]>
Commit | Line | Data |
---|---|---|
718e3744 | 1 | /* Routing Information Base. |
2 | * Copyright (C) 1997, 98, 99, 2001 Kunihiro Ishiguro | |
3 | * | |
4 | * This file is part of GNU Zebra. | |
5 | * | |
6 | * GNU Zebra is free software; you can redistribute it and/or modify it | |
7 | * under the terms of the GNU General Public License as published by the | |
8 | * Free Software Foundation; either version 2, or (at your option) any | |
9 | * later version. | |
10 | * | |
11 | * GNU Zebra is distributed in the hope that it will be useful, but | |
12 | * WITHOUT ANY WARRANTY; without even the implied warranty of | |
13 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU | |
14 | * General Public License for more details. | |
15 | * | |
896014f4 DL |
16 | * You should have received a copy of the GNU General Public License along |
17 | * with this program; see the file COPYING; if not, write to the Free Software | |
18 | * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA | |
718e3744 | 19 | */ |
20 | ||
21 | #include <zebra.h> | |
22 | ||
b892f1dd | 23 | #include "if.h" |
718e3744 | 24 | #include "prefix.h" |
25 | #include "table.h" | |
26 | #include "memory.h" | |
4a1ab8e4 | 27 | #include "zebra_memory.h" |
718e3744 | 28 | #include "command.h" |
718e3744 | 29 | #include "log.h" |
deaa50db | 30 | #include "log_int.h" |
718e3744 | 31 | #include "sockunion.h" |
4d38fdb4 | 32 | #include "linklist.h" |
33 | #include "thread.h" | |
34 | #include "workqueue.h" | |
7514fb77 PJ |
35 | #include "prefix.h" |
36 | #include "routemap.h" | |
fb018d25 | 37 | #include "nexthop.h" |
b72ede27 | 38 | #include "vrf.h" |
40c7bdb0 | 39 | #include "mpls.h" |
05737783 | 40 | #include "srcdest_table.h" |
718e3744 | 41 | |
42 | #include "zebra/rib.h" | |
43 | #include "zebra/rt.h" | |
7c551956 | 44 | #include "zebra/zebra_ns.h" |
718e3744 | 45 | #include "zebra/zserv.h" |
7c551956 | 46 | #include "zebra/zebra_vrf.h" |
718e3744 | 47 | #include "zebra/redistribute.h" |
8902474b | 48 | #include "zebra/zebra_routemap.h" |
718e3744 | 49 | #include "zebra/debug.h" |
fb018d25 | 50 | #include "zebra/zebra_rnh.h" |
88177fe3 | 51 | #include "zebra/interface.h" |
d44ca835 | 52 | #include "zebra/connected.h" |
718e3744 | 53 | |
d62a17ae | 54 | DEFINE_HOOK(rib_update, (struct route_node * rn, const char *reason), |
55 | (rn, reason)) | |
4f8ea50c | 56 | |
6baf7bb8 DS |
57 | /* Should we allow non Quagga processes to delete our routes */ |
58 | extern int allow_delete; | |
59 | ||
457eb9af PJ |
60 | /* Hold time for RIB process, should be very minimal. |
61 | * it is useful to able to set it otherwise for testing, hence exported | |
62 | * as global here for test-rig code. | |
63 | */ | |
64 | int rib_process_hold_time = 10; | |
65 | ||
718e3744 | 66 | /* Each route type's string and default distance value. */ |
d62a17ae | 67 | static const struct { |
68 | int key; | |
69 | int distance; | |
70 | } route_info[ZEBRA_ROUTE_MAX] = { | |
71 | [ZEBRA_ROUTE_SYSTEM] = {ZEBRA_ROUTE_SYSTEM, 0}, | |
72 | [ZEBRA_ROUTE_KERNEL] = {ZEBRA_ROUTE_KERNEL, 0}, | |
73 | [ZEBRA_ROUTE_CONNECT] = {ZEBRA_ROUTE_CONNECT, 0}, | |
74 | [ZEBRA_ROUTE_STATIC] = {ZEBRA_ROUTE_STATIC, 1}, | |
75 | [ZEBRA_ROUTE_RIP] = {ZEBRA_ROUTE_RIP, 120}, | |
76 | [ZEBRA_ROUTE_RIPNG] = {ZEBRA_ROUTE_RIPNG, 120}, | |
77 | [ZEBRA_ROUTE_OSPF] = {ZEBRA_ROUTE_OSPF, 110}, | |
78 | [ZEBRA_ROUTE_OSPF6] = {ZEBRA_ROUTE_OSPF6, 110}, | |
79 | [ZEBRA_ROUTE_ISIS] = {ZEBRA_ROUTE_ISIS, 115}, | |
80 | [ZEBRA_ROUTE_BGP] = {ZEBRA_ROUTE_BGP, 20 /* IBGP is 200. */}, | |
81 | [ZEBRA_ROUTE_NHRP] = {ZEBRA_ROUTE_NHRP, 10}, | |
82 | /* no entry/default: 150 */ | |
718e3744 | 83 | }; |
6b0655a2 | 84 | |
4623d897 DL |
85 | /* RPF lookup behaviour */ |
86 | static enum multicast_mode ipv4_multicast_mode = MCAST_NO_CONFIG; | |
87 | ||
6c4f4e6e | 88 | |
d62a17ae | 89 | static void __attribute__((format(printf, 5, 6))) |
90 | _rnode_zlog(const char *_func, vrf_id_t vrf_id, struct route_node *rn, | |
91 | int priority, const char *msgfmt, ...) | |
2263a412 | 92 | { |
d62a17ae | 93 | char buf[SRCDEST2STR_BUFFER + sizeof(" (MRIB)")]; |
94 | char msgbuf[512]; | |
95 | va_list ap; | |
2263a412 | 96 | |
d62a17ae | 97 | va_start(ap, msgfmt); |
98 | vsnprintf(msgbuf, sizeof(msgbuf), msgfmt, ap); | |
99 | va_end(ap); | |
2263a412 | 100 | |
d62a17ae | 101 | if (rn) { |
102 | rib_table_info_t *info = srcdest_rnode_table_info(rn); | |
103 | srcdest_rnode2str(rn, buf, sizeof(buf)); | |
cb653491 | 104 | |
d62a17ae | 105 | if (info->safi == SAFI_MULTICAST) |
106 | strcat(buf, " (MRIB)"); | |
107 | } else { | |
108 | snprintf(buf, sizeof(buf), "{(route_node *) NULL}"); | |
109 | } | |
2263a412 | 110 | |
d62a17ae | 111 | zlog(priority, "%s: %d:%s: %s", _func, vrf_id, buf, msgbuf); |
2263a412 DL |
112 | } |
113 | ||
d62a17ae | 114 | #define rnode_debug(node, vrf_id, ...) \ |
2263a412 | 115 | _rnode_zlog(__func__, vrf_id, node, LOG_DEBUG, __VA_ARGS__) |
d62a17ae | 116 | #define rnode_info(node, ...) \ |
2263a412 DL |
117 | _rnode_zlog(__func__, vrf_id, node, LOG_INFO, __VA_ARGS__) |
118 | ||
d62a17ae | 119 | u_char route_distance(int type) |
40c7bdb0 | 120 | { |
d62a17ae | 121 | u_char distance; |
40c7bdb0 | 122 | |
d62a17ae | 123 | if ((unsigned)type >= array_size(route_info)) |
124 | distance = 150; | |
125 | else | |
126 | distance = route_info[type].distance; | |
40c7bdb0 | 127 | |
d62a17ae | 128 | return distance; |
40c7bdb0 | 129 | } |
130 | ||
d62a17ae | 131 | int is_zebra_valid_kernel_table(u_int32_t table_id) |
7a4bb9c5 | 132 | { |
d62a17ae | 133 | if ((table_id > ZEBRA_KERNEL_TABLE_MAX)) |
134 | return 0; | |
8f500a1c RW |
135 | |
136 | #ifdef linux | |
d62a17ae | 137 | if ((table_id == RT_TABLE_UNSPEC) || (table_id == RT_TABLE_LOCAL) |
138 | || (table_id == RT_TABLE_COMPAT)) | |
139 | return 0; | |
8f500a1c RW |
140 | #endif |
141 | ||
d62a17ae | 142 | return 1; |
7a4bb9c5 DS |
143 | } |
144 | ||
d62a17ae | 145 | int is_zebra_main_routing_table(u_int32_t table_id) |
7a4bb9c5 | 146 | { |
d62a17ae | 147 | if ((table_id == RT_TABLE_MAIN) |
148 | || (table_id == zebrad.rtm_table_default)) | |
149 | return 1; | |
150 | return 0; | |
7a4bb9c5 DS |
151 | } |
152 | ||
d62a17ae | 153 | int zebra_check_addr(struct prefix *p) |
0aabccc0 | 154 | { |
d62a17ae | 155 | if (p->family == AF_INET) { |
156 | u_int32_t addr; | |
0aabccc0 | 157 | |
d62a17ae | 158 | addr = p->u.prefix4.s_addr; |
159 | addr = ntohl(addr); | |
0aabccc0 | 160 | |
d62a17ae | 161 | if (IPV4_NET127(addr) || IN_CLASSD(addr) |
162 | || IPV4_LINKLOCAL(addr)) | |
163 | return 0; | |
164 | } | |
165 | if (p->family == AF_INET6) { | |
166 | if (IN6_IS_ADDR_LOOPBACK(&p->u.prefix6)) | |
167 | return 0; | |
168 | if (IN6_IS_ADDR_LINKLOCAL(&p->u.prefix6)) | |
169 | return 0; | |
170 | } | |
171 | return 1; | |
0aabccc0 DD |
172 | } |
173 | ||
fa713d9e | 174 | /* Add nexthop to the end of a rib node's nexthop list */ |
d62a17ae | 175 | void route_entry_nexthop_add(struct route_entry *re, struct nexthop *nexthop) |
fa713d9e | 176 | { |
d62a17ae | 177 | nexthop_add(&re->nexthop, nexthop); |
178 | re->nexthop_num++; | |
718e3744 | 179 | } |
180 | ||
6e26278c | 181 | |
6e26278c DS |
182 | /** |
183 | * copy_nexthop - copy a nexthop to the rib structure. | |
184 | */ | |
d62a17ae | 185 | void route_entry_copy_nexthops(struct route_entry *re, struct nexthop *nh) |
6e26278c | 186 | { |
d62a17ae | 187 | assert(!re->nexthop); |
188 | copy_nexthops(&re->nexthop, nh, NULL); | |
189 | for (struct nexthop *nexthop = nh; nexthop; nexthop = nexthop->next) | |
190 | re->nexthop_num++; | |
6e26278c DS |
191 | } |
192 | ||
718e3744 | 193 | /* Delete specified nexthop from the list. */ |
d62a17ae | 194 | void route_entry_nexthop_delete(struct route_entry *re, struct nexthop *nexthop) |
718e3744 | 195 | { |
d62a17ae | 196 | if (nexthop->next) |
197 | nexthop->next->prev = nexthop->prev; | |
198 | if (nexthop->prev) | |
199 | nexthop->prev->next = nexthop->next; | |
200 | else | |
201 | re->nexthop = nexthop->next; | |
202 | re->nexthop_num--; | |
718e3744 | 203 | } |
204 | ||
fa713d9e | 205 | |
d62a17ae | 206 | struct nexthop *route_entry_nexthop_ifindex_add(struct route_entry *re, |
207 | ifindex_t ifindex) | |
718e3744 | 208 | { |
d62a17ae | 209 | struct nexthop *nexthop; |
718e3744 | 210 | |
d62a17ae | 211 | nexthop = nexthop_new(); |
212 | nexthop->type = NEXTHOP_TYPE_IFINDEX; | |
213 | nexthop->ifindex = ifindex; | |
718e3744 | 214 | |
d62a17ae | 215 | route_entry_nexthop_add(re, nexthop); |
718e3744 | 216 | |
d62a17ae | 217 | return nexthop; |
718e3744 | 218 | } |
219 | ||
d62a17ae | 220 | struct nexthop *route_entry_nexthop_ipv4_add(struct route_entry *re, |
221 | struct in_addr *ipv4, | |
222 | struct in_addr *src) | |
718e3744 | 223 | { |
d62a17ae | 224 | struct nexthop *nexthop; |
718e3744 | 225 | |
d62a17ae | 226 | nexthop = nexthop_new(); |
227 | nexthop->type = NEXTHOP_TYPE_IPV4; | |
228 | nexthop->gate.ipv4 = *ipv4; | |
229 | if (src) | |
230 | nexthop->src.ipv4 = *src; | |
718e3744 | 231 | |
d62a17ae | 232 | route_entry_nexthop_add(re, nexthop); |
718e3744 | 233 | |
d62a17ae | 234 | return nexthop; |
718e3744 | 235 | } |
236 | ||
d62a17ae | 237 | struct nexthop *route_entry_nexthop_ipv4_ifindex_add(struct route_entry *re, |
238 | struct in_addr *ipv4, | |
239 | struct in_addr *src, | |
240 | ifindex_t ifindex) | |
718e3744 | 241 | { |
d62a17ae | 242 | struct nexthop *nexthop; |
243 | struct interface *ifp; | |
718e3744 | 244 | |
d62a17ae | 245 | nexthop = nexthop_new(); |
246 | nexthop->type = NEXTHOP_TYPE_IPV4_IFINDEX; | |
247 | nexthop->gate.ipv4 = *ipv4; | |
248 | if (src) | |
249 | nexthop->src.ipv4 = *src; | |
250 | nexthop->ifindex = ifindex; | |
251 | ifp = if_lookup_by_index(nexthop->ifindex, VRF_DEFAULT); | |
252 | /*Pending: need to think if null ifp here is ok during bootup? | |
253 | There was a crash because ifp here was coming to be NULL */ | |
254 | if (ifp) | |
255 | if (connected_is_unnumbered(ifp)) { | |
256 | SET_FLAG(nexthop->flags, NEXTHOP_FLAG_ONLINK); | |
257 | } | |
718e3744 | 258 | |
d62a17ae | 259 | route_entry_nexthop_add(re, nexthop); |
718e3744 | 260 | |
d62a17ae | 261 | return nexthop; |
718e3744 | 262 | } |
263 | ||
d62a17ae | 264 | struct nexthop *route_entry_nexthop_ipv6_add(struct route_entry *re, |
265 | struct in6_addr *ipv6) | |
718e3744 | 266 | { |
d62a17ae | 267 | struct nexthop *nexthop; |
718e3744 | 268 | |
d62a17ae | 269 | nexthop = nexthop_new(); |
270 | nexthop->type = NEXTHOP_TYPE_IPV6; | |
271 | nexthop->gate.ipv6 = *ipv6; | |
718e3744 | 272 | |
d62a17ae | 273 | route_entry_nexthop_add(re, nexthop); |
718e3744 | 274 | |
d62a17ae | 275 | return nexthop; |
718e3744 | 276 | } |
277 | ||
d62a17ae | 278 | struct nexthop *route_entry_nexthop_ipv6_ifindex_add(struct route_entry *re, |
279 | struct in6_addr *ipv6, | |
280 | ifindex_t ifindex) | |
718e3744 | 281 | { |
d62a17ae | 282 | struct nexthop *nexthop; |
718e3744 | 283 | |
d62a17ae | 284 | nexthop = nexthop_new(); |
285 | nexthop->type = NEXTHOP_TYPE_IPV6_IFINDEX; | |
286 | nexthop->gate.ipv6 = *ipv6; | |
287 | nexthop->ifindex = ifindex; | |
718e3744 | 288 | |
d62a17ae | 289 | route_entry_nexthop_add(re, nexthop); |
718e3744 | 290 | |
d62a17ae | 291 | return nexthop; |
718e3744 | 292 | } |
718e3744 | 293 | |
d62a17ae | 294 | struct nexthop *route_entry_nexthop_blackhole_add(struct route_entry *re) |
595db7f1 | 295 | { |
d62a17ae | 296 | struct nexthop *nexthop; |
595db7f1 | 297 | |
d62a17ae | 298 | nexthop = nexthop_new(); |
299 | nexthop->type = NEXTHOP_TYPE_BLACKHOLE; | |
300 | SET_FLAG(re->flags, ZEBRA_FLAG_BLACKHOLE); | |
595db7f1 | 301 | |
d62a17ae | 302 | route_entry_nexthop_add(re, nexthop); |
595db7f1 | 303 | |
d62a17ae | 304 | return nexthop; |
595db7f1 | 305 | } |
306 | ||
d62a17ae | 307 | static void nexthop_set_resolved(afi_t afi, struct nexthop *newhop, |
308 | struct nexthop *nexthop) | |
4491a88f | 309 | { |
d62a17ae | 310 | struct nexthop *resolved_hop; |
4491a88f | 311 | |
d62a17ae | 312 | resolved_hop = nexthop_new(); |
313 | SET_FLAG(resolved_hop->flags, NEXTHOP_FLAG_ACTIVE); | |
314 | /* If the resolving route specifies a gateway, use it */ | |
315 | if (newhop->type == NEXTHOP_TYPE_IPV4 | |
316 | || newhop->type == NEXTHOP_TYPE_IPV4_IFINDEX) { | |
317 | resolved_hop->type = newhop->type; | |
318 | resolved_hop->gate.ipv4 = newhop->gate.ipv4; | |
4491a88f | 319 | |
d62a17ae | 320 | if (newhop->ifindex) { |
321 | resolved_hop->type = NEXTHOP_TYPE_IPV4_IFINDEX; | |
322 | resolved_hop->ifindex = newhop->ifindex; | |
323 | if (newhop->flags & NEXTHOP_FLAG_ONLINK) | |
324 | resolved_hop->flags |= NEXTHOP_FLAG_ONLINK; | |
325 | } | |
4491a88f | 326 | } |
d62a17ae | 327 | if (newhop->type == NEXTHOP_TYPE_IPV6 |
328 | || newhop->type == NEXTHOP_TYPE_IPV6_IFINDEX) { | |
329 | resolved_hop->type = newhop->type; | |
330 | resolved_hop->gate.ipv6 = newhop->gate.ipv6; | |
331 | ||
332 | if (newhop->ifindex) { | |
333 | resolved_hop->type = NEXTHOP_TYPE_IPV6_IFINDEX; | |
334 | resolved_hop->ifindex = newhop->ifindex; | |
335 | } | |
4491a88f | 336 | } |
4491a88f | 337 | |
d62a17ae | 338 | /* If the resolving route is an interface route, |
339 | * it means the gateway we are looking up is connected | |
340 | * to that interface. (The actual network is _not_ onlink). | |
341 | * Therefore, the resolved route should have the original | |
342 | * gateway as nexthop as it is directly connected. | |
343 | * | |
344 | * On Linux, we have to set the onlink netlink flag because | |
345 | * otherwise, the kernel won't accept the route. | |
346 | */ | |
347 | if (newhop->type == NEXTHOP_TYPE_IFINDEX) { | |
348 | resolved_hop->flags |= NEXTHOP_FLAG_ONLINK; | |
349 | if (afi == AFI_IP) { | |
350 | resolved_hop->type = NEXTHOP_TYPE_IPV4_IFINDEX; | |
351 | resolved_hop->gate.ipv4 = nexthop->gate.ipv4; | |
352 | } else if (afi == AFI_IP6) { | |
353 | resolved_hop->type = NEXTHOP_TYPE_IPV6_IFINDEX; | |
354 | resolved_hop->gate.ipv6 = nexthop->gate.ipv6; | |
355 | } | |
356 | resolved_hop->ifindex = newhop->ifindex; | |
4491a88f | 357 | } |
4491a88f | 358 | |
d62a17ae | 359 | resolved_hop->rparent = nexthop; |
360 | nexthop_add(&nexthop->resolved, resolved_hop); | |
4491a88f DS |
361 | } |
362 | ||
718e3744 | 363 | /* If force flag is not set, do not modify falgs at all for uninstall |
364 | the route from FIB. */ | |
d62a17ae | 365 | static int nexthop_active(afi_t afi, struct route_entry *re, |
366 | struct nexthop *nexthop, int set, | |
367 | struct route_node *top) | |
368 | { | |
369 | struct prefix p; | |
370 | struct route_table *table; | |
371 | struct route_node *rn; | |
372 | struct route_entry *match; | |
373 | int resolved; | |
374 | struct nexthop *newhop; | |
375 | struct interface *ifp; | |
376 | ||
377 | if ((nexthop->type == NEXTHOP_TYPE_IPV4) | |
378 | || nexthop->type == NEXTHOP_TYPE_IPV6) | |
379 | nexthop->ifindex = 0; | |
380 | ||
381 | if (set) { | |
382 | UNSET_FLAG(nexthop->flags, NEXTHOP_FLAG_RECURSIVE); | |
383 | zebra_deregister_rnh_static_nexthops(re->vrf_id, | |
384 | nexthop->resolved, top); | |
385 | nexthops_free(nexthop->resolved); | |
386 | nexthop->resolved = NULL; | |
387 | re->nexthop_mtu = 0; | |
d44ca835 | 388 | } |
18ff3edd | 389 | |
d62a17ae | 390 | /* Skip nexthops that have been filtered out due to route-map */ |
391 | /* The nexthops are specific to this route and so the same */ | |
392 | /* nexthop for a different route may not have this flag set */ | |
393 | if (CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_FILTERED)) | |
394 | return 0; | |
395 | ||
396 | /* | |
397 | * Check to see if we should trust the passed in information | |
398 | * for UNNUMBERED interfaces as that we won't find the GW | |
399 | * address in the routing table. | |
400 | */ | |
401 | if (CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_ONLINK)) { | |
402 | ifp = if_lookup_by_index(nexthop->ifindex, VRF_DEFAULT); | |
403 | if (ifp && connected_is_unnumbered(ifp)) { | |
404 | if (if_is_operative(ifp)) | |
405 | return 1; | |
406 | else | |
407 | return 0; | |
408 | } else | |
409 | return 0; | |
16814f96 | 410 | } |
718e3744 | 411 | |
d62a17ae | 412 | /* Make lookup prefix. */ |
413 | memset(&p, 0, sizeof(struct prefix)); | |
414 | switch (afi) { | |
415 | case AFI_IP: | |
416 | p.family = AF_INET; | |
417 | p.prefixlen = IPV4_MAX_PREFIXLEN; | |
418 | p.u.prefix4 = nexthop->gate.ipv4; | |
419 | break; | |
420 | case AFI_IP6: | |
421 | p.family = AF_INET6; | |
422 | p.prefixlen = IPV6_MAX_PREFIXLEN; | |
423 | p.u.prefix6 = nexthop->gate.ipv6; | |
424 | break; | |
425 | default: | |
426 | assert(afi != AFI_IP && afi != AFI_IP6); | |
427 | break; | |
718e3744 | 428 | } |
d62a17ae | 429 | /* Lookup table. */ |
430 | table = zebra_vrf_table(afi, SAFI_UNICAST, re->vrf_id); | |
431 | if (!table) | |
432 | return 0; | |
433 | ||
434 | rn = route_node_match(table, (struct prefix *)&p); | |
435 | while (rn) { | |
436 | route_unlock_node(rn); | |
437 | ||
438 | /* If lookup self prefix return immediately. */ | |
439 | if (rn == top) | |
440 | return 0; | |
441 | ||
442 | /* Pick up selected route. */ | |
443 | /* However, do not resolve over default route unless explicitly | |
444 | * allowed. */ | |
445 | if (is_default_prefix(&rn->p) | |
446 | && !nh_resolve_via_default(p.family)) | |
447 | return 0; | |
448 | ||
449 | RNODE_FOREACH_RE(rn, match) | |
450 | { | |
451 | if (CHECK_FLAG(match->status, ROUTE_ENTRY_REMOVED)) | |
452 | continue; | |
453 | ||
454 | /* if the next hop is imported from another table, skip | |
455 | * it */ | |
456 | if (match->type == ZEBRA_ROUTE_TABLE) | |
457 | continue; | |
458 | if (CHECK_FLAG(match->status, ROUTE_ENTRY_SELECTED_FIB)) | |
459 | break; | |
460 | } | |
48a53dc7 | 461 | |
d62a17ae | 462 | /* If there is no selected route or matched route is EGP, go up |
463 | tree. */ | |
464 | if (!match) { | |
465 | do { | |
466 | rn = rn->parent; | |
467 | } while (rn && rn->info == NULL); | |
468 | if (rn) | |
469 | route_lock_node(rn); | |
c87bdd2b | 470 | |
d62a17ae | 471 | continue; |
472 | } | |
473 | ||
474 | /* If the longest prefix match for the nexthop yields | |
475 | * a blackhole, mark it as inactive. */ | |
476 | if (CHECK_FLAG(match->flags, ZEBRA_FLAG_BLACKHOLE) | |
477 | || CHECK_FLAG(match->flags, ZEBRA_FLAG_REJECT)) | |
478 | return 0; | |
479 | ||
480 | if (match->type == ZEBRA_ROUTE_CONNECT) { | |
481 | /* Directly point connected route. */ | |
482 | newhop = match->nexthop; | |
483 | if (newhop) { | |
484 | if (nexthop->type == NEXTHOP_TYPE_IPV4 | |
485 | || nexthop->type == NEXTHOP_TYPE_IPV6) | |
486 | nexthop->ifindex = newhop->ifindex; | |
487 | } | |
488 | return 1; | |
489 | } else if (CHECK_FLAG(re->flags, ZEBRA_FLAG_INTERNAL)) { | |
490 | resolved = 0; | |
491 | for (ALL_NEXTHOPS(match->nexthop, newhop)) | |
492 | if (CHECK_FLAG(newhop->flags, NEXTHOP_FLAG_FIB) | |
493 | && !CHECK_FLAG(newhop->flags, | |
494 | NEXTHOP_FLAG_RECURSIVE)) { | |
495 | if (set) { | |
496 | SET_FLAG( | |
497 | nexthop->flags, | |
498 | NEXTHOP_FLAG_RECURSIVE); | |
499 | SET_FLAG( | |
500 | re->status, | |
501 | ROUTE_ENTRY_NEXTHOPS_CHANGED); | |
502 | ||
503 | nexthop_set_resolved( | |
504 | afi, newhop, nexthop); | |
505 | } | |
506 | resolved = 1; | |
507 | } | |
508 | return resolved; | |
509 | } else if (re->type == ZEBRA_ROUTE_STATIC) { | |
510 | resolved = 0; | |
511 | for (ALL_NEXTHOPS(match->nexthop, newhop)) | |
512 | if (CHECK_FLAG(newhop->flags, | |
513 | NEXTHOP_FLAG_FIB)) { | |
514 | if (set) { | |
515 | SET_FLAG( | |
516 | nexthop->flags, | |
517 | NEXTHOP_FLAG_RECURSIVE); | |
518 | ||
519 | nexthop_set_resolved( | |
520 | afi, newhop, nexthop); | |
521 | } | |
522 | resolved = 1; | |
523 | } | |
524 | if (resolved && set) | |
525 | re->nexthop_mtu = match->mtu; | |
526 | return resolved; | |
527 | } else { | |
528 | return 0; | |
529 | } | |
718e3744 | 530 | } |
d62a17ae | 531 | return 0; |
718e3744 | 532 | } |
718e3744 | 533 | |
d62a17ae | 534 | struct route_entry *rib_match(afi_t afi, safi_t safi, vrf_id_t vrf_id, |
535 | union g_addr *addr, struct route_node **rn_out) | |
536 | { | |
537 | struct prefix p; | |
538 | struct route_table *table; | |
539 | struct route_node *rn; | |
540 | struct route_entry *match; | |
541 | struct nexthop *newhop; | |
542 | ||
543 | /* Lookup table. */ | |
544 | table = zebra_vrf_table(afi, safi, vrf_id); | |
545 | if (!table) | |
546 | return 0; | |
547 | ||
548 | memset(&p, 0, sizeof(struct prefix)); | |
549 | p.family = afi; | |
550 | if (afi == AFI_IP) { | |
551 | p.u.prefix4 = addr->ipv4; | |
552 | p.prefixlen = IPV4_MAX_PREFIXLEN; | |
553 | } else { | |
554 | p.u.prefix6 = addr->ipv6; | |
555 | p.prefixlen = IPV6_MAX_PREFIXLEN; | |
556 | } | |
718e3744 | 557 | |
d62a17ae | 558 | rn = route_node_match(table, (struct prefix *)&p); |
718e3744 | 559 | |
d62a17ae | 560 | while (rn) { |
561 | route_unlock_node(rn); | |
718e3744 | 562 | |
d62a17ae | 563 | /* Pick up selected route. */ |
564 | RNODE_FOREACH_RE(rn, match) | |
565 | { | |
566 | if (CHECK_FLAG(match->status, ROUTE_ENTRY_REMOVED)) | |
567 | continue; | |
568 | if (CHECK_FLAG(match->status, ROUTE_ENTRY_SELECTED_FIB)) | |
569 | break; | |
570 | } | |
718e3744 | 571 | |
d62a17ae | 572 | /* If there is no selected route or matched route is EGP, go up |
573 | tree. */ | |
574 | if (!match) { | |
575 | do { | |
576 | rn = rn->parent; | |
577 | } while (rn && rn->info == NULL); | |
578 | if (rn) | |
579 | route_lock_node(rn); | |
580 | } else { | |
581 | if (match->type != ZEBRA_ROUTE_CONNECT) { | |
582 | int found = 0; | |
583 | for (ALL_NEXTHOPS(match->nexthop, newhop)) | |
584 | if (CHECK_FLAG(newhop->flags, | |
585 | NEXTHOP_FLAG_FIB)) { | |
586 | found = 1; | |
587 | break; | |
588 | } | |
589 | if (!found) | |
590 | return NULL; | |
591 | } | |
592 | ||
593 | if (rn_out) | |
594 | *rn_out = rn; | |
595 | return match; | |
596 | } | |
16814f96 | 597 | } |
d62a17ae | 598 | return NULL; |
599 | } | |
600 | ||
601 | struct route_entry *rib_match_ipv4_multicast(vrf_id_t vrf_id, | |
602 | struct in_addr addr, | |
603 | struct route_node **rn_out) | |
604 | { | |
605 | struct route_entry *re = NULL, *mre = NULL, *ure = NULL; | |
606 | struct route_node *m_rn = NULL, *u_rn = NULL; | |
607 | union g_addr gaddr = {.ipv4 = addr}; | |
608 | ||
609 | switch (ipv4_multicast_mode) { | |
610 | case MCAST_MRIB_ONLY: | |
611 | return rib_match(AFI_IP, SAFI_MULTICAST, vrf_id, &gaddr, | |
612 | rn_out); | |
613 | case MCAST_URIB_ONLY: | |
614 | return rib_match(AFI_IP, SAFI_UNICAST, vrf_id, &gaddr, rn_out); | |
615 | case MCAST_NO_CONFIG: | |
616 | case MCAST_MIX_MRIB_FIRST: | |
617 | re = mre = rib_match(AFI_IP, SAFI_MULTICAST, vrf_id, &gaddr, | |
618 | &m_rn); | |
619 | if (!mre) | |
620 | re = ure = rib_match(AFI_IP, SAFI_UNICAST, vrf_id, | |
621 | &gaddr, &u_rn); | |
622 | break; | |
623 | case MCAST_MIX_DISTANCE: | |
624 | mre = rib_match(AFI_IP, SAFI_MULTICAST, vrf_id, &gaddr, &m_rn); | |
625 | ure = rib_match(AFI_IP, SAFI_UNICAST, vrf_id, &gaddr, &u_rn); | |
626 | if (mre && ure) | |
627 | re = ure->distance < mre->distance ? ure : mre; | |
628 | else if (mre) | |
629 | re = mre; | |
630 | else if (ure) | |
631 | re = ure; | |
632 | break; | |
633 | case MCAST_MIX_PFXLEN: | |
634 | mre = rib_match(AFI_IP, SAFI_MULTICAST, vrf_id, &gaddr, &m_rn); | |
635 | ure = rib_match(AFI_IP, SAFI_UNICAST, vrf_id, &gaddr, &u_rn); | |
636 | if (mre && ure) | |
637 | re = u_rn->p.prefixlen > m_rn->p.prefixlen ? ure : mre; | |
638 | else if (mre) | |
639 | re = mre; | |
640 | else if (ure) | |
641 | re = ure; | |
642 | break; | |
718e3744 | 643 | } |
33550aa8 | 644 | |
d62a17ae | 645 | if (rn_out) |
646 | *rn_out = (re == mre) ? m_rn : u_rn; | |
718e3744 | 647 | |
d62a17ae | 648 | if (IS_ZEBRA_DEBUG_RIB) { |
649 | char buf[BUFSIZ]; | |
650 | inet_ntop(AF_INET, &addr, buf, BUFSIZ); | |
4623d897 | 651 | |
d62a17ae | 652 | zlog_debug("%s: %s: found %s, using %s", __func__, buf, |
653 | mre ? (ure ? "MRIB+URIB" : "MRIB") | |
654 | : ure ? "URIB" : "nothing", | |
655 | re == ure ? "URIB" : re == mre ? "MRIB" : "none"); | |
656 | } | |
657 | return re; | |
4623d897 DL |
658 | } |
659 | ||
d62a17ae | 660 | void multicast_mode_ipv4_set(enum multicast_mode mode) |
4623d897 | 661 | { |
d62a17ae | 662 | if (IS_ZEBRA_DEBUG_RIB) |
663 | zlog_debug("%s: multicast lookup mode set (%d)", __func__, | |
664 | mode); | |
665 | ipv4_multicast_mode = mode; | |
4623d897 DL |
666 | } |
667 | ||
d62a17ae | 668 | enum multicast_mode multicast_mode_ipv4_get(void) |
4623d897 | 669 | { |
d62a17ae | 670 | return ipv4_multicast_mode; |
4623d897 DL |
671 | } |
672 | ||
d62a17ae | 673 | struct route_entry *rib_lookup_ipv4(struct prefix_ipv4 *p, vrf_id_t vrf_id) |
718e3744 | 674 | { |
d62a17ae | 675 | struct route_table *table; |
676 | struct route_node *rn; | |
677 | struct route_entry *match; | |
678 | struct nexthop *nexthop; | |
718e3744 | 679 | |
d62a17ae | 680 | /* Lookup table. */ |
681 | table = zebra_vrf_table(AFI_IP, SAFI_UNICAST, vrf_id); | |
682 | if (!table) | |
683 | return 0; | |
718e3744 | 684 | |
d62a17ae | 685 | rn = route_node_lookup(table, (struct prefix *)p); |
718e3744 | 686 | |
d62a17ae | 687 | /* No route for this prefix. */ |
688 | if (!rn) | |
689 | return NULL; | |
718e3744 | 690 | |
d62a17ae | 691 | /* Unlock node. */ |
692 | route_unlock_node(rn); | |
718e3744 | 693 | |
d62a17ae | 694 | RNODE_FOREACH_RE(rn, match) |
695 | { | |
696 | if (CHECK_FLAG(match->status, ROUTE_ENTRY_REMOVED)) | |
697 | continue; | |
698 | if (CHECK_FLAG(match->status, ROUTE_ENTRY_SELECTED_FIB)) | |
699 | break; | |
700 | } | |
718e3744 | 701 | |
d62a17ae | 702 | if (!match) |
703 | return NULL; | |
718e3744 | 704 | |
d62a17ae | 705 | if (match->type == ZEBRA_ROUTE_CONNECT) |
706 | return match; | |
f9e1b38e | 707 | |
d62a17ae | 708 | for (ALL_NEXTHOPS(match->nexthop, nexthop)) |
709 | if (CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_FIB)) | |
710 | return match; | |
718e3744 | 711 | |
d62a17ae | 712 | return NULL; |
718e3744 | 713 | } |
714 | ||
dc95824a DO |
715 | /* |
716 | * This clone function, unlike its original rib_lookup_ipv4(), checks | |
717 | * if specified IPv4 route record (prefix/mask -> gate) exists in | |
f0f77c9a | 718 | * the whole RIB and has ROUTE_ENTRY_SELECTED_FIB set. |
dc95824a DO |
719 | * |
720 | * Return values: | |
721 | * -1: error | |
722 | * 0: exact match found | |
723 | * 1: a match was found with a different gate | |
724 | * 2: connected route found | |
725 | * 3: no matches found | |
726 | */ | |
d62a17ae | 727 | int rib_lookup_ipv4_route(struct prefix_ipv4 *p, union sockunion *qgate, |
728 | vrf_id_t vrf_id) | |
dc95824a | 729 | { |
d62a17ae | 730 | struct route_table *table; |
731 | struct route_node *rn; | |
732 | struct route_entry *match; | |
733 | struct nexthop *nexthop; | |
734 | int nexthops_active; | |
dc95824a | 735 | |
d62a17ae | 736 | /* Lookup table. */ |
737 | table = zebra_vrf_table(AFI_IP, SAFI_UNICAST, vrf_id); | |
738 | if (!table) | |
739 | return ZEBRA_RIB_LOOKUP_ERROR; | |
dc95824a | 740 | |
d62a17ae | 741 | /* Scan the RIB table for exactly matching RIB entry. */ |
742 | rn = route_node_lookup(table, (struct prefix *)p); | |
dc95824a | 743 | |
d62a17ae | 744 | /* No route for this prefix. */ |
745 | if (!rn) | |
746 | return ZEBRA_RIB_NOTFOUND; | |
dc95824a | 747 | |
d62a17ae | 748 | /* Unlock node. */ |
749 | route_unlock_node(rn); | |
dc95824a | 750 | |
d62a17ae | 751 | /* Find out if a "selected" RR for the discovered RIB entry exists ever. |
752 | */ | |
753 | RNODE_FOREACH_RE(rn, match) | |
754 | { | |
755 | if (CHECK_FLAG(match->status, ROUTE_ENTRY_REMOVED)) | |
756 | continue; | |
757 | if (CHECK_FLAG(match->status, ROUTE_ENTRY_SELECTED_FIB)) | |
758 | break; | |
759 | } | |
dc95824a | 760 | |
d62a17ae | 761 | /* None such found :( */ |
762 | if (!match) | |
763 | return ZEBRA_RIB_NOTFOUND; | |
764 | ||
765 | if (match->type == ZEBRA_ROUTE_CONNECT) | |
766 | return ZEBRA_RIB_FOUND_CONNECTED; | |
767 | ||
768 | /* Ok, we have a cood candidate, let's check it's nexthop list... */ | |
769 | nexthops_active = 0; | |
770 | for (ALL_NEXTHOPS(match->nexthop, nexthop)) | |
771 | if (CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_FIB)) { | |
772 | nexthops_active = 1; | |
773 | if (nexthop->gate.ipv4.s_addr == sockunion2ip(qgate)) | |
774 | return ZEBRA_RIB_FOUND_EXACT; | |
775 | if (IS_ZEBRA_DEBUG_RIB) { | |
776 | char gate_buf[INET_ADDRSTRLEN], | |
777 | qgate_buf[INET_ADDRSTRLEN]; | |
778 | inet_ntop(AF_INET, &nexthop->gate.ipv4.s_addr, | |
779 | gate_buf, INET_ADDRSTRLEN); | |
780 | inet_ntop(AF_INET, &sockunion2ip(qgate), | |
781 | qgate_buf, INET_ADDRSTRLEN); | |
782 | zlog_debug("%s: qgate == %s, %s == %s", | |
783 | __func__, qgate_buf, | |
784 | nexthop->rparent ? "rgate" : "gate", | |
785 | gate_buf); | |
786 | } | |
787 | } | |
788 | ||
789 | if (nexthops_active) | |
790 | return ZEBRA_RIB_FOUND_NOGATE; | |
791 | ||
792 | return ZEBRA_RIB_NOTFOUND; | |
793 | } | |
794 | ||
795 | #define RIB_SYSTEM_ROUTE(R) \ | |
796 | ((R)->type == ZEBRA_ROUTE_KERNEL || (R)->type == ZEBRA_ROUTE_CONNECT) | |
7514fb77 | 797 | |
dc95824a DO |
798 | /* This function verifies reachability of one given nexthop, which can be |
799 | * numbered or unnumbered, IPv4 or IPv6. The result is unconditionally stored | |
800 | * in nexthop->flags field. If the 4th parameter, 'set', is non-zero, | |
801 | * nexthop->ifindex will be updated appropriately as well. | |
802 | * An existing route map can turn (otherwise active) nexthop into inactive, but | |
803 | * not vice versa. | |
804 | * | |
805 | * The return value is the final value of 'ACTIVE' flag. | |
806 | */ | |
807 | ||
d62a17ae | 808 | static unsigned nexthop_active_check(struct route_node *rn, |
809 | struct route_entry *re, | |
810 | struct nexthop *nexthop, int set) | |
811 | { | |
812 | struct interface *ifp; | |
813 | route_map_result_t ret = RMAP_MATCH; | |
814 | int family; | |
815 | char buf[SRCDEST2STR_BUFFER]; | |
816 | struct prefix *p, *src_p; | |
817 | srcdest_rnode_prefixes(rn, &p, &src_p); | |
818 | ||
819 | if (rn->p.family == AF_INET) | |
820 | family = AFI_IP; | |
821 | else if (rn->p.family == AF_INET6) | |
822 | family = AFI_IP6; | |
823 | else | |
824 | family = 0; | |
825 | switch (nexthop->type) { | |
826 | case NEXTHOP_TYPE_IFINDEX: | |
827 | ifp = if_lookup_by_index(nexthop->ifindex, re->vrf_id); | |
828 | if (ifp && if_is_operative(ifp)) | |
829 | SET_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE); | |
830 | else | |
831 | UNSET_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE); | |
832 | break; | |
833 | case NEXTHOP_TYPE_IPV4: | |
834 | case NEXTHOP_TYPE_IPV4_IFINDEX: | |
835 | family = AFI_IP; | |
836 | if (nexthop_active(AFI_IP, re, nexthop, set, rn)) | |
837 | SET_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE); | |
838 | else | |
839 | UNSET_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE); | |
840 | break; | |
841 | case NEXTHOP_TYPE_IPV6: | |
842 | family = AFI_IP6; | |
843 | if (nexthop_active(AFI_IP6, re, nexthop, set, rn)) | |
844 | SET_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE); | |
845 | else | |
846 | UNSET_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE); | |
847 | break; | |
848 | case NEXTHOP_TYPE_IPV6_IFINDEX: | |
849 | /* RFC 5549, v4 prefix with v6 NH */ | |
850 | if (rn->p.family != AF_INET) | |
851 | family = AFI_IP6; | |
852 | if (IN6_IS_ADDR_LINKLOCAL(&nexthop->gate.ipv6)) { | |
853 | ifp = if_lookup_by_index(nexthop->ifindex, re->vrf_id); | |
854 | if (ifp && if_is_operative(ifp)) | |
855 | SET_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE); | |
856 | else | |
857 | UNSET_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE); | |
858 | } else { | |
859 | if (nexthop_active(AFI_IP6, re, nexthop, set, rn)) | |
860 | SET_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE); | |
861 | else | |
862 | UNSET_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE); | |
863 | } | |
864 | break; | |
865 | case NEXTHOP_TYPE_BLACKHOLE: | |
866 | SET_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE); | |
867 | break; | |
868 | default: | |
869 | break; | |
870 | } | |
871 | if (!CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE)) | |
872 | return 0; | |
873 | ||
874 | /* XXX: What exactly do those checks do? Do we support | |
875 | * e.g. IPv4 routes with IPv6 nexthops or vice versa? */ | |
876 | if (RIB_SYSTEM_ROUTE(re) || (family == AFI_IP && p->family != AF_INET) | |
877 | || (family == AFI_IP6 && p->family != AF_INET6)) | |
878 | return CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE); | |
879 | ||
880 | /* The original code didn't determine the family correctly | |
881 | * e.g. for NEXTHOP_TYPE_IFINDEX. Retrieve the correct afi | |
882 | * from the rib_table_info in those cases. | |
883 | * Possibly it may be better to use only the rib_table_info | |
884 | * in every case. | |
885 | */ | |
886 | if (!family) { | |
887 | rib_table_info_t *info; | |
888 | ||
889 | info = srcdest_rnode_table_info(rn); | |
890 | family = info->afi; | |
718e3744 | 891 | } |
c52ef59f | 892 | |
d62a17ae | 893 | memset(&nexthop->rmap_src.ipv6, 0, sizeof(union g_addr)); |
894 | ||
895 | /* It'll get set if required inside */ | |
896 | ret = zebra_route_map_check(family, re->type, p, nexthop, re->vrf_id, | |
897 | re->tag); | |
898 | if (ret == RMAP_DENYMATCH) { | |
899 | if (IS_ZEBRA_DEBUG_RIB) { | |
900 | srcdest_rnode2str(rn, buf, sizeof(buf)); | |
901 | zlog_debug( | |
902 | "%u:%s: Filtering out with NH out %s due to route map", | |
903 | re->vrf_id, buf, | |
904 | ifindex2ifname(nexthop->ifindex, re->vrf_id)); | |
905 | } | |
906 | UNSET_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE); | |
518f0eb1 | 907 | } |
d62a17ae | 908 | return CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE); |
718e3744 | 909 | } |
910 | ||
03e232a4 | 911 | /* Iterate over all nexthops of the given RIB entry and refresh their |
f0f77c9a DS |
912 | * ACTIVE flag. re->nexthop_active_num is updated accordingly. If any |
913 | * nexthop is found to toggle the ACTIVE flag, the whole re structure | |
914 | * is flagged with ROUTE_ENTRY_CHANGED. The 4th 'set' argument is | |
03e232a4 DO |
915 | * transparently passed to nexthop_active_check(). |
916 | * | |
917 | * Return value is the new number of active nexthops. | |
918 | */ | |
919 | ||
d62a17ae | 920 | static int nexthop_active_update(struct route_node *rn, struct route_entry *re, |
921 | int set) | |
922 | { | |
923 | struct nexthop *nexthop; | |
924 | union g_addr prev_src; | |
925 | unsigned int prev_active, new_active, old_num_nh; | |
926 | ifindex_t prev_index; | |
927 | old_num_nh = re->nexthop_active_num; | |
928 | ||
929 | re->nexthop_active_num = 0; | |
930 | UNSET_FLAG(re->status, ROUTE_ENTRY_CHANGED); | |
931 | ||
932 | for (nexthop = re->nexthop; nexthop; nexthop = nexthop->next) { | |
933 | /* No protocol daemon provides src and so we're skipping | |
934 | * tracking it */ | |
935 | prev_src = nexthop->rmap_src; | |
936 | prev_active = CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE); | |
937 | prev_index = nexthop->ifindex; | |
938 | if ((new_active = nexthop_active_check(rn, re, nexthop, set))) | |
939 | re->nexthop_active_num++; | |
940 | /* Don't allow src setting on IPv6 addr for now */ | |
941 | if (prev_active != new_active || prev_index != nexthop->ifindex | |
942 | || ((nexthop->type >= NEXTHOP_TYPE_IFINDEX | |
943 | && nexthop->type < NEXTHOP_TYPE_IPV6) | |
944 | && prev_src.ipv4.s_addr | |
945 | != nexthop->rmap_src.ipv4.s_addr) | |
946 | || ((nexthop->type >= NEXTHOP_TYPE_IPV6 | |
947 | && nexthop->type < NEXTHOP_TYPE_BLACKHOLE) | |
948 | && !(IPV6_ADDR_SAME(&prev_src.ipv6, | |
949 | &nexthop->rmap_src.ipv6)))) { | |
950 | SET_FLAG(re->status, ROUTE_ENTRY_CHANGED); | |
951 | SET_FLAG(re->status, ROUTE_ENTRY_NEXTHOPS_CHANGED); | |
952 | } | |
953 | } | |
954 | ||
955 | if (old_num_nh != re->nexthop_active_num) | |
956 | SET_FLAG(re->status, ROUTE_ENTRY_CHANGED); | |
6e26278c | 957 | |
d62a17ae | 958 | if (CHECK_FLAG(re->status, ROUTE_ENTRY_CHANGED)) { |
959 | SET_FLAG(re->status, ROUTE_ENTRY_NEXTHOPS_CHANGED); | |
960 | } | |
961 | ||
962 | return re->nexthop_active_num; | |
718e3744 | 963 | } |
6baeb988 | 964 | |
a64448ba DS |
965 | /* |
966 | * Is this RIB labeled-unicast? It must be of type BGP and all paths | |
967 | * (nexthops) must have a label. | |
968 | */ | |
d62a17ae | 969 | int zebra_rib_labeled_unicast(struct route_entry *re) |
a64448ba | 970 | { |
d62a17ae | 971 | struct nexthop *nexthop = NULL; |
a64448ba | 972 | |
d62a17ae | 973 | if (re->type != ZEBRA_ROUTE_BGP) |
974 | return 0; | |
a64448ba | 975 | |
d62a17ae | 976 | for (ALL_NEXTHOPS(re->nexthop, nexthop)) |
977 | if (!nexthop->nh_label || !nexthop->nh_label->num_labels) | |
978 | return 0; | |
6b0655a2 | 979 | |
d62a17ae | 980 | return 1; |
a64448ba | 981 | } |
718e3744 | 982 | |
6ae24471 DS |
983 | /* Update flag indicates whether this is a "replace" or not. Currently, this |
984 | * is only used for IPv4. | |
985 | */ | |
d62a17ae | 986 | int rib_install_kernel(struct route_node *rn, struct route_entry *re, |
987 | struct route_entry *old) | |
718e3744 | 988 | { |
d62a17ae | 989 | int ret = 0; |
990 | struct nexthop *nexthop; | |
991 | rib_table_info_t *info = srcdest_rnode_table_info(rn); | |
992 | struct prefix *p, *src_p; | |
993 | struct zebra_vrf *zvrf = vrf_info_lookup(re->vrf_id); | |
718e3744 | 994 | |
d62a17ae | 995 | srcdest_rnode_prefixes(rn, &p, &src_p); |
416ec78d | 996 | |
d62a17ae | 997 | if (info->safi != SAFI_UNICAST) { |
998 | for (ALL_NEXTHOPS(re->nexthop, nexthop)) | |
999 | SET_FLAG(nexthop->flags, NEXTHOP_FLAG_FIB); | |
1000 | return ret; | |
1001 | } | |
718e3744 | 1002 | |
d62a17ae | 1003 | /* |
1004 | * Make sure we update the FPM any time we send new information to | |
1005 | * the kernel. | |
1006 | */ | |
1007 | hook_call(rib_update, rn, "installing in kernel"); | |
1008 | ret = kernel_route_rib(p, src_p, old, re); | |
1009 | zvrf->installs++; | |
1010 | ||
1011 | /* If install succeeds, update FIB flag for nexthops. */ | |
1012 | if (!ret) { | |
1013 | for (ALL_NEXTHOPS(re->nexthop, nexthop)) { | |
1014 | if (CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_RECURSIVE)) | |
1015 | continue; | |
1016 | ||
1017 | if (CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE)) | |
1018 | SET_FLAG(nexthop->flags, NEXTHOP_FLAG_FIB); | |
1019 | else | |
1020 | UNSET_FLAG(nexthop->flags, NEXTHOP_FLAG_FIB); | |
1021 | } | |
1022 | } | |
3e5c6e00 | 1023 | |
d62a17ae | 1024 | return ret; |
718e3744 | 1025 | } |
1026 | ||
1027 | /* Uninstall the route from kernel. */ | |
d62a17ae | 1028 | int rib_uninstall_kernel(struct route_node *rn, struct route_entry *re) |
718e3744 | 1029 | { |
d62a17ae | 1030 | int ret = 0; |
1031 | struct nexthop *nexthop; | |
1032 | rib_table_info_t *info = srcdest_rnode_table_info(rn); | |
1033 | struct prefix *p, *src_p; | |
1034 | struct zebra_vrf *zvrf = vrf_info_lookup(re->vrf_id); | |
05737783 | 1035 | |
d62a17ae | 1036 | srcdest_rnode_prefixes(rn, &p, &src_p); |
718e3744 | 1037 | |
d62a17ae | 1038 | if (info->safi != SAFI_UNICAST) { |
1039 | for (ALL_NEXTHOPS(re->nexthop, nexthop)) | |
1040 | SET_FLAG(nexthop->flags, NEXTHOP_FLAG_FIB); | |
1041 | return ret; | |
1042 | } | |
416ec78d | 1043 | |
d62a17ae | 1044 | /* |
1045 | * Make sure we update the FPM any time we send new information to | |
1046 | * the kernel. | |
1047 | */ | |
1048 | hook_call(rib_update, rn, "uninstalling from kernel"); | |
1049 | ret = kernel_route_rib(p, src_p, re, NULL); | |
1050 | zvrf->removals++; | |
718e3744 | 1051 | |
d62a17ae | 1052 | for (ALL_NEXTHOPS(re->nexthop, nexthop)) |
1053 | UNSET_FLAG(nexthop->flags, NEXTHOP_FLAG_FIB); | |
718e3744 | 1054 | |
d62a17ae | 1055 | return ret; |
718e3744 | 1056 | } |
1057 | ||
1058 | /* Uninstall the route from kernel. */ | |
d62a17ae | 1059 | static void rib_uninstall(struct route_node *rn, struct route_entry *re) |
718e3744 | 1060 | { |
d62a17ae | 1061 | rib_table_info_t *info = srcdest_rnode_table_info(rn); |
416ec78d | 1062 | |
d62a17ae | 1063 | if (CHECK_FLAG(re->status, ROUTE_ENTRY_SELECTED_FIB)) { |
1064 | if (info->safi == SAFI_UNICAST) | |
1065 | hook_call(rib_update, rn, "rib_uninstall"); | |
5adc2528 | 1066 | |
d62a17ae | 1067 | if (!RIB_SYSTEM_ROUTE(re)) |
1068 | rib_uninstall_kernel(rn, re); | |
a64448ba | 1069 | |
d62a17ae | 1070 | /* If labeled-unicast route, uninstall transit LSP. */ |
1071 | if (zebra_rib_labeled_unicast(re)) | |
1072 | zebra_mpls_lsp_uninstall(info->zvrf, rn, re); | |
a64448ba | 1073 | |
d62a17ae | 1074 | UNSET_FLAG(re->status, ROUTE_ENTRY_SELECTED_FIB); |
1075 | } | |
446bb95e | 1076 | |
d62a17ae | 1077 | if (CHECK_FLAG(re->flags, ZEBRA_FLAG_SELECTED)) { |
1078 | struct prefix *p, *src_p; | |
1079 | srcdest_rnode_prefixes(rn, &p, &src_p); | |
05737783 | 1080 | |
d62a17ae | 1081 | redistribute_delete(p, src_p, re); |
1082 | UNSET_FLAG(re->flags, ZEBRA_FLAG_SELECTED); | |
1083 | } | |
718e3744 | 1084 | } |
1085 | ||
9fd92e3c AS |
1086 | /* |
1087 | * rib_can_delete_dest | |
1088 | * | |
1089 | * Returns TRUE if the given dest can be deleted from the table. | |
1090 | */ | |
d62a17ae | 1091 | static int rib_can_delete_dest(rib_dest_t *dest) |
9fd92e3c | 1092 | { |
d62a17ae | 1093 | if (dest->routes) { |
1094 | return 0; | |
1095 | } | |
9fd92e3c | 1096 | |
d62a17ae | 1097 | /* |
1098 | * Don't delete the dest if we have to update the FPM about this | |
1099 | * prefix. | |
1100 | */ | |
1101 | if (CHECK_FLAG(dest->flags, RIB_DEST_UPDATE_FPM) | |
1102 | || CHECK_FLAG(dest->flags, RIB_DEST_SENT_TO_FPM)) | |
1103 | return 0; | |
5adc2528 | 1104 | |
d62a17ae | 1105 | return 1; |
9fd92e3c AS |
1106 | } |
1107 | ||
1108 | /* | |
1109 | * rib_gc_dest | |
1110 | * | |
1111 | * Garbage collect the rib dest corresponding to the given route node | |
1112 | * if appropriate. | |
1113 | * | |
1114 | * Returns TRUE if the dest was deleted, FALSE otherwise. | |
1115 | */ | |
d62a17ae | 1116 | int rib_gc_dest(struct route_node *rn) |
9fd92e3c | 1117 | { |
d62a17ae | 1118 | rib_dest_t *dest; |
1119 | struct zebra_vrf *zvrf; | |
9fd92e3c | 1120 | |
d62a17ae | 1121 | dest = rib_dest_from_rnode(rn); |
1122 | if (!dest) | |
1123 | return 0; | |
9fd92e3c | 1124 | |
d62a17ae | 1125 | if (!rib_can_delete_dest(dest)) |
1126 | return 0; | |
9fd92e3c | 1127 | |
d62a17ae | 1128 | zvrf = rib_dest_vrf(dest); |
1129 | if (IS_ZEBRA_DEBUG_RIB) | |
1130 | rnode_debug(rn, zvrf_id(zvrf), "removing dest from table"); | |
9fd92e3c | 1131 | |
d62a17ae | 1132 | dest->rnode = NULL; |
1133 | XFREE(MTYPE_RIB_DEST, dest); | |
1134 | rn->info = NULL; | |
9fd92e3c | 1135 | |
d62a17ae | 1136 | /* |
1137 | * Release the one reference that we keep on the route node. | |
1138 | */ | |
1139 | route_unlock_node(rn); | |
1140 | return 1; | |
9fd92e3c AS |
1141 | } |
1142 | ||
d62a17ae | 1143 | static void rib_process_add_fib(struct zebra_vrf *zvrf, struct route_node *rn, |
1144 | struct route_entry *new) | |
3e5c6e00 | 1145 | { |
d62a17ae | 1146 | hook_call(rib_update, rn, "new route selected"); |
3e5c6e00 | 1147 | |
d62a17ae | 1148 | /* Update real nexthop. This may actually determine if nexthop is active |
1149 | * or not. */ | |
1150 | if (!nexthop_active_update(rn, new, 1)) { | |
1151 | UNSET_FLAG(new->status, ROUTE_ENTRY_CHANGED); | |
1152 | return; | |
1153 | } | |
3e5c6e00 | 1154 | |
d62a17ae | 1155 | SET_FLAG(new->status, ROUTE_ENTRY_SELECTED_FIB); |
1156 | if (IS_ZEBRA_DEBUG_RIB) { | |
1157 | char buf[SRCDEST2STR_BUFFER]; | |
1158 | srcdest_rnode2str(rn, buf, sizeof(buf)); | |
1159 | zlog_debug("%u:%s: Adding route rn %p, re %p (type %d)", | |
1160 | zvrf_id(zvrf), buf, rn, new, new->type); | |
1161 | } | |
3e5c6e00 | 1162 | |
d62a17ae | 1163 | /* If labeled-unicast route, install transit LSP. */ |
1164 | if (zebra_rib_labeled_unicast(new)) | |
1165 | zebra_mpls_lsp_install(zvrf, rn, new); | |
a64448ba | 1166 | |
d62a17ae | 1167 | if (!RIB_SYSTEM_ROUTE(new)) { |
1168 | if (rib_install_kernel(rn, new, NULL)) { | |
1169 | char buf[SRCDEST2STR_BUFFER]; | |
1170 | srcdest_rnode2str(rn, buf, sizeof(buf)); | |
1171 | zlog_warn("%u:%s: Route install failed", zvrf_id(zvrf), | |
1172 | buf); | |
1173 | } | |
1174 | } | |
3e5c6e00 | 1175 | |
d62a17ae | 1176 | UNSET_FLAG(new->status, ROUTE_ENTRY_CHANGED); |
3e5c6e00 | 1177 | } |
1178 | ||
d62a17ae | 1179 | static void rib_process_del_fib(struct zebra_vrf *zvrf, struct route_node *rn, |
1180 | struct route_entry *old) | |
3e5c6e00 | 1181 | { |
d62a17ae | 1182 | hook_call(rib_update, rn, "removing existing route"); |
3e5c6e00 | 1183 | |
d62a17ae | 1184 | /* Uninstall from kernel. */ |
1185 | if (IS_ZEBRA_DEBUG_RIB) { | |
1186 | char buf[SRCDEST2STR_BUFFER]; | |
1187 | srcdest_rnode2str(rn, buf, sizeof(buf)); | |
1188 | zlog_debug("%u:%s: Deleting route rn %p, re %p (type %d)", | |
1189 | zvrf_id(zvrf), buf, rn, old, old->type); | |
1190 | } | |
3e5c6e00 | 1191 | |
d62a17ae | 1192 | /* If labeled-unicast route, uninstall transit LSP. */ |
1193 | if (zebra_rib_labeled_unicast(old)) | |
1194 | zebra_mpls_lsp_uninstall(zvrf, rn, old); | |
1195 | ||
1196 | if (!RIB_SYSTEM_ROUTE(old)) | |
1197 | rib_uninstall_kernel(rn, old); | |
1198 | ||
1199 | UNSET_FLAG(old->status, ROUTE_ENTRY_SELECTED_FIB); | |
1200 | ||
1201 | /* Update nexthop for route, reset changed flag. */ | |
1202 | nexthop_active_update(rn, old, 1); | |
1203 | UNSET_FLAG(old->status, ROUTE_ENTRY_CHANGED); | |
1204 | } | |
1205 | ||
1206 | static void rib_process_update_fib(struct zebra_vrf *zvrf, | |
1207 | struct route_node *rn, | |
1208 | struct route_entry *old, | |
1209 | struct route_entry *new) | |
1210 | { | |
1211 | struct nexthop *nexthop = NULL; | |
1212 | int nh_active = 0; | |
1213 | int installed = 1; | |
1214 | ||
1215 | /* | |
1216 | * We have to install or update if a new route has been selected or | |
1217 | * something has changed. | |
1218 | */ | |
1219 | if (new != old || CHECK_FLAG(new->status, ROUTE_ENTRY_CHANGED)) { | |
1220 | hook_call(rib_update, rn, "updating existing route"); | |
1221 | ||
1222 | /* Update the nexthop; we could determine here that nexthop is | |
1223 | * inactive. */ | |
1224 | if (nexthop_active_update(rn, new, 1)) | |
1225 | nh_active = 1; | |
1226 | ||
1227 | /* If nexthop is active, install the selected route, if | |
1228 | * appropriate. If | |
1229 | * the install succeeds, cleanup flags for prior route, if | |
1230 | * different from | |
1231 | * newly selected. | |
1232 | */ | |
1233 | if (nh_active) { | |
1234 | if (IS_ZEBRA_DEBUG_RIB) { | |
1235 | char buf[SRCDEST2STR_BUFFER]; | |
1236 | srcdest_rnode2str(rn, buf, sizeof(buf)); | |
1237 | if (new != old) | |
1238 | zlog_debug( | |
1239 | "%u:%s: Updating route rn %p, re %p (type %d) " | |
1240 | "old %p (type %d)", | |
1241 | zvrf_id(zvrf), buf, rn, new, | |
1242 | new->type, old, old->type); | |
1243 | else | |
1244 | zlog_debug( | |
1245 | "%u:%s: Updating route rn %p, re %p (type %d)", | |
1246 | zvrf_id(zvrf), buf, rn, new, | |
1247 | new->type); | |
1248 | } | |
1249 | ||
1250 | /* If labeled-unicast route, uninstall transit LSP. */ | |
1251 | if (zebra_rib_labeled_unicast(old)) | |
1252 | zebra_mpls_lsp_uninstall(zvrf, rn, old); | |
1253 | ||
1254 | /* Non-system route should be installed. */ | |
1255 | if (!RIB_SYSTEM_ROUTE(new)) { | |
1256 | /* If labeled-unicast route, install transit | |
1257 | * LSP. */ | |
1258 | if (zebra_rib_labeled_unicast(new)) | |
1259 | zebra_mpls_lsp_install(zvrf, rn, new); | |
1260 | ||
1261 | if (rib_install_kernel(rn, new, old)) { | |
1262 | char buf[SRCDEST2STR_BUFFER]; | |
1263 | srcdest_rnode2str(rn, buf, sizeof(buf)); | |
1264 | installed = 0; | |
1265 | zlog_warn("%u:%s: Route install failed", | |
1266 | zvrf_id(zvrf), buf); | |
1267 | } | |
1268 | } | |
1269 | ||
1270 | /* If install succeeded or system route, cleanup flags | |
1271 | * for prior route. */ | |
1272 | if (installed && new != old) { | |
1273 | if (RIB_SYSTEM_ROUTE(new)) { | |
1274 | if (!RIB_SYSTEM_ROUTE(old)) | |
1275 | rib_uninstall_kernel(rn, old); | |
1276 | } else { | |
1277 | for (nexthop = old->nexthop; nexthop; | |
1278 | nexthop = nexthop->next) | |
1279 | UNSET_FLAG(nexthop->flags, | |
1280 | NEXTHOP_FLAG_FIB); | |
1281 | } | |
1282 | } | |
1283 | ||
1284 | /* Update for redistribution. */ | |
1285 | if (installed) | |
1286 | SET_FLAG(new->status, ROUTE_ENTRY_SELECTED_FIB); | |
1287 | } | |
a64448ba | 1288 | |
d62a17ae | 1289 | /* |
1290 | * If nexthop for selected route is not active or install | |
1291 | * failed, we | |
1292 | * may need to uninstall and delete for redistribution. | |
1293 | */ | |
1294 | if (!nh_active || !installed) { | |
1295 | if (IS_ZEBRA_DEBUG_RIB) { | |
1296 | char buf[SRCDEST2STR_BUFFER]; | |
1297 | srcdest_rnode2str(rn, buf, sizeof(buf)); | |
1298 | if (new != old) | |
1299 | zlog_debug( | |
1300 | "%u:%s: Deleting route rn %p, re %p (type %d) " | |
1301 | "old %p (type %d) - %s", | |
1302 | zvrf_id(zvrf), buf, rn, new, | |
1303 | new->type, old, old->type, | |
1304 | nh_active ? "install failed" | |
1305 | : "nexthop inactive"); | |
1306 | else | |
1307 | zlog_debug( | |
1308 | "%u:%s: Deleting route rn %p, re %p (type %d) - %s", | |
1309 | zvrf_id(zvrf), buf, rn, new, | |
1310 | new->type, | |
1311 | nh_active ? "install failed" | |
1312 | : "nexthop inactive"); | |
1313 | } | |
1314 | ||
1315 | /* If labeled-unicast route, uninstall transit LSP. */ | |
1316 | if (zebra_rib_labeled_unicast(old)) | |
1317 | zebra_mpls_lsp_uninstall(zvrf, rn, old); | |
1318 | ||
1319 | if (!RIB_SYSTEM_ROUTE(old)) | |
1320 | rib_uninstall_kernel(rn, old); | |
1321 | UNSET_FLAG(new->status, ROUTE_ENTRY_SELECTED_FIB); | |
1322 | } | |
1323 | } else { | |
1324 | /* | |
1325 | * Same route selected; check if in the FIB and if not, | |
1326 | * re-install. This | |
1327 | * is housekeeping code to deal with race conditions in kernel | |
1328 | * with linux | |
1329 | * netlink reporting interface up before IPv4 or IPv6 protocol | |
1330 | * is ready | |
1331 | * to add routes. | |
1332 | */ | |
1333 | if (!RIB_SYSTEM_ROUTE(new)) { | |
1334 | int in_fib = 0; | |
1335 | ||
1336 | for (ALL_NEXTHOPS(new->nexthop, nexthop)) | |
1337 | if (CHECK_FLAG(nexthop->flags, | |
1338 | NEXTHOP_FLAG_FIB)) { | |
1339 | in_fib = 1; | |
1340 | break; | |
1341 | } | |
1342 | if (!in_fib) | |
1343 | rib_install_kernel(rn, new, NULL); | |
1344 | } | |
1345 | } | |
1346 | ||
1347 | /* Update prior route. */ | |
1348 | if (new != old) { | |
1349 | UNSET_FLAG(old->status, ROUTE_ENTRY_SELECTED_FIB); | |
3e5c6e00 | 1350 | |
d62a17ae | 1351 | /* Set real nexthop. */ |
1352 | nexthop_active_update(rn, old, 1); | |
1353 | UNSET_FLAG(old->status, ROUTE_ENTRY_CHANGED); | |
1354 | } | |
3e5c6e00 | 1355 | |
d62a17ae | 1356 | /* Clear changed flag. */ |
1357 | UNSET_FLAG(new->status, ROUTE_ENTRY_CHANGED); | |
3e5c6e00 | 1358 | } |
1359 | ||
d62a17ae | 1360 | /* Check if 'alternate' RIB entry is better than 'current'. */ |
1361 | static struct route_entry *rib_choose_best(struct route_entry *current, | |
1362 | struct route_entry *alternate) | |
1363 | { | |
1364 | if (current == NULL) | |
1365 | return alternate; | |
1366 | ||
1367 | /* filter route selection in following order: | |
1368 | * - connected beats other types | |
1369 | * - lower distance beats higher | |
1370 | * - lower metric beats higher for equal distance | |
1371 | * - last, hence oldest, route wins tie break. | |
1372 | */ | |
1373 | ||
1374 | /* Connected routes. Pick the last connected | |
1375 | * route of the set of lowest metric connected routes. | |
1376 | */ | |
1377 | if (alternate->type == ZEBRA_ROUTE_CONNECT) { | |
1378 | if (current->type != ZEBRA_ROUTE_CONNECT | |
1379 | || alternate->metric <= current->metric) | |
1380 | return alternate; | |
1381 | ||
1382 | return current; | |
1383 | } | |
3e5c6e00 | 1384 | |
d62a17ae | 1385 | if (current->type == ZEBRA_ROUTE_CONNECT) |
1386 | return current; | |
3e5c6e00 | 1387 | |
d62a17ae | 1388 | /* higher distance loses */ |
1389 | if (alternate->distance < current->distance) | |
1390 | return alternate; | |
1391 | if (current->distance < alternate->distance) | |
1392 | return current; | |
3e5c6e00 | 1393 | |
d62a17ae | 1394 | /* metric tie-breaks equal distance */ |
1395 | if (alternate->metric <= current->metric) | |
1396 | return alternate; | |
3e5c6e00 | 1397 | |
d62a17ae | 1398 | return current; |
3e5c6e00 | 1399 | } |
1400 | ||
d62a17ae | 1401 | /* Core function for processing routing information base. */ |
1402 | static void rib_process(struct route_node *rn) | |
1403 | { | |
1404 | struct route_entry *re; | |
1405 | struct route_entry *next; | |
1406 | struct route_entry *old_selected = NULL; | |
1407 | struct route_entry *new_selected = NULL; | |
1408 | struct route_entry *old_fib = NULL; | |
1409 | struct route_entry *new_fib = NULL; | |
1410 | struct route_entry *best = NULL; | |
1411 | char buf[SRCDEST2STR_BUFFER]; | |
1412 | rib_dest_t *dest; | |
1413 | struct zebra_vrf *zvrf = NULL; | |
1414 | struct prefix *p, *src_p; | |
1415 | srcdest_rnode_prefixes(rn, &p, &src_p); | |
1416 | vrf_id_t vrf_id = VRF_UNKNOWN; | |
1417 | ||
1418 | assert(rn); | |
1419 | ||
1420 | dest = rib_dest_from_rnode(rn); | |
1421 | if (dest) { | |
1422 | zvrf = rib_dest_vrf(dest); | |
1423 | vrf_id = zvrf_id(zvrf); | |
1424 | } | |
bab85d4f | 1425 | |
d62a17ae | 1426 | if (IS_ZEBRA_DEBUG_RIB) |
1427 | srcdest_rnode2str(rn, buf, sizeof(buf)); | |
bab85d4f | 1428 | |
d62a17ae | 1429 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) |
1430 | zlog_debug("%u:%s: Processing rn %p", vrf_id, buf, rn); | |
bab85d4f | 1431 | |
d62a17ae | 1432 | RNODE_FOREACH_RE_SAFE(rn, re, next) |
1433 | { | |
1434 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) | |
1435 | zlog_debug( | |
1436 | "%u:%s: Examine re %p (type %d) status %x flags %x " | |
1437 | "dist %d metric %d", | |
1438 | vrf_id, buf, re, re->type, re->status, | |
1439 | re->flags, re->distance, re->metric); | |
1440 | ||
1441 | UNSET_FLAG(re->status, ROUTE_ENTRY_NEXTHOPS_CHANGED); | |
1442 | ||
1443 | /* Currently selected re. */ | |
1444 | if (CHECK_FLAG(re->flags, ZEBRA_FLAG_SELECTED)) { | |
1445 | assert(old_selected == NULL); | |
1446 | old_selected = re; | |
1447 | } | |
1448 | /* Currently in fib */ | |
1449 | if (CHECK_FLAG(re->status, ROUTE_ENTRY_SELECTED_FIB)) { | |
1450 | assert(old_fib == NULL); | |
1451 | old_fib = re; | |
1452 | } | |
bab85d4f | 1453 | |
d62a17ae | 1454 | /* Skip deleted entries from selection */ |
1455 | if (CHECK_FLAG(re->status, ROUTE_ENTRY_REMOVED)) | |
1456 | continue; | |
1457 | ||
1458 | /* Skip unreachable nexthop. */ | |
1459 | /* This first call to nexthop_active_update is merely to | |
1460 | * determine if | |
1461 | * there's any change to nexthops associated with this RIB | |
1462 | * entry. Now, | |
1463 | * rib_process() can be invoked due to an external event such as | |
1464 | * link | |
1465 | * down or due to next-hop-tracking evaluation. In the latter | |
1466 | * case, | |
1467 | * a decision has already been made that the NHs have changed. | |
1468 | * So, no | |
1469 | * need to invoke a potentially expensive call again. Further, | |
1470 | * since | |
1471 | * the change might be in a recursive NH which is not caught in | |
1472 | * the nexthop_active_update() code. Thus, we might miss changes | |
1473 | * to | |
1474 | * recursive NHs. | |
1475 | */ | |
1476 | if (!CHECK_FLAG(re->status, ROUTE_ENTRY_CHANGED) | |
1477 | && !nexthop_active_update(rn, re, 0)) { | |
1478 | if (re->type == ZEBRA_ROUTE_TABLE) { | |
1479 | /* XXX: HERE BE DRAGONS!!!!! | |
1480 | * In all honesty, I have not yet figured out | |
1481 | * what this part | |
1482 | * does or why the ROUTE_ENTRY_CHANGED test | |
1483 | * above is correct | |
1484 | * or why we need to delete a route here, and | |
1485 | * also not whether | |
1486 | * this concerns both selected and fib route, or | |
1487 | * only selected | |
1488 | * or only fib */ | |
1489 | /* This entry was denied by the 'ip protocol | |
1490 | * table' route-map, we | |
1491 | * need to delete it */ | |
1492 | if (re != old_selected) { | |
1493 | if (IS_ZEBRA_DEBUG_RIB) | |
1494 | zlog_debug( | |
1495 | "%s: %s: imported via import-table but denied " | |
1496 | "by the ip protocol table route-map", | |
1497 | __func__, buf); | |
1498 | rib_unlink(rn, re); | |
1499 | } else | |
1500 | SET_FLAG(re->status, | |
1501 | ROUTE_ENTRY_REMOVED); | |
1502 | } | |
1503 | ||
1504 | continue; | |
1505 | } | |
bab85d4f | 1506 | |
d62a17ae | 1507 | /* Infinite distance. */ |
1508 | if (re->distance == DISTANCE_INFINITY) { | |
1509 | UNSET_FLAG(re->status, ROUTE_ENTRY_CHANGED); | |
1510 | continue; | |
1511 | } | |
bab85d4f | 1512 | |
d62a17ae | 1513 | if (CHECK_FLAG(re->flags, ZEBRA_FLAG_FIB_OVERRIDE)) { |
1514 | best = rib_choose_best(new_fib, re); | |
1515 | if (new_fib && best != new_fib) | |
1516 | UNSET_FLAG(new_fib->status, | |
1517 | ROUTE_ENTRY_CHANGED); | |
1518 | new_fib = best; | |
1519 | } else { | |
1520 | best = rib_choose_best(new_selected, re); | |
1521 | if (new_selected && best != new_selected) | |
1522 | UNSET_FLAG(new_selected->status, | |
1523 | ROUTE_ENTRY_CHANGED); | |
1524 | new_selected = best; | |
1525 | } | |
1526 | if (best != re) | |
1527 | UNSET_FLAG(re->status, ROUTE_ENTRY_CHANGED); | |
1528 | } /* RNODE_FOREACH_RE */ | |
1529 | ||
1530 | /* If no FIB override route, use the selected route also for FIB */ | |
1531 | if (new_fib == NULL) | |
1532 | new_fib = new_selected; | |
1533 | ||
1534 | /* After the cycle is finished, the following pointers will be set: | |
1535 | * old_selected --- RE entry currently having SELECTED | |
1536 | * new_selected --- RE entry that is newly SELECTED | |
1537 | * old_fib --- RE entry currently in kernel FIB | |
1538 | * new_fib --- RE entry that is newly to be in kernel FIB | |
1539 | * | |
1540 | * new_selected will get SELECTED flag, and is going to be redistributed | |
1541 | * the zclients. new_fib (which can be new_selected) will be installed | |
1542 | * in kernel. | |
1543 | */ | |
1544 | ||
1545 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) { | |
1546 | zlog_debug( | |
1547 | "%u:%s: After processing: old_selected %p new_selected %p old_fib %p new_fib %p", | |
1548 | vrf_id, buf, (void *)old_selected, (void *)new_selected, | |
1549 | (void *)old_fib, (void *)new_fib); | |
1550 | } | |
446bb95e | 1551 | |
d62a17ae | 1552 | /* Buffer ROUTE_ENTRY_CHANGED here, because it will get cleared if |
1553 | * fib == selected */ | |
1554 | bool selected_changed = new_selected && CHECK_FLAG(new_selected->status, | |
1555 | ROUTE_ENTRY_CHANGED); | |
1556 | ||
1557 | /* Update fib according to selection results */ | |
1558 | if (new_fib && old_fib) | |
1559 | rib_process_update_fib(zvrf, rn, old_fib, new_fib); | |
1560 | else if (new_fib) | |
1561 | rib_process_add_fib(zvrf, rn, new_fib); | |
1562 | else if (old_fib) | |
1563 | rib_process_del_fib(zvrf, rn, old_fib); | |
1564 | ||
1565 | /* Redistribute SELECTED entry */ | |
1566 | if (old_selected != new_selected || selected_changed) { | |
1567 | struct nexthop *nexthop; | |
1568 | ||
1569 | /* Check if we have a FIB route for the destination, otherwise, | |
1570 | * don't redistribute it */ | |
1571 | for (ALL_NEXTHOPS(new_fib ? new_fib->nexthop : NULL, nexthop)) { | |
1572 | if (CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_FIB)) { | |
1573 | break; | |
1574 | } | |
1575 | } | |
1576 | if (!nexthop) | |
1577 | new_selected = NULL; | |
93bdadae | 1578 | |
d62a17ae | 1579 | if (new_selected && new_selected != new_fib) { |
1580 | nexthop_active_update(rn, new_selected, 1); | |
1581 | UNSET_FLAG(new_selected->status, ROUTE_ENTRY_CHANGED); | |
1582 | } | |
41ec9222 | 1583 | |
d62a17ae | 1584 | if (old_selected) { |
1585 | if (!new_selected) | |
1586 | redistribute_delete(p, src_p, old_selected); | |
1587 | if (old_selected != new_selected) | |
1588 | UNSET_FLAG(old_selected->flags, | |
1589 | ZEBRA_FLAG_SELECTED); | |
f857321e | 1590 | } |
446bb95e | 1591 | |
d62a17ae | 1592 | if (new_selected) { |
1593 | /* Install new or replace existing redistributed entry | |
1594 | */ | |
1595 | SET_FLAG(new_selected->flags, ZEBRA_FLAG_SELECTED); | |
1596 | redistribute_update(p, src_p, new_selected, | |
1597 | old_selected); | |
1598 | } | |
1599 | } | |
3e5c6e00 | 1600 | |
d62a17ae | 1601 | /* Remove all RE entries queued for removal */ |
1602 | RNODE_FOREACH_RE_SAFE(rn, re, next) | |
1603 | { | |
1604 | if (CHECK_FLAG(re->status, ROUTE_ENTRY_REMOVED)) { | |
1605 | if (IS_ZEBRA_DEBUG_RIB) { | |
1606 | rnode_debug(rn, vrf_id, "rn %p, removing re %p", | |
1607 | (void *)rn, (void *)re); | |
1608 | } | |
1609 | rib_unlink(rn, re); | |
1610 | } | |
1611 | } | |
4d38fdb4 | 1612 | |
d62a17ae | 1613 | /* |
1614 | * Check if the dest can be deleted now. | |
1615 | */ | |
1616 | rib_gc_dest(rn); | |
e96f9203 DO |
1617 | } |
1618 | ||
5110a0c6 | 1619 | /* Take a list of route_node structs and return 1, if there was a record |
d62a17ae | 1620 | * picked from it and processed by rib_process(). Don't process more, |
5110a0c6 | 1621 | * than one RN record; operate only in the specified sub-queue. |
e96f9203 | 1622 | */ |
d62a17ae | 1623 | static unsigned int process_subq(struct list *subq, u_char qindex) |
e96f9203 | 1624 | { |
d62a17ae | 1625 | struct listnode *lnode = listhead(subq); |
1626 | struct route_node *rnode; | |
1627 | rib_dest_t *dest; | |
1628 | struct zebra_vrf *zvrf = NULL; | |
5110a0c6 | 1629 | |
d62a17ae | 1630 | if (!lnode) |
1631 | return 0; | |
5110a0c6 | 1632 | |
d62a17ae | 1633 | rnode = listgetdata(lnode); |
1634 | dest = rib_dest_from_rnode(rnode); | |
1635 | if (dest) | |
1636 | zvrf = rib_dest_vrf(dest); | |
41ec9222 | 1637 | |
d62a17ae | 1638 | rib_process(rnode); |
5110a0c6 | 1639 | |
d62a17ae | 1640 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) { |
1641 | char buf[SRCDEST2STR_BUFFER]; | |
1642 | srcdest_rnode2str(rnode, buf, sizeof(buf)); | |
1643 | zlog_debug("%u:%s: rn %p dequeued from sub-queue %u", | |
1644 | zvrf ? zvrf_id(zvrf) : 0, buf, rnode, qindex); | |
1645 | } | |
41ec9222 | 1646 | |
d62a17ae | 1647 | if (rnode->info) |
1648 | UNSET_FLAG(rib_dest_from_rnode(rnode)->flags, | |
1649 | RIB_ROUTE_QUEUED(qindex)); | |
9fd92e3c | 1650 | |
67b9467f | 1651 | #if 0 |
5110a0c6 SH |
1652 | else |
1653 | { | |
1654 | zlog_debug ("%s: called for route_node (%p, %d) with no ribs", | |
1655 | __func__, rnode, rnode->lock); | |
1656 | zlog_backtrace(LOG_DEBUG); | |
1657 | } | |
67b9467f | 1658 | #endif |
d62a17ae | 1659 | route_unlock_node(rnode); |
1660 | list_delete_node(subq, lnode); | |
1661 | return 1; | |
e96f9203 DO |
1662 | } |
1663 | ||
fb018d25 DS |
1664 | /* |
1665 | * All meta queues have been processed. Trigger next-hop evaluation. | |
1666 | */ | |
d62a17ae | 1667 | static void meta_queue_process_complete(struct work_queue *dummy) |
fb018d25 | 1668 | { |
d62a17ae | 1669 | struct vrf *vrf; |
1670 | struct zebra_vrf *zvrf; | |
9ec6b0bb | 1671 | |
d62a17ae | 1672 | /* Evaluate nexthops for those VRFs which underwent route processing. |
1673 | * This | |
1674 | * should limit the evaluation to the necessary VRFs in most common | |
1675 | * situations. | |
1676 | */ | |
1677 | RB_FOREACH(vrf, vrf_id_head, &vrfs_by_id) | |
1678 | { | |
1679 | zvrf = vrf->info; | |
1680 | if (zvrf == NULL || !(zvrf->flags & ZEBRA_VRF_RIB_SCHEDULED)) | |
1681 | continue; | |
1682 | ||
1683 | zvrf->flags &= ~ZEBRA_VRF_RIB_SCHEDULED; | |
1684 | zebra_evaluate_rnh(zvrf_id(zvrf), AF_INET, 0, RNH_NEXTHOP_TYPE, | |
1685 | NULL); | |
1686 | zebra_evaluate_rnh(zvrf_id(zvrf), AF_INET, 0, | |
1687 | RNH_IMPORT_CHECK_TYPE, NULL); | |
1688 | zebra_evaluate_rnh(zvrf_id(zvrf), AF_INET6, 0, RNH_NEXTHOP_TYPE, | |
1689 | NULL); | |
1690 | zebra_evaluate_rnh(zvrf_id(zvrf), AF_INET6, 0, | |
1691 | RNH_IMPORT_CHECK_TYPE, NULL); | |
1692 | } | |
939fba27 | 1693 | |
d62a17ae | 1694 | /* Schedule LSPs for processing, if needed. */ |
1695 | zvrf = vrf_info_lookup(VRF_DEFAULT); | |
1696 | if (mpls_should_lsps_be_processed(zvrf)) { | |
1697 | if (IS_ZEBRA_DEBUG_MPLS) | |
1698 | zlog_debug( | |
1699 | "%u: Scheduling all LSPs upon RIB completion", | |
1700 | zvrf_id(zvrf)); | |
1701 | zebra_mpls_lsp_schedule(zvrf); | |
1702 | mpls_unmark_lsps_for_processing(zvrf); | |
1703 | } | |
fb018d25 DS |
1704 | } |
1705 | ||
e96f9203 | 1706 | /* Dispatch the meta queue by picking, processing and unlocking the next RN from |
d62a17ae | 1707 | * a non-empty sub-queue with lowest priority. wq is equal to zebra->ribq and |
1708 | * data | |
e96f9203 DO |
1709 | * is pointed to the meta queue structure. |
1710 | */ | |
d62a17ae | 1711 | static wq_item_status meta_queue_process(struct work_queue *dummy, void *data) |
e96f9203 | 1712 | { |
d62a17ae | 1713 | struct meta_queue *mq = data; |
1714 | unsigned i; | |
5110a0c6 | 1715 | |
d62a17ae | 1716 | for (i = 0; i < MQ_SIZE; i++) |
1717 | if (process_subq(mq->subq[i], i)) { | |
1718 | mq->size--; | |
1719 | break; | |
1720 | } | |
1721 | return mq->size ? WQ_REQUEUE : WQ_SUCCESS; | |
e96f9203 DO |
1722 | } |
1723 | ||
9fd92e3c AS |
1724 | /* |
1725 | * Map from rib types to queue type (priority) in meta queue | |
1726 | */ | |
5110a0c6 | 1727 | static const u_char meta_queue_map[ZEBRA_ROUTE_MAX] = { |
d62a17ae | 1728 | [ZEBRA_ROUTE_SYSTEM] = 4, |
1729 | [ZEBRA_ROUTE_KERNEL] = 0, | |
1730 | [ZEBRA_ROUTE_CONNECT] = 0, | |
1731 | [ZEBRA_ROUTE_STATIC] = 1, | |
1732 | [ZEBRA_ROUTE_RIP] = 2, | |
1733 | [ZEBRA_ROUTE_RIPNG] = 2, | |
1734 | [ZEBRA_ROUTE_OSPF] = 2, | |
1735 | [ZEBRA_ROUTE_OSPF6] = 2, | |
1736 | [ZEBRA_ROUTE_ISIS] = 2, | |
1737 | [ZEBRA_ROUTE_BGP] = 3, | |
1738 | [ZEBRA_ROUTE_PIM] = 4, // Shouldn't happen but for safety | |
1739 | [ZEBRA_ROUTE_EIGRP] = 2, | |
1740 | [ZEBRA_ROUTE_NHRP] = 2, | |
1741 | [ZEBRA_ROUTE_HSLS] = 4, | |
1742 | [ZEBRA_ROUTE_OLSR] = 4, | |
1743 | [ZEBRA_ROUTE_TABLE] = 1, | |
1744 | [ZEBRA_ROUTE_LDP] = 4, | |
1745 | [ZEBRA_ROUTE_VNC] = 3, | |
1746 | [ZEBRA_ROUTE_VNC_DIRECT] = 3, | |
1747 | [ZEBRA_ROUTE_VNC_DIRECT_RH] = 3, | |
1748 | [ZEBRA_ROUTE_BGP_DIRECT] = 3, | |
1749 | [ZEBRA_ROUTE_BGP_DIRECT_EXT] = 3, | |
1750 | [ZEBRA_ROUTE_BABEL] = 2, | |
1751 | [ZEBRA_ROUTE_ALL] = 4, // Shouldn't happen but for safety | |
5110a0c6 SH |
1752 | }; |
1753 | ||
1754 | /* Look into the RN and queue it into one or more priority queues, | |
1755 | * increasing the size for each data push done. | |
e96f9203 | 1756 | */ |
d62a17ae | 1757 | static void rib_meta_queue_add(struct meta_queue *mq, struct route_node *rn) |
e96f9203 | 1758 | { |
d62a17ae | 1759 | struct route_entry *re; |
5110a0c6 | 1760 | |
d62a17ae | 1761 | RNODE_FOREACH_RE(rn, re) |
2263a412 | 1762 | { |
d62a17ae | 1763 | u_char qindex = meta_queue_map[re->type]; |
1764 | struct zebra_vrf *zvrf; | |
1765 | ||
1766 | /* Invariant: at this point we always have rn->info set. */ | |
1767 | if (CHECK_FLAG(rib_dest_from_rnode(rn)->flags, | |
1768 | RIB_ROUTE_QUEUED(qindex))) { | |
1769 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) | |
1770 | rnode_debug( | |
1771 | rn, re->vrf_id, | |
1772 | "rn %p is already queued in sub-queue %u", | |
1773 | (void *)rn, qindex); | |
1774 | continue; | |
1775 | } | |
5110a0c6 | 1776 | |
d62a17ae | 1777 | SET_FLAG(rib_dest_from_rnode(rn)->flags, |
1778 | RIB_ROUTE_QUEUED(qindex)); | |
1779 | listnode_add(mq->subq[qindex], rn); | |
1780 | route_lock_node(rn); | |
1781 | mq->size++; | |
5110a0c6 | 1782 | |
d62a17ae | 1783 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) |
1784 | rnode_debug(rn, re->vrf_id, | |
1785 | "queued rn %p into sub-queue %u", | |
1786 | (void *)rn, qindex); | |
9ec6b0bb | 1787 | |
d62a17ae | 1788 | zvrf = zebra_vrf_lookup_by_id(re->vrf_id); |
1789 | if (zvrf) | |
1790 | zvrf->flags |= ZEBRA_VRF_RIB_SCHEDULED; | |
1791 | } | |
4d38fdb4 | 1792 | } |
1793 | ||
6d691129 | 1794 | /* Add route_node to work queue and schedule processing */ |
d62a17ae | 1795 | void rib_queue_add(struct route_node *rn) |
4d38fdb4 | 1796 | { |
d62a17ae | 1797 | assert(rn); |
fc328ac9 | 1798 | |
d62a17ae | 1799 | /* Pointless to queue a route_node with no RIB entries to add or remove |
1800 | */ | |
1801 | if (!rnode_to_ribs(rn)) { | |
1802 | zlog_debug("%s: called for route_node (%p, %d) with no ribs", | |
1803 | __func__, (void *)rn, rn->lock); | |
1804 | zlog_backtrace(LOG_DEBUG); | |
1805 | return; | |
1806 | } | |
4d38fdb4 | 1807 | |
d62a17ae | 1808 | if (zebrad.ribq == NULL) { |
1809 | zlog_err("%s: work_queue does not exist!", __func__); | |
1810 | return; | |
1811 | } | |
1812 | ||
1813 | /* | |
1814 | * The RIB queue should normally be either empty or holding the only | |
1815 | * work_queue_item element. In the latter case this element would | |
1816 | * hold a pointer to the meta queue structure, which must be used to | |
1817 | * actually queue the route nodes to process. So create the MQ | |
1818 | * holder, if necessary, then push the work into it in any case. | |
1819 | * This semantics was introduced after 0.99.9 release. | |
1820 | */ | |
1821 | if (!zebrad.ribq->items->count) | |
1822 | work_queue_add(zebrad.ribq, zebrad.mq); | |
e96f9203 | 1823 | |
d62a17ae | 1824 | rib_meta_queue_add(zebrad.mq, rn); |
fc328ac9 | 1825 | |
d62a17ae | 1826 | return; |
4d38fdb4 | 1827 | } |
1828 | ||
5110a0c6 SH |
1829 | /* Create new meta queue. |
1830 | A destructor function doesn't seem to be necessary here. | |
1831 | */ | |
d62a17ae | 1832 | static struct meta_queue *meta_queue_new(void) |
e96f9203 | 1833 | { |
d62a17ae | 1834 | struct meta_queue *new; |
1835 | unsigned i; | |
5110a0c6 | 1836 | |
d62a17ae | 1837 | new = XCALLOC(MTYPE_WORK_QUEUE, sizeof(struct meta_queue)); |
1838 | assert(new); | |
e96f9203 | 1839 | |
d62a17ae | 1840 | for (i = 0; i < MQ_SIZE; i++) { |
1841 | new->subq[i] = list_new(); | |
1842 | assert(new->subq[i]); | |
1843 | } | |
5110a0c6 | 1844 | |
d62a17ae | 1845 | return new; |
e96f9203 DO |
1846 | } |
1847 | ||
d62a17ae | 1848 | void meta_queue_free(struct meta_queue *mq) |
5a8dfcd8 | 1849 | { |
d62a17ae | 1850 | unsigned i; |
5a8dfcd8 | 1851 | |
d62a17ae | 1852 | for (i = 0; i < MQ_SIZE; i++) |
1853 | list_delete(mq->subq[i]); | |
5a8dfcd8 | 1854 | |
d62a17ae | 1855 | XFREE(MTYPE_WORK_QUEUE, mq); |
5a8dfcd8 RW |
1856 | } |
1857 | ||
4d38fdb4 | 1858 | /* initialise zebra rib work queue */ |
d62a17ae | 1859 | static void rib_queue_init(struct zebra_t *zebra) |
4d38fdb4 | 1860 | { |
d62a17ae | 1861 | assert(zebra); |
1862 | ||
1863 | if (!(zebra->ribq = | |
1864 | work_queue_new(zebra->master, "route_node processing"))) { | |
1865 | zlog_err("%s: could not initialise work queue!", __func__); | |
1866 | return; | |
1867 | } | |
4d38fdb4 | 1868 | |
d62a17ae | 1869 | /* fill in the work queue spec */ |
1870 | zebra->ribq->spec.workfunc = &meta_queue_process; | |
1871 | zebra->ribq->spec.errorfunc = NULL; | |
1872 | zebra->ribq->spec.completion_func = &meta_queue_process_complete; | |
1873 | /* XXX: TODO: These should be runtime configurable via vty */ | |
1874 | zebra->ribq->spec.max_retries = 3; | |
1875 | zebra->ribq->spec.hold = rib_process_hold_time; | |
1876 | ||
1877 | if (!(zebra->mq = meta_queue_new())) { | |
1878 | zlog_err("%s: could not initialise meta queue!", __func__); | |
1879 | return; | |
1880 | } | |
1881 | return; | |
718e3744 | 1882 | } |
1883 | ||
6d691129 PJ |
1884 | /* RIB updates are processed via a queue of pointers to route_nodes. |
1885 | * | |
1886 | * The queue length is bounded by the maximal size of the routing table, | |
1887 | * as a route_node will not be requeued, if already queued. | |
1888 | * | |
f0f77c9a DS |
1889 | * REs are submitted via rib_addnode or rib_delnode which set minimal |
1890 | * state, or static_install_route (when an existing RE is updated) | |
3c0755dc | 1891 | * and then submit route_node to queue for best-path selection later. |
f0f77c9a | 1892 | * Order of add/delete state changes are preserved for any given RE. |
6d691129 | 1893 | * |
f0f77c9a | 1894 | * Deleted REs are reaped during best-path selection. |
6d691129 PJ |
1895 | * |
1896 | * rib_addnode | |
f0f77c9a DS |
1897 | * |-> rib_link or unset ROUTE_ENTRY_REMOVE |->Update kernel with |
1898 | * |-------->| | best RE, if required | |
3c0755dc PJ |
1899 | * | | |
1900 | * static_install->|->rib_addqueue...... -> rib_process | |
1901 | * | | | |
1902 | * |-------->| |-> rib_unlink | |
f0f77c9a DS |
1903 | * |-> set ROUTE_ENTRY_REMOVE | |
1904 | * rib_delnode (RE freed) | |
6d691129 | 1905 | * |
9fd92e3c AS |
1906 | * The 'info' pointer of a route_node points to a rib_dest_t |
1907 | * ('dest'). Queueing state for a route_node is kept on the dest. The | |
1908 | * dest is created on-demand by rib_link() and is kept around at least | |
1909 | * as long as there are ribs hanging off it (@see rib_gc_dest()). | |
d62a17ae | 1910 | * |
6d691129 PJ |
1911 | * Refcounting (aka "locking" throughout the GNU Zebra and Quagga code): |
1912 | * | |
1913 | * - route_nodes: refcounted by: | |
9fd92e3c AS |
1914 | * - dest attached to route_node: |
1915 | * - managed by: rib_link/rib_gc_dest | |
6d691129 PJ |
1916 | * - route_node processing queue |
1917 | * - managed by: rib_addqueue, rib_process. | |
1918 | * | |
1919 | */ | |
d62a17ae | 1920 | |
f0f77c9a | 1921 | /* Add RE to head of the route node. */ |
d62a17ae | 1922 | static void rib_link(struct route_node *rn, struct route_entry *re, int process) |
1923 | { | |
1924 | struct route_entry *head; | |
1925 | rib_dest_t *dest; | |
1926 | afi_t afi; | |
1927 | const char *rmap_name; | |
9fd92e3c | 1928 | |
d62a17ae | 1929 | assert(re && rn); |
9fd92e3c | 1930 | |
d62a17ae | 1931 | dest = rib_dest_from_rnode(rn); |
1932 | if (!dest) { | |
1933 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) | |
1934 | rnode_debug(rn, re->vrf_id, "rn %p adding dest", rn); | |
7a4bb9c5 | 1935 | |
d62a17ae | 1936 | dest = XCALLOC(MTYPE_RIB_DEST, sizeof(rib_dest_t)); |
1937 | route_lock_node(rn); /* rn route table reference */ | |
1938 | rn->info = dest; | |
1939 | dest->rnode = rn; | |
1940 | } | |
2263a412 | 1941 | |
d62a17ae | 1942 | head = dest->routes; |
1943 | if (head) { | |
1944 | head->prev = re; | |
1945 | } | |
1946 | re->next = head; | |
1947 | dest->routes = re; | |
1948 | ||
1949 | afi = (rn->p.family == AF_INET) | |
1950 | ? AFI_IP | |
1951 | : (rn->p.family == AF_INET6) ? AFI_IP6 : AFI_MAX; | |
1952 | if (is_zebra_import_table_enabled(afi, re->table)) { | |
1953 | rmap_name = zebra_get_import_table_route_map(afi, re->table); | |
1954 | zebra_add_import_table_entry(rn, re, rmap_name); | |
1955 | } else if (process) | |
1956 | rib_queue_add(rn); | |
1957 | } | |
1958 | ||
1959 | void rib_addnode(struct route_node *rn, struct route_entry *re, int process) | |
1960 | { | |
1961 | /* RE node has been un-removed before route-node is processed. | |
1962 | * route_node must hence already be on the queue for processing.. | |
1963 | */ | |
1964 | if (CHECK_FLAG(re->status, ROUTE_ENTRY_REMOVED)) { | |
1965 | if (IS_ZEBRA_DEBUG_RIB) | |
1966 | rnode_debug(rn, re->vrf_id, "rn %p, un-removed re %p", | |
1967 | (void *)rn, (void *)re); | |
1968 | ||
1969 | UNSET_FLAG(re->status, ROUTE_ENTRY_REMOVED); | |
1970 | return; | |
1971 | } | |
1972 | rib_link(rn, re, process); | |
6d691129 PJ |
1973 | } |
1974 | ||
9fd92e3c AS |
1975 | /* |
1976 | * rib_unlink | |
1977 | * | |
1978 | * Detach a rib structure from a route_node. | |
1979 | * | |
1980 | * Note that a call to rib_unlink() should be followed by a call to | |
1981 | * rib_gc_dest() at some point. This allows a rib_dest_t that is no | |
1982 | * longer required to be deleted. | |
1983 | */ | |
d62a17ae | 1984 | void rib_unlink(struct route_node *rn, struct route_entry *re) |
6d691129 | 1985 | { |
d62a17ae | 1986 | rib_dest_t *dest; |
9fd92e3c | 1987 | |
d62a17ae | 1988 | assert(rn && re); |
6d691129 | 1989 | |
d62a17ae | 1990 | if (IS_ZEBRA_DEBUG_RIB) |
1991 | rnode_debug(rn, re->vrf_id, "rn %p, re %p", (void *)rn, | |
1992 | (void *)re); | |
6d691129 | 1993 | |
d62a17ae | 1994 | dest = rib_dest_from_rnode(rn); |
6d691129 | 1995 | |
d62a17ae | 1996 | if (re->next) |
1997 | re->next->prev = re->prev; | |
6d691129 | 1998 | |
d62a17ae | 1999 | if (re->prev) |
2000 | re->prev->next = re->next; | |
2001 | else { | |
2002 | dest->routes = re->next; | |
2003 | } | |
7a4bb9c5 | 2004 | |
d62a17ae | 2005 | /* free RE and nexthops */ |
2006 | zebra_deregister_rnh_static_nexthops(re->vrf_id, re->nexthop, rn); | |
2007 | nexthops_free(re->nexthop); | |
2008 | XFREE(MTYPE_RE, re); | |
2009 | } | |
2010 | ||
2011 | void rib_delnode(struct route_node *rn, struct route_entry *re) | |
2012 | { | |
2013 | afi_t afi; | |
2014 | ||
2015 | if (IS_ZEBRA_DEBUG_RIB) | |
2016 | rnode_debug(rn, re->vrf_id, "rn %p, re %p, removing", | |
2017 | (void *)rn, (void *)re); | |
2018 | SET_FLAG(re->status, ROUTE_ENTRY_REMOVED); | |
2019 | ||
2020 | afi = (rn->p.family == AF_INET) | |
2021 | ? AFI_IP | |
2022 | : (rn->p.family == AF_INET6) ? AFI_IP6 : AFI_MAX; | |
2023 | if (is_zebra_import_table_enabled(afi, re->table)) { | |
2024 | zebra_del_import_table_entry(rn, re); | |
2025 | /* Just clean up if non main table */ | |
2026 | if (IS_ZEBRA_DEBUG_RIB) { | |
2027 | char buf[SRCDEST2STR_BUFFER]; | |
2028 | srcdest_rnode2str(rn, buf, sizeof(buf)); | |
2029 | zlog_debug( | |
2030 | "%u:%s: Freeing route rn %p, re %p (type %d)", | |
2031 | re->vrf_id, buf, rn, re, re->type); | |
2032 | } | |
7a4bb9c5 | 2033 | |
d62a17ae | 2034 | rib_unlink(rn, re); |
2035 | } else { | |
2036 | rib_queue_add(rn); | |
2037 | } | |
718e3744 | 2038 | } |
2039 | ||
f0f77c9a | 2040 | /* This function dumps the contents of a given RE entry into |
dc95824a DO |
2041 | * standard debug log. Calling function name and IP prefix in |
2042 | * question are passed as 1st and 2nd arguments. | |
2043 | */ | |
2044 | ||
d62a17ae | 2045 | void _route_entry_dump(const char *func, union prefixconstptr pp, |
2046 | union prefixconstptr src_pp, | |
2047 | const struct route_entry *re) | |
2048 | { | |
2049 | const struct prefix *p = pp.p; | |
2050 | const struct prefix *src_p = src_pp.p; | |
2051 | bool is_srcdst = src_p && src_p->prefixlen; | |
2052 | char straddr[PREFIX_STRLEN]; | |
2053 | char srcaddr[PREFIX_STRLEN]; | |
2054 | struct nexthop *nexthop; | |
2055 | ||
2056 | zlog_debug("%s: dumping RE entry %p for %s%s%s vrf %u", func, | |
2057 | (const void *)re, prefix2str(pp, straddr, sizeof(straddr)), | |
2058 | is_srcdst ? " from " : "", | |
2059 | is_srcdst ? prefix2str(src_pp, srcaddr, sizeof(srcaddr)) | |
2060 | : "", | |
2061 | re->vrf_id); | |
2062 | zlog_debug( | |
2063 | "%s: refcnt == %lu, uptime == %lu, type == %u, instance == %d, table == %d", | |
2064 | func, re->refcnt, (unsigned long)re->uptime, re->type, | |
2065 | re->instance, re->table); | |
2066 | zlog_debug( | |
2067 | "%s: metric == %u, mtu == %u, distance == %u, flags == %u, status == %u", | |
2068 | func, re->metric, re->mtu, re->distance, re->flags, re->status); | |
2069 | zlog_debug("%s: nexthop_num == %u, nexthop_active_num == %u", func, | |
2070 | re->nexthop_num, re->nexthop_active_num); | |
2071 | ||
2072 | for (ALL_NEXTHOPS(re->nexthop, nexthop)) { | |
2073 | inet_ntop(p->family, &nexthop->gate, straddr, INET6_ADDRSTRLEN); | |
2074 | zlog_debug("%s: %s %s with flags %s%s%s", func, | |
2075 | (nexthop->rparent ? " NH" : "NH"), straddr, | |
2076 | (CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE) | |
2077 | ? "ACTIVE " | |
2078 | : ""), | |
2079 | (CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_FIB) | |
2080 | ? "FIB " | |
2081 | : ""), | |
2082 | (CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_RECURSIVE) | |
2083 | ? "RECURSIVE" | |
2084 | : "")); | |
2085 | } | |
2086 | zlog_debug("%s: dump complete", func); | |
dc95824a DO |
2087 | } |
2088 | ||
2089 | /* This is an exported helper to rtm_read() to dump the strange | |
f0f77c9a | 2090 | * RE entry found by rib_lookup_ipv4_route() |
dc95824a DO |
2091 | */ |
2092 | ||
d62a17ae | 2093 | void rib_lookup_and_dump(struct prefix_ipv4 *p, vrf_id_t vrf_id) |
2094 | { | |
2095 | struct route_table *table; | |
2096 | struct route_node *rn; | |
2097 | struct route_entry *re; | |
2098 | char prefix_buf[INET_ADDRSTRLEN]; | |
2099 | ||
2100 | /* Lookup table. */ | |
2101 | table = zebra_vrf_table(AFI_IP, SAFI_UNICAST, vrf_id); | |
2102 | if (!table) { | |
2103 | zlog_err("%s: zebra_vrf_table() returned NULL", __func__); | |
2104 | return; | |
2105 | } | |
2106 | ||
2107 | /* Scan the RIB table for exactly matching RE entry. */ | |
2108 | rn = route_node_lookup(table, (struct prefix *)p); | |
2109 | ||
2110 | /* No route for this prefix. */ | |
2111 | if (!rn) { | |
2112 | zlog_debug("%s: lookup failed for %s", __func__, | |
2113 | prefix2str((struct prefix *)p, prefix_buf, | |
2114 | sizeof(prefix_buf))); | |
2115 | return; | |
2116 | } | |
2117 | ||
2118 | /* Unlock node. */ | |
2119 | route_unlock_node(rn); | |
2120 | ||
2121 | /* let's go */ | |
2122 | RNODE_FOREACH_RE(rn, re) | |
2123 | { | |
2124 | zlog_debug("%s: rn %p, re %p: %s, %s", __func__, (void *)rn, | |
2125 | (void *)re, | |
2126 | (CHECK_FLAG(re->status, ROUTE_ENTRY_REMOVED) | |
2127 | ? "removed" | |
2128 | : "NOT removed"), | |
2129 | (CHECK_FLAG(re->flags, ZEBRA_FLAG_SELECTED) | |
2130 | ? "selected" | |
2131 | : "NOT selected")); | |
2132 | route_entry_dump(p, NULL, re); | |
2133 | } | |
dc95824a DO |
2134 | } |
2135 | ||
20e5ff0a DO |
2136 | /* Check if requested address assignment will fail due to another |
2137 | * route being installed by zebra in FIB already. Take necessary | |
2138 | * actions, if needed: remove such a route from FIB and deSELECT | |
f0f77c9a | 2139 | * corresponding RE entry. Then put affected RN into RIBQ head. |
20e5ff0a | 2140 | */ |
d62a17ae | 2141 | void rib_lookup_and_pushup(struct prefix_ipv4 *p, vrf_id_t vrf_id) |
2142 | { | |
2143 | struct route_table *table; | |
2144 | struct route_node *rn; | |
2145 | struct route_entry *re; | |
2146 | unsigned changed = 0; | |
2147 | ||
2148 | if (NULL == (table = zebra_vrf_table(AFI_IP, SAFI_UNICAST, vrf_id))) { | |
2149 | zlog_err("%s: zebra_vrf_table() returned NULL", __func__); | |
2150 | return; | |
2151 | } | |
2152 | ||
2153 | /* No matches would be the simplest case. */ | |
2154 | if (NULL == (rn = route_node_lookup(table, (struct prefix *)p))) | |
2155 | return; | |
2156 | ||
2157 | /* Unlock node. */ | |
2158 | route_unlock_node(rn); | |
2159 | ||
2160 | /* Check all RE entries. In case any changes have to be done, requeue | |
2161 | * the RN into RIBQ head. If the routing message about the new connected | |
2162 | * route (generated by the IP address we are going to assign very soon) | |
2163 | * comes before the RIBQ is processed, the new RE entry will join | |
2164 | * RIBQ record already on head. This is necessary for proper | |
2165 | * revalidation | |
2166 | * of the rest of the RE. | |
2167 | */ | |
2168 | RNODE_FOREACH_RE(rn, re) | |
2169 | { | |
2170 | if (CHECK_FLAG(re->status, ROUTE_ENTRY_SELECTED_FIB) | |
2171 | && !RIB_SYSTEM_ROUTE(re)) { | |
2172 | changed = 1; | |
2173 | if (IS_ZEBRA_DEBUG_RIB) { | |
2174 | char buf[PREFIX_STRLEN]; | |
2175 | zlog_debug( | |
2176 | "%u:%s: freeing way for connected prefix", | |
2177 | re->vrf_id, | |
2178 | prefix2str(&rn->p, buf, sizeof(buf))); | |
2179 | route_entry_dump(&rn->p, NULL, re); | |
2180 | } | |
2181 | rib_uninstall(rn, re); | |
2182 | } | |
2183 | } | |
2184 | if (changed) | |
2185 | rib_queue_add(rn); | |
20e5ff0a DO |
2186 | } |
2187 | ||
d62a17ae | 2188 | int rib_add_multipath(afi_t afi, safi_t safi, struct prefix *p, |
2189 | struct prefix_ipv6 *src_p, struct route_entry *re) | |
718e3744 | 2190 | { |
d62a17ae | 2191 | struct route_table *table; |
2192 | struct route_node *rn; | |
2193 | struct route_entry *same; | |
2194 | struct nexthop *nexthop; | |
2195 | int ret = 0; | |
2196 | int family; | |
b4c034b0 | 2197 | |
d62a17ae | 2198 | if (!re) |
2199 | return 0; | |
b4c034b0 | 2200 | |
d62a17ae | 2201 | if (p->family == AF_INET) |
2202 | family = AFI_IP; | |
2203 | else | |
2204 | family = AFI_IP6; | |
b4c034b0 | 2205 | |
d62a17ae | 2206 | assert(!src_p || family == AFI_IP6); |
05737783 | 2207 | |
d62a17ae | 2208 | /* Lookup table. */ |
2209 | table = zebra_vrf_table_with_table_id(family, safi, re->vrf_id, | |
2210 | re->table); | |
2211 | if (!table) | |
2212 | return 0; | |
cddf391b | 2213 | |
d62a17ae | 2214 | /* Make it sure prefixlen is applied to the prefix. */ |
2215 | apply_mask(p); | |
2216 | if (src_p) | |
2217 | apply_mask_ipv6(src_p); | |
718e3744 | 2218 | |
d62a17ae | 2219 | /* Set default distance by route type. */ |
2220 | if (re->distance == 0) { | |
2221 | re->distance = route_info[re->type].distance; | |
718e3744 | 2222 | |
d62a17ae | 2223 | /* iBGP distance is 200. */ |
2224 | if (re->type == ZEBRA_ROUTE_BGP | |
2225 | && CHECK_FLAG(re->flags, ZEBRA_FLAG_IBGP)) | |
2226 | re->distance = 200; | |
2227 | } | |
718e3744 | 2228 | |
d62a17ae | 2229 | /* Lookup route node.*/ |
2230 | rn = srcdest_rnode_get(table, p, src_p); | |
718e3744 | 2231 | |
d62a17ae | 2232 | /* If same type of route are installed, treat it as a implicit |
2233 | withdraw. */ | |
2234 | RNODE_FOREACH_RE(rn, same) | |
2235 | { | |
2236 | if (CHECK_FLAG(same->status, ROUTE_ENTRY_REMOVED)) | |
2237 | continue; | |
41ec9222 | 2238 | |
d62a17ae | 2239 | if (same->type == re->type && same->instance == re->instance |
2240 | && same->table == re->table | |
2241 | && same->type != ZEBRA_ROUTE_CONNECT) | |
2242 | break; | |
2243 | } | |
718e3744 | 2244 | |
d62a17ae | 2245 | /* If this route is kernel route, set FIB flag to the route. */ |
2246 | if (re->type == ZEBRA_ROUTE_KERNEL || re->type == ZEBRA_ROUTE_CONNECT) | |
2247 | for (nexthop = re->nexthop; nexthop; nexthop = nexthop->next) | |
2248 | SET_FLAG(nexthop->flags, NEXTHOP_FLAG_FIB); | |
718e3744 | 2249 | |
d62a17ae | 2250 | /* Link new re to node.*/ |
2251 | if (IS_ZEBRA_DEBUG_RIB) { | |
2252 | rnode_debug( | |
2253 | rn, re->vrf_id, | |
2254 | "Inserting route rn %p, re %p (type %d) existing %p", | |
2255 | (void *)rn, (void *)re, re->type, (void *)same); | |
718e3744 | 2256 | |
d62a17ae | 2257 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) |
2258 | route_entry_dump(p, src_p, re); | |
718e3744 | 2259 | } |
d62a17ae | 2260 | rib_addnode(rn, re, 1); |
2261 | ret = 1; | |
6b0655a2 | 2262 | |
d62a17ae | 2263 | /* Free implicit route.*/ |
2264 | if (same) { | |
2265 | rib_delnode(rn, same); | |
2266 | ret = -1; | |
2267 | } | |
718e3744 | 2268 | |
d62a17ae | 2269 | route_unlock_node(rn); |
2270 | return ret; | |
2271 | } | |
2272 | ||
2273 | void rib_delete(afi_t afi, safi_t safi, vrf_id_t vrf_id, int type, | |
2274 | u_short instance, int flags, struct prefix *p, | |
2275 | struct prefix_ipv6 *src_p, union g_addr *gate, | |
2276 | ifindex_t ifindex, u_int32_t table_id) | |
2277 | { | |
2278 | struct route_table *table; | |
2279 | struct route_node *rn; | |
2280 | struct route_entry *re; | |
2281 | struct route_entry *fib = NULL; | |
2282 | struct route_entry *same = NULL; | |
2283 | struct nexthop *nexthop; | |
2284 | char buf2[INET6_ADDRSTRLEN]; | |
2285 | ||
2286 | assert(!src_p || afi == AFI_IP6); | |
2287 | ||
2288 | /* Lookup table. */ | |
2289 | table = zebra_vrf_table_with_table_id(afi, safi, vrf_id, table_id); | |
2290 | if (!table) | |
2291 | return; | |
2292 | ||
2293 | /* Apply mask. */ | |
2294 | apply_mask(p); | |
2295 | if (src_p) | |
2296 | apply_mask_ipv6(src_p); | |
2297 | ||
2298 | /* Lookup route node. */ | |
2299 | rn = srcdest_rnode_lookup(table, p, src_p); | |
2300 | if (!rn) { | |
2301 | char dst_buf[PREFIX_STRLEN], src_buf[PREFIX_STRLEN]; | |
2302 | ||
2303 | prefix2str(p, dst_buf, sizeof(dst_buf)); | |
2304 | if (src_p && src_p->prefixlen) | |
2305 | prefix2str(src_p, src_buf, sizeof(src_buf)); | |
2306 | else | |
2307 | src_buf[0] = '\0'; | |
2308 | ||
2309 | if (IS_ZEBRA_DEBUG_RIB) | |
2310 | zlog_debug("%u:%s%s%s doesn't exist in rib", vrf_id, | |
2311 | dst_buf, | |
2312 | (src_buf[0] != '\0') ? " from " : "", | |
2313 | src_buf); | |
2314 | return; | |
2315 | } | |
718e3744 | 2316 | |
d62a17ae | 2317 | /* Lookup same type route. */ |
2318 | RNODE_FOREACH_RE(rn, re) | |
2319 | { | |
2320 | if (CHECK_FLAG(re->status, ROUTE_ENTRY_REMOVED)) | |
2321 | continue; | |
2322 | ||
2323 | if (CHECK_FLAG(re->status, ROUTE_ENTRY_SELECTED_FIB)) | |
2324 | fib = re; | |
2325 | ||
2326 | if (re->type != type) | |
2327 | continue; | |
2328 | if (re->instance != instance) | |
2329 | continue; | |
2330 | if (re->type == ZEBRA_ROUTE_CONNECT && (nexthop = re->nexthop) | |
2331 | && nexthop->type == NEXTHOP_TYPE_IFINDEX) { | |
2332 | if (nexthop->ifindex != ifindex) | |
2333 | continue; | |
2334 | if (re->refcnt) { | |
2335 | re->refcnt--; | |
2336 | route_unlock_node(rn); | |
2337 | route_unlock_node(rn); | |
2338 | return; | |
2339 | } | |
2340 | same = re; | |
2341 | break; | |
2342 | } | |
2343 | /* Make sure that the route found has the same gateway. */ | |
2344 | else { | |
2345 | if (gate == NULL) { | |
2346 | same = re; | |
2347 | break; | |
2348 | } | |
2349 | for (ALL_NEXTHOPS(re->nexthop, nexthop)) | |
2350 | if (IPV4_ADDR_SAME(&nexthop->gate.ipv4, gate) | |
2351 | || IPV6_ADDR_SAME(&nexthop->gate.ipv6, | |
2352 | gate)) { | |
2353 | same = re; | |
2354 | break; | |
2355 | } | |
2356 | if (same) | |
2357 | break; | |
2358 | } | |
2359 | } | |
2360 | /* If same type of route can't be found and this message is from | |
2361 | kernel. */ | |
2362 | if (!same) { | |
2363 | if (fib && type == ZEBRA_ROUTE_KERNEL | |
2364 | && CHECK_FLAG(flags, ZEBRA_FLAG_SELFROUTE)) { | |
2365 | if (IS_ZEBRA_DEBUG_RIB) { | |
2366 | rnode_debug( | |
2367 | rn, vrf_id, | |
2368 | "rn %p, re %p (type %d) was deleted from kernel, adding", | |
2369 | rn, fib, fib->type); | |
2370 | } | |
2371 | if (allow_delete) { | |
2372 | /* Unset flags. */ | |
2373 | for (nexthop = fib->nexthop; nexthop; | |
2374 | nexthop = nexthop->next) | |
2375 | UNSET_FLAG(nexthop->flags, | |
2376 | NEXTHOP_FLAG_FIB); | |
2377 | ||
2378 | UNSET_FLAG(fib->status, | |
2379 | ROUTE_ENTRY_SELECTED_FIB); | |
2380 | } else { | |
2381 | /* This means someone else, other than Zebra, | |
2382 | * has deleted | |
2383 | * a Zebra router from the kernel. We will add | |
2384 | * it back */ | |
2385 | rib_install_kernel(rn, fib, NULL); | |
2386 | } | |
2387 | } else { | |
2388 | if (IS_ZEBRA_DEBUG_RIB) { | |
2389 | if (gate) | |
2390 | rnode_debug( | |
2391 | rn, vrf_id, | |
2392 | "via %s ifindex %d type %d " | |
2393 | "doesn't exist in rib", | |
2394 | inet_ntop( | |
2395 | family2afi(afi), gate, | |
2396 | buf2, | |
2397 | INET_ADDRSTRLEN), /* FIXME | |
2398 | */ | |
2399 | ifindex, type); | |
2400 | else | |
2401 | rnode_debug( | |
2402 | rn, vrf_id, | |
2403 | "ifindex %d type %d doesn't exist in rib", | |
2404 | ifindex, type); | |
2405 | } | |
2406 | route_unlock_node(rn); | |
2407 | return; | |
2408 | } | |
2409 | } | |
718e3744 | 2410 | |
d62a17ae | 2411 | if (same) |
2412 | rib_delnode(rn, same); | |
05737783 | 2413 | |
d62a17ae | 2414 | route_unlock_node(rn); |
2415 | return; | |
2416 | } | |
718e3744 | 2417 | |
718e3744 | 2418 | |
d62a17ae | 2419 | int rib_add(afi_t afi, safi_t safi, vrf_id_t vrf_id, int type, u_short instance, |
2420 | int flags, struct prefix *p, struct prefix_ipv6 *src_p, | |
2421 | union g_addr *gate, union g_addr *src, ifindex_t ifindex, | |
2422 | u_int32_t table_id, u_int32_t metric, u_int32_t mtu, | |
2423 | u_char distance) | |
2424 | { | |
2425 | struct route_entry *re; | |
2426 | struct route_entry *same = NULL; | |
2427 | struct route_table *table; | |
2428 | struct route_node *rn; | |
2429 | struct nexthop *nexthop; | |
2430 | ||
2431 | assert(!src_p || afi == AFI_IP6); | |
2432 | ||
2433 | /* Lookup table. */ | |
2434 | table = zebra_vrf_table_with_table_id(afi, safi, vrf_id, table_id); | |
2435 | if (!table) | |
2436 | return 0; | |
2437 | ||
2438 | /* Make sure mask is applied. */ | |
2439 | apply_mask(p); | |
2440 | if (src_p) | |
2441 | apply_mask_ipv6(src_p); | |
2442 | ||
2443 | /* Set default distance by route type. */ | |
2444 | if (distance == 0) { | |
2445 | if ((unsigned)type >= array_size(route_info)) | |
2446 | distance = 150; | |
2447 | else | |
2448 | distance = route_info[type].distance; | |
2449 | ||
2450 | /* iBGP distance is 200. */ | |
2451 | if (type == ZEBRA_ROUTE_BGP | |
2452 | && CHECK_FLAG(flags, ZEBRA_FLAG_IBGP)) | |
2453 | distance = 200; | |
2454 | } | |
718e3744 | 2455 | |
d62a17ae | 2456 | /* Lookup route node.*/ |
2457 | rn = srcdest_rnode_get(table, p, src_p); | |
718e3744 | 2458 | |
d62a17ae | 2459 | /* If same type of route are installed, treat it as a implicit |
2460 | withdraw. */ | |
2461 | RNODE_FOREACH_RE(rn, re) | |
ebf1ead0 | 2462 | { |
d62a17ae | 2463 | if (CHECK_FLAG(re->status, ROUTE_ENTRY_REMOVED)) |
2464 | continue; | |
2465 | ||
2466 | if (re->type != type) | |
2467 | continue; | |
2468 | if (re->instance != instance) | |
2469 | continue; | |
2470 | if (re->type != ZEBRA_ROUTE_CONNECT) { | |
2471 | same = re; | |
2472 | break; | |
2473 | } | |
2474 | /* Duplicate connected route comes in. */ | |
2475 | else if ((nexthop = re->nexthop) | |
2476 | && nexthop->type == NEXTHOP_TYPE_IFINDEX | |
2477 | && nexthop->ifindex == ifindex | |
2478 | && !CHECK_FLAG(re->status, ROUTE_ENTRY_REMOVED)) { | |
2479 | re->refcnt++; | |
2480 | return 0; | |
2481 | } | |
ebf1ead0 | 2482 | } |
718e3744 | 2483 | |
d62a17ae | 2484 | /* Allocate new re structure. */ |
2485 | re = XCALLOC(MTYPE_RE, sizeof(struct route_entry)); | |
2486 | ||
2487 | re->type = type; | |
2488 | re->instance = instance; | |
2489 | re->distance = distance; | |
2490 | re->flags = flags; | |
2491 | re->metric = metric; | |
2492 | re->mtu = mtu; | |
2493 | re->table = table_id; | |
2494 | re->vrf_id = vrf_id; | |
2495 | re->nexthop_num = 0; | |
2496 | re->uptime = time(NULL); | |
2497 | ||
2498 | /* Nexthop settings. */ | |
2499 | if (gate) { | |
2500 | if (afi == AFI_IP6) { | |
2501 | if (ifindex) | |
2502 | route_entry_nexthop_ipv6_ifindex_add( | |
2503 | re, &gate->ipv6, ifindex); | |
2504 | else | |
2505 | route_entry_nexthop_ipv6_add(re, &gate->ipv6); | |
2506 | } else { | |
2507 | if (ifindex) | |
2508 | route_entry_nexthop_ipv4_ifindex_add( | |
2509 | re, &gate->ipv4, &src->ipv4, ifindex); | |
2510 | else | |
2511 | route_entry_nexthop_ipv4_add(re, &gate->ipv4, | |
2512 | &src->ipv4); | |
2513 | } | |
2514 | } else | |
2515 | route_entry_nexthop_ifindex_add(re, ifindex); | |
2516 | ||
2517 | /* If this route is kernel route, set FIB flag to the route. */ | |
2518 | if (type == ZEBRA_ROUTE_KERNEL || type == ZEBRA_ROUTE_CONNECT) | |
2519 | for (nexthop = re->nexthop; nexthop; nexthop = nexthop->next) | |
2520 | SET_FLAG(nexthop->flags, NEXTHOP_FLAG_FIB); | |
2521 | ||
2522 | /* Link new rib to node.*/ | |
2523 | if (IS_ZEBRA_DEBUG_RIB) { | |
2524 | rnode_debug( | |
2525 | rn, vrf_id, | |
2526 | "Inserting route rn %p, re %p (type %d) existing %p", | |
2527 | (void *)rn, (void *)re, re->type, (void *)same); | |
2528 | ||
2529 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) | |
2530 | route_entry_dump(p, src_p, re); | |
3b1098be | 2531 | } |
d62a17ae | 2532 | rib_addnode(rn, re, 1); |
718e3744 | 2533 | |
d62a17ae | 2534 | /* Free implicit route.*/ |
2535 | if (same) | |
2536 | rib_delnode(rn, same); | |
718e3744 | 2537 | |
d62a17ae | 2538 | route_unlock_node(rn); |
2539 | return 0; | |
718e3744 | 2540 | } |
2541 | ||
1c848137 | 2542 | /* Schedule routes of a particular table (address-family) based on event. */ |
d62a17ae | 2543 | static void rib_update_table(struct route_table *table, |
2544 | rib_update_event_t event) | |
2545 | { | |
2546 | struct route_node *rn; | |
2547 | struct route_entry *re, *next; | |
2548 | ||
2549 | /* Walk all routes and queue for processing, if appropriate for | |
2550 | * the trigger event. | |
2551 | */ | |
2552 | for (rn = route_top(table); rn; rn = srcdest_route_next(rn)) { | |
2553 | switch (event) { | |
2554 | case RIB_UPDATE_IF_CHANGE: | |
2555 | /* Examine all routes that won't get processed by the | |
2556 | * protocol or | |
2557 | * triggered by nexthop evaluation (NHT). This would be | |
2558 | * system, | |
2559 | * kernel and certain static routes. Note that NHT will | |
2560 | * get | |
2561 | * triggered upon an interface event as connected routes | |
2562 | * always | |
2563 | * get queued for processing. | |
2564 | */ | |
2565 | RNODE_FOREACH_RE_SAFE(rn, re, next) | |
2566 | { | |
2567 | if (re->type == ZEBRA_ROUTE_OSPF | |
2568 | || re->type == ZEBRA_ROUTE_OSPF6 | |
2569 | || re->type == ZEBRA_ROUTE_BGP) | |
2570 | continue; /* protocol will handle. */ | |
2571 | else if (re->type == ZEBRA_ROUTE_STATIC) { | |
2572 | struct nexthop *nh; | |
2573 | for (nh = re->nexthop; nh; | |
2574 | nh = nh->next) | |
2575 | if (!(nh->type | |
2576 | == NEXTHOP_TYPE_IPV4 | |
2577 | || nh->type | |
2578 | == NEXTHOP_TYPE_IPV6)) | |
2579 | break; | |
2580 | ||
2581 | /* If we only have nexthops to a | |
2582 | * gateway, NHT will | |
2583 | * take care. | |
2584 | */ | |
2585 | if (nh) | |
2586 | rib_queue_add(rn); | |
2587 | } else | |
2588 | rib_queue_add(rn); | |
2589 | } | |
2590 | break; | |
2591 | ||
2592 | case RIB_UPDATE_RMAP_CHANGE: | |
2593 | case RIB_UPDATE_OTHER: | |
2594 | /* Right now, examine all routes. Can restrict to a | |
2595 | * protocol in | |
2596 | * some cases (TODO). | |
2597 | */ | |
2598 | if (rnode_to_ribs(rn)) | |
2599 | rib_queue_add(rn); | |
2600 | break; | |
2601 | ||
2602 | default: | |
2603 | break; | |
2604 | } | |
2605 | } | |
b84c7253 | 2606 | } |
2607 | ||
718e3744 | 2608 | /* RIB update function. */ |
d62a17ae | 2609 | void rib_update(vrf_id_t vrf_id, rib_update_event_t event) |
718e3744 | 2610 | { |
d62a17ae | 2611 | struct route_table *table; |
1c848137 | 2612 | |
d62a17ae | 2613 | /* Process routes of interested address-families. */ |
2614 | table = zebra_vrf_table(AFI_IP, SAFI_UNICAST, vrf_id); | |
2615 | if (table) | |
2616 | rib_update_table(table, event); | |
718e3744 | 2617 | |
d62a17ae | 2618 | table = zebra_vrf_table(AFI_IP6, SAFI_UNICAST, vrf_id); |
2619 | if (table) | |
2620 | rib_update_table(table, event); | |
718e3744 | 2621 | } |
2622 | ||
718e3744 | 2623 | /* Remove all routes which comes from non main table. */ |
d62a17ae | 2624 | static void rib_weed_table(struct route_table *table) |
718e3744 | 2625 | { |
d62a17ae | 2626 | struct route_node *rn; |
2627 | struct route_entry *re; | |
2628 | struct route_entry *next; | |
718e3744 | 2629 | |
d62a17ae | 2630 | if (table) |
2631 | for (rn = route_top(table); rn; rn = srcdest_route_next(rn)) | |
2632 | RNODE_FOREACH_RE_SAFE(rn, re, next) | |
2633 | { | |
2634 | if (CHECK_FLAG(re->status, ROUTE_ENTRY_REMOVED)) | |
2635 | continue; | |
6d691129 | 2636 | |
d62a17ae | 2637 | if (re->table != zebrad.rtm_table_default |
2638 | && re->table != RT_TABLE_MAIN) | |
2639 | rib_delnode(rn, re); | |
2640 | } | |
718e3744 | 2641 | } |
2642 | ||
2643 | /* Delete all routes from non main table. */ | |
d62a17ae | 2644 | void rib_weed_tables(void) |
718e3744 | 2645 | { |
d62a17ae | 2646 | struct vrf *vrf; |
2647 | struct zebra_vrf *zvrf; | |
78104b9b | 2648 | |
d62a17ae | 2649 | RB_FOREACH(vrf, vrf_id_head, &vrfs_by_id) |
2650 | if ((zvrf = vrf->info) != NULL) { | |
2651 | rib_weed_table(zvrf->table[AFI_IP][SAFI_UNICAST]); | |
2652 | rib_weed_table(zvrf->table[AFI_IP6][SAFI_UNICAST]); | |
2653 | } | |
718e3744 | 2654 | } |
6b0655a2 | 2655 | |
718e3744 | 2656 | /* Delete self installed routes after zebra is relaunched. */ |
d62a17ae | 2657 | static void rib_sweep_table(struct route_table *table) |
2658 | { | |
2659 | struct route_node *rn; | |
2660 | struct route_entry *re; | |
2661 | struct route_entry *next; | |
2662 | int ret = 0; | |
2663 | ||
2664 | if (table) | |
2665 | for (rn = route_top(table); rn; rn = srcdest_route_next(rn)) | |
2666 | RNODE_FOREACH_RE_SAFE(rn, re, next) | |
2667 | { | |
2668 | if (CHECK_FLAG(re->status, ROUTE_ENTRY_REMOVED)) | |
2669 | continue; | |
2670 | ||
2671 | if (re->type == ZEBRA_ROUTE_KERNEL | |
2672 | && CHECK_FLAG(re->flags, | |
2673 | ZEBRA_FLAG_SELFROUTE)) { | |
2674 | ret = rib_uninstall_kernel(rn, re); | |
2675 | if (!ret) | |
2676 | rib_delnode(rn, re); | |
2677 | } | |
2678 | } | |
718e3744 | 2679 | } |
2680 | ||
2681 | /* Sweep all RIB tables. */ | |
d62a17ae | 2682 | void rib_sweep_route(void) |
718e3744 | 2683 | { |
d62a17ae | 2684 | struct vrf *vrf; |
2685 | struct zebra_vrf *zvrf; | |
78104b9b | 2686 | |
d62a17ae | 2687 | RB_FOREACH(vrf, vrf_id_head, &vrfs_by_id) |
2688 | if ((zvrf = vrf->info) != NULL) { | |
2689 | rib_sweep_table(zvrf->table[AFI_IP][SAFI_UNICAST]); | |
2690 | rib_sweep_table(zvrf->table[AFI_IP6][SAFI_UNICAST]); | |
2691 | } | |
718e3744 | 2692 | } |
2ea1ab1c VT |
2693 | |
2694 | /* Remove specific by protocol routes from 'table'. */ | |
d62a17ae | 2695 | static unsigned long rib_score_proto_table(u_char proto, u_short instance, |
2696 | struct route_table *table) | |
2697 | { | |
2698 | struct route_node *rn; | |
2699 | struct route_entry *re; | |
2700 | struct route_entry *next; | |
2701 | unsigned long n = 0; | |
2702 | ||
2703 | if (table) | |
2704 | for (rn = route_top(table); rn; rn = srcdest_route_next(rn)) | |
2705 | RNODE_FOREACH_RE_SAFE(rn, re, next) | |
2706 | { | |
2707 | if (CHECK_FLAG(re->status, ROUTE_ENTRY_REMOVED)) | |
2708 | continue; | |
2709 | if (re->type == proto | |
2710 | && re->instance == instance) { | |
2711 | rib_delnode(rn, re); | |
2712 | n++; | |
2713 | } | |
2714 | } | |
2715 | return n; | |
2ea1ab1c VT |
2716 | } |
2717 | ||
2718 | /* Remove specific by protocol routes. */ | |
d62a17ae | 2719 | unsigned long rib_score_proto(u_char proto, u_short instance) |
2ea1ab1c | 2720 | { |
d62a17ae | 2721 | struct vrf *vrf; |
2722 | struct zebra_vrf *zvrf; | |
2723 | unsigned long cnt = 0; | |
78104b9b | 2724 | |
d62a17ae | 2725 | RB_FOREACH(vrf, vrf_id_head, &vrfs_by_id) |
2726 | if ((zvrf = vrf->info) != NULL) | |
2727 | cnt += rib_score_proto_table(proto, instance, | |
2728 | zvrf->table[AFI_IP][SAFI_UNICAST]) | |
2729 | + rib_score_proto_table( | |
2730 | proto, instance, | |
2731 | zvrf->table[AFI_IP6][SAFI_UNICAST]); | |
78104b9b | 2732 | |
d62a17ae | 2733 | return cnt; |
2ea1ab1c VT |
2734 | } |
2735 | ||
718e3744 | 2736 | /* Close RIB and clean up kernel routes. */ |
d62a17ae | 2737 | void rib_close_table(struct route_table *table) |
718e3744 | 2738 | { |
d62a17ae | 2739 | struct route_node *rn; |
2740 | rib_table_info_t *info = table->info; | |
2741 | struct route_entry *re; | |
718e3744 | 2742 | |
d62a17ae | 2743 | if (table) |
2744 | for (rn = route_top(table); rn; rn = srcdest_route_next(rn)) | |
2745 | RNODE_FOREACH_RE(rn, re) | |
2746 | { | |
2747 | if (!CHECK_FLAG(re->status, | |
2748 | ROUTE_ENTRY_SELECTED_FIB)) | |
2749 | continue; | |
9fd92e3c | 2750 | |
d62a17ae | 2751 | if (info->safi == SAFI_UNICAST) |
2752 | hook_call(rib_update, rn, NULL); | |
5adc2528 | 2753 | |
d62a17ae | 2754 | if (!RIB_SYSTEM_ROUTE(re)) |
2755 | rib_uninstall_kernel(rn, re); | |
2756 | } | |
718e3744 | 2757 | } |
2758 | ||
718e3744 | 2759 | /* Routing information base initialize. */ |
d62a17ae | 2760 | void rib_init(void) |
718e3744 | 2761 | { |
d62a17ae | 2762 | rib_queue_init(&zebrad); |
718e3744 | 2763 | } |
0915bb0c AS |
2764 | |
2765 | /* | |
2766 | * vrf_id_get_next | |
2767 | * | |
2768 | * Get the first vrf id that is greater than the given vrf id if any. | |
2769 | * | |
2770 | * Returns TRUE if a vrf id was found, FALSE otherwise. | |
2771 | */ | |
d62a17ae | 2772 | static inline int vrf_id_get_next(vrf_id_t vrf_id, vrf_id_t *next_id_p) |
0915bb0c | 2773 | { |
d62a17ae | 2774 | struct vrf *vrf; |
b72ede27 | 2775 | |
d62a17ae | 2776 | vrf = vrf_lookup_by_id(vrf_id); |
2777 | if (vrf) { | |
2778 | vrf = RB_NEXT(vrf_id_head, vrf); | |
2779 | if (vrf) { | |
2780 | *next_id_p = vrf->vrf_id; | |
2781 | return 1; | |
2782 | } | |
2783 | } | |
0915bb0c | 2784 | |
d62a17ae | 2785 | return 0; |
0915bb0c AS |
2786 | } |
2787 | ||
2788 | /* | |
2789 | * rib_tables_iter_next | |
2790 | * | |
2791 | * Returns the next table in the iteration. | |
2792 | */ | |
d62a17ae | 2793 | struct route_table *rib_tables_iter_next(rib_tables_iter_t *iter) |
2794 | { | |
2795 | struct route_table *table; | |
2796 | ||
2797 | /* | |
2798 | * Array that helps us go over all AFI/SAFI combinations via one | |
2799 | * index. | |
2800 | */ | |
2801 | static struct { | |
2802 | afi_t afi; | |
2803 | safi_t safi; | |
2804 | } afi_safis[] = { | |
2805 | {AFI_IP, SAFI_UNICAST}, {AFI_IP, SAFI_MULTICAST}, | |
2806 | {AFI_IP, SAFI_LABELED_UNICAST}, {AFI_IP6, SAFI_UNICAST}, | |
2807 | {AFI_IP6, SAFI_MULTICAST}, {AFI_IP6, SAFI_LABELED_UNICAST}, | |
2808 | }; | |
2809 | ||
2810 | table = NULL; | |
2811 | ||
2812 | switch (iter->state) { | |
2813 | ||
2814 | case RIB_TABLES_ITER_S_INIT: | |
2815 | iter->vrf_id = VRF_DEFAULT; | |
2816 | iter->afi_safi_ix = -1; | |
2817 | ||
2818 | /* Fall through */ | |
2819 | ||
2820 | case RIB_TABLES_ITER_S_ITERATING: | |
2821 | iter->afi_safi_ix++; | |
2822 | while (1) { | |
2823 | ||
2824 | while (iter->afi_safi_ix | |
2825 | < (int)ZEBRA_NUM_OF(afi_safis)) { | |
2826 | table = zebra_vrf_table( | |
2827 | afi_safis[iter->afi_safi_ix].afi, | |
2828 | afi_safis[iter->afi_safi_ix].safi, | |
2829 | iter->vrf_id); | |
2830 | if (table) | |
2831 | break; | |
2832 | ||
2833 | iter->afi_safi_ix++; | |
2834 | } | |
2835 | ||
2836 | /* | |
2837 | * Found another table in this vrf. | |
2838 | */ | |
2839 | if (table) | |
2840 | break; | |
2841 | ||
2842 | /* | |
2843 | * Done with all tables in the current vrf, go to the | |
2844 | * next | |
2845 | * one. | |
2846 | */ | |
2847 | if (!vrf_id_get_next(iter->vrf_id, &iter->vrf_id)) | |
2848 | break; | |
2849 | ||
2850 | iter->afi_safi_ix = 0; | |
2851 | } | |
0915bb0c | 2852 | |
0915bb0c AS |
2853 | break; |
2854 | ||
d62a17ae | 2855 | case RIB_TABLES_ITER_S_DONE: |
2856 | return NULL; | |
0915bb0c AS |
2857 | } |
2858 | ||
d62a17ae | 2859 | if (table) |
2860 | iter->state = RIB_TABLES_ITER_S_ITERATING; | |
2861 | else | |
2862 | iter->state = RIB_TABLES_ITER_S_DONE; | |
0915bb0c | 2863 | |
d62a17ae | 2864 | return table; |
0915bb0c | 2865 | } |