]>
Commit | Line | Data |
---|---|---|
1 | /* | |
2 | * BIRD -- Routing Tables | |
3 | * | |
4 | * (c) 1998--2000 Martin Mares <mj@ucw.cz> | |
5 | * | |
6 | * Can be freely distributed and used under the terms of the GNU GPL. | |
7 | */ | |
8 | ||
9 | /** | |
10 | * DOC: Routing tables | |
11 | * | |
12 | * Routing tables are probably the most important structures BIRD uses. They | |
13 | * hold all the information about known networks, the associated routes and | |
14 | * their attributes. | |
15 | * | |
16 | * There are multiple routing tables (a primary one together with any | |
17 | * number of secondary ones if requested by the configuration). Each table | |
18 | * is basically a FIB containing entries describing the individual | |
19 | * destination networks. For each network (represented by structure &net), | |
20 | * there is a one-way linked list of route entries (&rte), the first entry | |
21 | * on the list being the best one (i.e., the one we currently use | |
22 | * for routing), the order of the other ones is undetermined. | |
23 | * | |
24 | * The &rte contains information about the route. There are net and src, which | |
25 | * together forms a key identifying the route in a routing table. There is a | |
26 | * pointer to a &rta structure (see the route attribute module for a precise | |
27 | * explanation) holding the route attributes, which are primary data about the | |
28 | * route. There are several technical fields used by routing table code (route | |
29 | * id, REF_* flags), There is also the pflags field, holding protocol-specific | |
30 | * flags. They are not used by routing table code, but by protocol-specific | |
31 | * hooks. In contrast to route attributes, they are not primary data and their | |
32 | * validity is also limited to the routing table. | |
33 | * | |
34 | * There are several mechanisms that allow automatic update of routes in one | |
35 | * routing table (dst) as a result of changes in another routing table (src). | |
36 | * They handle issues of recursive next hop resolving, flowspec validation and | |
37 | * RPKI validation. | |
38 | * | |
39 | * The first such mechanism is handling of recursive next hops. A route in the | |
40 | * dst table has an indirect next hop address, which is resolved through a route | |
41 | * in the src table (which may also be the same table) to get an immediate next | |
42 | * hop. This is implemented using structure &hostcache attached to the src | |
43 | * table, which contains &hostentry structures for each tracked next hop | |
44 | * address. These structures are linked from recursive routes in dst tables, | |
45 | * possibly multiple routes sharing one hostentry (as many routes may have the | |
46 | * same indirect next hop). There is also a trie in the hostcache, which matches | |
47 | * all prefixes that may influence resolving of tracked next hops. | |
48 | * | |
49 | * When a best route changes in the src table, the hostcache is notified using | |
50 | * rt_notify_hostcache(), which immediately checks using the trie whether the | |
51 | * change is relevant and if it is, then it schedules asynchronous hostcache | |
52 | * recomputation. The recomputation is done by rt_update_hostcache() (called | |
53 | * from rt_event() of src table), it walks through all hostentries and resolves | |
54 | * them (by rt_update_hostentry()). It also updates the trie. If a change in | |
55 | * hostentry resolution was found, then it schedules asynchronous nexthop | |
56 | * recomputation of associated dst table. That is done by rt_next_hop_update() | |
57 | * (called from rt_event() of dst table), it iterates over all routes in the dst | |
58 | * table and re-examines their hostentries for changes. Note that in contrast to | |
59 | * hostcache update, next hop update can be interrupted by main loop. These two | |
60 | * full-table walks (over hostcache and dst table) are necessary due to absence | |
61 | * of direct lookups (route -> affected nexthop, nexthop -> its route). | |
62 | * | |
63 | * The second mechanism is for flowspec validation, where validity of flowspec | |
64 | * routes depends of resolving their network prefixes in IP routing tables. This | |
65 | * is similar to the recursive next hop mechanism, but simpler as there are no | |
66 | * intermediate hostcache and hostentries (because flows are less likely to | |
67 | * share common net prefix than routes sharing a common next hop). In src table, | |
68 | * there is a list of dst tables (list flowspec_links), this list is updated by | |
69 | * flowpsec channels (by rt_flowspec_link() and rt_flowspec_unlink() during | |
70 | * channel start/stop). Each dst table has its own trie of prefixes that may | |
71 | * influence validation of flowspec routes in it (flowspec_trie). | |
72 | * | |
73 | * When a best route changes in the src table, rt_flowspec_notify() immediately | |
74 | * checks all dst tables from the list using their tries to see whether the | |
75 | * change is relevant for them. If it is, then an asynchronous re-validation of | |
76 | * flowspec routes in the dst table is scheduled. That is also done by function | |
77 | * rt_next_hop_update(), like nexthop recomputation above. It iterates over all | |
78 | * flowspec routes and re-validates them. It also recalculates the trie. | |
79 | * | |
80 | * Note that in contrast to the hostcache update, here the trie is recalculated | |
81 | * during the rt_next_hop_update(), which may be interleaved with IP route | |
82 | * updates. The trie is flushed at the beginning of recalculation, which means | |
83 | * that such updates may use partial trie to see if they are relevant. But it | |
84 | * works anyway! Either affected flowspec was already re-validated and added to | |
85 | * the trie, then IP route change would match the trie and trigger a next round | |
86 | * of re-validation, or it was not yet re-validated and added to the trie, but | |
87 | * will be re-validated later in this round anyway. | |
88 | * | |
89 | * The third mechanism is used for RPKI re-validation of IP routes and it is the | |
90 | * simplest. It is just a list of subscribers in src table, who are notified | |
91 | * when any change happened, but only after a settle time. Also, in RPKI case | |
92 | * the dst is not a table, but a channel, who refeeds routes through a filter. | |
93 | */ | |
94 | ||
95 | #undef LOCAL_DEBUG | |
96 | ||
97 | #include "nest/bird.h" | |
98 | #include "nest/route.h" | |
99 | #include "nest/protocol.h" | |
100 | #include "nest/iface.h" | |
101 | #include "nest/mpls.h" | |
102 | #include "lib/resource.h" | |
103 | #include "lib/event.h" | |
104 | #include "lib/timer.h" | |
105 | #include "lib/string.h" | |
106 | #include "conf/conf.h" | |
107 | #include "filter/filter.h" | |
108 | #include "filter/data.h" | |
109 | #include "lib/hash.h" | |
110 | #include "lib/string.h" | |
111 | #include "lib/alloca.h" | |
112 | #include "lib/flowspec.h" | |
113 | ||
114 | #ifdef CONFIG_BGP | |
115 | #include "proto/bgp/bgp.h" | |
116 | #endif | |
117 | ||
118 | pool *rt_table_pool; | |
119 | ||
120 | static slab *rte_slab; | |
121 | linpool *rte_update_pool; | |
122 | ||
123 | list routing_tables; | |
124 | ||
125 | static void rt_free_hostcache(rtable *tab); | |
126 | static void rt_notify_hostcache(rtable *tab, net *net); | |
127 | static void rt_update_hostcache(rtable *tab); | |
128 | static void rt_next_hop_update(rtable *tab); | |
129 | static inline void rt_prune_table(rtable *tab); | |
130 | static inline void rt_schedule_notify(rtable *tab); | |
131 | static void rt_flowspec_notify(rtable *tab, net *net); | |
132 | static void rt_kick_prune_timer(rtable *tab); | |
133 | ||
134 | ||
135 | static void | |
136 | net_init_with_trie(struct fib *f, void *N) | |
137 | { | |
138 | rtable *tab = SKIP_BACK(rtable, fib, f); | |
139 | net *n = N; | |
140 | ||
141 | if (tab->trie) | |
142 | trie_add_prefix(tab->trie, n->n.addr, n->n.addr->pxlen, n->n.addr->pxlen); | |
143 | ||
144 | if (tab->trie_new) | |
145 | trie_add_prefix(tab->trie_new, n->n.addr, n->n.addr->pxlen, n->n.addr->pxlen); | |
146 | } | |
147 | ||
148 | static inline void * | |
149 | net_route_ip6_sadr_trie(rtable *t, const net_addr_ip6_sadr *n0) | |
150 | { | |
151 | TRIE_WALK_TO_ROOT_IP6(t->trie, (const net_addr_ip6 *) n0, px) | |
152 | { | |
153 | net_addr_ip6_sadr n = NET_ADDR_IP6_SADR(px.prefix, px.pxlen, n0->src_prefix, n0->src_pxlen); | |
154 | net *best = NULL; | |
155 | int best_pxlen = 0; | |
156 | ||
157 | /* We need to do dst first matching. Since sadr addresses are hashed on dst | |
158 | prefix only, find the hash table chain and go through it to find the | |
159 | match with the longest matching src prefix. */ | |
160 | for (struct fib_node *fn = fib_get_chain(&t->fib, (net_addr *) &n); fn; fn = fn->next) | |
161 | { | |
162 | net_addr_ip6_sadr *a = (void *) fn->addr; | |
163 | ||
164 | if (net_equal_dst_ip6_sadr(&n, a) && | |
165 | net_in_net_src_ip6_sadr(&n, a) && | |
166 | (a->src_pxlen >= best_pxlen)) | |
167 | { | |
168 | best = fib_node_to_user(&t->fib, fn); | |
169 | best_pxlen = a->src_pxlen; | |
170 | } | |
171 | } | |
172 | ||
173 | if (best) | |
174 | return best; | |
175 | } | |
176 | TRIE_WALK_TO_ROOT_END; | |
177 | ||
178 | return NULL; | |
179 | } | |
180 | ||
181 | ||
182 | static inline void * | |
183 | net_route_ip6_sadr_fib(rtable *t, const net_addr_ip6_sadr *n0) | |
184 | { | |
185 | net_addr_ip6_sadr n; | |
186 | net_copy_ip6_sadr(&n, n0); | |
187 | ||
188 | while (1) | |
189 | { | |
190 | net *best = NULL; | |
191 | int best_pxlen = 0; | |
192 | ||
193 | /* We need to do dst first matching. Since sadr addresses are hashed on dst | |
194 | prefix only, find the hash table chain and go through it to find the | |
195 | match with the longest matching src prefix. */ | |
196 | for (struct fib_node *fn = fib_get_chain(&t->fib, (net_addr *) &n); fn; fn = fn->next) | |
197 | { | |
198 | net_addr_ip6_sadr *a = (void *) fn->addr; | |
199 | ||
200 | if (net_equal_dst_ip6_sadr(&n, a) && | |
201 | net_in_net_src_ip6_sadr(&n, a) && | |
202 | (a->src_pxlen >= best_pxlen)) | |
203 | { | |
204 | best = fib_node_to_user(&t->fib, fn); | |
205 | best_pxlen = a->src_pxlen; | |
206 | } | |
207 | } | |
208 | ||
209 | if (best) | |
210 | return best; | |
211 | ||
212 | if (!n.dst_pxlen) | |
213 | break; | |
214 | ||
215 | n.dst_pxlen--; | |
216 | ip6_clrbit(&n.dst_prefix, n.dst_pxlen); | |
217 | } | |
218 | ||
219 | return NULL; | |
220 | } | |
221 | ||
222 | net * | |
223 | net_route(rtable *tab, const net_addr *n) | |
224 | { | |
225 | ASSERT(tab->addr_type == n->type); | |
226 | net_addr_union *nu = SKIP_BACK(net_addr_union, n, n); | |
227 | ||
228 | #define TW(ipv, what) \ | |
229 | TRIE_WALK_TO_ROOT_IP##ipv(tab->trie, &(nu->ip##ipv), var) \ | |
230 | { what(ipv, var); } \ | |
231 | TRIE_WALK_TO_ROOT_END; return NULL; | |
232 | ||
233 | #define FW(ipv, what) do { \ | |
234 | net_addr_union nuc; net_copy(&nuc.n, n); \ | |
235 | while (1) { \ | |
236 | what(ipv, nuc.ip##ipv); if (!nuc.n.pxlen) return NULL; \ | |
237 | nuc.n.pxlen--; ip##ipv##_clrbit(&nuc.ip##ipv.prefix, nuc.ip##ipv.pxlen); \ | |
238 | } \ | |
239 | } while(0); return NULL; | |
240 | ||
241 | #define FVR_IP(ipv, var) \ | |
242 | net *r; if (r = net_find_valid(tab, (net_addr *) &var)) return r; | |
243 | ||
244 | #define FVR_VPN(ipv, var) \ | |
245 | net_addr_vpn##ipv _var0 = NET_ADDR_VPN##ipv(var.prefix, var.pxlen, nu->vpn##ipv.rd); FVR_IP(ipv, _var0); | |
246 | ||
247 | if (tab->trie) | |
248 | switch (n->type) { | |
249 | case NET_IP4: TW(4, FVR_IP); | |
250 | case NET_VPN4: TW(4, FVR_VPN); | |
251 | case NET_IP6: TW(6, FVR_IP); | |
252 | case NET_VPN6: TW(6, FVR_VPN); | |
253 | ||
254 | case NET_IP6_SADR: | |
255 | return net_route_ip6_sadr_trie(tab, (net_addr_ip6_sadr *) n); | |
256 | default: | |
257 | return NULL; | |
258 | } | |
259 | else | |
260 | switch (n->type) { | |
261 | case NET_IP4: FW(4, FVR_IP); | |
262 | case NET_VPN4: FW(4, FVR_VPN); | |
263 | case NET_IP6: FW(6, FVR_IP); | |
264 | case NET_VPN6: FW(6, FVR_VPN); | |
265 | ||
266 | case NET_IP6_SADR: | |
267 | return net_route_ip6_sadr_fib (tab, (net_addr_ip6_sadr *) n); | |
268 | default: | |
269 | return NULL; | |
270 | } | |
271 | ||
272 | #undef TW | |
273 | #undef FW | |
274 | #undef FVR_IP | |
275 | #undef FVR_VPN | |
276 | } | |
277 | ||
278 | ||
279 | /** | |
280 | * roa_check - check validity of route origination in a ROA table | |
281 | * @tab: ROA table | |
282 | * @n: network prefix to check | |
283 | * @asn: AS number of network prefix | |
284 | * | |
285 | * Implements RFC 6483 route validation for the given network prefix. The | |
286 | * procedure is to find all candidate ROAs - ROAs whose prefixes cover the given | |
287 | * network prefix. If there is no candidate ROA, return ROA_UNKNOWN. If there is | |
288 | * a candidate ROA with matching ASN and maxlen field greater than or equal to | |
289 | * the given prefix length, return ROA_VALID. Otherwise, return ROA_INVALID. If | |
290 | * caller cannot determine origin AS, 0 could be used (in that case ROA_VALID | |
291 | * cannot happen). Table @tab must have type NET_ROA4 or NET_ROA6, network @n | |
292 | * must have type NET_IP4 or NET_IP6, respectively. | |
293 | */ | |
294 | int | |
295 | net_roa_check(rtable *tab, const net_addr *n, u32 asn) | |
296 | { | |
297 | net_addr_union *nu = SKIP_BACK(net_addr_union, n, n); | |
298 | int anything = 0; | |
299 | struct fib_node *fn; | |
300 | ||
301 | #define TW(ipv) do { \ | |
302 | TRIE_WALK_TO_ROOT_IP##ipv(tab->trie, &(nu->ip##ipv), var) { \ | |
303 | net_addr_roa##ipv roa0 = NET_ADDR_ROA##ipv(var.prefix, var.pxlen, 0, 0); \ | |
304 | ROA_PARTIAL_CHECK(ipv); \ | |
305 | } TRIE_WALK_TO_ROOT_END; \ | |
306 | return anything ? ROA_INVALID : ROA_UNKNOWN; \ | |
307 | } while (0) | |
308 | ||
309 | #define FW(ipv) do { \ | |
310 | net_addr_roa##ipv roa0 = NET_ADDR_ROA##ipv(nu->ip##ipv.prefix, nu->ip##ipv.pxlen, 0, 0);\ | |
311 | while (1) { \ | |
312 | ROA_PARTIAL_CHECK(ipv); \ | |
313 | if (roa0.pxlen == 0) break; \ | |
314 | roa0.pxlen--; ip##ipv##_clrbit(&roa0.prefix, roa0.pxlen); \ | |
315 | } \ | |
316 | } while (0) | |
317 | ||
318 | #define ROA_PARTIAL_CHECK(ipv) do { \ | |
319 | for (fn = fib_get_chain(&tab->fib, (net_addr *) &roa0); fn; fn = fn->next) \ | |
320 | { \ | |
321 | net_addr_roa##ipv *roa = (void *) fn->addr; \ | |
322 | net *r = fib_node_to_user(&tab->fib, fn); \ | |
323 | if (net_equal_prefix_roa##ipv(roa, &roa0) && rte_is_valid(r->routes)) \ | |
324 | { \ | |
325 | anything = 1; \ | |
326 | if (asn && (roa->asn == asn) && (roa->max_pxlen >= nu->ip##ipv.pxlen)) \ | |
327 | return ROA_VALID; \ | |
328 | } \ | |
329 | } \ | |
330 | } while (0) | |
331 | ||
332 | if ((tab->addr_type == NET_ROA4) && (n->type == NET_IP4)) | |
333 | { | |
334 | if (tab->trie) TW(4); | |
335 | else FW(4); | |
336 | } | |
337 | else if ((tab->addr_type == NET_ROA6) && (n->type == NET_IP6)) | |
338 | { | |
339 | if (tab->trie) TW(6); | |
340 | else FW(6); | |
341 | } | |
342 | ||
343 | return anything ? ROA_INVALID : ROA_UNKNOWN; | |
344 | #undef ROA_PARTIAL_CHECK | |
345 | #undef TW | |
346 | #undef FW | |
347 | } | |
348 | ||
349 | /** | |
350 | * aspa_check - check validity of AS Path in an ASPA table | |
351 | * @tab: ASPA table | |
352 | * @path: AS Path to check | |
353 | * | |
354 | * Implements draft-ietf-sidrops-aspa-verification-16. | |
355 | */ | |
356 | enum aspa_result aspa_check(rtable *tab, const adata *path, bool force_upstream) | |
357 | { | |
358 | struct lp_state lps; | |
359 | lp_save(tmp_linpool, &lps); | |
360 | ||
361 | /* No support for confed paths */ | |
362 | if (as_path_contains_confed(path)) | |
363 | return ASPA_INVALID; | |
364 | ||
365 | /* Check path length */ | |
366 | uint len = as_path_getlen(path); | |
367 | if (len == 0) | |
368 | return ASPA_INVALID; | |
369 | ||
370 | /* Normalize the AS Path: drop stuffings */ | |
371 | u32 *asns = alloca(sizeof(u32) * len); | |
372 | uint ppos = 0; | |
373 | uint nsz = 0; | |
374 | while (as_path_walk(path, &ppos, &asns[nsz])) | |
375 | if ((nsz == 0) || (asns[nsz] != asns[nsz-1])) | |
376 | nsz++; | |
377 | ||
378 | /* Find the provider blocks for every AS on the path | |
379 | * and check allowed directions */ | |
380 | uint max_up = 0, min_up = 0, max_down = 0, min_down = 0; | |
381 | ||
382 | for (uint ap=0; ap<nsz; ap++) | |
383 | { | |
384 | net_addr_union nau = { .aspa = NET_ADDR_ASPA(asns[ap]), }; | |
385 | net *n = net_find(tab, &nau.n); | |
386 | ||
387 | bool found = false, down = false, up = false; | |
388 | ||
389 | for (rte *e = (n ? n->routes: NULL); e; e = e->next) | |
390 | { | |
391 | if (!rte_is_valid(e)) | |
392 | continue; | |
393 | ||
394 | eattr *ea = ea_find(e->attrs->eattrs, EA_ASPA_PROVIDERS); | |
395 | if (!ea) | |
396 | continue; | |
397 | ||
398 | /* Actually found some ASPA */ | |
399 | found = true; | |
400 | ||
401 | for (uint i=0; i * sizeof(u32) < ea->u.ptr->length; i++) | |
402 | { | |
403 | if ((ap > 0) && ((u32 *) ea->u.ptr->data)[i] == asns[ap-1]) | |
404 | up = true; | |
405 | if ((ap + 1 < nsz) && ((u32 *) ea->u.ptr->data)[i] == asns[ap+1]) | |
406 | down = true; | |
407 | ||
408 | if (down && up) | |
409 | /* Both peers found */ | |
410 | goto end_of_aspa; | |
411 | } | |
412 | } | |
413 | end_of_aspa:; | |
414 | ||
415 | /* Fast path for the upstream check */ | |
416 | if (force_upstream) | |
417 | { | |
418 | if (!found) | |
419 | /* Move min-upstream */ | |
420 | min_up = ap; | |
421 | else if (ap && !up) | |
422 | /* Exists but doesn't allow this upstream */ | |
423 | return ASPA_INVALID; | |
424 | } | |
425 | ||
426 | /* Fast path for no ASPA here */ | |
427 | else if (!found) | |
428 | { | |
429 | /* Extend max-downstream (min-downstream is stopped by unknown) */ | |
430 | max_down = ap+1; | |
431 | ||
432 | /* Move min-upstream (can't include unknown) */ | |
433 | min_up = ap; | |
434 | } | |
435 | ||
436 | /* ASPA exists and downstream may be extended */ | |
437 | else if (down) | |
438 | { | |
439 | /* Extending max-downstream always */ | |
440 | max_down = ap+1; | |
441 | ||
442 | /* Extending min-downstream unless unknown seen */ | |
443 | if (min_down == ap) | |
444 | min_down = ap+1; | |
445 | ||
446 | /* Downstream only */ | |
447 | if (!up) | |
448 | min_up = max_up = ap; | |
449 | } | |
450 | ||
451 | /* No extension for downstream, force upstream only from now */ | |
452 | else | |
453 | { | |
454 | force_upstream = 1; | |
455 | ||
456 | /* Not even upstream, move the ending here */ | |
457 | if (!up) | |
458 | min_up = max_up = ap; | |
459 | } | |
460 | } | |
461 | ||
462 | /* Is the path surely valid? */ | |
463 | if (min_up <= min_down) | |
464 | return ASPA_VALID; | |
465 | ||
466 | /* Is the path maybe valid? */ | |
467 | if (max_up <= max_down) | |
468 | return ASPA_UNKNOWN; | |
469 | ||
470 | /* Now there is surely a valley there. */ | |
471 | return ASPA_INVALID; | |
472 | } | |
473 | ||
474 | /** | |
475 | * rte_find - find a route | |
476 | * @net: network node | |
477 | * @src: route source | |
478 | * | |
479 | * The rte_find() function returns a route for destination @net | |
480 | * which is from route source @src. | |
481 | */ | |
482 | rte * | |
483 | rte_find(net *net, struct rte_src *src) | |
484 | { | |
485 | rte *e = net->routes; | |
486 | ||
487 | while (e && e->src != src) | |
488 | e = e->next; | |
489 | return e; | |
490 | } | |
491 | ||
492 | /** | |
493 | * rte_get_temp - get a temporary &rte | |
494 | * @a: attributes to assign to the new route (a &rta; in case it's | |
495 | * un-cached, rte_update() will create a cached copy automatically) | |
496 | * @src: route source | |
497 | * | |
498 | * Create a temporary &rte and bind it with the attributes @a. | |
499 | */ | |
500 | rte * | |
501 | rte_get_temp(rta *a, struct rte_src *src) | |
502 | { | |
503 | rte *e = sl_alloc(rte_slab); | |
504 | ||
505 | e->attrs = a; | |
506 | e->id = 0; | |
507 | e->flags = 0; | |
508 | e->pflags = 0; | |
509 | rt_lock_source(e->src = src); | |
510 | return e; | |
511 | } | |
512 | ||
513 | rte * | |
514 | rte_do_cow(rte *r) | |
515 | { | |
516 | rte *e = sl_alloc(rte_slab); | |
517 | ||
518 | memcpy(e, r, sizeof(rte)); | |
519 | ||
520 | rt_lock_source(e->src); | |
521 | e->attrs = rta_clone(r->attrs); | |
522 | e->flags = 0; | |
523 | return e; | |
524 | } | |
525 | ||
526 | /** | |
527 | * rte_cow_rta - get a private writable copy of &rte with writable &rta | |
528 | * @r: a route entry to be copied | |
529 | * @lp: a linpool from which to allocate &rta | |
530 | * | |
531 | * rte_cow_rta() takes a &rte and prepares it and associated &rta for | |
532 | * modification. There are three possibilities: First, both &rte and &rta are | |
533 | * private copies, in that case they are returned unchanged. Second, &rte is | |
534 | * private copy, but &rta is cached, in that case &rta is duplicated using | |
535 | * rta_do_cow(). Third, both &rte is shared and &rta is cached, in that case | |
536 | * both structures are duplicated by rte_do_cow() and rta_do_cow(). | |
537 | * | |
538 | * Note that in the second case, cached &rta loses one reference, while private | |
539 | * copy created by rta_do_cow() is a shallow copy sharing indirect data (eattrs, | |
540 | * nexthops, ...) with it. To work properly, original shared &rta should have | |
541 | * another reference during the life of created private copy. | |
542 | * | |
543 | * Result: a pointer to the new writable &rte with writable &rta. | |
544 | */ | |
545 | rte * | |
546 | rte_cow_rta(rte *r, linpool *lp) | |
547 | { | |
548 | if (!rta_is_cached(r->attrs)) | |
549 | return r; | |
550 | ||
551 | r = rte_cow(r); | |
552 | rta *a = rta_do_cow(r->attrs, lp); | |
553 | rta_free(r->attrs); | |
554 | r->attrs = a; | |
555 | return r; | |
556 | } | |
557 | ||
558 | static int /* Actually better or at least as good as */ | |
559 | rte_better(rte *new, rte *old) | |
560 | { | |
561 | int (*better)(rte *, rte *); | |
562 | ||
563 | if (!rte_is_valid(old)) | |
564 | return 1; | |
565 | if (!rte_is_valid(new)) | |
566 | return 0; | |
567 | ||
568 | if (new->attrs->pref > old->attrs->pref) | |
569 | return 1; | |
570 | if (new->attrs->pref < old->attrs->pref) | |
571 | return 0; | |
572 | if (new->src->proto->proto != old->src->proto->proto) | |
573 | { | |
574 | /* | |
575 | * If the user has configured protocol preferences, so that two different protocols | |
576 | * have the same preference, try to break the tie by comparing addresses. Not too | |
577 | * useful, but keeps the ordering of routes unambiguous. | |
578 | */ | |
579 | return new->src->proto->proto > old->src->proto->proto; | |
580 | } | |
581 | if (better = new->src->proto->rte_better) | |
582 | return better(new, old); | |
583 | return 0; | |
584 | } | |
585 | ||
586 | static int | |
587 | rte_mergable(rte *pri, rte *sec) | |
588 | { | |
589 | int (*mergable)(rte *, rte *); | |
590 | ||
591 | if (!rte_is_valid(pri) || !rte_is_valid(sec)) | |
592 | return 0; | |
593 | ||
594 | if (pri->attrs->pref != sec->attrs->pref) | |
595 | return 0; | |
596 | ||
597 | if (pri->src->proto->proto != sec->src->proto->proto) | |
598 | return 0; | |
599 | ||
600 | if (mergable = pri->src->proto->rte_mergable) | |
601 | return mergable(pri, sec); | |
602 | ||
603 | return 0; | |
604 | } | |
605 | ||
606 | static void | |
607 | rte_trace(struct channel *c, rte *e, int dir, char *msg) | |
608 | { | |
609 | log(L_TRACE "%s.%s %c %s %N %luL %uG %s", | |
610 | c->proto->name, c->name ?: "?", dir, msg, e->net->n.addr, e->src->private_id, e->src->global_id, | |
611 | rta_dest_name(e->attrs->dest)); | |
612 | } | |
613 | ||
614 | static inline void | |
615 | rte_trace_in(uint flag, struct channel *c, rte *e, char *msg) | |
616 | { | |
617 | if ((c->debug & flag) || (c->proto->debug & flag)) | |
618 | rte_trace(c, e, '>', msg); | |
619 | } | |
620 | ||
621 | static inline void | |
622 | rte_trace_out(uint flag, struct channel *c, rte *e, char *msg) | |
623 | { | |
624 | if ((c->debug & flag) || (c->proto->debug & flag)) | |
625 | rte_trace(c, e, '<', msg); | |
626 | } | |
627 | ||
628 | static rte * | |
629 | export_filter_(struct channel *c, rte *rt0, rte **rt_free, linpool *pool, int silent) | |
630 | { | |
631 | struct proto *p = c->proto; | |
632 | const struct filter *filter = c->out_filter; | |
633 | struct proto_stats *stats = &c->stats; | |
634 | rte *rt; | |
635 | int v; | |
636 | ||
637 | rt = rt0; | |
638 | *rt_free = NULL; | |
639 | ||
640 | v = p->preexport ? p->preexport(c, rt) : 0; | |
641 | if (v < 0) | |
642 | { | |
643 | if (silent) | |
644 | goto reject; | |
645 | ||
646 | stats->exp_updates_rejected++; | |
647 | if (v == RIC_REJECT) | |
648 | rte_trace_out(D_FILTERS, c, rt, "rejected by protocol"); | |
649 | goto reject; | |
650 | } | |
651 | if (v > 0) | |
652 | { | |
653 | if (!silent) | |
654 | rte_trace_out(D_FILTERS, c, rt, "forced accept by protocol"); | |
655 | goto accept; | |
656 | } | |
657 | ||
658 | v = filter && ((filter == FILTER_REJECT) || | |
659 | (f_run(filter, &rt, pool, | |
660 | (silent ? FF_SILENT : 0)) > F_ACCEPT)); | |
661 | if (v) | |
662 | { | |
663 | if (silent) | |
664 | goto reject; | |
665 | ||
666 | stats->exp_updates_filtered++; | |
667 | rte_trace_out(D_FILTERS, c, rt, "filtered out"); | |
668 | goto reject; | |
669 | } | |
670 | ||
671 | accept: | |
672 | if (rt != rt0) | |
673 | *rt_free = rt; | |
674 | return rt; | |
675 | ||
676 | reject: | |
677 | /* Discard temporary rte */ | |
678 | if (rt != rt0) | |
679 | rte_free(rt); | |
680 | return NULL; | |
681 | } | |
682 | ||
683 | static inline rte * | |
684 | export_filter(struct channel *c, rte *rt0, rte **rt_free, int silent) | |
685 | { | |
686 | return export_filter_(c, rt0, rt_free, rte_update_pool, silent); | |
687 | } | |
688 | ||
689 | static void | |
690 | do_rt_notify(struct channel *c, net *net, rte *new, rte *old, int refeed) | |
691 | { | |
692 | struct proto *p = c->proto; | |
693 | struct proto_stats *stats = &c->stats; | |
694 | ||
695 | if (refeed && new) | |
696 | c->refeed_count++; | |
697 | ||
698 | /* Apply export limit */ | |
699 | struct channel_limit *l = &c->out_limit; | |
700 | if (l->action && !old && new) | |
701 | { | |
702 | if (stats->exp_routes >= l->limit) | |
703 | channel_notify_limit(c, l, PLD_OUT, stats->exp_routes); | |
704 | ||
705 | if (l->state == PLS_BLOCKED) | |
706 | { | |
707 | stats->exp_updates_rejected++; | |
708 | rte_trace_out(D_FILTERS, c, new, "rejected [limit]"); | |
709 | return; | |
710 | } | |
711 | } | |
712 | ||
713 | /* Apply export table */ | |
714 | if (c->out_table && !rte_update_out(c, net->n.addr, new, old, refeed)) | |
715 | return; | |
716 | ||
717 | if (new) | |
718 | stats->exp_updates_accepted++; | |
719 | else | |
720 | stats->exp_withdraws_accepted++; | |
721 | ||
722 | if (old) | |
723 | { | |
724 | bmap_clear(&c->export_map, old->id); | |
725 | stats->exp_routes--; | |
726 | } | |
727 | ||
728 | if (new) | |
729 | { | |
730 | bmap_set(&c->export_map, new->id); | |
731 | stats->exp_routes++; | |
732 | } | |
733 | ||
734 | if (p->debug & D_ROUTES) | |
735 | { | |
736 | if (new && old) | |
737 | rte_trace_out(D_ROUTES, c, new, "replaced"); | |
738 | else if (new) | |
739 | rte_trace_out(D_ROUTES, c, new, "added"); | |
740 | else if (old) | |
741 | rte_trace_out(D_ROUTES, c, old, "removed"); | |
742 | } | |
743 | ||
744 | p->rt_notify(p, c, net, new, old); | |
745 | } | |
746 | ||
747 | static void | |
748 | rt_notify_basic(struct channel *c, net *net, rte *new, rte *old, int refeed) | |
749 | { | |
750 | // struct proto *p = c->proto; | |
751 | rte *new_free = NULL; | |
752 | ||
753 | if (new) | |
754 | c->stats.exp_updates_received++; | |
755 | else | |
756 | c->stats.exp_withdraws_received++; | |
757 | ||
758 | if (new) | |
759 | new = export_filter(c, new, &new_free, 0); | |
760 | ||
761 | if (old && !bmap_test(&c->export_map, old->id)) | |
762 | old = NULL; | |
763 | ||
764 | if (!new && !old) | |
765 | return; | |
766 | ||
767 | do_rt_notify(c, net, new, old, refeed); | |
768 | ||
769 | /* Discard temporary rte */ | |
770 | if (new_free) | |
771 | rte_free(new_free); | |
772 | } | |
773 | ||
774 | static void | |
775 | rt_notify_accepted(struct channel *c, net *net, rte *new_changed, rte *old_changed, int refeed) | |
776 | { | |
777 | // struct proto *p = c->proto; | |
778 | rte *new_best = NULL; | |
779 | rte *old_best = NULL; | |
780 | rte *new_free = NULL; | |
781 | int new_first = 0; | |
782 | ||
783 | /* | |
784 | * We assume that there are no changes in net route order except (added) | |
785 | * new_changed and (removed) old_changed. Therefore, the function is not | |
786 | * compatible with deterministic_med (where nontrivial reordering can happen | |
787 | * as a result of a route change) and with recomputation of recursive routes | |
788 | * due to next hop update (where many routes can be changed in one step). | |
789 | * | |
790 | * Note that we need this assumption just for optimizations, we could just | |
791 | * run full new_best recomputation otherwise. | |
792 | * | |
793 | * There are three cases: | |
794 | * feed or old_best is old_changed -> we need to recompute new_best | |
795 | * old_best is before new_changed -> new_best is old_best, ignore | |
796 | * old_best is after new_changed -> try new_changed, otherwise old_best | |
797 | */ | |
798 | ||
799 | if (net->routes) | |
800 | c->stats.exp_updates_received++; | |
801 | else | |
802 | c->stats.exp_withdraws_received++; | |
803 | ||
804 | /* Find old_best - either old_changed, or route for net->routes */ | |
805 | if (old_changed && bmap_test(&c->export_map, old_changed->id)) | |
806 | old_best = old_changed; | |
807 | else | |
808 | { | |
809 | for (rte *r = net->routes; rte_is_valid(r); r = r->next) | |
810 | { | |
811 | if (bmap_test(&c->export_map, r->id)) | |
812 | { | |
813 | old_best = r; | |
814 | break; | |
815 | } | |
816 | ||
817 | /* Note if new_changed found before old_best */ | |
818 | if (r == new_changed) | |
819 | new_first = 1; | |
820 | } | |
821 | } | |
822 | ||
823 | /* Find new_best */ | |
824 | if ((new_changed == old_changed) || (old_best == old_changed)) | |
825 | { | |
826 | /* Feed or old_best changed -> find first accepted by filters */ | |
827 | for (rte *r = net->routes; rte_is_valid(r); r = r->next) | |
828 | if (new_best = export_filter(c, r, &new_free, 0)) | |
829 | break; | |
830 | } | |
831 | else | |
832 | { | |
833 | /* Other cases -> either new_changed, or old_best (and nothing changed) */ | |
834 | if (new_first && (new_changed = export_filter(c, new_changed, &new_free, 0))) | |
835 | new_best = new_changed; | |
836 | else | |
837 | return; | |
838 | } | |
839 | ||
840 | if (!new_best && !old_best) | |
841 | return; | |
842 | ||
843 | do_rt_notify(c, net, new_best, old_best, refeed); | |
844 | ||
845 | /* Discard temporary rte */ | |
846 | if (new_free) | |
847 | rte_free(new_free); | |
848 | } | |
849 | ||
850 | ||
851 | static struct nexthop * | |
852 | nexthop_merge_rta(struct nexthop *nhs, rta *a, linpool *pool, int max) | |
853 | { | |
854 | return nexthop_merge(nhs, &(a->nh), 1, 0, max, pool); | |
855 | } | |
856 | ||
857 | rte * | |
858 | rt_export_merged(struct channel *c, net *net, rte **rt_free, linpool *pool, int silent) | |
859 | { | |
860 | // struct proto *p = c->proto; | |
861 | struct nexthop *nhs = NULL; | |
862 | rte *best0, *best, *rt0, *rt, *tmp; | |
863 | ||
864 | best0 = net->routes; | |
865 | *rt_free = NULL; | |
866 | ||
867 | if (!rte_is_valid(best0)) | |
868 | return NULL; | |
869 | ||
870 | best = export_filter_(c, best0, rt_free, pool, silent); | |
871 | ||
872 | if (!best || !rte_is_reachable(best)) | |
873 | return best; | |
874 | ||
875 | for (rt0 = best0->next; rt0; rt0 = rt0->next) | |
876 | { | |
877 | if (!rte_mergable(best0, rt0)) | |
878 | continue; | |
879 | ||
880 | rt = export_filter_(c, rt0, &tmp, pool, 1); | |
881 | ||
882 | if (!rt) | |
883 | continue; | |
884 | ||
885 | if (rte_is_reachable(rt)) | |
886 | nhs = nexthop_merge_rta(nhs, rt->attrs, pool, c->merge_limit); | |
887 | ||
888 | if (tmp) | |
889 | rte_free(tmp); | |
890 | } | |
891 | ||
892 | if (nhs) | |
893 | { | |
894 | nhs = nexthop_merge_rta(nhs, best->attrs, pool, c->merge_limit); | |
895 | ||
896 | if (nhs->next) | |
897 | { | |
898 | best = rte_cow_rta(best, pool); | |
899 | nexthop_link(best->attrs, nhs); | |
900 | } | |
901 | } | |
902 | ||
903 | if (best != best0) | |
904 | *rt_free = best; | |
905 | ||
906 | return best; | |
907 | } | |
908 | ||
909 | static void | |
910 | rt_notify_merged(struct channel *c, net *net, rte *new_changed, rte *old_changed, | |
911 | rte *new_best, rte *old_best, int refeed) | |
912 | { | |
913 | // struct proto *p = c->proto; | |
914 | rte *new_free = NULL; | |
915 | ||
916 | /* We assume that all rte arguments are either NULL or rte_is_valid() */ | |
917 | ||
918 | /* This check should be done by the caller */ | |
919 | if (!new_best && !old_best) | |
920 | return; | |
921 | ||
922 | /* Check whether the change is relevant to the merged route */ | |
923 | if ((new_best == old_best) && | |
924 | (new_changed != old_changed) && | |
925 | !rte_mergable(new_best, new_changed) && | |
926 | !rte_mergable(old_best, old_changed)) | |
927 | return; | |
928 | ||
929 | if (new_best) | |
930 | c->stats.exp_updates_received++; | |
931 | else | |
932 | c->stats.exp_withdraws_received++; | |
933 | ||
934 | /* Prepare new merged route */ | |
935 | if (new_best) | |
936 | new_best = rt_export_merged(c, net, &new_free, rte_update_pool, 0); | |
937 | ||
938 | /* Check old merged route */ | |
939 | if (old_best && !bmap_test(&c->export_map, old_best->id)) | |
940 | old_best = NULL; | |
941 | ||
942 | if (!new_best && !old_best) | |
943 | return; | |
944 | ||
945 | do_rt_notify(c, net, new_best, old_best, refeed); | |
946 | ||
947 | /* Discard temporary rte */ | |
948 | if (new_free) | |
949 | rte_free(new_free); | |
950 | } | |
951 | ||
952 | ||
953 | /** | |
954 | * rte_announce - announce a routing table change | |
955 | * @tab: table the route has been added to | |
956 | * @type: type of route announcement (RA_UNDEF or RA_ANY) | |
957 | * @net: network in question | |
958 | * @new: the new or changed route | |
959 | * @old: the previous route replaced by the new one | |
960 | * @new_best: the new best route for the same network | |
961 | * @old_best: the previous best route for the same network | |
962 | * | |
963 | * This function gets a routing table update and announces it to all protocols | |
964 | * that are connected to the same table by their channels. | |
965 | * | |
966 | * There are two ways of how routing table changes are announced. First, there | |
967 | * is a change of just one route in @net (which may caused a change of the best | |
968 | * route of the network). In this case @new and @old describes the changed route | |
969 | * and @new_best and @old_best describes best routes. Other routes are not | |
970 | * affected, but in sorted table the order of other routes might change. | |
971 | * | |
972 | * Second, There is a bulk change of multiple routes in @net, with shared best | |
973 | * route selection. In such case separate route changes are described using | |
974 | * @type of %RA_ANY, with @new and @old specifying the changed route, while | |
975 | * @new_best and @old_best are NULL. After that, another notification is done | |
976 | * where @new_best and @old_best are filled (may be the same), but @new and @old | |
977 | * are NULL. | |
978 | * | |
979 | * The function announces the change to all associated channels. For each | |
980 | * channel, an appropriate preprocessing is done according to channel &ra_mode. | |
981 | * For example, %RA_OPTIMAL channels receive just changes of best routes. | |
982 | * | |
983 | * In general, we first call preexport() hook of a protocol, which performs | |
984 | * basic checks on the route (each protocol has a right to veto or force accept | |
985 | * of the route before any filter is asked). Then we consult an export filter | |
986 | * of the channel and verify the old route in an export map of the channel. | |
987 | * Finally, the rt_notify() hook of the protocol gets called. | |
988 | * | |
989 | * Note that there are also calls of rt_notify() hooks due to feed, but that is | |
990 | * done outside of scope of rte_announce(). | |
991 | */ | |
992 | static void | |
993 | rte_announce(rtable *tab, uint type, net *net, rte *new, rte *old, | |
994 | rte *new_best, rte *old_best) | |
995 | { | |
996 | if (!rte_is_valid(new)) | |
997 | new = NULL; | |
998 | ||
999 | if (!rte_is_valid(old)) | |
1000 | old = NULL; | |
1001 | ||
1002 | if (!rte_is_valid(new_best)) | |
1003 | new_best = NULL; | |
1004 | ||
1005 | if (!rte_is_valid(old_best)) | |
1006 | old_best = NULL; | |
1007 | ||
1008 | if (!new && !old && !new_best && !old_best) | |
1009 | return; | |
1010 | ||
1011 | if (new_best != old_best) | |
1012 | { | |
1013 | if (new_best) | |
1014 | new_best->sender->stats.pref_routes++; | |
1015 | if (old_best) | |
1016 | old_best->sender->stats.pref_routes--; | |
1017 | ||
1018 | if (tab->hostcache) | |
1019 | rt_notify_hostcache(tab, net); | |
1020 | ||
1021 | if (!EMPTY_LIST(tab->flowspec_links)) | |
1022 | rt_flowspec_notify(tab, net); | |
1023 | } | |
1024 | ||
1025 | rt_schedule_notify(tab); | |
1026 | ||
1027 | struct channel *c; node *n; | |
1028 | WALK_LIST2(c, n, tab->channels, table_node) | |
1029 | { | |
1030 | if (c->export_state == ES_DOWN) | |
1031 | continue; | |
1032 | ||
1033 | if (type && (type != c->ra_mode)) | |
1034 | continue; | |
1035 | ||
1036 | switch (c->ra_mode) | |
1037 | { | |
1038 | case RA_OPTIMAL: | |
1039 | if (new_best != old_best) | |
1040 | rt_notify_basic(c, net, new_best, old_best, 0); | |
1041 | break; | |
1042 | ||
1043 | case RA_ANY: | |
1044 | if (new != old) | |
1045 | rt_notify_basic(c, net, new, old, 0); | |
1046 | break; | |
1047 | ||
1048 | case RA_ACCEPTED: | |
1049 | /* | |
1050 | * The (new != old) condition is problematic here, as it would break | |
1051 | * the second usage pattern (announcement after bulk change, used in | |
1052 | * rt_next_hop_update_net(), which sends both new and old as NULL). | |
1053 | * | |
1054 | * But recursive next hops do not work with sorted tables anyways, | |
1055 | * such configuration is forbidden in BGP and not supported in | |
1056 | * rt_notify_accepted(). | |
1057 | * | |
1058 | * The condition is needed to eliminate spurious announcements where | |
1059 | * both old and new routes are not valid (so they are NULL). | |
1060 | */ | |
1061 | if (new != old) | |
1062 | rt_notify_accepted(c, net, new, old, 0); | |
1063 | break; | |
1064 | ||
1065 | case RA_MERGED: | |
1066 | rt_notify_merged(c, net, new, old, new_best, old_best, 0); | |
1067 | break; | |
1068 | } | |
1069 | } | |
1070 | } | |
1071 | ||
1072 | static inline int | |
1073 | rte_validate(rte *e) | |
1074 | { | |
1075 | int c; | |
1076 | net *n = e->net; | |
1077 | ||
1078 | if (!net_validate(n->n.addr)) | |
1079 | { | |
1080 | log(L_WARN "Ignoring bogus prefix %N received via %s", | |
1081 | n->n.addr, e->sender->proto->name); | |
1082 | return 0; | |
1083 | } | |
1084 | ||
1085 | /* FIXME: better handling different nettypes */ | |
1086 | c = !net_is_flow(n->n.addr) ? | |
1087 | net_classify(n->n.addr): (IADDR_HOST | SCOPE_UNIVERSE); | |
1088 | if ((c < 0) || !(c & IADDR_HOST) || ((c & IADDR_SCOPE_MASK) <= SCOPE_LINK)) | |
1089 | { | |
1090 | log(L_WARN "Ignoring bogus route %N received via %s", | |
1091 | n->n.addr, e->sender->proto->name); | |
1092 | return 0; | |
1093 | } | |
1094 | ||
1095 | if (net_type_match(n->n.addr, NB_DEST) == !e->attrs->dest) | |
1096 | { | |
1097 | /* Exception for flowspec that failed validation */ | |
1098 | if (net_is_flow(n->n.addr) && (e->attrs->dest == RTD_UNREACHABLE)) | |
1099 | return 1; | |
1100 | ||
1101 | log(L_WARN "Ignoring route %N with invalid dest %d received via %s", | |
1102 | n->n.addr, e->attrs->dest, e->sender->proto->name); | |
1103 | return 0; | |
1104 | } | |
1105 | ||
1106 | if ((e->attrs->dest == RTD_UNICAST) && !nexthop_is_sorted(&(e->attrs->nh))) | |
1107 | { | |
1108 | log(L_WARN "Ignoring unsorted multipath route %N received via %s", | |
1109 | n->n.addr, e->sender->proto->name); | |
1110 | return 0; | |
1111 | } | |
1112 | ||
1113 | return 1; | |
1114 | } | |
1115 | ||
1116 | /** | |
1117 | * rte_free - delete a &rte | |
1118 | * @e: &rte to be deleted | |
1119 | * | |
1120 | * rte_free() deletes the given &rte from the routing table it's linked to. | |
1121 | */ | |
1122 | void | |
1123 | rte_free(rte *e) | |
1124 | { | |
1125 | rt_unlock_source(e->src); | |
1126 | if (rta_is_cached(e->attrs)) | |
1127 | rta_free(e->attrs); | |
1128 | sl_free(e); | |
1129 | } | |
1130 | ||
1131 | static inline void | |
1132 | rte_free_quick(rte *e) | |
1133 | { | |
1134 | rt_unlock_source(e->src); | |
1135 | rta_free(e->attrs); | |
1136 | sl_free(e); | |
1137 | } | |
1138 | ||
1139 | int | |
1140 | rte_same(rte *x, rte *y) | |
1141 | { | |
1142 | /* rte.flags / rte.pflags are not checked, as they are internal to rtable */ | |
1143 | return | |
1144 | x->attrs == y->attrs && | |
1145 | x->src == y->src && | |
1146 | rte_is_filtered(x) == rte_is_filtered(y); | |
1147 | } | |
1148 | ||
1149 | static inline int rte_is_ok(rte *e) { return e && !rte_is_filtered(e); } | |
1150 | ||
1151 | static void | |
1152 | rte_recalculate(struct channel *c, net *net, rte *new, struct rte_src *src) | |
1153 | { | |
1154 | struct proto *p = c->proto; | |
1155 | struct rtable *table = c->table; | |
1156 | struct proto_stats *stats = &c->stats; | |
1157 | static struct tbf rl_pipe = TBF_DEFAULT_LOG_LIMITS; | |
1158 | rte *before_old = NULL; | |
1159 | rte *old_best = net->routes; | |
1160 | rte *old = NULL; | |
1161 | rte **k; | |
1162 | ||
1163 | k = &net->routes; /* Find and remove original route from the same protocol */ | |
1164 | while (old = *k) | |
1165 | { | |
1166 | if (old->src == src) | |
1167 | { | |
1168 | /* If there is the same route in the routing table but from | |
1169 | * a different sender, then there are two paths from the | |
1170 | * source protocol to this routing table through transparent | |
1171 | * pipes, which is not allowed. | |
1172 | * | |
1173 | * We log that and ignore the route. If it is withdraw, we | |
1174 | * ignore it completely (there might be 'spurious withdraws', | |
1175 | * see FIXME in do_rte_announce()) | |
1176 | */ | |
1177 | if (old->sender->proto != p) | |
1178 | { | |
1179 | if (new) | |
1180 | { | |
1181 | log_rl(&rl_pipe, L_ERR "Pipe collision detected when sending %N to table %s", | |
1182 | net->n.addr, table->name); | |
1183 | rte_free_quick(new); | |
1184 | } | |
1185 | return; | |
1186 | } | |
1187 | ||
1188 | if (new && rte_same(old, new)) | |
1189 | { | |
1190 | /* No changes, ignore the new route and refresh the old one */ | |
1191 | ||
1192 | old->flags &= ~(REF_STALE | REF_DISCARD | REF_MODIFY); | |
1193 | ||
1194 | if (!rte_is_filtered(new)) | |
1195 | { | |
1196 | stats->imp_updates_ignored++; | |
1197 | rte_trace_in(D_ROUTES, c, new, "ignored"); | |
1198 | } | |
1199 | ||
1200 | rte_free_quick(new); | |
1201 | return; | |
1202 | } | |
1203 | *k = old->next; | |
1204 | table->rt_count--; | |
1205 | break; | |
1206 | } | |
1207 | k = &old->next; | |
1208 | before_old = old; | |
1209 | } | |
1210 | ||
1211 | /* Save the last accessed position */ | |
1212 | rte **pos = k; | |
1213 | ||
1214 | if (!old) | |
1215 | before_old = NULL; | |
1216 | ||
1217 | if (!old && !new) | |
1218 | { | |
1219 | stats->imp_withdraws_ignored++; | |
1220 | return; | |
1221 | } | |
1222 | ||
1223 | int new_ok = rte_is_ok(new); | |
1224 | int old_ok = rte_is_ok(old); | |
1225 | ||
1226 | struct channel_limit *l = &c->rx_limit; | |
1227 | if (l->action && !old && new && !c->in_table) | |
1228 | { | |
1229 | u32 all_routes = stats->imp_routes + stats->filt_routes; | |
1230 | ||
1231 | if (all_routes >= l->limit) | |
1232 | channel_notify_limit(c, l, PLD_RX, all_routes); | |
1233 | ||
1234 | if (l->state == PLS_BLOCKED) | |
1235 | { | |
1236 | /* In receive limit the situation is simple, old is NULL so | |
1237 | we just free new and exit like nothing happened */ | |
1238 | ||
1239 | stats->imp_updates_ignored++; | |
1240 | rte_trace_in(D_FILTERS, c, new, "ignored [limit]"); | |
1241 | rte_free_quick(new); | |
1242 | return; | |
1243 | } | |
1244 | } | |
1245 | ||
1246 | l = &c->in_limit; | |
1247 | if (l->action && !old_ok && new_ok) | |
1248 | { | |
1249 | if (stats->imp_routes >= l->limit) | |
1250 | channel_notify_limit(c, l, PLD_IN, stats->imp_routes); | |
1251 | ||
1252 | if (l->state == PLS_BLOCKED) | |
1253 | { | |
1254 | /* In import limit the situation is more complicated. We | |
1255 | shouldn't just drop the route, we should handle it like | |
1256 | it was filtered. We also have to continue the route | |
1257 | processing if old or new is non-NULL, but we should exit | |
1258 | if both are NULL as this case is probably assumed to be | |
1259 | already handled. */ | |
1260 | ||
1261 | stats->imp_updates_ignored++; | |
1262 | rte_trace_in(D_FILTERS, c, new, "ignored [limit]"); | |
1263 | ||
1264 | if (c->in_keep_filtered) | |
1265 | new->flags |= REF_FILTERED; | |
1266 | else | |
1267 | { rte_free_quick(new); new = NULL; } | |
1268 | ||
1269 | /* Note that old && !new could be possible when | |
1270 | c->in_keep_filtered changed in the recent past. */ | |
1271 | ||
1272 | if (!old && !new) | |
1273 | return; | |
1274 | ||
1275 | new_ok = 0; | |
1276 | goto skip_stats1; | |
1277 | } | |
1278 | } | |
1279 | ||
1280 | if (new_ok) | |
1281 | stats->imp_updates_accepted++; | |
1282 | else if (old_ok) | |
1283 | stats->imp_withdraws_accepted++; | |
1284 | else | |
1285 | stats->imp_withdraws_ignored++; | |
1286 | ||
1287 | if (old_ok || new_ok) | |
1288 | table->last_rt_change = current_time(); | |
1289 | ||
1290 | skip_stats1: | |
1291 | ||
1292 | if (new) | |
1293 | rte_is_filtered(new) ? stats->filt_routes++ : stats->imp_routes++; | |
1294 | if (old) | |
1295 | rte_is_filtered(old) ? stats->filt_routes-- : stats->imp_routes--; | |
1296 | ||
1297 | if (table->config->sorted) | |
1298 | { | |
1299 | /* If routes are sorted, just insert new route to appropriate position */ | |
1300 | if (new) | |
1301 | { | |
1302 | if (before_old && !rte_better(new, before_old)) | |
1303 | k = &before_old->next; | |
1304 | else | |
1305 | k = &net->routes; | |
1306 | ||
1307 | for (; *k; k=&(*k)->next) | |
1308 | if (rte_better(new, *k)) | |
1309 | break; | |
1310 | ||
1311 | new->next = *k; | |
1312 | *k = new; | |
1313 | ||
1314 | table->rt_count++; | |
1315 | } | |
1316 | } | |
1317 | else | |
1318 | { | |
1319 | /* If routes are not sorted, find the best route and move it on | |
1320 | the first position. There are several optimized cases. */ | |
1321 | ||
1322 | if (src->proto->rte_recalculate && src->proto->rte_recalculate(table, net, new, old, old_best)) | |
1323 | goto do_recalculate; | |
1324 | ||
1325 | if (new && rte_better(new, old_best)) | |
1326 | { | |
1327 | /* The first case - the new route is cleary optimal, | |
1328 | we link it at the first position */ | |
1329 | ||
1330 | new->next = net->routes; | |
1331 | net->routes = new; | |
1332 | ||
1333 | table->rt_count++; | |
1334 | } | |
1335 | else if (old == old_best) | |
1336 | { | |
1337 | /* The second case - the old best route disappeared, we add the | |
1338 | new route (if we have any) to the list (we don't care about | |
1339 | position) and then we elect the new optimal route and relink | |
1340 | that route at the first position and announce it. New optimal | |
1341 | route might be NULL if there is no more routes */ | |
1342 | ||
1343 | do_recalculate: | |
1344 | /* Add the new route to the list */ | |
1345 | if (new) | |
1346 | { | |
1347 | new->next = *pos; | |
1348 | *pos = new; | |
1349 | ||
1350 | table->rt_count++; | |
1351 | } | |
1352 | ||
1353 | /* Find a new optimal route (if there is any) */ | |
1354 | if (net->routes) | |
1355 | { | |
1356 | rte **bp = &net->routes; | |
1357 | for (k=&(*bp)->next; *k; k=&(*k)->next) | |
1358 | if (rte_better(*k, *bp)) | |
1359 | bp = k; | |
1360 | ||
1361 | /* And relink it */ | |
1362 | rte *best = *bp; | |
1363 | *bp = best->next; | |
1364 | best->next = net->routes; | |
1365 | net->routes = best; | |
1366 | } | |
1367 | } | |
1368 | else if (new) | |
1369 | { | |
1370 | /* The third case - the new route is not better than the old | |
1371 | best route (therefore old_best != NULL) and the old best | |
1372 | route was not removed (therefore old_best == net->routes). | |
1373 | We just link the new route to the old/last position. */ | |
1374 | ||
1375 | new->next = *pos; | |
1376 | *pos = new; | |
1377 | ||
1378 | table->rt_count++; | |
1379 | } | |
1380 | /* The fourth (empty) case - suboptimal route was removed, nothing to do */ | |
1381 | } | |
1382 | ||
1383 | if (new) | |
1384 | { | |
1385 | new->lastmod = current_time(); | |
1386 | ||
1387 | if (!old) | |
1388 | { | |
1389 | new->id = hmap_first_zero(&table->id_map); | |
1390 | hmap_set(&table->id_map, new->id); | |
1391 | } | |
1392 | else | |
1393 | new->id = old->id; | |
1394 | } | |
1395 | ||
1396 | /* Log the route change */ | |
1397 | if ((c->debug & D_ROUTES) || (p->debug & D_ROUTES)) | |
1398 | { | |
1399 | if (new_ok) | |
1400 | rte_trace(c, new, '>', new == net->routes ? "added [best]" : "added"); | |
1401 | else if (old_ok) | |
1402 | { | |
1403 | if (old != old_best) | |
1404 | rte_trace(c, old, '>', "removed"); | |
1405 | else if (rte_is_ok(net->routes)) | |
1406 | rte_trace(c, old, '>', "removed [replaced]"); | |
1407 | else | |
1408 | rte_trace(c, old, '>', "removed [sole]"); | |
1409 | } | |
1410 | } | |
1411 | ||
1412 | /* Propagate the route change */ | |
1413 | rte_announce(table, RA_UNDEF, net, new, old, net->routes, old_best); | |
1414 | ||
1415 | if (!net->routes && | |
1416 | (table->gc_counter++ >= table->config->gc_threshold)) | |
1417 | rt_kick_prune_timer(table); | |
1418 | ||
1419 | if (old_ok && p->rte_remove) | |
1420 | p->rte_remove(net, old); | |
1421 | if (new_ok && p->rte_insert) | |
1422 | p->rte_insert(net, new); | |
1423 | ||
1424 | if (old) | |
1425 | { | |
1426 | if (!new) | |
1427 | hmap_clear(&table->id_map, old->id); | |
1428 | ||
1429 | rte_free_quick(old); | |
1430 | } | |
1431 | } | |
1432 | ||
1433 | static int rte_update_nest_cnt; /* Nesting counter to allow recursive updates */ | |
1434 | ||
1435 | static inline void | |
1436 | rte_update_lock(void) | |
1437 | { | |
1438 | rte_update_nest_cnt++; | |
1439 | } | |
1440 | ||
1441 | static inline void | |
1442 | rte_update_unlock(void) | |
1443 | { | |
1444 | if (!--rte_update_nest_cnt) | |
1445 | lp_flush(rte_update_pool); | |
1446 | } | |
1447 | ||
1448 | /** | |
1449 | * rte_update - enter a new update to a routing table | |
1450 | * @table: table to be updated | |
1451 | * @c: channel doing the update | |
1452 | * @net: network node | |
1453 | * @p: protocol submitting the update | |
1454 | * @src: protocol originating the update | |
1455 | * @new: a &rte representing the new route or %NULL for route removal. | |
1456 | * | |
1457 | * This function is called by the routing protocols whenever they discover | |
1458 | * a new route or wish to update/remove an existing route. The right announcement | |
1459 | * sequence is to build route attributes first (either un-cached with @aflags set | |
1460 | * to zero or a cached one using rta_lookup(); in this case please note that | |
1461 | * you need to increase the use count of the attributes yourself by calling | |
1462 | * rta_clone()), call rte_get_temp() to obtain a temporary &rte, fill in all | |
1463 | * the appropriate data and finally submit the new &rte by calling rte_update(). | |
1464 | * | |
1465 | * @src specifies the protocol that originally created the route and the meaning | |
1466 | * of protocol-dependent data of @new. If @new is not %NULL, @src have to be the | |
1467 | * same value as @new->attrs->proto. @p specifies the protocol that called | |
1468 | * rte_update(). In most cases it is the same protocol as @src. rte_update() | |
1469 | * stores @p in @new->sender; | |
1470 | * | |
1471 | * When rte_update() gets any route, it automatically validates it (checks, | |
1472 | * whether the network and next hop address are valid IP addresses and also | |
1473 | * whether a normal routing protocol doesn't try to smuggle a host or link | |
1474 | * scope route to the table), converts all protocol dependent attributes stored | |
1475 | * in the &rte to temporary extended attributes, consults import filters of the | |
1476 | * protocol to see if the route should be accepted and/or its attributes modified, | |
1477 | * stores the temporary attributes back to the &rte. | |
1478 | * | |
1479 | * Now, having a "public" version of the route, we | |
1480 | * automatically find any old route defined by the protocol @src | |
1481 | * for network @n, replace it by the new one (or removing it if @new is %NULL), | |
1482 | * recalculate the optimal route for this destination and finally broadcast | |
1483 | * the change (if any) to all routing protocols by calling rte_announce(). | |
1484 | * | |
1485 | * All memory used for attribute lists and other temporary allocations is taken | |
1486 | * from a special linear pool @rte_update_pool and freed when rte_update() | |
1487 | * finishes. | |
1488 | */ | |
1489 | ||
1490 | void | |
1491 | rte_update2(struct channel *c, const net_addr *n, rte *new, struct rte_src *src) | |
1492 | { | |
1493 | struct proto *p = c->proto; | |
1494 | struct proto_stats *stats = &c->stats; | |
1495 | const struct filter *filter = c->in_filter; | |
1496 | struct mpls_fec *fec = NULL; | |
1497 | net *nn; | |
1498 | ||
1499 | ASSERT(c->channel_state == CS_UP); | |
1500 | ||
1501 | rte_update_lock(); | |
1502 | if (new) | |
1503 | { | |
1504 | /* Create a temporary table node */ | |
1505 | nn = alloca(sizeof(net) + n->length); | |
1506 | memset(nn, 0, sizeof(net) + n->length); | |
1507 | net_copy(nn->n.addr, n); | |
1508 | ||
1509 | new->net = nn; | |
1510 | new->sender = c; | |
1511 | ||
1512 | stats->imp_updates_received++; | |
1513 | if (!rte_validate(new)) | |
1514 | { | |
1515 | rte_trace_in(D_FILTERS, c, new, "invalid"); | |
1516 | stats->imp_updates_invalid++; | |
1517 | goto drop; | |
1518 | } | |
1519 | ||
1520 | if (filter == FILTER_REJECT) | |
1521 | { | |
1522 | stats->imp_updates_filtered++; | |
1523 | rte_trace_in(D_FILTERS, c, new, "filtered out"); | |
1524 | ||
1525 | if (! c->in_keep_filtered) | |
1526 | goto drop; | |
1527 | ||
1528 | /* new is a private copy, i could modify it */ | |
1529 | new->flags |= REF_FILTERED; | |
1530 | } | |
1531 | else if (filter) | |
1532 | { | |
1533 | int fr = f_run(filter, &new, rte_update_pool, 0); | |
1534 | if (fr > F_ACCEPT) | |
1535 | { | |
1536 | stats->imp_updates_filtered++; | |
1537 | rte_trace_in(D_FILTERS, c, new, "filtered out"); | |
1538 | ||
1539 | if (! c->in_keep_filtered) | |
1540 | goto drop; | |
1541 | ||
1542 | new->flags |= REF_FILTERED; | |
1543 | } | |
1544 | } | |
1545 | ||
1546 | if (p->mpls_map) | |
1547 | { | |
1548 | if (mpls_handle_rte(p->mpls_map, n, new, rte_update_pool, &fec) < 0) | |
1549 | { | |
1550 | rte_trace_in(D_FILTERS, c, new, "invalid"); | |
1551 | stats->imp_updates_invalid++; | |
1552 | goto drop; | |
1553 | } | |
1554 | } | |
1555 | ||
1556 | if (!rta_is_cached(new->attrs)) /* Need to copy attributes */ | |
1557 | new->attrs = rta_lookup(new->attrs); | |
1558 | new->flags |= REF_COW; | |
1559 | ||
1560 | /* Use the actual struct network, not the dummy one */ | |
1561 | nn = net_get(c->table, n); | |
1562 | new->net = nn; | |
1563 | } | |
1564 | else | |
1565 | { | |
1566 | stats->imp_withdraws_received++; | |
1567 | ||
1568 | if (!(nn = net_find(c->table, n)) || !src) | |
1569 | { | |
1570 | stats->imp_withdraws_ignored++; | |
1571 | rte_update_unlock(); | |
1572 | return; | |
1573 | } | |
1574 | } | |
1575 | ||
1576 | recalc: | |
1577 | /* And recalculate the best route */ | |
1578 | rte_recalculate(c, nn, new, src); | |
1579 | ||
1580 | if (p->mpls_map) | |
1581 | mpls_handle_rte_cleanup(p->mpls_map, &fec); | |
1582 | ||
1583 | rte_update_unlock(); | |
1584 | return; | |
1585 | ||
1586 | drop: | |
1587 | rte_free(new); | |
1588 | new = NULL; | |
1589 | if (nn = net_find(c->table, n)) | |
1590 | goto recalc; | |
1591 | ||
1592 | rte_update_unlock(); | |
1593 | } | |
1594 | ||
1595 | /* Independent call to rte_announce(), used from next hop | |
1596 | recalculation, outside of rte_update(). new must be non-NULL */ | |
1597 | static inline void | |
1598 | rte_announce_i(rtable *tab, uint type, net *net, rte *new, rte *old, | |
1599 | rte *new_best, rte *old_best) | |
1600 | { | |
1601 | rte_update_lock(); | |
1602 | rte_announce(tab, type, net, new, old, new_best, old_best); | |
1603 | rte_update_unlock(); | |
1604 | } | |
1605 | ||
1606 | static inline void | |
1607 | rte_discard(rte *old) /* Non-filtered route deletion, used during garbage collection */ | |
1608 | { | |
1609 | rte_update_lock(); | |
1610 | rte_recalculate(old->sender, old->net, NULL, old->src); | |
1611 | rte_update_unlock(); | |
1612 | } | |
1613 | ||
1614 | /* Modify existing route by protocol hook, used for long-lived graceful restart */ | |
1615 | static inline void | |
1616 | rte_modify(rte *old) | |
1617 | { | |
1618 | rte_update_lock(); | |
1619 | ||
1620 | rte *new = old->sender->proto->rte_modify(old, rte_update_pool); | |
1621 | if (new != old) | |
1622 | { | |
1623 | if (new) | |
1624 | { | |
1625 | if (!rta_is_cached(new->attrs)) | |
1626 | new->attrs = rta_lookup(new->attrs); | |
1627 | new->flags = (old->flags & ~REF_MODIFY) | REF_COW; | |
1628 | } | |
1629 | ||
1630 | rte_recalculate(old->sender, old->net, new, old->src); | |
1631 | } | |
1632 | ||
1633 | rte_update_unlock(); | |
1634 | } | |
1635 | ||
1636 | /* Check rtable for best route to given net whether it would be exported do p */ | |
1637 | int | |
1638 | rt_examine(rtable *t, net_addr *a, struct channel *c, const struct filter *filter) | |
1639 | { | |
1640 | struct proto *p = c->proto; | |
1641 | net *n = net_find(t, a); | |
1642 | rte *rt = n ? n->routes : NULL; | |
1643 | ||
1644 | if (!rte_is_valid(rt)) | |
1645 | return 0; | |
1646 | ||
1647 | rte_update_lock(); | |
1648 | ||
1649 | /* Rest is stripped down export_filter() */ | |
1650 | int v = p->preexport ? p->preexport(c, rt) : 0; | |
1651 | if (v == RIC_PROCESS) | |
1652 | v = (f_run(filter, &rt, rte_update_pool, FF_SILENT) <= F_ACCEPT); | |
1653 | ||
1654 | /* Discard temporary rte */ | |
1655 | if (rt != n->routes) | |
1656 | rte_free(rt); | |
1657 | ||
1658 | rte_update_unlock(); | |
1659 | ||
1660 | return v > 0; | |
1661 | } | |
1662 | ||
1663 | ||
1664 | /** | |
1665 | * rt_refresh_begin - start a refresh cycle | |
1666 | * @t: related routing table | |
1667 | * @c related channel | |
1668 | * | |
1669 | * This function starts a refresh cycle for given routing table and announce | |
1670 | * hook. The refresh cycle is a sequence where the protocol sends all its valid | |
1671 | * routes to the routing table (by rte_update()). After that, all protocol | |
1672 | * routes (more precisely routes with @c as @sender) not sent during the | |
1673 | * refresh cycle but still in the table from the past are pruned. This is | |
1674 | * implemented by marking all related routes as stale by REF_STALE flag in | |
1675 | * rt_refresh_begin(), then marking all related stale routes with REF_DISCARD | |
1676 | * flag in rt_refresh_end() and then removing such routes in the prune loop. | |
1677 | */ | |
1678 | void | |
1679 | rt_refresh_begin(rtable *t, struct channel *c) | |
1680 | { | |
1681 | if (c->debug & D_EVENTS) | |
1682 | log(L_TRACE "%s.%s: Route refresh begin", c->proto->name, c->name); | |
1683 | ||
1684 | FIB_WALK(&t->fib, net, n) | |
1685 | { | |
1686 | rte *e; | |
1687 | for (e = n->routes; e; e = e->next) | |
1688 | if (e->sender == c) | |
1689 | e->flags |= REF_STALE; | |
1690 | } | |
1691 | FIB_WALK_END; | |
1692 | } | |
1693 | ||
1694 | /** | |
1695 | * rt_refresh_end - end a refresh cycle | |
1696 | * @t: related routing table | |
1697 | * @c: related channel | |
1698 | * | |
1699 | * This function ends a refresh cycle for given routing table and announce | |
1700 | * hook. See rt_refresh_begin() for description of refresh cycles. | |
1701 | */ | |
1702 | void | |
1703 | rt_refresh_end(rtable *t, struct channel *c) | |
1704 | { | |
1705 | if (c->debug & D_EVENTS) | |
1706 | log(L_TRACE "%s.%s: Route refresh end", c->proto->name, c->name); | |
1707 | ||
1708 | int prune = 0; | |
1709 | ||
1710 | FIB_WALK(&t->fib, net, n) | |
1711 | { | |
1712 | rte *e; | |
1713 | for (e = n->routes; e; e = e->next) | |
1714 | if ((e->sender == c) && (e->flags & REF_STALE)) | |
1715 | { | |
1716 | e->flags |= REF_DISCARD; | |
1717 | prune = 1; | |
1718 | } | |
1719 | } | |
1720 | FIB_WALK_END; | |
1721 | ||
1722 | if (prune) | |
1723 | rt_schedule_prune(t); | |
1724 | } | |
1725 | ||
1726 | void | |
1727 | rt_modify_stale(rtable *t, struct channel *c) | |
1728 | { | |
1729 | int prune = 0; | |
1730 | ||
1731 | FIB_WALK(&t->fib, net, n) | |
1732 | { | |
1733 | rte *e; | |
1734 | for (e = n->routes; e; e = e->next) | |
1735 | if ((e->sender == c) && (e->flags & REF_STALE) && !(e->flags & REF_FILTERED)) | |
1736 | { | |
1737 | e->flags |= REF_MODIFY; | |
1738 | prune = 1; | |
1739 | } | |
1740 | } | |
1741 | FIB_WALK_END; | |
1742 | ||
1743 | if (prune) | |
1744 | rt_schedule_prune(t); | |
1745 | } | |
1746 | ||
1747 | /** | |
1748 | * rte_dump - dump a route | |
1749 | * @e: &rte to be dumped | |
1750 | * | |
1751 | * This functions dumps contents of a &rte to debug output. | |
1752 | */ | |
1753 | void | |
1754 | rte_dump(struct dump_request *dreq, rte *e) | |
1755 | { | |
1756 | net *n = e->net; | |
1757 | RDUMP("%-1N ", n->n.addr); | |
1758 | RDUMP("PF=%02x ", e->pflags); | |
1759 | rta_dump(dreq, e->attrs); | |
1760 | RDUMP("\n"); | |
1761 | } | |
1762 | ||
1763 | /** | |
1764 | * rt_dump - dump a routing table | |
1765 | * @t: routing table to be dumped | |
1766 | * | |
1767 | * This function dumps contents of a given routing table to debug output. | |
1768 | */ | |
1769 | void | |
1770 | rt_dump(struct dump_request *dreq, rtable *t) | |
1771 | { | |
1772 | RDUMP("Dump of routing table <%s>\n", t->name); | |
1773 | #ifdef DEBUGGING | |
1774 | fib_check(&t->fib); | |
1775 | #endif | |
1776 | FIB_WALK(&t->fib, net, n) | |
1777 | { | |
1778 | rte *e; | |
1779 | for(e=n->routes; e; e=e->next) | |
1780 | rte_dump(dreq, e); | |
1781 | } | |
1782 | FIB_WALK_END; | |
1783 | RDUMP("\n"); | |
1784 | } | |
1785 | ||
1786 | /** | |
1787 | * rt_dump_all - dump all routing tables | |
1788 | * | |
1789 | * This function dumps contents of all routing tables to debug output. | |
1790 | */ | |
1791 | void | |
1792 | rt_dump_all(struct dump_request *dreq) | |
1793 | { | |
1794 | rtable *t; | |
1795 | node *n; | |
1796 | ||
1797 | WALK_LIST2(t, n, routing_tables, n) | |
1798 | rt_dump(dreq, t); | |
1799 | } | |
1800 | ||
1801 | static inline void | |
1802 | rt_schedule_hcu(rtable *tab) | |
1803 | { | |
1804 | if (tab->hcu_scheduled) | |
1805 | return; | |
1806 | ||
1807 | tab->hcu_scheduled = 1; | |
1808 | ev_schedule(tab->rt_event); | |
1809 | } | |
1810 | ||
1811 | static inline void | |
1812 | rt_schedule_nhu(rtable *tab) | |
1813 | { | |
1814 | if (tab->nhu_state == NHU_CLEAN) | |
1815 | ev_schedule(tab->rt_event); | |
1816 | ||
1817 | /* state change: | |
1818 | * NHU_CLEAN -> NHU_SCHEDULED | |
1819 | * NHU_RUNNING -> NHU_DIRTY | |
1820 | */ | |
1821 | tab->nhu_state |= NHU_SCHEDULED; | |
1822 | } | |
1823 | ||
1824 | void | |
1825 | rt_schedule_prune(rtable *tab) | |
1826 | { | |
1827 | if (tab->prune_state == 0) | |
1828 | ev_schedule(tab->rt_event); | |
1829 | ||
1830 | /* state change 0->1, 2->3 */ | |
1831 | tab->prune_state |= 1; | |
1832 | } | |
1833 | ||
1834 | ||
1835 | static void | |
1836 | rt_event(void *ptr) | |
1837 | { | |
1838 | rtable *tab = ptr; | |
1839 | ||
1840 | rt_lock_table(tab); | |
1841 | ||
1842 | if (tab->hcu_scheduled) | |
1843 | rt_update_hostcache(tab); | |
1844 | ||
1845 | if (tab->nhu_state) | |
1846 | rt_next_hop_update(tab); | |
1847 | ||
1848 | if (tab->prune_state) | |
1849 | rt_prune_table(tab); | |
1850 | ||
1851 | rt_unlock_table(tab); | |
1852 | } | |
1853 | ||
1854 | ||
1855 | static void | |
1856 | rt_prune_timer(timer *t) | |
1857 | { | |
1858 | rtable *tab = t->data; | |
1859 | ||
1860 | if (tab->gc_counter >= tab->config->gc_threshold) | |
1861 | rt_schedule_prune(tab); | |
1862 | } | |
1863 | ||
1864 | static void | |
1865 | rt_kick_prune_timer(rtable *tab) | |
1866 | { | |
1867 | /* Return if prune is already scheduled */ | |
1868 | if (tm_active(tab->prune_timer) || (tab->prune_state & 1)) | |
1869 | return; | |
1870 | ||
1871 | /* Randomize GC period to +/- 50% */ | |
1872 | btime gc_period = tab->config->gc_period; | |
1873 | gc_period = (gc_period / 2) + (random_u32() % (uint) gc_period); | |
1874 | tm_start(tab->prune_timer, gc_period); | |
1875 | } | |
1876 | ||
1877 | ||
1878 | static inline btime | |
1879 | rt_settled_time(rtable *tab) | |
1880 | { | |
1881 | ASSUME(tab->base_settle_time != 0); | |
1882 | ||
1883 | return MIN(tab->last_rt_change + tab->config->min_settle_time, | |
1884 | tab->base_settle_time + tab->config->max_settle_time); | |
1885 | } | |
1886 | ||
1887 | static void | |
1888 | rt_settle_timer(timer *t) | |
1889 | { | |
1890 | rtable *tab = t->data; | |
1891 | ||
1892 | if (!tab->base_settle_time) | |
1893 | return; | |
1894 | ||
1895 | btime settled_time = rt_settled_time(tab); | |
1896 | if (current_time() < settled_time) | |
1897 | { | |
1898 | tm_set(tab->settle_timer, settled_time); | |
1899 | return; | |
1900 | } | |
1901 | ||
1902 | /* Settled */ | |
1903 | tab->base_settle_time = 0; | |
1904 | ||
1905 | struct rt_subscription *s; | |
1906 | WALK_LIST(s, tab->subscribers) | |
1907 | s->hook(s); | |
1908 | } | |
1909 | ||
1910 | static void | |
1911 | rt_kick_settle_timer(rtable *tab) | |
1912 | { | |
1913 | tab->base_settle_time = current_time(); | |
1914 | ||
1915 | if (!tab->settle_timer) | |
1916 | tab->settle_timer = tm_new_init(tab->rp, rt_settle_timer, tab, 0, 0); | |
1917 | ||
1918 | if (!tm_active(tab->settle_timer)) | |
1919 | tm_set(tab->settle_timer, rt_settled_time(tab)); | |
1920 | } | |
1921 | ||
1922 | static inline void | |
1923 | rt_schedule_notify(rtable *tab) | |
1924 | { | |
1925 | if (EMPTY_LIST(tab->subscribers)) | |
1926 | return; | |
1927 | ||
1928 | if (tab->base_settle_time) | |
1929 | return; | |
1930 | ||
1931 | rt_kick_settle_timer(tab); | |
1932 | } | |
1933 | ||
1934 | void | |
1935 | rt_subscribe(rtable *tab, struct rt_subscription *s) | |
1936 | { | |
1937 | s->tab = tab; | |
1938 | rt_lock_table(tab); | |
1939 | add_tail(&tab->subscribers, &s->n); | |
1940 | } | |
1941 | ||
1942 | void | |
1943 | rt_unsubscribe(struct rt_subscription *s) | |
1944 | { | |
1945 | rem_node(&s->n); | |
1946 | rt_unlock_table(s->tab); | |
1947 | } | |
1948 | ||
1949 | static struct rt_flowspec_link * | |
1950 | rt_flowspec_find_link(rtable *src, rtable *dst) | |
1951 | { | |
1952 | struct rt_flowspec_link *ln; | |
1953 | WALK_LIST(ln, src->flowspec_links) | |
1954 | if ((ln->src == src) && (ln->dst == dst)) | |
1955 | return ln; | |
1956 | ||
1957 | return NULL; | |
1958 | } | |
1959 | ||
1960 | void | |
1961 | rt_flowspec_link(rtable *src, rtable *dst) | |
1962 | { | |
1963 | ASSERT(rt_is_ip(src)); | |
1964 | ASSERT(rt_is_flow(dst)); | |
1965 | ||
1966 | struct rt_flowspec_link *ln = rt_flowspec_find_link(src, dst); | |
1967 | ||
1968 | if (!ln) | |
1969 | { | |
1970 | rt_lock_table(src); | |
1971 | rt_lock_table(dst); | |
1972 | ||
1973 | ln = mb_allocz(src->rp, sizeof(struct rt_flowspec_link)); | |
1974 | ln->src = src; | |
1975 | ln->dst = dst; | |
1976 | add_tail(&src->flowspec_links, &ln->n); | |
1977 | } | |
1978 | ||
1979 | ln->uc++; | |
1980 | } | |
1981 | ||
1982 | void | |
1983 | rt_flowspec_unlink(rtable *src, rtable *dst) | |
1984 | { | |
1985 | struct rt_flowspec_link *ln = rt_flowspec_find_link(src, dst); | |
1986 | ||
1987 | ASSERT(ln && (ln->uc > 0)); | |
1988 | ||
1989 | ln->uc--; | |
1990 | ||
1991 | if (!ln->uc) | |
1992 | { | |
1993 | rem_node(&ln->n); | |
1994 | mb_free(ln); | |
1995 | ||
1996 | rt_unlock_table(src); | |
1997 | rt_unlock_table(dst); | |
1998 | } | |
1999 | } | |
2000 | ||
2001 | static void | |
2002 | rt_flowspec_notify(rtable *src, net *net) | |
2003 | { | |
2004 | /* Only IP tables are src links */ | |
2005 | ASSERT(rt_is_ip(src)); | |
2006 | ||
2007 | struct rt_flowspec_link *ln; | |
2008 | WALK_LIST(ln, src->flowspec_links) | |
2009 | { | |
2010 | rtable *dst = ln->dst; | |
2011 | ASSERT(rt_is_flow(dst)); | |
2012 | ||
2013 | /* No need to inspect it further if recalculation is already active */ | |
2014 | if ((dst->nhu_state == NHU_SCHEDULED) || (dst->nhu_state == NHU_DIRTY)) | |
2015 | continue; | |
2016 | ||
2017 | if (trie_match_net(dst->flowspec_trie, net->n.addr)) | |
2018 | rt_schedule_nhu(dst); | |
2019 | } | |
2020 | } | |
2021 | ||
2022 | static void | |
2023 | rt_flowspec_reset_trie(rtable *tab) | |
2024 | { | |
2025 | linpool *lp = tab->flowspec_trie->lp; | |
2026 | int ipv4 = tab->flowspec_trie->ipv4; | |
2027 | ||
2028 | lp_flush(lp); | |
2029 | tab->flowspec_trie = f_new_trie(lp, 0); | |
2030 | tab->flowspec_trie->ipv4 = ipv4; | |
2031 | } | |
2032 | ||
2033 | static void | |
2034 | rt_free(resource *_r) | |
2035 | { | |
2036 | rtable *r = (rtable *) _r; | |
2037 | ||
2038 | DBG("Deleting routing table %s\n", r->name); | |
2039 | ASSERT_DIE(r->use_count == 0); | |
2040 | ||
2041 | if (r->internal) | |
2042 | return; | |
2043 | ||
2044 | r->config->table = NULL; | |
2045 | rem_node(&r->n); | |
2046 | ||
2047 | if (r->hostcache) | |
2048 | rt_free_hostcache(r); | |
2049 | ||
2050 | /* Freed automagically by the resource pool | |
2051 | fib_free(&r->fib); | |
2052 | hmap_free(&r->id_map); | |
2053 | rfree(r->rt_event); | |
2054 | rfree(r->settle_timer); | |
2055 | mb_free(r); | |
2056 | */ | |
2057 | } | |
2058 | ||
2059 | static void | |
2060 | rt_res_dump(struct dump_request *dreq, resource *_r) | |
2061 | { | |
2062 | rtable *r = (rtable *) _r; | |
2063 | RDUMP("name \"%s\", addr_type=%s, rt_count=%u, use_count=%d\n", | |
2064 | r->name, net_label[r->addr_type], r->rt_count, r->use_count); | |
2065 | } | |
2066 | ||
2067 | static struct resclass rt_class = { | |
2068 | .name = "Routing table", | |
2069 | .size = sizeof(struct rtable), | |
2070 | .free = rt_free, | |
2071 | .dump = rt_res_dump, | |
2072 | .lookup = NULL, | |
2073 | .memsize = NULL, | |
2074 | }; | |
2075 | ||
2076 | rtable * | |
2077 | rt_setup(pool *pp, struct rtable_config *cf) | |
2078 | { | |
2079 | pool *p = rp_newf(pp, "Routing table %s", cf->name); | |
2080 | ||
2081 | rtable *t = ralloc(p, &rt_class); | |
2082 | t->rp = p; | |
2083 | ||
2084 | t->name = cf->name; | |
2085 | t->config = cf; | |
2086 | t->addr_type = cf->addr_type; | |
2087 | t->debug = cf->debug; | |
2088 | ||
2089 | fib_init(&t->fib, p, t->addr_type, sizeof(net), OFFSETOF(net, n), 0, NULL); | |
2090 | ||
2091 | if (cf->trie_used) | |
2092 | { | |
2093 | t->trie = f_new_trie(lp_new_default(p), 0); | |
2094 | t->trie->ipv4 = net_val_match(t->addr_type, NB_IP4 | NB_VPN4 | NB_ROA4); | |
2095 | ||
2096 | t->fib.init = net_init_with_trie; | |
2097 | } | |
2098 | ||
2099 | init_list(&t->channels); | |
2100 | init_list(&t->flowspec_links); | |
2101 | init_list(&t->subscribers); | |
2102 | ||
2103 | hmap_init(&t->id_map, p, 1024); | |
2104 | hmap_set(&t->id_map, 0); | |
2105 | ||
2106 | if (!(t->internal = cf->internal)) | |
2107 | { | |
2108 | t->rt_event = ev_new_init(p, rt_event, t); | |
2109 | t->prune_timer = tm_new_init(p, rt_prune_timer, t, 0, 0); | |
2110 | t->last_rt_change = t->gc_time = current_time(); | |
2111 | ||
2112 | if (rt_is_flow(t)) | |
2113 | { | |
2114 | t->flowspec_trie = f_new_trie(lp_new_default(p), 0); | |
2115 | t->flowspec_trie->ipv4 = (t->addr_type == NET_FLOW4); | |
2116 | } | |
2117 | } | |
2118 | ||
2119 | return t; | |
2120 | } | |
2121 | ||
2122 | /** | |
2123 | * rt_init - initialize routing tables | |
2124 | * | |
2125 | * This function is called during BIRD startup. It initializes the | |
2126 | * routing table module. | |
2127 | */ | |
2128 | void | |
2129 | rt_init(void) | |
2130 | { | |
2131 | rta_init(); | |
2132 | rt_table_pool = rp_new(&root_pool, "Routing tables"); | |
2133 | rte_update_pool = lp_new_default(rt_table_pool); | |
2134 | rte_slab = sl_new(rt_table_pool, sizeof(rte)); | |
2135 | init_list(&routing_tables); | |
2136 | } | |
2137 | ||
2138 | ||
2139 | /** | |
2140 | * rt_prune_table - prune a routing table | |
2141 | * | |
2142 | * The prune loop scans routing tables and removes routes belonging to flushing | |
2143 | * protocols, discarded routes and also stale network entries. It is called from | |
2144 | * rt_event(). The event is rescheduled if the current iteration do not finish | |
2145 | * the table. The pruning is directed by the prune state (@prune_state), | |
2146 | * specifying whether the prune cycle is scheduled or running, and there | |
2147 | * is also a persistent pruning iterator (@prune_fit). | |
2148 | * | |
2149 | * The prune loop is used also for channel flushing. For this purpose, the | |
2150 | * channels to flush are marked before the iteration and notified after the | |
2151 | * iteration. | |
2152 | */ | |
2153 | static void | |
2154 | rt_prune_table(rtable *tab) | |
2155 | { | |
2156 | struct fib_iterator *fit = &tab->prune_fit; | |
2157 | int limit = 2000; | |
2158 | ||
2159 | struct channel *c; | |
2160 | node *n, *x; | |
2161 | ||
2162 | DBG("Pruning route table %s\n", tab->name); | |
2163 | #ifdef DEBUGGING | |
2164 | fib_check(&tab->fib); | |
2165 | #endif | |
2166 | ||
2167 | if (tab->prune_state == 0) | |
2168 | return; | |
2169 | ||
2170 | if (tab->prune_state == 1) | |
2171 | { | |
2172 | /* Mark channels to flush */ | |
2173 | WALK_LIST2(c, n, tab->channels, table_node) | |
2174 | if (c->channel_state == CS_FLUSHING) | |
2175 | c->flush_active = 1; | |
2176 | ||
2177 | FIB_ITERATE_INIT(fit, &tab->fib); | |
2178 | tab->prune_state = 2; | |
2179 | ||
2180 | tab->gc_counter = 0; | |
2181 | tab->gc_time = current_time(); | |
2182 | ||
2183 | if (tab->prune_trie) | |
2184 | { | |
2185 | /* Init prefix trie pruning */ | |
2186 | tab->trie_new = f_new_trie(lp_new_default(tab->rp), 0); | |
2187 | tab->trie_new->ipv4 = tab->trie->ipv4; | |
2188 | } | |
2189 | } | |
2190 | ||
2191 | again: | |
2192 | FIB_ITERATE_START(&tab->fib, fit, net, n) | |
2193 | { | |
2194 | rte *e; | |
2195 | ||
2196 | rescan: | |
2197 | if (limit <= 0) | |
2198 | { | |
2199 | FIB_ITERATE_PUT(fit); | |
2200 | ev_schedule(tab->rt_event); | |
2201 | return; | |
2202 | } | |
2203 | ||
2204 | for (e=n->routes; e; e=e->next) | |
2205 | { | |
2206 | if (e->sender->flush_active || (e->flags & REF_DISCARD)) | |
2207 | { | |
2208 | rte_discard(e); | |
2209 | limit--; | |
2210 | ||
2211 | goto rescan; | |
2212 | } | |
2213 | ||
2214 | if (e->flags & REF_MODIFY) | |
2215 | { | |
2216 | rte_modify(e); | |
2217 | limit--; | |
2218 | ||
2219 | goto rescan; | |
2220 | } | |
2221 | } | |
2222 | ||
2223 | if (!n->routes) /* Orphaned FIB entry */ | |
2224 | { | |
2225 | FIB_ITERATE_PUT(fit); | |
2226 | fib_delete(&tab->fib, n); | |
2227 | goto again; | |
2228 | } | |
2229 | ||
2230 | if (tab->trie_new) | |
2231 | { | |
2232 | trie_add_prefix(tab->trie_new, n->n.addr, n->n.addr->pxlen, n->n.addr->pxlen); | |
2233 | limit--; | |
2234 | } | |
2235 | } | |
2236 | FIB_ITERATE_END; | |
2237 | ||
2238 | #ifdef DEBUGGING | |
2239 | fib_check(&tab->fib); | |
2240 | #endif | |
2241 | ||
2242 | /* state change 2->0, 3->1 */ | |
2243 | tab->prune_state &= 1; | |
2244 | ||
2245 | if (tab->trie_new) | |
2246 | { | |
2247 | /* Finish prefix trie pruning */ | |
2248 | ||
2249 | if (!tab->trie_lock_count) | |
2250 | { | |
2251 | rfree(tab->trie->lp); | |
2252 | } | |
2253 | else | |
2254 | { | |
2255 | ASSERT(!tab->trie_old); | |
2256 | tab->trie_old = tab->trie; | |
2257 | tab->trie_old_lock_count = tab->trie_lock_count; | |
2258 | tab->trie_lock_count = 0; | |
2259 | } | |
2260 | ||
2261 | tab->trie = tab->trie_new; | |
2262 | tab->trie_new = NULL; | |
2263 | tab->prune_trie = 0; | |
2264 | } | |
2265 | else | |
2266 | { | |
2267 | /* Schedule prefix trie pruning */ | |
2268 | if (tab->trie && !tab->trie_old && (tab->trie->prefix_count > (2 * tab->fib.entries))) | |
2269 | { | |
2270 | /* state change 0->1, 2->3 */ | |
2271 | tab->prune_state |= 1; | |
2272 | tab->prune_trie = 1; | |
2273 | } | |
2274 | } | |
2275 | ||
2276 | if (tab->prune_state > 0) | |
2277 | ev_schedule(tab->rt_event); | |
2278 | ||
2279 | /* FIXME: This should be handled in a better way */ | |
2280 | rt_prune_sources(); | |
2281 | ||
2282 | /* Close flushed channels */ | |
2283 | WALK_LIST2_DELSAFE(c, n, x, tab->channels, table_node) | |
2284 | if (c->flush_active) | |
2285 | { | |
2286 | c->flush_active = 0; | |
2287 | channel_set_state(c, CS_DOWN); | |
2288 | } | |
2289 | ||
2290 | return; | |
2291 | } | |
2292 | ||
2293 | /** | |
2294 | * rt_lock_trie - lock a prefix trie of a routing table | |
2295 | * @tab: routing table with prefix trie to be locked | |
2296 | * | |
2297 | * The prune loop may rebuild the prefix trie and invalidate f_trie_walk_state | |
2298 | * structures. Therefore, asynchronous walks should lock the prefix trie using | |
2299 | * this function. That allows the prune loop to rebuild the trie, but postpones | |
2300 | * its freeing until all walks are done (unlocked by rt_unlock_trie()). | |
2301 | * | |
2302 | * Return a current trie that will be locked, the value should be passed back to | |
2303 | * rt_unlock_trie() for unlocking. | |
2304 | * | |
2305 | */ | |
2306 | struct f_trie * | |
2307 | rt_lock_trie(rtable *tab) | |
2308 | { | |
2309 | ASSERT(tab->trie); | |
2310 | ||
2311 | tab->trie_lock_count++; | |
2312 | return tab->trie; | |
2313 | } | |
2314 | ||
2315 | /** | |
2316 | * rt_unlock_trie - unlock a prefix trie of a routing table | |
2317 | * @tab: routing table with prefix trie to be locked | |
2318 | * @trie: value returned by matching rt_lock_trie() | |
2319 | * | |
2320 | * Done for trie locked by rt_lock_trie() after walk over the trie is done. | |
2321 | * It may free the trie and schedule next trie pruning. | |
2322 | */ | |
2323 | void | |
2324 | rt_unlock_trie(rtable *tab, struct f_trie *trie) | |
2325 | { | |
2326 | ASSERT(trie); | |
2327 | ||
2328 | if (trie == tab->trie) | |
2329 | { | |
2330 | /* Unlock the current prefix trie */ | |
2331 | ASSERT(tab->trie_lock_count); | |
2332 | tab->trie_lock_count--; | |
2333 | } | |
2334 | else if (trie == tab->trie_old) | |
2335 | { | |
2336 | /* Unlock the old prefix trie */ | |
2337 | ASSERT(tab->trie_old_lock_count); | |
2338 | tab->trie_old_lock_count--; | |
2339 | ||
2340 | /* Free old prefix trie that is no longer needed */ | |
2341 | if (!tab->trie_old_lock_count) | |
2342 | { | |
2343 | rfree(tab->trie_old->lp); | |
2344 | tab->trie_old = NULL; | |
2345 | ||
2346 | /* Kick prefix trie pruning that was postponed */ | |
2347 | if (tab->trie && (tab->trie->prefix_count > (2 * tab->fib.entries))) | |
2348 | { | |
2349 | tab->prune_trie = 1; | |
2350 | rt_schedule_prune(tab); | |
2351 | } | |
2352 | } | |
2353 | } | |
2354 | else | |
2355 | log(L_BUG "Invalid arg to rt_unlock_trie()"); | |
2356 | } | |
2357 | ||
2358 | ||
2359 | void | |
2360 | rt_preconfig(struct config *c) | |
2361 | { | |
2362 | init_list(&c->tables); | |
2363 | ||
2364 | rt_new_table(cf_get_symbol(c, "master4"), NET_IP4); | |
2365 | rt_new_table(cf_get_symbol(c, "master6"), NET_IP6); | |
2366 | } | |
2367 | ||
2368 | void | |
2369 | rt_postconfig(struct config *c) | |
2370 | { | |
2371 | uint num_tables = list_length(&c->tables); | |
2372 | btime def_gc_period = 400 MS * num_tables; | |
2373 | def_gc_period = MAX(def_gc_period, 10 S); | |
2374 | def_gc_period = MIN(def_gc_period, 600 S); | |
2375 | ||
2376 | struct rtable_config *rc; | |
2377 | WALK_LIST(rc, c->tables) | |
2378 | if (rc->gc_period == (uint) -1) | |
2379 | rc->gc_period = (uint) def_gc_period; | |
2380 | } | |
2381 | ||
2382 | ||
2383 | /* | |
2384 | * Some functions for handing internal next hop updates | |
2385 | * triggered by rt_schedule_nhu(). | |
2386 | */ | |
2387 | ||
2388 | void | |
2389 | rta_apply_hostentry(rta *a, struct hostentry *he, mpls_label_stack *mls) | |
2390 | { | |
2391 | a->hostentry = he; | |
2392 | a->dest = he->dest; | |
2393 | a->igp_metric = he->igp_metric; | |
2394 | ||
2395 | if (a->dest != RTD_UNICAST) | |
2396 | { | |
2397 | /* No nexthop */ | |
2398 | no_nexthop: | |
2399 | a->nh = (struct nexthop) {}; | |
2400 | if (mls) | |
2401 | { /* Store the label stack for later changes */ | |
2402 | a->nh.labels_orig = a->nh.labels = mls->len; | |
2403 | memcpy(a->nh.label, mls->stack, mls->len * sizeof(u32)); | |
2404 | } | |
2405 | return; | |
2406 | } | |
2407 | ||
2408 | if (((!mls) || (!mls->len)) && he->nexthop_linkable) | |
2409 | { /* Just link the nexthop chain, no label append happens. */ | |
2410 | memcpy(&(a->nh), &(he->src->nh), nexthop_size(&(he->src->nh))); | |
2411 | return; | |
2412 | } | |
2413 | ||
2414 | struct nexthop *nhp = NULL, *nhr = NULL; | |
2415 | int skip_nexthop = 0; | |
2416 | ||
2417 | for (struct nexthop *nh = &(he->src->nh); nh; nh = nh->next) | |
2418 | { | |
2419 | if (skip_nexthop) | |
2420 | skip_nexthop--; | |
2421 | else | |
2422 | { | |
2423 | nhr = nhp; | |
2424 | nhp = (nhp ? (nhp->next = lp_alloc(rte_update_pool, NEXTHOP_MAX_SIZE)) : &(a->nh)); | |
2425 | } | |
2426 | ||
2427 | memset(nhp, 0, NEXTHOP_MAX_SIZE); | |
2428 | nhp->iface = nh->iface; | |
2429 | nhp->weight = nh->weight; | |
2430 | ||
2431 | if (mls) | |
2432 | { | |
2433 | nhp->labels = nh->labels + mls->len; | |
2434 | nhp->labels_orig = mls->len; | |
2435 | if (nhp->labels <= MPLS_MAX_LABEL_STACK) | |
2436 | { | |
2437 | memcpy(nhp->label, nh->label, nh->labels * sizeof(u32)); /* First the hostentry labels */ | |
2438 | memcpy(&(nhp->label[nh->labels]), mls->stack, mls->len * sizeof(u32)); /* Then the bottom labels */ | |
2439 | } | |
2440 | else | |
2441 | { | |
2442 | log(L_WARN "Sum of label stack sizes %d + %d = %d exceedes allowed maximum (%d)", | |
2443 | nh->labels, mls->len, nhp->labels, MPLS_MAX_LABEL_STACK); | |
2444 | skip_nexthop++; | |
2445 | continue; | |
2446 | } | |
2447 | } | |
2448 | else if (nh->labels) | |
2449 | { | |
2450 | nhp->labels = nh->labels; | |
2451 | nhp->labels_orig = 0; | |
2452 | memcpy(nhp->label, nh->label, nh->labels * sizeof(u32)); | |
2453 | } | |
2454 | ||
2455 | if (ipa_nonzero(nh->gw)) | |
2456 | { | |
2457 | nhp->gw = nh->gw; /* Router nexthop */ | |
2458 | nhp->flags |= (nh->flags & RNF_ONLINK); | |
2459 | } | |
2460 | else if (!(nh->iface->flags & IF_MULTIACCESS) || (nh->iface->flags & IF_LOOPBACK)) | |
2461 | nhp->gw = IPA_NONE; /* PtP link - no need for nexthop */ | |
2462 | else if (ipa_nonzero(he->link)) | |
2463 | nhp->gw = he->link; /* Device nexthop with link-local address known */ | |
2464 | else | |
2465 | nhp->gw = he->addr; /* Device nexthop with link-local address unknown */ | |
2466 | } | |
2467 | ||
2468 | if (skip_nexthop) | |
2469 | if (nhr) | |
2470 | nhr->next = NULL; | |
2471 | else | |
2472 | { | |
2473 | a->dest = RTD_UNREACHABLE; | |
2474 | log(L_WARN "No valid nexthop remaining, setting route unreachable"); | |
2475 | goto no_nexthop; | |
2476 | } | |
2477 | } | |
2478 | ||
2479 | static inline int | |
2480 | rta_next_hop_outdated(rta *a) | |
2481 | { | |
2482 | struct hostentry *he = a->hostentry; | |
2483 | ||
2484 | if (!he) | |
2485 | return 0; | |
2486 | ||
2487 | if (!he->src) | |
2488 | return a->dest != RTD_UNREACHABLE; | |
2489 | ||
2490 | return (a->dest != he->dest) || (a->igp_metric != he->igp_metric) || | |
2491 | (!he->nexthop_linkable) || !nexthop_same(&(a->nh), &(he->src->nh)); | |
2492 | } | |
2493 | ||
2494 | static inline rte * | |
2495 | rt_next_hop_update_rte(rtable *tab UNUSED, rte *old) | |
2496 | { | |
2497 | if (!rta_next_hop_outdated(old->attrs)) | |
2498 | return NULL; | |
2499 | ||
2500 | rta *a = alloca(RTA_MAX_SIZE); | |
2501 | memcpy(a, old->attrs, rta_size(old->attrs)); | |
2502 | ||
2503 | mpls_label_stack mls = { .len = a->nh.labels_orig }; | |
2504 | memcpy(mls.stack, &a->nh.label[a->nh.labels - mls.len], mls.len * sizeof(u32)); | |
2505 | ||
2506 | rta_apply_hostentry(a, old->attrs->hostentry, &mls); | |
2507 | a->cached = 0; | |
2508 | ||
2509 | rte *e = sl_alloc(rte_slab); | |
2510 | memcpy(e, old, sizeof(rte)); | |
2511 | e->attrs = rta_lookup(a); | |
2512 | rt_lock_source(e->src); | |
2513 | ||
2514 | return e; | |
2515 | } | |
2516 | ||
2517 | ||
2518 | #ifdef CONFIG_BGP | |
2519 | ||
2520 | static inline int | |
2521 | net_flow_has_dst_prefix(const net_addr *n) | |
2522 | { | |
2523 | ASSUME(net_is_flow(n)); | |
2524 | ||
2525 | if (n->pxlen) | |
2526 | return 1; | |
2527 | ||
2528 | if (n->type == NET_FLOW4) | |
2529 | { | |
2530 | const net_addr_flow4 *n4 = (void *) n; | |
2531 | return (n4->length > sizeof(net_addr_flow4)) && (n4->data[0] == FLOW_TYPE_DST_PREFIX); | |
2532 | } | |
2533 | else | |
2534 | { | |
2535 | const net_addr_flow6 *n6 = (void *) n; | |
2536 | return (n6->length > sizeof(net_addr_flow6)) && (n6->data[0] == FLOW_TYPE_DST_PREFIX); | |
2537 | } | |
2538 | } | |
2539 | ||
2540 | static inline int | |
2541 | rta_as_path_is_empty(rta *a) | |
2542 | { | |
2543 | eattr *e = ea_find(a->eattrs, EA_CODE(PROTOCOL_BGP, BA_AS_PATH)); | |
2544 | return !e || (as_path_getlen(e->u.ptr) == 0); | |
2545 | } | |
2546 | ||
2547 | static inline u32 | |
2548 | rta_get_first_asn(rta *a) | |
2549 | { | |
2550 | eattr *e = ea_find(a->eattrs, EA_CODE(PROTOCOL_BGP, BA_AS_PATH)); | |
2551 | u32 asn; | |
2552 | ||
2553 | return (e && as_path_get_first_regular(e->u.ptr, &asn)) ? asn : 0; | |
2554 | } | |
2555 | ||
2556 | int | |
2557 | rt_flowspec_check(rtable *tab_ip, rtable *tab_flow, const net_addr *n, rta *a, int interior) | |
2558 | { | |
2559 | ASSERT(rt_is_ip(tab_ip)); | |
2560 | ASSERT(rt_is_flow(tab_flow)); | |
2561 | ASSERT(tab_ip->trie); | |
2562 | ||
2563 | /* RFC 8955 6. a) Flowspec has defined dst prefix */ | |
2564 | if (!net_flow_has_dst_prefix(n)) | |
2565 | return 0; | |
2566 | ||
2567 | /* RFC 9117 4.1. Accept AS_PATH is empty (fr */ | |
2568 | if (interior && rta_as_path_is_empty(a)) | |
2569 | return 1; | |
2570 | ||
2571 | ||
2572 | /* RFC 8955 6. b) Flowspec and its best-match route have the same originator */ | |
2573 | ||
2574 | /* Find flowspec dst prefix */ | |
2575 | net_addr dst; | |
2576 | if (n->type == NET_FLOW4) | |
2577 | net_fill_ip4(&dst, net4_prefix(n), net4_pxlen(n)); | |
2578 | else | |
2579 | net_fill_ip6(&dst, net6_prefix(n), net6_pxlen(n)); | |
2580 | ||
2581 | /* Find best-match BGP unicast route for flowspec dst prefix */ | |
2582 | net *nb = net_route(tab_ip, &dst); | |
2583 | rte *rb = nb ? nb->routes : NULL; | |
2584 | ||
2585 | /* Register prefix to trie for tracking further changes */ | |
2586 | int max_pxlen = (n->type == NET_FLOW4) ? IP4_MAX_PREFIX_LENGTH : IP6_MAX_PREFIX_LENGTH; | |
2587 | trie_add_prefix(tab_flow->flowspec_trie, &dst, (nb ? nb->n.addr->pxlen : 0), max_pxlen); | |
2588 | ||
2589 | /* No best-match BGP route -> no flowspec */ | |
2590 | if (!rb || (rb->attrs->source != RTS_BGP)) | |
2591 | return 0; | |
2592 | ||
2593 | /* Find ORIGINATOR_ID values */ | |
2594 | u32 orig_a = ea_get_int(a->eattrs, EA_CODE(PROTOCOL_BGP, BA_ORIGINATOR_ID), 0); | |
2595 | u32 orig_b = ea_get_int(rb->attrs->eattrs, EA_CODE(PROTOCOL_BGP, BA_ORIGINATOR_ID), 0); | |
2596 | ||
2597 | /* Originator is either ORIGINATOR_ID (if present), or BGP neighbor address (if not) */ | |
2598 | if ((orig_a != orig_b) || (!orig_a && !orig_b && !ipa_equal(a->from, rb->attrs->from))) | |
2599 | return 0; | |
2600 | ||
2601 | ||
2602 | /* Find ASN of the best-match route, for use in next checks */ | |
2603 | u32 asn_b = rta_get_first_asn(rb->attrs); | |
2604 | if (!asn_b) | |
2605 | return 0; | |
2606 | ||
2607 | /* RFC 9117 4.2. For EBGP, flowspec and its best-match route are from the same AS */ | |
2608 | if (!interior && (rta_get_first_asn(a) != asn_b)) | |
2609 | return 0; | |
2610 | ||
2611 | /* RFC 8955 6. c) More-specific routes are from the same AS as the best-match route */ | |
2612 | TRIE_WALK(tab_ip->trie, subnet, &dst) | |
2613 | { | |
2614 | net *nc = net_find_valid(tab_ip, &subnet); | |
2615 | if (!nc) | |
2616 | continue; | |
2617 | ||
2618 | rte *rc = nc->routes; | |
2619 | if (rc->attrs->source != RTS_BGP) | |
2620 | return 0; | |
2621 | ||
2622 | if (rta_get_first_asn(rc->attrs) != asn_b) | |
2623 | return 0; | |
2624 | } | |
2625 | TRIE_WALK_END; | |
2626 | ||
2627 | return 1; | |
2628 | } | |
2629 | ||
2630 | #endif /* CONFIG_BGP */ | |
2631 | ||
2632 | static rte * | |
2633 | rt_flowspec_update_rte(rtable *tab, rte *r) | |
2634 | { | |
2635 | #ifdef CONFIG_BGP | |
2636 | if ((r->attrs->source != RTS_BGP) || (r->sender->proto != r->src->proto)) | |
2637 | return NULL; | |
2638 | ||
2639 | struct bgp_channel *bc = (struct bgp_channel *) r->sender; | |
2640 | if (!bc->base_table) | |
2641 | return NULL; | |
2642 | ||
2643 | const net_addr *n = r->net->n.addr; | |
2644 | struct bgp_proto *p = (void *) r->src->proto; | |
2645 | int valid = rt_flowspec_check(bc->base_table, tab, n, r->attrs, p->is_interior); | |
2646 | int dest = valid ? RTD_NONE : RTD_UNREACHABLE; | |
2647 | ||
2648 | if (dest == r->attrs->dest) | |
2649 | return NULL; | |
2650 | ||
2651 | rta *a = alloca(RTA_MAX_SIZE); | |
2652 | memcpy(a, r->attrs, rta_size(r->attrs)); | |
2653 | a->dest = dest; | |
2654 | a->cached = 0; | |
2655 | ||
2656 | rte *new = sl_alloc(rte_slab); | |
2657 | memcpy(new, r, sizeof(rte)); | |
2658 | new->attrs = rta_lookup(a); | |
2659 | rt_lock_source(new->src); | |
2660 | ||
2661 | return new; | |
2662 | #else | |
2663 | return NULL; | |
2664 | #endif | |
2665 | } | |
2666 | ||
2667 | ||
2668 | static inline int | |
2669 | rt_next_hop_update_net(rtable *tab, net *n) | |
2670 | { | |
2671 | rte **k, *e, *new, *old_best, **new_best; | |
2672 | int count = 0; | |
2673 | int free_old_best = 0; | |
2674 | ||
2675 | old_best = n->routes; | |
2676 | if (!old_best) | |
2677 | return 0; | |
2678 | ||
2679 | for (k = &n->routes; e = *k; k = &e->next) | |
2680 | { | |
2681 | if (!net_is_flow(n->n.addr)) | |
2682 | new = rt_next_hop_update_rte(tab, e); | |
2683 | else | |
2684 | new = rt_flowspec_update_rte(tab, e); | |
2685 | ||
2686 | if (new) | |
2687 | { | |
2688 | *k = new; | |
2689 | ||
2690 | rte_trace_in(D_ROUTES, new->sender, new, "updated"); | |
2691 | rte_announce_i(tab, RA_ANY, n, new, e, NULL, NULL); | |
2692 | ||
2693 | /* Call a pre-comparison hook */ | |
2694 | /* Not really an efficient way to compute this */ | |
2695 | if (e->src->proto->rte_recalculate) | |
2696 | e->src->proto->rte_recalculate(tab, n, new, e, NULL); | |
2697 | ||
2698 | if (e != old_best) | |
2699 | rte_free_quick(e); | |
2700 | else /* Freeing of the old best rte is postponed */ | |
2701 | free_old_best = 1; | |
2702 | ||
2703 | e = new; | |
2704 | count++; | |
2705 | } | |
2706 | } | |
2707 | ||
2708 | if (!count) | |
2709 | return 0; | |
2710 | ||
2711 | /* Find the new best route */ | |
2712 | new_best = NULL; | |
2713 | for (k = &n->routes; e = *k; k = &e->next) | |
2714 | { | |
2715 | if (!new_best || rte_better(e, *new_best)) | |
2716 | new_best = k; | |
2717 | } | |
2718 | ||
2719 | /* Relink the new best route to the first position */ | |
2720 | new = *new_best; | |
2721 | if (new != n->routes) | |
2722 | { | |
2723 | *new_best = new->next; | |
2724 | new->next = n->routes; | |
2725 | n->routes = new; | |
2726 | } | |
2727 | ||
2728 | /* Announce the new best route */ | |
2729 | if (new != old_best) | |
2730 | rte_trace_in(D_ROUTES, new->sender, new, "updated [best]"); | |
2731 | ||
2732 | /* Propagate changes */ | |
2733 | rte_announce_i(tab, RA_UNDEF, n, NULL, NULL, n->routes, old_best); | |
2734 | ||
2735 | if (free_old_best) | |
2736 | rte_free_quick(old_best); | |
2737 | ||
2738 | return count; | |
2739 | } | |
2740 | ||
2741 | static void | |
2742 | rt_next_hop_update(rtable *tab) | |
2743 | { | |
2744 | struct fib_iterator *fit = &tab->nhu_fit; | |
2745 | int max_feed = 32; | |
2746 | ||
2747 | if (tab->nhu_state == NHU_CLEAN) | |
2748 | return; | |
2749 | ||
2750 | if (tab->nhu_state == NHU_SCHEDULED) | |
2751 | { | |
2752 | FIB_ITERATE_INIT(fit, &tab->fib); | |
2753 | tab->nhu_state = NHU_RUNNING; | |
2754 | ||
2755 | if (tab->flowspec_trie) | |
2756 | rt_flowspec_reset_trie(tab); | |
2757 | } | |
2758 | ||
2759 | FIB_ITERATE_START(&tab->fib, fit, net, n) | |
2760 | { | |
2761 | if (max_feed <= 0) | |
2762 | { | |
2763 | FIB_ITERATE_PUT(fit); | |
2764 | ev_schedule(tab->rt_event); | |
2765 | return; | |
2766 | } | |
2767 | max_feed -= rt_next_hop_update_net(tab, n); | |
2768 | } | |
2769 | FIB_ITERATE_END; | |
2770 | ||
2771 | /* State change: | |
2772 | * NHU_DIRTY -> NHU_SCHEDULED | |
2773 | * NHU_RUNNING -> NHU_CLEAN | |
2774 | */ | |
2775 | tab->nhu_state &= 1; | |
2776 | ||
2777 | if (tab->nhu_state != NHU_CLEAN) | |
2778 | ev_schedule(tab->rt_event); | |
2779 | } | |
2780 | ||
2781 | ||
2782 | struct rtable_config * | |
2783 | rt_new_table(struct symbol *s, uint addr_type) | |
2784 | { | |
2785 | /* Hack that allows to 'redefine' the master table */ | |
2786 | if ((s->class == SYM_TABLE) && | |
2787 | (s->table == new_config->def_tables[addr_type]) && | |
2788 | ((addr_type == NET_IP4) || (addr_type == NET_IP6))) | |
2789 | return s->table; | |
2790 | ||
2791 | struct rtable_config *c = cfg_allocz(sizeof(struct rtable_config)); | |
2792 | ||
2793 | cf_define_symbol(new_config, s, SYM_TABLE, table, c); | |
2794 | c->name = s->name; | |
2795 | c->addr_type = addr_type; | |
2796 | c->gc_threshold = 1000; | |
2797 | c->gc_period = (uint) -1; /* set in rt_postconfig() */ | |
2798 | c->min_settle_time = 1 S; | |
2799 | c->max_settle_time = 20 S; | |
2800 | c->debug = new_config->table_default_debug; | |
2801 | ||
2802 | add_tail(&new_config->tables, &c->n); | |
2803 | ||
2804 | /* First table of each type is kept as default */ | |
2805 | if (! new_config->def_tables[addr_type]) | |
2806 | new_config->def_tables[addr_type] = c; | |
2807 | ||
2808 | return c; | |
2809 | } | |
2810 | ||
2811 | /** | |
2812 | * rt_lock_table - lock a routing table | |
2813 | * @r: routing table to be locked | |
2814 | * | |
2815 | * Lock a routing table, because it's in use by a protocol, | |
2816 | * preventing it from being freed when it gets undefined in a new | |
2817 | * configuration. | |
2818 | */ | |
2819 | void | |
2820 | rt_lock_table(rtable *r) | |
2821 | { | |
2822 | r->use_count++; | |
2823 | } | |
2824 | ||
2825 | /** | |
2826 | * rt_unlock_table - unlock a routing table | |
2827 | * @r: routing table to be unlocked | |
2828 | * | |
2829 | * Unlock a routing table formerly locked by rt_lock_table(), | |
2830 | * that is decrease its use count and delete it if it's scheduled | |
2831 | * for deletion by configuration changes. | |
2832 | */ | |
2833 | void | |
2834 | rt_unlock_table(rtable *r) | |
2835 | { | |
2836 | if (!--r->use_count && r->deleted) | |
2837 | { | |
2838 | struct config *conf = r->deleted; | |
2839 | ||
2840 | /* Delete the routing table by freeing its pool */ | |
2841 | rt_shutdown(r); | |
2842 | config_del_obstacle(conf); | |
2843 | } | |
2844 | } | |
2845 | ||
2846 | static int | |
2847 | rt_reconfigure(rtable *tab, struct rtable_config *new, struct rtable_config *old) | |
2848 | { | |
2849 | if ((new->addr_type != old->addr_type) || | |
2850 | (new->sorted != old->sorted) || | |
2851 | (new->trie_used != old->trie_used)) | |
2852 | return 0; | |
2853 | ||
2854 | DBG("\t%s: same\n", new->name); | |
2855 | new->table = tab; | |
2856 | tab->name = new->name; | |
2857 | tab->config = new; | |
2858 | tab->debug = new->debug; | |
2859 | ||
2860 | return 1; | |
2861 | } | |
2862 | ||
2863 | static struct rtable_config * | |
2864 | rt_find_table_config(struct config *cf, char *name) | |
2865 | { | |
2866 | struct symbol *sym = cf_find_symbol(cf, name); | |
2867 | return (sym && (sym->class == SYM_TABLE)) ? sym->table : NULL; | |
2868 | } | |
2869 | ||
2870 | /** | |
2871 | * rt_commit - commit new routing table configuration | |
2872 | * @new: new configuration | |
2873 | * @old: original configuration or %NULL if it's boot time config | |
2874 | * | |
2875 | * Scan differences between @old and @new configuration and modify | |
2876 | * the routing tables according to these changes. If @new defines a | |
2877 | * previously unknown table, create it, if it omits a table existing | |
2878 | * in @old, schedule it for deletion (it gets deleted when all protocols | |
2879 | * disconnect from it by calling rt_unlock_table()), if it exists | |
2880 | * in both configurations, leave it unchanged. | |
2881 | */ | |
2882 | void | |
2883 | rt_commit(struct config *new, struct config *old) | |
2884 | { | |
2885 | struct rtable_config *o, *r; | |
2886 | ||
2887 | DBG("rt_commit:\n"); | |
2888 | if (old) | |
2889 | { | |
2890 | WALK_LIST(o, old->tables) | |
2891 | { | |
2892 | rtable *tab = o->table; | |
2893 | if (tab->deleted) | |
2894 | continue; | |
2895 | ||
2896 | r = rt_find_table_config(new, o->name); | |
2897 | if (r && !new->shutdown && rt_reconfigure(tab, r, o)) | |
2898 | continue; | |
2899 | ||
2900 | DBG("\t%s: deleted\n", o->name); | |
2901 | tab->deleted = old; | |
2902 | config_add_obstacle(old); | |
2903 | rt_lock_table(tab); | |
2904 | rt_unlock_table(tab); | |
2905 | } | |
2906 | } | |
2907 | ||
2908 | WALK_LIST(r, new->tables) | |
2909 | if (!r->table) | |
2910 | { | |
2911 | r->table = rt_setup(rt_table_pool, r); | |
2912 | DBG("\t%s: created\n", r->name); | |
2913 | add_tail(&routing_tables, &r->table->n); | |
2914 | } | |
2915 | DBG("\tdone\n"); | |
2916 | } | |
2917 | ||
2918 | static inline void | |
2919 | do_feed_channel(struct channel *c, net *n, rte *e) | |
2920 | { | |
2921 | rte_update_lock(); | |
2922 | if (c->ra_mode == RA_ACCEPTED) | |
2923 | rt_notify_accepted(c, n, NULL, NULL, c->refeeding); | |
2924 | else if (c->ra_mode == RA_MERGED) | |
2925 | rt_notify_merged(c, n, NULL, NULL, e, e, c->refeeding); | |
2926 | else /* RA_BASIC */ | |
2927 | rt_notify_basic(c, n, e, e, c->refeeding); | |
2928 | rte_update_unlock(); | |
2929 | } | |
2930 | ||
2931 | /** | |
2932 | * rt_feed_channel - advertise all routes to a channel | |
2933 | * @c: channel to be fed | |
2934 | * | |
2935 | * This function performs one pass of advertisement of routes to a channel that | |
2936 | * is in the ES_FEEDING state. It is called by the protocol code as long as it | |
2937 | * has something to do. (We avoid transferring all the routes in single pass in | |
2938 | * order not to monopolize CPU time.) | |
2939 | */ | |
2940 | int | |
2941 | rt_feed_channel(struct channel *c) | |
2942 | { | |
2943 | struct fib_iterator *fit = &c->feed_fit; | |
2944 | int max_feed = 256; | |
2945 | ||
2946 | ASSERT(c->export_state == ES_FEEDING); | |
2947 | ||
2948 | if (!c->feed_active) | |
2949 | { | |
2950 | FIB_ITERATE_INIT(fit, &c->table->fib); | |
2951 | c->feed_active = 1; | |
2952 | } | |
2953 | ||
2954 | FIB_ITERATE_START(&c->table->fib, fit, net, n) | |
2955 | { | |
2956 | rte *e = n->routes; | |
2957 | if (max_feed <= 0) | |
2958 | { | |
2959 | FIB_ITERATE_PUT(fit); | |
2960 | return 0; | |
2961 | } | |
2962 | ||
2963 | if ((c->ra_mode == RA_OPTIMAL) || | |
2964 | (c->ra_mode == RA_ACCEPTED) || | |
2965 | (c->ra_mode == RA_MERGED)) | |
2966 | if (rte_is_valid(e)) | |
2967 | { | |
2968 | /* In the meantime, the protocol may fell down */ | |
2969 | if (c->export_state != ES_FEEDING) | |
2970 | goto done; | |
2971 | ||
2972 | do_feed_channel(c, n, e); | |
2973 | max_feed--; | |
2974 | } | |
2975 | ||
2976 | if (c->ra_mode == RA_ANY) | |
2977 | for(e = n->routes; e; e = e->next) | |
2978 | { | |
2979 | /* In the meantime, the protocol may fell down */ | |
2980 | if (c->export_state != ES_FEEDING) | |
2981 | goto done; | |
2982 | ||
2983 | if (!rte_is_valid(e)) | |
2984 | continue; | |
2985 | ||
2986 | do_feed_channel(c, n, e); | |
2987 | max_feed--; | |
2988 | } | |
2989 | } | |
2990 | FIB_ITERATE_END; | |
2991 | ||
2992 | done: | |
2993 | c->feed_active = 0; | |
2994 | return 1; | |
2995 | } | |
2996 | ||
2997 | /** | |
2998 | * rt_feed_baby_abort - abort protocol feeding | |
2999 | * @c: channel | |
3000 | * | |
3001 | * This function is called by the protocol code when the protocol stops or | |
3002 | * ceases to exist during the feeding. | |
3003 | */ | |
3004 | void | |
3005 | rt_feed_channel_abort(struct channel *c) | |
3006 | { | |
3007 | if (c->feed_active) | |
3008 | { | |
3009 | /* Unlink the iterator */ | |
3010 | fit_get(&c->table->fib, &c->feed_fit); | |
3011 | c->feed_active = 0; | |
3012 | } | |
3013 | } | |
3014 | ||
3015 | ||
3016 | /* | |
3017 | * Import table | |
3018 | */ | |
3019 | ||
3020 | int | |
3021 | rte_update_in(struct channel *c, const net_addr *n, rte *new, struct rte_src *src) | |
3022 | { | |
3023 | struct rtable *tab = c->in_table; | |
3024 | rte *old, **pos; | |
3025 | net *net; | |
3026 | ||
3027 | if (new) | |
3028 | { | |
3029 | net = net_get(tab, n); | |
3030 | ||
3031 | if (!rta_is_cached(new->attrs)) | |
3032 | new->attrs = rta_lookup(new->attrs); | |
3033 | } | |
3034 | else | |
3035 | { | |
3036 | net = net_find(tab, n); | |
3037 | ||
3038 | if (!net) | |
3039 | goto drop_withdraw; | |
3040 | } | |
3041 | ||
3042 | /* Find the old rte */ | |
3043 | for (pos = &net->routes; old = *pos; pos = &old->next) | |
3044 | if (old->src == src) | |
3045 | { | |
3046 | if (new && rte_same(old, new)) | |
3047 | { | |
3048 | /* Refresh the old rte, continue with update to main rtable */ | |
3049 | if (old->flags & (REF_STALE | REF_DISCARD | REF_MODIFY)) | |
3050 | { | |
3051 | old->flags &= ~(REF_STALE | REF_DISCARD | REF_MODIFY); | |
3052 | ||
3053 | return 1; | |
3054 | } | |
3055 | ||
3056 | goto drop_update; | |
3057 | } | |
3058 | ||
3059 | /* Move iterator if needed */ | |
3060 | if (old == c->reload_next_rte) | |
3061 | c->reload_next_rte = old->next; | |
3062 | ||
3063 | /* Remove the old rte */ | |
3064 | *pos = old->next; | |
3065 | tab->rt_count--; | |
3066 | break; | |
3067 | } | |
3068 | ||
3069 | if (!old && !new) | |
3070 | goto drop_withdraw; | |
3071 | ||
3072 | struct channel_limit *l = &c->rx_limit; | |
3073 | if (l->action && !old && new) | |
3074 | { | |
3075 | if (tab->rt_count >= l->limit) | |
3076 | channel_notify_limit(c, l, PLD_RX, tab->rt_count); | |
3077 | ||
3078 | if (l->state == PLS_BLOCKED) | |
3079 | { | |
3080 | /* Required by rte_trace_in() */ | |
3081 | new->net = net; | |
3082 | ||
3083 | rte_trace_in(D_FILTERS, c, new, "ignored [limit]"); | |
3084 | goto drop_update; | |
3085 | } | |
3086 | } | |
3087 | ||
3088 | if (new) | |
3089 | { | |
3090 | /* Insert the new rte */ | |
3091 | rte *e = rte_do_cow(new); | |
3092 | e->flags |= REF_COW; | |
3093 | e->net = net; | |
3094 | e->sender = c; | |
3095 | e->lastmod = current_time(); | |
3096 | e->next = *pos; | |
3097 | *pos = new = e; | |
3098 | tab->rt_count++; | |
3099 | ||
3100 | if (!old) | |
3101 | { | |
3102 | new->id = hmap_first_zero(&tab->id_map); | |
3103 | hmap_set(&tab->id_map, new->id); | |
3104 | } | |
3105 | else | |
3106 | new->id = old->id; | |
3107 | } | |
3108 | ||
3109 | rte_announce(tab, RA_ANY, net, new, old, NULL, NULL); | |
3110 | ||
3111 | if (old) | |
3112 | { | |
3113 | if (!new) | |
3114 | hmap_clear(&tab->id_map, old->id); | |
3115 | ||
3116 | rte_free_quick(old); | |
3117 | } | |
3118 | ||
3119 | if (!net->routes) | |
3120 | fib_delete(&tab->fib, net); | |
3121 | ||
3122 | return 1; | |
3123 | ||
3124 | drop_update: | |
3125 | c->stats.imp_updates_received++; | |
3126 | c->stats.imp_updates_ignored++; | |
3127 | rte_free(new); | |
3128 | ||
3129 | if (!net->routes) | |
3130 | fib_delete(&tab->fib, net); | |
3131 | ||
3132 | return 0; | |
3133 | ||
3134 | drop_withdraw: | |
3135 | c->stats.imp_withdraws_received++; | |
3136 | c->stats.imp_withdraws_ignored++; | |
3137 | return 0; | |
3138 | } | |
3139 | ||
3140 | int | |
3141 | rt_reload_channel(struct channel *c) | |
3142 | { | |
3143 | struct rtable *tab = c->in_table; | |
3144 | struct fib_iterator *fit = &c->reload_fit; | |
3145 | int max_feed = 64; | |
3146 | ||
3147 | ASSERT(c->channel_state == CS_UP); | |
3148 | ||
3149 | if (!c->reload_active) | |
3150 | { | |
3151 | FIB_ITERATE_INIT(fit, &tab->fib); | |
3152 | c->reload_active = 1; | |
3153 | } | |
3154 | ||
3155 | do { | |
3156 | for (rte *e = c->reload_next_rte; e; e = e->next) | |
3157 | { | |
3158 | if (max_feed-- <= 0) | |
3159 | { | |
3160 | c->reload_next_rte = e; | |
3161 | debug("%s channel reload burst split (max_feed=%d)", c->proto->name, max_feed); | |
3162 | return 0; | |
3163 | } | |
3164 | ||
3165 | rte_update2(c, e->net->n.addr, rte_do_cow(e), e->src); | |
3166 | } | |
3167 | ||
3168 | c->reload_next_rte = NULL; | |
3169 | ||
3170 | FIB_ITERATE_START(&tab->fib, fit, net, n) | |
3171 | { | |
3172 | if (c->reload_next_rte = n->routes) | |
3173 | { | |
3174 | FIB_ITERATE_PUT_NEXT(fit, &tab->fib); | |
3175 | break; | |
3176 | } | |
3177 | } | |
3178 | FIB_ITERATE_END; | |
3179 | } | |
3180 | while (c->reload_next_rte); | |
3181 | ||
3182 | c->reload_active = 0; | |
3183 | return 1; | |
3184 | } | |
3185 | ||
3186 | void | |
3187 | rt_reload_channel_abort(struct channel *c) | |
3188 | { | |
3189 | if (c->reload_active) | |
3190 | { | |
3191 | /* Unlink the iterator */ | |
3192 | fit_get(&c->in_table->fib, &c->reload_fit); | |
3193 | c->reload_next_rte = NULL; | |
3194 | c->reload_active = 0; | |
3195 | } | |
3196 | } | |
3197 | ||
3198 | void | |
3199 | rt_prune_sync(rtable *t, int all) | |
3200 | { | |
3201 | struct fib_iterator fit; | |
3202 | ||
3203 | FIB_ITERATE_INIT(&fit, &t->fib); | |
3204 | ||
3205 | again: | |
3206 | FIB_ITERATE_START(&t->fib, &fit, net, n) | |
3207 | { | |
3208 | rte *e, **ee = &n->routes; | |
3209 | ||
3210 | while (e = *ee) | |
3211 | { | |
3212 | if (all || (e->flags & (REF_STALE | REF_DISCARD))) | |
3213 | { | |
3214 | *ee = e->next; | |
3215 | rte_free_quick(e); | |
3216 | t->rt_count--; | |
3217 | } | |
3218 | else | |
3219 | ee = &e->next; | |
3220 | } | |
3221 | ||
3222 | if (all || !n->routes) | |
3223 | { | |
3224 | FIB_ITERATE_PUT(&fit); | |
3225 | fib_delete(&t->fib, n); | |
3226 | goto again; | |
3227 | } | |
3228 | } | |
3229 | FIB_ITERATE_END; | |
3230 | } | |
3231 | ||
3232 | ||
3233 | /* | |
3234 | * Export table | |
3235 | */ | |
3236 | ||
3237 | int | |
3238 | rte_update_out(struct channel *c, const net_addr *n, rte *new, rte *old0, int refeed) | |
3239 | { | |
3240 | struct rtable *tab = c->out_table; | |
3241 | struct rte_src *src; | |
3242 | rte *old, **pos; | |
3243 | net *net; | |
3244 | ||
3245 | if (new) | |
3246 | { | |
3247 | net = net_get(tab, n); | |
3248 | src = new->src; | |
3249 | ||
3250 | if (!rta_is_cached(new->attrs)) | |
3251 | new->attrs = rta_lookup(new->attrs); | |
3252 | } | |
3253 | else | |
3254 | { | |
3255 | net = net_find(tab, n); | |
3256 | src = old0->src; | |
3257 | ||
3258 | if (!net) | |
3259 | goto drop_withdraw; | |
3260 | } | |
3261 | ||
3262 | /* Find the old rte */ | |
3263 | for (pos = &net->routes; old = *pos; pos = &old->next) | |
3264 | if ((c->ra_mode != RA_ANY) || (old->src == src)) | |
3265 | { | |
3266 | if (new && rte_same(old, new)) | |
3267 | { | |
3268 | /* REF_STALE / REF_DISCARD not used in export table */ | |
3269 | /* | |
3270 | if (old->flags & (REF_STALE | REF_DISCARD | REF_MODIFY)) | |
3271 | { | |
3272 | old->flags &= ~(REF_STALE | REF_DISCARD | REF_MODIFY); | |
3273 | return 1; | |
3274 | } | |
3275 | */ | |
3276 | ||
3277 | goto drop_update; | |
3278 | } | |
3279 | ||
3280 | /* Remove the old rte */ | |
3281 | *pos = old->next; | |
3282 | rte_free_quick(old); | |
3283 | tab->rt_count--; | |
3284 | ||
3285 | break; | |
3286 | } | |
3287 | ||
3288 | if (!new) | |
3289 | { | |
3290 | if (!old) | |
3291 | goto drop_withdraw; | |
3292 | ||
3293 | if (!net->routes) | |
3294 | fib_delete(&tab->fib, net); | |
3295 | ||
3296 | return 1; | |
3297 | } | |
3298 | ||
3299 | /* Insert the new rte */ | |
3300 | rte *e = rte_do_cow(new); | |
3301 | e->flags |= REF_COW; | |
3302 | e->net = net; | |
3303 | e->sender = c; | |
3304 | e->lastmod = current_time(); | |
3305 | e->next = *pos; | |
3306 | *pos = e; | |
3307 | tab->rt_count++; | |
3308 | return 1; | |
3309 | ||
3310 | drop_update: | |
3311 | return refeed; | |
3312 | ||
3313 | drop_withdraw: | |
3314 | return 0; | |
3315 | } | |
3316 | ||
3317 | ||
3318 | /* | |
3319 | * Hostcache | |
3320 | */ | |
3321 | ||
3322 | static inline u32 | |
3323 | hc_hash(ip_addr a, rtable *dep) | |
3324 | { | |
3325 | return ipa_hash(a) ^ ptr_hash(dep); | |
3326 | } | |
3327 | ||
3328 | static inline void | |
3329 | hc_insert(struct hostcache *hc, struct hostentry *he) | |
3330 | { | |
3331 | uint k = he->hash_key >> hc->hash_shift; | |
3332 | he->next = hc->hash_table[k]; | |
3333 | hc->hash_table[k] = he; | |
3334 | } | |
3335 | ||
3336 | static inline void | |
3337 | hc_remove(struct hostcache *hc, struct hostentry *he) | |
3338 | { | |
3339 | struct hostentry **hep; | |
3340 | uint k = he->hash_key >> hc->hash_shift; | |
3341 | ||
3342 | for (hep = &hc->hash_table[k]; *hep != he; hep = &(*hep)->next); | |
3343 | *hep = he->next; | |
3344 | } | |
3345 | ||
3346 | #define HC_DEF_ORDER 10 | |
3347 | #define HC_HI_MARK *4 | |
3348 | #define HC_HI_STEP 2 | |
3349 | #define HC_HI_ORDER 16 /* Must be at most 16 */ | |
3350 | #define HC_LO_MARK /5 | |
3351 | #define HC_LO_STEP 2 | |
3352 | #define HC_LO_ORDER 10 | |
3353 | ||
3354 | static void | |
3355 | hc_alloc_table(struct hostcache *hc, pool *p, unsigned order) | |
3356 | { | |
3357 | uint hsize = 1 << order; | |
3358 | hc->hash_order = order; | |
3359 | hc->hash_shift = 32 - order; | |
3360 | hc->hash_max = (order >= HC_HI_ORDER) ? ~0U : (hsize HC_HI_MARK); | |
3361 | hc->hash_min = (order <= HC_LO_ORDER) ? 0U : (hsize HC_LO_MARK); | |
3362 | ||
3363 | hc->hash_table = mb_allocz(p, hsize * sizeof(struct hostentry *)); | |
3364 | } | |
3365 | ||
3366 | static void | |
3367 | hc_resize(struct hostcache *hc, pool *p, unsigned new_order) | |
3368 | { | |
3369 | struct hostentry **old_table = hc->hash_table; | |
3370 | struct hostentry *he, *hen; | |
3371 | uint old_size = 1 << hc->hash_order; | |
3372 | uint i; | |
3373 | ||
3374 | hc_alloc_table(hc, p, new_order); | |
3375 | for (i = 0; i < old_size; i++) | |
3376 | for (he = old_table[i]; he != NULL; he=hen) | |
3377 | { | |
3378 | hen = he->next; | |
3379 | hc_insert(hc, he); | |
3380 | } | |
3381 | mb_free(old_table); | |
3382 | } | |
3383 | ||
3384 | static struct hostentry * | |
3385 | hc_new_hostentry(struct hostcache *hc, pool *p, ip_addr a, ip_addr ll, rtable *dep, unsigned k) | |
3386 | { | |
3387 | struct hostentry *he = sl_alloc(hc->slab); | |
3388 | ||
3389 | *he = (struct hostentry) { | |
3390 | .addr = a, | |
3391 | .link = ll, | |
3392 | .tab = dep, | |
3393 | .hash_key = k, | |
3394 | }; | |
3395 | ||
3396 | add_tail(&hc->hostentries, &he->ln); | |
3397 | hc_insert(hc, he); | |
3398 | ||
3399 | hc->hash_items++; | |
3400 | if (hc->hash_items > hc->hash_max) | |
3401 | hc_resize(hc, p, hc->hash_order + HC_HI_STEP); | |
3402 | ||
3403 | return he; | |
3404 | } | |
3405 | ||
3406 | static void | |
3407 | hc_delete_hostentry(struct hostcache *hc, pool *p, struct hostentry *he) | |
3408 | { | |
3409 | rta_free(he->src); | |
3410 | ||
3411 | rem_node(&he->ln); | |
3412 | hc_remove(hc, he); | |
3413 | sl_free(he); | |
3414 | ||
3415 | hc->hash_items--; | |
3416 | if (hc->hash_items < hc->hash_min) | |
3417 | hc_resize(hc, p, hc->hash_order - HC_LO_STEP); | |
3418 | } | |
3419 | ||
3420 | static void | |
3421 | rt_init_hostcache(rtable *tab) | |
3422 | { | |
3423 | struct hostcache *hc = mb_allocz(tab->rp, sizeof(struct hostcache)); | |
3424 | init_list(&hc->hostentries); | |
3425 | ||
3426 | hc->hash_items = 0; | |
3427 | hc_alloc_table(hc, tab->rp, HC_DEF_ORDER); | |
3428 | hc->slab = sl_new(tab->rp, sizeof(struct hostentry)); | |
3429 | ||
3430 | hc->lp = lp_new(tab->rp); | |
3431 | hc->trie = f_new_trie(hc->lp, 0); | |
3432 | ||
3433 | tab->hostcache = hc; | |
3434 | } | |
3435 | ||
3436 | static void | |
3437 | rt_free_hostcache(rtable *tab) | |
3438 | { | |
3439 | struct hostcache *hc = tab->hostcache; | |
3440 | ||
3441 | node *n; | |
3442 | WALK_LIST(n, hc->hostentries) | |
3443 | { | |
3444 | struct hostentry *he = SKIP_BACK(struct hostentry, ln, n); | |
3445 | rta_free(he->src); | |
3446 | ||
3447 | if (he->uc) | |
3448 | log(L_ERR "Hostcache is not empty in table %s", tab->name); | |
3449 | } | |
3450 | ||
3451 | /* Freed automagically by the resource pool | |
3452 | rfree(hc->slab); | |
3453 | rfree(hc->lp); | |
3454 | mb_free(hc->hash_table); | |
3455 | mb_free(hc); | |
3456 | */ | |
3457 | } | |
3458 | ||
3459 | static void | |
3460 | rt_notify_hostcache(rtable *tab, net *net) | |
3461 | { | |
3462 | if (tab->hcu_scheduled) | |
3463 | return; | |
3464 | ||
3465 | if (trie_match_net(tab->hostcache->trie, net->n.addr)) | |
3466 | rt_schedule_hcu(tab); | |
3467 | } | |
3468 | ||
3469 | static int | |
3470 | if_local_addr(ip_addr a, struct iface *i) | |
3471 | { | |
3472 | struct ifa *b; | |
3473 | ||
3474 | WALK_LIST(b, i->addrs) | |
3475 | if (ipa_equal(a, b->ip)) | |
3476 | return 1; | |
3477 | ||
3478 | return 0; | |
3479 | } | |
3480 | ||
3481 | u32 | |
3482 | rt_get_igp_metric(rte *rt) | |
3483 | { | |
3484 | eattr *ea = ea_find(rt->attrs->eattrs, EA_GEN_IGP_METRIC); | |
3485 | ||
3486 | if (ea) | |
3487 | return ea->u.data; | |
3488 | ||
3489 | if (rt->attrs->source == RTS_DEVICE) | |
3490 | return 0; | |
3491 | ||
3492 | if (rt->src->proto->rte_igp_metric) | |
3493 | return rt->src->proto->rte_igp_metric(rt); | |
3494 | ||
3495 | return IGP_METRIC_UNKNOWN; | |
3496 | } | |
3497 | ||
3498 | static int | |
3499 | rt_update_hostentry(rtable *tab, struct hostentry *he) | |
3500 | { | |
3501 | rta *old_src = he->src; | |
3502 | int direct = 0; | |
3503 | int pxlen = 0; | |
3504 | ||
3505 | /* Reset the hostentry */ | |
3506 | he->src = NULL; | |
3507 | he->dest = RTD_UNREACHABLE; | |
3508 | he->nexthop_linkable = 0; | |
3509 | he->igp_metric = 0; | |
3510 | ||
3511 | net_addr he_addr; | |
3512 | net_fill_ip_host(&he_addr, he->addr); | |
3513 | net *n = net_route(tab, &he_addr); | |
3514 | if (n) | |
3515 | { | |
3516 | rte *e = n->routes; | |
3517 | rta *a = e->attrs; | |
3518 | word pref = a->pref; | |
3519 | ||
3520 | for (rte *ee = n->routes; ee; ee = ee->next) | |
3521 | if ((ee->attrs->pref >= pref) && ee->attrs->hostentry) | |
3522 | { | |
3523 | /* Recursive route should not depend on another recursive route */ | |
3524 | log(L_WARN "Next hop address %I resolvable through recursive route for %N", | |
3525 | he->addr, n->n.addr); | |
3526 | goto done; | |
3527 | } | |
3528 | ||
3529 | pxlen = n->n.addr->pxlen; | |
3530 | ||
3531 | if (a->dest == RTD_UNICAST) | |
3532 | { | |
3533 | for (struct nexthop *nh = &(a->nh); nh; nh = nh->next) | |
3534 | if (ipa_zero(nh->gw)) | |
3535 | { | |
3536 | if (if_local_addr(he->addr, nh->iface)) | |
3537 | { | |
3538 | /* The host address is a local address, this is not valid */ | |
3539 | log(L_WARN "Next hop address %I is a local address of iface %s", | |
3540 | he->addr, nh->iface->name); | |
3541 | goto done; | |
3542 | } | |
3543 | ||
3544 | direct++; | |
3545 | } | |
3546 | } | |
3547 | ||
3548 | he->src = rta_clone(a); | |
3549 | he->dest = a->dest; | |
3550 | he->nexthop_linkable = !direct; | |
3551 | he->igp_metric = rt_get_igp_metric(e); | |
3552 | } | |
3553 | ||
3554 | done: | |
3555 | /* Add a prefix range to the trie */ | |
3556 | trie_add_prefix(tab->hostcache->trie, &he_addr, pxlen, he_addr.pxlen); | |
3557 | ||
3558 | rta_free(old_src); | |
3559 | return old_src != he->src; | |
3560 | } | |
3561 | ||
3562 | static void | |
3563 | rt_update_hostcache(rtable *tab) | |
3564 | { | |
3565 | struct hostcache *hc = tab->hostcache; | |
3566 | struct hostentry *he; | |
3567 | node *n, *x; | |
3568 | ||
3569 | /* Reset the trie */ | |
3570 | lp_flush(hc->lp); | |
3571 | hc->trie = f_new_trie(hc->lp, 0); | |
3572 | ||
3573 | WALK_LIST_DELSAFE(n, x, hc->hostentries) | |
3574 | { | |
3575 | he = SKIP_BACK(struct hostentry, ln, n); | |
3576 | if (!he->uc) | |
3577 | { | |
3578 | hc_delete_hostentry(hc, tab->rp, he); | |
3579 | continue; | |
3580 | } | |
3581 | ||
3582 | if (rt_update_hostentry(tab, he)) | |
3583 | rt_schedule_nhu(he->tab); | |
3584 | } | |
3585 | ||
3586 | tab->hcu_scheduled = 0; | |
3587 | } | |
3588 | ||
3589 | struct hostentry * | |
3590 | rt_get_hostentry(rtable *tab, ip_addr a, ip_addr ll, rtable *dep) | |
3591 | { | |
3592 | ip_addr link = ipa_zero(ll) ? a : ll; | |
3593 | struct hostentry *he; | |
3594 | ||
3595 | if (!tab->hostcache) | |
3596 | rt_init_hostcache(tab); | |
3597 | ||
3598 | u32 k = hc_hash(a, dep); | |
3599 | struct hostcache *hc = tab->hostcache; | |
3600 | for (he = hc->hash_table[k >> hc->hash_shift]; he != NULL; he = he->next) | |
3601 | if (ipa_equal(he->addr, a) && ipa_equal(he->link, link) && (he->tab == dep)) | |
3602 | return he; | |
3603 | ||
3604 | he = hc_new_hostentry(hc, tab->rp, a, link, dep, k); | |
3605 | he->owner = tab; | |
3606 | rt_update_hostentry(tab, he); | |
3607 | return he; | |
3608 | } | |
3609 | ||
3610 | ||
3611 | /* | |
3612 | * Documentation for functions declared inline in route.h | |
3613 | */ | |
3614 | #if 0 | |
3615 | ||
3616 | /** | |
3617 | * net_find - find a network entry | |
3618 | * @tab: a routing table | |
3619 | * @addr: address of the network | |
3620 | * | |
3621 | * net_find() looks up the given network in routing table @tab and | |
3622 | * returns a pointer to its &net entry or %NULL if no such network | |
3623 | * exists. | |
3624 | */ | |
3625 | static inline net *net_find(rtable *tab, net_addr *addr) | |
3626 | { DUMMY; } | |
3627 | ||
3628 | /** | |
3629 | * net_get - obtain a network entry | |
3630 | * @tab: a routing table | |
3631 | * @addr: address of the network | |
3632 | * | |
3633 | * net_get() looks up the given network in routing table @tab and | |
3634 | * returns a pointer to its &net entry. If no such entry exists, it's | |
3635 | * created. | |
3636 | */ | |
3637 | static inline net *net_get(rtable *tab, net_addr *addr) | |
3638 | { DUMMY; } | |
3639 | ||
3640 | /** | |
3641 | * rte_cow - copy a route for writing | |
3642 | * @r: a route entry to be copied | |
3643 | * | |
3644 | * rte_cow() takes a &rte and prepares it for modification. The exact action | |
3645 | * taken depends on the flags of the &rte -- if it's a temporary entry, it's | |
3646 | * just returned unchanged, else a new temporary entry with the same contents | |
3647 | * is created. | |
3648 | * | |
3649 | * The primary use of this function is inside the filter machinery -- when | |
3650 | * a filter wants to modify &rte contents (to change the preference or to | |
3651 | * attach another set of attributes), it must ensure that the &rte is not | |
3652 | * shared with anyone else (and especially that it isn't stored in any routing | |
3653 | * table). | |
3654 | * | |
3655 | * Result: a pointer to the new writable &rte. | |
3656 | */ | |
3657 | static inline rte * rte_cow(rte *r) | |
3658 | { DUMMY; } | |
3659 | ||
3660 | #endif |