Annotation of embedaddon/bird/nest/rt-table.c, revision 1.1.1.2
1.1 misho 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 specific to the route (preference, protocol
25: * metrics, time of last modification etc.) and a pointer to a &rta structure
26: * (see the route attribute module for a precise explanation) holding the
27: * remaining route attributes which are expected to be shared by multiple
28: * routes in order to conserve memory.
29: */
30:
31: #undef LOCAL_DEBUG
32:
33: #include "nest/bird.h"
34: #include "nest/route.h"
35: #include "nest/protocol.h"
36: #include "nest/cli.h"
37: #include "nest/iface.h"
38: #include "lib/resource.h"
39: #include "lib/event.h"
40: #include "lib/string.h"
41: #include "conf/conf.h"
42: #include "filter/filter.h"
43: #include "lib/string.h"
44: #include "lib/alloca.h"
45:
46: pool *rt_table_pool;
47:
48: static slab *rte_slab;
49: static linpool *rte_update_pool;
50:
1.1.1.2 ! misho 51: list routing_tables;
1.1 misho 52:
53: static byte *rt_format_via(rte *e);
54: static void rt_free_hostcache(rtable *tab);
55: static void rt_notify_hostcache(rtable *tab, net *net);
56: static void rt_update_hostcache(rtable *tab);
57: static void rt_next_hop_update(rtable *tab);
58: static inline int rt_prune_table(rtable *tab);
59: static inline void rt_schedule_gc(rtable *tab);
60: static inline void rt_schedule_prune(rtable *tab);
61:
62:
63: /* Like fib_route(), but skips empty net entries */
64: static net *
65: net_route(rtable *tab, ip_addr a, int len)
66: {
67: ip_addr a0;
68: net *n;
69:
70: while (len >= 0)
71: {
72: a0 = ipa_and(a, ipa_mkmask(len));
73: n = fib_find(&tab->fib, &a0, len);
74: if (n && rte_is_valid(n->routes))
75: return n;
76: len--;
77: }
78: return NULL;
79: }
80:
81: static void
82: rte_init(struct fib_node *N)
83: {
84: net *n = (net *) N;
85:
86: N->flags = 0;
87: n->routes = NULL;
88: }
89:
90: /**
91: * rte_find - find a route
92: * @net: network node
93: * @src: route source
94: *
95: * The rte_find() function returns a route for destination @net
96: * which is from route source @src.
97: */
98: rte *
99: rte_find(net *net, struct rte_src *src)
100: {
101: rte *e = net->routes;
102:
103: while (e && e->attrs->src != src)
104: e = e->next;
105: return e;
106: }
107:
108: /**
109: * rte_get_temp - get a temporary &rte
110: * @a: attributes to assign to the new route (a &rta; in case it's
111: * un-cached, rte_update() will create a cached copy automatically)
112: *
113: * Create a temporary &rte and bind it with the attributes @a.
114: * Also set route preference to the default preference set for
115: * the protocol.
116: */
117: rte *
118: rte_get_temp(rta *a)
119: {
120: rte *e = sl_alloc(rte_slab);
121:
122: e->attrs = a;
123: e->flags = 0;
124: e->pref = a->src->proto->preference;
125: return e;
126: }
127:
128: rte *
129: rte_do_cow(rte *r)
130: {
131: rte *e = sl_alloc(rte_slab);
132:
133: memcpy(e, r, sizeof(rte));
134: e->attrs = rta_clone(r->attrs);
135: e->flags = 0;
136: return e;
137: }
138:
139: /**
140: * rte_cow_rta - get a private writable copy of &rte with writable &rta
141: * @r: a route entry to be copied
142: * @lp: a linpool from which to allocate &rta
143: *
144: * rte_cow_rta() takes a &rte and prepares it and associated &rta for
145: * modification. There are three possibilities: First, both &rte and &rta are
146: * private copies, in that case they are returned unchanged. Second, &rte is
147: * private copy, but &rta is cached, in that case &rta is duplicated using
148: * rta_do_cow(). Third, both &rte is shared and &rta is cached, in that case
149: * both structures are duplicated by rte_do_cow() and rta_do_cow().
150: *
151: * Note that in the second case, cached &rta loses one reference, while private
152: * copy created by rta_do_cow() is a shallow copy sharing indirect data (eattrs,
153: * nexthops, ...) with it. To work properly, original shared &rta should have
154: * another reference during the life of created private copy.
155: *
156: * Result: a pointer to the new writable &rte with writable &rta.
157: */
158: rte *
159: rte_cow_rta(rte *r, linpool *lp)
160: {
161: if (!rta_is_cached(r->attrs))
162: return r;
163:
164: rte *e = rte_cow(r);
165: rta *a = rta_do_cow(r->attrs, lp);
166: rta_free(e->attrs);
167: e->attrs = a;
168: return e;
169: }
170:
171: static int /* Actually better or at least as good as */
172: rte_better(rte *new, rte *old)
173: {
174: int (*better)(rte *, rte *);
175:
176: if (!rte_is_valid(old))
177: return 1;
178: if (!rte_is_valid(new))
179: return 0;
180:
181: if (new->pref > old->pref)
182: return 1;
183: if (new->pref < old->pref)
184: return 0;
185: if (new->attrs->src->proto->proto != old->attrs->src->proto->proto)
186: {
187: /*
188: * If the user has configured protocol preferences, so that two different protocols
189: * have the same preference, try to break the tie by comparing addresses. Not too
190: * useful, but keeps the ordering of routes unambiguous.
191: */
192: return new->attrs->src->proto->proto > old->attrs->src->proto->proto;
193: }
194: if (better = new->attrs->src->proto->rte_better)
195: return better(new, old);
196: return 0;
197: }
198:
199: static int
200: rte_mergable(rte *pri, rte *sec)
201: {
202: int (*mergable)(rte *, rte *);
203:
204: if (!rte_is_valid(pri) || !rte_is_valid(sec))
205: return 0;
206:
207: if (pri->pref != sec->pref)
208: return 0;
209:
210: if (pri->attrs->src->proto->proto != sec->attrs->src->proto->proto)
211: return 0;
212:
213: if (mergable = pri->attrs->src->proto->rte_mergable)
214: return mergable(pri, sec);
215:
216: return 0;
217: }
218:
219: static void
220: rte_trace(struct proto *p, rte *e, int dir, char *msg)
221: {
222: log(L_TRACE "%s %c %s %I/%d %s", p->name, dir, msg, e->net->n.prefix, e->net->n.pxlen, rt_format_via(e));
223: }
224:
225: static inline void
226: rte_trace_in(uint flag, struct proto *p, rte *e, char *msg)
227: {
228: if (p->debug & flag)
229: rte_trace(p, e, '>', msg);
230: }
231:
232: static inline void
233: rte_trace_out(uint flag, struct proto *p, rte *e, char *msg)
234: {
235: if (p->debug & flag)
236: rte_trace(p, e, '<', msg);
237: }
238:
239: static rte *
240: export_filter_(struct announce_hook *ah, rte *rt0, rte **rt_free, ea_list **tmpa, linpool *pool, int silent)
241: {
242: struct proto *p = ah->proto;
243: struct filter *filter = ah->out_filter;
244: struct proto_stats *stats = ah->stats;
245: ea_list *tmpb = NULL;
246: rte *rt;
247: int v;
248:
249: rt = rt0;
250: *rt_free = NULL;
251:
252: if (!tmpa)
253: tmpa = &tmpb;
254:
1.1.1.2 ! misho 255: *tmpa = rte_make_tmp_attrs(rt, pool);
1.1 misho 256:
257: v = p->import_control ? p->import_control(p, &rt, tmpa, pool) : 0;
258: if (v < 0)
259: {
260: if (silent)
261: goto reject;
262:
263: stats->exp_updates_rejected++;
264: if (v == RIC_REJECT)
265: rte_trace_out(D_FILTERS, p, rt, "rejected by protocol");
266: goto reject;
267: }
268: if (v > 0)
269: {
270: if (!silent)
271: rte_trace_out(D_FILTERS, p, rt, "forced accept by protocol");
272: goto accept;
273: }
274:
275: v = filter && ((filter == FILTER_REJECT) ||
1.1.1.2 ! misho 276: (f_run(filter, &rt, tmpa, pool,
! 277: FF_FORCE_TMPATTR | (silent ? FF_SILENT : 0)) > F_ACCEPT));
1.1 misho 278: if (v)
279: {
280: if (silent)
281: goto reject;
282:
283: stats->exp_updates_filtered++;
284: rte_trace_out(D_FILTERS, p, rt, "filtered out");
285: goto reject;
286: }
287:
288: accept:
289: if (rt != rt0)
290: *rt_free = rt;
291: return rt;
292:
293: reject:
294: /* Discard temporary rte */
295: if (rt != rt0)
296: rte_free(rt);
297: return NULL;
298: }
299:
300: static inline rte *
301: export_filter(struct announce_hook *ah, rte *rt0, rte **rt_free, ea_list **tmpa, int silent)
302: {
303: return export_filter_(ah, rt0, rt_free, tmpa, rte_update_pool, silent);
304: }
305:
306: static void
307: do_rt_notify(struct announce_hook *ah, net *net, rte *new, rte *old, ea_list *tmpa, int refeed)
308: {
309: struct proto *p = ah->proto;
310: struct proto_stats *stats = ah->stats;
311:
312:
313: /*
314: * First, apply export limit.
315: *
316: * Export route limits has several problems. Because exp_routes
317: * counter is reset before refeed, we don't really know whether
318: * limit is breached and whether the update is new or not. Therefore
319: * the number of really exported routes may exceed the limit
320: * temporarily (routes exported before and new routes in refeed).
321: *
322: * Minor advantage is that if the limit is decreased and refeed is
323: * requested, the number of exported routes really decrease.
324: *
325: * Second problem is that with export limits, we don't know whether
326: * old was really exported (it might be blocked by limit). When a
327: * withdraw is exported, we announce it even when the previous
328: * update was blocked. This is not a big issue, but the same problem
329: * is in updating exp_routes counter. Therefore, to be consistent in
330: * increases and decreases of exp_routes, we count exported routes
331: * regardless of blocking by limits.
332: *
333: * Similar problem is in handling updates - when a new route is
334: * received and blocking is active, the route would be blocked, but
335: * when an update for the route will be received later, the update
336: * would be propagated (as old != NULL). Therefore, we have to block
337: * also non-new updates (contrary to import blocking).
338: */
339:
340: struct proto_limit *l = ah->out_limit;
341: if (l && new)
342: {
343: if ((!old || refeed) && (stats->exp_routes >= l->limit))
344: proto_notify_limit(ah, l, PLD_OUT, stats->exp_routes);
345:
346: if (l->state == PLS_BLOCKED)
347: {
348: stats->exp_routes++; /* see note above */
349: stats->exp_updates_rejected++;
350: rte_trace_out(D_FILTERS, p, new, "rejected [limit]");
351: new = NULL;
352:
353: if (!old)
354: return;
355: }
356: }
357:
358:
359: if (new)
360: stats->exp_updates_accepted++;
361: else
362: stats->exp_withdraws_accepted++;
363:
364: /* Hack: We do not decrease exp_routes during refeed, we instead
365: reset exp_routes at the start of refeed. */
366: if (new)
367: stats->exp_routes++;
368: if (old && !refeed)
369: stats->exp_routes--;
370:
371: if (p->debug & D_ROUTES)
372: {
373: if (new && old)
374: rte_trace_out(D_ROUTES, p, new, "replaced");
375: else if (new)
376: rte_trace_out(D_ROUTES, p, new, "added");
377: else if (old)
378: rte_trace_out(D_ROUTES, p, old, "removed");
379: }
380: if (!new)
381: p->rt_notify(p, ah->table, net, NULL, old, NULL);
382: else if (tmpa)
383: {
384: ea_list *t = tmpa;
385: while (t->next)
386: t = t->next;
387: t->next = new->attrs->eattrs;
388: p->rt_notify(p, ah->table, net, new, old, tmpa);
389: t->next = NULL;
390: }
391: else
392: p->rt_notify(p, ah->table, net, new, old, new->attrs->eattrs);
393: }
394:
395: static void
396: rt_notify_basic(struct announce_hook *ah, net *net, rte *new0, rte *old0, int refeed)
397: {
398: struct proto *p = ah->proto;
399: struct proto_stats *stats = ah->stats;
400:
401: rte *new = new0;
402: rte *old = old0;
403: rte *new_free = NULL;
404: rte *old_free = NULL;
405: ea_list *tmpa = NULL;
406:
407: if (new)
408: stats->exp_updates_received++;
409: else
410: stats->exp_withdraws_received++;
411:
412: /*
413: * This is a tricky part - we don't know whether route 'old' was
414: * exported to protocol 'p' or was filtered by the export filter.
415: * We try to run the export filter to know this to have a correct
416: * value in 'old' argument of rte_update (and proper filter value)
417: *
418: * FIXME - this is broken because 'configure soft' may change
419: * filters but keep routes. Refeed is expected to be called after
420: * change of the filters and with old == new, therefore we do not
421: * even try to run the filter on an old route, This may lead to
422: * 'spurious withdraws' but ensure that there are no 'missing
423: * withdraws'.
424: *
425: * This is not completely safe as there is a window between
426: * reconfiguration and the end of refeed - if a newly filtered
427: * route disappears during this period, proper withdraw is not
428: * sent (because old would be also filtered) and the route is
1.1.1.2 ! misho 429: * not refeeded (because it disappeared before that). This is
! 430: * handled below as a special case.
1.1 misho 431: */
432:
433: if (new)
434: new = export_filter(ah, new, &new_free, &tmpa, 0);
435:
436: if (old && !refeed)
437: old = export_filter(ah, old, &old_free, NULL, 1);
438:
439: if (!new && !old)
440: {
441: /*
442: * As mentioned above, 'old' value may be incorrect in some race conditions.
1.1.1.2 ! misho 443: * We generally ignore it with two exceptions:
! 444: *
! 445: * First, withdraw to pipe protocol. In that case we rather propagate
! 446: * unfiltered withdraws regardless of export filters to ensure that when a
! 447: * protocol is flushed, its routes are removed from all tables. Possible
! 448: * spurious unfiltered withdraws are not problem here as they are ignored if
! 449: * there is no corresponding route at the other end of the pipe.
! 450: *
! 451: * Second, recent filter change. If old route is older than filter change,
! 452: * then it was previously evaluated by a different filter and we do not know
! 453: * whether it was really propagated. In that case we rather send spurious
! 454: * withdraw than do nothing and possibly cause phantom routes.
! 455: *
! 456: * In both cases wqe directly call rt_notify() hook instead of
1.1 misho 457: * do_rt_notify() to avoid logging and stat counters.
458: */
459:
1.1.1.2 ! misho 460: int pipe_withdraw = 0, filter_change = 0;
1.1 misho 461: #ifdef CONFIG_PIPE
1.1.1.2 ! misho 462: pipe_withdraw = (p->proto == &proto_pipe) && !new0;
1.1 misho 463: #endif
1.1.1.2 ! misho 464: filter_change = old0 && (old0->lastmod <= ah->last_out_filter_change);
! 465:
! 466: if ((pipe_withdraw || filter_change) && (p != old0->sender->proto))
! 467: {
! 468: stats->exp_withdraws_accepted++;
! 469: p->rt_notify(p, ah->table, net, NULL, old0, NULL);
! 470: }
1.1 misho 471:
472: return;
473: }
474:
475: do_rt_notify(ah, net, new, old, tmpa, refeed);
476:
477: /* Discard temporary rte's */
478: if (new_free)
479: rte_free(new_free);
480: if (old_free)
481: rte_free(old_free);
482: }
483:
484: static void
485: rt_notify_accepted(struct announce_hook *ah, net *net, rte *new_changed, rte *old_changed, rte *before_old, int feed)
486: {
1.1.1.2 ! misho 487: struct proto *p = ah->proto;
1.1 misho 488: struct proto_stats *stats = ah->stats;
489:
490: rte *r;
491: rte *new_best = NULL;
492: rte *old_best = NULL;
493: rte *new_free = NULL;
494: rte *old_free = NULL;
495: ea_list *tmpa = NULL;
496:
497: /* Used to track whether we met old_changed position. If before_old is NULL
498: old_changed was the first and we met it implicitly before current best route. */
499: int old_meet = old_changed && !before_old;
500:
501: /* Note that before_old is either NULL or valid (not rejected) route.
502: If old_changed is valid, before_old have to be too. If old changed route
503: was not valid, caller must use NULL for both old_changed and before_old. */
504:
505: if (new_changed)
506: stats->exp_updates_received++;
507: else
508: stats->exp_withdraws_received++;
509:
510: /* First, find the new_best route - first accepted by filters */
511: for (r=net->routes; rte_is_valid(r); r=r->next)
512: {
513: if (new_best = export_filter(ah, r, &new_free, &tmpa, 0))
514: break;
515:
516: /* Note if we walked around the position of old_changed route */
517: if (r == before_old)
518: old_meet = 1;
519: }
520:
521: /*
522: * Second, handle the feed case. That means we do not care for
523: * old_best. It is NULL for feed, and the new_best for refeed.
524: * For refeed, there is a hack similar to one in rt_notify_basic()
525: * to ensure withdraws in case of changed filters
526: */
527: if (feed)
528: {
529: if (feed == 2) /* refeed */
530: old_best = new_best ? new_best :
531: (rte_is_valid(net->routes) ? net->routes : NULL);
532: else
533: old_best = NULL;
534:
535: if (!new_best && !old_best)
536: return;
537:
538: goto found;
539: }
540:
541: /*
542: * Now, we find the old_best route. Generally, it is the same as the
543: * new_best, unless new_best is the same as new_changed or
544: * old_changed is accepted before new_best.
545: *
546: * There are four cases:
547: *
548: * - We would find and accept old_changed before new_best, therefore
549: * old_changed is old_best. In remaining cases we suppose this
550: * is not true.
551: *
552: * - We found no new_best, therefore there is also no old_best and
553: * we ignore this withdraw.
554: *
555: * - We found new_best different than new_changed, therefore
556: * old_best is the same as new_best and we ignore this update.
557: *
558: * - We found new_best the same as new_changed, therefore it cannot
559: * be old_best and we have to continue search for old_best.
1.1.1.2 ! misho 560: *
! 561: * There is also a hack to ensure consistency in case of changed filters.
! 562: * It does not find the proper old_best, just selects a non-NULL route.
1.1 misho 563: */
564:
1.1.1.2 ! misho 565: /* Hack for changed filters */
! 566: if (old_changed &&
! 567: (p != old_changed->sender->proto) &&
! 568: (old_changed->lastmod <= ah->last_out_filter_change))
! 569: {
! 570: old_best = old_changed;
! 571: goto found;
! 572: }
! 573:
1.1 misho 574: /* First case */
575: if (old_meet)
576: if (old_best = export_filter(ah, old_changed, &old_free, NULL, 1))
577: goto found;
578:
579: /* Second case */
580: if (!new_best)
581: return;
582:
583: /* Third case, we use r instead of new_best, because export_filter() could change it */
584: if (r != new_changed)
585: {
586: if (new_free)
587: rte_free(new_free);
588: return;
589: }
590:
591: /* Fourth case */
592: for (r=r->next; rte_is_valid(r); r=r->next)
593: {
594: if (old_best = export_filter(ah, r, &old_free, NULL, 1))
595: goto found;
596:
597: if (r == before_old)
598: if (old_best = export_filter(ah, old_changed, &old_free, NULL, 1))
599: goto found;
600: }
601:
602: /* Implicitly, old_best is NULL and new_best is non-NULL */
603:
604: found:
605: do_rt_notify(ah, net, new_best, old_best, tmpa, (feed == 2));
606:
607: /* Discard temporary rte's */
608: if (new_free)
609: rte_free(new_free);
610: if (old_free)
611: rte_free(old_free);
612: }
613:
614:
615: static struct mpnh *
616: mpnh_merge_rta(struct mpnh *nhs, rta *a, linpool *pool, int max)
617: {
618: struct mpnh nh = { .gw = a->gw, .iface = a->iface };
619: struct mpnh *nh2 = (a->dest == RTD_MULTIPATH) ? a->nexthops : &nh;
620: return mpnh_merge(nhs, nh2, 1, 0, max, pool);
621: }
622:
623: rte *
624: rt_export_merged(struct announce_hook *ah, net *net, rte **rt_free, ea_list **tmpa, linpool *pool, int silent)
625: {
626: // struct proto *p = ah->proto;
627: struct mpnh *nhs = NULL;
628: rte *best0, *best, *rt0, *rt, *tmp;
629:
630: best0 = net->routes;
631: *rt_free = NULL;
632:
633: if (!rte_is_valid(best0))
634: return NULL;
635:
636: best = export_filter_(ah, best0, rt_free, tmpa, pool, silent);
637:
638: if (!best || !rte_is_reachable(best))
639: return best;
640:
641: for (rt0 = best0->next; rt0; rt0 = rt0->next)
642: {
643: if (!rte_mergable(best0, rt0))
644: continue;
645:
646: rt = export_filter_(ah, rt0, &tmp, NULL, pool, 1);
647:
648: if (!rt)
649: continue;
650:
651: if (rte_is_reachable(rt))
652: nhs = mpnh_merge_rta(nhs, rt->attrs, pool, ah->proto->merge_limit);
653:
654: if (tmp)
655: rte_free(tmp);
656: }
657:
658: if (nhs)
659: {
660: nhs = mpnh_merge_rta(nhs, best->attrs, pool, ah->proto->merge_limit);
661:
662: if (nhs->next)
663: {
664: best = rte_cow_rta(best, pool);
665: best->attrs->dest = RTD_MULTIPATH;
666: best->attrs->nexthops = nhs;
667: }
668: }
669:
670: if (best != best0)
671: *rt_free = best;
672:
673: return best;
674: }
675:
676:
677: static void
678: rt_notify_merged(struct announce_hook *ah, net *net, rte *new_changed, rte *old_changed,
679: rte *new_best, rte*old_best, int refeed)
680: {
681: // struct proto *p = ah->proto;
682:
683: rte *new_best_free = NULL;
684: rte *old_best_free = NULL;
685: rte *new_changed_free = NULL;
686: rte *old_changed_free = NULL;
687: ea_list *tmpa = NULL;
688:
689: /* We assume that all rte arguments are either NULL or rte_is_valid() */
690:
691: /* This check should be done by the caller */
692: if (!new_best && !old_best)
693: return;
694:
695: /* Check whether the change is relevant to the merged route */
696: if ((new_best == old_best) && !refeed)
697: {
698: new_changed = rte_mergable(new_best, new_changed) ?
699: export_filter(ah, new_changed, &new_changed_free, NULL, 1) : NULL;
700:
701: old_changed = rte_mergable(old_best, old_changed) ?
702: export_filter(ah, old_changed, &old_changed_free, NULL, 1) : NULL;
703:
704: if (!new_changed && !old_changed)
705: return;
706: }
707:
708: if (new_best)
709: ah->stats->exp_updates_received++;
710: else
711: ah->stats->exp_withdraws_received++;
712:
713: /* Prepare new merged route */
714: if (new_best)
715: new_best = rt_export_merged(ah, net, &new_best_free, &tmpa, rte_update_pool, 0);
716:
717: /* Prepare old merged route (without proper merged next hops) */
718: /* There are some issues with running filter on old route - see rt_notify_basic() */
719: if (old_best && !refeed)
720: old_best = export_filter(ah, old_best, &old_best_free, NULL, 1);
721:
722: if (new_best || old_best)
723: do_rt_notify(ah, net, new_best, old_best, tmpa, refeed);
724:
725: /* Discard temporary rte's */
726: if (new_best_free)
727: rte_free(new_best_free);
728: if (old_best_free)
729: rte_free(old_best_free);
730: if (new_changed_free)
731: rte_free(new_changed_free);
732: if (old_changed_free)
733: rte_free(old_changed_free);
734: }
735:
736:
737: /**
738: * rte_announce - announce a routing table change
739: * @tab: table the route has been added to
740: * @type: type of route announcement (RA_OPTIMAL or RA_ANY)
741: * @net: network in question
742: * @new: the new route to be announced
743: * @old: the previous route for the same network
744: * @new_best: the new best route for the same network
745: * @old_best: the previous best route for the same network
746: * @before_old: The previous route before @old for the same network.
747: * If @before_old is NULL @old was the first.
748: *
749: * This function gets a routing table update and announces it
750: * to all protocols that acccepts given type of route announcement
751: * and are connected to the same table by their announcement hooks.
752: *
753: * Route announcement of type %RA_OPTIMAL si generated when optimal
754: * route (in routing table @tab) changes. In that case @old stores the
755: * old optimal route.
756: *
757: * Route announcement of type %RA_ANY si generated when any route (in
758: * routing table @tab) changes In that case @old stores the old route
759: * from the same protocol.
760: *
761: * For each appropriate protocol, we first call its import_control()
762: * hook which performs basic checks on the route (each protocol has a
763: * right to veto or force accept of the route before any filter is
764: * asked) and adds default values of attributes specific to the new
765: * protocol (metrics, tags etc.). Then it consults the protocol's
766: * export filter and if it accepts the route, the rt_notify() hook of
767: * the protocol gets called.
768: */
769: static void
770: rte_announce(rtable *tab, unsigned type, net *net, rte *new, rte *old,
771: rte *new_best, rte *old_best, rte *before_old)
772: {
773: if (!rte_is_valid(new))
774: new = NULL;
775:
776: if (!rte_is_valid(old))
777: old = before_old = NULL;
778:
779: if (!rte_is_valid(new_best))
780: new_best = NULL;
781:
782: if (!rte_is_valid(old_best))
783: old_best = NULL;
784:
785: if (!old && !new)
786: return;
787:
788: if (type == RA_OPTIMAL)
789: {
790: if (new)
791: new->attrs->src->proto->stats.pref_routes++;
792: if (old)
793: old->attrs->src->proto->stats.pref_routes--;
794:
795: if (tab->hostcache)
796: rt_notify_hostcache(tab, net);
797: }
798:
799: struct announce_hook *a;
800: WALK_LIST(a, tab->hooks)
801: {
802: ASSERT(a->proto->export_state != ES_DOWN);
803: if (a->proto->accept_ra_types == type)
804: if (type == RA_ACCEPTED)
805: rt_notify_accepted(a, net, new, old, before_old, 0);
806: else if (type == RA_MERGED)
807: rt_notify_merged(a, net, new, old, new_best, old_best, 0);
808: else
809: rt_notify_basic(a, net, new, old, 0);
810: }
811: }
812:
813: static inline int
814: rte_validate(rte *e)
815: {
816: int c;
817: net *n = e->net;
818:
819: if ((n->n.pxlen > BITS_PER_IP_ADDRESS) || !ip_is_prefix(n->n.prefix,n->n.pxlen))
820: {
821: log(L_WARN "Ignoring bogus prefix %I/%d received via %s",
822: n->n.prefix, n->n.pxlen, e->sender->proto->name);
823: return 0;
824: }
825:
826: c = ipa_classify_net(n->n.prefix);
827: if ((c < 0) || !(c & IADDR_HOST) || ((c & IADDR_SCOPE_MASK) <= SCOPE_LINK))
828: {
829: log(L_WARN "Ignoring bogus route %I/%d received via %s",
830: n->n.prefix, n->n.pxlen, e->sender->proto->name);
831: return 0;
832: }
833:
834: if ((e->attrs->dest == RTD_MULTIPATH) && !mpnh_is_sorted(e->attrs->nexthops))
835: {
836: log(L_WARN "Ignoring unsorted multipath route %I/%d received via %s",
837: n->n.prefix, n->n.pxlen, e->sender->proto->name);
838: return 0;
839: }
840:
841: return 1;
842: }
843:
844: /**
845: * rte_free - delete a &rte
846: * @e: &rte to be deleted
847: *
848: * rte_free() deletes the given &rte from the routing table it's linked to.
849: */
850: void
851: rte_free(rte *e)
852: {
853: if (rta_is_cached(e->attrs))
854: rta_free(e->attrs);
855: sl_free(rte_slab, e);
856: }
857:
858: static inline void
859: rte_free_quick(rte *e)
860: {
861: rta_free(e->attrs);
862: sl_free(rte_slab, e);
863: }
864:
865: static int
866: rte_same(rte *x, rte *y)
867: {
1.1.1.2 ! misho 868: /* rte.flags are not checked, as they are mostly internal to rtable */
1.1 misho 869: return
870: x->attrs == y->attrs &&
871: x->pflags == y->pflags &&
872: x->pref == y->pref &&
1.1.1.2 ! misho 873: (!x->attrs->src->proto->rte_same || x->attrs->src->proto->rte_same(x, y)) &&
! 874: rte_is_filtered(x) == rte_is_filtered(y);
1.1 misho 875: }
876:
877: static inline int rte_is_ok(rte *e) { return e && !rte_is_filtered(e); }
878:
879: static void
880: rte_recalculate(struct announce_hook *ah, net *net, rte *new, struct rte_src *src)
881: {
882: struct proto *p = ah->proto;
883: struct rtable *table = ah->table;
884: struct proto_stats *stats = ah->stats;
885: static struct tbf rl_pipe = TBF_DEFAULT_LOG_LIMITS;
886: rte *before_old = NULL;
887: rte *old_best = net->routes;
888: rte *old = NULL;
889: rte **k;
890:
891: k = &net->routes; /* Find and remove original route from the same protocol */
892: while (old = *k)
893: {
894: if (old->attrs->src == src)
895: {
896: /* If there is the same route in the routing table but from
897: * a different sender, then there are two paths from the
898: * source protocol to this routing table through transparent
899: * pipes, which is not allowed.
900: *
901: * We log that and ignore the route. If it is withdraw, we
902: * ignore it completely (there might be 'spurious withdraws',
903: * see FIXME in do_rte_announce())
904: */
905: if (old->sender->proto != p)
906: {
907: if (new)
908: {
909: log_rl(&rl_pipe, L_ERR "Pipe collision detected when sending %I/%d to table %s",
910: net->n.prefix, net->n.pxlen, table->name);
911: rte_free_quick(new);
912: }
913: return;
914: }
915:
916: if (new && rte_same(old, new))
917: {
1.1.1.2 ! misho 918: /* No changes, ignore the new route and refresh the old one */
! 919:
! 920: old->flags &= ~(REF_STALE | REF_DISCARD | REF_MODIFY);
1.1 misho 921:
922: if (!rte_is_filtered(new))
923: {
924: stats->imp_updates_ignored++;
925: rte_trace_in(D_ROUTES, p, new, "ignored");
926: }
927:
928: rte_free_quick(new);
929: return;
930: }
931: *k = old->next;
932: break;
933: }
934: k = &old->next;
935: before_old = old;
936: }
937:
938: if (!old)
939: before_old = NULL;
940:
941: if (!old && !new)
942: {
943: stats->imp_withdraws_ignored++;
944: return;
945: }
946:
947: int new_ok = rte_is_ok(new);
948: int old_ok = rte_is_ok(old);
949:
950: struct proto_limit *l = ah->rx_limit;
951: if (l && !old && new)
952: {
953: u32 all_routes = stats->imp_routes + stats->filt_routes;
954:
955: if (all_routes >= l->limit)
956: proto_notify_limit(ah, l, PLD_RX, all_routes);
957:
958: if (l->state == PLS_BLOCKED)
959: {
960: /* In receive limit the situation is simple, old is NULL so
961: we just free new and exit like nothing happened */
962:
963: stats->imp_updates_ignored++;
964: rte_trace_in(D_FILTERS, p, new, "ignored [limit]");
965: rte_free_quick(new);
966: return;
967: }
968: }
969:
970: l = ah->in_limit;
971: if (l && !old_ok && new_ok)
972: {
973: if (stats->imp_routes >= l->limit)
974: proto_notify_limit(ah, l, PLD_IN, stats->imp_routes);
975:
976: if (l->state == PLS_BLOCKED)
977: {
978: /* In import limit the situation is more complicated. We
979: shouldn't just drop the route, we should handle it like
980: it was filtered. We also have to continue the route
981: processing if old or new is non-NULL, but we should exit
982: if both are NULL as this case is probably assumed to be
983: already handled. */
984:
985: stats->imp_updates_ignored++;
986: rte_trace_in(D_FILTERS, p, new, "ignored [limit]");
987:
988: if (ah->in_keep_filtered)
989: new->flags |= REF_FILTERED;
990: else
991: { rte_free_quick(new); new = NULL; }
992:
993: /* Note that old && !new could be possible when
994: ah->in_keep_filtered changed in the recent past. */
995:
996: if (!old && !new)
997: return;
998:
999: new_ok = 0;
1000: goto skip_stats1;
1001: }
1002: }
1003:
1004: if (new_ok)
1005: stats->imp_updates_accepted++;
1006: else if (old_ok)
1007: stats->imp_withdraws_accepted++;
1008: else
1009: stats->imp_withdraws_ignored++;
1010:
1011: skip_stats1:
1012:
1013: if (new)
1014: rte_is_filtered(new) ? stats->filt_routes++ : stats->imp_routes++;
1015: if (old)
1016: rte_is_filtered(old) ? stats->filt_routes-- : stats->imp_routes--;
1017:
1018: if (table->config->sorted)
1019: {
1020: /* If routes are sorted, just insert new route to appropriate position */
1021: if (new)
1022: {
1023: if (before_old && !rte_better(new, before_old))
1024: k = &before_old->next;
1025: else
1026: k = &net->routes;
1027:
1028: for (; *k; k=&(*k)->next)
1029: if (rte_better(new, *k))
1030: break;
1031:
1032: new->next = *k;
1033: *k = new;
1034: }
1035: }
1036: else
1037: {
1038: /* If routes are not sorted, find the best route and move it on
1039: the first position. There are several optimized cases. */
1040:
1041: if (src->proto->rte_recalculate && src->proto->rte_recalculate(table, net, new, old, old_best))
1042: goto do_recalculate;
1043:
1044: if (new && rte_better(new, old_best))
1045: {
1046: /* The first case - the new route is cleary optimal,
1047: we link it at the first position */
1048:
1049: new->next = net->routes;
1050: net->routes = new;
1051: }
1052: else if (old == old_best)
1053: {
1054: /* The second case - the old best route disappeared, we add the
1055: new route (if we have any) to the list (we don't care about
1056: position) and then we elect the new optimal route and relink
1057: that route at the first position and announce it. New optimal
1058: route might be NULL if there is no more routes */
1059:
1060: do_recalculate:
1061: /* Add the new route to the list */
1062: if (new)
1063: {
1064: new->next = net->routes;
1065: net->routes = new;
1066: }
1067:
1068: /* Find a new optimal route (if there is any) */
1069: if (net->routes)
1070: {
1071: rte **bp = &net->routes;
1072: for (k=&(*bp)->next; *k; k=&(*k)->next)
1073: if (rte_better(*k, *bp))
1074: bp = k;
1075:
1076: /* And relink it */
1077: rte *best = *bp;
1078: *bp = best->next;
1079: best->next = net->routes;
1080: net->routes = best;
1081: }
1082: }
1083: else if (new)
1084: {
1085: /* The third case - the new route is not better than the old
1086: best route (therefore old_best != NULL) and the old best
1087: route was not removed (therefore old_best == net->routes).
1088: We just link the new route after the old best route. */
1089:
1090: ASSERT(net->routes != NULL);
1091: new->next = net->routes->next;
1092: net->routes->next = new;
1093: }
1094: /* The fourth (empty) case - suboptimal route was removed, nothing to do */
1095: }
1096:
1097: if (new)
1098: new->lastmod = now;
1099:
1100: /* Log the route change */
1101: if (p->debug & D_ROUTES)
1102: {
1103: if (new_ok)
1104: rte_trace(p, new, '>', new == net->routes ? "added [best]" : "added");
1105: else if (old_ok)
1106: {
1107: if (old != old_best)
1108: rte_trace(p, old, '>', "removed");
1109: else if (rte_is_ok(net->routes))
1110: rte_trace(p, old, '>', "removed [replaced]");
1111: else
1112: rte_trace(p, old, '>', "removed [sole]");
1113: }
1114: }
1115:
1116: /* Propagate the route change */
1117: rte_announce(table, RA_ANY, net, new, old, NULL, NULL, NULL);
1118: if (net->routes != old_best)
1119: rte_announce(table, RA_OPTIMAL, net, net->routes, old_best, NULL, NULL, NULL);
1120: if (table->config->sorted)
1121: rte_announce(table, RA_ACCEPTED, net, new, old, NULL, NULL, before_old);
1122: rte_announce(table, RA_MERGED, net, new, old, net->routes, old_best, NULL);
1123:
1124: if (!net->routes &&
1125: (table->gc_counter++ >= table->config->gc_max_ops) &&
1126: (table->gc_time + table->config->gc_min_time <= now))
1127: rt_schedule_gc(table);
1128:
1129: if (old_ok && p->rte_remove)
1130: p->rte_remove(net, old);
1131: if (new_ok && p->rte_insert)
1132: p->rte_insert(net, new);
1133:
1134: if (old)
1135: rte_free_quick(old);
1136: }
1137:
1138: static int rte_update_nest_cnt; /* Nesting counter to allow recursive updates */
1139:
1140: static inline void
1141: rte_update_lock(void)
1142: {
1143: rte_update_nest_cnt++;
1144: }
1145:
1146: static inline void
1147: rte_update_unlock(void)
1148: {
1149: if (!--rte_update_nest_cnt)
1150: lp_flush(rte_update_pool);
1151: }
1152:
1153: static inline void
1154: rte_hide_dummy_routes(net *net, rte **dummy)
1155: {
1156: if (net->routes && net->routes->attrs->source == RTS_DUMMY)
1157: {
1158: *dummy = net->routes;
1159: net->routes = (*dummy)->next;
1160: }
1161: }
1162:
1163: static inline void
1164: rte_unhide_dummy_routes(net *net, rte **dummy)
1165: {
1166: if (*dummy)
1167: {
1168: (*dummy)->next = net->routes;
1169: net->routes = *dummy;
1170: }
1171: }
1172:
1173: /**
1174: * rte_update - enter a new update to a routing table
1175: * @table: table to be updated
1176: * @ah: pointer to table announce hook
1177: * @net: network node
1178: * @p: protocol submitting the update
1179: * @src: protocol originating the update
1180: * @new: a &rte representing the new route or %NULL for route removal.
1181: *
1182: * This function is called by the routing protocols whenever they discover
1183: * a new route or wish to update/remove an existing route. The right announcement
1184: * sequence is to build route attributes first (either un-cached with @aflags set
1185: * to zero or a cached one using rta_lookup(); in this case please note that
1186: * you need to increase the use count of the attributes yourself by calling
1187: * rta_clone()), call rte_get_temp() to obtain a temporary &rte, fill in all
1188: * the appropriate data and finally submit the new &rte by calling rte_update().
1189: *
1190: * @src specifies the protocol that originally created the route and the meaning
1191: * of protocol-dependent data of @new. If @new is not %NULL, @src have to be the
1192: * same value as @new->attrs->proto. @p specifies the protocol that called
1193: * rte_update(). In most cases it is the same protocol as @src. rte_update()
1194: * stores @p in @new->sender;
1195: *
1196: * When rte_update() gets any route, it automatically validates it (checks,
1197: * whether the network and next hop address are valid IP addresses and also
1198: * whether a normal routing protocol doesn't try to smuggle a host or link
1199: * scope route to the table), converts all protocol dependent attributes stored
1200: * in the &rte to temporary extended attributes, consults import filters of the
1201: * protocol to see if the route should be accepted and/or its attributes modified,
1202: * stores the temporary attributes back to the &rte.
1203: *
1204: * Now, having a "public" version of the route, we
1205: * automatically find any old route defined by the protocol @src
1206: * for network @n, replace it by the new one (or removing it if @new is %NULL),
1207: * recalculate the optimal route for this destination and finally broadcast
1208: * the change (if any) to all routing protocols by calling rte_announce().
1209: *
1210: * All memory used for attribute lists and other temporary allocations is taken
1211: * from a special linear pool @rte_update_pool and freed when rte_update()
1212: * finishes.
1213: */
1214:
1215: void
1216: rte_update2(struct announce_hook *ah, net *net, rte *new, struct rte_src *src)
1217: {
1218: struct proto *p = ah->proto;
1219: struct proto_stats *stats = ah->stats;
1220: struct filter *filter = ah->in_filter;
1221: ea_list *tmpa = NULL;
1222: rte *dummy = NULL;
1223:
1224: rte_update_lock();
1225: if (new)
1226: {
1227: new->sender = ah;
1228:
1229: stats->imp_updates_received++;
1230: if (!rte_validate(new))
1231: {
1232: rte_trace_in(D_FILTERS, p, new, "invalid");
1233: stats->imp_updates_invalid++;
1234: goto drop;
1235: }
1236:
1237: if (filter == FILTER_REJECT)
1238: {
1239: stats->imp_updates_filtered++;
1240: rte_trace_in(D_FILTERS, p, new, "filtered out");
1241:
1242: if (! ah->in_keep_filtered)
1243: goto drop;
1244:
1245: /* new is a private copy, i could modify it */
1246: new->flags |= REF_FILTERED;
1247: }
1248: else
1249: {
1.1.1.2 ! misho 1250: tmpa = rte_make_tmp_attrs(new, rte_update_pool);
1.1 misho 1251: if (filter && (filter != FILTER_REJECT))
1252: {
1253: ea_list *old_tmpa = tmpa;
1254: int fr = f_run(filter, &new, &tmpa, rte_update_pool, 0);
1255: if (fr > F_ACCEPT)
1256: {
1257: stats->imp_updates_filtered++;
1258: rte_trace_in(D_FILTERS, p, new, "filtered out");
1259:
1260: if (! ah->in_keep_filtered)
1261: goto drop;
1262:
1263: new->flags |= REF_FILTERED;
1264: }
1265: if (tmpa != old_tmpa && src->proto->store_tmp_attrs)
1266: src->proto->store_tmp_attrs(new, tmpa);
1267: }
1268: }
1269: if (!rta_is_cached(new->attrs)) /* Need to copy attributes */
1270: new->attrs = rta_lookup(new->attrs);
1271: new->flags |= REF_COW;
1272: }
1273: else
1274: {
1275: stats->imp_withdraws_received++;
1276:
1277: if (!net || !src)
1278: {
1279: stats->imp_withdraws_ignored++;
1280: rte_update_unlock();
1281: return;
1282: }
1283: }
1284:
1285: recalc:
1286: rte_hide_dummy_routes(net, &dummy);
1287: rte_recalculate(ah, net, new, src);
1288: rte_unhide_dummy_routes(net, &dummy);
1289: rte_update_unlock();
1290: return;
1291:
1292: drop:
1293: rte_free(new);
1294: new = NULL;
1295: goto recalc;
1296: }
1297:
1298: /* Independent call to rte_announce(), used from next hop
1299: recalculation, outside of rte_update(). new must be non-NULL */
1300: static inline void
1301: rte_announce_i(rtable *tab, unsigned type, net *net, rte *new, rte *old,
1302: rte *new_best, rte *old_best)
1303: {
1304: rte_update_lock();
1305: rte_announce(tab, type, net, new, old, new_best, old_best, NULL);
1306: rte_update_unlock();
1307: }
1308:
1309: static inline void
1310: rte_discard(rte *old) /* Non-filtered route deletion, used during garbage collection */
1311: {
1312: rte_update_lock();
1313: rte_recalculate(old->sender, old->net, NULL, old->attrs->src);
1314: rte_update_unlock();
1315: }
1316:
1.1.1.2 ! misho 1317: /* Modify existing route by protocol hook, used for long-lived graceful restart */
! 1318: static inline void
! 1319: rte_modify(rte *old)
! 1320: {
! 1321: rte_update_lock();
! 1322:
! 1323: rte *new = old->sender->proto->rte_modify(old, rte_update_pool);
! 1324: if (new != old)
! 1325: {
! 1326: if (new)
! 1327: {
! 1328: if (!rta_is_cached(new->attrs))
! 1329: new->attrs = rta_lookup(new->attrs);
! 1330: new->flags = (old->flags & ~REF_MODIFY) | REF_COW;
! 1331: }
! 1332:
! 1333: rte_recalculate(old->sender, old->net, new, old->attrs->src);
! 1334: }
! 1335:
! 1336: rte_update_unlock();
! 1337: }
! 1338:
1.1 misho 1339: /* Check rtable for best route to given net whether it would be exported do p */
1340: int
1341: rt_examine(rtable *t, ip_addr prefix, int pxlen, struct proto *p, struct filter *filter)
1342: {
1343: net *n = net_find(t, prefix, pxlen);
1344: rte *rt = n ? n->routes : NULL;
1345:
1346: if (!rte_is_valid(rt))
1347: return 0;
1348:
1349: rte_update_lock();
1350:
1351: /* Rest is stripped down export_filter() */
1.1.1.2 ! misho 1352: ea_list *tmpa = rte_make_tmp_attrs(rt, rte_update_pool);
1.1 misho 1353: int v = p->import_control ? p->import_control(p, &rt, &tmpa, rte_update_pool) : 0;
1354: if (v == RIC_PROCESS)
1.1.1.2 ! misho 1355: v = (f_run(filter, &rt, &tmpa, rte_update_pool,
! 1356: FF_FORCE_TMPATTR | FF_SILENT) <= F_ACCEPT);
1.1 misho 1357:
1358: /* Discard temporary rte */
1359: if (rt != n->routes)
1360: rte_free(rt);
1361:
1362: rte_update_unlock();
1363:
1364: return v > 0;
1365: }
1366:
1367:
1368: /**
1369: * rt_refresh_begin - start a refresh cycle
1370: * @t: related routing table
1371: * @ah: related announce hook
1372: *
1373: * This function starts a refresh cycle for given routing table and announce
1374: * hook. The refresh cycle is a sequence where the protocol sends all its valid
1375: * routes to the routing table (by rte_update()). After that, all protocol
1376: * routes (more precisely routes with @ah as @sender) not sent during the
1377: * refresh cycle but still in the table from the past are pruned. This is
1378: * implemented by marking all related routes as stale by REF_STALE flag in
1379: * rt_refresh_begin(), then marking all related stale routes with REF_DISCARD
1380: * flag in rt_refresh_end() and then removing such routes in the prune loop.
1381: */
1382: void
1383: rt_refresh_begin(rtable *t, struct announce_hook *ah)
1384: {
1385: net *n;
1386: rte *e;
1387:
1388: FIB_WALK(&t->fib, fn)
1389: {
1390: n = (net *) fn;
1391: for (e = n->routes; e; e = e->next)
1392: if (e->sender == ah)
1393: e->flags |= REF_STALE;
1394: }
1395: FIB_WALK_END;
1396: }
1397:
1398: /**
1399: * rt_refresh_end - end a refresh cycle
1400: * @t: related routing table
1401: * @ah: related announce hook
1402: *
1403: * This function starts a refresh cycle for given routing table and announce
1404: * hook. See rt_refresh_begin() for description of refresh cycles.
1405: */
1406: void
1407: rt_refresh_end(rtable *t, struct announce_hook *ah)
1408: {
1409: int prune = 0;
1410: net *n;
1411: rte *e;
1412:
1413: FIB_WALK(&t->fib, fn)
1414: {
1415: n = (net *) fn;
1416: for (e = n->routes; e; e = e->next)
1417: if ((e->sender == ah) && (e->flags & REF_STALE))
1418: {
1419: e->flags |= REF_DISCARD;
1420: prune = 1;
1421: }
1422: }
1423: FIB_WALK_END;
1424:
1425: if (prune)
1426: rt_schedule_prune(t);
1427: }
1428:
1.1.1.2 ! misho 1429: void
! 1430: rt_modify_stale(rtable *t, struct announce_hook *ah)
! 1431: {
! 1432: int prune = 0;
! 1433: net *n;
! 1434: rte *e;
! 1435:
! 1436: FIB_WALK(&t->fib, fn)
! 1437: {
! 1438: n = (net *) fn;
! 1439: for (e = n->routes; e; e = e->next)
! 1440: if ((e->sender == ah) && (e->flags & REF_STALE) && !(e->flags & REF_FILTERED))
! 1441: {
! 1442: e->flags |= REF_MODIFY;
! 1443: prune = 1;
! 1444: }
! 1445: }
! 1446: FIB_WALK_END;
! 1447:
! 1448: if (prune)
! 1449: rt_schedule_prune(t);
! 1450: }
! 1451:
1.1 misho 1452:
1453: /**
1454: * rte_dump - dump a route
1455: * @e: &rte to be dumped
1456: *
1457: * This functions dumps contents of a &rte to debug output.
1458: */
1459: void
1460: rte_dump(rte *e)
1461: {
1462: net *n = e->net;
1463: debug("%-1I/%2d ", n->n.prefix, n->n.pxlen);
1464: debug("KF=%02x PF=%02x pref=%d lm=%d ", n->n.flags, e->pflags, e->pref, now-e->lastmod);
1465: rta_dump(e->attrs);
1466: if (e->attrs->src->proto->proto->dump_attrs)
1467: e->attrs->src->proto->proto->dump_attrs(e);
1468: debug("\n");
1469: }
1470:
1471: /**
1472: * rt_dump - dump a routing table
1473: * @t: routing table to be dumped
1474: *
1475: * This function dumps contents of a given routing table to debug output.
1476: */
1477: void
1478: rt_dump(rtable *t)
1479: {
1480: rte *e;
1481: net *n;
1482: struct announce_hook *a;
1483:
1484: debug("Dump of routing table <%s>\n", t->name);
1485: #ifdef DEBUGGING
1486: fib_check(&t->fib);
1487: #endif
1488: FIB_WALK(&t->fib, fn)
1489: {
1490: n = (net *) fn;
1491: for(e=n->routes; e; e=e->next)
1492: rte_dump(e);
1493: }
1494: FIB_WALK_END;
1495: WALK_LIST(a, t->hooks)
1496: debug("\tAnnounces routes to protocol %s\n", a->proto->name);
1497: debug("\n");
1498: }
1499:
1500: /**
1501: * rt_dump_all - dump all routing tables
1502: *
1503: * This function dumps contents of all routing tables to debug output.
1504: */
1505: void
1506: rt_dump_all(void)
1507: {
1508: rtable *t;
1509:
1510: WALK_LIST(t, routing_tables)
1511: rt_dump(t);
1512: }
1513:
1514: static inline void
1515: rt_schedule_prune(rtable *tab)
1516: {
1517: rt_mark_for_prune(tab);
1518: ev_schedule(tab->rt_event);
1519: }
1520:
1521: static inline void
1522: rt_schedule_gc(rtable *tab)
1523: {
1524: if (tab->gc_scheduled)
1525: return;
1526:
1527: tab->gc_scheduled = 1;
1528: ev_schedule(tab->rt_event);
1529: }
1530:
1531: static inline void
1532: rt_schedule_hcu(rtable *tab)
1533: {
1534: if (tab->hcu_scheduled)
1535: return;
1536:
1537: tab->hcu_scheduled = 1;
1538: ev_schedule(tab->rt_event);
1539: }
1540:
1541: static inline void
1542: rt_schedule_nhu(rtable *tab)
1543: {
1544: if (tab->nhu_state == 0)
1545: ev_schedule(tab->rt_event);
1546:
1547: /* state change 0->1, 2->3 */
1548: tab->nhu_state |= 1;
1549: }
1550:
1551:
1552: static void
1553: rt_prune_nets(rtable *tab)
1554: {
1555: struct fib_iterator fit;
1556: int ncnt = 0, ndel = 0;
1557:
1558: #ifdef DEBUGGING
1559: fib_check(&tab->fib);
1560: #endif
1561:
1562: FIB_ITERATE_INIT(&fit, &tab->fib);
1563: again:
1564: FIB_ITERATE_START(&tab->fib, &fit, f)
1565: {
1566: net *n = (net *) f;
1567: ncnt++;
1568: if (!n->routes) /* Orphaned FIB entry */
1569: {
1570: FIB_ITERATE_PUT(&fit, f);
1571: fib_delete(&tab->fib, f);
1572: ndel++;
1573: goto again;
1574: }
1575: }
1576: FIB_ITERATE_END(f);
1577: DBG("Pruned %d of %d networks\n", ndel, ncnt);
1578:
1579: tab->gc_counter = 0;
1580: tab->gc_time = now;
1581: tab->gc_scheduled = 0;
1582: }
1583:
1584: static void
1585: rt_event(void *ptr)
1586: {
1587: rtable *tab = ptr;
1588:
1589: if (tab->hcu_scheduled)
1590: rt_update_hostcache(tab);
1591:
1592: if (tab->nhu_state)
1593: rt_next_hop_update(tab);
1594:
1595: if (tab->prune_state)
1596: if (!rt_prune_table(tab))
1597: {
1598: /* Table prune unfinished */
1599: ev_schedule(tab->rt_event);
1600: return;
1601: }
1602:
1603: if (tab->gc_scheduled)
1604: {
1605: rt_prune_nets(tab);
1606: rt_prune_sources(); // FIXME this should be moved to independent event
1607: }
1608: }
1609:
1610: void
1611: rt_setup(pool *p, rtable *t, char *name, struct rtable_config *cf)
1612: {
1613: bzero(t, sizeof(*t));
1614: fib_init(&t->fib, p, sizeof(net), 0, rte_init);
1615: t->name = name;
1616: t->config = cf;
1617: init_list(&t->hooks);
1618: if (cf)
1619: {
1620: t->rt_event = ev_new(p);
1621: t->rt_event->hook = rt_event;
1622: t->rt_event->data = t;
1623: t->gc_time = now;
1624: }
1625: }
1626:
1627: /**
1628: * rt_init - initialize routing tables
1629: *
1630: * This function is called during BIRD startup. It initializes the
1631: * routing table module.
1632: */
1633: void
1634: rt_init(void)
1635: {
1636: rta_init();
1637: rt_table_pool = rp_new(&root_pool, "Routing tables");
1638: rte_update_pool = lp_new(rt_table_pool, 4080);
1639: rte_slab = sl_new(rt_table_pool, sizeof(rte));
1640: init_list(&routing_tables);
1641: }
1642:
1643:
1644: static int
1645: rt_prune_step(rtable *tab, int *limit)
1646: {
1647: struct fib_iterator *fit = &tab->prune_fit;
1648:
1649: DBG("Pruning route table %s\n", tab->name);
1650: #ifdef DEBUGGING
1651: fib_check(&tab->fib);
1652: #endif
1653:
1654: if (tab->prune_state == RPS_NONE)
1655: return 1;
1656:
1657: if (tab->prune_state == RPS_SCHEDULED)
1658: {
1659: FIB_ITERATE_INIT(fit, &tab->fib);
1660: tab->prune_state = RPS_RUNNING;
1661: }
1662:
1663: again:
1664: FIB_ITERATE_START(&tab->fib, fit, fn)
1665: {
1666: net *n = (net *) fn;
1667: rte *e;
1668:
1669: rescan:
1670: for (e=n->routes; e; e=e->next)
1.1.1.2 ! misho 1671: {
1.1 misho 1672: if (e->sender->proto->flushing || (e->flags & REF_DISCARD))
1673: {
1674: if (*limit <= 0)
1675: {
1676: FIB_ITERATE_PUT(fit, fn);
1677: return 0;
1678: }
1679:
1680: rte_discard(e);
1681: (*limit)--;
1682:
1683: goto rescan;
1684: }
1.1.1.2 ! misho 1685:
! 1686: if (e->flags & REF_MODIFY)
! 1687: {
! 1688: if (*limit <= 0)
! 1689: {
! 1690: FIB_ITERATE_PUT(fit, fn);
! 1691: return 0;
! 1692: }
! 1693:
! 1694: rte_modify(e);
! 1695: (*limit)--;
! 1696:
! 1697: goto rescan;
! 1698: }
! 1699: }
! 1700:
1.1 misho 1701: if (!n->routes) /* Orphaned FIB entry */
1702: {
1703: FIB_ITERATE_PUT(fit, fn);
1704: fib_delete(&tab->fib, fn);
1705: goto again;
1706: }
1707: }
1708: FIB_ITERATE_END(fn);
1709:
1710: #ifdef DEBUGGING
1711: fib_check(&tab->fib);
1712: #endif
1713:
1714: tab->prune_state = RPS_NONE;
1715: return 1;
1716: }
1717:
1718: /**
1719: * rt_prune_table - prune a routing table
1720: * @tab: a routing table for pruning
1721: *
1722: * This function scans the routing table @tab and removes routes belonging to
1723: * flushing protocols, discarded routes and also stale network entries, in a
1724: * similar fashion like rt_prune_loop(). Returns 1 when all such routes are
1725: * pruned. Contrary to rt_prune_loop(), this function is not a part of the
1726: * protocol flushing loop, but it is called from rt_event() for just one routing
1727: * table.
1728: *
1729: * Note that rt_prune_table() and rt_prune_loop() share (for each table) the
1730: * prune state (@prune_state) and also the pruning iterator (@prune_fit).
1731: */
1732: static inline int
1733: rt_prune_table(rtable *tab)
1734: {
1735: int limit = 512;
1736: return rt_prune_step(tab, &limit);
1737: }
1738:
1739: /**
1740: * rt_prune_loop - prune routing tables
1741: *
1742: * The prune loop scans routing tables and removes routes belonging to flushing
1743: * protocols, discarded routes and also stale network entries. Returns 1 when
1744: * all such routes are pruned. It is a part of the protocol flushing loop.
1745: */
1746: int
1747: rt_prune_loop(void)
1748: {
1749: int limit = 512;
1750: rtable *t;
1751:
1752: WALK_LIST(t, routing_tables)
1753: if (! rt_prune_step(t, &limit))
1754: return 0;
1755:
1756: return 1;
1757: }
1758:
1759: void
1760: rt_preconfig(struct config *c)
1761: {
1762: struct symbol *s = cf_get_symbol("master");
1763:
1764: init_list(&c->tables);
1765: c->master_rtc = rt_new_table(s);
1766: }
1767:
1768:
1769: /*
1770: * Some functions for handing internal next hop updates
1771: * triggered by rt_schedule_nhu().
1772: */
1773:
1774: static inline int
1775: rta_next_hop_outdated(rta *a)
1776: {
1777: struct hostentry *he = a->hostentry;
1778:
1779: if (!he)
1780: return 0;
1781:
1782: if (!he->src)
1783: return a->dest != RTD_UNREACHABLE;
1784:
1785: return (a->iface != he->src->iface) || !ipa_equal(a->gw, he->gw) ||
1786: (a->dest != he->dest) || (a->igp_metric != he->igp_metric) ||
1787: !mpnh_same(a->nexthops, he->src->nexthops);
1788: }
1789:
1790: static inline void
1791: rta_apply_hostentry(rta *a, struct hostentry *he)
1792: {
1793: a->hostentry = he;
1794: a->iface = he->src ? he->src->iface : NULL;
1795: a->gw = he->gw;
1796: a->dest = he->dest;
1797: a->igp_metric = he->igp_metric;
1798: a->nexthops = he->src ? he->src->nexthops : NULL;
1799: }
1800:
1801: static inline rte *
1802: rt_next_hop_update_rte(rtable *tab UNUSED, rte *old)
1803: {
1804: rta a;
1805: memcpy(&a, old->attrs, sizeof(rta));
1806: rta_apply_hostentry(&a, old->attrs->hostentry);
1807: a.aflags = 0;
1808:
1809: rte *e = sl_alloc(rte_slab);
1810: memcpy(e, old, sizeof(rte));
1811: e->attrs = rta_lookup(&a);
1812:
1813: return e;
1814: }
1815:
1816: static inline int
1817: rt_next_hop_update_net(rtable *tab, net *n)
1818: {
1819: rte **k, *e, *new, *old_best, **new_best;
1820: int count = 0;
1821: int free_old_best = 0;
1822:
1823: old_best = n->routes;
1824: if (!old_best)
1825: return 0;
1826:
1827: for (k = &n->routes; e = *k; k = &e->next)
1828: if (rta_next_hop_outdated(e->attrs))
1829: {
1830: new = rt_next_hop_update_rte(tab, e);
1831: *k = new;
1832:
1833: rte_announce_i(tab, RA_ANY, n, new, e, NULL, NULL);
1834: rte_trace_in(D_ROUTES, new->sender->proto, new, "updated");
1835:
1836: /* Call a pre-comparison hook */
1837: /* Not really an efficient way to compute this */
1838: if (e->attrs->src->proto->rte_recalculate)
1839: e->attrs->src->proto->rte_recalculate(tab, n, new, e, NULL);
1840:
1841: if (e != old_best)
1842: rte_free_quick(e);
1843: else /* Freeing of the old best rte is postponed */
1844: free_old_best = 1;
1845:
1846: e = new;
1847: count++;
1848: }
1849:
1850: if (!count)
1851: return 0;
1852:
1853: /* Find the new best route */
1854: new_best = NULL;
1855: for (k = &n->routes; e = *k; k = &e->next)
1856: {
1857: if (!new_best || rte_better(e, *new_best))
1858: new_best = k;
1859: }
1860:
1861: /* Relink the new best route to the first position */
1862: new = *new_best;
1863: if (new != n->routes)
1864: {
1865: *new_best = new->next;
1866: new->next = n->routes;
1867: n->routes = new;
1868: }
1869:
1870: /* Announce the new best route */
1871: if (new != old_best)
1872: {
1873: rte_announce_i(tab, RA_OPTIMAL, n, new, old_best, NULL, NULL);
1874: rte_trace_in(D_ROUTES, new->sender->proto, new, "updated [best]");
1875: }
1876:
1877: /* FIXME: Better announcement of merged routes */
1878: rte_announce_i(tab, RA_MERGED, n, new, old_best, new, old_best);
1879:
1880: if (free_old_best)
1881: rte_free_quick(old_best);
1882:
1883: return count;
1884: }
1885:
1886: static void
1887: rt_next_hop_update(rtable *tab)
1888: {
1889: struct fib_iterator *fit = &tab->nhu_fit;
1890: int max_feed = 32;
1891:
1892: if (tab->nhu_state == 0)
1893: return;
1894:
1895: if (tab->nhu_state == 1)
1896: {
1897: FIB_ITERATE_INIT(fit, &tab->fib);
1898: tab->nhu_state = 2;
1899: }
1900:
1901: FIB_ITERATE_START(&tab->fib, fit, fn)
1902: {
1903: if (max_feed <= 0)
1904: {
1905: FIB_ITERATE_PUT(fit, fn);
1906: ev_schedule(tab->rt_event);
1907: return;
1908: }
1909: max_feed -= rt_next_hop_update_net(tab, (net *) fn);
1910: }
1911: FIB_ITERATE_END(fn);
1912:
1913: /* state change 2->0, 3->1 */
1914: tab->nhu_state &= 1;
1915:
1916: if (tab->nhu_state > 0)
1917: ev_schedule(tab->rt_event);
1918: }
1919:
1920:
1921: struct rtable_config *
1922: rt_new_table(struct symbol *s)
1923: {
1924: /* Hack that allows to 'redefine' the master table */
1925: if ((s->class == SYM_TABLE) && (s->def == new_config->master_rtc))
1926: return s->def;
1927:
1928: struct rtable_config *c = cfg_allocz(sizeof(struct rtable_config));
1929:
1930: cf_define_symbol(s, SYM_TABLE, c);
1931: c->name = s->name;
1932: add_tail(&new_config->tables, &c->n);
1933: c->gc_max_ops = 1000;
1934: c->gc_min_time = 5;
1935: return c;
1936: }
1937:
1938: /**
1939: * rt_lock_table - lock a routing table
1940: * @r: routing table to be locked
1941: *
1942: * Lock a routing table, because it's in use by a protocol,
1943: * preventing it from being freed when it gets undefined in a new
1944: * configuration.
1945: */
1946: void
1947: rt_lock_table(rtable *r)
1948: {
1949: r->use_count++;
1950: }
1951:
1952: /**
1953: * rt_unlock_table - unlock a routing table
1954: * @r: routing table to be unlocked
1955: *
1956: * Unlock a routing table formerly locked by rt_lock_table(),
1957: * that is decrease its use count and delete it if it's scheduled
1958: * for deletion by configuration changes.
1959: */
1960: void
1961: rt_unlock_table(rtable *r)
1962: {
1963: if (!--r->use_count && r->deleted)
1964: {
1965: struct config *conf = r->deleted;
1966: DBG("Deleting routing table %s\n", r->name);
1967: r->config->table = NULL;
1968: if (r->hostcache)
1969: rt_free_hostcache(r);
1970: rem_node(&r->n);
1971: fib_free(&r->fib);
1972: rfree(r->rt_event);
1973: mb_free(r);
1974: config_del_obstacle(conf);
1975: }
1976: }
1977:
1978: /**
1979: * rt_commit - commit new routing table configuration
1980: * @new: new configuration
1981: * @old: original configuration or %NULL if it's boot time config
1982: *
1983: * Scan differences between @old and @new configuration and modify
1984: * the routing tables according to these changes. If @new defines a
1985: * previously unknown table, create it, if it omits a table existing
1986: * in @old, schedule it for deletion (it gets deleted when all protocols
1987: * disconnect from it by calling rt_unlock_table()), if it exists
1988: * in both configurations, leave it unchanged.
1989: */
1990: void
1991: rt_commit(struct config *new, struct config *old)
1992: {
1993: struct rtable_config *o, *r;
1994:
1995: DBG("rt_commit:\n");
1996: if (old)
1997: {
1998: WALK_LIST(o, old->tables)
1999: {
2000: rtable *ot = o->table;
2001: if (!ot->deleted)
2002: {
2003: struct symbol *sym = cf_find_symbol(new, o->name);
2004: if (sym && sym->class == SYM_TABLE && !new->shutdown)
2005: {
2006: DBG("\t%s: same\n", o->name);
2007: r = sym->def;
2008: r->table = ot;
2009: ot->name = r->name;
2010: ot->config = r;
2011: if (o->sorted != r->sorted)
2012: log(L_WARN "Reconfiguration of rtable sorted flag not implemented");
2013: }
2014: else
2015: {
2016: DBG("\t%s: deleted\n", o->name);
2017: ot->deleted = old;
2018: config_add_obstacle(old);
2019: rt_lock_table(ot);
2020: rt_unlock_table(ot);
2021: }
2022: }
2023: }
2024: }
2025:
2026: WALK_LIST(r, new->tables)
2027: if (!r->table)
2028: {
2029: rtable *t = mb_alloc(rt_table_pool, sizeof(struct rtable));
2030: DBG("\t%s: created\n", r->name);
2031: rt_setup(rt_table_pool, t, r->name, r);
2032: add_tail(&routing_tables, &t->n);
2033: r->table = t;
2034: }
2035: DBG("\tdone\n");
2036: }
2037:
2038: static inline void
2039: do_feed_baby(struct proto *p, int type, struct announce_hook *h, net *n, rte *e)
2040: {
2041: rte_update_lock();
2042: if (type == RA_ACCEPTED)
2043: rt_notify_accepted(h, n, e, NULL, NULL, p->refeeding ? 2 : 1);
2044: else if (type == RA_MERGED)
2045: rt_notify_merged(h, n, NULL, NULL, e, p->refeeding ? e : NULL, p->refeeding);
2046: else
2047: rt_notify_basic(h, n, e, p->refeeding ? e : NULL, p->refeeding);
2048: rte_update_unlock();
2049: }
2050:
2051: /**
2052: * rt_feed_baby - advertise routes to a new protocol
2053: * @p: protocol to be fed
2054: *
2055: * This function performs one pass of advertisement of routes to a newly
2056: * initialized protocol. It's called by the protocol code as long as it
2057: * has something to do. (We avoid transferring all the routes in single
2058: * pass in order not to monopolize CPU time.)
2059: */
2060: int
2061: rt_feed_baby(struct proto *p)
2062: {
2063: struct announce_hook *h;
2064: struct fib_iterator *fit;
2065: int max_feed = 256;
2066:
2067: if (!p->feed_ahook) /* Need to initialize first */
2068: {
2069: if (!p->ahooks)
2070: return 1;
2071: DBG("Announcing routes to new protocol %s\n", p->name);
2072: p->feed_ahook = p->ahooks;
2073: fit = p->feed_iterator = mb_alloc(p->pool, sizeof(struct fib_iterator));
2074: goto next_hook;
2075: }
2076: fit = p->feed_iterator;
2077:
2078: again:
2079: h = p->feed_ahook;
2080: FIB_ITERATE_START(&h->table->fib, fit, fn)
2081: {
2082: net *n = (net *) fn;
2083: rte *e = n->routes;
2084: if (max_feed <= 0)
2085: {
2086: FIB_ITERATE_PUT(fit, fn);
2087: return 0;
2088: }
2089:
2090: /* XXXX perhaps we should change feed for RA_ACCEPTED to not use 'new' */
2091:
2092: if ((p->accept_ra_types == RA_OPTIMAL) ||
2093: (p->accept_ra_types == RA_ACCEPTED) ||
2094: (p->accept_ra_types == RA_MERGED))
2095: if (rte_is_valid(e))
2096: {
2097: if (p->export_state != ES_FEEDING)
2098: return 1; /* In the meantime, the protocol fell down. */
2099:
2100: do_feed_baby(p, p->accept_ra_types, h, n, e);
2101: max_feed--;
2102: }
2103:
2104: if (p->accept_ra_types == RA_ANY)
2105: for(e = n->routes; e; e = e->next)
2106: {
2107: if (p->export_state != ES_FEEDING)
2108: return 1; /* In the meantime, the protocol fell down. */
2109:
2110: if (!rte_is_valid(e))
2111: continue;
2112:
2113: do_feed_baby(p, RA_ANY, h, n, e);
2114: max_feed--;
2115: }
2116: }
2117: FIB_ITERATE_END(fn);
2118: p->feed_ahook = h->next;
2119: if (!p->feed_ahook)
2120: {
2121: mb_free(p->feed_iterator);
2122: p->feed_iterator = NULL;
2123: return 1;
2124: }
2125:
2126: next_hook:
2127: h = p->feed_ahook;
2128: FIB_ITERATE_INIT(fit, &h->table->fib);
2129: goto again;
2130: }
2131:
2132: /**
2133: * rt_feed_baby_abort - abort protocol feeding
2134: * @p: protocol
2135: *
2136: * This function is called by the protocol code when the protocol
2137: * stops or ceases to exist before the last iteration of rt_feed_baby()
2138: * has finished.
2139: */
2140: void
2141: rt_feed_baby_abort(struct proto *p)
2142: {
2143: if (p->feed_ahook)
2144: {
2145: /* Unlink the iterator and exit */
2146: fit_get(&p->feed_ahook->table->fib, p->feed_iterator);
2147: p->feed_ahook = NULL;
2148: }
2149: }
2150:
2151:
2152: static inline unsigned
2153: ptr_hash(void *ptr)
2154: {
2155: uintptr_t p = (uintptr_t) ptr;
2156: return p ^ (p << 8) ^ (p >> 16);
2157: }
2158:
2159: static inline unsigned
2160: hc_hash(ip_addr a, rtable *dep)
2161: {
2162: return (ipa_hash(a) ^ ptr_hash(dep)) & 0xffff;
2163: }
2164:
2165: static inline void
2166: hc_insert(struct hostcache *hc, struct hostentry *he)
2167: {
2168: uint k = he->hash_key >> hc->hash_shift;
2169: he->next = hc->hash_table[k];
2170: hc->hash_table[k] = he;
2171: }
2172:
2173: static inline void
2174: hc_remove(struct hostcache *hc, struct hostentry *he)
2175: {
2176: struct hostentry **hep;
2177: uint k = he->hash_key >> hc->hash_shift;
2178:
2179: for (hep = &hc->hash_table[k]; *hep != he; hep = &(*hep)->next);
2180: *hep = he->next;
2181: }
2182:
2183: #define HC_DEF_ORDER 10
2184: #define HC_HI_MARK *4
2185: #define HC_HI_STEP 2
2186: #define HC_HI_ORDER 16 /* Must be at most 16 */
2187: #define HC_LO_MARK /5
2188: #define HC_LO_STEP 2
2189: #define HC_LO_ORDER 10
2190:
2191: static void
2192: hc_alloc_table(struct hostcache *hc, unsigned order)
2193: {
2194: uint hsize = 1 << order;
2195: hc->hash_order = order;
2196: hc->hash_shift = 16 - order;
2197: hc->hash_max = (order >= HC_HI_ORDER) ? ~0U : (hsize HC_HI_MARK);
2198: hc->hash_min = (order <= HC_LO_ORDER) ? 0U : (hsize HC_LO_MARK);
2199:
2200: hc->hash_table = mb_allocz(rt_table_pool, hsize * sizeof(struct hostentry *));
2201: }
2202:
2203: static void
2204: hc_resize(struct hostcache *hc, unsigned new_order)
2205: {
2206: struct hostentry **old_table = hc->hash_table;
2207: struct hostentry *he, *hen;
2208: uint old_size = 1 << hc->hash_order;
2209: uint i;
2210:
2211: hc_alloc_table(hc, new_order);
2212: for (i = 0; i < old_size; i++)
2213: for (he = old_table[i]; he != NULL; he=hen)
2214: {
2215: hen = he->next;
2216: hc_insert(hc, he);
2217: }
2218: mb_free(old_table);
2219: }
2220:
2221: static struct hostentry *
2222: hc_new_hostentry(struct hostcache *hc, ip_addr a, ip_addr ll, rtable *dep, unsigned k)
2223: {
2224: struct hostentry *he = sl_alloc(hc->slab);
2225:
2226: he->addr = a;
2227: he->link = ll;
2228: he->tab = dep;
2229: he->hash_key = k;
2230: he->uc = 0;
2231: he->src = NULL;
2232:
2233: add_tail(&hc->hostentries, &he->ln);
2234: hc_insert(hc, he);
2235:
2236: hc->hash_items++;
2237: if (hc->hash_items > hc->hash_max)
2238: hc_resize(hc, hc->hash_order + HC_HI_STEP);
2239:
2240: return he;
2241: }
2242:
2243: static void
2244: hc_delete_hostentry(struct hostcache *hc, struct hostentry *he)
2245: {
2246: rta_free(he->src);
2247:
2248: rem_node(&he->ln);
2249: hc_remove(hc, he);
2250: sl_free(hc->slab, he);
2251:
2252: hc->hash_items--;
2253: if (hc->hash_items < hc->hash_min)
2254: hc_resize(hc, hc->hash_order - HC_LO_STEP);
2255: }
2256:
2257: static void
2258: rt_init_hostcache(rtable *tab)
2259: {
2260: struct hostcache *hc = mb_allocz(rt_table_pool, sizeof(struct hostcache));
2261: init_list(&hc->hostentries);
2262:
2263: hc->hash_items = 0;
2264: hc_alloc_table(hc, HC_DEF_ORDER);
2265: hc->slab = sl_new(rt_table_pool, sizeof(struct hostentry));
2266:
2267: hc->lp = lp_new(rt_table_pool, 1008);
2268: hc->trie = f_new_trie(hc->lp, sizeof(struct f_trie_node));
2269:
2270: tab->hostcache = hc;
2271: }
2272:
2273: static void
2274: rt_free_hostcache(rtable *tab)
2275: {
2276: struct hostcache *hc = tab->hostcache;
2277:
2278: node *n;
2279: WALK_LIST(n, hc->hostentries)
2280: {
2281: struct hostentry *he = SKIP_BACK(struct hostentry, ln, n);
2282: rta_free(he->src);
2283:
2284: if (he->uc)
2285: log(L_ERR "Hostcache is not empty in table %s", tab->name);
2286: }
2287:
2288: rfree(hc->slab);
2289: rfree(hc->lp);
2290: mb_free(hc->hash_table);
2291: mb_free(hc);
2292: }
2293:
2294: static void
2295: rt_notify_hostcache(rtable *tab, net *net)
2296: {
2297: struct hostcache *hc = tab->hostcache;
2298:
2299: if (tab->hcu_scheduled)
2300: return;
2301:
2302: if (trie_match_prefix(hc->trie, net->n.prefix, net->n.pxlen))
2303: rt_schedule_hcu(tab);
2304: }
2305:
2306: static int
2307: if_local_addr(ip_addr a, struct iface *i)
2308: {
2309: struct ifa *b;
2310:
2311: WALK_LIST(b, i->addrs)
2312: if (ipa_equal(a, b->ip))
2313: return 1;
2314:
2315: return 0;
2316: }
2317:
2318: static u32
2319: rt_get_igp_metric(rte *rt)
2320: {
2321: eattr *ea = ea_find(rt->attrs->eattrs, EA_GEN_IGP_METRIC);
2322:
2323: if (ea)
2324: return ea->u.data;
2325:
2326: rta *a = rt->attrs;
2327:
2328: #ifdef CONFIG_OSPF
2329: if ((a->source == RTS_OSPF) ||
2330: (a->source == RTS_OSPF_IA) ||
2331: (a->source == RTS_OSPF_EXT1))
2332: return rt->u.ospf.metric1;
2333: #endif
2334:
2335: #ifdef CONFIG_RIP
2336: if (a->source == RTS_RIP)
2337: return rt->u.rip.metric;
2338: #endif
2339:
2340: /* Device routes */
2341: if ((a->dest != RTD_ROUTER) && (a->dest != RTD_MULTIPATH))
2342: return 0;
2343:
2344: return IGP_METRIC_UNKNOWN;
2345: }
2346:
2347: static int
2348: rt_update_hostentry(rtable *tab, struct hostentry *he)
2349: {
2350: rta *old_src = he->src;
2351: int pxlen = 0;
2352:
2353: /* Reset the hostentry */
2354: he->src = NULL;
2355: he->gw = IPA_NONE;
2356: he->dest = RTD_UNREACHABLE;
2357: he->igp_metric = 0;
2358:
2359: net *n = net_route(tab, he->addr, MAX_PREFIX_LENGTH);
2360: if (n)
2361: {
2362: rte *e = n->routes;
2363: rta *a = e->attrs;
2364: pxlen = n->n.pxlen;
2365:
2366: if (a->hostentry)
2367: {
2368: /* Recursive route should not depend on another recursive route */
2369: log(L_WARN "Next hop address %I resolvable through recursive route for %I/%d",
2370: he->addr, n->n.prefix, pxlen);
2371: goto done;
2372: }
2373:
2374: if (a->dest == RTD_DEVICE)
2375: {
2376: if (if_local_addr(he->addr, a->iface))
2377: {
2378: /* The host address is a local address, this is not valid */
2379: log(L_WARN "Next hop address %I is a local address of iface %s",
2380: he->addr, a->iface->name);
2381: goto done;
2382: }
2383:
2384: /* The host is directly reachable, use link as a gateway */
2385: he->gw = he->link;
2386: he->dest = RTD_ROUTER;
2387: }
2388: else
2389: {
2390: /* The host is reachable through some route entry */
2391: he->gw = a->gw;
2392: he->dest = a->dest;
2393: }
2394:
2395: he->src = rta_clone(a);
2396: he->igp_metric = rt_get_igp_metric(e);
2397: }
2398:
2399: done:
2400: /* Add a prefix range to the trie */
2401: trie_add_prefix(tab->hostcache->trie, he->addr, MAX_PREFIX_LENGTH, pxlen, MAX_PREFIX_LENGTH);
2402:
2403: rta_free(old_src);
2404: return old_src != he->src;
2405: }
2406:
2407: static void
2408: rt_update_hostcache(rtable *tab)
2409: {
2410: struct hostcache *hc = tab->hostcache;
2411: struct hostentry *he;
2412: node *n, *x;
2413:
2414: /* Reset the trie */
2415: lp_flush(hc->lp);
2416: hc->trie = f_new_trie(hc->lp, sizeof(struct f_trie_node));
2417:
2418: WALK_LIST_DELSAFE(n, x, hc->hostentries)
2419: {
2420: he = SKIP_BACK(struct hostentry, ln, n);
2421: if (!he->uc)
2422: {
2423: hc_delete_hostentry(hc, he);
2424: continue;
2425: }
2426:
2427: if (rt_update_hostentry(tab, he))
2428: rt_schedule_nhu(he->tab);
2429: }
2430:
2431: tab->hcu_scheduled = 0;
2432: }
2433:
2434: static struct hostentry *
2435: rt_get_hostentry(rtable *tab, ip_addr a, ip_addr ll, rtable *dep)
2436: {
2437: struct hostentry *he;
2438:
2439: if (!tab->hostcache)
2440: rt_init_hostcache(tab);
2441:
2442: uint k = hc_hash(a, dep);
2443: struct hostcache *hc = tab->hostcache;
2444: for (he = hc->hash_table[k >> hc->hash_shift]; he != NULL; he = he->next)
2445: if (ipa_equal(he->addr, a) && (he->tab == dep))
2446: return he;
2447:
2448: he = hc_new_hostentry(hc, a, ll, dep, k);
2449: rt_update_hostentry(tab, he);
2450: return he;
2451: }
2452:
2453: void
2454: rta_set_recursive_next_hop(rtable *dep, rta *a, rtable *tab, ip_addr *gw, ip_addr *ll)
2455: {
2456: rta_apply_hostentry(a, rt_get_hostentry(tab, *gw, *ll, dep));
2457: }
2458:
2459:
2460: /*
2461: * CLI commands
2462: */
2463:
2464: static byte *
2465: rt_format_via(rte *e)
2466: {
2467: rta *a = e->attrs;
2468:
2469: /* Max text length w/o IP addr and interface name is 16 */
2470: static byte via[STD_ADDRESS_P_LENGTH+sizeof(a->iface->name)+16];
2471:
2472: switch (a->dest)
2473: {
2474: case RTD_ROUTER: bsprintf(via, "via %I on %s", a->gw, a->iface->name); break;
2475: case RTD_DEVICE: bsprintf(via, "dev %s", a->iface->name); break;
2476: case RTD_BLACKHOLE: bsprintf(via, "blackhole"); break;
2477: case RTD_UNREACHABLE: bsprintf(via, "unreachable"); break;
2478: case RTD_PROHIBIT: bsprintf(via, "prohibited"); break;
2479: case RTD_MULTIPATH: bsprintf(via, "multipath"); break;
2480: default: bsprintf(via, "???");
2481: }
2482: return via;
2483: }
2484:
2485: static void
2486: rt_show_rte(struct cli *c, byte *ia, rte *e, struct rt_show_data *d, ea_list *tmpa)
2487: {
2488: byte from[STD_ADDRESS_P_LENGTH+8];
2489: byte tm[TM_DATETIME_BUFFER_SIZE], info[256];
2490: rta *a = e->attrs;
2491: int primary = (e->net->routes == e);
2492: int sync_error = (e->net->n.flags & KRF_SYNC_ERROR);
2493: void (*get_route_info)(struct rte *, byte *buf, struct ea_list *attrs);
2494: struct mpnh *nh;
2495:
2496: tm_format_datetime(tm, &config->tf_route, e->lastmod);
2497: if (ipa_nonzero(a->from) && !ipa_equal(a->from, a->gw))
2498: bsprintf(from, " from %I", a->from);
2499: else
2500: from[0] = 0;
2501:
2502: get_route_info = a->src->proto->proto->get_route_info;
2503: if (get_route_info || d->verbose)
2504: {
2505: /* Need to normalize the extended attributes */
2506: ea_list *t = tmpa;
2507: t = ea_append(t, a->eattrs);
2508: tmpa = alloca(ea_scan(t));
2509: ea_merge(t, tmpa);
2510: ea_sort(tmpa);
2511: }
2512: if (get_route_info)
2513: get_route_info(e, info, tmpa);
2514: else
2515: bsprintf(info, " (%d)", e->pref);
2516: cli_printf(c, -1007, "%-18s %s [%s %s%s]%s%s", ia, rt_format_via(e), a->src->proto->name,
2517: tm, from, primary ? (sync_error ? " !" : " *") : "", info);
2518: for (nh = a->nexthops; nh; nh = nh->next)
2519: cli_printf(c, -1007, "\tvia %I on %s weight %d", nh->gw, nh->iface->name, nh->weight + 1);
2520: if (d->verbose)
2521: rta_show(c, a, tmpa);
2522: }
2523:
2524: static void
2525: rt_show_net(struct cli *c, net *n, struct rt_show_data *d)
2526: {
2527: rte *e, *ee;
2528: byte ia[STD_ADDRESS_P_LENGTH+8];
2529: struct ea_list *tmpa;
2530: struct announce_hook *a = NULL;
2531: int first = 1;
2532: int pass = 0;
2533:
2534: bsprintf(ia, "%I/%d", n->n.prefix, n->n.pxlen);
2535:
2536: if (d->export_mode)
2537: {
2538: if (! d->export_protocol->rt_notify)
2539: return;
2540:
2541: a = proto_find_announce_hook(d->export_protocol, d->table);
2542: if (!a)
2543: return;
2544: }
2545:
2546: for (e = n->routes; e; e = e->next)
2547: {
2548: if (rte_is_filtered(e) != d->filtered)
2549: continue;
2550:
2551: d->rt_counter++;
2552: d->net_counter += first;
2553: first = 0;
2554:
2555: if (pass)
2556: continue;
2557:
2558: ee = e;
2559: rte_update_lock(); /* We use the update buffer for filtering */
1.1.1.2 ! misho 2560: tmpa = rte_make_tmp_attrs(e, rte_update_pool);
1.1 misho 2561:
2562: /* Special case for merged export */
2563: if ((d->export_mode == RSEM_EXPORT) && (d->export_protocol->accept_ra_types == RA_MERGED))
2564: {
2565: rte *rt_free;
2566: e = rt_export_merged(a, n, &rt_free, &tmpa, rte_update_pool, 1);
2567: pass = 1;
2568:
2569: if (!e)
2570: { e = ee; goto skip; }
2571: }
2572: else if (d->export_mode)
2573: {
2574: struct proto *ep = d->export_protocol;
2575: int ic = ep->import_control ? ep->import_control(ep, &e, &tmpa, rte_update_pool) : 0;
2576:
2577: if (ep->accept_ra_types == RA_OPTIMAL || ep->accept_ra_types == RA_MERGED)
2578: pass = 1;
2579:
2580: if (ic < 0)
2581: goto skip;
2582:
2583: if (d->export_mode > RSEM_PREEXPORT)
2584: {
2585: /*
2586: * FIXME - This shows what should be exported according to current
2587: * filters, but not what was really exported. 'configure soft'
2588: * command may change the export filter and do not update routes.
2589: */
2590: int do_export = (ic > 0) ||
1.1.1.2 ! misho 2591: (f_run(a->out_filter, &e, &tmpa, rte_update_pool,
! 2592: FF_FORCE_TMPATTR | FF_SILENT) <= F_ACCEPT);
1.1 misho 2593:
2594: if (do_export != (d->export_mode == RSEM_EXPORT))
2595: goto skip;
2596:
2597: if ((d->export_mode == RSEM_EXPORT) && (ep->accept_ra_types == RA_ACCEPTED))
2598: pass = 1;
2599: }
2600: }
2601:
2602: if (d->show_protocol && (d->show_protocol != e->attrs->src->proto))
2603: goto skip;
2604:
2605: if (f_run(d->filter, &e, &tmpa, rte_update_pool, FF_FORCE_TMPATTR) > F_ACCEPT)
2606: goto skip;
2607:
2608: d->show_counter++;
2609: if (d->stats < 2)
2610: rt_show_rte(c, ia, e, d, tmpa);
2611: ia[0] = 0;
2612:
2613: skip:
2614: if (e != ee)
2615: {
2616: rte_free(e);
2617: e = ee;
2618: }
2619: rte_update_unlock();
2620:
2621: if (d->primary_only)
2622: break;
2623: }
2624: }
2625:
2626: static void
2627: rt_show_cont(struct cli *c)
2628: {
2629: struct rt_show_data *d = c->rover;
2630: #ifdef DEBUGGING
2631: unsigned max = 4;
2632: #else
2633: unsigned max = 64;
2634: #endif
2635: struct fib *fib = &d->table->fib;
2636: struct fib_iterator *it = &d->fit;
2637:
2638: FIB_ITERATE_START(fib, it, f)
2639: {
2640: net *n = (net *) f;
2641: if (d->running_on_config && d->running_on_config != config)
2642: {
2643: cli_printf(c, 8004, "Stopped due to reconfiguration");
2644: goto done;
2645: }
2646: if (d->export_protocol && (d->export_protocol->export_state == ES_DOWN))
2647: {
2648: cli_printf(c, 8005, "Protocol is down");
2649: goto done;
2650: }
2651: if (!max--)
2652: {
2653: FIB_ITERATE_PUT(it, f);
2654: return;
2655: }
2656: rt_show_net(c, n, d);
2657: }
2658: FIB_ITERATE_END(f);
2659: if (d->stats)
2660: cli_printf(c, 14, "%d of %d routes for %d networks", d->show_counter, d->rt_counter, d->net_counter);
2661: else
2662: cli_printf(c, 0, "");
2663: done:
2664: c->cont = c->cleanup = NULL;
2665: }
2666:
2667: static void
2668: rt_show_cleanup(struct cli *c)
2669: {
2670: struct rt_show_data *d = c->rover;
2671:
2672: /* Unlink the iterator */
2673: fit_get(&d->table->fib, &d->fit);
2674: }
2675:
2676: void
2677: rt_show(struct rt_show_data *d)
2678: {
2679: net *n;
2680:
2681: /* Default is either a master table or a table related to a respective protocol */
2682: if (!d->table && d->export_protocol) d->table = d->export_protocol->table;
2683: if (!d->table && d->show_protocol) d->table = d->show_protocol->table;
2684: if (!d->table) d->table = config->master_rtc->table;
2685:
2686: /* Filtered routes are neither exported nor have sensible ordering */
2687: if (d->filtered && (d->export_mode || d->primary_only))
2688: cli_msg(0, "");
2689:
2690: if (d->pxlen == 256)
2691: {
2692: FIB_ITERATE_INIT(&d->fit, &d->table->fib);
2693: this_cli->cont = rt_show_cont;
2694: this_cli->cleanup = rt_show_cleanup;
2695: this_cli->rover = d;
2696: }
2697: else
2698: {
2699: if (d->show_for)
2700: n = net_route(d->table, d->prefix, d->pxlen);
2701: else
2702: n = net_find(d->table, d->prefix, d->pxlen);
2703:
2704: if (n)
2705: rt_show_net(this_cli, n, d);
2706:
2707: if (d->rt_counter)
2708: cli_msg(0, "");
2709: else
2710: cli_msg(8001, "Network not in table");
2711: }
2712: }
2713:
2714: /*
2715: * Documentation for functions declared inline in route.h
2716: */
2717: #if 0
2718:
2719: /**
2720: * net_find - find a network entry
2721: * @tab: a routing table
2722: * @addr: address of the network
2723: * @len: length of the network prefix
2724: *
2725: * net_find() looks up the given network in routing table @tab and
2726: * returns a pointer to its &net entry or %NULL if no such network
2727: * exists.
2728: */
2729: static inline net *net_find(rtable *tab, ip_addr addr, unsigned len)
2730: { DUMMY; }
2731:
2732: /**
2733: * net_get - obtain a network entry
2734: * @tab: a routing table
2735: * @addr: address of the network
2736: * @len: length of the network prefix
2737: *
2738: * net_get() looks up the given network in routing table @tab and
2739: * returns a pointer to its &net entry. If no such entry exists, it's
2740: * created.
2741: */
2742: static inline net *net_get(rtable *tab, ip_addr addr, unsigned len)
2743: { DUMMY; }
2744:
2745: /**
2746: * rte_cow - copy a route for writing
2747: * @r: a route entry to be copied
2748: *
2749: * rte_cow() takes a &rte and prepares it for modification. The exact action
2750: * taken depends on the flags of the &rte -- if it's a temporary entry, it's
2751: * just returned unchanged, else a new temporary entry with the same contents
2752: * is created.
2753: *
2754: * The primary use of this function is inside the filter machinery -- when
2755: * a filter wants to modify &rte contents (to change the preference or to
2756: * attach another set of attributes), it must ensure that the &rte is not
2757: * shared with anyone else (and especially that it isn't stored in any routing
2758: * table).
2759: *
2760: * Result: a pointer to the new writable &rte.
2761: */
2762: static inline rte * rte_cow(rte *r)
2763: { DUMMY; }
2764:
2765: #endif
FreeBSD-CVSweb <freebsd-cvsweb@FreeBSD.org>
![[BACK]](/icons/cvsweb/back.gif){kind=icon}
Return to rt-table.c CVS log ![[TXT]](/icons/cvsweb/text.gif){kind=icon}
![[DIR]](/icons/cvsweb/dir.gif){kind=icon}
Up to [ELWIX - Embedded LightWeight unIX -] / embedaddon / bird / nest