]> git.proxmox.com Git - mirror_frr.git/blame - bgpd/bgp_mpath.c
lib: add frr-isisd to the native models
[mirror_frr.git] / bgpd / bgp_mpath.c
CommitLineData
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 *
896014f4
DL
17 * You should have received a copy of the GNU General Public License along
18 * with this program; see the file COPYING; if not, write to the Free Software
19 * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
165b5fff
JB
20 */
21
22#include <zebra.h>
23
24#include "command.h"
96450faf
JB
25#include "prefix.h"
26#include "linklist.h"
27#include "sockunion.h"
de8d5dff 28#include "memory.h"
3f9c7369 29#include "queue.h"
039f3a34 30#include "filter.h"
165b5fff
JB
31
32#include "bgpd/bgpd.h"
96450faf
JB
33#include "bgpd/bgp_table.h"
34#include "bgpd/bgp_route.h"
35#include "bgpd/bgp_attr.h"
de8d5dff 36#include "bgpd/bgp_debug.h"
0b597ef0
JB
37#include "bgpd/bgp_aspath.h"
38#include "bgpd/bgp_community.h"
39#include "bgpd/bgp_ecommunity.h"
57d187bc 40#include "bgpd/bgp_lcommunity.h"
165b5fff
JB
41#include "bgpd/bgp_mpath.h"
42
43/*
44 * bgp_maximum_paths_set
45 *
46 * Record maximum-paths configuration for BGP instance
47 */
d62a17ae 48int bgp_maximum_paths_set(struct bgp *bgp, afi_t afi, safi_t safi, int peertype,
d7c0a89a 49 uint16_t maxpaths, uint16_t options)
165b5fff 50{
d62a17ae 51 if (!bgp || (afi >= AFI_MAX) || (safi >= SAFI_MAX))
52 return -1;
53
54 switch (peertype) {
55 case BGP_PEER_IBGP:
56 bgp->maxpaths[afi][safi].maxpaths_ibgp = maxpaths;
57 bgp->maxpaths[afi][safi].ibgp_flags |= options;
58 break;
59 case BGP_PEER_EBGP:
60 bgp->maxpaths[afi][safi].maxpaths_ebgp = maxpaths;
61 break;
62 default:
63 return -1;
64 }
65
66 return 0;
165b5fff
JB
67}
68
69/*
70 * bgp_maximum_paths_unset
71 *
72 * Remove maximum-paths configuration from BGP instance
73 */
d62a17ae 74int bgp_maximum_paths_unset(struct bgp *bgp, afi_t afi, safi_t safi,
75 int peertype)
165b5fff 76{
d62a17ae 77 if (!bgp || (afi >= AFI_MAX) || (safi >= SAFI_MAX))
78 return -1;
79
80 switch (peertype) {
81 case BGP_PEER_IBGP:
82 bgp->maxpaths[afi][safi].maxpaths_ibgp = multipath_num;
83 bgp->maxpaths[afi][safi].ibgp_flags = 0;
84 break;
85 case BGP_PEER_EBGP:
86 bgp->maxpaths[afi][safi].maxpaths_ebgp = multipath_num;
87 break;
88 default:
89 return -1;
90 }
91
92 return 0;
165b5fff 93}
96450faf 94
194a4f2c
DW
95/*
96 * bgp_interface_same
97 *
98 * Return true if ifindex for ifp1 and ifp2 are the same, else return false.
99 */
100static int bgp_interface_same(struct interface *ifp1, struct interface *ifp2)
101{
102 if (!ifp1 && !ifp2)
103 return 1;
104
105 if (!ifp1 && ifp2)
106 return 0;
107
108 if (ifp1 && !ifp2)
109 return 0;
110
111 return (ifp1->ifindex == ifp2->ifindex);
112}
113
114
96450faf 115/*
18ee8310 116 * bgp_path_info_nexthop_cmp
96450faf
JB
117 *
118 * Compare the nexthops of two paths. Return value is less than, equal to,
40381db7
DS
119 * or greater than zero if bpi1 is respectively less than, equal to,
120 * or greater than bpi2.
96450faf 121 */
40381db7
DS
122int bgp_path_info_nexthop_cmp(struct bgp_path_info *bpi1,
123 struct bgp_path_info *bpi2)
96450faf 124{
d62a17ae 125 int compare;
05a38c0c 126 struct in6_addr addr1, addr2;
d62a17ae 127
40381db7 128 compare = IPV4_ADDR_CMP(&bpi1->attr->nexthop, &bpi2->attr->nexthop);
d62a17ae 129 if (!compare) {
40381db7
DS
130 if (bpi1->attr->mp_nexthop_len == bpi2->attr->mp_nexthop_len) {
131 switch (bpi1->attr->mp_nexthop_len) {
d62a17ae 132 case BGP_ATTR_NHLEN_IPV4:
133 case BGP_ATTR_NHLEN_VPNV4:
134 compare = IPV4_ADDR_CMP(
40381db7
DS
135 &bpi1->attr->mp_nexthop_global_in,
136 &bpi2->attr->mp_nexthop_global_in);
d62a17ae 137 break;
138 case BGP_ATTR_NHLEN_IPV6_GLOBAL:
139 case BGP_ATTR_NHLEN_VPNV6_GLOBAL:
140 compare = IPV6_ADDR_CMP(
40381db7
DS
141 &bpi1->attr->mp_nexthop_global,
142 &bpi2->attr->mp_nexthop_global);
d62a17ae 143 break;
144 case BGP_ATTR_NHLEN_IPV6_GLOBAL_AND_LL:
40381db7
DS
145 addr1 = (bpi1->attr->mp_nexthop_prefer_global)
146 ? bpi1->attr->mp_nexthop_global
147 : bpi1->attr->mp_nexthop_local;
148 addr2 = (bpi2->attr->mp_nexthop_prefer_global)
149 ? bpi2->attr->mp_nexthop_global
150 : bpi2->attr->mp_nexthop_local;
151
152 if (!bpi1->attr->mp_nexthop_prefer_global
153 && !bpi2->attr->mp_nexthop_prefer_global)
996c9314 154 compare = !bgp_interface_same(
40381db7
DS
155 bpi1->peer->ifp,
156 bpi2->peer->ifp);
194a4f2c 157
d62a17ae 158 if (!compare)
05a38c0c 159 compare = IPV6_ADDR_CMP(&addr1, &addr2);
d62a17ae 160 break;
161 }
162 }
163
164 /* This can happen if one IPv6 peer sends you global and
165 * link-local
166 * nexthops but another IPv6 peer only sends you global
167 */
40381db7
DS
168 else if (bpi1->attr->mp_nexthop_len
169 == BGP_ATTR_NHLEN_IPV6_GLOBAL
170 || bpi1->attr->mp_nexthop_len
d62a17ae 171 == BGP_ATTR_NHLEN_IPV6_GLOBAL_AND_LL) {
40381db7
DS
172 compare = IPV6_ADDR_CMP(&bpi1->attr->mp_nexthop_global,
173 &bpi2->attr->mp_nexthop_global);
d62a17ae 174 if (!compare) {
40381db7
DS
175 if (bpi1->attr->mp_nexthop_len
176 < bpi2->attr->mp_nexthop_len)
d62a17ae 177 compare = -1;
178 else
179 compare = 1;
180 }
181 }
182 }
183
184 return compare;
96450faf
JB
185}
186
187/*
18ee8310 188 * bgp_path_info_mpath_cmp
96450faf
JB
189 *
190 * This function determines our multipath list ordering. By ordering
191 * the list we can deterministically select which paths are included
192 * in the multipath set. The ordering also helps in detecting changes
193 * in the multipath selection so we can detect whether to send an
194 * update to zebra.
195 *
196 * The order of paths is determined first by received nexthop, and then
197 * by peer address if the nexthops are the same.
198 */
18ee8310 199static int bgp_path_info_mpath_cmp(void *val1, void *val2)
96450faf 200{
40381db7 201 struct bgp_path_info *bpi1, *bpi2;
d62a17ae 202 int compare;
203
40381db7
DS
204 bpi1 = val1;
205 bpi2 = val2;
d62a17ae 206
40381db7 207 compare = bgp_path_info_nexthop_cmp(bpi1, bpi2);
d62a17ae 208
209 if (!compare) {
40381db7 210 if (!bpi1->peer->su_remote && !bpi2->peer->su_remote)
d62a17ae 211 compare = 0;
40381db7 212 else if (!bpi1->peer->su_remote)
d62a17ae 213 compare = 1;
40381db7 214 else if (!bpi2->peer->su_remote)
d62a17ae 215 compare = -1;
216 else
40381db7
DS
217 compare = sockunion_cmp(bpi1->peer->su_remote,
218 bpi2->peer->su_remote);
d62a17ae 219 }
220
221 return compare;
96450faf
JB
222}
223
224/*
225 * bgp_mp_list_init
226 *
227 * Initialize the mp_list, which holds the list of multipaths
228 * selected by bgp_best_selection
229 */
d62a17ae 230void bgp_mp_list_init(struct list *mp_list)
96450faf 231{
d62a17ae 232 assert(mp_list);
233 memset(mp_list, 0, sizeof(struct list));
18ee8310 234 mp_list->cmp = bgp_path_info_mpath_cmp;
96450faf
JB
235}
236
237/*
238 * bgp_mp_list_clear
239 *
240 * Clears all entries out of the mp_list
241 */
d62a17ae 242void bgp_mp_list_clear(struct list *mp_list)
96450faf 243{
d62a17ae 244 assert(mp_list);
245 list_delete_all_node(mp_list);
96450faf
JB
246}
247
248/*
249 * bgp_mp_list_add
250 *
251 * Adds a multipath entry to the mp_list
252 */
4b7e6066 253void bgp_mp_list_add(struct list *mp_list, struct bgp_path_info *mpinfo)
96450faf 254{
d62a17ae 255 assert(mp_list && mpinfo);
256 listnode_add_sort(mp_list, mpinfo);
96450faf 257}
de8d5dff
JB
258
259/*
18ee8310 260 * bgp_path_info_mpath_new
de8d5dff 261 *
18ee8310 262 * Allocate and zero memory for a new bgp_path_info_mpath element
de8d5dff 263 */
18ee8310 264static struct bgp_path_info_mpath *bgp_path_info_mpath_new(void)
de8d5dff 265{
4b7e6066
DS
266 struct bgp_path_info_mpath *new_mpath;
267 new_mpath = XCALLOC(MTYPE_BGP_MPATH_INFO,
268 sizeof(struct bgp_path_info_mpath));
d62a17ae 269 return new_mpath;
de8d5dff
JB
270}
271
272/*
18ee8310 273 * bgp_path_info_mpath_free
de8d5dff 274 *
18ee8310 275 * Release resources for a bgp_path_info_mpath element and zero out pointer
de8d5dff 276 */
18ee8310 277void bgp_path_info_mpath_free(struct bgp_path_info_mpath **mpath)
de8d5dff 278{
d62a17ae 279 if (mpath && *mpath) {
280 if ((*mpath)->mp_attr)
281 bgp_attr_unintern(&(*mpath)->mp_attr);
282 XFREE(MTYPE_BGP_MPATH_INFO, *mpath);
283 *mpath = NULL;
284 }
de8d5dff
JB
285}
286
287/*
18ee8310 288 * bgp_path_info_mpath_get
de8d5dff 289 *
18ee8310 290 * Fetch the mpath element for the given bgp_path_info. Used for
de8d5dff
JB
291 * doing lazy allocation.
292 */
4b7e6066 293static struct bgp_path_info_mpath *
9b6d8fcf 294bgp_path_info_mpath_get(struct bgp_path_info *path)
de8d5dff 295{
4b7e6066 296 struct bgp_path_info_mpath *mpath;
9b6d8fcf 297 if (!path->mpath) {
18ee8310 298 mpath = bgp_path_info_mpath_new();
d62a17ae 299 if (!mpath)
300 return NULL;
9b6d8fcf
DS
301 path->mpath = mpath;
302 mpath->mp_info = path;
d62a17ae 303 }
9b6d8fcf 304 return path->mpath;
de8d5dff
JB
305}
306
307/*
18ee8310 308 * bgp_path_info_mpath_enqueue
de8d5dff
JB
309 *
310 * Enqueue a path onto the multipath list given the previous multipath
311 * list entry
312 */
18ee8310 313static void bgp_path_info_mpath_enqueue(struct bgp_path_info *prev_info,
9b6d8fcf 314 struct bgp_path_info *path)
de8d5dff 315{
4b7e6066 316 struct bgp_path_info_mpath *prev, *mpath;
de8d5dff 317
18ee8310 318 prev = bgp_path_info_mpath_get(prev_info);
9b6d8fcf 319 mpath = bgp_path_info_mpath_get(path);
d62a17ae 320 if (!prev || !mpath)
321 return;
de8d5dff 322
d62a17ae 323 mpath->mp_next = prev->mp_next;
324 mpath->mp_prev = prev;
325 if (prev->mp_next)
326 prev->mp_next->mp_prev = mpath;
327 prev->mp_next = mpath;
de8d5dff 328
9b6d8fcf 329 SET_FLAG(path->flags, BGP_PATH_MULTIPATH);
de8d5dff
JB
330}
331
332/*
18ee8310 333 * bgp_path_info_mpath_dequeue
de8d5dff
JB
334 *
335 * Remove a path from the multipath list
336 */
9b6d8fcf 337void bgp_path_info_mpath_dequeue(struct bgp_path_info *path)
de8d5dff 338{
9b6d8fcf 339 struct bgp_path_info_mpath *mpath = path->mpath;
d62a17ae 340 if (!mpath)
341 return;
342 if (mpath->mp_prev)
343 mpath->mp_prev->mp_next = mpath->mp_next;
344 if (mpath->mp_next)
345 mpath->mp_next->mp_prev = mpath->mp_prev;
346 mpath->mp_next = mpath->mp_prev = NULL;
9b6d8fcf 347 UNSET_FLAG(path->flags, BGP_PATH_MULTIPATH);
de8d5dff
JB
348}
349
350/*
18ee8310 351 * bgp_path_info_mpath_next
de8d5dff 352 *
18ee8310 353 * Given a bgp_path_info, return the next multipath entry
de8d5dff 354 */
9b6d8fcf 355struct bgp_path_info *bgp_path_info_mpath_next(struct bgp_path_info *path)
de8d5dff 356{
9b6d8fcf 357 if (!path->mpath || !path->mpath->mp_next)
d62a17ae 358 return NULL;
9b6d8fcf 359 return path->mpath->mp_next->mp_info;
de8d5dff
JB
360}
361
362/*
18ee8310 363 * bgp_path_info_mpath_first
de8d5dff 364 *
18ee8310 365 * Given bestpath bgp_path_info, return the first multipath entry.
de8d5dff 366 */
9b6d8fcf 367struct bgp_path_info *bgp_path_info_mpath_first(struct bgp_path_info *path)
de8d5dff 368{
9b6d8fcf 369 return bgp_path_info_mpath_next(path);
de8d5dff
JB
370}
371
372/*
18ee8310 373 * bgp_path_info_mpath_count
de8d5dff 374 *
18ee8310 375 * Given the bestpath bgp_path_info, return the number of multipath entries
de8d5dff 376 */
9b6d8fcf 377uint32_t bgp_path_info_mpath_count(struct bgp_path_info *path)
de8d5dff 378{
9b6d8fcf 379 if (!path->mpath)
d62a17ae 380 return 0;
9b6d8fcf 381 return path->mpath->mp_count;
de8d5dff
JB
382}
383
384/*
18ee8310 385 * bgp_path_info_mpath_count_set
de8d5dff
JB
386 *
387 * Sets the count of multipaths into bestpath's mpath element
388 */
9b6d8fcf 389static void bgp_path_info_mpath_count_set(struct bgp_path_info *path,
18ee8310 390 uint32_t count)
de8d5dff 391{
4b7e6066 392 struct bgp_path_info_mpath *mpath;
9b6d8fcf 393 if (!count && !path->mpath)
d62a17ae 394 return;
9b6d8fcf 395 mpath = bgp_path_info_mpath_get(path);
d62a17ae 396 if (!mpath)
397 return;
398 mpath->mp_count = count;
de8d5dff
JB
399}
400
0b597ef0 401/*
18ee8310 402 * bgp_path_info_mpath_attr
0b597ef0 403 *
18ee8310 404 * Given bestpath bgp_path_info, return aggregated attribute set used
0b597ef0
JB
405 * for advertising the multipath route
406 */
9b6d8fcf 407struct attr *bgp_path_info_mpath_attr(struct bgp_path_info *path)
0b597ef0 408{
9b6d8fcf 409 if (!path->mpath)
d62a17ae 410 return NULL;
9b6d8fcf 411 return path->mpath->mp_attr;
0b597ef0
JB
412}
413
414/*
18ee8310 415 * bgp_path_info_mpath_attr_set
0b597ef0
JB
416 *
417 * Sets the aggregated attribute into bestpath's mpath element
418 */
9b6d8fcf 419static void bgp_path_info_mpath_attr_set(struct bgp_path_info *path,
18ee8310 420 struct attr *attr)
0b597ef0 421{
4b7e6066 422 struct bgp_path_info_mpath *mpath;
9b6d8fcf 423 if (!attr && !path->mpath)
d62a17ae 424 return;
9b6d8fcf 425 mpath = bgp_path_info_mpath_get(path);
d62a17ae 426 if (!mpath)
427 return;
428 mpath->mp_attr = attr;
0b597ef0
JB
429}
430
de8d5dff 431/*
18ee8310 432 * bgp_path_info_mpath_update
de8d5dff
JB
433 *
434 * Compare and sync up the multipath list with the mp_list generated by
435 * bgp_best_selection
436 */
18ee8310
DS
437void bgp_path_info_mpath_update(struct bgp_node *rn,
438 struct bgp_path_info *new_best,
439 struct bgp_path_info *old_best,
440 struct list *mp_list,
441 struct bgp_maxpaths_cfg *mpath_cfg)
de8d5dff 442{
d7c0a89a 443 uint16_t maxpaths, mpath_count, old_mpath_count;
d62a17ae 444 struct listnode *mp_node, *mp_next_node;
4b7e6066 445 struct bgp_path_info *cur_mpath, *new_mpath, *next_mpath, *prev_mpath;
d62a17ae 446 int mpath_changed, debug;
447 char pfx_buf[PREFIX2STR_BUFFER], nh_buf[2][INET6_ADDRSTRLEN];
448 char path_buf[PATH_ADDPATH_STR_BUFFER];
449
450 mpath_changed = 0;
451 maxpaths = multipath_num;
452 mpath_count = 0;
453 cur_mpath = NULL;
454 old_mpath_count = 0;
455 prev_mpath = new_best;
456 mp_node = listhead(mp_list);
457 debug = bgp_debug_bestpath(&rn->p);
458
459 if (debug)
460 prefix2str(&rn->p, pfx_buf, sizeof(pfx_buf));
461
462 if (new_best) {
463 mpath_count++;
464 if (new_best != old_best)
18ee8310 465 bgp_path_info_mpath_dequeue(new_best);
d62a17ae 466 maxpaths = (new_best->peer->sort == BGP_PEER_IBGP)
467 ? mpath_cfg->maxpaths_ibgp
468 : mpath_cfg->maxpaths_ebgp;
469 }
470
471 if (old_best) {
18ee8310
DS
472 cur_mpath = bgp_path_info_mpath_first(old_best);
473 old_mpath_count = bgp_path_info_mpath_count(old_best);
474 bgp_path_info_mpath_count_set(old_best, 0);
475 bgp_path_info_mpath_dequeue(old_best);
d62a17ae 476 }
477
478 if (debug)
479 zlog_debug(
480 "%s: starting mpath update, newbest %s num candidates %d old-mpath-count %d",
481 pfx_buf, new_best ? new_best->peer->host : "NONE",
cc35bc7c 482 mp_list ? listcount(mp_list) : 0, old_mpath_count);
d62a17ae 483
484 /*
485 * We perform an ordered walk through both lists in parallel.
486 * The reason for the ordered walk is that if there are paths
487 * that were previously multipaths and are still multipaths, the walk
488 * should encounter them in both lists at the same time. Otherwise
489 * there will be paths that are in one list or another, and we
490 * will deal with these separately.
491 *
492 * Note that new_best might be somewhere in the mp_list, so we need
493 * to skip over it
494 */
495 while (mp_node || cur_mpath) {
4b7e6066 496 struct bgp_path_info *tmp_info;
d62a17ae 497
498 /*
499 * We can bail out of this loop if all existing paths on the
500 * multipath list have been visited (for cleanup purposes) and
501 * the maxpath requirement is fulfulled
502 */
503 if (!cur_mpath && (mpath_count >= maxpaths))
504 break;
505
506 mp_next_node = mp_node ? listnextnode(mp_node) : NULL;
18ee8310
DS
507 next_mpath =
508 cur_mpath ? bgp_path_info_mpath_next(cur_mpath) : NULL;
d62a17ae 509 tmp_info = mp_node ? listgetdata(mp_node) : NULL;
510
511 if (debug)
512 zlog_debug(
513 "%s: comparing candidate %s with existing mpath %s",
514 pfx_buf,
515 tmp_info ? tmp_info->peer->host : "NONE",
516 cur_mpath ? cur_mpath->peer->host : "NONE");
517
518 /*
519 * If equal, the path was a multipath and is still a multipath.
520 * Insert onto new multipath list if maxpaths allows.
521 */
522 if (mp_node && (listgetdata(mp_node) == cur_mpath)) {
523 list_delete_node(mp_list, mp_node);
18ee8310 524 bgp_path_info_mpath_dequeue(cur_mpath);
d62a17ae 525 if ((mpath_count < maxpaths)
18ee8310
DS
526 && bgp_path_info_nexthop_cmp(prev_mpath,
527 cur_mpath)) {
528 bgp_path_info_mpath_enqueue(prev_mpath,
529 cur_mpath);
d62a17ae 530 prev_mpath = cur_mpath;
531 mpath_count++;
532 if (debug) {
18ee8310 533 bgp_path_info_path_with_addpath_rx_str(
d62a17ae 534 cur_mpath, path_buf);
535 zlog_debug(
536 "%s: %s is still multipath, cur count %d",
537 pfx_buf, path_buf, mpath_count);
538 }
539 } else {
540 mpath_changed = 1;
541 if (debug) {
18ee8310 542 bgp_path_info_path_with_addpath_rx_str(
d62a17ae 543 cur_mpath, path_buf);
544 zlog_debug(
545 "%s: remove mpath %s nexthop %s, cur count %d",
546 pfx_buf, path_buf,
547 inet_ntop(AF_INET,
548 &cur_mpath->attr
549 ->nexthop,
550 nh_buf[0],
551 sizeof(nh_buf[0])),
552 mpath_count);
553 }
554 }
555 mp_node = mp_next_node;
556 cur_mpath = next_mpath;
557 continue;
558 }
559
560 if (cur_mpath
561 && (!mp_node
18ee8310
DS
562 || (bgp_path_info_mpath_cmp(cur_mpath,
563 listgetdata(mp_node))
d62a17ae 564 < 0))) {
565 /*
566 * If here, we have an old multipath and either the
567 * mp_list
568 * is finished or the next mp_node points to a later
569 * multipath, so we need to purge this path from the
570 * multipath list
571 */
18ee8310 572 bgp_path_info_mpath_dequeue(cur_mpath);
d62a17ae 573 mpath_changed = 1;
574 if (debug) {
18ee8310
DS
575 bgp_path_info_path_with_addpath_rx_str(
576 cur_mpath, path_buf);
d62a17ae 577 zlog_debug(
578 "%s: remove mpath %s nexthop %s, cur count %d",
579 pfx_buf, path_buf,
580 inet_ntop(AF_INET,
581 &cur_mpath->attr->nexthop,
582 nh_buf[0], sizeof(nh_buf[0])),
583 mpath_count);
584 }
585 cur_mpath = next_mpath;
586 } else {
587 /*
588 * If here, we have a path on the mp_list that was not
589 * previously
590 * a multipath (due to non-equivalance or maxpaths
591 * exceeded),
592 * or the matching multipath is sorted later in the
593 * multipath
594 * list. Before we enqueue the path on the new multipath
595 * list,
596 * make sure its not on the old_best multipath list or
597 * referenced
598 * via next_mpath:
599 * - If next_mpath points to this new path, update
600 * next_mpath to
601 * point to the multipath after this one
602 * - Dequeue the path from the multipath list just to
603 * make sure
604 */
605 new_mpath = listgetdata(mp_node);
606 list_delete_node(mp_list, mp_node);
2165e2ea
QY
607 assert(new_mpath);
608 assert(prev_mpath);
d62a17ae 609 if ((mpath_count < maxpaths) && (new_mpath != new_best)
18ee8310
DS
610 && bgp_path_info_nexthop_cmp(prev_mpath,
611 new_mpath)) {
d62a17ae 612 if (new_mpath == next_mpath)
18ee8310
DS
613 bgp_path_info_mpath_next(new_mpath);
614 bgp_path_info_mpath_dequeue(new_mpath);
d62a17ae 615
18ee8310
DS
616 bgp_path_info_mpath_enqueue(prev_mpath,
617 new_mpath);
d62a17ae 618 prev_mpath = new_mpath;
619 mpath_changed = 1;
620 mpath_count++;
621 if (debug) {
18ee8310 622 bgp_path_info_path_with_addpath_rx_str(
d62a17ae 623 new_mpath, path_buf);
624 zlog_debug(
625 "%s: add mpath %s nexthop %s, cur count %d",
626 pfx_buf, path_buf,
627 inet_ntop(AF_INET,
628 &new_mpath->attr
629 ->nexthop,
630 nh_buf[0],
631 sizeof(nh_buf[0])),
632 mpath_count);
633 }
634 }
635 mp_node = mp_next_node;
636 }
637 }
638
639 if (new_best) {
640 if (debug)
641 zlog_debug(
642 "%s: New mpath count (incl newbest) %d mpath-change %s",
643 pfx_buf, mpath_count,
644 mpath_changed ? "YES" : "NO");
645
18ee8310 646 bgp_path_info_mpath_count_set(new_best, mpath_count - 1);
d62a17ae 647 if (mpath_changed
18ee8310 648 || (bgp_path_info_mpath_count(new_best) != old_mpath_count))
1defdda8 649 SET_FLAG(new_best->flags, BGP_PATH_MULTIPATH_CHG);
d62a17ae 650 }
de8d5dff 651}
6918e74b
JB
652
653/*
654 * bgp_mp_dmed_deselect
655 *
1defdda8 656 * Clean up multipath information for BGP_PATH_DMED_SELECTED path that
6918e74b
JB
657 * is not selected as best path
658 */
4b7e6066 659void bgp_mp_dmed_deselect(struct bgp_path_info *dmed_best)
6918e74b 660{
4b7e6066 661 struct bgp_path_info *mpinfo, *mpnext;
6918e74b 662
d62a17ae 663 if (!dmed_best)
664 return;
6918e74b 665
18ee8310 666 for (mpinfo = bgp_path_info_mpath_first(dmed_best); mpinfo;
d62a17ae 667 mpinfo = mpnext) {
18ee8310
DS
668 mpnext = bgp_path_info_mpath_next(mpinfo);
669 bgp_path_info_mpath_dequeue(mpinfo);
d62a17ae 670 }
6918e74b 671
18ee8310 672 bgp_path_info_mpath_count_set(dmed_best, 0);
1defdda8 673 UNSET_FLAG(dmed_best->flags, BGP_PATH_MULTIPATH_CHG);
18ee8310 674 assert(bgp_path_info_mpath_first(dmed_best) == 0);
6918e74b 675}
0b597ef0
JB
676
677/*
18ee8310 678 * bgp_path_info_mpath_aggregate_update
0b597ef0
JB
679 *
680 * Set the multipath aggregate attribute. We need to see if the
681 * aggregate has changed and then set the ATTR_CHANGED flag on the
682 * bestpath info so that a peer update will be generated. The
683 * change is detected by generating the current attribute,
684 * interning it, and then comparing the interned pointer with the
685 * current value. We can skip this generate/compare step if there
686 * is no change in multipath selection and no attribute change in
687 * any multipath.
688 */
18ee8310
DS
689void bgp_path_info_mpath_aggregate_update(struct bgp_path_info *new_best,
690 struct bgp_path_info *old_best)
0b597ef0 691{
4b7e6066 692 struct bgp_path_info *mpinfo;
d62a17ae 693 struct aspath *aspath;
694 struct aspath *asmerge;
695 struct attr *new_attr, *old_attr;
d7c0a89a 696 uint8_t origin;
d62a17ae 697 struct community *community, *commerge;
698 struct ecommunity *ecomm, *ecommerge;
699 struct lcommunity *lcomm, *lcommerge;
700 struct attr attr = {0};
701
702 if (old_best && (old_best != new_best)
18ee8310 703 && (old_attr = bgp_path_info_mpath_attr(old_best))) {
d62a17ae 704 bgp_attr_unintern(&old_attr);
18ee8310 705 bgp_path_info_mpath_attr_set(old_best, NULL);
d62a17ae 706 }
707
708 if (!new_best)
709 return;
710
18ee8310
DS
711 if (!bgp_path_info_mpath_count(new_best)) {
712 if ((new_attr = bgp_path_info_mpath_attr(new_best))) {
d62a17ae 713 bgp_attr_unintern(&new_attr);
18ee8310 714 bgp_path_info_mpath_attr_set(new_best, NULL);
1defdda8 715 SET_FLAG(new_best->flags, BGP_PATH_ATTR_CHANGED);
d62a17ae 716 }
717 return;
718 }
719
720 bgp_attr_dup(&attr, new_best->attr);
721
9d303b37
DL
722 if (new_best->peer && bgp_flag_check(new_best->peer->bgp,
723 BGP_FLAG_MULTIPATH_RELAX_AS_SET)) {
d62a17ae 724
725 /* aggregate attribute from multipath constituents */
726 aspath = aspath_dup(attr.aspath);
727 origin = attr.origin;
728 community =
729 attr.community ? community_dup(attr.community) : NULL;
730 ecomm = (attr.ecommunity) ? ecommunity_dup(attr.ecommunity)
731 : NULL;
732 lcomm = (attr.lcommunity) ? lcommunity_dup(attr.lcommunity)
733 : NULL;
734
18ee8310
DS
735 for (mpinfo = bgp_path_info_mpath_first(new_best); mpinfo;
736 mpinfo = bgp_path_info_mpath_next(mpinfo)) {
d62a17ae 737 asmerge =
738 aspath_aggregate(aspath, mpinfo->attr->aspath);
739 aspath_free(aspath);
740 aspath = asmerge;
741
742 if (origin < mpinfo->attr->origin)
743 origin = mpinfo->attr->origin;
744
745 if (mpinfo->attr->community) {
746 if (community) {
747 commerge = community_merge(
748 community,
749 mpinfo->attr->community);
750 community =
751 community_uniq_sort(commerge);
3c1f53de 752 community_free(&commerge);
d62a17ae 753 } else
754 community = community_dup(
755 mpinfo->attr->community);
756 }
757
758 if (mpinfo->attr->ecommunity) {
759 if (ecomm) {
760 ecommerge = ecommunity_merge(
761 ecomm,
762 mpinfo->attr->ecommunity);
763 ecomm = ecommunity_uniq_sort(ecommerge);
764 ecommunity_free(&ecommerge);
765 } else
766 ecomm = ecommunity_dup(
767 mpinfo->attr->ecommunity);
768 }
769 if (mpinfo->attr->lcommunity) {
770 if (lcomm) {
771 lcommerge = lcommunity_merge(
772 lcomm,
773 mpinfo->attr->lcommunity);
774 lcomm = lcommunity_uniq_sort(lcommerge);
775 lcommunity_free(&lcommerge);
776 } else
777 lcomm = lcommunity_dup(
778 mpinfo->attr->lcommunity);
779 }
780 }
781
782 attr.aspath = aspath;
783 attr.origin = origin;
784 if (community) {
785 attr.community = community;
786 attr.flag |= ATTR_FLAG_BIT(BGP_ATTR_COMMUNITIES);
787 }
788 if (ecomm) {
789 attr.ecommunity = ecomm;
790 attr.flag |= ATTR_FLAG_BIT(BGP_ATTR_EXT_COMMUNITIES);
791 }
44f12f20
RW
792 if (lcomm) {
793 attr.lcommunity = lcomm;
794 attr.flag |= ATTR_FLAG_BIT(BGP_ATTR_LARGE_COMMUNITIES);
795 }
d62a17ae 796
797 /* Zap multipath attr nexthop so we set nexthop to self */
798 attr.nexthop.s_addr = 0;
799 memset(&attr.mp_nexthop_global, 0, sizeof(struct in6_addr));
800
801 /* TODO: should we set ATOMIC_AGGREGATE and AGGREGATOR? */
802 }
803
804 new_attr = bgp_attr_intern(&attr);
805
18ee8310
DS
806 if (new_attr != bgp_path_info_mpath_attr(new_best)) {
807 if ((old_attr = bgp_path_info_mpath_attr(new_best)))
d62a17ae 808 bgp_attr_unintern(&old_attr);
18ee8310 809 bgp_path_info_mpath_attr_set(new_best, new_attr);
1defdda8 810 SET_FLAG(new_best->flags, BGP_PATH_ATTR_CHANGED);
d62a17ae 811 } else
812 bgp_attr_unintern(&new_attr);
0b597ef0 813}