]>
Commit | Line | Data |
---|---|---|
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 | * | |
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 | |
19 | */ | |
20 | ||
21 | #include <zebra.h> | |
22 | ||
23 | #include "command.h" | |
24 | #include "if.h" | |
25 | #include "linklist.h" | |
26 | #include "log.h" | |
27 | #include "log_int.h" | |
28 | #include "memory.h" | |
29 | #include "mpls.h" | |
30 | #include "nexthop.h" | |
31 | #include "prefix.h" | |
32 | #include "prefix.h" | |
33 | #include "routemap.h" | |
34 | #include "sockunion.h" | |
35 | #include "srcdest_table.h" | |
36 | #include "table.h" | |
37 | #include "thread.h" | |
38 | #include "vrf.h" | |
39 | #include "workqueue.h" | |
40 | ||
41 | #include "zebra/connected.h" | |
42 | #include "zebra/debug.h" | |
43 | #include "zebra/interface.h" | |
44 | #include "zebra/redistribute.h" | |
45 | #include "zebra/rib.h" | |
46 | #include "zebra/rt.h" | |
47 | #include "zebra/zapi_msg.h" | |
48 | #include "zebra/zebra_errors.h" | |
49 | #include "zebra/zebra_memory.h" | |
50 | #include "zebra/zebra_ns.h" | |
51 | #include "zebra/zebra_rnh.h" | |
52 | #include "zebra/zebra_routemap.h" | |
53 | #include "zebra/zebra_vrf.h" | |
54 | #include "zebra/zebra_vxlan.h" | |
55 | ||
56 | DEFINE_HOOK(rib_update, (struct route_node * rn, const char *reason), | |
57 | (rn, reason)) | |
58 | ||
59 | /* Should we allow non Quagga processes to delete our routes */ | |
60 | extern int allow_delete; | |
61 | ||
62 | /* Each route type's string and default distance value. */ | |
63 | static const struct { | |
64 | int key; | |
65 | int distance; | |
66 | } route_info[ZEBRA_ROUTE_MAX] = { | |
67 | [ZEBRA_ROUTE_SYSTEM] = {ZEBRA_ROUTE_SYSTEM, 0}, | |
68 | [ZEBRA_ROUTE_KERNEL] = {ZEBRA_ROUTE_KERNEL, 0}, | |
69 | [ZEBRA_ROUTE_CONNECT] = {ZEBRA_ROUTE_CONNECT, 0}, | |
70 | [ZEBRA_ROUTE_STATIC] = {ZEBRA_ROUTE_STATIC, 1}, | |
71 | [ZEBRA_ROUTE_RIP] = {ZEBRA_ROUTE_RIP, 120}, | |
72 | [ZEBRA_ROUTE_RIPNG] = {ZEBRA_ROUTE_RIPNG, 120}, | |
73 | [ZEBRA_ROUTE_OSPF] = {ZEBRA_ROUTE_OSPF, 110}, | |
74 | [ZEBRA_ROUTE_OSPF6] = {ZEBRA_ROUTE_OSPF6, 110}, | |
75 | [ZEBRA_ROUTE_ISIS] = {ZEBRA_ROUTE_ISIS, 115}, | |
76 | [ZEBRA_ROUTE_BGP] = {ZEBRA_ROUTE_BGP, 20 /* IBGP is 200. */}, | |
77 | [ZEBRA_ROUTE_PIM] = {ZEBRA_ROUTE_PIM, 255}, | |
78 | [ZEBRA_ROUTE_EIGRP] = {ZEBRA_ROUTE_EIGRP, 90}, | |
79 | [ZEBRA_ROUTE_NHRP] = {ZEBRA_ROUTE_NHRP, 10}, | |
80 | [ZEBRA_ROUTE_HSLS] = {ZEBRA_ROUTE_HSLS, 255}, | |
81 | [ZEBRA_ROUTE_OLSR] = {ZEBRA_ROUTE_OLSR, 255}, | |
82 | [ZEBRA_ROUTE_TABLE] = {ZEBRA_ROUTE_TABLE, 150}, | |
83 | [ZEBRA_ROUTE_LDP] = {ZEBRA_ROUTE_LDP, 150}, | |
84 | [ZEBRA_ROUTE_VNC] = {ZEBRA_ROUTE_VNC, 20}, | |
85 | [ZEBRA_ROUTE_VNC_DIRECT] = {ZEBRA_ROUTE_VNC_DIRECT, 20}, | |
86 | [ZEBRA_ROUTE_VNC_DIRECT_RH] = {ZEBRA_ROUTE_VNC_DIRECT_RH, 20}, | |
87 | [ZEBRA_ROUTE_BGP_DIRECT] = {ZEBRA_ROUTE_BGP_DIRECT, 20}, | |
88 | [ZEBRA_ROUTE_BGP_DIRECT_EXT] = {ZEBRA_ROUTE_BGP_DIRECT_EXT, 20}, | |
89 | [ZEBRA_ROUTE_BABEL] = {ZEBRA_ROUTE_BABEL, 100}, | |
90 | [ZEBRA_ROUTE_SHARP] = {ZEBRA_ROUTE_SHARP, 150}, | |
91 | ||
92 | /* no entry/default: 150 */ | |
93 | }; | |
94 | ||
95 | /* RPF lookup behaviour */ | |
96 | static enum multicast_mode ipv4_multicast_mode = MCAST_NO_CONFIG; | |
97 | ||
98 | ||
99 | static void __attribute__((format(printf, 5, 6))) | |
100 | _rnode_zlog(const char *_func, vrf_id_t vrf_id, struct route_node *rn, | |
101 | int priority, const char *msgfmt, ...) | |
102 | { | |
103 | char buf[SRCDEST2STR_BUFFER + sizeof(" (MRIB)")]; | |
104 | char msgbuf[512]; | |
105 | va_list ap; | |
106 | ||
107 | va_start(ap, msgfmt); | |
108 | vsnprintf(msgbuf, sizeof(msgbuf), msgfmt, ap); | |
109 | va_end(ap); | |
110 | ||
111 | if (rn) { | |
112 | rib_table_info_t *info = srcdest_rnode_table_info(rn); | |
113 | srcdest_rnode2str(rn, buf, sizeof(buf)); | |
114 | ||
115 | if (info->safi == SAFI_MULTICAST) | |
116 | strcat(buf, " (MRIB)"); | |
117 | } else { | |
118 | snprintf(buf, sizeof(buf), "{(route_node *) NULL}"); | |
119 | } | |
120 | ||
121 | zlog(priority, "%s: %d:%s: %s", _func, vrf_id, buf, msgbuf); | |
122 | } | |
123 | ||
124 | #define rnode_debug(node, vrf_id, ...) \ | |
125 | _rnode_zlog(__func__, vrf_id, node, LOG_DEBUG, __VA_ARGS__) | |
126 | #define rnode_info(node, ...) \ | |
127 | _rnode_zlog(__func__, vrf_id, node, LOG_INFO, __VA_ARGS__) | |
128 | ||
129 | uint8_t route_distance(int type) | |
130 | { | |
131 | uint8_t distance; | |
132 | ||
133 | if ((unsigned)type >= array_size(route_info)) | |
134 | distance = 150; | |
135 | else | |
136 | distance = route_info[type].distance; | |
137 | ||
138 | return distance; | |
139 | } | |
140 | ||
141 | int is_zebra_valid_kernel_table(uint32_t table_id) | |
142 | { | |
143 | #ifdef linux | |
144 | if ((table_id == RT_TABLE_UNSPEC) || (table_id == RT_TABLE_LOCAL) | |
145 | || (table_id == RT_TABLE_COMPAT)) | |
146 | return 0; | |
147 | #endif | |
148 | ||
149 | return 1; | |
150 | } | |
151 | ||
152 | int is_zebra_main_routing_table(uint32_t table_id) | |
153 | { | |
154 | if ((table_id == RT_TABLE_MAIN) | |
155 | || (table_id == zebrad.rtm_table_default)) | |
156 | return 1; | |
157 | return 0; | |
158 | } | |
159 | ||
160 | int zebra_check_addr(const struct prefix *p) | |
161 | { | |
162 | if (p->family == AF_INET) { | |
163 | uint32_t addr; | |
164 | ||
165 | addr = p->u.prefix4.s_addr; | |
166 | addr = ntohl(addr); | |
167 | ||
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; | |
179 | } | |
180 | ||
181 | /* Add nexthop to the end of a rib node's nexthop list */ | |
182 | void route_entry_nexthop_add(struct route_entry *re, struct nexthop *nexthop) | |
183 | { | |
184 | nexthop_add(&re->ng.nexthop, nexthop); | |
185 | re->nexthop_num++; | |
186 | } | |
187 | ||
188 | ||
189 | /** | |
190 | * copy_nexthop - copy a nexthop to the rib structure. | |
191 | */ | |
192 | void route_entry_copy_nexthops(struct route_entry *re, struct nexthop *nh) | |
193 | { | |
194 | assert(!re->ng.nexthop); | |
195 | copy_nexthops(&re->ng.nexthop, nh, NULL); | |
196 | for (struct nexthop *nexthop = nh; nexthop; nexthop = nexthop->next) | |
197 | re->nexthop_num++; | |
198 | } | |
199 | ||
200 | /* Delete specified nexthop from the list. */ | |
201 | void route_entry_nexthop_delete(struct route_entry *re, struct nexthop *nexthop) | |
202 | { | |
203 | if (nexthop->next) | |
204 | nexthop->next->prev = nexthop->prev; | |
205 | if (nexthop->prev) | |
206 | nexthop->prev->next = nexthop->next; | |
207 | else | |
208 | re->ng.nexthop = nexthop->next; | |
209 | re->nexthop_num--; | |
210 | } | |
211 | ||
212 | ||
213 | struct nexthop *route_entry_nexthop_ifindex_add(struct route_entry *re, | |
214 | ifindex_t ifindex, | |
215 | vrf_id_t nh_vrf_id) | |
216 | { | |
217 | struct nexthop *nexthop; | |
218 | ||
219 | nexthop = nexthop_new(); | |
220 | nexthop->type = NEXTHOP_TYPE_IFINDEX; | |
221 | nexthop->ifindex = ifindex; | |
222 | nexthop->vrf_id = nh_vrf_id; | |
223 | ||
224 | route_entry_nexthop_add(re, nexthop); | |
225 | ||
226 | return nexthop; | |
227 | } | |
228 | ||
229 | struct nexthop *route_entry_nexthop_ipv4_add(struct route_entry *re, | |
230 | struct in_addr *ipv4, | |
231 | struct in_addr *src, | |
232 | vrf_id_t nh_vrf_id) | |
233 | { | |
234 | struct nexthop *nexthop; | |
235 | ||
236 | nexthop = nexthop_new(); | |
237 | nexthop->type = NEXTHOP_TYPE_IPV4; | |
238 | nexthop->vrf_id = nh_vrf_id; | |
239 | nexthop->gate.ipv4 = *ipv4; | |
240 | if (src) | |
241 | nexthop->src.ipv4 = *src; | |
242 | ||
243 | route_entry_nexthop_add(re, nexthop); | |
244 | ||
245 | return nexthop; | |
246 | } | |
247 | ||
248 | struct nexthop *route_entry_nexthop_ipv4_ifindex_add(struct route_entry *re, | |
249 | struct in_addr *ipv4, | |
250 | struct in_addr *src, | |
251 | ifindex_t ifindex, | |
252 | vrf_id_t nh_vrf_id) | |
253 | { | |
254 | struct nexthop *nexthop; | |
255 | struct interface *ifp; | |
256 | ||
257 | nexthop = nexthop_new(); | |
258 | nexthop->vrf_id = nh_vrf_id; | |
259 | nexthop->type = NEXTHOP_TYPE_IPV4_IFINDEX; | |
260 | nexthop->gate.ipv4 = *ipv4; | |
261 | if (src) | |
262 | nexthop->src.ipv4 = *src; | |
263 | nexthop->ifindex = ifindex; | |
264 | ifp = if_lookup_by_index(nexthop->ifindex, nh_vrf_id); | |
265 | /*Pending: need to think if null ifp here is ok during bootup? | |
266 | There was a crash because ifp here was coming to be NULL */ | |
267 | if (ifp) | |
268 | if (connected_is_unnumbered(ifp) | |
269 | || CHECK_FLAG(re->flags, ZEBRA_FLAG_EVPN_ROUTE)) { | |
270 | SET_FLAG(nexthop->flags, NEXTHOP_FLAG_ONLINK); | |
271 | } | |
272 | ||
273 | route_entry_nexthop_add(re, nexthop); | |
274 | ||
275 | return nexthop; | |
276 | } | |
277 | ||
278 | struct nexthop *route_entry_nexthop_ipv6_add(struct route_entry *re, | |
279 | struct in6_addr *ipv6, | |
280 | vrf_id_t nh_vrf_id) | |
281 | { | |
282 | struct nexthop *nexthop; | |
283 | ||
284 | nexthop = nexthop_new(); | |
285 | nexthop->vrf_id = nh_vrf_id; | |
286 | nexthop->type = NEXTHOP_TYPE_IPV6; | |
287 | nexthop->gate.ipv6 = *ipv6; | |
288 | ||
289 | route_entry_nexthop_add(re, nexthop); | |
290 | ||
291 | return nexthop; | |
292 | } | |
293 | ||
294 | struct nexthop *route_entry_nexthop_ipv6_ifindex_add(struct route_entry *re, | |
295 | struct in6_addr *ipv6, | |
296 | ifindex_t ifindex, | |
297 | vrf_id_t nh_vrf_id) | |
298 | { | |
299 | struct nexthop *nexthop; | |
300 | ||
301 | nexthop = nexthop_new(); | |
302 | nexthop->vrf_id = nh_vrf_id; | |
303 | nexthop->type = NEXTHOP_TYPE_IPV6_IFINDEX; | |
304 | nexthop->gate.ipv6 = *ipv6; | |
305 | nexthop->ifindex = ifindex; | |
306 | if (CHECK_FLAG(re->flags, ZEBRA_FLAG_EVPN_ROUTE)) | |
307 | SET_FLAG(nexthop->flags, NEXTHOP_FLAG_ONLINK); | |
308 | ||
309 | route_entry_nexthop_add(re, nexthop); | |
310 | ||
311 | return nexthop; | |
312 | } | |
313 | ||
314 | struct nexthop *route_entry_nexthop_blackhole_add(struct route_entry *re, | |
315 | enum blackhole_type bh_type) | |
316 | { | |
317 | struct nexthop *nexthop; | |
318 | ||
319 | nexthop = nexthop_new(); | |
320 | nexthop->vrf_id = VRF_DEFAULT; | |
321 | nexthop->type = NEXTHOP_TYPE_BLACKHOLE; | |
322 | nexthop->bh_type = bh_type; | |
323 | ||
324 | route_entry_nexthop_add(re, nexthop); | |
325 | ||
326 | return nexthop; | |
327 | } | |
328 | ||
329 | static void nexthop_set_resolved(afi_t afi, const struct nexthop *newhop, | |
330 | struct nexthop *nexthop) | |
331 | { | |
332 | struct nexthop *resolved_hop; | |
333 | ||
334 | resolved_hop = nexthop_new(); | |
335 | SET_FLAG(resolved_hop->flags, NEXTHOP_FLAG_ACTIVE); | |
336 | ||
337 | resolved_hop->vrf_id = nexthop->vrf_id; | |
338 | switch (newhop->type) { | |
339 | case NEXTHOP_TYPE_IPV4: | |
340 | case NEXTHOP_TYPE_IPV4_IFINDEX: | |
341 | /* If the resolving route specifies a gateway, use it */ | |
342 | resolved_hop->type = newhop->type; | |
343 | resolved_hop->gate.ipv4 = newhop->gate.ipv4; | |
344 | ||
345 | if (newhop->ifindex) { | |
346 | resolved_hop->type = NEXTHOP_TYPE_IPV4_IFINDEX; | |
347 | resolved_hop->ifindex = newhop->ifindex; | |
348 | if (newhop->flags & NEXTHOP_FLAG_ONLINK) | |
349 | resolved_hop->flags |= NEXTHOP_FLAG_ONLINK; | |
350 | } | |
351 | break; | |
352 | case NEXTHOP_TYPE_IPV6: | |
353 | case NEXTHOP_TYPE_IPV6_IFINDEX: | |
354 | resolved_hop->type = newhop->type; | |
355 | resolved_hop->gate.ipv6 = newhop->gate.ipv6; | |
356 | ||
357 | if (newhop->ifindex) { | |
358 | resolved_hop->type = NEXTHOP_TYPE_IPV6_IFINDEX; | |
359 | resolved_hop->ifindex = newhop->ifindex; | |
360 | } | |
361 | break; | |
362 | case NEXTHOP_TYPE_IFINDEX: | |
363 | /* If the resolving route is an interface route, | |
364 | * it means the gateway we are looking up is connected | |
365 | * to that interface. (The actual network is _not_ onlink). | |
366 | * Therefore, the resolved route should have the original | |
367 | * gateway as nexthop as it is directly connected. | |
368 | * | |
369 | * On Linux, we have to set the onlink netlink flag because | |
370 | * otherwise, the kernel won't accept the route. | |
371 | */ | |
372 | resolved_hop->flags |= NEXTHOP_FLAG_ONLINK; | |
373 | if (afi == AFI_IP) { | |
374 | resolved_hop->type = NEXTHOP_TYPE_IPV4_IFINDEX; | |
375 | resolved_hop->gate.ipv4 = nexthop->gate.ipv4; | |
376 | } else if (afi == AFI_IP6) { | |
377 | resolved_hop->type = NEXTHOP_TYPE_IPV6_IFINDEX; | |
378 | resolved_hop->gate.ipv6 = nexthop->gate.ipv6; | |
379 | } | |
380 | resolved_hop->ifindex = newhop->ifindex; | |
381 | break; | |
382 | case NEXTHOP_TYPE_BLACKHOLE: | |
383 | resolved_hop->type = NEXTHOP_TYPE_BLACKHOLE; | |
384 | resolved_hop->bh_type = nexthop->bh_type; | |
385 | break; | |
386 | } | |
387 | ||
388 | /* Copy labels of the resolved route */ | |
389 | if (newhop->nh_label) | |
390 | nexthop_add_labels(resolved_hop, newhop->nh_label_type, | |
391 | newhop->nh_label->num_labels, | |
392 | &newhop->nh_label->label[0]); | |
393 | ||
394 | resolved_hop->rparent = nexthop; | |
395 | nexthop_add(&nexthop->resolved, resolved_hop); | |
396 | } | |
397 | ||
398 | /* If force flag is not set, do not modify falgs at all for uninstall | |
399 | the route from FIB. */ | |
400 | static int nexthop_active(afi_t afi, struct route_entry *re, | |
401 | struct nexthop *nexthop, int set, | |
402 | struct route_node *top) | |
403 | { | |
404 | struct prefix p; | |
405 | struct route_table *table; | |
406 | struct route_node *rn; | |
407 | struct route_entry *match = NULL; | |
408 | int resolved; | |
409 | struct nexthop *newhop; | |
410 | struct interface *ifp; | |
411 | rib_dest_t *dest; | |
412 | ||
413 | if ((nexthop->type == NEXTHOP_TYPE_IPV4) | |
414 | || nexthop->type == NEXTHOP_TYPE_IPV6) | |
415 | nexthop->ifindex = 0; | |
416 | ||
417 | if (set) { | |
418 | UNSET_FLAG(nexthop->flags, NEXTHOP_FLAG_RECURSIVE); | |
419 | nexthops_free(nexthop->resolved); | |
420 | nexthop->resolved = NULL; | |
421 | re->nexthop_mtu = 0; | |
422 | } | |
423 | ||
424 | /* Next hops (remote VTEPs) for EVPN routes are fully resolved. */ | |
425 | if (CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_EVPN_RVTEP)) | |
426 | return 1; | |
427 | ||
428 | /* Skip nexthops that have been filtered out due to route-map */ | |
429 | /* The nexthops are specific to this route and so the same */ | |
430 | /* nexthop for a different route may not have this flag set */ | |
431 | if (CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_FILTERED)) | |
432 | return 0; | |
433 | ||
434 | /* | |
435 | * Check to see if we should trust the passed in information | |
436 | * for UNNUMBERED interfaces as that we won't find the GW | |
437 | * address in the routing table. | |
438 | */ | |
439 | if (CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_ONLINK)) { | |
440 | ifp = if_lookup_by_index(nexthop->ifindex, nexthop->vrf_id); | |
441 | if (ifp && connected_is_unnumbered(ifp)) { | |
442 | if (if_is_operative(ifp)) | |
443 | return 1; | |
444 | else | |
445 | return 0; | |
446 | } else | |
447 | return 0; | |
448 | } | |
449 | ||
450 | /* Make lookup prefix. */ | |
451 | memset(&p, 0, sizeof(struct prefix)); | |
452 | switch (afi) { | |
453 | case AFI_IP: | |
454 | p.family = AF_INET; | |
455 | p.prefixlen = IPV4_MAX_PREFIXLEN; | |
456 | p.u.prefix4 = nexthop->gate.ipv4; | |
457 | break; | |
458 | case AFI_IP6: | |
459 | p.family = AF_INET6; | |
460 | p.prefixlen = IPV6_MAX_PREFIXLEN; | |
461 | p.u.prefix6 = nexthop->gate.ipv6; | |
462 | break; | |
463 | default: | |
464 | assert(afi != AFI_IP && afi != AFI_IP6); | |
465 | break; | |
466 | } | |
467 | /* Lookup table. */ | |
468 | table = zebra_vrf_table(afi, SAFI_UNICAST, nexthop->vrf_id); | |
469 | if (!table) | |
470 | return 0; | |
471 | ||
472 | rn = route_node_match(table, (struct prefix *)&p); | |
473 | while (rn) { | |
474 | route_unlock_node(rn); | |
475 | ||
476 | /* Lookup should halt if we've matched against ourselves ('top', | |
477 | * if specified) - i.e., we cannot have a nexthop NH1 is | |
478 | * resolved by a route NH1. The exception is if the route is a | |
479 | * host route. | |
480 | */ | |
481 | if (top && rn == top) | |
482 | if (((afi == AFI_IP) && (rn->p.prefixlen != 32)) | |
483 | || ((afi == AFI_IP6) && (rn->p.prefixlen != 128))) | |
484 | return 0; | |
485 | ||
486 | /* Pick up selected route. */ | |
487 | /* However, do not resolve over default route unless explicitly | |
488 | * allowed. */ | |
489 | if (is_default_prefix(&rn->p) | |
490 | && !rnh_resolve_via_default(p.family)) | |
491 | return 0; | |
492 | ||
493 | dest = rib_dest_from_rnode(rn); | |
494 | if (dest && dest->selected_fib | |
495 | && !CHECK_FLAG(dest->selected_fib->status, | |
496 | ROUTE_ENTRY_REMOVED) | |
497 | && dest->selected_fib->type != ZEBRA_ROUTE_TABLE) | |
498 | match = dest->selected_fib; | |
499 | ||
500 | /* If there is no selected route or matched route is EGP, go up | |
501 | tree. */ | |
502 | if (!match) { | |
503 | do { | |
504 | rn = rn->parent; | |
505 | } while (rn && rn->info == NULL); | |
506 | if (rn) | |
507 | route_lock_node(rn); | |
508 | ||
509 | continue; | |
510 | } | |
511 | ||
512 | if (match->type == ZEBRA_ROUTE_CONNECT) { | |
513 | /* Directly point connected route. */ | |
514 | newhop = match->ng.nexthop; | |
515 | if (newhop) { | |
516 | if (nexthop->type == NEXTHOP_TYPE_IPV4 | |
517 | || nexthop->type == NEXTHOP_TYPE_IPV6) | |
518 | nexthop->ifindex = newhop->ifindex; | |
519 | } | |
520 | return 1; | |
521 | } else if (CHECK_FLAG(re->flags, ZEBRA_FLAG_ALLOW_RECURSION)) { | |
522 | resolved = 0; | |
523 | for (ALL_NEXTHOPS(match->ng, newhop)) { | |
524 | if (!CHECK_FLAG(newhop->flags, | |
525 | NEXTHOP_FLAG_FIB)) | |
526 | continue; | |
527 | if (CHECK_FLAG(newhop->flags, | |
528 | NEXTHOP_FLAG_RECURSIVE)) | |
529 | continue; | |
530 | ||
531 | if (set) { | |
532 | SET_FLAG(nexthop->flags, | |
533 | NEXTHOP_FLAG_RECURSIVE); | |
534 | SET_FLAG(re->status, | |
535 | ROUTE_ENTRY_NEXTHOPS_CHANGED); | |
536 | nexthop_set_resolved(afi, newhop, | |
537 | nexthop); | |
538 | } | |
539 | resolved = 1; | |
540 | } | |
541 | if (resolved && set) | |
542 | re->nexthop_mtu = match->mtu; | |
543 | return resolved; | |
544 | } else if (re->type == ZEBRA_ROUTE_STATIC) { | |
545 | resolved = 0; | |
546 | for (ALL_NEXTHOPS(match->ng, newhop)) { | |
547 | if (!CHECK_FLAG(newhop->flags, | |
548 | NEXTHOP_FLAG_FIB)) | |
549 | continue; | |
550 | ||
551 | if (set) { | |
552 | SET_FLAG(nexthop->flags, | |
553 | NEXTHOP_FLAG_RECURSIVE); | |
554 | nexthop_set_resolved(afi, newhop, | |
555 | nexthop); | |
556 | } | |
557 | resolved = 1; | |
558 | } | |
559 | if (resolved && set) | |
560 | re->nexthop_mtu = match->mtu; | |
561 | return resolved; | |
562 | } else { | |
563 | return 0; | |
564 | } | |
565 | } | |
566 | return 0; | |
567 | } | |
568 | ||
569 | struct route_entry *rib_match(afi_t afi, safi_t safi, vrf_id_t vrf_id, | |
570 | union g_addr *addr, struct route_node **rn_out) | |
571 | { | |
572 | struct prefix p; | |
573 | struct route_table *table; | |
574 | struct route_node *rn; | |
575 | struct route_entry *match = NULL; | |
576 | struct nexthop *newhop; | |
577 | ||
578 | /* Lookup table. */ | |
579 | table = zebra_vrf_table(afi, safi, vrf_id); | |
580 | if (!table) | |
581 | return 0; | |
582 | ||
583 | memset(&p, 0, sizeof(struct prefix)); | |
584 | p.family = afi; | |
585 | if (afi == AFI_IP) { | |
586 | p.u.prefix4 = addr->ipv4; | |
587 | p.prefixlen = IPV4_MAX_PREFIXLEN; | |
588 | } else { | |
589 | p.u.prefix6 = addr->ipv6; | |
590 | p.prefixlen = IPV6_MAX_PREFIXLEN; | |
591 | } | |
592 | ||
593 | rn = route_node_match(table, (struct prefix *)&p); | |
594 | ||
595 | while (rn) { | |
596 | rib_dest_t *dest; | |
597 | ||
598 | route_unlock_node(rn); | |
599 | ||
600 | dest = rib_dest_from_rnode(rn); | |
601 | if (dest && dest->selected_fib | |
602 | && !CHECK_FLAG(dest->selected_fib->status, | |
603 | ROUTE_ENTRY_REMOVED)) | |
604 | match = dest->selected_fib; | |
605 | ||
606 | /* If there is no selected route or matched route is EGP, go up | |
607 | tree. */ | |
608 | if (!match) { | |
609 | do { | |
610 | rn = rn->parent; | |
611 | } while (rn && rn->info == NULL); | |
612 | if (rn) | |
613 | route_lock_node(rn); | |
614 | } else { | |
615 | if (match->type != ZEBRA_ROUTE_CONNECT) { | |
616 | int found = 0; | |
617 | for (ALL_NEXTHOPS(match->ng, newhop)) | |
618 | if (CHECK_FLAG(newhop->flags, | |
619 | NEXTHOP_FLAG_FIB)) { | |
620 | found = 1; | |
621 | break; | |
622 | } | |
623 | if (!found) | |
624 | return NULL; | |
625 | } | |
626 | ||
627 | if (rn_out) | |
628 | *rn_out = rn; | |
629 | return match; | |
630 | } | |
631 | } | |
632 | return NULL; | |
633 | } | |
634 | ||
635 | struct route_entry *rib_match_ipv4_multicast(vrf_id_t vrf_id, | |
636 | struct in_addr addr, | |
637 | struct route_node **rn_out) | |
638 | { | |
639 | struct route_entry *re = NULL, *mre = NULL, *ure = NULL; | |
640 | struct route_node *m_rn = NULL, *u_rn = NULL; | |
641 | union g_addr gaddr = {.ipv4 = addr}; | |
642 | ||
643 | switch (ipv4_multicast_mode) { | |
644 | case MCAST_MRIB_ONLY: | |
645 | return rib_match(AFI_IP, SAFI_MULTICAST, vrf_id, &gaddr, | |
646 | rn_out); | |
647 | case MCAST_URIB_ONLY: | |
648 | return rib_match(AFI_IP, SAFI_UNICAST, vrf_id, &gaddr, rn_out); | |
649 | case MCAST_NO_CONFIG: | |
650 | case MCAST_MIX_MRIB_FIRST: | |
651 | re = mre = rib_match(AFI_IP, SAFI_MULTICAST, vrf_id, &gaddr, | |
652 | &m_rn); | |
653 | if (!mre) | |
654 | re = ure = rib_match(AFI_IP, SAFI_UNICAST, vrf_id, | |
655 | &gaddr, &u_rn); | |
656 | break; | |
657 | case MCAST_MIX_DISTANCE: | |
658 | mre = rib_match(AFI_IP, SAFI_MULTICAST, vrf_id, &gaddr, &m_rn); | |
659 | ure = rib_match(AFI_IP, SAFI_UNICAST, vrf_id, &gaddr, &u_rn); | |
660 | if (mre && ure) | |
661 | re = ure->distance < mre->distance ? ure : mre; | |
662 | else if (mre) | |
663 | re = mre; | |
664 | else if (ure) | |
665 | re = ure; | |
666 | break; | |
667 | case MCAST_MIX_PFXLEN: | |
668 | mre = rib_match(AFI_IP, SAFI_MULTICAST, vrf_id, &gaddr, &m_rn); | |
669 | ure = rib_match(AFI_IP, SAFI_UNICAST, vrf_id, &gaddr, &u_rn); | |
670 | if (mre && ure) | |
671 | re = u_rn->p.prefixlen > m_rn->p.prefixlen ? ure : mre; | |
672 | else if (mre) | |
673 | re = mre; | |
674 | else if (ure) | |
675 | re = ure; | |
676 | break; | |
677 | } | |
678 | ||
679 | if (rn_out) | |
680 | *rn_out = (re == mre) ? m_rn : u_rn; | |
681 | ||
682 | if (IS_ZEBRA_DEBUG_RIB) { | |
683 | char buf[BUFSIZ]; | |
684 | inet_ntop(AF_INET, &addr, buf, BUFSIZ); | |
685 | ||
686 | zlog_debug("%s: %s: vrf: %u found %s, using %s", | |
687 | __func__, buf, vrf_id, | |
688 | mre ? (ure ? "MRIB+URIB" : "MRIB") | |
689 | : ure ? "URIB" : "nothing", | |
690 | re == ure ? "URIB" : re == mre ? "MRIB" : "none"); | |
691 | } | |
692 | return re; | |
693 | } | |
694 | ||
695 | void multicast_mode_ipv4_set(enum multicast_mode mode) | |
696 | { | |
697 | if (IS_ZEBRA_DEBUG_RIB) | |
698 | zlog_debug("%s: multicast lookup mode set (%d)", __func__, | |
699 | mode); | |
700 | ipv4_multicast_mode = mode; | |
701 | } | |
702 | ||
703 | enum multicast_mode multicast_mode_ipv4_get(void) | |
704 | { | |
705 | return ipv4_multicast_mode; | |
706 | } | |
707 | ||
708 | struct route_entry *rib_lookup_ipv4(struct prefix_ipv4 *p, vrf_id_t vrf_id) | |
709 | { | |
710 | struct route_table *table; | |
711 | struct route_node *rn; | |
712 | struct route_entry *match = NULL; | |
713 | struct nexthop *nexthop; | |
714 | rib_dest_t *dest; | |
715 | ||
716 | /* Lookup table. */ | |
717 | table = zebra_vrf_table(AFI_IP, SAFI_UNICAST, vrf_id); | |
718 | if (!table) | |
719 | return 0; | |
720 | ||
721 | rn = route_node_lookup(table, (struct prefix *)p); | |
722 | ||
723 | /* No route for this prefix. */ | |
724 | if (!rn) | |
725 | return NULL; | |
726 | ||
727 | /* Unlock node. */ | |
728 | route_unlock_node(rn); | |
729 | dest = rib_dest_from_rnode(rn); | |
730 | ||
731 | if (dest && dest->selected_fib | |
732 | && !CHECK_FLAG(dest->selected_fib->status, ROUTE_ENTRY_REMOVED)) | |
733 | match = dest->selected_fib; | |
734 | ||
735 | if (!match) | |
736 | return NULL; | |
737 | ||
738 | if (match->type == ZEBRA_ROUTE_CONNECT) | |
739 | return match; | |
740 | ||
741 | for (ALL_NEXTHOPS(match->ng, nexthop)) | |
742 | if (CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_FIB)) | |
743 | return match; | |
744 | ||
745 | return NULL; | |
746 | } | |
747 | ||
748 | /* | |
749 | * This clone function, unlike its original rib_lookup_ipv4(), checks | |
750 | * if specified IPv4 route record (prefix/mask -> gate) exists in | |
751 | * the whole RIB and has ROUTE_ENTRY_SELECTED_FIB set. | |
752 | * | |
753 | * Return values: | |
754 | * -1: error | |
755 | * 0: exact match found | |
756 | * 1: a match was found with a different gate | |
757 | * 2: connected route found | |
758 | * 3: no matches found | |
759 | */ | |
760 | int rib_lookup_ipv4_route(struct prefix_ipv4 *p, union sockunion *qgate, | |
761 | vrf_id_t vrf_id) | |
762 | { | |
763 | struct route_table *table; | |
764 | struct route_node *rn; | |
765 | struct route_entry *match = NULL; | |
766 | struct nexthop *nexthop; | |
767 | int nexthops_active; | |
768 | rib_dest_t *dest; | |
769 | ||
770 | /* Lookup table. */ | |
771 | table = zebra_vrf_table(AFI_IP, SAFI_UNICAST, vrf_id); | |
772 | if (!table) | |
773 | return ZEBRA_RIB_LOOKUP_ERROR; | |
774 | ||
775 | /* Scan the RIB table for exactly matching RIB entry. */ | |
776 | rn = route_node_lookup(table, (struct prefix *)p); | |
777 | ||
778 | /* No route for this prefix. */ | |
779 | if (!rn) | |
780 | return ZEBRA_RIB_NOTFOUND; | |
781 | ||
782 | /* Unlock node. */ | |
783 | route_unlock_node(rn); | |
784 | dest = rib_dest_from_rnode(rn); | |
785 | ||
786 | /* Find out if a "selected" RR for the discovered RIB entry exists ever. | |
787 | */ | |
788 | if (dest && dest->selected_fib | |
789 | && !CHECK_FLAG(dest->selected_fib->status, ROUTE_ENTRY_REMOVED)) | |
790 | match = dest->selected_fib; | |
791 | ||
792 | /* None such found :( */ | |
793 | if (!match) | |
794 | return ZEBRA_RIB_NOTFOUND; | |
795 | ||
796 | if (match->type == ZEBRA_ROUTE_CONNECT) | |
797 | return ZEBRA_RIB_FOUND_CONNECTED; | |
798 | ||
799 | /* Ok, we have a cood candidate, let's check it's nexthop list... */ | |
800 | nexthops_active = 0; | |
801 | for (ALL_NEXTHOPS(match->ng, nexthop)) | |
802 | if (CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_FIB)) { | |
803 | nexthops_active = 1; | |
804 | if (nexthop->gate.ipv4.s_addr == sockunion2ip(qgate)) | |
805 | return ZEBRA_RIB_FOUND_EXACT; | |
806 | if (IS_ZEBRA_DEBUG_RIB) { | |
807 | char gate_buf[INET_ADDRSTRLEN], | |
808 | qgate_buf[INET_ADDRSTRLEN]; | |
809 | inet_ntop(AF_INET, &nexthop->gate.ipv4.s_addr, | |
810 | gate_buf, INET_ADDRSTRLEN); | |
811 | inet_ntop(AF_INET, &sockunion2ip(qgate), | |
812 | qgate_buf, INET_ADDRSTRLEN); | |
813 | zlog_debug("%s: qgate == %s, %s == %s", | |
814 | __func__, qgate_buf, | |
815 | nexthop->rparent ? "rgate" : "gate", | |
816 | gate_buf); | |
817 | } | |
818 | } | |
819 | ||
820 | if (nexthops_active) | |
821 | return ZEBRA_RIB_FOUND_NOGATE; | |
822 | ||
823 | return ZEBRA_RIB_NOTFOUND; | |
824 | } | |
825 | ||
826 | #define RIB_SYSTEM_ROUTE(R) \ | |
827 | ((R)->type == ZEBRA_ROUTE_KERNEL || (R)->type == ZEBRA_ROUTE_CONNECT) | |
828 | ||
829 | /* This function verifies reachability of one given nexthop, which can be | |
830 | * numbered or unnumbered, IPv4 or IPv6. The result is unconditionally stored | |
831 | * in nexthop->flags field. If the 4th parameter, 'set', is non-zero, | |
832 | * nexthop->ifindex will be updated appropriately as well. | |
833 | * An existing route map can turn (otherwise active) nexthop into inactive, but | |
834 | * not vice versa. | |
835 | * | |
836 | * The return value is the final value of 'ACTIVE' flag. | |
837 | */ | |
838 | ||
839 | static unsigned nexthop_active_check(struct route_node *rn, | |
840 | struct route_entry *re, | |
841 | struct nexthop *nexthop, int set) | |
842 | { | |
843 | struct interface *ifp; | |
844 | route_map_result_t ret = RMAP_MATCH; | |
845 | int family; | |
846 | char buf[SRCDEST2STR_BUFFER]; | |
847 | const struct prefix *p, *src_p; | |
848 | srcdest_rnode_prefixes(rn, &p, &src_p); | |
849 | ||
850 | if (rn->p.family == AF_INET) | |
851 | family = AFI_IP; | |
852 | else if (rn->p.family == AF_INET6) | |
853 | family = AFI_IP6; | |
854 | else | |
855 | family = 0; | |
856 | switch (nexthop->type) { | |
857 | case NEXTHOP_TYPE_IFINDEX: | |
858 | ifp = if_lookup_by_index(nexthop->ifindex, nexthop->vrf_id); | |
859 | if (ifp && if_is_operative(ifp)) | |
860 | SET_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE); | |
861 | else | |
862 | UNSET_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE); | |
863 | break; | |
864 | case NEXTHOP_TYPE_IPV4: | |
865 | case NEXTHOP_TYPE_IPV4_IFINDEX: | |
866 | family = AFI_IP; | |
867 | if (nexthop_active(AFI_IP, re, nexthop, set, rn)) | |
868 | SET_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE); | |
869 | else | |
870 | UNSET_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE); | |
871 | break; | |
872 | case NEXTHOP_TYPE_IPV6: | |
873 | family = AFI_IP6; | |
874 | if (nexthop_active(AFI_IP6, re, nexthop, set, rn)) | |
875 | SET_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE); | |
876 | else | |
877 | UNSET_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE); | |
878 | break; | |
879 | case NEXTHOP_TYPE_IPV6_IFINDEX: | |
880 | /* RFC 5549, v4 prefix with v6 NH */ | |
881 | if (rn->p.family != AF_INET) | |
882 | family = AFI_IP6; | |
883 | if (IN6_IS_ADDR_LINKLOCAL(&nexthop->gate.ipv6)) { | |
884 | ifp = if_lookup_by_index(nexthop->ifindex, | |
885 | nexthop->vrf_id); | |
886 | if (ifp && if_is_operative(ifp)) | |
887 | SET_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE); | |
888 | else | |
889 | UNSET_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE); | |
890 | } else { | |
891 | if (nexthop_active(AFI_IP6, re, nexthop, set, rn)) | |
892 | SET_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE); | |
893 | else | |
894 | UNSET_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE); | |
895 | } | |
896 | break; | |
897 | case NEXTHOP_TYPE_BLACKHOLE: | |
898 | SET_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE); | |
899 | break; | |
900 | default: | |
901 | break; | |
902 | } | |
903 | if (!CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE)) | |
904 | return 0; | |
905 | ||
906 | /* XXX: What exactly do those checks do? Do we support | |
907 | * e.g. IPv4 routes with IPv6 nexthops or vice versa? */ | |
908 | if (RIB_SYSTEM_ROUTE(re) || (family == AFI_IP && p->family != AF_INET) | |
909 | || (family == AFI_IP6 && p->family != AF_INET6)) | |
910 | return CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE); | |
911 | ||
912 | /* The original code didn't determine the family correctly | |
913 | * e.g. for NEXTHOP_TYPE_IFINDEX. Retrieve the correct afi | |
914 | * from the rib_table_info in those cases. | |
915 | * Possibly it may be better to use only the rib_table_info | |
916 | * in every case. | |
917 | */ | |
918 | if (!family) { | |
919 | rib_table_info_t *info; | |
920 | ||
921 | info = srcdest_rnode_table_info(rn); | |
922 | family = info->afi; | |
923 | } | |
924 | ||
925 | memset(&nexthop->rmap_src.ipv6, 0, sizeof(union g_addr)); | |
926 | ||
927 | /* It'll get set if required inside */ | |
928 | ret = zebra_route_map_check(family, re->type, re->instance, p, nexthop, | |
929 | nexthop->vrf_id, re->tag); | |
930 | if (ret == RMAP_DENYMATCH) { | |
931 | if (IS_ZEBRA_DEBUG_RIB) { | |
932 | srcdest_rnode2str(rn, buf, sizeof(buf)); | |
933 | zlog_debug( | |
934 | "%u:%s: Filtering out with NH out %s due to route map", | |
935 | re->vrf_id, buf, | |
936 | ifindex2ifname(nexthop->ifindex, | |
937 | nexthop->vrf_id)); | |
938 | } | |
939 | UNSET_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE); | |
940 | } | |
941 | return CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE); | |
942 | } | |
943 | ||
944 | /* Iterate over all nexthops of the given RIB entry and refresh their | |
945 | * ACTIVE flag. re->nexthop_active_num is updated accordingly. If any | |
946 | * nexthop is found to toggle the ACTIVE flag, the whole re structure | |
947 | * is flagged with ROUTE_ENTRY_CHANGED. The 4th 'set' argument is | |
948 | * transparently passed to nexthop_active_check(). | |
949 | * | |
950 | * Return value is the new number of active nexthops. | |
951 | */ | |
952 | ||
953 | static int nexthop_active_update(struct route_node *rn, struct route_entry *re, | |
954 | int set) | |
955 | { | |
956 | struct nexthop *nexthop; | |
957 | union g_addr prev_src; | |
958 | unsigned int prev_active, new_active, old_num_nh; | |
959 | ifindex_t prev_index; | |
960 | old_num_nh = re->nexthop_active_num; | |
961 | ||
962 | re->nexthop_active_num = 0; | |
963 | UNSET_FLAG(re->status, ROUTE_ENTRY_CHANGED); | |
964 | ||
965 | for (nexthop = re->ng.nexthop; nexthop; nexthop = nexthop->next) { | |
966 | /* No protocol daemon provides src and so we're skipping | |
967 | * tracking it */ | |
968 | prev_src = nexthop->rmap_src; | |
969 | prev_active = CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE); | |
970 | prev_index = nexthop->ifindex; | |
971 | if ((new_active = nexthop_active_check(rn, re, nexthop, set))) | |
972 | re->nexthop_active_num++; | |
973 | /* Don't allow src setting on IPv6 addr for now */ | |
974 | if (prev_active != new_active || prev_index != nexthop->ifindex | |
975 | || ((nexthop->type >= NEXTHOP_TYPE_IFINDEX | |
976 | && nexthop->type < NEXTHOP_TYPE_IPV6) | |
977 | && prev_src.ipv4.s_addr | |
978 | != nexthop->rmap_src.ipv4.s_addr) | |
979 | || ((nexthop->type >= NEXTHOP_TYPE_IPV6 | |
980 | && nexthop->type < NEXTHOP_TYPE_BLACKHOLE) | |
981 | && !(IPV6_ADDR_SAME(&prev_src.ipv6, | |
982 | &nexthop->rmap_src.ipv6)))) { | |
983 | SET_FLAG(re->status, ROUTE_ENTRY_CHANGED); | |
984 | SET_FLAG(re->status, ROUTE_ENTRY_NEXTHOPS_CHANGED); | |
985 | } | |
986 | } | |
987 | ||
988 | if (old_num_nh != re->nexthop_active_num) | |
989 | SET_FLAG(re->status, ROUTE_ENTRY_CHANGED); | |
990 | ||
991 | if (CHECK_FLAG(re->status, ROUTE_ENTRY_CHANGED)) { | |
992 | SET_FLAG(re->status, ROUTE_ENTRY_NEXTHOPS_CHANGED); | |
993 | } | |
994 | ||
995 | return re->nexthop_active_num; | |
996 | } | |
997 | ||
998 | /* | |
999 | * Is this RIB labeled-unicast? It must be of type BGP and all paths | |
1000 | * (nexthops) must have a label. | |
1001 | */ | |
1002 | int zebra_rib_labeled_unicast(struct route_entry *re) | |
1003 | { | |
1004 | struct nexthop *nexthop = NULL; | |
1005 | ||
1006 | if (re->type != ZEBRA_ROUTE_BGP) | |
1007 | return 0; | |
1008 | ||
1009 | for (ALL_NEXTHOPS(re->ng, nexthop)) | |
1010 | if (!nexthop->nh_label || !nexthop->nh_label->num_labels) | |
1011 | return 0; | |
1012 | ||
1013 | return 1; | |
1014 | } | |
1015 | ||
1016 | void kernel_route_rib_pass_fail(struct route_node *rn, const struct prefix *p, | |
1017 | struct route_entry *re, | |
1018 | enum dp_results res) | |
1019 | { | |
1020 | struct nexthop *nexthop; | |
1021 | char buf[PREFIX_STRLEN]; | |
1022 | rib_dest_t *dest; | |
1023 | ||
1024 | dest = rib_dest_from_rnode(rn); | |
1025 | ||
1026 | switch (res) { | |
1027 | case DP_INSTALL_SUCCESS: | |
1028 | dest->selected_fib = re; | |
1029 | for (ALL_NEXTHOPS(re->ng, nexthop)) { | |
1030 | if (CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_RECURSIVE)) | |
1031 | continue; | |
1032 | ||
1033 | if (CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE)) | |
1034 | SET_FLAG(nexthop->flags, NEXTHOP_FLAG_FIB); | |
1035 | else | |
1036 | UNSET_FLAG(nexthop->flags, NEXTHOP_FLAG_FIB); | |
1037 | } | |
1038 | zsend_route_notify_owner(re, p, ZAPI_ROUTE_INSTALLED); | |
1039 | break; | |
1040 | case DP_INSTALL_FAILURE: | |
1041 | /* | |
1042 | * I am not sure this is the right thing to do here | |
1043 | * but the code always set selected_fib before | |
1044 | * this assignment was moved here. | |
1045 | */ | |
1046 | dest->selected_fib = re; | |
1047 | ||
1048 | zsend_route_notify_owner(re, p, ZAPI_ROUTE_FAIL_INSTALL); | |
1049 | zlog_warn("%u:%s: Route install failed", re->vrf_id, | |
1050 | prefix2str(p, buf, sizeof(buf))); | |
1051 | break; | |
1052 | case DP_DELETE_SUCCESS: | |
1053 | /* | |
1054 | * The case where selected_fib is not re is | |
1055 | * when we have received a system route | |
1056 | * that is overriding our installed route | |
1057 | * as such we should leave the selected_fib | |
1058 | * pointer alone | |
1059 | */ | |
1060 | if (dest->selected_fib == re) | |
1061 | dest->selected_fib = NULL; | |
1062 | for (ALL_NEXTHOPS(re->ng, nexthop)) | |
1063 | UNSET_FLAG(nexthop->flags, NEXTHOP_FLAG_FIB); | |
1064 | ||
1065 | zsend_route_notify_owner(re, p, ZAPI_ROUTE_REMOVED); | |
1066 | break; | |
1067 | case DP_DELETE_FAILURE: | |
1068 | /* | |
1069 | * Should we set this to NULL if the | |
1070 | * delete fails? | |
1071 | */ | |
1072 | dest->selected_fib = NULL; | |
1073 | zlog_warn("%u:%s: Route Deletion failure", re->vrf_id, | |
1074 | prefix2str(p, buf, sizeof(buf))); | |
1075 | ||
1076 | zsend_route_notify_owner(re, p, ZAPI_ROUTE_REMOVE_FAIL); | |
1077 | break; | |
1078 | } | |
1079 | } | |
1080 | ||
1081 | /* Update flag indicates whether this is a "replace" or not. Currently, this | |
1082 | * is only used for IPv4. | |
1083 | */ | |
1084 | void rib_install_kernel(struct route_node *rn, struct route_entry *re, | |
1085 | struct route_entry *old) | |
1086 | { | |
1087 | struct nexthop *nexthop; | |
1088 | rib_table_info_t *info = srcdest_rnode_table_info(rn); | |
1089 | const struct prefix *p, *src_p; | |
1090 | struct zebra_vrf *zvrf = vrf_info_lookup(re->vrf_id); | |
1091 | ||
1092 | srcdest_rnode_prefixes(rn, &p, &src_p); | |
1093 | ||
1094 | if (info->safi != SAFI_UNICAST) { | |
1095 | for (ALL_NEXTHOPS(re->ng, nexthop)) | |
1096 | SET_FLAG(nexthop->flags, NEXTHOP_FLAG_FIB); | |
1097 | return; | |
1098 | } else { | |
1099 | struct nexthop *prev; | |
1100 | ||
1101 | for (ALL_NEXTHOPS(re->ng, nexthop)) { | |
1102 | UNSET_FLAG (nexthop->flags, NEXTHOP_FLAG_DUPLICATE); | |
1103 | for (ALL_NEXTHOPS(re->ng, prev)) { | |
1104 | if (prev == nexthop) | |
1105 | break; | |
1106 | if (nexthop_same_firsthop(nexthop, prev)) { | |
1107 | SET_FLAG(nexthop->flags, | |
1108 | NEXTHOP_FLAG_DUPLICATE); | |
1109 | break; | |
1110 | } | |
1111 | } | |
1112 | } | |
1113 | } | |
1114 | ||
1115 | /* | |
1116 | * If this is a replace to a new RE let the originator of the RE | |
1117 | * know that they've lost | |
1118 | */ | |
1119 | if (old && (old != re) && (old->type != re->type)) | |
1120 | zsend_route_notify_owner(old, p, ZAPI_ROUTE_BETTER_ADMIN_WON); | |
1121 | ||
1122 | /* | |
1123 | * Make sure we update the FPM any time we send new information to | |
1124 | * the kernel. | |
1125 | */ | |
1126 | hook_call(rib_update, rn, "installing in kernel"); | |
1127 | switch (kernel_route_rib(rn, p, src_p, old, re)) { | |
1128 | case DP_REQUEST_QUEUED: | |
1129 | flog_err( | |
1130 | ZEBRA_ERR_DP_INVALID_RC, | |
1131 | "No current known DataPlane interfaces can return this, please fix"); | |
1132 | break; | |
1133 | case DP_REQUEST_FAILURE: | |
1134 | flog_err( | |
1135 | ZEBRA_ERR_DP_INSTALL_FAIL, | |
1136 | "No current known Rib Install Failure cases, please fix"); | |
1137 | break; | |
1138 | case DP_REQUEST_SUCCESS: | |
1139 | zvrf->installs++; | |
1140 | break; | |
1141 | } | |
1142 | ||
1143 | return; | |
1144 | } | |
1145 | ||
1146 | /* Uninstall the route from kernel. */ | |
1147 | void rib_uninstall_kernel(struct route_node *rn, struct route_entry *re) | |
1148 | { | |
1149 | struct nexthop *nexthop; | |
1150 | rib_table_info_t *info = srcdest_rnode_table_info(rn); | |
1151 | const struct prefix *p, *src_p; | |
1152 | struct zebra_vrf *zvrf = vrf_info_lookup(re->vrf_id); | |
1153 | ||
1154 | srcdest_rnode_prefixes(rn, &p, &src_p); | |
1155 | ||
1156 | if (info->safi != SAFI_UNICAST) { | |
1157 | for (ALL_NEXTHOPS(re->ng, nexthop)) | |
1158 | UNSET_FLAG(nexthop->flags, NEXTHOP_FLAG_FIB); | |
1159 | return; | |
1160 | } | |
1161 | ||
1162 | /* | |
1163 | * Make sure we update the FPM any time we send new information to | |
1164 | * the kernel. | |
1165 | */ | |
1166 | hook_call(rib_update, rn, "uninstalling from kernel"); | |
1167 | switch (kernel_route_rib(rn, p, src_p, re, NULL)) { | |
1168 | case DP_REQUEST_QUEUED: | |
1169 | flog_err( | |
1170 | ZEBRA_ERR_DP_INVALID_RC, | |
1171 | "No current known DataPlane interfaces can return this, please fix"); | |
1172 | break; | |
1173 | case DP_REQUEST_FAILURE: | |
1174 | flog_err( | |
1175 | ZEBRA_ERR_DP_INSTALL_FAIL, | |
1176 | "No current known RIB Install Failure cases, please fix"); | |
1177 | break; | |
1178 | case DP_REQUEST_SUCCESS: | |
1179 | if (zvrf) | |
1180 | zvrf->removals++; | |
1181 | break; | |
1182 | } | |
1183 | ||
1184 | return; | |
1185 | } | |
1186 | ||
1187 | /* Uninstall the route from kernel. */ | |
1188 | static void rib_uninstall(struct route_node *rn, struct route_entry *re) | |
1189 | { | |
1190 | rib_table_info_t *info = srcdest_rnode_table_info(rn); | |
1191 | rib_dest_t *dest = rib_dest_from_rnode(rn); | |
1192 | ||
1193 | if (dest && dest->selected_fib == re) { | |
1194 | if (info->safi == SAFI_UNICAST) | |
1195 | hook_call(rib_update, rn, "rib_uninstall"); | |
1196 | ||
1197 | if (!RIB_SYSTEM_ROUTE(re)) | |
1198 | rib_uninstall_kernel(rn, re); | |
1199 | ||
1200 | /* If labeled-unicast route, uninstall transit LSP. */ | |
1201 | if (zebra_rib_labeled_unicast(re)) | |
1202 | zebra_mpls_lsp_uninstall(info->zvrf, rn, re); | |
1203 | } | |
1204 | ||
1205 | if (CHECK_FLAG(re->flags, ZEBRA_FLAG_SELECTED)) { | |
1206 | const struct prefix *p, *src_p; | |
1207 | ||
1208 | srcdest_rnode_prefixes(rn, &p, &src_p); | |
1209 | ||
1210 | redistribute_delete(p, src_p, re); | |
1211 | UNSET_FLAG(re->flags, ZEBRA_FLAG_SELECTED); | |
1212 | } | |
1213 | } | |
1214 | ||
1215 | /* | |
1216 | * rib_can_delete_dest | |
1217 | * | |
1218 | * Returns TRUE if the given dest can be deleted from the table. | |
1219 | */ | |
1220 | static int rib_can_delete_dest(rib_dest_t *dest) | |
1221 | { | |
1222 | if (dest->routes) { | |
1223 | return 0; | |
1224 | } | |
1225 | ||
1226 | /* | |
1227 | * Don't delete the dest if we have to update the FPM about this | |
1228 | * prefix. | |
1229 | */ | |
1230 | if (CHECK_FLAG(dest->flags, RIB_DEST_UPDATE_FPM) | |
1231 | || CHECK_FLAG(dest->flags, RIB_DEST_SENT_TO_FPM)) | |
1232 | return 0; | |
1233 | ||
1234 | return 1; | |
1235 | } | |
1236 | ||
1237 | /* | |
1238 | * rib_gc_dest | |
1239 | * | |
1240 | * Garbage collect the rib dest corresponding to the given route node | |
1241 | * if appropriate. | |
1242 | * | |
1243 | * Returns TRUE if the dest was deleted, FALSE otherwise. | |
1244 | */ | |
1245 | int rib_gc_dest(struct route_node *rn) | |
1246 | { | |
1247 | rib_dest_t *dest; | |
1248 | ||
1249 | dest = rib_dest_from_rnode(rn); | |
1250 | if (!dest) | |
1251 | return 0; | |
1252 | ||
1253 | if (!rib_can_delete_dest(dest)) | |
1254 | return 0; | |
1255 | ||
1256 | if (IS_ZEBRA_DEBUG_RIB) { | |
1257 | struct zebra_vrf *zvrf; | |
1258 | ||
1259 | zvrf = rib_dest_vrf(dest); | |
1260 | rnode_debug(rn, zvrf_id(zvrf), "removing dest from table"); | |
1261 | } | |
1262 | ||
1263 | dest->rnode = NULL; | |
1264 | XFREE(MTYPE_RIB_DEST, dest); | |
1265 | rn->info = NULL; | |
1266 | ||
1267 | /* | |
1268 | * Release the one reference that we keep on the route node. | |
1269 | */ | |
1270 | route_unlock_node(rn); | |
1271 | return 1; | |
1272 | } | |
1273 | ||
1274 | static void rib_process_add_fib(struct zebra_vrf *zvrf, struct route_node *rn, | |
1275 | struct route_entry *new) | |
1276 | { | |
1277 | rib_dest_t *dest = rib_dest_from_rnode(rn); | |
1278 | ||
1279 | hook_call(rib_update, rn, "new route selected"); | |
1280 | ||
1281 | /* Update real nexthop. This may actually determine if nexthop is active | |
1282 | * or not. */ | |
1283 | if (!nexthop_active_update(rn, new, 1)) { | |
1284 | UNSET_FLAG(new->status, ROUTE_ENTRY_CHANGED); | |
1285 | return; | |
1286 | } | |
1287 | ||
1288 | if (IS_ZEBRA_DEBUG_RIB) { | |
1289 | char buf[SRCDEST2STR_BUFFER]; | |
1290 | srcdest_rnode2str(rn, buf, sizeof(buf)); | |
1291 | zlog_debug("%u:%s: Adding route rn %p, re %p (type %d)", | |
1292 | zvrf_id(zvrf), buf, rn, new, new->type); | |
1293 | } | |
1294 | ||
1295 | /* If labeled-unicast route, install transit LSP. */ | |
1296 | if (zebra_rib_labeled_unicast(new)) | |
1297 | zebra_mpls_lsp_install(zvrf, rn, new); | |
1298 | ||
1299 | if (!RIB_SYSTEM_ROUTE(new)) | |
1300 | rib_install_kernel(rn, new, NULL); | |
1301 | else | |
1302 | dest->selected_fib = new; | |
1303 | ||
1304 | UNSET_FLAG(new->status, ROUTE_ENTRY_CHANGED); | |
1305 | } | |
1306 | ||
1307 | static void rib_process_del_fib(struct zebra_vrf *zvrf, struct route_node *rn, | |
1308 | struct route_entry *old) | |
1309 | { | |
1310 | rib_dest_t *dest = rib_dest_from_rnode(rn); | |
1311 | hook_call(rib_update, rn, "removing existing route"); | |
1312 | ||
1313 | /* Uninstall from kernel. */ | |
1314 | if (IS_ZEBRA_DEBUG_RIB) { | |
1315 | char buf[SRCDEST2STR_BUFFER]; | |
1316 | srcdest_rnode2str(rn, buf, sizeof(buf)); | |
1317 | zlog_debug("%u:%s: Deleting route rn %p, re %p (type %d)", | |
1318 | zvrf_id(zvrf), buf, rn, old, old->type); | |
1319 | } | |
1320 | ||
1321 | /* If labeled-unicast route, uninstall transit LSP. */ | |
1322 | if (zebra_rib_labeled_unicast(old)) | |
1323 | zebra_mpls_lsp_uninstall(zvrf, rn, old); | |
1324 | ||
1325 | if (!RIB_SYSTEM_ROUTE(old)) | |
1326 | rib_uninstall_kernel(rn, old); | |
1327 | else { | |
1328 | /* | |
1329 | * We are setting this to NULL here | |
1330 | * because that is what we traditionally | |
1331 | * have been doing. I am not positive | |
1332 | * that this is the right thing to do | |
1333 | * but let's leave the code alone | |
1334 | * for the RIB_SYSTEM_ROUTE case | |
1335 | */ | |
1336 | dest->selected_fib = NULL; | |
1337 | } | |
1338 | ||
1339 | /* Update nexthop for route, reset changed flag. */ | |
1340 | nexthop_active_update(rn, old, 1); | |
1341 | UNSET_FLAG(old->status, ROUTE_ENTRY_CHANGED); | |
1342 | } | |
1343 | ||
1344 | static void rib_process_update_fib(struct zebra_vrf *zvrf, | |
1345 | struct route_node *rn, | |
1346 | struct route_entry *old, | |
1347 | struct route_entry *new) | |
1348 | { | |
1349 | struct nexthop *nexthop = NULL; | |
1350 | int nh_active = 0; | |
1351 | rib_dest_t *dest = rib_dest_from_rnode(rn); | |
1352 | ||
1353 | /* | |
1354 | * We have to install or update if a new route has been selected or | |
1355 | * something has changed. | |
1356 | */ | |
1357 | if (new != old || CHECK_FLAG(new->status, ROUTE_ENTRY_CHANGED)) { | |
1358 | hook_call(rib_update, rn, "updating existing route"); | |
1359 | ||
1360 | /* Update the nexthop; we could determine here that nexthop is | |
1361 | * inactive. */ | |
1362 | if (nexthop_active_update(rn, new, 1)) | |
1363 | nh_active = 1; | |
1364 | ||
1365 | /* If nexthop is active, install the selected route, if | |
1366 | * appropriate. If | |
1367 | * the install succeeds, cleanup flags for prior route, if | |
1368 | * different from | |
1369 | * newly selected. | |
1370 | */ | |
1371 | if (nh_active) { | |
1372 | if (IS_ZEBRA_DEBUG_RIB) { | |
1373 | char buf[SRCDEST2STR_BUFFER]; | |
1374 | srcdest_rnode2str(rn, buf, sizeof(buf)); | |
1375 | if (new != old) | |
1376 | zlog_debug( | |
1377 | "%u:%s: Updating route rn %p, re %p (type %d) " | |
1378 | "old %p (type %d)", | |
1379 | zvrf_id(zvrf), buf, rn, new, | |
1380 | new->type, old, old->type); | |
1381 | else | |
1382 | zlog_debug( | |
1383 | "%u:%s: Updating route rn %p, re %p (type %d)", | |
1384 | zvrf_id(zvrf), buf, rn, new, | |
1385 | new->type); | |
1386 | } | |
1387 | ||
1388 | /* If labeled-unicast route, uninstall transit LSP. */ | |
1389 | if (zebra_rib_labeled_unicast(old)) | |
1390 | zebra_mpls_lsp_uninstall(zvrf, rn, old); | |
1391 | ||
1392 | /* Non-system route should be installed. */ | |
1393 | if (!RIB_SYSTEM_ROUTE(new)) { | |
1394 | /* If labeled-unicast route, install transit | |
1395 | * LSP. */ | |
1396 | if (zebra_rib_labeled_unicast(new)) | |
1397 | zebra_mpls_lsp_install(zvrf, rn, new); | |
1398 | ||
1399 | rib_install_kernel(rn, new, old); | |
1400 | } else { | |
1401 | /* | |
1402 | * We do not need to install the | |
1403 | * selected route because it | |
1404 | * is already isntalled by | |
1405 | * the system( ie not us ) | |
1406 | * so just mark it as winning | |
1407 | * we do need to ensure that | |
1408 | * if we uninstall a route | |
1409 | * from ourselves we don't | |
1410 | * over write this pointer | |
1411 | */ | |
1412 | dest->selected_fib = NULL; | |
1413 | } | |
1414 | /* If install succeeded or system route, cleanup flags | |
1415 | * for prior route. */ | |
1416 | if (new != old) { | |
1417 | if (RIB_SYSTEM_ROUTE(new)) { | |
1418 | if (!RIB_SYSTEM_ROUTE(old)) | |
1419 | rib_uninstall_kernel(rn, old); | |
1420 | } else { | |
1421 | for (nexthop = old->ng.nexthop; nexthop; | |
1422 | nexthop = nexthop->next) | |
1423 | UNSET_FLAG(nexthop->flags, | |
1424 | NEXTHOP_FLAG_FIB); | |
1425 | } | |
1426 | } | |
1427 | } | |
1428 | ||
1429 | /* | |
1430 | * If nexthop for selected route is not active or install | |
1431 | * failed, we | |
1432 | * may need to uninstall and delete for redistribution. | |
1433 | */ | |
1434 | if (!nh_active) { | |
1435 | if (IS_ZEBRA_DEBUG_RIB) { | |
1436 | char buf[SRCDEST2STR_BUFFER]; | |
1437 | srcdest_rnode2str(rn, buf, sizeof(buf)); | |
1438 | if (new != old) | |
1439 | zlog_debug( | |
1440 | "%u:%s: Deleting route rn %p, re %p (type %d) " | |
1441 | "old %p (type %d) - nexthop inactive", | |
1442 | zvrf_id(zvrf), buf, rn, new, | |
1443 | new->type, old, old->type); | |
1444 | else | |
1445 | zlog_debug( | |
1446 | "%u:%s: Deleting route rn %p, re %p (type %d) - nexthop inactive", | |
1447 | zvrf_id(zvrf), buf, rn, new, | |
1448 | new->type); | |
1449 | } | |
1450 | ||
1451 | /* If labeled-unicast route, uninstall transit LSP. */ | |
1452 | if (zebra_rib_labeled_unicast(old)) | |
1453 | zebra_mpls_lsp_uninstall(zvrf, rn, old); | |
1454 | ||
1455 | if (!RIB_SYSTEM_ROUTE(old)) | |
1456 | rib_uninstall_kernel(rn, old); | |
1457 | else | |
1458 | dest->selected_fib = NULL; | |
1459 | } | |
1460 | } else { | |
1461 | /* | |
1462 | * Same route selected; check if in the FIB and if not, | |
1463 | * re-install. This | |
1464 | * is housekeeping code to deal with race conditions in kernel | |
1465 | * with linux | |
1466 | * netlink reporting interface up before IPv4 or IPv6 protocol | |
1467 | * is ready | |
1468 | * to add routes. | |
1469 | */ | |
1470 | if (!RIB_SYSTEM_ROUTE(new)) { | |
1471 | bool in_fib = false; | |
1472 | ||
1473 | for (ALL_NEXTHOPS(new->ng, nexthop)) | |
1474 | if (CHECK_FLAG(nexthop->flags, | |
1475 | NEXTHOP_FLAG_FIB)) { | |
1476 | in_fib = true; | |
1477 | break; | |
1478 | } | |
1479 | if (!in_fib) | |
1480 | rib_install_kernel(rn, new, NULL); | |
1481 | } | |
1482 | } | |
1483 | ||
1484 | /* Update prior route. */ | |
1485 | if (new != old) { | |
1486 | /* Set real nexthop. */ | |
1487 | nexthop_active_update(rn, old, 1); | |
1488 | UNSET_FLAG(old->status, ROUTE_ENTRY_CHANGED); | |
1489 | } | |
1490 | ||
1491 | /* Clear changed flag. */ | |
1492 | UNSET_FLAG(new->status, ROUTE_ENTRY_CHANGED); | |
1493 | } | |
1494 | ||
1495 | /* Check if 'alternate' RIB entry is better than 'current'. */ | |
1496 | static struct route_entry *rib_choose_best(struct route_entry *current, | |
1497 | struct route_entry *alternate) | |
1498 | { | |
1499 | if (current == NULL) | |
1500 | return alternate; | |
1501 | ||
1502 | /* filter route selection in following order: | |
1503 | * - connected beats other types | |
1504 | * - if both connected, loopback or vrf wins | |
1505 | * - lower distance beats higher | |
1506 | * - lower metric beats higher for equal distance | |
1507 | * - last, hence oldest, route wins tie break. | |
1508 | */ | |
1509 | ||
1510 | /* Connected routes. Check to see if either are a vrf | |
1511 | * or loopback interface. If not, pick the last connected | |
1512 | * route of the set of lowest metric connected routes. | |
1513 | */ | |
1514 | if (alternate->type == ZEBRA_ROUTE_CONNECT) { | |
1515 | if (current->type != ZEBRA_ROUTE_CONNECT) | |
1516 | return alternate; | |
1517 | ||
1518 | /* both are connected. are either loop or vrf? */ | |
1519 | struct nexthop *nexthop = NULL; | |
1520 | ||
1521 | for (ALL_NEXTHOPS(alternate->ng, nexthop)) { | |
1522 | if (if_is_loopback_or_vrf(if_lookup_by_index( | |
1523 | nexthop->ifindex, alternate->vrf_id))) | |
1524 | return alternate; | |
1525 | } | |
1526 | ||
1527 | for (ALL_NEXTHOPS(current->ng, nexthop)) { | |
1528 | if (if_is_loopback_or_vrf(if_lookup_by_index( | |
1529 | nexthop->ifindex, current->vrf_id))) | |
1530 | return current; | |
1531 | } | |
1532 | ||
1533 | /* Neither are loop or vrf so pick best metric */ | |
1534 | if (alternate->metric <= current->metric) | |
1535 | return alternate; | |
1536 | ||
1537 | return current; | |
1538 | } | |
1539 | ||
1540 | if (current->type == ZEBRA_ROUTE_CONNECT) | |
1541 | return current; | |
1542 | ||
1543 | /* higher distance loses */ | |
1544 | if (alternate->distance < current->distance) | |
1545 | return alternate; | |
1546 | if (current->distance < alternate->distance) | |
1547 | return current; | |
1548 | ||
1549 | /* metric tie-breaks equal distance */ | |
1550 | if (alternate->metric <= current->metric) | |
1551 | return alternate; | |
1552 | ||
1553 | return current; | |
1554 | } | |
1555 | ||
1556 | /* Core function for processing routing information base. */ | |
1557 | static void rib_process(struct route_node *rn) | |
1558 | { | |
1559 | struct route_entry *re; | |
1560 | struct route_entry *next; | |
1561 | struct route_entry *old_selected = NULL; | |
1562 | struct route_entry *new_selected = NULL; | |
1563 | struct route_entry *old_fib = NULL; | |
1564 | struct route_entry *new_fib = NULL; | |
1565 | struct route_entry *best = NULL; | |
1566 | char buf[SRCDEST2STR_BUFFER]; | |
1567 | rib_dest_t *dest; | |
1568 | struct zebra_vrf *zvrf = NULL; | |
1569 | const struct prefix *p, *src_p; | |
1570 | ||
1571 | srcdest_rnode_prefixes(rn, &p, &src_p); | |
1572 | vrf_id_t vrf_id = VRF_UNKNOWN; | |
1573 | ||
1574 | assert(rn); | |
1575 | ||
1576 | dest = rib_dest_from_rnode(rn); | |
1577 | if (dest) { | |
1578 | zvrf = rib_dest_vrf(dest); | |
1579 | vrf_id = zvrf_id(zvrf); | |
1580 | } | |
1581 | ||
1582 | if (IS_ZEBRA_DEBUG_RIB) | |
1583 | srcdest_rnode2str(rn, buf, sizeof(buf)); | |
1584 | ||
1585 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) | |
1586 | zlog_debug("%u:%s: Processing rn %p", vrf_id, buf, rn); | |
1587 | ||
1588 | /* | |
1589 | * we can have rn's that have a NULL info pointer | |
1590 | * (dest). As such let's not let the deref happen | |
1591 | * additionally we know RNODE_FOREACH_RE_SAFE | |
1592 | * will not iterate so we are ok. | |
1593 | */ | |
1594 | if (dest) | |
1595 | old_fib = dest->selected_fib; | |
1596 | ||
1597 | RNODE_FOREACH_RE_SAFE (rn, re, next) { | |
1598 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) | |
1599 | zlog_debug( | |
1600 | "%u:%s: Examine re %p (type %d) status %x flags %x " | |
1601 | "dist %d metric %d", | |
1602 | vrf_id, buf, re, re->type, re->status, | |
1603 | re->flags, re->distance, re->metric); | |
1604 | ||
1605 | UNSET_FLAG(re->status, ROUTE_ENTRY_NEXTHOPS_CHANGED); | |
1606 | ||
1607 | /* Currently selected re. */ | |
1608 | if (CHECK_FLAG(re->flags, ZEBRA_FLAG_SELECTED)) { | |
1609 | assert(old_selected == NULL); | |
1610 | old_selected = re; | |
1611 | } | |
1612 | ||
1613 | /* Skip deleted entries from selection */ | |
1614 | if (CHECK_FLAG(re->status, ROUTE_ENTRY_REMOVED)) | |
1615 | continue; | |
1616 | ||
1617 | /* Skip unreachable nexthop. */ | |
1618 | /* This first call to nexthop_active_update is merely to | |
1619 | * determine if | |
1620 | * there's any change to nexthops associated with this RIB | |
1621 | * entry. Now, | |
1622 | * rib_process() can be invoked due to an external event such as | |
1623 | * link | |
1624 | * down or due to next-hop-tracking evaluation. In the latter | |
1625 | * case, | |
1626 | * a decision has already been made that the NHs have changed. | |
1627 | * So, no | |
1628 | * need to invoke a potentially expensive call again. Further, | |
1629 | * since | |
1630 | * the change might be in a recursive NH which is not caught in | |
1631 | * the nexthop_active_update() code. Thus, we might miss changes | |
1632 | * to | |
1633 | * recursive NHs. | |
1634 | */ | |
1635 | if (!CHECK_FLAG(re->status, ROUTE_ENTRY_CHANGED) | |
1636 | && !nexthop_active_update(rn, re, 0)) { | |
1637 | if (re->type == ZEBRA_ROUTE_TABLE) { | |
1638 | /* XXX: HERE BE DRAGONS!!!!! | |
1639 | * In all honesty, I have not yet figured out | |
1640 | * what this part | |
1641 | * does or why the ROUTE_ENTRY_CHANGED test | |
1642 | * above is correct | |
1643 | * or why we need to delete a route here, and | |
1644 | * also not whether | |
1645 | * this concerns both selected and fib route, or | |
1646 | * only selected | |
1647 | * or only fib */ | |
1648 | /* This entry was denied by the 'ip protocol | |
1649 | * table' route-map, we | |
1650 | * need to delete it */ | |
1651 | if (re != old_selected) { | |
1652 | if (IS_ZEBRA_DEBUG_RIB) | |
1653 | zlog_debug( | |
1654 | "%s: %u:%s: imported via import-table but denied " | |
1655 | "by the ip protocol table route-map", | |
1656 | __func__, vrf_id, buf); | |
1657 | rib_unlink(rn, re); | |
1658 | } else | |
1659 | SET_FLAG(re->status, | |
1660 | ROUTE_ENTRY_REMOVED); | |
1661 | } | |
1662 | ||
1663 | continue; | |
1664 | } | |
1665 | ||
1666 | /* Infinite distance. */ | |
1667 | if (re->distance == DISTANCE_INFINITY) { | |
1668 | UNSET_FLAG(re->status, ROUTE_ENTRY_CHANGED); | |
1669 | continue; | |
1670 | } | |
1671 | ||
1672 | if (CHECK_FLAG(re->flags, ZEBRA_FLAG_FIB_OVERRIDE)) { | |
1673 | best = rib_choose_best(new_fib, re); | |
1674 | if (new_fib && best != new_fib) | |
1675 | UNSET_FLAG(new_fib->status, | |
1676 | ROUTE_ENTRY_CHANGED); | |
1677 | new_fib = best; | |
1678 | } else { | |
1679 | best = rib_choose_best(new_selected, re); | |
1680 | if (new_selected && best != new_selected) | |
1681 | UNSET_FLAG(new_selected->status, | |
1682 | ROUTE_ENTRY_CHANGED); | |
1683 | new_selected = best; | |
1684 | } | |
1685 | if (best != re) | |
1686 | UNSET_FLAG(re->status, ROUTE_ENTRY_CHANGED); | |
1687 | } /* RNODE_FOREACH_RE */ | |
1688 | ||
1689 | /* If no FIB override route, use the selected route also for FIB */ | |
1690 | if (new_fib == NULL) | |
1691 | new_fib = new_selected; | |
1692 | ||
1693 | /* After the cycle is finished, the following pointers will be set: | |
1694 | * old_selected --- RE entry currently having SELECTED | |
1695 | * new_selected --- RE entry that is newly SELECTED | |
1696 | * old_fib --- RE entry currently in kernel FIB | |
1697 | * new_fib --- RE entry that is newly to be in kernel FIB | |
1698 | * | |
1699 | * new_selected will get SELECTED flag, and is going to be redistributed | |
1700 | * the zclients. new_fib (which can be new_selected) will be installed | |
1701 | * in kernel. | |
1702 | */ | |
1703 | ||
1704 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) { | |
1705 | zlog_debug( | |
1706 | "%u:%s: After processing: old_selected %p new_selected %p old_fib %p new_fib %p", | |
1707 | vrf_id, buf, (void *)old_selected, (void *)new_selected, | |
1708 | (void *)old_fib, (void *)new_fib); | |
1709 | } | |
1710 | ||
1711 | /* Buffer ROUTE_ENTRY_CHANGED here, because it will get cleared if | |
1712 | * fib == selected */ | |
1713 | bool selected_changed = new_selected && CHECK_FLAG(new_selected->status, | |
1714 | ROUTE_ENTRY_CHANGED); | |
1715 | ||
1716 | /* Update fib according to selection results */ | |
1717 | if (new_fib && old_fib) | |
1718 | rib_process_update_fib(zvrf, rn, old_fib, new_fib); | |
1719 | else if (new_fib) | |
1720 | rib_process_add_fib(zvrf, rn, new_fib); | |
1721 | else if (old_fib) | |
1722 | rib_process_del_fib(zvrf, rn, old_fib); | |
1723 | ||
1724 | /* Redistribute SELECTED entry */ | |
1725 | if (old_selected != new_selected || selected_changed) { | |
1726 | struct nexthop *nexthop = NULL; | |
1727 | ||
1728 | /* Check if we have a FIB route for the destination, otherwise, | |
1729 | * don't redistribute it */ | |
1730 | if (new_fib) { | |
1731 | for (ALL_NEXTHOPS(new_fib->ng, nexthop)) { | |
1732 | if (CHECK_FLAG(nexthop->flags, | |
1733 | NEXTHOP_FLAG_FIB)) { | |
1734 | break; | |
1735 | } | |
1736 | } | |
1737 | } | |
1738 | if (!nexthop) | |
1739 | new_selected = NULL; | |
1740 | ||
1741 | if (new_selected && new_selected != new_fib) { | |
1742 | nexthop_active_update(rn, new_selected, 1); | |
1743 | UNSET_FLAG(new_selected->status, ROUTE_ENTRY_CHANGED); | |
1744 | } | |
1745 | ||
1746 | if (old_selected) { | |
1747 | if (!new_selected) | |
1748 | redistribute_delete(p, src_p, old_selected); | |
1749 | if (old_selected != new_selected) | |
1750 | UNSET_FLAG(old_selected->flags, | |
1751 | ZEBRA_FLAG_SELECTED); | |
1752 | } | |
1753 | ||
1754 | if (new_selected) { | |
1755 | /* Install new or replace existing redistributed entry | |
1756 | */ | |
1757 | SET_FLAG(new_selected->flags, ZEBRA_FLAG_SELECTED); | |
1758 | redistribute_update(p, src_p, new_selected, | |
1759 | old_selected); | |
1760 | } | |
1761 | } | |
1762 | ||
1763 | /* Remove all RE entries queued for removal */ | |
1764 | RNODE_FOREACH_RE_SAFE (rn, re, next) { | |
1765 | if (CHECK_FLAG(re->status, ROUTE_ENTRY_REMOVED)) { | |
1766 | if (IS_ZEBRA_DEBUG_RIB) { | |
1767 | rnode_debug(rn, vrf_id, "rn %p, removing re %p", | |
1768 | (void *)rn, (void *)re); | |
1769 | } | |
1770 | rib_unlink(rn, re); | |
1771 | } | |
1772 | } | |
1773 | ||
1774 | /* | |
1775 | * Check if the dest can be deleted now. | |
1776 | */ | |
1777 | rib_gc_dest(rn); | |
1778 | } | |
1779 | ||
1780 | /* Take a list of route_node structs and return 1, if there was a record | |
1781 | * picked from it and processed by rib_process(). Don't process more, | |
1782 | * than one RN record; operate only in the specified sub-queue. | |
1783 | */ | |
1784 | static unsigned int process_subq(struct list *subq, uint8_t qindex) | |
1785 | { | |
1786 | struct listnode *lnode = listhead(subq); | |
1787 | struct route_node *rnode; | |
1788 | rib_dest_t *dest; | |
1789 | struct zebra_vrf *zvrf = NULL; | |
1790 | ||
1791 | if (!lnode) | |
1792 | return 0; | |
1793 | ||
1794 | rnode = listgetdata(lnode); | |
1795 | dest = rib_dest_from_rnode(rnode); | |
1796 | if (dest) | |
1797 | zvrf = rib_dest_vrf(dest); | |
1798 | ||
1799 | rib_process(rnode); | |
1800 | ||
1801 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) { | |
1802 | char buf[SRCDEST2STR_BUFFER]; | |
1803 | srcdest_rnode2str(rnode, buf, sizeof(buf)); | |
1804 | zlog_debug("%u:%s: rn %p dequeued from sub-queue %u", | |
1805 | zvrf ? zvrf_id(zvrf) : 0, buf, rnode, qindex); | |
1806 | } | |
1807 | ||
1808 | if (rnode->info) | |
1809 | UNSET_FLAG(rib_dest_from_rnode(rnode)->flags, | |
1810 | RIB_ROUTE_QUEUED(qindex)); | |
1811 | ||
1812 | #if 0 | |
1813 | else | |
1814 | { | |
1815 | zlog_debug ("%s: called for route_node (%p, %d) with no ribs", | |
1816 | __func__, rnode, rnode->lock); | |
1817 | zlog_backtrace(LOG_DEBUG); | |
1818 | } | |
1819 | #endif | |
1820 | route_unlock_node(rnode); | |
1821 | list_delete_node(subq, lnode); | |
1822 | return 1; | |
1823 | } | |
1824 | ||
1825 | /* | |
1826 | * All meta queues have been processed. Trigger next-hop evaluation. | |
1827 | */ | |
1828 | static void meta_queue_process_complete(struct work_queue *dummy) | |
1829 | { | |
1830 | struct vrf *vrf; | |
1831 | struct zebra_vrf *zvrf; | |
1832 | ||
1833 | /* Evaluate nexthops for those VRFs which underwent route processing. | |
1834 | * This | |
1835 | * should limit the evaluation to the necessary VRFs in most common | |
1836 | * situations. | |
1837 | */ | |
1838 | RB_FOREACH (vrf, vrf_id_head, &vrfs_by_id) { | |
1839 | zvrf = vrf->info; | |
1840 | if (zvrf == NULL || !(zvrf->flags & ZEBRA_VRF_RIB_SCHEDULED)) | |
1841 | continue; | |
1842 | ||
1843 | zvrf->flags &= ~ZEBRA_VRF_RIB_SCHEDULED; | |
1844 | zebra_evaluate_rnh(zvrf_id(zvrf), AF_INET, 0, RNH_NEXTHOP_TYPE, | |
1845 | NULL); | |
1846 | zebra_evaluate_rnh(zvrf_id(zvrf), AF_INET, 0, | |
1847 | RNH_IMPORT_CHECK_TYPE, NULL); | |
1848 | zebra_evaluate_rnh(zvrf_id(zvrf), AF_INET6, 0, RNH_NEXTHOP_TYPE, | |
1849 | NULL); | |
1850 | zebra_evaluate_rnh(zvrf_id(zvrf), AF_INET6, 0, | |
1851 | RNH_IMPORT_CHECK_TYPE, NULL); | |
1852 | } | |
1853 | ||
1854 | /* Schedule LSPs for processing, if needed. */ | |
1855 | zvrf = vrf_info_lookup(VRF_DEFAULT); | |
1856 | if (mpls_should_lsps_be_processed(zvrf)) { | |
1857 | if (IS_ZEBRA_DEBUG_MPLS) | |
1858 | zlog_debug( | |
1859 | "%u: Scheduling all LSPs upon RIB completion", | |
1860 | zvrf_id(zvrf)); | |
1861 | zebra_mpls_lsp_schedule(zvrf); | |
1862 | mpls_unmark_lsps_for_processing(zvrf); | |
1863 | } | |
1864 | } | |
1865 | ||
1866 | /* Dispatch the meta queue by picking, processing and unlocking the next RN from | |
1867 | * a non-empty sub-queue with lowest priority. wq is equal to zebra->ribq and | |
1868 | * data | |
1869 | * is pointed to the meta queue structure. | |
1870 | */ | |
1871 | static wq_item_status meta_queue_process(struct work_queue *dummy, void *data) | |
1872 | { | |
1873 | struct meta_queue *mq = data; | |
1874 | unsigned i; | |
1875 | ||
1876 | for (i = 0; i < MQ_SIZE; i++) | |
1877 | if (process_subq(mq->subq[i], i)) { | |
1878 | mq->size--; | |
1879 | break; | |
1880 | } | |
1881 | return mq->size ? WQ_REQUEUE : WQ_SUCCESS; | |
1882 | } | |
1883 | ||
1884 | /* | |
1885 | * Map from rib types to queue type (priority) in meta queue | |
1886 | */ | |
1887 | static const uint8_t meta_queue_map[ZEBRA_ROUTE_MAX] = { | |
1888 | [ZEBRA_ROUTE_SYSTEM] = 4, | |
1889 | [ZEBRA_ROUTE_KERNEL] = 0, | |
1890 | [ZEBRA_ROUTE_CONNECT] = 0, | |
1891 | [ZEBRA_ROUTE_STATIC] = 1, | |
1892 | [ZEBRA_ROUTE_RIP] = 2, | |
1893 | [ZEBRA_ROUTE_RIPNG] = 2, | |
1894 | [ZEBRA_ROUTE_OSPF] = 2, | |
1895 | [ZEBRA_ROUTE_OSPF6] = 2, | |
1896 | [ZEBRA_ROUTE_ISIS] = 2, | |
1897 | [ZEBRA_ROUTE_BGP] = 3, | |
1898 | [ZEBRA_ROUTE_PIM] = 4, // Shouldn't happen but for safety | |
1899 | [ZEBRA_ROUTE_EIGRP] = 2, | |
1900 | [ZEBRA_ROUTE_NHRP] = 2, | |
1901 | [ZEBRA_ROUTE_HSLS] = 4, | |
1902 | [ZEBRA_ROUTE_OLSR] = 4, | |
1903 | [ZEBRA_ROUTE_TABLE] = 1, | |
1904 | [ZEBRA_ROUTE_LDP] = 4, | |
1905 | [ZEBRA_ROUTE_VNC] = 3, | |
1906 | [ZEBRA_ROUTE_VNC_DIRECT] = 3, | |
1907 | [ZEBRA_ROUTE_VNC_DIRECT_RH] = 3, | |
1908 | [ZEBRA_ROUTE_BGP_DIRECT] = 3, | |
1909 | [ZEBRA_ROUTE_BGP_DIRECT_EXT] = 3, | |
1910 | [ZEBRA_ROUTE_BABEL] = 2, | |
1911 | [ZEBRA_ROUTE_ALL] = 4, // Shouldn't happen but for safety | |
1912 | }; | |
1913 | ||
1914 | /* Look into the RN and queue it into one or more priority queues, | |
1915 | * increasing the size for each data push done. | |
1916 | */ | |
1917 | static void rib_meta_queue_add(struct meta_queue *mq, struct route_node *rn) | |
1918 | { | |
1919 | struct route_entry *re; | |
1920 | ||
1921 | RNODE_FOREACH_RE (rn, re) { | |
1922 | uint8_t qindex = meta_queue_map[re->type]; | |
1923 | struct zebra_vrf *zvrf; | |
1924 | ||
1925 | /* Invariant: at this point we always have rn->info set. */ | |
1926 | if (CHECK_FLAG(rib_dest_from_rnode(rn)->flags, | |
1927 | RIB_ROUTE_QUEUED(qindex))) { | |
1928 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) | |
1929 | rnode_debug( | |
1930 | rn, re->vrf_id, | |
1931 | "rn %p is already queued in sub-queue %u", | |
1932 | (void *)rn, qindex); | |
1933 | continue; | |
1934 | } | |
1935 | ||
1936 | SET_FLAG(rib_dest_from_rnode(rn)->flags, | |
1937 | RIB_ROUTE_QUEUED(qindex)); | |
1938 | listnode_add(mq->subq[qindex], rn); | |
1939 | route_lock_node(rn); | |
1940 | mq->size++; | |
1941 | ||
1942 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) | |
1943 | rnode_debug(rn, re->vrf_id, | |
1944 | "queued rn %p into sub-queue %u", | |
1945 | (void *)rn, qindex); | |
1946 | ||
1947 | zvrf = zebra_vrf_lookup_by_id(re->vrf_id); | |
1948 | if (zvrf) | |
1949 | zvrf->flags |= ZEBRA_VRF_RIB_SCHEDULED; | |
1950 | } | |
1951 | } | |
1952 | ||
1953 | /* Add route_node to work queue and schedule processing */ | |
1954 | void rib_queue_add(struct route_node *rn) | |
1955 | { | |
1956 | assert(rn); | |
1957 | ||
1958 | /* Pointless to queue a route_node with no RIB entries to add or remove | |
1959 | */ | |
1960 | if (!rnode_to_ribs(rn)) { | |
1961 | zlog_debug("%s: called for route_node (%p, %d) with no ribs", | |
1962 | __func__, (void *)rn, rn->lock); | |
1963 | zlog_backtrace(LOG_DEBUG); | |
1964 | return; | |
1965 | } | |
1966 | ||
1967 | if (zebrad.ribq == NULL) { | |
1968 | flog_err(ZEBRA_ERR_WQ_NONEXISTENT, | |
1969 | "%s: work_queue does not exist!", __func__); | |
1970 | return; | |
1971 | } | |
1972 | ||
1973 | /* | |
1974 | * The RIB queue should normally be either empty or holding the only | |
1975 | * work_queue_item element. In the latter case this element would | |
1976 | * hold a pointer to the meta queue structure, which must be used to | |
1977 | * actually queue the route nodes to process. So create the MQ | |
1978 | * holder, if necessary, then push the work into it in any case. | |
1979 | * This semantics was introduced after 0.99.9 release. | |
1980 | */ | |
1981 | if (work_queue_empty(zebrad.ribq)) | |
1982 | work_queue_add(zebrad.ribq, zebrad.mq); | |
1983 | ||
1984 | rib_meta_queue_add(zebrad.mq, rn); | |
1985 | ||
1986 | return; | |
1987 | } | |
1988 | ||
1989 | /* Create new meta queue. | |
1990 | A destructor function doesn't seem to be necessary here. | |
1991 | */ | |
1992 | static struct meta_queue *meta_queue_new(void) | |
1993 | { | |
1994 | struct meta_queue *new; | |
1995 | unsigned i; | |
1996 | ||
1997 | new = XCALLOC(MTYPE_WORK_QUEUE, sizeof(struct meta_queue)); | |
1998 | ||
1999 | for (i = 0; i < MQ_SIZE; i++) { | |
2000 | new->subq[i] = list_new(); | |
2001 | assert(new->subq[i]); | |
2002 | } | |
2003 | ||
2004 | return new; | |
2005 | } | |
2006 | ||
2007 | void meta_queue_free(struct meta_queue *mq) | |
2008 | { | |
2009 | unsigned i; | |
2010 | ||
2011 | for (i = 0; i < MQ_SIZE; i++) | |
2012 | list_delete_and_null(&mq->subq[i]); | |
2013 | ||
2014 | XFREE(MTYPE_WORK_QUEUE, mq); | |
2015 | } | |
2016 | ||
2017 | /* initialise zebra rib work queue */ | |
2018 | static void rib_queue_init(struct zebra_t *zebra) | |
2019 | { | |
2020 | assert(zebra); | |
2021 | ||
2022 | if (!(zebra->ribq = | |
2023 | work_queue_new(zebra->master, "route_node processing"))) { | |
2024 | flog_err(ZEBRA_ERR_WQ_NONEXISTENT, | |
2025 | "%s: could not initialise work queue!", __func__); | |
2026 | return; | |
2027 | } | |
2028 | ||
2029 | /* fill in the work queue spec */ | |
2030 | zebra->ribq->spec.workfunc = &meta_queue_process; | |
2031 | zebra->ribq->spec.errorfunc = NULL; | |
2032 | zebra->ribq->spec.completion_func = &meta_queue_process_complete; | |
2033 | /* XXX: TODO: These should be runtime configurable via vty */ | |
2034 | zebra->ribq->spec.max_retries = 3; | |
2035 | zebra->ribq->spec.hold = ZEBRA_RIB_PROCESS_HOLD_TIME; | |
2036 | ||
2037 | if (!(zebra->mq = meta_queue_new())) { | |
2038 | flog_err(ZEBRA_ERR_WQ_NONEXISTENT, | |
2039 | "%s: could not initialise meta queue!", __func__); | |
2040 | return; | |
2041 | } | |
2042 | return; | |
2043 | } | |
2044 | ||
2045 | /* RIB updates are processed via a queue of pointers to route_nodes. | |
2046 | * | |
2047 | * The queue length is bounded by the maximal size of the routing table, | |
2048 | * as a route_node will not be requeued, if already queued. | |
2049 | * | |
2050 | * REs are submitted via rib_addnode or rib_delnode which set minimal | |
2051 | * state, or static_install_route (when an existing RE is updated) | |
2052 | * and then submit route_node to queue for best-path selection later. | |
2053 | * Order of add/delete state changes are preserved for any given RE. | |
2054 | * | |
2055 | * Deleted REs are reaped during best-path selection. | |
2056 | * | |
2057 | * rib_addnode | |
2058 | * |-> rib_link or unset ROUTE_ENTRY_REMOVE |->Update kernel with | |
2059 | * |-------->| | best RE, if required | |
2060 | * | | | |
2061 | * static_install->|->rib_addqueue...... -> rib_process | |
2062 | * | | | |
2063 | * |-------->| |-> rib_unlink | |
2064 | * |-> set ROUTE_ENTRY_REMOVE | | |
2065 | * rib_delnode (RE freed) | |
2066 | * | |
2067 | * The 'info' pointer of a route_node points to a rib_dest_t | |
2068 | * ('dest'). Queueing state for a route_node is kept on the dest. The | |
2069 | * dest is created on-demand by rib_link() and is kept around at least | |
2070 | * as long as there are ribs hanging off it (@see rib_gc_dest()). | |
2071 | * | |
2072 | * Refcounting (aka "locking" throughout the GNU Zebra and Quagga code): | |
2073 | * | |
2074 | * - route_nodes: refcounted by: | |
2075 | * - dest attached to route_node: | |
2076 | * - managed by: rib_link/rib_gc_dest | |
2077 | * - route_node processing queue | |
2078 | * - managed by: rib_addqueue, rib_process. | |
2079 | * | |
2080 | */ | |
2081 | ||
2082 | /* Add RE to head of the route node. */ | |
2083 | static void rib_link(struct route_node *rn, struct route_entry *re, int process) | |
2084 | { | |
2085 | struct route_entry *head; | |
2086 | rib_dest_t *dest; | |
2087 | afi_t afi; | |
2088 | const char *rmap_name; | |
2089 | ||
2090 | assert(re && rn); | |
2091 | ||
2092 | dest = rib_dest_from_rnode(rn); | |
2093 | if (!dest) { | |
2094 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) | |
2095 | rnode_debug(rn, re->vrf_id, "rn %p adding dest", rn); | |
2096 | ||
2097 | dest = XCALLOC(MTYPE_RIB_DEST, sizeof(rib_dest_t)); | |
2098 | route_lock_node(rn); /* rn route table reference */ | |
2099 | rn->info = dest; | |
2100 | dest->rnode = rn; | |
2101 | } | |
2102 | ||
2103 | head = dest->routes; | |
2104 | if (head) { | |
2105 | head->prev = re; | |
2106 | } | |
2107 | re->next = head; | |
2108 | dest->routes = re; | |
2109 | ||
2110 | afi = (rn->p.family == AF_INET) | |
2111 | ? AFI_IP | |
2112 | : (rn->p.family == AF_INET6) ? AFI_IP6 : AFI_MAX; | |
2113 | if (is_zebra_import_table_enabled(afi, re->table)) { | |
2114 | rmap_name = zebra_get_import_table_route_map(afi, re->table); | |
2115 | zebra_add_import_table_entry(rn, re, rmap_name); | |
2116 | } else if (process) | |
2117 | rib_queue_add(rn); | |
2118 | } | |
2119 | ||
2120 | static void rib_addnode(struct route_node *rn, | |
2121 | struct route_entry *re, int process) | |
2122 | { | |
2123 | /* RE node has been un-removed before route-node is processed. | |
2124 | * route_node must hence already be on the queue for processing.. | |
2125 | */ | |
2126 | if (CHECK_FLAG(re->status, ROUTE_ENTRY_REMOVED)) { | |
2127 | if (IS_ZEBRA_DEBUG_RIB) | |
2128 | rnode_debug(rn, re->vrf_id, "rn %p, un-removed re %p", | |
2129 | (void *)rn, (void *)re); | |
2130 | ||
2131 | UNSET_FLAG(re->status, ROUTE_ENTRY_REMOVED); | |
2132 | return; | |
2133 | } | |
2134 | rib_link(rn, re, process); | |
2135 | } | |
2136 | ||
2137 | /* | |
2138 | * rib_unlink | |
2139 | * | |
2140 | * Detach a rib structure from a route_node. | |
2141 | * | |
2142 | * Note that a call to rib_unlink() should be followed by a call to | |
2143 | * rib_gc_dest() at some point. This allows a rib_dest_t that is no | |
2144 | * longer required to be deleted. | |
2145 | */ | |
2146 | void rib_unlink(struct route_node *rn, struct route_entry *re) | |
2147 | { | |
2148 | rib_dest_t *dest; | |
2149 | ||
2150 | assert(rn && re); | |
2151 | ||
2152 | if (IS_ZEBRA_DEBUG_RIB) | |
2153 | rnode_debug(rn, re->vrf_id, "rn %p, re %p", (void *)rn, | |
2154 | (void *)re); | |
2155 | ||
2156 | dest = rib_dest_from_rnode(rn); | |
2157 | ||
2158 | if (re->next) | |
2159 | re->next->prev = re->prev; | |
2160 | ||
2161 | if (re->prev) | |
2162 | re->prev->next = re->next; | |
2163 | else { | |
2164 | dest->routes = re->next; | |
2165 | } | |
2166 | ||
2167 | if (dest->selected_fib == re) | |
2168 | dest->selected_fib = NULL; | |
2169 | ||
2170 | nexthops_free(re->ng.nexthop); | |
2171 | XFREE(MTYPE_RE, re); | |
2172 | } | |
2173 | ||
2174 | void rib_delnode(struct route_node *rn, struct route_entry *re) | |
2175 | { | |
2176 | afi_t afi; | |
2177 | ||
2178 | if (IS_ZEBRA_DEBUG_RIB) | |
2179 | rnode_debug(rn, re->vrf_id, "rn %p, re %p, removing", | |
2180 | (void *)rn, (void *)re); | |
2181 | SET_FLAG(re->status, ROUTE_ENTRY_REMOVED); | |
2182 | ||
2183 | afi = (rn->p.family == AF_INET) | |
2184 | ? AFI_IP | |
2185 | : (rn->p.family == AF_INET6) ? AFI_IP6 : AFI_MAX; | |
2186 | if (is_zebra_import_table_enabled(afi, re->table)) { | |
2187 | zebra_del_import_table_entry(rn, re); | |
2188 | /* Just clean up if non main table */ | |
2189 | if (IS_ZEBRA_DEBUG_RIB) { | |
2190 | char buf[SRCDEST2STR_BUFFER]; | |
2191 | srcdest_rnode2str(rn, buf, sizeof(buf)); | |
2192 | zlog_debug( | |
2193 | "%u:%s: Freeing route rn %p, re %p (type %d)", | |
2194 | re->vrf_id, buf, rn, re, re->type); | |
2195 | } | |
2196 | ||
2197 | rib_unlink(rn, re); | |
2198 | } else { | |
2199 | rib_queue_add(rn); | |
2200 | } | |
2201 | } | |
2202 | ||
2203 | /* This function dumps the contents of a given RE entry into | |
2204 | * standard debug log. Calling function name and IP prefix in | |
2205 | * question are passed as 1st and 2nd arguments. | |
2206 | */ | |
2207 | ||
2208 | void _route_entry_dump(const char *func, union prefixconstptr pp, | |
2209 | union prefixconstptr src_pp, | |
2210 | const struct route_entry *re) | |
2211 | { | |
2212 | const struct prefix *p = pp.p; | |
2213 | const struct prefix *src_p = src_pp.p; | |
2214 | bool is_srcdst = src_p && src_p->prefixlen; | |
2215 | char straddr[PREFIX_STRLEN]; | |
2216 | char srcaddr[PREFIX_STRLEN]; | |
2217 | struct nexthop *nexthop; | |
2218 | ||
2219 | zlog_debug("%s: dumping RE entry %p for %s%s%s vrf %u", func, | |
2220 | (const void *)re, prefix2str(pp, straddr, sizeof(straddr)), | |
2221 | is_srcdst ? " from " : "", | |
2222 | is_srcdst ? prefix2str(src_pp, srcaddr, sizeof(srcaddr)) | |
2223 | : "", | |
2224 | re->vrf_id); | |
2225 | zlog_debug("%s: uptime == %lu, type == %u, instance == %d, table == %d", | |
2226 | func, (unsigned long)re->uptime, re->type, re->instance, | |
2227 | re->table); | |
2228 | zlog_debug( | |
2229 | "%s: metric == %u, mtu == %u, distance == %u, flags == %u, status == %u", | |
2230 | func, re->metric, re->mtu, re->distance, re->flags, re->status); | |
2231 | zlog_debug("%s: nexthop_num == %u, nexthop_active_num == %u", func, | |
2232 | re->nexthop_num, re->nexthop_active_num); | |
2233 | ||
2234 | for (ALL_NEXTHOPS(re->ng, nexthop)) { | |
2235 | inet_ntop(p->family, &nexthop->gate, straddr, INET6_ADDRSTRLEN); | |
2236 | zlog_debug("%s: %s %s[%u] vrf %u with flags %s%s%s", func, | |
2237 | (nexthop->rparent ? " NH" : "NH"), straddr, | |
2238 | nexthop->ifindex, nexthop->vrf_id, | |
2239 | (CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE) | |
2240 | ? "ACTIVE " | |
2241 | : ""), | |
2242 | (CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_FIB) | |
2243 | ? "FIB " | |
2244 | : ""), | |
2245 | (CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_RECURSIVE) | |
2246 | ? "RECURSIVE" | |
2247 | : "")); | |
2248 | } | |
2249 | zlog_debug("%s: dump complete", func); | |
2250 | } | |
2251 | ||
2252 | /* This is an exported helper to rtm_read() to dump the strange | |
2253 | * RE entry found by rib_lookup_ipv4_route() | |
2254 | */ | |
2255 | ||
2256 | void rib_lookup_and_dump(struct prefix_ipv4 *p, vrf_id_t vrf_id) | |
2257 | { | |
2258 | struct route_table *table; | |
2259 | struct route_node *rn; | |
2260 | struct route_entry *re; | |
2261 | char prefix_buf[INET_ADDRSTRLEN]; | |
2262 | ||
2263 | /* Lookup table. */ | |
2264 | table = zebra_vrf_table(AFI_IP, SAFI_UNICAST, vrf_id); | |
2265 | if (!table) { | |
2266 | flog_err(ZEBRA_ERR_TABLE_LOOKUP_FAILED, | |
2267 | "%s:%u zebra_vrf_table() returned NULL", __func__, | |
2268 | vrf_id); | |
2269 | return; | |
2270 | } | |
2271 | ||
2272 | /* Scan the RIB table for exactly matching RE entry. */ | |
2273 | rn = route_node_lookup(table, (struct prefix *)p); | |
2274 | ||
2275 | /* No route for this prefix. */ | |
2276 | if (!rn) { | |
2277 | zlog_debug("%s:%u lookup failed for %s", __func__, vrf_id, | |
2278 | prefix2str((struct prefix *)p, prefix_buf, | |
2279 | sizeof(prefix_buf))); | |
2280 | return; | |
2281 | } | |
2282 | ||
2283 | /* Unlock node. */ | |
2284 | route_unlock_node(rn); | |
2285 | ||
2286 | /* let's go */ | |
2287 | RNODE_FOREACH_RE (rn, re) { | |
2288 | zlog_debug("%s:%u rn %p, re %p: %s, %s", | |
2289 | __func__, vrf_id, | |
2290 | (void *)rn, (void *)re, | |
2291 | (CHECK_FLAG(re->status, ROUTE_ENTRY_REMOVED) | |
2292 | ? "removed" | |
2293 | : "NOT removed"), | |
2294 | (CHECK_FLAG(re->flags, ZEBRA_FLAG_SELECTED) | |
2295 | ? "selected" | |
2296 | : "NOT selected")); | |
2297 | route_entry_dump(p, NULL, re); | |
2298 | } | |
2299 | } | |
2300 | ||
2301 | /* Check if requested address assignment will fail due to another | |
2302 | * route being installed by zebra in FIB already. Take necessary | |
2303 | * actions, if needed: remove such a route from FIB and deSELECT | |
2304 | * corresponding RE entry. Then put affected RN into RIBQ head. | |
2305 | */ | |
2306 | void rib_lookup_and_pushup(struct prefix_ipv4 *p, vrf_id_t vrf_id) | |
2307 | { | |
2308 | struct route_table *table; | |
2309 | struct route_node *rn; | |
2310 | unsigned changed = 0; | |
2311 | rib_dest_t *dest; | |
2312 | ||
2313 | if (NULL == (table = zebra_vrf_table(AFI_IP, SAFI_UNICAST, vrf_id))) { | |
2314 | flog_err(ZEBRA_ERR_TABLE_LOOKUP_FAILED, | |
2315 | "%s:%u zebra_vrf_table() returned NULL", __func__, | |
2316 | vrf_id); | |
2317 | return; | |
2318 | } | |
2319 | ||
2320 | /* No matches would be the simplest case. */ | |
2321 | if (NULL == (rn = route_node_lookup(table, (struct prefix *)p))) | |
2322 | return; | |
2323 | ||
2324 | /* Unlock node. */ | |
2325 | route_unlock_node(rn); | |
2326 | ||
2327 | dest = rib_dest_from_rnode(rn); | |
2328 | /* Check all RE entries. In case any changes have to be done, requeue | |
2329 | * the RN into RIBQ head. If the routing message about the new connected | |
2330 | * route (generated by the IP address we are going to assign very soon) | |
2331 | * comes before the RIBQ is processed, the new RE entry will join | |
2332 | * RIBQ record already on head. This is necessary for proper | |
2333 | * revalidation | |
2334 | * of the rest of the RE. | |
2335 | */ | |
2336 | if (dest->selected_fib && !RIB_SYSTEM_ROUTE(dest->selected_fib)) { | |
2337 | changed = 1; | |
2338 | if (IS_ZEBRA_DEBUG_RIB) { | |
2339 | char buf[PREFIX_STRLEN]; | |
2340 | ||
2341 | zlog_debug("%u:%s: freeing way for connected prefix", | |
2342 | dest->selected_fib->vrf_id, | |
2343 | prefix2str(&rn->p, buf, sizeof(buf))); | |
2344 | route_entry_dump(&rn->p, NULL, dest->selected_fib); | |
2345 | } | |
2346 | rib_uninstall(rn, dest->selected_fib); | |
2347 | } | |
2348 | if (changed) | |
2349 | rib_queue_add(rn); | |
2350 | } | |
2351 | ||
2352 | int rib_add_multipath(afi_t afi, safi_t safi, struct prefix *p, | |
2353 | struct prefix_ipv6 *src_p, struct route_entry *re) | |
2354 | { | |
2355 | struct route_table *table; | |
2356 | struct route_node *rn; | |
2357 | struct route_entry *same = NULL; | |
2358 | struct nexthop *nexthop; | |
2359 | int ret = 0; | |
2360 | ||
2361 | if (!re) | |
2362 | return 0; | |
2363 | ||
2364 | assert(!src_p || !src_p->prefixlen || afi == AFI_IP6); | |
2365 | ||
2366 | /* Lookup table. */ | |
2367 | table = zebra_vrf_table_with_table_id(afi, safi, re->vrf_id, re->table); | |
2368 | if (!table) { | |
2369 | XFREE(MTYPE_RE, re); | |
2370 | return 0; | |
2371 | } | |
2372 | ||
2373 | /* Make it sure prefixlen is applied to the prefix. */ | |
2374 | apply_mask(p); | |
2375 | if (src_p) | |
2376 | apply_mask_ipv6(src_p); | |
2377 | ||
2378 | /* Set default distance by route type. */ | |
2379 | if (re->distance == 0) { | |
2380 | re->distance = route_distance(re->type); | |
2381 | ||
2382 | /* iBGP distance is 200. */ | |
2383 | if (re->type == ZEBRA_ROUTE_BGP | |
2384 | && CHECK_FLAG(re->flags, ZEBRA_FLAG_IBGP)) | |
2385 | re->distance = 200; | |
2386 | } | |
2387 | ||
2388 | /* Lookup route node.*/ | |
2389 | rn = srcdest_rnode_get(table, p, src_p); | |
2390 | ||
2391 | /* | |
2392 | * If same type of route are installed, treat it as a implicit | |
2393 | * withdraw. | |
2394 | * If the user has specified the No route replace semantics | |
2395 | * for the install don't do a route replace. | |
2396 | */ | |
2397 | RNODE_FOREACH_RE (rn, same) { | |
2398 | if (CHECK_FLAG(same->status, ROUTE_ENTRY_REMOVED)) | |
2399 | continue; | |
2400 | ||
2401 | if (same->type != re->type) | |
2402 | continue; | |
2403 | if (same->instance != re->instance) | |
2404 | continue; | |
2405 | if (same->type == ZEBRA_ROUTE_KERNEL | |
2406 | && same->metric != re->metric) | |
2407 | continue; | |
2408 | ||
2409 | if (CHECK_FLAG(re->flags, ZEBRA_FLAG_RR_USE_DISTANCE) && | |
2410 | same->distance != re->distance) | |
2411 | continue; | |
2412 | ||
2413 | /* | |
2414 | * We should allow duplicate connected routes | |
2415 | * because of IPv6 link-local routes and unnumbered | |
2416 | * interfaces on Linux. | |
2417 | */ | |
2418 | if (same->type != ZEBRA_ROUTE_CONNECT) | |
2419 | break; | |
2420 | } | |
2421 | ||
2422 | /* If this route is kernel route, set FIB flag to the route. */ | |
2423 | if (RIB_SYSTEM_ROUTE(re)) | |
2424 | for (nexthop = re->ng.nexthop; nexthop; nexthop = nexthop->next) | |
2425 | SET_FLAG(nexthop->flags, NEXTHOP_FLAG_FIB); | |
2426 | ||
2427 | /* Link new re to node.*/ | |
2428 | if (IS_ZEBRA_DEBUG_RIB) { | |
2429 | rnode_debug( | |
2430 | rn, re->vrf_id, | |
2431 | "Inserting route rn %p, re %p (type %d) existing %p", | |
2432 | (void *)rn, (void *)re, re->type, (void *)same); | |
2433 | ||
2434 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) | |
2435 | route_entry_dump(p, src_p, re); | |
2436 | } | |
2437 | rib_addnode(rn, re, 1); | |
2438 | ret = 1; | |
2439 | ||
2440 | /* Free implicit route.*/ | |
2441 | if (same) { | |
2442 | rib_delnode(rn, same); | |
2443 | ret = -1; | |
2444 | } | |
2445 | ||
2446 | route_unlock_node(rn); | |
2447 | return ret; | |
2448 | } | |
2449 | ||
2450 | void rib_delete(afi_t afi, safi_t safi, vrf_id_t vrf_id, int type, | |
2451 | unsigned short instance, int flags, struct prefix *p, | |
2452 | struct prefix_ipv6 *src_p, const struct nexthop *nh, | |
2453 | uint32_t table_id, uint32_t metric, uint8_t distance, | |
2454 | bool fromkernel) | |
2455 | { | |
2456 | struct route_table *table; | |
2457 | struct route_node *rn; | |
2458 | struct route_entry *re; | |
2459 | struct route_entry *fib = NULL; | |
2460 | struct route_entry *same = NULL; | |
2461 | struct nexthop *rtnh; | |
2462 | char buf2[INET6_ADDRSTRLEN]; | |
2463 | rib_dest_t *dest; | |
2464 | ||
2465 | assert(!src_p || !src_p->prefixlen || afi == AFI_IP6); | |
2466 | ||
2467 | /* Lookup table. */ | |
2468 | table = zebra_vrf_table_with_table_id(afi, safi, vrf_id, table_id); | |
2469 | if (!table) | |
2470 | return; | |
2471 | ||
2472 | /* Apply mask. */ | |
2473 | apply_mask(p); | |
2474 | if (src_p) | |
2475 | apply_mask_ipv6(src_p); | |
2476 | ||
2477 | /* Lookup route node. */ | |
2478 | rn = srcdest_rnode_lookup(table, p, src_p); | |
2479 | if (!rn) { | |
2480 | char dst_buf[PREFIX_STRLEN], src_buf[PREFIX_STRLEN]; | |
2481 | ||
2482 | prefix2str(p, dst_buf, sizeof(dst_buf)); | |
2483 | if (src_p && src_p->prefixlen) | |
2484 | prefix2str(src_p, src_buf, sizeof(src_buf)); | |
2485 | else | |
2486 | src_buf[0] = '\0'; | |
2487 | ||
2488 | if (IS_ZEBRA_DEBUG_RIB) | |
2489 | zlog_debug("%u:%s%s%s doesn't exist in rib", vrf_id, | |
2490 | dst_buf, | |
2491 | (src_buf[0] != '\0') ? " from " : "", | |
2492 | src_buf); | |
2493 | return; | |
2494 | } | |
2495 | ||
2496 | dest = rib_dest_from_rnode(rn); | |
2497 | fib = dest->selected_fib; | |
2498 | ||
2499 | /* Lookup same type route. */ | |
2500 | RNODE_FOREACH_RE (rn, re) { | |
2501 | if (CHECK_FLAG(re->status, ROUTE_ENTRY_REMOVED)) | |
2502 | continue; | |
2503 | ||
2504 | if (re->type != type) | |
2505 | continue; | |
2506 | if (re->instance != instance) | |
2507 | continue; | |
2508 | if (CHECK_FLAG(re->flags, ZEBRA_FLAG_RR_USE_DISTANCE) && | |
2509 | distance != re->distance) | |
2510 | continue; | |
2511 | ||
2512 | if (re->type == ZEBRA_ROUTE_KERNEL && re->metric != metric) | |
2513 | continue; | |
2514 | if (re->type == ZEBRA_ROUTE_CONNECT && (rtnh = re->ng.nexthop) | |
2515 | && rtnh->type == NEXTHOP_TYPE_IFINDEX && nh) { | |
2516 | if (rtnh->ifindex != nh->ifindex) | |
2517 | continue; | |
2518 | same = re; | |
2519 | break; | |
2520 | } | |
2521 | /* Make sure that the route found has the same gateway. */ | |
2522 | else { | |
2523 | if (nh == NULL) { | |
2524 | same = re; | |
2525 | break; | |
2526 | } | |
2527 | for (ALL_NEXTHOPS(re->ng, rtnh)) | |
2528 | if (nexthop_same_no_recurse(rtnh, nh)) { | |
2529 | same = re; | |
2530 | break; | |
2531 | } | |
2532 | if (same) | |
2533 | break; | |
2534 | } | |
2535 | } | |
2536 | /* If same type of route can't be found and this message is from | |
2537 | kernel. */ | |
2538 | if (!same) { | |
2539 | /* | |
2540 | * In the past(HA!) we could get here because | |
2541 | * we were receiving a route delete from the | |
2542 | * kernel and we're not marking the proto | |
2543 | * as coming from it's appropriate originator. | |
2544 | * Now that we are properly noticing the fact | |
2545 | * that the kernel has deleted our route we | |
2546 | * are not going to get called in this path | |
2547 | * I am going to leave this here because | |
2548 | * this might still work this way on non-linux | |
2549 | * platforms as well as some weird state I have | |
2550 | * not properly thought of yet. | |
2551 | * If we can show that this code path is | |
2552 | * dead then we can remove it. | |
2553 | */ | |
2554 | if (fib && CHECK_FLAG(flags, ZEBRA_FLAG_SELFROUTE)) { | |
2555 | if (IS_ZEBRA_DEBUG_RIB) { | |
2556 | rnode_debug( | |
2557 | rn, vrf_id, | |
2558 | "rn %p, re %p (type %d) was deleted from kernel, adding", | |
2559 | rn, fib, fib->type); | |
2560 | } | |
2561 | if (allow_delete) { | |
2562 | /* Unset flags. */ | |
2563 | for (rtnh = fib->ng.nexthop; rtnh; | |
2564 | rtnh = rtnh->next) | |
2565 | UNSET_FLAG(rtnh->flags, | |
2566 | NEXTHOP_FLAG_FIB); | |
2567 | ||
2568 | /* | |
2569 | * This is a non FRR route | |
2570 | * as such we should mark | |
2571 | * it as deleted | |
2572 | */ | |
2573 | dest->selected_fib = NULL; | |
2574 | } else { | |
2575 | /* This means someone else, other than Zebra, | |
2576 | * has deleted | |
2577 | * a Zebra router from the kernel. We will add | |
2578 | * it back */ | |
2579 | rib_install_kernel(rn, fib, NULL); | |
2580 | } | |
2581 | } else { | |
2582 | if (IS_ZEBRA_DEBUG_RIB) { | |
2583 | if (nh) | |
2584 | rnode_debug( | |
2585 | rn, vrf_id, | |
2586 | "via %s ifindex %d type %d " | |
2587 | "doesn't exist in rib", | |
2588 | inet_ntop(afi2family(afi), | |
2589 | &nh->gate, buf2, | |
2590 | sizeof(buf2)), | |
2591 | nh->ifindex, type); | |
2592 | else | |
2593 | rnode_debug( | |
2594 | rn, vrf_id, | |
2595 | "type %d doesn't exist in rib", | |
2596 | type); | |
2597 | } | |
2598 | route_unlock_node(rn); | |
2599 | return; | |
2600 | } | |
2601 | } | |
2602 | ||
2603 | if (same) { | |
2604 | if (fromkernel && CHECK_FLAG(flags, ZEBRA_FLAG_SELFROUTE) | |
2605 | && !allow_delete) { | |
2606 | rib_install_kernel(rn, same, NULL); | |
2607 | route_unlock_node(rn); | |
2608 | ||
2609 | return; | |
2610 | } | |
2611 | ||
2612 | if (CHECK_FLAG(flags, ZEBRA_FLAG_EVPN_ROUTE)) { | |
2613 | struct nexthop *tmp_nh; | |
2614 | ||
2615 | for (ALL_NEXTHOPS(re->ng, tmp_nh)) { | |
2616 | struct ipaddr vtep_ip; | |
2617 | ||
2618 | memset(&vtep_ip, 0, sizeof(struct ipaddr)); | |
2619 | if (afi == AFI_IP) { | |
2620 | vtep_ip.ipa_type = IPADDR_V4; | |
2621 | memcpy(&(vtep_ip.ipaddr_v4), | |
2622 | &(tmp_nh->gate.ipv4), | |
2623 | sizeof(struct in_addr)); | |
2624 | } else { | |
2625 | vtep_ip.ipa_type = IPADDR_V6; | |
2626 | memcpy(&(vtep_ip.ipaddr_v6), | |
2627 | &(tmp_nh->gate.ipv6), | |
2628 | sizeof(struct in6_addr)); | |
2629 | } | |
2630 | zebra_vxlan_evpn_vrf_route_del(re->vrf_id, | |
2631 | &vtep_ip, p); | |
2632 | } | |
2633 | } | |
2634 | rib_delnode(rn, same); | |
2635 | } | |
2636 | ||
2637 | route_unlock_node(rn); | |
2638 | return; | |
2639 | } | |
2640 | ||
2641 | ||
2642 | int rib_add(afi_t afi, safi_t safi, vrf_id_t vrf_id, int type, | |
2643 | unsigned short instance, int flags, struct prefix *p, | |
2644 | struct prefix_ipv6 *src_p, const struct nexthop *nh, | |
2645 | uint32_t table_id, uint32_t metric, uint32_t mtu, uint8_t distance, | |
2646 | route_tag_t tag) | |
2647 | { | |
2648 | struct route_entry *re; | |
2649 | struct nexthop *nexthop; | |
2650 | ||
2651 | /* Allocate new route_entry structure. */ | |
2652 | re = XCALLOC(MTYPE_RE, sizeof(struct route_entry)); | |
2653 | re->type = type; | |
2654 | re->instance = instance; | |
2655 | re->distance = distance; | |
2656 | re->flags = flags; | |
2657 | re->metric = metric; | |
2658 | re->mtu = mtu; | |
2659 | re->table = table_id; | |
2660 | re->vrf_id = vrf_id; | |
2661 | re->nexthop_num = 0; | |
2662 | re->uptime = time(NULL); | |
2663 | re->tag = tag; | |
2664 | ||
2665 | /* Add nexthop. */ | |
2666 | nexthop = nexthop_new(); | |
2667 | *nexthop = *nh; | |
2668 | route_entry_nexthop_add(re, nexthop); | |
2669 | ||
2670 | return rib_add_multipath(afi, safi, p, src_p, re); | |
2671 | } | |
2672 | ||
2673 | /* Schedule routes of a particular table (address-family) based on event. */ | |
2674 | void rib_update_table(struct route_table *table, rib_update_event_t event) | |
2675 | { | |
2676 | struct route_node *rn; | |
2677 | struct route_entry *re, *next; | |
2678 | ||
2679 | /* Walk all routes and queue for processing, if appropriate for | |
2680 | * the trigger event. | |
2681 | */ | |
2682 | for (rn = route_top(table); rn; rn = srcdest_route_next(rn)) { | |
2683 | /* | |
2684 | * If we are looking at a route node and the node | |
2685 | * has already been queued we don't | |
2686 | * need to queue it up again | |
2687 | */ | |
2688 | if (rn->info && CHECK_FLAG(rib_dest_from_rnode(rn)->flags, | |
2689 | RIB_ROUTE_ANY_QUEUED)) | |
2690 | continue; | |
2691 | switch (event) { | |
2692 | case RIB_UPDATE_IF_CHANGE: | |
2693 | /* Examine all routes that won't get processed by the | |
2694 | * protocol or | |
2695 | * triggered by nexthop evaluation (NHT). This would be | |
2696 | * system, | |
2697 | * kernel and certain static routes. Note that NHT will | |
2698 | * get | |
2699 | * triggered upon an interface event as connected routes | |
2700 | * always | |
2701 | * get queued for processing. | |
2702 | */ | |
2703 | RNODE_FOREACH_RE_SAFE (rn, re, next) { | |
2704 | struct nexthop *nh; | |
2705 | ||
2706 | if (re->type != ZEBRA_ROUTE_SYSTEM | |
2707 | && re->type != ZEBRA_ROUTE_KERNEL | |
2708 | && re->type != ZEBRA_ROUTE_CONNECT | |
2709 | && re->type != ZEBRA_ROUTE_STATIC) | |
2710 | continue; | |
2711 | ||
2712 | if (re->type != ZEBRA_ROUTE_STATIC) { | |
2713 | rib_queue_add(rn); | |
2714 | continue; | |
2715 | } | |
2716 | ||
2717 | for (nh = re->ng.nexthop; nh; nh = nh->next) | |
2718 | if (!(nh->type == NEXTHOP_TYPE_IPV4 | |
2719 | || nh->type == NEXTHOP_TYPE_IPV6)) | |
2720 | break; | |
2721 | ||
2722 | /* If we only have nexthops to a | |
2723 | * gateway, NHT will | |
2724 | * take care. | |
2725 | */ | |
2726 | if (nh) | |
2727 | rib_queue_add(rn); | |
2728 | } | |
2729 | break; | |
2730 | ||
2731 | case RIB_UPDATE_RMAP_CHANGE: | |
2732 | case RIB_UPDATE_OTHER: | |
2733 | /* Right now, examine all routes. Can restrict to a | |
2734 | * protocol in | |
2735 | * some cases (TODO). | |
2736 | */ | |
2737 | if (rnode_to_ribs(rn)) | |
2738 | rib_queue_add(rn); | |
2739 | break; | |
2740 | ||
2741 | default: | |
2742 | break; | |
2743 | } | |
2744 | } | |
2745 | } | |
2746 | ||
2747 | /* RIB update function. */ | |
2748 | void rib_update(vrf_id_t vrf_id, rib_update_event_t event) | |
2749 | { | |
2750 | struct route_table *table; | |
2751 | ||
2752 | /* Process routes of interested address-families. */ | |
2753 | table = zebra_vrf_table(AFI_IP, SAFI_UNICAST, vrf_id); | |
2754 | if (table) { | |
2755 | if (IS_ZEBRA_DEBUG_EVENT) | |
2756 | zlog_debug("%s : AFI_IP event %d", __func__, event); | |
2757 | rib_update_table(table, event); | |
2758 | } | |
2759 | ||
2760 | table = zebra_vrf_table(AFI_IP6, SAFI_UNICAST, vrf_id); | |
2761 | if (table) { | |
2762 | if (IS_ZEBRA_DEBUG_EVENT) | |
2763 | zlog_debug("%s : AFI_IP6 event %d", __func__, event); | |
2764 | rib_update_table(table, event); | |
2765 | } | |
2766 | } | |
2767 | ||
2768 | /* Delete self installed routes after zebra is relaunched. */ | |
2769 | void rib_sweep_table(struct route_table *table) | |
2770 | { | |
2771 | struct route_node *rn; | |
2772 | struct route_entry *re; | |
2773 | struct route_entry *next; | |
2774 | struct nexthop *nexthop; | |
2775 | ||
2776 | if (!table) | |
2777 | return; | |
2778 | ||
2779 | for (rn = route_top(table); rn; rn = srcdest_route_next(rn)) { | |
2780 | RNODE_FOREACH_RE_SAFE (rn, re, next) { | |
2781 | if (IS_ZEBRA_DEBUG_RIB) | |
2782 | route_entry_dump(&rn->p, NULL, re); | |
2783 | ||
2784 | if (CHECK_FLAG(re->status, ROUTE_ENTRY_REMOVED)) | |
2785 | continue; | |
2786 | ||
2787 | if (!CHECK_FLAG(re->flags, ZEBRA_FLAG_SELFROUTE)) | |
2788 | continue; | |
2789 | ||
2790 | /* | |
2791 | * So we are starting up and have received | |
2792 | * routes from the kernel that we have installed | |
2793 | * from a previous run of zebra but not cleaned | |
2794 | * up ( say a kill -9 ) | |
2795 | * But since we haven't actually installed | |
2796 | * them yet( we received them from the kernel ) | |
2797 | * we don't think they are active. | |
2798 | * So let's pretend they are active to actually | |
2799 | * remove them. | |
2800 | * In all honesty I'm not sure if we should | |
2801 | * mark them as active when we receive them | |
2802 | * This is startup only so probably ok. | |
2803 | * | |
2804 | * If we ever decide to move rib_sweep_table | |
2805 | * to a different spot (ie startup ) | |
2806 | * this decision needs to be revisited | |
2807 | */ | |
2808 | for (ALL_NEXTHOPS(re->ng, nexthop)) | |
2809 | SET_FLAG(nexthop->flags, NEXTHOP_FLAG_FIB); | |
2810 | ||
2811 | rib_uninstall_kernel(rn, re); | |
2812 | rib_delnode(rn, re); | |
2813 | } | |
2814 | } | |
2815 | } | |
2816 | ||
2817 | /* Sweep all RIB tables. */ | |
2818 | void rib_sweep_route(void) | |
2819 | { | |
2820 | struct vrf *vrf; | |
2821 | struct zebra_vrf *zvrf; | |
2822 | ||
2823 | RB_FOREACH (vrf, vrf_id_head, &vrfs_by_id) { | |
2824 | if ((zvrf = vrf->info) == NULL) | |
2825 | continue; | |
2826 | ||
2827 | rib_sweep_table(zvrf->table[AFI_IP][SAFI_UNICAST]); | |
2828 | rib_sweep_table(zvrf->table[AFI_IP6][SAFI_UNICAST]); | |
2829 | } | |
2830 | ||
2831 | zebra_ns_sweep_route(); | |
2832 | } | |
2833 | ||
2834 | /* Remove specific by protocol routes from 'table'. */ | |
2835 | unsigned long rib_score_proto_table(uint8_t proto, unsigned short instance, | |
2836 | struct route_table *table) | |
2837 | { | |
2838 | struct route_node *rn; | |
2839 | struct route_entry *re; | |
2840 | struct route_entry *next; | |
2841 | unsigned long n = 0; | |
2842 | ||
2843 | if (table) | |
2844 | for (rn = route_top(table); rn; rn = srcdest_route_next(rn)) | |
2845 | RNODE_FOREACH_RE_SAFE (rn, re, next) { | |
2846 | if (CHECK_FLAG(re->status, ROUTE_ENTRY_REMOVED)) | |
2847 | continue; | |
2848 | if (re->type == proto | |
2849 | && re->instance == instance) { | |
2850 | rib_delnode(rn, re); | |
2851 | n++; | |
2852 | } | |
2853 | } | |
2854 | return n; | |
2855 | } | |
2856 | ||
2857 | /* Remove specific by protocol routes. */ | |
2858 | unsigned long rib_score_proto(uint8_t proto, unsigned short instance) | |
2859 | { | |
2860 | struct vrf *vrf; | |
2861 | struct zebra_vrf *zvrf; | |
2862 | unsigned long cnt = 0; | |
2863 | ||
2864 | RB_FOREACH (vrf, vrf_id_head, &vrfs_by_id) | |
2865 | if ((zvrf = vrf->info) != NULL) | |
2866 | cnt += rib_score_proto_table( | |
2867 | proto, instance, | |
2868 | zvrf->table[AFI_IP][SAFI_UNICAST]) | |
2869 | + rib_score_proto_table( | |
2870 | proto, instance, | |
2871 | zvrf->table[AFI_IP6][SAFI_UNICAST]); | |
2872 | ||
2873 | cnt += zebra_ns_score_proto(proto, instance); | |
2874 | ||
2875 | return cnt; | |
2876 | } | |
2877 | ||
2878 | /* Close RIB and clean up kernel routes. */ | |
2879 | void rib_close_table(struct route_table *table) | |
2880 | { | |
2881 | struct route_node *rn; | |
2882 | rib_table_info_t *info; | |
2883 | rib_dest_t *dest; | |
2884 | ||
2885 | if (!table) | |
2886 | return; | |
2887 | ||
2888 | info = table->info; | |
2889 | ||
2890 | for (rn = route_top(table); rn; rn = srcdest_route_next(rn)) { | |
2891 | dest = rib_dest_from_rnode(rn); | |
2892 | ||
2893 | if (dest && dest->selected_fib) { | |
2894 | if (info->safi == SAFI_UNICAST) | |
2895 | hook_call(rib_update, rn, NULL); | |
2896 | ||
2897 | if (!RIB_SYSTEM_ROUTE(dest->selected_fib)) | |
2898 | rib_uninstall_kernel(rn, dest->selected_fib); | |
2899 | } | |
2900 | } | |
2901 | } | |
2902 | ||
2903 | /* Routing information base initialize. */ | |
2904 | void rib_init(void) | |
2905 | { | |
2906 | rib_queue_init(&zebrad); | |
2907 | } | |
2908 | ||
2909 | /* | |
2910 | * vrf_id_get_next | |
2911 | * | |
2912 | * Get the first vrf id that is greater than the given vrf id if any. | |
2913 | * | |
2914 | * Returns TRUE if a vrf id was found, FALSE otherwise. | |
2915 | */ | |
2916 | static inline int vrf_id_get_next(vrf_id_t vrf_id, vrf_id_t *next_id_p) | |
2917 | { | |
2918 | struct vrf *vrf; | |
2919 | ||
2920 | vrf = vrf_lookup_by_id(vrf_id); | |
2921 | if (vrf) { | |
2922 | vrf = RB_NEXT(vrf_id_head, vrf); | |
2923 | if (vrf) { | |
2924 | *next_id_p = vrf->vrf_id; | |
2925 | return 1; | |
2926 | } | |
2927 | } | |
2928 | ||
2929 | return 0; | |
2930 | } | |
2931 | ||
2932 | /* | |
2933 | * rib_tables_iter_next | |
2934 | * | |
2935 | * Returns the next table in the iteration. | |
2936 | */ | |
2937 | struct route_table *rib_tables_iter_next(rib_tables_iter_t *iter) | |
2938 | { | |
2939 | struct route_table *table; | |
2940 | ||
2941 | /* | |
2942 | * Array that helps us go over all AFI/SAFI combinations via one | |
2943 | * index. | |
2944 | */ | |
2945 | static struct { | |
2946 | afi_t afi; | |
2947 | safi_t safi; | |
2948 | } afi_safis[] = { | |
2949 | {AFI_IP, SAFI_UNICAST}, {AFI_IP, SAFI_MULTICAST}, | |
2950 | {AFI_IP, SAFI_LABELED_UNICAST}, {AFI_IP6, SAFI_UNICAST}, | |
2951 | {AFI_IP6, SAFI_MULTICAST}, {AFI_IP6, SAFI_LABELED_UNICAST}, | |
2952 | }; | |
2953 | ||
2954 | table = NULL; | |
2955 | ||
2956 | switch (iter->state) { | |
2957 | ||
2958 | case RIB_TABLES_ITER_S_INIT: | |
2959 | iter->vrf_id = VRF_DEFAULT; | |
2960 | iter->afi_safi_ix = -1; | |
2961 | ||
2962 | /* Fall through */ | |
2963 | ||
2964 | case RIB_TABLES_ITER_S_ITERATING: | |
2965 | iter->afi_safi_ix++; | |
2966 | while (1) { | |
2967 | ||
2968 | while (iter->afi_safi_ix | |
2969 | < (int)ZEBRA_NUM_OF(afi_safis)) { | |
2970 | table = zebra_vrf_table( | |
2971 | afi_safis[iter->afi_safi_ix].afi, | |
2972 | afi_safis[iter->afi_safi_ix].safi, | |
2973 | iter->vrf_id); | |
2974 | if (table) | |
2975 | break; | |
2976 | ||
2977 | iter->afi_safi_ix++; | |
2978 | } | |
2979 | ||
2980 | /* | |
2981 | * Found another table in this vrf. | |
2982 | */ | |
2983 | if (table) | |
2984 | break; | |
2985 | ||
2986 | /* | |
2987 | * Done with all tables in the current vrf, go to the | |
2988 | * next | |
2989 | * one. | |
2990 | */ | |
2991 | if (!vrf_id_get_next(iter->vrf_id, &iter->vrf_id)) | |
2992 | break; | |
2993 | ||
2994 | iter->afi_safi_ix = 0; | |
2995 | } | |
2996 | ||
2997 | break; | |
2998 | ||
2999 | case RIB_TABLES_ITER_S_DONE: | |
3000 | return NULL; | |
3001 | } | |
3002 | ||
3003 | if (table) | |
3004 | iter->state = RIB_TABLES_ITER_S_ITERATING; | |
3005 | else | |
3006 | iter->state = RIB_TABLES_ITER_S_DONE; | |
3007 | ||
3008 | return table; | |
3009 | } |