649 | * change is detected by generating the current attribute, | |
650 | * interning it, and then comparing the interned pointer with the | |
651 | * current value. We can skip this generate/compare step if there | |
652 | * is no change in multipath selection and no attribute change in | |
653 | * any multipath. | |
654 | */ | |
655 | void | |
656 | bgp_info_mpath_aggregate_update (struct bgp_info *new_best, | |
657 | struct bgp_info *old_best) | |
658 | { | |
659 | struct bgp_info *mpinfo; | |
660 | struct aspath *aspath; | |
661 | struct aspath *asmerge; | |
662 | struct attr *new_attr, *old_attr; | |
219178b6 | 663 | u_char origin; |
0b597ef0 JB |
664 | struct community *community, *commerge; |
665 | struct ecommunity *ecomm, *ecommerge; | |
57d187bc | 666 | struct lcommunity *lcomm, *lcommerge; |
0b597ef0 JB |
667 | struct attr_extra *ae; |
668 | struct attr attr = { 0 }; | |
669 | ||
670 | if (old_best && (old_best != new_best) && | |
671 | (old_attr = bgp_info_mpath_attr (old_best))) | |
672 | { | |
fac3c245 | 673 | bgp_attr_unintern (&old_attr); |
0b597ef0 JB |
674 | bgp_info_mpath_attr_set (old_best, NULL); |
675 | } | |
676 | ||
677 | if (!new_best) | |
678 | return; | |
679 | ||
680 | if (!bgp_info_mpath_count (new_best)) | |
681 | { | |
682 | if ((new_attr = bgp_info_mpath_attr (new_best))) | |
683 | { | |
fac3c245 | 684 | bgp_attr_unintern (&new_attr); |
0b597ef0 JB |
685 | bgp_info_mpath_attr_set (new_best, NULL); |
686 | SET_FLAG (new_best->flags, BGP_INFO_ATTR_CHANGED); | |
687 | } | |
688 | return; | |
689 | } | |
690 | ||
0b597ef0 JB |
691 | bgp_attr_dup (&attr, new_best->attr); |
692 | ||
16fc1eec | 693 | if (new_best->peer && |
219178b6 | 694 | bgp_flag_check (new_best->peer->bgp, BGP_FLAG_MULTIPATH_RELAX_AS_SET)) |
0b597ef0 | 695 | { |
0b597ef0 | 696 | |
16fc1eec DS |
697 | /* aggregate attribute from multipath constituents */ |
698 | aspath = aspath_dup (attr.aspath); | |
699 | origin = attr.origin; | |
700 | community = attr.community ? community_dup (attr.community) : NULL; | |
701 | ae = attr.extra; | |
702 | ecomm = (ae && ae->ecommunity) ? ecommunity_dup (ae->ecommunity) : NULL; | |
57d187bc | 703 | lcomm = (ae && ae->lcommunity) ? lcommunity_dup (ae->lcommunity) : NULL; |
0b597ef0 | 704 | |
16fc1eec DS |
705 | for (mpinfo = bgp_info_mpath_first (new_best); mpinfo; |
706 | mpinfo = bgp_info_mpath_next (mpinfo)) | |
0b597ef0 | 707 | { |
16fc1eec DS |
708 | asmerge = aspath_aggregate (aspath, mpinfo->attr->aspath); |
709 | aspath_free (aspath); | |
710 | aspath = asmerge; | |
711 | ||
712 | if (origin < mpinfo->attr->origin) | |
713 | origin = mpinfo->attr->origin; | |
714 | ||
715 | if (mpinfo->attr->community) | |
0b597ef0 | 716 | { |
16fc1eec DS |
717 | if (community) |
718 | { | |
719 | commerge = community_merge (community, mpinfo->attr->community); | |
720 | community = community_uniq_sort (commerge); | |
721 | community_free (commerge); | |
722 | } | |
723 | else | |
724 | community = community_dup (mpinfo->attr->community); | |
0b597ef0 | 725 | } |
0b597ef0 | 726 | |
16fc1eec DS |
727 | ae = mpinfo->attr->extra; |
728 | if (ae && ae->ecommunity) | |
0b597ef0 | 729 | { |
16fc1eec DS |
730 | if (ecomm) |
731 | { | |
732 | ecommerge = ecommunity_merge (ecomm, ae->ecommunity); | |
733 | ecomm = ecommunity_uniq_sort (ecommerge); | |
734 | ecommunity_free (&ecommerge); | |
735 | } | |
736 | else | |
737 | ecomm = ecommunity_dup (ae->ecommunity); | |
0b597ef0 | 738 | } |
57d187bc JS |
739 | if (ae && ae->lcommunity) |
740 | { | |
741 | if (lcomm) | |
742 | { | |
743 | lcommerge = lcommunity_merge (lcomm, ae->lcommunity); | |
744 | lcomm = lcommunity_uniq_sort (lcommerge); | |
745 | lcommunity_free (&lcommerge); | |
746 | } | |
747 | else | |
748 | lcomm = lcommunity_dup (ae->lcommunity); | |
749 | } | |
0b597ef0 | 750 | } |
0b597ef0 | 751 | |
16fc1eec DS |
752 | attr.aspath = aspath; |
753 | attr.origin = origin; | |
754 | if (community) | |
755 | { | |
756 | attr.community = community; | |
757 | attr.flag |= ATTR_FLAG_BIT (BGP_ATTR_COMMUNITIES); | |
758 | } | |
759 | if (ecomm) | |
760 | { | |
761 | ae = bgp_attr_extra_get (&attr); | |
762 | ae->ecommunity = ecomm; | |
763 | attr.flag |= ATTR_FLAG_BIT (BGP_ATTR_EXT_COMMUNITIES); | |
764 | } | |
0b597ef0 | 765 | |
16fc1eec DS |
766 | /* Zap multipath attr nexthop so we set nexthop to self */ |
767 | attr.nexthop.s_addr = 0; | |
16fc1eec DS |
768 | if (attr.extra) |
769 | memset (&attr.extra->mp_nexthop_global, 0, sizeof (struct in6_addr)); | |
0b597ef0 | 770 | |
16fc1eec DS |
771 | /* TODO: should we set ATOMIC_AGGREGATE and AGGREGATOR? */ |
772 | } | |
0b597ef0 JB |
773 | |
774 | new_attr = bgp_attr_intern (&attr); | |
775 | bgp_attr_extra_free (&attr); | |
776 | ||
777 | if (new_attr != bgp_info_mpath_attr (new_best)) | |
778 | { | |
779 | if ((old_attr = bgp_info_mpath_attr (new_best))) | |
fac3c245 | 780 | bgp_attr_unintern (&old_attr); |
0b597ef0 JB |
781 | bgp_info_mpath_attr_set (new_best, new_attr); |
782 | SET_FLAG (new_best->flags, BGP_INFO_ATTR_CHANGED); | |
783 | } | |
784 | else | |
fac3c245 | 785 | bgp_attr_unintern (&new_attr); |
0b597ef0 | 786 | } |