xref: /openbsd/usr.sbin/ospf6d/kroute.c (revision 097a140d)
1 /*	$OpenBSD: kroute.c,v 1.65 2020/09/10 05:18:16 jmatthew Exp $ */
2 
3 /*
4  * Copyright (c) 2004 Esben Norby <norby@openbsd.org>
5  * Copyright (c) 2003, 2004 Henning Brauer <henning@openbsd.org>
6  *
7  * Permission to use, copy, modify, and distribute this software for any
8  * purpose with or without fee is hereby granted, provided that the above
9  * copyright notice and this permission notice appear in all copies.
10  *
11  * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
12  * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
13  * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
14  * ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
15  * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN
16  * ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
17  * OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
18  */
19 
20 #include <sys/types.h>
21 #include <sys/socket.h>
22 #include <sys/sysctl.h>
23 #include <sys/tree.h>
24 #include <sys/uio.h>
25 #include <netinet/in.h>
26 #include <arpa/inet.h>
27 #include <net/if.h>
28 #include <net/if_dl.h>
29 #include <net/if_types.h>
30 #include <net/route.h>
31 #include <err.h>
32 #include <errno.h>
33 #include <fcntl.h>
34 #include <stdio.h>
35 #include <stdlib.h>
36 #include <string.h>
37 #include <unistd.h>
38 #include <limits.h>
39 
40 #include "ospf6d.h"
41 #include "ospfe.h"
42 #include "log.h"
43 
44 struct {
45 	u_int32_t		rtseq;
46 	pid_t			pid;
47 	int			fib_sync;
48 	u_int8_t		fib_prio;
49 	int			fd;
50 	struct event		ev;
51 	u_int			rdomain;
52 } kr_state;
53 
54 struct kroute_node {
55 	RB_ENTRY(kroute_node)	 entry;
56 	struct kroute_node	*next;
57 	struct kroute		 r;
58 };
59 
60 void	kr_redist_remove(struct kroute_node *, struct kroute_node *);
61 int	kr_redist_eval(struct kroute *, struct kroute *);
62 void	kr_redistribute(struct kroute_node *);
63 int	kroute_compare(struct kroute_node *, struct kroute_node *);
64 int	kr_change_fib(struct kroute_node *, struct kroute *, int, int);
65 int	kr_delete_fib(struct kroute_node *);
66 
67 struct kroute_node	*kroute_find(const struct in6_addr *, u_int8_t,
68 			    u_int8_t);
69 struct kroute_node	*kroute_matchgw(struct kroute_node *,
70 			    struct in6_addr *, unsigned int);
71 int			 kroute_insert(struct kroute_node *);
72 int			 kroute_remove(struct kroute_node *);
73 void			 kroute_clear(void);
74 
75 struct iface		*kif_update(u_short, int, struct if_data *,
76 			   struct sockaddr_dl *);
77 int			 kif_validate(u_short);
78 
79 struct kroute_node	*kroute_match(struct in6_addr *);
80 
81 int		protect_lo(void);
82 void		get_rtaddrs(int, struct sockaddr *, struct sockaddr **);
83 void		if_change(u_short, int, struct if_data *, struct sockaddr_dl *);
84 void		if_newaddr(u_short, struct sockaddr_in6 *,
85 		    struct sockaddr_in6 *, struct sockaddr_in6 *);
86 void		if_deladdr(u_short, struct sockaddr_in6 *,
87 		    struct sockaddr_in6 *, struct sockaddr_in6 *);
88 void		if_announce(void *);
89 
90 int		send_rtmsg(int, int, struct kroute *);
91 int		dispatch_rtmsg(void);
92 int		fetchtable(void);
93 int		rtmsg_process(char *, size_t);
94 
95 RB_HEAD(kroute_tree, kroute_node)	krt;
96 RB_PROTOTYPE(kroute_tree, kroute_node, entry, kroute_compare)
97 RB_GENERATE(kroute_tree, kroute_node, entry, kroute_compare)
98 
99 int
100 kr_init(int fs, u_int rdomain, int redis_label_or_prefix, u_int8_t fib_prio)
101 {
102 	int		opt = 0, rcvbuf, default_rcvbuf;
103 	socklen_t	optlen;
104 	int		filter_prio = fib_prio;
105 	int		filter_flags = RTF_LLINFO | RTF_BROADCAST;
106 
107 	kr_state.fib_sync = fs;
108 	kr_state.rdomain = rdomain;
109 	kr_state.fib_prio = fib_prio;
110 
111 	if ((kr_state.fd = socket(AF_ROUTE,
112 	    SOCK_RAW | SOCK_CLOEXEC | SOCK_NONBLOCK, AF_INET6)) == -1) {
113 		log_warn("kr_init: socket");
114 		return (-1);
115 	}
116 
117 	/* not interested in my own messages */
118 	if (setsockopt(kr_state.fd, SOL_SOCKET, SO_USELOOPBACK,
119 	    &opt, sizeof(opt)) == -1)
120 		log_warn("kr_init: setsockopt");	/* not fatal */
121 
122 	if (redis_label_or_prefix) {
123 		filter_prio = 0;
124 		log_info("%s: priority filter disabled", __func__);
125 	} else
126 		log_debug("%s: priority filter enabled", __func__);
127 
128 	if (setsockopt(kr_state.fd, AF_ROUTE, ROUTE_PRIOFILTER, &filter_prio,
129 	    sizeof(filter_prio)) == -1) {
130 		log_warn("%s: setsockopt AF_ROUTE ROUTE_PRIOFILTER", __func__);
131 		/* not fatal */
132 	}
133 
134 	if (setsockopt(kr_state.fd, AF_ROUTE, ROUTE_FLAGFILTER, &filter_flags,
135 	    sizeof(filter_flags)) == -1) {
136 		log_warn("%s: setsockopt AF_ROUTE ROUTE_FLAGFILTER", __func__);
137 		/* not fatal */
138 	}
139 
140 	/* grow receive buffer, don't wanna miss messages */
141 	optlen = sizeof(default_rcvbuf);
142 	if (getsockopt(kr_state.fd, SOL_SOCKET, SO_RCVBUF,
143 	    &default_rcvbuf, &optlen) == -1)
144 		log_warn("kr_init getsockopt SOL_SOCKET SO_RCVBUF");
145 	else
146 		for (rcvbuf = MAX_RTSOCK_BUF;
147 		    rcvbuf > default_rcvbuf &&
148 		    setsockopt(kr_state.fd, SOL_SOCKET, SO_RCVBUF,
149 		    &rcvbuf, sizeof(rcvbuf)) == -1 && errno == ENOBUFS;
150 		    rcvbuf /= 2)
151 			;	/* nothing */
152 
153 	kr_state.pid = getpid();
154 	kr_state.rtseq = 1;
155 
156 	RB_INIT(&krt);
157 
158 	if (fetchtable() == -1)
159 		return (-1);
160 
161 	if (protect_lo() == -1)
162 		return (-1);
163 
164 	event_set(&kr_state.ev, kr_state.fd, EV_READ | EV_PERSIST,
165 	    kr_dispatch_msg, NULL);
166 	event_add(&kr_state.ev, NULL);
167 
168 	return (0);
169 }
170 
171 int
172 kr_change_fib(struct kroute_node *kr, struct kroute *kroute, int krcount,
173     int action)
174 {
175 	int			 i;
176 	struct kroute_node	*kn, *nkn;
177 
178 	if (action == RTM_ADD) {
179 		/*
180 		 * First remove all stale multipath routes.
181 		 * This step must be skipped when the action is RTM_CHANGE
182 		 * because it is already a single path route that will be
183 		 * changed.
184 		 */
185 		for (kn = kr; kn != NULL; kn = nkn) {
186 			for (i = 0; i < krcount; i++) {
187 				if (kn->r.scope == kroute[i].scope &&
188 				    IN6_ARE_ADDR_EQUAL(&kn->r.nexthop,
189 				    &kroute[i].nexthop))
190 					break;
191 			}
192 			nkn = kn->next;
193 			if (i == krcount) {
194 				/* stale route */
195 				if (kr_delete_fib(kn) == -1)
196 					log_warnx("kr_delete_fib failed");
197 				/*
198 				 * if head element was removed we need to adjust
199 				 * the head
200 				 */
201 				if (kr == kn)
202 					kr = nkn;
203 			}
204 		}
205 	}
206 
207 	/*
208 	 * now add or change the route
209 	 */
210 	for (i = 0; i < krcount; i++) {
211 		/* nexthop ::1 -> ignore silently */
212 		if (IN6_IS_ADDR_LOOPBACK(&kroute[i].nexthop))
213 			continue;
214 
215 		if (action == RTM_ADD && kr) {
216 			for (kn = kr; kn != NULL; kn = kn->next) {
217 				if (kn->r.scope == kroute[i].scope &&
218 				    IN6_ARE_ADDR_EQUAL(&kn->r.nexthop,
219 				    &kroute[i].nexthop))
220 					break;
221 			}
222 
223 			if (kn != NULL)
224 				/* nexthop already present, skip it */
225 				continue;
226 		} else
227 			/* modify first entry */
228 			kn = kr;
229 
230 		/* send update */
231 		if (send_rtmsg(kr_state.fd, action, &kroute[i]) == -1)
232 			return (-1);
233 
234 		/* create new entry unless we are changing the first entry */
235 		if (action == RTM_ADD)
236 			if ((kn = calloc(1, sizeof(*kn))) == NULL)
237 				fatal(NULL);
238 
239 		kn->r.prefix = kroute[i].prefix;
240 		kn->r.prefixlen = kroute[i].prefixlen;
241 		kn->r.nexthop = kroute[i].nexthop;
242 		kn->r.scope = kroute[i].scope;
243 		kn->r.flags = kroute[i].flags | F_OSPFD_INSERTED;
244 		kn->r.priority = kr_state.fib_prio;
245 		kn->r.ext_tag = kroute[i].ext_tag;
246 		rtlabel_unref(kn->r.rtlabel);	/* for RTM_CHANGE */
247 		kn->r.rtlabel = kroute[i].rtlabel;
248 
249 		if (action == RTM_ADD)
250 			if (kroute_insert(kn) == -1) {
251 				log_debug("kr_update_fib: cannot insert %s",
252 				    log_in6addr(&kn->r.nexthop));
253 				free(kn);
254 			}
255 		action = RTM_ADD;
256 	}
257 	return  (0);
258 }
259 
260 int
261 kr_change(struct kroute *kroute, int krcount)
262 {
263 	struct kroute_node	*kr;
264 	int			 action = RTM_ADD;
265 
266 	kroute->rtlabel = rtlabel_tag2id(kroute->ext_tag);
267 
268 	kr = kroute_find(&kroute->prefix, kroute->prefixlen, kr_state.fib_prio);
269 	if (kr != NULL && kr->next == NULL && krcount == 1) {
270 		/*
271 		 * single path OSPF route.
272 		 * The kernel does not allow to change a gateway route to a
273 		 * cloning route or contrary. In this case remove and add the
274 		 * route, otherwise change the existing one.
275 		 */
276 		if ((IN6_IS_ADDR_UNSPECIFIED(&kroute->nexthop) &&
277 		    !IN6_IS_ADDR_UNSPECIFIED(&kr->r.nexthop)) ||
278 		    (!IN6_IS_ADDR_UNSPECIFIED(&kroute->nexthop) &&
279 		    IN6_IS_ADDR_UNSPECIFIED(&kr->r.nexthop))) {
280 			if (kr_delete_fib(kr) == 0)
281 				kr = NULL;
282 			else {
283 				log_warn("kr_change: failed to remove route: "
284 				    "%s/%d", log_in6addr(&kr->r.prefix),
285 				    kr->r.prefixlen);
286 				return (-1);
287 			}
288 		} else
289 			action = RTM_CHANGE;
290 	}
291 
292 	return (kr_change_fib(kr, kroute, krcount, action));
293 }
294 
295 int
296 kr_delete_fib(struct kroute_node *kr)
297 {
298 	if (kr->r.priority != kr_state.fib_prio)
299 		log_warn("kr_delete_fib: %s/%d has wrong priority %d",
300 		    log_in6addr(&kr->r.prefix), kr->r.prefixlen,
301 		    kr->r.priority);
302 
303 	if (send_rtmsg(kr_state.fd, RTM_DELETE, &kr->r) == -1)
304 		return (-1);
305 
306 	if (kroute_remove(kr) == -1)
307 		return (-1);
308 
309 	return (0);
310 }
311 
312 int
313 kr_delete(struct kroute *kroute)
314 {
315 	struct kroute_node	*kr, *nkr;
316 
317 	if ((kr = kroute_find(&kroute->prefix, kroute->prefixlen,
318 	    kr_state.fib_prio)) == NULL)
319 		return (0);
320 
321 	while (kr != NULL) {
322 		nkr = kr->next;
323 		if (kr_delete_fib(kr) == -1)
324 			return (-1);
325 		kr = nkr;
326 	}
327 
328 	return (0);
329 }
330 
331 void
332 kr_shutdown(void)
333 {
334 	kr_fib_decouple();
335 	kroute_clear();
336 }
337 
338 void
339 kr_fib_couple(void)
340 {
341 	struct kroute_node	*kr;
342 	struct kroute_node	*kn;
343 
344 	if (kr_state.fib_sync == 1)	/* already coupled */
345 		return;
346 
347 	kr_state.fib_sync = 1;
348 
349 	RB_FOREACH(kr, kroute_tree, &krt)
350 		if (kr->r.priority == kr_state.fib_prio)
351 			for (kn = kr; kn != NULL; kn = kn->next)
352 				send_rtmsg(kr_state.fd, RTM_ADD, &kn->r);
353 
354 	log_info("kernel routing table coupled");
355 }
356 
357 void
358 kr_fib_decouple(void)
359 {
360 	struct kroute_node	*kr;
361 	struct kroute_node	*kn;
362 
363 	if (kr_state.fib_sync == 0)	/* already decoupled */
364 		return;
365 
366 	RB_FOREACH(kr, kroute_tree, &krt)
367 		if (kr->r.priority == kr_state.fib_prio)
368 			for (kn = kr; kn != NULL; kn = kn->next)
369 				send_rtmsg(kr_state.fd, RTM_DELETE, &kn->r);
370 
371 	kr_state.fib_sync = 0;
372 
373 	log_info("kernel routing table decoupled");
374 }
375 
376 void
377 kr_fib_update_prio(u_int8_t fib_prio)
378 {
379 	struct kroute_node	*kr;
380 
381 	RB_FOREACH(kr, kroute_tree, &krt)
382 		if ((kr->r.flags & F_OSPFD_INSERTED))
383 			kr->r.priority = fib_prio;
384 
385 	log_info("fib priority changed from %hhu to %hhu", kr_state.fib_prio,
386 	    fib_prio);
387 
388 	kr_state.fib_prio = fib_prio;
389 }
390 
391 /* ARGSUSED */
392 void
393 kr_dispatch_msg(int fd, short event, void *bula)
394 {
395 	/* XXX this is stupid */
396 	dispatch_rtmsg();
397 }
398 
399 void
400 kr_show_route(struct imsg *imsg)
401 {
402 	struct kroute_node	*kr;
403 	struct kroute_node	*kn;
404 	int			 flags;
405 	struct in6_addr		 addr;
406 
407 	switch (imsg->hdr.type) {
408 	case IMSG_CTL_KROUTE:
409 		if (imsg->hdr.len != IMSG_HEADER_SIZE + sizeof(flags)) {
410 			log_warnx("kr_show_route: wrong imsg len");
411 			return;
412 		}
413 		memcpy(&flags, imsg->data, sizeof(flags));
414 		RB_FOREACH(kr, kroute_tree, &krt)
415 			if (!flags || kr->r.flags & flags) {
416 				kn = kr;
417 				do {
418 					main_imsg_compose_ospfe(IMSG_CTL_KROUTE,
419 					    imsg->hdr.pid,
420 					    &kn->r, sizeof(kn->r));
421 				} while ((kn = kn->next) != NULL);
422 			}
423 		break;
424 	case IMSG_CTL_KROUTE_ADDR:
425 		if (imsg->hdr.len != IMSG_HEADER_SIZE +
426 		    sizeof(struct in6_addr)) {
427 			log_warnx("kr_show_route: wrong imsg len");
428 			return;
429 		}
430 		memcpy(&addr, imsg->data, sizeof(addr));
431 		kr = kroute_match(&addr);
432 		if (kr != NULL)
433 			main_imsg_compose_ospfe(IMSG_CTL_KROUTE, imsg->hdr.pid,
434 			    &kr->r, sizeof(kr->r));
435 		break;
436 	default:
437 		log_debug("kr_show_route: error handling imsg");
438 		break;
439 	}
440 
441 	main_imsg_compose_ospfe(IMSG_CTL_END, imsg->hdr.pid, NULL, 0);
442 }
443 
444 void
445 kr_redist_remove(struct kroute_node *kh, struct kroute_node *kn)
446 {
447 	struct kroute	 *kr;
448 
449 	/* was the route redistributed? */
450 	if ((kn->r.flags & F_REDISTRIBUTED) == 0)
451 		return;
452 
453 	/* remove redistributed flag */
454 	kn->r.flags &= ~F_REDISTRIBUTED;
455 	kr = &kn->r;
456 
457 	/* probably inform the RDE (check if no other path is redistributed) */
458 	for (kn = kh; kn; kn = kn->next)
459 		if (kn->r.flags & F_REDISTRIBUTED)
460 			break;
461 
462 	if (kn == NULL)
463 		main_imsg_compose_rde(IMSG_NETWORK_DEL, 0, kr,
464 		    sizeof(struct kroute));
465 }
466 
467 int
468 kr_redist_eval(struct kroute *kr, struct kroute *new_kr)
469 {
470 	u_int32_t	 metric = 0;
471 
472 	/* Only non-ospfd routes are considered for redistribution. */
473 	if (!(kr->flags & F_KERNEL))
474 		goto dont_redistribute;
475 
476 	/* Dynamic routes are not redistributable. */
477 	if (kr->flags & F_DYNAMIC)
478 		goto dont_redistribute;
479 
480 	/* interface is not up and running so don't announce */
481 	if (kr->flags & F_DOWN)
482 		goto dont_redistribute;
483 
484 	/*
485 	 * We consider loopback, multicast, link- and site-local,
486 	 * IPv4 mapped and IPv4 compatible addresses as not redistributable.
487 	 */
488 	if (IN6_IS_ADDR_LOOPBACK(&kr->prefix) ||
489 	    IN6_IS_ADDR_MULTICAST(&kr->prefix) ||
490 	    IN6_IS_ADDR_LINKLOCAL(&kr->prefix) ||
491 	    IN6_IS_ADDR_SITELOCAL(&kr->prefix) ||
492 	    IN6_IS_ADDR_V4MAPPED(&kr->prefix) ||
493 	    IN6_IS_ADDR_V4COMPAT(&kr->prefix))
494 		goto dont_redistribute;
495 	/*
496 	 * Consider networks with nexthop loopback as not redistributable
497 	 * unless it is a reject or blackhole route.
498 	 */
499 	if (IN6_IS_ADDR_LOOPBACK(&kr->nexthop) &&
500 	    !(kr->flags & (F_BLACKHOLE|F_REJECT)))
501 		goto dont_redistribute;
502 
503 	/* Should we redistribute this route? */
504 	if (!ospf_redistribute(kr, &metric))
505 		goto dont_redistribute;
506 
507 	/* prefix should be redistributed */
508 	kr->flags |= F_REDISTRIBUTED;
509 	/*
510 	 * only one of all multipath routes can be redistributed so
511 	 * redistribute the best one.
512 	 */
513 	if (new_kr->metric > metric) {
514 		*new_kr = *kr;
515 		new_kr->metric = metric;
516 	}
517 
518 	return (1);
519 
520 dont_redistribute:
521 	/* was the route redistributed? */
522 	if ((kr->flags & F_REDISTRIBUTED) == 0)
523 		return (0);
524 
525 	kr->flags &= ~F_REDISTRIBUTED;
526 	return (1);
527 }
528 
529 void
530 kr_redistribute(struct kroute_node *kh)
531 {
532 	struct kroute_node	*kn;
533 	struct kroute		 kr;
534 	int			 redistribute = 0;
535 
536 	/* only the highest prio route can be redistributed */
537 	if (kroute_find(&kh->r.prefix, kh->r.prefixlen, RTP_ANY) != kh)
538 		return;
539 
540 	bzero(&kr, sizeof(kr));
541 	kr.metric = UINT_MAX;
542 	for (kn = kh; kn; kn = kn->next)
543 		if (kr_redist_eval(&kn->r, &kr))
544 			redistribute = 1;
545 
546 	if (!redistribute)
547 		return;
548 
549 	if (kr.flags & F_REDISTRIBUTED) {
550 		main_imsg_compose_rde(IMSG_NETWORK_ADD, 0, &kr,
551 		    sizeof(struct kroute));
552 	} else {
553 		kr = kh->r;
554 		main_imsg_compose_rde(IMSG_NETWORK_DEL, 0, &kr,
555 		    sizeof(struct kroute));
556 	}
557 }
558 
559 void
560 kr_reload(int redis_label_or_prefix)
561 {
562 	struct kroute_node	*kr, *kn;
563 	u_int32_t		 dummy;
564 	int			 r;
565 	int			 filter_prio = kr_state.fib_prio;
566 
567 	/* update the priority filter */
568 	if (redis_label_or_prefix) {
569 		filter_prio = 0;
570 		log_info("%s: priority filter disabled", __func__);
571 	} else
572 		log_debug("%s: priority filter enabled", __func__);
573 
574 	if (setsockopt(kr_state.fd, AF_ROUTE, ROUTE_PRIOFILTER, &filter_prio,
575 	    sizeof(filter_prio)) == -1) {
576 		log_warn("%s: setsockopt AF_ROUTE ROUTE_PRIOFILTER", __func__);
577 		/* not fatal */
578 	}
579 
580 	RB_FOREACH(kr, kroute_tree, &krt) {
581 		for (kn = kr; kn; kn = kn->next) {
582 			r = ospf_redistribute(&kn->r, &dummy);
583 			/*
584 			 * if it is redistributed, redistribute again metric
585 			 * may have changed.
586 			 */
587 			if ((kn->r.flags & F_REDISTRIBUTED && !r) || r)
588 				break;
589 		}
590 		if (kn) {
591 			/*
592 			 * kr_redistribute copes with removes and RDE with
593 			 * duplicates
594 			 */
595 			kr_redistribute(kr);
596 		}
597 	}
598 }
599 
600 /* rb-tree compare */
601 int
602 kroute_compare(struct kroute_node *a, struct kroute_node *b)
603 {
604 	int	i;
605 
606 	/* XXX maybe switch a & b */
607 	i = memcmp(&a->r.prefix, &b->r.prefix, sizeof(a->r.prefix));
608 	if (i)
609 		return (i);
610 	if (a->r.prefixlen < b->r.prefixlen)
611 		return (-1);
612 	if (a->r.prefixlen > b->r.prefixlen)
613 		return (1);
614 
615 	/* if the priority is RTP_ANY finish on the first address hit */
616 	if (a->r.priority == RTP_ANY || b->r.priority == RTP_ANY)
617 		return (0);
618 	if (a->r.priority < b->r.priority)
619 		return (-1);
620 	if (a->r.priority > b->r.priority)
621 		return (1);
622 	return (0);
623 }
624 
625 /* tree management */
626 struct kroute_node *
627 kroute_find(const struct in6_addr *prefix, u_int8_t prefixlen, u_int8_t prio)
628 {
629 	struct kroute_node	s;
630 	struct kroute_node	*kn, *tmp;
631 
632 	s.r.prefix = *prefix;
633 	s.r.prefixlen = prefixlen;
634 	s.r.priority = prio;
635 
636 	kn = RB_FIND(kroute_tree, &krt, &s);
637 	if (kn && prio == RTP_ANY) {
638 		tmp = RB_PREV(kroute_tree, &krt, kn);
639 		while (tmp) {
640 			if (kroute_compare(&s, tmp) == 0)
641 				kn = tmp;
642 			else
643 				break;
644 			tmp = RB_PREV(kroute_tree, &krt, kn);
645 		}
646 	}
647 	return (kn);
648 }
649 
650 struct kroute_node *
651 kroute_matchgw(struct kroute_node *kr, struct in6_addr *nh, unsigned int scope)
652 {
653 	while (kr) {
654 		if (scope == kr->r.scope &&
655 		    IN6_ARE_ADDR_EQUAL(&kr->r.nexthop, nh))
656 			return (kr);
657 		kr = kr->next;
658 	}
659 
660 	return (NULL);
661 }
662 
663 int
664 kroute_insert(struct kroute_node *kr)
665 {
666 	struct kroute_node	*krm, *krh;
667 
668 	if ((krh = RB_INSERT(kroute_tree, &krt, kr)) != NULL) {
669 		/*
670 		 * Multipath route, add at end of list.
671 		 */
672 		krm = krh;
673 		while (krm->next != NULL)
674 			krm = krm->next;
675 		krm->next = kr;
676 		kr->next = NULL; /* to be sure */
677 	} else
678 		krh = kr;
679 
680 	if (!(kr->r.flags & F_KERNEL)) {
681 		/* don't validate or redistribute ospf route */
682 		kr->r.flags &= ~F_DOWN;
683 		return (0);
684 	}
685 
686 	if (kif_validate(kr->r.ifindex))
687 		kr->r.flags &= ~F_DOWN;
688 	else
689 		kr->r.flags |= F_DOWN;
690 
691 	kr_redistribute(krh);
692 	return (0);
693 }
694 
695 int
696 kroute_remove(struct kroute_node *kr)
697 {
698 	struct kroute_node	*krm;
699 
700 	if ((krm = RB_FIND(kroute_tree, &krt, kr)) == NULL) {
701 		log_warnx("kroute_remove failed to find %s/%u",
702 		    log_in6addr(&kr->r.prefix), kr->r.prefixlen);
703 		return (-1);
704 	}
705 
706 	if (krm == kr) {
707 		/* head element */
708 		if (RB_REMOVE(kroute_tree, &krt, kr) == NULL) {
709 			log_warnx("kroute_remove failed for %s/%u",
710 			    log_in6addr(&kr->r.prefix), kr->r.prefixlen);
711 			return (-1);
712 		}
713 		if (kr->next != NULL) {
714 			if (RB_INSERT(kroute_tree, &krt, kr->next) != NULL) {
715 				log_warnx("kroute_remove failed to add %s/%u",
716 				    log_in6addr(&kr->r.prefix),
717 				    kr->r.prefixlen);
718 				return (-1);
719 			}
720 		}
721 	} else {
722 		/* somewhere in the list */
723 		while (krm->next != kr && krm->next != NULL)
724 			krm = krm->next;
725 		if (krm->next == NULL) {
726 			log_warnx("kroute_remove multipath list corrupted "
727 			    "for %s/%u", log_in6addr(&kr->r.prefix),
728 			    kr->r.prefixlen);
729 			return (-1);
730 		}
731 		krm->next = kr->next;
732 	}
733 
734 	kr_redist_remove(krm, kr);
735 	rtlabel_unref(kr->r.rtlabel);
736 
737 	free(kr);
738 	return (0);
739 }
740 
741 void
742 kroute_clear(void)
743 {
744 	struct kroute_node	*kr;
745 
746 	while ((kr = RB_MIN(kroute_tree, &krt)) != NULL)
747 		kroute_remove(kr);
748 }
749 
750 struct iface *
751 kif_update(u_short ifindex, int flags, struct if_data *ifd,
752     struct sockaddr_dl *sdl)
753 {
754 	struct iface	*iface;
755 	char		 ifname[IF_NAMESIZE];
756 
757 	if ((iface = if_find(ifindex)) == NULL) {
758 		bzero(ifname, sizeof(ifname));
759 		if (sdl && sdl->sdl_family == AF_LINK) {
760 			if (sdl->sdl_nlen >= sizeof(ifname))
761 				memcpy(ifname, sdl->sdl_data,
762 				    sizeof(ifname) - 1);
763 			else if (sdl->sdl_nlen > 0)
764 				memcpy(ifname, sdl->sdl_data, sdl->sdl_nlen);
765 			else
766 				return (NULL);
767 		} else
768 			return (NULL);
769 		if ((iface = if_new(ifindex, ifname)) == NULL)
770 			return (NULL);
771 	}
772 
773 	if_update(iface, ifd->ifi_mtu, flags, ifd->ifi_type,
774 	    ifd->ifi_link_state, ifd->ifi_baudrate, ifd->ifi_rdomain);
775 
776 	return (iface);
777 }
778 
779 int
780 kif_validate(u_short ifindex)
781 {
782 	struct iface	*iface;
783 
784 	if ((iface = if_find(ifindex)) == NULL) {
785 		log_warnx("interface with index %u not found", ifindex);
786 		return (-1);
787 	}
788 
789 	return ((iface->flags & IFF_UP) && LINK_STATE_IS_UP(iface->linkstate));
790 }
791 
792 struct kroute_node *
793 kroute_match(struct in6_addr *key)
794 {
795 	int			 i;
796 	struct kroute_node	*kr;
797 	struct in6_addr		 ina;
798 
799 	/* we will never match the default route */
800 	for (i = 128; i > 0; i--) {
801 		inet6applymask(&ina, key, i);
802 		if ((kr = kroute_find(&ina, i, RTP_ANY)) != NULL)
803 			return (kr);
804 	}
805 
806 	/* if we don't have a match yet, try to find a default route */
807 	if ((kr = kroute_find(&in6addr_any, 0, RTP_ANY)) != NULL)
808 			return (kr);
809 
810 	return (NULL);
811 }
812 
813 /* misc */
814 int
815 protect_lo(void)
816 {
817 	struct kroute_node	*kr;
818 
819 	/* special protection for loopback */
820 	if ((kr = calloc(1, sizeof(struct kroute_node))) == NULL) {
821 		log_warn("protect_lo");
822 		return (-1);
823 	}
824 	memcpy(&kr->r.prefix, &in6addr_loopback, sizeof(kr->r.prefix));
825 	kr->r.prefixlen = 128;
826 	kr->r.flags = F_KERNEL|F_CONNECTED;
827 
828 	if (RB_INSERT(kroute_tree, &krt, kr) != NULL)
829 		free(kr);	/* kernel route already there, no problem */
830 
831 	return (0);
832 }
833 
834 #define ROUNDUP(a) \
835 	((a) > 0 ? (1 + (((a) - 1) | (sizeof(long) - 1))) : sizeof(long))
836 
837 void
838 get_rtaddrs(int addrs, struct sockaddr *sa, struct sockaddr **rti_info)
839 {
840 	int	i;
841 
842 	for (i = 0; i < RTAX_MAX; i++) {
843 		if (addrs & (1 << i)) {
844 			rti_info[i] = sa;
845 			sa = (struct sockaddr *)((char *)(sa) +
846 			    ROUNDUP(sa->sa_len));
847 		} else
848 			rti_info[i] = NULL;
849 	}
850 }
851 
852 void
853 if_change(u_short ifindex, int flags, struct if_data *ifd,
854     struct sockaddr_dl *sdl)
855 {
856 	struct kroute_node	*kr, *tkr;
857 	struct iface		*iface;
858 	u_int8_t		 wasvalid, isvalid;
859 
860 	wasvalid = kif_validate(ifindex);
861 
862 	if ((iface = kif_update(ifindex, flags, ifd, sdl)) == NULL) {
863 		log_warn("if_change: kif_update(%u)", ifindex);
864 		return;
865 	}
866 
867 	/* inform engine and rde about state change */
868 	main_imsg_compose_rde(IMSG_IFINFO, 0, iface, sizeof(struct iface));
869 	main_imsg_compose_ospfe(IMSG_IFINFO, 0, iface, sizeof(struct iface));
870 
871 	isvalid = (iface->flags & IFF_UP) &&
872 	    LINK_STATE_IS_UP(iface->linkstate);
873 
874 	if (wasvalid == isvalid)
875 		return;		/* nothing changed wrt validity */
876 
877 	/* update redistribute list */
878 	RB_FOREACH(kr, kroute_tree, &krt) {
879 		for (tkr = kr; tkr != NULL; tkr = tkr->next) {
880 			if (tkr->r.ifindex == ifindex) {
881 				if (isvalid)
882 					tkr->r.flags &= ~F_DOWN;
883 				else
884 					tkr->r.flags |= F_DOWN;
885 
886 			}
887 		}
888 		kr_redistribute(kr);
889 	}
890 }
891 
892 void
893 if_newaddr(u_short ifindex, struct sockaddr_in6 *ifa, struct sockaddr_in6 *mask,
894     struct sockaddr_in6 *brd)
895 {
896 	struct iface		*iface;
897 	struct iface_addr	*ia;
898 	struct ifaddrchange	 ifc;
899 
900 	if (ifa == NULL || ifa->sin6_family != AF_INET6)
901 		return;
902 	if ((iface = if_find(ifindex)) == NULL) {
903 		log_warnx("if_newaddr: corresponding if %d not found", ifindex);
904 		return;
905 	}
906 
907 	/* We only care about link-local and global-scope. */
908 	if (IN6_IS_ADDR_UNSPECIFIED(&ifa->sin6_addr) ||
909 	    IN6_IS_ADDR_LOOPBACK(&ifa->sin6_addr) ||
910 	    IN6_IS_ADDR_MULTICAST(&ifa->sin6_addr) ||
911 	    IN6_IS_ADDR_SITELOCAL(&ifa->sin6_addr) ||
912 	    IN6_IS_ADDR_V4MAPPED(&ifa->sin6_addr) ||
913 	    IN6_IS_ADDR_V4COMPAT(&ifa->sin6_addr))
914 		return;
915 
916 	clearscope(&ifa->sin6_addr);
917 
918 	if (IN6_IS_ADDR_LINKLOCAL(&ifa->sin6_addr) ||
919 	    iface->flags & IFF_LOOPBACK)
920 		iface->addr = ifa->sin6_addr;
921 
922 	if ((ia = calloc(1, sizeof(struct iface_addr))) == NULL)
923 		fatal("if_newaddr");
924 
925 	ia->addr = ifa->sin6_addr;
926 
927 	if (mask)
928 		ia->prefixlen = mask2prefixlen(mask);
929 	else
930 		ia->prefixlen = 0;
931 	if (brd && brd->sin6_family == AF_INET6)
932 		ia->dstbrd = brd->sin6_addr;
933 	else
934 		bzero(&ia->dstbrd, sizeof(ia->dstbrd));
935 
936 	switch (iface->type) {
937 	case IF_TYPE_BROADCAST:
938 	case IF_TYPE_NBMA:
939 		log_debug("if_newaddr: ifindex %u, addr %s/%d",
940 		    ifindex, log_in6addr(&ia->addr), ia->prefixlen);
941 		break;
942 	case IF_TYPE_VIRTUALLINK:	/* FIXME */
943 		break;
944 	case IF_TYPE_POINTOPOINT:
945 	case IF_TYPE_POINTOMULTIPOINT:
946 		log_debug("if_newaddr: ifindex %u, addr %s/%d, "
947 		    "dest %s", ifindex, log_in6addr(&ia->addr),
948 		    ia->prefixlen, log_in6addr(&ia->dstbrd));
949 		break;
950 	default:
951 		fatalx("if_newaddr: unknown interface type");
952 	}
953 
954 	TAILQ_INSERT_TAIL(&iface->ifa_list, ia, entry);
955 	/* inform engine and rde if interface is used */
956 	if (iface->cflags & F_IFACE_CONFIGURED) {
957 		ifc.addr = ia->addr;
958 		ifc.dstbrd = ia->dstbrd;
959 		ifc.prefixlen = ia->prefixlen;
960 		ifc.ifindex = ifindex;
961 		main_imsg_compose_ospfe(IMSG_IFADDRNEW, 0, &ifc, sizeof(ifc));
962 		main_imsg_compose_rde(IMSG_IFADDRNEW, 0, &ifc, sizeof(ifc));
963 	}
964 }
965 
966 void
967 if_deladdr(u_short ifindex, struct sockaddr_in6 *ifa, struct sockaddr_in6 *mask,
968     struct sockaddr_in6 *brd)
969 {
970 	struct iface		*iface;
971 	struct iface_addr	*ia, *nia;
972 	struct ifaddrchange	 ifc;
973 
974 	if (ifa == NULL || ifa->sin6_family != AF_INET6)
975 		return;
976 	if ((iface = if_find(ifindex)) == NULL) {
977 		log_warnx("if_deladdr: corresponding if %d not found", ifindex);
978 		return;
979 	}
980 
981 	/* We only care about link-local and global-scope. */
982 	if (IN6_IS_ADDR_UNSPECIFIED(&ifa->sin6_addr) ||
983 	    IN6_IS_ADDR_LOOPBACK(&ifa->sin6_addr) ||
984 	    IN6_IS_ADDR_MULTICAST(&ifa->sin6_addr) ||
985 	    IN6_IS_ADDR_SITELOCAL(&ifa->sin6_addr) ||
986 	    IN6_IS_ADDR_V4MAPPED(&ifa->sin6_addr) ||
987 	    IN6_IS_ADDR_V4COMPAT(&ifa->sin6_addr))
988 		return;
989 
990 	clearscope(&ifa->sin6_addr);
991 
992 	for (ia = TAILQ_FIRST(&iface->ifa_list); ia != NULL; ia = nia) {
993 		nia = TAILQ_NEXT(ia, entry);
994 
995 		if (IN6_ARE_ADDR_EQUAL(&ia->addr, &ifa->sin6_addr)) {
996 			log_debug("if_deladdr: ifindex %u, addr %s/%d",
997 			    ifindex, log_in6addr(&ia->addr), ia->prefixlen);
998 			TAILQ_REMOVE(&iface->ifa_list, ia, entry);
999 			/* inform engine and rde if interface is used */
1000 			if (iface->cflags & F_IFACE_CONFIGURED) {
1001 				ifc.addr = ia->addr;
1002 				ifc.dstbrd = ia->dstbrd;
1003 				ifc.prefixlen = ia->prefixlen;
1004 				ifc.ifindex = ifindex;
1005 				main_imsg_compose_ospfe(IMSG_IFADDRDEL, 0, &ifc,
1006 				    sizeof(ifc));
1007 				main_imsg_compose_rde(IMSG_IFADDRDEL, 0, &ifc,
1008 				    sizeof(ifc));
1009 			}
1010 			free(ia);
1011 			return;
1012 		}
1013 	}
1014 }
1015 
1016 void
1017 if_announce(void *msg)
1018 {
1019 	struct if_announcemsghdr	*ifan;
1020 	struct iface			*iface;
1021 
1022 	ifan = msg;
1023 
1024 	switch (ifan->ifan_what) {
1025 	case IFAN_ARRIVAL:
1026 		if ((iface = if_new(ifan->ifan_index, ifan->ifan_name)) == NULL)
1027 			fatal("if_announce failed");
1028 		break;
1029 	case IFAN_DEPARTURE:
1030 		iface = if_find(ifan->ifan_index);
1031 		if_del(iface);
1032 		break;
1033 	}
1034 }
1035 
1036 /* rtsock */
1037 int
1038 send_rtmsg(int fd, int action, struct kroute *kroute)
1039 {
1040 	struct iovec		iov[5];
1041 	struct rt_msghdr	hdr;
1042 	struct pad {
1043 		struct sockaddr_in6	addr;
1044 		char			pad[sizeof(long)]; /* thank you IPv6 */
1045 	} prefix, nexthop, mask;
1046 	struct {
1047 		struct sockaddr_dl	addr;
1048 		char			pad[sizeof(long)];
1049 	} ifp;
1050 	struct sockaddr_rtlabel	sa_rl;
1051 	int			iovcnt = 0;
1052 	const char		*label;
1053 
1054 	if (kr_state.fib_sync == 0)
1055 		return (0);
1056 
1057 	/* initialize header */
1058 	bzero(&hdr, sizeof(hdr));
1059 	hdr.rtm_version = RTM_VERSION;
1060 	hdr.rtm_type = action;
1061 	hdr.rtm_priority = kr_state.fib_prio;
1062 	hdr.rtm_tableid = kr_state.rdomain;	/* rtableid */
1063 	if (action == RTM_CHANGE)
1064 		hdr.rtm_fmask = RTF_REJECT|RTF_BLACKHOLE;
1065 	else
1066 		hdr.rtm_flags = RTF_MPATH;
1067 	hdr.rtm_seq = kr_state.rtseq++;	/* overflow doesn't matter */
1068 	hdr.rtm_hdrlen = sizeof(hdr);
1069 	hdr.rtm_msglen = sizeof(hdr);
1070 	/* adjust iovec */
1071 	iov[iovcnt].iov_base = &hdr;
1072 	iov[iovcnt++].iov_len = sizeof(hdr);
1073 
1074 	bzero(&prefix, sizeof(prefix));
1075 	prefix.addr.sin6_len = sizeof(struct sockaddr_in6);
1076 	prefix.addr.sin6_family = AF_INET6;
1077 	prefix.addr.sin6_addr = kroute->prefix;
1078 	/* adjust header */
1079 	hdr.rtm_addrs |= RTA_DST;
1080 	hdr.rtm_msglen += ROUNDUP(sizeof(struct sockaddr_in6));
1081 	/* adjust iovec */
1082 	iov[iovcnt].iov_base = &prefix;
1083 	iov[iovcnt++].iov_len = ROUNDUP(sizeof(struct sockaddr_in6));
1084 
1085 	if (!IN6_IS_ADDR_UNSPECIFIED(&kroute->nexthop)) {
1086 		bzero(&nexthop, sizeof(nexthop));
1087 		nexthop.addr.sin6_len = sizeof(struct sockaddr_in6);
1088 		nexthop.addr.sin6_family = AF_INET6;
1089 		nexthop.addr.sin6_addr = kroute->nexthop;
1090 		nexthop.addr.sin6_scope_id = kroute->scope;
1091 		/*
1092 		 * XXX we should set the sin6_scope_id but the kernel
1093 		 * XXX does not expect it that way. It must be fiddled
1094 		 * XXX into the sin6_addr. Welcome to the typical
1095 		 * XXX IPv6 insanity and all without wine bottles.
1096 		 */
1097 		embedscope(&nexthop.addr);
1098 
1099 		/* adjust header */
1100 		hdr.rtm_flags |= RTF_GATEWAY;
1101 		hdr.rtm_addrs |= RTA_GATEWAY;
1102 		hdr.rtm_msglen += ROUNDUP(sizeof(struct sockaddr_in6));
1103 		/* adjust iovec */
1104 		iov[iovcnt].iov_base = &nexthop;
1105 		iov[iovcnt++].iov_len = ROUNDUP(sizeof(struct sockaddr_in6));
1106 	} else if (kroute->ifindex) {
1107 		/*
1108 		 * We don't have an interface address in that network,
1109 		 * so we install a cloning route.  The kernel will then
1110 		 * do neigbor discovery.
1111 		 */
1112 		bzero(&ifp, sizeof(ifp));
1113 		ifp.addr.sdl_len = sizeof(struct sockaddr_dl);
1114 		ifp.addr.sdl_family = AF_LINK;
1115 
1116 		ifp.addr.sdl_index  = kroute->ifindex;
1117 		/* adjust header */
1118 		hdr.rtm_flags |= RTF_CLONING;
1119 		hdr.rtm_addrs |= RTA_GATEWAY;
1120 		hdr.rtm_msglen += ROUNDUP(sizeof(struct sockaddr_dl));
1121 		/* adjust iovec */
1122 		iov[iovcnt].iov_base = &ifp;
1123 		iov[iovcnt++].iov_len = ROUNDUP(sizeof(struct sockaddr_dl));
1124 	}
1125 
1126 	bzero(&mask, sizeof(mask));
1127 	mask.addr.sin6_len = sizeof(struct sockaddr_in6);
1128 	mask.addr.sin6_family = AF_INET6;
1129 	mask.addr.sin6_addr = *prefixlen2mask(kroute->prefixlen);
1130 	/* adjust header */
1131 	if (kroute->prefixlen == 128)
1132 		hdr.rtm_flags |= RTF_HOST;
1133 	hdr.rtm_addrs |= RTA_NETMASK;
1134 	hdr.rtm_msglen += ROUNDUP(sizeof(struct sockaddr_in6));
1135 	/* adjust iovec */
1136 	iov[iovcnt].iov_base = &mask;
1137 	iov[iovcnt++].iov_len = ROUNDUP(sizeof(struct sockaddr_in6));
1138 
1139 	if (kroute->rtlabel != 0) {
1140 		sa_rl.sr_len = sizeof(sa_rl);
1141 		sa_rl.sr_family = AF_UNSPEC;
1142 		label = rtlabel_id2name(kroute->rtlabel);
1143 		if (strlcpy(sa_rl.sr_label, label,
1144 		    sizeof(sa_rl.sr_label)) >= sizeof(sa_rl.sr_label)) {
1145 			log_warnx("send_rtmsg: invalid rtlabel");
1146 			return (-1);
1147 		}
1148 		/* adjust header */
1149 		hdr.rtm_addrs |= RTA_LABEL;
1150 		hdr.rtm_msglen += sizeof(sa_rl);
1151 		/* adjust iovec */
1152 		iov[iovcnt].iov_base = &sa_rl;
1153 		iov[iovcnt++].iov_len = sizeof(sa_rl);
1154 	}
1155 
1156 retry:
1157 	if (writev(fd, iov, iovcnt) == -1) {
1158 		if (errno == ESRCH) {
1159 			if (hdr.rtm_type == RTM_CHANGE) {
1160 				hdr.rtm_type = RTM_ADD;
1161 				goto retry;
1162 			} else if (hdr.rtm_type == RTM_DELETE) {
1163 				log_info("route %s/%u vanished before delete",
1164 				    log_sockaddr(&prefix), kroute->prefixlen);
1165 				return (0);
1166 			}
1167 		}
1168 		log_warn("send_rtmsg: action %u, prefix %s/%u", hdr.rtm_type,
1169 		    log_sockaddr(&prefix), kroute->prefixlen);
1170 		return (0);
1171 	}
1172 
1173 	return (0);
1174 }
1175 
1176 int
1177 fetchtable(void)
1178 {
1179 	size_t			 len;
1180 	int			 mib[7];
1181 	char			*buf;
1182 	int			 rv;
1183 
1184 	mib[0] = CTL_NET;
1185 	mib[1] = PF_ROUTE;
1186 	mib[2] = 0;
1187 	mib[3] = AF_INET6;
1188 	mib[4] = NET_RT_DUMP;
1189 	mib[5] = 0;
1190 	mib[6] = kr_state.rdomain;	/* rtableid */
1191 
1192 	if (sysctl(mib, 7, NULL, &len, NULL, 0) == -1) {
1193 		log_warn("sysctl");
1194 		return (-1);
1195 	}
1196 	if ((buf = malloc(len)) == NULL) {
1197 		log_warn("fetchtable");
1198 		return (-1);
1199 	}
1200 	if (sysctl(mib, 7, buf, &len, NULL, 0) == -1) {
1201 		log_warn("sysctl");
1202 		free(buf);
1203 		return (-1);
1204 	}
1205 
1206 	rv = rtmsg_process(buf, len);
1207 	free(buf);
1208 
1209 	return (rv);
1210 }
1211 
1212 int
1213 fetchifs(u_short ifindex)
1214 {
1215 	size_t			 len;
1216 	int			 mib[6];
1217 	char			*buf;
1218 	int			 rv;
1219 
1220 	mib[0] = CTL_NET;
1221 	mib[1] = PF_ROUTE;
1222 	mib[2] = 0;
1223 	mib[3] = AF_INET6;
1224 	mib[4] = NET_RT_IFLIST;
1225 	mib[5] = ifindex;
1226 
1227 	if (sysctl(mib, 6, NULL, &len, NULL, 0) == -1) {
1228 		log_warn("sysctl");
1229 		return (-1);
1230 	}
1231 	if ((buf = malloc(len)) == NULL) {
1232 		log_warn("fetchifs");
1233 		return (-1);
1234 	}
1235 	if (sysctl(mib, 6, buf, &len, NULL, 0) == -1) {
1236 		log_warn("sysctl");
1237 		free(buf);
1238 		return (-1);
1239 	}
1240 
1241 	rv = rtmsg_process(buf, len);
1242 	free(buf);
1243 
1244 	return (rv);
1245 }
1246 
1247 int
1248 dispatch_rtmsg(void)
1249 {
1250 	char			 buf[RT_BUF_SIZE];
1251 	ssize_t			 n;
1252 
1253 	if ((n = read(kr_state.fd, &buf, sizeof(buf))) == -1) {
1254 		if (errno == EAGAIN || errno == EINTR)
1255 			return (0);
1256 		log_warn("dispatch_rtmsg: read error");
1257 		return (-1);
1258 	}
1259 
1260 	if (n == 0) {
1261 		log_warnx("routing socket closed");
1262 		return (-1);
1263 	}
1264 
1265 	return (rtmsg_process(buf, n));
1266 }
1267 
1268 int
1269 rtmsg_process(char *buf, size_t len)
1270 {
1271 	struct rt_msghdr	*rtm;
1272 	struct if_msghdr	 ifm;
1273 	struct ifa_msghdr	*ifam;
1274 	struct sockaddr		*sa, *rti_info[RTAX_MAX];
1275 	struct sockaddr_in6	*sa_in6;
1276 	struct sockaddr_rtlabel	*label;
1277 	struct kroute_node	*kr, *okr;
1278 	struct in6_addr		 prefix, nexthop;
1279 	u_int8_t		 prefixlen, prio;
1280 	int			 flags, mpath;
1281 	unsigned int		 scope;
1282 	u_short			 ifindex = 0;
1283 	int			 rv;
1284 	size_t			 offset;
1285 	char			*next;
1286 
1287 	for (offset = 0; offset < len; offset += rtm->rtm_msglen) {
1288 		next = buf + offset;
1289 		rtm = (struct rt_msghdr *)next;
1290 		if (len < offset + sizeof(u_short) ||
1291 		    len < offset + rtm->rtm_msglen)
1292 			fatalx("rtmsg_process: partial rtm in buffer");
1293 		if (rtm->rtm_version != RTM_VERSION)
1294 			continue;
1295 
1296 		bzero(&prefix, sizeof(prefix));
1297 		bzero(&nexthop, sizeof(nexthop));
1298 		scope = 0;
1299 		prefixlen = 0;
1300 		flags = F_KERNEL;
1301 		mpath = 0;
1302 		prio = 0;
1303 
1304 		sa = (struct sockaddr *)(next + rtm->rtm_hdrlen);
1305 		get_rtaddrs(rtm->rtm_addrs, sa, rti_info);
1306 
1307 		switch (rtm->rtm_type) {
1308 		case RTM_ADD:
1309 		case RTM_GET:
1310 		case RTM_CHANGE:
1311 		case RTM_DELETE:
1312 			if (rtm->rtm_errno)		/* failed attempts... */
1313 				continue;
1314 
1315 			if (rtm->rtm_tableid != kr_state.rdomain)
1316 				continue;
1317 
1318 			if (rtm->rtm_type == RTM_GET &&
1319 			    rtm->rtm_pid != kr_state.pid) /* caused by us */
1320 				continue;
1321 
1322 			if ((sa = rti_info[RTAX_DST]) == NULL)
1323 				continue;
1324 
1325 			/* Skip ARP/ND cache and broadcast routes. */
1326 			if (rtm->rtm_flags & (RTF_LLINFO|RTF_BROADCAST))
1327 				continue;
1328 
1329 			if (rtm->rtm_flags & RTF_MPATH)
1330 				mpath = 1;
1331 			prio = rtm->rtm_priority;
1332 			flags = (prio == kr_state.fib_prio) ?
1333 			    F_OSPFD_INSERTED : F_KERNEL;
1334 
1335 			switch (sa->sa_family) {
1336 			case AF_INET6:
1337 				prefix =
1338 				    ((struct sockaddr_in6 *)sa)->sin6_addr;
1339 				sa_in6 = (struct sockaddr_in6 *)
1340 				    rti_info[RTAX_NETMASK];
1341 				if (sa_in6 != NULL) {
1342 					if (sa_in6->sin6_len != 0)
1343 						prefixlen = mask2prefixlen(
1344 						    sa_in6);
1345 				} else if (rtm->rtm_flags & RTF_HOST)
1346 					prefixlen = 128;
1347 				else
1348 					fatalx("classful IPv6 address?!!");
1349 				if (rtm->rtm_flags & RTF_STATIC)
1350 					flags |= F_STATIC;
1351 				if (rtm->rtm_flags & RTF_BLACKHOLE)
1352 					flags |= F_BLACKHOLE;
1353 				if (rtm->rtm_flags & RTF_REJECT)
1354 					flags |= F_REJECT;
1355 				if (rtm->rtm_flags & RTF_DYNAMIC)
1356 					flags |= F_DYNAMIC;
1357 				break;
1358 			default:
1359 				continue;
1360 			}
1361 
1362 			ifindex = rtm->rtm_index;
1363 			if ((sa = rti_info[RTAX_GATEWAY]) != NULL) {
1364 				switch (sa->sa_family) {
1365 				case AF_INET6:
1366 					if (rtm->rtm_flags & RTF_CONNECTED)
1367 						flags |= F_CONNECTED;
1368 
1369 					sa_in6 = (struct sockaddr_in6 *)sa;
1370 					/*
1371 					 * XXX The kernel provides the scope
1372 					 * XXX via the kame hack instead of
1373 					 * XXX the scope_id field.
1374 					 */
1375 					recoverscope(sa_in6);
1376 					nexthop = sa_in6->sin6_addr;
1377 					scope = sa_in6->sin6_scope_id;
1378 					break;
1379 				case AF_LINK:
1380 					flags |= F_CONNECTED;
1381 					break;
1382 				}
1383 			}
1384 		}
1385 
1386 		switch (rtm->rtm_type) {
1387 		case RTM_ADD:
1388 		case RTM_GET:
1389 		case RTM_CHANGE:
1390 			if (IN6_IS_ADDR_UNSPECIFIED(&nexthop) &&
1391 			    !(flags & F_CONNECTED)) {
1392 				log_warnx("rtmsg_process no nexthop for %s/%u",
1393 				    log_in6addr(&prefix), prefixlen);
1394 				continue;
1395 			}
1396 
1397 			if ((okr = kroute_find(&prefix, prefixlen, prio))
1398 			    != NULL) {
1399 				/* just add new multipath routes */
1400 				if (mpath && rtm->rtm_type == RTM_ADD)
1401 					goto add;
1402 				/* get the correct route */
1403 				kr = okr;
1404 				if (mpath && (kr = kroute_matchgw(okr,
1405 				    &nexthop, scope)) == NULL) {
1406 					log_warnx("rtmsg_process: mpath route"
1407 					    " not found");
1408 					/* add routes we missed out earlier */
1409 					goto add;
1410 				}
1411 
1412 				if (kr->r.flags & F_REDISTRIBUTED)
1413 					flags |= F_REDISTRIBUTED;
1414 				kr->r.nexthop = nexthop;
1415 				kr->r.scope = scope;
1416 				kr->r.flags = flags;
1417 				kr->r.ifindex = ifindex;
1418 
1419 				rtlabel_unref(kr->r.rtlabel);
1420 				kr->r.rtlabel = 0;
1421 				kr->r.ext_tag = 0;
1422 				if ((label = (struct sockaddr_rtlabel *)
1423 				    rti_info[RTAX_LABEL]) != NULL) {
1424 					kr->r.rtlabel =
1425 					    rtlabel_name2id(label->sr_label);
1426 					kr->r.ext_tag =
1427 					    rtlabel_id2tag(kr->r.rtlabel);
1428 				}
1429 
1430 				if (kif_validate(kr->r.ifindex))
1431 					kr->r.flags &= ~F_DOWN;
1432 				else
1433 					kr->r.flags |= F_DOWN;
1434 
1435 				/* just readd, the RDE will care */
1436 				kr_redistribute(okr);
1437 			} else {
1438 add:
1439 				if ((kr = calloc(1,
1440 				    sizeof(struct kroute_node))) == NULL) {
1441 					log_warn("rtmsg_process calloc");
1442 					return (-1);
1443 				}
1444 				kr->r.prefix = prefix;
1445 				kr->r.prefixlen = prefixlen;
1446 				kr->r.nexthop = nexthop;
1447 				kr->r.scope = scope;
1448 				kr->r.flags = flags;
1449 				kr->r.ifindex = ifindex;
1450 				kr->r.priority = prio;
1451 
1452 				if (rtm->rtm_priority == kr_state.fib_prio) {
1453 					log_warnx("alien OSPF route %s/%d",
1454 					    log_in6addr(&prefix), prefixlen);
1455 					rv = send_rtmsg(kr_state.fd,
1456 					    RTM_DELETE, &kr->r);
1457 					free(kr);
1458 					if (rv == -1)
1459 						return (-1);
1460 				} else {
1461 					if ((label = (struct sockaddr_rtlabel *)
1462 					    rti_info[RTAX_LABEL]) != NULL) {
1463 						kr->r.rtlabel =
1464 						    rtlabel_name2id(
1465 						    label->sr_label);
1466 						kr->r.ext_tag =
1467 						    rtlabel_id2tag(
1468 						    kr->r.rtlabel);
1469 					}
1470 
1471 					kroute_insert(kr);
1472 				}
1473 			}
1474 			break;
1475 		case RTM_DELETE:
1476 			if ((kr = kroute_find(&prefix, prefixlen, prio)) ==
1477 			    NULL)
1478 				continue;
1479 			if (!(kr->r.flags & F_KERNEL))
1480 				continue;
1481 			/* get the correct route */
1482 			okr = kr;
1483 			if (mpath && (kr = kroute_matchgw(kr, &nexthop,
1484 			    scope)) == NULL) {
1485 				log_warnx("rtmsg_process mpath route"
1486 				    " not found");
1487 				return (-1);
1488 			}
1489 			if (kroute_remove(kr) == -1)
1490 				return (-1);
1491 			break;
1492 		case RTM_IFINFO:
1493 			memcpy(&ifm, next, sizeof(ifm));
1494 			if_change(ifm.ifm_index, ifm.ifm_flags, &ifm.ifm_data,
1495 			    (struct sockaddr_dl *)rti_info[RTAX_IFP]);
1496 			break;
1497 		case RTM_NEWADDR:
1498 			ifam = (struct ifa_msghdr *)rtm;
1499 			if ((ifam->ifam_addrs & (RTA_NETMASK | RTA_IFA |
1500 			    RTA_BRD)) == 0)
1501 				break;
1502 
1503 			if_newaddr(ifam->ifam_index,
1504 			    (struct sockaddr_in6 *)rti_info[RTAX_IFA],
1505 			    (struct sockaddr_in6 *)rti_info[RTAX_NETMASK],
1506 			    (struct sockaddr_in6 *)rti_info[RTAX_BRD]);
1507 			break;
1508 		case RTM_DELADDR:
1509 			ifam = (struct ifa_msghdr *)rtm;
1510 			if ((ifam->ifam_addrs & (RTA_NETMASK | RTA_IFA |
1511 			    RTA_BRD)) == 0)
1512 				break;
1513 
1514 			if_deladdr(ifam->ifam_index,
1515 			    (struct sockaddr_in6 *)rti_info[RTAX_IFA],
1516 			    (struct sockaddr_in6 *)rti_info[RTAX_NETMASK],
1517 			    (struct sockaddr_in6 *)rti_info[RTAX_BRD]);
1518 			break;
1519 		case RTM_IFANNOUNCE:
1520 			if_announce(next);
1521 			break;
1522 		default:
1523 			/* ignore for now */
1524 			break;
1525 		}
1526 	}
1527 	return (offset);
1528 }
1529