]>
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 | ||
718e3744 | 60 | /* Each route type's string and default distance value. */ |
d62a17ae | 61 | static const struct { |
62 | int key; | |
63 | int distance; | |
64 | } route_info[ZEBRA_ROUTE_MAX] = { | |
9d303b37 DL |
65 | [ZEBRA_ROUTE_SYSTEM] = {ZEBRA_ROUTE_SYSTEM, 0}, |
66 | [ZEBRA_ROUTE_KERNEL] = {ZEBRA_ROUTE_KERNEL, 0}, | |
67 | [ZEBRA_ROUTE_CONNECT] = {ZEBRA_ROUTE_CONNECT, 0}, | |
68 | [ZEBRA_ROUTE_STATIC] = {ZEBRA_ROUTE_STATIC, 1}, | |
69 | [ZEBRA_ROUTE_RIP] = {ZEBRA_ROUTE_RIP, 120}, | |
70 | [ZEBRA_ROUTE_RIPNG] = {ZEBRA_ROUTE_RIPNG, 120}, | |
71 | [ZEBRA_ROUTE_OSPF] = {ZEBRA_ROUTE_OSPF, 110}, | |
72 | [ZEBRA_ROUTE_OSPF6] = {ZEBRA_ROUTE_OSPF6, 110}, | |
73 | [ZEBRA_ROUTE_ISIS] = {ZEBRA_ROUTE_ISIS, 115}, | |
74 | [ZEBRA_ROUTE_BGP] = {ZEBRA_ROUTE_BGP, 20 /* IBGP is 200. */}, | |
c710b277 DS |
75 | [ZEBRA_ROUTE_PIM] = {ZEBRA_ROUTE_PIM, 255}, |
76 | [ZEBRA_ROUTE_EIGRP] = {ZEBRA_ROUTE_EIGRP, 90}, | |
9d303b37 | 77 | [ZEBRA_ROUTE_NHRP] = {ZEBRA_ROUTE_NHRP, 10}, |
c710b277 DS |
78 | [ZEBRA_ROUTE_HSLS] = {ZEBRA_ROUTE_HSLS, 255}, |
79 | [ZEBRA_ROUTE_OLSR] = {ZEBRA_ROUTE_OLSR, 255}, | |
80 | [ZEBRA_ROUTE_TABLE] = {ZEBRA_ROUTE_TABLE, 150}, | |
81 | [ZEBRA_ROUTE_LDP] = {ZEBRA_ROUTE_LDP, 150}, | |
82 | [ZEBRA_ROUTE_VNC] = {ZEBRA_ROUTE_VNC, 20}, | |
83 | [ZEBRA_ROUTE_VNC_DIRECT] = {ZEBRA_ROUTE_VNC_DIRECT, 20}, | |
84 | [ZEBRA_ROUTE_VNC_DIRECT_RH] = {ZEBRA_ROUTE_VNC_DIRECT_RH, 20}, | |
85 | [ZEBRA_ROUTE_BGP_DIRECT] = {ZEBRA_ROUTE_BGP_DIRECT, 20}, | |
86 | [ZEBRA_ROUTE_BGP_DIRECT_EXT] = {ZEBRA_ROUTE_BGP_DIRECT_EXT, 20}, | |
87 | [ZEBRA_ROUTE_BABEL] = {ZEBRA_ROUTE_BABEL, 100}, | |
88 | ||
d62a17ae | 89 | /* no entry/default: 150 */ |
718e3744 | 90 | }; |
6b0655a2 | 91 | |
4623d897 DL |
92 | /* RPF lookup behaviour */ |
93 | static enum multicast_mode ipv4_multicast_mode = MCAST_NO_CONFIG; | |
94 | ||
6c4f4e6e | 95 | |
d62a17ae | 96 | static void __attribute__((format(printf, 5, 6))) |
97 | _rnode_zlog(const char *_func, vrf_id_t vrf_id, struct route_node *rn, | |
98 | int priority, const char *msgfmt, ...) | |
2263a412 | 99 | { |
d62a17ae | 100 | char buf[SRCDEST2STR_BUFFER + sizeof(" (MRIB)")]; |
101 | char msgbuf[512]; | |
102 | va_list ap; | |
2263a412 | 103 | |
d62a17ae | 104 | va_start(ap, msgfmt); |
105 | vsnprintf(msgbuf, sizeof(msgbuf), msgfmt, ap); | |
106 | va_end(ap); | |
2263a412 | 107 | |
d62a17ae | 108 | if (rn) { |
109 | rib_table_info_t *info = srcdest_rnode_table_info(rn); | |
110 | srcdest_rnode2str(rn, buf, sizeof(buf)); | |
cb653491 | 111 | |
d62a17ae | 112 | if (info->safi == SAFI_MULTICAST) |
113 | strcat(buf, " (MRIB)"); | |
114 | } else { | |
115 | snprintf(buf, sizeof(buf), "{(route_node *) NULL}"); | |
116 | } | |
2263a412 | 117 | |
d62a17ae | 118 | zlog(priority, "%s: %d:%s: %s", _func, vrf_id, buf, msgbuf); |
2263a412 DL |
119 | } |
120 | ||
d62a17ae | 121 | #define rnode_debug(node, vrf_id, ...) \ |
2263a412 | 122 | _rnode_zlog(__func__, vrf_id, node, LOG_DEBUG, __VA_ARGS__) |
d62a17ae | 123 | #define rnode_info(node, ...) \ |
2263a412 DL |
124 | _rnode_zlog(__func__, vrf_id, node, LOG_INFO, __VA_ARGS__) |
125 | ||
fd289fc8 | 126 | uint8_t route_distance(int type) |
40c7bdb0 | 127 | { |
fd289fc8 | 128 | uint8_t distance; |
40c7bdb0 | 129 | |
d62a17ae | 130 | if ((unsigned)type >= array_size(route_info)) |
131 | distance = 150; | |
132 | else | |
133 | distance = route_info[type].distance; | |
40c7bdb0 | 134 | |
d62a17ae | 135 | return distance; |
40c7bdb0 | 136 | } |
137 | ||
d62a17ae | 138 | int is_zebra_valid_kernel_table(u_int32_t table_id) |
7a4bb9c5 | 139 | { |
d62a17ae | 140 | if ((table_id > ZEBRA_KERNEL_TABLE_MAX)) |
141 | return 0; | |
8f500a1c RW |
142 | |
143 | #ifdef linux | |
d62a17ae | 144 | if ((table_id == RT_TABLE_UNSPEC) || (table_id == RT_TABLE_LOCAL) |
145 | || (table_id == RT_TABLE_COMPAT)) | |
146 | return 0; | |
8f500a1c RW |
147 | #endif |
148 | ||
d62a17ae | 149 | return 1; |
7a4bb9c5 DS |
150 | } |
151 | ||
d62a17ae | 152 | int is_zebra_main_routing_table(u_int32_t table_id) |
7a4bb9c5 | 153 | { |
d62a17ae | 154 | if ((table_id == RT_TABLE_MAIN) |
155 | || (table_id == zebrad.rtm_table_default)) | |
156 | return 1; | |
157 | return 0; | |
7a4bb9c5 DS |
158 | } |
159 | ||
d62a17ae | 160 | int zebra_check_addr(struct prefix *p) |
0aabccc0 | 161 | { |
d62a17ae | 162 | if (p->family == AF_INET) { |
163 | u_int32_t addr; | |
0aabccc0 | 164 | |
d62a17ae | 165 | addr = p->u.prefix4.s_addr; |
166 | addr = ntohl(addr); | |
0aabccc0 | 167 | |
d62a17ae | 168 | if (IPV4_NET127(addr) || IN_CLASSD(addr) |
169 | || IPV4_LINKLOCAL(addr)) | |
170 | return 0; | |
171 | } | |
172 | if (p->family == AF_INET6) { | |
173 | if (IN6_IS_ADDR_LOOPBACK(&p->u.prefix6)) | |
174 | return 0; | |
175 | if (IN6_IS_ADDR_LINKLOCAL(&p->u.prefix6)) | |
176 | return 0; | |
177 | } | |
178 | return 1; | |
0aabccc0 DD |
179 | } |
180 | ||
fa713d9e | 181 | /* Add nexthop to the end of a rib node's nexthop list */ |
d62a17ae | 182 | void route_entry_nexthop_add(struct route_entry *re, struct nexthop *nexthop) |
fa713d9e | 183 | { |
d62a17ae | 184 | nexthop_add(&re->nexthop, nexthop); |
185 | re->nexthop_num++; | |
718e3744 | 186 | } |
187 | ||
6e26278c | 188 | |
6e26278c DS |
189 | /** |
190 | * copy_nexthop - copy a nexthop to the rib structure. | |
191 | */ | |
d62a17ae | 192 | void route_entry_copy_nexthops(struct route_entry *re, struct nexthop *nh) |
6e26278c | 193 | { |
d62a17ae | 194 | assert(!re->nexthop); |
195 | copy_nexthops(&re->nexthop, nh, NULL); | |
196 | for (struct nexthop *nexthop = nh; nexthop; nexthop = nexthop->next) | |
197 | re->nexthop_num++; | |
6e26278c DS |
198 | } |
199 | ||
718e3744 | 200 | /* Delete specified nexthop from the list. */ |
d62a17ae | 201 | void route_entry_nexthop_delete(struct route_entry *re, struct nexthop *nexthop) |
718e3744 | 202 | { |
d62a17ae | 203 | if (nexthop->next) |
204 | nexthop->next->prev = nexthop->prev; | |
205 | if (nexthop->prev) | |
206 | nexthop->prev->next = nexthop->next; | |
207 | else | |
208 | re->nexthop = nexthop->next; | |
209 | re->nexthop_num--; | |
718e3744 | 210 | } |
211 | ||
fa713d9e | 212 | |
d62a17ae | 213 | struct nexthop *route_entry_nexthop_ifindex_add(struct route_entry *re, |
214 | ifindex_t ifindex) | |
718e3744 | 215 | { |
d62a17ae | 216 | struct nexthop *nexthop; |
718e3744 | 217 | |
d62a17ae | 218 | nexthop = nexthop_new(); |
219 | nexthop->type = NEXTHOP_TYPE_IFINDEX; | |
220 | nexthop->ifindex = ifindex; | |
718e3744 | 221 | |
d62a17ae | 222 | route_entry_nexthop_add(re, nexthop); |
718e3744 | 223 | |
d62a17ae | 224 | return nexthop; |
718e3744 | 225 | } |
226 | ||
d62a17ae | 227 | struct nexthop *route_entry_nexthop_ipv4_add(struct route_entry *re, |
228 | struct in_addr *ipv4, | |
229 | struct in_addr *src) | |
718e3744 | 230 | { |
d62a17ae | 231 | struct nexthop *nexthop; |
718e3744 | 232 | |
d62a17ae | 233 | nexthop = nexthop_new(); |
234 | nexthop->type = NEXTHOP_TYPE_IPV4; | |
235 | nexthop->gate.ipv4 = *ipv4; | |
236 | if (src) | |
237 | nexthop->src.ipv4 = *src; | |
718e3744 | 238 | |
d62a17ae | 239 | route_entry_nexthop_add(re, nexthop); |
718e3744 | 240 | |
d62a17ae | 241 | return nexthop; |
718e3744 | 242 | } |
243 | ||
d62a17ae | 244 | struct nexthop *route_entry_nexthop_ipv4_ifindex_add(struct route_entry *re, |
245 | struct in_addr *ipv4, | |
246 | struct in_addr *src, | |
247 | ifindex_t ifindex) | |
718e3744 | 248 | { |
d62a17ae | 249 | struct nexthop *nexthop; |
250 | struct interface *ifp; | |
718e3744 | 251 | |
d62a17ae | 252 | nexthop = nexthop_new(); |
253 | nexthop->type = NEXTHOP_TYPE_IPV4_IFINDEX; | |
254 | nexthop->gate.ipv4 = *ipv4; | |
255 | if (src) | |
256 | nexthop->src.ipv4 = *src; | |
257 | nexthop->ifindex = ifindex; | |
eec2e592 | 258 | ifp = if_lookup_by_index(nexthop->ifindex, re->vrf_id); |
d62a17ae | 259 | /*Pending: need to think if null ifp here is ok during bootup? |
260 | There was a crash because ifp here was coming to be NULL */ | |
261 | if (ifp) | |
262 | if (connected_is_unnumbered(ifp)) { | |
263 | SET_FLAG(nexthop->flags, NEXTHOP_FLAG_ONLINK); | |
264 | } | |
718e3744 | 265 | |
d62a17ae | 266 | route_entry_nexthop_add(re, nexthop); |
718e3744 | 267 | |
d62a17ae | 268 | return nexthop; |
718e3744 | 269 | } |
270 | ||
d62a17ae | 271 | struct nexthop *route_entry_nexthop_ipv6_add(struct route_entry *re, |
272 | struct in6_addr *ipv6) | |
718e3744 | 273 | { |
d62a17ae | 274 | struct nexthop *nexthop; |
718e3744 | 275 | |
d62a17ae | 276 | nexthop = nexthop_new(); |
277 | nexthop->type = NEXTHOP_TYPE_IPV6; | |
278 | nexthop->gate.ipv6 = *ipv6; | |
718e3744 | 279 | |
d62a17ae | 280 | route_entry_nexthop_add(re, nexthop); |
718e3744 | 281 | |
d62a17ae | 282 | return nexthop; |
718e3744 | 283 | } |
284 | ||
d62a17ae | 285 | struct nexthop *route_entry_nexthop_ipv6_ifindex_add(struct route_entry *re, |
286 | struct in6_addr *ipv6, | |
287 | ifindex_t ifindex) | |
718e3744 | 288 | { |
d62a17ae | 289 | struct nexthop *nexthop; |
718e3744 | 290 | |
d62a17ae | 291 | nexthop = nexthop_new(); |
292 | nexthop->type = NEXTHOP_TYPE_IPV6_IFINDEX; | |
293 | nexthop->gate.ipv6 = *ipv6; | |
294 | nexthop->ifindex = ifindex; | |
718e3744 | 295 | |
d62a17ae | 296 | route_entry_nexthop_add(re, nexthop); |
718e3744 | 297 | |
d62a17ae | 298 | return nexthop; |
718e3744 | 299 | } |
718e3744 | 300 | |
a8309422 | 301 | struct nexthop *route_entry_nexthop_blackhole_add(struct route_entry *re, |
60466a63 | 302 | enum blackhole_type bh_type) |
595db7f1 | 303 | { |
d62a17ae | 304 | struct nexthop *nexthop; |
595db7f1 | 305 | |
d62a17ae | 306 | nexthop = nexthop_new(); |
307 | nexthop->type = NEXTHOP_TYPE_BLACKHOLE; | |
a8309422 | 308 | nexthop->bh_type = bh_type; |
595db7f1 | 309 | |
d62a17ae | 310 | route_entry_nexthop_add(re, nexthop); |
595db7f1 | 311 | |
d62a17ae | 312 | return nexthop; |
595db7f1 | 313 | } |
314 | ||
d62a17ae | 315 | static void nexthop_set_resolved(afi_t afi, struct nexthop *newhop, |
316 | struct nexthop *nexthop) | |
4491a88f | 317 | { |
d62a17ae | 318 | struct nexthop *resolved_hop; |
4491a88f | 319 | |
d62a17ae | 320 | resolved_hop = nexthop_new(); |
321 | SET_FLAG(resolved_hop->flags, NEXTHOP_FLAG_ACTIVE); | |
d855d11f RW |
322 | |
323 | switch (newhop->type) { | |
324 | case NEXTHOP_TYPE_IPV4: | |
325 | case NEXTHOP_TYPE_IPV4_IFINDEX: | |
326 | /* If the resolving route specifies a gateway, use it */ | |
d62a17ae | 327 | resolved_hop->type = newhop->type; |
328 | resolved_hop->gate.ipv4 = newhop->gate.ipv4; | |
4491a88f | 329 | |
d62a17ae | 330 | if (newhop->ifindex) { |
331 | resolved_hop->type = NEXTHOP_TYPE_IPV4_IFINDEX; | |
332 | resolved_hop->ifindex = newhop->ifindex; | |
333 | if (newhop->flags & NEXTHOP_FLAG_ONLINK) | |
334 | resolved_hop->flags |= NEXTHOP_FLAG_ONLINK; | |
335 | } | |
d855d11f RW |
336 | break; |
337 | case NEXTHOP_TYPE_IPV6: | |
338 | case NEXTHOP_TYPE_IPV6_IFINDEX: | |
d62a17ae | 339 | resolved_hop->type = newhop->type; |
340 | resolved_hop->gate.ipv6 = newhop->gate.ipv6; | |
341 | ||
342 | if (newhop->ifindex) { | |
343 | resolved_hop->type = NEXTHOP_TYPE_IPV6_IFINDEX; | |
344 | resolved_hop->ifindex = newhop->ifindex; | |
345 | } | |
d855d11f RW |
346 | break; |
347 | case NEXTHOP_TYPE_IFINDEX: | |
348 | /* If the resolving route is an interface route, | |
349 | * it means the gateway we are looking up is connected | |
350 | * to that interface. (The actual network is _not_ onlink). | |
351 | * Therefore, the resolved route should have the original | |
352 | * gateway as nexthop as it is directly connected. | |
353 | * | |
354 | * On Linux, we have to set the onlink netlink flag because | |
355 | * otherwise, the kernel won't accept the route. | |
356 | */ | |
d62a17ae | 357 | resolved_hop->flags |= NEXTHOP_FLAG_ONLINK; |
358 | if (afi == AFI_IP) { | |
359 | resolved_hop->type = NEXTHOP_TYPE_IPV4_IFINDEX; | |
360 | resolved_hop->gate.ipv4 = nexthop->gate.ipv4; | |
361 | } else if (afi == AFI_IP6) { | |
362 | resolved_hop->type = NEXTHOP_TYPE_IPV6_IFINDEX; | |
363 | resolved_hop->gate.ipv6 = nexthop->gate.ipv6; | |
364 | } | |
365 | resolved_hop->ifindex = newhop->ifindex; | |
d855d11f RW |
366 | break; |
367 | case NEXTHOP_TYPE_BLACKHOLE: | |
59693377 DS |
368 | resolved_hop->type = NEXTHOP_TYPE_BLACKHOLE; |
369 | resolved_hop->bh_type = nexthop->bh_type; | |
d855d11f | 370 | break; |
59693377 | 371 | } |
d855d11f | 372 | |
d62a17ae | 373 | resolved_hop->rparent = nexthop; |
374 | nexthop_add(&nexthop->resolved, resolved_hop); | |
4491a88f DS |
375 | } |
376 | ||
718e3744 | 377 | /* If force flag is not set, do not modify falgs at all for uninstall |
378 | the route from FIB. */ | |
d62a17ae | 379 | static int nexthop_active(afi_t afi, struct route_entry *re, |
380 | struct nexthop *nexthop, int set, | |
381 | struct route_node *top) | |
382 | { | |
383 | struct prefix p; | |
384 | struct route_table *table; | |
385 | struct route_node *rn; | |
386 | struct route_entry *match; | |
387 | int resolved; | |
388 | struct nexthop *newhop; | |
389 | struct interface *ifp; | |
390 | ||
391 | if ((nexthop->type == NEXTHOP_TYPE_IPV4) | |
392 | || nexthop->type == NEXTHOP_TYPE_IPV6) | |
393 | nexthop->ifindex = 0; | |
394 | ||
395 | if (set) { | |
396 | UNSET_FLAG(nexthop->flags, NEXTHOP_FLAG_RECURSIVE); | |
397 | zebra_deregister_rnh_static_nexthops(re->vrf_id, | |
398 | nexthop->resolved, top); | |
399 | nexthops_free(nexthop->resolved); | |
400 | nexthop->resolved = NULL; | |
401 | re->nexthop_mtu = 0; | |
d44ca835 | 402 | } |
18ff3edd | 403 | |
d62a17ae | 404 | /* Skip nexthops that have been filtered out due to route-map */ |
405 | /* The nexthops are specific to this route and so the same */ | |
406 | /* nexthop for a different route may not have this flag set */ | |
407 | if (CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_FILTERED)) | |
408 | return 0; | |
409 | ||
410 | /* | |
411 | * Check to see if we should trust the passed in information | |
412 | * for UNNUMBERED interfaces as that we won't find the GW | |
413 | * address in the routing table. | |
414 | */ | |
415 | if (CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_ONLINK)) { | |
eec2e592 | 416 | ifp = if_lookup_by_index(nexthop->ifindex, re->vrf_id); |
d62a17ae | 417 | if (ifp && connected_is_unnumbered(ifp)) { |
418 | if (if_is_operative(ifp)) | |
419 | return 1; | |
420 | else | |
421 | return 0; | |
422 | } else | |
423 | return 0; | |
16814f96 | 424 | } |
718e3744 | 425 | |
d62a17ae | 426 | /* Make lookup prefix. */ |
427 | memset(&p, 0, sizeof(struct prefix)); | |
428 | switch (afi) { | |
429 | case AFI_IP: | |
430 | p.family = AF_INET; | |
431 | p.prefixlen = IPV4_MAX_PREFIXLEN; | |
432 | p.u.prefix4 = nexthop->gate.ipv4; | |
433 | break; | |
434 | case AFI_IP6: | |
435 | p.family = AF_INET6; | |
436 | p.prefixlen = IPV6_MAX_PREFIXLEN; | |
437 | p.u.prefix6 = nexthop->gate.ipv6; | |
438 | break; | |
439 | default: | |
440 | assert(afi != AFI_IP && afi != AFI_IP6); | |
441 | break; | |
718e3744 | 442 | } |
d62a17ae | 443 | /* Lookup table. */ |
444 | table = zebra_vrf_table(afi, SAFI_UNICAST, re->vrf_id); | |
445 | if (!table) | |
446 | return 0; | |
447 | ||
448 | rn = route_node_match(table, (struct prefix *)&p); | |
449 | while (rn) { | |
450 | route_unlock_node(rn); | |
451 | ||
452 | /* If lookup self prefix return immediately. */ | |
453 | if (rn == top) | |
454 | return 0; | |
455 | ||
456 | /* Pick up selected route. */ | |
457 | /* However, do not resolve over default route unless explicitly | |
458 | * allowed. */ | |
459 | if (is_default_prefix(&rn->p) | |
460 | && !nh_resolve_via_default(p.family)) | |
461 | return 0; | |
462 | ||
a2addae8 | 463 | RNODE_FOREACH_RE (rn, match) { |
d62a17ae | 464 | if (CHECK_FLAG(match->status, ROUTE_ENTRY_REMOVED)) |
465 | continue; | |
466 | ||
467 | /* if the next hop is imported from another table, skip | |
468 | * it */ | |
469 | if (match->type == ZEBRA_ROUTE_TABLE) | |
470 | continue; | |
471 | if (CHECK_FLAG(match->status, ROUTE_ENTRY_SELECTED_FIB)) | |
472 | break; | |
473 | } | |
48a53dc7 | 474 | |
d62a17ae | 475 | /* If there is no selected route or matched route is EGP, go up |
476 | tree. */ | |
477 | if (!match) { | |
478 | do { | |
479 | rn = rn->parent; | |
480 | } while (rn && rn->info == NULL); | |
481 | if (rn) | |
482 | route_lock_node(rn); | |
c87bdd2b | 483 | |
d62a17ae | 484 | continue; |
485 | } | |
486 | ||
d62a17ae | 487 | if (match->type == ZEBRA_ROUTE_CONNECT) { |
488 | /* Directly point connected route. */ | |
489 | newhop = match->nexthop; | |
490 | if (newhop) { | |
491 | if (nexthop->type == NEXTHOP_TYPE_IPV4 | |
492 | || nexthop->type == NEXTHOP_TYPE_IPV6) | |
493 | nexthop->ifindex = newhop->ifindex; | |
494 | } | |
495 | return 1; | |
496 | } else if (CHECK_FLAG(re->flags, ZEBRA_FLAG_INTERNAL)) { | |
497 | resolved = 0; | |
a8309422 | 498 | for (ALL_NEXTHOPS(match->nexthop, newhop)) { |
a8309422 DL |
499 | if (!CHECK_FLAG(newhop->flags, |
500 | NEXTHOP_FLAG_FIB)) | |
501 | continue; | |
502 | if (CHECK_FLAG(newhop->flags, | |
503 | NEXTHOP_FLAG_RECURSIVE)) | |
504 | continue; | |
505 | ||
506 | if (set) { | |
507 | SET_FLAG(nexthop->flags, | |
508 | NEXTHOP_FLAG_RECURSIVE); | |
509 | SET_FLAG(re->status, | |
510 | ROUTE_ENTRY_NEXTHOPS_CHANGED); | |
511 | nexthop_set_resolved(afi, newhop, | |
512 | nexthop); | |
d62a17ae | 513 | } |
a8309422 DL |
514 | resolved = 1; |
515 | } | |
43e31305 JB |
516 | if (resolved && set) |
517 | re->nexthop_mtu = match->mtu; | |
d62a17ae | 518 | return resolved; |
519 | } else if (re->type == ZEBRA_ROUTE_STATIC) { | |
520 | resolved = 0; | |
a8309422 | 521 | for (ALL_NEXTHOPS(match->nexthop, newhop)) { |
a8309422 DL |
522 | if (!CHECK_FLAG(newhop->flags, |
523 | NEXTHOP_FLAG_FIB)) | |
524 | continue; | |
525 | ||
526 | if (set) { | |
527 | SET_FLAG(nexthop->flags, | |
528 | NEXTHOP_FLAG_RECURSIVE); | |
529 | nexthop_set_resolved(afi, newhop, | |
530 | nexthop); | |
d62a17ae | 531 | } |
a8309422 DL |
532 | resolved = 1; |
533 | } | |
d62a17ae | 534 | if (resolved && set) |
535 | re->nexthop_mtu = match->mtu; | |
536 | return resolved; | |
537 | } else { | |
538 | return 0; | |
539 | } | |
718e3744 | 540 | } |
d62a17ae | 541 | return 0; |
718e3744 | 542 | } |
718e3744 | 543 | |
d62a17ae | 544 | struct route_entry *rib_match(afi_t afi, safi_t safi, vrf_id_t vrf_id, |
545 | union g_addr *addr, struct route_node **rn_out) | |
546 | { | |
547 | struct prefix p; | |
548 | struct route_table *table; | |
549 | struct route_node *rn; | |
550 | struct route_entry *match; | |
551 | struct nexthop *newhop; | |
552 | ||
553 | /* Lookup table. */ | |
554 | table = zebra_vrf_table(afi, safi, vrf_id); | |
555 | if (!table) | |
556 | return 0; | |
557 | ||
558 | memset(&p, 0, sizeof(struct prefix)); | |
559 | p.family = afi; | |
560 | if (afi == AFI_IP) { | |
561 | p.u.prefix4 = addr->ipv4; | |
562 | p.prefixlen = IPV4_MAX_PREFIXLEN; | |
563 | } else { | |
564 | p.u.prefix6 = addr->ipv6; | |
565 | p.prefixlen = IPV6_MAX_PREFIXLEN; | |
566 | } | |
718e3744 | 567 | |
d62a17ae | 568 | rn = route_node_match(table, (struct prefix *)&p); |
718e3744 | 569 | |
d62a17ae | 570 | while (rn) { |
571 | route_unlock_node(rn); | |
718e3744 | 572 | |
d62a17ae | 573 | /* Pick up selected route. */ |
a2addae8 | 574 | RNODE_FOREACH_RE (rn, match) { |
d62a17ae | 575 | if (CHECK_FLAG(match->status, ROUTE_ENTRY_REMOVED)) |
576 | continue; | |
577 | if (CHECK_FLAG(match->status, ROUTE_ENTRY_SELECTED_FIB)) | |
578 | break; | |
579 | } | |
718e3744 | 580 | |
d62a17ae | 581 | /* If there is no selected route or matched route is EGP, go up |
582 | tree. */ | |
583 | if (!match) { | |
584 | do { | |
585 | rn = rn->parent; | |
586 | } while (rn && rn->info == NULL); | |
587 | if (rn) | |
588 | route_lock_node(rn); | |
589 | } else { | |
590 | if (match->type != ZEBRA_ROUTE_CONNECT) { | |
591 | int found = 0; | |
592 | for (ALL_NEXTHOPS(match->nexthop, newhop)) | |
593 | if (CHECK_FLAG(newhop->flags, | |
594 | NEXTHOP_FLAG_FIB)) { | |
595 | found = 1; | |
596 | break; | |
597 | } | |
598 | if (!found) | |
599 | return NULL; | |
600 | } | |
601 | ||
602 | if (rn_out) | |
603 | *rn_out = rn; | |
604 | return match; | |
605 | } | |
16814f96 | 606 | } |
d62a17ae | 607 | return NULL; |
608 | } | |
609 | ||
610 | struct route_entry *rib_match_ipv4_multicast(vrf_id_t vrf_id, | |
611 | struct in_addr addr, | |
612 | struct route_node **rn_out) | |
613 | { | |
614 | struct route_entry *re = NULL, *mre = NULL, *ure = NULL; | |
615 | struct route_node *m_rn = NULL, *u_rn = NULL; | |
616 | union g_addr gaddr = {.ipv4 = addr}; | |
617 | ||
618 | switch (ipv4_multicast_mode) { | |
619 | case MCAST_MRIB_ONLY: | |
620 | return rib_match(AFI_IP, SAFI_MULTICAST, vrf_id, &gaddr, | |
621 | rn_out); | |
622 | case MCAST_URIB_ONLY: | |
623 | return rib_match(AFI_IP, SAFI_UNICAST, vrf_id, &gaddr, rn_out); | |
624 | case MCAST_NO_CONFIG: | |
625 | case MCAST_MIX_MRIB_FIRST: | |
626 | re = mre = rib_match(AFI_IP, SAFI_MULTICAST, vrf_id, &gaddr, | |
627 | &m_rn); | |
628 | if (!mre) | |
629 | re = ure = rib_match(AFI_IP, SAFI_UNICAST, vrf_id, | |
630 | &gaddr, &u_rn); | |
631 | break; | |
632 | case MCAST_MIX_DISTANCE: | |
633 | mre = rib_match(AFI_IP, SAFI_MULTICAST, vrf_id, &gaddr, &m_rn); | |
634 | ure = rib_match(AFI_IP, SAFI_UNICAST, vrf_id, &gaddr, &u_rn); | |
635 | if (mre && ure) | |
636 | re = ure->distance < mre->distance ? ure : mre; | |
637 | else if (mre) | |
638 | re = mre; | |
639 | else if (ure) | |
640 | re = ure; | |
641 | break; | |
642 | case MCAST_MIX_PFXLEN: | |
643 | mre = rib_match(AFI_IP, SAFI_MULTICAST, vrf_id, &gaddr, &m_rn); | |
644 | ure = rib_match(AFI_IP, SAFI_UNICAST, vrf_id, &gaddr, &u_rn); | |
645 | if (mre && ure) | |
646 | re = u_rn->p.prefixlen > m_rn->p.prefixlen ? ure : mre; | |
647 | else if (mre) | |
648 | re = mre; | |
649 | else if (ure) | |
650 | re = ure; | |
651 | break; | |
718e3744 | 652 | } |
33550aa8 | 653 | |
d62a17ae | 654 | if (rn_out) |
655 | *rn_out = (re == mre) ? m_rn : u_rn; | |
718e3744 | 656 | |
d62a17ae | 657 | if (IS_ZEBRA_DEBUG_RIB) { |
658 | char buf[BUFSIZ]; | |
659 | inet_ntop(AF_INET, &addr, buf, BUFSIZ); | |
4623d897 | 660 | |
d62a17ae | 661 | zlog_debug("%s: %s: found %s, using %s", __func__, buf, |
662 | mre ? (ure ? "MRIB+URIB" : "MRIB") | |
663 | : ure ? "URIB" : "nothing", | |
664 | re == ure ? "URIB" : re == mre ? "MRIB" : "none"); | |
665 | } | |
666 | return re; | |
4623d897 DL |
667 | } |
668 | ||
d62a17ae | 669 | void multicast_mode_ipv4_set(enum multicast_mode mode) |
4623d897 | 670 | { |
d62a17ae | 671 | if (IS_ZEBRA_DEBUG_RIB) |
672 | zlog_debug("%s: multicast lookup mode set (%d)", __func__, | |
673 | mode); | |
674 | ipv4_multicast_mode = mode; | |
4623d897 DL |
675 | } |
676 | ||
d62a17ae | 677 | enum multicast_mode multicast_mode_ipv4_get(void) |
4623d897 | 678 | { |
d62a17ae | 679 | return ipv4_multicast_mode; |
4623d897 DL |
680 | } |
681 | ||
d62a17ae | 682 | struct route_entry *rib_lookup_ipv4(struct prefix_ipv4 *p, vrf_id_t vrf_id) |
718e3744 | 683 | { |
d62a17ae | 684 | struct route_table *table; |
685 | struct route_node *rn; | |
686 | struct route_entry *match; | |
687 | struct nexthop *nexthop; | |
718e3744 | 688 | |
d62a17ae | 689 | /* Lookup table. */ |
690 | table = zebra_vrf_table(AFI_IP, SAFI_UNICAST, vrf_id); | |
691 | if (!table) | |
692 | return 0; | |
718e3744 | 693 | |
d62a17ae | 694 | rn = route_node_lookup(table, (struct prefix *)p); |
718e3744 | 695 | |
d62a17ae | 696 | /* No route for this prefix. */ |
697 | if (!rn) | |
698 | return NULL; | |
718e3744 | 699 | |
d62a17ae | 700 | /* Unlock node. */ |
701 | route_unlock_node(rn); | |
718e3744 | 702 | |
a2addae8 | 703 | RNODE_FOREACH_RE (rn, match) { |
d62a17ae | 704 | if (CHECK_FLAG(match->status, ROUTE_ENTRY_REMOVED)) |
705 | continue; | |
706 | if (CHECK_FLAG(match->status, ROUTE_ENTRY_SELECTED_FIB)) | |
707 | break; | |
708 | } | |
718e3744 | 709 | |
d62a17ae | 710 | if (!match) |
711 | return NULL; | |
718e3744 | 712 | |
d62a17ae | 713 | if (match->type == ZEBRA_ROUTE_CONNECT) |
714 | return match; | |
f9e1b38e | 715 | |
d62a17ae | 716 | for (ALL_NEXTHOPS(match->nexthop, nexthop)) |
717 | if (CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_FIB)) | |
718 | return match; | |
718e3744 | 719 | |
d62a17ae | 720 | return NULL; |
718e3744 | 721 | } |
722 | ||
dc95824a DO |
723 | /* |
724 | * This clone function, unlike its original rib_lookup_ipv4(), checks | |
725 | * if specified IPv4 route record (prefix/mask -> gate) exists in | |
f0f77c9a | 726 | * the whole RIB and has ROUTE_ENTRY_SELECTED_FIB set. |
dc95824a DO |
727 | * |
728 | * Return values: | |
729 | * -1: error | |
730 | * 0: exact match found | |
731 | * 1: a match was found with a different gate | |
732 | * 2: connected route found | |
733 | * 3: no matches found | |
734 | */ | |
d62a17ae | 735 | int rib_lookup_ipv4_route(struct prefix_ipv4 *p, union sockunion *qgate, |
736 | vrf_id_t vrf_id) | |
dc95824a | 737 | { |
d62a17ae | 738 | struct route_table *table; |
739 | struct route_node *rn; | |
740 | struct route_entry *match; | |
741 | struct nexthop *nexthop; | |
742 | int nexthops_active; | |
dc95824a | 743 | |
d62a17ae | 744 | /* Lookup table. */ |
745 | table = zebra_vrf_table(AFI_IP, SAFI_UNICAST, vrf_id); | |
746 | if (!table) | |
747 | return ZEBRA_RIB_LOOKUP_ERROR; | |
dc95824a | 748 | |
d62a17ae | 749 | /* Scan the RIB table for exactly matching RIB entry. */ |
750 | rn = route_node_lookup(table, (struct prefix *)p); | |
dc95824a | 751 | |
d62a17ae | 752 | /* No route for this prefix. */ |
753 | if (!rn) | |
754 | return ZEBRA_RIB_NOTFOUND; | |
dc95824a | 755 | |
d62a17ae | 756 | /* Unlock node. */ |
757 | route_unlock_node(rn); | |
dc95824a | 758 | |
d62a17ae | 759 | /* Find out if a "selected" RR for the discovered RIB entry exists ever. |
760 | */ | |
a2addae8 | 761 | RNODE_FOREACH_RE (rn, match) { |
d62a17ae | 762 | if (CHECK_FLAG(match->status, ROUTE_ENTRY_REMOVED)) |
763 | continue; | |
764 | if (CHECK_FLAG(match->status, ROUTE_ENTRY_SELECTED_FIB)) | |
765 | break; | |
766 | } | |
dc95824a | 767 | |
d62a17ae | 768 | /* None such found :( */ |
769 | if (!match) | |
770 | return ZEBRA_RIB_NOTFOUND; | |
771 | ||
772 | if (match->type == ZEBRA_ROUTE_CONNECT) | |
773 | return ZEBRA_RIB_FOUND_CONNECTED; | |
774 | ||
775 | /* Ok, we have a cood candidate, let's check it's nexthop list... */ | |
776 | nexthops_active = 0; | |
777 | for (ALL_NEXTHOPS(match->nexthop, nexthop)) | |
778 | if (CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_FIB)) { | |
779 | nexthops_active = 1; | |
780 | if (nexthop->gate.ipv4.s_addr == sockunion2ip(qgate)) | |
781 | return ZEBRA_RIB_FOUND_EXACT; | |
782 | if (IS_ZEBRA_DEBUG_RIB) { | |
783 | char gate_buf[INET_ADDRSTRLEN], | |
784 | qgate_buf[INET_ADDRSTRLEN]; | |
785 | inet_ntop(AF_INET, &nexthop->gate.ipv4.s_addr, | |
786 | gate_buf, INET_ADDRSTRLEN); | |
787 | inet_ntop(AF_INET, &sockunion2ip(qgate), | |
788 | qgate_buf, INET_ADDRSTRLEN); | |
789 | zlog_debug("%s: qgate == %s, %s == %s", | |
790 | __func__, qgate_buf, | |
791 | nexthop->rparent ? "rgate" : "gate", | |
792 | gate_buf); | |
793 | } | |
794 | } | |
795 | ||
796 | if (nexthops_active) | |
797 | return ZEBRA_RIB_FOUND_NOGATE; | |
798 | ||
799 | return ZEBRA_RIB_NOTFOUND; | |
800 | } | |
801 | ||
802 | #define RIB_SYSTEM_ROUTE(R) \ | |
803 | ((R)->type == ZEBRA_ROUTE_KERNEL || (R)->type == ZEBRA_ROUTE_CONNECT) | |
7514fb77 | 804 | |
dc95824a DO |
805 | /* This function verifies reachability of one given nexthop, which can be |
806 | * numbered or unnumbered, IPv4 or IPv6. The result is unconditionally stored | |
807 | * in nexthop->flags field. If the 4th parameter, 'set', is non-zero, | |
808 | * nexthop->ifindex will be updated appropriately as well. | |
809 | * An existing route map can turn (otherwise active) nexthop into inactive, but | |
810 | * not vice versa. | |
811 | * | |
812 | * The return value is the final value of 'ACTIVE' flag. | |
813 | */ | |
814 | ||
d62a17ae | 815 | static unsigned nexthop_active_check(struct route_node *rn, |
816 | struct route_entry *re, | |
817 | struct nexthop *nexthop, int set) | |
818 | { | |
819 | struct interface *ifp; | |
820 | route_map_result_t ret = RMAP_MATCH; | |
821 | int family; | |
822 | char buf[SRCDEST2STR_BUFFER]; | |
823 | struct prefix *p, *src_p; | |
824 | srcdest_rnode_prefixes(rn, &p, &src_p); | |
825 | ||
826 | if (rn->p.family == AF_INET) | |
827 | family = AFI_IP; | |
828 | else if (rn->p.family == AF_INET6) | |
829 | family = AFI_IP6; | |
830 | else | |
831 | family = 0; | |
832 | switch (nexthop->type) { | |
833 | case NEXTHOP_TYPE_IFINDEX: | |
834 | ifp = if_lookup_by_index(nexthop->ifindex, re->vrf_id); | |
835 | if (ifp && if_is_operative(ifp)) | |
836 | SET_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE); | |
837 | else | |
838 | UNSET_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE); | |
839 | break; | |
840 | case NEXTHOP_TYPE_IPV4: | |
841 | case NEXTHOP_TYPE_IPV4_IFINDEX: | |
842 | family = AFI_IP; | |
843 | if (nexthop_active(AFI_IP, re, nexthop, set, rn)) | |
844 | SET_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE); | |
845 | else | |
846 | UNSET_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE); | |
847 | break; | |
848 | case NEXTHOP_TYPE_IPV6: | |
849 | family = AFI_IP6; | |
850 | if (nexthop_active(AFI_IP6, re, nexthop, set, rn)) | |
851 | SET_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE); | |
852 | else | |
853 | UNSET_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE); | |
854 | break; | |
855 | case NEXTHOP_TYPE_IPV6_IFINDEX: | |
856 | /* RFC 5549, v4 prefix with v6 NH */ | |
857 | if (rn->p.family != AF_INET) | |
858 | family = AFI_IP6; | |
859 | if (IN6_IS_ADDR_LINKLOCAL(&nexthop->gate.ipv6)) { | |
860 | ifp = if_lookup_by_index(nexthop->ifindex, re->vrf_id); | |
861 | if (ifp && if_is_operative(ifp)) | |
862 | SET_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE); | |
863 | else | |
864 | UNSET_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE); | |
865 | } else { | |
866 | if (nexthop_active(AFI_IP6, re, nexthop, set, rn)) | |
867 | SET_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE); | |
868 | else | |
869 | UNSET_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE); | |
870 | } | |
871 | break; | |
872 | case NEXTHOP_TYPE_BLACKHOLE: | |
873 | SET_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE); | |
874 | break; | |
875 | default: | |
876 | break; | |
877 | } | |
878 | if (!CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE)) | |
879 | return 0; | |
880 | ||
881 | /* XXX: What exactly do those checks do? Do we support | |
882 | * e.g. IPv4 routes with IPv6 nexthops or vice versa? */ | |
883 | if (RIB_SYSTEM_ROUTE(re) || (family == AFI_IP && p->family != AF_INET) | |
884 | || (family == AFI_IP6 && p->family != AF_INET6)) | |
885 | return CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE); | |
886 | ||
887 | /* The original code didn't determine the family correctly | |
888 | * e.g. for NEXTHOP_TYPE_IFINDEX. Retrieve the correct afi | |
889 | * from the rib_table_info in those cases. | |
890 | * Possibly it may be better to use only the rib_table_info | |
891 | * in every case. | |
892 | */ | |
893 | if (!family) { | |
894 | rib_table_info_t *info; | |
895 | ||
896 | info = srcdest_rnode_table_info(rn); | |
897 | family = info->afi; | |
718e3744 | 898 | } |
c52ef59f | 899 | |
d62a17ae | 900 | memset(&nexthop->rmap_src.ipv6, 0, sizeof(union g_addr)); |
901 | ||
902 | /* It'll get set if required inside */ | |
903 | ret = zebra_route_map_check(family, re->type, p, nexthop, re->vrf_id, | |
904 | re->tag); | |
905 | if (ret == RMAP_DENYMATCH) { | |
906 | if (IS_ZEBRA_DEBUG_RIB) { | |
907 | srcdest_rnode2str(rn, buf, sizeof(buf)); | |
908 | zlog_debug( | |
909 | "%u:%s: Filtering out with NH out %s due to route map", | |
910 | re->vrf_id, buf, | |
911 | ifindex2ifname(nexthop->ifindex, re->vrf_id)); | |
912 | } | |
913 | UNSET_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE); | |
518f0eb1 | 914 | } |
d62a17ae | 915 | return CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE); |
718e3744 | 916 | } |
917 | ||
03e232a4 | 918 | /* Iterate over all nexthops of the given RIB entry and refresh their |
f0f77c9a DS |
919 | * ACTIVE flag. re->nexthop_active_num is updated accordingly. If any |
920 | * nexthop is found to toggle the ACTIVE flag, the whole re structure | |
921 | * is flagged with ROUTE_ENTRY_CHANGED. The 4th 'set' argument is | |
03e232a4 DO |
922 | * transparently passed to nexthop_active_check(). |
923 | * | |
924 | * Return value is the new number of active nexthops. | |
925 | */ | |
926 | ||
d62a17ae | 927 | static int nexthop_active_update(struct route_node *rn, struct route_entry *re, |
928 | int set) | |
929 | { | |
930 | struct nexthop *nexthop; | |
931 | union g_addr prev_src; | |
932 | unsigned int prev_active, new_active, old_num_nh; | |
933 | ifindex_t prev_index; | |
934 | old_num_nh = re->nexthop_active_num; | |
935 | ||
936 | re->nexthop_active_num = 0; | |
937 | UNSET_FLAG(re->status, ROUTE_ENTRY_CHANGED); | |
938 | ||
939 | for (nexthop = re->nexthop; nexthop; nexthop = nexthop->next) { | |
940 | /* No protocol daemon provides src and so we're skipping | |
941 | * tracking it */ | |
942 | prev_src = nexthop->rmap_src; | |
943 | prev_active = CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE); | |
944 | prev_index = nexthop->ifindex; | |
945 | if ((new_active = nexthop_active_check(rn, re, nexthop, set))) | |
946 | re->nexthop_active_num++; | |
947 | /* Don't allow src setting on IPv6 addr for now */ | |
948 | if (prev_active != new_active || prev_index != nexthop->ifindex | |
949 | || ((nexthop->type >= NEXTHOP_TYPE_IFINDEX | |
950 | && nexthop->type < NEXTHOP_TYPE_IPV6) | |
951 | && prev_src.ipv4.s_addr | |
952 | != nexthop->rmap_src.ipv4.s_addr) | |
953 | || ((nexthop->type >= NEXTHOP_TYPE_IPV6 | |
954 | && nexthop->type < NEXTHOP_TYPE_BLACKHOLE) | |
955 | && !(IPV6_ADDR_SAME(&prev_src.ipv6, | |
956 | &nexthop->rmap_src.ipv6)))) { | |
957 | SET_FLAG(re->status, ROUTE_ENTRY_CHANGED); | |
958 | SET_FLAG(re->status, ROUTE_ENTRY_NEXTHOPS_CHANGED); | |
959 | } | |
960 | } | |
961 | ||
962 | if (old_num_nh != re->nexthop_active_num) | |
963 | SET_FLAG(re->status, ROUTE_ENTRY_CHANGED); | |
6e26278c | 964 | |
d62a17ae | 965 | if (CHECK_FLAG(re->status, ROUTE_ENTRY_CHANGED)) { |
966 | SET_FLAG(re->status, ROUTE_ENTRY_NEXTHOPS_CHANGED); | |
967 | } | |
968 | ||
969 | return re->nexthop_active_num; | |
718e3744 | 970 | } |
6baeb988 | 971 | |
a64448ba DS |
972 | /* |
973 | * Is this RIB labeled-unicast? It must be of type BGP and all paths | |
974 | * (nexthops) must have a label. | |
975 | */ | |
d62a17ae | 976 | int zebra_rib_labeled_unicast(struct route_entry *re) |
a64448ba | 977 | { |
d62a17ae | 978 | struct nexthop *nexthop = NULL; |
a64448ba | 979 | |
d62a17ae | 980 | if (re->type != ZEBRA_ROUTE_BGP) |
981 | return 0; | |
a64448ba | 982 | |
d62a17ae | 983 | for (ALL_NEXTHOPS(re->nexthop, nexthop)) |
984 | if (!nexthop->nh_label || !nexthop->nh_label->num_labels) | |
985 | return 0; | |
6b0655a2 | 986 | |
d62a17ae | 987 | return 1; |
a64448ba | 988 | } |
718e3744 | 989 | |
6ae24471 DS |
990 | /* Update flag indicates whether this is a "replace" or not. Currently, this |
991 | * is only used for IPv4. | |
992 | */ | |
d62a17ae | 993 | int rib_install_kernel(struct route_node *rn, struct route_entry *re, |
994 | struct route_entry *old) | |
718e3744 | 995 | { |
d62a17ae | 996 | int ret = 0; |
997 | struct nexthop *nexthop; | |
998 | rib_table_info_t *info = srcdest_rnode_table_info(rn); | |
999 | struct prefix *p, *src_p; | |
1000 | struct zebra_vrf *zvrf = vrf_info_lookup(re->vrf_id); | |
718e3744 | 1001 | |
d62a17ae | 1002 | srcdest_rnode_prefixes(rn, &p, &src_p); |
416ec78d | 1003 | |
d62a17ae | 1004 | if (info->safi != SAFI_UNICAST) { |
1005 | for (ALL_NEXTHOPS(re->nexthop, nexthop)) | |
1006 | SET_FLAG(nexthop->flags, NEXTHOP_FLAG_FIB); | |
1007 | return ret; | |
25b9cb0c DL |
1008 | } else { |
1009 | struct nexthop *prev; | |
1010 | ||
1011 | for (ALL_NEXTHOPS(re->nexthop, nexthop)) { | |
1012 | UNSET_FLAG (nexthop->flags, NEXTHOP_FLAG_DUPLICATE); | |
1013 | for (ALL_NEXTHOPS(re->nexthop, prev)) { | |
1014 | if (prev == nexthop) | |
1015 | break; | |
1016 | if (nexthop_same_firsthop (nexthop, prev)) | |
1017 | { | |
1018 | SET_FLAG (nexthop->flags, NEXTHOP_FLAG_DUPLICATE); | |
1019 | break; | |
1020 | } | |
1021 | } | |
1022 | } | |
d62a17ae | 1023 | } |
718e3744 | 1024 | |
25b9cb0c | 1025 | |
d62a17ae | 1026 | /* |
1027 | * Make sure we update the FPM any time we send new information to | |
1028 | * the kernel. | |
1029 | */ | |
1030 | hook_call(rib_update, rn, "installing in kernel"); | |
1031 | ret = kernel_route_rib(p, src_p, old, re); | |
1032 | zvrf->installs++; | |
1033 | ||
1034 | /* If install succeeds, update FIB flag for nexthops. */ | |
1035 | if (!ret) { | |
1036 | for (ALL_NEXTHOPS(re->nexthop, nexthop)) { | |
1037 | if (CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_RECURSIVE)) | |
1038 | continue; | |
1039 | ||
1040 | if (CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE)) | |
1041 | SET_FLAG(nexthop->flags, NEXTHOP_FLAG_FIB); | |
1042 | else | |
1043 | UNSET_FLAG(nexthop->flags, NEXTHOP_FLAG_FIB); | |
1044 | } | |
1045 | } | |
3e5c6e00 | 1046 | |
d62a17ae | 1047 | return ret; |
718e3744 | 1048 | } |
1049 | ||
1050 | /* Uninstall the route from kernel. */ | |
d62a17ae | 1051 | int rib_uninstall_kernel(struct route_node *rn, struct route_entry *re) |
718e3744 | 1052 | { |
d62a17ae | 1053 | int ret = 0; |
1054 | struct nexthop *nexthop; | |
1055 | rib_table_info_t *info = srcdest_rnode_table_info(rn); | |
1056 | struct prefix *p, *src_p; | |
1057 | struct zebra_vrf *zvrf = vrf_info_lookup(re->vrf_id); | |
05737783 | 1058 | |
d62a17ae | 1059 | srcdest_rnode_prefixes(rn, &p, &src_p); |
718e3744 | 1060 | |
d62a17ae | 1061 | if (info->safi != SAFI_UNICAST) { |
1062 | for (ALL_NEXTHOPS(re->nexthop, nexthop)) | |
1063 | SET_FLAG(nexthop->flags, NEXTHOP_FLAG_FIB); | |
1064 | return ret; | |
1065 | } | |
416ec78d | 1066 | |
d62a17ae | 1067 | /* |
1068 | * Make sure we update the FPM any time we send new information to | |
1069 | * the kernel. | |
1070 | */ | |
1071 | hook_call(rib_update, rn, "uninstalling from kernel"); | |
1072 | ret = kernel_route_rib(p, src_p, re, NULL); | |
1073 | zvrf->removals++; | |
718e3744 | 1074 | |
d62a17ae | 1075 | for (ALL_NEXTHOPS(re->nexthop, nexthop)) |
1076 | UNSET_FLAG(nexthop->flags, NEXTHOP_FLAG_FIB); | |
718e3744 | 1077 | |
d62a17ae | 1078 | return ret; |
718e3744 | 1079 | } |
1080 | ||
1081 | /* Uninstall the route from kernel. */ | |
d62a17ae | 1082 | static void rib_uninstall(struct route_node *rn, struct route_entry *re) |
718e3744 | 1083 | { |
d62a17ae | 1084 | rib_table_info_t *info = srcdest_rnode_table_info(rn); |
416ec78d | 1085 | |
d62a17ae | 1086 | if (CHECK_FLAG(re->status, ROUTE_ENTRY_SELECTED_FIB)) { |
1087 | if (info->safi == SAFI_UNICAST) | |
1088 | hook_call(rib_update, rn, "rib_uninstall"); | |
5adc2528 | 1089 | |
d62a17ae | 1090 | if (!RIB_SYSTEM_ROUTE(re)) |
1091 | rib_uninstall_kernel(rn, re); | |
a64448ba | 1092 | |
d62a17ae | 1093 | /* If labeled-unicast route, uninstall transit LSP. */ |
1094 | if (zebra_rib_labeled_unicast(re)) | |
1095 | zebra_mpls_lsp_uninstall(info->zvrf, rn, re); | |
a64448ba | 1096 | |
d62a17ae | 1097 | UNSET_FLAG(re->status, ROUTE_ENTRY_SELECTED_FIB); |
1098 | } | |
446bb95e | 1099 | |
d62a17ae | 1100 | if (CHECK_FLAG(re->flags, ZEBRA_FLAG_SELECTED)) { |
1101 | struct prefix *p, *src_p; | |
1102 | srcdest_rnode_prefixes(rn, &p, &src_p); | |
05737783 | 1103 | |
d62a17ae | 1104 | redistribute_delete(p, src_p, re); |
1105 | UNSET_FLAG(re->flags, ZEBRA_FLAG_SELECTED); | |
1106 | } | |
718e3744 | 1107 | } |
1108 | ||
9fd92e3c AS |
1109 | /* |
1110 | * rib_can_delete_dest | |
1111 | * | |
1112 | * Returns TRUE if the given dest can be deleted from the table. | |
1113 | */ | |
d62a17ae | 1114 | static int rib_can_delete_dest(rib_dest_t *dest) |
9fd92e3c | 1115 | { |
d62a17ae | 1116 | if (dest->routes) { |
1117 | return 0; | |
1118 | } | |
9fd92e3c | 1119 | |
d62a17ae | 1120 | /* |
1121 | * Don't delete the dest if we have to update the FPM about this | |
1122 | * prefix. | |
1123 | */ | |
1124 | if (CHECK_FLAG(dest->flags, RIB_DEST_UPDATE_FPM) | |
1125 | || CHECK_FLAG(dest->flags, RIB_DEST_SENT_TO_FPM)) | |
1126 | return 0; | |
5adc2528 | 1127 | |
d62a17ae | 1128 | return 1; |
9fd92e3c AS |
1129 | } |
1130 | ||
1131 | /* | |
1132 | * rib_gc_dest | |
1133 | * | |
1134 | * Garbage collect the rib dest corresponding to the given route node | |
1135 | * if appropriate. | |
1136 | * | |
1137 | * Returns TRUE if the dest was deleted, FALSE otherwise. | |
1138 | */ | |
d62a17ae | 1139 | int rib_gc_dest(struct route_node *rn) |
9fd92e3c | 1140 | { |
d62a17ae | 1141 | rib_dest_t *dest; |
9fd92e3c | 1142 | |
d62a17ae | 1143 | dest = rib_dest_from_rnode(rn); |
1144 | if (!dest) | |
1145 | return 0; | |
9fd92e3c | 1146 | |
d62a17ae | 1147 | if (!rib_can_delete_dest(dest)) |
1148 | return 0; | |
9fd92e3c | 1149 | |
c9abf558 DS |
1150 | if (IS_ZEBRA_DEBUG_RIB) { |
1151 | struct zebra_vrf *zvrf; | |
1152 | ||
1153 | zvrf = rib_dest_vrf(dest); | |
d62a17ae | 1154 | rnode_debug(rn, zvrf_id(zvrf), "removing dest from table"); |
c9abf558 | 1155 | } |
9fd92e3c | 1156 | |
d62a17ae | 1157 | dest->rnode = NULL; |
1158 | XFREE(MTYPE_RIB_DEST, dest); | |
1159 | rn->info = NULL; | |
9fd92e3c | 1160 | |
d62a17ae | 1161 | /* |
1162 | * Release the one reference that we keep on the route node. | |
1163 | */ | |
1164 | route_unlock_node(rn); | |
1165 | return 1; | |
9fd92e3c AS |
1166 | } |
1167 | ||
d62a17ae | 1168 | static void rib_process_add_fib(struct zebra_vrf *zvrf, struct route_node *rn, |
1169 | struct route_entry *new) | |
3e5c6e00 | 1170 | { |
d62a17ae | 1171 | hook_call(rib_update, rn, "new route selected"); |
3e5c6e00 | 1172 | |
d62a17ae | 1173 | /* Update real nexthop. This may actually determine if nexthop is active |
1174 | * or not. */ | |
1175 | if (!nexthop_active_update(rn, new, 1)) { | |
1176 | UNSET_FLAG(new->status, ROUTE_ENTRY_CHANGED); | |
1177 | return; | |
1178 | } | |
3e5c6e00 | 1179 | |
d62a17ae | 1180 | SET_FLAG(new->status, ROUTE_ENTRY_SELECTED_FIB); |
1181 | if (IS_ZEBRA_DEBUG_RIB) { | |
1182 | char buf[SRCDEST2STR_BUFFER]; | |
1183 | srcdest_rnode2str(rn, buf, sizeof(buf)); | |
1184 | zlog_debug("%u:%s: Adding route rn %p, re %p (type %d)", | |
1185 | zvrf_id(zvrf), buf, rn, new, new->type); | |
1186 | } | |
3e5c6e00 | 1187 | |
d62a17ae | 1188 | /* If labeled-unicast route, install transit LSP. */ |
1189 | if (zebra_rib_labeled_unicast(new)) | |
1190 | zebra_mpls_lsp_install(zvrf, rn, new); | |
a64448ba | 1191 | |
d62a17ae | 1192 | if (!RIB_SYSTEM_ROUTE(new)) { |
1193 | if (rib_install_kernel(rn, new, NULL)) { | |
1194 | char buf[SRCDEST2STR_BUFFER]; | |
1195 | srcdest_rnode2str(rn, buf, sizeof(buf)); | |
1196 | zlog_warn("%u:%s: Route install failed", zvrf_id(zvrf), | |
1197 | buf); | |
1198 | } | |
1199 | } | |
3e5c6e00 | 1200 | |
d62a17ae | 1201 | UNSET_FLAG(new->status, ROUTE_ENTRY_CHANGED); |
3e5c6e00 | 1202 | } |
1203 | ||
d62a17ae | 1204 | static void rib_process_del_fib(struct zebra_vrf *zvrf, struct route_node *rn, |
1205 | struct route_entry *old) | |
3e5c6e00 | 1206 | { |
d62a17ae | 1207 | hook_call(rib_update, rn, "removing existing route"); |
3e5c6e00 | 1208 | |
d62a17ae | 1209 | /* Uninstall from kernel. */ |
1210 | if (IS_ZEBRA_DEBUG_RIB) { | |
1211 | char buf[SRCDEST2STR_BUFFER]; | |
1212 | srcdest_rnode2str(rn, buf, sizeof(buf)); | |
1213 | zlog_debug("%u:%s: Deleting route rn %p, re %p (type %d)", | |
1214 | zvrf_id(zvrf), buf, rn, old, old->type); | |
1215 | } | |
3e5c6e00 | 1216 | |
d62a17ae | 1217 | /* If labeled-unicast route, uninstall transit LSP. */ |
1218 | if (zebra_rib_labeled_unicast(old)) | |
1219 | zebra_mpls_lsp_uninstall(zvrf, rn, old); | |
1220 | ||
1221 | if (!RIB_SYSTEM_ROUTE(old)) | |
1222 | rib_uninstall_kernel(rn, old); | |
1223 | ||
1224 | UNSET_FLAG(old->status, ROUTE_ENTRY_SELECTED_FIB); | |
1225 | ||
1226 | /* Update nexthop for route, reset changed flag. */ | |
1227 | nexthop_active_update(rn, old, 1); | |
1228 | UNSET_FLAG(old->status, ROUTE_ENTRY_CHANGED); | |
1229 | } | |
1230 | ||
1231 | static void rib_process_update_fib(struct zebra_vrf *zvrf, | |
1232 | struct route_node *rn, | |
1233 | struct route_entry *old, | |
1234 | struct route_entry *new) | |
1235 | { | |
1236 | struct nexthop *nexthop = NULL; | |
1237 | int nh_active = 0; | |
1238 | int installed = 1; | |
1239 | ||
1240 | /* | |
1241 | * We have to install or update if a new route has been selected or | |
1242 | * something has changed. | |
1243 | */ | |
1244 | if (new != old || CHECK_FLAG(new->status, ROUTE_ENTRY_CHANGED)) { | |
1245 | hook_call(rib_update, rn, "updating existing route"); | |
1246 | ||
1247 | /* Update the nexthop; we could determine here that nexthop is | |
1248 | * inactive. */ | |
1249 | if (nexthop_active_update(rn, new, 1)) | |
1250 | nh_active = 1; | |
1251 | ||
1252 | /* If nexthop is active, install the selected route, if | |
1253 | * appropriate. If | |
1254 | * the install succeeds, cleanup flags for prior route, if | |
1255 | * different from | |
1256 | * newly selected. | |
1257 | */ | |
1258 | if (nh_active) { | |
1259 | if (IS_ZEBRA_DEBUG_RIB) { | |
1260 | char buf[SRCDEST2STR_BUFFER]; | |
1261 | srcdest_rnode2str(rn, buf, sizeof(buf)); | |
1262 | if (new != old) | |
1263 | zlog_debug( | |
1264 | "%u:%s: Updating route rn %p, re %p (type %d) " | |
1265 | "old %p (type %d)", | |
1266 | zvrf_id(zvrf), buf, rn, new, | |
1267 | new->type, old, old->type); | |
1268 | else | |
1269 | zlog_debug( | |
1270 | "%u:%s: Updating route rn %p, re %p (type %d)", | |
1271 | zvrf_id(zvrf), buf, rn, new, | |
1272 | new->type); | |
1273 | } | |
1274 | ||
1275 | /* If labeled-unicast route, uninstall transit LSP. */ | |
1276 | if (zebra_rib_labeled_unicast(old)) | |
1277 | zebra_mpls_lsp_uninstall(zvrf, rn, old); | |
1278 | ||
1279 | /* Non-system route should be installed. */ | |
1280 | if (!RIB_SYSTEM_ROUTE(new)) { | |
1281 | /* If labeled-unicast route, install transit | |
1282 | * LSP. */ | |
1283 | if (zebra_rib_labeled_unicast(new)) | |
1284 | zebra_mpls_lsp_install(zvrf, rn, new); | |
1285 | ||
1286 | if (rib_install_kernel(rn, new, old)) { | |
1287 | char buf[SRCDEST2STR_BUFFER]; | |
1288 | srcdest_rnode2str(rn, buf, sizeof(buf)); | |
1289 | installed = 0; | |
1290 | zlog_warn("%u:%s: Route install failed", | |
1291 | zvrf_id(zvrf), buf); | |
1292 | } | |
1293 | } | |
1294 | ||
1295 | /* If install succeeded or system route, cleanup flags | |
1296 | * for prior route. */ | |
1297 | if (installed && new != old) { | |
1298 | if (RIB_SYSTEM_ROUTE(new)) { | |
1299 | if (!RIB_SYSTEM_ROUTE(old)) | |
1300 | rib_uninstall_kernel(rn, old); | |
1301 | } else { | |
1302 | for (nexthop = old->nexthop; nexthop; | |
1303 | nexthop = nexthop->next) | |
1304 | UNSET_FLAG(nexthop->flags, | |
1305 | NEXTHOP_FLAG_FIB); | |
1306 | } | |
1307 | } | |
1308 | ||
1309 | /* Update for redistribution. */ | |
1310 | if (installed) | |
1311 | SET_FLAG(new->status, ROUTE_ENTRY_SELECTED_FIB); | |
1312 | } | |
a64448ba | 1313 | |
d62a17ae | 1314 | /* |
1315 | * If nexthop for selected route is not active or install | |
1316 | * failed, we | |
1317 | * may need to uninstall and delete for redistribution. | |
1318 | */ | |
1319 | if (!nh_active || !installed) { | |
1320 | if (IS_ZEBRA_DEBUG_RIB) { | |
1321 | char buf[SRCDEST2STR_BUFFER]; | |
1322 | srcdest_rnode2str(rn, buf, sizeof(buf)); | |
1323 | if (new != old) | |
1324 | zlog_debug( | |
1325 | "%u:%s: Deleting route rn %p, re %p (type %d) " | |
1326 | "old %p (type %d) - %s", | |
1327 | zvrf_id(zvrf), buf, rn, new, | |
1328 | new->type, old, old->type, | |
1329 | nh_active ? "install failed" | |
1330 | : "nexthop inactive"); | |
1331 | else | |
1332 | zlog_debug( | |
1333 | "%u:%s: Deleting route rn %p, re %p (type %d) - %s", | |
1334 | zvrf_id(zvrf), buf, rn, new, | |
1335 | new->type, | |
1336 | nh_active ? "install failed" | |
1337 | : "nexthop inactive"); | |
1338 | } | |
1339 | ||
1340 | /* If labeled-unicast route, uninstall transit LSP. */ | |
1341 | if (zebra_rib_labeled_unicast(old)) | |
1342 | zebra_mpls_lsp_uninstall(zvrf, rn, old); | |
1343 | ||
1344 | if (!RIB_SYSTEM_ROUTE(old)) | |
1345 | rib_uninstall_kernel(rn, old); | |
1346 | UNSET_FLAG(new->status, ROUTE_ENTRY_SELECTED_FIB); | |
1347 | } | |
1348 | } else { | |
1349 | /* | |
1350 | * Same route selected; check if in the FIB and if not, | |
1351 | * re-install. This | |
1352 | * is housekeeping code to deal with race conditions in kernel | |
1353 | * with linux | |
1354 | * netlink reporting interface up before IPv4 or IPv6 protocol | |
1355 | * is ready | |
1356 | * to add routes. | |
1357 | */ | |
1358 | if (!RIB_SYSTEM_ROUTE(new)) { | |
1359 | int in_fib = 0; | |
1360 | ||
1361 | for (ALL_NEXTHOPS(new->nexthop, nexthop)) | |
1362 | if (CHECK_FLAG(nexthop->flags, | |
1363 | NEXTHOP_FLAG_FIB)) { | |
1364 | in_fib = 1; | |
1365 | break; | |
1366 | } | |
1367 | if (!in_fib) | |
1368 | rib_install_kernel(rn, new, NULL); | |
1369 | } | |
1370 | } | |
1371 | ||
1372 | /* Update prior route. */ | |
1373 | if (new != old) { | |
1374 | UNSET_FLAG(old->status, ROUTE_ENTRY_SELECTED_FIB); | |
3e5c6e00 | 1375 | |
d62a17ae | 1376 | /* Set real nexthop. */ |
1377 | nexthop_active_update(rn, old, 1); | |
1378 | UNSET_FLAG(old->status, ROUTE_ENTRY_CHANGED); | |
1379 | } | |
3e5c6e00 | 1380 | |
d62a17ae | 1381 | /* Clear changed flag. */ |
1382 | UNSET_FLAG(new->status, ROUTE_ENTRY_CHANGED); | |
3e5c6e00 | 1383 | } |
1384 | ||
d62a17ae | 1385 | /* Check if 'alternate' RIB entry is better than 'current'. */ |
1386 | static struct route_entry *rib_choose_best(struct route_entry *current, | |
1387 | struct route_entry *alternate) | |
1388 | { | |
1389 | if (current == NULL) | |
1390 | return alternate; | |
1391 | ||
1392 | /* filter route selection in following order: | |
1393 | * - connected beats other types | |
1394 | * - lower distance beats higher | |
1395 | * - lower metric beats higher for equal distance | |
1396 | * - last, hence oldest, route wins tie break. | |
1397 | */ | |
1398 | ||
1399 | /* Connected routes. Pick the last connected | |
1400 | * route of the set of lowest metric connected routes. | |
1401 | */ | |
1402 | if (alternate->type == ZEBRA_ROUTE_CONNECT) { | |
1403 | if (current->type != ZEBRA_ROUTE_CONNECT | |
1404 | || alternate->metric <= current->metric) | |
1405 | return alternate; | |
1406 | ||
1407 | return current; | |
1408 | } | |
3e5c6e00 | 1409 | |
d62a17ae | 1410 | if (current->type == ZEBRA_ROUTE_CONNECT) |
1411 | return current; | |
3e5c6e00 | 1412 | |
d62a17ae | 1413 | /* higher distance loses */ |
1414 | if (alternate->distance < current->distance) | |
1415 | return alternate; | |
1416 | if (current->distance < alternate->distance) | |
1417 | return current; | |
3e5c6e00 | 1418 | |
d62a17ae | 1419 | /* metric tie-breaks equal distance */ |
1420 | if (alternate->metric <= current->metric) | |
1421 | return alternate; | |
3e5c6e00 | 1422 | |
d62a17ae | 1423 | return current; |
3e5c6e00 | 1424 | } |
1425 | ||
d62a17ae | 1426 | /* Core function for processing routing information base. */ |
1427 | static void rib_process(struct route_node *rn) | |
1428 | { | |
1429 | struct route_entry *re; | |
1430 | struct route_entry *next; | |
1431 | struct route_entry *old_selected = NULL; | |
1432 | struct route_entry *new_selected = NULL; | |
1433 | struct route_entry *old_fib = NULL; | |
1434 | struct route_entry *new_fib = NULL; | |
1435 | struct route_entry *best = NULL; | |
1436 | char buf[SRCDEST2STR_BUFFER]; | |
1437 | rib_dest_t *dest; | |
1438 | struct zebra_vrf *zvrf = NULL; | |
1439 | struct prefix *p, *src_p; | |
1440 | srcdest_rnode_prefixes(rn, &p, &src_p); | |
1441 | vrf_id_t vrf_id = VRF_UNKNOWN; | |
1442 | ||
1443 | assert(rn); | |
1444 | ||
1445 | dest = rib_dest_from_rnode(rn); | |
1446 | if (dest) { | |
1447 | zvrf = rib_dest_vrf(dest); | |
1448 | vrf_id = zvrf_id(zvrf); | |
1449 | } | |
bab85d4f | 1450 | |
d62a17ae | 1451 | if (IS_ZEBRA_DEBUG_RIB) |
1452 | srcdest_rnode2str(rn, buf, sizeof(buf)); | |
bab85d4f | 1453 | |
d62a17ae | 1454 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) |
1455 | zlog_debug("%u:%s: Processing rn %p", vrf_id, buf, rn); | |
bab85d4f | 1456 | |
a2addae8 | 1457 | RNODE_FOREACH_RE_SAFE (rn, re, next) { |
d62a17ae | 1458 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) |
1459 | zlog_debug( | |
1460 | "%u:%s: Examine re %p (type %d) status %x flags %x " | |
1461 | "dist %d metric %d", | |
1462 | vrf_id, buf, re, re->type, re->status, | |
1463 | re->flags, re->distance, re->metric); | |
1464 | ||
1465 | UNSET_FLAG(re->status, ROUTE_ENTRY_NEXTHOPS_CHANGED); | |
1466 | ||
1467 | /* Currently selected re. */ | |
1468 | if (CHECK_FLAG(re->flags, ZEBRA_FLAG_SELECTED)) { | |
1469 | assert(old_selected == NULL); | |
1470 | old_selected = re; | |
1471 | } | |
1472 | /* Currently in fib */ | |
1473 | if (CHECK_FLAG(re->status, ROUTE_ENTRY_SELECTED_FIB)) { | |
1474 | assert(old_fib == NULL); | |
1475 | old_fib = re; | |
1476 | } | |
bab85d4f | 1477 | |
d62a17ae | 1478 | /* Skip deleted entries from selection */ |
1479 | if (CHECK_FLAG(re->status, ROUTE_ENTRY_REMOVED)) | |
1480 | continue; | |
1481 | ||
1482 | /* Skip unreachable nexthop. */ | |
1483 | /* This first call to nexthop_active_update is merely to | |
1484 | * determine if | |
1485 | * there's any change to nexthops associated with this RIB | |
1486 | * entry. Now, | |
1487 | * rib_process() can be invoked due to an external event such as | |
1488 | * link | |
1489 | * down or due to next-hop-tracking evaluation. In the latter | |
1490 | * case, | |
1491 | * a decision has already been made that the NHs have changed. | |
1492 | * So, no | |
1493 | * need to invoke a potentially expensive call again. Further, | |
1494 | * since | |
1495 | * the change might be in a recursive NH which is not caught in | |
1496 | * the nexthop_active_update() code. Thus, we might miss changes | |
1497 | * to | |
1498 | * recursive NHs. | |
1499 | */ | |
1500 | if (!CHECK_FLAG(re->status, ROUTE_ENTRY_CHANGED) | |
1501 | && !nexthop_active_update(rn, re, 0)) { | |
1502 | if (re->type == ZEBRA_ROUTE_TABLE) { | |
1503 | /* XXX: HERE BE DRAGONS!!!!! | |
1504 | * In all honesty, I have not yet figured out | |
1505 | * what this part | |
1506 | * does or why the ROUTE_ENTRY_CHANGED test | |
1507 | * above is correct | |
1508 | * or why we need to delete a route here, and | |
1509 | * also not whether | |
1510 | * this concerns both selected and fib route, or | |
1511 | * only selected | |
1512 | * or only fib */ | |
1513 | /* This entry was denied by the 'ip protocol | |
1514 | * table' route-map, we | |
1515 | * need to delete it */ | |
1516 | if (re != old_selected) { | |
1517 | if (IS_ZEBRA_DEBUG_RIB) | |
1518 | zlog_debug( | |
1519 | "%s: %s: imported via import-table but denied " | |
1520 | "by the ip protocol table route-map", | |
1521 | __func__, buf); | |
1522 | rib_unlink(rn, re); | |
1523 | } else | |
1524 | SET_FLAG(re->status, | |
1525 | ROUTE_ENTRY_REMOVED); | |
1526 | } | |
1527 | ||
1528 | continue; | |
1529 | } | |
bab85d4f | 1530 | |
d62a17ae | 1531 | /* Infinite distance. */ |
1532 | if (re->distance == DISTANCE_INFINITY) { | |
1533 | UNSET_FLAG(re->status, ROUTE_ENTRY_CHANGED); | |
1534 | continue; | |
1535 | } | |
bab85d4f | 1536 | |
d62a17ae | 1537 | if (CHECK_FLAG(re->flags, ZEBRA_FLAG_FIB_OVERRIDE)) { |
1538 | best = rib_choose_best(new_fib, re); | |
1539 | if (new_fib && best != new_fib) | |
1540 | UNSET_FLAG(new_fib->status, | |
1541 | ROUTE_ENTRY_CHANGED); | |
1542 | new_fib = best; | |
1543 | } else { | |
1544 | best = rib_choose_best(new_selected, re); | |
1545 | if (new_selected && best != new_selected) | |
1546 | UNSET_FLAG(new_selected->status, | |
1547 | ROUTE_ENTRY_CHANGED); | |
1548 | new_selected = best; | |
1549 | } | |
1550 | if (best != re) | |
1551 | UNSET_FLAG(re->status, ROUTE_ENTRY_CHANGED); | |
1552 | } /* RNODE_FOREACH_RE */ | |
1553 | ||
1554 | /* If no FIB override route, use the selected route also for FIB */ | |
1555 | if (new_fib == NULL) | |
1556 | new_fib = new_selected; | |
1557 | ||
1558 | /* After the cycle is finished, the following pointers will be set: | |
1559 | * old_selected --- RE entry currently having SELECTED | |
1560 | * new_selected --- RE entry that is newly SELECTED | |
1561 | * old_fib --- RE entry currently in kernel FIB | |
1562 | * new_fib --- RE entry that is newly to be in kernel FIB | |
1563 | * | |
1564 | * new_selected will get SELECTED flag, and is going to be redistributed | |
1565 | * the zclients. new_fib (which can be new_selected) will be installed | |
1566 | * in kernel. | |
1567 | */ | |
1568 | ||
1569 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) { | |
1570 | zlog_debug( | |
1571 | "%u:%s: After processing: old_selected %p new_selected %p old_fib %p new_fib %p", | |
1572 | vrf_id, buf, (void *)old_selected, (void *)new_selected, | |
1573 | (void *)old_fib, (void *)new_fib); | |
1574 | } | |
446bb95e | 1575 | |
d62a17ae | 1576 | /* Buffer ROUTE_ENTRY_CHANGED here, because it will get cleared if |
1577 | * fib == selected */ | |
9d303b37 DL |
1578 | bool selected_changed = new_selected && CHECK_FLAG(new_selected->status, |
1579 | ROUTE_ENTRY_CHANGED); | |
d62a17ae | 1580 | |
1581 | /* Update fib according to selection results */ | |
1582 | if (new_fib && old_fib) | |
1583 | rib_process_update_fib(zvrf, rn, old_fib, new_fib); | |
1584 | else if (new_fib) | |
1585 | rib_process_add_fib(zvrf, rn, new_fib); | |
1586 | else if (old_fib) | |
1587 | rib_process_del_fib(zvrf, rn, old_fib); | |
1588 | ||
1589 | /* Redistribute SELECTED entry */ | |
1590 | if (old_selected != new_selected || selected_changed) { | |
1591 | struct nexthop *nexthop; | |
1592 | ||
1593 | /* Check if we have a FIB route for the destination, otherwise, | |
1594 | * don't redistribute it */ | |
1595 | for (ALL_NEXTHOPS(new_fib ? new_fib->nexthop : NULL, nexthop)) { | |
1596 | if (CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_FIB)) { | |
1597 | break; | |
1598 | } | |
1599 | } | |
1600 | if (!nexthop) | |
1601 | new_selected = NULL; | |
93bdadae | 1602 | |
d62a17ae | 1603 | if (new_selected && new_selected != new_fib) { |
1604 | nexthop_active_update(rn, new_selected, 1); | |
1605 | UNSET_FLAG(new_selected->status, ROUTE_ENTRY_CHANGED); | |
1606 | } | |
41ec9222 | 1607 | |
d62a17ae | 1608 | if (old_selected) { |
1609 | if (!new_selected) | |
1610 | redistribute_delete(p, src_p, old_selected); | |
1611 | if (old_selected != new_selected) | |
1612 | UNSET_FLAG(old_selected->flags, | |
1613 | ZEBRA_FLAG_SELECTED); | |
f857321e | 1614 | } |
446bb95e | 1615 | |
d62a17ae | 1616 | if (new_selected) { |
1617 | /* Install new or replace existing redistributed entry | |
1618 | */ | |
1619 | SET_FLAG(new_selected->flags, ZEBRA_FLAG_SELECTED); | |
1620 | redistribute_update(p, src_p, new_selected, | |
1621 | old_selected); | |
1622 | } | |
1623 | } | |
3e5c6e00 | 1624 | |
d62a17ae | 1625 | /* Remove all RE entries queued for removal */ |
a2addae8 | 1626 | RNODE_FOREACH_RE_SAFE (rn, re, next) { |
d62a17ae | 1627 | if (CHECK_FLAG(re->status, ROUTE_ENTRY_REMOVED)) { |
1628 | if (IS_ZEBRA_DEBUG_RIB) { | |
1629 | rnode_debug(rn, vrf_id, "rn %p, removing re %p", | |
1630 | (void *)rn, (void *)re); | |
1631 | } | |
1632 | rib_unlink(rn, re); | |
1633 | } | |
1634 | } | |
4d38fdb4 | 1635 | |
d62a17ae | 1636 | /* |
1637 | * Check if the dest can be deleted now. | |
1638 | */ | |
1639 | rib_gc_dest(rn); | |
e96f9203 DO |
1640 | } |
1641 | ||
5110a0c6 | 1642 | /* Take a list of route_node structs and return 1, if there was a record |
d62a17ae | 1643 | * picked from it and processed by rib_process(). Don't process more, |
5110a0c6 | 1644 | * than one RN record; operate only in the specified sub-queue. |
e96f9203 | 1645 | */ |
d62a17ae | 1646 | static unsigned int process_subq(struct list *subq, u_char qindex) |
e96f9203 | 1647 | { |
d62a17ae | 1648 | struct listnode *lnode = listhead(subq); |
1649 | struct route_node *rnode; | |
1650 | rib_dest_t *dest; | |
1651 | struct zebra_vrf *zvrf = NULL; | |
5110a0c6 | 1652 | |
d62a17ae | 1653 | if (!lnode) |
1654 | return 0; | |
5110a0c6 | 1655 | |
d62a17ae | 1656 | rnode = listgetdata(lnode); |
1657 | dest = rib_dest_from_rnode(rnode); | |
1658 | if (dest) | |
1659 | zvrf = rib_dest_vrf(dest); | |
41ec9222 | 1660 | |
d62a17ae | 1661 | rib_process(rnode); |
5110a0c6 | 1662 | |
d62a17ae | 1663 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) { |
1664 | char buf[SRCDEST2STR_BUFFER]; | |
1665 | srcdest_rnode2str(rnode, buf, sizeof(buf)); | |
1666 | zlog_debug("%u:%s: rn %p dequeued from sub-queue %u", | |
1667 | zvrf ? zvrf_id(zvrf) : 0, buf, rnode, qindex); | |
1668 | } | |
41ec9222 | 1669 | |
d62a17ae | 1670 | if (rnode->info) |
1671 | UNSET_FLAG(rib_dest_from_rnode(rnode)->flags, | |
1672 | RIB_ROUTE_QUEUED(qindex)); | |
9fd92e3c | 1673 | |
67b9467f | 1674 | #if 0 |
5110a0c6 SH |
1675 | else |
1676 | { | |
1677 | zlog_debug ("%s: called for route_node (%p, %d) with no ribs", | |
1678 | __func__, rnode, rnode->lock); | |
1679 | zlog_backtrace(LOG_DEBUG); | |
1680 | } | |
67b9467f | 1681 | #endif |
d62a17ae | 1682 | route_unlock_node(rnode); |
1683 | list_delete_node(subq, lnode); | |
1684 | return 1; | |
e96f9203 DO |
1685 | } |
1686 | ||
fb018d25 DS |
1687 | /* |
1688 | * All meta queues have been processed. Trigger next-hop evaluation. | |
1689 | */ | |
d62a17ae | 1690 | static void meta_queue_process_complete(struct work_queue *dummy) |
fb018d25 | 1691 | { |
d62a17ae | 1692 | struct vrf *vrf; |
1693 | struct zebra_vrf *zvrf; | |
9ec6b0bb | 1694 | |
d62a17ae | 1695 | /* Evaluate nexthops for those VRFs which underwent route processing. |
1696 | * This | |
1697 | * should limit the evaluation to the necessary VRFs in most common | |
1698 | * situations. | |
1699 | */ | |
a2addae8 | 1700 | RB_FOREACH (vrf, vrf_id_head, &vrfs_by_id) { |
d62a17ae | 1701 | zvrf = vrf->info; |
1702 | if (zvrf == NULL || !(zvrf->flags & ZEBRA_VRF_RIB_SCHEDULED)) | |
1703 | continue; | |
1704 | ||
1705 | zvrf->flags &= ~ZEBRA_VRF_RIB_SCHEDULED; | |
1706 | zebra_evaluate_rnh(zvrf_id(zvrf), AF_INET, 0, RNH_NEXTHOP_TYPE, | |
1707 | NULL); | |
1708 | zebra_evaluate_rnh(zvrf_id(zvrf), AF_INET, 0, | |
1709 | RNH_IMPORT_CHECK_TYPE, NULL); | |
1710 | zebra_evaluate_rnh(zvrf_id(zvrf), AF_INET6, 0, RNH_NEXTHOP_TYPE, | |
1711 | NULL); | |
1712 | zebra_evaluate_rnh(zvrf_id(zvrf), AF_INET6, 0, | |
1713 | RNH_IMPORT_CHECK_TYPE, NULL); | |
1714 | } | |
939fba27 | 1715 | |
d62a17ae | 1716 | /* Schedule LSPs for processing, if needed. */ |
1717 | zvrf = vrf_info_lookup(VRF_DEFAULT); | |
1718 | if (mpls_should_lsps_be_processed(zvrf)) { | |
1719 | if (IS_ZEBRA_DEBUG_MPLS) | |
1720 | zlog_debug( | |
1721 | "%u: Scheduling all LSPs upon RIB completion", | |
1722 | zvrf_id(zvrf)); | |
1723 | zebra_mpls_lsp_schedule(zvrf); | |
1724 | mpls_unmark_lsps_for_processing(zvrf); | |
1725 | } | |
fb018d25 DS |
1726 | } |
1727 | ||
e96f9203 | 1728 | /* Dispatch the meta queue by picking, processing and unlocking the next RN from |
d62a17ae | 1729 | * a non-empty sub-queue with lowest priority. wq is equal to zebra->ribq and |
1730 | * data | |
e96f9203 DO |
1731 | * is pointed to the meta queue structure. |
1732 | */ | |
d62a17ae | 1733 | static wq_item_status meta_queue_process(struct work_queue *dummy, void *data) |
e96f9203 | 1734 | { |
d62a17ae | 1735 | struct meta_queue *mq = data; |
1736 | unsigned i; | |
5110a0c6 | 1737 | |
d62a17ae | 1738 | for (i = 0; i < MQ_SIZE; i++) |
1739 | if (process_subq(mq->subq[i], i)) { | |
1740 | mq->size--; | |
1741 | break; | |
1742 | } | |
1743 | return mq->size ? WQ_REQUEUE : WQ_SUCCESS; | |
e96f9203 DO |
1744 | } |
1745 | ||
9fd92e3c AS |
1746 | /* |
1747 | * Map from rib types to queue type (priority) in meta queue | |
1748 | */ | |
5110a0c6 | 1749 | static const u_char meta_queue_map[ZEBRA_ROUTE_MAX] = { |
9d303b37 DL |
1750 | [ZEBRA_ROUTE_SYSTEM] = 4, |
1751 | [ZEBRA_ROUTE_KERNEL] = 0, | |
1752 | [ZEBRA_ROUTE_CONNECT] = 0, | |
1753 | [ZEBRA_ROUTE_STATIC] = 1, | |
1754 | [ZEBRA_ROUTE_RIP] = 2, | |
1755 | [ZEBRA_ROUTE_RIPNG] = 2, | |
1756 | [ZEBRA_ROUTE_OSPF] = 2, | |
1757 | [ZEBRA_ROUTE_OSPF6] = 2, | |
1758 | [ZEBRA_ROUTE_ISIS] = 2, | |
1759 | [ZEBRA_ROUTE_BGP] = 3, | |
1760 | [ZEBRA_ROUTE_PIM] = 4, // Shouldn't happen but for safety | |
1761 | [ZEBRA_ROUTE_EIGRP] = 2, | |
1762 | [ZEBRA_ROUTE_NHRP] = 2, | |
1763 | [ZEBRA_ROUTE_HSLS] = 4, | |
1764 | [ZEBRA_ROUTE_OLSR] = 4, | |
1765 | [ZEBRA_ROUTE_TABLE] = 1, | |
1766 | [ZEBRA_ROUTE_LDP] = 4, | |
1767 | [ZEBRA_ROUTE_VNC] = 3, | |
1768 | [ZEBRA_ROUTE_VNC_DIRECT] = 3, | |
1769 | [ZEBRA_ROUTE_VNC_DIRECT_RH] = 3, | |
1770 | [ZEBRA_ROUTE_BGP_DIRECT] = 3, | |
1771 | [ZEBRA_ROUTE_BGP_DIRECT_EXT] = 3, | |
1772 | [ZEBRA_ROUTE_BABEL] = 2, | |
1773 | [ZEBRA_ROUTE_ALL] = 4, // Shouldn't happen but for safety | |
5110a0c6 SH |
1774 | }; |
1775 | ||
1776 | /* Look into the RN and queue it into one or more priority queues, | |
1777 | * increasing the size for each data push done. | |
e96f9203 | 1778 | */ |
d62a17ae | 1779 | static void rib_meta_queue_add(struct meta_queue *mq, struct route_node *rn) |
e96f9203 | 1780 | { |
d62a17ae | 1781 | struct route_entry *re; |
5110a0c6 | 1782 | |
a2addae8 | 1783 | RNODE_FOREACH_RE (rn, re) { |
d62a17ae | 1784 | u_char qindex = meta_queue_map[re->type]; |
1785 | struct zebra_vrf *zvrf; | |
1786 | ||
1787 | /* Invariant: at this point we always have rn->info set. */ | |
1788 | if (CHECK_FLAG(rib_dest_from_rnode(rn)->flags, | |
1789 | RIB_ROUTE_QUEUED(qindex))) { | |
1790 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) | |
1791 | rnode_debug( | |
1792 | rn, re->vrf_id, | |
1793 | "rn %p is already queued in sub-queue %u", | |
1794 | (void *)rn, qindex); | |
1795 | continue; | |
1796 | } | |
5110a0c6 | 1797 | |
d62a17ae | 1798 | SET_FLAG(rib_dest_from_rnode(rn)->flags, |
1799 | RIB_ROUTE_QUEUED(qindex)); | |
1800 | listnode_add(mq->subq[qindex], rn); | |
1801 | route_lock_node(rn); | |
1802 | mq->size++; | |
5110a0c6 | 1803 | |
d62a17ae | 1804 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) |
1805 | rnode_debug(rn, re->vrf_id, | |
1806 | "queued rn %p into sub-queue %u", | |
1807 | (void *)rn, qindex); | |
9ec6b0bb | 1808 | |
d62a17ae | 1809 | zvrf = zebra_vrf_lookup_by_id(re->vrf_id); |
1810 | if (zvrf) | |
1811 | zvrf->flags |= ZEBRA_VRF_RIB_SCHEDULED; | |
1812 | } | |
4d38fdb4 | 1813 | } |
1814 | ||
6d691129 | 1815 | /* Add route_node to work queue and schedule processing */ |
d62a17ae | 1816 | void rib_queue_add(struct route_node *rn) |
4d38fdb4 | 1817 | { |
d62a17ae | 1818 | assert(rn); |
fc328ac9 | 1819 | |
d62a17ae | 1820 | /* Pointless to queue a route_node with no RIB entries to add or remove |
1821 | */ | |
1822 | if (!rnode_to_ribs(rn)) { | |
1823 | zlog_debug("%s: called for route_node (%p, %d) with no ribs", | |
1824 | __func__, (void *)rn, rn->lock); | |
1825 | zlog_backtrace(LOG_DEBUG); | |
1826 | return; | |
1827 | } | |
4d38fdb4 | 1828 | |
d62a17ae | 1829 | if (zebrad.ribq == NULL) { |
1830 | zlog_err("%s: work_queue does not exist!", __func__); | |
1831 | return; | |
1832 | } | |
1833 | ||
1834 | /* | |
1835 | * The RIB queue should normally be either empty or holding the only | |
1836 | * work_queue_item element. In the latter case this element would | |
1837 | * hold a pointer to the meta queue structure, which must be used to | |
1838 | * actually queue the route nodes to process. So create the MQ | |
1839 | * holder, if necessary, then push the work into it in any case. | |
1840 | * This semantics was introduced after 0.99.9 release. | |
1841 | */ | |
f104f6c1 | 1842 | if (work_queue_empty(zebrad.ribq)) |
d62a17ae | 1843 | work_queue_add(zebrad.ribq, zebrad.mq); |
e96f9203 | 1844 | |
d62a17ae | 1845 | rib_meta_queue_add(zebrad.mq, rn); |
fc328ac9 | 1846 | |
d62a17ae | 1847 | return; |
4d38fdb4 | 1848 | } |
1849 | ||
5110a0c6 SH |
1850 | /* Create new meta queue. |
1851 | A destructor function doesn't seem to be necessary here. | |
1852 | */ | |
d62a17ae | 1853 | static struct meta_queue *meta_queue_new(void) |
e96f9203 | 1854 | { |
d62a17ae | 1855 | struct meta_queue *new; |
1856 | unsigned i; | |
5110a0c6 | 1857 | |
d62a17ae | 1858 | new = XCALLOC(MTYPE_WORK_QUEUE, sizeof(struct meta_queue)); |
1859 | assert(new); | |
e96f9203 | 1860 | |
d62a17ae | 1861 | for (i = 0; i < MQ_SIZE; i++) { |
1862 | new->subq[i] = list_new(); | |
1863 | assert(new->subq[i]); | |
1864 | } | |
5110a0c6 | 1865 | |
d62a17ae | 1866 | return new; |
e96f9203 DO |
1867 | } |
1868 | ||
d62a17ae | 1869 | void meta_queue_free(struct meta_queue *mq) |
5a8dfcd8 | 1870 | { |
d62a17ae | 1871 | unsigned i; |
5a8dfcd8 | 1872 | |
d62a17ae | 1873 | for (i = 0; i < MQ_SIZE; i++) |
affe9e99 | 1874 | list_delete_and_null(&mq->subq[i]); |
5a8dfcd8 | 1875 | |
d62a17ae | 1876 | XFREE(MTYPE_WORK_QUEUE, mq); |
5a8dfcd8 RW |
1877 | } |
1878 | ||
4d38fdb4 | 1879 | /* initialise zebra rib work queue */ |
d62a17ae | 1880 | static void rib_queue_init(struct zebra_t *zebra) |
4d38fdb4 | 1881 | { |
d62a17ae | 1882 | assert(zebra); |
1883 | ||
1884 | if (!(zebra->ribq = | |
1885 | work_queue_new(zebra->master, "route_node processing"))) { | |
1886 | zlog_err("%s: could not initialise work queue!", __func__); | |
1887 | return; | |
1888 | } | |
4d38fdb4 | 1889 | |
d62a17ae | 1890 | /* fill in the work queue spec */ |
1891 | zebra->ribq->spec.workfunc = &meta_queue_process; | |
1892 | zebra->ribq->spec.errorfunc = NULL; | |
1893 | zebra->ribq->spec.completion_func = &meta_queue_process_complete; | |
1894 | /* XXX: TODO: These should be runtime configurable via vty */ | |
1895 | zebra->ribq->spec.max_retries = 3; | |
3a30f50f | 1896 | zebra->ribq->spec.hold = ZEBRA_RIB_PROCESS_HOLD_TIME; |
d62a17ae | 1897 | |
1898 | if (!(zebra->mq = meta_queue_new())) { | |
1899 | zlog_err("%s: could not initialise meta queue!", __func__); | |
1900 | return; | |
1901 | } | |
1902 | return; | |
718e3744 | 1903 | } |
1904 | ||
6d691129 PJ |
1905 | /* RIB updates are processed via a queue of pointers to route_nodes. |
1906 | * | |
1907 | * The queue length is bounded by the maximal size of the routing table, | |
1908 | * as a route_node will not be requeued, if already queued. | |
1909 | * | |
f0f77c9a DS |
1910 | * REs are submitted via rib_addnode or rib_delnode which set minimal |
1911 | * state, or static_install_route (when an existing RE is updated) | |
3c0755dc | 1912 | * and then submit route_node to queue for best-path selection later. |
f0f77c9a | 1913 | * Order of add/delete state changes are preserved for any given RE. |
6d691129 | 1914 | * |
f0f77c9a | 1915 | * Deleted REs are reaped during best-path selection. |
6d691129 PJ |
1916 | * |
1917 | * rib_addnode | |
f0f77c9a DS |
1918 | * |-> rib_link or unset ROUTE_ENTRY_REMOVE |->Update kernel with |
1919 | * |-------->| | best RE, if required | |
3c0755dc PJ |
1920 | * | | |
1921 | * static_install->|->rib_addqueue...... -> rib_process | |
1922 | * | | | |
1923 | * |-------->| |-> rib_unlink | |
f0f77c9a DS |
1924 | * |-> set ROUTE_ENTRY_REMOVE | |
1925 | * rib_delnode (RE freed) | |
6d691129 | 1926 | * |
9fd92e3c AS |
1927 | * The 'info' pointer of a route_node points to a rib_dest_t |
1928 | * ('dest'). Queueing state for a route_node is kept on the dest. The | |
1929 | * dest is created on-demand by rib_link() and is kept around at least | |
1930 | * as long as there are ribs hanging off it (@see rib_gc_dest()). | |
d62a17ae | 1931 | * |
6d691129 PJ |
1932 | * Refcounting (aka "locking" throughout the GNU Zebra and Quagga code): |
1933 | * | |
1934 | * - route_nodes: refcounted by: | |
9fd92e3c AS |
1935 | * - dest attached to route_node: |
1936 | * - managed by: rib_link/rib_gc_dest | |
6d691129 PJ |
1937 | * - route_node processing queue |
1938 | * - managed by: rib_addqueue, rib_process. | |
1939 | * | |
1940 | */ | |
d62a17ae | 1941 | |
f0f77c9a | 1942 | /* Add RE to head of the route node. */ |
d62a17ae | 1943 | static void rib_link(struct route_node *rn, struct route_entry *re, int process) |
1944 | { | |
1945 | struct route_entry *head; | |
1946 | rib_dest_t *dest; | |
1947 | afi_t afi; | |
1948 | const char *rmap_name; | |
9fd92e3c | 1949 | |
d62a17ae | 1950 | assert(re && rn); |
9fd92e3c | 1951 | |
d62a17ae | 1952 | dest = rib_dest_from_rnode(rn); |
1953 | if (!dest) { | |
1954 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) | |
1955 | rnode_debug(rn, re->vrf_id, "rn %p adding dest", rn); | |
7a4bb9c5 | 1956 | |
d62a17ae | 1957 | dest = XCALLOC(MTYPE_RIB_DEST, sizeof(rib_dest_t)); |
1958 | route_lock_node(rn); /* rn route table reference */ | |
1959 | rn->info = dest; | |
1960 | dest->rnode = rn; | |
1961 | } | |
2263a412 | 1962 | |
d62a17ae | 1963 | head = dest->routes; |
1964 | if (head) { | |
1965 | head->prev = re; | |
1966 | } | |
1967 | re->next = head; | |
1968 | dest->routes = re; | |
1969 | ||
1970 | afi = (rn->p.family == AF_INET) | |
1971 | ? AFI_IP | |
1972 | : (rn->p.family == AF_INET6) ? AFI_IP6 : AFI_MAX; | |
1973 | if (is_zebra_import_table_enabled(afi, re->table)) { | |
1974 | rmap_name = zebra_get_import_table_route_map(afi, re->table); | |
1975 | zebra_add_import_table_entry(rn, re, rmap_name); | |
1976 | } else if (process) | |
1977 | rib_queue_add(rn); | |
1978 | } | |
1979 | ||
1980 | void rib_addnode(struct route_node *rn, struct route_entry *re, int process) | |
1981 | { | |
1982 | /* RE node has been un-removed before route-node is processed. | |
1983 | * route_node must hence already be on the queue for processing.. | |
1984 | */ | |
1985 | if (CHECK_FLAG(re->status, ROUTE_ENTRY_REMOVED)) { | |
1986 | if (IS_ZEBRA_DEBUG_RIB) | |
1987 | rnode_debug(rn, re->vrf_id, "rn %p, un-removed re %p", | |
1988 | (void *)rn, (void *)re); | |
1989 | ||
1990 | UNSET_FLAG(re->status, ROUTE_ENTRY_REMOVED); | |
1991 | return; | |
1992 | } | |
1993 | rib_link(rn, re, process); | |
6d691129 PJ |
1994 | } |
1995 | ||
9fd92e3c AS |
1996 | /* |
1997 | * rib_unlink | |
1998 | * | |
1999 | * Detach a rib structure from a route_node. | |
2000 | * | |
2001 | * Note that a call to rib_unlink() should be followed by a call to | |
2002 | * rib_gc_dest() at some point. This allows a rib_dest_t that is no | |
2003 | * longer required to be deleted. | |
2004 | */ | |
d62a17ae | 2005 | void rib_unlink(struct route_node *rn, struct route_entry *re) |
6d691129 | 2006 | { |
d62a17ae | 2007 | rib_dest_t *dest; |
9fd92e3c | 2008 | |
d62a17ae | 2009 | assert(rn && re); |
6d691129 | 2010 | |
d62a17ae | 2011 | if (IS_ZEBRA_DEBUG_RIB) |
2012 | rnode_debug(rn, re->vrf_id, "rn %p, re %p", (void *)rn, | |
2013 | (void *)re); | |
6d691129 | 2014 | |
d62a17ae | 2015 | dest = rib_dest_from_rnode(rn); |
6d691129 | 2016 | |
d62a17ae | 2017 | if (re->next) |
2018 | re->next->prev = re->prev; | |
6d691129 | 2019 | |
d62a17ae | 2020 | if (re->prev) |
2021 | re->prev->next = re->next; | |
2022 | else { | |
2023 | dest->routes = re->next; | |
2024 | } | |
7a4bb9c5 | 2025 | |
d62a17ae | 2026 | /* free RE and nexthops */ |
2027 | zebra_deregister_rnh_static_nexthops(re->vrf_id, re->nexthop, rn); | |
2028 | nexthops_free(re->nexthop); | |
2029 | XFREE(MTYPE_RE, re); | |
2030 | } | |
2031 | ||
2032 | void rib_delnode(struct route_node *rn, struct route_entry *re) | |
2033 | { | |
2034 | afi_t afi; | |
2035 | ||
2036 | if (IS_ZEBRA_DEBUG_RIB) | |
2037 | rnode_debug(rn, re->vrf_id, "rn %p, re %p, removing", | |
2038 | (void *)rn, (void *)re); | |
2039 | SET_FLAG(re->status, ROUTE_ENTRY_REMOVED); | |
2040 | ||
2041 | afi = (rn->p.family == AF_INET) | |
2042 | ? AFI_IP | |
2043 | : (rn->p.family == AF_INET6) ? AFI_IP6 : AFI_MAX; | |
2044 | if (is_zebra_import_table_enabled(afi, re->table)) { | |
2045 | zebra_del_import_table_entry(rn, re); | |
2046 | /* Just clean up if non main table */ | |
2047 | if (IS_ZEBRA_DEBUG_RIB) { | |
2048 | char buf[SRCDEST2STR_BUFFER]; | |
2049 | srcdest_rnode2str(rn, buf, sizeof(buf)); | |
2050 | zlog_debug( | |
2051 | "%u:%s: Freeing route rn %p, re %p (type %d)", | |
2052 | re->vrf_id, buf, rn, re, re->type); | |
2053 | } | |
7a4bb9c5 | 2054 | |
d62a17ae | 2055 | rib_unlink(rn, re); |
2056 | } else { | |
2057 | rib_queue_add(rn); | |
2058 | } | |
718e3744 | 2059 | } |
2060 | ||
f0f77c9a | 2061 | /* This function dumps the contents of a given RE entry into |
dc95824a DO |
2062 | * standard debug log. Calling function name and IP prefix in |
2063 | * question are passed as 1st and 2nd arguments. | |
2064 | */ | |
2065 | ||
d62a17ae | 2066 | void _route_entry_dump(const char *func, union prefixconstptr pp, |
2067 | union prefixconstptr src_pp, | |
2068 | const struct route_entry *re) | |
2069 | { | |
2070 | const struct prefix *p = pp.p; | |
2071 | const struct prefix *src_p = src_pp.p; | |
2072 | bool is_srcdst = src_p && src_p->prefixlen; | |
2073 | char straddr[PREFIX_STRLEN]; | |
2074 | char srcaddr[PREFIX_STRLEN]; | |
2075 | struct nexthop *nexthop; | |
2076 | ||
2077 | zlog_debug("%s: dumping RE entry %p for %s%s%s vrf %u", func, | |
2078 | (const void *)re, prefix2str(pp, straddr, sizeof(straddr)), | |
2079 | is_srcdst ? " from " : "", | |
2080 | is_srcdst ? prefix2str(src_pp, srcaddr, sizeof(srcaddr)) | |
2081 | : "", | |
2082 | re->vrf_id); | |
cc54cfee RW |
2083 | zlog_debug("%s: uptime == %lu, type == %u, instance == %d, table == %d", |
2084 | func, (unsigned long)re->uptime, re->type, re->instance, | |
2085 | re->table); | |
d62a17ae | 2086 | zlog_debug( |
2087 | "%s: metric == %u, mtu == %u, distance == %u, flags == %u, status == %u", | |
2088 | func, re->metric, re->mtu, re->distance, re->flags, re->status); | |
2089 | zlog_debug("%s: nexthop_num == %u, nexthop_active_num == %u", func, | |
2090 | re->nexthop_num, re->nexthop_active_num); | |
2091 | ||
2092 | for (ALL_NEXTHOPS(re->nexthop, nexthop)) { | |
2093 | inet_ntop(p->family, &nexthop->gate, straddr, INET6_ADDRSTRLEN); | |
915902cb | 2094 | zlog_debug("%s: %s %s[%u] with flags %s%s%s", func, |
d62a17ae | 2095 | (nexthop->rparent ? " NH" : "NH"), straddr, |
915902cb | 2096 | nexthop->ifindex, |
d62a17ae | 2097 | (CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE) |
2098 | ? "ACTIVE " | |
2099 | : ""), | |
2100 | (CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_FIB) | |
2101 | ? "FIB " | |
2102 | : ""), | |
2103 | (CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_RECURSIVE) | |
2104 | ? "RECURSIVE" | |
2105 | : "")); | |
2106 | } | |
2107 | zlog_debug("%s: dump complete", func); | |
dc95824a DO |
2108 | } |
2109 | ||
2110 | /* This is an exported helper to rtm_read() to dump the strange | |
f0f77c9a | 2111 | * RE entry found by rib_lookup_ipv4_route() |
dc95824a DO |
2112 | */ |
2113 | ||
d62a17ae | 2114 | void rib_lookup_and_dump(struct prefix_ipv4 *p, vrf_id_t vrf_id) |
2115 | { | |
2116 | struct route_table *table; | |
2117 | struct route_node *rn; | |
2118 | struct route_entry *re; | |
2119 | char prefix_buf[INET_ADDRSTRLEN]; | |
2120 | ||
2121 | /* Lookup table. */ | |
2122 | table = zebra_vrf_table(AFI_IP, SAFI_UNICAST, vrf_id); | |
2123 | if (!table) { | |
2124 | zlog_err("%s: zebra_vrf_table() returned NULL", __func__); | |
2125 | return; | |
2126 | } | |
2127 | ||
2128 | /* Scan the RIB table for exactly matching RE entry. */ | |
2129 | rn = route_node_lookup(table, (struct prefix *)p); | |
2130 | ||
2131 | /* No route for this prefix. */ | |
2132 | if (!rn) { | |
2133 | zlog_debug("%s: lookup failed for %s", __func__, | |
2134 | prefix2str((struct prefix *)p, prefix_buf, | |
2135 | sizeof(prefix_buf))); | |
2136 | return; | |
2137 | } | |
2138 | ||
2139 | /* Unlock node. */ | |
2140 | route_unlock_node(rn); | |
2141 | ||
2142 | /* let's go */ | |
a2addae8 | 2143 | RNODE_FOREACH_RE (rn, re) { |
d62a17ae | 2144 | zlog_debug("%s: rn %p, re %p: %s, %s", __func__, (void *)rn, |
2145 | (void *)re, | |
2146 | (CHECK_FLAG(re->status, ROUTE_ENTRY_REMOVED) | |
2147 | ? "removed" | |
2148 | : "NOT removed"), | |
2149 | (CHECK_FLAG(re->flags, ZEBRA_FLAG_SELECTED) | |
2150 | ? "selected" | |
2151 | : "NOT selected")); | |
2152 | route_entry_dump(p, NULL, re); | |
2153 | } | |
dc95824a DO |
2154 | } |
2155 | ||
20e5ff0a DO |
2156 | /* Check if requested address assignment will fail due to another |
2157 | * route being installed by zebra in FIB already. Take necessary | |
2158 | * actions, if needed: remove such a route from FIB and deSELECT | |
f0f77c9a | 2159 | * corresponding RE entry. Then put affected RN into RIBQ head. |
20e5ff0a | 2160 | */ |
d62a17ae | 2161 | void rib_lookup_and_pushup(struct prefix_ipv4 *p, vrf_id_t vrf_id) |
2162 | { | |
2163 | struct route_table *table; | |
2164 | struct route_node *rn; | |
2165 | struct route_entry *re; | |
2166 | unsigned changed = 0; | |
2167 | ||
2168 | if (NULL == (table = zebra_vrf_table(AFI_IP, SAFI_UNICAST, vrf_id))) { | |
2169 | zlog_err("%s: zebra_vrf_table() returned NULL", __func__); | |
2170 | return; | |
2171 | } | |
2172 | ||
2173 | /* No matches would be the simplest case. */ | |
2174 | if (NULL == (rn = route_node_lookup(table, (struct prefix *)p))) | |
2175 | return; | |
2176 | ||
2177 | /* Unlock node. */ | |
2178 | route_unlock_node(rn); | |
2179 | ||
2180 | /* Check all RE entries. In case any changes have to be done, requeue | |
2181 | * the RN into RIBQ head. If the routing message about the new connected | |
2182 | * route (generated by the IP address we are going to assign very soon) | |
2183 | * comes before the RIBQ is processed, the new RE entry will join | |
2184 | * RIBQ record already on head. This is necessary for proper | |
2185 | * revalidation | |
2186 | * of the rest of the RE. | |
2187 | */ | |
a2addae8 | 2188 | RNODE_FOREACH_RE (rn, re) { |
d62a17ae | 2189 | if (CHECK_FLAG(re->status, ROUTE_ENTRY_SELECTED_FIB) |
2190 | && !RIB_SYSTEM_ROUTE(re)) { | |
2191 | changed = 1; | |
2192 | if (IS_ZEBRA_DEBUG_RIB) { | |
2193 | char buf[PREFIX_STRLEN]; | |
2194 | zlog_debug( | |
2195 | "%u:%s: freeing way for connected prefix", | |
2196 | re->vrf_id, | |
2197 | prefix2str(&rn->p, buf, sizeof(buf))); | |
2198 | route_entry_dump(&rn->p, NULL, re); | |
2199 | } | |
2200 | rib_uninstall(rn, re); | |
2201 | } | |
2202 | } | |
2203 | if (changed) | |
2204 | rib_queue_add(rn); | |
20e5ff0a DO |
2205 | } |
2206 | ||
d62a17ae | 2207 | int rib_add_multipath(afi_t afi, safi_t safi, struct prefix *p, |
2208 | struct prefix_ipv6 *src_p, struct route_entry *re) | |
718e3744 | 2209 | { |
d62a17ae | 2210 | struct route_table *table; |
2211 | struct route_node *rn; | |
2212 | struct route_entry *same; | |
2213 | struct nexthop *nexthop; | |
2214 | int ret = 0; | |
b4c034b0 | 2215 | |
d62a17ae | 2216 | if (!re) |
2217 | return 0; | |
b4c034b0 | 2218 | |
7990990e | 2219 | assert(!src_p || afi == AFI_IP6); |
05737783 | 2220 | |
d62a17ae | 2221 | /* Lookup table. */ |
7865c65d RW |
2222 | table = zebra_vrf_table_with_table_id(afi, safi, re->vrf_id, re->table); |
2223 | if (!table) { | |
2224 | XFREE(MTYPE_RE, re); | |
d62a17ae | 2225 | return 0; |
7865c65d | 2226 | } |
cddf391b | 2227 | |
d62a17ae | 2228 | /* Make it sure prefixlen is applied to the prefix. */ |
2229 | apply_mask(p); | |
2230 | if (src_p) | |
2231 | apply_mask_ipv6(src_p); | |
718e3744 | 2232 | |
d62a17ae | 2233 | /* Set default distance by route type. */ |
2234 | if (re->distance == 0) { | |
0492eea0 | 2235 | re->distance = route_distance(re->type); |
718e3744 | 2236 | |
d62a17ae | 2237 | /* iBGP distance is 200. */ |
2238 | if (re->type == ZEBRA_ROUTE_BGP | |
2239 | && CHECK_FLAG(re->flags, ZEBRA_FLAG_IBGP)) | |
2240 | re->distance = 200; | |
2241 | } | |
718e3744 | 2242 | |
d62a17ae | 2243 | /* Lookup route node.*/ |
2244 | rn = srcdest_rnode_get(table, p, src_p); | |
718e3744 | 2245 | |
d62a17ae | 2246 | /* If same type of route are installed, treat it as a implicit |
2247 | withdraw. */ | |
a2addae8 | 2248 | RNODE_FOREACH_RE (rn, same) { |
d62a17ae | 2249 | if (CHECK_FLAG(same->status, ROUTE_ENTRY_REMOVED)) |
2250 | continue; | |
41ec9222 | 2251 | |
eb327fa5 RW |
2252 | if (same->type != re->type) |
2253 | continue; | |
2254 | if (same->instance != re->instance) | |
2255 | continue; | |
2256 | if (same->type == ZEBRA_ROUTE_KERNEL && | |
2257 | same->metric != re->metric) | |
2258 | continue; | |
844b3a87 RW |
2259 | /* |
2260 | * We should allow duplicate connected routes because of | |
2261 | * IPv6 link-local routes and unnumbered interfaces on Linux. | |
2262 | */ | |
2263 | if (same->type != ZEBRA_ROUTE_CONNECT) | |
d62a17ae | 2264 | break; |
2265 | } | |
718e3744 | 2266 | |
d62a17ae | 2267 | /* If this route is kernel route, set FIB flag to the route. */ |
8628fc61 | 2268 | if (RIB_SYSTEM_ROUTE(re)) |
d62a17ae | 2269 | for (nexthop = re->nexthop; nexthop; nexthop = nexthop->next) |
2270 | SET_FLAG(nexthop->flags, NEXTHOP_FLAG_FIB); | |
718e3744 | 2271 | |
d62a17ae | 2272 | /* Link new re to node.*/ |
2273 | if (IS_ZEBRA_DEBUG_RIB) { | |
2274 | rnode_debug( | |
2275 | rn, re->vrf_id, | |
2276 | "Inserting route rn %p, re %p (type %d) existing %p", | |
2277 | (void *)rn, (void *)re, re->type, (void *)same); | |
718e3744 | 2278 | |
d62a17ae | 2279 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) |
2280 | route_entry_dump(p, src_p, re); | |
718e3744 | 2281 | } |
d62a17ae | 2282 | rib_addnode(rn, re, 1); |
2283 | ret = 1; | |
6b0655a2 | 2284 | |
d62a17ae | 2285 | /* Free implicit route.*/ |
2286 | if (same) { | |
2287 | rib_delnode(rn, same); | |
2288 | ret = -1; | |
2289 | } | |
718e3744 | 2290 | |
d62a17ae | 2291 | route_unlock_node(rn); |
2292 | return ret; | |
2293 | } | |
2294 | ||
2295 | void rib_delete(afi_t afi, safi_t safi, vrf_id_t vrf_id, int type, | |
2296 | u_short instance, int flags, struct prefix *p, | |
fd36be7e | 2297 | struct prefix_ipv6 *src_p, const struct nexthop *nh, |
5dfeba19 | 2298 | u_int32_t table_id, u_int32_t metric, bool fromkernel) |
d62a17ae | 2299 | { |
2300 | struct route_table *table; | |
2301 | struct route_node *rn; | |
2302 | struct route_entry *re; | |
2303 | struct route_entry *fib = NULL; | |
2304 | struct route_entry *same = NULL; | |
fd36be7e | 2305 | struct nexthop *rtnh; |
d62a17ae | 2306 | char buf2[INET6_ADDRSTRLEN]; |
2307 | ||
2308 | assert(!src_p || afi == AFI_IP6); | |
2309 | ||
2310 | /* Lookup table. */ | |
2311 | table = zebra_vrf_table_with_table_id(afi, safi, vrf_id, table_id); | |
2312 | if (!table) | |
2313 | return; | |
2314 | ||
2315 | /* Apply mask. */ | |
2316 | apply_mask(p); | |
2317 | if (src_p) | |
2318 | apply_mask_ipv6(src_p); | |
2319 | ||
2320 | /* Lookup route node. */ | |
2321 | rn = srcdest_rnode_lookup(table, p, src_p); | |
2322 | if (!rn) { | |
2323 | char dst_buf[PREFIX_STRLEN], src_buf[PREFIX_STRLEN]; | |
2324 | ||
2325 | prefix2str(p, dst_buf, sizeof(dst_buf)); | |
2326 | if (src_p && src_p->prefixlen) | |
2327 | prefix2str(src_p, src_buf, sizeof(src_buf)); | |
2328 | else | |
2329 | src_buf[0] = '\0'; | |
2330 | ||
2331 | if (IS_ZEBRA_DEBUG_RIB) | |
2332 | zlog_debug("%u:%s%s%s doesn't exist in rib", vrf_id, | |
2333 | dst_buf, | |
2334 | (src_buf[0] != '\0') ? " from " : "", | |
2335 | src_buf); | |
2336 | return; | |
2337 | } | |
718e3744 | 2338 | |
d62a17ae | 2339 | /* Lookup same type route. */ |
a2addae8 | 2340 | RNODE_FOREACH_RE (rn, re) { |
d62a17ae | 2341 | if (CHECK_FLAG(re->status, ROUTE_ENTRY_REMOVED)) |
2342 | continue; | |
2343 | ||
2344 | if (CHECK_FLAG(re->status, ROUTE_ENTRY_SELECTED_FIB)) | |
2345 | fib = re; | |
2346 | ||
2347 | if (re->type != type) | |
2348 | continue; | |
2349 | if (re->instance != instance) | |
2350 | continue; | |
f19435a8 DS |
2351 | if (re->type == ZEBRA_ROUTE_KERNEL && |
2352 | re->metric != metric) | |
2353 | continue; | |
fd36be7e DL |
2354 | if (re->type == ZEBRA_ROUTE_CONNECT && (rtnh = re->nexthop) |
2355 | && rtnh->type == NEXTHOP_TYPE_IFINDEX && nh) { | |
2356 | if (rtnh->ifindex != nh->ifindex) | |
d62a17ae | 2357 | continue; |
d62a17ae | 2358 | same = re; |
2359 | break; | |
2360 | } | |
2361 | /* Make sure that the route found has the same gateway. */ | |
2362 | else { | |
fd36be7e | 2363 | if (nh == NULL) { |
d62a17ae | 2364 | same = re; |
2365 | break; | |
2366 | } | |
fd36be7e DL |
2367 | for (ALL_NEXTHOPS(re->nexthop, rtnh)) |
2368 | if (nexthop_same_no_recurse(rtnh, nh)) { | |
d62a17ae | 2369 | same = re; |
2370 | break; | |
2371 | } | |
2372 | if (same) | |
2373 | break; | |
2374 | } | |
2375 | } | |
2376 | /* If same type of route can't be found and this message is from | |
2377 | kernel. */ | |
2378 | if (!same) { | |
5dfeba19 DS |
2379 | /* |
2380 | * In the past(HA!) we could get here because | |
2381 | * we were receiving a route delete from the | |
2382 | * kernel and we're not marking the proto | |
2383 | * as coming from it's appropriate originator. | |
2384 | * Now that we are properly noticing the fact | |
2385 | * that the kernel has deleted our route we | |
2386 | * are not going to get called in this path | |
2387 | * I am going to leave this here because | |
2388 | * this might still work this way on non-linux | |
2389 | * platforms as well as some weird state I have | |
2390 | * not properly thought of yet. | |
2391 | * If we can show that this code path is | |
2392 | * dead then we can remove it. | |
2393 | */ | |
d62a17ae | 2394 | if (fib && type == ZEBRA_ROUTE_KERNEL |
2395 | && CHECK_FLAG(flags, ZEBRA_FLAG_SELFROUTE)) { | |
2396 | if (IS_ZEBRA_DEBUG_RIB) { | |
2397 | rnode_debug( | |
2398 | rn, vrf_id, | |
2399 | "rn %p, re %p (type %d) was deleted from kernel, adding", | |
2400 | rn, fib, fib->type); | |
2401 | } | |
2402 | if (allow_delete) { | |
2403 | /* Unset flags. */ | |
fd36be7e DL |
2404 | for (rtnh = fib->nexthop; rtnh; |
2405 | rtnh = rtnh->next) | |
2406 | UNSET_FLAG(rtnh->flags, | |
d62a17ae | 2407 | NEXTHOP_FLAG_FIB); |
2408 | ||
2409 | UNSET_FLAG(fib->status, | |
2410 | ROUTE_ENTRY_SELECTED_FIB); | |
2411 | } else { | |
2412 | /* This means someone else, other than Zebra, | |
2413 | * has deleted | |
2414 | * a Zebra router from the kernel. We will add | |
2415 | * it back */ | |
2416 | rib_install_kernel(rn, fib, NULL); | |
2417 | } | |
2418 | } else { | |
2419 | if (IS_ZEBRA_DEBUG_RIB) { | |
fd36be7e | 2420 | if (nh) |
d62a17ae | 2421 | rnode_debug( |
2422 | rn, vrf_id, | |
2423 | "via %s ifindex %d type %d " | |
2424 | "doesn't exist in rib", | |
2425 | inet_ntop( | |
60466a63 QY |
2426 | family2afi(afi), |
2427 | &nh->gate, buf2, | |
d62a17ae | 2428 | INET_ADDRSTRLEN), /* FIXME |
9d303b37 | 2429 | */ |
fd36be7e | 2430 | nh->ifindex, type); |
d62a17ae | 2431 | else |
2432 | rnode_debug( | |
2433 | rn, vrf_id, | |
fd36be7e DL |
2434 | "type %d doesn't exist in rib", |
2435 | type); | |
d62a17ae | 2436 | } |
2437 | route_unlock_node(rn); | |
2438 | return; | |
2439 | } | |
2440 | } | |
718e3744 | 2441 | |
5dfeba19 DS |
2442 | if (same) { |
2443 | if (fromkernel && | |
2444 | CHECK_FLAG(flags, ZEBRA_FLAG_SELFROUTE) && | |
2445 | !allow_delete) { | |
2446 | rib_install_kernel(rn, same, NULL); | |
2447 | route_unlock_node(rn); | |
2448 | ||
2449 | return; | |
2450 | } | |
d62a17ae | 2451 | rib_delnode(rn, same); |
5dfeba19 | 2452 | } |
05737783 | 2453 | |
d62a17ae | 2454 | route_unlock_node(rn); |
2455 | return; | |
2456 | } | |
718e3744 | 2457 | |
718e3744 | 2458 | |
d62a17ae | 2459 | int rib_add(afi_t afi, safi_t safi, vrf_id_t vrf_id, int type, u_short instance, |
2460 | int flags, struct prefix *p, struct prefix_ipv6 *src_p, | |
fd36be7e | 2461 | const struct nexthop *nh, u_int32_t table_id, u_int32_t metric, |
fd289fc8 | 2462 | u_int32_t mtu, uint8_t distance) |
d62a17ae | 2463 | { |
2464 | struct route_entry *re; | |
66af6845 | 2465 | struct nexthop *nexthop; |
718e3744 | 2466 | |
66af6845 | 2467 | /* Allocate new route_entry structure. */ |
d62a17ae | 2468 | re = XCALLOC(MTYPE_RE, sizeof(struct route_entry)); |
d62a17ae | 2469 | re->type = type; |
2470 | re->instance = instance; | |
2471 | re->distance = distance; | |
2472 | re->flags = flags; | |
2473 | re->metric = metric; | |
2474 | re->mtu = mtu; | |
2475 | re->table = table_id; | |
2476 | re->vrf_id = vrf_id; | |
2477 | re->nexthop_num = 0; | |
2478 | re->uptime = time(NULL); | |
2479 | ||
66af6845 RW |
2480 | /* Add nexthop. */ |
2481 | nexthop = nexthop_new(); | |
2482 | *nexthop = *nh; | |
2483 | route_entry_nexthop_add(re, nexthop); | |
718e3744 | 2484 | |
66af6845 | 2485 | return rib_add_multipath(afi, safi, p, src_p, re); |
718e3744 | 2486 | } |
2487 | ||
1c848137 | 2488 | /* Schedule routes of a particular table (address-family) based on event. */ |
d62a17ae | 2489 | static void rib_update_table(struct route_table *table, |
2490 | rib_update_event_t event) | |
2491 | { | |
2492 | struct route_node *rn; | |
2493 | struct route_entry *re, *next; | |
2494 | ||
2495 | /* Walk all routes and queue for processing, if appropriate for | |
2496 | * the trigger event. | |
2497 | */ | |
2498 | for (rn = route_top(table); rn; rn = srcdest_route_next(rn)) { | |
1ca60f2c DS |
2499 | /* |
2500 | * If we are looking at a route node and the node | |
2501 | * has already been queued we don't | |
2502 | * need to queue it up again | |
2503 | */ | |
2504 | if (rn->info | |
2505 | && CHECK_FLAG(rib_dest_from_rnode(rn)->flags, | |
2506 | RIB_ROUTE_ANY_QUEUED)) | |
2507 | continue; | |
d62a17ae | 2508 | switch (event) { |
2509 | case RIB_UPDATE_IF_CHANGE: | |
2510 | /* Examine all routes that won't get processed by the | |
2511 | * protocol or | |
2512 | * triggered by nexthop evaluation (NHT). This would be | |
2513 | * system, | |
2514 | * kernel and certain static routes. Note that NHT will | |
2515 | * get | |
2516 | * triggered upon an interface event as connected routes | |
2517 | * always | |
2518 | * get queued for processing. | |
2519 | */ | |
a2addae8 | 2520 | RNODE_FOREACH_RE_SAFE (rn, re, next) { |
0a16efff DS |
2521 | struct nexthop *nh; |
2522 | ||
2523 | if (re->type != ZEBRA_ROUTE_SYSTEM && | |
2524 | re->type != ZEBRA_ROUTE_KERNEL && | |
2525 | re->type != ZEBRA_ROUTE_CONNECT && | |
2526 | re->type != ZEBRA_ROUTE_STATIC) | |
2527 | continue; | |
2528 | ||
2529 | if (re->type != ZEBRA_ROUTE_STATIC) { | |
2530 | rib_queue_add(rn); | |
2531 | continue; | |
2532 | } | |
2533 | ||
2534 | for (nh = re->nexthop; nh; nh = nh->next) | |
2535 | if (!(nh->type == NEXTHOP_TYPE_IPV4 | |
2536 | || nh->type == NEXTHOP_TYPE_IPV6)) | |
2537 | break; | |
2538 | ||
2539 | /* If we only have nexthops to a | |
2540 | * gateway, NHT will | |
2541 | * take care. | |
2542 | */ | |
2543 | if (nh) | |
d62a17ae | 2544 | rib_queue_add(rn); |
2545 | } | |
2546 | break; | |
2547 | ||
2548 | case RIB_UPDATE_RMAP_CHANGE: | |
2549 | case RIB_UPDATE_OTHER: | |
2550 | /* Right now, examine all routes. Can restrict to a | |
2551 | * protocol in | |
2552 | * some cases (TODO). | |
2553 | */ | |
2554 | if (rnode_to_ribs(rn)) | |
2555 | rib_queue_add(rn); | |
2556 | break; | |
2557 | ||
2558 | default: | |
2559 | break; | |
2560 | } | |
2561 | } | |
b84c7253 | 2562 | } |
2563 | ||
718e3744 | 2564 | /* RIB update function. */ |
d62a17ae | 2565 | void rib_update(vrf_id_t vrf_id, rib_update_event_t event) |
718e3744 | 2566 | { |
d62a17ae | 2567 | struct route_table *table; |
1c848137 | 2568 | |
d62a17ae | 2569 | /* Process routes of interested address-families. */ |
2570 | table = zebra_vrf_table(AFI_IP, SAFI_UNICAST, vrf_id); | |
2571 | if (table) | |
2572 | rib_update_table(table, event); | |
718e3744 | 2573 | |
d62a17ae | 2574 | table = zebra_vrf_table(AFI_IP6, SAFI_UNICAST, vrf_id); |
2575 | if (table) | |
2576 | rib_update_table(table, event); | |
718e3744 | 2577 | } |
2578 | ||
718e3744 | 2579 | /* Remove all routes which comes from non main table. */ |
d62a17ae | 2580 | static void rib_weed_table(struct route_table *table) |
718e3744 | 2581 | { |
d62a17ae | 2582 | struct route_node *rn; |
2583 | struct route_entry *re; | |
2584 | struct route_entry *next; | |
718e3744 | 2585 | |
d62a17ae | 2586 | if (table) |
2587 | for (rn = route_top(table); rn; rn = srcdest_route_next(rn)) | |
a2addae8 | 2588 | RNODE_FOREACH_RE_SAFE (rn, re, next) { |
d62a17ae | 2589 | if (CHECK_FLAG(re->status, ROUTE_ENTRY_REMOVED)) |
2590 | continue; | |
6d691129 | 2591 | |
d62a17ae | 2592 | if (re->table != zebrad.rtm_table_default |
2593 | && re->table != RT_TABLE_MAIN) | |
2594 | rib_delnode(rn, re); | |
2595 | } | |
718e3744 | 2596 | } |
2597 | ||
2598 | /* Delete all routes from non main table. */ | |
d62a17ae | 2599 | void rib_weed_tables(void) |
718e3744 | 2600 | { |
d62a17ae | 2601 | struct vrf *vrf; |
2602 | struct zebra_vrf *zvrf; | |
78104b9b | 2603 | |
a2addae8 RW |
2604 | RB_FOREACH (vrf, vrf_id_head, &vrfs_by_id) |
2605 | if ((zvrf = vrf->info) != NULL) { | |
2606 | rib_weed_table(zvrf->table[AFI_IP][SAFI_UNICAST]); | |
2607 | rib_weed_table(zvrf->table[AFI_IP6][SAFI_UNICAST]); | |
2608 | } | |
718e3744 | 2609 | } |
6b0655a2 | 2610 | |
718e3744 | 2611 | /* Delete self installed routes after zebra is relaunched. */ |
d62a17ae | 2612 | static void rib_sweep_table(struct route_table *table) |
2613 | { | |
2614 | struct route_node *rn; | |
2615 | struct route_entry *re; | |
2616 | struct route_entry *next; | |
915902cb | 2617 | struct nexthop *nexthop; |
d62a17ae | 2618 | int ret = 0; |
2619 | ||
915902cb DS |
2620 | if (!table) |
2621 | return; | |
d62a17ae | 2622 | |
915902cb | 2623 | for (rn = route_top(table); rn; rn = srcdest_route_next(rn)) { |
a2addae8 | 2624 | RNODE_FOREACH_RE_SAFE (rn, re, next) { |
915902cb DS |
2625 | if (IS_ZEBRA_DEBUG_RIB) |
2626 | route_entry_dump(&rn->p, NULL, re); | |
2627 | ||
2628 | if (CHECK_FLAG(re->status, ROUTE_ENTRY_REMOVED)) | |
2629 | continue; | |
2630 | ||
2631 | if (!CHECK_FLAG(re->flags, ZEBRA_FLAG_SELFROUTE)) | |
2632 | continue; | |
2633 | ||
2634 | /* | |
2635 | * So we are starting up and have received | |
2636 | * routes from the kernel that we have installed | |
2637 | * from a previous run of zebra but not cleaned | |
2638 | * up ( say a kill -9 ) | |
2639 | * But since we haven't actually installed | |
2640 | * them yet( we received them from the kernel ) | |
2641 | * we don't think they are active. | |
2642 | * So let's pretend they are active to actually | |
2643 | * remove them. | |
2644 | * In all honesty I'm not sure if we should | |
2645 | * mark them as active when we receive them | |
2646 | * This is startup only so probably ok. | |
2647 | * | |
2648 | * If we ever decide to move rib_sweep_table | |
2649 | * to a different spot (ie startup ) | |
2650 | * this decision needs to be revisited | |
2651 | */ | |
2652 | for (ALL_NEXTHOPS(re->nexthop, nexthop)) | |
2653 | SET_FLAG(nexthop->flags, NEXTHOP_FLAG_FIB); | |
2654 | ||
2655 | ret = rib_uninstall_kernel(rn, re); | |
2656 | if (!ret) | |
2657 | rib_delnode(rn, re); | |
2658 | } | |
2659 | } | |
718e3744 | 2660 | } |
2661 | ||
2662 | /* Sweep all RIB tables. */ | |
d62a17ae | 2663 | void rib_sweep_route(void) |
718e3744 | 2664 | { |
d62a17ae | 2665 | struct vrf *vrf; |
2666 | struct zebra_vrf *zvrf; | |
78104b9b | 2667 | |
a2addae8 | 2668 | RB_FOREACH (vrf, vrf_id_head, &vrfs_by_id) { |
915902cb DS |
2669 | if ((zvrf = vrf->info) == NULL) |
2670 | continue; | |
2671 | ||
d62a17ae | 2672 | rib_sweep_table(zvrf->table[AFI_IP][SAFI_UNICAST]); |
2673 | rib_sweep_table(zvrf->table[AFI_IP6][SAFI_UNICAST]); | |
2674 | } | |
718e3744 | 2675 | } |
2ea1ab1c VT |
2676 | |
2677 | /* Remove specific by protocol routes from 'table'. */ | |
d62a17ae | 2678 | static unsigned long rib_score_proto_table(u_char proto, u_short instance, |
2679 | struct route_table *table) | |
2680 | { | |
2681 | struct route_node *rn; | |
2682 | struct route_entry *re; | |
2683 | struct route_entry *next; | |
2684 | unsigned long n = 0; | |
2685 | ||
2686 | if (table) | |
2687 | for (rn = route_top(table); rn; rn = srcdest_route_next(rn)) | |
a2addae8 | 2688 | RNODE_FOREACH_RE_SAFE (rn, re, next) { |
d62a17ae | 2689 | if (CHECK_FLAG(re->status, ROUTE_ENTRY_REMOVED)) |
2690 | continue; | |
2691 | if (re->type == proto | |
2692 | && re->instance == instance) { | |
2693 | rib_delnode(rn, re); | |
2694 | n++; | |
2695 | } | |
2696 | } | |
2697 | return n; | |
2ea1ab1c VT |
2698 | } |
2699 | ||
2700 | /* Remove specific by protocol routes. */ | |
d62a17ae | 2701 | unsigned long rib_score_proto(u_char proto, u_short instance) |
2ea1ab1c | 2702 | { |
d62a17ae | 2703 | struct vrf *vrf; |
2704 | struct zebra_vrf *zvrf; | |
2705 | unsigned long cnt = 0; | |
78104b9b | 2706 | |
a2addae8 RW |
2707 | RB_FOREACH (vrf, vrf_id_head, &vrfs_by_id) |
2708 | if ((zvrf = vrf->info) != NULL) | |
2709 | cnt += rib_score_proto_table( | |
2710 | proto, instance, | |
2711 | zvrf->table[AFI_IP][SAFI_UNICAST]) | |
2712 | + rib_score_proto_table( | |
2713 | proto, instance, | |
2714 | zvrf->table[AFI_IP6][SAFI_UNICAST]); | |
78104b9b | 2715 | |
d62a17ae | 2716 | return cnt; |
2ea1ab1c VT |
2717 | } |
2718 | ||
718e3744 | 2719 | /* Close RIB and clean up kernel routes. */ |
d62a17ae | 2720 | void rib_close_table(struct route_table *table) |
718e3744 | 2721 | { |
d62a17ae | 2722 | struct route_node *rn; |
1e9f448f | 2723 | rib_table_info_t *info; |
d62a17ae | 2724 | struct route_entry *re; |
718e3744 | 2725 | |
1e9f448f DS |
2726 | if (!table) |
2727 | return; | |
9fd92e3c | 2728 | |
1e9f448f | 2729 | info = table->info; |
5adc2528 | 2730 | |
1e9f448f | 2731 | for (rn = route_top(table); rn; rn = srcdest_route_next(rn)) |
a2addae8 RW |
2732 | RNODE_FOREACH_RE (rn, re) { |
2733 | if (!CHECK_FLAG(re->status, ROUTE_ENTRY_SELECTED_FIB)) | |
1e9f448f DS |
2734 | continue; |
2735 | ||
2736 | if (info->safi == SAFI_UNICAST) | |
2737 | hook_call(rib_update, rn, NULL); | |
2738 | ||
2739 | if (!RIB_SYSTEM_ROUTE(re)) | |
2740 | rib_uninstall_kernel(rn, re); | |
2741 | } | |
718e3744 | 2742 | } |
2743 | ||
718e3744 | 2744 | /* Routing information base initialize. */ |
d62a17ae | 2745 | void rib_init(void) |
718e3744 | 2746 | { |
d62a17ae | 2747 | rib_queue_init(&zebrad); |
718e3744 | 2748 | } |
0915bb0c AS |
2749 | |
2750 | /* | |
2751 | * vrf_id_get_next | |
2752 | * | |
2753 | * Get the first vrf id that is greater than the given vrf id if any. | |
2754 | * | |
2755 | * Returns TRUE if a vrf id was found, FALSE otherwise. | |
2756 | */ | |
d62a17ae | 2757 | static inline int vrf_id_get_next(vrf_id_t vrf_id, vrf_id_t *next_id_p) |
0915bb0c | 2758 | { |
d62a17ae | 2759 | struct vrf *vrf; |
b72ede27 | 2760 | |
d62a17ae | 2761 | vrf = vrf_lookup_by_id(vrf_id); |
2762 | if (vrf) { | |
2763 | vrf = RB_NEXT(vrf_id_head, vrf); | |
2764 | if (vrf) { | |
2765 | *next_id_p = vrf->vrf_id; | |
2766 | return 1; | |
2767 | } | |
2768 | } | |
0915bb0c | 2769 | |
d62a17ae | 2770 | return 0; |
0915bb0c AS |
2771 | } |
2772 | ||
2773 | /* | |
2774 | * rib_tables_iter_next | |
2775 | * | |
2776 | * Returns the next table in the iteration. | |
2777 | */ | |
d62a17ae | 2778 | struct route_table *rib_tables_iter_next(rib_tables_iter_t *iter) |
2779 | { | |
2780 | struct route_table *table; | |
2781 | ||
2782 | /* | |
2783 | * Array that helps us go over all AFI/SAFI combinations via one | |
2784 | * index. | |
2785 | */ | |
2786 | static struct { | |
2787 | afi_t afi; | |
2788 | safi_t safi; | |
2789 | } afi_safis[] = { | |
2790 | {AFI_IP, SAFI_UNICAST}, {AFI_IP, SAFI_MULTICAST}, | |
2791 | {AFI_IP, SAFI_LABELED_UNICAST}, {AFI_IP6, SAFI_UNICAST}, | |
2792 | {AFI_IP6, SAFI_MULTICAST}, {AFI_IP6, SAFI_LABELED_UNICAST}, | |
2793 | }; | |
2794 | ||
2795 | table = NULL; | |
2796 | ||
2797 | switch (iter->state) { | |
2798 | ||
2799 | case RIB_TABLES_ITER_S_INIT: | |
2800 | iter->vrf_id = VRF_DEFAULT; | |
2801 | iter->afi_safi_ix = -1; | |
2802 | ||
2803 | /* Fall through */ | |
2804 | ||
2805 | case RIB_TABLES_ITER_S_ITERATING: | |
2806 | iter->afi_safi_ix++; | |
2807 | while (1) { | |
2808 | ||
2809 | while (iter->afi_safi_ix | |
2810 | < (int)ZEBRA_NUM_OF(afi_safis)) { | |
2811 | table = zebra_vrf_table( | |
2812 | afi_safis[iter->afi_safi_ix].afi, | |
2813 | afi_safis[iter->afi_safi_ix].safi, | |
2814 | iter->vrf_id); | |
2815 | if (table) | |
2816 | break; | |
2817 | ||
2818 | iter->afi_safi_ix++; | |
2819 | } | |
2820 | ||
2821 | /* | |
2822 | * Found another table in this vrf. | |
2823 | */ | |
2824 | if (table) | |
2825 | break; | |
2826 | ||
2827 | /* | |
2828 | * Done with all tables in the current vrf, go to the | |
2829 | * next | |
2830 | * one. | |
2831 | */ | |
2832 | if (!vrf_id_get_next(iter->vrf_id, &iter->vrf_id)) | |
2833 | break; | |
2834 | ||
2835 | iter->afi_safi_ix = 0; | |
2836 | } | |
0915bb0c | 2837 | |
0915bb0c AS |
2838 | break; |
2839 | ||
d62a17ae | 2840 | case RIB_TABLES_ITER_S_DONE: |
2841 | return NULL; | |
0915bb0c AS |
2842 | } |
2843 | ||
d62a17ae | 2844 | if (table) |
2845 | iter->state = RIB_TABLES_ITER_S_ITERATING; | |
2846 | else | |
2847 | iter->state = RIB_TABLES_ITER_S_DONE; | |
0915bb0c | 2848 | |
d62a17ae | 2849 | return table; |
0915bb0c | 2850 | } |