]>
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] = { | |
9d303b37 DL |
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}, | |
d62a17ae | 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 | |
a8309422 | 294 | struct nexthop *route_entry_nexthop_blackhole_add(struct route_entry *re, |
60466a63 | 295 | enum blackhole_type bh_type) |
595db7f1 | 296 | { |
d62a17ae | 297 | struct nexthop *nexthop; |
595db7f1 | 298 | |
d62a17ae | 299 | nexthop = nexthop_new(); |
300 | nexthop->type = NEXTHOP_TYPE_BLACKHOLE; | |
a8309422 | 301 | nexthop->bh_type = bh_type; |
595db7f1 | 302 | |
d62a17ae | 303 | route_entry_nexthop_add(re, nexthop); |
595db7f1 | 304 | |
d62a17ae | 305 | return nexthop; |
595db7f1 | 306 | } |
307 | ||
d62a17ae | 308 | static void nexthop_set_resolved(afi_t afi, struct nexthop *newhop, |
309 | struct nexthop *nexthop) | |
4491a88f | 310 | { |
d62a17ae | 311 | struct nexthop *resolved_hop; |
4491a88f | 312 | |
d62a17ae | 313 | resolved_hop = nexthop_new(); |
314 | SET_FLAG(resolved_hop->flags, NEXTHOP_FLAG_ACTIVE); | |
315 | /* If the resolving route specifies a gateway, use it */ | |
316 | if (newhop->type == NEXTHOP_TYPE_IPV4 | |
317 | || newhop->type == NEXTHOP_TYPE_IPV4_IFINDEX) { | |
318 | resolved_hop->type = newhop->type; | |
319 | resolved_hop->gate.ipv4 = newhop->gate.ipv4; | |
4491a88f | 320 | |
d62a17ae | 321 | if (newhop->ifindex) { |
322 | resolved_hop->type = NEXTHOP_TYPE_IPV4_IFINDEX; | |
323 | resolved_hop->ifindex = newhop->ifindex; | |
324 | if (newhop->flags & NEXTHOP_FLAG_ONLINK) | |
325 | resolved_hop->flags |= NEXTHOP_FLAG_ONLINK; | |
326 | } | |
4491a88f | 327 | } |
d62a17ae | 328 | if (newhop->type == NEXTHOP_TYPE_IPV6 |
329 | || newhop->type == NEXTHOP_TYPE_IPV6_IFINDEX) { | |
330 | resolved_hop->type = newhop->type; | |
331 | resolved_hop->gate.ipv6 = newhop->gate.ipv6; | |
332 | ||
333 | if (newhop->ifindex) { | |
334 | resolved_hop->type = NEXTHOP_TYPE_IPV6_IFINDEX; | |
335 | resolved_hop->ifindex = newhop->ifindex; | |
336 | } | |
4491a88f | 337 | } |
4491a88f | 338 | |
d62a17ae | 339 | /* If the resolving route is an interface route, |
340 | * it means the gateway we are looking up is connected | |
341 | * to that interface. (The actual network is _not_ onlink). | |
342 | * Therefore, the resolved route should have the original | |
343 | * gateway as nexthop as it is directly connected. | |
344 | * | |
345 | * On Linux, we have to set the onlink netlink flag because | |
346 | * otherwise, the kernel won't accept the route. | |
347 | */ | |
348 | if (newhop->type == NEXTHOP_TYPE_IFINDEX) { | |
349 | resolved_hop->flags |= NEXTHOP_FLAG_ONLINK; | |
350 | if (afi == AFI_IP) { | |
351 | resolved_hop->type = NEXTHOP_TYPE_IPV4_IFINDEX; | |
352 | resolved_hop->gate.ipv4 = nexthop->gate.ipv4; | |
353 | } else if (afi == AFI_IP6) { | |
354 | resolved_hop->type = NEXTHOP_TYPE_IPV6_IFINDEX; | |
355 | resolved_hop->gate.ipv6 = nexthop->gate.ipv6; | |
356 | } | |
357 | resolved_hop->ifindex = newhop->ifindex; | |
4491a88f | 358 | } |
4491a88f | 359 | |
d62a17ae | 360 | resolved_hop->rparent = nexthop; |
361 | nexthop_add(&nexthop->resolved, resolved_hop); | |
4491a88f DS |
362 | } |
363 | ||
718e3744 | 364 | /* If force flag is not set, do not modify falgs at all for uninstall |
365 | the route from FIB. */ | |
d62a17ae | 366 | static int nexthop_active(afi_t afi, struct route_entry *re, |
367 | struct nexthop *nexthop, int set, | |
368 | struct route_node *top) | |
369 | { | |
370 | struct prefix p; | |
371 | struct route_table *table; | |
372 | struct route_node *rn; | |
373 | struct route_entry *match; | |
374 | int resolved; | |
375 | struct nexthop *newhop; | |
376 | struct interface *ifp; | |
377 | ||
378 | if ((nexthop->type == NEXTHOP_TYPE_IPV4) | |
379 | || nexthop->type == NEXTHOP_TYPE_IPV6) | |
380 | nexthop->ifindex = 0; | |
381 | ||
382 | if (set) { | |
383 | UNSET_FLAG(nexthop->flags, NEXTHOP_FLAG_RECURSIVE); | |
384 | zebra_deregister_rnh_static_nexthops(re->vrf_id, | |
385 | nexthop->resolved, top); | |
386 | nexthops_free(nexthop->resolved); | |
387 | nexthop->resolved = NULL; | |
388 | re->nexthop_mtu = 0; | |
d44ca835 | 389 | } |
18ff3edd | 390 | |
d62a17ae | 391 | /* Skip nexthops that have been filtered out due to route-map */ |
392 | /* The nexthops are specific to this route and so the same */ | |
393 | /* nexthop for a different route may not have this flag set */ | |
394 | if (CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_FILTERED)) | |
395 | return 0; | |
396 | ||
397 | /* | |
398 | * Check to see if we should trust the passed in information | |
399 | * for UNNUMBERED interfaces as that we won't find the GW | |
400 | * address in the routing table. | |
401 | */ | |
402 | if (CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_ONLINK)) { | |
403 | ifp = if_lookup_by_index(nexthop->ifindex, VRF_DEFAULT); | |
404 | if (ifp && connected_is_unnumbered(ifp)) { | |
405 | if (if_is_operative(ifp)) | |
406 | return 1; | |
407 | else | |
408 | return 0; | |
409 | } else | |
410 | return 0; | |
16814f96 | 411 | } |
718e3744 | 412 | |
d62a17ae | 413 | /* Make lookup prefix. */ |
414 | memset(&p, 0, sizeof(struct prefix)); | |
415 | switch (afi) { | |
416 | case AFI_IP: | |
417 | p.family = AF_INET; | |
418 | p.prefixlen = IPV4_MAX_PREFIXLEN; | |
419 | p.u.prefix4 = nexthop->gate.ipv4; | |
420 | break; | |
421 | case AFI_IP6: | |
422 | p.family = AF_INET6; | |
423 | p.prefixlen = IPV6_MAX_PREFIXLEN; | |
424 | p.u.prefix6 = nexthop->gate.ipv6; | |
425 | break; | |
426 | default: | |
427 | assert(afi != AFI_IP && afi != AFI_IP6); | |
428 | break; | |
718e3744 | 429 | } |
d62a17ae | 430 | /* Lookup table. */ |
431 | table = zebra_vrf_table(afi, SAFI_UNICAST, re->vrf_id); | |
432 | if (!table) | |
433 | return 0; | |
434 | ||
435 | rn = route_node_match(table, (struct prefix *)&p); | |
436 | while (rn) { | |
437 | route_unlock_node(rn); | |
438 | ||
439 | /* If lookup self prefix return immediately. */ | |
440 | if (rn == top) | |
441 | return 0; | |
442 | ||
443 | /* Pick up selected route. */ | |
444 | /* However, do not resolve over default route unless explicitly | |
445 | * allowed. */ | |
446 | if (is_default_prefix(&rn->p) | |
447 | && !nh_resolve_via_default(p.family)) | |
448 | return 0; | |
449 | ||
450 | RNODE_FOREACH_RE(rn, match) | |
451 | { | |
452 | if (CHECK_FLAG(match->status, ROUTE_ENTRY_REMOVED)) | |
453 | continue; | |
454 | ||
455 | /* if the next hop is imported from another table, skip | |
456 | * it */ | |
457 | if (match->type == ZEBRA_ROUTE_TABLE) | |
458 | continue; | |
459 | if (CHECK_FLAG(match->status, ROUTE_ENTRY_SELECTED_FIB)) | |
460 | break; | |
461 | } | |
48a53dc7 | 462 | |
d62a17ae | 463 | /* If there is no selected route or matched route is EGP, go up |
464 | tree. */ | |
465 | if (!match) { | |
466 | do { | |
467 | rn = rn->parent; | |
468 | } while (rn && rn->info == NULL); | |
469 | if (rn) | |
470 | route_lock_node(rn); | |
c87bdd2b | 471 | |
d62a17ae | 472 | continue; |
473 | } | |
474 | ||
d62a17ae | 475 | if (match->type == ZEBRA_ROUTE_CONNECT) { |
476 | /* Directly point connected route. */ | |
477 | newhop = match->nexthop; | |
478 | if (newhop) { | |
479 | if (nexthop->type == NEXTHOP_TYPE_IPV4 | |
480 | || nexthop->type == NEXTHOP_TYPE_IPV6) | |
481 | nexthop->ifindex = newhop->ifindex; | |
482 | } | |
483 | return 1; | |
484 | } else if (CHECK_FLAG(re->flags, ZEBRA_FLAG_INTERNAL)) { | |
485 | resolved = 0; | |
a8309422 DL |
486 | for (ALL_NEXTHOPS(match->nexthop, newhop)) { |
487 | if (newhop->type == NEXTHOP_TYPE_BLACKHOLE) | |
488 | continue; | |
489 | if (!CHECK_FLAG(newhop->flags, | |
490 | NEXTHOP_FLAG_FIB)) | |
491 | continue; | |
492 | if (CHECK_FLAG(newhop->flags, | |
493 | NEXTHOP_FLAG_RECURSIVE)) | |
494 | continue; | |
495 | ||
496 | if (set) { | |
497 | SET_FLAG(nexthop->flags, | |
498 | NEXTHOP_FLAG_RECURSIVE); | |
499 | SET_FLAG(re->status, | |
500 | ROUTE_ENTRY_NEXTHOPS_CHANGED); | |
501 | nexthop_set_resolved(afi, newhop, | |
502 | nexthop); | |
d62a17ae | 503 | } |
a8309422 DL |
504 | resolved = 1; |
505 | } | |
43e31305 JB |
506 | if (resolved && set) |
507 | re->nexthop_mtu = match->mtu; | |
d62a17ae | 508 | return resolved; |
509 | } else if (re->type == ZEBRA_ROUTE_STATIC) { | |
510 | resolved = 0; | |
a8309422 DL |
511 | for (ALL_NEXTHOPS(match->nexthop, newhop)) { |
512 | if (newhop->type == NEXTHOP_TYPE_BLACKHOLE) | |
513 | continue; | |
514 | if (!CHECK_FLAG(newhop->flags, | |
515 | NEXTHOP_FLAG_FIB)) | |
516 | continue; | |
517 | ||
518 | if (set) { | |
519 | SET_FLAG(nexthop->flags, | |
520 | NEXTHOP_FLAG_RECURSIVE); | |
521 | nexthop_set_resolved(afi, newhop, | |
522 | nexthop); | |
d62a17ae | 523 | } |
a8309422 DL |
524 | resolved = 1; |
525 | } | |
d62a17ae | 526 | if (resolved && set) |
527 | re->nexthop_mtu = match->mtu; | |
528 | return resolved; | |
529 | } else { | |
530 | return 0; | |
531 | } | |
718e3744 | 532 | } |
d62a17ae | 533 | return 0; |
718e3744 | 534 | } |
718e3744 | 535 | |
d62a17ae | 536 | struct route_entry *rib_match(afi_t afi, safi_t safi, vrf_id_t vrf_id, |
537 | union g_addr *addr, struct route_node **rn_out) | |
538 | { | |
539 | struct prefix p; | |
540 | struct route_table *table; | |
541 | struct route_node *rn; | |
542 | struct route_entry *match; | |
543 | struct nexthop *newhop; | |
544 | ||
545 | /* Lookup table. */ | |
546 | table = zebra_vrf_table(afi, safi, vrf_id); | |
547 | if (!table) | |
548 | return 0; | |
549 | ||
550 | memset(&p, 0, sizeof(struct prefix)); | |
551 | p.family = afi; | |
552 | if (afi == AFI_IP) { | |
553 | p.u.prefix4 = addr->ipv4; | |
554 | p.prefixlen = IPV4_MAX_PREFIXLEN; | |
555 | } else { | |
556 | p.u.prefix6 = addr->ipv6; | |
557 | p.prefixlen = IPV6_MAX_PREFIXLEN; | |
558 | } | |
718e3744 | 559 | |
d62a17ae | 560 | rn = route_node_match(table, (struct prefix *)&p); |
718e3744 | 561 | |
d62a17ae | 562 | while (rn) { |
563 | route_unlock_node(rn); | |
718e3744 | 564 | |
d62a17ae | 565 | /* Pick up selected route. */ |
566 | RNODE_FOREACH_RE(rn, match) | |
567 | { | |
568 | if (CHECK_FLAG(match->status, ROUTE_ENTRY_REMOVED)) | |
569 | continue; | |
570 | if (CHECK_FLAG(match->status, ROUTE_ENTRY_SELECTED_FIB)) | |
571 | break; | |
572 | } | |
718e3744 | 573 | |
d62a17ae | 574 | /* If there is no selected route or matched route is EGP, go up |
575 | tree. */ | |
576 | if (!match) { | |
577 | do { | |
578 | rn = rn->parent; | |
579 | } while (rn && rn->info == NULL); | |
580 | if (rn) | |
581 | route_lock_node(rn); | |
582 | } else { | |
583 | if (match->type != ZEBRA_ROUTE_CONNECT) { | |
584 | int found = 0; | |
585 | for (ALL_NEXTHOPS(match->nexthop, newhop)) | |
586 | if (CHECK_FLAG(newhop->flags, | |
587 | NEXTHOP_FLAG_FIB)) { | |
588 | found = 1; | |
589 | break; | |
590 | } | |
591 | if (!found) | |
592 | return NULL; | |
593 | } | |
594 | ||
595 | if (rn_out) | |
596 | *rn_out = rn; | |
597 | return match; | |
598 | } | |
16814f96 | 599 | } |
d62a17ae | 600 | return NULL; |
601 | } | |
602 | ||
603 | struct route_entry *rib_match_ipv4_multicast(vrf_id_t vrf_id, | |
604 | struct in_addr addr, | |
605 | struct route_node **rn_out) | |
606 | { | |
607 | struct route_entry *re = NULL, *mre = NULL, *ure = NULL; | |
608 | struct route_node *m_rn = NULL, *u_rn = NULL; | |
609 | union g_addr gaddr = {.ipv4 = addr}; | |
610 | ||
611 | switch (ipv4_multicast_mode) { | |
612 | case MCAST_MRIB_ONLY: | |
613 | return rib_match(AFI_IP, SAFI_MULTICAST, vrf_id, &gaddr, | |
614 | rn_out); | |
615 | case MCAST_URIB_ONLY: | |
616 | return rib_match(AFI_IP, SAFI_UNICAST, vrf_id, &gaddr, rn_out); | |
617 | case MCAST_NO_CONFIG: | |
618 | case MCAST_MIX_MRIB_FIRST: | |
619 | re = mre = rib_match(AFI_IP, SAFI_MULTICAST, vrf_id, &gaddr, | |
620 | &m_rn); | |
621 | if (!mre) | |
622 | re = ure = rib_match(AFI_IP, SAFI_UNICAST, vrf_id, | |
623 | &gaddr, &u_rn); | |
624 | break; | |
625 | case MCAST_MIX_DISTANCE: | |
626 | mre = rib_match(AFI_IP, SAFI_MULTICAST, vrf_id, &gaddr, &m_rn); | |
627 | ure = rib_match(AFI_IP, SAFI_UNICAST, vrf_id, &gaddr, &u_rn); | |
628 | if (mre && ure) | |
629 | re = ure->distance < mre->distance ? ure : mre; | |
630 | else if (mre) | |
631 | re = mre; | |
632 | else if (ure) | |
633 | re = ure; | |
634 | break; | |
635 | case MCAST_MIX_PFXLEN: | |
636 | mre = rib_match(AFI_IP, SAFI_MULTICAST, vrf_id, &gaddr, &m_rn); | |
637 | ure = rib_match(AFI_IP, SAFI_UNICAST, vrf_id, &gaddr, &u_rn); | |
638 | if (mre && ure) | |
639 | re = u_rn->p.prefixlen > m_rn->p.prefixlen ? ure : mre; | |
640 | else if (mre) | |
641 | re = mre; | |
642 | else if (ure) | |
643 | re = ure; | |
644 | break; | |
718e3744 | 645 | } |
33550aa8 | 646 | |
d62a17ae | 647 | if (rn_out) |
648 | *rn_out = (re == mre) ? m_rn : u_rn; | |
718e3744 | 649 | |
d62a17ae | 650 | if (IS_ZEBRA_DEBUG_RIB) { |
651 | char buf[BUFSIZ]; | |
652 | inet_ntop(AF_INET, &addr, buf, BUFSIZ); | |
4623d897 | 653 | |
d62a17ae | 654 | zlog_debug("%s: %s: found %s, using %s", __func__, buf, |
655 | mre ? (ure ? "MRIB+URIB" : "MRIB") | |
656 | : ure ? "URIB" : "nothing", | |
657 | re == ure ? "URIB" : re == mre ? "MRIB" : "none"); | |
658 | } | |
659 | return re; | |
4623d897 DL |
660 | } |
661 | ||
d62a17ae | 662 | void multicast_mode_ipv4_set(enum multicast_mode mode) |
4623d897 | 663 | { |
d62a17ae | 664 | if (IS_ZEBRA_DEBUG_RIB) |
665 | zlog_debug("%s: multicast lookup mode set (%d)", __func__, | |
666 | mode); | |
667 | ipv4_multicast_mode = mode; | |
4623d897 DL |
668 | } |
669 | ||
d62a17ae | 670 | enum multicast_mode multicast_mode_ipv4_get(void) |
4623d897 | 671 | { |
d62a17ae | 672 | return ipv4_multicast_mode; |
4623d897 DL |
673 | } |
674 | ||
d62a17ae | 675 | struct route_entry *rib_lookup_ipv4(struct prefix_ipv4 *p, vrf_id_t vrf_id) |
718e3744 | 676 | { |
d62a17ae | 677 | struct route_table *table; |
678 | struct route_node *rn; | |
679 | struct route_entry *match; | |
680 | struct nexthop *nexthop; | |
718e3744 | 681 | |
d62a17ae | 682 | /* Lookup table. */ |
683 | table = zebra_vrf_table(AFI_IP, SAFI_UNICAST, vrf_id); | |
684 | if (!table) | |
685 | return 0; | |
718e3744 | 686 | |
d62a17ae | 687 | rn = route_node_lookup(table, (struct prefix *)p); |
718e3744 | 688 | |
d62a17ae | 689 | /* No route for this prefix. */ |
690 | if (!rn) | |
691 | return NULL; | |
718e3744 | 692 | |
d62a17ae | 693 | /* Unlock node. */ |
694 | route_unlock_node(rn); | |
718e3744 | 695 | |
d62a17ae | 696 | RNODE_FOREACH_RE(rn, match) |
697 | { | |
698 | if (CHECK_FLAG(match->status, ROUTE_ENTRY_REMOVED)) | |
699 | continue; | |
700 | if (CHECK_FLAG(match->status, ROUTE_ENTRY_SELECTED_FIB)) | |
701 | break; | |
702 | } | |
718e3744 | 703 | |
d62a17ae | 704 | if (!match) |
705 | return NULL; | |
718e3744 | 706 | |
d62a17ae | 707 | if (match->type == ZEBRA_ROUTE_CONNECT) |
708 | return match; | |
f9e1b38e | 709 | |
d62a17ae | 710 | for (ALL_NEXTHOPS(match->nexthop, nexthop)) |
711 | if (CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_FIB)) | |
712 | return match; | |
718e3744 | 713 | |
d62a17ae | 714 | return NULL; |
718e3744 | 715 | } |
716 | ||
dc95824a DO |
717 | /* |
718 | * This clone function, unlike its original rib_lookup_ipv4(), checks | |
719 | * if specified IPv4 route record (prefix/mask -> gate) exists in | |
f0f77c9a | 720 | * the whole RIB and has ROUTE_ENTRY_SELECTED_FIB set. |
dc95824a DO |
721 | * |
722 | * Return values: | |
723 | * -1: error | |
724 | * 0: exact match found | |
725 | * 1: a match was found with a different gate | |
726 | * 2: connected route found | |
727 | * 3: no matches found | |
728 | */ | |
d62a17ae | 729 | int rib_lookup_ipv4_route(struct prefix_ipv4 *p, union sockunion *qgate, |
730 | vrf_id_t vrf_id) | |
dc95824a | 731 | { |
d62a17ae | 732 | struct route_table *table; |
733 | struct route_node *rn; | |
734 | struct route_entry *match; | |
735 | struct nexthop *nexthop; | |
736 | int nexthops_active; | |
dc95824a | 737 | |
d62a17ae | 738 | /* Lookup table. */ |
739 | table = zebra_vrf_table(AFI_IP, SAFI_UNICAST, vrf_id); | |
740 | if (!table) | |
741 | return ZEBRA_RIB_LOOKUP_ERROR; | |
dc95824a | 742 | |
d62a17ae | 743 | /* Scan the RIB table for exactly matching RIB entry. */ |
744 | rn = route_node_lookup(table, (struct prefix *)p); | |
dc95824a | 745 | |
d62a17ae | 746 | /* No route for this prefix. */ |
747 | if (!rn) | |
748 | return ZEBRA_RIB_NOTFOUND; | |
dc95824a | 749 | |
d62a17ae | 750 | /* Unlock node. */ |
751 | route_unlock_node(rn); | |
dc95824a | 752 | |
d62a17ae | 753 | /* Find out if a "selected" RR for the discovered RIB entry exists ever. |
754 | */ | |
755 | RNODE_FOREACH_RE(rn, match) | |
756 | { | |
757 | if (CHECK_FLAG(match->status, ROUTE_ENTRY_REMOVED)) | |
758 | continue; | |
759 | if (CHECK_FLAG(match->status, ROUTE_ENTRY_SELECTED_FIB)) | |
760 | break; | |
761 | } | |
dc95824a | 762 | |
d62a17ae | 763 | /* None such found :( */ |
764 | if (!match) | |
765 | return ZEBRA_RIB_NOTFOUND; | |
766 | ||
767 | if (match->type == ZEBRA_ROUTE_CONNECT) | |
768 | return ZEBRA_RIB_FOUND_CONNECTED; | |
769 | ||
770 | /* Ok, we have a cood candidate, let's check it's nexthop list... */ | |
771 | nexthops_active = 0; | |
772 | for (ALL_NEXTHOPS(match->nexthop, nexthop)) | |
773 | if (CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_FIB)) { | |
774 | nexthops_active = 1; | |
775 | if (nexthop->gate.ipv4.s_addr == sockunion2ip(qgate)) | |
776 | return ZEBRA_RIB_FOUND_EXACT; | |
777 | if (IS_ZEBRA_DEBUG_RIB) { | |
778 | char gate_buf[INET_ADDRSTRLEN], | |
779 | qgate_buf[INET_ADDRSTRLEN]; | |
780 | inet_ntop(AF_INET, &nexthop->gate.ipv4.s_addr, | |
781 | gate_buf, INET_ADDRSTRLEN); | |
782 | inet_ntop(AF_INET, &sockunion2ip(qgate), | |
783 | qgate_buf, INET_ADDRSTRLEN); | |
784 | zlog_debug("%s: qgate == %s, %s == %s", | |
785 | __func__, qgate_buf, | |
786 | nexthop->rparent ? "rgate" : "gate", | |
787 | gate_buf); | |
788 | } | |
789 | } | |
790 | ||
791 | if (nexthops_active) | |
792 | return ZEBRA_RIB_FOUND_NOGATE; | |
793 | ||
794 | return ZEBRA_RIB_NOTFOUND; | |
795 | } | |
796 | ||
797 | #define RIB_SYSTEM_ROUTE(R) \ | |
798 | ((R)->type == ZEBRA_ROUTE_KERNEL || (R)->type == ZEBRA_ROUTE_CONNECT) | |
7514fb77 | 799 | |
dc95824a DO |
800 | /* This function verifies reachability of one given nexthop, which can be |
801 | * numbered or unnumbered, IPv4 or IPv6. The result is unconditionally stored | |
802 | * in nexthop->flags field. If the 4th parameter, 'set', is non-zero, | |
803 | * nexthop->ifindex will be updated appropriately as well. | |
804 | * An existing route map can turn (otherwise active) nexthop into inactive, but | |
805 | * not vice versa. | |
806 | * | |
807 | * The return value is the final value of 'ACTIVE' flag. | |
808 | */ | |
809 | ||
d62a17ae | 810 | static unsigned nexthop_active_check(struct route_node *rn, |
811 | struct route_entry *re, | |
812 | struct nexthop *nexthop, int set) | |
813 | { | |
814 | struct interface *ifp; | |
815 | route_map_result_t ret = RMAP_MATCH; | |
816 | int family; | |
817 | char buf[SRCDEST2STR_BUFFER]; | |
818 | struct prefix *p, *src_p; | |
819 | srcdest_rnode_prefixes(rn, &p, &src_p); | |
820 | ||
821 | if (rn->p.family == AF_INET) | |
822 | family = AFI_IP; | |
823 | else if (rn->p.family == AF_INET6) | |
824 | family = AFI_IP6; | |
825 | else | |
826 | family = 0; | |
827 | switch (nexthop->type) { | |
828 | case NEXTHOP_TYPE_IFINDEX: | |
829 | ifp = if_lookup_by_index(nexthop->ifindex, re->vrf_id); | |
830 | if (ifp && if_is_operative(ifp)) | |
831 | SET_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE); | |
832 | else | |
833 | UNSET_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE); | |
834 | break; | |
835 | case NEXTHOP_TYPE_IPV4: | |
836 | case NEXTHOP_TYPE_IPV4_IFINDEX: | |
837 | family = AFI_IP; | |
838 | if (nexthop_active(AFI_IP, re, nexthop, set, rn)) | |
839 | SET_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE); | |
840 | else | |
841 | UNSET_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE); | |
842 | break; | |
843 | case NEXTHOP_TYPE_IPV6: | |
844 | family = AFI_IP6; | |
845 | if (nexthop_active(AFI_IP6, re, nexthop, set, rn)) | |
846 | SET_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE); | |
847 | else | |
848 | UNSET_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE); | |
849 | break; | |
850 | case NEXTHOP_TYPE_IPV6_IFINDEX: | |
851 | /* RFC 5549, v4 prefix with v6 NH */ | |
852 | if (rn->p.family != AF_INET) | |
853 | family = AFI_IP6; | |
854 | if (IN6_IS_ADDR_LINKLOCAL(&nexthop->gate.ipv6)) { | |
855 | ifp = if_lookup_by_index(nexthop->ifindex, re->vrf_id); | |
856 | if (ifp && if_is_operative(ifp)) | |
857 | SET_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE); | |
858 | else | |
859 | UNSET_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE); | |
860 | } else { | |
861 | if (nexthop_active(AFI_IP6, re, nexthop, set, rn)) | |
862 | SET_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE); | |
863 | else | |
864 | UNSET_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE); | |
865 | } | |
866 | break; | |
867 | case NEXTHOP_TYPE_BLACKHOLE: | |
868 | SET_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE); | |
869 | break; | |
870 | default: | |
871 | break; | |
872 | } | |
873 | if (!CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE)) | |
874 | return 0; | |
875 | ||
876 | /* XXX: What exactly do those checks do? Do we support | |
877 | * e.g. IPv4 routes with IPv6 nexthops or vice versa? */ | |
878 | if (RIB_SYSTEM_ROUTE(re) || (family == AFI_IP && p->family != AF_INET) | |
879 | || (family == AFI_IP6 && p->family != AF_INET6)) | |
880 | return CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE); | |
881 | ||
882 | /* The original code didn't determine the family correctly | |
883 | * e.g. for NEXTHOP_TYPE_IFINDEX. Retrieve the correct afi | |
884 | * from the rib_table_info in those cases. | |
885 | * Possibly it may be better to use only the rib_table_info | |
886 | * in every case. | |
887 | */ | |
888 | if (!family) { | |
889 | rib_table_info_t *info; | |
890 | ||
891 | info = srcdest_rnode_table_info(rn); | |
892 | family = info->afi; | |
718e3744 | 893 | } |
c52ef59f | 894 | |
d62a17ae | 895 | memset(&nexthop->rmap_src.ipv6, 0, sizeof(union g_addr)); |
896 | ||
897 | /* It'll get set if required inside */ | |
898 | ret = zebra_route_map_check(family, re->type, p, nexthop, re->vrf_id, | |
899 | re->tag); | |
900 | if (ret == RMAP_DENYMATCH) { | |
901 | if (IS_ZEBRA_DEBUG_RIB) { | |
902 | srcdest_rnode2str(rn, buf, sizeof(buf)); | |
903 | zlog_debug( | |
904 | "%u:%s: Filtering out with NH out %s due to route map", | |
905 | re->vrf_id, buf, | |
906 | ifindex2ifname(nexthop->ifindex, re->vrf_id)); | |
907 | } | |
908 | UNSET_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE); | |
518f0eb1 | 909 | } |
d62a17ae | 910 | return CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE); |
718e3744 | 911 | } |
912 | ||
03e232a4 | 913 | /* Iterate over all nexthops of the given RIB entry and refresh their |
f0f77c9a DS |
914 | * ACTIVE flag. re->nexthop_active_num is updated accordingly. If any |
915 | * nexthop is found to toggle the ACTIVE flag, the whole re structure | |
916 | * is flagged with ROUTE_ENTRY_CHANGED. The 4th 'set' argument is | |
03e232a4 DO |
917 | * transparently passed to nexthop_active_check(). |
918 | * | |
919 | * Return value is the new number of active nexthops. | |
920 | */ | |
921 | ||
d62a17ae | 922 | static int nexthop_active_update(struct route_node *rn, struct route_entry *re, |
923 | int set) | |
924 | { | |
925 | struct nexthop *nexthop; | |
926 | union g_addr prev_src; | |
927 | unsigned int prev_active, new_active, old_num_nh; | |
928 | ifindex_t prev_index; | |
929 | old_num_nh = re->nexthop_active_num; | |
930 | ||
931 | re->nexthop_active_num = 0; | |
932 | UNSET_FLAG(re->status, ROUTE_ENTRY_CHANGED); | |
933 | ||
934 | for (nexthop = re->nexthop; nexthop; nexthop = nexthop->next) { | |
935 | /* No protocol daemon provides src and so we're skipping | |
936 | * tracking it */ | |
937 | prev_src = nexthop->rmap_src; | |
938 | prev_active = CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE); | |
939 | prev_index = nexthop->ifindex; | |
940 | if ((new_active = nexthop_active_check(rn, re, nexthop, set))) | |
941 | re->nexthop_active_num++; | |
942 | /* Don't allow src setting on IPv6 addr for now */ | |
943 | if (prev_active != new_active || prev_index != nexthop->ifindex | |
944 | || ((nexthop->type >= NEXTHOP_TYPE_IFINDEX | |
945 | && nexthop->type < NEXTHOP_TYPE_IPV6) | |
946 | && prev_src.ipv4.s_addr | |
947 | != nexthop->rmap_src.ipv4.s_addr) | |
948 | || ((nexthop->type >= NEXTHOP_TYPE_IPV6 | |
949 | && nexthop->type < NEXTHOP_TYPE_BLACKHOLE) | |
950 | && !(IPV6_ADDR_SAME(&prev_src.ipv6, | |
951 | &nexthop->rmap_src.ipv6)))) { | |
952 | SET_FLAG(re->status, ROUTE_ENTRY_CHANGED); | |
953 | SET_FLAG(re->status, ROUTE_ENTRY_NEXTHOPS_CHANGED); | |
954 | } | |
955 | } | |
956 | ||
957 | if (old_num_nh != re->nexthop_active_num) | |
958 | SET_FLAG(re->status, ROUTE_ENTRY_CHANGED); | |
6e26278c | 959 | |
d62a17ae | 960 | if (CHECK_FLAG(re->status, ROUTE_ENTRY_CHANGED)) { |
961 | SET_FLAG(re->status, ROUTE_ENTRY_NEXTHOPS_CHANGED); | |
962 | } | |
963 | ||
964 | return re->nexthop_active_num; | |
718e3744 | 965 | } |
6baeb988 | 966 | |
a64448ba DS |
967 | /* |
968 | * Is this RIB labeled-unicast? It must be of type BGP and all paths | |
969 | * (nexthops) must have a label. | |
970 | */ | |
d62a17ae | 971 | int zebra_rib_labeled_unicast(struct route_entry *re) |
a64448ba | 972 | { |
d62a17ae | 973 | struct nexthop *nexthop = NULL; |
a64448ba | 974 | |
d62a17ae | 975 | if (re->type != ZEBRA_ROUTE_BGP) |
976 | return 0; | |
a64448ba | 977 | |
d62a17ae | 978 | for (ALL_NEXTHOPS(re->nexthop, nexthop)) |
979 | if (!nexthop->nh_label || !nexthop->nh_label->num_labels) | |
980 | return 0; | |
6b0655a2 | 981 | |
d62a17ae | 982 | return 1; |
a64448ba | 983 | } |
718e3744 | 984 | |
6ae24471 DS |
985 | /* Update flag indicates whether this is a "replace" or not. Currently, this |
986 | * is only used for IPv4. | |
987 | */ | |
d62a17ae | 988 | int rib_install_kernel(struct route_node *rn, struct route_entry *re, |
989 | struct route_entry *old) | |
718e3744 | 990 | { |
d62a17ae | 991 | int ret = 0; |
992 | struct nexthop *nexthop; | |
993 | rib_table_info_t *info = srcdest_rnode_table_info(rn); | |
994 | struct prefix *p, *src_p; | |
995 | struct zebra_vrf *zvrf = vrf_info_lookup(re->vrf_id); | |
718e3744 | 996 | |
d62a17ae | 997 | srcdest_rnode_prefixes(rn, &p, &src_p); |
416ec78d | 998 | |
d62a17ae | 999 | if (info->safi != SAFI_UNICAST) { |
1000 | for (ALL_NEXTHOPS(re->nexthop, nexthop)) | |
1001 | SET_FLAG(nexthop->flags, NEXTHOP_FLAG_FIB); | |
1002 | return ret; | |
1003 | } | |
718e3744 | 1004 | |
d62a17ae | 1005 | /* |
1006 | * Make sure we update the FPM any time we send new information to | |
1007 | * the kernel. | |
1008 | */ | |
1009 | hook_call(rib_update, rn, "installing in kernel"); | |
1010 | ret = kernel_route_rib(p, src_p, old, re); | |
1011 | zvrf->installs++; | |
1012 | ||
1013 | /* If install succeeds, update FIB flag for nexthops. */ | |
1014 | if (!ret) { | |
1015 | for (ALL_NEXTHOPS(re->nexthop, nexthop)) { | |
1016 | if (CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_RECURSIVE)) | |
1017 | continue; | |
1018 | ||
1019 | if (CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE)) | |
1020 | SET_FLAG(nexthop->flags, NEXTHOP_FLAG_FIB); | |
1021 | else | |
1022 | UNSET_FLAG(nexthop->flags, NEXTHOP_FLAG_FIB); | |
1023 | } | |
1024 | } | |
3e5c6e00 | 1025 | |
d62a17ae | 1026 | return ret; |
718e3744 | 1027 | } |
1028 | ||
1029 | /* Uninstall the route from kernel. */ | |
d62a17ae | 1030 | int rib_uninstall_kernel(struct route_node *rn, struct route_entry *re) |
718e3744 | 1031 | { |
d62a17ae | 1032 | int ret = 0; |
1033 | struct nexthop *nexthop; | |
1034 | rib_table_info_t *info = srcdest_rnode_table_info(rn); | |
1035 | struct prefix *p, *src_p; | |
1036 | struct zebra_vrf *zvrf = vrf_info_lookup(re->vrf_id); | |
05737783 | 1037 | |
d62a17ae | 1038 | srcdest_rnode_prefixes(rn, &p, &src_p); |
718e3744 | 1039 | |
d62a17ae | 1040 | if (info->safi != SAFI_UNICAST) { |
1041 | for (ALL_NEXTHOPS(re->nexthop, nexthop)) | |
1042 | SET_FLAG(nexthop->flags, NEXTHOP_FLAG_FIB); | |
1043 | return ret; | |
1044 | } | |
416ec78d | 1045 | |
d62a17ae | 1046 | /* |
1047 | * Make sure we update the FPM any time we send new information to | |
1048 | * the kernel. | |
1049 | */ | |
1050 | hook_call(rib_update, rn, "uninstalling from kernel"); | |
1051 | ret = kernel_route_rib(p, src_p, re, NULL); | |
1052 | zvrf->removals++; | |
718e3744 | 1053 | |
d62a17ae | 1054 | for (ALL_NEXTHOPS(re->nexthop, nexthop)) |
1055 | UNSET_FLAG(nexthop->flags, NEXTHOP_FLAG_FIB); | |
718e3744 | 1056 | |
d62a17ae | 1057 | return ret; |
718e3744 | 1058 | } |
1059 | ||
1060 | /* Uninstall the route from kernel. */ | |
d62a17ae | 1061 | static void rib_uninstall(struct route_node *rn, struct route_entry *re) |
718e3744 | 1062 | { |
d62a17ae | 1063 | rib_table_info_t *info = srcdest_rnode_table_info(rn); |
416ec78d | 1064 | |
d62a17ae | 1065 | if (CHECK_FLAG(re->status, ROUTE_ENTRY_SELECTED_FIB)) { |
1066 | if (info->safi == SAFI_UNICAST) | |
1067 | hook_call(rib_update, rn, "rib_uninstall"); | |
5adc2528 | 1068 | |
d62a17ae | 1069 | if (!RIB_SYSTEM_ROUTE(re)) |
1070 | rib_uninstall_kernel(rn, re); | |
a64448ba | 1071 | |
d62a17ae | 1072 | /* If labeled-unicast route, uninstall transit LSP. */ |
1073 | if (zebra_rib_labeled_unicast(re)) | |
1074 | zebra_mpls_lsp_uninstall(info->zvrf, rn, re); | |
a64448ba | 1075 | |
d62a17ae | 1076 | UNSET_FLAG(re->status, ROUTE_ENTRY_SELECTED_FIB); |
1077 | } | |
446bb95e | 1078 | |
d62a17ae | 1079 | if (CHECK_FLAG(re->flags, ZEBRA_FLAG_SELECTED)) { |
1080 | struct prefix *p, *src_p; | |
1081 | srcdest_rnode_prefixes(rn, &p, &src_p); | |
05737783 | 1082 | |
d62a17ae | 1083 | redistribute_delete(p, src_p, re); |
1084 | UNSET_FLAG(re->flags, ZEBRA_FLAG_SELECTED); | |
1085 | } | |
718e3744 | 1086 | } |
1087 | ||
9fd92e3c AS |
1088 | /* |
1089 | * rib_can_delete_dest | |
1090 | * | |
1091 | * Returns TRUE if the given dest can be deleted from the table. | |
1092 | */ | |
d62a17ae | 1093 | static int rib_can_delete_dest(rib_dest_t *dest) |
9fd92e3c | 1094 | { |
d62a17ae | 1095 | if (dest->routes) { |
1096 | return 0; | |
1097 | } | |
9fd92e3c | 1098 | |
d62a17ae | 1099 | /* |
1100 | * Don't delete the dest if we have to update the FPM about this | |
1101 | * prefix. | |
1102 | */ | |
1103 | if (CHECK_FLAG(dest->flags, RIB_DEST_UPDATE_FPM) | |
1104 | || CHECK_FLAG(dest->flags, RIB_DEST_SENT_TO_FPM)) | |
1105 | return 0; | |
5adc2528 | 1106 | |
d62a17ae | 1107 | return 1; |
9fd92e3c AS |
1108 | } |
1109 | ||
1110 | /* | |
1111 | * rib_gc_dest | |
1112 | * | |
1113 | * Garbage collect the rib dest corresponding to the given route node | |
1114 | * if appropriate. | |
1115 | * | |
1116 | * Returns TRUE if the dest was deleted, FALSE otherwise. | |
1117 | */ | |
d62a17ae | 1118 | int rib_gc_dest(struct route_node *rn) |
9fd92e3c | 1119 | { |
d62a17ae | 1120 | rib_dest_t *dest; |
1121 | struct zebra_vrf *zvrf; | |
9fd92e3c | 1122 | |
d62a17ae | 1123 | dest = rib_dest_from_rnode(rn); |
1124 | if (!dest) | |
1125 | return 0; | |
9fd92e3c | 1126 | |
d62a17ae | 1127 | if (!rib_can_delete_dest(dest)) |
1128 | return 0; | |
9fd92e3c | 1129 | |
d62a17ae | 1130 | zvrf = rib_dest_vrf(dest); |
1131 | if (IS_ZEBRA_DEBUG_RIB) | |
1132 | rnode_debug(rn, zvrf_id(zvrf), "removing dest from table"); | |
9fd92e3c | 1133 | |
d62a17ae | 1134 | dest->rnode = NULL; |
1135 | XFREE(MTYPE_RIB_DEST, dest); | |
1136 | rn->info = NULL; | |
9fd92e3c | 1137 | |
d62a17ae | 1138 | /* |
1139 | * Release the one reference that we keep on the route node. | |
1140 | */ | |
1141 | route_unlock_node(rn); | |
1142 | return 1; | |
9fd92e3c AS |
1143 | } |
1144 | ||
d62a17ae | 1145 | static void rib_process_add_fib(struct zebra_vrf *zvrf, struct route_node *rn, |
1146 | struct route_entry *new) | |
3e5c6e00 | 1147 | { |
d62a17ae | 1148 | hook_call(rib_update, rn, "new route selected"); |
3e5c6e00 | 1149 | |
d62a17ae | 1150 | /* Update real nexthop. This may actually determine if nexthop is active |
1151 | * or not. */ | |
1152 | if (!nexthop_active_update(rn, new, 1)) { | |
1153 | UNSET_FLAG(new->status, ROUTE_ENTRY_CHANGED); | |
1154 | return; | |
1155 | } | |
3e5c6e00 | 1156 | |
d62a17ae | 1157 | SET_FLAG(new->status, ROUTE_ENTRY_SELECTED_FIB); |
1158 | if (IS_ZEBRA_DEBUG_RIB) { | |
1159 | char buf[SRCDEST2STR_BUFFER]; | |
1160 | srcdest_rnode2str(rn, buf, sizeof(buf)); | |
1161 | zlog_debug("%u:%s: Adding route rn %p, re %p (type %d)", | |
1162 | zvrf_id(zvrf), buf, rn, new, new->type); | |
1163 | } | |
3e5c6e00 | 1164 | |
d62a17ae | 1165 | /* If labeled-unicast route, install transit LSP. */ |
1166 | if (zebra_rib_labeled_unicast(new)) | |
1167 | zebra_mpls_lsp_install(zvrf, rn, new); | |
a64448ba | 1168 | |
d62a17ae | 1169 | if (!RIB_SYSTEM_ROUTE(new)) { |
1170 | if (rib_install_kernel(rn, new, NULL)) { | |
1171 | char buf[SRCDEST2STR_BUFFER]; | |
1172 | srcdest_rnode2str(rn, buf, sizeof(buf)); | |
1173 | zlog_warn("%u:%s: Route install failed", zvrf_id(zvrf), | |
1174 | buf); | |
1175 | } | |
1176 | } | |
3e5c6e00 | 1177 | |
d62a17ae | 1178 | UNSET_FLAG(new->status, ROUTE_ENTRY_CHANGED); |
3e5c6e00 | 1179 | } |
1180 | ||
d62a17ae | 1181 | static void rib_process_del_fib(struct zebra_vrf *zvrf, struct route_node *rn, |
1182 | struct route_entry *old) | |
3e5c6e00 | 1183 | { |
d62a17ae | 1184 | hook_call(rib_update, rn, "removing existing route"); |
3e5c6e00 | 1185 | |
d62a17ae | 1186 | /* Uninstall from kernel. */ |
1187 | if (IS_ZEBRA_DEBUG_RIB) { | |
1188 | char buf[SRCDEST2STR_BUFFER]; | |
1189 | srcdest_rnode2str(rn, buf, sizeof(buf)); | |
1190 | zlog_debug("%u:%s: Deleting route rn %p, re %p (type %d)", | |
1191 | zvrf_id(zvrf), buf, rn, old, old->type); | |
1192 | } | |
3e5c6e00 | 1193 | |
d62a17ae | 1194 | /* If labeled-unicast route, uninstall transit LSP. */ |
1195 | if (zebra_rib_labeled_unicast(old)) | |
1196 | zebra_mpls_lsp_uninstall(zvrf, rn, old); | |
1197 | ||
1198 | if (!RIB_SYSTEM_ROUTE(old)) | |
1199 | rib_uninstall_kernel(rn, old); | |
1200 | ||
1201 | UNSET_FLAG(old->status, ROUTE_ENTRY_SELECTED_FIB); | |
1202 | ||
1203 | /* Update nexthop for route, reset changed flag. */ | |
1204 | nexthop_active_update(rn, old, 1); | |
1205 | UNSET_FLAG(old->status, ROUTE_ENTRY_CHANGED); | |
1206 | } | |
1207 | ||
1208 | static void rib_process_update_fib(struct zebra_vrf *zvrf, | |
1209 | struct route_node *rn, | |
1210 | struct route_entry *old, | |
1211 | struct route_entry *new) | |
1212 | { | |
1213 | struct nexthop *nexthop = NULL; | |
1214 | int nh_active = 0; | |
1215 | int installed = 1; | |
1216 | ||
1217 | /* | |
1218 | * We have to install or update if a new route has been selected or | |
1219 | * something has changed. | |
1220 | */ | |
1221 | if (new != old || CHECK_FLAG(new->status, ROUTE_ENTRY_CHANGED)) { | |
1222 | hook_call(rib_update, rn, "updating existing route"); | |
1223 | ||
1224 | /* Update the nexthop; we could determine here that nexthop is | |
1225 | * inactive. */ | |
1226 | if (nexthop_active_update(rn, new, 1)) | |
1227 | nh_active = 1; | |
1228 | ||
1229 | /* If nexthop is active, install the selected route, if | |
1230 | * appropriate. If | |
1231 | * the install succeeds, cleanup flags for prior route, if | |
1232 | * different from | |
1233 | * newly selected. | |
1234 | */ | |
1235 | if (nh_active) { | |
1236 | if (IS_ZEBRA_DEBUG_RIB) { | |
1237 | char buf[SRCDEST2STR_BUFFER]; | |
1238 | srcdest_rnode2str(rn, buf, sizeof(buf)); | |
1239 | if (new != old) | |
1240 | zlog_debug( | |
1241 | "%u:%s: Updating route rn %p, re %p (type %d) " | |
1242 | "old %p (type %d)", | |
1243 | zvrf_id(zvrf), buf, rn, new, | |
1244 | new->type, old, old->type); | |
1245 | else | |
1246 | zlog_debug( | |
1247 | "%u:%s: Updating route rn %p, re %p (type %d)", | |
1248 | zvrf_id(zvrf), buf, rn, new, | |
1249 | new->type); | |
1250 | } | |
1251 | ||
1252 | /* If labeled-unicast route, uninstall transit LSP. */ | |
1253 | if (zebra_rib_labeled_unicast(old)) | |
1254 | zebra_mpls_lsp_uninstall(zvrf, rn, old); | |
1255 | ||
1256 | /* Non-system route should be installed. */ | |
1257 | if (!RIB_SYSTEM_ROUTE(new)) { | |
1258 | /* If labeled-unicast route, install transit | |
1259 | * LSP. */ | |
1260 | if (zebra_rib_labeled_unicast(new)) | |
1261 | zebra_mpls_lsp_install(zvrf, rn, new); | |
1262 | ||
1263 | if (rib_install_kernel(rn, new, old)) { | |
1264 | char buf[SRCDEST2STR_BUFFER]; | |
1265 | srcdest_rnode2str(rn, buf, sizeof(buf)); | |
1266 | installed = 0; | |
1267 | zlog_warn("%u:%s: Route install failed", | |
1268 | zvrf_id(zvrf), buf); | |
1269 | } | |
1270 | } | |
1271 | ||
1272 | /* If install succeeded or system route, cleanup flags | |
1273 | * for prior route. */ | |
1274 | if (installed && new != old) { | |
1275 | if (RIB_SYSTEM_ROUTE(new)) { | |
1276 | if (!RIB_SYSTEM_ROUTE(old)) | |
1277 | rib_uninstall_kernel(rn, old); | |
1278 | } else { | |
1279 | for (nexthop = old->nexthop; nexthop; | |
1280 | nexthop = nexthop->next) | |
1281 | UNSET_FLAG(nexthop->flags, | |
1282 | NEXTHOP_FLAG_FIB); | |
1283 | } | |
1284 | } | |
1285 | ||
1286 | /* Update for redistribution. */ | |
1287 | if (installed) | |
1288 | SET_FLAG(new->status, ROUTE_ENTRY_SELECTED_FIB); | |
1289 | } | |
a64448ba | 1290 | |
d62a17ae | 1291 | /* |
1292 | * If nexthop for selected route is not active or install | |
1293 | * failed, we | |
1294 | * may need to uninstall and delete for redistribution. | |
1295 | */ | |
1296 | if (!nh_active || !installed) { | |
1297 | if (IS_ZEBRA_DEBUG_RIB) { | |
1298 | char buf[SRCDEST2STR_BUFFER]; | |
1299 | srcdest_rnode2str(rn, buf, sizeof(buf)); | |
1300 | if (new != old) | |
1301 | zlog_debug( | |
1302 | "%u:%s: Deleting route rn %p, re %p (type %d) " | |
1303 | "old %p (type %d) - %s", | |
1304 | zvrf_id(zvrf), buf, rn, new, | |
1305 | new->type, old, old->type, | |
1306 | nh_active ? "install failed" | |
1307 | : "nexthop inactive"); | |
1308 | else | |
1309 | zlog_debug( | |
1310 | "%u:%s: Deleting route rn %p, re %p (type %d) - %s", | |
1311 | zvrf_id(zvrf), buf, rn, new, | |
1312 | new->type, | |
1313 | nh_active ? "install failed" | |
1314 | : "nexthop inactive"); | |
1315 | } | |
1316 | ||
1317 | /* If labeled-unicast route, uninstall transit LSP. */ | |
1318 | if (zebra_rib_labeled_unicast(old)) | |
1319 | zebra_mpls_lsp_uninstall(zvrf, rn, old); | |
1320 | ||
1321 | if (!RIB_SYSTEM_ROUTE(old)) | |
1322 | rib_uninstall_kernel(rn, old); | |
1323 | UNSET_FLAG(new->status, ROUTE_ENTRY_SELECTED_FIB); | |
1324 | } | |
1325 | } else { | |
1326 | /* | |
1327 | * Same route selected; check if in the FIB and if not, | |
1328 | * re-install. This | |
1329 | * is housekeeping code to deal with race conditions in kernel | |
1330 | * with linux | |
1331 | * netlink reporting interface up before IPv4 or IPv6 protocol | |
1332 | * is ready | |
1333 | * to add routes. | |
1334 | */ | |
1335 | if (!RIB_SYSTEM_ROUTE(new)) { | |
1336 | int in_fib = 0; | |
1337 | ||
1338 | for (ALL_NEXTHOPS(new->nexthop, nexthop)) | |
1339 | if (CHECK_FLAG(nexthop->flags, | |
1340 | NEXTHOP_FLAG_FIB)) { | |
1341 | in_fib = 1; | |
1342 | break; | |
1343 | } | |
1344 | if (!in_fib) | |
1345 | rib_install_kernel(rn, new, NULL); | |
1346 | } | |
1347 | } | |
1348 | ||
1349 | /* Update prior route. */ | |
1350 | if (new != old) { | |
1351 | UNSET_FLAG(old->status, ROUTE_ENTRY_SELECTED_FIB); | |
3e5c6e00 | 1352 | |
d62a17ae | 1353 | /* Set real nexthop. */ |
1354 | nexthop_active_update(rn, old, 1); | |
1355 | UNSET_FLAG(old->status, ROUTE_ENTRY_CHANGED); | |
1356 | } | |
3e5c6e00 | 1357 | |
d62a17ae | 1358 | /* Clear changed flag. */ |
1359 | UNSET_FLAG(new->status, ROUTE_ENTRY_CHANGED); | |
3e5c6e00 | 1360 | } |
1361 | ||
d62a17ae | 1362 | /* Check if 'alternate' RIB entry is better than 'current'. */ |
1363 | static struct route_entry *rib_choose_best(struct route_entry *current, | |
1364 | struct route_entry *alternate) | |
1365 | { | |
1366 | if (current == NULL) | |
1367 | return alternate; | |
1368 | ||
1369 | /* filter route selection in following order: | |
1370 | * - connected beats other types | |
1371 | * - lower distance beats higher | |
1372 | * - lower metric beats higher for equal distance | |
1373 | * - last, hence oldest, route wins tie break. | |
1374 | */ | |
1375 | ||
1376 | /* Connected routes. Pick the last connected | |
1377 | * route of the set of lowest metric connected routes. | |
1378 | */ | |
1379 | if (alternate->type == ZEBRA_ROUTE_CONNECT) { | |
1380 | if (current->type != ZEBRA_ROUTE_CONNECT | |
1381 | || alternate->metric <= current->metric) | |
1382 | return alternate; | |
1383 | ||
1384 | return current; | |
1385 | } | |
3e5c6e00 | 1386 | |
d62a17ae | 1387 | if (current->type == ZEBRA_ROUTE_CONNECT) |
1388 | return current; | |
3e5c6e00 | 1389 | |
d62a17ae | 1390 | /* higher distance loses */ |
1391 | if (alternate->distance < current->distance) | |
1392 | return alternate; | |
1393 | if (current->distance < alternate->distance) | |
1394 | return current; | |
3e5c6e00 | 1395 | |
d62a17ae | 1396 | /* metric tie-breaks equal distance */ |
1397 | if (alternate->metric <= current->metric) | |
1398 | return alternate; | |
3e5c6e00 | 1399 | |
d62a17ae | 1400 | return current; |
3e5c6e00 | 1401 | } |
1402 | ||
d62a17ae | 1403 | /* Core function for processing routing information base. */ |
1404 | static void rib_process(struct route_node *rn) | |
1405 | { | |
1406 | struct route_entry *re; | |
1407 | struct route_entry *next; | |
1408 | struct route_entry *old_selected = NULL; | |
1409 | struct route_entry *new_selected = NULL; | |
1410 | struct route_entry *old_fib = NULL; | |
1411 | struct route_entry *new_fib = NULL; | |
1412 | struct route_entry *best = NULL; | |
1413 | char buf[SRCDEST2STR_BUFFER]; | |
1414 | rib_dest_t *dest; | |
1415 | struct zebra_vrf *zvrf = NULL; | |
1416 | struct prefix *p, *src_p; | |
1417 | srcdest_rnode_prefixes(rn, &p, &src_p); | |
1418 | vrf_id_t vrf_id = VRF_UNKNOWN; | |
1419 | ||
1420 | assert(rn); | |
1421 | ||
1422 | dest = rib_dest_from_rnode(rn); | |
1423 | if (dest) { | |
1424 | zvrf = rib_dest_vrf(dest); | |
1425 | vrf_id = zvrf_id(zvrf); | |
1426 | } | |
bab85d4f | 1427 | |
d62a17ae | 1428 | if (IS_ZEBRA_DEBUG_RIB) |
1429 | srcdest_rnode2str(rn, buf, sizeof(buf)); | |
bab85d4f | 1430 | |
d62a17ae | 1431 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) |
1432 | zlog_debug("%u:%s: Processing rn %p", vrf_id, buf, rn); | |
bab85d4f | 1433 | |
d62a17ae | 1434 | RNODE_FOREACH_RE_SAFE(rn, re, next) |
1435 | { | |
1436 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) | |
1437 | zlog_debug( | |
1438 | "%u:%s: Examine re %p (type %d) status %x flags %x " | |
1439 | "dist %d metric %d", | |
1440 | vrf_id, buf, re, re->type, re->status, | |
1441 | re->flags, re->distance, re->metric); | |
1442 | ||
1443 | UNSET_FLAG(re->status, ROUTE_ENTRY_NEXTHOPS_CHANGED); | |
1444 | ||
1445 | /* Currently selected re. */ | |
1446 | if (CHECK_FLAG(re->flags, ZEBRA_FLAG_SELECTED)) { | |
1447 | assert(old_selected == NULL); | |
1448 | old_selected = re; | |
1449 | } | |
1450 | /* Currently in fib */ | |
1451 | if (CHECK_FLAG(re->status, ROUTE_ENTRY_SELECTED_FIB)) { | |
1452 | assert(old_fib == NULL); | |
1453 | old_fib = re; | |
1454 | } | |
bab85d4f | 1455 | |
d62a17ae | 1456 | /* Skip deleted entries from selection */ |
1457 | if (CHECK_FLAG(re->status, ROUTE_ENTRY_REMOVED)) | |
1458 | continue; | |
1459 | ||
1460 | /* Skip unreachable nexthop. */ | |
1461 | /* This first call to nexthop_active_update is merely to | |
1462 | * determine if | |
1463 | * there's any change to nexthops associated with this RIB | |
1464 | * entry. Now, | |
1465 | * rib_process() can be invoked due to an external event such as | |
1466 | * link | |
1467 | * down or due to next-hop-tracking evaluation. In the latter | |
1468 | * case, | |
1469 | * a decision has already been made that the NHs have changed. | |
1470 | * So, no | |
1471 | * need to invoke a potentially expensive call again. Further, | |
1472 | * since | |
1473 | * the change might be in a recursive NH which is not caught in | |
1474 | * the nexthop_active_update() code. Thus, we might miss changes | |
1475 | * to | |
1476 | * recursive NHs. | |
1477 | */ | |
1478 | if (!CHECK_FLAG(re->status, ROUTE_ENTRY_CHANGED) | |
1479 | && !nexthop_active_update(rn, re, 0)) { | |
1480 | if (re->type == ZEBRA_ROUTE_TABLE) { | |
1481 | /* XXX: HERE BE DRAGONS!!!!! | |
1482 | * In all honesty, I have not yet figured out | |
1483 | * what this part | |
1484 | * does or why the ROUTE_ENTRY_CHANGED test | |
1485 | * above is correct | |
1486 | * or why we need to delete a route here, and | |
1487 | * also not whether | |
1488 | * this concerns both selected and fib route, or | |
1489 | * only selected | |
1490 | * or only fib */ | |
1491 | /* This entry was denied by the 'ip protocol | |
1492 | * table' route-map, we | |
1493 | * need to delete it */ | |
1494 | if (re != old_selected) { | |
1495 | if (IS_ZEBRA_DEBUG_RIB) | |
1496 | zlog_debug( | |
1497 | "%s: %s: imported via import-table but denied " | |
1498 | "by the ip protocol table route-map", | |
1499 | __func__, buf); | |
1500 | rib_unlink(rn, re); | |
1501 | } else | |
1502 | SET_FLAG(re->status, | |
1503 | ROUTE_ENTRY_REMOVED); | |
1504 | } | |
1505 | ||
1506 | continue; | |
1507 | } | |
bab85d4f | 1508 | |
d62a17ae | 1509 | /* Infinite distance. */ |
1510 | if (re->distance == DISTANCE_INFINITY) { | |
1511 | UNSET_FLAG(re->status, ROUTE_ENTRY_CHANGED); | |
1512 | continue; | |
1513 | } | |
bab85d4f | 1514 | |
d62a17ae | 1515 | if (CHECK_FLAG(re->flags, ZEBRA_FLAG_FIB_OVERRIDE)) { |
1516 | best = rib_choose_best(new_fib, re); | |
1517 | if (new_fib && best != new_fib) | |
1518 | UNSET_FLAG(new_fib->status, | |
1519 | ROUTE_ENTRY_CHANGED); | |
1520 | new_fib = best; | |
1521 | } else { | |
1522 | best = rib_choose_best(new_selected, re); | |
1523 | if (new_selected && best != new_selected) | |
1524 | UNSET_FLAG(new_selected->status, | |
1525 | ROUTE_ENTRY_CHANGED); | |
1526 | new_selected = best; | |
1527 | } | |
1528 | if (best != re) | |
1529 | UNSET_FLAG(re->status, ROUTE_ENTRY_CHANGED); | |
1530 | } /* RNODE_FOREACH_RE */ | |
1531 | ||
1532 | /* If no FIB override route, use the selected route also for FIB */ | |
1533 | if (new_fib == NULL) | |
1534 | new_fib = new_selected; | |
1535 | ||
1536 | /* After the cycle is finished, the following pointers will be set: | |
1537 | * old_selected --- RE entry currently having SELECTED | |
1538 | * new_selected --- RE entry that is newly SELECTED | |
1539 | * old_fib --- RE entry currently in kernel FIB | |
1540 | * new_fib --- RE entry that is newly to be in kernel FIB | |
1541 | * | |
1542 | * new_selected will get SELECTED flag, and is going to be redistributed | |
1543 | * the zclients. new_fib (which can be new_selected) will be installed | |
1544 | * in kernel. | |
1545 | */ | |
1546 | ||
1547 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) { | |
1548 | zlog_debug( | |
1549 | "%u:%s: After processing: old_selected %p new_selected %p old_fib %p new_fib %p", | |
1550 | vrf_id, buf, (void *)old_selected, (void *)new_selected, | |
1551 | (void *)old_fib, (void *)new_fib); | |
1552 | } | |
446bb95e | 1553 | |
d62a17ae | 1554 | /* Buffer ROUTE_ENTRY_CHANGED here, because it will get cleared if |
1555 | * fib == selected */ | |
9d303b37 DL |
1556 | bool selected_changed = new_selected && CHECK_FLAG(new_selected->status, |
1557 | ROUTE_ENTRY_CHANGED); | |
d62a17ae | 1558 | |
1559 | /* Update fib according to selection results */ | |
1560 | if (new_fib && old_fib) | |
1561 | rib_process_update_fib(zvrf, rn, old_fib, new_fib); | |
1562 | else if (new_fib) | |
1563 | rib_process_add_fib(zvrf, rn, new_fib); | |
1564 | else if (old_fib) | |
1565 | rib_process_del_fib(zvrf, rn, old_fib); | |
1566 | ||
1567 | /* Redistribute SELECTED entry */ | |
1568 | if (old_selected != new_selected || selected_changed) { | |
1569 | struct nexthop *nexthop; | |
1570 | ||
1571 | /* Check if we have a FIB route for the destination, otherwise, | |
1572 | * don't redistribute it */ | |
1573 | for (ALL_NEXTHOPS(new_fib ? new_fib->nexthop : NULL, nexthop)) { | |
1574 | if (CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_FIB)) { | |
1575 | break; | |
1576 | } | |
1577 | } | |
1578 | if (!nexthop) | |
1579 | new_selected = NULL; | |
93bdadae | 1580 | |
d62a17ae | 1581 | if (new_selected && new_selected != new_fib) { |
1582 | nexthop_active_update(rn, new_selected, 1); | |
1583 | UNSET_FLAG(new_selected->status, ROUTE_ENTRY_CHANGED); | |
1584 | } | |
41ec9222 | 1585 | |
d62a17ae | 1586 | if (old_selected) { |
1587 | if (!new_selected) | |
1588 | redistribute_delete(p, src_p, old_selected); | |
1589 | if (old_selected != new_selected) | |
1590 | UNSET_FLAG(old_selected->flags, | |
1591 | ZEBRA_FLAG_SELECTED); | |
f857321e | 1592 | } |
446bb95e | 1593 | |
d62a17ae | 1594 | if (new_selected) { |
1595 | /* Install new or replace existing redistributed entry | |
1596 | */ | |
1597 | SET_FLAG(new_selected->flags, ZEBRA_FLAG_SELECTED); | |
1598 | redistribute_update(p, src_p, new_selected, | |
1599 | old_selected); | |
1600 | } | |
1601 | } | |
3e5c6e00 | 1602 | |
d62a17ae | 1603 | /* Remove all RE entries queued for removal */ |
1604 | RNODE_FOREACH_RE_SAFE(rn, re, next) | |
1605 | { | |
1606 | if (CHECK_FLAG(re->status, ROUTE_ENTRY_REMOVED)) { | |
1607 | if (IS_ZEBRA_DEBUG_RIB) { | |
1608 | rnode_debug(rn, vrf_id, "rn %p, removing re %p", | |
1609 | (void *)rn, (void *)re); | |
1610 | } | |
1611 | rib_unlink(rn, re); | |
1612 | } | |
1613 | } | |
4d38fdb4 | 1614 | |
d62a17ae | 1615 | /* |
1616 | * Check if the dest can be deleted now. | |
1617 | */ | |
1618 | rib_gc_dest(rn); | |
e96f9203 DO |
1619 | } |
1620 | ||
5110a0c6 | 1621 | /* Take a list of route_node structs and return 1, if there was a record |
d62a17ae | 1622 | * picked from it and processed by rib_process(). Don't process more, |
5110a0c6 | 1623 | * than one RN record; operate only in the specified sub-queue. |
e96f9203 | 1624 | */ |
d62a17ae | 1625 | static unsigned int process_subq(struct list *subq, u_char qindex) |
e96f9203 | 1626 | { |
d62a17ae | 1627 | struct listnode *lnode = listhead(subq); |
1628 | struct route_node *rnode; | |
1629 | rib_dest_t *dest; | |
1630 | struct zebra_vrf *zvrf = NULL; | |
5110a0c6 | 1631 | |
d62a17ae | 1632 | if (!lnode) |
1633 | return 0; | |
5110a0c6 | 1634 | |
d62a17ae | 1635 | rnode = listgetdata(lnode); |
1636 | dest = rib_dest_from_rnode(rnode); | |
1637 | if (dest) | |
1638 | zvrf = rib_dest_vrf(dest); | |
41ec9222 | 1639 | |
d62a17ae | 1640 | rib_process(rnode); |
5110a0c6 | 1641 | |
d62a17ae | 1642 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) { |
1643 | char buf[SRCDEST2STR_BUFFER]; | |
1644 | srcdest_rnode2str(rnode, buf, sizeof(buf)); | |
1645 | zlog_debug("%u:%s: rn %p dequeued from sub-queue %u", | |
1646 | zvrf ? zvrf_id(zvrf) : 0, buf, rnode, qindex); | |
1647 | } | |
41ec9222 | 1648 | |
d62a17ae | 1649 | if (rnode->info) |
1650 | UNSET_FLAG(rib_dest_from_rnode(rnode)->flags, | |
1651 | RIB_ROUTE_QUEUED(qindex)); | |
9fd92e3c | 1652 | |
67b9467f | 1653 | #if 0 |
5110a0c6 SH |
1654 | else |
1655 | { | |
1656 | zlog_debug ("%s: called for route_node (%p, %d) with no ribs", | |
1657 | __func__, rnode, rnode->lock); | |
1658 | zlog_backtrace(LOG_DEBUG); | |
1659 | } | |
67b9467f | 1660 | #endif |
d62a17ae | 1661 | route_unlock_node(rnode); |
1662 | list_delete_node(subq, lnode); | |
1663 | return 1; | |
e96f9203 DO |
1664 | } |
1665 | ||
fb018d25 DS |
1666 | /* |
1667 | * All meta queues have been processed. Trigger next-hop evaluation. | |
1668 | */ | |
d62a17ae | 1669 | static void meta_queue_process_complete(struct work_queue *dummy) |
fb018d25 | 1670 | { |
d62a17ae | 1671 | struct vrf *vrf; |
1672 | struct zebra_vrf *zvrf; | |
9ec6b0bb | 1673 | |
d62a17ae | 1674 | /* Evaluate nexthops for those VRFs which underwent route processing. |
1675 | * This | |
1676 | * should limit the evaluation to the necessary VRFs in most common | |
1677 | * situations. | |
1678 | */ | |
1679 | RB_FOREACH(vrf, vrf_id_head, &vrfs_by_id) | |
1680 | { | |
1681 | zvrf = vrf->info; | |
1682 | if (zvrf == NULL || !(zvrf->flags & ZEBRA_VRF_RIB_SCHEDULED)) | |
1683 | continue; | |
1684 | ||
1685 | zvrf->flags &= ~ZEBRA_VRF_RIB_SCHEDULED; | |
1686 | zebra_evaluate_rnh(zvrf_id(zvrf), AF_INET, 0, RNH_NEXTHOP_TYPE, | |
1687 | NULL); | |
1688 | zebra_evaluate_rnh(zvrf_id(zvrf), AF_INET, 0, | |
1689 | RNH_IMPORT_CHECK_TYPE, NULL); | |
1690 | zebra_evaluate_rnh(zvrf_id(zvrf), AF_INET6, 0, RNH_NEXTHOP_TYPE, | |
1691 | NULL); | |
1692 | zebra_evaluate_rnh(zvrf_id(zvrf), AF_INET6, 0, | |
1693 | RNH_IMPORT_CHECK_TYPE, NULL); | |
1694 | } | |
939fba27 | 1695 | |
d62a17ae | 1696 | /* Schedule LSPs for processing, if needed. */ |
1697 | zvrf = vrf_info_lookup(VRF_DEFAULT); | |
1698 | if (mpls_should_lsps_be_processed(zvrf)) { | |
1699 | if (IS_ZEBRA_DEBUG_MPLS) | |
1700 | zlog_debug( | |
1701 | "%u: Scheduling all LSPs upon RIB completion", | |
1702 | zvrf_id(zvrf)); | |
1703 | zebra_mpls_lsp_schedule(zvrf); | |
1704 | mpls_unmark_lsps_for_processing(zvrf); | |
1705 | } | |
fb018d25 DS |
1706 | } |
1707 | ||
e96f9203 | 1708 | /* Dispatch the meta queue by picking, processing and unlocking the next RN from |
d62a17ae | 1709 | * a non-empty sub-queue with lowest priority. wq is equal to zebra->ribq and |
1710 | * data | |
e96f9203 DO |
1711 | * is pointed to the meta queue structure. |
1712 | */ | |
d62a17ae | 1713 | static wq_item_status meta_queue_process(struct work_queue *dummy, void *data) |
e96f9203 | 1714 | { |
d62a17ae | 1715 | struct meta_queue *mq = data; |
1716 | unsigned i; | |
5110a0c6 | 1717 | |
d62a17ae | 1718 | for (i = 0; i < MQ_SIZE; i++) |
1719 | if (process_subq(mq->subq[i], i)) { | |
1720 | mq->size--; | |
1721 | break; | |
1722 | } | |
1723 | return mq->size ? WQ_REQUEUE : WQ_SUCCESS; | |
e96f9203 DO |
1724 | } |
1725 | ||
9fd92e3c AS |
1726 | /* |
1727 | * Map from rib types to queue type (priority) in meta queue | |
1728 | */ | |
5110a0c6 | 1729 | static const u_char meta_queue_map[ZEBRA_ROUTE_MAX] = { |
9d303b37 DL |
1730 | [ZEBRA_ROUTE_SYSTEM] = 4, |
1731 | [ZEBRA_ROUTE_KERNEL] = 0, | |
1732 | [ZEBRA_ROUTE_CONNECT] = 0, | |
1733 | [ZEBRA_ROUTE_STATIC] = 1, | |
1734 | [ZEBRA_ROUTE_RIP] = 2, | |
1735 | [ZEBRA_ROUTE_RIPNG] = 2, | |
1736 | [ZEBRA_ROUTE_OSPF] = 2, | |
1737 | [ZEBRA_ROUTE_OSPF6] = 2, | |
1738 | [ZEBRA_ROUTE_ISIS] = 2, | |
1739 | [ZEBRA_ROUTE_BGP] = 3, | |
1740 | [ZEBRA_ROUTE_PIM] = 4, // Shouldn't happen but for safety | |
1741 | [ZEBRA_ROUTE_EIGRP] = 2, | |
1742 | [ZEBRA_ROUTE_NHRP] = 2, | |
1743 | [ZEBRA_ROUTE_HSLS] = 4, | |
1744 | [ZEBRA_ROUTE_OLSR] = 4, | |
1745 | [ZEBRA_ROUTE_TABLE] = 1, | |
1746 | [ZEBRA_ROUTE_LDP] = 4, | |
1747 | [ZEBRA_ROUTE_VNC] = 3, | |
1748 | [ZEBRA_ROUTE_VNC_DIRECT] = 3, | |
1749 | [ZEBRA_ROUTE_VNC_DIRECT_RH] = 3, | |
1750 | [ZEBRA_ROUTE_BGP_DIRECT] = 3, | |
1751 | [ZEBRA_ROUTE_BGP_DIRECT_EXT] = 3, | |
1752 | [ZEBRA_ROUTE_BABEL] = 2, | |
1753 | [ZEBRA_ROUTE_ALL] = 4, // Shouldn't happen but for safety | |
5110a0c6 SH |
1754 | }; |
1755 | ||
1756 | /* Look into the RN and queue it into one or more priority queues, | |
1757 | * increasing the size for each data push done. | |
e96f9203 | 1758 | */ |
d62a17ae | 1759 | static void rib_meta_queue_add(struct meta_queue *mq, struct route_node *rn) |
e96f9203 | 1760 | { |
d62a17ae | 1761 | struct route_entry *re; |
5110a0c6 | 1762 | |
d62a17ae | 1763 | RNODE_FOREACH_RE(rn, re) |
2263a412 | 1764 | { |
d62a17ae | 1765 | u_char qindex = meta_queue_map[re->type]; |
1766 | struct zebra_vrf *zvrf; | |
1767 | ||
1768 | /* Invariant: at this point we always have rn->info set. */ | |
1769 | if (CHECK_FLAG(rib_dest_from_rnode(rn)->flags, | |
1770 | RIB_ROUTE_QUEUED(qindex))) { | |
1771 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) | |
1772 | rnode_debug( | |
1773 | rn, re->vrf_id, | |
1774 | "rn %p is already queued in sub-queue %u", | |
1775 | (void *)rn, qindex); | |
1776 | continue; | |
1777 | } | |
5110a0c6 | 1778 | |
d62a17ae | 1779 | SET_FLAG(rib_dest_from_rnode(rn)->flags, |
1780 | RIB_ROUTE_QUEUED(qindex)); | |
1781 | listnode_add(mq->subq[qindex], rn); | |
1782 | route_lock_node(rn); | |
1783 | mq->size++; | |
5110a0c6 | 1784 | |
d62a17ae | 1785 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) |
1786 | rnode_debug(rn, re->vrf_id, | |
1787 | "queued rn %p into sub-queue %u", | |
1788 | (void *)rn, qindex); | |
9ec6b0bb | 1789 | |
d62a17ae | 1790 | zvrf = zebra_vrf_lookup_by_id(re->vrf_id); |
1791 | if (zvrf) | |
1792 | zvrf->flags |= ZEBRA_VRF_RIB_SCHEDULED; | |
1793 | } | |
4d38fdb4 | 1794 | } |
1795 | ||
6d691129 | 1796 | /* Add route_node to work queue and schedule processing */ |
d62a17ae | 1797 | void rib_queue_add(struct route_node *rn) |
4d38fdb4 | 1798 | { |
d62a17ae | 1799 | assert(rn); |
fc328ac9 | 1800 | |
d62a17ae | 1801 | /* Pointless to queue a route_node with no RIB entries to add or remove |
1802 | */ | |
1803 | if (!rnode_to_ribs(rn)) { | |
1804 | zlog_debug("%s: called for route_node (%p, %d) with no ribs", | |
1805 | __func__, (void *)rn, rn->lock); | |
1806 | zlog_backtrace(LOG_DEBUG); | |
1807 | return; | |
1808 | } | |
4d38fdb4 | 1809 | |
d62a17ae | 1810 | if (zebrad.ribq == NULL) { |
1811 | zlog_err("%s: work_queue does not exist!", __func__); | |
1812 | return; | |
1813 | } | |
1814 | ||
1815 | /* | |
1816 | * The RIB queue should normally be either empty or holding the only | |
1817 | * work_queue_item element. In the latter case this element would | |
1818 | * hold a pointer to the meta queue structure, which must be used to | |
1819 | * actually queue the route nodes to process. So create the MQ | |
1820 | * holder, if necessary, then push the work into it in any case. | |
1821 | * This semantics was introduced after 0.99.9 release. | |
1822 | */ | |
f104f6c1 | 1823 | if (work_queue_empty(zebrad.ribq)) |
d62a17ae | 1824 | work_queue_add(zebrad.ribq, zebrad.mq); |
e96f9203 | 1825 | |
d62a17ae | 1826 | rib_meta_queue_add(zebrad.mq, rn); |
fc328ac9 | 1827 | |
d62a17ae | 1828 | return; |
4d38fdb4 | 1829 | } |
1830 | ||
5110a0c6 SH |
1831 | /* Create new meta queue. |
1832 | A destructor function doesn't seem to be necessary here. | |
1833 | */ | |
d62a17ae | 1834 | static struct meta_queue *meta_queue_new(void) |
e96f9203 | 1835 | { |
d62a17ae | 1836 | struct meta_queue *new; |
1837 | unsigned i; | |
5110a0c6 | 1838 | |
d62a17ae | 1839 | new = XCALLOC(MTYPE_WORK_QUEUE, sizeof(struct meta_queue)); |
1840 | assert(new); | |
e96f9203 | 1841 | |
d62a17ae | 1842 | for (i = 0; i < MQ_SIZE; i++) { |
1843 | new->subq[i] = list_new(); | |
1844 | assert(new->subq[i]); | |
1845 | } | |
5110a0c6 | 1846 | |
d62a17ae | 1847 | return new; |
e96f9203 DO |
1848 | } |
1849 | ||
d62a17ae | 1850 | void meta_queue_free(struct meta_queue *mq) |
5a8dfcd8 | 1851 | { |
d62a17ae | 1852 | unsigned i; |
5a8dfcd8 | 1853 | |
d62a17ae | 1854 | for (i = 0; i < MQ_SIZE; i++) |
1855 | list_delete(mq->subq[i]); | |
5a8dfcd8 | 1856 | |
d62a17ae | 1857 | XFREE(MTYPE_WORK_QUEUE, mq); |
5a8dfcd8 RW |
1858 | } |
1859 | ||
4d38fdb4 | 1860 | /* initialise zebra rib work queue */ |
d62a17ae | 1861 | static void rib_queue_init(struct zebra_t *zebra) |
4d38fdb4 | 1862 | { |
d62a17ae | 1863 | assert(zebra); |
1864 | ||
1865 | if (!(zebra->ribq = | |
1866 | work_queue_new(zebra->master, "route_node processing"))) { | |
1867 | zlog_err("%s: could not initialise work queue!", __func__); | |
1868 | return; | |
1869 | } | |
4d38fdb4 | 1870 | |
d62a17ae | 1871 | /* fill in the work queue spec */ |
1872 | zebra->ribq->spec.workfunc = &meta_queue_process; | |
1873 | zebra->ribq->spec.errorfunc = NULL; | |
1874 | zebra->ribq->spec.completion_func = &meta_queue_process_complete; | |
1875 | /* XXX: TODO: These should be runtime configurable via vty */ | |
1876 | zebra->ribq->spec.max_retries = 3; | |
1877 | zebra->ribq->spec.hold = rib_process_hold_time; | |
1878 | ||
1879 | if (!(zebra->mq = meta_queue_new())) { | |
1880 | zlog_err("%s: could not initialise meta queue!", __func__); | |
1881 | return; | |
1882 | } | |
1883 | return; | |
718e3744 | 1884 | } |
1885 | ||
6d691129 PJ |
1886 | /* RIB updates are processed via a queue of pointers to route_nodes. |
1887 | * | |
1888 | * The queue length is bounded by the maximal size of the routing table, | |
1889 | * as a route_node will not be requeued, if already queued. | |
1890 | * | |
f0f77c9a DS |
1891 | * REs are submitted via rib_addnode or rib_delnode which set minimal |
1892 | * state, or static_install_route (when an existing RE is updated) | |
3c0755dc | 1893 | * and then submit route_node to queue for best-path selection later. |
f0f77c9a | 1894 | * Order of add/delete state changes are preserved for any given RE. |
6d691129 | 1895 | * |
f0f77c9a | 1896 | * Deleted REs are reaped during best-path selection. |
6d691129 PJ |
1897 | * |
1898 | * rib_addnode | |
f0f77c9a DS |
1899 | * |-> rib_link or unset ROUTE_ENTRY_REMOVE |->Update kernel with |
1900 | * |-------->| | best RE, if required | |
3c0755dc PJ |
1901 | * | | |
1902 | * static_install->|->rib_addqueue...... -> rib_process | |
1903 | * | | | |
1904 | * |-------->| |-> rib_unlink | |
f0f77c9a DS |
1905 | * |-> set ROUTE_ENTRY_REMOVE | |
1906 | * rib_delnode (RE freed) | |
6d691129 | 1907 | * |
9fd92e3c AS |
1908 | * The 'info' pointer of a route_node points to a rib_dest_t |
1909 | * ('dest'). Queueing state for a route_node is kept on the dest. The | |
1910 | * dest is created on-demand by rib_link() and is kept around at least | |
1911 | * as long as there are ribs hanging off it (@see rib_gc_dest()). | |
d62a17ae | 1912 | * |
6d691129 PJ |
1913 | * Refcounting (aka "locking" throughout the GNU Zebra and Quagga code): |
1914 | * | |
1915 | * - route_nodes: refcounted by: | |
9fd92e3c AS |
1916 | * - dest attached to route_node: |
1917 | * - managed by: rib_link/rib_gc_dest | |
6d691129 PJ |
1918 | * - route_node processing queue |
1919 | * - managed by: rib_addqueue, rib_process. | |
1920 | * | |
1921 | */ | |
d62a17ae | 1922 | |
f0f77c9a | 1923 | /* Add RE to head of the route node. */ |
d62a17ae | 1924 | static void rib_link(struct route_node *rn, struct route_entry *re, int process) |
1925 | { | |
1926 | struct route_entry *head; | |
1927 | rib_dest_t *dest; | |
1928 | afi_t afi; | |
1929 | const char *rmap_name; | |
9fd92e3c | 1930 | |
d62a17ae | 1931 | assert(re && rn); |
9fd92e3c | 1932 | |
d62a17ae | 1933 | dest = rib_dest_from_rnode(rn); |
1934 | if (!dest) { | |
1935 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) | |
1936 | rnode_debug(rn, re->vrf_id, "rn %p adding dest", rn); | |
7a4bb9c5 | 1937 | |
d62a17ae | 1938 | dest = XCALLOC(MTYPE_RIB_DEST, sizeof(rib_dest_t)); |
1939 | route_lock_node(rn); /* rn route table reference */ | |
1940 | rn->info = dest; | |
1941 | dest->rnode = rn; | |
1942 | } | |
2263a412 | 1943 | |
d62a17ae | 1944 | head = dest->routes; |
1945 | if (head) { | |
1946 | head->prev = re; | |
1947 | } | |
1948 | re->next = head; | |
1949 | dest->routes = re; | |
1950 | ||
1951 | afi = (rn->p.family == AF_INET) | |
1952 | ? AFI_IP | |
1953 | : (rn->p.family == AF_INET6) ? AFI_IP6 : AFI_MAX; | |
1954 | if (is_zebra_import_table_enabled(afi, re->table)) { | |
1955 | rmap_name = zebra_get_import_table_route_map(afi, re->table); | |
1956 | zebra_add_import_table_entry(rn, re, rmap_name); | |
1957 | } else if (process) | |
1958 | rib_queue_add(rn); | |
1959 | } | |
1960 | ||
1961 | void rib_addnode(struct route_node *rn, struct route_entry *re, int process) | |
1962 | { | |
1963 | /* RE node has been un-removed before route-node is processed. | |
1964 | * route_node must hence already be on the queue for processing.. | |
1965 | */ | |
1966 | if (CHECK_FLAG(re->status, ROUTE_ENTRY_REMOVED)) { | |
1967 | if (IS_ZEBRA_DEBUG_RIB) | |
1968 | rnode_debug(rn, re->vrf_id, "rn %p, un-removed re %p", | |
1969 | (void *)rn, (void *)re); | |
1970 | ||
1971 | UNSET_FLAG(re->status, ROUTE_ENTRY_REMOVED); | |
1972 | return; | |
1973 | } | |
1974 | rib_link(rn, re, process); | |
6d691129 PJ |
1975 | } |
1976 | ||
9fd92e3c AS |
1977 | /* |
1978 | * rib_unlink | |
1979 | * | |
1980 | * Detach a rib structure from a route_node. | |
1981 | * | |
1982 | * Note that a call to rib_unlink() should be followed by a call to | |
1983 | * rib_gc_dest() at some point. This allows a rib_dest_t that is no | |
1984 | * longer required to be deleted. | |
1985 | */ | |
d62a17ae | 1986 | void rib_unlink(struct route_node *rn, struct route_entry *re) |
6d691129 | 1987 | { |
d62a17ae | 1988 | rib_dest_t *dest; |
9fd92e3c | 1989 | |
d62a17ae | 1990 | assert(rn && re); |
6d691129 | 1991 | |
d62a17ae | 1992 | if (IS_ZEBRA_DEBUG_RIB) |
1993 | rnode_debug(rn, re->vrf_id, "rn %p, re %p", (void *)rn, | |
1994 | (void *)re); | |
6d691129 | 1995 | |
d62a17ae | 1996 | dest = rib_dest_from_rnode(rn); |
6d691129 | 1997 | |
d62a17ae | 1998 | if (re->next) |
1999 | re->next->prev = re->prev; | |
6d691129 | 2000 | |
d62a17ae | 2001 | if (re->prev) |
2002 | re->prev->next = re->next; | |
2003 | else { | |
2004 | dest->routes = re->next; | |
2005 | } | |
7a4bb9c5 | 2006 | |
d62a17ae | 2007 | /* free RE and nexthops */ |
2008 | zebra_deregister_rnh_static_nexthops(re->vrf_id, re->nexthop, rn); | |
2009 | nexthops_free(re->nexthop); | |
2010 | XFREE(MTYPE_RE, re); | |
2011 | } | |
2012 | ||
2013 | void rib_delnode(struct route_node *rn, struct route_entry *re) | |
2014 | { | |
2015 | afi_t afi; | |
2016 | ||
2017 | if (IS_ZEBRA_DEBUG_RIB) | |
2018 | rnode_debug(rn, re->vrf_id, "rn %p, re %p, removing", | |
2019 | (void *)rn, (void *)re); | |
2020 | SET_FLAG(re->status, ROUTE_ENTRY_REMOVED); | |
2021 | ||
2022 | afi = (rn->p.family == AF_INET) | |
2023 | ? AFI_IP | |
2024 | : (rn->p.family == AF_INET6) ? AFI_IP6 : AFI_MAX; | |
2025 | if (is_zebra_import_table_enabled(afi, re->table)) { | |
2026 | zebra_del_import_table_entry(rn, re); | |
2027 | /* Just clean up if non main table */ | |
2028 | if (IS_ZEBRA_DEBUG_RIB) { | |
2029 | char buf[SRCDEST2STR_BUFFER]; | |
2030 | srcdest_rnode2str(rn, buf, sizeof(buf)); | |
2031 | zlog_debug( | |
2032 | "%u:%s: Freeing route rn %p, re %p (type %d)", | |
2033 | re->vrf_id, buf, rn, re, re->type); | |
2034 | } | |
7a4bb9c5 | 2035 | |
d62a17ae | 2036 | rib_unlink(rn, re); |
2037 | } else { | |
2038 | rib_queue_add(rn); | |
2039 | } | |
718e3744 | 2040 | } |
2041 | ||
f0f77c9a | 2042 | /* This function dumps the contents of a given RE entry into |
dc95824a DO |
2043 | * standard debug log. Calling function name and IP prefix in |
2044 | * question are passed as 1st and 2nd arguments. | |
2045 | */ | |
2046 | ||
d62a17ae | 2047 | void _route_entry_dump(const char *func, union prefixconstptr pp, |
2048 | union prefixconstptr src_pp, | |
2049 | const struct route_entry *re) | |
2050 | { | |
2051 | const struct prefix *p = pp.p; | |
2052 | const struct prefix *src_p = src_pp.p; | |
2053 | bool is_srcdst = src_p && src_p->prefixlen; | |
2054 | char straddr[PREFIX_STRLEN]; | |
2055 | char srcaddr[PREFIX_STRLEN]; | |
2056 | struct nexthop *nexthop; | |
2057 | ||
2058 | zlog_debug("%s: dumping RE entry %p for %s%s%s vrf %u", func, | |
2059 | (const void *)re, prefix2str(pp, straddr, sizeof(straddr)), | |
2060 | is_srcdst ? " from " : "", | |
2061 | is_srcdst ? prefix2str(src_pp, srcaddr, sizeof(srcaddr)) | |
2062 | : "", | |
2063 | re->vrf_id); | |
cc54cfee RW |
2064 | zlog_debug("%s: uptime == %lu, type == %u, instance == %d, table == %d", |
2065 | func, (unsigned long)re->uptime, re->type, re->instance, | |
2066 | re->table); | |
d62a17ae | 2067 | zlog_debug( |
2068 | "%s: metric == %u, mtu == %u, distance == %u, flags == %u, status == %u", | |
2069 | func, re->metric, re->mtu, re->distance, re->flags, re->status); | |
2070 | zlog_debug("%s: nexthop_num == %u, nexthop_active_num == %u", func, | |
2071 | re->nexthop_num, re->nexthop_active_num); | |
2072 | ||
2073 | for (ALL_NEXTHOPS(re->nexthop, nexthop)) { | |
2074 | inet_ntop(p->family, &nexthop->gate, straddr, INET6_ADDRSTRLEN); | |
915902cb | 2075 | zlog_debug("%s: %s %s[%u] with flags %s%s%s", func, |
d62a17ae | 2076 | (nexthop->rparent ? " NH" : "NH"), straddr, |
915902cb | 2077 | nexthop->ifindex, |
d62a17ae | 2078 | (CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE) |
2079 | ? "ACTIVE " | |
2080 | : ""), | |
2081 | (CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_FIB) | |
2082 | ? "FIB " | |
2083 | : ""), | |
2084 | (CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_RECURSIVE) | |
2085 | ? "RECURSIVE" | |
2086 | : "")); | |
2087 | } | |
2088 | zlog_debug("%s: dump complete", func); | |
dc95824a DO |
2089 | } |
2090 | ||
2091 | /* This is an exported helper to rtm_read() to dump the strange | |
f0f77c9a | 2092 | * RE entry found by rib_lookup_ipv4_route() |
dc95824a DO |
2093 | */ |
2094 | ||
d62a17ae | 2095 | void rib_lookup_and_dump(struct prefix_ipv4 *p, vrf_id_t vrf_id) |
2096 | { | |
2097 | struct route_table *table; | |
2098 | struct route_node *rn; | |
2099 | struct route_entry *re; | |
2100 | char prefix_buf[INET_ADDRSTRLEN]; | |
2101 | ||
2102 | /* Lookup table. */ | |
2103 | table = zebra_vrf_table(AFI_IP, SAFI_UNICAST, vrf_id); | |
2104 | if (!table) { | |
2105 | zlog_err("%s: zebra_vrf_table() returned NULL", __func__); | |
2106 | return; | |
2107 | } | |
2108 | ||
2109 | /* Scan the RIB table for exactly matching RE entry. */ | |
2110 | rn = route_node_lookup(table, (struct prefix *)p); | |
2111 | ||
2112 | /* No route for this prefix. */ | |
2113 | if (!rn) { | |
2114 | zlog_debug("%s: lookup failed for %s", __func__, | |
2115 | prefix2str((struct prefix *)p, prefix_buf, | |
2116 | sizeof(prefix_buf))); | |
2117 | return; | |
2118 | } | |
2119 | ||
2120 | /* Unlock node. */ | |
2121 | route_unlock_node(rn); | |
2122 | ||
2123 | /* let's go */ | |
2124 | RNODE_FOREACH_RE(rn, re) | |
2125 | { | |
2126 | zlog_debug("%s: rn %p, re %p: %s, %s", __func__, (void *)rn, | |
2127 | (void *)re, | |
2128 | (CHECK_FLAG(re->status, ROUTE_ENTRY_REMOVED) | |
2129 | ? "removed" | |
2130 | : "NOT removed"), | |
2131 | (CHECK_FLAG(re->flags, ZEBRA_FLAG_SELECTED) | |
2132 | ? "selected" | |
2133 | : "NOT selected")); | |
2134 | route_entry_dump(p, NULL, re); | |
2135 | } | |
dc95824a DO |
2136 | } |
2137 | ||
20e5ff0a DO |
2138 | /* Check if requested address assignment will fail due to another |
2139 | * route being installed by zebra in FIB already. Take necessary | |
2140 | * actions, if needed: remove such a route from FIB and deSELECT | |
f0f77c9a | 2141 | * corresponding RE entry. Then put affected RN into RIBQ head. |
20e5ff0a | 2142 | */ |
d62a17ae | 2143 | void rib_lookup_and_pushup(struct prefix_ipv4 *p, vrf_id_t vrf_id) |
2144 | { | |
2145 | struct route_table *table; | |
2146 | struct route_node *rn; | |
2147 | struct route_entry *re; | |
2148 | unsigned changed = 0; | |
2149 | ||
2150 | if (NULL == (table = zebra_vrf_table(AFI_IP, SAFI_UNICAST, vrf_id))) { | |
2151 | zlog_err("%s: zebra_vrf_table() returned NULL", __func__); | |
2152 | return; | |
2153 | } | |
2154 | ||
2155 | /* No matches would be the simplest case. */ | |
2156 | if (NULL == (rn = route_node_lookup(table, (struct prefix *)p))) | |
2157 | return; | |
2158 | ||
2159 | /* Unlock node. */ | |
2160 | route_unlock_node(rn); | |
2161 | ||
2162 | /* Check all RE entries. In case any changes have to be done, requeue | |
2163 | * the RN into RIBQ head. If the routing message about the new connected | |
2164 | * route (generated by the IP address we are going to assign very soon) | |
2165 | * comes before the RIBQ is processed, the new RE entry will join | |
2166 | * RIBQ record already on head. This is necessary for proper | |
2167 | * revalidation | |
2168 | * of the rest of the RE. | |
2169 | */ | |
2170 | RNODE_FOREACH_RE(rn, re) | |
2171 | { | |
2172 | if (CHECK_FLAG(re->status, ROUTE_ENTRY_SELECTED_FIB) | |
2173 | && !RIB_SYSTEM_ROUTE(re)) { | |
2174 | changed = 1; | |
2175 | if (IS_ZEBRA_DEBUG_RIB) { | |
2176 | char buf[PREFIX_STRLEN]; | |
2177 | zlog_debug( | |
2178 | "%u:%s: freeing way for connected prefix", | |
2179 | re->vrf_id, | |
2180 | prefix2str(&rn->p, buf, sizeof(buf))); | |
2181 | route_entry_dump(&rn->p, NULL, re); | |
2182 | } | |
2183 | rib_uninstall(rn, re); | |
2184 | } | |
2185 | } | |
2186 | if (changed) | |
2187 | rib_queue_add(rn); | |
20e5ff0a DO |
2188 | } |
2189 | ||
d62a17ae | 2190 | int rib_add_multipath(afi_t afi, safi_t safi, struct prefix *p, |
2191 | struct prefix_ipv6 *src_p, struct route_entry *re) | |
718e3744 | 2192 | { |
d62a17ae | 2193 | struct route_table *table; |
2194 | struct route_node *rn; | |
2195 | struct route_entry *same; | |
2196 | struct nexthop *nexthop; | |
2197 | int ret = 0; | |
2198 | int family; | |
b4c034b0 | 2199 | |
d62a17ae | 2200 | if (!re) |
2201 | return 0; | |
b4c034b0 | 2202 | |
d62a17ae | 2203 | if (p->family == AF_INET) |
2204 | family = AFI_IP; | |
2205 | else | |
2206 | family = AFI_IP6; | |
b4c034b0 | 2207 | |
d62a17ae | 2208 | assert(!src_p || family == AFI_IP6); |
05737783 | 2209 | |
d62a17ae | 2210 | /* Lookup table. */ |
2211 | table = zebra_vrf_table_with_table_id(family, safi, re->vrf_id, | |
2212 | re->table); | |
2213 | if (!table) | |
2214 | return 0; | |
cddf391b | 2215 | |
d62a17ae | 2216 | /* Make it sure prefixlen is applied to the prefix. */ |
2217 | apply_mask(p); | |
2218 | if (src_p) | |
2219 | apply_mask_ipv6(src_p); | |
718e3744 | 2220 | |
d62a17ae | 2221 | /* Set default distance by route type. */ |
2222 | if (re->distance == 0) { | |
2223 | re->distance = route_info[re->type].distance; | |
718e3744 | 2224 | |
d62a17ae | 2225 | /* iBGP distance is 200. */ |
2226 | if (re->type == ZEBRA_ROUTE_BGP | |
2227 | && CHECK_FLAG(re->flags, ZEBRA_FLAG_IBGP)) | |
2228 | re->distance = 200; | |
2229 | } | |
718e3744 | 2230 | |
d62a17ae | 2231 | /* Lookup route node.*/ |
2232 | rn = srcdest_rnode_get(table, p, src_p); | |
718e3744 | 2233 | |
d62a17ae | 2234 | /* If same type of route are installed, treat it as a implicit |
2235 | withdraw. */ | |
2236 | RNODE_FOREACH_RE(rn, same) | |
2237 | { | |
2238 | if (CHECK_FLAG(same->status, ROUTE_ENTRY_REMOVED)) | |
2239 | continue; | |
41ec9222 | 2240 | |
d62a17ae | 2241 | if (same->type == re->type && same->instance == re->instance |
8628fc61 | 2242 | && same->table == re->table && !RIB_SYSTEM_ROUTE(same)) |
d62a17ae | 2243 | break; |
2244 | } | |
718e3744 | 2245 | |
d62a17ae | 2246 | /* If this route is kernel route, set FIB flag to the route. */ |
8628fc61 | 2247 | if (RIB_SYSTEM_ROUTE(re)) |
d62a17ae | 2248 | for (nexthop = re->nexthop; nexthop; nexthop = nexthop->next) |
2249 | SET_FLAG(nexthop->flags, NEXTHOP_FLAG_FIB); | |
718e3744 | 2250 | |
d62a17ae | 2251 | /* Link new re to node.*/ |
2252 | if (IS_ZEBRA_DEBUG_RIB) { | |
2253 | rnode_debug( | |
2254 | rn, re->vrf_id, | |
2255 | "Inserting route rn %p, re %p (type %d) existing %p", | |
2256 | (void *)rn, (void *)re, re->type, (void *)same); | |
718e3744 | 2257 | |
d62a17ae | 2258 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) |
2259 | route_entry_dump(p, src_p, re); | |
718e3744 | 2260 | } |
d62a17ae | 2261 | rib_addnode(rn, re, 1); |
2262 | ret = 1; | |
6b0655a2 | 2263 | |
d62a17ae | 2264 | /* Free implicit route.*/ |
2265 | if (same) { | |
2266 | rib_delnode(rn, same); | |
2267 | ret = -1; | |
2268 | } | |
718e3744 | 2269 | |
d62a17ae | 2270 | route_unlock_node(rn); |
2271 | return ret; | |
2272 | } | |
2273 | ||
2274 | void rib_delete(afi_t afi, safi_t safi, vrf_id_t vrf_id, int type, | |
2275 | u_short instance, int flags, struct prefix *p, | |
fd36be7e DL |
2276 | struct prefix_ipv6 *src_p, const struct nexthop *nh, |
2277 | u_int32_t table_id, u_int32_t metric) | |
d62a17ae | 2278 | { |
2279 | struct route_table *table; | |
2280 | struct route_node *rn; | |
2281 | struct route_entry *re; | |
2282 | struct route_entry *fib = NULL; | |
2283 | struct route_entry *same = NULL; | |
fd36be7e | 2284 | struct nexthop *rtnh; |
d62a17ae | 2285 | char buf2[INET6_ADDRSTRLEN]; |
2286 | ||
2287 | assert(!src_p || afi == AFI_IP6); | |
2288 | ||
2289 | /* Lookup table. */ | |
2290 | table = zebra_vrf_table_with_table_id(afi, safi, vrf_id, table_id); | |
2291 | if (!table) | |
2292 | return; | |
2293 | ||
2294 | /* Apply mask. */ | |
2295 | apply_mask(p); | |
2296 | if (src_p) | |
2297 | apply_mask_ipv6(src_p); | |
2298 | ||
2299 | /* Lookup route node. */ | |
2300 | rn = srcdest_rnode_lookup(table, p, src_p); | |
2301 | if (!rn) { | |
2302 | char dst_buf[PREFIX_STRLEN], src_buf[PREFIX_STRLEN]; | |
2303 | ||
2304 | prefix2str(p, dst_buf, sizeof(dst_buf)); | |
2305 | if (src_p && src_p->prefixlen) | |
2306 | prefix2str(src_p, src_buf, sizeof(src_buf)); | |
2307 | else | |
2308 | src_buf[0] = '\0'; | |
2309 | ||
2310 | if (IS_ZEBRA_DEBUG_RIB) | |
2311 | zlog_debug("%u:%s%s%s doesn't exist in rib", vrf_id, | |
2312 | dst_buf, | |
2313 | (src_buf[0] != '\0') ? " from " : "", | |
2314 | src_buf); | |
2315 | return; | |
2316 | } | |
718e3744 | 2317 | |
d62a17ae | 2318 | /* Lookup same type route. */ |
2319 | RNODE_FOREACH_RE(rn, re) | |
2320 | { | |
2321 | if (CHECK_FLAG(re->status, ROUTE_ENTRY_REMOVED)) | |
2322 | continue; | |
2323 | ||
2324 | if (CHECK_FLAG(re->status, ROUTE_ENTRY_SELECTED_FIB)) | |
2325 | fib = re; | |
2326 | ||
2327 | if (re->type != type) | |
2328 | continue; | |
2329 | if (re->instance != instance) | |
2330 | continue; | |
f19435a8 DS |
2331 | if (re->type == ZEBRA_ROUTE_KERNEL && |
2332 | re->metric != metric) | |
2333 | continue; | |
fd36be7e DL |
2334 | if (re->type == ZEBRA_ROUTE_CONNECT && (rtnh = re->nexthop) |
2335 | && rtnh->type == NEXTHOP_TYPE_IFINDEX && nh) { | |
2336 | if (rtnh->ifindex != nh->ifindex) | |
d62a17ae | 2337 | continue; |
d62a17ae | 2338 | same = re; |
2339 | break; | |
2340 | } | |
2341 | /* Make sure that the route found has the same gateway. */ | |
2342 | else { | |
fd36be7e | 2343 | if (nh == NULL) { |
d62a17ae | 2344 | same = re; |
2345 | break; | |
2346 | } | |
fd36be7e DL |
2347 | for (ALL_NEXTHOPS(re->nexthop, rtnh)) |
2348 | if (nexthop_same_no_recurse(rtnh, nh)) { | |
d62a17ae | 2349 | same = re; |
2350 | break; | |
2351 | } | |
2352 | if (same) | |
2353 | break; | |
2354 | } | |
2355 | } | |
2356 | /* If same type of route can't be found and this message is from | |
2357 | kernel. */ | |
2358 | if (!same) { | |
2359 | if (fib && type == ZEBRA_ROUTE_KERNEL | |
2360 | && CHECK_FLAG(flags, ZEBRA_FLAG_SELFROUTE)) { | |
2361 | if (IS_ZEBRA_DEBUG_RIB) { | |
2362 | rnode_debug( | |
2363 | rn, vrf_id, | |
2364 | "rn %p, re %p (type %d) was deleted from kernel, adding", | |
2365 | rn, fib, fib->type); | |
2366 | } | |
2367 | if (allow_delete) { | |
2368 | /* Unset flags. */ | |
fd36be7e DL |
2369 | for (rtnh = fib->nexthop; rtnh; |
2370 | rtnh = rtnh->next) | |
2371 | UNSET_FLAG(rtnh->flags, | |
d62a17ae | 2372 | NEXTHOP_FLAG_FIB); |
2373 | ||
2374 | UNSET_FLAG(fib->status, | |
2375 | ROUTE_ENTRY_SELECTED_FIB); | |
2376 | } else { | |
2377 | /* This means someone else, other than Zebra, | |
2378 | * has deleted | |
2379 | * a Zebra router from the kernel. We will add | |
2380 | * it back */ | |
2381 | rib_install_kernel(rn, fib, NULL); | |
2382 | } | |
2383 | } else { | |
2384 | if (IS_ZEBRA_DEBUG_RIB) { | |
fd36be7e | 2385 | if (nh) |
d62a17ae | 2386 | rnode_debug( |
2387 | rn, vrf_id, | |
2388 | "via %s ifindex %d type %d " | |
2389 | "doesn't exist in rib", | |
2390 | inet_ntop( | |
60466a63 QY |
2391 | family2afi(afi), |
2392 | &nh->gate, buf2, | |
d62a17ae | 2393 | INET_ADDRSTRLEN), /* FIXME |
9d303b37 | 2394 | */ |
fd36be7e | 2395 | nh->ifindex, type); |
d62a17ae | 2396 | else |
2397 | rnode_debug( | |
2398 | rn, vrf_id, | |
fd36be7e DL |
2399 | "type %d doesn't exist in rib", |
2400 | type); | |
d62a17ae | 2401 | } |
2402 | route_unlock_node(rn); | |
2403 | return; | |
2404 | } | |
2405 | } | |
718e3744 | 2406 | |
d62a17ae | 2407 | if (same) |
2408 | rib_delnode(rn, same); | |
05737783 | 2409 | |
d62a17ae | 2410 | route_unlock_node(rn); |
2411 | return; | |
2412 | } | |
718e3744 | 2413 | |
718e3744 | 2414 | |
d62a17ae | 2415 | int rib_add(afi_t afi, safi_t safi, vrf_id_t vrf_id, int type, u_short instance, |
2416 | int flags, struct prefix *p, struct prefix_ipv6 *src_p, | |
fd36be7e DL |
2417 | const struct nexthop *nh, u_int32_t table_id, u_int32_t metric, |
2418 | u_int32_t mtu, u_char distance) | |
d62a17ae | 2419 | { |
2420 | struct route_entry *re; | |
2421 | struct route_entry *same = NULL; | |
2422 | struct route_table *table; | |
2423 | struct route_node *rn; | |
fd36be7e | 2424 | struct nexthop *rtnh; |
d62a17ae | 2425 | |
2426 | assert(!src_p || afi == AFI_IP6); | |
2427 | ||
2428 | /* Lookup table. */ | |
2429 | table = zebra_vrf_table_with_table_id(afi, safi, vrf_id, table_id); | |
2430 | if (!table) | |
2431 | return 0; | |
2432 | ||
2433 | /* Make sure mask is applied. */ | |
2434 | apply_mask(p); | |
2435 | if (src_p) | |
2436 | apply_mask_ipv6(src_p); | |
2437 | ||
2438 | /* Set default distance by route type. */ | |
2439 | if (distance == 0) { | |
2440 | if ((unsigned)type >= array_size(route_info)) | |
2441 | distance = 150; | |
2442 | else | |
2443 | distance = route_info[type].distance; | |
2444 | ||
2445 | /* iBGP distance is 200. */ | |
2446 | if (type == ZEBRA_ROUTE_BGP | |
2447 | && CHECK_FLAG(flags, ZEBRA_FLAG_IBGP)) | |
2448 | distance = 200; | |
2449 | } | |
718e3744 | 2450 | |
d62a17ae | 2451 | /* Lookup route node.*/ |
2452 | rn = srcdest_rnode_get(table, p, src_p); | |
718e3744 | 2453 | |
d62a17ae | 2454 | /* If same type of route are installed, treat it as a implicit |
2455 | withdraw. */ | |
2456 | RNODE_FOREACH_RE(rn, re) | |
ebf1ead0 | 2457 | { |
d62a17ae | 2458 | if (CHECK_FLAG(re->status, ROUTE_ENTRY_REMOVED)) |
2459 | continue; | |
2460 | ||
2461 | if (re->type != type) | |
2462 | continue; | |
2463 | if (re->instance != instance) | |
2464 | continue; | |
f19435a8 DS |
2465 | if (re->type == ZEBRA_ROUTE_KERNEL && |
2466 | re->metric != metric) | |
2467 | continue; | |
3ce58880 | 2468 | if (!RIB_SYSTEM_ROUTE(re)) { |
d62a17ae | 2469 | same = re; |
2470 | break; | |
2471 | } | |
3ce58880 | 2472 | /* Duplicate system route comes in. */ |
a3d18ce6 RW |
2473 | rtnh = re->nexthop; |
2474 | if (nexthop_same_no_recurse(rtnh, nh)) | |
d62a17ae | 2475 | return 0; |
5f20552d RW |
2476 | /* |
2477 | * Nexthop is different. Remove the old route unless it's | |
c2713b2a RW |
2478 | * a connected route. This exception is necessary because |
2479 | * of IPv6 link-local routes and unnumbered interfaces on | |
2480 | * Linux. | |
5f20552d | 2481 | */ |
c2713b2a | 2482 | else if (type != ZEBRA_ROUTE_CONNECT) |
a3d18ce6 | 2483 | same = re; |
ebf1ead0 | 2484 | } |
718e3744 | 2485 | |
d62a17ae | 2486 | /* Allocate new re structure. */ |
2487 | re = XCALLOC(MTYPE_RE, sizeof(struct route_entry)); | |
2488 | ||
2489 | re->type = type; | |
2490 | re->instance = instance; | |
2491 | re->distance = distance; | |
2492 | re->flags = flags; | |
2493 | re->metric = metric; | |
2494 | re->mtu = mtu; | |
2495 | re->table = table_id; | |
2496 | re->vrf_id = vrf_id; | |
2497 | re->nexthop_num = 0; | |
2498 | re->uptime = time(NULL); | |
2499 | ||
fd36be7e DL |
2500 | rtnh = nexthop_new(); |
2501 | *rtnh = *nh; | |
2502 | route_entry_nexthop_add(re, rtnh); | |
d62a17ae | 2503 | |
2504 | /* If this route is kernel route, set FIB flag to the route. */ | |
8628fc61 | 2505 | if (RIB_SYSTEM_ROUTE(re)) |
fd36be7e DL |
2506 | for (rtnh = re->nexthop; rtnh; rtnh = rtnh->next) |
2507 | SET_FLAG(rtnh->flags, NEXTHOP_FLAG_FIB); | |
d62a17ae | 2508 | |
2509 | /* Link new rib to node.*/ | |
2510 | if (IS_ZEBRA_DEBUG_RIB) { | |
2511 | rnode_debug( | |
2512 | rn, vrf_id, | |
2513 | "Inserting route rn %p, re %p (type %d) existing %p", | |
2514 | (void *)rn, (void *)re, re->type, (void *)same); | |
2515 | ||
2516 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) | |
2517 | route_entry_dump(p, src_p, re); | |
3b1098be | 2518 | } |
d62a17ae | 2519 | rib_addnode(rn, re, 1); |
718e3744 | 2520 | |
d62a17ae | 2521 | /* Free implicit route.*/ |
2522 | if (same) | |
2523 | rib_delnode(rn, same); | |
718e3744 | 2524 | |
d62a17ae | 2525 | route_unlock_node(rn); |
2526 | return 0; | |
718e3744 | 2527 | } |
2528 | ||
1c848137 | 2529 | /* Schedule routes of a particular table (address-family) based on event. */ |
d62a17ae | 2530 | static void rib_update_table(struct route_table *table, |
2531 | rib_update_event_t event) | |
2532 | { | |
2533 | struct route_node *rn; | |
2534 | struct route_entry *re, *next; | |
2535 | ||
2536 | /* Walk all routes and queue for processing, if appropriate for | |
2537 | * the trigger event. | |
2538 | */ | |
2539 | for (rn = route_top(table); rn; rn = srcdest_route_next(rn)) { | |
2540 | switch (event) { | |
2541 | case RIB_UPDATE_IF_CHANGE: | |
2542 | /* Examine all routes that won't get processed by the | |
2543 | * protocol or | |
2544 | * triggered by nexthop evaluation (NHT). This would be | |
2545 | * system, | |
2546 | * kernel and certain static routes. Note that NHT will | |
2547 | * get | |
2548 | * triggered upon an interface event as connected routes | |
2549 | * always | |
2550 | * get queued for processing. | |
2551 | */ | |
2552 | RNODE_FOREACH_RE_SAFE(rn, re, next) | |
2553 | { | |
2554 | if (re->type == ZEBRA_ROUTE_OSPF | |
2555 | || re->type == ZEBRA_ROUTE_OSPF6 | |
2556 | || re->type == ZEBRA_ROUTE_BGP) | |
2557 | continue; /* protocol will handle. */ | |
2558 | else if (re->type == ZEBRA_ROUTE_STATIC) { | |
2559 | struct nexthop *nh; | |
2560 | for (nh = re->nexthop; nh; | |
2561 | nh = nh->next) | |
2562 | if (!(nh->type | |
2563 | == NEXTHOP_TYPE_IPV4 | |
2564 | || nh->type | |
2565 | == NEXTHOP_TYPE_IPV6)) | |
2566 | break; | |
2567 | ||
2568 | /* If we only have nexthops to a | |
2569 | * gateway, NHT will | |
2570 | * take care. | |
2571 | */ | |
2572 | if (nh) | |
2573 | rib_queue_add(rn); | |
2574 | } else | |
2575 | rib_queue_add(rn); | |
2576 | } | |
2577 | break; | |
2578 | ||
2579 | case RIB_UPDATE_RMAP_CHANGE: | |
2580 | case RIB_UPDATE_OTHER: | |
2581 | /* Right now, examine all routes. Can restrict to a | |
2582 | * protocol in | |
2583 | * some cases (TODO). | |
2584 | */ | |
2585 | if (rnode_to_ribs(rn)) | |
2586 | rib_queue_add(rn); | |
2587 | break; | |
2588 | ||
2589 | default: | |
2590 | break; | |
2591 | } | |
2592 | } | |
b84c7253 | 2593 | } |
2594 | ||
718e3744 | 2595 | /* RIB update function. */ |
d62a17ae | 2596 | void rib_update(vrf_id_t vrf_id, rib_update_event_t event) |
718e3744 | 2597 | { |
d62a17ae | 2598 | struct route_table *table; |
1c848137 | 2599 | |
d62a17ae | 2600 | /* Process routes of interested address-families. */ |
2601 | table = zebra_vrf_table(AFI_IP, SAFI_UNICAST, vrf_id); | |
2602 | if (table) | |
2603 | rib_update_table(table, event); | |
718e3744 | 2604 | |
d62a17ae | 2605 | table = zebra_vrf_table(AFI_IP6, SAFI_UNICAST, vrf_id); |
2606 | if (table) | |
2607 | rib_update_table(table, event); | |
718e3744 | 2608 | } |
2609 | ||
718e3744 | 2610 | /* Remove all routes which comes from non main table. */ |
d62a17ae | 2611 | static void rib_weed_table(struct route_table *table) |
718e3744 | 2612 | { |
d62a17ae | 2613 | struct route_node *rn; |
2614 | struct route_entry *re; | |
2615 | struct route_entry *next; | |
718e3744 | 2616 | |
d62a17ae | 2617 | if (table) |
2618 | for (rn = route_top(table); rn; rn = srcdest_route_next(rn)) | |
2619 | RNODE_FOREACH_RE_SAFE(rn, re, next) | |
2620 | { | |
2621 | if (CHECK_FLAG(re->status, ROUTE_ENTRY_REMOVED)) | |
2622 | continue; | |
6d691129 | 2623 | |
d62a17ae | 2624 | if (re->table != zebrad.rtm_table_default |
2625 | && re->table != RT_TABLE_MAIN) | |
2626 | rib_delnode(rn, re); | |
2627 | } | |
718e3744 | 2628 | } |
2629 | ||
2630 | /* Delete all routes from non main table. */ | |
d62a17ae | 2631 | void rib_weed_tables(void) |
718e3744 | 2632 | { |
d62a17ae | 2633 | struct vrf *vrf; |
2634 | struct zebra_vrf *zvrf; | |
78104b9b | 2635 | |
d62a17ae | 2636 | RB_FOREACH(vrf, vrf_id_head, &vrfs_by_id) |
2637 | if ((zvrf = vrf->info) != NULL) { | |
2638 | rib_weed_table(zvrf->table[AFI_IP][SAFI_UNICAST]); | |
2639 | rib_weed_table(zvrf->table[AFI_IP6][SAFI_UNICAST]); | |
2640 | } | |
718e3744 | 2641 | } |
6b0655a2 | 2642 | |
718e3744 | 2643 | /* Delete self installed routes after zebra is relaunched. */ |
d62a17ae | 2644 | static void rib_sweep_table(struct route_table *table) |
2645 | { | |
2646 | struct route_node *rn; | |
2647 | struct route_entry *re; | |
2648 | struct route_entry *next; | |
915902cb | 2649 | struct nexthop *nexthop; |
d62a17ae | 2650 | int ret = 0; |
2651 | ||
915902cb DS |
2652 | if (!table) |
2653 | return; | |
d62a17ae | 2654 | |
915902cb DS |
2655 | for (rn = route_top(table); rn; rn = srcdest_route_next(rn)) { |
2656 | RNODE_FOREACH_RE_SAFE(rn, re, next) | |
2657 | { | |
2658 | if (IS_ZEBRA_DEBUG_RIB) | |
2659 | route_entry_dump(&rn->p, NULL, re); | |
2660 | ||
2661 | if (CHECK_FLAG(re->status, ROUTE_ENTRY_REMOVED)) | |
2662 | continue; | |
2663 | ||
2664 | if (!CHECK_FLAG(re->flags, ZEBRA_FLAG_SELFROUTE)) | |
2665 | continue; | |
2666 | ||
2667 | /* | |
2668 | * So we are starting up and have received | |
2669 | * routes from the kernel that we have installed | |
2670 | * from a previous run of zebra but not cleaned | |
2671 | * up ( say a kill -9 ) | |
2672 | * But since we haven't actually installed | |
2673 | * them yet( we received them from the kernel ) | |
2674 | * we don't think they are active. | |
2675 | * So let's pretend they are active to actually | |
2676 | * remove them. | |
2677 | * In all honesty I'm not sure if we should | |
2678 | * mark them as active when we receive them | |
2679 | * This is startup only so probably ok. | |
2680 | * | |
2681 | * If we ever decide to move rib_sweep_table | |
2682 | * to a different spot (ie startup ) | |
2683 | * this decision needs to be revisited | |
2684 | */ | |
2685 | for (ALL_NEXTHOPS(re->nexthop, nexthop)) | |
2686 | SET_FLAG(nexthop->flags, NEXTHOP_FLAG_FIB); | |
2687 | ||
2688 | ret = rib_uninstall_kernel(rn, re); | |
2689 | if (!ret) | |
2690 | rib_delnode(rn, re); | |
2691 | } | |
2692 | } | |
718e3744 | 2693 | } |
2694 | ||
2695 | /* Sweep all RIB tables. */ | |
d62a17ae | 2696 | void rib_sweep_route(void) |
718e3744 | 2697 | { |
d62a17ae | 2698 | struct vrf *vrf; |
2699 | struct zebra_vrf *zvrf; | |
78104b9b | 2700 | |
915902cb DS |
2701 | RB_FOREACH(vrf, vrf_id_head, &vrfs_by_id) { |
2702 | if ((zvrf = vrf->info) == NULL) | |
2703 | continue; | |
2704 | ||
d62a17ae | 2705 | rib_sweep_table(zvrf->table[AFI_IP][SAFI_UNICAST]); |
2706 | rib_sweep_table(zvrf->table[AFI_IP6][SAFI_UNICAST]); | |
2707 | } | |
718e3744 | 2708 | } |
2ea1ab1c VT |
2709 | |
2710 | /* Remove specific by protocol routes from 'table'. */ | |
d62a17ae | 2711 | static unsigned long rib_score_proto_table(u_char proto, u_short instance, |
2712 | struct route_table *table) | |
2713 | { | |
2714 | struct route_node *rn; | |
2715 | struct route_entry *re; | |
2716 | struct route_entry *next; | |
2717 | unsigned long n = 0; | |
2718 | ||
2719 | if (table) | |
2720 | for (rn = route_top(table); rn; rn = srcdest_route_next(rn)) | |
2721 | RNODE_FOREACH_RE_SAFE(rn, re, next) | |
2722 | { | |
2723 | if (CHECK_FLAG(re->status, ROUTE_ENTRY_REMOVED)) | |
2724 | continue; | |
2725 | if (re->type == proto | |
2726 | && re->instance == instance) { | |
2727 | rib_delnode(rn, re); | |
2728 | n++; | |
2729 | } | |
2730 | } | |
2731 | return n; | |
2ea1ab1c VT |
2732 | } |
2733 | ||
2734 | /* Remove specific by protocol routes. */ | |
d62a17ae | 2735 | unsigned long rib_score_proto(u_char proto, u_short instance) |
2ea1ab1c | 2736 | { |
d62a17ae | 2737 | struct vrf *vrf; |
2738 | struct zebra_vrf *zvrf; | |
2739 | unsigned long cnt = 0; | |
78104b9b | 2740 | |
d62a17ae | 2741 | RB_FOREACH(vrf, vrf_id_head, &vrfs_by_id) |
2742 | if ((zvrf = vrf->info) != NULL) | |
2743 | cnt += rib_score_proto_table(proto, instance, | |
2744 | zvrf->table[AFI_IP][SAFI_UNICAST]) | |
2745 | + rib_score_proto_table( | |
2746 | proto, instance, | |
2747 | zvrf->table[AFI_IP6][SAFI_UNICAST]); | |
78104b9b | 2748 | |
d62a17ae | 2749 | return cnt; |
2ea1ab1c VT |
2750 | } |
2751 | ||
718e3744 | 2752 | /* Close RIB and clean up kernel routes. */ |
d62a17ae | 2753 | void rib_close_table(struct route_table *table) |
718e3744 | 2754 | { |
d62a17ae | 2755 | struct route_node *rn; |
1e9f448f | 2756 | rib_table_info_t *info; |
d62a17ae | 2757 | struct route_entry *re; |
718e3744 | 2758 | |
1e9f448f DS |
2759 | if (!table) |
2760 | return; | |
9fd92e3c | 2761 | |
1e9f448f | 2762 | info = table->info; |
5adc2528 | 2763 | |
1e9f448f DS |
2764 | for (rn = route_top(table); rn; rn = srcdest_route_next(rn)) |
2765 | RNODE_FOREACH_RE(rn, re) | |
2766 | { | |
2767 | if (!CHECK_FLAG(re->status, | |
2768 | ROUTE_ENTRY_SELECTED_FIB)) | |
2769 | continue; | |
2770 | ||
2771 | if (info->safi == SAFI_UNICAST) | |
2772 | hook_call(rib_update, rn, NULL); | |
2773 | ||
2774 | if (!RIB_SYSTEM_ROUTE(re)) | |
2775 | rib_uninstall_kernel(rn, re); | |
2776 | } | |
718e3744 | 2777 | } |
2778 | ||
718e3744 | 2779 | /* Routing information base initialize. */ |
d62a17ae | 2780 | void rib_init(void) |
718e3744 | 2781 | { |
d62a17ae | 2782 | rib_queue_init(&zebrad); |
718e3744 | 2783 | } |
0915bb0c AS |
2784 | |
2785 | /* | |
2786 | * vrf_id_get_next | |
2787 | * | |
2788 | * Get the first vrf id that is greater than the given vrf id if any. | |
2789 | * | |
2790 | * Returns TRUE if a vrf id was found, FALSE otherwise. | |
2791 | */ | |
d62a17ae | 2792 | static inline int vrf_id_get_next(vrf_id_t vrf_id, vrf_id_t *next_id_p) |
0915bb0c | 2793 | { |
d62a17ae | 2794 | struct vrf *vrf; |
b72ede27 | 2795 | |
d62a17ae | 2796 | vrf = vrf_lookup_by_id(vrf_id); |
2797 | if (vrf) { | |
2798 | vrf = RB_NEXT(vrf_id_head, vrf); | |
2799 | if (vrf) { | |
2800 | *next_id_p = vrf->vrf_id; | |
2801 | return 1; | |
2802 | } | |
2803 | } | |
0915bb0c | 2804 | |
d62a17ae | 2805 | return 0; |
0915bb0c AS |
2806 | } |
2807 | ||
2808 | /* | |
2809 | * rib_tables_iter_next | |
2810 | * | |
2811 | * Returns the next table in the iteration. | |
2812 | */ | |
d62a17ae | 2813 | struct route_table *rib_tables_iter_next(rib_tables_iter_t *iter) |
2814 | { | |
2815 | struct route_table *table; | |
2816 | ||
2817 | /* | |
2818 | * Array that helps us go over all AFI/SAFI combinations via one | |
2819 | * index. | |
2820 | */ | |
2821 | static struct { | |
2822 | afi_t afi; | |
2823 | safi_t safi; | |
2824 | } afi_safis[] = { | |
2825 | {AFI_IP, SAFI_UNICAST}, {AFI_IP, SAFI_MULTICAST}, | |
2826 | {AFI_IP, SAFI_LABELED_UNICAST}, {AFI_IP6, SAFI_UNICAST}, | |
2827 | {AFI_IP6, SAFI_MULTICAST}, {AFI_IP6, SAFI_LABELED_UNICAST}, | |
2828 | }; | |
2829 | ||
2830 | table = NULL; | |
2831 | ||
2832 | switch (iter->state) { | |
2833 | ||
2834 | case RIB_TABLES_ITER_S_INIT: | |
2835 | iter->vrf_id = VRF_DEFAULT; | |
2836 | iter->afi_safi_ix = -1; | |
2837 | ||
2838 | /* Fall through */ | |
2839 | ||
2840 | case RIB_TABLES_ITER_S_ITERATING: | |
2841 | iter->afi_safi_ix++; | |
2842 | while (1) { | |
2843 | ||
2844 | while (iter->afi_safi_ix | |
2845 | < (int)ZEBRA_NUM_OF(afi_safis)) { | |
2846 | table = zebra_vrf_table( | |
2847 | afi_safis[iter->afi_safi_ix].afi, | |
2848 | afi_safis[iter->afi_safi_ix].safi, | |
2849 | iter->vrf_id); | |
2850 | if (table) | |
2851 | break; | |
2852 | ||
2853 | iter->afi_safi_ix++; | |
2854 | } | |
2855 | ||
2856 | /* | |
2857 | * Found another table in this vrf. | |
2858 | */ | |
2859 | if (table) | |
2860 | break; | |
2861 | ||
2862 | /* | |
2863 | * Done with all tables in the current vrf, go to the | |
2864 | * next | |
2865 | * one. | |
2866 | */ | |
2867 | if (!vrf_id_get_next(iter->vrf_id, &iter->vrf_id)) | |
2868 | break; | |
2869 | ||
2870 | iter->afi_safi_ix = 0; | |
2871 | } | |
0915bb0c | 2872 | |
0915bb0c AS |
2873 | break; |
2874 | ||
d62a17ae | 2875 | case RIB_TABLES_ITER_S_DONE: |
2876 | return NULL; | |
0915bb0c AS |
2877 | } |
2878 | ||
d62a17ae | 2879 | if (table) |
2880 | iter->state = RIB_TABLES_ITER_S_ITERATING; | |
2881 | else | |
2882 | iter->state = RIB_TABLES_ITER_S_DONE; | |
0915bb0c | 2883 | |
d62a17ae | 2884 | return table; |
0915bb0c | 2885 | } |